FILE
ice_sriov.c
File preview with syntax highlighting.
linux-6.19.14/drivers/net/ethernet/intel/ice/ice_sriov.c
1// SPDX-License-Identifier: GPL-2.0 2/* Copyright (c) 2018, Intel Corporation. */ 3 4#include "ice.h" 5#include "ice_vf_lib_private.h" 6#include "ice_base.h" 7#include "ice_lib.h" 8#include "ice_fltr.h" 9#include "ice_dcb_lib.h" 10#include "ice_flow.h" 11#include "ice_eswitch.h" 12#include "virt/allowlist.h" 13#include "ice_flex_pipe.h" 14#include "ice_vf_vsi_vlan_ops.h" 15#include "ice_vlan.h" 16 17/** 18 * ice_free_vf_entries - Free all VF entries from the hash table 19 * @pf: pointer to the PF structure 20 * 21 * Iterate over the VF hash table, removing and releasing all VF entries. 22 * Called during VF teardown or as cleanup during failed VF initialization. 23 */ 24static void ice_free_vf_entries(struct ice_pf *pf) 25{ 26 struct ice_vfs *vfs = &pf->vfs; 27 struct hlist_node *tmp; 28 struct ice_vf *vf; 29 unsigned int bkt; 30 31 /* Remove all VFs from the hash table and release their main 32 * reference. Once all references to the VF are dropped, ice_put_vf() 33 * will call ice_release_vf which will remove the VF memory. 34 */ 35 lockdep_assert_held(&vfs->table_lock); 36 37 hash_for_each_safe(vfs->table, bkt, tmp, vf, entry) { 38 hash_del_rcu(&vf->entry); 39 ice_deinitialize_vf_entry(vf); 40 ice_put_vf(vf); 41 } 42} 43 44/** 45 * ice_free_vf_res - Free a VF's resources 46 * @vf: pointer to the VF info 47 */ 48static void ice_free_vf_res(struct ice_vf *vf) 49{ 50 struct ice_pf *pf = vf->pf; 51 int i, last_vector_idx; 52 53 /* First, disable VF's configuration API to prevent OS from 54 * accessing the VF's VSI after it's freed or invalidated. 55 */ 56 clear_bit(ICE_VF_STATE_INIT, vf->vf_states); 57 ice_vf_fdir_exit(vf); 58 /* free VF control VSI */ 59 if (vf->ctrl_vsi_idx != ICE_NO_VSI) 60 ice_vf_ctrl_vsi_release(vf); 61 62 /* free VSI and disconnect it from the parent uplink */ 63 if (vf->lan_vsi_idx != ICE_NO_VSI) { 64 ice_vf_vsi_release(vf); 65 vf->num_mac = 0; 66 vf->num_mac_lldp = 0; 67 } 68 69 last_vector_idx = vf->first_vector_idx + vf->num_msix - 1; 70 71 /* clear VF MDD event information */ 72 memset(&vf->mdd_tx_events, 0, sizeof(vf->mdd_tx_events)); 73 memset(&vf->mdd_rx_events, 0, sizeof(vf->mdd_rx_events)); 74 75 /* Disable interrupts so that VF starts in a known state */ 76 for (i = vf->first_vector_idx; i <= last_vector_idx; i++) { 77 wr32(&pf->hw, GLINT_DYN_CTL(i), GLINT_DYN_CTL_CLEARPBA_M); 78 ice_flush(&pf->hw); 79 } 80 /* reset some of the state variables keeping track of the resources */ 81 clear_bit(ICE_VF_STATE_MC_PROMISC, vf->vf_states); 82 clear_bit(ICE_VF_STATE_UC_PROMISC, vf->vf_states); 83} 84 85/** 86 * ice_dis_vf_mappings 87 * @vf: pointer to the VF structure 88 */ 89static void ice_dis_vf_mappings(struct ice_vf *vf) 90{ 91 struct ice_pf *pf = vf->pf; 92 struct ice_vsi *vsi; 93 struct device *dev; 94 int first, last, v; 95 struct ice_hw *hw; 96 97 hw = &pf->hw; 98 vsi = ice_get_vf_vsi(vf); 99 if (WARN_ON(!vsi)) 100 return; 101 102 dev = ice_pf_to_dev(pf); 103 wr32(hw, VPINT_ALLOC(vf->vf_id), 0); 104 wr32(hw, VPINT_ALLOC_PCI(vf->vf_id), 0); 105 106 first = vf->first_vector_idx; 107 last = first + vf->num_msix - 1; 108 for (v = first; v <= last; v++) { 109 u32 reg; 110 111 reg = FIELD_PREP(GLINT_VECT2FUNC_IS_PF_M, 1) | 112 FIELD_PREP(GLINT_VECT2FUNC_PF_NUM_M, hw->pf_id); 113 wr32(hw, GLINT_VECT2FUNC(v), reg); 114 } 115 116 if (vsi->tx_mapping_mode == ICE_VSI_MAP_CONTIG) 117 wr32(hw, VPLAN_TX_QBASE(vf->vf_id), 0); 118 else 119 dev_err(dev, "Scattered mode for VF Tx queues is not yet implemented\n"); 120 121 if (vsi->rx_mapping_mode == ICE_VSI_MAP_CONTIG) 122 wr32(hw, VPLAN_RX_QBASE(vf->vf_id), 0); 123 else 124 dev_err(dev, "Scattered mode for VF Rx queues is not yet implemented\n"); 125} 126 127/** 128 * ice_free_vfs - Free all VFs 129 * @pf: pointer to the PF structure 130 */ 131void ice_free_vfs(struct ice_pf *pf) 132{ 133 struct device *dev = ice_pf_to_dev(pf); 134 struct ice_vfs *vfs = &pf->vfs; 135 struct ice_hw *hw = &pf->hw; 136 struct ice_vf *vf; 137 unsigned int bkt; 138 139 if (!ice_has_vfs(pf)) 140 return; 141 142 while (test_and_set_bit(ICE_VF_DIS, pf->state)) 143 usleep_range(1000, 2000); 144 145 /* Disable IOV before freeing resources. This lets any VF drivers 146 * running in the host get themselves cleaned up before we yank 147 * the carpet out from underneath their feet. 148 */ 149 if (!pci_vfs_assigned(pf->pdev)) 150 pci_disable_sriov(pf->pdev); 151 else 152 dev_warn(dev, "VFs are assigned - not disabling SR-IOV\n"); 153 154 mutex_lock(&vfs->table_lock); 155 156 ice_for_each_vf(pf, bkt, vf) { 157 mutex_lock(&vf->cfg_lock); 158 159 ice_eswitch_detach_vf(pf, vf); 160 ice_dis_vf_qs(vf); 161 ice_virt_free_irqs(pf, vf->first_vector_idx, vf->num_msix); 162 163 if (test_bit(ICE_VF_STATE_INIT, vf->vf_states)) { 164 /* disable VF qp mappings and set VF disable state */ 165 ice_dis_vf_mappings(vf); 166 set_bit(ICE_VF_STATE_DIS, vf->vf_states); 167 ice_free_vf_res(vf); 168 } 169 170 if (!pci_vfs_assigned(pf->pdev)) { 171 u32 reg_idx, bit_idx; 172 173 reg_idx = (hw->func_caps.vf_base_id + vf->vf_id) / 32; 174 bit_idx = (hw->func_caps.vf_base_id + vf->vf_id) % 32; 175 wr32(hw, GLGEN_VFLRSTAT(reg_idx), BIT(bit_idx)); 176 } 177 178 mutex_unlock(&vf->cfg_lock); 179 } 180 181 vfs->num_qps_per = 0; 182 ice_free_vf_entries(pf); 183 184 mutex_unlock(&vfs->table_lock); 185 186 clear_bit(ICE_VF_DIS, pf->state); 187 clear_bit(ICE_FLAG_SRIOV_ENA, pf->flags); 188} 189 190/** 191 * ice_vf_vsi_setup - Set up a VF VSI 192 * @vf: VF to setup VSI for 193 * 194 * Returns pointer to the successfully allocated VSI struct on success, 195 * otherwise returns NULL on failure. 196 */ 197static struct ice_vsi *ice_vf_vsi_setup(struct ice_vf *vf) 198{ 199 struct ice_vsi_cfg_params params = {}; 200 struct ice_pf *pf = vf->pf; 201 struct ice_vsi *vsi; 202 203 params.type = ICE_VSI_VF; 204 params.port_info = ice_vf_get_port_info(vf); 205 params.vf = vf; 206 params.flags = ICE_VSI_FLAG_INIT; 207 208 vsi = ice_vsi_setup(pf, ¶ms); 209 210 if (!vsi) { 211 dev_err(ice_pf_to_dev(pf), "Failed to create VF VSI\n"); 212 ice_vf_invalidate_vsi(vf); 213 return NULL; 214 } 215 216 vf->lan_vsi_idx = vsi->idx; 217 218 return vsi; 219} 220 221 222/** 223 * ice_ena_vf_msix_mappings - enable VF MSIX mappings in hardware 224 * @vf: VF to enable MSIX mappings for 225 * 226 * Some of the registers need to be indexed/configured using hardware global 227 * device values and other registers need 0-based values, which represent PF 228 * based values. 229 */ 230static void ice_ena_vf_msix_mappings(struct ice_vf *vf) 231{ 232 int device_based_first_msix, device_based_last_msix; 233 int pf_based_first_msix, pf_based_last_msix, v; 234 struct ice_pf *pf = vf->pf; 235 int device_based_vf_id; 236 struct ice_hw *hw; 237 u32 reg; 238 239 hw = &pf->hw; 240 pf_based_first_msix = vf->first_vector_idx; 241 pf_based_last_msix = (pf_based_first_msix + vf->num_msix) - 1; 242 243 device_based_first_msix = pf_based_first_msix + 244 pf->hw.func_caps.common_cap.msix_vector_first_id; 245 device_based_last_msix = 246 (device_based_first_msix + vf->num_msix) - 1; 247 device_based_vf_id = vf->vf_id + hw->func_caps.vf_base_id; 248 249 reg = FIELD_PREP(VPINT_ALLOC_FIRST_M, device_based_first_msix) | 250 FIELD_PREP(VPINT_ALLOC_LAST_M, device_based_last_msix) | 251 VPINT_ALLOC_VALID_M; 252 wr32(hw, VPINT_ALLOC(vf->vf_id), reg); 253 254 reg = FIELD_PREP(VPINT_ALLOC_PCI_FIRST_M, device_based_first_msix) | 255 FIELD_PREP(VPINT_ALLOC_PCI_LAST_M, device_based_last_msix) | 256 VPINT_ALLOC_PCI_VALID_M; 257 wr32(hw, VPINT_ALLOC_PCI(vf->vf_id), reg); 258 259 /* map the interrupts to its functions */ 260 for (v = pf_based_first_msix; v <= pf_based_last_msix; v++) { 261 reg = FIELD_PREP(GLINT_VECT2FUNC_VF_NUM_M, device_based_vf_id) | 262 FIELD_PREP(GLINT_VECT2FUNC_PF_NUM_M, hw->pf_id); 263 wr32(hw, GLINT_VECT2FUNC(v), reg); 264 } 265 266 /* Map mailbox interrupt to VF MSI-X vector 0 */ 267 wr32(hw, VPINT_MBX_CTL(device_based_vf_id), VPINT_MBX_CTL_CAUSE_ENA_M); 268} 269 270/** 271 * ice_ena_vf_q_mappings - enable Rx/Tx queue mappings for a VF 272 * @vf: VF to enable the mappings for 273 * @max_txq: max Tx queues allowed on the VF's VSI 274 * @max_rxq: max Rx queues allowed on the VF's VSI 275 */ 276static void ice_ena_vf_q_mappings(struct ice_vf *vf, u16 max_txq, u16 max_rxq) 277{ 278 struct device *dev = ice_pf_to_dev(vf->pf); 279 struct ice_vsi *vsi = ice_get_vf_vsi(vf); 280 struct ice_hw *hw = &vf->pf->hw; 281 u32 reg; 282 283 if (WARN_ON(!vsi)) 284 return; 285 286 /* set regardless of mapping mode */ 287 wr32(hw, VPLAN_TXQ_MAPENA(vf->vf_id), VPLAN_TXQ_MAPENA_TX_ENA_M); 288 289 /* VF Tx queues allocation */ 290 if (vsi->tx_mapping_mode == ICE_VSI_MAP_CONTIG) { 291 /* set the VF PF Tx queue range 292 * VFNUMQ value should be set to (number of queues - 1). A value 293 * of 0 means 1 queue and a value of 255 means 256 queues 294 */ 295 reg = FIELD_PREP(VPLAN_TX_QBASE_VFFIRSTQ_M, vsi->txq_map[0]) | 296 FIELD_PREP(VPLAN_TX_QBASE_VFNUMQ_M, max_txq - 1); 297 wr32(hw, VPLAN_TX_QBASE(vf->vf_id), reg); 298 } else { 299 dev_err(dev, "Scattered mode for VF Tx queues is not yet implemented\n"); 300 } 301 302 /* set regardless of mapping mode */ 303 wr32(hw, VPLAN_RXQ_MAPENA(vf->vf_id), VPLAN_RXQ_MAPENA_RX_ENA_M); 304 305 /* VF Rx queues allocation */ 306 if (vsi->rx_mapping_mode == ICE_VSI_MAP_CONTIG) { 307 /* set the VF PF Rx queue range 308 * VFNUMQ value should be set to (number of queues - 1). A value 309 * of 0 means 1 queue and a value of 255 means 256 queues 310 */ 311 reg = FIELD_PREP(VPLAN_RX_QBASE_VFFIRSTQ_M, vsi->rxq_map[0]) | 312 FIELD_PREP(VPLAN_RX_QBASE_VFNUMQ_M, max_rxq - 1); 313 wr32(hw, VPLAN_RX_QBASE(vf->vf_id), reg); 314 } else { 315 dev_err(dev, "Scattered mode for VF Rx queues is not yet implemented\n"); 316 } 317} 318 319/** 320 * ice_ena_vf_mappings - enable VF MSIX and queue mapping 321 * @vf: pointer to the VF structure 322 */ 323static void ice_ena_vf_mappings(struct ice_vf *vf) 324{ 325 struct ice_vsi *vsi = ice_get_vf_vsi(vf); 326 327 if (WARN_ON(!vsi)) 328 return; 329 330 ice_ena_vf_msix_mappings(vf); 331 ice_ena_vf_q_mappings(vf, vsi->alloc_txq, vsi->alloc_rxq); 332} 333 334/** 335 * ice_calc_vf_reg_idx - Calculate the VF's register index in the PF space 336 * @vf: VF to calculate the register index for 337 * @q_vector: a q_vector associated to the VF 338 */ 339void ice_calc_vf_reg_idx(struct ice_vf *vf, struct ice_q_vector *q_vector) 340{ 341 if (!vf || !q_vector) 342 return; 343 344 /* always add one to account for the OICR being the first MSIX */ 345 q_vector->vf_reg_idx = q_vector->v_idx + ICE_NONQ_VECS_VF; 346 q_vector->reg_idx = vf->first_vector_idx + q_vector->vf_reg_idx; 347} 348 349/** 350 * ice_set_per_vf_res - check if vectors and queues are available 351 * @pf: pointer to the PF structure 352 * @num_vfs: the number of SR-IOV VFs being configured 353 * 354 * First, determine HW interrupts from common pool. If we allocate fewer VFs, we 355 * get more vectors and can enable more queues per VF. Note that this does not 356 * grab any vectors from the SW pool already allocated. Also note, that all 357 * vector counts include one for each VF's miscellaneous interrupt vector 358 * (i.e. OICR). 359 * 360 * Minimum VFs - 2 vectors, 1 queue pair 361 * Small VFs - 5 vectors, 4 queue pairs 362 * Medium VFs - 17 vectors, 16 queue pairs 363 * 364 * Second, determine number of queue pairs per VF by starting with a pre-defined 365 * maximum each VF supports. If this is not possible, then we adjust based on 366 * queue pairs available on the device. 367 * 368 * Lastly, set queue and MSI-X VF variables tracked by the PF so it can be used 369 * by each VF during VF initialization and reset. 370 */ 371static int ice_set_per_vf_res(struct ice_pf *pf, u16 num_vfs) 372{ 373 u16 num_msix_per_vf, num_txq, num_rxq, avail_qs; 374 int msix_avail_per_vf, msix_avail_for_sriov; 375 struct device *dev = ice_pf_to_dev(pf); 376 377 lockdep_assert_held(&pf->vfs.table_lock); 378 379 if (!num_vfs) 380 return -EINVAL; 381 382 /* determine MSI-X resources per VF */ 383 msix_avail_for_sriov = pf->virt_irq_tracker.num_entries; 384 msix_avail_per_vf = msix_avail_for_sriov / num_vfs; 385 if (msix_avail_per_vf >= ICE_NUM_VF_MSIX_MED) { 386 num_msix_per_vf = ICE_NUM_VF_MSIX_MED; 387 } else if (msix_avail_per_vf >= ICE_NUM_VF_MSIX_SMALL) { 388 num_msix_per_vf = ICE_NUM_VF_MSIX_SMALL; 389 } else if (msix_avail_per_vf >= ICE_NUM_VF_MSIX_MULTIQ_MIN) { 390 num_msix_per_vf = ICE_NUM_VF_MSIX_MULTIQ_MIN; 391 } else if (msix_avail_per_vf >= ICE_MIN_INTR_PER_VF) { 392 num_msix_per_vf = ICE_MIN_INTR_PER_VF; 393 } else { 394 dev_err(dev, "Only %d MSI-X interrupts available for SR-IOV. Not enough to support minimum of %d MSI-X interrupts per VF for %d VFs\n", 395 msix_avail_for_sriov, ICE_MIN_INTR_PER_VF, 396 num_vfs); 397 return -ENOSPC; 398 } 399 400 num_txq = min_t(u16, num_msix_per_vf - ICE_NONQ_VECS_VF, 401 ICE_MAX_RSS_QS_PER_VF); 402 avail_qs = ice_get_avail_txq_count(pf) / num_vfs; 403 if (!avail_qs) 404 num_txq = 0; 405 else if (num_txq > avail_qs) 406 num_txq = rounddown_pow_of_two(avail_qs); 407 408 num_rxq = min_t(u16, num_msix_per_vf - ICE_NONQ_VECS_VF, 409 ICE_MAX_RSS_QS_PER_VF); 410 avail_qs = ice_get_avail_rxq_count(pf) / num_vfs; 411 if (!avail_qs) 412 num_rxq = 0; 413 else if (num_rxq > avail_qs) 414 num_rxq = rounddown_pow_of_two(avail_qs); 415 416 if (num_txq < ICE_MIN_QS_PER_VF || num_rxq < ICE_MIN_QS_PER_VF) { 417 dev_err(dev, "Not enough queues to support minimum of %d queue pairs per VF for %d VFs\n", 418 ICE_MIN_QS_PER_VF, num_vfs); 419 return -ENOSPC; 420 } 421 422 /* only allow equal Tx/Rx queue count (i.e. queue pairs) */ 423 pf->vfs.num_qps_per = min_t(int, num_txq, num_rxq); 424 pf->vfs.num_msix_per = num_msix_per_vf; 425 dev_info(dev, "Enabling %d VFs with %d vectors and %d queues per VF\n", 426 num_vfs, pf->vfs.num_msix_per, pf->vfs.num_qps_per); 427 428 return 0; 429} 430 431/** 432 * ice_init_vf_vsi_res - initialize/setup VF VSI resources 433 * @vf: VF to initialize/setup the VSI for 434 * 435 * This function creates a VSI for the VF, adds a VLAN 0 filter, and sets up the 436 * VF VSI's broadcast filter and is only used during initial VF creation. 437 */ 438static int ice_init_vf_vsi_res(struct ice_vf *vf) 439{ 440 struct ice_pf *pf = vf->pf; 441 struct ice_vsi *vsi; 442 int err; 443 444 vf->first_vector_idx = ice_virt_get_irqs(pf, vf->num_msix); 445 if (vf->first_vector_idx < 0) 446 return -ENOMEM; 447 448 vsi = ice_vf_vsi_setup(vf); 449 if (!vsi) 450 return -ENOMEM; 451 452 err = ice_vf_init_host_cfg(vf, vsi); 453 if (err) 454 goto release_vsi; 455 456 return 0; 457 458release_vsi: 459 ice_vf_vsi_release(vf); 460 return err; 461} 462 463/** 464 * ice_start_vfs - start VFs so they are ready to be used by SR-IOV 465 * @pf: PF the VFs are associated with 466 */ 467static int ice_start_vfs(struct ice_pf *pf) 468{ 469 struct ice_hw *hw = &pf->hw; 470 unsigned int bkt, it_cnt; 471 struct ice_vf *vf; 472 int retval; 473 474 lockdep_assert_held(&pf->vfs.table_lock); 475 476 it_cnt = 0; 477 ice_for_each_vf(pf, bkt, vf) { 478 vf->vf_ops->clear_reset_trigger(vf); 479 480 retval = ice_init_vf_vsi_res(vf); 481 if (retval) { 482 dev_err(ice_pf_to_dev(pf), "Failed to initialize VSI resources for VF %d, error %d\n", 483 vf->vf_id, retval); 484 goto teardown; 485 } 486 487 retval = ice_eswitch_attach_vf(pf, vf); 488 if (retval) { 489 dev_err(ice_pf_to_dev(pf), "Failed to attach VF %d to eswitch, error %d", 490 vf->vf_id, retval); 491 ice_vf_vsi_release(vf); 492 goto teardown; 493 } 494 495 set_bit(ICE_VF_STATE_INIT, vf->vf_states); 496 ice_ena_vf_mappings(vf); 497 wr32(hw, VFGEN_RSTAT(vf->vf_id), VIRTCHNL_VFR_VFACTIVE); 498 it_cnt++; 499 } 500 501 ice_flush(hw); 502 return 0; 503 504teardown: 505 ice_for_each_vf(pf, bkt, vf) { 506 if (it_cnt == 0) 507 break; 508 509 ice_dis_vf_mappings(vf); 510 ice_vf_vsi_release(vf); 511 it_cnt--; 512 } 513 514 return retval; 515} 516 517/** 518 * ice_sriov_free_vf - Free VF memory after all references are dropped 519 * @vf: pointer to VF to free 520 * 521 * Called by ice_put_vf through ice_release_vf once the last reference to a VF 522 * structure has been dropped. 523 */ 524static void ice_sriov_free_vf(struct ice_vf *vf) 525{ 526 mutex_destroy(&vf->cfg_lock); 527 528 kfree_rcu(vf, rcu); 529} 530 531/** 532 * ice_sriov_clear_reset_state - clears VF Reset status register 533 * @vf: the vf to configure 534 */ 535static void ice_sriov_clear_reset_state(struct ice_vf *vf) 536{ 537 struct ice_hw *hw = &vf->pf->hw; 538 539 /* Clear the reset status register so that VF immediately sees that 540 * the device is resetting, even if hardware hasn't yet gotten around 541 * to clearing VFGEN_RSTAT for us. 542 */ 543 wr32(hw, VFGEN_RSTAT(vf->vf_id), VIRTCHNL_VFR_INPROGRESS); 544} 545 546/** 547 * ice_sriov_clear_mbx_register - clears SRIOV VF's mailbox registers 548 * @vf: the vf to configure 549 */ 550static void ice_sriov_clear_mbx_register(struct ice_vf *vf) 551{ 552 struct ice_pf *pf = vf->pf; 553 554 wr32(&pf->hw, VF_MBX_ARQLEN(vf->vf_id), 0); 555 wr32(&pf->hw, VF_MBX_ATQLEN(vf->vf_id), 0); 556} 557 558/** 559 * ice_sriov_trigger_reset_register - trigger VF reset for SRIOV VF 560 * @vf: pointer to VF structure 561 * @is_vflr: true if reset occurred due to VFLR 562 * 563 * Trigger and cleanup after a VF reset for a SR-IOV VF. 564 */ 565static void ice_sriov_trigger_reset_register(struct ice_vf *vf, bool is_vflr) 566{ 567 struct ice_pf *pf = vf->pf; 568 u32 reg, reg_idx, bit_idx; 569 unsigned int vf_abs_id, i; 570 struct device *dev; 571 struct ice_hw *hw; 572 573 dev = ice_pf_to_dev(pf); 574 hw = &pf->hw; 575 vf_abs_id = vf->vf_id + hw->func_caps.vf_base_id; 576 577 /* In the case of a VFLR, HW has already reset the VF and we just need 578 * to clean up. Otherwise we must first trigger the reset using the 579 * VFRTRIG register. 580 */ 581 if (!is_vflr) { 582 reg = rd32(hw, VPGEN_VFRTRIG(vf->vf_id)); 583 reg |= VPGEN_VFRTRIG_VFSWR_M; 584 wr32(hw, VPGEN_VFRTRIG(vf->vf_id), reg); 585 } 586 587 /* clear the VFLR bit in GLGEN_VFLRSTAT */ 588 reg_idx = (vf_abs_id) / 32; 589 bit_idx = (vf_abs_id) % 32; 590 wr32(hw, GLGEN_VFLRSTAT(reg_idx), BIT(bit_idx)); 591 ice_flush(hw); 592 593 wr32(hw, PF_PCI_CIAA, 594 VF_DEVICE_STATUS | (vf_abs_id << PF_PCI_CIAA_VF_NUM_S)); 595 for (i = 0; i < ICE_PCI_CIAD_WAIT_COUNT; i++) { 596 reg = rd32(hw, PF_PCI_CIAD); 597 /* no transactions pending so stop polling */ 598 if ((reg & VF_TRANS_PENDING_M) == 0) 599 break; 600 601 dev_err(dev, "VF %u PCI transactions stuck\n", vf->vf_id); 602 udelay(ICE_PCI_CIAD_WAIT_DELAY_US); 603 } 604} 605 606/** 607 * ice_sriov_poll_reset_status - poll SRIOV VF reset status 608 * @vf: pointer to VF structure 609 * 610 * Returns true when reset is successful, else returns false 611 */ 612static bool ice_sriov_poll_reset_status(struct ice_vf *vf) 613{ 614 struct ice_pf *pf = vf->pf; 615 unsigned int i; 616 u32 reg; 617 618 for (i = 0; i < 10; i++) { 619 /* VF reset requires driver to first reset the VF and then 620 * poll the status register to make sure that the reset 621 * completed successfully. 622 */ 623 reg = rd32(&pf->hw, VPGEN_VFRSTAT(vf->vf_id)); 624 if (reg & VPGEN_VFRSTAT_VFRD_M) 625 return true; 626 627 /* only sleep if the reset is not done */ 628 usleep_range(10, 20); 629 } 630 return false; 631} 632 633/** 634 * ice_sriov_clear_reset_trigger - enable VF to access hardware 635 * @vf: VF to enabled hardware access for 636 */ 637static void ice_sriov_clear_reset_trigger(struct ice_vf *vf) 638{ 639 struct ice_hw *hw = &vf->pf->hw; 640 u32 reg; 641 642 reg = rd32(hw, VPGEN_VFRTRIG(vf->vf_id)); 643 reg &= ~VPGEN_VFRTRIG_VFSWR_M; 644 wr32(hw, VPGEN_VFRTRIG(vf->vf_id), reg); 645 ice_flush(hw); 646} 647 648/** 649 * ice_sriov_post_vsi_rebuild - tasks to do after the VF's VSI have been rebuilt 650 * @vf: VF to perform tasks on 651 */ 652static void ice_sriov_post_vsi_rebuild(struct ice_vf *vf) 653{ 654 ice_ena_vf_mappings(vf); 655 wr32(&vf->pf->hw, VFGEN_RSTAT(vf->vf_id), VIRTCHNL_VFR_VFACTIVE); 656} 657 658static const struct ice_vf_ops ice_sriov_vf_ops = { 659 .reset_type = ICE_VF_RESET, 660 .free = ice_sriov_free_vf, 661 .clear_reset_state = ice_sriov_clear_reset_state, 662 .clear_mbx_register = ice_sriov_clear_mbx_register, 663 .trigger_reset_register = ice_sriov_trigger_reset_register, 664 .poll_reset_status = ice_sriov_poll_reset_status, 665 .clear_reset_trigger = ice_sriov_clear_reset_trigger, 666 .irq_close = NULL, 667 .post_vsi_rebuild = ice_sriov_post_vsi_rebuild, 668}; 669 670/** 671 * ice_create_vf_entries - Allocate and insert VF entries 672 * @pf: pointer to the PF structure 673 * @num_vfs: the number of VFs to allocate 674 * 675 * Allocate new VF entries and insert them into the hash table. Set some 676 * basic default fields for initializing the new VFs. 677 * 678 * After this function exits, the hash table will have num_vfs entries 679 * inserted. 680 * 681 * Returns 0 on success or an integer error code on failure. 682 */ 683static int ice_create_vf_entries(struct ice_pf *pf, u16 num_vfs) 684{ 685 struct pci_dev *pdev = pf->pdev; 686 struct ice_vfs *vfs = &pf->vfs; 687 struct pci_dev *vfdev = NULL; 688 struct ice_vf *vf; 689 u16 vf_pdev_id; 690 int err, pos; 691 692 lockdep_assert_held(&vfs->table_lock); 693 694 pos = pci_find_ext_capability(pdev, PCI_EXT_CAP_ID_SRIOV); 695 pci_read_config_word(pdev, pos + PCI_SRIOV_VF_DID, &vf_pdev_id); 696 697 for (u16 vf_id = 0; vf_id < num_vfs; vf_id++) { 698 vf = kzalloc(sizeof(*vf), GFP_KERNEL); 699 if (!vf) { 700 err = -ENOMEM; 701 goto err_free_entries; 702 } 703 kref_init(&vf->refcnt); 704 705 vf->pf = pf; 706 vf->vf_id = vf_id; 707 708 /* set sriov vf ops for VFs created during SRIOV flow */ 709 vf->vf_ops = &ice_sriov_vf_ops; 710 711 ice_initialize_vf_entry(vf); 712 713 do { 714 vfdev = pci_get_device(pdev->vendor, vf_pdev_id, vfdev); 715 } while (vfdev && vfdev->physfn != pdev); 716 vf->vfdev = vfdev; 717 vf->vf_sw_id = pf->first_sw; 718 719 pci_dev_get(vfdev); 720 721 hash_add_rcu(vfs->table, &vf->entry, vf_id); 722 } 723 724 /* Decrement of refcount done by pci_get_device() inside the loop does 725 * not touch the last iteration's vfdev, so it has to be done manually 726 * to balance pci_dev_get() added within the loop. 727 */ 728 pci_dev_put(vfdev); 729 730 return 0; 731 732err_free_entries: 733 ice_free_vf_entries(pf); 734 return err; 735} 736 737/** 738 * ice_ena_vfs - enable VFs so they are ready to be used 739 * @pf: pointer to the PF structure 740 * @num_vfs: number of VFs to enable 741 */ 742static int ice_ena_vfs(struct ice_pf *pf, u16 num_vfs) 743{ 744 struct device *dev = ice_pf_to_dev(pf); 745 struct ice_hw *hw = &pf->hw; 746 int ret; 747 748 /* Disable global interrupt 0 so we don't try to handle the VFLR. */ 749 wr32(hw, GLINT_DYN_CTL(pf->oicr_irq.index), 750 ICE_ITR_NONE << GLINT_DYN_CTL_ITR_INDX_S); 751 set_bit(ICE_OICR_INTR_DIS, pf->state); 752 ice_flush(hw); 753 754 ret = pci_enable_sriov(pf->pdev, num_vfs); 755 if (ret) 756 goto err_unroll_intr; 757 758 mutex_lock(&pf->vfs.table_lock); 759 760 ret = ice_set_per_vf_res(pf, num_vfs); 761 if (ret) { 762 dev_err(dev, "Not enough resources for %d VFs, err %d. Try with fewer number of VFs\n", 763 num_vfs, ret); 764 goto err_unroll_sriov; 765 } 766 767 ret = ice_create_vf_entries(pf, num_vfs); 768 if (ret) { 769 dev_err(dev, "Failed to allocate VF entries for %d VFs\n", 770 num_vfs); 771 goto err_unroll_sriov; 772 } 773 774 ret = ice_start_vfs(pf); 775 if (ret) { 776 dev_err(dev, "Failed to start %d VFs, err %d\n", num_vfs, ret); 777 ret = -EAGAIN; 778 goto err_unroll_vf_entries; 779 } 780 781 clear_bit(ICE_VF_DIS, pf->state); 782 783 /* rearm global interrupts */ 784 if (test_and_clear_bit(ICE_OICR_INTR_DIS, pf->state)) 785 ice_irq_dynamic_ena(hw, NULL, NULL); 786 787 mutex_unlock(&pf->vfs.table_lock); 788 789 return 0; 790 791err_unroll_vf_entries: 792 ice_free_vf_entries(pf); 793err_unroll_sriov: 794 mutex_unlock(&pf->vfs.table_lock); 795 pci_disable_sriov(pf->pdev); 796err_unroll_intr: 797 /* rearm interrupts here */ 798 ice_irq_dynamic_ena(hw, NULL, NULL); 799 clear_bit(ICE_OICR_INTR_DIS, pf->state); 800 return ret; 801} 802 803/** 804 * ice_pci_sriov_ena - Enable or change number of VFs 805 * @pf: pointer to the PF structure 806 * @num_vfs: number of VFs to allocate 807 * 808 * Returns 0 on success and negative on failure 809 */ 810static int ice_pci_sriov_ena(struct ice_pf *pf, int num_vfs) 811{ 812 struct device *dev = ice_pf_to_dev(pf); 813 int err; 814 815 if (!num_vfs) { 816 ice_free_vfs(pf); 817 return 0; 818 } 819 820 if (num_vfs > pf->vfs.num_supported) { 821 dev_err(dev, "Can't enable %d VFs, max VFs supported is %d\n", 822 num_vfs, pf->vfs.num_supported); 823 return -EOPNOTSUPP; 824 } 825 826 dev_info(dev, "Enabling %d VFs\n", num_vfs); 827 err = ice_ena_vfs(pf, num_vfs); 828 if (err) { 829 dev_err(dev, "Failed to enable SR-IOV: %d\n", err); 830 return err; 831 } 832 833 set_bit(ICE_FLAG_SRIOV_ENA, pf->flags); 834 return 0; 835} 836 837/** 838 * ice_check_sriov_allowed - check if SR-IOV is allowed based on various checks 839 * @pf: PF to enabled SR-IOV on 840 */ 841static int ice_check_sriov_allowed(struct ice_pf *pf) 842{ 843 struct device *dev = ice_pf_to_dev(pf); 844 845 if (!test_bit(ICE_FLAG_SRIOV_CAPABLE, pf->flags)) { 846 dev_err(dev, "This device is not capable of SR-IOV\n"); 847 return -EOPNOTSUPP; 848 } 849 850 if (ice_is_safe_mode(pf)) { 851 dev_err(dev, "SR-IOV cannot be configured - Device is in Safe Mode\n"); 852 return -EOPNOTSUPP; 853 } 854 855 if (!ice_pf_state_is_nominal(pf)) { 856 dev_err(dev, "Cannot enable SR-IOV, device not ready\n"); 857 return -EBUSY; 858 } 859 860 return 0; 861} 862 863/** 864 * ice_sriov_get_vf_total_msix - return number of MSI-X used by VFs 865 * @pdev: pointer to pci_dev struct 866 * 867 * The function is called via sysfs ops 868 */ 869u32 ice_sriov_get_vf_total_msix(struct pci_dev *pdev) 870{ 871 struct ice_pf *pf = pci_get_drvdata(pdev); 872 873 return pf->virt_irq_tracker.num_entries; 874} 875 876static void ice_sriov_remap_vectors(struct ice_pf *pf, u16 restricted_id) 877{ 878 u16 vf_ids[ICE_MAX_SRIOV_VFS]; 879 struct ice_vf *tmp_vf; 880 int to_remap = 0, bkt; 881 882 /* For better irqs usage try to remap irqs of VFs 883 * that aren't running yet 884 */ 885 ice_for_each_vf(pf, bkt, tmp_vf) { 886 /* skip VF which is changing the number of MSI-X */ 887 if (restricted_id == tmp_vf->vf_id || 888 test_bit(ICE_VF_STATE_ACTIVE, tmp_vf->vf_states)) 889 continue; 890 891 ice_dis_vf_mappings(tmp_vf); 892 ice_virt_free_irqs(pf, tmp_vf->first_vector_idx, 893 tmp_vf->num_msix); 894 895 vf_ids[to_remap] = tmp_vf->vf_id; 896 to_remap += 1; 897 } 898 899 for (int i = 0; i < to_remap; i++) { 900 tmp_vf = ice_get_vf_by_id(pf, vf_ids[i]); 901 if (!tmp_vf) 902 continue; 903 904 tmp_vf->first_vector_idx = 905 ice_virt_get_irqs(pf, tmp_vf->num_msix); 906 /* there is no need to rebuild VSI as we are only changing the 907 * vector indexes not amount of MSI-X or queues 908 */ 909 ice_ena_vf_mappings(tmp_vf); 910 ice_put_vf(tmp_vf); 911 } 912} 913 914/** 915 * ice_sriov_set_msix_vec_count 916 * @vf_dev: pointer to pci_dev struct of VF device 917 * @msix_vec_count: new value for MSI-X amount on this VF 918 * 919 * Set requested MSI-X, queues and registers for @vf_dev. 920 * 921 * First do some sanity checks like if there are any VFs, if the new value 922 * is correct etc. Then disable old mapping (MSI-X and queues registers), change 923 * MSI-X and queues, rebuild VSI and enable new mapping. 924 * 925 * If it is possible (driver not binded to VF) try to remap also other VFs to 926 * linearize irqs register usage. 927 */ 928int ice_sriov_set_msix_vec_count(struct pci_dev *vf_dev, int msix_vec_count) 929{ 930 struct pci_dev *pdev = pci_physfn(vf_dev); 931 struct ice_pf *pf = pci_get_drvdata(pdev); 932 u16 prev_msix, prev_queues, queues; 933 bool needs_rebuild = false; 934 struct ice_vsi *vsi; 935 struct ice_vf *vf; 936 937 if (!ice_get_num_vfs(pf)) 938 return -ENOENT; 939 940 if (!msix_vec_count) 941 return 0; 942 943 queues = msix_vec_count; 944 /* add 1 MSI-X for OICR */ 945 msix_vec_count += 1; 946 947 if (queues > min(ice_get_avail_txq_count(pf), 948 ice_get_avail_rxq_count(pf))) 949 return -EINVAL; 950 951 if (msix_vec_count < ICE_MIN_INTR_PER_VF) 952 return -EINVAL; 953 954 vf = ice_get_vf_by_dev(pf, vf_dev); 955 if (!vf) 956 return -ENOENT; 957 958 vsi = ice_get_vf_vsi(vf); 959 if (!vsi) { 960 ice_put_vf(vf); 961 return -ENOENT; 962 } 963 964 /* No need to rebuild if we're setting to the same value */ 965 if (msix_vec_count == vf->num_msix) { 966 ice_put_vf(vf); 967 return 0; 968 } 969 970 prev_msix = vf->num_msix; 971 prev_queues = vf->num_vf_qs; 972 973 ice_dis_vf_mappings(vf); 974 ice_virt_free_irqs(pf, vf->first_vector_idx, vf->num_msix); 975 976 /* Remap all VFs beside the one is now configured */ 977 ice_sriov_remap_vectors(pf, vf->vf_id); 978 979 vf->num_msix = msix_vec_count; 980 vf->num_vf_qs = queues; 981 vf->first_vector_idx = ice_virt_get_irqs(pf, vf->num_msix); 982 if (vf->first_vector_idx < 0) 983 goto unroll; 984 985 vsi->req_txq = queues; 986 vsi->req_rxq = queues; 987 988 if (ice_vsi_rebuild(vsi, ICE_VSI_FLAG_NO_INIT)) { 989 /* Try to rebuild with previous values */ 990 needs_rebuild = true; 991 goto unroll; 992 } 993 994 dev_info(ice_pf_to_dev(pf), 995 "Changing VF %d resources to %d vectors and %d queues\n", 996 vf->vf_id, vf->num_msix, vf->num_vf_qs); 997 998 ice_ena_vf_mappings(vf); 999 ice_put_vf(vf); 1000 1001 return 0; 1002 1003unroll: 1004 dev_info(ice_pf_to_dev(pf), 1005 "Can't set %d vectors on VF %d, falling back to %d\n", 1006 vf->num_msix, vf->vf_id, prev_msix); 1007 1008 vf->num_msix = prev_msix; 1009 vf->num_vf_qs = prev_queues; 1010 1011 vf->first_vector_idx = ice_virt_get_irqs(pf, vf->num_msix); 1012 if (vf->first_vector_idx < 0) { 1013 ice_put_vf(vf); 1014 return -EINVAL; 1015 } 1016 1017 if (needs_rebuild) { 1018 vsi->req_txq = prev_queues; 1019 vsi->req_rxq = prev_queues; 1020 1021 ice_vsi_rebuild(vsi, ICE_VSI_FLAG_NO_INIT); 1022 } 1023 1024 ice_ena_vf_mappings(vf); 1025 ice_put_vf(vf); 1026 1027 return -EINVAL; 1028} 1029 1030/** 1031 * ice_sriov_configure - Enable or change number of VFs via sysfs 1032 * @pdev: pointer to a pci_dev structure 1033 * @num_vfs: number of VFs to allocate or 0 to free VFs 1034 * 1035 * This function is called when the user updates the number of VFs in sysfs. On 1036 * success return whatever num_vfs was set to by the caller. Return negative on 1037 * failure. 1038 */ 1039int ice_sriov_configure(struct pci_dev *pdev, int num_vfs) 1040{ 1041 struct ice_pf *pf = pci_get_drvdata(pdev); 1042 struct device *dev = ice_pf_to_dev(pf); 1043 int err; 1044 1045 err = ice_check_sriov_allowed(pf); 1046 if (err) 1047 return err; 1048 1049 if (!num_vfs) { 1050 if (!pci_vfs_assigned(pdev)) { 1051 ice_free_vfs(pf); 1052 return 0; 1053 } 1054 1055 dev_err(dev, "can't free VFs because some are assigned to VMs.\n"); 1056 return -EBUSY; 1057 } 1058 1059 err = ice_pci_sriov_ena(pf, num_vfs); 1060 if (err) 1061 return err; 1062 1063 return num_vfs; 1064} 1065 1066/** 1067 * ice_process_vflr_event - Free VF resources via IRQ calls 1068 * @pf: pointer to the PF structure 1069 * 1070 * called from the VFLR IRQ handler to 1071 * free up VF resources and state variables 1072 */ 1073void ice_process_vflr_event(struct ice_pf *pf) 1074{ 1075 struct ice_hw *hw = &pf->hw; 1076 struct ice_vf *vf; 1077 unsigned int bkt; 1078 u32 reg; 1079 1080 if (!test_and_clear_bit(ICE_VFLR_EVENT_PENDING, pf->state) || 1081 !ice_has_vfs(pf)) 1082 return; 1083 1084 mutex_lock(&pf->vfs.table_lock); 1085 ice_for_each_vf(pf, bkt, vf) { 1086 u32 reg_idx, bit_idx; 1087 1088 reg_idx = (hw->func_caps.vf_base_id + vf->vf_id) / 32; 1089 bit_idx = (hw->func_caps.vf_base_id + vf->vf_id) % 32; 1090 /* read GLGEN_VFLRSTAT register to find out the flr VFs */ 1091 reg = rd32(hw, GLGEN_VFLRSTAT(reg_idx)); 1092 if (reg & BIT(bit_idx)) 1093 /* GLGEN_VFLRSTAT bit will be cleared in ice_reset_vf */ 1094 ice_reset_vf(vf, ICE_VF_RESET_VFLR | ICE_VF_RESET_LOCK); 1095 } 1096 mutex_unlock(&pf->vfs.table_lock); 1097} 1098 1099/** 1100 * ice_get_vf_from_pfq - get the VF who owns the PF space queue passed in 1101 * @pf: PF used to index all VFs 1102 * @pfq: queue index relative to the PF's function space 1103 * 1104 * If no VF is found who owns the pfq then return NULL, otherwise return a 1105 * pointer to the VF who owns the pfq 1106 * 1107 * If this function returns non-NULL, it acquires a reference count of the VF 1108 * structure. The caller is responsible for calling ice_put_vf() to drop this 1109 * reference. 1110 */ 1111static struct ice_vf *ice_get_vf_from_pfq(struct ice_pf *pf, u16 pfq) 1112{ 1113 struct ice_vf *vf; 1114 unsigned int bkt; 1115 1116 rcu_read_lock(); 1117 ice_for_each_vf_rcu(pf, bkt, vf) { 1118 struct ice_vsi *vsi; 1119 u16 rxq_idx; 1120 1121 vsi = ice_get_vf_vsi(vf); 1122 if (!vsi) 1123 continue; 1124 1125 ice_for_each_rxq(vsi, rxq_idx) 1126 if (vsi->rxq_map[rxq_idx] == pfq) { 1127 struct ice_vf *found; 1128 1129 if (kref_get_unless_zero(&vf->refcnt)) 1130 found = vf; 1131 else 1132 found = NULL; 1133 rcu_read_unlock(); 1134 return found; 1135 } 1136 } 1137 rcu_read_unlock(); 1138 1139 return NULL; 1140} 1141 1142/** 1143 * ice_globalq_to_pfq - convert from global queue index to PF space queue index 1144 * @pf: PF used for conversion 1145 * @globalq: global queue index used to convert to PF space queue index 1146 */ 1147static u32 ice_globalq_to_pfq(struct ice_pf *pf, u32 globalq) 1148{ 1149 return globalq - pf->hw.func_caps.common_cap.rxq_first_id; 1150} 1151 1152/** 1153 * ice_vf_lan_overflow_event - handle LAN overflow event for a VF 1154 * @pf: PF that the LAN overflow event happened on 1155 * @event: structure holding the event information for the LAN overflow event 1156 * 1157 * Determine if the LAN overflow event was caused by a VF queue. If it was not 1158 * caused by a VF, do nothing. If a VF caused this LAN overflow event trigger a 1159 * reset on the offending VF. 1160 */ 1161void 1162ice_vf_lan_overflow_event(struct ice_pf *pf, struct ice_rq_event_info *event) 1163{ 1164 struct ice_aqc_event_lan_overflow *cmd; 1165 u32 gldcb_rtctq, queue; 1166 struct ice_vf *vf; 1167 1168 cmd = libie_aq_raw(&event->desc); 1169 gldcb_rtctq = le32_to_cpu(cmd->prtdcb_ruptq); 1170 dev_dbg(ice_pf_to_dev(pf), "GLDCB_RTCTQ: 0x%08x\n", gldcb_rtctq); 1171 1172 /* event returns device global Rx queue number */ 1173 queue = FIELD_GET(GLDCB_RTCTQ_RXQNUM_M, gldcb_rtctq); 1174 1175 vf = ice_get_vf_from_pfq(pf, ice_globalq_to_pfq(pf, queue)); 1176 if (!vf) 1177 return; 1178 1179 ice_reset_vf(vf, ICE_VF_RESET_NOTIFY | ICE_VF_RESET_LOCK); 1180 ice_put_vf(vf); 1181} 1182 1183/** 1184 * ice_set_vf_spoofchk 1185 * @netdev: network interface device structure 1186 * @vf_id: VF identifier 1187 * @ena: flag to enable or disable feature 1188 * 1189 * Enable or disable VF spoof checking 1190 */ 1191int ice_set_vf_spoofchk(struct net_device *netdev, int vf_id, bool ena) 1192{ 1193 struct ice_pf *pf = ice_netdev_to_pf(netdev); 1194 struct ice_vsi *vf_vsi; 1195 struct device *dev; 1196 struct ice_vf *vf; 1197 int ret; 1198 1199 dev = ice_pf_to_dev(pf); 1200 1201 vf = ice_get_vf_by_id(pf, vf_id); 1202 if (!vf) 1203 return -EINVAL; 1204 1205 ret = ice_check_vf_ready_for_cfg(vf); 1206 if (ret) 1207 goto out_put_vf; 1208 1209 vf_vsi = ice_get_vf_vsi(vf); 1210 if (!vf_vsi) { 1211 netdev_err(netdev, "VSI %d for VF %d is null\n", 1212 vf->lan_vsi_idx, vf->vf_id); 1213 ret = -EINVAL; 1214 goto out_put_vf; 1215 } 1216 1217 if (vf_vsi->type != ICE_VSI_VF) { 1218 netdev_err(netdev, "Type %d of VSI %d for VF %d is no ICE_VSI_VF\n", 1219 vf_vsi->type, vf_vsi->vsi_num, vf->vf_id); 1220 ret = -ENODEV; 1221 goto out_put_vf; 1222 } 1223 1224 if (ena == vf->spoofchk) { 1225 dev_dbg(dev, "VF spoofchk already %s\n", ena ? "ON" : "OFF"); 1226 ret = 0; 1227 goto out_put_vf; 1228 } 1229 1230 ret = ice_vsi_apply_spoofchk(vf_vsi, ena); 1231 if (ret) 1232 dev_err(dev, "Failed to set spoofchk %s for VF %d VSI %d\n error %d\n", 1233 ena ? "ON" : "OFF", vf->vf_id, vf_vsi->vsi_num, ret); 1234 else 1235 vf->spoofchk = ena; 1236 1237out_put_vf: 1238 ice_put_vf(vf); 1239 return ret; 1240} 1241 1242/** 1243 * ice_get_vf_cfg 1244 * @netdev: network interface device structure 1245 * @vf_id: VF identifier 1246 * @ivi: VF configuration structure 1247 * 1248 * return VF configuration 1249 */ 1250int 1251ice_get_vf_cfg(struct net_device *netdev, int vf_id, struct ifla_vf_info *ivi) 1252{ 1253 struct ice_pf *pf = ice_netdev_to_pf(netdev); 1254 struct ice_vf *vf; 1255 int ret; 1256 1257 vf = ice_get_vf_by_id(pf, vf_id); 1258 if (!vf) 1259 return -EINVAL; 1260 1261 ret = ice_check_vf_ready_for_cfg(vf); 1262 if (ret) 1263 goto out_put_vf; 1264 1265 ivi->vf = vf_id; 1266 ether_addr_copy(ivi->mac, vf->hw_lan_addr); 1267 1268 /* VF configuration for VLAN and applicable QoS */ 1269 ivi->vlan = ice_vf_get_port_vlan_id(vf); 1270 ivi->qos = ice_vf_get_port_vlan_prio(vf); 1271 if (ice_vf_is_port_vlan_ena(vf)) 1272 ivi->vlan_proto = cpu_to_be16(ice_vf_get_port_vlan_tpid(vf)); 1273 1274 ivi->trusted = vf->trusted; 1275 ivi->spoofchk = vf->spoofchk; 1276 if (!vf->link_forced) 1277 ivi->linkstate = IFLA_VF_LINK_STATE_AUTO; 1278 else if (vf->link_up) 1279 ivi->linkstate = IFLA_VF_LINK_STATE_ENABLE; 1280 else 1281 ivi->linkstate = IFLA_VF_LINK_STATE_DISABLE; 1282 ivi->max_tx_rate = vf->max_tx_rate; 1283 ivi->min_tx_rate = vf->min_tx_rate; 1284 1285out_put_vf: 1286 ice_put_vf(vf); 1287 return ret; 1288} 1289 1290/** 1291 * __ice_set_vf_mac - program VF MAC address 1292 * @pf: PF to be configure 1293 * @vf_id: VF identifier 1294 * @mac: MAC address 1295 * 1296 * program VF MAC address 1297 * Return: zero on success or an error code on failure 1298 */ 1299int __ice_set_vf_mac(struct ice_pf *pf, u16 vf_id, const u8 *mac) 1300{ 1301 struct device *dev; 1302 struct ice_vf *vf; 1303 int ret; 1304 1305 dev = ice_pf_to_dev(pf); 1306 if (is_multicast_ether_addr(mac)) { 1307 dev_err(dev, "%pM not a valid unicast address\n", mac); 1308 return -EINVAL; 1309 } 1310 1311 vf = ice_get_vf_by_id(pf, vf_id); 1312 if (!vf) 1313 return -EINVAL; 1314 1315 /* nothing left to do, unicast MAC already set */ 1316 if (ether_addr_equal(vf->dev_lan_addr, mac) && 1317 ether_addr_equal(vf->hw_lan_addr, mac)) { 1318 ret = 0; 1319 goto out_put_vf; 1320 } 1321 1322 ret = ice_check_vf_ready_for_cfg(vf); 1323 if (ret) 1324 goto out_put_vf; 1325 1326 mutex_lock(&vf->cfg_lock); 1327 1328 /* VF is notified of its new MAC via the PF's response to the 1329 * VIRTCHNL_OP_GET_VF_RESOURCES message after the VF has been reset 1330 */ 1331 ether_addr_copy(vf->dev_lan_addr, mac); 1332 ether_addr_copy(vf->hw_lan_addr, mac); 1333 if (is_zero_ether_addr(mac)) { 1334 /* VF will send VIRTCHNL_OP_ADD_ETH_ADDR message with its MAC */ 1335 vf->pf_set_mac = false; 1336 dev_info(dev, "Removing MAC on VF %d. VF driver will be reinitialized\n", 1337 vf->vf_id); 1338 } else { 1339 /* PF will add MAC rule for the VF */ 1340 vf->pf_set_mac = true; 1341 dev_info(dev, "Setting MAC %pM on VF %d. VF driver will be reinitialized\n", 1342 mac, vf_id); 1343 } 1344 1345 ice_reset_vf(vf, ICE_VF_RESET_NOTIFY); 1346 mutex_unlock(&vf->cfg_lock); 1347 1348out_put_vf: 1349 ice_put_vf(vf); 1350 return ret; 1351} 1352 1353/** 1354 * ice_set_vf_mac - .ndo_set_vf_mac handler 1355 * @netdev: network interface device structure 1356 * @vf_id: VF identifier 1357 * @mac: MAC address 1358 * 1359 * program VF MAC address 1360 * Return: zero on success or an error code on failure 1361 */ 1362int ice_set_vf_mac(struct net_device *netdev, int vf_id, u8 *mac) 1363{ 1364 return __ice_set_vf_mac(ice_netdev_to_pf(netdev), vf_id, mac); 1365} 1366 1367/** 1368 * ice_set_vf_trust 1369 * @netdev: network interface device structure 1370 * @vf_id: VF identifier 1371 * @trusted: Boolean value to enable/disable trusted VF 1372 * 1373 * Enable or disable a given VF as trusted 1374 */ 1375int ice_set_vf_trust(struct net_device *netdev, int vf_id, bool trusted) 1376{ 1377 struct ice_pf *pf = ice_netdev_to_pf(netdev); 1378 struct ice_vf *vf; 1379 int ret; 1380 1381 vf = ice_get_vf_by_id(pf, vf_id); 1382 if (!vf) 1383 return -EINVAL; 1384 1385 if (ice_is_eswitch_mode_switchdev(pf)) { 1386 dev_info(ice_pf_to_dev(pf), "Trusted VF is forbidden in switchdev mode\n"); 1387 return -EOPNOTSUPP; 1388 } 1389 1390 ret = ice_check_vf_ready_for_cfg(vf); 1391 if (ret) 1392 goto out_put_vf; 1393 1394 /* Check if already trusted */ 1395 if (trusted == vf->trusted) { 1396 ret = 0; 1397 goto out_put_vf; 1398 } 1399 1400 mutex_lock(&vf->cfg_lock); 1401 1402 while (!trusted && vf->num_mac_lldp) 1403 ice_vf_update_mac_lldp_num(vf, ice_get_vf_vsi(vf), false); 1404 1405 vf->trusted = trusted; 1406 ice_reset_vf(vf, ICE_VF_RESET_NOTIFY); 1407 dev_info(ice_pf_to_dev(pf), "VF %u is now %strusted\n", 1408 vf_id, trusted ? "" : "un"); 1409 1410 mutex_unlock(&vf->cfg_lock); 1411 1412out_put_vf: 1413 ice_put_vf(vf); 1414 return ret; 1415} 1416 1417/** 1418 * ice_set_vf_link_state 1419 * @netdev: network interface device structure 1420 * @vf_id: VF identifier 1421 * @link_state: required link state 1422 * 1423 * Set VF's link state, irrespective of physical link state status 1424 */ 1425int ice_set_vf_link_state(struct net_device *netdev, int vf_id, int link_state) 1426{ 1427 struct ice_pf *pf = ice_netdev_to_pf(netdev); 1428 struct ice_vf *vf; 1429 int ret; 1430 1431 vf = ice_get_vf_by_id(pf, vf_id); 1432 if (!vf) 1433 return -EINVAL; 1434 1435 ret = ice_check_vf_ready_for_cfg(vf); 1436 if (ret) 1437 goto out_put_vf; 1438 1439 switch (link_state) { 1440 case IFLA_VF_LINK_STATE_AUTO: 1441 vf->link_forced = false; 1442 break; 1443 case IFLA_VF_LINK_STATE_ENABLE: 1444 vf->link_forced = true; 1445 vf->link_up = true; 1446 break; 1447 case IFLA_VF_LINK_STATE_DISABLE: 1448 vf->link_forced = true; 1449 vf->link_up = false; 1450 break; 1451 default: 1452 ret = -EINVAL; 1453 goto out_put_vf; 1454 } 1455 1456 ice_vc_notify_vf_link_state(vf); 1457 1458out_put_vf: 1459 ice_put_vf(vf); 1460 return ret; 1461} 1462 1463/** 1464 * ice_calc_all_vfs_min_tx_rate - calculate cumulative min Tx rate on all VFs 1465 * @pf: PF associated with VFs 1466 */ 1467static int ice_calc_all_vfs_min_tx_rate(struct ice_pf *pf) 1468{ 1469 struct ice_vf *vf; 1470 unsigned int bkt; 1471 int rate = 0; 1472 1473 rcu_read_lock(); 1474 ice_for_each_vf_rcu(pf, bkt, vf) 1475 rate += vf->min_tx_rate; 1476 rcu_read_unlock(); 1477 1478 return rate; 1479} 1480 1481/** 1482 * ice_min_tx_rate_oversubscribed - check if min Tx rate causes oversubscription 1483 * @vf: VF trying to configure min_tx_rate 1484 * @min_tx_rate: min Tx rate in Mbps 1485 * 1486 * Check if the min_tx_rate being passed in will cause oversubscription of total 1487 * min_tx_rate based on the current link speed and all other VFs configured 1488 * min_tx_rate 1489 * 1490 * Return true if the passed min_tx_rate would cause oversubscription, else 1491 * return false 1492 */ 1493static bool 1494ice_min_tx_rate_oversubscribed(struct ice_vf *vf, int min_tx_rate) 1495{ 1496 struct ice_vsi *vsi = ice_get_vf_vsi(vf); 1497 int all_vfs_min_tx_rate; 1498 int link_speed_mbps; 1499 1500 if (WARN_ON(!vsi)) 1501 return false; 1502 1503 link_speed_mbps = ice_get_link_speed_mbps(vsi); 1504 all_vfs_min_tx_rate = ice_calc_all_vfs_min_tx_rate(vf->pf); 1505 1506 /* this VF's previous rate is being overwritten */ 1507 all_vfs_min_tx_rate -= vf->min_tx_rate; 1508 1509 if (all_vfs_min_tx_rate + min_tx_rate > link_speed_mbps) { 1510 dev_err(ice_pf_to_dev(vf->pf), "min_tx_rate of %d Mbps on VF %u would cause oversubscription of %d Mbps based on the current link speed %d Mbps\n", 1511 min_tx_rate, vf->vf_id, 1512 all_vfs_min_tx_rate + min_tx_rate - link_speed_mbps, 1513 link_speed_mbps); 1514 return true; 1515 } 1516 1517 return false; 1518} 1519 1520/** 1521 * ice_set_vf_bw - set min/max VF bandwidth 1522 * @netdev: network interface device structure 1523 * @vf_id: VF identifier 1524 * @min_tx_rate: Minimum Tx rate in Mbps 1525 * @max_tx_rate: Maximum Tx rate in Mbps 1526 */ 1527int 1528ice_set_vf_bw(struct net_device *netdev, int vf_id, int min_tx_rate, 1529 int max_tx_rate) 1530{ 1531 struct ice_pf *pf = ice_netdev_to_pf(netdev); 1532 struct ice_vsi *vsi; 1533 struct device *dev; 1534 struct ice_vf *vf; 1535 int ret; 1536 1537 dev = ice_pf_to_dev(pf); 1538 1539 vf = ice_get_vf_by_id(pf, vf_id); 1540 if (!vf) 1541 return -EINVAL; 1542 1543 ret = ice_check_vf_ready_for_cfg(vf); 1544 if (ret) 1545 goto out_put_vf; 1546 1547 vsi = ice_get_vf_vsi(vf); 1548 if (!vsi) { 1549 ret = -EINVAL; 1550 goto out_put_vf; 1551 } 1552 1553 if (min_tx_rate && ice_is_dcb_active(pf)) { 1554 dev_err(dev, "DCB on PF is currently enabled. VF min Tx rate limiting not allowed on this PF.\n"); 1555 ret = -EOPNOTSUPP; 1556 goto out_put_vf; 1557 } 1558 1559 if (ice_min_tx_rate_oversubscribed(vf, min_tx_rate)) { 1560 ret = -EINVAL; 1561 goto out_put_vf; 1562 } 1563 1564 if (vf->min_tx_rate != (unsigned int)min_tx_rate) { 1565 ret = ice_set_min_bw_limit(vsi, (u64)min_tx_rate * 1000); 1566 if (ret) { 1567 dev_err(dev, "Unable to set min-tx-rate for VF %d\n", 1568 vf->vf_id); 1569 goto out_put_vf; 1570 } 1571 1572 vf->min_tx_rate = min_tx_rate; 1573 } 1574 1575 if (vf->max_tx_rate != (unsigned int)max_tx_rate) { 1576 ret = ice_set_max_bw_limit(vsi, (u64)max_tx_rate * 1000); 1577 if (ret) { 1578 dev_err(dev, "Unable to set max-tx-rate for VF %d\n", 1579 vf->vf_id); 1580 goto out_put_vf; 1581 } 1582 1583 vf->max_tx_rate = max_tx_rate; 1584 } 1585 1586out_put_vf: 1587 ice_put_vf(vf); 1588 return ret; 1589} 1590 1591/** 1592 * ice_get_vf_stats - populate some stats for the VF 1593 * @netdev: the netdev of the PF 1594 * @vf_id: the host OS identifier (0-255) 1595 * @vf_stats: pointer to the OS memory to be initialized 1596 */ 1597int ice_get_vf_stats(struct net_device *netdev, int vf_id, 1598 struct ifla_vf_stats *vf_stats) 1599{ 1600 struct ice_pf *pf = ice_netdev_to_pf(netdev); 1601 struct ice_eth_stats *stats; 1602 struct ice_vsi *vsi; 1603 struct ice_vf *vf; 1604 int ret; 1605 1606 vf = ice_get_vf_by_id(pf, vf_id); 1607 if (!vf) 1608 return -EINVAL; 1609 1610 ret = ice_check_vf_ready_for_cfg(vf); 1611 if (ret) 1612 goto out_put_vf; 1613 1614 vsi = ice_get_vf_vsi(vf); 1615 if (!vsi) { 1616 ret = -EINVAL; 1617 goto out_put_vf; 1618 } 1619 1620 ice_update_eth_stats(vsi); 1621 stats = &vsi->eth_stats; 1622 1623 memset(vf_stats, 0, sizeof(*vf_stats)); 1624 1625 vf_stats->rx_packets = stats->rx_unicast + stats->rx_broadcast + 1626 stats->rx_multicast; 1627 vf_stats->tx_packets = stats->tx_unicast + stats->tx_broadcast + 1628 stats->tx_multicast; 1629 vf_stats->rx_bytes = stats->rx_bytes; 1630 vf_stats->tx_bytes = stats->tx_bytes; 1631 vf_stats->broadcast = stats->rx_broadcast; 1632 vf_stats->multicast = stats->rx_multicast; 1633 vf_stats->rx_dropped = stats->rx_discards; 1634 vf_stats->tx_dropped = stats->tx_discards; 1635 1636out_put_vf: 1637 ice_put_vf(vf); 1638 return ret; 1639} 1640 1641/** 1642 * ice_is_supported_port_vlan_proto - make sure the vlan_proto is supported 1643 * @hw: hardware structure used to check the VLAN mode 1644 * @vlan_proto: VLAN TPID being checked 1645 * 1646 * If the device is configured in Double VLAN Mode (DVM), then both ETH_P_8021Q 1647 * and ETH_P_8021AD are supported. If the device is configured in Single VLAN 1648 * Mode (SVM), then only ETH_P_8021Q is supported. 1649 */ 1650static bool 1651ice_is_supported_port_vlan_proto(struct ice_hw *hw, u16 vlan_proto) 1652{ 1653 bool is_supported = false; 1654 1655 switch (vlan_proto) { 1656 case ETH_P_8021Q: 1657 is_supported = true; 1658 break; 1659 case ETH_P_8021AD: 1660 if (ice_is_dvm_ena(hw)) 1661 is_supported = true; 1662 break; 1663 } 1664 1665 return is_supported; 1666} 1667 1668/** 1669 * ice_set_vf_port_vlan 1670 * @netdev: network interface device structure 1671 * @vf_id: VF identifier 1672 * @vlan_id: VLAN ID being set 1673 * @qos: priority setting 1674 * @vlan_proto: VLAN protocol 1675 * 1676 * program VF Port VLAN ID and/or QoS 1677 */ 1678int 1679ice_set_vf_port_vlan(struct net_device *netdev, int vf_id, u16 vlan_id, u8 qos, 1680 __be16 vlan_proto) 1681{ 1682 struct ice_pf *pf = ice_netdev_to_pf(netdev); 1683 u16 local_vlan_proto = ntohs(vlan_proto); 1684 struct device *dev; 1685 struct ice_vf *vf; 1686 int ret; 1687 1688 dev = ice_pf_to_dev(pf); 1689 1690 if (vlan_id >= VLAN_N_VID || qos > 7) { 1691 dev_err(dev, "Invalid Port VLAN parameters for VF %d, ID %d, QoS %d\n", 1692 vf_id, vlan_id, qos); 1693 return -EINVAL; 1694 } 1695 1696 if (!ice_is_supported_port_vlan_proto(&pf->hw, local_vlan_proto)) { 1697 dev_err(dev, "VF VLAN protocol 0x%04x is not supported\n", 1698 local_vlan_proto); 1699 return -EPROTONOSUPPORT; 1700 } 1701 1702 vf = ice_get_vf_by_id(pf, vf_id); 1703 if (!vf) 1704 return -EINVAL; 1705 1706 ret = ice_check_vf_ready_for_cfg(vf); 1707 if (ret) 1708 goto out_put_vf; 1709 1710 if (ice_vf_get_port_vlan_prio(vf) == qos && 1711 ice_vf_get_port_vlan_tpid(vf) == local_vlan_proto && 1712 ice_vf_get_port_vlan_id(vf) == vlan_id) { 1713 /* duplicate request, so just return success */ 1714 dev_dbg(dev, "Duplicate port VLAN %u, QoS %u, TPID 0x%04x request\n", 1715 vlan_id, qos, local_vlan_proto); 1716 ret = 0; 1717 goto out_put_vf; 1718 } 1719 1720 mutex_lock(&vf->cfg_lock); 1721 1722 vf->port_vlan_info = ICE_VLAN(local_vlan_proto, vlan_id, qos); 1723 if (ice_vf_is_port_vlan_ena(vf)) 1724 dev_info(dev, "Setting VLAN %u, QoS %u, TPID 0x%04x on VF %d\n", 1725 vlan_id, qos, local_vlan_proto, vf_id); 1726 else 1727 dev_info(dev, "Clearing port VLAN on VF %d\n", vf_id); 1728 1729 ice_reset_vf(vf, ICE_VF_RESET_NOTIFY); 1730 mutex_unlock(&vf->cfg_lock); 1731 1732out_put_vf: 1733 ice_put_vf(vf); 1734 return ret; 1735} 1736 1737/** 1738 * ice_print_vf_rx_mdd_event - print VF Rx malicious driver detect event 1739 * @vf: pointer to the VF structure 1740 */ 1741void ice_print_vf_rx_mdd_event(struct ice_vf *vf) 1742{ 1743 struct ice_pf *pf = vf->pf; 1744 struct device *dev; 1745 1746 dev = ice_pf_to_dev(pf); 1747 1748 dev_info(dev, "%d Rx Malicious Driver Detection events detected on PF %d VF %d MAC %pM. mdd-auto-reset-vfs=%s\n", 1749 vf->mdd_rx_events.count, pf->hw.pf_id, vf->vf_id, 1750 vf->dev_lan_addr, 1751 test_bit(ICE_FLAG_MDD_AUTO_RESET_VF, pf->flags) 1752 ? "on" : "off"); 1753} 1754 1755/** 1756 * ice_print_vf_tx_mdd_event - print VF Tx malicious driver detect event 1757 * @vf: pointer to the VF structure 1758 */ 1759void ice_print_vf_tx_mdd_event(struct ice_vf *vf) 1760{ 1761 struct ice_pf *pf = vf->pf; 1762 struct device *dev; 1763 1764 dev = ice_pf_to_dev(pf); 1765 1766 dev_info(dev, "%d Tx Malicious Driver Detection events detected on PF %d VF %d MAC %pM. mdd-auto-reset-vfs=%s\n", 1767 vf->mdd_tx_events.count, pf->hw.pf_id, vf->vf_id, 1768 vf->dev_lan_addr, 1769 test_bit(ICE_FLAG_MDD_AUTO_RESET_VF, pf->flags) 1770 ? "on" : "off"); 1771} 1772 1773/** 1774 * ice_print_vfs_mdd_events - print VFs malicious driver detect event 1775 * @pf: pointer to the PF structure 1776 * 1777 * Called from ice_handle_mdd_event to rate limit and print VFs MDD events. 1778 */ 1779void ice_print_vfs_mdd_events(struct ice_pf *pf) 1780{ 1781 struct ice_vf *vf; 1782 unsigned int bkt; 1783 1784 /* check that there are pending MDD events to print */ 1785 if (!test_and_clear_bit(ICE_MDD_VF_PRINT_PENDING, pf->state)) 1786 return; 1787 1788 /* VF MDD event logs are rate limited to one second intervals */ 1789 if (time_is_after_jiffies(pf->vfs.last_printed_mdd_jiffies + HZ * 1)) 1790 return; 1791 1792 pf->vfs.last_printed_mdd_jiffies = jiffies; 1793 1794 mutex_lock(&pf->vfs.table_lock); 1795 ice_for_each_vf(pf, bkt, vf) { 1796 /* only print Rx MDD event message if there are new events */ 1797 if (vf->mdd_rx_events.count != vf->mdd_rx_events.last_printed) { 1798 vf->mdd_rx_events.last_printed = 1799 vf->mdd_rx_events.count; 1800 ice_print_vf_rx_mdd_event(vf); 1801 } 1802 1803 /* only print Tx MDD event message if there are new events */ 1804 if (vf->mdd_tx_events.count != vf->mdd_tx_events.last_printed) { 1805 vf->mdd_tx_events.last_printed = 1806 vf->mdd_tx_events.count; 1807 ice_print_vf_tx_mdd_event(vf); 1808 } 1809 } 1810 mutex_unlock(&pf->vfs.table_lock); 1811} 1812 1813/** 1814 * ice_restore_all_vfs_msi_state - restore VF MSI state after PF FLR 1815 * @pf: pointer to the PF structure 1816 * 1817 * Called when recovering from a PF FLR to restore interrupt capability to 1818 * the VFs. 1819 */ 1820void ice_restore_all_vfs_msi_state(struct ice_pf *pf) 1821{ 1822 struct ice_vf *vf; 1823 u32 bkt; 1824 1825 ice_for_each_vf(pf, bkt, vf) 1826 pci_restore_msi_state(vf->vfdev); 1827}
~/
Read-only file preview.