1 /*
2 * Copyright (c) 2015, Mellanox Technologies. All rights reserved.
3 *
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:
9 *
10 * Redistribution and use in source and binary forms, with or
11 * without modification, are permitted provided that the following
12 * conditions are met:
13 *
14 * - Redistributions of source code must retain the above
15 * copyright notice, this list of conditions and the following
16 * disclaimer.
17 *
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.
22 *
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
30 * SOFTWARE.
31 */
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"
39 #include "eswitch.h"
40 #include "fs_core.h"
42 #define UPLINK_VPORT 0xFFFF
44 enum {
45 MLX5_ACTION_NONE = 0,
46 MLX5_ACTION_ADD = 1,
47 MLX5_ACTION_DEL = 2,
48 };
50 /* Vport UC/MC hash node */
51 struct vport_addr {
52 struct l2addr_node node;
53 u8 action;
54 u32 vport;
55 struct mlx5_flow_handle *flow_rule;
56 bool mpfs; /* UC MAC was added to MPFs */
57 /* A flag indicating that mac was added due to mc promiscuous vport */
58 bool mc_promisc;
59 };
61 enum {
62 UC_ADDR_CHANGE = BIT(0),
63 MC_ADDR_CHANGE = BIT(1),
64 PROMISC_CHANGE = BIT(3),
65 };
67 /* Vport context events */
68 #define SRIOV_VPORT_EVENTS (UC_ADDR_CHANGE | \
69 MC_ADDR_CHANGE | \
70 PROMISC_CHANGE)
72 static int arm_vport_context_events_cmd(struct mlx5_core_dev *dev, u16 vport,
73 u32 events_mask)
74 {
75 int in[MLX5_ST_SZ_DW(modify_nic_vport_context_in)] = {0};
76 int out[MLX5_ST_SZ_DW(modify_nic_vport_context_out)] = {0};
77 void *nic_vport_ctx;
79 MLX5_SET(modify_nic_vport_context_in, in,
80 opcode, MLX5_CMD_OP_MODIFY_NIC_VPORT_CONTEXT);
81 MLX5_SET(modify_nic_vport_context_in, in, field_select.change_event, 1);
82 MLX5_SET(modify_nic_vport_context_in, in, vport_number, vport);
83 if (vport)
84 MLX5_SET(modify_nic_vport_context_in, in, other_vport, 1);
85 nic_vport_ctx = MLX5_ADDR_OF(modify_nic_vport_context_in,
86 in, nic_vport_context);
88 MLX5_SET(nic_vport_context, nic_vport_ctx, arm_change_event, 1);
90 if (events_mask & UC_ADDR_CHANGE)
91 MLX5_SET(nic_vport_context, nic_vport_ctx,
92 event_on_uc_address_change, 1);
93 if (events_mask & MC_ADDR_CHANGE)
94 MLX5_SET(nic_vport_context, nic_vport_ctx,
95 event_on_mc_address_change, 1);
96 if (events_mask & PROMISC_CHANGE)
97 MLX5_SET(nic_vport_context, nic_vport_ctx,
98 event_on_promisc_change, 1);
100 return mlx5_cmd_exec(dev, in, sizeof(in), out, sizeof(out));
101 }
103 /* E-Switch vport context HW commands */
104 static int modify_esw_vport_context_cmd(struct mlx5_core_dev *dev, u16 vport,
105 void *in, int inlen)
106 {
107 u32 out[MLX5_ST_SZ_DW(modify_esw_vport_context_out)] = {0};
109 MLX5_SET(modify_esw_vport_context_in, in, opcode,
110 MLX5_CMD_OP_MODIFY_ESW_VPORT_CONTEXT);
111 MLX5_SET(modify_esw_vport_context_in, in, vport_number, vport);
112 if (vport)
113 MLX5_SET(modify_esw_vport_context_in, in, other_vport, 1);
114 return mlx5_cmd_exec(dev, in, inlen, out, sizeof(out));
115 }
117 static int modify_esw_vport_cvlan(struct mlx5_core_dev *dev, u32 vport,
118 u16 vlan, u8 qos, u8 set_flags)
119 {
120 u32 in[MLX5_ST_SZ_DW(modify_esw_vport_context_in)] = {0};
122 if (!MLX5_CAP_ESW(dev, vport_cvlan_strip) ||
123 !MLX5_CAP_ESW(dev, vport_cvlan_insert_if_not_exist))
124 return -EOPNOTSUPP;
126 esw_debug(dev, "Set Vport[%d] VLAN %d qos %d set=%x\n",
127 vport, vlan, qos, set_flags);
129 if (set_flags & SET_VLAN_STRIP)
130 MLX5_SET(modify_esw_vport_context_in, in,
131 esw_vport_context.vport_cvlan_strip, 1);
133 if (set_flags & SET_VLAN_INSERT) {
134 /* insert only if no vlan in packet */
135 MLX5_SET(modify_esw_vport_context_in, in,
136 esw_vport_context.vport_cvlan_insert, 1);
138 MLX5_SET(modify_esw_vport_context_in, in,
139 esw_vport_context.cvlan_pcp, qos);
140 MLX5_SET(modify_esw_vport_context_in, in,
141 esw_vport_context.cvlan_id, vlan);
142 }
144 MLX5_SET(modify_esw_vport_context_in, in,
145 field_select.vport_cvlan_strip, 1);
146 MLX5_SET(modify_esw_vport_context_in, in,
147 field_select.vport_cvlan_insert, 1);
149 return modify_esw_vport_context_cmd(dev, vport, in, sizeof(in));
150 }
152 /* E-Switch FDB */
153 static struct mlx5_flow_handle *
154 __esw_fdb_set_vport_rule(struct mlx5_eswitch *esw, u32 vport, bool rx_rule,
155 u8 mac_c[ETH_ALEN], u8 mac_v[ETH_ALEN])
156 {
157 int match_header = (is_zero_ether_addr(mac_c) ? 0 :
158 MLX5_MATCH_OUTER_HEADERS);
159 struct mlx5_flow_handle *flow_rule = NULL;
160 struct mlx5_flow_act flow_act = {0};
161 struct mlx5_flow_destination dest = {};
162 struct mlx5_flow_spec *spec;
163 void *mv_misc = NULL;
164 void *mc_misc = NULL;
165 u8 *dmac_v = NULL;
166 u8 *dmac_c = NULL;
168 if (rx_rule)
169 match_header |= MLX5_MATCH_MISC_PARAMETERS;
171 spec = kvzalloc(sizeof(*spec), GFP_KERNEL);
172 if (!spec)
173 return NULL;
175 dmac_v = MLX5_ADDR_OF(fte_match_param, spec->match_value,
176 outer_headers.dmac_47_16);
177 dmac_c = MLX5_ADDR_OF(fte_match_param, spec->match_criteria,
178 outer_headers.dmac_47_16);
180 if (match_header & MLX5_MATCH_OUTER_HEADERS) {
181 ether_addr_copy(dmac_v, mac_v);
182 ether_addr_copy(dmac_c, mac_c);
183 }
185 if (match_header & MLX5_MATCH_MISC_PARAMETERS) {
186 mv_misc = MLX5_ADDR_OF(fte_match_param, spec->match_value,
187 misc_parameters);
188 mc_misc = MLX5_ADDR_OF(fte_match_param, spec->match_criteria,
189 misc_parameters);
190 MLX5_SET(fte_match_set_misc, mv_misc, source_port, UPLINK_VPORT);
191 MLX5_SET_TO_ONES(fte_match_set_misc, mc_misc, source_port);
192 }
194 dest.type = MLX5_FLOW_DESTINATION_TYPE_VPORT;
195 dest.vport.num = vport;
197 esw_debug(esw->dev,
198 "\tFDB add rule dmac_v(%pM) dmac_c(%pM) -> vport(%d)\n",
199 dmac_v, dmac_c, vport);
200 spec->match_criteria_enable = match_header;
201 flow_act.action = MLX5_FLOW_CONTEXT_ACTION_FWD_DEST;
202 flow_rule =
203 mlx5_add_flow_rules(esw->fdb_table.legacy.fdb, spec,
204 &flow_act, &dest, 1);
205 if (IS_ERR(flow_rule)) {
206 esw_warn(esw->dev,
207 "FDB: Failed to add flow rule: dmac_v(%pM) dmac_c(%pM) -> vport(%d), err(%ld)\n",
208 dmac_v, dmac_c, vport, PTR_ERR(flow_rule));
209 flow_rule = NULL;
210 }
212 kvfree(spec);
213 return flow_rule;
214 }
216 static struct mlx5_flow_handle *
217 esw_fdb_set_vport_rule(struct mlx5_eswitch *esw, u8 mac[ETH_ALEN], u32 vport)
218 {
219 u8 mac_c[ETH_ALEN];
221 eth_broadcast_addr(mac_c);
222 return __esw_fdb_set_vport_rule(esw, vport, false, mac_c, mac);
223 }
225 static struct mlx5_flow_handle *
226 esw_fdb_set_vport_allmulti_rule(struct mlx5_eswitch *esw, u32 vport)
227 {
228 u8 mac_c[ETH_ALEN];
229 u8 mac_v[ETH_ALEN];
231 eth_zero_addr(mac_c);
232 eth_zero_addr(mac_v);
233 mac_c[0] = 0x01;
234 mac_v[0] = 0x01;
235 return __esw_fdb_set_vport_rule(esw, vport, false, mac_c, mac_v);
236 }
238 static struct mlx5_flow_handle *
239 esw_fdb_set_vport_promisc_rule(struct mlx5_eswitch *esw, u32 vport)
240 {
241 u8 mac_c[ETH_ALEN];
242 u8 mac_v[ETH_ALEN];
244 eth_zero_addr(mac_c);
245 eth_zero_addr(mac_v);
246 return __esw_fdb_set_vport_rule(esw, vport, true, mac_c, mac_v);
247 }
249 static int esw_create_legacy_fdb_table(struct mlx5_eswitch *esw)
250 {
251 int inlen = MLX5_ST_SZ_BYTES(create_flow_group_in);
252 struct mlx5_flow_table_attr ft_attr = {};
253 struct mlx5_core_dev *dev = esw->dev;
254 struct mlx5_flow_namespace *root_ns;
255 struct mlx5_flow_table *fdb;
256 struct mlx5_flow_group *g;
257 void *match_criteria;
258 int table_size;
259 u32 *flow_group_in;
260 u8 *dmac;
261 int err = 0;
263 esw_debug(dev, "Create FDB log_max_size(%d)\n",
264 MLX5_CAP_ESW_FLOWTABLE_FDB(dev, log_max_ft_size));
266 root_ns = mlx5_get_flow_namespace(dev, MLX5_FLOW_NAMESPACE_FDB);
267 if (!root_ns) {
268 esw_warn(dev, "Failed to get FDB flow namespace\n");
269 return -EOPNOTSUPP;
270 }
272 flow_group_in = kvzalloc(inlen, GFP_KERNEL);
273 if (!flow_group_in)
274 return -ENOMEM;
276 table_size = BIT(MLX5_CAP_ESW_FLOWTABLE_FDB(dev, log_max_ft_size));
278 ft_attr.max_fte = table_size;
279 fdb = mlx5_create_flow_table(root_ns, &ft_attr);
280 if (IS_ERR(fdb)) {
281 err = PTR_ERR(fdb);
282 esw_warn(dev, "Failed to create FDB Table err %d\n", err);
283 goto out;
284 }
285 esw->fdb_table.legacy.fdb = fdb;
287 /* Addresses group : Full match unicast/multicast addresses */
288 MLX5_SET(create_flow_group_in, flow_group_in, match_criteria_enable,
289 MLX5_MATCH_OUTER_HEADERS);
290 match_criteria = MLX5_ADDR_OF(create_flow_group_in, flow_group_in, match_criteria);
291 dmac = MLX5_ADDR_OF(fte_match_param, match_criteria, outer_headers.dmac_47_16);
292 MLX5_SET(create_flow_group_in, flow_group_in, start_flow_index, 0);
293 /* Preserve 2 entries for allmulti and promisc rules*/
294 MLX5_SET(create_flow_group_in, flow_group_in, end_flow_index, table_size - 3);
295 eth_broadcast_addr(dmac);
296 g = mlx5_create_flow_group(fdb, flow_group_in);
297 if (IS_ERR(g)) {
298 err = PTR_ERR(g);
299 esw_warn(dev, "Failed to create flow group err(%d)\n", err);
300 goto out;
301 }
302 esw->fdb_table.legacy.addr_grp = g;
304 /* Allmulti group : One rule that forwards any mcast traffic */
305 MLX5_SET(create_flow_group_in, flow_group_in, match_criteria_enable,
306 MLX5_MATCH_OUTER_HEADERS);
307 MLX5_SET(create_flow_group_in, flow_group_in, start_flow_index, table_size - 2);
308 MLX5_SET(create_flow_group_in, flow_group_in, end_flow_index, table_size - 2);
309 eth_zero_addr(dmac);
310 dmac[0] = 0x01;
311 g = mlx5_create_flow_group(fdb, flow_group_in);
312 if (IS_ERR(g)) {
313 err = PTR_ERR(g);
314 esw_warn(dev, "Failed to create allmulti flow group err(%d)\n", err);
315 goto out;
316 }
317 esw->fdb_table.legacy.allmulti_grp = g;
319 /* Promiscuous group :
320 * One rule that forward all unmatched traffic from previous groups
321 */
322 eth_zero_addr(dmac);
323 MLX5_SET(create_flow_group_in, flow_group_in, match_criteria_enable,
324 MLX5_MATCH_MISC_PARAMETERS);
325 MLX5_SET_TO_ONES(fte_match_param, match_criteria, misc_parameters.source_port);
326 MLX5_SET(create_flow_group_in, flow_group_in, start_flow_index, table_size - 1);
327 MLX5_SET(create_flow_group_in, flow_group_in, end_flow_index, table_size - 1);
328 g = mlx5_create_flow_group(fdb, flow_group_in);
329 if (IS_ERR(g)) {
330 err = PTR_ERR(g);
331 esw_warn(dev, "Failed to create promisc flow group err(%d)\n", err);
332 goto out;
333 }
334 esw->fdb_table.legacy.promisc_grp = g;
336 out:
337 if (err) {
338 if (!IS_ERR_OR_NULL(esw->fdb_table.legacy.allmulti_grp)) {
339 mlx5_destroy_flow_group(esw->fdb_table.legacy.allmulti_grp);
340 esw->fdb_table.legacy.allmulti_grp = NULL;
341 }
342 if (!IS_ERR_OR_NULL(esw->fdb_table.legacy.addr_grp)) {
343 mlx5_destroy_flow_group(esw->fdb_table.legacy.addr_grp);
344 esw->fdb_table.legacy.addr_grp = NULL;
345 }
346 if (!IS_ERR_OR_NULL(esw->fdb_table.legacy.fdb)) {
347 mlx5_destroy_flow_table(esw->fdb_table.legacy.fdb);
348 esw->fdb_table.legacy.fdb = NULL;
349 }
350 }
352 kvfree(flow_group_in);
353 return err;
354 }
356 static void esw_destroy_legacy_fdb_table(struct mlx5_eswitch *esw)
357 {
358 if (!esw->fdb_table.legacy.fdb)
359 return;
361 esw_debug(esw->dev, "Destroy FDB Table\n");
362 mlx5_destroy_flow_group(esw->fdb_table.legacy.promisc_grp);
363 mlx5_destroy_flow_group(esw->fdb_table.legacy.allmulti_grp);
364 mlx5_destroy_flow_group(esw->fdb_table.legacy.addr_grp);
365 mlx5_destroy_flow_table(esw->fdb_table.legacy.fdb);
366 esw->fdb_table.legacy.fdb = NULL;
367 esw->fdb_table.legacy.addr_grp = NULL;
368 esw->fdb_table.legacy.allmulti_grp = NULL;
369 esw->fdb_table.legacy.promisc_grp = NULL;
370 }
372 /* E-Switch vport UC/MC lists management */
373 typedef int (*vport_addr_action)(struct mlx5_eswitch *esw,
374 struct vport_addr *vaddr);
376 static int esw_add_uc_addr(struct mlx5_eswitch *esw, struct vport_addr *vaddr)
377 {
378 u8 *mac = vaddr->node.addr;
379 u32 vport = vaddr->vport;
380 int err;
382 /* Skip mlx5_mpfs_add_mac for PFs,
383 * it is already done by the PF netdev in mlx5e_execute_l2_action
384 */
385 if (!vport)
386 goto fdb_add;
388 err = mlx5_mpfs_add_mac(esw->dev, mac);
389 if (err) {
390 esw_warn(esw->dev,
391 "Failed to add L2 table mac(%pM) for vport(%d), err(%d)\n",
392 mac, vport, err);
393 return err;
394 }
395 vaddr->mpfs = true;
397 fdb_add:
398 /* SRIOV is enabled: Forward UC MAC to vport */
399 if (esw->fdb_table.legacy.fdb && esw->mode == SRIOV_LEGACY)
400 vaddr->flow_rule = esw_fdb_set_vport_rule(esw, mac, vport);
402 esw_debug(esw->dev, "\tADDED UC MAC: vport[%d] %pM fr(%p)\n",
403 vport, mac, vaddr->flow_rule);
405 return 0;
406 }
408 static int esw_del_uc_addr(struct mlx5_eswitch *esw, struct vport_addr *vaddr)
409 {
410 u8 *mac = vaddr->node.addr;
411 u32 vport = vaddr->vport;
412 int err = 0;
414 /* Skip mlx5_mpfs_del_mac for PFs,
415 * it is already done by the PF netdev in mlx5e_execute_l2_action
416 */
417 if (!vport || !vaddr->mpfs)
418 goto fdb_del;
420 err = mlx5_mpfs_del_mac(esw->dev, mac);
421 if (err)
422 esw_warn(esw->dev,
423 "Failed to del L2 table mac(%pM) for vport(%d), err(%d)\n",
424 mac, vport, err);
425 vaddr->mpfs = false;
427 fdb_del:
428 if (vaddr->flow_rule)
429 mlx5_del_flow_rules(vaddr->flow_rule);
430 vaddr->flow_rule = NULL;
432 return 0;
433 }
435 static void update_allmulti_vports(struct mlx5_eswitch *esw,
436 struct vport_addr *vaddr,
437 struct esw_mc_addr *esw_mc)
438 {
439 u8 *mac = vaddr->node.addr;
440 u32 vport_idx = 0;
442 for (vport_idx = 0; vport_idx < esw->total_vports; vport_idx++) {
443 struct mlx5_vport *vport = &esw->vports[vport_idx];
444 struct hlist_head *vport_hash = vport->mc_list;
445 struct vport_addr *iter_vaddr =
446 l2addr_hash_find(vport_hash,
447 mac,
448 struct vport_addr);
449 if (IS_ERR_OR_NULL(vport->allmulti_rule) ||
450 vaddr->vport == vport_idx)
451 continue;
452 switch (vaddr->action) {
453 case MLX5_ACTION_ADD:
454 if (iter_vaddr)
455 continue;
456 iter_vaddr = l2addr_hash_add(vport_hash, mac,
457 struct vport_addr,
458 GFP_KERNEL);
459 if (!iter_vaddr) {
460 esw_warn(esw->dev,
461 "ALL-MULTI: Failed to add MAC(%pM) to vport[%d] DB\n",
462 mac, vport_idx);
463 continue;
464 }
465 iter_vaddr->vport = vport_idx;
466 iter_vaddr->flow_rule =
467 esw_fdb_set_vport_rule(esw,
468 mac,
469 vport_idx);
470 iter_vaddr->mc_promisc = true;
471 break;
472 case MLX5_ACTION_DEL:
473 if (!iter_vaddr)
474 continue;
475 mlx5_del_flow_rules(iter_vaddr->flow_rule);
476 l2addr_hash_del(iter_vaddr);
477 break;
478 }
479 }
480 }
482 static int esw_add_mc_addr(struct mlx5_eswitch *esw, struct vport_addr *vaddr)
483 {
484 struct hlist_head *hash = esw->mc_table;
485 struct esw_mc_addr *esw_mc;
486 u8 *mac = vaddr->node.addr;
487 u32 vport = vaddr->vport;
489 if (!esw->fdb_table.legacy.fdb)
490 return 0;
492 esw_mc = l2addr_hash_find(hash, mac, struct esw_mc_addr);
493 if (esw_mc)
494 goto add;
496 esw_mc = l2addr_hash_add(hash, mac, struct esw_mc_addr, GFP_KERNEL);
497 if (!esw_mc)
498 return -ENOMEM;
500 esw_mc->uplink_rule = /* Forward MC MAC to Uplink */
501 esw_fdb_set_vport_rule(esw, mac, UPLINK_VPORT);
503 /* Add this multicast mac to all the mc promiscuous vports */
504 update_allmulti_vports(esw, vaddr, esw_mc);
506 add:
507 /* If the multicast mac is added as a result of mc promiscuous vport,
508 * don't increment the multicast ref count
509 */
510 if (!vaddr->mc_promisc)
511 esw_mc->refcnt++;
513 /* Forward MC MAC to vport */
514 vaddr->flow_rule = esw_fdb_set_vport_rule(esw, mac, vport);
515 esw_debug(esw->dev,
516 "\tADDED MC MAC: vport[%d] %pM fr(%p) refcnt(%d) uplinkfr(%p)\n",
517 vport, mac, vaddr->flow_rule,
518 esw_mc->refcnt, esw_mc->uplink_rule);
519 return 0;
520 }
522 static int esw_del_mc_addr(struct mlx5_eswitch *esw, struct vport_addr *vaddr)
523 {
524 struct hlist_head *hash = esw->mc_table;
525 struct esw_mc_addr *esw_mc;
526 u8 *mac = vaddr->node.addr;
527 u32 vport = vaddr->vport;
529 if (!esw->fdb_table.legacy.fdb)
530 return 0;
532 esw_mc = l2addr_hash_find(hash, mac, struct esw_mc_addr);
533 if (!esw_mc) {
534 esw_warn(esw->dev,
535 "Failed to find eswitch MC addr for MAC(%pM) vport(%d)",
536 mac, vport);
537 return -EINVAL;
538 }
539 esw_debug(esw->dev,
540 "\tDELETE MC MAC: vport[%d] %pM fr(%p) refcnt(%d) uplinkfr(%p)\n",
541 vport, mac, vaddr->flow_rule, esw_mc->refcnt,
542 esw_mc->uplink_rule);
544 if (vaddr->flow_rule)
545 mlx5_del_flow_rules(vaddr->flow_rule);
546 vaddr->flow_rule = NULL;
548 /* If the multicast mac is added as a result of mc promiscuous vport,
549 * don't decrement the multicast ref count.
550 */
551 if (vaddr->mc_promisc || (--esw_mc->refcnt > 0))
552 return 0;
554 /* Remove this multicast mac from all the mc promiscuous vports */
555 update_allmulti_vports(esw, vaddr, esw_mc);
557 if (esw_mc->uplink_rule)
558 mlx5_del_flow_rules(esw_mc->uplink_rule);
560 l2addr_hash_del(esw_mc);
561 return 0;
562 }
564 /* Apply vport UC/MC list to HW l2 table and FDB table */
565 static void esw_apply_vport_addr_list(struct mlx5_eswitch *esw,
566 u32 vport_num, int list_type)
567 {
568 struct mlx5_vport *vport = &esw->vports[vport_num];
569 bool is_uc = list_type == MLX5_NVPRT_LIST_TYPE_UC;
570 vport_addr_action vport_addr_add;
571 vport_addr_action vport_addr_del;
572 struct vport_addr *addr;
573 struct l2addr_node *node;
574 struct hlist_head *hash;
575 struct hlist_node *tmp;
576 int hi;
578 vport_addr_add = is_uc ? esw_add_uc_addr :
579 esw_add_mc_addr;
580 vport_addr_del = is_uc ? esw_del_uc_addr :
581 esw_del_mc_addr;
583 hash = is_uc ? vport->uc_list : vport->mc_list;
584 for_each_l2hash_node(node, tmp, hash, hi) {
585 addr = container_of(node, struct vport_addr, node);
586 switch (addr->action) {
587 case MLX5_ACTION_ADD:
588 vport_addr_add(esw, addr);
589 addr->action = MLX5_ACTION_NONE;
590 break;
591 case MLX5_ACTION_DEL:
592 vport_addr_del(esw, addr);
593 l2addr_hash_del(addr);
594 break;
595 }
596 }
597 }
599 /* Sync vport UC/MC list from vport context */
600 static void esw_update_vport_addr_list(struct mlx5_eswitch *esw,
601 u32 vport_num, int list_type)
602 {
603 struct mlx5_vport *vport = &esw->vports[vport_num];
604 bool is_uc = list_type == MLX5_NVPRT_LIST_TYPE_UC;
605 u8 (*mac_list)[ETH_ALEN];
606 struct l2addr_node *node;
607 struct vport_addr *addr;
608 struct hlist_head *hash;
609 struct hlist_node *tmp;
610 int size;
611 int err;
612 int hi;
613 int i;
615 size = is_uc ? MLX5_MAX_UC_PER_VPORT(esw->dev) :
616 MLX5_MAX_MC_PER_VPORT(esw->dev);
618 mac_list = kcalloc(size, ETH_ALEN, GFP_KERNEL);
619 if (!mac_list)
620 return;
622 hash = is_uc ? vport->uc_list : vport->mc_list;
624 for_each_l2hash_node(node, tmp, hash, hi) {
625 addr = container_of(node, struct vport_addr, node);
626 addr->action = MLX5_ACTION_DEL;
627 }
629 if (!vport->enabled)
630 goto out;
632 err = mlx5_query_nic_vport_mac_list(esw->dev, vport_num, list_type,
633 mac_list, &size);
634 if (err)
635 goto out;
636 esw_debug(esw->dev, "vport[%d] context update %s list size (%d)\n",
637 vport_num, is_uc ? "UC" : "MC", size);
639 for (i = 0; i < size; i++) {
640 if (is_uc && !is_valid_ether_addr(mac_list[i]))
641 continue;
643 if (!is_uc && !is_multicast_ether_addr(mac_list[i]))
644 continue;
646 addr = l2addr_hash_find(hash, mac_list[i], struct vport_addr);
647 if (addr) {
648 addr->action = MLX5_ACTION_NONE;
649 /* If this mac was previously added because of allmulti
650 * promiscuous rx mode, its now converted to be original
651 * vport mac.
652 */
653 if (addr->mc_promisc) {
654 struct esw_mc_addr *esw_mc =
655 l2addr_hash_find(esw->mc_table,
656 mac_list[i],
657 struct esw_mc_addr);
658 if (!esw_mc) {
659 esw_warn(esw->dev,
660 "Failed to MAC(%pM) in mcast DB\n",
661 mac_list[i]);
662 continue;
663 }
664 esw_mc->refcnt++;
665 addr->mc_promisc = false;
666 }
667 continue;
668 }
670 addr = l2addr_hash_add(hash, mac_list[i], struct vport_addr,
671 GFP_KERNEL);
672 if (!addr) {
673 esw_warn(esw->dev,
674 "Failed to add MAC(%pM) to vport[%d] DB\n",
675 mac_list[i], vport_num);
676 continue;
677 }
678 addr->vport = vport_num;
679 addr->action = MLX5_ACTION_ADD;
680 }
681 out:
682 kfree(mac_list);
683 }
685 /* Sync vport UC/MC list from vport context
686 * Must be called after esw_update_vport_addr_list
687 */
688 static void esw_update_vport_mc_promisc(struct mlx5_eswitch *esw, u32 vport_num)
689 {
690 struct mlx5_vport *vport = &esw->vports[vport_num];
691 struct l2addr_node *node;
692 struct vport_addr *addr;
693 struct hlist_head *hash;
694 struct hlist_node *tmp;
695 int hi;
697 hash = vport->mc_list;
699 for_each_l2hash_node(node, tmp, esw->mc_table, hi) {
700 u8 *mac = node->addr;
702 addr = l2addr_hash_find(hash, mac, struct vport_addr);
703 if (addr) {
704 if (addr->action == MLX5_ACTION_DEL)
705 addr->action = MLX5_ACTION_NONE;
706 continue;
707 }
708 addr = l2addr_hash_add(hash, mac, struct vport_addr,
709 GFP_KERNEL);
710 if (!addr) {
711 esw_warn(esw->dev,
712 "Failed to add allmulti MAC(%pM) to vport[%d] DB\n",
713 mac, vport_num);
714 continue;
715 }
716 addr->vport = vport_num;
717 addr->action = MLX5_ACTION_ADD;
718 addr->mc_promisc = true;
719 }
720 }
722 /* Apply vport rx mode to HW FDB table */
723 static void esw_apply_vport_rx_mode(struct mlx5_eswitch *esw, u32 vport_num,
724 bool promisc, bool mc_promisc)
725 {
726 struct esw_mc_addr *allmulti_addr = &esw->mc_promisc;
727 struct mlx5_vport *vport = &esw->vports[vport_num];
729 if (IS_ERR_OR_NULL(vport->allmulti_rule) != mc_promisc)
730 goto promisc;
732 if (mc_promisc) {
733 vport->allmulti_rule =
734 esw_fdb_set_vport_allmulti_rule(esw, vport_num);
735 if (!allmulti_addr->uplink_rule)
736 allmulti_addr->uplink_rule =
737 esw_fdb_set_vport_allmulti_rule(esw,
738 UPLINK_VPORT);
739 allmulti_addr->refcnt++;
740 } else if (vport->allmulti_rule) {
741 mlx5_del_flow_rules(vport->allmulti_rule);
742 vport->allmulti_rule = NULL;
744 if (--allmulti_addr->refcnt > 0)
745 goto promisc;
747 if (allmulti_addr->uplink_rule)
748 mlx5_del_flow_rules(allmulti_addr->uplink_rule);
749 allmulti_addr->uplink_rule = NULL;
750 }
752 promisc:
753 if (IS_ERR_OR_NULL(vport->promisc_rule) != promisc)
754 return;
756 if (promisc) {
757 vport->promisc_rule = esw_fdb_set_vport_promisc_rule(esw,
758 vport_num);
759 } else if (vport->promisc_rule) {
760 mlx5_del_flow_rules(vport->promisc_rule);
761 vport->promisc_rule = NULL;
762 }
763 }
765 /* Sync vport rx mode from vport context */
766 static void esw_update_vport_rx_mode(struct mlx5_eswitch *esw, u32 vport_num)
767 {
768 struct mlx5_vport *vport = &esw->vports[vport_num];
769 int promisc_all = 0;
770 int promisc_uc = 0;
771 int promisc_mc = 0;
772 int err;
774 err = mlx5_query_nic_vport_promisc(esw->dev,
775 vport_num,
776 &promisc_uc,
777 &promisc_mc,
778 &promisc_all);
779 if (err)
780 return;
781 esw_debug(esw->dev, "vport[%d] context update rx mode promisc_all=%d, all_multi=%d\n",
782 vport_num, promisc_all, promisc_mc);
784 if (!vport->info.trusted || !vport->enabled) {
785 promisc_uc = 0;
786 promisc_mc = 0;
787 promisc_all = 0;
788 }
790 esw_apply_vport_rx_mode(esw, vport_num, promisc_all,
791 (promisc_all || promisc_mc));
792 }
794 static void esw_vport_change_handle_locked(struct mlx5_vport *vport)
795 {
796 struct mlx5_core_dev *dev = vport->dev;
797 struct mlx5_eswitch *esw = dev->priv.eswitch;
798 u8 mac[ETH_ALEN];
800 mlx5_query_nic_vport_mac_address(dev, vport->vport, mac);
801 esw_debug(dev, "vport[%d] Context Changed: perm mac: %pM\n",
802 vport->vport, mac);
804 if (vport->enabled_events & UC_ADDR_CHANGE) {
805 esw_update_vport_addr_list(esw, vport->vport,
806 MLX5_NVPRT_LIST_TYPE_UC);
807 esw_apply_vport_addr_list(esw, vport->vport,
808 MLX5_NVPRT_LIST_TYPE_UC);
809 }
811 if (vport->enabled_events & MC_ADDR_CHANGE) {
812 esw_update_vport_addr_list(esw, vport->vport,
813 MLX5_NVPRT_LIST_TYPE_MC);
814 }
816 if (vport->enabled_events & PROMISC_CHANGE) {
817 esw_update_vport_rx_mode(esw, vport->vport);
818 if (!IS_ERR_OR_NULL(vport->allmulti_rule))
819 esw_update_vport_mc_promisc(esw, vport->vport);
820 }
822 if (vport->enabled_events & (PROMISC_CHANGE | MC_ADDR_CHANGE)) {
823 esw_apply_vport_addr_list(esw, vport->vport,
824 MLX5_NVPRT_LIST_TYPE_MC);
825 }
827 esw_debug(esw->dev, "vport[%d] Context Changed: Done\n", vport->vport);
828 if (vport->enabled)
829 arm_vport_context_events_cmd(dev, vport->vport,
830 vport->enabled_events);
831 }
833 static void esw_vport_change_handler(struct work_struct *work)
834 {
835 struct mlx5_vport *vport =
836 container_of(work, struct mlx5_vport, vport_change_handler);
837 struct mlx5_eswitch *esw = vport->dev->priv.eswitch;
839 mutex_lock(&esw->state_lock);
840 esw_vport_change_handle_locked(vport);
841 mutex_unlock(&esw->state_lock);
842 }
844 static int esw_vport_enable_egress_acl(struct mlx5_eswitch *esw,
845 struct mlx5_vport *vport)
846 {
847 int inlen = MLX5_ST_SZ_BYTES(create_flow_group_in);
848 struct mlx5_flow_group *vlan_grp = NULL;
849 struct mlx5_flow_group *drop_grp = NULL;
850 struct mlx5_core_dev *dev = esw->dev;
851 struct mlx5_flow_namespace *root_ns;
852 struct mlx5_flow_table *acl;
853 void *match_criteria;
854 u32 *flow_group_in;
855 /* The egress acl table contains 2 rules:
856 * 1)Allow traffic with vlan_tag=vst_vlan_id
857 * 2)Drop all other traffic.
858 */
859 int table_size = 2;
860 int err = 0;
862 if (!MLX5_CAP_ESW_EGRESS_ACL(dev, ft_support))
863 return -EOPNOTSUPP;
865 if (!IS_ERR_OR_NULL(vport->egress.acl))
866 return 0;
868 esw_debug(dev, "Create vport[%d] egress ACL log_max_size(%d)\n",
869 vport->vport, MLX5_CAP_ESW_EGRESS_ACL(dev, log_max_ft_size));
871 root_ns = mlx5_get_flow_vport_acl_namespace(dev, MLX5_FLOW_NAMESPACE_ESW_EGRESS,
872 vport->vport);
873 if (!root_ns) {
874 esw_warn(dev, "Failed to get E-Switch egress flow namespace for vport (%d)\n", vport->vport);
875 return -EOPNOTSUPP;
876 }
878 flow_group_in = kvzalloc(inlen, GFP_KERNEL);
879 if (!flow_group_in)
880 return -ENOMEM;
882 acl = mlx5_create_vport_flow_table(root_ns, 0, table_size, 0, vport->vport);
883 if (IS_ERR(acl)) {
884 err = PTR_ERR(acl);
885 esw_warn(dev, "Failed to create E-Switch vport[%d] egress flow Table, err(%d)\n",
886 vport->vport, err);
887 goto out;
888 }
890 MLX5_SET(create_flow_group_in, flow_group_in, match_criteria_enable, MLX5_MATCH_OUTER_HEADERS);
891 match_criteria = MLX5_ADDR_OF(create_flow_group_in, flow_group_in, match_criteria);
892 MLX5_SET_TO_ONES(fte_match_param, match_criteria, outer_headers.cvlan_tag);
893 MLX5_SET_TO_ONES(fte_match_param, match_criteria, outer_headers.first_vid);
894 MLX5_SET(create_flow_group_in, flow_group_in, start_flow_index, 0);
895 MLX5_SET(create_flow_group_in, flow_group_in, end_flow_index, 0);
897 vlan_grp = mlx5_create_flow_group(acl, flow_group_in);
898 if (IS_ERR(vlan_grp)) {
899 err = PTR_ERR(vlan_grp);
900 esw_warn(dev, "Failed to create E-Switch vport[%d] egress allowed vlans flow group, err(%d)\n",
901 vport->vport, err);
902 goto out;
903 }
905 memset(flow_group_in, 0, inlen);
906 MLX5_SET(create_flow_group_in, flow_group_in, start_flow_index, 1);
907 MLX5_SET(create_flow_group_in, flow_group_in, end_flow_index, 1);
908 drop_grp = mlx5_create_flow_group(acl, flow_group_in);
909 if (IS_ERR(drop_grp)) {
910 err = PTR_ERR(drop_grp);
911 esw_warn(dev, "Failed to create E-Switch vport[%d] egress drop flow group, err(%d)\n",
912 vport->vport, err);
913 goto out;
914 }
916 vport->egress.acl = acl;
917 vport->egress.drop_grp = drop_grp;
918 vport->egress.allowed_vlans_grp = vlan_grp;
919 out:
920 kvfree(flow_group_in);
921 if (err && !IS_ERR_OR_NULL(vlan_grp))
922 mlx5_destroy_flow_group(vlan_grp);
923 if (err && !IS_ERR_OR_NULL(acl))
924 mlx5_destroy_flow_table(acl);
925 return err;
926 }
928 static void esw_vport_cleanup_egress_rules(struct mlx5_eswitch *esw,
929 struct mlx5_vport *vport)
930 {
931 if (!IS_ERR_OR_NULL(vport->egress.allowed_vlan))
932 mlx5_del_flow_rules(vport->egress.allowed_vlan);
934 if (!IS_ERR_OR_NULL(vport->egress.drop_rule))
935 mlx5_del_flow_rules(vport->egress.drop_rule);
937 vport->egress.allowed_vlan = NULL;
938 vport->egress.drop_rule = NULL;
939 }
941 static void esw_vport_disable_egress_acl(struct mlx5_eswitch *esw,
942 struct mlx5_vport *vport)
943 {
944 if (IS_ERR_OR_NULL(vport->egress.acl))
945 return;
947 esw_debug(esw->dev, "Destroy vport[%d] E-Switch egress ACL\n", vport->vport);
949 esw_vport_cleanup_egress_rules(esw, vport);
950 mlx5_destroy_flow_group(vport->egress.allowed_vlans_grp);
951 mlx5_destroy_flow_group(vport->egress.drop_grp);
952 mlx5_destroy_flow_table(vport->egress.acl);
953 vport->egress.allowed_vlans_grp = NULL;
954 vport->egress.drop_grp = NULL;
955 vport->egress.acl = NULL;
956 }
958 static int esw_vport_enable_ingress_acl(struct mlx5_eswitch *esw,
959 struct mlx5_vport *vport)
960 {
961 int inlen = MLX5_ST_SZ_BYTES(create_flow_group_in);
962 struct mlx5_core_dev *dev = esw->dev;
963 struct mlx5_flow_namespace *root_ns;
964 struct mlx5_flow_table *acl;
965 struct mlx5_flow_group *g;
966 void *match_criteria;
967 u32 *flow_group_in;
968 /* The ingress acl table contains 4 groups
969 * (2 active rules at the same time -
970 * 1 allow rule from one of the first 3 groups.
971 * 1 drop rule from the last group):
972 * 1)Allow untagged traffic with smac=original mac.
973 * 2)Allow untagged traffic.
974 * 3)Allow traffic with smac=original mac.
975 * 4)Drop all other traffic.
976 */
977 int table_size = 4;
978 int err = 0;
980 if (!MLX5_CAP_ESW_INGRESS_ACL(dev, ft_support))
981 return -EOPNOTSUPP;
983 if (!IS_ERR_OR_NULL(vport->ingress.acl))
984 return 0;
986 esw_debug(dev, "Create vport[%d] ingress ACL log_max_size(%d)\n",
987 vport->vport, MLX5_CAP_ESW_INGRESS_ACL(dev, log_max_ft_size));
989 root_ns = mlx5_get_flow_vport_acl_namespace(dev, MLX5_FLOW_NAMESPACE_ESW_INGRESS,
990 vport->vport);
991 if (!root_ns) {
992 esw_warn(dev, "Failed to get E-Switch ingress flow namespace for vport (%d)\n", vport->vport);
993 return -EOPNOTSUPP;
994 }
996 flow_group_in = kvzalloc(inlen, GFP_KERNEL);
997 if (!flow_group_in)
998 return -ENOMEM;
1000 acl = mlx5_create_vport_flow_table(root_ns, 0, table_size, 0, vport->vport);
1001 if (IS_ERR(acl)) {
1002 err = PTR_ERR(acl);
1003 esw_warn(dev, "Failed to create E-Switch vport[%d] ingress flow Table, err(%d)\n",
1004 vport->vport, err);
1005 goto out;
1006 }
1007 vport->ingress.acl = acl;
1009 match_criteria = MLX5_ADDR_OF(create_flow_group_in, flow_group_in, match_criteria);
1011 MLX5_SET(create_flow_group_in, flow_group_in, match_criteria_enable, MLX5_MATCH_OUTER_HEADERS);
1012 MLX5_SET_TO_ONES(fte_match_param, match_criteria, outer_headers.cvlan_tag);
1013 MLX5_SET_TO_ONES(fte_match_param, match_criteria, outer_headers.smac_47_16);
1014 MLX5_SET_TO_ONES(fte_match_param, match_criteria, outer_headers.smac_15_0);
1015 MLX5_SET(create_flow_group_in, flow_group_in, start_flow_index, 0);
1016 MLX5_SET(create_flow_group_in, flow_group_in, end_flow_index, 0);
1018 g = mlx5_create_flow_group(acl, flow_group_in);
1019 if (IS_ERR(g)) {
1020 err = PTR_ERR(g);
1021 esw_warn(dev, "Failed to create E-Switch vport[%d] ingress untagged spoofchk flow group, err(%d)\n",
1022 vport->vport, err);
1023 goto out;
1024 }
1025 vport->ingress.allow_untagged_spoofchk_grp = g;
1027 memset(flow_group_in, 0, inlen);
1028 MLX5_SET(create_flow_group_in, flow_group_in, match_criteria_enable, MLX5_MATCH_OUTER_HEADERS);
1029 MLX5_SET_TO_ONES(fte_match_param, match_criteria, outer_headers.cvlan_tag);
1030 MLX5_SET(create_flow_group_in, flow_group_in, start_flow_index, 1);
1031 MLX5_SET(create_flow_group_in, flow_group_in, end_flow_index, 1);
1033 g = mlx5_create_flow_group(acl, flow_group_in);
1034 if (IS_ERR(g)) {
1035 err = PTR_ERR(g);
1036 esw_warn(dev, "Failed to create E-Switch vport[%d] ingress untagged flow group, err(%d)\n",
1037 vport->vport, err);
1038 goto out;
1039 }
1040 vport->ingress.allow_untagged_only_grp = g;
1042 memset(flow_group_in, 0, inlen);
1043 MLX5_SET(create_flow_group_in, flow_group_in, match_criteria_enable, MLX5_MATCH_OUTER_HEADERS);
1044 MLX5_SET_TO_ONES(fte_match_param, match_criteria, outer_headers.smac_47_16);
1045 MLX5_SET_TO_ONES(fte_match_param, match_criteria, outer_headers.smac_15_0);
1046 MLX5_SET(create_flow_group_in, flow_group_in, start_flow_index, 2);
1047 MLX5_SET(create_flow_group_in, flow_group_in, end_flow_index, 2);
1049 g = mlx5_create_flow_group(acl, flow_group_in);
1050 if (IS_ERR(g)) {
1051 err = PTR_ERR(g);
1052 esw_warn(dev, "Failed to create E-Switch vport[%d] ingress spoofchk flow group, err(%d)\n",
1053 vport->vport, err);
1054 goto out;
1055 }
1056 vport->ingress.allow_spoofchk_only_grp = g;
1058 memset(flow_group_in, 0, inlen);
1059 MLX5_SET(create_flow_group_in, flow_group_in, start_flow_index, 3);
1060 MLX5_SET(create_flow_group_in, flow_group_in, end_flow_index, 3);
1062 g = mlx5_create_flow_group(acl, flow_group_in);
1063 if (IS_ERR(g)) {
1064 err = PTR_ERR(g);
1065 esw_warn(dev, "Failed to create E-Switch vport[%d] ingress drop flow group, err(%d)\n",
1066 vport->vport, err);
1067 goto out;
1068 }
1069 vport->ingress.drop_grp = g;
1071 out:
1072 if (err) {
1073 if (!IS_ERR_OR_NULL(vport->ingress.allow_spoofchk_only_grp))
1074 mlx5_destroy_flow_group(
1075 vport->ingress.allow_spoofchk_only_grp);
1076 if (!IS_ERR_OR_NULL(vport->ingress.allow_untagged_only_grp))
1077 mlx5_destroy_flow_group(
1078 vport->ingress.allow_untagged_only_grp);
1079 if (!IS_ERR_OR_NULL(vport->ingress.allow_untagged_spoofchk_grp))
1080 mlx5_destroy_flow_group(
1081 vport->ingress.allow_untagged_spoofchk_grp);
1082 if (!IS_ERR_OR_NULL(vport->ingress.acl))
1083 mlx5_destroy_flow_table(vport->ingress.acl);
1084 }
1086 kvfree(flow_group_in);
1087 return err;
1088 }
1090 static void esw_vport_cleanup_ingress_rules(struct mlx5_eswitch *esw,
1091 struct mlx5_vport *vport)
1092 {
1093 if (!IS_ERR_OR_NULL(vport->ingress.drop_rule))
1094 mlx5_del_flow_rules(vport->ingress.drop_rule);
1096 if (!IS_ERR_OR_NULL(vport->ingress.allow_rule))
1097 mlx5_del_flow_rules(vport->ingress.allow_rule);
1099 vport->ingress.drop_rule = NULL;
1100 vport->ingress.allow_rule = NULL;
1101 }
1103 static void esw_vport_disable_ingress_acl(struct mlx5_eswitch *esw,
1104 struct mlx5_vport *vport)
1105 {
1106 if (IS_ERR_OR_NULL(vport->ingress.acl))
1107 return;
1109 esw_debug(esw->dev, "Destroy vport[%d] E-Switch ingress ACL\n", vport->vport);
1111 esw_vport_cleanup_ingress_rules(esw, vport);
1112 mlx5_destroy_flow_group(vport->ingress.allow_spoofchk_only_grp);
1113 mlx5_destroy_flow_group(vport->ingress.allow_untagged_only_grp);
1114 mlx5_destroy_flow_group(vport->ingress.allow_untagged_spoofchk_grp);
1115 mlx5_destroy_flow_group(vport->ingress.drop_grp);
1116 mlx5_destroy_flow_table(vport->ingress.acl);
1117 vport->ingress.acl = NULL;
1118 vport->ingress.drop_grp = NULL;
1119 vport->ingress.allow_spoofchk_only_grp = NULL;
1120 vport->ingress.allow_untagged_only_grp = NULL;
1121 vport->ingress.allow_untagged_spoofchk_grp = NULL;
1122 }
1124 static int esw_vport_ingress_config(struct mlx5_eswitch *esw,
1125 struct mlx5_vport *vport)
1126 {
1127 struct mlx5_fc *counter = vport->ingress.drop_counter;
1128 struct mlx5_flow_destination drop_ctr_dst = {0};
1129 struct mlx5_flow_destination *dst = NULL;
1130 struct mlx5_flow_act flow_act = {0};
1131 struct mlx5_flow_spec *spec;
1132 int dest_num = 0;
1133 int err = 0;
1134 u8 *smac_v;
1136 if (vport->info.spoofchk && !is_valid_ether_addr(vport->info.mac)) {
1137 mlx5_core_warn(esw->dev,
1138 "vport[%d] configure ingress rules failed, illegal mac with spoofchk\n",
1139 vport->vport);
1140 return -EPERM;
1141 }
1143 esw_vport_cleanup_ingress_rules(esw, vport);
1145 if (!vport->info.vlan && !vport->info.qos && !vport->info.spoofchk) {
1146 esw_vport_disable_ingress_acl(esw, vport);
1147 return 0;
1148 }
1150 err = esw_vport_enable_ingress_acl(esw, vport);
1151 if (err) {
1152 mlx5_core_warn(esw->dev,
1153 "failed to enable ingress acl (%d) on vport[%d]\n",
1154 err, vport->vport);
1155 return err;
1156 }
1158 esw_debug(esw->dev,
1159 "vport[%d] configure ingress rules, vlan(%d) qos(%d)\n",
1160 vport->vport, vport->info.vlan, vport->info.qos);
1162 spec = kvzalloc(sizeof(*spec), GFP_KERNEL);
1163 if (!spec) {
1164 err = -ENOMEM;
1165 goto out;
1166 }
1168 if (vport->info.vlan || vport->info.qos)
1169 MLX5_SET_TO_ONES(fte_match_param, spec->match_criteria, outer_headers.cvlan_tag);
1171 if (vport->info.spoofchk) {
1172 MLX5_SET_TO_ONES(fte_match_param, spec->match_criteria, outer_headers.smac_47_16);
1173 MLX5_SET_TO_ONES(fte_match_param, spec->match_criteria, outer_headers.smac_15_0);
1174 smac_v = MLX5_ADDR_OF(fte_match_param,
1175 spec->match_value,
1176 outer_headers.smac_47_16);
1177 ether_addr_copy(smac_v, vport->info.mac);
1178 }
1180 spec->match_criteria_enable = MLX5_MATCH_OUTER_HEADERS;
1181 flow_act.action = MLX5_FLOW_CONTEXT_ACTION_ALLOW;
1182 vport->ingress.allow_rule =
1183 mlx5_add_flow_rules(vport->ingress.acl, spec,
1184 &flow_act, NULL, 0);
1185 if (IS_ERR(vport->ingress.allow_rule)) {
1186 err = PTR_ERR(vport->ingress.allow_rule);
1187 esw_warn(esw->dev,
1188 "vport[%d] configure ingress allow rule, err(%d)\n",
1189 vport->vport, err);
1190 vport->ingress.allow_rule = NULL;
1191 goto out;
1192 }
1194 memset(spec, 0, sizeof(*spec));
1195 flow_act.action = MLX5_FLOW_CONTEXT_ACTION_DROP;
1197 /* Attach drop flow counter */
1198 if (counter) {
1199 flow_act.action |= MLX5_FLOW_CONTEXT_ACTION_COUNT;
1200 drop_ctr_dst.type = MLX5_FLOW_DESTINATION_TYPE_COUNTER;
1201 drop_ctr_dst.counter = counter;
1202 dst = &drop_ctr_dst;
1203 dest_num++;
1204 }
1205 vport->ingress.drop_rule =
1206 mlx5_add_flow_rules(vport->ingress.acl, spec,
1207 &flow_act, dst, dest_num);
1208 if (IS_ERR(vport->ingress.drop_rule)) {
1209 err = PTR_ERR(vport->ingress.drop_rule);
1210 esw_warn(esw->dev,
1211 "vport[%d] configure ingress drop rule, err(%d)\n",
1212 vport->vport, err);
1213 vport->ingress.drop_rule = NULL;
1214 goto out;
1215 }
1217 out:
1218 if (err)
1219 esw_vport_cleanup_ingress_rules(esw, vport);
1220 kvfree(spec);
1221 return err;
1222 }
1224 static int esw_vport_egress_config(struct mlx5_eswitch *esw,
1225 struct mlx5_vport *vport)
1226 {
1227 struct mlx5_fc *counter = vport->egress.drop_counter;
1228 struct mlx5_flow_destination drop_ctr_dst = {0};
1229 struct mlx5_flow_destination *dst = NULL;
1230 struct mlx5_flow_act flow_act = {0};
1231 struct mlx5_flow_spec *spec;
1232 int dest_num = 0;
1233 int err = 0;
1235 esw_vport_cleanup_egress_rules(esw, vport);
1237 if (!vport->info.vlan && !vport->info.qos) {
1238 esw_vport_disable_egress_acl(esw, vport);
1239 return 0;
1240 }
1242 err = esw_vport_enable_egress_acl(esw, vport);
1243 if (err) {
1244 mlx5_core_warn(esw->dev,
1245 "failed to enable egress acl (%d) on vport[%d]\n",
1246 err, vport->vport);
1247 return err;
1248 }
1250 esw_debug(esw->dev,
1251 "vport[%d] configure egress rules, vlan(%d) qos(%d)\n",
1252 vport->vport, vport->info.vlan, vport->info.qos);
1254 spec = kvzalloc(sizeof(*spec), GFP_KERNEL);
1255 if (!spec) {
1256 err = -ENOMEM;
1257 goto out;
1258 }
1260 /* Allowed vlan rule */
1261 MLX5_SET_TO_ONES(fte_match_param, spec->match_criteria, outer_headers.cvlan_tag);
1262 MLX5_SET_TO_ONES(fte_match_param, spec->match_value, outer_headers.cvlan_tag);
1263 MLX5_SET_TO_ONES(fte_match_param, spec->match_criteria, outer_headers.first_vid);
1264 MLX5_SET(fte_match_param, spec->match_value, outer_headers.first_vid, vport->info.vlan);
1266 spec->match_criteria_enable = MLX5_MATCH_OUTER_HEADERS;
1267 flow_act.action = MLX5_FLOW_CONTEXT_ACTION_ALLOW;
1268 vport->egress.allowed_vlan =
1269 mlx5_add_flow_rules(vport->egress.acl, spec,
1270 &flow_act, NULL, 0);
1271 if (IS_ERR(vport->egress.allowed_vlan)) {
1272 err = PTR_ERR(vport->egress.allowed_vlan);
1273 esw_warn(esw->dev,
1274 "vport[%d] configure egress allowed vlan rule failed, err(%d)\n",
1275 vport->vport, err);
1276 vport->egress.allowed_vlan = NULL;
1277 goto out;
1278 }
1280 /* Drop others rule (star rule) */
1281 memset(spec, 0, sizeof(*spec));
1282 flow_act.action = MLX5_FLOW_CONTEXT_ACTION_DROP;
1284 /* Attach egress drop flow counter */
1285 if (counter) {
1286 flow_act.action |= MLX5_FLOW_CONTEXT_ACTION_COUNT;
1287 drop_ctr_dst.type = MLX5_FLOW_DESTINATION_TYPE_COUNTER;
1288 drop_ctr_dst.counter = counter;
1289 dst = &drop_ctr_dst;
1290 dest_num++;
1291 }
1292 vport->egress.drop_rule =
1293 mlx5_add_flow_rules(vport->egress.acl, spec,
1294 &flow_act, dst, dest_num);
1295 if (IS_ERR(vport->egress.drop_rule)) {
1296 err = PTR_ERR(vport->egress.drop_rule);
1297 esw_warn(esw->dev,
1298 "vport[%d] configure egress drop rule failed, err(%d)\n",
1299 vport->vport, err);
1300 vport->egress.drop_rule = NULL;
1301 }
1302 out:
1303 kvfree(spec);
1304 return err;
1305 }
1307 /* Vport QoS management */
1308 static int esw_create_tsar(struct mlx5_eswitch *esw)
1309 {
1310 u32 tsar_ctx[MLX5_ST_SZ_DW(scheduling_context)] = {0};
1311 struct mlx5_core_dev *dev = esw->dev;
1312 int err;
1314 if (!MLX5_CAP_GEN(dev, qos) || !MLX5_CAP_QOS(dev, esw_scheduling))
1315 return 0;
1317 if (esw->qos.enabled)
1318 return -EEXIST;
1320 err = mlx5_create_scheduling_element_cmd(dev,
1321 SCHEDULING_HIERARCHY_E_SWITCH,
1322 tsar_ctx,
1323 &esw->qos.root_tsar_id);
1324 if (err) {
1325 esw_warn(esw->dev, "E-Switch create TSAR failed (%d)\n", err);
1326 return err;
1327 }
1329 esw->qos.enabled = true;
1330 return 0;
1331 }
1333 static void esw_destroy_tsar(struct mlx5_eswitch *esw)
1334 {
1335 int err;
1337 if (!esw->qos.enabled)
1338 return;
1340 err = mlx5_destroy_scheduling_element_cmd(esw->dev,
1341 SCHEDULING_HIERARCHY_E_SWITCH,
1342 esw->qos.root_tsar_id);
1343 if (err)
1344 esw_warn(esw->dev, "E-Switch destroy TSAR failed (%d)\n", err);
1346 esw->qos.enabled = false;
1347 }
1349 static int esw_vport_enable_qos(struct mlx5_eswitch *esw, int vport_num,
1350 u32 initial_max_rate, u32 initial_bw_share)
1351 {
1352 u32 sched_ctx[MLX5_ST_SZ_DW(scheduling_context)] = {0};
1353 struct mlx5_vport *vport = &esw->vports[vport_num];
1354 struct mlx5_core_dev *dev = esw->dev;
1355 void *vport_elem;
1356 int err = 0;
1358 if (!esw->qos.enabled || !MLX5_CAP_GEN(dev, qos) ||
1359 !MLX5_CAP_QOS(dev, esw_scheduling))
1360 return 0;
1362 if (vport->qos.enabled)
1363 return -EEXIST;
1365 MLX5_SET(scheduling_context, sched_ctx, element_type,
1366 SCHEDULING_CONTEXT_ELEMENT_TYPE_VPORT);
1367 vport_elem = MLX5_ADDR_OF(scheduling_context, sched_ctx,
1368 element_attributes);
1369 MLX5_SET(vport_element, vport_elem, vport_number, vport_num);
1370 MLX5_SET(scheduling_context, sched_ctx, parent_element_id,
1371 esw->qos.root_tsar_id);
1372 MLX5_SET(scheduling_context, sched_ctx, max_average_bw,
1373 initial_max_rate);
1374 MLX5_SET(scheduling_context, sched_ctx, bw_share, initial_bw_share);
1376 err = mlx5_create_scheduling_element_cmd(dev,
1377 SCHEDULING_HIERARCHY_E_SWITCH,
1378 sched_ctx,
1379 &vport->qos.esw_tsar_ix);
1380 if (err) {
1381 esw_warn(esw->dev, "E-Switch create TSAR vport element failed (vport=%d,err=%d)\n",
1382 vport_num, err);
1383 return err;
1384 }
1386 vport->qos.enabled = true;
1387 return 0;
1388 }
1390 static void esw_vport_disable_qos(struct mlx5_eswitch *esw, int vport_num)
1391 {
1392 struct mlx5_vport *vport = &esw->vports[vport_num];
1393 int err = 0;
1395 if (!vport->qos.enabled)
1396 return;
1398 err = mlx5_destroy_scheduling_element_cmd(esw->dev,
1399 SCHEDULING_HIERARCHY_E_SWITCH,
1400 vport->qos.esw_tsar_ix);
1401 if (err)
1402 esw_warn(esw->dev, "E-Switch destroy TSAR vport element failed (vport=%d,err=%d)\n",
1403 vport_num, err);
1405 vport->qos.enabled = false;
1406 }
1408 static int esw_vport_qos_config(struct mlx5_eswitch *esw, int vport_num,
1409 u32 max_rate, u32 bw_share)
1410 {
1411 u32 sched_ctx[MLX5_ST_SZ_DW(scheduling_context)] = {0};
1412 struct mlx5_vport *vport = &esw->vports[vport_num];
1413 struct mlx5_core_dev *dev = esw->dev;
1414 void *vport_elem;
1415 u32 bitmask = 0;
1416 int err = 0;
1418 if (!MLX5_CAP_GEN(dev, qos) || !MLX5_CAP_QOS(dev, esw_scheduling))
1419 return -EOPNOTSUPP;
1421 if (!vport->qos.enabled)
1422 return -EIO;
1424 MLX5_SET(scheduling_context, sched_ctx, element_type,
1425 SCHEDULING_CONTEXT_ELEMENT_TYPE_VPORT);
1426 vport_elem = MLX5_ADDR_OF(scheduling_context, sched_ctx,
1427 element_attributes);
1428 MLX5_SET(vport_element, vport_elem, vport_number, vport_num);
1429 MLX5_SET(scheduling_context, sched_ctx, parent_element_id,
1430 esw->qos.root_tsar_id);
1431 MLX5_SET(scheduling_context, sched_ctx, max_average_bw,
1432 max_rate);
1433 MLX5_SET(scheduling_context, sched_ctx, bw_share, bw_share);
1434 bitmask |= MODIFY_SCHEDULING_ELEMENT_IN_MODIFY_BITMASK_MAX_AVERAGE_BW;
1435 bitmask |= MODIFY_SCHEDULING_ELEMENT_IN_MODIFY_BITMASK_BW_SHARE;
1437 err = mlx5_modify_scheduling_element_cmd(dev,
1438 SCHEDULING_HIERARCHY_E_SWITCH,
1439 sched_ctx,
1440 vport->qos.esw_tsar_ix,
1441 bitmask);
1442 if (err) {
1443 esw_warn(esw->dev, "E-Switch modify TSAR vport element failed (vport=%d,err=%d)\n",
1444 vport_num, err);
1445 return err;
1446 }
1448 return 0;
1449 }
1451 static void node_guid_gen_from_mac(u64 *node_guid, u8 mac[ETH_ALEN])
1452 {
1453 ((u8 *)node_guid)[7] = mac[0];
1454 ((u8 *)node_guid)[6] = mac[1];
1455 ((u8 *)node_guid)[5] = mac[2];
1456 ((u8 *)node_guid)[4] = 0xff;
1457 ((u8 *)node_guid)[3] = 0xfe;
1458 ((u8 *)node_guid)[2] = mac[3];
1459 ((u8 *)node_guid)[1] = mac[4];
1460 ((u8 *)node_guid)[0] = mac[5];
1461 }
1463 static void esw_apply_vport_conf(struct mlx5_eswitch *esw,
1464 struct mlx5_vport *vport)
1465 {
1466 int vport_num = vport->vport;
1468 if (!vport_num)
1469 return;
1471 mlx5_modify_vport_admin_state(esw->dev,
1472 MLX5_VPORT_STATE_OP_MOD_ESW_VPORT,
1473 vport_num,
1474 vport->info.link_state);
1475 mlx5_modify_nic_vport_mac_address(esw->dev, vport_num, vport->info.mac);
1476 mlx5_modify_nic_vport_node_guid(esw->dev, vport_num, vport->info.node_guid);
1477 modify_esw_vport_cvlan(esw->dev, vport_num, vport->info.vlan, vport->info.qos,
1478 (vport->info.vlan || vport->info.qos));
1480 /* Only legacy mode needs ACLs */
1481 if (esw->mode == SRIOV_LEGACY) {
1482 esw_vport_ingress_config(esw, vport);
1483 esw_vport_egress_config(esw, vport);
1484 }
1485 }
1487 static void esw_vport_create_drop_counters(struct mlx5_vport *vport)
1488 {
1489 struct mlx5_core_dev *dev = vport->dev;
1491 if (MLX5_CAP_ESW_INGRESS_ACL(dev, flow_counter)) {
1492 vport->ingress.drop_counter = mlx5_fc_create(dev, false);
1493 if (IS_ERR(vport->ingress.drop_counter)) {
1494 esw_warn(dev,
1495 "vport[%d] configure ingress drop rule counter failed\n",
1496 vport->vport);
1497 vport->ingress.drop_counter = NULL;
1498 }
1499 }
1501 if (MLX5_CAP_ESW_EGRESS_ACL(dev, flow_counter)) {
1502 vport->egress.drop_counter = mlx5_fc_create(dev, false);
1503 if (IS_ERR(vport->egress.drop_counter)) {
1504 esw_warn(dev,
1505 "vport[%d] configure egress drop rule counter failed\n",
1506 vport->vport);
1507 vport->egress.drop_counter = NULL;
1508 }
1509 }
1510 }
1512 static void esw_vport_destroy_drop_counters(struct mlx5_vport *vport)
1513 {
1514 struct mlx5_core_dev *dev = vport->dev;
1516 if (vport->ingress.drop_counter)
1517 mlx5_fc_destroy(dev, vport->ingress.drop_counter);
1518 if (vport->egress.drop_counter)
1519 mlx5_fc_destroy(dev, vport->egress.drop_counter);
1520 }
1522 static void esw_enable_vport(struct mlx5_eswitch *esw, int vport_num,
1523 int enable_events)
1524 {
1525 struct mlx5_vport *vport = &esw->vports[vport_num];
1527 mutex_lock(&esw->state_lock);
1528 WARN_ON(vport->enabled);
1530 esw_debug(esw->dev, "Enabling VPORT(%d)\n", vport_num);
1532 /* Create steering drop counters for ingress and egress ACLs */
1533 if (vport_num && esw->mode == SRIOV_LEGACY)
1534 esw_vport_create_drop_counters(vport);
1536 /* Restore old vport configuration */
1537 esw_apply_vport_conf(esw, vport);
1539 /* Attach vport to the eswitch rate limiter */
1540 if (esw_vport_enable_qos(esw, vport_num, vport->info.max_rate,
1541 vport->qos.bw_share))
1542 esw_warn(esw->dev, "Failed to attach vport %d to eswitch rate limiter", vport_num);
1544 /* Sync with current vport context */
1545 vport->enabled_events = enable_events;
1546 vport->enabled = true;
1548 /* only PF is trusted by default */
1549 if (!vport_num)
1550 vport->info.trusted = true;
1552 esw_vport_change_handle_locked(vport);
1554 esw->enabled_vports++;
1555 esw_debug(esw->dev, "Enabled VPORT(%d)\n", vport_num);
1556 mutex_unlock(&esw->state_lock);
1557 }
1559 static void esw_disable_vport(struct mlx5_eswitch *esw, int vport_num)
1560 {
1561 struct mlx5_vport *vport = &esw->vports[vport_num];
1563 if (!vport->enabled)
1564 return;
1566 esw_debug(esw->dev, "Disabling vport(%d)\n", vport_num);
1567 /* Mark this vport as disabled to discard new events */
1568 vport->enabled = false;
1570 synchronize_irq(pci_irq_vector(esw->dev->pdev, MLX5_EQ_VEC_ASYNC));
1571 /* Wait for current already scheduled events to complete */
1572 flush_workqueue(esw->work_queue);
1573 /* Disable events from this vport */
1574 arm_vport_context_events_cmd(esw->dev, vport->vport, 0);
1575 mutex_lock(&esw->state_lock);
1576 /* We don't assume VFs will cleanup after themselves.
1577 * Calling vport change handler while vport is disabled will cleanup
1578 * the vport resources.
1579 */
1580 esw_vport_change_handle_locked(vport);
1581 vport->enabled_events = 0;
1582 esw_vport_disable_qos(esw, vport_num);
1583 if (vport_num && esw->mode == SRIOV_LEGACY) {
1584 mlx5_modify_vport_admin_state(esw->dev,
1585 MLX5_VPORT_STATE_OP_MOD_ESW_VPORT,
1586 vport_num,
1587 MLX5_VPORT_ADMIN_STATE_DOWN);
1588 esw_vport_disable_egress_acl(esw, vport);
1589 esw_vport_disable_ingress_acl(esw, vport);
1590 esw_vport_destroy_drop_counters(vport);
1591 }
1592 esw->enabled_vports--;
1593 mutex_unlock(&esw->state_lock);
1594 }
1596 /* Public E-Switch API */
1597 #define ESW_ALLOWED(esw) ((esw) && MLX5_ESWITCH_MANAGER((esw)->dev))
1600 int mlx5_eswitch_enable_sriov(struct mlx5_eswitch *esw, int nvfs, int mode)
1601 {
1602 int err;
1603 int i, enabled_events;
1605 if (!ESW_ALLOWED(esw) ||
1606 !MLX5_CAP_ESW_FLOWTABLE_FDB(esw->dev, ft_support)) {
1607 esw_warn(esw->dev, "E-Switch FDB is not supported, aborting ...\n");
1608 return -EOPNOTSUPP;
1609 }
1611 if (!MLX5_CAP_ESW_INGRESS_ACL(esw->dev, ft_support))
1612 esw_warn(esw->dev, "E-Switch ingress ACL is not supported by FW\n");
1614 if (!MLX5_CAP_ESW_EGRESS_ACL(esw->dev, ft_support))
1615 esw_warn(esw->dev, "E-Switch engress ACL is not supported by FW\n");
1617 esw_info(esw->dev, "E-Switch enable SRIOV: nvfs(%d) mode (%d)\n", nvfs, mode);
1618 esw->mode = mode;
1620 if (mode == SRIOV_LEGACY) {
1621 err = esw_create_legacy_fdb_table(esw);
1622 } else {
1623 mlx5_reload_interface(esw->dev, MLX5_INTERFACE_PROTOCOL_IB);
1625 err = esw_offloads_init(esw, nvfs + 1);
1626 }
1628 if (err)
1629 goto abort;
1631 err = esw_create_tsar(esw);
1632 if (err)
1633 esw_warn(esw->dev, "Failed to create eswitch TSAR");
1635 /* Don't enable vport events when in SRIOV_OFFLOADS mode, since:
1636 * 1. L2 table (MPFS) is programmed by PF/VF representors netdevs set_rx_mode
1637 * 2. FDB/Eswitch is programmed by user space tools
1638 */
1639 enabled_events = (mode == SRIOV_LEGACY) ? SRIOV_VPORT_EVENTS : 0;
1640 for (i = 0; i <= nvfs; i++)
1641 esw_enable_vport(esw, i, enabled_events);
1643 esw_info(esw->dev, "SRIOV enabled: active vports(%d)\n",
1644 esw->enabled_vports);
1645 return 0;
1647 abort:
1648 esw->mode = SRIOV_NONE;
1650 if (mode == SRIOV_OFFLOADS)
1651 mlx5_reload_interface(esw->dev, MLX5_INTERFACE_PROTOCOL_IB);
1653 return err;
1654 }
1656 void mlx5_eswitch_disable_sriov(struct mlx5_eswitch *esw)
1657 {
1658 struct esw_mc_addr *mc_promisc;
1659 int old_mode;
1660 int nvports;
1661 int i;
1663 if (!ESW_ALLOWED(esw) || esw->mode == SRIOV_NONE)
1664 return;
1666 esw_info(esw->dev, "disable SRIOV: active vports(%d) mode(%d)\n",
1667 esw->enabled_vports, esw->mode);
1669 mc_promisc = &esw->mc_promisc;
1670 nvports = esw->enabled_vports;
1672 for (i = 0; i < esw->total_vports; i++)
1673 esw_disable_vport(esw, i);
1675 if (mc_promisc && mc_promisc->uplink_rule)
1676 mlx5_del_flow_rules(mc_promisc->uplink_rule);
1678 esw_destroy_tsar(esw);
1680 if (esw->mode == SRIOV_LEGACY)
1681 esw_destroy_legacy_fdb_table(esw);
1682 else if (esw->mode == SRIOV_OFFLOADS)
1683 esw_offloads_cleanup(esw, nvports);
1685 old_mode = esw->mode;
1686 esw->mode = SRIOV_NONE;
1688 if (old_mode == SRIOV_OFFLOADS)
1689 mlx5_reload_interface(esw->dev, MLX5_INTERFACE_PROTOCOL_IB);
1690 }
1692 int mlx5_eswitch_init(struct mlx5_core_dev *dev)
1693 {
1694 int total_vports = MLX5_TOTAL_VPORTS(dev);
1695 struct mlx5_eswitch *esw;
1696 int vport_num;
1697 int err;
1699 if (!MLX5_ESWITCH_MANAGER(dev))
1700 return 0;
1702 esw_info(dev,
1703 "Total vports %d, per vport: max uc(%d) max mc(%d)\n",
1704 total_vports,
1705 MLX5_MAX_UC_PER_VPORT(dev),
1706 MLX5_MAX_MC_PER_VPORT(dev));
1708 esw = kzalloc(sizeof(*esw), GFP_KERNEL);
1709 if (!esw)
1710 return -ENOMEM;
1712 esw->dev = dev;
1714 esw->work_queue = create_singlethread_workqueue("mlx5_esw_wq");
1715 if (!esw->work_queue) {
1716 err = -ENOMEM;
1717 goto abort;
1718 }
1720 esw->vports = kcalloc(total_vports, sizeof(struct mlx5_vport),
1721 GFP_KERNEL);
1722 if (!esw->vports) {
1723 err = -ENOMEM;
1724 goto abort;
1725 }
1727 err = esw_offloads_init_reps(esw);
1728 if (err)
1729 goto abort;
1731 hash_init(esw->offloads.encap_tbl);
1732 hash_init(esw->offloads.mod_hdr_tbl);
1733 mutex_init(&esw->state_lock);
1735 for (vport_num = 0; vport_num < total_vports; vport_num++) {
1736 struct mlx5_vport *vport = &esw->vports[vport_num];
1738 vport->vport = vport_num;
1739 vport->info.link_state = MLX5_VPORT_ADMIN_STATE_AUTO;
1740 vport->dev = dev;
1741 INIT_WORK(&vport->vport_change_handler,
1742 esw_vport_change_handler);
1743 }
1745 esw->total_vports = total_vports;
1746 esw->enabled_vports = 0;
1747 esw->mode = SRIOV_NONE;
1748 esw->offloads.inline_mode = MLX5_INLINE_MODE_NONE;
1749 if (MLX5_CAP_ESW_FLOWTABLE_FDB(dev, encap) &&
1750 MLX5_CAP_ESW_FLOWTABLE_FDB(dev, decap))
1751 esw->offloads.encap = DEVLINK_ESWITCH_ENCAP_MODE_BASIC;
1752 else
1753 esw->offloads.encap = DEVLINK_ESWITCH_ENCAP_MODE_NONE;
1755 dev->priv.eswitch = esw;
1756 return 0;
1757 abort:
1758 if (esw->work_queue)
1759 destroy_workqueue(esw->work_queue);
1760 esw_offloads_cleanup_reps(esw);
1761 kfree(esw->vports);
1762 kfree(esw);
1763 return err;
1764 }
1766 void mlx5_eswitch_cleanup(struct mlx5_eswitch *esw)
1767 {
1768 if (!esw || !MLX5_ESWITCH_MANAGER(esw->dev))
1769 return;
1771 esw_info(esw->dev, "cleanup\n");
1773 esw->dev->priv.eswitch = NULL;
1774 destroy_workqueue(esw->work_queue);
1775 esw_offloads_cleanup_reps(esw);
1776 kfree(esw->vports);
1777 kfree(esw);
1778 }
1780 void mlx5_eswitch_vport_event(struct mlx5_eswitch *esw, struct mlx5_eqe *eqe)
1781 {
1782 struct mlx5_eqe_vport_change *vc_eqe = &eqe->data.vport_change;
1783 u16 vport_num = be16_to_cpu(vc_eqe->vport_num);
1784 struct mlx5_vport *vport;
1786 if (!esw) {
1787 pr_warn("MLX5 E-Switch: vport %d got an event while eswitch is not initialized\n",
1788 vport_num);
1789 return;
1790 }
1792 vport = &esw->vports[vport_num];
1793 if (vport->enabled)
1794 queue_work(esw->work_queue, &vport->vport_change_handler);
1795 }
1797 /* Vport Administration */
1798 #define LEGAL_VPORT(esw, vport) (vport >= 0 && vport < esw->total_vports)
1800 int mlx5_eswitch_set_vport_mac(struct mlx5_eswitch *esw,
1801 int vport, u8 mac[ETH_ALEN])
1802 {
1803 struct mlx5_vport *evport;
1804 u64 node_guid;
1805 int err = 0;
1807 if (!MLX5_CAP_GEN(esw->dev, vport_group_manager))
1808 return -EPERM;
1809 if (!LEGAL_VPORT(esw, vport) || is_multicast_ether_addr(mac))
1810 return -EINVAL;
1812 mutex_lock(&esw->state_lock);
1813 evport = &esw->vports[vport];
1815 if (evport->info.spoofchk && !is_valid_ether_addr(mac)) {
1816 mlx5_core_warn(esw->dev,
1817 "MAC invalidation is not allowed when spoofchk is on, vport(%d)\n",
1818 vport);
1819 err = -EPERM;
1820 goto unlock;
1821 }
1823 err = mlx5_modify_nic_vport_mac_address(esw->dev, vport, mac);
1824 if (err) {
1825 mlx5_core_warn(esw->dev,
1826 "Failed to mlx5_modify_nic_vport_mac vport(%d) err=(%d)\n",
1827 vport, err);
1828 goto unlock;
1829 }
1831 node_guid_gen_from_mac(&node_guid, mac);
1832 err = mlx5_modify_nic_vport_node_guid(esw->dev, vport, node_guid);
1833 if (err)
1834 mlx5_core_warn(esw->dev,
1835 "Failed to set vport %d node guid, err = %d. RDMA_CM will not function properly for this VF.\n",
1836 vport, err);
1838 ether_addr_copy(evport->info.mac, mac);
1839 evport->info.node_guid = node_guid;
1840 if (evport->enabled && esw->mode == SRIOV_LEGACY)
1841 err = esw_vport_ingress_config(esw, evport);
1843 unlock:
1844 mutex_unlock(&esw->state_lock);
1845 return err;
1846 }
1848 int mlx5_eswitch_set_vport_state(struct mlx5_eswitch *esw,
1849 int vport, int link_state)
1850 {
1851 struct mlx5_vport *evport;
1852 int err = 0;
1854 if (!ESW_ALLOWED(esw))
1855 return -EPERM;
1856 if (!LEGAL_VPORT(esw, vport))
1857 return -EINVAL;
1859 mutex_lock(&esw->state_lock);
1860 evport = &esw->vports[vport];
1862 err = mlx5_modify_vport_admin_state(esw->dev,
1863 MLX5_VPORT_STATE_OP_MOD_ESW_VPORT,
1864 vport, link_state);
1865 if (err) {
1866 mlx5_core_warn(esw->dev,
1867 "Failed to set vport %d link state, err = %d",
1868 vport, err);
1869 goto unlock;
1870 }
1872 evport->info.link_state = link_state;
1874 unlock:
1875 mutex_unlock(&esw->state_lock);
1876 return 0;
1877 }
1879 int mlx5_eswitch_get_vport_config(struct mlx5_eswitch *esw,
1880 int vport, struct ifla_vf_info *ivi)
1881 {
1882 struct mlx5_vport *evport;
1884 if (!MLX5_CAP_GEN(esw->dev, vport_group_manager))
1885 return -EPERM;
1886 if (!LEGAL_VPORT(esw, vport))
1887 return -EINVAL;
1889 evport = &esw->vports[vport];
1891 memset(ivi, 0, sizeof(*ivi));
1892 ivi->vf = vport - 1;
1894 mutex_lock(&esw->state_lock);
1895 ether_addr_copy(ivi->mac, evport->info.mac);
1896 ivi->linkstate = evport->info.link_state;
1897 ivi->vlan = evport->info.vlan;
1898 ivi->qos = evport->info.qos;
1899 ivi->spoofchk = evport->info.spoofchk;
1900 ivi->trusted = evport->info.trusted;
1901 ivi->min_tx_rate = evport->info.min_rate;
1902 ivi->max_tx_rate = evport->info.max_rate;
1903 mutex_unlock(&esw->state_lock);
1905 return 0;
1906 }
1908 int __mlx5_eswitch_set_vport_vlan(struct mlx5_eswitch *esw,
1909 int vport, u16 vlan, u8 qos, u8 set_flags)
1910 {
1911 struct mlx5_vport *evport;
1912 int err = 0;
1914 if (!ESW_ALLOWED(esw))
1915 return -EPERM;
1916 if (!LEGAL_VPORT(esw, vport) || (vlan > 4095) || (qos > 7))
1917 return -EINVAL;
1919 mutex_lock(&esw->state_lock);
1920 evport = &esw->vports[vport];
1922 err = modify_esw_vport_cvlan(esw->dev, vport, vlan, qos, set_flags);
1923 if (err)
1924 goto unlock;
1926 evport->info.vlan = vlan;
1927 evport->info.qos = qos;
1928 if (evport->enabled && esw->mode == SRIOV_LEGACY) {
1929 err = esw_vport_ingress_config(esw, evport);
1930 if (err)
1931 goto unlock;
1932 err = esw_vport_egress_config(esw, evport);
1933 }
1935 unlock:
1936 mutex_unlock(&esw->state_lock);
1937 return err;
1938 }
1940 int mlx5_eswitch_set_vport_vlan(struct mlx5_eswitch *esw,
1941 int vport, u16 vlan, u8 qos)
1942 {
1943 u8 set_flags = 0;
1945 if (vlan || qos)
1946 set_flags = SET_VLAN_STRIP | SET_VLAN_INSERT;
1948 return __mlx5_eswitch_set_vport_vlan(esw, vport, vlan, qos, set_flags);
1949 }
1951 int mlx5_eswitch_set_vport_spoofchk(struct mlx5_eswitch *esw,
1952 int vport, bool spoofchk)
1953 {
1954 struct mlx5_vport *evport;
1955 bool pschk;
1956 int err = 0;
1958 if (!ESW_ALLOWED(esw))
1959 return -EPERM;
1960 if (!LEGAL_VPORT(esw, vport))
1961 return -EINVAL;
1963 mutex_lock(&esw->state_lock);
1964 evport = &esw->vports[vport];
1965 pschk = evport->info.spoofchk;
1966 evport->info.spoofchk = spoofchk;
1967 if (evport->enabled && esw->mode == SRIOV_LEGACY)
1968 err = esw_vport_ingress_config(esw, evport);
1969 if (err)
1970 evport->info.spoofchk = pschk;
1971 mutex_unlock(&esw->state_lock);
1973 return err;
1974 }
1976 int mlx5_eswitch_set_vport_trust(struct mlx5_eswitch *esw,
1977 int vport, bool setting)
1978 {
1979 struct mlx5_vport *evport;
1981 if (!ESW_ALLOWED(esw))
1982 return -EPERM;
1983 if (!LEGAL_VPORT(esw, vport))
1984 return -EINVAL;
1986 mutex_lock(&esw->state_lock);
1987 evport = &esw->vports[vport];
1988 evport->info.trusted = setting;
1989 if (evport->enabled)
1990 esw_vport_change_handle_locked(evport);
1991 mutex_unlock(&esw->state_lock);
1993 return 0;
1994 }
1996 static u32 calculate_vports_min_rate_divider(struct mlx5_eswitch *esw)
1997 {
1998 u32 fw_max_bw_share = MLX5_CAP_QOS(esw->dev, max_tsar_bw_share);
1999 struct mlx5_vport *evport;
2000 u32 max_guarantee = 0;
2001 int i;
2003 for (i = 0; i < esw->total_vports; i++) {
2004 evport = &esw->vports[i];
2005 if (!evport->enabled || evport->info.min_rate < max_guarantee)
2006 continue;
2007 max_guarantee = evport->info.min_rate;
2008 }
2010 return max_t(u32, max_guarantee / fw_max_bw_share, 1);
2011 }
2013 static int normalize_vports_min_rate(struct mlx5_eswitch *esw, u32 divider)
2014 {
2015 u32 fw_max_bw_share = MLX5_CAP_QOS(esw->dev, max_tsar_bw_share);
2016 struct mlx5_vport *evport;
2017 u32 vport_max_rate;
2018 u32 vport_min_rate;
2019 u32 bw_share;
2020 int err;
2021 int i;
2023 for (i = 0; i < esw->total_vports; i++) {
2024 evport = &esw->vports[i];
2025 if (!evport->enabled)
2026 continue;
2027 vport_min_rate = evport->info.min_rate;
2028 vport_max_rate = evport->info.max_rate;
2029 bw_share = MLX5_MIN_BW_SHARE;
2031 if (vport_min_rate)
2032 bw_share = MLX5_RATE_TO_BW_SHARE(vport_min_rate,
2033 divider,
2034 fw_max_bw_share);
2036 if (bw_share == evport->qos.bw_share)
2037 continue;
2039 err = esw_vport_qos_config(esw, i, vport_max_rate,
2040 bw_share);
2041 if (!err)
2042 evport->qos.bw_share = bw_share;
2043 else
2044 return err;
2045 }
2047 return 0;
2048 }
2050 int mlx5_eswitch_set_vport_rate(struct mlx5_eswitch *esw, int vport,
2051 u32 max_rate, u32 min_rate)
2052 {
2053 u32 fw_max_bw_share = MLX5_CAP_QOS(esw->dev, max_tsar_bw_share);
2054 bool min_rate_supported = MLX5_CAP_QOS(esw->dev, esw_bw_share) &&
2055 fw_max_bw_share >= MLX5_MIN_BW_SHARE;
2056 bool max_rate_supported = MLX5_CAP_QOS(esw->dev, esw_rate_limit);
2057 struct mlx5_vport *evport;
2058 u32 previous_min_rate;
2059 u32 divider;
2060 int err = 0;
2062 if (!ESW_ALLOWED(esw))
2063 return -EPERM;
2064 if (!LEGAL_VPORT(esw, vport))
2065 return -EINVAL;
2066 if ((min_rate && !min_rate_supported) || (max_rate && !max_rate_supported))
2067 return -EOPNOTSUPP;
2069 mutex_lock(&esw->state_lock);
2070 evport = &esw->vports[vport];
2072 if (min_rate == evport->info.min_rate)
2073 goto set_max_rate;
2075 previous_min_rate = evport->info.min_rate;
2076 evport->info.min_rate = min_rate;
2077 divider = calculate_vports_min_rate_divider(esw);
2078 err = normalize_vports_min_rate(esw, divider);
2079 if (err) {
2080 evport->info.min_rate = previous_min_rate;
2081 goto unlock;
2082 }
2084 set_max_rate:
2085 if (max_rate == evport->info.max_rate)
2086 goto unlock;
2088 err = esw_vport_qos_config(esw, vport, max_rate, evport->qos.bw_share);
2089 if (!err)
2090 evport->info.max_rate = max_rate;
2092 unlock:
2093 mutex_unlock(&esw->state_lock);
2094 return err;
2095 }
2097 static int mlx5_eswitch_query_vport_drop_stats(struct mlx5_core_dev *dev,
2098 int vport_idx,
2099 struct mlx5_vport_drop_stats *stats)
2100 {
2101 struct mlx5_eswitch *esw = dev->priv.eswitch;
2102 struct mlx5_vport *vport = &esw->vports[vport_idx];
2103 u64 rx_discard_vport_down, tx_discard_vport_down;
2104 u64 bytes = 0;
2105 int err = 0;
2107 if (!vport->enabled || esw->mode != SRIOV_LEGACY)
2108 return 0;
2110 if (vport->egress.drop_counter)
2111 mlx5_fc_query(dev, vport->egress.drop_counter,
2112 &stats->rx_dropped, &bytes);
2114 if (vport->ingress.drop_counter)
2115 mlx5_fc_query(dev, vport->ingress.drop_counter,
2116 &stats->tx_dropped, &bytes);
2118 if (!MLX5_CAP_GEN(dev, receive_discard_vport_down) &&
2119 !MLX5_CAP_GEN(dev, transmit_discard_vport_down))
2120 return 0;
2122 err = mlx5_query_vport_down_stats(dev, vport_idx,
2123 &rx_discard_vport_down,
2124 &tx_discard_vport_down);
2125 if (err)
2126 return err;
2128 if (MLX5_CAP_GEN(dev, receive_discard_vport_down))
2129 stats->rx_dropped += rx_discard_vport_down;
2130 if (MLX5_CAP_GEN(dev, transmit_discard_vport_down))
2131 stats->tx_dropped += tx_discard_vport_down;
2133 return 0;
2134 }
2136 int mlx5_eswitch_get_vport_stats(struct mlx5_eswitch *esw,
2137 int vport,
2138 struct ifla_vf_stats *vf_stats)
2139 {
2140 int outlen = MLX5_ST_SZ_BYTES(query_vport_counter_out);
2141 u32 in[MLX5_ST_SZ_DW(query_vport_counter_in)] = {0};
2142 struct mlx5_vport_drop_stats stats = {0};
2143 int err = 0;
2144 u32 *out;
2146 if (!ESW_ALLOWED(esw))
2147 return -EPERM;
2148 if (!LEGAL_VPORT(esw, vport))
2149 return -EINVAL;
2151 out = kvzalloc(outlen, GFP_KERNEL);
2152 if (!out)
2153 return -ENOMEM;
2155 MLX5_SET(query_vport_counter_in, in, opcode,
2156 MLX5_CMD_OP_QUERY_VPORT_COUNTER);
2157 MLX5_SET(query_vport_counter_in, in, op_mod, 0);
2158 MLX5_SET(query_vport_counter_in, in, vport_number, vport);
2159 if (vport)
2160 MLX5_SET(query_vport_counter_in, in, other_vport, 1);
2162 memset(out, 0, outlen);
2163 err = mlx5_cmd_exec(esw->dev, in, sizeof(in), out, outlen);
2164 if (err)
2165 goto free_out;
2167 #define MLX5_GET_CTR(p, x) \
2168 MLX5_GET64(query_vport_counter_out, p, x)
2170 memset(vf_stats, 0, sizeof(*vf_stats));
2171 vf_stats->rx_packets =
2172 MLX5_GET_CTR(out, received_eth_unicast.packets) +
2173 MLX5_GET_CTR(out, received_ib_unicast.packets) +
2174 MLX5_GET_CTR(out, received_eth_multicast.packets) +
2175 MLX5_GET_CTR(out, received_ib_multicast.packets) +
2176 MLX5_GET_CTR(out, received_eth_broadcast.packets);
2178 vf_stats->rx_bytes =
2179 MLX5_GET_CTR(out, received_eth_unicast.octets) +
2180 MLX5_GET_CTR(out, received_ib_unicast.octets) +
2181 MLX5_GET_CTR(out, received_eth_multicast.octets) +
2182 MLX5_GET_CTR(out, received_ib_multicast.octets) +
2183 MLX5_GET_CTR(out, received_eth_broadcast.octets);
2185 vf_stats->tx_packets =
2186 MLX5_GET_CTR(out, transmitted_eth_unicast.packets) +
2187 MLX5_GET_CTR(out, transmitted_ib_unicast.packets) +
2188 MLX5_GET_CTR(out, transmitted_eth_multicast.packets) +
2189 MLX5_GET_CTR(out, transmitted_ib_multicast.packets) +
2190 MLX5_GET_CTR(out, transmitted_eth_broadcast.packets);
2192 vf_stats->tx_bytes =
2193 MLX5_GET_CTR(out, transmitted_eth_unicast.octets) +
2194 MLX5_GET_CTR(out, transmitted_ib_unicast.octets) +
2195 MLX5_GET_CTR(out, transmitted_eth_multicast.octets) +
2196 MLX5_GET_CTR(out, transmitted_ib_multicast.octets) +
2197 MLX5_GET_CTR(out, transmitted_eth_broadcast.octets);
2199 vf_stats->multicast =
2200 MLX5_GET_CTR(out, received_eth_multicast.packets) +
2201 MLX5_GET_CTR(out, received_ib_multicast.packets);
2203 vf_stats->broadcast =
2204 MLX5_GET_CTR(out, received_eth_broadcast.packets);
2206 err = mlx5_eswitch_query_vport_drop_stats(esw->dev, vport, &stats);
2207 if (err)
2208 goto free_out;
2209 vf_stats->rx_dropped = stats.rx_dropped;
2210 vf_stats->tx_dropped = stats.tx_dropped;
2212 free_out:
2213 kvfree(out);
2214 return err;
2215 }
2217 u8 mlx5_eswitch_mode(struct mlx5_eswitch *esw)
2218 {
2219 return ESW_ALLOWED(esw) ? esw->mode : SRIOV_NONE;
2220 }
2221 EXPORT_SYMBOL_GPL(mlx5_eswitch_mode);