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 *__must_check
76 mlx5_eswitch_get_vport(struct mlx5_eswitch *esw, u16 vport_num)
80 if (!esw || !MLX5_CAP_GEN(esw->dev, vport_group_manager))
81 return ERR_PTR(-EPERM);
83 idx = mlx5_eswitch_vport_num_to_index(esw, vport_num);
85 if (idx > esw->total_vports - 1) {
86 esw_debug(esw->dev, "vport out of range: num(0x%x), idx(0x%x)\n",
88 return ERR_PTR(-EINVAL);
91 return &esw->vports[idx];
94 static int arm_vport_context_events_cmd(struct mlx5_core_dev *dev, u16 vport,
97 int in[MLX5_ST_SZ_DW(modify_nic_vport_context_in)] = {0};
98 int out[MLX5_ST_SZ_DW(modify_nic_vport_context_out)] = {0};
101 MLX5_SET(modify_nic_vport_context_in, in,
102 opcode, MLX5_CMD_OP_MODIFY_NIC_VPORT_CONTEXT);
103 MLX5_SET(modify_nic_vport_context_in, in, field_select.change_event, 1);
104 MLX5_SET(modify_nic_vport_context_in, in, vport_number, vport);
105 MLX5_SET(modify_nic_vport_context_in, in, other_vport, 1);
106 nic_vport_ctx = MLX5_ADDR_OF(modify_nic_vport_context_in,
107 in, nic_vport_context);
109 MLX5_SET(nic_vport_context, nic_vport_ctx, arm_change_event, 1);
111 if (events_mask & UC_ADDR_CHANGE)
112 MLX5_SET(nic_vport_context, nic_vport_ctx,
113 event_on_uc_address_change, 1);
114 if (events_mask & MC_ADDR_CHANGE)
115 MLX5_SET(nic_vport_context, nic_vport_ctx,
116 event_on_mc_address_change, 1);
117 if (events_mask & PROMISC_CHANGE)
118 MLX5_SET(nic_vport_context, nic_vport_ctx,
119 event_on_promisc_change, 1);
121 return mlx5_cmd_exec(dev, in, sizeof(in), out, sizeof(out));
124 /* E-Switch vport context HW commands */
125 static int modify_esw_vport_context_cmd(struct mlx5_core_dev *dev, u16 vport,
128 u32 out[MLX5_ST_SZ_DW(modify_esw_vport_context_out)] = {0};
130 MLX5_SET(modify_esw_vport_context_in, in, opcode,
131 MLX5_CMD_OP_MODIFY_ESW_VPORT_CONTEXT);
132 MLX5_SET(modify_esw_vport_context_in, in, vport_number, vport);
133 MLX5_SET(modify_esw_vport_context_in, in, other_vport, 1);
134 return mlx5_cmd_exec(dev, in, inlen, out, sizeof(out));
137 static int modify_esw_vport_cvlan(struct mlx5_core_dev *dev, u16 vport,
138 u16 vlan, u8 qos, u8 set_flags)
140 u32 in[MLX5_ST_SZ_DW(modify_esw_vport_context_in)] = {0};
142 if (!MLX5_CAP_ESW(dev, vport_cvlan_strip) ||
143 !MLX5_CAP_ESW(dev, vport_cvlan_insert_if_not_exist))
146 esw_debug(dev, "Set Vport[%d] VLAN %d qos %d set=%x\n",
147 vport, vlan, qos, set_flags);
149 if (set_flags & SET_VLAN_STRIP)
150 MLX5_SET(modify_esw_vport_context_in, in,
151 esw_vport_context.vport_cvlan_strip, 1);
153 if (set_flags & SET_VLAN_INSERT) {
154 /* insert only if no vlan in packet */
155 MLX5_SET(modify_esw_vport_context_in, in,
156 esw_vport_context.vport_cvlan_insert, 1);
158 MLX5_SET(modify_esw_vport_context_in, in,
159 esw_vport_context.cvlan_pcp, qos);
160 MLX5_SET(modify_esw_vport_context_in, in,
161 esw_vport_context.cvlan_id, vlan);
164 MLX5_SET(modify_esw_vport_context_in, in,
165 field_select.vport_cvlan_strip, 1);
166 MLX5_SET(modify_esw_vport_context_in, in,
167 field_select.vport_cvlan_insert, 1);
169 return modify_esw_vport_context_cmd(dev, vport, in, sizeof(in));
173 static struct mlx5_flow_handle *
174 __esw_fdb_set_vport_rule(struct mlx5_eswitch *esw, u16 vport, bool rx_rule,
175 u8 mac_c[ETH_ALEN], u8 mac_v[ETH_ALEN])
177 int match_header = (is_zero_ether_addr(mac_c) ? 0 :
178 MLX5_MATCH_OUTER_HEADERS);
179 struct mlx5_flow_handle *flow_rule = NULL;
180 struct mlx5_flow_act flow_act = {0};
181 struct mlx5_flow_destination dest = {};
182 struct mlx5_flow_spec *spec;
183 void *mv_misc = NULL;
184 void *mc_misc = NULL;
189 match_header |= MLX5_MATCH_MISC_PARAMETERS;
191 spec = kvzalloc(sizeof(*spec), GFP_KERNEL);
195 dmac_v = MLX5_ADDR_OF(fte_match_param, spec->match_value,
196 outer_headers.dmac_47_16);
197 dmac_c = MLX5_ADDR_OF(fte_match_param, spec->match_criteria,
198 outer_headers.dmac_47_16);
200 if (match_header & MLX5_MATCH_OUTER_HEADERS) {
201 ether_addr_copy(dmac_v, mac_v);
202 ether_addr_copy(dmac_c, mac_c);
205 if (match_header & MLX5_MATCH_MISC_PARAMETERS) {
206 mv_misc = MLX5_ADDR_OF(fte_match_param, spec->match_value,
208 mc_misc = MLX5_ADDR_OF(fte_match_param, spec->match_criteria,
210 MLX5_SET(fte_match_set_misc, mv_misc, source_port, MLX5_VPORT_UPLINK);
211 MLX5_SET_TO_ONES(fte_match_set_misc, mc_misc, source_port);
214 dest.type = MLX5_FLOW_DESTINATION_TYPE_VPORT;
215 dest.vport.num = vport;
218 "\tFDB add rule dmac_v(%pM) dmac_c(%pM) -> vport(%d)\n",
219 dmac_v, dmac_c, vport);
220 spec->match_criteria_enable = match_header;
221 flow_act.action = MLX5_FLOW_CONTEXT_ACTION_FWD_DEST;
223 mlx5_add_flow_rules(esw->fdb_table.legacy.fdb, spec,
224 &flow_act, &dest, 1);
225 if (IS_ERR(flow_rule)) {
227 "FDB: Failed to add flow rule: dmac_v(%pM) dmac_c(%pM) -> vport(%d), err(%ld)\n",
228 dmac_v, dmac_c, vport, PTR_ERR(flow_rule));
236 static struct mlx5_flow_handle *
237 esw_fdb_set_vport_rule(struct mlx5_eswitch *esw, u8 mac[ETH_ALEN], u16 vport)
241 eth_broadcast_addr(mac_c);
242 return __esw_fdb_set_vport_rule(esw, vport, false, mac_c, mac);
245 static struct mlx5_flow_handle *
246 esw_fdb_set_vport_allmulti_rule(struct mlx5_eswitch *esw, u16 vport)
251 eth_zero_addr(mac_c);
252 eth_zero_addr(mac_v);
255 return __esw_fdb_set_vport_rule(esw, vport, false, mac_c, mac_v);
258 static struct mlx5_flow_handle *
259 esw_fdb_set_vport_promisc_rule(struct mlx5_eswitch *esw, u16 vport)
264 eth_zero_addr(mac_c);
265 eth_zero_addr(mac_v);
266 return __esw_fdb_set_vport_rule(esw, vport, true, mac_c, mac_v);
270 LEGACY_VEPA_PRIO = 0,
274 static int esw_create_legacy_vepa_table(struct mlx5_eswitch *esw)
276 struct mlx5_core_dev *dev = esw->dev;
277 struct mlx5_flow_namespace *root_ns;
278 struct mlx5_flow_table *fdb;
281 root_ns = mlx5_get_fdb_sub_ns(dev, 0);
283 esw_warn(dev, "Failed to get FDB flow namespace\n");
287 /* num FTE 2, num FG 2 */
288 fdb = mlx5_create_auto_grouped_flow_table(root_ns, LEGACY_VEPA_PRIO,
292 esw_warn(dev, "Failed to create VEPA FDB err %d\n", err);
295 esw->fdb_table.legacy.vepa_fdb = fdb;
300 static int esw_create_legacy_fdb_table(struct mlx5_eswitch *esw)
302 int inlen = MLX5_ST_SZ_BYTES(create_flow_group_in);
303 struct mlx5_flow_table_attr ft_attr = {};
304 struct mlx5_core_dev *dev = esw->dev;
305 struct mlx5_flow_namespace *root_ns;
306 struct mlx5_flow_table *fdb;
307 struct mlx5_flow_group *g;
308 void *match_criteria;
314 esw_debug(dev, "Create FDB log_max_size(%d)\n",
315 MLX5_CAP_ESW_FLOWTABLE_FDB(dev, log_max_ft_size));
317 root_ns = mlx5_get_fdb_sub_ns(dev, 0);
319 esw_warn(dev, "Failed to get FDB flow namespace\n");
323 flow_group_in = kvzalloc(inlen, GFP_KERNEL);
327 table_size = BIT(MLX5_CAP_ESW_FLOWTABLE_FDB(dev, log_max_ft_size));
328 ft_attr.max_fte = table_size;
329 ft_attr.prio = LEGACY_FDB_PRIO;
330 fdb = mlx5_create_flow_table(root_ns, &ft_attr);
333 esw_warn(dev, "Failed to create FDB Table err %d\n", err);
336 esw->fdb_table.legacy.fdb = fdb;
338 /* Addresses group : Full match unicast/multicast addresses */
339 MLX5_SET(create_flow_group_in, flow_group_in, match_criteria_enable,
340 MLX5_MATCH_OUTER_HEADERS);
341 match_criteria = MLX5_ADDR_OF(create_flow_group_in, flow_group_in, match_criteria);
342 dmac = MLX5_ADDR_OF(fte_match_param, match_criteria, outer_headers.dmac_47_16);
343 MLX5_SET(create_flow_group_in, flow_group_in, start_flow_index, 0);
344 /* Preserve 2 entries for allmulti and promisc rules*/
345 MLX5_SET(create_flow_group_in, flow_group_in, end_flow_index, table_size - 3);
346 eth_broadcast_addr(dmac);
347 g = mlx5_create_flow_group(fdb, flow_group_in);
350 esw_warn(dev, "Failed to create flow group err(%d)\n", err);
353 esw->fdb_table.legacy.addr_grp = g;
355 /* Allmulti group : One rule that forwards any mcast traffic */
356 MLX5_SET(create_flow_group_in, flow_group_in, match_criteria_enable,
357 MLX5_MATCH_OUTER_HEADERS);
358 MLX5_SET(create_flow_group_in, flow_group_in, start_flow_index, table_size - 2);
359 MLX5_SET(create_flow_group_in, flow_group_in, end_flow_index, table_size - 2);
362 g = mlx5_create_flow_group(fdb, flow_group_in);
365 esw_warn(dev, "Failed to create allmulti flow group err(%d)\n", err);
368 esw->fdb_table.legacy.allmulti_grp = g;
370 /* Promiscuous group :
371 * One rule that forward all unmatched traffic from previous groups
374 MLX5_SET(create_flow_group_in, flow_group_in, match_criteria_enable,
375 MLX5_MATCH_MISC_PARAMETERS);
376 MLX5_SET_TO_ONES(fte_match_param, match_criteria, misc_parameters.source_port);
377 MLX5_SET(create_flow_group_in, flow_group_in, start_flow_index, table_size - 1);
378 MLX5_SET(create_flow_group_in, flow_group_in, end_flow_index, table_size - 1);
379 g = mlx5_create_flow_group(fdb, flow_group_in);
382 esw_warn(dev, "Failed to create promisc flow group err(%d)\n", err);
385 esw->fdb_table.legacy.promisc_grp = g;
389 esw_destroy_legacy_fdb_table(esw);
391 kvfree(flow_group_in);
395 static void esw_destroy_legacy_vepa_table(struct mlx5_eswitch *esw)
397 esw_debug(esw->dev, "Destroy VEPA Table\n");
398 if (!esw->fdb_table.legacy.vepa_fdb)
401 mlx5_destroy_flow_table(esw->fdb_table.legacy.vepa_fdb);
402 esw->fdb_table.legacy.vepa_fdb = NULL;
405 static void esw_destroy_legacy_fdb_table(struct mlx5_eswitch *esw)
407 esw_debug(esw->dev, "Destroy FDB Table\n");
408 if (!esw->fdb_table.legacy.fdb)
411 if (esw->fdb_table.legacy.promisc_grp)
412 mlx5_destroy_flow_group(esw->fdb_table.legacy.promisc_grp);
413 if (esw->fdb_table.legacy.allmulti_grp)
414 mlx5_destroy_flow_group(esw->fdb_table.legacy.allmulti_grp);
415 if (esw->fdb_table.legacy.addr_grp)
416 mlx5_destroy_flow_group(esw->fdb_table.legacy.addr_grp);
417 mlx5_destroy_flow_table(esw->fdb_table.legacy.fdb);
419 esw->fdb_table.legacy.fdb = NULL;
420 esw->fdb_table.legacy.addr_grp = NULL;
421 esw->fdb_table.legacy.allmulti_grp = NULL;
422 esw->fdb_table.legacy.promisc_grp = NULL;
425 static int esw_create_legacy_table(struct mlx5_eswitch *esw)
429 memset(&esw->fdb_table.legacy, 0, sizeof(struct legacy_fdb));
431 err = esw_create_legacy_vepa_table(esw);
435 err = esw_create_legacy_fdb_table(esw);
437 esw_destroy_legacy_vepa_table(esw);
442 static void esw_destroy_legacy_table(struct mlx5_eswitch *esw)
444 esw_cleanup_vepa_rules(esw);
445 esw_destroy_legacy_fdb_table(esw);
446 esw_destroy_legacy_vepa_table(esw);
449 /* E-Switch vport UC/MC lists management */
450 typedef int (*vport_addr_action)(struct mlx5_eswitch *esw,
451 struct vport_addr *vaddr);
453 static int esw_add_uc_addr(struct mlx5_eswitch *esw, struct vport_addr *vaddr)
455 u8 *mac = vaddr->node.addr;
456 u16 vport = vaddr->vport;
459 /* Skip mlx5_mpfs_add_mac for eswitch_managers,
460 * it is already done by its netdev in mlx5e_execute_l2_action
462 if (esw->manager_vport == vport)
465 err = mlx5_mpfs_add_mac(esw->dev, mac);
468 "Failed to add L2 table mac(%pM) for vport(0x%x), err(%d)\n",
475 /* SRIOV is enabled: Forward UC MAC to vport */
476 if (esw->fdb_table.legacy.fdb && esw->mode == SRIOV_LEGACY)
477 vaddr->flow_rule = esw_fdb_set_vport_rule(esw, mac, vport);
479 esw_debug(esw->dev, "\tADDED UC MAC: vport[%d] %pM fr(%p)\n",
480 vport, mac, vaddr->flow_rule);
485 static int esw_del_uc_addr(struct mlx5_eswitch *esw, struct vport_addr *vaddr)
487 u8 *mac = vaddr->node.addr;
488 u16 vport = vaddr->vport;
491 /* Skip mlx5_mpfs_del_mac for eswitch managerss,
492 * it is already done by its netdev in mlx5e_execute_l2_action
494 if (!vaddr->mpfs || esw->manager_vport == vport)
497 err = mlx5_mpfs_del_mac(esw->dev, mac);
500 "Failed to del L2 table mac(%pM) for vport(%d), err(%d)\n",
505 if (vaddr->flow_rule)
506 mlx5_del_flow_rules(vaddr->flow_rule);
507 vaddr->flow_rule = NULL;
512 static void update_allmulti_vports(struct mlx5_eswitch *esw,
513 struct vport_addr *vaddr,
514 struct esw_mc_addr *esw_mc)
516 u8 *mac = vaddr->node.addr;
517 struct mlx5_vport *vport;
520 mlx5_esw_for_all_vports(esw, i, vport) {
521 struct hlist_head *vport_hash = vport->mc_list;
522 struct vport_addr *iter_vaddr =
523 l2addr_hash_find(vport_hash,
526 vport_num = vport->vport;
527 if (IS_ERR_OR_NULL(vport->allmulti_rule) ||
528 vaddr->vport == vport_num)
530 switch (vaddr->action) {
531 case MLX5_ACTION_ADD:
534 iter_vaddr = l2addr_hash_add(vport_hash, mac,
539 "ALL-MULTI: Failed to add MAC(%pM) to vport[%d] DB\n",
543 iter_vaddr->vport = vport_num;
544 iter_vaddr->flow_rule =
545 esw_fdb_set_vport_rule(esw,
548 iter_vaddr->mc_promisc = true;
550 case MLX5_ACTION_DEL:
553 mlx5_del_flow_rules(iter_vaddr->flow_rule);
554 l2addr_hash_del(iter_vaddr);
560 static int esw_add_mc_addr(struct mlx5_eswitch *esw, struct vport_addr *vaddr)
562 struct hlist_head *hash = esw->mc_table;
563 struct esw_mc_addr *esw_mc;
564 u8 *mac = vaddr->node.addr;
565 u16 vport = vaddr->vport;
567 if (!esw->fdb_table.legacy.fdb)
570 esw_mc = l2addr_hash_find(hash, mac, struct esw_mc_addr);
574 esw_mc = l2addr_hash_add(hash, mac, struct esw_mc_addr, GFP_KERNEL);
578 esw_mc->uplink_rule = /* Forward MC MAC to Uplink */
579 esw_fdb_set_vport_rule(esw, mac, MLX5_VPORT_UPLINK);
581 /* Add this multicast mac to all the mc promiscuous vports */
582 update_allmulti_vports(esw, vaddr, esw_mc);
585 /* If the multicast mac is added as a result of mc promiscuous vport,
586 * don't increment the multicast ref count
588 if (!vaddr->mc_promisc)
591 /* Forward MC MAC to vport */
592 vaddr->flow_rule = esw_fdb_set_vport_rule(esw, mac, vport);
594 "\tADDED MC MAC: vport[%d] %pM fr(%p) refcnt(%d) uplinkfr(%p)\n",
595 vport, mac, vaddr->flow_rule,
596 esw_mc->refcnt, esw_mc->uplink_rule);
600 static int esw_del_mc_addr(struct mlx5_eswitch *esw, struct vport_addr *vaddr)
602 struct hlist_head *hash = esw->mc_table;
603 struct esw_mc_addr *esw_mc;
604 u8 *mac = vaddr->node.addr;
605 u16 vport = vaddr->vport;
607 if (!esw->fdb_table.legacy.fdb)
610 esw_mc = l2addr_hash_find(hash, mac, struct esw_mc_addr);
613 "Failed to find eswitch MC addr for MAC(%pM) vport(%d)",
618 "\tDELETE MC MAC: vport[%d] %pM fr(%p) refcnt(%d) uplinkfr(%p)\n",
619 vport, mac, vaddr->flow_rule, esw_mc->refcnt,
620 esw_mc->uplink_rule);
622 if (vaddr->flow_rule)
623 mlx5_del_flow_rules(vaddr->flow_rule);
624 vaddr->flow_rule = NULL;
626 /* If the multicast mac is added as a result of mc promiscuous vport,
627 * don't decrement the multicast ref count.
629 if (vaddr->mc_promisc || (--esw_mc->refcnt > 0))
632 /* Remove this multicast mac from all the mc promiscuous vports */
633 update_allmulti_vports(esw, vaddr, esw_mc);
635 if (esw_mc->uplink_rule)
636 mlx5_del_flow_rules(esw_mc->uplink_rule);
638 l2addr_hash_del(esw_mc);
642 /* Apply vport UC/MC list to HW l2 table and FDB table */
643 static void esw_apply_vport_addr_list(struct mlx5_eswitch *esw,
644 struct mlx5_vport *vport, int list_type)
646 bool is_uc = list_type == MLX5_NVPRT_LIST_TYPE_UC;
647 vport_addr_action vport_addr_add;
648 vport_addr_action vport_addr_del;
649 struct vport_addr *addr;
650 struct l2addr_node *node;
651 struct hlist_head *hash;
652 struct hlist_node *tmp;
655 vport_addr_add = is_uc ? esw_add_uc_addr :
657 vport_addr_del = is_uc ? esw_del_uc_addr :
660 hash = is_uc ? vport->uc_list : vport->mc_list;
661 for_each_l2hash_node(node, tmp, hash, hi) {
662 addr = container_of(node, struct vport_addr, node);
663 switch (addr->action) {
664 case MLX5_ACTION_ADD:
665 vport_addr_add(esw, addr);
666 addr->action = MLX5_ACTION_NONE;
668 case MLX5_ACTION_DEL:
669 vport_addr_del(esw, addr);
670 l2addr_hash_del(addr);
676 /* Sync vport UC/MC list from vport context */
677 static void esw_update_vport_addr_list(struct mlx5_eswitch *esw,
678 struct mlx5_vport *vport, int list_type)
680 bool is_uc = list_type == MLX5_NVPRT_LIST_TYPE_UC;
681 u8 (*mac_list)[ETH_ALEN];
682 struct l2addr_node *node;
683 struct vport_addr *addr;
684 struct hlist_head *hash;
685 struct hlist_node *tmp;
691 size = is_uc ? MLX5_MAX_UC_PER_VPORT(esw->dev) :
692 MLX5_MAX_MC_PER_VPORT(esw->dev);
694 mac_list = kcalloc(size, ETH_ALEN, GFP_KERNEL);
698 hash = is_uc ? vport->uc_list : vport->mc_list;
700 for_each_l2hash_node(node, tmp, hash, hi) {
701 addr = container_of(node, struct vport_addr, node);
702 addr->action = MLX5_ACTION_DEL;
708 err = mlx5_query_nic_vport_mac_list(esw->dev, vport->vport, list_type,
712 esw_debug(esw->dev, "vport[%d] context update %s list size (%d)\n",
713 vport->vport, is_uc ? "UC" : "MC", size);
715 for (i = 0; i < size; i++) {
716 if (is_uc && !is_valid_ether_addr(mac_list[i]))
719 if (!is_uc && !is_multicast_ether_addr(mac_list[i]))
722 addr = l2addr_hash_find(hash, mac_list[i], struct vport_addr);
724 addr->action = MLX5_ACTION_NONE;
725 /* If this mac was previously added because of allmulti
726 * promiscuous rx mode, its now converted to be original
729 if (addr->mc_promisc) {
730 struct esw_mc_addr *esw_mc =
731 l2addr_hash_find(esw->mc_table,
736 "Failed to MAC(%pM) in mcast DB\n",
741 addr->mc_promisc = false;
746 addr = l2addr_hash_add(hash, mac_list[i], struct vport_addr,
750 "Failed to add MAC(%pM) to vport[%d] DB\n",
751 mac_list[i], vport->vport);
754 addr->vport = vport->vport;
755 addr->action = MLX5_ACTION_ADD;
761 /* Sync vport UC/MC list from vport context
762 * Must be called after esw_update_vport_addr_list
764 static void esw_update_vport_mc_promisc(struct mlx5_eswitch *esw,
765 struct mlx5_vport *vport)
767 struct l2addr_node *node;
768 struct vport_addr *addr;
769 struct hlist_head *hash;
770 struct hlist_node *tmp;
773 hash = vport->mc_list;
775 for_each_l2hash_node(node, tmp, esw->mc_table, hi) {
776 u8 *mac = node->addr;
778 addr = l2addr_hash_find(hash, mac, struct vport_addr);
780 if (addr->action == MLX5_ACTION_DEL)
781 addr->action = MLX5_ACTION_NONE;
784 addr = l2addr_hash_add(hash, mac, struct vport_addr,
788 "Failed to add allmulti MAC(%pM) to vport[%d] DB\n",
792 addr->vport = vport->vport;
793 addr->action = MLX5_ACTION_ADD;
794 addr->mc_promisc = true;
798 /* Apply vport rx mode to HW FDB table */
799 static void esw_apply_vport_rx_mode(struct mlx5_eswitch *esw,
800 struct mlx5_vport *vport,
801 bool promisc, bool mc_promisc)
803 struct esw_mc_addr *allmulti_addr = &esw->mc_promisc;
805 if (IS_ERR_OR_NULL(vport->allmulti_rule) != mc_promisc)
809 vport->allmulti_rule =
810 esw_fdb_set_vport_allmulti_rule(esw, vport->vport);
811 if (!allmulti_addr->uplink_rule)
812 allmulti_addr->uplink_rule =
813 esw_fdb_set_vport_allmulti_rule(esw,
815 allmulti_addr->refcnt++;
816 } else if (vport->allmulti_rule) {
817 mlx5_del_flow_rules(vport->allmulti_rule);
818 vport->allmulti_rule = NULL;
820 if (--allmulti_addr->refcnt > 0)
823 if (allmulti_addr->uplink_rule)
824 mlx5_del_flow_rules(allmulti_addr->uplink_rule);
825 allmulti_addr->uplink_rule = NULL;
829 if (IS_ERR_OR_NULL(vport->promisc_rule) != promisc)
833 vport->promisc_rule =
834 esw_fdb_set_vport_promisc_rule(esw, vport->vport);
835 } else if (vport->promisc_rule) {
836 mlx5_del_flow_rules(vport->promisc_rule);
837 vport->promisc_rule = NULL;
841 /* Sync vport rx mode from vport context */
842 static void esw_update_vport_rx_mode(struct mlx5_eswitch *esw,
843 struct mlx5_vport *vport)
850 err = mlx5_query_nic_vport_promisc(esw->dev,
857 esw_debug(esw->dev, "vport[%d] context update rx mode promisc_all=%d, all_multi=%d\n",
858 vport->vport, promisc_all, promisc_mc);
860 if (!vport->info.trusted || !vport->enabled) {
866 esw_apply_vport_rx_mode(esw, vport, promisc_all,
867 (promisc_all || promisc_mc));
870 static void esw_vport_change_handle_locked(struct mlx5_vport *vport)
872 struct mlx5_core_dev *dev = vport->dev;
873 struct mlx5_eswitch *esw = dev->priv.eswitch;
876 mlx5_query_nic_vport_mac_address(dev, vport->vport, mac);
877 esw_debug(dev, "vport[%d] Context Changed: perm mac: %pM\n",
880 if (vport->enabled_events & UC_ADDR_CHANGE) {
881 esw_update_vport_addr_list(esw, vport, MLX5_NVPRT_LIST_TYPE_UC);
882 esw_apply_vport_addr_list(esw, vport, MLX5_NVPRT_LIST_TYPE_UC);
885 if (vport->enabled_events & MC_ADDR_CHANGE)
886 esw_update_vport_addr_list(esw, vport, MLX5_NVPRT_LIST_TYPE_MC);
888 if (vport->enabled_events & PROMISC_CHANGE) {
889 esw_update_vport_rx_mode(esw, vport);
890 if (!IS_ERR_OR_NULL(vport->allmulti_rule))
891 esw_update_vport_mc_promisc(esw, vport);
894 if (vport->enabled_events & (PROMISC_CHANGE | MC_ADDR_CHANGE))
895 esw_apply_vport_addr_list(esw, vport, MLX5_NVPRT_LIST_TYPE_MC);
897 esw_debug(esw->dev, "vport[%d] Context Changed: Done\n", vport->vport);
899 arm_vport_context_events_cmd(dev, vport->vport,
900 vport->enabled_events);
903 static void esw_vport_change_handler(struct work_struct *work)
905 struct mlx5_vport *vport =
906 container_of(work, struct mlx5_vport, vport_change_handler);
907 struct mlx5_eswitch *esw = vport->dev->priv.eswitch;
909 mutex_lock(&esw->state_lock);
910 esw_vport_change_handle_locked(vport);
911 mutex_unlock(&esw->state_lock);
914 int esw_vport_enable_egress_acl(struct mlx5_eswitch *esw,
915 struct mlx5_vport *vport)
917 int inlen = MLX5_ST_SZ_BYTES(create_flow_group_in);
918 struct mlx5_flow_group *vlan_grp = NULL;
919 struct mlx5_flow_group *drop_grp = NULL;
920 struct mlx5_core_dev *dev = esw->dev;
921 struct mlx5_flow_namespace *root_ns;
922 struct mlx5_flow_table *acl;
923 void *match_criteria;
925 /* The egress acl table contains 2 rules:
926 * 1)Allow traffic with vlan_tag=vst_vlan_id
927 * 2)Drop all other traffic.
932 if (!MLX5_CAP_ESW_EGRESS_ACL(dev, ft_support))
935 if (!IS_ERR_OR_NULL(vport->egress.acl))
938 esw_debug(dev, "Create vport[%d] egress ACL log_max_size(%d)\n",
939 vport->vport, MLX5_CAP_ESW_EGRESS_ACL(dev, log_max_ft_size));
941 root_ns = mlx5_get_flow_vport_acl_namespace(dev, MLX5_FLOW_NAMESPACE_ESW_EGRESS,
944 esw_warn(dev, "Failed to get E-Switch egress flow namespace for vport (%d)\n", vport->vport);
948 flow_group_in = kvzalloc(inlen, GFP_KERNEL);
952 acl = mlx5_create_vport_flow_table(root_ns, 0, table_size, 0, vport->vport);
955 esw_warn(dev, "Failed to create E-Switch vport[%d] egress flow Table, err(%d)\n",
960 MLX5_SET(create_flow_group_in, flow_group_in, match_criteria_enable, MLX5_MATCH_OUTER_HEADERS);
961 match_criteria = MLX5_ADDR_OF(create_flow_group_in, flow_group_in, match_criteria);
962 MLX5_SET_TO_ONES(fte_match_param, match_criteria, outer_headers.cvlan_tag);
963 MLX5_SET_TO_ONES(fte_match_param, match_criteria, outer_headers.first_vid);
964 MLX5_SET(create_flow_group_in, flow_group_in, start_flow_index, 0);
965 MLX5_SET(create_flow_group_in, flow_group_in, end_flow_index, 0);
967 vlan_grp = mlx5_create_flow_group(acl, flow_group_in);
968 if (IS_ERR(vlan_grp)) {
969 err = PTR_ERR(vlan_grp);
970 esw_warn(dev, "Failed to create E-Switch vport[%d] egress allowed vlans flow group, err(%d)\n",
975 memset(flow_group_in, 0, inlen);
976 MLX5_SET(create_flow_group_in, flow_group_in, start_flow_index, 1);
977 MLX5_SET(create_flow_group_in, flow_group_in, end_flow_index, 1);
978 drop_grp = mlx5_create_flow_group(acl, flow_group_in);
979 if (IS_ERR(drop_grp)) {
980 err = PTR_ERR(drop_grp);
981 esw_warn(dev, "Failed to create E-Switch vport[%d] egress drop flow group, err(%d)\n",
986 vport->egress.acl = acl;
987 vport->egress.drop_grp = drop_grp;
988 vport->egress.allowed_vlans_grp = vlan_grp;
990 kvfree(flow_group_in);
991 if (err && !IS_ERR_OR_NULL(vlan_grp))
992 mlx5_destroy_flow_group(vlan_grp);
993 if (err && !IS_ERR_OR_NULL(acl))
994 mlx5_destroy_flow_table(acl);
998 void esw_vport_cleanup_egress_rules(struct mlx5_eswitch *esw,
999 struct mlx5_vport *vport)
1001 if (!IS_ERR_OR_NULL(vport->egress.allowed_vlan))
1002 mlx5_del_flow_rules(vport->egress.allowed_vlan);
1004 if (!IS_ERR_OR_NULL(vport->egress.drop_rule))
1005 mlx5_del_flow_rules(vport->egress.drop_rule);
1007 vport->egress.allowed_vlan = NULL;
1008 vport->egress.drop_rule = NULL;
1011 void esw_vport_disable_egress_acl(struct mlx5_eswitch *esw,
1012 struct mlx5_vport *vport)
1014 if (IS_ERR_OR_NULL(vport->egress.acl))
1017 esw_debug(esw->dev, "Destroy vport[%d] E-Switch egress ACL\n", vport->vport);
1019 esw_vport_cleanup_egress_rules(esw, vport);
1020 mlx5_destroy_flow_group(vport->egress.allowed_vlans_grp);
1021 mlx5_destroy_flow_group(vport->egress.drop_grp);
1022 mlx5_destroy_flow_table(vport->egress.acl);
1023 vport->egress.allowed_vlans_grp = NULL;
1024 vport->egress.drop_grp = NULL;
1025 vport->egress.acl = NULL;
1028 int esw_vport_enable_ingress_acl(struct mlx5_eswitch *esw,
1029 struct mlx5_vport *vport)
1031 int inlen = MLX5_ST_SZ_BYTES(create_flow_group_in);
1032 struct mlx5_core_dev *dev = esw->dev;
1033 struct mlx5_flow_namespace *root_ns;
1034 struct mlx5_flow_table *acl;
1035 struct mlx5_flow_group *g;
1036 void *match_criteria;
1038 /* The ingress acl table contains 4 groups
1039 * (2 active rules at the same time -
1040 * 1 allow rule from one of the first 3 groups.
1041 * 1 drop rule from the last group):
1042 * 1)Allow untagged traffic with smac=original mac.
1043 * 2)Allow untagged traffic.
1044 * 3)Allow traffic with smac=original mac.
1045 * 4)Drop all other traffic.
1050 if (!MLX5_CAP_ESW_INGRESS_ACL(dev, ft_support))
1053 if (!IS_ERR_OR_NULL(vport->ingress.acl))
1056 esw_debug(dev, "Create vport[%d] ingress ACL log_max_size(%d)\n",
1057 vport->vport, MLX5_CAP_ESW_INGRESS_ACL(dev, log_max_ft_size));
1059 root_ns = mlx5_get_flow_vport_acl_namespace(dev, MLX5_FLOW_NAMESPACE_ESW_INGRESS,
1062 esw_warn(dev, "Failed to get E-Switch ingress flow namespace for vport (%d)\n", vport->vport);
1066 flow_group_in = kvzalloc(inlen, GFP_KERNEL);
1070 acl = mlx5_create_vport_flow_table(root_ns, 0, table_size, 0, vport->vport);
1073 esw_warn(dev, "Failed to create E-Switch vport[%d] ingress flow Table, err(%d)\n",
1077 vport->ingress.acl = acl;
1079 match_criteria = MLX5_ADDR_OF(create_flow_group_in, flow_group_in, match_criteria);
1081 MLX5_SET(create_flow_group_in, flow_group_in, match_criteria_enable, MLX5_MATCH_OUTER_HEADERS);
1082 MLX5_SET_TO_ONES(fte_match_param, match_criteria, outer_headers.cvlan_tag);
1083 MLX5_SET_TO_ONES(fte_match_param, match_criteria, outer_headers.smac_47_16);
1084 MLX5_SET_TO_ONES(fte_match_param, match_criteria, outer_headers.smac_15_0);
1085 MLX5_SET(create_flow_group_in, flow_group_in, start_flow_index, 0);
1086 MLX5_SET(create_flow_group_in, flow_group_in, end_flow_index, 0);
1088 g = mlx5_create_flow_group(acl, flow_group_in);
1091 esw_warn(dev, "Failed to create E-Switch vport[%d] ingress untagged spoofchk flow group, err(%d)\n",
1095 vport->ingress.allow_untagged_spoofchk_grp = g;
1097 memset(flow_group_in, 0, inlen);
1098 MLX5_SET(create_flow_group_in, flow_group_in, match_criteria_enable, MLX5_MATCH_OUTER_HEADERS);
1099 MLX5_SET_TO_ONES(fte_match_param, match_criteria, outer_headers.cvlan_tag);
1100 MLX5_SET(create_flow_group_in, flow_group_in, start_flow_index, 1);
1101 MLX5_SET(create_flow_group_in, flow_group_in, end_flow_index, 1);
1103 g = mlx5_create_flow_group(acl, flow_group_in);
1106 esw_warn(dev, "Failed to create E-Switch vport[%d] ingress untagged flow group, err(%d)\n",
1110 vport->ingress.allow_untagged_only_grp = g;
1112 memset(flow_group_in, 0, inlen);
1113 MLX5_SET(create_flow_group_in, flow_group_in, match_criteria_enable, MLX5_MATCH_OUTER_HEADERS);
1114 MLX5_SET_TO_ONES(fte_match_param, match_criteria, outer_headers.smac_47_16);
1115 MLX5_SET_TO_ONES(fte_match_param, match_criteria, outer_headers.smac_15_0);
1116 MLX5_SET(create_flow_group_in, flow_group_in, start_flow_index, 2);
1117 MLX5_SET(create_flow_group_in, flow_group_in, end_flow_index, 2);
1119 g = mlx5_create_flow_group(acl, flow_group_in);
1122 esw_warn(dev, "Failed to create E-Switch vport[%d] ingress spoofchk flow group, err(%d)\n",
1126 vport->ingress.allow_spoofchk_only_grp = g;
1128 memset(flow_group_in, 0, inlen);
1129 MLX5_SET(create_flow_group_in, flow_group_in, start_flow_index, 3);
1130 MLX5_SET(create_flow_group_in, flow_group_in, end_flow_index, 3);
1132 g = mlx5_create_flow_group(acl, flow_group_in);
1135 esw_warn(dev, "Failed to create E-Switch vport[%d] ingress drop flow group, err(%d)\n",
1139 vport->ingress.drop_grp = g;
1143 if (!IS_ERR_OR_NULL(vport->ingress.allow_spoofchk_only_grp))
1144 mlx5_destroy_flow_group(
1145 vport->ingress.allow_spoofchk_only_grp);
1146 if (!IS_ERR_OR_NULL(vport->ingress.allow_untagged_only_grp))
1147 mlx5_destroy_flow_group(
1148 vport->ingress.allow_untagged_only_grp);
1149 if (!IS_ERR_OR_NULL(vport->ingress.allow_untagged_spoofchk_grp))
1150 mlx5_destroy_flow_group(
1151 vport->ingress.allow_untagged_spoofchk_grp);
1152 if (!IS_ERR_OR_NULL(vport->ingress.acl))
1153 mlx5_destroy_flow_table(vport->ingress.acl);
1156 kvfree(flow_group_in);
1160 void esw_vport_cleanup_ingress_rules(struct mlx5_eswitch *esw,
1161 struct mlx5_vport *vport)
1163 if (!IS_ERR_OR_NULL(vport->ingress.drop_rule))
1164 mlx5_del_flow_rules(vport->ingress.drop_rule);
1166 if (!IS_ERR_OR_NULL(vport->ingress.allow_rule))
1167 mlx5_del_flow_rules(vport->ingress.allow_rule);
1169 vport->ingress.drop_rule = NULL;
1170 vport->ingress.allow_rule = NULL;
1173 void esw_vport_disable_ingress_acl(struct mlx5_eswitch *esw,
1174 struct mlx5_vport *vport)
1176 if (IS_ERR_OR_NULL(vport->ingress.acl))
1179 esw_debug(esw->dev, "Destroy vport[%d] E-Switch ingress ACL\n", vport->vport);
1181 esw_vport_cleanup_ingress_rules(esw, vport);
1182 mlx5_destroy_flow_group(vport->ingress.allow_spoofchk_only_grp);
1183 mlx5_destroy_flow_group(vport->ingress.allow_untagged_only_grp);
1184 mlx5_destroy_flow_group(vport->ingress.allow_untagged_spoofchk_grp);
1185 mlx5_destroy_flow_group(vport->ingress.drop_grp);
1186 mlx5_destroy_flow_table(vport->ingress.acl);
1187 vport->ingress.acl = NULL;
1188 vport->ingress.drop_grp = NULL;
1189 vport->ingress.allow_spoofchk_only_grp = NULL;
1190 vport->ingress.allow_untagged_only_grp = NULL;
1191 vport->ingress.allow_untagged_spoofchk_grp = NULL;
1194 static int esw_vport_ingress_config(struct mlx5_eswitch *esw,
1195 struct mlx5_vport *vport)
1197 struct mlx5_fc *counter = vport->ingress.drop_counter;
1198 struct mlx5_flow_destination drop_ctr_dst = {0};
1199 struct mlx5_flow_destination *dst = NULL;
1200 struct mlx5_flow_act flow_act = {0};
1201 struct mlx5_flow_spec *spec;
1206 esw_vport_cleanup_ingress_rules(esw, vport);
1208 if (!vport->info.vlan && !vport->info.qos && !vport->info.spoofchk) {
1209 esw_vport_disable_ingress_acl(esw, vport);
1213 err = esw_vport_enable_ingress_acl(esw, vport);
1215 mlx5_core_warn(esw->dev,
1216 "failed to enable ingress acl (%d) on vport[%d]\n",
1222 "vport[%d] configure ingress rules, vlan(%d) qos(%d)\n",
1223 vport->vport, vport->info.vlan, vport->info.qos);
1225 spec = kvzalloc(sizeof(*spec), GFP_KERNEL);
1231 if (vport->info.vlan || vport->info.qos)
1232 MLX5_SET_TO_ONES(fte_match_param, spec->match_criteria, outer_headers.cvlan_tag);
1234 if (vport->info.spoofchk) {
1235 MLX5_SET_TO_ONES(fte_match_param, spec->match_criteria, outer_headers.smac_47_16);
1236 MLX5_SET_TO_ONES(fte_match_param, spec->match_criteria, outer_headers.smac_15_0);
1237 smac_v = MLX5_ADDR_OF(fte_match_param,
1239 outer_headers.smac_47_16);
1240 ether_addr_copy(smac_v, vport->info.mac);
1243 spec->match_criteria_enable = MLX5_MATCH_OUTER_HEADERS;
1244 flow_act.action = MLX5_FLOW_CONTEXT_ACTION_ALLOW;
1245 vport->ingress.allow_rule =
1246 mlx5_add_flow_rules(vport->ingress.acl, spec,
1247 &flow_act, NULL, 0);
1248 if (IS_ERR(vport->ingress.allow_rule)) {
1249 err = PTR_ERR(vport->ingress.allow_rule);
1251 "vport[%d] configure ingress allow rule, err(%d)\n",
1253 vport->ingress.allow_rule = NULL;
1257 memset(spec, 0, sizeof(*spec));
1258 flow_act.action = MLX5_FLOW_CONTEXT_ACTION_DROP;
1260 /* Attach drop flow counter */
1262 flow_act.action |= MLX5_FLOW_CONTEXT_ACTION_COUNT;
1263 drop_ctr_dst.type = MLX5_FLOW_DESTINATION_TYPE_COUNTER;
1264 drop_ctr_dst.counter_id = mlx5_fc_id(counter);
1265 dst = &drop_ctr_dst;
1268 vport->ingress.drop_rule =
1269 mlx5_add_flow_rules(vport->ingress.acl, spec,
1270 &flow_act, dst, dest_num);
1271 if (IS_ERR(vport->ingress.drop_rule)) {
1272 err = PTR_ERR(vport->ingress.drop_rule);
1274 "vport[%d] configure ingress drop rule, err(%d)\n",
1276 vport->ingress.drop_rule = NULL;
1282 esw_vport_cleanup_ingress_rules(esw, vport);
1287 static int esw_vport_egress_config(struct mlx5_eswitch *esw,
1288 struct mlx5_vport *vport)
1290 struct mlx5_fc *counter = vport->egress.drop_counter;
1291 struct mlx5_flow_destination drop_ctr_dst = {0};
1292 struct mlx5_flow_destination *dst = NULL;
1293 struct mlx5_flow_act flow_act = {0};
1294 struct mlx5_flow_spec *spec;
1298 esw_vport_cleanup_egress_rules(esw, vport);
1300 if (!vport->info.vlan && !vport->info.qos) {
1301 esw_vport_disable_egress_acl(esw, vport);
1305 err = esw_vport_enable_egress_acl(esw, vport);
1307 mlx5_core_warn(esw->dev,
1308 "failed to enable egress acl (%d) on vport[%d]\n",
1314 "vport[%d] configure egress rules, vlan(%d) qos(%d)\n",
1315 vport->vport, vport->info.vlan, vport->info.qos);
1317 spec = kvzalloc(sizeof(*spec), GFP_KERNEL);
1323 /* Allowed vlan rule */
1324 MLX5_SET_TO_ONES(fte_match_param, spec->match_criteria, outer_headers.cvlan_tag);
1325 MLX5_SET_TO_ONES(fte_match_param, spec->match_value, outer_headers.cvlan_tag);
1326 MLX5_SET_TO_ONES(fte_match_param, spec->match_criteria, outer_headers.first_vid);
1327 MLX5_SET(fte_match_param, spec->match_value, outer_headers.first_vid, vport->info.vlan);
1329 spec->match_criteria_enable = MLX5_MATCH_OUTER_HEADERS;
1330 flow_act.action = MLX5_FLOW_CONTEXT_ACTION_ALLOW;
1331 vport->egress.allowed_vlan =
1332 mlx5_add_flow_rules(vport->egress.acl, spec,
1333 &flow_act, NULL, 0);
1334 if (IS_ERR(vport->egress.allowed_vlan)) {
1335 err = PTR_ERR(vport->egress.allowed_vlan);
1337 "vport[%d] configure egress allowed vlan rule failed, err(%d)\n",
1339 vport->egress.allowed_vlan = NULL;
1343 /* Drop others rule (star rule) */
1344 memset(spec, 0, sizeof(*spec));
1345 flow_act.action = MLX5_FLOW_CONTEXT_ACTION_DROP;
1347 /* Attach egress drop flow counter */
1349 flow_act.action |= MLX5_FLOW_CONTEXT_ACTION_COUNT;
1350 drop_ctr_dst.type = MLX5_FLOW_DESTINATION_TYPE_COUNTER;
1351 drop_ctr_dst.counter_id = mlx5_fc_id(counter);
1352 dst = &drop_ctr_dst;
1355 vport->egress.drop_rule =
1356 mlx5_add_flow_rules(vport->egress.acl, spec,
1357 &flow_act, dst, dest_num);
1358 if (IS_ERR(vport->egress.drop_rule)) {
1359 err = PTR_ERR(vport->egress.drop_rule);
1361 "vport[%d] configure egress drop rule failed, err(%d)\n",
1363 vport->egress.drop_rule = NULL;
1370 /* Vport QoS management */
1371 static int esw_create_tsar(struct mlx5_eswitch *esw)
1373 u32 tsar_ctx[MLX5_ST_SZ_DW(scheduling_context)] = {0};
1374 struct mlx5_core_dev *dev = esw->dev;
1377 if (!MLX5_CAP_GEN(dev, qos) || !MLX5_CAP_QOS(dev, esw_scheduling))
1380 if (esw->qos.enabled)
1383 err = mlx5_create_scheduling_element_cmd(dev,
1384 SCHEDULING_HIERARCHY_E_SWITCH,
1386 &esw->qos.root_tsar_id);
1388 esw_warn(esw->dev, "E-Switch create TSAR failed (%d)\n", err);
1392 esw->qos.enabled = true;
1396 static void esw_destroy_tsar(struct mlx5_eswitch *esw)
1400 if (!esw->qos.enabled)
1403 err = mlx5_destroy_scheduling_element_cmd(esw->dev,
1404 SCHEDULING_HIERARCHY_E_SWITCH,
1405 esw->qos.root_tsar_id);
1407 esw_warn(esw->dev, "E-Switch destroy TSAR failed (%d)\n", err);
1409 esw->qos.enabled = false;
1412 static int esw_vport_enable_qos(struct mlx5_eswitch *esw,
1413 struct mlx5_vport *vport,
1414 u32 initial_max_rate, u32 initial_bw_share)
1416 u32 sched_ctx[MLX5_ST_SZ_DW(scheduling_context)] = {0};
1417 struct mlx5_core_dev *dev = esw->dev;
1421 if (!esw->qos.enabled || !MLX5_CAP_GEN(dev, qos) ||
1422 !MLX5_CAP_QOS(dev, esw_scheduling))
1425 if (vport->qos.enabled)
1428 MLX5_SET(scheduling_context, sched_ctx, element_type,
1429 SCHEDULING_CONTEXT_ELEMENT_TYPE_VPORT);
1430 vport_elem = MLX5_ADDR_OF(scheduling_context, sched_ctx,
1431 element_attributes);
1432 MLX5_SET(vport_element, vport_elem, vport_number, vport->vport);
1433 MLX5_SET(scheduling_context, sched_ctx, parent_element_id,
1434 esw->qos.root_tsar_id);
1435 MLX5_SET(scheduling_context, sched_ctx, max_average_bw,
1437 MLX5_SET(scheduling_context, sched_ctx, bw_share, initial_bw_share);
1439 err = mlx5_create_scheduling_element_cmd(dev,
1440 SCHEDULING_HIERARCHY_E_SWITCH,
1442 &vport->qos.esw_tsar_ix);
1444 esw_warn(esw->dev, "E-Switch create TSAR vport element failed (vport=%d,err=%d)\n",
1449 vport->qos.enabled = true;
1453 static void esw_vport_disable_qos(struct mlx5_eswitch *esw,
1454 struct mlx5_vport *vport)
1458 if (!vport->qos.enabled)
1461 err = mlx5_destroy_scheduling_element_cmd(esw->dev,
1462 SCHEDULING_HIERARCHY_E_SWITCH,
1463 vport->qos.esw_tsar_ix);
1465 esw_warn(esw->dev, "E-Switch destroy TSAR vport element failed (vport=%d,err=%d)\n",
1468 vport->qos.enabled = false;
1471 static int esw_vport_qos_config(struct mlx5_eswitch *esw,
1472 struct mlx5_vport *vport,
1473 u32 max_rate, u32 bw_share)
1475 u32 sched_ctx[MLX5_ST_SZ_DW(scheduling_context)] = {0};
1476 struct mlx5_core_dev *dev = esw->dev;
1481 if (!MLX5_CAP_GEN(dev, qos) || !MLX5_CAP_QOS(dev, esw_scheduling))
1484 if (!vport->qos.enabled)
1487 MLX5_SET(scheduling_context, sched_ctx, element_type,
1488 SCHEDULING_CONTEXT_ELEMENT_TYPE_VPORT);
1489 vport_elem = MLX5_ADDR_OF(scheduling_context, sched_ctx,
1490 element_attributes);
1491 MLX5_SET(vport_element, vport_elem, vport_number, vport->vport);
1492 MLX5_SET(scheduling_context, sched_ctx, parent_element_id,
1493 esw->qos.root_tsar_id);
1494 MLX5_SET(scheduling_context, sched_ctx, max_average_bw,
1496 MLX5_SET(scheduling_context, sched_ctx, bw_share, bw_share);
1497 bitmask |= MODIFY_SCHEDULING_ELEMENT_IN_MODIFY_BITMASK_MAX_AVERAGE_BW;
1498 bitmask |= MODIFY_SCHEDULING_ELEMENT_IN_MODIFY_BITMASK_BW_SHARE;
1500 err = mlx5_modify_scheduling_element_cmd(dev,
1501 SCHEDULING_HIERARCHY_E_SWITCH,
1503 vport->qos.esw_tsar_ix,
1506 esw_warn(esw->dev, "E-Switch modify TSAR vport element failed (vport=%d,err=%d)\n",
1514 static void node_guid_gen_from_mac(u64 *node_guid, u8 mac[ETH_ALEN])
1516 ((u8 *)node_guid)[7] = mac[0];
1517 ((u8 *)node_guid)[6] = mac[1];
1518 ((u8 *)node_guid)[5] = mac[2];
1519 ((u8 *)node_guid)[4] = 0xff;
1520 ((u8 *)node_guid)[3] = 0xfe;
1521 ((u8 *)node_guid)[2] = mac[3];
1522 ((u8 *)node_guid)[1] = mac[4];
1523 ((u8 *)node_guid)[0] = mac[5];
1526 static void esw_apply_vport_conf(struct mlx5_eswitch *esw,
1527 struct mlx5_vport *vport)
1529 u16 vport_num = vport->vport;
1531 if (esw->manager_vport == vport_num)
1534 mlx5_modify_vport_admin_state(esw->dev,
1535 MLX5_VPORT_STATE_OP_MOD_ESW_VPORT,
1537 vport->info.link_state);
1539 /* Host PF has its own mac/guid. */
1541 mlx5_modify_nic_vport_mac_address(esw->dev, vport_num,
1543 mlx5_modify_nic_vport_node_guid(esw->dev, vport_num,
1544 vport->info.node_guid);
1547 modify_esw_vport_cvlan(esw->dev, vport_num, vport->info.vlan, vport->info.qos,
1548 (vport->info.vlan || vport->info.qos));
1550 /* Only legacy mode needs ACLs */
1551 if (esw->mode == SRIOV_LEGACY) {
1552 esw_vport_ingress_config(esw, vport);
1553 esw_vport_egress_config(esw, vport);
1557 static void esw_vport_create_drop_counters(struct mlx5_vport *vport)
1559 struct mlx5_core_dev *dev = vport->dev;
1561 if (MLX5_CAP_ESW_INGRESS_ACL(dev, flow_counter)) {
1562 vport->ingress.drop_counter = mlx5_fc_create(dev, false);
1563 if (IS_ERR(vport->ingress.drop_counter)) {
1565 "vport[%d] configure ingress drop rule counter failed\n",
1567 vport->ingress.drop_counter = NULL;
1571 if (MLX5_CAP_ESW_EGRESS_ACL(dev, flow_counter)) {
1572 vport->egress.drop_counter = mlx5_fc_create(dev, false);
1573 if (IS_ERR(vport->egress.drop_counter)) {
1575 "vport[%d] configure egress drop rule counter failed\n",
1577 vport->egress.drop_counter = NULL;
1582 static void esw_vport_destroy_drop_counters(struct mlx5_vport *vport)
1584 struct mlx5_core_dev *dev = vport->dev;
1586 if (vport->ingress.drop_counter)
1587 mlx5_fc_destroy(dev, vport->ingress.drop_counter);
1588 if (vport->egress.drop_counter)
1589 mlx5_fc_destroy(dev, vport->egress.drop_counter);
1592 static void esw_enable_vport(struct mlx5_eswitch *esw, struct mlx5_vport *vport,
1595 u16 vport_num = vport->vport;
1597 mutex_lock(&esw->state_lock);
1598 WARN_ON(vport->enabled);
1600 esw_debug(esw->dev, "Enabling VPORT(%d)\n", vport_num);
1602 /* Create steering drop counters for ingress and egress ACLs */
1603 if (vport_num && esw->mode == SRIOV_LEGACY)
1604 esw_vport_create_drop_counters(vport);
1606 /* Restore old vport configuration */
1607 esw_apply_vport_conf(esw, vport);
1609 /* Attach vport to the eswitch rate limiter */
1610 if (esw_vport_enable_qos(esw, vport, vport->info.max_rate,
1611 vport->qos.bw_share))
1612 esw_warn(esw->dev, "Failed to attach vport %d to eswitch rate limiter", vport_num);
1614 /* Sync with current vport context */
1615 vport->enabled_events = enable_events;
1616 vport->enabled = true;
1618 /* Esw manager is trusted by default. Host PF (vport 0) is trusted as well
1619 * in smartNIC as it's a vport group manager.
1621 if (esw->manager_vport == vport_num ||
1622 (!vport_num && mlx5_core_is_ecpf(esw->dev)))
1623 vport->info.trusted = true;
1625 esw_vport_change_handle_locked(vport);
1627 esw->enabled_vports++;
1628 esw_debug(esw->dev, "Enabled VPORT(%d)\n", vport_num);
1629 mutex_unlock(&esw->state_lock);
1632 static void esw_disable_vport(struct mlx5_eswitch *esw,
1633 struct mlx5_vport *vport)
1635 u16 vport_num = vport->vport;
1637 if (!vport->enabled)
1640 esw_debug(esw->dev, "Disabling vport(%d)\n", vport_num);
1641 /* Mark this vport as disabled to discard new events */
1642 vport->enabled = false;
1644 /* Wait for current already scheduled events to complete */
1645 flush_workqueue(esw->work_queue);
1646 /* Disable events from this vport */
1647 arm_vport_context_events_cmd(esw->dev, vport->vport, 0);
1648 mutex_lock(&esw->state_lock);
1649 /* We don't assume VFs will cleanup after themselves.
1650 * Calling vport change handler while vport is disabled will cleanup
1651 * the vport resources.
1653 esw_vport_change_handle_locked(vport);
1654 vport->enabled_events = 0;
1655 esw_vport_disable_qos(esw, vport);
1656 if (esw->manager_vport != vport_num &&
1657 esw->mode == SRIOV_LEGACY) {
1658 mlx5_modify_vport_admin_state(esw->dev,
1659 MLX5_VPORT_STATE_OP_MOD_ESW_VPORT,
1661 MLX5_VPORT_ADMIN_STATE_DOWN);
1662 esw_vport_disable_egress_acl(esw, vport);
1663 esw_vport_disable_ingress_acl(esw, vport);
1664 esw_vport_destroy_drop_counters(vport);
1666 esw->enabled_vports--;
1667 mutex_unlock(&esw->state_lock);
1670 static int eswitch_vport_event(struct notifier_block *nb,
1671 unsigned long type, void *data)
1673 struct mlx5_eswitch *esw = mlx5_nb_cof(nb, struct mlx5_eswitch, nb);
1674 struct mlx5_eqe *eqe = data;
1675 struct mlx5_vport *vport;
1678 vport_num = be16_to_cpu(eqe->data.vport_change.vport_num);
1679 vport = mlx5_eswitch_get_vport(esw, vport_num);
1684 queue_work(esw->work_queue, &vport->vport_change_handler);
1689 static int query_esw_functions(struct mlx5_core_dev *dev,
1690 u32 *out, int outlen)
1692 u32 in[MLX5_ST_SZ_DW(query_esw_functions_in)] = {0};
1694 MLX5_SET(query_esw_functions_in, in, opcode,
1695 MLX5_CMD_OP_QUERY_ESW_FUNCTIONS);
1697 return mlx5_cmd_exec(dev, in, sizeof(in), out, outlen);
1700 int mlx5_esw_query_functions(struct mlx5_core_dev *dev, u16 *num_vfs)
1702 u32 out[MLX5_ST_SZ_DW(query_esw_functions_out)] = {0};
1705 err = query_esw_functions(dev, out, sizeof(out));
1709 *num_vfs = MLX5_GET(query_esw_functions_out, out,
1710 host_params_context.host_num_of_vfs);
1711 esw_debug(dev, "host_num_of_vfs=%d\n", *num_vfs);
1716 /* Public E-Switch API */
1717 #define ESW_ALLOWED(esw) ((esw) && MLX5_ESWITCH_MANAGER((esw)->dev))
1719 int mlx5_eswitch_enable_sriov(struct mlx5_eswitch *esw, int nvfs, int mode)
1721 struct mlx5_vport *vport;
1722 int total_nvports = 0;
1725 int i, enabled_events;
1727 if (!ESW_ALLOWED(esw) ||
1728 !MLX5_CAP_ESW_FLOWTABLE_FDB(esw->dev, ft_support)) {
1729 esw_warn(esw->dev, "E-Switch FDB is not supported, aborting ...\n");
1733 if (!MLX5_CAP_ESW_INGRESS_ACL(esw->dev, ft_support))
1734 esw_warn(esw->dev, "E-Switch ingress ACL is not supported by FW\n");
1736 if (!MLX5_CAP_ESW_EGRESS_ACL(esw->dev, ft_support))
1737 esw_warn(esw->dev, "E-Switch engress ACL is not supported by FW\n");
1739 esw_info(esw->dev, "E-Switch enable SRIOV: nvfs(%d) mode (%d)\n", nvfs, mode);
1741 if (mode == SRIOV_OFFLOADS) {
1742 if (mlx5_core_is_ecpf_esw_manager(esw->dev)) {
1743 err = mlx5_esw_query_functions(esw->dev, &vf_nvports);
1746 total_nvports = esw->total_vports;
1749 total_nvports = nvfs + MLX5_SPECIAL_VPORTS(esw->dev);
1755 mlx5_lag_update(esw->dev);
1757 if (mode == SRIOV_LEGACY) {
1758 err = esw_create_legacy_table(esw);
1762 mlx5_reload_interface(esw->dev, MLX5_INTERFACE_PROTOCOL_ETH);
1763 mlx5_reload_interface(esw->dev, MLX5_INTERFACE_PROTOCOL_IB);
1764 err = esw_offloads_init(esw, vf_nvports, total_nvports);
1770 err = esw_create_tsar(esw);
1772 esw_warn(esw->dev, "Failed to create eswitch TSAR");
1774 /* Don't enable vport events when in SRIOV_OFFLOADS mode, since:
1775 * 1. L2 table (MPFS) is programmed by PF/VF representors netdevs set_rx_mode
1776 * 2. FDB/Eswitch is programmed by user space tools
1778 enabled_events = (mode == SRIOV_LEGACY) ? SRIOV_VPORT_EVENTS : 0;
1780 /* Enable PF vport */
1781 vport = mlx5_eswitch_get_vport(esw, MLX5_VPORT_PF);
1782 esw_enable_vport(esw, vport, enabled_events);
1784 /* Enable ECPF vports */
1785 if (mlx5_ecpf_vport_exists(esw->dev)) {
1786 vport = mlx5_eswitch_get_vport(esw, MLX5_VPORT_ECPF);
1787 esw_enable_vport(esw, vport, enabled_events);
1790 /* Enable VF vports */
1791 mlx5_esw_for_each_vf_vport(esw, i, vport, nvfs)
1792 esw_enable_vport(esw, vport, enabled_events);
1794 if (mode == SRIOV_LEGACY) {
1795 MLX5_NB_INIT(&esw->nb, eswitch_vport_event, NIC_VPORT_CHANGE);
1796 mlx5_eq_notifier_register(esw->dev, &esw->nb);
1799 esw_info(esw->dev, "SRIOV enabled: active vports(%d)\n",
1800 esw->enabled_vports);
1804 esw->mode = SRIOV_NONE;
1806 if (mode == SRIOV_OFFLOADS) {
1807 mlx5_reload_interface(esw->dev, MLX5_INTERFACE_PROTOCOL_IB);
1808 mlx5_reload_interface(esw->dev, MLX5_INTERFACE_PROTOCOL_ETH);
1814 void mlx5_eswitch_disable_sriov(struct mlx5_eswitch *esw)
1816 struct esw_mc_addr *mc_promisc;
1817 struct mlx5_vport *vport;
1821 if (!ESW_ALLOWED(esw) || esw->mode == SRIOV_NONE)
1824 esw_info(esw->dev, "disable SRIOV: active vports(%d) mode(%d)\n",
1825 esw->enabled_vports, esw->mode);
1827 mc_promisc = &esw->mc_promisc;
1829 if (esw->mode == SRIOV_LEGACY)
1830 mlx5_eq_notifier_unregister(esw->dev, &esw->nb);
1832 mlx5_esw_for_all_vports(esw, i, vport)
1833 esw_disable_vport(esw, vport);
1835 if (mc_promisc && mc_promisc->uplink_rule)
1836 mlx5_del_flow_rules(mc_promisc->uplink_rule);
1838 esw_destroy_tsar(esw);
1840 if (esw->mode == SRIOV_LEGACY)
1841 esw_destroy_legacy_table(esw);
1842 else if (esw->mode == SRIOV_OFFLOADS)
1843 esw_offloads_cleanup(esw);
1845 old_mode = esw->mode;
1846 esw->mode = SRIOV_NONE;
1848 mlx5_lag_update(esw->dev);
1850 if (old_mode == SRIOV_OFFLOADS) {
1851 mlx5_reload_interface(esw->dev, MLX5_INTERFACE_PROTOCOL_IB);
1852 mlx5_reload_interface(esw->dev, MLX5_INTERFACE_PROTOCOL_ETH);
1856 int mlx5_eswitch_init(struct mlx5_core_dev *dev)
1858 int total_vports = MLX5_TOTAL_VPORTS(dev);
1859 struct mlx5_eswitch *esw;
1860 struct mlx5_vport *vport;
1863 if (!MLX5_VPORT_MANAGER(dev))
1867 "Total vports %d, per vport: max uc(%d) max mc(%d)\n",
1869 MLX5_MAX_UC_PER_VPORT(dev),
1870 MLX5_MAX_MC_PER_VPORT(dev));
1872 esw = kzalloc(sizeof(*esw), GFP_KERNEL);
1877 esw->manager_vport = mlx5_eswitch_manager_vport(dev);
1879 esw->work_queue = create_singlethread_workqueue("mlx5_esw_wq");
1880 if (!esw->work_queue) {
1885 esw->vports = kcalloc(total_vports, sizeof(struct mlx5_vport),
1892 esw->total_vports = total_vports;
1894 err = esw_offloads_init_reps(esw);
1898 hash_init(esw->offloads.encap_tbl);
1899 hash_init(esw->offloads.mod_hdr_tbl);
1900 mutex_init(&esw->state_lock);
1902 mlx5_esw_for_all_vports(esw, i, vport) {
1903 vport->vport = mlx5_eswitch_index_to_vport_num(esw, i);
1904 vport->info.link_state = MLX5_VPORT_ADMIN_STATE_AUTO;
1906 INIT_WORK(&vport->vport_change_handler,
1907 esw_vport_change_handler);
1910 esw->enabled_vports = 0;
1911 esw->mode = SRIOV_NONE;
1912 esw->offloads.inline_mode = MLX5_INLINE_MODE_NONE;
1913 if (MLX5_CAP_ESW_FLOWTABLE_FDB(dev, reformat) &&
1914 MLX5_CAP_ESW_FLOWTABLE_FDB(dev, decap))
1915 esw->offloads.encap = DEVLINK_ESWITCH_ENCAP_MODE_BASIC;
1917 esw->offloads.encap = DEVLINK_ESWITCH_ENCAP_MODE_NONE;
1919 dev->priv.eswitch = esw;
1922 if (esw->work_queue)
1923 destroy_workqueue(esw->work_queue);
1924 esw_offloads_cleanup_reps(esw);
1930 void mlx5_eswitch_cleanup(struct mlx5_eswitch *esw)
1932 if (!esw || !MLX5_VPORT_MANAGER(esw->dev))
1935 esw_info(esw->dev, "cleanup\n");
1937 esw->dev->priv.eswitch = NULL;
1938 destroy_workqueue(esw->work_queue);
1939 esw_offloads_cleanup_reps(esw);
1944 /* Vport Administration */
1945 int mlx5_eswitch_set_vport_mac(struct mlx5_eswitch *esw,
1946 u16 vport, u8 mac[ETH_ALEN])
1948 struct mlx5_vport *evport = mlx5_eswitch_get_vport(esw, vport);
1953 return PTR_ERR(evport);
1954 if (is_multicast_ether_addr(mac))
1957 mutex_lock(&esw->state_lock);
1959 if (evport->info.spoofchk && !is_valid_ether_addr(mac))
1960 mlx5_core_warn(esw->dev,
1961 "Set invalid MAC while spoofchk is on, vport(%d)\n",
1964 err = mlx5_modify_nic_vport_mac_address(esw->dev, vport, mac);
1966 mlx5_core_warn(esw->dev,
1967 "Failed to mlx5_modify_nic_vport_mac vport(%d) err=(%d)\n",
1972 node_guid_gen_from_mac(&node_guid, mac);
1973 err = mlx5_modify_nic_vport_node_guid(esw->dev, vport, node_guid);
1975 mlx5_core_warn(esw->dev,
1976 "Failed to set vport %d node guid, err = %d. RDMA_CM will not function properly for this VF.\n",
1979 ether_addr_copy(evport->info.mac, mac);
1980 evport->info.node_guid = node_guid;
1981 if (evport->enabled && esw->mode == SRIOV_LEGACY)
1982 err = esw_vport_ingress_config(esw, evport);
1985 mutex_unlock(&esw->state_lock);
1989 int mlx5_eswitch_set_vport_state(struct mlx5_eswitch *esw,
1990 u16 vport, int link_state)
1992 struct mlx5_vport *evport = mlx5_eswitch_get_vport(esw, vport);
1995 if (!ESW_ALLOWED(esw))
1998 return PTR_ERR(evport);
2000 mutex_lock(&esw->state_lock);
2002 err = mlx5_modify_vport_admin_state(esw->dev,
2003 MLX5_VPORT_STATE_OP_MOD_ESW_VPORT,
2004 vport, 1, link_state);
2006 mlx5_core_warn(esw->dev,
2007 "Failed to set vport %d link state, err = %d",
2012 evport->info.link_state = link_state;
2015 mutex_unlock(&esw->state_lock);
2019 int mlx5_eswitch_get_vport_config(struct mlx5_eswitch *esw,
2020 u16 vport, struct ifla_vf_info *ivi)
2022 struct mlx5_vport *evport = mlx5_eswitch_get_vport(esw, vport);
2025 return PTR_ERR(evport);
2027 memset(ivi, 0, sizeof(*ivi));
2028 ivi->vf = vport - 1;
2030 mutex_lock(&esw->state_lock);
2031 ether_addr_copy(ivi->mac, evport->info.mac);
2032 ivi->linkstate = evport->info.link_state;
2033 ivi->vlan = evport->info.vlan;
2034 ivi->qos = evport->info.qos;
2035 ivi->spoofchk = evport->info.spoofchk;
2036 ivi->trusted = evport->info.trusted;
2037 ivi->min_tx_rate = evport->info.min_rate;
2038 ivi->max_tx_rate = evport->info.max_rate;
2039 mutex_unlock(&esw->state_lock);
2044 int __mlx5_eswitch_set_vport_vlan(struct mlx5_eswitch *esw,
2045 u16 vport, u16 vlan, u8 qos, u8 set_flags)
2047 struct mlx5_vport *evport = mlx5_eswitch_get_vport(esw, vport);
2050 if (!ESW_ALLOWED(esw))
2053 return PTR_ERR(evport);
2054 if (vlan > 4095 || qos > 7)
2057 mutex_lock(&esw->state_lock);
2059 err = modify_esw_vport_cvlan(esw->dev, vport, vlan, qos, set_flags);
2063 evport->info.vlan = vlan;
2064 evport->info.qos = qos;
2065 if (evport->enabled && esw->mode == SRIOV_LEGACY) {
2066 err = esw_vport_ingress_config(esw, evport);
2069 err = esw_vport_egress_config(esw, evport);
2073 mutex_unlock(&esw->state_lock);
2077 int mlx5_eswitch_set_vport_vlan(struct mlx5_eswitch *esw,
2078 u16 vport, u16 vlan, u8 qos)
2083 set_flags = SET_VLAN_STRIP | SET_VLAN_INSERT;
2085 return __mlx5_eswitch_set_vport_vlan(esw, vport, vlan, qos, set_flags);
2088 int mlx5_eswitch_set_vport_spoofchk(struct mlx5_eswitch *esw,
2089 u16 vport, bool spoofchk)
2091 struct mlx5_vport *evport = mlx5_eswitch_get_vport(esw, vport);
2095 if (!ESW_ALLOWED(esw))
2098 return PTR_ERR(evport);
2100 mutex_lock(&esw->state_lock);
2101 pschk = evport->info.spoofchk;
2102 evport->info.spoofchk = spoofchk;
2103 if (pschk && !is_valid_ether_addr(evport->info.mac))
2104 mlx5_core_warn(esw->dev,
2105 "Spoofchk in set while MAC is invalid, vport(%d)\n",
2107 if (evport->enabled && esw->mode == SRIOV_LEGACY)
2108 err = esw_vport_ingress_config(esw, evport);
2110 evport->info.spoofchk = pschk;
2111 mutex_unlock(&esw->state_lock);
2116 static void esw_cleanup_vepa_rules(struct mlx5_eswitch *esw)
2118 if (esw->fdb_table.legacy.vepa_uplink_rule)
2119 mlx5_del_flow_rules(esw->fdb_table.legacy.vepa_uplink_rule);
2121 if (esw->fdb_table.legacy.vepa_star_rule)
2122 mlx5_del_flow_rules(esw->fdb_table.legacy.vepa_star_rule);
2124 esw->fdb_table.legacy.vepa_uplink_rule = NULL;
2125 esw->fdb_table.legacy.vepa_star_rule = NULL;
2128 static int _mlx5_eswitch_set_vepa_locked(struct mlx5_eswitch *esw,
2131 struct mlx5_flow_destination dest = {};
2132 struct mlx5_flow_act flow_act = {};
2133 struct mlx5_flow_handle *flow_rule;
2134 struct mlx5_flow_spec *spec;
2139 esw_cleanup_vepa_rules(esw);
2143 if (esw->fdb_table.legacy.vepa_uplink_rule)
2146 spec = kvzalloc(sizeof(*spec), GFP_KERNEL);
2150 /* Uplink rule forward uplink traffic to FDB */
2151 misc = MLX5_ADDR_OF(fte_match_param, spec->match_value, misc_parameters);
2152 MLX5_SET(fte_match_set_misc, misc, source_port, MLX5_VPORT_UPLINK);
2154 misc = MLX5_ADDR_OF(fte_match_param, spec->match_criteria, misc_parameters);
2155 MLX5_SET_TO_ONES(fte_match_set_misc, misc, source_port);
2157 spec->match_criteria_enable = MLX5_MATCH_MISC_PARAMETERS;
2158 dest.type = MLX5_FLOW_DESTINATION_TYPE_FLOW_TABLE;
2159 dest.ft = esw->fdb_table.legacy.fdb;
2160 flow_act.action = MLX5_FLOW_CONTEXT_ACTION_FWD_DEST;
2161 flow_rule = mlx5_add_flow_rules(esw->fdb_table.legacy.vepa_fdb, spec,
2162 &flow_act, &dest, 1);
2163 if (IS_ERR(flow_rule)) {
2164 err = PTR_ERR(flow_rule);
2167 esw->fdb_table.legacy.vepa_uplink_rule = flow_rule;
2170 /* Star rule to forward all traffic to uplink vport */
2171 memset(spec, 0, sizeof(*spec));
2172 memset(&dest, 0, sizeof(dest));
2173 dest.type = MLX5_FLOW_DESTINATION_TYPE_VPORT;
2174 dest.vport.num = MLX5_VPORT_UPLINK;
2175 flow_act.action = MLX5_FLOW_CONTEXT_ACTION_FWD_DEST;
2176 flow_rule = mlx5_add_flow_rules(esw->fdb_table.legacy.vepa_fdb, spec,
2177 &flow_act, &dest, 1);
2178 if (IS_ERR(flow_rule)) {
2179 err = PTR_ERR(flow_rule);
2182 esw->fdb_table.legacy.vepa_star_rule = flow_rule;
2188 esw_cleanup_vepa_rules(esw);
2192 int mlx5_eswitch_set_vepa(struct mlx5_eswitch *esw, u8 setting)
2199 if (!ESW_ALLOWED(esw))
2202 mutex_lock(&esw->state_lock);
2203 if (esw->mode != SRIOV_LEGACY) {
2208 err = _mlx5_eswitch_set_vepa_locked(esw, setting);
2211 mutex_unlock(&esw->state_lock);
2215 int mlx5_eswitch_get_vepa(struct mlx5_eswitch *esw, u8 *setting)
2222 if (!ESW_ALLOWED(esw))
2225 mutex_lock(&esw->state_lock);
2226 if (esw->mode != SRIOV_LEGACY) {
2231 *setting = esw->fdb_table.legacy.vepa_uplink_rule ? 1 : 0;
2234 mutex_unlock(&esw->state_lock);
2238 int mlx5_eswitch_set_vport_trust(struct mlx5_eswitch *esw,
2239 u16 vport, bool setting)
2241 struct mlx5_vport *evport = mlx5_eswitch_get_vport(esw, vport);
2243 if (!ESW_ALLOWED(esw))
2246 return PTR_ERR(evport);
2248 mutex_lock(&esw->state_lock);
2249 evport->info.trusted = setting;
2250 if (evport->enabled)
2251 esw_vport_change_handle_locked(evport);
2252 mutex_unlock(&esw->state_lock);
2257 static u32 calculate_vports_min_rate_divider(struct mlx5_eswitch *esw)
2259 u32 fw_max_bw_share = MLX5_CAP_QOS(esw->dev, max_tsar_bw_share);
2260 struct mlx5_vport *evport;
2261 u32 max_guarantee = 0;
2264 mlx5_esw_for_all_vports(esw, i, evport) {
2265 if (!evport->enabled || evport->info.min_rate < max_guarantee)
2267 max_guarantee = evport->info.min_rate;
2270 return max_t(u32, max_guarantee / fw_max_bw_share, 1);
2273 static int normalize_vports_min_rate(struct mlx5_eswitch *esw, u32 divider)
2275 u32 fw_max_bw_share = MLX5_CAP_QOS(esw->dev, max_tsar_bw_share);
2276 struct mlx5_vport *evport;
2283 mlx5_esw_for_all_vports(esw, i, evport) {
2284 if (!evport->enabled)
2286 vport_min_rate = evport->info.min_rate;
2287 vport_max_rate = evport->info.max_rate;
2288 bw_share = MLX5_MIN_BW_SHARE;
2291 bw_share = MLX5_RATE_TO_BW_SHARE(vport_min_rate,
2295 if (bw_share == evport->qos.bw_share)
2298 err = esw_vport_qos_config(esw, evport, vport_max_rate,
2301 evport->qos.bw_share = bw_share;
2309 int mlx5_eswitch_set_vport_rate(struct mlx5_eswitch *esw, u16 vport,
2310 u32 max_rate, u32 min_rate)
2312 struct mlx5_vport *evport = mlx5_eswitch_get_vport(esw, vport);
2313 u32 fw_max_bw_share;
2314 u32 previous_min_rate;
2316 bool min_rate_supported;
2317 bool max_rate_supported;
2320 if (!ESW_ALLOWED(esw))
2323 return PTR_ERR(evport);
2325 fw_max_bw_share = MLX5_CAP_QOS(esw->dev, max_tsar_bw_share);
2326 min_rate_supported = MLX5_CAP_QOS(esw->dev, esw_bw_share) &&
2327 fw_max_bw_share >= MLX5_MIN_BW_SHARE;
2328 max_rate_supported = MLX5_CAP_QOS(esw->dev, esw_rate_limit);
2330 if ((min_rate && !min_rate_supported) || (max_rate && !max_rate_supported))
2333 mutex_lock(&esw->state_lock);
2335 if (min_rate == evport->info.min_rate)
2338 previous_min_rate = evport->info.min_rate;
2339 evport->info.min_rate = min_rate;
2340 divider = calculate_vports_min_rate_divider(esw);
2341 err = normalize_vports_min_rate(esw, divider);
2343 evport->info.min_rate = previous_min_rate;
2348 if (max_rate == evport->info.max_rate)
2351 err = esw_vport_qos_config(esw, evport, max_rate, evport->qos.bw_share);
2353 evport->info.max_rate = max_rate;
2356 mutex_unlock(&esw->state_lock);
2360 static int mlx5_eswitch_query_vport_drop_stats(struct mlx5_core_dev *dev,
2361 struct mlx5_vport *vport,
2362 struct mlx5_vport_drop_stats *stats)
2364 struct mlx5_eswitch *esw = dev->priv.eswitch;
2365 u64 rx_discard_vport_down, tx_discard_vport_down;
2369 if (!vport->enabled || esw->mode != SRIOV_LEGACY)
2372 if (vport->egress.drop_counter)
2373 mlx5_fc_query(dev, vport->egress.drop_counter,
2374 &stats->rx_dropped, &bytes);
2376 if (vport->ingress.drop_counter)
2377 mlx5_fc_query(dev, vport->ingress.drop_counter,
2378 &stats->tx_dropped, &bytes);
2380 if (!MLX5_CAP_GEN(dev, receive_discard_vport_down) &&
2381 !MLX5_CAP_GEN(dev, transmit_discard_vport_down))
2384 err = mlx5_query_vport_down_stats(dev, vport->vport, 1,
2385 &rx_discard_vport_down,
2386 &tx_discard_vport_down);
2390 if (MLX5_CAP_GEN(dev, receive_discard_vport_down))
2391 stats->rx_dropped += rx_discard_vport_down;
2392 if (MLX5_CAP_GEN(dev, transmit_discard_vport_down))
2393 stats->tx_dropped += tx_discard_vport_down;
2398 int mlx5_eswitch_get_vport_stats(struct mlx5_eswitch *esw,
2400 struct ifla_vf_stats *vf_stats)
2402 struct mlx5_vport *vport = mlx5_eswitch_get_vport(esw, vport_num);
2403 int outlen = MLX5_ST_SZ_BYTES(query_vport_counter_out);
2404 u32 in[MLX5_ST_SZ_DW(query_vport_counter_in)] = {0};
2405 struct mlx5_vport_drop_stats stats = {0};
2410 return PTR_ERR(vport);
2412 out = kvzalloc(outlen, GFP_KERNEL);
2416 MLX5_SET(query_vport_counter_in, in, opcode,
2417 MLX5_CMD_OP_QUERY_VPORT_COUNTER);
2418 MLX5_SET(query_vport_counter_in, in, op_mod, 0);
2419 MLX5_SET(query_vport_counter_in, in, vport_number, vport->vport);
2420 MLX5_SET(query_vport_counter_in, in, other_vport, 1);
2422 memset(out, 0, outlen);
2423 err = mlx5_cmd_exec(esw->dev, in, sizeof(in), out, outlen);
2427 #define MLX5_GET_CTR(p, x) \
2428 MLX5_GET64(query_vport_counter_out, p, x)
2430 memset(vf_stats, 0, sizeof(*vf_stats));
2431 vf_stats->rx_packets =
2432 MLX5_GET_CTR(out, received_eth_unicast.packets) +
2433 MLX5_GET_CTR(out, received_ib_unicast.packets) +
2434 MLX5_GET_CTR(out, received_eth_multicast.packets) +
2435 MLX5_GET_CTR(out, received_ib_multicast.packets) +
2436 MLX5_GET_CTR(out, received_eth_broadcast.packets);
2438 vf_stats->rx_bytes =
2439 MLX5_GET_CTR(out, received_eth_unicast.octets) +
2440 MLX5_GET_CTR(out, received_ib_unicast.octets) +
2441 MLX5_GET_CTR(out, received_eth_multicast.octets) +
2442 MLX5_GET_CTR(out, received_ib_multicast.octets) +
2443 MLX5_GET_CTR(out, received_eth_broadcast.octets);
2445 vf_stats->tx_packets =
2446 MLX5_GET_CTR(out, transmitted_eth_unicast.packets) +
2447 MLX5_GET_CTR(out, transmitted_ib_unicast.packets) +
2448 MLX5_GET_CTR(out, transmitted_eth_multicast.packets) +
2449 MLX5_GET_CTR(out, transmitted_ib_multicast.packets) +
2450 MLX5_GET_CTR(out, transmitted_eth_broadcast.packets);
2452 vf_stats->tx_bytes =
2453 MLX5_GET_CTR(out, transmitted_eth_unicast.octets) +
2454 MLX5_GET_CTR(out, transmitted_ib_unicast.octets) +
2455 MLX5_GET_CTR(out, transmitted_eth_multicast.octets) +
2456 MLX5_GET_CTR(out, transmitted_ib_multicast.octets) +
2457 MLX5_GET_CTR(out, transmitted_eth_broadcast.octets);
2459 vf_stats->multicast =
2460 MLX5_GET_CTR(out, received_eth_multicast.packets) +
2461 MLX5_GET_CTR(out, received_ib_multicast.packets);
2463 vf_stats->broadcast =
2464 MLX5_GET_CTR(out, received_eth_broadcast.packets);
2466 err = mlx5_eswitch_query_vport_drop_stats(esw->dev, vport, &stats);
2469 vf_stats->rx_dropped = stats.rx_dropped;
2470 vf_stats->tx_dropped = stats.tx_dropped;
2477 u8 mlx5_eswitch_mode(struct mlx5_eswitch *esw)
2479 return ESW_ALLOWED(esw) ? esw->mode : SRIOV_NONE;
2481 EXPORT_SYMBOL_GPL(mlx5_eswitch_mode);
2483 bool mlx5_esw_lag_prereq(struct mlx5_core_dev *dev0, struct mlx5_core_dev *dev1)
2485 if ((dev0->priv.eswitch->mode == SRIOV_NONE &&
2486 dev1->priv.eswitch->mode == SRIOV_NONE) ||
2487 (dev0->priv.eswitch->mode == SRIOV_OFFLOADS &&
2488 dev1->priv.eswitch->mode == SRIOV_OFFLOADS))
2494 bool mlx5_esw_multipath_prereq(struct mlx5_core_dev *dev0,
2495 struct mlx5_core_dev *dev1)
2497 return (dev0->priv.eswitch->mode == SRIOV_OFFLOADS &&
2498 dev1->priv.eswitch->mode == SRIOV_OFFLOADS);