1 /* SPDX-License-Identifier: BSD-3-Clause */ 2 /* Copyright (c) 2020, Intel Corporation 3 * All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions are met: 7 * 8 * 1. Redistributions of source code must retain the above copyright notice, 9 * this list of conditions and the following disclaimer. 10 * 11 * 2. Redistributions in binary form must reproduce the above copyright 12 * notice, this list of conditions and the following disclaimer in the 13 * documentation and/or other materials provided with the distribution. 14 * 15 * 3. Neither the name of the Intel Corporation nor the names of its 16 * contributors may be used to endorse or promote products derived from 17 * this software without specific prior written permission. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" 20 * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 22 * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE 23 * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 24 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 25 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 26 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 27 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 28 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 29 * POSSIBILITY OF SUCH DAMAGE. 30 */ 31 /*$FreeBSD$*/ 32 33 /** 34 * @file if_ice_iflib.c 35 * @brief iflib driver implementation 36 * 37 * Contains the main entry point for the iflib driver implementation. It 38 * implements the various ifdi driver methods, and sets up the module and 39 * driver values to load an iflib driver. 40 */ 41 42 #include "ice_iflib.h" 43 #include "ice_drv_info.h" 44 #include "ice_switch.h" 45 #include "ice_sched.h" 46 47 #include <sys/module.h> 48 #include <sys/sockio.h> 49 #include <sys/smp.h> 50 #include <dev/pci/pcivar.h> 51 #include <dev/pci/pcireg.h> 52 53 /* 54 * Device method prototypes 55 */ 56 57 static void *ice_register(device_t); 58 static int ice_if_attach_pre(if_ctx_t); 59 static int ice_attach_pre_recovery_mode(struct ice_softc *sc); 60 static int ice_if_attach_post(if_ctx_t); 61 static void ice_attach_post_recovery_mode(struct ice_softc *sc); 62 static int ice_if_detach(if_ctx_t); 63 static int ice_if_tx_queues_alloc(if_ctx_t ctx, caddr_t *vaddrs, uint64_t *paddrs, int ntxqs, int ntxqsets); 64 static int ice_if_rx_queues_alloc(if_ctx_t ctx, caddr_t *vaddrs, uint64_t *paddrs, int nqs, int nqsets); 65 static int ice_if_msix_intr_assign(if_ctx_t ctx, int msix); 66 static void ice_if_queues_free(if_ctx_t ctx); 67 static int ice_if_mtu_set(if_ctx_t ctx, uint32_t mtu); 68 static void ice_if_intr_enable(if_ctx_t ctx); 69 static void ice_if_intr_disable(if_ctx_t ctx); 70 static int ice_if_rx_queue_intr_enable(if_ctx_t ctx, uint16_t rxqid); 71 static int ice_if_tx_queue_intr_enable(if_ctx_t ctx, uint16_t txqid); 72 static int ice_if_promisc_set(if_ctx_t ctx, int flags); 73 static void ice_if_media_status(if_ctx_t ctx, struct ifmediareq *ifmr); 74 static int ice_if_media_change(if_ctx_t ctx); 75 static void ice_if_init(if_ctx_t ctx); 76 static void ice_if_timer(if_ctx_t ctx, uint16_t qid); 77 static void ice_if_update_admin_status(if_ctx_t ctx); 78 static void ice_if_multi_set(if_ctx_t ctx); 79 static void ice_if_vlan_register(if_ctx_t ctx, u16 vtag); 80 static void ice_if_vlan_unregister(if_ctx_t ctx, u16 vtag); 81 static void ice_if_stop(if_ctx_t ctx); 82 static uint64_t ice_if_get_counter(if_ctx_t ctx, ift_counter counter); 83 static int ice_if_priv_ioctl(if_ctx_t ctx, u_long command, caddr_t data); 84 static int ice_if_i2c_req(if_ctx_t ctx, struct ifi2creq *req); 85 static int ice_if_suspend(if_ctx_t ctx); 86 static int ice_if_resume(if_ctx_t ctx); 87 88 static int ice_msix_que(void *arg); 89 static int ice_msix_admin(void *arg); 90 91 /* 92 * Helper function prototypes 93 */ 94 static int ice_pci_mapping(struct ice_softc *sc); 95 static void ice_free_pci_mapping(struct ice_softc *sc); 96 static void ice_update_link_status(struct ice_softc *sc, bool update_media); 97 static void ice_init_device_features(struct ice_softc *sc); 98 static void ice_init_tx_tracking(struct ice_vsi *vsi); 99 static void ice_handle_reset_event(struct ice_softc *sc); 100 static void ice_handle_pf_reset_request(struct ice_softc *sc); 101 static void ice_prepare_for_reset(struct ice_softc *sc); 102 static int ice_rebuild_pf_vsi_qmap(struct ice_softc *sc); 103 static void ice_rebuild(struct ice_softc *sc); 104 static void ice_rebuild_recovery_mode(struct ice_softc *sc); 105 static void ice_free_irqvs(struct ice_softc *sc); 106 static void ice_update_rx_mbuf_sz(struct ice_softc *sc); 107 static void ice_poll_for_media_avail(struct ice_softc *sc); 108 static void ice_setup_scctx(struct ice_softc *sc); 109 static int ice_allocate_msix(struct ice_softc *sc); 110 static void ice_admin_timer(void *arg); 111 static void ice_transition_recovery_mode(struct ice_softc *sc); 112 static void ice_transition_safe_mode(struct ice_softc *sc); 113 114 /* 115 * Device Interface Declaration 116 */ 117 118 /** 119 * @var ice_methods 120 * @brief ice driver method entry points 121 * 122 * List of device methods implementing the generic device interface used by 123 * the device stack to interact with the ice driver. Since this is an iflib 124 * driver, most of the methods point to the generic iflib implementation. 125 */ 126 static device_method_t ice_methods[] = { 127 /* Device interface */ 128 DEVMETHOD(device_register, ice_register), 129 DEVMETHOD(device_probe, iflib_device_probe_vendor), 130 DEVMETHOD(device_attach, iflib_device_attach), 131 DEVMETHOD(device_detach, iflib_device_detach), 132 DEVMETHOD(device_shutdown, iflib_device_shutdown), 133 DEVMETHOD(device_suspend, iflib_device_suspend), 134 DEVMETHOD(device_resume, iflib_device_resume), 135 DEVMETHOD_END 136 }; 137 138 /** 139 * @var ice_iflib_methods 140 * @brief iflib method entry points 141 * 142 * List of device methods used by the iflib stack to interact with this 143 * driver. These are the real main entry points used to interact with this 144 * driver. 145 */ 146 static device_method_t ice_iflib_methods[] = { 147 DEVMETHOD(ifdi_attach_pre, ice_if_attach_pre), 148 DEVMETHOD(ifdi_attach_post, ice_if_attach_post), 149 DEVMETHOD(ifdi_detach, ice_if_detach), 150 DEVMETHOD(ifdi_tx_queues_alloc, ice_if_tx_queues_alloc), 151 DEVMETHOD(ifdi_rx_queues_alloc, ice_if_rx_queues_alloc), 152 DEVMETHOD(ifdi_msix_intr_assign, ice_if_msix_intr_assign), 153 DEVMETHOD(ifdi_queues_free, ice_if_queues_free), 154 DEVMETHOD(ifdi_mtu_set, ice_if_mtu_set), 155 DEVMETHOD(ifdi_intr_enable, ice_if_intr_enable), 156 DEVMETHOD(ifdi_intr_disable, ice_if_intr_disable), 157 DEVMETHOD(ifdi_rx_queue_intr_enable, ice_if_rx_queue_intr_enable), 158 DEVMETHOD(ifdi_tx_queue_intr_enable, ice_if_tx_queue_intr_enable), 159 DEVMETHOD(ifdi_promisc_set, ice_if_promisc_set), 160 DEVMETHOD(ifdi_media_status, ice_if_media_status), 161 DEVMETHOD(ifdi_media_change, ice_if_media_change), 162 DEVMETHOD(ifdi_init, ice_if_init), 163 DEVMETHOD(ifdi_stop, ice_if_stop), 164 DEVMETHOD(ifdi_timer, ice_if_timer), 165 DEVMETHOD(ifdi_update_admin_status, ice_if_update_admin_status), 166 DEVMETHOD(ifdi_multi_set, ice_if_multi_set), 167 DEVMETHOD(ifdi_vlan_register, ice_if_vlan_register), 168 DEVMETHOD(ifdi_vlan_unregister, ice_if_vlan_unregister), 169 DEVMETHOD(ifdi_get_counter, ice_if_get_counter), 170 DEVMETHOD(ifdi_priv_ioctl, ice_if_priv_ioctl), 171 DEVMETHOD(ifdi_i2c_req, ice_if_i2c_req), 172 DEVMETHOD(ifdi_suspend, ice_if_suspend), 173 DEVMETHOD(ifdi_resume, ice_if_resume), 174 DEVMETHOD_END 175 }; 176 177 /** 178 * @var ice_driver 179 * @brief driver structure for the generic device stack 180 * 181 * driver_t definition used to setup the generic device methods. 182 */ 183 static driver_t ice_driver = { 184 .name = "ice", 185 .methods = ice_methods, 186 .size = sizeof(struct ice_softc), 187 }; 188 189 /** 190 * @var ice_iflib_driver 191 * @brief driver structure for the iflib stack 192 * 193 * driver_t definition used to setup the iflib device methods. 194 */ 195 static driver_t ice_iflib_driver = { 196 .name = "ice", 197 .methods = ice_iflib_methods, 198 .size = sizeof(struct ice_softc), 199 }; 200 201 extern struct if_txrx ice_txrx; 202 extern struct if_txrx ice_recovery_txrx; 203 204 /** 205 * @var ice_sctx 206 * @brief ice driver shared context 207 * 208 * Structure defining shared values (context) that is used by all instances of 209 * the device. Primarily used to setup details about how the iflib stack 210 * should treat this driver. Also defines the default, minimum, and maximum 211 * number of descriptors in each ring. 212 */ 213 static struct if_shared_ctx ice_sctx = { 214 .isc_magic = IFLIB_MAGIC, 215 .isc_q_align = PAGE_SIZE, 216 217 .isc_tx_maxsize = ICE_MAX_FRAME_SIZE, 218 /* We could technically set this as high as ICE_MAX_DMA_SEG_SIZE, but 219 * that doesn't make sense since that would be larger than the maximum 220 * size of a single packet. 221 */ 222 .isc_tx_maxsegsize = ICE_MAX_FRAME_SIZE, 223 224 /* XXX: This is only used by iflib to ensure that 225 * scctx->isc_tx_tso_size_max + the VLAN header is a valid size. 226 */ 227 .isc_tso_maxsize = ICE_TSO_SIZE + sizeof(struct ether_vlan_header), 228 /* XXX: This is used by iflib to set the number of segments in the TSO 229 * DMA tag. However, scctx->isc_tx_tso_segsize_max is used to set the 230 * related ifnet parameter. 231 */ 232 .isc_tso_maxsegsize = ICE_MAX_DMA_SEG_SIZE, 233 234 .isc_rx_maxsize = ICE_MAX_FRAME_SIZE, 235 .isc_rx_nsegments = ICE_MAX_RX_SEGS, 236 .isc_rx_maxsegsize = ICE_MAX_FRAME_SIZE, 237 238 .isc_nfl = 1, 239 .isc_ntxqs = 1, 240 .isc_nrxqs = 1, 241 242 .isc_admin_intrcnt = 1, 243 .isc_vendor_info = ice_vendor_info_array, 244 .isc_driver_version = __DECONST(char *, ice_driver_version), 245 .isc_driver = &ice_iflib_driver, 246 247 /* 248 * IFLIB_NEED_SCRATCH ensures that mbufs have scratch space available 249 * for hardware checksum offload 250 * 251 * IFLIB_TSO_INIT_IP ensures that the TSO packets have zeroed out the 252 * IP sum field, required by our hardware to calculate valid TSO 253 * checksums. 254 * 255 * IFLIB_ADMIN_ALWAYS_RUN ensures that the administrative task runs 256 * even when the interface is down. 257 * 258 * IFLIB_SKIP_MSIX allows the driver to handle allocating MSI-X 259 * vectors manually instead of relying on iflib code to do this. 260 */ 261 .isc_flags = IFLIB_NEED_SCRATCH | IFLIB_TSO_INIT_IP | 262 IFLIB_ADMIN_ALWAYS_RUN | IFLIB_SKIP_MSIX, 263 264 .isc_nrxd_min = {ICE_MIN_DESC_COUNT}, 265 .isc_ntxd_min = {ICE_MIN_DESC_COUNT}, 266 .isc_nrxd_max = {ICE_IFLIB_MAX_DESC_COUNT}, 267 .isc_ntxd_max = {ICE_IFLIB_MAX_DESC_COUNT}, 268 .isc_nrxd_default = {ICE_DEFAULT_DESC_COUNT}, 269 .isc_ntxd_default = {ICE_DEFAULT_DESC_COUNT}, 270 }; 271 272 /** 273 * @var ice_devclass 274 * @brief ice driver device class 275 * 276 * device class used to setup the ice driver module kobject class. 277 */ 278 devclass_t ice_devclass; 279 DRIVER_MODULE(ice, pci, ice_driver, ice_devclass, ice_module_event_handler, 0); 280 281 MODULE_VERSION(ice, 1); 282 MODULE_DEPEND(ice, pci, 1, 1, 1); 283 MODULE_DEPEND(ice, ether, 1, 1, 1); 284 MODULE_DEPEND(ice, iflib, 1, 1, 1); 285 286 IFLIB_PNP_INFO(pci, ice, ice_vendor_info_array); 287 288 /* Static driver-wide sysctls */ 289 #include "ice_iflib_sysctls.h" 290 291 /** 292 * ice_pci_mapping - Map PCI BAR memory 293 * @sc: device private softc 294 * 295 * Map PCI BAR 0 for device operation. 296 */ 297 static int 298 ice_pci_mapping(struct ice_softc *sc) 299 { 300 int rc; 301 302 /* Map BAR0 */ 303 rc = ice_map_bar(sc->dev, &sc->bar0, 0); 304 if (rc) 305 return rc; 306 307 return 0; 308 } 309 310 /** 311 * ice_free_pci_mapping - Release PCI BAR memory 312 * @sc: device private softc 313 * 314 * Release PCI BARs which were previously mapped by ice_pci_mapping(). 315 */ 316 static void 317 ice_free_pci_mapping(struct ice_softc *sc) 318 { 319 /* Free BAR0 */ 320 ice_free_bar(sc->dev, &sc->bar0); 321 } 322 323 /* 324 * Device methods 325 */ 326 327 /** 328 * ice_register - register device method callback 329 * @dev: the device being registered 330 * 331 * Returns a pointer to the shared context structure, which is used by iflib. 332 */ 333 static void * 334 ice_register(device_t dev __unused) 335 { 336 return &ice_sctx; 337 } /* ice_register */ 338 339 /** 340 * ice_setup_scctx - Setup the iflib softc context structure 341 * @sc: the device private structure 342 * 343 * Setup the parameters in if_softc_ctx_t structure used by the iflib stack 344 * when loading. 345 */ 346 static void 347 ice_setup_scctx(struct ice_softc *sc) 348 { 349 if_softc_ctx_t scctx = sc->scctx; 350 struct ice_hw *hw = &sc->hw; 351 bool safe_mode, recovery_mode; 352 353 safe_mode = ice_is_bit_set(sc->feat_en, ICE_FEATURE_SAFE_MODE); 354 recovery_mode = ice_test_state(&sc->state, ICE_STATE_RECOVERY_MODE); 355 356 /* 357 * If the driver loads in Safe mode or Recovery mode, limit iflib to 358 * a single queue pair. 359 */ 360 if (safe_mode || recovery_mode) { 361 scctx->isc_ntxqsets = scctx->isc_nrxqsets = 1; 362 scctx->isc_ntxqsets_max = 1; 363 scctx->isc_nrxqsets_max = 1; 364 } else { 365 /* 366 * iflib initially sets the isc_ntxqsets and isc_nrxqsets to 367 * the values of the override sysctls. Cache these initial 368 * values so that the driver can be aware of what the iflib 369 * sysctl value is when setting up MSI-X vectors. 370 */ 371 sc->ifc_sysctl_ntxqs = scctx->isc_ntxqsets; 372 sc->ifc_sysctl_nrxqs = scctx->isc_nrxqsets; 373 374 if (scctx->isc_ntxqsets == 0) 375 scctx->isc_ntxqsets = hw->func_caps.common_cap.rss_table_size; 376 if (scctx->isc_nrxqsets == 0) 377 scctx->isc_nrxqsets = hw->func_caps.common_cap.rss_table_size; 378 379 scctx->isc_ntxqsets_max = hw->func_caps.common_cap.num_txq; 380 scctx->isc_nrxqsets_max = hw->func_caps.common_cap.num_rxq; 381 382 /* 383 * Sanity check that the iflib sysctl values are within the 384 * maximum supported range. 385 */ 386 if (sc->ifc_sysctl_ntxqs > scctx->isc_ntxqsets_max) 387 sc->ifc_sysctl_ntxqs = scctx->isc_ntxqsets_max; 388 if (sc->ifc_sysctl_nrxqs > scctx->isc_nrxqsets_max) 389 sc->ifc_sysctl_nrxqs = scctx->isc_nrxqsets_max; 390 } 391 392 scctx->isc_txqsizes[0] = roundup2(scctx->isc_ntxd[0] 393 * sizeof(struct ice_tx_desc), DBA_ALIGN); 394 scctx->isc_rxqsizes[0] = roundup2(scctx->isc_nrxd[0] 395 * sizeof(union ice_32b_rx_flex_desc), DBA_ALIGN); 396 397 scctx->isc_tx_nsegments = ICE_MAX_TX_SEGS; 398 scctx->isc_tx_tso_segments_max = ICE_MAX_TSO_SEGS; 399 scctx->isc_tx_tso_size_max = ICE_TSO_SIZE; 400 scctx->isc_tx_tso_segsize_max = ICE_MAX_DMA_SEG_SIZE; 401 402 scctx->isc_msix_bar = PCIR_BAR(ICE_MSIX_BAR); 403 scctx->isc_rss_table_size = hw->func_caps.common_cap.rss_table_size; 404 405 /* 406 * If the driver loads in recovery mode, disable Tx/Rx functionality 407 */ 408 if (recovery_mode) 409 scctx->isc_txrx = &ice_recovery_txrx; 410 else 411 scctx->isc_txrx = &ice_txrx; 412 413 /* 414 * If the driver loads in Safe mode or Recovery mode, disable 415 * advanced features including hardware offloads. 416 */ 417 if (safe_mode || recovery_mode) { 418 scctx->isc_capenable = ICE_SAFE_CAPS; 419 scctx->isc_tx_csum_flags = 0; 420 } else { 421 scctx->isc_capenable = ICE_FULL_CAPS; 422 scctx->isc_tx_csum_flags = ICE_CSUM_OFFLOAD; 423 } 424 425 scctx->isc_capabilities = scctx->isc_capenable; 426 } /* ice_setup_scctx */ 427 428 /** 429 * ice_if_attach_pre - Early device attach logic 430 * @ctx: the iflib context structure 431 * 432 * Called by iflib during the attach process. Earliest main driver entry 433 * point which performs necessary hardware and driver initialization. Called 434 * before the Tx and Rx queues are allocated. 435 */ 436 static int 437 ice_if_attach_pre(if_ctx_t ctx) 438 { 439 struct ice_softc *sc = (struct ice_softc *)iflib_get_softc(ctx); 440 enum ice_fw_modes fw_mode; 441 enum ice_status status; 442 if_softc_ctx_t scctx; 443 struct ice_hw *hw; 444 device_t dev; 445 int err; 446 447 device_printf(iflib_get_dev(ctx), "Loading the iflib ice driver\n"); 448 449 sc->ctx = ctx; 450 sc->media = iflib_get_media(ctx); 451 sc->sctx = iflib_get_sctx(ctx); 452 sc->iflib_ctx_lock = iflib_ctx_lock_get(ctx); 453 454 dev = sc->dev = iflib_get_dev(ctx); 455 scctx = sc->scctx = iflib_get_softc_ctx(ctx); 456 457 hw = &sc->hw; 458 hw->back = sc; 459 460 snprintf(sc->admin_mtx_name, sizeof(sc->admin_mtx_name), 461 "%s:admin", device_get_nameunit(dev)); 462 mtx_init(&sc->admin_mtx, sc->admin_mtx_name, NULL, MTX_DEF); 463 callout_init_mtx(&sc->admin_timer, &sc->admin_mtx, 0); 464 465 ASSERT_CTX_LOCKED(sc); 466 467 if (ice_pci_mapping(sc)) { 468 err = (ENXIO); 469 goto destroy_admin_timer; 470 } 471 472 /* Save off the PCI information */ 473 ice_save_pci_info(hw, dev); 474 475 /* create tunables as early as possible */ 476 ice_add_device_tunables(sc); 477 478 /* Setup ControlQ lengths */ 479 ice_set_ctrlq_len(hw); 480 481 fw_mode = ice_get_fw_mode(hw); 482 if (fw_mode == ICE_FW_MODE_REC) { 483 device_printf(dev, "Firmware recovery mode detected. Limiting functionality. Refer to Intel(R) Ethernet Adapters and Devices User Guide for details on firmware recovery mode.\n"); 484 485 err = ice_attach_pre_recovery_mode(sc); 486 if (err) 487 goto free_pci_mapping; 488 489 return (0); 490 } 491 492 /* Initialize the hw data structure */ 493 status = ice_init_hw(hw); 494 if (status) { 495 if (status == ICE_ERR_FW_API_VER) { 496 /* Enter recovery mode, so that the driver remains 497 * loaded. This way, if the system administrator 498 * cannot update the driver, they may still attempt to 499 * downgrade the NVM. 500 */ 501 err = ice_attach_pre_recovery_mode(sc); 502 if (err) 503 goto free_pci_mapping; 504 505 return (0); 506 } else { 507 err = EIO; 508 device_printf(dev, "Unable to initialize hw, err %s aq_err %s\n", 509 ice_status_str(status), 510 ice_aq_str(hw->adminq.sq_last_status)); 511 } 512 goto free_pci_mapping; 513 } 514 515 /* Notify firmware of the device driver version */ 516 err = ice_send_version(sc); 517 if (err) 518 goto deinit_hw; 519 520 ice_load_pkg_file(sc); 521 522 err = ice_init_link_events(sc); 523 if (err) { 524 device_printf(dev, "ice_init_link_events failed: %s\n", 525 ice_err_str(err)); 526 goto deinit_hw; 527 } 528 529 ice_print_nvm_version(sc); 530 531 ice_init_device_features(sc); 532 533 /* Setup the MAC address */ 534 iflib_set_mac(ctx, hw->port_info->mac.lan_addr); 535 536 /* Setup the iflib softc context structure */ 537 ice_setup_scctx(sc); 538 539 /* Initialize the Tx queue manager */ 540 err = ice_resmgr_init(&sc->tx_qmgr, hw->func_caps.common_cap.num_txq); 541 if (err) { 542 device_printf(dev, "Unable to initialize Tx queue manager: %s\n", 543 ice_err_str(err)); 544 goto deinit_hw; 545 } 546 547 /* Initialize the Rx queue manager */ 548 err = ice_resmgr_init(&sc->rx_qmgr, hw->func_caps.common_cap.num_rxq); 549 if (err) { 550 device_printf(dev, "Unable to initialize Rx queue manager: %s\n", 551 ice_err_str(err)); 552 goto free_tx_qmgr; 553 } 554 555 /* Initialize the interrupt resource manager */ 556 err = ice_alloc_intr_tracking(sc); 557 if (err) 558 /* Errors are already printed */ 559 goto free_rx_qmgr; 560 561 /* Determine maximum number of VSIs we'll prepare for */ 562 sc->num_available_vsi = min(ICE_MAX_VSI_AVAILABLE, 563 hw->func_caps.guar_num_vsi); 564 565 if (!sc->num_available_vsi) { 566 err = EIO; 567 device_printf(dev, "No VSIs allocated to host\n"); 568 goto free_intr_tracking; 569 } 570 571 /* Allocate storage for the VSI pointers */ 572 sc->all_vsi = (struct ice_vsi **) 573 malloc(sizeof(struct ice_vsi *) * sc->num_available_vsi, 574 M_ICE, M_WAITOK | M_ZERO); 575 if (!sc->all_vsi) { 576 err = ENOMEM; 577 device_printf(dev, "Unable to allocate VSI array\n"); 578 goto free_intr_tracking; 579 } 580 581 /* 582 * Prepare the statically allocated primary PF VSI in the softc 583 * structure. Other VSIs will be dynamically allocated as needed. 584 */ 585 ice_setup_pf_vsi(sc); 586 587 err = ice_alloc_vsi_qmap(&sc->pf_vsi, scctx->isc_ntxqsets_max, 588 scctx->isc_nrxqsets_max); 589 if (err) { 590 device_printf(dev, "Unable to allocate VSI Queue maps\n"); 591 goto free_main_vsi; 592 } 593 594 /* Allocate MSI-X vectors (due to isc_flags IFLIB_SKIP_MSIX) */ 595 err = ice_allocate_msix(sc); 596 if (err) 597 goto free_main_vsi; 598 599 return 0; 600 601 free_main_vsi: 602 /* ice_release_vsi will free the queue maps if they were allocated */ 603 ice_release_vsi(&sc->pf_vsi); 604 free(sc->all_vsi, M_ICE); 605 sc->all_vsi = NULL; 606 free_intr_tracking: 607 ice_free_intr_tracking(sc); 608 free_rx_qmgr: 609 ice_resmgr_destroy(&sc->rx_qmgr); 610 free_tx_qmgr: 611 ice_resmgr_destroy(&sc->tx_qmgr); 612 deinit_hw: 613 ice_deinit_hw(hw); 614 free_pci_mapping: 615 ice_free_pci_mapping(sc); 616 destroy_admin_timer: 617 mtx_lock(&sc->admin_mtx); 618 callout_stop(&sc->admin_timer); 619 mtx_unlock(&sc->admin_mtx); 620 mtx_destroy(&sc->admin_mtx); 621 return err; 622 } /* ice_if_attach_pre */ 623 624 /** 625 * ice_attach_pre_recovery_mode - Limited driver attach_pre for FW recovery 626 * @sc: the device private softc 627 * 628 * Loads the device driver in limited Firmware Recovery mode, intended to 629 * allow users to update the firmware to attempt to recover the device. 630 * 631 * @remark We may enter recovery mode in case either (a) the firmware is 632 * detected to be in an invalid state and must be re-programmed, or (b) the 633 * driver detects that the loaded firmware has a non-compatible API version 634 * that the driver cannot operate with. 635 */ 636 static int 637 ice_attach_pre_recovery_mode(struct ice_softc *sc) 638 { 639 ice_set_state(&sc->state, ICE_STATE_RECOVERY_MODE); 640 641 /* Setup the iflib softc context */ 642 ice_setup_scctx(sc); 643 644 /* Setup the PF VSI back pointer */ 645 sc->pf_vsi.sc = sc; 646 647 /* 648 * We still need to allocate MSI-X vectors since we need one vector to 649 * run the administrative admin interrupt 650 */ 651 return ice_allocate_msix(sc); 652 } 653 654 /** 655 * ice_update_link_status - notify OS of link state change 656 * @sc: device private softc structure 657 * @update_media: true if we should update media even if link didn't change 658 * 659 * Called to notify iflib core of link status changes. Should be called once 660 * during attach_post, and whenever link status changes during runtime. 661 * 662 * This call only updates the currently supported media types if the link 663 * status changed, or if update_media is set to true. 664 */ 665 static void 666 ice_update_link_status(struct ice_softc *sc, bool update_media) 667 { 668 struct ice_hw *hw = &sc->hw; 669 enum ice_status status; 670 671 /* Never report link up when in recovery mode */ 672 if (ice_test_state(&sc->state, ICE_STATE_RECOVERY_MODE)) 673 return; 674 675 /* Report link status to iflib only once each time it changes */ 676 if (!ice_testandset_state(&sc->state, ICE_STATE_LINK_STATUS_REPORTED)) { 677 if (sc->link_up) { /* link is up */ 678 uint64_t baudrate = ice_aq_speed_to_rate(sc->hw.port_info); 679 680 ice_set_default_local_lldp_mib(sc); 681 682 iflib_link_state_change(sc->ctx, LINK_STATE_UP, baudrate); 683 684 ice_link_up_msg(sc); 685 686 update_media = true; 687 } else { /* link is down */ 688 iflib_link_state_change(sc->ctx, LINK_STATE_DOWN, 0); 689 690 update_media = true; 691 } 692 } 693 694 /* Update the supported media types */ 695 if (update_media) { 696 status = ice_add_media_types(sc, sc->media); 697 if (status) 698 device_printf(sc->dev, "Error adding device media types: %s aq_err %s\n", 699 ice_status_str(status), 700 ice_aq_str(hw->adminq.sq_last_status)); 701 } 702 703 /* TODO: notify VFs of link state change */ 704 } 705 706 /** 707 * ice_if_attach_post - Late device attach logic 708 * @ctx: the iflib context structure 709 * 710 * Called by iflib to finish up attaching the device. Performs any attach 711 * logic which must wait until after the Tx and Rx queues have been 712 * allocated. 713 */ 714 static int 715 ice_if_attach_post(if_ctx_t ctx) 716 { 717 struct ice_softc *sc = (struct ice_softc *)iflib_get_softc(ctx); 718 if_t ifp = iflib_get_ifp(ctx); 719 int err; 720 721 ASSERT_CTX_LOCKED(sc); 722 723 /* We don't yet support loading if MSI-X is not supported */ 724 if (sc->scctx->isc_intr != IFLIB_INTR_MSIX) { 725 device_printf(sc->dev, "The ice driver does not support loading without MSI-X\n"); 726 return (ENOTSUP); 727 } 728 729 /* The ifnet structure hasn't yet been initialized when the attach_pre 730 * handler is called, so wait until attach_post to setup the 731 * isc_max_frame_size. 732 */ 733 734 sc->ifp = ifp; 735 sc->scctx->isc_max_frame_size = ifp->if_mtu + 736 ETHER_HDR_LEN + ETHER_CRC_LEN + ETHER_VLAN_ENCAP_LEN; 737 738 /* 739 * If we are in recovery mode, only perform a limited subset of 740 * initialization to support NVM recovery. 741 */ 742 if (ice_test_state(&sc->state, ICE_STATE_RECOVERY_MODE)) { 743 ice_attach_post_recovery_mode(sc); 744 return (0); 745 } 746 747 sc->pf_vsi.max_frame_size = sc->scctx->isc_max_frame_size; 748 749 err = ice_initialize_vsi(&sc->pf_vsi); 750 if (err) { 751 device_printf(sc->dev, "Unable to initialize Main VSI: %s\n", 752 ice_err_str(err)); 753 return err; 754 } 755 756 /* Configure the main PF VSI for RSS */ 757 err = ice_config_rss(&sc->pf_vsi); 758 if (err) { 759 device_printf(sc->dev, 760 "Unable to configure RSS for the main VSI, err %s\n", 761 ice_err_str(err)); 762 return err; 763 } 764 765 /* Configure switch to drop transmitted LLDP and PAUSE frames */ 766 err = ice_cfg_pf_ethertype_filters(sc); 767 if (err) 768 return err; 769 770 ice_get_and_print_bus_info(sc); 771 772 ice_set_link_management_mode(sc); 773 774 ice_init_saved_phy_cfg(sc); 775 776 ice_add_device_sysctls(sc); 777 778 /* Get DCBX/LLDP state and start DCBX agent */ 779 ice_init_dcb_setup(sc); 780 781 /* Setup link configuration parameters */ 782 ice_init_link_configuration(sc); 783 ice_update_link_status(sc, true); 784 785 /* Configure interrupt causes for the administrative interrupt */ 786 ice_configure_misc_interrupts(sc); 787 788 /* Enable ITR 0 right away, so that we can handle admin interrupts */ 789 ice_enable_intr(&sc->hw, sc->irqvs[0].me); 790 791 /* Start the admin timer */ 792 mtx_lock(&sc->admin_mtx); 793 callout_reset(&sc->admin_timer, hz/2, ice_admin_timer, sc); 794 mtx_unlock(&sc->admin_mtx); 795 796 return 0; 797 } /* ice_if_attach_post */ 798 799 /** 800 * ice_attach_post_recovery_mode - Limited driver attach_post for FW recovery 801 * @sc: the device private softc 802 * 803 * Performs minimal work to prepare the driver to recover an NVM in case the 804 * firmware is in recovery mode. 805 */ 806 static void 807 ice_attach_post_recovery_mode(struct ice_softc *sc) 808 { 809 /* Configure interrupt causes for the administrative interrupt */ 810 ice_configure_misc_interrupts(sc); 811 812 /* Enable ITR 0 right away, so that we can handle admin interrupts */ 813 ice_enable_intr(&sc->hw, sc->irqvs[0].me); 814 815 /* Start the admin timer */ 816 mtx_lock(&sc->admin_mtx); 817 callout_reset(&sc->admin_timer, hz/2, ice_admin_timer, sc); 818 mtx_unlock(&sc->admin_mtx); 819 } 820 821 /** 822 * ice_free_irqvs - Free IRQ vector memory 823 * @sc: the device private softc structure 824 * 825 * Free IRQ vector memory allocated during ice_if_msix_intr_assign. 826 */ 827 static void 828 ice_free_irqvs(struct ice_softc *sc) 829 { 830 struct ice_vsi *vsi = &sc->pf_vsi; 831 if_ctx_t ctx = sc->ctx; 832 int i; 833 834 /* If the irqvs array is NULL, then there are no vectors to free */ 835 if (sc->irqvs == NULL) 836 return; 837 838 /* Free the IRQ vectors */ 839 for (i = 0; i < sc->num_irq_vectors; i++) 840 iflib_irq_free(ctx, &sc->irqvs[i].irq); 841 842 /* Clear the irqv pointers */ 843 for (i = 0; i < vsi->num_rx_queues; i++) 844 vsi->rx_queues[i].irqv = NULL; 845 846 for (i = 0; i < vsi->num_tx_queues; i++) 847 vsi->tx_queues[i].irqv = NULL; 848 849 /* Release the vector array memory */ 850 free(sc->irqvs, M_ICE); 851 sc->irqvs = NULL; 852 sc->num_irq_vectors = 0; 853 } 854 855 /** 856 * ice_if_detach - Device driver detach logic 857 * @ctx: iflib context structure 858 * 859 * Perform device shutdown logic to detach the device driver. 860 * 861 * Note that there is no guarantee of the ordering of ice_if_queues_free() and 862 * ice_if_detach(). It is possible for the functions to be called in either 863 * order, and they must not assume to have a strict ordering. 864 */ 865 static int 866 ice_if_detach(if_ctx_t ctx) 867 { 868 struct ice_softc *sc = (struct ice_softc *)iflib_get_softc(ctx); 869 struct ice_vsi *vsi = &sc->pf_vsi; 870 int i; 871 872 ASSERT_CTX_LOCKED(sc); 873 874 /* Indicate that we're detaching */ 875 ice_set_state(&sc->state, ICE_STATE_DETACHING); 876 877 /* Stop the admin timer */ 878 mtx_lock(&sc->admin_mtx); 879 callout_stop(&sc->admin_timer); 880 mtx_unlock(&sc->admin_mtx); 881 mtx_destroy(&sc->admin_mtx); 882 883 /* Free allocated media types */ 884 ifmedia_removeall(sc->media); 885 886 /* Free the Tx and Rx sysctl contexts, and assign NULL to the node 887 * pointers. Note, the calls here and those in ice_if_queues_free() 888 * are *BOTH* necessary, as we cannot guarantee which path will be 889 * run first 890 */ 891 ice_vsi_del_txqs_ctx(vsi); 892 ice_vsi_del_rxqs_ctx(vsi); 893 894 /* Release MSI-X resources */ 895 ice_free_irqvs(sc); 896 897 for (i = 0; i < sc->num_available_vsi; i++) { 898 if (sc->all_vsi[i]) 899 ice_release_vsi(sc->all_vsi[i]); 900 } 901 902 if (sc->all_vsi) { 903 free(sc->all_vsi, M_ICE); 904 sc->all_vsi = NULL; 905 } 906 907 /* Release MSI-X memory */ 908 pci_release_msi(sc->dev); 909 910 if (sc->msix_table != NULL) { 911 bus_release_resource(sc->dev, SYS_RES_MEMORY, 912 rman_get_rid(sc->msix_table), 913 sc->msix_table); 914 sc->msix_table = NULL; 915 } 916 917 ice_free_intr_tracking(sc); 918 919 /* Destroy the queue managers */ 920 ice_resmgr_destroy(&sc->tx_qmgr); 921 ice_resmgr_destroy(&sc->rx_qmgr); 922 923 if (!ice_test_state(&sc->state, ICE_STATE_RECOVERY_MODE)) 924 ice_deinit_hw(&sc->hw); 925 926 ice_free_pci_mapping(sc); 927 928 return 0; 929 } /* ice_if_detach */ 930 931 /** 932 * ice_if_tx_queues_alloc - Allocate Tx queue memory 933 * @ctx: iflib context structure 934 * @vaddrs: virtual addresses for the queue memory 935 * @paddrs: physical addresses for the queue memory 936 * @ntxqs: the number of Tx queues per set (should always be 1) 937 * @ntxqsets: the number of Tx queue sets to allocate 938 * 939 * Called by iflib to allocate Tx queues for the device. Allocates driver 940 * memory to track each queue, the status arrays used for descriptor 941 * status reporting, and Tx queue sysctls. 942 */ 943 static int 944 ice_if_tx_queues_alloc(if_ctx_t ctx, caddr_t *vaddrs, uint64_t *paddrs, 945 int __invariant_only ntxqs, int ntxqsets) 946 { 947 struct ice_softc *sc = (struct ice_softc *)iflib_get_softc(ctx); 948 struct ice_vsi *vsi = &sc->pf_vsi; 949 struct ice_tx_queue *txq; 950 int err, i, j; 951 952 MPASS(ntxqs == 1); 953 MPASS(sc->scctx->isc_ntxd[0] <= ICE_MAX_DESC_COUNT); 954 ASSERT_CTX_LOCKED(sc); 955 956 /* Do not bother allocating queues if we're in recovery mode */ 957 if (ice_test_state(&sc->state, ICE_STATE_RECOVERY_MODE)) 958 return (0); 959 960 /* Allocate queue structure memory */ 961 if (!(vsi->tx_queues = 962 (struct ice_tx_queue *) malloc(sizeof(struct ice_tx_queue) * ntxqsets, M_ICE, M_WAITOK | M_ZERO))) { 963 device_printf(sc->dev, "Unable to allocate Tx queue memory\n"); 964 return (ENOMEM); 965 } 966 967 /* Allocate report status arrays */ 968 for (i = 0, txq = vsi->tx_queues; i < ntxqsets; i++, txq++) { 969 if (!(txq->tx_rsq = 970 (uint16_t *) malloc(sizeof(uint16_t) * sc->scctx->isc_ntxd[0], M_ICE, M_WAITOK))) { 971 device_printf(sc->dev, "Unable to allocate tx_rsq memory\n"); 972 err = ENOMEM; 973 goto free_tx_queues; 974 } 975 /* Initialize report status array */ 976 for (j = 0; j < sc->scctx->isc_ntxd[0]; j++) 977 txq->tx_rsq[j] = QIDX_INVALID; 978 } 979 980 /* Assign queues from PF space to the main VSI */ 981 err = ice_resmgr_assign_contiguous(&sc->tx_qmgr, vsi->tx_qmap, ntxqsets); 982 if (err) { 983 device_printf(sc->dev, "Unable to assign PF queues: %s\n", 984 ice_err_str(err)); 985 goto free_tx_queues; 986 } 987 vsi->qmap_type = ICE_RESMGR_ALLOC_CONTIGUOUS; 988 989 /* Add Tx queue sysctls context */ 990 ice_vsi_add_txqs_ctx(vsi); 991 992 for (i = 0, txq = vsi->tx_queues; i < ntxqsets; i++, txq++) { 993 txq->me = i; 994 txq->vsi = vsi; 995 996 /* store the queue size for easier access */ 997 txq->desc_count = sc->scctx->isc_ntxd[0]; 998 999 /* get the virtual and physical address of the hardware queues */ 1000 txq->tail = QTX_COMM_DBELL(vsi->tx_qmap[i]); 1001 txq->tx_base = (struct ice_tx_desc *)vaddrs[i]; 1002 txq->tx_paddr = paddrs[i]; 1003 1004 ice_add_txq_sysctls(txq); 1005 } 1006 1007 vsi->num_tx_queues = ntxqsets; 1008 1009 return (0); 1010 1011 free_tx_queues: 1012 for (i = 0, txq = vsi->tx_queues; i < ntxqsets; i++, txq++) { 1013 if (txq->tx_rsq != NULL) { 1014 free(txq->tx_rsq, M_ICE); 1015 txq->tx_rsq = NULL; 1016 } 1017 } 1018 free(vsi->tx_queues, M_ICE); 1019 vsi->tx_queues = NULL; 1020 return err; 1021 } 1022 1023 /** 1024 * ice_if_rx_queues_alloc - Allocate Rx queue memory 1025 * @ctx: iflib context structure 1026 * @vaddrs: virtual addresses for the queue memory 1027 * @paddrs: physical addresses for the queue memory 1028 * @nrxqs: number of Rx queues per set (should always be 1) 1029 * @nrxqsets: number of Rx queue sets to allocate 1030 * 1031 * Called by iflib to allocate Rx queues for the device. Allocates driver 1032 * memory to track each queue, as well as sets up the Rx queue sysctls. 1033 */ 1034 static int 1035 ice_if_rx_queues_alloc(if_ctx_t ctx, caddr_t *vaddrs, uint64_t *paddrs, 1036 int __invariant_only nrxqs, int nrxqsets) 1037 { 1038 struct ice_softc *sc = (struct ice_softc *)iflib_get_softc(ctx); 1039 struct ice_vsi *vsi = &sc->pf_vsi; 1040 struct ice_rx_queue *rxq; 1041 int err, i; 1042 1043 MPASS(nrxqs == 1); 1044 MPASS(sc->scctx->isc_nrxd[0] <= ICE_MAX_DESC_COUNT); 1045 ASSERT_CTX_LOCKED(sc); 1046 1047 /* Do not bother allocating queues if we're in recovery mode */ 1048 if (ice_test_state(&sc->state, ICE_STATE_RECOVERY_MODE)) 1049 return (0); 1050 1051 /* Allocate queue structure memory */ 1052 if (!(vsi->rx_queues = 1053 (struct ice_rx_queue *) malloc(sizeof(struct ice_rx_queue) * nrxqsets, M_ICE, M_WAITOK | M_ZERO))) { 1054 device_printf(sc->dev, "Unable to allocate Rx queue memory\n"); 1055 return (ENOMEM); 1056 } 1057 1058 /* Assign queues from PF space to the main VSI */ 1059 err = ice_resmgr_assign_contiguous(&sc->rx_qmgr, vsi->rx_qmap, nrxqsets); 1060 if (err) { 1061 device_printf(sc->dev, "Unable to assign PF queues: %s\n", 1062 ice_err_str(err)); 1063 goto free_rx_queues; 1064 } 1065 vsi->qmap_type = ICE_RESMGR_ALLOC_CONTIGUOUS; 1066 1067 /* Add Rx queue sysctls context */ 1068 ice_vsi_add_rxqs_ctx(vsi); 1069 1070 for (i = 0, rxq = vsi->rx_queues; i < nrxqsets; i++, rxq++) { 1071 rxq->me = i; 1072 rxq->vsi = vsi; 1073 1074 /* store the queue size for easier access */ 1075 rxq->desc_count = sc->scctx->isc_nrxd[0]; 1076 1077 /* get the virtual and physical address of the hardware queues */ 1078 rxq->tail = QRX_TAIL(vsi->rx_qmap[i]); 1079 rxq->rx_base = (union ice_32b_rx_flex_desc *)vaddrs[i]; 1080 rxq->rx_paddr = paddrs[i]; 1081 1082 ice_add_rxq_sysctls(rxq); 1083 } 1084 1085 vsi->num_rx_queues = nrxqsets; 1086 1087 return (0); 1088 1089 free_rx_queues: 1090 free(vsi->rx_queues, M_ICE); 1091 vsi->rx_queues = NULL; 1092 return err; 1093 } 1094 1095 /** 1096 * ice_if_queues_free - Free queue memory 1097 * @ctx: the iflib context structure 1098 * 1099 * Free queue memory allocated by ice_if_tx_queues_alloc() and 1100 * ice_if_rx_queues_alloc(). 1101 * 1102 * There is no guarantee that ice_if_queues_free() and ice_if_detach() will be 1103 * called in the same order. It's possible for ice_if_queues_free() to be 1104 * called prior to ice_if_detach(), and vice versa. 1105 * 1106 * For this reason, the main VSI is a static member of the ice_softc, which is 1107 * not free'd until after iflib finishes calling both of these functions. 1108 * 1109 * Thus, care must be taken in how we manage the memory being freed by this 1110 * function, and in what tasks it can and must perform. 1111 */ 1112 static void 1113 ice_if_queues_free(if_ctx_t ctx) 1114 { 1115 struct ice_softc *sc = (struct ice_softc *)iflib_get_softc(ctx); 1116 struct ice_vsi *vsi = &sc->pf_vsi; 1117 struct ice_tx_queue *txq; 1118 int i; 1119 1120 /* Free the Tx and Rx sysctl contexts, and assign NULL to the node 1121 * pointers. Note, the calls here and those in ice_if_detach() 1122 * are *BOTH* necessary, as we cannot guarantee which path will be 1123 * run first 1124 */ 1125 ice_vsi_del_txqs_ctx(vsi); 1126 ice_vsi_del_rxqs_ctx(vsi); 1127 1128 /* Release MSI-X IRQ vectors, if not yet released in ice_if_detach */ 1129 ice_free_irqvs(sc); 1130 1131 if (vsi->tx_queues != NULL) { 1132 /* free the tx_rsq arrays */ 1133 for (i = 0, txq = vsi->tx_queues; i < vsi->num_tx_queues; i++, txq++) { 1134 if (txq->tx_rsq != NULL) { 1135 free(txq->tx_rsq, M_ICE); 1136 txq->tx_rsq = NULL; 1137 } 1138 } 1139 free(vsi->tx_queues, M_ICE); 1140 vsi->tx_queues = NULL; 1141 vsi->num_tx_queues = 0; 1142 } 1143 if (vsi->rx_queues != NULL) { 1144 free(vsi->rx_queues, M_ICE); 1145 vsi->rx_queues = NULL; 1146 vsi->num_rx_queues = 0; 1147 } 1148 } 1149 1150 /** 1151 * ice_msix_que - Fast interrupt handler for MSI-X receive queues 1152 * @arg: The Rx queue memory 1153 * 1154 * Interrupt filter function for iflib MSI-X interrupts. Called by iflib when 1155 * an MSI-X interrupt for a given queue is triggered. Currently this just asks 1156 * iflib to schedule the main Rx thread. 1157 */ 1158 static int 1159 ice_msix_que(void *arg) 1160 { 1161 struct ice_rx_queue __unused *rxq = (struct ice_rx_queue *)arg; 1162 1163 /* TODO: dynamic ITR algorithm?? */ 1164 1165 return (FILTER_SCHEDULE_THREAD); 1166 } 1167 1168 /** 1169 * ice_msix_admin - Fast interrupt handler for MSI-X admin interrupt 1170 * @arg: pointer to device softc memory 1171 * 1172 * Called by iflib when an administrative interrupt occurs. Should perform any 1173 * fast logic for handling the interrupt cause, and then indicate whether the 1174 * admin task needs to be queued. 1175 */ 1176 static int 1177 ice_msix_admin(void *arg) 1178 { 1179 struct ice_softc *sc = (struct ice_softc *)arg; 1180 struct ice_hw *hw = &sc->hw; 1181 device_t dev = sc->dev; 1182 u32 oicr; 1183 1184 /* There is no safe way to modify the enabled miscellaneous causes of 1185 * the OICR vector at runtime, as doing so would be prone to race 1186 * conditions. Reading PFINT_OICR will unmask the associated interrupt 1187 * causes and allow future interrupts to occur. The admin interrupt 1188 * vector will not be re-enabled until after we exit this function, 1189 * but any delayed tasks must be resilient against possible "late 1190 * arrival" interrupts that occur while we're already handling the 1191 * task. This is done by using state bits and serializing these 1192 * delayed tasks via the admin status task function. 1193 */ 1194 oicr = rd32(hw, PFINT_OICR); 1195 1196 /* Processing multiple controlq interrupts on a single vector does not 1197 * provide an indication of which controlq triggered the interrupt. 1198 * We might try reading the INTEVENT bit of the respective PFINT_*_CTL 1199 * registers. However, the INTEVENT bit is not guaranteed to be set as 1200 * it gets automatically cleared when the hardware acknowledges the 1201 * interrupt. 1202 * 1203 * This means we don't really have a good indication of whether or 1204 * which controlq triggered this interrupt. We'll just notify the 1205 * admin task that it should check all the controlqs. 1206 */ 1207 ice_set_state(&sc->state, ICE_STATE_CONTROLQ_EVENT_PENDING); 1208 1209 if (oicr & PFINT_OICR_VFLR_M) { 1210 ice_set_state(&sc->state, ICE_STATE_VFLR_PENDING); 1211 } 1212 1213 if (oicr & PFINT_OICR_MAL_DETECT_M) { 1214 ice_set_state(&sc->state, ICE_STATE_MDD_PENDING); 1215 } 1216 1217 if (oicr & PFINT_OICR_GRST_M) { 1218 u32 reset; 1219 1220 reset = (rd32(hw, GLGEN_RSTAT) & GLGEN_RSTAT_RESET_TYPE_M) >> 1221 GLGEN_RSTAT_RESET_TYPE_S; 1222 1223 if (reset == ICE_RESET_CORER) 1224 sc->soft_stats.corer_count++; 1225 else if (reset == ICE_RESET_GLOBR) 1226 sc->soft_stats.globr_count++; 1227 else 1228 sc->soft_stats.empr_count++; 1229 1230 /* There are a couple of bits at play for handling resets. 1231 * First, the ICE_STATE_RESET_OICR_RECV bit is used to 1232 * indicate that the driver has received an OICR with a reset 1233 * bit active, indicating that a CORER/GLOBR/EMPR is about to 1234 * happen. Second, we set hw->reset_ongoing to indicate that 1235 * the hardware is in reset. We will set this back to false as 1236 * soon as the driver has determined that the hardware is out 1237 * of reset. 1238 * 1239 * If the driver wishes to trigger a reqest, it can set one of 1240 * the ICE_STATE_RESET_*_REQ bits, which will trigger the 1241 * correct type of reset. 1242 */ 1243 if (!ice_testandset_state(&sc->state, ICE_STATE_RESET_OICR_RECV)) 1244 hw->reset_ongoing = true; 1245 } 1246 1247 if (oicr & PFINT_OICR_ECC_ERR_M) { 1248 device_printf(dev, "ECC Error detected!\n"); 1249 ice_set_state(&sc->state, ICE_STATE_RESET_PFR_REQ); 1250 } 1251 1252 if (oicr & PFINT_OICR_PE_CRITERR_M) { 1253 device_printf(dev, "Critical Protocol Engine Error detected!\n"); 1254 ice_set_state(&sc->state, ICE_STATE_RESET_PFR_REQ); 1255 } 1256 1257 if (oicr & PFINT_OICR_PCI_EXCEPTION_M) { 1258 device_printf(dev, "PCI Exception detected!\n"); 1259 ice_set_state(&sc->state, ICE_STATE_RESET_PFR_REQ); 1260 } 1261 1262 if (oicr & PFINT_OICR_HMC_ERR_M) { 1263 /* Log the HMC errors, but don't disable the interrupt cause */ 1264 ice_log_hmc_error(hw, dev); 1265 } 1266 1267 return (FILTER_SCHEDULE_THREAD); 1268 } 1269 1270 /** 1271 * ice_allocate_msix - Allocate MSI-X vectors for the interface 1272 * @sc: the device private softc 1273 * 1274 * Map the MSI-X bar, and then request MSI-X vectors in a two-stage process. 1275 * 1276 * First, determine a suitable total number of vectors based on the number 1277 * of CPUs, RSS buckets, the administrative vector, and other demands such as 1278 * RDMA. 1279 * 1280 * Request the desired amount of vectors, and see how many we obtain. If we 1281 * don't obtain as many as desired, reduce the demands by lowering the number 1282 * of requested queues or reducing the demand from other features such as 1283 * RDMA. 1284 * 1285 * @remark This function is required because the driver sets the 1286 * IFLIB_SKIP_MSIX flag indicating that the driver will manage MSI-X vectors 1287 * manually. 1288 * 1289 * @remark This driver will only use MSI-X vectors. If this is not possible, 1290 * neither MSI or legacy interrupts will be tried. 1291 * 1292 * @post on success this function must set the following scctx parameters: 1293 * isc_vectors, isc_nrxqsets, isc_ntxqsets, and isc_intr. 1294 * 1295 * @returns zero on success or an error code on failure. 1296 */ 1297 static int 1298 ice_allocate_msix(struct ice_softc *sc) 1299 { 1300 bool iflib_override_queue_count = false; 1301 if_softc_ctx_t scctx = sc->scctx; 1302 device_t dev = sc->dev; 1303 cpuset_t cpus; 1304 int bar, queues, vectors, requested; 1305 int err = 0; 1306 1307 /* Allocate the MSI-X bar */ 1308 bar = scctx->isc_msix_bar; 1309 sc->msix_table = bus_alloc_resource_any(dev, SYS_RES_MEMORY, &bar, RF_ACTIVE); 1310 if (!sc->msix_table) { 1311 device_printf(dev, "Unable to map MSI-X table\n"); 1312 return (ENOMEM); 1313 } 1314 1315 /* Check if the iflib queue count sysctls have been set */ 1316 if (sc->ifc_sysctl_ntxqs || sc->ifc_sysctl_nrxqs) 1317 iflib_override_queue_count = true; 1318 1319 err = bus_get_cpus(dev, INTR_CPUS, sizeof(cpus), &cpus); 1320 if (err) { 1321 device_printf(dev, "%s: Unable to fetch the CPU list: %s\n", 1322 __func__, ice_err_str(err)); 1323 CPU_COPY(&all_cpus, &cpus); 1324 } 1325 1326 /* Attempt to mimic behavior of iflib_msix_init */ 1327 if (iflib_override_queue_count) { 1328 /* 1329 * If the override sysctls have been set, limit the queues to 1330 * the number of logical CPUs. 1331 */ 1332 queues = mp_ncpus; 1333 } else { 1334 /* 1335 * Otherwise, limit the queue count to the CPUs associated 1336 * with the NUMA node the device is associated with. 1337 */ 1338 queues = CPU_COUNT(&cpus); 1339 } 1340 1341 /* Clamp to the number of RSS buckets */ 1342 queues = imin(queues, rss_getnumbuckets()); 1343 1344 /* 1345 * Clamp the number of queue pairs to the minimum of the requested Tx 1346 * and Rx queues. 1347 */ 1348 queues = imin(queues, sc->ifc_sysctl_ntxqs ?: scctx->isc_ntxqsets); 1349 queues = imin(queues, sc->ifc_sysctl_nrxqs ?: scctx->isc_nrxqsets); 1350 1351 /* 1352 * Determine the number of vectors to request. Note that we also need 1353 * to allocate one vector for administrative tasks. 1354 */ 1355 requested = queues + 1; 1356 1357 vectors = requested; 1358 1359 err = pci_alloc_msix(dev, &vectors); 1360 if (err) { 1361 device_printf(dev, "Failed to allocate %d MSI-X vectors, err %s\n", 1362 vectors, ice_err_str(err)); 1363 goto err_free_msix_table; 1364 } 1365 1366 /* If we don't receive enough vectors, reduce demands */ 1367 if (vectors < requested) { 1368 int diff = requested - vectors; 1369 1370 device_printf(dev, "Requested %d MSI-X vectors, but got only %d\n", 1371 requested, vectors); 1372 1373 /* 1374 * If we still have a difference, we need to reduce the number 1375 * of queue pairs. 1376 * 1377 * However, we still need at least one vector for the admin 1378 * interrupt and one queue pair. 1379 */ 1380 if (queues <= diff) { 1381 device_printf(dev, "Unable to allocate sufficient MSI-X vectors\n"); 1382 err = (ERANGE); 1383 goto err_pci_release_msi; 1384 } 1385 1386 queues -= diff; 1387 } 1388 1389 device_printf(dev, "Using %d Tx and Rx queues\n", queues); 1390 device_printf(dev, "Using MSI-X interrupts with %d vectors\n", 1391 vectors); 1392 1393 scctx->isc_vectors = vectors; 1394 scctx->isc_nrxqsets = queues; 1395 scctx->isc_ntxqsets = queues; 1396 scctx->isc_intr = IFLIB_INTR_MSIX; 1397 1398 /* Interrupt allocation tracking isn't required in recovery mode, 1399 * since neither RDMA nor VFs are enabled. 1400 */ 1401 if (ice_test_state(&sc->state, ICE_STATE_RECOVERY_MODE)) 1402 return (0); 1403 1404 /* Keep track of which interrupt indices are being used for what */ 1405 sc->lan_vectors = vectors; 1406 err = ice_resmgr_assign_contiguous(&sc->imgr, sc->pf_imap, sc->lan_vectors); 1407 if (err) { 1408 device_printf(dev, "Unable to assign PF interrupt mapping: %s\n", 1409 ice_err_str(err)); 1410 goto err_pci_release_msi; 1411 } 1412 1413 return (0); 1414 1415 err_pci_release_msi: 1416 pci_release_msi(dev); 1417 err_free_msix_table: 1418 if (sc->msix_table != NULL) { 1419 bus_release_resource(sc->dev, SYS_RES_MEMORY, 1420 rman_get_rid(sc->msix_table), 1421 sc->msix_table); 1422 sc->msix_table = NULL; 1423 } 1424 1425 return (err); 1426 } 1427 1428 /** 1429 * ice_if_msix_intr_assign - Assign MSI-X interrupt vectors to queues 1430 * @ctx: the iflib context structure 1431 * @msix: the number of vectors we were assigned 1432 * 1433 * Called by iflib to assign MSI-X vectors to queues. Currently requires that 1434 * we get at least the same number of vectors as we have queues, and that we 1435 * always have the same number of Tx and Rx queues. 1436 * 1437 * Tx queues use a softirq instead of using their own hardware interrupt. 1438 */ 1439 static int 1440 ice_if_msix_intr_assign(if_ctx_t ctx, int msix) 1441 { 1442 struct ice_softc *sc = (struct ice_softc *)iflib_get_softc(ctx); 1443 struct ice_vsi *vsi = &sc->pf_vsi; 1444 int err, i, vector; 1445 1446 ASSERT_CTX_LOCKED(sc); 1447 1448 if (vsi->num_rx_queues != vsi->num_tx_queues) { 1449 device_printf(sc->dev, 1450 "iflib requested %d Tx queues, and %d Rx queues, but the driver isn't able to support a differing number of Tx and Rx queues\n", 1451 vsi->num_tx_queues, vsi->num_rx_queues); 1452 return (EOPNOTSUPP); 1453 } 1454 1455 if (msix < (vsi->num_rx_queues + 1)) { 1456 device_printf(sc->dev, 1457 "Not enough MSI-X vectors to assign one vector to each queue pair\n"); 1458 return (EOPNOTSUPP); 1459 } 1460 1461 /* Save the number of vectors for future use */ 1462 sc->num_irq_vectors = vsi->num_rx_queues + 1; 1463 1464 /* Allocate space to store the IRQ vector data */ 1465 if (!(sc->irqvs = 1466 (struct ice_irq_vector *) malloc(sizeof(struct ice_irq_vector) * (sc->num_irq_vectors), 1467 M_ICE, M_NOWAIT))) { 1468 device_printf(sc->dev, 1469 "Unable to allocate irqv memory\n"); 1470 return (ENOMEM); 1471 } 1472 1473 /* Administrative interrupt events will use vector 0 */ 1474 err = iflib_irq_alloc_generic(ctx, &sc->irqvs[0].irq, 1, IFLIB_INTR_ADMIN, 1475 ice_msix_admin, sc, 0, "admin"); 1476 if (err) { 1477 device_printf(sc->dev, 1478 "Failed to register Admin queue handler: %s\n", 1479 ice_err_str(err)); 1480 goto free_irqvs; 1481 } 1482 sc->irqvs[0].me = 0; 1483 1484 /* Do not allocate queue interrupts when in recovery mode */ 1485 if (ice_test_state(&sc->state, ICE_STATE_RECOVERY_MODE)) 1486 return (0); 1487 1488 for (i = 0, vector = 1; i < vsi->num_rx_queues; i++, vector++) { 1489 struct ice_rx_queue *rxq = &vsi->rx_queues[i]; 1490 struct ice_tx_queue *txq = &vsi->tx_queues[i]; 1491 int rid = vector + 1; 1492 char irq_name[16]; 1493 1494 snprintf(irq_name, sizeof(irq_name), "rxq%d", i); 1495 err = iflib_irq_alloc_generic(ctx, &sc->irqvs[vector].irq, rid, 1496 IFLIB_INTR_RX, ice_msix_que, 1497 rxq, rxq->me, irq_name); 1498 if (err) { 1499 device_printf(sc->dev, 1500 "Failed to allocate q int %d err: %s\n", 1501 i, ice_err_str(err)); 1502 vector--; 1503 i--; 1504 goto fail; 1505 } 1506 sc->irqvs[vector].me = vector; 1507 rxq->irqv = &sc->irqvs[vector]; 1508 1509 bzero(irq_name, sizeof(irq_name)); 1510 1511 snprintf(irq_name, sizeof(irq_name), "txq%d", i); 1512 iflib_softirq_alloc_generic(ctx, &sc->irqvs[vector].irq, 1513 IFLIB_INTR_TX, txq, 1514 txq->me, irq_name); 1515 txq->irqv = &sc->irqvs[vector]; 1516 } 1517 1518 return (0); 1519 fail: 1520 for (; i >= 0; i--, vector--) 1521 iflib_irq_free(ctx, &sc->irqvs[vector].irq); 1522 iflib_irq_free(ctx, &sc->irqvs[0].irq); 1523 free_irqvs: 1524 free(sc->irqvs, M_ICE); 1525 sc->irqvs = NULL; 1526 return err; 1527 } 1528 1529 /** 1530 * ice_if_mtu_set - Set the device MTU 1531 * @ctx: iflib context structure 1532 * @mtu: the MTU requested 1533 * 1534 * Called by iflib to configure the device's Maximum Transmission Unit (MTU). 1535 * 1536 * @pre assumes the caller holds the iflib CTX lock 1537 */ 1538 static int 1539 ice_if_mtu_set(if_ctx_t ctx, uint32_t mtu) 1540 { 1541 struct ice_softc *sc = (struct ice_softc *)iflib_get_softc(ctx); 1542 1543 ASSERT_CTX_LOCKED(sc); 1544 1545 /* Do not support configuration when in recovery mode */ 1546 if (ice_test_state(&sc->state, ICE_STATE_RECOVERY_MODE)) 1547 return (ENOSYS); 1548 1549 if (mtu < ICE_MIN_MTU || mtu > ICE_MAX_MTU) 1550 return (EINVAL); 1551 1552 sc->scctx->isc_max_frame_size = mtu + 1553 ETHER_HDR_LEN + ETHER_CRC_LEN + ETHER_VLAN_ENCAP_LEN; 1554 1555 sc->pf_vsi.max_frame_size = sc->scctx->isc_max_frame_size; 1556 1557 return (0); 1558 } 1559 1560 /** 1561 * ice_if_intr_enable - Enable device interrupts 1562 * @ctx: iflib context structure 1563 * 1564 * Called by iflib to request enabling device interrupts. 1565 */ 1566 static void 1567 ice_if_intr_enable(if_ctx_t ctx) 1568 { 1569 struct ice_softc *sc = (struct ice_softc *)iflib_get_softc(ctx); 1570 struct ice_vsi *vsi = &sc->pf_vsi; 1571 struct ice_hw *hw = &sc->hw; 1572 1573 ASSERT_CTX_LOCKED(sc); 1574 1575 /* Enable ITR 0 */ 1576 ice_enable_intr(hw, sc->irqvs[0].me); 1577 1578 /* Do not enable queue interrupts in recovery mode */ 1579 if (ice_test_state(&sc->state, ICE_STATE_RECOVERY_MODE)) 1580 return; 1581 1582 /* Enable all queue interrupts */ 1583 for (int i = 0; i < vsi->num_rx_queues; i++) 1584 ice_enable_intr(hw, vsi->rx_queues[i].irqv->me); 1585 } 1586 1587 /** 1588 * ice_if_intr_disable - Disable device interrupts 1589 * @ctx: iflib context structure 1590 * 1591 * Called by iflib to request disabling device interrupts. 1592 */ 1593 static void 1594 ice_if_intr_disable(if_ctx_t ctx) 1595 { 1596 struct ice_softc *sc = (struct ice_softc *)iflib_get_softc(ctx); 1597 struct ice_hw *hw = &sc->hw; 1598 unsigned int i; 1599 1600 ASSERT_CTX_LOCKED(sc); 1601 1602 /* IFDI_INTR_DISABLE may be called prior to interrupts actually being 1603 * assigned to queues. Instead of assuming that the interrupt 1604 * assignment in the rx_queues structure is valid, just disable all 1605 * possible interrupts 1606 * 1607 * Note that we choose not to disable ITR 0 because this handles the 1608 * AdminQ interrupts, and we want to keep processing these even when 1609 * the interface is offline. 1610 */ 1611 for (i = 1; i < hw->func_caps.common_cap.num_msix_vectors; i++) 1612 ice_disable_intr(hw, i); 1613 } 1614 1615 /** 1616 * ice_if_rx_queue_intr_enable - Enable a specific Rx queue interrupt 1617 * @ctx: iflib context structure 1618 * @rxqid: the Rx queue to enable 1619 * 1620 * Enable a specific Rx queue interrupt. 1621 * 1622 * This function is not protected by the iflib CTX lock. 1623 */ 1624 static int 1625 ice_if_rx_queue_intr_enable(if_ctx_t ctx, uint16_t rxqid) 1626 { 1627 struct ice_softc *sc = (struct ice_softc *)iflib_get_softc(ctx); 1628 struct ice_vsi *vsi = &sc->pf_vsi; 1629 struct ice_hw *hw = &sc->hw; 1630 1631 /* Do not enable queue interrupts in recovery mode */ 1632 if (ice_test_state(&sc->state, ICE_STATE_RECOVERY_MODE)) 1633 return (ENOSYS); 1634 1635 ice_enable_intr(hw, vsi->rx_queues[rxqid].irqv->me); 1636 return (0); 1637 } 1638 1639 /** 1640 * ice_if_tx_queue_intr_enable - Enable a specific Tx queue interrupt 1641 * @ctx: iflib context structure 1642 * @txqid: the Tx queue to enable 1643 * 1644 * Enable a specific Tx queue interrupt. 1645 * 1646 * This function is not protected by the iflib CTX lock. 1647 */ 1648 static int 1649 ice_if_tx_queue_intr_enable(if_ctx_t ctx, uint16_t txqid) 1650 { 1651 struct ice_softc *sc = (struct ice_softc *)iflib_get_softc(ctx); 1652 struct ice_vsi *vsi = &sc->pf_vsi; 1653 struct ice_hw *hw = &sc->hw; 1654 1655 /* Do not enable queue interrupts in recovery mode */ 1656 if (ice_test_state(&sc->state, ICE_STATE_RECOVERY_MODE)) 1657 return (ENOSYS); 1658 1659 ice_enable_intr(hw, vsi->tx_queues[txqid].irqv->me); 1660 return (0); 1661 } 1662 1663 /** 1664 * ice_if_promisc_set - Set device promiscuous mode 1665 * @ctx: iflib context structure 1666 * @flags: promiscuous flags to configure 1667 * 1668 * Called by iflib to configure device promiscuous mode. 1669 * 1670 * @remark Calls to this function will always overwrite the previous setting 1671 */ 1672 static int 1673 ice_if_promisc_set(if_ctx_t ctx, int flags) 1674 { 1675 struct ice_softc *sc = (struct ice_softc *)iflib_get_softc(ctx); 1676 struct ice_hw *hw = &sc->hw; 1677 device_t dev = sc->dev; 1678 enum ice_status status; 1679 bool promisc_enable = flags & IFF_PROMISC; 1680 bool multi_enable = flags & IFF_ALLMULTI; 1681 1682 /* Do not support configuration when in recovery mode */ 1683 if (ice_test_state(&sc->state, ICE_STATE_RECOVERY_MODE)) 1684 return (ENOSYS); 1685 1686 if (multi_enable) 1687 return (EOPNOTSUPP); 1688 1689 if (promisc_enable) { 1690 status = ice_set_vsi_promisc(hw, sc->pf_vsi.idx, 1691 ICE_VSI_PROMISC_MASK, 0); 1692 if (status && status != ICE_ERR_ALREADY_EXISTS) { 1693 device_printf(dev, 1694 "Failed to enable promiscuous mode for PF VSI, err %s aq_err %s\n", 1695 ice_status_str(status), 1696 ice_aq_str(hw->adminq.sq_last_status)); 1697 return (EIO); 1698 } 1699 } else { 1700 status = ice_clear_vsi_promisc(hw, sc->pf_vsi.idx, 1701 ICE_VSI_PROMISC_MASK, 0); 1702 if (status) { 1703 device_printf(dev, 1704 "Failed to disable promiscuous mode for PF VSI, err %s aq_err %s\n", 1705 ice_status_str(status), 1706 ice_aq_str(hw->adminq.sq_last_status)); 1707 return (EIO); 1708 } 1709 } 1710 1711 return (0); 1712 } 1713 1714 /** 1715 * ice_if_media_change - Change device media 1716 * @ctx: device ctx structure 1717 * 1718 * Called by iflib when a media change is requested. This operation is not 1719 * supported by the hardware, so we just return an error code. 1720 */ 1721 static int 1722 ice_if_media_change(if_ctx_t ctx) 1723 { 1724 struct ice_softc *sc = (struct ice_softc *)iflib_get_softc(ctx); 1725 1726 device_printf(sc->dev, "Media change is not supported.\n"); 1727 return (ENODEV); 1728 } 1729 1730 /** 1731 * ice_if_media_status - Report current device media 1732 * @ctx: iflib context structure 1733 * @ifmr: ifmedia request structure to update 1734 * 1735 * Updates the provided ifmr with current device media status, including link 1736 * status and media type. 1737 */ 1738 static void 1739 ice_if_media_status(if_ctx_t ctx, struct ifmediareq *ifmr) 1740 { 1741 struct ice_softc *sc = (struct ice_softc *)iflib_get_softc(ctx); 1742 struct ice_link_status *li = &sc->hw.port_info->phy.link_info; 1743 1744 ifmr->ifm_status = IFM_AVALID; 1745 ifmr->ifm_active = IFM_ETHER; 1746 1747 /* Never report link up or media types when in recovery mode */ 1748 if (ice_test_state(&sc->state, ICE_STATE_RECOVERY_MODE)) 1749 return; 1750 1751 if (!sc->link_up) 1752 return; 1753 1754 ifmr->ifm_status |= IFM_ACTIVE; 1755 ifmr->ifm_active |= IFM_FDX; 1756 1757 if (li->phy_type_low) 1758 ifmr->ifm_active |= ice_get_phy_type_low(li->phy_type_low); 1759 else if (li->phy_type_high) 1760 ifmr->ifm_active |= ice_get_phy_type_high(li->phy_type_high); 1761 else 1762 ifmr->ifm_active |= IFM_UNKNOWN; 1763 1764 /* Report flow control status as well */ 1765 if (li->an_info & ICE_AQ_LINK_PAUSE_TX) 1766 ifmr->ifm_active |= IFM_ETH_TXPAUSE; 1767 if (li->an_info & ICE_AQ_LINK_PAUSE_RX) 1768 ifmr->ifm_active |= IFM_ETH_RXPAUSE; 1769 } 1770 1771 /** 1772 * ice_init_tx_tracking - Initialize Tx queue software tracking values 1773 * @vsi: the VSI to initialize 1774 * 1775 * Initialize Tx queue software tracking values, including the Report Status 1776 * queue, and related software tracking values. 1777 */ 1778 static void 1779 ice_init_tx_tracking(struct ice_vsi *vsi) 1780 { 1781 struct ice_tx_queue *txq; 1782 size_t j; 1783 int i; 1784 1785 for (i = 0, txq = vsi->tx_queues; i < vsi->num_tx_queues; i++, txq++) { 1786 1787 txq->tx_rs_cidx = txq->tx_rs_pidx = 0; 1788 1789 /* Initialize the last processed descriptor to be the end of 1790 * the ring, rather than the start, so that we avoid an 1791 * off-by-one error in ice_ift_txd_credits_update for the 1792 * first packet. 1793 */ 1794 txq->tx_cidx_processed = txq->desc_count - 1; 1795 1796 for (j = 0; j < txq->desc_count; j++) 1797 txq->tx_rsq[j] = QIDX_INVALID; 1798 } 1799 } 1800 1801 /** 1802 * ice_update_rx_mbuf_sz - Update the Rx buffer size for all queues 1803 * @sc: the device softc 1804 * 1805 * Called to update the Rx queue mbuf_sz parameter for configuring the receive 1806 * buffer sizes when programming hardware. 1807 */ 1808 static void 1809 ice_update_rx_mbuf_sz(struct ice_softc *sc) 1810 { 1811 uint32_t mbuf_sz = iflib_get_rx_mbuf_sz(sc->ctx); 1812 struct ice_vsi *vsi = &sc->pf_vsi; 1813 1814 MPASS(mbuf_sz <= UINT16_MAX); 1815 vsi->mbuf_sz = mbuf_sz; 1816 } 1817 1818 /** 1819 * ice_if_init - Initialize the device 1820 * @ctx: iflib ctx structure 1821 * 1822 * Called by iflib to bring the device up, i.e. ifconfig ice0 up. Initializes 1823 * device filters and prepares the Tx and Rx engines. 1824 * 1825 * @pre assumes the caller holds the iflib CTX lock 1826 */ 1827 static void 1828 ice_if_init(if_ctx_t ctx) 1829 { 1830 struct ice_softc *sc = (struct ice_softc *)iflib_get_softc(ctx); 1831 device_t dev = sc->dev; 1832 int err; 1833 1834 ASSERT_CTX_LOCKED(sc); 1835 1836 /* 1837 * We've seen an issue with 11.3/12.1 where sideband routines are 1838 * called after detach is called. This would call routines after 1839 * if_stop, causing issues with the teardown process. This has 1840 * seemingly been fixed in STABLE snapshots, but it seems like a 1841 * good idea to have this guard here regardless. 1842 */ 1843 if (ice_driver_is_detaching(sc)) 1844 return; 1845 1846 if (ice_test_state(&sc->state, ICE_STATE_RECOVERY_MODE)) 1847 return; 1848 1849 if (ice_test_state(&sc->state, ICE_STATE_RESET_FAILED)) { 1850 device_printf(sc->dev, "request to start interface cannot be completed as the device failed to reset\n"); 1851 return; 1852 } 1853 1854 if (ice_test_state(&sc->state, ICE_STATE_PREPARED_FOR_RESET)) { 1855 device_printf(sc->dev, "request to start interface while device is prepared for impending reset\n"); 1856 return; 1857 } 1858 1859 ice_update_rx_mbuf_sz(sc); 1860 1861 /* Update the MAC address... User might use a LAA */ 1862 err = ice_update_laa_mac(sc); 1863 if (err) { 1864 device_printf(dev, 1865 "LAA address change failed, err %s\n", 1866 ice_err_str(err)); 1867 return; 1868 } 1869 1870 /* Initialize software Tx tracking values */ 1871 ice_init_tx_tracking(&sc->pf_vsi); 1872 1873 err = ice_cfg_vsi_for_tx(&sc->pf_vsi); 1874 if (err) { 1875 device_printf(dev, 1876 "Unable to configure the main VSI for Tx: %s\n", 1877 ice_err_str(err)); 1878 return; 1879 } 1880 1881 err = ice_cfg_vsi_for_rx(&sc->pf_vsi); 1882 if (err) { 1883 device_printf(dev, 1884 "Unable to configure the main VSI for Rx: %s\n", 1885 ice_err_str(err)); 1886 goto err_cleanup_tx; 1887 } 1888 1889 err = ice_control_rx_queues(&sc->pf_vsi, true); 1890 if (err) { 1891 device_printf(dev, 1892 "Unable to enable Rx rings for transmit: %s\n", 1893 ice_err_str(err)); 1894 goto err_cleanup_tx; 1895 } 1896 1897 err = ice_cfg_pf_default_mac_filters(sc); 1898 if (err) { 1899 device_printf(dev, 1900 "Unable to configure default MAC filters: %s\n", 1901 ice_err_str(err)); 1902 goto err_stop_rx; 1903 } 1904 1905 /* We use software interrupts for Tx, so we only program the hardware 1906 * interrupts for Rx. 1907 */ 1908 ice_configure_rxq_interrupts(&sc->pf_vsi); 1909 ice_configure_rx_itr(&sc->pf_vsi); 1910 1911 /* Configure promiscuous mode */ 1912 ice_if_promisc_set(ctx, if_getflags(sc->ifp)); 1913 1914 ice_set_state(&sc->state, ICE_STATE_DRIVER_INITIALIZED); 1915 return; 1916 1917 err_stop_rx: 1918 ice_control_rx_queues(&sc->pf_vsi, false); 1919 err_cleanup_tx: 1920 ice_vsi_disable_tx(&sc->pf_vsi); 1921 } 1922 1923 /** 1924 * ice_poll_for_media_avail - Re-enable link if media is detected 1925 * @sc: device private structure 1926 * 1927 * Intended to be called from the driver's timer function, this function 1928 * sends the Get Link Status AQ command and re-enables HW link if the 1929 * command says that media is available. 1930 * 1931 * If the driver doesn't have the "NO_MEDIA" state set, then this does nothing, 1932 * since media removal events are supposed to be sent to the driver through 1933 * a link status event. 1934 */ 1935 static void 1936 ice_poll_for_media_avail(struct ice_softc *sc) 1937 { 1938 struct ice_hw *hw = &sc->hw; 1939 struct ice_port_info *pi = hw->port_info; 1940 1941 if (ice_test_state(&sc->state, ICE_STATE_NO_MEDIA)) { 1942 pi->phy.get_link_info = true; 1943 ice_get_link_status(pi, &sc->link_up); 1944 1945 if (pi->phy.link_info.link_info & ICE_AQ_MEDIA_AVAILABLE) { 1946 enum ice_status status; 1947 1948 /* Re-enable link and re-apply user link settings */ 1949 ice_apply_saved_phy_cfg(sc); 1950 1951 /* Update the OS about changes in media capability */ 1952 status = ice_add_media_types(sc, sc->media); 1953 if (status) 1954 device_printf(sc->dev, "Error adding device media types: %s aq_err %s\n", 1955 ice_status_str(status), 1956 ice_aq_str(hw->adminq.sq_last_status)); 1957 1958 ice_clear_state(&sc->state, ICE_STATE_NO_MEDIA); 1959 } 1960 } 1961 } 1962 1963 /** 1964 * ice_if_timer - called by iflib periodically 1965 * @ctx: iflib ctx structure 1966 * @qid: the queue this timer was called for 1967 * 1968 * This callback is triggered by iflib periodically. We use it to update the 1969 * hw statistics. 1970 * 1971 * @remark this function is not protected by the iflib CTX lock. 1972 */ 1973 static void 1974 ice_if_timer(if_ctx_t ctx, uint16_t qid) 1975 { 1976 struct ice_softc *sc = (struct ice_softc *)iflib_get_softc(ctx); 1977 uint64_t prev_link_xoff_rx = sc->stats.cur.link_xoff_rx; 1978 1979 if (qid != 0) 1980 return; 1981 1982 /* Do not attempt to update stats when in recovery mode */ 1983 if (ice_test_state(&sc->state, ICE_STATE_RECOVERY_MODE)) 1984 return; 1985 1986 /* Update device statistics */ 1987 ice_update_pf_stats(sc); 1988 1989 /* 1990 * For proper watchdog management, the iflib stack needs to know if 1991 * we've been paused during the last interval. Check if the 1992 * link_xoff_rx stat changed, and set the isc_pause_frames, if so. 1993 */ 1994 if (sc->stats.cur.link_xoff_rx != prev_link_xoff_rx) 1995 sc->scctx->isc_pause_frames = 1; 1996 1997 /* Update the primary VSI stats */ 1998 ice_update_vsi_hw_stats(&sc->pf_vsi); 1999 } 2000 2001 /** 2002 * ice_admin_timer - called periodically to trigger the admin task 2003 * @arg: callout(9) argument pointing to the device private softc structure 2004 * 2005 * Timer function used as part of a callout(9) timer that will periodically 2006 * trigger the admin task, even when the interface is down. 2007 * 2008 * @remark this function is not called by iflib and is not protected by the 2009 * iflib CTX lock. 2010 * 2011 * @remark because this is a callout function, it cannot sleep and should not 2012 * attempt taking the iflib CTX lock. 2013 */ 2014 static void 2015 ice_admin_timer(void *arg) 2016 { 2017 struct ice_softc *sc = (struct ice_softc *)arg; 2018 2019 /* Fire off the admin task */ 2020 iflib_admin_intr_deferred(sc->ctx); 2021 2022 /* Reschedule the admin timer */ 2023 callout_schedule(&sc->admin_timer, hz/2); 2024 } 2025 2026 /** 2027 * ice_transition_recovery_mode - Transition to recovery mode 2028 * @sc: the device private softc 2029 * 2030 * Called when the driver detects that the firmware has entered recovery mode 2031 * at run time. 2032 */ 2033 static void 2034 ice_transition_recovery_mode(struct ice_softc *sc) 2035 { 2036 struct ice_vsi *vsi = &sc->pf_vsi; 2037 int i; 2038 2039 device_printf(sc->dev, "Firmware recovery mode detected. Limiting functionality. Refer to Intel(R) Ethernet Adapters and Devices User Guide for details on firmware recovery mode.\n"); 2040 2041 /* Tell the stack that the link has gone down */ 2042 iflib_link_state_change(sc->ctx, LINK_STATE_DOWN, 0); 2043 2044 /* Request that the device be re-initialized */ 2045 ice_request_stack_reinit(sc); 2046 2047 ice_clear_bit(ICE_FEATURE_SRIOV, sc->feat_en); 2048 ice_clear_bit(ICE_FEATURE_SRIOV, sc->feat_cap); 2049 2050 ice_vsi_del_txqs_ctx(vsi); 2051 ice_vsi_del_rxqs_ctx(vsi); 2052 2053 for (i = 0; i < sc->num_available_vsi; i++) { 2054 if (sc->all_vsi[i]) 2055 ice_release_vsi(sc->all_vsi[i]); 2056 } 2057 sc->num_available_vsi = 0; 2058 2059 if (sc->all_vsi) { 2060 free(sc->all_vsi, M_ICE); 2061 sc->all_vsi = NULL; 2062 } 2063 2064 /* Destroy the interrupt manager */ 2065 ice_resmgr_destroy(&sc->imgr); 2066 /* Destroy the queue managers */ 2067 ice_resmgr_destroy(&sc->tx_qmgr); 2068 ice_resmgr_destroy(&sc->rx_qmgr); 2069 2070 ice_deinit_hw(&sc->hw); 2071 } 2072 2073 /** 2074 * ice_transition_safe_mode - Transition to safe mode 2075 * @sc: the device private softc 2076 * 2077 * Called when the driver attempts to reload the DDP package during a device 2078 * reset, and the new download fails. If so, we must transition to safe mode 2079 * at run time. 2080 * 2081 * @remark although safe mode normally allocates only a single queue, we can't 2082 * change the number of queues dynamically when using iflib. Due to this, we 2083 * do not attempt to reduce the number of queues. 2084 */ 2085 static void 2086 ice_transition_safe_mode(struct ice_softc *sc) 2087 { 2088 /* Indicate that we are in Safe mode */ 2089 ice_set_bit(ICE_FEATURE_SAFE_MODE, sc->feat_cap); 2090 ice_set_bit(ICE_FEATURE_SAFE_MODE, sc->feat_en); 2091 2092 ice_clear_bit(ICE_FEATURE_SRIOV, sc->feat_en); 2093 ice_clear_bit(ICE_FEATURE_SRIOV, sc->feat_cap); 2094 2095 ice_clear_bit(ICE_FEATURE_RSS, sc->feat_cap); 2096 ice_clear_bit(ICE_FEATURE_RSS, sc->feat_en); 2097 } 2098 2099 /** 2100 * ice_if_update_admin_status - update admin status 2101 * @ctx: iflib ctx structure 2102 * 2103 * Called by iflib to update the admin status. For our purposes, this means 2104 * check the adminq, and update the link status. It's ultimately triggered by 2105 * our admin interrupt, or by the ice_if_timer periodically. 2106 * 2107 * @pre assumes the caller holds the iflib CTX lock 2108 */ 2109 static void 2110 ice_if_update_admin_status(if_ctx_t ctx) 2111 { 2112 struct ice_softc *sc = (struct ice_softc *)iflib_get_softc(ctx); 2113 enum ice_fw_modes fw_mode; 2114 bool reschedule = false; 2115 u16 pending = 0; 2116 2117 ASSERT_CTX_LOCKED(sc); 2118 2119 /* Check if the firmware entered recovery mode at run time */ 2120 fw_mode = ice_get_fw_mode(&sc->hw); 2121 if (fw_mode == ICE_FW_MODE_REC) { 2122 if (!ice_testandset_state(&sc->state, ICE_STATE_RECOVERY_MODE)) { 2123 /* If we just entered recovery mode, log a warning to 2124 * the system administrator and deinit driver state 2125 * that is no longer functional. 2126 */ 2127 ice_transition_recovery_mode(sc); 2128 } 2129 } else if (fw_mode == ICE_FW_MODE_ROLLBACK) { 2130 if (!ice_testandset_state(&sc->state, ICE_STATE_ROLLBACK_MODE)) { 2131 /* Rollback mode isn't fatal, but we don't want to 2132 * repeatedly post a message about it. 2133 */ 2134 ice_print_rollback_msg(&sc->hw); 2135 } 2136 } 2137 2138 /* Handle global reset events */ 2139 ice_handle_reset_event(sc); 2140 2141 /* Handle PF reset requests */ 2142 ice_handle_pf_reset_request(sc); 2143 2144 /* Handle MDD events */ 2145 ice_handle_mdd_event(sc); 2146 2147 if (ice_test_state(&sc->state, ICE_STATE_RESET_FAILED) || 2148 ice_test_state(&sc->state, ICE_STATE_PREPARED_FOR_RESET) || 2149 ice_test_state(&sc->state, ICE_STATE_RECOVERY_MODE)) { 2150 /* 2151 * If we know the control queues are disabled, skip processing 2152 * the control queues entirely. 2153 */ 2154 ; 2155 } else if (ice_testandclear_state(&sc->state, ICE_STATE_CONTROLQ_EVENT_PENDING)) { 2156 ice_process_ctrlq(sc, ICE_CTL_Q_ADMIN, &pending); 2157 if (pending > 0) 2158 reschedule = true; 2159 2160 ice_process_ctrlq(sc, ICE_CTL_Q_MAILBOX, &pending); 2161 if (pending > 0) 2162 reschedule = true; 2163 } 2164 2165 /* Poll for link up */ 2166 ice_poll_for_media_avail(sc); 2167 2168 /* Check and update link status */ 2169 ice_update_link_status(sc, false); 2170 2171 /* 2172 * If there are still messages to process, we need to reschedule 2173 * ourselves. Otherwise, we can just re-enable the interrupt. We'll be 2174 * woken up at the next interrupt or timer event. 2175 */ 2176 if (reschedule) { 2177 ice_set_state(&sc->state, ICE_STATE_CONTROLQ_EVENT_PENDING); 2178 iflib_admin_intr_deferred(ctx); 2179 } else { 2180 ice_enable_intr(&sc->hw, sc->irqvs[0].me); 2181 } 2182 } 2183 2184 /** 2185 * ice_prepare_for_reset - Prepare device for an impending reset 2186 * @sc: The device private softc 2187 * 2188 * Prepare the driver for an impending reset, shutting down VSIs, clearing the 2189 * scheduler setup, and shutting down controlqs. Uses the 2190 * ICE_STATE_PREPARED_FOR_RESET to indicate whether we've already prepared the 2191 * driver for reset or not. 2192 */ 2193 static void 2194 ice_prepare_for_reset(struct ice_softc *sc) 2195 { 2196 struct ice_hw *hw = &sc->hw; 2197 2198 /* If we're already prepared, there's nothing to do */ 2199 if (ice_testandset_state(&sc->state, ICE_STATE_PREPARED_FOR_RESET)) 2200 return; 2201 2202 log(LOG_INFO, "%s: preparing to reset device logic\n", sc->ifp->if_xname); 2203 2204 /* In recovery mode, hardware is not initialized */ 2205 if (ice_test_state(&sc->state, ICE_STATE_RECOVERY_MODE)) 2206 return; 2207 2208 /* Release the main PF VSI queue mappings */ 2209 ice_resmgr_release_map(&sc->tx_qmgr, sc->pf_vsi.tx_qmap, 2210 sc->pf_vsi.num_tx_queues); 2211 ice_resmgr_release_map(&sc->rx_qmgr, sc->pf_vsi.rx_qmap, 2212 sc->pf_vsi.num_rx_queues); 2213 2214 ice_clear_hw_tbls(hw); 2215 2216 if (hw->port_info) 2217 ice_sched_clear_port(hw->port_info); 2218 2219 ice_shutdown_all_ctrlq(hw); 2220 } 2221 2222 /** 2223 * ice_rebuild_pf_vsi_qmap - Rebuild the main PF VSI queue mapping 2224 * @sc: the device softc pointer 2225 * 2226 * Loops over the Tx and Rx queues for the main PF VSI and reassigns the queue 2227 * mapping after a reset occurred. 2228 */ 2229 static int 2230 ice_rebuild_pf_vsi_qmap(struct ice_softc *sc) 2231 { 2232 struct ice_vsi *vsi = &sc->pf_vsi; 2233 struct ice_tx_queue *txq; 2234 struct ice_rx_queue *rxq; 2235 int err, i; 2236 2237 /* Re-assign Tx queues from PF space to the main VSI */ 2238 err = ice_resmgr_assign_contiguous(&sc->tx_qmgr, vsi->tx_qmap, 2239 vsi->num_tx_queues); 2240 if (err) { 2241 device_printf(sc->dev, "Unable to re-assign PF Tx queues: %s\n", 2242 ice_err_str(err)); 2243 return (err); 2244 } 2245 2246 /* Re-assign Rx queues from PF space to this VSI */ 2247 err = ice_resmgr_assign_contiguous(&sc->rx_qmgr, vsi->rx_qmap, 2248 vsi->num_rx_queues); 2249 if (err) { 2250 device_printf(sc->dev, "Unable to re-assign PF Rx queues: %s\n", 2251 ice_err_str(err)); 2252 goto err_release_tx_queues; 2253 } 2254 2255 vsi->qmap_type = ICE_RESMGR_ALLOC_CONTIGUOUS; 2256 2257 /* Re-assign Tx queue tail pointers */ 2258 for (i = 0, txq = vsi->tx_queues; i < vsi->num_tx_queues; i++, txq++) 2259 txq->tail = QTX_COMM_DBELL(vsi->tx_qmap[i]); 2260 2261 /* Re-assign Rx queue tail pointers */ 2262 for (i = 0, rxq = vsi->rx_queues; i < vsi->num_rx_queues; i++, rxq++) 2263 rxq->tail = QRX_TAIL(vsi->rx_qmap[i]); 2264 2265 return (0); 2266 2267 err_release_tx_queues: 2268 ice_resmgr_release_map(&sc->tx_qmgr, sc->pf_vsi.tx_qmap, 2269 sc->pf_vsi.num_tx_queues); 2270 2271 return (err); 2272 } 2273 2274 /* determine if the iflib context is active */ 2275 #define CTX_ACTIVE(ctx) ((if_getdrvflags(iflib_get_ifp(ctx)) & IFF_DRV_RUNNING)) 2276 2277 /** 2278 * ice_rebuild_recovery_mode - Rebuild driver state while in recovery mode 2279 * @sc: The device private softc 2280 * 2281 * Handle a driver rebuild while in recovery mode. This will only rebuild the 2282 * limited functionality supported while in recovery mode. 2283 */ 2284 static void 2285 ice_rebuild_recovery_mode(struct ice_softc *sc) 2286 { 2287 device_t dev = sc->dev; 2288 2289 /* enable PCIe bus master */ 2290 pci_enable_busmaster(dev); 2291 2292 /* Configure interrupt causes for the administrative interrupt */ 2293 ice_configure_misc_interrupts(sc); 2294 2295 /* Enable ITR 0 right away, so that we can handle admin interrupts */ 2296 ice_enable_intr(&sc->hw, sc->irqvs[0].me); 2297 2298 /* Now that the rebuild is finished, we're no longer prepared to reset */ 2299 ice_clear_state(&sc->state, ICE_STATE_PREPARED_FOR_RESET); 2300 2301 log(LOG_INFO, "%s: device rebuild successful\n", sc->ifp->if_xname); 2302 2303 /* In order to completely restore device functionality, the iflib core 2304 * needs to be reset. We need to request an iflib reset. Additionally, 2305 * because the state of IFC_DO_RESET is cached within task_fn_admin in 2306 * the iflib core, we also want re-run the admin task so that iflib 2307 * resets immediately instead of waiting for the next interrupt. 2308 */ 2309 ice_request_stack_reinit(sc); 2310 2311 return; 2312 } 2313 2314 /** 2315 * ice_rebuild - Rebuild driver state post reset 2316 * @sc: The device private softc 2317 * 2318 * Restore driver state after a reset occurred. Restart the controlqs, setup 2319 * the hardware port, and re-enable the VSIs. 2320 */ 2321 static void 2322 ice_rebuild(struct ice_softc *sc) 2323 { 2324 struct ice_hw *hw = &sc->hw; 2325 device_t dev = sc->dev; 2326 enum ice_status status; 2327 int err; 2328 2329 sc->rebuild_ticks = ticks; 2330 2331 /* If we're rebuilding, then a reset has succeeded. */ 2332 ice_clear_state(&sc->state, ICE_STATE_RESET_FAILED); 2333 2334 /* 2335 * If the firmware is in recovery mode, only restore the limited 2336 * functionality supported by recovery mode. 2337 */ 2338 if (ice_test_state(&sc->state, ICE_STATE_RECOVERY_MODE)) { 2339 ice_rebuild_recovery_mode(sc); 2340 return; 2341 } 2342 2343 /* enable PCIe bus master */ 2344 pci_enable_busmaster(dev); 2345 2346 status = ice_init_all_ctrlq(hw); 2347 if (status) { 2348 device_printf(dev, "failed to re-init controlqs, err %s\n", 2349 ice_status_str(status)); 2350 goto err_shutdown_ctrlq; 2351 } 2352 2353 /* Query the allocated resources for Tx scheduler */ 2354 status = ice_sched_query_res_alloc(hw); 2355 if (status) { 2356 device_printf(dev, 2357 "Failed to query scheduler resources, err %s aq_err %s\n", 2358 ice_status_str(status), 2359 ice_aq_str(hw->adminq.sq_last_status)); 2360 goto err_shutdown_ctrlq; 2361 } 2362 2363 err = ice_send_version(sc); 2364 if (err) 2365 goto err_shutdown_ctrlq; 2366 2367 err = ice_init_link_events(sc); 2368 if (err) { 2369 device_printf(dev, "ice_init_link_events failed: %s\n", 2370 ice_err_str(err)); 2371 goto err_shutdown_ctrlq; 2372 } 2373 2374 status = ice_clear_pf_cfg(hw); 2375 if (status) { 2376 device_printf(dev, "failed to clear PF configuration, err %s\n", 2377 ice_status_str(status)); 2378 goto err_shutdown_ctrlq; 2379 } 2380 2381 ice_clear_pxe_mode(hw); 2382 2383 status = ice_get_caps(hw); 2384 if (status) { 2385 device_printf(dev, "failed to get capabilities, err %s\n", 2386 ice_status_str(status)); 2387 goto err_shutdown_ctrlq; 2388 } 2389 2390 status = ice_sched_init_port(hw->port_info); 2391 if (status) { 2392 device_printf(dev, "failed to initialize port, err %s\n", 2393 ice_status_str(status)); 2394 goto err_sched_cleanup; 2395 } 2396 2397 /* If we previously loaded the package, it needs to be reloaded now */ 2398 if (!ice_is_bit_set(sc->feat_en, ICE_FEATURE_SAFE_MODE)) { 2399 status = ice_init_pkg(hw, hw->pkg_copy, hw->pkg_size); 2400 if (status) { 2401 ice_log_pkg_init(sc, &status); 2402 2403 ice_transition_safe_mode(sc); 2404 } 2405 } 2406 2407 ice_reset_pf_stats(sc); 2408 2409 err = ice_rebuild_pf_vsi_qmap(sc); 2410 if (err) { 2411 device_printf(sc->dev, "Unable to re-assign main VSI queues, err %s\n", 2412 ice_err_str(err)); 2413 goto err_sched_cleanup; 2414 } 2415 err = ice_initialize_vsi(&sc->pf_vsi); 2416 if (err) { 2417 device_printf(sc->dev, "Unable to re-initialize Main VSI, err %s\n", 2418 ice_err_str(err)); 2419 goto err_release_queue_allocations; 2420 } 2421 2422 /* Replay all VSI configuration */ 2423 err = ice_replay_all_vsi_cfg(sc); 2424 if (err) 2425 goto err_deinit_pf_vsi; 2426 2427 /* Reconfigure the main PF VSI for RSS */ 2428 err = ice_config_rss(&sc->pf_vsi); 2429 if (err) { 2430 device_printf(sc->dev, 2431 "Unable to reconfigure RSS for the main VSI, err %s\n", 2432 ice_err_str(err)); 2433 goto err_deinit_pf_vsi; 2434 } 2435 2436 /* Refresh link status */ 2437 ice_clear_state(&sc->state, ICE_STATE_LINK_STATUS_REPORTED); 2438 sc->hw.port_info->phy.get_link_info = true; 2439 ice_get_link_status(sc->hw.port_info, &sc->link_up); 2440 ice_update_link_status(sc, true); 2441 2442 /* Configure interrupt causes for the administrative interrupt */ 2443 ice_configure_misc_interrupts(sc); 2444 2445 /* Enable ITR 0 right away, so that we can handle admin interrupts */ 2446 ice_enable_intr(&sc->hw, sc->irqvs[0].me); 2447 2448 /* Now that the rebuild is finished, we're no longer prepared to reset */ 2449 ice_clear_state(&sc->state, ICE_STATE_PREPARED_FOR_RESET); 2450 2451 log(LOG_INFO, "%s: device rebuild successful\n", sc->ifp->if_xname); 2452 2453 /* In order to completely restore device functionality, the iflib core 2454 * needs to be reset. We need to request an iflib reset. Additionally, 2455 * because the state of IFC_DO_RESET is cached within task_fn_admin in 2456 * the iflib core, we also want re-run the admin task so that iflib 2457 * resets immediately instead of waiting for the next interrupt. 2458 */ 2459 ice_request_stack_reinit(sc); 2460 2461 return; 2462 2463 err_deinit_pf_vsi: 2464 ice_deinit_vsi(&sc->pf_vsi); 2465 err_release_queue_allocations: 2466 ice_resmgr_release_map(&sc->tx_qmgr, sc->pf_vsi.tx_qmap, 2467 sc->pf_vsi.num_tx_queues); 2468 ice_resmgr_release_map(&sc->rx_qmgr, sc->pf_vsi.rx_qmap, 2469 sc->pf_vsi.num_rx_queues); 2470 err_sched_cleanup: 2471 ice_sched_cleanup_all(hw); 2472 err_shutdown_ctrlq: 2473 ice_shutdown_all_ctrlq(hw); 2474 ice_set_state(&sc->state, ICE_STATE_RESET_FAILED); 2475 device_printf(dev, "Driver rebuild failed, please reload the device driver\n"); 2476 } 2477 2478 /** 2479 * ice_handle_reset_event - Handle reset events triggered by OICR 2480 * @sc: The device private softc 2481 * 2482 * Handle reset events triggered by an OICR notification. This includes CORER, 2483 * GLOBR, and EMPR resets triggered by software on this or any other PF or by 2484 * firmware. 2485 * 2486 * @pre assumes the iflib context lock is held, and will unlock it while 2487 * waiting for the hardware to finish reset. 2488 */ 2489 static void 2490 ice_handle_reset_event(struct ice_softc *sc) 2491 { 2492 struct ice_hw *hw = &sc->hw; 2493 enum ice_status status; 2494 device_t dev = sc->dev; 2495 2496 /* When a CORER, GLOBR, or EMPR is about to happen, the hardware will 2497 * trigger an OICR interrupt. Our OICR handler will determine when 2498 * this occurs and set the ICE_STATE_RESET_OICR_RECV bit as 2499 * appropriate. 2500 */ 2501 if (!ice_testandclear_state(&sc->state, ICE_STATE_RESET_OICR_RECV)) 2502 return; 2503 2504 ice_prepare_for_reset(sc); 2505 2506 /* 2507 * Release the iflib context lock and wait for the device to finish 2508 * resetting. 2509 */ 2510 IFLIB_CTX_UNLOCK(sc); 2511 status = ice_check_reset(hw); 2512 IFLIB_CTX_LOCK(sc); 2513 if (status) { 2514 device_printf(dev, "Device never came out of reset, err %s\n", 2515 ice_status_str(status)); 2516 ice_set_state(&sc->state, ICE_STATE_RESET_FAILED); 2517 return; 2518 } 2519 2520 /* We're done with the reset, so we can rebuild driver state */ 2521 sc->hw.reset_ongoing = false; 2522 ice_rebuild(sc); 2523 2524 /* In the unlikely event that a PF reset request occurs at the same 2525 * time as a global reset, clear the request now. This avoids 2526 * resetting a second time right after we reset due to a global event. 2527 */ 2528 if (ice_testandclear_state(&sc->state, ICE_STATE_RESET_PFR_REQ)) 2529 device_printf(dev, "Ignoring PFR request that occurred while a reset was ongoing\n"); 2530 } 2531 2532 /** 2533 * ice_handle_pf_reset_request - Initiate PF reset requested by software 2534 * @sc: The device private softc 2535 * 2536 * Initiate a PF reset requested by software. We handle this in the admin task 2537 * so that only one thread actually handles driver preparation and cleanup, 2538 * rather than having multiple threads possibly attempt to run this code 2539 * simultaneously. 2540 * 2541 * @pre assumes the iflib context lock is held and will unlock it while 2542 * waiting for the PF reset to complete. 2543 */ 2544 static void 2545 ice_handle_pf_reset_request(struct ice_softc *sc) 2546 { 2547 struct ice_hw *hw = &sc->hw; 2548 enum ice_status status; 2549 2550 /* Check for PF reset requests */ 2551 if (!ice_testandclear_state(&sc->state, ICE_STATE_RESET_PFR_REQ)) 2552 return; 2553 2554 /* Make sure we're prepared for reset */ 2555 ice_prepare_for_reset(sc); 2556 2557 /* 2558 * Release the iflib context lock and wait for the device to finish 2559 * resetting. 2560 */ 2561 IFLIB_CTX_UNLOCK(sc); 2562 status = ice_reset(hw, ICE_RESET_PFR); 2563 IFLIB_CTX_LOCK(sc); 2564 if (status) { 2565 device_printf(sc->dev, "device PF reset failed, err %s\n", 2566 ice_status_str(status)); 2567 ice_set_state(&sc->state, ICE_STATE_RESET_FAILED); 2568 return; 2569 } 2570 2571 sc->soft_stats.pfr_count++; 2572 ice_rebuild(sc); 2573 } 2574 2575 /** 2576 * ice_init_device_features - Init device driver features 2577 * @sc: driver softc structure 2578 * 2579 * @pre assumes that the function capabilities bits have been set up by 2580 * ice_init_hw(). 2581 */ 2582 static void 2583 ice_init_device_features(struct ice_softc *sc) 2584 { 2585 /* 2586 * A failed pkg file download triggers safe mode, disabling advanced 2587 * device feature support 2588 */ 2589 if (ice_is_bit_set(sc->feat_en, ICE_FEATURE_SAFE_MODE)) 2590 return; 2591 2592 /* Set capabilities that all devices support */ 2593 ice_set_bit(ICE_FEATURE_SRIOV, sc->feat_cap); 2594 ice_set_bit(ICE_FEATURE_RSS, sc->feat_cap); 2595 ice_set_bit(ICE_FEATURE_LENIENT_LINK_MODE, sc->feat_cap); 2596 ice_set_bit(ICE_FEATURE_DEFAULT_OVERRIDE, sc->feat_cap); 2597 2598 /* Disable features due to hardware limitations... */ 2599 if (!sc->hw.func_caps.common_cap.rss_table_size) 2600 ice_clear_bit(ICE_FEATURE_RSS, sc->feat_cap); 2601 2602 /* Disable capabilities not supported by the OS */ 2603 ice_disable_unsupported_features(sc->feat_cap); 2604 2605 /* RSS is always enabled for iflib */ 2606 if (ice_is_bit_set(sc->feat_cap, ICE_FEATURE_RSS)) 2607 ice_set_bit(ICE_FEATURE_RSS, sc->feat_en); 2608 } 2609 2610 /** 2611 * ice_if_multi_set - Callback to update Multicast filters in HW 2612 * @ctx: iflib ctx structure 2613 * 2614 * Called by iflib in response to SIOCDELMULTI and SIOCADDMULTI. Must search 2615 * the if_multiaddrs list and determine which filters have been added or 2616 * removed from the list, and update HW programming to reflect the new list. 2617 * 2618 * @pre assumes the caller holds the iflib CTX lock 2619 */ 2620 static void 2621 ice_if_multi_set(if_ctx_t ctx) 2622 { 2623 struct ice_softc *sc = (struct ice_softc *)iflib_get_softc(ctx); 2624 int err; 2625 2626 ASSERT_CTX_LOCKED(sc); 2627 2628 /* Do not handle multicast configuration in recovery mode */ 2629 if (ice_test_state(&sc->state, ICE_STATE_RECOVERY_MODE)) 2630 return; 2631 2632 err = ice_sync_multicast_filters(sc); 2633 if (err) { 2634 device_printf(sc->dev, 2635 "Failed to synchronize multicast filter list: %s\n", 2636 ice_err_str(err)); 2637 return; 2638 } 2639 } 2640 2641 /** 2642 * ice_if_vlan_register - Register a VLAN with the hardware 2643 * @ctx: iflib ctx pointer 2644 * @vtag: VLAN to add 2645 * 2646 * Programs the main PF VSI with a hardware filter for the given VLAN. 2647 * 2648 * @pre assumes the caller holds the iflib CTX lock 2649 */ 2650 static void 2651 ice_if_vlan_register(if_ctx_t ctx, u16 vtag) 2652 { 2653 struct ice_softc *sc = (struct ice_softc *)iflib_get_softc(ctx); 2654 enum ice_status status; 2655 2656 ASSERT_CTX_LOCKED(sc); 2657 2658 /* Do not handle VLAN configuration in recovery mode */ 2659 if (ice_test_state(&sc->state, ICE_STATE_RECOVERY_MODE)) 2660 return; 2661 2662 status = ice_add_vlan_hw_filter(&sc->pf_vsi, vtag); 2663 if (status) { 2664 device_printf(sc->dev, 2665 "Failure adding VLAN %d to main VSI, err %s aq_err %s\n", 2666 vtag, ice_status_str(status), 2667 ice_aq_str(sc->hw.adminq.sq_last_status)); 2668 } 2669 } 2670 2671 /** 2672 * ice_if_vlan_unregister - Remove a VLAN filter from the hardware 2673 * @ctx: iflib ctx pointer 2674 * @vtag: VLAN to add 2675 * 2676 * Removes the previously programmed VLAN filter from the main PF VSI. 2677 * 2678 * @pre assumes the caller holds the iflib CTX lock 2679 */ 2680 static void 2681 ice_if_vlan_unregister(if_ctx_t ctx, u16 vtag) 2682 { 2683 struct ice_softc *sc = (struct ice_softc *)iflib_get_softc(ctx); 2684 enum ice_status status; 2685 2686 ASSERT_CTX_LOCKED(sc); 2687 2688 /* Do not handle VLAN configuration in recovery mode */ 2689 if (ice_test_state(&sc->state, ICE_STATE_RECOVERY_MODE)) 2690 return; 2691 2692 status = ice_remove_vlan_hw_filter(&sc->pf_vsi, vtag); 2693 if (status) { 2694 device_printf(sc->dev, 2695 "Failure removing VLAN %d from main VSI, err %s aq_err %s\n", 2696 vtag, ice_status_str(status), 2697 ice_aq_str(sc->hw.adminq.sq_last_status)); 2698 } 2699 } 2700 2701 /** 2702 * ice_if_stop - Stop the device 2703 * @ctx: iflib context structure 2704 * 2705 * Called by iflib to stop the device and bring it down. (i.e. ifconfig ice0 2706 * down) 2707 * 2708 * @pre assumes the caller holds the iflib CTX lock 2709 */ 2710 static void 2711 ice_if_stop(if_ctx_t ctx) 2712 { 2713 struct ice_softc *sc = (struct ice_softc *)iflib_get_softc(ctx); 2714 2715 ASSERT_CTX_LOCKED(sc); 2716 2717 /* 2718 * The iflib core may call IFDI_STOP prior to the first call to 2719 * IFDI_INIT. This will cause us to attempt to remove MAC filters we 2720 * don't have, and disable Tx queues which aren't yet configured. 2721 * Although it is likely these extra operations are harmless, they do 2722 * cause spurious warning messages to be displayed, which may confuse 2723 * users. 2724 * 2725 * To avoid these messages, we use a state bit indicating if we've 2726 * been initialized. It will be set when ice_if_init is called, and 2727 * cleared here in ice_if_stop. 2728 */ 2729 if (!ice_testandclear_state(&sc->state, ICE_STATE_DRIVER_INITIALIZED)) 2730 return; 2731 2732 if (ice_test_state(&sc->state, ICE_STATE_RESET_FAILED)) { 2733 device_printf(sc->dev, "request to stop interface cannot be completed as the device failed to reset\n"); 2734 return; 2735 } 2736 2737 if (ice_test_state(&sc->state, ICE_STATE_PREPARED_FOR_RESET)) { 2738 device_printf(sc->dev, "request to stop interface while device is prepared for impending reset\n"); 2739 return; 2740 } 2741 2742 /* Remove the MAC filters, stop Tx, and stop Rx. We don't check the 2743 * return of these functions because there's nothing we can really do 2744 * if they fail, and the functions already print error messages. 2745 * Just try to shut down as much as we can. 2746 */ 2747 ice_rm_pf_default_mac_filters(sc); 2748 2749 /* Dissociate the Tx and Rx queues from the interrupts */ 2750 ice_flush_txq_interrupts(&sc->pf_vsi); 2751 ice_flush_rxq_interrupts(&sc->pf_vsi); 2752 2753 /* Disable the Tx and Rx queues */ 2754 ice_vsi_disable_tx(&sc->pf_vsi); 2755 ice_control_rx_queues(&sc->pf_vsi, false); 2756 } 2757 2758 /** 2759 * ice_if_get_counter - Get current value of an ifnet statistic 2760 * @ctx: iflib context pointer 2761 * @counter: ifnet counter to read 2762 * 2763 * Reads the current value of an ifnet counter for the device. 2764 * 2765 * This function is not protected by the iflib CTX lock. 2766 */ 2767 static uint64_t 2768 ice_if_get_counter(if_ctx_t ctx, ift_counter counter) 2769 { 2770 struct ice_softc *sc = (struct ice_softc *)iflib_get_softc(ctx); 2771 2772 /* Return the counter for the main PF VSI */ 2773 return ice_get_ifnet_counter(&sc->pf_vsi, counter); 2774 } 2775 2776 /** 2777 * ice_request_stack_reinit - Request that iflib re-initialize 2778 * @sc: the device private softc 2779 * 2780 * Request that the device be brought down and up, to re-initialize. For 2781 * example, this may be called when a device reset occurs, or when Tx and Rx 2782 * queues need to be re-initialized. 2783 * 2784 * This is required because the iflib state is outside the driver, and must be 2785 * re-initialized if we need to resart Tx and Rx queues. 2786 */ 2787 void 2788 ice_request_stack_reinit(struct ice_softc *sc) 2789 { 2790 if (CTX_ACTIVE(sc->ctx)) { 2791 iflib_request_reset(sc->ctx); 2792 iflib_admin_intr_deferred(sc->ctx); 2793 } 2794 } 2795 2796 /** 2797 * ice_driver_is_detaching - Check if the driver is detaching/unloading 2798 * @sc: device private softc 2799 * 2800 * Returns true if the driver is detaching, false otherwise. 2801 * 2802 * @remark on newer kernels, take advantage of iflib_in_detach in order to 2803 * report detachment correctly as early as possible. 2804 * 2805 * @remark this function is used by various code paths that want to avoid 2806 * running if the driver is about to be removed. This includes sysctls and 2807 * other driver access points. Note that it does not fully resolve 2808 * detach-based race conditions as it is possible for a thread to race with 2809 * iflib_in_detach. 2810 */ 2811 bool 2812 ice_driver_is_detaching(struct ice_softc *sc) 2813 { 2814 return (ice_test_state(&sc->state, ICE_STATE_DETACHING) || 2815 iflib_in_detach(sc->ctx)); 2816 } 2817 2818 /** 2819 * ice_if_priv_ioctl - Device private ioctl handler 2820 * @ctx: iflib context pointer 2821 * @command: The ioctl command issued 2822 * @data: ioctl specific data 2823 * 2824 * iflib callback for handling custom driver specific ioctls. 2825 * 2826 * @pre Assumes that the iflib context lock is held. 2827 */ 2828 static int 2829 ice_if_priv_ioctl(if_ctx_t ctx, u_long command, caddr_t data) 2830 { 2831 struct ice_softc *sc = (struct ice_softc *)iflib_get_softc(ctx); 2832 struct ifdrv *ifd; 2833 device_t dev = sc->dev; 2834 2835 if (data == NULL) 2836 return (EINVAL); 2837 2838 ASSERT_CTX_LOCKED(sc); 2839 2840 /* Make sure the command type is valid */ 2841 switch (command) { 2842 case SIOCSDRVSPEC: 2843 case SIOCGDRVSPEC: 2844 /* Accepted commands */ 2845 break; 2846 case SIOCGPRIVATE_0: 2847 /* 2848 * Although we do not support this ioctl command, it's 2849 * expected that iflib will forward it to the IFDI_PRIV_IOCTL 2850 * handler. Do not print a message in this case 2851 */ 2852 return (ENOTSUP); 2853 default: 2854 /* 2855 * If we get a different command for this function, it's 2856 * definitely unexpected, so log a message indicating what 2857 * command we got for debugging purposes. 2858 */ 2859 device_printf(dev, "%s: unexpected ioctl command %08lx\n", 2860 __func__, command); 2861 return (EINVAL); 2862 } 2863 2864 ifd = (struct ifdrv *)data; 2865 2866 switch (ifd->ifd_cmd) { 2867 case ICE_NVM_ACCESS: 2868 return ice_handle_nvm_access_ioctl(sc, ifd); 2869 default: 2870 return EINVAL; 2871 } 2872 } 2873 2874 /** 2875 * ice_if_i2c_req - I2C request handler for iflib 2876 * @ctx: iflib context pointer 2877 * @req: The I2C parameters to use 2878 * 2879 * Read from the port's I2C eeprom using the parameters from the ioctl. 2880 * 2881 * @remark The iflib-only part is pretty simple. 2882 */ 2883 static int 2884 ice_if_i2c_req(if_ctx_t ctx, struct ifi2creq *req) 2885 { 2886 struct ice_softc *sc = (struct ice_softc *)iflib_get_softc(ctx); 2887 2888 return ice_handle_i2c_req(sc, req); 2889 } 2890 2891 /** 2892 * ice_if_suspend - PCI device suspend handler for iflib 2893 * @ctx: iflib context pointer 2894 * 2895 * Deinitializes the driver and clears HW resources in preparation for 2896 * suspend or an FLR. 2897 * 2898 * @returns 0; this return value is ignored 2899 */ 2900 static int 2901 ice_if_suspend(if_ctx_t ctx) 2902 { 2903 struct ice_softc *sc = (struct ice_softc *)iflib_get_softc(ctx); 2904 2905 /* At least a PFR is always going to happen after this; 2906 * either via FLR or during the D3->D0 transition. 2907 */ 2908 ice_clear_state(&sc->state, ICE_STATE_RESET_PFR_REQ); 2909 2910 ice_prepare_for_reset(sc); 2911 2912 return (0); 2913 } 2914 2915 /** 2916 * ice_if_resume - PCI device resume handler for iflib 2917 * @ctx: iflib context pointer 2918 * 2919 * Reinitializes the driver and the HW after PCI resume or after 2920 * an FLR. An init is performed by iflib after this function is finished. 2921 * 2922 * @returns 0; this return value is ignored 2923 */ 2924 static int 2925 ice_if_resume(if_ctx_t ctx) 2926 { 2927 struct ice_softc *sc = (struct ice_softc *)iflib_get_softc(ctx); 2928 2929 ice_rebuild(sc); 2930 2931 return (0); 2932 } 2933 2934