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 ice_eswitch_reserve_cp_queues(pf, -ice_get_num_vfs(pf));
175 mutex_lock(&vfs->table_lock);
177 ice_for_each_vf(pf, bkt, vf) {
178 mutex_lock(&vf->cfg_lock);
180 ice_eswitch_detach(pf, vf);
183 if (test_bit(ICE_VF_STATE_INIT, vf->vf_states)) {
184 /* disable VF qp mappings and set VF disable state */
185 ice_dis_vf_mappings(vf);
186 set_bit(ICE_VF_STATE_DIS, vf->vf_states);
190 if (!pci_vfs_assigned(pf->pdev)) {
191 u32 reg_idx, bit_idx;
193 reg_idx = (hw->func_caps.vf_base_id + vf->vf_id) / 32;
194 bit_idx = (hw->func_caps.vf_base_id + vf->vf_id) % 32;
195 wr32(hw, GLGEN_VFLRSTAT(reg_idx), BIT(bit_idx));
198 /* clear malicious info since the VF is getting released */
199 list_del(&vf->mbx_info.list_entry);
201 mutex_unlock(&vf->cfg_lock);
204 if (ice_sriov_free_msix_res(pf))
205 dev_err(dev, "Failed to free MSIX resources used by SR-IOV\n");
207 vfs->num_qps_per = 0;
208 ice_free_vf_entries(pf);
210 mutex_unlock(&vfs->table_lock);
212 clear_bit(ICE_VF_DIS, pf->state);
213 clear_bit(ICE_FLAG_SRIOV_ENA, pf->flags);
217 * ice_vf_vsi_setup - Set up a VF VSI
218 * @vf: VF to setup VSI for
220 * Returns pointer to the successfully allocated VSI struct on success,
221 * otherwise returns NULL on failure.
223 static struct ice_vsi *ice_vf_vsi_setup(struct ice_vf *vf)
225 struct ice_vsi_cfg_params params = {};
226 struct ice_pf *pf = vf->pf;
229 params.type = ICE_VSI_VF;
230 params.pi = ice_vf_get_port_info(vf);
232 params.flags = ICE_VSI_FLAG_INIT;
234 vsi = ice_vsi_setup(pf, ¶ms);
237 dev_err(ice_pf_to_dev(pf), "Failed to create VF VSI\n");
238 ice_vf_invalidate_vsi(vf);
242 vf->lan_vsi_idx = vsi->idx;
243 vf->lan_vsi_num = vsi->vsi_num;
250 * ice_ena_vf_msix_mappings - enable VF MSIX mappings in hardware
251 * @vf: VF to enable MSIX mappings for
253 * Some of the registers need to be indexed/configured using hardware global
254 * device values and other registers need 0-based values, which represent PF
257 static void ice_ena_vf_msix_mappings(struct ice_vf *vf)
259 int device_based_first_msix, device_based_last_msix;
260 int pf_based_first_msix, pf_based_last_msix, v;
261 struct ice_pf *pf = vf->pf;
262 int device_based_vf_id;
267 pf_based_first_msix = vf->first_vector_idx;
268 pf_based_last_msix = (pf_based_first_msix + vf->num_msix) - 1;
270 device_based_first_msix = pf_based_first_msix +
271 pf->hw.func_caps.common_cap.msix_vector_first_id;
272 device_based_last_msix =
273 (device_based_first_msix + vf->num_msix) - 1;
274 device_based_vf_id = vf->vf_id + hw->func_caps.vf_base_id;
276 reg = FIELD_PREP(VPINT_ALLOC_FIRST_M, device_based_first_msix) |
277 FIELD_PREP(VPINT_ALLOC_LAST_M, device_based_last_msix) |
279 wr32(hw, VPINT_ALLOC(vf->vf_id), reg);
281 reg = FIELD_PREP(VPINT_ALLOC_PCI_FIRST_M, device_based_first_msix) |
282 FIELD_PREP(VPINT_ALLOC_PCI_LAST_M, device_based_last_msix) |
283 VPINT_ALLOC_PCI_VALID_M;
284 wr32(hw, VPINT_ALLOC_PCI(vf->vf_id), reg);
286 /* map the interrupts to its functions */
287 for (v = pf_based_first_msix; v <= pf_based_last_msix; v++) {
288 reg = FIELD_PREP(GLINT_VECT2FUNC_VF_NUM_M, device_based_vf_id) |
289 FIELD_PREP(GLINT_VECT2FUNC_PF_NUM_M, hw->pf_id);
290 wr32(hw, GLINT_VECT2FUNC(v), reg);
293 /* Map mailbox interrupt to VF MSI-X vector 0 */
294 wr32(hw, VPINT_MBX_CTL(device_based_vf_id), VPINT_MBX_CTL_CAUSE_ENA_M);
298 * ice_ena_vf_q_mappings - enable Rx/Tx queue mappings for a VF
299 * @vf: VF to enable the mappings for
300 * @max_txq: max Tx queues allowed on the VF's VSI
301 * @max_rxq: max Rx queues allowed on the VF's VSI
303 static void ice_ena_vf_q_mappings(struct ice_vf *vf, u16 max_txq, u16 max_rxq)
305 struct device *dev = ice_pf_to_dev(vf->pf);
306 struct ice_vsi *vsi = ice_get_vf_vsi(vf);
307 struct ice_hw *hw = &vf->pf->hw;
313 /* set regardless of mapping mode */
314 wr32(hw, VPLAN_TXQ_MAPENA(vf->vf_id), VPLAN_TXQ_MAPENA_TX_ENA_M);
316 /* VF Tx queues allocation */
317 if (vsi->tx_mapping_mode == ICE_VSI_MAP_CONTIG) {
318 /* set the VF PF Tx queue range
319 * VFNUMQ value should be set to (number of queues - 1). A value
320 * of 0 means 1 queue and a value of 255 means 256 queues
322 reg = FIELD_PREP(VPLAN_TX_QBASE_VFFIRSTQ_M, vsi->txq_map[0]) |
323 FIELD_PREP(VPLAN_TX_QBASE_VFNUMQ_M, max_txq - 1);
324 wr32(hw, VPLAN_TX_QBASE(vf->vf_id), reg);
326 dev_err(dev, "Scattered mode for VF Tx queues is not yet implemented\n");
329 /* set regardless of mapping mode */
330 wr32(hw, VPLAN_RXQ_MAPENA(vf->vf_id), VPLAN_RXQ_MAPENA_RX_ENA_M);
332 /* VF Rx queues allocation */
333 if (vsi->rx_mapping_mode == ICE_VSI_MAP_CONTIG) {
334 /* set the VF PF Rx queue range
335 * VFNUMQ value should be set to (number of queues - 1). A value
336 * of 0 means 1 queue and a value of 255 means 256 queues
338 reg = FIELD_PREP(VPLAN_RX_QBASE_VFFIRSTQ_M, vsi->rxq_map[0]) |
339 FIELD_PREP(VPLAN_RX_QBASE_VFNUMQ_M, max_rxq - 1);
340 wr32(hw, VPLAN_RX_QBASE(vf->vf_id), reg);
342 dev_err(dev, "Scattered mode for VF Rx queues is not yet implemented\n");
347 * ice_ena_vf_mappings - enable VF MSIX and queue mapping
348 * @vf: pointer to the VF structure
350 static void ice_ena_vf_mappings(struct ice_vf *vf)
352 struct ice_vsi *vsi = ice_get_vf_vsi(vf);
357 ice_ena_vf_msix_mappings(vf);
358 ice_ena_vf_q_mappings(vf, vsi->alloc_txq, vsi->alloc_rxq);
362 * ice_calc_vf_reg_idx - Calculate the VF's register index in the PF space
363 * @vf: VF to calculate the register index for
364 * @q_vector: a q_vector associated to the VF
366 int ice_calc_vf_reg_idx(struct ice_vf *vf, struct ice_q_vector *q_vector)
368 if (!vf || !q_vector)
371 /* always add one to account for the OICR being the first MSIX */
372 return vf->first_vector_idx + q_vector->v_idx + 1;
376 * ice_sriov_set_msix_res - Set any used MSIX resources
377 * @pf: pointer to PF structure
378 * @num_msix_needed: number of MSIX vectors needed for all SR-IOV VFs
380 * This function allows SR-IOV resources to be taken from the end of the PF's
381 * allowed HW MSIX vectors so that the irq_tracker will not be affected. We
382 * just set the pf->sriov_base_vector and return success.
384 * If there are not enough resources available, return an error. This should
385 * always be caught by ice_set_per_vf_res().
387 * Return 0 on success, and -EINVAL when there are not enough MSIX vectors
388 * in the PF's space available for SR-IOV.
390 static int ice_sriov_set_msix_res(struct ice_pf *pf, u16 num_msix_needed)
392 u16 total_vectors = pf->hw.func_caps.common_cap.num_msix_vectors;
393 int vectors_used = ice_get_max_used_msix_vector(pf);
394 int sriov_base_vector;
396 sriov_base_vector = total_vectors - num_msix_needed;
398 /* make sure we only grab irq_tracker entries from the list end and
399 * that we have enough available MSIX vectors
401 if (sriov_base_vector < vectors_used)
404 pf->sriov_base_vector = sriov_base_vector;
410 * ice_set_per_vf_res - check if vectors and queues are available
411 * @pf: pointer to the PF structure
412 * @num_vfs: the number of SR-IOV VFs being configured
414 * First, determine HW interrupts from common pool. If we allocate fewer VFs, we
415 * get more vectors and can enable more queues per VF. Note that this does not
416 * grab any vectors from the SW pool already allocated. Also note, that all
417 * vector counts include one for each VF's miscellaneous interrupt vector
420 * Minimum VFs - 2 vectors, 1 queue pair
421 * Small VFs - 5 vectors, 4 queue pairs
422 * Medium VFs - 17 vectors, 16 queue pairs
424 * Second, determine number of queue pairs per VF by starting with a pre-defined
425 * maximum each VF supports. If this is not possible, then we adjust based on
426 * queue pairs available on the device.
428 * Lastly, set queue and MSI-X VF variables tracked by the PF so it can be used
429 * by each VF during VF initialization and reset.
431 static int ice_set_per_vf_res(struct ice_pf *pf, u16 num_vfs)
433 int vectors_used = ice_get_max_used_msix_vector(pf);
434 u16 num_msix_per_vf, num_txq, num_rxq, avail_qs;
435 int msix_avail_per_vf, msix_avail_for_sriov;
436 struct device *dev = ice_pf_to_dev(pf);
439 lockdep_assert_held(&pf->vfs.table_lock);
444 /* determine MSI-X resources per VF */
445 msix_avail_for_sriov = pf->hw.func_caps.common_cap.num_msix_vectors -
447 msix_avail_per_vf = msix_avail_for_sriov / num_vfs;
448 if (msix_avail_per_vf >= ICE_NUM_VF_MSIX_MED) {
449 num_msix_per_vf = ICE_NUM_VF_MSIX_MED;
450 } else if (msix_avail_per_vf >= ICE_NUM_VF_MSIX_SMALL) {
451 num_msix_per_vf = ICE_NUM_VF_MSIX_SMALL;
452 } else if (msix_avail_per_vf >= ICE_NUM_VF_MSIX_MULTIQ_MIN) {
453 num_msix_per_vf = ICE_NUM_VF_MSIX_MULTIQ_MIN;
454 } else if (msix_avail_per_vf >= ICE_MIN_INTR_PER_VF) {
455 num_msix_per_vf = ICE_MIN_INTR_PER_VF;
457 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",
458 msix_avail_for_sriov, ICE_MIN_INTR_PER_VF,
463 num_txq = min_t(u16, num_msix_per_vf - ICE_NONQ_VECS_VF,
464 ICE_MAX_RSS_QS_PER_VF);
465 avail_qs = ice_get_avail_txq_count(pf) / num_vfs;
468 else if (num_txq > avail_qs)
469 num_txq = rounddown_pow_of_two(avail_qs);
471 num_rxq = min_t(u16, num_msix_per_vf - ICE_NONQ_VECS_VF,
472 ICE_MAX_RSS_QS_PER_VF);
473 avail_qs = ice_get_avail_rxq_count(pf) / num_vfs;
476 else if (num_rxq > avail_qs)
477 num_rxq = rounddown_pow_of_two(avail_qs);
479 if (num_txq < ICE_MIN_QS_PER_VF || num_rxq < ICE_MIN_QS_PER_VF) {
480 dev_err(dev, "Not enough queues to support minimum of %d queue pairs per VF for %d VFs\n",
481 ICE_MIN_QS_PER_VF, num_vfs);
485 err = ice_sriov_set_msix_res(pf, num_msix_per_vf * num_vfs);
487 dev_err(dev, "Unable to set MSI-X resources for %d VFs, err %d\n",
492 /* only allow equal Tx/Rx queue count (i.e. queue pairs) */
493 pf->vfs.num_qps_per = min_t(int, num_txq, num_rxq);
494 pf->vfs.num_msix_per = num_msix_per_vf;
495 dev_info(dev, "Enabling %d VFs with %d vectors and %d queues per VF\n",
496 num_vfs, pf->vfs.num_msix_per, pf->vfs.num_qps_per);
502 * ice_sriov_get_irqs - get irqs for SR-IOV usacase
503 * @pf: pointer to PF structure
504 * @needed: number of irqs to get
506 * This returns the first MSI-X vector index in PF space that is used by this
507 * VF. This index is used when accessing PF relative registers such as
508 * GLINT_VECT2FUNC and GLINT_DYN_CTL.
509 * This will always be the OICR index in the AVF driver so any functionality
510 * using vf->first_vector_idx for queue configuration_id: id of VF which will
513 * Only SRIOV specific vectors are tracked in sriov_irq_bm. SRIOV vectors are
514 * allocated from the end of global irq index. First bit in sriov_irq_bm means
515 * last irq index etc. It simplifies extension of SRIOV vectors.
516 * They will be always located from sriov_base_vector to the last irq
517 * index. While increasing/decreasing sriov_base_vector can be moved.
519 static int ice_sriov_get_irqs(struct ice_pf *pf, u16 needed)
521 int res = bitmap_find_next_zero_area(pf->sriov_irq_bm,
522 pf->sriov_irq_size, 0, needed, 0);
523 /* conversion from number in bitmap to global irq index */
524 int index = pf->sriov_irq_size - res - needed;
526 if (res >= pf->sriov_irq_size || index < pf->sriov_base_vector)
529 bitmap_set(pf->sriov_irq_bm, res, needed);
534 * ice_sriov_free_irqs - free irqs used by the VF
535 * @pf: pointer to PF structure
536 * @vf: pointer to VF structure
538 static void ice_sriov_free_irqs(struct ice_pf *pf, struct ice_vf *vf)
540 /* Move back from first vector index to first index in bitmap */
541 int bm_i = pf->sriov_irq_size - vf->first_vector_idx - vf->num_msix;
543 bitmap_clear(pf->sriov_irq_bm, bm_i, vf->num_msix);
544 vf->first_vector_idx = 0;
548 * ice_init_vf_vsi_res - initialize/setup VF VSI resources
549 * @vf: VF to initialize/setup the VSI for
551 * This function creates a VSI for the VF, adds a VLAN 0 filter, and sets up the
552 * VF VSI's broadcast filter and is only used during initial VF creation.
554 static int ice_init_vf_vsi_res(struct ice_vf *vf)
556 struct ice_pf *pf = vf->pf;
560 vf->first_vector_idx = ice_sriov_get_irqs(pf, vf->num_msix);
561 if (vf->first_vector_idx < 0)
564 vsi = ice_vf_vsi_setup(vf);
568 err = ice_vf_init_host_cfg(vf, vsi);
575 ice_vf_vsi_release(vf);
580 * ice_start_vfs - start VFs so they are ready to be used by SR-IOV
581 * @pf: PF the VFs are associated with
583 static int ice_start_vfs(struct ice_pf *pf)
585 struct ice_hw *hw = &pf->hw;
586 unsigned int bkt, it_cnt;
590 lockdep_assert_held(&pf->vfs.table_lock);
593 ice_for_each_vf(pf, bkt, vf) {
594 vf->vf_ops->clear_reset_trigger(vf);
596 retval = ice_init_vf_vsi_res(vf);
598 dev_err(ice_pf_to_dev(pf), "Failed to initialize VSI resources for VF %d, error %d\n",
603 retval = ice_eswitch_attach(pf, vf);
605 dev_err(ice_pf_to_dev(pf), "Failed to attach VF %d to eswitch, error %d",
607 ice_vf_vsi_release(vf);
611 set_bit(ICE_VF_STATE_INIT, vf->vf_states);
612 ice_ena_vf_mappings(vf);
613 wr32(hw, VFGEN_RSTAT(vf->vf_id), VIRTCHNL_VFR_VFACTIVE);
621 ice_for_each_vf(pf, bkt, vf) {
625 ice_dis_vf_mappings(vf);
626 ice_vf_vsi_release(vf);
634 * ice_sriov_free_vf - Free VF memory after all references are dropped
635 * @vf: pointer to VF to free
637 * Called by ice_put_vf through ice_release_vf once the last reference to a VF
638 * structure has been dropped.
640 static void ice_sriov_free_vf(struct ice_vf *vf)
642 mutex_destroy(&vf->cfg_lock);
648 * ice_sriov_clear_reset_state - clears VF Reset status register
649 * @vf: the vf to configure
651 static void ice_sriov_clear_reset_state(struct ice_vf *vf)
653 struct ice_hw *hw = &vf->pf->hw;
655 /* Clear the reset status register so that VF immediately sees that
656 * the device is resetting, even if hardware hasn't yet gotten around
657 * to clearing VFGEN_RSTAT for us.
659 wr32(hw, VFGEN_RSTAT(vf->vf_id), VIRTCHNL_VFR_INPROGRESS);
663 * ice_sriov_clear_mbx_register - clears SRIOV VF's mailbox registers
664 * @vf: the vf to configure
666 static void ice_sriov_clear_mbx_register(struct ice_vf *vf)
668 struct ice_pf *pf = vf->pf;
670 wr32(&pf->hw, VF_MBX_ARQLEN(vf->vf_id), 0);
671 wr32(&pf->hw, VF_MBX_ATQLEN(vf->vf_id), 0);
675 * ice_sriov_trigger_reset_register - trigger VF reset for SRIOV VF
676 * @vf: pointer to VF structure
677 * @is_vflr: true if reset occurred due to VFLR
679 * Trigger and cleanup after a VF reset for a SR-IOV VF.
681 static void ice_sriov_trigger_reset_register(struct ice_vf *vf, bool is_vflr)
683 struct ice_pf *pf = vf->pf;
684 u32 reg, reg_idx, bit_idx;
685 unsigned int vf_abs_id, i;
689 dev = ice_pf_to_dev(pf);
691 vf_abs_id = vf->vf_id + hw->func_caps.vf_base_id;
693 /* In the case of a VFLR, HW has already reset the VF and we just need
694 * to clean up. Otherwise we must first trigger the reset using the
698 reg = rd32(hw, VPGEN_VFRTRIG(vf->vf_id));
699 reg |= VPGEN_VFRTRIG_VFSWR_M;
700 wr32(hw, VPGEN_VFRTRIG(vf->vf_id), reg);
703 /* clear the VFLR bit in GLGEN_VFLRSTAT */
704 reg_idx = (vf_abs_id) / 32;
705 bit_idx = (vf_abs_id) % 32;
706 wr32(hw, GLGEN_VFLRSTAT(reg_idx), BIT(bit_idx));
709 wr32(hw, PF_PCI_CIAA,
710 VF_DEVICE_STATUS | (vf_abs_id << PF_PCI_CIAA_VF_NUM_S));
711 for (i = 0; i < ICE_PCI_CIAD_WAIT_COUNT; i++) {
712 reg = rd32(hw, PF_PCI_CIAD);
713 /* no transactions pending so stop polling */
714 if ((reg & VF_TRANS_PENDING_M) == 0)
717 dev_err(dev, "VF %u PCI transactions stuck\n", vf->vf_id);
718 udelay(ICE_PCI_CIAD_WAIT_DELAY_US);
723 * ice_sriov_poll_reset_status - poll SRIOV VF reset status
724 * @vf: pointer to VF structure
726 * Returns true when reset is successful, else returns false
728 static bool ice_sriov_poll_reset_status(struct ice_vf *vf)
730 struct ice_pf *pf = vf->pf;
734 for (i = 0; i < 10; i++) {
735 /* VF reset requires driver to first reset the VF and then
736 * poll the status register to make sure that the reset
737 * completed successfully.
739 reg = rd32(&pf->hw, VPGEN_VFRSTAT(vf->vf_id));
740 if (reg & VPGEN_VFRSTAT_VFRD_M)
743 /* only sleep if the reset is not done */
744 usleep_range(10, 20);
750 * ice_sriov_clear_reset_trigger - enable VF to access hardware
751 * @vf: VF to enabled hardware access for
753 static void ice_sriov_clear_reset_trigger(struct ice_vf *vf)
755 struct ice_hw *hw = &vf->pf->hw;
758 reg = rd32(hw, VPGEN_VFRTRIG(vf->vf_id));
759 reg &= ~VPGEN_VFRTRIG_VFSWR_M;
760 wr32(hw, VPGEN_VFRTRIG(vf->vf_id), reg);
765 * ice_sriov_post_vsi_rebuild - tasks to do after the VF's VSI have been rebuilt
766 * @vf: VF to perform tasks on
768 static void ice_sriov_post_vsi_rebuild(struct ice_vf *vf)
770 ice_ena_vf_mappings(vf);
771 wr32(&vf->pf->hw, VFGEN_RSTAT(vf->vf_id), VIRTCHNL_VFR_VFACTIVE);
774 static const struct ice_vf_ops ice_sriov_vf_ops = {
775 .reset_type = ICE_VF_RESET,
776 .free = ice_sriov_free_vf,
777 .clear_reset_state = ice_sriov_clear_reset_state,
778 .clear_mbx_register = ice_sriov_clear_mbx_register,
779 .trigger_reset_register = ice_sriov_trigger_reset_register,
780 .poll_reset_status = ice_sriov_poll_reset_status,
781 .clear_reset_trigger = ice_sriov_clear_reset_trigger,
783 .post_vsi_rebuild = ice_sriov_post_vsi_rebuild,
787 * ice_create_vf_entries - Allocate and insert VF entries
788 * @pf: pointer to the PF structure
789 * @num_vfs: the number of VFs to allocate
791 * Allocate new VF entries and insert them into the hash table. Set some
792 * basic default fields for initializing the new VFs.
794 * After this function exits, the hash table will have num_vfs entries
797 * Returns 0 on success or an integer error code on failure.
799 static int ice_create_vf_entries(struct ice_pf *pf, u16 num_vfs)
801 struct pci_dev *pdev = pf->pdev;
802 struct ice_vfs *vfs = &pf->vfs;
803 struct pci_dev *vfdev = NULL;
808 lockdep_assert_held(&vfs->table_lock);
810 pos = pci_find_ext_capability(pdev, PCI_EXT_CAP_ID_SRIOV);
811 pci_read_config_word(pdev, pos + PCI_SRIOV_VF_DID, &vf_pdev_id);
813 for (u16 vf_id = 0; vf_id < num_vfs; vf_id++) {
814 vf = kzalloc(sizeof(*vf), GFP_KERNEL);
817 goto err_free_entries;
819 kref_init(&vf->refcnt);
824 /* set sriov vf ops for VFs created during SRIOV flow */
825 vf->vf_ops = &ice_sriov_vf_ops;
827 ice_initialize_vf_entry(vf);
830 vfdev = pci_get_device(pdev->vendor, vf_pdev_id, vfdev);
831 } while (vfdev && vfdev->physfn != pdev);
833 vf->vf_sw_id = pf->first_sw;
837 /* set default number of MSI-X */
838 vf->num_msix = pf->vfs.num_msix_per;
839 vf->num_vf_qs = pf->vfs.num_qps_per;
840 ice_vc_set_default_allowlist(vf);
842 hash_add_rcu(vfs->table, &vf->entry, vf_id);
845 /* Decrement of refcount done by pci_get_device() inside the loop does
846 * not touch the last iteration's vfdev, so it has to be done manually
847 * to balance pci_dev_get() added within the loop.
854 ice_free_vf_entries(pf);
859 * ice_ena_vfs - enable VFs so they are ready to be used
860 * @pf: pointer to the PF structure
861 * @num_vfs: number of VFs to enable
863 static int ice_ena_vfs(struct ice_pf *pf, u16 num_vfs)
865 int total_vectors = pf->hw.func_caps.common_cap.num_msix_vectors;
866 struct device *dev = ice_pf_to_dev(pf);
867 struct ice_hw *hw = &pf->hw;
870 pf->sriov_irq_bm = bitmap_zalloc(total_vectors, GFP_KERNEL);
871 if (!pf->sriov_irq_bm)
873 pf->sriov_irq_size = total_vectors;
875 /* Disable global interrupt 0 so we don't try to handle the VFLR. */
876 wr32(hw, GLINT_DYN_CTL(pf->oicr_irq.index),
877 ICE_ITR_NONE << GLINT_DYN_CTL_ITR_INDX_S);
878 set_bit(ICE_OICR_INTR_DIS, pf->state);
881 ret = pci_enable_sriov(pf->pdev, num_vfs);
883 goto err_unroll_intr;
885 mutex_lock(&pf->vfs.table_lock);
887 ret = ice_set_per_vf_res(pf, num_vfs);
889 dev_err(dev, "Not enough resources for %d VFs, err %d. Try with fewer number of VFs\n",
891 goto err_unroll_sriov;
894 ret = ice_create_vf_entries(pf, num_vfs);
896 dev_err(dev, "Failed to allocate VF entries for %d VFs\n",
898 goto err_unroll_sriov;
901 ice_eswitch_reserve_cp_queues(pf, num_vfs);
902 ret = ice_start_vfs(pf);
904 dev_err(dev, "Failed to start %d VFs, err %d\n", num_vfs, ret);
906 goto err_unroll_vf_entries;
909 clear_bit(ICE_VF_DIS, pf->state);
911 /* rearm global interrupts */
912 if (test_and_clear_bit(ICE_OICR_INTR_DIS, pf->state))
913 ice_irq_dynamic_ena(hw, NULL, NULL);
915 mutex_unlock(&pf->vfs.table_lock);
919 err_unroll_vf_entries:
920 ice_free_vf_entries(pf);
922 mutex_unlock(&pf->vfs.table_lock);
923 pci_disable_sriov(pf->pdev);
925 /* rearm interrupts here */
926 ice_irq_dynamic_ena(hw, NULL, NULL);
927 clear_bit(ICE_OICR_INTR_DIS, pf->state);
928 bitmap_free(pf->sriov_irq_bm);
933 * ice_pci_sriov_ena - Enable or change number of VFs
934 * @pf: pointer to the PF structure
935 * @num_vfs: number of VFs to allocate
937 * Returns 0 on success and negative on failure
939 static int ice_pci_sriov_ena(struct ice_pf *pf, int num_vfs)
941 struct device *dev = ice_pf_to_dev(pf);
949 if (num_vfs > pf->vfs.num_supported) {
950 dev_err(dev, "Can't enable %d VFs, max VFs supported is %d\n",
951 num_vfs, pf->vfs.num_supported);
955 dev_info(dev, "Enabling %d VFs\n", num_vfs);
956 err = ice_ena_vfs(pf, num_vfs);
958 dev_err(dev, "Failed to enable SR-IOV: %d\n", err);
962 set_bit(ICE_FLAG_SRIOV_ENA, pf->flags);
967 * ice_check_sriov_allowed - check if SR-IOV is allowed based on various checks
968 * @pf: PF to enabled SR-IOV on
970 static int ice_check_sriov_allowed(struct ice_pf *pf)
972 struct device *dev = ice_pf_to_dev(pf);
974 if (!test_bit(ICE_FLAG_SRIOV_CAPABLE, pf->flags)) {
975 dev_err(dev, "This device is not capable of SR-IOV\n");
979 if (ice_is_safe_mode(pf)) {
980 dev_err(dev, "SR-IOV cannot be configured - Device is in Safe Mode\n");
984 if (!ice_pf_state_is_nominal(pf)) {
985 dev_err(dev, "Cannot enable SR-IOV, device not ready\n");
993 * ice_sriov_get_vf_total_msix - return number of MSI-X used by VFs
994 * @pdev: pointer to pci_dev struct
996 * The function is called via sysfs ops
998 u32 ice_sriov_get_vf_total_msix(struct pci_dev *pdev)
1000 struct ice_pf *pf = pci_get_drvdata(pdev);
1002 return pf->sriov_irq_size - ice_get_max_used_msix_vector(pf);
1005 static int ice_sriov_move_base_vector(struct ice_pf *pf, int move)
1007 if (pf->sriov_base_vector - move < ice_get_max_used_msix_vector(pf))
1010 pf->sriov_base_vector -= move;
1014 static void ice_sriov_remap_vectors(struct ice_pf *pf, u16 restricted_id)
1016 u16 vf_ids[ICE_MAX_SRIOV_VFS];
1017 struct ice_vf *tmp_vf;
1018 int to_remap = 0, bkt;
1020 /* For better irqs usage try to remap irqs of VFs
1021 * that aren't running yet
1023 ice_for_each_vf(pf, bkt, tmp_vf) {
1024 /* skip VF which is changing the number of MSI-X */
1025 if (restricted_id == tmp_vf->vf_id ||
1026 test_bit(ICE_VF_STATE_ACTIVE, tmp_vf->vf_states))
1029 ice_dis_vf_mappings(tmp_vf);
1030 ice_sriov_free_irqs(pf, tmp_vf);
1032 vf_ids[to_remap] = tmp_vf->vf_id;
1036 for (int i = 0; i < to_remap; i++) {
1037 tmp_vf = ice_get_vf_by_id(pf, vf_ids[i]);
1041 tmp_vf->first_vector_idx =
1042 ice_sriov_get_irqs(pf, tmp_vf->num_msix);
1043 /* there is no need to rebuild VSI as we are only changing the
1044 * vector indexes not amount of MSI-X or queues
1046 ice_ena_vf_mappings(tmp_vf);
1052 * ice_sriov_set_msix_vec_count
1053 * @vf_dev: pointer to pci_dev struct of VF device
1054 * @msix_vec_count: new value for MSI-X amount on this VF
1056 * Set requested MSI-X, queues and registers for @vf_dev.
1058 * First do some sanity checks like if there are any VFs, if the new value
1059 * is correct etc. Then disable old mapping (MSI-X and queues registers), change
1060 * MSI-X and queues, rebuild VSI and enable new mapping.
1062 * If it is possible (driver not binded to VF) try to remap also other VFs to
1063 * linearize irqs register usage.
1065 int ice_sriov_set_msix_vec_count(struct pci_dev *vf_dev, int msix_vec_count)
1067 struct pci_dev *pdev = pci_physfn(vf_dev);
1068 struct ice_pf *pf = pci_get_drvdata(pdev);
1069 u16 prev_msix, prev_queues, queues;
1070 bool needs_rebuild = false;
1071 struct ice_vsi *vsi;
1075 if (!ice_get_num_vfs(pf))
1078 if (!msix_vec_count)
1081 queues = msix_vec_count;
1082 /* add 1 MSI-X for OICR */
1083 msix_vec_count += 1;
1085 if (queues > min(ice_get_avail_txq_count(pf),
1086 ice_get_avail_rxq_count(pf)))
1089 if (msix_vec_count < ICE_MIN_INTR_PER_VF)
1092 /* Transition of PCI VF function number to function_id */
1093 for (id = 0; id < pci_num_vf(pdev); id++) {
1094 if (vf_dev->devfn == pci_iov_virtfn_devfn(pdev, id))
1098 if (id == pci_num_vf(pdev))
1101 vf = ice_get_vf_by_id(pf, id);
1106 vsi = ice_get_vf_vsi(vf);
1110 prev_msix = vf->num_msix;
1111 prev_queues = vf->num_vf_qs;
1113 if (ice_sriov_move_base_vector(pf, msix_vec_count - prev_msix)) {
1118 ice_dis_vf_mappings(vf);
1119 ice_sriov_free_irqs(pf, vf);
1121 /* Remap all VFs beside the one is now configured */
1122 ice_sriov_remap_vectors(pf, vf->vf_id);
1124 vf->num_msix = msix_vec_count;
1125 vf->num_vf_qs = queues;
1126 vf->first_vector_idx = ice_sriov_get_irqs(pf, vf->num_msix);
1127 if (vf->first_vector_idx < 0)
1130 if (ice_vf_reconfig_vsi(vf) || ice_vf_init_host_cfg(vf, vsi)) {
1131 /* Try to rebuild with previous values */
1132 needs_rebuild = true;
1136 dev_info(ice_pf_to_dev(pf),
1137 "Changing VF %d resources to %d vectors and %d queues\n",
1138 vf->vf_id, vf->num_msix, vf->num_vf_qs);
1140 ice_ena_vf_mappings(vf);
1146 dev_info(ice_pf_to_dev(pf),
1147 "Can't set %d vectors on VF %d, falling back to %d\n",
1148 vf->num_msix, vf->vf_id, prev_msix);
1150 vf->num_msix = prev_msix;
1151 vf->num_vf_qs = prev_queues;
1152 vf->first_vector_idx = ice_sriov_get_irqs(pf, vf->num_msix);
1153 if (vf->first_vector_idx < 0)
1156 if (needs_rebuild) {
1157 ice_vf_reconfig_vsi(vf);
1158 ice_vf_init_host_cfg(vf, vsi);
1161 ice_ena_vf_mappings(vf);
1168 * ice_sriov_configure - Enable or change number of VFs via sysfs
1169 * @pdev: pointer to a pci_dev structure
1170 * @num_vfs: number of VFs to allocate or 0 to free VFs
1172 * This function is called when the user updates the number of VFs in sysfs. On
1173 * success return whatever num_vfs was set to by the caller. Return negative on
1176 int ice_sriov_configure(struct pci_dev *pdev, int num_vfs)
1178 struct ice_pf *pf = pci_get_drvdata(pdev);
1179 struct device *dev = ice_pf_to_dev(pf);
1182 err = ice_check_sriov_allowed(pf);
1187 if (!pci_vfs_assigned(pdev)) {
1192 dev_err(dev, "can't free VFs because some are assigned to VMs.\n");
1196 err = ice_pci_sriov_ena(pf, num_vfs);
1204 * ice_process_vflr_event - Free VF resources via IRQ calls
1205 * @pf: pointer to the PF structure
1207 * called from the VFLR IRQ handler to
1208 * free up VF resources and state variables
1210 void ice_process_vflr_event(struct ice_pf *pf)
1212 struct ice_hw *hw = &pf->hw;
1217 if (!test_and_clear_bit(ICE_VFLR_EVENT_PENDING, pf->state) ||
1221 mutex_lock(&pf->vfs.table_lock);
1222 ice_for_each_vf(pf, bkt, vf) {
1223 u32 reg_idx, bit_idx;
1225 reg_idx = (hw->func_caps.vf_base_id + vf->vf_id) / 32;
1226 bit_idx = (hw->func_caps.vf_base_id + vf->vf_id) % 32;
1227 /* read GLGEN_VFLRSTAT register to find out the flr VFs */
1228 reg = rd32(hw, GLGEN_VFLRSTAT(reg_idx));
1229 if (reg & BIT(bit_idx))
1230 /* GLGEN_VFLRSTAT bit will be cleared in ice_reset_vf */
1231 ice_reset_vf(vf, ICE_VF_RESET_VFLR | ICE_VF_RESET_LOCK);
1233 mutex_unlock(&pf->vfs.table_lock);
1237 * ice_get_vf_from_pfq - get the VF who owns the PF space queue passed in
1238 * @pf: PF used to index all VFs
1239 * @pfq: queue index relative to the PF's function space
1241 * If no VF is found who owns the pfq then return NULL, otherwise return a
1242 * pointer to the VF who owns the pfq
1244 * If this function returns non-NULL, it acquires a reference count of the VF
1245 * structure. The caller is responsible for calling ice_put_vf() to drop this
1248 static struct ice_vf *ice_get_vf_from_pfq(struct ice_pf *pf, u16 pfq)
1254 ice_for_each_vf_rcu(pf, bkt, vf) {
1255 struct ice_vsi *vsi;
1258 vsi = ice_get_vf_vsi(vf);
1262 ice_for_each_rxq(vsi, rxq_idx)
1263 if (vsi->rxq_map[rxq_idx] == pfq) {
1264 struct ice_vf *found;
1266 if (kref_get_unless_zero(&vf->refcnt))
1280 * ice_globalq_to_pfq - convert from global queue index to PF space queue index
1281 * @pf: PF used for conversion
1282 * @globalq: global queue index used to convert to PF space queue index
1284 static u32 ice_globalq_to_pfq(struct ice_pf *pf, u32 globalq)
1286 return globalq - pf->hw.func_caps.common_cap.rxq_first_id;
1290 * ice_vf_lan_overflow_event - handle LAN overflow event for a VF
1291 * @pf: PF that the LAN overflow event happened on
1292 * @event: structure holding the event information for the LAN overflow event
1294 * Determine if the LAN overflow event was caused by a VF queue. If it was not
1295 * caused by a VF, do nothing. If a VF caused this LAN overflow event trigger a
1296 * reset on the offending VF.
1299 ice_vf_lan_overflow_event(struct ice_pf *pf, struct ice_rq_event_info *event)
1301 u32 gldcb_rtctq, queue;
1304 gldcb_rtctq = le32_to_cpu(event->desc.params.lan_overflow.prtdcb_ruptq);
1305 dev_dbg(ice_pf_to_dev(pf), "GLDCB_RTCTQ: 0x%08x\n", gldcb_rtctq);
1307 /* event returns device global Rx queue number */
1308 queue = FIELD_GET(GLDCB_RTCTQ_RXQNUM_M, gldcb_rtctq);
1310 vf = ice_get_vf_from_pfq(pf, ice_globalq_to_pfq(pf, queue));
1314 ice_reset_vf(vf, ICE_VF_RESET_NOTIFY | ICE_VF_RESET_LOCK);
1319 * ice_set_vf_spoofchk
1320 * @netdev: network interface device structure
1321 * @vf_id: VF identifier
1322 * @ena: flag to enable or disable feature
1324 * Enable or disable VF spoof checking
1326 int ice_set_vf_spoofchk(struct net_device *netdev, int vf_id, bool ena)
1328 struct ice_netdev_priv *np = netdev_priv(netdev);
1329 struct ice_pf *pf = np->vsi->back;
1330 struct ice_vsi *vf_vsi;
1335 dev = ice_pf_to_dev(pf);
1337 vf = ice_get_vf_by_id(pf, vf_id);
1341 ret = ice_check_vf_ready_for_cfg(vf);
1345 vf_vsi = ice_get_vf_vsi(vf);
1347 netdev_err(netdev, "VSI %d for VF %d is null\n",
1348 vf->lan_vsi_idx, vf->vf_id);
1353 if (vf_vsi->type != ICE_VSI_VF) {
1354 netdev_err(netdev, "Type %d of VSI %d for VF %d is no ICE_VSI_VF\n",
1355 vf_vsi->type, vf_vsi->vsi_num, vf->vf_id);
1360 if (ena == vf->spoofchk) {
1361 dev_dbg(dev, "VF spoofchk already %s\n", ena ? "ON" : "OFF");
1366 ret = ice_vsi_apply_spoofchk(vf_vsi, ena);
1368 dev_err(dev, "Failed to set spoofchk %s for VF %d VSI %d\n error %d\n",
1369 ena ? "ON" : "OFF", vf->vf_id, vf_vsi->vsi_num, ret);
1380 * @netdev: network interface device structure
1381 * @vf_id: VF identifier
1382 * @ivi: VF configuration structure
1384 * return VF configuration
1387 ice_get_vf_cfg(struct net_device *netdev, int vf_id, struct ifla_vf_info *ivi)
1389 struct ice_pf *pf = ice_netdev_to_pf(netdev);
1393 vf = ice_get_vf_by_id(pf, vf_id);
1397 ret = ice_check_vf_ready_for_cfg(vf);
1402 ether_addr_copy(ivi->mac, vf->hw_lan_addr);
1404 /* VF configuration for VLAN and applicable QoS */
1405 ivi->vlan = ice_vf_get_port_vlan_id(vf);
1406 ivi->qos = ice_vf_get_port_vlan_prio(vf);
1407 if (ice_vf_is_port_vlan_ena(vf))
1408 ivi->vlan_proto = cpu_to_be16(ice_vf_get_port_vlan_tpid(vf));
1410 ivi->trusted = vf->trusted;
1411 ivi->spoofchk = vf->spoofchk;
1412 if (!vf->link_forced)
1413 ivi->linkstate = IFLA_VF_LINK_STATE_AUTO;
1414 else if (vf->link_up)
1415 ivi->linkstate = IFLA_VF_LINK_STATE_ENABLE;
1417 ivi->linkstate = IFLA_VF_LINK_STATE_DISABLE;
1418 ivi->max_tx_rate = vf->max_tx_rate;
1419 ivi->min_tx_rate = vf->min_tx_rate;
1428 * @netdev: network interface device structure
1429 * @vf_id: VF identifier
1432 * program VF MAC address
1434 int ice_set_vf_mac(struct net_device *netdev, int vf_id, u8 *mac)
1436 struct ice_pf *pf = ice_netdev_to_pf(netdev);
1440 if (is_multicast_ether_addr(mac)) {
1441 netdev_err(netdev, "%pM not a valid unicast address\n", mac);
1445 vf = ice_get_vf_by_id(pf, vf_id);
1449 /* nothing left to do, unicast MAC already set */
1450 if (ether_addr_equal(vf->dev_lan_addr, mac) &&
1451 ether_addr_equal(vf->hw_lan_addr, mac)) {
1456 ret = ice_check_vf_ready_for_cfg(vf);
1460 mutex_lock(&vf->cfg_lock);
1462 /* VF is notified of its new MAC via the PF's response to the
1463 * VIRTCHNL_OP_GET_VF_RESOURCES message after the VF has been reset
1465 ether_addr_copy(vf->dev_lan_addr, mac);
1466 ether_addr_copy(vf->hw_lan_addr, mac);
1467 if (is_zero_ether_addr(mac)) {
1468 /* VF will send VIRTCHNL_OP_ADD_ETH_ADDR message with its MAC */
1469 vf->pf_set_mac = false;
1470 netdev_info(netdev, "Removing MAC on VF %d. VF driver will be reinitialized\n",
1473 /* PF will add MAC rule for the VF */
1474 vf->pf_set_mac = true;
1475 netdev_info(netdev, "Setting MAC %pM on VF %d. VF driver will be reinitialized\n",
1479 ice_reset_vf(vf, ICE_VF_RESET_NOTIFY);
1480 mutex_unlock(&vf->cfg_lock);
1489 * @netdev: network interface device structure
1490 * @vf_id: VF identifier
1491 * @trusted: Boolean value to enable/disable trusted VF
1493 * Enable or disable a given VF as trusted
1495 int ice_set_vf_trust(struct net_device *netdev, int vf_id, bool trusted)
1497 struct ice_pf *pf = ice_netdev_to_pf(netdev);
1501 vf = ice_get_vf_by_id(pf, vf_id);
1505 if (ice_is_eswitch_mode_switchdev(pf)) {
1506 dev_info(ice_pf_to_dev(pf), "Trusted VF is forbidden in switchdev mode\n");
1510 ret = ice_check_vf_ready_for_cfg(vf);
1514 /* Check if already trusted */
1515 if (trusted == vf->trusted) {
1520 mutex_lock(&vf->cfg_lock);
1522 vf->trusted = trusted;
1523 ice_reset_vf(vf, ICE_VF_RESET_NOTIFY);
1524 dev_info(ice_pf_to_dev(pf), "VF %u is now %strusted\n",
1525 vf_id, trusted ? "" : "un");
1527 mutex_unlock(&vf->cfg_lock);
1535 * ice_set_vf_link_state
1536 * @netdev: network interface device structure
1537 * @vf_id: VF identifier
1538 * @link_state: required link state
1540 * Set VF's link state, irrespective of physical link state status
1542 int ice_set_vf_link_state(struct net_device *netdev, int vf_id, int link_state)
1544 struct ice_pf *pf = ice_netdev_to_pf(netdev);
1548 vf = ice_get_vf_by_id(pf, vf_id);
1552 ret = ice_check_vf_ready_for_cfg(vf);
1556 switch (link_state) {
1557 case IFLA_VF_LINK_STATE_AUTO:
1558 vf->link_forced = false;
1560 case IFLA_VF_LINK_STATE_ENABLE:
1561 vf->link_forced = true;
1564 case IFLA_VF_LINK_STATE_DISABLE:
1565 vf->link_forced = true;
1566 vf->link_up = false;
1573 ice_vc_notify_vf_link_state(vf);
1581 * ice_calc_all_vfs_min_tx_rate - calculate cumulative min Tx rate on all VFs
1582 * @pf: PF associated with VFs
1584 static int ice_calc_all_vfs_min_tx_rate(struct ice_pf *pf)
1591 ice_for_each_vf_rcu(pf, bkt, vf)
1592 rate += vf->min_tx_rate;
1599 * ice_min_tx_rate_oversubscribed - check if min Tx rate causes oversubscription
1600 * @vf: VF trying to configure min_tx_rate
1601 * @min_tx_rate: min Tx rate in Mbps
1603 * Check if the min_tx_rate being passed in will cause oversubscription of total
1604 * min_tx_rate based on the current link speed and all other VFs configured
1607 * Return true if the passed min_tx_rate would cause oversubscription, else
1611 ice_min_tx_rate_oversubscribed(struct ice_vf *vf, int min_tx_rate)
1613 struct ice_vsi *vsi = ice_get_vf_vsi(vf);
1614 int all_vfs_min_tx_rate;
1615 int link_speed_mbps;
1620 link_speed_mbps = ice_get_link_speed_mbps(vsi);
1621 all_vfs_min_tx_rate = ice_calc_all_vfs_min_tx_rate(vf->pf);
1623 /* this VF's previous rate is being overwritten */
1624 all_vfs_min_tx_rate -= vf->min_tx_rate;
1626 if (all_vfs_min_tx_rate + min_tx_rate > link_speed_mbps) {
1627 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",
1628 min_tx_rate, vf->vf_id,
1629 all_vfs_min_tx_rate + min_tx_rate - link_speed_mbps,
1638 * ice_set_vf_bw - set min/max VF bandwidth
1639 * @netdev: network interface device structure
1640 * @vf_id: VF identifier
1641 * @min_tx_rate: Minimum Tx rate in Mbps
1642 * @max_tx_rate: Maximum Tx rate in Mbps
1645 ice_set_vf_bw(struct net_device *netdev, int vf_id, int min_tx_rate,
1648 struct ice_pf *pf = ice_netdev_to_pf(netdev);
1649 struct ice_vsi *vsi;
1654 dev = ice_pf_to_dev(pf);
1656 vf = ice_get_vf_by_id(pf, vf_id);
1660 ret = ice_check_vf_ready_for_cfg(vf);
1664 vsi = ice_get_vf_vsi(vf);
1670 if (min_tx_rate && ice_is_dcb_active(pf)) {
1671 dev_err(dev, "DCB on PF is currently enabled. VF min Tx rate limiting not allowed on this PF.\n");
1676 if (ice_min_tx_rate_oversubscribed(vf, min_tx_rate)) {
1681 if (vf->min_tx_rate != (unsigned int)min_tx_rate) {
1682 ret = ice_set_min_bw_limit(vsi, (u64)min_tx_rate * 1000);
1684 dev_err(dev, "Unable to set min-tx-rate for VF %d\n",
1689 vf->min_tx_rate = min_tx_rate;
1692 if (vf->max_tx_rate != (unsigned int)max_tx_rate) {
1693 ret = ice_set_max_bw_limit(vsi, (u64)max_tx_rate * 1000);
1695 dev_err(dev, "Unable to set max-tx-rate for VF %d\n",
1700 vf->max_tx_rate = max_tx_rate;
1709 * ice_get_vf_stats - populate some stats for the VF
1710 * @netdev: the netdev of the PF
1711 * @vf_id: the host OS identifier (0-255)
1712 * @vf_stats: pointer to the OS memory to be initialized
1714 int ice_get_vf_stats(struct net_device *netdev, int vf_id,
1715 struct ifla_vf_stats *vf_stats)
1717 struct ice_pf *pf = ice_netdev_to_pf(netdev);
1718 struct ice_eth_stats *stats;
1719 struct ice_vsi *vsi;
1723 vf = ice_get_vf_by_id(pf, vf_id);
1727 ret = ice_check_vf_ready_for_cfg(vf);
1731 vsi = ice_get_vf_vsi(vf);
1737 ice_update_eth_stats(vsi);
1738 stats = &vsi->eth_stats;
1740 memset(vf_stats, 0, sizeof(*vf_stats));
1742 vf_stats->rx_packets = stats->rx_unicast + stats->rx_broadcast +
1743 stats->rx_multicast;
1744 vf_stats->tx_packets = stats->tx_unicast + stats->tx_broadcast +
1745 stats->tx_multicast;
1746 vf_stats->rx_bytes = stats->rx_bytes;
1747 vf_stats->tx_bytes = stats->tx_bytes;
1748 vf_stats->broadcast = stats->rx_broadcast;
1749 vf_stats->multicast = stats->rx_multicast;
1750 vf_stats->rx_dropped = stats->rx_discards;
1751 vf_stats->tx_dropped = stats->tx_discards;
1759 * ice_is_supported_port_vlan_proto - make sure the vlan_proto is supported
1760 * @hw: hardware structure used to check the VLAN mode
1761 * @vlan_proto: VLAN TPID being checked
1763 * If the device is configured in Double VLAN Mode (DVM), then both ETH_P_8021Q
1764 * and ETH_P_8021AD are supported. If the device is configured in Single VLAN
1765 * Mode (SVM), then only ETH_P_8021Q is supported.
1768 ice_is_supported_port_vlan_proto(struct ice_hw *hw, u16 vlan_proto)
1770 bool is_supported = false;
1772 switch (vlan_proto) {
1774 is_supported = true;
1777 if (ice_is_dvm_ena(hw))
1778 is_supported = true;
1782 return is_supported;
1786 * ice_set_vf_port_vlan
1787 * @netdev: network interface device structure
1788 * @vf_id: VF identifier
1789 * @vlan_id: VLAN ID being set
1790 * @qos: priority setting
1791 * @vlan_proto: VLAN protocol
1793 * program VF Port VLAN ID and/or QoS
1796 ice_set_vf_port_vlan(struct net_device *netdev, int vf_id, u16 vlan_id, u8 qos,
1799 struct ice_pf *pf = ice_netdev_to_pf(netdev);
1800 u16 local_vlan_proto = ntohs(vlan_proto);
1805 dev = ice_pf_to_dev(pf);
1807 if (vlan_id >= VLAN_N_VID || qos > 7) {
1808 dev_err(dev, "Invalid Port VLAN parameters for VF %d, ID %d, QoS %d\n",
1809 vf_id, vlan_id, qos);
1813 if (!ice_is_supported_port_vlan_proto(&pf->hw, local_vlan_proto)) {
1814 dev_err(dev, "VF VLAN protocol 0x%04x is not supported\n",
1816 return -EPROTONOSUPPORT;
1819 vf = ice_get_vf_by_id(pf, vf_id);
1823 ret = ice_check_vf_ready_for_cfg(vf);
1827 if (ice_vf_get_port_vlan_prio(vf) == qos &&
1828 ice_vf_get_port_vlan_tpid(vf) == local_vlan_proto &&
1829 ice_vf_get_port_vlan_id(vf) == vlan_id) {
1830 /* duplicate request, so just return success */
1831 dev_dbg(dev, "Duplicate port VLAN %u, QoS %u, TPID 0x%04x request\n",
1832 vlan_id, qos, local_vlan_proto);
1837 mutex_lock(&vf->cfg_lock);
1839 vf->port_vlan_info = ICE_VLAN(local_vlan_proto, vlan_id, qos);
1840 if (ice_vf_is_port_vlan_ena(vf))
1841 dev_info(dev, "Setting VLAN %u, QoS %u, TPID 0x%04x on VF %d\n",
1842 vlan_id, qos, local_vlan_proto, vf_id);
1844 dev_info(dev, "Clearing port VLAN on VF %d\n", vf_id);
1846 ice_reset_vf(vf, ICE_VF_RESET_NOTIFY);
1847 mutex_unlock(&vf->cfg_lock);
1855 * ice_print_vf_rx_mdd_event - print VF Rx malicious driver detect event
1856 * @vf: pointer to the VF structure
1858 void ice_print_vf_rx_mdd_event(struct ice_vf *vf)
1860 struct ice_pf *pf = vf->pf;
1863 dev = ice_pf_to_dev(pf);
1865 dev_info(dev, "%d Rx Malicious Driver Detection events detected on PF %d VF %d MAC %pM. mdd-auto-reset-vfs=%s\n",
1866 vf->mdd_rx_events.count, pf->hw.pf_id, vf->vf_id,
1868 test_bit(ICE_FLAG_MDD_AUTO_RESET_VF, pf->flags)
1873 * ice_print_vfs_mdd_events - print VFs malicious driver detect event
1874 * @pf: pointer to the PF structure
1876 * Called from ice_handle_mdd_event to rate limit and print VFs MDD events.
1878 void ice_print_vfs_mdd_events(struct ice_pf *pf)
1880 struct device *dev = ice_pf_to_dev(pf);
1881 struct ice_hw *hw = &pf->hw;
1885 /* check that there are pending MDD events to print */
1886 if (!test_and_clear_bit(ICE_MDD_VF_PRINT_PENDING, pf->state))
1889 /* VF MDD event logs are rate limited to one second intervals */
1890 if (time_is_after_jiffies(pf->vfs.last_printed_mdd_jiffies + HZ * 1))
1893 pf->vfs.last_printed_mdd_jiffies = jiffies;
1895 mutex_lock(&pf->vfs.table_lock);
1896 ice_for_each_vf(pf, bkt, vf) {
1897 /* only print Rx MDD event message if there are new events */
1898 if (vf->mdd_rx_events.count != vf->mdd_rx_events.last_printed) {
1899 vf->mdd_rx_events.last_printed =
1900 vf->mdd_rx_events.count;
1901 ice_print_vf_rx_mdd_event(vf);
1904 /* only print Tx MDD event message if there are new events */
1905 if (vf->mdd_tx_events.count != vf->mdd_tx_events.last_printed) {
1906 vf->mdd_tx_events.last_printed =
1907 vf->mdd_tx_events.count;
1909 dev_info(dev, "%d Tx Malicious Driver Detection events detected on PF %d VF %d MAC %pM.\n",
1910 vf->mdd_tx_events.count, hw->pf_id, vf->vf_id,
1914 mutex_unlock(&pf->vfs.table_lock);
1918 * ice_restore_all_vfs_msi_state - restore VF MSI state after PF FLR
1919 * @pf: pointer to the PF structure
1921 * Called when recovering from a PF FLR to restore interrupt capability to
1924 void ice_restore_all_vfs_msi_state(struct ice_pf *pf)
1929 ice_for_each_vf(pf, bkt, vf)
1930 pci_restore_msi_state(vf->vfdev);