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"
50 /* Vport UC/MC hash node */
52 struct l2addr_node node;
55 struct mlx5_flow_handle *flow_rule;
56 bool mpfs; /* UC MAC was added to MPFs */
57 /* A flag indicating that mac was added due to mc promiscuous vport */
62 UC_ADDR_CHANGE = BIT(0),
63 MC_ADDR_CHANGE = BIT(1),
64 PROMISC_CHANGE = BIT(3),
67 static void esw_destroy_legacy_fdb_table(struct mlx5_eswitch *esw);
68 static void esw_cleanup_vepa_rules(struct mlx5_eswitch *esw);
70 /* Vport context events */
71 #define SRIOV_VPORT_EVENTS (UC_ADDR_CHANGE | \
75 struct mlx5_vport *mlx5_eswitch_get_vport(struct mlx5_eswitch *esw,
78 u16 idx = mlx5_eswitch_vport_num_to_index(esw, vport_num);
80 WARN_ON(idx > esw->total_vports - 1);
81 return &esw->vports[idx];
84 static int arm_vport_context_events_cmd(struct mlx5_core_dev *dev, u16 vport,
87 int in[MLX5_ST_SZ_DW(modify_nic_vport_context_in)] = {0};
88 int out[MLX5_ST_SZ_DW(modify_nic_vport_context_out)] = {0};
91 MLX5_SET(modify_nic_vport_context_in, in,
92 opcode, MLX5_CMD_OP_MODIFY_NIC_VPORT_CONTEXT);
93 MLX5_SET(modify_nic_vport_context_in, in, field_select.change_event, 1);
94 MLX5_SET(modify_nic_vport_context_in, in, vport_number, vport);
95 MLX5_SET(modify_nic_vport_context_in, in, other_vport, 1);
96 nic_vport_ctx = MLX5_ADDR_OF(modify_nic_vport_context_in,
97 in, nic_vport_context);
99 MLX5_SET(nic_vport_context, nic_vport_ctx, arm_change_event, 1);
101 if (events_mask & UC_ADDR_CHANGE)
102 MLX5_SET(nic_vport_context, nic_vport_ctx,
103 event_on_uc_address_change, 1);
104 if (events_mask & MC_ADDR_CHANGE)
105 MLX5_SET(nic_vport_context, nic_vport_ctx,
106 event_on_mc_address_change, 1);
107 if (events_mask & PROMISC_CHANGE)
108 MLX5_SET(nic_vport_context, nic_vport_ctx,
109 event_on_promisc_change, 1);
111 return mlx5_cmd_exec(dev, in, sizeof(in), out, sizeof(out));
114 /* E-Switch vport context HW commands */
115 static int modify_esw_vport_context_cmd(struct mlx5_core_dev *dev, u16 vport,
118 u32 out[MLX5_ST_SZ_DW(modify_esw_vport_context_out)] = {0};
120 MLX5_SET(modify_esw_vport_context_in, in, opcode,
121 MLX5_CMD_OP_MODIFY_ESW_VPORT_CONTEXT);
122 MLX5_SET(modify_esw_vport_context_in, in, vport_number, vport);
123 MLX5_SET(modify_esw_vport_context_in, in, other_vport, 1);
124 return mlx5_cmd_exec(dev, in, inlen, out, sizeof(out));
127 static int modify_esw_vport_cvlan(struct mlx5_core_dev *dev, u16 vport,
128 u16 vlan, u8 qos, u8 set_flags)
130 u32 in[MLX5_ST_SZ_DW(modify_esw_vport_context_in)] = {0};
132 if (!MLX5_CAP_ESW(dev, vport_cvlan_strip) ||
133 !MLX5_CAP_ESW(dev, vport_cvlan_insert_if_not_exist))
136 esw_debug(dev, "Set Vport[%d] VLAN %d qos %d set=%x\n",
137 vport, vlan, qos, set_flags);
139 if (set_flags & SET_VLAN_STRIP)
140 MLX5_SET(modify_esw_vport_context_in, in,
141 esw_vport_context.vport_cvlan_strip, 1);
143 if (set_flags & SET_VLAN_INSERT) {
144 /* insert only if no vlan in packet */
145 MLX5_SET(modify_esw_vport_context_in, in,
146 esw_vport_context.vport_cvlan_insert, 1);
148 MLX5_SET(modify_esw_vport_context_in, in,
149 esw_vport_context.cvlan_pcp, qos);
150 MLX5_SET(modify_esw_vport_context_in, in,
151 esw_vport_context.cvlan_id, vlan);
154 MLX5_SET(modify_esw_vport_context_in, in,
155 field_select.vport_cvlan_strip, 1);
156 MLX5_SET(modify_esw_vport_context_in, in,
157 field_select.vport_cvlan_insert, 1);
159 return modify_esw_vport_context_cmd(dev, vport, in, sizeof(in));
163 static struct mlx5_flow_handle *
164 __esw_fdb_set_vport_rule(struct mlx5_eswitch *esw, u16 vport, bool rx_rule,
165 u8 mac_c[ETH_ALEN], u8 mac_v[ETH_ALEN])
167 int match_header = (is_zero_ether_addr(mac_c) ? 0 :
168 MLX5_MATCH_OUTER_HEADERS);
169 struct mlx5_flow_handle *flow_rule = NULL;
170 struct mlx5_flow_act flow_act = {0};
171 struct mlx5_flow_destination dest = {};
172 struct mlx5_flow_spec *spec;
173 void *mv_misc = NULL;
174 void *mc_misc = NULL;
179 match_header |= MLX5_MATCH_MISC_PARAMETERS;
181 spec = kvzalloc(sizeof(*spec), GFP_KERNEL);
185 dmac_v = MLX5_ADDR_OF(fte_match_param, spec->match_value,
186 outer_headers.dmac_47_16);
187 dmac_c = MLX5_ADDR_OF(fte_match_param, spec->match_criteria,
188 outer_headers.dmac_47_16);
190 if (match_header & MLX5_MATCH_OUTER_HEADERS) {
191 ether_addr_copy(dmac_v, mac_v);
192 ether_addr_copy(dmac_c, mac_c);
195 if (match_header & MLX5_MATCH_MISC_PARAMETERS) {
196 mv_misc = MLX5_ADDR_OF(fte_match_param, spec->match_value,
198 mc_misc = MLX5_ADDR_OF(fte_match_param, spec->match_criteria,
200 MLX5_SET(fte_match_set_misc, mv_misc, source_port, MLX5_VPORT_UPLINK);
201 MLX5_SET_TO_ONES(fte_match_set_misc, mc_misc, source_port);
204 dest.type = MLX5_FLOW_DESTINATION_TYPE_VPORT;
205 dest.vport.num = vport;
208 "\tFDB add rule dmac_v(%pM) dmac_c(%pM) -> vport(%d)\n",
209 dmac_v, dmac_c, vport);
210 spec->match_criteria_enable = match_header;
211 flow_act.action = MLX5_FLOW_CONTEXT_ACTION_FWD_DEST;
213 mlx5_add_flow_rules(esw->fdb_table.legacy.fdb, spec,
214 &flow_act, &dest, 1);
215 if (IS_ERR(flow_rule)) {
217 "FDB: Failed to add flow rule: dmac_v(%pM) dmac_c(%pM) -> vport(%d), err(%ld)\n",
218 dmac_v, dmac_c, vport, PTR_ERR(flow_rule));
226 static struct mlx5_flow_handle *
227 esw_fdb_set_vport_rule(struct mlx5_eswitch *esw, u8 mac[ETH_ALEN], u16 vport)
231 eth_broadcast_addr(mac_c);
232 return __esw_fdb_set_vport_rule(esw, vport, false, mac_c, mac);
235 static struct mlx5_flow_handle *
236 esw_fdb_set_vport_allmulti_rule(struct mlx5_eswitch *esw, u16 vport)
241 eth_zero_addr(mac_c);
242 eth_zero_addr(mac_v);
245 return __esw_fdb_set_vport_rule(esw, vport, false, mac_c, mac_v);
248 static struct mlx5_flow_handle *
249 esw_fdb_set_vport_promisc_rule(struct mlx5_eswitch *esw, u16 vport)
254 eth_zero_addr(mac_c);
255 eth_zero_addr(mac_v);
256 return __esw_fdb_set_vport_rule(esw, vport, true, mac_c, mac_v);
260 LEGACY_VEPA_PRIO = 0,
264 static int esw_create_legacy_vepa_table(struct mlx5_eswitch *esw)
266 struct mlx5_core_dev *dev = esw->dev;
267 struct mlx5_flow_namespace *root_ns;
268 struct mlx5_flow_table *fdb;
271 root_ns = mlx5_get_fdb_sub_ns(dev, 0);
273 esw_warn(dev, "Failed to get FDB flow namespace\n");
277 /* num FTE 2, num FG 2 */
278 fdb = mlx5_create_auto_grouped_flow_table(root_ns, LEGACY_VEPA_PRIO,
282 esw_warn(dev, "Failed to create VEPA FDB err %d\n", err);
285 esw->fdb_table.legacy.vepa_fdb = fdb;
290 static int esw_create_legacy_fdb_table(struct mlx5_eswitch *esw)
292 int inlen = MLX5_ST_SZ_BYTES(create_flow_group_in);
293 struct mlx5_flow_table_attr ft_attr = {};
294 struct mlx5_core_dev *dev = esw->dev;
295 struct mlx5_flow_namespace *root_ns;
296 struct mlx5_flow_table *fdb;
297 struct mlx5_flow_group *g;
298 void *match_criteria;
304 esw_debug(dev, "Create FDB log_max_size(%d)\n",
305 MLX5_CAP_ESW_FLOWTABLE_FDB(dev, log_max_ft_size));
307 root_ns = mlx5_get_fdb_sub_ns(dev, 0);
309 esw_warn(dev, "Failed to get FDB flow namespace\n");
313 flow_group_in = kvzalloc(inlen, GFP_KERNEL);
317 table_size = BIT(MLX5_CAP_ESW_FLOWTABLE_FDB(dev, log_max_ft_size));
318 ft_attr.max_fte = table_size;
319 ft_attr.prio = LEGACY_FDB_PRIO;
320 fdb = mlx5_create_flow_table(root_ns, &ft_attr);
323 esw_warn(dev, "Failed to create FDB Table err %d\n", err);
326 esw->fdb_table.legacy.fdb = fdb;
328 /* Addresses group : Full match unicast/multicast addresses */
329 MLX5_SET(create_flow_group_in, flow_group_in, match_criteria_enable,
330 MLX5_MATCH_OUTER_HEADERS);
331 match_criteria = MLX5_ADDR_OF(create_flow_group_in, flow_group_in, match_criteria);
332 dmac = MLX5_ADDR_OF(fte_match_param, match_criteria, outer_headers.dmac_47_16);
333 MLX5_SET(create_flow_group_in, flow_group_in, start_flow_index, 0);
334 /* Preserve 2 entries for allmulti and promisc rules*/
335 MLX5_SET(create_flow_group_in, flow_group_in, end_flow_index, table_size - 3);
336 eth_broadcast_addr(dmac);
337 g = mlx5_create_flow_group(fdb, flow_group_in);
340 esw_warn(dev, "Failed to create flow group err(%d)\n", err);
343 esw->fdb_table.legacy.addr_grp = g;
345 /* Allmulti group : One rule that forwards any mcast traffic */
346 MLX5_SET(create_flow_group_in, flow_group_in, match_criteria_enable,
347 MLX5_MATCH_OUTER_HEADERS);
348 MLX5_SET(create_flow_group_in, flow_group_in, start_flow_index, table_size - 2);
349 MLX5_SET(create_flow_group_in, flow_group_in, end_flow_index, table_size - 2);
352 g = mlx5_create_flow_group(fdb, flow_group_in);
355 esw_warn(dev, "Failed to create allmulti flow group err(%d)\n", err);
358 esw->fdb_table.legacy.allmulti_grp = g;
360 /* Promiscuous group :
361 * One rule that forward all unmatched traffic from previous groups
364 MLX5_SET(create_flow_group_in, flow_group_in, match_criteria_enable,
365 MLX5_MATCH_MISC_PARAMETERS);
366 MLX5_SET_TO_ONES(fte_match_param, match_criteria, misc_parameters.source_port);
367 MLX5_SET(create_flow_group_in, flow_group_in, start_flow_index, table_size - 1);
368 MLX5_SET(create_flow_group_in, flow_group_in, end_flow_index, table_size - 1);
369 g = mlx5_create_flow_group(fdb, flow_group_in);
372 esw_warn(dev, "Failed to create promisc flow group err(%d)\n", err);
375 esw->fdb_table.legacy.promisc_grp = g;
379 esw_destroy_legacy_fdb_table(esw);
381 kvfree(flow_group_in);
385 static void esw_destroy_legacy_vepa_table(struct mlx5_eswitch *esw)
387 esw_debug(esw->dev, "Destroy VEPA Table\n");
388 if (!esw->fdb_table.legacy.vepa_fdb)
391 mlx5_destroy_flow_table(esw->fdb_table.legacy.vepa_fdb);
392 esw->fdb_table.legacy.vepa_fdb = NULL;
395 static void esw_destroy_legacy_fdb_table(struct mlx5_eswitch *esw)
397 esw_debug(esw->dev, "Destroy FDB Table\n");
398 if (!esw->fdb_table.legacy.fdb)
401 if (esw->fdb_table.legacy.promisc_grp)
402 mlx5_destroy_flow_group(esw->fdb_table.legacy.promisc_grp);
403 if (esw->fdb_table.legacy.allmulti_grp)
404 mlx5_destroy_flow_group(esw->fdb_table.legacy.allmulti_grp);
405 if (esw->fdb_table.legacy.addr_grp)
406 mlx5_destroy_flow_group(esw->fdb_table.legacy.addr_grp);
407 mlx5_destroy_flow_table(esw->fdb_table.legacy.fdb);
409 esw->fdb_table.legacy.fdb = NULL;
410 esw->fdb_table.legacy.addr_grp = NULL;
411 esw->fdb_table.legacy.allmulti_grp = NULL;
412 esw->fdb_table.legacy.promisc_grp = NULL;
415 static int esw_create_legacy_table(struct mlx5_eswitch *esw)
419 memset(&esw->fdb_table.legacy, 0, sizeof(struct legacy_fdb));
421 err = esw_create_legacy_vepa_table(esw);
425 err = esw_create_legacy_fdb_table(esw);
427 esw_destroy_legacy_vepa_table(esw);
432 static void esw_destroy_legacy_table(struct mlx5_eswitch *esw)
434 esw_cleanup_vepa_rules(esw);
435 esw_destroy_legacy_fdb_table(esw);
436 esw_destroy_legacy_vepa_table(esw);
439 /* E-Switch vport UC/MC lists management */
440 typedef int (*vport_addr_action)(struct mlx5_eswitch *esw,
441 struct vport_addr *vaddr);
443 static int esw_add_uc_addr(struct mlx5_eswitch *esw, struct vport_addr *vaddr)
445 u8 *mac = vaddr->node.addr;
446 u16 vport = vaddr->vport;
449 /* Skip mlx5_mpfs_add_mac for eswitch_managers,
450 * it is already done by its netdev in mlx5e_execute_l2_action
452 if (esw->manager_vport == vport)
455 err = mlx5_mpfs_add_mac(esw->dev, mac);
458 "Failed to add L2 table mac(%pM) for vport(0x%x), err(%d)\n",
465 /* SRIOV is enabled: Forward UC MAC to vport */
466 if (esw->fdb_table.legacy.fdb && esw->mode == SRIOV_LEGACY)
467 vaddr->flow_rule = esw_fdb_set_vport_rule(esw, mac, vport);
469 esw_debug(esw->dev, "\tADDED UC MAC: vport[%d] %pM fr(%p)\n",
470 vport, mac, vaddr->flow_rule);
475 static int esw_del_uc_addr(struct mlx5_eswitch *esw, struct vport_addr *vaddr)
477 u8 *mac = vaddr->node.addr;
478 u16 vport = vaddr->vport;
481 /* Skip mlx5_mpfs_del_mac for eswitch managerss,
482 * it is already done by its netdev in mlx5e_execute_l2_action
484 if (!vaddr->mpfs || esw->manager_vport == vport)
487 err = mlx5_mpfs_del_mac(esw->dev, mac);
490 "Failed to del L2 table mac(%pM) for vport(%d), err(%d)\n",
495 if (vaddr->flow_rule)
496 mlx5_del_flow_rules(vaddr->flow_rule);
497 vaddr->flow_rule = NULL;
502 static void update_allmulti_vports(struct mlx5_eswitch *esw,
503 struct vport_addr *vaddr,
504 struct esw_mc_addr *esw_mc)
506 u8 *mac = vaddr->node.addr;
507 struct mlx5_vport *vport;
510 mlx5_esw_for_all_vports(esw, i, vport) {
511 struct hlist_head *vport_hash = vport->mc_list;
512 struct vport_addr *iter_vaddr =
513 l2addr_hash_find(vport_hash,
516 vport_num = vport->vport;
517 if (IS_ERR_OR_NULL(vport->allmulti_rule) ||
518 vaddr->vport == vport_num)
520 switch (vaddr->action) {
521 case MLX5_ACTION_ADD:
524 iter_vaddr = l2addr_hash_add(vport_hash, mac,
529 "ALL-MULTI: Failed to add MAC(%pM) to vport[%d] DB\n",
533 iter_vaddr->vport = vport_num;
534 iter_vaddr->flow_rule =
535 esw_fdb_set_vport_rule(esw,
538 iter_vaddr->mc_promisc = true;
540 case MLX5_ACTION_DEL:
543 mlx5_del_flow_rules(iter_vaddr->flow_rule);
544 l2addr_hash_del(iter_vaddr);
550 static int esw_add_mc_addr(struct mlx5_eswitch *esw, struct vport_addr *vaddr)
552 struct hlist_head *hash = esw->mc_table;
553 struct esw_mc_addr *esw_mc;
554 u8 *mac = vaddr->node.addr;
555 u16 vport = vaddr->vport;
557 if (!esw->fdb_table.legacy.fdb)
560 esw_mc = l2addr_hash_find(hash, mac, struct esw_mc_addr);
564 esw_mc = l2addr_hash_add(hash, mac, struct esw_mc_addr, GFP_KERNEL);
568 esw_mc->uplink_rule = /* Forward MC MAC to Uplink */
569 esw_fdb_set_vport_rule(esw, mac, MLX5_VPORT_UPLINK);
571 /* Add this multicast mac to all the mc promiscuous vports */
572 update_allmulti_vports(esw, vaddr, esw_mc);
575 /* If the multicast mac is added as a result of mc promiscuous vport,
576 * don't increment the multicast ref count
578 if (!vaddr->mc_promisc)
581 /* Forward MC MAC to vport */
582 vaddr->flow_rule = esw_fdb_set_vport_rule(esw, mac, vport);
584 "\tADDED MC MAC: vport[%d] %pM fr(%p) refcnt(%d) uplinkfr(%p)\n",
585 vport, mac, vaddr->flow_rule,
586 esw_mc->refcnt, esw_mc->uplink_rule);
590 static int esw_del_mc_addr(struct mlx5_eswitch *esw, struct vport_addr *vaddr)
592 struct hlist_head *hash = esw->mc_table;
593 struct esw_mc_addr *esw_mc;
594 u8 *mac = vaddr->node.addr;
595 u16 vport = vaddr->vport;
597 if (!esw->fdb_table.legacy.fdb)
600 esw_mc = l2addr_hash_find(hash, mac, struct esw_mc_addr);
603 "Failed to find eswitch MC addr for MAC(%pM) vport(%d)",
608 "\tDELETE MC MAC: vport[%d] %pM fr(%p) refcnt(%d) uplinkfr(%p)\n",
609 vport, mac, vaddr->flow_rule, esw_mc->refcnt,
610 esw_mc->uplink_rule);
612 if (vaddr->flow_rule)
613 mlx5_del_flow_rules(vaddr->flow_rule);
614 vaddr->flow_rule = NULL;
616 /* If the multicast mac is added as a result of mc promiscuous vport,
617 * don't decrement the multicast ref count.
619 if (vaddr->mc_promisc || (--esw_mc->refcnt > 0))
622 /* Remove this multicast mac from all the mc promiscuous vports */
623 update_allmulti_vports(esw, vaddr, esw_mc);
625 if (esw_mc->uplink_rule)
626 mlx5_del_flow_rules(esw_mc->uplink_rule);
628 l2addr_hash_del(esw_mc);
632 /* Apply vport UC/MC list to HW l2 table and FDB table */
633 static void esw_apply_vport_addr_list(struct mlx5_eswitch *esw,
634 struct mlx5_vport *vport, int list_type)
636 bool is_uc = list_type == MLX5_NVPRT_LIST_TYPE_UC;
637 vport_addr_action vport_addr_add;
638 vport_addr_action vport_addr_del;
639 struct vport_addr *addr;
640 struct l2addr_node *node;
641 struct hlist_head *hash;
642 struct hlist_node *tmp;
645 vport_addr_add = is_uc ? esw_add_uc_addr :
647 vport_addr_del = is_uc ? esw_del_uc_addr :
650 hash = is_uc ? vport->uc_list : vport->mc_list;
651 for_each_l2hash_node(node, tmp, hash, hi) {
652 addr = container_of(node, struct vport_addr, node);
653 switch (addr->action) {
654 case MLX5_ACTION_ADD:
655 vport_addr_add(esw, addr);
656 addr->action = MLX5_ACTION_NONE;
658 case MLX5_ACTION_DEL:
659 vport_addr_del(esw, addr);
660 l2addr_hash_del(addr);
666 /* Sync vport UC/MC list from vport context */
667 static void esw_update_vport_addr_list(struct mlx5_eswitch *esw,
668 struct mlx5_vport *vport, int list_type)
670 bool is_uc = list_type == MLX5_NVPRT_LIST_TYPE_UC;
671 u8 (*mac_list)[ETH_ALEN];
672 struct l2addr_node *node;
673 struct vport_addr *addr;
674 struct hlist_head *hash;
675 struct hlist_node *tmp;
681 size = is_uc ? MLX5_MAX_UC_PER_VPORT(esw->dev) :
682 MLX5_MAX_MC_PER_VPORT(esw->dev);
684 mac_list = kcalloc(size, ETH_ALEN, GFP_KERNEL);
688 hash = is_uc ? vport->uc_list : vport->mc_list;
690 for_each_l2hash_node(node, tmp, hash, hi) {
691 addr = container_of(node, struct vport_addr, node);
692 addr->action = MLX5_ACTION_DEL;
698 err = mlx5_query_nic_vport_mac_list(esw->dev, vport->vport, list_type,
702 esw_debug(esw->dev, "vport[%d] context update %s list size (%d)\n",
703 vport->vport, is_uc ? "UC" : "MC", size);
705 for (i = 0; i < size; i++) {
706 if (is_uc && !is_valid_ether_addr(mac_list[i]))
709 if (!is_uc && !is_multicast_ether_addr(mac_list[i]))
712 addr = l2addr_hash_find(hash, mac_list[i], struct vport_addr);
714 addr->action = MLX5_ACTION_NONE;
715 /* If this mac was previously added because of allmulti
716 * promiscuous rx mode, its now converted to be original
719 if (addr->mc_promisc) {
720 struct esw_mc_addr *esw_mc =
721 l2addr_hash_find(esw->mc_table,
726 "Failed to MAC(%pM) in mcast DB\n",
731 addr->mc_promisc = false;
736 addr = l2addr_hash_add(hash, mac_list[i], struct vport_addr,
740 "Failed to add MAC(%pM) to vport[%d] DB\n",
741 mac_list[i], vport->vport);
744 addr->vport = vport->vport;
745 addr->action = MLX5_ACTION_ADD;
751 /* Sync vport UC/MC list from vport context
752 * Must be called after esw_update_vport_addr_list
754 static void esw_update_vport_mc_promisc(struct mlx5_eswitch *esw,
755 struct mlx5_vport *vport)
757 struct l2addr_node *node;
758 struct vport_addr *addr;
759 struct hlist_head *hash;
760 struct hlist_node *tmp;
763 hash = vport->mc_list;
765 for_each_l2hash_node(node, tmp, esw->mc_table, hi) {
766 u8 *mac = node->addr;
768 addr = l2addr_hash_find(hash, mac, struct vport_addr);
770 if (addr->action == MLX5_ACTION_DEL)
771 addr->action = MLX5_ACTION_NONE;
774 addr = l2addr_hash_add(hash, mac, struct vport_addr,
778 "Failed to add allmulti MAC(%pM) to vport[%d] DB\n",
782 addr->vport = vport->vport;
783 addr->action = MLX5_ACTION_ADD;
784 addr->mc_promisc = true;
788 /* Apply vport rx mode to HW FDB table */
789 static void esw_apply_vport_rx_mode(struct mlx5_eswitch *esw,
790 struct mlx5_vport *vport,
791 bool promisc, bool mc_promisc)
793 struct esw_mc_addr *allmulti_addr = &esw->mc_promisc;
795 if (IS_ERR_OR_NULL(vport->allmulti_rule) != mc_promisc)
799 vport->allmulti_rule =
800 esw_fdb_set_vport_allmulti_rule(esw, vport->vport);
801 if (!allmulti_addr->uplink_rule)
802 allmulti_addr->uplink_rule =
803 esw_fdb_set_vport_allmulti_rule(esw,
805 allmulti_addr->refcnt++;
806 } else if (vport->allmulti_rule) {
807 mlx5_del_flow_rules(vport->allmulti_rule);
808 vport->allmulti_rule = NULL;
810 if (--allmulti_addr->refcnt > 0)
813 if (allmulti_addr->uplink_rule)
814 mlx5_del_flow_rules(allmulti_addr->uplink_rule);
815 allmulti_addr->uplink_rule = NULL;
819 if (IS_ERR_OR_NULL(vport->promisc_rule) != promisc)
823 vport->promisc_rule =
824 esw_fdb_set_vport_promisc_rule(esw, vport->vport);
825 } else if (vport->promisc_rule) {
826 mlx5_del_flow_rules(vport->promisc_rule);
827 vport->promisc_rule = NULL;
831 /* Sync vport rx mode from vport context */
832 static void esw_update_vport_rx_mode(struct mlx5_eswitch *esw,
833 struct mlx5_vport *vport)
840 err = mlx5_query_nic_vport_promisc(esw->dev,
847 esw_debug(esw->dev, "vport[%d] context update rx mode promisc_all=%d, all_multi=%d\n",
848 vport->vport, promisc_all, promisc_mc);
850 if (!vport->info.trusted || !vport->enabled) {
856 esw_apply_vport_rx_mode(esw, vport, promisc_all,
857 (promisc_all || promisc_mc));
860 static void esw_vport_change_handle_locked(struct mlx5_vport *vport)
862 struct mlx5_core_dev *dev = vport->dev;
863 struct mlx5_eswitch *esw = dev->priv.eswitch;
866 mlx5_query_nic_vport_mac_address(dev, vport->vport, mac);
867 esw_debug(dev, "vport[%d] Context Changed: perm mac: %pM\n",
870 if (vport->enabled_events & UC_ADDR_CHANGE) {
871 esw_update_vport_addr_list(esw, vport, MLX5_NVPRT_LIST_TYPE_UC);
872 esw_apply_vport_addr_list(esw, vport, MLX5_NVPRT_LIST_TYPE_UC);
875 if (vport->enabled_events & MC_ADDR_CHANGE)
876 esw_update_vport_addr_list(esw, vport, MLX5_NVPRT_LIST_TYPE_MC);
878 if (vport->enabled_events & PROMISC_CHANGE) {
879 esw_update_vport_rx_mode(esw, vport);
880 if (!IS_ERR_OR_NULL(vport->allmulti_rule))
881 esw_update_vport_mc_promisc(esw, vport);
884 if (vport->enabled_events & (PROMISC_CHANGE | MC_ADDR_CHANGE))
885 esw_apply_vport_addr_list(esw, vport, MLX5_NVPRT_LIST_TYPE_MC);
887 esw_debug(esw->dev, "vport[%d] Context Changed: Done\n", vport->vport);
889 arm_vport_context_events_cmd(dev, vport->vport,
890 vport->enabled_events);
893 static void esw_vport_change_handler(struct work_struct *work)
895 struct mlx5_vport *vport =
896 container_of(work, struct mlx5_vport, vport_change_handler);
897 struct mlx5_eswitch *esw = vport->dev->priv.eswitch;
899 mutex_lock(&esw->state_lock);
900 esw_vport_change_handle_locked(vport);
901 mutex_unlock(&esw->state_lock);
904 int esw_vport_enable_egress_acl(struct mlx5_eswitch *esw,
905 struct mlx5_vport *vport)
907 int inlen = MLX5_ST_SZ_BYTES(create_flow_group_in);
908 struct mlx5_flow_group *vlan_grp = NULL;
909 struct mlx5_flow_group *drop_grp = NULL;
910 struct mlx5_core_dev *dev = esw->dev;
911 struct mlx5_flow_namespace *root_ns;
912 struct mlx5_flow_table *acl;
913 void *match_criteria;
915 /* The egress acl table contains 2 rules:
916 * 1)Allow traffic with vlan_tag=vst_vlan_id
917 * 2)Drop all other traffic.
922 if (!MLX5_CAP_ESW_EGRESS_ACL(dev, ft_support))
925 if (!IS_ERR_OR_NULL(vport->egress.acl))
928 esw_debug(dev, "Create vport[%d] egress ACL log_max_size(%d)\n",
929 vport->vport, MLX5_CAP_ESW_EGRESS_ACL(dev, log_max_ft_size));
931 root_ns = mlx5_get_flow_vport_acl_namespace(dev, MLX5_FLOW_NAMESPACE_ESW_EGRESS,
934 esw_warn(dev, "Failed to get E-Switch egress flow namespace for vport (%d)\n", vport->vport);
938 flow_group_in = kvzalloc(inlen, GFP_KERNEL);
942 acl = mlx5_create_vport_flow_table(root_ns, 0, table_size, 0, vport->vport);
945 esw_warn(dev, "Failed to create E-Switch vport[%d] egress flow Table, err(%d)\n",
950 MLX5_SET(create_flow_group_in, flow_group_in, match_criteria_enable, MLX5_MATCH_OUTER_HEADERS);
951 match_criteria = MLX5_ADDR_OF(create_flow_group_in, flow_group_in, match_criteria);
952 MLX5_SET_TO_ONES(fte_match_param, match_criteria, outer_headers.cvlan_tag);
953 MLX5_SET_TO_ONES(fte_match_param, match_criteria, outer_headers.first_vid);
954 MLX5_SET(create_flow_group_in, flow_group_in, start_flow_index, 0);
955 MLX5_SET(create_flow_group_in, flow_group_in, end_flow_index, 0);
957 vlan_grp = mlx5_create_flow_group(acl, flow_group_in);
958 if (IS_ERR(vlan_grp)) {
959 err = PTR_ERR(vlan_grp);
960 esw_warn(dev, "Failed to create E-Switch vport[%d] egress allowed vlans flow group, err(%d)\n",
965 memset(flow_group_in, 0, inlen);
966 MLX5_SET(create_flow_group_in, flow_group_in, start_flow_index, 1);
967 MLX5_SET(create_flow_group_in, flow_group_in, end_flow_index, 1);
968 drop_grp = mlx5_create_flow_group(acl, flow_group_in);
969 if (IS_ERR(drop_grp)) {
970 err = PTR_ERR(drop_grp);
971 esw_warn(dev, "Failed to create E-Switch vport[%d] egress drop flow group, err(%d)\n",
976 vport->egress.acl = acl;
977 vport->egress.drop_grp = drop_grp;
978 vport->egress.allowed_vlans_grp = vlan_grp;
980 kvfree(flow_group_in);
981 if (err && !IS_ERR_OR_NULL(vlan_grp))
982 mlx5_destroy_flow_group(vlan_grp);
983 if (err && !IS_ERR_OR_NULL(acl))
984 mlx5_destroy_flow_table(acl);
988 void esw_vport_cleanup_egress_rules(struct mlx5_eswitch *esw,
989 struct mlx5_vport *vport)
991 if (!IS_ERR_OR_NULL(vport->egress.allowed_vlan))
992 mlx5_del_flow_rules(vport->egress.allowed_vlan);
994 if (!IS_ERR_OR_NULL(vport->egress.drop_rule))
995 mlx5_del_flow_rules(vport->egress.drop_rule);
997 vport->egress.allowed_vlan = NULL;
998 vport->egress.drop_rule = NULL;
1001 void esw_vport_disable_egress_acl(struct mlx5_eswitch *esw,
1002 struct mlx5_vport *vport)
1004 if (IS_ERR_OR_NULL(vport->egress.acl))
1007 esw_debug(esw->dev, "Destroy vport[%d] E-Switch egress ACL\n", vport->vport);
1009 esw_vport_cleanup_egress_rules(esw, vport);
1010 mlx5_destroy_flow_group(vport->egress.allowed_vlans_grp);
1011 mlx5_destroy_flow_group(vport->egress.drop_grp);
1012 mlx5_destroy_flow_table(vport->egress.acl);
1013 vport->egress.allowed_vlans_grp = NULL;
1014 vport->egress.drop_grp = NULL;
1015 vport->egress.acl = NULL;
1018 int esw_vport_enable_ingress_acl(struct mlx5_eswitch *esw,
1019 struct mlx5_vport *vport)
1021 int inlen = MLX5_ST_SZ_BYTES(create_flow_group_in);
1022 struct mlx5_core_dev *dev = esw->dev;
1023 struct mlx5_flow_namespace *root_ns;
1024 struct mlx5_flow_table *acl;
1025 struct mlx5_flow_group *g;
1026 void *match_criteria;
1028 /* The ingress acl table contains 4 groups
1029 * (2 active rules at the same time -
1030 * 1 allow rule from one of the first 3 groups.
1031 * 1 drop rule from the last group):
1032 * 1)Allow untagged traffic with smac=original mac.
1033 * 2)Allow untagged traffic.
1034 * 3)Allow traffic with smac=original mac.
1035 * 4)Drop all other traffic.
1040 if (!MLX5_CAP_ESW_INGRESS_ACL(dev, ft_support))
1043 if (!IS_ERR_OR_NULL(vport->ingress.acl))
1046 esw_debug(dev, "Create vport[%d] ingress ACL log_max_size(%d)\n",
1047 vport->vport, MLX5_CAP_ESW_INGRESS_ACL(dev, log_max_ft_size));
1049 root_ns = mlx5_get_flow_vport_acl_namespace(dev, MLX5_FLOW_NAMESPACE_ESW_INGRESS,
1052 esw_warn(dev, "Failed to get E-Switch ingress flow namespace for vport (%d)\n", vport->vport);
1056 flow_group_in = kvzalloc(inlen, GFP_KERNEL);
1060 acl = mlx5_create_vport_flow_table(root_ns, 0, table_size, 0, vport->vport);
1063 esw_warn(dev, "Failed to create E-Switch vport[%d] ingress flow Table, err(%d)\n",
1067 vport->ingress.acl = acl;
1069 match_criteria = MLX5_ADDR_OF(create_flow_group_in, flow_group_in, match_criteria);
1071 MLX5_SET(create_flow_group_in, flow_group_in, match_criteria_enable, MLX5_MATCH_OUTER_HEADERS);
1072 MLX5_SET_TO_ONES(fte_match_param, match_criteria, outer_headers.cvlan_tag);
1073 MLX5_SET_TO_ONES(fte_match_param, match_criteria, outer_headers.smac_47_16);
1074 MLX5_SET_TO_ONES(fte_match_param, match_criteria, outer_headers.smac_15_0);
1075 MLX5_SET(create_flow_group_in, flow_group_in, start_flow_index, 0);
1076 MLX5_SET(create_flow_group_in, flow_group_in, end_flow_index, 0);
1078 g = mlx5_create_flow_group(acl, flow_group_in);
1081 esw_warn(dev, "Failed to create E-Switch vport[%d] ingress untagged spoofchk flow group, err(%d)\n",
1085 vport->ingress.allow_untagged_spoofchk_grp = g;
1087 memset(flow_group_in, 0, inlen);
1088 MLX5_SET(create_flow_group_in, flow_group_in, match_criteria_enable, MLX5_MATCH_OUTER_HEADERS);
1089 MLX5_SET_TO_ONES(fte_match_param, match_criteria, outer_headers.cvlan_tag);
1090 MLX5_SET(create_flow_group_in, flow_group_in, start_flow_index, 1);
1091 MLX5_SET(create_flow_group_in, flow_group_in, end_flow_index, 1);
1093 g = mlx5_create_flow_group(acl, flow_group_in);
1096 esw_warn(dev, "Failed to create E-Switch vport[%d] ingress untagged flow group, err(%d)\n",
1100 vport->ingress.allow_untagged_only_grp = g;
1102 memset(flow_group_in, 0, inlen);
1103 MLX5_SET(create_flow_group_in, flow_group_in, match_criteria_enable, MLX5_MATCH_OUTER_HEADERS);
1104 MLX5_SET_TO_ONES(fte_match_param, match_criteria, outer_headers.smac_47_16);
1105 MLX5_SET_TO_ONES(fte_match_param, match_criteria, outer_headers.smac_15_0);
1106 MLX5_SET(create_flow_group_in, flow_group_in, start_flow_index, 2);
1107 MLX5_SET(create_flow_group_in, flow_group_in, end_flow_index, 2);
1109 g = mlx5_create_flow_group(acl, flow_group_in);
1112 esw_warn(dev, "Failed to create E-Switch vport[%d] ingress spoofchk flow group, err(%d)\n",
1116 vport->ingress.allow_spoofchk_only_grp = g;
1118 memset(flow_group_in, 0, inlen);
1119 MLX5_SET(create_flow_group_in, flow_group_in, start_flow_index, 3);
1120 MLX5_SET(create_flow_group_in, flow_group_in, end_flow_index, 3);
1122 g = mlx5_create_flow_group(acl, flow_group_in);
1125 esw_warn(dev, "Failed to create E-Switch vport[%d] ingress drop flow group, err(%d)\n",
1129 vport->ingress.drop_grp = g;
1133 if (!IS_ERR_OR_NULL(vport->ingress.allow_spoofchk_only_grp))
1134 mlx5_destroy_flow_group(
1135 vport->ingress.allow_spoofchk_only_grp);
1136 if (!IS_ERR_OR_NULL(vport->ingress.allow_untagged_only_grp))
1137 mlx5_destroy_flow_group(
1138 vport->ingress.allow_untagged_only_grp);
1139 if (!IS_ERR_OR_NULL(vport->ingress.allow_untagged_spoofchk_grp))
1140 mlx5_destroy_flow_group(
1141 vport->ingress.allow_untagged_spoofchk_grp);
1142 if (!IS_ERR_OR_NULL(vport->ingress.acl))
1143 mlx5_destroy_flow_table(vport->ingress.acl);
1146 kvfree(flow_group_in);
1150 void esw_vport_cleanup_ingress_rules(struct mlx5_eswitch *esw,
1151 struct mlx5_vport *vport)
1153 if (!IS_ERR_OR_NULL(vport->ingress.drop_rule))
1154 mlx5_del_flow_rules(vport->ingress.drop_rule);
1156 if (!IS_ERR_OR_NULL(vport->ingress.allow_rule))
1157 mlx5_del_flow_rules(vport->ingress.allow_rule);
1159 vport->ingress.drop_rule = NULL;
1160 vport->ingress.allow_rule = NULL;
1163 void esw_vport_disable_ingress_acl(struct mlx5_eswitch *esw,
1164 struct mlx5_vport *vport)
1166 if (IS_ERR_OR_NULL(vport->ingress.acl))
1169 esw_debug(esw->dev, "Destroy vport[%d] E-Switch ingress ACL\n", vport->vport);
1171 esw_vport_cleanup_ingress_rules(esw, vport);
1172 mlx5_destroy_flow_group(vport->ingress.allow_spoofchk_only_grp);
1173 mlx5_destroy_flow_group(vport->ingress.allow_untagged_only_grp);
1174 mlx5_destroy_flow_group(vport->ingress.allow_untagged_spoofchk_grp);
1175 mlx5_destroy_flow_group(vport->ingress.drop_grp);
1176 mlx5_destroy_flow_table(vport->ingress.acl);
1177 vport->ingress.acl = NULL;
1178 vport->ingress.drop_grp = NULL;
1179 vport->ingress.allow_spoofchk_only_grp = NULL;
1180 vport->ingress.allow_untagged_only_grp = NULL;
1181 vport->ingress.allow_untagged_spoofchk_grp = NULL;
1184 static int esw_vport_ingress_config(struct mlx5_eswitch *esw,
1185 struct mlx5_vport *vport)
1187 struct mlx5_fc *counter = vport->ingress.drop_counter;
1188 struct mlx5_flow_destination drop_ctr_dst = {0};
1189 struct mlx5_flow_destination *dst = NULL;
1190 struct mlx5_flow_act flow_act = {0};
1191 struct mlx5_flow_spec *spec;
1196 esw_vport_cleanup_ingress_rules(esw, vport);
1198 if (!vport->info.vlan && !vport->info.qos && !vport->info.spoofchk) {
1199 esw_vport_disable_ingress_acl(esw, vport);
1203 err = esw_vport_enable_ingress_acl(esw, vport);
1205 mlx5_core_warn(esw->dev,
1206 "failed to enable ingress acl (%d) on vport[%d]\n",
1212 "vport[%d] configure ingress rules, vlan(%d) qos(%d)\n",
1213 vport->vport, vport->info.vlan, vport->info.qos);
1215 spec = kvzalloc(sizeof(*spec), GFP_KERNEL);
1221 if (vport->info.vlan || vport->info.qos)
1222 MLX5_SET_TO_ONES(fte_match_param, spec->match_criteria, outer_headers.cvlan_tag);
1224 if (vport->info.spoofchk) {
1225 MLX5_SET_TO_ONES(fte_match_param, spec->match_criteria, outer_headers.smac_47_16);
1226 MLX5_SET_TO_ONES(fte_match_param, spec->match_criteria, outer_headers.smac_15_0);
1227 smac_v = MLX5_ADDR_OF(fte_match_param,
1229 outer_headers.smac_47_16);
1230 ether_addr_copy(smac_v, vport->info.mac);
1233 spec->match_criteria_enable = MLX5_MATCH_OUTER_HEADERS;
1234 flow_act.action = MLX5_FLOW_CONTEXT_ACTION_ALLOW;
1235 vport->ingress.allow_rule =
1236 mlx5_add_flow_rules(vport->ingress.acl, spec,
1237 &flow_act, NULL, 0);
1238 if (IS_ERR(vport->ingress.allow_rule)) {
1239 err = PTR_ERR(vport->ingress.allow_rule);
1241 "vport[%d] configure ingress allow rule, err(%d)\n",
1243 vport->ingress.allow_rule = NULL;
1247 memset(spec, 0, sizeof(*spec));
1248 flow_act.action = MLX5_FLOW_CONTEXT_ACTION_DROP;
1250 /* Attach drop flow counter */
1252 flow_act.action |= MLX5_FLOW_CONTEXT_ACTION_COUNT;
1253 drop_ctr_dst.type = MLX5_FLOW_DESTINATION_TYPE_COUNTER;
1254 drop_ctr_dst.counter_id = mlx5_fc_id(counter);
1255 dst = &drop_ctr_dst;
1258 vport->ingress.drop_rule =
1259 mlx5_add_flow_rules(vport->ingress.acl, spec,
1260 &flow_act, dst, dest_num);
1261 if (IS_ERR(vport->ingress.drop_rule)) {
1262 err = PTR_ERR(vport->ingress.drop_rule);
1264 "vport[%d] configure ingress drop rule, err(%d)\n",
1266 vport->ingress.drop_rule = NULL;
1272 esw_vport_cleanup_ingress_rules(esw, vport);
1277 static int esw_vport_egress_config(struct mlx5_eswitch *esw,
1278 struct mlx5_vport *vport)
1280 struct mlx5_fc *counter = vport->egress.drop_counter;
1281 struct mlx5_flow_destination drop_ctr_dst = {0};
1282 struct mlx5_flow_destination *dst = NULL;
1283 struct mlx5_flow_act flow_act = {0};
1284 struct mlx5_flow_spec *spec;
1288 esw_vport_cleanup_egress_rules(esw, vport);
1290 if (!vport->info.vlan && !vport->info.qos) {
1291 esw_vport_disable_egress_acl(esw, vport);
1295 err = esw_vport_enable_egress_acl(esw, vport);
1297 mlx5_core_warn(esw->dev,
1298 "failed to enable egress acl (%d) on vport[%d]\n",
1304 "vport[%d] configure egress rules, vlan(%d) qos(%d)\n",
1305 vport->vport, vport->info.vlan, vport->info.qos);
1307 spec = kvzalloc(sizeof(*spec), GFP_KERNEL);
1313 /* Allowed vlan rule */
1314 MLX5_SET_TO_ONES(fte_match_param, spec->match_criteria, outer_headers.cvlan_tag);
1315 MLX5_SET_TO_ONES(fte_match_param, spec->match_value, outer_headers.cvlan_tag);
1316 MLX5_SET_TO_ONES(fte_match_param, spec->match_criteria, outer_headers.first_vid);
1317 MLX5_SET(fte_match_param, spec->match_value, outer_headers.first_vid, vport->info.vlan);
1319 spec->match_criteria_enable = MLX5_MATCH_OUTER_HEADERS;
1320 flow_act.action = MLX5_FLOW_CONTEXT_ACTION_ALLOW;
1321 vport->egress.allowed_vlan =
1322 mlx5_add_flow_rules(vport->egress.acl, spec,
1323 &flow_act, NULL, 0);
1324 if (IS_ERR(vport->egress.allowed_vlan)) {
1325 err = PTR_ERR(vport->egress.allowed_vlan);
1327 "vport[%d] configure egress allowed vlan rule failed, err(%d)\n",
1329 vport->egress.allowed_vlan = NULL;
1333 /* Drop others rule (star rule) */
1334 memset(spec, 0, sizeof(*spec));
1335 flow_act.action = MLX5_FLOW_CONTEXT_ACTION_DROP;
1337 /* Attach egress drop flow counter */
1339 flow_act.action |= MLX5_FLOW_CONTEXT_ACTION_COUNT;
1340 drop_ctr_dst.type = MLX5_FLOW_DESTINATION_TYPE_COUNTER;
1341 drop_ctr_dst.counter_id = mlx5_fc_id(counter);
1342 dst = &drop_ctr_dst;
1345 vport->egress.drop_rule =
1346 mlx5_add_flow_rules(vport->egress.acl, spec,
1347 &flow_act, dst, dest_num);
1348 if (IS_ERR(vport->egress.drop_rule)) {
1349 err = PTR_ERR(vport->egress.drop_rule);
1351 "vport[%d] configure egress drop rule failed, err(%d)\n",
1353 vport->egress.drop_rule = NULL;
1360 /* Vport QoS management */
1361 static int esw_create_tsar(struct mlx5_eswitch *esw)
1363 u32 tsar_ctx[MLX5_ST_SZ_DW(scheduling_context)] = {0};
1364 struct mlx5_core_dev *dev = esw->dev;
1367 if (!MLX5_CAP_GEN(dev, qos) || !MLX5_CAP_QOS(dev, esw_scheduling))
1370 if (esw->qos.enabled)
1373 err = mlx5_create_scheduling_element_cmd(dev,
1374 SCHEDULING_HIERARCHY_E_SWITCH,
1376 &esw->qos.root_tsar_id);
1378 esw_warn(esw->dev, "E-Switch create TSAR failed (%d)\n", err);
1382 esw->qos.enabled = true;
1386 static void esw_destroy_tsar(struct mlx5_eswitch *esw)
1390 if (!esw->qos.enabled)
1393 err = mlx5_destroy_scheduling_element_cmd(esw->dev,
1394 SCHEDULING_HIERARCHY_E_SWITCH,
1395 esw->qos.root_tsar_id);
1397 esw_warn(esw->dev, "E-Switch destroy TSAR failed (%d)\n", err);
1399 esw->qos.enabled = false;
1402 static int esw_vport_enable_qos(struct mlx5_eswitch *esw,
1403 struct mlx5_vport *vport,
1404 u32 initial_max_rate, u32 initial_bw_share)
1406 u32 sched_ctx[MLX5_ST_SZ_DW(scheduling_context)] = {0};
1407 struct mlx5_core_dev *dev = esw->dev;
1411 if (!esw->qos.enabled || !MLX5_CAP_GEN(dev, qos) ||
1412 !MLX5_CAP_QOS(dev, esw_scheduling))
1415 if (vport->qos.enabled)
1418 MLX5_SET(scheduling_context, sched_ctx, element_type,
1419 SCHEDULING_CONTEXT_ELEMENT_TYPE_VPORT);
1420 vport_elem = MLX5_ADDR_OF(scheduling_context, sched_ctx,
1421 element_attributes);
1422 MLX5_SET(vport_element, vport_elem, vport_number, vport->vport);
1423 MLX5_SET(scheduling_context, sched_ctx, parent_element_id,
1424 esw->qos.root_tsar_id);
1425 MLX5_SET(scheduling_context, sched_ctx, max_average_bw,
1427 MLX5_SET(scheduling_context, sched_ctx, bw_share, initial_bw_share);
1429 err = mlx5_create_scheduling_element_cmd(dev,
1430 SCHEDULING_HIERARCHY_E_SWITCH,
1432 &vport->qos.esw_tsar_ix);
1434 esw_warn(esw->dev, "E-Switch create TSAR vport element failed (vport=%d,err=%d)\n",
1439 vport->qos.enabled = true;
1443 static void esw_vport_disable_qos(struct mlx5_eswitch *esw,
1444 struct mlx5_vport *vport)
1448 if (!vport->qos.enabled)
1451 err = mlx5_destroy_scheduling_element_cmd(esw->dev,
1452 SCHEDULING_HIERARCHY_E_SWITCH,
1453 vport->qos.esw_tsar_ix);
1455 esw_warn(esw->dev, "E-Switch destroy TSAR vport element failed (vport=%d,err=%d)\n",
1458 vport->qos.enabled = false;
1461 static int esw_vport_qos_config(struct mlx5_eswitch *esw,
1462 struct mlx5_vport *vport,
1463 u32 max_rate, u32 bw_share)
1465 u32 sched_ctx[MLX5_ST_SZ_DW(scheduling_context)] = {0};
1466 struct mlx5_core_dev *dev = esw->dev;
1471 if (!MLX5_CAP_GEN(dev, qos) || !MLX5_CAP_QOS(dev, esw_scheduling))
1474 if (!vport->qos.enabled)
1477 MLX5_SET(scheduling_context, sched_ctx, element_type,
1478 SCHEDULING_CONTEXT_ELEMENT_TYPE_VPORT);
1479 vport_elem = MLX5_ADDR_OF(scheduling_context, sched_ctx,
1480 element_attributes);
1481 MLX5_SET(vport_element, vport_elem, vport_number, vport->vport);
1482 MLX5_SET(scheduling_context, sched_ctx, parent_element_id,
1483 esw->qos.root_tsar_id);
1484 MLX5_SET(scheduling_context, sched_ctx, max_average_bw,
1486 MLX5_SET(scheduling_context, sched_ctx, bw_share, bw_share);
1487 bitmask |= MODIFY_SCHEDULING_ELEMENT_IN_MODIFY_BITMASK_MAX_AVERAGE_BW;
1488 bitmask |= MODIFY_SCHEDULING_ELEMENT_IN_MODIFY_BITMASK_BW_SHARE;
1490 err = mlx5_modify_scheduling_element_cmd(dev,
1491 SCHEDULING_HIERARCHY_E_SWITCH,
1493 vport->qos.esw_tsar_ix,
1496 esw_warn(esw->dev, "E-Switch modify TSAR vport element failed (vport=%d,err=%d)\n",
1504 static void node_guid_gen_from_mac(u64 *node_guid, u8 mac[ETH_ALEN])
1506 ((u8 *)node_guid)[7] = mac[0];
1507 ((u8 *)node_guid)[6] = mac[1];
1508 ((u8 *)node_guid)[5] = mac[2];
1509 ((u8 *)node_guid)[4] = 0xff;
1510 ((u8 *)node_guid)[3] = 0xfe;
1511 ((u8 *)node_guid)[2] = mac[3];
1512 ((u8 *)node_guid)[1] = mac[4];
1513 ((u8 *)node_guid)[0] = mac[5];
1516 static void esw_apply_vport_conf(struct mlx5_eswitch *esw,
1517 struct mlx5_vport *vport)
1519 int vport_num = vport->vport;
1521 if (esw->manager_vport == vport_num)
1524 mlx5_modify_vport_admin_state(esw->dev,
1525 MLX5_VPORT_STATE_OP_MOD_ESW_VPORT,
1527 vport->info.link_state);
1529 /* Host PF has its own mac/guid. */
1531 mlx5_modify_nic_vport_mac_address(esw->dev, vport_num,
1533 mlx5_modify_nic_vport_node_guid(esw->dev, vport_num,
1534 vport->info.node_guid);
1537 modify_esw_vport_cvlan(esw->dev, vport_num, vport->info.vlan, vport->info.qos,
1538 (vport->info.vlan || vport->info.qos));
1540 /* Only legacy mode needs ACLs */
1541 if (esw->mode == SRIOV_LEGACY) {
1542 esw_vport_ingress_config(esw, vport);
1543 esw_vport_egress_config(esw, vport);
1547 static void esw_vport_create_drop_counters(struct mlx5_vport *vport)
1549 struct mlx5_core_dev *dev = vport->dev;
1551 if (MLX5_CAP_ESW_INGRESS_ACL(dev, flow_counter)) {
1552 vport->ingress.drop_counter = mlx5_fc_create(dev, false);
1553 if (IS_ERR(vport->ingress.drop_counter)) {
1555 "vport[%d] configure ingress drop rule counter failed\n",
1557 vport->ingress.drop_counter = NULL;
1561 if (MLX5_CAP_ESW_EGRESS_ACL(dev, flow_counter)) {
1562 vport->egress.drop_counter = mlx5_fc_create(dev, false);
1563 if (IS_ERR(vport->egress.drop_counter)) {
1565 "vport[%d] configure egress drop rule counter failed\n",
1567 vport->egress.drop_counter = NULL;
1572 static void esw_vport_destroy_drop_counters(struct mlx5_vport *vport)
1574 struct mlx5_core_dev *dev = vport->dev;
1576 if (vport->ingress.drop_counter)
1577 mlx5_fc_destroy(dev, vport->ingress.drop_counter);
1578 if (vport->egress.drop_counter)
1579 mlx5_fc_destroy(dev, vport->egress.drop_counter);
1582 static void esw_enable_vport(struct mlx5_eswitch *esw, struct mlx5_vport *vport,
1585 u16 vport_num = vport->vport;
1587 mutex_lock(&esw->state_lock);
1588 WARN_ON(vport->enabled);
1590 esw_debug(esw->dev, "Enabling VPORT(%d)\n", vport_num);
1592 /* Create steering drop counters for ingress and egress ACLs */
1593 if (vport_num && esw->mode == SRIOV_LEGACY)
1594 esw_vport_create_drop_counters(vport);
1596 /* Restore old vport configuration */
1597 esw_apply_vport_conf(esw, vport);
1599 /* Attach vport to the eswitch rate limiter */
1600 if (esw_vport_enable_qos(esw, vport, vport->info.max_rate,
1601 vport->qos.bw_share))
1602 esw_warn(esw->dev, "Failed to attach vport %d to eswitch rate limiter", vport_num);
1604 /* Sync with current vport context */
1605 vport->enabled_events = enable_events;
1606 vport->enabled = true;
1608 /* Esw manager is trusted by default. Host PF (vport 0) is trusted as well
1609 * in smartNIC as it's a vport group manager.
1611 if (esw->manager_vport == vport_num ||
1612 (!vport_num && mlx5_core_is_ecpf(esw->dev)))
1613 vport->info.trusted = true;
1615 esw_vport_change_handle_locked(vport);
1617 esw->enabled_vports++;
1618 esw_debug(esw->dev, "Enabled VPORT(%d)\n", vport_num);
1619 mutex_unlock(&esw->state_lock);
1622 static void esw_disable_vport(struct mlx5_eswitch *esw,
1623 struct mlx5_vport *vport)
1625 u16 vport_num = vport->vport;
1627 if (!vport->enabled)
1630 esw_debug(esw->dev, "Disabling vport(%d)\n", vport_num);
1631 /* Mark this vport as disabled to discard new events */
1632 vport->enabled = false;
1634 /* Wait for current already scheduled events to complete */
1635 flush_workqueue(esw->work_queue);
1636 /* Disable events from this vport */
1637 arm_vport_context_events_cmd(esw->dev, vport->vport, 0);
1638 mutex_lock(&esw->state_lock);
1639 /* We don't assume VFs will cleanup after themselves.
1640 * Calling vport change handler while vport is disabled will cleanup
1641 * the vport resources.
1643 esw_vport_change_handle_locked(vport);
1644 vport->enabled_events = 0;
1645 esw_vport_disable_qos(esw, vport);
1646 if (esw->manager_vport != vport_num &&
1647 esw->mode == SRIOV_LEGACY) {
1648 mlx5_modify_vport_admin_state(esw->dev,
1649 MLX5_VPORT_STATE_OP_MOD_ESW_VPORT,
1651 MLX5_VPORT_ADMIN_STATE_DOWN);
1652 esw_vport_disable_egress_acl(esw, vport);
1653 esw_vport_disable_ingress_acl(esw, vport);
1654 esw_vport_destroy_drop_counters(vport);
1656 esw->enabled_vports--;
1657 mutex_unlock(&esw->state_lock);
1660 static int eswitch_vport_event(struct notifier_block *nb,
1661 unsigned long type, void *data)
1663 struct mlx5_eswitch *esw = mlx5_nb_cof(nb, struct mlx5_eswitch, nb);
1664 struct mlx5_eqe *eqe = data;
1665 struct mlx5_vport *vport;
1668 vport_num = be16_to_cpu(eqe->data.vport_change.vport_num);
1669 vport = mlx5_eswitch_get_vport(esw, vport_num);
1671 queue_work(esw->work_queue, &vport->vport_change_handler);
1676 /* Public E-Switch API */
1677 #define ESW_ALLOWED(esw) ((esw) && MLX5_ESWITCH_MANAGER((esw)->dev))
1679 int mlx5_eswitch_enable_sriov(struct mlx5_eswitch *esw, int nvfs, int mode)
1681 int vf_nvports = 0, total_nvports = 0;
1682 struct mlx5_vport *vport;
1684 int i, enabled_events;
1686 if (!ESW_ALLOWED(esw) ||
1687 !MLX5_CAP_ESW_FLOWTABLE_FDB(esw->dev, ft_support)) {
1688 esw_warn(esw->dev, "E-Switch FDB is not supported, aborting ...\n");
1692 if (!MLX5_CAP_ESW_INGRESS_ACL(esw->dev, ft_support))
1693 esw_warn(esw->dev, "E-Switch ingress ACL is not supported by FW\n");
1695 if (!MLX5_CAP_ESW_EGRESS_ACL(esw->dev, ft_support))
1696 esw_warn(esw->dev, "E-Switch engress ACL is not supported by FW\n");
1698 esw_info(esw->dev, "E-Switch enable SRIOV: nvfs(%d) mode (%d)\n", nvfs, mode);
1700 if (mode == SRIOV_OFFLOADS) {
1701 if (mlx5_core_is_ecpf_esw_manager(esw->dev)) {
1702 err = mlx5_query_host_params_num_vfs(esw->dev, &vf_nvports);
1705 total_nvports = esw->total_vports;
1708 total_nvports = nvfs + MLX5_SPECIAL_VPORTS(esw->dev);
1714 mlx5_lag_update(esw->dev);
1716 if (mode == SRIOV_LEGACY) {
1717 err = esw_create_legacy_table(esw);
1721 mlx5_reload_interface(esw->dev, MLX5_INTERFACE_PROTOCOL_ETH);
1722 mlx5_reload_interface(esw->dev, MLX5_INTERFACE_PROTOCOL_IB);
1723 err = esw_offloads_init(esw, vf_nvports, total_nvports);
1729 err = esw_create_tsar(esw);
1731 esw_warn(esw->dev, "Failed to create eswitch TSAR");
1733 /* Don't enable vport events when in SRIOV_OFFLOADS mode, since:
1734 * 1. L2 table (MPFS) is programmed by PF/VF representors netdevs set_rx_mode
1735 * 2. FDB/Eswitch is programmed by user space tools
1737 enabled_events = (mode == SRIOV_LEGACY) ? SRIOV_VPORT_EVENTS : 0;
1739 /* Enable PF vport */
1740 vport = mlx5_eswitch_get_vport(esw, MLX5_VPORT_PF);
1741 esw_enable_vport(esw, vport, enabled_events);
1743 /* Enable ECPF vports */
1744 if (mlx5_ecpf_vport_exists(esw->dev)) {
1745 vport = mlx5_eswitch_get_vport(esw, MLX5_VPORT_ECPF);
1746 esw_enable_vport(esw, vport, enabled_events);
1749 /* Enable VF vports */
1750 mlx5_esw_for_each_vf_vport(esw, i, vport, nvfs)
1751 esw_enable_vport(esw, vport, enabled_events);
1753 if (mode == SRIOV_LEGACY) {
1754 MLX5_NB_INIT(&esw->nb, eswitch_vport_event, NIC_VPORT_CHANGE);
1755 mlx5_eq_notifier_register(esw->dev, &esw->nb);
1758 esw_info(esw->dev, "SRIOV enabled: active vports(%d)\n",
1759 esw->enabled_vports);
1763 esw->mode = SRIOV_NONE;
1765 if (mode == SRIOV_OFFLOADS) {
1766 mlx5_reload_interface(esw->dev, MLX5_INTERFACE_PROTOCOL_IB);
1767 mlx5_reload_interface(esw->dev, MLX5_INTERFACE_PROTOCOL_ETH);
1773 void mlx5_eswitch_disable_sriov(struct mlx5_eswitch *esw)
1775 struct esw_mc_addr *mc_promisc;
1776 struct mlx5_vport *vport;
1780 if (!ESW_ALLOWED(esw) || esw->mode == SRIOV_NONE)
1783 esw_info(esw->dev, "disable SRIOV: active vports(%d) mode(%d)\n",
1784 esw->enabled_vports, esw->mode);
1786 mc_promisc = &esw->mc_promisc;
1788 if (esw->mode == SRIOV_LEGACY)
1789 mlx5_eq_notifier_unregister(esw->dev, &esw->nb);
1791 mlx5_esw_for_all_vports(esw, i, vport)
1792 esw_disable_vport(esw, vport);
1794 if (mc_promisc && mc_promisc->uplink_rule)
1795 mlx5_del_flow_rules(mc_promisc->uplink_rule);
1797 esw_destroy_tsar(esw);
1799 if (esw->mode == SRIOV_LEGACY)
1800 esw_destroy_legacy_table(esw);
1801 else if (esw->mode == SRIOV_OFFLOADS)
1802 esw_offloads_cleanup(esw);
1804 old_mode = esw->mode;
1805 esw->mode = SRIOV_NONE;
1807 mlx5_lag_update(esw->dev);
1809 if (old_mode == SRIOV_OFFLOADS) {
1810 mlx5_reload_interface(esw->dev, MLX5_INTERFACE_PROTOCOL_IB);
1811 mlx5_reload_interface(esw->dev, MLX5_INTERFACE_PROTOCOL_ETH);
1815 int mlx5_eswitch_init(struct mlx5_core_dev *dev)
1817 int total_vports = MLX5_TOTAL_VPORTS(dev);
1818 struct mlx5_eswitch *esw;
1819 struct mlx5_vport *vport;
1822 if (!MLX5_VPORT_MANAGER(dev))
1826 "Total vports %d, per vport: max uc(%d) max mc(%d)\n",
1828 MLX5_MAX_UC_PER_VPORT(dev),
1829 MLX5_MAX_MC_PER_VPORT(dev));
1831 esw = kzalloc(sizeof(*esw), GFP_KERNEL);
1836 esw->manager_vport = mlx5_eswitch_manager_vport(dev);
1838 esw->work_queue = create_singlethread_workqueue("mlx5_esw_wq");
1839 if (!esw->work_queue) {
1844 esw->vports = kcalloc(total_vports, sizeof(struct mlx5_vport),
1851 esw->total_vports = total_vports;
1853 err = esw_offloads_init_reps(esw);
1857 hash_init(esw->offloads.encap_tbl);
1858 hash_init(esw->offloads.mod_hdr_tbl);
1859 mutex_init(&esw->state_lock);
1861 mlx5_esw_for_all_vports(esw, i, vport) {
1862 vport->vport = mlx5_eswitch_index_to_vport_num(esw, i);
1863 vport->info.link_state = MLX5_VPORT_ADMIN_STATE_AUTO;
1865 INIT_WORK(&vport->vport_change_handler,
1866 esw_vport_change_handler);
1869 esw->enabled_vports = 0;
1870 esw->mode = SRIOV_NONE;
1871 esw->offloads.inline_mode = MLX5_INLINE_MODE_NONE;
1872 if (MLX5_CAP_ESW_FLOWTABLE_FDB(dev, reformat) &&
1873 MLX5_CAP_ESW_FLOWTABLE_FDB(dev, decap))
1874 esw->offloads.encap = DEVLINK_ESWITCH_ENCAP_MODE_BASIC;
1876 esw->offloads.encap = DEVLINK_ESWITCH_ENCAP_MODE_NONE;
1878 dev->priv.eswitch = esw;
1881 if (esw->work_queue)
1882 destroy_workqueue(esw->work_queue);
1883 esw_offloads_cleanup_reps(esw);
1889 void mlx5_eswitch_cleanup(struct mlx5_eswitch *esw)
1891 if (!esw || !MLX5_VPORT_MANAGER(esw->dev))
1894 esw_info(esw->dev, "cleanup\n");
1896 esw->dev->priv.eswitch = NULL;
1897 destroy_workqueue(esw->work_queue);
1898 esw_offloads_cleanup_reps(esw);
1903 /* Vport Administration */
1904 #define LEGAL_VPORT(esw, vport) (vport >= 0 && vport < esw->total_vports)
1906 int mlx5_eswitch_set_vport_mac(struct mlx5_eswitch *esw,
1907 int vport, u8 mac[ETH_ALEN])
1909 struct mlx5_vport *evport;
1913 if (!esw || !MLX5_CAP_GEN(esw->dev, vport_group_manager))
1915 if (!LEGAL_VPORT(esw, vport) || is_multicast_ether_addr(mac))
1918 mutex_lock(&esw->state_lock);
1919 evport = mlx5_eswitch_get_vport(esw, vport);
1921 if (evport->info.spoofchk && !is_valid_ether_addr(mac))
1922 mlx5_core_warn(esw->dev,
1923 "Set invalid MAC while spoofchk is on, vport(%d)\n",
1926 err = mlx5_modify_nic_vport_mac_address(esw->dev, vport, mac);
1928 mlx5_core_warn(esw->dev,
1929 "Failed to mlx5_modify_nic_vport_mac vport(%d) err=(%d)\n",
1934 node_guid_gen_from_mac(&node_guid, mac);
1935 err = mlx5_modify_nic_vport_node_guid(esw->dev, vport, node_guid);
1937 mlx5_core_warn(esw->dev,
1938 "Failed to set vport %d node guid, err = %d. RDMA_CM will not function properly for this VF.\n",
1941 ether_addr_copy(evport->info.mac, mac);
1942 evport->info.node_guid = node_guid;
1943 if (evport->enabled && esw->mode == SRIOV_LEGACY)
1944 err = esw_vport_ingress_config(esw, evport);
1947 mutex_unlock(&esw->state_lock);
1951 int mlx5_eswitch_set_vport_state(struct mlx5_eswitch *esw,
1952 int vport, int link_state)
1954 struct mlx5_vport *evport;
1957 if (!ESW_ALLOWED(esw))
1959 if (!LEGAL_VPORT(esw, vport))
1962 mutex_lock(&esw->state_lock);
1963 evport = mlx5_eswitch_get_vport(esw, vport);
1965 err = mlx5_modify_vport_admin_state(esw->dev,
1966 MLX5_VPORT_STATE_OP_MOD_ESW_VPORT,
1967 vport, 1, link_state);
1969 mlx5_core_warn(esw->dev,
1970 "Failed to set vport %d link state, err = %d",
1975 evport->info.link_state = link_state;
1978 mutex_unlock(&esw->state_lock);
1982 int mlx5_eswitch_get_vport_config(struct mlx5_eswitch *esw,
1983 int vport, struct ifla_vf_info *ivi)
1985 struct mlx5_vport *evport;
1987 if (!esw || !MLX5_CAP_GEN(esw->dev, vport_group_manager))
1989 if (!LEGAL_VPORT(esw, vport))
1992 evport = mlx5_eswitch_get_vport(esw, vport);
1994 memset(ivi, 0, sizeof(*ivi));
1995 ivi->vf = vport - 1;
1997 mutex_lock(&esw->state_lock);
1998 ether_addr_copy(ivi->mac, evport->info.mac);
1999 ivi->linkstate = evport->info.link_state;
2000 ivi->vlan = evport->info.vlan;
2001 ivi->qos = evport->info.qos;
2002 ivi->spoofchk = evport->info.spoofchk;
2003 ivi->trusted = evport->info.trusted;
2004 ivi->min_tx_rate = evport->info.min_rate;
2005 ivi->max_tx_rate = evport->info.max_rate;
2006 mutex_unlock(&esw->state_lock);
2011 int __mlx5_eswitch_set_vport_vlan(struct mlx5_eswitch *esw,
2012 int vport, u16 vlan, u8 qos, u8 set_flags)
2014 struct mlx5_vport *evport;
2017 if (!ESW_ALLOWED(esw))
2019 if (!LEGAL_VPORT(esw, vport) || (vlan > 4095) || (qos > 7))
2022 mutex_lock(&esw->state_lock);
2023 evport = mlx5_eswitch_get_vport(esw, vport);
2025 err = modify_esw_vport_cvlan(esw->dev, vport, vlan, qos, set_flags);
2029 evport->info.vlan = vlan;
2030 evport->info.qos = qos;
2031 if (evport->enabled && esw->mode == SRIOV_LEGACY) {
2032 err = esw_vport_ingress_config(esw, evport);
2035 err = esw_vport_egress_config(esw, evport);
2039 mutex_unlock(&esw->state_lock);
2043 int mlx5_eswitch_set_vport_vlan(struct mlx5_eswitch *esw,
2044 int vport, u16 vlan, u8 qos)
2049 set_flags = SET_VLAN_STRIP | SET_VLAN_INSERT;
2051 return __mlx5_eswitch_set_vport_vlan(esw, vport, vlan, qos, set_flags);
2054 int mlx5_eswitch_set_vport_spoofchk(struct mlx5_eswitch *esw,
2055 int vport, bool spoofchk)
2057 struct mlx5_vport *evport;
2061 if (!ESW_ALLOWED(esw))
2063 if (!LEGAL_VPORT(esw, vport))
2066 mutex_lock(&esw->state_lock);
2067 evport = mlx5_eswitch_get_vport(esw, vport);
2068 pschk = evport->info.spoofchk;
2069 evport->info.spoofchk = spoofchk;
2070 if (pschk && !is_valid_ether_addr(evport->info.mac))
2071 mlx5_core_warn(esw->dev,
2072 "Spoofchk in set while MAC is invalid, vport(%d)\n",
2074 if (evport->enabled && esw->mode == SRIOV_LEGACY)
2075 err = esw_vport_ingress_config(esw, evport);
2077 evport->info.spoofchk = pschk;
2078 mutex_unlock(&esw->state_lock);
2083 static void esw_cleanup_vepa_rules(struct mlx5_eswitch *esw)
2085 if (esw->fdb_table.legacy.vepa_uplink_rule)
2086 mlx5_del_flow_rules(esw->fdb_table.legacy.vepa_uplink_rule);
2088 if (esw->fdb_table.legacy.vepa_star_rule)
2089 mlx5_del_flow_rules(esw->fdb_table.legacy.vepa_star_rule);
2091 esw->fdb_table.legacy.vepa_uplink_rule = NULL;
2092 esw->fdb_table.legacy.vepa_star_rule = NULL;
2095 static int _mlx5_eswitch_set_vepa_locked(struct mlx5_eswitch *esw,
2098 struct mlx5_flow_destination dest = {};
2099 struct mlx5_flow_act flow_act = {};
2100 struct mlx5_flow_handle *flow_rule;
2101 struct mlx5_flow_spec *spec;
2106 esw_cleanup_vepa_rules(esw);
2110 if (esw->fdb_table.legacy.vepa_uplink_rule)
2113 spec = kvzalloc(sizeof(*spec), GFP_KERNEL);
2117 /* Uplink rule forward uplink traffic to FDB */
2118 misc = MLX5_ADDR_OF(fte_match_param, spec->match_value, misc_parameters);
2119 MLX5_SET(fte_match_set_misc, misc, source_port, MLX5_VPORT_UPLINK);
2121 misc = MLX5_ADDR_OF(fte_match_param, spec->match_criteria, misc_parameters);
2122 MLX5_SET_TO_ONES(fte_match_set_misc, misc, source_port);
2124 spec->match_criteria_enable = MLX5_MATCH_MISC_PARAMETERS;
2125 dest.type = MLX5_FLOW_DESTINATION_TYPE_FLOW_TABLE;
2126 dest.ft = esw->fdb_table.legacy.fdb;
2127 flow_act.action = MLX5_FLOW_CONTEXT_ACTION_FWD_DEST;
2128 flow_rule = mlx5_add_flow_rules(esw->fdb_table.legacy.vepa_fdb, spec,
2129 &flow_act, &dest, 1);
2130 if (IS_ERR(flow_rule)) {
2131 err = PTR_ERR(flow_rule);
2134 esw->fdb_table.legacy.vepa_uplink_rule = flow_rule;
2137 /* Star rule to forward all traffic to uplink vport */
2138 memset(spec, 0, sizeof(*spec));
2139 memset(&dest, 0, sizeof(dest));
2140 dest.type = MLX5_FLOW_DESTINATION_TYPE_VPORT;
2141 dest.vport.num = MLX5_VPORT_UPLINK;
2142 flow_act.action = MLX5_FLOW_CONTEXT_ACTION_FWD_DEST;
2143 flow_rule = mlx5_add_flow_rules(esw->fdb_table.legacy.vepa_fdb, spec,
2144 &flow_act, &dest, 1);
2145 if (IS_ERR(flow_rule)) {
2146 err = PTR_ERR(flow_rule);
2149 esw->fdb_table.legacy.vepa_star_rule = flow_rule;
2155 esw_cleanup_vepa_rules(esw);
2159 int mlx5_eswitch_set_vepa(struct mlx5_eswitch *esw, u8 setting)
2166 if (!ESW_ALLOWED(esw))
2169 mutex_lock(&esw->state_lock);
2170 if (esw->mode != SRIOV_LEGACY) {
2175 err = _mlx5_eswitch_set_vepa_locked(esw, setting);
2178 mutex_unlock(&esw->state_lock);
2182 int mlx5_eswitch_get_vepa(struct mlx5_eswitch *esw, u8 *setting)
2189 if (!ESW_ALLOWED(esw))
2192 mutex_lock(&esw->state_lock);
2193 if (esw->mode != SRIOV_LEGACY) {
2198 *setting = esw->fdb_table.legacy.vepa_uplink_rule ? 1 : 0;
2201 mutex_unlock(&esw->state_lock);
2205 int mlx5_eswitch_set_vport_trust(struct mlx5_eswitch *esw,
2206 int vport, bool setting)
2208 struct mlx5_vport *evport;
2210 if (!ESW_ALLOWED(esw))
2212 if (!LEGAL_VPORT(esw, vport))
2215 mutex_lock(&esw->state_lock);
2216 evport = mlx5_eswitch_get_vport(esw, vport);
2217 evport->info.trusted = setting;
2218 if (evport->enabled)
2219 esw_vport_change_handle_locked(evport);
2220 mutex_unlock(&esw->state_lock);
2225 static u32 calculate_vports_min_rate_divider(struct mlx5_eswitch *esw)
2227 u32 fw_max_bw_share = MLX5_CAP_QOS(esw->dev, max_tsar_bw_share);
2228 struct mlx5_vport *evport;
2229 u32 max_guarantee = 0;
2232 mlx5_esw_for_all_vports(esw, i, evport) {
2233 if (!evport->enabled || evport->info.min_rate < max_guarantee)
2235 max_guarantee = evport->info.min_rate;
2238 return max_t(u32, max_guarantee / fw_max_bw_share, 1);
2241 static int normalize_vports_min_rate(struct mlx5_eswitch *esw, u32 divider)
2243 u32 fw_max_bw_share = MLX5_CAP_QOS(esw->dev, max_tsar_bw_share);
2244 struct mlx5_vport *evport;
2251 mlx5_esw_for_all_vports(esw, i, evport) {
2252 if (!evport->enabled)
2254 vport_min_rate = evport->info.min_rate;
2255 vport_max_rate = evport->info.max_rate;
2256 bw_share = MLX5_MIN_BW_SHARE;
2259 bw_share = MLX5_RATE_TO_BW_SHARE(vport_min_rate,
2263 if (bw_share == evport->qos.bw_share)
2266 err = esw_vport_qos_config(esw, evport, vport_max_rate,
2269 evport->qos.bw_share = bw_share;
2277 int mlx5_eswitch_set_vport_rate(struct mlx5_eswitch *esw, int vport,
2278 u32 max_rate, u32 min_rate)
2280 struct mlx5_vport *evport;
2281 u32 fw_max_bw_share;
2282 u32 previous_min_rate;
2284 bool min_rate_supported;
2285 bool max_rate_supported;
2288 if (!ESW_ALLOWED(esw))
2290 if (!LEGAL_VPORT(esw, vport))
2293 fw_max_bw_share = MLX5_CAP_QOS(esw->dev, max_tsar_bw_share);
2294 min_rate_supported = MLX5_CAP_QOS(esw->dev, esw_bw_share) &&
2295 fw_max_bw_share >= MLX5_MIN_BW_SHARE;
2296 max_rate_supported = MLX5_CAP_QOS(esw->dev, esw_rate_limit);
2298 if ((min_rate && !min_rate_supported) || (max_rate && !max_rate_supported))
2301 mutex_lock(&esw->state_lock);
2302 evport = mlx5_eswitch_get_vport(esw, vport);
2304 if (min_rate == evport->info.min_rate)
2307 previous_min_rate = evport->info.min_rate;
2308 evport->info.min_rate = min_rate;
2309 divider = calculate_vports_min_rate_divider(esw);
2310 err = normalize_vports_min_rate(esw, divider);
2312 evport->info.min_rate = previous_min_rate;
2317 if (max_rate == evport->info.max_rate)
2320 err = esw_vport_qos_config(esw, evport, max_rate, evport->qos.bw_share);
2322 evport->info.max_rate = max_rate;
2325 mutex_unlock(&esw->state_lock);
2329 static int mlx5_eswitch_query_vport_drop_stats(struct mlx5_core_dev *dev,
2330 struct mlx5_vport *vport,
2331 struct mlx5_vport_drop_stats *stats)
2333 struct mlx5_eswitch *esw = dev->priv.eswitch;
2334 u64 rx_discard_vport_down, tx_discard_vport_down;
2338 if (!vport->enabled || esw->mode != SRIOV_LEGACY)
2341 if (vport->egress.drop_counter)
2342 mlx5_fc_query(dev, vport->egress.drop_counter,
2343 &stats->rx_dropped, &bytes);
2345 if (vport->ingress.drop_counter)
2346 mlx5_fc_query(dev, vport->ingress.drop_counter,
2347 &stats->tx_dropped, &bytes);
2349 if (!MLX5_CAP_GEN(dev, receive_discard_vport_down) &&
2350 !MLX5_CAP_GEN(dev, transmit_discard_vport_down))
2353 err = mlx5_query_vport_down_stats(dev, vport->vport, 1,
2354 &rx_discard_vport_down,
2355 &tx_discard_vport_down);
2359 if (MLX5_CAP_GEN(dev, receive_discard_vport_down))
2360 stats->rx_dropped += rx_discard_vport_down;
2361 if (MLX5_CAP_GEN(dev, transmit_discard_vport_down))
2362 stats->tx_dropped += tx_discard_vport_down;
2367 int mlx5_eswitch_get_vport_stats(struct mlx5_eswitch *esw,
2369 struct ifla_vf_stats *vf_stats)
2371 struct mlx5_vport *vport;
2372 int outlen = MLX5_ST_SZ_BYTES(query_vport_counter_out);
2373 u32 in[MLX5_ST_SZ_DW(query_vport_counter_in)] = {0};
2374 struct mlx5_vport_drop_stats stats = {0};
2378 if (!ESW_ALLOWED(esw))
2380 if (!LEGAL_VPORT(esw, vport_num))
2383 vport = mlx5_eswitch_get_vport(esw, vport_num);
2384 out = kvzalloc(outlen, GFP_KERNEL);
2388 MLX5_SET(query_vport_counter_in, in, opcode,
2389 MLX5_CMD_OP_QUERY_VPORT_COUNTER);
2390 MLX5_SET(query_vport_counter_in, in, op_mod, 0);
2391 MLX5_SET(query_vport_counter_in, in, vport_number, vport->vport);
2392 MLX5_SET(query_vport_counter_in, in, other_vport, 1);
2394 memset(out, 0, outlen);
2395 err = mlx5_cmd_exec(esw->dev, in, sizeof(in), out, outlen);
2399 #define MLX5_GET_CTR(p, x) \
2400 MLX5_GET64(query_vport_counter_out, p, x)
2402 memset(vf_stats, 0, sizeof(*vf_stats));
2403 vf_stats->rx_packets =
2404 MLX5_GET_CTR(out, received_eth_unicast.packets) +
2405 MLX5_GET_CTR(out, received_ib_unicast.packets) +
2406 MLX5_GET_CTR(out, received_eth_multicast.packets) +
2407 MLX5_GET_CTR(out, received_ib_multicast.packets) +
2408 MLX5_GET_CTR(out, received_eth_broadcast.packets);
2410 vf_stats->rx_bytes =
2411 MLX5_GET_CTR(out, received_eth_unicast.octets) +
2412 MLX5_GET_CTR(out, received_ib_unicast.octets) +
2413 MLX5_GET_CTR(out, received_eth_multicast.octets) +
2414 MLX5_GET_CTR(out, received_ib_multicast.octets) +
2415 MLX5_GET_CTR(out, received_eth_broadcast.octets);
2417 vf_stats->tx_packets =
2418 MLX5_GET_CTR(out, transmitted_eth_unicast.packets) +
2419 MLX5_GET_CTR(out, transmitted_ib_unicast.packets) +
2420 MLX5_GET_CTR(out, transmitted_eth_multicast.packets) +
2421 MLX5_GET_CTR(out, transmitted_ib_multicast.packets) +
2422 MLX5_GET_CTR(out, transmitted_eth_broadcast.packets);
2424 vf_stats->tx_bytes =
2425 MLX5_GET_CTR(out, transmitted_eth_unicast.octets) +
2426 MLX5_GET_CTR(out, transmitted_ib_unicast.octets) +
2427 MLX5_GET_CTR(out, transmitted_eth_multicast.octets) +
2428 MLX5_GET_CTR(out, transmitted_ib_multicast.octets) +
2429 MLX5_GET_CTR(out, transmitted_eth_broadcast.octets);
2431 vf_stats->multicast =
2432 MLX5_GET_CTR(out, received_eth_multicast.packets) +
2433 MLX5_GET_CTR(out, received_ib_multicast.packets);
2435 vf_stats->broadcast =
2436 MLX5_GET_CTR(out, received_eth_broadcast.packets);
2438 err = mlx5_eswitch_query_vport_drop_stats(esw->dev, vport, &stats);
2441 vf_stats->rx_dropped = stats.rx_dropped;
2442 vf_stats->tx_dropped = stats.tx_dropped;
2449 u8 mlx5_eswitch_mode(struct mlx5_eswitch *esw)
2451 return ESW_ALLOWED(esw) ? esw->mode : SRIOV_NONE;
2453 EXPORT_SYMBOL_GPL(mlx5_eswitch_mode);
2455 bool mlx5_esw_lag_prereq(struct mlx5_core_dev *dev0, struct mlx5_core_dev *dev1)
2457 if ((dev0->priv.eswitch->mode == SRIOV_NONE &&
2458 dev1->priv.eswitch->mode == SRIOV_NONE) ||
2459 (dev0->priv.eswitch->mode == SRIOV_OFFLOADS &&
2460 dev1->priv.eswitch->mode == SRIOV_OFFLOADS))
2466 bool mlx5_esw_multipath_prereq(struct mlx5_core_dev *dev0,
2467 struct mlx5_core_dev *dev1)
2469 return (dev0->priv.eswitch->mode == SRIOV_OFFLOADS &&
2470 dev1->priv.eswitch->mode == SRIOV_OFFLOADS);