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_create_offloads_fdb_table(struct mlx5_eswitch *esw, int nvports);
85 void esw_destroy_offloads_fdb_table(struct mlx5_eswitch *esw);
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)];
91 int out[MLX5_ST_SZ_DW(modify_nic_vport_context_out)];
95 memset(out, 0, sizeof(out));
96 memset(in, 0, sizeof(in));
98 MLX5_SET(modify_nic_vport_context_in, in,
99 opcode, MLX5_CMD_OP_MODIFY_NIC_VPORT_CONTEXT);
100 MLX5_SET(modify_nic_vport_context_in, in, field_select.change_event, 1);
101 MLX5_SET(modify_nic_vport_context_in, in, vport_number, vport);
103 MLX5_SET(modify_nic_vport_context_in, in, other_vport, 1);
104 nic_vport_ctx = MLX5_ADDR_OF(modify_nic_vport_context_in,
105 in, nic_vport_context);
107 MLX5_SET(nic_vport_context, nic_vport_ctx, arm_change_event, 1);
109 if (events_mask & UC_ADDR_CHANGE)
110 MLX5_SET(nic_vport_context, nic_vport_ctx,
111 event_on_uc_address_change, 1);
112 if (events_mask & MC_ADDR_CHANGE)
113 MLX5_SET(nic_vport_context, nic_vport_ctx,
114 event_on_mc_address_change, 1);
115 if (events_mask & PROMISC_CHANGE)
116 MLX5_SET(nic_vport_context, nic_vport_ctx,
117 event_on_promisc_change, 1);
119 err = mlx5_cmd_exec(dev, in, sizeof(in), out, sizeof(out));
122 err = mlx5_cmd_status_to_err_v2(out);
130 /* E-Switch vport context HW commands */
131 static int query_esw_vport_context_cmd(struct mlx5_core_dev *mdev, u32 vport,
132 u32 *out, int outlen)
134 u32 in[MLX5_ST_SZ_DW(query_esw_vport_context_in)];
136 memset(in, 0, sizeof(in));
138 MLX5_SET(query_nic_vport_context_in, in, opcode,
139 MLX5_CMD_OP_QUERY_ESW_VPORT_CONTEXT);
141 MLX5_SET(query_esw_vport_context_in, in, vport_number, vport);
143 MLX5_SET(query_esw_vport_context_in, in, other_vport, 1);
145 return mlx5_cmd_exec_check_status(mdev, in, sizeof(in), out, outlen);
148 static int query_esw_vport_cvlan(struct mlx5_core_dev *dev, u32 vport,
151 u32 out[MLX5_ST_SZ_DW(query_esw_vport_context_out)];
156 memset(out, 0, sizeof(out));
161 if (!MLX5_CAP_ESW(dev, vport_cvlan_strip) ||
162 !MLX5_CAP_ESW(dev, vport_cvlan_insert_if_not_exist))
165 err = query_esw_vport_context_cmd(dev, vport, out, sizeof(out));
169 cvlan_strip = MLX5_GET(query_esw_vport_context_out, out,
170 esw_vport_context.vport_cvlan_strip);
172 cvlan_insert = MLX5_GET(query_esw_vport_context_out, out,
173 esw_vport_context.vport_cvlan_insert);
175 if (cvlan_strip || cvlan_insert) {
176 *vlan = MLX5_GET(query_esw_vport_context_out, out,
177 esw_vport_context.cvlan_id);
178 *qos = MLX5_GET(query_esw_vport_context_out, out,
179 esw_vport_context.cvlan_pcp);
182 esw_debug(dev, "Query Vport[%d] cvlan: VLAN %d qos=%d\n",
188 static int modify_esw_vport_context_cmd(struct mlx5_core_dev *dev, u16 vport,
191 u32 out[MLX5_ST_SZ_DW(modify_esw_vport_context_out)];
193 memset(out, 0, sizeof(out));
195 MLX5_SET(modify_esw_vport_context_in, in, vport_number, vport);
197 MLX5_SET(modify_esw_vport_context_in, in, other_vport, 1);
199 MLX5_SET(modify_esw_vport_context_in, in, opcode,
200 MLX5_CMD_OP_MODIFY_ESW_VPORT_CONTEXT);
202 return mlx5_cmd_exec_check_status(dev, in, inlen,
206 static int modify_esw_vport_cvlan(struct mlx5_core_dev *dev, u32 vport,
207 u16 vlan, u8 qos, bool set)
209 u32 in[MLX5_ST_SZ_DW(modify_esw_vport_context_in)];
211 memset(in, 0, sizeof(in));
213 if (!MLX5_CAP_ESW(dev, vport_cvlan_strip) ||
214 !MLX5_CAP_ESW(dev, vport_cvlan_insert_if_not_exist))
217 esw_debug(dev, "Set Vport[%d] VLAN %d qos %d set=%d\n",
218 vport, vlan, qos, set);
221 MLX5_SET(modify_esw_vport_context_in, in,
222 esw_vport_context.vport_cvlan_strip, 1);
223 /* insert only if no vlan in packet */
224 MLX5_SET(modify_esw_vport_context_in, in,
225 esw_vport_context.vport_cvlan_insert, 1);
226 MLX5_SET(modify_esw_vport_context_in, in,
227 esw_vport_context.cvlan_pcp, qos);
228 MLX5_SET(modify_esw_vport_context_in, in,
229 esw_vport_context.cvlan_id, vlan);
232 MLX5_SET(modify_esw_vport_context_in, in,
233 field_select.vport_cvlan_strip, 1);
234 MLX5_SET(modify_esw_vport_context_in, in,
235 field_select.vport_cvlan_insert, 1);
237 return modify_esw_vport_context_cmd(dev, vport, in, sizeof(in));
240 /* HW L2 Table (MPFS) management */
241 static int set_l2_table_entry_cmd(struct mlx5_core_dev *dev, u32 index,
242 u8 *mac, u8 vlan_valid, u16 vlan)
244 u32 in[MLX5_ST_SZ_DW(set_l2_table_entry_in)];
245 u32 out[MLX5_ST_SZ_DW(set_l2_table_entry_out)];
248 memset(in, 0, sizeof(in));
249 memset(out, 0, sizeof(out));
251 MLX5_SET(set_l2_table_entry_in, in, opcode,
252 MLX5_CMD_OP_SET_L2_TABLE_ENTRY);
253 MLX5_SET(set_l2_table_entry_in, in, table_index, index);
254 MLX5_SET(set_l2_table_entry_in, in, vlan_valid, vlan_valid);
255 MLX5_SET(set_l2_table_entry_in, in, vlan, vlan);
257 in_mac_addr = MLX5_ADDR_OF(set_l2_table_entry_in, in, mac_address);
258 ether_addr_copy(&in_mac_addr[2], mac);
260 return mlx5_cmd_exec_check_status(dev, in, sizeof(in),
264 static int del_l2_table_entry_cmd(struct mlx5_core_dev *dev, u32 index)
266 u32 in[MLX5_ST_SZ_DW(delete_l2_table_entry_in)];
267 u32 out[MLX5_ST_SZ_DW(delete_l2_table_entry_out)];
269 memset(in, 0, sizeof(in));
270 memset(out, 0, sizeof(out));
272 MLX5_SET(delete_l2_table_entry_in, in, opcode,
273 MLX5_CMD_OP_DELETE_L2_TABLE_ENTRY);
274 MLX5_SET(delete_l2_table_entry_in, in, table_index, index);
275 return mlx5_cmd_exec_check_status(dev, in, sizeof(in),
279 static int alloc_l2_table_index(struct mlx5_l2_table *l2_table, u32 *ix)
283 *ix = find_first_zero_bit(l2_table->bitmap, l2_table->size);
284 if (*ix >= l2_table->size)
287 __set_bit(*ix, l2_table->bitmap);
292 static void free_l2_table_index(struct mlx5_l2_table *l2_table, u32 ix)
294 __clear_bit(ix, l2_table->bitmap);
297 static int set_l2_table_entry(struct mlx5_core_dev *dev, u8 *mac,
298 u8 vlan_valid, u16 vlan,
301 struct mlx5_l2_table *l2_table = &dev->priv.eswitch->l2_table;
304 err = alloc_l2_table_index(l2_table, index);
308 err = set_l2_table_entry_cmd(dev, *index, mac, vlan_valid, vlan);
310 free_l2_table_index(l2_table, *index);
315 static void del_l2_table_entry(struct mlx5_core_dev *dev, u32 index)
317 struct mlx5_l2_table *l2_table = &dev->priv.eswitch->l2_table;
319 del_l2_table_entry_cmd(dev, index);
320 free_l2_table_index(l2_table, index);
324 static struct mlx5_flow_rule *
325 __esw_fdb_set_vport_rule(struct mlx5_eswitch *esw, u32 vport, bool rx_rule,
326 u8 mac_c[ETH_ALEN], u8 mac_v[ETH_ALEN])
328 int match_header = (is_zero_ether_addr(mac_c) ? 0 :
329 MLX5_MATCH_OUTER_HEADERS);
330 struct mlx5_flow_rule *flow_rule = NULL;
331 struct mlx5_flow_destination dest;
332 void *mv_misc = NULL;
333 void *mc_misc = NULL;
340 match_header |= MLX5_MATCH_MISC_PARAMETERS;
341 match_v = kzalloc(MLX5_ST_SZ_BYTES(fte_match_param), GFP_KERNEL);
342 match_c = kzalloc(MLX5_ST_SZ_BYTES(fte_match_param), GFP_KERNEL);
343 if (!match_v || !match_c) {
344 pr_warn("FDB: Failed to alloc match parameters\n");
348 dmac_v = MLX5_ADDR_OF(fte_match_param, match_v,
349 outer_headers.dmac_47_16);
350 dmac_c = MLX5_ADDR_OF(fte_match_param, match_c,
351 outer_headers.dmac_47_16);
353 if (match_header & MLX5_MATCH_OUTER_HEADERS) {
354 ether_addr_copy(dmac_v, mac_v);
355 ether_addr_copy(dmac_c, mac_c);
358 if (match_header & MLX5_MATCH_MISC_PARAMETERS) {
359 mv_misc = MLX5_ADDR_OF(fte_match_param, match_v, misc_parameters);
360 mc_misc = MLX5_ADDR_OF(fte_match_param, match_c, misc_parameters);
361 MLX5_SET(fte_match_set_misc, mv_misc, source_port, UPLINK_VPORT);
362 MLX5_SET_TO_ONES(fte_match_set_misc, mc_misc, source_port);
365 dest.type = MLX5_FLOW_DESTINATION_TYPE_VPORT;
366 dest.vport_num = vport;
369 "\tFDB add rule dmac_v(%pM) dmac_c(%pM) -> vport(%d)\n",
370 dmac_v, dmac_c, vport);
372 mlx5_add_flow_rule(esw->fdb_table.fdb,
376 MLX5_FLOW_CONTEXT_ACTION_FWD_DEST,
378 if (IS_ERR(flow_rule)) {
380 "FDB: Failed to add flow rule: dmac_v(%pM) dmac_c(%pM) -> vport(%d), err(%ld)\n",
381 dmac_v, dmac_c, vport, PTR_ERR(flow_rule));
390 static struct mlx5_flow_rule *
391 esw_fdb_set_vport_rule(struct mlx5_eswitch *esw, u8 mac[ETH_ALEN], u32 vport)
395 eth_broadcast_addr(mac_c);
396 return __esw_fdb_set_vport_rule(esw, vport, false, mac_c, mac);
399 static struct mlx5_flow_rule *
400 esw_fdb_set_vport_allmulti_rule(struct mlx5_eswitch *esw, u32 vport)
405 eth_zero_addr(mac_c);
406 eth_zero_addr(mac_v);
409 return __esw_fdb_set_vport_rule(esw, vport, false, mac_c, mac_v);
412 static struct mlx5_flow_rule *
413 esw_fdb_set_vport_promisc_rule(struct mlx5_eswitch *esw, u32 vport)
418 eth_zero_addr(mac_c);
419 eth_zero_addr(mac_v);
420 return __esw_fdb_set_vport_rule(esw, vport, true, mac_c, mac_v);
423 static int esw_create_legacy_fdb_table(struct mlx5_eswitch *esw, int nvports)
425 int inlen = MLX5_ST_SZ_BYTES(create_flow_group_in);
426 struct mlx5_core_dev *dev = esw->dev;
427 struct mlx5_flow_namespace *root_ns;
428 struct mlx5_flow_table *fdb;
429 struct mlx5_flow_group *g;
430 void *match_criteria;
436 esw_debug(dev, "Create FDB log_max_size(%d)\n",
437 MLX5_CAP_ESW_FLOWTABLE_FDB(dev, log_max_ft_size));
439 root_ns = mlx5_get_flow_namespace(dev, MLX5_FLOW_NAMESPACE_FDB);
441 esw_warn(dev, "Failed to get FDB flow namespace\n");
445 flow_group_in = mlx5_vzalloc(inlen);
448 memset(flow_group_in, 0, inlen);
450 table_size = BIT(MLX5_CAP_ESW_FLOWTABLE_FDB(dev, log_max_ft_size));
451 fdb = mlx5_create_flow_table(root_ns, 0, table_size, 0);
454 esw_warn(dev, "Failed to create FDB Table err %d\n", err);
457 esw->fdb_table.fdb = fdb;
459 /* Addresses group : Full match unicast/multicast addresses */
460 MLX5_SET(create_flow_group_in, flow_group_in, match_criteria_enable,
461 MLX5_MATCH_OUTER_HEADERS);
462 match_criteria = MLX5_ADDR_OF(create_flow_group_in, flow_group_in, match_criteria);
463 dmac = MLX5_ADDR_OF(fte_match_param, match_criteria, outer_headers.dmac_47_16);
464 MLX5_SET(create_flow_group_in, flow_group_in, start_flow_index, 0);
465 /* Preserve 2 entries for allmulti and promisc rules*/
466 MLX5_SET(create_flow_group_in, flow_group_in, end_flow_index, table_size - 3);
467 eth_broadcast_addr(dmac);
468 g = mlx5_create_flow_group(fdb, flow_group_in);
471 esw_warn(dev, "Failed to create flow group err(%d)\n", err);
474 esw->fdb_table.legacy.addr_grp = g;
476 /* Allmulti group : One rule that forwards any mcast traffic */
477 MLX5_SET(create_flow_group_in, flow_group_in, match_criteria_enable,
478 MLX5_MATCH_OUTER_HEADERS);
479 MLX5_SET(create_flow_group_in, flow_group_in, start_flow_index, table_size - 2);
480 MLX5_SET(create_flow_group_in, flow_group_in, end_flow_index, table_size - 2);
483 g = mlx5_create_flow_group(fdb, flow_group_in);
486 esw_warn(dev, "Failed to create allmulti flow group err(%d)\n", err);
489 esw->fdb_table.legacy.allmulti_grp = g;
491 /* Promiscuous group :
492 * One rule that forward all unmatched traffic from previous groups
495 MLX5_SET(create_flow_group_in, flow_group_in, match_criteria_enable,
496 MLX5_MATCH_MISC_PARAMETERS);
497 MLX5_SET_TO_ONES(fte_match_param, match_criteria, misc_parameters.source_port);
498 MLX5_SET(create_flow_group_in, flow_group_in, start_flow_index, table_size - 1);
499 MLX5_SET(create_flow_group_in, flow_group_in, end_flow_index, table_size - 1);
500 g = mlx5_create_flow_group(fdb, flow_group_in);
503 esw_warn(dev, "Failed to create promisc flow group err(%d)\n", err);
506 esw->fdb_table.legacy.promisc_grp = g;
510 if (!IS_ERR_OR_NULL(esw->fdb_table.legacy.allmulti_grp)) {
511 mlx5_destroy_flow_group(esw->fdb_table.legacy.allmulti_grp);
512 esw->fdb_table.legacy.allmulti_grp = NULL;
514 if (!IS_ERR_OR_NULL(esw->fdb_table.legacy.addr_grp)) {
515 mlx5_destroy_flow_group(esw->fdb_table.legacy.addr_grp);
516 esw->fdb_table.legacy.addr_grp = NULL;
518 if (!IS_ERR_OR_NULL(esw->fdb_table.fdb)) {
519 mlx5_destroy_flow_table(esw->fdb_table.fdb);
520 esw->fdb_table.fdb = NULL;
524 kvfree(flow_group_in);
528 static void esw_destroy_legacy_fdb_table(struct mlx5_eswitch *esw)
530 if (!esw->fdb_table.fdb)
533 esw_debug(esw->dev, "Destroy FDB Table\n");
534 mlx5_destroy_flow_group(esw->fdb_table.legacy.promisc_grp);
535 mlx5_destroy_flow_group(esw->fdb_table.legacy.allmulti_grp);
536 mlx5_destroy_flow_group(esw->fdb_table.legacy.addr_grp);
537 mlx5_destroy_flow_table(esw->fdb_table.fdb);
538 esw->fdb_table.fdb = NULL;
539 esw->fdb_table.legacy.addr_grp = NULL;
540 esw->fdb_table.legacy.allmulti_grp = NULL;
541 esw->fdb_table.legacy.promisc_grp = NULL;
544 /* E-Switch vport UC/MC lists management */
545 typedef int (*vport_addr_action)(struct mlx5_eswitch *esw,
546 struct vport_addr *vaddr);
548 static int esw_add_uc_addr(struct mlx5_eswitch *esw, struct vport_addr *vaddr)
550 struct hlist_head *hash = esw->l2_table.l2_hash;
551 struct esw_uc_addr *esw_uc;
552 u8 *mac = vaddr->node.addr;
553 u32 vport = vaddr->vport;
556 esw_uc = l2addr_hash_find(hash, mac, struct esw_uc_addr);
559 "Failed to set L2 mac(%pM) for vport(%d), mac is already in use by vport(%d)\n",
560 mac, vport, esw_uc->vport);
564 esw_uc = l2addr_hash_add(hash, mac, struct esw_uc_addr, GFP_KERNEL);
567 esw_uc->vport = vport;
569 err = set_l2_table_entry(esw->dev, mac, 0, 0, &esw_uc->table_index);
573 /* SRIOV is enabled: Forward UC MAC to vport */
574 if (esw->fdb_table.fdb && esw->mode == SRIOV_LEGACY)
575 vaddr->flow_rule = esw_fdb_set_vport_rule(esw, mac, vport);
577 esw_debug(esw->dev, "\tADDED UC MAC: vport[%d] %pM index:%d fr(%p)\n",
578 vport, mac, esw_uc->table_index, vaddr->flow_rule);
581 l2addr_hash_del(esw_uc);
585 static int esw_del_uc_addr(struct mlx5_eswitch *esw, struct vport_addr *vaddr)
587 struct hlist_head *hash = esw->l2_table.l2_hash;
588 struct esw_uc_addr *esw_uc;
589 u8 *mac = vaddr->node.addr;
590 u32 vport = vaddr->vport;
592 esw_uc = l2addr_hash_find(hash, mac, struct esw_uc_addr);
593 if (!esw_uc || esw_uc->vport != vport) {
595 "MAC(%pM) doesn't belong to vport (%d)\n",
599 esw_debug(esw->dev, "\tDELETE UC MAC: vport[%d] %pM index:%d fr(%p)\n",
600 vport, mac, esw_uc->table_index, vaddr->flow_rule);
602 del_l2_table_entry(esw->dev, esw_uc->table_index);
604 if (vaddr->flow_rule)
605 mlx5_del_flow_rule(vaddr->flow_rule);
606 vaddr->flow_rule = NULL;
608 l2addr_hash_del(esw_uc);
612 static void update_allmulti_vports(struct mlx5_eswitch *esw,
613 struct vport_addr *vaddr,
614 struct esw_mc_addr *esw_mc)
616 u8 *mac = vaddr->node.addr;
619 for (vport_idx = 0; vport_idx < esw->total_vports; vport_idx++) {
620 struct mlx5_vport *vport = &esw->vports[vport_idx];
621 struct hlist_head *vport_hash = vport->mc_list;
622 struct vport_addr *iter_vaddr =
623 l2addr_hash_find(vport_hash,
626 if (IS_ERR_OR_NULL(vport->allmulti_rule) ||
627 vaddr->vport == vport_idx)
629 switch (vaddr->action) {
630 case MLX5_ACTION_ADD:
633 iter_vaddr = l2addr_hash_add(vport_hash, mac,
638 "ALL-MULTI: Failed to add MAC(%pM) to vport[%d] DB\n",
642 iter_vaddr->vport = vport_idx;
643 iter_vaddr->flow_rule =
644 esw_fdb_set_vport_rule(esw,
647 iter_vaddr->mc_promisc = true;
649 case MLX5_ACTION_DEL:
652 mlx5_del_flow_rule(iter_vaddr->flow_rule);
653 l2addr_hash_del(iter_vaddr);
659 static int esw_add_mc_addr(struct mlx5_eswitch *esw, struct vport_addr *vaddr)
661 struct hlist_head *hash = esw->mc_table;
662 struct esw_mc_addr *esw_mc;
663 u8 *mac = vaddr->node.addr;
664 u32 vport = vaddr->vport;
666 if (!esw->fdb_table.fdb)
669 esw_mc = l2addr_hash_find(hash, mac, struct esw_mc_addr);
673 esw_mc = l2addr_hash_add(hash, mac, struct esw_mc_addr, GFP_KERNEL);
677 esw_mc->uplink_rule = /* Forward MC MAC to Uplink */
678 esw_fdb_set_vport_rule(esw, mac, UPLINK_VPORT);
680 /* Add this multicast mac to all the mc promiscuous vports */
681 update_allmulti_vports(esw, vaddr, esw_mc);
684 /* If the multicast mac is added as a result of mc promiscuous vport,
685 * don't increment the multicast ref count
687 if (!vaddr->mc_promisc)
690 /* Forward MC MAC to vport */
691 vaddr->flow_rule = esw_fdb_set_vport_rule(esw, mac, vport);
693 "\tADDED MC MAC: vport[%d] %pM fr(%p) refcnt(%d) uplinkfr(%p)\n",
694 vport, mac, vaddr->flow_rule,
695 esw_mc->refcnt, esw_mc->uplink_rule);
699 static int esw_del_mc_addr(struct mlx5_eswitch *esw, struct vport_addr *vaddr)
701 struct hlist_head *hash = esw->mc_table;
702 struct esw_mc_addr *esw_mc;
703 u8 *mac = vaddr->node.addr;
704 u32 vport = vaddr->vport;
706 if (!esw->fdb_table.fdb)
709 esw_mc = l2addr_hash_find(hash, mac, struct esw_mc_addr);
712 "Failed to find eswitch MC addr for MAC(%pM) vport(%d)",
717 "\tDELETE MC MAC: vport[%d] %pM fr(%p) refcnt(%d) uplinkfr(%p)\n",
718 vport, mac, vaddr->flow_rule, esw_mc->refcnt,
719 esw_mc->uplink_rule);
721 if (vaddr->flow_rule)
722 mlx5_del_flow_rule(vaddr->flow_rule);
723 vaddr->flow_rule = NULL;
725 /* If the multicast mac is added as a result of mc promiscuous vport,
726 * don't decrement the multicast ref count.
728 if (vaddr->mc_promisc || (--esw_mc->refcnt > 0))
731 /* Remove this multicast mac from all the mc promiscuous vports */
732 update_allmulti_vports(esw, vaddr, esw_mc);
734 if (esw_mc->uplink_rule)
735 mlx5_del_flow_rule(esw_mc->uplink_rule);
737 l2addr_hash_del(esw_mc);
741 /* Apply vport UC/MC list to HW l2 table and FDB table */
742 static void esw_apply_vport_addr_list(struct mlx5_eswitch *esw,
743 u32 vport_num, int list_type)
745 struct mlx5_vport *vport = &esw->vports[vport_num];
746 bool is_uc = list_type == MLX5_NVPRT_LIST_TYPE_UC;
747 vport_addr_action vport_addr_add;
748 vport_addr_action vport_addr_del;
749 struct vport_addr *addr;
750 struct l2addr_node *node;
751 struct hlist_head *hash;
752 struct hlist_node *tmp;
755 vport_addr_add = is_uc ? esw_add_uc_addr :
757 vport_addr_del = is_uc ? esw_del_uc_addr :
760 hash = is_uc ? vport->uc_list : vport->mc_list;
761 for_each_l2hash_node(node, tmp, hash, hi) {
762 addr = container_of(node, struct vport_addr, node);
763 switch (addr->action) {
764 case MLX5_ACTION_ADD:
765 vport_addr_add(esw, addr);
766 addr->action = MLX5_ACTION_NONE;
768 case MLX5_ACTION_DEL:
769 vport_addr_del(esw, addr);
770 l2addr_hash_del(addr);
776 /* Sync vport UC/MC list from vport context */
777 static void esw_update_vport_addr_list(struct mlx5_eswitch *esw,
778 u32 vport_num, int list_type)
780 struct mlx5_vport *vport = &esw->vports[vport_num];
781 bool is_uc = list_type == MLX5_NVPRT_LIST_TYPE_UC;
782 u8 (*mac_list)[ETH_ALEN];
783 struct l2addr_node *node;
784 struct vport_addr *addr;
785 struct hlist_head *hash;
786 struct hlist_node *tmp;
792 size = is_uc ? MLX5_MAX_UC_PER_VPORT(esw->dev) :
793 MLX5_MAX_MC_PER_VPORT(esw->dev);
795 mac_list = kcalloc(size, ETH_ALEN, GFP_KERNEL);
799 hash = is_uc ? vport->uc_list : vport->mc_list;
801 for_each_l2hash_node(node, tmp, hash, hi) {
802 addr = container_of(node, struct vport_addr, node);
803 addr->action = MLX5_ACTION_DEL;
809 err = mlx5_query_nic_vport_mac_list(esw->dev, vport_num, list_type,
813 esw_debug(esw->dev, "vport[%d] context update %s list size (%d)\n",
814 vport_num, is_uc ? "UC" : "MC", size);
816 for (i = 0; i < size; i++) {
817 if (is_uc && !is_valid_ether_addr(mac_list[i]))
820 if (!is_uc && !is_multicast_ether_addr(mac_list[i]))
823 addr = l2addr_hash_find(hash, mac_list[i], struct vport_addr);
825 addr->action = MLX5_ACTION_NONE;
826 /* If this mac was previously added because of allmulti
827 * promiscuous rx mode, its now converted to be original
830 if (addr->mc_promisc) {
831 struct esw_mc_addr *esw_mc =
832 l2addr_hash_find(esw->mc_table,
837 "Failed to MAC(%pM) in mcast DB\n",
842 addr->mc_promisc = false;
847 addr = l2addr_hash_add(hash, mac_list[i], struct vport_addr,
851 "Failed to add MAC(%pM) to vport[%d] DB\n",
852 mac_list[i], vport_num);
855 addr->vport = vport_num;
856 addr->action = MLX5_ACTION_ADD;
862 /* Sync vport UC/MC list from vport context
863 * Must be called after esw_update_vport_addr_list
865 static void esw_update_vport_mc_promisc(struct mlx5_eswitch *esw, u32 vport_num)
867 struct mlx5_vport *vport = &esw->vports[vport_num];
868 struct l2addr_node *node;
869 struct vport_addr *addr;
870 struct hlist_head *hash;
871 struct hlist_node *tmp;
874 hash = vport->mc_list;
876 for_each_l2hash_node(node, tmp, esw->mc_table, hi) {
877 u8 *mac = node->addr;
879 addr = l2addr_hash_find(hash, mac, struct vport_addr);
881 if (addr->action == MLX5_ACTION_DEL)
882 addr->action = MLX5_ACTION_NONE;
885 addr = l2addr_hash_add(hash, mac, struct vport_addr,
889 "Failed to add allmulti MAC(%pM) to vport[%d] DB\n",
893 addr->vport = vport_num;
894 addr->action = MLX5_ACTION_ADD;
895 addr->mc_promisc = true;
899 /* Apply vport rx mode to HW FDB table */
900 static void esw_apply_vport_rx_mode(struct mlx5_eswitch *esw, u32 vport_num,
901 bool promisc, bool mc_promisc)
903 struct esw_mc_addr *allmulti_addr = esw->mc_promisc;
904 struct mlx5_vport *vport = &esw->vports[vport_num];
906 if (IS_ERR_OR_NULL(vport->allmulti_rule) != mc_promisc)
910 vport->allmulti_rule =
911 esw_fdb_set_vport_allmulti_rule(esw, vport_num);
912 if (!allmulti_addr->uplink_rule)
913 allmulti_addr->uplink_rule =
914 esw_fdb_set_vport_allmulti_rule(esw,
916 allmulti_addr->refcnt++;
917 } else if (vport->allmulti_rule) {
918 mlx5_del_flow_rule(vport->allmulti_rule);
919 vport->allmulti_rule = NULL;
921 if (--allmulti_addr->refcnt > 0)
924 if (allmulti_addr->uplink_rule)
925 mlx5_del_flow_rule(allmulti_addr->uplink_rule);
926 allmulti_addr->uplink_rule = NULL;
930 if (IS_ERR_OR_NULL(vport->promisc_rule) != promisc)
934 vport->promisc_rule = esw_fdb_set_vport_promisc_rule(esw,
936 } else if (vport->promisc_rule) {
937 mlx5_del_flow_rule(vport->promisc_rule);
938 vport->promisc_rule = NULL;
942 /* Sync vport rx mode from vport context */
943 static void esw_update_vport_rx_mode(struct mlx5_eswitch *esw, u32 vport_num)
945 struct mlx5_vport *vport = &esw->vports[vport_num];
951 err = mlx5_query_nic_vport_promisc(esw->dev,
958 esw_debug(esw->dev, "vport[%d] context update rx mode promisc_all=%d, all_multi=%d\n",
959 vport_num, promisc_all, promisc_mc);
961 if (!vport->trusted || !vport->enabled) {
967 esw_apply_vport_rx_mode(esw, vport_num, promisc_all,
968 (promisc_all || promisc_mc));
971 static void esw_vport_change_handle_locked(struct mlx5_vport *vport)
973 struct mlx5_core_dev *dev = vport->dev;
974 struct mlx5_eswitch *esw = dev->priv.eswitch;
977 mlx5_query_nic_vport_mac_address(dev, vport->vport, mac);
978 esw_debug(dev, "vport[%d] Context Changed: perm mac: %pM\n",
981 if (vport->enabled_events & UC_ADDR_CHANGE) {
982 esw_update_vport_addr_list(esw, vport->vport,
983 MLX5_NVPRT_LIST_TYPE_UC);
984 esw_apply_vport_addr_list(esw, vport->vport,
985 MLX5_NVPRT_LIST_TYPE_UC);
988 if (vport->enabled_events & MC_ADDR_CHANGE) {
989 esw_update_vport_addr_list(esw, vport->vport,
990 MLX5_NVPRT_LIST_TYPE_MC);
993 if (vport->enabled_events & PROMISC_CHANGE) {
994 esw_update_vport_rx_mode(esw, vport->vport);
995 if (!IS_ERR_OR_NULL(vport->allmulti_rule))
996 esw_update_vport_mc_promisc(esw, vport->vport);
999 if (vport->enabled_events & (PROMISC_CHANGE | MC_ADDR_CHANGE)) {
1000 esw_apply_vport_addr_list(esw, vport->vport,
1001 MLX5_NVPRT_LIST_TYPE_MC);
1004 esw_debug(esw->dev, "vport[%d] Context Changed: Done\n", vport->vport);
1006 arm_vport_context_events_cmd(dev, vport->vport,
1007 vport->enabled_events);
1010 static void esw_vport_change_handler(struct work_struct *work)
1012 struct mlx5_vport *vport =
1013 container_of(work, struct mlx5_vport, vport_change_handler);
1014 struct mlx5_eswitch *esw = vport->dev->priv.eswitch;
1016 mutex_lock(&esw->state_lock);
1017 esw_vport_change_handle_locked(vport);
1018 mutex_unlock(&esw->state_lock);
1021 static void esw_vport_enable_egress_acl(struct mlx5_eswitch *esw,
1022 struct mlx5_vport *vport)
1024 int inlen = MLX5_ST_SZ_BYTES(create_flow_group_in);
1025 struct mlx5_flow_group *vlan_grp = NULL;
1026 struct mlx5_flow_group *drop_grp = NULL;
1027 struct mlx5_core_dev *dev = esw->dev;
1028 struct mlx5_flow_namespace *root_ns;
1029 struct mlx5_flow_table *acl;
1030 void *match_criteria;
1032 /* The egress acl table contains 2 rules:
1033 * 1)Allow traffic with vlan_tag=vst_vlan_id
1034 * 2)Drop all other traffic.
1039 if (!MLX5_CAP_ESW_EGRESS_ACL(dev, ft_support) ||
1040 !IS_ERR_OR_NULL(vport->egress.acl))
1043 esw_debug(dev, "Create vport[%d] egress ACL log_max_size(%d)\n",
1044 vport->vport, MLX5_CAP_ESW_EGRESS_ACL(dev, log_max_ft_size));
1046 root_ns = mlx5_get_flow_namespace(dev, MLX5_FLOW_NAMESPACE_ESW_EGRESS);
1048 esw_warn(dev, "Failed to get E-Switch egress flow namespace\n");
1052 flow_group_in = mlx5_vzalloc(inlen);
1056 acl = mlx5_create_vport_flow_table(root_ns, 0, table_size, 0, vport->vport);
1059 esw_warn(dev, "Failed to create E-Switch vport[%d] egress flow Table, err(%d)\n",
1064 MLX5_SET(create_flow_group_in, flow_group_in, match_criteria_enable, MLX5_MATCH_OUTER_HEADERS);
1065 match_criteria = MLX5_ADDR_OF(create_flow_group_in, flow_group_in, match_criteria);
1066 MLX5_SET_TO_ONES(fte_match_param, match_criteria, outer_headers.vlan_tag);
1067 MLX5_SET_TO_ONES(fte_match_param, match_criteria, outer_headers.first_vid);
1068 MLX5_SET(create_flow_group_in, flow_group_in, start_flow_index, 0);
1069 MLX5_SET(create_flow_group_in, flow_group_in, end_flow_index, 0);
1071 vlan_grp = mlx5_create_flow_group(acl, flow_group_in);
1072 if (IS_ERR(vlan_grp)) {
1073 err = PTR_ERR(vlan_grp);
1074 esw_warn(dev, "Failed to create E-Switch vport[%d] egress allowed vlans flow group, err(%d)\n",
1079 memset(flow_group_in, 0, inlen);
1080 MLX5_SET(create_flow_group_in, flow_group_in, start_flow_index, 1);
1081 MLX5_SET(create_flow_group_in, flow_group_in, end_flow_index, 1);
1082 drop_grp = mlx5_create_flow_group(acl, flow_group_in);
1083 if (IS_ERR(drop_grp)) {
1084 err = PTR_ERR(drop_grp);
1085 esw_warn(dev, "Failed to create E-Switch vport[%d] egress drop flow group, err(%d)\n",
1090 vport->egress.acl = acl;
1091 vport->egress.drop_grp = drop_grp;
1092 vport->egress.allowed_vlans_grp = vlan_grp;
1094 kvfree(flow_group_in);
1095 if (err && !IS_ERR_OR_NULL(vlan_grp))
1096 mlx5_destroy_flow_group(vlan_grp);
1097 if (err && !IS_ERR_OR_NULL(acl))
1098 mlx5_destroy_flow_table(acl);
1101 static void esw_vport_cleanup_egress_rules(struct mlx5_eswitch *esw,
1102 struct mlx5_vport *vport)
1104 if (!IS_ERR_OR_NULL(vport->egress.allowed_vlan))
1105 mlx5_del_flow_rule(vport->egress.allowed_vlan);
1107 if (!IS_ERR_OR_NULL(vport->egress.drop_rule))
1108 mlx5_del_flow_rule(vport->egress.drop_rule);
1110 vport->egress.allowed_vlan = NULL;
1111 vport->egress.drop_rule = NULL;
1114 static void esw_vport_disable_egress_acl(struct mlx5_eswitch *esw,
1115 struct mlx5_vport *vport)
1117 if (IS_ERR_OR_NULL(vport->egress.acl))
1120 esw_debug(esw->dev, "Destroy vport[%d] E-Switch egress ACL\n", vport->vport);
1122 esw_vport_cleanup_egress_rules(esw, vport);
1123 mlx5_destroy_flow_group(vport->egress.allowed_vlans_grp);
1124 mlx5_destroy_flow_group(vport->egress.drop_grp);
1125 mlx5_destroy_flow_table(vport->egress.acl);
1126 vport->egress.allowed_vlans_grp = NULL;
1127 vport->egress.drop_grp = NULL;
1128 vport->egress.acl = NULL;
1131 static void esw_vport_enable_ingress_acl(struct mlx5_eswitch *esw,
1132 struct mlx5_vport *vport)
1134 int inlen = MLX5_ST_SZ_BYTES(create_flow_group_in);
1135 struct mlx5_core_dev *dev = esw->dev;
1136 struct mlx5_flow_namespace *root_ns;
1137 struct mlx5_flow_table *acl;
1138 struct mlx5_flow_group *g;
1139 void *match_criteria;
1141 /* The ingress acl table contains 4 groups
1142 * (2 active rules at the same time -
1143 * 1 allow rule from one of the first 3 groups.
1144 * 1 drop rule from the last group):
1145 * 1)Allow untagged traffic with smac=original mac.
1146 * 2)Allow untagged traffic.
1147 * 3)Allow traffic with smac=original mac.
1148 * 4)Drop all other traffic.
1153 if (!MLX5_CAP_ESW_INGRESS_ACL(dev, ft_support) ||
1154 !IS_ERR_OR_NULL(vport->ingress.acl))
1157 esw_debug(dev, "Create vport[%d] ingress ACL log_max_size(%d)\n",
1158 vport->vport, MLX5_CAP_ESW_INGRESS_ACL(dev, log_max_ft_size));
1160 root_ns = mlx5_get_flow_namespace(dev, MLX5_FLOW_NAMESPACE_ESW_INGRESS);
1162 esw_warn(dev, "Failed to get E-Switch ingress flow namespace\n");
1166 flow_group_in = mlx5_vzalloc(inlen);
1170 acl = mlx5_create_vport_flow_table(root_ns, 0, table_size, 0, vport->vport);
1173 esw_warn(dev, "Failed to create E-Switch vport[%d] ingress flow Table, err(%d)\n",
1177 vport->ingress.acl = acl;
1179 match_criteria = MLX5_ADDR_OF(create_flow_group_in, flow_group_in, match_criteria);
1181 MLX5_SET(create_flow_group_in, flow_group_in, match_criteria_enable, MLX5_MATCH_OUTER_HEADERS);
1182 MLX5_SET_TO_ONES(fte_match_param, match_criteria, outer_headers.vlan_tag);
1183 MLX5_SET_TO_ONES(fte_match_param, match_criteria, outer_headers.smac_47_16);
1184 MLX5_SET_TO_ONES(fte_match_param, match_criteria, outer_headers.smac_15_0);
1185 MLX5_SET(create_flow_group_in, flow_group_in, start_flow_index, 0);
1186 MLX5_SET(create_flow_group_in, flow_group_in, end_flow_index, 0);
1188 g = mlx5_create_flow_group(acl, flow_group_in);
1191 esw_warn(dev, "Failed to create E-Switch vport[%d] ingress untagged spoofchk flow group, err(%d)\n",
1195 vport->ingress.allow_untagged_spoofchk_grp = g;
1197 memset(flow_group_in, 0, inlen);
1198 MLX5_SET(create_flow_group_in, flow_group_in, match_criteria_enable, MLX5_MATCH_OUTER_HEADERS);
1199 MLX5_SET_TO_ONES(fte_match_param, match_criteria, outer_headers.vlan_tag);
1200 MLX5_SET(create_flow_group_in, flow_group_in, start_flow_index, 1);
1201 MLX5_SET(create_flow_group_in, flow_group_in, end_flow_index, 1);
1203 g = mlx5_create_flow_group(acl, flow_group_in);
1206 esw_warn(dev, "Failed to create E-Switch vport[%d] ingress untagged flow group, err(%d)\n",
1210 vport->ingress.allow_untagged_only_grp = g;
1212 memset(flow_group_in, 0, inlen);
1213 MLX5_SET(create_flow_group_in, flow_group_in, match_criteria_enable, MLX5_MATCH_OUTER_HEADERS);
1214 MLX5_SET_TO_ONES(fte_match_param, match_criteria, outer_headers.smac_47_16);
1215 MLX5_SET_TO_ONES(fte_match_param, match_criteria, outer_headers.smac_15_0);
1216 MLX5_SET(create_flow_group_in, flow_group_in, start_flow_index, 2);
1217 MLX5_SET(create_flow_group_in, flow_group_in, end_flow_index, 2);
1219 g = mlx5_create_flow_group(acl, flow_group_in);
1222 esw_warn(dev, "Failed to create E-Switch vport[%d] ingress spoofchk flow group, err(%d)\n",
1226 vport->ingress.allow_spoofchk_only_grp = g;
1228 memset(flow_group_in, 0, inlen);
1229 MLX5_SET(create_flow_group_in, flow_group_in, start_flow_index, 3);
1230 MLX5_SET(create_flow_group_in, flow_group_in, end_flow_index, 3);
1232 g = mlx5_create_flow_group(acl, flow_group_in);
1235 esw_warn(dev, "Failed to create E-Switch vport[%d] ingress drop flow group, err(%d)\n",
1239 vport->ingress.drop_grp = g;
1243 if (!IS_ERR_OR_NULL(vport->ingress.allow_spoofchk_only_grp))
1244 mlx5_destroy_flow_group(
1245 vport->ingress.allow_spoofchk_only_grp);
1246 if (!IS_ERR_OR_NULL(vport->ingress.allow_untagged_only_grp))
1247 mlx5_destroy_flow_group(
1248 vport->ingress.allow_untagged_only_grp);
1249 if (!IS_ERR_OR_NULL(vport->ingress.allow_untagged_spoofchk_grp))
1250 mlx5_destroy_flow_group(
1251 vport->ingress.allow_untagged_spoofchk_grp);
1252 if (!IS_ERR_OR_NULL(vport->ingress.acl))
1253 mlx5_destroy_flow_table(vport->ingress.acl);
1256 kvfree(flow_group_in);
1259 static void esw_vport_cleanup_ingress_rules(struct mlx5_eswitch *esw,
1260 struct mlx5_vport *vport)
1262 if (!IS_ERR_OR_NULL(vport->ingress.drop_rule))
1263 mlx5_del_flow_rule(vport->ingress.drop_rule);
1265 if (!IS_ERR_OR_NULL(vport->ingress.allow_rule))
1266 mlx5_del_flow_rule(vport->ingress.allow_rule);
1268 vport->ingress.drop_rule = NULL;
1269 vport->ingress.allow_rule = NULL;
1272 static void esw_vport_disable_ingress_acl(struct mlx5_eswitch *esw,
1273 struct mlx5_vport *vport)
1275 if (IS_ERR_OR_NULL(vport->ingress.acl))
1278 esw_debug(esw->dev, "Destroy vport[%d] E-Switch ingress ACL\n", vport->vport);
1280 esw_vport_cleanup_ingress_rules(esw, vport);
1281 mlx5_destroy_flow_group(vport->ingress.allow_spoofchk_only_grp);
1282 mlx5_destroy_flow_group(vport->ingress.allow_untagged_only_grp);
1283 mlx5_destroy_flow_group(vport->ingress.allow_untagged_spoofchk_grp);
1284 mlx5_destroy_flow_group(vport->ingress.drop_grp);
1285 mlx5_destroy_flow_table(vport->ingress.acl);
1286 vport->ingress.acl = NULL;
1287 vport->ingress.drop_grp = NULL;
1288 vport->ingress.allow_spoofchk_only_grp = NULL;
1289 vport->ingress.allow_untagged_only_grp = NULL;
1290 vport->ingress.allow_untagged_spoofchk_grp = NULL;
1293 static int esw_vport_ingress_config(struct mlx5_eswitch *esw,
1294 struct mlx5_vport *vport)
1302 if (vport->spoofchk) {
1303 err = mlx5_query_nic_vport_mac_address(esw->dev, vport->vport, smac);
1306 "vport[%d] configure ingress rules failed, query smac failed, err(%d)\n",
1311 if (!is_valid_ether_addr(smac)) {
1312 mlx5_core_warn(esw->dev,
1313 "vport[%d] configure ingress rules failed, illegal mac with spoofchk\n",
1319 esw_vport_cleanup_ingress_rules(esw, vport);
1321 if (!vport->vlan && !vport->qos && !vport->spoofchk) {
1322 esw_vport_disable_ingress_acl(esw, vport);
1326 esw_vport_enable_ingress_acl(esw, vport);
1329 "vport[%d] configure ingress rules, vlan(%d) qos(%d)\n",
1330 vport->vport, vport->vlan, vport->qos);
1332 match_v = kzalloc(MLX5_ST_SZ_BYTES(fte_match_param), GFP_KERNEL);
1333 match_c = kzalloc(MLX5_ST_SZ_BYTES(fte_match_param), GFP_KERNEL);
1334 if (!match_v || !match_c) {
1336 esw_warn(esw->dev, "vport[%d] configure ingress rules failed, err(%d)\n",
1341 if (vport->vlan || vport->qos)
1342 MLX5_SET_TO_ONES(fte_match_param, match_c, outer_headers.vlan_tag);
1344 if (vport->spoofchk) {
1345 MLX5_SET_TO_ONES(fte_match_param, match_c, outer_headers.smac_47_16);
1346 MLX5_SET_TO_ONES(fte_match_param, match_c, outer_headers.smac_15_0);
1347 smac_v = MLX5_ADDR_OF(fte_match_param,
1349 outer_headers.smac_47_16);
1350 ether_addr_copy(smac_v, smac);
1353 vport->ingress.allow_rule =
1354 mlx5_add_flow_rule(vport->ingress.acl,
1355 MLX5_MATCH_OUTER_HEADERS,
1358 MLX5_FLOW_CONTEXT_ACTION_ALLOW,
1360 if (IS_ERR(vport->ingress.allow_rule)) {
1361 err = PTR_ERR(vport->ingress.allow_rule);
1362 pr_warn("vport[%d] configure ingress allow rule, err(%d)\n",
1364 vport->ingress.allow_rule = NULL;
1368 memset(match_c, 0, MLX5_ST_SZ_BYTES(fte_match_param));
1369 memset(match_v, 0, MLX5_ST_SZ_BYTES(fte_match_param));
1370 vport->ingress.drop_rule =
1371 mlx5_add_flow_rule(vport->ingress.acl,
1375 MLX5_FLOW_CONTEXT_ACTION_DROP,
1377 if (IS_ERR(vport->ingress.drop_rule)) {
1378 err = PTR_ERR(vport->ingress.drop_rule);
1379 pr_warn("vport[%d] configure ingress drop rule, err(%d)\n",
1381 vport->ingress.drop_rule = NULL;
1387 esw_vport_cleanup_ingress_rules(esw, vport);
1394 static int esw_vport_egress_config(struct mlx5_eswitch *esw,
1395 struct mlx5_vport *vport)
1401 esw_vport_cleanup_egress_rules(esw, vport);
1403 if (!vport->vlan && !vport->qos) {
1404 esw_vport_disable_egress_acl(esw, vport);
1408 esw_vport_enable_egress_acl(esw, vport);
1411 "vport[%d] configure egress rules, vlan(%d) qos(%d)\n",
1412 vport->vport, vport->vlan, vport->qos);
1414 match_v = kzalloc(MLX5_ST_SZ_BYTES(fte_match_param), GFP_KERNEL);
1415 match_c = kzalloc(MLX5_ST_SZ_BYTES(fte_match_param), GFP_KERNEL);
1416 if (!match_v || !match_c) {
1418 esw_warn(esw->dev, "vport[%d] configure egress rules failed, err(%d)\n",
1423 /* Allowed vlan rule */
1424 MLX5_SET_TO_ONES(fte_match_param, match_c, outer_headers.vlan_tag);
1425 MLX5_SET_TO_ONES(fte_match_param, match_v, outer_headers.vlan_tag);
1426 MLX5_SET_TO_ONES(fte_match_param, match_c, outer_headers.first_vid);
1427 MLX5_SET(fte_match_param, match_v, outer_headers.first_vid, vport->vlan);
1429 vport->egress.allowed_vlan =
1430 mlx5_add_flow_rule(vport->egress.acl,
1431 MLX5_MATCH_OUTER_HEADERS,
1434 MLX5_FLOW_CONTEXT_ACTION_ALLOW,
1436 if (IS_ERR(vport->egress.allowed_vlan)) {
1437 err = PTR_ERR(vport->egress.allowed_vlan);
1438 pr_warn("vport[%d] configure egress allowed vlan rule failed, err(%d)\n",
1440 vport->egress.allowed_vlan = NULL;
1444 /* Drop others rule (star rule) */
1445 memset(match_c, 0, MLX5_ST_SZ_BYTES(fte_match_param));
1446 memset(match_v, 0, MLX5_ST_SZ_BYTES(fte_match_param));
1447 vport->egress.drop_rule =
1448 mlx5_add_flow_rule(vport->egress.acl,
1452 MLX5_FLOW_CONTEXT_ACTION_DROP,
1454 if (IS_ERR(vport->egress.drop_rule)) {
1455 err = PTR_ERR(vport->egress.drop_rule);
1456 pr_warn("vport[%d] configure egress drop rule failed, err(%d)\n",
1458 vport->egress.drop_rule = NULL;
1466 static void esw_enable_vport(struct mlx5_eswitch *esw, int vport_num,
1469 struct mlx5_vport *vport = &esw->vports[vport_num];
1471 mutex_lock(&esw->state_lock);
1472 WARN_ON(vport->enabled);
1474 esw_debug(esw->dev, "Enabling VPORT(%d)\n", vport_num);
1476 if (vport_num) { /* Only VFs need ACLs for VST and spoofchk filtering */
1477 esw_vport_ingress_config(esw, vport);
1478 esw_vport_egress_config(esw, vport);
1481 mlx5_modify_vport_admin_state(esw->dev,
1482 MLX5_QUERY_VPORT_STATE_IN_OP_MOD_ESW_VPORT,
1484 MLX5_ESW_VPORT_ADMIN_STATE_AUTO);
1486 /* Sync with current vport context */
1487 vport->enabled_events = enable_events;
1488 vport->enabled = true;
1490 /* only PF is trusted by default */
1491 vport->trusted = (vport_num) ? false : true;
1492 esw_vport_change_handle_locked(vport);
1494 esw->enabled_vports++;
1495 esw_debug(esw->dev, "Enabled VPORT(%d)\n", vport_num);
1496 mutex_unlock(&esw->state_lock);
1499 static void esw_disable_vport(struct mlx5_eswitch *esw, int vport_num)
1501 struct mlx5_vport *vport = &esw->vports[vport_num];
1503 if (!vport->enabled)
1506 esw_debug(esw->dev, "Disabling vport(%d)\n", vport_num);
1507 /* Mark this vport as disabled to discard new events */
1508 vport->enabled = false;
1510 synchronize_irq(mlx5_get_msix_vec(esw->dev, MLX5_EQ_VEC_ASYNC));
1512 mlx5_modify_vport_admin_state(esw->dev,
1513 MLX5_QUERY_VPORT_STATE_IN_OP_MOD_ESW_VPORT,
1515 MLX5_ESW_VPORT_ADMIN_STATE_DOWN);
1516 /* Wait for current already scheduled events to complete */
1517 flush_workqueue(esw->work_queue);
1518 /* Disable events from this vport */
1519 arm_vport_context_events_cmd(esw->dev, vport->vport, 0);
1520 mutex_lock(&esw->state_lock);
1521 /* We don't assume VFs will cleanup after themselves.
1522 * Calling vport change handler while vport is disabled will cleanup
1523 * the vport resources.
1525 esw_vport_change_handle_locked(vport);
1526 vport->enabled_events = 0;
1528 esw_vport_disable_egress_acl(esw, vport);
1529 esw_vport_disable_ingress_acl(esw, vport);
1531 esw->enabled_vports--;
1532 mutex_unlock(&esw->state_lock);
1535 /* Public E-Switch API */
1536 int mlx5_eswitch_enable_sriov(struct mlx5_eswitch *esw, int nvfs, int mode)
1539 int i, enabled_events;
1541 if (!esw || !MLX5_CAP_GEN(esw->dev, vport_group_manager) ||
1542 MLX5_CAP_GEN(esw->dev, port_type) != MLX5_CAP_PORT_TYPE_ETH)
1545 if (!MLX5_CAP_GEN(esw->dev, eswitch_flow_table) ||
1546 !MLX5_CAP_ESW_FLOWTABLE_FDB(esw->dev, ft_support)) {
1547 esw_warn(esw->dev, "E-Switch FDB is not supported, aborting ...\n");
1551 if (!MLX5_CAP_ESW_INGRESS_ACL(esw->dev, ft_support))
1552 esw_warn(esw->dev, "E-Switch ingress ACL is not supported by FW\n");
1554 if (!MLX5_CAP_ESW_EGRESS_ACL(esw->dev, ft_support))
1555 esw_warn(esw->dev, "E-Switch engress ACL is not supported by FW\n");
1557 esw_info(esw->dev, "E-Switch enable SRIOV: nvfs(%d) mode (%d)\n", nvfs, mode);
1559 esw_disable_vport(esw, 0);
1561 if (mode == SRIOV_LEGACY)
1562 err = esw_create_legacy_fdb_table(esw, nvfs + 1);
1564 err = esw_create_offloads_fdb_table(esw, nvfs + 1);
1568 enabled_events = (mode == SRIOV_LEGACY) ? SRIOV_VPORT_EVENTS : UC_ADDR_CHANGE;
1569 for (i = 0; i <= nvfs; i++)
1570 esw_enable_vport(esw, i, enabled_events);
1572 esw_info(esw->dev, "SRIOV enabled: active vports(%d)\n",
1573 esw->enabled_vports);
1577 esw_enable_vport(esw, 0, UC_ADDR_CHANGE);
1581 void mlx5_eswitch_disable_sriov(struct mlx5_eswitch *esw)
1583 struct esw_mc_addr *mc_promisc;
1586 if (!esw || !MLX5_CAP_GEN(esw->dev, vport_group_manager) ||
1587 MLX5_CAP_GEN(esw->dev, port_type) != MLX5_CAP_PORT_TYPE_ETH)
1590 esw_info(esw->dev, "disable SRIOV: active vports(%d) mode(%d)\n",
1591 esw->enabled_vports, esw->mode);
1593 mc_promisc = esw->mc_promisc;
1595 for (i = 0; i < esw->total_vports; i++)
1596 esw_disable_vport(esw, i);
1598 if (mc_promisc && mc_promisc->uplink_rule)
1599 mlx5_del_flow_rule(mc_promisc->uplink_rule);
1601 if (esw->mode == SRIOV_LEGACY)
1602 esw_destroy_legacy_fdb_table(esw);
1604 esw_destroy_offloads_fdb_table(esw);
1606 esw->mode = SRIOV_NONE;
1607 /* VPORT 0 (PF) must be enabled back with non-sriov configuration */
1608 esw_enable_vport(esw, 0, UC_ADDR_CHANGE);
1611 int mlx5_eswitch_init(struct mlx5_core_dev *dev)
1613 int l2_table_size = 1 << MLX5_CAP_GEN(dev, log_max_l2_table);
1614 int total_vports = MLX5_TOTAL_VPORTS(dev);
1615 struct esw_mc_addr *mc_promisc;
1616 struct mlx5_eswitch *esw;
1620 if (!MLX5_CAP_GEN(dev, vport_group_manager) ||
1621 MLX5_CAP_GEN(dev, port_type) != MLX5_CAP_PORT_TYPE_ETH)
1625 "Total vports %d, l2 table size(%d), per vport: max uc(%d) max mc(%d)\n",
1626 total_vports, l2_table_size,
1627 MLX5_MAX_UC_PER_VPORT(dev),
1628 MLX5_MAX_MC_PER_VPORT(dev));
1630 esw = kzalloc(sizeof(*esw), GFP_KERNEL);
1636 esw->l2_table.bitmap = kcalloc(BITS_TO_LONGS(l2_table_size),
1637 sizeof(uintptr_t), GFP_KERNEL);
1638 if (!esw->l2_table.bitmap) {
1642 esw->l2_table.size = l2_table_size;
1644 mc_promisc = kzalloc(sizeof(*mc_promisc), GFP_KERNEL);
1649 esw->mc_promisc = mc_promisc;
1651 esw->work_queue = create_singlethread_workqueue("mlx5_esw_wq");
1652 if (!esw->work_queue) {
1657 esw->vports = kcalloc(total_vports, sizeof(struct mlx5_vport),
1664 mutex_init(&esw->state_lock);
1666 for (vport_num = 0; vport_num < total_vports; vport_num++) {
1667 struct mlx5_vport *vport = &esw->vports[vport_num];
1669 vport->vport = vport_num;
1671 INIT_WORK(&vport->vport_change_handler,
1672 esw_vport_change_handler);
1675 esw->total_vports = total_vports;
1676 esw->enabled_vports = 0;
1677 esw->mode = SRIOV_NONE;
1679 dev->priv.eswitch = esw;
1680 esw_enable_vport(esw, 0, UC_ADDR_CHANGE);
1681 /* VF Vports will be enabled when SRIOV is enabled */
1684 if (esw->work_queue)
1685 destroy_workqueue(esw->work_queue);
1686 kfree(esw->l2_table.bitmap);
1692 void mlx5_eswitch_cleanup(struct mlx5_eswitch *esw)
1694 if (!esw || !MLX5_CAP_GEN(esw->dev, vport_group_manager) ||
1695 MLX5_CAP_GEN(esw->dev, port_type) != MLX5_CAP_PORT_TYPE_ETH)
1698 esw_info(esw->dev, "cleanup\n");
1699 esw_disable_vport(esw, 0);
1701 esw->dev->priv.eswitch = NULL;
1702 destroy_workqueue(esw->work_queue);
1703 kfree(esw->l2_table.bitmap);
1704 kfree(esw->mc_promisc);
1709 void mlx5_eswitch_vport_event(struct mlx5_eswitch *esw, struct mlx5_eqe *eqe)
1711 struct mlx5_eqe_vport_change *vc_eqe = &eqe->data.vport_change;
1712 u16 vport_num = be16_to_cpu(vc_eqe->vport_num);
1713 struct mlx5_vport *vport;
1716 pr_warn("MLX5 E-Switch: vport %d got an event while eswitch is not initialized\n",
1721 vport = &esw->vports[vport_num];
1723 queue_work(esw->work_queue, &vport->vport_change_handler);
1726 /* Vport Administration */
1727 #define ESW_ALLOWED(esw) \
1728 (esw && MLX5_CAP_GEN(esw->dev, vport_group_manager) && mlx5_core_is_pf(esw->dev))
1729 #define LEGAL_VPORT(esw, vport) (vport >= 0 && vport < esw->total_vports)
1731 static void node_guid_gen_from_mac(u64 *node_guid, u8 mac[ETH_ALEN])
1733 ((u8 *)node_guid)[7] = mac[0];
1734 ((u8 *)node_guid)[6] = mac[1];
1735 ((u8 *)node_guid)[5] = mac[2];
1736 ((u8 *)node_guid)[4] = 0xff;
1737 ((u8 *)node_guid)[3] = 0xfe;
1738 ((u8 *)node_guid)[2] = mac[3];
1739 ((u8 *)node_guid)[1] = mac[4];
1740 ((u8 *)node_guid)[0] = mac[5];
1743 int mlx5_eswitch_set_vport_mac(struct mlx5_eswitch *esw,
1744 int vport, u8 mac[ETH_ALEN])
1746 struct mlx5_vport *evport;
1750 if (!ESW_ALLOWED(esw))
1752 if (!LEGAL_VPORT(esw, vport))
1755 evport = &esw->vports[vport];
1757 if (evport->spoofchk && !is_valid_ether_addr(mac)) {
1758 mlx5_core_warn(esw->dev,
1759 "MAC invalidation is not allowed when spoofchk is on, vport(%d)\n",
1764 err = mlx5_modify_nic_vport_mac_address(esw->dev, vport, mac);
1766 mlx5_core_warn(esw->dev,
1767 "Failed to mlx5_modify_nic_vport_mac vport(%d) err=(%d)\n",
1772 node_guid_gen_from_mac(&node_guid, mac);
1773 err = mlx5_modify_nic_vport_node_guid(esw->dev, vport, node_guid);
1775 mlx5_core_warn(esw->dev,
1776 "Failed to set vport %d node guid, err = %d. RDMA_CM will not function properly for this VF.\n",
1779 mutex_lock(&esw->state_lock);
1780 if (evport->enabled)
1781 err = esw_vport_ingress_config(esw, evport);
1782 mutex_unlock(&esw->state_lock);
1786 int mlx5_eswitch_set_vport_state(struct mlx5_eswitch *esw,
1787 int vport, int link_state)
1789 if (!ESW_ALLOWED(esw))
1791 if (!LEGAL_VPORT(esw, vport))
1794 return mlx5_modify_vport_admin_state(esw->dev,
1795 MLX5_QUERY_VPORT_STATE_IN_OP_MOD_ESW_VPORT,
1799 int mlx5_eswitch_get_vport_config(struct mlx5_eswitch *esw,
1800 int vport, struct ifla_vf_info *ivi)
1802 struct mlx5_vport *evport;
1806 if (!ESW_ALLOWED(esw))
1808 if (!LEGAL_VPORT(esw, vport))
1811 evport = &esw->vports[vport];
1813 memset(ivi, 0, sizeof(*ivi));
1814 ivi->vf = vport - 1;
1816 mlx5_query_nic_vport_mac_address(esw->dev, vport, ivi->mac);
1817 ivi->linkstate = mlx5_query_vport_admin_state(esw->dev,
1818 MLX5_QUERY_VPORT_STATE_IN_OP_MOD_ESW_VPORT,
1820 query_esw_vport_cvlan(esw->dev, vport, &vlan, &qos);
1823 ivi->spoofchk = evport->spoofchk;
1828 int mlx5_eswitch_set_vport_vlan(struct mlx5_eswitch *esw,
1829 int vport, u16 vlan, u8 qos)
1831 struct mlx5_vport *evport;
1835 if (!ESW_ALLOWED(esw))
1837 if (!LEGAL_VPORT(esw, vport) || (vlan > 4095) || (qos > 7))
1843 evport = &esw->vports[vport];
1845 err = modify_esw_vport_cvlan(esw->dev, vport, vlan, qos, set);
1849 mutex_lock(&esw->state_lock);
1850 evport->vlan = vlan;
1852 if (evport->enabled) {
1853 err = esw_vport_ingress_config(esw, evport);
1856 err = esw_vport_egress_config(esw, evport);
1860 mutex_unlock(&esw->state_lock);
1864 int mlx5_eswitch_set_vport_spoofchk(struct mlx5_eswitch *esw,
1865 int vport, bool spoofchk)
1867 struct mlx5_vport *evport;
1871 if (!ESW_ALLOWED(esw))
1873 if (!LEGAL_VPORT(esw, vport))
1876 evport = &esw->vports[vport];
1878 mutex_lock(&esw->state_lock);
1879 pschk = evport->spoofchk;
1880 evport->spoofchk = spoofchk;
1881 if (evport->enabled)
1882 err = esw_vport_ingress_config(esw, evport);
1884 evport->spoofchk = pschk;
1885 mutex_unlock(&esw->state_lock);
1890 int mlx5_eswitch_set_vport_trust(struct mlx5_eswitch *esw,
1891 int vport, bool setting)
1893 struct mlx5_vport *evport;
1895 if (!ESW_ALLOWED(esw))
1897 if (!LEGAL_VPORT(esw, vport))
1900 evport = &esw->vports[vport];
1902 mutex_lock(&esw->state_lock);
1903 evport->trusted = setting;
1904 if (evport->enabled)
1905 esw_vport_change_handle_locked(evport);
1906 mutex_unlock(&esw->state_lock);
1911 int mlx5_eswitch_get_vport_stats(struct mlx5_eswitch *esw,
1913 struct ifla_vf_stats *vf_stats)
1915 int outlen = MLX5_ST_SZ_BYTES(query_vport_counter_out);
1916 u32 in[MLX5_ST_SZ_DW(query_vport_counter_in)];
1920 if (!ESW_ALLOWED(esw))
1922 if (!LEGAL_VPORT(esw, vport))
1925 out = mlx5_vzalloc(outlen);
1929 memset(in, 0, sizeof(in));
1931 MLX5_SET(query_vport_counter_in, in, opcode,
1932 MLX5_CMD_OP_QUERY_VPORT_COUNTER);
1933 MLX5_SET(query_vport_counter_in, in, op_mod, 0);
1934 MLX5_SET(query_vport_counter_in, in, vport_number, vport);
1936 MLX5_SET(query_vport_counter_in, in, other_vport, 1);
1938 memset(out, 0, outlen);
1939 err = mlx5_cmd_exec(esw->dev, in, sizeof(in), out, outlen);
1943 #define MLX5_GET_CTR(p, x) \
1944 MLX5_GET64(query_vport_counter_out, p, x)
1946 memset(vf_stats, 0, sizeof(*vf_stats));
1947 vf_stats->rx_packets =
1948 MLX5_GET_CTR(out, received_eth_unicast.packets) +
1949 MLX5_GET_CTR(out, received_eth_multicast.packets) +
1950 MLX5_GET_CTR(out, received_eth_broadcast.packets);
1952 vf_stats->rx_bytes =
1953 MLX5_GET_CTR(out, received_eth_unicast.octets) +
1954 MLX5_GET_CTR(out, received_eth_multicast.octets) +
1955 MLX5_GET_CTR(out, received_eth_broadcast.octets);
1957 vf_stats->tx_packets =
1958 MLX5_GET_CTR(out, transmitted_eth_unicast.packets) +
1959 MLX5_GET_CTR(out, transmitted_eth_multicast.packets) +
1960 MLX5_GET_CTR(out, transmitted_eth_broadcast.packets);
1962 vf_stats->tx_bytes =
1963 MLX5_GET_CTR(out, transmitted_eth_unicast.octets) +
1964 MLX5_GET_CTR(out, transmitted_eth_multicast.octets) +
1965 MLX5_GET_CTR(out, transmitted_eth_broadcast.octets);
1967 vf_stats->multicast =
1968 MLX5_GET_CTR(out, received_eth_multicast.packets);
1970 vf_stats->broadcast =
1971 MLX5_GET_CTR(out, received_eth_broadcast.packets);