1 // SPDX-License-Identifier: GPL-2.0
2 /* Copyright (c) 2018, Intel Corporation. */
9 #include "ice_dcb_lib.h"
10 #include "ice_devlink.h"
11 #include "ice_vsi_vlan_ops.h"
14 * ice_vsi_type_str - maps VSI type enum to string equivalents
15 * @vsi_type: VSI type enum
17 const char *ice_vsi_type_str(enum ice_vsi_type vsi_type)
25 return "ICE_VSI_CTRL";
27 return "ICE_VSI_CHNL";
30 case ICE_VSI_SWITCHDEV_CTRL:
31 return "ICE_VSI_SWITCHDEV_CTRL";
38 * ice_vsi_ctrl_all_rx_rings - Start or stop a VSI's Rx rings
39 * @vsi: the VSI being configured
40 * @ena: start or stop the Rx rings
42 * First enable/disable all of the Rx rings, flush any remaining writes, and
43 * then verify that they have all been enabled/disabled successfully. This will
44 * let all of the register writes complete when enabling/disabling the Rx rings
45 * before waiting for the change in hardware to complete.
47 static int ice_vsi_ctrl_all_rx_rings(struct ice_vsi *vsi, bool ena)
52 ice_for_each_rxq(vsi, i)
53 ice_vsi_ctrl_one_rx_ring(vsi, ena, i, false);
55 ice_flush(&vsi->back->hw);
57 ice_for_each_rxq(vsi, i) {
58 ret = ice_vsi_wait_one_rx_ring(vsi, ena, i);
67 * ice_vsi_alloc_arrays - Allocate queue and vector pointer arrays for the VSI
70 * On error: returns error code (negative)
71 * On success: returns 0
73 static int ice_vsi_alloc_arrays(struct ice_vsi *vsi)
75 struct ice_pf *pf = vsi->back;
78 dev = ice_pf_to_dev(pf);
79 if (vsi->type == ICE_VSI_CHNL)
82 /* allocate memory for both Tx and Rx ring pointers */
83 vsi->tx_rings = devm_kcalloc(dev, vsi->alloc_txq,
84 sizeof(*vsi->tx_rings), GFP_KERNEL);
88 vsi->rx_rings = devm_kcalloc(dev, vsi->alloc_rxq,
89 sizeof(*vsi->rx_rings), GFP_KERNEL);
93 /* txq_map needs to have enough space to track both Tx (stack) rings
94 * and XDP rings; at this point vsi->num_xdp_txq might not be set,
95 * so use num_possible_cpus() as we want to always provide XDP ring
96 * per CPU, regardless of queue count settings from user that might
97 * have come from ethtool's set_channels() callback;
99 vsi->txq_map = devm_kcalloc(dev, (vsi->alloc_txq + num_possible_cpus()),
100 sizeof(*vsi->txq_map), GFP_KERNEL);
105 vsi->rxq_map = devm_kcalloc(dev, vsi->alloc_rxq,
106 sizeof(*vsi->rxq_map), GFP_KERNEL);
110 /* There is no need to allocate q_vectors for a loopback VSI. */
111 if (vsi->type == ICE_VSI_LB)
114 /* allocate memory for q_vector pointers */
115 vsi->q_vectors = devm_kcalloc(dev, vsi->num_q_vectors,
116 sizeof(*vsi->q_vectors), GFP_KERNEL);
120 vsi->af_xdp_zc_qps = bitmap_zalloc(max_t(int, vsi->alloc_txq, vsi->alloc_rxq), GFP_KERNEL);
121 if (!vsi->af_xdp_zc_qps)
127 devm_kfree(dev, vsi->q_vectors);
129 devm_kfree(dev, vsi->rxq_map);
131 devm_kfree(dev, vsi->txq_map);
133 devm_kfree(dev, vsi->rx_rings);
135 devm_kfree(dev, vsi->tx_rings);
140 * ice_vsi_set_num_desc - Set number of descriptors for queues on this VSI
141 * @vsi: the VSI being configured
143 static void ice_vsi_set_num_desc(struct ice_vsi *vsi)
147 case ICE_VSI_SWITCHDEV_CTRL:
150 /* a user could change the values of num_[tr]x_desc using
151 * ethtool -G so we should keep those values instead of
152 * overwriting them with the defaults.
154 if (!vsi->num_rx_desc)
155 vsi->num_rx_desc = ICE_DFLT_NUM_RX_DESC;
156 if (!vsi->num_tx_desc)
157 vsi->num_tx_desc = ICE_DFLT_NUM_TX_DESC;
160 dev_dbg(ice_pf_to_dev(vsi->back), "Not setting number of Tx/Rx descriptors for VSI type %d\n",
167 * ice_vsi_set_num_qs - Set number of queues, descriptors and vectors for a VSI
168 * @vsi: the VSI being configured
170 * Return 0 on success and a negative value on error
172 static void ice_vsi_set_num_qs(struct ice_vsi *vsi)
174 enum ice_vsi_type vsi_type = vsi->type;
175 struct ice_pf *pf = vsi->back;
176 struct ice_vf *vf = vsi->vf;
178 if (WARN_ON(vsi_type == ICE_VSI_VF && !vf))
184 vsi->alloc_txq = vsi->req_txq;
185 vsi->num_txq = vsi->req_txq;
187 vsi->alloc_txq = min3(pf->num_lan_msix,
188 ice_get_avail_txq_count(pf),
189 (u16)num_online_cpus());
192 pf->num_lan_tx = vsi->alloc_txq;
194 /* only 1 Rx queue unless RSS is enabled */
195 if (!test_bit(ICE_FLAG_RSS_ENA, pf->flags)) {
199 vsi->alloc_rxq = vsi->req_rxq;
200 vsi->num_rxq = vsi->req_rxq;
202 vsi->alloc_rxq = min3(pf->num_lan_msix,
203 ice_get_avail_rxq_count(pf),
204 (u16)num_online_cpus());
208 pf->num_lan_rx = vsi->alloc_rxq;
210 vsi->num_q_vectors = min_t(int, pf->num_lan_msix,
211 max_t(int, vsi->alloc_rxq,
214 case ICE_VSI_SWITCHDEV_CTRL:
215 /* The number of queues for ctrl VSI is equal to number of PRs
216 * Each ring is associated to the corresponding VF_PR netdev.
217 * Tx and Rx rings are always equal
219 if (vsi->req_txq && vsi->req_rxq) {
220 vsi->alloc_txq = vsi->req_txq;
221 vsi->alloc_rxq = vsi->req_rxq;
227 vsi->num_q_vectors = 1;
231 vf->num_vf_qs = vf->num_req_qs;
232 vsi->alloc_txq = vf->num_vf_qs;
233 vsi->alloc_rxq = vf->num_vf_qs;
234 /* pf->vfs.num_msix_per includes (VF miscellaneous vector +
235 * data queue interrupts). Since vsi->num_q_vectors is number
236 * of queues vectors, subtract 1 (ICE_NONQ_VECS_VF) from the
237 * original vector count
239 vsi->num_q_vectors = vf->num_msix - ICE_NONQ_VECS_VF;
244 vsi->num_q_vectors = 1;
255 dev_warn(ice_pf_to_dev(pf), "Unknown VSI type %d\n", vsi_type);
259 ice_vsi_set_num_desc(vsi);
263 * ice_get_free_slot - get the next non-NULL location index in array
264 * @array: array to search
265 * @size: size of the array
266 * @curr: last known occupied index to be used as a search hint
268 * void * is being used to keep the functionality generic. This lets us use this
269 * function on any array of pointers.
271 static int ice_get_free_slot(void *array, int size, int curr)
273 int **tmp_array = (int **)array;
276 if (curr < (size - 1) && !tmp_array[curr + 1]) {
281 while ((i < size) && (tmp_array[i]))
292 * ice_vsi_delete_from_hw - delete a VSI from the switch
293 * @vsi: pointer to VSI being removed
295 static void ice_vsi_delete_from_hw(struct ice_vsi *vsi)
297 struct ice_pf *pf = vsi->back;
298 struct ice_vsi_ctx *ctxt;
301 ice_fltr_remove_all(vsi);
302 ctxt = kzalloc(sizeof(*ctxt), GFP_KERNEL);
306 if (vsi->type == ICE_VSI_VF)
307 ctxt->vf_num = vsi->vf->vf_id;
308 ctxt->vsi_num = vsi->vsi_num;
310 memcpy(&ctxt->info, &vsi->info, sizeof(ctxt->info));
312 status = ice_free_vsi(&pf->hw, vsi->idx, ctxt, false, NULL);
314 dev_err(ice_pf_to_dev(pf), "Failed to delete VSI %i in FW - error: %d\n",
315 vsi->vsi_num, status);
321 * ice_vsi_free_arrays - De-allocate queue and vector pointer arrays for the VSI
322 * @vsi: pointer to VSI being cleared
324 static void ice_vsi_free_arrays(struct ice_vsi *vsi)
326 struct ice_pf *pf = vsi->back;
329 dev = ice_pf_to_dev(pf);
331 bitmap_free(vsi->af_xdp_zc_qps);
332 vsi->af_xdp_zc_qps = NULL;
333 /* free the ring and vector containers */
334 devm_kfree(dev, vsi->q_vectors);
335 vsi->q_vectors = NULL;
336 devm_kfree(dev, vsi->tx_rings);
337 vsi->tx_rings = NULL;
338 devm_kfree(dev, vsi->rx_rings);
339 vsi->rx_rings = NULL;
340 devm_kfree(dev, vsi->txq_map);
342 devm_kfree(dev, vsi->rxq_map);
347 * ice_vsi_free_stats - Free the ring statistics structures
350 static void ice_vsi_free_stats(struct ice_vsi *vsi)
352 struct ice_vsi_stats *vsi_stat;
353 struct ice_pf *pf = vsi->back;
356 if (vsi->type == ICE_VSI_CHNL)
361 vsi_stat = pf->vsi_stats[vsi->idx];
365 ice_for_each_alloc_txq(vsi, i) {
366 if (vsi_stat->tx_ring_stats[i]) {
367 kfree_rcu(vsi_stat->tx_ring_stats[i], rcu);
368 WRITE_ONCE(vsi_stat->tx_ring_stats[i], NULL);
372 ice_for_each_alloc_rxq(vsi, i) {
373 if (vsi_stat->rx_ring_stats[i]) {
374 kfree_rcu(vsi_stat->rx_ring_stats[i], rcu);
375 WRITE_ONCE(vsi_stat->rx_ring_stats[i], NULL);
379 kfree(vsi_stat->tx_ring_stats);
380 kfree(vsi_stat->rx_ring_stats);
382 pf->vsi_stats[vsi->idx] = NULL;
386 * ice_vsi_alloc_ring_stats - Allocates Tx and Rx ring stats for the VSI
387 * @vsi: VSI which is having stats allocated
389 static int ice_vsi_alloc_ring_stats(struct ice_vsi *vsi)
391 struct ice_ring_stats **tx_ring_stats;
392 struct ice_ring_stats **rx_ring_stats;
393 struct ice_vsi_stats *vsi_stats;
394 struct ice_pf *pf = vsi->back;
397 vsi_stats = pf->vsi_stats[vsi->idx];
398 tx_ring_stats = vsi_stats->tx_ring_stats;
399 rx_ring_stats = vsi_stats->rx_ring_stats;
401 /* Allocate Tx ring stats */
402 ice_for_each_alloc_txq(vsi, i) {
403 struct ice_ring_stats *ring_stats;
404 struct ice_tx_ring *ring;
406 ring = vsi->tx_rings[i];
407 ring_stats = tx_ring_stats[i];
410 ring_stats = kzalloc(sizeof(*ring_stats), GFP_KERNEL);
414 WRITE_ONCE(tx_ring_stats[i], ring_stats);
417 ring->ring_stats = ring_stats;
420 /* Allocate Rx ring stats */
421 ice_for_each_alloc_rxq(vsi, i) {
422 struct ice_ring_stats *ring_stats;
423 struct ice_rx_ring *ring;
425 ring = vsi->rx_rings[i];
426 ring_stats = rx_ring_stats[i];
429 ring_stats = kzalloc(sizeof(*ring_stats), GFP_KERNEL);
433 WRITE_ONCE(rx_ring_stats[i], ring_stats);
436 ring->ring_stats = ring_stats;
442 ice_vsi_free_stats(vsi);
447 * ice_vsi_free - clean up and deallocate the provided VSI
448 * @vsi: pointer to VSI being cleared
450 * This deallocates the VSI's queue resources, removes it from the PF's
451 * VSI array if necessary, and deallocates the VSI
453 static void ice_vsi_free(struct ice_vsi *vsi)
455 struct ice_pf *pf = NULL;
458 if (!vsi || !vsi->back)
462 dev = ice_pf_to_dev(pf);
464 if (!pf->vsi[vsi->idx] || pf->vsi[vsi->idx] != vsi) {
465 dev_dbg(dev, "vsi does not exist at pf->vsi[%d]\n", vsi->idx);
469 mutex_lock(&pf->sw_mutex);
470 /* updates the PF for this cleared VSI */
472 pf->vsi[vsi->idx] = NULL;
473 pf->next_vsi = vsi->idx;
475 ice_vsi_free_stats(vsi);
476 ice_vsi_free_arrays(vsi);
477 mutex_unlock(&pf->sw_mutex);
478 devm_kfree(dev, vsi);
481 void ice_vsi_delete(struct ice_vsi *vsi)
483 ice_vsi_delete_from_hw(vsi);
488 * ice_msix_clean_ctrl_vsi - MSIX mode interrupt handler for ctrl VSI
489 * @irq: interrupt number
490 * @data: pointer to a q_vector
492 static irqreturn_t ice_msix_clean_ctrl_vsi(int __always_unused irq, void *data)
494 struct ice_q_vector *q_vector = (struct ice_q_vector *)data;
496 if (!q_vector->tx.tx_ring)
499 #define FDIR_RX_DESC_CLEAN_BUDGET 64
500 ice_clean_rx_irq(q_vector->rx.rx_ring, FDIR_RX_DESC_CLEAN_BUDGET);
501 ice_clean_ctrl_tx_irq(q_vector->tx.tx_ring);
507 * ice_msix_clean_rings - MSIX mode Interrupt Handler
508 * @irq: interrupt number
509 * @data: pointer to a q_vector
511 static irqreturn_t ice_msix_clean_rings(int __always_unused irq, void *data)
513 struct ice_q_vector *q_vector = (struct ice_q_vector *)data;
515 if (!q_vector->tx.tx_ring && !q_vector->rx.rx_ring)
518 q_vector->total_events++;
520 napi_schedule(&q_vector->napi);
525 static irqreturn_t ice_eswitch_msix_clean_rings(int __always_unused irq, void *data)
527 struct ice_q_vector *q_vector = (struct ice_q_vector *)data;
528 struct ice_pf *pf = q_vector->vsi->back;
529 struct ice_repr *repr;
532 if (!q_vector->tx.tx_ring && !q_vector->rx.rx_ring)
535 xa_for_each(&pf->eswitch.reprs, id, repr)
536 napi_schedule(&repr->q_vector->napi);
542 * ice_vsi_alloc_stat_arrays - Allocate statistics arrays
545 static int ice_vsi_alloc_stat_arrays(struct ice_vsi *vsi)
547 struct ice_vsi_stats *vsi_stat;
548 struct ice_pf *pf = vsi->back;
550 if (vsi->type == ICE_VSI_CHNL)
555 if (pf->vsi_stats[vsi->idx])
556 /* realloc will happen in rebuild path */
559 vsi_stat = kzalloc(sizeof(*vsi_stat), GFP_KERNEL);
563 vsi_stat->tx_ring_stats =
564 kcalloc(vsi->alloc_txq, sizeof(*vsi_stat->tx_ring_stats),
566 if (!vsi_stat->tx_ring_stats)
569 vsi_stat->rx_ring_stats =
570 kcalloc(vsi->alloc_rxq, sizeof(*vsi_stat->rx_ring_stats),
572 if (!vsi_stat->rx_ring_stats)
575 pf->vsi_stats[vsi->idx] = vsi_stat;
580 kfree(vsi_stat->rx_ring_stats);
582 kfree(vsi_stat->tx_ring_stats);
584 pf->vsi_stats[vsi->idx] = NULL;
589 * ice_vsi_alloc_def - set default values for already allocated VSI
591 * @ch: ptr to channel
594 ice_vsi_alloc_def(struct ice_vsi *vsi, struct ice_channel *ch)
596 if (vsi->type != ICE_VSI_CHNL) {
597 ice_vsi_set_num_qs(vsi);
598 if (ice_vsi_alloc_arrays(vsi))
603 case ICE_VSI_SWITCHDEV_CTRL:
604 /* Setup eswitch MSIX irq handler for VSI */
605 vsi->irq_handler = ice_eswitch_msix_clean_rings;
608 /* Setup default MSIX irq handler for VSI */
609 vsi->irq_handler = ice_msix_clean_rings;
612 /* Setup ctrl VSI MSIX irq handler */
613 vsi->irq_handler = ice_msix_clean_ctrl_vsi;
619 vsi->num_rxq = ch->num_rxq;
620 vsi->num_txq = ch->num_txq;
621 vsi->next_base_q = ch->base_q;
627 ice_vsi_free_arrays(vsi);
635 * ice_vsi_alloc - Allocates the next available struct VSI in the PF
636 * @pf: board private structure
638 * Reserves a VSI index from the PF and allocates an empty VSI structure
639 * without a type. The VSI structure must later be initialized by calling
642 * returns a pointer to a VSI on success, NULL on failure.
644 static struct ice_vsi *ice_vsi_alloc(struct ice_pf *pf)
646 struct device *dev = ice_pf_to_dev(pf);
647 struct ice_vsi *vsi = NULL;
649 /* Need to protect the allocation of the VSIs at the PF level */
650 mutex_lock(&pf->sw_mutex);
652 /* If we have already allocated our maximum number of VSIs,
653 * pf->next_vsi will be ICE_NO_VSI. If not, pf->next_vsi index
654 * is available to be populated
656 if (pf->next_vsi == ICE_NO_VSI) {
657 dev_dbg(dev, "out of VSI slots!\n");
661 vsi = devm_kzalloc(dev, sizeof(*vsi), GFP_KERNEL);
666 set_bit(ICE_VSI_DOWN, vsi->state);
668 /* fill slot and make note of the index */
669 vsi->idx = pf->next_vsi;
670 pf->vsi[pf->next_vsi] = vsi;
672 /* prepare pf->next_vsi for next use */
673 pf->next_vsi = ice_get_free_slot(pf->vsi, pf->num_alloc_vsi,
677 mutex_unlock(&pf->sw_mutex);
682 * ice_alloc_fd_res - Allocate FD resource for a VSI
683 * @vsi: pointer to the ice_vsi
685 * This allocates the FD resources
687 * Returns 0 on success, -EPERM on no-op or -EIO on failure
689 static int ice_alloc_fd_res(struct ice_vsi *vsi)
691 struct ice_pf *pf = vsi->back;
694 /* Flow Director filters are only allocated/assigned to the PF VSI or
695 * CHNL VSI which passes the traffic. The CTRL VSI is only used to
696 * add/delete filters so resources are not allocated to it
698 if (!test_bit(ICE_FLAG_FD_ENA, pf->flags))
701 if (!(vsi->type == ICE_VSI_PF || vsi->type == ICE_VSI_VF ||
702 vsi->type == ICE_VSI_CHNL))
705 /* FD filters from guaranteed pool per VSI */
706 g_val = pf->hw.func_caps.fd_fltr_guar;
710 /* FD filters from best effort pool */
711 b_val = pf->hw.func_caps.fd_fltr_best_effort;
715 /* PF main VSI gets only 64 FD resources from guaranteed pool
716 * when ADQ is configured.
718 #define ICE_PF_VSI_GFLTR 64
720 /* determine FD filter resources per VSI from shared(best effort) and
723 if (vsi->type == ICE_VSI_PF) {
724 vsi->num_gfltr = g_val;
725 /* if MQPRIO is configured, main VSI doesn't get all FD
726 * resources from guaranteed pool. PF VSI gets 64 FD resources
728 if (test_bit(ICE_FLAG_TC_MQPRIO, pf->flags)) {
729 if (g_val < ICE_PF_VSI_GFLTR)
731 /* allow bare minimum entries for PF VSI */
732 vsi->num_gfltr = ICE_PF_VSI_GFLTR;
735 /* each VSI gets same "best_effort" quota */
736 vsi->num_bfltr = b_val;
737 } else if (vsi->type == ICE_VSI_VF) {
740 /* each VSI gets same "best_effort" quota */
741 vsi->num_bfltr = b_val;
743 struct ice_vsi *main_vsi;
746 main_vsi = ice_get_main_vsi(pf);
750 if (!main_vsi->all_numtc)
753 /* figure out ADQ numtc */
754 numtc = main_vsi->all_numtc - ICE_CHNL_START_TC;
756 /* only one TC but still asking resources for channels,
759 if (numtc < ICE_CHNL_START_TC)
762 g_val -= ICE_PF_VSI_GFLTR;
763 /* channel VSIs gets equal share from guaranteed pool */
764 vsi->num_gfltr = g_val / numtc;
766 /* each VSI gets same "best_effort" quota */
767 vsi->num_bfltr = b_val;
774 * ice_vsi_get_qs - Assign queues from PF to VSI
775 * @vsi: the VSI to assign queues to
777 * Returns 0 on success and a negative value on error
779 static int ice_vsi_get_qs(struct ice_vsi *vsi)
781 struct ice_pf *pf = vsi->back;
782 struct ice_qs_cfg tx_qs_cfg = {
783 .qs_mutex = &pf->avail_q_mutex,
784 .pf_map = pf->avail_txqs,
785 .pf_map_size = pf->max_pf_txqs,
786 .q_count = vsi->alloc_txq,
787 .scatter_count = ICE_MAX_SCATTER_TXQS,
788 .vsi_map = vsi->txq_map,
790 .mapping_mode = ICE_VSI_MAP_CONTIG
792 struct ice_qs_cfg rx_qs_cfg = {
793 .qs_mutex = &pf->avail_q_mutex,
794 .pf_map = pf->avail_rxqs,
795 .pf_map_size = pf->max_pf_rxqs,
796 .q_count = vsi->alloc_rxq,
797 .scatter_count = ICE_MAX_SCATTER_RXQS,
798 .vsi_map = vsi->rxq_map,
800 .mapping_mode = ICE_VSI_MAP_CONTIG
804 if (vsi->type == ICE_VSI_CHNL)
807 ret = __ice_vsi_get_qs(&tx_qs_cfg);
810 vsi->tx_mapping_mode = tx_qs_cfg.mapping_mode;
812 ret = __ice_vsi_get_qs(&rx_qs_cfg);
815 vsi->rx_mapping_mode = rx_qs_cfg.mapping_mode;
821 * ice_vsi_put_qs - Release queues from VSI to PF
822 * @vsi: the VSI that is going to release queues
824 static void ice_vsi_put_qs(struct ice_vsi *vsi)
826 struct ice_pf *pf = vsi->back;
829 mutex_lock(&pf->avail_q_mutex);
831 ice_for_each_alloc_txq(vsi, i) {
832 clear_bit(vsi->txq_map[i], pf->avail_txqs);
833 vsi->txq_map[i] = ICE_INVAL_Q_INDEX;
836 ice_for_each_alloc_rxq(vsi, i) {
837 clear_bit(vsi->rxq_map[i], pf->avail_rxqs);
838 vsi->rxq_map[i] = ICE_INVAL_Q_INDEX;
841 mutex_unlock(&pf->avail_q_mutex);
846 * @pf: pointer to the PF struct
848 * returns true if driver is in safe mode, false otherwise
850 bool ice_is_safe_mode(struct ice_pf *pf)
852 return !test_bit(ICE_FLAG_ADV_FEATURES, pf->flags);
857 * @pf: pointer to the PF struct
859 * returns true if RDMA is currently supported, false otherwise
861 bool ice_is_rdma_ena(struct ice_pf *pf)
863 return test_bit(ICE_FLAG_RDMA_ENA, pf->flags);
867 * ice_vsi_clean_rss_flow_fld - Delete RSS configuration
868 * @vsi: the VSI being cleaned up
870 * This function deletes RSS input set for all flows that were configured
873 static void ice_vsi_clean_rss_flow_fld(struct ice_vsi *vsi)
875 struct ice_pf *pf = vsi->back;
878 if (ice_is_safe_mode(pf))
881 status = ice_rem_vsi_rss_cfg(&pf->hw, vsi->idx);
883 dev_dbg(ice_pf_to_dev(pf), "ice_rem_vsi_rss_cfg failed for vsi = %d, error = %d\n",
884 vsi->vsi_num, status);
888 * ice_rss_clean - Delete RSS related VSI structures and configuration
889 * @vsi: the VSI being removed
891 static void ice_rss_clean(struct ice_vsi *vsi)
893 struct ice_pf *pf = vsi->back;
896 dev = ice_pf_to_dev(pf);
898 devm_kfree(dev, vsi->rss_hkey_user);
899 devm_kfree(dev, vsi->rss_lut_user);
901 ice_vsi_clean_rss_flow_fld(vsi);
902 /* remove RSS replay list */
903 if (!ice_is_safe_mode(pf))
904 ice_rem_vsi_rss_list(&pf->hw, vsi->idx);
908 * ice_vsi_set_rss_params - Setup RSS capabilities per VSI type
909 * @vsi: the VSI being configured
911 static void ice_vsi_set_rss_params(struct ice_vsi *vsi)
913 struct ice_hw_common_caps *cap;
914 struct ice_pf *pf = vsi->back;
917 if (!test_bit(ICE_FLAG_RSS_ENA, pf->flags)) {
922 cap = &pf->hw.func_caps.common_cap;
923 max_rss_size = BIT(cap->rss_table_entry_width);
927 /* PF VSI will inherit RSS instance of PF */
928 vsi->rss_table_size = (u16)cap->rss_table_size;
929 if (vsi->type == ICE_VSI_CHNL)
930 vsi->rss_size = min_t(u16, vsi->num_rxq, max_rss_size);
932 vsi->rss_size = min_t(u16, num_online_cpus(),
934 vsi->rss_lut_type = ICE_LUT_PF;
936 case ICE_VSI_SWITCHDEV_CTRL:
937 vsi->rss_table_size = ICE_LUT_VSI_SIZE;
938 vsi->rss_size = min_t(u16, num_online_cpus(), max_rss_size);
939 vsi->rss_lut_type = ICE_LUT_VSI;
942 /* VF VSI will get a small RSS table.
943 * For VSI_LUT, LUT size should be set to 64 bytes.
945 vsi->rss_table_size = ICE_LUT_VSI_SIZE;
946 vsi->rss_size = ICE_MAX_RSS_QS_PER_VF;
947 vsi->rss_lut_type = ICE_LUT_VSI;
952 dev_dbg(ice_pf_to_dev(pf), "Unsupported VSI type %s\n",
953 ice_vsi_type_str(vsi->type));
959 * ice_set_dflt_vsi_ctx - Set default VSI context before adding a VSI
960 * @hw: HW structure used to determine the VLAN mode of the device
961 * @ctxt: the VSI context being set
963 * This initializes a default VSI context for all sections except the Queues.
965 static void ice_set_dflt_vsi_ctx(struct ice_hw *hw, struct ice_vsi_ctx *ctxt)
969 memset(&ctxt->info, 0, sizeof(ctxt->info));
970 /* VSI's should be allocated from shared pool */
971 ctxt->alloc_from_pool = true;
972 /* Src pruning enabled by default */
973 ctxt->info.sw_flags = ICE_AQ_VSI_SW_FLAG_SRC_PRUNE;
974 /* Traffic from VSI can be sent to LAN */
975 ctxt->info.sw_flags2 = ICE_AQ_VSI_SW_FLAG_LAN_ENA;
976 /* allow all untagged/tagged packets by default on Tx */
977 ctxt->info.inner_vlan_flags = FIELD_PREP(ICE_AQ_VSI_INNER_VLAN_TX_MODE_M,
978 ICE_AQ_VSI_INNER_VLAN_TX_MODE_ALL);
979 /* SVM - by default bits 3 and 4 in inner_vlan_flags are 0's which
980 * results in legacy behavior (show VLAN, DEI, and UP) in descriptor.
982 * DVM - leave inner VLAN in packet by default
984 if (ice_is_dvm_ena(hw)) {
985 ctxt->info.inner_vlan_flags |=
986 FIELD_PREP(ICE_AQ_VSI_INNER_VLAN_EMODE_M,
987 ICE_AQ_VSI_INNER_VLAN_EMODE_NOTHING);
988 ctxt->info.outer_vlan_flags =
989 FIELD_PREP(ICE_AQ_VSI_OUTER_VLAN_TX_MODE_M,
990 ICE_AQ_VSI_OUTER_VLAN_TX_MODE_ALL);
991 ctxt->info.outer_vlan_flags |=
992 FIELD_PREP(ICE_AQ_VSI_OUTER_TAG_TYPE_M,
993 ICE_AQ_VSI_OUTER_TAG_VLAN_8100);
994 ctxt->info.outer_vlan_flags |=
995 FIELD_PREP(ICE_AQ_VSI_OUTER_VLAN_EMODE_M,
996 ICE_AQ_VSI_OUTER_VLAN_EMODE_NOTHING);
998 /* Have 1:1 UP mapping for both ingress/egress tables */
999 table |= ICE_UP_TABLE_TRANSLATE(0, 0);
1000 table |= ICE_UP_TABLE_TRANSLATE(1, 1);
1001 table |= ICE_UP_TABLE_TRANSLATE(2, 2);
1002 table |= ICE_UP_TABLE_TRANSLATE(3, 3);
1003 table |= ICE_UP_TABLE_TRANSLATE(4, 4);
1004 table |= ICE_UP_TABLE_TRANSLATE(5, 5);
1005 table |= ICE_UP_TABLE_TRANSLATE(6, 6);
1006 table |= ICE_UP_TABLE_TRANSLATE(7, 7);
1007 ctxt->info.ingress_table = cpu_to_le32(table);
1008 ctxt->info.egress_table = cpu_to_le32(table);
1009 /* Have 1:1 UP mapping for outer to inner UP table */
1010 ctxt->info.outer_up_table = cpu_to_le32(table);
1011 /* No Outer tag support outer_tag_flags remains to zero */
1015 * ice_vsi_setup_q_map - Setup a VSI queue map
1016 * @vsi: the VSI being configured
1017 * @ctxt: VSI context structure
1019 static int ice_vsi_setup_q_map(struct ice_vsi *vsi, struct ice_vsi_ctx *ctxt)
1021 u16 offset = 0, qmap = 0, tx_count = 0, rx_count = 0, pow = 0;
1022 u16 num_txq_per_tc, num_rxq_per_tc;
1023 u16 qcount_tx = vsi->alloc_txq;
1024 u16 qcount_rx = vsi->alloc_rxq;
1028 if (!vsi->tc_cfg.numtc) {
1029 /* at least TC0 should be enabled by default */
1030 vsi->tc_cfg.numtc = 1;
1031 vsi->tc_cfg.ena_tc = 1;
1034 num_rxq_per_tc = min_t(u16, qcount_rx / vsi->tc_cfg.numtc, ICE_MAX_RXQS_PER_TC);
1035 if (!num_rxq_per_tc)
1037 num_txq_per_tc = qcount_tx / vsi->tc_cfg.numtc;
1038 if (!num_txq_per_tc)
1041 /* find the (rounded up) power-of-2 of qcount */
1042 pow = (u16)order_base_2(num_rxq_per_tc);
1044 /* TC mapping is a function of the number of Rx queues assigned to the
1045 * VSI for each traffic class and the offset of these queues.
1046 * The first 10 bits are for queue offset for TC0, next 4 bits for no:of
1047 * queues allocated to TC0. No:of queues is a power-of-2.
1049 * If TC is not enabled, the queue offset is set to 0, and allocate one
1050 * queue, this way, traffic for the given TC will be sent to the default
1053 * Setup number and offset of Rx queues for all TCs for the VSI
1055 ice_for_each_traffic_class(i) {
1056 if (!(vsi->tc_cfg.ena_tc & BIT(i))) {
1057 /* TC is not enabled */
1058 vsi->tc_cfg.tc_info[i].qoffset = 0;
1059 vsi->tc_cfg.tc_info[i].qcount_rx = 1;
1060 vsi->tc_cfg.tc_info[i].qcount_tx = 1;
1061 vsi->tc_cfg.tc_info[i].netdev_tc = 0;
1062 ctxt->info.tc_mapping[i] = 0;
1067 vsi->tc_cfg.tc_info[i].qoffset = offset;
1068 vsi->tc_cfg.tc_info[i].qcount_rx = num_rxq_per_tc;
1069 vsi->tc_cfg.tc_info[i].qcount_tx = num_txq_per_tc;
1070 vsi->tc_cfg.tc_info[i].netdev_tc = netdev_tc++;
1072 qmap = FIELD_PREP(ICE_AQ_VSI_TC_Q_OFFSET_M, offset);
1073 qmap |= FIELD_PREP(ICE_AQ_VSI_TC_Q_NUM_M, pow);
1074 offset += num_rxq_per_tc;
1075 tx_count += num_txq_per_tc;
1076 ctxt->info.tc_mapping[i] = cpu_to_le16(qmap);
1079 /* if offset is non-zero, means it is calculated correctly based on
1080 * enabled TCs for a given VSI otherwise qcount_rx will always
1081 * be correct and non-zero because it is based off - VSI's
1082 * allocated Rx queues which is at least 1 (hence qcount_tx will be
1088 rx_count = num_rxq_per_tc;
1090 if (rx_count > vsi->alloc_rxq) {
1091 dev_err(ice_pf_to_dev(vsi->back), "Trying to use more Rx queues (%u), than were allocated (%u)!\n",
1092 rx_count, vsi->alloc_rxq);
1096 if (tx_count > vsi->alloc_txq) {
1097 dev_err(ice_pf_to_dev(vsi->back), "Trying to use more Tx queues (%u), than were allocated (%u)!\n",
1098 tx_count, vsi->alloc_txq);
1102 vsi->num_txq = tx_count;
1103 vsi->num_rxq = rx_count;
1105 if (vsi->type == ICE_VSI_VF && vsi->num_txq != vsi->num_rxq) {
1106 dev_dbg(ice_pf_to_dev(vsi->back), "VF VSI should have same number of Tx and Rx queues. Hence making them equal\n");
1107 /* since there is a chance that num_rxq could have been changed
1108 * in the above for loop, make num_txq equal to num_rxq.
1110 vsi->num_txq = vsi->num_rxq;
1113 /* Rx queue mapping */
1114 ctxt->info.mapping_flags |= cpu_to_le16(ICE_AQ_VSI_Q_MAP_CONTIG);
1115 /* q_mapping buffer holds the info for the first queue allocated for
1116 * this VSI in the PF space and also the number of queues associated
1119 ctxt->info.q_mapping[0] = cpu_to_le16(vsi->rxq_map[0]);
1120 ctxt->info.q_mapping[1] = cpu_to_le16(vsi->num_rxq);
1126 * ice_set_fd_vsi_ctx - Set FD VSI context before adding a VSI
1127 * @ctxt: the VSI context being set
1128 * @vsi: the VSI being configured
1130 static void ice_set_fd_vsi_ctx(struct ice_vsi_ctx *ctxt, struct ice_vsi *vsi)
1132 u8 dflt_q_group, dflt_q_prio;
1133 u16 dflt_q, report_q, val;
1135 if (vsi->type != ICE_VSI_PF && vsi->type != ICE_VSI_CTRL &&
1136 vsi->type != ICE_VSI_VF && vsi->type != ICE_VSI_CHNL)
1139 val = ICE_AQ_VSI_PROP_FLOW_DIR_VALID;
1140 ctxt->info.valid_sections |= cpu_to_le16(val);
1146 /* enable flow director filtering/programming */
1147 val = ICE_AQ_VSI_FD_ENABLE | ICE_AQ_VSI_FD_PROG_ENABLE;
1148 ctxt->info.fd_options = cpu_to_le16(val);
1149 /* max of allocated flow director filters */
1150 ctxt->info.max_fd_fltr_dedicated =
1151 cpu_to_le16(vsi->num_gfltr);
1152 /* max of shared flow director filters any VSI may program */
1153 ctxt->info.max_fd_fltr_shared =
1154 cpu_to_le16(vsi->num_bfltr);
1155 /* default queue index within the VSI of the default FD */
1156 val = FIELD_PREP(ICE_AQ_VSI_FD_DEF_Q_M, dflt_q);
1157 /* target queue or queue group to the FD filter */
1158 val |= FIELD_PREP(ICE_AQ_VSI_FD_DEF_GRP_M, dflt_q_group);
1159 ctxt->info.fd_def_q = cpu_to_le16(val);
1160 /* queue index on which FD filter completion is reported */
1161 val = FIELD_PREP(ICE_AQ_VSI_FD_REPORT_Q_M, report_q);
1162 /* priority of the default qindex action */
1163 val |= FIELD_PREP(ICE_AQ_VSI_FD_DEF_PRIORITY_M, dflt_q_prio);
1164 ctxt->info.fd_report_opt = cpu_to_le16(val);
1168 * ice_set_rss_vsi_ctx - Set RSS VSI context before adding a VSI
1169 * @ctxt: the VSI context being set
1170 * @vsi: the VSI being configured
1172 static void ice_set_rss_vsi_ctx(struct ice_vsi_ctx *ctxt, struct ice_vsi *vsi)
1174 u8 lut_type, hash_type;
1179 dev = ice_pf_to_dev(pf);
1181 switch (vsi->type) {
1184 /* PF VSI will inherit RSS instance of PF */
1185 lut_type = ICE_AQ_VSI_Q_OPT_RSS_LUT_PF;
1188 /* VF VSI will gets a small RSS table which is a VSI LUT type */
1189 lut_type = ICE_AQ_VSI_Q_OPT_RSS_LUT_VSI;
1192 dev_dbg(dev, "Unsupported VSI type %s\n",
1193 ice_vsi_type_str(vsi->type));
1197 hash_type = ICE_AQ_VSI_Q_OPT_RSS_HASH_TPLZ;
1198 vsi->rss_hfunc = hash_type;
1200 ctxt->info.q_opt_rss =
1201 FIELD_PREP(ICE_AQ_VSI_Q_OPT_RSS_LUT_M, lut_type) |
1202 FIELD_PREP(ICE_AQ_VSI_Q_OPT_RSS_HASH_M, hash_type);
1206 ice_chnl_vsi_setup_q_map(struct ice_vsi *vsi, struct ice_vsi_ctx *ctxt)
1208 struct ice_pf *pf = vsi->back;
1213 qcount = min_t(int, vsi->num_rxq, pf->num_lan_msix);
1215 pow = order_base_2(qcount);
1216 qmap = FIELD_PREP(ICE_AQ_VSI_TC_Q_OFFSET_M, offset);
1217 qmap |= FIELD_PREP(ICE_AQ_VSI_TC_Q_NUM_M, pow);
1219 ctxt->info.tc_mapping[0] = cpu_to_le16(qmap);
1220 ctxt->info.mapping_flags |= cpu_to_le16(ICE_AQ_VSI_Q_MAP_CONTIG);
1221 ctxt->info.q_mapping[0] = cpu_to_le16(vsi->next_base_q);
1222 ctxt->info.q_mapping[1] = cpu_to_le16(qcount);
1226 * ice_vsi_is_vlan_pruning_ena - check if VLAN pruning is enabled or not
1227 * @vsi: VSI to check whether or not VLAN pruning is enabled.
1229 * returns true if Rx VLAN pruning is enabled and false otherwise.
1231 static bool ice_vsi_is_vlan_pruning_ena(struct ice_vsi *vsi)
1233 return vsi->info.sw_flags2 & ICE_AQ_VSI_SW_FLAG_RX_VLAN_PRUNE_ENA;
1237 * ice_vsi_init - Create and initialize a VSI
1238 * @vsi: the VSI being configured
1239 * @vsi_flags: VSI configuration flags
1241 * Set ICE_FLAG_VSI_INIT to initialize a new VSI context, clear it to
1242 * reconfigure an existing context.
1244 * This initializes a VSI context depending on the VSI type to be added and
1245 * passes it down to the add_vsi aq command to create a new VSI.
1247 static int ice_vsi_init(struct ice_vsi *vsi, u32 vsi_flags)
1249 struct ice_pf *pf = vsi->back;
1250 struct ice_hw *hw = &pf->hw;
1251 struct ice_vsi_ctx *ctxt;
1255 dev = ice_pf_to_dev(pf);
1256 ctxt = kzalloc(sizeof(*ctxt), GFP_KERNEL);
1260 switch (vsi->type) {
1264 ctxt->flags = ICE_AQ_VSI_TYPE_PF;
1266 case ICE_VSI_SWITCHDEV_CTRL:
1268 ctxt->flags = ICE_AQ_VSI_TYPE_VMDQ2;
1271 ctxt->flags = ICE_AQ_VSI_TYPE_VF;
1272 /* VF number here is the absolute VF number (0-255) */
1273 ctxt->vf_num = vsi->vf->vf_id + hw->func_caps.vf_base_id;
1280 /* Handle VLAN pruning for channel VSI if main VSI has VLAN
1283 if (vsi->type == ICE_VSI_CHNL) {
1284 struct ice_vsi *main_vsi;
1286 main_vsi = ice_get_main_vsi(pf);
1287 if (main_vsi && ice_vsi_is_vlan_pruning_ena(main_vsi))
1288 ctxt->info.sw_flags2 |=
1289 ICE_AQ_VSI_SW_FLAG_RX_VLAN_PRUNE_ENA;
1291 ctxt->info.sw_flags2 &=
1292 ~ICE_AQ_VSI_SW_FLAG_RX_VLAN_PRUNE_ENA;
1295 ice_set_dflt_vsi_ctx(hw, ctxt);
1296 if (test_bit(ICE_FLAG_FD_ENA, pf->flags))
1297 ice_set_fd_vsi_ctx(ctxt, vsi);
1298 /* if the switch is in VEB mode, allow VSI loopback */
1299 if (vsi->vsw->bridge_mode == BRIDGE_MODE_VEB)
1300 ctxt->info.sw_flags |= ICE_AQ_VSI_SW_FLAG_ALLOW_LB;
1302 /* Set LUT type and HASH type if RSS is enabled */
1303 if (test_bit(ICE_FLAG_RSS_ENA, pf->flags) &&
1304 vsi->type != ICE_VSI_CTRL) {
1305 ice_set_rss_vsi_ctx(ctxt, vsi);
1306 /* if updating VSI context, make sure to set valid_section:
1307 * to indicate which section of VSI context being updated
1309 if (!(vsi_flags & ICE_VSI_FLAG_INIT))
1310 ctxt->info.valid_sections |=
1311 cpu_to_le16(ICE_AQ_VSI_PROP_Q_OPT_VALID);
1314 ctxt->info.sw_id = vsi->port_info->sw_id;
1315 if (vsi->type == ICE_VSI_CHNL) {
1316 ice_chnl_vsi_setup_q_map(vsi, ctxt);
1318 ret = ice_vsi_setup_q_map(vsi, ctxt);
1322 if (!(vsi_flags & ICE_VSI_FLAG_INIT))
1323 /* means VSI being updated */
1324 /* must to indicate which section of VSI context are
1327 ctxt->info.valid_sections |=
1328 cpu_to_le16(ICE_AQ_VSI_PROP_RXQ_MAP_VALID);
1331 /* Allow control frames out of main VSI */
1332 if (vsi->type == ICE_VSI_PF) {
1333 ctxt->info.sec_flags |= ICE_AQ_VSI_SEC_FLAG_ALLOW_DEST_OVRD;
1334 ctxt->info.valid_sections |=
1335 cpu_to_le16(ICE_AQ_VSI_PROP_SECURITY_VALID);
1338 if (vsi_flags & ICE_VSI_FLAG_INIT) {
1339 ret = ice_add_vsi(hw, vsi->idx, ctxt, NULL);
1341 dev_err(dev, "Add VSI failed, err %d\n", ret);
1346 ret = ice_update_vsi(hw, vsi->idx, ctxt, NULL);
1348 dev_err(dev, "Update VSI failed, err %d\n", ret);
1354 /* keep context for update VSI operations */
1355 vsi->info = ctxt->info;
1357 /* record VSI number returned */
1358 vsi->vsi_num = ctxt->vsi_num;
1366 * ice_vsi_clear_rings - Deallocates the Tx and Rx rings for VSI
1367 * @vsi: the VSI having rings deallocated
1369 static void ice_vsi_clear_rings(struct ice_vsi *vsi)
1373 /* Avoid stale references by clearing map from vector to ring */
1374 if (vsi->q_vectors) {
1375 ice_for_each_q_vector(vsi, i) {
1376 struct ice_q_vector *q_vector = vsi->q_vectors[i];
1379 q_vector->tx.tx_ring = NULL;
1380 q_vector->rx.rx_ring = NULL;
1385 if (vsi->tx_rings) {
1386 ice_for_each_alloc_txq(vsi, i) {
1387 if (vsi->tx_rings[i]) {
1388 kfree_rcu(vsi->tx_rings[i], rcu);
1389 WRITE_ONCE(vsi->tx_rings[i], NULL);
1393 if (vsi->rx_rings) {
1394 ice_for_each_alloc_rxq(vsi, i) {
1395 if (vsi->rx_rings[i]) {
1396 kfree_rcu(vsi->rx_rings[i], rcu);
1397 WRITE_ONCE(vsi->rx_rings[i], NULL);
1404 * ice_vsi_alloc_rings - Allocates Tx and Rx rings for the VSI
1405 * @vsi: VSI which is having rings allocated
1407 static int ice_vsi_alloc_rings(struct ice_vsi *vsi)
1409 bool dvm_ena = ice_is_dvm_ena(&vsi->back->hw);
1410 struct ice_pf *pf = vsi->back;
1414 dev = ice_pf_to_dev(pf);
1415 /* Allocate Tx rings */
1416 ice_for_each_alloc_txq(vsi, i) {
1417 struct ice_tx_ring *ring;
1419 /* allocate with kzalloc(), free with kfree_rcu() */
1420 ring = kzalloc(sizeof(*ring), GFP_KERNEL);
1426 ring->reg_idx = vsi->txq_map[i];
1428 ring->tx_tstamps = &pf->ptp.port.tx;
1430 ring->count = vsi->num_tx_desc;
1431 ring->txq_teid = ICE_INVAL_TEID;
1433 ring->flags |= ICE_TX_FLAGS_RING_VLAN_L2TAG2;
1435 ring->flags |= ICE_TX_FLAGS_RING_VLAN_L2TAG1;
1436 WRITE_ONCE(vsi->tx_rings[i], ring);
1439 /* Allocate Rx rings */
1440 ice_for_each_alloc_rxq(vsi, i) {
1441 struct ice_rx_ring *ring;
1443 /* allocate with kzalloc(), free with kfree_rcu() */
1444 ring = kzalloc(sizeof(*ring), GFP_KERNEL);
1449 ring->reg_idx = vsi->rxq_map[i];
1451 ring->netdev = vsi->netdev;
1453 ring->count = vsi->num_rx_desc;
1454 ring->cached_phctime = pf->ptp.cached_phc_time;
1455 WRITE_ONCE(vsi->rx_rings[i], ring);
1461 ice_vsi_clear_rings(vsi);
1466 * ice_vsi_manage_rss_lut - disable/enable RSS
1467 * @vsi: the VSI being changed
1468 * @ena: boolean value indicating if this is an enable or disable request
1470 * In the event of disable request for RSS, this function will zero out RSS
1471 * LUT, while in the event of enable request for RSS, it will reconfigure RSS
1474 void ice_vsi_manage_rss_lut(struct ice_vsi *vsi, bool ena)
1478 lut = kzalloc(vsi->rss_table_size, GFP_KERNEL);
1483 if (vsi->rss_lut_user)
1484 memcpy(lut, vsi->rss_lut_user, vsi->rss_table_size);
1486 ice_fill_rss_lut(lut, vsi->rss_table_size,
1490 ice_set_rss_lut(vsi, lut, vsi->rss_table_size);
1495 * ice_vsi_cfg_crc_strip - Configure CRC stripping for a VSI
1496 * @vsi: VSI to be configured
1497 * @disable: set to true to have FCS / CRC in the frame data
1499 void ice_vsi_cfg_crc_strip(struct ice_vsi *vsi, bool disable)
1503 ice_for_each_rxq(vsi, i)
1505 vsi->rx_rings[i]->flags |= ICE_RX_FLAGS_CRC_STRIP_DIS;
1507 vsi->rx_rings[i]->flags &= ~ICE_RX_FLAGS_CRC_STRIP_DIS;
1511 * ice_vsi_cfg_rss_lut_key - Configure RSS params for a VSI
1512 * @vsi: VSI to be configured
1514 int ice_vsi_cfg_rss_lut_key(struct ice_vsi *vsi)
1516 struct ice_pf *pf = vsi->back;
1521 dev = ice_pf_to_dev(pf);
1522 if (vsi->type == ICE_VSI_PF && vsi->ch_rss_size &&
1523 (test_bit(ICE_FLAG_TC_MQPRIO, pf->flags))) {
1524 vsi->rss_size = min_t(u16, vsi->rss_size, vsi->ch_rss_size);
1526 vsi->rss_size = min_t(u16, vsi->rss_size, vsi->num_rxq);
1528 /* If orig_rss_size is valid and it is less than determined
1529 * main VSI's rss_size, update main VSI's rss_size to be
1530 * orig_rss_size so that when tc-qdisc is deleted, main VSI
1531 * RSS table gets programmed to be correct (whatever it was
1532 * to begin with (prior to setup-tc for ADQ config)
1534 if (vsi->orig_rss_size && vsi->rss_size < vsi->orig_rss_size &&
1535 vsi->orig_rss_size <= vsi->num_rxq) {
1536 vsi->rss_size = vsi->orig_rss_size;
1537 /* now orig_rss_size is used, reset it to zero */
1538 vsi->orig_rss_size = 0;
1542 lut = kzalloc(vsi->rss_table_size, GFP_KERNEL);
1546 if (vsi->rss_lut_user)
1547 memcpy(lut, vsi->rss_lut_user, vsi->rss_table_size);
1549 ice_fill_rss_lut(lut, vsi->rss_table_size, vsi->rss_size);
1551 err = ice_set_rss_lut(vsi, lut, vsi->rss_table_size);
1553 dev_err(dev, "set_rss_lut failed, error %d\n", err);
1554 goto ice_vsi_cfg_rss_exit;
1557 key = kzalloc(ICE_GET_SET_RSS_KEY_EXTEND_KEY_SIZE, GFP_KERNEL);
1560 goto ice_vsi_cfg_rss_exit;
1563 if (vsi->rss_hkey_user)
1564 memcpy(key, vsi->rss_hkey_user, ICE_GET_SET_RSS_KEY_EXTEND_KEY_SIZE);
1566 netdev_rss_key_fill((void *)key, ICE_GET_SET_RSS_KEY_EXTEND_KEY_SIZE);
1568 err = ice_set_rss_key(vsi, key);
1570 dev_err(dev, "set_rss_key failed, error %d\n", err);
1573 ice_vsi_cfg_rss_exit:
1579 * ice_vsi_set_vf_rss_flow_fld - Sets VF VSI RSS input set for different flows
1580 * @vsi: VSI to be configured
1582 * This function will only be called during the VF VSI setup. Upon successful
1583 * completion of package download, this function will configure default RSS
1584 * input sets for VF VSI.
1586 static void ice_vsi_set_vf_rss_flow_fld(struct ice_vsi *vsi)
1588 struct ice_pf *pf = vsi->back;
1592 dev = ice_pf_to_dev(pf);
1593 if (ice_is_safe_mode(pf)) {
1594 dev_dbg(dev, "Advanced RSS disabled. Package download failed, vsi num = %d\n",
1599 status = ice_add_avf_rss_cfg(&pf->hw, vsi, ICE_DEFAULT_RSS_HENA);
1601 dev_dbg(dev, "ice_add_avf_rss_cfg failed for vsi = %d, error = %d\n",
1602 vsi->vsi_num, status);
1605 static const struct ice_rss_hash_cfg default_rss_cfgs[] = {
1606 /* configure RSS for IPv4 with input set IP src/dst */
1607 {ICE_FLOW_SEG_HDR_IPV4, ICE_FLOW_HASH_IPV4, ICE_RSS_ANY_HEADERS, false},
1608 /* configure RSS for IPv6 with input set IPv6 src/dst */
1609 {ICE_FLOW_SEG_HDR_IPV6, ICE_FLOW_HASH_IPV6, ICE_RSS_ANY_HEADERS, false},
1610 /* configure RSS for tcp4 with input set IP src/dst, TCP src/dst */
1611 {ICE_FLOW_SEG_HDR_TCP | ICE_FLOW_SEG_HDR_IPV4,
1612 ICE_HASH_TCP_IPV4, ICE_RSS_ANY_HEADERS, false},
1613 /* configure RSS for udp4 with input set IP src/dst, UDP src/dst */
1614 {ICE_FLOW_SEG_HDR_UDP | ICE_FLOW_SEG_HDR_IPV4,
1615 ICE_HASH_UDP_IPV4, ICE_RSS_ANY_HEADERS, false},
1616 /* configure RSS for sctp4 with input set IP src/dst - only support
1617 * RSS on SCTPv4 on outer headers (non-tunneled)
1619 {ICE_FLOW_SEG_HDR_SCTP | ICE_FLOW_SEG_HDR_IPV4,
1620 ICE_HASH_SCTP_IPV4, ICE_RSS_OUTER_HEADERS, false},
1621 /* configure RSS for tcp6 with input set IPv6 src/dst, TCP src/dst */
1622 {ICE_FLOW_SEG_HDR_TCP | ICE_FLOW_SEG_HDR_IPV6,
1623 ICE_HASH_TCP_IPV6, ICE_RSS_ANY_HEADERS, false},
1624 /* configure RSS for udp6 with input set IPv6 src/dst, UDP src/dst */
1625 {ICE_FLOW_SEG_HDR_UDP | ICE_FLOW_SEG_HDR_IPV6,
1626 ICE_HASH_UDP_IPV6, ICE_RSS_ANY_HEADERS, false},
1627 /* configure RSS for sctp6 with input set IPv6 src/dst - only support
1628 * RSS on SCTPv6 on outer headers (non-tunneled)
1630 {ICE_FLOW_SEG_HDR_SCTP | ICE_FLOW_SEG_HDR_IPV6,
1631 ICE_HASH_SCTP_IPV6, ICE_RSS_OUTER_HEADERS, false},
1632 /* configure RSS for IPSEC ESP SPI with input set MAC_IPV4_SPI */
1633 {ICE_FLOW_SEG_HDR_ESP,
1634 ICE_FLOW_HASH_ESP_SPI, ICE_RSS_OUTER_HEADERS, false},
1638 * ice_vsi_set_rss_flow_fld - Sets RSS input set for different flows
1639 * @vsi: VSI to be configured
1641 * This function will only be called after successful download package call
1642 * during initialization of PF. Since the downloaded package will erase the
1643 * RSS section, this function will configure RSS input sets for different
1644 * flow types. The last profile added has the highest priority, therefore 2
1645 * tuple profiles (i.e. IPv4 src/dst) are added before 4 tuple profiles
1646 * (i.e. IPv4 src/dst TCP src/dst port).
1648 static void ice_vsi_set_rss_flow_fld(struct ice_vsi *vsi)
1650 u16 vsi_num = vsi->vsi_num;
1651 struct ice_pf *pf = vsi->back;
1652 struct ice_hw *hw = &pf->hw;
1657 dev = ice_pf_to_dev(pf);
1658 if (ice_is_safe_mode(pf)) {
1659 dev_dbg(dev, "Advanced RSS disabled. Package download failed, vsi num = %d\n",
1663 for (i = 0; i < ARRAY_SIZE(default_rss_cfgs); i++) {
1664 const struct ice_rss_hash_cfg *cfg = &default_rss_cfgs[i];
1666 status = ice_add_rss_cfg(hw, vsi, cfg);
1668 dev_dbg(dev, "ice_add_rss_cfg failed, addl_hdrs = %x, hash_flds = %llx, hdr_type = %d, symm = %d\n",
1669 cfg->addl_hdrs, cfg->hash_flds,
1670 cfg->hdr_type, cfg->symm);
1675 * ice_vsi_cfg_frame_size - setup max frame size and Rx buffer length
1678 static void ice_vsi_cfg_frame_size(struct ice_vsi *vsi)
1680 if (!vsi->netdev || test_bit(ICE_FLAG_LEGACY_RX, vsi->back->flags)) {
1681 vsi->max_frame = ICE_MAX_FRAME_LEGACY_RX;
1682 vsi->rx_buf_len = ICE_RXBUF_1664;
1683 #if (PAGE_SIZE < 8192)
1684 } else if (!ICE_2K_TOO_SMALL_WITH_PADDING &&
1685 (vsi->netdev->mtu <= ETH_DATA_LEN)) {
1686 vsi->max_frame = ICE_RXBUF_1536 - NET_IP_ALIGN;
1687 vsi->rx_buf_len = ICE_RXBUF_1536 - NET_IP_ALIGN;
1690 vsi->max_frame = ICE_AQ_SET_MAC_FRAME_SIZE_MAX;
1691 vsi->rx_buf_len = ICE_RXBUF_3072;
1696 * ice_pf_state_is_nominal - checks the PF for nominal state
1697 * @pf: pointer to PF to check
1699 * Check the PF's state for a collection of bits that would indicate
1700 * the PF is in a state that would inhibit normal operation for
1701 * driver functionality.
1703 * Returns true if PF is in a nominal state, false otherwise
1705 bool ice_pf_state_is_nominal(struct ice_pf *pf)
1707 DECLARE_BITMAP(check_bits, ICE_STATE_NBITS) = { 0 };
1712 bitmap_set(check_bits, 0, ICE_STATE_NOMINAL_CHECK_BITS);
1713 if (bitmap_intersects(pf->state, check_bits, ICE_STATE_NBITS))
1720 * ice_update_eth_stats - Update VSI-specific ethernet statistics counters
1721 * @vsi: the VSI to be updated
1723 void ice_update_eth_stats(struct ice_vsi *vsi)
1725 struct ice_eth_stats *prev_es, *cur_es;
1726 struct ice_hw *hw = &vsi->back->hw;
1727 struct ice_pf *pf = vsi->back;
1728 u16 vsi_num = vsi->vsi_num; /* HW absolute index of a VSI */
1730 prev_es = &vsi->eth_stats_prev;
1731 cur_es = &vsi->eth_stats;
1733 if (ice_is_reset_in_progress(pf->state))
1734 vsi->stat_offsets_loaded = false;
1736 ice_stat_update40(hw, GLV_GORCL(vsi_num), vsi->stat_offsets_loaded,
1737 &prev_es->rx_bytes, &cur_es->rx_bytes);
1739 ice_stat_update40(hw, GLV_UPRCL(vsi_num), vsi->stat_offsets_loaded,
1740 &prev_es->rx_unicast, &cur_es->rx_unicast);
1742 ice_stat_update40(hw, GLV_MPRCL(vsi_num), vsi->stat_offsets_loaded,
1743 &prev_es->rx_multicast, &cur_es->rx_multicast);
1745 ice_stat_update40(hw, GLV_BPRCL(vsi_num), vsi->stat_offsets_loaded,
1746 &prev_es->rx_broadcast, &cur_es->rx_broadcast);
1748 ice_stat_update32(hw, GLV_RDPC(vsi_num), vsi->stat_offsets_loaded,
1749 &prev_es->rx_discards, &cur_es->rx_discards);
1751 ice_stat_update40(hw, GLV_GOTCL(vsi_num), vsi->stat_offsets_loaded,
1752 &prev_es->tx_bytes, &cur_es->tx_bytes);
1754 ice_stat_update40(hw, GLV_UPTCL(vsi_num), vsi->stat_offsets_loaded,
1755 &prev_es->tx_unicast, &cur_es->tx_unicast);
1757 ice_stat_update40(hw, GLV_MPTCL(vsi_num), vsi->stat_offsets_loaded,
1758 &prev_es->tx_multicast, &cur_es->tx_multicast);
1760 ice_stat_update40(hw, GLV_BPTCL(vsi_num), vsi->stat_offsets_loaded,
1761 &prev_es->tx_broadcast, &cur_es->tx_broadcast);
1763 ice_stat_update32(hw, GLV_TEPC(vsi_num), vsi->stat_offsets_loaded,
1764 &prev_es->tx_errors, &cur_es->tx_errors);
1766 vsi->stat_offsets_loaded = true;
1770 * ice_write_qrxflxp_cntxt - write/configure QRXFLXP_CNTXT register
1772 * @pf_q: index of the Rx queue in the PF's queue space
1773 * @rxdid: flexible descriptor RXDID
1774 * @prio: priority for the RXDID for this queue
1775 * @ena_ts: true to enable timestamp and false to disable timestamp
1778 ice_write_qrxflxp_cntxt(struct ice_hw *hw, u16 pf_q, u32 rxdid, u32 prio,
1781 int regval = rd32(hw, QRXFLXP_CNTXT(pf_q));
1783 /* clear any previous values */
1784 regval &= ~(QRXFLXP_CNTXT_RXDID_IDX_M |
1785 QRXFLXP_CNTXT_RXDID_PRIO_M |
1786 QRXFLXP_CNTXT_TS_M);
1788 regval |= FIELD_PREP(QRXFLXP_CNTXT_RXDID_IDX_M, rxdid);
1789 regval |= FIELD_PREP(QRXFLXP_CNTXT_RXDID_PRIO_M, prio);
1792 /* Enable TimeSync on this queue */
1793 regval |= QRXFLXP_CNTXT_TS_M;
1795 wr32(hw, QRXFLXP_CNTXT(pf_q), regval);
1798 int ice_vsi_cfg_single_rxq(struct ice_vsi *vsi, u16 q_idx)
1800 if (q_idx >= vsi->num_rxq)
1803 return ice_vsi_cfg_rxq(vsi->rx_rings[q_idx]);
1806 int ice_vsi_cfg_single_txq(struct ice_vsi *vsi, struct ice_tx_ring **tx_rings, u16 q_idx)
1808 DEFINE_FLEX(struct ice_aqc_add_tx_qgrp, qg_buf, txqs, 1);
1810 if (q_idx >= vsi->alloc_txq || !tx_rings || !tx_rings[q_idx])
1813 qg_buf->num_txqs = 1;
1815 return ice_vsi_cfg_txq(vsi, tx_rings[q_idx], qg_buf);
1819 * ice_vsi_cfg_rxqs - Configure the VSI for Rx
1820 * @vsi: the VSI being configured
1822 * Return 0 on success and a negative value on error
1823 * Configure the Rx VSI for operation.
1825 int ice_vsi_cfg_rxqs(struct ice_vsi *vsi)
1829 if (vsi->type == ICE_VSI_VF)
1832 ice_vsi_cfg_frame_size(vsi);
1834 /* set up individual rings */
1835 ice_for_each_rxq(vsi, i) {
1836 int err = ice_vsi_cfg_rxq(vsi->rx_rings[i]);
1846 * ice_vsi_cfg_txqs - Configure the VSI for Tx
1847 * @vsi: the VSI being configured
1848 * @rings: Tx ring array to be configured
1849 * @count: number of Tx ring array elements
1851 * Return 0 on success and a negative value on error
1852 * Configure the Tx VSI for operation.
1855 ice_vsi_cfg_txqs(struct ice_vsi *vsi, struct ice_tx_ring **rings, u16 count)
1857 DEFINE_FLEX(struct ice_aqc_add_tx_qgrp, qg_buf, txqs, 1);
1861 qg_buf->num_txqs = 1;
1863 for (q_idx = 0; q_idx < count; q_idx++) {
1864 err = ice_vsi_cfg_txq(vsi, rings[q_idx], qg_buf);
1873 * ice_vsi_cfg_lan_txqs - Configure the VSI for Tx
1874 * @vsi: the VSI being configured
1876 * Return 0 on success and a negative value on error
1877 * Configure the Tx VSI for operation.
1879 int ice_vsi_cfg_lan_txqs(struct ice_vsi *vsi)
1881 return ice_vsi_cfg_txqs(vsi, vsi->tx_rings, vsi->num_txq);
1885 * ice_vsi_cfg_xdp_txqs - Configure Tx queues dedicated for XDP in given VSI
1886 * @vsi: the VSI being configured
1888 * Return 0 on success and a negative value on error
1889 * Configure the Tx queues dedicated for XDP in given VSI for operation.
1891 int ice_vsi_cfg_xdp_txqs(struct ice_vsi *vsi)
1896 ret = ice_vsi_cfg_txqs(vsi, vsi->xdp_rings, vsi->num_xdp_txq);
1900 ice_for_each_rxq(vsi, i)
1901 ice_tx_xsk_pool(vsi, i);
1907 * ice_intrl_usec_to_reg - convert interrupt rate limit to register value
1908 * @intrl: interrupt rate limit in usecs
1909 * @gran: interrupt rate limit granularity in usecs
1911 * This function converts a decimal interrupt rate limit in usecs to the format
1912 * expected by firmware.
1914 static u32 ice_intrl_usec_to_reg(u8 intrl, u8 gran)
1916 u32 val = intrl / gran;
1919 return val | GLINT_RATE_INTRL_ENA_M;
1924 * ice_write_intrl - write throttle rate limit to interrupt specific register
1925 * @q_vector: pointer to interrupt specific structure
1926 * @intrl: throttle rate limit in microseconds to write
1928 void ice_write_intrl(struct ice_q_vector *q_vector, u8 intrl)
1930 struct ice_hw *hw = &q_vector->vsi->back->hw;
1932 wr32(hw, GLINT_RATE(q_vector->reg_idx),
1933 ice_intrl_usec_to_reg(intrl, ICE_INTRL_GRAN_ABOVE_25));
1936 static struct ice_q_vector *ice_pull_qvec_from_rc(struct ice_ring_container *rc)
1939 case ICE_RX_CONTAINER:
1941 return rc->rx_ring->q_vector;
1943 case ICE_TX_CONTAINER:
1945 return rc->tx_ring->q_vector;
1955 * __ice_write_itr - write throttle rate to register
1956 * @q_vector: pointer to interrupt data structure
1957 * @rc: pointer to ring container
1958 * @itr: throttle rate in microseconds to write
1960 static void __ice_write_itr(struct ice_q_vector *q_vector,
1961 struct ice_ring_container *rc, u16 itr)
1963 struct ice_hw *hw = &q_vector->vsi->back->hw;
1965 wr32(hw, GLINT_ITR(rc->itr_idx, q_vector->reg_idx),
1966 ITR_REG_ALIGN(itr) >> ICE_ITR_GRAN_S);
1970 * ice_write_itr - write throttle rate to queue specific register
1971 * @rc: pointer to ring container
1972 * @itr: throttle rate in microseconds to write
1974 void ice_write_itr(struct ice_ring_container *rc, u16 itr)
1976 struct ice_q_vector *q_vector;
1978 q_vector = ice_pull_qvec_from_rc(rc);
1982 __ice_write_itr(q_vector, rc, itr);
1986 * ice_set_q_vector_intrl - set up interrupt rate limiting
1987 * @q_vector: the vector to be configured
1989 * Interrupt rate limiting is local to the vector, not per-queue so we must
1990 * detect if either ring container has dynamic moderation enabled to decide
1991 * what to set the interrupt rate limit to via INTRL settings. In the case that
1992 * dynamic moderation is disabled on both, write the value with the cached
1993 * setting to make sure INTRL register matches the user visible value.
1995 void ice_set_q_vector_intrl(struct ice_q_vector *q_vector)
1997 if (ITR_IS_DYNAMIC(&q_vector->tx) || ITR_IS_DYNAMIC(&q_vector->rx)) {
1998 /* in the case of dynamic enabled, cap each vector to no more
1999 * than (4 us) 250,000 ints/sec, which allows low latency
2000 * but still less than 500,000 interrupts per second, which
2001 * reduces CPU a bit in the case of the lowest latency
2002 * setting. The 4 here is a value in microseconds.
2004 ice_write_intrl(q_vector, 4);
2006 ice_write_intrl(q_vector, q_vector->intrl);
2011 * ice_vsi_cfg_msix - MSIX mode Interrupt Config in the HW
2012 * @vsi: the VSI being configured
2014 * This configures MSIX mode interrupts for the PF VSI, and should not be used
2017 void ice_vsi_cfg_msix(struct ice_vsi *vsi)
2019 struct ice_pf *pf = vsi->back;
2020 struct ice_hw *hw = &pf->hw;
2021 u16 txq = 0, rxq = 0;
2024 ice_for_each_q_vector(vsi, i) {
2025 struct ice_q_vector *q_vector = vsi->q_vectors[i];
2026 u16 reg_idx = q_vector->reg_idx;
2028 ice_cfg_itr(hw, q_vector);
2030 /* Both Transmit Queue Interrupt Cause Control register
2031 * and Receive Queue Interrupt Cause control register
2032 * expects MSIX_INDX field to be the vector index
2033 * within the function space and not the absolute
2034 * vector index across PF or across device.
2035 * For SR-IOV VF VSIs queue vector index always starts
2036 * with 1 since first vector index(0) is used for OICR
2037 * in VF space. Since VMDq and other PF VSIs are within
2038 * the PF function space, use the vector index that is
2039 * tracked for this PF.
2041 for (q = 0; q < q_vector->num_ring_tx; q++) {
2042 ice_cfg_txq_interrupt(vsi, txq, reg_idx,
2043 q_vector->tx.itr_idx);
2047 for (q = 0; q < q_vector->num_ring_rx; q++) {
2048 ice_cfg_rxq_interrupt(vsi, rxq, reg_idx,
2049 q_vector->rx.itr_idx);
2056 * ice_vsi_start_all_rx_rings - start/enable all of a VSI's Rx rings
2057 * @vsi: the VSI whose rings are to be enabled
2059 * Returns 0 on success and a negative value on error
2061 int ice_vsi_start_all_rx_rings(struct ice_vsi *vsi)
2063 return ice_vsi_ctrl_all_rx_rings(vsi, true);
2067 * ice_vsi_stop_all_rx_rings - stop/disable all of a VSI's Rx rings
2068 * @vsi: the VSI whose rings are to be disabled
2070 * Returns 0 on success and a negative value on error
2072 int ice_vsi_stop_all_rx_rings(struct ice_vsi *vsi)
2074 return ice_vsi_ctrl_all_rx_rings(vsi, false);
2078 * ice_vsi_stop_tx_rings - Disable Tx rings
2079 * @vsi: the VSI being configured
2080 * @rst_src: reset source
2081 * @rel_vmvf_num: Relative ID of VF/VM
2082 * @rings: Tx ring array to be stopped
2083 * @count: number of Tx ring array elements
2086 ice_vsi_stop_tx_rings(struct ice_vsi *vsi, enum ice_disq_rst_src rst_src,
2087 u16 rel_vmvf_num, struct ice_tx_ring **rings, u16 count)
2091 if (vsi->num_txq > ICE_LAN_TXQ_MAX_QDIS)
2094 for (q_idx = 0; q_idx < count; q_idx++) {
2095 struct ice_txq_meta txq_meta = { };
2098 if (!rings || !rings[q_idx])
2101 ice_fill_txq_meta(vsi, rings[q_idx], &txq_meta);
2102 status = ice_vsi_stop_tx_ring(vsi, rst_src, rel_vmvf_num,
2103 rings[q_idx], &txq_meta);
2113 * ice_vsi_stop_lan_tx_rings - Disable LAN Tx rings
2114 * @vsi: the VSI being configured
2115 * @rst_src: reset source
2116 * @rel_vmvf_num: Relative ID of VF/VM
2119 ice_vsi_stop_lan_tx_rings(struct ice_vsi *vsi, enum ice_disq_rst_src rst_src,
2122 return ice_vsi_stop_tx_rings(vsi, rst_src, rel_vmvf_num, vsi->tx_rings, vsi->num_txq);
2126 * ice_vsi_stop_xdp_tx_rings - Disable XDP Tx rings
2127 * @vsi: the VSI being configured
2129 int ice_vsi_stop_xdp_tx_rings(struct ice_vsi *vsi)
2131 return ice_vsi_stop_tx_rings(vsi, ICE_NO_RESET, 0, vsi->xdp_rings, vsi->num_xdp_txq);
2135 * ice_vsi_is_rx_queue_active
2136 * @vsi: the VSI being configured
2138 * Return true if at least one queue is active.
2140 bool ice_vsi_is_rx_queue_active(struct ice_vsi *vsi)
2142 struct ice_pf *pf = vsi->back;
2143 struct ice_hw *hw = &pf->hw;
2146 ice_for_each_rxq(vsi, i) {
2150 pf_q = vsi->rxq_map[i];
2151 rx_reg = rd32(hw, QRX_CTRL(pf_q));
2152 if (rx_reg & QRX_CTRL_QENA_STAT_M)
2159 static void ice_vsi_set_tc_cfg(struct ice_vsi *vsi)
2161 if (!test_bit(ICE_FLAG_DCB_ENA, vsi->back->flags)) {
2162 vsi->tc_cfg.ena_tc = ICE_DFLT_TRAFFIC_CLASS;
2163 vsi->tc_cfg.numtc = 1;
2167 /* set VSI TC information based on DCB config */
2168 ice_vsi_set_dcb_tc_cfg(vsi);
2172 * ice_cfg_sw_lldp - Config switch rules for LLDP packet handling
2173 * @vsi: the VSI being configured
2174 * @tx: bool to determine Tx or Rx rule
2175 * @create: bool to determine create or remove Rule
2177 void ice_cfg_sw_lldp(struct ice_vsi *vsi, bool tx, bool create)
2179 int (*eth_fltr)(struct ice_vsi *v, u16 type, u16 flag,
2180 enum ice_sw_fwd_act_type act);
2181 struct ice_pf *pf = vsi->back;
2185 dev = ice_pf_to_dev(pf);
2186 eth_fltr = create ? ice_fltr_add_eth : ice_fltr_remove_eth;
2189 status = eth_fltr(vsi, ETH_P_LLDP, ICE_FLTR_TX,
2192 if (ice_fw_supports_lldp_fltr_ctrl(&pf->hw)) {
2193 status = ice_lldp_fltr_add_remove(&pf->hw, vsi->vsi_num,
2196 status = eth_fltr(vsi, ETH_P_LLDP, ICE_FLTR_RX,
2202 dev_dbg(dev, "Fail %s %s LLDP rule on VSI %i error: %d\n",
2203 create ? "adding" : "removing", tx ? "TX" : "RX",
2204 vsi->vsi_num, status);
2208 * ice_set_agg_vsi - sets up scheduler aggregator node and move VSI into it
2209 * @vsi: pointer to the VSI
2211 * This function will allocate new scheduler aggregator now if needed and will
2212 * move specified VSI into it.
2214 static void ice_set_agg_vsi(struct ice_vsi *vsi)
2216 struct device *dev = ice_pf_to_dev(vsi->back);
2217 struct ice_agg_node *agg_node_iter = NULL;
2218 u32 agg_id = ICE_INVALID_AGG_NODE_ID;
2219 struct ice_agg_node *agg_node = NULL;
2220 int node_offset, max_agg_nodes = 0;
2221 struct ice_port_info *port_info;
2222 struct ice_pf *pf = vsi->back;
2223 u32 agg_node_id_start = 0;
2226 /* create (as needed) scheduler aggregator node and move VSI into
2227 * corresponding aggregator node
2228 * - PF aggregator node to contains VSIs of type _PF and _CTRL
2229 * - VF aggregator nodes will contain VF VSI
2231 port_info = pf->hw.port_info;
2235 switch (vsi->type) {
2240 case ICE_VSI_SWITCHDEV_CTRL:
2241 max_agg_nodes = ICE_MAX_PF_AGG_NODES;
2242 agg_node_id_start = ICE_PF_AGG_NODE_ID_START;
2243 agg_node_iter = &pf->pf_agg_node[0];
2246 /* user can create 'n' VFs on a given PF, but since max children
2247 * per aggregator node can be only 64. Following code handles
2248 * aggregator(s) for VF VSIs, either selects a agg_node which
2249 * was already created provided num_vsis < 64, otherwise
2250 * select next available node, which will be created
2252 max_agg_nodes = ICE_MAX_VF_AGG_NODES;
2253 agg_node_id_start = ICE_VF_AGG_NODE_ID_START;
2254 agg_node_iter = &pf->vf_agg_node[0];
2257 /* other VSI type, handle later if needed */
2258 dev_dbg(dev, "unexpected VSI type %s\n",
2259 ice_vsi_type_str(vsi->type));
2263 /* find the appropriate aggregator node */
2264 for (node_offset = 0; node_offset < max_agg_nodes; node_offset++) {
2265 /* see if we can find space in previously created
2266 * node if num_vsis < 64, otherwise skip
2268 if (agg_node_iter->num_vsis &&
2269 agg_node_iter->num_vsis == ICE_MAX_VSIS_IN_AGG_NODE) {
2274 if (agg_node_iter->valid &&
2275 agg_node_iter->agg_id != ICE_INVALID_AGG_NODE_ID) {
2276 agg_id = agg_node_iter->agg_id;
2277 agg_node = agg_node_iter;
2281 /* find unclaimed agg_id */
2282 if (agg_node_iter->agg_id == ICE_INVALID_AGG_NODE_ID) {
2283 agg_id = node_offset + agg_node_id_start;
2284 agg_node = agg_node_iter;
2287 /* move to next agg_node */
2294 /* if selected aggregator node was not created, create it */
2295 if (!agg_node->valid) {
2296 status = ice_cfg_agg(port_info, agg_id, ICE_AGG_TYPE_AGG,
2297 (u8)vsi->tc_cfg.ena_tc);
2299 dev_err(dev, "unable to create aggregator node with agg_id %u\n",
2303 /* aggregator node is created, store the needed info */
2304 agg_node->valid = true;
2305 agg_node->agg_id = agg_id;
2308 /* move VSI to corresponding aggregator node */
2309 status = ice_move_vsi_to_agg(port_info, agg_id, vsi->idx,
2310 (u8)vsi->tc_cfg.ena_tc);
2312 dev_err(dev, "unable to move VSI idx %u into aggregator %u node",
2317 /* keep active children count for aggregator node */
2318 agg_node->num_vsis++;
2320 /* cache the 'agg_id' in VSI, so that after reset - VSI will be moved
2321 * to aggregator node
2323 vsi->agg_node = agg_node;
2324 dev_dbg(dev, "successfully moved VSI idx %u tc_bitmap 0x%x) into aggregator node %d which has num_vsis %u\n",
2325 vsi->idx, vsi->tc_cfg.ena_tc, vsi->agg_node->agg_id,
2326 vsi->agg_node->num_vsis);
2329 static int ice_vsi_cfg_tc_lan(struct ice_pf *pf, struct ice_vsi *vsi)
2331 u16 max_txqs[ICE_MAX_TRAFFIC_CLASS] = { 0 };
2332 struct device *dev = ice_pf_to_dev(pf);
2335 /* configure VSI nodes based on number of queues and TC's */
2336 ice_for_each_traffic_class(i) {
2337 if (!(vsi->tc_cfg.ena_tc & BIT(i)))
2340 if (vsi->type == ICE_VSI_CHNL) {
2341 if (!vsi->alloc_txq && vsi->num_txq)
2342 max_txqs[i] = vsi->num_txq;
2344 max_txqs[i] = pf->num_lan_tx;
2346 max_txqs[i] = vsi->alloc_txq;
2349 if (vsi->type == ICE_VSI_PF)
2350 max_txqs[i] += vsi->num_xdp_txq;
2353 dev_dbg(dev, "vsi->tc_cfg.ena_tc = %d\n", vsi->tc_cfg.ena_tc);
2354 ret = ice_cfg_vsi_lan(vsi->port_info, vsi->idx, vsi->tc_cfg.ena_tc,
2357 dev_err(dev, "VSI %d failed lan queue config, error %d\n",
2366 * ice_vsi_cfg_def - configure default VSI based on the type
2367 * @vsi: pointer to VSI
2368 * @params: the parameters to configure this VSI with
2371 ice_vsi_cfg_def(struct ice_vsi *vsi, struct ice_vsi_cfg_params *params)
2373 struct device *dev = ice_pf_to_dev(vsi->back);
2374 struct ice_pf *pf = vsi->back;
2377 vsi->vsw = pf->first_sw;
2379 ret = ice_vsi_alloc_def(vsi, params->ch);
2383 /* allocate memory for Tx/Rx ring stat pointers */
2384 ret = ice_vsi_alloc_stat_arrays(vsi);
2386 goto unroll_vsi_alloc;
2388 ice_alloc_fd_res(vsi);
2390 ret = ice_vsi_get_qs(vsi);
2392 dev_err(dev, "Failed to allocate queues. vsi->idx = %d\n",
2394 goto unroll_vsi_alloc_stat;
2397 /* set RSS capabilities */
2398 ice_vsi_set_rss_params(vsi);
2400 /* set TC configuration */
2401 ice_vsi_set_tc_cfg(vsi);
2403 /* create the VSI */
2404 ret = ice_vsi_init(vsi, params->flags);
2408 ice_vsi_init_vlan_ops(vsi);
2410 switch (vsi->type) {
2412 case ICE_VSI_SWITCHDEV_CTRL:
2414 ret = ice_vsi_alloc_q_vectors(vsi);
2416 goto unroll_vsi_init;
2418 ret = ice_vsi_alloc_rings(vsi);
2420 goto unroll_vector_base;
2422 ret = ice_vsi_alloc_ring_stats(vsi);
2424 goto unroll_vector_base;
2426 ice_vsi_map_rings_to_vectors(vsi);
2428 /* Associate q_vector rings to napi */
2429 ice_vsi_set_napi_queues(vsi, true);
2431 vsi->stat_offsets_loaded = false;
2433 if (ice_is_xdp_ena_vsi(vsi)) {
2434 ret = ice_vsi_determine_xdp_res(vsi);
2436 goto unroll_vector_base;
2437 ret = ice_prepare_xdp_rings(vsi, vsi->xdp_prog);
2439 goto unroll_vector_base;
2442 /* ICE_VSI_CTRL does not need RSS so skip RSS processing */
2443 if (vsi->type != ICE_VSI_CTRL)
2444 /* Do not exit if configuring RSS had an issue, at
2445 * least receive traffic on first queue. Hence no
2446 * need to capture return value
2448 if (test_bit(ICE_FLAG_RSS_ENA, pf->flags)) {
2449 ice_vsi_cfg_rss_lut_key(vsi);
2450 ice_vsi_set_rss_flow_fld(vsi);
2455 if (test_bit(ICE_FLAG_RSS_ENA, pf->flags)) {
2456 ice_vsi_cfg_rss_lut_key(vsi);
2457 ice_vsi_set_rss_flow_fld(vsi);
2461 /* VF driver will take care of creating netdev for this type and
2462 * map queues to vectors through Virtchnl, PF driver only
2463 * creates a VSI and corresponding structures for bookkeeping
2466 ret = ice_vsi_alloc_q_vectors(vsi);
2468 goto unroll_vsi_init;
2470 ret = ice_vsi_alloc_rings(vsi);
2472 goto unroll_alloc_q_vector;
2474 ret = ice_vsi_alloc_ring_stats(vsi);
2476 goto unroll_vector_base;
2478 vsi->stat_offsets_loaded = false;
2480 /* Do not exit if configuring RSS had an issue, at least
2481 * receive traffic on first queue. Hence no need to capture
2484 if (test_bit(ICE_FLAG_RSS_ENA, pf->flags)) {
2485 ice_vsi_cfg_rss_lut_key(vsi);
2486 ice_vsi_set_vf_rss_flow_fld(vsi);
2490 ret = ice_vsi_alloc_rings(vsi);
2492 goto unroll_vsi_init;
2494 ret = ice_vsi_alloc_ring_stats(vsi);
2496 goto unroll_vector_base;
2500 /* clean up the resources and exit */
2502 goto unroll_vsi_init;
2508 /* reclaim SW interrupts back to the common pool */
2509 unroll_alloc_q_vector:
2510 ice_vsi_free_q_vectors(vsi);
2512 ice_vsi_delete_from_hw(vsi);
2514 ice_vsi_put_qs(vsi);
2515 unroll_vsi_alloc_stat:
2516 ice_vsi_free_stats(vsi);
2518 ice_vsi_free_arrays(vsi);
2523 * ice_vsi_cfg - configure a previously allocated VSI
2524 * @vsi: pointer to VSI
2525 * @params: parameters used to configure this VSI
2527 int ice_vsi_cfg(struct ice_vsi *vsi, struct ice_vsi_cfg_params *params)
2529 struct ice_pf *pf = vsi->back;
2532 if (WARN_ON(params->type == ICE_VSI_VF && !params->vf))
2535 vsi->type = params->type;
2536 vsi->port_info = params->pi;
2538 /* For VSIs which don't have a connected VF, this will be NULL */
2539 vsi->vf = params->vf;
2541 ret = ice_vsi_cfg_def(vsi, params);
2545 ret = ice_vsi_cfg_tc_lan(vsi->back, vsi);
2549 if (vsi->type == ICE_VSI_CTRL) {
2551 WARN_ON(vsi->vf->ctrl_vsi_idx != ICE_NO_VSI);
2552 vsi->vf->ctrl_vsi_idx = vsi->idx;
2554 WARN_ON(pf->ctrl_vsi_idx != ICE_NO_VSI);
2555 pf->ctrl_vsi_idx = vsi->idx;
2563 * ice_vsi_decfg - remove all VSI configuration
2564 * @vsi: pointer to VSI
2566 void ice_vsi_decfg(struct ice_vsi *vsi)
2568 struct ice_pf *pf = vsi->back;
2571 /* The Rx rule will only exist to remove if the LLDP FW
2572 * engine is currently stopped
2574 if (!ice_is_safe_mode(pf) && vsi->type == ICE_VSI_PF &&
2575 !test_bit(ICE_FLAG_FW_LLDP_AGENT, pf->flags))
2576 ice_cfg_sw_lldp(vsi, false, false);
2578 ice_rm_vsi_lan_cfg(vsi->port_info, vsi->idx);
2579 err = ice_rm_vsi_rdma_cfg(vsi->port_info, vsi->idx);
2581 dev_err(ice_pf_to_dev(pf), "Failed to remove RDMA scheduler config for VSI %u, err %d\n",
2584 if (ice_is_xdp_ena_vsi(vsi))
2585 /* return value check can be skipped here, it always returns
2586 * 0 if reset is in progress
2588 ice_destroy_xdp_rings(vsi);
2590 ice_vsi_clear_rings(vsi);
2591 ice_vsi_free_q_vectors(vsi);
2592 ice_vsi_put_qs(vsi);
2593 ice_vsi_free_arrays(vsi);
2595 /* SR-IOV determines needed MSIX resources all at once instead of per
2596 * VSI since when VFs are spawned we know how many VFs there are and how
2597 * many interrupts each VF needs. SR-IOV MSIX resources are also
2598 * cleared in the same manner.
2601 if (vsi->type == ICE_VSI_VF &&
2602 vsi->agg_node && vsi->agg_node->valid)
2603 vsi->agg_node->num_vsis--;
2607 * ice_vsi_setup - Set up a VSI by a given type
2608 * @pf: board private structure
2609 * @params: parameters to use when creating the VSI
2611 * This allocates the sw VSI structure and its queue resources.
2613 * Returns pointer to the successfully allocated and configured VSI sw struct on
2614 * success, NULL on failure.
2617 ice_vsi_setup(struct ice_pf *pf, struct ice_vsi_cfg_params *params)
2619 struct device *dev = ice_pf_to_dev(pf);
2620 struct ice_vsi *vsi;
2623 /* ice_vsi_setup can only initialize a new VSI, and we must have
2624 * a port_info structure for it.
2626 if (WARN_ON(!(params->flags & ICE_VSI_FLAG_INIT)) ||
2627 WARN_ON(!params->pi))
2630 vsi = ice_vsi_alloc(pf);
2632 dev_err(dev, "could not allocate VSI\n");
2636 ret = ice_vsi_cfg(vsi, params);
2640 /* Add switch rule to drop all Tx Flow Control Frames, of look up
2641 * type ETHERTYPE from VSIs, and restrict malicious VF from sending
2642 * out PAUSE or PFC frames. If enabled, FW can still send FC frames.
2643 * The rule is added once for PF VSI in order to create appropriate
2644 * recipe, since VSI/VSI list is ignored with drop action...
2645 * Also add rules to handle LLDP Tx packets. Tx LLDP packets need to
2646 * be dropped so that VFs cannot send LLDP packets to reconfig DCB
2647 * settings in the HW.
2649 if (!ice_is_safe_mode(pf) && vsi->type == ICE_VSI_PF) {
2650 ice_fltr_add_eth(vsi, ETH_P_PAUSE, ICE_FLTR_TX,
2652 ice_cfg_sw_lldp(vsi, true, true);
2656 ice_set_agg_vsi(vsi);
2667 * ice_vsi_release_msix - Clear the queue to Interrupt mapping in HW
2668 * @vsi: the VSI being cleaned up
2670 static void ice_vsi_release_msix(struct ice_vsi *vsi)
2672 struct ice_pf *pf = vsi->back;
2673 struct ice_hw *hw = &pf->hw;
2678 ice_for_each_q_vector(vsi, i) {
2679 struct ice_q_vector *q_vector = vsi->q_vectors[i];
2681 ice_write_intrl(q_vector, 0);
2682 for (q = 0; q < q_vector->num_ring_tx; q++) {
2683 ice_write_itr(&q_vector->tx, 0);
2684 wr32(hw, QINT_TQCTL(vsi->txq_map[txq]), 0);
2685 if (ice_is_xdp_ena_vsi(vsi)) {
2686 u32 xdp_txq = txq + vsi->num_xdp_txq;
2688 wr32(hw, QINT_TQCTL(vsi->txq_map[xdp_txq]), 0);
2693 for (q = 0; q < q_vector->num_ring_rx; q++) {
2694 ice_write_itr(&q_vector->rx, 0);
2695 wr32(hw, QINT_RQCTL(vsi->rxq_map[rxq]), 0);
2704 * ice_vsi_free_irq - Free the IRQ association with the OS
2705 * @vsi: the VSI being configured
2707 void ice_vsi_free_irq(struct ice_vsi *vsi)
2709 struct ice_pf *pf = vsi->back;
2712 if (!vsi->q_vectors || !vsi->irqs_ready)
2715 ice_vsi_release_msix(vsi);
2716 if (vsi->type == ICE_VSI_VF)
2719 vsi->irqs_ready = false;
2720 ice_free_cpu_rx_rmap(vsi);
2722 ice_for_each_q_vector(vsi, i) {
2725 irq_num = vsi->q_vectors[i]->irq.virq;
2727 /* free only the irqs that were actually requested */
2728 if (!vsi->q_vectors[i] ||
2729 !(vsi->q_vectors[i]->num_ring_tx ||
2730 vsi->q_vectors[i]->num_ring_rx))
2733 /* clear the affinity notifier in the IRQ descriptor */
2734 if (!IS_ENABLED(CONFIG_RFS_ACCEL))
2735 irq_set_affinity_notifier(irq_num, NULL);
2737 /* clear the affinity_mask in the IRQ descriptor */
2738 irq_set_affinity_hint(irq_num, NULL);
2739 synchronize_irq(irq_num);
2740 devm_free_irq(ice_pf_to_dev(pf), irq_num, vsi->q_vectors[i]);
2745 * ice_vsi_free_tx_rings - Free Tx resources for VSI queues
2746 * @vsi: the VSI having resources freed
2748 void ice_vsi_free_tx_rings(struct ice_vsi *vsi)
2755 ice_for_each_txq(vsi, i)
2756 if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc)
2757 ice_free_tx_ring(vsi->tx_rings[i]);
2761 * ice_vsi_free_rx_rings - Free Rx resources for VSI queues
2762 * @vsi: the VSI having resources freed
2764 void ice_vsi_free_rx_rings(struct ice_vsi *vsi)
2771 ice_for_each_rxq(vsi, i)
2772 if (vsi->rx_rings[i] && vsi->rx_rings[i]->desc)
2773 ice_free_rx_ring(vsi->rx_rings[i]);
2777 * ice_vsi_close - Shut down a VSI
2778 * @vsi: the VSI being shut down
2780 void ice_vsi_close(struct ice_vsi *vsi)
2782 if (!test_and_set_bit(ICE_VSI_DOWN, vsi->state))
2785 ice_vsi_free_irq(vsi);
2786 ice_vsi_free_tx_rings(vsi);
2787 ice_vsi_free_rx_rings(vsi);
2791 * ice_ena_vsi - resume a VSI
2792 * @vsi: the VSI being resume
2793 * @locked: is the rtnl_lock already held
2795 int ice_ena_vsi(struct ice_vsi *vsi, bool locked)
2799 if (!test_bit(ICE_VSI_NEEDS_RESTART, vsi->state))
2802 clear_bit(ICE_VSI_NEEDS_RESTART, vsi->state);
2804 if (vsi->netdev && vsi->type == ICE_VSI_PF) {
2805 if (netif_running(vsi->netdev)) {
2809 err = ice_open_internal(vsi->netdev);
2814 } else if (vsi->type == ICE_VSI_CTRL) {
2815 err = ice_vsi_open_ctrl(vsi);
2822 * ice_dis_vsi - pause a VSI
2823 * @vsi: the VSI being paused
2824 * @locked: is the rtnl_lock already held
2826 void ice_dis_vsi(struct ice_vsi *vsi, bool locked)
2828 if (test_bit(ICE_VSI_DOWN, vsi->state))
2831 set_bit(ICE_VSI_NEEDS_RESTART, vsi->state);
2833 if (vsi->type == ICE_VSI_PF && vsi->netdev) {
2834 if (netif_running(vsi->netdev)) {
2845 } else if (vsi->type == ICE_VSI_CTRL ||
2846 vsi->type == ICE_VSI_SWITCHDEV_CTRL) {
2852 * ice_vsi_dis_irq - Mask off queue interrupt generation on the VSI
2853 * @vsi: the VSI being un-configured
2855 void ice_vsi_dis_irq(struct ice_vsi *vsi)
2857 struct ice_pf *pf = vsi->back;
2858 struct ice_hw *hw = &pf->hw;
2862 /* disable interrupt causation from each queue */
2863 if (vsi->tx_rings) {
2864 ice_for_each_txq(vsi, i) {
2865 if (vsi->tx_rings[i]) {
2868 reg = vsi->tx_rings[i]->reg_idx;
2869 val = rd32(hw, QINT_TQCTL(reg));
2870 val &= ~QINT_TQCTL_CAUSE_ENA_M;
2871 wr32(hw, QINT_TQCTL(reg), val);
2876 if (vsi->rx_rings) {
2877 ice_for_each_rxq(vsi, i) {
2878 if (vsi->rx_rings[i]) {
2881 reg = vsi->rx_rings[i]->reg_idx;
2882 val = rd32(hw, QINT_RQCTL(reg));
2883 val &= ~QINT_RQCTL_CAUSE_ENA_M;
2884 wr32(hw, QINT_RQCTL(reg), val);
2889 /* disable each interrupt */
2890 ice_for_each_q_vector(vsi, i) {
2891 if (!vsi->q_vectors[i])
2893 wr32(hw, GLINT_DYN_CTL(vsi->q_vectors[i]->reg_idx), 0);
2898 /* don't call synchronize_irq() for VF's from the host */
2899 if (vsi->type == ICE_VSI_VF)
2902 ice_for_each_q_vector(vsi, i)
2903 synchronize_irq(vsi->q_vectors[i]->irq.virq);
2907 * ice_queue_set_napi - Set the napi instance for the queue
2908 * @dev: device to which NAPI and queue belong
2909 * @queue_index: Index of queue
2910 * @type: queue type as RX or TX
2911 * @napi: NAPI context
2912 * @locked: is the rtnl_lock already held
2914 * Set the napi instance for the queue
2917 ice_queue_set_napi(struct net_device *dev, unsigned int queue_index,
2918 enum netdev_queue_type type, struct napi_struct *napi,
2923 netif_queue_set_napi(dev, queue_index, type, napi);
2929 * ice_q_vector_set_napi_queues - Map queue[s] associated with the napi
2930 * @q_vector: q_vector pointer
2931 * @locked: is the rtnl_lock already held
2933 * Associate the q_vector napi with all the queue[s] on the vector
2935 void ice_q_vector_set_napi_queues(struct ice_q_vector *q_vector, bool locked)
2937 struct ice_rx_ring *rx_ring;
2938 struct ice_tx_ring *tx_ring;
2940 ice_for_each_rx_ring(rx_ring, q_vector->rx)
2941 ice_queue_set_napi(q_vector->vsi->netdev, rx_ring->q_index,
2942 NETDEV_QUEUE_TYPE_RX, &q_vector->napi,
2945 ice_for_each_tx_ring(tx_ring, q_vector->tx)
2946 ice_queue_set_napi(q_vector->vsi->netdev, tx_ring->q_index,
2947 NETDEV_QUEUE_TYPE_TX, &q_vector->napi,
2949 /* Also set the interrupt number for the NAPI */
2950 netif_napi_set_irq(&q_vector->napi, q_vector->irq.virq);
2954 * ice_vsi_set_napi_queues
2956 * @locked: is the rtnl_lock already held
2958 * Associate queue[s] with napi for all vectors
2960 void ice_vsi_set_napi_queues(struct ice_vsi *vsi, bool locked)
2967 ice_for_each_q_vector(vsi, i)
2968 ice_q_vector_set_napi_queues(vsi->q_vectors[i], locked);
2972 * ice_vsi_release - Delete a VSI and free its resources
2973 * @vsi: the VSI being removed
2975 * Returns 0 on success or < 0 on error
2977 int ice_vsi_release(struct ice_vsi *vsi)
2985 if (test_bit(ICE_FLAG_RSS_ENA, pf->flags))
2991 /* retain SW VSI data structure since it is needed to unregister and
2992 * free VSI netdev when PF is not in reset recovery pending state,\
2993 * for ex: during rmmod.
2995 if (!ice_is_reset_in_progress(pf->state))
2996 ice_vsi_delete(vsi);
3002 * ice_vsi_rebuild_get_coalesce - get coalesce from all q_vectors
3003 * @vsi: VSI connected with q_vectors
3004 * @coalesce: array of struct with stored coalesce
3006 * Returns array size.
3009 ice_vsi_rebuild_get_coalesce(struct ice_vsi *vsi,
3010 struct ice_coalesce_stored *coalesce)
3014 ice_for_each_q_vector(vsi, i) {
3015 struct ice_q_vector *q_vector = vsi->q_vectors[i];
3017 coalesce[i].itr_tx = q_vector->tx.itr_settings;
3018 coalesce[i].itr_rx = q_vector->rx.itr_settings;
3019 coalesce[i].intrl = q_vector->intrl;
3021 if (i < vsi->num_txq)
3022 coalesce[i].tx_valid = true;
3023 if (i < vsi->num_rxq)
3024 coalesce[i].rx_valid = true;
3027 return vsi->num_q_vectors;
3031 * ice_vsi_rebuild_set_coalesce - set coalesce from earlier saved arrays
3032 * @vsi: VSI connected with q_vectors
3033 * @coalesce: pointer to array of struct with stored coalesce
3034 * @size: size of coalesce array
3036 * Before this function, ice_vsi_rebuild_get_coalesce should be called to save
3037 * ITR params in arrays. If size is 0 or coalesce wasn't stored set coalesce
3041 ice_vsi_rebuild_set_coalesce(struct ice_vsi *vsi,
3042 struct ice_coalesce_stored *coalesce, int size)
3044 struct ice_ring_container *rc;
3047 if ((size && !coalesce) || !vsi)
3050 /* There are a couple of cases that have to be handled here:
3051 * 1. The case where the number of queue vectors stays the same, but
3052 * the number of Tx or Rx rings changes (the first for loop)
3053 * 2. The case where the number of queue vectors increased (the
3056 for (i = 0; i < size && i < vsi->num_q_vectors; i++) {
3057 /* There are 2 cases to handle here and they are the same for
3059 * if the entry was valid previously (coalesce[i].[tr]x_valid
3060 * and the loop variable is less than the number of rings
3061 * allocated, then write the previous values
3063 * if the entry was not valid previously, but the number of
3064 * rings is less than are allocated (this means the number of
3065 * rings increased from previously), then write out the
3066 * values in the first element
3068 * Also, always write the ITR, even if in ITR_IS_DYNAMIC
3069 * as there is no harm because the dynamic algorithm
3070 * will just overwrite.
3072 if (i < vsi->alloc_rxq && coalesce[i].rx_valid) {
3073 rc = &vsi->q_vectors[i]->rx;
3074 rc->itr_settings = coalesce[i].itr_rx;
3075 ice_write_itr(rc, rc->itr_setting);
3076 } else if (i < vsi->alloc_rxq) {
3077 rc = &vsi->q_vectors[i]->rx;
3078 rc->itr_settings = coalesce[0].itr_rx;
3079 ice_write_itr(rc, rc->itr_setting);
3082 if (i < vsi->alloc_txq && coalesce[i].tx_valid) {
3083 rc = &vsi->q_vectors[i]->tx;
3084 rc->itr_settings = coalesce[i].itr_tx;
3085 ice_write_itr(rc, rc->itr_setting);
3086 } else if (i < vsi->alloc_txq) {
3087 rc = &vsi->q_vectors[i]->tx;
3088 rc->itr_settings = coalesce[0].itr_tx;
3089 ice_write_itr(rc, rc->itr_setting);
3092 vsi->q_vectors[i]->intrl = coalesce[i].intrl;
3093 ice_set_q_vector_intrl(vsi->q_vectors[i]);
3096 /* the number of queue vectors increased so write whatever is in
3099 for (; i < vsi->num_q_vectors; i++) {
3101 rc = &vsi->q_vectors[i]->tx;
3102 rc->itr_settings = coalesce[0].itr_tx;
3103 ice_write_itr(rc, rc->itr_setting);
3106 rc = &vsi->q_vectors[i]->rx;
3107 rc->itr_settings = coalesce[0].itr_rx;
3108 ice_write_itr(rc, rc->itr_setting);
3110 vsi->q_vectors[i]->intrl = coalesce[0].intrl;
3111 ice_set_q_vector_intrl(vsi->q_vectors[i]);
3116 * ice_vsi_realloc_stat_arrays - Frees unused stat structures or alloc new ones
3120 ice_vsi_realloc_stat_arrays(struct ice_vsi *vsi)
3122 u16 req_txq = vsi->req_txq ? vsi->req_txq : vsi->alloc_txq;
3123 u16 req_rxq = vsi->req_rxq ? vsi->req_rxq : vsi->alloc_rxq;
3124 struct ice_ring_stats **tx_ring_stats;
3125 struct ice_ring_stats **rx_ring_stats;
3126 struct ice_vsi_stats *vsi_stat;
3127 struct ice_pf *pf = vsi->back;
3128 u16 prev_txq = vsi->alloc_txq;
3129 u16 prev_rxq = vsi->alloc_rxq;
3132 vsi_stat = pf->vsi_stats[vsi->idx];
3134 if (req_txq < prev_txq) {
3135 for (i = req_txq; i < prev_txq; i++) {
3136 if (vsi_stat->tx_ring_stats[i]) {
3137 kfree_rcu(vsi_stat->tx_ring_stats[i], rcu);
3138 WRITE_ONCE(vsi_stat->tx_ring_stats[i], NULL);
3143 tx_ring_stats = vsi_stat->rx_ring_stats;
3144 vsi_stat->tx_ring_stats =
3145 krealloc_array(vsi_stat->tx_ring_stats, req_txq,
3146 sizeof(*vsi_stat->tx_ring_stats),
3147 GFP_KERNEL | __GFP_ZERO);
3148 if (!vsi_stat->tx_ring_stats) {
3149 vsi_stat->tx_ring_stats = tx_ring_stats;
3153 if (req_rxq < prev_rxq) {
3154 for (i = req_rxq; i < prev_rxq; i++) {
3155 if (vsi_stat->rx_ring_stats[i]) {
3156 kfree_rcu(vsi_stat->rx_ring_stats[i], rcu);
3157 WRITE_ONCE(vsi_stat->rx_ring_stats[i], NULL);
3162 rx_ring_stats = vsi_stat->rx_ring_stats;
3163 vsi_stat->rx_ring_stats =
3164 krealloc_array(vsi_stat->rx_ring_stats, req_rxq,
3165 sizeof(*vsi_stat->rx_ring_stats),
3166 GFP_KERNEL | __GFP_ZERO);
3167 if (!vsi_stat->rx_ring_stats) {
3168 vsi_stat->rx_ring_stats = rx_ring_stats;
3176 * ice_vsi_rebuild - Rebuild VSI after reset
3177 * @vsi: VSI to be rebuild
3178 * @vsi_flags: flags used for VSI rebuild flow
3180 * Set vsi_flags to ICE_VSI_FLAG_INIT to initialize a new VSI, or
3181 * ICE_VSI_FLAG_NO_INIT to rebuild an existing VSI in hardware.
3183 * Returns 0 on success and negative value on failure
3185 int ice_vsi_rebuild(struct ice_vsi *vsi, u32 vsi_flags)
3187 struct ice_vsi_cfg_params params = {};
3188 struct ice_coalesce_stored *coalesce;
3189 int prev_num_q_vectors = 0;
3196 params = ice_vsi_to_params(vsi);
3197 params.flags = vsi_flags;
3200 if (WARN_ON(vsi->type == ICE_VSI_VF && !vsi->vf))
3203 coalesce = kcalloc(vsi->num_q_vectors,
3204 sizeof(struct ice_coalesce_stored), GFP_KERNEL);
3208 prev_num_q_vectors = ice_vsi_rebuild_get_coalesce(vsi, coalesce);
3210 ret = ice_vsi_realloc_stat_arrays(vsi);
3215 ret = ice_vsi_cfg_def(vsi, ¶ms);
3219 ret = ice_vsi_cfg_tc_lan(pf, vsi);
3221 if (vsi_flags & ICE_VSI_FLAG_INIT) {
3223 goto err_vsi_cfg_tc_lan;
3227 return ice_schedule_reset(pf, ICE_RESET_PFR);
3230 ice_vsi_rebuild_set_coalesce(vsi, coalesce, prev_num_q_vectors);
3243 * ice_is_reset_in_progress - check for a reset in progress
3244 * @state: PF state field
3246 bool ice_is_reset_in_progress(unsigned long *state)
3248 return test_bit(ICE_RESET_OICR_RECV, state) ||
3249 test_bit(ICE_PFR_REQ, state) ||
3250 test_bit(ICE_CORER_REQ, state) ||
3251 test_bit(ICE_GLOBR_REQ, state);
3255 * ice_wait_for_reset - Wait for driver to finish reset and rebuild
3256 * @pf: pointer to the PF structure
3257 * @timeout: length of time to wait, in jiffies
3259 * Wait (sleep) for a short time until the driver finishes cleaning up from
3260 * a device reset. The caller must be able to sleep. Use this to delay
3261 * operations that could fail while the driver is cleaning up after a device
3264 * Returns 0 on success, -EBUSY if the reset is not finished within the
3265 * timeout, and -ERESTARTSYS if the thread was interrupted.
3267 int ice_wait_for_reset(struct ice_pf *pf, unsigned long timeout)
3271 ret = wait_event_interruptible_timeout(pf->reset_wait_queue,
3272 !ice_is_reset_in_progress(pf->state),
3283 * ice_vsi_update_q_map - update our copy of the VSI info with new queue map
3284 * @vsi: VSI being configured
3285 * @ctx: the context buffer returned from AQ VSI update command
3287 static void ice_vsi_update_q_map(struct ice_vsi *vsi, struct ice_vsi_ctx *ctx)
3289 vsi->info.mapping_flags = ctx->info.mapping_flags;
3290 memcpy(&vsi->info.q_mapping, &ctx->info.q_mapping,
3291 sizeof(vsi->info.q_mapping));
3292 memcpy(&vsi->info.tc_mapping, ctx->info.tc_mapping,
3293 sizeof(vsi->info.tc_mapping));
3297 * ice_vsi_cfg_netdev_tc - Setup the netdev TC configuration
3298 * @vsi: the VSI being configured
3299 * @ena_tc: TC map to be enabled
3301 void ice_vsi_cfg_netdev_tc(struct ice_vsi *vsi, u8 ena_tc)
3303 struct net_device *netdev = vsi->netdev;
3304 struct ice_pf *pf = vsi->back;
3305 int numtc = vsi->tc_cfg.numtc;
3306 struct ice_dcbx_cfg *dcbcfg;
3313 /* CHNL VSI doesn't have it's own netdev, hence, no netdev_tc */
3314 if (vsi->type == ICE_VSI_CHNL)
3318 netdev_reset_tc(netdev);
3322 if (vsi->type == ICE_VSI_PF && ice_is_adq_active(pf))
3323 numtc = vsi->all_numtc;
3325 if (netdev_set_num_tc(netdev, numtc))
3328 dcbcfg = &pf->hw.port_info->qos_cfg.local_dcbx_cfg;
3330 ice_for_each_traffic_class(i)
3331 if (vsi->tc_cfg.ena_tc & BIT(i))
3332 netdev_set_tc_queue(netdev,
3333 vsi->tc_cfg.tc_info[i].netdev_tc,
3334 vsi->tc_cfg.tc_info[i].qcount_tx,
3335 vsi->tc_cfg.tc_info[i].qoffset);
3336 /* setup TC queue map for CHNL TCs */
3337 ice_for_each_chnl_tc(i) {
3338 if (!(vsi->all_enatc & BIT(i)))
3340 if (!vsi->mqprio_qopt.qopt.count[i])
3342 netdev_set_tc_queue(netdev, i,
3343 vsi->mqprio_qopt.qopt.count[i],
3344 vsi->mqprio_qopt.qopt.offset[i]);
3347 if (test_bit(ICE_FLAG_TC_MQPRIO, pf->flags))
3350 for (i = 0; i < ICE_MAX_USER_PRIORITY; i++) {
3351 u8 ets_tc = dcbcfg->etscfg.prio_table[i];
3353 /* Get the mapped netdev TC# for the UP */
3354 netdev_tc = vsi->tc_cfg.tc_info[ets_tc].netdev_tc;
3355 netdev_set_prio_tc_map(netdev, i, netdev_tc);
3360 * ice_vsi_setup_q_map_mqprio - Prepares mqprio based tc_config
3361 * @vsi: the VSI being configured,
3362 * @ctxt: VSI context structure
3363 * @ena_tc: number of traffic classes to enable
3365 * Prepares VSI tc_config to have queue configurations based on MQPRIO options.
3368 ice_vsi_setup_q_map_mqprio(struct ice_vsi *vsi, struct ice_vsi_ctx *ctxt,
3371 u16 pow, offset = 0, qcount_tx = 0, qcount_rx = 0, qmap;
3372 u16 tc0_offset = vsi->mqprio_qopt.qopt.offset[0];
3373 int tc0_qcount = vsi->mqprio_qopt.qopt.count[0];
3374 u16 new_txq, new_rxq;
3378 vsi->tc_cfg.ena_tc = ena_tc ? ena_tc : 1;
3380 pow = order_base_2(tc0_qcount);
3381 qmap = FIELD_PREP(ICE_AQ_VSI_TC_Q_OFFSET_M, tc0_offset);
3382 qmap |= FIELD_PREP(ICE_AQ_VSI_TC_Q_NUM_M, pow);
3384 ice_for_each_traffic_class(i) {
3385 if (!(vsi->tc_cfg.ena_tc & BIT(i))) {
3386 /* TC is not enabled */
3387 vsi->tc_cfg.tc_info[i].qoffset = 0;
3388 vsi->tc_cfg.tc_info[i].qcount_rx = 1;
3389 vsi->tc_cfg.tc_info[i].qcount_tx = 1;
3390 vsi->tc_cfg.tc_info[i].netdev_tc = 0;
3391 ctxt->info.tc_mapping[i] = 0;
3395 offset = vsi->mqprio_qopt.qopt.offset[i];
3396 qcount_rx = vsi->mqprio_qopt.qopt.count[i];
3397 qcount_tx = vsi->mqprio_qopt.qopt.count[i];
3398 vsi->tc_cfg.tc_info[i].qoffset = offset;
3399 vsi->tc_cfg.tc_info[i].qcount_rx = qcount_rx;
3400 vsi->tc_cfg.tc_info[i].qcount_tx = qcount_tx;
3401 vsi->tc_cfg.tc_info[i].netdev_tc = netdev_tc++;
3404 if (vsi->all_numtc && vsi->all_numtc != vsi->tc_cfg.numtc) {
3405 ice_for_each_chnl_tc(i) {
3406 if (!(vsi->all_enatc & BIT(i)))
3408 offset = vsi->mqprio_qopt.qopt.offset[i];
3409 qcount_rx = vsi->mqprio_qopt.qopt.count[i];
3410 qcount_tx = vsi->mqprio_qopt.qopt.count[i];
3414 new_txq = offset + qcount_tx;
3415 if (new_txq > vsi->alloc_txq) {
3416 dev_err(ice_pf_to_dev(vsi->back), "Trying to use more Tx queues (%u), than were allocated (%u)!\n",
3417 new_txq, vsi->alloc_txq);
3421 new_rxq = offset + qcount_rx;
3422 if (new_rxq > vsi->alloc_rxq) {
3423 dev_err(ice_pf_to_dev(vsi->back), "Trying to use more Rx queues (%u), than were allocated (%u)!\n",
3424 new_rxq, vsi->alloc_rxq);
3428 /* Set actual Tx/Rx queue pairs */
3429 vsi->num_txq = new_txq;
3430 vsi->num_rxq = new_rxq;
3432 /* Setup queue TC[0].qmap for given VSI context */
3433 ctxt->info.tc_mapping[0] = cpu_to_le16(qmap);
3434 ctxt->info.q_mapping[0] = cpu_to_le16(vsi->rxq_map[0]);
3435 ctxt->info.q_mapping[1] = cpu_to_le16(tc0_qcount);
3437 /* Find queue count available for channel VSIs and starting offset
3440 if (tc0_qcount && tc0_qcount < vsi->num_rxq) {
3441 vsi->cnt_q_avail = vsi->num_rxq - tc0_qcount;
3442 vsi->next_base_q = tc0_qcount;
3444 dev_dbg(ice_pf_to_dev(vsi->back), "vsi->num_txq = %d\n", vsi->num_txq);
3445 dev_dbg(ice_pf_to_dev(vsi->back), "vsi->num_rxq = %d\n", vsi->num_rxq);
3446 dev_dbg(ice_pf_to_dev(vsi->back), "all_numtc %u, all_enatc: 0x%04x, tc_cfg.numtc %u\n",
3447 vsi->all_numtc, vsi->all_enatc, vsi->tc_cfg.numtc);
3453 * ice_vsi_cfg_tc - Configure VSI Tx Sched for given TC map
3454 * @vsi: VSI to be configured
3455 * @ena_tc: TC bitmap
3457 * VSI queues expected to be quiesced before calling this function
3459 int ice_vsi_cfg_tc(struct ice_vsi *vsi, u8 ena_tc)
3461 u16 max_txqs[ICE_MAX_TRAFFIC_CLASS] = { 0 };
3462 struct ice_pf *pf = vsi->back;
3463 struct ice_tc_cfg old_tc_cfg;
3464 struct ice_vsi_ctx *ctx;
3469 dev = ice_pf_to_dev(pf);
3470 if (vsi->tc_cfg.ena_tc == ena_tc &&
3471 vsi->mqprio_qopt.mode != TC_MQPRIO_MODE_CHANNEL)
3474 ice_for_each_traffic_class(i) {
3475 /* build bitmap of enabled TCs */
3476 if (ena_tc & BIT(i))
3478 /* populate max_txqs per TC */
3479 max_txqs[i] = vsi->alloc_txq;
3480 /* Update max_txqs if it is CHNL VSI, because alloc_t[r]xq are
3481 * zero for CHNL VSI, hence use num_txq instead as max_txqs
3483 if (vsi->type == ICE_VSI_CHNL &&
3484 test_bit(ICE_FLAG_TC_MQPRIO, pf->flags))
3485 max_txqs[i] = vsi->num_txq;
3488 memcpy(&old_tc_cfg, &vsi->tc_cfg, sizeof(old_tc_cfg));
3489 vsi->tc_cfg.ena_tc = ena_tc;
3490 vsi->tc_cfg.numtc = num_tc;
3492 ctx = kzalloc(sizeof(*ctx), GFP_KERNEL);
3497 ctx->info = vsi->info;
3499 if (vsi->type == ICE_VSI_PF &&
3500 test_bit(ICE_FLAG_TC_MQPRIO, pf->flags))
3501 ret = ice_vsi_setup_q_map_mqprio(vsi, ctx, ena_tc);
3503 ret = ice_vsi_setup_q_map(vsi, ctx);
3506 memcpy(&vsi->tc_cfg, &old_tc_cfg, sizeof(vsi->tc_cfg));
3510 /* must to indicate which section of VSI context are being modified */
3511 ctx->info.valid_sections = cpu_to_le16(ICE_AQ_VSI_PROP_RXQ_MAP_VALID);
3512 ret = ice_update_vsi(&pf->hw, vsi->idx, ctx, NULL);
3514 dev_info(dev, "Failed VSI Update\n");
3518 if (vsi->type == ICE_VSI_PF &&
3519 test_bit(ICE_FLAG_TC_MQPRIO, pf->flags))
3520 ret = ice_cfg_vsi_lan(vsi->port_info, vsi->idx, 1, max_txqs);
3522 ret = ice_cfg_vsi_lan(vsi->port_info, vsi->idx,
3523 vsi->tc_cfg.ena_tc, max_txqs);
3526 dev_err(dev, "VSI %d failed TC config, error %d\n",
3530 ice_vsi_update_q_map(vsi, ctx);
3531 vsi->info.valid_sections = 0;
3533 ice_vsi_cfg_netdev_tc(vsi, ena_tc);
3540 * ice_update_ring_stats - Update ring statistics
3541 * @stats: stats to be updated
3542 * @pkts: number of processed packets
3543 * @bytes: number of processed bytes
3545 * This function assumes that caller has acquired a u64_stats_sync lock.
3547 static void ice_update_ring_stats(struct ice_q_stats *stats, u64 pkts, u64 bytes)
3549 stats->bytes += bytes;
3550 stats->pkts += pkts;
3554 * ice_update_tx_ring_stats - Update Tx ring specific counters
3555 * @tx_ring: ring to update
3556 * @pkts: number of processed packets
3557 * @bytes: number of processed bytes
3559 void ice_update_tx_ring_stats(struct ice_tx_ring *tx_ring, u64 pkts, u64 bytes)
3561 u64_stats_update_begin(&tx_ring->ring_stats->syncp);
3562 ice_update_ring_stats(&tx_ring->ring_stats->stats, pkts, bytes);
3563 u64_stats_update_end(&tx_ring->ring_stats->syncp);
3567 * ice_update_rx_ring_stats - Update Rx ring specific counters
3568 * @rx_ring: ring to update
3569 * @pkts: number of processed packets
3570 * @bytes: number of processed bytes
3572 void ice_update_rx_ring_stats(struct ice_rx_ring *rx_ring, u64 pkts, u64 bytes)
3574 u64_stats_update_begin(&rx_ring->ring_stats->syncp);
3575 ice_update_ring_stats(&rx_ring->ring_stats->stats, pkts, bytes);
3576 u64_stats_update_end(&rx_ring->ring_stats->syncp);
3580 * ice_is_dflt_vsi_in_use - check if the default forwarding VSI is being used
3581 * @pi: port info of the switch with default VSI
3583 * Return true if the there is a single VSI in default forwarding VSI list
3585 bool ice_is_dflt_vsi_in_use(struct ice_port_info *pi)
3587 bool exists = false;
3589 ice_check_if_dflt_vsi(pi, 0, &exists);
3594 * ice_is_vsi_dflt_vsi - check if the VSI passed in is the default VSI
3595 * @vsi: VSI to compare against default forwarding VSI
3597 * If this VSI passed in is the default forwarding VSI then return true, else
3600 bool ice_is_vsi_dflt_vsi(struct ice_vsi *vsi)
3602 return ice_check_if_dflt_vsi(vsi->port_info, vsi->idx, NULL);
3606 * ice_set_dflt_vsi - set the default forwarding VSI
3607 * @vsi: VSI getting set as the default forwarding VSI on the switch
3609 * If the VSI passed in is already the default VSI and it's enabled just return
3612 * Otherwise try to set the VSI passed in as the switch's default VSI and
3613 * return the result.
3615 int ice_set_dflt_vsi(struct ice_vsi *vsi)
3623 dev = ice_pf_to_dev(vsi->back);
3625 if (ice_lag_is_switchdev_running(vsi->back)) {
3626 dev_dbg(dev, "VSI %d passed is a part of LAG containing interfaces in switchdev mode, nothing to do\n",
3631 /* the VSI passed in is already the default VSI */
3632 if (ice_is_vsi_dflt_vsi(vsi)) {
3633 dev_dbg(dev, "VSI %d passed in is already the default forwarding VSI, nothing to do\n",
3638 status = ice_cfg_dflt_vsi(vsi->port_info, vsi->idx, true, ICE_FLTR_RX);
3640 dev_err(dev, "Failed to set VSI %d as the default forwarding VSI, error %d\n",
3641 vsi->vsi_num, status);
3649 * ice_clear_dflt_vsi - clear the default forwarding VSI
3650 * @vsi: VSI to remove from filter list
3652 * If the switch has no default VSI or it's not enabled then return error.
3654 * Otherwise try to clear the default VSI and return the result.
3656 int ice_clear_dflt_vsi(struct ice_vsi *vsi)
3664 dev = ice_pf_to_dev(vsi->back);
3666 /* there is no default VSI configured */
3667 if (!ice_is_dflt_vsi_in_use(vsi->port_info))
3670 status = ice_cfg_dflt_vsi(vsi->port_info, vsi->idx, false,
3673 dev_err(dev, "Failed to clear the default forwarding VSI %d, error %d\n",
3674 vsi->vsi_num, status);
3682 * ice_get_link_speed_mbps - get link speed in Mbps
3683 * @vsi: the VSI whose link speed is being queried
3685 * Return current VSI link speed and 0 if the speed is unknown.
3687 int ice_get_link_speed_mbps(struct ice_vsi *vsi)
3689 unsigned int link_speed;
3691 link_speed = vsi->port_info->phy.link_info.link_speed;
3693 return (int)ice_get_link_speed(fls(link_speed) - 1);
3697 * ice_get_link_speed_kbps - get link speed in Kbps
3698 * @vsi: the VSI whose link speed is being queried
3700 * Return current VSI link speed and 0 if the speed is unknown.
3702 int ice_get_link_speed_kbps(struct ice_vsi *vsi)
3706 speed_mbps = ice_get_link_speed_mbps(vsi);
3708 return speed_mbps * 1000;
3712 * ice_set_min_bw_limit - setup minimum BW limit for Tx based on min_tx_rate
3713 * @vsi: VSI to be configured
3714 * @min_tx_rate: min Tx rate in Kbps to be configured as BW limit
3716 * If the min_tx_rate is specified as 0 that means to clear the minimum BW limit
3717 * profile, otherwise a non-zero value will force a minimum BW limit for the VSI
3720 int ice_set_min_bw_limit(struct ice_vsi *vsi, u64 min_tx_rate)
3722 struct ice_pf *pf = vsi->back;
3727 dev = ice_pf_to_dev(pf);
3728 if (!vsi->port_info) {
3729 dev_dbg(dev, "VSI %d, type %u specified doesn't have valid port_info\n",
3730 vsi->idx, vsi->type);
3734 speed = ice_get_link_speed_kbps(vsi);
3735 if (min_tx_rate > (u64)speed) {
3736 dev_err(dev, "invalid min Tx rate %llu Kbps specified for %s %d is greater than current link speed %u Kbps\n",
3737 min_tx_rate, ice_vsi_type_str(vsi->type), vsi->idx,
3742 /* Configure min BW for VSI limit */
3744 status = ice_cfg_vsi_bw_lmt_per_tc(vsi->port_info, vsi->idx, 0,
3745 ICE_MIN_BW, min_tx_rate);
3747 dev_err(dev, "failed to set min Tx rate(%llu Kbps) for %s %d\n",
3748 min_tx_rate, ice_vsi_type_str(vsi->type),
3753 dev_dbg(dev, "set min Tx rate(%llu Kbps) for %s\n",
3754 min_tx_rate, ice_vsi_type_str(vsi->type));
3756 status = ice_cfg_vsi_bw_dflt_lmt_per_tc(vsi->port_info,
3760 dev_err(dev, "failed to clear min Tx rate configuration for %s %d\n",
3761 ice_vsi_type_str(vsi->type), vsi->idx);
3765 dev_dbg(dev, "cleared min Tx rate configuration for %s %d\n",
3766 ice_vsi_type_str(vsi->type), vsi->idx);
3773 * ice_set_max_bw_limit - setup maximum BW limit for Tx based on max_tx_rate
3774 * @vsi: VSI to be configured
3775 * @max_tx_rate: max Tx rate in Kbps to be configured as BW limit
3777 * If the max_tx_rate is specified as 0 that means to clear the maximum BW limit
3778 * profile, otherwise a non-zero value will force a maximum BW limit for the VSI
3781 int ice_set_max_bw_limit(struct ice_vsi *vsi, u64 max_tx_rate)
3783 struct ice_pf *pf = vsi->back;
3788 dev = ice_pf_to_dev(pf);
3789 if (!vsi->port_info) {
3790 dev_dbg(dev, "VSI %d, type %u specified doesn't have valid port_info\n",
3791 vsi->idx, vsi->type);
3795 speed = ice_get_link_speed_kbps(vsi);
3796 if (max_tx_rate > (u64)speed) {
3797 dev_err(dev, "invalid max Tx rate %llu Kbps specified for %s %d is greater than current link speed %u Kbps\n",
3798 max_tx_rate, ice_vsi_type_str(vsi->type), vsi->idx,
3803 /* Configure max BW for VSI limit */
3805 status = ice_cfg_vsi_bw_lmt_per_tc(vsi->port_info, vsi->idx, 0,
3806 ICE_MAX_BW, max_tx_rate);
3808 dev_err(dev, "failed setting max Tx rate(%llu Kbps) for %s %d\n",
3809 max_tx_rate, ice_vsi_type_str(vsi->type),
3814 dev_dbg(dev, "set max Tx rate(%llu Kbps) for %s %d\n",
3815 max_tx_rate, ice_vsi_type_str(vsi->type), vsi->idx);
3817 status = ice_cfg_vsi_bw_dflt_lmt_per_tc(vsi->port_info,
3821 dev_err(dev, "failed clearing max Tx rate configuration for %s %d\n",
3822 ice_vsi_type_str(vsi->type), vsi->idx);
3826 dev_dbg(dev, "cleared max Tx rate configuration for %s %d\n",
3827 ice_vsi_type_str(vsi->type), vsi->idx);
3834 * ice_set_link - turn on/off physical link
3835 * @vsi: VSI to modify physical link on
3836 * @ena: turn on/off physical link
3838 int ice_set_link(struct ice_vsi *vsi, bool ena)
3840 struct device *dev = ice_pf_to_dev(vsi->back);
3841 struct ice_port_info *pi = vsi->port_info;
3842 struct ice_hw *hw = pi->hw;
3845 if (vsi->type != ICE_VSI_PF)
3848 status = ice_aq_set_link_restart_an(pi, ena, NULL);
3850 /* if link is owned by manageability, FW will return ICE_AQ_RC_EMODE.
3851 * this is not a fatal error, so print a warning message and return
3852 * a success code. Return an error if FW returns an error code other
3853 * than ICE_AQ_RC_EMODE
3855 if (status == -EIO) {
3856 if (hw->adminq.sq_last_status == ICE_AQ_RC_EMODE)
3857 dev_dbg(dev, "can't set link to %s, err %d aq_err %s. not fatal, continuing\n",
3858 (ena ? "ON" : "OFF"), status,
3859 ice_aq_str(hw->adminq.sq_last_status));
3860 } else if (status) {
3861 dev_err(dev, "can't set link to %s, err %d aq_err %s\n",
3862 (ena ? "ON" : "OFF"), status,
3863 ice_aq_str(hw->adminq.sq_last_status));
3871 * ice_vsi_add_vlan_zero - add VLAN 0 filter(s) for this VSI
3872 * @vsi: VSI used to add VLAN filters
3874 * In Single VLAN Mode (SVM), single VLAN filters via ICE_SW_LKUP_VLAN are based
3875 * on the inner VLAN ID, so the VLAN TPID (i.e. 0x8100 or 0x888a8) doesn't
3876 * matter. In Double VLAN Mode (DVM), outer/single VLAN filters via
3877 * ICE_SW_LKUP_VLAN are based on the outer/single VLAN ID + VLAN TPID.
3879 * For both modes add a VLAN 0 + no VLAN TPID filter to handle untagged traffic
3880 * when VLAN pruning is enabled. Also, this handles VLAN 0 priority tagged
3881 * traffic in SVM, since the VLAN TPID isn't part of filtering.
3883 * If DVM is enabled then an explicit VLAN 0 + VLAN TPID filter needs to be
3884 * added to allow VLAN 0 priority tagged traffic in DVM, since the VLAN TPID is
3885 * part of filtering.
3887 int ice_vsi_add_vlan_zero(struct ice_vsi *vsi)
3889 struct ice_vsi_vlan_ops *vlan_ops = ice_get_compat_vsi_vlan_ops(vsi);
3890 struct ice_vlan vlan;
3893 vlan = ICE_VLAN(0, 0, 0);
3894 err = vlan_ops->add_vlan(vsi, &vlan);
3895 if (err && err != -EEXIST)
3898 /* in SVM both VLAN 0 filters are identical */
3899 if (!ice_is_dvm_ena(&vsi->back->hw))
3902 vlan = ICE_VLAN(ETH_P_8021Q, 0, 0);
3903 err = vlan_ops->add_vlan(vsi, &vlan);
3904 if (err && err != -EEXIST)
3911 * ice_vsi_del_vlan_zero - delete VLAN 0 filter(s) for this VSI
3912 * @vsi: VSI used to add VLAN filters
3914 * Delete the VLAN 0 filters in the same manner that they were added in
3915 * ice_vsi_add_vlan_zero.
3917 int ice_vsi_del_vlan_zero(struct ice_vsi *vsi)
3919 struct ice_vsi_vlan_ops *vlan_ops = ice_get_compat_vsi_vlan_ops(vsi);
3920 struct ice_vlan vlan;
3923 vlan = ICE_VLAN(0, 0, 0);
3924 err = vlan_ops->del_vlan(vsi, &vlan);
3925 if (err && err != -EEXIST)
3928 /* in SVM both VLAN 0 filters are identical */
3929 if (!ice_is_dvm_ena(&vsi->back->hw))
3932 vlan = ICE_VLAN(ETH_P_8021Q, 0, 0);
3933 err = vlan_ops->del_vlan(vsi, &vlan);
3934 if (err && err != -EEXIST)
3937 /* when deleting the last VLAN filter, make sure to disable the VLAN
3938 * promisc mode so the filter isn't left by accident
3940 return ice_clear_vsi_promisc(&vsi->back->hw, vsi->idx,
3941 ICE_MCAST_VLAN_PROMISC_BITS, 0);
3945 * ice_vsi_num_zero_vlans - get number of VLAN 0 filters based on VLAN mode
3946 * @vsi: VSI used to get the VLAN mode
3948 * If DVM is enabled then 2 VLAN 0 filters are added, else if SVM is enabled
3949 * then 1 VLAN 0 filter is added. See ice_vsi_add_vlan_zero for more details.
3951 static u16 ice_vsi_num_zero_vlans(struct ice_vsi *vsi)
3953 #define ICE_DVM_NUM_ZERO_VLAN_FLTRS 2
3954 #define ICE_SVM_NUM_ZERO_VLAN_FLTRS 1
3955 /* no VLAN 0 filter is created when a port VLAN is active */
3956 if (vsi->type == ICE_VSI_VF) {
3957 if (WARN_ON(!vsi->vf))
3960 if (ice_vf_is_port_vlan_ena(vsi->vf))
3964 if (ice_is_dvm_ena(&vsi->back->hw))
3965 return ICE_DVM_NUM_ZERO_VLAN_FLTRS;
3967 return ICE_SVM_NUM_ZERO_VLAN_FLTRS;
3971 * ice_vsi_has_non_zero_vlans - check if VSI has any non-zero VLANs
3972 * @vsi: VSI used to determine if any non-zero VLANs have been added
3974 bool ice_vsi_has_non_zero_vlans(struct ice_vsi *vsi)
3976 return (vsi->num_vlan > ice_vsi_num_zero_vlans(vsi));
3980 * ice_vsi_num_non_zero_vlans - get the number of non-zero VLANs for this VSI
3981 * @vsi: VSI used to get the number of non-zero VLANs added
3983 u16 ice_vsi_num_non_zero_vlans(struct ice_vsi *vsi)
3985 return (vsi->num_vlan - ice_vsi_num_zero_vlans(vsi));
3989 * ice_is_feature_supported
3990 * @pf: pointer to the struct ice_pf instance
3991 * @f: feature enum to be checked
3993 * returns true if feature is supported, false otherwise
3995 bool ice_is_feature_supported(struct ice_pf *pf, enum ice_feature f)
3997 if (f < 0 || f >= ICE_F_MAX)
4000 return test_bit(f, pf->features);
4004 * ice_set_feature_support
4005 * @pf: pointer to the struct ice_pf instance
4006 * @f: feature enum to set
4008 void ice_set_feature_support(struct ice_pf *pf, enum ice_feature f)
4010 if (f < 0 || f >= ICE_F_MAX)
4013 set_bit(f, pf->features);
4017 * ice_clear_feature_support
4018 * @pf: pointer to the struct ice_pf instance
4019 * @f: feature enum to clear
4021 void ice_clear_feature_support(struct ice_pf *pf, enum ice_feature f)
4023 if (f < 0 || f >= ICE_F_MAX)
4026 clear_bit(f, pf->features);
4030 * ice_init_feature_support
4031 * @pf: pointer to the struct ice_pf instance
4033 * called during init to setup supported feature
4035 void ice_init_feature_support(struct ice_pf *pf)
4037 switch (pf->hw.device_id) {
4038 case ICE_DEV_ID_E810C_BACKPLANE:
4039 case ICE_DEV_ID_E810C_QSFP:
4040 case ICE_DEV_ID_E810C_SFP:
4041 case ICE_DEV_ID_E810_XXV_BACKPLANE:
4042 case ICE_DEV_ID_E810_XXV_QSFP:
4043 case ICE_DEV_ID_E810_XXV_SFP:
4044 ice_set_feature_support(pf, ICE_F_DSCP);
4045 if (ice_is_phy_rclk_in_netlist(&pf->hw))
4046 ice_set_feature_support(pf, ICE_F_PHY_RCLK);
4047 /* If we don't own the timer - don't enable other caps */
4048 if (!ice_pf_src_tmr_owned(pf))
4050 if (ice_is_cgu_in_netlist(&pf->hw))
4051 ice_set_feature_support(pf, ICE_F_CGU);
4052 if (ice_is_clock_mux_in_netlist(&pf->hw))
4053 ice_set_feature_support(pf, ICE_F_SMA_CTRL);
4054 if (ice_gnss_is_gps_present(&pf->hw))
4055 ice_set_feature_support(pf, ICE_F_GNSS);
4063 * ice_vsi_update_security - update security block in VSI
4064 * @vsi: pointer to VSI structure
4065 * @fill: function pointer to fill ctx
4068 ice_vsi_update_security(struct ice_vsi *vsi, void (*fill)(struct ice_vsi_ctx *))
4070 struct ice_vsi_ctx ctx = { 0 };
4072 ctx.info = vsi->info;
4073 ctx.info.valid_sections = cpu_to_le16(ICE_AQ_VSI_PROP_SECURITY_VALID);
4076 if (ice_update_vsi(&vsi->back->hw, vsi->idx, &ctx, NULL))
4079 vsi->info = ctx.info;
4084 * ice_vsi_ctx_set_antispoof - set antispoof function in VSI ctx
4085 * @ctx: pointer to VSI ctx structure
4087 void ice_vsi_ctx_set_antispoof(struct ice_vsi_ctx *ctx)
4089 ctx->info.sec_flags |= ICE_AQ_VSI_SEC_FLAG_ENA_MAC_ANTI_SPOOF |
4090 (ICE_AQ_VSI_SEC_TX_VLAN_PRUNE_ENA <<
4091 ICE_AQ_VSI_SEC_TX_PRUNE_ENA_S);
4095 * ice_vsi_ctx_clear_antispoof - clear antispoof function in VSI ctx
4096 * @ctx: pointer to VSI ctx structure
4098 void ice_vsi_ctx_clear_antispoof(struct ice_vsi_ctx *ctx)
4100 ctx->info.sec_flags &= ~ICE_AQ_VSI_SEC_FLAG_ENA_MAC_ANTI_SPOOF &
4101 ~(ICE_AQ_VSI_SEC_TX_VLAN_PRUNE_ENA <<
4102 ICE_AQ_VSI_SEC_TX_PRUNE_ENA_S);
4106 * ice_vsi_ctx_set_allow_override - allow destination override on VSI
4107 * @ctx: pointer to VSI ctx structure
4109 void ice_vsi_ctx_set_allow_override(struct ice_vsi_ctx *ctx)
4111 ctx->info.sec_flags |= ICE_AQ_VSI_SEC_FLAG_ALLOW_DEST_OVRD;
4115 * ice_vsi_ctx_clear_allow_override - turn off destination override on VSI
4116 * @ctx: pointer to VSI ctx structure
4118 void ice_vsi_ctx_clear_allow_override(struct ice_vsi_ctx *ctx)
4120 ctx->info.sec_flags &= ~ICE_AQ_VSI_SEC_FLAG_ALLOW_DEST_OVRD;
4124 * ice_vsi_update_local_lb - update sw block in VSI with local loopback bit
4125 * @vsi: pointer to VSI structure
4126 * @set: set or unset the bit
4129 ice_vsi_update_local_lb(struct ice_vsi *vsi, bool set)
4131 struct ice_vsi_ctx ctx = {
4135 ctx.info.valid_sections = cpu_to_le16(ICE_AQ_VSI_PROP_SW_VALID);
4137 ctx.info.sw_flags |= ICE_AQ_VSI_SW_FLAG_LOCAL_LB;
4139 ctx.info.sw_flags &= ~ICE_AQ_VSI_SW_FLAG_LOCAL_LB;
4141 if (ice_update_vsi(&vsi->back->hw, vsi->idx, &ctx, NULL))
4144 vsi->info = ctx.info;