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 VFs.
216 * Each ring is associated to the corresponding VF_PR netdev.
218 vsi->alloc_txq = ice_get_num_vfs(pf);
219 vsi->alloc_rxq = vsi->alloc_txq;
220 vsi->num_q_vectors = 1;
224 vf->num_vf_qs = vf->num_req_qs;
225 vsi->alloc_txq = vf->num_vf_qs;
226 vsi->alloc_rxq = vf->num_vf_qs;
227 /* pf->vfs.num_msix_per includes (VF miscellaneous vector +
228 * data queue interrupts). Since vsi->num_q_vectors is number
229 * of queues vectors, subtract 1 (ICE_NONQ_VECS_VF) from the
230 * original vector count
232 vsi->num_q_vectors = vf->num_msix - ICE_NONQ_VECS_VF;
237 vsi->num_q_vectors = 1;
248 dev_warn(ice_pf_to_dev(pf), "Unknown VSI type %d\n", vsi_type);
252 ice_vsi_set_num_desc(vsi);
256 * ice_get_free_slot - get the next non-NULL location index in array
257 * @array: array to search
258 * @size: size of the array
259 * @curr: last known occupied index to be used as a search hint
261 * void * is being used to keep the functionality generic. This lets us use this
262 * function on any array of pointers.
264 static int ice_get_free_slot(void *array, int size, int curr)
266 int **tmp_array = (int **)array;
269 if (curr < (size - 1) && !tmp_array[curr + 1]) {
274 while ((i < size) && (tmp_array[i]))
285 * ice_vsi_delete_from_hw - delete a VSI from the switch
286 * @vsi: pointer to VSI being removed
288 static void ice_vsi_delete_from_hw(struct ice_vsi *vsi)
290 struct ice_pf *pf = vsi->back;
291 struct ice_vsi_ctx *ctxt;
294 ice_fltr_remove_all(vsi);
295 ctxt = kzalloc(sizeof(*ctxt), GFP_KERNEL);
299 if (vsi->type == ICE_VSI_VF)
300 ctxt->vf_num = vsi->vf->vf_id;
301 ctxt->vsi_num = vsi->vsi_num;
303 memcpy(&ctxt->info, &vsi->info, sizeof(ctxt->info));
305 status = ice_free_vsi(&pf->hw, vsi->idx, ctxt, false, NULL);
307 dev_err(ice_pf_to_dev(pf), "Failed to delete VSI %i in FW - error: %d\n",
308 vsi->vsi_num, status);
314 * ice_vsi_free_arrays - De-allocate queue and vector pointer arrays for the VSI
315 * @vsi: pointer to VSI being cleared
317 static void ice_vsi_free_arrays(struct ice_vsi *vsi)
319 struct ice_pf *pf = vsi->back;
322 dev = ice_pf_to_dev(pf);
324 bitmap_free(vsi->af_xdp_zc_qps);
325 vsi->af_xdp_zc_qps = NULL;
326 /* free the ring and vector containers */
327 devm_kfree(dev, vsi->q_vectors);
328 vsi->q_vectors = NULL;
329 devm_kfree(dev, vsi->tx_rings);
330 vsi->tx_rings = NULL;
331 devm_kfree(dev, vsi->rx_rings);
332 vsi->rx_rings = NULL;
333 devm_kfree(dev, vsi->txq_map);
335 devm_kfree(dev, vsi->rxq_map);
340 * ice_vsi_free_stats - Free the ring statistics structures
343 static void ice_vsi_free_stats(struct ice_vsi *vsi)
345 struct ice_vsi_stats *vsi_stat;
346 struct ice_pf *pf = vsi->back;
349 if (vsi->type == ICE_VSI_CHNL)
354 vsi_stat = pf->vsi_stats[vsi->idx];
358 ice_for_each_alloc_txq(vsi, i) {
359 if (vsi_stat->tx_ring_stats[i]) {
360 kfree_rcu(vsi_stat->tx_ring_stats[i], rcu);
361 WRITE_ONCE(vsi_stat->tx_ring_stats[i], NULL);
365 ice_for_each_alloc_rxq(vsi, i) {
366 if (vsi_stat->rx_ring_stats[i]) {
367 kfree_rcu(vsi_stat->rx_ring_stats[i], rcu);
368 WRITE_ONCE(vsi_stat->rx_ring_stats[i], NULL);
372 kfree(vsi_stat->tx_ring_stats);
373 kfree(vsi_stat->rx_ring_stats);
375 pf->vsi_stats[vsi->idx] = NULL;
379 * ice_vsi_alloc_ring_stats - Allocates Tx and Rx ring stats for the VSI
380 * @vsi: VSI which is having stats allocated
382 static int ice_vsi_alloc_ring_stats(struct ice_vsi *vsi)
384 struct ice_ring_stats **tx_ring_stats;
385 struct ice_ring_stats **rx_ring_stats;
386 struct ice_vsi_stats *vsi_stats;
387 struct ice_pf *pf = vsi->back;
390 vsi_stats = pf->vsi_stats[vsi->idx];
391 tx_ring_stats = vsi_stats->tx_ring_stats;
392 rx_ring_stats = vsi_stats->rx_ring_stats;
394 /* Allocate Tx ring stats */
395 ice_for_each_alloc_txq(vsi, i) {
396 struct ice_ring_stats *ring_stats;
397 struct ice_tx_ring *ring;
399 ring = vsi->tx_rings[i];
400 ring_stats = tx_ring_stats[i];
403 ring_stats = kzalloc(sizeof(*ring_stats), GFP_KERNEL);
407 WRITE_ONCE(tx_ring_stats[i], ring_stats);
410 ring->ring_stats = ring_stats;
413 /* Allocate Rx ring stats */
414 ice_for_each_alloc_rxq(vsi, i) {
415 struct ice_ring_stats *ring_stats;
416 struct ice_rx_ring *ring;
418 ring = vsi->rx_rings[i];
419 ring_stats = rx_ring_stats[i];
422 ring_stats = kzalloc(sizeof(*ring_stats), GFP_KERNEL);
426 WRITE_ONCE(rx_ring_stats[i], ring_stats);
429 ring->ring_stats = ring_stats;
435 ice_vsi_free_stats(vsi);
440 * ice_vsi_free - clean up and deallocate the provided VSI
441 * @vsi: pointer to VSI being cleared
443 * This deallocates the VSI's queue resources, removes it from the PF's
444 * VSI array if necessary, and deallocates the VSI
446 static void ice_vsi_free(struct ice_vsi *vsi)
448 struct ice_pf *pf = NULL;
451 if (!vsi || !vsi->back)
455 dev = ice_pf_to_dev(pf);
457 if (!pf->vsi[vsi->idx] || pf->vsi[vsi->idx] != vsi) {
458 dev_dbg(dev, "vsi does not exist at pf->vsi[%d]\n", vsi->idx);
462 mutex_lock(&pf->sw_mutex);
463 /* updates the PF for this cleared VSI */
465 pf->vsi[vsi->idx] = NULL;
466 pf->next_vsi = vsi->idx;
468 ice_vsi_free_stats(vsi);
469 ice_vsi_free_arrays(vsi);
470 mutex_unlock(&pf->sw_mutex);
471 devm_kfree(dev, vsi);
474 void ice_vsi_delete(struct ice_vsi *vsi)
476 ice_vsi_delete_from_hw(vsi);
481 * ice_msix_clean_ctrl_vsi - MSIX mode interrupt handler for ctrl VSI
482 * @irq: interrupt number
483 * @data: pointer to a q_vector
485 static irqreturn_t ice_msix_clean_ctrl_vsi(int __always_unused irq, void *data)
487 struct ice_q_vector *q_vector = (struct ice_q_vector *)data;
489 if (!q_vector->tx.tx_ring)
492 #define FDIR_RX_DESC_CLEAN_BUDGET 64
493 ice_clean_rx_irq(q_vector->rx.rx_ring, FDIR_RX_DESC_CLEAN_BUDGET);
494 ice_clean_ctrl_tx_irq(q_vector->tx.tx_ring);
500 * ice_msix_clean_rings - MSIX mode Interrupt Handler
501 * @irq: interrupt number
502 * @data: pointer to a q_vector
504 static irqreturn_t ice_msix_clean_rings(int __always_unused irq, void *data)
506 struct ice_q_vector *q_vector = (struct ice_q_vector *)data;
508 if (!q_vector->tx.tx_ring && !q_vector->rx.rx_ring)
511 q_vector->total_events++;
513 napi_schedule(&q_vector->napi);
518 static irqreturn_t ice_eswitch_msix_clean_rings(int __always_unused irq, void *data)
520 struct ice_q_vector *q_vector = (struct ice_q_vector *)data;
521 struct ice_pf *pf = q_vector->vsi->back;
525 if (!q_vector->tx.tx_ring && !q_vector->rx.rx_ring)
529 ice_for_each_vf_rcu(pf, bkt, vf)
530 napi_schedule(&vf->repr->q_vector->napi);
537 * ice_vsi_alloc_stat_arrays - Allocate statistics arrays
540 static int ice_vsi_alloc_stat_arrays(struct ice_vsi *vsi)
542 struct ice_vsi_stats *vsi_stat;
543 struct ice_pf *pf = vsi->back;
545 if (vsi->type == ICE_VSI_CHNL)
550 if (pf->vsi_stats[vsi->idx])
551 /* realloc will happen in rebuild path */
554 vsi_stat = kzalloc(sizeof(*vsi_stat), GFP_KERNEL);
558 vsi_stat->tx_ring_stats =
559 kcalloc(vsi->alloc_txq, sizeof(*vsi_stat->tx_ring_stats),
561 if (!vsi_stat->tx_ring_stats)
564 vsi_stat->rx_ring_stats =
565 kcalloc(vsi->alloc_rxq, sizeof(*vsi_stat->rx_ring_stats),
567 if (!vsi_stat->rx_ring_stats)
570 pf->vsi_stats[vsi->idx] = vsi_stat;
575 kfree(vsi_stat->rx_ring_stats);
577 kfree(vsi_stat->tx_ring_stats);
579 pf->vsi_stats[vsi->idx] = NULL;
584 * ice_vsi_alloc_def - set default values for already allocated VSI
586 * @ch: ptr to channel
589 ice_vsi_alloc_def(struct ice_vsi *vsi, struct ice_channel *ch)
591 if (vsi->type != ICE_VSI_CHNL) {
592 ice_vsi_set_num_qs(vsi);
593 if (ice_vsi_alloc_arrays(vsi))
598 case ICE_VSI_SWITCHDEV_CTRL:
599 /* Setup eswitch MSIX irq handler for VSI */
600 vsi->irq_handler = ice_eswitch_msix_clean_rings;
603 /* Setup default MSIX irq handler for VSI */
604 vsi->irq_handler = ice_msix_clean_rings;
607 /* Setup ctrl VSI MSIX irq handler */
608 vsi->irq_handler = ice_msix_clean_ctrl_vsi;
614 vsi->num_rxq = ch->num_rxq;
615 vsi->num_txq = ch->num_txq;
616 vsi->next_base_q = ch->base_q;
622 ice_vsi_free_arrays(vsi);
630 * ice_vsi_alloc - Allocates the next available struct VSI in the PF
631 * @pf: board private structure
633 * Reserves a VSI index from the PF and allocates an empty VSI structure
634 * without a type. The VSI structure must later be initialized by calling
637 * returns a pointer to a VSI on success, NULL on failure.
639 static struct ice_vsi *ice_vsi_alloc(struct ice_pf *pf)
641 struct device *dev = ice_pf_to_dev(pf);
642 struct ice_vsi *vsi = NULL;
644 /* Need to protect the allocation of the VSIs at the PF level */
645 mutex_lock(&pf->sw_mutex);
647 /* If we have already allocated our maximum number of VSIs,
648 * pf->next_vsi will be ICE_NO_VSI. If not, pf->next_vsi index
649 * is available to be populated
651 if (pf->next_vsi == ICE_NO_VSI) {
652 dev_dbg(dev, "out of VSI slots!\n");
656 vsi = devm_kzalloc(dev, sizeof(*vsi), GFP_KERNEL);
661 set_bit(ICE_VSI_DOWN, vsi->state);
663 /* fill slot and make note of the index */
664 vsi->idx = pf->next_vsi;
665 pf->vsi[pf->next_vsi] = vsi;
667 /* prepare pf->next_vsi for next use */
668 pf->next_vsi = ice_get_free_slot(pf->vsi, pf->num_alloc_vsi,
672 mutex_unlock(&pf->sw_mutex);
677 * ice_alloc_fd_res - Allocate FD resource for a VSI
678 * @vsi: pointer to the ice_vsi
680 * This allocates the FD resources
682 * Returns 0 on success, -EPERM on no-op or -EIO on failure
684 static int ice_alloc_fd_res(struct ice_vsi *vsi)
686 struct ice_pf *pf = vsi->back;
689 /* Flow Director filters are only allocated/assigned to the PF VSI or
690 * CHNL VSI which passes the traffic. The CTRL VSI is only used to
691 * add/delete filters so resources are not allocated to it
693 if (!test_bit(ICE_FLAG_FD_ENA, pf->flags))
696 if (!(vsi->type == ICE_VSI_PF || vsi->type == ICE_VSI_VF ||
697 vsi->type == ICE_VSI_CHNL))
700 /* FD filters from guaranteed pool per VSI */
701 g_val = pf->hw.func_caps.fd_fltr_guar;
705 /* FD filters from best effort pool */
706 b_val = pf->hw.func_caps.fd_fltr_best_effort;
710 /* PF main VSI gets only 64 FD resources from guaranteed pool
711 * when ADQ is configured.
713 #define ICE_PF_VSI_GFLTR 64
715 /* determine FD filter resources per VSI from shared(best effort) and
718 if (vsi->type == ICE_VSI_PF) {
719 vsi->num_gfltr = g_val;
720 /* if MQPRIO is configured, main VSI doesn't get all FD
721 * resources from guaranteed pool. PF VSI gets 64 FD resources
723 if (test_bit(ICE_FLAG_TC_MQPRIO, pf->flags)) {
724 if (g_val < ICE_PF_VSI_GFLTR)
726 /* allow bare minimum entries for PF VSI */
727 vsi->num_gfltr = ICE_PF_VSI_GFLTR;
730 /* each VSI gets same "best_effort" quota */
731 vsi->num_bfltr = b_val;
732 } else if (vsi->type == ICE_VSI_VF) {
735 /* each VSI gets same "best_effort" quota */
736 vsi->num_bfltr = b_val;
738 struct ice_vsi *main_vsi;
741 main_vsi = ice_get_main_vsi(pf);
745 if (!main_vsi->all_numtc)
748 /* figure out ADQ numtc */
749 numtc = main_vsi->all_numtc - ICE_CHNL_START_TC;
751 /* only one TC but still asking resources for channels,
754 if (numtc < ICE_CHNL_START_TC)
757 g_val -= ICE_PF_VSI_GFLTR;
758 /* channel VSIs gets equal share from guaranteed pool */
759 vsi->num_gfltr = g_val / numtc;
761 /* each VSI gets same "best_effort" quota */
762 vsi->num_bfltr = b_val;
769 * ice_vsi_get_qs - Assign queues from PF to VSI
770 * @vsi: the VSI to assign queues to
772 * Returns 0 on success and a negative value on error
774 static int ice_vsi_get_qs(struct ice_vsi *vsi)
776 struct ice_pf *pf = vsi->back;
777 struct ice_qs_cfg tx_qs_cfg = {
778 .qs_mutex = &pf->avail_q_mutex,
779 .pf_map = pf->avail_txqs,
780 .pf_map_size = pf->max_pf_txqs,
781 .q_count = vsi->alloc_txq,
782 .scatter_count = ICE_MAX_SCATTER_TXQS,
783 .vsi_map = vsi->txq_map,
785 .mapping_mode = ICE_VSI_MAP_CONTIG
787 struct ice_qs_cfg rx_qs_cfg = {
788 .qs_mutex = &pf->avail_q_mutex,
789 .pf_map = pf->avail_rxqs,
790 .pf_map_size = pf->max_pf_rxqs,
791 .q_count = vsi->alloc_rxq,
792 .scatter_count = ICE_MAX_SCATTER_RXQS,
793 .vsi_map = vsi->rxq_map,
795 .mapping_mode = ICE_VSI_MAP_CONTIG
799 if (vsi->type == ICE_VSI_CHNL)
802 ret = __ice_vsi_get_qs(&tx_qs_cfg);
805 vsi->tx_mapping_mode = tx_qs_cfg.mapping_mode;
807 ret = __ice_vsi_get_qs(&rx_qs_cfg);
810 vsi->rx_mapping_mode = rx_qs_cfg.mapping_mode;
816 * ice_vsi_put_qs - Release queues from VSI to PF
817 * @vsi: the VSI that is going to release queues
819 static void ice_vsi_put_qs(struct ice_vsi *vsi)
821 struct ice_pf *pf = vsi->back;
824 mutex_lock(&pf->avail_q_mutex);
826 ice_for_each_alloc_txq(vsi, i) {
827 clear_bit(vsi->txq_map[i], pf->avail_txqs);
828 vsi->txq_map[i] = ICE_INVAL_Q_INDEX;
831 ice_for_each_alloc_rxq(vsi, i) {
832 clear_bit(vsi->rxq_map[i], pf->avail_rxqs);
833 vsi->rxq_map[i] = ICE_INVAL_Q_INDEX;
836 mutex_unlock(&pf->avail_q_mutex);
841 * @pf: pointer to the PF struct
843 * returns true if driver is in safe mode, false otherwise
845 bool ice_is_safe_mode(struct ice_pf *pf)
847 return !test_bit(ICE_FLAG_ADV_FEATURES, pf->flags);
852 * @pf: pointer to the PF struct
854 * returns true if RDMA is currently supported, false otherwise
856 bool ice_is_rdma_ena(struct ice_pf *pf)
858 return test_bit(ICE_FLAG_RDMA_ENA, pf->flags);
862 * ice_vsi_clean_rss_flow_fld - Delete RSS configuration
863 * @vsi: the VSI being cleaned up
865 * This function deletes RSS input set for all flows that were configured
868 static void ice_vsi_clean_rss_flow_fld(struct ice_vsi *vsi)
870 struct ice_pf *pf = vsi->back;
873 if (ice_is_safe_mode(pf))
876 status = ice_rem_vsi_rss_cfg(&pf->hw, vsi->idx);
878 dev_dbg(ice_pf_to_dev(pf), "ice_rem_vsi_rss_cfg failed for vsi = %d, error = %d\n",
879 vsi->vsi_num, status);
883 * ice_rss_clean - Delete RSS related VSI structures and configuration
884 * @vsi: the VSI being removed
886 static void ice_rss_clean(struct ice_vsi *vsi)
888 struct ice_pf *pf = vsi->back;
891 dev = ice_pf_to_dev(pf);
893 devm_kfree(dev, vsi->rss_hkey_user);
894 devm_kfree(dev, vsi->rss_lut_user);
896 ice_vsi_clean_rss_flow_fld(vsi);
897 /* remove RSS replay list */
898 if (!ice_is_safe_mode(pf))
899 ice_rem_vsi_rss_list(&pf->hw, vsi->idx);
903 * ice_vsi_set_rss_params - Setup RSS capabilities per VSI type
904 * @vsi: the VSI being configured
906 static void ice_vsi_set_rss_params(struct ice_vsi *vsi)
908 struct ice_hw_common_caps *cap;
909 struct ice_pf *pf = vsi->back;
912 if (!test_bit(ICE_FLAG_RSS_ENA, pf->flags)) {
917 cap = &pf->hw.func_caps.common_cap;
918 max_rss_size = BIT(cap->rss_table_entry_width);
922 /* PF VSI will inherit RSS instance of PF */
923 vsi->rss_table_size = (u16)cap->rss_table_size;
924 if (vsi->type == ICE_VSI_CHNL)
925 vsi->rss_size = min_t(u16, vsi->num_rxq, max_rss_size);
927 vsi->rss_size = min_t(u16, num_online_cpus(),
929 vsi->rss_lut_type = ICE_LUT_PF;
931 case ICE_VSI_SWITCHDEV_CTRL:
932 vsi->rss_table_size = ICE_LUT_VSI_SIZE;
933 vsi->rss_size = min_t(u16, num_online_cpus(), max_rss_size);
934 vsi->rss_lut_type = ICE_LUT_VSI;
937 /* VF VSI will get a small RSS table.
938 * For VSI_LUT, LUT size should be set to 64 bytes.
940 vsi->rss_table_size = ICE_LUT_VSI_SIZE;
941 vsi->rss_size = ICE_MAX_RSS_QS_PER_VF;
942 vsi->rss_lut_type = ICE_LUT_VSI;
947 dev_dbg(ice_pf_to_dev(pf), "Unsupported VSI type %s\n",
948 ice_vsi_type_str(vsi->type));
954 * ice_set_dflt_vsi_ctx - Set default VSI context before adding a VSI
955 * @hw: HW structure used to determine the VLAN mode of the device
956 * @ctxt: the VSI context being set
958 * This initializes a default VSI context for all sections except the Queues.
960 static void ice_set_dflt_vsi_ctx(struct ice_hw *hw, struct ice_vsi_ctx *ctxt)
964 memset(&ctxt->info, 0, sizeof(ctxt->info));
965 /* VSI's should be allocated from shared pool */
966 ctxt->alloc_from_pool = true;
967 /* Src pruning enabled by default */
968 ctxt->info.sw_flags = ICE_AQ_VSI_SW_FLAG_SRC_PRUNE;
969 /* Traffic from VSI can be sent to LAN */
970 ctxt->info.sw_flags2 = ICE_AQ_VSI_SW_FLAG_LAN_ENA;
971 /* allow all untagged/tagged packets by default on Tx */
972 ctxt->info.inner_vlan_flags = ((ICE_AQ_VSI_INNER_VLAN_TX_MODE_ALL &
973 ICE_AQ_VSI_INNER_VLAN_TX_MODE_M) >>
974 ICE_AQ_VSI_INNER_VLAN_TX_MODE_S);
975 /* SVM - by default bits 3 and 4 in inner_vlan_flags are 0's which
976 * results in legacy behavior (show VLAN, DEI, and UP) in descriptor.
978 * DVM - leave inner VLAN in packet by default
980 if (ice_is_dvm_ena(hw)) {
981 ctxt->info.inner_vlan_flags |=
982 ICE_AQ_VSI_INNER_VLAN_EMODE_NOTHING;
983 ctxt->info.outer_vlan_flags =
984 (ICE_AQ_VSI_OUTER_VLAN_TX_MODE_ALL <<
985 ICE_AQ_VSI_OUTER_VLAN_TX_MODE_S) &
986 ICE_AQ_VSI_OUTER_VLAN_TX_MODE_M;
987 ctxt->info.outer_vlan_flags |=
988 (ICE_AQ_VSI_OUTER_TAG_VLAN_8100 <<
989 ICE_AQ_VSI_OUTER_TAG_TYPE_S) &
990 ICE_AQ_VSI_OUTER_TAG_TYPE_M;
991 ctxt->info.outer_vlan_flags |=
992 FIELD_PREP(ICE_AQ_VSI_OUTER_VLAN_EMODE_M,
993 ICE_AQ_VSI_OUTER_VLAN_EMODE_NOTHING);
995 /* Have 1:1 UP mapping for both ingress/egress tables */
996 table |= ICE_UP_TABLE_TRANSLATE(0, 0);
997 table |= ICE_UP_TABLE_TRANSLATE(1, 1);
998 table |= ICE_UP_TABLE_TRANSLATE(2, 2);
999 table |= ICE_UP_TABLE_TRANSLATE(3, 3);
1000 table |= ICE_UP_TABLE_TRANSLATE(4, 4);
1001 table |= ICE_UP_TABLE_TRANSLATE(5, 5);
1002 table |= ICE_UP_TABLE_TRANSLATE(6, 6);
1003 table |= ICE_UP_TABLE_TRANSLATE(7, 7);
1004 ctxt->info.ingress_table = cpu_to_le32(table);
1005 ctxt->info.egress_table = cpu_to_le32(table);
1006 /* Have 1:1 UP mapping for outer to inner UP table */
1007 ctxt->info.outer_up_table = cpu_to_le32(table);
1008 /* No Outer tag support outer_tag_flags remains to zero */
1012 * ice_vsi_setup_q_map - Setup a VSI queue map
1013 * @vsi: the VSI being configured
1014 * @ctxt: VSI context structure
1016 static int ice_vsi_setup_q_map(struct ice_vsi *vsi, struct ice_vsi_ctx *ctxt)
1018 u16 offset = 0, qmap = 0, tx_count = 0, rx_count = 0, pow = 0;
1019 u16 num_txq_per_tc, num_rxq_per_tc;
1020 u16 qcount_tx = vsi->alloc_txq;
1021 u16 qcount_rx = vsi->alloc_rxq;
1025 if (!vsi->tc_cfg.numtc) {
1026 /* at least TC0 should be enabled by default */
1027 vsi->tc_cfg.numtc = 1;
1028 vsi->tc_cfg.ena_tc = 1;
1031 num_rxq_per_tc = min_t(u16, qcount_rx / vsi->tc_cfg.numtc, ICE_MAX_RXQS_PER_TC);
1032 if (!num_rxq_per_tc)
1034 num_txq_per_tc = qcount_tx / vsi->tc_cfg.numtc;
1035 if (!num_txq_per_tc)
1038 /* find the (rounded up) power-of-2 of qcount */
1039 pow = (u16)order_base_2(num_rxq_per_tc);
1041 /* TC mapping is a function of the number of Rx queues assigned to the
1042 * VSI for each traffic class and the offset of these queues.
1043 * The first 10 bits are for queue offset for TC0, next 4 bits for no:of
1044 * queues allocated to TC0. No:of queues is a power-of-2.
1046 * If TC is not enabled, the queue offset is set to 0, and allocate one
1047 * queue, this way, traffic for the given TC will be sent to the default
1050 * Setup number and offset of Rx queues for all TCs for the VSI
1052 ice_for_each_traffic_class(i) {
1053 if (!(vsi->tc_cfg.ena_tc & BIT(i))) {
1054 /* TC is not enabled */
1055 vsi->tc_cfg.tc_info[i].qoffset = 0;
1056 vsi->tc_cfg.tc_info[i].qcount_rx = 1;
1057 vsi->tc_cfg.tc_info[i].qcount_tx = 1;
1058 vsi->tc_cfg.tc_info[i].netdev_tc = 0;
1059 ctxt->info.tc_mapping[i] = 0;
1064 vsi->tc_cfg.tc_info[i].qoffset = offset;
1065 vsi->tc_cfg.tc_info[i].qcount_rx = num_rxq_per_tc;
1066 vsi->tc_cfg.tc_info[i].qcount_tx = num_txq_per_tc;
1067 vsi->tc_cfg.tc_info[i].netdev_tc = netdev_tc++;
1069 qmap = ((offset << ICE_AQ_VSI_TC_Q_OFFSET_S) &
1070 ICE_AQ_VSI_TC_Q_OFFSET_M) |
1071 ((pow << ICE_AQ_VSI_TC_Q_NUM_S) &
1072 ICE_AQ_VSI_TC_Q_NUM_M);
1073 offset += num_rxq_per_tc;
1074 tx_count += num_txq_per_tc;
1075 ctxt->info.tc_mapping[i] = cpu_to_le16(qmap);
1078 /* if offset is non-zero, means it is calculated correctly based on
1079 * enabled TCs for a given VSI otherwise qcount_rx will always
1080 * be correct and non-zero because it is based off - VSI's
1081 * allocated Rx queues which is at least 1 (hence qcount_tx will be
1087 rx_count = num_rxq_per_tc;
1089 if (rx_count > vsi->alloc_rxq) {
1090 dev_err(ice_pf_to_dev(vsi->back), "Trying to use more Rx queues (%u), than were allocated (%u)!\n",
1091 rx_count, vsi->alloc_rxq);
1095 if (tx_count > vsi->alloc_txq) {
1096 dev_err(ice_pf_to_dev(vsi->back), "Trying to use more Tx queues (%u), than were allocated (%u)!\n",
1097 tx_count, vsi->alloc_txq);
1101 vsi->num_txq = tx_count;
1102 vsi->num_rxq = rx_count;
1104 if (vsi->type == ICE_VSI_VF && vsi->num_txq != vsi->num_rxq) {
1105 dev_dbg(ice_pf_to_dev(vsi->back), "VF VSI should have same number of Tx and Rx queues. Hence making them equal\n");
1106 /* since there is a chance that num_rxq could have been changed
1107 * in the above for loop, make num_txq equal to num_rxq.
1109 vsi->num_txq = vsi->num_rxq;
1112 /* Rx queue mapping */
1113 ctxt->info.mapping_flags |= cpu_to_le16(ICE_AQ_VSI_Q_MAP_CONTIG);
1114 /* q_mapping buffer holds the info for the first queue allocated for
1115 * this VSI in the PF space and also the number of queues associated
1118 ctxt->info.q_mapping[0] = cpu_to_le16(vsi->rxq_map[0]);
1119 ctxt->info.q_mapping[1] = cpu_to_le16(vsi->num_rxq);
1125 * ice_set_fd_vsi_ctx - Set FD VSI context before adding a VSI
1126 * @ctxt: the VSI context being set
1127 * @vsi: the VSI being configured
1129 static void ice_set_fd_vsi_ctx(struct ice_vsi_ctx *ctxt, struct ice_vsi *vsi)
1131 u8 dflt_q_group, dflt_q_prio;
1132 u16 dflt_q, report_q, val;
1134 if (vsi->type != ICE_VSI_PF && vsi->type != ICE_VSI_CTRL &&
1135 vsi->type != ICE_VSI_VF && vsi->type != ICE_VSI_CHNL)
1138 val = ICE_AQ_VSI_PROP_FLOW_DIR_VALID;
1139 ctxt->info.valid_sections |= cpu_to_le16(val);
1145 /* enable flow director filtering/programming */
1146 val = ICE_AQ_VSI_FD_ENABLE | ICE_AQ_VSI_FD_PROG_ENABLE;
1147 ctxt->info.fd_options = cpu_to_le16(val);
1148 /* max of allocated flow director filters */
1149 ctxt->info.max_fd_fltr_dedicated =
1150 cpu_to_le16(vsi->num_gfltr);
1151 /* max of shared flow director filters any VSI may program */
1152 ctxt->info.max_fd_fltr_shared =
1153 cpu_to_le16(vsi->num_bfltr);
1154 /* default queue index within the VSI of the default FD */
1155 val = ((dflt_q << ICE_AQ_VSI_FD_DEF_Q_S) &
1156 ICE_AQ_VSI_FD_DEF_Q_M);
1157 /* target queue or queue group to the FD filter */
1158 val |= ((dflt_q_group << ICE_AQ_VSI_FD_DEF_GRP_S) &
1159 ICE_AQ_VSI_FD_DEF_GRP_M);
1160 ctxt->info.fd_def_q = cpu_to_le16(val);
1161 /* queue index on which FD filter completion is reported */
1162 val = ((report_q << ICE_AQ_VSI_FD_REPORT_Q_S) &
1163 ICE_AQ_VSI_FD_REPORT_Q_M);
1164 /* priority of the default qindex action */
1165 val |= ((dflt_q_prio << ICE_AQ_VSI_FD_DEF_PRIORITY_S) &
1166 ICE_AQ_VSI_FD_DEF_PRIORITY_M);
1167 ctxt->info.fd_report_opt = cpu_to_le16(val);
1171 * ice_set_rss_vsi_ctx - Set RSS VSI context before adding a VSI
1172 * @ctxt: the VSI context being set
1173 * @vsi: the VSI being configured
1175 static void ice_set_rss_vsi_ctx(struct ice_vsi_ctx *ctxt, struct ice_vsi *vsi)
1177 u8 lut_type, hash_type;
1182 dev = ice_pf_to_dev(pf);
1184 switch (vsi->type) {
1187 /* PF VSI will inherit RSS instance of PF */
1188 lut_type = ICE_AQ_VSI_Q_OPT_RSS_LUT_PF;
1189 hash_type = ICE_AQ_VSI_Q_OPT_RSS_TPLZ;
1192 /* VF VSI will gets a small RSS table which is a VSI LUT type */
1193 lut_type = ICE_AQ_VSI_Q_OPT_RSS_LUT_VSI;
1194 hash_type = ICE_AQ_VSI_Q_OPT_RSS_TPLZ;
1197 dev_dbg(dev, "Unsupported VSI type %s\n",
1198 ice_vsi_type_str(vsi->type));
1202 ctxt->info.q_opt_rss = ((lut_type << ICE_AQ_VSI_Q_OPT_RSS_LUT_S) &
1203 ICE_AQ_VSI_Q_OPT_RSS_LUT_M) |
1204 (hash_type & ICE_AQ_VSI_Q_OPT_RSS_HASH_M);
1208 ice_chnl_vsi_setup_q_map(struct ice_vsi *vsi, struct ice_vsi_ctx *ctxt)
1210 struct ice_pf *pf = vsi->back;
1215 qcount = min_t(int, vsi->num_rxq, pf->num_lan_msix);
1217 pow = order_base_2(qcount);
1218 qmap = ((offset << ICE_AQ_VSI_TC_Q_OFFSET_S) &
1219 ICE_AQ_VSI_TC_Q_OFFSET_M) |
1220 ((pow << ICE_AQ_VSI_TC_Q_NUM_S) &
1221 ICE_AQ_VSI_TC_Q_NUM_M);
1223 ctxt->info.tc_mapping[0] = cpu_to_le16(qmap);
1224 ctxt->info.mapping_flags |= cpu_to_le16(ICE_AQ_VSI_Q_MAP_CONTIG);
1225 ctxt->info.q_mapping[0] = cpu_to_le16(vsi->next_base_q);
1226 ctxt->info.q_mapping[1] = cpu_to_le16(qcount);
1230 * ice_vsi_is_vlan_pruning_ena - check if VLAN pruning is enabled or not
1231 * @vsi: VSI to check whether or not VLAN pruning is enabled.
1233 * returns true if Rx VLAN pruning is enabled and false otherwise.
1235 static bool ice_vsi_is_vlan_pruning_ena(struct ice_vsi *vsi)
1237 return vsi->info.sw_flags2 & ICE_AQ_VSI_SW_FLAG_RX_VLAN_PRUNE_ENA;
1241 * ice_vsi_init - Create and initialize a VSI
1242 * @vsi: the VSI being configured
1243 * @vsi_flags: VSI configuration flags
1245 * Set ICE_FLAG_VSI_INIT to initialize a new VSI context, clear it to
1246 * reconfigure an existing context.
1248 * This initializes a VSI context depending on the VSI type to be added and
1249 * passes it down to the add_vsi aq command to create a new VSI.
1251 static int ice_vsi_init(struct ice_vsi *vsi, u32 vsi_flags)
1253 struct ice_pf *pf = vsi->back;
1254 struct ice_hw *hw = &pf->hw;
1255 struct ice_vsi_ctx *ctxt;
1259 dev = ice_pf_to_dev(pf);
1260 ctxt = kzalloc(sizeof(*ctxt), GFP_KERNEL);
1264 switch (vsi->type) {
1268 ctxt->flags = ICE_AQ_VSI_TYPE_PF;
1270 case ICE_VSI_SWITCHDEV_CTRL:
1272 ctxt->flags = ICE_AQ_VSI_TYPE_VMDQ2;
1275 ctxt->flags = ICE_AQ_VSI_TYPE_VF;
1276 /* VF number here is the absolute VF number (0-255) */
1277 ctxt->vf_num = vsi->vf->vf_id + hw->func_caps.vf_base_id;
1284 /* Handle VLAN pruning for channel VSI if main VSI has VLAN
1287 if (vsi->type == ICE_VSI_CHNL) {
1288 struct ice_vsi *main_vsi;
1290 main_vsi = ice_get_main_vsi(pf);
1291 if (main_vsi && ice_vsi_is_vlan_pruning_ena(main_vsi))
1292 ctxt->info.sw_flags2 |=
1293 ICE_AQ_VSI_SW_FLAG_RX_VLAN_PRUNE_ENA;
1295 ctxt->info.sw_flags2 &=
1296 ~ICE_AQ_VSI_SW_FLAG_RX_VLAN_PRUNE_ENA;
1299 ice_set_dflt_vsi_ctx(hw, ctxt);
1300 if (test_bit(ICE_FLAG_FD_ENA, pf->flags))
1301 ice_set_fd_vsi_ctx(ctxt, vsi);
1302 /* if the switch is in VEB mode, allow VSI loopback */
1303 if (vsi->vsw->bridge_mode == BRIDGE_MODE_VEB)
1304 ctxt->info.sw_flags |= ICE_AQ_VSI_SW_FLAG_ALLOW_LB;
1306 /* Set LUT type and HASH type if RSS is enabled */
1307 if (test_bit(ICE_FLAG_RSS_ENA, pf->flags) &&
1308 vsi->type != ICE_VSI_CTRL) {
1309 ice_set_rss_vsi_ctx(ctxt, vsi);
1310 /* if updating VSI context, make sure to set valid_section:
1311 * to indicate which section of VSI context being updated
1313 if (!(vsi_flags & ICE_VSI_FLAG_INIT))
1314 ctxt->info.valid_sections |=
1315 cpu_to_le16(ICE_AQ_VSI_PROP_Q_OPT_VALID);
1318 ctxt->info.sw_id = vsi->port_info->sw_id;
1319 if (vsi->type == ICE_VSI_CHNL) {
1320 ice_chnl_vsi_setup_q_map(vsi, ctxt);
1322 ret = ice_vsi_setup_q_map(vsi, ctxt);
1326 if (!(vsi_flags & ICE_VSI_FLAG_INIT))
1327 /* means VSI being updated */
1328 /* must to indicate which section of VSI context are
1331 ctxt->info.valid_sections |=
1332 cpu_to_le16(ICE_AQ_VSI_PROP_RXQ_MAP_VALID);
1335 /* Allow control frames out of main VSI */
1336 if (vsi->type == ICE_VSI_PF) {
1337 ctxt->info.sec_flags |= ICE_AQ_VSI_SEC_FLAG_ALLOW_DEST_OVRD;
1338 ctxt->info.valid_sections |=
1339 cpu_to_le16(ICE_AQ_VSI_PROP_SECURITY_VALID);
1342 if (vsi_flags & ICE_VSI_FLAG_INIT) {
1343 ret = ice_add_vsi(hw, vsi->idx, ctxt, NULL);
1345 dev_err(dev, "Add VSI failed, err %d\n", ret);
1350 ret = ice_update_vsi(hw, vsi->idx, ctxt, NULL);
1352 dev_err(dev, "Update VSI failed, err %d\n", ret);
1358 /* keep context for update VSI operations */
1359 vsi->info = ctxt->info;
1361 /* record VSI number returned */
1362 vsi->vsi_num = ctxt->vsi_num;
1370 * ice_vsi_clear_rings - Deallocates the Tx and Rx rings for VSI
1371 * @vsi: the VSI having rings deallocated
1373 static void ice_vsi_clear_rings(struct ice_vsi *vsi)
1377 /* Avoid stale references by clearing map from vector to ring */
1378 if (vsi->q_vectors) {
1379 ice_for_each_q_vector(vsi, i) {
1380 struct ice_q_vector *q_vector = vsi->q_vectors[i];
1383 q_vector->tx.tx_ring = NULL;
1384 q_vector->rx.rx_ring = NULL;
1389 if (vsi->tx_rings) {
1390 ice_for_each_alloc_txq(vsi, i) {
1391 if (vsi->tx_rings[i]) {
1392 kfree_rcu(vsi->tx_rings[i], rcu);
1393 WRITE_ONCE(vsi->tx_rings[i], NULL);
1397 if (vsi->rx_rings) {
1398 ice_for_each_alloc_rxq(vsi, i) {
1399 if (vsi->rx_rings[i]) {
1400 kfree_rcu(vsi->rx_rings[i], rcu);
1401 WRITE_ONCE(vsi->rx_rings[i], NULL);
1408 * ice_vsi_alloc_rings - Allocates Tx and Rx rings for the VSI
1409 * @vsi: VSI which is having rings allocated
1411 static int ice_vsi_alloc_rings(struct ice_vsi *vsi)
1413 bool dvm_ena = ice_is_dvm_ena(&vsi->back->hw);
1414 struct ice_pf *pf = vsi->back;
1418 dev = ice_pf_to_dev(pf);
1419 /* Allocate Tx rings */
1420 ice_for_each_alloc_txq(vsi, i) {
1421 struct ice_tx_ring *ring;
1423 /* allocate with kzalloc(), free with kfree_rcu() */
1424 ring = kzalloc(sizeof(*ring), GFP_KERNEL);
1430 ring->reg_idx = vsi->txq_map[i];
1432 ring->tx_tstamps = &pf->ptp.port.tx;
1434 ring->count = vsi->num_tx_desc;
1435 ring->txq_teid = ICE_INVAL_TEID;
1437 ring->flags |= ICE_TX_FLAGS_RING_VLAN_L2TAG2;
1439 ring->flags |= ICE_TX_FLAGS_RING_VLAN_L2TAG1;
1440 WRITE_ONCE(vsi->tx_rings[i], ring);
1443 /* Allocate Rx rings */
1444 ice_for_each_alloc_rxq(vsi, i) {
1445 struct ice_rx_ring *ring;
1447 /* allocate with kzalloc(), free with kfree_rcu() */
1448 ring = kzalloc(sizeof(*ring), GFP_KERNEL);
1453 ring->reg_idx = vsi->rxq_map[i];
1455 ring->netdev = vsi->netdev;
1457 ring->count = vsi->num_rx_desc;
1458 ring->cached_phctime = pf->ptp.cached_phc_time;
1459 WRITE_ONCE(vsi->rx_rings[i], ring);
1465 ice_vsi_clear_rings(vsi);
1470 * ice_vsi_manage_rss_lut - disable/enable RSS
1471 * @vsi: the VSI being changed
1472 * @ena: boolean value indicating if this is an enable or disable request
1474 * In the event of disable request for RSS, this function will zero out RSS
1475 * LUT, while in the event of enable request for RSS, it will reconfigure RSS
1478 void ice_vsi_manage_rss_lut(struct ice_vsi *vsi, bool ena)
1482 lut = kzalloc(vsi->rss_table_size, GFP_KERNEL);
1487 if (vsi->rss_lut_user)
1488 memcpy(lut, vsi->rss_lut_user, vsi->rss_table_size);
1490 ice_fill_rss_lut(lut, vsi->rss_table_size,
1494 ice_set_rss_lut(vsi, lut, vsi->rss_table_size);
1499 * ice_vsi_cfg_crc_strip - Configure CRC stripping for a VSI
1500 * @vsi: VSI to be configured
1501 * @disable: set to true to have FCS / CRC in the frame data
1503 void ice_vsi_cfg_crc_strip(struct ice_vsi *vsi, bool disable)
1507 ice_for_each_rxq(vsi, i)
1509 vsi->rx_rings[i]->flags |= ICE_RX_FLAGS_CRC_STRIP_DIS;
1511 vsi->rx_rings[i]->flags &= ~ICE_RX_FLAGS_CRC_STRIP_DIS;
1515 * ice_vsi_cfg_rss_lut_key - Configure RSS params for a VSI
1516 * @vsi: VSI to be configured
1518 int ice_vsi_cfg_rss_lut_key(struct ice_vsi *vsi)
1520 struct ice_pf *pf = vsi->back;
1525 dev = ice_pf_to_dev(pf);
1526 if (vsi->type == ICE_VSI_PF && vsi->ch_rss_size &&
1527 (test_bit(ICE_FLAG_TC_MQPRIO, pf->flags))) {
1528 vsi->rss_size = min_t(u16, vsi->rss_size, vsi->ch_rss_size);
1530 vsi->rss_size = min_t(u16, vsi->rss_size, vsi->num_rxq);
1532 /* If orig_rss_size is valid and it is less than determined
1533 * main VSI's rss_size, update main VSI's rss_size to be
1534 * orig_rss_size so that when tc-qdisc is deleted, main VSI
1535 * RSS table gets programmed to be correct (whatever it was
1536 * to begin with (prior to setup-tc for ADQ config)
1538 if (vsi->orig_rss_size && vsi->rss_size < vsi->orig_rss_size &&
1539 vsi->orig_rss_size <= vsi->num_rxq) {
1540 vsi->rss_size = vsi->orig_rss_size;
1541 /* now orig_rss_size is used, reset it to zero */
1542 vsi->orig_rss_size = 0;
1546 lut = kzalloc(vsi->rss_table_size, GFP_KERNEL);
1550 if (vsi->rss_lut_user)
1551 memcpy(lut, vsi->rss_lut_user, vsi->rss_table_size);
1553 ice_fill_rss_lut(lut, vsi->rss_table_size, vsi->rss_size);
1555 err = ice_set_rss_lut(vsi, lut, vsi->rss_table_size);
1557 dev_err(dev, "set_rss_lut failed, error %d\n", err);
1558 goto ice_vsi_cfg_rss_exit;
1561 key = kzalloc(ICE_GET_SET_RSS_KEY_EXTEND_KEY_SIZE, GFP_KERNEL);
1564 goto ice_vsi_cfg_rss_exit;
1567 if (vsi->rss_hkey_user)
1568 memcpy(key, vsi->rss_hkey_user, ICE_GET_SET_RSS_KEY_EXTEND_KEY_SIZE);
1570 netdev_rss_key_fill((void *)key, ICE_GET_SET_RSS_KEY_EXTEND_KEY_SIZE);
1572 err = ice_set_rss_key(vsi, key);
1574 dev_err(dev, "set_rss_key failed, error %d\n", err);
1577 ice_vsi_cfg_rss_exit:
1583 * ice_vsi_set_vf_rss_flow_fld - Sets VF VSI RSS input set for different flows
1584 * @vsi: VSI to be configured
1586 * This function will only be called during the VF VSI setup. Upon successful
1587 * completion of package download, this function will configure default RSS
1588 * input sets for VF VSI.
1590 static void ice_vsi_set_vf_rss_flow_fld(struct ice_vsi *vsi)
1592 struct ice_pf *pf = vsi->back;
1596 dev = ice_pf_to_dev(pf);
1597 if (ice_is_safe_mode(pf)) {
1598 dev_dbg(dev, "Advanced RSS disabled. Package download failed, vsi num = %d\n",
1603 status = ice_add_avf_rss_cfg(&pf->hw, vsi->idx, ICE_DEFAULT_RSS_HENA);
1605 dev_dbg(dev, "ice_add_avf_rss_cfg failed for vsi = %d, error = %d\n",
1606 vsi->vsi_num, status);
1610 * ice_vsi_set_rss_flow_fld - Sets RSS input set for different flows
1611 * @vsi: VSI to be configured
1613 * This function will only be called after successful download package call
1614 * during initialization of PF. Since the downloaded package will erase the
1615 * RSS section, this function will configure RSS input sets for different
1616 * flow types. The last profile added has the highest priority, therefore 2
1617 * tuple profiles (i.e. IPv4 src/dst) are added before 4 tuple profiles
1618 * (i.e. IPv4 src/dst TCP src/dst port).
1620 static void ice_vsi_set_rss_flow_fld(struct ice_vsi *vsi)
1622 u16 vsi_handle = vsi->idx, vsi_num = vsi->vsi_num;
1623 struct ice_pf *pf = vsi->back;
1624 struct ice_hw *hw = &pf->hw;
1628 dev = ice_pf_to_dev(pf);
1629 if (ice_is_safe_mode(pf)) {
1630 dev_dbg(dev, "Advanced RSS disabled. Package download failed, vsi num = %d\n",
1634 /* configure RSS for IPv4 with input set IP src/dst */
1635 status = ice_add_rss_cfg(hw, vsi_handle, ICE_FLOW_HASH_IPV4,
1636 ICE_FLOW_SEG_HDR_IPV4);
1638 dev_dbg(dev, "ice_add_rss_cfg failed for ipv4 flow, vsi = %d, error = %d\n",
1641 /* configure RSS for IPv6 with input set IPv6 src/dst */
1642 status = ice_add_rss_cfg(hw, vsi_handle, ICE_FLOW_HASH_IPV6,
1643 ICE_FLOW_SEG_HDR_IPV6);
1645 dev_dbg(dev, "ice_add_rss_cfg failed for ipv6 flow, vsi = %d, error = %d\n",
1648 /* configure RSS for tcp4 with input set IP src/dst, TCP src/dst */
1649 status = ice_add_rss_cfg(hw, vsi_handle, ICE_HASH_TCP_IPV4,
1650 ICE_FLOW_SEG_HDR_TCP | ICE_FLOW_SEG_HDR_IPV4);
1652 dev_dbg(dev, "ice_add_rss_cfg failed for tcp4 flow, vsi = %d, error = %d\n",
1655 /* configure RSS for udp4 with input set IP src/dst, UDP src/dst */
1656 status = ice_add_rss_cfg(hw, vsi_handle, ICE_HASH_UDP_IPV4,
1657 ICE_FLOW_SEG_HDR_UDP | ICE_FLOW_SEG_HDR_IPV4);
1659 dev_dbg(dev, "ice_add_rss_cfg failed for udp4 flow, vsi = %d, error = %d\n",
1662 /* configure RSS for sctp4 with input set IP src/dst */
1663 status = ice_add_rss_cfg(hw, vsi_handle, ICE_FLOW_HASH_IPV4,
1664 ICE_FLOW_SEG_HDR_SCTP | ICE_FLOW_SEG_HDR_IPV4);
1666 dev_dbg(dev, "ice_add_rss_cfg failed for sctp4 flow, vsi = %d, error = %d\n",
1669 /* configure RSS for tcp6 with input set IPv6 src/dst, TCP src/dst */
1670 status = ice_add_rss_cfg(hw, vsi_handle, ICE_HASH_TCP_IPV6,
1671 ICE_FLOW_SEG_HDR_TCP | ICE_FLOW_SEG_HDR_IPV6);
1673 dev_dbg(dev, "ice_add_rss_cfg failed for tcp6 flow, vsi = %d, error = %d\n",
1676 /* configure RSS for udp6 with input set IPv6 src/dst, UDP src/dst */
1677 status = ice_add_rss_cfg(hw, vsi_handle, ICE_HASH_UDP_IPV6,
1678 ICE_FLOW_SEG_HDR_UDP | ICE_FLOW_SEG_HDR_IPV6);
1680 dev_dbg(dev, "ice_add_rss_cfg failed for udp6 flow, vsi = %d, error = %d\n",
1683 /* configure RSS for sctp6 with input set IPv6 src/dst */
1684 status = ice_add_rss_cfg(hw, vsi_handle, ICE_FLOW_HASH_IPV6,
1685 ICE_FLOW_SEG_HDR_SCTP | ICE_FLOW_SEG_HDR_IPV6);
1687 dev_dbg(dev, "ice_add_rss_cfg failed for sctp6 flow, vsi = %d, error = %d\n",
1690 status = ice_add_rss_cfg(hw, vsi_handle, ICE_FLOW_HASH_ESP_SPI,
1691 ICE_FLOW_SEG_HDR_ESP);
1693 dev_dbg(dev, "ice_add_rss_cfg failed for esp/spi flow, vsi = %d, error = %d\n",
1698 * ice_vsi_cfg_frame_size - setup max frame size and Rx buffer length
1701 static void ice_vsi_cfg_frame_size(struct ice_vsi *vsi)
1703 if (!vsi->netdev || test_bit(ICE_FLAG_LEGACY_RX, vsi->back->flags)) {
1704 vsi->max_frame = ICE_MAX_FRAME_LEGACY_RX;
1705 vsi->rx_buf_len = ICE_RXBUF_1664;
1706 #if (PAGE_SIZE < 8192)
1707 } else if (!ICE_2K_TOO_SMALL_WITH_PADDING &&
1708 (vsi->netdev->mtu <= ETH_DATA_LEN)) {
1709 vsi->max_frame = ICE_RXBUF_1536 - NET_IP_ALIGN;
1710 vsi->rx_buf_len = ICE_RXBUF_1536 - NET_IP_ALIGN;
1713 vsi->max_frame = ICE_AQ_SET_MAC_FRAME_SIZE_MAX;
1714 vsi->rx_buf_len = ICE_RXBUF_3072;
1719 * ice_pf_state_is_nominal - checks the PF for nominal state
1720 * @pf: pointer to PF to check
1722 * Check the PF's state for a collection of bits that would indicate
1723 * the PF is in a state that would inhibit normal operation for
1724 * driver functionality.
1726 * Returns true if PF is in a nominal state, false otherwise
1728 bool ice_pf_state_is_nominal(struct ice_pf *pf)
1730 DECLARE_BITMAP(check_bits, ICE_STATE_NBITS) = { 0 };
1735 bitmap_set(check_bits, 0, ICE_STATE_NOMINAL_CHECK_BITS);
1736 if (bitmap_intersects(pf->state, check_bits, ICE_STATE_NBITS))
1743 * ice_update_eth_stats - Update VSI-specific ethernet statistics counters
1744 * @vsi: the VSI to be updated
1746 void ice_update_eth_stats(struct ice_vsi *vsi)
1748 struct ice_eth_stats *prev_es, *cur_es;
1749 struct ice_hw *hw = &vsi->back->hw;
1750 struct ice_pf *pf = vsi->back;
1751 u16 vsi_num = vsi->vsi_num; /* HW absolute index of a VSI */
1753 prev_es = &vsi->eth_stats_prev;
1754 cur_es = &vsi->eth_stats;
1756 if (ice_is_reset_in_progress(pf->state))
1757 vsi->stat_offsets_loaded = false;
1759 ice_stat_update40(hw, GLV_GORCL(vsi_num), vsi->stat_offsets_loaded,
1760 &prev_es->rx_bytes, &cur_es->rx_bytes);
1762 ice_stat_update40(hw, GLV_UPRCL(vsi_num), vsi->stat_offsets_loaded,
1763 &prev_es->rx_unicast, &cur_es->rx_unicast);
1765 ice_stat_update40(hw, GLV_MPRCL(vsi_num), vsi->stat_offsets_loaded,
1766 &prev_es->rx_multicast, &cur_es->rx_multicast);
1768 ice_stat_update40(hw, GLV_BPRCL(vsi_num), vsi->stat_offsets_loaded,
1769 &prev_es->rx_broadcast, &cur_es->rx_broadcast);
1771 ice_stat_update32(hw, GLV_RDPC(vsi_num), vsi->stat_offsets_loaded,
1772 &prev_es->rx_discards, &cur_es->rx_discards);
1774 ice_stat_update40(hw, GLV_GOTCL(vsi_num), vsi->stat_offsets_loaded,
1775 &prev_es->tx_bytes, &cur_es->tx_bytes);
1777 ice_stat_update40(hw, GLV_UPTCL(vsi_num), vsi->stat_offsets_loaded,
1778 &prev_es->tx_unicast, &cur_es->tx_unicast);
1780 ice_stat_update40(hw, GLV_MPTCL(vsi_num), vsi->stat_offsets_loaded,
1781 &prev_es->tx_multicast, &cur_es->tx_multicast);
1783 ice_stat_update40(hw, GLV_BPTCL(vsi_num), vsi->stat_offsets_loaded,
1784 &prev_es->tx_broadcast, &cur_es->tx_broadcast);
1786 ice_stat_update32(hw, GLV_TEPC(vsi_num), vsi->stat_offsets_loaded,
1787 &prev_es->tx_errors, &cur_es->tx_errors);
1789 vsi->stat_offsets_loaded = true;
1793 * ice_write_qrxflxp_cntxt - write/configure QRXFLXP_CNTXT register
1795 * @pf_q: index of the Rx queue in the PF's queue space
1796 * @rxdid: flexible descriptor RXDID
1797 * @prio: priority for the RXDID for this queue
1798 * @ena_ts: true to enable timestamp and false to disable timestamp
1801 ice_write_qrxflxp_cntxt(struct ice_hw *hw, u16 pf_q, u32 rxdid, u32 prio,
1804 int regval = rd32(hw, QRXFLXP_CNTXT(pf_q));
1806 /* clear any previous values */
1807 regval &= ~(QRXFLXP_CNTXT_RXDID_IDX_M |
1808 QRXFLXP_CNTXT_RXDID_PRIO_M |
1809 QRXFLXP_CNTXT_TS_M);
1811 regval |= (rxdid << QRXFLXP_CNTXT_RXDID_IDX_S) &
1812 QRXFLXP_CNTXT_RXDID_IDX_M;
1814 regval |= (prio << QRXFLXP_CNTXT_RXDID_PRIO_S) &
1815 QRXFLXP_CNTXT_RXDID_PRIO_M;
1818 /* Enable TimeSync on this queue */
1819 regval |= QRXFLXP_CNTXT_TS_M;
1821 wr32(hw, QRXFLXP_CNTXT(pf_q), regval);
1824 int ice_vsi_cfg_single_rxq(struct ice_vsi *vsi, u16 q_idx)
1826 if (q_idx >= vsi->num_rxq)
1829 return ice_vsi_cfg_rxq(vsi->rx_rings[q_idx]);
1832 int ice_vsi_cfg_single_txq(struct ice_vsi *vsi, struct ice_tx_ring **tx_rings, u16 q_idx)
1834 DEFINE_FLEX(struct ice_aqc_add_tx_qgrp, qg_buf, txqs, 1);
1836 if (q_idx >= vsi->alloc_txq || !tx_rings || !tx_rings[q_idx])
1839 qg_buf->num_txqs = 1;
1841 return ice_vsi_cfg_txq(vsi, tx_rings[q_idx], qg_buf);
1845 * ice_vsi_cfg_rxqs - Configure the VSI for Rx
1846 * @vsi: the VSI being configured
1848 * Return 0 on success and a negative value on error
1849 * Configure the Rx VSI for operation.
1851 int ice_vsi_cfg_rxqs(struct ice_vsi *vsi)
1855 if (vsi->type == ICE_VSI_VF)
1858 ice_vsi_cfg_frame_size(vsi);
1860 /* set up individual rings */
1861 ice_for_each_rxq(vsi, i) {
1862 int err = ice_vsi_cfg_rxq(vsi->rx_rings[i]);
1872 * ice_vsi_cfg_txqs - Configure the VSI for Tx
1873 * @vsi: the VSI being configured
1874 * @rings: Tx ring array to be configured
1875 * @count: number of Tx ring array elements
1877 * Return 0 on success and a negative value on error
1878 * Configure the Tx VSI for operation.
1881 ice_vsi_cfg_txqs(struct ice_vsi *vsi, struct ice_tx_ring **rings, u16 count)
1883 DEFINE_FLEX(struct ice_aqc_add_tx_qgrp, qg_buf, txqs, 1);
1887 qg_buf->num_txqs = 1;
1889 for (q_idx = 0; q_idx < count; q_idx++) {
1890 err = ice_vsi_cfg_txq(vsi, rings[q_idx], qg_buf);
1899 * ice_vsi_cfg_lan_txqs - Configure the VSI for Tx
1900 * @vsi: the VSI being configured
1902 * Return 0 on success and a negative value on error
1903 * Configure the Tx VSI for operation.
1905 int ice_vsi_cfg_lan_txqs(struct ice_vsi *vsi)
1907 return ice_vsi_cfg_txqs(vsi, vsi->tx_rings, vsi->num_txq);
1911 * ice_vsi_cfg_xdp_txqs - Configure Tx queues dedicated for XDP in given VSI
1912 * @vsi: the VSI being configured
1914 * Return 0 on success and a negative value on error
1915 * Configure the Tx queues dedicated for XDP in given VSI for operation.
1917 int ice_vsi_cfg_xdp_txqs(struct ice_vsi *vsi)
1922 ret = ice_vsi_cfg_txqs(vsi, vsi->xdp_rings, vsi->num_xdp_txq);
1926 ice_for_each_rxq(vsi, i)
1927 ice_tx_xsk_pool(vsi, i);
1933 * ice_intrl_usec_to_reg - convert interrupt rate limit to register value
1934 * @intrl: interrupt rate limit in usecs
1935 * @gran: interrupt rate limit granularity in usecs
1937 * This function converts a decimal interrupt rate limit in usecs to the format
1938 * expected by firmware.
1940 static u32 ice_intrl_usec_to_reg(u8 intrl, u8 gran)
1942 u32 val = intrl / gran;
1945 return val | GLINT_RATE_INTRL_ENA_M;
1950 * ice_write_intrl - write throttle rate limit to interrupt specific register
1951 * @q_vector: pointer to interrupt specific structure
1952 * @intrl: throttle rate limit in microseconds to write
1954 void ice_write_intrl(struct ice_q_vector *q_vector, u8 intrl)
1956 struct ice_hw *hw = &q_vector->vsi->back->hw;
1958 wr32(hw, GLINT_RATE(q_vector->reg_idx),
1959 ice_intrl_usec_to_reg(intrl, ICE_INTRL_GRAN_ABOVE_25));
1962 static struct ice_q_vector *ice_pull_qvec_from_rc(struct ice_ring_container *rc)
1965 case ICE_RX_CONTAINER:
1967 return rc->rx_ring->q_vector;
1969 case ICE_TX_CONTAINER:
1971 return rc->tx_ring->q_vector;
1981 * __ice_write_itr - write throttle rate to register
1982 * @q_vector: pointer to interrupt data structure
1983 * @rc: pointer to ring container
1984 * @itr: throttle rate in microseconds to write
1986 static void __ice_write_itr(struct ice_q_vector *q_vector,
1987 struct ice_ring_container *rc, u16 itr)
1989 struct ice_hw *hw = &q_vector->vsi->back->hw;
1991 wr32(hw, GLINT_ITR(rc->itr_idx, q_vector->reg_idx),
1992 ITR_REG_ALIGN(itr) >> ICE_ITR_GRAN_S);
1996 * ice_write_itr - write throttle rate to queue specific register
1997 * @rc: pointer to ring container
1998 * @itr: throttle rate in microseconds to write
2000 void ice_write_itr(struct ice_ring_container *rc, u16 itr)
2002 struct ice_q_vector *q_vector;
2004 q_vector = ice_pull_qvec_from_rc(rc);
2008 __ice_write_itr(q_vector, rc, itr);
2012 * ice_set_q_vector_intrl - set up interrupt rate limiting
2013 * @q_vector: the vector to be configured
2015 * Interrupt rate limiting is local to the vector, not per-queue so we must
2016 * detect if either ring container has dynamic moderation enabled to decide
2017 * what to set the interrupt rate limit to via INTRL settings. In the case that
2018 * dynamic moderation is disabled on both, write the value with the cached
2019 * setting to make sure INTRL register matches the user visible value.
2021 void ice_set_q_vector_intrl(struct ice_q_vector *q_vector)
2023 if (ITR_IS_DYNAMIC(&q_vector->tx) || ITR_IS_DYNAMIC(&q_vector->rx)) {
2024 /* in the case of dynamic enabled, cap each vector to no more
2025 * than (4 us) 250,000 ints/sec, which allows low latency
2026 * but still less than 500,000 interrupts per second, which
2027 * reduces CPU a bit in the case of the lowest latency
2028 * setting. The 4 here is a value in microseconds.
2030 ice_write_intrl(q_vector, 4);
2032 ice_write_intrl(q_vector, q_vector->intrl);
2037 * ice_vsi_cfg_msix - MSIX mode Interrupt Config in the HW
2038 * @vsi: the VSI being configured
2040 * This configures MSIX mode interrupts for the PF VSI, and should not be used
2043 void ice_vsi_cfg_msix(struct ice_vsi *vsi)
2045 struct ice_pf *pf = vsi->back;
2046 struct ice_hw *hw = &pf->hw;
2047 u16 txq = 0, rxq = 0;
2050 ice_for_each_q_vector(vsi, i) {
2051 struct ice_q_vector *q_vector = vsi->q_vectors[i];
2052 u16 reg_idx = q_vector->reg_idx;
2054 ice_cfg_itr(hw, q_vector);
2056 /* Both Transmit Queue Interrupt Cause Control register
2057 * and Receive Queue Interrupt Cause control register
2058 * expects MSIX_INDX field to be the vector index
2059 * within the function space and not the absolute
2060 * vector index across PF or across device.
2061 * For SR-IOV VF VSIs queue vector index always starts
2062 * with 1 since first vector index(0) is used for OICR
2063 * in VF space. Since VMDq and other PF VSIs are within
2064 * the PF function space, use the vector index that is
2065 * tracked for this PF.
2067 for (q = 0; q < q_vector->num_ring_tx; q++) {
2068 ice_cfg_txq_interrupt(vsi, txq, reg_idx,
2069 q_vector->tx.itr_idx);
2073 for (q = 0; q < q_vector->num_ring_rx; q++) {
2074 ice_cfg_rxq_interrupt(vsi, rxq, reg_idx,
2075 q_vector->rx.itr_idx);
2082 * ice_vsi_start_all_rx_rings - start/enable all of a VSI's Rx rings
2083 * @vsi: the VSI whose rings are to be enabled
2085 * Returns 0 on success and a negative value on error
2087 int ice_vsi_start_all_rx_rings(struct ice_vsi *vsi)
2089 return ice_vsi_ctrl_all_rx_rings(vsi, true);
2093 * ice_vsi_stop_all_rx_rings - stop/disable all of a VSI's Rx rings
2094 * @vsi: the VSI whose rings are to be disabled
2096 * Returns 0 on success and a negative value on error
2098 int ice_vsi_stop_all_rx_rings(struct ice_vsi *vsi)
2100 return ice_vsi_ctrl_all_rx_rings(vsi, false);
2104 * ice_vsi_stop_tx_rings - Disable Tx rings
2105 * @vsi: the VSI being configured
2106 * @rst_src: reset source
2107 * @rel_vmvf_num: Relative ID of VF/VM
2108 * @rings: Tx ring array to be stopped
2109 * @count: number of Tx ring array elements
2112 ice_vsi_stop_tx_rings(struct ice_vsi *vsi, enum ice_disq_rst_src rst_src,
2113 u16 rel_vmvf_num, struct ice_tx_ring **rings, u16 count)
2117 if (vsi->num_txq > ICE_LAN_TXQ_MAX_QDIS)
2120 for (q_idx = 0; q_idx < count; q_idx++) {
2121 struct ice_txq_meta txq_meta = { };
2124 if (!rings || !rings[q_idx])
2127 ice_fill_txq_meta(vsi, rings[q_idx], &txq_meta);
2128 status = ice_vsi_stop_tx_ring(vsi, rst_src, rel_vmvf_num,
2129 rings[q_idx], &txq_meta);
2139 * ice_vsi_stop_lan_tx_rings - Disable LAN Tx rings
2140 * @vsi: the VSI being configured
2141 * @rst_src: reset source
2142 * @rel_vmvf_num: Relative ID of VF/VM
2145 ice_vsi_stop_lan_tx_rings(struct ice_vsi *vsi, enum ice_disq_rst_src rst_src,
2148 return ice_vsi_stop_tx_rings(vsi, rst_src, rel_vmvf_num, vsi->tx_rings, vsi->num_txq);
2152 * ice_vsi_stop_xdp_tx_rings - Disable XDP Tx rings
2153 * @vsi: the VSI being configured
2155 int ice_vsi_stop_xdp_tx_rings(struct ice_vsi *vsi)
2157 return ice_vsi_stop_tx_rings(vsi, ICE_NO_RESET, 0, vsi->xdp_rings, vsi->num_xdp_txq);
2161 * ice_vsi_is_rx_queue_active
2162 * @vsi: the VSI being configured
2164 * Return true if at least one queue is active.
2166 bool ice_vsi_is_rx_queue_active(struct ice_vsi *vsi)
2168 struct ice_pf *pf = vsi->back;
2169 struct ice_hw *hw = &pf->hw;
2172 ice_for_each_rxq(vsi, i) {
2176 pf_q = vsi->rxq_map[i];
2177 rx_reg = rd32(hw, QRX_CTRL(pf_q));
2178 if (rx_reg & QRX_CTRL_QENA_STAT_M)
2185 static void ice_vsi_set_tc_cfg(struct ice_vsi *vsi)
2187 if (!test_bit(ICE_FLAG_DCB_ENA, vsi->back->flags)) {
2188 vsi->tc_cfg.ena_tc = ICE_DFLT_TRAFFIC_CLASS;
2189 vsi->tc_cfg.numtc = 1;
2193 /* set VSI TC information based on DCB config */
2194 ice_vsi_set_dcb_tc_cfg(vsi);
2198 * ice_cfg_sw_lldp - Config switch rules for LLDP packet handling
2199 * @vsi: the VSI being configured
2200 * @tx: bool to determine Tx or Rx rule
2201 * @create: bool to determine create or remove Rule
2203 void ice_cfg_sw_lldp(struct ice_vsi *vsi, bool tx, bool create)
2205 int (*eth_fltr)(struct ice_vsi *v, u16 type, u16 flag,
2206 enum ice_sw_fwd_act_type act);
2207 struct ice_pf *pf = vsi->back;
2211 dev = ice_pf_to_dev(pf);
2212 eth_fltr = create ? ice_fltr_add_eth : ice_fltr_remove_eth;
2215 status = eth_fltr(vsi, ETH_P_LLDP, ICE_FLTR_TX,
2218 if (ice_fw_supports_lldp_fltr_ctrl(&pf->hw)) {
2219 status = ice_lldp_fltr_add_remove(&pf->hw, vsi->vsi_num,
2222 status = eth_fltr(vsi, ETH_P_LLDP, ICE_FLTR_RX,
2228 dev_dbg(dev, "Fail %s %s LLDP rule on VSI %i error: %d\n",
2229 create ? "adding" : "removing", tx ? "TX" : "RX",
2230 vsi->vsi_num, status);
2234 * ice_set_agg_vsi - sets up scheduler aggregator node and move VSI into it
2235 * @vsi: pointer to the VSI
2237 * This function will allocate new scheduler aggregator now if needed and will
2238 * move specified VSI into it.
2240 static void ice_set_agg_vsi(struct ice_vsi *vsi)
2242 struct device *dev = ice_pf_to_dev(vsi->back);
2243 struct ice_agg_node *agg_node_iter = NULL;
2244 u32 agg_id = ICE_INVALID_AGG_NODE_ID;
2245 struct ice_agg_node *agg_node = NULL;
2246 int node_offset, max_agg_nodes = 0;
2247 struct ice_port_info *port_info;
2248 struct ice_pf *pf = vsi->back;
2249 u32 agg_node_id_start = 0;
2252 /* create (as needed) scheduler aggregator node and move VSI into
2253 * corresponding aggregator node
2254 * - PF aggregator node to contains VSIs of type _PF and _CTRL
2255 * - VF aggregator nodes will contain VF VSI
2257 port_info = pf->hw.port_info;
2261 switch (vsi->type) {
2266 case ICE_VSI_SWITCHDEV_CTRL:
2267 max_agg_nodes = ICE_MAX_PF_AGG_NODES;
2268 agg_node_id_start = ICE_PF_AGG_NODE_ID_START;
2269 agg_node_iter = &pf->pf_agg_node[0];
2272 /* user can create 'n' VFs on a given PF, but since max children
2273 * per aggregator node can be only 64. Following code handles
2274 * aggregator(s) for VF VSIs, either selects a agg_node which
2275 * was already created provided num_vsis < 64, otherwise
2276 * select next available node, which will be created
2278 max_agg_nodes = ICE_MAX_VF_AGG_NODES;
2279 agg_node_id_start = ICE_VF_AGG_NODE_ID_START;
2280 agg_node_iter = &pf->vf_agg_node[0];
2283 /* other VSI type, handle later if needed */
2284 dev_dbg(dev, "unexpected VSI type %s\n",
2285 ice_vsi_type_str(vsi->type));
2289 /* find the appropriate aggregator node */
2290 for (node_offset = 0; node_offset < max_agg_nodes; node_offset++) {
2291 /* see if we can find space in previously created
2292 * node if num_vsis < 64, otherwise skip
2294 if (agg_node_iter->num_vsis &&
2295 agg_node_iter->num_vsis == ICE_MAX_VSIS_IN_AGG_NODE) {
2300 if (agg_node_iter->valid &&
2301 agg_node_iter->agg_id != ICE_INVALID_AGG_NODE_ID) {
2302 agg_id = agg_node_iter->agg_id;
2303 agg_node = agg_node_iter;
2307 /* find unclaimed agg_id */
2308 if (agg_node_iter->agg_id == ICE_INVALID_AGG_NODE_ID) {
2309 agg_id = node_offset + agg_node_id_start;
2310 agg_node = agg_node_iter;
2313 /* move to next agg_node */
2320 /* if selected aggregator node was not created, create it */
2321 if (!agg_node->valid) {
2322 status = ice_cfg_agg(port_info, agg_id, ICE_AGG_TYPE_AGG,
2323 (u8)vsi->tc_cfg.ena_tc);
2325 dev_err(dev, "unable to create aggregator node with agg_id %u\n",
2329 /* aggregator node is created, store the needed info */
2330 agg_node->valid = true;
2331 agg_node->agg_id = agg_id;
2334 /* move VSI to corresponding aggregator node */
2335 status = ice_move_vsi_to_agg(port_info, agg_id, vsi->idx,
2336 (u8)vsi->tc_cfg.ena_tc);
2338 dev_err(dev, "unable to move VSI idx %u into aggregator %u node",
2343 /* keep active children count for aggregator node */
2344 agg_node->num_vsis++;
2346 /* cache the 'agg_id' in VSI, so that after reset - VSI will be moved
2347 * to aggregator node
2349 vsi->agg_node = agg_node;
2350 dev_dbg(dev, "successfully moved VSI idx %u tc_bitmap 0x%x) into aggregator node %d which has num_vsis %u\n",
2351 vsi->idx, vsi->tc_cfg.ena_tc, vsi->agg_node->agg_id,
2352 vsi->agg_node->num_vsis);
2355 static int ice_vsi_cfg_tc_lan(struct ice_pf *pf, struct ice_vsi *vsi)
2357 u16 max_txqs[ICE_MAX_TRAFFIC_CLASS] = { 0 };
2358 struct device *dev = ice_pf_to_dev(pf);
2361 /* configure VSI nodes based on number of queues and TC's */
2362 ice_for_each_traffic_class(i) {
2363 if (!(vsi->tc_cfg.ena_tc & BIT(i)))
2366 if (vsi->type == ICE_VSI_CHNL) {
2367 if (!vsi->alloc_txq && vsi->num_txq)
2368 max_txqs[i] = vsi->num_txq;
2370 max_txqs[i] = pf->num_lan_tx;
2372 max_txqs[i] = vsi->alloc_txq;
2376 dev_dbg(dev, "vsi->tc_cfg.ena_tc = %d\n", vsi->tc_cfg.ena_tc);
2377 ret = ice_cfg_vsi_lan(vsi->port_info, vsi->idx, vsi->tc_cfg.ena_tc,
2380 dev_err(dev, "VSI %d failed lan queue config, error %d\n",
2389 * ice_vsi_cfg_def - configure default VSI based on the type
2390 * @vsi: pointer to VSI
2391 * @params: the parameters to configure this VSI with
2394 ice_vsi_cfg_def(struct ice_vsi *vsi, struct ice_vsi_cfg_params *params)
2396 struct device *dev = ice_pf_to_dev(vsi->back);
2397 struct ice_pf *pf = vsi->back;
2400 vsi->vsw = pf->first_sw;
2402 ret = ice_vsi_alloc_def(vsi, params->ch);
2406 /* allocate memory for Tx/Rx ring stat pointers */
2407 ret = ice_vsi_alloc_stat_arrays(vsi);
2409 goto unroll_vsi_alloc;
2411 ice_alloc_fd_res(vsi);
2413 ret = ice_vsi_get_qs(vsi);
2415 dev_err(dev, "Failed to allocate queues. vsi->idx = %d\n",
2417 goto unroll_vsi_alloc_stat;
2420 /* set RSS capabilities */
2421 ice_vsi_set_rss_params(vsi);
2423 /* set TC configuration */
2424 ice_vsi_set_tc_cfg(vsi);
2426 /* create the VSI */
2427 ret = ice_vsi_init(vsi, params->flags);
2431 ice_vsi_init_vlan_ops(vsi);
2433 switch (vsi->type) {
2435 case ICE_VSI_SWITCHDEV_CTRL:
2437 ret = ice_vsi_alloc_q_vectors(vsi);
2439 goto unroll_vsi_init;
2441 ret = ice_vsi_alloc_rings(vsi);
2443 goto unroll_vector_base;
2445 ret = ice_vsi_alloc_ring_stats(vsi);
2447 goto unroll_vector_base;
2449 ice_vsi_map_rings_to_vectors(vsi);
2450 vsi->stat_offsets_loaded = false;
2452 if (ice_is_xdp_ena_vsi(vsi)) {
2453 ret = ice_vsi_determine_xdp_res(vsi);
2455 goto unroll_vector_base;
2456 ret = ice_prepare_xdp_rings(vsi, vsi->xdp_prog);
2458 goto unroll_vector_base;
2461 /* ICE_VSI_CTRL does not need RSS so skip RSS processing */
2462 if (vsi->type != ICE_VSI_CTRL)
2463 /* Do not exit if configuring RSS had an issue, at
2464 * least receive traffic on first queue. Hence no
2465 * need to capture return value
2467 if (test_bit(ICE_FLAG_RSS_ENA, pf->flags)) {
2468 ice_vsi_cfg_rss_lut_key(vsi);
2469 ice_vsi_set_rss_flow_fld(vsi);
2474 if (test_bit(ICE_FLAG_RSS_ENA, pf->flags)) {
2475 ice_vsi_cfg_rss_lut_key(vsi);
2476 ice_vsi_set_rss_flow_fld(vsi);
2480 /* VF driver will take care of creating netdev for this type and
2481 * map queues to vectors through Virtchnl, PF driver only
2482 * creates a VSI and corresponding structures for bookkeeping
2485 ret = ice_vsi_alloc_q_vectors(vsi);
2487 goto unroll_vsi_init;
2489 ret = ice_vsi_alloc_rings(vsi);
2491 goto unroll_alloc_q_vector;
2493 ret = ice_vsi_alloc_ring_stats(vsi);
2495 goto unroll_vector_base;
2497 vsi->stat_offsets_loaded = false;
2499 /* Do not exit if configuring RSS had an issue, at least
2500 * receive traffic on first queue. Hence no need to capture
2503 if (test_bit(ICE_FLAG_RSS_ENA, pf->flags)) {
2504 ice_vsi_cfg_rss_lut_key(vsi);
2505 ice_vsi_set_vf_rss_flow_fld(vsi);
2509 ret = ice_vsi_alloc_rings(vsi);
2511 goto unroll_vsi_init;
2513 ret = ice_vsi_alloc_ring_stats(vsi);
2515 goto unroll_vector_base;
2519 /* clean up the resources and exit */
2521 goto unroll_vsi_init;
2527 /* reclaim SW interrupts back to the common pool */
2528 unroll_alloc_q_vector:
2529 ice_vsi_free_q_vectors(vsi);
2531 ice_vsi_delete_from_hw(vsi);
2533 ice_vsi_put_qs(vsi);
2534 unroll_vsi_alloc_stat:
2535 ice_vsi_free_stats(vsi);
2537 ice_vsi_free_arrays(vsi);
2542 * ice_vsi_cfg - configure a previously allocated VSI
2543 * @vsi: pointer to VSI
2544 * @params: parameters used to configure this VSI
2546 int ice_vsi_cfg(struct ice_vsi *vsi, struct ice_vsi_cfg_params *params)
2548 struct ice_pf *pf = vsi->back;
2551 if (WARN_ON(params->type == ICE_VSI_VF && !params->vf))
2554 vsi->type = params->type;
2555 vsi->port_info = params->pi;
2557 /* For VSIs which don't have a connected VF, this will be NULL */
2558 vsi->vf = params->vf;
2560 ret = ice_vsi_cfg_def(vsi, params);
2564 ret = ice_vsi_cfg_tc_lan(vsi->back, vsi);
2568 if (vsi->type == ICE_VSI_CTRL) {
2570 WARN_ON(vsi->vf->ctrl_vsi_idx != ICE_NO_VSI);
2571 vsi->vf->ctrl_vsi_idx = vsi->idx;
2573 WARN_ON(pf->ctrl_vsi_idx != ICE_NO_VSI);
2574 pf->ctrl_vsi_idx = vsi->idx;
2582 * ice_vsi_decfg - remove all VSI configuration
2583 * @vsi: pointer to VSI
2585 void ice_vsi_decfg(struct ice_vsi *vsi)
2587 struct ice_pf *pf = vsi->back;
2590 /* The Rx rule will only exist to remove if the LLDP FW
2591 * engine is currently stopped
2593 if (!ice_is_safe_mode(pf) && vsi->type == ICE_VSI_PF &&
2594 !test_bit(ICE_FLAG_FW_LLDP_AGENT, pf->flags))
2595 ice_cfg_sw_lldp(vsi, false, false);
2597 ice_rm_vsi_lan_cfg(vsi->port_info, vsi->idx);
2598 err = ice_rm_vsi_rdma_cfg(vsi->port_info, vsi->idx);
2600 dev_err(ice_pf_to_dev(pf), "Failed to remove RDMA scheduler config for VSI %u, err %d\n",
2603 if (ice_is_xdp_ena_vsi(vsi))
2604 /* return value check can be skipped here, it always returns
2605 * 0 if reset is in progress
2607 ice_destroy_xdp_rings(vsi);
2609 ice_vsi_clear_rings(vsi);
2610 ice_vsi_free_q_vectors(vsi);
2611 ice_vsi_put_qs(vsi);
2612 ice_vsi_free_arrays(vsi);
2614 /* SR-IOV determines needed MSIX resources all at once instead of per
2615 * VSI since when VFs are spawned we know how many VFs there are and how
2616 * many interrupts each VF needs. SR-IOV MSIX resources are also
2617 * cleared in the same manner.
2620 if (vsi->type == ICE_VSI_VF &&
2621 vsi->agg_node && vsi->agg_node->valid)
2622 vsi->agg_node->num_vsis--;
2623 if (vsi->agg_node) {
2624 vsi->agg_node->valid = false;
2625 vsi->agg_node->agg_id = 0;
2630 * ice_vsi_setup - Set up a VSI by a given type
2631 * @pf: board private structure
2632 * @params: parameters to use when creating the VSI
2634 * This allocates the sw VSI structure and its queue resources.
2636 * Returns pointer to the successfully allocated and configured VSI sw struct on
2637 * success, NULL on failure.
2640 ice_vsi_setup(struct ice_pf *pf, struct ice_vsi_cfg_params *params)
2642 struct device *dev = ice_pf_to_dev(pf);
2643 struct ice_vsi *vsi;
2646 /* ice_vsi_setup can only initialize a new VSI, and we must have
2647 * a port_info structure for it.
2649 if (WARN_ON(!(params->flags & ICE_VSI_FLAG_INIT)) ||
2650 WARN_ON(!params->pi))
2653 vsi = ice_vsi_alloc(pf);
2655 dev_err(dev, "could not allocate VSI\n");
2659 ret = ice_vsi_cfg(vsi, params);
2663 /* Add switch rule to drop all Tx Flow Control Frames, of look up
2664 * type ETHERTYPE from VSIs, and restrict malicious VF from sending
2665 * out PAUSE or PFC frames. If enabled, FW can still send FC frames.
2666 * The rule is added once for PF VSI in order to create appropriate
2667 * recipe, since VSI/VSI list is ignored with drop action...
2668 * Also add rules to handle LLDP Tx packets. Tx LLDP packets need to
2669 * be dropped so that VFs cannot send LLDP packets to reconfig DCB
2670 * settings in the HW.
2672 if (!ice_is_safe_mode(pf) && vsi->type == ICE_VSI_PF) {
2673 ice_fltr_add_eth(vsi, ETH_P_PAUSE, ICE_FLTR_TX,
2675 ice_cfg_sw_lldp(vsi, true, true);
2679 ice_set_agg_vsi(vsi);
2690 * ice_vsi_release_msix - Clear the queue to Interrupt mapping in HW
2691 * @vsi: the VSI being cleaned up
2693 static void ice_vsi_release_msix(struct ice_vsi *vsi)
2695 struct ice_pf *pf = vsi->back;
2696 struct ice_hw *hw = &pf->hw;
2701 ice_for_each_q_vector(vsi, i) {
2702 struct ice_q_vector *q_vector = vsi->q_vectors[i];
2704 ice_write_intrl(q_vector, 0);
2705 for (q = 0; q < q_vector->num_ring_tx; q++) {
2706 ice_write_itr(&q_vector->tx, 0);
2707 wr32(hw, QINT_TQCTL(vsi->txq_map[txq]), 0);
2708 if (ice_is_xdp_ena_vsi(vsi)) {
2709 u32 xdp_txq = txq + vsi->num_xdp_txq;
2711 wr32(hw, QINT_TQCTL(vsi->txq_map[xdp_txq]), 0);
2716 for (q = 0; q < q_vector->num_ring_rx; q++) {
2717 ice_write_itr(&q_vector->rx, 0);
2718 wr32(hw, QINT_RQCTL(vsi->rxq_map[rxq]), 0);
2727 * ice_vsi_free_irq - Free the IRQ association with the OS
2728 * @vsi: the VSI being configured
2730 void ice_vsi_free_irq(struct ice_vsi *vsi)
2732 struct ice_pf *pf = vsi->back;
2735 if (!vsi->q_vectors || !vsi->irqs_ready)
2738 ice_vsi_release_msix(vsi);
2739 if (vsi->type == ICE_VSI_VF)
2742 vsi->irqs_ready = false;
2743 ice_free_cpu_rx_rmap(vsi);
2745 ice_for_each_q_vector(vsi, i) {
2748 irq_num = vsi->q_vectors[i]->irq.virq;
2750 /* free only the irqs that were actually requested */
2751 if (!vsi->q_vectors[i] ||
2752 !(vsi->q_vectors[i]->num_ring_tx ||
2753 vsi->q_vectors[i]->num_ring_rx))
2756 /* clear the affinity notifier in the IRQ descriptor */
2757 if (!IS_ENABLED(CONFIG_RFS_ACCEL))
2758 irq_set_affinity_notifier(irq_num, NULL);
2760 /* clear the affinity_mask in the IRQ descriptor */
2761 irq_set_affinity_hint(irq_num, NULL);
2762 synchronize_irq(irq_num);
2763 devm_free_irq(ice_pf_to_dev(pf), irq_num, vsi->q_vectors[i]);
2768 * ice_vsi_free_tx_rings - Free Tx resources for VSI queues
2769 * @vsi: the VSI having resources freed
2771 void ice_vsi_free_tx_rings(struct ice_vsi *vsi)
2778 ice_for_each_txq(vsi, i)
2779 if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc)
2780 ice_free_tx_ring(vsi->tx_rings[i]);
2784 * ice_vsi_free_rx_rings - Free Rx resources for VSI queues
2785 * @vsi: the VSI having resources freed
2787 void ice_vsi_free_rx_rings(struct ice_vsi *vsi)
2794 ice_for_each_rxq(vsi, i)
2795 if (vsi->rx_rings[i] && vsi->rx_rings[i]->desc)
2796 ice_free_rx_ring(vsi->rx_rings[i]);
2800 * ice_vsi_close - Shut down a VSI
2801 * @vsi: the VSI being shut down
2803 void ice_vsi_close(struct ice_vsi *vsi)
2805 if (!test_and_set_bit(ICE_VSI_DOWN, vsi->state))
2808 ice_vsi_free_irq(vsi);
2809 ice_vsi_free_tx_rings(vsi);
2810 ice_vsi_free_rx_rings(vsi);
2814 * ice_ena_vsi - resume a VSI
2815 * @vsi: the VSI being resume
2816 * @locked: is the rtnl_lock already held
2818 int ice_ena_vsi(struct ice_vsi *vsi, bool locked)
2822 if (!test_bit(ICE_VSI_NEEDS_RESTART, vsi->state))
2825 clear_bit(ICE_VSI_NEEDS_RESTART, vsi->state);
2827 if (vsi->netdev && vsi->type == ICE_VSI_PF) {
2828 if (netif_running(vsi->netdev)) {
2832 err = ice_open_internal(vsi->netdev);
2837 } else if (vsi->type == ICE_VSI_CTRL) {
2838 err = ice_vsi_open_ctrl(vsi);
2845 * ice_dis_vsi - pause a VSI
2846 * @vsi: the VSI being paused
2847 * @locked: is the rtnl_lock already held
2849 void ice_dis_vsi(struct ice_vsi *vsi, bool locked)
2851 if (test_bit(ICE_VSI_DOWN, vsi->state))
2854 set_bit(ICE_VSI_NEEDS_RESTART, vsi->state);
2856 if (vsi->type == ICE_VSI_PF && vsi->netdev) {
2857 if (netif_running(vsi->netdev)) {
2868 } else if (vsi->type == ICE_VSI_CTRL ||
2869 vsi->type == ICE_VSI_SWITCHDEV_CTRL) {
2875 * ice_vsi_dis_irq - Mask off queue interrupt generation on the VSI
2876 * @vsi: the VSI being un-configured
2878 void ice_vsi_dis_irq(struct ice_vsi *vsi)
2880 struct ice_pf *pf = vsi->back;
2881 struct ice_hw *hw = &pf->hw;
2885 /* disable interrupt causation from each queue */
2886 if (vsi->tx_rings) {
2887 ice_for_each_txq(vsi, i) {
2888 if (vsi->tx_rings[i]) {
2891 reg = vsi->tx_rings[i]->reg_idx;
2892 val = rd32(hw, QINT_TQCTL(reg));
2893 val &= ~QINT_TQCTL_CAUSE_ENA_M;
2894 wr32(hw, QINT_TQCTL(reg), val);
2899 if (vsi->rx_rings) {
2900 ice_for_each_rxq(vsi, i) {
2901 if (vsi->rx_rings[i]) {
2904 reg = vsi->rx_rings[i]->reg_idx;
2905 val = rd32(hw, QINT_RQCTL(reg));
2906 val &= ~QINT_RQCTL_CAUSE_ENA_M;
2907 wr32(hw, QINT_RQCTL(reg), val);
2912 /* disable each interrupt */
2913 ice_for_each_q_vector(vsi, i) {
2914 if (!vsi->q_vectors[i])
2916 wr32(hw, GLINT_DYN_CTL(vsi->q_vectors[i]->reg_idx), 0);
2921 /* don't call synchronize_irq() for VF's from the host */
2922 if (vsi->type == ICE_VSI_VF)
2925 ice_for_each_q_vector(vsi, i)
2926 synchronize_irq(vsi->q_vectors[i]->irq.virq);
2930 * ice_vsi_release - Delete a VSI and free its resources
2931 * @vsi: the VSI being removed
2933 * Returns 0 on success or < 0 on error
2935 int ice_vsi_release(struct ice_vsi *vsi)
2943 if (test_bit(ICE_FLAG_RSS_ENA, pf->flags))
2949 /* retain SW VSI data structure since it is needed to unregister and
2950 * free VSI netdev when PF is not in reset recovery pending state,\
2951 * for ex: during rmmod.
2953 if (!ice_is_reset_in_progress(pf->state))
2954 ice_vsi_delete(vsi);
2960 * ice_vsi_rebuild_get_coalesce - get coalesce from all q_vectors
2961 * @vsi: VSI connected with q_vectors
2962 * @coalesce: array of struct with stored coalesce
2964 * Returns array size.
2967 ice_vsi_rebuild_get_coalesce(struct ice_vsi *vsi,
2968 struct ice_coalesce_stored *coalesce)
2972 ice_for_each_q_vector(vsi, i) {
2973 struct ice_q_vector *q_vector = vsi->q_vectors[i];
2975 coalesce[i].itr_tx = q_vector->tx.itr_settings;
2976 coalesce[i].itr_rx = q_vector->rx.itr_settings;
2977 coalesce[i].intrl = q_vector->intrl;
2979 if (i < vsi->num_txq)
2980 coalesce[i].tx_valid = true;
2981 if (i < vsi->num_rxq)
2982 coalesce[i].rx_valid = true;
2985 return vsi->num_q_vectors;
2989 * ice_vsi_rebuild_set_coalesce - set coalesce from earlier saved arrays
2990 * @vsi: VSI connected with q_vectors
2991 * @coalesce: pointer to array of struct with stored coalesce
2992 * @size: size of coalesce array
2994 * Before this function, ice_vsi_rebuild_get_coalesce should be called to save
2995 * ITR params in arrays. If size is 0 or coalesce wasn't stored set coalesce
2999 ice_vsi_rebuild_set_coalesce(struct ice_vsi *vsi,
3000 struct ice_coalesce_stored *coalesce, int size)
3002 struct ice_ring_container *rc;
3005 if ((size && !coalesce) || !vsi)
3008 /* There are a couple of cases that have to be handled here:
3009 * 1. The case where the number of queue vectors stays the same, but
3010 * the number of Tx or Rx rings changes (the first for loop)
3011 * 2. The case where the number of queue vectors increased (the
3014 for (i = 0; i < size && i < vsi->num_q_vectors; i++) {
3015 /* There are 2 cases to handle here and they are the same for
3017 * if the entry was valid previously (coalesce[i].[tr]x_valid
3018 * and the loop variable is less than the number of rings
3019 * allocated, then write the previous values
3021 * if the entry was not valid previously, but the number of
3022 * rings is less than are allocated (this means the number of
3023 * rings increased from previously), then write out the
3024 * values in the first element
3026 * Also, always write the ITR, even if in ITR_IS_DYNAMIC
3027 * as there is no harm because the dynamic algorithm
3028 * will just overwrite.
3030 if (i < vsi->alloc_rxq && coalesce[i].rx_valid) {
3031 rc = &vsi->q_vectors[i]->rx;
3032 rc->itr_settings = coalesce[i].itr_rx;
3033 ice_write_itr(rc, rc->itr_setting);
3034 } else if (i < vsi->alloc_rxq) {
3035 rc = &vsi->q_vectors[i]->rx;
3036 rc->itr_settings = coalesce[0].itr_rx;
3037 ice_write_itr(rc, rc->itr_setting);
3040 if (i < vsi->alloc_txq && coalesce[i].tx_valid) {
3041 rc = &vsi->q_vectors[i]->tx;
3042 rc->itr_settings = coalesce[i].itr_tx;
3043 ice_write_itr(rc, rc->itr_setting);
3044 } else if (i < vsi->alloc_txq) {
3045 rc = &vsi->q_vectors[i]->tx;
3046 rc->itr_settings = coalesce[0].itr_tx;
3047 ice_write_itr(rc, rc->itr_setting);
3050 vsi->q_vectors[i]->intrl = coalesce[i].intrl;
3051 ice_set_q_vector_intrl(vsi->q_vectors[i]);
3054 /* the number of queue vectors increased so write whatever is in
3057 for (; i < vsi->num_q_vectors; i++) {
3059 rc = &vsi->q_vectors[i]->tx;
3060 rc->itr_settings = coalesce[0].itr_tx;
3061 ice_write_itr(rc, rc->itr_setting);
3064 rc = &vsi->q_vectors[i]->rx;
3065 rc->itr_settings = coalesce[0].itr_rx;
3066 ice_write_itr(rc, rc->itr_setting);
3068 vsi->q_vectors[i]->intrl = coalesce[0].intrl;
3069 ice_set_q_vector_intrl(vsi->q_vectors[i]);
3074 * ice_vsi_realloc_stat_arrays - Frees unused stat structures
3076 * @prev_txq: Number of Tx rings before ring reallocation
3077 * @prev_rxq: Number of Rx rings before ring reallocation
3080 ice_vsi_realloc_stat_arrays(struct ice_vsi *vsi, int prev_txq, int prev_rxq)
3082 struct ice_vsi_stats *vsi_stat;
3083 struct ice_pf *pf = vsi->back;
3086 if (!prev_txq || !prev_rxq)
3088 if (vsi->type == ICE_VSI_CHNL)
3091 vsi_stat = pf->vsi_stats[vsi->idx];
3093 if (vsi->num_txq < prev_txq) {
3094 for (i = vsi->num_txq; i < prev_txq; i++) {
3095 if (vsi_stat->tx_ring_stats[i]) {
3096 kfree_rcu(vsi_stat->tx_ring_stats[i], rcu);
3097 WRITE_ONCE(vsi_stat->tx_ring_stats[i], NULL);
3102 if (vsi->num_rxq < prev_rxq) {
3103 for (i = vsi->num_rxq; i < prev_rxq; i++) {
3104 if (vsi_stat->rx_ring_stats[i]) {
3105 kfree_rcu(vsi_stat->rx_ring_stats[i], rcu);
3106 WRITE_ONCE(vsi_stat->rx_ring_stats[i], NULL);
3113 * ice_vsi_rebuild - Rebuild VSI after reset
3114 * @vsi: VSI to be rebuild
3115 * @vsi_flags: flags used for VSI rebuild flow
3117 * Set vsi_flags to ICE_VSI_FLAG_INIT to initialize a new VSI, or
3118 * ICE_VSI_FLAG_NO_INIT to rebuild an existing VSI in hardware.
3120 * Returns 0 on success and negative value on failure
3122 int ice_vsi_rebuild(struct ice_vsi *vsi, u32 vsi_flags)
3124 struct ice_vsi_cfg_params params = {};
3125 struct ice_coalesce_stored *coalesce;
3126 int ret, prev_txq, prev_rxq;
3127 int prev_num_q_vectors = 0;
3133 params = ice_vsi_to_params(vsi);
3134 params.flags = vsi_flags;
3137 if (WARN_ON(vsi->type == ICE_VSI_VF && !vsi->vf))
3140 coalesce = kcalloc(vsi->num_q_vectors,
3141 sizeof(struct ice_coalesce_stored), GFP_KERNEL);
3145 prev_num_q_vectors = ice_vsi_rebuild_get_coalesce(vsi, coalesce);
3147 prev_txq = vsi->num_txq;
3148 prev_rxq = vsi->num_rxq;
3151 ret = ice_vsi_cfg_def(vsi, ¶ms);
3155 ret = ice_vsi_cfg_tc_lan(pf, vsi);
3157 if (vsi_flags & ICE_VSI_FLAG_INIT) {
3159 goto err_vsi_cfg_tc_lan;
3163 return ice_schedule_reset(pf, ICE_RESET_PFR);
3166 ice_vsi_realloc_stat_arrays(vsi, prev_txq, prev_rxq);
3168 ice_vsi_rebuild_set_coalesce(vsi, coalesce, prev_num_q_vectors);
3181 * ice_is_reset_in_progress - check for a reset in progress
3182 * @state: PF state field
3184 bool ice_is_reset_in_progress(unsigned long *state)
3186 return test_bit(ICE_RESET_OICR_RECV, state) ||
3187 test_bit(ICE_PFR_REQ, state) ||
3188 test_bit(ICE_CORER_REQ, state) ||
3189 test_bit(ICE_GLOBR_REQ, state);
3193 * ice_wait_for_reset - Wait for driver to finish reset and rebuild
3194 * @pf: pointer to the PF structure
3195 * @timeout: length of time to wait, in jiffies
3197 * Wait (sleep) for a short time until the driver finishes cleaning up from
3198 * a device reset. The caller must be able to sleep. Use this to delay
3199 * operations that could fail while the driver is cleaning up after a device
3202 * Returns 0 on success, -EBUSY if the reset is not finished within the
3203 * timeout, and -ERESTARTSYS if the thread was interrupted.
3205 int ice_wait_for_reset(struct ice_pf *pf, unsigned long timeout)
3209 ret = wait_event_interruptible_timeout(pf->reset_wait_queue,
3210 !ice_is_reset_in_progress(pf->state),
3221 * ice_vsi_update_q_map - update our copy of the VSI info with new queue map
3222 * @vsi: VSI being configured
3223 * @ctx: the context buffer returned from AQ VSI update command
3225 static void ice_vsi_update_q_map(struct ice_vsi *vsi, struct ice_vsi_ctx *ctx)
3227 vsi->info.mapping_flags = ctx->info.mapping_flags;
3228 memcpy(&vsi->info.q_mapping, &ctx->info.q_mapping,
3229 sizeof(vsi->info.q_mapping));
3230 memcpy(&vsi->info.tc_mapping, ctx->info.tc_mapping,
3231 sizeof(vsi->info.tc_mapping));
3235 * ice_vsi_cfg_netdev_tc - Setup the netdev TC configuration
3236 * @vsi: the VSI being configured
3237 * @ena_tc: TC map to be enabled
3239 void ice_vsi_cfg_netdev_tc(struct ice_vsi *vsi, u8 ena_tc)
3241 struct net_device *netdev = vsi->netdev;
3242 struct ice_pf *pf = vsi->back;
3243 int numtc = vsi->tc_cfg.numtc;
3244 struct ice_dcbx_cfg *dcbcfg;
3251 /* CHNL VSI doesn't have it's own netdev, hence, no netdev_tc */
3252 if (vsi->type == ICE_VSI_CHNL)
3256 netdev_reset_tc(netdev);
3260 if (vsi->type == ICE_VSI_PF && ice_is_adq_active(pf))
3261 numtc = vsi->all_numtc;
3263 if (netdev_set_num_tc(netdev, numtc))
3266 dcbcfg = &pf->hw.port_info->qos_cfg.local_dcbx_cfg;
3268 ice_for_each_traffic_class(i)
3269 if (vsi->tc_cfg.ena_tc & BIT(i))
3270 netdev_set_tc_queue(netdev,
3271 vsi->tc_cfg.tc_info[i].netdev_tc,
3272 vsi->tc_cfg.tc_info[i].qcount_tx,
3273 vsi->tc_cfg.tc_info[i].qoffset);
3274 /* setup TC queue map for CHNL TCs */
3275 ice_for_each_chnl_tc(i) {
3276 if (!(vsi->all_enatc & BIT(i)))
3278 if (!vsi->mqprio_qopt.qopt.count[i])
3280 netdev_set_tc_queue(netdev, i,
3281 vsi->mqprio_qopt.qopt.count[i],
3282 vsi->mqprio_qopt.qopt.offset[i]);
3285 if (test_bit(ICE_FLAG_TC_MQPRIO, pf->flags))
3288 for (i = 0; i < ICE_MAX_USER_PRIORITY; i++) {
3289 u8 ets_tc = dcbcfg->etscfg.prio_table[i];
3291 /* Get the mapped netdev TC# for the UP */
3292 netdev_tc = vsi->tc_cfg.tc_info[ets_tc].netdev_tc;
3293 netdev_set_prio_tc_map(netdev, i, netdev_tc);
3298 * ice_vsi_setup_q_map_mqprio - Prepares mqprio based tc_config
3299 * @vsi: the VSI being configured,
3300 * @ctxt: VSI context structure
3301 * @ena_tc: number of traffic classes to enable
3303 * Prepares VSI tc_config to have queue configurations based on MQPRIO options.
3306 ice_vsi_setup_q_map_mqprio(struct ice_vsi *vsi, struct ice_vsi_ctx *ctxt,
3309 u16 pow, offset = 0, qcount_tx = 0, qcount_rx = 0, qmap;
3310 u16 tc0_offset = vsi->mqprio_qopt.qopt.offset[0];
3311 int tc0_qcount = vsi->mqprio_qopt.qopt.count[0];
3312 u16 new_txq, new_rxq;
3316 vsi->tc_cfg.ena_tc = ena_tc ? ena_tc : 1;
3318 pow = order_base_2(tc0_qcount);
3319 qmap = ((tc0_offset << ICE_AQ_VSI_TC_Q_OFFSET_S) &
3320 ICE_AQ_VSI_TC_Q_OFFSET_M) |
3321 ((pow << ICE_AQ_VSI_TC_Q_NUM_S) & ICE_AQ_VSI_TC_Q_NUM_M);
3323 ice_for_each_traffic_class(i) {
3324 if (!(vsi->tc_cfg.ena_tc & BIT(i))) {
3325 /* TC is not enabled */
3326 vsi->tc_cfg.tc_info[i].qoffset = 0;
3327 vsi->tc_cfg.tc_info[i].qcount_rx = 1;
3328 vsi->tc_cfg.tc_info[i].qcount_tx = 1;
3329 vsi->tc_cfg.tc_info[i].netdev_tc = 0;
3330 ctxt->info.tc_mapping[i] = 0;
3334 offset = vsi->mqprio_qopt.qopt.offset[i];
3335 qcount_rx = vsi->mqprio_qopt.qopt.count[i];
3336 qcount_tx = vsi->mqprio_qopt.qopt.count[i];
3337 vsi->tc_cfg.tc_info[i].qoffset = offset;
3338 vsi->tc_cfg.tc_info[i].qcount_rx = qcount_rx;
3339 vsi->tc_cfg.tc_info[i].qcount_tx = qcount_tx;
3340 vsi->tc_cfg.tc_info[i].netdev_tc = netdev_tc++;
3343 if (vsi->all_numtc && vsi->all_numtc != vsi->tc_cfg.numtc) {
3344 ice_for_each_chnl_tc(i) {
3345 if (!(vsi->all_enatc & BIT(i)))
3347 offset = vsi->mqprio_qopt.qopt.offset[i];
3348 qcount_rx = vsi->mqprio_qopt.qopt.count[i];
3349 qcount_tx = vsi->mqprio_qopt.qopt.count[i];
3353 new_txq = offset + qcount_tx;
3354 if (new_txq > vsi->alloc_txq) {
3355 dev_err(ice_pf_to_dev(vsi->back), "Trying to use more Tx queues (%u), than were allocated (%u)!\n",
3356 new_txq, vsi->alloc_txq);
3360 new_rxq = offset + qcount_rx;
3361 if (new_rxq > vsi->alloc_rxq) {
3362 dev_err(ice_pf_to_dev(vsi->back), "Trying to use more Rx queues (%u), than were allocated (%u)!\n",
3363 new_rxq, vsi->alloc_rxq);
3367 /* Set actual Tx/Rx queue pairs */
3368 vsi->num_txq = new_txq;
3369 vsi->num_rxq = new_rxq;
3371 /* Setup queue TC[0].qmap for given VSI context */
3372 ctxt->info.tc_mapping[0] = cpu_to_le16(qmap);
3373 ctxt->info.q_mapping[0] = cpu_to_le16(vsi->rxq_map[0]);
3374 ctxt->info.q_mapping[1] = cpu_to_le16(tc0_qcount);
3376 /* Find queue count available for channel VSIs and starting offset
3379 if (tc0_qcount && tc0_qcount < vsi->num_rxq) {
3380 vsi->cnt_q_avail = vsi->num_rxq - tc0_qcount;
3381 vsi->next_base_q = tc0_qcount;
3383 dev_dbg(ice_pf_to_dev(vsi->back), "vsi->num_txq = %d\n", vsi->num_txq);
3384 dev_dbg(ice_pf_to_dev(vsi->back), "vsi->num_rxq = %d\n", vsi->num_rxq);
3385 dev_dbg(ice_pf_to_dev(vsi->back), "all_numtc %u, all_enatc: 0x%04x, tc_cfg.numtc %u\n",
3386 vsi->all_numtc, vsi->all_enatc, vsi->tc_cfg.numtc);
3392 * ice_vsi_cfg_tc - Configure VSI Tx Sched for given TC map
3393 * @vsi: VSI to be configured
3394 * @ena_tc: TC bitmap
3396 * VSI queues expected to be quiesced before calling this function
3398 int ice_vsi_cfg_tc(struct ice_vsi *vsi, u8 ena_tc)
3400 u16 max_txqs[ICE_MAX_TRAFFIC_CLASS] = { 0 };
3401 struct ice_pf *pf = vsi->back;
3402 struct ice_tc_cfg old_tc_cfg;
3403 struct ice_vsi_ctx *ctx;
3408 dev = ice_pf_to_dev(pf);
3409 if (vsi->tc_cfg.ena_tc == ena_tc &&
3410 vsi->mqprio_qopt.mode != TC_MQPRIO_MODE_CHANNEL)
3413 ice_for_each_traffic_class(i) {
3414 /* build bitmap of enabled TCs */
3415 if (ena_tc & BIT(i))
3417 /* populate max_txqs per TC */
3418 max_txqs[i] = vsi->alloc_txq;
3419 /* Update max_txqs if it is CHNL VSI, because alloc_t[r]xq are
3420 * zero for CHNL VSI, hence use num_txq instead as max_txqs
3422 if (vsi->type == ICE_VSI_CHNL &&
3423 test_bit(ICE_FLAG_TC_MQPRIO, pf->flags))
3424 max_txqs[i] = vsi->num_txq;
3427 memcpy(&old_tc_cfg, &vsi->tc_cfg, sizeof(old_tc_cfg));
3428 vsi->tc_cfg.ena_tc = ena_tc;
3429 vsi->tc_cfg.numtc = num_tc;
3431 ctx = kzalloc(sizeof(*ctx), GFP_KERNEL);
3436 ctx->info = vsi->info;
3438 if (vsi->type == ICE_VSI_PF &&
3439 test_bit(ICE_FLAG_TC_MQPRIO, pf->flags))
3440 ret = ice_vsi_setup_q_map_mqprio(vsi, ctx, ena_tc);
3442 ret = ice_vsi_setup_q_map(vsi, ctx);
3445 memcpy(&vsi->tc_cfg, &old_tc_cfg, sizeof(vsi->tc_cfg));
3449 /* must to indicate which section of VSI context are being modified */
3450 ctx->info.valid_sections = cpu_to_le16(ICE_AQ_VSI_PROP_RXQ_MAP_VALID);
3451 ret = ice_update_vsi(&pf->hw, vsi->idx, ctx, NULL);
3453 dev_info(dev, "Failed VSI Update\n");
3457 if (vsi->type == ICE_VSI_PF &&
3458 test_bit(ICE_FLAG_TC_MQPRIO, pf->flags))
3459 ret = ice_cfg_vsi_lan(vsi->port_info, vsi->idx, 1, max_txqs);
3461 ret = ice_cfg_vsi_lan(vsi->port_info, vsi->idx,
3462 vsi->tc_cfg.ena_tc, max_txqs);
3465 dev_err(dev, "VSI %d failed TC config, error %d\n",
3469 ice_vsi_update_q_map(vsi, ctx);
3470 vsi->info.valid_sections = 0;
3472 ice_vsi_cfg_netdev_tc(vsi, ena_tc);
3479 * ice_update_ring_stats - Update ring statistics
3480 * @stats: stats to be updated
3481 * @pkts: number of processed packets
3482 * @bytes: number of processed bytes
3484 * This function assumes that caller has acquired a u64_stats_sync lock.
3486 static void ice_update_ring_stats(struct ice_q_stats *stats, u64 pkts, u64 bytes)
3488 stats->bytes += bytes;
3489 stats->pkts += pkts;
3493 * ice_update_tx_ring_stats - Update Tx ring specific counters
3494 * @tx_ring: ring to update
3495 * @pkts: number of processed packets
3496 * @bytes: number of processed bytes
3498 void ice_update_tx_ring_stats(struct ice_tx_ring *tx_ring, u64 pkts, u64 bytes)
3500 u64_stats_update_begin(&tx_ring->ring_stats->syncp);
3501 ice_update_ring_stats(&tx_ring->ring_stats->stats, pkts, bytes);
3502 u64_stats_update_end(&tx_ring->ring_stats->syncp);
3506 * ice_update_rx_ring_stats - Update Rx ring specific counters
3507 * @rx_ring: ring to update
3508 * @pkts: number of processed packets
3509 * @bytes: number of processed bytes
3511 void ice_update_rx_ring_stats(struct ice_rx_ring *rx_ring, u64 pkts, u64 bytes)
3513 u64_stats_update_begin(&rx_ring->ring_stats->syncp);
3514 ice_update_ring_stats(&rx_ring->ring_stats->stats, pkts, bytes);
3515 u64_stats_update_end(&rx_ring->ring_stats->syncp);
3519 * ice_is_dflt_vsi_in_use - check if the default forwarding VSI is being used
3520 * @pi: port info of the switch with default VSI
3522 * Return true if the there is a single VSI in default forwarding VSI list
3524 bool ice_is_dflt_vsi_in_use(struct ice_port_info *pi)
3526 bool exists = false;
3528 ice_check_if_dflt_vsi(pi, 0, &exists);
3533 * ice_is_vsi_dflt_vsi - check if the VSI passed in is the default VSI
3534 * @vsi: VSI to compare against default forwarding VSI
3536 * If this VSI passed in is the default forwarding VSI then return true, else
3539 bool ice_is_vsi_dflt_vsi(struct ice_vsi *vsi)
3541 return ice_check_if_dflt_vsi(vsi->port_info, vsi->idx, NULL);
3545 * ice_set_dflt_vsi - set the default forwarding VSI
3546 * @vsi: VSI getting set as the default forwarding VSI on the switch
3548 * If the VSI passed in is already the default VSI and it's enabled just return
3551 * Otherwise try to set the VSI passed in as the switch's default VSI and
3552 * return the result.
3554 int ice_set_dflt_vsi(struct ice_vsi *vsi)
3562 dev = ice_pf_to_dev(vsi->back);
3564 if (ice_lag_is_switchdev_running(vsi->back)) {
3565 dev_dbg(dev, "VSI %d passed is a part of LAG containing interfaces in switchdev mode, nothing to do\n",
3570 /* the VSI passed in is already the default VSI */
3571 if (ice_is_vsi_dflt_vsi(vsi)) {
3572 dev_dbg(dev, "VSI %d passed in is already the default forwarding VSI, nothing to do\n",
3577 status = ice_cfg_dflt_vsi(vsi->port_info, vsi->idx, true, ICE_FLTR_RX);
3579 dev_err(dev, "Failed to set VSI %d as the default forwarding VSI, error %d\n",
3580 vsi->vsi_num, status);
3588 * ice_clear_dflt_vsi - clear the default forwarding VSI
3589 * @vsi: VSI to remove from filter list
3591 * If the switch has no default VSI or it's not enabled then return error.
3593 * Otherwise try to clear the default VSI and return the result.
3595 int ice_clear_dflt_vsi(struct ice_vsi *vsi)
3603 dev = ice_pf_to_dev(vsi->back);
3605 /* there is no default VSI configured */
3606 if (!ice_is_dflt_vsi_in_use(vsi->port_info))
3609 status = ice_cfg_dflt_vsi(vsi->port_info, vsi->idx, false,
3612 dev_err(dev, "Failed to clear the default forwarding VSI %d, error %d\n",
3613 vsi->vsi_num, status);
3621 * ice_get_link_speed_mbps - get link speed in Mbps
3622 * @vsi: the VSI whose link speed is being queried
3624 * Return current VSI link speed and 0 if the speed is unknown.
3626 int ice_get_link_speed_mbps(struct ice_vsi *vsi)
3628 unsigned int link_speed;
3630 link_speed = vsi->port_info->phy.link_info.link_speed;
3632 return (int)ice_get_link_speed(fls(link_speed) - 1);
3636 * ice_get_link_speed_kbps - get link speed in Kbps
3637 * @vsi: the VSI whose link speed is being queried
3639 * Return current VSI link speed and 0 if the speed is unknown.
3641 int ice_get_link_speed_kbps(struct ice_vsi *vsi)
3645 speed_mbps = ice_get_link_speed_mbps(vsi);
3647 return speed_mbps * 1000;
3651 * ice_set_min_bw_limit - setup minimum BW limit for Tx based on min_tx_rate
3652 * @vsi: VSI to be configured
3653 * @min_tx_rate: min Tx rate in Kbps to be configured as BW limit
3655 * If the min_tx_rate is specified as 0 that means to clear the minimum BW limit
3656 * profile, otherwise a non-zero value will force a minimum BW limit for the VSI
3659 int ice_set_min_bw_limit(struct ice_vsi *vsi, u64 min_tx_rate)
3661 struct ice_pf *pf = vsi->back;
3666 dev = ice_pf_to_dev(pf);
3667 if (!vsi->port_info) {
3668 dev_dbg(dev, "VSI %d, type %u specified doesn't have valid port_info\n",
3669 vsi->idx, vsi->type);
3673 speed = ice_get_link_speed_kbps(vsi);
3674 if (min_tx_rate > (u64)speed) {
3675 dev_err(dev, "invalid min Tx rate %llu Kbps specified for %s %d is greater than current link speed %u Kbps\n",
3676 min_tx_rate, ice_vsi_type_str(vsi->type), vsi->idx,
3681 /* Configure min BW for VSI limit */
3683 status = ice_cfg_vsi_bw_lmt_per_tc(vsi->port_info, vsi->idx, 0,
3684 ICE_MIN_BW, min_tx_rate);
3686 dev_err(dev, "failed to set min Tx rate(%llu Kbps) for %s %d\n",
3687 min_tx_rate, ice_vsi_type_str(vsi->type),
3692 dev_dbg(dev, "set min Tx rate(%llu Kbps) for %s\n",
3693 min_tx_rate, ice_vsi_type_str(vsi->type));
3695 status = ice_cfg_vsi_bw_dflt_lmt_per_tc(vsi->port_info,
3699 dev_err(dev, "failed to clear min Tx rate configuration for %s %d\n",
3700 ice_vsi_type_str(vsi->type), vsi->idx);
3704 dev_dbg(dev, "cleared min Tx rate configuration for %s %d\n",
3705 ice_vsi_type_str(vsi->type), vsi->idx);
3712 * ice_set_max_bw_limit - setup maximum BW limit for Tx based on max_tx_rate
3713 * @vsi: VSI to be configured
3714 * @max_tx_rate: max Tx rate in Kbps to be configured as BW limit
3716 * If the max_tx_rate is specified as 0 that means to clear the maximum BW limit
3717 * profile, otherwise a non-zero value will force a maximum BW limit for the VSI
3720 int ice_set_max_bw_limit(struct ice_vsi *vsi, u64 max_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 (max_tx_rate > (u64)speed) {
3736 dev_err(dev, "invalid max Tx rate %llu Kbps specified for %s %d is greater than current link speed %u Kbps\n",
3737 max_tx_rate, ice_vsi_type_str(vsi->type), vsi->idx,
3742 /* Configure max BW for VSI limit */
3744 status = ice_cfg_vsi_bw_lmt_per_tc(vsi->port_info, vsi->idx, 0,
3745 ICE_MAX_BW, max_tx_rate);
3747 dev_err(dev, "failed setting max Tx rate(%llu Kbps) for %s %d\n",
3748 max_tx_rate, ice_vsi_type_str(vsi->type),
3753 dev_dbg(dev, "set max Tx rate(%llu Kbps) for %s %d\n",
3754 max_tx_rate, ice_vsi_type_str(vsi->type), vsi->idx);
3756 status = ice_cfg_vsi_bw_dflt_lmt_per_tc(vsi->port_info,
3760 dev_err(dev, "failed clearing max Tx rate configuration for %s %d\n",
3761 ice_vsi_type_str(vsi->type), vsi->idx);
3765 dev_dbg(dev, "cleared max Tx rate configuration for %s %d\n",
3766 ice_vsi_type_str(vsi->type), vsi->idx);
3773 * ice_set_link - turn on/off physical link
3774 * @vsi: VSI to modify physical link on
3775 * @ena: turn on/off physical link
3777 int ice_set_link(struct ice_vsi *vsi, bool ena)
3779 struct device *dev = ice_pf_to_dev(vsi->back);
3780 struct ice_port_info *pi = vsi->port_info;
3781 struct ice_hw *hw = pi->hw;
3784 if (vsi->type != ICE_VSI_PF)
3787 status = ice_aq_set_link_restart_an(pi, ena, NULL);
3789 /* if link is owned by manageability, FW will return ICE_AQ_RC_EMODE.
3790 * this is not a fatal error, so print a warning message and return
3791 * a success code. Return an error if FW returns an error code other
3792 * than ICE_AQ_RC_EMODE
3794 if (status == -EIO) {
3795 if (hw->adminq.sq_last_status == ICE_AQ_RC_EMODE)
3796 dev_dbg(dev, "can't set link to %s, err %d aq_err %s. not fatal, continuing\n",
3797 (ena ? "ON" : "OFF"), status,
3798 ice_aq_str(hw->adminq.sq_last_status));
3799 } else if (status) {
3800 dev_err(dev, "can't set link to %s, err %d aq_err %s\n",
3801 (ena ? "ON" : "OFF"), status,
3802 ice_aq_str(hw->adminq.sq_last_status));
3810 * ice_vsi_add_vlan_zero - add VLAN 0 filter(s) for this VSI
3811 * @vsi: VSI used to add VLAN filters
3813 * In Single VLAN Mode (SVM), single VLAN filters via ICE_SW_LKUP_VLAN are based
3814 * on the inner VLAN ID, so the VLAN TPID (i.e. 0x8100 or 0x888a8) doesn't
3815 * matter. In Double VLAN Mode (DVM), outer/single VLAN filters via
3816 * ICE_SW_LKUP_VLAN are based on the outer/single VLAN ID + VLAN TPID.
3818 * For both modes add a VLAN 0 + no VLAN TPID filter to handle untagged traffic
3819 * when VLAN pruning is enabled. Also, this handles VLAN 0 priority tagged
3820 * traffic in SVM, since the VLAN TPID isn't part of filtering.
3822 * If DVM is enabled then an explicit VLAN 0 + VLAN TPID filter needs to be
3823 * added to allow VLAN 0 priority tagged traffic in DVM, since the VLAN TPID is
3824 * part of filtering.
3826 int ice_vsi_add_vlan_zero(struct ice_vsi *vsi)
3828 struct ice_vsi_vlan_ops *vlan_ops = ice_get_compat_vsi_vlan_ops(vsi);
3829 struct ice_vlan vlan;
3832 vlan = ICE_VLAN(0, 0, 0);
3833 err = vlan_ops->add_vlan(vsi, &vlan);
3834 if (err && err != -EEXIST)
3837 /* in SVM both VLAN 0 filters are identical */
3838 if (!ice_is_dvm_ena(&vsi->back->hw))
3841 vlan = ICE_VLAN(ETH_P_8021Q, 0, 0);
3842 err = vlan_ops->add_vlan(vsi, &vlan);
3843 if (err && err != -EEXIST)
3850 * ice_vsi_del_vlan_zero - delete VLAN 0 filter(s) for this VSI
3851 * @vsi: VSI used to add VLAN filters
3853 * Delete the VLAN 0 filters in the same manner that they were added in
3854 * ice_vsi_add_vlan_zero.
3856 int ice_vsi_del_vlan_zero(struct ice_vsi *vsi)
3858 struct ice_vsi_vlan_ops *vlan_ops = ice_get_compat_vsi_vlan_ops(vsi);
3859 struct ice_vlan vlan;
3862 vlan = ICE_VLAN(0, 0, 0);
3863 err = vlan_ops->del_vlan(vsi, &vlan);
3864 if (err && err != -EEXIST)
3867 /* in SVM both VLAN 0 filters are identical */
3868 if (!ice_is_dvm_ena(&vsi->back->hw))
3871 vlan = ICE_VLAN(ETH_P_8021Q, 0, 0);
3872 err = vlan_ops->del_vlan(vsi, &vlan);
3873 if (err && err != -EEXIST)
3876 /* when deleting the last VLAN filter, make sure to disable the VLAN
3877 * promisc mode so the filter isn't left by accident
3879 return ice_clear_vsi_promisc(&vsi->back->hw, vsi->idx,
3880 ICE_MCAST_VLAN_PROMISC_BITS, 0);
3884 * ice_vsi_num_zero_vlans - get number of VLAN 0 filters based on VLAN mode
3885 * @vsi: VSI used to get the VLAN mode
3887 * If DVM is enabled then 2 VLAN 0 filters are added, else if SVM is enabled
3888 * then 1 VLAN 0 filter is added. See ice_vsi_add_vlan_zero for more details.
3890 static u16 ice_vsi_num_zero_vlans(struct ice_vsi *vsi)
3892 #define ICE_DVM_NUM_ZERO_VLAN_FLTRS 2
3893 #define ICE_SVM_NUM_ZERO_VLAN_FLTRS 1
3894 /* no VLAN 0 filter is created when a port VLAN is active */
3895 if (vsi->type == ICE_VSI_VF) {
3896 if (WARN_ON(!vsi->vf))
3899 if (ice_vf_is_port_vlan_ena(vsi->vf))
3903 if (ice_is_dvm_ena(&vsi->back->hw))
3904 return ICE_DVM_NUM_ZERO_VLAN_FLTRS;
3906 return ICE_SVM_NUM_ZERO_VLAN_FLTRS;
3910 * ice_vsi_has_non_zero_vlans - check if VSI has any non-zero VLANs
3911 * @vsi: VSI used to determine if any non-zero VLANs have been added
3913 bool ice_vsi_has_non_zero_vlans(struct ice_vsi *vsi)
3915 return (vsi->num_vlan > ice_vsi_num_zero_vlans(vsi));
3919 * ice_vsi_num_non_zero_vlans - get the number of non-zero VLANs for this VSI
3920 * @vsi: VSI used to get the number of non-zero VLANs added
3922 u16 ice_vsi_num_non_zero_vlans(struct ice_vsi *vsi)
3924 return (vsi->num_vlan - ice_vsi_num_zero_vlans(vsi));
3928 * ice_is_feature_supported
3929 * @pf: pointer to the struct ice_pf instance
3930 * @f: feature enum to be checked
3932 * returns true if feature is supported, false otherwise
3934 bool ice_is_feature_supported(struct ice_pf *pf, enum ice_feature f)
3936 if (f < 0 || f >= ICE_F_MAX)
3939 return test_bit(f, pf->features);
3943 * ice_set_feature_support
3944 * @pf: pointer to the struct ice_pf instance
3945 * @f: feature enum to set
3947 void ice_set_feature_support(struct ice_pf *pf, enum ice_feature f)
3949 if (f < 0 || f >= ICE_F_MAX)
3952 set_bit(f, pf->features);
3956 * ice_clear_feature_support
3957 * @pf: pointer to the struct ice_pf instance
3958 * @f: feature enum to clear
3960 void ice_clear_feature_support(struct ice_pf *pf, enum ice_feature f)
3962 if (f < 0 || f >= ICE_F_MAX)
3965 clear_bit(f, pf->features);
3969 * ice_init_feature_support
3970 * @pf: pointer to the struct ice_pf instance
3972 * called during init to setup supported feature
3974 void ice_init_feature_support(struct ice_pf *pf)
3976 switch (pf->hw.device_id) {
3977 case ICE_DEV_ID_E810C_BACKPLANE:
3978 case ICE_DEV_ID_E810C_QSFP:
3979 case ICE_DEV_ID_E810C_SFP:
3980 case ICE_DEV_ID_E810_XXV_BACKPLANE:
3981 case ICE_DEV_ID_E810_XXV_QSFP:
3982 case ICE_DEV_ID_E810_XXV_SFP:
3983 ice_set_feature_support(pf, ICE_F_DSCP);
3984 if (ice_is_phy_rclk_in_netlist(&pf->hw))
3985 ice_set_feature_support(pf, ICE_F_PHY_RCLK);
3986 /* If we don't own the timer - don't enable other caps */
3987 if (!ice_pf_src_tmr_owned(pf))
3989 if (ice_is_cgu_in_netlist(&pf->hw))
3990 ice_set_feature_support(pf, ICE_F_CGU);
3991 if (ice_is_clock_mux_in_netlist(&pf->hw))
3992 ice_set_feature_support(pf, ICE_F_SMA_CTRL);
3993 if (ice_gnss_is_gps_present(&pf->hw))
3994 ice_set_feature_support(pf, ICE_F_GNSS);
4002 * ice_vsi_update_security - update security block in VSI
4003 * @vsi: pointer to VSI structure
4004 * @fill: function pointer to fill ctx
4007 ice_vsi_update_security(struct ice_vsi *vsi, void (*fill)(struct ice_vsi_ctx *))
4009 struct ice_vsi_ctx ctx = { 0 };
4011 ctx.info = vsi->info;
4012 ctx.info.valid_sections = cpu_to_le16(ICE_AQ_VSI_PROP_SECURITY_VALID);
4015 if (ice_update_vsi(&vsi->back->hw, vsi->idx, &ctx, NULL))
4018 vsi->info = ctx.info;
4023 * ice_vsi_ctx_set_antispoof - set antispoof function in VSI ctx
4024 * @ctx: pointer to VSI ctx structure
4026 void ice_vsi_ctx_set_antispoof(struct ice_vsi_ctx *ctx)
4028 ctx->info.sec_flags |= ICE_AQ_VSI_SEC_FLAG_ENA_MAC_ANTI_SPOOF |
4029 (ICE_AQ_VSI_SEC_TX_VLAN_PRUNE_ENA <<
4030 ICE_AQ_VSI_SEC_TX_PRUNE_ENA_S);
4034 * ice_vsi_ctx_clear_antispoof - clear antispoof function in VSI ctx
4035 * @ctx: pointer to VSI ctx structure
4037 void ice_vsi_ctx_clear_antispoof(struct ice_vsi_ctx *ctx)
4039 ctx->info.sec_flags &= ~ICE_AQ_VSI_SEC_FLAG_ENA_MAC_ANTI_SPOOF &
4040 ~(ICE_AQ_VSI_SEC_TX_VLAN_PRUNE_ENA <<
4041 ICE_AQ_VSI_SEC_TX_PRUNE_ENA_S);
4045 * ice_vsi_ctx_set_allow_override - allow destination override on VSI
4046 * @ctx: pointer to VSI ctx structure
4048 void ice_vsi_ctx_set_allow_override(struct ice_vsi_ctx *ctx)
4050 ctx->info.sec_flags |= ICE_AQ_VSI_SEC_FLAG_ALLOW_DEST_OVRD;
4054 * ice_vsi_ctx_clear_allow_override - turn off destination override on VSI
4055 * @ctx: pointer to VSI ctx structure
4057 void ice_vsi_ctx_clear_allow_override(struct ice_vsi_ctx *ctx)
4059 ctx->info.sec_flags &= ~ICE_AQ_VSI_SEC_FLAG_ALLOW_DEST_OVRD;
4063 * ice_vsi_update_local_lb - update sw block in VSI with local loopback bit
4064 * @vsi: pointer to VSI structure
4065 * @set: set or unset the bit
4068 ice_vsi_update_local_lb(struct ice_vsi *vsi, bool set)
4070 struct ice_vsi_ctx ctx = {
4074 ctx.info.valid_sections = cpu_to_le16(ICE_AQ_VSI_PROP_SW_VALID);
4076 ctx.info.sw_flags |= ICE_AQ_VSI_SW_FLAG_LOCAL_LB;
4078 ctx.info.sw_flags &= ~ICE_AQ_VSI_SW_FLAG_LOCAL_LB;
4080 if (ice_update_vsi(&vsi->back->hw, vsi->idx, &ctx, NULL))
4083 vsi->info = ctx.info;