2 * Copyright (c) 2015, Mellanox Technologies. All rights reserved.
4 * This software is available to you under a choice of one of two
5 * licenses. You may choose to be licensed under the terms of the GNU
6 * General Public License (GPL) Version 2, available from the file
7 * COPYING in the main directory of this source tree, or the
8 * OpenIB.org BSD license below:
10 * Redistribution and use in source and binary forms, with or
11 * without modification, are permitted provided that the following
14 * - Redistributions of source code must retain the above
15 * copyright notice, this list of conditions and the following
18 * - Redistributions in binary form must reproduce the above
19 * copyright notice, this list of conditions and the following
20 * disclaimer in the documentation and/or other materials
21 * provided with the distribution.
23 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
24 * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
25 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
26 * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
27 * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
28 * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
29 * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
33 #include <linux/etherdevice.h>
34 #include <linux/mlx5/driver.h>
35 #include <linux/mlx5/mlx5_ifc.h>
36 #include <linux/mlx5/vport.h>
37 #include <linux/mlx5/fs.h>
38 #include "mlx5_core.h"
41 #define UPLINK_VPORT 0xFFFF
49 /* Vport UC/MC hash node */
51 struct l2addr_node node;
54 struct mlx5_flow_handle *flow_rule;
55 bool mpfs; /* UC MAC was added to MPFs */
56 /* A flag indicating that mac was added due to mc promiscuous vport */
61 UC_ADDR_CHANGE = BIT(0),
62 MC_ADDR_CHANGE = BIT(1),
63 PROMISC_CHANGE = BIT(3),
66 /* Vport context events */
67 #define SRIOV_VPORT_EVENTS (UC_ADDR_CHANGE | \
71 static int arm_vport_context_events_cmd(struct mlx5_core_dev *dev, u16 vport,
74 int in[MLX5_ST_SZ_DW(modify_nic_vport_context_in)] = {0};
75 int out[MLX5_ST_SZ_DW(modify_nic_vport_context_out)] = {0};
78 MLX5_SET(modify_nic_vport_context_in, in,
79 opcode, MLX5_CMD_OP_MODIFY_NIC_VPORT_CONTEXT);
80 MLX5_SET(modify_nic_vport_context_in, in, field_select.change_event, 1);
81 MLX5_SET(modify_nic_vport_context_in, in, vport_number, vport);
82 MLX5_SET(modify_nic_vport_context_in, in, other_vport, 1);
83 nic_vport_ctx = MLX5_ADDR_OF(modify_nic_vport_context_in,
84 in, nic_vport_context);
86 MLX5_SET(nic_vport_context, nic_vport_ctx, arm_change_event, 1);
88 if (events_mask & UC_ADDR_CHANGE)
89 MLX5_SET(nic_vport_context, nic_vport_ctx,
90 event_on_uc_address_change, 1);
91 if (events_mask & MC_ADDR_CHANGE)
92 MLX5_SET(nic_vport_context, nic_vport_ctx,
93 event_on_mc_address_change, 1);
94 if (events_mask & PROMISC_CHANGE)
95 MLX5_SET(nic_vport_context, nic_vport_ctx,
96 event_on_promisc_change, 1);
98 return mlx5_cmd_exec(dev, in, sizeof(in), out, sizeof(out));
101 /* E-Switch vport context HW commands */
102 static int modify_esw_vport_context_cmd(struct mlx5_core_dev *dev, u16 vport,
105 u32 out[MLX5_ST_SZ_DW(modify_esw_vport_context_out)] = {0};
107 MLX5_SET(modify_esw_vport_context_in, in, opcode,
108 MLX5_CMD_OP_MODIFY_ESW_VPORT_CONTEXT);
109 MLX5_SET(modify_esw_vport_context_in, in, vport_number, vport);
110 MLX5_SET(modify_esw_vport_context_in, in, other_vport, 1);
111 return mlx5_cmd_exec(dev, in, inlen, out, sizeof(out));
114 static int modify_esw_vport_cvlan(struct mlx5_core_dev *dev, u32 vport,
115 u16 vlan, u8 qos, u8 set_flags)
117 u32 in[MLX5_ST_SZ_DW(modify_esw_vport_context_in)] = {0};
119 if (!MLX5_CAP_ESW(dev, vport_cvlan_strip) ||
120 !MLX5_CAP_ESW(dev, vport_cvlan_insert_if_not_exist))
123 esw_debug(dev, "Set Vport[%d] VLAN %d qos %d set=%x\n",
124 vport, vlan, qos, set_flags);
126 if (set_flags & SET_VLAN_STRIP)
127 MLX5_SET(modify_esw_vport_context_in, in,
128 esw_vport_context.vport_cvlan_strip, 1);
130 if (set_flags & SET_VLAN_INSERT) {
131 /* insert only if no vlan in packet */
132 MLX5_SET(modify_esw_vport_context_in, in,
133 esw_vport_context.vport_cvlan_insert, 1);
135 MLX5_SET(modify_esw_vport_context_in, in,
136 esw_vport_context.cvlan_pcp, qos);
137 MLX5_SET(modify_esw_vport_context_in, in,
138 esw_vport_context.cvlan_id, vlan);
141 MLX5_SET(modify_esw_vport_context_in, in,
142 field_select.vport_cvlan_strip, 1);
143 MLX5_SET(modify_esw_vport_context_in, in,
144 field_select.vport_cvlan_insert, 1);
146 return modify_esw_vport_context_cmd(dev, vport, in, sizeof(in));
150 static struct mlx5_flow_handle *
151 __esw_fdb_set_vport_rule(struct mlx5_eswitch *esw, u32 vport, bool rx_rule,
152 u8 mac_c[ETH_ALEN], u8 mac_v[ETH_ALEN])
154 int match_header = (is_zero_ether_addr(mac_c) ? 0 :
155 MLX5_MATCH_OUTER_HEADERS);
156 struct mlx5_flow_handle *flow_rule = NULL;
157 struct mlx5_flow_act flow_act = {0};
158 struct mlx5_flow_destination dest;
159 struct mlx5_flow_spec *spec;
160 void *mv_misc = NULL;
161 void *mc_misc = NULL;
166 match_header |= MLX5_MATCH_MISC_PARAMETERS;
168 spec = kvzalloc(sizeof(*spec), GFP_KERNEL);
172 dmac_v = MLX5_ADDR_OF(fte_match_param, spec->match_value,
173 outer_headers.dmac_47_16);
174 dmac_c = MLX5_ADDR_OF(fte_match_param, spec->match_criteria,
175 outer_headers.dmac_47_16);
177 if (match_header & MLX5_MATCH_OUTER_HEADERS) {
178 ether_addr_copy(dmac_v, mac_v);
179 ether_addr_copy(dmac_c, mac_c);
182 if (match_header & MLX5_MATCH_MISC_PARAMETERS) {
183 mv_misc = MLX5_ADDR_OF(fte_match_param, spec->match_value,
185 mc_misc = MLX5_ADDR_OF(fte_match_param, spec->match_criteria,
187 MLX5_SET(fte_match_set_misc, mv_misc, source_port, UPLINK_VPORT);
188 MLX5_SET_TO_ONES(fte_match_set_misc, mc_misc, source_port);
191 dest.type = MLX5_FLOW_DESTINATION_TYPE_VPORT;
192 dest.vport_num = vport;
195 "\tFDB add rule dmac_v(%pM) dmac_c(%pM) -> vport(%d)\n",
196 dmac_v, dmac_c, vport);
197 spec->match_criteria_enable = match_header;
198 flow_act.action = MLX5_FLOW_CONTEXT_ACTION_FWD_DEST;
200 mlx5_add_flow_rules(esw->fdb_table.fdb, spec,
201 &flow_act, &dest, 1);
202 if (IS_ERR(flow_rule)) {
204 "FDB: Failed to add flow rule: dmac_v(%pM) dmac_c(%pM) -> vport(%d), err(%ld)\n",
205 dmac_v, dmac_c, vport, PTR_ERR(flow_rule));
213 static struct mlx5_flow_handle *
214 esw_fdb_set_vport_rule(struct mlx5_eswitch *esw, u8 mac[ETH_ALEN], u32 vport)
218 eth_broadcast_addr(mac_c);
219 return __esw_fdb_set_vport_rule(esw, vport, false, mac_c, mac);
222 static struct mlx5_flow_handle *
223 esw_fdb_set_vport_allmulti_rule(struct mlx5_eswitch *esw, u32 vport)
228 eth_zero_addr(mac_c);
229 eth_zero_addr(mac_v);
232 return __esw_fdb_set_vport_rule(esw, vport, false, mac_c, mac_v);
235 static struct mlx5_flow_handle *
236 esw_fdb_set_vport_promisc_rule(struct mlx5_eswitch *esw, u32 vport)
241 eth_zero_addr(mac_c);
242 eth_zero_addr(mac_v);
243 return __esw_fdb_set_vport_rule(esw, vport, true, mac_c, mac_v);
246 static int esw_create_legacy_fdb_table(struct mlx5_eswitch *esw, int nvports)
248 int inlen = MLX5_ST_SZ_BYTES(create_flow_group_in);
249 struct mlx5_flow_table_attr ft_attr = {};
250 struct mlx5_core_dev *dev = esw->dev;
251 struct mlx5_flow_namespace *root_ns;
252 struct mlx5_flow_table *fdb;
253 struct mlx5_flow_group *g;
254 void *match_criteria;
260 esw_debug(dev, "Create FDB log_max_size(%d)\n",
261 MLX5_CAP_ESW_FLOWTABLE_FDB(dev, log_max_ft_size));
263 root_ns = mlx5_get_flow_namespace(dev, MLX5_FLOW_NAMESPACE_FDB);
265 esw_warn(dev, "Failed to get FDB flow namespace\n");
269 flow_group_in = kvzalloc(inlen, GFP_KERNEL);
273 table_size = BIT(MLX5_CAP_ESW_FLOWTABLE_FDB(dev, log_max_ft_size));
275 ft_attr.max_fte = table_size;
276 fdb = mlx5_create_flow_table(root_ns, &ft_attr);
279 esw_warn(dev, "Failed to create FDB Table err %d\n", err);
282 esw->fdb_table.fdb = fdb;
284 /* Addresses group : Full match unicast/multicast addresses */
285 MLX5_SET(create_flow_group_in, flow_group_in, match_criteria_enable,
286 MLX5_MATCH_OUTER_HEADERS);
287 match_criteria = MLX5_ADDR_OF(create_flow_group_in, flow_group_in, match_criteria);
288 dmac = MLX5_ADDR_OF(fte_match_param, match_criteria, outer_headers.dmac_47_16);
289 MLX5_SET(create_flow_group_in, flow_group_in, start_flow_index, 0);
290 /* Preserve 2 entries for allmulti and promisc rules*/
291 MLX5_SET(create_flow_group_in, flow_group_in, end_flow_index, table_size - 3);
292 eth_broadcast_addr(dmac);
293 g = mlx5_create_flow_group(fdb, flow_group_in);
296 esw_warn(dev, "Failed to create flow group err(%d)\n", err);
299 esw->fdb_table.legacy.addr_grp = g;
301 /* Allmulti group : One rule that forwards any mcast traffic */
302 MLX5_SET(create_flow_group_in, flow_group_in, match_criteria_enable,
303 MLX5_MATCH_OUTER_HEADERS);
304 MLX5_SET(create_flow_group_in, flow_group_in, start_flow_index, table_size - 2);
305 MLX5_SET(create_flow_group_in, flow_group_in, end_flow_index, table_size - 2);
308 g = mlx5_create_flow_group(fdb, flow_group_in);
311 esw_warn(dev, "Failed to create allmulti flow group err(%d)\n", err);
314 esw->fdb_table.legacy.allmulti_grp = g;
316 /* Promiscuous group :
317 * One rule that forward all unmatched traffic from previous groups
320 MLX5_SET(create_flow_group_in, flow_group_in, match_criteria_enable,
321 MLX5_MATCH_MISC_PARAMETERS);
322 MLX5_SET_TO_ONES(fte_match_param, match_criteria, misc_parameters.source_port);
323 MLX5_SET(create_flow_group_in, flow_group_in, start_flow_index, table_size - 1);
324 MLX5_SET(create_flow_group_in, flow_group_in, end_flow_index, table_size - 1);
325 g = mlx5_create_flow_group(fdb, flow_group_in);
328 esw_warn(dev, "Failed to create promisc flow group err(%d)\n", err);
331 esw->fdb_table.legacy.promisc_grp = g;
335 if (!IS_ERR_OR_NULL(esw->fdb_table.legacy.allmulti_grp)) {
336 mlx5_destroy_flow_group(esw->fdb_table.legacy.allmulti_grp);
337 esw->fdb_table.legacy.allmulti_grp = NULL;
339 if (!IS_ERR_OR_NULL(esw->fdb_table.legacy.addr_grp)) {
340 mlx5_destroy_flow_group(esw->fdb_table.legacy.addr_grp);
341 esw->fdb_table.legacy.addr_grp = NULL;
343 if (!IS_ERR_OR_NULL(esw->fdb_table.fdb)) {
344 mlx5_destroy_flow_table(esw->fdb_table.fdb);
345 esw->fdb_table.fdb = NULL;
349 kvfree(flow_group_in);
353 static void esw_destroy_legacy_fdb_table(struct mlx5_eswitch *esw)
355 if (!esw->fdb_table.fdb)
358 esw_debug(esw->dev, "Destroy FDB Table\n");
359 mlx5_destroy_flow_group(esw->fdb_table.legacy.promisc_grp);
360 mlx5_destroy_flow_group(esw->fdb_table.legacy.allmulti_grp);
361 mlx5_destroy_flow_group(esw->fdb_table.legacy.addr_grp);
362 mlx5_destroy_flow_table(esw->fdb_table.fdb);
363 esw->fdb_table.fdb = NULL;
364 esw->fdb_table.legacy.addr_grp = NULL;
365 esw->fdb_table.legacy.allmulti_grp = NULL;
366 esw->fdb_table.legacy.promisc_grp = NULL;
369 /* E-Switch vport UC/MC lists management */
370 typedef int (*vport_addr_action)(struct mlx5_eswitch *esw,
371 struct vport_addr *vaddr);
373 static int esw_add_uc_addr(struct mlx5_eswitch *esw, struct vport_addr *vaddr)
375 u8 *mac = vaddr->node.addr;
376 u32 vport = vaddr->vport;
379 /* Skip mlx5_mpfs_add_mac for PFs,
380 * it is already done by the PF netdev in mlx5e_execute_l2_action
385 err = mlx5_mpfs_add_mac(esw->dev, mac);
388 "Failed to add L2 table mac(%pM) for vport(%d), err(%d)\n",
395 /* SRIOV is enabled: Forward UC MAC to vport */
396 if (esw->fdb_table.fdb && esw->mode == SRIOV_LEGACY)
397 vaddr->flow_rule = esw_fdb_set_vport_rule(esw, mac, vport);
399 esw_debug(esw->dev, "\tADDED UC MAC: vport[%d] %pM fr(%p)\n",
400 vport, mac, vaddr->flow_rule);
405 static int esw_del_uc_addr(struct mlx5_eswitch *esw, struct vport_addr *vaddr)
407 u8 *mac = vaddr->node.addr;
408 u32 vport = vaddr->vport;
411 /* Skip mlx5_mpfs_del_mac for PFs,
412 * it is already done by the PF netdev in mlx5e_execute_l2_action
414 if (!vport || !vaddr->mpfs)
417 err = mlx5_mpfs_del_mac(esw->dev, mac);
420 "Failed to del L2 table mac(%pM) for vport(%d), err(%d)\n",
425 if (vaddr->flow_rule)
426 mlx5_del_flow_rules(vaddr->flow_rule);
427 vaddr->flow_rule = NULL;
432 static void update_allmulti_vports(struct mlx5_eswitch *esw,
433 struct vport_addr *vaddr,
434 struct esw_mc_addr *esw_mc)
436 u8 *mac = vaddr->node.addr;
439 for (vport_idx = 0; vport_idx < esw->total_vports; vport_idx++) {
440 struct mlx5_vport *vport = &esw->vports[vport_idx];
441 struct hlist_head *vport_hash = vport->mc_list;
442 struct vport_addr *iter_vaddr =
443 l2addr_hash_find(vport_hash,
446 if (IS_ERR_OR_NULL(vport->allmulti_rule) ||
447 vaddr->vport == vport_idx)
449 switch (vaddr->action) {
450 case MLX5_ACTION_ADD:
453 iter_vaddr = l2addr_hash_add(vport_hash, mac,
458 "ALL-MULTI: Failed to add MAC(%pM) to vport[%d] DB\n",
462 iter_vaddr->vport = vport_idx;
463 iter_vaddr->flow_rule =
464 esw_fdb_set_vport_rule(esw,
467 iter_vaddr->mc_promisc = true;
469 case MLX5_ACTION_DEL:
472 mlx5_del_flow_rules(iter_vaddr->flow_rule);
473 l2addr_hash_del(iter_vaddr);
479 static int esw_add_mc_addr(struct mlx5_eswitch *esw, struct vport_addr *vaddr)
481 struct hlist_head *hash = esw->mc_table;
482 struct esw_mc_addr *esw_mc;
483 u8 *mac = vaddr->node.addr;
484 u32 vport = vaddr->vport;
486 if (!esw->fdb_table.fdb)
489 esw_mc = l2addr_hash_find(hash, mac, struct esw_mc_addr);
493 esw_mc = l2addr_hash_add(hash, mac, struct esw_mc_addr, GFP_KERNEL);
497 esw_mc->uplink_rule = /* Forward MC MAC to Uplink */
498 esw_fdb_set_vport_rule(esw, mac, UPLINK_VPORT);
500 /* Add this multicast mac to all the mc promiscuous vports */
501 update_allmulti_vports(esw, vaddr, esw_mc);
504 /* If the multicast mac is added as a result of mc promiscuous vport,
505 * don't increment the multicast ref count
507 if (!vaddr->mc_promisc)
510 /* Forward MC MAC to vport */
511 vaddr->flow_rule = esw_fdb_set_vport_rule(esw, mac, vport);
513 "\tADDED MC MAC: vport[%d] %pM fr(%p) refcnt(%d) uplinkfr(%p)\n",
514 vport, mac, vaddr->flow_rule,
515 esw_mc->refcnt, esw_mc->uplink_rule);
519 static int esw_del_mc_addr(struct mlx5_eswitch *esw, struct vport_addr *vaddr)
521 struct hlist_head *hash = esw->mc_table;
522 struct esw_mc_addr *esw_mc;
523 u8 *mac = vaddr->node.addr;
524 u32 vport = vaddr->vport;
526 if (!esw->fdb_table.fdb)
529 esw_mc = l2addr_hash_find(hash, mac, struct esw_mc_addr);
532 "Failed to find eswitch MC addr for MAC(%pM) vport(%d)",
537 "\tDELETE MC MAC: vport[%d] %pM fr(%p) refcnt(%d) uplinkfr(%p)\n",
538 vport, mac, vaddr->flow_rule, esw_mc->refcnt,
539 esw_mc->uplink_rule);
541 if (vaddr->flow_rule)
542 mlx5_del_flow_rules(vaddr->flow_rule);
543 vaddr->flow_rule = NULL;
545 /* If the multicast mac is added as a result of mc promiscuous vport,
546 * don't decrement the multicast ref count.
548 if (vaddr->mc_promisc || (--esw_mc->refcnt > 0))
551 /* Remove this multicast mac from all the mc promiscuous vports */
552 update_allmulti_vports(esw, vaddr, esw_mc);
554 if (esw_mc->uplink_rule)
555 mlx5_del_flow_rules(esw_mc->uplink_rule);
557 l2addr_hash_del(esw_mc);
561 /* Apply vport UC/MC list to HW l2 table and FDB table */
562 static void esw_apply_vport_addr_list(struct mlx5_eswitch *esw,
563 u32 vport_num, int list_type)
565 struct mlx5_vport *vport = &esw->vports[vport_num];
566 bool is_uc = list_type == MLX5_NVPRT_LIST_TYPE_UC;
567 vport_addr_action vport_addr_add;
568 vport_addr_action vport_addr_del;
569 struct vport_addr *addr;
570 struct l2addr_node *node;
571 struct hlist_head *hash;
572 struct hlist_node *tmp;
575 vport_addr_add = is_uc ? esw_add_uc_addr :
577 vport_addr_del = is_uc ? esw_del_uc_addr :
580 hash = is_uc ? vport->uc_list : vport->mc_list;
581 for_each_l2hash_node(node, tmp, hash, hi) {
582 addr = container_of(node, struct vport_addr, node);
583 switch (addr->action) {
584 case MLX5_ACTION_ADD:
585 vport_addr_add(esw, addr);
586 addr->action = MLX5_ACTION_NONE;
588 case MLX5_ACTION_DEL:
589 vport_addr_del(esw, addr);
590 l2addr_hash_del(addr);
596 /* Sync vport UC/MC list from vport context */
597 static void esw_update_vport_addr_list(struct mlx5_eswitch *esw,
598 u32 vport_num, int list_type)
600 struct mlx5_vport *vport = &esw->vports[vport_num];
601 bool is_uc = list_type == MLX5_NVPRT_LIST_TYPE_UC;
602 u8 (*mac_list)[ETH_ALEN];
603 struct l2addr_node *node;
604 struct vport_addr *addr;
605 struct hlist_head *hash;
606 struct hlist_node *tmp;
612 size = is_uc ? MLX5_MAX_UC_PER_VPORT(esw->dev) :
613 MLX5_MAX_MC_PER_VPORT(esw->dev);
615 mac_list = kcalloc(size, ETH_ALEN, GFP_KERNEL);
619 hash = is_uc ? vport->uc_list : vport->mc_list;
621 for_each_l2hash_node(node, tmp, hash, hi) {
622 addr = container_of(node, struct vport_addr, node);
623 addr->action = MLX5_ACTION_DEL;
629 err = mlx5_query_nic_vport_mac_list(esw->dev, vport_num, list_type,
633 esw_debug(esw->dev, "vport[%d] context update %s list size (%d)\n",
634 vport_num, is_uc ? "UC" : "MC", size);
636 for (i = 0; i < size; i++) {
637 if (is_uc && !is_valid_ether_addr(mac_list[i]))
640 if (!is_uc && !is_multicast_ether_addr(mac_list[i]))
643 addr = l2addr_hash_find(hash, mac_list[i], struct vport_addr);
645 addr->action = MLX5_ACTION_NONE;
646 /* If this mac was previously added because of allmulti
647 * promiscuous rx mode, its now converted to be original
650 if (addr->mc_promisc) {
651 struct esw_mc_addr *esw_mc =
652 l2addr_hash_find(esw->mc_table,
657 "Failed to MAC(%pM) in mcast DB\n",
662 addr->mc_promisc = false;
667 addr = l2addr_hash_add(hash, mac_list[i], struct vport_addr,
671 "Failed to add MAC(%pM) to vport[%d] DB\n",
672 mac_list[i], vport_num);
675 addr->vport = vport_num;
676 addr->action = MLX5_ACTION_ADD;
682 /* Sync vport UC/MC list from vport context
683 * Must be called after esw_update_vport_addr_list
685 static void esw_update_vport_mc_promisc(struct mlx5_eswitch *esw, u32 vport_num)
687 struct mlx5_vport *vport = &esw->vports[vport_num];
688 struct l2addr_node *node;
689 struct vport_addr *addr;
690 struct hlist_head *hash;
691 struct hlist_node *tmp;
694 hash = vport->mc_list;
696 for_each_l2hash_node(node, tmp, esw->mc_table, hi) {
697 u8 *mac = node->addr;
699 addr = l2addr_hash_find(hash, mac, struct vport_addr);
701 if (addr->action == MLX5_ACTION_DEL)
702 addr->action = MLX5_ACTION_NONE;
705 addr = l2addr_hash_add(hash, mac, struct vport_addr,
709 "Failed to add allmulti MAC(%pM) to vport[%d] DB\n",
713 addr->vport = vport_num;
714 addr->action = MLX5_ACTION_ADD;
715 addr->mc_promisc = true;
719 /* Apply vport rx mode to HW FDB table */
720 static void esw_apply_vport_rx_mode(struct mlx5_eswitch *esw, u32 vport_num,
721 bool promisc, bool mc_promisc)
723 struct esw_mc_addr *allmulti_addr = &esw->mc_promisc;
724 struct mlx5_vport *vport = &esw->vports[vport_num];
726 if (IS_ERR_OR_NULL(vport->allmulti_rule) != mc_promisc)
730 vport->allmulti_rule =
731 esw_fdb_set_vport_allmulti_rule(esw, vport_num);
732 if (!allmulti_addr->uplink_rule)
733 allmulti_addr->uplink_rule =
734 esw_fdb_set_vport_allmulti_rule(esw,
736 allmulti_addr->refcnt++;
737 } else if (vport->allmulti_rule) {
738 mlx5_del_flow_rules(vport->allmulti_rule);
739 vport->allmulti_rule = NULL;
741 if (--allmulti_addr->refcnt > 0)
744 if (allmulti_addr->uplink_rule)
745 mlx5_del_flow_rules(allmulti_addr->uplink_rule);
746 allmulti_addr->uplink_rule = NULL;
750 if (IS_ERR_OR_NULL(vport->promisc_rule) != promisc)
754 vport->promisc_rule = esw_fdb_set_vport_promisc_rule(esw,
756 } else if (vport->promisc_rule) {
757 mlx5_del_flow_rules(vport->promisc_rule);
758 vport->promisc_rule = NULL;
762 /* Sync vport rx mode from vport context */
763 static void esw_update_vport_rx_mode(struct mlx5_eswitch *esw, u32 vport_num)
765 struct mlx5_vport *vport = &esw->vports[vport_num];
771 err = mlx5_query_nic_vport_promisc(esw->dev,
778 esw_debug(esw->dev, "vport[%d] context update rx mode promisc_all=%d, all_multi=%d\n",
779 vport_num, promisc_all, promisc_mc);
781 if (!vport->info.trusted || !vport->enabled) {
787 esw_apply_vport_rx_mode(esw, vport_num, promisc_all,
788 (promisc_all || promisc_mc));
791 static void esw_vport_change_handle_locked(struct mlx5_vport *vport)
793 struct mlx5_core_dev *dev = vport->dev;
794 struct mlx5_eswitch *esw = dev->priv.eswitch;
797 mlx5_query_nic_vport_mac_address(dev, vport->vport, mac);
798 esw_debug(dev, "vport[%d] Context Changed: perm mac: %pM\n",
801 if (vport->enabled_events & UC_ADDR_CHANGE) {
802 esw_update_vport_addr_list(esw, vport->vport,
803 MLX5_NVPRT_LIST_TYPE_UC);
804 esw_apply_vport_addr_list(esw, vport->vport,
805 MLX5_NVPRT_LIST_TYPE_UC);
808 if (vport->enabled_events & MC_ADDR_CHANGE) {
809 esw_update_vport_addr_list(esw, vport->vport,
810 MLX5_NVPRT_LIST_TYPE_MC);
813 if (vport->enabled_events & PROMISC_CHANGE) {
814 esw_update_vport_rx_mode(esw, vport->vport);
815 if (!IS_ERR_OR_NULL(vport->allmulti_rule))
816 esw_update_vport_mc_promisc(esw, vport->vport);
819 if (vport->enabled_events & (PROMISC_CHANGE | MC_ADDR_CHANGE)) {
820 esw_apply_vport_addr_list(esw, vport->vport,
821 MLX5_NVPRT_LIST_TYPE_MC);
824 esw_debug(esw->dev, "vport[%d] Context Changed: Done\n", vport->vport);
826 arm_vport_context_events_cmd(dev, vport->vport,
827 vport->enabled_events);
830 static void esw_vport_change_handler(struct work_struct *work)
832 struct mlx5_vport *vport =
833 container_of(work, struct mlx5_vport, vport_change_handler);
834 struct mlx5_eswitch *esw = vport->dev->priv.eswitch;
836 mutex_lock(&esw->state_lock);
837 esw_vport_change_handle_locked(vport);
838 mutex_unlock(&esw->state_lock);
841 static int esw_vport_enable_egress_acl(struct mlx5_eswitch *esw,
842 struct mlx5_vport *vport)
844 int inlen = MLX5_ST_SZ_BYTES(create_flow_group_in);
845 struct mlx5_flow_group *vlan_grp = NULL;
846 struct mlx5_flow_group *drop_grp = NULL;
847 struct mlx5_core_dev *dev = esw->dev;
848 struct mlx5_flow_namespace *root_ns;
849 struct mlx5_flow_table *acl;
850 void *match_criteria;
852 /* The egress acl table contains 2 rules:
853 * 1)Allow traffic with vlan_tag=vst_vlan_id
854 * 2)Drop all other traffic.
859 if (!MLX5_CAP_ESW_EGRESS_ACL(dev, ft_support))
862 if (!IS_ERR_OR_NULL(vport->egress.acl))
865 esw_debug(dev, "Create vport[%d] egress ACL log_max_size(%d)\n",
866 vport->vport, MLX5_CAP_ESW_EGRESS_ACL(dev, log_max_ft_size));
868 root_ns = mlx5_get_flow_namespace(dev, MLX5_FLOW_NAMESPACE_ESW_EGRESS);
870 esw_warn(dev, "Failed to get E-Switch egress flow namespace\n");
874 flow_group_in = kvzalloc(inlen, GFP_KERNEL);
878 acl = mlx5_create_vport_flow_table(root_ns, 0, table_size, 0, vport->vport);
881 esw_warn(dev, "Failed to create E-Switch vport[%d] egress flow Table, err(%d)\n",
886 MLX5_SET(create_flow_group_in, flow_group_in, match_criteria_enable, MLX5_MATCH_OUTER_HEADERS);
887 match_criteria = MLX5_ADDR_OF(create_flow_group_in, flow_group_in, match_criteria);
888 MLX5_SET_TO_ONES(fte_match_param, match_criteria, outer_headers.cvlan_tag);
889 MLX5_SET_TO_ONES(fte_match_param, match_criteria, outer_headers.first_vid);
890 MLX5_SET(create_flow_group_in, flow_group_in, start_flow_index, 0);
891 MLX5_SET(create_flow_group_in, flow_group_in, end_flow_index, 0);
893 vlan_grp = mlx5_create_flow_group(acl, flow_group_in);
894 if (IS_ERR(vlan_grp)) {
895 err = PTR_ERR(vlan_grp);
896 esw_warn(dev, "Failed to create E-Switch vport[%d] egress allowed vlans flow group, err(%d)\n",
901 memset(flow_group_in, 0, inlen);
902 MLX5_SET(create_flow_group_in, flow_group_in, start_flow_index, 1);
903 MLX5_SET(create_flow_group_in, flow_group_in, end_flow_index, 1);
904 drop_grp = mlx5_create_flow_group(acl, flow_group_in);
905 if (IS_ERR(drop_grp)) {
906 err = PTR_ERR(drop_grp);
907 esw_warn(dev, "Failed to create E-Switch vport[%d] egress drop flow group, err(%d)\n",
912 vport->egress.acl = acl;
913 vport->egress.drop_grp = drop_grp;
914 vport->egress.allowed_vlans_grp = vlan_grp;
916 kvfree(flow_group_in);
917 if (err && !IS_ERR_OR_NULL(vlan_grp))
918 mlx5_destroy_flow_group(vlan_grp);
919 if (err && !IS_ERR_OR_NULL(acl))
920 mlx5_destroy_flow_table(acl);
924 static void esw_vport_cleanup_egress_rules(struct mlx5_eswitch *esw,
925 struct mlx5_vport *vport)
927 if (!IS_ERR_OR_NULL(vport->egress.allowed_vlan))
928 mlx5_del_flow_rules(vport->egress.allowed_vlan);
930 if (!IS_ERR_OR_NULL(vport->egress.drop_rule))
931 mlx5_del_flow_rules(vport->egress.drop_rule);
933 vport->egress.allowed_vlan = NULL;
934 vport->egress.drop_rule = NULL;
937 static void esw_vport_disable_egress_acl(struct mlx5_eswitch *esw,
938 struct mlx5_vport *vport)
940 if (IS_ERR_OR_NULL(vport->egress.acl))
943 esw_debug(esw->dev, "Destroy vport[%d] E-Switch egress ACL\n", vport->vport);
945 esw_vport_cleanup_egress_rules(esw, vport);
946 mlx5_destroy_flow_group(vport->egress.allowed_vlans_grp);
947 mlx5_destroy_flow_group(vport->egress.drop_grp);
948 mlx5_destroy_flow_table(vport->egress.acl);
949 vport->egress.allowed_vlans_grp = NULL;
950 vport->egress.drop_grp = NULL;
951 vport->egress.acl = NULL;
954 static int esw_vport_enable_ingress_acl(struct mlx5_eswitch *esw,
955 struct mlx5_vport *vport)
957 int inlen = MLX5_ST_SZ_BYTES(create_flow_group_in);
958 struct mlx5_core_dev *dev = esw->dev;
959 struct mlx5_flow_namespace *root_ns;
960 struct mlx5_flow_table *acl;
961 struct mlx5_flow_group *g;
962 void *match_criteria;
964 /* The ingress acl table contains 4 groups
965 * (2 active rules at the same time -
966 * 1 allow rule from one of the first 3 groups.
967 * 1 drop rule from the last group):
968 * 1)Allow untagged traffic with smac=original mac.
969 * 2)Allow untagged traffic.
970 * 3)Allow traffic with smac=original mac.
971 * 4)Drop all other traffic.
976 if (!MLX5_CAP_ESW_INGRESS_ACL(dev, ft_support))
979 if (!IS_ERR_OR_NULL(vport->ingress.acl))
982 esw_debug(dev, "Create vport[%d] ingress ACL log_max_size(%d)\n",
983 vport->vport, MLX5_CAP_ESW_INGRESS_ACL(dev, log_max_ft_size));
985 root_ns = mlx5_get_flow_namespace(dev, MLX5_FLOW_NAMESPACE_ESW_INGRESS);
987 esw_warn(dev, "Failed to get E-Switch ingress flow namespace\n");
991 flow_group_in = kvzalloc(inlen, GFP_KERNEL);
995 acl = mlx5_create_vport_flow_table(root_ns, 0, table_size, 0, vport->vport);
998 esw_warn(dev, "Failed to create E-Switch vport[%d] ingress flow Table, err(%d)\n",
1002 vport->ingress.acl = acl;
1004 match_criteria = MLX5_ADDR_OF(create_flow_group_in, flow_group_in, match_criteria);
1006 MLX5_SET(create_flow_group_in, flow_group_in, match_criteria_enable, MLX5_MATCH_OUTER_HEADERS);
1007 MLX5_SET_TO_ONES(fte_match_param, match_criteria, outer_headers.cvlan_tag);
1008 MLX5_SET_TO_ONES(fte_match_param, match_criteria, outer_headers.smac_47_16);
1009 MLX5_SET_TO_ONES(fte_match_param, match_criteria, outer_headers.smac_15_0);
1010 MLX5_SET(create_flow_group_in, flow_group_in, start_flow_index, 0);
1011 MLX5_SET(create_flow_group_in, flow_group_in, end_flow_index, 0);
1013 g = mlx5_create_flow_group(acl, flow_group_in);
1016 esw_warn(dev, "Failed to create E-Switch vport[%d] ingress untagged spoofchk flow group, err(%d)\n",
1020 vport->ingress.allow_untagged_spoofchk_grp = g;
1022 memset(flow_group_in, 0, inlen);
1023 MLX5_SET(create_flow_group_in, flow_group_in, match_criteria_enable, MLX5_MATCH_OUTER_HEADERS);
1024 MLX5_SET_TO_ONES(fte_match_param, match_criteria, outer_headers.cvlan_tag);
1025 MLX5_SET(create_flow_group_in, flow_group_in, start_flow_index, 1);
1026 MLX5_SET(create_flow_group_in, flow_group_in, end_flow_index, 1);
1028 g = mlx5_create_flow_group(acl, flow_group_in);
1031 esw_warn(dev, "Failed to create E-Switch vport[%d] ingress untagged flow group, err(%d)\n",
1035 vport->ingress.allow_untagged_only_grp = g;
1037 memset(flow_group_in, 0, inlen);
1038 MLX5_SET(create_flow_group_in, flow_group_in, match_criteria_enable, MLX5_MATCH_OUTER_HEADERS);
1039 MLX5_SET_TO_ONES(fte_match_param, match_criteria, outer_headers.smac_47_16);
1040 MLX5_SET_TO_ONES(fte_match_param, match_criteria, outer_headers.smac_15_0);
1041 MLX5_SET(create_flow_group_in, flow_group_in, start_flow_index, 2);
1042 MLX5_SET(create_flow_group_in, flow_group_in, end_flow_index, 2);
1044 g = mlx5_create_flow_group(acl, flow_group_in);
1047 esw_warn(dev, "Failed to create E-Switch vport[%d] ingress spoofchk flow group, err(%d)\n",
1051 vport->ingress.allow_spoofchk_only_grp = g;
1053 memset(flow_group_in, 0, inlen);
1054 MLX5_SET(create_flow_group_in, flow_group_in, start_flow_index, 3);
1055 MLX5_SET(create_flow_group_in, flow_group_in, end_flow_index, 3);
1057 g = mlx5_create_flow_group(acl, flow_group_in);
1060 esw_warn(dev, "Failed to create E-Switch vport[%d] ingress drop flow group, err(%d)\n",
1064 vport->ingress.drop_grp = g;
1068 if (!IS_ERR_OR_NULL(vport->ingress.allow_spoofchk_only_grp))
1069 mlx5_destroy_flow_group(
1070 vport->ingress.allow_spoofchk_only_grp);
1071 if (!IS_ERR_OR_NULL(vport->ingress.allow_untagged_only_grp))
1072 mlx5_destroy_flow_group(
1073 vport->ingress.allow_untagged_only_grp);
1074 if (!IS_ERR_OR_NULL(vport->ingress.allow_untagged_spoofchk_grp))
1075 mlx5_destroy_flow_group(
1076 vport->ingress.allow_untagged_spoofchk_grp);
1077 if (!IS_ERR_OR_NULL(vport->ingress.acl))
1078 mlx5_destroy_flow_table(vport->ingress.acl);
1081 kvfree(flow_group_in);
1085 static void esw_vport_cleanup_ingress_rules(struct mlx5_eswitch *esw,
1086 struct mlx5_vport *vport)
1088 if (!IS_ERR_OR_NULL(vport->ingress.drop_rule))
1089 mlx5_del_flow_rules(vport->ingress.drop_rule);
1091 if (!IS_ERR_OR_NULL(vport->ingress.allow_rule))
1092 mlx5_del_flow_rules(vport->ingress.allow_rule);
1094 vport->ingress.drop_rule = NULL;
1095 vport->ingress.allow_rule = NULL;
1098 static void esw_vport_disable_ingress_acl(struct mlx5_eswitch *esw,
1099 struct mlx5_vport *vport)
1101 if (IS_ERR_OR_NULL(vport->ingress.acl))
1104 esw_debug(esw->dev, "Destroy vport[%d] E-Switch ingress ACL\n", vport->vport);
1106 esw_vport_cleanup_ingress_rules(esw, vport);
1107 mlx5_destroy_flow_group(vport->ingress.allow_spoofchk_only_grp);
1108 mlx5_destroy_flow_group(vport->ingress.allow_untagged_only_grp);
1109 mlx5_destroy_flow_group(vport->ingress.allow_untagged_spoofchk_grp);
1110 mlx5_destroy_flow_group(vport->ingress.drop_grp);
1111 mlx5_destroy_flow_table(vport->ingress.acl);
1112 vport->ingress.acl = NULL;
1113 vport->ingress.drop_grp = NULL;
1114 vport->ingress.allow_spoofchk_only_grp = NULL;
1115 vport->ingress.allow_untagged_only_grp = NULL;
1116 vport->ingress.allow_untagged_spoofchk_grp = NULL;
1119 static int esw_vport_ingress_config(struct mlx5_eswitch *esw,
1120 struct mlx5_vport *vport)
1122 struct mlx5_flow_act flow_act = {0};
1123 struct mlx5_flow_spec *spec;
1127 esw_vport_cleanup_ingress_rules(esw, vport);
1129 if (!vport->info.vlan && !vport->info.qos && !vport->info.spoofchk) {
1130 esw_vport_disable_ingress_acl(esw, vport);
1134 err = esw_vport_enable_ingress_acl(esw, vport);
1136 mlx5_core_warn(esw->dev,
1137 "failed to enable ingress acl (%d) on vport[%d]\n",
1143 "vport[%d] configure ingress rules, vlan(%d) qos(%d)\n",
1144 vport->vport, vport->info.vlan, vport->info.qos);
1146 spec = kvzalloc(sizeof(*spec), GFP_KERNEL);
1152 if (vport->info.vlan || vport->info.qos)
1153 MLX5_SET_TO_ONES(fte_match_param, spec->match_criteria, outer_headers.cvlan_tag);
1155 if (vport->info.spoofchk) {
1156 MLX5_SET_TO_ONES(fte_match_param, spec->match_criteria, outer_headers.smac_47_16);
1157 MLX5_SET_TO_ONES(fte_match_param, spec->match_criteria, outer_headers.smac_15_0);
1158 smac_v = MLX5_ADDR_OF(fte_match_param,
1160 outer_headers.smac_47_16);
1161 ether_addr_copy(smac_v, vport->info.mac);
1164 spec->match_criteria_enable = MLX5_MATCH_OUTER_HEADERS;
1165 flow_act.action = MLX5_FLOW_CONTEXT_ACTION_ALLOW;
1166 vport->ingress.allow_rule =
1167 mlx5_add_flow_rules(vport->ingress.acl, spec,
1168 &flow_act, NULL, 0);
1169 if (IS_ERR(vport->ingress.allow_rule)) {
1170 err = PTR_ERR(vport->ingress.allow_rule);
1172 "vport[%d] configure ingress allow rule, err(%d)\n",
1174 vport->ingress.allow_rule = NULL;
1178 memset(spec, 0, sizeof(*spec));
1179 flow_act.action = MLX5_FLOW_CONTEXT_ACTION_DROP;
1180 vport->ingress.drop_rule =
1181 mlx5_add_flow_rules(vport->ingress.acl, spec,
1182 &flow_act, NULL, 0);
1183 if (IS_ERR(vport->ingress.drop_rule)) {
1184 err = PTR_ERR(vport->ingress.drop_rule);
1186 "vport[%d] configure ingress drop rule, err(%d)\n",
1188 vport->ingress.drop_rule = NULL;
1194 esw_vport_cleanup_ingress_rules(esw, vport);
1199 static int esw_vport_egress_config(struct mlx5_eswitch *esw,
1200 struct mlx5_vport *vport)
1202 struct mlx5_flow_act flow_act = {0};
1203 struct mlx5_flow_spec *spec;
1206 esw_vport_cleanup_egress_rules(esw, vport);
1208 if (!vport->info.vlan && !vport->info.qos) {
1209 esw_vport_disable_egress_acl(esw, vport);
1213 err = esw_vport_enable_egress_acl(esw, vport);
1215 mlx5_core_warn(esw->dev,
1216 "failed to enable egress acl (%d) on vport[%d]\n",
1222 "vport[%d] configure egress rules, vlan(%d) qos(%d)\n",
1223 vport->vport, vport->info.vlan, vport->info.qos);
1225 spec = kvzalloc(sizeof(*spec), GFP_KERNEL);
1231 /* Allowed vlan rule */
1232 MLX5_SET_TO_ONES(fte_match_param, spec->match_criteria, outer_headers.cvlan_tag);
1233 MLX5_SET_TO_ONES(fte_match_param, spec->match_value, outer_headers.cvlan_tag);
1234 MLX5_SET_TO_ONES(fte_match_param, spec->match_criteria, outer_headers.first_vid);
1235 MLX5_SET(fte_match_param, spec->match_value, outer_headers.first_vid, vport->info.vlan);
1237 spec->match_criteria_enable = MLX5_MATCH_OUTER_HEADERS;
1238 flow_act.action = MLX5_FLOW_CONTEXT_ACTION_ALLOW;
1239 vport->egress.allowed_vlan =
1240 mlx5_add_flow_rules(vport->egress.acl, spec,
1241 &flow_act, NULL, 0);
1242 if (IS_ERR(vport->egress.allowed_vlan)) {
1243 err = PTR_ERR(vport->egress.allowed_vlan);
1245 "vport[%d] configure egress allowed vlan rule failed, err(%d)\n",
1247 vport->egress.allowed_vlan = NULL;
1251 /* Drop others rule (star rule) */
1252 memset(spec, 0, sizeof(*spec));
1253 flow_act.action = MLX5_FLOW_CONTEXT_ACTION_DROP;
1254 vport->egress.drop_rule =
1255 mlx5_add_flow_rules(vport->egress.acl, spec,
1256 &flow_act, NULL, 0);
1257 if (IS_ERR(vport->egress.drop_rule)) {
1258 err = PTR_ERR(vport->egress.drop_rule);
1260 "vport[%d] configure egress drop rule failed, err(%d)\n",
1262 vport->egress.drop_rule = NULL;
1269 /* Vport QoS management */
1270 static int esw_create_tsar(struct mlx5_eswitch *esw)
1272 u32 tsar_ctx[MLX5_ST_SZ_DW(scheduling_context)] = {0};
1273 struct mlx5_core_dev *dev = esw->dev;
1276 if (!MLX5_CAP_GEN(dev, qos) || !MLX5_CAP_QOS(dev, esw_scheduling))
1279 if (esw->qos.enabled)
1282 err = mlx5_create_scheduling_element_cmd(dev,
1283 SCHEDULING_HIERARCHY_E_SWITCH,
1285 &esw->qos.root_tsar_id);
1287 esw_warn(esw->dev, "E-Switch create TSAR failed (%d)\n", err);
1291 esw->qos.enabled = true;
1295 static void esw_destroy_tsar(struct mlx5_eswitch *esw)
1299 if (!esw->qos.enabled)
1302 err = mlx5_destroy_scheduling_element_cmd(esw->dev,
1303 SCHEDULING_HIERARCHY_E_SWITCH,
1304 esw->qos.root_tsar_id);
1306 esw_warn(esw->dev, "E-Switch destroy TSAR failed (%d)\n", err);
1308 esw->qos.enabled = false;
1311 static int esw_vport_enable_qos(struct mlx5_eswitch *esw, int vport_num,
1312 u32 initial_max_rate, u32 initial_bw_share)
1314 u32 sched_ctx[MLX5_ST_SZ_DW(scheduling_context)] = {0};
1315 struct mlx5_vport *vport = &esw->vports[vport_num];
1316 struct mlx5_core_dev *dev = esw->dev;
1320 if (!esw->qos.enabled || !MLX5_CAP_GEN(dev, qos) ||
1321 !MLX5_CAP_QOS(dev, esw_scheduling))
1324 if (vport->qos.enabled)
1327 MLX5_SET(scheduling_context, &sched_ctx, element_type,
1328 SCHEDULING_CONTEXT_ELEMENT_TYPE_VPORT);
1329 vport_elem = MLX5_ADDR_OF(scheduling_context, &sched_ctx,
1330 element_attributes);
1331 MLX5_SET(vport_element, vport_elem, vport_number, vport_num);
1332 MLX5_SET(scheduling_context, &sched_ctx, parent_element_id,
1333 esw->qos.root_tsar_id);
1334 MLX5_SET(scheduling_context, &sched_ctx, max_average_bw,
1336 MLX5_SET(scheduling_context, &sched_ctx, bw_share, initial_bw_share);
1338 err = mlx5_create_scheduling_element_cmd(dev,
1339 SCHEDULING_HIERARCHY_E_SWITCH,
1341 &vport->qos.esw_tsar_ix);
1343 esw_warn(esw->dev, "E-Switch create TSAR vport element failed (vport=%d,err=%d)\n",
1348 vport->qos.enabled = true;
1352 static void esw_vport_disable_qos(struct mlx5_eswitch *esw, int vport_num)
1354 struct mlx5_vport *vport = &esw->vports[vport_num];
1357 if (!vport->qos.enabled)
1360 err = mlx5_destroy_scheduling_element_cmd(esw->dev,
1361 SCHEDULING_HIERARCHY_E_SWITCH,
1362 vport->qos.esw_tsar_ix);
1364 esw_warn(esw->dev, "E-Switch destroy TSAR vport element failed (vport=%d,err=%d)\n",
1367 vport->qos.enabled = false;
1370 static int esw_vport_qos_config(struct mlx5_eswitch *esw, int vport_num,
1371 u32 max_rate, u32 bw_share)
1373 u32 sched_ctx[MLX5_ST_SZ_DW(scheduling_context)] = {0};
1374 struct mlx5_vport *vport = &esw->vports[vport_num];
1375 struct mlx5_core_dev *dev = esw->dev;
1380 if (!MLX5_CAP_GEN(dev, qos) || !MLX5_CAP_QOS(dev, esw_scheduling))
1383 if (!vport->qos.enabled)
1386 MLX5_SET(scheduling_context, &sched_ctx, element_type,
1387 SCHEDULING_CONTEXT_ELEMENT_TYPE_VPORT);
1388 vport_elem = MLX5_ADDR_OF(scheduling_context, &sched_ctx,
1389 element_attributes);
1390 MLX5_SET(vport_element, vport_elem, vport_number, vport_num);
1391 MLX5_SET(scheduling_context, &sched_ctx, parent_element_id,
1392 esw->qos.root_tsar_id);
1393 MLX5_SET(scheduling_context, &sched_ctx, max_average_bw,
1395 MLX5_SET(scheduling_context, &sched_ctx, bw_share, bw_share);
1396 bitmask |= MODIFY_SCHEDULING_ELEMENT_IN_MODIFY_BITMASK_MAX_AVERAGE_BW;
1397 bitmask |= MODIFY_SCHEDULING_ELEMENT_IN_MODIFY_BITMASK_BW_SHARE;
1399 err = mlx5_modify_scheduling_element_cmd(dev,
1400 SCHEDULING_HIERARCHY_E_SWITCH,
1402 vport->qos.esw_tsar_ix,
1405 esw_warn(esw->dev, "E-Switch modify TSAR vport element failed (vport=%d,err=%d)\n",
1413 static void node_guid_gen_from_mac(u64 *node_guid, u8 mac[ETH_ALEN])
1415 ((u8 *)node_guid)[7] = mac[0];
1416 ((u8 *)node_guid)[6] = mac[1];
1417 ((u8 *)node_guid)[5] = mac[2];
1418 ((u8 *)node_guid)[4] = 0xff;
1419 ((u8 *)node_guid)[3] = 0xfe;
1420 ((u8 *)node_guid)[2] = mac[3];
1421 ((u8 *)node_guid)[1] = mac[4];
1422 ((u8 *)node_guid)[0] = mac[5];
1425 static void esw_apply_vport_conf(struct mlx5_eswitch *esw,
1426 struct mlx5_vport *vport)
1428 int vport_num = vport->vport;
1433 mlx5_modify_vport_admin_state(esw->dev,
1434 MLX5_QUERY_VPORT_STATE_IN_OP_MOD_ESW_VPORT,
1436 vport->info.link_state);
1437 mlx5_modify_nic_vport_mac_address(esw->dev, vport_num, vport->info.mac);
1438 mlx5_modify_nic_vport_node_guid(esw->dev, vport_num, vport->info.node_guid);
1439 modify_esw_vport_cvlan(esw->dev, vport_num, vport->info.vlan, vport->info.qos,
1440 (vport->info.vlan || vport->info.qos));
1442 /* Only legacy mode needs ACLs */
1443 if (esw->mode == SRIOV_LEGACY) {
1444 esw_vport_ingress_config(esw, vport);
1445 esw_vport_egress_config(esw, vport);
1449 static void esw_enable_vport(struct mlx5_eswitch *esw, int vport_num,
1452 struct mlx5_vport *vport = &esw->vports[vport_num];
1454 mutex_lock(&esw->state_lock);
1455 WARN_ON(vport->enabled);
1457 esw_debug(esw->dev, "Enabling VPORT(%d)\n", vport_num);
1459 /* Restore old vport configuration */
1460 esw_apply_vport_conf(esw, vport);
1462 /* Attach vport to the eswitch rate limiter */
1463 if (esw_vport_enable_qos(esw, vport_num, vport->info.max_rate,
1464 vport->qos.bw_share))
1465 esw_warn(esw->dev, "Failed to attach vport %d to eswitch rate limiter", vport_num);
1467 /* Sync with current vport context */
1468 vport->enabled_events = enable_events;
1469 vport->enabled = true;
1471 /* only PF is trusted by default */
1473 vport->info.trusted = true;
1475 esw_vport_change_handle_locked(vport);
1477 esw->enabled_vports++;
1478 esw_debug(esw->dev, "Enabled VPORT(%d)\n", vport_num);
1479 mutex_unlock(&esw->state_lock);
1482 static void esw_disable_vport(struct mlx5_eswitch *esw, int vport_num)
1484 struct mlx5_vport *vport = &esw->vports[vport_num];
1486 if (!vport->enabled)
1489 esw_debug(esw->dev, "Disabling vport(%d)\n", vport_num);
1490 /* Mark this vport as disabled to discard new events */
1491 vport->enabled = false;
1493 synchronize_irq(pci_irq_vector(esw->dev->pdev, MLX5_EQ_VEC_ASYNC));
1494 /* Wait for current already scheduled events to complete */
1495 flush_workqueue(esw->work_queue);
1496 /* Disable events from this vport */
1497 arm_vport_context_events_cmd(esw->dev, vport->vport, 0);
1498 mutex_lock(&esw->state_lock);
1499 /* We don't assume VFs will cleanup after themselves.
1500 * Calling vport change handler while vport is disabled will cleanup
1501 * the vport resources.
1503 esw_vport_change_handle_locked(vport);
1504 vport->enabled_events = 0;
1505 esw_vport_disable_qos(esw, vport_num);
1506 if (vport_num && esw->mode == SRIOV_LEGACY) {
1507 mlx5_modify_vport_admin_state(esw->dev,
1508 MLX5_QUERY_VPORT_STATE_IN_OP_MOD_ESW_VPORT,
1510 MLX5_ESW_VPORT_ADMIN_STATE_DOWN);
1511 esw_vport_disable_egress_acl(esw, vport);
1512 esw_vport_disable_ingress_acl(esw, vport);
1514 esw->enabled_vports--;
1515 mutex_unlock(&esw->state_lock);
1518 /* Public E-Switch API */
1519 #define ESW_ALLOWED(esw) ((esw) && MLX5_ESWITCH_MANAGER((esw)->dev))
1522 int mlx5_eswitch_enable_sriov(struct mlx5_eswitch *esw, int nvfs, int mode)
1525 int i, enabled_events;
1527 if (!ESW_ALLOWED(esw) ||
1528 !MLX5_CAP_ESW_FLOWTABLE_FDB(esw->dev, ft_support)) {
1529 esw_warn(esw->dev, "E-Switch FDB is not supported, aborting ...\n");
1533 if (!MLX5_CAP_ESW_INGRESS_ACL(esw->dev, ft_support))
1534 esw_warn(esw->dev, "E-Switch ingress ACL is not supported by FW\n");
1536 if (!MLX5_CAP_ESW_EGRESS_ACL(esw->dev, ft_support))
1537 esw_warn(esw->dev, "E-Switch engress ACL is not supported by FW\n");
1539 esw_info(esw->dev, "E-Switch enable SRIOV: nvfs(%d) mode (%d)\n", nvfs, mode);
1542 if (mode == SRIOV_LEGACY)
1543 err = esw_create_legacy_fdb_table(esw, nvfs + 1);
1545 err = esw_offloads_init(esw, nvfs + 1);
1549 err = esw_create_tsar(esw);
1551 esw_warn(esw->dev, "Failed to create eswitch TSAR");
1553 /* Don't enable vport events when in SRIOV_OFFLOADS mode, since:
1554 * 1. L2 table (MPFS) is programmed by PF/VF representors netdevs set_rx_mode
1555 * 2. FDB/Eswitch is programmed by user space tools
1557 enabled_events = (mode == SRIOV_LEGACY) ? SRIOV_VPORT_EVENTS : 0;
1558 for (i = 0; i <= nvfs; i++)
1559 esw_enable_vport(esw, i, enabled_events);
1561 esw_info(esw->dev, "SRIOV enabled: active vports(%d)\n",
1562 esw->enabled_vports);
1566 esw->mode = SRIOV_NONE;
1570 void mlx5_eswitch_disable_sriov(struct mlx5_eswitch *esw)
1572 struct esw_mc_addr *mc_promisc;
1576 if (!ESW_ALLOWED(esw) || esw->mode == SRIOV_NONE)
1579 esw_info(esw->dev, "disable SRIOV: active vports(%d) mode(%d)\n",
1580 esw->enabled_vports, esw->mode);
1582 mc_promisc = &esw->mc_promisc;
1583 nvports = esw->enabled_vports;
1585 for (i = 0; i < esw->total_vports; i++)
1586 esw_disable_vport(esw, i);
1588 if (mc_promisc && mc_promisc->uplink_rule)
1589 mlx5_del_flow_rules(mc_promisc->uplink_rule);
1591 esw_destroy_tsar(esw);
1593 if (esw->mode == SRIOV_LEGACY)
1594 esw_destroy_legacy_fdb_table(esw);
1595 else if (esw->mode == SRIOV_OFFLOADS)
1596 esw_offloads_cleanup(esw, nvports);
1598 esw->mode = SRIOV_NONE;
1601 int mlx5_eswitch_init(struct mlx5_core_dev *dev)
1603 int total_vports = MLX5_TOTAL_VPORTS(dev);
1604 struct mlx5_eswitch *esw;
1608 if (!MLX5_VPORT_MANAGER(dev))
1612 "Total vports %d, per vport: max uc(%d) max mc(%d)\n",
1614 MLX5_MAX_UC_PER_VPORT(dev),
1615 MLX5_MAX_MC_PER_VPORT(dev));
1617 esw = kzalloc(sizeof(*esw), GFP_KERNEL);
1623 esw->work_queue = create_singlethread_workqueue("mlx5_esw_wq");
1624 if (!esw->work_queue) {
1629 esw->vports = kcalloc(total_vports, sizeof(struct mlx5_vport),
1636 esw->offloads.vport_reps =
1637 kzalloc(total_vports * sizeof(struct mlx5_eswitch_rep),
1639 if (!esw->offloads.vport_reps) {
1644 hash_init(esw->offloads.encap_tbl);
1645 hash_init(esw->offloads.mod_hdr_tbl);
1646 mutex_init(&esw->state_lock);
1648 for (vport_num = 0; vport_num < total_vports; vport_num++) {
1649 struct mlx5_vport *vport = &esw->vports[vport_num];
1651 vport->vport = vport_num;
1652 vport->info.link_state = MLX5_ESW_VPORT_ADMIN_STATE_AUTO;
1654 INIT_WORK(&vport->vport_change_handler,
1655 esw_vport_change_handler);
1658 esw->total_vports = total_vports;
1659 esw->enabled_vports = 0;
1660 esw->mode = SRIOV_NONE;
1661 esw->offloads.inline_mode = MLX5_INLINE_MODE_NONE;
1662 if (MLX5_CAP_ESW_FLOWTABLE_FDB(dev, encap) &&
1663 MLX5_CAP_ESW_FLOWTABLE_FDB(dev, decap))
1664 esw->offloads.encap = DEVLINK_ESWITCH_ENCAP_MODE_BASIC;
1666 esw->offloads.encap = DEVLINK_ESWITCH_ENCAP_MODE_NONE;
1668 dev->priv.eswitch = esw;
1671 if (esw->work_queue)
1672 destroy_workqueue(esw->work_queue);
1674 kfree(esw->offloads.vport_reps);
1679 void mlx5_eswitch_cleanup(struct mlx5_eswitch *esw)
1681 if (!esw || !MLX5_VPORT_MANAGER(esw->dev))
1684 esw_info(esw->dev, "cleanup\n");
1686 esw->dev->priv.eswitch = NULL;
1687 destroy_workqueue(esw->work_queue);
1688 kfree(esw->offloads.vport_reps);
1693 void mlx5_eswitch_vport_event(struct mlx5_eswitch *esw, struct mlx5_eqe *eqe)
1695 struct mlx5_eqe_vport_change *vc_eqe = &eqe->data.vport_change;
1696 u16 vport_num = be16_to_cpu(vc_eqe->vport_num);
1697 struct mlx5_vport *vport;
1700 pr_warn("MLX5 E-Switch: vport %d got an event while eswitch is not initialized\n",
1705 vport = &esw->vports[vport_num];
1707 queue_work(esw->work_queue, &vport->vport_change_handler);
1710 /* Vport Administration */
1711 #define LEGAL_VPORT(esw, vport) (vport >= 0 && vport < esw->total_vports)
1713 int mlx5_eswitch_set_vport_mac(struct mlx5_eswitch *esw,
1714 int vport, u8 mac[ETH_ALEN])
1716 struct mlx5_vport *evport;
1720 if (!esw || !MLX5_CAP_GEN(esw->dev, vport_group_manager))
1722 if (!LEGAL_VPORT(esw, vport) || is_multicast_ether_addr(mac))
1725 mutex_lock(&esw->state_lock);
1726 evport = &esw->vports[vport];
1728 if (evport->info.spoofchk && !is_valid_ether_addr(mac))
1729 mlx5_core_warn(esw->dev,
1730 "Set invalid MAC while spoofchk is on, vport(%d)\n",
1733 err = mlx5_modify_nic_vport_mac_address(esw->dev, vport, mac);
1735 mlx5_core_warn(esw->dev,
1736 "Failed to mlx5_modify_nic_vport_mac vport(%d) err=(%d)\n",
1741 node_guid_gen_from_mac(&node_guid, mac);
1742 err = mlx5_modify_nic_vport_node_guid(esw->dev, vport, node_guid);
1744 mlx5_core_warn(esw->dev,
1745 "Failed to set vport %d node guid, err = %d. RDMA_CM will not function properly for this VF.\n",
1748 ether_addr_copy(evport->info.mac, mac);
1749 evport->info.node_guid = node_guid;
1750 if (evport->enabled && esw->mode == SRIOV_LEGACY)
1751 err = esw_vport_ingress_config(esw, evport);
1754 mutex_unlock(&esw->state_lock);
1758 int mlx5_eswitch_set_vport_state(struct mlx5_eswitch *esw,
1759 int vport, int link_state)
1761 struct mlx5_vport *evport;
1764 if (!ESW_ALLOWED(esw))
1766 if (!LEGAL_VPORT(esw, vport))
1769 mutex_lock(&esw->state_lock);
1770 evport = &esw->vports[vport];
1772 err = mlx5_modify_vport_admin_state(esw->dev,
1773 MLX5_QUERY_VPORT_STATE_IN_OP_MOD_ESW_VPORT,
1776 mlx5_core_warn(esw->dev,
1777 "Failed to set vport %d link state, err = %d",
1782 evport->info.link_state = link_state;
1785 mutex_unlock(&esw->state_lock);
1789 int mlx5_eswitch_get_vport_config(struct mlx5_eswitch *esw,
1790 int vport, struct ifla_vf_info *ivi)
1792 struct mlx5_vport *evport;
1794 if (!esw || !MLX5_CAP_GEN(esw->dev, vport_group_manager))
1796 if (!LEGAL_VPORT(esw, vport))
1799 evport = &esw->vports[vport];
1801 memset(ivi, 0, sizeof(*ivi));
1802 ivi->vf = vport - 1;
1804 mutex_lock(&esw->state_lock);
1805 ether_addr_copy(ivi->mac, evport->info.mac);
1806 ivi->linkstate = evport->info.link_state;
1807 ivi->vlan = evport->info.vlan;
1808 ivi->qos = evport->info.qos;
1809 ivi->spoofchk = evport->info.spoofchk;
1810 ivi->trusted = evport->info.trusted;
1811 ivi->min_tx_rate = evport->info.min_rate;
1812 ivi->max_tx_rate = evport->info.max_rate;
1813 mutex_unlock(&esw->state_lock);
1818 int __mlx5_eswitch_set_vport_vlan(struct mlx5_eswitch *esw,
1819 int vport, u16 vlan, u8 qos, u8 set_flags)
1821 struct mlx5_vport *evport;
1824 if (!ESW_ALLOWED(esw))
1826 if (!LEGAL_VPORT(esw, vport) || (vlan > 4095) || (qos > 7))
1829 mutex_lock(&esw->state_lock);
1830 evport = &esw->vports[vport];
1832 err = modify_esw_vport_cvlan(esw->dev, vport, vlan, qos, set_flags);
1836 evport->info.vlan = vlan;
1837 evport->info.qos = qos;
1838 if (evport->enabled && esw->mode == SRIOV_LEGACY) {
1839 err = esw_vport_ingress_config(esw, evport);
1842 err = esw_vport_egress_config(esw, evport);
1846 mutex_unlock(&esw->state_lock);
1850 int mlx5_eswitch_set_vport_vlan(struct mlx5_eswitch *esw,
1851 int vport, u16 vlan, u8 qos)
1856 set_flags = SET_VLAN_STRIP | SET_VLAN_INSERT;
1858 return __mlx5_eswitch_set_vport_vlan(esw, vport, vlan, qos, set_flags);
1861 int mlx5_eswitch_set_vport_spoofchk(struct mlx5_eswitch *esw,
1862 int vport, bool spoofchk)
1864 struct mlx5_vport *evport;
1868 if (!ESW_ALLOWED(esw))
1870 if (!LEGAL_VPORT(esw, vport))
1873 mutex_lock(&esw->state_lock);
1874 evport = &esw->vports[vport];
1875 pschk = evport->info.spoofchk;
1876 evport->info.spoofchk = spoofchk;
1877 if (pschk && !is_valid_ether_addr(evport->info.mac))
1878 mlx5_core_warn(esw->dev,
1879 "Spoofchk in set while MAC is invalid, vport(%d)\n",
1881 if (evport->enabled && esw->mode == SRIOV_LEGACY)
1882 err = esw_vport_ingress_config(esw, evport);
1884 evport->info.spoofchk = pschk;
1885 mutex_unlock(&esw->state_lock);
1890 int mlx5_eswitch_set_vport_trust(struct mlx5_eswitch *esw,
1891 int vport, bool setting)
1893 struct mlx5_vport *evport;
1895 if (!ESW_ALLOWED(esw))
1897 if (!LEGAL_VPORT(esw, vport))
1900 mutex_lock(&esw->state_lock);
1901 evport = &esw->vports[vport];
1902 evport->info.trusted = setting;
1903 if (evport->enabled)
1904 esw_vport_change_handle_locked(evport);
1905 mutex_unlock(&esw->state_lock);
1910 static u32 calculate_vports_min_rate_divider(struct mlx5_eswitch *esw)
1912 u32 fw_max_bw_share = MLX5_CAP_QOS(esw->dev, max_tsar_bw_share);
1913 struct mlx5_vport *evport;
1914 u32 max_guarantee = 0;
1917 for (i = 0; i < esw->total_vports; i++) {
1918 evport = &esw->vports[i];
1919 if (!evport->enabled || evport->info.min_rate < max_guarantee)
1921 max_guarantee = evport->info.min_rate;
1925 return max_t(u32, max_guarantee / fw_max_bw_share, 1);
1929 static int normalize_vports_min_rate(struct mlx5_eswitch *esw)
1931 u32 fw_max_bw_share = MLX5_CAP_QOS(esw->dev, max_tsar_bw_share);
1932 u32 divider = calculate_vports_min_rate_divider(esw);
1933 struct mlx5_vport *evport;
1940 for (i = 0; i < esw->total_vports; i++) {
1941 evport = &esw->vports[i];
1942 if (!evport->enabled)
1944 vport_min_rate = evport->info.min_rate;
1945 vport_max_rate = evport->info.max_rate;
1949 bw_share = MLX5_RATE_TO_BW_SHARE(vport_min_rate,
1953 if (bw_share == evport->qos.bw_share)
1956 err = esw_vport_qos_config(esw, i, vport_max_rate,
1959 evport->qos.bw_share = bw_share;
1967 int mlx5_eswitch_set_vport_rate(struct mlx5_eswitch *esw, int vport,
1968 u32 max_rate, u32 min_rate)
1970 struct mlx5_vport *evport;
1971 u32 fw_max_bw_share;
1972 u32 previous_min_rate;
1973 bool min_rate_supported;
1974 bool max_rate_supported;
1977 if (!ESW_ALLOWED(esw))
1979 if (!LEGAL_VPORT(esw, vport))
1982 fw_max_bw_share = MLX5_CAP_QOS(esw->dev, max_tsar_bw_share);
1983 min_rate_supported = MLX5_CAP_QOS(esw->dev, esw_bw_share) &&
1984 fw_max_bw_share >= MLX5_MIN_BW_SHARE;
1985 max_rate_supported = MLX5_CAP_QOS(esw->dev, esw_rate_limit);
1987 if ((min_rate && !min_rate_supported) || (max_rate && !max_rate_supported))
1990 mutex_lock(&esw->state_lock);
1991 evport = &esw->vports[vport];
1993 if (min_rate == evport->info.min_rate)
1996 previous_min_rate = evport->info.min_rate;
1997 evport->info.min_rate = min_rate;
1998 err = normalize_vports_min_rate(esw);
2000 evport->info.min_rate = previous_min_rate;
2005 if (max_rate == evport->info.max_rate)
2008 err = esw_vport_qos_config(esw, vport, max_rate, evport->qos.bw_share);
2010 evport->info.max_rate = max_rate;
2013 mutex_unlock(&esw->state_lock);
2017 int mlx5_eswitch_get_vport_stats(struct mlx5_eswitch *esw,
2019 struct ifla_vf_stats *vf_stats)
2021 int outlen = MLX5_ST_SZ_BYTES(query_vport_counter_out);
2022 u32 in[MLX5_ST_SZ_DW(query_vport_counter_in)] = {0};
2026 if (!ESW_ALLOWED(esw))
2028 if (!LEGAL_VPORT(esw, vport))
2031 out = kvzalloc(outlen, GFP_KERNEL);
2035 MLX5_SET(query_vport_counter_in, in, opcode,
2036 MLX5_CMD_OP_QUERY_VPORT_COUNTER);
2037 MLX5_SET(query_vport_counter_in, in, op_mod, 0);
2038 MLX5_SET(query_vport_counter_in, in, vport_number, vport);
2040 MLX5_SET(query_vport_counter_in, in, other_vport, 1);
2042 memset(out, 0, outlen);
2043 err = mlx5_cmd_exec(esw->dev, in, sizeof(in), out, outlen);
2047 #define MLX5_GET_CTR(p, x) \
2048 MLX5_GET64(query_vport_counter_out, p, x)
2050 memset(vf_stats, 0, sizeof(*vf_stats));
2051 vf_stats->rx_packets =
2052 MLX5_GET_CTR(out, received_eth_unicast.packets) +
2053 MLX5_GET_CTR(out, received_ib_unicast.packets) +
2054 MLX5_GET_CTR(out, received_eth_multicast.packets) +
2055 MLX5_GET_CTR(out, received_ib_multicast.packets) +
2056 MLX5_GET_CTR(out, received_eth_broadcast.packets);
2058 vf_stats->rx_bytes =
2059 MLX5_GET_CTR(out, received_eth_unicast.octets) +
2060 MLX5_GET_CTR(out, received_ib_unicast.octets) +
2061 MLX5_GET_CTR(out, received_eth_multicast.octets) +
2062 MLX5_GET_CTR(out, received_ib_multicast.octets) +
2063 MLX5_GET_CTR(out, received_eth_broadcast.octets);
2065 vf_stats->tx_packets =
2066 MLX5_GET_CTR(out, transmitted_eth_unicast.packets) +
2067 MLX5_GET_CTR(out, transmitted_ib_unicast.packets) +
2068 MLX5_GET_CTR(out, transmitted_eth_multicast.packets) +
2069 MLX5_GET_CTR(out, transmitted_ib_multicast.packets) +
2070 MLX5_GET_CTR(out, transmitted_eth_broadcast.packets);
2072 vf_stats->tx_bytes =
2073 MLX5_GET_CTR(out, transmitted_eth_unicast.octets) +
2074 MLX5_GET_CTR(out, transmitted_ib_unicast.octets) +
2075 MLX5_GET_CTR(out, transmitted_eth_multicast.octets) +
2076 MLX5_GET_CTR(out, transmitted_ib_multicast.octets) +
2077 MLX5_GET_CTR(out, transmitted_eth_broadcast.octets);
2079 vf_stats->multicast =
2080 MLX5_GET_CTR(out, received_eth_multicast.packets) +
2081 MLX5_GET_CTR(out, received_ib_multicast.packets);
2083 vf_stats->broadcast =
2084 MLX5_GET_CTR(out, received_eth_broadcast.packets);