1 // SPDX-License-Identifier: GPL-2.0
2 /* Copyright (c) 2018, Intel Corporation. */
5 #include "ice_vf_lib_private.h"
9 #include "ice_dcb_lib.h"
11 #include "ice_eswitch.h"
12 #include "ice_virtchnl_allowlist.h"
13 #include "ice_flex_pipe.h"
14 #include "ice_vf_vsi_vlan_ops.h"
18 * ice_free_vf_entries - Free all VF entries from the hash table
19 * @pf: pointer to the PF structure
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.
24 static void ice_free_vf_entries(struct ice_pf *pf)
26 struct ice_vfs *vfs = &pf->vfs;
27 struct hlist_node *tmp;
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.
35 lockdep_assert_held(&vfs->table_lock);
37 hash_for_each_safe(vfs->table, bkt, tmp, vf, entry) {
38 hash_del_rcu(&vf->entry);
44 * ice_free_vf_res - Free a VF's resources
45 * @vf: pointer to the VF info
47 static void ice_free_vf_res(struct ice_vf *vf)
49 struct ice_pf *pf = vf->pf;
50 int i, last_vector_idx;
52 /* First, disable VF's configuration API to prevent OS from
53 * accessing the VF's VSI after it's freed or invalidated.
55 clear_bit(ICE_VF_STATE_INIT, vf->vf_states);
57 /* free VF control VSI */
58 if (vf->ctrl_vsi_idx != ICE_NO_VSI)
59 ice_vf_ctrl_vsi_release(vf);
61 /* free VSI and disconnect it from the parent uplink */
62 if (vf->lan_vsi_idx != ICE_NO_VSI) {
63 ice_vf_vsi_release(vf);
67 last_vector_idx = vf->first_vector_idx + vf->num_msix - 1;
69 /* clear VF MDD event information */
70 memset(&vf->mdd_tx_events, 0, sizeof(vf->mdd_tx_events));
71 memset(&vf->mdd_rx_events, 0, sizeof(vf->mdd_rx_events));
73 /* Disable interrupts so that VF starts in a known state */
74 for (i = vf->first_vector_idx; i <= last_vector_idx; i++) {
75 wr32(&pf->hw, GLINT_DYN_CTL(i), GLINT_DYN_CTL_CLEARPBA_M);
78 /* reset some of the state variables keeping track of the resources */
79 clear_bit(ICE_VF_STATE_MC_PROMISC, vf->vf_states);
80 clear_bit(ICE_VF_STATE_UC_PROMISC, vf->vf_states);
85 * @vf: pointer to the VF structure
87 static void ice_dis_vf_mappings(struct ice_vf *vf)
89 struct ice_pf *pf = vf->pf;
96 vsi = ice_get_vf_vsi(vf);
100 dev = ice_pf_to_dev(pf);
101 wr32(hw, VPINT_ALLOC(vf->vf_id), 0);
102 wr32(hw, VPINT_ALLOC_PCI(vf->vf_id), 0);
104 first = vf->first_vector_idx;
105 last = first + vf->num_msix - 1;
106 for (v = first; v <= last; v++) {
109 reg = FIELD_PREP(GLINT_VECT2FUNC_IS_PF_M, 1) |
110 FIELD_PREP(GLINT_VECT2FUNC_PF_NUM_M, hw->pf_id);
111 wr32(hw, GLINT_VECT2FUNC(v), reg);
114 if (vsi->tx_mapping_mode == ICE_VSI_MAP_CONTIG)
115 wr32(hw, VPLAN_TX_QBASE(vf->vf_id), 0);
117 dev_err(dev, "Scattered mode for VF Tx queues is not yet implemented\n");
119 if (vsi->rx_mapping_mode == ICE_VSI_MAP_CONTIG)
120 wr32(hw, VPLAN_RX_QBASE(vf->vf_id), 0);
122 dev_err(dev, "Scattered mode for VF Rx queues is not yet implemented\n");
126 * ice_sriov_free_msix_res - Reset/free any used MSIX resources
127 * @pf: pointer to the PF structure
129 * Since no MSIX entries are taken from the pf->irq_tracker then just clear
130 * the pf->sriov_base_vector.
132 * Returns 0 on success, and -EINVAL on error.
134 static int ice_sriov_free_msix_res(struct ice_pf *pf)
139 bitmap_free(pf->sriov_irq_bm);
140 pf->sriov_irq_size = 0;
141 pf->sriov_base_vector = 0;
147 * ice_free_vfs - Free all VFs
148 * @pf: pointer to the PF structure
150 void ice_free_vfs(struct ice_pf *pf)
152 struct device *dev = ice_pf_to_dev(pf);
153 struct ice_vfs *vfs = &pf->vfs;
154 struct ice_hw *hw = &pf->hw;
158 if (!ice_has_vfs(pf))
161 while (test_and_set_bit(ICE_VF_DIS, pf->state))
162 usleep_range(1000, 2000);
164 /* Disable IOV before freeing resources. This lets any VF drivers
165 * running in the host get themselves cleaned up before we yank
166 * the carpet out from underneath their feet.
168 if (!pci_vfs_assigned(pf->pdev))
169 pci_disable_sriov(pf->pdev);
171 dev_warn(dev, "VFs are assigned - not disabling SR-IOV\n");
173 mutex_lock(&vfs->table_lock);
175 ice_for_each_vf(pf, bkt, vf) {
176 mutex_lock(&vf->cfg_lock);
178 ice_eswitch_detach(pf, vf);
181 if (test_bit(ICE_VF_STATE_INIT, vf->vf_states)) {
182 /* disable VF qp mappings and set VF disable state */
183 ice_dis_vf_mappings(vf);
184 set_bit(ICE_VF_STATE_DIS, vf->vf_states);
188 if (!pci_vfs_assigned(pf->pdev)) {
189 u32 reg_idx, bit_idx;
191 reg_idx = (hw->func_caps.vf_base_id + vf->vf_id) / 32;
192 bit_idx = (hw->func_caps.vf_base_id + vf->vf_id) % 32;
193 wr32(hw, GLGEN_VFLRSTAT(reg_idx), BIT(bit_idx));
196 /* clear malicious info since the VF is getting released */
197 list_del(&vf->mbx_info.list_entry);
199 mutex_unlock(&vf->cfg_lock);
202 if (ice_sriov_free_msix_res(pf))
203 dev_err(dev, "Failed to free MSIX resources used by SR-IOV\n");
205 vfs->num_qps_per = 0;
206 ice_free_vf_entries(pf);
208 mutex_unlock(&vfs->table_lock);
210 clear_bit(ICE_VF_DIS, pf->state);
211 clear_bit(ICE_FLAG_SRIOV_ENA, pf->flags);
215 * ice_vf_vsi_setup - Set up a VF VSI
216 * @vf: VF to setup VSI for
218 * Returns pointer to the successfully allocated VSI struct on success,
219 * otherwise returns NULL on failure.
221 static struct ice_vsi *ice_vf_vsi_setup(struct ice_vf *vf)
223 struct ice_vsi_cfg_params params = {};
224 struct ice_pf *pf = vf->pf;
227 params.type = ICE_VSI_VF;
228 params.pi = ice_vf_get_port_info(vf);
230 params.flags = ICE_VSI_FLAG_INIT;
232 vsi = ice_vsi_setup(pf, ¶ms);
235 dev_err(ice_pf_to_dev(pf), "Failed to create VF VSI\n");
236 ice_vf_invalidate_vsi(vf);
240 vf->lan_vsi_idx = vsi->idx;
247 * ice_ena_vf_msix_mappings - enable VF MSIX mappings in hardware
248 * @vf: VF to enable MSIX mappings for
250 * Some of the registers need to be indexed/configured using hardware global
251 * device values and other registers need 0-based values, which represent PF
254 static void ice_ena_vf_msix_mappings(struct ice_vf *vf)
256 int device_based_first_msix, device_based_last_msix;
257 int pf_based_first_msix, pf_based_last_msix, v;
258 struct ice_pf *pf = vf->pf;
259 int device_based_vf_id;
264 pf_based_first_msix = vf->first_vector_idx;
265 pf_based_last_msix = (pf_based_first_msix + vf->num_msix) - 1;
267 device_based_first_msix = pf_based_first_msix +
268 pf->hw.func_caps.common_cap.msix_vector_first_id;
269 device_based_last_msix =
270 (device_based_first_msix + vf->num_msix) - 1;
271 device_based_vf_id = vf->vf_id + hw->func_caps.vf_base_id;
273 reg = FIELD_PREP(VPINT_ALLOC_FIRST_M, device_based_first_msix) |
274 FIELD_PREP(VPINT_ALLOC_LAST_M, device_based_last_msix) |
276 wr32(hw, VPINT_ALLOC(vf->vf_id), reg);
278 reg = FIELD_PREP(VPINT_ALLOC_PCI_FIRST_M, device_based_first_msix) |
279 FIELD_PREP(VPINT_ALLOC_PCI_LAST_M, device_based_last_msix) |
280 VPINT_ALLOC_PCI_VALID_M;
281 wr32(hw, VPINT_ALLOC_PCI(vf->vf_id), reg);
283 /* map the interrupts to its functions */
284 for (v = pf_based_first_msix; v <= pf_based_last_msix; v++) {
285 reg = FIELD_PREP(GLINT_VECT2FUNC_VF_NUM_M, device_based_vf_id) |
286 FIELD_PREP(GLINT_VECT2FUNC_PF_NUM_M, hw->pf_id);
287 wr32(hw, GLINT_VECT2FUNC(v), reg);
290 /* Map mailbox interrupt to VF MSI-X vector 0 */
291 wr32(hw, VPINT_MBX_CTL(device_based_vf_id), VPINT_MBX_CTL_CAUSE_ENA_M);
295 * ice_ena_vf_q_mappings - enable Rx/Tx queue mappings for a VF
296 * @vf: VF to enable the mappings for
297 * @max_txq: max Tx queues allowed on the VF's VSI
298 * @max_rxq: max Rx queues allowed on the VF's VSI
300 static void ice_ena_vf_q_mappings(struct ice_vf *vf, u16 max_txq, u16 max_rxq)
302 struct device *dev = ice_pf_to_dev(vf->pf);
303 struct ice_vsi *vsi = ice_get_vf_vsi(vf);
304 struct ice_hw *hw = &vf->pf->hw;
310 /* set regardless of mapping mode */
311 wr32(hw, VPLAN_TXQ_MAPENA(vf->vf_id), VPLAN_TXQ_MAPENA_TX_ENA_M);
313 /* VF Tx queues allocation */
314 if (vsi->tx_mapping_mode == ICE_VSI_MAP_CONTIG) {
315 /* set the VF PF Tx queue range
316 * VFNUMQ value should be set to (number of queues - 1). A value
317 * of 0 means 1 queue and a value of 255 means 256 queues
319 reg = FIELD_PREP(VPLAN_TX_QBASE_VFFIRSTQ_M, vsi->txq_map[0]) |
320 FIELD_PREP(VPLAN_TX_QBASE_VFNUMQ_M, max_txq - 1);
321 wr32(hw, VPLAN_TX_QBASE(vf->vf_id), reg);
323 dev_err(dev, "Scattered mode for VF Tx queues is not yet implemented\n");
326 /* set regardless of mapping mode */
327 wr32(hw, VPLAN_RXQ_MAPENA(vf->vf_id), VPLAN_RXQ_MAPENA_RX_ENA_M);
329 /* VF Rx queues allocation */
330 if (vsi->rx_mapping_mode == ICE_VSI_MAP_CONTIG) {
331 /* set the VF PF Rx queue range
332 * VFNUMQ value should be set to (number of queues - 1). A value
333 * of 0 means 1 queue and a value of 255 means 256 queues
335 reg = FIELD_PREP(VPLAN_RX_QBASE_VFFIRSTQ_M, vsi->rxq_map[0]) |
336 FIELD_PREP(VPLAN_RX_QBASE_VFNUMQ_M, max_rxq - 1);
337 wr32(hw, VPLAN_RX_QBASE(vf->vf_id), reg);
339 dev_err(dev, "Scattered mode for VF Rx queues is not yet implemented\n");
344 * ice_ena_vf_mappings - enable VF MSIX and queue mapping
345 * @vf: pointer to the VF structure
347 static void ice_ena_vf_mappings(struct ice_vf *vf)
349 struct ice_vsi *vsi = ice_get_vf_vsi(vf);
354 ice_ena_vf_msix_mappings(vf);
355 ice_ena_vf_q_mappings(vf, vsi->alloc_txq, vsi->alloc_rxq);
359 * ice_calc_vf_reg_idx - Calculate the VF's register index in the PF space
360 * @vf: VF to calculate the register index for
361 * @q_vector: a q_vector associated to the VF
363 int ice_calc_vf_reg_idx(struct ice_vf *vf, struct ice_q_vector *q_vector)
365 if (!vf || !q_vector)
368 /* always add one to account for the OICR being the first MSIX */
369 return vf->first_vector_idx + q_vector->v_idx + 1;
373 * ice_sriov_set_msix_res - Set any used MSIX resources
374 * @pf: pointer to PF structure
375 * @num_msix_needed: number of MSIX vectors needed for all SR-IOV VFs
377 * This function allows SR-IOV resources to be taken from the end of the PF's
378 * allowed HW MSIX vectors so that the irq_tracker will not be affected. We
379 * just set the pf->sriov_base_vector and return success.
381 * If there are not enough resources available, return an error. This should
382 * always be caught by ice_set_per_vf_res().
384 * Return 0 on success, and -EINVAL when there are not enough MSIX vectors
385 * in the PF's space available for SR-IOV.
387 static int ice_sriov_set_msix_res(struct ice_pf *pf, u16 num_msix_needed)
389 u16 total_vectors = pf->hw.func_caps.common_cap.num_msix_vectors;
390 int vectors_used = ice_get_max_used_msix_vector(pf);
391 int sriov_base_vector;
393 sriov_base_vector = total_vectors - num_msix_needed;
395 /* make sure we only grab irq_tracker entries from the list end and
396 * that we have enough available MSIX vectors
398 if (sriov_base_vector < vectors_used)
401 pf->sriov_base_vector = sriov_base_vector;
407 * ice_set_per_vf_res - check if vectors and queues are available
408 * @pf: pointer to the PF structure
409 * @num_vfs: the number of SR-IOV VFs being configured
411 * First, determine HW interrupts from common pool. If we allocate fewer VFs, we
412 * get more vectors and can enable more queues per VF. Note that this does not
413 * grab any vectors from the SW pool already allocated. Also note, that all
414 * vector counts include one for each VF's miscellaneous interrupt vector
417 * Minimum VFs - 2 vectors, 1 queue pair
418 * Small VFs - 5 vectors, 4 queue pairs
419 * Medium VFs - 17 vectors, 16 queue pairs
421 * Second, determine number of queue pairs per VF by starting with a pre-defined
422 * maximum each VF supports. If this is not possible, then we adjust based on
423 * queue pairs available on the device.
425 * Lastly, set queue and MSI-X VF variables tracked by the PF so it can be used
426 * by each VF during VF initialization and reset.
428 static int ice_set_per_vf_res(struct ice_pf *pf, u16 num_vfs)
430 int vectors_used = ice_get_max_used_msix_vector(pf);
431 u16 num_msix_per_vf, num_txq, num_rxq, avail_qs;
432 int msix_avail_per_vf, msix_avail_for_sriov;
433 struct device *dev = ice_pf_to_dev(pf);
436 lockdep_assert_held(&pf->vfs.table_lock);
441 /* determine MSI-X resources per VF */
442 msix_avail_for_sriov = pf->hw.func_caps.common_cap.num_msix_vectors -
444 msix_avail_per_vf = msix_avail_for_sriov / num_vfs;
445 if (msix_avail_per_vf >= ICE_NUM_VF_MSIX_MED) {
446 num_msix_per_vf = ICE_NUM_VF_MSIX_MED;
447 } else if (msix_avail_per_vf >= ICE_NUM_VF_MSIX_SMALL) {
448 num_msix_per_vf = ICE_NUM_VF_MSIX_SMALL;
449 } else if (msix_avail_per_vf >= ICE_NUM_VF_MSIX_MULTIQ_MIN) {
450 num_msix_per_vf = ICE_NUM_VF_MSIX_MULTIQ_MIN;
451 } else if (msix_avail_per_vf >= ICE_MIN_INTR_PER_VF) {
452 num_msix_per_vf = ICE_MIN_INTR_PER_VF;
454 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",
455 msix_avail_for_sriov, ICE_MIN_INTR_PER_VF,
460 num_txq = min_t(u16, num_msix_per_vf - ICE_NONQ_VECS_VF,
461 ICE_MAX_RSS_QS_PER_VF);
462 avail_qs = ice_get_avail_txq_count(pf) / num_vfs;
465 else if (num_txq > avail_qs)
466 num_txq = rounddown_pow_of_two(avail_qs);
468 num_rxq = min_t(u16, num_msix_per_vf - ICE_NONQ_VECS_VF,
469 ICE_MAX_RSS_QS_PER_VF);
470 avail_qs = ice_get_avail_rxq_count(pf) / num_vfs;
473 else if (num_rxq > avail_qs)
474 num_rxq = rounddown_pow_of_two(avail_qs);
476 if (num_txq < ICE_MIN_QS_PER_VF || num_rxq < ICE_MIN_QS_PER_VF) {
477 dev_err(dev, "Not enough queues to support minimum of %d queue pairs per VF for %d VFs\n",
478 ICE_MIN_QS_PER_VF, num_vfs);
482 err = ice_sriov_set_msix_res(pf, num_msix_per_vf * num_vfs);
484 dev_err(dev, "Unable to set MSI-X resources for %d VFs, err %d\n",
489 /* only allow equal Tx/Rx queue count (i.e. queue pairs) */
490 pf->vfs.num_qps_per = min_t(int, num_txq, num_rxq);
491 pf->vfs.num_msix_per = num_msix_per_vf;
492 dev_info(dev, "Enabling %d VFs with %d vectors and %d queues per VF\n",
493 num_vfs, pf->vfs.num_msix_per, pf->vfs.num_qps_per);
499 * ice_sriov_get_irqs - get irqs for SR-IOV usacase
500 * @pf: pointer to PF structure
501 * @needed: number of irqs to get
503 * This returns the first MSI-X vector index in PF space that is used by this
504 * VF. This index is used when accessing PF relative registers such as
505 * GLINT_VECT2FUNC and GLINT_DYN_CTL.
506 * This will always be the OICR index in the AVF driver so any functionality
507 * using vf->first_vector_idx for queue configuration_id: id of VF which will
510 * Only SRIOV specific vectors are tracked in sriov_irq_bm. SRIOV vectors are
511 * allocated from the end of global irq index. First bit in sriov_irq_bm means
512 * last irq index etc. It simplifies extension of SRIOV vectors.
513 * They will be always located from sriov_base_vector to the last irq
514 * index. While increasing/decreasing sriov_base_vector can be moved.
516 static int ice_sriov_get_irqs(struct ice_pf *pf, u16 needed)
518 int res = bitmap_find_next_zero_area(pf->sriov_irq_bm,
519 pf->sriov_irq_size, 0, needed, 0);
520 /* conversion from number in bitmap to global irq index */
521 int index = pf->sriov_irq_size - res - needed;
523 if (res >= pf->sriov_irq_size || index < pf->sriov_base_vector)
526 bitmap_set(pf->sriov_irq_bm, res, needed);
531 * ice_sriov_free_irqs - free irqs used by the VF
532 * @pf: pointer to PF structure
533 * @vf: pointer to VF structure
535 static void ice_sriov_free_irqs(struct ice_pf *pf, struct ice_vf *vf)
537 /* Move back from first vector index to first index in bitmap */
538 int bm_i = pf->sriov_irq_size - vf->first_vector_idx - vf->num_msix;
540 bitmap_clear(pf->sriov_irq_bm, bm_i, vf->num_msix);
541 vf->first_vector_idx = 0;
545 * ice_init_vf_vsi_res - initialize/setup VF VSI resources
546 * @vf: VF to initialize/setup the VSI for
548 * This function creates a VSI for the VF, adds a VLAN 0 filter, and sets up the
549 * VF VSI's broadcast filter and is only used during initial VF creation.
551 static int ice_init_vf_vsi_res(struct ice_vf *vf)
553 struct ice_pf *pf = vf->pf;
557 vf->first_vector_idx = ice_sriov_get_irqs(pf, vf->num_msix);
558 if (vf->first_vector_idx < 0)
561 vsi = ice_vf_vsi_setup(vf);
565 err = ice_vf_init_host_cfg(vf, vsi);
572 ice_vf_vsi_release(vf);
577 * ice_start_vfs - start VFs so they are ready to be used by SR-IOV
578 * @pf: PF the VFs are associated with
580 static int ice_start_vfs(struct ice_pf *pf)
582 struct ice_hw *hw = &pf->hw;
583 unsigned int bkt, it_cnt;
587 lockdep_assert_held(&pf->vfs.table_lock);
590 ice_for_each_vf(pf, bkt, vf) {
591 vf->vf_ops->clear_reset_trigger(vf);
593 retval = ice_init_vf_vsi_res(vf);
595 dev_err(ice_pf_to_dev(pf), "Failed to initialize VSI resources for VF %d, error %d\n",
600 retval = ice_eswitch_attach(pf, vf);
602 dev_err(ice_pf_to_dev(pf), "Failed to attach VF %d to eswitch, error %d",
604 ice_vf_vsi_release(vf);
608 set_bit(ICE_VF_STATE_INIT, vf->vf_states);
609 ice_ena_vf_mappings(vf);
610 wr32(hw, VFGEN_RSTAT(vf->vf_id), VIRTCHNL_VFR_VFACTIVE);
618 ice_for_each_vf(pf, bkt, vf) {
622 ice_dis_vf_mappings(vf);
623 ice_vf_vsi_release(vf);
631 * ice_sriov_free_vf - Free VF memory after all references are dropped
632 * @vf: pointer to VF to free
634 * Called by ice_put_vf through ice_release_vf once the last reference to a VF
635 * structure has been dropped.
637 static void ice_sriov_free_vf(struct ice_vf *vf)
639 mutex_destroy(&vf->cfg_lock);
645 * ice_sriov_clear_reset_state - clears VF Reset status register
646 * @vf: the vf to configure
648 static void ice_sriov_clear_reset_state(struct ice_vf *vf)
650 struct ice_hw *hw = &vf->pf->hw;
652 /* Clear the reset status register so that VF immediately sees that
653 * the device is resetting, even if hardware hasn't yet gotten around
654 * to clearing VFGEN_RSTAT for us.
656 wr32(hw, VFGEN_RSTAT(vf->vf_id), VIRTCHNL_VFR_INPROGRESS);
660 * ice_sriov_clear_mbx_register - clears SRIOV VF's mailbox registers
661 * @vf: the vf to configure
663 static void ice_sriov_clear_mbx_register(struct ice_vf *vf)
665 struct ice_pf *pf = vf->pf;
667 wr32(&pf->hw, VF_MBX_ARQLEN(vf->vf_id), 0);
668 wr32(&pf->hw, VF_MBX_ATQLEN(vf->vf_id), 0);
672 * ice_sriov_trigger_reset_register - trigger VF reset for SRIOV VF
673 * @vf: pointer to VF structure
674 * @is_vflr: true if reset occurred due to VFLR
676 * Trigger and cleanup after a VF reset for a SR-IOV VF.
678 static void ice_sriov_trigger_reset_register(struct ice_vf *vf, bool is_vflr)
680 struct ice_pf *pf = vf->pf;
681 u32 reg, reg_idx, bit_idx;
682 unsigned int vf_abs_id, i;
686 dev = ice_pf_to_dev(pf);
688 vf_abs_id = vf->vf_id + hw->func_caps.vf_base_id;
690 /* In the case of a VFLR, HW has already reset the VF and we just need
691 * to clean up. Otherwise we must first trigger the reset using the
695 reg = rd32(hw, VPGEN_VFRTRIG(vf->vf_id));
696 reg |= VPGEN_VFRTRIG_VFSWR_M;
697 wr32(hw, VPGEN_VFRTRIG(vf->vf_id), reg);
700 /* clear the VFLR bit in GLGEN_VFLRSTAT */
701 reg_idx = (vf_abs_id) / 32;
702 bit_idx = (vf_abs_id) % 32;
703 wr32(hw, GLGEN_VFLRSTAT(reg_idx), BIT(bit_idx));
706 wr32(hw, PF_PCI_CIAA,
707 VF_DEVICE_STATUS | (vf_abs_id << PF_PCI_CIAA_VF_NUM_S));
708 for (i = 0; i < ICE_PCI_CIAD_WAIT_COUNT; i++) {
709 reg = rd32(hw, PF_PCI_CIAD);
710 /* no transactions pending so stop polling */
711 if ((reg & VF_TRANS_PENDING_M) == 0)
714 dev_err(dev, "VF %u PCI transactions stuck\n", vf->vf_id);
715 udelay(ICE_PCI_CIAD_WAIT_DELAY_US);
720 * ice_sriov_poll_reset_status - poll SRIOV VF reset status
721 * @vf: pointer to VF structure
723 * Returns true when reset is successful, else returns false
725 static bool ice_sriov_poll_reset_status(struct ice_vf *vf)
727 struct ice_pf *pf = vf->pf;
731 for (i = 0; i < 10; i++) {
732 /* VF reset requires driver to first reset the VF and then
733 * poll the status register to make sure that the reset
734 * completed successfully.
736 reg = rd32(&pf->hw, VPGEN_VFRSTAT(vf->vf_id));
737 if (reg & VPGEN_VFRSTAT_VFRD_M)
740 /* only sleep if the reset is not done */
741 usleep_range(10, 20);
747 * ice_sriov_clear_reset_trigger - enable VF to access hardware
748 * @vf: VF to enabled hardware access for
750 static void ice_sriov_clear_reset_trigger(struct ice_vf *vf)
752 struct ice_hw *hw = &vf->pf->hw;
755 reg = rd32(hw, VPGEN_VFRTRIG(vf->vf_id));
756 reg &= ~VPGEN_VFRTRIG_VFSWR_M;
757 wr32(hw, VPGEN_VFRTRIG(vf->vf_id), reg);
762 * ice_sriov_post_vsi_rebuild - tasks to do after the VF's VSI have been rebuilt
763 * @vf: VF to perform tasks on
765 static void ice_sriov_post_vsi_rebuild(struct ice_vf *vf)
767 ice_ena_vf_mappings(vf);
768 wr32(&vf->pf->hw, VFGEN_RSTAT(vf->vf_id), VIRTCHNL_VFR_VFACTIVE);
771 static const struct ice_vf_ops ice_sriov_vf_ops = {
772 .reset_type = ICE_VF_RESET,
773 .free = ice_sriov_free_vf,
774 .clear_reset_state = ice_sriov_clear_reset_state,
775 .clear_mbx_register = ice_sriov_clear_mbx_register,
776 .trigger_reset_register = ice_sriov_trigger_reset_register,
777 .poll_reset_status = ice_sriov_poll_reset_status,
778 .clear_reset_trigger = ice_sriov_clear_reset_trigger,
780 .post_vsi_rebuild = ice_sriov_post_vsi_rebuild,
784 * ice_create_vf_entries - Allocate and insert VF entries
785 * @pf: pointer to the PF structure
786 * @num_vfs: the number of VFs to allocate
788 * Allocate new VF entries and insert them into the hash table. Set some
789 * basic default fields for initializing the new VFs.
791 * After this function exits, the hash table will have num_vfs entries
794 * Returns 0 on success or an integer error code on failure.
796 static int ice_create_vf_entries(struct ice_pf *pf, u16 num_vfs)
798 struct pci_dev *pdev = pf->pdev;
799 struct ice_vfs *vfs = &pf->vfs;
800 struct pci_dev *vfdev = NULL;
805 lockdep_assert_held(&vfs->table_lock);
807 pos = pci_find_ext_capability(pdev, PCI_EXT_CAP_ID_SRIOV);
808 pci_read_config_word(pdev, pos + PCI_SRIOV_VF_DID, &vf_pdev_id);
810 for (u16 vf_id = 0; vf_id < num_vfs; vf_id++) {
811 vf = kzalloc(sizeof(*vf), GFP_KERNEL);
814 goto err_free_entries;
816 kref_init(&vf->refcnt);
821 /* set sriov vf ops for VFs created during SRIOV flow */
822 vf->vf_ops = &ice_sriov_vf_ops;
824 ice_initialize_vf_entry(vf);
827 vfdev = pci_get_device(pdev->vendor, vf_pdev_id, vfdev);
828 } while (vfdev && vfdev->physfn != pdev);
830 vf->vf_sw_id = pf->first_sw;
834 hash_add_rcu(vfs->table, &vf->entry, vf_id);
837 /* Decrement of refcount done by pci_get_device() inside the loop does
838 * not touch the last iteration's vfdev, so it has to be done manually
839 * to balance pci_dev_get() added within the loop.
846 ice_free_vf_entries(pf);
851 * ice_ena_vfs - enable VFs so they are ready to be used
852 * @pf: pointer to the PF structure
853 * @num_vfs: number of VFs to enable
855 static int ice_ena_vfs(struct ice_pf *pf, u16 num_vfs)
857 int total_vectors = pf->hw.func_caps.common_cap.num_msix_vectors;
858 struct device *dev = ice_pf_to_dev(pf);
859 struct ice_hw *hw = &pf->hw;
862 pf->sriov_irq_bm = bitmap_zalloc(total_vectors, GFP_KERNEL);
863 if (!pf->sriov_irq_bm)
865 pf->sriov_irq_size = total_vectors;
867 /* Disable global interrupt 0 so we don't try to handle the VFLR. */
868 wr32(hw, GLINT_DYN_CTL(pf->oicr_irq.index),
869 ICE_ITR_NONE << GLINT_DYN_CTL_ITR_INDX_S);
870 set_bit(ICE_OICR_INTR_DIS, pf->state);
873 ret = pci_enable_sriov(pf->pdev, num_vfs);
875 goto err_unroll_intr;
877 mutex_lock(&pf->vfs.table_lock);
879 ret = ice_set_per_vf_res(pf, num_vfs);
881 dev_err(dev, "Not enough resources for %d VFs, err %d. Try with fewer number of VFs\n",
883 goto err_unroll_sriov;
886 ret = ice_create_vf_entries(pf, num_vfs);
888 dev_err(dev, "Failed to allocate VF entries for %d VFs\n",
890 goto err_unroll_sriov;
893 ret = ice_start_vfs(pf);
895 dev_err(dev, "Failed to start %d VFs, err %d\n", num_vfs, ret);
897 goto err_unroll_vf_entries;
900 clear_bit(ICE_VF_DIS, pf->state);
902 /* rearm global interrupts */
903 if (test_and_clear_bit(ICE_OICR_INTR_DIS, pf->state))
904 ice_irq_dynamic_ena(hw, NULL, NULL);
906 mutex_unlock(&pf->vfs.table_lock);
910 err_unroll_vf_entries:
911 ice_free_vf_entries(pf);
913 mutex_unlock(&pf->vfs.table_lock);
914 pci_disable_sriov(pf->pdev);
916 /* rearm interrupts here */
917 ice_irq_dynamic_ena(hw, NULL, NULL);
918 clear_bit(ICE_OICR_INTR_DIS, pf->state);
919 bitmap_free(pf->sriov_irq_bm);
924 * ice_pci_sriov_ena - Enable or change number of VFs
925 * @pf: pointer to the PF structure
926 * @num_vfs: number of VFs to allocate
928 * Returns 0 on success and negative on failure
930 static int ice_pci_sriov_ena(struct ice_pf *pf, int num_vfs)
932 struct device *dev = ice_pf_to_dev(pf);
940 if (num_vfs > pf->vfs.num_supported) {
941 dev_err(dev, "Can't enable %d VFs, max VFs supported is %d\n",
942 num_vfs, pf->vfs.num_supported);
946 dev_info(dev, "Enabling %d VFs\n", num_vfs);
947 err = ice_ena_vfs(pf, num_vfs);
949 dev_err(dev, "Failed to enable SR-IOV: %d\n", err);
953 set_bit(ICE_FLAG_SRIOV_ENA, pf->flags);
958 * ice_check_sriov_allowed - check if SR-IOV is allowed based on various checks
959 * @pf: PF to enabled SR-IOV on
961 static int ice_check_sriov_allowed(struct ice_pf *pf)
963 struct device *dev = ice_pf_to_dev(pf);
965 if (!test_bit(ICE_FLAG_SRIOV_CAPABLE, pf->flags)) {
966 dev_err(dev, "This device is not capable of SR-IOV\n");
970 if (ice_is_safe_mode(pf)) {
971 dev_err(dev, "SR-IOV cannot be configured - Device is in Safe Mode\n");
975 if (!ice_pf_state_is_nominal(pf)) {
976 dev_err(dev, "Cannot enable SR-IOV, device not ready\n");
984 * ice_sriov_get_vf_total_msix - return number of MSI-X used by VFs
985 * @pdev: pointer to pci_dev struct
987 * The function is called via sysfs ops
989 u32 ice_sriov_get_vf_total_msix(struct pci_dev *pdev)
991 struct ice_pf *pf = pci_get_drvdata(pdev);
993 return pf->sriov_irq_size - ice_get_max_used_msix_vector(pf);
996 static int ice_sriov_move_base_vector(struct ice_pf *pf, int move)
998 if (pf->sriov_base_vector - move < ice_get_max_used_msix_vector(pf))
1001 pf->sriov_base_vector -= move;
1005 static void ice_sriov_remap_vectors(struct ice_pf *pf, u16 restricted_id)
1007 u16 vf_ids[ICE_MAX_SRIOV_VFS];
1008 struct ice_vf *tmp_vf;
1009 int to_remap = 0, bkt;
1011 /* For better irqs usage try to remap irqs of VFs
1012 * that aren't running yet
1014 ice_for_each_vf(pf, bkt, tmp_vf) {
1015 /* skip VF which is changing the number of MSI-X */
1016 if (restricted_id == tmp_vf->vf_id ||
1017 test_bit(ICE_VF_STATE_ACTIVE, tmp_vf->vf_states))
1020 ice_dis_vf_mappings(tmp_vf);
1021 ice_sriov_free_irqs(pf, tmp_vf);
1023 vf_ids[to_remap] = tmp_vf->vf_id;
1027 for (int i = 0; i < to_remap; i++) {
1028 tmp_vf = ice_get_vf_by_id(pf, vf_ids[i]);
1032 tmp_vf->first_vector_idx =
1033 ice_sriov_get_irqs(pf, tmp_vf->num_msix);
1034 /* there is no need to rebuild VSI as we are only changing the
1035 * vector indexes not amount of MSI-X or queues
1037 ice_ena_vf_mappings(tmp_vf);
1043 * ice_sriov_set_msix_vec_count
1044 * @vf_dev: pointer to pci_dev struct of VF device
1045 * @msix_vec_count: new value for MSI-X amount on this VF
1047 * Set requested MSI-X, queues and registers for @vf_dev.
1049 * First do some sanity checks like if there are any VFs, if the new value
1050 * is correct etc. Then disable old mapping (MSI-X and queues registers), change
1051 * MSI-X and queues, rebuild VSI and enable new mapping.
1053 * If it is possible (driver not binded to VF) try to remap also other VFs to
1054 * linearize irqs register usage.
1056 int ice_sriov_set_msix_vec_count(struct pci_dev *vf_dev, int msix_vec_count)
1058 struct pci_dev *pdev = pci_physfn(vf_dev);
1059 struct ice_pf *pf = pci_get_drvdata(pdev);
1060 u16 prev_msix, prev_queues, queues;
1061 bool needs_rebuild = false;
1062 struct ice_vsi *vsi;
1066 if (!ice_get_num_vfs(pf))
1069 if (!msix_vec_count)
1072 queues = msix_vec_count;
1073 /* add 1 MSI-X for OICR */
1074 msix_vec_count += 1;
1076 if (queues > min(ice_get_avail_txq_count(pf),
1077 ice_get_avail_rxq_count(pf)))
1080 if (msix_vec_count < ICE_MIN_INTR_PER_VF)
1083 /* Transition of PCI VF function number to function_id */
1084 for (id = 0; id < pci_num_vf(pdev); id++) {
1085 if (vf_dev->devfn == pci_iov_virtfn_devfn(pdev, id))
1089 if (id == pci_num_vf(pdev))
1092 vf = ice_get_vf_by_id(pf, id);
1097 vsi = ice_get_vf_vsi(vf);
1101 prev_msix = vf->num_msix;
1102 prev_queues = vf->num_vf_qs;
1104 if (ice_sriov_move_base_vector(pf, msix_vec_count - prev_msix)) {
1109 ice_dis_vf_mappings(vf);
1110 ice_sriov_free_irqs(pf, vf);
1112 /* Remap all VFs beside the one is now configured */
1113 ice_sriov_remap_vectors(pf, vf->vf_id);
1115 vf->num_msix = msix_vec_count;
1116 vf->num_vf_qs = queues;
1117 vf->first_vector_idx = ice_sriov_get_irqs(pf, vf->num_msix);
1118 if (vf->first_vector_idx < 0)
1121 if (ice_vf_reconfig_vsi(vf) || ice_vf_init_host_cfg(vf, vsi)) {
1122 /* Try to rebuild with previous values */
1123 needs_rebuild = true;
1127 dev_info(ice_pf_to_dev(pf),
1128 "Changing VF %d resources to %d vectors and %d queues\n",
1129 vf->vf_id, vf->num_msix, vf->num_vf_qs);
1131 ice_ena_vf_mappings(vf);
1137 dev_info(ice_pf_to_dev(pf),
1138 "Can't set %d vectors on VF %d, falling back to %d\n",
1139 vf->num_msix, vf->vf_id, prev_msix);
1141 vf->num_msix = prev_msix;
1142 vf->num_vf_qs = prev_queues;
1143 vf->first_vector_idx = ice_sriov_get_irqs(pf, vf->num_msix);
1144 if (vf->first_vector_idx < 0)
1147 if (needs_rebuild) {
1148 ice_vf_reconfig_vsi(vf);
1149 ice_vf_init_host_cfg(vf, vsi);
1152 ice_ena_vf_mappings(vf);
1159 * ice_sriov_configure - Enable or change number of VFs via sysfs
1160 * @pdev: pointer to a pci_dev structure
1161 * @num_vfs: number of VFs to allocate or 0 to free VFs
1163 * This function is called when the user updates the number of VFs in sysfs. On
1164 * success return whatever num_vfs was set to by the caller. Return negative on
1167 int ice_sriov_configure(struct pci_dev *pdev, int num_vfs)
1169 struct ice_pf *pf = pci_get_drvdata(pdev);
1170 struct device *dev = ice_pf_to_dev(pf);
1173 err = ice_check_sriov_allowed(pf);
1178 if (!pci_vfs_assigned(pdev)) {
1183 dev_err(dev, "can't free VFs because some are assigned to VMs.\n");
1187 err = ice_pci_sriov_ena(pf, num_vfs);
1195 * ice_process_vflr_event - Free VF resources via IRQ calls
1196 * @pf: pointer to the PF structure
1198 * called from the VFLR IRQ handler to
1199 * free up VF resources and state variables
1201 void ice_process_vflr_event(struct ice_pf *pf)
1203 struct ice_hw *hw = &pf->hw;
1208 if (!test_and_clear_bit(ICE_VFLR_EVENT_PENDING, pf->state) ||
1212 mutex_lock(&pf->vfs.table_lock);
1213 ice_for_each_vf(pf, bkt, vf) {
1214 u32 reg_idx, bit_idx;
1216 reg_idx = (hw->func_caps.vf_base_id + vf->vf_id) / 32;
1217 bit_idx = (hw->func_caps.vf_base_id + vf->vf_id) % 32;
1218 /* read GLGEN_VFLRSTAT register to find out the flr VFs */
1219 reg = rd32(hw, GLGEN_VFLRSTAT(reg_idx));
1220 if (reg & BIT(bit_idx))
1221 /* GLGEN_VFLRSTAT bit will be cleared in ice_reset_vf */
1222 ice_reset_vf(vf, ICE_VF_RESET_VFLR | ICE_VF_RESET_LOCK);
1224 mutex_unlock(&pf->vfs.table_lock);
1228 * ice_get_vf_from_pfq - get the VF who owns the PF space queue passed in
1229 * @pf: PF used to index all VFs
1230 * @pfq: queue index relative to the PF's function space
1232 * If no VF is found who owns the pfq then return NULL, otherwise return a
1233 * pointer to the VF who owns the pfq
1235 * If this function returns non-NULL, it acquires a reference count of the VF
1236 * structure. The caller is responsible for calling ice_put_vf() to drop this
1239 static struct ice_vf *ice_get_vf_from_pfq(struct ice_pf *pf, u16 pfq)
1245 ice_for_each_vf_rcu(pf, bkt, vf) {
1246 struct ice_vsi *vsi;
1249 vsi = ice_get_vf_vsi(vf);
1253 ice_for_each_rxq(vsi, rxq_idx)
1254 if (vsi->rxq_map[rxq_idx] == pfq) {
1255 struct ice_vf *found;
1257 if (kref_get_unless_zero(&vf->refcnt))
1271 * ice_globalq_to_pfq - convert from global queue index to PF space queue index
1272 * @pf: PF used for conversion
1273 * @globalq: global queue index used to convert to PF space queue index
1275 static u32 ice_globalq_to_pfq(struct ice_pf *pf, u32 globalq)
1277 return globalq - pf->hw.func_caps.common_cap.rxq_first_id;
1281 * ice_vf_lan_overflow_event - handle LAN overflow event for a VF
1282 * @pf: PF that the LAN overflow event happened on
1283 * @event: structure holding the event information for the LAN overflow event
1285 * Determine if the LAN overflow event was caused by a VF queue. If it was not
1286 * caused by a VF, do nothing. If a VF caused this LAN overflow event trigger a
1287 * reset on the offending VF.
1290 ice_vf_lan_overflow_event(struct ice_pf *pf, struct ice_rq_event_info *event)
1292 u32 gldcb_rtctq, queue;
1295 gldcb_rtctq = le32_to_cpu(event->desc.params.lan_overflow.prtdcb_ruptq);
1296 dev_dbg(ice_pf_to_dev(pf), "GLDCB_RTCTQ: 0x%08x\n", gldcb_rtctq);
1298 /* event returns device global Rx queue number */
1299 queue = FIELD_GET(GLDCB_RTCTQ_RXQNUM_M, gldcb_rtctq);
1301 vf = ice_get_vf_from_pfq(pf, ice_globalq_to_pfq(pf, queue));
1305 ice_reset_vf(vf, ICE_VF_RESET_NOTIFY | ICE_VF_RESET_LOCK);
1310 * ice_set_vf_spoofchk
1311 * @netdev: network interface device structure
1312 * @vf_id: VF identifier
1313 * @ena: flag to enable or disable feature
1315 * Enable or disable VF spoof checking
1317 int ice_set_vf_spoofchk(struct net_device *netdev, int vf_id, bool ena)
1319 struct ice_netdev_priv *np = netdev_priv(netdev);
1320 struct ice_pf *pf = np->vsi->back;
1321 struct ice_vsi *vf_vsi;
1326 dev = ice_pf_to_dev(pf);
1328 vf = ice_get_vf_by_id(pf, vf_id);
1332 ret = ice_check_vf_ready_for_cfg(vf);
1336 vf_vsi = ice_get_vf_vsi(vf);
1338 netdev_err(netdev, "VSI %d for VF %d is null\n",
1339 vf->lan_vsi_idx, vf->vf_id);
1344 if (vf_vsi->type != ICE_VSI_VF) {
1345 netdev_err(netdev, "Type %d of VSI %d for VF %d is no ICE_VSI_VF\n",
1346 vf_vsi->type, vf_vsi->vsi_num, vf->vf_id);
1351 if (ena == vf->spoofchk) {
1352 dev_dbg(dev, "VF spoofchk already %s\n", ena ? "ON" : "OFF");
1357 ret = ice_vsi_apply_spoofchk(vf_vsi, ena);
1359 dev_err(dev, "Failed to set spoofchk %s for VF %d VSI %d\n error %d\n",
1360 ena ? "ON" : "OFF", vf->vf_id, vf_vsi->vsi_num, ret);
1371 * @netdev: network interface device structure
1372 * @vf_id: VF identifier
1373 * @ivi: VF configuration structure
1375 * return VF configuration
1378 ice_get_vf_cfg(struct net_device *netdev, int vf_id, struct ifla_vf_info *ivi)
1380 struct ice_pf *pf = ice_netdev_to_pf(netdev);
1384 vf = ice_get_vf_by_id(pf, vf_id);
1388 ret = ice_check_vf_ready_for_cfg(vf);
1393 ether_addr_copy(ivi->mac, vf->hw_lan_addr);
1395 /* VF configuration for VLAN and applicable QoS */
1396 ivi->vlan = ice_vf_get_port_vlan_id(vf);
1397 ivi->qos = ice_vf_get_port_vlan_prio(vf);
1398 if (ice_vf_is_port_vlan_ena(vf))
1399 ivi->vlan_proto = cpu_to_be16(ice_vf_get_port_vlan_tpid(vf));
1401 ivi->trusted = vf->trusted;
1402 ivi->spoofchk = vf->spoofchk;
1403 if (!vf->link_forced)
1404 ivi->linkstate = IFLA_VF_LINK_STATE_AUTO;
1405 else if (vf->link_up)
1406 ivi->linkstate = IFLA_VF_LINK_STATE_ENABLE;
1408 ivi->linkstate = IFLA_VF_LINK_STATE_DISABLE;
1409 ivi->max_tx_rate = vf->max_tx_rate;
1410 ivi->min_tx_rate = vf->min_tx_rate;
1419 * @netdev: network interface device structure
1420 * @vf_id: VF identifier
1423 * program VF MAC address
1425 int ice_set_vf_mac(struct net_device *netdev, int vf_id, u8 *mac)
1427 struct ice_pf *pf = ice_netdev_to_pf(netdev);
1431 if (is_multicast_ether_addr(mac)) {
1432 netdev_err(netdev, "%pM not a valid unicast address\n", mac);
1436 vf = ice_get_vf_by_id(pf, vf_id);
1440 /* nothing left to do, unicast MAC already set */
1441 if (ether_addr_equal(vf->dev_lan_addr, mac) &&
1442 ether_addr_equal(vf->hw_lan_addr, mac)) {
1447 ret = ice_check_vf_ready_for_cfg(vf);
1451 mutex_lock(&vf->cfg_lock);
1453 /* VF is notified of its new MAC via the PF's response to the
1454 * VIRTCHNL_OP_GET_VF_RESOURCES message after the VF has been reset
1456 ether_addr_copy(vf->dev_lan_addr, mac);
1457 ether_addr_copy(vf->hw_lan_addr, mac);
1458 if (is_zero_ether_addr(mac)) {
1459 /* VF will send VIRTCHNL_OP_ADD_ETH_ADDR message with its MAC */
1460 vf->pf_set_mac = false;
1461 netdev_info(netdev, "Removing MAC on VF %d. VF driver will be reinitialized\n",
1464 /* PF will add MAC rule for the VF */
1465 vf->pf_set_mac = true;
1466 netdev_info(netdev, "Setting MAC %pM on VF %d. VF driver will be reinitialized\n",
1470 ice_reset_vf(vf, ICE_VF_RESET_NOTIFY);
1471 mutex_unlock(&vf->cfg_lock);
1480 * @netdev: network interface device structure
1481 * @vf_id: VF identifier
1482 * @trusted: Boolean value to enable/disable trusted VF
1484 * Enable or disable a given VF as trusted
1486 int ice_set_vf_trust(struct net_device *netdev, int vf_id, bool trusted)
1488 struct ice_pf *pf = ice_netdev_to_pf(netdev);
1492 vf = ice_get_vf_by_id(pf, vf_id);
1496 if (ice_is_eswitch_mode_switchdev(pf)) {
1497 dev_info(ice_pf_to_dev(pf), "Trusted VF is forbidden in switchdev mode\n");
1501 ret = ice_check_vf_ready_for_cfg(vf);
1505 /* Check if already trusted */
1506 if (trusted == vf->trusted) {
1511 mutex_lock(&vf->cfg_lock);
1513 vf->trusted = trusted;
1514 ice_reset_vf(vf, ICE_VF_RESET_NOTIFY);
1515 dev_info(ice_pf_to_dev(pf), "VF %u is now %strusted\n",
1516 vf_id, trusted ? "" : "un");
1518 mutex_unlock(&vf->cfg_lock);
1526 * ice_set_vf_link_state
1527 * @netdev: network interface device structure
1528 * @vf_id: VF identifier
1529 * @link_state: required link state
1531 * Set VF's link state, irrespective of physical link state status
1533 int ice_set_vf_link_state(struct net_device *netdev, int vf_id, int link_state)
1535 struct ice_pf *pf = ice_netdev_to_pf(netdev);
1539 vf = ice_get_vf_by_id(pf, vf_id);
1543 ret = ice_check_vf_ready_for_cfg(vf);
1547 switch (link_state) {
1548 case IFLA_VF_LINK_STATE_AUTO:
1549 vf->link_forced = false;
1551 case IFLA_VF_LINK_STATE_ENABLE:
1552 vf->link_forced = true;
1555 case IFLA_VF_LINK_STATE_DISABLE:
1556 vf->link_forced = true;
1557 vf->link_up = false;
1564 ice_vc_notify_vf_link_state(vf);
1572 * ice_calc_all_vfs_min_tx_rate - calculate cumulative min Tx rate on all VFs
1573 * @pf: PF associated with VFs
1575 static int ice_calc_all_vfs_min_tx_rate(struct ice_pf *pf)
1582 ice_for_each_vf_rcu(pf, bkt, vf)
1583 rate += vf->min_tx_rate;
1590 * ice_min_tx_rate_oversubscribed - check if min Tx rate causes oversubscription
1591 * @vf: VF trying to configure min_tx_rate
1592 * @min_tx_rate: min Tx rate in Mbps
1594 * Check if the min_tx_rate being passed in will cause oversubscription of total
1595 * min_tx_rate based on the current link speed and all other VFs configured
1598 * Return true if the passed min_tx_rate would cause oversubscription, else
1602 ice_min_tx_rate_oversubscribed(struct ice_vf *vf, int min_tx_rate)
1604 struct ice_vsi *vsi = ice_get_vf_vsi(vf);
1605 int all_vfs_min_tx_rate;
1606 int link_speed_mbps;
1611 link_speed_mbps = ice_get_link_speed_mbps(vsi);
1612 all_vfs_min_tx_rate = ice_calc_all_vfs_min_tx_rate(vf->pf);
1614 /* this VF's previous rate is being overwritten */
1615 all_vfs_min_tx_rate -= vf->min_tx_rate;
1617 if (all_vfs_min_tx_rate + min_tx_rate > link_speed_mbps) {
1618 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",
1619 min_tx_rate, vf->vf_id,
1620 all_vfs_min_tx_rate + min_tx_rate - link_speed_mbps,
1629 * ice_set_vf_bw - set min/max VF bandwidth
1630 * @netdev: network interface device structure
1631 * @vf_id: VF identifier
1632 * @min_tx_rate: Minimum Tx rate in Mbps
1633 * @max_tx_rate: Maximum Tx rate in Mbps
1636 ice_set_vf_bw(struct net_device *netdev, int vf_id, int min_tx_rate,
1639 struct ice_pf *pf = ice_netdev_to_pf(netdev);
1640 struct ice_vsi *vsi;
1645 dev = ice_pf_to_dev(pf);
1647 vf = ice_get_vf_by_id(pf, vf_id);
1651 ret = ice_check_vf_ready_for_cfg(vf);
1655 vsi = ice_get_vf_vsi(vf);
1661 if (min_tx_rate && ice_is_dcb_active(pf)) {
1662 dev_err(dev, "DCB on PF is currently enabled. VF min Tx rate limiting not allowed on this PF.\n");
1667 if (ice_min_tx_rate_oversubscribed(vf, min_tx_rate)) {
1672 if (vf->min_tx_rate != (unsigned int)min_tx_rate) {
1673 ret = ice_set_min_bw_limit(vsi, (u64)min_tx_rate * 1000);
1675 dev_err(dev, "Unable to set min-tx-rate for VF %d\n",
1680 vf->min_tx_rate = min_tx_rate;
1683 if (vf->max_tx_rate != (unsigned int)max_tx_rate) {
1684 ret = ice_set_max_bw_limit(vsi, (u64)max_tx_rate * 1000);
1686 dev_err(dev, "Unable to set max-tx-rate for VF %d\n",
1691 vf->max_tx_rate = max_tx_rate;
1700 * ice_get_vf_stats - populate some stats for the VF
1701 * @netdev: the netdev of the PF
1702 * @vf_id: the host OS identifier (0-255)
1703 * @vf_stats: pointer to the OS memory to be initialized
1705 int ice_get_vf_stats(struct net_device *netdev, int vf_id,
1706 struct ifla_vf_stats *vf_stats)
1708 struct ice_pf *pf = ice_netdev_to_pf(netdev);
1709 struct ice_eth_stats *stats;
1710 struct ice_vsi *vsi;
1714 vf = ice_get_vf_by_id(pf, vf_id);
1718 ret = ice_check_vf_ready_for_cfg(vf);
1722 vsi = ice_get_vf_vsi(vf);
1728 ice_update_eth_stats(vsi);
1729 stats = &vsi->eth_stats;
1731 memset(vf_stats, 0, sizeof(*vf_stats));
1733 vf_stats->rx_packets = stats->rx_unicast + stats->rx_broadcast +
1734 stats->rx_multicast;
1735 vf_stats->tx_packets = stats->tx_unicast + stats->tx_broadcast +
1736 stats->tx_multicast;
1737 vf_stats->rx_bytes = stats->rx_bytes;
1738 vf_stats->tx_bytes = stats->tx_bytes;
1739 vf_stats->broadcast = stats->rx_broadcast;
1740 vf_stats->multicast = stats->rx_multicast;
1741 vf_stats->rx_dropped = stats->rx_discards;
1742 vf_stats->tx_dropped = stats->tx_discards;
1750 * ice_is_supported_port_vlan_proto - make sure the vlan_proto is supported
1751 * @hw: hardware structure used to check the VLAN mode
1752 * @vlan_proto: VLAN TPID being checked
1754 * If the device is configured in Double VLAN Mode (DVM), then both ETH_P_8021Q
1755 * and ETH_P_8021AD are supported. If the device is configured in Single VLAN
1756 * Mode (SVM), then only ETH_P_8021Q is supported.
1759 ice_is_supported_port_vlan_proto(struct ice_hw *hw, u16 vlan_proto)
1761 bool is_supported = false;
1763 switch (vlan_proto) {
1765 is_supported = true;
1768 if (ice_is_dvm_ena(hw))
1769 is_supported = true;
1773 return is_supported;
1777 * ice_set_vf_port_vlan
1778 * @netdev: network interface device structure
1779 * @vf_id: VF identifier
1780 * @vlan_id: VLAN ID being set
1781 * @qos: priority setting
1782 * @vlan_proto: VLAN protocol
1784 * program VF Port VLAN ID and/or QoS
1787 ice_set_vf_port_vlan(struct net_device *netdev, int vf_id, u16 vlan_id, u8 qos,
1790 struct ice_pf *pf = ice_netdev_to_pf(netdev);
1791 u16 local_vlan_proto = ntohs(vlan_proto);
1796 dev = ice_pf_to_dev(pf);
1798 if (vlan_id >= VLAN_N_VID || qos > 7) {
1799 dev_err(dev, "Invalid Port VLAN parameters for VF %d, ID %d, QoS %d\n",
1800 vf_id, vlan_id, qos);
1804 if (!ice_is_supported_port_vlan_proto(&pf->hw, local_vlan_proto)) {
1805 dev_err(dev, "VF VLAN protocol 0x%04x is not supported\n",
1807 return -EPROTONOSUPPORT;
1810 vf = ice_get_vf_by_id(pf, vf_id);
1814 ret = ice_check_vf_ready_for_cfg(vf);
1818 if (ice_vf_get_port_vlan_prio(vf) == qos &&
1819 ice_vf_get_port_vlan_tpid(vf) == local_vlan_proto &&
1820 ice_vf_get_port_vlan_id(vf) == vlan_id) {
1821 /* duplicate request, so just return success */
1822 dev_dbg(dev, "Duplicate port VLAN %u, QoS %u, TPID 0x%04x request\n",
1823 vlan_id, qos, local_vlan_proto);
1828 mutex_lock(&vf->cfg_lock);
1830 vf->port_vlan_info = ICE_VLAN(local_vlan_proto, vlan_id, qos);
1831 if (ice_vf_is_port_vlan_ena(vf))
1832 dev_info(dev, "Setting VLAN %u, QoS %u, TPID 0x%04x on VF %d\n",
1833 vlan_id, qos, local_vlan_proto, vf_id);
1835 dev_info(dev, "Clearing port VLAN on VF %d\n", vf_id);
1837 ice_reset_vf(vf, ICE_VF_RESET_NOTIFY);
1838 mutex_unlock(&vf->cfg_lock);
1846 * ice_print_vf_rx_mdd_event - print VF Rx malicious driver detect event
1847 * @vf: pointer to the VF structure
1849 void ice_print_vf_rx_mdd_event(struct ice_vf *vf)
1851 struct ice_pf *pf = vf->pf;
1854 dev = ice_pf_to_dev(pf);
1856 dev_info(dev, "%d Rx Malicious Driver Detection events detected on PF %d VF %d MAC %pM. mdd-auto-reset-vfs=%s\n",
1857 vf->mdd_rx_events.count, pf->hw.pf_id, vf->vf_id,
1859 test_bit(ICE_FLAG_MDD_AUTO_RESET_VF, pf->flags)
1864 * ice_print_vfs_mdd_events - print VFs malicious driver detect event
1865 * @pf: pointer to the PF structure
1867 * Called from ice_handle_mdd_event to rate limit and print VFs MDD events.
1869 void ice_print_vfs_mdd_events(struct ice_pf *pf)
1871 struct device *dev = ice_pf_to_dev(pf);
1872 struct ice_hw *hw = &pf->hw;
1876 /* check that there are pending MDD events to print */
1877 if (!test_and_clear_bit(ICE_MDD_VF_PRINT_PENDING, pf->state))
1880 /* VF MDD event logs are rate limited to one second intervals */
1881 if (time_is_after_jiffies(pf->vfs.last_printed_mdd_jiffies + HZ * 1))
1884 pf->vfs.last_printed_mdd_jiffies = jiffies;
1886 mutex_lock(&pf->vfs.table_lock);
1887 ice_for_each_vf(pf, bkt, vf) {
1888 /* only print Rx MDD event message if there are new events */
1889 if (vf->mdd_rx_events.count != vf->mdd_rx_events.last_printed) {
1890 vf->mdd_rx_events.last_printed =
1891 vf->mdd_rx_events.count;
1892 ice_print_vf_rx_mdd_event(vf);
1895 /* only print Tx MDD event message if there are new events */
1896 if (vf->mdd_tx_events.count != vf->mdd_tx_events.last_printed) {
1897 vf->mdd_tx_events.last_printed =
1898 vf->mdd_tx_events.count;
1900 dev_info(dev, "%d Tx Malicious Driver Detection events detected on PF %d VF %d MAC %pM.\n",
1901 vf->mdd_tx_events.count, hw->pf_id, vf->vf_id,
1905 mutex_unlock(&pf->vfs.table_lock);
1909 * ice_restore_all_vfs_msi_state - restore VF MSI state after PF FLR
1910 * @pf: pointer to the PF structure
1912 * Called when recovering from a PF FLR to restore interrupt capability to
1915 void ice_restore_all_vfs_msi_state(struct ice_pf *pf)
1920 ice_for_each_vf(pf, bkt, vf)
1921 pci_restore_msi_state(vf->vfdev);