1 // SPDX-License-Identifier: GPL-2.0
2 /* Copyright (c) 2018, Intel Corporation. */
9 #include "ice_dcb_lib.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)
27 return "ICE_VSI_CTRL";
29 return "ICE_VSI_CHNL";
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);
123 devm_kfree(dev, vsi->rxq_map);
125 devm_kfree(dev, vsi->txq_map);
127 devm_kfree(dev, vsi->rx_rings);
129 devm_kfree(dev, vsi->tx_rings);
134 * ice_vsi_set_num_desc - Set number of descriptors for queues on this VSI
135 * @vsi: the VSI being configured
137 static void ice_vsi_set_num_desc(struct ice_vsi *vsi)
144 /* a user could change the values of num_[tr]x_desc using
145 * ethtool -G so we should keep those values instead of
146 * overwriting them with the defaults.
148 if (!vsi->num_rx_desc)
149 vsi->num_rx_desc = ICE_DFLT_NUM_RX_DESC;
150 if (!vsi->num_tx_desc)
151 vsi->num_tx_desc = ICE_DFLT_NUM_TX_DESC;
154 dev_dbg(ice_pf_to_dev(vsi->back), "Not setting number of Tx/Rx descriptors for VSI type %d\n",
161 * ice_vsi_set_num_qs - Set number of queues, descriptors and vectors for a VSI
162 * @vsi: the VSI being configured
164 * Return 0 on success and a negative value on error
166 static void ice_vsi_set_num_qs(struct ice_vsi *vsi)
168 enum ice_vsi_type vsi_type = vsi->type;
169 struct ice_pf *pf = vsi->back;
170 struct ice_vf *vf = vsi->vf;
172 if (WARN_ON(vsi_type == ICE_VSI_VF && !vf))
178 vsi->alloc_txq = vsi->req_txq;
179 vsi->num_txq = vsi->req_txq;
181 vsi->alloc_txq = min3(pf->num_lan_msix,
182 ice_get_avail_txq_count(pf),
183 (u16)num_online_cpus());
186 pf->num_lan_tx = vsi->alloc_txq;
188 /* only 1 Rx queue unless RSS is enabled */
189 if (!test_bit(ICE_FLAG_RSS_ENA, pf->flags)) {
193 vsi->alloc_rxq = vsi->req_rxq;
194 vsi->num_rxq = vsi->req_rxq;
196 vsi->alloc_rxq = min3(pf->num_lan_msix,
197 ice_get_avail_rxq_count(pf),
198 (u16)num_online_cpus());
202 pf->num_lan_rx = vsi->alloc_rxq;
204 vsi->num_q_vectors = min_t(int, pf->num_lan_msix,
205 max_t(int, vsi->alloc_rxq,
211 vsi->num_q_vectors = 1;
212 vsi->irq_dyn_alloc = true;
216 vf->num_vf_qs = vf->num_req_qs;
217 vsi->alloc_txq = vf->num_vf_qs;
218 vsi->alloc_rxq = vf->num_vf_qs;
219 /* pf->vfs.num_msix_per includes (VF miscellaneous vector +
220 * data queue interrupts). Since vsi->num_q_vectors is number
221 * of queues vectors, subtract 1 (ICE_NONQ_VECS_VF) from the
222 * original vector count
224 vsi->num_q_vectors = vf->num_msix - ICE_NONQ_VECS_VF;
229 vsi->num_q_vectors = 1;
240 dev_warn(ice_pf_to_dev(pf), "Unknown VSI type %d\n", vsi_type);
244 ice_vsi_set_num_desc(vsi);
248 * ice_get_free_slot - get the next non-NULL location index in array
249 * @array: array to search
250 * @size: size of the array
251 * @curr: last known occupied index to be used as a search hint
253 * void * is being used to keep the functionality generic. This lets us use this
254 * function on any array of pointers.
256 static int ice_get_free_slot(void *array, int size, int curr)
258 int **tmp_array = (int **)array;
261 if (curr < (size - 1) && !tmp_array[curr + 1]) {
266 while ((i < size) && (tmp_array[i]))
277 * ice_vsi_delete_from_hw - delete a VSI from the switch
278 * @vsi: pointer to VSI being removed
280 static void ice_vsi_delete_from_hw(struct ice_vsi *vsi)
282 struct ice_pf *pf = vsi->back;
283 struct ice_vsi_ctx *ctxt;
286 ice_fltr_remove_all(vsi);
287 ctxt = kzalloc(sizeof(*ctxt), GFP_KERNEL);
291 if (vsi->type == ICE_VSI_VF)
292 ctxt->vf_num = vsi->vf->vf_id;
293 ctxt->vsi_num = vsi->vsi_num;
295 memcpy(&ctxt->info, &vsi->info, sizeof(ctxt->info));
297 status = ice_free_vsi(&pf->hw, vsi->idx, ctxt, false, NULL);
299 dev_err(ice_pf_to_dev(pf), "Failed to delete VSI %i in FW - error: %d\n",
300 vsi->vsi_num, status);
306 * ice_vsi_free_arrays - De-allocate queue and vector pointer arrays for the VSI
307 * @vsi: pointer to VSI being cleared
309 static void ice_vsi_free_arrays(struct ice_vsi *vsi)
311 struct ice_pf *pf = vsi->back;
314 dev = ice_pf_to_dev(pf);
316 /* free the ring and vector containers */
317 devm_kfree(dev, vsi->q_vectors);
318 vsi->q_vectors = NULL;
319 devm_kfree(dev, vsi->tx_rings);
320 vsi->tx_rings = NULL;
321 devm_kfree(dev, vsi->rx_rings);
322 vsi->rx_rings = NULL;
323 devm_kfree(dev, vsi->txq_map);
325 devm_kfree(dev, vsi->rxq_map);
330 * ice_vsi_free_stats - Free the ring statistics structures
333 static void ice_vsi_free_stats(struct ice_vsi *vsi)
335 struct ice_vsi_stats *vsi_stat;
336 struct ice_pf *pf = vsi->back;
339 if (vsi->type == ICE_VSI_CHNL)
344 vsi_stat = pf->vsi_stats[vsi->idx];
348 ice_for_each_alloc_txq(vsi, i) {
349 if (vsi_stat->tx_ring_stats[i]) {
350 kfree_rcu(vsi_stat->tx_ring_stats[i], rcu);
351 WRITE_ONCE(vsi_stat->tx_ring_stats[i], NULL);
355 ice_for_each_alloc_rxq(vsi, i) {
356 if (vsi_stat->rx_ring_stats[i]) {
357 kfree_rcu(vsi_stat->rx_ring_stats[i], rcu);
358 WRITE_ONCE(vsi_stat->rx_ring_stats[i], NULL);
362 kfree(vsi_stat->tx_ring_stats);
363 kfree(vsi_stat->rx_ring_stats);
365 pf->vsi_stats[vsi->idx] = NULL;
369 * ice_vsi_alloc_ring_stats - Allocates Tx and Rx ring stats for the VSI
370 * @vsi: VSI which is having stats allocated
372 static int ice_vsi_alloc_ring_stats(struct ice_vsi *vsi)
374 struct ice_ring_stats **tx_ring_stats;
375 struct ice_ring_stats **rx_ring_stats;
376 struct ice_vsi_stats *vsi_stats;
377 struct ice_pf *pf = vsi->back;
380 vsi_stats = pf->vsi_stats[vsi->idx];
381 tx_ring_stats = vsi_stats->tx_ring_stats;
382 rx_ring_stats = vsi_stats->rx_ring_stats;
384 /* Allocate Tx ring stats */
385 ice_for_each_alloc_txq(vsi, i) {
386 struct ice_ring_stats *ring_stats;
387 struct ice_tx_ring *ring;
389 ring = vsi->tx_rings[i];
390 ring_stats = tx_ring_stats[i];
393 ring_stats = kzalloc(sizeof(*ring_stats), GFP_KERNEL);
397 WRITE_ONCE(tx_ring_stats[i], ring_stats);
400 ring->ring_stats = ring_stats;
403 /* Allocate Rx ring stats */
404 ice_for_each_alloc_rxq(vsi, i) {
405 struct ice_ring_stats *ring_stats;
406 struct ice_rx_ring *ring;
408 ring = vsi->rx_rings[i];
409 ring_stats = rx_ring_stats[i];
412 ring_stats = kzalloc(sizeof(*ring_stats), GFP_KERNEL);
416 WRITE_ONCE(rx_ring_stats[i], ring_stats);
419 ring->ring_stats = ring_stats;
425 ice_vsi_free_stats(vsi);
430 * ice_vsi_free - clean up and deallocate the provided VSI
431 * @vsi: pointer to VSI being cleared
433 * This deallocates the VSI's queue resources, removes it from the PF's
434 * VSI array if necessary, and deallocates the VSI
436 void ice_vsi_free(struct ice_vsi *vsi)
438 struct ice_pf *pf = NULL;
441 if (!vsi || !vsi->back)
445 dev = ice_pf_to_dev(pf);
447 if (!pf->vsi[vsi->idx] || pf->vsi[vsi->idx] != vsi) {
448 dev_dbg(dev, "vsi does not exist at pf->vsi[%d]\n", vsi->idx);
452 mutex_lock(&pf->sw_mutex);
453 /* updates the PF for this cleared VSI */
455 pf->vsi[vsi->idx] = NULL;
456 pf->next_vsi = vsi->idx;
458 ice_vsi_free_stats(vsi);
459 ice_vsi_free_arrays(vsi);
460 mutex_destroy(&vsi->xdp_state_lock);
461 mutex_unlock(&pf->sw_mutex);
462 devm_kfree(dev, vsi);
465 void ice_vsi_delete(struct ice_vsi *vsi)
467 ice_vsi_delete_from_hw(vsi);
472 * ice_msix_clean_ctrl_vsi - MSIX mode interrupt handler for ctrl VSI
473 * @irq: interrupt number
474 * @data: pointer to a q_vector
476 static irqreturn_t ice_msix_clean_ctrl_vsi(int __always_unused irq, void *data)
478 struct ice_q_vector *q_vector = (struct ice_q_vector *)data;
480 if (!q_vector->tx.tx_ring)
483 #define FDIR_RX_DESC_CLEAN_BUDGET 64
484 ice_clean_rx_irq(q_vector->rx.rx_ring, FDIR_RX_DESC_CLEAN_BUDGET);
485 ice_clean_ctrl_tx_irq(q_vector->tx.tx_ring);
491 * ice_msix_clean_rings - MSIX mode Interrupt Handler
492 * @irq: interrupt number
493 * @data: pointer to a q_vector
495 static irqreturn_t ice_msix_clean_rings(int __always_unused irq, void *data)
497 struct ice_q_vector *q_vector = (struct ice_q_vector *)data;
499 if (!q_vector->tx.tx_ring && !q_vector->rx.rx_ring)
502 q_vector->total_events++;
504 napi_schedule(&q_vector->napi);
510 * ice_vsi_alloc_stat_arrays - Allocate statistics arrays
513 static int ice_vsi_alloc_stat_arrays(struct ice_vsi *vsi)
515 struct ice_vsi_stats *vsi_stat;
516 struct ice_pf *pf = vsi->back;
518 if (vsi->type == ICE_VSI_CHNL)
523 if (pf->vsi_stats[vsi->idx])
524 /* realloc will happen in rebuild path */
527 vsi_stat = kzalloc(sizeof(*vsi_stat), GFP_KERNEL);
531 vsi_stat->tx_ring_stats =
532 kcalloc(vsi->alloc_txq, sizeof(*vsi_stat->tx_ring_stats),
534 if (!vsi_stat->tx_ring_stats)
537 vsi_stat->rx_ring_stats =
538 kcalloc(vsi->alloc_rxq, sizeof(*vsi_stat->rx_ring_stats),
540 if (!vsi_stat->rx_ring_stats)
543 pf->vsi_stats[vsi->idx] = vsi_stat;
548 kfree(vsi_stat->rx_ring_stats);
550 kfree(vsi_stat->tx_ring_stats);
552 pf->vsi_stats[vsi->idx] = NULL;
557 * ice_vsi_alloc_def - set default values for already allocated VSI
559 * @ch: ptr to channel
562 ice_vsi_alloc_def(struct ice_vsi *vsi, struct ice_channel *ch)
564 if (vsi->type != ICE_VSI_CHNL) {
565 ice_vsi_set_num_qs(vsi);
566 if (ice_vsi_alloc_arrays(vsi))
573 /* Setup default MSIX irq handler for VSI */
574 vsi->irq_handler = ice_msix_clean_rings;
577 /* Setup ctrl VSI MSIX irq handler */
578 vsi->irq_handler = ice_msix_clean_ctrl_vsi;
584 vsi->num_rxq = ch->num_rxq;
585 vsi->num_txq = ch->num_txq;
586 vsi->next_base_q = ch->base_q;
592 ice_vsi_free_arrays(vsi);
600 * ice_vsi_alloc - Allocates the next available struct VSI in the PF
601 * @pf: board private structure
603 * Reserves a VSI index from the PF and allocates an empty VSI structure
604 * without a type. The VSI structure must later be initialized by calling
607 * returns a pointer to a VSI on success, NULL on failure.
609 struct ice_vsi *ice_vsi_alloc(struct ice_pf *pf)
611 struct device *dev = ice_pf_to_dev(pf);
612 struct ice_vsi *vsi = NULL;
614 /* Need to protect the allocation of the VSIs at the PF level */
615 mutex_lock(&pf->sw_mutex);
617 /* If we have already allocated our maximum number of VSIs,
618 * pf->next_vsi will be ICE_NO_VSI. If not, pf->next_vsi index
619 * is available to be populated
621 if (pf->next_vsi == ICE_NO_VSI) {
622 dev_dbg(dev, "out of VSI slots!\n");
626 vsi = devm_kzalloc(dev, sizeof(*vsi), GFP_KERNEL);
631 set_bit(ICE_VSI_DOWN, vsi->state);
633 /* fill slot and make note of the index */
634 vsi->idx = pf->next_vsi;
635 pf->vsi[pf->next_vsi] = vsi;
637 /* prepare pf->next_vsi for next use */
638 pf->next_vsi = ice_get_free_slot(pf->vsi, pf->num_alloc_vsi,
641 mutex_init(&vsi->xdp_state_lock);
644 mutex_unlock(&pf->sw_mutex);
649 * ice_alloc_fd_res - Allocate FD resource for a VSI
650 * @vsi: pointer to the ice_vsi
652 * This allocates the FD resources
654 * Returns 0 on success, -EPERM on no-op or -EIO on failure
656 static int ice_alloc_fd_res(struct ice_vsi *vsi)
658 struct ice_pf *pf = vsi->back;
661 /* Flow Director filters are only allocated/assigned to the PF VSI or
662 * CHNL VSI which passes the traffic. The CTRL VSI is only used to
663 * add/delete filters so resources are not allocated to it
665 if (!test_bit(ICE_FLAG_FD_ENA, pf->flags))
668 if (!(vsi->type == ICE_VSI_PF || vsi->type == ICE_VSI_VF ||
669 vsi->type == ICE_VSI_CHNL))
672 /* FD filters from guaranteed pool per VSI */
673 g_val = pf->hw.func_caps.fd_fltr_guar;
677 /* FD filters from best effort pool */
678 b_val = pf->hw.func_caps.fd_fltr_best_effort;
682 /* PF main VSI gets only 64 FD resources from guaranteed pool
683 * when ADQ is configured.
685 #define ICE_PF_VSI_GFLTR 64
687 /* determine FD filter resources per VSI from shared(best effort) and
690 if (vsi->type == ICE_VSI_PF) {
691 vsi->num_gfltr = g_val;
692 /* if MQPRIO is configured, main VSI doesn't get all FD
693 * resources from guaranteed pool. PF VSI gets 64 FD resources
695 if (test_bit(ICE_FLAG_TC_MQPRIO, pf->flags)) {
696 if (g_val < ICE_PF_VSI_GFLTR)
698 /* allow bare minimum entries for PF VSI */
699 vsi->num_gfltr = ICE_PF_VSI_GFLTR;
702 /* each VSI gets same "best_effort" quota */
703 vsi->num_bfltr = b_val;
704 } else if (vsi->type == ICE_VSI_VF) {
707 /* each VSI gets same "best_effort" quota */
708 vsi->num_bfltr = b_val;
710 struct ice_vsi *main_vsi;
713 main_vsi = ice_get_main_vsi(pf);
717 if (!main_vsi->all_numtc)
720 /* figure out ADQ numtc */
721 numtc = main_vsi->all_numtc - ICE_CHNL_START_TC;
723 /* only one TC but still asking resources for channels,
726 if (numtc < ICE_CHNL_START_TC)
729 g_val -= ICE_PF_VSI_GFLTR;
730 /* channel VSIs gets equal share from guaranteed pool */
731 vsi->num_gfltr = g_val / numtc;
733 /* each VSI gets same "best_effort" quota */
734 vsi->num_bfltr = b_val;
741 * ice_vsi_get_qs - Assign queues from PF to VSI
742 * @vsi: the VSI to assign queues to
744 * Returns 0 on success and a negative value on error
746 static int ice_vsi_get_qs(struct ice_vsi *vsi)
748 struct ice_pf *pf = vsi->back;
749 struct ice_qs_cfg tx_qs_cfg = {
750 .qs_mutex = &pf->avail_q_mutex,
751 .pf_map = pf->avail_txqs,
752 .pf_map_size = pf->max_pf_txqs,
753 .q_count = vsi->alloc_txq,
754 .scatter_count = ICE_MAX_SCATTER_TXQS,
755 .vsi_map = vsi->txq_map,
757 .mapping_mode = ICE_VSI_MAP_CONTIG
759 struct ice_qs_cfg rx_qs_cfg = {
760 .qs_mutex = &pf->avail_q_mutex,
761 .pf_map = pf->avail_rxqs,
762 .pf_map_size = pf->max_pf_rxqs,
763 .q_count = vsi->alloc_rxq,
764 .scatter_count = ICE_MAX_SCATTER_RXQS,
765 .vsi_map = vsi->rxq_map,
767 .mapping_mode = ICE_VSI_MAP_CONTIG
771 if (vsi->type == ICE_VSI_CHNL)
774 ret = __ice_vsi_get_qs(&tx_qs_cfg);
777 vsi->tx_mapping_mode = tx_qs_cfg.mapping_mode;
779 ret = __ice_vsi_get_qs(&rx_qs_cfg);
782 vsi->rx_mapping_mode = rx_qs_cfg.mapping_mode;
788 * ice_vsi_put_qs - Release queues from VSI to PF
789 * @vsi: the VSI that is going to release queues
791 static void ice_vsi_put_qs(struct ice_vsi *vsi)
793 struct ice_pf *pf = vsi->back;
796 mutex_lock(&pf->avail_q_mutex);
798 ice_for_each_alloc_txq(vsi, i) {
799 clear_bit(vsi->txq_map[i], pf->avail_txqs);
800 vsi->txq_map[i] = ICE_INVAL_Q_INDEX;
803 ice_for_each_alloc_rxq(vsi, i) {
804 clear_bit(vsi->rxq_map[i], pf->avail_rxqs);
805 vsi->rxq_map[i] = ICE_INVAL_Q_INDEX;
808 mutex_unlock(&pf->avail_q_mutex);
813 * @pf: pointer to the PF struct
815 * returns true if driver is in safe mode, false otherwise
817 bool ice_is_safe_mode(struct ice_pf *pf)
819 return !test_bit(ICE_FLAG_ADV_FEATURES, pf->flags);
824 * @pf: pointer to the PF struct
826 * returns true if RDMA is currently supported, false otherwise
828 bool ice_is_rdma_ena(struct ice_pf *pf)
830 return test_bit(ICE_FLAG_RDMA_ENA, pf->flags);
834 * ice_vsi_clean_rss_flow_fld - Delete RSS configuration
835 * @vsi: the VSI being cleaned up
837 * This function deletes RSS input set for all flows that were configured
840 static void ice_vsi_clean_rss_flow_fld(struct ice_vsi *vsi)
842 struct ice_pf *pf = vsi->back;
845 if (ice_is_safe_mode(pf))
848 status = ice_rem_vsi_rss_cfg(&pf->hw, vsi->idx);
850 dev_dbg(ice_pf_to_dev(pf), "ice_rem_vsi_rss_cfg failed for vsi = %d, error = %d\n",
851 vsi->vsi_num, status);
855 * ice_rss_clean - Delete RSS related VSI structures and configuration
856 * @vsi: the VSI being removed
858 static void ice_rss_clean(struct ice_vsi *vsi)
860 struct ice_pf *pf = vsi->back;
863 dev = ice_pf_to_dev(pf);
865 devm_kfree(dev, vsi->rss_hkey_user);
866 devm_kfree(dev, vsi->rss_lut_user);
868 ice_vsi_clean_rss_flow_fld(vsi);
869 /* remove RSS replay list */
870 if (!ice_is_safe_mode(pf))
871 ice_rem_vsi_rss_list(&pf->hw, vsi->idx);
875 * ice_vsi_set_rss_params - Setup RSS capabilities per VSI type
876 * @vsi: the VSI being configured
878 static void ice_vsi_set_rss_params(struct ice_vsi *vsi)
880 struct ice_hw_common_caps *cap;
881 struct ice_pf *pf = vsi->back;
884 if (!test_bit(ICE_FLAG_RSS_ENA, pf->flags)) {
889 cap = &pf->hw.func_caps.common_cap;
890 max_rss_size = BIT(cap->rss_table_entry_width);
894 /* PF VSI will inherit RSS instance of PF */
895 vsi->rss_table_size = (u16)cap->rss_table_size;
896 if (vsi->type == ICE_VSI_CHNL)
897 vsi->rss_size = min_t(u16, vsi->num_rxq, max_rss_size);
899 vsi->rss_size = min_t(u16, num_online_cpus(),
901 vsi->rss_lut_type = ICE_LUT_PF;
904 vsi->rss_table_size = ICE_LUT_VSI_SIZE;
905 vsi->rss_size = min_t(u16, num_online_cpus(), max_rss_size);
906 vsi->rss_lut_type = ICE_LUT_VSI;
909 /* VF VSI will get a small RSS table.
910 * For VSI_LUT, LUT size should be set to 64 bytes.
912 vsi->rss_table_size = ICE_LUT_VSI_SIZE;
913 vsi->rss_size = ICE_MAX_RSS_QS_PER_VF;
914 vsi->rss_lut_type = ICE_LUT_VSI;
919 dev_dbg(ice_pf_to_dev(pf), "Unsupported VSI type %s\n",
920 ice_vsi_type_str(vsi->type));
926 * ice_set_dflt_vsi_ctx - Set default VSI context before adding a VSI
927 * @hw: HW structure used to determine the VLAN mode of the device
928 * @ctxt: the VSI context being set
930 * This initializes a default VSI context for all sections except the Queues.
932 static void ice_set_dflt_vsi_ctx(struct ice_hw *hw, struct ice_vsi_ctx *ctxt)
936 memset(&ctxt->info, 0, sizeof(ctxt->info));
937 /* VSI's should be allocated from shared pool */
938 ctxt->alloc_from_pool = true;
939 /* Src pruning enabled by default */
940 ctxt->info.sw_flags = ICE_AQ_VSI_SW_FLAG_SRC_PRUNE;
941 /* Traffic from VSI can be sent to LAN */
942 ctxt->info.sw_flags2 = ICE_AQ_VSI_SW_FLAG_LAN_ENA;
943 /* allow all untagged/tagged packets by default on Tx */
944 ctxt->info.inner_vlan_flags = FIELD_PREP(ICE_AQ_VSI_INNER_VLAN_TX_MODE_M,
945 ICE_AQ_VSI_INNER_VLAN_TX_MODE_ALL);
946 /* SVM - by default bits 3 and 4 in inner_vlan_flags are 0's which
947 * results in legacy behavior (show VLAN, DEI, and UP) in descriptor.
949 * DVM - leave inner VLAN in packet by default
951 if (ice_is_dvm_ena(hw)) {
952 ctxt->info.inner_vlan_flags |=
953 FIELD_PREP(ICE_AQ_VSI_INNER_VLAN_EMODE_M,
954 ICE_AQ_VSI_INNER_VLAN_EMODE_NOTHING);
955 ctxt->info.outer_vlan_flags =
956 FIELD_PREP(ICE_AQ_VSI_OUTER_VLAN_TX_MODE_M,
957 ICE_AQ_VSI_OUTER_VLAN_TX_MODE_ALL);
958 ctxt->info.outer_vlan_flags |=
959 FIELD_PREP(ICE_AQ_VSI_OUTER_TAG_TYPE_M,
960 ICE_AQ_VSI_OUTER_TAG_VLAN_8100);
961 ctxt->info.outer_vlan_flags |=
962 FIELD_PREP(ICE_AQ_VSI_OUTER_VLAN_EMODE_M,
963 ICE_AQ_VSI_OUTER_VLAN_EMODE_NOTHING);
965 /* Have 1:1 UP mapping for both ingress/egress tables */
966 table |= ICE_UP_TABLE_TRANSLATE(0, 0);
967 table |= ICE_UP_TABLE_TRANSLATE(1, 1);
968 table |= ICE_UP_TABLE_TRANSLATE(2, 2);
969 table |= ICE_UP_TABLE_TRANSLATE(3, 3);
970 table |= ICE_UP_TABLE_TRANSLATE(4, 4);
971 table |= ICE_UP_TABLE_TRANSLATE(5, 5);
972 table |= ICE_UP_TABLE_TRANSLATE(6, 6);
973 table |= ICE_UP_TABLE_TRANSLATE(7, 7);
974 ctxt->info.ingress_table = cpu_to_le32(table);
975 ctxt->info.egress_table = cpu_to_le32(table);
976 /* Have 1:1 UP mapping for outer to inner UP table */
977 ctxt->info.outer_up_table = cpu_to_le32(table);
978 /* No Outer tag support outer_tag_flags remains to zero */
982 * ice_vsi_setup_q_map - Setup a VSI queue map
983 * @vsi: the VSI being configured
984 * @ctxt: VSI context structure
986 static int ice_vsi_setup_q_map(struct ice_vsi *vsi, struct ice_vsi_ctx *ctxt)
988 u16 offset = 0, qmap = 0, tx_count = 0, rx_count = 0, pow = 0;
989 u16 num_txq_per_tc, num_rxq_per_tc;
990 u16 qcount_tx = vsi->alloc_txq;
991 u16 qcount_rx = vsi->alloc_rxq;
995 if (!vsi->tc_cfg.numtc) {
996 /* at least TC0 should be enabled by default */
997 vsi->tc_cfg.numtc = 1;
998 vsi->tc_cfg.ena_tc = 1;
1001 num_rxq_per_tc = min_t(u16, qcount_rx / vsi->tc_cfg.numtc, ICE_MAX_RXQS_PER_TC);
1002 if (!num_rxq_per_tc)
1004 num_txq_per_tc = qcount_tx / vsi->tc_cfg.numtc;
1005 if (!num_txq_per_tc)
1008 /* find the (rounded up) power-of-2 of qcount */
1009 pow = (u16)order_base_2(num_rxq_per_tc);
1011 /* TC mapping is a function of the number of Rx queues assigned to the
1012 * VSI for each traffic class and the offset of these queues.
1013 * The first 10 bits are for queue offset for TC0, next 4 bits for no:of
1014 * queues allocated to TC0. No:of queues is a power-of-2.
1016 * If TC is not enabled, the queue offset is set to 0, and allocate one
1017 * queue, this way, traffic for the given TC will be sent to the default
1020 * Setup number and offset of Rx queues for all TCs for the VSI
1022 ice_for_each_traffic_class(i) {
1023 if (!(vsi->tc_cfg.ena_tc & BIT(i))) {
1024 /* TC is not enabled */
1025 vsi->tc_cfg.tc_info[i].qoffset = 0;
1026 vsi->tc_cfg.tc_info[i].qcount_rx = 1;
1027 vsi->tc_cfg.tc_info[i].qcount_tx = 1;
1028 vsi->tc_cfg.tc_info[i].netdev_tc = 0;
1029 ctxt->info.tc_mapping[i] = 0;
1034 vsi->tc_cfg.tc_info[i].qoffset = offset;
1035 vsi->tc_cfg.tc_info[i].qcount_rx = num_rxq_per_tc;
1036 vsi->tc_cfg.tc_info[i].qcount_tx = num_txq_per_tc;
1037 vsi->tc_cfg.tc_info[i].netdev_tc = netdev_tc++;
1039 qmap = FIELD_PREP(ICE_AQ_VSI_TC_Q_OFFSET_M, offset);
1040 qmap |= FIELD_PREP(ICE_AQ_VSI_TC_Q_NUM_M, pow);
1041 offset += num_rxq_per_tc;
1042 tx_count += num_txq_per_tc;
1043 ctxt->info.tc_mapping[i] = cpu_to_le16(qmap);
1046 /* if offset is non-zero, means it is calculated correctly based on
1047 * enabled TCs for a given VSI otherwise qcount_rx will always
1048 * be correct and non-zero because it is based off - VSI's
1049 * allocated Rx queues which is at least 1 (hence qcount_tx will be
1055 rx_count = num_rxq_per_tc;
1057 if (rx_count > vsi->alloc_rxq) {
1058 dev_err(ice_pf_to_dev(vsi->back), "Trying to use more Rx queues (%u), than were allocated (%u)!\n",
1059 rx_count, vsi->alloc_rxq);
1063 if (tx_count > vsi->alloc_txq) {
1064 dev_err(ice_pf_to_dev(vsi->back), "Trying to use more Tx queues (%u), than were allocated (%u)!\n",
1065 tx_count, vsi->alloc_txq);
1069 vsi->num_txq = tx_count;
1070 vsi->num_rxq = rx_count;
1072 if (vsi->type == ICE_VSI_VF && vsi->num_txq != vsi->num_rxq) {
1073 dev_dbg(ice_pf_to_dev(vsi->back), "VF VSI should have same number of Tx and Rx queues. Hence making them equal\n");
1074 /* since there is a chance that num_rxq could have been changed
1075 * in the above for loop, make num_txq equal to num_rxq.
1077 vsi->num_txq = vsi->num_rxq;
1080 /* Rx queue mapping */
1081 ctxt->info.mapping_flags |= cpu_to_le16(ICE_AQ_VSI_Q_MAP_CONTIG);
1082 /* q_mapping buffer holds the info for the first queue allocated for
1083 * this VSI in the PF space and also the number of queues associated
1086 ctxt->info.q_mapping[0] = cpu_to_le16(vsi->rxq_map[0]);
1087 ctxt->info.q_mapping[1] = cpu_to_le16(vsi->num_rxq);
1093 * ice_set_fd_vsi_ctx - Set FD VSI context before adding a VSI
1094 * @ctxt: the VSI context being set
1095 * @vsi: the VSI being configured
1097 static void ice_set_fd_vsi_ctx(struct ice_vsi_ctx *ctxt, struct ice_vsi *vsi)
1099 u8 dflt_q_group, dflt_q_prio;
1100 u16 dflt_q, report_q, val;
1102 if (vsi->type != ICE_VSI_PF && vsi->type != ICE_VSI_CTRL &&
1103 vsi->type != ICE_VSI_VF && vsi->type != ICE_VSI_CHNL)
1106 val = ICE_AQ_VSI_PROP_FLOW_DIR_VALID;
1107 ctxt->info.valid_sections |= cpu_to_le16(val);
1113 /* enable flow director filtering/programming */
1114 val = ICE_AQ_VSI_FD_ENABLE | ICE_AQ_VSI_FD_PROG_ENABLE;
1115 ctxt->info.fd_options = cpu_to_le16(val);
1116 /* max of allocated flow director filters */
1117 ctxt->info.max_fd_fltr_dedicated =
1118 cpu_to_le16(vsi->num_gfltr);
1119 /* max of shared flow director filters any VSI may program */
1120 ctxt->info.max_fd_fltr_shared =
1121 cpu_to_le16(vsi->num_bfltr);
1122 /* default queue index within the VSI of the default FD */
1123 val = FIELD_PREP(ICE_AQ_VSI_FD_DEF_Q_M, dflt_q);
1124 /* target queue or queue group to the FD filter */
1125 val |= FIELD_PREP(ICE_AQ_VSI_FD_DEF_GRP_M, dflt_q_group);
1126 ctxt->info.fd_def_q = cpu_to_le16(val);
1127 /* queue index on which FD filter completion is reported */
1128 val = FIELD_PREP(ICE_AQ_VSI_FD_REPORT_Q_M, report_q);
1129 /* priority of the default qindex action */
1130 val |= FIELD_PREP(ICE_AQ_VSI_FD_DEF_PRIORITY_M, dflt_q_prio);
1131 ctxt->info.fd_report_opt = cpu_to_le16(val);
1135 * ice_set_rss_vsi_ctx - Set RSS VSI context before adding a VSI
1136 * @ctxt: the VSI context being set
1137 * @vsi: the VSI being configured
1139 static void ice_set_rss_vsi_ctx(struct ice_vsi_ctx *ctxt, struct ice_vsi *vsi)
1141 u8 lut_type, hash_type;
1146 dev = ice_pf_to_dev(pf);
1148 switch (vsi->type) {
1151 /* PF VSI will inherit RSS instance of PF */
1152 lut_type = ICE_AQ_VSI_Q_OPT_RSS_LUT_PF;
1156 /* VF VSI will gets a small RSS table which is a VSI LUT type */
1157 lut_type = ICE_AQ_VSI_Q_OPT_RSS_LUT_VSI;
1160 dev_dbg(dev, "Unsupported VSI type %s\n",
1161 ice_vsi_type_str(vsi->type));
1165 hash_type = ICE_AQ_VSI_Q_OPT_RSS_HASH_TPLZ;
1166 vsi->rss_hfunc = hash_type;
1168 ctxt->info.q_opt_rss =
1169 FIELD_PREP(ICE_AQ_VSI_Q_OPT_RSS_LUT_M, lut_type) |
1170 FIELD_PREP(ICE_AQ_VSI_Q_OPT_RSS_HASH_M, hash_type);
1174 ice_chnl_vsi_setup_q_map(struct ice_vsi *vsi, struct ice_vsi_ctx *ctxt)
1176 struct ice_pf *pf = vsi->back;
1181 qcount = min_t(int, vsi->num_rxq, pf->num_lan_msix);
1183 pow = order_base_2(qcount);
1184 qmap = FIELD_PREP(ICE_AQ_VSI_TC_Q_OFFSET_M, offset);
1185 qmap |= FIELD_PREP(ICE_AQ_VSI_TC_Q_NUM_M, pow);
1187 ctxt->info.tc_mapping[0] = cpu_to_le16(qmap);
1188 ctxt->info.mapping_flags |= cpu_to_le16(ICE_AQ_VSI_Q_MAP_CONTIG);
1189 ctxt->info.q_mapping[0] = cpu_to_le16(vsi->next_base_q);
1190 ctxt->info.q_mapping[1] = cpu_to_le16(qcount);
1194 * ice_vsi_is_vlan_pruning_ena - check if VLAN pruning is enabled or not
1195 * @vsi: VSI to check whether or not VLAN pruning is enabled.
1197 * returns true if Rx VLAN pruning is enabled and false otherwise.
1199 static bool ice_vsi_is_vlan_pruning_ena(struct ice_vsi *vsi)
1201 return vsi->info.sw_flags2 & ICE_AQ_VSI_SW_FLAG_RX_VLAN_PRUNE_ENA;
1205 * ice_vsi_init - Create and initialize a VSI
1206 * @vsi: the VSI being configured
1207 * @vsi_flags: VSI configuration flags
1209 * Set ICE_FLAG_VSI_INIT to initialize a new VSI context, clear it to
1210 * reconfigure an existing context.
1212 * This initializes a VSI context depending on the VSI type to be added and
1213 * passes it down to the add_vsi aq command to create a new VSI.
1215 static int ice_vsi_init(struct ice_vsi *vsi, u32 vsi_flags)
1217 struct ice_pf *pf = vsi->back;
1218 struct ice_hw *hw = &pf->hw;
1219 struct ice_vsi_ctx *ctxt;
1223 dev = ice_pf_to_dev(pf);
1224 ctxt = kzalloc(sizeof(*ctxt), GFP_KERNEL);
1228 switch (vsi->type) {
1232 ctxt->flags = ICE_AQ_VSI_TYPE_PF;
1236 ctxt->flags = ICE_AQ_VSI_TYPE_VMDQ2;
1239 ctxt->flags = ICE_AQ_VSI_TYPE_VF;
1240 /* VF number here is the absolute VF number (0-255) */
1241 ctxt->vf_num = vsi->vf->vf_id + hw->func_caps.vf_base_id;
1248 /* Handle VLAN pruning for channel VSI if main VSI has VLAN
1251 if (vsi->type == ICE_VSI_CHNL) {
1252 struct ice_vsi *main_vsi;
1254 main_vsi = ice_get_main_vsi(pf);
1255 if (main_vsi && ice_vsi_is_vlan_pruning_ena(main_vsi))
1256 ctxt->info.sw_flags2 |=
1257 ICE_AQ_VSI_SW_FLAG_RX_VLAN_PRUNE_ENA;
1259 ctxt->info.sw_flags2 &=
1260 ~ICE_AQ_VSI_SW_FLAG_RX_VLAN_PRUNE_ENA;
1263 ice_set_dflt_vsi_ctx(hw, ctxt);
1264 if (test_bit(ICE_FLAG_FD_ENA, pf->flags))
1265 ice_set_fd_vsi_ctx(ctxt, vsi);
1266 /* if the switch is in VEB mode, allow VSI loopback */
1267 if (vsi->vsw->bridge_mode == BRIDGE_MODE_VEB)
1268 ctxt->info.sw_flags |= ICE_AQ_VSI_SW_FLAG_ALLOW_LB;
1270 /* Set LUT type and HASH type if RSS is enabled */
1271 if (test_bit(ICE_FLAG_RSS_ENA, pf->flags) &&
1272 vsi->type != ICE_VSI_CTRL) {
1273 ice_set_rss_vsi_ctx(ctxt, vsi);
1274 /* if updating VSI context, make sure to set valid_section:
1275 * to indicate which section of VSI context being updated
1277 if (!(vsi_flags & ICE_VSI_FLAG_INIT))
1278 ctxt->info.valid_sections |=
1279 cpu_to_le16(ICE_AQ_VSI_PROP_Q_OPT_VALID);
1282 ctxt->info.sw_id = vsi->port_info->sw_id;
1283 if (vsi->type == ICE_VSI_CHNL) {
1284 ice_chnl_vsi_setup_q_map(vsi, ctxt);
1286 ret = ice_vsi_setup_q_map(vsi, ctxt);
1290 if (!(vsi_flags & ICE_VSI_FLAG_INIT))
1291 /* means VSI being updated */
1292 /* must to indicate which section of VSI context are
1295 ctxt->info.valid_sections |=
1296 cpu_to_le16(ICE_AQ_VSI_PROP_RXQ_MAP_VALID);
1299 /* Allow control frames out of main VSI */
1300 if (vsi->type == ICE_VSI_PF) {
1301 ctxt->info.sec_flags |= ICE_AQ_VSI_SEC_FLAG_ALLOW_DEST_OVRD;
1302 ctxt->info.valid_sections |=
1303 cpu_to_le16(ICE_AQ_VSI_PROP_SECURITY_VALID);
1306 if (vsi_flags & ICE_VSI_FLAG_INIT) {
1307 ret = ice_add_vsi(hw, vsi->idx, ctxt, NULL);
1309 dev_err(dev, "Add VSI failed, err %d\n", ret);
1314 ret = ice_update_vsi(hw, vsi->idx, ctxt, NULL);
1316 dev_err(dev, "Update VSI failed, err %d\n", ret);
1322 /* keep context for update VSI operations */
1323 vsi->info = ctxt->info;
1325 /* record VSI number returned */
1326 vsi->vsi_num = ctxt->vsi_num;
1334 * ice_vsi_clear_rings - Deallocates the Tx and Rx rings for VSI
1335 * @vsi: the VSI having rings deallocated
1337 static void ice_vsi_clear_rings(struct ice_vsi *vsi)
1341 /* Avoid stale references by clearing map from vector to ring */
1342 if (vsi->q_vectors) {
1343 ice_for_each_q_vector(vsi, i) {
1344 struct ice_q_vector *q_vector = vsi->q_vectors[i];
1347 q_vector->tx.tx_ring = NULL;
1348 q_vector->rx.rx_ring = NULL;
1353 if (vsi->tx_rings) {
1354 ice_for_each_alloc_txq(vsi, i) {
1355 if (vsi->tx_rings[i]) {
1356 kfree_rcu(vsi->tx_rings[i], rcu);
1357 WRITE_ONCE(vsi->tx_rings[i], NULL);
1361 if (vsi->rx_rings) {
1362 ice_for_each_alloc_rxq(vsi, i) {
1363 if (vsi->rx_rings[i]) {
1364 kfree_rcu(vsi->rx_rings[i], rcu);
1365 WRITE_ONCE(vsi->rx_rings[i], NULL);
1372 * ice_vsi_alloc_rings - Allocates Tx and Rx rings for the VSI
1373 * @vsi: VSI which is having rings allocated
1375 static int ice_vsi_alloc_rings(struct ice_vsi *vsi)
1377 bool dvm_ena = ice_is_dvm_ena(&vsi->back->hw);
1378 struct ice_pf *pf = vsi->back;
1382 dev = ice_pf_to_dev(pf);
1383 /* Allocate Tx rings */
1384 ice_for_each_alloc_txq(vsi, i) {
1385 struct ice_tx_ring *ring;
1387 /* allocate with kzalloc(), free with kfree_rcu() */
1388 ring = kzalloc(sizeof(*ring), GFP_KERNEL);
1394 ring->reg_idx = vsi->txq_map[i];
1396 ring->tx_tstamps = &pf->ptp.port.tx;
1398 ring->count = vsi->num_tx_desc;
1399 ring->txq_teid = ICE_INVAL_TEID;
1401 ring->flags |= ICE_TX_FLAGS_RING_VLAN_L2TAG2;
1403 ring->flags |= ICE_TX_FLAGS_RING_VLAN_L2TAG1;
1404 WRITE_ONCE(vsi->tx_rings[i], ring);
1407 /* Allocate Rx rings */
1408 ice_for_each_alloc_rxq(vsi, i) {
1409 struct ice_rx_ring *ring;
1411 /* allocate with kzalloc(), free with kfree_rcu() */
1412 ring = kzalloc(sizeof(*ring), GFP_KERNEL);
1417 ring->reg_idx = vsi->rxq_map[i];
1419 ring->netdev = vsi->netdev;
1421 ring->count = vsi->num_rx_desc;
1422 ring->cached_phctime = pf->ptp.cached_phc_time;
1423 WRITE_ONCE(vsi->rx_rings[i], ring);
1429 ice_vsi_clear_rings(vsi);
1434 * ice_vsi_manage_rss_lut - disable/enable RSS
1435 * @vsi: the VSI being changed
1436 * @ena: boolean value indicating if this is an enable or disable request
1438 * In the event of disable request for RSS, this function will zero out RSS
1439 * LUT, while in the event of enable request for RSS, it will reconfigure RSS
1442 void ice_vsi_manage_rss_lut(struct ice_vsi *vsi, bool ena)
1446 lut = kzalloc(vsi->rss_table_size, GFP_KERNEL);
1451 if (vsi->rss_lut_user)
1452 memcpy(lut, vsi->rss_lut_user, vsi->rss_table_size);
1454 ice_fill_rss_lut(lut, vsi->rss_table_size,
1458 ice_set_rss_lut(vsi, lut, vsi->rss_table_size);
1463 * ice_vsi_cfg_crc_strip - Configure CRC stripping for a VSI
1464 * @vsi: VSI to be configured
1465 * @disable: set to true to have FCS / CRC in the frame data
1467 void ice_vsi_cfg_crc_strip(struct ice_vsi *vsi, bool disable)
1471 ice_for_each_rxq(vsi, i)
1473 vsi->rx_rings[i]->flags |= ICE_RX_FLAGS_CRC_STRIP_DIS;
1475 vsi->rx_rings[i]->flags &= ~ICE_RX_FLAGS_CRC_STRIP_DIS;
1479 * ice_vsi_cfg_rss_lut_key - Configure RSS params for a VSI
1480 * @vsi: VSI to be configured
1482 int ice_vsi_cfg_rss_lut_key(struct ice_vsi *vsi)
1484 struct ice_pf *pf = vsi->back;
1489 dev = ice_pf_to_dev(pf);
1490 if (vsi->type == ICE_VSI_PF && vsi->ch_rss_size &&
1491 (test_bit(ICE_FLAG_TC_MQPRIO, pf->flags))) {
1492 vsi->rss_size = min_t(u16, vsi->rss_size, vsi->ch_rss_size);
1494 vsi->rss_size = min_t(u16, vsi->rss_size, vsi->num_rxq);
1496 /* If orig_rss_size is valid and it is less than determined
1497 * main VSI's rss_size, update main VSI's rss_size to be
1498 * orig_rss_size so that when tc-qdisc is deleted, main VSI
1499 * RSS table gets programmed to be correct (whatever it was
1500 * to begin with (prior to setup-tc for ADQ config)
1502 if (vsi->orig_rss_size && vsi->rss_size < vsi->orig_rss_size &&
1503 vsi->orig_rss_size <= vsi->num_rxq) {
1504 vsi->rss_size = vsi->orig_rss_size;
1505 /* now orig_rss_size is used, reset it to zero */
1506 vsi->orig_rss_size = 0;
1510 lut = kzalloc(vsi->rss_table_size, GFP_KERNEL);
1514 if (vsi->rss_lut_user)
1515 memcpy(lut, vsi->rss_lut_user, vsi->rss_table_size);
1517 ice_fill_rss_lut(lut, vsi->rss_table_size, vsi->rss_size);
1519 err = ice_set_rss_lut(vsi, lut, vsi->rss_table_size);
1521 dev_err(dev, "set_rss_lut failed, error %d\n", err);
1522 goto ice_vsi_cfg_rss_exit;
1525 key = kzalloc(ICE_GET_SET_RSS_KEY_EXTEND_KEY_SIZE, GFP_KERNEL);
1528 goto ice_vsi_cfg_rss_exit;
1531 if (vsi->rss_hkey_user)
1532 memcpy(key, vsi->rss_hkey_user, ICE_GET_SET_RSS_KEY_EXTEND_KEY_SIZE);
1534 netdev_rss_key_fill((void *)key, ICE_GET_SET_RSS_KEY_EXTEND_KEY_SIZE);
1536 err = ice_set_rss_key(vsi, key);
1538 dev_err(dev, "set_rss_key failed, error %d\n", err);
1541 ice_vsi_cfg_rss_exit:
1547 * ice_vsi_set_vf_rss_flow_fld - Sets VF VSI RSS input set for different flows
1548 * @vsi: VSI to be configured
1550 * This function will only be called during the VF VSI setup. Upon successful
1551 * completion of package download, this function will configure default RSS
1552 * input sets for VF VSI.
1554 static void ice_vsi_set_vf_rss_flow_fld(struct ice_vsi *vsi)
1556 struct ice_pf *pf = vsi->back;
1560 dev = ice_pf_to_dev(pf);
1561 if (ice_is_safe_mode(pf)) {
1562 dev_dbg(dev, "Advanced RSS disabled. Package download failed, vsi num = %d\n",
1567 status = ice_add_avf_rss_cfg(&pf->hw, vsi, ICE_DEFAULT_RSS_HENA);
1569 dev_dbg(dev, "ice_add_avf_rss_cfg failed for vsi = %d, error = %d\n",
1570 vsi->vsi_num, status);
1573 static const struct ice_rss_hash_cfg default_rss_cfgs[] = {
1574 /* configure RSS for IPv4 with input set IP src/dst */
1575 {ICE_FLOW_SEG_HDR_IPV4, ICE_FLOW_HASH_IPV4, ICE_RSS_ANY_HEADERS, false},
1576 /* configure RSS for IPv6 with input set IPv6 src/dst */
1577 {ICE_FLOW_SEG_HDR_IPV6, ICE_FLOW_HASH_IPV6, ICE_RSS_ANY_HEADERS, false},
1578 /* configure RSS for tcp4 with input set IP src/dst, TCP src/dst */
1579 {ICE_FLOW_SEG_HDR_TCP | ICE_FLOW_SEG_HDR_IPV4,
1580 ICE_HASH_TCP_IPV4, ICE_RSS_ANY_HEADERS, false},
1581 /* configure RSS for udp4 with input set IP src/dst, UDP src/dst */
1582 {ICE_FLOW_SEG_HDR_UDP | ICE_FLOW_SEG_HDR_IPV4,
1583 ICE_HASH_UDP_IPV4, ICE_RSS_ANY_HEADERS, false},
1584 /* configure RSS for sctp4 with input set IP src/dst - only support
1585 * RSS on SCTPv4 on outer headers (non-tunneled)
1587 {ICE_FLOW_SEG_HDR_SCTP | ICE_FLOW_SEG_HDR_IPV4,
1588 ICE_HASH_SCTP_IPV4, ICE_RSS_OUTER_HEADERS, false},
1589 /* configure RSS for gtpc4 with input set IPv4 src/dst */
1590 {ICE_FLOW_SEG_HDR_GTPC | ICE_FLOW_SEG_HDR_IPV4,
1591 ICE_FLOW_HASH_IPV4, ICE_RSS_OUTER_HEADERS, false},
1592 /* configure RSS for gtpc4t with input set IPv4 src/dst */
1593 {ICE_FLOW_SEG_HDR_GTPC_TEID | ICE_FLOW_SEG_HDR_IPV4,
1594 ICE_FLOW_HASH_GTP_C_IPV4_TEID, ICE_RSS_OUTER_HEADERS, false},
1595 /* configure RSS for gtpu4 with input set IPv4 src/dst */
1596 {ICE_FLOW_SEG_HDR_GTPU_IP | ICE_FLOW_SEG_HDR_IPV4,
1597 ICE_FLOW_HASH_GTP_U_IPV4_TEID, ICE_RSS_OUTER_HEADERS, false},
1598 /* configure RSS for gtpu4e with input set IPv4 src/dst */
1599 {ICE_FLOW_SEG_HDR_GTPU_EH | ICE_FLOW_SEG_HDR_IPV4,
1600 ICE_FLOW_HASH_GTP_U_IPV4_EH, ICE_RSS_OUTER_HEADERS, false},
1601 /* configure RSS for gtpu4u with input set IPv4 src/dst */
1602 { ICE_FLOW_SEG_HDR_GTPU_UP | ICE_FLOW_SEG_HDR_IPV4,
1603 ICE_FLOW_HASH_GTP_U_IPV4_UP, ICE_RSS_OUTER_HEADERS, false},
1604 /* configure RSS for gtpu4d with input set IPv4 src/dst */
1605 {ICE_FLOW_SEG_HDR_GTPU_DWN | ICE_FLOW_SEG_HDR_IPV4,
1606 ICE_FLOW_HASH_GTP_U_IPV4_DWN, ICE_RSS_OUTER_HEADERS, false},
1608 /* configure RSS for tcp6 with input set IPv6 src/dst, TCP src/dst */
1609 {ICE_FLOW_SEG_HDR_TCP | ICE_FLOW_SEG_HDR_IPV6,
1610 ICE_HASH_TCP_IPV6, ICE_RSS_ANY_HEADERS, false},
1611 /* configure RSS for udp6 with input set IPv6 src/dst, UDP src/dst */
1612 {ICE_FLOW_SEG_HDR_UDP | ICE_FLOW_SEG_HDR_IPV6,
1613 ICE_HASH_UDP_IPV6, ICE_RSS_ANY_HEADERS, false},
1614 /* configure RSS for sctp6 with input set IPv6 src/dst - only support
1615 * RSS on SCTPv6 on outer headers (non-tunneled)
1617 {ICE_FLOW_SEG_HDR_SCTP | ICE_FLOW_SEG_HDR_IPV6,
1618 ICE_HASH_SCTP_IPV6, ICE_RSS_OUTER_HEADERS, false},
1619 /* configure RSS for IPSEC ESP SPI with input set MAC_IPV4_SPI */
1620 {ICE_FLOW_SEG_HDR_ESP,
1621 ICE_FLOW_HASH_ESP_SPI, ICE_RSS_OUTER_HEADERS, false},
1622 /* configure RSS for gtpc6 with input set IPv6 src/dst */
1623 {ICE_FLOW_SEG_HDR_GTPC | ICE_FLOW_SEG_HDR_IPV6,
1624 ICE_FLOW_HASH_IPV6, ICE_RSS_OUTER_HEADERS, false},
1625 /* configure RSS for gtpc6t with input set IPv6 src/dst */
1626 {ICE_FLOW_SEG_HDR_GTPC_TEID | ICE_FLOW_SEG_HDR_IPV6,
1627 ICE_FLOW_HASH_GTP_C_IPV6_TEID, ICE_RSS_OUTER_HEADERS, false},
1628 /* configure RSS for gtpu6 with input set IPv6 src/dst */
1629 {ICE_FLOW_SEG_HDR_GTPU_IP | ICE_FLOW_SEG_HDR_IPV6,
1630 ICE_FLOW_HASH_GTP_U_IPV6_TEID, ICE_RSS_OUTER_HEADERS, false},
1631 /* configure RSS for gtpu6e with input set IPv6 src/dst */
1632 {ICE_FLOW_SEG_HDR_GTPU_EH | ICE_FLOW_SEG_HDR_IPV6,
1633 ICE_FLOW_HASH_GTP_U_IPV6_EH, ICE_RSS_OUTER_HEADERS, false},
1634 /* configure RSS for gtpu6u with input set IPv6 src/dst */
1635 { ICE_FLOW_SEG_HDR_GTPU_UP | ICE_FLOW_SEG_HDR_IPV6,
1636 ICE_FLOW_HASH_GTP_U_IPV6_UP, ICE_RSS_OUTER_HEADERS, false},
1637 /* configure RSS for gtpu6d with input set IPv6 src/dst */
1638 {ICE_FLOW_SEG_HDR_GTPU_DWN | ICE_FLOW_SEG_HDR_IPV6,
1639 ICE_FLOW_HASH_GTP_U_IPV6_DWN, ICE_RSS_OUTER_HEADERS, false},
1643 * ice_vsi_set_rss_flow_fld - Sets RSS input set for different flows
1644 * @vsi: VSI to be configured
1646 * This function will only be called after successful download package call
1647 * during initialization of PF. Since the downloaded package will erase the
1648 * RSS section, this function will configure RSS input sets for different
1649 * flow types. The last profile added has the highest priority, therefore 2
1650 * tuple profiles (i.e. IPv4 src/dst) are added before 4 tuple profiles
1651 * (i.e. IPv4 src/dst TCP src/dst port).
1653 static void ice_vsi_set_rss_flow_fld(struct ice_vsi *vsi)
1655 u16 vsi_num = vsi->vsi_num;
1656 struct ice_pf *pf = vsi->back;
1657 struct ice_hw *hw = &pf->hw;
1662 dev = ice_pf_to_dev(pf);
1663 if (ice_is_safe_mode(pf)) {
1664 dev_dbg(dev, "Advanced RSS disabled. Package download failed, vsi num = %d\n",
1668 for (i = 0; i < ARRAY_SIZE(default_rss_cfgs); i++) {
1669 const struct ice_rss_hash_cfg *cfg = &default_rss_cfgs[i];
1671 status = ice_add_rss_cfg(hw, vsi, cfg);
1673 dev_dbg(dev, "ice_add_rss_cfg failed, addl_hdrs = %x, hash_flds = %llx, hdr_type = %d, symm = %d\n",
1674 cfg->addl_hdrs, cfg->hash_flds,
1675 cfg->hdr_type, cfg->symm);
1680 * ice_pf_state_is_nominal - checks the PF for nominal state
1681 * @pf: pointer to PF to check
1683 * Check the PF's state for a collection of bits that would indicate
1684 * the PF is in a state that would inhibit normal operation for
1685 * driver functionality.
1687 * Returns true if PF is in a nominal state, false otherwise
1689 bool ice_pf_state_is_nominal(struct ice_pf *pf)
1691 DECLARE_BITMAP(check_bits, ICE_STATE_NBITS) = { 0 };
1696 bitmap_set(check_bits, 0, ICE_STATE_NOMINAL_CHECK_BITS);
1697 if (bitmap_intersects(pf->state, check_bits, ICE_STATE_NBITS))
1704 * ice_update_eth_stats - Update VSI-specific ethernet statistics counters
1705 * @vsi: the VSI to be updated
1707 void ice_update_eth_stats(struct ice_vsi *vsi)
1709 struct ice_eth_stats *prev_es, *cur_es;
1710 struct ice_hw *hw = &vsi->back->hw;
1711 struct ice_pf *pf = vsi->back;
1712 u16 vsi_num = vsi->vsi_num; /* HW absolute index of a VSI */
1714 prev_es = &vsi->eth_stats_prev;
1715 cur_es = &vsi->eth_stats;
1717 if (ice_is_reset_in_progress(pf->state))
1718 vsi->stat_offsets_loaded = false;
1720 ice_stat_update40(hw, GLV_GORCL(vsi_num), vsi->stat_offsets_loaded,
1721 &prev_es->rx_bytes, &cur_es->rx_bytes);
1723 ice_stat_update40(hw, GLV_UPRCL(vsi_num), vsi->stat_offsets_loaded,
1724 &prev_es->rx_unicast, &cur_es->rx_unicast);
1726 ice_stat_update40(hw, GLV_MPRCL(vsi_num), vsi->stat_offsets_loaded,
1727 &prev_es->rx_multicast, &cur_es->rx_multicast);
1729 ice_stat_update40(hw, GLV_BPRCL(vsi_num), vsi->stat_offsets_loaded,
1730 &prev_es->rx_broadcast, &cur_es->rx_broadcast);
1732 ice_stat_update32(hw, GLV_RDPC(vsi_num), vsi->stat_offsets_loaded,
1733 &prev_es->rx_discards, &cur_es->rx_discards);
1735 ice_stat_update40(hw, GLV_GOTCL(vsi_num), vsi->stat_offsets_loaded,
1736 &prev_es->tx_bytes, &cur_es->tx_bytes);
1738 ice_stat_update40(hw, GLV_UPTCL(vsi_num), vsi->stat_offsets_loaded,
1739 &prev_es->tx_unicast, &cur_es->tx_unicast);
1741 ice_stat_update40(hw, GLV_MPTCL(vsi_num), vsi->stat_offsets_loaded,
1742 &prev_es->tx_multicast, &cur_es->tx_multicast);
1744 ice_stat_update40(hw, GLV_BPTCL(vsi_num), vsi->stat_offsets_loaded,
1745 &prev_es->tx_broadcast, &cur_es->tx_broadcast);
1747 ice_stat_update32(hw, GLV_TEPC(vsi_num), vsi->stat_offsets_loaded,
1748 &prev_es->tx_errors, &cur_es->tx_errors);
1750 vsi->stat_offsets_loaded = true;
1754 * ice_write_qrxflxp_cntxt - write/configure QRXFLXP_CNTXT register
1756 * @pf_q: index of the Rx queue in the PF's queue space
1757 * @rxdid: flexible descriptor RXDID
1758 * @prio: priority for the RXDID for this queue
1759 * @ena_ts: true to enable timestamp and false to disable timestamp
1762 ice_write_qrxflxp_cntxt(struct ice_hw *hw, u16 pf_q, u32 rxdid, u32 prio,
1765 int regval = rd32(hw, QRXFLXP_CNTXT(pf_q));
1767 /* clear any previous values */
1768 regval &= ~(QRXFLXP_CNTXT_RXDID_IDX_M |
1769 QRXFLXP_CNTXT_RXDID_PRIO_M |
1770 QRXFLXP_CNTXT_TS_M);
1772 regval |= FIELD_PREP(QRXFLXP_CNTXT_RXDID_IDX_M, rxdid);
1773 regval |= FIELD_PREP(QRXFLXP_CNTXT_RXDID_PRIO_M, prio);
1776 /* Enable TimeSync on this queue */
1777 regval |= QRXFLXP_CNTXT_TS_M;
1779 wr32(hw, QRXFLXP_CNTXT(pf_q), regval);
1783 * ice_intrl_usec_to_reg - convert interrupt rate limit to register value
1784 * @intrl: interrupt rate limit in usecs
1785 * @gran: interrupt rate limit granularity in usecs
1787 * This function converts a decimal interrupt rate limit in usecs to the format
1788 * expected by firmware.
1790 static u32 ice_intrl_usec_to_reg(u8 intrl, u8 gran)
1792 u32 val = intrl / gran;
1795 return val | GLINT_RATE_INTRL_ENA_M;
1800 * ice_write_intrl - write throttle rate limit to interrupt specific register
1801 * @q_vector: pointer to interrupt specific structure
1802 * @intrl: throttle rate limit in microseconds to write
1804 void ice_write_intrl(struct ice_q_vector *q_vector, u8 intrl)
1806 struct ice_hw *hw = &q_vector->vsi->back->hw;
1808 wr32(hw, GLINT_RATE(q_vector->reg_idx),
1809 ice_intrl_usec_to_reg(intrl, ICE_INTRL_GRAN_ABOVE_25));
1812 static struct ice_q_vector *ice_pull_qvec_from_rc(struct ice_ring_container *rc)
1815 case ICE_RX_CONTAINER:
1817 return rc->rx_ring->q_vector;
1819 case ICE_TX_CONTAINER:
1821 return rc->tx_ring->q_vector;
1831 * __ice_write_itr - write throttle rate to register
1832 * @q_vector: pointer to interrupt data structure
1833 * @rc: pointer to ring container
1834 * @itr: throttle rate in microseconds to write
1836 static void __ice_write_itr(struct ice_q_vector *q_vector,
1837 struct ice_ring_container *rc, u16 itr)
1839 struct ice_hw *hw = &q_vector->vsi->back->hw;
1841 wr32(hw, GLINT_ITR(rc->itr_idx, q_vector->reg_idx),
1842 ITR_REG_ALIGN(itr) >> ICE_ITR_GRAN_S);
1846 * ice_write_itr - write throttle rate to queue specific register
1847 * @rc: pointer to ring container
1848 * @itr: throttle rate in microseconds to write
1850 void ice_write_itr(struct ice_ring_container *rc, u16 itr)
1852 struct ice_q_vector *q_vector;
1854 q_vector = ice_pull_qvec_from_rc(rc);
1858 __ice_write_itr(q_vector, rc, itr);
1862 * ice_set_q_vector_intrl - set up interrupt rate limiting
1863 * @q_vector: the vector to be configured
1865 * Interrupt rate limiting is local to the vector, not per-queue so we must
1866 * detect if either ring container has dynamic moderation enabled to decide
1867 * what to set the interrupt rate limit to via INTRL settings. In the case that
1868 * dynamic moderation is disabled on both, write the value with the cached
1869 * setting to make sure INTRL register matches the user visible value.
1871 void ice_set_q_vector_intrl(struct ice_q_vector *q_vector)
1873 if (ITR_IS_DYNAMIC(&q_vector->tx) || ITR_IS_DYNAMIC(&q_vector->rx)) {
1874 /* in the case of dynamic enabled, cap each vector to no more
1875 * than (4 us) 250,000 ints/sec, which allows low latency
1876 * but still less than 500,000 interrupts per second, which
1877 * reduces CPU a bit in the case of the lowest latency
1878 * setting. The 4 here is a value in microseconds.
1880 ice_write_intrl(q_vector, 4);
1882 ice_write_intrl(q_vector, q_vector->intrl);
1887 * ice_vsi_cfg_msix - MSIX mode Interrupt Config in the HW
1888 * @vsi: the VSI being configured
1890 * This configures MSIX mode interrupts for the PF VSI, and should not be used
1893 void ice_vsi_cfg_msix(struct ice_vsi *vsi)
1895 struct ice_pf *pf = vsi->back;
1896 struct ice_hw *hw = &pf->hw;
1897 u16 txq = 0, rxq = 0;
1900 ice_for_each_q_vector(vsi, i) {
1901 struct ice_q_vector *q_vector = vsi->q_vectors[i];
1902 u16 reg_idx = q_vector->reg_idx;
1904 ice_cfg_itr(hw, q_vector);
1906 /* Both Transmit Queue Interrupt Cause Control register
1907 * and Receive Queue Interrupt Cause control register
1908 * expects MSIX_INDX field to be the vector index
1909 * within the function space and not the absolute
1910 * vector index across PF or across device.
1911 * For SR-IOV VF VSIs queue vector index always starts
1912 * with 1 since first vector index(0) is used for OICR
1913 * in VF space. Since VMDq and other PF VSIs are within
1914 * the PF function space, use the vector index that is
1915 * tracked for this PF.
1917 for (q = 0; q < q_vector->num_ring_tx; q++) {
1918 ice_cfg_txq_interrupt(vsi, txq, reg_idx,
1919 q_vector->tx.itr_idx);
1923 for (q = 0; q < q_vector->num_ring_rx; q++) {
1924 ice_cfg_rxq_interrupt(vsi, rxq, reg_idx,
1925 q_vector->rx.itr_idx);
1932 * ice_vsi_start_all_rx_rings - start/enable all of a VSI's Rx rings
1933 * @vsi: the VSI whose rings are to be enabled
1935 * Returns 0 on success and a negative value on error
1937 int ice_vsi_start_all_rx_rings(struct ice_vsi *vsi)
1939 return ice_vsi_ctrl_all_rx_rings(vsi, true);
1943 * ice_vsi_stop_all_rx_rings - stop/disable all of a VSI's Rx rings
1944 * @vsi: the VSI whose rings are to be disabled
1946 * Returns 0 on success and a negative value on error
1948 int ice_vsi_stop_all_rx_rings(struct ice_vsi *vsi)
1950 return ice_vsi_ctrl_all_rx_rings(vsi, false);
1954 * ice_vsi_stop_tx_rings - Disable Tx rings
1955 * @vsi: the VSI being configured
1956 * @rst_src: reset source
1957 * @rel_vmvf_num: Relative ID of VF/VM
1958 * @rings: Tx ring array to be stopped
1959 * @count: number of Tx ring array elements
1962 ice_vsi_stop_tx_rings(struct ice_vsi *vsi, enum ice_disq_rst_src rst_src,
1963 u16 rel_vmvf_num, struct ice_tx_ring **rings, u16 count)
1967 if (vsi->num_txq > ICE_LAN_TXQ_MAX_QDIS)
1970 for (q_idx = 0; q_idx < count; q_idx++) {
1971 struct ice_txq_meta txq_meta = { };
1974 if (!rings || !rings[q_idx])
1977 ice_fill_txq_meta(vsi, rings[q_idx], &txq_meta);
1978 status = ice_vsi_stop_tx_ring(vsi, rst_src, rel_vmvf_num,
1979 rings[q_idx], &txq_meta);
1989 * ice_vsi_stop_lan_tx_rings - Disable LAN Tx rings
1990 * @vsi: the VSI being configured
1991 * @rst_src: reset source
1992 * @rel_vmvf_num: Relative ID of VF/VM
1995 ice_vsi_stop_lan_tx_rings(struct ice_vsi *vsi, enum ice_disq_rst_src rst_src,
1998 return ice_vsi_stop_tx_rings(vsi, rst_src, rel_vmvf_num, vsi->tx_rings, vsi->num_txq);
2002 * ice_vsi_stop_xdp_tx_rings - Disable XDP Tx rings
2003 * @vsi: the VSI being configured
2005 int ice_vsi_stop_xdp_tx_rings(struct ice_vsi *vsi)
2007 return ice_vsi_stop_tx_rings(vsi, ICE_NO_RESET, 0, vsi->xdp_rings, vsi->num_xdp_txq);
2011 * ice_vsi_is_rx_queue_active
2012 * @vsi: the VSI being configured
2014 * Return true if at least one queue is active.
2016 bool ice_vsi_is_rx_queue_active(struct ice_vsi *vsi)
2018 struct ice_pf *pf = vsi->back;
2019 struct ice_hw *hw = &pf->hw;
2022 ice_for_each_rxq(vsi, i) {
2026 pf_q = vsi->rxq_map[i];
2027 rx_reg = rd32(hw, QRX_CTRL(pf_q));
2028 if (rx_reg & QRX_CTRL_QENA_STAT_M)
2035 static void ice_vsi_set_tc_cfg(struct ice_vsi *vsi)
2037 if (!test_bit(ICE_FLAG_DCB_ENA, vsi->back->flags)) {
2038 vsi->tc_cfg.ena_tc = ICE_DFLT_TRAFFIC_CLASS;
2039 vsi->tc_cfg.numtc = 1;
2043 /* set VSI TC information based on DCB config */
2044 ice_vsi_set_dcb_tc_cfg(vsi);
2048 * ice_cfg_sw_lldp - Config switch rules for LLDP packet handling
2049 * @vsi: the VSI being configured
2050 * @tx: bool to determine Tx or Rx rule
2051 * @create: bool to determine create or remove Rule
2053 void ice_cfg_sw_lldp(struct ice_vsi *vsi, bool tx, bool create)
2055 int (*eth_fltr)(struct ice_vsi *v, u16 type, u16 flag,
2056 enum ice_sw_fwd_act_type act);
2057 struct ice_pf *pf = vsi->back;
2061 dev = ice_pf_to_dev(pf);
2062 eth_fltr = create ? ice_fltr_add_eth : ice_fltr_remove_eth;
2065 status = eth_fltr(vsi, ETH_P_LLDP, ICE_FLTR_TX,
2068 if (ice_fw_supports_lldp_fltr_ctrl(&pf->hw)) {
2069 status = ice_lldp_fltr_add_remove(&pf->hw, vsi->vsi_num,
2072 status = eth_fltr(vsi, ETH_P_LLDP, ICE_FLTR_RX,
2078 dev_dbg(dev, "Fail %s %s LLDP rule on VSI %i error: %d\n",
2079 create ? "adding" : "removing", tx ? "TX" : "RX",
2080 vsi->vsi_num, status);
2084 * ice_set_agg_vsi - sets up scheduler aggregator node and move VSI into it
2085 * @vsi: pointer to the VSI
2087 * This function will allocate new scheduler aggregator now if needed and will
2088 * move specified VSI into it.
2090 static void ice_set_agg_vsi(struct ice_vsi *vsi)
2092 struct device *dev = ice_pf_to_dev(vsi->back);
2093 struct ice_agg_node *agg_node_iter = NULL;
2094 u32 agg_id = ICE_INVALID_AGG_NODE_ID;
2095 struct ice_agg_node *agg_node = NULL;
2096 int node_offset, max_agg_nodes = 0;
2097 struct ice_port_info *port_info;
2098 struct ice_pf *pf = vsi->back;
2099 u32 agg_node_id_start = 0;
2102 /* create (as needed) scheduler aggregator node and move VSI into
2103 * corresponding aggregator node
2104 * - PF aggregator node to contains VSIs of type _PF and _CTRL
2105 * - VF aggregator nodes will contain VF VSI
2107 port_info = pf->hw.port_info;
2111 switch (vsi->type) {
2117 max_agg_nodes = ICE_MAX_PF_AGG_NODES;
2118 agg_node_id_start = ICE_PF_AGG_NODE_ID_START;
2119 agg_node_iter = &pf->pf_agg_node[0];
2122 /* user can create 'n' VFs on a given PF, but since max children
2123 * per aggregator node can be only 64. Following code handles
2124 * aggregator(s) for VF VSIs, either selects a agg_node which
2125 * was already created provided num_vsis < 64, otherwise
2126 * select next available node, which will be created
2128 max_agg_nodes = ICE_MAX_VF_AGG_NODES;
2129 agg_node_id_start = ICE_VF_AGG_NODE_ID_START;
2130 agg_node_iter = &pf->vf_agg_node[0];
2133 /* other VSI type, handle later if needed */
2134 dev_dbg(dev, "unexpected VSI type %s\n",
2135 ice_vsi_type_str(vsi->type));
2139 /* find the appropriate aggregator node */
2140 for (node_offset = 0; node_offset < max_agg_nodes; node_offset++) {
2141 /* see if we can find space in previously created
2142 * node if num_vsis < 64, otherwise skip
2144 if (agg_node_iter->num_vsis &&
2145 agg_node_iter->num_vsis == ICE_MAX_VSIS_IN_AGG_NODE) {
2150 if (agg_node_iter->valid &&
2151 agg_node_iter->agg_id != ICE_INVALID_AGG_NODE_ID) {
2152 agg_id = agg_node_iter->agg_id;
2153 agg_node = agg_node_iter;
2157 /* find unclaimed agg_id */
2158 if (agg_node_iter->agg_id == ICE_INVALID_AGG_NODE_ID) {
2159 agg_id = node_offset + agg_node_id_start;
2160 agg_node = agg_node_iter;
2163 /* move to next agg_node */
2170 /* if selected aggregator node was not created, create it */
2171 if (!agg_node->valid) {
2172 status = ice_cfg_agg(port_info, agg_id, ICE_AGG_TYPE_AGG,
2173 (u8)vsi->tc_cfg.ena_tc);
2175 dev_err(dev, "unable to create aggregator node with agg_id %u\n",
2179 /* aggregator node is created, store the needed info */
2180 agg_node->valid = true;
2181 agg_node->agg_id = agg_id;
2184 /* move VSI to corresponding aggregator node */
2185 status = ice_move_vsi_to_agg(port_info, agg_id, vsi->idx,
2186 (u8)vsi->tc_cfg.ena_tc);
2188 dev_err(dev, "unable to move VSI idx %u into aggregator %u node",
2193 /* keep active children count for aggregator node */
2194 agg_node->num_vsis++;
2196 /* cache the 'agg_id' in VSI, so that after reset - VSI will be moved
2197 * to aggregator node
2199 vsi->agg_node = agg_node;
2200 dev_dbg(dev, "successfully moved VSI idx %u tc_bitmap 0x%x) into aggregator node %d which has num_vsis %u\n",
2201 vsi->idx, vsi->tc_cfg.ena_tc, vsi->agg_node->agg_id,
2202 vsi->agg_node->num_vsis);
2205 static int ice_vsi_cfg_tc_lan(struct ice_pf *pf, struct ice_vsi *vsi)
2207 u16 max_txqs[ICE_MAX_TRAFFIC_CLASS] = { 0 };
2208 struct device *dev = ice_pf_to_dev(pf);
2211 /* configure VSI nodes based on number of queues and TC's */
2212 ice_for_each_traffic_class(i) {
2213 if (!(vsi->tc_cfg.ena_tc & BIT(i)))
2216 if (vsi->type == ICE_VSI_CHNL) {
2217 if (!vsi->alloc_txq && vsi->num_txq)
2218 max_txqs[i] = vsi->num_txq;
2220 max_txqs[i] = pf->num_lan_tx;
2222 max_txqs[i] = vsi->alloc_txq;
2225 if (vsi->type == ICE_VSI_PF)
2226 max_txqs[i] += vsi->num_xdp_txq;
2229 dev_dbg(dev, "vsi->tc_cfg.ena_tc = %d\n", vsi->tc_cfg.ena_tc);
2230 ret = ice_cfg_vsi_lan(vsi->port_info, vsi->idx, vsi->tc_cfg.ena_tc,
2233 dev_err(dev, "VSI %d failed lan queue config, error %d\n",
2242 * ice_vsi_cfg_def - configure default VSI based on the type
2243 * @vsi: pointer to VSI
2245 static int ice_vsi_cfg_def(struct ice_vsi *vsi)
2247 struct device *dev = ice_pf_to_dev(vsi->back);
2248 struct ice_pf *pf = vsi->back;
2251 vsi->vsw = pf->first_sw;
2253 ret = ice_vsi_alloc_def(vsi, vsi->ch);
2257 /* allocate memory for Tx/Rx ring stat pointers */
2258 ret = ice_vsi_alloc_stat_arrays(vsi);
2260 goto unroll_vsi_alloc;
2262 ice_alloc_fd_res(vsi);
2264 ret = ice_vsi_get_qs(vsi);
2266 dev_err(dev, "Failed to allocate queues. vsi->idx = %d\n",
2268 goto unroll_vsi_alloc_stat;
2271 /* set RSS capabilities */
2272 ice_vsi_set_rss_params(vsi);
2274 /* set TC configuration */
2275 ice_vsi_set_tc_cfg(vsi);
2277 /* create the VSI */
2278 ret = ice_vsi_init(vsi, vsi->flags);
2282 ice_vsi_init_vlan_ops(vsi);
2284 switch (vsi->type) {
2288 ret = ice_vsi_alloc_q_vectors(vsi);
2290 goto unroll_vsi_init;
2292 ret = ice_vsi_alloc_rings(vsi);
2294 goto unroll_vector_base;
2296 ret = ice_vsi_alloc_ring_stats(vsi);
2298 goto unroll_vector_base;
2300 if (ice_is_xdp_ena_vsi(vsi)) {
2301 ret = ice_vsi_determine_xdp_res(vsi);
2303 goto unroll_vector_base;
2304 ret = ice_prepare_xdp_rings(vsi, vsi->xdp_prog,
2307 goto unroll_vector_base;
2310 ice_vsi_map_rings_to_vectors(vsi);
2312 vsi->stat_offsets_loaded = false;
2314 /* ICE_VSI_CTRL does not need RSS so skip RSS processing */
2315 if (vsi->type != ICE_VSI_CTRL)
2316 /* Do not exit if configuring RSS had an issue, at
2317 * least receive traffic on first queue. Hence no
2318 * need to capture return value
2320 if (test_bit(ICE_FLAG_RSS_ENA, pf->flags)) {
2321 ice_vsi_cfg_rss_lut_key(vsi);
2322 ice_vsi_set_rss_flow_fld(vsi);
2327 if (test_bit(ICE_FLAG_RSS_ENA, pf->flags)) {
2328 ice_vsi_cfg_rss_lut_key(vsi);
2329 ice_vsi_set_rss_flow_fld(vsi);
2333 /* VF driver will take care of creating netdev for this type and
2334 * map queues to vectors through Virtchnl, PF driver only
2335 * creates a VSI and corresponding structures for bookkeeping
2338 ret = ice_vsi_alloc_q_vectors(vsi);
2340 goto unroll_vsi_init;
2342 ret = ice_vsi_alloc_rings(vsi);
2344 goto unroll_alloc_q_vector;
2346 ret = ice_vsi_alloc_ring_stats(vsi);
2348 goto unroll_vector_base;
2350 vsi->stat_offsets_loaded = false;
2352 /* Do not exit if configuring RSS had an issue, at least
2353 * receive traffic on first queue. Hence no need to capture
2356 if (test_bit(ICE_FLAG_RSS_ENA, pf->flags)) {
2357 ice_vsi_cfg_rss_lut_key(vsi);
2358 ice_vsi_set_vf_rss_flow_fld(vsi);
2362 ret = ice_vsi_alloc_rings(vsi);
2364 goto unroll_vsi_init;
2366 ret = ice_vsi_alloc_ring_stats(vsi);
2368 goto unroll_vector_base;
2372 /* clean up the resources and exit */
2374 goto unroll_vsi_init;
2380 /* reclaim SW interrupts back to the common pool */
2381 unroll_alloc_q_vector:
2382 ice_vsi_free_q_vectors(vsi);
2384 ice_vsi_delete_from_hw(vsi);
2386 ice_vsi_put_qs(vsi);
2387 unroll_vsi_alloc_stat:
2388 ice_vsi_free_stats(vsi);
2390 ice_vsi_free_arrays(vsi);
2395 * ice_vsi_cfg - configure a previously allocated VSI
2396 * @vsi: pointer to VSI
2398 int ice_vsi_cfg(struct ice_vsi *vsi)
2400 struct ice_pf *pf = vsi->back;
2403 if (WARN_ON(vsi->type == ICE_VSI_VF && !vsi->vf))
2406 ret = ice_vsi_cfg_def(vsi);
2410 ret = ice_vsi_cfg_tc_lan(vsi->back, vsi);
2414 if (vsi->type == ICE_VSI_CTRL) {
2416 WARN_ON(vsi->vf->ctrl_vsi_idx != ICE_NO_VSI);
2417 vsi->vf->ctrl_vsi_idx = vsi->idx;
2419 WARN_ON(pf->ctrl_vsi_idx != ICE_NO_VSI);
2420 pf->ctrl_vsi_idx = vsi->idx;
2428 * ice_vsi_decfg - remove all VSI configuration
2429 * @vsi: pointer to VSI
2431 void ice_vsi_decfg(struct ice_vsi *vsi)
2433 struct ice_pf *pf = vsi->back;
2436 ice_rm_vsi_lan_cfg(vsi->port_info, vsi->idx);
2437 err = ice_rm_vsi_rdma_cfg(vsi->port_info, vsi->idx);
2439 dev_err(ice_pf_to_dev(pf), "Failed to remove RDMA scheduler config for VSI %u, err %d\n",
2443 /* return value check can be skipped here, it always returns
2444 * 0 if reset is in progress
2446 ice_destroy_xdp_rings(vsi, ICE_XDP_CFG_PART);
2448 ice_vsi_clear_rings(vsi);
2449 ice_vsi_free_q_vectors(vsi);
2450 ice_vsi_put_qs(vsi);
2451 ice_vsi_free_arrays(vsi);
2453 /* SR-IOV determines needed MSIX resources all at once instead of per
2454 * VSI since when VFs are spawned we know how many VFs there are and how
2455 * many interrupts each VF needs. SR-IOV MSIX resources are also
2456 * cleared in the same manner.
2459 if (vsi->type == ICE_VSI_VF &&
2460 vsi->agg_node && vsi->agg_node->valid)
2461 vsi->agg_node->num_vsis--;
2465 * ice_vsi_setup - Set up a VSI by a given type
2466 * @pf: board private structure
2467 * @params: parameters to use when creating the VSI
2469 * This allocates the sw VSI structure and its queue resources.
2471 * Returns pointer to the successfully allocated and configured VSI sw struct on
2472 * success, NULL on failure.
2475 ice_vsi_setup(struct ice_pf *pf, struct ice_vsi_cfg_params *params)
2477 struct device *dev = ice_pf_to_dev(pf);
2478 struct ice_vsi *vsi;
2481 /* ice_vsi_setup can only initialize a new VSI, and we must have
2482 * a port_info structure for it.
2484 if (WARN_ON(!(params->flags & ICE_VSI_FLAG_INIT)) ||
2485 WARN_ON(!params->port_info))
2488 vsi = ice_vsi_alloc(pf);
2490 dev_err(dev, "could not allocate VSI\n");
2494 vsi->params = *params;
2495 ret = ice_vsi_cfg(vsi);
2499 /* Add switch rule to drop all Tx Flow Control Frames, of look up
2500 * type ETHERTYPE from VSIs, and restrict malicious VF from sending
2501 * out PAUSE or PFC frames. If enabled, FW can still send FC frames.
2502 * The rule is added once for PF VSI in order to create appropriate
2503 * recipe, since VSI/VSI list is ignored with drop action...
2504 * Also add rules to handle LLDP Tx packets. Tx LLDP packets need to
2505 * be dropped so that VFs cannot send LLDP packets to reconfig DCB
2506 * settings in the HW.
2508 if (!ice_is_safe_mode(pf) && vsi->type == ICE_VSI_PF) {
2509 ice_fltr_add_eth(vsi, ETH_P_PAUSE, ICE_FLTR_TX,
2511 ice_cfg_sw_lldp(vsi, true, true);
2515 ice_set_agg_vsi(vsi);
2526 * ice_vsi_release_msix - Clear the queue to Interrupt mapping in HW
2527 * @vsi: the VSI being cleaned up
2529 static void ice_vsi_release_msix(struct ice_vsi *vsi)
2531 struct ice_pf *pf = vsi->back;
2532 struct ice_hw *hw = &pf->hw;
2537 ice_for_each_q_vector(vsi, i) {
2538 struct ice_q_vector *q_vector = vsi->q_vectors[i];
2540 ice_write_intrl(q_vector, 0);
2541 for (q = 0; q < q_vector->num_ring_tx; q++) {
2542 ice_write_itr(&q_vector->tx, 0);
2543 wr32(hw, QINT_TQCTL(vsi->txq_map[txq]), 0);
2544 if (vsi->xdp_rings) {
2545 u32 xdp_txq = txq + vsi->num_xdp_txq;
2547 wr32(hw, QINT_TQCTL(vsi->txq_map[xdp_txq]), 0);
2552 for (q = 0; q < q_vector->num_ring_rx; q++) {
2553 ice_write_itr(&q_vector->rx, 0);
2554 wr32(hw, QINT_RQCTL(vsi->rxq_map[rxq]), 0);
2563 * ice_vsi_free_irq - Free the IRQ association with the OS
2564 * @vsi: the VSI being configured
2566 void ice_vsi_free_irq(struct ice_vsi *vsi)
2568 struct ice_pf *pf = vsi->back;
2571 if (!vsi->q_vectors || !vsi->irqs_ready)
2574 ice_vsi_release_msix(vsi);
2575 if (vsi->type == ICE_VSI_VF)
2578 vsi->irqs_ready = false;
2579 ice_free_cpu_rx_rmap(vsi);
2581 ice_for_each_q_vector(vsi, i) {
2584 irq_num = vsi->q_vectors[i]->irq.virq;
2586 /* free only the irqs that were actually requested */
2587 if (!vsi->q_vectors[i] ||
2588 !(vsi->q_vectors[i]->num_ring_tx ||
2589 vsi->q_vectors[i]->num_ring_rx))
2592 /* clear the affinity notifier in the IRQ descriptor */
2593 if (!IS_ENABLED(CONFIG_RFS_ACCEL))
2594 irq_set_affinity_notifier(irq_num, NULL);
2596 /* clear the affinity_hint in the IRQ descriptor */
2597 irq_update_affinity_hint(irq_num, NULL);
2598 synchronize_irq(irq_num);
2599 devm_free_irq(ice_pf_to_dev(pf), irq_num, vsi->q_vectors[i]);
2604 * ice_vsi_free_tx_rings - Free Tx resources for VSI queues
2605 * @vsi: the VSI having resources freed
2607 void ice_vsi_free_tx_rings(struct ice_vsi *vsi)
2614 ice_for_each_txq(vsi, i)
2615 if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc)
2616 ice_free_tx_ring(vsi->tx_rings[i]);
2620 * ice_vsi_free_rx_rings - Free Rx resources for VSI queues
2621 * @vsi: the VSI having resources freed
2623 void ice_vsi_free_rx_rings(struct ice_vsi *vsi)
2630 ice_for_each_rxq(vsi, i)
2631 if (vsi->rx_rings[i] && vsi->rx_rings[i]->desc)
2632 ice_free_rx_ring(vsi->rx_rings[i]);
2636 * ice_vsi_close - Shut down a VSI
2637 * @vsi: the VSI being shut down
2639 void ice_vsi_close(struct ice_vsi *vsi)
2641 if (!test_and_set_bit(ICE_VSI_DOWN, vsi->state))
2644 ice_vsi_clear_napi_queues(vsi);
2645 ice_vsi_free_irq(vsi);
2646 ice_vsi_free_tx_rings(vsi);
2647 ice_vsi_free_rx_rings(vsi);
2651 * ice_ena_vsi - resume a VSI
2652 * @vsi: the VSI being resume
2653 * @locked: is the rtnl_lock already held
2655 int ice_ena_vsi(struct ice_vsi *vsi, bool locked)
2659 if (!test_bit(ICE_VSI_NEEDS_RESTART, vsi->state))
2662 clear_bit(ICE_VSI_NEEDS_RESTART, vsi->state);
2664 if (vsi->netdev && (vsi->type == ICE_VSI_PF ||
2665 vsi->type == ICE_VSI_SF)) {
2666 if (netif_running(vsi->netdev)) {
2670 err = ice_open_internal(vsi->netdev);
2675 } else if (vsi->type == ICE_VSI_CTRL) {
2676 err = ice_vsi_open_ctrl(vsi);
2683 * ice_dis_vsi - pause a VSI
2684 * @vsi: the VSI being paused
2685 * @locked: is the rtnl_lock already held
2687 void ice_dis_vsi(struct ice_vsi *vsi, bool locked)
2689 bool already_down = test_bit(ICE_VSI_DOWN, vsi->state);
2691 set_bit(ICE_VSI_NEEDS_RESTART, vsi->state);
2693 if (vsi->netdev && (vsi->type == ICE_VSI_PF ||
2694 vsi->type == ICE_VSI_SF)) {
2695 if (netif_running(vsi->netdev)) {
2698 already_down = test_bit(ICE_VSI_DOWN, vsi->state);
2704 } else if (!already_down) {
2707 } else if (vsi->type == ICE_VSI_CTRL && !already_down) {
2713 * ice_vsi_set_napi_queues - associate netdev queues with napi
2716 * Associate queue[s] with napi for all vectors.
2717 * The caller must hold rtnl_lock.
2719 void ice_vsi_set_napi_queues(struct ice_vsi *vsi)
2721 struct net_device *netdev = vsi->netdev;
2727 ice_for_each_rxq(vsi, q_idx)
2728 netif_queue_set_napi(netdev, q_idx, NETDEV_QUEUE_TYPE_RX,
2729 &vsi->rx_rings[q_idx]->q_vector->napi);
2731 ice_for_each_txq(vsi, q_idx)
2732 netif_queue_set_napi(netdev, q_idx, NETDEV_QUEUE_TYPE_TX,
2733 &vsi->tx_rings[q_idx]->q_vector->napi);
2734 /* Also set the interrupt number for the NAPI */
2735 ice_for_each_q_vector(vsi, v_idx) {
2736 struct ice_q_vector *q_vector = vsi->q_vectors[v_idx];
2738 netif_napi_set_irq(&q_vector->napi, q_vector->irq.virq);
2743 * ice_vsi_clear_napi_queues - dissociate netdev queues from napi
2746 * Clear the association between all VSI queues queue[s] and napi.
2747 * The caller must hold rtnl_lock.
2749 void ice_vsi_clear_napi_queues(struct ice_vsi *vsi)
2751 struct net_device *netdev = vsi->netdev;
2757 ice_for_each_txq(vsi, q_idx)
2758 netif_queue_set_napi(netdev, q_idx, NETDEV_QUEUE_TYPE_TX, NULL);
2760 ice_for_each_rxq(vsi, q_idx)
2761 netif_queue_set_napi(netdev, q_idx, NETDEV_QUEUE_TYPE_RX, NULL);
2765 * ice_napi_add - register NAPI handler for the VSI
2766 * @vsi: VSI for which NAPI handler is to be registered
2768 * This function is only called in the driver's load path. Registering the NAPI
2769 * handler is done in ice_vsi_alloc_q_vector() for all other cases (i.e. resume,
2770 * reset/rebuild, etc.)
2772 void ice_napi_add(struct ice_vsi *vsi)
2779 ice_for_each_q_vector(vsi, v_idx)
2780 netif_napi_add_config(vsi->netdev,
2781 &vsi->q_vectors[v_idx]->napi,
2787 * ice_vsi_release - Delete a VSI and free its resources
2788 * @vsi: the VSI being removed
2790 * Returns 0 on success or < 0 on error
2792 int ice_vsi_release(struct ice_vsi *vsi)
2800 if (test_bit(ICE_FLAG_RSS_ENA, pf->flags))
2805 /* The Rx rule will only exist to remove if the LLDP FW
2806 * engine is currently stopped
2808 if (!ice_is_safe_mode(pf) && vsi->type == ICE_VSI_PF &&
2809 !test_bit(ICE_FLAG_FW_LLDP_AGENT, pf->flags))
2810 ice_cfg_sw_lldp(vsi, false, false);
2814 /* retain SW VSI data structure since it is needed to unregister and
2815 * free VSI netdev when PF is not in reset recovery pending state,\
2816 * for ex: during rmmod.
2818 if (!ice_is_reset_in_progress(pf->state))
2819 ice_vsi_delete(vsi);
2825 * ice_vsi_rebuild_get_coalesce - get coalesce from all q_vectors
2826 * @vsi: VSI connected with q_vectors
2827 * @coalesce: array of struct with stored coalesce
2829 * Returns array size.
2832 ice_vsi_rebuild_get_coalesce(struct ice_vsi *vsi,
2833 struct ice_coalesce_stored *coalesce)
2837 ice_for_each_q_vector(vsi, i) {
2838 struct ice_q_vector *q_vector = vsi->q_vectors[i];
2840 coalesce[i].itr_tx = q_vector->tx.itr_settings;
2841 coalesce[i].itr_rx = q_vector->rx.itr_settings;
2842 coalesce[i].intrl = q_vector->intrl;
2844 if (i < vsi->num_txq)
2845 coalesce[i].tx_valid = true;
2846 if (i < vsi->num_rxq)
2847 coalesce[i].rx_valid = true;
2850 return vsi->num_q_vectors;
2854 * ice_vsi_rebuild_set_coalesce - set coalesce from earlier saved arrays
2855 * @vsi: VSI connected with q_vectors
2856 * @coalesce: pointer to array of struct with stored coalesce
2857 * @size: size of coalesce array
2859 * Before this function, ice_vsi_rebuild_get_coalesce should be called to save
2860 * ITR params in arrays. If size is 0 or coalesce wasn't stored set coalesce
2864 ice_vsi_rebuild_set_coalesce(struct ice_vsi *vsi,
2865 struct ice_coalesce_stored *coalesce, int size)
2867 struct ice_ring_container *rc;
2870 if ((size && !coalesce) || !vsi)
2873 /* There are a couple of cases that have to be handled here:
2874 * 1. The case where the number of queue vectors stays the same, but
2875 * the number of Tx or Rx rings changes (the first for loop)
2876 * 2. The case where the number of queue vectors increased (the
2879 for (i = 0; i < size && i < vsi->num_q_vectors; i++) {
2880 /* There are 2 cases to handle here and they are the same for
2882 * if the entry was valid previously (coalesce[i].[tr]x_valid
2883 * and the loop variable is less than the number of rings
2884 * allocated, then write the previous values
2886 * if the entry was not valid previously, but the number of
2887 * rings is less than are allocated (this means the number of
2888 * rings increased from previously), then write out the
2889 * values in the first element
2891 * Also, always write the ITR, even if in ITR_IS_DYNAMIC
2892 * as there is no harm because the dynamic algorithm
2893 * will just overwrite.
2895 if (i < vsi->alloc_rxq && coalesce[i].rx_valid) {
2896 rc = &vsi->q_vectors[i]->rx;
2897 rc->itr_settings = coalesce[i].itr_rx;
2898 ice_write_itr(rc, rc->itr_setting);
2899 } else if (i < vsi->alloc_rxq) {
2900 rc = &vsi->q_vectors[i]->rx;
2901 rc->itr_settings = coalesce[0].itr_rx;
2902 ice_write_itr(rc, rc->itr_setting);
2905 if (i < vsi->alloc_txq && coalesce[i].tx_valid) {
2906 rc = &vsi->q_vectors[i]->tx;
2907 rc->itr_settings = coalesce[i].itr_tx;
2908 ice_write_itr(rc, rc->itr_setting);
2909 } else if (i < vsi->alloc_txq) {
2910 rc = &vsi->q_vectors[i]->tx;
2911 rc->itr_settings = coalesce[0].itr_tx;
2912 ice_write_itr(rc, rc->itr_setting);
2915 vsi->q_vectors[i]->intrl = coalesce[i].intrl;
2916 ice_set_q_vector_intrl(vsi->q_vectors[i]);
2919 /* the number of queue vectors increased so write whatever is in
2922 for (; i < vsi->num_q_vectors; i++) {
2924 rc = &vsi->q_vectors[i]->tx;
2925 rc->itr_settings = coalesce[0].itr_tx;
2926 ice_write_itr(rc, rc->itr_setting);
2929 rc = &vsi->q_vectors[i]->rx;
2930 rc->itr_settings = coalesce[0].itr_rx;
2931 ice_write_itr(rc, rc->itr_setting);
2933 vsi->q_vectors[i]->intrl = coalesce[0].intrl;
2934 ice_set_q_vector_intrl(vsi->q_vectors[i]);
2939 * ice_vsi_realloc_stat_arrays - Frees unused stat structures or alloc new ones
2943 ice_vsi_realloc_stat_arrays(struct ice_vsi *vsi)
2945 u16 req_txq = vsi->req_txq ? vsi->req_txq : vsi->alloc_txq;
2946 u16 req_rxq = vsi->req_rxq ? vsi->req_rxq : vsi->alloc_rxq;
2947 struct ice_ring_stats **tx_ring_stats;
2948 struct ice_ring_stats **rx_ring_stats;
2949 struct ice_vsi_stats *vsi_stat;
2950 struct ice_pf *pf = vsi->back;
2951 u16 prev_txq = vsi->alloc_txq;
2952 u16 prev_rxq = vsi->alloc_rxq;
2955 vsi_stat = pf->vsi_stats[vsi->idx];
2957 if (req_txq < prev_txq) {
2958 for (i = req_txq; i < prev_txq; i++) {
2959 if (vsi_stat->tx_ring_stats[i]) {
2960 kfree_rcu(vsi_stat->tx_ring_stats[i], rcu);
2961 WRITE_ONCE(vsi_stat->tx_ring_stats[i], NULL);
2966 tx_ring_stats = vsi_stat->tx_ring_stats;
2967 vsi_stat->tx_ring_stats =
2968 krealloc_array(vsi_stat->tx_ring_stats, req_txq,
2969 sizeof(*vsi_stat->tx_ring_stats),
2970 GFP_KERNEL | __GFP_ZERO);
2971 if (!vsi_stat->tx_ring_stats) {
2972 vsi_stat->tx_ring_stats = tx_ring_stats;
2976 if (req_rxq < prev_rxq) {
2977 for (i = req_rxq; i < prev_rxq; i++) {
2978 if (vsi_stat->rx_ring_stats[i]) {
2979 kfree_rcu(vsi_stat->rx_ring_stats[i], rcu);
2980 WRITE_ONCE(vsi_stat->rx_ring_stats[i], NULL);
2985 rx_ring_stats = vsi_stat->rx_ring_stats;
2986 vsi_stat->rx_ring_stats =
2987 krealloc_array(vsi_stat->rx_ring_stats, req_rxq,
2988 sizeof(*vsi_stat->rx_ring_stats),
2989 GFP_KERNEL | __GFP_ZERO);
2990 if (!vsi_stat->rx_ring_stats) {
2991 vsi_stat->rx_ring_stats = rx_ring_stats;
2999 * ice_vsi_rebuild - Rebuild VSI after reset
3000 * @vsi: VSI to be rebuild
3001 * @vsi_flags: flags used for VSI rebuild flow
3003 * Set vsi_flags to ICE_VSI_FLAG_INIT to initialize a new VSI, or
3004 * ICE_VSI_FLAG_NO_INIT to rebuild an existing VSI in hardware.
3006 * Returns 0 on success and negative value on failure
3008 int ice_vsi_rebuild(struct ice_vsi *vsi, u32 vsi_flags)
3010 struct ice_coalesce_stored *coalesce;
3011 int prev_num_q_vectors;
3018 vsi->flags = vsi_flags;
3020 if (WARN_ON(vsi->type == ICE_VSI_VF && !vsi->vf))
3023 mutex_lock(&vsi->xdp_state_lock);
3025 ret = ice_vsi_realloc_stat_arrays(vsi);
3030 ret = ice_vsi_cfg_def(vsi);
3034 coalesce = kcalloc(vsi->num_q_vectors,
3035 sizeof(struct ice_coalesce_stored), GFP_KERNEL);
3041 prev_num_q_vectors = ice_vsi_rebuild_get_coalesce(vsi, coalesce);
3043 ret = ice_vsi_cfg_tc_lan(pf, vsi);
3045 if (vsi_flags & ICE_VSI_FLAG_INIT) {
3050 ret = ice_schedule_reset(pf, ICE_RESET_PFR);
3054 ice_vsi_rebuild_set_coalesce(vsi, coalesce, prev_num_q_vectors);
3055 clear_bit(ICE_VSI_REBUILD_PENDING, vsi->state);
3063 mutex_unlock(&vsi->xdp_state_lock);
3068 * ice_is_reset_in_progress - check for a reset in progress
3069 * @state: PF state field
3071 bool ice_is_reset_in_progress(unsigned long *state)
3073 return test_bit(ICE_RESET_OICR_RECV, state) ||
3074 test_bit(ICE_PFR_REQ, state) ||
3075 test_bit(ICE_CORER_REQ, state) ||
3076 test_bit(ICE_GLOBR_REQ, state);
3080 * ice_wait_for_reset - Wait for driver to finish reset and rebuild
3081 * @pf: pointer to the PF structure
3082 * @timeout: length of time to wait, in jiffies
3084 * Wait (sleep) for a short time until the driver finishes cleaning up from
3085 * a device reset. The caller must be able to sleep. Use this to delay
3086 * operations that could fail while the driver is cleaning up after a device
3089 * Returns 0 on success, -EBUSY if the reset is not finished within the
3090 * timeout, and -ERESTARTSYS if the thread was interrupted.
3092 int ice_wait_for_reset(struct ice_pf *pf, unsigned long timeout)
3096 ret = wait_event_interruptible_timeout(pf->reset_wait_queue,
3097 !ice_is_reset_in_progress(pf->state),
3108 * ice_vsi_update_q_map - update our copy of the VSI info with new queue map
3109 * @vsi: VSI being configured
3110 * @ctx: the context buffer returned from AQ VSI update command
3112 static void ice_vsi_update_q_map(struct ice_vsi *vsi, struct ice_vsi_ctx *ctx)
3114 vsi->info.mapping_flags = ctx->info.mapping_flags;
3115 memcpy(&vsi->info.q_mapping, &ctx->info.q_mapping,
3116 sizeof(vsi->info.q_mapping));
3117 memcpy(&vsi->info.tc_mapping, ctx->info.tc_mapping,
3118 sizeof(vsi->info.tc_mapping));
3122 * ice_vsi_cfg_netdev_tc - Setup the netdev TC configuration
3123 * @vsi: the VSI being configured
3124 * @ena_tc: TC map to be enabled
3126 void ice_vsi_cfg_netdev_tc(struct ice_vsi *vsi, u8 ena_tc)
3128 struct net_device *netdev = vsi->netdev;
3129 struct ice_pf *pf = vsi->back;
3130 int numtc = vsi->tc_cfg.numtc;
3131 struct ice_dcbx_cfg *dcbcfg;
3138 /* CHNL VSI doesn't have it's own netdev, hence, no netdev_tc */
3139 if (vsi->type == ICE_VSI_CHNL)
3143 netdev_reset_tc(netdev);
3147 if (vsi->type == ICE_VSI_PF && ice_is_adq_active(pf))
3148 numtc = vsi->all_numtc;
3150 if (netdev_set_num_tc(netdev, numtc))
3153 dcbcfg = &pf->hw.port_info->qos_cfg.local_dcbx_cfg;
3155 ice_for_each_traffic_class(i)
3156 if (vsi->tc_cfg.ena_tc & BIT(i))
3157 netdev_set_tc_queue(netdev,
3158 vsi->tc_cfg.tc_info[i].netdev_tc,
3159 vsi->tc_cfg.tc_info[i].qcount_tx,
3160 vsi->tc_cfg.tc_info[i].qoffset);
3161 /* setup TC queue map for CHNL TCs */
3162 ice_for_each_chnl_tc(i) {
3163 if (!(vsi->all_enatc & BIT(i)))
3165 if (!vsi->mqprio_qopt.qopt.count[i])
3167 netdev_set_tc_queue(netdev, i,
3168 vsi->mqprio_qopt.qopt.count[i],
3169 vsi->mqprio_qopt.qopt.offset[i]);
3172 if (test_bit(ICE_FLAG_TC_MQPRIO, pf->flags))
3175 for (i = 0; i < ICE_MAX_USER_PRIORITY; i++) {
3176 u8 ets_tc = dcbcfg->etscfg.prio_table[i];
3178 /* Get the mapped netdev TC# for the UP */
3179 netdev_tc = vsi->tc_cfg.tc_info[ets_tc].netdev_tc;
3180 netdev_set_prio_tc_map(netdev, i, netdev_tc);
3185 * ice_vsi_setup_q_map_mqprio - Prepares mqprio based tc_config
3186 * @vsi: the VSI being configured,
3187 * @ctxt: VSI context structure
3188 * @ena_tc: number of traffic classes to enable
3190 * Prepares VSI tc_config to have queue configurations based on MQPRIO options.
3193 ice_vsi_setup_q_map_mqprio(struct ice_vsi *vsi, struct ice_vsi_ctx *ctxt,
3196 u16 pow, offset = 0, qcount_tx = 0, qcount_rx = 0, qmap;
3197 u16 tc0_offset = vsi->mqprio_qopt.qopt.offset[0];
3198 int tc0_qcount = vsi->mqprio_qopt.qopt.count[0];
3199 u16 new_txq, new_rxq;
3203 vsi->tc_cfg.ena_tc = ena_tc ? ena_tc : 1;
3205 pow = order_base_2(tc0_qcount);
3206 qmap = FIELD_PREP(ICE_AQ_VSI_TC_Q_OFFSET_M, tc0_offset);
3207 qmap |= FIELD_PREP(ICE_AQ_VSI_TC_Q_NUM_M, pow);
3209 ice_for_each_traffic_class(i) {
3210 if (!(vsi->tc_cfg.ena_tc & BIT(i))) {
3211 /* TC is not enabled */
3212 vsi->tc_cfg.tc_info[i].qoffset = 0;
3213 vsi->tc_cfg.tc_info[i].qcount_rx = 1;
3214 vsi->tc_cfg.tc_info[i].qcount_tx = 1;
3215 vsi->tc_cfg.tc_info[i].netdev_tc = 0;
3216 ctxt->info.tc_mapping[i] = 0;
3220 offset = vsi->mqprio_qopt.qopt.offset[i];
3221 qcount_rx = vsi->mqprio_qopt.qopt.count[i];
3222 qcount_tx = vsi->mqprio_qopt.qopt.count[i];
3223 vsi->tc_cfg.tc_info[i].qoffset = offset;
3224 vsi->tc_cfg.tc_info[i].qcount_rx = qcount_rx;
3225 vsi->tc_cfg.tc_info[i].qcount_tx = qcount_tx;
3226 vsi->tc_cfg.tc_info[i].netdev_tc = netdev_tc++;
3229 if (vsi->all_numtc && vsi->all_numtc != vsi->tc_cfg.numtc) {
3230 ice_for_each_chnl_tc(i) {
3231 if (!(vsi->all_enatc & BIT(i)))
3233 offset = vsi->mqprio_qopt.qopt.offset[i];
3234 qcount_rx = vsi->mqprio_qopt.qopt.count[i];
3235 qcount_tx = vsi->mqprio_qopt.qopt.count[i];
3239 new_txq = offset + qcount_tx;
3240 if (new_txq > vsi->alloc_txq) {
3241 dev_err(ice_pf_to_dev(vsi->back), "Trying to use more Tx queues (%u), than were allocated (%u)!\n",
3242 new_txq, vsi->alloc_txq);
3246 new_rxq = offset + qcount_rx;
3247 if (new_rxq > vsi->alloc_rxq) {
3248 dev_err(ice_pf_to_dev(vsi->back), "Trying to use more Rx queues (%u), than were allocated (%u)!\n",
3249 new_rxq, vsi->alloc_rxq);
3253 /* Set actual Tx/Rx queue pairs */
3254 vsi->num_txq = new_txq;
3255 vsi->num_rxq = new_rxq;
3257 /* Setup queue TC[0].qmap for given VSI context */
3258 ctxt->info.tc_mapping[0] = cpu_to_le16(qmap);
3259 ctxt->info.q_mapping[0] = cpu_to_le16(vsi->rxq_map[0]);
3260 ctxt->info.q_mapping[1] = cpu_to_le16(tc0_qcount);
3262 /* Find queue count available for channel VSIs and starting offset
3265 if (tc0_qcount && tc0_qcount < vsi->num_rxq) {
3266 vsi->cnt_q_avail = vsi->num_rxq - tc0_qcount;
3267 vsi->next_base_q = tc0_qcount;
3269 dev_dbg(ice_pf_to_dev(vsi->back), "vsi->num_txq = %d\n", vsi->num_txq);
3270 dev_dbg(ice_pf_to_dev(vsi->back), "vsi->num_rxq = %d\n", vsi->num_rxq);
3271 dev_dbg(ice_pf_to_dev(vsi->back), "all_numtc %u, all_enatc: 0x%04x, tc_cfg.numtc %u\n",
3272 vsi->all_numtc, vsi->all_enatc, vsi->tc_cfg.numtc);
3278 * ice_vsi_cfg_tc - Configure VSI Tx Sched for given TC map
3279 * @vsi: VSI to be configured
3280 * @ena_tc: TC bitmap
3282 * VSI queues expected to be quiesced before calling this function
3284 int ice_vsi_cfg_tc(struct ice_vsi *vsi, u8 ena_tc)
3286 u16 max_txqs[ICE_MAX_TRAFFIC_CLASS] = { 0 };
3287 struct ice_pf *pf = vsi->back;
3288 struct ice_tc_cfg old_tc_cfg;
3289 struct ice_vsi_ctx *ctx;
3294 dev = ice_pf_to_dev(pf);
3295 if (vsi->tc_cfg.ena_tc == ena_tc &&
3296 vsi->mqprio_qopt.mode != TC_MQPRIO_MODE_CHANNEL)
3299 ice_for_each_traffic_class(i) {
3300 /* build bitmap of enabled TCs */
3301 if (ena_tc & BIT(i))
3303 /* populate max_txqs per TC */
3304 max_txqs[i] = vsi->alloc_txq;
3305 /* Update max_txqs if it is CHNL VSI, because alloc_t[r]xq are
3306 * zero for CHNL VSI, hence use num_txq instead as max_txqs
3308 if (vsi->type == ICE_VSI_CHNL &&
3309 test_bit(ICE_FLAG_TC_MQPRIO, pf->flags))
3310 max_txqs[i] = vsi->num_txq;
3313 memcpy(&old_tc_cfg, &vsi->tc_cfg, sizeof(old_tc_cfg));
3314 vsi->tc_cfg.ena_tc = ena_tc;
3315 vsi->tc_cfg.numtc = num_tc;
3317 ctx = kzalloc(sizeof(*ctx), GFP_KERNEL);
3322 ctx->info = vsi->info;
3324 if (vsi->type == ICE_VSI_PF &&
3325 test_bit(ICE_FLAG_TC_MQPRIO, pf->flags))
3326 ret = ice_vsi_setup_q_map_mqprio(vsi, ctx, ena_tc);
3328 ret = ice_vsi_setup_q_map(vsi, ctx);
3331 memcpy(&vsi->tc_cfg, &old_tc_cfg, sizeof(vsi->tc_cfg));
3335 /* must to indicate which section of VSI context are being modified */
3336 ctx->info.valid_sections = cpu_to_le16(ICE_AQ_VSI_PROP_RXQ_MAP_VALID);
3337 ret = ice_update_vsi(&pf->hw, vsi->idx, ctx, NULL);
3339 dev_info(dev, "Failed VSI Update\n");
3343 if (vsi->type == ICE_VSI_PF &&
3344 test_bit(ICE_FLAG_TC_MQPRIO, pf->flags))
3345 ret = ice_cfg_vsi_lan(vsi->port_info, vsi->idx, 1, max_txqs);
3347 ret = ice_cfg_vsi_lan(vsi->port_info, vsi->idx,
3348 vsi->tc_cfg.ena_tc, max_txqs);
3351 dev_err(dev, "VSI %d failed TC config, error %d\n",
3355 ice_vsi_update_q_map(vsi, ctx);
3356 vsi->info.valid_sections = 0;
3358 ice_vsi_cfg_netdev_tc(vsi, ena_tc);
3365 * ice_update_ring_stats - Update ring statistics
3366 * @stats: stats to be updated
3367 * @pkts: number of processed packets
3368 * @bytes: number of processed bytes
3370 * This function assumes that caller has acquired a u64_stats_sync lock.
3372 static void ice_update_ring_stats(struct ice_q_stats *stats, u64 pkts, u64 bytes)
3374 stats->bytes += bytes;
3375 stats->pkts += pkts;
3379 * ice_update_tx_ring_stats - Update Tx ring specific counters
3380 * @tx_ring: ring to update
3381 * @pkts: number of processed packets
3382 * @bytes: number of processed bytes
3384 void ice_update_tx_ring_stats(struct ice_tx_ring *tx_ring, u64 pkts, u64 bytes)
3386 u64_stats_update_begin(&tx_ring->ring_stats->syncp);
3387 ice_update_ring_stats(&tx_ring->ring_stats->stats, pkts, bytes);
3388 u64_stats_update_end(&tx_ring->ring_stats->syncp);
3392 * ice_update_rx_ring_stats - Update Rx ring specific counters
3393 * @rx_ring: ring to update
3394 * @pkts: number of processed packets
3395 * @bytes: number of processed bytes
3397 void ice_update_rx_ring_stats(struct ice_rx_ring *rx_ring, u64 pkts, u64 bytes)
3399 u64_stats_update_begin(&rx_ring->ring_stats->syncp);
3400 ice_update_ring_stats(&rx_ring->ring_stats->stats, pkts, bytes);
3401 u64_stats_update_end(&rx_ring->ring_stats->syncp);
3405 * ice_is_dflt_vsi_in_use - check if the default forwarding VSI is being used
3406 * @pi: port info of the switch with default VSI
3408 * Return true if the there is a single VSI in default forwarding VSI list
3410 bool ice_is_dflt_vsi_in_use(struct ice_port_info *pi)
3412 bool exists = false;
3414 ice_check_if_dflt_vsi(pi, 0, &exists);
3419 * ice_is_vsi_dflt_vsi - check if the VSI passed in is the default VSI
3420 * @vsi: VSI to compare against default forwarding VSI
3422 * If this VSI passed in is the default forwarding VSI then return true, else
3425 bool ice_is_vsi_dflt_vsi(struct ice_vsi *vsi)
3427 return ice_check_if_dflt_vsi(vsi->port_info, vsi->idx, NULL);
3431 * ice_set_dflt_vsi - set the default forwarding VSI
3432 * @vsi: VSI getting set as the default forwarding VSI on the switch
3434 * If the VSI passed in is already the default VSI and it's enabled just return
3437 * Otherwise try to set the VSI passed in as the switch's default VSI and
3438 * return the result.
3440 int ice_set_dflt_vsi(struct ice_vsi *vsi)
3448 dev = ice_pf_to_dev(vsi->back);
3450 if (ice_lag_is_switchdev_running(vsi->back)) {
3451 dev_dbg(dev, "VSI %d passed is a part of LAG containing interfaces in switchdev mode, nothing to do\n",
3456 /* the VSI passed in is already the default VSI */
3457 if (ice_is_vsi_dflt_vsi(vsi)) {
3458 dev_dbg(dev, "VSI %d passed in is already the default forwarding VSI, nothing to do\n",
3463 status = ice_cfg_dflt_vsi(vsi->port_info, vsi->idx, true, ICE_FLTR_RX);
3465 dev_err(dev, "Failed to set VSI %d as the default forwarding VSI, error %d\n",
3466 vsi->vsi_num, status);
3474 * ice_clear_dflt_vsi - clear the default forwarding VSI
3475 * @vsi: VSI to remove from filter list
3477 * If the switch has no default VSI or it's not enabled then return error.
3479 * Otherwise try to clear the default VSI and return the result.
3481 int ice_clear_dflt_vsi(struct ice_vsi *vsi)
3489 dev = ice_pf_to_dev(vsi->back);
3491 /* there is no default VSI configured */
3492 if (!ice_is_dflt_vsi_in_use(vsi->port_info))
3495 status = ice_cfg_dflt_vsi(vsi->port_info, vsi->idx, false,
3498 dev_err(dev, "Failed to clear the default forwarding VSI %d, error %d\n",
3499 vsi->vsi_num, status);
3507 * ice_get_link_speed_mbps - get link speed in Mbps
3508 * @vsi: the VSI whose link speed is being queried
3510 * Return current VSI link speed and 0 if the speed is unknown.
3512 int ice_get_link_speed_mbps(struct ice_vsi *vsi)
3514 unsigned int link_speed;
3516 link_speed = vsi->port_info->phy.link_info.link_speed;
3518 return (int)ice_get_link_speed(fls(link_speed) - 1);
3522 * ice_get_link_speed_kbps - get link speed in Kbps
3523 * @vsi: the VSI whose link speed is being queried
3525 * Return current VSI link speed and 0 if the speed is unknown.
3527 int ice_get_link_speed_kbps(struct ice_vsi *vsi)
3531 speed_mbps = ice_get_link_speed_mbps(vsi);
3533 return speed_mbps * 1000;
3537 * ice_set_min_bw_limit - setup minimum BW limit for Tx based on min_tx_rate
3538 * @vsi: VSI to be configured
3539 * @min_tx_rate: min Tx rate in Kbps to be configured as BW limit
3541 * If the min_tx_rate is specified as 0 that means to clear the minimum BW limit
3542 * profile, otherwise a non-zero value will force a minimum BW limit for the VSI
3545 int ice_set_min_bw_limit(struct ice_vsi *vsi, u64 min_tx_rate)
3547 struct ice_pf *pf = vsi->back;
3552 dev = ice_pf_to_dev(pf);
3553 if (!vsi->port_info) {
3554 dev_dbg(dev, "VSI %d, type %u specified doesn't have valid port_info\n",
3555 vsi->idx, vsi->type);
3559 speed = ice_get_link_speed_kbps(vsi);
3560 if (min_tx_rate > (u64)speed) {
3561 dev_err(dev, "invalid min Tx rate %llu Kbps specified for %s %d is greater than current link speed %u Kbps\n",
3562 min_tx_rate, ice_vsi_type_str(vsi->type), vsi->idx,
3567 /* Configure min BW for VSI limit */
3569 status = ice_cfg_vsi_bw_lmt_per_tc(vsi->port_info, vsi->idx, 0,
3570 ICE_MIN_BW, min_tx_rate);
3572 dev_err(dev, "failed to set min Tx rate(%llu Kbps) for %s %d\n",
3573 min_tx_rate, ice_vsi_type_str(vsi->type),
3578 dev_dbg(dev, "set min Tx rate(%llu Kbps) for %s\n",
3579 min_tx_rate, ice_vsi_type_str(vsi->type));
3581 status = ice_cfg_vsi_bw_dflt_lmt_per_tc(vsi->port_info,
3585 dev_err(dev, "failed to clear min Tx rate configuration for %s %d\n",
3586 ice_vsi_type_str(vsi->type), vsi->idx);
3590 dev_dbg(dev, "cleared min Tx rate configuration for %s %d\n",
3591 ice_vsi_type_str(vsi->type), vsi->idx);
3598 * ice_set_max_bw_limit - setup maximum BW limit for Tx based on max_tx_rate
3599 * @vsi: VSI to be configured
3600 * @max_tx_rate: max Tx rate in Kbps to be configured as BW limit
3602 * If the max_tx_rate is specified as 0 that means to clear the maximum BW limit
3603 * profile, otherwise a non-zero value will force a maximum BW limit for the VSI
3606 int ice_set_max_bw_limit(struct ice_vsi *vsi, u64 max_tx_rate)
3608 struct ice_pf *pf = vsi->back;
3613 dev = ice_pf_to_dev(pf);
3614 if (!vsi->port_info) {
3615 dev_dbg(dev, "VSI %d, type %u specified doesn't have valid port_info\n",
3616 vsi->idx, vsi->type);
3620 speed = ice_get_link_speed_kbps(vsi);
3621 if (max_tx_rate > (u64)speed) {
3622 dev_err(dev, "invalid max Tx rate %llu Kbps specified for %s %d is greater than current link speed %u Kbps\n",
3623 max_tx_rate, ice_vsi_type_str(vsi->type), vsi->idx,
3628 /* Configure max BW for VSI limit */
3630 status = ice_cfg_vsi_bw_lmt_per_tc(vsi->port_info, vsi->idx, 0,
3631 ICE_MAX_BW, max_tx_rate);
3633 dev_err(dev, "failed setting max Tx rate(%llu Kbps) for %s %d\n",
3634 max_tx_rate, ice_vsi_type_str(vsi->type),
3639 dev_dbg(dev, "set max Tx rate(%llu Kbps) for %s %d\n",
3640 max_tx_rate, ice_vsi_type_str(vsi->type), vsi->idx);
3642 status = ice_cfg_vsi_bw_dflt_lmt_per_tc(vsi->port_info,
3646 dev_err(dev, "failed clearing max Tx rate configuration for %s %d\n",
3647 ice_vsi_type_str(vsi->type), vsi->idx);
3651 dev_dbg(dev, "cleared max Tx rate configuration for %s %d\n",
3652 ice_vsi_type_str(vsi->type), vsi->idx);
3659 * ice_set_link - turn on/off physical link
3660 * @vsi: VSI to modify physical link on
3661 * @ena: turn on/off physical link
3663 int ice_set_link(struct ice_vsi *vsi, bool ena)
3665 struct device *dev = ice_pf_to_dev(vsi->back);
3666 struct ice_port_info *pi = vsi->port_info;
3667 struct ice_hw *hw = pi->hw;
3670 if (vsi->type != ICE_VSI_PF)
3673 status = ice_aq_set_link_restart_an(pi, ena, NULL);
3675 /* if link is owned by manageability, FW will return ICE_AQ_RC_EMODE.
3676 * this is not a fatal error, so print a warning message and return
3677 * a success code. Return an error if FW returns an error code other
3678 * than ICE_AQ_RC_EMODE
3680 if (status == -EIO) {
3681 if (hw->adminq.sq_last_status == ICE_AQ_RC_EMODE)
3682 dev_dbg(dev, "can't set link to %s, err %d aq_err %s. not fatal, continuing\n",
3683 (ena ? "ON" : "OFF"), status,
3684 ice_aq_str(hw->adminq.sq_last_status));
3685 } else if (status) {
3686 dev_err(dev, "can't set link to %s, err %d aq_err %s\n",
3687 (ena ? "ON" : "OFF"), status,
3688 ice_aq_str(hw->adminq.sq_last_status));
3696 * ice_vsi_add_vlan_zero - add VLAN 0 filter(s) for this VSI
3697 * @vsi: VSI used to add VLAN filters
3699 * In Single VLAN Mode (SVM), single VLAN filters via ICE_SW_LKUP_VLAN are based
3700 * on the inner VLAN ID, so the VLAN TPID (i.e. 0x8100 or 0x888a8) doesn't
3701 * matter. In Double VLAN Mode (DVM), outer/single VLAN filters via
3702 * ICE_SW_LKUP_VLAN are based on the outer/single VLAN ID + VLAN TPID.
3704 * For both modes add a VLAN 0 + no VLAN TPID filter to handle untagged traffic
3705 * when VLAN pruning is enabled. Also, this handles VLAN 0 priority tagged
3706 * traffic in SVM, since the VLAN TPID isn't part of filtering.
3708 * If DVM is enabled then an explicit VLAN 0 + VLAN TPID filter needs to be
3709 * added to allow VLAN 0 priority tagged traffic in DVM, since the VLAN TPID is
3710 * part of filtering.
3712 int ice_vsi_add_vlan_zero(struct ice_vsi *vsi)
3714 struct ice_vsi_vlan_ops *vlan_ops = ice_get_compat_vsi_vlan_ops(vsi);
3715 struct ice_vlan vlan;
3718 vlan = ICE_VLAN(0, 0, 0);
3719 err = vlan_ops->add_vlan(vsi, &vlan);
3720 if (err && err != -EEXIST)
3723 /* in SVM both VLAN 0 filters are identical */
3724 if (!ice_is_dvm_ena(&vsi->back->hw))
3727 vlan = ICE_VLAN(ETH_P_8021Q, 0, 0);
3728 err = vlan_ops->add_vlan(vsi, &vlan);
3729 if (err && err != -EEXIST)
3736 * ice_vsi_del_vlan_zero - delete VLAN 0 filter(s) for this VSI
3737 * @vsi: VSI used to add VLAN filters
3739 * Delete the VLAN 0 filters in the same manner that they were added in
3740 * ice_vsi_add_vlan_zero.
3742 int ice_vsi_del_vlan_zero(struct ice_vsi *vsi)
3744 struct ice_vsi_vlan_ops *vlan_ops = ice_get_compat_vsi_vlan_ops(vsi);
3745 struct ice_vlan vlan;
3748 vlan = ICE_VLAN(0, 0, 0);
3749 err = vlan_ops->del_vlan(vsi, &vlan);
3750 if (err && err != -EEXIST)
3753 /* in SVM both VLAN 0 filters are identical */
3754 if (!ice_is_dvm_ena(&vsi->back->hw))
3757 vlan = ICE_VLAN(ETH_P_8021Q, 0, 0);
3758 err = vlan_ops->del_vlan(vsi, &vlan);
3759 if (err && err != -EEXIST)
3762 /* when deleting the last VLAN filter, make sure to disable the VLAN
3763 * promisc mode so the filter isn't left by accident
3765 return ice_clear_vsi_promisc(&vsi->back->hw, vsi->idx,
3766 ICE_MCAST_VLAN_PROMISC_BITS, 0);
3770 * ice_vsi_num_zero_vlans - get number of VLAN 0 filters based on VLAN mode
3771 * @vsi: VSI used to get the VLAN mode
3773 * If DVM is enabled then 2 VLAN 0 filters are added, else if SVM is enabled
3774 * then 1 VLAN 0 filter is added. See ice_vsi_add_vlan_zero for more details.
3776 static u16 ice_vsi_num_zero_vlans(struct ice_vsi *vsi)
3778 #define ICE_DVM_NUM_ZERO_VLAN_FLTRS 2
3779 #define ICE_SVM_NUM_ZERO_VLAN_FLTRS 1
3780 /* no VLAN 0 filter is created when a port VLAN is active */
3781 if (vsi->type == ICE_VSI_VF) {
3782 if (WARN_ON(!vsi->vf))
3785 if (ice_vf_is_port_vlan_ena(vsi->vf))
3789 if (ice_is_dvm_ena(&vsi->back->hw))
3790 return ICE_DVM_NUM_ZERO_VLAN_FLTRS;
3792 return ICE_SVM_NUM_ZERO_VLAN_FLTRS;
3796 * ice_vsi_has_non_zero_vlans - check if VSI has any non-zero VLANs
3797 * @vsi: VSI used to determine if any non-zero VLANs have been added
3799 bool ice_vsi_has_non_zero_vlans(struct ice_vsi *vsi)
3801 return (vsi->num_vlan > ice_vsi_num_zero_vlans(vsi));
3805 * ice_vsi_num_non_zero_vlans - get the number of non-zero VLANs for this VSI
3806 * @vsi: VSI used to get the number of non-zero VLANs added
3808 u16 ice_vsi_num_non_zero_vlans(struct ice_vsi *vsi)
3810 return (vsi->num_vlan - ice_vsi_num_zero_vlans(vsi));
3814 * ice_is_feature_supported
3815 * @pf: pointer to the struct ice_pf instance
3816 * @f: feature enum to be checked
3818 * returns true if feature is supported, false otherwise
3820 bool ice_is_feature_supported(struct ice_pf *pf, enum ice_feature f)
3822 if (f < 0 || f >= ICE_F_MAX)
3825 return test_bit(f, pf->features);
3829 * ice_set_feature_support
3830 * @pf: pointer to the struct ice_pf instance
3831 * @f: feature enum to set
3833 void ice_set_feature_support(struct ice_pf *pf, enum ice_feature f)
3835 if (f < 0 || f >= ICE_F_MAX)
3838 set_bit(f, pf->features);
3842 * ice_clear_feature_support
3843 * @pf: pointer to the struct ice_pf instance
3844 * @f: feature enum to clear
3846 void ice_clear_feature_support(struct ice_pf *pf, enum ice_feature f)
3848 if (f < 0 || f >= ICE_F_MAX)
3851 clear_bit(f, pf->features);
3855 * ice_init_feature_support
3856 * @pf: pointer to the struct ice_pf instance
3858 * called during init to setup supported feature
3860 void ice_init_feature_support(struct ice_pf *pf)
3862 switch (pf->hw.device_id) {
3863 case ICE_DEV_ID_E810C_BACKPLANE:
3864 case ICE_DEV_ID_E810C_QSFP:
3865 case ICE_DEV_ID_E810C_SFP:
3866 case ICE_DEV_ID_E810_XXV_BACKPLANE:
3867 case ICE_DEV_ID_E810_XXV_QSFP:
3868 case ICE_DEV_ID_E810_XXV_SFP:
3869 ice_set_feature_support(pf, ICE_F_DSCP);
3870 if (ice_is_phy_rclk_in_netlist(&pf->hw))
3871 ice_set_feature_support(pf, ICE_F_PHY_RCLK);
3872 /* If we don't own the timer - don't enable other caps */
3873 if (!ice_pf_src_tmr_owned(pf))
3875 if (ice_is_cgu_in_netlist(&pf->hw))
3876 ice_set_feature_support(pf, ICE_F_CGU);
3877 if (ice_is_clock_mux_in_netlist(&pf->hw))
3878 ice_set_feature_support(pf, ICE_F_SMA_CTRL);
3879 if (ice_gnss_is_gps_present(&pf->hw))
3880 ice_set_feature_support(pf, ICE_F_GNSS);
3886 if (pf->hw.mac_type == ICE_MAC_E830)
3887 ice_set_feature_support(pf, ICE_F_MBX_LIMIT);
3891 * ice_vsi_update_security - update security block in VSI
3892 * @vsi: pointer to VSI structure
3893 * @fill: function pointer to fill ctx
3896 ice_vsi_update_security(struct ice_vsi *vsi, void (*fill)(struct ice_vsi_ctx *))
3898 struct ice_vsi_ctx ctx = { 0 };
3900 ctx.info = vsi->info;
3901 ctx.info.valid_sections = cpu_to_le16(ICE_AQ_VSI_PROP_SECURITY_VALID);
3904 if (ice_update_vsi(&vsi->back->hw, vsi->idx, &ctx, NULL))
3907 vsi->info = ctx.info;
3912 * ice_vsi_ctx_set_antispoof - set antispoof function in VSI ctx
3913 * @ctx: pointer to VSI ctx structure
3915 void ice_vsi_ctx_set_antispoof(struct ice_vsi_ctx *ctx)
3917 ctx->info.sec_flags |= ICE_AQ_VSI_SEC_FLAG_ENA_MAC_ANTI_SPOOF |
3918 (ICE_AQ_VSI_SEC_TX_VLAN_PRUNE_ENA <<
3919 ICE_AQ_VSI_SEC_TX_PRUNE_ENA_S);
3923 * ice_vsi_ctx_clear_antispoof - clear antispoof function in VSI ctx
3924 * @ctx: pointer to VSI ctx structure
3926 void ice_vsi_ctx_clear_antispoof(struct ice_vsi_ctx *ctx)
3928 ctx->info.sec_flags &= ~ICE_AQ_VSI_SEC_FLAG_ENA_MAC_ANTI_SPOOF &
3929 ~(ICE_AQ_VSI_SEC_TX_VLAN_PRUNE_ENA <<
3930 ICE_AQ_VSI_SEC_TX_PRUNE_ENA_S);
3934 * ice_vsi_ctx_set_allow_override - allow destination override on VSI
3935 * @ctx: pointer to VSI ctx structure
3937 void ice_vsi_ctx_set_allow_override(struct ice_vsi_ctx *ctx)
3939 ctx->info.sec_flags |= ICE_AQ_VSI_SEC_FLAG_ALLOW_DEST_OVRD;
3943 * ice_vsi_ctx_clear_allow_override - turn off destination override on VSI
3944 * @ctx: pointer to VSI ctx structure
3946 void ice_vsi_ctx_clear_allow_override(struct ice_vsi_ctx *ctx)
3948 ctx->info.sec_flags &= ~ICE_AQ_VSI_SEC_FLAG_ALLOW_DEST_OVRD;
3952 * ice_vsi_update_local_lb - update sw block in VSI with local loopback bit
3953 * @vsi: pointer to VSI structure
3954 * @set: set or unset the bit
3957 ice_vsi_update_local_lb(struct ice_vsi *vsi, bool set)
3959 struct ice_vsi_ctx ctx = {
3963 ctx.info.valid_sections = cpu_to_le16(ICE_AQ_VSI_PROP_SW_VALID);
3965 ctx.info.sw_flags |= ICE_AQ_VSI_SW_FLAG_LOCAL_LB;
3967 ctx.info.sw_flags &= ~ICE_AQ_VSI_SW_FLAG_LOCAL_LB;
3969 if (ice_update_vsi(&vsi->back->hw, vsi->idx, &ctx, NULL))
3972 vsi->info = ctx.info;