2 * Copyright (c) 2015, Mellanox Technologies. All rights reserved.
4 * This software is available to you under a choice of one of two
5 * licenses. You may choose to be licensed under the terms of the GNU
6 * General Public License (GPL) Version 2, available from the file
7 * COPYING in the main directory of this source tree, or the
8 * OpenIB.org BSD license below:
10 * Redistribution and use in source and binary forms, with or
11 * without modification, are permitted provided that the following
14 * - Redistributions of source code must retain the above
15 * copyright notice, this list of conditions and the following
18 * - Redistributions in binary form must reproduce the above
19 * copyright notice, this list of conditions and the following
20 * disclaimer in the documentation and/or other materials
21 * provided with the distribution.
23 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
24 * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
25 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
26 * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
27 * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
28 * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
29 * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
33 #include <linux/etherdevice.h>
34 #include <linux/mlx5/driver.h>
35 #include <linux/mlx5/mlx5_ifc.h>
36 #include <linux/mlx5/vport.h>
37 #include <linux/mlx5/fs.h>
38 #include "mlx5_core.h"
41 #define UPLINK_VPORT 0xFFFF
49 /* E-Switch UC L2 table hash node */
51 struct l2addr_node node;
56 /* E-Switch MC FDB table hash node */
57 struct esw_mc_addr { /* SRIOV only */
58 struct l2addr_node node;
59 struct mlx5_flow_rule *uplink_rule; /* Forward to uplink rule */
63 /* Vport UC/MC hash node */
65 struct l2addr_node node;
68 struct mlx5_flow_rule *flow_rule; /* SRIOV only */
69 /* A flag indicating that mac was added due to mc promiscuous vport */
74 UC_ADDR_CHANGE = BIT(0),
75 MC_ADDR_CHANGE = BIT(1),
76 PROMISC_CHANGE = BIT(3),
79 /* Vport context events */
80 #define SRIOV_VPORT_EVENTS (UC_ADDR_CHANGE | \
84 int esw_offloads_init(struct mlx5_eswitch *esw, int nvports);
85 void esw_offloads_cleanup(struct mlx5_eswitch *esw, int nvports);
87 static int arm_vport_context_events_cmd(struct mlx5_core_dev *dev, u16 vport,
90 int in[MLX5_ST_SZ_DW(modify_nic_vport_context_in)] = {0};
91 int out[MLX5_ST_SZ_DW(modify_nic_vport_context_out)] = {0};
94 MLX5_SET(modify_nic_vport_context_in, in,
95 opcode, MLX5_CMD_OP_MODIFY_NIC_VPORT_CONTEXT);
96 MLX5_SET(modify_nic_vport_context_in, in, field_select.change_event, 1);
97 MLX5_SET(modify_nic_vport_context_in, in, vport_number, vport);
99 MLX5_SET(modify_nic_vport_context_in, in, other_vport, 1);
100 nic_vport_ctx = MLX5_ADDR_OF(modify_nic_vport_context_in,
101 in, nic_vport_context);
103 MLX5_SET(nic_vport_context, nic_vport_ctx, arm_change_event, 1);
105 if (events_mask & UC_ADDR_CHANGE)
106 MLX5_SET(nic_vport_context, nic_vport_ctx,
107 event_on_uc_address_change, 1);
108 if (events_mask & MC_ADDR_CHANGE)
109 MLX5_SET(nic_vport_context, nic_vport_ctx,
110 event_on_mc_address_change, 1);
111 if (events_mask & PROMISC_CHANGE)
112 MLX5_SET(nic_vport_context, nic_vport_ctx,
113 event_on_promisc_change, 1);
115 return mlx5_cmd_exec(dev, in, sizeof(in), out, sizeof(out));
118 /* E-Switch vport context HW commands */
119 static int modify_esw_vport_context_cmd(struct mlx5_core_dev *dev, u16 vport,
122 u32 out[MLX5_ST_SZ_DW(modify_esw_vport_context_out)] = {0};
124 MLX5_SET(modify_esw_vport_context_in, in, opcode,
125 MLX5_CMD_OP_MODIFY_ESW_VPORT_CONTEXT);
126 MLX5_SET(modify_esw_vport_context_in, in, vport_number, vport);
128 MLX5_SET(modify_esw_vport_context_in, in, other_vport, 1);
129 return mlx5_cmd_exec(dev, in, inlen, out, sizeof(out));
132 static int modify_esw_vport_cvlan(struct mlx5_core_dev *dev, u32 vport,
133 u16 vlan, u8 qos, bool set)
135 u32 in[MLX5_ST_SZ_DW(modify_esw_vport_context_in)] = {0};
137 if (!MLX5_CAP_ESW(dev, vport_cvlan_strip) ||
138 !MLX5_CAP_ESW(dev, vport_cvlan_insert_if_not_exist))
141 esw_debug(dev, "Set Vport[%d] VLAN %d qos %d set=%d\n",
142 vport, vlan, qos, set);
144 MLX5_SET(modify_esw_vport_context_in, in,
145 esw_vport_context.vport_cvlan_strip, 1);
146 /* insert only if no vlan in packet */
147 MLX5_SET(modify_esw_vport_context_in, in,
148 esw_vport_context.vport_cvlan_insert, 1);
149 MLX5_SET(modify_esw_vport_context_in, in,
150 esw_vport_context.cvlan_pcp, qos);
151 MLX5_SET(modify_esw_vport_context_in, in,
152 esw_vport_context.cvlan_id, vlan);
155 MLX5_SET(modify_esw_vport_context_in, in,
156 field_select.vport_cvlan_strip, 1);
157 MLX5_SET(modify_esw_vport_context_in, in,
158 field_select.vport_cvlan_insert, 1);
160 return modify_esw_vport_context_cmd(dev, vport, in, sizeof(in));
163 /* HW L2 Table (MPFS) management */
164 static int set_l2_table_entry_cmd(struct mlx5_core_dev *dev, u32 index,
165 u8 *mac, u8 vlan_valid, u16 vlan)
167 u32 in[MLX5_ST_SZ_DW(set_l2_table_entry_in)] = {0};
168 u32 out[MLX5_ST_SZ_DW(set_l2_table_entry_out)] = {0};
171 MLX5_SET(set_l2_table_entry_in, in, opcode,
172 MLX5_CMD_OP_SET_L2_TABLE_ENTRY);
173 MLX5_SET(set_l2_table_entry_in, in, table_index, index);
174 MLX5_SET(set_l2_table_entry_in, in, vlan_valid, vlan_valid);
175 MLX5_SET(set_l2_table_entry_in, in, vlan, vlan);
177 in_mac_addr = MLX5_ADDR_OF(set_l2_table_entry_in, in, mac_address);
178 ether_addr_copy(&in_mac_addr[2], mac);
180 return mlx5_cmd_exec(dev, in, sizeof(in), out, sizeof(out));
183 static int del_l2_table_entry_cmd(struct mlx5_core_dev *dev, u32 index)
185 u32 in[MLX5_ST_SZ_DW(delete_l2_table_entry_in)] = {0};
186 u32 out[MLX5_ST_SZ_DW(delete_l2_table_entry_out)] = {0};
188 MLX5_SET(delete_l2_table_entry_in, in, opcode,
189 MLX5_CMD_OP_DELETE_L2_TABLE_ENTRY);
190 MLX5_SET(delete_l2_table_entry_in, in, table_index, index);
191 return mlx5_cmd_exec(dev, in, sizeof(in), out, sizeof(out));
194 static int alloc_l2_table_index(struct mlx5_l2_table *l2_table, u32 *ix)
198 *ix = find_first_zero_bit(l2_table->bitmap, l2_table->size);
199 if (*ix >= l2_table->size)
202 __set_bit(*ix, l2_table->bitmap);
207 static void free_l2_table_index(struct mlx5_l2_table *l2_table, u32 ix)
209 __clear_bit(ix, l2_table->bitmap);
212 static int set_l2_table_entry(struct mlx5_core_dev *dev, u8 *mac,
213 u8 vlan_valid, u16 vlan,
216 struct mlx5_l2_table *l2_table = &dev->priv.eswitch->l2_table;
219 err = alloc_l2_table_index(l2_table, index);
223 err = set_l2_table_entry_cmd(dev, *index, mac, vlan_valid, vlan);
225 free_l2_table_index(l2_table, *index);
230 static void del_l2_table_entry(struct mlx5_core_dev *dev, u32 index)
232 struct mlx5_l2_table *l2_table = &dev->priv.eswitch->l2_table;
234 del_l2_table_entry_cmd(dev, index);
235 free_l2_table_index(l2_table, index);
239 static struct mlx5_flow_rule *
240 __esw_fdb_set_vport_rule(struct mlx5_eswitch *esw, u32 vport, bool rx_rule,
241 u8 mac_c[ETH_ALEN], u8 mac_v[ETH_ALEN])
243 int match_header = (is_zero_ether_addr(mac_c) ? 0 :
244 MLX5_MATCH_OUTER_HEADERS);
245 struct mlx5_flow_rule *flow_rule = NULL;
246 struct mlx5_flow_destination dest;
247 struct mlx5_flow_spec *spec;
248 void *mv_misc = NULL;
249 void *mc_misc = NULL;
254 match_header |= MLX5_MATCH_MISC_PARAMETERS;
256 spec = mlx5_vzalloc(sizeof(*spec));
258 esw_warn(esw->dev, "FDB: Failed to alloc match parameters\n");
261 dmac_v = MLX5_ADDR_OF(fte_match_param, spec->match_value,
262 outer_headers.dmac_47_16);
263 dmac_c = MLX5_ADDR_OF(fte_match_param, spec->match_criteria,
264 outer_headers.dmac_47_16);
266 if (match_header & MLX5_MATCH_OUTER_HEADERS) {
267 ether_addr_copy(dmac_v, mac_v);
268 ether_addr_copy(dmac_c, mac_c);
271 if (match_header & MLX5_MATCH_MISC_PARAMETERS) {
272 mv_misc = MLX5_ADDR_OF(fte_match_param, spec->match_value,
274 mc_misc = MLX5_ADDR_OF(fte_match_param, spec->match_criteria,
276 MLX5_SET(fte_match_set_misc, mv_misc, source_port, UPLINK_VPORT);
277 MLX5_SET_TO_ONES(fte_match_set_misc, mc_misc, source_port);
280 dest.type = MLX5_FLOW_DESTINATION_TYPE_VPORT;
281 dest.vport_num = vport;
284 "\tFDB add rule dmac_v(%pM) dmac_c(%pM) -> vport(%d)\n",
285 dmac_v, dmac_c, vport);
286 spec->match_criteria_enable = match_header;
288 mlx5_add_flow_rule(esw->fdb_table.fdb, spec,
289 MLX5_FLOW_CONTEXT_ACTION_FWD_DEST,
291 if (IS_ERR(flow_rule)) {
293 "FDB: Failed to add flow rule: dmac_v(%pM) dmac_c(%pM) -> vport(%d), err(%ld)\n",
294 dmac_v, dmac_c, vport, PTR_ERR(flow_rule));
302 static struct mlx5_flow_rule *
303 esw_fdb_set_vport_rule(struct mlx5_eswitch *esw, u8 mac[ETH_ALEN], u32 vport)
307 eth_broadcast_addr(mac_c);
308 return __esw_fdb_set_vport_rule(esw, vport, false, mac_c, mac);
311 static struct mlx5_flow_rule *
312 esw_fdb_set_vport_allmulti_rule(struct mlx5_eswitch *esw, u32 vport)
317 eth_zero_addr(mac_c);
318 eth_zero_addr(mac_v);
321 return __esw_fdb_set_vport_rule(esw, vport, false, mac_c, mac_v);
324 static struct mlx5_flow_rule *
325 esw_fdb_set_vport_promisc_rule(struct mlx5_eswitch *esw, u32 vport)
330 eth_zero_addr(mac_c);
331 eth_zero_addr(mac_v);
332 return __esw_fdb_set_vport_rule(esw, vport, true, mac_c, mac_v);
335 static int esw_create_legacy_fdb_table(struct mlx5_eswitch *esw, int nvports)
337 int inlen = MLX5_ST_SZ_BYTES(create_flow_group_in);
338 struct mlx5_core_dev *dev = esw->dev;
339 struct mlx5_flow_namespace *root_ns;
340 struct mlx5_flow_table *fdb;
341 struct mlx5_flow_group *g;
342 void *match_criteria;
348 esw_debug(dev, "Create FDB log_max_size(%d)\n",
349 MLX5_CAP_ESW_FLOWTABLE_FDB(dev, log_max_ft_size));
351 root_ns = mlx5_get_flow_namespace(dev, MLX5_FLOW_NAMESPACE_FDB);
353 esw_warn(dev, "Failed to get FDB flow namespace\n");
357 flow_group_in = mlx5_vzalloc(inlen);
360 memset(flow_group_in, 0, inlen);
362 table_size = BIT(MLX5_CAP_ESW_FLOWTABLE_FDB(dev, log_max_ft_size));
363 fdb = mlx5_create_flow_table(root_ns, 0, table_size, 0);
366 esw_warn(dev, "Failed to create FDB Table err %d\n", err);
369 esw->fdb_table.fdb = fdb;
371 /* Addresses group : Full match unicast/multicast addresses */
372 MLX5_SET(create_flow_group_in, flow_group_in, match_criteria_enable,
373 MLX5_MATCH_OUTER_HEADERS);
374 match_criteria = MLX5_ADDR_OF(create_flow_group_in, flow_group_in, match_criteria);
375 dmac = MLX5_ADDR_OF(fte_match_param, match_criteria, outer_headers.dmac_47_16);
376 MLX5_SET(create_flow_group_in, flow_group_in, start_flow_index, 0);
377 /* Preserve 2 entries for allmulti and promisc rules*/
378 MLX5_SET(create_flow_group_in, flow_group_in, end_flow_index, table_size - 3);
379 eth_broadcast_addr(dmac);
380 g = mlx5_create_flow_group(fdb, flow_group_in);
383 esw_warn(dev, "Failed to create flow group err(%d)\n", err);
386 esw->fdb_table.legacy.addr_grp = g;
388 /* Allmulti group : One rule that forwards any mcast traffic */
389 MLX5_SET(create_flow_group_in, flow_group_in, match_criteria_enable,
390 MLX5_MATCH_OUTER_HEADERS);
391 MLX5_SET(create_flow_group_in, flow_group_in, start_flow_index, table_size - 2);
392 MLX5_SET(create_flow_group_in, flow_group_in, end_flow_index, table_size - 2);
395 g = mlx5_create_flow_group(fdb, flow_group_in);
398 esw_warn(dev, "Failed to create allmulti flow group err(%d)\n", err);
401 esw->fdb_table.legacy.allmulti_grp = g;
403 /* Promiscuous group :
404 * One rule that forward all unmatched traffic from previous groups
407 MLX5_SET(create_flow_group_in, flow_group_in, match_criteria_enable,
408 MLX5_MATCH_MISC_PARAMETERS);
409 MLX5_SET_TO_ONES(fte_match_param, match_criteria, misc_parameters.source_port);
410 MLX5_SET(create_flow_group_in, flow_group_in, start_flow_index, table_size - 1);
411 MLX5_SET(create_flow_group_in, flow_group_in, end_flow_index, table_size - 1);
412 g = mlx5_create_flow_group(fdb, flow_group_in);
415 esw_warn(dev, "Failed to create promisc flow group err(%d)\n", err);
418 esw->fdb_table.legacy.promisc_grp = g;
422 if (!IS_ERR_OR_NULL(esw->fdb_table.legacy.allmulti_grp)) {
423 mlx5_destroy_flow_group(esw->fdb_table.legacy.allmulti_grp);
424 esw->fdb_table.legacy.allmulti_grp = NULL;
426 if (!IS_ERR_OR_NULL(esw->fdb_table.legacy.addr_grp)) {
427 mlx5_destroy_flow_group(esw->fdb_table.legacy.addr_grp);
428 esw->fdb_table.legacy.addr_grp = NULL;
430 if (!IS_ERR_OR_NULL(esw->fdb_table.fdb)) {
431 mlx5_destroy_flow_table(esw->fdb_table.fdb);
432 esw->fdb_table.fdb = NULL;
436 kvfree(flow_group_in);
440 static void esw_destroy_legacy_fdb_table(struct mlx5_eswitch *esw)
442 if (!esw->fdb_table.fdb)
445 esw_debug(esw->dev, "Destroy FDB Table\n");
446 mlx5_destroy_flow_group(esw->fdb_table.legacy.promisc_grp);
447 mlx5_destroy_flow_group(esw->fdb_table.legacy.allmulti_grp);
448 mlx5_destroy_flow_group(esw->fdb_table.legacy.addr_grp);
449 mlx5_destroy_flow_table(esw->fdb_table.fdb);
450 esw->fdb_table.fdb = NULL;
451 esw->fdb_table.legacy.addr_grp = NULL;
452 esw->fdb_table.legacy.allmulti_grp = NULL;
453 esw->fdb_table.legacy.promisc_grp = NULL;
456 /* E-Switch vport UC/MC lists management */
457 typedef int (*vport_addr_action)(struct mlx5_eswitch *esw,
458 struct vport_addr *vaddr);
460 static int esw_add_uc_addr(struct mlx5_eswitch *esw, struct vport_addr *vaddr)
462 struct hlist_head *hash = esw->l2_table.l2_hash;
463 struct esw_uc_addr *esw_uc;
464 u8 *mac = vaddr->node.addr;
465 u32 vport = vaddr->vport;
468 esw_uc = l2addr_hash_find(hash, mac, struct esw_uc_addr);
471 "Failed to set L2 mac(%pM) for vport(%d), mac is already in use by vport(%d)\n",
472 mac, vport, esw_uc->vport);
476 esw_uc = l2addr_hash_add(hash, mac, struct esw_uc_addr, GFP_KERNEL);
479 esw_uc->vport = vport;
481 err = set_l2_table_entry(esw->dev, mac, 0, 0, &esw_uc->table_index);
485 /* SRIOV is enabled: Forward UC MAC to vport */
486 if (esw->fdb_table.fdb && esw->mode == SRIOV_LEGACY)
487 vaddr->flow_rule = esw_fdb_set_vport_rule(esw, mac, vport);
489 esw_debug(esw->dev, "\tADDED UC MAC: vport[%d] %pM index:%d fr(%p)\n",
490 vport, mac, esw_uc->table_index, vaddr->flow_rule);
493 l2addr_hash_del(esw_uc);
497 static int esw_del_uc_addr(struct mlx5_eswitch *esw, struct vport_addr *vaddr)
499 struct hlist_head *hash = esw->l2_table.l2_hash;
500 struct esw_uc_addr *esw_uc;
501 u8 *mac = vaddr->node.addr;
502 u32 vport = vaddr->vport;
504 esw_uc = l2addr_hash_find(hash, mac, struct esw_uc_addr);
505 if (!esw_uc || esw_uc->vport != vport) {
507 "MAC(%pM) doesn't belong to vport (%d)\n",
511 esw_debug(esw->dev, "\tDELETE UC MAC: vport[%d] %pM index:%d fr(%p)\n",
512 vport, mac, esw_uc->table_index, vaddr->flow_rule);
514 del_l2_table_entry(esw->dev, esw_uc->table_index);
516 if (vaddr->flow_rule)
517 mlx5_del_flow_rule(vaddr->flow_rule);
518 vaddr->flow_rule = NULL;
520 l2addr_hash_del(esw_uc);
524 static void update_allmulti_vports(struct mlx5_eswitch *esw,
525 struct vport_addr *vaddr,
526 struct esw_mc_addr *esw_mc)
528 u8 *mac = vaddr->node.addr;
531 for (vport_idx = 0; vport_idx < esw->total_vports; vport_idx++) {
532 struct mlx5_vport *vport = &esw->vports[vport_idx];
533 struct hlist_head *vport_hash = vport->mc_list;
534 struct vport_addr *iter_vaddr =
535 l2addr_hash_find(vport_hash,
538 if (IS_ERR_OR_NULL(vport->allmulti_rule) ||
539 vaddr->vport == vport_idx)
541 switch (vaddr->action) {
542 case MLX5_ACTION_ADD:
545 iter_vaddr = l2addr_hash_add(vport_hash, mac,
550 "ALL-MULTI: Failed to add MAC(%pM) to vport[%d] DB\n",
554 iter_vaddr->vport = vport_idx;
555 iter_vaddr->flow_rule =
556 esw_fdb_set_vport_rule(esw,
559 iter_vaddr->mc_promisc = true;
561 case MLX5_ACTION_DEL:
564 mlx5_del_flow_rule(iter_vaddr->flow_rule);
565 l2addr_hash_del(iter_vaddr);
571 static int esw_add_mc_addr(struct mlx5_eswitch *esw, struct vport_addr *vaddr)
573 struct hlist_head *hash = esw->mc_table;
574 struct esw_mc_addr *esw_mc;
575 u8 *mac = vaddr->node.addr;
576 u32 vport = vaddr->vport;
578 if (!esw->fdb_table.fdb)
581 esw_mc = l2addr_hash_find(hash, mac, struct esw_mc_addr);
585 esw_mc = l2addr_hash_add(hash, mac, struct esw_mc_addr, GFP_KERNEL);
589 esw_mc->uplink_rule = /* Forward MC MAC to Uplink */
590 esw_fdb_set_vport_rule(esw, mac, UPLINK_VPORT);
592 /* Add this multicast mac to all the mc promiscuous vports */
593 update_allmulti_vports(esw, vaddr, esw_mc);
596 /* If the multicast mac is added as a result of mc promiscuous vport,
597 * don't increment the multicast ref count
599 if (!vaddr->mc_promisc)
602 /* Forward MC MAC to vport */
603 vaddr->flow_rule = esw_fdb_set_vport_rule(esw, mac, vport);
605 "\tADDED MC MAC: vport[%d] %pM fr(%p) refcnt(%d) uplinkfr(%p)\n",
606 vport, mac, vaddr->flow_rule,
607 esw_mc->refcnt, esw_mc->uplink_rule);
611 static int esw_del_mc_addr(struct mlx5_eswitch *esw, struct vport_addr *vaddr)
613 struct hlist_head *hash = esw->mc_table;
614 struct esw_mc_addr *esw_mc;
615 u8 *mac = vaddr->node.addr;
616 u32 vport = vaddr->vport;
618 if (!esw->fdb_table.fdb)
621 esw_mc = l2addr_hash_find(hash, mac, struct esw_mc_addr);
624 "Failed to find eswitch MC addr for MAC(%pM) vport(%d)",
629 "\tDELETE MC MAC: vport[%d] %pM fr(%p) refcnt(%d) uplinkfr(%p)\n",
630 vport, mac, vaddr->flow_rule, esw_mc->refcnt,
631 esw_mc->uplink_rule);
633 if (vaddr->flow_rule)
634 mlx5_del_flow_rule(vaddr->flow_rule);
635 vaddr->flow_rule = NULL;
637 /* If the multicast mac is added as a result of mc promiscuous vport,
638 * don't decrement the multicast ref count.
640 if (vaddr->mc_promisc || (--esw_mc->refcnt > 0))
643 /* Remove this multicast mac from all the mc promiscuous vports */
644 update_allmulti_vports(esw, vaddr, esw_mc);
646 if (esw_mc->uplink_rule)
647 mlx5_del_flow_rule(esw_mc->uplink_rule);
649 l2addr_hash_del(esw_mc);
653 /* Apply vport UC/MC list to HW l2 table and FDB table */
654 static void esw_apply_vport_addr_list(struct mlx5_eswitch *esw,
655 u32 vport_num, int list_type)
657 struct mlx5_vport *vport = &esw->vports[vport_num];
658 bool is_uc = list_type == MLX5_NVPRT_LIST_TYPE_UC;
659 vport_addr_action vport_addr_add;
660 vport_addr_action vport_addr_del;
661 struct vport_addr *addr;
662 struct l2addr_node *node;
663 struct hlist_head *hash;
664 struct hlist_node *tmp;
667 vport_addr_add = is_uc ? esw_add_uc_addr :
669 vport_addr_del = is_uc ? esw_del_uc_addr :
672 hash = is_uc ? vport->uc_list : vport->mc_list;
673 for_each_l2hash_node(node, tmp, hash, hi) {
674 addr = container_of(node, struct vport_addr, node);
675 switch (addr->action) {
676 case MLX5_ACTION_ADD:
677 vport_addr_add(esw, addr);
678 addr->action = MLX5_ACTION_NONE;
680 case MLX5_ACTION_DEL:
681 vport_addr_del(esw, addr);
682 l2addr_hash_del(addr);
688 /* Sync vport UC/MC list from vport context */
689 static void esw_update_vport_addr_list(struct mlx5_eswitch *esw,
690 u32 vport_num, int list_type)
692 struct mlx5_vport *vport = &esw->vports[vport_num];
693 bool is_uc = list_type == MLX5_NVPRT_LIST_TYPE_UC;
694 u8 (*mac_list)[ETH_ALEN];
695 struct l2addr_node *node;
696 struct vport_addr *addr;
697 struct hlist_head *hash;
698 struct hlist_node *tmp;
704 size = is_uc ? MLX5_MAX_UC_PER_VPORT(esw->dev) :
705 MLX5_MAX_MC_PER_VPORT(esw->dev);
707 mac_list = kcalloc(size, ETH_ALEN, GFP_KERNEL);
711 hash = is_uc ? vport->uc_list : vport->mc_list;
713 for_each_l2hash_node(node, tmp, hash, hi) {
714 addr = container_of(node, struct vport_addr, node);
715 addr->action = MLX5_ACTION_DEL;
721 err = mlx5_query_nic_vport_mac_list(esw->dev, vport_num, list_type,
725 esw_debug(esw->dev, "vport[%d] context update %s list size (%d)\n",
726 vport_num, is_uc ? "UC" : "MC", size);
728 for (i = 0; i < size; i++) {
729 if (is_uc && !is_valid_ether_addr(mac_list[i]))
732 if (!is_uc && !is_multicast_ether_addr(mac_list[i]))
735 addr = l2addr_hash_find(hash, mac_list[i], struct vport_addr);
737 addr->action = MLX5_ACTION_NONE;
738 /* If this mac was previously added because of allmulti
739 * promiscuous rx mode, its now converted to be original
742 if (addr->mc_promisc) {
743 struct esw_mc_addr *esw_mc =
744 l2addr_hash_find(esw->mc_table,
749 "Failed to MAC(%pM) in mcast DB\n",
754 addr->mc_promisc = false;
759 addr = l2addr_hash_add(hash, mac_list[i], struct vport_addr,
763 "Failed to add MAC(%pM) to vport[%d] DB\n",
764 mac_list[i], vport_num);
767 addr->vport = vport_num;
768 addr->action = MLX5_ACTION_ADD;
774 /* Sync vport UC/MC list from vport context
775 * Must be called after esw_update_vport_addr_list
777 static void esw_update_vport_mc_promisc(struct mlx5_eswitch *esw, u32 vport_num)
779 struct mlx5_vport *vport = &esw->vports[vport_num];
780 struct l2addr_node *node;
781 struct vport_addr *addr;
782 struct hlist_head *hash;
783 struct hlist_node *tmp;
786 hash = vport->mc_list;
788 for_each_l2hash_node(node, tmp, esw->mc_table, hi) {
789 u8 *mac = node->addr;
791 addr = l2addr_hash_find(hash, mac, struct vport_addr);
793 if (addr->action == MLX5_ACTION_DEL)
794 addr->action = MLX5_ACTION_NONE;
797 addr = l2addr_hash_add(hash, mac, struct vport_addr,
801 "Failed to add allmulti MAC(%pM) to vport[%d] DB\n",
805 addr->vport = vport_num;
806 addr->action = MLX5_ACTION_ADD;
807 addr->mc_promisc = true;
811 /* Apply vport rx mode to HW FDB table */
812 static void esw_apply_vport_rx_mode(struct mlx5_eswitch *esw, u32 vport_num,
813 bool promisc, bool mc_promisc)
815 struct esw_mc_addr *allmulti_addr = esw->mc_promisc;
816 struct mlx5_vport *vport = &esw->vports[vport_num];
818 if (IS_ERR_OR_NULL(vport->allmulti_rule) != mc_promisc)
822 vport->allmulti_rule =
823 esw_fdb_set_vport_allmulti_rule(esw, vport_num);
824 if (!allmulti_addr->uplink_rule)
825 allmulti_addr->uplink_rule =
826 esw_fdb_set_vport_allmulti_rule(esw,
828 allmulti_addr->refcnt++;
829 } else if (vport->allmulti_rule) {
830 mlx5_del_flow_rule(vport->allmulti_rule);
831 vport->allmulti_rule = NULL;
833 if (--allmulti_addr->refcnt > 0)
836 if (allmulti_addr->uplink_rule)
837 mlx5_del_flow_rule(allmulti_addr->uplink_rule);
838 allmulti_addr->uplink_rule = NULL;
842 if (IS_ERR_OR_NULL(vport->promisc_rule) != promisc)
846 vport->promisc_rule = esw_fdb_set_vport_promisc_rule(esw,
848 } else if (vport->promisc_rule) {
849 mlx5_del_flow_rule(vport->promisc_rule);
850 vport->promisc_rule = NULL;
854 /* Sync vport rx mode from vport context */
855 static void esw_update_vport_rx_mode(struct mlx5_eswitch *esw, u32 vport_num)
857 struct mlx5_vport *vport = &esw->vports[vport_num];
863 err = mlx5_query_nic_vport_promisc(esw->dev,
870 esw_debug(esw->dev, "vport[%d] context update rx mode promisc_all=%d, all_multi=%d\n",
871 vport_num, promisc_all, promisc_mc);
873 if (!vport->info.trusted || !vport->enabled) {
879 esw_apply_vport_rx_mode(esw, vport_num, promisc_all,
880 (promisc_all || promisc_mc));
883 static void esw_vport_change_handle_locked(struct mlx5_vport *vport)
885 struct mlx5_core_dev *dev = vport->dev;
886 struct mlx5_eswitch *esw = dev->priv.eswitch;
889 mlx5_query_nic_vport_mac_address(dev, vport->vport, mac);
890 esw_debug(dev, "vport[%d] Context Changed: perm mac: %pM\n",
893 if (vport->enabled_events & UC_ADDR_CHANGE) {
894 esw_update_vport_addr_list(esw, vport->vport,
895 MLX5_NVPRT_LIST_TYPE_UC);
896 esw_apply_vport_addr_list(esw, vport->vport,
897 MLX5_NVPRT_LIST_TYPE_UC);
900 if (vport->enabled_events & MC_ADDR_CHANGE) {
901 esw_update_vport_addr_list(esw, vport->vport,
902 MLX5_NVPRT_LIST_TYPE_MC);
905 if (vport->enabled_events & PROMISC_CHANGE) {
906 esw_update_vport_rx_mode(esw, vport->vport);
907 if (!IS_ERR_OR_NULL(vport->allmulti_rule))
908 esw_update_vport_mc_promisc(esw, vport->vport);
911 if (vport->enabled_events & (PROMISC_CHANGE | MC_ADDR_CHANGE)) {
912 esw_apply_vport_addr_list(esw, vport->vport,
913 MLX5_NVPRT_LIST_TYPE_MC);
916 esw_debug(esw->dev, "vport[%d] Context Changed: Done\n", vport->vport);
918 arm_vport_context_events_cmd(dev, vport->vport,
919 vport->enabled_events);
922 static void esw_vport_change_handler(struct work_struct *work)
924 struct mlx5_vport *vport =
925 container_of(work, struct mlx5_vport, vport_change_handler);
926 struct mlx5_eswitch *esw = vport->dev->priv.eswitch;
928 mutex_lock(&esw->state_lock);
929 esw_vport_change_handle_locked(vport);
930 mutex_unlock(&esw->state_lock);
933 static void esw_vport_enable_egress_acl(struct mlx5_eswitch *esw,
934 struct mlx5_vport *vport)
936 int inlen = MLX5_ST_SZ_BYTES(create_flow_group_in);
937 struct mlx5_flow_group *vlan_grp = NULL;
938 struct mlx5_flow_group *drop_grp = NULL;
939 struct mlx5_core_dev *dev = esw->dev;
940 struct mlx5_flow_namespace *root_ns;
941 struct mlx5_flow_table *acl;
942 void *match_criteria;
944 /* The egress acl table contains 2 rules:
945 * 1)Allow traffic with vlan_tag=vst_vlan_id
946 * 2)Drop all other traffic.
951 if (!MLX5_CAP_ESW_EGRESS_ACL(dev, ft_support) ||
952 !IS_ERR_OR_NULL(vport->egress.acl))
955 esw_debug(dev, "Create vport[%d] egress ACL log_max_size(%d)\n",
956 vport->vport, MLX5_CAP_ESW_EGRESS_ACL(dev, log_max_ft_size));
958 root_ns = mlx5_get_flow_namespace(dev, MLX5_FLOW_NAMESPACE_ESW_EGRESS);
960 esw_warn(dev, "Failed to get E-Switch egress flow namespace\n");
964 flow_group_in = mlx5_vzalloc(inlen);
968 acl = mlx5_create_vport_flow_table(root_ns, 0, table_size, 0, vport->vport);
971 esw_warn(dev, "Failed to create E-Switch vport[%d] egress flow Table, err(%d)\n",
976 MLX5_SET(create_flow_group_in, flow_group_in, match_criteria_enable, MLX5_MATCH_OUTER_HEADERS);
977 match_criteria = MLX5_ADDR_OF(create_flow_group_in, flow_group_in, match_criteria);
978 MLX5_SET_TO_ONES(fte_match_param, match_criteria, outer_headers.vlan_tag);
979 MLX5_SET_TO_ONES(fte_match_param, match_criteria, outer_headers.first_vid);
980 MLX5_SET(create_flow_group_in, flow_group_in, start_flow_index, 0);
981 MLX5_SET(create_flow_group_in, flow_group_in, end_flow_index, 0);
983 vlan_grp = mlx5_create_flow_group(acl, flow_group_in);
984 if (IS_ERR(vlan_grp)) {
985 err = PTR_ERR(vlan_grp);
986 esw_warn(dev, "Failed to create E-Switch vport[%d] egress allowed vlans flow group, err(%d)\n",
991 memset(flow_group_in, 0, inlen);
992 MLX5_SET(create_flow_group_in, flow_group_in, start_flow_index, 1);
993 MLX5_SET(create_flow_group_in, flow_group_in, end_flow_index, 1);
994 drop_grp = mlx5_create_flow_group(acl, flow_group_in);
995 if (IS_ERR(drop_grp)) {
996 err = PTR_ERR(drop_grp);
997 esw_warn(dev, "Failed to create E-Switch vport[%d] egress drop flow group, err(%d)\n",
1002 vport->egress.acl = acl;
1003 vport->egress.drop_grp = drop_grp;
1004 vport->egress.allowed_vlans_grp = vlan_grp;
1006 kvfree(flow_group_in);
1007 if (err && !IS_ERR_OR_NULL(vlan_grp))
1008 mlx5_destroy_flow_group(vlan_grp);
1009 if (err && !IS_ERR_OR_NULL(acl))
1010 mlx5_destroy_flow_table(acl);
1013 static void esw_vport_cleanup_egress_rules(struct mlx5_eswitch *esw,
1014 struct mlx5_vport *vport)
1016 if (!IS_ERR_OR_NULL(vport->egress.allowed_vlan))
1017 mlx5_del_flow_rule(vport->egress.allowed_vlan);
1019 if (!IS_ERR_OR_NULL(vport->egress.drop_rule))
1020 mlx5_del_flow_rule(vport->egress.drop_rule);
1022 vport->egress.allowed_vlan = NULL;
1023 vport->egress.drop_rule = NULL;
1026 static void esw_vport_disable_egress_acl(struct mlx5_eswitch *esw,
1027 struct mlx5_vport *vport)
1029 if (IS_ERR_OR_NULL(vport->egress.acl))
1032 esw_debug(esw->dev, "Destroy vport[%d] E-Switch egress ACL\n", vport->vport);
1034 esw_vport_cleanup_egress_rules(esw, vport);
1035 mlx5_destroy_flow_group(vport->egress.allowed_vlans_grp);
1036 mlx5_destroy_flow_group(vport->egress.drop_grp);
1037 mlx5_destroy_flow_table(vport->egress.acl);
1038 vport->egress.allowed_vlans_grp = NULL;
1039 vport->egress.drop_grp = NULL;
1040 vport->egress.acl = NULL;
1043 static void esw_vport_enable_ingress_acl(struct mlx5_eswitch *esw,
1044 struct mlx5_vport *vport)
1046 int inlen = MLX5_ST_SZ_BYTES(create_flow_group_in);
1047 struct mlx5_core_dev *dev = esw->dev;
1048 struct mlx5_flow_namespace *root_ns;
1049 struct mlx5_flow_table *acl;
1050 struct mlx5_flow_group *g;
1051 void *match_criteria;
1053 /* The ingress acl table contains 4 groups
1054 * (2 active rules at the same time -
1055 * 1 allow rule from one of the first 3 groups.
1056 * 1 drop rule from the last group):
1057 * 1)Allow untagged traffic with smac=original mac.
1058 * 2)Allow untagged traffic.
1059 * 3)Allow traffic with smac=original mac.
1060 * 4)Drop all other traffic.
1065 if (!MLX5_CAP_ESW_INGRESS_ACL(dev, ft_support) ||
1066 !IS_ERR_OR_NULL(vport->ingress.acl))
1069 esw_debug(dev, "Create vport[%d] ingress ACL log_max_size(%d)\n",
1070 vport->vport, MLX5_CAP_ESW_INGRESS_ACL(dev, log_max_ft_size));
1072 root_ns = mlx5_get_flow_namespace(dev, MLX5_FLOW_NAMESPACE_ESW_INGRESS);
1074 esw_warn(dev, "Failed to get E-Switch ingress flow namespace\n");
1078 flow_group_in = mlx5_vzalloc(inlen);
1082 acl = mlx5_create_vport_flow_table(root_ns, 0, table_size, 0, vport->vport);
1085 esw_warn(dev, "Failed to create E-Switch vport[%d] ingress flow Table, err(%d)\n",
1089 vport->ingress.acl = acl;
1091 match_criteria = MLX5_ADDR_OF(create_flow_group_in, flow_group_in, match_criteria);
1093 MLX5_SET(create_flow_group_in, flow_group_in, match_criteria_enable, MLX5_MATCH_OUTER_HEADERS);
1094 MLX5_SET_TO_ONES(fte_match_param, match_criteria, outer_headers.vlan_tag);
1095 MLX5_SET_TO_ONES(fte_match_param, match_criteria, outer_headers.smac_47_16);
1096 MLX5_SET_TO_ONES(fte_match_param, match_criteria, outer_headers.smac_15_0);
1097 MLX5_SET(create_flow_group_in, flow_group_in, start_flow_index, 0);
1098 MLX5_SET(create_flow_group_in, flow_group_in, end_flow_index, 0);
1100 g = mlx5_create_flow_group(acl, flow_group_in);
1103 esw_warn(dev, "Failed to create E-Switch vport[%d] ingress untagged spoofchk flow group, err(%d)\n",
1107 vport->ingress.allow_untagged_spoofchk_grp = g;
1109 memset(flow_group_in, 0, inlen);
1110 MLX5_SET(create_flow_group_in, flow_group_in, match_criteria_enable, MLX5_MATCH_OUTER_HEADERS);
1111 MLX5_SET_TO_ONES(fte_match_param, match_criteria, outer_headers.vlan_tag);
1112 MLX5_SET(create_flow_group_in, flow_group_in, start_flow_index, 1);
1113 MLX5_SET(create_flow_group_in, flow_group_in, end_flow_index, 1);
1115 g = mlx5_create_flow_group(acl, flow_group_in);
1118 esw_warn(dev, "Failed to create E-Switch vport[%d] ingress untagged flow group, err(%d)\n",
1122 vport->ingress.allow_untagged_only_grp = g;
1124 memset(flow_group_in, 0, inlen);
1125 MLX5_SET(create_flow_group_in, flow_group_in, match_criteria_enable, MLX5_MATCH_OUTER_HEADERS);
1126 MLX5_SET_TO_ONES(fte_match_param, match_criteria, outer_headers.smac_47_16);
1127 MLX5_SET_TO_ONES(fte_match_param, match_criteria, outer_headers.smac_15_0);
1128 MLX5_SET(create_flow_group_in, flow_group_in, start_flow_index, 2);
1129 MLX5_SET(create_flow_group_in, flow_group_in, end_flow_index, 2);
1131 g = mlx5_create_flow_group(acl, flow_group_in);
1134 esw_warn(dev, "Failed to create E-Switch vport[%d] ingress spoofchk flow group, err(%d)\n",
1138 vport->ingress.allow_spoofchk_only_grp = g;
1140 memset(flow_group_in, 0, inlen);
1141 MLX5_SET(create_flow_group_in, flow_group_in, start_flow_index, 3);
1142 MLX5_SET(create_flow_group_in, flow_group_in, end_flow_index, 3);
1144 g = mlx5_create_flow_group(acl, flow_group_in);
1147 esw_warn(dev, "Failed to create E-Switch vport[%d] ingress drop flow group, err(%d)\n",
1151 vport->ingress.drop_grp = g;
1155 if (!IS_ERR_OR_NULL(vport->ingress.allow_spoofchk_only_grp))
1156 mlx5_destroy_flow_group(
1157 vport->ingress.allow_spoofchk_only_grp);
1158 if (!IS_ERR_OR_NULL(vport->ingress.allow_untagged_only_grp))
1159 mlx5_destroy_flow_group(
1160 vport->ingress.allow_untagged_only_grp);
1161 if (!IS_ERR_OR_NULL(vport->ingress.allow_untagged_spoofchk_grp))
1162 mlx5_destroy_flow_group(
1163 vport->ingress.allow_untagged_spoofchk_grp);
1164 if (!IS_ERR_OR_NULL(vport->ingress.acl))
1165 mlx5_destroy_flow_table(vport->ingress.acl);
1168 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_rule(vport->ingress.drop_rule);
1177 if (!IS_ERR_OR_NULL(vport->ingress.allow_rule))
1178 mlx5_del_flow_rule(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_flow_spec *spec;
1212 if (vport->info.spoofchk && !is_valid_ether_addr(vport->info.mac)) {
1213 mlx5_core_warn(esw->dev,
1214 "vport[%d] configure ingress rules failed, illegal mac with spoofchk\n",
1220 esw_vport_cleanup_ingress_rules(esw, vport);
1222 if (!vport->info.vlan && !vport->info.qos && !vport->info.spoofchk) {
1223 esw_vport_disable_ingress_acl(esw, vport);
1227 esw_vport_enable_ingress_acl(esw, vport);
1230 "vport[%d] configure ingress rules, vlan(%d) qos(%d)\n",
1231 vport->vport, vport->info.vlan, vport->info.qos);
1233 spec = mlx5_vzalloc(sizeof(*spec));
1236 esw_warn(esw->dev, "vport[%d] configure ingress rules failed, err(%d)\n",
1241 if (vport->info.vlan || vport->info.qos)
1242 MLX5_SET_TO_ONES(fte_match_param, spec->match_criteria, outer_headers.vlan_tag);
1244 if (vport->info.spoofchk) {
1245 MLX5_SET_TO_ONES(fte_match_param, spec->match_criteria, outer_headers.smac_47_16);
1246 MLX5_SET_TO_ONES(fte_match_param, spec->match_criteria, outer_headers.smac_15_0);
1247 smac_v = MLX5_ADDR_OF(fte_match_param,
1249 outer_headers.smac_47_16);
1250 ether_addr_copy(smac_v, vport->info.mac);
1253 spec->match_criteria_enable = MLX5_MATCH_OUTER_HEADERS;
1254 vport->ingress.allow_rule =
1255 mlx5_add_flow_rule(vport->ingress.acl, spec,
1256 MLX5_FLOW_CONTEXT_ACTION_ALLOW,
1258 if (IS_ERR(vport->ingress.allow_rule)) {
1259 err = PTR_ERR(vport->ingress.allow_rule);
1261 "vport[%d] configure ingress allow rule, err(%d)\n",
1263 vport->ingress.allow_rule = NULL;
1267 memset(spec, 0, sizeof(*spec));
1268 vport->ingress.drop_rule =
1269 mlx5_add_flow_rule(vport->ingress.acl, spec,
1270 MLX5_FLOW_CONTEXT_ACTION_DROP,
1272 if (IS_ERR(vport->ingress.drop_rule)) {
1273 err = PTR_ERR(vport->ingress.drop_rule);
1275 "vport[%d] configure ingress drop rule, err(%d)\n",
1277 vport->ingress.drop_rule = NULL;
1283 esw_vport_cleanup_ingress_rules(esw, vport);
1288 static int esw_vport_egress_config(struct mlx5_eswitch *esw,
1289 struct mlx5_vport *vport)
1291 struct mlx5_flow_spec *spec;
1294 esw_vport_cleanup_egress_rules(esw, vport);
1296 if (!vport->info.vlan && !vport->info.qos) {
1297 esw_vport_disable_egress_acl(esw, vport);
1301 esw_vport_enable_egress_acl(esw, vport);
1304 "vport[%d] configure egress rules, vlan(%d) qos(%d)\n",
1305 vport->vport, vport->info.vlan, vport->info.qos);
1307 spec = mlx5_vzalloc(sizeof(*spec));
1310 esw_warn(esw->dev, "vport[%d] configure egress rules failed, err(%d)\n",
1315 /* Allowed vlan rule */
1316 MLX5_SET_TO_ONES(fte_match_param, spec->match_criteria, outer_headers.vlan_tag);
1317 MLX5_SET_TO_ONES(fte_match_param, spec->match_value, outer_headers.vlan_tag);
1318 MLX5_SET_TO_ONES(fte_match_param, spec->match_criteria, outer_headers.first_vid);
1319 MLX5_SET(fte_match_param, spec->match_value, outer_headers.first_vid, vport->info.vlan);
1321 spec->match_criteria_enable = MLX5_MATCH_OUTER_HEADERS;
1322 vport->egress.allowed_vlan =
1323 mlx5_add_flow_rule(vport->egress.acl, spec,
1324 MLX5_FLOW_CONTEXT_ACTION_ALLOW,
1326 if (IS_ERR(vport->egress.allowed_vlan)) {
1327 err = PTR_ERR(vport->egress.allowed_vlan);
1329 "vport[%d] configure egress allowed vlan rule failed, err(%d)\n",
1331 vport->egress.allowed_vlan = NULL;
1335 /* Drop others rule (star rule) */
1336 memset(spec, 0, sizeof(*spec));
1337 vport->egress.drop_rule =
1338 mlx5_add_flow_rule(vport->egress.acl, spec,
1339 MLX5_FLOW_CONTEXT_ACTION_DROP,
1341 if (IS_ERR(vport->egress.drop_rule)) {
1342 err = PTR_ERR(vport->egress.drop_rule);
1344 "vport[%d] configure egress drop rule failed, err(%d)\n",
1346 vport->egress.drop_rule = NULL;
1353 static void node_guid_gen_from_mac(u64 *node_guid, u8 mac[ETH_ALEN])
1355 ((u8 *)node_guid)[7] = mac[0];
1356 ((u8 *)node_guid)[6] = mac[1];
1357 ((u8 *)node_guid)[5] = mac[2];
1358 ((u8 *)node_guid)[4] = 0xff;
1359 ((u8 *)node_guid)[3] = 0xfe;
1360 ((u8 *)node_guid)[2] = mac[3];
1361 ((u8 *)node_guid)[1] = mac[4];
1362 ((u8 *)node_guid)[0] = mac[5];
1365 static void esw_apply_vport_conf(struct mlx5_eswitch *esw,
1366 struct mlx5_vport *vport)
1368 int vport_num = vport->vport;
1373 mlx5_modify_vport_admin_state(esw->dev,
1374 MLX5_QUERY_VPORT_STATE_IN_OP_MOD_ESW_VPORT,
1376 vport->info.link_state);
1377 mlx5_modify_nic_vport_mac_address(esw->dev, vport_num, vport->info.mac);
1378 mlx5_modify_nic_vport_node_guid(esw->dev, vport_num, vport->info.node_guid);
1379 modify_esw_vport_cvlan(esw->dev, vport_num, vport->info.vlan, vport->info.qos,
1380 (vport->info.vlan || vport->info.qos));
1382 /* Only legacy mode needs ACLs */
1383 if (esw->mode == SRIOV_LEGACY) {
1384 esw_vport_ingress_config(esw, vport);
1385 esw_vport_egress_config(esw, vport);
1388 static void esw_enable_vport(struct mlx5_eswitch *esw, int vport_num,
1391 struct mlx5_vport *vport = &esw->vports[vport_num];
1393 mutex_lock(&esw->state_lock);
1394 WARN_ON(vport->enabled);
1396 esw_debug(esw->dev, "Enabling VPORT(%d)\n", vport_num);
1398 /* Restore old vport configuration */
1399 esw_apply_vport_conf(esw, vport);
1401 /* Sync with current vport context */
1402 vport->enabled_events = enable_events;
1403 vport->enabled = true;
1405 /* only PF is trusted by default */
1407 vport->info.trusted = true;
1409 esw_vport_change_handle_locked(vport);
1411 esw->enabled_vports++;
1412 esw_debug(esw->dev, "Enabled VPORT(%d)\n", vport_num);
1413 mutex_unlock(&esw->state_lock);
1416 static void esw_disable_vport(struct mlx5_eswitch *esw, int vport_num)
1418 struct mlx5_vport *vport = &esw->vports[vport_num];
1420 if (!vport->enabled)
1423 esw_debug(esw->dev, "Disabling vport(%d)\n", vport_num);
1424 /* Mark this vport as disabled to discard new events */
1425 vport->enabled = false;
1427 synchronize_irq(mlx5_get_msix_vec(esw->dev, MLX5_EQ_VEC_ASYNC));
1428 /* Wait for current already scheduled events to complete */
1429 flush_workqueue(esw->work_queue);
1430 /* Disable events from this vport */
1431 arm_vport_context_events_cmd(esw->dev, vport->vport, 0);
1432 mutex_lock(&esw->state_lock);
1433 /* We don't assume VFs will cleanup after themselves.
1434 * Calling vport change handler while vport is disabled will cleanup
1435 * the vport resources.
1437 esw_vport_change_handle_locked(vport);
1438 vport->enabled_events = 0;
1440 if (vport_num && esw->mode == SRIOV_LEGACY) {
1441 mlx5_modify_vport_admin_state(esw->dev,
1442 MLX5_QUERY_VPORT_STATE_IN_OP_MOD_ESW_VPORT,
1444 MLX5_ESW_VPORT_ADMIN_STATE_DOWN);
1445 esw_vport_disable_egress_acl(esw, vport);
1446 esw_vport_disable_ingress_acl(esw, vport);
1448 esw->enabled_vports--;
1449 mutex_unlock(&esw->state_lock);
1452 /* Public E-Switch API */
1453 int mlx5_eswitch_enable_sriov(struct mlx5_eswitch *esw, int nvfs, int mode)
1456 int i, enabled_events;
1458 if (!esw || !MLX5_CAP_GEN(esw->dev, vport_group_manager) ||
1459 MLX5_CAP_GEN(esw->dev, port_type) != MLX5_CAP_PORT_TYPE_ETH)
1462 if (!MLX5_CAP_GEN(esw->dev, eswitch_flow_table) ||
1463 !MLX5_CAP_ESW_FLOWTABLE_FDB(esw->dev, ft_support)) {
1464 esw_warn(esw->dev, "E-Switch FDB is not supported, aborting ...\n");
1468 if (!MLX5_CAP_ESW_INGRESS_ACL(esw->dev, ft_support))
1469 esw_warn(esw->dev, "E-Switch ingress ACL is not supported by FW\n");
1471 if (!MLX5_CAP_ESW_EGRESS_ACL(esw->dev, ft_support))
1472 esw_warn(esw->dev, "E-Switch engress ACL is not supported by FW\n");
1474 esw_info(esw->dev, "E-Switch enable SRIOV: nvfs(%d) mode (%d)\n", nvfs, mode);
1476 esw_disable_vport(esw, 0);
1478 if (mode == SRIOV_LEGACY)
1479 err = esw_create_legacy_fdb_table(esw, nvfs + 1);
1481 err = esw_offloads_init(esw, nvfs + 1);
1485 enabled_events = (mode == SRIOV_LEGACY) ? SRIOV_VPORT_EVENTS : UC_ADDR_CHANGE;
1486 for (i = 0; i <= nvfs; i++)
1487 esw_enable_vport(esw, i, enabled_events);
1489 esw_info(esw->dev, "SRIOV enabled: active vports(%d)\n",
1490 esw->enabled_vports);
1494 esw_enable_vport(esw, 0, UC_ADDR_CHANGE);
1498 void mlx5_eswitch_disable_sriov(struct mlx5_eswitch *esw)
1500 struct esw_mc_addr *mc_promisc;
1504 if (!esw || !MLX5_CAP_GEN(esw->dev, vport_group_manager) ||
1505 MLX5_CAP_GEN(esw->dev, port_type) != MLX5_CAP_PORT_TYPE_ETH)
1508 esw_info(esw->dev, "disable SRIOV: active vports(%d) mode(%d)\n",
1509 esw->enabled_vports, esw->mode);
1511 mc_promisc = esw->mc_promisc;
1512 nvports = esw->enabled_vports;
1514 for (i = 0; i < esw->total_vports; i++)
1515 esw_disable_vport(esw, i);
1517 if (mc_promisc && mc_promisc->uplink_rule)
1518 mlx5_del_flow_rule(mc_promisc->uplink_rule);
1520 if (esw->mode == SRIOV_LEGACY)
1521 esw_destroy_legacy_fdb_table(esw);
1522 else if (esw->mode == SRIOV_OFFLOADS)
1523 esw_offloads_cleanup(esw, nvports);
1525 esw->mode = SRIOV_NONE;
1526 /* VPORT 0 (PF) must be enabled back with non-sriov configuration */
1527 esw_enable_vport(esw, 0, UC_ADDR_CHANGE);
1530 void mlx5_eswitch_attach(struct mlx5_eswitch *esw)
1532 if (!esw || !MLX5_CAP_GEN(esw->dev, vport_group_manager) ||
1533 MLX5_CAP_GEN(esw->dev, port_type) != MLX5_CAP_PORT_TYPE_ETH)
1536 esw_enable_vport(esw, 0, UC_ADDR_CHANGE);
1537 /* VF Vports will be enabled when SRIOV is enabled */
1540 void mlx5_eswitch_detach(struct mlx5_eswitch *esw)
1542 if (!esw || !MLX5_CAP_GEN(esw->dev, vport_group_manager) ||
1543 MLX5_CAP_GEN(esw->dev, port_type) != MLX5_CAP_PORT_TYPE_ETH)
1546 esw_disable_vport(esw, 0);
1549 int mlx5_eswitch_init(struct mlx5_core_dev *dev)
1551 int l2_table_size = 1 << MLX5_CAP_GEN(dev, log_max_l2_table);
1552 int total_vports = MLX5_TOTAL_VPORTS(dev);
1553 struct esw_mc_addr *mc_promisc;
1554 struct mlx5_eswitch *esw;
1558 if (!MLX5_CAP_GEN(dev, vport_group_manager) ||
1559 MLX5_CAP_GEN(dev, port_type) != MLX5_CAP_PORT_TYPE_ETH)
1563 "Total vports %d, l2 table size(%d), per vport: max uc(%d) max mc(%d)\n",
1564 total_vports, l2_table_size,
1565 MLX5_MAX_UC_PER_VPORT(dev),
1566 MLX5_MAX_MC_PER_VPORT(dev));
1568 esw = kzalloc(sizeof(*esw), GFP_KERNEL);
1574 esw->l2_table.bitmap = kcalloc(BITS_TO_LONGS(l2_table_size),
1575 sizeof(uintptr_t), GFP_KERNEL);
1576 if (!esw->l2_table.bitmap) {
1580 esw->l2_table.size = l2_table_size;
1582 mc_promisc = kzalloc(sizeof(*mc_promisc), GFP_KERNEL);
1587 esw->mc_promisc = mc_promisc;
1589 esw->work_queue = create_singlethread_workqueue("mlx5_esw_wq");
1590 if (!esw->work_queue) {
1595 esw->vports = kcalloc(total_vports, sizeof(struct mlx5_vport),
1602 esw->offloads.vport_reps =
1603 kzalloc(total_vports * sizeof(struct mlx5_eswitch_rep),
1605 if (!esw->offloads.vport_reps) {
1610 mutex_init(&esw->state_lock);
1612 for (vport_num = 0; vport_num < total_vports; vport_num++) {
1613 struct mlx5_vport *vport = &esw->vports[vport_num];
1615 vport->vport = vport_num;
1616 vport->info.link_state = MLX5_ESW_VPORT_ADMIN_STATE_AUTO;
1618 INIT_WORK(&vport->vport_change_handler,
1619 esw_vport_change_handler);
1622 esw->total_vports = total_vports;
1623 esw->enabled_vports = 0;
1624 esw->mode = SRIOV_NONE;
1626 dev->priv.eswitch = esw;
1629 if (esw->work_queue)
1630 destroy_workqueue(esw->work_queue);
1631 kfree(esw->l2_table.bitmap);
1633 kfree(esw->offloads.vport_reps);
1638 void mlx5_eswitch_cleanup(struct mlx5_eswitch *esw)
1640 if (!esw || !MLX5_CAP_GEN(esw->dev, vport_group_manager) ||
1641 MLX5_CAP_GEN(esw->dev, port_type) != MLX5_CAP_PORT_TYPE_ETH)
1644 esw_info(esw->dev, "cleanup\n");
1646 esw->dev->priv.eswitch = NULL;
1647 destroy_workqueue(esw->work_queue);
1648 kfree(esw->l2_table.bitmap);
1649 kfree(esw->mc_promisc);
1650 kfree(esw->offloads.vport_reps);
1655 void mlx5_eswitch_vport_event(struct mlx5_eswitch *esw, struct mlx5_eqe *eqe)
1657 struct mlx5_eqe_vport_change *vc_eqe = &eqe->data.vport_change;
1658 u16 vport_num = be16_to_cpu(vc_eqe->vport_num);
1659 struct mlx5_vport *vport;
1662 pr_warn("MLX5 E-Switch: vport %d got an event while eswitch is not initialized\n",
1667 vport = &esw->vports[vport_num];
1669 queue_work(esw->work_queue, &vport->vport_change_handler);
1672 /* Vport Administration */
1673 #define ESW_ALLOWED(esw) \
1674 (esw && MLX5_CAP_GEN(esw->dev, vport_group_manager) && mlx5_core_is_pf(esw->dev))
1675 #define LEGAL_VPORT(esw, vport) (vport >= 0 && vport < esw->total_vports)
1677 int mlx5_eswitch_set_vport_mac(struct mlx5_eswitch *esw,
1678 int vport, u8 mac[ETH_ALEN])
1680 struct mlx5_vport *evport;
1684 if (!ESW_ALLOWED(esw))
1686 if (!LEGAL_VPORT(esw, vport))
1689 mutex_lock(&esw->state_lock);
1690 evport = &esw->vports[vport];
1692 if (evport->info.spoofchk && !is_valid_ether_addr(mac)) {
1693 mlx5_core_warn(esw->dev,
1694 "MAC invalidation is not allowed when spoofchk is on, vport(%d)\n",
1700 err = mlx5_modify_nic_vport_mac_address(esw->dev, vport, mac);
1702 mlx5_core_warn(esw->dev,
1703 "Failed to mlx5_modify_nic_vport_mac vport(%d) err=(%d)\n",
1708 node_guid_gen_from_mac(&node_guid, mac);
1709 err = mlx5_modify_nic_vport_node_guid(esw->dev, vport, node_guid);
1711 mlx5_core_warn(esw->dev,
1712 "Failed to set vport %d node guid, err = %d. RDMA_CM will not function properly for this VF.\n",
1715 ether_addr_copy(evport->info.mac, mac);
1716 evport->info.node_guid = node_guid;
1717 if (evport->enabled && esw->mode == SRIOV_LEGACY)
1718 err = esw_vport_ingress_config(esw, evport);
1721 mutex_unlock(&esw->state_lock);
1725 int mlx5_eswitch_set_vport_state(struct mlx5_eswitch *esw,
1726 int vport, int link_state)
1728 struct mlx5_vport *evport;
1731 if (!ESW_ALLOWED(esw))
1733 if (!LEGAL_VPORT(esw, vport))
1736 mutex_lock(&esw->state_lock);
1737 evport = &esw->vports[vport];
1739 err = mlx5_modify_vport_admin_state(esw->dev,
1740 MLX5_QUERY_VPORT_STATE_IN_OP_MOD_ESW_VPORT,
1743 mlx5_core_warn(esw->dev,
1744 "Failed to set vport %d link state, err = %d",
1749 evport->info.link_state = link_state;
1752 mutex_unlock(&esw->state_lock);
1756 int mlx5_eswitch_get_vport_config(struct mlx5_eswitch *esw,
1757 int vport, struct ifla_vf_info *ivi)
1759 struct mlx5_vport *evport;
1761 if (!ESW_ALLOWED(esw))
1763 if (!LEGAL_VPORT(esw, vport))
1766 evport = &esw->vports[vport];
1768 memset(ivi, 0, sizeof(*ivi));
1769 ivi->vf = vport - 1;
1771 mutex_lock(&esw->state_lock);
1772 ether_addr_copy(ivi->mac, evport->info.mac);
1773 ivi->linkstate = evport->info.link_state;
1774 ivi->vlan = evport->info.vlan;
1775 ivi->qos = evport->info.qos;
1776 ivi->spoofchk = evport->info.spoofchk;
1777 ivi->trusted = evport->info.trusted;
1778 mutex_unlock(&esw->state_lock);
1783 int mlx5_eswitch_set_vport_vlan(struct mlx5_eswitch *esw,
1784 int vport, u16 vlan, u8 qos)
1786 struct mlx5_vport *evport;
1790 if (!ESW_ALLOWED(esw))
1792 if (!LEGAL_VPORT(esw, vport) || (vlan > 4095) || (qos > 7))
1798 mutex_lock(&esw->state_lock);
1799 evport = &esw->vports[vport];
1801 err = modify_esw_vport_cvlan(esw->dev, vport, vlan, qos, set);
1805 evport->info.vlan = vlan;
1806 evport->info.qos = qos;
1807 if (evport->enabled && esw->mode == SRIOV_LEGACY) {
1808 err = esw_vport_ingress_config(esw, evport);
1811 err = esw_vport_egress_config(esw, evport);
1815 mutex_unlock(&esw->state_lock);
1819 int mlx5_eswitch_set_vport_spoofchk(struct mlx5_eswitch *esw,
1820 int vport, bool spoofchk)
1822 struct mlx5_vport *evport;
1826 if (!ESW_ALLOWED(esw))
1828 if (!LEGAL_VPORT(esw, vport))
1831 mutex_lock(&esw->state_lock);
1832 evport = &esw->vports[vport];
1833 pschk = evport->info.spoofchk;
1834 evport->info.spoofchk = spoofchk;
1835 if (evport->enabled && esw->mode == SRIOV_LEGACY)
1836 err = esw_vport_ingress_config(esw, evport);
1838 evport->info.spoofchk = pschk;
1839 mutex_unlock(&esw->state_lock);
1844 int mlx5_eswitch_set_vport_trust(struct mlx5_eswitch *esw,
1845 int vport, bool setting)
1847 struct mlx5_vport *evport;
1849 if (!ESW_ALLOWED(esw))
1851 if (!LEGAL_VPORT(esw, vport))
1854 mutex_lock(&esw->state_lock);
1855 evport = &esw->vports[vport];
1856 evport->info.trusted = setting;
1857 if (evport->enabled)
1858 esw_vport_change_handle_locked(evport);
1859 mutex_unlock(&esw->state_lock);
1864 int mlx5_eswitch_get_vport_stats(struct mlx5_eswitch *esw,
1866 struct ifla_vf_stats *vf_stats)
1868 int outlen = MLX5_ST_SZ_BYTES(query_vport_counter_out);
1869 u32 in[MLX5_ST_SZ_DW(query_vport_counter_in)] = {0};
1873 if (!ESW_ALLOWED(esw))
1875 if (!LEGAL_VPORT(esw, vport))
1878 out = mlx5_vzalloc(outlen);
1882 MLX5_SET(query_vport_counter_in, in, opcode,
1883 MLX5_CMD_OP_QUERY_VPORT_COUNTER);
1884 MLX5_SET(query_vport_counter_in, in, op_mod, 0);
1885 MLX5_SET(query_vport_counter_in, in, vport_number, vport);
1887 MLX5_SET(query_vport_counter_in, in, other_vport, 1);
1889 memset(out, 0, outlen);
1890 err = mlx5_cmd_exec(esw->dev, in, sizeof(in), out, outlen);
1894 #define MLX5_GET_CTR(p, x) \
1895 MLX5_GET64(query_vport_counter_out, p, x)
1897 memset(vf_stats, 0, sizeof(*vf_stats));
1898 vf_stats->rx_packets =
1899 MLX5_GET_CTR(out, received_eth_unicast.packets) +
1900 MLX5_GET_CTR(out, received_eth_multicast.packets) +
1901 MLX5_GET_CTR(out, received_eth_broadcast.packets);
1903 vf_stats->rx_bytes =
1904 MLX5_GET_CTR(out, received_eth_unicast.octets) +
1905 MLX5_GET_CTR(out, received_eth_multicast.octets) +
1906 MLX5_GET_CTR(out, received_eth_broadcast.octets);
1908 vf_stats->tx_packets =
1909 MLX5_GET_CTR(out, transmitted_eth_unicast.packets) +
1910 MLX5_GET_CTR(out, transmitted_eth_multicast.packets) +
1911 MLX5_GET_CTR(out, transmitted_eth_broadcast.packets);
1913 vf_stats->tx_bytes =
1914 MLX5_GET_CTR(out, transmitted_eth_unicast.octets) +
1915 MLX5_GET_CTR(out, transmitted_eth_multicast.octets) +
1916 MLX5_GET_CTR(out, transmitted_eth_broadcast.octets);
1918 vf_stats->multicast =
1919 MLX5_GET_CTR(out, received_eth_multicast.packets);
1921 vf_stats->broadcast =
1922 MLX5_GET_CTR(out, received_eth_broadcast.packets);