| 1 | // SPDX-License-Identifier: GPL-2.0 |
| 2 | /* Copyright (c) 2018, Intel Corporation. */ |
| 3 | |
| 4 | /* Intel(R) Ethernet Connection E800 Series Linux Driver */ |
| 5 | |
| 6 | #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt |
| 7 | |
| 8 | #include "ice.h" |
| 9 | #include "ice_lib.h" |
| 10 | #include "ice_dcb_lib.h" |
| 11 | |
| 12 | #define DRV_VERSION "0.7.4-k" |
| 13 | #define DRV_SUMMARY "Intel(R) Ethernet Connection E800 Series Linux Driver" |
| 14 | const char ice_drv_ver[] = DRV_VERSION; |
| 15 | static const char ice_driver_string[] = DRV_SUMMARY; |
| 16 | static const char ice_copyright[] = "Copyright (c) 2018, Intel Corporation."; |
| 17 | |
| 18 | MODULE_AUTHOR("Intel Corporation, <linux.nics@intel.com>"); |
| 19 | MODULE_DESCRIPTION(DRV_SUMMARY); |
| 20 | MODULE_LICENSE("GPL v2"); |
| 21 | MODULE_VERSION(DRV_VERSION); |
| 22 | |
| 23 | static int debug = -1; |
| 24 | module_param(debug, int, 0644); |
| 25 | #ifndef CONFIG_DYNAMIC_DEBUG |
| 26 | MODULE_PARM_DESC(debug, "netif level (0=none,...,16=all), hw debug_mask (0x8XXXXXXX)"); |
| 27 | #else |
| 28 | MODULE_PARM_DESC(debug, "netif level (0=none,...,16=all)"); |
| 29 | #endif /* !CONFIG_DYNAMIC_DEBUG */ |
| 30 | |
| 31 | static struct workqueue_struct *ice_wq; |
| 32 | static const struct net_device_ops ice_netdev_ops; |
| 33 | |
| 34 | static void ice_rebuild(struct ice_pf *pf); |
| 35 | |
| 36 | static void ice_vsi_release_all(struct ice_pf *pf); |
| 37 | static void ice_update_vsi_stats(struct ice_vsi *vsi); |
| 38 | static void ice_update_pf_stats(struct ice_pf *pf); |
| 39 | |
| 40 | /** |
| 41 | * ice_get_tx_pending - returns number of Tx descriptors not processed |
| 42 | * @ring: the ring of descriptors |
| 43 | */ |
| 44 | static u32 ice_get_tx_pending(struct ice_ring *ring) |
| 45 | { |
| 46 | u32 head, tail; |
| 47 | |
| 48 | head = ring->next_to_clean; |
| 49 | tail = readl(ring->tail); |
| 50 | |
| 51 | if (head != tail) |
| 52 | return (head < tail) ? |
| 53 | tail - head : (tail + ring->count - head); |
| 54 | return 0; |
| 55 | } |
| 56 | |
| 57 | /** |
| 58 | * ice_check_for_hang_subtask - check for and recover hung queues |
| 59 | * @pf: pointer to PF struct |
| 60 | */ |
| 61 | static void ice_check_for_hang_subtask(struct ice_pf *pf) |
| 62 | { |
| 63 | struct ice_vsi *vsi = NULL; |
| 64 | unsigned int i; |
| 65 | u32 v, v_idx; |
| 66 | int packets; |
| 67 | |
| 68 | ice_for_each_vsi(pf, v) |
| 69 | if (pf->vsi[v] && pf->vsi[v]->type == ICE_VSI_PF) { |
| 70 | vsi = pf->vsi[v]; |
| 71 | break; |
| 72 | } |
| 73 | |
| 74 | if (!vsi || test_bit(__ICE_DOWN, vsi->state)) |
| 75 | return; |
| 76 | |
| 77 | if (!(vsi->netdev && netif_carrier_ok(vsi->netdev))) |
| 78 | return; |
| 79 | |
| 80 | for (i = 0; i < vsi->num_txq; i++) { |
| 81 | struct ice_ring *tx_ring = vsi->tx_rings[i]; |
| 82 | |
| 83 | if (tx_ring && tx_ring->desc) { |
| 84 | int itr = ICE_ITR_NONE; |
| 85 | |
| 86 | /* If packet counter has not changed the queue is |
| 87 | * likely stalled, so force an interrupt for this |
| 88 | * queue. |
| 89 | * |
| 90 | * prev_pkt would be negative if there was no |
| 91 | * pending work. |
| 92 | */ |
| 93 | packets = tx_ring->stats.pkts & INT_MAX; |
| 94 | if (tx_ring->tx_stats.prev_pkt == packets) { |
| 95 | /* Trigger sw interrupt to revive the queue */ |
| 96 | v_idx = tx_ring->q_vector->v_idx; |
| 97 | wr32(&vsi->back->hw, |
| 98 | GLINT_DYN_CTL(vsi->hw_base_vector + v_idx), |
| 99 | (itr << GLINT_DYN_CTL_ITR_INDX_S) | |
| 100 | GLINT_DYN_CTL_SWINT_TRIG_M | |
| 101 | GLINT_DYN_CTL_INTENA_MSK_M); |
| 102 | continue; |
| 103 | } |
| 104 | |
| 105 | /* Memory barrier between read of packet count and call |
| 106 | * to ice_get_tx_pending() |
| 107 | */ |
| 108 | smp_rmb(); |
| 109 | tx_ring->tx_stats.prev_pkt = |
| 110 | ice_get_tx_pending(tx_ring) ? packets : -1; |
| 111 | } |
| 112 | } |
| 113 | } |
| 114 | |
| 115 | /** |
| 116 | * ice_add_mac_to_sync_list - creates list of MAC addresses to be synced |
| 117 | * @netdev: the net device on which the sync is happening |
| 118 | * @addr: MAC address to sync |
| 119 | * |
| 120 | * This is a callback function which is called by the in kernel device sync |
| 121 | * functions (like __dev_uc_sync, __dev_mc_sync, etc). This function only |
| 122 | * populates the tmp_sync_list, which is later used by ice_add_mac to add the |
| 123 | * MAC filters from the hardware. |
| 124 | */ |
| 125 | static int ice_add_mac_to_sync_list(struct net_device *netdev, const u8 *addr) |
| 126 | { |
| 127 | struct ice_netdev_priv *np = netdev_priv(netdev); |
| 128 | struct ice_vsi *vsi = np->vsi; |
| 129 | |
| 130 | if (ice_add_mac_to_list(vsi, &vsi->tmp_sync_list, addr)) |
| 131 | return -EINVAL; |
| 132 | |
| 133 | return 0; |
| 134 | } |
| 135 | |
| 136 | /** |
| 137 | * ice_add_mac_to_unsync_list - creates list of MAC addresses to be unsynced |
| 138 | * @netdev: the net device on which the unsync is happening |
| 139 | * @addr: MAC address to unsync |
| 140 | * |
| 141 | * This is a callback function which is called by the in kernel device unsync |
| 142 | * functions (like __dev_uc_unsync, __dev_mc_unsync, etc). This function only |
| 143 | * populates the tmp_unsync_list, which is later used by ice_remove_mac to |
| 144 | * delete the MAC filters from the hardware. |
| 145 | */ |
| 146 | static int ice_add_mac_to_unsync_list(struct net_device *netdev, const u8 *addr) |
| 147 | { |
| 148 | struct ice_netdev_priv *np = netdev_priv(netdev); |
| 149 | struct ice_vsi *vsi = np->vsi; |
| 150 | |
| 151 | if (ice_add_mac_to_list(vsi, &vsi->tmp_unsync_list, addr)) |
| 152 | return -EINVAL; |
| 153 | |
| 154 | return 0; |
| 155 | } |
| 156 | |
| 157 | /** |
| 158 | * ice_vsi_fltr_changed - check if filter state changed |
| 159 | * @vsi: VSI to be checked |
| 160 | * |
| 161 | * returns true if filter state has changed, false otherwise. |
| 162 | */ |
| 163 | static bool ice_vsi_fltr_changed(struct ice_vsi *vsi) |
| 164 | { |
| 165 | return test_bit(ICE_VSI_FLAG_UMAC_FLTR_CHANGED, vsi->flags) || |
| 166 | test_bit(ICE_VSI_FLAG_MMAC_FLTR_CHANGED, vsi->flags) || |
| 167 | test_bit(ICE_VSI_FLAG_VLAN_FLTR_CHANGED, vsi->flags); |
| 168 | } |
| 169 | |
| 170 | /** |
| 171 | * ice_cfg_promisc - Enable or disable promiscuous mode for a given PF |
| 172 | * @vsi: the VSI being configured |
| 173 | * @promisc_m: mask of promiscuous config bits |
| 174 | * @set_promisc: enable or disable promisc flag request |
| 175 | * |
| 176 | */ |
| 177 | static int ice_cfg_promisc(struct ice_vsi *vsi, u8 promisc_m, bool set_promisc) |
| 178 | { |
| 179 | struct ice_hw *hw = &vsi->back->hw; |
| 180 | enum ice_status status = 0; |
| 181 | |
| 182 | if (vsi->type != ICE_VSI_PF) |
| 183 | return 0; |
| 184 | |
| 185 | if (vsi->vlan_ena) { |
| 186 | status = ice_set_vlan_vsi_promisc(hw, vsi->idx, promisc_m, |
| 187 | set_promisc); |
| 188 | } else { |
| 189 | if (set_promisc) |
| 190 | status = ice_set_vsi_promisc(hw, vsi->idx, promisc_m, |
| 191 | 0); |
| 192 | else |
| 193 | status = ice_clear_vsi_promisc(hw, vsi->idx, promisc_m, |
| 194 | 0); |
| 195 | } |
| 196 | |
| 197 | if (status) |
| 198 | return -EIO; |
| 199 | |
| 200 | return 0; |
| 201 | } |
| 202 | |
| 203 | /** |
| 204 | * ice_vsi_sync_fltr - Update the VSI filter list to the HW |
| 205 | * @vsi: ptr to the VSI |
| 206 | * |
| 207 | * Push any outstanding VSI filter changes through the AdminQ. |
| 208 | */ |
| 209 | static int ice_vsi_sync_fltr(struct ice_vsi *vsi) |
| 210 | { |
| 211 | struct device *dev = &vsi->back->pdev->dev; |
| 212 | struct net_device *netdev = vsi->netdev; |
| 213 | bool promisc_forced_on = false; |
| 214 | struct ice_pf *pf = vsi->back; |
| 215 | struct ice_hw *hw = &pf->hw; |
| 216 | enum ice_status status = 0; |
| 217 | u32 changed_flags = 0; |
| 218 | u8 promisc_m; |
| 219 | int err = 0; |
| 220 | |
| 221 | if (!vsi->netdev) |
| 222 | return -EINVAL; |
| 223 | |
| 224 | while (test_and_set_bit(__ICE_CFG_BUSY, vsi->state)) |
| 225 | usleep_range(1000, 2000); |
| 226 | |
| 227 | changed_flags = vsi->current_netdev_flags ^ vsi->netdev->flags; |
| 228 | vsi->current_netdev_flags = vsi->netdev->flags; |
| 229 | |
| 230 | INIT_LIST_HEAD(&vsi->tmp_sync_list); |
| 231 | INIT_LIST_HEAD(&vsi->tmp_unsync_list); |
| 232 | |
| 233 | if (ice_vsi_fltr_changed(vsi)) { |
| 234 | clear_bit(ICE_VSI_FLAG_UMAC_FLTR_CHANGED, vsi->flags); |
| 235 | clear_bit(ICE_VSI_FLAG_MMAC_FLTR_CHANGED, vsi->flags); |
| 236 | clear_bit(ICE_VSI_FLAG_VLAN_FLTR_CHANGED, vsi->flags); |
| 237 | |
| 238 | /* grab the netdev's addr_list_lock */ |
| 239 | netif_addr_lock_bh(netdev); |
| 240 | __dev_uc_sync(netdev, ice_add_mac_to_sync_list, |
| 241 | ice_add_mac_to_unsync_list); |
| 242 | __dev_mc_sync(netdev, ice_add_mac_to_sync_list, |
| 243 | ice_add_mac_to_unsync_list); |
| 244 | /* our temp lists are populated. release lock */ |
| 245 | netif_addr_unlock_bh(netdev); |
| 246 | } |
| 247 | |
| 248 | /* Remove MAC addresses in the unsync list */ |
| 249 | status = ice_remove_mac(hw, &vsi->tmp_unsync_list); |
| 250 | ice_free_fltr_list(dev, &vsi->tmp_unsync_list); |
| 251 | if (status) { |
| 252 | netdev_err(netdev, "Failed to delete MAC filters\n"); |
| 253 | /* if we failed because of alloc failures, just bail */ |
| 254 | if (status == ICE_ERR_NO_MEMORY) { |
| 255 | err = -ENOMEM; |
| 256 | goto out; |
| 257 | } |
| 258 | } |
| 259 | |
| 260 | /* Add MAC addresses in the sync list */ |
| 261 | status = ice_add_mac(hw, &vsi->tmp_sync_list); |
| 262 | ice_free_fltr_list(dev, &vsi->tmp_sync_list); |
| 263 | /* If filter is added successfully or already exists, do not go into |
| 264 | * 'if' condition and report it as error. Instead continue processing |
| 265 | * rest of the function. |
| 266 | */ |
| 267 | if (status && status != ICE_ERR_ALREADY_EXISTS) { |
| 268 | netdev_err(netdev, "Failed to add MAC filters\n"); |
| 269 | /* If there is no more space for new umac filters, VSI |
| 270 | * should go into promiscuous mode. There should be some |
| 271 | * space reserved for promiscuous filters. |
| 272 | */ |
| 273 | if (hw->adminq.sq_last_status == ICE_AQ_RC_ENOSPC && |
| 274 | !test_and_set_bit(__ICE_FLTR_OVERFLOW_PROMISC, |
| 275 | vsi->state)) { |
| 276 | promisc_forced_on = true; |
| 277 | netdev_warn(netdev, |
| 278 | "Reached MAC filter limit, forcing promisc mode on VSI %d\n", |
| 279 | vsi->vsi_num); |
| 280 | } else { |
| 281 | err = -EIO; |
| 282 | goto out; |
| 283 | } |
| 284 | } |
| 285 | /* check for changes in promiscuous modes */ |
| 286 | if (changed_flags & IFF_ALLMULTI) { |
| 287 | if (vsi->current_netdev_flags & IFF_ALLMULTI) { |
| 288 | if (vsi->vlan_ena) |
| 289 | promisc_m = ICE_MCAST_VLAN_PROMISC_BITS; |
| 290 | else |
| 291 | promisc_m = ICE_MCAST_PROMISC_BITS; |
| 292 | |
| 293 | err = ice_cfg_promisc(vsi, promisc_m, true); |
| 294 | if (err) { |
| 295 | netdev_err(netdev, "Error setting Multicast promiscuous mode on VSI %i\n", |
| 296 | vsi->vsi_num); |
| 297 | vsi->current_netdev_flags &= ~IFF_ALLMULTI; |
| 298 | goto out_promisc; |
| 299 | } |
| 300 | } else if (!(vsi->current_netdev_flags & IFF_ALLMULTI)) { |
| 301 | if (vsi->vlan_ena) |
| 302 | promisc_m = ICE_MCAST_VLAN_PROMISC_BITS; |
| 303 | else |
| 304 | promisc_m = ICE_MCAST_PROMISC_BITS; |
| 305 | |
| 306 | err = ice_cfg_promisc(vsi, promisc_m, false); |
| 307 | if (err) { |
| 308 | netdev_err(netdev, "Error clearing Multicast promiscuous mode on VSI %i\n", |
| 309 | vsi->vsi_num); |
| 310 | vsi->current_netdev_flags |= IFF_ALLMULTI; |
| 311 | goto out_promisc; |
| 312 | } |
| 313 | } |
| 314 | } |
| 315 | |
| 316 | if (((changed_flags & IFF_PROMISC) || promisc_forced_on) || |
| 317 | test_bit(ICE_VSI_FLAG_PROMISC_CHANGED, vsi->flags)) { |
| 318 | clear_bit(ICE_VSI_FLAG_PROMISC_CHANGED, vsi->flags); |
| 319 | if (vsi->current_netdev_flags & IFF_PROMISC) { |
| 320 | /* Apply Tx filter rule to get traffic from VMs */ |
| 321 | status = ice_cfg_dflt_vsi(hw, vsi->idx, true, |
| 322 | ICE_FLTR_TX); |
| 323 | if (status) { |
| 324 | netdev_err(netdev, "Error setting default VSI %i tx rule\n", |
| 325 | vsi->vsi_num); |
| 326 | vsi->current_netdev_flags &= ~IFF_PROMISC; |
| 327 | err = -EIO; |
| 328 | goto out_promisc; |
| 329 | } |
| 330 | /* Apply Rx filter rule to get traffic from wire */ |
| 331 | status = ice_cfg_dflt_vsi(hw, vsi->idx, true, |
| 332 | ICE_FLTR_RX); |
| 333 | if (status) { |
| 334 | netdev_err(netdev, "Error setting default VSI %i rx rule\n", |
| 335 | vsi->vsi_num); |
| 336 | vsi->current_netdev_flags &= ~IFF_PROMISC; |
| 337 | err = -EIO; |
| 338 | goto out_promisc; |
| 339 | } |
| 340 | } else { |
| 341 | /* Clear Tx filter rule to stop traffic from VMs */ |
| 342 | status = ice_cfg_dflt_vsi(hw, vsi->idx, false, |
| 343 | ICE_FLTR_TX); |
| 344 | if (status) { |
| 345 | netdev_err(netdev, "Error clearing default VSI %i tx rule\n", |
| 346 | vsi->vsi_num); |
| 347 | vsi->current_netdev_flags |= IFF_PROMISC; |
| 348 | err = -EIO; |
| 349 | goto out_promisc; |
| 350 | } |
| 351 | /* Clear Rx filter to remove traffic from wire */ |
| 352 | status = ice_cfg_dflt_vsi(hw, vsi->idx, false, |
| 353 | ICE_FLTR_RX); |
| 354 | if (status) { |
| 355 | netdev_err(netdev, "Error clearing default VSI %i rx rule\n", |
| 356 | vsi->vsi_num); |
| 357 | vsi->current_netdev_flags |= IFF_PROMISC; |
| 358 | err = -EIO; |
| 359 | goto out_promisc; |
| 360 | } |
| 361 | } |
| 362 | } |
| 363 | goto exit; |
| 364 | |
| 365 | out_promisc: |
| 366 | set_bit(ICE_VSI_FLAG_PROMISC_CHANGED, vsi->flags); |
| 367 | goto exit; |
| 368 | out: |
| 369 | /* if something went wrong then set the changed flag so we try again */ |
| 370 | set_bit(ICE_VSI_FLAG_UMAC_FLTR_CHANGED, vsi->flags); |
| 371 | set_bit(ICE_VSI_FLAG_MMAC_FLTR_CHANGED, vsi->flags); |
| 372 | exit: |
| 373 | clear_bit(__ICE_CFG_BUSY, vsi->state); |
| 374 | return err; |
| 375 | } |
| 376 | |
| 377 | /** |
| 378 | * ice_sync_fltr_subtask - Sync the VSI filter list with HW |
| 379 | * @pf: board private structure |
| 380 | */ |
| 381 | static void ice_sync_fltr_subtask(struct ice_pf *pf) |
| 382 | { |
| 383 | int v; |
| 384 | |
| 385 | if (!pf || !(test_bit(ICE_FLAG_FLTR_SYNC, pf->flags))) |
| 386 | return; |
| 387 | |
| 388 | clear_bit(ICE_FLAG_FLTR_SYNC, pf->flags); |
| 389 | |
| 390 | ice_for_each_vsi(pf, v) |
| 391 | if (pf->vsi[v] && ice_vsi_fltr_changed(pf->vsi[v]) && |
| 392 | ice_vsi_sync_fltr(pf->vsi[v])) { |
| 393 | /* come back and try again later */ |
| 394 | set_bit(ICE_FLAG_FLTR_SYNC, pf->flags); |
| 395 | break; |
| 396 | } |
| 397 | } |
| 398 | |
| 399 | /** |
| 400 | * ice_dis_vsi - pause a VSI |
| 401 | * @vsi: the VSI being paused |
| 402 | * @locked: is the rtnl_lock already held |
| 403 | */ |
| 404 | static void ice_dis_vsi(struct ice_vsi *vsi, bool locked) |
| 405 | { |
| 406 | if (test_bit(__ICE_DOWN, vsi->state)) |
| 407 | return; |
| 408 | |
| 409 | set_bit(__ICE_NEEDS_RESTART, vsi->state); |
| 410 | |
| 411 | if (vsi->type == ICE_VSI_PF && vsi->netdev) { |
| 412 | if (netif_running(vsi->netdev)) { |
| 413 | if (!locked) { |
| 414 | rtnl_lock(); |
| 415 | vsi->netdev->netdev_ops->ndo_stop(vsi->netdev); |
| 416 | rtnl_unlock(); |
| 417 | } else { |
| 418 | vsi->netdev->netdev_ops->ndo_stop(vsi->netdev); |
| 419 | } |
| 420 | } else { |
| 421 | ice_vsi_close(vsi); |
| 422 | } |
| 423 | } |
| 424 | } |
| 425 | |
| 426 | /** |
| 427 | * ice_pf_dis_all_vsi - Pause all VSIs on a PF |
| 428 | * @pf: the PF |
| 429 | * @locked: is the rtnl_lock already held |
| 430 | */ |
| 431 | #ifdef CONFIG_DCB |
| 432 | void ice_pf_dis_all_vsi(struct ice_pf *pf, bool locked) |
| 433 | #else |
| 434 | static void ice_pf_dis_all_vsi(struct ice_pf *pf, bool locked) |
| 435 | #endif /* CONFIG_DCB */ |
| 436 | { |
| 437 | int v; |
| 438 | |
| 439 | ice_for_each_vsi(pf, v) |
| 440 | if (pf->vsi[v]) |
| 441 | ice_dis_vsi(pf->vsi[v], locked); |
| 442 | } |
| 443 | |
| 444 | /** |
| 445 | * ice_prepare_for_reset - prep for the core to reset |
| 446 | * @pf: board private structure |
| 447 | * |
| 448 | * Inform or close all dependent features in prep for reset. |
| 449 | */ |
| 450 | static void |
| 451 | ice_prepare_for_reset(struct ice_pf *pf) |
| 452 | { |
| 453 | struct ice_hw *hw = &pf->hw; |
| 454 | |
| 455 | /* already prepared for reset */ |
| 456 | if (test_bit(__ICE_PREPARED_FOR_RESET, pf->state)) |
| 457 | return; |
| 458 | |
| 459 | /* Notify VFs of impending reset */ |
| 460 | if (ice_check_sq_alive(hw, &hw->mailboxq)) |
| 461 | ice_vc_notify_reset(pf); |
| 462 | |
| 463 | /* disable the VSIs and their queues that are not already DOWN */ |
| 464 | ice_pf_dis_all_vsi(pf, false); |
| 465 | |
| 466 | if (hw->port_info) |
| 467 | ice_sched_clear_port(hw->port_info); |
| 468 | |
| 469 | ice_shutdown_all_ctrlq(hw); |
| 470 | |
| 471 | set_bit(__ICE_PREPARED_FOR_RESET, pf->state); |
| 472 | } |
| 473 | |
| 474 | /** |
| 475 | * ice_do_reset - Initiate one of many types of resets |
| 476 | * @pf: board private structure |
| 477 | * @reset_type: reset type requested |
| 478 | * before this function was called. |
| 479 | */ |
| 480 | static void ice_do_reset(struct ice_pf *pf, enum ice_reset_req reset_type) |
| 481 | { |
| 482 | struct device *dev = &pf->pdev->dev; |
| 483 | struct ice_hw *hw = &pf->hw; |
| 484 | |
| 485 | dev_dbg(dev, "reset_type 0x%x requested\n", reset_type); |
| 486 | WARN_ON(in_interrupt()); |
| 487 | |
| 488 | ice_prepare_for_reset(pf); |
| 489 | |
| 490 | /* trigger the reset */ |
| 491 | if (ice_reset(hw, reset_type)) { |
| 492 | dev_err(dev, "reset %d failed\n", reset_type); |
| 493 | set_bit(__ICE_RESET_FAILED, pf->state); |
| 494 | clear_bit(__ICE_RESET_OICR_RECV, pf->state); |
| 495 | clear_bit(__ICE_PREPARED_FOR_RESET, pf->state); |
| 496 | clear_bit(__ICE_PFR_REQ, pf->state); |
| 497 | clear_bit(__ICE_CORER_REQ, pf->state); |
| 498 | clear_bit(__ICE_GLOBR_REQ, pf->state); |
| 499 | return; |
| 500 | } |
| 501 | |
| 502 | /* PFR is a bit of a special case because it doesn't result in an OICR |
| 503 | * interrupt. So for PFR, rebuild after the reset and clear the reset- |
| 504 | * associated state bits. |
| 505 | */ |
| 506 | if (reset_type == ICE_RESET_PFR) { |
| 507 | pf->pfr_count++; |
| 508 | ice_rebuild(pf); |
| 509 | clear_bit(__ICE_PREPARED_FOR_RESET, pf->state); |
| 510 | clear_bit(__ICE_PFR_REQ, pf->state); |
| 511 | ice_reset_all_vfs(pf, true); |
| 512 | } |
| 513 | } |
| 514 | |
| 515 | /** |
| 516 | * ice_reset_subtask - Set up for resetting the device and driver |
| 517 | * @pf: board private structure |
| 518 | */ |
| 519 | static void ice_reset_subtask(struct ice_pf *pf) |
| 520 | { |
| 521 | enum ice_reset_req reset_type = ICE_RESET_INVAL; |
| 522 | |
| 523 | /* When a CORER/GLOBR/EMPR is about to happen, the hardware triggers an |
| 524 | * OICR interrupt. The OICR handler (ice_misc_intr) determines what type |
| 525 | * of reset is pending and sets bits in pf->state indicating the reset |
| 526 | * type and __ICE_RESET_OICR_RECV. So, if the latter bit is set |
| 527 | * prepare for pending reset if not already (for PF software-initiated |
| 528 | * global resets the software should already be prepared for it as |
| 529 | * indicated by __ICE_PREPARED_FOR_RESET; for global resets initiated |
| 530 | * by firmware or software on other PFs, that bit is not set so prepare |
| 531 | * for the reset now), poll for reset done, rebuild and return. |
| 532 | */ |
| 533 | if (test_bit(__ICE_RESET_OICR_RECV, pf->state)) { |
| 534 | /* Perform the largest reset requested */ |
| 535 | if (test_and_clear_bit(__ICE_CORER_RECV, pf->state)) |
| 536 | reset_type = ICE_RESET_CORER; |
| 537 | if (test_and_clear_bit(__ICE_GLOBR_RECV, pf->state)) |
| 538 | reset_type = ICE_RESET_GLOBR; |
| 539 | /* return if no valid reset type requested */ |
| 540 | if (reset_type == ICE_RESET_INVAL) |
| 541 | return; |
| 542 | ice_prepare_for_reset(pf); |
| 543 | |
| 544 | /* make sure we are ready to rebuild */ |
| 545 | if (ice_check_reset(&pf->hw)) { |
| 546 | set_bit(__ICE_RESET_FAILED, pf->state); |
| 547 | } else { |
| 548 | /* done with reset. start rebuild */ |
| 549 | pf->hw.reset_ongoing = false; |
| 550 | ice_rebuild(pf); |
| 551 | /* clear bit to resume normal operations, but |
| 552 | * ICE_NEEDS_RESTART bit is set in case rebuild failed |
| 553 | */ |
| 554 | clear_bit(__ICE_RESET_OICR_RECV, pf->state); |
| 555 | clear_bit(__ICE_PREPARED_FOR_RESET, pf->state); |
| 556 | clear_bit(__ICE_PFR_REQ, pf->state); |
| 557 | clear_bit(__ICE_CORER_REQ, pf->state); |
| 558 | clear_bit(__ICE_GLOBR_REQ, pf->state); |
| 559 | ice_reset_all_vfs(pf, true); |
| 560 | } |
| 561 | |
| 562 | return; |
| 563 | } |
| 564 | |
| 565 | /* No pending resets to finish processing. Check for new resets */ |
| 566 | if (test_bit(__ICE_PFR_REQ, pf->state)) |
| 567 | reset_type = ICE_RESET_PFR; |
| 568 | if (test_bit(__ICE_CORER_REQ, pf->state)) |
| 569 | reset_type = ICE_RESET_CORER; |
| 570 | if (test_bit(__ICE_GLOBR_REQ, pf->state)) |
| 571 | reset_type = ICE_RESET_GLOBR; |
| 572 | /* If no valid reset type requested just return */ |
| 573 | if (reset_type == ICE_RESET_INVAL) |
| 574 | return; |
| 575 | |
| 576 | /* reset if not already down or busy */ |
| 577 | if (!test_bit(__ICE_DOWN, pf->state) && |
| 578 | !test_bit(__ICE_CFG_BUSY, pf->state)) { |
| 579 | ice_do_reset(pf, reset_type); |
| 580 | } |
| 581 | } |
| 582 | |
| 583 | /** |
| 584 | * ice_print_link_msg - print link up or down message |
| 585 | * @vsi: the VSI whose link status is being queried |
| 586 | * @isup: boolean for if the link is now up or down |
| 587 | */ |
| 588 | void ice_print_link_msg(struct ice_vsi *vsi, bool isup) |
| 589 | { |
| 590 | const char *speed; |
| 591 | const char *fc; |
| 592 | |
| 593 | if (vsi->current_isup == isup) |
| 594 | return; |
| 595 | |
| 596 | vsi->current_isup = isup; |
| 597 | |
| 598 | if (!isup) { |
| 599 | netdev_info(vsi->netdev, "NIC Link is Down\n"); |
| 600 | return; |
| 601 | } |
| 602 | |
| 603 | switch (vsi->port_info->phy.link_info.link_speed) { |
| 604 | case ICE_AQ_LINK_SPEED_40GB: |
| 605 | speed = "40 G"; |
| 606 | break; |
| 607 | case ICE_AQ_LINK_SPEED_25GB: |
| 608 | speed = "25 G"; |
| 609 | break; |
| 610 | case ICE_AQ_LINK_SPEED_20GB: |
| 611 | speed = "20 G"; |
| 612 | break; |
| 613 | case ICE_AQ_LINK_SPEED_10GB: |
| 614 | speed = "10 G"; |
| 615 | break; |
| 616 | case ICE_AQ_LINK_SPEED_5GB: |
| 617 | speed = "5 G"; |
| 618 | break; |
| 619 | case ICE_AQ_LINK_SPEED_2500MB: |
| 620 | speed = "2.5 G"; |
| 621 | break; |
| 622 | case ICE_AQ_LINK_SPEED_1000MB: |
| 623 | speed = "1 G"; |
| 624 | break; |
| 625 | case ICE_AQ_LINK_SPEED_100MB: |
| 626 | speed = "100 M"; |
| 627 | break; |
| 628 | default: |
| 629 | speed = "Unknown"; |
| 630 | break; |
| 631 | } |
| 632 | |
| 633 | switch (vsi->port_info->fc.current_mode) { |
| 634 | case ICE_FC_FULL: |
| 635 | fc = "RX/TX"; |
| 636 | break; |
| 637 | case ICE_FC_TX_PAUSE: |
| 638 | fc = "TX"; |
| 639 | break; |
| 640 | case ICE_FC_RX_PAUSE: |
| 641 | fc = "RX"; |
| 642 | break; |
| 643 | case ICE_FC_NONE: |
| 644 | fc = "None"; |
| 645 | break; |
| 646 | default: |
| 647 | fc = "Unknown"; |
| 648 | break; |
| 649 | } |
| 650 | |
| 651 | netdev_info(vsi->netdev, "NIC Link is up %sbps, Flow Control: %s\n", |
| 652 | speed, fc); |
| 653 | } |
| 654 | |
| 655 | /** |
| 656 | * ice_vsi_link_event - update the VSI's netdev |
| 657 | * @vsi: the VSI on which the link event occurred |
| 658 | * @link_up: whether or not the VSI needs to be set up or down |
| 659 | */ |
| 660 | static void ice_vsi_link_event(struct ice_vsi *vsi, bool link_up) |
| 661 | { |
| 662 | if (!vsi || test_bit(__ICE_DOWN, vsi->state)) |
| 663 | return; |
| 664 | |
| 665 | if (vsi->type == ICE_VSI_PF) { |
| 666 | if (!vsi->netdev) { |
| 667 | dev_dbg(&vsi->back->pdev->dev, |
| 668 | "vsi->netdev is not initialized!\n"); |
| 669 | return; |
| 670 | } |
| 671 | if (link_up) { |
| 672 | netif_carrier_on(vsi->netdev); |
| 673 | netif_tx_wake_all_queues(vsi->netdev); |
| 674 | } else { |
| 675 | netif_carrier_off(vsi->netdev); |
| 676 | netif_tx_stop_all_queues(vsi->netdev); |
| 677 | } |
| 678 | } |
| 679 | } |
| 680 | |
| 681 | /** |
| 682 | * ice_link_event - process the link event |
| 683 | * @pf: pf that the link event is associated with |
| 684 | * @pi: port_info for the port that the link event is associated with |
| 685 | * |
| 686 | * Returns -EIO if ice_get_link_status() fails |
| 687 | * Returns 0 on success |
| 688 | */ |
| 689 | static int |
| 690 | ice_link_event(struct ice_pf *pf, struct ice_port_info *pi) |
| 691 | { |
| 692 | u8 new_link_speed, old_link_speed; |
| 693 | struct ice_phy_info *phy_info; |
| 694 | bool new_link_same_as_old; |
| 695 | bool new_link, old_link; |
| 696 | u8 lport; |
| 697 | u16 v; |
| 698 | |
| 699 | phy_info = &pi->phy; |
| 700 | phy_info->link_info_old = phy_info->link_info; |
| 701 | /* Force ice_get_link_status() to update link info */ |
| 702 | phy_info->get_link_info = true; |
| 703 | |
| 704 | old_link = (phy_info->link_info_old.link_info & ICE_AQ_LINK_UP); |
| 705 | old_link_speed = phy_info->link_info_old.link_speed; |
| 706 | |
| 707 | lport = pi->lport; |
| 708 | if (ice_get_link_status(pi, &new_link)) { |
| 709 | dev_dbg(&pf->pdev->dev, |
| 710 | "Could not get link status for port %d\n", lport); |
| 711 | return -EIO; |
| 712 | } |
| 713 | |
| 714 | new_link_speed = phy_info->link_info.link_speed; |
| 715 | |
| 716 | new_link_same_as_old = (new_link == old_link && |
| 717 | new_link_speed == old_link_speed); |
| 718 | |
| 719 | ice_for_each_vsi(pf, v) { |
| 720 | struct ice_vsi *vsi = pf->vsi[v]; |
| 721 | |
| 722 | if (!vsi || !vsi->port_info) |
| 723 | continue; |
| 724 | |
| 725 | if (new_link_same_as_old && |
| 726 | (test_bit(__ICE_DOWN, vsi->state) || |
| 727 | new_link == netif_carrier_ok(vsi->netdev))) |
| 728 | continue; |
| 729 | |
| 730 | if (vsi->port_info->lport == lport) { |
| 731 | ice_print_link_msg(vsi, new_link); |
| 732 | ice_vsi_link_event(vsi, new_link); |
| 733 | } |
| 734 | } |
| 735 | |
| 736 | if (!new_link_same_as_old && pf->num_alloc_vfs) |
| 737 | ice_vc_notify_link_state(pf); |
| 738 | |
| 739 | return 0; |
| 740 | } |
| 741 | |
| 742 | /** |
| 743 | * ice_watchdog_subtask - periodic tasks not using event driven scheduling |
| 744 | * @pf: board private structure |
| 745 | */ |
| 746 | static void ice_watchdog_subtask(struct ice_pf *pf) |
| 747 | { |
| 748 | int i; |
| 749 | |
| 750 | /* if interface is down do nothing */ |
| 751 | if (test_bit(__ICE_DOWN, pf->state) || |
| 752 | test_bit(__ICE_CFG_BUSY, pf->state)) |
| 753 | return; |
| 754 | |
| 755 | /* make sure we don't do these things too often */ |
| 756 | if (time_before(jiffies, |
| 757 | pf->serv_tmr_prev + pf->serv_tmr_period)) |
| 758 | return; |
| 759 | |
| 760 | pf->serv_tmr_prev = jiffies; |
| 761 | |
| 762 | /* Update the stats for active netdevs so the network stack |
| 763 | * can look at updated numbers whenever it cares to |
| 764 | */ |
| 765 | ice_update_pf_stats(pf); |
| 766 | ice_for_each_vsi(pf, i) |
| 767 | if (pf->vsi[i] && pf->vsi[i]->netdev) |
| 768 | ice_update_vsi_stats(pf->vsi[i]); |
| 769 | } |
| 770 | |
| 771 | /** |
| 772 | * ice_init_link_events - enable/initialize link events |
| 773 | * @pi: pointer to the port_info instance |
| 774 | * |
| 775 | * Returns -EIO on failure, 0 on success |
| 776 | */ |
| 777 | static int ice_init_link_events(struct ice_port_info *pi) |
| 778 | { |
| 779 | u16 mask; |
| 780 | |
| 781 | mask = ~((u16)(ICE_AQ_LINK_EVENT_UPDOWN | ICE_AQ_LINK_EVENT_MEDIA_NA | |
| 782 | ICE_AQ_LINK_EVENT_MODULE_QUAL_FAIL)); |
| 783 | |
| 784 | if (ice_aq_set_event_mask(pi->hw, pi->lport, mask, NULL)) { |
| 785 | dev_dbg(ice_hw_to_dev(pi->hw), |
| 786 | "Failed to set link event mask for port %d\n", |
| 787 | pi->lport); |
| 788 | return -EIO; |
| 789 | } |
| 790 | |
| 791 | if (ice_aq_get_link_info(pi, true, NULL, NULL)) { |
| 792 | dev_dbg(ice_hw_to_dev(pi->hw), |
| 793 | "Failed to enable link events for port %d\n", |
| 794 | pi->lport); |
| 795 | return -EIO; |
| 796 | } |
| 797 | |
| 798 | return 0; |
| 799 | } |
| 800 | |
| 801 | /** |
| 802 | * ice_handle_link_event - handle link event via ARQ |
| 803 | * @pf: pf that the link event is associated with |
| 804 | * |
| 805 | * Return -EINVAL if port_info is null |
| 806 | * Return status on success |
| 807 | */ |
| 808 | static int ice_handle_link_event(struct ice_pf *pf) |
| 809 | { |
| 810 | struct ice_port_info *port_info; |
| 811 | int status; |
| 812 | |
| 813 | port_info = pf->hw.port_info; |
| 814 | if (!port_info) |
| 815 | return -EINVAL; |
| 816 | |
| 817 | status = ice_link_event(pf, port_info); |
| 818 | if (status) |
| 819 | dev_dbg(&pf->pdev->dev, |
| 820 | "Could not process link event, error %d\n", status); |
| 821 | |
| 822 | return status; |
| 823 | } |
| 824 | |
| 825 | /** |
| 826 | * __ice_clean_ctrlq - helper function to clean controlq rings |
| 827 | * @pf: ptr to struct ice_pf |
| 828 | * @q_type: specific Control queue type |
| 829 | */ |
| 830 | static int __ice_clean_ctrlq(struct ice_pf *pf, enum ice_ctl_q q_type) |
| 831 | { |
| 832 | struct ice_rq_event_info event; |
| 833 | struct ice_hw *hw = &pf->hw; |
| 834 | struct ice_ctl_q_info *cq; |
| 835 | u16 pending, i = 0; |
| 836 | const char *qtype; |
| 837 | u32 oldval, val; |
| 838 | |
| 839 | /* Do not clean control queue if/when PF reset fails */ |
| 840 | if (test_bit(__ICE_RESET_FAILED, pf->state)) |
| 841 | return 0; |
| 842 | |
| 843 | switch (q_type) { |
| 844 | case ICE_CTL_Q_ADMIN: |
| 845 | cq = &hw->adminq; |
| 846 | qtype = "Admin"; |
| 847 | break; |
| 848 | case ICE_CTL_Q_MAILBOX: |
| 849 | cq = &hw->mailboxq; |
| 850 | qtype = "Mailbox"; |
| 851 | break; |
| 852 | default: |
| 853 | dev_warn(&pf->pdev->dev, "Unknown control queue type 0x%x\n", |
| 854 | q_type); |
| 855 | return 0; |
| 856 | } |
| 857 | |
| 858 | /* check for error indications - PF_xx_AxQLEN register layout for |
| 859 | * FW/MBX/SB are identical so just use defines for PF_FW_AxQLEN. |
| 860 | */ |
| 861 | val = rd32(hw, cq->rq.len); |
| 862 | if (val & (PF_FW_ARQLEN_ARQVFE_M | PF_FW_ARQLEN_ARQOVFL_M | |
| 863 | PF_FW_ARQLEN_ARQCRIT_M)) { |
| 864 | oldval = val; |
| 865 | if (val & PF_FW_ARQLEN_ARQVFE_M) |
| 866 | dev_dbg(&pf->pdev->dev, |
| 867 | "%s Receive Queue VF Error detected\n", qtype); |
| 868 | if (val & PF_FW_ARQLEN_ARQOVFL_M) { |
| 869 | dev_dbg(&pf->pdev->dev, |
| 870 | "%s Receive Queue Overflow Error detected\n", |
| 871 | qtype); |
| 872 | } |
| 873 | if (val & PF_FW_ARQLEN_ARQCRIT_M) |
| 874 | dev_dbg(&pf->pdev->dev, |
| 875 | "%s Receive Queue Critical Error detected\n", |
| 876 | qtype); |
| 877 | val &= ~(PF_FW_ARQLEN_ARQVFE_M | PF_FW_ARQLEN_ARQOVFL_M | |
| 878 | PF_FW_ARQLEN_ARQCRIT_M); |
| 879 | if (oldval != val) |
| 880 | wr32(hw, cq->rq.len, val); |
| 881 | } |
| 882 | |
| 883 | val = rd32(hw, cq->sq.len); |
| 884 | if (val & (PF_FW_ATQLEN_ATQVFE_M | PF_FW_ATQLEN_ATQOVFL_M | |
| 885 | PF_FW_ATQLEN_ATQCRIT_M)) { |
| 886 | oldval = val; |
| 887 | if (val & PF_FW_ATQLEN_ATQVFE_M) |
| 888 | dev_dbg(&pf->pdev->dev, |
| 889 | "%s Send Queue VF Error detected\n", qtype); |
| 890 | if (val & PF_FW_ATQLEN_ATQOVFL_M) { |
| 891 | dev_dbg(&pf->pdev->dev, |
| 892 | "%s Send Queue Overflow Error detected\n", |
| 893 | qtype); |
| 894 | } |
| 895 | if (val & PF_FW_ATQLEN_ATQCRIT_M) |
| 896 | dev_dbg(&pf->pdev->dev, |
| 897 | "%s Send Queue Critical Error detected\n", |
| 898 | qtype); |
| 899 | val &= ~(PF_FW_ATQLEN_ATQVFE_M | PF_FW_ATQLEN_ATQOVFL_M | |
| 900 | PF_FW_ATQLEN_ATQCRIT_M); |
| 901 | if (oldval != val) |
| 902 | wr32(hw, cq->sq.len, val); |
| 903 | } |
| 904 | |
| 905 | event.buf_len = cq->rq_buf_size; |
| 906 | event.msg_buf = devm_kzalloc(&pf->pdev->dev, event.buf_len, |
| 907 | GFP_KERNEL); |
| 908 | if (!event.msg_buf) |
| 909 | return 0; |
| 910 | |
| 911 | do { |
| 912 | enum ice_status ret; |
| 913 | u16 opcode; |
| 914 | |
| 915 | ret = ice_clean_rq_elem(hw, cq, &event, &pending); |
| 916 | if (ret == ICE_ERR_AQ_NO_WORK) |
| 917 | break; |
| 918 | if (ret) { |
| 919 | dev_err(&pf->pdev->dev, |
| 920 | "%s Receive Queue event error %d\n", qtype, |
| 921 | ret); |
| 922 | break; |
| 923 | } |
| 924 | |
| 925 | opcode = le16_to_cpu(event.desc.opcode); |
| 926 | |
| 927 | switch (opcode) { |
| 928 | case ice_aqc_opc_get_link_status: |
| 929 | if (ice_handle_link_event(pf)) |
| 930 | dev_err(&pf->pdev->dev, |
| 931 | "Could not handle link event\n"); |
| 932 | break; |
| 933 | case ice_mbx_opc_send_msg_to_pf: |
| 934 | ice_vc_process_vf_msg(pf, &event); |
| 935 | break; |
| 936 | case ice_aqc_opc_fw_logging: |
| 937 | ice_output_fw_log(hw, &event.desc, event.msg_buf); |
| 938 | break; |
| 939 | case ice_aqc_opc_lldp_set_mib_change: |
| 940 | ice_dcb_process_lldp_set_mib_change(pf, &event); |
| 941 | break; |
| 942 | default: |
| 943 | dev_dbg(&pf->pdev->dev, |
| 944 | "%s Receive Queue unknown event 0x%04x ignored\n", |
| 945 | qtype, opcode); |
| 946 | break; |
| 947 | } |
| 948 | } while (pending && (i++ < ICE_DFLT_IRQ_WORK)); |
| 949 | |
| 950 | devm_kfree(&pf->pdev->dev, event.msg_buf); |
| 951 | |
| 952 | return pending && (i == ICE_DFLT_IRQ_WORK); |
| 953 | } |
| 954 | |
| 955 | /** |
| 956 | * ice_ctrlq_pending - check if there is a difference between ntc and ntu |
| 957 | * @hw: pointer to hardware info |
| 958 | * @cq: control queue information |
| 959 | * |
| 960 | * returns true if there are pending messages in a queue, false if there aren't |
| 961 | */ |
| 962 | static bool ice_ctrlq_pending(struct ice_hw *hw, struct ice_ctl_q_info *cq) |
| 963 | { |
| 964 | u16 ntu; |
| 965 | |
| 966 | ntu = (u16)(rd32(hw, cq->rq.head) & cq->rq.head_mask); |
| 967 | return cq->rq.next_to_clean != ntu; |
| 968 | } |
| 969 | |
| 970 | /** |
| 971 | * ice_clean_adminq_subtask - clean the AdminQ rings |
| 972 | * @pf: board private structure |
| 973 | */ |
| 974 | static void ice_clean_adminq_subtask(struct ice_pf *pf) |
| 975 | { |
| 976 | struct ice_hw *hw = &pf->hw; |
| 977 | |
| 978 | if (!test_bit(__ICE_ADMINQ_EVENT_PENDING, pf->state)) |
| 979 | return; |
| 980 | |
| 981 | if (__ice_clean_ctrlq(pf, ICE_CTL_Q_ADMIN)) |
| 982 | return; |
| 983 | |
| 984 | clear_bit(__ICE_ADMINQ_EVENT_PENDING, pf->state); |
| 985 | |
| 986 | /* There might be a situation where new messages arrive to a control |
| 987 | * queue between processing the last message and clearing the |
| 988 | * EVENT_PENDING bit. So before exiting, check queue head again (using |
| 989 | * ice_ctrlq_pending) and process new messages if any. |
| 990 | */ |
| 991 | if (ice_ctrlq_pending(hw, &hw->adminq)) |
| 992 | __ice_clean_ctrlq(pf, ICE_CTL_Q_ADMIN); |
| 993 | |
| 994 | ice_flush(hw); |
| 995 | } |
| 996 | |
| 997 | /** |
| 998 | * ice_clean_mailboxq_subtask - clean the MailboxQ rings |
| 999 | * @pf: board private structure |
| 1000 | */ |
| 1001 | static void ice_clean_mailboxq_subtask(struct ice_pf *pf) |
| 1002 | { |
| 1003 | struct ice_hw *hw = &pf->hw; |
| 1004 | |
| 1005 | if (!test_bit(__ICE_MAILBOXQ_EVENT_PENDING, pf->state)) |
| 1006 | return; |
| 1007 | |
| 1008 | if (__ice_clean_ctrlq(pf, ICE_CTL_Q_MAILBOX)) |
| 1009 | return; |
| 1010 | |
| 1011 | clear_bit(__ICE_MAILBOXQ_EVENT_PENDING, pf->state); |
| 1012 | |
| 1013 | if (ice_ctrlq_pending(hw, &hw->mailboxq)) |
| 1014 | __ice_clean_ctrlq(pf, ICE_CTL_Q_MAILBOX); |
| 1015 | |
| 1016 | ice_flush(hw); |
| 1017 | } |
| 1018 | |
| 1019 | /** |
| 1020 | * ice_service_task_schedule - schedule the service task to wake up |
| 1021 | * @pf: board private structure |
| 1022 | * |
| 1023 | * If not already scheduled, this puts the task into the work queue. |
| 1024 | */ |
| 1025 | static void ice_service_task_schedule(struct ice_pf *pf) |
| 1026 | { |
| 1027 | if (!test_bit(__ICE_SERVICE_DIS, pf->state) && |
| 1028 | !test_and_set_bit(__ICE_SERVICE_SCHED, pf->state) && |
| 1029 | !test_bit(__ICE_NEEDS_RESTART, pf->state)) |
| 1030 | queue_work(ice_wq, &pf->serv_task); |
| 1031 | } |
| 1032 | |
| 1033 | /** |
| 1034 | * ice_service_task_complete - finish up the service task |
| 1035 | * @pf: board private structure |
| 1036 | */ |
| 1037 | static void ice_service_task_complete(struct ice_pf *pf) |
| 1038 | { |
| 1039 | WARN_ON(!test_bit(__ICE_SERVICE_SCHED, pf->state)); |
| 1040 | |
| 1041 | /* force memory (pf->state) to sync before next service task */ |
| 1042 | smp_mb__before_atomic(); |
| 1043 | clear_bit(__ICE_SERVICE_SCHED, pf->state); |
| 1044 | } |
| 1045 | |
| 1046 | /** |
| 1047 | * ice_service_task_stop - stop service task and cancel works |
| 1048 | * @pf: board private structure |
| 1049 | */ |
| 1050 | static void ice_service_task_stop(struct ice_pf *pf) |
| 1051 | { |
| 1052 | set_bit(__ICE_SERVICE_DIS, pf->state); |
| 1053 | |
| 1054 | if (pf->serv_tmr.function) |
| 1055 | del_timer_sync(&pf->serv_tmr); |
| 1056 | if (pf->serv_task.func) |
| 1057 | cancel_work_sync(&pf->serv_task); |
| 1058 | |
| 1059 | clear_bit(__ICE_SERVICE_SCHED, pf->state); |
| 1060 | } |
| 1061 | |
| 1062 | /** |
| 1063 | * ice_service_task_restart - restart service task and schedule works |
| 1064 | * @pf: board private structure |
| 1065 | * |
| 1066 | * This function is needed for suspend and resume works (e.g WoL scenario) |
| 1067 | */ |
| 1068 | static void ice_service_task_restart(struct ice_pf *pf) |
| 1069 | { |
| 1070 | clear_bit(__ICE_SERVICE_DIS, pf->state); |
| 1071 | ice_service_task_schedule(pf); |
| 1072 | } |
| 1073 | |
| 1074 | /** |
| 1075 | * ice_service_timer - timer callback to schedule service task |
| 1076 | * @t: pointer to timer_list |
| 1077 | */ |
| 1078 | static void ice_service_timer(struct timer_list *t) |
| 1079 | { |
| 1080 | struct ice_pf *pf = from_timer(pf, t, serv_tmr); |
| 1081 | |
| 1082 | mod_timer(&pf->serv_tmr, round_jiffies(pf->serv_tmr_period + jiffies)); |
| 1083 | ice_service_task_schedule(pf); |
| 1084 | } |
| 1085 | |
| 1086 | /** |
| 1087 | * ice_handle_mdd_event - handle malicious driver detect event |
| 1088 | * @pf: pointer to the PF structure |
| 1089 | * |
| 1090 | * Called from service task. OICR interrupt handler indicates MDD event |
| 1091 | */ |
| 1092 | static void ice_handle_mdd_event(struct ice_pf *pf) |
| 1093 | { |
| 1094 | struct ice_hw *hw = &pf->hw; |
| 1095 | bool mdd_detected = false; |
| 1096 | u32 reg; |
| 1097 | int i; |
| 1098 | |
| 1099 | if (!test_bit(__ICE_MDD_EVENT_PENDING, pf->state)) |
| 1100 | return; |
| 1101 | |
| 1102 | /* find what triggered the MDD event */ |
| 1103 | reg = rd32(hw, GL_MDET_TX_PQM); |
| 1104 | if (reg & GL_MDET_TX_PQM_VALID_M) { |
| 1105 | u8 pf_num = (reg & GL_MDET_TX_PQM_PF_NUM_M) >> |
| 1106 | GL_MDET_TX_PQM_PF_NUM_S; |
| 1107 | u16 vf_num = (reg & GL_MDET_TX_PQM_VF_NUM_M) >> |
| 1108 | GL_MDET_TX_PQM_VF_NUM_S; |
| 1109 | u8 event = (reg & GL_MDET_TX_PQM_MAL_TYPE_M) >> |
| 1110 | GL_MDET_TX_PQM_MAL_TYPE_S; |
| 1111 | u16 queue = ((reg & GL_MDET_TX_PQM_QNUM_M) >> |
| 1112 | GL_MDET_TX_PQM_QNUM_S); |
| 1113 | |
| 1114 | if (netif_msg_tx_err(pf)) |
| 1115 | dev_info(&pf->pdev->dev, "Malicious Driver Detection event %d on TX queue %d PF# %d VF# %d\n", |
| 1116 | event, queue, pf_num, vf_num); |
| 1117 | wr32(hw, GL_MDET_TX_PQM, 0xffffffff); |
| 1118 | mdd_detected = true; |
| 1119 | } |
| 1120 | |
| 1121 | reg = rd32(hw, GL_MDET_TX_TCLAN); |
| 1122 | if (reg & GL_MDET_TX_TCLAN_VALID_M) { |
| 1123 | u8 pf_num = (reg & GL_MDET_TX_TCLAN_PF_NUM_M) >> |
| 1124 | GL_MDET_TX_TCLAN_PF_NUM_S; |
| 1125 | u16 vf_num = (reg & GL_MDET_TX_TCLAN_VF_NUM_M) >> |
| 1126 | GL_MDET_TX_TCLAN_VF_NUM_S; |
| 1127 | u8 event = (reg & GL_MDET_TX_TCLAN_MAL_TYPE_M) >> |
| 1128 | GL_MDET_TX_TCLAN_MAL_TYPE_S; |
| 1129 | u16 queue = ((reg & GL_MDET_TX_TCLAN_QNUM_M) >> |
| 1130 | GL_MDET_TX_TCLAN_QNUM_S); |
| 1131 | |
| 1132 | if (netif_msg_rx_err(pf)) |
| 1133 | dev_info(&pf->pdev->dev, "Malicious Driver Detection event %d on TX queue %d PF# %d VF# %d\n", |
| 1134 | event, queue, pf_num, vf_num); |
| 1135 | wr32(hw, GL_MDET_TX_TCLAN, 0xffffffff); |
| 1136 | mdd_detected = true; |
| 1137 | } |
| 1138 | |
| 1139 | reg = rd32(hw, GL_MDET_RX); |
| 1140 | if (reg & GL_MDET_RX_VALID_M) { |
| 1141 | u8 pf_num = (reg & GL_MDET_RX_PF_NUM_M) >> |
| 1142 | GL_MDET_RX_PF_NUM_S; |
| 1143 | u16 vf_num = (reg & GL_MDET_RX_VF_NUM_M) >> |
| 1144 | GL_MDET_RX_VF_NUM_S; |
| 1145 | u8 event = (reg & GL_MDET_RX_MAL_TYPE_M) >> |
| 1146 | GL_MDET_RX_MAL_TYPE_S; |
| 1147 | u16 queue = ((reg & GL_MDET_RX_QNUM_M) >> |
| 1148 | GL_MDET_RX_QNUM_S); |
| 1149 | |
| 1150 | if (netif_msg_rx_err(pf)) |
| 1151 | dev_info(&pf->pdev->dev, "Malicious Driver Detection event %d on RX queue %d PF# %d VF# %d\n", |
| 1152 | event, queue, pf_num, vf_num); |
| 1153 | wr32(hw, GL_MDET_RX, 0xffffffff); |
| 1154 | mdd_detected = true; |
| 1155 | } |
| 1156 | |
| 1157 | if (mdd_detected) { |
| 1158 | bool pf_mdd_detected = false; |
| 1159 | |
| 1160 | reg = rd32(hw, PF_MDET_TX_PQM); |
| 1161 | if (reg & PF_MDET_TX_PQM_VALID_M) { |
| 1162 | wr32(hw, PF_MDET_TX_PQM, 0xFFFF); |
| 1163 | dev_info(&pf->pdev->dev, "TX driver issue detected, PF reset issued\n"); |
| 1164 | pf_mdd_detected = true; |
| 1165 | } |
| 1166 | |
| 1167 | reg = rd32(hw, PF_MDET_TX_TCLAN); |
| 1168 | if (reg & PF_MDET_TX_TCLAN_VALID_M) { |
| 1169 | wr32(hw, PF_MDET_TX_TCLAN, 0xFFFF); |
| 1170 | dev_info(&pf->pdev->dev, "TX driver issue detected, PF reset issued\n"); |
| 1171 | pf_mdd_detected = true; |
| 1172 | } |
| 1173 | |
| 1174 | reg = rd32(hw, PF_MDET_RX); |
| 1175 | if (reg & PF_MDET_RX_VALID_M) { |
| 1176 | wr32(hw, PF_MDET_RX, 0xFFFF); |
| 1177 | dev_info(&pf->pdev->dev, "RX driver issue detected, PF reset issued\n"); |
| 1178 | pf_mdd_detected = true; |
| 1179 | } |
| 1180 | /* Queue belongs to the PF initiate a reset */ |
| 1181 | if (pf_mdd_detected) { |
| 1182 | set_bit(__ICE_NEEDS_RESTART, pf->state); |
| 1183 | ice_service_task_schedule(pf); |
| 1184 | } |
| 1185 | } |
| 1186 | |
| 1187 | /* see if one of the VFs needs to be reset */ |
| 1188 | for (i = 0; i < pf->num_alloc_vfs && mdd_detected; i++) { |
| 1189 | struct ice_vf *vf = &pf->vf[i]; |
| 1190 | |
| 1191 | reg = rd32(hw, VP_MDET_TX_PQM(i)); |
| 1192 | if (reg & VP_MDET_TX_PQM_VALID_M) { |
| 1193 | wr32(hw, VP_MDET_TX_PQM(i), 0xFFFF); |
| 1194 | vf->num_mdd_events++; |
| 1195 | dev_info(&pf->pdev->dev, "TX driver issue detected on VF %d\n", |
| 1196 | i); |
| 1197 | } |
| 1198 | |
| 1199 | reg = rd32(hw, VP_MDET_TX_TCLAN(i)); |
| 1200 | if (reg & VP_MDET_TX_TCLAN_VALID_M) { |
| 1201 | wr32(hw, VP_MDET_TX_TCLAN(i), 0xFFFF); |
| 1202 | vf->num_mdd_events++; |
| 1203 | dev_info(&pf->pdev->dev, "TX driver issue detected on VF %d\n", |
| 1204 | i); |
| 1205 | } |
| 1206 | |
| 1207 | reg = rd32(hw, VP_MDET_TX_TDPU(i)); |
| 1208 | if (reg & VP_MDET_TX_TDPU_VALID_M) { |
| 1209 | wr32(hw, VP_MDET_TX_TDPU(i), 0xFFFF); |
| 1210 | vf->num_mdd_events++; |
| 1211 | dev_info(&pf->pdev->dev, "TX driver issue detected on VF %d\n", |
| 1212 | i); |
| 1213 | } |
| 1214 | |
| 1215 | reg = rd32(hw, VP_MDET_RX(i)); |
| 1216 | if (reg & VP_MDET_RX_VALID_M) { |
| 1217 | wr32(hw, VP_MDET_RX(i), 0xFFFF); |
| 1218 | vf->num_mdd_events++; |
| 1219 | dev_info(&pf->pdev->dev, "RX driver issue detected on VF %d\n", |
| 1220 | i); |
| 1221 | } |
| 1222 | |
| 1223 | if (vf->num_mdd_events > ICE_DFLT_NUM_MDD_EVENTS_ALLOWED) { |
| 1224 | dev_info(&pf->pdev->dev, |
| 1225 | "Too many MDD events on VF %d, disabled\n", i); |
| 1226 | dev_info(&pf->pdev->dev, |
| 1227 | "Use PF Control I/F to re-enable the VF\n"); |
| 1228 | set_bit(ICE_VF_STATE_DIS, vf->vf_states); |
| 1229 | } |
| 1230 | } |
| 1231 | |
| 1232 | /* re-enable MDD interrupt cause */ |
| 1233 | clear_bit(__ICE_MDD_EVENT_PENDING, pf->state); |
| 1234 | reg = rd32(hw, PFINT_OICR_ENA); |
| 1235 | reg |= PFINT_OICR_MAL_DETECT_M; |
| 1236 | wr32(hw, PFINT_OICR_ENA, reg); |
| 1237 | ice_flush(hw); |
| 1238 | } |
| 1239 | |
| 1240 | /** |
| 1241 | * ice_service_task - manage and run subtasks |
| 1242 | * @work: pointer to work_struct contained by the PF struct |
| 1243 | */ |
| 1244 | static void ice_service_task(struct work_struct *work) |
| 1245 | { |
| 1246 | struct ice_pf *pf = container_of(work, struct ice_pf, serv_task); |
| 1247 | unsigned long start_time = jiffies; |
| 1248 | |
| 1249 | /* subtasks */ |
| 1250 | |
| 1251 | /* process reset requests first */ |
| 1252 | ice_reset_subtask(pf); |
| 1253 | |
| 1254 | /* bail if a reset/recovery cycle is pending or rebuild failed */ |
| 1255 | if (ice_is_reset_in_progress(pf->state) || |
| 1256 | test_bit(__ICE_SUSPENDED, pf->state) || |
| 1257 | test_bit(__ICE_NEEDS_RESTART, pf->state)) { |
| 1258 | ice_service_task_complete(pf); |
| 1259 | return; |
| 1260 | } |
| 1261 | |
| 1262 | ice_check_for_hang_subtask(pf); |
| 1263 | ice_sync_fltr_subtask(pf); |
| 1264 | ice_handle_mdd_event(pf); |
| 1265 | ice_process_vflr_event(pf); |
| 1266 | ice_watchdog_subtask(pf); |
| 1267 | ice_clean_adminq_subtask(pf); |
| 1268 | ice_clean_mailboxq_subtask(pf); |
| 1269 | |
| 1270 | /* Clear __ICE_SERVICE_SCHED flag to allow scheduling next event */ |
| 1271 | ice_service_task_complete(pf); |
| 1272 | |
| 1273 | /* If the tasks have taken longer than one service timer period |
| 1274 | * or there is more work to be done, reset the service timer to |
| 1275 | * schedule the service task now. |
| 1276 | */ |
| 1277 | if (time_after(jiffies, (start_time + pf->serv_tmr_period)) || |
| 1278 | test_bit(__ICE_MDD_EVENT_PENDING, pf->state) || |
| 1279 | test_bit(__ICE_VFLR_EVENT_PENDING, pf->state) || |
| 1280 | test_bit(__ICE_MAILBOXQ_EVENT_PENDING, pf->state) || |
| 1281 | test_bit(__ICE_ADMINQ_EVENT_PENDING, pf->state)) |
| 1282 | mod_timer(&pf->serv_tmr, jiffies); |
| 1283 | } |
| 1284 | |
| 1285 | /** |
| 1286 | * ice_set_ctrlq_len - helper function to set controlq length |
| 1287 | * @hw: pointer to the HW instance |
| 1288 | */ |
| 1289 | static void ice_set_ctrlq_len(struct ice_hw *hw) |
| 1290 | { |
| 1291 | hw->adminq.num_rq_entries = ICE_AQ_LEN; |
| 1292 | hw->adminq.num_sq_entries = ICE_AQ_LEN; |
| 1293 | hw->adminq.rq_buf_size = ICE_AQ_MAX_BUF_LEN; |
| 1294 | hw->adminq.sq_buf_size = ICE_AQ_MAX_BUF_LEN; |
| 1295 | hw->mailboxq.num_rq_entries = ICE_MBXQ_LEN; |
| 1296 | hw->mailboxq.num_sq_entries = ICE_MBXQ_LEN; |
| 1297 | hw->mailboxq.rq_buf_size = ICE_MBXQ_MAX_BUF_LEN; |
| 1298 | hw->mailboxq.sq_buf_size = ICE_MBXQ_MAX_BUF_LEN; |
| 1299 | } |
| 1300 | |
| 1301 | /** |
| 1302 | * ice_irq_affinity_notify - Callback for affinity changes |
| 1303 | * @notify: context as to what irq was changed |
| 1304 | * @mask: the new affinity mask |
| 1305 | * |
| 1306 | * This is a callback function used by the irq_set_affinity_notifier function |
| 1307 | * so that we may register to receive changes to the irq affinity masks. |
| 1308 | */ |
| 1309 | static void |
| 1310 | ice_irq_affinity_notify(struct irq_affinity_notify *notify, |
| 1311 | const cpumask_t *mask) |
| 1312 | { |
| 1313 | struct ice_q_vector *q_vector = |
| 1314 | container_of(notify, struct ice_q_vector, affinity_notify); |
| 1315 | |
| 1316 | cpumask_copy(&q_vector->affinity_mask, mask); |
| 1317 | } |
| 1318 | |
| 1319 | /** |
| 1320 | * ice_irq_affinity_release - Callback for affinity notifier release |
| 1321 | * @ref: internal core kernel usage |
| 1322 | * |
| 1323 | * This is a callback function used by the irq_set_affinity_notifier function |
| 1324 | * to inform the current notification subscriber that they will no longer |
| 1325 | * receive notifications. |
| 1326 | */ |
| 1327 | static void ice_irq_affinity_release(struct kref __always_unused *ref) {} |
| 1328 | |
| 1329 | /** |
| 1330 | * ice_vsi_ena_irq - Enable IRQ for the given VSI |
| 1331 | * @vsi: the VSI being configured |
| 1332 | */ |
| 1333 | static int ice_vsi_ena_irq(struct ice_vsi *vsi) |
| 1334 | { |
| 1335 | struct ice_pf *pf = vsi->back; |
| 1336 | struct ice_hw *hw = &pf->hw; |
| 1337 | |
| 1338 | if (test_bit(ICE_FLAG_MSIX_ENA, pf->flags)) { |
| 1339 | int i; |
| 1340 | |
| 1341 | ice_for_each_q_vector(vsi, i) |
| 1342 | ice_irq_dynamic_ena(hw, vsi, vsi->q_vectors[i]); |
| 1343 | } |
| 1344 | |
| 1345 | ice_flush(hw); |
| 1346 | return 0; |
| 1347 | } |
| 1348 | |
| 1349 | /** |
| 1350 | * ice_vsi_req_irq_msix - get MSI-X vectors from the OS for the VSI |
| 1351 | * @vsi: the VSI being configured |
| 1352 | * @basename: name for the vector |
| 1353 | */ |
| 1354 | static int ice_vsi_req_irq_msix(struct ice_vsi *vsi, char *basename) |
| 1355 | { |
| 1356 | int q_vectors = vsi->num_q_vectors; |
| 1357 | struct ice_pf *pf = vsi->back; |
| 1358 | int base = vsi->sw_base_vector; |
| 1359 | int rx_int_idx = 0; |
| 1360 | int tx_int_idx = 0; |
| 1361 | int vector, err; |
| 1362 | int irq_num; |
| 1363 | |
| 1364 | for (vector = 0; vector < q_vectors; vector++) { |
| 1365 | struct ice_q_vector *q_vector = vsi->q_vectors[vector]; |
| 1366 | |
| 1367 | irq_num = pf->msix_entries[base + vector].vector; |
| 1368 | |
| 1369 | if (q_vector->tx.ring && q_vector->rx.ring) { |
| 1370 | snprintf(q_vector->name, sizeof(q_vector->name) - 1, |
| 1371 | "%s-%s-%d", basename, "TxRx", rx_int_idx++); |
| 1372 | tx_int_idx++; |
| 1373 | } else if (q_vector->rx.ring) { |
| 1374 | snprintf(q_vector->name, sizeof(q_vector->name) - 1, |
| 1375 | "%s-%s-%d", basename, "rx", rx_int_idx++); |
| 1376 | } else if (q_vector->tx.ring) { |
| 1377 | snprintf(q_vector->name, sizeof(q_vector->name) - 1, |
| 1378 | "%s-%s-%d", basename, "tx", tx_int_idx++); |
| 1379 | } else { |
| 1380 | /* skip this unused q_vector */ |
| 1381 | continue; |
| 1382 | } |
| 1383 | err = devm_request_irq(&pf->pdev->dev, irq_num, |
| 1384 | vsi->irq_handler, 0, |
| 1385 | q_vector->name, q_vector); |
| 1386 | if (err) { |
| 1387 | netdev_err(vsi->netdev, |
| 1388 | "MSIX request_irq failed, error: %d\n", err); |
| 1389 | goto free_q_irqs; |
| 1390 | } |
| 1391 | |
| 1392 | /* register for affinity change notifications */ |
| 1393 | q_vector->affinity_notify.notify = ice_irq_affinity_notify; |
| 1394 | q_vector->affinity_notify.release = ice_irq_affinity_release; |
| 1395 | irq_set_affinity_notifier(irq_num, &q_vector->affinity_notify); |
| 1396 | |
| 1397 | /* assign the mask for this irq */ |
| 1398 | irq_set_affinity_hint(irq_num, &q_vector->affinity_mask); |
| 1399 | } |
| 1400 | |
| 1401 | vsi->irqs_ready = true; |
| 1402 | return 0; |
| 1403 | |
| 1404 | free_q_irqs: |
| 1405 | while (vector) { |
| 1406 | vector--; |
| 1407 | irq_num = pf->msix_entries[base + vector].vector, |
| 1408 | irq_set_affinity_notifier(irq_num, NULL); |
| 1409 | irq_set_affinity_hint(irq_num, NULL); |
| 1410 | devm_free_irq(&pf->pdev->dev, irq_num, &vsi->q_vectors[vector]); |
| 1411 | } |
| 1412 | return err; |
| 1413 | } |
| 1414 | |
| 1415 | /** |
| 1416 | * ice_ena_misc_vector - enable the non-queue interrupts |
| 1417 | * @pf: board private structure |
| 1418 | */ |
| 1419 | static void ice_ena_misc_vector(struct ice_pf *pf) |
| 1420 | { |
| 1421 | struct ice_hw *hw = &pf->hw; |
| 1422 | u32 val; |
| 1423 | |
| 1424 | /* clear things first */ |
| 1425 | wr32(hw, PFINT_OICR_ENA, 0); /* disable all */ |
| 1426 | rd32(hw, PFINT_OICR); /* read to clear */ |
| 1427 | |
| 1428 | val = (PFINT_OICR_ECC_ERR_M | |
| 1429 | PFINT_OICR_MAL_DETECT_M | |
| 1430 | PFINT_OICR_GRST_M | |
| 1431 | PFINT_OICR_PCI_EXCEPTION_M | |
| 1432 | PFINT_OICR_VFLR_M | |
| 1433 | PFINT_OICR_HMC_ERR_M | |
| 1434 | PFINT_OICR_PE_CRITERR_M); |
| 1435 | |
| 1436 | wr32(hw, PFINT_OICR_ENA, val); |
| 1437 | |
| 1438 | /* SW_ITR_IDX = 0, but don't change INTENA */ |
| 1439 | wr32(hw, GLINT_DYN_CTL(pf->hw_oicr_idx), |
| 1440 | GLINT_DYN_CTL_SW_ITR_INDX_M | GLINT_DYN_CTL_INTENA_MSK_M); |
| 1441 | } |
| 1442 | |
| 1443 | /** |
| 1444 | * ice_misc_intr - misc interrupt handler |
| 1445 | * @irq: interrupt number |
| 1446 | * @data: pointer to a q_vector |
| 1447 | */ |
| 1448 | static irqreturn_t ice_misc_intr(int __always_unused irq, void *data) |
| 1449 | { |
| 1450 | struct ice_pf *pf = (struct ice_pf *)data; |
| 1451 | struct ice_hw *hw = &pf->hw; |
| 1452 | irqreturn_t ret = IRQ_NONE; |
| 1453 | u32 oicr, ena_mask; |
| 1454 | |
| 1455 | set_bit(__ICE_ADMINQ_EVENT_PENDING, pf->state); |
| 1456 | set_bit(__ICE_MAILBOXQ_EVENT_PENDING, pf->state); |
| 1457 | |
| 1458 | oicr = rd32(hw, PFINT_OICR); |
| 1459 | ena_mask = rd32(hw, PFINT_OICR_ENA); |
| 1460 | |
| 1461 | if (oicr & PFINT_OICR_MAL_DETECT_M) { |
| 1462 | ena_mask &= ~PFINT_OICR_MAL_DETECT_M; |
| 1463 | set_bit(__ICE_MDD_EVENT_PENDING, pf->state); |
| 1464 | } |
| 1465 | if (oicr & PFINT_OICR_VFLR_M) { |
| 1466 | ena_mask &= ~PFINT_OICR_VFLR_M; |
| 1467 | set_bit(__ICE_VFLR_EVENT_PENDING, pf->state); |
| 1468 | } |
| 1469 | |
| 1470 | if (oicr & PFINT_OICR_GRST_M) { |
| 1471 | u32 reset; |
| 1472 | |
| 1473 | /* we have a reset warning */ |
| 1474 | ena_mask &= ~PFINT_OICR_GRST_M; |
| 1475 | reset = (rd32(hw, GLGEN_RSTAT) & GLGEN_RSTAT_RESET_TYPE_M) >> |
| 1476 | GLGEN_RSTAT_RESET_TYPE_S; |
| 1477 | |
| 1478 | if (reset == ICE_RESET_CORER) |
| 1479 | pf->corer_count++; |
| 1480 | else if (reset == ICE_RESET_GLOBR) |
| 1481 | pf->globr_count++; |
| 1482 | else if (reset == ICE_RESET_EMPR) |
| 1483 | pf->empr_count++; |
| 1484 | else |
| 1485 | dev_dbg(&pf->pdev->dev, "Invalid reset type %d\n", |
| 1486 | reset); |
| 1487 | |
| 1488 | /* If a reset cycle isn't already in progress, we set a bit in |
| 1489 | * pf->state so that the service task can start a reset/rebuild. |
| 1490 | * We also make note of which reset happened so that peer |
| 1491 | * devices/drivers can be informed. |
| 1492 | */ |
| 1493 | if (!test_and_set_bit(__ICE_RESET_OICR_RECV, pf->state)) { |
| 1494 | if (reset == ICE_RESET_CORER) |
| 1495 | set_bit(__ICE_CORER_RECV, pf->state); |
| 1496 | else if (reset == ICE_RESET_GLOBR) |
| 1497 | set_bit(__ICE_GLOBR_RECV, pf->state); |
| 1498 | else |
| 1499 | set_bit(__ICE_EMPR_RECV, pf->state); |
| 1500 | |
| 1501 | /* There are couple of different bits at play here. |
| 1502 | * hw->reset_ongoing indicates whether the hardware is |
| 1503 | * in reset. This is set to true when a reset interrupt |
| 1504 | * is received and set back to false after the driver |
| 1505 | * has determined that the hardware is out of reset. |
| 1506 | * |
| 1507 | * __ICE_RESET_OICR_RECV in pf->state indicates |
| 1508 | * that a post reset rebuild is required before the |
| 1509 | * driver is operational again. This is set above. |
| 1510 | * |
| 1511 | * As this is the start of the reset/rebuild cycle, set |
| 1512 | * both to indicate that. |
| 1513 | */ |
| 1514 | hw->reset_ongoing = true; |
| 1515 | } |
| 1516 | } |
| 1517 | |
| 1518 | if (oicr & PFINT_OICR_HMC_ERR_M) { |
| 1519 | ena_mask &= ~PFINT_OICR_HMC_ERR_M; |
| 1520 | dev_dbg(&pf->pdev->dev, |
| 1521 | "HMC Error interrupt - info 0x%x, data 0x%x\n", |
| 1522 | rd32(hw, PFHMC_ERRORINFO), |
| 1523 | rd32(hw, PFHMC_ERRORDATA)); |
| 1524 | } |
| 1525 | |
| 1526 | /* Report and mask off any remaining unexpected interrupts */ |
| 1527 | oicr &= ena_mask; |
| 1528 | if (oicr) { |
| 1529 | dev_dbg(&pf->pdev->dev, "unhandled interrupt oicr=0x%08x\n", |
| 1530 | oicr); |
| 1531 | /* If a critical error is pending there is no choice but to |
| 1532 | * reset the device. |
| 1533 | */ |
| 1534 | if (oicr & (PFINT_OICR_PE_CRITERR_M | |
| 1535 | PFINT_OICR_PCI_EXCEPTION_M | |
| 1536 | PFINT_OICR_ECC_ERR_M)) { |
| 1537 | set_bit(__ICE_PFR_REQ, pf->state); |
| 1538 | ice_service_task_schedule(pf); |
| 1539 | } |
| 1540 | ena_mask &= ~oicr; |
| 1541 | } |
| 1542 | ret = IRQ_HANDLED; |
| 1543 | |
| 1544 | /* re-enable interrupt causes that are not handled during this pass */ |
| 1545 | wr32(hw, PFINT_OICR_ENA, ena_mask); |
| 1546 | if (!test_bit(__ICE_DOWN, pf->state)) { |
| 1547 | ice_service_task_schedule(pf); |
| 1548 | ice_irq_dynamic_ena(hw, NULL, NULL); |
| 1549 | } |
| 1550 | |
| 1551 | return ret; |
| 1552 | } |
| 1553 | |
| 1554 | /** |
| 1555 | * ice_dis_ctrlq_interrupts - disable control queue interrupts |
| 1556 | * @hw: pointer to HW structure |
| 1557 | */ |
| 1558 | static void ice_dis_ctrlq_interrupts(struct ice_hw *hw) |
| 1559 | { |
| 1560 | /* disable Admin queue Interrupt causes */ |
| 1561 | wr32(hw, PFINT_FW_CTL, |
| 1562 | rd32(hw, PFINT_FW_CTL) & ~PFINT_FW_CTL_CAUSE_ENA_M); |
| 1563 | |
| 1564 | /* disable Mailbox queue Interrupt causes */ |
| 1565 | wr32(hw, PFINT_MBX_CTL, |
| 1566 | rd32(hw, PFINT_MBX_CTL) & ~PFINT_MBX_CTL_CAUSE_ENA_M); |
| 1567 | |
| 1568 | /* disable Control queue Interrupt causes */ |
| 1569 | wr32(hw, PFINT_OICR_CTL, |
| 1570 | rd32(hw, PFINT_OICR_CTL) & ~PFINT_OICR_CTL_CAUSE_ENA_M); |
| 1571 | |
| 1572 | ice_flush(hw); |
| 1573 | } |
| 1574 | |
| 1575 | /** |
| 1576 | * ice_free_irq_msix_misc - Unroll misc vector setup |
| 1577 | * @pf: board private structure |
| 1578 | */ |
| 1579 | static void ice_free_irq_msix_misc(struct ice_pf *pf) |
| 1580 | { |
| 1581 | struct ice_hw *hw = &pf->hw; |
| 1582 | |
| 1583 | ice_dis_ctrlq_interrupts(hw); |
| 1584 | |
| 1585 | /* disable OICR interrupt */ |
| 1586 | wr32(hw, PFINT_OICR_ENA, 0); |
| 1587 | ice_flush(hw); |
| 1588 | |
| 1589 | if (test_bit(ICE_FLAG_MSIX_ENA, pf->flags) && pf->msix_entries) { |
| 1590 | synchronize_irq(pf->msix_entries[pf->sw_oicr_idx].vector); |
| 1591 | devm_free_irq(&pf->pdev->dev, |
| 1592 | pf->msix_entries[pf->sw_oicr_idx].vector, pf); |
| 1593 | } |
| 1594 | |
| 1595 | pf->num_avail_sw_msix += 1; |
| 1596 | ice_free_res(pf->sw_irq_tracker, pf->sw_oicr_idx, ICE_RES_MISC_VEC_ID); |
| 1597 | pf->num_avail_hw_msix += 1; |
| 1598 | ice_free_res(pf->hw_irq_tracker, pf->hw_oicr_idx, ICE_RES_MISC_VEC_ID); |
| 1599 | } |
| 1600 | |
| 1601 | /** |
| 1602 | * ice_ena_ctrlq_interrupts - enable control queue interrupts |
| 1603 | * @hw: pointer to HW structure |
| 1604 | * @v_idx: HW vector index to associate the control queue interrupts with |
| 1605 | */ |
| 1606 | static void ice_ena_ctrlq_interrupts(struct ice_hw *hw, u16 v_idx) |
| 1607 | { |
| 1608 | u32 val; |
| 1609 | |
| 1610 | val = ((v_idx & PFINT_OICR_CTL_MSIX_INDX_M) | |
| 1611 | PFINT_OICR_CTL_CAUSE_ENA_M); |
| 1612 | wr32(hw, PFINT_OICR_CTL, val); |
| 1613 | |
| 1614 | /* enable Admin queue Interrupt causes */ |
| 1615 | val = ((v_idx & PFINT_FW_CTL_MSIX_INDX_M) | |
| 1616 | PFINT_FW_CTL_CAUSE_ENA_M); |
| 1617 | wr32(hw, PFINT_FW_CTL, val); |
| 1618 | |
| 1619 | /* enable Mailbox queue Interrupt causes */ |
| 1620 | val = ((v_idx & PFINT_MBX_CTL_MSIX_INDX_M) | |
| 1621 | PFINT_MBX_CTL_CAUSE_ENA_M); |
| 1622 | wr32(hw, PFINT_MBX_CTL, val); |
| 1623 | |
| 1624 | ice_flush(hw); |
| 1625 | } |
| 1626 | |
| 1627 | /** |
| 1628 | * ice_req_irq_msix_misc - Setup the misc vector to handle non queue events |
| 1629 | * @pf: board private structure |
| 1630 | * |
| 1631 | * This sets up the handler for MSIX 0, which is used to manage the |
| 1632 | * non-queue interrupts, e.g. AdminQ and errors. This is not used |
| 1633 | * when in MSI or Legacy interrupt mode. |
| 1634 | */ |
| 1635 | static int ice_req_irq_msix_misc(struct ice_pf *pf) |
| 1636 | { |
| 1637 | struct ice_hw *hw = &pf->hw; |
| 1638 | int oicr_idx, err = 0; |
| 1639 | |
| 1640 | if (!pf->int_name[0]) |
| 1641 | snprintf(pf->int_name, sizeof(pf->int_name) - 1, "%s-%s:misc", |
| 1642 | dev_driver_string(&pf->pdev->dev), |
| 1643 | dev_name(&pf->pdev->dev)); |
| 1644 | |
| 1645 | /* Do not request IRQ but do enable OICR interrupt since settings are |
| 1646 | * lost during reset. Note that this function is called only during |
| 1647 | * rebuild path and not while reset is in progress. |
| 1648 | */ |
| 1649 | if (ice_is_reset_in_progress(pf->state)) |
| 1650 | goto skip_req_irq; |
| 1651 | |
| 1652 | /* reserve one vector in sw_irq_tracker for misc interrupts */ |
| 1653 | oicr_idx = ice_get_res(pf, pf->sw_irq_tracker, 1, ICE_RES_MISC_VEC_ID); |
| 1654 | if (oicr_idx < 0) |
| 1655 | return oicr_idx; |
| 1656 | |
| 1657 | pf->num_avail_sw_msix -= 1; |
| 1658 | pf->sw_oicr_idx = oicr_idx; |
| 1659 | |
| 1660 | /* reserve one vector in hw_irq_tracker for misc interrupts */ |
| 1661 | oicr_idx = ice_get_res(pf, pf->hw_irq_tracker, 1, ICE_RES_MISC_VEC_ID); |
| 1662 | if (oicr_idx < 0) { |
| 1663 | ice_free_res(pf->sw_irq_tracker, 1, ICE_RES_MISC_VEC_ID); |
| 1664 | pf->num_avail_sw_msix += 1; |
| 1665 | return oicr_idx; |
| 1666 | } |
| 1667 | pf->num_avail_hw_msix -= 1; |
| 1668 | pf->hw_oicr_idx = oicr_idx; |
| 1669 | |
| 1670 | err = devm_request_irq(&pf->pdev->dev, |
| 1671 | pf->msix_entries[pf->sw_oicr_idx].vector, |
| 1672 | ice_misc_intr, 0, pf->int_name, pf); |
| 1673 | if (err) { |
| 1674 | dev_err(&pf->pdev->dev, |
| 1675 | "devm_request_irq for %s failed: %d\n", |
| 1676 | pf->int_name, err); |
| 1677 | ice_free_res(pf->sw_irq_tracker, 1, ICE_RES_MISC_VEC_ID); |
| 1678 | pf->num_avail_sw_msix += 1; |
| 1679 | ice_free_res(pf->hw_irq_tracker, 1, ICE_RES_MISC_VEC_ID); |
| 1680 | pf->num_avail_hw_msix += 1; |
| 1681 | return err; |
| 1682 | } |
| 1683 | |
| 1684 | skip_req_irq: |
| 1685 | ice_ena_misc_vector(pf); |
| 1686 | |
| 1687 | ice_ena_ctrlq_interrupts(hw, pf->hw_oicr_idx); |
| 1688 | wr32(hw, GLINT_ITR(ICE_RX_ITR, pf->hw_oicr_idx), |
| 1689 | ITR_REG_ALIGN(ICE_ITR_8K) >> ICE_ITR_GRAN_S); |
| 1690 | |
| 1691 | ice_flush(hw); |
| 1692 | ice_irq_dynamic_ena(hw, NULL, NULL); |
| 1693 | |
| 1694 | return 0; |
| 1695 | } |
| 1696 | |
| 1697 | /** |
| 1698 | * ice_napi_del - Remove NAPI handler for the VSI |
| 1699 | * @vsi: VSI for which NAPI handler is to be removed |
| 1700 | */ |
| 1701 | void ice_napi_del(struct ice_vsi *vsi) |
| 1702 | { |
| 1703 | int v_idx; |
| 1704 | |
| 1705 | if (!vsi->netdev) |
| 1706 | return; |
| 1707 | |
| 1708 | ice_for_each_q_vector(vsi, v_idx) |
| 1709 | netif_napi_del(&vsi->q_vectors[v_idx]->napi); |
| 1710 | } |
| 1711 | |
| 1712 | /** |
| 1713 | * ice_napi_add - register NAPI handler for the VSI |
| 1714 | * @vsi: VSI for which NAPI handler is to be registered |
| 1715 | * |
| 1716 | * This function is only called in the driver's load path. Registering the NAPI |
| 1717 | * handler is done in ice_vsi_alloc_q_vector() for all other cases (i.e. resume, |
| 1718 | * reset/rebuild, etc.) |
| 1719 | */ |
| 1720 | static void ice_napi_add(struct ice_vsi *vsi) |
| 1721 | { |
| 1722 | int v_idx; |
| 1723 | |
| 1724 | if (!vsi->netdev) |
| 1725 | return; |
| 1726 | |
| 1727 | ice_for_each_q_vector(vsi, v_idx) |
| 1728 | netif_napi_add(vsi->netdev, &vsi->q_vectors[v_idx]->napi, |
| 1729 | ice_napi_poll, NAPI_POLL_WEIGHT); |
| 1730 | } |
| 1731 | |
| 1732 | /** |
| 1733 | * ice_cfg_netdev - Allocate, configure and register a netdev |
| 1734 | * @vsi: the VSI associated with the new netdev |
| 1735 | * |
| 1736 | * Returns 0 on success, negative value on failure |
| 1737 | */ |
| 1738 | static int ice_cfg_netdev(struct ice_vsi *vsi) |
| 1739 | { |
| 1740 | netdev_features_t csumo_features; |
| 1741 | netdev_features_t vlano_features; |
| 1742 | netdev_features_t dflt_features; |
| 1743 | netdev_features_t tso_features; |
| 1744 | struct ice_netdev_priv *np; |
| 1745 | struct net_device *netdev; |
| 1746 | u8 mac_addr[ETH_ALEN]; |
| 1747 | int err; |
| 1748 | |
| 1749 | netdev = alloc_etherdev_mqs(sizeof(*np), vsi->alloc_txq, |
| 1750 | vsi->alloc_rxq); |
| 1751 | if (!netdev) |
| 1752 | return -ENOMEM; |
| 1753 | |
| 1754 | vsi->netdev = netdev; |
| 1755 | np = netdev_priv(netdev); |
| 1756 | np->vsi = vsi; |
| 1757 | |
| 1758 | dflt_features = NETIF_F_SG | |
| 1759 | NETIF_F_HIGHDMA | |
| 1760 | NETIF_F_RXHASH; |
| 1761 | |
| 1762 | csumo_features = NETIF_F_RXCSUM | |
| 1763 | NETIF_F_IP_CSUM | |
| 1764 | NETIF_F_SCTP_CRC | |
| 1765 | NETIF_F_IPV6_CSUM; |
| 1766 | |
| 1767 | vlano_features = NETIF_F_HW_VLAN_CTAG_FILTER | |
| 1768 | NETIF_F_HW_VLAN_CTAG_TX | |
| 1769 | NETIF_F_HW_VLAN_CTAG_RX; |
| 1770 | |
| 1771 | tso_features = NETIF_F_TSO; |
| 1772 | |
| 1773 | /* set features that user can change */ |
| 1774 | netdev->hw_features = dflt_features | csumo_features | |
| 1775 | vlano_features | tso_features; |
| 1776 | |
| 1777 | /* enable features */ |
| 1778 | netdev->features |= netdev->hw_features; |
| 1779 | /* encap and VLAN devices inherit default, csumo and tso features */ |
| 1780 | netdev->hw_enc_features |= dflt_features | csumo_features | |
| 1781 | tso_features; |
| 1782 | netdev->vlan_features |= dflt_features | csumo_features | |
| 1783 | tso_features; |
| 1784 | |
| 1785 | if (vsi->type == ICE_VSI_PF) { |
| 1786 | SET_NETDEV_DEV(netdev, &vsi->back->pdev->dev); |
| 1787 | ether_addr_copy(mac_addr, vsi->port_info->mac.perm_addr); |
| 1788 | |
| 1789 | ether_addr_copy(netdev->dev_addr, mac_addr); |
| 1790 | ether_addr_copy(netdev->perm_addr, mac_addr); |
| 1791 | } |
| 1792 | |
| 1793 | netdev->priv_flags |= IFF_UNICAST_FLT; |
| 1794 | |
| 1795 | /* assign netdev_ops */ |
| 1796 | netdev->netdev_ops = &ice_netdev_ops; |
| 1797 | |
| 1798 | /* setup watchdog timeout value to be 5 second */ |
| 1799 | netdev->watchdog_timeo = 5 * HZ; |
| 1800 | |
| 1801 | ice_set_ethtool_ops(netdev); |
| 1802 | |
| 1803 | netdev->min_mtu = ETH_MIN_MTU; |
| 1804 | netdev->max_mtu = ICE_MAX_MTU; |
| 1805 | |
| 1806 | err = register_netdev(vsi->netdev); |
| 1807 | if (err) |
| 1808 | return err; |
| 1809 | |
| 1810 | netif_carrier_off(vsi->netdev); |
| 1811 | |
| 1812 | /* make sure transmit queues start off as stopped */ |
| 1813 | netif_tx_stop_all_queues(vsi->netdev); |
| 1814 | |
| 1815 | return 0; |
| 1816 | } |
| 1817 | |
| 1818 | /** |
| 1819 | * ice_fill_rss_lut - Fill the RSS lookup table with default values |
| 1820 | * @lut: Lookup table |
| 1821 | * @rss_table_size: Lookup table size |
| 1822 | * @rss_size: Range of queue number for hashing |
| 1823 | */ |
| 1824 | void ice_fill_rss_lut(u8 *lut, u16 rss_table_size, u16 rss_size) |
| 1825 | { |
| 1826 | u16 i; |
| 1827 | |
| 1828 | for (i = 0; i < rss_table_size; i++) |
| 1829 | lut[i] = i % rss_size; |
| 1830 | } |
| 1831 | |
| 1832 | /** |
| 1833 | * ice_pf_vsi_setup - Set up a PF VSI |
| 1834 | * @pf: board private structure |
| 1835 | * @pi: pointer to the port_info instance |
| 1836 | * |
| 1837 | * Returns pointer to the successfully allocated VSI sw struct on success, |
| 1838 | * otherwise returns NULL on failure. |
| 1839 | */ |
| 1840 | static struct ice_vsi * |
| 1841 | ice_pf_vsi_setup(struct ice_pf *pf, struct ice_port_info *pi) |
| 1842 | { |
| 1843 | return ice_vsi_setup(pf, pi, ICE_VSI_PF, ICE_INVAL_VFID); |
| 1844 | } |
| 1845 | |
| 1846 | /** |
| 1847 | * ice_vlan_rx_add_vid - Add a VLAN ID filter to HW offload |
| 1848 | * @netdev: network interface to be adjusted |
| 1849 | * @proto: unused protocol |
| 1850 | * @vid: VLAN ID to be added |
| 1851 | * |
| 1852 | * net_device_ops implementation for adding VLAN IDs |
| 1853 | */ |
| 1854 | static int |
| 1855 | ice_vlan_rx_add_vid(struct net_device *netdev, __always_unused __be16 proto, |
| 1856 | u16 vid) |
| 1857 | { |
| 1858 | struct ice_netdev_priv *np = netdev_priv(netdev); |
| 1859 | struct ice_vsi *vsi = np->vsi; |
| 1860 | int ret; |
| 1861 | |
| 1862 | if (vid >= VLAN_N_VID) { |
| 1863 | netdev_err(netdev, "VLAN id requested %d is out of range %d\n", |
| 1864 | vid, VLAN_N_VID); |
| 1865 | return -EINVAL; |
| 1866 | } |
| 1867 | |
| 1868 | if (vsi->info.pvid) |
| 1869 | return -EINVAL; |
| 1870 | |
| 1871 | /* Enable VLAN pruning when VLAN 0 is added */ |
| 1872 | if (unlikely(!vid)) { |
| 1873 | ret = ice_cfg_vlan_pruning(vsi, true, false); |
| 1874 | if (ret) |
| 1875 | return ret; |
| 1876 | } |
| 1877 | |
| 1878 | /* Add all VLAN IDs including 0 to the switch filter. VLAN ID 0 is |
| 1879 | * needed to continue allowing all untagged packets since VLAN prune |
| 1880 | * list is applied to all packets by the switch |
| 1881 | */ |
| 1882 | ret = ice_vsi_add_vlan(vsi, vid); |
| 1883 | if (!ret) { |
| 1884 | vsi->vlan_ena = true; |
| 1885 | set_bit(ICE_VSI_FLAG_VLAN_FLTR_CHANGED, vsi->flags); |
| 1886 | } |
| 1887 | |
| 1888 | return ret; |
| 1889 | } |
| 1890 | |
| 1891 | /** |
| 1892 | * ice_vlan_rx_kill_vid - Remove a VLAN ID filter from HW offload |
| 1893 | * @netdev: network interface to be adjusted |
| 1894 | * @proto: unused protocol |
| 1895 | * @vid: VLAN ID to be removed |
| 1896 | * |
| 1897 | * net_device_ops implementation for removing VLAN IDs |
| 1898 | */ |
| 1899 | static int |
| 1900 | ice_vlan_rx_kill_vid(struct net_device *netdev, __always_unused __be16 proto, |
| 1901 | u16 vid) |
| 1902 | { |
| 1903 | struct ice_netdev_priv *np = netdev_priv(netdev); |
| 1904 | struct ice_vsi *vsi = np->vsi; |
| 1905 | int ret; |
| 1906 | |
| 1907 | if (vsi->info.pvid) |
| 1908 | return -EINVAL; |
| 1909 | |
| 1910 | /* Make sure ice_vsi_kill_vlan is successful before updating VLAN |
| 1911 | * information |
| 1912 | */ |
| 1913 | ret = ice_vsi_kill_vlan(vsi, vid); |
| 1914 | if (ret) |
| 1915 | return ret; |
| 1916 | |
| 1917 | /* Disable VLAN pruning when VLAN 0 is removed */ |
| 1918 | if (unlikely(!vid)) |
| 1919 | ret = ice_cfg_vlan_pruning(vsi, false, false); |
| 1920 | |
| 1921 | vsi->vlan_ena = false; |
| 1922 | set_bit(ICE_VSI_FLAG_VLAN_FLTR_CHANGED, vsi->flags); |
| 1923 | return ret; |
| 1924 | } |
| 1925 | |
| 1926 | /** |
| 1927 | * ice_setup_pf_sw - Setup the HW switch on startup or after reset |
| 1928 | * @pf: board private structure |
| 1929 | * |
| 1930 | * Returns 0 on success, negative value on failure |
| 1931 | */ |
| 1932 | static int ice_setup_pf_sw(struct ice_pf *pf) |
| 1933 | { |
| 1934 | LIST_HEAD(tmp_add_list); |
| 1935 | u8 broadcast[ETH_ALEN]; |
| 1936 | struct ice_vsi *vsi; |
| 1937 | int status = 0; |
| 1938 | |
| 1939 | if (ice_is_reset_in_progress(pf->state)) |
| 1940 | return -EBUSY; |
| 1941 | |
| 1942 | vsi = ice_pf_vsi_setup(pf, pf->hw.port_info); |
| 1943 | if (!vsi) { |
| 1944 | status = -ENOMEM; |
| 1945 | goto unroll_vsi_setup; |
| 1946 | } |
| 1947 | |
| 1948 | status = ice_cfg_netdev(vsi); |
| 1949 | if (status) { |
| 1950 | status = -ENODEV; |
| 1951 | goto unroll_vsi_setup; |
| 1952 | } |
| 1953 | |
| 1954 | /* registering the NAPI handler requires both the queues and |
| 1955 | * netdev to be created, which are done in ice_pf_vsi_setup() |
| 1956 | * and ice_cfg_netdev() respectively |
| 1957 | */ |
| 1958 | ice_napi_add(vsi); |
| 1959 | |
| 1960 | /* To add a MAC filter, first add the MAC to a list and then |
| 1961 | * pass the list to ice_add_mac. |
| 1962 | */ |
| 1963 | |
| 1964 | /* Add a unicast MAC filter so the VSI can get its packets */ |
| 1965 | status = ice_add_mac_to_list(vsi, &tmp_add_list, |
| 1966 | vsi->port_info->mac.perm_addr); |
| 1967 | if (status) |
| 1968 | goto unroll_napi_add; |
| 1969 | |
| 1970 | /* VSI needs to receive broadcast traffic, so add the broadcast |
| 1971 | * MAC address to the list as well. |
| 1972 | */ |
| 1973 | eth_broadcast_addr(broadcast); |
| 1974 | status = ice_add_mac_to_list(vsi, &tmp_add_list, broadcast); |
| 1975 | if (status) |
| 1976 | goto free_mac_list; |
| 1977 | |
| 1978 | /* program MAC filters for entries in tmp_add_list */ |
| 1979 | status = ice_add_mac(&pf->hw, &tmp_add_list); |
| 1980 | if (status) { |
| 1981 | dev_err(&pf->pdev->dev, "Could not add MAC filters\n"); |
| 1982 | status = -ENOMEM; |
| 1983 | goto free_mac_list; |
| 1984 | } |
| 1985 | |
| 1986 | ice_free_fltr_list(&pf->pdev->dev, &tmp_add_list); |
| 1987 | return status; |
| 1988 | |
| 1989 | free_mac_list: |
| 1990 | ice_free_fltr_list(&pf->pdev->dev, &tmp_add_list); |
| 1991 | |
| 1992 | unroll_napi_add: |
| 1993 | if (vsi) { |
| 1994 | ice_napi_del(vsi); |
| 1995 | if (vsi->netdev) { |
| 1996 | if (vsi->netdev->reg_state == NETREG_REGISTERED) |
| 1997 | unregister_netdev(vsi->netdev); |
| 1998 | free_netdev(vsi->netdev); |
| 1999 | vsi->netdev = NULL; |
| 2000 | } |
| 2001 | } |
| 2002 | |
| 2003 | unroll_vsi_setup: |
| 2004 | if (vsi) { |
| 2005 | ice_vsi_free_q_vectors(vsi); |
| 2006 | ice_vsi_delete(vsi); |
| 2007 | ice_vsi_put_qs(vsi); |
| 2008 | pf->q_left_tx += vsi->alloc_txq; |
| 2009 | pf->q_left_rx += vsi->alloc_rxq; |
| 2010 | ice_vsi_clear(vsi); |
| 2011 | } |
| 2012 | return status; |
| 2013 | } |
| 2014 | |
| 2015 | /** |
| 2016 | * ice_determine_q_usage - Calculate queue distribution |
| 2017 | * @pf: board private structure |
| 2018 | * |
| 2019 | * Return -ENOMEM if we don't get enough queues for all ports |
| 2020 | */ |
| 2021 | static void ice_determine_q_usage(struct ice_pf *pf) |
| 2022 | { |
| 2023 | u16 q_left_tx, q_left_rx; |
| 2024 | |
| 2025 | q_left_tx = pf->hw.func_caps.common_cap.num_txq; |
| 2026 | q_left_rx = pf->hw.func_caps.common_cap.num_rxq; |
| 2027 | |
| 2028 | pf->num_lan_tx = min_t(int, q_left_tx, num_online_cpus()); |
| 2029 | |
| 2030 | /* only 1 Rx queue unless RSS is enabled */ |
| 2031 | if (!test_bit(ICE_FLAG_RSS_ENA, pf->flags)) |
| 2032 | pf->num_lan_rx = 1; |
| 2033 | else |
| 2034 | pf->num_lan_rx = min_t(int, q_left_rx, num_online_cpus()); |
| 2035 | |
| 2036 | pf->q_left_tx = q_left_tx - pf->num_lan_tx; |
| 2037 | pf->q_left_rx = q_left_rx - pf->num_lan_rx; |
| 2038 | } |
| 2039 | |
| 2040 | /** |
| 2041 | * ice_deinit_pf - Unrolls initialziations done by ice_init_pf |
| 2042 | * @pf: board private structure to initialize |
| 2043 | */ |
| 2044 | static void ice_deinit_pf(struct ice_pf *pf) |
| 2045 | { |
| 2046 | ice_service_task_stop(pf); |
| 2047 | mutex_destroy(&pf->sw_mutex); |
| 2048 | mutex_destroy(&pf->avail_q_mutex); |
| 2049 | } |
| 2050 | |
| 2051 | /** |
| 2052 | * ice_init_pf - Initialize general software structures (struct ice_pf) |
| 2053 | * @pf: board private structure to initialize |
| 2054 | */ |
| 2055 | static void ice_init_pf(struct ice_pf *pf) |
| 2056 | { |
| 2057 | bitmap_zero(pf->flags, ICE_PF_FLAGS_NBITS); |
| 2058 | set_bit(ICE_FLAG_MSIX_ENA, pf->flags); |
| 2059 | #ifdef CONFIG_PCI_IOV |
| 2060 | if (pf->hw.func_caps.common_cap.sr_iov_1_1) { |
| 2061 | struct ice_hw *hw = &pf->hw; |
| 2062 | |
| 2063 | set_bit(ICE_FLAG_SRIOV_CAPABLE, pf->flags); |
| 2064 | pf->num_vfs_supported = min_t(int, hw->func_caps.num_allocd_vfs, |
| 2065 | ICE_MAX_VF_COUNT); |
| 2066 | } |
| 2067 | #endif /* CONFIG_PCI_IOV */ |
| 2068 | |
| 2069 | mutex_init(&pf->sw_mutex); |
| 2070 | mutex_init(&pf->avail_q_mutex); |
| 2071 | |
| 2072 | /* Clear avail_[t|r]x_qs bitmaps (set all to avail) */ |
| 2073 | mutex_lock(&pf->avail_q_mutex); |
| 2074 | bitmap_zero(pf->avail_txqs, ICE_MAX_TXQS); |
| 2075 | bitmap_zero(pf->avail_rxqs, ICE_MAX_RXQS); |
| 2076 | mutex_unlock(&pf->avail_q_mutex); |
| 2077 | |
| 2078 | if (pf->hw.func_caps.common_cap.rss_table_size) |
| 2079 | set_bit(ICE_FLAG_RSS_ENA, pf->flags); |
| 2080 | |
| 2081 | /* setup service timer and periodic service task */ |
| 2082 | timer_setup(&pf->serv_tmr, ice_service_timer, 0); |
| 2083 | pf->serv_tmr_period = HZ; |
| 2084 | INIT_WORK(&pf->serv_task, ice_service_task); |
| 2085 | clear_bit(__ICE_SERVICE_SCHED, pf->state); |
| 2086 | } |
| 2087 | |
| 2088 | /** |
| 2089 | * ice_ena_msix_range - Request a range of MSIX vectors from the OS |
| 2090 | * @pf: board private structure |
| 2091 | * |
| 2092 | * compute the number of MSIX vectors required (v_budget) and request from |
| 2093 | * the OS. Return the number of vectors reserved or negative on failure |
| 2094 | */ |
| 2095 | static int ice_ena_msix_range(struct ice_pf *pf) |
| 2096 | { |
| 2097 | int v_left, v_actual, v_budget = 0; |
| 2098 | int needed, err, i; |
| 2099 | |
| 2100 | v_left = pf->hw.func_caps.common_cap.num_msix_vectors; |
| 2101 | |
| 2102 | /* reserve one vector for miscellaneous handler */ |
| 2103 | needed = 1; |
| 2104 | v_budget += needed; |
| 2105 | v_left -= needed; |
| 2106 | |
| 2107 | /* reserve vectors for LAN traffic */ |
| 2108 | pf->num_lan_msix = min_t(int, num_online_cpus(), v_left); |
| 2109 | v_budget += pf->num_lan_msix; |
| 2110 | v_left -= pf->num_lan_msix; |
| 2111 | |
| 2112 | pf->msix_entries = devm_kcalloc(&pf->pdev->dev, v_budget, |
| 2113 | sizeof(*pf->msix_entries), GFP_KERNEL); |
| 2114 | |
| 2115 | if (!pf->msix_entries) { |
| 2116 | err = -ENOMEM; |
| 2117 | goto exit_err; |
| 2118 | } |
| 2119 | |
| 2120 | for (i = 0; i < v_budget; i++) |
| 2121 | pf->msix_entries[i].entry = i; |
| 2122 | |
| 2123 | /* actually reserve the vectors */ |
| 2124 | v_actual = pci_enable_msix_range(pf->pdev, pf->msix_entries, |
| 2125 | ICE_MIN_MSIX, v_budget); |
| 2126 | |
| 2127 | if (v_actual < 0) { |
| 2128 | dev_err(&pf->pdev->dev, "unable to reserve MSI-X vectors\n"); |
| 2129 | err = v_actual; |
| 2130 | goto msix_err; |
| 2131 | } |
| 2132 | |
| 2133 | if (v_actual < v_budget) { |
| 2134 | dev_warn(&pf->pdev->dev, |
| 2135 | "not enough vectors. requested = %d, obtained = %d\n", |
| 2136 | v_budget, v_actual); |
| 2137 | if (v_actual >= (pf->num_lan_msix + 1)) { |
| 2138 | pf->num_avail_sw_msix = v_actual - |
| 2139 | (pf->num_lan_msix + 1); |
| 2140 | } else if (v_actual >= 2) { |
| 2141 | pf->num_lan_msix = 1; |
| 2142 | pf->num_avail_sw_msix = v_actual - 2; |
| 2143 | } else { |
| 2144 | pci_disable_msix(pf->pdev); |
| 2145 | err = -ERANGE; |
| 2146 | goto msix_err; |
| 2147 | } |
| 2148 | } |
| 2149 | |
| 2150 | return v_actual; |
| 2151 | |
| 2152 | msix_err: |
| 2153 | devm_kfree(&pf->pdev->dev, pf->msix_entries); |
| 2154 | goto exit_err; |
| 2155 | |
| 2156 | exit_err: |
| 2157 | pf->num_lan_msix = 0; |
| 2158 | clear_bit(ICE_FLAG_MSIX_ENA, pf->flags); |
| 2159 | return err; |
| 2160 | } |
| 2161 | |
| 2162 | /** |
| 2163 | * ice_dis_msix - Disable MSI-X interrupt setup in OS |
| 2164 | * @pf: board private structure |
| 2165 | */ |
| 2166 | static void ice_dis_msix(struct ice_pf *pf) |
| 2167 | { |
| 2168 | pci_disable_msix(pf->pdev); |
| 2169 | devm_kfree(&pf->pdev->dev, pf->msix_entries); |
| 2170 | pf->msix_entries = NULL; |
| 2171 | clear_bit(ICE_FLAG_MSIX_ENA, pf->flags); |
| 2172 | } |
| 2173 | |
| 2174 | /** |
| 2175 | * ice_clear_interrupt_scheme - Undo things done by ice_init_interrupt_scheme |
| 2176 | * @pf: board private structure |
| 2177 | */ |
| 2178 | static void ice_clear_interrupt_scheme(struct ice_pf *pf) |
| 2179 | { |
| 2180 | if (test_bit(ICE_FLAG_MSIX_ENA, pf->flags)) |
| 2181 | ice_dis_msix(pf); |
| 2182 | |
| 2183 | if (pf->sw_irq_tracker) { |
| 2184 | devm_kfree(&pf->pdev->dev, pf->sw_irq_tracker); |
| 2185 | pf->sw_irq_tracker = NULL; |
| 2186 | } |
| 2187 | |
| 2188 | if (pf->hw_irq_tracker) { |
| 2189 | devm_kfree(&pf->pdev->dev, pf->hw_irq_tracker); |
| 2190 | pf->hw_irq_tracker = NULL; |
| 2191 | } |
| 2192 | } |
| 2193 | |
| 2194 | /** |
| 2195 | * ice_init_interrupt_scheme - Determine proper interrupt scheme |
| 2196 | * @pf: board private structure to initialize |
| 2197 | */ |
| 2198 | static int ice_init_interrupt_scheme(struct ice_pf *pf) |
| 2199 | { |
| 2200 | int vectors = 0, hw_vectors = 0; |
| 2201 | |
| 2202 | if (test_bit(ICE_FLAG_MSIX_ENA, pf->flags)) |
| 2203 | vectors = ice_ena_msix_range(pf); |
| 2204 | else |
| 2205 | return -ENODEV; |
| 2206 | |
| 2207 | if (vectors < 0) |
| 2208 | return vectors; |
| 2209 | |
| 2210 | /* set up vector assignment tracking */ |
| 2211 | pf->sw_irq_tracker = |
| 2212 | devm_kzalloc(&pf->pdev->dev, sizeof(*pf->sw_irq_tracker) + |
| 2213 | (sizeof(u16) * vectors), GFP_KERNEL); |
| 2214 | if (!pf->sw_irq_tracker) { |
| 2215 | ice_dis_msix(pf); |
| 2216 | return -ENOMEM; |
| 2217 | } |
| 2218 | |
| 2219 | /* populate SW interrupts pool with number of OS granted IRQs. */ |
| 2220 | pf->num_avail_sw_msix = vectors; |
| 2221 | pf->sw_irq_tracker->num_entries = vectors; |
| 2222 | |
| 2223 | /* set up HW vector assignment tracking */ |
| 2224 | hw_vectors = pf->hw.func_caps.common_cap.num_msix_vectors; |
| 2225 | pf->hw_irq_tracker = |
| 2226 | devm_kzalloc(&pf->pdev->dev, sizeof(*pf->hw_irq_tracker) + |
| 2227 | (sizeof(u16) * hw_vectors), GFP_KERNEL); |
| 2228 | if (!pf->hw_irq_tracker) { |
| 2229 | ice_clear_interrupt_scheme(pf); |
| 2230 | return -ENOMEM; |
| 2231 | } |
| 2232 | |
| 2233 | /* populate HW interrupts pool with number of HW supported irqs. */ |
| 2234 | pf->num_avail_hw_msix = hw_vectors; |
| 2235 | pf->hw_irq_tracker->num_entries = hw_vectors; |
| 2236 | |
| 2237 | return 0; |
| 2238 | } |
| 2239 | |
| 2240 | /** |
| 2241 | * ice_verify_cacheline_size - verify driver's assumption of 64 Byte cache lines |
| 2242 | * @pf: pointer to the PF structure |
| 2243 | * |
| 2244 | * There is no error returned here because the driver should be able to handle |
| 2245 | * 128 Byte cache lines, so we only print a warning in case issues are seen, |
| 2246 | * specifically with Tx. |
| 2247 | */ |
| 2248 | static void ice_verify_cacheline_size(struct ice_pf *pf) |
| 2249 | { |
| 2250 | if (rd32(&pf->hw, GLPCI_CNF2) & GLPCI_CNF2_CACHELINE_SIZE_M) |
| 2251 | dev_warn(&pf->pdev->dev, |
| 2252 | "%d Byte cache line assumption is invalid, driver may have Tx timeouts!\n", |
| 2253 | ICE_CACHE_LINE_BYTES); |
| 2254 | } |
| 2255 | |
| 2256 | /** |
| 2257 | * ice_probe - Device initialization routine |
| 2258 | * @pdev: PCI device information struct |
| 2259 | * @ent: entry in ice_pci_tbl |
| 2260 | * |
| 2261 | * Returns 0 on success, negative on failure |
| 2262 | */ |
| 2263 | static int |
| 2264 | ice_probe(struct pci_dev *pdev, const struct pci_device_id __always_unused *ent) |
| 2265 | { |
| 2266 | struct device *dev = &pdev->dev; |
| 2267 | struct ice_pf *pf; |
| 2268 | struct ice_hw *hw; |
| 2269 | int err; |
| 2270 | |
| 2271 | /* this driver uses devres, see Documentation/driver-model/devres.txt */ |
| 2272 | err = pcim_enable_device(pdev); |
| 2273 | if (err) |
| 2274 | return err; |
| 2275 | |
| 2276 | err = pcim_iomap_regions(pdev, BIT(ICE_BAR0), pci_name(pdev)); |
| 2277 | if (err) { |
| 2278 | dev_err(dev, "BAR0 I/O map error %d\n", err); |
| 2279 | return err; |
| 2280 | } |
| 2281 | |
| 2282 | pf = devm_kzalloc(dev, sizeof(*pf), GFP_KERNEL); |
| 2283 | if (!pf) |
| 2284 | return -ENOMEM; |
| 2285 | |
| 2286 | /* set up for high or low dma */ |
| 2287 | err = dma_set_mask_and_coherent(dev, DMA_BIT_MASK(64)); |
| 2288 | if (err) |
| 2289 | err = dma_set_mask_and_coherent(dev, DMA_BIT_MASK(32)); |
| 2290 | if (err) { |
| 2291 | dev_err(dev, "DMA configuration failed: 0x%x\n", err); |
| 2292 | return err; |
| 2293 | } |
| 2294 | |
| 2295 | pci_enable_pcie_error_reporting(pdev); |
| 2296 | pci_set_master(pdev); |
| 2297 | |
| 2298 | pf->pdev = pdev; |
| 2299 | pci_set_drvdata(pdev, pf); |
| 2300 | set_bit(__ICE_DOWN, pf->state); |
| 2301 | /* Disable service task until DOWN bit is cleared */ |
| 2302 | set_bit(__ICE_SERVICE_DIS, pf->state); |
| 2303 | |
| 2304 | hw = &pf->hw; |
| 2305 | hw->hw_addr = pcim_iomap_table(pdev)[ICE_BAR0]; |
| 2306 | hw->back = pf; |
| 2307 | hw->vendor_id = pdev->vendor; |
| 2308 | hw->device_id = pdev->device; |
| 2309 | pci_read_config_byte(pdev, PCI_REVISION_ID, &hw->revision_id); |
| 2310 | hw->subsystem_vendor_id = pdev->subsystem_vendor; |
| 2311 | hw->subsystem_device_id = pdev->subsystem_device; |
| 2312 | hw->bus.device = PCI_SLOT(pdev->devfn); |
| 2313 | hw->bus.func = PCI_FUNC(pdev->devfn); |
| 2314 | ice_set_ctrlq_len(hw); |
| 2315 | |
| 2316 | pf->msg_enable = netif_msg_init(debug, ICE_DFLT_NETIF_M); |
| 2317 | |
| 2318 | #ifndef CONFIG_DYNAMIC_DEBUG |
| 2319 | if (debug < -1) |
| 2320 | hw->debug_mask = debug; |
| 2321 | #endif |
| 2322 | |
| 2323 | err = ice_init_hw(hw); |
| 2324 | if (err) { |
| 2325 | dev_err(dev, "ice_init_hw failed: %d\n", err); |
| 2326 | err = -EIO; |
| 2327 | goto err_exit_unroll; |
| 2328 | } |
| 2329 | |
| 2330 | dev_info(dev, "firmware %d.%d.%05d api %d.%d\n", |
| 2331 | hw->fw_maj_ver, hw->fw_min_ver, hw->fw_build, |
| 2332 | hw->api_maj_ver, hw->api_min_ver); |
| 2333 | |
| 2334 | ice_init_pf(pf); |
| 2335 | |
| 2336 | err = ice_init_pf_dcb(pf); |
| 2337 | if (err) { |
| 2338 | clear_bit(ICE_FLAG_DCB_CAPABLE, pf->flags); |
| 2339 | clear_bit(ICE_FLAG_DCB_ENA, pf->flags); |
| 2340 | |
| 2341 | /* do not fail overall init if DCB init fails */ |
| 2342 | err = 0; |
| 2343 | } |
| 2344 | |
| 2345 | ice_determine_q_usage(pf); |
| 2346 | |
| 2347 | pf->num_alloc_vsi = hw->func_caps.guar_num_vsi; |
| 2348 | if (!pf->num_alloc_vsi) { |
| 2349 | err = -EIO; |
| 2350 | goto err_init_pf_unroll; |
| 2351 | } |
| 2352 | |
| 2353 | pf->vsi = devm_kcalloc(dev, pf->num_alloc_vsi, sizeof(*pf->vsi), |
| 2354 | GFP_KERNEL); |
| 2355 | if (!pf->vsi) { |
| 2356 | err = -ENOMEM; |
| 2357 | goto err_init_pf_unroll; |
| 2358 | } |
| 2359 | |
| 2360 | err = ice_init_interrupt_scheme(pf); |
| 2361 | if (err) { |
| 2362 | dev_err(dev, "ice_init_interrupt_scheme failed: %d\n", err); |
| 2363 | err = -EIO; |
| 2364 | goto err_init_interrupt_unroll; |
| 2365 | } |
| 2366 | |
| 2367 | /* Driver is mostly up */ |
| 2368 | clear_bit(__ICE_DOWN, pf->state); |
| 2369 | |
| 2370 | /* In case of MSIX we are going to setup the misc vector right here |
| 2371 | * to handle admin queue events etc. In case of legacy and MSI |
| 2372 | * the misc functionality and queue processing is combined in |
| 2373 | * the same vector and that gets setup at open. |
| 2374 | */ |
| 2375 | if (test_bit(ICE_FLAG_MSIX_ENA, pf->flags)) { |
| 2376 | err = ice_req_irq_msix_misc(pf); |
| 2377 | if (err) { |
| 2378 | dev_err(dev, "setup of misc vector failed: %d\n", err); |
| 2379 | goto err_init_interrupt_unroll; |
| 2380 | } |
| 2381 | } |
| 2382 | |
| 2383 | /* create switch struct for the switch element created by FW on boot */ |
| 2384 | pf->first_sw = devm_kzalloc(dev, sizeof(*pf->first_sw), GFP_KERNEL); |
| 2385 | if (!pf->first_sw) { |
| 2386 | err = -ENOMEM; |
| 2387 | goto err_msix_misc_unroll; |
| 2388 | } |
| 2389 | |
| 2390 | if (hw->evb_veb) |
| 2391 | pf->first_sw->bridge_mode = BRIDGE_MODE_VEB; |
| 2392 | else |
| 2393 | pf->first_sw->bridge_mode = BRIDGE_MODE_VEPA; |
| 2394 | |
| 2395 | pf->first_sw->pf = pf; |
| 2396 | |
| 2397 | /* record the sw_id available for later use */ |
| 2398 | pf->first_sw->sw_id = hw->port_info->sw_id; |
| 2399 | |
| 2400 | err = ice_setup_pf_sw(pf); |
| 2401 | if (err) { |
| 2402 | dev_err(dev, "probe failed due to setup pf switch:%d\n", err); |
| 2403 | goto err_alloc_sw_unroll; |
| 2404 | } |
| 2405 | |
| 2406 | clear_bit(__ICE_SERVICE_DIS, pf->state); |
| 2407 | |
| 2408 | /* since everything is good, start the service timer */ |
| 2409 | mod_timer(&pf->serv_tmr, round_jiffies(jiffies + pf->serv_tmr_period)); |
| 2410 | |
| 2411 | err = ice_init_link_events(pf->hw.port_info); |
| 2412 | if (err) { |
| 2413 | dev_err(dev, "ice_init_link_events failed: %d\n", err); |
| 2414 | goto err_alloc_sw_unroll; |
| 2415 | } |
| 2416 | |
| 2417 | ice_verify_cacheline_size(pf); |
| 2418 | |
| 2419 | return 0; |
| 2420 | |
| 2421 | err_alloc_sw_unroll: |
| 2422 | set_bit(__ICE_SERVICE_DIS, pf->state); |
| 2423 | set_bit(__ICE_DOWN, pf->state); |
| 2424 | devm_kfree(&pf->pdev->dev, pf->first_sw); |
| 2425 | err_msix_misc_unroll: |
| 2426 | ice_free_irq_msix_misc(pf); |
| 2427 | err_init_interrupt_unroll: |
| 2428 | ice_clear_interrupt_scheme(pf); |
| 2429 | devm_kfree(dev, pf->vsi); |
| 2430 | err_init_pf_unroll: |
| 2431 | ice_deinit_pf(pf); |
| 2432 | ice_deinit_hw(hw); |
| 2433 | err_exit_unroll: |
| 2434 | pci_disable_pcie_error_reporting(pdev); |
| 2435 | return err; |
| 2436 | } |
| 2437 | |
| 2438 | /** |
| 2439 | * ice_remove - Device removal routine |
| 2440 | * @pdev: PCI device information struct |
| 2441 | */ |
| 2442 | static void ice_remove(struct pci_dev *pdev) |
| 2443 | { |
| 2444 | struct ice_pf *pf = pci_get_drvdata(pdev); |
| 2445 | int i; |
| 2446 | |
| 2447 | if (!pf) |
| 2448 | return; |
| 2449 | |
| 2450 | for (i = 0; i < ICE_MAX_RESET_WAIT; i++) { |
| 2451 | if (!ice_is_reset_in_progress(pf->state)) |
| 2452 | break; |
| 2453 | msleep(100); |
| 2454 | } |
| 2455 | |
| 2456 | set_bit(__ICE_DOWN, pf->state); |
| 2457 | ice_service_task_stop(pf); |
| 2458 | |
| 2459 | if (test_bit(ICE_FLAG_SRIOV_ENA, pf->flags)) |
| 2460 | ice_free_vfs(pf); |
| 2461 | ice_vsi_release_all(pf); |
| 2462 | ice_free_irq_msix_misc(pf); |
| 2463 | ice_for_each_vsi(pf, i) { |
| 2464 | if (!pf->vsi[i]) |
| 2465 | continue; |
| 2466 | ice_vsi_free_q_vectors(pf->vsi[i]); |
| 2467 | } |
| 2468 | ice_clear_interrupt_scheme(pf); |
| 2469 | ice_deinit_pf(pf); |
| 2470 | ice_deinit_hw(&pf->hw); |
| 2471 | pci_disable_pcie_error_reporting(pdev); |
| 2472 | } |
| 2473 | |
| 2474 | /** |
| 2475 | * ice_pci_err_detected - warning that PCI error has been detected |
| 2476 | * @pdev: PCI device information struct |
| 2477 | * @err: the type of PCI error |
| 2478 | * |
| 2479 | * Called to warn that something happened on the PCI bus and the error handling |
| 2480 | * is in progress. Allows the driver to gracefully prepare/handle PCI errors. |
| 2481 | */ |
| 2482 | static pci_ers_result_t |
| 2483 | ice_pci_err_detected(struct pci_dev *pdev, enum pci_channel_state err) |
| 2484 | { |
| 2485 | struct ice_pf *pf = pci_get_drvdata(pdev); |
| 2486 | |
| 2487 | if (!pf) { |
| 2488 | dev_err(&pdev->dev, "%s: unrecoverable device error %d\n", |
| 2489 | __func__, err); |
| 2490 | return PCI_ERS_RESULT_DISCONNECT; |
| 2491 | } |
| 2492 | |
| 2493 | if (!test_bit(__ICE_SUSPENDED, pf->state)) { |
| 2494 | ice_service_task_stop(pf); |
| 2495 | |
| 2496 | if (!test_bit(__ICE_PREPARED_FOR_RESET, pf->state)) { |
| 2497 | set_bit(__ICE_PFR_REQ, pf->state); |
| 2498 | ice_prepare_for_reset(pf); |
| 2499 | } |
| 2500 | } |
| 2501 | |
| 2502 | return PCI_ERS_RESULT_NEED_RESET; |
| 2503 | } |
| 2504 | |
| 2505 | /** |
| 2506 | * ice_pci_err_slot_reset - a PCI slot reset has just happened |
| 2507 | * @pdev: PCI device information struct |
| 2508 | * |
| 2509 | * Called to determine if the driver can recover from the PCI slot reset by |
| 2510 | * using a register read to determine if the device is recoverable. |
| 2511 | */ |
| 2512 | static pci_ers_result_t ice_pci_err_slot_reset(struct pci_dev *pdev) |
| 2513 | { |
| 2514 | struct ice_pf *pf = pci_get_drvdata(pdev); |
| 2515 | pci_ers_result_t result; |
| 2516 | int err; |
| 2517 | u32 reg; |
| 2518 | |
| 2519 | err = pci_enable_device_mem(pdev); |
| 2520 | if (err) { |
| 2521 | dev_err(&pdev->dev, |
| 2522 | "Cannot re-enable PCI device after reset, error %d\n", |
| 2523 | err); |
| 2524 | result = PCI_ERS_RESULT_DISCONNECT; |
| 2525 | } else { |
| 2526 | pci_set_master(pdev); |
| 2527 | pci_restore_state(pdev); |
| 2528 | pci_save_state(pdev); |
| 2529 | pci_wake_from_d3(pdev, false); |
| 2530 | |
| 2531 | /* Check for life */ |
| 2532 | reg = rd32(&pf->hw, GLGEN_RTRIG); |
| 2533 | if (!reg) |
| 2534 | result = PCI_ERS_RESULT_RECOVERED; |
| 2535 | else |
| 2536 | result = PCI_ERS_RESULT_DISCONNECT; |
| 2537 | } |
| 2538 | |
| 2539 | err = pci_cleanup_aer_uncorrect_error_status(pdev); |
| 2540 | if (err) |
| 2541 | dev_dbg(&pdev->dev, |
| 2542 | "pci_cleanup_aer_uncorrect_error_status failed, error %d\n", |
| 2543 | err); |
| 2544 | /* non-fatal, continue */ |
| 2545 | |
| 2546 | return result; |
| 2547 | } |
| 2548 | |
| 2549 | /** |
| 2550 | * ice_pci_err_resume - restart operations after PCI error recovery |
| 2551 | * @pdev: PCI device information struct |
| 2552 | * |
| 2553 | * Called to allow the driver to bring things back up after PCI error and/or |
| 2554 | * reset recovery have finished |
| 2555 | */ |
| 2556 | static void ice_pci_err_resume(struct pci_dev *pdev) |
| 2557 | { |
| 2558 | struct ice_pf *pf = pci_get_drvdata(pdev); |
| 2559 | |
| 2560 | if (!pf) { |
| 2561 | dev_err(&pdev->dev, |
| 2562 | "%s failed, device is unrecoverable\n", __func__); |
| 2563 | return; |
| 2564 | } |
| 2565 | |
| 2566 | if (test_bit(__ICE_SUSPENDED, pf->state)) { |
| 2567 | dev_dbg(&pdev->dev, "%s failed to resume normal operations!\n", |
| 2568 | __func__); |
| 2569 | return; |
| 2570 | } |
| 2571 | |
| 2572 | ice_do_reset(pf, ICE_RESET_PFR); |
| 2573 | ice_service_task_restart(pf); |
| 2574 | mod_timer(&pf->serv_tmr, round_jiffies(jiffies + pf->serv_tmr_period)); |
| 2575 | } |
| 2576 | |
| 2577 | /** |
| 2578 | * ice_pci_err_reset_prepare - prepare device driver for PCI reset |
| 2579 | * @pdev: PCI device information struct |
| 2580 | */ |
| 2581 | static void ice_pci_err_reset_prepare(struct pci_dev *pdev) |
| 2582 | { |
| 2583 | struct ice_pf *pf = pci_get_drvdata(pdev); |
| 2584 | |
| 2585 | if (!test_bit(__ICE_SUSPENDED, pf->state)) { |
| 2586 | ice_service_task_stop(pf); |
| 2587 | |
| 2588 | if (!test_bit(__ICE_PREPARED_FOR_RESET, pf->state)) { |
| 2589 | set_bit(__ICE_PFR_REQ, pf->state); |
| 2590 | ice_prepare_for_reset(pf); |
| 2591 | } |
| 2592 | } |
| 2593 | } |
| 2594 | |
| 2595 | /** |
| 2596 | * ice_pci_err_reset_done - PCI reset done, device driver reset can begin |
| 2597 | * @pdev: PCI device information struct |
| 2598 | */ |
| 2599 | static void ice_pci_err_reset_done(struct pci_dev *pdev) |
| 2600 | { |
| 2601 | ice_pci_err_resume(pdev); |
| 2602 | } |
| 2603 | |
| 2604 | /* ice_pci_tbl - PCI Device ID Table |
| 2605 | * |
| 2606 | * Wildcard entries (PCI_ANY_ID) should come last |
| 2607 | * Last entry must be all 0s |
| 2608 | * |
| 2609 | * { Vendor ID, Device ID, SubVendor ID, SubDevice ID, |
| 2610 | * Class, Class Mask, private data (not used) } |
| 2611 | */ |
| 2612 | static const struct pci_device_id ice_pci_tbl[] = { |
| 2613 | { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810C_BACKPLANE), 0 }, |
| 2614 | { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810C_QSFP), 0 }, |
| 2615 | { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810C_SFP), 0 }, |
| 2616 | /* required last entry */ |
| 2617 | { 0, } |
| 2618 | }; |
| 2619 | MODULE_DEVICE_TABLE(pci, ice_pci_tbl); |
| 2620 | |
| 2621 | static const struct pci_error_handlers ice_pci_err_handler = { |
| 2622 | .error_detected = ice_pci_err_detected, |
| 2623 | .slot_reset = ice_pci_err_slot_reset, |
| 2624 | .reset_prepare = ice_pci_err_reset_prepare, |
| 2625 | .reset_done = ice_pci_err_reset_done, |
| 2626 | .resume = ice_pci_err_resume |
| 2627 | }; |
| 2628 | |
| 2629 | static struct pci_driver ice_driver = { |
| 2630 | .name = KBUILD_MODNAME, |
| 2631 | .id_table = ice_pci_tbl, |
| 2632 | .probe = ice_probe, |
| 2633 | .remove = ice_remove, |
| 2634 | .sriov_configure = ice_sriov_configure, |
| 2635 | .err_handler = &ice_pci_err_handler |
| 2636 | }; |
| 2637 | |
| 2638 | /** |
| 2639 | * ice_module_init - Driver registration routine |
| 2640 | * |
| 2641 | * ice_module_init is the first routine called when the driver is |
| 2642 | * loaded. All it does is register with the PCI subsystem. |
| 2643 | */ |
| 2644 | static int __init ice_module_init(void) |
| 2645 | { |
| 2646 | int status; |
| 2647 | |
| 2648 | pr_info("%s - version %s\n", ice_driver_string, ice_drv_ver); |
| 2649 | pr_info("%s\n", ice_copyright); |
| 2650 | |
| 2651 | ice_wq = alloc_workqueue("%s", WQ_MEM_RECLAIM, 0, KBUILD_MODNAME); |
| 2652 | if (!ice_wq) { |
| 2653 | pr_err("Failed to create workqueue\n"); |
| 2654 | return -ENOMEM; |
| 2655 | } |
| 2656 | |
| 2657 | status = pci_register_driver(&ice_driver); |
| 2658 | if (status) { |
| 2659 | pr_err("failed to register pci driver, err %d\n", status); |
| 2660 | destroy_workqueue(ice_wq); |
| 2661 | } |
| 2662 | |
| 2663 | return status; |
| 2664 | } |
| 2665 | module_init(ice_module_init); |
| 2666 | |
| 2667 | /** |
| 2668 | * ice_module_exit - Driver exit cleanup routine |
| 2669 | * |
| 2670 | * ice_module_exit is called just before the driver is removed |
| 2671 | * from memory. |
| 2672 | */ |
| 2673 | static void __exit ice_module_exit(void) |
| 2674 | { |
| 2675 | pci_unregister_driver(&ice_driver); |
| 2676 | destroy_workqueue(ice_wq); |
| 2677 | pr_info("module unloaded\n"); |
| 2678 | } |
| 2679 | module_exit(ice_module_exit); |
| 2680 | |
| 2681 | /** |
| 2682 | * ice_set_mac_address - NDO callback to set MAC address |
| 2683 | * @netdev: network interface device structure |
| 2684 | * @pi: pointer to an address structure |
| 2685 | * |
| 2686 | * Returns 0 on success, negative on failure |
| 2687 | */ |
| 2688 | static int ice_set_mac_address(struct net_device *netdev, void *pi) |
| 2689 | { |
| 2690 | struct ice_netdev_priv *np = netdev_priv(netdev); |
| 2691 | struct ice_vsi *vsi = np->vsi; |
| 2692 | struct ice_pf *pf = vsi->back; |
| 2693 | struct ice_hw *hw = &pf->hw; |
| 2694 | struct sockaddr *addr = pi; |
| 2695 | enum ice_status status; |
| 2696 | LIST_HEAD(a_mac_list); |
| 2697 | LIST_HEAD(r_mac_list); |
| 2698 | u8 flags = 0; |
| 2699 | int err; |
| 2700 | u8 *mac; |
| 2701 | |
| 2702 | mac = (u8 *)addr->sa_data; |
| 2703 | |
| 2704 | if (!is_valid_ether_addr(mac)) |
| 2705 | return -EADDRNOTAVAIL; |
| 2706 | |
| 2707 | if (ether_addr_equal(netdev->dev_addr, mac)) { |
| 2708 | netdev_warn(netdev, "already using mac %pM\n", mac); |
| 2709 | return 0; |
| 2710 | } |
| 2711 | |
| 2712 | if (test_bit(__ICE_DOWN, pf->state) || |
| 2713 | ice_is_reset_in_progress(pf->state)) { |
| 2714 | netdev_err(netdev, "can't set mac %pM. device not ready\n", |
| 2715 | mac); |
| 2716 | return -EBUSY; |
| 2717 | } |
| 2718 | |
| 2719 | /* When we change the MAC address we also have to change the MAC address |
| 2720 | * based filter rules that were created previously for the old MAC |
| 2721 | * address. So first, we remove the old filter rule using ice_remove_mac |
| 2722 | * and then create a new filter rule using ice_add_mac. Note that for |
| 2723 | * both these operations, we first need to form a "list" of MAC |
| 2724 | * addresses (even though in this case, we have only 1 MAC address to be |
| 2725 | * added/removed) and this done using ice_add_mac_to_list. Depending on |
| 2726 | * the ensuing operation this "list" of MAC addresses is either to be |
| 2727 | * added or removed from the filter. |
| 2728 | */ |
| 2729 | err = ice_add_mac_to_list(vsi, &r_mac_list, netdev->dev_addr); |
| 2730 | if (err) { |
| 2731 | err = -EADDRNOTAVAIL; |
| 2732 | goto free_lists; |
| 2733 | } |
| 2734 | |
| 2735 | status = ice_remove_mac(hw, &r_mac_list); |
| 2736 | if (status) { |
| 2737 | err = -EADDRNOTAVAIL; |
| 2738 | goto free_lists; |
| 2739 | } |
| 2740 | |
| 2741 | err = ice_add_mac_to_list(vsi, &a_mac_list, mac); |
| 2742 | if (err) { |
| 2743 | err = -EADDRNOTAVAIL; |
| 2744 | goto free_lists; |
| 2745 | } |
| 2746 | |
| 2747 | status = ice_add_mac(hw, &a_mac_list); |
| 2748 | if (status) { |
| 2749 | err = -EADDRNOTAVAIL; |
| 2750 | goto free_lists; |
| 2751 | } |
| 2752 | |
| 2753 | free_lists: |
| 2754 | /* free list entries */ |
| 2755 | ice_free_fltr_list(&pf->pdev->dev, &r_mac_list); |
| 2756 | ice_free_fltr_list(&pf->pdev->dev, &a_mac_list); |
| 2757 | |
| 2758 | if (err) { |
| 2759 | netdev_err(netdev, "can't set mac %pM. filter update failed\n", |
| 2760 | mac); |
| 2761 | return err; |
| 2762 | } |
| 2763 | |
| 2764 | /* change the netdev's MAC address */ |
| 2765 | memcpy(netdev->dev_addr, mac, netdev->addr_len); |
| 2766 | netdev_dbg(vsi->netdev, "updated mac address to %pM\n", |
| 2767 | netdev->dev_addr); |
| 2768 | |
| 2769 | /* write new MAC address to the firmware */ |
| 2770 | flags = ICE_AQC_MAN_MAC_UPDATE_LAA_WOL; |
| 2771 | status = ice_aq_manage_mac_write(hw, mac, flags, NULL); |
| 2772 | if (status) { |
| 2773 | netdev_err(netdev, "can't set mac %pM. write to firmware failed.\n", |
| 2774 | mac); |
| 2775 | } |
| 2776 | return 0; |
| 2777 | } |
| 2778 | |
| 2779 | /** |
| 2780 | * ice_set_rx_mode - NDO callback to set the netdev filters |
| 2781 | * @netdev: network interface device structure |
| 2782 | */ |
| 2783 | static void ice_set_rx_mode(struct net_device *netdev) |
| 2784 | { |
| 2785 | struct ice_netdev_priv *np = netdev_priv(netdev); |
| 2786 | struct ice_vsi *vsi = np->vsi; |
| 2787 | |
| 2788 | if (!vsi) |
| 2789 | return; |
| 2790 | |
| 2791 | /* Set the flags to synchronize filters |
| 2792 | * ndo_set_rx_mode may be triggered even without a change in netdev |
| 2793 | * flags |
| 2794 | */ |
| 2795 | set_bit(ICE_VSI_FLAG_UMAC_FLTR_CHANGED, vsi->flags); |
| 2796 | set_bit(ICE_VSI_FLAG_MMAC_FLTR_CHANGED, vsi->flags); |
| 2797 | set_bit(ICE_FLAG_FLTR_SYNC, vsi->back->flags); |
| 2798 | |
| 2799 | /* schedule our worker thread which will take care of |
| 2800 | * applying the new filter changes |
| 2801 | */ |
| 2802 | ice_service_task_schedule(vsi->back); |
| 2803 | } |
| 2804 | |
| 2805 | /** |
| 2806 | * ice_fdb_add - add an entry to the hardware database |
| 2807 | * @ndm: the input from the stack |
| 2808 | * @tb: pointer to array of nladdr (unused) |
| 2809 | * @dev: the net device pointer |
| 2810 | * @addr: the MAC address entry being added |
| 2811 | * @vid: VLAN ID |
| 2812 | * @flags: instructions from stack about fdb operation |
| 2813 | * @extack: netlink extended ack |
| 2814 | */ |
| 2815 | static int |
| 2816 | ice_fdb_add(struct ndmsg *ndm, struct nlattr __always_unused *tb[], |
| 2817 | struct net_device *dev, const unsigned char *addr, u16 vid, |
| 2818 | u16 flags, struct netlink_ext_ack __always_unused *extack) |
| 2819 | { |
| 2820 | int err; |
| 2821 | |
| 2822 | if (vid) { |
| 2823 | netdev_err(dev, "VLANs aren't supported yet for dev_uc|mc_add()\n"); |
| 2824 | return -EINVAL; |
| 2825 | } |
| 2826 | if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) { |
| 2827 | netdev_err(dev, "FDB only supports static addresses\n"); |
| 2828 | return -EINVAL; |
| 2829 | } |
| 2830 | |
| 2831 | if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr)) |
| 2832 | err = dev_uc_add_excl(dev, addr); |
| 2833 | else if (is_multicast_ether_addr(addr)) |
| 2834 | err = dev_mc_add_excl(dev, addr); |
| 2835 | else |
| 2836 | err = -EINVAL; |
| 2837 | |
| 2838 | /* Only return duplicate errors if NLM_F_EXCL is set */ |
| 2839 | if (err == -EEXIST && !(flags & NLM_F_EXCL)) |
| 2840 | err = 0; |
| 2841 | |
| 2842 | return err; |
| 2843 | } |
| 2844 | |
| 2845 | /** |
| 2846 | * ice_fdb_del - delete an entry from the hardware database |
| 2847 | * @ndm: the input from the stack |
| 2848 | * @tb: pointer to array of nladdr (unused) |
| 2849 | * @dev: the net device pointer |
| 2850 | * @addr: the MAC address entry being added |
| 2851 | * @vid: VLAN ID |
| 2852 | */ |
| 2853 | static int |
| 2854 | ice_fdb_del(struct ndmsg *ndm, __always_unused struct nlattr *tb[], |
| 2855 | struct net_device *dev, const unsigned char *addr, |
| 2856 | __always_unused u16 vid) |
| 2857 | { |
| 2858 | int err; |
| 2859 | |
| 2860 | if (ndm->ndm_state & NUD_PERMANENT) { |
| 2861 | netdev_err(dev, "FDB only supports static addresses\n"); |
| 2862 | return -EINVAL; |
| 2863 | } |
| 2864 | |
| 2865 | if (is_unicast_ether_addr(addr)) |
| 2866 | err = dev_uc_del(dev, addr); |
| 2867 | else if (is_multicast_ether_addr(addr)) |
| 2868 | err = dev_mc_del(dev, addr); |
| 2869 | else |
| 2870 | err = -EINVAL; |
| 2871 | |
| 2872 | return err; |
| 2873 | } |
| 2874 | |
| 2875 | /** |
| 2876 | * ice_set_features - set the netdev feature flags |
| 2877 | * @netdev: ptr to the netdev being adjusted |
| 2878 | * @features: the feature set that the stack is suggesting |
| 2879 | */ |
| 2880 | static int |
| 2881 | ice_set_features(struct net_device *netdev, netdev_features_t features) |
| 2882 | { |
| 2883 | struct ice_netdev_priv *np = netdev_priv(netdev); |
| 2884 | struct ice_vsi *vsi = np->vsi; |
| 2885 | int ret = 0; |
| 2886 | |
| 2887 | if (features & NETIF_F_RXHASH && !(netdev->features & NETIF_F_RXHASH)) |
| 2888 | ret = ice_vsi_manage_rss_lut(vsi, true); |
| 2889 | else if (!(features & NETIF_F_RXHASH) && |
| 2890 | netdev->features & NETIF_F_RXHASH) |
| 2891 | ret = ice_vsi_manage_rss_lut(vsi, false); |
| 2892 | |
| 2893 | if ((features & NETIF_F_HW_VLAN_CTAG_RX) && |
| 2894 | !(netdev->features & NETIF_F_HW_VLAN_CTAG_RX)) |
| 2895 | ret = ice_vsi_manage_vlan_stripping(vsi, true); |
| 2896 | else if (!(features & NETIF_F_HW_VLAN_CTAG_RX) && |
| 2897 | (netdev->features & NETIF_F_HW_VLAN_CTAG_RX)) |
| 2898 | ret = ice_vsi_manage_vlan_stripping(vsi, false); |
| 2899 | else if ((features & NETIF_F_HW_VLAN_CTAG_TX) && |
| 2900 | !(netdev->features & NETIF_F_HW_VLAN_CTAG_TX)) |
| 2901 | ret = ice_vsi_manage_vlan_insertion(vsi); |
| 2902 | else if (!(features & NETIF_F_HW_VLAN_CTAG_TX) && |
| 2903 | (netdev->features & NETIF_F_HW_VLAN_CTAG_TX)) |
| 2904 | ret = ice_vsi_manage_vlan_insertion(vsi); |
| 2905 | |
| 2906 | return ret; |
| 2907 | } |
| 2908 | |
| 2909 | /** |
| 2910 | * ice_vsi_vlan_setup - Setup VLAN offload properties on a VSI |
| 2911 | * @vsi: VSI to setup VLAN properties for |
| 2912 | */ |
| 2913 | static int ice_vsi_vlan_setup(struct ice_vsi *vsi) |
| 2914 | { |
| 2915 | int ret = 0; |
| 2916 | |
| 2917 | if (vsi->netdev->features & NETIF_F_HW_VLAN_CTAG_RX) |
| 2918 | ret = ice_vsi_manage_vlan_stripping(vsi, true); |
| 2919 | if (vsi->netdev->features & NETIF_F_HW_VLAN_CTAG_TX) |
| 2920 | ret = ice_vsi_manage_vlan_insertion(vsi); |
| 2921 | |
| 2922 | return ret; |
| 2923 | } |
| 2924 | |
| 2925 | /** |
| 2926 | * ice_vsi_cfg - Setup the VSI |
| 2927 | * @vsi: the VSI being configured |
| 2928 | * |
| 2929 | * Return 0 on success and negative value on error |
| 2930 | */ |
| 2931 | static int ice_vsi_cfg(struct ice_vsi *vsi) |
| 2932 | { |
| 2933 | int err; |
| 2934 | |
| 2935 | if (vsi->netdev) { |
| 2936 | ice_set_rx_mode(vsi->netdev); |
| 2937 | |
| 2938 | err = ice_vsi_vlan_setup(vsi); |
| 2939 | |
| 2940 | if (err) |
| 2941 | return err; |
| 2942 | } |
| 2943 | ice_vsi_cfg_dcb_rings(vsi); |
| 2944 | |
| 2945 | err = ice_vsi_cfg_lan_txqs(vsi); |
| 2946 | if (!err) |
| 2947 | err = ice_vsi_cfg_rxqs(vsi); |
| 2948 | |
| 2949 | return err; |
| 2950 | } |
| 2951 | |
| 2952 | /** |
| 2953 | * ice_napi_enable_all - Enable NAPI for all q_vectors in the VSI |
| 2954 | * @vsi: the VSI being configured |
| 2955 | */ |
| 2956 | static void ice_napi_enable_all(struct ice_vsi *vsi) |
| 2957 | { |
| 2958 | int q_idx; |
| 2959 | |
| 2960 | if (!vsi->netdev) |
| 2961 | return; |
| 2962 | |
| 2963 | ice_for_each_q_vector(vsi, q_idx) { |
| 2964 | struct ice_q_vector *q_vector = vsi->q_vectors[q_idx]; |
| 2965 | |
| 2966 | if (q_vector->rx.ring || q_vector->tx.ring) |
| 2967 | napi_enable(&q_vector->napi); |
| 2968 | } |
| 2969 | } |
| 2970 | |
| 2971 | /** |
| 2972 | * ice_up_complete - Finish the last steps of bringing up a connection |
| 2973 | * @vsi: The VSI being configured |
| 2974 | * |
| 2975 | * Return 0 on success and negative value on error |
| 2976 | */ |
| 2977 | static int ice_up_complete(struct ice_vsi *vsi) |
| 2978 | { |
| 2979 | struct ice_pf *pf = vsi->back; |
| 2980 | int err; |
| 2981 | |
| 2982 | if (test_bit(ICE_FLAG_MSIX_ENA, pf->flags)) |
| 2983 | ice_vsi_cfg_msix(vsi); |
| 2984 | else |
| 2985 | return -ENOTSUPP; |
| 2986 | |
| 2987 | /* Enable only Rx rings, Tx rings were enabled by the FW when the |
| 2988 | * Tx queue group list was configured and the context bits were |
| 2989 | * programmed using ice_vsi_cfg_txqs |
| 2990 | */ |
| 2991 | err = ice_vsi_start_rx_rings(vsi); |
| 2992 | if (err) |
| 2993 | return err; |
| 2994 | |
| 2995 | clear_bit(__ICE_DOWN, vsi->state); |
| 2996 | ice_napi_enable_all(vsi); |
| 2997 | ice_vsi_ena_irq(vsi); |
| 2998 | |
| 2999 | if (vsi->port_info && |
| 3000 | (vsi->port_info->phy.link_info.link_info & ICE_AQ_LINK_UP) && |
| 3001 | vsi->netdev) { |
| 3002 | ice_print_link_msg(vsi, true); |
| 3003 | netif_tx_start_all_queues(vsi->netdev); |
| 3004 | netif_carrier_on(vsi->netdev); |
| 3005 | } |
| 3006 | |
| 3007 | ice_service_task_schedule(pf); |
| 3008 | |
| 3009 | return 0; |
| 3010 | } |
| 3011 | |
| 3012 | /** |
| 3013 | * ice_up - Bring the connection back up after being down |
| 3014 | * @vsi: VSI being configured |
| 3015 | */ |
| 3016 | int ice_up(struct ice_vsi *vsi) |
| 3017 | { |
| 3018 | int err; |
| 3019 | |
| 3020 | err = ice_vsi_cfg(vsi); |
| 3021 | if (!err) |
| 3022 | err = ice_up_complete(vsi); |
| 3023 | |
| 3024 | return err; |
| 3025 | } |
| 3026 | |
| 3027 | /** |
| 3028 | * ice_fetch_u64_stats_per_ring - get packets and bytes stats per ring |
| 3029 | * @ring: Tx or Rx ring to read stats from |
| 3030 | * @pkts: packets stats counter |
| 3031 | * @bytes: bytes stats counter |
| 3032 | * |
| 3033 | * This function fetches stats from the ring considering the atomic operations |
| 3034 | * that needs to be performed to read u64 values in 32 bit machine. |
| 3035 | */ |
| 3036 | static void |
| 3037 | ice_fetch_u64_stats_per_ring(struct ice_ring *ring, u64 *pkts, u64 *bytes) |
| 3038 | { |
| 3039 | unsigned int start; |
| 3040 | *pkts = 0; |
| 3041 | *bytes = 0; |
| 3042 | |
| 3043 | if (!ring) |
| 3044 | return; |
| 3045 | do { |
| 3046 | start = u64_stats_fetch_begin_irq(&ring->syncp); |
| 3047 | *pkts = ring->stats.pkts; |
| 3048 | *bytes = ring->stats.bytes; |
| 3049 | } while (u64_stats_fetch_retry_irq(&ring->syncp, start)); |
| 3050 | } |
| 3051 | |
| 3052 | /** |
| 3053 | * ice_update_vsi_ring_stats - Update VSI stats counters |
| 3054 | * @vsi: the VSI to be updated |
| 3055 | */ |
| 3056 | static void ice_update_vsi_ring_stats(struct ice_vsi *vsi) |
| 3057 | { |
| 3058 | struct rtnl_link_stats64 *vsi_stats = &vsi->net_stats; |
| 3059 | struct ice_ring *ring; |
| 3060 | u64 pkts, bytes; |
| 3061 | int i; |
| 3062 | |
| 3063 | /* reset netdev stats */ |
| 3064 | vsi_stats->tx_packets = 0; |
| 3065 | vsi_stats->tx_bytes = 0; |
| 3066 | vsi_stats->rx_packets = 0; |
| 3067 | vsi_stats->rx_bytes = 0; |
| 3068 | |
| 3069 | /* reset non-netdev (extended) stats */ |
| 3070 | vsi->tx_restart = 0; |
| 3071 | vsi->tx_busy = 0; |
| 3072 | vsi->tx_linearize = 0; |
| 3073 | vsi->rx_buf_failed = 0; |
| 3074 | vsi->rx_page_failed = 0; |
| 3075 | |
| 3076 | rcu_read_lock(); |
| 3077 | |
| 3078 | /* update Tx rings counters */ |
| 3079 | ice_for_each_txq(vsi, i) { |
| 3080 | ring = READ_ONCE(vsi->tx_rings[i]); |
| 3081 | ice_fetch_u64_stats_per_ring(ring, &pkts, &bytes); |
| 3082 | vsi_stats->tx_packets += pkts; |
| 3083 | vsi_stats->tx_bytes += bytes; |
| 3084 | vsi->tx_restart += ring->tx_stats.restart_q; |
| 3085 | vsi->tx_busy += ring->tx_stats.tx_busy; |
| 3086 | vsi->tx_linearize += ring->tx_stats.tx_linearize; |
| 3087 | } |
| 3088 | |
| 3089 | /* update Rx rings counters */ |
| 3090 | ice_for_each_rxq(vsi, i) { |
| 3091 | ring = READ_ONCE(vsi->rx_rings[i]); |
| 3092 | ice_fetch_u64_stats_per_ring(ring, &pkts, &bytes); |
| 3093 | vsi_stats->rx_packets += pkts; |
| 3094 | vsi_stats->rx_bytes += bytes; |
| 3095 | vsi->rx_buf_failed += ring->rx_stats.alloc_buf_failed; |
| 3096 | vsi->rx_page_failed += ring->rx_stats.alloc_page_failed; |
| 3097 | } |
| 3098 | |
| 3099 | rcu_read_unlock(); |
| 3100 | } |
| 3101 | |
| 3102 | /** |
| 3103 | * ice_update_vsi_stats - Update VSI stats counters |
| 3104 | * @vsi: the VSI to be updated |
| 3105 | */ |
| 3106 | static void ice_update_vsi_stats(struct ice_vsi *vsi) |
| 3107 | { |
| 3108 | struct rtnl_link_stats64 *cur_ns = &vsi->net_stats; |
| 3109 | struct ice_eth_stats *cur_es = &vsi->eth_stats; |
| 3110 | struct ice_pf *pf = vsi->back; |
| 3111 | |
| 3112 | if (test_bit(__ICE_DOWN, vsi->state) || |
| 3113 | test_bit(__ICE_CFG_BUSY, pf->state)) |
| 3114 | return; |
| 3115 | |
| 3116 | /* get stats as recorded by Tx/Rx rings */ |
| 3117 | ice_update_vsi_ring_stats(vsi); |
| 3118 | |
| 3119 | /* get VSI stats as recorded by the hardware */ |
| 3120 | ice_update_eth_stats(vsi); |
| 3121 | |
| 3122 | cur_ns->tx_errors = cur_es->tx_errors; |
| 3123 | cur_ns->rx_dropped = cur_es->rx_discards; |
| 3124 | cur_ns->tx_dropped = cur_es->tx_discards; |
| 3125 | cur_ns->multicast = cur_es->rx_multicast; |
| 3126 | |
| 3127 | /* update some more netdev stats if this is main VSI */ |
| 3128 | if (vsi->type == ICE_VSI_PF) { |
| 3129 | cur_ns->rx_crc_errors = pf->stats.crc_errors; |
| 3130 | cur_ns->rx_errors = pf->stats.crc_errors + |
| 3131 | pf->stats.illegal_bytes; |
| 3132 | cur_ns->rx_length_errors = pf->stats.rx_len_errors; |
| 3133 | } |
| 3134 | } |
| 3135 | |
| 3136 | /** |
| 3137 | * ice_update_pf_stats - Update PF port stats counters |
| 3138 | * @pf: PF whose stats needs to be updated |
| 3139 | */ |
| 3140 | static void ice_update_pf_stats(struct ice_pf *pf) |
| 3141 | { |
| 3142 | struct ice_hw_port_stats *prev_ps, *cur_ps; |
| 3143 | struct ice_hw *hw = &pf->hw; |
| 3144 | u8 pf_id; |
| 3145 | |
| 3146 | prev_ps = &pf->stats_prev; |
| 3147 | cur_ps = &pf->stats; |
| 3148 | pf_id = hw->pf_id; |
| 3149 | |
| 3150 | ice_stat_update40(hw, GLPRT_GORCH(pf_id), GLPRT_GORCL(pf_id), |
| 3151 | pf->stat_prev_loaded, &prev_ps->eth.rx_bytes, |
| 3152 | &cur_ps->eth.rx_bytes); |
| 3153 | |
| 3154 | ice_stat_update40(hw, GLPRT_UPRCH(pf_id), GLPRT_UPRCL(pf_id), |
| 3155 | pf->stat_prev_loaded, &prev_ps->eth.rx_unicast, |
| 3156 | &cur_ps->eth.rx_unicast); |
| 3157 | |
| 3158 | ice_stat_update40(hw, GLPRT_MPRCH(pf_id), GLPRT_MPRCL(pf_id), |
| 3159 | pf->stat_prev_loaded, &prev_ps->eth.rx_multicast, |
| 3160 | &cur_ps->eth.rx_multicast); |
| 3161 | |
| 3162 | ice_stat_update40(hw, GLPRT_BPRCH(pf_id), GLPRT_BPRCL(pf_id), |
| 3163 | pf->stat_prev_loaded, &prev_ps->eth.rx_broadcast, |
| 3164 | &cur_ps->eth.rx_broadcast); |
| 3165 | |
| 3166 | ice_stat_update40(hw, GLPRT_GOTCH(pf_id), GLPRT_GOTCL(pf_id), |
| 3167 | pf->stat_prev_loaded, &prev_ps->eth.tx_bytes, |
| 3168 | &cur_ps->eth.tx_bytes); |
| 3169 | |
| 3170 | ice_stat_update40(hw, GLPRT_UPTCH(pf_id), GLPRT_UPTCL(pf_id), |
| 3171 | pf->stat_prev_loaded, &prev_ps->eth.tx_unicast, |
| 3172 | &cur_ps->eth.tx_unicast); |
| 3173 | |
| 3174 | ice_stat_update40(hw, GLPRT_MPTCH(pf_id), GLPRT_MPTCL(pf_id), |
| 3175 | pf->stat_prev_loaded, &prev_ps->eth.tx_multicast, |
| 3176 | &cur_ps->eth.tx_multicast); |
| 3177 | |
| 3178 | ice_stat_update40(hw, GLPRT_BPTCH(pf_id), GLPRT_BPTCL(pf_id), |
| 3179 | pf->stat_prev_loaded, &prev_ps->eth.tx_broadcast, |
| 3180 | &cur_ps->eth.tx_broadcast); |
| 3181 | |
| 3182 | ice_stat_update32(hw, GLPRT_TDOLD(pf_id), pf->stat_prev_loaded, |
| 3183 | &prev_ps->tx_dropped_link_down, |
| 3184 | &cur_ps->tx_dropped_link_down); |
| 3185 | |
| 3186 | ice_stat_update40(hw, GLPRT_PRC64H(pf_id), GLPRT_PRC64L(pf_id), |
| 3187 | pf->stat_prev_loaded, &prev_ps->rx_size_64, |
| 3188 | &cur_ps->rx_size_64); |
| 3189 | |
| 3190 | ice_stat_update40(hw, GLPRT_PRC127H(pf_id), GLPRT_PRC127L(pf_id), |
| 3191 | pf->stat_prev_loaded, &prev_ps->rx_size_127, |
| 3192 | &cur_ps->rx_size_127); |
| 3193 | |
| 3194 | ice_stat_update40(hw, GLPRT_PRC255H(pf_id), GLPRT_PRC255L(pf_id), |
| 3195 | pf->stat_prev_loaded, &prev_ps->rx_size_255, |
| 3196 | &cur_ps->rx_size_255); |
| 3197 | |
| 3198 | ice_stat_update40(hw, GLPRT_PRC511H(pf_id), GLPRT_PRC511L(pf_id), |
| 3199 | pf->stat_prev_loaded, &prev_ps->rx_size_511, |
| 3200 | &cur_ps->rx_size_511); |
| 3201 | |
| 3202 | ice_stat_update40(hw, GLPRT_PRC1023H(pf_id), |
| 3203 | GLPRT_PRC1023L(pf_id), pf->stat_prev_loaded, |
| 3204 | &prev_ps->rx_size_1023, &cur_ps->rx_size_1023); |
| 3205 | |
| 3206 | ice_stat_update40(hw, GLPRT_PRC1522H(pf_id), |
| 3207 | GLPRT_PRC1522L(pf_id), pf->stat_prev_loaded, |
| 3208 | &prev_ps->rx_size_1522, &cur_ps->rx_size_1522); |
| 3209 | |
| 3210 | ice_stat_update40(hw, GLPRT_PRC9522H(pf_id), |
| 3211 | GLPRT_PRC9522L(pf_id), pf->stat_prev_loaded, |
| 3212 | &prev_ps->rx_size_big, &cur_ps->rx_size_big); |
| 3213 | |
| 3214 | ice_stat_update40(hw, GLPRT_PTC64H(pf_id), GLPRT_PTC64L(pf_id), |
| 3215 | pf->stat_prev_loaded, &prev_ps->tx_size_64, |
| 3216 | &cur_ps->tx_size_64); |
| 3217 | |
| 3218 | ice_stat_update40(hw, GLPRT_PTC127H(pf_id), GLPRT_PTC127L(pf_id), |
| 3219 | pf->stat_prev_loaded, &prev_ps->tx_size_127, |
| 3220 | &cur_ps->tx_size_127); |
| 3221 | |
| 3222 | ice_stat_update40(hw, GLPRT_PTC255H(pf_id), GLPRT_PTC255L(pf_id), |
| 3223 | pf->stat_prev_loaded, &prev_ps->tx_size_255, |
| 3224 | &cur_ps->tx_size_255); |
| 3225 | |
| 3226 | ice_stat_update40(hw, GLPRT_PTC511H(pf_id), GLPRT_PTC511L(pf_id), |
| 3227 | pf->stat_prev_loaded, &prev_ps->tx_size_511, |
| 3228 | &cur_ps->tx_size_511); |
| 3229 | |
| 3230 | ice_stat_update40(hw, GLPRT_PTC1023H(pf_id), |
| 3231 | GLPRT_PTC1023L(pf_id), pf->stat_prev_loaded, |
| 3232 | &prev_ps->tx_size_1023, &cur_ps->tx_size_1023); |
| 3233 | |
| 3234 | ice_stat_update40(hw, GLPRT_PTC1522H(pf_id), |
| 3235 | GLPRT_PTC1522L(pf_id), pf->stat_prev_loaded, |
| 3236 | &prev_ps->tx_size_1522, &cur_ps->tx_size_1522); |
| 3237 | |
| 3238 | ice_stat_update40(hw, GLPRT_PTC9522H(pf_id), |
| 3239 | GLPRT_PTC9522L(pf_id), pf->stat_prev_loaded, |
| 3240 | &prev_ps->tx_size_big, &cur_ps->tx_size_big); |
| 3241 | |
| 3242 | ice_stat_update32(hw, GLPRT_LXONRXC(pf_id), pf->stat_prev_loaded, |
| 3243 | &prev_ps->link_xon_rx, &cur_ps->link_xon_rx); |
| 3244 | |
| 3245 | ice_stat_update32(hw, GLPRT_LXOFFRXC(pf_id), pf->stat_prev_loaded, |
| 3246 | &prev_ps->link_xoff_rx, &cur_ps->link_xoff_rx); |
| 3247 | |
| 3248 | ice_stat_update32(hw, GLPRT_LXONTXC(pf_id), pf->stat_prev_loaded, |
| 3249 | &prev_ps->link_xon_tx, &cur_ps->link_xon_tx); |
| 3250 | |
| 3251 | ice_stat_update32(hw, GLPRT_LXOFFTXC(pf_id), pf->stat_prev_loaded, |
| 3252 | &prev_ps->link_xoff_tx, &cur_ps->link_xoff_tx); |
| 3253 | |
| 3254 | ice_update_dcb_stats(pf); |
| 3255 | |
| 3256 | ice_stat_update32(hw, GLPRT_CRCERRS(pf_id), pf->stat_prev_loaded, |
| 3257 | &prev_ps->crc_errors, &cur_ps->crc_errors); |
| 3258 | |
| 3259 | ice_stat_update32(hw, GLPRT_ILLERRC(pf_id), pf->stat_prev_loaded, |
| 3260 | &prev_ps->illegal_bytes, &cur_ps->illegal_bytes); |
| 3261 | |
| 3262 | ice_stat_update32(hw, GLPRT_MLFC(pf_id), pf->stat_prev_loaded, |
| 3263 | &prev_ps->mac_local_faults, |
| 3264 | &cur_ps->mac_local_faults); |
| 3265 | |
| 3266 | ice_stat_update32(hw, GLPRT_MRFC(pf_id), pf->stat_prev_loaded, |
| 3267 | &prev_ps->mac_remote_faults, |
| 3268 | &cur_ps->mac_remote_faults); |
| 3269 | |
| 3270 | ice_stat_update32(hw, GLPRT_RLEC(pf_id), pf->stat_prev_loaded, |
| 3271 | &prev_ps->rx_len_errors, &cur_ps->rx_len_errors); |
| 3272 | |
| 3273 | ice_stat_update32(hw, GLPRT_RUC(pf_id), pf->stat_prev_loaded, |
| 3274 | &prev_ps->rx_undersize, &cur_ps->rx_undersize); |
| 3275 | |
| 3276 | ice_stat_update32(hw, GLPRT_RFC(pf_id), pf->stat_prev_loaded, |
| 3277 | &prev_ps->rx_fragments, &cur_ps->rx_fragments); |
| 3278 | |
| 3279 | ice_stat_update32(hw, GLPRT_ROC(pf_id), pf->stat_prev_loaded, |
| 3280 | &prev_ps->rx_oversize, &cur_ps->rx_oversize); |
| 3281 | |
| 3282 | ice_stat_update32(hw, GLPRT_RJC(pf_id), pf->stat_prev_loaded, |
| 3283 | &prev_ps->rx_jabber, &cur_ps->rx_jabber); |
| 3284 | |
| 3285 | pf->stat_prev_loaded = true; |
| 3286 | } |
| 3287 | |
| 3288 | /** |
| 3289 | * ice_get_stats64 - get statistics for network device structure |
| 3290 | * @netdev: network interface device structure |
| 3291 | * @stats: main device statistics structure |
| 3292 | */ |
| 3293 | static |
| 3294 | void ice_get_stats64(struct net_device *netdev, struct rtnl_link_stats64 *stats) |
| 3295 | { |
| 3296 | struct ice_netdev_priv *np = netdev_priv(netdev); |
| 3297 | struct rtnl_link_stats64 *vsi_stats; |
| 3298 | struct ice_vsi *vsi = np->vsi; |
| 3299 | |
| 3300 | vsi_stats = &vsi->net_stats; |
| 3301 | |
| 3302 | if (test_bit(__ICE_DOWN, vsi->state) || !vsi->num_txq || !vsi->num_rxq) |
| 3303 | return; |
| 3304 | /* netdev packet/byte stats come from ring counter. These are obtained |
| 3305 | * by summing up ring counters (done by ice_update_vsi_ring_stats). |
| 3306 | */ |
| 3307 | ice_update_vsi_ring_stats(vsi); |
| 3308 | stats->tx_packets = vsi_stats->tx_packets; |
| 3309 | stats->tx_bytes = vsi_stats->tx_bytes; |
| 3310 | stats->rx_packets = vsi_stats->rx_packets; |
| 3311 | stats->rx_bytes = vsi_stats->rx_bytes; |
| 3312 | |
| 3313 | /* The rest of the stats can be read from the hardware but instead we |
| 3314 | * just return values that the watchdog task has already obtained from |
| 3315 | * the hardware. |
| 3316 | */ |
| 3317 | stats->multicast = vsi_stats->multicast; |
| 3318 | stats->tx_errors = vsi_stats->tx_errors; |
| 3319 | stats->tx_dropped = vsi_stats->tx_dropped; |
| 3320 | stats->rx_errors = vsi_stats->rx_errors; |
| 3321 | stats->rx_dropped = vsi_stats->rx_dropped; |
| 3322 | stats->rx_crc_errors = vsi_stats->rx_crc_errors; |
| 3323 | stats->rx_length_errors = vsi_stats->rx_length_errors; |
| 3324 | } |
| 3325 | |
| 3326 | /** |
| 3327 | * ice_napi_disable_all - Disable NAPI for all q_vectors in the VSI |
| 3328 | * @vsi: VSI having NAPI disabled |
| 3329 | */ |
| 3330 | static void ice_napi_disable_all(struct ice_vsi *vsi) |
| 3331 | { |
| 3332 | int q_idx; |
| 3333 | |
| 3334 | if (!vsi->netdev) |
| 3335 | return; |
| 3336 | |
| 3337 | ice_for_each_q_vector(vsi, q_idx) { |
| 3338 | struct ice_q_vector *q_vector = vsi->q_vectors[q_idx]; |
| 3339 | |
| 3340 | if (q_vector->rx.ring || q_vector->tx.ring) |
| 3341 | napi_disable(&q_vector->napi); |
| 3342 | } |
| 3343 | } |
| 3344 | |
| 3345 | /** |
| 3346 | * ice_force_phys_link_state - Force the physical link state |
| 3347 | * @vsi: VSI to force the physical link state to up/down |
| 3348 | * @link_up: true/false indicates to set the physical link to up/down |
| 3349 | * |
| 3350 | * Force the physical link state by getting the current PHY capabilities from |
| 3351 | * hardware and setting the PHY config based on the determined capabilities. If |
| 3352 | * link changes a link event will be triggered because both the Enable Automatic |
| 3353 | * Link Update and LESM Enable bits are set when setting the PHY capabilities. |
| 3354 | * |
| 3355 | * Returns 0 on success, negative on failure |
| 3356 | */ |
| 3357 | static int ice_force_phys_link_state(struct ice_vsi *vsi, bool link_up) |
| 3358 | { |
| 3359 | struct ice_aqc_get_phy_caps_data *pcaps; |
| 3360 | struct ice_aqc_set_phy_cfg_data *cfg; |
| 3361 | struct ice_port_info *pi; |
| 3362 | struct device *dev; |
| 3363 | int retcode; |
| 3364 | |
| 3365 | if (!vsi || !vsi->port_info || !vsi->back) |
| 3366 | return -EINVAL; |
| 3367 | if (vsi->type != ICE_VSI_PF) |
| 3368 | return 0; |
| 3369 | |
| 3370 | dev = &vsi->back->pdev->dev; |
| 3371 | |
| 3372 | pi = vsi->port_info; |
| 3373 | |
| 3374 | pcaps = devm_kzalloc(dev, sizeof(*pcaps), GFP_KERNEL); |
| 3375 | if (!pcaps) |
| 3376 | return -ENOMEM; |
| 3377 | |
| 3378 | retcode = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_SW_CFG, pcaps, |
| 3379 | NULL); |
| 3380 | if (retcode) { |
| 3381 | dev_err(dev, |
| 3382 | "Failed to get phy capabilities, VSI %d error %d\n", |
| 3383 | vsi->vsi_num, retcode); |
| 3384 | retcode = -EIO; |
| 3385 | goto out; |
| 3386 | } |
| 3387 | |
| 3388 | /* No change in link */ |
| 3389 | if (link_up == !!(pcaps->caps & ICE_AQC_PHY_EN_LINK) && |
| 3390 | link_up == !!(pi->phy.link_info.link_info & ICE_AQ_LINK_UP)) |
| 3391 | goto out; |
| 3392 | |
| 3393 | cfg = devm_kzalloc(dev, sizeof(*cfg), GFP_KERNEL); |
| 3394 | if (!cfg) { |
| 3395 | retcode = -ENOMEM; |
| 3396 | goto out; |
| 3397 | } |
| 3398 | |
| 3399 | cfg->phy_type_low = pcaps->phy_type_low; |
| 3400 | cfg->phy_type_high = pcaps->phy_type_high; |
| 3401 | cfg->caps = pcaps->caps | ICE_AQ_PHY_ENA_AUTO_LINK_UPDT; |
| 3402 | cfg->low_power_ctrl = pcaps->low_power_ctrl; |
| 3403 | cfg->eee_cap = pcaps->eee_cap; |
| 3404 | cfg->eeer_value = pcaps->eeer_value; |
| 3405 | cfg->link_fec_opt = pcaps->link_fec_options; |
| 3406 | if (link_up) |
| 3407 | cfg->caps |= ICE_AQ_PHY_ENA_LINK; |
| 3408 | else |
| 3409 | cfg->caps &= ~ICE_AQ_PHY_ENA_LINK; |
| 3410 | |
| 3411 | retcode = ice_aq_set_phy_cfg(&vsi->back->hw, pi->lport, cfg, NULL); |
| 3412 | if (retcode) { |
| 3413 | dev_err(dev, "Failed to set phy config, VSI %d error %d\n", |
| 3414 | vsi->vsi_num, retcode); |
| 3415 | retcode = -EIO; |
| 3416 | } |
| 3417 | |
| 3418 | devm_kfree(dev, cfg); |
| 3419 | out: |
| 3420 | devm_kfree(dev, pcaps); |
| 3421 | return retcode; |
| 3422 | } |
| 3423 | |
| 3424 | /** |
| 3425 | * ice_down - Shutdown the connection |
| 3426 | * @vsi: The VSI being stopped |
| 3427 | */ |
| 3428 | int ice_down(struct ice_vsi *vsi) |
| 3429 | { |
| 3430 | int i, tx_err, rx_err, link_err = 0; |
| 3431 | |
| 3432 | /* Caller of this function is expected to set the |
| 3433 | * vsi->state __ICE_DOWN bit |
| 3434 | */ |
| 3435 | if (vsi->netdev) { |
| 3436 | netif_carrier_off(vsi->netdev); |
| 3437 | netif_tx_disable(vsi->netdev); |
| 3438 | } |
| 3439 | |
| 3440 | ice_vsi_dis_irq(vsi); |
| 3441 | |
| 3442 | tx_err = ice_vsi_stop_lan_tx_rings(vsi, ICE_NO_RESET, 0); |
| 3443 | if (tx_err) |
| 3444 | netdev_err(vsi->netdev, |
| 3445 | "Failed stop Tx rings, VSI %d error %d\n", |
| 3446 | vsi->vsi_num, tx_err); |
| 3447 | |
| 3448 | rx_err = ice_vsi_stop_rx_rings(vsi); |
| 3449 | if (rx_err) |
| 3450 | netdev_err(vsi->netdev, |
| 3451 | "Failed stop Rx rings, VSI %d error %d\n", |
| 3452 | vsi->vsi_num, rx_err); |
| 3453 | |
| 3454 | ice_napi_disable_all(vsi); |
| 3455 | |
| 3456 | if (test_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, vsi->back->flags)) { |
| 3457 | link_err = ice_force_phys_link_state(vsi, false); |
| 3458 | if (link_err) |
| 3459 | netdev_err(vsi->netdev, |
| 3460 | "Failed to set physical link down, VSI %d error %d\n", |
| 3461 | vsi->vsi_num, link_err); |
| 3462 | } |
| 3463 | |
| 3464 | ice_for_each_txq(vsi, i) |
| 3465 | ice_clean_tx_ring(vsi->tx_rings[i]); |
| 3466 | |
| 3467 | ice_for_each_rxq(vsi, i) |
| 3468 | ice_clean_rx_ring(vsi->rx_rings[i]); |
| 3469 | |
| 3470 | if (tx_err || rx_err || link_err) { |
| 3471 | netdev_err(vsi->netdev, |
| 3472 | "Failed to close VSI 0x%04X on switch 0x%04X\n", |
| 3473 | vsi->vsi_num, vsi->vsw->sw_id); |
| 3474 | return -EIO; |
| 3475 | } |
| 3476 | |
| 3477 | return 0; |
| 3478 | } |
| 3479 | |
| 3480 | /** |
| 3481 | * ice_vsi_setup_tx_rings - Allocate VSI Tx queue resources |
| 3482 | * @vsi: VSI having resources allocated |
| 3483 | * |
| 3484 | * Return 0 on success, negative on failure |
| 3485 | */ |
| 3486 | static int ice_vsi_setup_tx_rings(struct ice_vsi *vsi) |
| 3487 | { |
| 3488 | int i, err = 0; |
| 3489 | |
| 3490 | if (!vsi->num_txq) { |
| 3491 | dev_err(&vsi->back->pdev->dev, "VSI %d has 0 Tx queues\n", |
| 3492 | vsi->vsi_num); |
| 3493 | return -EINVAL; |
| 3494 | } |
| 3495 | |
| 3496 | ice_for_each_txq(vsi, i) { |
| 3497 | vsi->tx_rings[i]->netdev = vsi->netdev; |
| 3498 | err = ice_setup_tx_ring(vsi->tx_rings[i]); |
| 3499 | if (err) |
| 3500 | break; |
| 3501 | } |
| 3502 | |
| 3503 | return err; |
| 3504 | } |
| 3505 | |
| 3506 | /** |
| 3507 | * ice_vsi_setup_rx_rings - Allocate VSI Rx queue resources |
| 3508 | * @vsi: VSI having resources allocated |
| 3509 | * |
| 3510 | * Return 0 on success, negative on failure |
| 3511 | */ |
| 3512 | static int ice_vsi_setup_rx_rings(struct ice_vsi *vsi) |
| 3513 | { |
| 3514 | int i, err = 0; |
| 3515 | |
| 3516 | if (!vsi->num_rxq) { |
| 3517 | dev_err(&vsi->back->pdev->dev, "VSI %d has 0 Rx queues\n", |
| 3518 | vsi->vsi_num); |
| 3519 | return -EINVAL; |
| 3520 | } |
| 3521 | |
| 3522 | ice_for_each_rxq(vsi, i) { |
| 3523 | vsi->rx_rings[i]->netdev = vsi->netdev; |
| 3524 | err = ice_setup_rx_ring(vsi->rx_rings[i]); |
| 3525 | if (err) |
| 3526 | break; |
| 3527 | } |
| 3528 | |
| 3529 | return err; |
| 3530 | } |
| 3531 | |
| 3532 | /** |
| 3533 | * ice_vsi_req_irq - Request IRQ from the OS |
| 3534 | * @vsi: The VSI IRQ is being requested for |
| 3535 | * @basename: name for the vector |
| 3536 | * |
| 3537 | * Return 0 on success and a negative value on error |
| 3538 | */ |
| 3539 | static int ice_vsi_req_irq(struct ice_vsi *vsi, char *basename) |
| 3540 | { |
| 3541 | struct ice_pf *pf = vsi->back; |
| 3542 | int err = -EINVAL; |
| 3543 | |
| 3544 | if (test_bit(ICE_FLAG_MSIX_ENA, pf->flags)) |
| 3545 | err = ice_vsi_req_irq_msix(vsi, basename); |
| 3546 | |
| 3547 | return err; |
| 3548 | } |
| 3549 | |
| 3550 | /** |
| 3551 | * ice_vsi_open - Called when a network interface is made active |
| 3552 | * @vsi: the VSI to open |
| 3553 | * |
| 3554 | * Initialization of the VSI |
| 3555 | * |
| 3556 | * Returns 0 on success, negative value on error |
| 3557 | */ |
| 3558 | static int ice_vsi_open(struct ice_vsi *vsi) |
| 3559 | { |
| 3560 | char int_name[ICE_INT_NAME_STR_LEN]; |
| 3561 | struct ice_pf *pf = vsi->back; |
| 3562 | int err; |
| 3563 | |
| 3564 | /* allocate descriptors */ |
| 3565 | err = ice_vsi_setup_tx_rings(vsi); |
| 3566 | if (err) |
| 3567 | goto err_setup_tx; |
| 3568 | |
| 3569 | err = ice_vsi_setup_rx_rings(vsi); |
| 3570 | if (err) |
| 3571 | goto err_setup_rx; |
| 3572 | |
| 3573 | err = ice_vsi_cfg(vsi); |
| 3574 | if (err) |
| 3575 | goto err_setup_rx; |
| 3576 | |
| 3577 | snprintf(int_name, sizeof(int_name) - 1, "%s-%s", |
| 3578 | dev_driver_string(&pf->pdev->dev), vsi->netdev->name); |
| 3579 | err = ice_vsi_req_irq(vsi, int_name); |
| 3580 | if (err) |
| 3581 | goto err_setup_rx; |
| 3582 | |
| 3583 | /* Notify the stack of the actual queue counts. */ |
| 3584 | err = netif_set_real_num_tx_queues(vsi->netdev, vsi->num_txq); |
| 3585 | if (err) |
| 3586 | goto err_set_qs; |
| 3587 | |
| 3588 | err = netif_set_real_num_rx_queues(vsi->netdev, vsi->num_rxq); |
| 3589 | if (err) |
| 3590 | goto err_set_qs; |
| 3591 | |
| 3592 | err = ice_up_complete(vsi); |
| 3593 | if (err) |
| 3594 | goto err_up_complete; |
| 3595 | |
| 3596 | return 0; |
| 3597 | |
| 3598 | err_up_complete: |
| 3599 | ice_down(vsi); |
| 3600 | err_set_qs: |
| 3601 | ice_vsi_free_irq(vsi); |
| 3602 | err_setup_rx: |
| 3603 | ice_vsi_free_rx_rings(vsi); |
| 3604 | err_setup_tx: |
| 3605 | ice_vsi_free_tx_rings(vsi); |
| 3606 | |
| 3607 | return err; |
| 3608 | } |
| 3609 | |
| 3610 | /** |
| 3611 | * ice_vsi_release_all - Delete all VSIs |
| 3612 | * @pf: PF from which all VSIs are being removed |
| 3613 | */ |
| 3614 | static void ice_vsi_release_all(struct ice_pf *pf) |
| 3615 | { |
| 3616 | int err, i; |
| 3617 | |
| 3618 | if (!pf->vsi) |
| 3619 | return; |
| 3620 | |
| 3621 | ice_for_each_vsi(pf, i) { |
| 3622 | if (!pf->vsi[i]) |
| 3623 | continue; |
| 3624 | |
| 3625 | err = ice_vsi_release(pf->vsi[i]); |
| 3626 | if (err) |
| 3627 | dev_dbg(&pf->pdev->dev, |
| 3628 | "Failed to release pf->vsi[%d], err %d, vsi_num = %d\n", |
| 3629 | i, err, pf->vsi[i]->vsi_num); |
| 3630 | } |
| 3631 | } |
| 3632 | |
| 3633 | /** |
| 3634 | * ice_ena_vsi - resume a VSI |
| 3635 | * @vsi: the VSI being resume |
| 3636 | * @locked: is the rtnl_lock already held |
| 3637 | */ |
| 3638 | static int ice_ena_vsi(struct ice_vsi *vsi, bool locked) |
| 3639 | { |
| 3640 | int err = 0; |
| 3641 | |
| 3642 | if (!test_bit(__ICE_NEEDS_RESTART, vsi->state)) |
| 3643 | return err; |
| 3644 | |
| 3645 | clear_bit(__ICE_NEEDS_RESTART, vsi->state); |
| 3646 | |
| 3647 | if (vsi->netdev && vsi->type == ICE_VSI_PF) { |
| 3648 | struct net_device *netd = vsi->netdev; |
| 3649 | |
| 3650 | if (netif_running(vsi->netdev)) { |
| 3651 | if (locked) { |
| 3652 | err = netd->netdev_ops->ndo_open(netd); |
| 3653 | } else { |
| 3654 | rtnl_lock(); |
| 3655 | err = netd->netdev_ops->ndo_open(netd); |
| 3656 | rtnl_unlock(); |
| 3657 | } |
| 3658 | } else { |
| 3659 | err = ice_vsi_open(vsi); |
| 3660 | } |
| 3661 | } |
| 3662 | |
| 3663 | return err; |
| 3664 | } |
| 3665 | |
| 3666 | /** |
| 3667 | * ice_pf_ena_all_vsi - Resume all VSIs on a PF |
| 3668 | * @pf: the PF |
| 3669 | * @locked: is the rtnl_lock already held |
| 3670 | */ |
| 3671 | #ifdef CONFIG_DCB |
| 3672 | int ice_pf_ena_all_vsi(struct ice_pf *pf, bool locked) |
| 3673 | #else |
| 3674 | static int ice_pf_ena_all_vsi(struct ice_pf *pf, bool locked) |
| 3675 | #endif /* CONFIG_DCB */ |
| 3676 | { |
| 3677 | int v; |
| 3678 | |
| 3679 | ice_for_each_vsi(pf, v) |
| 3680 | if (pf->vsi[v]) |
| 3681 | if (ice_ena_vsi(pf->vsi[v], locked)) |
| 3682 | return -EIO; |
| 3683 | |
| 3684 | return 0; |
| 3685 | } |
| 3686 | |
| 3687 | /** |
| 3688 | * ice_vsi_rebuild_all - rebuild all VSIs in pf |
| 3689 | * @pf: the PF |
| 3690 | */ |
| 3691 | static int ice_vsi_rebuild_all(struct ice_pf *pf) |
| 3692 | { |
| 3693 | int i; |
| 3694 | |
| 3695 | /* loop through pf->vsi array and reinit the VSI if found */ |
| 3696 | ice_for_each_vsi(pf, i) { |
| 3697 | int err; |
| 3698 | |
| 3699 | if (!pf->vsi[i]) |
| 3700 | continue; |
| 3701 | |
| 3702 | err = ice_vsi_rebuild(pf->vsi[i]); |
| 3703 | if (err) { |
| 3704 | dev_err(&pf->pdev->dev, |
| 3705 | "VSI at index %d rebuild failed\n", |
| 3706 | pf->vsi[i]->idx); |
| 3707 | return err; |
| 3708 | } |
| 3709 | |
| 3710 | dev_info(&pf->pdev->dev, |
| 3711 | "VSI at index %d rebuilt. vsi_num = 0x%x\n", |
| 3712 | pf->vsi[i]->idx, pf->vsi[i]->vsi_num); |
| 3713 | } |
| 3714 | |
| 3715 | return 0; |
| 3716 | } |
| 3717 | |
| 3718 | /** |
| 3719 | * ice_vsi_replay_all - replay all VSIs configuration in the PF |
| 3720 | * @pf: the PF |
| 3721 | */ |
| 3722 | static int ice_vsi_replay_all(struct ice_pf *pf) |
| 3723 | { |
| 3724 | struct ice_hw *hw = &pf->hw; |
| 3725 | enum ice_status ret; |
| 3726 | int i; |
| 3727 | |
| 3728 | /* loop through pf->vsi array and replay the VSI if found */ |
| 3729 | ice_for_each_vsi(pf, i) { |
| 3730 | if (!pf->vsi[i]) |
| 3731 | continue; |
| 3732 | |
| 3733 | ret = ice_replay_vsi(hw, pf->vsi[i]->idx); |
| 3734 | if (ret) { |
| 3735 | dev_err(&pf->pdev->dev, |
| 3736 | "VSI at index %d replay failed %d\n", |
| 3737 | pf->vsi[i]->idx, ret); |
| 3738 | return -EIO; |
| 3739 | } |
| 3740 | |
| 3741 | /* Re-map HW VSI number, using VSI handle that has been |
| 3742 | * previously validated in ice_replay_vsi() call above |
| 3743 | */ |
| 3744 | pf->vsi[i]->vsi_num = ice_get_hw_vsi_num(hw, pf->vsi[i]->idx); |
| 3745 | |
| 3746 | dev_info(&pf->pdev->dev, |
| 3747 | "VSI at index %d filter replayed successfully - vsi_num %i\n", |
| 3748 | pf->vsi[i]->idx, pf->vsi[i]->vsi_num); |
| 3749 | } |
| 3750 | |
| 3751 | /* Clean up replay filter after successful re-configuration */ |
| 3752 | ice_replay_post(hw); |
| 3753 | return 0; |
| 3754 | } |
| 3755 | |
| 3756 | /** |
| 3757 | * ice_rebuild - rebuild after reset |
| 3758 | * @pf: pf to rebuild |
| 3759 | */ |
| 3760 | static void ice_rebuild(struct ice_pf *pf) |
| 3761 | { |
| 3762 | struct device *dev = &pf->pdev->dev; |
| 3763 | struct ice_hw *hw = &pf->hw; |
| 3764 | enum ice_status ret; |
| 3765 | int err, i; |
| 3766 | |
| 3767 | if (test_bit(__ICE_DOWN, pf->state)) |
| 3768 | goto clear_recovery; |
| 3769 | |
| 3770 | dev_dbg(dev, "rebuilding pf\n"); |
| 3771 | |
| 3772 | ret = ice_init_all_ctrlq(hw); |
| 3773 | if (ret) { |
| 3774 | dev_err(dev, "control queues init failed %d\n", ret); |
| 3775 | goto err_init_ctrlq; |
| 3776 | } |
| 3777 | |
| 3778 | ret = ice_clear_pf_cfg(hw); |
| 3779 | if (ret) { |
| 3780 | dev_err(dev, "clear PF configuration failed %d\n", ret); |
| 3781 | goto err_init_ctrlq; |
| 3782 | } |
| 3783 | |
| 3784 | ice_clear_pxe_mode(hw); |
| 3785 | |
| 3786 | ret = ice_get_caps(hw); |
| 3787 | if (ret) { |
| 3788 | dev_err(dev, "ice_get_caps failed %d\n", ret); |
| 3789 | goto err_init_ctrlq; |
| 3790 | } |
| 3791 | |
| 3792 | err = ice_sched_init_port(hw->port_info); |
| 3793 | if (err) |
| 3794 | goto err_sched_init_port; |
| 3795 | |
| 3796 | ice_dcb_rebuild(pf); |
| 3797 | |
| 3798 | /* reset search_hint of irq_trackers to 0 since interrupts are |
| 3799 | * reclaimed and could be allocated from beginning during VSI rebuild |
| 3800 | */ |
| 3801 | pf->sw_irq_tracker->search_hint = 0; |
| 3802 | pf->hw_irq_tracker->search_hint = 0; |
| 3803 | |
| 3804 | err = ice_vsi_rebuild_all(pf); |
| 3805 | if (err) { |
| 3806 | dev_err(dev, "ice_vsi_rebuild_all failed\n"); |
| 3807 | goto err_vsi_rebuild; |
| 3808 | } |
| 3809 | |
| 3810 | err = ice_update_link_info(hw->port_info); |
| 3811 | if (err) |
| 3812 | dev_err(&pf->pdev->dev, "Get link status error %d\n", err); |
| 3813 | |
| 3814 | /* Replay all VSIs Configuration, including filters after reset */ |
| 3815 | if (ice_vsi_replay_all(pf)) { |
| 3816 | dev_err(&pf->pdev->dev, |
| 3817 | "error replaying VSI configurations with switch filter rules\n"); |
| 3818 | goto err_vsi_rebuild; |
| 3819 | } |
| 3820 | |
| 3821 | /* start misc vector */ |
| 3822 | if (test_bit(ICE_FLAG_MSIX_ENA, pf->flags)) { |
| 3823 | err = ice_req_irq_msix_misc(pf); |
| 3824 | if (err) { |
| 3825 | dev_err(dev, "misc vector setup failed: %d\n", err); |
| 3826 | goto err_vsi_rebuild; |
| 3827 | } |
| 3828 | } |
| 3829 | |
| 3830 | /* restart the VSIs that were rebuilt and running before the reset */ |
| 3831 | err = ice_pf_ena_all_vsi(pf, false); |
| 3832 | if (err) { |
| 3833 | dev_err(&pf->pdev->dev, "error enabling VSIs\n"); |
| 3834 | /* no need to disable VSIs in tear down path in ice_rebuild() |
| 3835 | * since its already taken care in ice_vsi_open() |
| 3836 | */ |
| 3837 | goto err_vsi_rebuild; |
| 3838 | } |
| 3839 | |
| 3840 | ice_for_each_vsi(pf, i) { |
| 3841 | bool link_up; |
| 3842 | |
| 3843 | if (!pf->vsi[i] || pf->vsi[i]->type != ICE_VSI_PF) |
| 3844 | continue; |
| 3845 | ice_get_link_status(pf->vsi[i]->port_info, &link_up); |
| 3846 | if (link_up) { |
| 3847 | netif_carrier_on(pf->vsi[i]->netdev); |
| 3848 | netif_tx_wake_all_queues(pf->vsi[i]->netdev); |
| 3849 | } else { |
| 3850 | netif_carrier_off(pf->vsi[i]->netdev); |
| 3851 | netif_tx_stop_all_queues(pf->vsi[i]->netdev); |
| 3852 | } |
| 3853 | } |
| 3854 | |
| 3855 | /* if we get here, reset flow is successful */ |
| 3856 | clear_bit(__ICE_RESET_FAILED, pf->state); |
| 3857 | return; |
| 3858 | |
| 3859 | err_vsi_rebuild: |
| 3860 | ice_vsi_release_all(pf); |
| 3861 | err_sched_init_port: |
| 3862 | ice_sched_cleanup_all(hw); |
| 3863 | err_init_ctrlq: |
| 3864 | ice_shutdown_all_ctrlq(hw); |
| 3865 | set_bit(__ICE_RESET_FAILED, pf->state); |
| 3866 | clear_recovery: |
| 3867 | /* set this bit in PF state to control service task scheduling */ |
| 3868 | set_bit(__ICE_NEEDS_RESTART, pf->state); |
| 3869 | dev_err(dev, "Rebuild failed, unload and reload driver\n"); |
| 3870 | } |
| 3871 | |
| 3872 | /** |
| 3873 | * ice_change_mtu - NDO callback to change the MTU |
| 3874 | * @netdev: network interface device structure |
| 3875 | * @new_mtu: new value for maximum frame size |
| 3876 | * |
| 3877 | * Returns 0 on success, negative on failure |
| 3878 | */ |
| 3879 | static int ice_change_mtu(struct net_device *netdev, int new_mtu) |
| 3880 | { |
| 3881 | struct ice_netdev_priv *np = netdev_priv(netdev); |
| 3882 | struct ice_vsi *vsi = np->vsi; |
| 3883 | struct ice_pf *pf = vsi->back; |
| 3884 | u8 count = 0; |
| 3885 | |
| 3886 | if (new_mtu == netdev->mtu) { |
| 3887 | netdev_warn(netdev, "mtu is already %u\n", netdev->mtu); |
| 3888 | return 0; |
| 3889 | } |
| 3890 | |
| 3891 | if (new_mtu < netdev->min_mtu) { |
| 3892 | netdev_err(netdev, "new mtu invalid. min_mtu is %d\n", |
| 3893 | netdev->min_mtu); |
| 3894 | return -EINVAL; |
| 3895 | } else if (new_mtu > netdev->max_mtu) { |
| 3896 | netdev_err(netdev, "new mtu invalid. max_mtu is %d\n", |
| 3897 | netdev->min_mtu); |
| 3898 | return -EINVAL; |
| 3899 | } |
| 3900 | /* if a reset is in progress, wait for some time for it to complete */ |
| 3901 | do { |
| 3902 | if (ice_is_reset_in_progress(pf->state)) { |
| 3903 | count++; |
| 3904 | usleep_range(1000, 2000); |
| 3905 | } else { |
| 3906 | break; |
| 3907 | } |
| 3908 | |
| 3909 | } while (count < 100); |
| 3910 | |
| 3911 | if (count == 100) { |
| 3912 | netdev_err(netdev, "can't change mtu. Device is busy\n"); |
| 3913 | return -EBUSY; |
| 3914 | } |
| 3915 | |
| 3916 | netdev->mtu = new_mtu; |
| 3917 | |
| 3918 | /* if VSI is up, bring it down and then back up */ |
| 3919 | if (!test_and_set_bit(__ICE_DOWN, vsi->state)) { |
| 3920 | int err; |
| 3921 | |
| 3922 | err = ice_down(vsi); |
| 3923 | if (err) { |
| 3924 | netdev_err(netdev, "change mtu if_up err %d\n", err); |
| 3925 | return err; |
| 3926 | } |
| 3927 | |
| 3928 | err = ice_up(vsi); |
| 3929 | if (err) { |
| 3930 | netdev_err(netdev, "change mtu if_up err %d\n", err); |
| 3931 | return err; |
| 3932 | } |
| 3933 | } |
| 3934 | |
| 3935 | netdev_dbg(netdev, "changed mtu to %d\n", new_mtu); |
| 3936 | return 0; |
| 3937 | } |
| 3938 | |
| 3939 | /** |
| 3940 | * ice_set_rss - Set RSS keys and lut |
| 3941 | * @vsi: Pointer to VSI structure |
| 3942 | * @seed: RSS hash seed |
| 3943 | * @lut: Lookup table |
| 3944 | * @lut_size: Lookup table size |
| 3945 | * |
| 3946 | * Returns 0 on success, negative on failure |
| 3947 | */ |
| 3948 | int ice_set_rss(struct ice_vsi *vsi, u8 *seed, u8 *lut, u16 lut_size) |
| 3949 | { |
| 3950 | struct ice_pf *pf = vsi->back; |
| 3951 | struct ice_hw *hw = &pf->hw; |
| 3952 | enum ice_status status; |
| 3953 | |
| 3954 | if (seed) { |
| 3955 | struct ice_aqc_get_set_rss_keys *buf = |
| 3956 | (struct ice_aqc_get_set_rss_keys *)seed; |
| 3957 | |
| 3958 | status = ice_aq_set_rss_key(hw, vsi->idx, buf); |
| 3959 | |
| 3960 | if (status) { |
| 3961 | dev_err(&pf->pdev->dev, |
| 3962 | "Cannot set RSS key, err %d aq_err %d\n", |
| 3963 | status, hw->adminq.rq_last_status); |
| 3964 | return -EIO; |
| 3965 | } |
| 3966 | } |
| 3967 | |
| 3968 | if (lut) { |
| 3969 | status = ice_aq_set_rss_lut(hw, vsi->idx, vsi->rss_lut_type, |
| 3970 | lut, lut_size); |
| 3971 | if (status) { |
| 3972 | dev_err(&pf->pdev->dev, |
| 3973 | "Cannot set RSS lut, err %d aq_err %d\n", |
| 3974 | status, hw->adminq.rq_last_status); |
| 3975 | return -EIO; |
| 3976 | } |
| 3977 | } |
| 3978 | |
| 3979 | return 0; |
| 3980 | } |
| 3981 | |
| 3982 | /** |
| 3983 | * ice_get_rss - Get RSS keys and lut |
| 3984 | * @vsi: Pointer to VSI structure |
| 3985 | * @seed: Buffer to store the keys |
| 3986 | * @lut: Buffer to store the lookup table entries |
| 3987 | * @lut_size: Size of buffer to store the lookup table entries |
| 3988 | * |
| 3989 | * Returns 0 on success, negative on failure |
| 3990 | */ |
| 3991 | int ice_get_rss(struct ice_vsi *vsi, u8 *seed, u8 *lut, u16 lut_size) |
| 3992 | { |
| 3993 | struct ice_pf *pf = vsi->back; |
| 3994 | struct ice_hw *hw = &pf->hw; |
| 3995 | enum ice_status status; |
| 3996 | |
| 3997 | if (seed) { |
| 3998 | struct ice_aqc_get_set_rss_keys *buf = |
| 3999 | (struct ice_aqc_get_set_rss_keys *)seed; |
| 4000 | |
| 4001 | status = ice_aq_get_rss_key(hw, vsi->idx, buf); |
| 4002 | if (status) { |
| 4003 | dev_err(&pf->pdev->dev, |
| 4004 | "Cannot get RSS key, err %d aq_err %d\n", |
| 4005 | status, hw->adminq.rq_last_status); |
| 4006 | return -EIO; |
| 4007 | } |
| 4008 | } |
| 4009 | |
| 4010 | if (lut) { |
| 4011 | status = ice_aq_get_rss_lut(hw, vsi->idx, vsi->rss_lut_type, |
| 4012 | lut, lut_size); |
| 4013 | if (status) { |
| 4014 | dev_err(&pf->pdev->dev, |
| 4015 | "Cannot get RSS lut, err %d aq_err %d\n", |
| 4016 | status, hw->adminq.rq_last_status); |
| 4017 | return -EIO; |
| 4018 | } |
| 4019 | } |
| 4020 | |
| 4021 | return 0; |
| 4022 | } |
| 4023 | |
| 4024 | /** |
| 4025 | * ice_bridge_getlink - Get the hardware bridge mode |
| 4026 | * @skb: skb buff |
| 4027 | * @pid: process ID |
| 4028 | * @seq: RTNL message seq |
| 4029 | * @dev: the netdev being configured |
| 4030 | * @filter_mask: filter mask passed in |
| 4031 | * @nlflags: netlink flags passed in |
| 4032 | * |
| 4033 | * Return the bridge mode (VEB/VEPA) |
| 4034 | */ |
| 4035 | static int |
| 4036 | ice_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq, |
| 4037 | struct net_device *dev, u32 filter_mask, int nlflags) |
| 4038 | { |
| 4039 | struct ice_netdev_priv *np = netdev_priv(dev); |
| 4040 | struct ice_vsi *vsi = np->vsi; |
| 4041 | struct ice_pf *pf = vsi->back; |
| 4042 | u16 bmode; |
| 4043 | |
| 4044 | bmode = pf->first_sw->bridge_mode; |
| 4045 | |
| 4046 | return ndo_dflt_bridge_getlink(skb, pid, seq, dev, bmode, 0, 0, nlflags, |
| 4047 | filter_mask, NULL); |
| 4048 | } |
| 4049 | |
| 4050 | /** |
| 4051 | * ice_vsi_update_bridge_mode - Update VSI for switching bridge mode (VEB/VEPA) |
| 4052 | * @vsi: Pointer to VSI structure |
| 4053 | * @bmode: Hardware bridge mode (VEB/VEPA) |
| 4054 | * |
| 4055 | * Returns 0 on success, negative on failure |
| 4056 | */ |
| 4057 | static int ice_vsi_update_bridge_mode(struct ice_vsi *vsi, u16 bmode) |
| 4058 | { |
| 4059 | struct device *dev = &vsi->back->pdev->dev; |
| 4060 | struct ice_aqc_vsi_props *vsi_props; |
| 4061 | struct ice_hw *hw = &vsi->back->hw; |
| 4062 | struct ice_vsi_ctx *ctxt; |
| 4063 | enum ice_status status; |
| 4064 | int ret = 0; |
| 4065 | |
| 4066 | vsi_props = &vsi->info; |
| 4067 | |
| 4068 | ctxt = devm_kzalloc(dev, sizeof(*ctxt), GFP_KERNEL); |
| 4069 | if (!ctxt) |
| 4070 | return -ENOMEM; |
| 4071 | |
| 4072 | ctxt->info = vsi->info; |
| 4073 | |
| 4074 | if (bmode == BRIDGE_MODE_VEB) |
| 4075 | /* change from VEPA to VEB mode */ |
| 4076 | ctxt->info.sw_flags |= ICE_AQ_VSI_SW_FLAG_ALLOW_LB; |
| 4077 | else |
| 4078 | /* change from VEB to VEPA mode */ |
| 4079 | ctxt->info.sw_flags &= ~ICE_AQ_VSI_SW_FLAG_ALLOW_LB; |
| 4080 | ctxt->info.valid_sections = cpu_to_le16(ICE_AQ_VSI_PROP_SW_VALID); |
| 4081 | |
| 4082 | status = ice_update_vsi(hw, vsi->idx, ctxt, NULL); |
| 4083 | if (status) { |
| 4084 | dev_err(dev, "update VSI for bridge mode failed, bmode = %d err %d aq_err %d\n", |
| 4085 | bmode, status, hw->adminq.sq_last_status); |
| 4086 | ret = -EIO; |
| 4087 | goto out; |
| 4088 | } |
| 4089 | /* Update sw flags for book keeping */ |
| 4090 | vsi_props->sw_flags = ctxt->info.sw_flags; |
| 4091 | |
| 4092 | out: |
| 4093 | devm_kfree(dev, ctxt); |
| 4094 | return ret; |
| 4095 | } |
| 4096 | |
| 4097 | /** |
| 4098 | * ice_bridge_setlink - Set the hardware bridge mode |
| 4099 | * @dev: the netdev being configured |
| 4100 | * @nlh: RTNL message |
| 4101 | * @flags: bridge setlink flags |
| 4102 | * @extack: netlink extended ack |
| 4103 | * |
| 4104 | * Sets the bridge mode (VEB/VEPA) of the switch to which the netdev (VSI) is |
| 4105 | * hooked up to. Iterates through the PF VSI list and sets the loopback mode (if |
| 4106 | * not already set for all VSIs connected to this switch. And also update the |
| 4107 | * unicast switch filter rules for the corresponding switch of the netdev. |
| 4108 | */ |
| 4109 | static int |
| 4110 | ice_bridge_setlink(struct net_device *dev, struct nlmsghdr *nlh, |
| 4111 | u16 __always_unused flags, |
| 4112 | struct netlink_ext_ack __always_unused *extack) |
| 4113 | { |
| 4114 | struct ice_netdev_priv *np = netdev_priv(dev); |
| 4115 | struct ice_pf *pf = np->vsi->back; |
| 4116 | struct nlattr *attr, *br_spec; |
| 4117 | struct ice_hw *hw = &pf->hw; |
| 4118 | enum ice_status status; |
| 4119 | struct ice_sw *pf_sw; |
| 4120 | int rem, v, err = 0; |
| 4121 | |
| 4122 | pf_sw = pf->first_sw; |
| 4123 | /* find the attribute in the netlink message */ |
| 4124 | br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC); |
| 4125 | |
| 4126 | nla_for_each_nested(attr, br_spec, rem) { |
| 4127 | __u16 mode; |
| 4128 | |
| 4129 | if (nla_type(attr) != IFLA_BRIDGE_MODE) |
| 4130 | continue; |
| 4131 | mode = nla_get_u16(attr); |
| 4132 | if (mode != BRIDGE_MODE_VEPA && mode != BRIDGE_MODE_VEB) |
| 4133 | return -EINVAL; |
| 4134 | /* Continue if bridge mode is not being flipped */ |
| 4135 | if (mode == pf_sw->bridge_mode) |
| 4136 | continue; |
| 4137 | /* Iterates through the PF VSI list and update the loopback |
| 4138 | * mode of the VSI |
| 4139 | */ |
| 4140 | ice_for_each_vsi(pf, v) { |
| 4141 | if (!pf->vsi[v]) |
| 4142 | continue; |
| 4143 | err = ice_vsi_update_bridge_mode(pf->vsi[v], mode); |
| 4144 | if (err) |
| 4145 | return err; |
| 4146 | } |
| 4147 | |
| 4148 | hw->evb_veb = (mode == BRIDGE_MODE_VEB); |
| 4149 | /* Update the unicast switch filter rules for the corresponding |
| 4150 | * switch of the netdev |
| 4151 | */ |
| 4152 | status = ice_update_sw_rule_bridge_mode(hw); |
| 4153 | if (status) { |
| 4154 | netdev_err(dev, "switch rule update failed, mode = %d err %d aq_err %d\n", |
| 4155 | mode, status, hw->adminq.sq_last_status); |
| 4156 | /* revert hw->evb_veb */ |
| 4157 | hw->evb_veb = (pf_sw->bridge_mode == BRIDGE_MODE_VEB); |
| 4158 | return -EIO; |
| 4159 | } |
| 4160 | |
| 4161 | pf_sw->bridge_mode = mode; |
| 4162 | } |
| 4163 | |
| 4164 | return 0; |
| 4165 | } |
| 4166 | |
| 4167 | /** |
| 4168 | * ice_tx_timeout - Respond to a Tx Hang |
| 4169 | * @netdev: network interface device structure |
| 4170 | */ |
| 4171 | static void ice_tx_timeout(struct net_device *netdev) |
| 4172 | { |
| 4173 | struct ice_netdev_priv *np = netdev_priv(netdev); |
| 4174 | struct ice_ring *tx_ring = NULL; |
| 4175 | struct ice_vsi *vsi = np->vsi; |
| 4176 | struct ice_pf *pf = vsi->back; |
| 4177 | int hung_queue = -1; |
| 4178 | u32 i; |
| 4179 | |
| 4180 | pf->tx_timeout_count++; |
| 4181 | |
| 4182 | /* find the stopped queue the same way dev_watchdog() does */ |
| 4183 | for (i = 0; i < netdev->num_tx_queues; i++) { |
| 4184 | unsigned long trans_start; |
| 4185 | struct netdev_queue *q; |
| 4186 | |
| 4187 | q = netdev_get_tx_queue(netdev, i); |
| 4188 | trans_start = q->trans_start; |
| 4189 | if (netif_xmit_stopped(q) && |
| 4190 | time_after(jiffies, |
| 4191 | trans_start + netdev->watchdog_timeo)) { |
| 4192 | hung_queue = i; |
| 4193 | break; |
| 4194 | } |
| 4195 | } |
| 4196 | |
| 4197 | if (i == netdev->num_tx_queues) |
| 4198 | netdev_info(netdev, "tx_timeout: no netdev hung queue found\n"); |
| 4199 | else |
| 4200 | /* now that we have an index, find the tx_ring struct */ |
| 4201 | for (i = 0; i < vsi->num_txq; i++) |
| 4202 | if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc) |
| 4203 | if (hung_queue == vsi->tx_rings[i]->q_index) { |
| 4204 | tx_ring = vsi->tx_rings[i]; |
| 4205 | break; |
| 4206 | } |
| 4207 | |
| 4208 | /* Reset recovery level if enough time has elapsed after last timeout. |
| 4209 | * Also ensure no new reset action happens before next timeout period. |
| 4210 | */ |
| 4211 | if (time_after(jiffies, (pf->tx_timeout_last_recovery + HZ * 20))) |
| 4212 | pf->tx_timeout_recovery_level = 1; |
| 4213 | else if (time_before(jiffies, (pf->tx_timeout_last_recovery + |
| 4214 | netdev->watchdog_timeo))) |
| 4215 | return; |
| 4216 | |
| 4217 | if (tx_ring) { |
| 4218 | struct ice_hw *hw = &pf->hw; |
| 4219 | u32 head, val = 0; |
| 4220 | |
| 4221 | head = (rd32(hw, QTX_COMM_HEAD(vsi->txq_map[hung_queue])) & |
| 4222 | QTX_COMM_HEAD_HEAD_M) >> QTX_COMM_HEAD_HEAD_S; |
| 4223 | /* Read interrupt register */ |
| 4224 | if (test_bit(ICE_FLAG_MSIX_ENA, pf->flags)) |
| 4225 | val = rd32(hw, |
| 4226 | GLINT_DYN_CTL(tx_ring->q_vector->v_idx + |
| 4227 | tx_ring->vsi->hw_base_vector)); |
| 4228 | |
| 4229 | netdev_info(netdev, "tx_timeout: VSI_num: %d, Q %d, NTC: 0x%x, HW_HEAD: 0x%x, NTU: 0x%x, INT: 0x%x\n", |
| 4230 | vsi->vsi_num, hung_queue, tx_ring->next_to_clean, |
| 4231 | head, tx_ring->next_to_use, val); |
| 4232 | } |
| 4233 | |
| 4234 | pf->tx_timeout_last_recovery = jiffies; |
| 4235 | netdev_info(netdev, "tx_timeout recovery level %d, hung_queue %d\n", |
| 4236 | pf->tx_timeout_recovery_level, hung_queue); |
| 4237 | |
| 4238 | switch (pf->tx_timeout_recovery_level) { |
| 4239 | case 1: |
| 4240 | set_bit(__ICE_PFR_REQ, pf->state); |
| 4241 | break; |
| 4242 | case 2: |
| 4243 | set_bit(__ICE_CORER_REQ, pf->state); |
| 4244 | break; |
| 4245 | case 3: |
| 4246 | set_bit(__ICE_GLOBR_REQ, pf->state); |
| 4247 | break; |
| 4248 | default: |
| 4249 | netdev_err(netdev, "tx_timeout recovery unsuccessful, device is in unrecoverable state.\n"); |
| 4250 | set_bit(__ICE_DOWN, pf->state); |
| 4251 | set_bit(__ICE_NEEDS_RESTART, vsi->state); |
| 4252 | set_bit(__ICE_SERVICE_DIS, pf->state); |
| 4253 | break; |
| 4254 | } |
| 4255 | |
| 4256 | ice_service_task_schedule(pf); |
| 4257 | pf->tx_timeout_recovery_level++; |
| 4258 | } |
| 4259 | |
| 4260 | /** |
| 4261 | * ice_open - Called when a network interface becomes active |
| 4262 | * @netdev: network interface device structure |
| 4263 | * |
| 4264 | * The open entry point is called when a network interface is made |
| 4265 | * active by the system (IFF_UP). At this point all resources needed |
| 4266 | * for transmit and receive operations are allocated, the interrupt |
| 4267 | * handler is registered with the OS, the netdev watchdog is enabled, |
| 4268 | * and the stack is notified that the interface is ready. |
| 4269 | * |
| 4270 | * Returns 0 on success, negative value on failure |
| 4271 | */ |
| 4272 | static int ice_open(struct net_device *netdev) |
| 4273 | { |
| 4274 | struct ice_netdev_priv *np = netdev_priv(netdev); |
| 4275 | struct ice_vsi *vsi = np->vsi; |
| 4276 | int err; |
| 4277 | |
| 4278 | if (test_bit(__ICE_NEEDS_RESTART, vsi->back->state)) { |
| 4279 | netdev_err(netdev, "driver needs to be unloaded and reloaded\n"); |
| 4280 | return -EIO; |
| 4281 | } |
| 4282 | |
| 4283 | netif_carrier_off(netdev); |
| 4284 | |
| 4285 | err = ice_force_phys_link_state(vsi, true); |
| 4286 | if (err) { |
| 4287 | netdev_err(netdev, |
| 4288 | "Failed to set physical link up, error %d\n", err); |
| 4289 | return err; |
| 4290 | } |
| 4291 | |
| 4292 | err = ice_vsi_open(vsi); |
| 4293 | if (err) |
| 4294 | netdev_err(netdev, "Failed to open VSI 0x%04X on switch 0x%04X\n", |
| 4295 | vsi->vsi_num, vsi->vsw->sw_id); |
| 4296 | return err; |
| 4297 | } |
| 4298 | |
| 4299 | /** |
| 4300 | * ice_stop - Disables a network interface |
| 4301 | * @netdev: network interface device structure |
| 4302 | * |
| 4303 | * The stop entry point is called when an interface is de-activated by the OS, |
| 4304 | * and the netdevice enters the DOWN state. The hardware is still under the |
| 4305 | * driver's control, but the netdev interface is disabled. |
| 4306 | * |
| 4307 | * Returns success only - not allowed to fail |
| 4308 | */ |
| 4309 | static int ice_stop(struct net_device *netdev) |
| 4310 | { |
| 4311 | struct ice_netdev_priv *np = netdev_priv(netdev); |
| 4312 | struct ice_vsi *vsi = np->vsi; |
| 4313 | |
| 4314 | ice_vsi_close(vsi); |
| 4315 | |
| 4316 | return 0; |
| 4317 | } |
| 4318 | |
| 4319 | /** |
| 4320 | * ice_features_check - Validate encapsulated packet conforms to limits |
| 4321 | * @skb: skb buffer |
| 4322 | * @netdev: This port's netdev |
| 4323 | * @features: Offload features that the stack believes apply |
| 4324 | */ |
| 4325 | static netdev_features_t |
| 4326 | ice_features_check(struct sk_buff *skb, |
| 4327 | struct net_device __always_unused *netdev, |
| 4328 | netdev_features_t features) |
| 4329 | { |
| 4330 | size_t len; |
| 4331 | |
| 4332 | /* No point in doing any of this if neither checksum nor GSO are |
| 4333 | * being requested for this frame. We can rule out both by just |
| 4334 | * checking for CHECKSUM_PARTIAL |
| 4335 | */ |
| 4336 | if (skb->ip_summed != CHECKSUM_PARTIAL) |
| 4337 | return features; |
| 4338 | |
| 4339 | /* We cannot support GSO if the MSS is going to be less than |
| 4340 | * 64 bytes. If it is then we need to drop support for GSO. |
| 4341 | */ |
| 4342 | if (skb_is_gso(skb) && (skb_shinfo(skb)->gso_size < 64)) |
| 4343 | features &= ~NETIF_F_GSO_MASK; |
| 4344 | |
| 4345 | len = skb_network_header(skb) - skb->data; |
| 4346 | if (len & ~(ICE_TXD_MACLEN_MAX)) |
| 4347 | goto out_rm_features; |
| 4348 | |
| 4349 | len = skb_transport_header(skb) - skb_network_header(skb); |
| 4350 | if (len & ~(ICE_TXD_IPLEN_MAX)) |
| 4351 | goto out_rm_features; |
| 4352 | |
| 4353 | if (skb->encapsulation) { |
| 4354 | len = skb_inner_network_header(skb) - skb_transport_header(skb); |
| 4355 | if (len & ~(ICE_TXD_L4LEN_MAX)) |
| 4356 | goto out_rm_features; |
| 4357 | |
| 4358 | len = skb_inner_transport_header(skb) - |
| 4359 | skb_inner_network_header(skb); |
| 4360 | if (len & ~(ICE_TXD_IPLEN_MAX)) |
| 4361 | goto out_rm_features; |
| 4362 | } |
| 4363 | |
| 4364 | return features; |
| 4365 | out_rm_features: |
| 4366 | return features & ~(NETIF_F_CSUM_MASK | NETIF_F_GSO_MASK); |
| 4367 | } |
| 4368 | |
| 4369 | static const struct net_device_ops ice_netdev_ops = { |
| 4370 | .ndo_open = ice_open, |
| 4371 | .ndo_stop = ice_stop, |
| 4372 | .ndo_start_xmit = ice_start_xmit, |
| 4373 | .ndo_features_check = ice_features_check, |
| 4374 | .ndo_set_rx_mode = ice_set_rx_mode, |
| 4375 | .ndo_set_mac_address = ice_set_mac_address, |
| 4376 | .ndo_validate_addr = eth_validate_addr, |
| 4377 | .ndo_change_mtu = ice_change_mtu, |
| 4378 | .ndo_get_stats64 = ice_get_stats64, |
| 4379 | .ndo_set_vf_spoofchk = ice_set_vf_spoofchk, |
| 4380 | .ndo_set_vf_mac = ice_set_vf_mac, |
| 4381 | .ndo_get_vf_config = ice_get_vf_cfg, |
| 4382 | .ndo_set_vf_trust = ice_set_vf_trust, |
| 4383 | .ndo_set_vf_vlan = ice_set_vf_port_vlan, |
| 4384 | .ndo_set_vf_link_state = ice_set_vf_link_state, |
| 4385 | .ndo_vlan_rx_add_vid = ice_vlan_rx_add_vid, |
| 4386 | .ndo_vlan_rx_kill_vid = ice_vlan_rx_kill_vid, |
| 4387 | .ndo_set_features = ice_set_features, |
| 4388 | .ndo_bridge_getlink = ice_bridge_getlink, |
| 4389 | .ndo_bridge_setlink = ice_bridge_setlink, |
| 4390 | .ndo_fdb_add = ice_fdb_add, |
| 4391 | .ndo_fdb_del = ice_fdb_del, |
| 4392 | .ndo_tx_timeout = ice_tx_timeout, |
| 4393 | }; |