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 /* The vport getter/iterator are only valid after esw->total_vports
76 * and vport->vport are initialized in mlx5_eswitch_init.
78 #define mlx5_esw_for_all_vports(esw, i, vport) \
79 for ((i) = MLX5_VPORT_PF; \
80 (vport) = &(esw)->vports[i], \
81 (i) < (esw)->total_vports; (i)++)
83 #define mlx5_esw_for_each_vf_vport(esw, i, vport, nvfs) \
84 for ((i) = MLX5_VPORT_FIRST_VF; \
85 (vport) = &(esw)->vports[i], \
88 static struct mlx5_vport *mlx5_eswitch_get_vport(struct mlx5_eswitch *esw,
91 u16 idx = mlx5_eswitch_vport_num_to_index(esw, vport_num);
93 WARN_ON(idx > esw->total_vports - 1);
94 return &esw->vports[idx];
97 static int arm_vport_context_events_cmd(struct mlx5_core_dev *dev, u16 vport,
100 int in[MLX5_ST_SZ_DW(modify_nic_vport_context_in)] = {0};
101 int out[MLX5_ST_SZ_DW(modify_nic_vport_context_out)] = {0};
104 MLX5_SET(modify_nic_vport_context_in, in,
105 opcode, MLX5_CMD_OP_MODIFY_NIC_VPORT_CONTEXT);
106 MLX5_SET(modify_nic_vport_context_in, in, field_select.change_event, 1);
107 MLX5_SET(modify_nic_vport_context_in, in, vport_number, vport);
108 MLX5_SET(modify_nic_vport_context_in, in, other_vport, 1);
109 nic_vport_ctx = MLX5_ADDR_OF(modify_nic_vport_context_in,
110 in, nic_vport_context);
112 MLX5_SET(nic_vport_context, nic_vport_ctx, arm_change_event, 1);
114 if (events_mask & UC_ADDR_CHANGE)
115 MLX5_SET(nic_vport_context, nic_vport_ctx,
116 event_on_uc_address_change, 1);
117 if (events_mask & MC_ADDR_CHANGE)
118 MLX5_SET(nic_vport_context, nic_vport_ctx,
119 event_on_mc_address_change, 1);
120 if (events_mask & PROMISC_CHANGE)
121 MLX5_SET(nic_vport_context, nic_vport_ctx,
122 event_on_promisc_change, 1);
124 return mlx5_cmd_exec(dev, in, sizeof(in), out, sizeof(out));
127 /* E-Switch vport context HW commands */
128 static int modify_esw_vport_context_cmd(struct mlx5_core_dev *dev, u16 vport,
131 u32 out[MLX5_ST_SZ_DW(modify_esw_vport_context_out)] = {0};
133 MLX5_SET(modify_esw_vport_context_in, in, opcode,
134 MLX5_CMD_OP_MODIFY_ESW_VPORT_CONTEXT);
135 MLX5_SET(modify_esw_vport_context_in, in, vport_number, vport);
136 MLX5_SET(modify_esw_vport_context_in, in, other_vport, 1);
137 return mlx5_cmd_exec(dev, in, inlen, out, sizeof(out));
140 static int modify_esw_vport_cvlan(struct mlx5_core_dev *dev, u16 vport,
141 u16 vlan, u8 qos, u8 set_flags)
143 u32 in[MLX5_ST_SZ_DW(modify_esw_vport_context_in)] = {0};
145 if (!MLX5_CAP_ESW(dev, vport_cvlan_strip) ||
146 !MLX5_CAP_ESW(dev, vport_cvlan_insert_if_not_exist))
149 esw_debug(dev, "Set Vport[%d] VLAN %d qos %d set=%x\n",
150 vport, vlan, qos, set_flags);
152 if (set_flags & SET_VLAN_STRIP)
153 MLX5_SET(modify_esw_vport_context_in, in,
154 esw_vport_context.vport_cvlan_strip, 1);
156 if (set_flags & SET_VLAN_INSERT) {
157 /* insert only if no vlan in packet */
158 MLX5_SET(modify_esw_vport_context_in, in,
159 esw_vport_context.vport_cvlan_insert, 1);
161 MLX5_SET(modify_esw_vport_context_in, in,
162 esw_vport_context.cvlan_pcp, qos);
163 MLX5_SET(modify_esw_vport_context_in, in,
164 esw_vport_context.cvlan_id, vlan);
167 MLX5_SET(modify_esw_vport_context_in, in,
168 field_select.vport_cvlan_strip, 1);
169 MLX5_SET(modify_esw_vport_context_in, in,
170 field_select.vport_cvlan_insert, 1);
172 return modify_esw_vport_context_cmd(dev, vport, in, sizeof(in));
176 static struct mlx5_flow_handle *
177 __esw_fdb_set_vport_rule(struct mlx5_eswitch *esw, u16 vport, bool rx_rule,
178 u8 mac_c[ETH_ALEN], u8 mac_v[ETH_ALEN])
180 int match_header = (is_zero_ether_addr(mac_c) ? 0 :
181 MLX5_MATCH_OUTER_HEADERS);
182 struct mlx5_flow_handle *flow_rule = NULL;
183 struct mlx5_flow_act flow_act = {0};
184 struct mlx5_flow_destination dest = {};
185 struct mlx5_flow_spec *spec;
186 void *mv_misc = NULL;
187 void *mc_misc = NULL;
192 match_header |= MLX5_MATCH_MISC_PARAMETERS;
194 spec = kvzalloc(sizeof(*spec), GFP_KERNEL);
198 dmac_v = MLX5_ADDR_OF(fte_match_param, spec->match_value,
199 outer_headers.dmac_47_16);
200 dmac_c = MLX5_ADDR_OF(fte_match_param, spec->match_criteria,
201 outer_headers.dmac_47_16);
203 if (match_header & MLX5_MATCH_OUTER_HEADERS) {
204 ether_addr_copy(dmac_v, mac_v);
205 ether_addr_copy(dmac_c, mac_c);
208 if (match_header & MLX5_MATCH_MISC_PARAMETERS) {
209 mv_misc = MLX5_ADDR_OF(fte_match_param, spec->match_value,
211 mc_misc = MLX5_ADDR_OF(fte_match_param, spec->match_criteria,
213 MLX5_SET(fte_match_set_misc, mv_misc, source_port, MLX5_VPORT_UPLINK);
214 MLX5_SET_TO_ONES(fte_match_set_misc, mc_misc, source_port);
217 dest.type = MLX5_FLOW_DESTINATION_TYPE_VPORT;
218 dest.vport.num = vport;
221 "\tFDB add rule dmac_v(%pM) dmac_c(%pM) -> vport(%d)\n",
222 dmac_v, dmac_c, vport);
223 spec->match_criteria_enable = match_header;
224 flow_act.action = MLX5_FLOW_CONTEXT_ACTION_FWD_DEST;
226 mlx5_add_flow_rules(esw->fdb_table.legacy.fdb, spec,
227 &flow_act, &dest, 1);
228 if (IS_ERR(flow_rule)) {
230 "FDB: Failed to add flow rule: dmac_v(%pM) dmac_c(%pM) -> vport(%d), err(%ld)\n",
231 dmac_v, dmac_c, vport, PTR_ERR(flow_rule));
239 static struct mlx5_flow_handle *
240 esw_fdb_set_vport_rule(struct mlx5_eswitch *esw, u8 mac[ETH_ALEN], u16 vport)
244 eth_broadcast_addr(mac_c);
245 return __esw_fdb_set_vport_rule(esw, vport, false, mac_c, mac);
248 static struct mlx5_flow_handle *
249 esw_fdb_set_vport_allmulti_rule(struct mlx5_eswitch *esw, u16 vport)
254 eth_zero_addr(mac_c);
255 eth_zero_addr(mac_v);
258 return __esw_fdb_set_vport_rule(esw, vport, false, mac_c, mac_v);
261 static struct mlx5_flow_handle *
262 esw_fdb_set_vport_promisc_rule(struct mlx5_eswitch *esw, u16 vport)
267 eth_zero_addr(mac_c);
268 eth_zero_addr(mac_v);
269 return __esw_fdb_set_vport_rule(esw, vport, true, mac_c, mac_v);
273 LEGACY_VEPA_PRIO = 0,
277 static int esw_create_legacy_vepa_table(struct mlx5_eswitch *esw)
279 struct mlx5_core_dev *dev = esw->dev;
280 struct mlx5_flow_namespace *root_ns;
281 struct mlx5_flow_table *fdb;
284 root_ns = mlx5_get_fdb_sub_ns(dev, 0);
286 esw_warn(dev, "Failed to get FDB flow namespace\n");
290 /* num FTE 2, num FG 2 */
291 fdb = mlx5_create_auto_grouped_flow_table(root_ns, LEGACY_VEPA_PRIO,
295 esw_warn(dev, "Failed to create VEPA FDB err %d\n", err);
298 esw->fdb_table.legacy.vepa_fdb = fdb;
303 static int esw_create_legacy_fdb_table(struct mlx5_eswitch *esw)
305 int inlen = MLX5_ST_SZ_BYTES(create_flow_group_in);
306 struct mlx5_flow_table_attr ft_attr = {};
307 struct mlx5_core_dev *dev = esw->dev;
308 struct mlx5_flow_namespace *root_ns;
309 struct mlx5_flow_table *fdb;
310 struct mlx5_flow_group *g;
311 void *match_criteria;
317 esw_debug(dev, "Create FDB log_max_size(%d)\n",
318 MLX5_CAP_ESW_FLOWTABLE_FDB(dev, log_max_ft_size));
320 root_ns = mlx5_get_fdb_sub_ns(dev, 0);
322 esw_warn(dev, "Failed to get FDB flow namespace\n");
326 flow_group_in = kvzalloc(inlen, GFP_KERNEL);
330 table_size = BIT(MLX5_CAP_ESW_FLOWTABLE_FDB(dev, log_max_ft_size));
331 ft_attr.max_fte = table_size;
332 ft_attr.prio = LEGACY_FDB_PRIO;
333 fdb = mlx5_create_flow_table(root_ns, &ft_attr);
336 esw_warn(dev, "Failed to create FDB Table err %d\n", err);
339 esw->fdb_table.legacy.fdb = fdb;
341 /* Addresses group : Full match unicast/multicast addresses */
342 MLX5_SET(create_flow_group_in, flow_group_in, match_criteria_enable,
343 MLX5_MATCH_OUTER_HEADERS);
344 match_criteria = MLX5_ADDR_OF(create_flow_group_in, flow_group_in, match_criteria);
345 dmac = MLX5_ADDR_OF(fte_match_param, match_criteria, outer_headers.dmac_47_16);
346 MLX5_SET(create_flow_group_in, flow_group_in, start_flow_index, 0);
347 /* Preserve 2 entries for allmulti and promisc rules*/
348 MLX5_SET(create_flow_group_in, flow_group_in, end_flow_index, table_size - 3);
349 eth_broadcast_addr(dmac);
350 g = mlx5_create_flow_group(fdb, flow_group_in);
353 esw_warn(dev, "Failed to create flow group err(%d)\n", err);
356 esw->fdb_table.legacy.addr_grp = g;
358 /* Allmulti group : One rule that forwards any mcast traffic */
359 MLX5_SET(create_flow_group_in, flow_group_in, match_criteria_enable,
360 MLX5_MATCH_OUTER_HEADERS);
361 MLX5_SET(create_flow_group_in, flow_group_in, start_flow_index, table_size - 2);
362 MLX5_SET(create_flow_group_in, flow_group_in, end_flow_index, table_size - 2);
365 g = mlx5_create_flow_group(fdb, flow_group_in);
368 esw_warn(dev, "Failed to create allmulti flow group err(%d)\n", err);
371 esw->fdb_table.legacy.allmulti_grp = g;
373 /* Promiscuous group :
374 * One rule that forward all unmatched traffic from previous groups
377 MLX5_SET(create_flow_group_in, flow_group_in, match_criteria_enable,
378 MLX5_MATCH_MISC_PARAMETERS);
379 MLX5_SET_TO_ONES(fte_match_param, match_criteria, misc_parameters.source_port);
380 MLX5_SET(create_flow_group_in, flow_group_in, start_flow_index, table_size - 1);
381 MLX5_SET(create_flow_group_in, flow_group_in, end_flow_index, table_size - 1);
382 g = mlx5_create_flow_group(fdb, flow_group_in);
385 esw_warn(dev, "Failed to create promisc flow group err(%d)\n", err);
388 esw->fdb_table.legacy.promisc_grp = g;
392 esw_destroy_legacy_fdb_table(esw);
394 kvfree(flow_group_in);
398 static void esw_destroy_legacy_vepa_table(struct mlx5_eswitch *esw)
400 esw_debug(esw->dev, "Destroy VEPA Table\n");
401 if (!esw->fdb_table.legacy.vepa_fdb)
404 mlx5_destroy_flow_table(esw->fdb_table.legacy.vepa_fdb);
405 esw->fdb_table.legacy.vepa_fdb = NULL;
408 static void esw_destroy_legacy_fdb_table(struct mlx5_eswitch *esw)
410 esw_debug(esw->dev, "Destroy FDB Table\n");
411 if (!esw->fdb_table.legacy.fdb)
414 if (esw->fdb_table.legacy.promisc_grp)
415 mlx5_destroy_flow_group(esw->fdb_table.legacy.promisc_grp);
416 if (esw->fdb_table.legacy.allmulti_grp)
417 mlx5_destroy_flow_group(esw->fdb_table.legacy.allmulti_grp);
418 if (esw->fdb_table.legacy.addr_grp)
419 mlx5_destroy_flow_group(esw->fdb_table.legacy.addr_grp);
420 mlx5_destroy_flow_table(esw->fdb_table.legacy.fdb);
422 esw->fdb_table.legacy.fdb = NULL;
423 esw->fdb_table.legacy.addr_grp = NULL;
424 esw->fdb_table.legacy.allmulti_grp = NULL;
425 esw->fdb_table.legacy.promisc_grp = NULL;
428 static int esw_create_legacy_table(struct mlx5_eswitch *esw)
432 memset(&esw->fdb_table.legacy, 0, sizeof(struct legacy_fdb));
434 err = esw_create_legacy_vepa_table(esw);
438 err = esw_create_legacy_fdb_table(esw);
440 esw_destroy_legacy_vepa_table(esw);
445 static void esw_destroy_legacy_table(struct mlx5_eswitch *esw)
447 esw_cleanup_vepa_rules(esw);
448 esw_destroy_legacy_fdb_table(esw);
449 esw_destroy_legacy_vepa_table(esw);
452 /* E-Switch vport UC/MC lists management */
453 typedef int (*vport_addr_action)(struct mlx5_eswitch *esw,
454 struct vport_addr *vaddr);
456 static int esw_add_uc_addr(struct mlx5_eswitch *esw, struct vport_addr *vaddr)
458 u8 *mac = vaddr->node.addr;
459 u16 vport = vaddr->vport;
462 /* Skip mlx5_mpfs_add_mac for eswitch_managers,
463 * it is already done by its netdev in mlx5e_execute_l2_action
465 if (esw->manager_vport == vport)
468 err = mlx5_mpfs_add_mac(esw->dev, mac);
471 "Failed to add L2 table mac(%pM) for vport(0x%x), err(%d)\n",
478 /* SRIOV is enabled: Forward UC MAC to vport */
479 if (esw->fdb_table.legacy.fdb && esw->mode == SRIOV_LEGACY)
480 vaddr->flow_rule = esw_fdb_set_vport_rule(esw, mac, vport);
482 esw_debug(esw->dev, "\tADDED UC MAC: vport[%d] %pM fr(%p)\n",
483 vport, mac, vaddr->flow_rule);
488 static int esw_del_uc_addr(struct mlx5_eswitch *esw, struct vport_addr *vaddr)
490 u8 *mac = vaddr->node.addr;
491 u16 vport = vaddr->vport;
494 /* Skip mlx5_mpfs_del_mac for eswitch managerss,
495 * it is already done by its netdev in mlx5e_execute_l2_action
497 if (!vaddr->mpfs || esw->manager_vport == vport)
500 err = mlx5_mpfs_del_mac(esw->dev, mac);
503 "Failed to del L2 table mac(%pM) for vport(%d), err(%d)\n",
508 if (vaddr->flow_rule)
509 mlx5_del_flow_rules(vaddr->flow_rule);
510 vaddr->flow_rule = NULL;
515 static void update_allmulti_vports(struct mlx5_eswitch *esw,
516 struct vport_addr *vaddr,
517 struct esw_mc_addr *esw_mc)
519 u8 *mac = vaddr->node.addr;
520 struct mlx5_vport *vport;
523 mlx5_esw_for_all_vports(esw, i, vport) {
524 struct hlist_head *vport_hash = vport->mc_list;
525 struct vport_addr *iter_vaddr =
526 l2addr_hash_find(vport_hash,
529 vport_num = vport->vport;
530 if (IS_ERR_OR_NULL(vport->allmulti_rule) ||
531 vaddr->vport == vport_num)
533 switch (vaddr->action) {
534 case MLX5_ACTION_ADD:
537 iter_vaddr = l2addr_hash_add(vport_hash, mac,
542 "ALL-MULTI: Failed to add MAC(%pM) to vport[%d] DB\n",
546 iter_vaddr->vport = vport_num;
547 iter_vaddr->flow_rule =
548 esw_fdb_set_vport_rule(esw,
551 iter_vaddr->mc_promisc = true;
553 case MLX5_ACTION_DEL:
556 mlx5_del_flow_rules(iter_vaddr->flow_rule);
557 l2addr_hash_del(iter_vaddr);
563 static int esw_add_mc_addr(struct mlx5_eswitch *esw, struct vport_addr *vaddr)
565 struct hlist_head *hash = esw->mc_table;
566 struct esw_mc_addr *esw_mc;
567 u8 *mac = vaddr->node.addr;
568 u16 vport = vaddr->vport;
570 if (!esw->fdb_table.legacy.fdb)
573 esw_mc = l2addr_hash_find(hash, mac, struct esw_mc_addr);
577 esw_mc = l2addr_hash_add(hash, mac, struct esw_mc_addr, GFP_KERNEL);
581 esw_mc->uplink_rule = /* Forward MC MAC to Uplink */
582 esw_fdb_set_vport_rule(esw, mac, MLX5_VPORT_UPLINK);
584 /* Add this multicast mac to all the mc promiscuous vports */
585 update_allmulti_vports(esw, vaddr, esw_mc);
588 /* If the multicast mac is added as a result of mc promiscuous vport,
589 * don't increment the multicast ref count
591 if (!vaddr->mc_promisc)
594 /* Forward MC MAC to vport */
595 vaddr->flow_rule = esw_fdb_set_vport_rule(esw, mac, vport);
597 "\tADDED MC MAC: vport[%d] %pM fr(%p) refcnt(%d) uplinkfr(%p)\n",
598 vport, mac, vaddr->flow_rule,
599 esw_mc->refcnt, esw_mc->uplink_rule);
603 static int esw_del_mc_addr(struct mlx5_eswitch *esw, struct vport_addr *vaddr)
605 struct hlist_head *hash = esw->mc_table;
606 struct esw_mc_addr *esw_mc;
607 u8 *mac = vaddr->node.addr;
608 u16 vport = vaddr->vport;
610 if (!esw->fdb_table.legacy.fdb)
613 esw_mc = l2addr_hash_find(hash, mac, struct esw_mc_addr);
616 "Failed to find eswitch MC addr for MAC(%pM) vport(%d)",
621 "\tDELETE MC MAC: vport[%d] %pM fr(%p) refcnt(%d) uplinkfr(%p)\n",
622 vport, mac, vaddr->flow_rule, esw_mc->refcnt,
623 esw_mc->uplink_rule);
625 if (vaddr->flow_rule)
626 mlx5_del_flow_rules(vaddr->flow_rule);
627 vaddr->flow_rule = NULL;
629 /* If the multicast mac is added as a result of mc promiscuous vport,
630 * don't decrement the multicast ref count.
632 if (vaddr->mc_promisc || (--esw_mc->refcnt > 0))
635 /* Remove this multicast mac from all the mc promiscuous vports */
636 update_allmulti_vports(esw, vaddr, esw_mc);
638 if (esw_mc->uplink_rule)
639 mlx5_del_flow_rules(esw_mc->uplink_rule);
641 l2addr_hash_del(esw_mc);
645 /* Apply vport UC/MC list to HW l2 table and FDB table */
646 static void esw_apply_vport_addr_list(struct mlx5_eswitch *esw,
647 u16 vport_num, int list_type)
649 struct mlx5_vport *vport = mlx5_eswitch_get_vport(esw, vport_num);
650 bool is_uc = list_type == MLX5_NVPRT_LIST_TYPE_UC;
651 vport_addr_action vport_addr_add;
652 vport_addr_action vport_addr_del;
653 struct vport_addr *addr;
654 struct l2addr_node *node;
655 struct hlist_head *hash;
656 struct hlist_node *tmp;
659 vport_addr_add = is_uc ? esw_add_uc_addr :
661 vport_addr_del = is_uc ? esw_del_uc_addr :
664 hash = is_uc ? vport->uc_list : vport->mc_list;
665 for_each_l2hash_node(node, tmp, hash, hi) {
666 addr = container_of(node, struct vport_addr, node);
667 switch (addr->action) {
668 case MLX5_ACTION_ADD:
669 vport_addr_add(esw, addr);
670 addr->action = MLX5_ACTION_NONE;
672 case MLX5_ACTION_DEL:
673 vport_addr_del(esw, addr);
674 l2addr_hash_del(addr);
680 /* Sync vport UC/MC list from vport context */
681 static void esw_update_vport_addr_list(struct mlx5_eswitch *esw,
682 u16 vport_num, int list_type)
684 struct mlx5_vport *vport = mlx5_eswitch_get_vport(esw, vport_num);
685 bool is_uc = list_type == MLX5_NVPRT_LIST_TYPE_UC;
686 u8 (*mac_list)[ETH_ALEN];
687 struct l2addr_node *node;
688 struct vport_addr *addr;
689 struct hlist_head *hash;
690 struct hlist_node *tmp;
696 size = is_uc ? MLX5_MAX_UC_PER_VPORT(esw->dev) :
697 MLX5_MAX_MC_PER_VPORT(esw->dev);
699 mac_list = kcalloc(size, ETH_ALEN, GFP_KERNEL);
703 hash = is_uc ? vport->uc_list : vport->mc_list;
705 for_each_l2hash_node(node, tmp, hash, hi) {
706 addr = container_of(node, struct vport_addr, node);
707 addr->action = MLX5_ACTION_DEL;
713 err = mlx5_query_nic_vport_mac_list(esw->dev, vport_num, list_type,
717 esw_debug(esw->dev, "vport[%d] context update %s list size (%d)\n",
718 vport_num, is_uc ? "UC" : "MC", size);
720 for (i = 0; i < size; i++) {
721 if (is_uc && !is_valid_ether_addr(mac_list[i]))
724 if (!is_uc && !is_multicast_ether_addr(mac_list[i]))
727 addr = l2addr_hash_find(hash, mac_list[i], struct vport_addr);
729 addr->action = MLX5_ACTION_NONE;
730 /* If this mac was previously added because of allmulti
731 * promiscuous rx mode, its now converted to be original
734 if (addr->mc_promisc) {
735 struct esw_mc_addr *esw_mc =
736 l2addr_hash_find(esw->mc_table,
741 "Failed to MAC(%pM) in mcast DB\n",
746 addr->mc_promisc = false;
751 addr = l2addr_hash_add(hash, mac_list[i], struct vport_addr,
755 "Failed to add MAC(%pM) to vport[%d] DB\n",
756 mac_list[i], vport_num);
759 addr->vport = vport_num;
760 addr->action = MLX5_ACTION_ADD;
766 /* Sync vport UC/MC list from vport context
767 * Must be called after esw_update_vport_addr_list
769 static void esw_update_vport_mc_promisc(struct mlx5_eswitch *esw, u16 vport_num)
771 struct mlx5_vport *vport = mlx5_eswitch_get_vport(esw, vport_num);
772 struct l2addr_node *node;
773 struct vport_addr *addr;
774 struct hlist_head *hash;
775 struct hlist_node *tmp;
778 hash = vport->mc_list;
780 for_each_l2hash_node(node, tmp, esw->mc_table, hi) {
781 u8 *mac = node->addr;
783 addr = l2addr_hash_find(hash, mac, struct vport_addr);
785 if (addr->action == MLX5_ACTION_DEL)
786 addr->action = MLX5_ACTION_NONE;
789 addr = l2addr_hash_add(hash, mac, struct vport_addr,
793 "Failed to add allmulti MAC(%pM) to vport[%d] DB\n",
797 addr->vport = vport_num;
798 addr->action = MLX5_ACTION_ADD;
799 addr->mc_promisc = true;
803 /* Apply vport rx mode to HW FDB table */
804 static void esw_apply_vport_rx_mode(struct mlx5_eswitch *esw, u16 vport_num,
805 bool promisc, bool mc_promisc)
807 struct mlx5_vport *vport = mlx5_eswitch_get_vport(esw, vport_num);
808 struct esw_mc_addr *allmulti_addr = &esw->mc_promisc;
810 if (IS_ERR_OR_NULL(vport->allmulti_rule) != mc_promisc)
814 vport->allmulti_rule =
815 esw_fdb_set_vport_allmulti_rule(esw, vport_num);
816 if (!allmulti_addr->uplink_rule)
817 allmulti_addr->uplink_rule =
818 esw_fdb_set_vport_allmulti_rule(esw,
820 allmulti_addr->refcnt++;
821 } else if (vport->allmulti_rule) {
822 mlx5_del_flow_rules(vport->allmulti_rule);
823 vport->allmulti_rule = NULL;
825 if (--allmulti_addr->refcnt > 0)
828 if (allmulti_addr->uplink_rule)
829 mlx5_del_flow_rules(allmulti_addr->uplink_rule);
830 allmulti_addr->uplink_rule = NULL;
834 if (IS_ERR_OR_NULL(vport->promisc_rule) != promisc)
838 vport->promisc_rule = esw_fdb_set_vport_promisc_rule(esw,
840 } else if (vport->promisc_rule) {
841 mlx5_del_flow_rules(vport->promisc_rule);
842 vport->promisc_rule = NULL;
846 /* Sync vport rx mode from vport context */
847 static void esw_update_vport_rx_mode(struct mlx5_eswitch *esw, u16 vport_num)
849 struct mlx5_vport *vport = mlx5_eswitch_get_vport(esw, vport_num);
855 err = mlx5_query_nic_vport_promisc(esw->dev,
862 esw_debug(esw->dev, "vport[%d] context update rx mode promisc_all=%d, all_multi=%d\n",
863 vport_num, promisc_all, promisc_mc);
865 if (!vport->info.trusted || !vport->enabled) {
871 esw_apply_vport_rx_mode(esw, vport_num, promisc_all,
872 (promisc_all || promisc_mc));
875 static void esw_vport_change_handle_locked(struct mlx5_vport *vport)
877 struct mlx5_core_dev *dev = vport->dev;
878 struct mlx5_eswitch *esw = dev->priv.eswitch;
881 mlx5_query_nic_vport_mac_address(dev, vport->vport, mac);
882 esw_debug(dev, "vport[%d] Context Changed: perm mac: %pM\n",
885 if (vport->enabled_events & UC_ADDR_CHANGE) {
886 esw_update_vport_addr_list(esw, vport->vport,
887 MLX5_NVPRT_LIST_TYPE_UC);
888 esw_apply_vport_addr_list(esw, vport->vport,
889 MLX5_NVPRT_LIST_TYPE_UC);
892 if (vport->enabled_events & MC_ADDR_CHANGE) {
893 esw_update_vport_addr_list(esw, vport->vport,
894 MLX5_NVPRT_LIST_TYPE_MC);
897 if (vport->enabled_events & PROMISC_CHANGE) {
898 esw_update_vport_rx_mode(esw, vport->vport);
899 if (!IS_ERR_OR_NULL(vport->allmulti_rule))
900 esw_update_vport_mc_promisc(esw, vport->vport);
903 if (vport->enabled_events & (PROMISC_CHANGE | MC_ADDR_CHANGE)) {
904 esw_apply_vport_addr_list(esw, vport->vport,
905 MLX5_NVPRT_LIST_TYPE_MC);
908 esw_debug(esw->dev, "vport[%d] Context Changed: Done\n", vport->vport);
910 arm_vport_context_events_cmd(dev, vport->vport,
911 vport->enabled_events);
914 static void esw_vport_change_handler(struct work_struct *work)
916 struct mlx5_vport *vport =
917 container_of(work, struct mlx5_vport, vport_change_handler);
918 struct mlx5_eswitch *esw = vport->dev->priv.eswitch;
920 mutex_lock(&esw->state_lock);
921 esw_vport_change_handle_locked(vport);
922 mutex_unlock(&esw->state_lock);
925 static int esw_vport_enable_egress_acl(struct mlx5_eswitch *esw,
926 struct mlx5_vport *vport)
928 int inlen = MLX5_ST_SZ_BYTES(create_flow_group_in);
929 struct mlx5_flow_group *vlan_grp = NULL;
930 struct mlx5_flow_group *drop_grp = NULL;
931 struct mlx5_core_dev *dev = esw->dev;
932 struct mlx5_flow_namespace *root_ns;
933 struct mlx5_flow_table *acl;
934 void *match_criteria;
936 /* The egress acl table contains 2 rules:
937 * 1)Allow traffic with vlan_tag=vst_vlan_id
938 * 2)Drop all other traffic.
943 if (!MLX5_CAP_ESW_EGRESS_ACL(dev, ft_support))
946 if (!IS_ERR_OR_NULL(vport->egress.acl))
949 esw_debug(dev, "Create vport[%d] egress ACL log_max_size(%d)\n",
950 vport->vport, MLX5_CAP_ESW_EGRESS_ACL(dev, log_max_ft_size));
952 root_ns = mlx5_get_flow_vport_acl_namespace(dev, MLX5_FLOW_NAMESPACE_ESW_EGRESS,
955 esw_warn(dev, "Failed to get E-Switch egress flow namespace for vport (%d)\n", vport->vport);
959 flow_group_in = kvzalloc(inlen, GFP_KERNEL);
963 acl = mlx5_create_vport_flow_table(root_ns, 0, table_size, 0, vport->vport);
966 esw_warn(dev, "Failed to create E-Switch vport[%d] egress flow Table, err(%d)\n",
971 MLX5_SET(create_flow_group_in, flow_group_in, match_criteria_enable, MLX5_MATCH_OUTER_HEADERS);
972 match_criteria = MLX5_ADDR_OF(create_flow_group_in, flow_group_in, match_criteria);
973 MLX5_SET_TO_ONES(fte_match_param, match_criteria, outer_headers.cvlan_tag);
974 MLX5_SET_TO_ONES(fte_match_param, match_criteria, outer_headers.first_vid);
975 MLX5_SET(create_flow_group_in, flow_group_in, start_flow_index, 0);
976 MLX5_SET(create_flow_group_in, flow_group_in, end_flow_index, 0);
978 vlan_grp = mlx5_create_flow_group(acl, flow_group_in);
979 if (IS_ERR(vlan_grp)) {
980 err = PTR_ERR(vlan_grp);
981 esw_warn(dev, "Failed to create E-Switch vport[%d] egress allowed vlans flow group, err(%d)\n",
986 memset(flow_group_in, 0, inlen);
987 MLX5_SET(create_flow_group_in, flow_group_in, start_flow_index, 1);
988 MLX5_SET(create_flow_group_in, flow_group_in, end_flow_index, 1);
989 drop_grp = mlx5_create_flow_group(acl, flow_group_in);
990 if (IS_ERR(drop_grp)) {
991 err = PTR_ERR(drop_grp);
992 esw_warn(dev, "Failed to create E-Switch vport[%d] egress drop flow group, err(%d)\n",
997 vport->egress.acl = acl;
998 vport->egress.drop_grp = drop_grp;
999 vport->egress.allowed_vlans_grp = vlan_grp;
1001 kvfree(flow_group_in);
1002 if (err && !IS_ERR_OR_NULL(vlan_grp))
1003 mlx5_destroy_flow_group(vlan_grp);
1004 if (err && !IS_ERR_OR_NULL(acl))
1005 mlx5_destroy_flow_table(acl);
1009 static void esw_vport_cleanup_egress_rules(struct mlx5_eswitch *esw,
1010 struct mlx5_vport *vport)
1012 if (!IS_ERR_OR_NULL(vport->egress.allowed_vlan))
1013 mlx5_del_flow_rules(vport->egress.allowed_vlan);
1015 if (!IS_ERR_OR_NULL(vport->egress.drop_rule))
1016 mlx5_del_flow_rules(vport->egress.drop_rule);
1018 vport->egress.allowed_vlan = NULL;
1019 vport->egress.drop_rule = NULL;
1022 static void esw_vport_disable_egress_acl(struct mlx5_eswitch *esw,
1023 struct mlx5_vport *vport)
1025 if (IS_ERR_OR_NULL(vport->egress.acl))
1028 esw_debug(esw->dev, "Destroy vport[%d] E-Switch egress ACL\n", vport->vport);
1030 esw_vport_cleanup_egress_rules(esw, vport);
1031 mlx5_destroy_flow_group(vport->egress.allowed_vlans_grp);
1032 mlx5_destroy_flow_group(vport->egress.drop_grp);
1033 mlx5_destroy_flow_table(vport->egress.acl);
1034 vport->egress.allowed_vlans_grp = NULL;
1035 vport->egress.drop_grp = NULL;
1036 vport->egress.acl = NULL;
1039 static int esw_vport_enable_ingress_acl(struct mlx5_eswitch *esw,
1040 struct mlx5_vport *vport)
1042 int inlen = MLX5_ST_SZ_BYTES(create_flow_group_in);
1043 struct mlx5_core_dev *dev = esw->dev;
1044 struct mlx5_flow_namespace *root_ns;
1045 struct mlx5_flow_table *acl;
1046 struct mlx5_flow_group *g;
1047 void *match_criteria;
1049 /* The ingress acl table contains 4 groups
1050 * (2 active rules at the same time -
1051 * 1 allow rule from one of the first 3 groups.
1052 * 1 drop rule from the last group):
1053 * 1)Allow untagged traffic with smac=original mac.
1054 * 2)Allow untagged traffic.
1055 * 3)Allow traffic with smac=original mac.
1056 * 4)Drop all other traffic.
1061 if (!MLX5_CAP_ESW_INGRESS_ACL(dev, ft_support))
1064 if (!IS_ERR_OR_NULL(vport->ingress.acl))
1067 esw_debug(dev, "Create vport[%d] ingress ACL log_max_size(%d)\n",
1068 vport->vport, MLX5_CAP_ESW_INGRESS_ACL(dev, log_max_ft_size));
1070 root_ns = mlx5_get_flow_vport_acl_namespace(dev, MLX5_FLOW_NAMESPACE_ESW_INGRESS,
1073 esw_warn(dev, "Failed to get E-Switch ingress flow namespace for vport (%d)\n", vport->vport);
1077 flow_group_in = kvzalloc(inlen, GFP_KERNEL);
1081 acl = mlx5_create_vport_flow_table(root_ns, 0, table_size, 0, vport->vport);
1084 esw_warn(dev, "Failed to create E-Switch vport[%d] ingress flow Table, err(%d)\n",
1088 vport->ingress.acl = acl;
1090 match_criteria = MLX5_ADDR_OF(create_flow_group_in, flow_group_in, match_criteria);
1092 MLX5_SET(create_flow_group_in, flow_group_in, match_criteria_enable, MLX5_MATCH_OUTER_HEADERS);
1093 MLX5_SET_TO_ONES(fte_match_param, match_criteria, outer_headers.cvlan_tag);
1094 MLX5_SET_TO_ONES(fte_match_param, match_criteria, outer_headers.smac_47_16);
1095 MLX5_SET_TO_ONES(fte_match_param, match_criteria, outer_headers.smac_15_0);
1096 MLX5_SET(create_flow_group_in, flow_group_in, start_flow_index, 0);
1097 MLX5_SET(create_flow_group_in, flow_group_in, end_flow_index, 0);
1099 g = mlx5_create_flow_group(acl, flow_group_in);
1102 esw_warn(dev, "Failed to create E-Switch vport[%d] ingress untagged spoofchk flow group, err(%d)\n",
1106 vport->ingress.allow_untagged_spoofchk_grp = g;
1108 memset(flow_group_in, 0, inlen);
1109 MLX5_SET(create_flow_group_in, flow_group_in, match_criteria_enable, MLX5_MATCH_OUTER_HEADERS);
1110 MLX5_SET_TO_ONES(fte_match_param, match_criteria, outer_headers.cvlan_tag);
1111 MLX5_SET(create_flow_group_in, flow_group_in, start_flow_index, 1);
1112 MLX5_SET(create_flow_group_in, flow_group_in, end_flow_index, 1);
1114 g = mlx5_create_flow_group(acl, flow_group_in);
1117 esw_warn(dev, "Failed to create E-Switch vport[%d] ingress untagged flow group, err(%d)\n",
1121 vport->ingress.allow_untagged_only_grp = g;
1123 memset(flow_group_in, 0, inlen);
1124 MLX5_SET(create_flow_group_in, flow_group_in, match_criteria_enable, MLX5_MATCH_OUTER_HEADERS);
1125 MLX5_SET_TO_ONES(fte_match_param, match_criteria, outer_headers.smac_47_16);
1126 MLX5_SET_TO_ONES(fte_match_param, match_criteria, outer_headers.smac_15_0);
1127 MLX5_SET(create_flow_group_in, flow_group_in, start_flow_index, 2);
1128 MLX5_SET(create_flow_group_in, flow_group_in, end_flow_index, 2);
1130 g = mlx5_create_flow_group(acl, flow_group_in);
1133 esw_warn(dev, "Failed to create E-Switch vport[%d] ingress spoofchk flow group, err(%d)\n",
1137 vport->ingress.allow_spoofchk_only_grp = g;
1139 memset(flow_group_in, 0, inlen);
1140 MLX5_SET(create_flow_group_in, flow_group_in, start_flow_index, 3);
1141 MLX5_SET(create_flow_group_in, flow_group_in, end_flow_index, 3);
1143 g = mlx5_create_flow_group(acl, flow_group_in);
1146 esw_warn(dev, "Failed to create E-Switch vport[%d] ingress drop flow group, err(%d)\n",
1150 vport->ingress.drop_grp = g;
1154 if (!IS_ERR_OR_NULL(vport->ingress.allow_spoofchk_only_grp))
1155 mlx5_destroy_flow_group(
1156 vport->ingress.allow_spoofchk_only_grp);
1157 if (!IS_ERR_OR_NULL(vport->ingress.allow_untagged_only_grp))
1158 mlx5_destroy_flow_group(
1159 vport->ingress.allow_untagged_only_grp);
1160 if (!IS_ERR_OR_NULL(vport->ingress.allow_untagged_spoofchk_grp))
1161 mlx5_destroy_flow_group(
1162 vport->ingress.allow_untagged_spoofchk_grp);
1163 if (!IS_ERR_OR_NULL(vport->ingress.acl))
1164 mlx5_destroy_flow_table(vport->ingress.acl);
1167 kvfree(flow_group_in);
1171 static void esw_vport_cleanup_ingress_rules(struct mlx5_eswitch *esw,
1172 struct mlx5_vport *vport)
1174 if (!IS_ERR_OR_NULL(vport->ingress.drop_rule))
1175 mlx5_del_flow_rules(vport->ingress.drop_rule);
1177 if (!IS_ERR_OR_NULL(vport->ingress.allow_rule))
1178 mlx5_del_flow_rules(vport->ingress.allow_rule);
1180 vport->ingress.drop_rule = NULL;
1181 vport->ingress.allow_rule = NULL;
1184 static void esw_vport_disable_ingress_acl(struct mlx5_eswitch *esw,
1185 struct mlx5_vport *vport)
1187 if (IS_ERR_OR_NULL(vport->ingress.acl))
1190 esw_debug(esw->dev, "Destroy vport[%d] E-Switch ingress ACL\n", vport->vport);
1192 esw_vport_cleanup_ingress_rules(esw, vport);
1193 mlx5_destroy_flow_group(vport->ingress.allow_spoofchk_only_grp);
1194 mlx5_destroy_flow_group(vport->ingress.allow_untagged_only_grp);
1195 mlx5_destroy_flow_group(vport->ingress.allow_untagged_spoofchk_grp);
1196 mlx5_destroy_flow_group(vport->ingress.drop_grp);
1197 mlx5_destroy_flow_table(vport->ingress.acl);
1198 vport->ingress.acl = NULL;
1199 vport->ingress.drop_grp = NULL;
1200 vport->ingress.allow_spoofchk_only_grp = NULL;
1201 vport->ingress.allow_untagged_only_grp = NULL;
1202 vport->ingress.allow_untagged_spoofchk_grp = NULL;
1205 static int esw_vport_ingress_config(struct mlx5_eswitch *esw,
1206 struct mlx5_vport *vport)
1208 struct mlx5_fc *counter = vport->ingress.drop_counter;
1209 struct mlx5_flow_destination drop_ctr_dst = {0};
1210 struct mlx5_flow_destination *dst = NULL;
1211 struct mlx5_flow_act flow_act = {0};
1212 struct mlx5_flow_spec *spec;
1217 esw_vport_cleanup_ingress_rules(esw, vport);
1219 if (!vport->info.vlan && !vport->info.qos && !vport->info.spoofchk) {
1220 esw_vport_disable_ingress_acl(esw, vport);
1224 err = esw_vport_enable_ingress_acl(esw, vport);
1226 mlx5_core_warn(esw->dev,
1227 "failed to enable ingress acl (%d) on vport[%d]\n",
1233 "vport[%d] configure ingress rules, vlan(%d) qos(%d)\n",
1234 vport->vport, vport->info.vlan, vport->info.qos);
1236 spec = kvzalloc(sizeof(*spec), GFP_KERNEL);
1242 if (vport->info.vlan || vport->info.qos)
1243 MLX5_SET_TO_ONES(fte_match_param, spec->match_criteria, outer_headers.cvlan_tag);
1245 if (vport->info.spoofchk) {
1246 MLX5_SET_TO_ONES(fte_match_param, spec->match_criteria, outer_headers.smac_47_16);
1247 MLX5_SET_TO_ONES(fte_match_param, spec->match_criteria, outer_headers.smac_15_0);
1248 smac_v = MLX5_ADDR_OF(fte_match_param,
1250 outer_headers.smac_47_16);
1251 ether_addr_copy(smac_v, vport->info.mac);
1254 spec->match_criteria_enable = MLX5_MATCH_OUTER_HEADERS;
1255 flow_act.action = MLX5_FLOW_CONTEXT_ACTION_ALLOW;
1256 vport->ingress.allow_rule =
1257 mlx5_add_flow_rules(vport->ingress.acl, spec,
1258 &flow_act, NULL, 0);
1259 if (IS_ERR(vport->ingress.allow_rule)) {
1260 err = PTR_ERR(vport->ingress.allow_rule);
1262 "vport[%d] configure ingress allow rule, err(%d)\n",
1264 vport->ingress.allow_rule = NULL;
1268 memset(spec, 0, sizeof(*spec));
1269 flow_act.action = MLX5_FLOW_CONTEXT_ACTION_DROP;
1271 /* Attach drop flow counter */
1273 flow_act.action |= MLX5_FLOW_CONTEXT_ACTION_COUNT;
1274 drop_ctr_dst.type = MLX5_FLOW_DESTINATION_TYPE_COUNTER;
1275 drop_ctr_dst.counter_id = mlx5_fc_id(counter);
1276 dst = &drop_ctr_dst;
1279 vport->ingress.drop_rule =
1280 mlx5_add_flow_rules(vport->ingress.acl, spec,
1281 &flow_act, dst, dest_num);
1282 if (IS_ERR(vport->ingress.drop_rule)) {
1283 err = PTR_ERR(vport->ingress.drop_rule);
1285 "vport[%d] configure ingress drop rule, err(%d)\n",
1287 vport->ingress.drop_rule = NULL;
1293 esw_vport_cleanup_ingress_rules(esw, vport);
1298 static int esw_vport_egress_config(struct mlx5_eswitch *esw,
1299 struct mlx5_vport *vport)
1301 struct mlx5_fc *counter = vport->egress.drop_counter;
1302 struct mlx5_flow_destination drop_ctr_dst = {0};
1303 struct mlx5_flow_destination *dst = NULL;
1304 struct mlx5_flow_act flow_act = {0};
1305 struct mlx5_flow_spec *spec;
1309 esw_vport_cleanup_egress_rules(esw, vport);
1311 if (!vport->info.vlan && !vport->info.qos) {
1312 esw_vport_disable_egress_acl(esw, vport);
1316 err = esw_vport_enable_egress_acl(esw, vport);
1318 mlx5_core_warn(esw->dev,
1319 "failed to enable egress acl (%d) on vport[%d]\n",
1325 "vport[%d] configure egress rules, vlan(%d) qos(%d)\n",
1326 vport->vport, vport->info.vlan, vport->info.qos);
1328 spec = kvzalloc(sizeof(*spec), GFP_KERNEL);
1334 /* Allowed vlan rule */
1335 MLX5_SET_TO_ONES(fte_match_param, spec->match_criteria, outer_headers.cvlan_tag);
1336 MLX5_SET_TO_ONES(fte_match_param, spec->match_value, outer_headers.cvlan_tag);
1337 MLX5_SET_TO_ONES(fte_match_param, spec->match_criteria, outer_headers.first_vid);
1338 MLX5_SET(fte_match_param, spec->match_value, outer_headers.first_vid, vport->info.vlan);
1340 spec->match_criteria_enable = MLX5_MATCH_OUTER_HEADERS;
1341 flow_act.action = MLX5_FLOW_CONTEXT_ACTION_ALLOW;
1342 vport->egress.allowed_vlan =
1343 mlx5_add_flow_rules(vport->egress.acl, spec,
1344 &flow_act, NULL, 0);
1345 if (IS_ERR(vport->egress.allowed_vlan)) {
1346 err = PTR_ERR(vport->egress.allowed_vlan);
1348 "vport[%d] configure egress allowed vlan rule failed, err(%d)\n",
1350 vport->egress.allowed_vlan = NULL;
1354 /* Drop others rule (star rule) */
1355 memset(spec, 0, sizeof(*spec));
1356 flow_act.action = MLX5_FLOW_CONTEXT_ACTION_DROP;
1358 /* Attach egress drop flow counter */
1360 flow_act.action |= MLX5_FLOW_CONTEXT_ACTION_COUNT;
1361 drop_ctr_dst.type = MLX5_FLOW_DESTINATION_TYPE_COUNTER;
1362 drop_ctr_dst.counter_id = mlx5_fc_id(counter);
1363 dst = &drop_ctr_dst;
1366 vport->egress.drop_rule =
1367 mlx5_add_flow_rules(vport->egress.acl, spec,
1368 &flow_act, dst, dest_num);
1369 if (IS_ERR(vport->egress.drop_rule)) {
1370 err = PTR_ERR(vport->egress.drop_rule);
1372 "vport[%d] configure egress drop rule failed, err(%d)\n",
1374 vport->egress.drop_rule = NULL;
1381 /* Vport QoS management */
1382 static int esw_create_tsar(struct mlx5_eswitch *esw)
1384 u32 tsar_ctx[MLX5_ST_SZ_DW(scheduling_context)] = {0};
1385 struct mlx5_core_dev *dev = esw->dev;
1388 if (!MLX5_CAP_GEN(dev, qos) || !MLX5_CAP_QOS(dev, esw_scheduling))
1391 if (esw->qos.enabled)
1394 err = mlx5_create_scheduling_element_cmd(dev,
1395 SCHEDULING_HIERARCHY_E_SWITCH,
1397 &esw->qos.root_tsar_id);
1399 esw_warn(esw->dev, "E-Switch create TSAR failed (%d)\n", err);
1403 esw->qos.enabled = true;
1407 static void esw_destroy_tsar(struct mlx5_eswitch *esw)
1411 if (!esw->qos.enabled)
1414 err = mlx5_destroy_scheduling_element_cmd(esw->dev,
1415 SCHEDULING_HIERARCHY_E_SWITCH,
1416 esw->qos.root_tsar_id);
1418 esw_warn(esw->dev, "E-Switch destroy TSAR failed (%d)\n", err);
1420 esw->qos.enabled = false;
1423 static int esw_vport_enable_qos(struct mlx5_eswitch *esw, int vport_num,
1424 u32 initial_max_rate, u32 initial_bw_share)
1426 struct mlx5_vport *vport = mlx5_eswitch_get_vport(esw, vport_num);
1427 u32 sched_ctx[MLX5_ST_SZ_DW(scheduling_context)] = {0};
1428 struct mlx5_core_dev *dev = esw->dev;
1432 if (!esw->qos.enabled || !MLX5_CAP_GEN(dev, qos) ||
1433 !MLX5_CAP_QOS(dev, esw_scheduling))
1436 if (vport->qos.enabled)
1439 MLX5_SET(scheduling_context, sched_ctx, element_type,
1440 SCHEDULING_CONTEXT_ELEMENT_TYPE_VPORT);
1441 vport_elem = MLX5_ADDR_OF(scheduling_context, sched_ctx,
1442 element_attributes);
1443 MLX5_SET(vport_element, vport_elem, vport_number, vport_num);
1444 MLX5_SET(scheduling_context, sched_ctx, parent_element_id,
1445 esw->qos.root_tsar_id);
1446 MLX5_SET(scheduling_context, sched_ctx, max_average_bw,
1448 MLX5_SET(scheduling_context, sched_ctx, bw_share, initial_bw_share);
1450 err = mlx5_create_scheduling_element_cmd(dev,
1451 SCHEDULING_HIERARCHY_E_SWITCH,
1453 &vport->qos.esw_tsar_ix);
1455 esw_warn(esw->dev, "E-Switch create TSAR vport element failed (vport=%d,err=%d)\n",
1460 vport->qos.enabled = true;
1464 static void esw_vport_disable_qos(struct mlx5_eswitch *esw, int vport_num)
1466 struct mlx5_vport *vport = mlx5_eswitch_get_vport(esw, vport_num);
1469 if (!vport->qos.enabled)
1472 err = mlx5_destroy_scheduling_element_cmd(esw->dev,
1473 SCHEDULING_HIERARCHY_E_SWITCH,
1474 vport->qos.esw_tsar_ix);
1476 esw_warn(esw->dev, "E-Switch destroy TSAR vport element failed (vport=%d,err=%d)\n",
1479 vport->qos.enabled = false;
1482 static int esw_vport_qos_config(struct mlx5_eswitch *esw, int vport_num,
1483 u32 max_rate, u32 bw_share)
1485 struct mlx5_vport *vport = mlx5_eswitch_get_vport(esw, vport_num);
1486 u32 sched_ctx[MLX5_ST_SZ_DW(scheduling_context)] = {0};
1487 struct mlx5_core_dev *dev = esw->dev;
1492 if (!MLX5_CAP_GEN(dev, qos) || !MLX5_CAP_QOS(dev, esw_scheduling))
1495 if (!vport->qos.enabled)
1498 MLX5_SET(scheduling_context, sched_ctx, element_type,
1499 SCHEDULING_CONTEXT_ELEMENT_TYPE_VPORT);
1500 vport_elem = MLX5_ADDR_OF(scheduling_context, sched_ctx,
1501 element_attributes);
1502 MLX5_SET(vport_element, vport_elem, vport_number, vport_num);
1503 MLX5_SET(scheduling_context, sched_ctx, parent_element_id,
1504 esw->qos.root_tsar_id);
1505 MLX5_SET(scheduling_context, sched_ctx, max_average_bw,
1507 MLX5_SET(scheduling_context, sched_ctx, bw_share, bw_share);
1508 bitmask |= MODIFY_SCHEDULING_ELEMENT_IN_MODIFY_BITMASK_MAX_AVERAGE_BW;
1509 bitmask |= MODIFY_SCHEDULING_ELEMENT_IN_MODIFY_BITMASK_BW_SHARE;
1511 err = mlx5_modify_scheduling_element_cmd(dev,
1512 SCHEDULING_HIERARCHY_E_SWITCH,
1514 vport->qos.esw_tsar_ix,
1517 esw_warn(esw->dev, "E-Switch modify TSAR vport element failed (vport=%d,err=%d)\n",
1525 static void node_guid_gen_from_mac(u64 *node_guid, u8 mac[ETH_ALEN])
1527 ((u8 *)node_guid)[7] = mac[0];
1528 ((u8 *)node_guid)[6] = mac[1];
1529 ((u8 *)node_guid)[5] = mac[2];
1530 ((u8 *)node_guid)[4] = 0xff;
1531 ((u8 *)node_guid)[3] = 0xfe;
1532 ((u8 *)node_guid)[2] = mac[3];
1533 ((u8 *)node_guid)[1] = mac[4];
1534 ((u8 *)node_guid)[0] = mac[5];
1537 static void esw_apply_vport_conf(struct mlx5_eswitch *esw,
1538 struct mlx5_vport *vport)
1540 int vport_num = vport->vport;
1542 if (esw->manager_vport == vport_num)
1545 mlx5_modify_vport_admin_state(esw->dev,
1546 MLX5_VPORT_STATE_OP_MOD_ESW_VPORT,
1548 vport->info.link_state);
1550 /* Host PF has its own mac/guid. */
1552 mlx5_modify_nic_vport_mac_address(esw->dev, vport_num,
1554 mlx5_modify_nic_vport_node_guid(esw->dev, vport_num,
1555 vport->info.node_guid);
1558 modify_esw_vport_cvlan(esw->dev, vport_num, vport->info.vlan, vport->info.qos,
1559 (vport->info.vlan || vport->info.qos));
1561 /* Only legacy mode needs ACLs */
1562 if (esw->mode == SRIOV_LEGACY) {
1563 esw_vport_ingress_config(esw, vport);
1564 esw_vport_egress_config(esw, vport);
1568 static void esw_vport_create_drop_counters(struct mlx5_vport *vport)
1570 struct mlx5_core_dev *dev = vport->dev;
1572 if (MLX5_CAP_ESW_INGRESS_ACL(dev, flow_counter)) {
1573 vport->ingress.drop_counter = mlx5_fc_create(dev, false);
1574 if (IS_ERR(vport->ingress.drop_counter)) {
1576 "vport[%d] configure ingress drop rule counter failed\n",
1578 vport->ingress.drop_counter = NULL;
1582 if (MLX5_CAP_ESW_EGRESS_ACL(dev, flow_counter)) {
1583 vport->egress.drop_counter = mlx5_fc_create(dev, false);
1584 if (IS_ERR(vport->egress.drop_counter)) {
1586 "vport[%d] configure egress drop rule counter failed\n",
1588 vport->egress.drop_counter = NULL;
1593 static void esw_vport_destroy_drop_counters(struct mlx5_vport *vport)
1595 struct mlx5_core_dev *dev = vport->dev;
1597 if (vport->ingress.drop_counter)
1598 mlx5_fc_destroy(dev, vport->ingress.drop_counter);
1599 if (vport->egress.drop_counter)
1600 mlx5_fc_destroy(dev, vport->egress.drop_counter);
1603 static void esw_enable_vport(struct mlx5_eswitch *esw, struct mlx5_vport *vport,
1606 u16 vport_num = vport->vport;
1608 mutex_lock(&esw->state_lock);
1609 WARN_ON(vport->enabled);
1611 esw_debug(esw->dev, "Enabling VPORT(%d)\n", vport_num);
1613 /* Create steering drop counters for ingress and egress ACLs */
1614 if (vport_num && esw->mode == SRIOV_LEGACY)
1615 esw_vport_create_drop_counters(vport);
1617 /* Restore old vport configuration */
1618 esw_apply_vport_conf(esw, vport);
1620 /* Attach vport to the eswitch rate limiter */
1621 if (esw_vport_enable_qos(esw, vport_num, vport->info.max_rate,
1622 vport->qos.bw_share))
1623 esw_warn(esw->dev, "Failed to attach vport %d to eswitch rate limiter", vport_num);
1625 /* Sync with current vport context */
1626 vport->enabled_events = enable_events;
1627 vport->enabled = true;
1629 /* Esw manager is trusted by default. Host PF (vport 0) is trusted as well
1630 * in smartNIC as it's a vport group manager.
1632 if (esw->manager_vport == vport_num ||
1633 (!vport_num && mlx5_core_is_ecpf(esw->dev)))
1634 vport->info.trusted = true;
1636 esw_vport_change_handle_locked(vport);
1638 esw->enabled_vports++;
1639 esw_debug(esw->dev, "Enabled VPORT(%d)\n", vport_num);
1640 mutex_unlock(&esw->state_lock);
1643 static void esw_disable_vport(struct mlx5_eswitch *esw,
1644 struct mlx5_vport *vport)
1646 u16 vport_num = vport->vport;
1648 if (!vport->enabled)
1651 esw_debug(esw->dev, "Disabling vport(%d)\n", vport_num);
1652 /* Mark this vport as disabled to discard new events */
1653 vport->enabled = false;
1655 /* Wait for current already scheduled events to complete */
1656 flush_workqueue(esw->work_queue);
1657 /* Disable events from this vport */
1658 arm_vport_context_events_cmd(esw->dev, vport->vport, 0);
1659 mutex_lock(&esw->state_lock);
1660 /* We don't assume VFs will cleanup after themselves.
1661 * Calling vport change handler while vport is disabled will cleanup
1662 * the vport resources.
1664 esw_vport_change_handle_locked(vport);
1665 vport->enabled_events = 0;
1666 esw_vport_disable_qos(esw, vport_num);
1667 if (esw->manager_vport != vport_num &&
1668 esw->mode == SRIOV_LEGACY) {
1669 mlx5_modify_vport_admin_state(esw->dev,
1670 MLX5_VPORT_STATE_OP_MOD_ESW_VPORT,
1672 MLX5_VPORT_ADMIN_STATE_DOWN);
1673 esw_vport_disable_egress_acl(esw, vport);
1674 esw_vport_disable_ingress_acl(esw, vport);
1675 esw_vport_destroy_drop_counters(vport);
1677 esw->enabled_vports--;
1678 mutex_unlock(&esw->state_lock);
1681 static int eswitch_vport_event(struct notifier_block *nb,
1682 unsigned long type, void *data)
1684 struct mlx5_eswitch *esw = mlx5_nb_cof(nb, struct mlx5_eswitch, nb);
1685 struct mlx5_eqe *eqe = data;
1686 struct mlx5_vport *vport;
1689 vport_num = be16_to_cpu(eqe->data.vport_change.vport_num);
1690 vport = mlx5_eswitch_get_vport(esw, vport_num);
1692 queue_work(esw->work_queue, &vport->vport_change_handler);
1697 /* Public E-Switch API */
1698 #define ESW_ALLOWED(esw) ((esw) && MLX5_ESWITCH_MANAGER((esw)->dev))
1700 int mlx5_eswitch_enable_sriov(struct mlx5_eswitch *esw, int nvfs, int mode)
1702 int vf_nvports = 0, total_nvports = 0;
1703 struct mlx5_vport *vport;
1705 int i, enabled_events;
1707 if (!ESW_ALLOWED(esw) ||
1708 !MLX5_CAP_ESW_FLOWTABLE_FDB(esw->dev, ft_support)) {
1709 esw_warn(esw->dev, "E-Switch FDB is not supported, aborting ...\n");
1713 if (!MLX5_CAP_ESW_INGRESS_ACL(esw->dev, ft_support))
1714 esw_warn(esw->dev, "E-Switch ingress ACL is not supported by FW\n");
1716 if (!MLX5_CAP_ESW_EGRESS_ACL(esw->dev, ft_support))
1717 esw_warn(esw->dev, "E-Switch engress ACL is not supported by FW\n");
1719 esw_info(esw->dev, "E-Switch enable SRIOV: nvfs(%d) mode (%d)\n", nvfs, mode);
1721 if (mode == SRIOV_OFFLOADS) {
1722 if (mlx5_core_is_ecpf_esw_manager(esw->dev)) {
1723 err = mlx5_query_host_params_num_vfs(esw->dev, &vf_nvports);
1726 total_nvports = esw->total_vports;
1729 total_nvports = nvfs + MLX5_SPECIAL_VPORTS(esw->dev);
1735 mlx5_lag_update(esw->dev);
1737 if (mode == SRIOV_LEGACY) {
1738 err = esw_create_legacy_table(esw);
1742 mlx5_reload_interface(esw->dev, MLX5_INTERFACE_PROTOCOL_ETH);
1743 mlx5_reload_interface(esw->dev, MLX5_INTERFACE_PROTOCOL_IB);
1744 err = esw_offloads_init(esw, vf_nvports, total_nvports);
1750 err = esw_create_tsar(esw);
1752 esw_warn(esw->dev, "Failed to create eswitch TSAR");
1754 /* Don't enable vport events when in SRIOV_OFFLOADS mode, since:
1755 * 1. L2 table (MPFS) is programmed by PF/VF representors netdevs set_rx_mode
1756 * 2. FDB/Eswitch is programmed by user space tools
1758 enabled_events = (mode == SRIOV_LEGACY) ? SRIOV_VPORT_EVENTS : 0;
1760 /* Enable PF vport */
1761 vport = mlx5_eswitch_get_vport(esw, MLX5_VPORT_PF);
1762 esw_enable_vport(esw, vport, enabled_events);
1764 /* Enable ECPF vports */
1765 if (mlx5_ecpf_vport_exists(esw->dev)) {
1766 vport = mlx5_eswitch_get_vport(esw, MLX5_VPORT_ECPF);
1767 esw_enable_vport(esw, vport, enabled_events);
1770 /* Enable VF vports */
1771 mlx5_esw_for_each_vf_vport(esw, i, vport, nvfs)
1772 esw_enable_vport(esw, vport, enabled_events);
1774 if (mode == SRIOV_LEGACY) {
1775 MLX5_NB_INIT(&esw->nb, eswitch_vport_event, NIC_VPORT_CHANGE);
1776 mlx5_eq_notifier_register(esw->dev, &esw->nb);
1779 esw_info(esw->dev, "SRIOV enabled: active vports(%d)\n",
1780 esw->enabled_vports);
1784 esw->mode = SRIOV_NONE;
1786 if (mode == SRIOV_OFFLOADS) {
1787 mlx5_reload_interface(esw->dev, MLX5_INTERFACE_PROTOCOL_IB);
1788 mlx5_reload_interface(esw->dev, MLX5_INTERFACE_PROTOCOL_ETH);
1794 void mlx5_eswitch_disable_sriov(struct mlx5_eswitch *esw)
1796 struct esw_mc_addr *mc_promisc;
1797 struct mlx5_vport *vport;
1801 if (!ESW_ALLOWED(esw) || esw->mode == SRIOV_NONE)
1804 esw_info(esw->dev, "disable SRIOV: active vports(%d) mode(%d)\n",
1805 esw->enabled_vports, esw->mode);
1807 mc_promisc = &esw->mc_promisc;
1809 if (esw->mode == SRIOV_LEGACY)
1810 mlx5_eq_notifier_unregister(esw->dev, &esw->nb);
1812 mlx5_esw_for_all_vports(esw, i, vport)
1813 esw_disable_vport(esw, vport);
1815 if (mc_promisc && mc_promisc->uplink_rule)
1816 mlx5_del_flow_rules(mc_promisc->uplink_rule);
1818 esw_destroy_tsar(esw);
1820 if (esw->mode == SRIOV_LEGACY)
1821 esw_destroy_legacy_table(esw);
1822 else if (esw->mode == SRIOV_OFFLOADS)
1823 esw_offloads_cleanup(esw);
1825 old_mode = esw->mode;
1826 esw->mode = SRIOV_NONE;
1828 mlx5_lag_update(esw->dev);
1830 if (old_mode == SRIOV_OFFLOADS) {
1831 mlx5_reload_interface(esw->dev, MLX5_INTERFACE_PROTOCOL_IB);
1832 mlx5_reload_interface(esw->dev, MLX5_INTERFACE_PROTOCOL_ETH);
1836 int mlx5_eswitch_init(struct mlx5_core_dev *dev)
1838 int total_vports = MLX5_TOTAL_VPORTS(dev);
1839 struct mlx5_eswitch *esw;
1840 struct mlx5_vport *vport;
1843 if (!MLX5_VPORT_MANAGER(dev))
1847 "Total vports %d, per vport: max uc(%d) max mc(%d)\n",
1849 MLX5_MAX_UC_PER_VPORT(dev),
1850 MLX5_MAX_MC_PER_VPORT(dev));
1852 esw = kzalloc(sizeof(*esw), GFP_KERNEL);
1857 esw->manager_vport = mlx5_eswitch_manager_vport(dev);
1859 esw->work_queue = create_singlethread_workqueue("mlx5_esw_wq");
1860 if (!esw->work_queue) {
1865 esw->vports = kcalloc(total_vports, sizeof(struct mlx5_vport),
1872 esw->total_vports = total_vports;
1874 err = esw_offloads_init_reps(esw);
1878 hash_init(esw->offloads.encap_tbl);
1879 hash_init(esw->offloads.mod_hdr_tbl);
1880 mutex_init(&esw->state_lock);
1882 mlx5_esw_for_all_vports(esw, i, vport) {
1883 vport->vport = mlx5_eswitch_index_to_vport_num(esw, i);
1884 vport->info.link_state = MLX5_VPORT_ADMIN_STATE_AUTO;
1886 INIT_WORK(&vport->vport_change_handler,
1887 esw_vport_change_handler);
1890 esw->enabled_vports = 0;
1891 esw->mode = SRIOV_NONE;
1892 esw->offloads.inline_mode = MLX5_INLINE_MODE_NONE;
1893 if (MLX5_CAP_ESW_FLOWTABLE_FDB(dev, reformat) &&
1894 MLX5_CAP_ESW_FLOWTABLE_FDB(dev, decap))
1895 esw->offloads.encap = DEVLINK_ESWITCH_ENCAP_MODE_BASIC;
1897 esw->offloads.encap = DEVLINK_ESWITCH_ENCAP_MODE_NONE;
1899 dev->priv.eswitch = esw;
1902 if (esw->work_queue)
1903 destroy_workqueue(esw->work_queue);
1904 esw_offloads_cleanup_reps(esw);
1910 void mlx5_eswitch_cleanup(struct mlx5_eswitch *esw)
1912 if (!esw || !MLX5_VPORT_MANAGER(esw->dev))
1915 esw_info(esw->dev, "cleanup\n");
1917 esw->dev->priv.eswitch = NULL;
1918 destroy_workqueue(esw->work_queue);
1919 esw_offloads_cleanup_reps(esw);
1924 /* Vport Administration */
1925 #define LEGAL_VPORT(esw, vport) (vport >= 0 && vport < esw->total_vports)
1927 int mlx5_eswitch_set_vport_mac(struct mlx5_eswitch *esw,
1928 int vport, u8 mac[ETH_ALEN])
1930 struct mlx5_vport *evport;
1934 if (!esw || !MLX5_CAP_GEN(esw->dev, vport_group_manager))
1936 if (!LEGAL_VPORT(esw, vport) || is_multicast_ether_addr(mac))
1939 mutex_lock(&esw->state_lock);
1940 evport = &esw->vports[vport];
1942 if (evport->info.spoofchk && !is_valid_ether_addr(mac))
1943 mlx5_core_warn(esw->dev,
1944 "Set invalid MAC while spoofchk is on, vport(%d)\n",
1947 err = mlx5_modify_nic_vport_mac_address(esw->dev, vport, mac);
1949 mlx5_core_warn(esw->dev,
1950 "Failed to mlx5_modify_nic_vport_mac vport(%d) err=(%d)\n",
1955 node_guid_gen_from_mac(&node_guid, mac);
1956 err = mlx5_modify_nic_vport_node_guid(esw->dev, vport, node_guid);
1958 mlx5_core_warn(esw->dev,
1959 "Failed to set vport %d node guid, err = %d. RDMA_CM will not function properly for this VF.\n",
1962 ether_addr_copy(evport->info.mac, mac);
1963 evport->info.node_guid = node_guid;
1964 if (evport->enabled && esw->mode == SRIOV_LEGACY)
1965 err = esw_vport_ingress_config(esw, evport);
1968 mutex_unlock(&esw->state_lock);
1972 int mlx5_eswitch_set_vport_state(struct mlx5_eswitch *esw,
1973 int vport, int link_state)
1975 struct mlx5_vport *evport;
1978 if (!ESW_ALLOWED(esw))
1980 if (!LEGAL_VPORT(esw, vport))
1983 mutex_lock(&esw->state_lock);
1984 evport = &esw->vports[vport];
1986 err = mlx5_modify_vport_admin_state(esw->dev,
1987 MLX5_VPORT_STATE_OP_MOD_ESW_VPORT,
1988 vport, 1, link_state);
1990 mlx5_core_warn(esw->dev,
1991 "Failed to set vport %d link state, err = %d",
1996 evport->info.link_state = link_state;
1999 mutex_unlock(&esw->state_lock);
2003 int mlx5_eswitch_get_vport_config(struct mlx5_eswitch *esw,
2004 int vport, struct ifla_vf_info *ivi)
2006 struct mlx5_vport *evport;
2008 if (!esw || !MLX5_CAP_GEN(esw->dev, vport_group_manager))
2010 if (!LEGAL_VPORT(esw, vport))
2013 evport = &esw->vports[vport];
2015 memset(ivi, 0, sizeof(*ivi));
2016 ivi->vf = vport - 1;
2018 mutex_lock(&esw->state_lock);
2019 ether_addr_copy(ivi->mac, evport->info.mac);
2020 ivi->linkstate = evport->info.link_state;
2021 ivi->vlan = evport->info.vlan;
2022 ivi->qos = evport->info.qos;
2023 ivi->spoofchk = evport->info.spoofchk;
2024 ivi->trusted = evport->info.trusted;
2025 ivi->min_tx_rate = evport->info.min_rate;
2026 ivi->max_tx_rate = evport->info.max_rate;
2027 mutex_unlock(&esw->state_lock);
2032 int __mlx5_eswitch_set_vport_vlan(struct mlx5_eswitch *esw,
2033 int vport, u16 vlan, u8 qos, u8 set_flags)
2035 struct mlx5_vport *evport;
2038 if (!ESW_ALLOWED(esw))
2040 if (!LEGAL_VPORT(esw, vport) || (vlan > 4095) || (qos > 7))
2043 mutex_lock(&esw->state_lock);
2044 evport = &esw->vports[vport];
2046 err = modify_esw_vport_cvlan(esw->dev, vport, vlan, qos, set_flags);
2050 evport->info.vlan = vlan;
2051 evport->info.qos = qos;
2052 if (evport->enabled && esw->mode == SRIOV_LEGACY) {
2053 err = esw_vport_ingress_config(esw, evport);
2056 err = esw_vport_egress_config(esw, evport);
2060 mutex_unlock(&esw->state_lock);
2064 int mlx5_eswitch_set_vport_vlan(struct mlx5_eswitch *esw,
2065 int vport, u16 vlan, u8 qos)
2070 set_flags = SET_VLAN_STRIP | SET_VLAN_INSERT;
2072 return __mlx5_eswitch_set_vport_vlan(esw, vport, vlan, qos, set_flags);
2075 int mlx5_eswitch_set_vport_spoofchk(struct mlx5_eswitch *esw,
2076 int vport, bool spoofchk)
2078 struct mlx5_vport *evport;
2082 if (!ESW_ALLOWED(esw))
2084 if (!LEGAL_VPORT(esw, vport))
2087 mutex_lock(&esw->state_lock);
2088 evport = &esw->vports[vport];
2089 pschk = evport->info.spoofchk;
2090 evport->info.spoofchk = spoofchk;
2091 if (pschk && !is_valid_ether_addr(evport->info.mac))
2092 mlx5_core_warn(esw->dev,
2093 "Spoofchk in set while MAC is invalid, vport(%d)\n",
2095 if (evport->enabled && esw->mode == SRIOV_LEGACY)
2096 err = esw_vport_ingress_config(esw, evport);
2098 evport->info.spoofchk = pschk;
2099 mutex_unlock(&esw->state_lock);
2104 static void esw_cleanup_vepa_rules(struct mlx5_eswitch *esw)
2106 if (esw->fdb_table.legacy.vepa_uplink_rule)
2107 mlx5_del_flow_rules(esw->fdb_table.legacy.vepa_uplink_rule);
2109 if (esw->fdb_table.legacy.vepa_star_rule)
2110 mlx5_del_flow_rules(esw->fdb_table.legacy.vepa_star_rule);
2112 esw->fdb_table.legacy.vepa_uplink_rule = NULL;
2113 esw->fdb_table.legacy.vepa_star_rule = NULL;
2116 static int _mlx5_eswitch_set_vepa_locked(struct mlx5_eswitch *esw,
2119 struct mlx5_flow_destination dest = {};
2120 struct mlx5_flow_act flow_act = {};
2121 struct mlx5_flow_handle *flow_rule;
2122 struct mlx5_flow_spec *spec;
2127 esw_cleanup_vepa_rules(esw);
2131 if (esw->fdb_table.legacy.vepa_uplink_rule)
2134 spec = kvzalloc(sizeof(*spec), GFP_KERNEL);
2138 /* Uplink rule forward uplink traffic to FDB */
2139 misc = MLX5_ADDR_OF(fte_match_param, spec->match_value, misc_parameters);
2140 MLX5_SET(fte_match_set_misc, misc, source_port, MLX5_VPORT_UPLINK);
2142 misc = MLX5_ADDR_OF(fte_match_param, spec->match_criteria, misc_parameters);
2143 MLX5_SET_TO_ONES(fte_match_set_misc, misc, source_port);
2145 spec->match_criteria_enable = MLX5_MATCH_MISC_PARAMETERS;
2146 dest.type = MLX5_FLOW_DESTINATION_TYPE_FLOW_TABLE;
2147 dest.ft = esw->fdb_table.legacy.fdb;
2148 flow_act.action = MLX5_FLOW_CONTEXT_ACTION_FWD_DEST;
2149 flow_rule = mlx5_add_flow_rules(esw->fdb_table.legacy.vepa_fdb, spec,
2150 &flow_act, &dest, 1);
2151 if (IS_ERR(flow_rule)) {
2152 err = PTR_ERR(flow_rule);
2155 esw->fdb_table.legacy.vepa_uplink_rule = flow_rule;
2158 /* Star rule to forward all traffic to uplink vport */
2159 memset(spec, 0, sizeof(*spec));
2160 memset(&dest, 0, sizeof(dest));
2161 dest.type = MLX5_FLOW_DESTINATION_TYPE_VPORT;
2162 dest.vport.num = MLX5_VPORT_UPLINK;
2163 flow_act.action = MLX5_FLOW_CONTEXT_ACTION_FWD_DEST;
2164 flow_rule = mlx5_add_flow_rules(esw->fdb_table.legacy.vepa_fdb, spec,
2165 &flow_act, &dest, 1);
2166 if (IS_ERR(flow_rule)) {
2167 err = PTR_ERR(flow_rule);
2170 esw->fdb_table.legacy.vepa_star_rule = flow_rule;
2176 esw_cleanup_vepa_rules(esw);
2180 int mlx5_eswitch_set_vepa(struct mlx5_eswitch *esw, u8 setting)
2187 if (!ESW_ALLOWED(esw))
2190 mutex_lock(&esw->state_lock);
2191 if (esw->mode != SRIOV_LEGACY) {
2196 err = _mlx5_eswitch_set_vepa_locked(esw, setting);
2199 mutex_unlock(&esw->state_lock);
2203 int mlx5_eswitch_get_vepa(struct mlx5_eswitch *esw, u8 *setting)
2210 if (!ESW_ALLOWED(esw))
2213 mutex_lock(&esw->state_lock);
2214 if (esw->mode != SRIOV_LEGACY) {
2219 *setting = esw->fdb_table.legacy.vepa_uplink_rule ? 1 : 0;
2222 mutex_unlock(&esw->state_lock);
2226 int mlx5_eswitch_set_vport_trust(struct mlx5_eswitch *esw,
2227 int vport, bool setting)
2229 struct mlx5_vport *evport;
2231 if (!ESW_ALLOWED(esw))
2233 if (!LEGAL_VPORT(esw, vport))
2236 mutex_lock(&esw->state_lock);
2237 evport = &esw->vports[vport];
2238 evport->info.trusted = setting;
2239 if (evport->enabled)
2240 esw_vport_change_handle_locked(evport);
2241 mutex_unlock(&esw->state_lock);
2246 static u32 calculate_vports_min_rate_divider(struct mlx5_eswitch *esw)
2248 u32 fw_max_bw_share = MLX5_CAP_QOS(esw->dev, max_tsar_bw_share);
2249 struct mlx5_vport *evport;
2250 u32 max_guarantee = 0;
2253 mlx5_esw_for_all_vports(esw, i, evport) {
2254 if (!evport->enabled || evport->info.min_rate < max_guarantee)
2256 max_guarantee = evport->info.min_rate;
2259 return max_t(u32, max_guarantee / fw_max_bw_share, 1);
2262 static int normalize_vports_min_rate(struct mlx5_eswitch *esw, u32 divider)
2264 u32 fw_max_bw_share = MLX5_CAP_QOS(esw->dev, max_tsar_bw_share);
2265 struct mlx5_vport *evport;
2272 mlx5_esw_for_all_vports(esw, i, evport) {
2273 if (!evport->enabled)
2275 vport_min_rate = evport->info.min_rate;
2276 vport_max_rate = evport->info.max_rate;
2277 bw_share = MLX5_MIN_BW_SHARE;
2280 bw_share = MLX5_RATE_TO_BW_SHARE(vport_min_rate,
2284 if (bw_share == evport->qos.bw_share)
2287 err = esw_vport_qos_config(esw, evport->vport, vport_max_rate,
2290 evport->qos.bw_share = bw_share;
2298 int mlx5_eswitch_set_vport_rate(struct mlx5_eswitch *esw, int vport,
2299 u32 max_rate, u32 min_rate)
2301 struct mlx5_vport *evport;
2302 u32 fw_max_bw_share;
2303 u32 previous_min_rate;
2305 bool min_rate_supported;
2306 bool max_rate_supported;
2309 if (!ESW_ALLOWED(esw))
2311 if (!LEGAL_VPORT(esw, vport))
2314 fw_max_bw_share = MLX5_CAP_QOS(esw->dev, max_tsar_bw_share);
2315 min_rate_supported = MLX5_CAP_QOS(esw->dev, esw_bw_share) &&
2316 fw_max_bw_share >= MLX5_MIN_BW_SHARE;
2317 max_rate_supported = MLX5_CAP_QOS(esw->dev, esw_rate_limit);
2319 if ((min_rate && !min_rate_supported) || (max_rate && !max_rate_supported))
2322 mutex_lock(&esw->state_lock);
2323 evport = &esw->vports[vport];
2325 if (min_rate == evport->info.min_rate)
2328 previous_min_rate = evport->info.min_rate;
2329 evport->info.min_rate = min_rate;
2330 divider = calculate_vports_min_rate_divider(esw);
2331 err = normalize_vports_min_rate(esw, divider);
2333 evport->info.min_rate = previous_min_rate;
2338 if (max_rate == evport->info.max_rate)
2341 err = esw_vport_qos_config(esw, vport, max_rate, evport->qos.bw_share);
2343 evport->info.max_rate = max_rate;
2346 mutex_unlock(&esw->state_lock);
2350 static int mlx5_eswitch_query_vport_drop_stats(struct mlx5_core_dev *dev,
2352 struct mlx5_vport_drop_stats *stats)
2354 struct mlx5_eswitch *esw = dev->priv.eswitch;
2355 struct mlx5_vport *vport = &esw->vports[vport_idx];
2356 u64 rx_discard_vport_down, tx_discard_vport_down;
2360 if (!vport->enabled || esw->mode != SRIOV_LEGACY)
2363 if (vport->egress.drop_counter)
2364 mlx5_fc_query(dev, vport->egress.drop_counter,
2365 &stats->rx_dropped, &bytes);
2367 if (vport->ingress.drop_counter)
2368 mlx5_fc_query(dev, vport->ingress.drop_counter,
2369 &stats->tx_dropped, &bytes);
2371 if (!MLX5_CAP_GEN(dev, receive_discard_vport_down) &&
2372 !MLX5_CAP_GEN(dev, transmit_discard_vport_down))
2375 err = mlx5_query_vport_down_stats(dev, vport_idx, 1,
2376 &rx_discard_vport_down,
2377 &tx_discard_vport_down);
2381 if (MLX5_CAP_GEN(dev, receive_discard_vport_down))
2382 stats->rx_dropped += rx_discard_vport_down;
2383 if (MLX5_CAP_GEN(dev, transmit_discard_vport_down))
2384 stats->tx_dropped += tx_discard_vport_down;
2389 int mlx5_eswitch_get_vport_stats(struct mlx5_eswitch *esw,
2391 struct ifla_vf_stats *vf_stats)
2393 int outlen = MLX5_ST_SZ_BYTES(query_vport_counter_out);
2394 u32 in[MLX5_ST_SZ_DW(query_vport_counter_in)] = {0};
2395 struct mlx5_vport_drop_stats stats = {0};
2399 if (!ESW_ALLOWED(esw))
2401 if (!LEGAL_VPORT(esw, vport))
2404 out = kvzalloc(outlen, GFP_KERNEL);
2408 MLX5_SET(query_vport_counter_in, in, opcode,
2409 MLX5_CMD_OP_QUERY_VPORT_COUNTER);
2410 MLX5_SET(query_vport_counter_in, in, op_mod, 0);
2411 MLX5_SET(query_vport_counter_in, in, vport_number, vport);
2412 MLX5_SET(query_vport_counter_in, in, other_vport, 1);
2414 memset(out, 0, outlen);
2415 err = mlx5_cmd_exec(esw->dev, in, sizeof(in), out, outlen);
2419 #define MLX5_GET_CTR(p, x) \
2420 MLX5_GET64(query_vport_counter_out, p, x)
2422 memset(vf_stats, 0, sizeof(*vf_stats));
2423 vf_stats->rx_packets =
2424 MLX5_GET_CTR(out, received_eth_unicast.packets) +
2425 MLX5_GET_CTR(out, received_ib_unicast.packets) +
2426 MLX5_GET_CTR(out, received_eth_multicast.packets) +
2427 MLX5_GET_CTR(out, received_ib_multicast.packets) +
2428 MLX5_GET_CTR(out, received_eth_broadcast.packets);
2430 vf_stats->rx_bytes =
2431 MLX5_GET_CTR(out, received_eth_unicast.octets) +
2432 MLX5_GET_CTR(out, received_ib_unicast.octets) +
2433 MLX5_GET_CTR(out, received_eth_multicast.octets) +
2434 MLX5_GET_CTR(out, received_ib_multicast.octets) +
2435 MLX5_GET_CTR(out, received_eth_broadcast.octets);
2437 vf_stats->tx_packets =
2438 MLX5_GET_CTR(out, transmitted_eth_unicast.packets) +
2439 MLX5_GET_CTR(out, transmitted_ib_unicast.packets) +
2440 MLX5_GET_CTR(out, transmitted_eth_multicast.packets) +
2441 MLX5_GET_CTR(out, transmitted_ib_multicast.packets) +
2442 MLX5_GET_CTR(out, transmitted_eth_broadcast.packets);
2444 vf_stats->tx_bytes =
2445 MLX5_GET_CTR(out, transmitted_eth_unicast.octets) +
2446 MLX5_GET_CTR(out, transmitted_ib_unicast.octets) +
2447 MLX5_GET_CTR(out, transmitted_eth_multicast.octets) +
2448 MLX5_GET_CTR(out, transmitted_ib_multicast.octets) +
2449 MLX5_GET_CTR(out, transmitted_eth_broadcast.octets);
2451 vf_stats->multicast =
2452 MLX5_GET_CTR(out, received_eth_multicast.packets) +
2453 MLX5_GET_CTR(out, received_ib_multicast.packets);
2455 vf_stats->broadcast =
2456 MLX5_GET_CTR(out, received_eth_broadcast.packets);
2458 err = mlx5_eswitch_query_vport_drop_stats(esw->dev, vport, &stats);
2461 vf_stats->rx_dropped = stats.rx_dropped;
2462 vf_stats->tx_dropped = stats.tx_dropped;
2469 u8 mlx5_eswitch_mode(struct mlx5_eswitch *esw)
2471 return ESW_ALLOWED(esw) ? esw->mode : SRIOV_NONE;
2473 EXPORT_SYMBOL_GPL(mlx5_eswitch_mode);
2475 bool mlx5_esw_lag_prereq(struct mlx5_core_dev *dev0, struct mlx5_core_dev *dev1)
2477 if ((dev0->priv.eswitch->mode == SRIOV_NONE &&
2478 dev1->priv.eswitch->mode == SRIOV_NONE) ||
2479 (dev0->priv.eswitch->mode == SRIOV_OFFLOADS &&
2480 dev1->priv.eswitch->mode == SRIOV_OFFLOADS))
2486 bool mlx5_esw_multipath_prereq(struct mlx5_core_dev *dev0,
2487 struct mlx5_core_dev *dev1)
2489 return (dev0->priv.eswitch->mode == SRIOV_OFFLOADS &&
2490 dev1->priv.eswitch->mode == SRIOV_OFFLOADS);