1 // SPDX-License-Identifier: GPL-2.0 2 /* Copyright (c) 2018-2023, Intel Corporation. */ 3 4 /* Intel(R) Ethernet Connection E800 Series Linux Driver */ 5 6 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt 7 8 #include <generated/utsrelease.h> 9 #include <linux/crash_dump.h> 10 #include "ice.h" 11 #include "ice_base.h" 12 #include "ice_lib.h" 13 #include "ice_fltr.h" 14 #include "ice_dcb_lib.h" 15 #include "ice_dcb_nl.h" 16 #include "devlink/devlink.h" 17 #include "devlink/devlink_port.h" 18 #include "ice_sf_eth.h" 19 #include "ice_hwmon.h" 20 /* Including ice_trace.h with CREATE_TRACE_POINTS defined will generate the 21 * ice tracepoint functions. This must be done exactly once across the 22 * ice driver. 23 */ 24 #define CREATE_TRACE_POINTS 25 #include "ice_trace.h" 26 #include "ice_eswitch.h" 27 #include "ice_tc_lib.h" 28 #include "ice_vsi_vlan_ops.h" 29 #include <net/xdp_sock_drv.h> 30 31 #define DRV_SUMMARY "Intel(R) Ethernet Connection E800 Series Linux Driver" 32 static const char ice_driver_string[] = DRV_SUMMARY; 33 static const char ice_copyright[] = "Copyright (c) 2018, Intel Corporation."; 34 35 /* DDP Package file located in firmware search paths (e.g. /lib/firmware/) */ 36 #define ICE_DDP_PKG_PATH "intel/ice/ddp/" 37 #define ICE_DDP_PKG_FILE ICE_DDP_PKG_PATH "ice.pkg" 38 39 MODULE_DESCRIPTION(DRV_SUMMARY); 40 MODULE_IMPORT_NS(LIBIE); 41 MODULE_LICENSE("GPL v2"); 42 MODULE_FIRMWARE(ICE_DDP_PKG_FILE); 43 44 static int debug = -1; 45 module_param(debug, int, 0644); 46 #ifndef CONFIG_DYNAMIC_DEBUG 47 MODULE_PARM_DESC(debug, "netif level (0=none,...,16=all), hw debug_mask (0x8XXXXXXX)"); 48 #else 49 MODULE_PARM_DESC(debug, "netif level (0=none,...,16=all)"); 50 #endif /* !CONFIG_DYNAMIC_DEBUG */ 51 52 DEFINE_STATIC_KEY_FALSE(ice_xdp_locking_key); 53 EXPORT_SYMBOL(ice_xdp_locking_key); 54 55 /** 56 * ice_hw_to_dev - Get device pointer from the hardware structure 57 * @hw: pointer to the device HW structure 58 * 59 * Used to access the device pointer from compilation units which can't easily 60 * include the definition of struct ice_pf without leading to circular header 61 * dependencies. 62 */ 63 struct device *ice_hw_to_dev(struct ice_hw *hw) 64 { 65 struct ice_pf *pf = container_of(hw, struct ice_pf, hw); 66 67 return &pf->pdev->dev; 68 } 69 70 static struct workqueue_struct *ice_wq; 71 struct workqueue_struct *ice_lag_wq; 72 static const struct net_device_ops ice_netdev_safe_mode_ops; 73 static const struct net_device_ops ice_netdev_ops; 74 75 static void ice_rebuild(struct ice_pf *pf, enum ice_reset_req reset_type); 76 77 static void ice_vsi_release_all(struct ice_pf *pf); 78 79 static int ice_rebuild_channels(struct ice_pf *pf); 80 static void ice_remove_q_channels(struct ice_vsi *vsi, bool rem_adv_fltr); 81 82 static int 83 ice_indr_setup_tc_cb(struct net_device *netdev, struct Qdisc *sch, 84 void *cb_priv, enum tc_setup_type type, void *type_data, 85 void *data, 86 void (*cleanup)(struct flow_block_cb *block_cb)); 87 88 bool netif_is_ice(const struct net_device *dev) 89 { 90 return dev && (dev->netdev_ops == &ice_netdev_ops); 91 } 92 93 /** 94 * ice_get_tx_pending - returns number of Tx descriptors not processed 95 * @ring: the ring of descriptors 96 */ 97 static u16 ice_get_tx_pending(struct ice_tx_ring *ring) 98 { 99 u16 head, tail; 100 101 head = ring->next_to_clean; 102 tail = ring->next_to_use; 103 104 if (head != tail) 105 return (head < tail) ? 106 tail - head : (tail + ring->count - head); 107 return 0; 108 } 109 110 /** 111 * ice_check_for_hang_subtask - check for and recover hung queues 112 * @pf: pointer to PF struct 113 */ 114 static void ice_check_for_hang_subtask(struct ice_pf *pf) 115 { 116 struct ice_vsi *vsi = NULL; 117 struct ice_hw *hw; 118 unsigned int i; 119 int packets; 120 u32 v; 121 122 ice_for_each_vsi(pf, v) 123 if (pf->vsi[v] && pf->vsi[v]->type == ICE_VSI_PF) { 124 vsi = pf->vsi[v]; 125 break; 126 } 127 128 if (!vsi || test_bit(ICE_VSI_DOWN, vsi->state)) 129 return; 130 131 if (!(vsi->netdev && netif_carrier_ok(vsi->netdev))) 132 return; 133 134 hw = &vsi->back->hw; 135 136 ice_for_each_txq(vsi, i) { 137 struct ice_tx_ring *tx_ring = vsi->tx_rings[i]; 138 struct ice_ring_stats *ring_stats; 139 140 if (!tx_ring) 141 continue; 142 if (ice_ring_ch_enabled(tx_ring)) 143 continue; 144 145 ring_stats = tx_ring->ring_stats; 146 if (!ring_stats) 147 continue; 148 149 if (tx_ring->desc) { 150 /* If packet counter has not changed the queue is 151 * likely stalled, so force an interrupt for this 152 * queue. 153 * 154 * prev_pkt would be negative if there was no 155 * pending work. 156 */ 157 packets = ring_stats->stats.pkts & INT_MAX; 158 if (ring_stats->tx_stats.prev_pkt == packets) { 159 /* Trigger sw interrupt to revive the queue */ 160 ice_trigger_sw_intr(hw, tx_ring->q_vector); 161 continue; 162 } 163 164 /* Memory barrier between read of packet count and call 165 * to ice_get_tx_pending() 166 */ 167 smp_rmb(); 168 ring_stats->tx_stats.prev_pkt = 169 ice_get_tx_pending(tx_ring) ? packets : -1; 170 } 171 } 172 } 173 174 /** 175 * ice_init_mac_fltr - Set initial MAC filters 176 * @pf: board private structure 177 * 178 * Set initial set of MAC filters for PF VSI; configure filters for permanent 179 * address and broadcast address. If an error is encountered, netdevice will be 180 * unregistered. 181 */ 182 static int ice_init_mac_fltr(struct ice_pf *pf) 183 { 184 struct ice_vsi *vsi; 185 u8 *perm_addr; 186 187 vsi = ice_get_main_vsi(pf); 188 if (!vsi) 189 return -EINVAL; 190 191 perm_addr = vsi->port_info->mac.perm_addr; 192 return ice_fltr_add_mac_and_broadcast(vsi, perm_addr, ICE_FWD_TO_VSI); 193 } 194 195 /** 196 * ice_add_mac_to_sync_list - creates list of MAC addresses to be synced 197 * @netdev: the net device on which the sync is happening 198 * @addr: MAC address to sync 199 * 200 * This is a callback function which is called by the in kernel device sync 201 * functions (like __dev_uc_sync, __dev_mc_sync, etc). This function only 202 * populates the tmp_sync_list, which is later used by ice_add_mac to add the 203 * MAC filters from the hardware. 204 */ 205 static int ice_add_mac_to_sync_list(struct net_device *netdev, const u8 *addr) 206 { 207 struct ice_netdev_priv *np = netdev_priv(netdev); 208 struct ice_vsi *vsi = np->vsi; 209 210 if (ice_fltr_add_mac_to_list(vsi, &vsi->tmp_sync_list, addr, 211 ICE_FWD_TO_VSI)) 212 return -EINVAL; 213 214 return 0; 215 } 216 217 /** 218 * ice_add_mac_to_unsync_list - creates list of MAC addresses to be unsynced 219 * @netdev: the net device on which the unsync is happening 220 * @addr: MAC address to unsync 221 * 222 * This is a callback function which is called by the in kernel device unsync 223 * functions (like __dev_uc_unsync, __dev_mc_unsync, etc). This function only 224 * populates the tmp_unsync_list, which is later used by ice_remove_mac to 225 * delete the MAC filters from the hardware. 226 */ 227 static int ice_add_mac_to_unsync_list(struct net_device *netdev, const u8 *addr) 228 { 229 struct ice_netdev_priv *np = netdev_priv(netdev); 230 struct ice_vsi *vsi = np->vsi; 231 232 /* Under some circumstances, we might receive a request to delete our 233 * own device address from our uc list. Because we store the device 234 * address in the VSI's MAC filter list, we need to ignore such 235 * requests and not delete our device address from this list. 236 */ 237 if (ether_addr_equal(addr, netdev->dev_addr)) 238 return 0; 239 240 if (ice_fltr_add_mac_to_list(vsi, &vsi->tmp_unsync_list, addr, 241 ICE_FWD_TO_VSI)) 242 return -EINVAL; 243 244 return 0; 245 } 246 247 /** 248 * ice_vsi_fltr_changed - check if filter state changed 249 * @vsi: VSI to be checked 250 * 251 * returns true if filter state has changed, false otherwise. 252 */ 253 static bool ice_vsi_fltr_changed(struct ice_vsi *vsi) 254 { 255 return test_bit(ICE_VSI_UMAC_FLTR_CHANGED, vsi->state) || 256 test_bit(ICE_VSI_MMAC_FLTR_CHANGED, vsi->state); 257 } 258 259 /** 260 * ice_set_promisc - Enable promiscuous mode for a given PF 261 * @vsi: the VSI being configured 262 * @promisc_m: mask of promiscuous config bits 263 * 264 */ 265 static int ice_set_promisc(struct ice_vsi *vsi, u8 promisc_m) 266 { 267 int status; 268 269 if (vsi->type != ICE_VSI_PF) 270 return 0; 271 272 if (ice_vsi_has_non_zero_vlans(vsi)) { 273 promisc_m |= (ICE_PROMISC_VLAN_RX | ICE_PROMISC_VLAN_TX); 274 status = ice_fltr_set_vlan_vsi_promisc(&vsi->back->hw, vsi, 275 promisc_m); 276 } else { 277 status = ice_fltr_set_vsi_promisc(&vsi->back->hw, vsi->idx, 278 promisc_m, 0); 279 } 280 if (status && status != -EEXIST) 281 return status; 282 283 netdev_dbg(vsi->netdev, "set promisc filter bits for VSI %i: 0x%x\n", 284 vsi->vsi_num, promisc_m); 285 return 0; 286 } 287 288 /** 289 * ice_clear_promisc - Disable promiscuous mode for a given PF 290 * @vsi: the VSI being configured 291 * @promisc_m: mask of promiscuous config bits 292 * 293 */ 294 static int ice_clear_promisc(struct ice_vsi *vsi, u8 promisc_m) 295 { 296 int status; 297 298 if (vsi->type != ICE_VSI_PF) 299 return 0; 300 301 if (ice_vsi_has_non_zero_vlans(vsi)) { 302 promisc_m |= (ICE_PROMISC_VLAN_RX | ICE_PROMISC_VLAN_TX); 303 status = ice_fltr_clear_vlan_vsi_promisc(&vsi->back->hw, vsi, 304 promisc_m); 305 } else { 306 status = ice_fltr_clear_vsi_promisc(&vsi->back->hw, vsi->idx, 307 promisc_m, 0); 308 } 309 310 netdev_dbg(vsi->netdev, "clear promisc filter bits for VSI %i: 0x%x\n", 311 vsi->vsi_num, promisc_m); 312 return status; 313 } 314 315 /** 316 * ice_vsi_sync_fltr - Update the VSI filter list to the HW 317 * @vsi: ptr to the VSI 318 * 319 * Push any outstanding VSI filter changes through the AdminQ. 320 */ 321 static int ice_vsi_sync_fltr(struct ice_vsi *vsi) 322 { 323 struct ice_vsi_vlan_ops *vlan_ops = ice_get_compat_vsi_vlan_ops(vsi); 324 struct device *dev = ice_pf_to_dev(vsi->back); 325 struct net_device *netdev = vsi->netdev; 326 bool promisc_forced_on = false; 327 struct ice_pf *pf = vsi->back; 328 struct ice_hw *hw = &pf->hw; 329 u32 changed_flags = 0; 330 int err; 331 332 if (!vsi->netdev) 333 return -EINVAL; 334 335 while (test_and_set_bit(ICE_CFG_BUSY, vsi->state)) 336 usleep_range(1000, 2000); 337 338 changed_flags = vsi->current_netdev_flags ^ vsi->netdev->flags; 339 vsi->current_netdev_flags = vsi->netdev->flags; 340 341 INIT_LIST_HEAD(&vsi->tmp_sync_list); 342 INIT_LIST_HEAD(&vsi->tmp_unsync_list); 343 344 if (ice_vsi_fltr_changed(vsi)) { 345 clear_bit(ICE_VSI_UMAC_FLTR_CHANGED, vsi->state); 346 clear_bit(ICE_VSI_MMAC_FLTR_CHANGED, vsi->state); 347 348 /* grab the netdev's addr_list_lock */ 349 netif_addr_lock_bh(netdev); 350 __dev_uc_sync(netdev, ice_add_mac_to_sync_list, 351 ice_add_mac_to_unsync_list); 352 __dev_mc_sync(netdev, ice_add_mac_to_sync_list, 353 ice_add_mac_to_unsync_list); 354 /* our temp lists are populated. release lock */ 355 netif_addr_unlock_bh(netdev); 356 } 357 358 /* Remove MAC addresses in the unsync list */ 359 err = ice_fltr_remove_mac_list(vsi, &vsi->tmp_unsync_list); 360 ice_fltr_free_list(dev, &vsi->tmp_unsync_list); 361 if (err) { 362 netdev_err(netdev, "Failed to delete MAC filters\n"); 363 /* if we failed because of alloc failures, just bail */ 364 if (err == -ENOMEM) 365 goto out; 366 } 367 368 /* Add MAC addresses in the sync list */ 369 err = ice_fltr_add_mac_list(vsi, &vsi->tmp_sync_list); 370 ice_fltr_free_list(dev, &vsi->tmp_sync_list); 371 /* If filter is added successfully or already exists, do not go into 372 * 'if' condition and report it as error. Instead continue processing 373 * rest of the function. 374 */ 375 if (err && err != -EEXIST) { 376 netdev_err(netdev, "Failed to add MAC filters\n"); 377 /* If there is no more space for new umac filters, VSI 378 * should go into promiscuous mode. There should be some 379 * space reserved for promiscuous filters. 380 */ 381 if (hw->adminq.sq_last_status == ICE_AQ_RC_ENOSPC && 382 !test_and_set_bit(ICE_FLTR_OVERFLOW_PROMISC, 383 vsi->state)) { 384 promisc_forced_on = true; 385 netdev_warn(netdev, "Reached MAC filter limit, forcing promisc mode on VSI %d\n", 386 vsi->vsi_num); 387 } else { 388 goto out; 389 } 390 } 391 err = 0; 392 /* check for changes in promiscuous modes */ 393 if (changed_flags & IFF_ALLMULTI) { 394 if (vsi->current_netdev_flags & IFF_ALLMULTI) { 395 err = ice_set_promisc(vsi, ICE_MCAST_PROMISC_BITS); 396 if (err) { 397 vsi->current_netdev_flags &= ~IFF_ALLMULTI; 398 goto out_promisc; 399 } 400 } else { 401 /* !(vsi->current_netdev_flags & IFF_ALLMULTI) */ 402 err = ice_clear_promisc(vsi, ICE_MCAST_PROMISC_BITS); 403 if (err) { 404 vsi->current_netdev_flags |= IFF_ALLMULTI; 405 goto out_promisc; 406 } 407 } 408 } 409 410 if (((changed_flags & IFF_PROMISC) || promisc_forced_on) || 411 test_bit(ICE_VSI_PROMISC_CHANGED, vsi->state)) { 412 clear_bit(ICE_VSI_PROMISC_CHANGED, vsi->state); 413 if (vsi->current_netdev_flags & IFF_PROMISC) { 414 /* Apply Rx filter rule to get traffic from wire */ 415 if (!ice_is_dflt_vsi_in_use(vsi->port_info)) { 416 err = ice_set_dflt_vsi(vsi); 417 if (err && err != -EEXIST) { 418 netdev_err(netdev, "Error %d setting default VSI %i Rx rule\n", 419 err, vsi->vsi_num); 420 vsi->current_netdev_flags &= 421 ~IFF_PROMISC; 422 goto out_promisc; 423 } 424 err = 0; 425 vlan_ops->dis_rx_filtering(vsi); 426 427 /* promiscuous mode implies allmulticast so 428 * that VSIs that are in promiscuous mode are 429 * subscribed to multicast packets coming to 430 * the port 431 */ 432 err = ice_set_promisc(vsi, 433 ICE_MCAST_PROMISC_BITS); 434 if (err) 435 goto out_promisc; 436 } 437 } else { 438 /* Clear Rx filter to remove traffic from wire */ 439 if (ice_is_vsi_dflt_vsi(vsi)) { 440 err = ice_clear_dflt_vsi(vsi); 441 if (err) { 442 netdev_err(netdev, "Error %d clearing default VSI %i Rx rule\n", 443 err, vsi->vsi_num); 444 vsi->current_netdev_flags |= 445 IFF_PROMISC; 446 goto out_promisc; 447 } 448 if (vsi->netdev->features & 449 NETIF_F_HW_VLAN_CTAG_FILTER) 450 vlan_ops->ena_rx_filtering(vsi); 451 } 452 453 /* disable allmulti here, but only if allmulti is not 454 * still enabled for the netdev 455 */ 456 if (!(vsi->current_netdev_flags & IFF_ALLMULTI)) { 457 err = ice_clear_promisc(vsi, 458 ICE_MCAST_PROMISC_BITS); 459 if (err) { 460 netdev_err(netdev, "Error %d clearing multicast promiscuous on VSI %i\n", 461 err, vsi->vsi_num); 462 } 463 } 464 } 465 } 466 goto exit; 467 468 out_promisc: 469 set_bit(ICE_VSI_PROMISC_CHANGED, vsi->state); 470 goto exit; 471 out: 472 /* if something went wrong then set the changed flag so we try again */ 473 set_bit(ICE_VSI_UMAC_FLTR_CHANGED, vsi->state); 474 set_bit(ICE_VSI_MMAC_FLTR_CHANGED, vsi->state); 475 exit: 476 clear_bit(ICE_CFG_BUSY, vsi->state); 477 return err; 478 } 479 480 /** 481 * ice_sync_fltr_subtask - Sync the VSI filter list with HW 482 * @pf: board private structure 483 */ 484 static void ice_sync_fltr_subtask(struct ice_pf *pf) 485 { 486 int v; 487 488 if (!pf || !(test_bit(ICE_FLAG_FLTR_SYNC, pf->flags))) 489 return; 490 491 clear_bit(ICE_FLAG_FLTR_SYNC, pf->flags); 492 493 ice_for_each_vsi(pf, v) 494 if (pf->vsi[v] && ice_vsi_fltr_changed(pf->vsi[v]) && 495 ice_vsi_sync_fltr(pf->vsi[v])) { 496 /* come back and try again later */ 497 set_bit(ICE_FLAG_FLTR_SYNC, pf->flags); 498 break; 499 } 500 } 501 502 /** 503 * ice_pf_dis_all_vsi - Pause all VSIs on a PF 504 * @pf: the PF 505 * @locked: is the rtnl_lock already held 506 */ 507 static void ice_pf_dis_all_vsi(struct ice_pf *pf, bool locked) 508 { 509 int node; 510 int v; 511 512 ice_for_each_vsi(pf, v) 513 if (pf->vsi[v]) 514 ice_dis_vsi(pf->vsi[v], locked); 515 516 for (node = 0; node < ICE_MAX_PF_AGG_NODES; node++) 517 pf->pf_agg_node[node].num_vsis = 0; 518 519 for (node = 0; node < ICE_MAX_VF_AGG_NODES; node++) 520 pf->vf_agg_node[node].num_vsis = 0; 521 } 522 523 /** 524 * ice_clear_sw_switch_recipes - clear switch recipes 525 * @pf: board private structure 526 * 527 * Mark switch recipes as not created in sw structures. There are cases where 528 * rules (especially advanced rules) need to be restored, either re-read from 529 * hardware or added again. For example after the reset. 'recp_created' flag 530 * prevents from doing that and need to be cleared upfront. 531 */ 532 static void ice_clear_sw_switch_recipes(struct ice_pf *pf) 533 { 534 struct ice_sw_recipe *recp; 535 u8 i; 536 537 recp = pf->hw.switch_info->recp_list; 538 for (i = 0; i < ICE_MAX_NUM_RECIPES; i++) 539 recp[i].recp_created = false; 540 } 541 542 /** 543 * ice_prepare_for_reset - prep for reset 544 * @pf: board private structure 545 * @reset_type: reset type requested 546 * 547 * Inform or close all dependent features in prep for reset. 548 */ 549 static void 550 ice_prepare_for_reset(struct ice_pf *pf, enum ice_reset_req reset_type) 551 { 552 struct ice_hw *hw = &pf->hw; 553 struct ice_vsi *vsi; 554 struct ice_vf *vf; 555 unsigned int bkt; 556 557 dev_dbg(ice_pf_to_dev(pf), "reset_type=%d\n", reset_type); 558 559 /* already prepared for reset */ 560 if (test_bit(ICE_PREPARED_FOR_RESET, pf->state)) 561 return; 562 563 synchronize_irq(pf->oicr_irq.virq); 564 565 ice_unplug_aux_dev(pf); 566 567 /* Notify VFs of impending reset */ 568 if (ice_check_sq_alive(hw, &hw->mailboxq)) 569 ice_vc_notify_reset(pf); 570 571 /* Disable VFs until reset is completed */ 572 mutex_lock(&pf->vfs.table_lock); 573 ice_for_each_vf(pf, bkt, vf) 574 ice_set_vf_state_dis(vf); 575 mutex_unlock(&pf->vfs.table_lock); 576 577 if (ice_is_eswitch_mode_switchdev(pf)) { 578 if (reset_type != ICE_RESET_PFR) 579 ice_clear_sw_switch_recipes(pf); 580 } 581 582 /* release ADQ specific HW and SW resources */ 583 vsi = ice_get_main_vsi(pf); 584 if (!vsi) 585 goto skip; 586 587 /* to be on safe side, reset orig_rss_size so that normal flow 588 * of deciding rss_size can take precedence 589 */ 590 vsi->orig_rss_size = 0; 591 592 if (test_bit(ICE_FLAG_TC_MQPRIO, pf->flags)) { 593 if (reset_type == ICE_RESET_PFR) { 594 vsi->old_ena_tc = vsi->all_enatc; 595 vsi->old_numtc = vsi->all_numtc; 596 } else { 597 ice_remove_q_channels(vsi, true); 598 599 /* for other reset type, do not support channel rebuild 600 * hence reset needed info 601 */ 602 vsi->old_ena_tc = 0; 603 vsi->all_enatc = 0; 604 vsi->old_numtc = 0; 605 vsi->all_numtc = 0; 606 vsi->req_txq = 0; 607 vsi->req_rxq = 0; 608 clear_bit(ICE_FLAG_TC_MQPRIO, pf->flags); 609 memset(&vsi->mqprio_qopt, 0, sizeof(vsi->mqprio_qopt)); 610 } 611 } 612 613 if (vsi->netdev) 614 netif_device_detach(vsi->netdev); 615 skip: 616 617 /* clear SW filtering DB */ 618 ice_clear_hw_tbls(hw); 619 /* disable the VSIs and their queues that are not already DOWN */ 620 set_bit(ICE_VSI_REBUILD_PENDING, ice_get_main_vsi(pf)->state); 621 ice_pf_dis_all_vsi(pf, false); 622 623 if (test_bit(ICE_FLAG_PTP_SUPPORTED, pf->flags)) 624 ice_ptp_prepare_for_reset(pf, reset_type); 625 626 if (ice_is_feature_supported(pf, ICE_F_GNSS)) 627 ice_gnss_exit(pf); 628 629 if (hw->port_info) 630 ice_sched_clear_port(hw->port_info); 631 632 ice_shutdown_all_ctrlq(hw, false); 633 634 set_bit(ICE_PREPARED_FOR_RESET, pf->state); 635 } 636 637 /** 638 * ice_do_reset - Initiate one of many types of resets 639 * @pf: board private structure 640 * @reset_type: reset type requested before this function was called. 641 */ 642 static void ice_do_reset(struct ice_pf *pf, enum ice_reset_req reset_type) 643 { 644 struct device *dev = ice_pf_to_dev(pf); 645 struct ice_hw *hw = &pf->hw; 646 647 dev_dbg(dev, "reset_type 0x%x requested\n", reset_type); 648 649 if (pf->lag && pf->lag->bonded && reset_type == ICE_RESET_PFR) { 650 dev_dbg(dev, "PFR on a bonded interface, promoting to CORER\n"); 651 reset_type = ICE_RESET_CORER; 652 } 653 654 ice_prepare_for_reset(pf, reset_type); 655 656 /* trigger the reset */ 657 if (ice_reset(hw, reset_type)) { 658 dev_err(dev, "reset %d failed\n", reset_type); 659 set_bit(ICE_RESET_FAILED, pf->state); 660 clear_bit(ICE_RESET_OICR_RECV, pf->state); 661 clear_bit(ICE_PREPARED_FOR_RESET, pf->state); 662 clear_bit(ICE_PFR_REQ, pf->state); 663 clear_bit(ICE_CORER_REQ, pf->state); 664 clear_bit(ICE_GLOBR_REQ, pf->state); 665 wake_up(&pf->reset_wait_queue); 666 return; 667 } 668 669 /* PFR is a bit of a special case because it doesn't result in an OICR 670 * interrupt. So for PFR, rebuild after the reset and clear the reset- 671 * associated state bits. 672 */ 673 if (reset_type == ICE_RESET_PFR) { 674 pf->pfr_count++; 675 ice_rebuild(pf, reset_type); 676 clear_bit(ICE_PREPARED_FOR_RESET, pf->state); 677 clear_bit(ICE_PFR_REQ, pf->state); 678 wake_up(&pf->reset_wait_queue); 679 ice_reset_all_vfs(pf); 680 } 681 } 682 683 /** 684 * ice_reset_subtask - Set up for resetting the device and driver 685 * @pf: board private structure 686 */ 687 static void ice_reset_subtask(struct ice_pf *pf) 688 { 689 enum ice_reset_req reset_type = ICE_RESET_INVAL; 690 691 /* When a CORER/GLOBR/EMPR is about to happen, the hardware triggers an 692 * OICR interrupt. The OICR handler (ice_misc_intr) determines what type 693 * of reset is pending and sets bits in pf->state indicating the reset 694 * type and ICE_RESET_OICR_RECV. So, if the latter bit is set 695 * prepare for pending reset if not already (for PF software-initiated 696 * global resets the software should already be prepared for it as 697 * indicated by ICE_PREPARED_FOR_RESET; for global resets initiated 698 * by firmware or software on other PFs, that bit is not set so prepare 699 * for the reset now), poll for reset done, rebuild and return. 700 */ 701 if (test_bit(ICE_RESET_OICR_RECV, pf->state)) { 702 /* Perform the largest reset requested */ 703 if (test_and_clear_bit(ICE_CORER_RECV, pf->state)) 704 reset_type = ICE_RESET_CORER; 705 if (test_and_clear_bit(ICE_GLOBR_RECV, pf->state)) 706 reset_type = ICE_RESET_GLOBR; 707 if (test_and_clear_bit(ICE_EMPR_RECV, pf->state)) 708 reset_type = ICE_RESET_EMPR; 709 /* return if no valid reset type requested */ 710 if (reset_type == ICE_RESET_INVAL) 711 return; 712 ice_prepare_for_reset(pf, reset_type); 713 714 /* make sure we are ready to rebuild */ 715 if (ice_check_reset(&pf->hw)) { 716 set_bit(ICE_RESET_FAILED, pf->state); 717 } else { 718 /* done with reset. start rebuild */ 719 pf->hw.reset_ongoing = false; 720 ice_rebuild(pf, reset_type); 721 /* clear bit to resume normal operations, but 722 * ICE_NEEDS_RESTART bit is set in case rebuild failed 723 */ 724 clear_bit(ICE_RESET_OICR_RECV, pf->state); 725 clear_bit(ICE_PREPARED_FOR_RESET, pf->state); 726 clear_bit(ICE_PFR_REQ, pf->state); 727 clear_bit(ICE_CORER_REQ, pf->state); 728 clear_bit(ICE_GLOBR_REQ, pf->state); 729 wake_up(&pf->reset_wait_queue); 730 ice_reset_all_vfs(pf); 731 } 732 733 return; 734 } 735 736 /* No pending resets to finish processing. Check for new resets */ 737 if (test_bit(ICE_PFR_REQ, pf->state)) { 738 reset_type = ICE_RESET_PFR; 739 if (pf->lag && pf->lag->bonded) { 740 dev_dbg(ice_pf_to_dev(pf), "PFR on a bonded interface, promoting to CORER\n"); 741 reset_type = ICE_RESET_CORER; 742 } 743 } 744 if (test_bit(ICE_CORER_REQ, pf->state)) 745 reset_type = ICE_RESET_CORER; 746 if (test_bit(ICE_GLOBR_REQ, pf->state)) 747 reset_type = ICE_RESET_GLOBR; 748 /* If no valid reset type requested just return */ 749 if (reset_type == ICE_RESET_INVAL) 750 return; 751 752 /* reset if not already down or busy */ 753 if (!test_bit(ICE_DOWN, pf->state) && 754 !test_bit(ICE_CFG_BUSY, pf->state)) { 755 ice_do_reset(pf, reset_type); 756 } 757 } 758 759 /** 760 * ice_print_topo_conflict - print topology conflict message 761 * @vsi: the VSI whose topology status is being checked 762 */ 763 static void ice_print_topo_conflict(struct ice_vsi *vsi) 764 { 765 switch (vsi->port_info->phy.link_info.topo_media_conflict) { 766 case ICE_AQ_LINK_TOPO_CONFLICT: 767 case ICE_AQ_LINK_MEDIA_CONFLICT: 768 case ICE_AQ_LINK_TOPO_UNREACH_PRT: 769 case ICE_AQ_LINK_TOPO_UNDRUTIL_PRT: 770 case ICE_AQ_LINK_TOPO_UNDRUTIL_MEDIA: 771 netdev_info(vsi->netdev, "Potential misconfiguration of the Ethernet port detected. If it was not intended, please use the Intel (R) Ethernet Port Configuration Tool to address the issue.\n"); 772 break; 773 case ICE_AQ_LINK_TOPO_UNSUPP_MEDIA: 774 if (test_bit(ICE_FLAG_LINK_LENIENT_MODE_ENA, vsi->back->flags)) 775 netdev_warn(vsi->netdev, "An unsupported module type was detected. Refer to the Intel(R) Ethernet Adapters and Devices User Guide for a list of supported modules\n"); 776 else 777 netdev_err(vsi->netdev, "Rx/Tx is disabled on this device because an unsupported module type was detected. Refer to the Intel(R) Ethernet Adapters and Devices User Guide for a list of supported modules.\n"); 778 break; 779 default: 780 break; 781 } 782 } 783 784 /** 785 * ice_print_link_msg - print link up or down message 786 * @vsi: the VSI whose link status is being queried 787 * @isup: boolean for if the link is now up or down 788 */ 789 void ice_print_link_msg(struct ice_vsi *vsi, bool isup) 790 { 791 struct ice_aqc_get_phy_caps_data *caps; 792 const char *an_advertised; 793 const char *fec_req; 794 const char *speed; 795 const char *fec; 796 const char *fc; 797 const char *an; 798 int status; 799 800 if (!vsi) 801 return; 802 803 if (vsi->current_isup == isup) 804 return; 805 806 vsi->current_isup = isup; 807 808 if (!isup) { 809 netdev_info(vsi->netdev, "NIC Link is Down\n"); 810 return; 811 } 812 813 switch (vsi->port_info->phy.link_info.link_speed) { 814 case ICE_AQ_LINK_SPEED_200GB: 815 speed = "200 G"; 816 break; 817 case ICE_AQ_LINK_SPEED_100GB: 818 speed = "100 G"; 819 break; 820 case ICE_AQ_LINK_SPEED_50GB: 821 speed = "50 G"; 822 break; 823 case ICE_AQ_LINK_SPEED_40GB: 824 speed = "40 G"; 825 break; 826 case ICE_AQ_LINK_SPEED_25GB: 827 speed = "25 G"; 828 break; 829 case ICE_AQ_LINK_SPEED_20GB: 830 speed = "20 G"; 831 break; 832 case ICE_AQ_LINK_SPEED_10GB: 833 speed = "10 G"; 834 break; 835 case ICE_AQ_LINK_SPEED_5GB: 836 speed = "5 G"; 837 break; 838 case ICE_AQ_LINK_SPEED_2500MB: 839 speed = "2.5 G"; 840 break; 841 case ICE_AQ_LINK_SPEED_1000MB: 842 speed = "1 G"; 843 break; 844 case ICE_AQ_LINK_SPEED_100MB: 845 speed = "100 M"; 846 break; 847 default: 848 speed = "Unknown "; 849 break; 850 } 851 852 switch (vsi->port_info->fc.current_mode) { 853 case ICE_FC_FULL: 854 fc = "Rx/Tx"; 855 break; 856 case ICE_FC_TX_PAUSE: 857 fc = "Tx"; 858 break; 859 case ICE_FC_RX_PAUSE: 860 fc = "Rx"; 861 break; 862 case ICE_FC_NONE: 863 fc = "None"; 864 break; 865 default: 866 fc = "Unknown"; 867 break; 868 } 869 870 /* Get FEC mode based on negotiated link info */ 871 switch (vsi->port_info->phy.link_info.fec_info) { 872 case ICE_AQ_LINK_25G_RS_528_FEC_EN: 873 case ICE_AQ_LINK_25G_RS_544_FEC_EN: 874 fec = "RS-FEC"; 875 break; 876 case ICE_AQ_LINK_25G_KR_FEC_EN: 877 fec = "FC-FEC/BASE-R"; 878 break; 879 default: 880 fec = "NONE"; 881 break; 882 } 883 884 /* check if autoneg completed, might be false due to not supported */ 885 if (vsi->port_info->phy.link_info.an_info & ICE_AQ_AN_COMPLETED) 886 an = "True"; 887 else 888 an = "False"; 889 890 /* Get FEC mode requested based on PHY caps last SW configuration */ 891 caps = kzalloc(sizeof(*caps), GFP_KERNEL); 892 if (!caps) { 893 fec_req = "Unknown"; 894 an_advertised = "Unknown"; 895 goto done; 896 } 897 898 status = ice_aq_get_phy_caps(vsi->port_info, false, 899 ICE_AQC_REPORT_ACTIVE_CFG, caps, NULL); 900 if (status) 901 netdev_info(vsi->netdev, "Get phy capability failed.\n"); 902 903 an_advertised = ice_is_phy_caps_an_enabled(caps) ? "On" : "Off"; 904 905 if (caps->link_fec_options & ICE_AQC_PHY_FEC_25G_RS_528_REQ || 906 caps->link_fec_options & ICE_AQC_PHY_FEC_25G_RS_544_REQ) 907 fec_req = "RS-FEC"; 908 else if (caps->link_fec_options & ICE_AQC_PHY_FEC_10G_KR_40G_KR4_REQ || 909 caps->link_fec_options & ICE_AQC_PHY_FEC_25G_KR_REQ) 910 fec_req = "FC-FEC/BASE-R"; 911 else 912 fec_req = "NONE"; 913 914 kfree(caps); 915 916 done: 917 netdev_info(vsi->netdev, "NIC Link is up %sbps Full Duplex, Requested FEC: %s, Negotiated FEC: %s, Autoneg Advertised: %s, Autoneg Negotiated: %s, Flow Control: %s\n", 918 speed, fec_req, fec, an_advertised, an, fc); 919 ice_print_topo_conflict(vsi); 920 } 921 922 /** 923 * ice_vsi_link_event - update the VSI's netdev 924 * @vsi: the VSI on which the link event occurred 925 * @link_up: whether or not the VSI needs to be set up or down 926 */ 927 static void ice_vsi_link_event(struct ice_vsi *vsi, bool link_up) 928 { 929 if (!vsi) 930 return; 931 932 if (test_bit(ICE_VSI_DOWN, vsi->state) || !vsi->netdev) 933 return; 934 935 if (vsi->type == ICE_VSI_PF) { 936 if (link_up == netif_carrier_ok(vsi->netdev)) 937 return; 938 939 if (link_up) { 940 netif_carrier_on(vsi->netdev); 941 netif_tx_wake_all_queues(vsi->netdev); 942 } else { 943 netif_carrier_off(vsi->netdev); 944 netif_tx_stop_all_queues(vsi->netdev); 945 } 946 } 947 } 948 949 /** 950 * ice_set_dflt_mib - send a default config MIB to the FW 951 * @pf: private PF struct 952 * 953 * This function sends a default configuration MIB to the FW. 954 * 955 * If this function errors out at any point, the driver is still able to 956 * function. The main impact is that LFC may not operate as expected. 957 * Therefore an error state in this function should be treated with a DBG 958 * message and continue on with driver rebuild/reenable. 959 */ 960 static void ice_set_dflt_mib(struct ice_pf *pf) 961 { 962 struct device *dev = ice_pf_to_dev(pf); 963 u8 mib_type, *buf, *lldpmib = NULL; 964 u16 len, typelen, offset = 0; 965 struct ice_lldp_org_tlv *tlv; 966 struct ice_hw *hw = &pf->hw; 967 u32 ouisubtype; 968 969 mib_type = SET_LOCAL_MIB_TYPE_LOCAL_MIB; 970 lldpmib = kzalloc(ICE_LLDPDU_SIZE, GFP_KERNEL); 971 if (!lldpmib) { 972 dev_dbg(dev, "%s Failed to allocate MIB memory\n", 973 __func__); 974 return; 975 } 976 977 /* Add ETS CFG TLV */ 978 tlv = (struct ice_lldp_org_tlv *)lldpmib; 979 typelen = ((ICE_TLV_TYPE_ORG << ICE_LLDP_TLV_TYPE_S) | 980 ICE_IEEE_ETS_TLV_LEN); 981 tlv->typelen = htons(typelen); 982 ouisubtype = ((ICE_IEEE_8021QAZ_OUI << ICE_LLDP_TLV_OUI_S) | 983 ICE_IEEE_SUBTYPE_ETS_CFG); 984 tlv->ouisubtype = htonl(ouisubtype); 985 986 buf = tlv->tlvinfo; 987 buf[0] = 0; 988 989 /* ETS CFG all UPs map to TC 0. Next 4 (1 - 4) Octets = 0. 990 * Octets 5 - 12 are BW values, set octet 5 to 100% BW. 991 * Octets 13 - 20 are TSA values - leave as zeros 992 */ 993 buf[5] = 0x64; 994 len = FIELD_GET(ICE_LLDP_TLV_LEN_M, typelen); 995 offset += len + 2; 996 tlv = (struct ice_lldp_org_tlv *) 997 ((char *)tlv + sizeof(tlv->typelen) + len); 998 999 /* Add ETS REC TLV */ 1000 buf = tlv->tlvinfo; 1001 tlv->typelen = htons(typelen); 1002 1003 ouisubtype = ((ICE_IEEE_8021QAZ_OUI << ICE_LLDP_TLV_OUI_S) | 1004 ICE_IEEE_SUBTYPE_ETS_REC); 1005 tlv->ouisubtype = htonl(ouisubtype); 1006 1007 /* First octet of buf is reserved 1008 * Octets 1 - 4 map UP to TC - all UPs map to zero 1009 * Octets 5 - 12 are BW values - set TC 0 to 100%. 1010 * Octets 13 - 20 are TSA value - leave as zeros 1011 */ 1012 buf[5] = 0x64; 1013 offset += len + 2; 1014 tlv = (struct ice_lldp_org_tlv *) 1015 ((char *)tlv + sizeof(tlv->typelen) + len); 1016 1017 /* Add PFC CFG TLV */ 1018 typelen = ((ICE_TLV_TYPE_ORG << ICE_LLDP_TLV_TYPE_S) | 1019 ICE_IEEE_PFC_TLV_LEN); 1020 tlv->typelen = htons(typelen); 1021 1022 ouisubtype = ((ICE_IEEE_8021QAZ_OUI << ICE_LLDP_TLV_OUI_S) | 1023 ICE_IEEE_SUBTYPE_PFC_CFG); 1024 tlv->ouisubtype = htonl(ouisubtype); 1025 1026 /* Octet 1 left as all zeros - PFC disabled */ 1027 buf[0] = 0x08; 1028 len = FIELD_GET(ICE_LLDP_TLV_LEN_M, typelen); 1029 offset += len + 2; 1030 1031 if (ice_aq_set_lldp_mib(hw, mib_type, (void *)lldpmib, offset, NULL)) 1032 dev_dbg(dev, "%s Failed to set default LLDP MIB\n", __func__); 1033 1034 kfree(lldpmib); 1035 } 1036 1037 /** 1038 * ice_check_phy_fw_load - check if PHY FW load failed 1039 * @pf: pointer to PF struct 1040 * @link_cfg_err: bitmap from the link info structure 1041 * 1042 * check if external PHY FW load failed and print an error message if it did 1043 */ 1044 static void ice_check_phy_fw_load(struct ice_pf *pf, u8 link_cfg_err) 1045 { 1046 if (!(link_cfg_err & ICE_AQ_LINK_EXTERNAL_PHY_LOAD_FAILURE)) { 1047 clear_bit(ICE_FLAG_PHY_FW_LOAD_FAILED, pf->flags); 1048 return; 1049 } 1050 1051 if (test_bit(ICE_FLAG_PHY_FW_LOAD_FAILED, pf->flags)) 1052 return; 1053 1054 if (link_cfg_err & ICE_AQ_LINK_EXTERNAL_PHY_LOAD_FAILURE) { 1055 dev_err(ice_pf_to_dev(pf), "Device failed to load the FW for the external PHY. Please download and install the latest NVM for your device and try again\n"); 1056 set_bit(ICE_FLAG_PHY_FW_LOAD_FAILED, pf->flags); 1057 } 1058 } 1059 1060 /** 1061 * ice_check_module_power 1062 * @pf: pointer to PF struct 1063 * @link_cfg_err: bitmap from the link info structure 1064 * 1065 * check module power level returned by a previous call to aq_get_link_info 1066 * and print error messages if module power level is not supported 1067 */ 1068 static void ice_check_module_power(struct ice_pf *pf, u8 link_cfg_err) 1069 { 1070 /* if module power level is supported, clear the flag */ 1071 if (!(link_cfg_err & (ICE_AQ_LINK_INVAL_MAX_POWER_LIMIT | 1072 ICE_AQ_LINK_MODULE_POWER_UNSUPPORTED))) { 1073 clear_bit(ICE_FLAG_MOD_POWER_UNSUPPORTED, pf->flags); 1074 return; 1075 } 1076 1077 /* if ICE_FLAG_MOD_POWER_UNSUPPORTED was previously set and the 1078 * above block didn't clear this bit, there's nothing to do 1079 */ 1080 if (test_bit(ICE_FLAG_MOD_POWER_UNSUPPORTED, pf->flags)) 1081 return; 1082 1083 if (link_cfg_err & ICE_AQ_LINK_INVAL_MAX_POWER_LIMIT) { 1084 dev_err(ice_pf_to_dev(pf), "The installed module is incompatible with the device's NVM image. Cannot start link\n"); 1085 set_bit(ICE_FLAG_MOD_POWER_UNSUPPORTED, pf->flags); 1086 } else if (link_cfg_err & ICE_AQ_LINK_MODULE_POWER_UNSUPPORTED) { 1087 dev_err(ice_pf_to_dev(pf), "The module's power requirements exceed the device's power supply. Cannot start link\n"); 1088 set_bit(ICE_FLAG_MOD_POWER_UNSUPPORTED, pf->flags); 1089 } 1090 } 1091 1092 /** 1093 * ice_check_link_cfg_err - check if link configuration failed 1094 * @pf: pointer to the PF struct 1095 * @link_cfg_err: bitmap from the link info structure 1096 * 1097 * print if any link configuration failure happens due to the value in the 1098 * link_cfg_err parameter in the link info structure 1099 */ 1100 static void ice_check_link_cfg_err(struct ice_pf *pf, u8 link_cfg_err) 1101 { 1102 ice_check_module_power(pf, link_cfg_err); 1103 ice_check_phy_fw_load(pf, link_cfg_err); 1104 } 1105 1106 /** 1107 * ice_link_event - process the link event 1108 * @pf: PF that the link event is associated with 1109 * @pi: port_info for the port that the link event is associated with 1110 * @link_up: true if the physical link is up and false if it is down 1111 * @link_speed: current link speed received from the link event 1112 * 1113 * Returns 0 on success and negative on failure 1114 */ 1115 static int 1116 ice_link_event(struct ice_pf *pf, struct ice_port_info *pi, bool link_up, 1117 u16 link_speed) 1118 { 1119 struct device *dev = ice_pf_to_dev(pf); 1120 struct ice_phy_info *phy_info; 1121 struct ice_vsi *vsi; 1122 u16 old_link_speed; 1123 bool old_link; 1124 int status; 1125 1126 phy_info = &pi->phy; 1127 phy_info->link_info_old = phy_info->link_info; 1128 1129 old_link = !!(phy_info->link_info_old.link_info & ICE_AQ_LINK_UP); 1130 old_link_speed = phy_info->link_info_old.link_speed; 1131 1132 /* update the link info structures and re-enable link events, 1133 * don't bail on failure due to other book keeping needed 1134 */ 1135 status = ice_update_link_info(pi); 1136 if (status) 1137 dev_dbg(dev, "Failed to update link status on port %d, err %d aq_err %s\n", 1138 pi->lport, status, 1139 ice_aq_str(pi->hw->adminq.sq_last_status)); 1140 1141 ice_check_link_cfg_err(pf, pi->phy.link_info.link_cfg_err); 1142 1143 /* Check if the link state is up after updating link info, and treat 1144 * this event as an UP event since the link is actually UP now. 1145 */ 1146 if (phy_info->link_info.link_info & ICE_AQ_LINK_UP) 1147 link_up = true; 1148 1149 vsi = ice_get_main_vsi(pf); 1150 if (!vsi || !vsi->port_info) 1151 return -EINVAL; 1152 1153 /* turn off PHY if media was removed */ 1154 if (!test_bit(ICE_FLAG_NO_MEDIA, pf->flags) && 1155 !(pi->phy.link_info.link_info & ICE_AQ_MEDIA_AVAILABLE)) { 1156 set_bit(ICE_FLAG_NO_MEDIA, pf->flags); 1157 ice_set_link(vsi, false); 1158 } 1159 1160 /* if the old link up/down and speed is the same as the new */ 1161 if (link_up == old_link && link_speed == old_link_speed) 1162 return 0; 1163 1164 ice_ptp_link_change(pf, pf->hw.pf_id, link_up); 1165 1166 if (ice_is_dcb_active(pf)) { 1167 if (test_bit(ICE_FLAG_DCB_ENA, pf->flags)) 1168 ice_dcb_rebuild(pf); 1169 } else { 1170 if (link_up) 1171 ice_set_dflt_mib(pf); 1172 } 1173 ice_vsi_link_event(vsi, link_up); 1174 ice_print_link_msg(vsi, link_up); 1175 1176 ice_vc_notify_link_state(pf); 1177 1178 return 0; 1179 } 1180 1181 /** 1182 * ice_watchdog_subtask - periodic tasks not using event driven scheduling 1183 * @pf: board private structure 1184 */ 1185 static void ice_watchdog_subtask(struct ice_pf *pf) 1186 { 1187 int i; 1188 1189 /* if interface is down do nothing */ 1190 if (test_bit(ICE_DOWN, pf->state) || 1191 test_bit(ICE_CFG_BUSY, pf->state)) 1192 return; 1193 1194 /* make sure we don't do these things too often */ 1195 if (time_before(jiffies, 1196 pf->serv_tmr_prev + pf->serv_tmr_period)) 1197 return; 1198 1199 pf->serv_tmr_prev = jiffies; 1200 1201 /* Update the stats for active netdevs so the network stack 1202 * can look at updated numbers whenever it cares to 1203 */ 1204 ice_update_pf_stats(pf); 1205 ice_for_each_vsi(pf, i) 1206 if (pf->vsi[i] && pf->vsi[i]->netdev) 1207 ice_update_vsi_stats(pf->vsi[i]); 1208 } 1209 1210 /** 1211 * ice_init_link_events - enable/initialize link events 1212 * @pi: pointer to the port_info instance 1213 * 1214 * Returns -EIO on failure, 0 on success 1215 */ 1216 static int ice_init_link_events(struct ice_port_info *pi) 1217 { 1218 u16 mask; 1219 1220 mask = ~((u16)(ICE_AQ_LINK_EVENT_UPDOWN | ICE_AQ_LINK_EVENT_MEDIA_NA | 1221 ICE_AQ_LINK_EVENT_MODULE_QUAL_FAIL | 1222 ICE_AQ_LINK_EVENT_PHY_FW_LOAD_FAIL)); 1223 1224 if (ice_aq_set_event_mask(pi->hw, pi->lport, mask, NULL)) { 1225 dev_dbg(ice_hw_to_dev(pi->hw), "Failed to set link event mask for port %d\n", 1226 pi->lport); 1227 return -EIO; 1228 } 1229 1230 if (ice_aq_get_link_info(pi, true, NULL, NULL)) { 1231 dev_dbg(ice_hw_to_dev(pi->hw), "Failed to enable link events for port %d\n", 1232 pi->lport); 1233 return -EIO; 1234 } 1235 1236 return 0; 1237 } 1238 1239 /** 1240 * ice_handle_link_event - handle link event via ARQ 1241 * @pf: PF that the link event is associated with 1242 * @event: event structure containing link status info 1243 */ 1244 static int 1245 ice_handle_link_event(struct ice_pf *pf, struct ice_rq_event_info *event) 1246 { 1247 struct ice_aqc_get_link_status_data *link_data; 1248 struct ice_port_info *port_info; 1249 int status; 1250 1251 link_data = (struct ice_aqc_get_link_status_data *)event->msg_buf; 1252 port_info = pf->hw.port_info; 1253 if (!port_info) 1254 return -EINVAL; 1255 1256 status = ice_link_event(pf, port_info, 1257 !!(link_data->link_info & ICE_AQ_LINK_UP), 1258 le16_to_cpu(link_data->link_speed)); 1259 if (status) 1260 dev_dbg(ice_pf_to_dev(pf), "Could not process link event, error %d\n", 1261 status); 1262 1263 return status; 1264 } 1265 1266 /** 1267 * ice_get_fwlog_data - copy the FW log data from ARQ event 1268 * @pf: PF that the FW log event is associated with 1269 * @event: event structure containing FW log data 1270 */ 1271 static void 1272 ice_get_fwlog_data(struct ice_pf *pf, struct ice_rq_event_info *event) 1273 { 1274 struct ice_fwlog_data *fwlog; 1275 struct ice_hw *hw = &pf->hw; 1276 1277 fwlog = &hw->fwlog_ring.rings[hw->fwlog_ring.tail]; 1278 1279 memset(fwlog->data, 0, PAGE_SIZE); 1280 fwlog->data_size = le16_to_cpu(event->desc.datalen); 1281 1282 memcpy(fwlog->data, event->msg_buf, fwlog->data_size); 1283 ice_fwlog_ring_increment(&hw->fwlog_ring.tail, hw->fwlog_ring.size); 1284 1285 if (ice_fwlog_ring_full(&hw->fwlog_ring)) { 1286 /* the rings are full so bump the head to create room */ 1287 ice_fwlog_ring_increment(&hw->fwlog_ring.head, 1288 hw->fwlog_ring.size); 1289 } 1290 } 1291 1292 /** 1293 * ice_aq_prep_for_event - Prepare to wait for an AdminQ event from firmware 1294 * @pf: pointer to the PF private structure 1295 * @task: intermediate helper storage and identifier for waiting 1296 * @opcode: the opcode to wait for 1297 * 1298 * Prepares to wait for a specific AdminQ completion event on the ARQ for 1299 * a given PF. Actual wait would be done by a call to ice_aq_wait_for_event(). 1300 * 1301 * Calls are separated to allow caller registering for event before sending 1302 * the command, which mitigates a race between registering and FW responding. 1303 * 1304 * To obtain only the descriptor contents, pass an task->event with null 1305 * msg_buf. If the complete data buffer is desired, allocate the 1306 * task->event.msg_buf with enough space ahead of time. 1307 */ 1308 void ice_aq_prep_for_event(struct ice_pf *pf, struct ice_aq_task *task, 1309 u16 opcode) 1310 { 1311 INIT_HLIST_NODE(&task->entry); 1312 task->opcode = opcode; 1313 task->state = ICE_AQ_TASK_WAITING; 1314 1315 spin_lock_bh(&pf->aq_wait_lock); 1316 hlist_add_head(&task->entry, &pf->aq_wait_list); 1317 spin_unlock_bh(&pf->aq_wait_lock); 1318 } 1319 1320 /** 1321 * ice_aq_wait_for_event - Wait for an AdminQ event from firmware 1322 * @pf: pointer to the PF private structure 1323 * @task: ptr prepared by ice_aq_prep_for_event() 1324 * @timeout: how long to wait, in jiffies 1325 * 1326 * Waits for a specific AdminQ completion event on the ARQ for a given PF. The 1327 * current thread will be put to sleep until the specified event occurs or 1328 * until the given timeout is reached. 1329 * 1330 * Returns: zero on success, or a negative error code on failure. 1331 */ 1332 int ice_aq_wait_for_event(struct ice_pf *pf, struct ice_aq_task *task, 1333 unsigned long timeout) 1334 { 1335 enum ice_aq_task_state *state = &task->state; 1336 struct device *dev = ice_pf_to_dev(pf); 1337 unsigned long start = jiffies; 1338 long ret; 1339 int err; 1340 1341 ret = wait_event_interruptible_timeout(pf->aq_wait_queue, 1342 *state != ICE_AQ_TASK_WAITING, 1343 timeout); 1344 switch (*state) { 1345 case ICE_AQ_TASK_NOT_PREPARED: 1346 WARN(1, "call to %s without ice_aq_prep_for_event()", __func__); 1347 err = -EINVAL; 1348 break; 1349 case ICE_AQ_TASK_WAITING: 1350 err = ret < 0 ? ret : -ETIMEDOUT; 1351 break; 1352 case ICE_AQ_TASK_CANCELED: 1353 err = ret < 0 ? ret : -ECANCELED; 1354 break; 1355 case ICE_AQ_TASK_COMPLETE: 1356 err = ret < 0 ? ret : 0; 1357 break; 1358 default: 1359 WARN(1, "Unexpected AdminQ wait task state %u", *state); 1360 err = -EINVAL; 1361 break; 1362 } 1363 1364 dev_dbg(dev, "Waited %u msecs (max %u msecs) for firmware response to op 0x%04x\n", 1365 jiffies_to_msecs(jiffies - start), 1366 jiffies_to_msecs(timeout), 1367 task->opcode); 1368 1369 spin_lock_bh(&pf->aq_wait_lock); 1370 hlist_del(&task->entry); 1371 spin_unlock_bh(&pf->aq_wait_lock); 1372 1373 return err; 1374 } 1375 1376 /** 1377 * ice_aq_check_events - Check if any thread is waiting for an AdminQ event 1378 * @pf: pointer to the PF private structure 1379 * @opcode: the opcode of the event 1380 * @event: the event to check 1381 * 1382 * Loops over the current list of pending threads waiting for an AdminQ event. 1383 * For each matching task, copy the contents of the event into the task 1384 * structure and wake up the thread. 1385 * 1386 * If multiple threads wait for the same opcode, they will all be woken up. 1387 * 1388 * Note that event->msg_buf will only be duplicated if the event has a buffer 1389 * with enough space already allocated. Otherwise, only the descriptor and 1390 * message length will be copied. 1391 * 1392 * Returns: true if an event was found, false otherwise 1393 */ 1394 static void ice_aq_check_events(struct ice_pf *pf, u16 opcode, 1395 struct ice_rq_event_info *event) 1396 { 1397 struct ice_rq_event_info *task_ev; 1398 struct ice_aq_task *task; 1399 bool found = false; 1400 1401 spin_lock_bh(&pf->aq_wait_lock); 1402 hlist_for_each_entry(task, &pf->aq_wait_list, entry) { 1403 if (task->state != ICE_AQ_TASK_WAITING) 1404 continue; 1405 if (task->opcode != opcode) 1406 continue; 1407 1408 task_ev = &task->event; 1409 memcpy(&task_ev->desc, &event->desc, sizeof(event->desc)); 1410 task_ev->msg_len = event->msg_len; 1411 1412 /* Only copy the data buffer if a destination was set */ 1413 if (task_ev->msg_buf && task_ev->buf_len >= event->buf_len) { 1414 memcpy(task_ev->msg_buf, event->msg_buf, 1415 event->buf_len); 1416 task_ev->buf_len = event->buf_len; 1417 } 1418 1419 task->state = ICE_AQ_TASK_COMPLETE; 1420 found = true; 1421 } 1422 spin_unlock_bh(&pf->aq_wait_lock); 1423 1424 if (found) 1425 wake_up(&pf->aq_wait_queue); 1426 } 1427 1428 /** 1429 * ice_aq_cancel_waiting_tasks - Immediately cancel all waiting tasks 1430 * @pf: the PF private structure 1431 * 1432 * Set all waiting tasks to ICE_AQ_TASK_CANCELED, and wake up their threads. 1433 * This will then cause ice_aq_wait_for_event to exit with -ECANCELED. 1434 */ 1435 static void ice_aq_cancel_waiting_tasks(struct ice_pf *pf) 1436 { 1437 struct ice_aq_task *task; 1438 1439 spin_lock_bh(&pf->aq_wait_lock); 1440 hlist_for_each_entry(task, &pf->aq_wait_list, entry) 1441 task->state = ICE_AQ_TASK_CANCELED; 1442 spin_unlock_bh(&pf->aq_wait_lock); 1443 1444 wake_up(&pf->aq_wait_queue); 1445 } 1446 1447 #define ICE_MBX_OVERFLOW_WATERMARK 64 1448 1449 /** 1450 * __ice_clean_ctrlq - helper function to clean controlq rings 1451 * @pf: ptr to struct ice_pf 1452 * @q_type: specific Control queue type 1453 */ 1454 static int __ice_clean_ctrlq(struct ice_pf *pf, enum ice_ctl_q q_type) 1455 { 1456 struct device *dev = ice_pf_to_dev(pf); 1457 struct ice_rq_event_info event; 1458 struct ice_hw *hw = &pf->hw; 1459 struct ice_ctl_q_info *cq; 1460 u16 pending, i = 0; 1461 const char *qtype; 1462 u32 oldval, val; 1463 1464 /* Do not clean control queue if/when PF reset fails */ 1465 if (test_bit(ICE_RESET_FAILED, pf->state)) 1466 return 0; 1467 1468 switch (q_type) { 1469 case ICE_CTL_Q_ADMIN: 1470 cq = &hw->adminq; 1471 qtype = "Admin"; 1472 break; 1473 case ICE_CTL_Q_SB: 1474 cq = &hw->sbq; 1475 qtype = "Sideband"; 1476 break; 1477 case ICE_CTL_Q_MAILBOX: 1478 cq = &hw->mailboxq; 1479 qtype = "Mailbox"; 1480 /* we are going to try to detect a malicious VF, so set the 1481 * state to begin detection 1482 */ 1483 hw->mbx_snapshot.mbx_buf.state = ICE_MAL_VF_DETECT_STATE_NEW_SNAPSHOT; 1484 break; 1485 default: 1486 dev_warn(dev, "Unknown control queue type 0x%x\n", q_type); 1487 return 0; 1488 } 1489 1490 /* check for error indications - PF_xx_AxQLEN register layout for 1491 * FW/MBX/SB are identical so just use defines for PF_FW_AxQLEN. 1492 */ 1493 val = rd32(hw, cq->rq.len); 1494 if (val & (PF_FW_ARQLEN_ARQVFE_M | PF_FW_ARQLEN_ARQOVFL_M | 1495 PF_FW_ARQLEN_ARQCRIT_M)) { 1496 oldval = val; 1497 if (val & PF_FW_ARQLEN_ARQVFE_M) 1498 dev_dbg(dev, "%s Receive Queue VF Error detected\n", 1499 qtype); 1500 if (val & PF_FW_ARQLEN_ARQOVFL_M) { 1501 dev_dbg(dev, "%s Receive Queue Overflow Error detected\n", 1502 qtype); 1503 } 1504 if (val & PF_FW_ARQLEN_ARQCRIT_M) 1505 dev_dbg(dev, "%s Receive Queue Critical Error detected\n", 1506 qtype); 1507 val &= ~(PF_FW_ARQLEN_ARQVFE_M | PF_FW_ARQLEN_ARQOVFL_M | 1508 PF_FW_ARQLEN_ARQCRIT_M); 1509 if (oldval != val) 1510 wr32(hw, cq->rq.len, val); 1511 } 1512 1513 val = rd32(hw, cq->sq.len); 1514 if (val & (PF_FW_ATQLEN_ATQVFE_M | PF_FW_ATQLEN_ATQOVFL_M | 1515 PF_FW_ATQLEN_ATQCRIT_M)) { 1516 oldval = val; 1517 if (val & PF_FW_ATQLEN_ATQVFE_M) 1518 dev_dbg(dev, "%s Send Queue VF Error detected\n", 1519 qtype); 1520 if (val & PF_FW_ATQLEN_ATQOVFL_M) { 1521 dev_dbg(dev, "%s Send Queue Overflow Error detected\n", 1522 qtype); 1523 } 1524 if (val & PF_FW_ATQLEN_ATQCRIT_M) 1525 dev_dbg(dev, "%s Send Queue Critical Error detected\n", 1526 qtype); 1527 val &= ~(PF_FW_ATQLEN_ATQVFE_M | PF_FW_ATQLEN_ATQOVFL_M | 1528 PF_FW_ATQLEN_ATQCRIT_M); 1529 if (oldval != val) 1530 wr32(hw, cq->sq.len, val); 1531 } 1532 1533 event.buf_len = cq->rq_buf_size; 1534 event.msg_buf = kzalloc(event.buf_len, GFP_KERNEL); 1535 if (!event.msg_buf) 1536 return 0; 1537 1538 do { 1539 struct ice_mbx_data data = {}; 1540 u16 opcode; 1541 int ret; 1542 1543 ret = ice_clean_rq_elem(hw, cq, &event, &pending); 1544 if (ret == -EALREADY) 1545 break; 1546 if (ret) { 1547 dev_err(dev, "%s Receive Queue event error %d\n", qtype, 1548 ret); 1549 break; 1550 } 1551 1552 opcode = le16_to_cpu(event.desc.opcode); 1553 1554 /* Notify any thread that might be waiting for this event */ 1555 ice_aq_check_events(pf, opcode, &event); 1556 1557 switch (opcode) { 1558 case ice_aqc_opc_get_link_status: 1559 if (ice_handle_link_event(pf, &event)) 1560 dev_err(dev, "Could not handle link event\n"); 1561 break; 1562 case ice_aqc_opc_event_lan_overflow: 1563 ice_vf_lan_overflow_event(pf, &event); 1564 break; 1565 case ice_mbx_opc_send_msg_to_pf: 1566 data.num_msg_proc = i; 1567 data.num_pending_arq = pending; 1568 data.max_num_msgs_mbx = hw->mailboxq.num_rq_entries; 1569 data.async_watermark_val = ICE_MBX_OVERFLOW_WATERMARK; 1570 1571 ice_vc_process_vf_msg(pf, &event, &data); 1572 break; 1573 case ice_aqc_opc_fw_logs_event: 1574 ice_get_fwlog_data(pf, &event); 1575 break; 1576 case ice_aqc_opc_lldp_set_mib_change: 1577 ice_dcb_process_lldp_set_mib_change(pf, &event); 1578 break; 1579 default: 1580 dev_dbg(dev, "%s Receive Queue unknown event 0x%04x ignored\n", 1581 qtype, opcode); 1582 break; 1583 } 1584 } while (pending && (i++ < ICE_DFLT_IRQ_WORK)); 1585 1586 kfree(event.msg_buf); 1587 1588 return pending && (i == ICE_DFLT_IRQ_WORK); 1589 } 1590 1591 /** 1592 * ice_ctrlq_pending - check if there is a difference between ntc and ntu 1593 * @hw: pointer to hardware info 1594 * @cq: control queue information 1595 * 1596 * returns true if there are pending messages in a queue, false if there aren't 1597 */ 1598 static bool ice_ctrlq_pending(struct ice_hw *hw, struct ice_ctl_q_info *cq) 1599 { 1600 u16 ntu; 1601 1602 ntu = (u16)(rd32(hw, cq->rq.head) & cq->rq.head_mask); 1603 return cq->rq.next_to_clean != ntu; 1604 } 1605 1606 /** 1607 * ice_clean_adminq_subtask - clean the AdminQ rings 1608 * @pf: board private structure 1609 */ 1610 static void ice_clean_adminq_subtask(struct ice_pf *pf) 1611 { 1612 struct ice_hw *hw = &pf->hw; 1613 1614 if (!test_bit(ICE_ADMINQ_EVENT_PENDING, pf->state)) 1615 return; 1616 1617 if (__ice_clean_ctrlq(pf, ICE_CTL_Q_ADMIN)) 1618 return; 1619 1620 clear_bit(ICE_ADMINQ_EVENT_PENDING, pf->state); 1621 1622 /* There might be a situation where new messages arrive to a control 1623 * queue between processing the last message and clearing the 1624 * EVENT_PENDING bit. So before exiting, check queue head again (using 1625 * ice_ctrlq_pending) and process new messages if any. 1626 */ 1627 if (ice_ctrlq_pending(hw, &hw->adminq)) 1628 __ice_clean_ctrlq(pf, ICE_CTL_Q_ADMIN); 1629 1630 ice_flush(hw); 1631 } 1632 1633 /** 1634 * ice_clean_mailboxq_subtask - clean the MailboxQ rings 1635 * @pf: board private structure 1636 */ 1637 static void ice_clean_mailboxq_subtask(struct ice_pf *pf) 1638 { 1639 struct ice_hw *hw = &pf->hw; 1640 1641 if (!test_bit(ICE_MAILBOXQ_EVENT_PENDING, pf->state)) 1642 return; 1643 1644 if (__ice_clean_ctrlq(pf, ICE_CTL_Q_MAILBOX)) 1645 return; 1646 1647 clear_bit(ICE_MAILBOXQ_EVENT_PENDING, pf->state); 1648 1649 if (ice_ctrlq_pending(hw, &hw->mailboxq)) 1650 __ice_clean_ctrlq(pf, ICE_CTL_Q_MAILBOX); 1651 1652 ice_flush(hw); 1653 } 1654 1655 /** 1656 * ice_clean_sbq_subtask - clean the Sideband Queue rings 1657 * @pf: board private structure 1658 */ 1659 static void ice_clean_sbq_subtask(struct ice_pf *pf) 1660 { 1661 struct ice_hw *hw = &pf->hw; 1662 1663 /* if mac_type is not generic, sideband is not supported 1664 * and there's nothing to do here 1665 */ 1666 if (!ice_is_generic_mac(hw)) { 1667 clear_bit(ICE_SIDEBANDQ_EVENT_PENDING, pf->state); 1668 return; 1669 } 1670 1671 if (!test_bit(ICE_SIDEBANDQ_EVENT_PENDING, pf->state)) 1672 return; 1673 1674 if (__ice_clean_ctrlq(pf, ICE_CTL_Q_SB)) 1675 return; 1676 1677 clear_bit(ICE_SIDEBANDQ_EVENT_PENDING, pf->state); 1678 1679 if (ice_ctrlq_pending(hw, &hw->sbq)) 1680 __ice_clean_ctrlq(pf, ICE_CTL_Q_SB); 1681 1682 ice_flush(hw); 1683 } 1684 1685 /** 1686 * ice_service_task_schedule - schedule the service task to wake up 1687 * @pf: board private structure 1688 * 1689 * If not already scheduled, this puts the task into the work queue. 1690 */ 1691 void ice_service_task_schedule(struct ice_pf *pf) 1692 { 1693 if (!test_bit(ICE_SERVICE_DIS, pf->state) && 1694 !test_and_set_bit(ICE_SERVICE_SCHED, pf->state) && 1695 !test_bit(ICE_NEEDS_RESTART, pf->state)) 1696 queue_work(ice_wq, &pf->serv_task); 1697 } 1698 1699 /** 1700 * ice_service_task_complete - finish up the service task 1701 * @pf: board private structure 1702 */ 1703 static void ice_service_task_complete(struct ice_pf *pf) 1704 { 1705 WARN_ON(!test_bit(ICE_SERVICE_SCHED, pf->state)); 1706 1707 /* force memory (pf->state) to sync before next service task */ 1708 smp_mb__before_atomic(); 1709 clear_bit(ICE_SERVICE_SCHED, pf->state); 1710 } 1711 1712 /** 1713 * ice_service_task_stop - stop service task and cancel works 1714 * @pf: board private structure 1715 * 1716 * Return 0 if the ICE_SERVICE_DIS bit was not already set, 1717 * 1 otherwise. 1718 */ 1719 static int ice_service_task_stop(struct ice_pf *pf) 1720 { 1721 int ret; 1722 1723 ret = test_and_set_bit(ICE_SERVICE_DIS, pf->state); 1724 1725 if (pf->serv_tmr.function) 1726 del_timer_sync(&pf->serv_tmr); 1727 if (pf->serv_task.func) 1728 cancel_work_sync(&pf->serv_task); 1729 1730 clear_bit(ICE_SERVICE_SCHED, pf->state); 1731 return ret; 1732 } 1733 1734 /** 1735 * ice_service_task_restart - restart service task and schedule works 1736 * @pf: board private structure 1737 * 1738 * This function is needed for suspend and resume works (e.g WoL scenario) 1739 */ 1740 static void ice_service_task_restart(struct ice_pf *pf) 1741 { 1742 clear_bit(ICE_SERVICE_DIS, pf->state); 1743 ice_service_task_schedule(pf); 1744 } 1745 1746 /** 1747 * ice_service_timer - timer callback to schedule service task 1748 * @t: pointer to timer_list 1749 */ 1750 static void ice_service_timer(struct timer_list *t) 1751 { 1752 struct ice_pf *pf = from_timer(pf, t, serv_tmr); 1753 1754 mod_timer(&pf->serv_tmr, round_jiffies(pf->serv_tmr_period + jiffies)); 1755 ice_service_task_schedule(pf); 1756 } 1757 1758 /** 1759 * ice_mdd_maybe_reset_vf - reset VF after MDD event 1760 * @pf: pointer to the PF structure 1761 * @vf: pointer to the VF structure 1762 * @reset_vf_tx: whether Tx MDD has occurred 1763 * @reset_vf_rx: whether Rx MDD has occurred 1764 * 1765 * Since the queue can get stuck on VF MDD events, the PF can be configured to 1766 * automatically reset the VF by enabling the private ethtool flag 1767 * mdd-auto-reset-vf. 1768 */ 1769 static void ice_mdd_maybe_reset_vf(struct ice_pf *pf, struct ice_vf *vf, 1770 bool reset_vf_tx, bool reset_vf_rx) 1771 { 1772 struct device *dev = ice_pf_to_dev(pf); 1773 1774 if (!test_bit(ICE_FLAG_MDD_AUTO_RESET_VF, pf->flags)) 1775 return; 1776 1777 /* VF MDD event counters will be cleared by reset, so print the event 1778 * prior to reset. 1779 */ 1780 if (reset_vf_tx) 1781 ice_print_vf_tx_mdd_event(vf); 1782 1783 if (reset_vf_rx) 1784 ice_print_vf_rx_mdd_event(vf); 1785 1786 dev_info(dev, "PF-to-VF reset on PF %d VF %d due to MDD event\n", 1787 pf->hw.pf_id, vf->vf_id); 1788 ice_reset_vf(vf, ICE_VF_RESET_NOTIFY | ICE_VF_RESET_LOCK); 1789 } 1790 1791 /** 1792 * ice_handle_mdd_event - handle malicious driver detect event 1793 * @pf: pointer to the PF structure 1794 * 1795 * Called from service task. OICR interrupt handler indicates MDD event. 1796 * VF MDD logging is guarded by net_ratelimit. Additional PF and VF log 1797 * messages are wrapped by netif_msg_[rx|tx]_err. Since VF Rx MDD events 1798 * disable the queue, the PF can be configured to reset the VF using ethtool 1799 * private flag mdd-auto-reset-vf. 1800 */ 1801 static void ice_handle_mdd_event(struct ice_pf *pf) 1802 { 1803 struct device *dev = ice_pf_to_dev(pf); 1804 struct ice_hw *hw = &pf->hw; 1805 struct ice_vf *vf; 1806 unsigned int bkt; 1807 u32 reg; 1808 1809 if (!test_and_clear_bit(ICE_MDD_EVENT_PENDING, pf->state)) { 1810 /* Since the VF MDD event logging is rate limited, check if 1811 * there are pending MDD events. 1812 */ 1813 ice_print_vfs_mdd_events(pf); 1814 return; 1815 } 1816 1817 /* find what triggered an MDD event */ 1818 reg = rd32(hw, GL_MDET_TX_PQM); 1819 if (reg & GL_MDET_TX_PQM_VALID_M) { 1820 u8 pf_num = FIELD_GET(GL_MDET_TX_PQM_PF_NUM_M, reg); 1821 u16 vf_num = FIELD_GET(GL_MDET_TX_PQM_VF_NUM_M, reg); 1822 u8 event = FIELD_GET(GL_MDET_TX_PQM_MAL_TYPE_M, reg); 1823 u16 queue = FIELD_GET(GL_MDET_TX_PQM_QNUM_M, reg); 1824 1825 if (netif_msg_tx_err(pf)) 1826 dev_info(dev, "Malicious Driver Detection event %d on TX queue %d PF# %d VF# %d\n", 1827 event, queue, pf_num, vf_num); 1828 wr32(hw, GL_MDET_TX_PQM, 0xffffffff); 1829 } 1830 1831 reg = rd32(hw, GL_MDET_TX_TCLAN_BY_MAC(hw)); 1832 if (reg & GL_MDET_TX_TCLAN_VALID_M) { 1833 u8 pf_num = FIELD_GET(GL_MDET_TX_TCLAN_PF_NUM_M, reg); 1834 u16 vf_num = FIELD_GET(GL_MDET_TX_TCLAN_VF_NUM_M, reg); 1835 u8 event = FIELD_GET(GL_MDET_TX_TCLAN_MAL_TYPE_M, reg); 1836 u16 queue = FIELD_GET(GL_MDET_TX_TCLAN_QNUM_M, reg); 1837 1838 if (netif_msg_tx_err(pf)) 1839 dev_info(dev, "Malicious Driver Detection event %d on TX queue %d PF# %d VF# %d\n", 1840 event, queue, pf_num, vf_num); 1841 wr32(hw, GL_MDET_TX_TCLAN_BY_MAC(hw), U32_MAX); 1842 } 1843 1844 reg = rd32(hw, GL_MDET_RX); 1845 if (reg & GL_MDET_RX_VALID_M) { 1846 u8 pf_num = FIELD_GET(GL_MDET_RX_PF_NUM_M, reg); 1847 u16 vf_num = FIELD_GET(GL_MDET_RX_VF_NUM_M, reg); 1848 u8 event = FIELD_GET(GL_MDET_RX_MAL_TYPE_M, reg); 1849 u16 queue = FIELD_GET(GL_MDET_RX_QNUM_M, reg); 1850 1851 if (netif_msg_rx_err(pf)) 1852 dev_info(dev, "Malicious Driver Detection event %d on RX queue %d PF# %d VF# %d\n", 1853 event, queue, pf_num, vf_num); 1854 wr32(hw, GL_MDET_RX, 0xffffffff); 1855 } 1856 1857 /* check to see if this PF caused an MDD event */ 1858 reg = rd32(hw, PF_MDET_TX_PQM); 1859 if (reg & PF_MDET_TX_PQM_VALID_M) { 1860 wr32(hw, PF_MDET_TX_PQM, 0xFFFF); 1861 if (netif_msg_tx_err(pf)) 1862 dev_info(dev, "Malicious Driver Detection event TX_PQM detected on PF\n"); 1863 } 1864 1865 reg = rd32(hw, PF_MDET_TX_TCLAN_BY_MAC(hw)); 1866 if (reg & PF_MDET_TX_TCLAN_VALID_M) { 1867 wr32(hw, PF_MDET_TX_TCLAN_BY_MAC(hw), 0xffff); 1868 if (netif_msg_tx_err(pf)) 1869 dev_info(dev, "Malicious Driver Detection event TX_TCLAN detected on PF\n"); 1870 } 1871 1872 reg = rd32(hw, PF_MDET_RX); 1873 if (reg & PF_MDET_RX_VALID_M) { 1874 wr32(hw, PF_MDET_RX, 0xFFFF); 1875 if (netif_msg_rx_err(pf)) 1876 dev_info(dev, "Malicious Driver Detection event RX detected on PF\n"); 1877 } 1878 1879 /* Check to see if one of the VFs caused an MDD event, and then 1880 * increment counters and set print pending 1881 */ 1882 mutex_lock(&pf->vfs.table_lock); 1883 ice_for_each_vf(pf, bkt, vf) { 1884 bool reset_vf_tx = false, reset_vf_rx = false; 1885 1886 reg = rd32(hw, VP_MDET_TX_PQM(vf->vf_id)); 1887 if (reg & VP_MDET_TX_PQM_VALID_M) { 1888 wr32(hw, VP_MDET_TX_PQM(vf->vf_id), 0xFFFF); 1889 vf->mdd_tx_events.count++; 1890 set_bit(ICE_MDD_VF_PRINT_PENDING, pf->state); 1891 if (netif_msg_tx_err(pf)) 1892 dev_info(dev, "Malicious Driver Detection event TX_PQM detected on VF %d\n", 1893 vf->vf_id); 1894 1895 reset_vf_tx = true; 1896 } 1897 1898 reg = rd32(hw, VP_MDET_TX_TCLAN(vf->vf_id)); 1899 if (reg & VP_MDET_TX_TCLAN_VALID_M) { 1900 wr32(hw, VP_MDET_TX_TCLAN(vf->vf_id), 0xFFFF); 1901 vf->mdd_tx_events.count++; 1902 set_bit(ICE_MDD_VF_PRINT_PENDING, pf->state); 1903 if (netif_msg_tx_err(pf)) 1904 dev_info(dev, "Malicious Driver Detection event TX_TCLAN detected on VF %d\n", 1905 vf->vf_id); 1906 1907 reset_vf_tx = true; 1908 } 1909 1910 reg = rd32(hw, VP_MDET_TX_TDPU(vf->vf_id)); 1911 if (reg & VP_MDET_TX_TDPU_VALID_M) { 1912 wr32(hw, VP_MDET_TX_TDPU(vf->vf_id), 0xFFFF); 1913 vf->mdd_tx_events.count++; 1914 set_bit(ICE_MDD_VF_PRINT_PENDING, pf->state); 1915 if (netif_msg_tx_err(pf)) 1916 dev_info(dev, "Malicious Driver Detection event TX_TDPU detected on VF %d\n", 1917 vf->vf_id); 1918 1919 reset_vf_tx = true; 1920 } 1921 1922 reg = rd32(hw, VP_MDET_RX(vf->vf_id)); 1923 if (reg & VP_MDET_RX_VALID_M) { 1924 wr32(hw, VP_MDET_RX(vf->vf_id), 0xFFFF); 1925 vf->mdd_rx_events.count++; 1926 set_bit(ICE_MDD_VF_PRINT_PENDING, pf->state); 1927 if (netif_msg_rx_err(pf)) 1928 dev_info(dev, "Malicious Driver Detection event RX detected on VF %d\n", 1929 vf->vf_id); 1930 1931 reset_vf_rx = true; 1932 } 1933 1934 if (reset_vf_tx || reset_vf_rx) 1935 ice_mdd_maybe_reset_vf(pf, vf, reset_vf_tx, 1936 reset_vf_rx); 1937 } 1938 mutex_unlock(&pf->vfs.table_lock); 1939 1940 ice_print_vfs_mdd_events(pf); 1941 } 1942 1943 /** 1944 * ice_force_phys_link_state - Force the physical link state 1945 * @vsi: VSI to force the physical link state to up/down 1946 * @link_up: true/false indicates to set the physical link to up/down 1947 * 1948 * Force the physical link state by getting the current PHY capabilities from 1949 * hardware and setting the PHY config based on the determined capabilities. If 1950 * link changes a link event will be triggered because both the Enable Automatic 1951 * Link Update and LESM Enable bits are set when setting the PHY capabilities. 1952 * 1953 * Returns 0 on success, negative on failure 1954 */ 1955 static int ice_force_phys_link_state(struct ice_vsi *vsi, bool link_up) 1956 { 1957 struct ice_aqc_get_phy_caps_data *pcaps; 1958 struct ice_aqc_set_phy_cfg_data *cfg; 1959 struct ice_port_info *pi; 1960 struct device *dev; 1961 int retcode; 1962 1963 if (!vsi || !vsi->port_info || !vsi->back) 1964 return -EINVAL; 1965 if (vsi->type != ICE_VSI_PF) 1966 return 0; 1967 1968 dev = ice_pf_to_dev(vsi->back); 1969 1970 pi = vsi->port_info; 1971 1972 pcaps = kzalloc(sizeof(*pcaps), GFP_KERNEL); 1973 if (!pcaps) 1974 return -ENOMEM; 1975 1976 retcode = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_ACTIVE_CFG, pcaps, 1977 NULL); 1978 if (retcode) { 1979 dev_err(dev, "Failed to get phy capabilities, VSI %d error %d\n", 1980 vsi->vsi_num, retcode); 1981 retcode = -EIO; 1982 goto out; 1983 } 1984 1985 /* No change in link */ 1986 if (link_up == !!(pcaps->caps & ICE_AQC_PHY_EN_LINK) && 1987 link_up == !!(pi->phy.link_info.link_info & ICE_AQ_LINK_UP)) 1988 goto out; 1989 1990 /* Use the current user PHY configuration. The current user PHY 1991 * configuration is initialized during probe from PHY capabilities 1992 * software mode, and updated on set PHY configuration. 1993 */ 1994 cfg = kmemdup(&pi->phy.curr_user_phy_cfg, sizeof(*cfg), GFP_KERNEL); 1995 if (!cfg) { 1996 retcode = -ENOMEM; 1997 goto out; 1998 } 1999 2000 cfg->caps |= ICE_AQ_PHY_ENA_AUTO_LINK_UPDT; 2001 if (link_up) 2002 cfg->caps |= ICE_AQ_PHY_ENA_LINK; 2003 else 2004 cfg->caps &= ~ICE_AQ_PHY_ENA_LINK; 2005 2006 retcode = ice_aq_set_phy_cfg(&vsi->back->hw, pi, cfg, NULL); 2007 if (retcode) { 2008 dev_err(dev, "Failed to set phy config, VSI %d error %d\n", 2009 vsi->vsi_num, retcode); 2010 retcode = -EIO; 2011 } 2012 2013 kfree(cfg); 2014 out: 2015 kfree(pcaps); 2016 return retcode; 2017 } 2018 2019 /** 2020 * ice_init_nvm_phy_type - Initialize the NVM PHY type 2021 * @pi: port info structure 2022 * 2023 * Initialize nvm_phy_type_[low|high] for link lenient mode support 2024 */ 2025 static int ice_init_nvm_phy_type(struct ice_port_info *pi) 2026 { 2027 struct ice_aqc_get_phy_caps_data *pcaps; 2028 struct ice_pf *pf = pi->hw->back; 2029 int err; 2030 2031 pcaps = kzalloc(sizeof(*pcaps), GFP_KERNEL); 2032 if (!pcaps) 2033 return -ENOMEM; 2034 2035 err = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_TOPO_CAP_NO_MEDIA, 2036 pcaps, NULL); 2037 2038 if (err) { 2039 dev_err(ice_pf_to_dev(pf), "Get PHY capability failed.\n"); 2040 goto out; 2041 } 2042 2043 pf->nvm_phy_type_hi = pcaps->phy_type_high; 2044 pf->nvm_phy_type_lo = pcaps->phy_type_low; 2045 2046 out: 2047 kfree(pcaps); 2048 return err; 2049 } 2050 2051 /** 2052 * ice_init_link_dflt_override - Initialize link default override 2053 * @pi: port info structure 2054 * 2055 * Initialize link default override and PHY total port shutdown during probe 2056 */ 2057 static void ice_init_link_dflt_override(struct ice_port_info *pi) 2058 { 2059 struct ice_link_default_override_tlv *ldo; 2060 struct ice_pf *pf = pi->hw->back; 2061 2062 ldo = &pf->link_dflt_override; 2063 if (ice_get_link_default_override(ldo, pi)) 2064 return; 2065 2066 if (!(ldo->options & ICE_LINK_OVERRIDE_PORT_DIS)) 2067 return; 2068 2069 /* Enable Total Port Shutdown (override/replace link-down-on-close 2070 * ethtool private flag) for ports with Port Disable bit set. 2071 */ 2072 set_bit(ICE_FLAG_TOTAL_PORT_SHUTDOWN_ENA, pf->flags); 2073 set_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, pf->flags); 2074 } 2075 2076 /** 2077 * ice_init_phy_cfg_dflt_override - Initialize PHY cfg default override settings 2078 * @pi: port info structure 2079 * 2080 * If default override is enabled, initialize the user PHY cfg speed and FEC 2081 * settings using the default override mask from the NVM. 2082 * 2083 * The PHY should only be configured with the default override settings the 2084 * first time media is available. The ICE_LINK_DEFAULT_OVERRIDE_PENDING state 2085 * is used to indicate that the user PHY cfg default override is initialized 2086 * and the PHY has not been configured with the default override settings. The 2087 * state is set here, and cleared in ice_configure_phy the first time the PHY is 2088 * configured. 2089 * 2090 * This function should be called only if the FW doesn't support default 2091 * configuration mode, as reported by ice_fw_supports_report_dflt_cfg. 2092 */ 2093 static void ice_init_phy_cfg_dflt_override(struct ice_port_info *pi) 2094 { 2095 struct ice_link_default_override_tlv *ldo; 2096 struct ice_aqc_set_phy_cfg_data *cfg; 2097 struct ice_phy_info *phy = &pi->phy; 2098 struct ice_pf *pf = pi->hw->back; 2099 2100 ldo = &pf->link_dflt_override; 2101 2102 /* If link default override is enabled, use to mask NVM PHY capabilities 2103 * for speed and FEC default configuration. 2104 */ 2105 cfg = &phy->curr_user_phy_cfg; 2106 2107 if (ldo->phy_type_low || ldo->phy_type_high) { 2108 cfg->phy_type_low = pf->nvm_phy_type_lo & 2109 cpu_to_le64(ldo->phy_type_low); 2110 cfg->phy_type_high = pf->nvm_phy_type_hi & 2111 cpu_to_le64(ldo->phy_type_high); 2112 } 2113 cfg->link_fec_opt = ldo->fec_options; 2114 phy->curr_user_fec_req = ICE_FEC_AUTO; 2115 2116 set_bit(ICE_LINK_DEFAULT_OVERRIDE_PENDING, pf->state); 2117 } 2118 2119 /** 2120 * ice_init_phy_user_cfg - Initialize the PHY user configuration 2121 * @pi: port info structure 2122 * 2123 * Initialize the current user PHY configuration, speed, FEC, and FC requested 2124 * mode to default. The PHY defaults are from get PHY capabilities topology 2125 * with media so call when media is first available. An error is returned if 2126 * called when media is not available. The PHY initialization completed state is 2127 * set here. 2128 * 2129 * These configurations are used when setting PHY 2130 * configuration. The user PHY configuration is updated on set PHY 2131 * configuration. Returns 0 on success, negative on failure 2132 */ 2133 static int ice_init_phy_user_cfg(struct ice_port_info *pi) 2134 { 2135 struct ice_aqc_get_phy_caps_data *pcaps; 2136 struct ice_phy_info *phy = &pi->phy; 2137 struct ice_pf *pf = pi->hw->back; 2138 int err; 2139 2140 if (!(phy->link_info.link_info & ICE_AQ_MEDIA_AVAILABLE)) 2141 return -EIO; 2142 2143 pcaps = kzalloc(sizeof(*pcaps), GFP_KERNEL); 2144 if (!pcaps) 2145 return -ENOMEM; 2146 2147 if (ice_fw_supports_report_dflt_cfg(pi->hw)) 2148 err = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_DFLT_CFG, 2149 pcaps, NULL); 2150 else 2151 err = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_TOPO_CAP_MEDIA, 2152 pcaps, NULL); 2153 if (err) { 2154 dev_err(ice_pf_to_dev(pf), "Get PHY capability failed.\n"); 2155 goto err_out; 2156 } 2157 2158 ice_copy_phy_caps_to_cfg(pi, pcaps, &pi->phy.curr_user_phy_cfg); 2159 2160 /* check if lenient mode is supported and enabled */ 2161 if (ice_fw_supports_link_override(pi->hw) && 2162 !(pcaps->module_compliance_enforcement & 2163 ICE_AQC_MOD_ENFORCE_STRICT_MODE)) { 2164 set_bit(ICE_FLAG_LINK_LENIENT_MODE_ENA, pf->flags); 2165 2166 /* if the FW supports default PHY configuration mode, then the driver 2167 * does not have to apply link override settings. If not, 2168 * initialize user PHY configuration with link override values 2169 */ 2170 if (!ice_fw_supports_report_dflt_cfg(pi->hw) && 2171 (pf->link_dflt_override.options & ICE_LINK_OVERRIDE_EN)) { 2172 ice_init_phy_cfg_dflt_override(pi); 2173 goto out; 2174 } 2175 } 2176 2177 /* if link default override is not enabled, set user flow control and 2178 * FEC settings based on what get_phy_caps returned 2179 */ 2180 phy->curr_user_fec_req = ice_caps_to_fec_mode(pcaps->caps, 2181 pcaps->link_fec_options); 2182 phy->curr_user_fc_req = ice_caps_to_fc_mode(pcaps->caps); 2183 2184 out: 2185 phy->curr_user_speed_req = ICE_AQ_LINK_SPEED_M; 2186 set_bit(ICE_PHY_INIT_COMPLETE, pf->state); 2187 err_out: 2188 kfree(pcaps); 2189 return err; 2190 } 2191 2192 /** 2193 * ice_configure_phy - configure PHY 2194 * @vsi: VSI of PHY 2195 * 2196 * Set the PHY configuration. If the current PHY configuration is the same as 2197 * the curr_user_phy_cfg, then do nothing to avoid link flap. Otherwise 2198 * configure the based get PHY capabilities for topology with media. 2199 */ 2200 static int ice_configure_phy(struct ice_vsi *vsi) 2201 { 2202 struct device *dev = ice_pf_to_dev(vsi->back); 2203 struct ice_port_info *pi = vsi->port_info; 2204 struct ice_aqc_get_phy_caps_data *pcaps; 2205 struct ice_aqc_set_phy_cfg_data *cfg; 2206 struct ice_phy_info *phy = &pi->phy; 2207 struct ice_pf *pf = vsi->back; 2208 int err; 2209 2210 /* Ensure we have media as we cannot configure a medialess port */ 2211 if (!(phy->link_info.link_info & ICE_AQ_MEDIA_AVAILABLE)) 2212 return -ENOMEDIUM; 2213 2214 ice_print_topo_conflict(vsi); 2215 2216 if (!test_bit(ICE_FLAG_LINK_LENIENT_MODE_ENA, pf->flags) && 2217 phy->link_info.topo_media_conflict == ICE_AQ_LINK_TOPO_UNSUPP_MEDIA) 2218 return -EPERM; 2219 2220 if (test_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, pf->flags)) 2221 return ice_force_phys_link_state(vsi, true); 2222 2223 pcaps = kzalloc(sizeof(*pcaps), GFP_KERNEL); 2224 if (!pcaps) 2225 return -ENOMEM; 2226 2227 /* Get current PHY config */ 2228 err = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_ACTIVE_CFG, pcaps, 2229 NULL); 2230 if (err) { 2231 dev_err(dev, "Failed to get PHY configuration, VSI %d error %d\n", 2232 vsi->vsi_num, err); 2233 goto done; 2234 } 2235 2236 /* If PHY enable link is configured and configuration has not changed, 2237 * there's nothing to do 2238 */ 2239 if (pcaps->caps & ICE_AQC_PHY_EN_LINK && 2240 ice_phy_caps_equals_cfg(pcaps, &phy->curr_user_phy_cfg)) 2241 goto done; 2242 2243 /* Use PHY topology as baseline for configuration */ 2244 memset(pcaps, 0, sizeof(*pcaps)); 2245 if (ice_fw_supports_report_dflt_cfg(pi->hw)) 2246 err = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_DFLT_CFG, 2247 pcaps, NULL); 2248 else 2249 err = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_TOPO_CAP_MEDIA, 2250 pcaps, NULL); 2251 if (err) { 2252 dev_err(dev, "Failed to get PHY caps, VSI %d error %d\n", 2253 vsi->vsi_num, err); 2254 goto done; 2255 } 2256 2257 cfg = kzalloc(sizeof(*cfg), GFP_KERNEL); 2258 if (!cfg) { 2259 err = -ENOMEM; 2260 goto done; 2261 } 2262 2263 ice_copy_phy_caps_to_cfg(pi, pcaps, cfg); 2264 2265 /* Speed - If default override pending, use curr_user_phy_cfg set in 2266 * ice_init_phy_user_cfg_ldo. 2267 */ 2268 if (test_and_clear_bit(ICE_LINK_DEFAULT_OVERRIDE_PENDING, 2269 vsi->back->state)) { 2270 cfg->phy_type_low = phy->curr_user_phy_cfg.phy_type_low; 2271 cfg->phy_type_high = phy->curr_user_phy_cfg.phy_type_high; 2272 } else { 2273 u64 phy_low = 0, phy_high = 0; 2274 2275 ice_update_phy_type(&phy_low, &phy_high, 2276 pi->phy.curr_user_speed_req); 2277 cfg->phy_type_low = pcaps->phy_type_low & cpu_to_le64(phy_low); 2278 cfg->phy_type_high = pcaps->phy_type_high & 2279 cpu_to_le64(phy_high); 2280 } 2281 2282 /* Can't provide what was requested; use PHY capabilities */ 2283 if (!cfg->phy_type_low && !cfg->phy_type_high) { 2284 cfg->phy_type_low = pcaps->phy_type_low; 2285 cfg->phy_type_high = pcaps->phy_type_high; 2286 } 2287 2288 /* FEC */ 2289 ice_cfg_phy_fec(pi, cfg, phy->curr_user_fec_req); 2290 2291 /* Can't provide what was requested; use PHY capabilities */ 2292 if (cfg->link_fec_opt != 2293 (cfg->link_fec_opt & pcaps->link_fec_options)) { 2294 cfg->caps |= pcaps->caps & ICE_AQC_PHY_EN_AUTO_FEC; 2295 cfg->link_fec_opt = pcaps->link_fec_options; 2296 } 2297 2298 /* Flow Control - always supported; no need to check against 2299 * capabilities 2300 */ 2301 ice_cfg_phy_fc(pi, cfg, phy->curr_user_fc_req); 2302 2303 /* Enable link and link update */ 2304 cfg->caps |= ICE_AQ_PHY_ENA_AUTO_LINK_UPDT | ICE_AQ_PHY_ENA_LINK; 2305 2306 err = ice_aq_set_phy_cfg(&pf->hw, pi, cfg, NULL); 2307 if (err) 2308 dev_err(dev, "Failed to set phy config, VSI %d error %d\n", 2309 vsi->vsi_num, err); 2310 2311 kfree(cfg); 2312 done: 2313 kfree(pcaps); 2314 return err; 2315 } 2316 2317 /** 2318 * ice_check_media_subtask - Check for media 2319 * @pf: pointer to PF struct 2320 * 2321 * If media is available, then initialize PHY user configuration if it is not 2322 * been, and configure the PHY if the interface is up. 2323 */ 2324 static void ice_check_media_subtask(struct ice_pf *pf) 2325 { 2326 struct ice_port_info *pi; 2327 struct ice_vsi *vsi; 2328 int err; 2329 2330 /* No need to check for media if it's already present */ 2331 if (!test_bit(ICE_FLAG_NO_MEDIA, pf->flags)) 2332 return; 2333 2334 vsi = ice_get_main_vsi(pf); 2335 if (!vsi) 2336 return; 2337 2338 /* Refresh link info and check if media is present */ 2339 pi = vsi->port_info; 2340 err = ice_update_link_info(pi); 2341 if (err) 2342 return; 2343 2344 ice_check_link_cfg_err(pf, pi->phy.link_info.link_cfg_err); 2345 2346 if (pi->phy.link_info.link_info & ICE_AQ_MEDIA_AVAILABLE) { 2347 if (!test_bit(ICE_PHY_INIT_COMPLETE, pf->state)) 2348 ice_init_phy_user_cfg(pi); 2349 2350 /* PHY settings are reset on media insertion, reconfigure 2351 * PHY to preserve settings. 2352 */ 2353 if (test_bit(ICE_VSI_DOWN, vsi->state) && 2354 test_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, vsi->back->flags)) 2355 return; 2356 2357 err = ice_configure_phy(vsi); 2358 if (!err) 2359 clear_bit(ICE_FLAG_NO_MEDIA, pf->flags); 2360 2361 /* A Link Status Event will be generated; the event handler 2362 * will complete bringing the interface up 2363 */ 2364 } 2365 } 2366 2367 /** 2368 * ice_service_task - manage and run subtasks 2369 * @work: pointer to work_struct contained by the PF struct 2370 */ 2371 static void ice_service_task(struct work_struct *work) 2372 { 2373 struct ice_pf *pf = container_of(work, struct ice_pf, serv_task); 2374 unsigned long start_time = jiffies; 2375 2376 /* subtasks */ 2377 2378 /* process reset requests first */ 2379 ice_reset_subtask(pf); 2380 2381 /* bail if a reset/recovery cycle is pending or rebuild failed */ 2382 if (ice_is_reset_in_progress(pf->state) || 2383 test_bit(ICE_SUSPENDED, pf->state) || 2384 test_bit(ICE_NEEDS_RESTART, pf->state)) { 2385 ice_service_task_complete(pf); 2386 return; 2387 } 2388 2389 if (test_and_clear_bit(ICE_AUX_ERR_PENDING, pf->state)) { 2390 struct iidc_event *event; 2391 2392 event = kzalloc(sizeof(*event), GFP_KERNEL); 2393 if (event) { 2394 set_bit(IIDC_EVENT_CRIT_ERR, event->type); 2395 /* report the entire OICR value to AUX driver */ 2396 swap(event->reg, pf->oicr_err_reg); 2397 ice_send_event_to_aux(pf, event); 2398 kfree(event); 2399 } 2400 } 2401 2402 /* unplug aux dev per request, if an unplug request came in 2403 * while processing a plug request, this will handle it 2404 */ 2405 if (test_and_clear_bit(ICE_FLAG_UNPLUG_AUX_DEV, pf->flags)) 2406 ice_unplug_aux_dev(pf); 2407 2408 /* Plug aux device per request */ 2409 if (test_and_clear_bit(ICE_FLAG_PLUG_AUX_DEV, pf->flags)) 2410 ice_plug_aux_dev(pf); 2411 2412 if (test_and_clear_bit(ICE_FLAG_MTU_CHANGED, pf->flags)) { 2413 struct iidc_event *event; 2414 2415 event = kzalloc(sizeof(*event), GFP_KERNEL); 2416 if (event) { 2417 set_bit(IIDC_EVENT_AFTER_MTU_CHANGE, event->type); 2418 ice_send_event_to_aux(pf, event); 2419 kfree(event); 2420 } 2421 } 2422 2423 ice_clean_adminq_subtask(pf); 2424 ice_check_media_subtask(pf); 2425 ice_check_for_hang_subtask(pf); 2426 ice_sync_fltr_subtask(pf); 2427 ice_handle_mdd_event(pf); 2428 ice_watchdog_subtask(pf); 2429 2430 if (ice_is_safe_mode(pf)) { 2431 ice_service_task_complete(pf); 2432 return; 2433 } 2434 2435 ice_process_vflr_event(pf); 2436 ice_clean_mailboxq_subtask(pf); 2437 ice_clean_sbq_subtask(pf); 2438 ice_sync_arfs_fltrs(pf); 2439 ice_flush_fdir_ctx(pf); 2440 2441 /* Clear ICE_SERVICE_SCHED flag to allow scheduling next event */ 2442 ice_service_task_complete(pf); 2443 2444 /* If the tasks have taken longer than one service timer period 2445 * or there is more work to be done, reset the service timer to 2446 * schedule the service task now. 2447 */ 2448 if (time_after(jiffies, (start_time + pf->serv_tmr_period)) || 2449 test_bit(ICE_MDD_EVENT_PENDING, pf->state) || 2450 test_bit(ICE_VFLR_EVENT_PENDING, pf->state) || 2451 test_bit(ICE_MAILBOXQ_EVENT_PENDING, pf->state) || 2452 test_bit(ICE_FD_VF_FLUSH_CTX, pf->state) || 2453 test_bit(ICE_SIDEBANDQ_EVENT_PENDING, pf->state) || 2454 test_bit(ICE_ADMINQ_EVENT_PENDING, pf->state)) 2455 mod_timer(&pf->serv_tmr, jiffies); 2456 } 2457 2458 /** 2459 * ice_set_ctrlq_len - helper function to set controlq length 2460 * @hw: pointer to the HW instance 2461 */ 2462 static void ice_set_ctrlq_len(struct ice_hw *hw) 2463 { 2464 hw->adminq.num_rq_entries = ICE_AQ_LEN; 2465 hw->adminq.num_sq_entries = ICE_AQ_LEN; 2466 hw->adminq.rq_buf_size = ICE_AQ_MAX_BUF_LEN; 2467 hw->adminq.sq_buf_size = ICE_AQ_MAX_BUF_LEN; 2468 hw->mailboxq.num_rq_entries = PF_MBX_ARQLEN_ARQLEN_M; 2469 hw->mailboxq.num_sq_entries = ICE_MBXSQ_LEN; 2470 hw->mailboxq.rq_buf_size = ICE_MBXQ_MAX_BUF_LEN; 2471 hw->mailboxq.sq_buf_size = ICE_MBXQ_MAX_BUF_LEN; 2472 hw->sbq.num_rq_entries = ICE_SBQ_LEN; 2473 hw->sbq.num_sq_entries = ICE_SBQ_LEN; 2474 hw->sbq.rq_buf_size = ICE_SBQ_MAX_BUF_LEN; 2475 hw->sbq.sq_buf_size = ICE_SBQ_MAX_BUF_LEN; 2476 } 2477 2478 /** 2479 * ice_schedule_reset - schedule a reset 2480 * @pf: board private structure 2481 * @reset: reset being requested 2482 */ 2483 int ice_schedule_reset(struct ice_pf *pf, enum ice_reset_req reset) 2484 { 2485 struct device *dev = ice_pf_to_dev(pf); 2486 2487 /* bail out if earlier reset has failed */ 2488 if (test_bit(ICE_RESET_FAILED, pf->state)) { 2489 dev_dbg(dev, "earlier reset has failed\n"); 2490 return -EIO; 2491 } 2492 /* bail if reset/recovery already in progress */ 2493 if (ice_is_reset_in_progress(pf->state)) { 2494 dev_dbg(dev, "Reset already in progress\n"); 2495 return -EBUSY; 2496 } 2497 2498 switch (reset) { 2499 case ICE_RESET_PFR: 2500 set_bit(ICE_PFR_REQ, pf->state); 2501 break; 2502 case ICE_RESET_CORER: 2503 set_bit(ICE_CORER_REQ, pf->state); 2504 break; 2505 case ICE_RESET_GLOBR: 2506 set_bit(ICE_GLOBR_REQ, pf->state); 2507 break; 2508 default: 2509 return -EINVAL; 2510 } 2511 2512 ice_service_task_schedule(pf); 2513 return 0; 2514 } 2515 2516 /** 2517 * ice_irq_affinity_notify - Callback for affinity changes 2518 * @notify: context as to what irq was changed 2519 * @mask: the new affinity mask 2520 * 2521 * This is a callback function used by the irq_set_affinity_notifier function 2522 * so that we may register to receive changes to the irq affinity masks. 2523 */ 2524 static void 2525 ice_irq_affinity_notify(struct irq_affinity_notify *notify, 2526 const cpumask_t *mask) 2527 { 2528 struct ice_q_vector *q_vector = 2529 container_of(notify, struct ice_q_vector, affinity_notify); 2530 2531 cpumask_copy(&q_vector->affinity_mask, mask); 2532 } 2533 2534 /** 2535 * ice_irq_affinity_release - Callback for affinity notifier release 2536 * @ref: internal core kernel usage 2537 * 2538 * This is a callback function used by the irq_set_affinity_notifier function 2539 * to inform the current notification subscriber that they will no longer 2540 * receive notifications. 2541 */ 2542 static void ice_irq_affinity_release(struct kref __always_unused *ref) {} 2543 2544 /** 2545 * ice_vsi_ena_irq - Enable IRQ for the given VSI 2546 * @vsi: the VSI being configured 2547 */ 2548 static int ice_vsi_ena_irq(struct ice_vsi *vsi) 2549 { 2550 struct ice_hw *hw = &vsi->back->hw; 2551 int i; 2552 2553 ice_for_each_q_vector(vsi, i) 2554 ice_irq_dynamic_ena(hw, vsi, vsi->q_vectors[i]); 2555 2556 ice_flush(hw); 2557 return 0; 2558 } 2559 2560 /** 2561 * ice_vsi_req_irq_msix - get MSI-X vectors from the OS for the VSI 2562 * @vsi: the VSI being configured 2563 * @basename: name for the vector 2564 */ 2565 static int ice_vsi_req_irq_msix(struct ice_vsi *vsi, char *basename) 2566 { 2567 int q_vectors = vsi->num_q_vectors; 2568 struct ice_pf *pf = vsi->back; 2569 struct device *dev; 2570 int rx_int_idx = 0; 2571 int tx_int_idx = 0; 2572 int vector, err; 2573 int irq_num; 2574 2575 dev = ice_pf_to_dev(pf); 2576 for (vector = 0; vector < q_vectors; vector++) { 2577 struct ice_q_vector *q_vector = vsi->q_vectors[vector]; 2578 2579 irq_num = q_vector->irq.virq; 2580 2581 if (q_vector->tx.tx_ring && q_vector->rx.rx_ring) { 2582 snprintf(q_vector->name, sizeof(q_vector->name) - 1, 2583 "%s-%s-%d", basename, "TxRx", rx_int_idx++); 2584 tx_int_idx++; 2585 } else if (q_vector->rx.rx_ring) { 2586 snprintf(q_vector->name, sizeof(q_vector->name) - 1, 2587 "%s-%s-%d", basename, "rx", rx_int_idx++); 2588 } else if (q_vector->tx.tx_ring) { 2589 snprintf(q_vector->name, sizeof(q_vector->name) - 1, 2590 "%s-%s-%d", basename, "tx", tx_int_idx++); 2591 } else { 2592 /* skip this unused q_vector */ 2593 continue; 2594 } 2595 if (vsi->type == ICE_VSI_CTRL && vsi->vf) 2596 err = devm_request_irq(dev, irq_num, vsi->irq_handler, 2597 IRQF_SHARED, q_vector->name, 2598 q_vector); 2599 else 2600 err = devm_request_irq(dev, irq_num, vsi->irq_handler, 2601 0, q_vector->name, q_vector); 2602 if (err) { 2603 netdev_err(vsi->netdev, "MSIX request_irq failed, error: %d\n", 2604 err); 2605 goto free_q_irqs; 2606 } 2607 2608 /* register for affinity change notifications */ 2609 if (!IS_ENABLED(CONFIG_RFS_ACCEL)) { 2610 struct irq_affinity_notify *affinity_notify; 2611 2612 affinity_notify = &q_vector->affinity_notify; 2613 affinity_notify->notify = ice_irq_affinity_notify; 2614 affinity_notify->release = ice_irq_affinity_release; 2615 irq_set_affinity_notifier(irq_num, affinity_notify); 2616 } 2617 2618 /* assign the mask for this irq */ 2619 irq_update_affinity_hint(irq_num, &q_vector->affinity_mask); 2620 } 2621 2622 err = ice_set_cpu_rx_rmap(vsi); 2623 if (err) { 2624 netdev_err(vsi->netdev, "Failed to setup CPU RMAP on VSI %u: %pe\n", 2625 vsi->vsi_num, ERR_PTR(err)); 2626 goto free_q_irqs; 2627 } 2628 2629 vsi->irqs_ready = true; 2630 return 0; 2631 2632 free_q_irqs: 2633 while (vector--) { 2634 irq_num = vsi->q_vectors[vector]->irq.virq; 2635 if (!IS_ENABLED(CONFIG_RFS_ACCEL)) 2636 irq_set_affinity_notifier(irq_num, NULL); 2637 irq_update_affinity_hint(irq_num, NULL); 2638 devm_free_irq(dev, irq_num, &vsi->q_vectors[vector]); 2639 } 2640 return err; 2641 } 2642 2643 /** 2644 * ice_xdp_alloc_setup_rings - Allocate and setup Tx rings for XDP 2645 * @vsi: VSI to setup Tx rings used by XDP 2646 * 2647 * Return 0 on success and negative value on error 2648 */ 2649 static int ice_xdp_alloc_setup_rings(struct ice_vsi *vsi) 2650 { 2651 struct device *dev = ice_pf_to_dev(vsi->back); 2652 struct ice_tx_desc *tx_desc; 2653 int i, j; 2654 2655 ice_for_each_xdp_txq(vsi, i) { 2656 u16 xdp_q_idx = vsi->alloc_txq + i; 2657 struct ice_ring_stats *ring_stats; 2658 struct ice_tx_ring *xdp_ring; 2659 2660 xdp_ring = kzalloc(sizeof(*xdp_ring), GFP_KERNEL); 2661 if (!xdp_ring) 2662 goto free_xdp_rings; 2663 2664 ring_stats = kzalloc(sizeof(*ring_stats), GFP_KERNEL); 2665 if (!ring_stats) { 2666 ice_free_tx_ring(xdp_ring); 2667 goto free_xdp_rings; 2668 } 2669 2670 xdp_ring->ring_stats = ring_stats; 2671 xdp_ring->q_index = xdp_q_idx; 2672 xdp_ring->reg_idx = vsi->txq_map[xdp_q_idx]; 2673 xdp_ring->vsi = vsi; 2674 xdp_ring->netdev = NULL; 2675 xdp_ring->dev = dev; 2676 xdp_ring->count = vsi->num_tx_desc; 2677 WRITE_ONCE(vsi->xdp_rings[i], xdp_ring); 2678 if (ice_setup_tx_ring(xdp_ring)) 2679 goto free_xdp_rings; 2680 ice_set_ring_xdp(xdp_ring); 2681 spin_lock_init(&xdp_ring->tx_lock); 2682 for (j = 0; j < xdp_ring->count; j++) { 2683 tx_desc = ICE_TX_DESC(xdp_ring, j); 2684 tx_desc->cmd_type_offset_bsz = 0; 2685 } 2686 } 2687 2688 return 0; 2689 2690 free_xdp_rings: 2691 for (; i >= 0; i--) { 2692 if (vsi->xdp_rings[i] && vsi->xdp_rings[i]->desc) { 2693 kfree_rcu(vsi->xdp_rings[i]->ring_stats, rcu); 2694 vsi->xdp_rings[i]->ring_stats = NULL; 2695 ice_free_tx_ring(vsi->xdp_rings[i]); 2696 } 2697 } 2698 return -ENOMEM; 2699 } 2700 2701 /** 2702 * ice_vsi_assign_bpf_prog - set or clear bpf prog pointer on VSI 2703 * @vsi: VSI to set the bpf prog on 2704 * @prog: the bpf prog pointer 2705 */ 2706 static void ice_vsi_assign_bpf_prog(struct ice_vsi *vsi, struct bpf_prog *prog) 2707 { 2708 struct bpf_prog *old_prog; 2709 int i; 2710 2711 old_prog = xchg(&vsi->xdp_prog, prog); 2712 ice_for_each_rxq(vsi, i) 2713 WRITE_ONCE(vsi->rx_rings[i]->xdp_prog, vsi->xdp_prog); 2714 2715 if (old_prog) 2716 bpf_prog_put(old_prog); 2717 } 2718 2719 static struct ice_tx_ring *ice_xdp_ring_from_qid(struct ice_vsi *vsi, int qid) 2720 { 2721 struct ice_q_vector *q_vector; 2722 struct ice_tx_ring *ring; 2723 2724 if (static_key_enabled(&ice_xdp_locking_key)) 2725 return vsi->xdp_rings[qid % vsi->num_xdp_txq]; 2726 2727 q_vector = vsi->rx_rings[qid]->q_vector; 2728 ice_for_each_tx_ring(ring, q_vector->tx) 2729 if (ice_ring_is_xdp(ring)) 2730 return ring; 2731 2732 return NULL; 2733 } 2734 2735 /** 2736 * ice_map_xdp_rings - Map XDP rings to interrupt vectors 2737 * @vsi: the VSI with XDP rings being configured 2738 * 2739 * Map XDP rings to interrupt vectors and perform the configuration steps 2740 * dependent on the mapping. 2741 */ 2742 void ice_map_xdp_rings(struct ice_vsi *vsi) 2743 { 2744 int xdp_rings_rem = vsi->num_xdp_txq; 2745 int v_idx, q_idx; 2746 2747 /* follow the logic from ice_vsi_map_rings_to_vectors */ 2748 ice_for_each_q_vector(vsi, v_idx) { 2749 struct ice_q_vector *q_vector = vsi->q_vectors[v_idx]; 2750 int xdp_rings_per_v, q_id, q_base; 2751 2752 xdp_rings_per_v = DIV_ROUND_UP(xdp_rings_rem, 2753 vsi->num_q_vectors - v_idx); 2754 q_base = vsi->num_xdp_txq - xdp_rings_rem; 2755 2756 for (q_id = q_base; q_id < (q_base + xdp_rings_per_v); q_id++) { 2757 struct ice_tx_ring *xdp_ring = vsi->xdp_rings[q_id]; 2758 2759 xdp_ring->q_vector = q_vector; 2760 xdp_ring->next = q_vector->tx.tx_ring; 2761 q_vector->tx.tx_ring = xdp_ring; 2762 } 2763 xdp_rings_rem -= xdp_rings_per_v; 2764 } 2765 2766 ice_for_each_rxq(vsi, q_idx) { 2767 vsi->rx_rings[q_idx]->xdp_ring = ice_xdp_ring_from_qid(vsi, 2768 q_idx); 2769 ice_tx_xsk_pool(vsi, q_idx); 2770 } 2771 } 2772 2773 /** 2774 * ice_prepare_xdp_rings - Allocate, configure and setup Tx rings for XDP 2775 * @vsi: VSI to bring up Tx rings used by XDP 2776 * @prog: bpf program that will be assigned to VSI 2777 * @cfg_type: create from scratch or restore the existing configuration 2778 * 2779 * Return 0 on success and negative value on error 2780 */ 2781 int ice_prepare_xdp_rings(struct ice_vsi *vsi, struct bpf_prog *prog, 2782 enum ice_xdp_cfg cfg_type) 2783 { 2784 u16 max_txqs[ICE_MAX_TRAFFIC_CLASS] = { 0 }; 2785 struct ice_pf *pf = vsi->back; 2786 struct ice_qs_cfg xdp_qs_cfg = { 2787 .qs_mutex = &pf->avail_q_mutex, 2788 .pf_map = pf->avail_txqs, 2789 .pf_map_size = pf->max_pf_txqs, 2790 .q_count = vsi->num_xdp_txq, 2791 .scatter_count = ICE_MAX_SCATTER_TXQS, 2792 .vsi_map = vsi->txq_map, 2793 .vsi_map_offset = vsi->alloc_txq, 2794 .mapping_mode = ICE_VSI_MAP_CONTIG 2795 }; 2796 struct device *dev; 2797 int status, i; 2798 2799 dev = ice_pf_to_dev(pf); 2800 vsi->xdp_rings = devm_kcalloc(dev, vsi->num_xdp_txq, 2801 sizeof(*vsi->xdp_rings), GFP_KERNEL); 2802 if (!vsi->xdp_rings) 2803 return -ENOMEM; 2804 2805 vsi->xdp_mapping_mode = xdp_qs_cfg.mapping_mode; 2806 if (__ice_vsi_get_qs(&xdp_qs_cfg)) 2807 goto err_map_xdp; 2808 2809 if (static_key_enabled(&ice_xdp_locking_key)) 2810 netdev_warn(vsi->netdev, 2811 "Could not allocate one XDP Tx ring per CPU, XDP_TX/XDP_REDIRECT actions will be slower\n"); 2812 2813 if (ice_xdp_alloc_setup_rings(vsi)) 2814 goto clear_xdp_rings; 2815 2816 /* omit the scheduler update if in reset path; XDP queues will be 2817 * taken into account at the end of ice_vsi_rebuild, where 2818 * ice_cfg_vsi_lan is being called 2819 */ 2820 if (cfg_type == ICE_XDP_CFG_PART) 2821 return 0; 2822 2823 ice_map_xdp_rings(vsi); 2824 2825 /* tell the Tx scheduler that right now we have 2826 * additional queues 2827 */ 2828 for (i = 0; i < vsi->tc_cfg.numtc; i++) 2829 max_txqs[i] = vsi->num_txq + vsi->num_xdp_txq; 2830 2831 status = ice_cfg_vsi_lan(vsi->port_info, vsi->idx, vsi->tc_cfg.ena_tc, 2832 max_txqs); 2833 if (status) { 2834 dev_err(dev, "Failed VSI LAN queue config for XDP, error: %d\n", 2835 status); 2836 goto clear_xdp_rings; 2837 } 2838 2839 /* assign the prog only when it's not already present on VSI; 2840 * this flow is a subject of both ethtool -L and ndo_bpf flows; 2841 * VSI rebuild that happens under ethtool -L can expose us to 2842 * the bpf_prog refcount issues as we would be swapping same 2843 * bpf_prog pointers from vsi->xdp_prog and calling bpf_prog_put 2844 * on it as it would be treated as an 'old_prog'; for ndo_bpf 2845 * this is not harmful as dev_xdp_install bumps the refcount 2846 * before calling the op exposed by the driver; 2847 */ 2848 if (!ice_is_xdp_ena_vsi(vsi)) 2849 ice_vsi_assign_bpf_prog(vsi, prog); 2850 2851 return 0; 2852 clear_xdp_rings: 2853 ice_for_each_xdp_txq(vsi, i) 2854 if (vsi->xdp_rings[i]) { 2855 kfree_rcu(vsi->xdp_rings[i], rcu); 2856 vsi->xdp_rings[i] = NULL; 2857 } 2858 2859 err_map_xdp: 2860 mutex_lock(&pf->avail_q_mutex); 2861 ice_for_each_xdp_txq(vsi, i) { 2862 clear_bit(vsi->txq_map[i + vsi->alloc_txq], pf->avail_txqs); 2863 vsi->txq_map[i + vsi->alloc_txq] = ICE_INVAL_Q_INDEX; 2864 } 2865 mutex_unlock(&pf->avail_q_mutex); 2866 2867 devm_kfree(dev, vsi->xdp_rings); 2868 return -ENOMEM; 2869 } 2870 2871 /** 2872 * ice_destroy_xdp_rings - undo the configuration made by ice_prepare_xdp_rings 2873 * @vsi: VSI to remove XDP rings 2874 * @cfg_type: disable XDP permanently or allow it to be restored later 2875 * 2876 * Detach XDP rings from irq vectors, clean up the PF bitmap and free 2877 * resources 2878 */ 2879 int ice_destroy_xdp_rings(struct ice_vsi *vsi, enum ice_xdp_cfg cfg_type) 2880 { 2881 u16 max_txqs[ICE_MAX_TRAFFIC_CLASS] = { 0 }; 2882 struct ice_pf *pf = vsi->back; 2883 int i, v_idx; 2884 2885 /* q_vectors are freed in reset path so there's no point in detaching 2886 * rings 2887 */ 2888 if (cfg_type == ICE_XDP_CFG_PART) 2889 goto free_qmap; 2890 2891 ice_for_each_q_vector(vsi, v_idx) { 2892 struct ice_q_vector *q_vector = vsi->q_vectors[v_idx]; 2893 struct ice_tx_ring *ring; 2894 2895 ice_for_each_tx_ring(ring, q_vector->tx) 2896 if (!ring->tx_buf || !ice_ring_is_xdp(ring)) 2897 break; 2898 2899 /* restore the value of last node prior to XDP setup */ 2900 q_vector->tx.tx_ring = ring; 2901 } 2902 2903 free_qmap: 2904 mutex_lock(&pf->avail_q_mutex); 2905 ice_for_each_xdp_txq(vsi, i) { 2906 clear_bit(vsi->txq_map[i + vsi->alloc_txq], pf->avail_txqs); 2907 vsi->txq_map[i + vsi->alloc_txq] = ICE_INVAL_Q_INDEX; 2908 } 2909 mutex_unlock(&pf->avail_q_mutex); 2910 2911 ice_for_each_xdp_txq(vsi, i) 2912 if (vsi->xdp_rings[i]) { 2913 if (vsi->xdp_rings[i]->desc) { 2914 synchronize_rcu(); 2915 ice_free_tx_ring(vsi->xdp_rings[i]); 2916 } 2917 kfree_rcu(vsi->xdp_rings[i]->ring_stats, rcu); 2918 vsi->xdp_rings[i]->ring_stats = NULL; 2919 kfree_rcu(vsi->xdp_rings[i], rcu); 2920 vsi->xdp_rings[i] = NULL; 2921 } 2922 2923 devm_kfree(ice_pf_to_dev(pf), vsi->xdp_rings); 2924 vsi->xdp_rings = NULL; 2925 2926 if (static_key_enabled(&ice_xdp_locking_key)) 2927 static_branch_dec(&ice_xdp_locking_key); 2928 2929 if (cfg_type == ICE_XDP_CFG_PART) 2930 return 0; 2931 2932 ice_vsi_assign_bpf_prog(vsi, NULL); 2933 2934 /* notify Tx scheduler that we destroyed XDP queues and bring 2935 * back the old number of child nodes 2936 */ 2937 for (i = 0; i < vsi->tc_cfg.numtc; i++) 2938 max_txqs[i] = vsi->num_txq; 2939 2940 /* change number of XDP Tx queues to 0 */ 2941 vsi->num_xdp_txq = 0; 2942 2943 return ice_cfg_vsi_lan(vsi->port_info, vsi->idx, vsi->tc_cfg.ena_tc, 2944 max_txqs); 2945 } 2946 2947 /** 2948 * ice_vsi_rx_napi_schedule - Schedule napi on RX queues from VSI 2949 * @vsi: VSI to schedule napi on 2950 */ 2951 static void ice_vsi_rx_napi_schedule(struct ice_vsi *vsi) 2952 { 2953 int i; 2954 2955 ice_for_each_rxq(vsi, i) { 2956 struct ice_rx_ring *rx_ring = vsi->rx_rings[i]; 2957 2958 if (READ_ONCE(rx_ring->xsk_pool)) 2959 napi_schedule(&rx_ring->q_vector->napi); 2960 } 2961 } 2962 2963 /** 2964 * ice_vsi_determine_xdp_res - figure out how many Tx qs can XDP have 2965 * @vsi: VSI to determine the count of XDP Tx qs 2966 * 2967 * returns 0 if Tx qs count is higher than at least half of CPU count, 2968 * -ENOMEM otherwise 2969 */ 2970 int ice_vsi_determine_xdp_res(struct ice_vsi *vsi) 2971 { 2972 u16 avail = ice_get_avail_txq_count(vsi->back); 2973 u16 cpus = num_possible_cpus(); 2974 2975 if (avail < cpus / 2) 2976 return -ENOMEM; 2977 2978 if (vsi->type == ICE_VSI_SF) 2979 avail = vsi->alloc_txq; 2980 2981 vsi->num_xdp_txq = min_t(u16, avail, cpus); 2982 2983 if (vsi->num_xdp_txq < cpus) 2984 static_branch_inc(&ice_xdp_locking_key); 2985 2986 return 0; 2987 } 2988 2989 /** 2990 * ice_max_xdp_frame_size - returns the maximum allowed frame size for XDP 2991 * @vsi: Pointer to VSI structure 2992 */ 2993 static int ice_max_xdp_frame_size(struct ice_vsi *vsi) 2994 { 2995 if (test_bit(ICE_FLAG_LEGACY_RX, vsi->back->flags)) 2996 return ICE_RXBUF_1664; 2997 else 2998 return ICE_RXBUF_3072; 2999 } 3000 3001 /** 3002 * ice_xdp_setup_prog - Add or remove XDP eBPF program 3003 * @vsi: VSI to setup XDP for 3004 * @prog: XDP program 3005 * @extack: netlink extended ack 3006 */ 3007 static int 3008 ice_xdp_setup_prog(struct ice_vsi *vsi, struct bpf_prog *prog, 3009 struct netlink_ext_ack *extack) 3010 { 3011 unsigned int frame_size = vsi->netdev->mtu + ICE_ETH_PKT_HDR_PAD; 3012 int ret = 0, xdp_ring_err = 0; 3013 bool if_running; 3014 3015 if (prog && !prog->aux->xdp_has_frags) { 3016 if (frame_size > ice_max_xdp_frame_size(vsi)) { 3017 NL_SET_ERR_MSG_MOD(extack, 3018 "MTU is too large for linear frames and XDP prog does not support frags"); 3019 return -EOPNOTSUPP; 3020 } 3021 } 3022 3023 /* hot swap progs and avoid toggling link */ 3024 if (ice_is_xdp_ena_vsi(vsi) == !!prog || 3025 test_bit(ICE_VSI_REBUILD_PENDING, vsi->state)) { 3026 ice_vsi_assign_bpf_prog(vsi, prog); 3027 return 0; 3028 } 3029 3030 if_running = netif_running(vsi->netdev) && 3031 !test_and_set_bit(ICE_VSI_DOWN, vsi->state); 3032 3033 /* need to stop netdev while setting up the program for Rx rings */ 3034 if (if_running) { 3035 ret = ice_down(vsi); 3036 if (ret) { 3037 NL_SET_ERR_MSG_MOD(extack, "Preparing device for XDP attach failed"); 3038 return ret; 3039 } 3040 } 3041 3042 if (!ice_is_xdp_ena_vsi(vsi) && prog) { 3043 xdp_ring_err = ice_vsi_determine_xdp_res(vsi); 3044 if (xdp_ring_err) { 3045 NL_SET_ERR_MSG_MOD(extack, "Not enough Tx resources for XDP"); 3046 } else { 3047 xdp_ring_err = ice_prepare_xdp_rings(vsi, prog, 3048 ICE_XDP_CFG_FULL); 3049 if (xdp_ring_err) 3050 NL_SET_ERR_MSG_MOD(extack, "Setting up XDP Tx resources failed"); 3051 } 3052 xdp_features_set_redirect_target(vsi->netdev, true); 3053 /* reallocate Rx queues that are used for zero-copy */ 3054 xdp_ring_err = ice_realloc_zc_buf(vsi, true); 3055 if (xdp_ring_err) 3056 NL_SET_ERR_MSG_MOD(extack, "Setting up XDP Rx resources failed"); 3057 } else if (ice_is_xdp_ena_vsi(vsi) && !prog) { 3058 xdp_features_clear_redirect_target(vsi->netdev); 3059 xdp_ring_err = ice_destroy_xdp_rings(vsi, ICE_XDP_CFG_FULL); 3060 if (xdp_ring_err) 3061 NL_SET_ERR_MSG_MOD(extack, "Freeing XDP Tx resources failed"); 3062 /* reallocate Rx queues that were used for zero-copy */ 3063 xdp_ring_err = ice_realloc_zc_buf(vsi, false); 3064 if (xdp_ring_err) 3065 NL_SET_ERR_MSG_MOD(extack, "Freeing XDP Rx resources failed"); 3066 } 3067 3068 if (if_running) 3069 ret = ice_up(vsi); 3070 3071 if (!ret && prog) 3072 ice_vsi_rx_napi_schedule(vsi); 3073 3074 return (ret || xdp_ring_err) ? -ENOMEM : 0; 3075 } 3076 3077 /** 3078 * ice_xdp_safe_mode - XDP handler for safe mode 3079 * @dev: netdevice 3080 * @xdp: XDP command 3081 */ 3082 static int ice_xdp_safe_mode(struct net_device __always_unused *dev, 3083 struct netdev_bpf *xdp) 3084 { 3085 NL_SET_ERR_MSG_MOD(xdp->extack, 3086 "Please provide working DDP firmware package in order to use XDP\n" 3087 "Refer to Documentation/networking/device_drivers/ethernet/intel/ice.rst"); 3088 return -EOPNOTSUPP; 3089 } 3090 3091 /** 3092 * ice_xdp - implements XDP handler 3093 * @dev: netdevice 3094 * @xdp: XDP command 3095 */ 3096 int ice_xdp(struct net_device *dev, struct netdev_bpf *xdp) 3097 { 3098 struct ice_netdev_priv *np = netdev_priv(dev); 3099 struct ice_vsi *vsi = np->vsi; 3100 int ret; 3101 3102 if (vsi->type != ICE_VSI_PF && vsi->type != ICE_VSI_SF) { 3103 NL_SET_ERR_MSG_MOD(xdp->extack, "XDP can be loaded only on PF or SF VSI"); 3104 return -EINVAL; 3105 } 3106 3107 mutex_lock(&vsi->xdp_state_lock); 3108 3109 switch (xdp->command) { 3110 case XDP_SETUP_PROG: 3111 ret = ice_xdp_setup_prog(vsi, xdp->prog, xdp->extack); 3112 break; 3113 case XDP_SETUP_XSK_POOL: 3114 ret = ice_xsk_pool_setup(vsi, xdp->xsk.pool, xdp->xsk.queue_id); 3115 break; 3116 default: 3117 ret = -EINVAL; 3118 } 3119 3120 mutex_unlock(&vsi->xdp_state_lock); 3121 return ret; 3122 } 3123 3124 /** 3125 * ice_ena_misc_vector - enable the non-queue interrupts 3126 * @pf: board private structure 3127 */ 3128 static void ice_ena_misc_vector(struct ice_pf *pf) 3129 { 3130 struct ice_hw *hw = &pf->hw; 3131 u32 pf_intr_start_offset; 3132 u32 val; 3133 3134 /* Disable anti-spoof detection interrupt to prevent spurious event 3135 * interrupts during a function reset. Anti-spoof functionally is 3136 * still supported. 3137 */ 3138 val = rd32(hw, GL_MDCK_TX_TDPU); 3139 val |= GL_MDCK_TX_TDPU_RCU_ANTISPOOF_ITR_DIS_M; 3140 wr32(hw, GL_MDCK_TX_TDPU, val); 3141 3142 /* clear things first */ 3143 wr32(hw, PFINT_OICR_ENA, 0); /* disable all */ 3144 rd32(hw, PFINT_OICR); /* read to clear */ 3145 3146 val = (PFINT_OICR_ECC_ERR_M | 3147 PFINT_OICR_MAL_DETECT_M | 3148 PFINT_OICR_GRST_M | 3149 PFINT_OICR_PCI_EXCEPTION_M | 3150 PFINT_OICR_VFLR_M | 3151 PFINT_OICR_HMC_ERR_M | 3152 PFINT_OICR_PE_PUSH_M | 3153 PFINT_OICR_PE_CRITERR_M); 3154 3155 wr32(hw, PFINT_OICR_ENA, val); 3156 3157 /* SW_ITR_IDX = 0, but don't change INTENA */ 3158 wr32(hw, GLINT_DYN_CTL(pf->oicr_irq.index), 3159 GLINT_DYN_CTL_SW_ITR_INDX_M | GLINT_DYN_CTL_INTENA_MSK_M); 3160 3161 if (!pf->hw.dev_caps.ts_dev_info.ts_ll_int_read) 3162 return; 3163 pf_intr_start_offset = rd32(hw, PFINT_ALLOC) & PFINT_ALLOC_FIRST; 3164 wr32(hw, GLINT_DYN_CTL(pf->ll_ts_irq.index + pf_intr_start_offset), 3165 GLINT_DYN_CTL_SW_ITR_INDX_M | GLINT_DYN_CTL_INTENA_MSK_M); 3166 } 3167 3168 /** 3169 * ice_ll_ts_intr - ll_ts interrupt handler 3170 * @irq: interrupt number 3171 * @data: pointer to a q_vector 3172 */ 3173 static irqreturn_t ice_ll_ts_intr(int __always_unused irq, void *data) 3174 { 3175 struct ice_pf *pf = data; 3176 u32 pf_intr_start_offset; 3177 struct ice_ptp_tx *tx; 3178 unsigned long flags; 3179 struct ice_hw *hw; 3180 u32 val; 3181 u8 idx; 3182 3183 hw = &pf->hw; 3184 tx = &pf->ptp.port.tx; 3185 spin_lock_irqsave(&tx->lock, flags); 3186 ice_ptp_complete_tx_single_tstamp(tx); 3187 3188 idx = find_next_bit_wrap(tx->in_use, tx->len, 3189 tx->last_ll_ts_idx_read + 1); 3190 if (idx != tx->len) 3191 ice_ptp_req_tx_single_tstamp(tx, idx); 3192 spin_unlock_irqrestore(&tx->lock, flags); 3193 3194 val = GLINT_DYN_CTL_INTENA_M | GLINT_DYN_CTL_CLEARPBA_M | 3195 (ICE_ITR_NONE << GLINT_DYN_CTL_ITR_INDX_S); 3196 pf_intr_start_offset = rd32(hw, PFINT_ALLOC) & PFINT_ALLOC_FIRST; 3197 wr32(hw, GLINT_DYN_CTL(pf->ll_ts_irq.index + pf_intr_start_offset), 3198 val); 3199 3200 return IRQ_HANDLED; 3201 } 3202 3203 /** 3204 * ice_misc_intr - misc interrupt handler 3205 * @irq: interrupt number 3206 * @data: pointer to a q_vector 3207 */ 3208 static irqreturn_t ice_misc_intr(int __always_unused irq, void *data) 3209 { 3210 struct ice_pf *pf = (struct ice_pf *)data; 3211 irqreturn_t ret = IRQ_HANDLED; 3212 struct ice_hw *hw = &pf->hw; 3213 struct device *dev; 3214 u32 oicr, ena_mask; 3215 3216 dev = ice_pf_to_dev(pf); 3217 set_bit(ICE_ADMINQ_EVENT_PENDING, pf->state); 3218 set_bit(ICE_MAILBOXQ_EVENT_PENDING, pf->state); 3219 set_bit(ICE_SIDEBANDQ_EVENT_PENDING, pf->state); 3220 3221 oicr = rd32(hw, PFINT_OICR); 3222 ena_mask = rd32(hw, PFINT_OICR_ENA); 3223 3224 if (oicr & PFINT_OICR_SWINT_M) { 3225 ena_mask &= ~PFINT_OICR_SWINT_M; 3226 pf->sw_int_count++; 3227 } 3228 3229 if (oicr & PFINT_OICR_MAL_DETECT_M) { 3230 ena_mask &= ~PFINT_OICR_MAL_DETECT_M; 3231 set_bit(ICE_MDD_EVENT_PENDING, pf->state); 3232 } 3233 if (oicr & PFINT_OICR_VFLR_M) { 3234 /* disable any further VFLR event notifications */ 3235 if (test_bit(ICE_VF_RESETS_DISABLED, pf->state)) { 3236 u32 reg = rd32(hw, PFINT_OICR_ENA); 3237 3238 reg &= ~PFINT_OICR_VFLR_M; 3239 wr32(hw, PFINT_OICR_ENA, reg); 3240 } else { 3241 ena_mask &= ~PFINT_OICR_VFLR_M; 3242 set_bit(ICE_VFLR_EVENT_PENDING, pf->state); 3243 } 3244 } 3245 3246 if (oicr & PFINT_OICR_GRST_M) { 3247 u32 reset; 3248 3249 /* we have a reset warning */ 3250 ena_mask &= ~PFINT_OICR_GRST_M; 3251 reset = FIELD_GET(GLGEN_RSTAT_RESET_TYPE_M, 3252 rd32(hw, GLGEN_RSTAT)); 3253 3254 if (reset == ICE_RESET_CORER) 3255 pf->corer_count++; 3256 else if (reset == ICE_RESET_GLOBR) 3257 pf->globr_count++; 3258 else if (reset == ICE_RESET_EMPR) 3259 pf->empr_count++; 3260 else 3261 dev_dbg(dev, "Invalid reset type %d\n", reset); 3262 3263 /* If a reset cycle isn't already in progress, we set a bit in 3264 * pf->state so that the service task can start a reset/rebuild. 3265 */ 3266 if (!test_and_set_bit(ICE_RESET_OICR_RECV, pf->state)) { 3267 if (reset == ICE_RESET_CORER) 3268 set_bit(ICE_CORER_RECV, pf->state); 3269 else if (reset == ICE_RESET_GLOBR) 3270 set_bit(ICE_GLOBR_RECV, pf->state); 3271 else 3272 set_bit(ICE_EMPR_RECV, pf->state); 3273 3274 /* There are couple of different bits at play here. 3275 * hw->reset_ongoing indicates whether the hardware is 3276 * in reset. This is set to true when a reset interrupt 3277 * is received and set back to false after the driver 3278 * has determined that the hardware is out of reset. 3279 * 3280 * ICE_RESET_OICR_RECV in pf->state indicates 3281 * that a post reset rebuild is required before the 3282 * driver is operational again. This is set above. 3283 * 3284 * As this is the start of the reset/rebuild cycle, set 3285 * both to indicate that. 3286 */ 3287 hw->reset_ongoing = true; 3288 } 3289 } 3290 3291 if (oicr & PFINT_OICR_TSYN_TX_M) { 3292 ena_mask &= ~PFINT_OICR_TSYN_TX_M; 3293 if (ice_pf_state_is_nominal(pf) && 3294 pf->hw.dev_caps.ts_dev_info.ts_ll_int_read) { 3295 struct ice_ptp_tx *tx = &pf->ptp.port.tx; 3296 unsigned long flags; 3297 u8 idx; 3298 3299 spin_lock_irqsave(&tx->lock, flags); 3300 idx = find_next_bit_wrap(tx->in_use, tx->len, 3301 tx->last_ll_ts_idx_read + 1); 3302 if (idx != tx->len) 3303 ice_ptp_req_tx_single_tstamp(tx, idx); 3304 spin_unlock_irqrestore(&tx->lock, flags); 3305 } else if (ice_ptp_pf_handles_tx_interrupt(pf)) { 3306 set_bit(ICE_MISC_THREAD_TX_TSTAMP, pf->misc_thread); 3307 ret = IRQ_WAKE_THREAD; 3308 } 3309 } 3310 3311 if (oicr & PFINT_OICR_TSYN_EVNT_M) { 3312 u8 tmr_idx = hw->func_caps.ts_func_info.tmr_index_owned; 3313 u32 gltsyn_stat = rd32(hw, GLTSYN_STAT(tmr_idx)); 3314 3315 ena_mask &= ~PFINT_OICR_TSYN_EVNT_M; 3316 3317 if (ice_pf_src_tmr_owned(pf)) { 3318 /* Save EVENTs from GLTSYN register */ 3319 pf->ptp.ext_ts_irq |= gltsyn_stat & 3320 (GLTSYN_STAT_EVENT0_M | 3321 GLTSYN_STAT_EVENT1_M | 3322 GLTSYN_STAT_EVENT2_M); 3323 3324 ice_ptp_extts_event(pf); 3325 } 3326 } 3327 3328 #define ICE_AUX_CRIT_ERR (PFINT_OICR_PE_CRITERR_M | PFINT_OICR_HMC_ERR_M | PFINT_OICR_PE_PUSH_M) 3329 if (oicr & ICE_AUX_CRIT_ERR) { 3330 pf->oicr_err_reg |= oicr; 3331 set_bit(ICE_AUX_ERR_PENDING, pf->state); 3332 ena_mask &= ~ICE_AUX_CRIT_ERR; 3333 } 3334 3335 /* Report any remaining unexpected interrupts */ 3336 oicr &= ena_mask; 3337 if (oicr) { 3338 dev_dbg(dev, "unhandled interrupt oicr=0x%08x\n", oicr); 3339 /* If a critical error is pending there is no choice but to 3340 * reset the device. 3341 */ 3342 if (oicr & (PFINT_OICR_PCI_EXCEPTION_M | 3343 PFINT_OICR_ECC_ERR_M)) { 3344 set_bit(ICE_PFR_REQ, pf->state); 3345 } 3346 } 3347 ice_service_task_schedule(pf); 3348 if (ret == IRQ_HANDLED) 3349 ice_irq_dynamic_ena(hw, NULL, NULL); 3350 3351 return ret; 3352 } 3353 3354 /** 3355 * ice_misc_intr_thread_fn - misc interrupt thread function 3356 * @irq: interrupt number 3357 * @data: pointer to a q_vector 3358 */ 3359 static irqreturn_t ice_misc_intr_thread_fn(int __always_unused irq, void *data) 3360 { 3361 struct ice_pf *pf = data; 3362 struct ice_hw *hw; 3363 3364 hw = &pf->hw; 3365 3366 if (ice_is_reset_in_progress(pf->state)) 3367 goto skip_irq; 3368 3369 if (test_and_clear_bit(ICE_MISC_THREAD_TX_TSTAMP, pf->misc_thread)) { 3370 /* Process outstanding Tx timestamps. If there is more work, 3371 * re-arm the interrupt to trigger again. 3372 */ 3373 if (ice_ptp_process_ts(pf) == ICE_TX_TSTAMP_WORK_PENDING) { 3374 wr32(hw, PFINT_OICR, PFINT_OICR_TSYN_TX_M); 3375 ice_flush(hw); 3376 } 3377 } 3378 3379 skip_irq: 3380 ice_irq_dynamic_ena(hw, NULL, NULL); 3381 3382 return IRQ_HANDLED; 3383 } 3384 3385 /** 3386 * ice_dis_ctrlq_interrupts - disable control queue interrupts 3387 * @hw: pointer to HW structure 3388 */ 3389 static void ice_dis_ctrlq_interrupts(struct ice_hw *hw) 3390 { 3391 /* disable Admin queue Interrupt causes */ 3392 wr32(hw, PFINT_FW_CTL, 3393 rd32(hw, PFINT_FW_CTL) & ~PFINT_FW_CTL_CAUSE_ENA_M); 3394 3395 /* disable Mailbox queue Interrupt causes */ 3396 wr32(hw, PFINT_MBX_CTL, 3397 rd32(hw, PFINT_MBX_CTL) & ~PFINT_MBX_CTL_CAUSE_ENA_M); 3398 3399 wr32(hw, PFINT_SB_CTL, 3400 rd32(hw, PFINT_SB_CTL) & ~PFINT_SB_CTL_CAUSE_ENA_M); 3401 3402 /* disable Control queue Interrupt causes */ 3403 wr32(hw, PFINT_OICR_CTL, 3404 rd32(hw, PFINT_OICR_CTL) & ~PFINT_OICR_CTL_CAUSE_ENA_M); 3405 3406 ice_flush(hw); 3407 } 3408 3409 /** 3410 * ice_free_irq_msix_ll_ts- Unroll ll_ts vector setup 3411 * @pf: board private structure 3412 */ 3413 static void ice_free_irq_msix_ll_ts(struct ice_pf *pf) 3414 { 3415 int irq_num = pf->ll_ts_irq.virq; 3416 3417 synchronize_irq(irq_num); 3418 devm_free_irq(ice_pf_to_dev(pf), irq_num, pf); 3419 3420 ice_free_irq(pf, pf->ll_ts_irq); 3421 } 3422 3423 /** 3424 * ice_free_irq_msix_misc - Unroll misc vector setup 3425 * @pf: board private structure 3426 */ 3427 static void ice_free_irq_msix_misc(struct ice_pf *pf) 3428 { 3429 int misc_irq_num = pf->oicr_irq.virq; 3430 struct ice_hw *hw = &pf->hw; 3431 3432 ice_dis_ctrlq_interrupts(hw); 3433 3434 /* disable OICR interrupt */ 3435 wr32(hw, PFINT_OICR_ENA, 0); 3436 ice_flush(hw); 3437 3438 synchronize_irq(misc_irq_num); 3439 devm_free_irq(ice_pf_to_dev(pf), misc_irq_num, pf); 3440 3441 ice_free_irq(pf, pf->oicr_irq); 3442 if (pf->hw.dev_caps.ts_dev_info.ts_ll_int_read) 3443 ice_free_irq_msix_ll_ts(pf); 3444 } 3445 3446 /** 3447 * ice_ena_ctrlq_interrupts - enable control queue interrupts 3448 * @hw: pointer to HW structure 3449 * @reg_idx: HW vector index to associate the control queue interrupts with 3450 */ 3451 static void ice_ena_ctrlq_interrupts(struct ice_hw *hw, u16 reg_idx) 3452 { 3453 u32 val; 3454 3455 val = ((reg_idx & PFINT_OICR_CTL_MSIX_INDX_M) | 3456 PFINT_OICR_CTL_CAUSE_ENA_M); 3457 wr32(hw, PFINT_OICR_CTL, val); 3458 3459 /* enable Admin queue Interrupt causes */ 3460 val = ((reg_idx & PFINT_FW_CTL_MSIX_INDX_M) | 3461 PFINT_FW_CTL_CAUSE_ENA_M); 3462 wr32(hw, PFINT_FW_CTL, val); 3463 3464 /* enable Mailbox queue Interrupt causes */ 3465 val = ((reg_idx & PFINT_MBX_CTL_MSIX_INDX_M) | 3466 PFINT_MBX_CTL_CAUSE_ENA_M); 3467 wr32(hw, PFINT_MBX_CTL, val); 3468 3469 if (!hw->dev_caps.ts_dev_info.ts_ll_int_read) { 3470 /* enable Sideband queue Interrupt causes */ 3471 val = ((reg_idx & PFINT_SB_CTL_MSIX_INDX_M) | 3472 PFINT_SB_CTL_CAUSE_ENA_M); 3473 wr32(hw, PFINT_SB_CTL, val); 3474 } 3475 3476 ice_flush(hw); 3477 } 3478 3479 /** 3480 * ice_req_irq_msix_misc - Setup the misc vector to handle non queue events 3481 * @pf: board private structure 3482 * 3483 * This sets up the handler for MSIX 0, which is used to manage the 3484 * non-queue interrupts, e.g. AdminQ and errors. This is not used 3485 * when in MSI or Legacy interrupt mode. 3486 */ 3487 static int ice_req_irq_msix_misc(struct ice_pf *pf) 3488 { 3489 struct device *dev = ice_pf_to_dev(pf); 3490 struct ice_hw *hw = &pf->hw; 3491 u32 pf_intr_start_offset; 3492 struct msi_map irq; 3493 int err = 0; 3494 3495 if (!pf->int_name[0]) 3496 snprintf(pf->int_name, sizeof(pf->int_name) - 1, "%s-%s:misc", 3497 dev_driver_string(dev), dev_name(dev)); 3498 3499 if (!pf->int_name_ll_ts[0]) 3500 snprintf(pf->int_name_ll_ts, sizeof(pf->int_name_ll_ts) - 1, 3501 "%s-%s:ll_ts", dev_driver_string(dev), dev_name(dev)); 3502 /* Do not request IRQ but do enable OICR interrupt since settings are 3503 * lost during reset. Note that this function is called only during 3504 * rebuild path and not while reset is in progress. 3505 */ 3506 if (ice_is_reset_in_progress(pf->state)) 3507 goto skip_req_irq; 3508 3509 /* reserve one vector in irq_tracker for misc interrupts */ 3510 irq = ice_alloc_irq(pf, false); 3511 if (irq.index < 0) 3512 return irq.index; 3513 3514 pf->oicr_irq = irq; 3515 err = devm_request_threaded_irq(dev, pf->oicr_irq.virq, ice_misc_intr, 3516 ice_misc_intr_thread_fn, 0, 3517 pf->int_name, pf); 3518 if (err) { 3519 dev_err(dev, "devm_request_threaded_irq for %s failed: %d\n", 3520 pf->int_name, err); 3521 ice_free_irq(pf, pf->oicr_irq); 3522 return err; 3523 } 3524 3525 /* reserve one vector in irq_tracker for ll_ts interrupt */ 3526 if (!pf->hw.dev_caps.ts_dev_info.ts_ll_int_read) 3527 goto skip_req_irq; 3528 3529 irq = ice_alloc_irq(pf, false); 3530 if (irq.index < 0) 3531 return irq.index; 3532 3533 pf->ll_ts_irq = irq; 3534 err = devm_request_irq(dev, pf->ll_ts_irq.virq, ice_ll_ts_intr, 0, 3535 pf->int_name_ll_ts, pf); 3536 if (err) { 3537 dev_err(dev, "devm_request_irq for %s failed: %d\n", 3538 pf->int_name_ll_ts, err); 3539 ice_free_irq(pf, pf->ll_ts_irq); 3540 return err; 3541 } 3542 3543 skip_req_irq: 3544 ice_ena_misc_vector(pf); 3545 3546 ice_ena_ctrlq_interrupts(hw, pf->oicr_irq.index); 3547 /* This enables LL TS interrupt */ 3548 pf_intr_start_offset = rd32(hw, PFINT_ALLOC) & PFINT_ALLOC_FIRST; 3549 if (pf->hw.dev_caps.ts_dev_info.ts_ll_int_read) 3550 wr32(hw, PFINT_SB_CTL, 3551 ((pf->ll_ts_irq.index + pf_intr_start_offset) & 3552 PFINT_SB_CTL_MSIX_INDX_M) | PFINT_SB_CTL_CAUSE_ENA_M); 3553 wr32(hw, GLINT_ITR(ICE_RX_ITR, pf->oicr_irq.index), 3554 ITR_REG_ALIGN(ICE_ITR_8K) >> ICE_ITR_GRAN_S); 3555 3556 ice_flush(hw); 3557 ice_irq_dynamic_ena(hw, NULL, NULL); 3558 3559 return 0; 3560 } 3561 3562 /** 3563 * ice_set_ops - set netdev and ethtools ops for the given netdev 3564 * @vsi: the VSI associated with the new netdev 3565 */ 3566 static void ice_set_ops(struct ice_vsi *vsi) 3567 { 3568 struct net_device *netdev = vsi->netdev; 3569 struct ice_pf *pf = ice_netdev_to_pf(netdev); 3570 3571 if (ice_is_safe_mode(pf)) { 3572 netdev->netdev_ops = &ice_netdev_safe_mode_ops; 3573 ice_set_ethtool_safe_mode_ops(netdev); 3574 return; 3575 } 3576 3577 netdev->netdev_ops = &ice_netdev_ops; 3578 netdev->udp_tunnel_nic_info = &pf->hw.udp_tunnel_nic; 3579 netdev->xdp_metadata_ops = &ice_xdp_md_ops; 3580 ice_set_ethtool_ops(netdev); 3581 3582 if (vsi->type != ICE_VSI_PF) 3583 return; 3584 3585 netdev->xdp_features = NETDEV_XDP_ACT_BASIC | NETDEV_XDP_ACT_REDIRECT | 3586 NETDEV_XDP_ACT_XSK_ZEROCOPY | 3587 NETDEV_XDP_ACT_RX_SG; 3588 netdev->xdp_zc_max_segs = ICE_MAX_BUF_TXD; 3589 } 3590 3591 /** 3592 * ice_set_netdev_features - set features for the given netdev 3593 * @netdev: netdev instance 3594 */ 3595 void ice_set_netdev_features(struct net_device *netdev) 3596 { 3597 struct ice_pf *pf = ice_netdev_to_pf(netdev); 3598 bool is_dvm_ena = ice_is_dvm_ena(&pf->hw); 3599 netdev_features_t csumo_features; 3600 netdev_features_t vlano_features; 3601 netdev_features_t dflt_features; 3602 netdev_features_t tso_features; 3603 3604 if (ice_is_safe_mode(pf)) { 3605 /* safe mode */ 3606 netdev->features = NETIF_F_SG | NETIF_F_HIGHDMA; 3607 netdev->hw_features = netdev->features; 3608 return; 3609 } 3610 3611 dflt_features = NETIF_F_SG | 3612 NETIF_F_HIGHDMA | 3613 NETIF_F_NTUPLE | 3614 NETIF_F_RXHASH; 3615 3616 csumo_features = NETIF_F_RXCSUM | 3617 NETIF_F_IP_CSUM | 3618 NETIF_F_SCTP_CRC | 3619 NETIF_F_IPV6_CSUM; 3620 3621 vlano_features = NETIF_F_HW_VLAN_CTAG_FILTER | 3622 NETIF_F_HW_VLAN_CTAG_TX | 3623 NETIF_F_HW_VLAN_CTAG_RX; 3624 3625 /* Enable CTAG/STAG filtering by default in Double VLAN Mode (DVM) */ 3626 if (is_dvm_ena) 3627 vlano_features |= NETIF_F_HW_VLAN_STAG_FILTER; 3628 3629 tso_features = NETIF_F_TSO | 3630 NETIF_F_TSO_ECN | 3631 NETIF_F_TSO6 | 3632 NETIF_F_GSO_GRE | 3633 NETIF_F_GSO_UDP_TUNNEL | 3634 NETIF_F_GSO_GRE_CSUM | 3635 NETIF_F_GSO_UDP_TUNNEL_CSUM | 3636 NETIF_F_GSO_PARTIAL | 3637 NETIF_F_GSO_IPXIP4 | 3638 NETIF_F_GSO_IPXIP6 | 3639 NETIF_F_GSO_UDP_L4; 3640 3641 netdev->gso_partial_features |= NETIF_F_GSO_UDP_TUNNEL_CSUM | 3642 NETIF_F_GSO_GRE_CSUM; 3643 /* set features that user can change */ 3644 netdev->hw_features = dflt_features | csumo_features | 3645 vlano_features | tso_features; 3646 3647 /* add support for HW_CSUM on packets with MPLS header */ 3648 netdev->mpls_features = NETIF_F_HW_CSUM | 3649 NETIF_F_TSO | 3650 NETIF_F_TSO6; 3651 3652 /* enable features */ 3653 netdev->features |= netdev->hw_features; 3654 3655 netdev->hw_features |= NETIF_F_HW_TC; 3656 netdev->hw_features |= NETIF_F_LOOPBACK; 3657 3658 /* encap and VLAN devices inherit default, csumo and tso features */ 3659 netdev->hw_enc_features |= dflt_features | csumo_features | 3660 tso_features; 3661 netdev->vlan_features |= dflt_features | csumo_features | 3662 tso_features; 3663 3664 /* advertise support but don't enable by default since only one type of 3665 * VLAN offload can be enabled at a time (i.e. CTAG or STAG). When one 3666 * type turns on the other has to be turned off. This is enforced by the 3667 * ice_fix_features() ndo callback. 3668 */ 3669 if (is_dvm_ena) 3670 netdev->hw_features |= NETIF_F_HW_VLAN_STAG_RX | 3671 NETIF_F_HW_VLAN_STAG_TX; 3672 3673 /* Leave CRC / FCS stripping enabled by default, but allow the value to 3674 * be changed at runtime 3675 */ 3676 netdev->hw_features |= NETIF_F_RXFCS; 3677 3678 netif_set_tso_max_size(netdev, ICE_MAX_TSO_SIZE); 3679 } 3680 3681 /** 3682 * ice_fill_rss_lut - Fill the RSS lookup table with default values 3683 * @lut: Lookup table 3684 * @rss_table_size: Lookup table size 3685 * @rss_size: Range of queue number for hashing 3686 */ 3687 void ice_fill_rss_lut(u8 *lut, u16 rss_table_size, u16 rss_size) 3688 { 3689 u16 i; 3690 3691 for (i = 0; i < rss_table_size; i++) 3692 lut[i] = i % rss_size; 3693 } 3694 3695 /** 3696 * ice_pf_vsi_setup - Set up a PF VSI 3697 * @pf: board private structure 3698 * @pi: pointer to the port_info instance 3699 * 3700 * Returns pointer to the successfully allocated VSI software struct 3701 * on success, otherwise returns NULL on failure. 3702 */ 3703 static struct ice_vsi * 3704 ice_pf_vsi_setup(struct ice_pf *pf, struct ice_port_info *pi) 3705 { 3706 struct ice_vsi_cfg_params params = {}; 3707 3708 params.type = ICE_VSI_PF; 3709 params.port_info = pi; 3710 params.flags = ICE_VSI_FLAG_INIT; 3711 3712 return ice_vsi_setup(pf, ¶ms); 3713 } 3714 3715 static struct ice_vsi * 3716 ice_chnl_vsi_setup(struct ice_pf *pf, struct ice_port_info *pi, 3717 struct ice_channel *ch) 3718 { 3719 struct ice_vsi_cfg_params params = {}; 3720 3721 params.type = ICE_VSI_CHNL; 3722 params.port_info = pi; 3723 params.ch = ch; 3724 params.flags = ICE_VSI_FLAG_INIT; 3725 3726 return ice_vsi_setup(pf, ¶ms); 3727 } 3728 3729 /** 3730 * ice_ctrl_vsi_setup - Set up a control VSI 3731 * @pf: board private structure 3732 * @pi: pointer to the port_info instance 3733 * 3734 * Returns pointer to the successfully allocated VSI software struct 3735 * on success, otherwise returns NULL on failure. 3736 */ 3737 static struct ice_vsi * 3738 ice_ctrl_vsi_setup(struct ice_pf *pf, struct ice_port_info *pi) 3739 { 3740 struct ice_vsi_cfg_params params = {}; 3741 3742 params.type = ICE_VSI_CTRL; 3743 params.port_info = pi; 3744 params.flags = ICE_VSI_FLAG_INIT; 3745 3746 return ice_vsi_setup(pf, ¶ms); 3747 } 3748 3749 /** 3750 * ice_lb_vsi_setup - Set up a loopback VSI 3751 * @pf: board private structure 3752 * @pi: pointer to the port_info instance 3753 * 3754 * Returns pointer to the successfully allocated VSI software struct 3755 * on success, otherwise returns NULL on failure. 3756 */ 3757 struct ice_vsi * 3758 ice_lb_vsi_setup(struct ice_pf *pf, struct ice_port_info *pi) 3759 { 3760 struct ice_vsi_cfg_params params = {}; 3761 3762 params.type = ICE_VSI_LB; 3763 params.port_info = pi; 3764 params.flags = ICE_VSI_FLAG_INIT; 3765 3766 return ice_vsi_setup(pf, ¶ms); 3767 } 3768 3769 /** 3770 * ice_vlan_rx_add_vid - Add a VLAN ID filter to HW offload 3771 * @netdev: network interface to be adjusted 3772 * @proto: VLAN TPID 3773 * @vid: VLAN ID to be added 3774 * 3775 * net_device_ops implementation for adding VLAN IDs 3776 */ 3777 int ice_vlan_rx_add_vid(struct net_device *netdev, __be16 proto, u16 vid) 3778 { 3779 struct ice_netdev_priv *np = netdev_priv(netdev); 3780 struct ice_vsi_vlan_ops *vlan_ops; 3781 struct ice_vsi *vsi = np->vsi; 3782 struct ice_vlan vlan; 3783 int ret; 3784 3785 /* VLAN 0 is added by default during load/reset */ 3786 if (!vid) 3787 return 0; 3788 3789 while (test_and_set_bit(ICE_CFG_BUSY, vsi->state)) 3790 usleep_range(1000, 2000); 3791 3792 /* Add multicast promisc rule for the VLAN ID to be added if 3793 * all-multicast is currently enabled. 3794 */ 3795 if (vsi->current_netdev_flags & IFF_ALLMULTI) { 3796 ret = ice_fltr_set_vsi_promisc(&vsi->back->hw, vsi->idx, 3797 ICE_MCAST_VLAN_PROMISC_BITS, 3798 vid); 3799 if (ret) 3800 goto finish; 3801 } 3802 3803 vlan_ops = ice_get_compat_vsi_vlan_ops(vsi); 3804 3805 /* Add a switch rule for this VLAN ID so its corresponding VLAN tagged 3806 * packets aren't pruned by the device's internal switch on Rx 3807 */ 3808 vlan = ICE_VLAN(be16_to_cpu(proto), vid, 0); 3809 ret = vlan_ops->add_vlan(vsi, &vlan); 3810 if (ret) 3811 goto finish; 3812 3813 /* If all-multicast is currently enabled and this VLAN ID is only one 3814 * besides VLAN-0 we have to update look-up type of multicast promisc 3815 * rule for VLAN-0 from ICE_SW_LKUP_PROMISC to ICE_SW_LKUP_PROMISC_VLAN. 3816 */ 3817 if ((vsi->current_netdev_flags & IFF_ALLMULTI) && 3818 ice_vsi_num_non_zero_vlans(vsi) == 1) { 3819 ice_fltr_clear_vsi_promisc(&vsi->back->hw, vsi->idx, 3820 ICE_MCAST_PROMISC_BITS, 0); 3821 ice_fltr_set_vsi_promisc(&vsi->back->hw, vsi->idx, 3822 ICE_MCAST_VLAN_PROMISC_BITS, 0); 3823 } 3824 3825 finish: 3826 clear_bit(ICE_CFG_BUSY, vsi->state); 3827 3828 return ret; 3829 } 3830 3831 /** 3832 * ice_vlan_rx_kill_vid - Remove a VLAN ID filter from HW offload 3833 * @netdev: network interface to be adjusted 3834 * @proto: VLAN TPID 3835 * @vid: VLAN ID to be removed 3836 * 3837 * net_device_ops implementation for removing VLAN IDs 3838 */ 3839 int ice_vlan_rx_kill_vid(struct net_device *netdev, __be16 proto, u16 vid) 3840 { 3841 struct ice_netdev_priv *np = netdev_priv(netdev); 3842 struct ice_vsi_vlan_ops *vlan_ops; 3843 struct ice_vsi *vsi = np->vsi; 3844 struct ice_vlan vlan; 3845 int ret; 3846 3847 /* don't allow removal of VLAN 0 */ 3848 if (!vid) 3849 return 0; 3850 3851 while (test_and_set_bit(ICE_CFG_BUSY, vsi->state)) 3852 usleep_range(1000, 2000); 3853 3854 ret = ice_clear_vsi_promisc(&vsi->back->hw, vsi->idx, 3855 ICE_MCAST_VLAN_PROMISC_BITS, vid); 3856 if (ret) { 3857 netdev_err(netdev, "Error clearing multicast promiscuous mode on VSI %i\n", 3858 vsi->vsi_num); 3859 vsi->current_netdev_flags |= IFF_ALLMULTI; 3860 } 3861 3862 vlan_ops = ice_get_compat_vsi_vlan_ops(vsi); 3863 3864 /* Make sure VLAN delete is successful before updating VLAN 3865 * information 3866 */ 3867 vlan = ICE_VLAN(be16_to_cpu(proto), vid, 0); 3868 ret = vlan_ops->del_vlan(vsi, &vlan); 3869 if (ret) 3870 goto finish; 3871 3872 /* Remove multicast promisc rule for the removed VLAN ID if 3873 * all-multicast is enabled. 3874 */ 3875 if (vsi->current_netdev_flags & IFF_ALLMULTI) 3876 ice_fltr_clear_vsi_promisc(&vsi->back->hw, vsi->idx, 3877 ICE_MCAST_VLAN_PROMISC_BITS, vid); 3878 3879 if (!ice_vsi_has_non_zero_vlans(vsi)) { 3880 /* Update look-up type of multicast promisc rule for VLAN 0 3881 * from ICE_SW_LKUP_PROMISC_VLAN to ICE_SW_LKUP_PROMISC when 3882 * all-multicast is enabled and VLAN 0 is the only VLAN rule. 3883 */ 3884 if (vsi->current_netdev_flags & IFF_ALLMULTI) { 3885 ice_fltr_clear_vsi_promisc(&vsi->back->hw, vsi->idx, 3886 ICE_MCAST_VLAN_PROMISC_BITS, 3887 0); 3888 ice_fltr_set_vsi_promisc(&vsi->back->hw, vsi->idx, 3889 ICE_MCAST_PROMISC_BITS, 0); 3890 } 3891 } 3892 3893 finish: 3894 clear_bit(ICE_CFG_BUSY, vsi->state); 3895 3896 return ret; 3897 } 3898 3899 /** 3900 * ice_rep_indr_tc_block_unbind 3901 * @cb_priv: indirection block private data 3902 */ 3903 static void ice_rep_indr_tc_block_unbind(void *cb_priv) 3904 { 3905 struct ice_indr_block_priv *indr_priv = cb_priv; 3906 3907 list_del(&indr_priv->list); 3908 kfree(indr_priv); 3909 } 3910 3911 /** 3912 * ice_tc_indir_block_unregister - Unregister TC indirect block notifications 3913 * @vsi: VSI struct which has the netdev 3914 */ 3915 static void ice_tc_indir_block_unregister(struct ice_vsi *vsi) 3916 { 3917 struct ice_netdev_priv *np = netdev_priv(vsi->netdev); 3918 3919 flow_indr_dev_unregister(ice_indr_setup_tc_cb, np, 3920 ice_rep_indr_tc_block_unbind); 3921 } 3922 3923 /** 3924 * ice_tc_indir_block_register - Register TC indirect block notifications 3925 * @vsi: VSI struct which has the netdev 3926 * 3927 * Returns 0 on success, negative value on failure 3928 */ 3929 static int ice_tc_indir_block_register(struct ice_vsi *vsi) 3930 { 3931 struct ice_netdev_priv *np; 3932 3933 if (!vsi || !vsi->netdev) 3934 return -EINVAL; 3935 3936 np = netdev_priv(vsi->netdev); 3937 3938 INIT_LIST_HEAD(&np->tc_indr_block_priv_list); 3939 return flow_indr_dev_register(ice_indr_setup_tc_cb, np); 3940 } 3941 3942 /** 3943 * ice_get_avail_q_count - Get count of queues in use 3944 * @pf_qmap: bitmap to get queue use count from 3945 * @lock: pointer to a mutex that protects access to pf_qmap 3946 * @size: size of the bitmap 3947 */ 3948 static u16 3949 ice_get_avail_q_count(unsigned long *pf_qmap, struct mutex *lock, u16 size) 3950 { 3951 unsigned long bit; 3952 u16 count = 0; 3953 3954 mutex_lock(lock); 3955 for_each_clear_bit(bit, pf_qmap, size) 3956 count++; 3957 mutex_unlock(lock); 3958 3959 return count; 3960 } 3961 3962 /** 3963 * ice_get_avail_txq_count - Get count of Tx queues in use 3964 * @pf: pointer to an ice_pf instance 3965 */ 3966 u16 ice_get_avail_txq_count(struct ice_pf *pf) 3967 { 3968 return ice_get_avail_q_count(pf->avail_txqs, &pf->avail_q_mutex, 3969 pf->max_pf_txqs); 3970 } 3971 3972 /** 3973 * ice_get_avail_rxq_count - Get count of Rx queues in use 3974 * @pf: pointer to an ice_pf instance 3975 */ 3976 u16 ice_get_avail_rxq_count(struct ice_pf *pf) 3977 { 3978 return ice_get_avail_q_count(pf->avail_rxqs, &pf->avail_q_mutex, 3979 pf->max_pf_rxqs); 3980 } 3981 3982 /** 3983 * ice_deinit_pf - Unrolls initialziations done by ice_init_pf 3984 * @pf: board private structure to initialize 3985 */ 3986 static void ice_deinit_pf(struct ice_pf *pf) 3987 { 3988 ice_service_task_stop(pf); 3989 mutex_destroy(&pf->lag_mutex); 3990 mutex_destroy(&pf->adev_mutex); 3991 mutex_destroy(&pf->sw_mutex); 3992 mutex_destroy(&pf->tc_mutex); 3993 mutex_destroy(&pf->avail_q_mutex); 3994 mutex_destroy(&pf->vfs.table_lock); 3995 3996 if (pf->avail_txqs) { 3997 bitmap_free(pf->avail_txqs); 3998 pf->avail_txqs = NULL; 3999 } 4000 4001 if (pf->avail_rxqs) { 4002 bitmap_free(pf->avail_rxqs); 4003 pf->avail_rxqs = NULL; 4004 } 4005 4006 if (pf->ptp.clock) 4007 ptp_clock_unregister(pf->ptp.clock); 4008 4009 xa_destroy(&pf->dyn_ports); 4010 xa_destroy(&pf->sf_nums); 4011 } 4012 4013 /** 4014 * ice_set_pf_caps - set PFs capability flags 4015 * @pf: pointer to the PF instance 4016 */ 4017 static void ice_set_pf_caps(struct ice_pf *pf) 4018 { 4019 struct ice_hw_func_caps *func_caps = &pf->hw.func_caps; 4020 4021 clear_bit(ICE_FLAG_RDMA_ENA, pf->flags); 4022 if (func_caps->common_cap.rdma) 4023 set_bit(ICE_FLAG_RDMA_ENA, pf->flags); 4024 clear_bit(ICE_FLAG_DCB_CAPABLE, pf->flags); 4025 if (func_caps->common_cap.dcb) 4026 set_bit(ICE_FLAG_DCB_CAPABLE, pf->flags); 4027 clear_bit(ICE_FLAG_SRIOV_CAPABLE, pf->flags); 4028 if (func_caps->common_cap.sr_iov_1_1) { 4029 set_bit(ICE_FLAG_SRIOV_CAPABLE, pf->flags); 4030 pf->vfs.num_supported = min_t(int, func_caps->num_allocd_vfs, 4031 ICE_MAX_SRIOV_VFS); 4032 } 4033 clear_bit(ICE_FLAG_RSS_ENA, pf->flags); 4034 if (func_caps->common_cap.rss_table_size) 4035 set_bit(ICE_FLAG_RSS_ENA, pf->flags); 4036 4037 clear_bit(ICE_FLAG_FD_ENA, pf->flags); 4038 if (func_caps->fd_fltr_guar > 0 || func_caps->fd_fltr_best_effort > 0) { 4039 u16 unused; 4040 4041 /* ctrl_vsi_idx will be set to a valid value when flow director 4042 * is setup by ice_init_fdir 4043 */ 4044 pf->ctrl_vsi_idx = ICE_NO_VSI; 4045 set_bit(ICE_FLAG_FD_ENA, pf->flags); 4046 /* force guaranteed filter pool for PF */ 4047 ice_alloc_fd_guar_item(&pf->hw, &unused, 4048 func_caps->fd_fltr_guar); 4049 /* force shared filter pool for PF */ 4050 ice_alloc_fd_shrd_item(&pf->hw, &unused, 4051 func_caps->fd_fltr_best_effort); 4052 } 4053 4054 clear_bit(ICE_FLAG_PTP_SUPPORTED, pf->flags); 4055 if (func_caps->common_cap.ieee_1588 && 4056 !(pf->hw.mac_type == ICE_MAC_E830)) 4057 set_bit(ICE_FLAG_PTP_SUPPORTED, pf->flags); 4058 4059 pf->max_pf_txqs = func_caps->common_cap.num_txq; 4060 pf->max_pf_rxqs = func_caps->common_cap.num_rxq; 4061 } 4062 4063 /** 4064 * ice_init_pf - Initialize general software structures (struct ice_pf) 4065 * @pf: board private structure to initialize 4066 */ 4067 static int ice_init_pf(struct ice_pf *pf) 4068 { 4069 ice_set_pf_caps(pf); 4070 4071 mutex_init(&pf->sw_mutex); 4072 mutex_init(&pf->tc_mutex); 4073 mutex_init(&pf->adev_mutex); 4074 mutex_init(&pf->lag_mutex); 4075 4076 INIT_HLIST_HEAD(&pf->aq_wait_list); 4077 spin_lock_init(&pf->aq_wait_lock); 4078 init_waitqueue_head(&pf->aq_wait_queue); 4079 4080 init_waitqueue_head(&pf->reset_wait_queue); 4081 4082 /* setup service timer and periodic service task */ 4083 timer_setup(&pf->serv_tmr, ice_service_timer, 0); 4084 pf->serv_tmr_period = HZ; 4085 INIT_WORK(&pf->serv_task, ice_service_task); 4086 clear_bit(ICE_SERVICE_SCHED, pf->state); 4087 4088 mutex_init(&pf->avail_q_mutex); 4089 pf->avail_txqs = bitmap_zalloc(pf->max_pf_txqs, GFP_KERNEL); 4090 if (!pf->avail_txqs) 4091 return -ENOMEM; 4092 4093 pf->avail_rxqs = bitmap_zalloc(pf->max_pf_rxqs, GFP_KERNEL); 4094 if (!pf->avail_rxqs) { 4095 bitmap_free(pf->avail_txqs); 4096 pf->avail_txqs = NULL; 4097 return -ENOMEM; 4098 } 4099 4100 mutex_init(&pf->vfs.table_lock); 4101 hash_init(pf->vfs.table); 4102 ice_mbx_init_snapshot(&pf->hw); 4103 4104 xa_init(&pf->dyn_ports); 4105 xa_init(&pf->sf_nums); 4106 4107 return 0; 4108 } 4109 4110 /** 4111 * ice_is_wol_supported - check if WoL is supported 4112 * @hw: pointer to hardware info 4113 * 4114 * Check if WoL is supported based on the HW configuration. 4115 * Returns true if NVM supports and enables WoL for this port, false otherwise 4116 */ 4117 bool ice_is_wol_supported(struct ice_hw *hw) 4118 { 4119 u16 wol_ctrl; 4120 4121 /* A bit set to 1 in the NVM Software Reserved Word 2 (WoL control 4122 * word) indicates WoL is not supported on the corresponding PF ID. 4123 */ 4124 if (ice_read_sr_word(hw, ICE_SR_NVM_WOL_CFG, &wol_ctrl)) 4125 return false; 4126 4127 return !(BIT(hw->port_info->lport) & wol_ctrl); 4128 } 4129 4130 /** 4131 * ice_vsi_recfg_qs - Change the number of queues on a VSI 4132 * @vsi: VSI being changed 4133 * @new_rx: new number of Rx queues 4134 * @new_tx: new number of Tx queues 4135 * @locked: is adev device_lock held 4136 * 4137 * Only change the number of queues if new_tx, or new_rx is non-0. 4138 * 4139 * Returns 0 on success. 4140 */ 4141 int ice_vsi_recfg_qs(struct ice_vsi *vsi, int new_rx, int new_tx, bool locked) 4142 { 4143 struct ice_pf *pf = vsi->back; 4144 int i, err = 0, timeout = 50; 4145 4146 if (!new_rx && !new_tx) 4147 return -EINVAL; 4148 4149 while (test_and_set_bit(ICE_CFG_BUSY, pf->state)) { 4150 timeout--; 4151 if (!timeout) 4152 return -EBUSY; 4153 usleep_range(1000, 2000); 4154 } 4155 4156 if (new_tx) 4157 vsi->req_txq = (u16)new_tx; 4158 if (new_rx) 4159 vsi->req_rxq = (u16)new_rx; 4160 4161 /* set for the next time the netdev is started */ 4162 if (!netif_running(vsi->netdev)) { 4163 err = ice_vsi_rebuild(vsi, ICE_VSI_FLAG_NO_INIT); 4164 if (err) 4165 goto rebuild_err; 4166 dev_dbg(ice_pf_to_dev(pf), "Link is down, queue count change happens when link is brought up\n"); 4167 goto done; 4168 } 4169 4170 ice_vsi_close(vsi); 4171 err = ice_vsi_rebuild(vsi, ICE_VSI_FLAG_NO_INIT); 4172 if (err) 4173 goto rebuild_err; 4174 4175 ice_for_each_traffic_class(i) { 4176 if (vsi->tc_cfg.ena_tc & BIT(i)) 4177 netdev_set_tc_queue(vsi->netdev, 4178 vsi->tc_cfg.tc_info[i].netdev_tc, 4179 vsi->tc_cfg.tc_info[i].qcount_tx, 4180 vsi->tc_cfg.tc_info[i].qoffset); 4181 } 4182 ice_pf_dcb_recfg(pf, locked); 4183 ice_vsi_open(vsi); 4184 goto done; 4185 4186 rebuild_err: 4187 dev_err(ice_pf_to_dev(pf), "Error during VSI rebuild: %d. Unload and reload the driver.\n", 4188 err); 4189 done: 4190 clear_bit(ICE_CFG_BUSY, pf->state); 4191 return err; 4192 } 4193 4194 /** 4195 * ice_set_safe_mode_vlan_cfg - configure PF VSI to allow all VLANs in safe mode 4196 * @pf: PF to configure 4197 * 4198 * No VLAN offloads/filtering are advertised in safe mode so make sure the PF 4199 * VSI can still Tx/Rx VLAN tagged packets. 4200 */ 4201 static void ice_set_safe_mode_vlan_cfg(struct ice_pf *pf) 4202 { 4203 struct ice_vsi *vsi = ice_get_main_vsi(pf); 4204 struct ice_vsi_ctx *ctxt; 4205 struct ice_hw *hw; 4206 int status; 4207 4208 if (!vsi) 4209 return; 4210 4211 ctxt = kzalloc(sizeof(*ctxt), GFP_KERNEL); 4212 if (!ctxt) 4213 return; 4214 4215 hw = &pf->hw; 4216 ctxt->info = vsi->info; 4217 4218 ctxt->info.valid_sections = 4219 cpu_to_le16(ICE_AQ_VSI_PROP_VLAN_VALID | 4220 ICE_AQ_VSI_PROP_SECURITY_VALID | 4221 ICE_AQ_VSI_PROP_SW_VALID); 4222 4223 /* disable VLAN anti-spoof */ 4224 ctxt->info.sec_flags &= ~(ICE_AQ_VSI_SEC_TX_VLAN_PRUNE_ENA << 4225 ICE_AQ_VSI_SEC_TX_PRUNE_ENA_S); 4226 4227 /* disable VLAN pruning and keep all other settings */ 4228 ctxt->info.sw_flags2 &= ~ICE_AQ_VSI_SW_FLAG_RX_VLAN_PRUNE_ENA; 4229 4230 /* allow all VLANs on Tx and don't strip on Rx */ 4231 ctxt->info.inner_vlan_flags = ICE_AQ_VSI_INNER_VLAN_TX_MODE_ALL | 4232 ICE_AQ_VSI_INNER_VLAN_EMODE_NOTHING; 4233 4234 status = ice_update_vsi(hw, vsi->idx, ctxt, NULL); 4235 if (status) { 4236 dev_err(ice_pf_to_dev(vsi->back), "Failed to update VSI for safe mode VLANs, err %d aq_err %s\n", 4237 status, ice_aq_str(hw->adminq.sq_last_status)); 4238 } else { 4239 vsi->info.sec_flags = ctxt->info.sec_flags; 4240 vsi->info.sw_flags2 = ctxt->info.sw_flags2; 4241 vsi->info.inner_vlan_flags = ctxt->info.inner_vlan_flags; 4242 } 4243 4244 kfree(ctxt); 4245 } 4246 4247 /** 4248 * ice_log_pkg_init - log result of DDP package load 4249 * @hw: pointer to hardware info 4250 * @state: state of package load 4251 */ 4252 static void ice_log_pkg_init(struct ice_hw *hw, enum ice_ddp_state state) 4253 { 4254 struct ice_pf *pf = hw->back; 4255 struct device *dev; 4256 4257 dev = ice_pf_to_dev(pf); 4258 4259 switch (state) { 4260 case ICE_DDP_PKG_SUCCESS: 4261 dev_info(dev, "The DDP package was successfully loaded: %s version %d.%d.%d.%d\n", 4262 hw->active_pkg_name, 4263 hw->active_pkg_ver.major, 4264 hw->active_pkg_ver.minor, 4265 hw->active_pkg_ver.update, 4266 hw->active_pkg_ver.draft); 4267 break; 4268 case ICE_DDP_PKG_SAME_VERSION_ALREADY_LOADED: 4269 dev_info(dev, "DDP package already present on device: %s version %d.%d.%d.%d\n", 4270 hw->active_pkg_name, 4271 hw->active_pkg_ver.major, 4272 hw->active_pkg_ver.minor, 4273 hw->active_pkg_ver.update, 4274 hw->active_pkg_ver.draft); 4275 break; 4276 case ICE_DDP_PKG_ALREADY_LOADED_NOT_SUPPORTED: 4277 dev_err(dev, "The device has a DDP package that is not supported by the driver. The device has package '%s' version %d.%d.x.x. The driver requires version %d.%d.x.x. Entering Safe Mode.\n", 4278 hw->active_pkg_name, 4279 hw->active_pkg_ver.major, 4280 hw->active_pkg_ver.minor, 4281 ICE_PKG_SUPP_VER_MAJ, ICE_PKG_SUPP_VER_MNR); 4282 break; 4283 case ICE_DDP_PKG_COMPATIBLE_ALREADY_LOADED: 4284 dev_info(dev, "The driver could not load the DDP package file because a compatible DDP package is already present on the device. The device has package '%s' version %d.%d.%d.%d. The package file found by the driver: '%s' version %d.%d.%d.%d.\n", 4285 hw->active_pkg_name, 4286 hw->active_pkg_ver.major, 4287 hw->active_pkg_ver.minor, 4288 hw->active_pkg_ver.update, 4289 hw->active_pkg_ver.draft, 4290 hw->pkg_name, 4291 hw->pkg_ver.major, 4292 hw->pkg_ver.minor, 4293 hw->pkg_ver.update, 4294 hw->pkg_ver.draft); 4295 break; 4296 case ICE_DDP_PKG_FW_MISMATCH: 4297 dev_err(dev, "The firmware loaded on the device is not compatible with the DDP package. Please update the device's NVM. Entering safe mode.\n"); 4298 break; 4299 case ICE_DDP_PKG_INVALID_FILE: 4300 dev_err(dev, "The DDP package file is invalid. Entering Safe Mode.\n"); 4301 break; 4302 case ICE_DDP_PKG_FILE_VERSION_TOO_HIGH: 4303 dev_err(dev, "The DDP package file version is higher than the driver supports. Please use an updated driver. Entering Safe Mode.\n"); 4304 break; 4305 case ICE_DDP_PKG_FILE_VERSION_TOO_LOW: 4306 dev_err(dev, "The DDP package file version is lower than the driver supports. The driver requires version %d.%d.x.x. Please use an updated DDP Package file. Entering Safe Mode.\n", 4307 ICE_PKG_SUPP_VER_MAJ, ICE_PKG_SUPP_VER_MNR); 4308 break; 4309 case ICE_DDP_PKG_FILE_SIGNATURE_INVALID: 4310 dev_err(dev, "The DDP package could not be loaded because its signature is not valid. Please use a valid DDP Package. Entering Safe Mode.\n"); 4311 break; 4312 case ICE_DDP_PKG_FILE_REVISION_TOO_LOW: 4313 dev_err(dev, "The DDP Package could not be loaded because its security revision is too low. Please use an updated DDP Package. Entering Safe Mode.\n"); 4314 break; 4315 case ICE_DDP_PKG_LOAD_ERROR: 4316 dev_err(dev, "An error occurred on the device while loading the DDP package. The device will be reset.\n"); 4317 /* poll for reset to complete */ 4318 if (ice_check_reset(hw)) 4319 dev_err(dev, "Error resetting device. Please reload the driver\n"); 4320 break; 4321 case ICE_DDP_PKG_ERR: 4322 default: 4323 dev_err(dev, "An unknown error occurred when loading the DDP package. Entering Safe Mode.\n"); 4324 break; 4325 } 4326 } 4327 4328 /** 4329 * ice_load_pkg - load/reload the DDP Package file 4330 * @firmware: firmware structure when firmware requested or NULL for reload 4331 * @pf: pointer to the PF instance 4332 * 4333 * Called on probe and post CORER/GLOBR rebuild to load DDP Package and 4334 * initialize HW tables. 4335 */ 4336 static void 4337 ice_load_pkg(const struct firmware *firmware, struct ice_pf *pf) 4338 { 4339 enum ice_ddp_state state = ICE_DDP_PKG_ERR; 4340 struct device *dev = ice_pf_to_dev(pf); 4341 struct ice_hw *hw = &pf->hw; 4342 4343 /* Load DDP Package */ 4344 if (firmware && !hw->pkg_copy) { 4345 state = ice_copy_and_init_pkg(hw, firmware->data, 4346 firmware->size); 4347 ice_log_pkg_init(hw, state); 4348 } else if (!firmware && hw->pkg_copy) { 4349 /* Reload package during rebuild after CORER/GLOBR reset */ 4350 state = ice_init_pkg(hw, hw->pkg_copy, hw->pkg_size); 4351 ice_log_pkg_init(hw, state); 4352 } else { 4353 dev_err(dev, "The DDP package file failed to load. Entering Safe Mode.\n"); 4354 } 4355 4356 if (!ice_is_init_pkg_successful(state)) { 4357 /* Safe Mode */ 4358 clear_bit(ICE_FLAG_ADV_FEATURES, pf->flags); 4359 return; 4360 } 4361 4362 /* Successful download package is the precondition for advanced 4363 * features, hence setting the ICE_FLAG_ADV_FEATURES flag 4364 */ 4365 set_bit(ICE_FLAG_ADV_FEATURES, pf->flags); 4366 } 4367 4368 /** 4369 * ice_verify_cacheline_size - verify driver's assumption of 64 Byte cache lines 4370 * @pf: pointer to the PF structure 4371 * 4372 * There is no error returned here because the driver should be able to handle 4373 * 128 Byte cache lines, so we only print a warning in case issues are seen, 4374 * specifically with Tx. 4375 */ 4376 static void ice_verify_cacheline_size(struct ice_pf *pf) 4377 { 4378 if (rd32(&pf->hw, GLPCI_CNF2) & GLPCI_CNF2_CACHELINE_SIZE_M) 4379 dev_warn(ice_pf_to_dev(pf), "%d Byte cache line assumption is invalid, driver may have Tx timeouts!\n", 4380 ICE_CACHE_LINE_BYTES); 4381 } 4382 4383 /** 4384 * ice_send_version - update firmware with driver version 4385 * @pf: PF struct 4386 * 4387 * Returns 0 on success, else error code 4388 */ 4389 static int ice_send_version(struct ice_pf *pf) 4390 { 4391 struct ice_driver_ver dv; 4392 4393 dv.major_ver = 0xff; 4394 dv.minor_ver = 0xff; 4395 dv.build_ver = 0xff; 4396 dv.subbuild_ver = 0; 4397 strscpy((char *)dv.driver_string, UTS_RELEASE, 4398 sizeof(dv.driver_string)); 4399 return ice_aq_send_driver_ver(&pf->hw, &dv, NULL); 4400 } 4401 4402 /** 4403 * ice_init_fdir - Initialize flow director VSI and configuration 4404 * @pf: pointer to the PF instance 4405 * 4406 * returns 0 on success, negative on error 4407 */ 4408 static int ice_init_fdir(struct ice_pf *pf) 4409 { 4410 struct device *dev = ice_pf_to_dev(pf); 4411 struct ice_vsi *ctrl_vsi; 4412 int err; 4413 4414 /* Side Band Flow Director needs to have a control VSI. 4415 * Allocate it and store it in the PF. 4416 */ 4417 ctrl_vsi = ice_ctrl_vsi_setup(pf, pf->hw.port_info); 4418 if (!ctrl_vsi) { 4419 dev_dbg(dev, "could not create control VSI\n"); 4420 return -ENOMEM; 4421 } 4422 4423 err = ice_vsi_open_ctrl(ctrl_vsi); 4424 if (err) { 4425 dev_dbg(dev, "could not open control VSI\n"); 4426 goto err_vsi_open; 4427 } 4428 4429 mutex_init(&pf->hw.fdir_fltr_lock); 4430 4431 err = ice_fdir_create_dflt_rules(pf); 4432 if (err) 4433 goto err_fdir_rule; 4434 4435 return 0; 4436 4437 err_fdir_rule: 4438 ice_fdir_release_flows(&pf->hw); 4439 ice_vsi_close(ctrl_vsi); 4440 err_vsi_open: 4441 ice_vsi_release(ctrl_vsi); 4442 if (pf->ctrl_vsi_idx != ICE_NO_VSI) { 4443 pf->vsi[pf->ctrl_vsi_idx] = NULL; 4444 pf->ctrl_vsi_idx = ICE_NO_VSI; 4445 } 4446 return err; 4447 } 4448 4449 static void ice_deinit_fdir(struct ice_pf *pf) 4450 { 4451 struct ice_vsi *vsi = ice_get_ctrl_vsi(pf); 4452 4453 if (!vsi) 4454 return; 4455 4456 ice_vsi_manage_fdir(vsi, false); 4457 ice_vsi_release(vsi); 4458 if (pf->ctrl_vsi_idx != ICE_NO_VSI) { 4459 pf->vsi[pf->ctrl_vsi_idx] = NULL; 4460 pf->ctrl_vsi_idx = ICE_NO_VSI; 4461 } 4462 4463 mutex_destroy(&(&pf->hw)->fdir_fltr_lock); 4464 } 4465 4466 /** 4467 * ice_get_opt_fw_name - return optional firmware file name or NULL 4468 * @pf: pointer to the PF instance 4469 */ 4470 static char *ice_get_opt_fw_name(struct ice_pf *pf) 4471 { 4472 /* Optional firmware name same as default with additional dash 4473 * followed by a EUI-64 identifier (PCIe Device Serial Number) 4474 */ 4475 struct pci_dev *pdev = pf->pdev; 4476 char *opt_fw_filename; 4477 u64 dsn; 4478 4479 /* Determine the name of the optional file using the DSN (two 4480 * dwords following the start of the DSN Capability). 4481 */ 4482 dsn = pci_get_dsn(pdev); 4483 if (!dsn) 4484 return NULL; 4485 4486 opt_fw_filename = kzalloc(NAME_MAX, GFP_KERNEL); 4487 if (!opt_fw_filename) 4488 return NULL; 4489 4490 snprintf(opt_fw_filename, NAME_MAX, "%sice-%016llx.pkg", 4491 ICE_DDP_PKG_PATH, dsn); 4492 4493 return opt_fw_filename; 4494 } 4495 4496 /** 4497 * ice_request_fw - Device initialization routine 4498 * @pf: pointer to the PF instance 4499 * @firmware: double pointer to firmware struct 4500 * 4501 * Return: zero when successful, negative values otherwise. 4502 */ 4503 static int ice_request_fw(struct ice_pf *pf, const struct firmware **firmware) 4504 { 4505 char *opt_fw_filename = ice_get_opt_fw_name(pf); 4506 struct device *dev = ice_pf_to_dev(pf); 4507 int err = 0; 4508 4509 /* optional device-specific DDP (if present) overrides the default DDP 4510 * package file. kernel logs a debug message if the file doesn't exist, 4511 * and warning messages for other errors. 4512 */ 4513 if (opt_fw_filename) { 4514 err = firmware_request_nowarn(firmware, opt_fw_filename, dev); 4515 kfree(opt_fw_filename); 4516 if (!err) 4517 return err; 4518 } 4519 err = request_firmware(firmware, ICE_DDP_PKG_FILE, dev); 4520 if (err) 4521 dev_err(dev, "The DDP package file was not found or could not be read. Entering Safe Mode\n"); 4522 4523 return err; 4524 } 4525 4526 /** 4527 * ice_init_tx_topology - performs Tx topology initialization 4528 * @hw: pointer to the hardware structure 4529 * @firmware: pointer to firmware structure 4530 * 4531 * Return: zero when init was successful, negative values otherwise. 4532 */ 4533 static int 4534 ice_init_tx_topology(struct ice_hw *hw, const struct firmware *firmware) 4535 { 4536 u8 num_tx_sched_layers = hw->num_tx_sched_layers; 4537 struct ice_pf *pf = hw->back; 4538 struct device *dev; 4539 u8 *buf_copy; 4540 int err; 4541 4542 dev = ice_pf_to_dev(pf); 4543 /* ice_cfg_tx_topo buf argument is not a constant, 4544 * so we have to make a copy 4545 */ 4546 buf_copy = kmemdup(firmware->data, firmware->size, GFP_KERNEL); 4547 4548 err = ice_cfg_tx_topo(hw, buf_copy, firmware->size); 4549 if (!err) { 4550 if (hw->num_tx_sched_layers > num_tx_sched_layers) 4551 dev_info(dev, "Tx scheduling layers switching feature disabled\n"); 4552 else 4553 dev_info(dev, "Tx scheduling layers switching feature enabled\n"); 4554 /* if there was a change in topology ice_cfg_tx_topo triggered 4555 * a CORER and we need to re-init hw 4556 */ 4557 ice_deinit_hw(hw); 4558 err = ice_init_hw(hw); 4559 4560 return err; 4561 } else if (err == -EIO) { 4562 dev_info(dev, "DDP package does not support Tx scheduling layers switching feature - please update to the latest DDP package and try again\n"); 4563 } 4564 4565 return 0; 4566 } 4567 4568 /** 4569 * ice_init_ddp_config - DDP related configuration 4570 * @hw: pointer to the hardware structure 4571 * @pf: pointer to pf structure 4572 * 4573 * This function loads DDP file from the disk, then initializes Tx 4574 * topology. At the end DDP package is loaded on the card. 4575 * 4576 * Return: zero when init was successful, negative values otherwise. 4577 */ 4578 static int ice_init_ddp_config(struct ice_hw *hw, struct ice_pf *pf) 4579 { 4580 struct device *dev = ice_pf_to_dev(pf); 4581 const struct firmware *firmware = NULL; 4582 int err; 4583 4584 err = ice_request_fw(pf, &firmware); 4585 if (err) { 4586 dev_err(dev, "Fail during requesting FW: %d\n", err); 4587 return err; 4588 } 4589 4590 err = ice_init_tx_topology(hw, firmware); 4591 if (err) { 4592 dev_err(dev, "Fail during initialization of Tx topology: %d\n", 4593 err); 4594 release_firmware(firmware); 4595 return err; 4596 } 4597 4598 /* Download firmware to device */ 4599 ice_load_pkg(firmware, pf); 4600 release_firmware(firmware); 4601 4602 return 0; 4603 } 4604 4605 /** 4606 * ice_print_wake_reason - show the wake up cause in the log 4607 * @pf: pointer to the PF struct 4608 */ 4609 static void ice_print_wake_reason(struct ice_pf *pf) 4610 { 4611 u32 wus = pf->wakeup_reason; 4612 const char *wake_str; 4613 4614 /* if no wake event, nothing to print */ 4615 if (!wus) 4616 return; 4617 4618 if (wus & PFPM_WUS_LNKC_M) 4619 wake_str = "Link\n"; 4620 else if (wus & PFPM_WUS_MAG_M) 4621 wake_str = "Magic Packet\n"; 4622 else if (wus & PFPM_WUS_MNG_M) 4623 wake_str = "Management\n"; 4624 else if (wus & PFPM_WUS_FW_RST_WK_M) 4625 wake_str = "Firmware Reset\n"; 4626 else 4627 wake_str = "Unknown\n"; 4628 4629 dev_info(ice_pf_to_dev(pf), "Wake reason: %s", wake_str); 4630 } 4631 4632 /** 4633 * ice_pf_fwlog_update_module - update 1 module 4634 * @pf: pointer to the PF struct 4635 * @log_level: log_level to use for the @module 4636 * @module: module to update 4637 */ 4638 void ice_pf_fwlog_update_module(struct ice_pf *pf, int log_level, int module) 4639 { 4640 struct ice_hw *hw = &pf->hw; 4641 4642 hw->fwlog_cfg.module_entries[module].log_level = log_level; 4643 } 4644 4645 /** 4646 * ice_register_netdev - register netdev 4647 * @vsi: pointer to the VSI struct 4648 */ 4649 static int ice_register_netdev(struct ice_vsi *vsi) 4650 { 4651 int err; 4652 4653 if (!vsi || !vsi->netdev) 4654 return -EIO; 4655 4656 err = register_netdev(vsi->netdev); 4657 if (err) 4658 return err; 4659 4660 set_bit(ICE_VSI_NETDEV_REGISTERED, vsi->state); 4661 netif_carrier_off(vsi->netdev); 4662 netif_tx_stop_all_queues(vsi->netdev); 4663 4664 return 0; 4665 } 4666 4667 static void ice_unregister_netdev(struct ice_vsi *vsi) 4668 { 4669 if (!vsi || !vsi->netdev) 4670 return; 4671 4672 unregister_netdev(vsi->netdev); 4673 clear_bit(ICE_VSI_NETDEV_REGISTERED, vsi->state); 4674 } 4675 4676 /** 4677 * ice_cfg_netdev - Allocate, configure and register a netdev 4678 * @vsi: the VSI associated with the new netdev 4679 * 4680 * Returns 0 on success, negative value on failure 4681 */ 4682 static int ice_cfg_netdev(struct ice_vsi *vsi) 4683 { 4684 struct ice_netdev_priv *np; 4685 struct net_device *netdev; 4686 u8 mac_addr[ETH_ALEN]; 4687 4688 netdev = alloc_etherdev_mqs(sizeof(*np), vsi->alloc_txq, 4689 vsi->alloc_rxq); 4690 if (!netdev) 4691 return -ENOMEM; 4692 4693 set_bit(ICE_VSI_NETDEV_ALLOCD, vsi->state); 4694 vsi->netdev = netdev; 4695 np = netdev_priv(netdev); 4696 np->vsi = vsi; 4697 4698 ice_set_netdev_features(netdev); 4699 ice_set_ops(vsi); 4700 4701 if (vsi->type == ICE_VSI_PF) { 4702 SET_NETDEV_DEV(netdev, ice_pf_to_dev(vsi->back)); 4703 ether_addr_copy(mac_addr, vsi->port_info->mac.perm_addr); 4704 eth_hw_addr_set(netdev, mac_addr); 4705 } 4706 4707 netdev->priv_flags |= IFF_UNICAST_FLT; 4708 4709 /* Setup netdev TC information */ 4710 ice_vsi_cfg_netdev_tc(vsi, vsi->tc_cfg.ena_tc); 4711 4712 netdev->max_mtu = ICE_MAX_MTU; 4713 4714 return 0; 4715 } 4716 4717 static void ice_decfg_netdev(struct ice_vsi *vsi) 4718 { 4719 clear_bit(ICE_VSI_NETDEV_ALLOCD, vsi->state); 4720 free_netdev(vsi->netdev); 4721 vsi->netdev = NULL; 4722 } 4723 4724 /** 4725 * ice_wait_for_fw - wait for full FW readiness 4726 * @hw: pointer to the hardware structure 4727 * @timeout: milliseconds that can elapse before timing out 4728 */ 4729 static int ice_wait_for_fw(struct ice_hw *hw, u32 timeout) 4730 { 4731 int fw_loading; 4732 u32 elapsed = 0; 4733 4734 while (elapsed <= timeout) { 4735 fw_loading = rd32(hw, GL_MNG_FWSM) & GL_MNG_FWSM_FW_LOADING_M; 4736 4737 /* firmware was not yet loaded, we have to wait more */ 4738 if (fw_loading) { 4739 elapsed += 100; 4740 msleep(100); 4741 continue; 4742 } 4743 return 0; 4744 } 4745 4746 return -ETIMEDOUT; 4747 } 4748 4749 int ice_init_dev(struct ice_pf *pf) 4750 { 4751 struct device *dev = ice_pf_to_dev(pf); 4752 struct ice_hw *hw = &pf->hw; 4753 int err; 4754 4755 err = ice_init_hw(hw); 4756 if (err) { 4757 dev_err(dev, "ice_init_hw failed: %d\n", err); 4758 return err; 4759 } 4760 4761 /* Some cards require longer initialization times 4762 * due to necessity of loading FW from an external source. 4763 * This can take even half a minute. 4764 */ 4765 if (ice_is_pf_c827(hw)) { 4766 err = ice_wait_for_fw(hw, 30000); 4767 if (err) { 4768 dev_err(dev, "ice_wait_for_fw timed out"); 4769 return err; 4770 } 4771 } 4772 4773 ice_init_feature_support(pf); 4774 4775 err = ice_init_ddp_config(hw, pf); 4776 if (err) 4777 return err; 4778 4779 /* if ice_init_ddp_config fails, ICE_FLAG_ADV_FEATURES bit won't be 4780 * set in pf->state, which will cause ice_is_safe_mode to return 4781 * true 4782 */ 4783 if (ice_is_safe_mode(pf)) { 4784 /* we already got function/device capabilities but these don't 4785 * reflect what the driver needs to do in safe mode. Instead of 4786 * adding conditional logic everywhere to ignore these 4787 * device/function capabilities, override them. 4788 */ 4789 ice_set_safe_mode_caps(hw); 4790 } 4791 4792 err = ice_init_pf(pf); 4793 if (err) { 4794 dev_err(dev, "ice_init_pf failed: %d\n", err); 4795 goto err_init_pf; 4796 } 4797 4798 pf->hw.udp_tunnel_nic.set_port = ice_udp_tunnel_set_port; 4799 pf->hw.udp_tunnel_nic.unset_port = ice_udp_tunnel_unset_port; 4800 pf->hw.udp_tunnel_nic.flags = UDP_TUNNEL_NIC_INFO_MAY_SLEEP; 4801 pf->hw.udp_tunnel_nic.shared = &pf->hw.udp_tunnel_shared; 4802 if (pf->hw.tnl.valid_count[TNL_VXLAN]) { 4803 pf->hw.udp_tunnel_nic.tables[0].n_entries = 4804 pf->hw.tnl.valid_count[TNL_VXLAN]; 4805 pf->hw.udp_tunnel_nic.tables[0].tunnel_types = 4806 UDP_TUNNEL_TYPE_VXLAN; 4807 } 4808 if (pf->hw.tnl.valid_count[TNL_GENEVE]) { 4809 pf->hw.udp_tunnel_nic.tables[1].n_entries = 4810 pf->hw.tnl.valid_count[TNL_GENEVE]; 4811 pf->hw.udp_tunnel_nic.tables[1].tunnel_types = 4812 UDP_TUNNEL_TYPE_GENEVE; 4813 } 4814 4815 err = ice_init_interrupt_scheme(pf); 4816 if (err) { 4817 dev_err(dev, "ice_init_interrupt_scheme failed: %d\n", err); 4818 err = -EIO; 4819 goto err_init_interrupt_scheme; 4820 } 4821 4822 /* In case of MSIX we are going to setup the misc vector right here 4823 * to handle admin queue events etc. In case of legacy and MSI 4824 * the misc functionality and queue processing is combined in 4825 * the same vector and that gets setup at open. 4826 */ 4827 err = ice_req_irq_msix_misc(pf); 4828 if (err) { 4829 dev_err(dev, "setup of misc vector failed: %d\n", err); 4830 goto err_req_irq_msix_misc; 4831 } 4832 4833 return 0; 4834 4835 err_req_irq_msix_misc: 4836 ice_clear_interrupt_scheme(pf); 4837 err_init_interrupt_scheme: 4838 ice_deinit_pf(pf); 4839 err_init_pf: 4840 ice_deinit_hw(hw); 4841 return err; 4842 } 4843 4844 void ice_deinit_dev(struct ice_pf *pf) 4845 { 4846 ice_free_irq_msix_misc(pf); 4847 ice_deinit_pf(pf); 4848 ice_deinit_hw(&pf->hw); 4849 4850 /* Service task is already stopped, so call reset directly. */ 4851 ice_reset(&pf->hw, ICE_RESET_PFR); 4852 pci_wait_for_pending_transaction(pf->pdev); 4853 ice_clear_interrupt_scheme(pf); 4854 } 4855 4856 static void ice_init_features(struct ice_pf *pf) 4857 { 4858 struct device *dev = ice_pf_to_dev(pf); 4859 4860 if (ice_is_safe_mode(pf)) 4861 return; 4862 4863 /* initialize DDP driven features */ 4864 if (test_bit(ICE_FLAG_PTP_SUPPORTED, pf->flags)) 4865 ice_ptp_init(pf); 4866 4867 if (ice_is_feature_supported(pf, ICE_F_GNSS)) 4868 ice_gnss_init(pf); 4869 4870 if (ice_is_feature_supported(pf, ICE_F_CGU) || 4871 ice_is_feature_supported(pf, ICE_F_PHY_RCLK)) 4872 ice_dpll_init(pf); 4873 4874 /* Note: Flow director init failure is non-fatal to load */ 4875 if (ice_init_fdir(pf)) 4876 dev_err(dev, "could not initialize flow director\n"); 4877 4878 /* Note: DCB init failure is non-fatal to load */ 4879 if (ice_init_pf_dcb(pf, false)) { 4880 clear_bit(ICE_FLAG_DCB_CAPABLE, pf->flags); 4881 clear_bit(ICE_FLAG_DCB_ENA, pf->flags); 4882 } else { 4883 ice_cfg_lldp_mib_change(&pf->hw, true); 4884 } 4885 4886 if (ice_init_lag(pf)) 4887 dev_warn(dev, "Failed to init link aggregation support\n"); 4888 4889 ice_hwmon_init(pf); 4890 } 4891 4892 static void ice_deinit_features(struct ice_pf *pf) 4893 { 4894 if (ice_is_safe_mode(pf)) 4895 return; 4896 4897 ice_deinit_lag(pf); 4898 if (test_bit(ICE_FLAG_DCB_CAPABLE, pf->flags)) 4899 ice_cfg_lldp_mib_change(&pf->hw, false); 4900 ice_deinit_fdir(pf); 4901 if (ice_is_feature_supported(pf, ICE_F_GNSS)) 4902 ice_gnss_exit(pf); 4903 if (test_bit(ICE_FLAG_PTP_SUPPORTED, pf->flags)) 4904 ice_ptp_release(pf); 4905 if (test_bit(ICE_FLAG_DPLL, pf->flags)) 4906 ice_dpll_deinit(pf); 4907 if (pf->eswitch_mode == DEVLINK_ESWITCH_MODE_SWITCHDEV) 4908 xa_destroy(&pf->eswitch.reprs); 4909 } 4910 4911 static void ice_init_wakeup(struct ice_pf *pf) 4912 { 4913 /* Save wakeup reason register for later use */ 4914 pf->wakeup_reason = rd32(&pf->hw, PFPM_WUS); 4915 4916 /* check for a power management event */ 4917 ice_print_wake_reason(pf); 4918 4919 /* clear wake status, all bits */ 4920 wr32(&pf->hw, PFPM_WUS, U32_MAX); 4921 4922 /* Disable WoL at init, wait for user to enable */ 4923 device_set_wakeup_enable(ice_pf_to_dev(pf), false); 4924 } 4925 4926 static int ice_init_link(struct ice_pf *pf) 4927 { 4928 struct device *dev = ice_pf_to_dev(pf); 4929 int err; 4930 4931 err = ice_init_link_events(pf->hw.port_info); 4932 if (err) { 4933 dev_err(dev, "ice_init_link_events failed: %d\n", err); 4934 return err; 4935 } 4936 4937 /* not a fatal error if this fails */ 4938 err = ice_init_nvm_phy_type(pf->hw.port_info); 4939 if (err) 4940 dev_err(dev, "ice_init_nvm_phy_type failed: %d\n", err); 4941 4942 /* not a fatal error if this fails */ 4943 err = ice_update_link_info(pf->hw.port_info); 4944 if (err) 4945 dev_err(dev, "ice_update_link_info failed: %d\n", err); 4946 4947 ice_init_link_dflt_override(pf->hw.port_info); 4948 4949 ice_check_link_cfg_err(pf, 4950 pf->hw.port_info->phy.link_info.link_cfg_err); 4951 4952 /* if media available, initialize PHY settings */ 4953 if (pf->hw.port_info->phy.link_info.link_info & 4954 ICE_AQ_MEDIA_AVAILABLE) { 4955 /* not a fatal error if this fails */ 4956 err = ice_init_phy_user_cfg(pf->hw.port_info); 4957 if (err) 4958 dev_err(dev, "ice_init_phy_user_cfg failed: %d\n", err); 4959 4960 if (!test_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, pf->flags)) { 4961 struct ice_vsi *vsi = ice_get_main_vsi(pf); 4962 4963 if (vsi) 4964 ice_configure_phy(vsi); 4965 } 4966 } else { 4967 set_bit(ICE_FLAG_NO_MEDIA, pf->flags); 4968 } 4969 4970 return err; 4971 } 4972 4973 static int ice_init_pf_sw(struct ice_pf *pf) 4974 { 4975 bool dvm = ice_is_dvm_ena(&pf->hw); 4976 struct ice_vsi *vsi; 4977 int err; 4978 4979 /* create switch struct for the switch element created by FW on boot */ 4980 pf->first_sw = kzalloc(sizeof(*pf->first_sw), GFP_KERNEL); 4981 if (!pf->first_sw) 4982 return -ENOMEM; 4983 4984 if (pf->hw.evb_veb) 4985 pf->first_sw->bridge_mode = BRIDGE_MODE_VEB; 4986 else 4987 pf->first_sw->bridge_mode = BRIDGE_MODE_VEPA; 4988 4989 pf->first_sw->pf = pf; 4990 4991 /* record the sw_id available for later use */ 4992 pf->first_sw->sw_id = pf->hw.port_info->sw_id; 4993 4994 err = ice_aq_set_port_params(pf->hw.port_info, dvm, NULL); 4995 if (err) 4996 goto err_aq_set_port_params; 4997 4998 vsi = ice_pf_vsi_setup(pf, pf->hw.port_info); 4999 if (!vsi) { 5000 err = -ENOMEM; 5001 goto err_pf_vsi_setup; 5002 } 5003 5004 return 0; 5005 5006 err_pf_vsi_setup: 5007 err_aq_set_port_params: 5008 kfree(pf->first_sw); 5009 return err; 5010 } 5011 5012 static void ice_deinit_pf_sw(struct ice_pf *pf) 5013 { 5014 struct ice_vsi *vsi = ice_get_main_vsi(pf); 5015 5016 if (!vsi) 5017 return; 5018 5019 ice_vsi_release(vsi); 5020 kfree(pf->first_sw); 5021 } 5022 5023 static int ice_alloc_vsis(struct ice_pf *pf) 5024 { 5025 struct device *dev = ice_pf_to_dev(pf); 5026 5027 pf->num_alloc_vsi = pf->hw.func_caps.guar_num_vsi; 5028 if (!pf->num_alloc_vsi) 5029 return -EIO; 5030 5031 if (pf->num_alloc_vsi > UDP_TUNNEL_NIC_MAX_SHARING_DEVICES) { 5032 dev_warn(dev, 5033 "limiting the VSI count due to UDP tunnel limitation %d > %d\n", 5034 pf->num_alloc_vsi, UDP_TUNNEL_NIC_MAX_SHARING_DEVICES); 5035 pf->num_alloc_vsi = UDP_TUNNEL_NIC_MAX_SHARING_DEVICES; 5036 } 5037 5038 pf->vsi = devm_kcalloc(dev, pf->num_alloc_vsi, sizeof(*pf->vsi), 5039 GFP_KERNEL); 5040 if (!pf->vsi) 5041 return -ENOMEM; 5042 5043 pf->vsi_stats = devm_kcalloc(dev, pf->num_alloc_vsi, 5044 sizeof(*pf->vsi_stats), GFP_KERNEL); 5045 if (!pf->vsi_stats) { 5046 devm_kfree(dev, pf->vsi); 5047 return -ENOMEM; 5048 } 5049 5050 return 0; 5051 } 5052 5053 static void ice_dealloc_vsis(struct ice_pf *pf) 5054 { 5055 devm_kfree(ice_pf_to_dev(pf), pf->vsi_stats); 5056 pf->vsi_stats = NULL; 5057 5058 pf->num_alloc_vsi = 0; 5059 devm_kfree(ice_pf_to_dev(pf), pf->vsi); 5060 pf->vsi = NULL; 5061 } 5062 5063 static int ice_init_devlink(struct ice_pf *pf) 5064 { 5065 int err; 5066 5067 err = ice_devlink_register_params(pf); 5068 if (err) 5069 return err; 5070 5071 ice_devlink_init_regions(pf); 5072 ice_devlink_register(pf); 5073 5074 return 0; 5075 } 5076 5077 static void ice_deinit_devlink(struct ice_pf *pf) 5078 { 5079 ice_devlink_unregister(pf); 5080 ice_devlink_destroy_regions(pf); 5081 ice_devlink_unregister_params(pf); 5082 } 5083 5084 static int ice_init(struct ice_pf *pf) 5085 { 5086 int err; 5087 5088 err = ice_init_dev(pf); 5089 if (err) 5090 return err; 5091 5092 err = ice_alloc_vsis(pf); 5093 if (err) 5094 goto err_alloc_vsis; 5095 5096 err = ice_init_pf_sw(pf); 5097 if (err) 5098 goto err_init_pf_sw; 5099 5100 ice_init_wakeup(pf); 5101 5102 err = ice_init_link(pf); 5103 if (err) 5104 goto err_init_link; 5105 5106 err = ice_send_version(pf); 5107 if (err) 5108 goto err_init_link; 5109 5110 ice_verify_cacheline_size(pf); 5111 5112 if (ice_is_safe_mode(pf)) 5113 ice_set_safe_mode_vlan_cfg(pf); 5114 else 5115 /* print PCI link speed and width */ 5116 pcie_print_link_status(pf->pdev); 5117 5118 /* ready to go, so clear down state bit */ 5119 clear_bit(ICE_DOWN, pf->state); 5120 clear_bit(ICE_SERVICE_DIS, pf->state); 5121 5122 /* since everything is good, start the service timer */ 5123 mod_timer(&pf->serv_tmr, round_jiffies(jiffies + pf->serv_tmr_period)); 5124 5125 return 0; 5126 5127 err_init_link: 5128 ice_deinit_pf_sw(pf); 5129 err_init_pf_sw: 5130 ice_dealloc_vsis(pf); 5131 err_alloc_vsis: 5132 ice_deinit_dev(pf); 5133 return err; 5134 } 5135 5136 static void ice_deinit(struct ice_pf *pf) 5137 { 5138 set_bit(ICE_SERVICE_DIS, pf->state); 5139 set_bit(ICE_DOWN, pf->state); 5140 5141 ice_deinit_pf_sw(pf); 5142 ice_dealloc_vsis(pf); 5143 ice_deinit_dev(pf); 5144 } 5145 5146 /** 5147 * ice_load - load pf by init hw and starting VSI 5148 * @pf: pointer to the pf instance 5149 * 5150 * This function has to be called under devl_lock. 5151 */ 5152 int ice_load(struct ice_pf *pf) 5153 { 5154 struct ice_vsi *vsi; 5155 int err; 5156 5157 devl_assert_locked(priv_to_devlink(pf)); 5158 5159 vsi = ice_get_main_vsi(pf); 5160 5161 /* init channel list */ 5162 INIT_LIST_HEAD(&vsi->ch_list); 5163 5164 err = ice_cfg_netdev(vsi); 5165 if (err) 5166 return err; 5167 5168 /* Setup DCB netlink interface */ 5169 ice_dcbnl_setup(vsi); 5170 5171 err = ice_init_mac_fltr(pf); 5172 if (err) 5173 goto err_init_mac_fltr; 5174 5175 err = ice_devlink_create_pf_port(pf); 5176 if (err) 5177 goto err_devlink_create_pf_port; 5178 5179 SET_NETDEV_DEVLINK_PORT(vsi->netdev, &pf->devlink_port); 5180 5181 err = ice_register_netdev(vsi); 5182 if (err) 5183 goto err_register_netdev; 5184 5185 err = ice_tc_indir_block_register(vsi); 5186 if (err) 5187 goto err_tc_indir_block_register; 5188 5189 ice_napi_add(vsi); 5190 5191 err = ice_init_rdma(pf); 5192 if (err) 5193 goto err_init_rdma; 5194 5195 ice_init_features(pf); 5196 ice_service_task_restart(pf); 5197 5198 clear_bit(ICE_DOWN, pf->state); 5199 5200 return 0; 5201 5202 err_init_rdma: 5203 ice_tc_indir_block_unregister(vsi); 5204 err_tc_indir_block_register: 5205 ice_unregister_netdev(vsi); 5206 err_register_netdev: 5207 ice_devlink_destroy_pf_port(pf); 5208 err_devlink_create_pf_port: 5209 err_init_mac_fltr: 5210 ice_decfg_netdev(vsi); 5211 return err; 5212 } 5213 5214 /** 5215 * ice_unload - unload pf by stopping VSI and deinit hw 5216 * @pf: pointer to the pf instance 5217 * 5218 * This function has to be called under devl_lock. 5219 */ 5220 void ice_unload(struct ice_pf *pf) 5221 { 5222 struct ice_vsi *vsi = ice_get_main_vsi(pf); 5223 5224 devl_assert_locked(priv_to_devlink(pf)); 5225 5226 ice_deinit_features(pf); 5227 ice_deinit_rdma(pf); 5228 ice_tc_indir_block_unregister(vsi); 5229 ice_unregister_netdev(vsi); 5230 ice_devlink_destroy_pf_port(pf); 5231 ice_decfg_netdev(vsi); 5232 } 5233 5234 /** 5235 * ice_probe - Device initialization routine 5236 * @pdev: PCI device information struct 5237 * @ent: entry in ice_pci_tbl 5238 * 5239 * Returns 0 on success, negative on failure 5240 */ 5241 static int 5242 ice_probe(struct pci_dev *pdev, const struct pci_device_id __always_unused *ent) 5243 { 5244 struct device *dev = &pdev->dev; 5245 struct ice_adapter *adapter; 5246 struct ice_pf *pf; 5247 struct ice_hw *hw; 5248 int err; 5249 5250 if (pdev->is_virtfn) { 5251 dev_err(dev, "can't probe a virtual function\n"); 5252 return -EINVAL; 5253 } 5254 5255 /* when under a kdump kernel initiate a reset before enabling the 5256 * device in order to clear out any pending DMA transactions. These 5257 * transactions can cause some systems to machine check when doing 5258 * the pcim_enable_device() below. 5259 */ 5260 if (is_kdump_kernel()) { 5261 pci_save_state(pdev); 5262 pci_clear_master(pdev); 5263 err = pcie_flr(pdev); 5264 if (err) 5265 return err; 5266 pci_restore_state(pdev); 5267 } 5268 5269 /* this driver uses devres, see 5270 * Documentation/driver-api/driver-model/devres.rst 5271 */ 5272 err = pcim_enable_device(pdev); 5273 if (err) 5274 return err; 5275 5276 err = pcim_iomap_regions(pdev, BIT(ICE_BAR0), dev_driver_string(dev)); 5277 if (err) { 5278 dev_err(dev, "BAR0 I/O map error %d\n", err); 5279 return err; 5280 } 5281 5282 pf = ice_allocate_pf(dev); 5283 if (!pf) 5284 return -ENOMEM; 5285 5286 /* initialize Auxiliary index to invalid value */ 5287 pf->aux_idx = -1; 5288 5289 /* set up for high or low DMA */ 5290 err = dma_set_mask_and_coherent(dev, DMA_BIT_MASK(64)); 5291 if (err) { 5292 dev_err(dev, "DMA configuration failed: 0x%x\n", err); 5293 return err; 5294 } 5295 5296 pci_set_master(pdev); 5297 5298 adapter = ice_adapter_get(pdev); 5299 if (IS_ERR(adapter)) 5300 return PTR_ERR(adapter); 5301 5302 pf->pdev = pdev; 5303 pf->adapter = adapter; 5304 pci_set_drvdata(pdev, pf); 5305 set_bit(ICE_DOWN, pf->state); 5306 /* Disable service task until DOWN bit is cleared */ 5307 set_bit(ICE_SERVICE_DIS, pf->state); 5308 5309 hw = &pf->hw; 5310 hw->hw_addr = pcim_iomap_table(pdev)[ICE_BAR0]; 5311 pci_save_state(pdev); 5312 5313 hw->back = pf; 5314 hw->port_info = NULL; 5315 hw->vendor_id = pdev->vendor; 5316 hw->device_id = pdev->device; 5317 pci_read_config_byte(pdev, PCI_REVISION_ID, &hw->revision_id); 5318 hw->subsystem_vendor_id = pdev->subsystem_vendor; 5319 hw->subsystem_device_id = pdev->subsystem_device; 5320 hw->bus.device = PCI_SLOT(pdev->devfn); 5321 hw->bus.func = PCI_FUNC(pdev->devfn); 5322 ice_set_ctrlq_len(hw); 5323 5324 pf->msg_enable = netif_msg_init(debug, ICE_DFLT_NETIF_M); 5325 5326 #ifndef CONFIG_DYNAMIC_DEBUG 5327 if (debug < -1) 5328 hw->debug_mask = debug; 5329 #endif 5330 5331 err = ice_init(pf); 5332 if (err) 5333 goto err_init; 5334 5335 devl_lock(priv_to_devlink(pf)); 5336 err = ice_load(pf); 5337 if (err) 5338 goto err_load; 5339 5340 err = ice_init_devlink(pf); 5341 if (err) 5342 goto err_init_devlink; 5343 devl_unlock(priv_to_devlink(pf)); 5344 5345 return 0; 5346 5347 err_init_devlink: 5348 ice_unload(pf); 5349 err_load: 5350 devl_unlock(priv_to_devlink(pf)); 5351 ice_deinit(pf); 5352 err_init: 5353 ice_adapter_put(pdev); 5354 pci_disable_device(pdev); 5355 return err; 5356 } 5357 5358 /** 5359 * ice_set_wake - enable or disable Wake on LAN 5360 * @pf: pointer to the PF struct 5361 * 5362 * Simple helper for WoL control 5363 */ 5364 static void ice_set_wake(struct ice_pf *pf) 5365 { 5366 struct ice_hw *hw = &pf->hw; 5367 bool wol = pf->wol_ena; 5368 5369 /* clear wake state, otherwise new wake events won't fire */ 5370 wr32(hw, PFPM_WUS, U32_MAX); 5371 5372 /* enable / disable APM wake up, no RMW needed */ 5373 wr32(hw, PFPM_APM, wol ? PFPM_APM_APME_M : 0); 5374 5375 /* set magic packet filter enabled */ 5376 wr32(hw, PFPM_WUFC, wol ? PFPM_WUFC_MAG_M : 0); 5377 } 5378 5379 /** 5380 * ice_setup_mc_magic_wake - setup device to wake on multicast magic packet 5381 * @pf: pointer to the PF struct 5382 * 5383 * Issue firmware command to enable multicast magic wake, making 5384 * sure that any locally administered address (LAA) is used for 5385 * wake, and that PF reset doesn't undo the LAA. 5386 */ 5387 static void ice_setup_mc_magic_wake(struct ice_pf *pf) 5388 { 5389 struct device *dev = ice_pf_to_dev(pf); 5390 struct ice_hw *hw = &pf->hw; 5391 u8 mac_addr[ETH_ALEN]; 5392 struct ice_vsi *vsi; 5393 int status; 5394 u8 flags; 5395 5396 if (!pf->wol_ena) 5397 return; 5398 5399 vsi = ice_get_main_vsi(pf); 5400 if (!vsi) 5401 return; 5402 5403 /* Get current MAC address in case it's an LAA */ 5404 if (vsi->netdev) 5405 ether_addr_copy(mac_addr, vsi->netdev->dev_addr); 5406 else 5407 ether_addr_copy(mac_addr, vsi->port_info->mac.perm_addr); 5408 5409 flags = ICE_AQC_MAN_MAC_WR_MC_MAG_EN | 5410 ICE_AQC_MAN_MAC_UPDATE_LAA_WOL | 5411 ICE_AQC_MAN_MAC_WR_WOL_LAA_PFR_KEEP; 5412 5413 status = ice_aq_manage_mac_write(hw, mac_addr, flags, NULL); 5414 if (status) 5415 dev_err(dev, "Failed to enable Multicast Magic Packet wake, err %d aq_err %s\n", 5416 status, ice_aq_str(hw->adminq.sq_last_status)); 5417 } 5418 5419 /** 5420 * ice_remove - Device removal routine 5421 * @pdev: PCI device information struct 5422 */ 5423 static void ice_remove(struct pci_dev *pdev) 5424 { 5425 struct ice_pf *pf = pci_get_drvdata(pdev); 5426 int i; 5427 5428 for (i = 0; i < ICE_MAX_RESET_WAIT; i++) { 5429 if (!ice_is_reset_in_progress(pf->state)) 5430 break; 5431 msleep(100); 5432 } 5433 5434 if (test_bit(ICE_FLAG_SRIOV_ENA, pf->flags)) { 5435 set_bit(ICE_VF_RESETS_DISABLED, pf->state); 5436 ice_free_vfs(pf); 5437 } 5438 5439 ice_hwmon_exit(pf); 5440 5441 ice_service_task_stop(pf); 5442 ice_aq_cancel_waiting_tasks(pf); 5443 set_bit(ICE_DOWN, pf->state); 5444 5445 if (!ice_is_safe_mode(pf)) 5446 ice_remove_arfs(pf); 5447 5448 devl_lock(priv_to_devlink(pf)); 5449 ice_dealloc_all_dynamic_ports(pf); 5450 ice_deinit_devlink(pf); 5451 5452 ice_unload(pf); 5453 devl_unlock(priv_to_devlink(pf)); 5454 5455 ice_deinit(pf); 5456 ice_vsi_release_all(pf); 5457 5458 ice_setup_mc_magic_wake(pf); 5459 ice_set_wake(pf); 5460 5461 ice_adapter_put(pdev); 5462 pci_disable_device(pdev); 5463 } 5464 5465 /** 5466 * ice_shutdown - PCI callback for shutting down device 5467 * @pdev: PCI device information struct 5468 */ 5469 static void ice_shutdown(struct pci_dev *pdev) 5470 { 5471 struct ice_pf *pf = pci_get_drvdata(pdev); 5472 5473 ice_remove(pdev); 5474 5475 if (system_state == SYSTEM_POWER_OFF) { 5476 pci_wake_from_d3(pdev, pf->wol_ena); 5477 pci_set_power_state(pdev, PCI_D3hot); 5478 } 5479 } 5480 5481 /** 5482 * ice_prepare_for_shutdown - prep for PCI shutdown 5483 * @pf: board private structure 5484 * 5485 * Inform or close all dependent features in prep for PCI device shutdown 5486 */ 5487 static void ice_prepare_for_shutdown(struct ice_pf *pf) 5488 { 5489 struct ice_hw *hw = &pf->hw; 5490 u32 v; 5491 5492 /* Notify VFs of impending reset */ 5493 if (ice_check_sq_alive(hw, &hw->mailboxq)) 5494 ice_vc_notify_reset(pf); 5495 5496 dev_dbg(ice_pf_to_dev(pf), "Tearing down internal switch for shutdown\n"); 5497 5498 /* disable the VSIs and their queues that are not already DOWN */ 5499 ice_pf_dis_all_vsi(pf, false); 5500 5501 ice_for_each_vsi(pf, v) 5502 if (pf->vsi[v]) 5503 pf->vsi[v]->vsi_num = 0; 5504 5505 ice_shutdown_all_ctrlq(hw, true); 5506 } 5507 5508 /** 5509 * ice_reinit_interrupt_scheme - Reinitialize interrupt scheme 5510 * @pf: board private structure to reinitialize 5511 * 5512 * This routine reinitialize interrupt scheme that was cleared during 5513 * power management suspend callback. 5514 * 5515 * This should be called during resume routine to re-allocate the q_vectors 5516 * and reacquire interrupts. 5517 */ 5518 static int ice_reinit_interrupt_scheme(struct ice_pf *pf) 5519 { 5520 struct device *dev = ice_pf_to_dev(pf); 5521 int ret, v; 5522 5523 /* Since we clear MSIX flag during suspend, we need to 5524 * set it back during resume... 5525 */ 5526 5527 ret = ice_init_interrupt_scheme(pf); 5528 if (ret) { 5529 dev_err(dev, "Failed to re-initialize interrupt %d\n", ret); 5530 return ret; 5531 } 5532 5533 /* Remap vectors and rings, after successful re-init interrupts */ 5534 ice_for_each_vsi(pf, v) { 5535 if (!pf->vsi[v]) 5536 continue; 5537 5538 ret = ice_vsi_alloc_q_vectors(pf->vsi[v]); 5539 if (ret) 5540 goto err_reinit; 5541 ice_vsi_map_rings_to_vectors(pf->vsi[v]); 5542 rtnl_lock(); 5543 ice_vsi_set_napi_queues(pf->vsi[v]); 5544 rtnl_unlock(); 5545 } 5546 5547 ret = ice_req_irq_msix_misc(pf); 5548 if (ret) { 5549 dev_err(dev, "Setting up misc vector failed after device suspend %d\n", 5550 ret); 5551 goto err_reinit; 5552 } 5553 5554 return 0; 5555 5556 err_reinit: 5557 while (v--) 5558 if (pf->vsi[v]) { 5559 rtnl_lock(); 5560 ice_vsi_clear_napi_queues(pf->vsi[v]); 5561 rtnl_unlock(); 5562 ice_vsi_free_q_vectors(pf->vsi[v]); 5563 } 5564 5565 return ret; 5566 } 5567 5568 /** 5569 * ice_suspend 5570 * @dev: generic device information structure 5571 * 5572 * Power Management callback to quiesce the device and prepare 5573 * for D3 transition. 5574 */ 5575 static int ice_suspend(struct device *dev) 5576 { 5577 struct pci_dev *pdev = to_pci_dev(dev); 5578 struct ice_pf *pf; 5579 int disabled, v; 5580 5581 pf = pci_get_drvdata(pdev); 5582 5583 if (!ice_pf_state_is_nominal(pf)) { 5584 dev_err(dev, "Device is not ready, no need to suspend it\n"); 5585 return -EBUSY; 5586 } 5587 5588 /* Stop watchdog tasks until resume completion. 5589 * Even though it is most likely that the service task is 5590 * disabled if the device is suspended or down, the service task's 5591 * state is controlled by a different state bit, and we should 5592 * store and honor whatever state that bit is in at this point. 5593 */ 5594 disabled = ice_service_task_stop(pf); 5595 5596 ice_deinit_rdma(pf); 5597 5598 /* Already suspended?, then there is nothing to do */ 5599 if (test_and_set_bit(ICE_SUSPENDED, pf->state)) { 5600 if (!disabled) 5601 ice_service_task_restart(pf); 5602 return 0; 5603 } 5604 5605 if (test_bit(ICE_DOWN, pf->state) || 5606 ice_is_reset_in_progress(pf->state)) { 5607 dev_err(dev, "can't suspend device in reset or already down\n"); 5608 if (!disabled) 5609 ice_service_task_restart(pf); 5610 return 0; 5611 } 5612 5613 ice_setup_mc_magic_wake(pf); 5614 5615 ice_prepare_for_shutdown(pf); 5616 5617 ice_set_wake(pf); 5618 5619 /* Free vectors, clear the interrupt scheme and release IRQs 5620 * for proper hibernation, especially with large number of CPUs. 5621 * Otherwise hibernation might fail when mapping all the vectors back 5622 * to CPU0. 5623 */ 5624 ice_free_irq_msix_misc(pf); 5625 ice_for_each_vsi(pf, v) { 5626 if (!pf->vsi[v]) 5627 continue; 5628 rtnl_lock(); 5629 ice_vsi_clear_napi_queues(pf->vsi[v]); 5630 rtnl_unlock(); 5631 ice_vsi_free_q_vectors(pf->vsi[v]); 5632 } 5633 ice_clear_interrupt_scheme(pf); 5634 5635 pci_save_state(pdev); 5636 pci_wake_from_d3(pdev, pf->wol_ena); 5637 pci_set_power_state(pdev, PCI_D3hot); 5638 return 0; 5639 } 5640 5641 /** 5642 * ice_resume - PM callback for waking up from D3 5643 * @dev: generic device information structure 5644 */ 5645 static int ice_resume(struct device *dev) 5646 { 5647 struct pci_dev *pdev = to_pci_dev(dev); 5648 enum ice_reset_req reset_type; 5649 struct ice_pf *pf; 5650 struct ice_hw *hw; 5651 int ret; 5652 5653 pci_set_power_state(pdev, PCI_D0); 5654 pci_restore_state(pdev); 5655 pci_save_state(pdev); 5656 5657 if (!pci_device_is_present(pdev)) 5658 return -ENODEV; 5659 5660 ret = pci_enable_device_mem(pdev); 5661 if (ret) { 5662 dev_err(dev, "Cannot enable device after suspend\n"); 5663 return ret; 5664 } 5665 5666 pf = pci_get_drvdata(pdev); 5667 hw = &pf->hw; 5668 5669 pf->wakeup_reason = rd32(hw, PFPM_WUS); 5670 ice_print_wake_reason(pf); 5671 5672 /* We cleared the interrupt scheme when we suspended, so we need to 5673 * restore it now to resume device functionality. 5674 */ 5675 ret = ice_reinit_interrupt_scheme(pf); 5676 if (ret) 5677 dev_err(dev, "Cannot restore interrupt scheme: %d\n", ret); 5678 5679 ret = ice_init_rdma(pf); 5680 if (ret) 5681 dev_err(dev, "Reinitialize RDMA during resume failed: %d\n", 5682 ret); 5683 5684 clear_bit(ICE_DOWN, pf->state); 5685 /* Now perform PF reset and rebuild */ 5686 reset_type = ICE_RESET_PFR; 5687 /* re-enable service task for reset, but allow reset to schedule it */ 5688 clear_bit(ICE_SERVICE_DIS, pf->state); 5689 5690 if (ice_schedule_reset(pf, reset_type)) 5691 dev_err(dev, "Reset during resume failed.\n"); 5692 5693 clear_bit(ICE_SUSPENDED, pf->state); 5694 ice_service_task_restart(pf); 5695 5696 /* Restart the service task */ 5697 mod_timer(&pf->serv_tmr, round_jiffies(jiffies + pf->serv_tmr_period)); 5698 5699 return 0; 5700 } 5701 5702 /** 5703 * ice_pci_err_detected - warning that PCI error has been detected 5704 * @pdev: PCI device information struct 5705 * @err: the type of PCI error 5706 * 5707 * Called to warn that something happened on the PCI bus and the error handling 5708 * is in progress. Allows the driver to gracefully prepare/handle PCI errors. 5709 */ 5710 static pci_ers_result_t 5711 ice_pci_err_detected(struct pci_dev *pdev, pci_channel_state_t err) 5712 { 5713 struct ice_pf *pf = pci_get_drvdata(pdev); 5714 5715 if (!pf) { 5716 dev_err(&pdev->dev, "%s: unrecoverable device error %d\n", 5717 __func__, err); 5718 return PCI_ERS_RESULT_DISCONNECT; 5719 } 5720 5721 if (!test_bit(ICE_SUSPENDED, pf->state)) { 5722 ice_service_task_stop(pf); 5723 5724 if (!test_bit(ICE_PREPARED_FOR_RESET, pf->state)) { 5725 set_bit(ICE_PFR_REQ, pf->state); 5726 ice_prepare_for_reset(pf, ICE_RESET_PFR); 5727 } 5728 } 5729 5730 return PCI_ERS_RESULT_NEED_RESET; 5731 } 5732 5733 /** 5734 * ice_pci_err_slot_reset - a PCI slot reset has just happened 5735 * @pdev: PCI device information struct 5736 * 5737 * Called to determine if the driver can recover from the PCI slot reset by 5738 * using a register read to determine if the device is recoverable. 5739 */ 5740 static pci_ers_result_t ice_pci_err_slot_reset(struct pci_dev *pdev) 5741 { 5742 struct ice_pf *pf = pci_get_drvdata(pdev); 5743 pci_ers_result_t result; 5744 int err; 5745 u32 reg; 5746 5747 err = pci_enable_device_mem(pdev); 5748 if (err) { 5749 dev_err(&pdev->dev, "Cannot re-enable PCI device after reset, error %d\n", 5750 err); 5751 result = PCI_ERS_RESULT_DISCONNECT; 5752 } else { 5753 pci_set_master(pdev); 5754 pci_restore_state(pdev); 5755 pci_save_state(pdev); 5756 pci_wake_from_d3(pdev, false); 5757 5758 /* Check for life */ 5759 reg = rd32(&pf->hw, GLGEN_RTRIG); 5760 if (!reg) 5761 result = PCI_ERS_RESULT_RECOVERED; 5762 else 5763 result = PCI_ERS_RESULT_DISCONNECT; 5764 } 5765 5766 return result; 5767 } 5768 5769 /** 5770 * ice_pci_err_resume - restart operations after PCI error recovery 5771 * @pdev: PCI device information struct 5772 * 5773 * Called to allow the driver to bring things back up after PCI error and/or 5774 * reset recovery have finished 5775 */ 5776 static void ice_pci_err_resume(struct pci_dev *pdev) 5777 { 5778 struct ice_pf *pf = pci_get_drvdata(pdev); 5779 5780 if (!pf) { 5781 dev_err(&pdev->dev, "%s failed, device is unrecoverable\n", 5782 __func__); 5783 return; 5784 } 5785 5786 if (test_bit(ICE_SUSPENDED, pf->state)) { 5787 dev_dbg(&pdev->dev, "%s failed to resume normal operations!\n", 5788 __func__); 5789 return; 5790 } 5791 5792 ice_restore_all_vfs_msi_state(pf); 5793 5794 ice_do_reset(pf, ICE_RESET_PFR); 5795 ice_service_task_restart(pf); 5796 mod_timer(&pf->serv_tmr, round_jiffies(jiffies + pf->serv_tmr_period)); 5797 } 5798 5799 /** 5800 * ice_pci_err_reset_prepare - prepare device driver for PCI reset 5801 * @pdev: PCI device information struct 5802 */ 5803 static void ice_pci_err_reset_prepare(struct pci_dev *pdev) 5804 { 5805 struct ice_pf *pf = pci_get_drvdata(pdev); 5806 5807 if (!test_bit(ICE_SUSPENDED, pf->state)) { 5808 ice_service_task_stop(pf); 5809 5810 if (!test_bit(ICE_PREPARED_FOR_RESET, pf->state)) { 5811 set_bit(ICE_PFR_REQ, pf->state); 5812 ice_prepare_for_reset(pf, ICE_RESET_PFR); 5813 } 5814 } 5815 } 5816 5817 /** 5818 * ice_pci_err_reset_done - PCI reset done, device driver reset can begin 5819 * @pdev: PCI device information struct 5820 */ 5821 static void ice_pci_err_reset_done(struct pci_dev *pdev) 5822 { 5823 ice_pci_err_resume(pdev); 5824 } 5825 5826 /* ice_pci_tbl - PCI Device ID Table 5827 * 5828 * Wildcard entries (PCI_ANY_ID) should come last 5829 * Last entry must be all 0s 5830 * 5831 * { Vendor ID, Device ID, SubVendor ID, SubDevice ID, 5832 * Class, Class Mask, private data (not used) } 5833 */ 5834 static const struct pci_device_id ice_pci_tbl[] = { 5835 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810C_BACKPLANE) }, 5836 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810C_QSFP) }, 5837 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810C_SFP) }, 5838 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810_XXV_BACKPLANE) }, 5839 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810_XXV_QSFP) }, 5840 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810_XXV_SFP) }, 5841 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_BACKPLANE) }, 5842 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_QSFP) }, 5843 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_SFP) }, 5844 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_10G_BASE_T) }, 5845 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_SGMII) }, 5846 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_BACKPLANE) }, 5847 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_QSFP) }, 5848 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_SFP) }, 5849 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_10G_BASE_T) }, 5850 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_SGMII) }, 5851 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822L_BACKPLANE) }, 5852 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822L_SFP) }, 5853 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822L_10G_BASE_T) }, 5854 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822L_SGMII) }, 5855 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_BACKPLANE) }, 5856 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_SFP) }, 5857 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_10G_BASE_T) }, 5858 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_1GBE) }, 5859 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_QSFP) }, 5860 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822_SI_DFLT) }, 5861 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E825C_BACKPLANE), }, 5862 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E825C_QSFP), }, 5863 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E825C_SFP), }, 5864 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E825C_SGMII), }, 5865 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E830CC_BACKPLANE) }, 5866 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E830CC_QSFP56) }, 5867 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E830CC_SFP) }, 5868 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E830CC_SFP_DD) }, 5869 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E830C_BACKPLANE), }, 5870 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E830_XXV_BACKPLANE), }, 5871 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E830C_QSFP), }, 5872 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E830_XXV_QSFP), }, 5873 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E830C_SFP), }, 5874 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E830_XXV_SFP), }, 5875 /* required last entry */ 5876 {} 5877 }; 5878 MODULE_DEVICE_TABLE(pci, ice_pci_tbl); 5879 5880 static DEFINE_SIMPLE_DEV_PM_OPS(ice_pm_ops, ice_suspend, ice_resume); 5881 5882 static const struct pci_error_handlers ice_pci_err_handler = { 5883 .error_detected = ice_pci_err_detected, 5884 .slot_reset = ice_pci_err_slot_reset, 5885 .reset_prepare = ice_pci_err_reset_prepare, 5886 .reset_done = ice_pci_err_reset_done, 5887 .resume = ice_pci_err_resume 5888 }; 5889 5890 static struct pci_driver ice_driver = { 5891 .name = KBUILD_MODNAME, 5892 .id_table = ice_pci_tbl, 5893 .probe = ice_probe, 5894 .remove = ice_remove, 5895 .driver.pm = pm_sleep_ptr(&ice_pm_ops), 5896 .shutdown = ice_shutdown, 5897 .sriov_configure = ice_sriov_configure, 5898 .sriov_get_vf_total_msix = ice_sriov_get_vf_total_msix, 5899 .sriov_set_msix_vec_count = ice_sriov_set_msix_vec_count, 5900 .err_handler = &ice_pci_err_handler 5901 }; 5902 5903 /** 5904 * ice_module_init - Driver registration routine 5905 * 5906 * ice_module_init is the first routine called when the driver is 5907 * loaded. All it does is register with the PCI subsystem. 5908 */ 5909 static int __init ice_module_init(void) 5910 { 5911 int status = -ENOMEM; 5912 5913 pr_info("%s\n", ice_driver_string); 5914 pr_info("%s\n", ice_copyright); 5915 5916 ice_adv_lnk_speed_maps_init(); 5917 5918 ice_wq = alloc_workqueue("%s", 0, 0, KBUILD_MODNAME); 5919 if (!ice_wq) { 5920 pr_err("Failed to create workqueue\n"); 5921 return status; 5922 } 5923 5924 ice_lag_wq = alloc_ordered_workqueue("ice_lag_wq", 0); 5925 if (!ice_lag_wq) { 5926 pr_err("Failed to create LAG workqueue\n"); 5927 goto err_dest_wq; 5928 } 5929 5930 ice_debugfs_init(); 5931 5932 status = pci_register_driver(&ice_driver); 5933 if (status) { 5934 pr_err("failed to register PCI driver, err %d\n", status); 5935 goto err_dest_lag_wq; 5936 } 5937 5938 status = ice_sf_driver_register(); 5939 if (status) { 5940 pr_err("Failed to register SF driver, err %d\n", status); 5941 goto err_sf_driver; 5942 } 5943 5944 return 0; 5945 5946 err_sf_driver: 5947 pci_unregister_driver(&ice_driver); 5948 err_dest_lag_wq: 5949 destroy_workqueue(ice_lag_wq); 5950 ice_debugfs_exit(); 5951 err_dest_wq: 5952 destroy_workqueue(ice_wq); 5953 return status; 5954 } 5955 module_init(ice_module_init); 5956 5957 /** 5958 * ice_module_exit - Driver exit cleanup routine 5959 * 5960 * ice_module_exit is called just before the driver is removed 5961 * from memory. 5962 */ 5963 static void __exit ice_module_exit(void) 5964 { 5965 ice_sf_driver_unregister(); 5966 pci_unregister_driver(&ice_driver); 5967 ice_debugfs_exit(); 5968 destroy_workqueue(ice_wq); 5969 destroy_workqueue(ice_lag_wq); 5970 pr_info("module unloaded\n"); 5971 } 5972 module_exit(ice_module_exit); 5973 5974 /** 5975 * ice_set_mac_address - NDO callback to set MAC address 5976 * @netdev: network interface device structure 5977 * @pi: pointer to an address structure 5978 * 5979 * Returns 0 on success, negative on failure 5980 */ 5981 static int ice_set_mac_address(struct net_device *netdev, void *pi) 5982 { 5983 struct ice_netdev_priv *np = netdev_priv(netdev); 5984 struct ice_vsi *vsi = np->vsi; 5985 struct ice_pf *pf = vsi->back; 5986 struct ice_hw *hw = &pf->hw; 5987 struct sockaddr *addr = pi; 5988 u8 old_mac[ETH_ALEN]; 5989 u8 flags = 0; 5990 u8 *mac; 5991 int err; 5992 5993 mac = (u8 *)addr->sa_data; 5994 5995 if (!is_valid_ether_addr(mac)) 5996 return -EADDRNOTAVAIL; 5997 5998 if (test_bit(ICE_DOWN, pf->state) || 5999 ice_is_reset_in_progress(pf->state)) { 6000 netdev_err(netdev, "can't set mac %pM. device not ready\n", 6001 mac); 6002 return -EBUSY; 6003 } 6004 6005 if (ice_chnl_dmac_fltr_cnt(pf)) { 6006 netdev_err(netdev, "can't set mac %pM. Device has tc-flower filters, delete all of them and try again\n", 6007 mac); 6008 return -EAGAIN; 6009 } 6010 6011 netif_addr_lock_bh(netdev); 6012 ether_addr_copy(old_mac, netdev->dev_addr); 6013 /* change the netdev's MAC address */ 6014 eth_hw_addr_set(netdev, mac); 6015 netif_addr_unlock_bh(netdev); 6016 6017 /* Clean up old MAC filter. Not an error if old filter doesn't exist */ 6018 err = ice_fltr_remove_mac(vsi, old_mac, ICE_FWD_TO_VSI); 6019 if (err && err != -ENOENT) { 6020 err = -EADDRNOTAVAIL; 6021 goto err_update_filters; 6022 } 6023 6024 /* Add filter for new MAC. If filter exists, return success */ 6025 err = ice_fltr_add_mac(vsi, mac, ICE_FWD_TO_VSI); 6026 if (err == -EEXIST) { 6027 /* Although this MAC filter is already present in hardware it's 6028 * possible in some cases (e.g. bonding) that dev_addr was 6029 * modified outside of the driver and needs to be restored back 6030 * to this value. 6031 */ 6032 netdev_dbg(netdev, "filter for MAC %pM already exists\n", mac); 6033 6034 return 0; 6035 } else if (err) { 6036 /* error if the new filter addition failed */ 6037 err = -EADDRNOTAVAIL; 6038 } 6039 6040 err_update_filters: 6041 if (err) { 6042 netdev_err(netdev, "can't set MAC %pM. filter update failed\n", 6043 mac); 6044 netif_addr_lock_bh(netdev); 6045 eth_hw_addr_set(netdev, old_mac); 6046 netif_addr_unlock_bh(netdev); 6047 return err; 6048 } 6049 6050 netdev_dbg(vsi->netdev, "updated MAC address to %pM\n", 6051 netdev->dev_addr); 6052 6053 /* write new MAC address to the firmware */ 6054 flags = ICE_AQC_MAN_MAC_UPDATE_LAA_WOL; 6055 err = ice_aq_manage_mac_write(hw, mac, flags, NULL); 6056 if (err) { 6057 netdev_err(netdev, "can't set MAC %pM. write to firmware failed error %d\n", 6058 mac, err); 6059 } 6060 return 0; 6061 } 6062 6063 /** 6064 * ice_set_rx_mode - NDO callback to set the netdev filters 6065 * @netdev: network interface device structure 6066 */ 6067 static void ice_set_rx_mode(struct net_device *netdev) 6068 { 6069 struct ice_netdev_priv *np = netdev_priv(netdev); 6070 struct ice_vsi *vsi = np->vsi; 6071 6072 if (!vsi || ice_is_switchdev_running(vsi->back)) 6073 return; 6074 6075 /* Set the flags to synchronize filters 6076 * ndo_set_rx_mode may be triggered even without a change in netdev 6077 * flags 6078 */ 6079 set_bit(ICE_VSI_UMAC_FLTR_CHANGED, vsi->state); 6080 set_bit(ICE_VSI_MMAC_FLTR_CHANGED, vsi->state); 6081 set_bit(ICE_FLAG_FLTR_SYNC, vsi->back->flags); 6082 6083 /* schedule our worker thread which will take care of 6084 * applying the new filter changes 6085 */ 6086 ice_service_task_schedule(vsi->back); 6087 } 6088 6089 /** 6090 * ice_set_tx_maxrate - NDO callback to set the maximum per-queue bitrate 6091 * @netdev: network interface device structure 6092 * @queue_index: Queue ID 6093 * @maxrate: maximum bandwidth in Mbps 6094 */ 6095 static int 6096 ice_set_tx_maxrate(struct net_device *netdev, int queue_index, u32 maxrate) 6097 { 6098 struct ice_netdev_priv *np = netdev_priv(netdev); 6099 struct ice_vsi *vsi = np->vsi; 6100 u16 q_handle; 6101 int status; 6102 u8 tc; 6103 6104 /* Validate maxrate requested is within permitted range */ 6105 if (maxrate && (maxrate > (ICE_SCHED_MAX_BW / 1000))) { 6106 netdev_err(netdev, "Invalid max rate %d specified for the queue %d\n", 6107 maxrate, queue_index); 6108 return -EINVAL; 6109 } 6110 6111 q_handle = vsi->tx_rings[queue_index]->q_handle; 6112 tc = ice_dcb_get_tc(vsi, queue_index); 6113 6114 vsi = ice_locate_vsi_using_queue(vsi, queue_index); 6115 if (!vsi) { 6116 netdev_err(netdev, "Invalid VSI for given queue %d\n", 6117 queue_index); 6118 return -EINVAL; 6119 } 6120 6121 /* Set BW back to default, when user set maxrate to 0 */ 6122 if (!maxrate) 6123 status = ice_cfg_q_bw_dflt_lmt(vsi->port_info, vsi->idx, tc, 6124 q_handle, ICE_MAX_BW); 6125 else 6126 status = ice_cfg_q_bw_lmt(vsi->port_info, vsi->idx, tc, 6127 q_handle, ICE_MAX_BW, maxrate * 1000); 6128 if (status) 6129 netdev_err(netdev, "Unable to set Tx max rate, error %d\n", 6130 status); 6131 6132 return status; 6133 } 6134 6135 /** 6136 * ice_fdb_add - add an entry to the hardware database 6137 * @ndm: the input from the stack 6138 * @tb: pointer to array of nladdr (unused) 6139 * @dev: the net device pointer 6140 * @addr: the MAC address entry being added 6141 * @vid: VLAN ID 6142 * @flags: instructions from stack about fdb operation 6143 * @extack: netlink extended ack 6144 */ 6145 static int 6146 ice_fdb_add(struct ndmsg *ndm, struct nlattr __always_unused *tb[], 6147 struct net_device *dev, const unsigned char *addr, u16 vid, 6148 u16 flags, struct netlink_ext_ack __always_unused *extack) 6149 { 6150 int err; 6151 6152 if (vid) { 6153 netdev_err(dev, "VLANs aren't supported yet for dev_uc|mc_add()\n"); 6154 return -EINVAL; 6155 } 6156 if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) { 6157 netdev_err(dev, "FDB only supports static addresses\n"); 6158 return -EINVAL; 6159 } 6160 6161 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr)) 6162 err = dev_uc_add_excl(dev, addr); 6163 else if (is_multicast_ether_addr(addr)) 6164 err = dev_mc_add_excl(dev, addr); 6165 else 6166 err = -EINVAL; 6167 6168 /* Only return duplicate errors if NLM_F_EXCL is set */ 6169 if (err == -EEXIST && !(flags & NLM_F_EXCL)) 6170 err = 0; 6171 6172 return err; 6173 } 6174 6175 /** 6176 * ice_fdb_del - delete an entry from the hardware database 6177 * @ndm: the input from the stack 6178 * @tb: pointer to array of nladdr (unused) 6179 * @dev: the net device pointer 6180 * @addr: the MAC address entry being added 6181 * @vid: VLAN ID 6182 * @extack: netlink extended ack 6183 */ 6184 static int 6185 ice_fdb_del(struct ndmsg *ndm, __always_unused struct nlattr *tb[], 6186 struct net_device *dev, const unsigned char *addr, 6187 __always_unused u16 vid, struct netlink_ext_ack *extack) 6188 { 6189 int err; 6190 6191 if (ndm->ndm_state & NUD_PERMANENT) { 6192 netdev_err(dev, "FDB only supports static addresses\n"); 6193 return -EINVAL; 6194 } 6195 6196 if (is_unicast_ether_addr(addr)) 6197 err = dev_uc_del(dev, addr); 6198 else if (is_multicast_ether_addr(addr)) 6199 err = dev_mc_del(dev, addr); 6200 else 6201 err = -EINVAL; 6202 6203 return err; 6204 } 6205 6206 #define NETIF_VLAN_OFFLOAD_FEATURES (NETIF_F_HW_VLAN_CTAG_RX | \ 6207 NETIF_F_HW_VLAN_CTAG_TX | \ 6208 NETIF_F_HW_VLAN_STAG_RX | \ 6209 NETIF_F_HW_VLAN_STAG_TX) 6210 6211 #define NETIF_VLAN_STRIPPING_FEATURES (NETIF_F_HW_VLAN_CTAG_RX | \ 6212 NETIF_F_HW_VLAN_STAG_RX) 6213 6214 #define NETIF_VLAN_FILTERING_FEATURES (NETIF_F_HW_VLAN_CTAG_FILTER | \ 6215 NETIF_F_HW_VLAN_STAG_FILTER) 6216 6217 /** 6218 * ice_fix_features - fix the netdev features flags based on device limitations 6219 * @netdev: ptr to the netdev that flags are being fixed on 6220 * @features: features that need to be checked and possibly fixed 6221 * 6222 * Make sure any fixups are made to features in this callback. This enables the 6223 * driver to not have to check unsupported configurations throughout the driver 6224 * because that's the responsiblity of this callback. 6225 * 6226 * Single VLAN Mode (SVM) Supported Features: 6227 * NETIF_F_HW_VLAN_CTAG_FILTER 6228 * NETIF_F_HW_VLAN_CTAG_RX 6229 * NETIF_F_HW_VLAN_CTAG_TX 6230 * 6231 * Double VLAN Mode (DVM) Supported Features: 6232 * NETIF_F_HW_VLAN_CTAG_FILTER 6233 * NETIF_F_HW_VLAN_CTAG_RX 6234 * NETIF_F_HW_VLAN_CTAG_TX 6235 * 6236 * NETIF_F_HW_VLAN_STAG_FILTER 6237 * NETIF_HW_VLAN_STAG_RX 6238 * NETIF_HW_VLAN_STAG_TX 6239 * 6240 * Features that need fixing: 6241 * Cannot simultaneously enable CTAG and STAG stripping and/or insertion. 6242 * These are mutually exlusive as the VSI context cannot support multiple 6243 * VLAN ethertypes simultaneously for stripping and/or insertion. If this 6244 * is not done, then default to clearing the requested STAG offload 6245 * settings. 6246 * 6247 * All supported filtering has to be enabled or disabled together. For 6248 * example, in DVM, CTAG and STAG filtering have to be enabled and disabled 6249 * together. If this is not done, then default to VLAN filtering disabled. 6250 * These are mutually exclusive as there is currently no way to 6251 * enable/disable VLAN filtering based on VLAN ethertype when using VLAN 6252 * prune rules. 6253 */ 6254 static netdev_features_t 6255 ice_fix_features(struct net_device *netdev, netdev_features_t features) 6256 { 6257 struct ice_netdev_priv *np = netdev_priv(netdev); 6258 netdev_features_t req_vlan_fltr, cur_vlan_fltr; 6259 bool cur_ctag, cur_stag, req_ctag, req_stag; 6260 6261 cur_vlan_fltr = netdev->features & NETIF_VLAN_FILTERING_FEATURES; 6262 cur_ctag = cur_vlan_fltr & NETIF_F_HW_VLAN_CTAG_FILTER; 6263 cur_stag = cur_vlan_fltr & NETIF_F_HW_VLAN_STAG_FILTER; 6264 6265 req_vlan_fltr = features & NETIF_VLAN_FILTERING_FEATURES; 6266 req_ctag = req_vlan_fltr & NETIF_F_HW_VLAN_CTAG_FILTER; 6267 req_stag = req_vlan_fltr & NETIF_F_HW_VLAN_STAG_FILTER; 6268 6269 if (req_vlan_fltr != cur_vlan_fltr) { 6270 if (ice_is_dvm_ena(&np->vsi->back->hw)) { 6271 if (req_ctag && req_stag) { 6272 features |= NETIF_VLAN_FILTERING_FEATURES; 6273 } else if (!req_ctag && !req_stag) { 6274 features &= ~NETIF_VLAN_FILTERING_FEATURES; 6275 } else if ((!cur_ctag && req_ctag && !cur_stag) || 6276 (!cur_stag && req_stag && !cur_ctag)) { 6277 features |= NETIF_VLAN_FILTERING_FEATURES; 6278 netdev_warn(netdev, "802.1Q and 802.1ad VLAN filtering must be either both on or both off. VLAN filtering has been enabled for both types.\n"); 6279 } else if ((cur_ctag && !req_ctag && cur_stag) || 6280 (cur_stag && !req_stag && cur_ctag)) { 6281 features &= ~NETIF_VLAN_FILTERING_FEATURES; 6282 netdev_warn(netdev, "802.1Q and 802.1ad VLAN filtering must be either both on or both off. VLAN filtering has been disabled for both types.\n"); 6283 } 6284 } else { 6285 if (req_vlan_fltr & NETIF_F_HW_VLAN_STAG_FILTER) 6286 netdev_warn(netdev, "cannot support requested 802.1ad filtering setting in SVM mode\n"); 6287 6288 if (req_vlan_fltr & NETIF_F_HW_VLAN_CTAG_FILTER) 6289 features |= NETIF_F_HW_VLAN_CTAG_FILTER; 6290 } 6291 } 6292 6293 if ((features & (NETIF_F_HW_VLAN_CTAG_RX | NETIF_F_HW_VLAN_CTAG_TX)) && 6294 (features & (NETIF_F_HW_VLAN_STAG_RX | NETIF_F_HW_VLAN_STAG_TX))) { 6295 netdev_warn(netdev, "cannot support CTAG and STAG VLAN stripping and/or insertion simultaneously since CTAG and STAG offloads are mutually exclusive, clearing STAG offload settings\n"); 6296 features &= ~(NETIF_F_HW_VLAN_STAG_RX | 6297 NETIF_F_HW_VLAN_STAG_TX); 6298 } 6299 6300 if (!(netdev->features & NETIF_F_RXFCS) && 6301 (features & NETIF_F_RXFCS) && 6302 (features & NETIF_VLAN_STRIPPING_FEATURES) && 6303 !ice_vsi_has_non_zero_vlans(np->vsi)) { 6304 netdev_warn(netdev, "Disabling VLAN stripping as FCS/CRC stripping is also disabled and there is no VLAN configured\n"); 6305 features &= ~NETIF_VLAN_STRIPPING_FEATURES; 6306 } 6307 6308 return features; 6309 } 6310 6311 /** 6312 * ice_set_rx_rings_vlan_proto - update rings with new stripped VLAN proto 6313 * @vsi: PF's VSI 6314 * @vlan_ethertype: VLAN ethertype (802.1Q or 802.1ad) in network byte order 6315 * 6316 * Store current stripped VLAN proto in ring packet context, 6317 * so it can be accessed more efficiently by packet processing code. 6318 */ 6319 static void 6320 ice_set_rx_rings_vlan_proto(struct ice_vsi *vsi, __be16 vlan_ethertype) 6321 { 6322 u16 i; 6323 6324 ice_for_each_alloc_rxq(vsi, i) 6325 vsi->rx_rings[i]->pkt_ctx.vlan_proto = vlan_ethertype; 6326 } 6327 6328 /** 6329 * ice_set_vlan_offload_features - set VLAN offload features for the PF VSI 6330 * @vsi: PF's VSI 6331 * @features: features used to determine VLAN offload settings 6332 * 6333 * First, determine the vlan_ethertype based on the VLAN offload bits in 6334 * features. Then determine if stripping and insertion should be enabled or 6335 * disabled. Finally enable or disable VLAN stripping and insertion. 6336 */ 6337 static int 6338 ice_set_vlan_offload_features(struct ice_vsi *vsi, netdev_features_t features) 6339 { 6340 bool enable_stripping = true, enable_insertion = true; 6341 struct ice_vsi_vlan_ops *vlan_ops; 6342 int strip_err = 0, insert_err = 0; 6343 u16 vlan_ethertype = 0; 6344 6345 vlan_ops = ice_get_compat_vsi_vlan_ops(vsi); 6346 6347 if (features & (NETIF_F_HW_VLAN_STAG_RX | NETIF_F_HW_VLAN_STAG_TX)) 6348 vlan_ethertype = ETH_P_8021AD; 6349 else if (features & (NETIF_F_HW_VLAN_CTAG_RX | NETIF_F_HW_VLAN_CTAG_TX)) 6350 vlan_ethertype = ETH_P_8021Q; 6351 6352 if (!(features & (NETIF_F_HW_VLAN_STAG_RX | NETIF_F_HW_VLAN_CTAG_RX))) 6353 enable_stripping = false; 6354 if (!(features & (NETIF_F_HW_VLAN_STAG_TX | NETIF_F_HW_VLAN_CTAG_TX))) 6355 enable_insertion = false; 6356 6357 if (enable_stripping) 6358 strip_err = vlan_ops->ena_stripping(vsi, vlan_ethertype); 6359 else 6360 strip_err = vlan_ops->dis_stripping(vsi); 6361 6362 if (enable_insertion) 6363 insert_err = vlan_ops->ena_insertion(vsi, vlan_ethertype); 6364 else 6365 insert_err = vlan_ops->dis_insertion(vsi); 6366 6367 if (strip_err || insert_err) 6368 return -EIO; 6369 6370 ice_set_rx_rings_vlan_proto(vsi, enable_stripping ? 6371 htons(vlan_ethertype) : 0); 6372 6373 return 0; 6374 } 6375 6376 /** 6377 * ice_set_vlan_filtering_features - set VLAN filtering features for the PF VSI 6378 * @vsi: PF's VSI 6379 * @features: features used to determine VLAN filtering settings 6380 * 6381 * Enable or disable Rx VLAN filtering based on the VLAN filtering bits in the 6382 * features. 6383 */ 6384 static int 6385 ice_set_vlan_filtering_features(struct ice_vsi *vsi, netdev_features_t features) 6386 { 6387 struct ice_vsi_vlan_ops *vlan_ops = ice_get_compat_vsi_vlan_ops(vsi); 6388 int err = 0; 6389 6390 /* support Single VLAN Mode (SVM) and Double VLAN Mode (DVM) by checking 6391 * if either bit is set 6392 */ 6393 if (features & 6394 (NETIF_F_HW_VLAN_CTAG_FILTER | NETIF_F_HW_VLAN_STAG_FILTER)) 6395 err = vlan_ops->ena_rx_filtering(vsi); 6396 else 6397 err = vlan_ops->dis_rx_filtering(vsi); 6398 6399 return err; 6400 } 6401 6402 /** 6403 * ice_set_vlan_features - set VLAN settings based on suggested feature set 6404 * @netdev: ptr to the netdev being adjusted 6405 * @features: the feature set that the stack is suggesting 6406 * 6407 * Only update VLAN settings if the requested_vlan_features are different than 6408 * the current_vlan_features. 6409 */ 6410 static int 6411 ice_set_vlan_features(struct net_device *netdev, netdev_features_t features) 6412 { 6413 netdev_features_t current_vlan_features, requested_vlan_features; 6414 struct ice_netdev_priv *np = netdev_priv(netdev); 6415 struct ice_vsi *vsi = np->vsi; 6416 int err; 6417 6418 current_vlan_features = netdev->features & NETIF_VLAN_OFFLOAD_FEATURES; 6419 requested_vlan_features = features & NETIF_VLAN_OFFLOAD_FEATURES; 6420 if (current_vlan_features ^ requested_vlan_features) { 6421 if ((features & NETIF_F_RXFCS) && 6422 (features & NETIF_VLAN_STRIPPING_FEATURES)) { 6423 dev_err(ice_pf_to_dev(vsi->back), 6424 "To enable VLAN stripping, you must first enable FCS/CRC stripping\n"); 6425 return -EIO; 6426 } 6427 6428 err = ice_set_vlan_offload_features(vsi, features); 6429 if (err) 6430 return err; 6431 } 6432 6433 current_vlan_features = netdev->features & 6434 NETIF_VLAN_FILTERING_FEATURES; 6435 requested_vlan_features = features & NETIF_VLAN_FILTERING_FEATURES; 6436 if (current_vlan_features ^ requested_vlan_features) { 6437 err = ice_set_vlan_filtering_features(vsi, features); 6438 if (err) 6439 return err; 6440 } 6441 6442 return 0; 6443 } 6444 6445 /** 6446 * ice_set_loopback - turn on/off loopback mode on underlying PF 6447 * @vsi: ptr to VSI 6448 * @ena: flag to indicate the on/off setting 6449 */ 6450 static int ice_set_loopback(struct ice_vsi *vsi, bool ena) 6451 { 6452 bool if_running = netif_running(vsi->netdev); 6453 int ret; 6454 6455 if (if_running && !test_and_set_bit(ICE_VSI_DOWN, vsi->state)) { 6456 ret = ice_down(vsi); 6457 if (ret) { 6458 netdev_err(vsi->netdev, "Preparing device to toggle loopback failed\n"); 6459 return ret; 6460 } 6461 } 6462 ret = ice_aq_set_mac_loopback(&vsi->back->hw, ena, NULL); 6463 if (ret) 6464 netdev_err(vsi->netdev, "Failed to toggle loopback state\n"); 6465 if (if_running) 6466 ret = ice_up(vsi); 6467 6468 return ret; 6469 } 6470 6471 /** 6472 * ice_set_features - set the netdev feature flags 6473 * @netdev: ptr to the netdev being adjusted 6474 * @features: the feature set that the stack is suggesting 6475 */ 6476 static int 6477 ice_set_features(struct net_device *netdev, netdev_features_t features) 6478 { 6479 netdev_features_t changed = netdev->features ^ features; 6480 struct ice_netdev_priv *np = netdev_priv(netdev); 6481 struct ice_vsi *vsi = np->vsi; 6482 struct ice_pf *pf = vsi->back; 6483 int ret = 0; 6484 6485 /* Don't set any netdev advanced features with device in Safe Mode */ 6486 if (ice_is_safe_mode(pf)) { 6487 dev_err(ice_pf_to_dev(pf), 6488 "Device is in Safe Mode - not enabling advanced netdev features\n"); 6489 return ret; 6490 } 6491 6492 /* Do not change setting during reset */ 6493 if (ice_is_reset_in_progress(pf->state)) { 6494 dev_err(ice_pf_to_dev(pf), 6495 "Device is resetting, changing advanced netdev features temporarily unavailable.\n"); 6496 return -EBUSY; 6497 } 6498 6499 /* Multiple features can be changed in one call so keep features in 6500 * separate if/else statements to guarantee each feature is checked 6501 */ 6502 if (changed & NETIF_F_RXHASH) 6503 ice_vsi_manage_rss_lut(vsi, !!(features & NETIF_F_RXHASH)); 6504 6505 ret = ice_set_vlan_features(netdev, features); 6506 if (ret) 6507 return ret; 6508 6509 /* Turn on receive of FCS aka CRC, and after setting this 6510 * flag the packet data will have the 4 byte CRC appended 6511 */ 6512 if (changed & NETIF_F_RXFCS) { 6513 if ((features & NETIF_F_RXFCS) && 6514 (features & NETIF_VLAN_STRIPPING_FEATURES)) { 6515 dev_err(ice_pf_to_dev(vsi->back), 6516 "To disable FCS/CRC stripping, you must first disable VLAN stripping\n"); 6517 return -EIO; 6518 } 6519 6520 ice_vsi_cfg_crc_strip(vsi, !!(features & NETIF_F_RXFCS)); 6521 ret = ice_down_up(vsi); 6522 if (ret) 6523 return ret; 6524 } 6525 6526 if (changed & NETIF_F_NTUPLE) { 6527 bool ena = !!(features & NETIF_F_NTUPLE); 6528 6529 ice_vsi_manage_fdir(vsi, ena); 6530 ena ? ice_init_arfs(vsi) : ice_clear_arfs(vsi); 6531 } 6532 6533 /* don't turn off hw_tc_offload when ADQ is already enabled */ 6534 if (!(features & NETIF_F_HW_TC) && ice_is_adq_active(pf)) { 6535 dev_err(ice_pf_to_dev(pf), "ADQ is active, can't turn hw_tc_offload off\n"); 6536 return -EACCES; 6537 } 6538 6539 if (changed & NETIF_F_HW_TC) { 6540 bool ena = !!(features & NETIF_F_HW_TC); 6541 6542 ena ? set_bit(ICE_FLAG_CLS_FLOWER, pf->flags) : 6543 clear_bit(ICE_FLAG_CLS_FLOWER, pf->flags); 6544 } 6545 6546 if (changed & NETIF_F_LOOPBACK) 6547 ret = ice_set_loopback(vsi, !!(features & NETIF_F_LOOPBACK)); 6548 6549 return ret; 6550 } 6551 6552 /** 6553 * ice_vsi_vlan_setup - Setup VLAN offload properties on a PF VSI 6554 * @vsi: VSI to setup VLAN properties for 6555 */ 6556 static int ice_vsi_vlan_setup(struct ice_vsi *vsi) 6557 { 6558 int err; 6559 6560 err = ice_set_vlan_offload_features(vsi, vsi->netdev->features); 6561 if (err) 6562 return err; 6563 6564 err = ice_set_vlan_filtering_features(vsi, vsi->netdev->features); 6565 if (err) 6566 return err; 6567 6568 return ice_vsi_add_vlan_zero(vsi); 6569 } 6570 6571 /** 6572 * ice_vsi_cfg_lan - Setup the VSI lan related config 6573 * @vsi: the VSI being configured 6574 * 6575 * Return 0 on success and negative value on error 6576 */ 6577 int ice_vsi_cfg_lan(struct ice_vsi *vsi) 6578 { 6579 int err; 6580 6581 if (vsi->netdev && vsi->type == ICE_VSI_PF) { 6582 ice_set_rx_mode(vsi->netdev); 6583 6584 err = ice_vsi_vlan_setup(vsi); 6585 if (err) 6586 return err; 6587 } 6588 ice_vsi_cfg_dcb_rings(vsi); 6589 6590 err = ice_vsi_cfg_lan_txqs(vsi); 6591 if (!err && ice_is_xdp_ena_vsi(vsi)) 6592 err = ice_vsi_cfg_xdp_txqs(vsi); 6593 if (!err) 6594 err = ice_vsi_cfg_rxqs(vsi); 6595 6596 return err; 6597 } 6598 6599 /* THEORY OF MODERATION: 6600 * The ice driver hardware works differently than the hardware that DIMLIB was 6601 * originally made for. ice hardware doesn't have packet count limits that 6602 * can trigger an interrupt, but it *does* have interrupt rate limit support, 6603 * which is hard-coded to a limit of 250,000 ints/second. 6604 * If not using dynamic moderation, the INTRL value can be modified 6605 * by ethtool rx-usecs-high. 6606 */ 6607 struct ice_dim { 6608 /* the throttle rate for interrupts, basically worst case delay before 6609 * an initial interrupt fires, value is stored in microseconds. 6610 */ 6611 u16 itr; 6612 }; 6613 6614 /* Make a different profile for Rx that doesn't allow quite so aggressive 6615 * moderation at the high end (it maxes out at 126us or about 8k interrupts a 6616 * second. 6617 */ 6618 static const struct ice_dim rx_profile[] = { 6619 {2}, /* 500,000 ints/s, capped at 250K by INTRL */ 6620 {8}, /* 125,000 ints/s */ 6621 {16}, /* 62,500 ints/s */ 6622 {62}, /* 16,129 ints/s */ 6623 {126} /* 7,936 ints/s */ 6624 }; 6625 6626 /* The transmit profile, which has the same sorts of values 6627 * as the previous struct 6628 */ 6629 static const struct ice_dim tx_profile[] = { 6630 {2}, /* 500,000 ints/s, capped at 250K by INTRL */ 6631 {8}, /* 125,000 ints/s */ 6632 {40}, /* 16,125 ints/s */ 6633 {128}, /* 7,812 ints/s */ 6634 {256} /* 3,906 ints/s */ 6635 }; 6636 6637 static void ice_tx_dim_work(struct work_struct *work) 6638 { 6639 struct ice_ring_container *rc; 6640 struct dim *dim; 6641 u16 itr; 6642 6643 dim = container_of(work, struct dim, work); 6644 rc = dim->priv; 6645 6646 WARN_ON(dim->profile_ix >= ARRAY_SIZE(tx_profile)); 6647 6648 /* look up the values in our local table */ 6649 itr = tx_profile[dim->profile_ix].itr; 6650 6651 ice_trace(tx_dim_work, container_of(rc, struct ice_q_vector, tx), dim); 6652 ice_write_itr(rc, itr); 6653 6654 dim->state = DIM_START_MEASURE; 6655 } 6656 6657 static void ice_rx_dim_work(struct work_struct *work) 6658 { 6659 struct ice_ring_container *rc; 6660 struct dim *dim; 6661 u16 itr; 6662 6663 dim = container_of(work, struct dim, work); 6664 rc = dim->priv; 6665 6666 WARN_ON(dim->profile_ix >= ARRAY_SIZE(rx_profile)); 6667 6668 /* look up the values in our local table */ 6669 itr = rx_profile[dim->profile_ix].itr; 6670 6671 ice_trace(rx_dim_work, container_of(rc, struct ice_q_vector, rx), dim); 6672 ice_write_itr(rc, itr); 6673 6674 dim->state = DIM_START_MEASURE; 6675 } 6676 6677 #define ICE_DIM_DEFAULT_PROFILE_IX 1 6678 6679 /** 6680 * ice_init_moderation - set up interrupt moderation 6681 * @q_vector: the vector containing rings to be configured 6682 * 6683 * Set up interrupt moderation registers, with the intent to do the right thing 6684 * when called from reset or from probe, and whether or not dynamic moderation 6685 * is enabled or not. Take special care to write all the registers in both 6686 * dynamic moderation mode or not in order to make sure hardware is in a known 6687 * state. 6688 */ 6689 static void ice_init_moderation(struct ice_q_vector *q_vector) 6690 { 6691 struct ice_ring_container *rc; 6692 bool tx_dynamic, rx_dynamic; 6693 6694 rc = &q_vector->tx; 6695 INIT_WORK(&rc->dim.work, ice_tx_dim_work); 6696 rc->dim.mode = DIM_CQ_PERIOD_MODE_START_FROM_EQE; 6697 rc->dim.profile_ix = ICE_DIM_DEFAULT_PROFILE_IX; 6698 rc->dim.priv = rc; 6699 tx_dynamic = ITR_IS_DYNAMIC(rc); 6700 6701 /* set the initial TX ITR to match the above */ 6702 ice_write_itr(rc, tx_dynamic ? 6703 tx_profile[rc->dim.profile_ix].itr : rc->itr_setting); 6704 6705 rc = &q_vector->rx; 6706 INIT_WORK(&rc->dim.work, ice_rx_dim_work); 6707 rc->dim.mode = DIM_CQ_PERIOD_MODE_START_FROM_EQE; 6708 rc->dim.profile_ix = ICE_DIM_DEFAULT_PROFILE_IX; 6709 rc->dim.priv = rc; 6710 rx_dynamic = ITR_IS_DYNAMIC(rc); 6711 6712 /* set the initial RX ITR to match the above */ 6713 ice_write_itr(rc, rx_dynamic ? rx_profile[rc->dim.profile_ix].itr : 6714 rc->itr_setting); 6715 6716 ice_set_q_vector_intrl(q_vector); 6717 } 6718 6719 /** 6720 * ice_napi_enable_all - Enable NAPI for all q_vectors in the VSI 6721 * @vsi: the VSI being configured 6722 */ 6723 static void ice_napi_enable_all(struct ice_vsi *vsi) 6724 { 6725 int q_idx; 6726 6727 if (!vsi->netdev) 6728 return; 6729 6730 ice_for_each_q_vector(vsi, q_idx) { 6731 struct ice_q_vector *q_vector = vsi->q_vectors[q_idx]; 6732 6733 ice_init_moderation(q_vector); 6734 6735 if (q_vector->rx.rx_ring || q_vector->tx.tx_ring) 6736 napi_enable(&q_vector->napi); 6737 } 6738 } 6739 6740 /** 6741 * ice_up_complete - Finish the last steps of bringing up a connection 6742 * @vsi: The VSI being configured 6743 * 6744 * Return 0 on success and negative value on error 6745 */ 6746 static int ice_up_complete(struct ice_vsi *vsi) 6747 { 6748 struct ice_pf *pf = vsi->back; 6749 int err; 6750 6751 ice_vsi_cfg_msix(vsi); 6752 6753 /* Enable only Rx rings, Tx rings were enabled by the FW when the 6754 * Tx queue group list was configured and the context bits were 6755 * programmed using ice_vsi_cfg_txqs 6756 */ 6757 err = ice_vsi_start_all_rx_rings(vsi); 6758 if (err) 6759 return err; 6760 6761 clear_bit(ICE_VSI_DOWN, vsi->state); 6762 ice_napi_enable_all(vsi); 6763 ice_vsi_ena_irq(vsi); 6764 6765 if (vsi->port_info && 6766 (vsi->port_info->phy.link_info.link_info & ICE_AQ_LINK_UP) && 6767 ((vsi->netdev && (vsi->type == ICE_VSI_PF || 6768 vsi->type == ICE_VSI_SF)))) { 6769 ice_print_link_msg(vsi, true); 6770 netif_tx_start_all_queues(vsi->netdev); 6771 netif_carrier_on(vsi->netdev); 6772 ice_ptp_link_change(pf, pf->hw.pf_id, true); 6773 } 6774 6775 /* Perform an initial read of the statistics registers now to 6776 * set the baseline so counters are ready when interface is up 6777 */ 6778 ice_update_eth_stats(vsi); 6779 6780 if (vsi->type == ICE_VSI_PF) 6781 ice_service_task_schedule(pf); 6782 6783 return 0; 6784 } 6785 6786 /** 6787 * ice_up - Bring the connection back up after being down 6788 * @vsi: VSI being configured 6789 */ 6790 int ice_up(struct ice_vsi *vsi) 6791 { 6792 int err; 6793 6794 err = ice_vsi_cfg_lan(vsi); 6795 if (!err) 6796 err = ice_up_complete(vsi); 6797 6798 return err; 6799 } 6800 6801 /** 6802 * ice_fetch_u64_stats_per_ring - get packets and bytes stats per ring 6803 * @syncp: pointer to u64_stats_sync 6804 * @stats: stats that pkts and bytes count will be taken from 6805 * @pkts: packets stats counter 6806 * @bytes: bytes stats counter 6807 * 6808 * This function fetches stats from the ring considering the atomic operations 6809 * that needs to be performed to read u64 values in 32 bit machine. 6810 */ 6811 void 6812 ice_fetch_u64_stats_per_ring(struct u64_stats_sync *syncp, 6813 struct ice_q_stats stats, u64 *pkts, u64 *bytes) 6814 { 6815 unsigned int start; 6816 6817 do { 6818 start = u64_stats_fetch_begin(syncp); 6819 *pkts = stats.pkts; 6820 *bytes = stats.bytes; 6821 } while (u64_stats_fetch_retry(syncp, start)); 6822 } 6823 6824 /** 6825 * ice_update_vsi_tx_ring_stats - Update VSI Tx ring stats counters 6826 * @vsi: the VSI to be updated 6827 * @vsi_stats: the stats struct to be updated 6828 * @rings: rings to work on 6829 * @count: number of rings 6830 */ 6831 static void 6832 ice_update_vsi_tx_ring_stats(struct ice_vsi *vsi, 6833 struct rtnl_link_stats64 *vsi_stats, 6834 struct ice_tx_ring **rings, u16 count) 6835 { 6836 u16 i; 6837 6838 for (i = 0; i < count; i++) { 6839 struct ice_tx_ring *ring; 6840 u64 pkts = 0, bytes = 0; 6841 6842 ring = READ_ONCE(rings[i]); 6843 if (!ring || !ring->ring_stats) 6844 continue; 6845 ice_fetch_u64_stats_per_ring(&ring->ring_stats->syncp, 6846 ring->ring_stats->stats, &pkts, 6847 &bytes); 6848 vsi_stats->tx_packets += pkts; 6849 vsi_stats->tx_bytes += bytes; 6850 vsi->tx_restart += ring->ring_stats->tx_stats.restart_q; 6851 vsi->tx_busy += ring->ring_stats->tx_stats.tx_busy; 6852 vsi->tx_linearize += ring->ring_stats->tx_stats.tx_linearize; 6853 } 6854 } 6855 6856 /** 6857 * ice_update_vsi_ring_stats - Update VSI stats counters 6858 * @vsi: the VSI to be updated 6859 */ 6860 static void ice_update_vsi_ring_stats(struct ice_vsi *vsi) 6861 { 6862 struct rtnl_link_stats64 *net_stats, *stats_prev; 6863 struct rtnl_link_stats64 *vsi_stats; 6864 struct ice_pf *pf = vsi->back; 6865 u64 pkts, bytes; 6866 int i; 6867 6868 vsi_stats = kzalloc(sizeof(*vsi_stats), GFP_ATOMIC); 6869 if (!vsi_stats) 6870 return; 6871 6872 /* reset non-netdev (extended) stats */ 6873 vsi->tx_restart = 0; 6874 vsi->tx_busy = 0; 6875 vsi->tx_linearize = 0; 6876 vsi->rx_buf_failed = 0; 6877 vsi->rx_page_failed = 0; 6878 6879 rcu_read_lock(); 6880 6881 /* update Tx rings counters */ 6882 ice_update_vsi_tx_ring_stats(vsi, vsi_stats, vsi->tx_rings, 6883 vsi->num_txq); 6884 6885 /* update Rx rings counters */ 6886 ice_for_each_rxq(vsi, i) { 6887 struct ice_rx_ring *ring = READ_ONCE(vsi->rx_rings[i]); 6888 struct ice_ring_stats *ring_stats; 6889 6890 ring_stats = ring->ring_stats; 6891 ice_fetch_u64_stats_per_ring(&ring_stats->syncp, 6892 ring_stats->stats, &pkts, 6893 &bytes); 6894 vsi_stats->rx_packets += pkts; 6895 vsi_stats->rx_bytes += bytes; 6896 vsi->rx_buf_failed += ring_stats->rx_stats.alloc_buf_failed; 6897 vsi->rx_page_failed += ring_stats->rx_stats.alloc_page_failed; 6898 } 6899 6900 /* update XDP Tx rings counters */ 6901 if (ice_is_xdp_ena_vsi(vsi)) 6902 ice_update_vsi_tx_ring_stats(vsi, vsi_stats, vsi->xdp_rings, 6903 vsi->num_xdp_txq); 6904 6905 rcu_read_unlock(); 6906 6907 net_stats = &vsi->net_stats; 6908 stats_prev = &vsi->net_stats_prev; 6909 6910 /* Update netdev counters, but keep in mind that values could start at 6911 * random value after PF reset. And as we increase the reported stat by 6912 * diff of Prev-Cur, we need to be sure that Prev is valid. If it's not, 6913 * let's skip this round. 6914 */ 6915 if (likely(pf->stat_prev_loaded)) { 6916 net_stats->tx_packets += vsi_stats->tx_packets - stats_prev->tx_packets; 6917 net_stats->tx_bytes += vsi_stats->tx_bytes - stats_prev->tx_bytes; 6918 net_stats->rx_packets += vsi_stats->rx_packets - stats_prev->rx_packets; 6919 net_stats->rx_bytes += vsi_stats->rx_bytes - stats_prev->rx_bytes; 6920 } 6921 6922 stats_prev->tx_packets = vsi_stats->tx_packets; 6923 stats_prev->tx_bytes = vsi_stats->tx_bytes; 6924 stats_prev->rx_packets = vsi_stats->rx_packets; 6925 stats_prev->rx_bytes = vsi_stats->rx_bytes; 6926 6927 kfree(vsi_stats); 6928 } 6929 6930 /** 6931 * ice_update_vsi_stats - Update VSI stats counters 6932 * @vsi: the VSI to be updated 6933 */ 6934 void ice_update_vsi_stats(struct ice_vsi *vsi) 6935 { 6936 struct rtnl_link_stats64 *cur_ns = &vsi->net_stats; 6937 struct ice_eth_stats *cur_es = &vsi->eth_stats; 6938 struct ice_pf *pf = vsi->back; 6939 6940 if (test_bit(ICE_VSI_DOWN, vsi->state) || 6941 test_bit(ICE_CFG_BUSY, pf->state)) 6942 return; 6943 6944 /* get stats as recorded by Tx/Rx rings */ 6945 ice_update_vsi_ring_stats(vsi); 6946 6947 /* get VSI stats as recorded by the hardware */ 6948 ice_update_eth_stats(vsi); 6949 6950 cur_ns->tx_errors = cur_es->tx_errors; 6951 cur_ns->rx_dropped = cur_es->rx_discards; 6952 cur_ns->tx_dropped = cur_es->tx_discards; 6953 cur_ns->multicast = cur_es->rx_multicast; 6954 6955 /* update some more netdev stats if this is main VSI */ 6956 if (vsi->type == ICE_VSI_PF) { 6957 cur_ns->rx_crc_errors = pf->stats.crc_errors; 6958 cur_ns->rx_errors = pf->stats.crc_errors + 6959 pf->stats.illegal_bytes + 6960 pf->stats.rx_undersize + 6961 pf->hw_csum_rx_error + 6962 pf->stats.rx_jabber + 6963 pf->stats.rx_fragments + 6964 pf->stats.rx_oversize; 6965 /* record drops from the port level */ 6966 cur_ns->rx_missed_errors = pf->stats.eth.rx_discards; 6967 } 6968 } 6969 6970 /** 6971 * ice_update_pf_stats - Update PF port stats counters 6972 * @pf: PF whose stats needs to be updated 6973 */ 6974 void ice_update_pf_stats(struct ice_pf *pf) 6975 { 6976 struct ice_hw_port_stats *prev_ps, *cur_ps; 6977 struct ice_hw *hw = &pf->hw; 6978 u16 fd_ctr_base; 6979 u8 port; 6980 6981 port = hw->port_info->lport; 6982 prev_ps = &pf->stats_prev; 6983 cur_ps = &pf->stats; 6984 6985 if (ice_is_reset_in_progress(pf->state)) 6986 pf->stat_prev_loaded = false; 6987 6988 ice_stat_update40(hw, GLPRT_GORCL(port), pf->stat_prev_loaded, 6989 &prev_ps->eth.rx_bytes, 6990 &cur_ps->eth.rx_bytes); 6991 6992 ice_stat_update40(hw, GLPRT_UPRCL(port), pf->stat_prev_loaded, 6993 &prev_ps->eth.rx_unicast, 6994 &cur_ps->eth.rx_unicast); 6995 6996 ice_stat_update40(hw, GLPRT_MPRCL(port), pf->stat_prev_loaded, 6997 &prev_ps->eth.rx_multicast, 6998 &cur_ps->eth.rx_multicast); 6999 7000 ice_stat_update40(hw, GLPRT_BPRCL(port), pf->stat_prev_loaded, 7001 &prev_ps->eth.rx_broadcast, 7002 &cur_ps->eth.rx_broadcast); 7003 7004 ice_stat_update32(hw, PRTRPB_RDPC, pf->stat_prev_loaded, 7005 &prev_ps->eth.rx_discards, 7006 &cur_ps->eth.rx_discards); 7007 7008 ice_stat_update40(hw, GLPRT_GOTCL(port), pf->stat_prev_loaded, 7009 &prev_ps->eth.tx_bytes, 7010 &cur_ps->eth.tx_bytes); 7011 7012 ice_stat_update40(hw, GLPRT_UPTCL(port), pf->stat_prev_loaded, 7013 &prev_ps->eth.tx_unicast, 7014 &cur_ps->eth.tx_unicast); 7015 7016 ice_stat_update40(hw, GLPRT_MPTCL(port), pf->stat_prev_loaded, 7017 &prev_ps->eth.tx_multicast, 7018 &cur_ps->eth.tx_multicast); 7019 7020 ice_stat_update40(hw, GLPRT_BPTCL(port), pf->stat_prev_loaded, 7021 &prev_ps->eth.tx_broadcast, 7022 &cur_ps->eth.tx_broadcast); 7023 7024 ice_stat_update32(hw, GLPRT_TDOLD(port), pf->stat_prev_loaded, 7025 &prev_ps->tx_dropped_link_down, 7026 &cur_ps->tx_dropped_link_down); 7027 7028 ice_stat_update40(hw, GLPRT_PRC64L(port), pf->stat_prev_loaded, 7029 &prev_ps->rx_size_64, &cur_ps->rx_size_64); 7030 7031 ice_stat_update40(hw, GLPRT_PRC127L(port), pf->stat_prev_loaded, 7032 &prev_ps->rx_size_127, &cur_ps->rx_size_127); 7033 7034 ice_stat_update40(hw, GLPRT_PRC255L(port), pf->stat_prev_loaded, 7035 &prev_ps->rx_size_255, &cur_ps->rx_size_255); 7036 7037 ice_stat_update40(hw, GLPRT_PRC511L(port), pf->stat_prev_loaded, 7038 &prev_ps->rx_size_511, &cur_ps->rx_size_511); 7039 7040 ice_stat_update40(hw, GLPRT_PRC1023L(port), pf->stat_prev_loaded, 7041 &prev_ps->rx_size_1023, &cur_ps->rx_size_1023); 7042 7043 ice_stat_update40(hw, GLPRT_PRC1522L(port), pf->stat_prev_loaded, 7044 &prev_ps->rx_size_1522, &cur_ps->rx_size_1522); 7045 7046 ice_stat_update40(hw, GLPRT_PRC9522L(port), pf->stat_prev_loaded, 7047 &prev_ps->rx_size_big, &cur_ps->rx_size_big); 7048 7049 ice_stat_update40(hw, GLPRT_PTC64L(port), pf->stat_prev_loaded, 7050 &prev_ps->tx_size_64, &cur_ps->tx_size_64); 7051 7052 ice_stat_update40(hw, GLPRT_PTC127L(port), pf->stat_prev_loaded, 7053 &prev_ps->tx_size_127, &cur_ps->tx_size_127); 7054 7055 ice_stat_update40(hw, GLPRT_PTC255L(port), pf->stat_prev_loaded, 7056 &prev_ps->tx_size_255, &cur_ps->tx_size_255); 7057 7058 ice_stat_update40(hw, GLPRT_PTC511L(port), pf->stat_prev_loaded, 7059 &prev_ps->tx_size_511, &cur_ps->tx_size_511); 7060 7061 ice_stat_update40(hw, GLPRT_PTC1023L(port), pf->stat_prev_loaded, 7062 &prev_ps->tx_size_1023, &cur_ps->tx_size_1023); 7063 7064 ice_stat_update40(hw, GLPRT_PTC1522L(port), pf->stat_prev_loaded, 7065 &prev_ps->tx_size_1522, &cur_ps->tx_size_1522); 7066 7067 ice_stat_update40(hw, GLPRT_PTC9522L(port), pf->stat_prev_loaded, 7068 &prev_ps->tx_size_big, &cur_ps->tx_size_big); 7069 7070 fd_ctr_base = hw->fd_ctr_base; 7071 7072 ice_stat_update40(hw, 7073 GLSTAT_FD_CNT0L(ICE_FD_SB_STAT_IDX(fd_ctr_base)), 7074 pf->stat_prev_loaded, &prev_ps->fd_sb_match, 7075 &cur_ps->fd_sb_match); 7076 ice_stat_update32(hw, GLPRT_LXONRXC(port), pf->stat_prev_loaded, 7077 &prev_ps->link_xon_rx, &cur_ps->link_xon_rx); 7078 7079 ice_stat_update32(hw, GLPRT_LXOFFRXC(port), pf->stat_prev_loaded, 7080 &prev_ps->link_xoff_rx, &cur_ps->link_xoff_rx); 7081 7082 ice_stat_update32(hw, GLPRT_LXONTXC(port), pf->stat_prev_loaded, 7083 &prev_ps->link_xon_tx, &cur_ps->link_xon_tx); 7084 7085 ice_stat_update32(hw, GLPRT_LXOFFTXC(port), pf->stat_prev_loaded, 7086 &prev_ps->link_xoff_tx, &cur_ps->link_xoff_tx); 7087 7088 ice_update_dcb_stats(pf); 7089 7090 ice_stat_update32(hw, GLPRT_CRCERRS(port), pf->stat_prev_loaded, 7091 &prev_ps->crc_errors, &cur_ps->crc_errors); 7092 7093 ice_stat_update32(hw, GLPRT_ILLERRC(port), pf->stat_prev_loaded, 7094 &prev_ps->illegal_bytes, &cur_ps->illegal_bytes); 7095 7096 ice_stat_update32(hw, GLPRT_MLFC(port), pf->stat_prev_loaded, 7097 &prev_ps->mac_local_faults, 7098 &cur_ps->mac_local_faults); 7099 7100 ice_stat_update32(hw, GLPRT_MRFC(port), pf->stat_prev_loaded, 7101 &prev_ps->mac_remote_faults, 7102 &cur_ps->mac_remote_faults); 7103 7104 ice_stat_update32(hw, GLPRT_RUC(port), pf->stat_prev_loaded, 7105 &prev_ps->rx_undersize, &cur_ps->rx_undersize); 7106 7107 ice_stat_update32(hw, GLPRT_RFC(port), pf->stat_prev_loaded, 7108 &prev_ps->rx_fragments, &cur_ps->rx_fragments); 7109 7110 ice_stat_update32(hw, GLPRT_ROC(port), pf->stat_prev_loaded, 7111 &prev_ps->rx_oversize, &cur_ps->rx_oversize); 7112 7113 ice_stat_update32(hw, GLPRT_RJC(port), pf->stat_prev_loaded, 7114 &prev_ps->rx_jabber, &cur_ps->rx_jabber); 7115 7116 cur_ps->fd_sb_status = test_bit(ICE_FLAG_FD_ENA, pf->flags) ? 1 : 0; 7117 7118 pf->stat_prev_loaded = true; 7119 } 7120 7121 /** 7122 * ice_get_stats64 - get statistics for network device structure 7123 * @netdev: network interface device structure 7124 * @stats: main device statistics structure 7125 */ 7126 void ice_get_stats64(struct net_device *netdev, struct rtnl_link_stats64 *stats) 7127 { 7128 struct ice_netdev_priv *np = netdev_priv(netdev); 7129 struct rtnl_link_stats64 *vsi_stats; 7130 struct ice_vsi *vsi = np->vsi; 7131 7132 vsi_stats = &vsi->net_stats; 7133 7134 if (!vsi->num_txq || !vsi->num_rxq) 7135 return; 7136 7137 /* netdev packet/byte stats come from ring counter. These are obtained 7138 * by summing up ring counters (done by ice_update_vsi_ring_stats). 7139 * But, only call the update routine and read the registers if VSI is 7140 * not down. 7141 */ 7142 if (!test_bit(ICE_VSI_DOWN, vsi->state)) 7143 ice_update_vsi_ring_stats(vsi); 7144 stats->tx_packets = vsi_stats->tx_packets; 7145 stats->tx_bytes = vsi_stats->tx_bytes; 7146 stats->rx_packets = vsi_stats->rx_packets; 7147 stats->rx_bytes = vsi_stats->rx_bytes; 7148 7149 /* The rest of the stats can be read from the hardware but instead we 7150 * just return values that the watchdog task has already obtained from 7151 * the hardware. 7152 */ 7153 stats->multicast = vsi_stats->multicast; 7154 stats->tx_errors = vsi_stats->tx_errors; 7155 stats->tx_dropped = vsi_stats->tx_dropped; 7156 stats->rx_errors = vsi_stats->rx_errors; 7157 stats->rx_dropped = vsi_stats->rx_dropped; 7158 stats->rx_crc_errors = vsi_stats->rx_crc_errors; 7159 stats->rx_length_errors = vsi_stats->rx_length_errors; 7160 } 7161 7162 /** 7163 * ice_napi_disable_all - Disable NAPI for all q_vectors in the VSI 7164 * @vsi: VSI having NAPI disabled 7165 */ 7166 static void ice_napi_disable_all(struct ice_vsi *vsi) 7167 { 7168 int q_idx; 7169 7170 if (!vsi->netdev) 7171 return; 7172 7173 ice_for_each_q_vector(vsi, q_idx) { 7174 struct ice_q_vector *q_vector = vsi->q_vectors[q_idx]; 7175 7176 if (q_vector->rx.rx_ring || q_vector->tx.tx_ring) 7177 napi_disable(&q_vector->napi); 7178 7179 cancel_work_sync(&q_vector->tx.dim.work); 7180 cancel_work_sync(&q_vector->rx.dim.work); 7181 } 7182 } 7183 7184 /** 7185 * ice_vsi_dis_irq - Mask off queue interrupt generation on the VSI 7186 * @vsi: the VSI being un-configured 7187 */ 7188 static void ice_vsi_dis_irq(struct ice_vsi *vsi) 7189 { 7190 struct ice_pf *pf = vsi->back; 7191 struct ice_hw *hw = &pf->hw; 7192 u32 val; 7193 int i; 7194 7195 /* disable interrupt causation from each Rx queue; Tx queues are 7196 * handled in ice_vsi_stop_tx_ring() 7197 */ 7198 if (vsi->rx_rings) { 7199 ice_for_each_rxq(vsi, i) { 7200 if (vsi->rx_rings[i]) { 7201 u16 reg; 7202 7203 reg = vsi->rx_rings[i]->reg_idx; 7204 val = rd32(hw, QINT_RQCTL(reg)); 7205 val &= ~QINT_RQCTL_CAUSE_ENA_M; 7206 wr32(hw, QINT_RQCTL(reg), val); 7207 } 7208 } 7209 } 7210 7211 /* disable each interrupt */ 7212 ice_for_each_q_vector(vsi, i) { 7213 if (!vsi->q_vectors[i]) 7214 continue; 7215 wr32(hw, GLINT_DYN_CTL(vsi->q_vectors[i]->reg_idx), 0); 7216 } 7217 7218 ice_flush(hw); 7219 7220 /* don't call synchronize_irq() for VF's from the host */ 7221 if (vsi->type == ICE_VSI_VF) 7222 return; 7223 7224 ice_for_each_q_vector(vsi, i) 7225 synchronize_irq(vsi->q_vectors[i]->irq.virq); 7226 } 7227 7228 /** 7229 * ice_down - Shutdown the connection 7230 * @vsi: The VSI being stopped 7231 * 7232 * Caller of this function is expected to set the vsi->state ICE_DOWN bit 7233 */ 7234 int ice_down(struct ice_vsi *vsi) 7235 { 7236 int i, tx_err, rx_err, vlan_err = 0; 7237 7238 WARN_ON(!test_bit(ICE_VSI_DOWN, vsi->state)); 7239 7240 if (vsi->netdev) { 7241 vlan_err = ice_vsi_del_vlan_zero(vsi); 7242 ice_ptp_link_change(vsi->back, vsi->back->hw.pf_id, false); 7243 netif_carrier_off(vsi->netdev); 7244 netif_tx_disable(vsi->netdev); 7245 } 7246 7247 ice_vsi_dis_irq(vsi); 7248 7249 tx_err = ice_vsi_stop_lan_tx_rings(vsi, ICE_NO_RESET, 0); 7250 if (tx_err) 7251 netdev_err(vsi->netdev, "Failed stop Tx rings, VSI %d error %d\n", 7252 vsi->vsi_num, tx_err); 7253 if (!tx_err && vsi->xdp_rings) { 7254 tx_err = ice_vsi_stop_xdp_tx_rings(vsi); 7255 if (tx_err) 7256 netdev_err(vsi->netdev, "Failed stop XDP rings, VSI %d error %d\n", 7257 vsi->vsi_num, tx_err); 7258 } 7259 7260 rx_err = ice_vsi_stop_all_rx_rings(vsi); 7261 if (rx_err) 7262 netdev_err(vsi->netdev, "Failed stop Rx rings, VSI %d error %d\n", 7263 vsi->vsi_num, rx_err); 7264 7265 ice_napi_disable_all(vsi); 7266 7267 ice_for_each_txq(vsi, i) 7268 ice_clean_tx_ring(vsi->tx_rings[i]); 7269 7270 if (vsi->xdp_rings) 7271 ice_for_each_xdp_txq(vsi, i) 7272 ice_clean_tx_ring(vsi->xdp_rings[i]); 7273 7274 ice_for_each_rxq(vsi, i) 7275 ice_clean_rx_ring(vsi->rx_rings[i]); 7276 7277 if (tx_err || rx_err || vlan_err) { 7278 netdev_err(vsi->netdev, "Failed to close VSI 0x%04X on switch 0x%04X\n", 7279 vsi->vsi_num, vsi->vsw->sw_id); 7280 return -EIO; 7281 } 7282 7283 return 0; 7284 } 7285 7286 /** 7287 * ice_down_up - shutdown the VSI connection and bring it up 7288 * @vsi: the VSI to be reconnected 7289 */ 7290 int ice_down_up(struct ice_vsi *vsi) 7291 { 7292 int ret; 7293 7294 /* if DOWN already set, nothing to do */ 7295 if (test_and_set_bit(ICE_VSI_DOWN, vsi->state)) 7296 return 0; 7297 7298 ret = ice_down(vsi); 7299 if (ret) 7300 return ret; 7301 7302 ret = ice_up(vsi); 7303 if (ret) { 7304 netdev_err(vsi->netdev, "reallocating resources failed during netdev features change, may need to reload driver\n"); 7305 return ret; 7306 } 7307 7308 return 0; 7309 } 7310 7311 /** 7312 * ice_vsi_setup_tx_rings - Allocate VSI Tx queue resources 7313 * @vsi: VSI having resources allocated 7314 * 7315 * Return 0 on success, negative on failure 7316 */ 7317 int ice_vsi_setup_tx_rings(struct ice_vsi *vsi) 7318 { 7319 int i, err = 0; 7320 7321 if (!vsi->num_txq) { 7322 dev_err(ice_pf_to_dev(vsi->back), "VSI %d has 0 Tx queues\n", 7323 vsi->vsi_num); 7324 return -EINVAL; 7325 } 7326 7327 ice_for_each_txq(vsi, i) { 7328 struct ice_tx_ring *ring = vsi->tx_rings[i]; 7329 7330 if (!ring) 7331 return -EINVAL; 7332 7333 if (vsi->netdev) 7334 ring->netdev = vsi->netdev; 7335 err = ice_setup_tx_ring(ring); 7336 if (err) 7337 break; 7338 } 7339 7340 return err; 7341 } 7342 7343 /** 7344 * ice_vsi_setup_rx_rings - Allocate VSI Rx queue resources 7345 * @vsi: VSI having resources allocated 7346 * 7347 * Return 0 on success, negative on failure 7348 */ 7349 int ice_vsi_setup_rx_rings(struct ice_vsi *vsi) 7350 { 7351 int i, err = 0; 7352 7353 if (!vsi->num_rxq) { 7354 dev_err(ice_pf_to_dev(vsi->back), "VSI %d has 0 Rx queues\n", 7355 vsi->vsi_num); 7356 return -EINVAL; 7357 } 7358 7359 ice_for_each_rxq(vsi, i) { 7360 struct ice_rx_ring *ring = vsi->rx_rings[i]; 7361 7362 if (!ring) 7363 return -EINVAL; 7364 7365 if (vsi->netdev) 7366 ring->netdev = vsi->netdev; 7367 err = ice_setup_rx_ring(ring); 7368 if (err) 7369 break; 7370 } 7371 7372 return err; 7373 } 7374 7375 /** 7376 * ice_vsi_open_ctrl - open control VSI for use 7377 * @vsi: the VSI to open 7378 * 7379 * Initialization of the Control VSI 7380 * 7381 * Returns 0 on success, negative value on error 7382 */ 7383 int ice_vsi_open_ctrl(struct ice_vsi *vsi) 7384 { 7385 char int_name[ICE_INT_NAME_STR_LEN]; 7386 struct ice_pf *pf = vsi->back; 7387 struct device *dev; 7388 int err; 7389 7390 dev = ice_pf_to_dev(pf); 7391 /* allocate descriptors */ 7392 err = ice_vsi_setup_tx_rings(vsi); 7393 if (err) 7394 goto err_setup_tx; 7395 7396 err = ice_vsi_setup_rx_rings(vsi); 7397 if (err) 7398 goto err_setup_rx; 7399 7400 err = ice_vsi_cfg_lan(vsi); 7401 if (err) 7402 goto err_setup_rx; 7403 7404 snprintf(int_name, sizeof(int_name) - 1, "%s-%s:ctrl", 7405 dev_driver_string(dev), dev_name(dev)); 7406 err = ice_vsi_req_irq_msix(vsi, int_name); 7407 if (err) 7408 goto err_setup_rx; 7409 7410 ice_vsi_cfg_msix(vsi); 7411 7412 err = ice_vsi_start_all_rx_rings(vsi); 7413 if (err) 7414 goto err_up_complete; 7415 7416 clear_bit(ICE_VSI_DOWN, vsi->state); 7417 ice_vsi_ena_irq(vsi); 7418 7419 return 0; 7420 7421 err_up_complete: 7422 ice_down(vsi); 7423 err_setup_rx: 7424 ice_vsi_free_rx_rings(vsi); 7425 err_setup_tx: 7426 ice_vsi_free_tx_rings(vsi); 7427 7428 return err; 7429 } 7430 7431 /** 7432 * ice_vsi_open - Called when a network interface is made active 7433 * @vsi: the VSI to open 7434 * 7435 * Initialization of the VSI 7436 * 7437 * Returns 0 on success, negative value on error 7438 */ 7439 int ice_vsi_open(struct ice_vsi *vsi) 7440 { 7441 char int_name[ICE_INT_NAME_STR_LEN]; 7442 struct ice_pf *pf = vsi->back; 7443 int err; 7444 7445 /* allocate descriptors */ 7446 err = ice_vsi_setup_tx_rings(vsi); 7447 if (err) 7448 goto err_setup_tx; 7449 7450 err = ice_vsi_setup_rx_rings(vsi); 7451 if (err) 7452 goto err_setup_rx; 7453 7454 err = ice_vsi_cfg_lan(vsi); 7455 if (err) 7456 goto err_setup_rx; 7457 7458 snprintf(int_name, sizeof(int_name) - 1, "%s-%s", 7459 dev_driver_string(ice_pf_to_dev(pf)), vsi->netdev->name); 7460 err = ice_vsi_req_irq_msix(vsi, int_name); 7461 if (err) 7462 goto err_setup_rx; 7463 7464 ice_vsi_cfg_netdev_tc(vsi, vsi->tc_cfg.ena_tc); 7465 7466 if (vsi->type == ICE_VSI_PF || vsi->type == ICE_VSI_SF) { 7467 /* Notify the stack of the actual queue counts. */ 7468 err = netif_set_real_num_tx_queues(vsi->netdev, vsi->num_txq); 7469 if (err) 7470 goto err_set_qs; 7471 7472 err = netif_set_real_num_rx_queues(vsi->netdev, vsi->num_rxq); 7473 if (err) 7474 goto err_set_qs; 7475 7476 ice_vsi_set_napi_queues(vsi); 7477 } 7478 7479 err = ice_up_complete(vsi); 7480 if (err) 7481 goto err_up_complete; 7482 7483 return 0; 7484 7485 err_up_complete: 7486 ice_down(vsi); 7487 err_set_qs: 7488 ice_vsi_free_irq(vsi); 7489 err_setup_rx: 7490 ice_vsi_free_rx_rings(vsi); 7491 err_setup_tx: 7492 ice_vsi_free_tx_rings(vsi); 7493 7494 return err; 7495 } 7496 7497 /** 7498 * ice_vsi_release_all - Delete all VSIs 7499 * @pf: PF from which all VSIs are being removed 7500 */ 7501 static void ice_vsi_release_all(struct ice_pf *pf) 7502 { 7503 int err, i; 7504 7505 if (!pf->vsi) 7506 return; 7507 7508 ice_for_each_vsi(pf, i) { 7509 if (!pf->vsi[i]) 7510 continue; 7511 7512 if (pf->vsi[i]->type == ICE_VSI_CHNL) 7513 continue; 7514 7515 err = ice_vsi_release(pf->vsi[i]); 7516 if (err) 7517 dev_dbg(ice_pf_to_dev(pf), "Failed to release pf->vsi[%d], err %d, vsi_num = %d\n", 7518 i, err, pf->vsi[i]->vsi_num); 7519 } 7520 } 7521 7522 /** 7523 * ice_vsi_rebuild_by_type - Rebuild VSI of a given type 7524 * @pf: pointer to the PF instance 7525 * @type: VSI type to rebuild 7526 * 7527 * Iterates through the pf->vsi array and rebuilds VSIs of the requested type 7528 */ 7529 static int ice_vsi_rebuild_by_type(struct ice_pf *pf, enum ice_vsi_type type) 7530 { 7531 struct device *dev = ice_pf_to_dev(pf); 7532 int i, err; 7533 7534 ice_for_each_vsi(pf, i) { 7535 struct ice_vsi *vsi = pf->vsi[i]; 7536 7537 if (!vsi || vsi->type != type) 7538 continue; 7539 7540 /* rebuild the VSI */ 7541 err = ice_vsi_rebuild(vsi, ICE_VSI_FLAG_INIT); 7542 if (err) { 7543 dev_err(dev, "rebuild VSI failed, err %d, VSI index %d, type %s\n", 7544 err, vsi->idx, ice_vsi_type_str(type)); 7545 return err; 7546 } 7547 7548 /* replay filters for the VSI */ 7549 err = ice_replay_vsi(&pf->hw, vsi->idx); 7550 if (err) { 7551 dev_err(dev, "replay VSI failed, error %d, VSI index %d, type %s\n", 7552 err, vsi->idx, ice_vsi_type_str(type)); 7553 return err; 7554 } 7555 7556 /* Re-map HW VSI number, using VSI handle that has been 7557 * previously validated in ice_replay_vsi() call above 7558 */ 7559 vsi->vsi_num = ice_get_hw_vsi_num(&pf->hw, vsi->idx); 7560 7561 /* enable the VSI */ 7562 err = ice_ena_vsi(vsi, false); 7563 if (err) { 7564 dev_err(dev, "enable VSI failed, err %d, VSI index %d, type %s\n", 7565 err, vsi->idx, ice_vsi_type_str(type)); 7566 return err; 7567 } 7568 7569 dev_info(dev, "VSI rebuilt. VSI index %d, type %s\n", vsi->idx, 7570 ice_vsi_type_str(type)); 7571 } 7572 7573 return 0; 7574 } 7575 7576 /** 7577 * ice_update_pf_netdev_link - Update PF netdev link status 7578 * @pf: pointer to the PF instance 7579 */ 7580 static void ice_update_pf_netdev_link(struct ice_pf *pf) 7581 { 7582 bool link_up; 7583 int i; 7584 7585 ice_for_each_vsi(pf, i) { 7586 struct ice_vsi *vsi = pf->vsi[i]; 7587 7588 if (!vsi || vsi->type != ICE_VSI_PF) 7589 return; 7590 7591 ice_get_link_status(pf->vsi[i]->port_info, &link_up); 7592 if (link_up) { 7593 netif_carrier_on(pf->vsi[i]->netdev); 7594 netif_tx_wake_all_queues(pf->vsi[i]->netdev); 7595 } else { 7596 netif_carrier_off(pf->vsi[i]->netdev); 7597 netif_tx_stop_all_queues(pf->vsi[i]->netdev); 7598 } 7599 } 7600 } 7601 7602 /** 7603 * ice_rebuild - rebuild after reset 7604 * @pf: PF to rebuild 7605 * @reset_type: type of reset 7606 * 7607 * Do not rebuild VF VSI in this flow because that is already handled via 7608 * ice_reset_all_vfs(). This is because requirements for resetting a VF after a 7609 * PFR/CORER/GLOBER/etc. are different than the normal flow. Also, we don't want 7610 * to reset/rebuild all the VF VSI twice. 7611 */ 7612 static void ice_rebuild(struct ice_pf *pf, enum ice_reset_req reset_type) 7613 { 7614 struct ice_vsi *vsi = ice_get_main_vsi(pf); 7615 struct device *dev = ice_pf_to_dev(pf); 7616 struct ice_hw *hw = &pf->hw; 7617 bool dvm; 7618 int err; 7619 7620 if (test_bit(ICE_DOWN, pf->state)) 7621 goto clear_recovery; 7622 7623 dev_dbg(dev, "rebuilding PF after reset_type=%d\n", reset_type); 7624 7625 #define ICE_EMP_RESET_SLEEP_MS 5000 7626 if (reset_type == ICE_RESET_EMPR) { 7627 /* If an EMP reset has occurred, any previously pending flash 7628 * update will have completed. We no longer know whether or 7629 * not the NVM update EMP reset is restricted. 7630 */ 7631 pf->fw_emp_reset_disabled = false; 7632 7633 msleep(ICE_EMP_RESET_SLEEP_MS); 7634 } 7635 7636 err = ice_init_all_ctrlq(hw); 7637 if (err) { 7638 dev_err(dev, "control queues init failed %d\n", err); 7639 goto err_init_ctrlq; 7640 } 7641 7642 /* if DDP was previously loaded successfully */ 7643 if (!ice_is_safe_mode(pf)) { 7644 /* reload the SW DB of filter tables */ 7645 if (reset_type == ICE_RESET_PFR) 7646 ice_fill_blk_tbls(hw); 7647 else 7648 /* Reload DDP Package after CORER/GLOBR reset */ 7649 ice_load_pkg(NULL, pf); 7650 } 7651 7652 err = ice_clear_pf_cfg(hw); 7653 if (err) { 7654 dev_err(dev, "clear PF configuration failed %d\n", err); 7655 goto err_init_ctrlq; 7656 } 7657 7658 ice_clear_pxe_mode(hw); 7659 7660 err = ice_init_nvm(hw); 7661 if (err) { 7662 dev_err(dev, "ice_init_nvm failed %d\n", err); 7663 goto err_init_ctrlq; 7664 } 7665 7666 err = ice_get_caps(hw); 7667 if (err) { 7668 dev_err(dev, "ice_get_caps failed %d\n", err); 7669 goto err_init_ctrlq; 7670 } 7671 7672 err = ice_aq_set_mac_cfg(hw, ICE_AQ_SET_MAC_FRAME_SIZE_MAX, NULL); 7673 if (err) { 7674 dev_err(dev, "set_mac_cfg failed %d\n", err); 7675 goto err_init_ctrlq; 7676 } 7677 7678 dvm = ice_is_dvm_ena(hw); 7679 7680 err = ice_aq_set_port_params(pf->hw.port_info, dvm, NULL); 7681 if (err) 7682 goto err_init_ctrlq; 7683 7684 err = ice_sched_init_port(hw->port_info); 7685 if (err) 7686 goto err_sched_init_port; 7687 7688 /* start misc vector */ 7689 err = ice_req_irq_msix_misc(pf); 7690 if (err) { 7691 dev_err(dev, "misc vector setup failed: %d\n", err); 7692 goto err_sched_init_port; 7693 } 7694 7695 if (test_bit(ICE_FLAG_FD_ENA, pf->flags)) { 7696 wr32(hw, PFQF_FD_ENA, PFQF_FD_ENA_FD_ENA_M); 7697 if (!rd32(hw, PFQF_FD_SIZE)) { 7698 u16 unused, guar, b_effort; 7699 7700 guar = hw->func_caps.fd_fltr_guar; 7701 b_effort = hw->func_caps.fd_fltr_best_effort; 7702 7703 /* force guaranteed filter pool for PF */ 7704 ice_alloc_fd_guar_item(hw, &unused, guar); 7705 /* force shared filter pool for PF */ 7706 ice_alloc_fd_shrd_item(hw, &unused, b_effort); 7707 } 7708 } 7709 7710 if (test_bit(ICE_FLAG_DCB_ENA, pf->flags)) 7711 ice_dcb_rebuild(pf); 7712 7713 /* If the PF previously had enabled PTP, PTP init needs to happen before 7714 * the VSI rebuild. If not, this causes the PTP link status events to 7715 * fail. 7716 */ 7717 if (test_bit(ICE_FLAG_PTP_SUPPORTED, pf->flags)) 7718 ice_ptp_rebuild(pf, reset_type); 7719 7720 if (ice_is_feature_supported(pf, ICE_F_GNSS)) 7721 ice_gnss_init(pf); 7722 7723 /* rebuild PF VSI */ 7724 err = ice_vsi_rebuild_by_type(pf, ICE_VSI_PF); 7725 if (err) { 7726 dev_err(dev, "PF VSI rebuild failed: %d\n", err); 7727 goto err_vsi_rebuild; 7728 } 7729 7730 if (reset_type == ICE_RESET_PFR) { 7731 err = ice_rebuild_channels(pf); 7732 if (err) { 7733 dev_err(dev, "failed to rebuild and replay ADQ VSIs, err %d\n", 7734 err); 7735 goto err_vsi_rebuild; 7736 } 7737 } 7738 7739 /* If Flow Director is active */ 7740 if (test_bit(ICE_FLAG_FD_ENA, pf->flags)) { 7741 err = ice_vsi_rebuild_by_type(pf, ICE_VSI_CTRL); 7742 if (err) { 7743 dev_err(dev, "control VSI rebuild failed: %d\n", err); 7744 goto err_vsi_rebuild; 7745 } 7746 7747 /* replay HW Flow Director recipes */ 7748 if (hw->fdir_prof) 7749 ice_fdir_replay_flows(hw); 7750 7751 /* replay Flow Director filters */ 7752 ice_fdir_replay_fltrs(pf); 7753 7754 ice_rebuild_arfs(pf); 7755 } 7756 7757 if (vsi && vsi->netdev) 7758 netif_device_attach(vsi->netdev); 7759 7760 ice_update_pf_netdev_link(pf); 7761 7762 /* tell the firmware we are up */ 7763 err = ice_send_version(pf); 7764 if (err) { 7765 dev_err(dev, "Rebuild failed due to error sending driver version: %d\n", 7766 err); 7767 goto err_vsi_rebuild; 7768 } 7769 7770 ice_replay_post(hw); 7771 7772 /* if we get here, reset flow is successful */ 7773 clear_bit(ICE_RESET_FAILED, pf->state); 7774 7775 ice_plug_aux_dev(pf); 7776 if (ice_is_feature_supported(pf, ICE_F_SRIOV_LAG)) 7777 ice_lag_rebuild(pf); 7778 7779 /* Restore timestamp mode settings after VSI rebuild */ 7780 ice_ptp_restore_timestamp_mode(pf); 7781 return; 7782 7783 err_vsi_rebuild: 7784 err_sched_init_port: 7785 ice_sched_cleanup_all(hw); 7786 err_init_ctrlq: 7787 ice_shutdown_all_ctrlq(hw, false); 7788 set_bit(ICE_RESET_FAILED, pf->state); 7789 clear_recovery: 7790 /* set this bit in PF state to control service task scheduling */ 7791 set_bit(ICE_NEEDS_RESTART, pf->state); 7792 dev_err(dev, "Rebuild failed, unload and reload driver\n"); 7793 } 7794 7795 /** 7796 * ice_change_mtu - NDO callback to change the MTU 7797 * @netdev: network interface device structure 7798 * @new_mtu: new value for maximum frame size 7799 * 7800 * Returns 0 on success, negative on failure 7801 */ 7802 int ice_change_mtu(struct net_device *netdev, int new_mtu) 7803 { 7804 struct ice_netdev_priv *np = netdev_priv(netdev); 7805 struct ice_vsi *vsi = np->vsi; 7806 struct ice_pf *pf = vsi->back; 7807 struct bpf_prog *prog; 7808 u8 count = 0; 7809 int err = 0; 7810 7811 if (new_mtu == (int)netdev->mtu) { 7812 netdev_warn(netdev, "MTU is already %u\n", netdev->mtu); 7813 return 0; 7814 } 7815 7816 prog = vsi->xdp_prog; 7817 if (prog && !prog->aux->xdp_has_frags) { 7818 int frame_size = ice_max_xdp_frame_size(vsi); 7819 7820 if (new_mtu + ICE_ETH_PKT_HDR_PAD > frame_size) { 7821 netdev_err(netdev, "max MTU for XDP usage is %d\n", 7822 frame_size - ICE_ETH_PKT_HDR_PAD); 7823 return -EINVAL; 7824 } 7825 } else if (test_bit(ICE_FLAG_LEGACY_RX, pf->flags)) { 7826 if (new_mtu + ICE_ETH_PKT_HDR_PAD > ICE_MAX_FRAME_LEGACY_RX) { 7827 netdev_err(netdev, "Too big MTU for legacy-rx; Max is %d\n", 7828 ICE_MAX_FRAME_LEGACY_RX - ICE_ETH_PKT_HDR_PAD); 7829 return -EINVAL; 7830 } 7831 } 7832 7833 /* if a reset is in progress, wait for some time for it to complete */ 7834 do { 7835 if (ice_is_reset_in_progress(pf->state)) { 7836 count++; 7837 usleep_range(1000, 2000); 7838 } else { 7839 break; 7840 } 7841 7842 } while (count < 100); 7843 7844 if (count == 100) { 7845 netdev_err(netdev, "can't change MTU. Device is busy\n"); 7846 return -EBUSY; 7847 } 7848 7849 WRITE_ONCE(netdev->mtu, (unsigned int)new_mtu); 7850 err = ice_down_up(vsi); 7851 if (err) 7852 return err; 7853 7854 netdev_dbg(netdev, "changed MTU to %d\n", new_mtu); 7855 set_bit(ICE_FLAG_MTU_CHANGED, pf->flags); 7856 7857 return err; 7858 } 7859 7860 /** 7861 * ice_eth_ioctl - Access the hwtstamp interface 7862 * @netdev: network interface device structure 7863 * @ifr: interface request data 7864 * @cmd: ioctl command 7865 */ 7866 static int ice_eth_ioctl(struct net_device *netdev, struct ifreq *ifr, int cmd) 7867 { 7868 struct ice_netdev_priv *np = netdev_priv(netdev); 7869 struct ice_pf *pf = np->vsi->back; 7870 7871 switch (cmd) { 7872 case SIOCGHWTSTAMP: 7873 return ice_ptp_get_ts_config(pf, ifr); 7874 case SIOCSHWTSTAMP: 7875 return ice_ptp_set_ts_config(pf, ifr); 7876 default: 7877 return -EOPNOTSUPP; 7878 } 7879 } 7880 7881 /** 7882 * ice_aq_str - convert AQ err code to a string 7883 * @aq_err: the AQ error code to convert 7884 */ 7885 const char *ice_aq_str(enum ice_aq_err aq_err) 7886 { 7887 switch (aq_err) { 7888 case ICE_AQ_RC_OK: 7889 return "OK"; 7890 case ICE_AQ_RC_EPERM: 7891 return "ICE_AQ_RC_EPERM"; 7892 case ICE_AQ_RC_ENOENT: 7893 return "ICE_AQ_RC_ENOENT"; 7894 case ICE_AQ_RC_ENOMEM: 7895 return "ICE_AQ_RC_ENOMEM"; 7896 case ICE_AQ_RC_EBUSY: 7897 return "ICE_AQ_RC_EBUSY"; 7898 case ICE_AQ_RC_EEXIST: 7899 return "ICE_AQ_RC_EEXIST"; 7900 case ICE_AQ_RC_EINVAL: 7901 return "ICE_AQ_RC_EINVAL"; 7902 case ICE_AQ_RC_ENOSPC: 7903 return "ICE_AQ_RC_ENOSPC"; 7904 case ICE_AQ_RC_ENOSYS: 7905 return "ICE_AQ_RC_ENOSYS"; 7906 case ICE_AQ_RC_EMODE: 7907 return "ICE_AQ_RC_EMODE"; 7908 case ICE_AQ_RC_ENOSEC: 7909 return "ICE_AQ_RC_ENOSEC"; 7910 case ICE_AQ_RC_EBADSIG: 7911 return "ICE_AQ_RC_EBADSIG"; 7912 case ICE_AQ_RC_ESVN: 7913 return "ICE_AQ_RC_ESVN"; 7914 case ICE_AQ_RC_EBADMAN: 7915 return "ICE_AQ_RC_EBADMAN"; 7916 case ICE_AQ_RC_EBADBUF: 7917 return "ICE_AQ_RC_EBADBUF"; 7918 } 7919 7920 return "ICE_AQ_RC_UNKNOWN"; 7921 } 7922 7923 /** 7924 * ice_set_rss_lut - Set RSS LUT 7925 * @vsi: Pointer to VSI structure 7926 * @lut: Lookup table 7927 * @lut_size: Lookup table size 7928 * 7929 * Returns 0 on success, negative on failure 7930 */ 7931 int ice_set_rss_lut(struct ice_vsi *vsi, u8 *lut, u16 lut_size) 7932 { 7933 struct ice_aq_get_set_rss_lut_params params = {}; 7934 struct ice_hw *hw = &vsi->back->hw; 7935 int status; 7936 7937 if (!lut) 7938 return -EINVAL; 7939 7940 params.vsi_handle = vsi->idx; 7941 params.lut_size = lut_size; 7942 params.lut_type = vsi->rss_lut_type; 7943 params.lut = lut; 7944 7945 status = ice_aq_set_rss_lut(hw, ¶ms); 7946 if (status) 7947 dev_err(ice_pf_to_dev(vsi->back), "Cannot set RSS lut, err %d aq_err %s\n", 7948 status, ice_aq_str(hw->adminq.sq_last_status)); 7949 7950 return status; 7951 } 7952 7953 /** 7954 * ice_set_rss_key - Set RSS key 7955 * @vsi: Pointer to the VSI structure 7956 * @seed: RSS hash seed 7957 * 7958 * Returns 0 on success, negative on failure 7959 */ 7960 int ice_set_rss_key(struct ice_vsi *vsi, u8 *seed) 7961 { 7962 struct ice_hw *hw = &vsi->back->hw; 7963 int status; 7964 7965 if (!seed) 7966 return -EINVAL; 7967 7968 status = ice_aq_set_rss_key(hw, vsi->idx, (struct ice_aqc_get_set_rss_keys *)seed); 7969 if (status) 7970 dev_err(ice_pf_to_dev(vsi->back), "Cannot set RSS key, err %d aq_err %s\n", 7971 status, ice_aq_str(hw->adminq.sq_last_status)); 7972 7973 return status; 7974 } 7975 7976 /** 7977 * ice_get_rss_lut - Get RSS LUT 7978 * @vsi: Pointer to VSI structure 7979 * @lut: Buffer to store the lookup table entries 7980 * @lut_size: Size of buffer to store the lookup table entries 7981 * 7982 * Returns 0 on success, negative on failure 7983 */ 7984 int ice_get_rss_lut(struct ice_vsi *vsi, u8 *lut, u16 lut_size) 7985 { 7986 struct ice_aq_get_set_rss_lut_params params = {}; 7987 struct ice_hw *hw = &vsi->back->hw; 7988 int status; 7989 7990 if (!lut) 7991 return -EINVAL; 7992 7993 params.vsi_handle = vsi->idx; 7994 params.lut_size = lut_size; 7995 params.lut_type = vsi->rss_lut_type; 7996 params.lut = lut; 7997 7998 status = ice_aq_get_rss_lut(hw, ¶ms); 7999 if (status) 8000 dev_err(ice_pf_to_dev(vsi->back), "Cannot get RSS lut, err %d aq_err %s\n", 8001 status, ice_aq_str(hw->adminq.sq_last_status)); 8002 8003 return status; 8004 } 8005 8006 /** 8007 * ice_get_rss_key - Get RSS key 8008 * @vsi: Pointer to VSI structure 8009 * @seed: Buffer to store the key in 8010 * 8011 * Returns 0 on success, negative on failure 8012 */ 8013 int ice_get_rss_key(struct ice_vsi *vsi, u8 *seed) 8014 { 8015 struct ice_hw *hw = &vsi->back->hw; 8016 int status; 8017 8018 if (!seed) 8019 return -EINVAL; 8020 8021 status = ice_aq_get_rss_key(hw, vsi->idx, (struct ice_aqc_get_set_rss_keys *)seed); 8022 if (status) 8023 dev_err(ice_pf_to_dev(vsi->back), "Cannot get RSS key, err %d aq_err %s\n", 8024 status, ice_aq_str(hw->adminq.sq_last_status)); 8025 8026 return status; 8027 } 8028 8029 /** 8030 * ice_set_rss_hfunc - Set RSS HASH function 8031 * @vsi: Pointer to VSI structure 8032 * @hfunc: hash function (ICE_AQ_VSI_Q_OPT_RSS_*) 8033 * 8034 * Returns 0 on success, negative on failure 8035 */ 8036 int ice_set_rss_hfunc(struct ice_vsi *vsi, u8 hfunc) 8037 { 8038 struct ice_hw *hw = &vsi->back->hw; 8039 struct ice_vsi_ctx *ctx; 8040 bool symm; 8041 int err; 8042 8043 if (hfunc == vsi->rss_hfunc) 8044 return 0; 8045 8046 if (hfunc != ICE_AQ_VSI_Q_OPT_RSS_HASH_TPLZ && 8047 hfunc != ICE_AQ_VSI_Q_OPT_RSS_HASH_SYM_TPLZ) 8048 return -EOPNOTSUPP; 8049 8050 ctx = kzalloc(sizeof(*ctx), GFP_KERNEL); 8051 if (!ctx) 8052 return -ENOMEM; 8053 8054 ctx->info.valid_sections = cpu_to_le16(ICE_AQ_VSI_PROP_Q_OPT_VALID); 8055 ctx->info.q_opt_rss = vsi->info.q_opt_rss; 8056 ctx->info.q_opt_rss &= ~ICE_AQ_VSI_Q_OPT_RSS_HASH_M; 8057 ctx->info.q_opt_rss |= 8058 FIELD_PREP(ICE_AQ_VSI_Q_OPT_RSS_HASH_M, hfunc); 8059 ctx->info.q_opt_tc = vsi->info.q_opt_tc; 8060 ctx->info.q_opt_flags = vsi->info.q_opt_rss; 8061 8062 err = ice_update_vsi(hw, vsi->idx, ctx, NULL); 8063 if (err) { 8064 dev_err(ice_pf_to_dev(vsi->back), "Failed to configure RSS hash for VSI %d, error %d\n", 8065 vsi->vsi_num, err); 8066 } else { 8067 vsi->info.q_opt_rss = ctx->info.q_opt_rss; 8068 vsi->rss_hfunc = hfunc; 8069 netdev_info(vsi->netdev, "Hash function set to: %sToeplitz\n", 8070 hfunc == ICE_AQ_VSI_Q_OPT_RSS_HASH_SYM_TPLZ ? 8071 "Symmetric " : ""); 8072 } 8073 kfree(ctx); 8074 if (err) 8075 return err; 8076 8077 /* Fix the symmetry setting for all existing RSS configurations */ 8078 symm = !!(hfunc == ICE_AQ_VSI_Q_OPT_RSS_HASH_SYM_TPLZ); 8079 return ice_set_rss_cfg_symm(hw, vsi, symm); 8080 } 8081 8082 /** 8083 * ice_bridge_getlink - Get the hardware bridge mode 8084 * @skb: skb buff 8085 * @pid: process ID 8086 * @seq: RTNL message seq 8087 * @dev: the netdev being configured 8088 * @filter_mask: filter mask passed in 8089 * @nlflags: netlink flags passed in 8090 * 8091 * Return the bridge mode (VEB/VEPA) 8092 */ 8093 static int 8094 ice_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq, 8095 struct net_device *dev, u32 filter_mask, int nlflags) 8096 { 8097 struct ice_netdev_priv *np = netdev_priv(dev); 8098 struct ice_vsi *vsi = np->vsi; 8099 struct ice_pf *pf = vsi->back; 8100 u16 bmode; 8101 8102 bmode = pf->first_sw->bridge_mode; 8103 8104 return ndo_dflt_bridge_getlink(skb, pid, seq, dev, bmode, 0, 0, nlflags, 8105 filter_mask, NULL); 8106 } 8107 8108 /** 8109 * ice_vsi_update_bridge_mode - Update VSI for switching bridge mode (VEB/VEPA) 8110 * @vsi: Pointer to VSI structure 8111 * @bmode: Hardware bridge mode (VEB/VEPA) 8112 * 8113 * Returns 0 on success, negative on failure 8114 */ 8115 static int ice_vsi_update_bridge_mode(struct ice_vsi *vsi, u16 bmode) 8116 { 8117 struct ice_aqc_vsi_props *vsi_props; 8118 struct ice_hw *hw = &vsi->back->hw; 8119 struct ice_vsi_ctx *ctxt; 8120 int ret; 8121 8122 vsi_props = &vsi->info; 8123 8124 ctxt = kzalloc(sizeof(*ctxt), GFP_KERNEL); 8125 if (!ctxt) 8126 return -ENOMEM; 8127 8128 ctxt->info = vsi->info; 8129 8130 if (bmode == BRIDGE_MODE_VEB) 8131 /* change from VEPA to VEB mode */ 8132 ctxt->info.sw_flags |= ICE_AQ_VSI_SW_FLAG_ALLOW_LB; 8133 else 8134 /* change from VEB to VEPA mode */ 8135 ctxt->info.sw_flags &= ~ICE_AQ_VSI_SW_FLAG_ALLOW_LB; 8136 ctxt->info.valid_sections = cpu_to_le16(ICE_AQ_VSI_PROP_SW_VALID); 8137 8138 ret = ice_update_vsi(hw, vsi->idx, ctxt, NULL); 8139 if (ret) { 8140 dev_err(ice_pf_to_dev(vsi->back), "update VSI for bridge mode failed, bmode = %d err %d aq_err %s\n", 8141 bmode, ret, ice_aq_str(hw->adminq.sq_last_status)); 8142 goto out; 8143 } 8144 /* Update sw flags for book keeping */ 8145 vsi_props->sw_flags = ctxt->info.sw_flags; 8146 8147 out: 8148 kfree(ctxt); 8149 return ret; 8150 } 8151 8152 /** 8153 * ice_bridge_setlink - Set the hardware bridge mode 8154 * @dev: the netdev being configured 8155 * @nlh: RTNL message 8156 * @flags: bridge setlink flags 8157 * @extack: netlink extended ack 8158 * 8159 * Sets the bridge mode (VEB/VEPA) of the switch to which the netdev (VSI) is 8160 * hooked up to. Iterates through the PF VSI list and sets the loopback mode (if 8161 * not already set for all VSIs connected to this switch. And also update the 8162 * unicast switch filter rules for the corresponding switch of the netdev. 8163 */ 8164 static int 8165 ice_bridge_setlink(struct net_device *dev, struct nlmsghdr *nlh, 8166 u16 __always_unused flags, 8167 struct netlink_ext_ack __always_unused *extack) 8168 { 8169 struct ice_netdev_priv *np = netdev_priv(dev); 8170 struct ice_pf *pf = np->vsi->back; 8171 struct nlattr *attr, *br_spec; 8172 struct ice_hw *hw = &pf->hw; 8173 struct ice_sw *pf_sw; 8174 int rem, v, err = 0; 8175 8176 pf_sw = pf->first_sw; 8177 /* find the attribute in the netlink message */ 8178 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC); 8179 if (!br_spec) 8180 return -EINVAL; 8181 8182 nla_for_each_nested_type(attr, IFLA_BRIDGE_MODE, br_spec, rem) { 8183 __u16 mode = nla_get_u16(attr); 8184 8185 if (mode != BRIDGE_MODE_VEPA && mode != BRIDGE_MODE_VEB) 8186 return -EINVAL; 8187 /* Continue if bridge mode is not being flipped */ 8188 if (mode == pf_sw->bridge_mode) 8189 continue; 8190 /* Iterates through the PF VSI list and update the loopback 8191 * mode of the VSI 8192 */ 8193 ice_for_each_vsi(pf, v) { 8194 if (!pf->vsi[v]) 8195 continue; 8196 err = ice_vsi_update_bridge_mode(pf->vsi[v], mode); 8197 if (err) 8198 return err; 8199 } 8200 8201 hw->evb_veb = (mode == BRIDGE_MODE_VEB); 8202 /* Update the unicast switch filter rules for the corresponding 8203 * switch of the netdev 8204 */ 8205 err = ice_update_sw_rule_bridge_mode(hw); 8206 if (err) { 8207 netdev_err(dev, "switch rule update failed, mode = %d err %d aq_err %s\n", 8208 mode, err, 8209 ice_aq_str(hw->adminq.sq_last_status)); 8210 /* revert hw->evb_veb */ 8211 hw->evb_veb = (pf_sw->bridge_mode == BRIDGE_MODE_VEB); 8212 return err; 8213 } 8214 8215 pf_sw->bridge_mode = mode; 8216 } 8217 8218 return 0; 8219 } 8220 8221 /** 8222 * ice_tx_timeout - Respond to a Tx Hang 8223 * @netdev: network interface device structure 8224 * @txqueue: Tx queue 8225 */ 8226 void ice_tx_timeout(struct net_device *netdev, unsigned int txqueue) 8227 { 8228 struct ice_netdev_priv *np = netdev_priv(netdev); 8229 struct ice_tx_ring *tx_ring = NULL; 8230 struct ice_vsi *vsi = np->vsi; 8231 struct ice_pf *pf = vsi->back; 8232 u32 i; 8233 8234 pf->tx_timeout_count++; 8235 8236 /* Check if PFC is enabled for the TC to which the queue belongs 8237 * to. If yes then Tx timeout is not caused by a hung queue, no 8238 * need to reset and rebuild 8239 */ 8240 if (ice_is_pfc_causing_hung_q(pf, txqueue)) { 8241 dev_info(ice_pf_to_dev(pf), "Fake Tx hang detected on queue %u, timeout caused by PFC storm\n", 8242 txqueue); 8243 return; 8244 } 8245 8246 /* now that we have an index, find the tx_ring struct */ 8247 ice_for_each_txq(vsi, i) 8248 if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc) 8249 if (txqueue == vsi->tx_rings[i]->q_index) { 8250 tx_ring = vsi->tx_rings[i]; 8251 break; 8252 } 8253 8254 /* Reset recovery level if enough time has elapsed after last timeout. 8255 * Also ensure no new reset action happens before next timeout period. 8256 */ 8257 if (time_after(jiffies, (pf->tx_timeout_last_recovery + HZ * 20))) 8258 pf->tx_timeout_recovery_level = 1; 8259 else if (time_before(jiffies, (pf->tx_timeout_last_recovery + 8260 netdev->watchdog_timeo))) 8261 return; 8262 8263 if (tx_ring) { 8264 struct ice_hw *hw = &pf->hw; 8265 u32 head, val = 0; 8266 8267 head = FIELD_GET(QTX_COMM_HEAD_HEAD_M, 8268 rd32(hw, QTX_COMM_HEAD(vsi->txq_map[txqueue]))); 8269 /* Read interrupt register */ 8270 val = rd32(hw, GLINT_DYN_CTL(tx_ring->q_vector->reg_idx)); 8271 8272 netdev_info(netdev, "tx_timeout: VSI_num: %d, Q %u, NTC: 0x%x, HW_HEAD: 0x%x, NTU: 0x%x, INT: 0x%x\n", 8273 vsi->vsi_num, txqueue, tx_ring->next_to_clean, 8274 head, tx_ring->next_to_use, val); 8275 } 8276 8277 pf->tx_timeout_last_recovery = jiffies; 8278 netdev_info(netdev, "tx_timeout recovery level %d, txqueue %u\n", 8279 pf->tx_timeout_recovery_level, txqueue); 8280 8281 switch (pf->tx_timeout_recovery_level) { 8282 case 1: 8283 set_bit(ICE_PFR_REQ, pf->state); 8284 break; 8285 case 2: 8286 set_bit(ICE_CORER_REQ, pf->state); 8287 break; 8288 case 3: 8289 set_bit(ICE_GLOBR_REQ, pf->state); 8290 break; 8291 default: 8292 netdev_err(netdev, "tx_timeout recovery unsuccessful, device is in unrecoverable state.\n"); 8293 set_bit(ICE_DOWN, pf->state); 8294 set_bit(ICE_VSI_NEEDS_RESTART, vsi->state); 8295 set_bit(ICE_SERVICE_DIS, pf->state); 8296 break; 8297 } 8298 8299 ice_service_task_schedule(pf); 8300 pf->tx_timeout_recovery_level++; 8301 } 8302 8303 /** 8304 * ice_setup_tc_cls_flower - flower classifier offloads 8305 * @np: net device to configure 8306 * @filter_dev: device on which filter is added 8307 * @cls_flower: offload data 8308 */ 8309 static int 8310 ice_setup_tc_cls_flower(struct ice_netdev_priv *np, 8311 struct net_device *filter_dev, 8312 struct flow_cls_offload *cls_flower) 8313 { 8314 struct ice_vsi *vsi = np->vsi; 8315 8316 if (cls_flower->common.chain_index) 8317 return -EOPNOTSUPP; 8318 8319 switch (cls_flower->command) { 8320 case FLOW_CLS_REPLACE: 8321 return ice_add_cls_flower(filter_dev, vsi, cls_flower); 8322 case FLOW_CLS_DESTROY: 8323 return ice_del_cls_flower(vsi, cls_flower); 8324 default: 8325 return -EINVAL; 8326 } 8327 } 8328 8329 /** 8330 * ice_setup_tc_block_cb - callback handler registered for TC block 8331 * @type: TC SETUP type 8332 * @type_data: TC flower offload data that contains user input 8333 * @cb_priv: netdev private data 8334 */ 8335 static int 8336 ice_setup_tc_block_cb(enum tc_setup_type type, void *type_data, void *cb_priv) 8337 { 8338 struct ice_netdev_priv *np = cb_priv; 8339 8340 switch (type) { 8341 case TC_SETUP_CLSFLOWER: 8342 return ice_setup_tc_cls_flower(np, np->vsi->netdev, 8343 type_data); 8344 default: 8345 return -EOPNOTSUPP; 8346 } 8347 } 8348 8349 /** 8350 * ice_validate_mqprio_qopt - Validate TCF input parameters 8351 * @vsi: Pointer to VSI 8352 * @mqprio_qopt: input parameters for mqprio queue configuration 8353 * 8354 * This function validates MQPRIO params, such as qcount (power of 2 wherever 8355 * needed), and make sure user doesn't specify qcount and BW rate limit 8356 * for TCs, which are more than "num_tc" 8357 */ 8358 static int 8359 ice_validate_mqprio_qopt(struct ice_vsi *vsi, 8360 struct tc_mqprio_qopt_offload *mqprio_qopt) 8361 { 8362 int non_power_of_2_qcount = 0; 8363 struct ice_pf *pf = vsi->back; 8364 int max_rss_q_cnt = 0; 8365 u64 sum_min_rate = 0; 8366 struct device *dev; 8367 int i, speed; 8368 u8 num_tc; 8369 8370 if (vsi->type != ICE_VSI_PF) 8371 return -EINVAL; 8372 8373 if (mqprio_qopt->qopt.offset[0] != 0 || 8374 mqprio_qopt->qopt.num_tc < 1 || 8375 mqprio_qopt->qopt.num_tc > ICE_CHNL_MAX_TC) 8376 return -EINVAL; 8377 8378 dev = ice_pf_to_dev(pf); 8379 vsi->ch_rss_size = 0; 8380 num_tc = mqprio_qopt->qopt.num_tc; 8381 speed = ice_get_link_speed_kbps(vsi); 8382 8383 for (i = 0; num_tc; i++) { 8384 int qcount = mqprio_qopt->qopt.count[i]; 8385 u64 max_rate, min_rate, rem; 8386 8387 if (!qcount) 8388 return -EINVAL; 8389 8390 if (is_power_of_2(qcount)) { 8391 if (non_power_of_2_qcount && 8392 qcount > non_power_of_2_qcount) { 8393 dev_err(dev, "qcount[%d] cannot be greater than non power of 2 qcount[%d]\n", 8394 qcount, non_power_of_2_qcount); 8395 return -EINVAL; 8396 } 8397 if (qcount > max_rss_q_cnt) 8398 max_rss_q_cnt = qcount; 8399 } else { 8400 if (non_power_of_2_qcount && 8401 qcount != non_power_of_2_qcount) { 8402 dev_err(dev, "Only one non power of 2 qcount allowed[%d,%d]\n", 8403 qcount, non_power_of_2_qcount); 8404 return -EINVAL; 8405 } 8406 if (qcount < max_rss_q_cnt) { 8407 dev_err(dev, "non power of 2 qcount[%d] cannot be less than other qcount[%d]\n", 8408 qcount, max_rss_q_cnt); 8409 return -EINVAL; 8410 } 8411 max_rss_q_cnt = qcount; 8412 non_power_of_2_qcount = qcount; 8413 } 8414 8415 /* TC command takes input in K/N/Gbps or K/M/Gbit etc but 8416 * converts the bandwidth rate limit into Bytes/s when 8417 * passing it down to the driver. So convert input bandwidth 8418 * from Bytes/s to Kbps 8419 */ 8420 max_rate = mqprio_qopt->max_rate[i]; 8421 max_rate = div_u64(max_rate, ICE_BW_KBPS_DIVISOR); 8422 8423 /* min_rate is minimum guaranteed rate and it can't be zero */ 8424 min_rate = mqprio_qopt->min_rate[i]; 8425 min_rate = div_u64(min_rate, ICE_BW_KBPS_DIVISOR); 8426 sum_min_rate += min_rate; 8427 8428 if (min_rate && min_rate < ICE_MIN_BW_LIMIT) { 8429 dev_err(dev, "TC%d: min_rate(%llu Kbps) < %u Kbps\n", i, 8430 min_rate, ICE_MIN_BW_LIMIT); 8431 return -EINVAL; 8432 } 8433 8434 if (max_rate && max_rate > speed) { 8435 dev_err(dev, "TC%d: max_rate(%llu Kbps) > link speed of %u Kbps\n", 8436 i, max_rate, speed); 8437 return -EINVAL; 8438 } 8439 8440 iter_div_u64_rem(min_rate, ICE_MIN_BW_LIMIT, &rem); 8441 if (rem) { 8442 dev_err(dev, "TC%d: Min Rate not multiple of %u Kbps", 8443 i, ICE_MIN_BW_LIMIT); 8444 return -EINVAL; 8445 } 8446 8447 iter_div_u64_rem(max_rate, ICE_MIN_BW_LIMIT, &rem); 8448 if (rem) { 8449 dev_err(dev, "TC%d: Max Rate not multiple of %u Kbps", 8450 i, ICE_MIN_BW_LIMIT); 8451 return -EINVAL; 8452 } 8453 8454 /* min_rate can't be more than max_rate, except when max_rate 8455 * is zero (implies max_rate sought is max line rate). In such 8456 * a case min_rate can be more than max. 8457 */ 8458 if (max_rate && min_rate > max_rate) { 8459 dev_err(dev, "min_rate %llu Kbps can't be more than max_rate %llu Kbps\n", 8460 min_rate, max_rate); 8461 return -EINVAL; 8462 } 8463 8464 if (i >= mqprio_qopt->qopt.num_tc - 1) 8465 break; 8466 if (mqprio_qopt->qopt.offset[i + 1] != 8467 (mqprio_qopt->qopt.offset[i] + qcount)) 8468 return -EINVAL; 8469 } 8470 if (vsi->num_rxq < 8471 (mqprio_qopt->qopt.offset[i] + mqprio_qopt->qopt.count[i])) 8472 return -EINVAL; 8473 if (vsi->num_txq < 8474 (mqprio_qopt->qopt.offset[i] + mqprio_qopt->qopt.count[i])) 8475 return -EINVAL; 8476 8477 if (sum_min_rate && sum_min_rate > (u64)speed) { 8478 dev_err(dev, "Invalid min Tx rate(%llu) Kbps > speed (%u) Kbps specified\n", 8479 sum_min_rate, speed); 8480 return -EINVAL; 8481 } 8482 8483 /* make sure vsi->ch_rss_size is set correctly based on TC's qcount */ 8484 vsi->ch_rss_size = max_rss_q_cnt; 8485 8486 return 0; 8487 } 8488 8489 /** 8490 * ice_add_vsi_to_fdir - add a VSI to the flow director group for PF 8491 * @pf: ptr to PF device 8492 * @vsi: ptr to VSI 8493 */ 8494 static int ice_add_vsi_to_fdir(struct ice_pf *pf, struct ice_vsi *vsi) 8495 { 8496 struct device *dev = ice_pf_to_dev(pf); 8497 bool added = false; 8498 struct ice_hw *hw; 8499 int flow; 8500 8501 if (!(vsi->num_gfltr || vsi->num_bfltr)) 8502 return -EINVAL; 8503 8504 hw = &pf->hw; 8505 for (flow = 0; flow < ICE_FLTR_PTYPE_MAX; flow++) { 8506 struct ice_fd_hw_prof *prof; 8507 int tun, status; 8508 u64 entry_h; 8509 8510 if (!(hw->fdir_prof && hw->fdir_prof[flow] && 8511 hw->fdir_prof[flow]->cnt)) 8512 continue; 8513 8514 for (tun = 0; tun < ICE_FD_HW_SEG_MAX; tun++) { 8515 enum ice_flow_priority prio; 8516 8517 /* add this VSI to FDir profile for this flow */ 8518 prio = ICE_FLOW_PRIO_NORMAL; 8519 prof = hw->fdir_prof[flow]; 8520 status = ice_flow_add_entry(hw, ICE_BLK_FD, 8521 prof->prof_id[tun], 8522 prof->vsi_h[0], vsi->idx, 8523 prio, prof->fdir_seg[tun], 8524 &entry_h); 8525 if (status) { 8526 dev_err(dev, "channel VSI idx %d, not able to add to group %d\n", 8527 vsi->idx, flow); 8528 continue; 8529 } 8530 8531 prof->entry_h[prof->cnt][tun] = entry_h; 8532 } 8533 8534 /* store VSI for filter replay and delete */ 8535 prof->vsi_h[prof->cnt] = vsi->idx; 8536 prof->cnt++; 8537 8538 added = true; 8539 dev_dbg(dev, "VSI idx %d added to fdir group %d\n", vsi->idx, 8540 flow); 8541 } 8542 8543 if (!added) 8544 dev_dbg(dev, "VSI idx %d not added to fdir groups\n", vsi->idx); 8545 8546 return 0; 8547 } 8548 8549 /** 8550 * ice_add_channel - add a channel by adding VSI 8551 * @pf: ptr to PF device 8552 * @sw_id: underlying HW switching element ID 8553 * @ch: ptr to channel structure 8554 * 8555 * Add a channel (VSI) using add_vsi and queue_map 8556 */ 8557 static int ice_add_channel(struct ice_pf *pf, u16 sw_id, struct ice_channel *ch) 8558 { 8559 struct device *dev = ice_pf_to_dev(pf); 8560 struct ice_vsi *vsi; 8561 8562 if (ch->type != ICE_VSI_CHNL) { 8563 dev_err(dev, "add new VSI failed, ch->type %d\n", ch->type); 8564 return -EINVAL; 8565 } 8566 8567 vsi = ice_chnl_vsi_setup(pf, pf->hw.port_info, ch); 8568 if (!vsi || vsi->type != ICE_VSI_CHNL) { 8569 dev_err(dev, "create chnl VSI failure\n"); 8570 return -EINVAL; 8571 } 8572 8573 ice_add_vsi_to_fdir(pf, vsi); 8574 8575 ch->sw_id = sw_id; 8576 ch->vsi_num = vsi->vsi_num; 8577 ch->info.mapping_flags = vsi->info.mapping_flags; 8578 ch->ch_vsi = vsi; 8579 /* set the back pointer of channel for newly created VSI */ 8580 vsi->ch = ch; 8581 8582 memcpy(&ch->info.q_mapping, &vsi->info.q_mapping, 8583 sizeof(vsi->info.q_mapping)); 8584 memcpy(&ch->info.tc_mapping, vsi->info.tc_mapping, 8585 sizeof(vsi->info.tc_mapping)); 8586 8587 return 0; 8588 } 8589 8590 /** 8591 * ice_chnl_cfg_res 8592 * @vsi: the VSI being setup 8593 * @ch: ptr to channel structure 8594 * 8595 * Configure channel specific resources such as rings, vector. 8596 */ 8597 static void ice_chnl_cfg_res(struct ice_vsi *vsi, struct ice_channel *ch) 8598 { 8599 int i; 8600 8601 for (i = 0; i < ch->num_txq; i++) { 8602 struct ice_q_vector *tx_q_vector, *rx_q_vector; 8603 struct ice_ring_container *rc; 8604 struct ice_tx_ring *tx_ring; 8605 struct ice_rx_ring *rx_ring; 8606 8607 tx_ring = vsi->tx_rings[ch->base_q + i]; 8608 rx_ring = vsi->rx_rings[ch->base_q + i]; 8609 if (!tx_ring || !rx_ring) 8610 continue; 8611 8612 /* setup ring being channel enabled */ 8613 tx_ring->ch = ch; 8614 rx_ring->ch = ch; 8615 8616 /* following code block sets up vector specific attributes */ 8617 tx_q_vector = tx_ring->q_vector; 8618 rx_q_vector = rx_ring->q_vector; 8619 if (!tx_q_vector && !rx_q_vector) 8620 continue; 8621 8622 if (tx_q_vector) { 8623 tx_q_vector->ch = ch; 8624 /* setup Tx and Rx ITR setting if DIM is off */ 8625 rc = &tx_q_vector->tx; 8626 if (!ITR_IS_DYNAMIC(rc)) 8627 ice_write_itr(rc, rc->itr_setting); 8628 } 8629 if (rx_q_vector) { 8630 rx_q_vector->ch = ch; 8631 /* setup Tx and Rx ITR setting if DIM is off */ 8632 rc = &rx_q_vector->rx; 8633 if (!ITR_IS_DYNAMIC(rc)) 8634 ice_write_itr(rc, rc->itr_setting); 8635 } 8636 } 8637 8638 /* it is safe to assume that, if channel has non-zero num_t[r]xq, then 8639 * GLINT_ITR register would have written to perform in-context 8640 * update, hence perform flush 8641 */ 8642 if (ch->num_txq || ch->num_rxq) 8643 ice_flush(&vsi->back->hw); 8644 } 8645 8646 /** 8647 * ice_cfg_chnl_all_res - configure channel resources 8648 * @vsi: pte to main_vsi 8649 * @ch: ptr to channel structure 8650 * 8651 * This function configures channel specific resources such as flow-director 8652 * counter index, and other resources such as queues, vectors, ITR settings 8653 */ 8654 static void 8655 ice_cfg_chnl_all_res(struct ice_vsi *vsi, struct ice_channel *ch) 8656 { 8657 /* configure channel (aka ADQ) resources such as queues, vectors, 8658 * ITR settings for channel specific vectors and anything else 8659 */ 8660 ice_chnl_cfg_res(vsi, ch); 8661 } 8662 8663 /** 8664 * ice_setup_hw_channel - setup new channel 8665 * @pf: ptr to PF device 8666 * @vsi: the VSI being setup 8667 * @ch: ptr to channel structure 8668 * @sw_id: underlying HW switching element ID 8669 * @type: type of channel to be created (VMDq2/VF) 8670 * 8671 * Setup new channel (VSI) based on specified type (VMDq2/VF) 8672 * and configures Tx rings accordingly 8673 */ 8674 static int 8675 ice_setup_hw_channel(struct ice_pf *pf, struct ice_vsi *vsi, 8676 struct ice_channel *ch, u16 sw_id, u8 type) 8677 { 8678 struct device *dev = ice_pf_to_dev(pf); 8679 int ret; 8680 8681 ch->base_q = vsi->next_base_q; 8682 ch->type = type; 8683 8684 ret = ice_add_channel(pf, sw_id, ch); 8685 if (ret) { 8686 dev_err(dev, "failed to add_channel using sw_id %u\n", sw_id); 8687 return ret; 8688 } 8689 8690 /* configure/setup ADQ specific resources */ 8691 ice_cfg_chnl_all_res(vsi, ch); 8692 8693 /* make sure to update the next_base_q so that subsequent channel's 8694 * (aka ADQ) VSI queue map is correct 8695 */ 8696 vsi->next_base_q = vsi->next_base_q + ch->num_rxq; 8697 dev_dbg(dev, "added channel: vsi_num %u, num_rxq %u\n", ch->vsi_num, 8698 ch->num_rxq); 8699 8700 return 0; 8701 } 8702 8703 /** 8704 * ice_setup_channel - setup new channel using uplink element 8705 * @pf: ptr to PF device 8706 * @vsi: the VSI being setup 8707 * @ch: ptr to channel structure 8708 * 8709 * Setup new channel (VSI) based on specified type (VMDq2/VF) 8710 * and uplink switching element 8711 */ 8712 static bool 8713 ice_setup_channel(struct ice_pf *pf, struct ice_vsi *vsi, 8714 struct ice_channel *ch) 8715 { 8716 struct device *dev = ice_pf_to_dev(pf); 8717 u16 sw_id; 8718 int ret; 8719 8720 if (vsi->type != ICE_VSI_PF) { 8721 dev_err(dev, "unsupported parent VSI type(%d)\n", vsi->type); 8722 return false; 8723 } 8724 8725 sw_id = pf->first_sw->sw_id; 8726 8727 /* create channel (VSI) */ 8728 ret = ice_setup_hw_channel(pf, vsi, ch, sw_id, ICE_VSI_CHNL); 8729 if (ret) { 8730 dev_err(dev, "failed to setup hw_channel\n"); 8731 return false; 8732 } 8733 dev_dbg(dev, "successfully created channel()\n"); 8734 8735 return ch->ch_vsi ? true : false; 8736 } 8737 8738 /** 8739 * ice_set_bw_limit - setup BW limit for Tx traffic based on max_tx_rate 8740 * @vsi: VSI to be configured 8741 * @max_tx_rate: max Tx rate in Kbps to be configured as maximum BW limit 8742 * @min_tx_rate: min Tx rate in Kbps to be configured as minimum BW limit 8743 */ 8744 static int 8745 ice_set_bw_limit(struct ice_vsi *vsi, u64 max_tx_rate, u64 min_tx_rate) 8746 { 8747 int err; 8748 8749 err = ice_set_min_bw_limit(vsi, min_tx_rate); 8750 if (err) 8751 return err; 8752 8753 return ice_set_max_bw_limit(vsi, max_tx_rate); 8754 } 8755 8756 /** 8757 * ice_create_q_channel - function to create channel 8758 * @vsi: VSI to be configured 8759 * @ch: ptr to channel (it contains channel specific params) 8760 * 8761 * This function creates channel (VSI) using num_queues specified by user, 8762 * reconfigs RSS if needed. 8763 */ 8764 static int ice_create_q_channel(struct ice_vsi *vsi, struct ice_channel *ch) 8765 { 8766 struct ice_pf *pf = vsi->back; 8767 struct device *dev; 8768 8769 if (!ch) 8770 return -EINVAL; 8771 8772 dev = ice_pf_to_dev(pf); 8773 if (!ch->num_txq || !ch->num_rxq) { 8774 dev_err(dev, "Invalid num_queues requested: %d\n", ch->num_rxq); 8775 return -EINVAL; 8776 } 8777 8778 if (!vsi->cnt_q_avail || vsi->cnt_q_avail < ch->num_txq) { 8779 dev_err(dev, "cnt_q_avail (%u) less than num_queues %d\n", 8780 vsi->cnt_q_avail, ch->num_txq); 8781 return -EINVAL; 8782 } 8783 8784 if (!ice_setup_channel(pf, vsi, ch)) { 8785 dev_info(dev, "Failed to setup channel\n"); 8786 return -EINVAL; 8787 } 8788 /* configure BW rate limit */ 8789 if (ch->ch_vsi && (ch->max_tx_rate || ch->min_tx_rate)) { 8790 int ret; 8791 8792 ret = ice_set_bw_limit(ch->ch_vsi, ch->max_tx_rate, 8793 ch->min_tx_rate); 8794 if (ret) 8795 dev_err(dev, "failed to set Tx rate of %llu Kbps for VSI(%u)\n", 8796 ch->max_tx_rate, ch->ch_vsi->vsi_num); 8797 else 8798 dev_dbg(dev, "set Tx rate of %llu Kbps for VSI(%u)\n", 8799 ch->max_tx_rate, ch->ch_vsi->vsi_num); 8800 } 8801 8802 vsi->cnt_q_avail -= ch->num_txq; 8803 8804 return 0; 8805 } 8806 8807 /** 8808 * ice_rem_all_chnl_fltrs - removes all channel filters 8809 * @pf: ptr to PF, TC-flower based filter are tracked at PF level 8810 * 8811 * Remove all advanced switch filters only if they are channel specific 8812 * tc-flower based filter 8813 */ 8814 static void ice_rem_all_chnl_fltrs(struct ice_pf *pf) 8815 { 8816 struct ice_tc_flower_fltr *fltr; 8817 struct hlist_node *node; 8818 8819 /* to remove all channel filters, iterate an ordered list of filters */ 8820 hlist_for_each_entry_safe(fltr, node, 8821 &pf->tc_flower_fltr_list, 8822 tc_flower_node) { 8823 struct ice_rule_query_data rule; 8824 int status; 8825 8826 /* for now process only channel specific filters */ 8827 if (!ice_is_chnl_fltr(fltr)) 8828 continue; 8829 8830 rule.rid = fltr->rid; 8831 rule.rule_id = fltr->rule_id; 8832 rule.vsi_handle = fltr->dest_vsi_handle; 8833 status = ice_rem_adv_rule_by_id(&pf->hw, &rule); 8834 if (status) { 8835 if (status == -ENOENT) 8836 dev_dbg(ice_pf_to_dev(pf), "TC flower filter (rule_id %u) does not exist\n", 8837 rule.rule_id); 8838 else 8839 dev_err(ice_pf_to_dev(pf), "failed to delete TC flower filter, status %d\n", 8840 status); 8841 } else if (fltr->dest_vsi) { 8842 /* update advanced switch filter count */ 8843 if (fltr->dest_vsi->type == ICE_VSI_CHNL) { 8844 u32 flags = fltr->flags; 8845 8846 fltr->dest_vsi->num_chnl_fltr--; 8847 if (flags & (ICE_TC_FLWR_FIELD_DST_MAC | 8848 ICE_TC_FLWR_FIELD_ENC_DST_MAC)) 8849 pf->num_dmac_chnl_fltrs--; 8850 } 8851 } 8852 8853 hlist_del(&fltr->tc_flower_node); 8854 kfree(fltr); 8855 } 8856 } 8857 8858 /** 8859 * ice_remove_q_channels - Remove queue channels for the TCs 8860 * @vsi: VSI to be configured 8861 * @rem_fltr: delete advanced switch filter or not 8862 * 8863 * Remove queue channels for the TCs 8864 */ 8865 static void ice_remove_q_channels(struct ice_vsi *vsi, bool rem_fltr) 8866 { 8867 struct ice_channel *ch, *ch_tmp; 8868 struct ice_pf *pf = vsi->back; 8869 int i; 8870 8871 /* remove all tc-flower based filter if they are channel filters only */ 8872 if (rem_fltr) 8873 ice_rem_all_chnl_fltrs(pf); 8874 8875 /* remove ntuple filters since queue configuration is being changed */ 8876 if (vsi->netdev->features & NETIF_F_NTUPLE) { 8877 struct ice_hw *hw = &pf->hw; 8878 8879 mutex_lock(&hw->fdir_fltr_lock); 8880 ice_fdir_del_all_fltrs(vsi); 8881 mutex_unlock(&hw->fdir_fltr_lock); 8882 } 8883 8884 /* perform cleanup for channels if they exist */ 8885 list_for_each_entry_safe(ch, ch_tmp, &vsi->ch_list, list) { 8886 struct ice_vsi *ch_vsi; 8887 8888 list_del(&ch->list); 8889 ch_vsi = ch->ch_vsi; 8890 if (!ch_vsi) { 8891 kfree(ch); 8892 continue; 8893 } 8894 8895 /* Reset queue contexts */ 8896 for (i = 0; i < ch->num_rxq; i++) { 8897 struct ice_tx_ring *tx_ring; 8898 struct ice_rx_ring *rx_ring; 8899 8900 tx_ring = vsi->tx_rings[ch->base_q + i]; 8901 rx_ring = vsi->rx_rings[ch->base_q + i]; 8902 if (tx_ring) { 8903 tx_ring->ch = NULL; 8904 if (tx_ring->q_vector) 8905 tx_ring->q_vector->ch = NULL; 8906 } 8907 if (rx_ring) { 8908 rx_ring->ch = NULL; 8909 if (rx_ring->q_vector) 8910 rx_ring->q_vector->ch = NULL; 8911 } 8912 } 8913 8914 /* Release FD resources for the channel VSI */ 8915 ice_fdir_rem_adq_chnl(&pf->hw, ch->ch_vsi->idx); 8916 8917 /* clear the VSI from scheduler tree */ 8918 ice_rm_vsi_lan_cfg(ch->ch_vsi->port_info, ch->ch_vsi->idx); 8919 8920 /* Delete VSI from FW, PF and HW VSI arrays */ 8921 ice_vsi_delete(ch->ch_vsi); 8922 8923 /* free the channel */ 8924 kfree(ch); 8925 } 8926 8927 /* clear the channel VSI map which is stored in main VSI */ 8928 ice_for_each_chnl_tc(i) 8929 vsi->tc_map_vsi[i] = NULL; 8930 8931 /* reset main VSI's all TC information */ 8932 vsi->all_enatc = 0; 8933 vsi->all_numtc = 0; 8934 } 8935 8936 /** 8937 * ice_rebuild_channels - rebuild channel 8938 * @pf: ptr to PF 8939 * 8940 * Recreate channel VSIs and replay filters 8941 */ 8942 static int ice_rebuild_channels(struct ice_pf *pf) 8943 { 8944 struct device *dev = ice_pf_to_dev(pf); 8945 struct ice_vsi *main_vsi; 8946 bool rem_adv_fltr = true; 8947 struct ice_channel *ch; 8948 struct ice_vsi *vsi; 8949 int tc_idx = 1; 8950 int i, err; 8951 8952 main_vsi = ice_get_main_vsi(pf); 8953 if (!main_vsi) 8954 return 0; 8955 8956 if (!test_bit(ICE_FLAG_TC_MQPRIO, pf->flags) || 8957 main_vsi->old_numtc == 1) 8958 return 0; /* nothing to be done */ 8959 8960 /* reconfigure main VSI based on old value of TC and cached values 8961 * for MQPRIO opts 8962 */ 8963 err = ice_vsi_cfg_tc(main_vsi, main_vsi->old_ena_tc); 8964 if (err) { 8965 dev_err(dev, "failed configuring TC(ena_tc:0x%02x) for HW VSI=%u\n", 8966 main_vsi->old_ena_tc, main_vsi->vsi_num); 8967 return err; 8968 } 8969 8970 /* rebuild ADQ VSIs */ 8971 ice_for_each_vsi(pf, i) { 8972 enum ice_vsi_type type; 8973 8974 vsi = pf->vsi[i]; 8975 if (!vsi || vsi->type != ICE_VSI_CHNL) 8976 continue; 8977 8978 type = vsi->type; 8979 8980 /* rebuild ADQ VSI */ 8981 err = ice_vsi_rebuild(vsi, ICE_VSI_FLAG_INIT); 8982 if (err) { 8983 dev_err(dev, "VSI (type:%s) at index %d rebuild failed, err %d\n", 8984 ice_vsi_type_str(type), vsi->idx, err); 8985 goto cleanup; 8986 } 8987 8988 /* Re-map HW VSI number, using VSI handle that has been 8989 * previously validated in ice_replay_vsi() call above 8990 */ 8991 vsi->vsi_num = ice_get_hw_vsi_num(&pf->hw, vsi->idx); 8992 8993 /* replay filters for the VSI */ 8994 err = ice_replay_vsi(&pf->hw, vsi->idx); 8995 if (err) { 8996 dev_err(dev, "VSI (type:%s) replay failed, err %d, VSI index %d\n", 8997 ice_vsi_type_str(type), err, vsi->idx); 8998 rem_adv_fltr = false; 8999 goto cleanup; 9000 } 9001 dev_info(dev, "VSI (type:%s) at index %d rebuilt successfully\n", 9002 ice_vsi_type_str(type), vsi->idx); 9003 9004 /* store ADQ VSI at correct TC index in main VSI's 9005 * map of TC to VSI 9006 */ 9007 main_vsi->tc_map_vsi[tc_idx++] = vsi; 9008 } 9009 9010 /* ADQ VSI(s) has been rebuilt successfully, so setup 9011 * channel for main VSI's Tx and Rx rings 9012 */ 9013 list_for_each_entry(ch, &main_vsi->ch_list, list) { 9014 struct ice_vsi *ch_vsi; 9015 9016 ch_vsi = ch->ch_vsi; 9017 if (!ch_vsi) 9018 continue; 9019 9020 /* reconfig channel resources */ 9021 ice_cfg_chnl_all_res(main_vsi, ch); 9022 9023 /* replay BW rate limit if it is non-zero */ 9024 if (!ch->max_tx_rate && !ch->min_tx_rate) 9025 continue; 9026 9027 err = ice_set_bw_limit(ch_vsi, ch->max_tx_rate, 9028 ch->min_tx_rate); 9029 if (err) 9030 dev_err(dev, "failed (err:%d) to rebuild BW rate limit, max_tx_rate: %llu Kbps, min_tx_rate: %llu Kbps for VSI(%u)\n", 9031 err, ch->max_tx_rate, ch->min_tx_rate, 9032 ch_vsi->vsi_num); 9033 else 9034 dev_dbg(dev, "successfully rebuild BW rate limit, max_tx_rate: %llu Kbps, min_tx_rate: %llu Kbps for VSI(%u)\n", 9035 ch->max_tx_rate, ch->min_tx_rate, 9036 ch_vsi->vsi_num); 9037 } 9038 9039 /* reconfig RSS for main VSI */ 9040 if (main_vsi->ch_rss_size) 9041 ice_vsi_cfg_rss_lut_key(main_vsi); 9042 9043 return 0; 9044 9045 cleanup: 9046 ice_remove_q_channels(main_vsi, rem_adv_fltr); 9047 return err; 9048 } 9049 9050 /** 9051 * ice_create_q_channels - Add queue channel for the given TCs 9052 * @vsi: VSI to be configured 9053 * 9054 * Configures queue channel mapping to the given TCs 9055 */ 9056 static int ice_create_q_channels(struct ice_vsi *vsi) 9057 { 9058 struct ice_pf *pf = vsi->back; 9059 struct ice_channel *ch; 9060 int ret = 0, i; 9061 9062 ice_for_each_chnl_tc(i) { 9063 if (!(vsi->all_enatc & BIT(i))) 9064 continue; 9065 9066 ch = kzalloc(sizeof(*ch), GFP_KERNEL); 9067 if (!ch) { 9068 ret = -ENOMEM; 9069 goto err_free; 9070 } 9071 INIT_LIST_HEAD(&ch->list); 9072 ch->num_rxq = vsi->mqprio_qopt.qopt.count[i]; 9073 ch->num_txq = vsi->mqprio_qopt.qopt.count[i]; 9074 ch->base_q = vsi->mqprio_qopt.qopt.offset[i]; 9075 ch->max_tx_rate = vsi->mqprio_qopt.max_rate[i]; 9076 ch->min_tx_rate = vsi->mqprio_qopt.min_rate[i]; 9077 9078 /* convert to Kbits/s */ 9079 if (ch->max_tx_rate) 9080 ch->max_tx_rate = div_u64(ch->max_tx_rate, 9081 ICE_BW_KBPS_DIVISOR); 9082 if (ch->min_tx_rate) 9083 ch->min_tx_rate = div_u64(ch->min_tx_rate, 9084 ICE_BW_KBPS_DIVISOR); 9085 9086 ret = ice_create_q_channel(vsi, ch); 9087 if (ret) { 9088 dev_err(ice_pf_to_dev(pf), 9089 "failed creating channel TC:%d\n", i); 9090 kfree(ch); 9091 goto err_free; 9092 } 9093 list_add_tail(&ch->list, &vsi->ch_list); 9094 vsi->tc_map_vsi[i] = ch->ch_vsi; 9095 dev_dbg(ice_pf_to_dev(pf), 9096 "successfully created channel: VSI %pK\n", ch->ch_vsi); 9097 } 9098 return 0; 9099 9100 err_free: 9101 ice_remove_q_channels(vsi, false); 9102 9103 return ret; 9104 } 9105 9106 /** 9107 * ice_setup_tc_mqprio_qdisc - configure multiple traffic classes 9108 * @netdev: net device to configure 9109 * @type_data: TC offload data 9110 */ 9111 static int ice_setup_tc_mqprio_qdisc(struct net_device *netdev, void *type_data) 9112 { 9113 struct tc_mqprio_qopt_offload *mqprio_qopt = type_data; 9114 struct ice_netdev_priv *np = netdev_priv(netdev); 9115 struct ice_vsi *vsi = np->vsi; 9116 struct ice_pf *pf = vsi->back; 9117 u16 mode, ena_tc_qdisc = 0; 9118 int cur_txq, cur_rxq; 9119 u8 hw = 0, num_tcf; 9120 struct device *dev; 9121 int ret, i; 9122 9123 dev = ice_pf_to_dev(pf); 9124 num_tcf = mqprio_qopt->qopt.num_tc; 9125 hw = mqprio_qopt->qopt.hw; 9126 mode = mqprio_qopt->mode; 9127 if (!hw) { 9128 clear_bit(ICE_FLAG_TC_MQPRIO, pf->flags); 9129 vsi->ch_rss_size = 0; 9130 memcpy(&vsi->mqprio_qopt, mqprio_qopt, sizeof(*mqprio_qopt)); 9131 goto config_tcf; 9132 } 9133 9134 /* Generate queue region map for number of TCF requested */ 9135 for (i = 0; i < num_tcf; i++) 9136 ena_tc_qdisc |= BIT(i); 9137 9138 switch (mode) { 9139 case TC_MQPRIO_MODE_CHANNEL: 9140 9141 if (pf->hw.port_info->is_custom_tx_enabled) { 9142 dev_err(dev, "Custom Tx scheduler feature enabled, can't configure ADQ\n"); 9143 return -EBUSY; 9144 } 9145 ice_tear_down_devlink_rate_tree(pf); 9146 9147 ret = ice_validate_mqprio_qopt(vsi, mqprio_qopt); 9148 if (ret) { 9149 netdev_err(netdev, "failed to validate_mqprio_qopt(), ret %d\n", 9150 ret); 9151 return ret; 9152 } 9153 memcpy(&vsi->mqprio_qopt, mqprio_qopt, sizeof(*mqprio_qopt)); 9154 set_bit(ICE_FLAG_TC_MQPRIO, pf->flags); 9155 /* don't assume state of hw_tc_offload during driver load 9156 * and set the flag for TC flower filter if hw_tc_offload 9157 * already ON 9158 */ 9159 if (vsi->netdev->features & NETIF_F_HW_TC) 9160 set_bit(ICE_FLAG_CLS_FLOWER, pf->flags); 9161 break; 9162 default: 9163 return -EINVAL; 9164 } 9165 9166 config_tcf: 9167 9168 /* Requesting same TCF configuration as already enabled */ 9169 if (ena_tc_qdisc == vsi->tc_cfg.ena_tc && 9170 mode != TC_MQPRIO_MODE_CHANNEL) 9171 return 0; 9172 9173 /* Pause VSI queues */ 9174 ice_dis_vsi(vsi, true); 9175 9176 if (!hw && !test_bit(ICE_FLAG_TC_MQPRIO, pf->flags)) 9177 ice_remove_q_channels(vsi, true); 9178 9179 if (!hw && !test_bit(ICE_FLAG_TC_MQPRIO, pf->flags)) { 9180 vsi->req_txq = min_t(int, ice_get_avail_txq_count(pf), 9181 num_online_cpus()); 9182 vsi->req_rxq = min_t(int, ice_get_avail_rxq_count(pf), 9183 num_online_cpus()); 9184 } else { 9185 /* logic to rebuild VSI, same like ethtool -L */ 9186 u16 offset = 0, qcount_tx = 0, qcount_rx = 0; 9187 9188 for (i = 0; i < num_tcf; i++) { 9189 if (!(ena_tc_qdisc & BIT(i))) 9190 continue; 9191 9192 offset = vsi->mqprio_qopt.qopt.offset[i]; 9193 qcount_rx = vsi->mqprio_qopt.qopt.count[i]; 9194 qcount_tx = vsi->mqprio_qopt.qopt.count[i]; 9195 } 9196 vsi->req_txq = offset + qcount_tx; 9197 vsi->req_rxq = offset + qcount_rx; 9198 9199 /* store away original rss_size info, so that it gets reused 9200 * form ice_vsi_rebuild during tc-qdisc delete stage - to 9201 * determine, what should be the rss_sizefor main VSI 9202 */ 9203 vsi->orig_rss_size = vsi->rss_size; 9204 } 9205 9206 /* save current values of Tx and Rx queues before calling VSI rebuild 9207 * for fallback option 9208 */ 9209 cur_txq = vsi->num_txq; 9210 cur_rxq = vsi->num_rxq; 9211 9212 /* proceed with rebuild main VSI using correct number of queues */ 9213 ret = ice_vsi_rebuild(vsi, ICE_VSI_FLAG_NO_INIT); 9214 if (ret) { 9215 /* fallback to current number of queues */ 9216 dev_info(dev, "Rebuild failed with new queues, try with current number of queues\n"); 9217 vsi->req_txq = cur_txq; 9218 vsi->req_rxq = cur_rxq; 9219 clear_bit(ICE_RESET_FAILED, pf->state); 9220 if (ice_vsi_rebuild(vsi, ICE_VSI_FLAG_NO_INIT)) { 9221 dev_err(dev, "Rebuild of main VSI failed again\n"); 9222 return ret; 9223 } 9224 } 9225 9226 vsi->all_numtc = num_tcf; 9227 vsi->all_enatc = ena_tc_qdisc; 9228 ret = ice_vsi_cfg_tc(vsi, ena_tc_qdisc); 9229 if (ret) { 9230 netdev_err(netdev, "failed configuring TC for VSI id=%d\n", 9231 vsi->vsi_num); 9232 goto exit; 9233 } 9234 9235 if (test_bit(ICE_FLAG_TC_MQPRIO, pf->flags)) { 9236 u64 max_tx_rate = vsi->mqprio_qopt.max_rate[0]; 9237 u64 min_tx_rate = vsi->mqprio_qopt.min_rate[0]; 9238 9239 /* set TC0 rate limit if specified */ 9240 if (max_tx_rate || min_tx_rate) { 9241 /* convert to Kbits/s */ 9242 if (max_tx_rate) 9243 max_tx_rate = div_u64(max_tx_rate, ICE_BW_KBPS_DIVISOR); 9244 if (min_tx_rate) 9245 min_tx_rate = div_u64(min_tx_rate, ICE_BW_KBPS_DIVISOR); 9246 9247 ret = ice_set_bw_limit(vsi, max_tx_rate, min_tx_rate); 9248 if (!ret) { 9249 dev_dbg(dev, "set Tx rate max %llu min %llu for VSI(%u)\n", 9250 max_tx_rate, min_tx_rate, vsi->vsi_num); 9251 } else { 9252 dev_err(dev, "failed to set Tx rate max %llu min %llu for VSI(%u)\n", 9253 max_tx_rate, min_tx_rate, vsi->vsi_num); 9254 goto exit; 9255 } 9256 } 9257 ret = ice_create_q_channels(vsi); 9258 if (ret) { 9259 netdev_err(netdev, "failed configuring queue channels\n"); 9260 goto exit; 9261 } else { 9262 netdev_dbg(netdev, "successfully configured channels\n"); 9263 } 9264 } 9265 9266 if (vsi->ch_rss_size) 9267 ice_vsi_cfg_rss_lut_key(vsi); 9268 9269 exit: 9270 /* if error, reset the all_numtc and all_enatc */ 9271 if (ret) { 9272 vsi->all_numtc = 0; 9273 vsi->all_enatc = 0; 9274 } 9275 /* resume VSI */ 9276 ice_ena_vsi(vsi, true); 9277 9278 return ret; 9279 } 9280 9281 static LIST_HEAD(ice_block_cb_list); 9282 9283 static int 9284 ice_setup_tc(struct net_device *netdev, enum tc_setup_type type, 9285 void *type_data) 9286 { 9287 struct ice_netdev_priv *np = netdev_priv(netdev); 9288 struct ice_pf *pf = np->vsi->back; 9289 bool locked = false; 9290 int err; 9291 9292 switch (type) { 9293 case TC_SETUP_BLOCK: 9294 return flow_block_cb_setup_simple(type_data, 9295 &ice_block_cb_list, 9296 ice_setup_tc_block_cb, 9297 np, np, true); 9298 case TC_SETUP_QDISC_MQPRIO: 9299 if (ice_is_eswitch_mode_switchdev(pf)) { 9300 netdev_err(netdev, "TC MQPRIO offload not supported, switchdev is enabled\n"); 9301 return -EOPNOTSUPP; 9302 } 9303 9304 if (pf->adev) { 9305 mutex_lock(&pf->adev_mutex); 9306 device_lock(&pf->adev->dev); 9307 locked = true; 9308 if (pf->adev->dev.driver) { 9309 netdev_err(netdev, "Cannot change qdisc when RDMA is active\n"); 9310 err = -EBUSY; 9311 goto adev_unlock; 9312 } 9313 } 9314 9315 /* setup traffic classifier for receive side */ 9316 mutex_lock(&pf->tc_mutex); 9317 err = ice_setup_tc_mqprio_qdisc(netdev, type_data); 9318 mutex_unlock(&pf->tc_mutex); 9319 9320 adev_unlock: 9321 if (locked) { 9322 device_unlock(&pf->adev->dev); 9323 mutex_unlock(&pf->adev_mutex); 9324 } 9325 return err; 9326 default: 9327 return -EOPNOTSUPP; 9328 } 9329 return -EOPNOTSUPP; 9330 } 9331 9332 static struct ice_indr_block_priv * 9333 ice_indr_block_priv_lookup(struct ice_netdev_priv *np, 9334 struct net_device *netdev) 9335 { 9336 struct ice_indr_block_priv *cb_priv; 9337 9338 list_for_each_entry(cb_priv, &np->tc_indr_block_priv_list, list) { 9339 if (!cb_priv->netdev) 9340 return NULL; 9341 if (cb_priv->netdev == netdev) 9342 return cb_priv; 9343 } 9344 return NULL; 9345 } 9346 9347 static int 9348 ice_indr_setup_block_cb(enum tc_setup_type type, void *type_data, 9349 void *indr_priv) 9350 { 9351 struct ice_indr_block_priv *priv = indr_priv; 9352 struct ice_netdev_priv *np = priv->np; 9353 9354 switch (type) { 9355 case TC_SETUP_CLSFLOWER: 9356 return ice_setup_tc_cls_flower(np, priv->netdev, 9357 (struct flow_cls_offload *) 9358 type_data); 9359 default: 9360 return -EOPNOTSUPP; 9361 } 9362 } 9363 9364 static int 9365 ice_indr_setup_tc_block(struct net_device *netdev, struct Qdisc *sch, 9366 struct ice_netdev_priv *np, 9367 struct flow_block_offload *f, void *data, 9368 void (*cleanup)(struct flow_block_cb *block_cb)) 9369 { 9370 struct ice_indr_block_priv *indr_priv; 9371 struct flow_block_cb *block_cb; 9372 9373 if (!ice_is_tunnel_supported(netdev) && 9374 !(is_vlan_dev(netdev) && 9375 vlan_dev_real_dev(netdev) == np->vsi->netdev)) 9376 return -EOPNOTSUPP; 9377 9378 if (f->binder_type != FLOW_BLOCK_BINDER_TYPE_CLSACT_INGRESS) 9379 return -EOPNOTSUPP; 9380 9381 switch (f->command) { 9382 case FLOW_BLOCK_BIND: 9383 indr_priv = ice_indr_block_priv_lookup(np, netdev); 9384 if (indr_priv) 9385 return -EEXIST; 9386 9387 indr_priv = kzalloc(sizeof(*indr_priv), GFP_KERNEL); 9388 if (!indr_priv) 9389 return -ENOMEM; 9390 9391 indr_priv->netdev = netdev; 9392 indr_priv->np = np; 9393 list_add(&indr_priv->list, &np->tc_indr_block_priv_list); 9394 9395 block_cb = 9396 flow_indr_block_cb_alloc(ice_indr_setup_block_cb, 9397 indr_priv, indr_priv, 9398 ice_rep_indr_tc_block_unbind, 9399 f, netdev, sch, data, np, 9400 cleanup); 9401 9402 if (IS_ERR(block_cb)) { 9403 list_del(&indr_priv->list); 9404 kfree(indr_priv); 9405 return PTR_ERR(block_cb); 9406 } 9407 flow_block_cb_add(block_cb, f); 9408 list_add_tail(&block_cb->driver_list, &ice_block_cb_list); 9409 break; 9410 case FLOW_BLOCK_UNBIND: 9411 indr_priv = ice_indr_block_priv_lookup(np, netdev); 9412 if (!indr_priv) 9413 return -ENOENT; 9414 9415 block_cb = flow_block_cb_lookup(f->block, 9416 ice_indr_setup_block_cb, 9417 indr_priv); 9418 if (!block_cb) 9419 return -ENOENT; 9420 9421 flow_indr_block_cb_remove(block_cb, f); 9422 9423 list_del(&block_cb->driver_list); 9424 break; 9425 default: 9426 return -EOPNOTSUPP; 9427 } 9428 return 0; 9429 } 9430 9431 static int 9432 ice_indr_setup_tc_cb(struct net_device *netdev, struct Qdisc *sch, 9433 void *cb_priv, enum tc_setup_type type, void *type_data, 9434 void *data, 9435 void (*cleanup)(struct flow_block_cb *block_cb)) 9436 { 9437 switch (type) { 9438 case TC_SETUP_BLOCK: 9439 return ice_indr_setup_tc_block(netdev, sch, cb_priv, type_data, 9440 data, cleanup); 9441 9442 default: 9443 return -EOPNOTSUPP; 9444 } 9445 } 9446 9447 /** 9448 * ice_open - Called when a network interface becomes active 9449 * @netdev: network interface device structure 9450 * 9451 * The open entry point is called when a network interface is made 9452 * active by the system (IFF_UP). At this point all resources needed 9453 * for transmit and receive operations are allocated, the interrupt 9454 * handler is registered with the OS, the netdev watchdog is enabled, 9455 * and the stack is notified that the interface is ready. 9456 * 9457 * Returns 0 on success, negative value on failure 9458 */ 9459 int ice_open(struct net_device *netdev) 9460 { 9461 struct ice_netdev_priv *np = netdev_priv(netdev); 9462 struct ice_pf *pf = np->vsi->back; 9463 9464 if (ice_is_reset_in_progress(pf->state)) { 9465 netdev_err(netdev, "can't open net device while reset is in progress"); 9466 return -EBUSY; 9467 } 9468 9469 return ice_open_internal(netdev); 9470 } 9471 9472 /** 9473 * ice_open_internal - Called when a network interface becomes active 9474 * @netdev: network interface device structure 9475 * 9476 * Internal ice_open implementation. Should not be used directly except for ice_open and reset 9477 * handling routine 9478 * 9479 * Returns 0 on success, negative value on failure 9480 */ 9481 int ice_open_internal(struct net_device *netdev) 9482 { 9483 struct ice_netdev_priv *np = netdev_priv(netdev); 9484 struct ice_vsi *vsi = np->vsi; 9485 struct ice_pf *pf = vsi->back; 9486 struct ice_port_info *pi; 9487 int err; 9488 9489 if (test_bit(ICE_NEEDS_RESTART, pf->state)) { 9490 netdev_err(netdev, "driver needs to be unloaded and reloaded\n"); 9491 return -EIO; 9492 } 9493 9494 netif_carrier_off(netdev); 9495 9496 pi = vsi->port_info; 9497 err = ice_update_link_info(pi); 9498 if (err) { 9499 netdev_err(netdev, "Failed to get link info, error %d\n", err); 9500 return err; 9501 } 9502 9503 ice_check_link_cfg_err(pf, pi->phy.link_info.link_cfg_err); 9504 9505 /* Set PHY if there is media, otherwise, turn off PHY */ 9506 if (pi->phy.link_info.link_info & ICE_AQ_MEDIA_AVAILABLE) { 9507 clear_bit(ICE_FLAG_NO_MEDIA, pf->flags); 9508 if (!test_bit(ICE_PHY_INIT_COMPLETE, pf->state)) { 9509 err = ice_init_phy_user_cfg(pi); 9510 if (err) { 9511 netdev_err(netdev, "Failed to initialize PHY settings, error %d\n", 9512 err); 9513 return err; 9514 } 9515 } 9516 9517 err = ice_configure_phy(vsi); 9518 if (err) { 9519 netdev_err(netdev, "Failed to set physical link up, error %d\n", 9520 err); 9521 return err; 9522 } 9523 } else { 9524 set_bit(ICE_FLAG_NO_MEDIA, pf->flags); 9525 ice_set_link(vsi, false); 9526 } 9527 9528 err = ice_vsi_open(vsi); 9529 if (err) 9530 netdev_err(netdev, "Failed to open VSI 0x%04X on switch 0x%04X\n", 9531 vsi->vsi_num, vsi->vsw->sw_id); 9532 9533 /* Update existing tunnels information */ 9534 udp_tunnel_get_rx_info(netdev); 9535 9536 return err; 9537 } 9538 9539 /** 9540 * ice_stop - Disables a network interface 9541 * @netdev: network interface device structure 9542 * 9543 * The stop entry point is called when an interface is de-activated by the OS, 9544 * and the netdevice enters the DOWN state. The hardware is still under the 9545 * driver's control, but the netdev interface is disabled. 9546 * 9547 * Returns success only - not allowed to fail 9548 */ 9549 int ice_stop(struct net_device *netdev) 9550 { 9551 struct ice_netdev_priv *np = netdev_priv(netdev); 9552 struct ice_vsi *vsi = np->vsi; 9553 struct ice_pf *pf = vsi->back; 9554 9555 if (ice_is_reset_in_progress(pf->state)) { 9556 netdev_err(netdev, "can't stop net device while reset is in progress"); 9557 return -EBUSY; 9558 } 9559 9560 if (test_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, vsi->back->flags)) { 9561 int link_err = ice_force_phys_link_state(vsi, false); 9562 9563 if (link_err) { 9564 if (link_err == -ENOMEDIUM) 9565 netdev_info(vsi->netdev, "Skipping link reconfig - no media attached, VSI %d\n", 9566 vsi->vsi_num); 9567 else 9568 netdev_err(vsi->netdev, "Failed to set physical link down, VSI %d error %d\n", 9569 vsi->vsi_num, link_err); 9570 9571 ice_vsi_close(vsi); 9572 return -EIO; 9573 } 9574 } 9575 9576 ice_vsi_close(vsi); 9577 9578 return 0; 9579 } 9580 9581 /** 9582 * ice_features_check - Validate encapsulated packet conforms to limits 9583 * @skb: skb buffer 9584 * @netdev: This port's netdev 9585 * @features: Offload features that the stack believes apply 9586 */ 9587 static netdev_features_t 9588 ice_features_check(struct sk_buff *skb, 9589 struct net_device __always_unused *netdev, 9590 netdev_features_t features) 9591 { 9592 bool gso = skb_is_gso(skb); 9593 size_t len; 9594 9595 /* No point in doing any of this if neither checksum nor GSO are 9596 * being requested for this frame. We can rule out both by just 9597 * checking for CHECKSUM_PARTIAL 9598 */ 9599 if (skb->ip_summed != CHECKSUM_PARTIAL) 9600 return features; 9601 9602 /* We cannot support GSO if the MSS is going to be less than 9603 * 64 bytes. If it is then we need to drop support for GSO. 9604 */ 9605 if (gso && (skb_shinfo(skb)->gso_size < ICE_TXD_CTX_MIN_MSS)) 9606 features &= ~NETIF_F_GSO_MASK; 9607 9608 len = skb_network_offset(skb); 9609 if (len > ICE_TXD_MACLEN_MAX || len & 0x1) 9610 goto out_rm_features; 9611 9612 len = skb_network_header_len(skb); 9613 if (len > ICE_TXD_IPLEN_MAX || len & 0x1) 9614 goto out_rm_features; 9615 9616 if (skb->encapsulation) { 9617 /* this must work for VXLAN frames AND IPIP/SIT frames, and in 9618 * the case of IPIP frames, the transport header pointer is 9619 * after the inner header! So check to make sure that this 9620 * is a GRE or UDP_TUNNEL frame before doing that math. 9621 */ 9622 if (gso && (skb_shinfo(skb)->gso_type & 9623 (SKB_GSO_GRE | SKB_GSO_UDP_TUNNEL))) { 9624 len = skb_inner_network_header(skb) - 9625 skb_transport_header(skb); 9626 if (len > ICE_TXD_L4LEN_MAX || len & 0x1) 9627 goto out_rm_features; 9628 } 9629 9630 len = skb_inner_network_header_len(skb); 9631 if (len > ICE_TXD_IPLEN_MAX || len & 0x1) 9632 goto out_rm_features; 9633 } 9634 9635 return features; 9636 out_rm_features: 9637 return features & ~(NETIF_F_CSUM_MASK | NETIF_F_GSO_MASK); 9638 } 9639 9640 static const struct net_device_ops ice_netdev_safe_mode_ops = { 9641 .ndo_open = ice_open, 9642 .ndo_stop = ice_stop, 9643 .ndo_start_xmit = ice_start_xmit, 9644 .ndo_set_mac_address = ice_set_mac_address, 9645 .ndo_validate_addr = eth_validate_addr, 9646 .ndo_change_mtu = ice_change_mtu, 9647 .ndo_get_stats64 = ice_get_stats64, 9648 .ndo_tx_timeout = ice_tx_timeout, 9649 .ndo_bpf = ice_xdp_safe_mode, 9650 }; 9651 9652 static const struct net_device_ops ice_netdev_ops = { 9653 .ndo_open = ice_open, 9654 .ndo_stop = ice_stop, 9655 .ndo_start_xmit = ice_start_xmit, 9656 .ndo_select_queue = ice_select_queue, 9657 .ndo_features_check = ice_features_check, 9658 .ndo_fix_features = ice_fix_features, 9659 .ndo_set_rx_mode = ice_set_rx_mode, 9660 .ndo_set_mac_address = ice_set_mac_address, 9661 .ndo_validate_addr = eth_validate_addr, 9662 .ndo_change_mtu = ice_change_mtu, 9663 .ndo_get_stats64 = ice_get_stats64, 9664 .ndo_set_tx_maxrate = ice_set_tx_maxrate, 9665 .ndo_eth_ioctl = ice_eth_ioctl, 9666 .ndo_set_vf_spoofchk = ice_set_vf_spoofchk, 9667 .ndo_set_vf_mac = ice_set_vf_mac, 9668 .ndo_get_vf_config = ice_get_vf_cfg, 9669 .ndo_set_vf_trust = ice_set_vf_trust, 9670 .ndo_set_vf_vlan = ice_set_vf_port_vlan, 9671 .ndo_set_vf_link_state = ice_set_vf_link_state, 9672 .ndo_get_vf_stats = ice_get_vf_stats, 9673 .ndo_set_vf_rate = ice_set_vf_bw, 9674 .ndo_vlan_rx_add_vid = ice_vlan_rx_add_vid, 9675 .ndo_vlan_rx_kill_vid = ice_vlan_rx_kill_vid, 9676 .ndo_setup_tc = ice_setup_tc, 9677 .ndo_set_features = ice_set_features, 9678 .ndo_bridge_getlink = ice_bridge_getlink, 9679 .ndo_bridge_setlink = ice_bridge_setlink, 9680 .ndo_fdb_add = ice_fdb_add, 9681 .ndo_fdb_del = ice_fdb_del, 9682 #ifdef CONFIG_RFS_ACCEL 9683 .ndo_rx_flow_steer = ice_rx_flow_steer, 9684 #endif 9685 .ndo_tx_timeout = ice_tx_timeout, 9686 .ndo_bpf = ice_xdp, 9687 .ndo_xdp_xmit = ice_xdp_xmit, 9688 .ndo_xsk_wakeup = ice_xsk_wakeup, 9689 }; 9690