xref: /freebsd/sys/dev/mlx5/mlx5_core/mlx5_eswitch.c (revision 76ed99ed)
1 /*-
2  * Copyright (c) 2013-2017, Mellanox Technologies, Ltd.  All rights reserved.
3  *
4  * Redistribution and use in source and binary forms, with or without
5  * modification, are permitted provided that the following conditions
6  * are met:
7  * 1. Redistributions of source code must retain the above copyright
8  *    notice, this list of conditions and the following disclaimer.
9  * 2. Redistributions in binary form must reproduce the above copyright
10  *    notice, this list of conditions and the following disclaimer in the
11  *    documentation and/or other materials provided with the distribution.
12  *
13  * THIS SOFTWARE IS PROVIDED BY AUTHOR AND CONTRIBUTORS `AS IS' AND
14  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
15  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
16  * ARE DISCLAIMED.  IN NO EVENT SHALL AUTHOR OR CONTRIBUTORS BE LIABLE
17  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
18  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
19  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
20  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
21  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
22  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
23  * SUCH DAMAGE.
24  */
25 
26 #include "opt_rss.h"
27 #include "opt_ratelimit.h"
28 
29 #include <linux/etherdevice.h>
30 #include <dev/mlx5/driver.h>
31 #include <dev/mlx5/mlx5_ifc.h>
32 #include <dev/mlx5/vport.h>
33 #include <dev/mlx5/fs.h>
34 #include <dev/mlx5/mpfs.h>
35 #include <dev/mlx5/mlx5_core/mlx5_core.h>
36 #include <dev/mlx5/mlx5_core/eswitch.h>
37 
38 #define UPLINK_VPORT 0xFFFF
39 
40 #define MLX5_DEBUG_ESWITCH_MASK BIT(3)
41 
42 #define esw_info(dev, format, ...)				\
43 	printf("mlx5_core: INFO: ""(%s): E-Switch: " format, (dev)->priv.name, ##__VA_ARGS__)
44 
45 #define esw_warn(dev, format, ...)				\
46 	printf("mlx5_core: WARN: ""(%s): E-Switch: " format, (dev)->priv.name, ##__VA_ARGS__)
47 
48 #define esw_debug(dev, format, ...)				\
49 	mlx5_core_dbg_mask(dev, MLX5_DEBUG_ESWITCH_MASK, format, ##__VA_ARGS__)
50 
51 enum {
52 	MLX5_ACTION_NONE = 0,
53 	MLX5_ACTION_ADD  = 1,
54 	MLX5_ACTION_DEL  = 2,
55 };
56 
57 /* E-Switch UC L2 table hash node */
58 struct esw_uc_addr {
59 	struct l2addr_node node;
60 	u32                table_index;
61 	u32                vport;
62 };
63 
64 /* E-Switch MC FDB table hash node */
65 struct esw_mc_addr { /* SRIOV only */
66 	struct l2addr_node     node;
67 	struct mlx5_flow_rule *uplink_rule; /* Forward to uplink rule */
68 	u32                    refcnt;
69 };
70 
71 /* Vport UC/MC hash node */
72 struct vport_addr {
73 	struct l2addr_node     node;
74 	u8                     action;
75 	u32                    vport;
76 	struct mlx5_flow_rule *flow_rule; /* SRIOV only */
77 };
78 
79 enum {
80 	UC_ADDR_CHANGE = BIT(0),
81 	MC_ADDR_CHANGE = BIT(1),
82 };
83 
84 /* Vport context events */
85 #define SRIOV_VPORT_EVENTS (UC_ADDR_CHANGE | \
86 			    MC_ADDR_CHANGE)
87 
arm_vport_context_events_cmd(struct mlx5_core_dev * dev,u16 vport,u32 events_mask)88 static int arm_vport_context_events_cmd(struct mlx5_core_dev *dev, u16 vport,
89 					u32 events_mask)
90 {
91 	int in[MLX5_ST_SZ_DW(modify_nic_vport_context_in)] = {0};
92 	int out[MLX5_ST_SZ_DW(modify_nic_vport_context_out)] = {0};
93 	void *nic_vport_ctx;
94 
95 	MLX5_SET(modify_nic_vport_context_in, in,
96 		 opcode, MLX5_CMD_OP_MODIFY_NIC_VPORT_CONTEXT);
97 	MLX5_SET(modify_nic_vport_context_in, in, field_select.change_event, 1);
98 	MLX5_SET(modify_nic_vport_context_in, in, vport_number, vport);
99 	if (vport)
100 		MLX5_SET(modify_nic_vport_context_in, in, other_vport, 1);
101 	nic_vport_ctx = MLX5_ADDR_OF(modify_nic_vport_context_in,
102 				     in, nic_vport_context);
103 
104 	MLX5_SET(nic_vport_context, nic_vport_ctx, arm_change_event, 1);
105 
106 	if (events_mask & UC_ADDR_CHANGE)
107 		MLX5_SET(nic_vport_context, nic_vport_ctx,
108 			 event_on_uc_address_change, 1);
109 	if (events_mask & MC_ADDR_CHANGE)
110 		MLX5_SET(nic_vport_context, nic_vport_ctx,
111 			 event_on_mc_address_change, 1);
112 
113 	return mlx5_cmd_exec(dev, in, sizeof(in), out, sizeof(out));
114 }
115 
116 /* E-Switch vport context HW commands */
query_esw_vport_context_cmd(struct mlx5_core_dev * mdev,u32 vport,u32 * out,int outlen)117 static int query_esw_vport_context_cmd(struct mlx5_core_dev *mdev, u32 vport,
118 				       u32 *out, int outlen)
119 {
120 	u32 in[MLX5_ST_SZ_DW(query_esw_vport_context_in)] = {0};
121 
122 	MLX5_SET(query_nic_vport_context_in, in, opcode,
123 		 MLX5_CMD_OP_QUERY_ESW_VPORT_CONTEXT);
124 
125 	MLX5_SET(query_esw_vport_context_in, in, vport_number, vport);
126 	if (vport)
127 		MLX5_SET(query_esw_vport_context_in, in, other_vport, 1);
128 
129 	return mlx5_cmd_exec(mdev, in, sizeof(in), out, outlen);
130 }
131 
query_esw_vport_cvlan(struct mlx5_core_dev * dev,u32 vport,u16 * vlan,u8 * qos)132 static int query_esw_vport_cvlan(struct mlx5_core_dev *dev, u32 vport,
133 				 u16 *vlan, u8 *qos)
134 {
135 	u32 out[MLX5_ST_SZ_DW(query_esw_vport_context_out)] = {0};
136 	int err;
137 	bool cvlan_strip;
138 	bool cvlan_insert;
139 
140 	*vlan = 0;
141 	*qos = 0;
142 
143 	if (!MLX5_CAP_ESW(dev, vport_cvlan_strip) ||
144 	    !MLX5_CAP_ESW(dev, vport_cvlan_insert_if_not_exist))
145 		return -ENOTSUPP;
146 
147 	err = query_esw_vport_context_cmd(dev, vport, out, sizeof(out));
148 	if (err)
149 		goto out;
150 
151 	cvlan_strip = MLX5_GET(query_esw_vport_context_out, out,
152 			       esw_vport_context.vport_cvlan_strip);
153 
154 	cvlan_insert = MLX5_GET(query_esw_vport_context_out, out,
155 				esw_vport_context.vport_cvlan_insert);
156 
157 	if (cvlan_strip || cvlan_insert) {
158 		*vlan = MLX5_GET(query_esw_vport_context_out, out,
159 				 esw_vport_context.cvlan_id);
160 		*qos = MLX5_GET(query_esw_vport_context_out, out,
161 				esw_vport_context.cvlan_pcp);
162 	}
163 
164 	esw_debug(dev, "Query Vport[%d] cvlan: VLAN %d qos=%d\n",
165 		  vport, *vlan, *qos);
166 out:
167 	return err;
168 }
169 
modify_esw_vport_context_cmd(struct mlx5_core_dev * dev,u16 vport,void * in,int inlen)170 static int modify_esw_vport_context_cmd(struct mlx5_core_dev *dev, u16 vport,
171 					void *in, int inlen)
172 {
173 	u32 out[MLX5_ST_SZ_DW(modify_esw_vport_context_out)] = {0};
174 
175 	MLX5_SET(modify_esw_vport_context_in, in, vport_number, vport);
176 	if (vport)
177 		MLX5_SET(modify_esw_vport_context_in, in, other_vport, 1);
178 
179 	MLX5_SET(modify_esw_vport_context_in, in, opcode,
180 		 MLX5_CMD_OP_MODIFY_ESW_VPORT_CONTEXT);
181 
182 	return mlx5_cmd_exec(dev, in, inlen, out, sizeof(out));
183 }
184 
modify_esw_vport_cvlan(struct mlx5_core_dev * dev,u32 vport,u16 vlan,u8 qos,bool set)185 static int modify_esw_vport_cvlan(struct mlx5_core_dev *dev, u32 vport,
186 				  u16 vlan, u8 qos, bool set)
187 {
188 	u32 in[MLX5_ST_SZ_DW(modify_esw_vport_context_in)] = {0};
189 
190 	if (!MLX5_CAP_ESW(dev, vport_cvlan_strip) ||
191 	    !MLX5_CAP_ESW(dev, vport_cvlan_insert_if_not_exist))
192 		return -ENOTSUPP;
193 
194 	esw_debug(dev, "Set Vport[%d] VLAN %d qos %d set=%d\n",
195 		  vport, vlan, qos, set);
196 
197 	if (set) {
198 		MLX5_SET(modify_esw_vport_context_in, in,
199 			 esw_vport_context.vport_cvlan_strip, 1);
200 		/* insert only if no vlan in packet */
201 		MLX5_SET(modify_esw_vport_context_in, in,
202 			 esw_vport_context.vport_cvlan_insert, 1);
203 		MLX5_SET(modify_esw_vport_context_in, in,
204 			 esw_vport_context.cvlan_pcp, qos);
205 		MLX5_SET(modify_esw_vport_context_in, in,
206 			 esw_vport_context.cvlan_id, vlan);
207 	}
208 
209 	MLX5_SET(modify_esw_vport_context_in, in,
210 		 field_select.vport_cvlan_strip, 1);
211 	MLX5_SET(modify_esw_vport_context_in, in,
212 		 field_select.vport_cvlan_insert, 1);
213 
214 	return modify_esw_vport_context_cmd(dev, vport, in, sizeof(in));
215 }
216 
217 /* E-Switch FDB */
218 static struct mlx5_flow_rule *
esw_fdb_set_vport_rule(struct mlx5_eswitch * esw,u8 mac[ETH_ALEN],u32 vport)219 esw_fdb_set_vport_rule(struct mlx5_eswitch *esw, u8 mac[ETH_ALEN], u32 vport)
220 {
221 	int match_header = MLX5_MATCH_OUTER_HEADERS;
222 	struct mlx5_flow_destination dest;
223 	struct mlx5_flow_rule *flow_rule = NULL;
224 	struct mlx5_flow_act flow_act = {};
225 	u32 *match_v;
226 	u32 *match_c;
227 	u8 *dmac_v;
228 	u8 *dmac_c;
229 
230 	match_v = kzalloc(MLX5_ST_SZ_BYTES(fte_match_param), GFP_KERNEL);
231 	match_c = kzalloc(MLX5_ST_SZ_BYTES(fte_match_param), GFP_KERNEL);
232 	if (!match_v || !match_c) {
233 		printf("mlx5_core: WARN: ""FDB: Failed to alloc match parameters\n");
234 		goto out;
235 	}
236 	dmac_v = MLX5_ADDR_OF(fte_match_param, match_v,
237 			      outer_headers.dmac_47_16);
238 	dmac_c = MLX5_ADDR_OF(fte_match_param, match_c,
239 			      outer_headers.dmac_47_16);
240 
241 	ether_addr_copy(dmac_v, mac);
242 	/* Match criteria mask */
243 	memset(dmac_c, 0xff, 6);
244 
245 	dest.type = MLX5_FLOW_CONTEXT_DEST_TYPE_VPORT;
246 	dest.vport_num = vport;
247 
248 	esw_debug(esw->dev,
249 		  "\tFDB add rule dmac_v(%pM) dmac_c(%pM) -> vport(%d)\n",
250 		  dmac_v, dmac_c, vport);
251 	flow_rule =
252 		mlx5_add_flow_rule(esw->fdb_table.fdb,
253 				   match_header,
254 				   match_c,
255 				   match_v,
256 				   MLX5_FLOW_RULE_FWD_ACTION_DEST,
257 				   &flow_act, &dest);
258 	if (IS_ERR_OR_NULL(flow_rule)) {
259 		printf("mlx5_core: WARN: ""FDB: Failed to add flow rule: dmac_v(%pM) dmac_c(%pM) -> vport(%d), err(%ld)\n", dmac_v, dmac_c, vport, PTR_ERR(flow_rule));
260 		flow_rule = NULL;
261 	}
262 out:
263 	kfree(match_v);
264 	kfree(match_c);
265 	return flow_rule;
266 }
267 
esw_create_fdb_table(struct mlx5_eswitch * esw)268 static int esw_create_fdb_table(struct mlx5_eswitch *esw)
269 {
270 	int inlen = MLX5_ST_SZ_BYTES(create_flow_group_in);
271 	struct mlx5_core_dev *dev = esw->dev;
272 	struct mlx5_flow_namespace *root_ns;
273 	struct mlx5_flow_table *fdb;
274 	struct mlx5_flow_group *g;
275 	void *match_criteria;
276 	int table_size;
277 	u32 *flow_group_in;
278 	u8 *dmac;
279 	int err = 0;
280 
281 	esw_debug(dev, "Create FDB log_max_size(%d)\n",
282 		  MLX5_CAP_ESW_FLOWTABLE_FDB(dev, log_max_ft_size));
283 
284 	root_ns = mlx5_get_flow_namespace(dev, MLX5_FLOW_NAMESPACE_FDB);
285 	if (!root_ns) {
286 		esw_warn(dev, "Failed to get FDB flow namespace\n");
287 		return -ENOMEM;
288 	}
289 
290 	flow_group_in = mlx5_vzalloc(inlen);
291 	if (!flow_group_in)
292 		return -ENOMEM;
293 	memset(flow_group_in, 0, inlen);
294 
295 	/* (-2) Since MaorG said so .. */
296 	table_size = BIT(MLX5_CAP_ESW_FLOWTABLE_FDB(dev, log_max_ft_size)) - 2;
297 
298 	fdb = mlx5_create_flow_table(root_ns, 0, "FDB", table_size);
299 	if (IS_ERR_OR_NULL(fdb)) {
300 		err = PTR_ERR(fdb);
301 		esw_warn(dev, "Failed to create FDB Table err %d\n", err);
302 		goto out;
303 	}
304 
305 	MLX5_SET(create_flow_group_in, flow_group_in, match_criteria_enable,
306 		 MLX5_MATCH_OUTER_HEADERS);
307 	match_criteria = MLX5_ADDR_OF(create_flow_group_in, flow_group_in, match_criteria);
308 	dmac = MLX5_ADDR_OF(fte_match_param, match_criteria, outer_headers.dmac_47_16);
309 	MLX5_SET(create_flow_group_in, flow_group_in, start_flow_index, 0);
310 	MLX5_SET(create_flow_group_in, flow_group_in, end_flow_index, table_size - 1);
311 	eth_broadcast_addr(dmac);
312 
313 	g = mlx5_create_flow_group(fdb, flow_group_in);
314 	if (IS_ERR_OR_NULL(g)) {
315 		err = PTR_ERR(g);
316 		esw_warn(dev, "Failed to create flow group err(%d)\n", err);
317 		goto out;
318 	}
319 
320 	esw->fdb_table.addr_grp = g;
321 	esw->fdb_table.fdb = fdb;
322 out:
323 	kfree(flow_group_in);
324 	if (err && !IS_ERR_OR_NULL(fdb))
325 		mlx5_destroy_flow_table(fdb);
326 	return err;
327 }
328 
esw_destroy_fdb_table(struct mlx5_eswitch * esw)329 static void esw_destroy_fdb_table(struct mlx5_eswitch *esw)
330 {
331 	if (!esw->fdb_table.fdb)
332 		return;
333 
334 	esw_debug(esw->dev, "Destroy FDB Table\n");
335 	mlx5_destroy_flow_group(esw->fdb_table.addr_grp);
336 	mlx5_destroy_flow_table(esw->fdb_table.fdb);
337 	esw->fdb_table.fdb = NULL;
338 	esw->fdb_table.addr_grp = NULL;
339 }
340 
341 /* E-Switch vport UC/MC lists management */
342 typedef int (*vport_addr_action)(struct mlx5_eswitch *esw,
343 				 struct vport_addr *vaddr);
344 
esw_add_uc_addr(struct mlx5_eswitch * esw,struct vport_addr * vaddr)345 static int esw_add_uc_addr(struct mlx5_eswitch *esw, struct vport_addr *vaddr)
346 {
347 	struct hlist_head *hash = esw->l2_table.l2_hash;
348 	struct esw_uc_addr *esw_uc;
349 	u8 *mac = vaddr->node.addr;
350 	u32 vport = vaddr->vport;
351 	int err;
352 
353 	esw_uc = l2addr_hash_find(hash, mac, struct esw_uc_addr);
354 	if (esw_uc) {
355 		esw_warn(esw->dev,
356 			 "Failed to set L2 mac(%pM) for vport(%d), mac is already in use by vport(%d)\n",
357 			 mac, vport, esw_uc->vport);
358 		return -EEXIST;
359 	}
360 
361 	esw_uc = l2addr_hash_add(hash, mac, struct esw_uc_addr, GFP_KERNEL);
362 	if (!esw_uc)
363 		return -ENOMEM;
364 	esw_uc->vport = vport;
365 
366 	err = mlx5_mpfs_add_mac(esw->dev, &esw_uc->table_index, mac, 0, 0);
367 	if (err)
368 		goto abort;
369 
370 	if (esw->fdb_table.fdb) /* SRIOV is enabled: Forward UC MAC to vport */
371 		vaddr->flow_rule = esw_fdb_set_vport_rule(esw, mac, vport);
372 
373 	esw_debug(esw->dev, "\tADDED UC MAC: vport[%d] %pM index:%d fr(%p)\n",
374 		  vport, mac, esw_uc->table_index, vaddr->flow_rule);
375 	return err;
376 abort:
377 	l2addr_hash_del(esw_uc);
378 	return err;
379 }
380 
esw_del_uc_addr(struct mlx5_eswitch * esw,struct vport_addr * vaddr)381 static int esw_del_uc_addr(struct mlx5_eswitch *esw, struct vport_addr *vaddr)
382 {
383 	struct hlist_head *hash = esw->l2_table.l2_hash;
384 	struct esw_uc_addr *esw_uc;
385 	u8 *mac = vaddr->node.addr;
386 	u32 vport = vaddr->vport;
387 
388 	esw_uc = l2addr_hash_find(hash, mac, struct esw_uc_addr);
389 	if (!esw_uc || esw_uc->vport != vport) {
390 		esw_debug(esw->dev,
391 			  "MAC(%pM) doesn't belong to vport (%d)\n",
392 			  mac, vport);
393 		return -EINVAL;
394 	}
395 	esw_debug(esw->dev, "\tDELETE UC MAC: vport[%d] %pM index:%d fr(%p)\n",
396 		  vport, mac, esw_uc->table_index, vaddr->flow_rule);
397 
398 	mlx5_mpfs_del_mac(esw->dev, esw_uc->table_index);
399 
400 	mlx5_del_flow_rule(&vaddr->flow_rule);
401 
402 	l2addr_hash_del(esw_uc);
403 	return 0;
404 }
405 
esw_add_mc_addr(struct mlx5_eswitch * esw,struct vport_addr * vaddr)406 static int esw_add_mc_addr(struct mlx5_eswitch *esw, struct vport_addr *vaddr)
407 {
408 	struct hlist_head *hash = esw->mc_table;
409 	struct esw_mc_addr *esw_mc;
410 	u8 *mac = vaddr->node.addr;
411 	u32 vport = vaddr->vport;
412 
413 	if (!esw->fdb_table.fdb)
414 		return 0;
415 
416 	esw_mc = l2addr_hash_find(hash, mac, struct esw_mc_addr);
417 	if (esw_mc)
418 		goto add;
419 
420 	esw_mc = l2addr_hash_add(hash, mac, struct esw_mc_addr, GFP_KERNEL);
421 	if (!esw_mc)
422 		return -ENOMEM;
423 
424 	esw_mc->uplink_rule = /* Forward MC MAC to Uplink */
425 		esw_fdb_set_vport_rule(esw, mac, UPLINK_VPORT);
426 add:
427 	esw_mc->refcnt++;
428 	/* Forward MC MAC to vport */
429 	vaddr->flow_rule = esw_fdb_set_vport_rule(esw, mac, vport);
430 	esw_debug(esw->dev,
431 		  "\tADDED MC MAC: vport[%d] %pM fr(%p) refcnt(%d) uplinkfr(%p)\n",
432 		  vport, mac, vaddr->flow_rule,
433 		  esw_mc->refcnt, esw_mc->uplink_rule);
434 	return 0;
435 }
436 
esw_del_mc_addr(struct mlx5_eswitch * esw,struct vport_addr * vaddr)437 static int esw_del_mc_addr(struct mlx5_eswitch *esw, struct vport_addr *vaddr)
438 {
439 	struct hlist_head *hash = esw->mc_table;
440 	struct esw_mc_addr *esw_mc;
441 	u8 *mac = vaddr->node.addr;
442 	u32 vport = vaddr->vport;
443 
444 	if (!esw->fdb_table.fdb)
445 		return 0;
446 
447 	esw_mc = l2addr_hash_find(hash, mac, struct esw_mc_addr);
448 	if (!esw_mc) {
449 		esw_warn(esw->dev,
450 			 "Failed to find eswitch MC addr for MAC(%pM) vport(%d)",
451 			 mac, vport);
452 		return -EINVAL;
453 	}
454 	esw_debug(esw->dev,
455 		  "\tDELETE MC MAC: vport[%d] %pM fr(%p) refcnt(%d) uplinkfr(%p)\n",
456 		  vport, mac, vaddr->flow_rule, esw_mc->refcnt,
457 		  esw_mc->uplink_rule);
458 
459 	mlx5_del_flow_rule(&vaddr->flow_rule);
460 
461 	if (--esw_mc->refcnt)
462 		return 0;
463 
464 	mlx5_del_flow_rule(&esw_mc->uplink_rule);
465 
466 	l2addr_hash_del(esw_mc);
467 	return 0;
468 }
469 
470 /* Apply vport UC/MC list to HW l2 table and FDB table */
esw_apply_vport_addr_list(struct mlx5_eswitch * esw,u32 vport_num,int list_type)471 static void esw_apply_vport_addr_list(struct mlx5_eswitch *esw,
472 				      u32 vport_num, int list_type)
473 {
474 	struct mlx5_vport *vport = &esw->vports[vport_num];
475 	bool is_uc = list_type == MLX5_NIC_VPORT_LIST_TYPE_UC;
476 	vport_addr_action vport_addr_add;
477 	vport_addr_action vport_addr_del;
478 	struct vport_addr *addr;
479 	struct l2addr_node *node;
480 	struct hlist_head *hash;
481 	struct hlist_node *tmp;
482 	int hi;
483 
484 	vport_addr_add = is_uc ? esw_add_uc_addr :
485 				 esw_add_mc_addr;
486 	vport_addr_del = is_uc ? esw_del_uc_addr :
487 				 esw_del_mc_addr;
488 
489 	hash = is_uc ? vport->uc_list : vport->mc_list;
490 	for_each_l2hash_node(node, tmp, hash, hi) {
491 		addr = container_of(node, struct vport_addr, node);
492 		switch (addr->action) {
493 		case MLX5_ACTION_ADD:
494 			vport_addr_add(esw, addr);
495 			addr->action = MLX5_ACTION_NONE;
496 			break;
497 		case MLX5_ACTION_DEL:
498 			vport_addr_del(esw, addr);
499 			l2addr_hash_del(addr);
500 			break;
501 		}
502 	}
503 }
504 
505 /* Sync vport UC/MC list from vport context */
esw_update_vport_addr_list(struct mlx5_eswitch * esw,u32 vport_num,int list_type)506 static void esw_update_vport_addr_list(struct mlx5_eswitch *esw,
507 				       u32 vport_num, int list_type)
508 {
509 	struct mlx5_vport *vport = &esw->vports[vport_num];
510 	bool is_uc = list_type == MLX5_NIC_VPORT_LIST_TYPE_UC;
511 	u8 (*mac_list)[ETH_ALEN];
512 	struct l2addr_node *node;
513 	struct vport_addr *addr;
514 	struct hlist_head *hash;
515 	struct hlist_node *tmp;
516 	int size;
517 	int err;
518 	int hi;
519 	int i;
520 
521 	size = is_uc ? MLX5_MAX_UC_PER_VPORT(esw->dev) :
522 		       MLX5_MAX_MC_PER_VPORT(esw->dev);
523 
524 	mac_list = kcalloc(size, ETH_ALEN, GFP_KERNEL);
525 	if (!mac_list)
526 		return;
527 
528 	hash = is_uc ? vport->uc_list : vport->mc_list;
529 
530 	for_each_l2hash_node(node, tmp, hash, hi) {
531 		addr = container_of(node, struct vport_addr, node);
532 		addr->action = MLX5_ACTION_DEL;
533 	}
534 
535 	err = mlx5_query_nic_vport_mac_list(esw->dev, vport_num, list_type,
536 					    mac_list, &size);
537 	if (err)
538 		return;
539 	esw_debug(esw->dev, "vport[%d] context update %s list size (%d)\n",
540 		  vport_num, is_uc ? "UC" : "MC", size);
541 
542 	for (i = 0; i < size; i++) {
543 		if (is_uc && !is_valid_ether_addr(mac_list[i]))
544 			continue;
545 
546 		if (!is_uc && !is_multicast_ether_addr(mac_list[i]))
547 			continue;
548 
549 		addr = l2addr_hash_find(hash, mac_list[i], struct vport_addr);
550 		if (addr) {
551 			addr->action = MLX5_ACTION_NONE;
552 			continue;
553 		}
554 
555 		addr = l2addr_hash_add(hash, mac_list[i], struct vport_addr,
556 				       GFP_KERNEL);
557 		if (!addr) {
558 			esw_warn(esw->dev,
559 				 "Failed to add MAC(%pM) to vport[%d] DB\n",
560 				 mac_list[i], vport_num);
561 			continue;
562 		}
563 		addr->vport = vport_num;
564 		addr->action = MLX5_ACTION_ADD;
565 	}
566 	kfree(mac_list);
567 }
568 
esw_vport_change_handler(struct work_struct * work)569 static void esw_vport_change_handler(struct work_struct *work)
570 {
571 	struct mlx5_vport *vport =
572 		container_of(work, struct mlx5_vport, vport_change_handler);
573 	struct mlx5_core_dev *dev = vport->dev;
574 	struct mlx5_eswitch *esw = dev->priv.eswitch;
575 	u8 mac[ETH_ALEN];
576 
577 	mlx5_query_nic_vport_mac_address(dev, vport->vport, mac);
578 	esw_debug(dev, "vport[%d] Context Changed: perm mac: %pM\n",
579 		  vport->vport, mac);
580 
581 	if (vport->enabled_events & UC_ADDR_CHANGE) {
582 		esw_update_vport_addr_list(esw, vport->vport,
583 					   MLX5_NIC_VPORT_LIST_TYPE_UC);
584 		esw_apply_vport_addr_list(esw, vport->vport,
585 					  MLX5_NIC_VPORT_LIST_TYPE_UC);
586 	}
587 
588 	if (vport->enabled_events & MC_ADDR_CHANGE) {
589 		esw_update_vport_addr_list(esw, vport->vport,
590 					   MLX5_NIC_VPORT_LIST_TYPE_MC);
591 		esw_apply_vport_addr_list(esw, vport->vport,
592 					  MLX5_NIC_VPORT_LIST_TYPE_MC);
593 	}
594 
595 	esw_debug(esw->dev, "vport[%d] Context Changed: Done\n", vport->vport);
596 	if (vport->enabled)
597 		arm_vport_context_events_cmd(dev, vport->vport,
598 					     vport->enabled_events);
599 }
600 
esw_vport_enable_egress_acl(struct mlx5_eswitch * esw,struct mlx5_vport * vport)601 static void esw_vport_enable_egress_acl(struct mlx5_eswitch *esw,
602 					struct mlx5_vport *vport)
603 {
604 	int inlen = MLX5_ST_SZ_BYTES(create_flow_group_in);
605 	struct mlx5_flow_group *vlan_grp = NULL;
606 	struct mlx5_flow_group *drop_grp = NULL;
607 	struct mlx5_core_dev *dev = esw->dev;
608 	struct mlx5_flow_namespace *root_ns;
609 	struct mlx5_flow_table *acl;
610 	void *match_criteria;
611 	char table_name[32];
612 	u32 *flow_group_in;
613 	int table_size = 2;
614 	int err = 0;
615 
616 	if (!MLX5_CAP_ESW_EGRESS_ACL(dev, ft_support))
617 		return;
618 
619 	esw_debug(dev, "Create vport[%d] egress ACL log_max_size(%d)\n",
620 		  vport->vport, MLX5_CAP_ESW_EGRESS_ACL(dev, log_max_ft_size));
621 
622 	root_ns = mlx5_get_flow_namespace(dev, MLX5_FLOW_NAMESPACE_ESW_EGRESS);
623 	if (!root_ns) {
624 		esw_warn(dev, "Failed to get E-Switch egress flow namespace\n");
625 		return;
626 	}
627 
628 	flow_group_in = mlx5_vzalloc(inlen);
629 	if (!flow_group_in)
630 		return;
631 
632 	snprintf(table_name, 32, "egress_%d", vport->vport);
633 	acl = mlx5_create_vport_flow_table(root_ns, vport->vport, 0, table_name, table_size);
634 	if (IS_ERR_OR_NULL(acl)) {
635 		err = PTR_ERR(acl);
636 		esw_warn(dev, "Failed to create E-Switch vport[%d] egress flow Table, err(%d)\n",
637 			 vport->vport, err);
638 		goto out;
639 	}
640 
641 	MLX5_SET(create_flow_group_in, flow_group_in, match_criteria_enable, MLX5_MATCH_OUTER_HEADERS);
642 	match_criteria = MLX5_ADDR_OF(create_flow_group_in, flow_group_in, match_criteria);
643 	MLX5_SET_TO_ONES(fte_match_param, match_criteria, outer_headers.cvlan_tag);
644 	MLX5_SET_TO_ONES(fte_match_param, match_criteria, outer_headers.first_vid);
645 	MLX5_SET(create_flow_group_in, flow_group_in, start_flow_index, 0);
646 	MLX5_SET(create_flow_group_in, flow_group_in, end_flow_index, 0);
647 
648 	vlan_grp = mlx5_create_flow_group(acl, flow_group_in);
649 	if (IS_ERR_OR_NULL(vlan_grp)) {
650 		err = PTR_ERR(vlan_grp);
651 		esw_warn(dev, "Failed to create E-Switch vport[%d] egress allowed vlans flow group, err(%d)\n",
652 			 vport->vport, err);
653 		goto out;
654 	}
655 
656 	memset(flow_group_in, 0, inlen);
657 	MLX5_SET(create_flow_group_in, flow_group_in, start_flow_index, 1);
658 	MLX5_SET(create_flow_group_in, flow_group_in, end_flow_index, 1);
659 	drop_grp = mlx5_create_flow_group(acl, flow_group_in);
660 	if (IS_ERR_OR_NULL(drop_grp)) {
661 		err = PTR_ERR(drop_grp);
662 		esw_warn(dev, "Failed to create E-Switch vport[%d] egress drop flow group, err(%d)\n",
663 			 vport->vport, err);
664 		goto out;
665 	}
666 
667 	vport->egress.acl = acl;
668 	vport->egress.drop_grp = drop_grp;
669 	vport->egress.allowed_vlans_grp = vlan_grp;
670 out:
671 	kfree(flow_group_in);
672 	if (err && !IS_ERR_OR_NULL(vlan_grp))
673 		mlx5_destroy_flow_group(vlan_grp);
674 	if (err && !IS_ERR_OR_NULL(acl))
675 		mlx5_destroy_flow_table(acl);
676 }
677 
esw_vport_cleanup_egress_rules(struct mlx5_eswitch * esw,struct mlx5_vport * vport)678 static void esw_vport_cleanup_egress_rules(struct mlx5_eswitch *esw,
679 					   struct mlx5_vport *vport)
680 {
681 	mlx5_del_flow_rule(&vport->egress.allowed_vlan);
682 	mlx5_del_flow_rule(&vport->egress.drop_rule);
683 }
684 
esw_vport_disable_egress_acl(struct mlx5_eswitch * esw,struct mlx5_vport * vport)685 static void esw_vport_disable_egress_acl(struct mlx5_eswitch *esw,
686 					 struct mlx5_vport *vport)
687 {
688 	if (IS_ERR_OR_NULL(vport->egress.acl))
689 		return;
690 
691 	esw_debug(esw->dev, "Destroy vport[%d] E-Switch egress ACL\n", vport->vport);
692 
693 	esw_vport_cleanup_egress_rules(esw, vport);
694 	mlx5_destroy_flow_group(vport->egress.allowed_vlans_grp);
695 	mlx5_destroy_flow_group(vport->egress.drop_grp);
696 	mlx5_destroy_flow_table(vport->egress.acl);
697 	vport->egress.allowed_vlans_grp = NULL;
698 	vport->egress.drop_grp = NULL;
699 	vport->egress.acl = NULL;
700 }
701 
esw_vport_enable_ingress_acl(struct mlx5_eswitch * esw,struct mlx5_vport * vport)702 static void esw_vport_enable_ingress_acl(struct mlx5_eswitch *esw,
703 					 struct mlx5_vport *vport)
704 {
705 	int inlen = MLX5_ST_SZ_BYTES(create_flow_group_in);
706 	struct mlx5_core_dev *dev = esw->dev;
707 	struct mlx5_flow_namespace *root_ns;
708 	struct mlx5_flow_table *acl;
709 	struct mlx5_flow_group *g;
710 	void *match_criteria;
711 	char table_name[32];
712 	u32 *flow_group_in;
713 	int table_size = 1;
714 	int err = 0;
715 
716 	if (!MLX5_CAP_ESW_INGRESS_ACL(dev, ft_support))
717 		return;
718 
719 	esw_debug(dev, "Create vport[%d] ingress ACL log_max_size(%d)\n",
720 		  vport->vport, MLX5_CAP_ESW_INGRESS_ACL(dev, log_max_ft_size));
721 
722 	root_ns = mlx5_get_flow_namespace(dev, MLX5_FLOW_NAMESPACE_ESW_INGRESS);
723 	if (!root_ns) {
724 		esw_warn(dev, "Failed to get E-Switch ingress flow namespace\n");
725 		return;
726 	}
727 
728 	flow_group_in = mlx5_vzalloc(inlen);
729 	if (!flow_group_in)
730 		return;
731 
732 	snprintf(table_name, 32, "ingress_%d", vport->vport);
733 	acl = mlx5_create_vport_flow_table(root_ns, vport->vport, 0, table_name, table_size);
734 	if (IS_ERR_OR_NULL(acl)) {
735 		err = PTR_ERR(acl);
736 		esw_warn(dev, "Failed to create E-Switch vport[%d] ingress flow Table, err(%d)\n",
737 			 vport->vport, err);
738 		goto out;
739 	}
740 
741 	MLX5_SET(create_flow_group_in, flow_group_in, match_criteria_enable, MLX5_MATCH_OUTER_HEADERS);
742 	match_criteria = MLX5_ADDR_OF(create_flow_group_in, flow_group_in, match_criteria);
743 	MLX5_SET_TO_ONES(fte_match_param, match_criteria, outer_headers.cvlan_tag);
744 	MLX5_SET(create_flow_group_in, flow_group_in, start_flow_index, 0);
745 	MLX5_SET(create_flow_group_in, flow_group_in, end_flow_index, 0);
746 
747 	g = mlx5_create_flow_group(acl, flow_group_in);
748 	if (IS_ERR_OR_NULL(g)) {
749 		err = PTR_ERR(g);
750 		esw_warn(dev, "Failed to create E-Switch vport[%d] ingress flow group, err(%d)\n",
751 			 vport->vport, err);
752 		goto out;
753 	}
754 
755 	vport->ingress.acl = acl;
756 	vport->ingress.drop_grp = g;
757 out:
758 	kfree(flow_group_in);
759 	if (err && !IS_ERR_OR_NULL(acl))
760 		mlx5_destroy_flow_table(acl);
761 }
762 
esw_vport_cleanup_ingress_rules(struct mlx5_eswitch * esw,struct mlx5_vport * vport)763 static void esw_vport_cleanup_ingress_rules(struct mlx5_eswitch *esw,
764 					    struct mlx5_vport *vport)
765 {
766 	mlx5_del_flow_rule(&vport->ingress.drop_rule);
767 }
768 
esw_vport_disable_ingress_acl(struct mlx5_eswitch * esw,struct mlx5_vport * vport)769 static void esw_vport_disable_ingress_acl(struct mlx5_eswitch *esw,
770 					  struct mlx5_vport *vport)
771 {
772 	if (IS_ERR_OR_NULL(vport->ingress.acl))
773 		return;
774 
775 	esw_debug(esw->dev, "Destroy vport[%d] E-Switch ingress ACL\n", vport->vport);
776 
777 	esw_vport_cleanup_ingress_rules(esw, vport);
778 	mlx5_destroy_flow_group(vport->ingress.drop_grp);
779 	mlx5_destroy_flow_table(vport->ingress.acl);
780 	vport->ingress.acl = NULL;
781 	vport->ingress.drop_grp = NULL;
782 }
783 
esw_vport_ingress_config(struct mlx5_eswitch * esw,struct mlx5_vport * vport)784 static int esw_vport_ingress_config(struct mlx5_eswitch *esw,
785 				    struct mlx5_vport *vport)
786 {
787 	struct mlx5_flow_act flow_act = {};
788 	struct mlx5_flow_destination dest;
789 	u32 *match_v;
790 	u32 *match_c;
791 	int err = 0;
792 
793 	if (IS_ERR_OR_NULL(vport->ingress.acl)) {
794 		esw_warn(esw->dev,
795 			 "vport[%d] configure ingress rules failed, ingress acl is not initialized!\n",
796 			 vport->vport);
797 		return -EPERM;
798 	}
799 
800 	esw_vport_cleanup_ingress_rules(esw, vport);
801 
802 	if (!vport->vlan && !vport->qos)
803 		return 0;
804 
805 	esw_debug(esw->dev,
806 		  "vport[%d] configure ingress rules, vlan(%d) qos(%d)\n",
807 		  vport->vport, vport->vlan, vport->qos);
808 
809 	match_v = kzalloc(MLX5_ST_SZ_BYTES(fte_match_param), GFP_KERNEL);
810 	match_c = kzalloc(MLX5_ST_SZ_BYTES(fte_match_param), GFP_KERNEL);
811 	if (!match_v || !match_c) {
812 		err = -ENOMEM;
813 		esw_warn(esw->dev, "vport[%d] configure ingress rules failed, err(%d)\n",
814 			 vport->vport, err);
815 		goto out;
816 	}
817 	MLX5_SET_TO_ONES(fte_match_param, match_c, outer_headers.cvlan_tag);
818 	MLX5_SET_TO_ONES(fte_match_param, match_v, outer_headers.cvlan_tag);
819 
820 	dest.type = MLX5_FLOW_CONTEXT_DEST_TYPE_VPORT;
821 	dest.vport_num = vport->vport;
822 
823 	vport->ingress.drop_rule =
824 		mlx5_add_flow_rule(vport->ingress.acl,
825 				   MLX5_MATCH_OUTER_HEADERS,
826 				   match_c,
827 				   match_v,
828 				   MLX5_FLOW_RULE_FWD_ACTION_DROP,
829 				   &flow_act, &dest);
830 	if (IS_ERR_OR_NULL(vport->ingress.drop_rule)) {
831 		err = PTR_ERR(vport->ingress.drop_rule);
832 		printf("mlx5_core: WARN: ""vport[%d] configure ingress rules, err(%d)\n", vport->vport, err);
833 		vport->ingress.drop_rule = NULL;
834 	}
835 out:
836 	kfree(match_v);
837 	kfree(match_c);
838 	return err;
839 }
840 
esw_vport_egress_config(struct mlx5_eswitch * esw,struct mlx5_vport * vport)841 static int esw_vport_egress_config(struct mlx5_eswitch *esw,
842 				   struct mlx5_vport *vport)
843 {
844 	struct mlx5_flow_act flow_act = {};
845 	struct mlx5_flow_destination dest;
846 	u32 *match_v;
847 	u32 *match_c;
848 	int err = 0;
849 
850 	if (IS_ERR_OR_NULL(vport->egress.acl)) {
851 		esw_warn(esw->dev, "vport[%d] configure rgress rules failed, egress acl is not initialized!\n",
852 			 vport->vport);
853 		return -EPERM;
854 	}
855 
856 	esw_vport_cleanup_egress_rules(esw, vport);
857 
858 	if (!vport->vlan && !vport->qos)
859 		return 0;
860 
861 	esw_debug(esw->dev,
862 		  "vport[%d] configure egress rules, vlan(%d) qos(%d)\n",
863 		  vport->vport, vport->vlan, vport->qos);
864 
865 	match_v = kzalloc(MLX5_ST_SZ_BYTES(fte_match_param), GFP_KERNEL);
866 	match_c = kzalloc(MLX5_ST_SZ_BYTES(fte_match_param), GFP_KERNEL);
867 	if (!match_v || !match_c) {
868 		err = -ENOMEM;
869 		esw_warn(esw->dev, "vport[%d] configure egress rules failed, err(%d)\n",
870 			 vport->vport, err);
871 		goto out;
872 	}
873 
874 	/* Allowed vlan rule */
875 	MLX5_SET_TO_ONES(fte_match_param, match_c, outer_headers.cvlan_tag);
876 	MLX5_SET_TO_ONES(fte_match_param, match_v, outer_headers.cvlan_tag);
877 	MLX5_SET_TO_ONES(fte_match_param, match_c, outer_headers.first_vid);
878 	MLX5_SET(fte_match_param, match_v, outer_headers.first_vid, vport->vlan);
879 
880 	dest.type = MLX5_FLOW_CONTEXT_DEST_TYPE_VPORT;
881 	dest.vport_num = vport->vport;
882 
883 	vport->egress.allowed_vlan =
884 		mlx5_add_flow_rule(vport->egress.acl,
885 				   MLX5_MATCH_OUTER_HEADERS,
886 				   match_c,
887 				   match_v,
888 				   MLX5_FLOW_RULE_FWD_ACTION_ALLOW,
889 				   &flow_act, &dest);
890 	if (IS_ERR_OR_NULL(vport->egress.allowed_vlan)) {
891 		err = PTR_ERR(vport->egress.allowed_vlan);
892 		printf("mlx5_core: WARN: ""vport[%d] configure egress allowed vlan rule failed, err(%d)\n", vport->vport, err);
893 		vport->egress.allowed_vlan = NULL;
894 		goto out;
895 	}
896 
897 	/* Drop others rule (star rule) */
898 	memset(match_c, 0, MLX5_ST_SZ_BYTES(fte_match_param));
899 	memset(match_v, 0, MLX5_ST_SZ_BYTES(fte_match_param));
900 	vport->egress.drop_rule =
901 		mlx5_add_flow_rule(vport->egress.acl,
902 				   0,
903 				   match_c,
904 				   match_v,
905 				   MLX5_FLOW_RULE_FWD_ACTION_DROP,
906 				   &flow_act, &dest);
907 	if (IS_ERR_OR_NULL(vport->egress.drop_rule)) {
908 		err = PTR_ERR(vport->egress.drop_rule);
909 		printf("mlx5_core: WARN: ""vport[%d] configure egress drop rule failed, err(%d)\n", vport->vport, err);
910 		vport->egress.drop_rule = NULL;
911 	}
912 out:
913 	kfree(match_v);
914 	kfree(match_c);
915 	return err;
916 }
917 
esw_enable_vport(struct mlx5_eswitch * esw,int vport_num,int enable_events)918 static void esw_enable_vport(struct mlx5_eswitch *esw, int vport_num,
919 			     int enable_events)
920 {
921 	struct mlx5_vport *vport = &esw->vports[vport_num];
922 	unsigned long flags;
923 
924 	mutex_lock(&vport->state_lock);
925 	WARN_ON(vport->enabled);
926 
927 	esw_debug(esw->dev, "Enabling VPORT(%d)\n", vport_num);
928 
929 	if (vport_num) { /* Only VFs need ACLs for VST and spoofchk filtering */
930 		esw_vport_enable_ingress_acl(esw, vport);
931 		esw_vport_enable_egress_acl(esw, vport);
932 		esw_vport_ingress_config(esw, vport);
933 		esw_vport_egress_config(esw, vport);
934 	}
935 
936 	mlx5_modify_vport_admin_state(esw->dev,
937 				      MLX5_QUERY_VPORT_STATE_IN_OP_MOD_ESW_VPORT,
938 				      vport_num,
939 				      MLX5_ESW_VPORT_ADMIN_STATE_AUTO);
940 
941 	/* Sync with current vport context */
942 	vport->enabled_events = enable_events;
943 	esw_vport_change_handler(&vport->vport_change_handler);
944 
945 	spin_lock_irqsave(&vport->lock, flags);
946 	vport->enabled = true;
947 	spin_unlock_irqrestore(&vport->lock, flags);
948 
949 	arm_vport_context_events_cmd(esw->dev, vport_num, enable_events);
950 
951 	esw->enabled_vports++;
952 	esw_debug(esw->dev, "Enabled VPORT(%d)\n", vport_num);
953 	mutex_unlock(&vport->state_lock);
954 }
955 
esw_cleanup_vport(struct mlx5_eswitch * esw,u16 vport_num)956 static void esw_cleanup_vport(struct mlx5_eswitch *esw, u16 vport_num)
957 {
958 	struct mlx5_vport *vport = &esw->vports[vport_num];
959 	struct l2addr_node *node;
960 	struct vport_addr *addr;
961 	struct hlist_node *tmp;
962 	int hi;
963 
964 	for_each_l2hash_node(node, tmp, vport->uc_list, hi) {
965 		addr = container_of(node, struct vport_addr, node);
966 		addr->action = MLX5_ACTION_DEL;
967 	}
968 	esw_apply_vport_addr_list(esw, vport_num, MLX5_NIC_VPORT_LIST_TYPE_UC);
969 
970 	for_each_l2hash_node(node, tmp, vport->mc_list, hi) {
971 		addr = container_of(node, struct vport_addr, node);
972 		addr->action = MLX5_ACTION_DEL;
973 	}
974 	esw_apply_vport_addr_list(esw, vport_num, MLX5_NIC_VPORT_LIST_TYPE_MC);
975 }
976 
esw_disable_vport(struct mlx5_eswitch * esw,int vport_num)977 static void esw_disable_vport(struct mlx5_eswitch *esw, int vport_num)
978 {
979 	struct mlx5_vport *vport = &esw->vports[vport_num];
980 	unsigned long flags;
981 
982 	mutex_lock(&vport->state_lock);
983 	if (!vport->enabled) {
984 		mutex_unlock(&vport->state_lock);
985 		return;
986 	}
987 
988 	esw_debug(esw->dev, "Disabling vport(%d)\n", vport_num);
989 	/* Mark this vport as disabled to discard new events */
990 	spin_lock_irqsave(&vport->lock, flags);
991 	vport->enabled = false;
992 	vport->enabled_events = 0;
993 	spin_unlock_irqrestore(&vport->lock, flags);
994 
995 	mlx5_modify_vport_admin_state(esw->dev,
996 				      MLX5_QUERY_VPORT_STATE_IN_OP_MOD_ESW_VPORT,
997 				      vport_num,
998 				      MLX5_ESW_VPORT_ADMIN_STATE_DOWN);
999 	/* Wait for current already scheduled events to complete */
1000 	flush_workqueue(esw->work_queue);
1001 	/* Disable events from this vport */
1002 	arm_vport_context_events_cmd(esw->dev, vport->vport, 0);
1003 	/* We don't assume VFs will cleanup after themselves */
1004 	esw_cleanup_vport(esw, vport_num);
1005 	if (vport_num) {
1006 		esw_vport_disable_egress_acl(esw, vport);
1007 		esw_vport_disable_ingress_acl(esw, vport);
1008 	}
1009 	esw->enabled_vports--;
1010 	mutex_unlock(&vport->state_lock);
1011 }
1012 
1013 /* Public E-Switch API */
mlx5_eswitch_enable_sriov(struct mlx5_eswitch * esw,int nvfs)1014 int mlx5_eswitch_enable_sriov(struct mlx5_eswitch *esw, int nvfs)
1015 {
1016 	int err;
1017 	int i;
1018 
1019 	if (!esw || !MLX5_CAP_GEN(esw->dev, vport_group_manager) ||
1020 	    MLX5_CAP_GEN(esw->dev, port_type) != MLX5_CAP_PORT_TYPE_ETH)
1021 		return 0;
1022 
1023 	if (!MLX5_CAP_GEN(esw->dev, eswitch_flow_table) ||
1024 	    !MLX5_CAP_ESW_FLOWTABLE_FDB(esw->dev, ft_support)) {
1025 		esw_warn(esw->dev, "E-Switch FDB is not supported, aborting ...\n");
1026 		return -ENOTSUPP;
1027 	}
1028 
1029 	if (!MLX5_CAP_ESW_INGRESS_ACL(esw->dev, ft_support))
1030 		esw_warn(esw->dev, "E-Switch ingress ACL is not supported by FW\n");
1031 
1032 	if (!MLX5_CAP_ESW_EGRESS_ACL(esw->dev, ft_support))
1033 		esw_warn(esw->dev, "E-Switch engress ACL is not supported by FW\n");
1034 
1035 	esw_info(esw->dev, "E-Switch enable SRIOV: nvfs(%d)\n", nvfs);
1036 
1037 	esw_disable_vport(esw, 0);
1038 
1039 	err = esw_create_fdb_table(esw);
1040 	if (err)
1041 		goto abort;
1042 
1043 	for (i = 0; i <= nvfs; i++)
1044 		esw_enable_vport(esw, i, SRIOV_VPORT_EVENTS);
1045 
1046 	esw_info(esw->dev, "SRIOV enabled: active vports(%d)\n",
1047 		 esw->enabled_vports);
1048 	return 0;
1049 
1050 abort:
1051 	esw_enable_vport(esw, 0, UC_ADDR_CHANGE);
1052 	return err;
1053 }
1054 
mlx5_eswitch_disable_sriov(struct mlx5_eswitch * esw)1055 void mlx5_eswitch_disable_sriov(struct mlx5_eswitch *esw)
1056 {
1057 	int i;
1058 
1059 	if (!esw || !MLX5_CAP_GEN(esw->dev, vport_group_manager) ||
1060 	    MLX5_CAP_GEN(esw->dev, port_type) != MLX5_CAP_PORT_TYPE_ETH)
1061 		return;
1062 
1063 	esw_info(esw->dev, "disable SRIOV: active vports(%d)\n",
1064 		 esw->enabled_vports);
1065 
1066 	for (i = 0; i < esw->total_vports; i++)
1067 		esw_disable_vport(esw, i);
1068 
1069 	esw_destroy_fdb_table(esw);
1070 
1071 	/* VPORT 0 (PF) must be enabled back with non-sriov configuration */
1072 	esw_enable_vport(esw, 0, UC_ADDR_CHANGE);
1073 }
1074 
mlx5_eswitch_init(struct mlx5_core_dev * dev,int total_vports)1075 int mlx5_eswitch_init(struct mlx5_core_dev *dev, int total_vports)
1076 {
1077 	int l2_table_size = 1 << MLX5_CAP_GEN(dev, log_max_l2_table);
1078 	struct mlx5_eswitch *esw;
1079 	int vport_num;
1080 	int err;
1081 
1082 	if (!MLX5_CAP_GEN(dev, vport_group_manager) ||
1083 	    MLX5_CAP_GEN(dev, port_type) != MLX5_CAP_PORT_TYPE_ETH)
1084 		return 0;
1085 
1086 	esw_info(dev,
1087 		 "Total vports %d, l2 table size(%d), per vport: max uc(%d) max mc(%d)\n",
1088 		 total_vports, l2_table_size,
1089 		 MLX5_MAX_UC_PER_VPORT(dev),
1090 		 MLX5_MAX_MC_PER_VPORT(dev));
1091 
1092 	esw = kzalloc(sizeof(*esw), GFP_KERNEL);
1093 	if (!esw)
1094 		return -ENOMEM;
1095 
1096 	esw->dev = dev;
1097 
1098 	esw->l2_table.bitmap = kcalloc(BITS_TO_LONGS(l2_table_size),
1099 				   sizeof(uintptr_t), GFP_KERNEL);
1100 	if (!esw->l2_table.bitmap) {
1101 		err = -ENOMEM;
1102 		goto abort;
1103 	}
1104 	esw->l2_table.size = l2_table_size;
1105 
1106 	esw->work_queue = create_singlethread_workqueue("mlx5_esw_wq");
1107 	if (!esw->work_queue) {
1108 		err = -ENOMEM;
1109 		goto abort;
1110 	}
1111 
1112 	esw->vports = kcalloc(total_vports, sizeof(struct mlx5_vport),
1113 			      GFP_KERNEL);
1114 	if (!esw->vports) {
1115 		err = -ENOMEM;
1116 		goto abort;
1117 	}
1118 
1119 	for (vport_num = 0; vport_num < total_vports; vport_num++) {
1120 		struct mlx5_vport *vport = &esw->vports[vport_num];
1121 
1122 		vport->vport = vport_num;
1123 		vport->dev = dev;
1124 		INIT_WORK(&vport->vport_change_handler,
1125 			  esw_vport_change_handler);
1126 		spin_lock_init(&vport->lock);
1127 		mutex_init(&vport->state_lock);
1128 	}
1129 
1130 	esw->total_vports = total_vports;
1131 	esw->enabled_vports = 0;
1132 
1133 	dev->priv.eswitch = esw;
1134 	esw_enable_vport(esw, 0, UC_ADDR_CHANGE);
1135 	/* VF Vports will be enabled when SRIOV is enabled */
1136 	return 0;
1137 abort:
1138 	if (esw->work_queue)
1139 		destroy_workqueue(esw->work_queue);
1140 	kfree(esw->l2_table.bitmap);
1141 	kfree(esw->vports);
1142 	kfree(esw);
1143 	return err;
1144 }
1145 
mlx5_eswitch_cleanup(struct mlx5_eswitch * esw)1146 void mlx5_eswitch_cleanup(struct mlx5_eswitch *esw)
1147 {
1148 	if (!esw || !MLX5_CAP_GEN(esw->dev, vport_group_manager) ||
1149 	    MLX5_CAP_GEN(esw->dev, port_type) != MLX5_CAP_PORT_TYPE_ETH)
1150 		return;
1151 
1152 	esw_info(esw->dev, "cleanup\n");
1153 	esw_disable_vport(esw, 0);
1154 
1155 	esw->dev->priv.eswitch = NULL;
1156 	destroy_workqueue(esw->work_queue);
1157 	kfree(esw->l2_table.bitmap);
1158 	kfree(esw->vports);
1159 	kfree(esw);
1160 }
1161 
mlx5_eswitch_vport_event(struct mlx5_eswitch * esw,struct mlx5_eqe * eqe)1162 void mlx5_eswitch_vport_event(struct mlx5_eswitch *esw, struct mlx5_eqe *eqe)
1163 {
1164 	struct mlx5_eqe_vport_change *vc_eqe = &eqe->data.vport_change;
1165 	u16 vport_num = be16_to_cpu(vc_eqe->vport_num);
1166 	struct mlx5_vport *vport;
1167 
1168 	if (!esw) {
1169 		printf("mlx5_core: WARN: ""MLX5 E-Switch: vport %d got an event while eswitch is not initialized\n", vport_num);
1170 		return;
1171 	}
1172 
1173 	vport = &esw->vports[vport_num];
1174 	spin_lock(&vport->lock);
1175 	if (vport->enabled)
1176 		queue_work(esw->work_queue, &vport->vport_change_handler);
1177 	spin_unlock(&vport->lock);
1178 }
1179 
1180 /* Vport Administration */
1181 #define ESW_ALLOWED(esw) \
1182 	(esw && MLX5_CAP_GEN(esw->dev, vport_group_manager) && mlx5_core_is_pf(esw->dev))
1183 #define LEGAL_VPORT(esw, vport) (vport >= 0 && vport < esw->total_vports)
1184 
node_guid_gen_from_mac(u64 * node_guid,u8 mac[ETH_ALEN])1185 static void node_guid_gen_from_mac(u64 *node_guid, u8 mac[ETH_ALEN])
1186 {
1187 	((u8 *)node_guid)[7] = mac[0];
1188 	((u8 *)node_guid)[6] = mac[1];
1189 	((u8 *)node_guid)[5] = mac[2];
1190 	((u8 *)node_guid)[4] = 0xff;
1191 	((u8 *)node_guid)[3] = 0xfe;
1192 	((u8 *)node_guid)[2] = mac[3];
1193 	((u8 *)node_guid)[1] = mac[4];
1194 	((u8 *)node_guid)[0] = mac[5];
1195 }
1196 
mlx5_eswitch_set_vport_mac(struct mlx5_eswitch * esw,int vport,u8 mac[ETH_ALEN])1197 int mlx5_eswitch_set_vport_mac(struct mlx5_eswitch *esw,
1198 			       int vport, u8 mac[ETH_ALEN])
1199 {
1200 	int err = 0;
1201 	u64 node_guid;
1202 
1203 	if (!ESW_ALLOWED(esw))
1204 		return -EPERM;
1205 	if (!LEGAL_VPORT(esw, vport))
1206 		return -EINVAL;
1207 
1208 	err = mlx5_modify_nic_vport_mac_address(esw->dev, vport, mac);
1209 	if (err) {
1210 		mlx5_core_warn(esw->dev,
1211 			       "Failed to mlx5_modify_nic_vport_mac vport(%d) err=(%d)\n",
1212 			       vport, err);
1213 		return err;
1214 	}
1215 
1216 	node_guid_gen_from_mac(&node_guid, mac);
1217 	err = mlx5_modify_nic_vport_node_guid(esw->dev, vport, node_guid);
1218 	if (err) {
1219 		mlx5_core_warn(esw->dev,
1220 			       "Failed to mlx5_modify_nic_vport_node_guid vport(%d) err=(%d)\n",
1221 			       vport, err);
1222 		return err;
1223 	}
1224 
1225 	return err;
1226 }
1227 
mlx5_eswitch_set_vport_state(struct mlx5_eswitch * esw,int vport,int link_state)1228 int mlx5_eswitch_set_vport_state(struct mlx5_eswitch *esw,
1229 				 int vport, int link_state)
1230 {
1231 	if (!ESW_ALLOWED(esw))
1232 		return -EPERM;
1233 	if (!LEGAL_VPORT(esw, vport))
1234 		return -EINVAL;
1235 
1236 	return mlx5_modify_vport_admin_state(esw->dev,
1237 					     MLX5_QUERY_VPORT_STATE_IN_OP_MOD_ESW_VPORT,
1238 					     vport, link_state);
1239 }
1240 
mlx5_eswitch_get_vport_config(struct mlx5_eswitch * esw,int vport,struct mlx5_esw_vport_info * ivi)1241 int mlx5_eswitch_get_vport_config(struct mlx5_eswitch *esw,
1242 				  int vport, struct mlx5_esw_vport_info *ivi)
1243 {
1244 	u16 vlan;
1245 	u8 qos;
1246 
1247 	if (!ESW_ALLOWED(esw))
1248 		return -EPERM;
1249 	if (!LEGAL_VPORT(esw, vport))
1250 		return -EINVAL;
1251 
1252 	memset(ivi, 0, sizeof(*ivi));
1253 	ivi->vf = vport - 1;
1254 
1255 	mlx5_query_nic_vport_mac_address(esw->dev, vport, ivi->mac);
1256 	ivi->linkstate = mlx5_query_vport_admin_state(esw->dev,
1257 						      MLX5_QUERY_VPORT_STATE_IN_OP_MOD_ESW_VPORT,
1258 						      vport);
1259 	query_esw_vport_cvlan(esw->dev, vport, &vlan, &qos);
1260 	ivi->vlan = vlan;
1261 	ivi->qos = qos;
1262 	ivi->spoofchk = 0;
1263 
1264 	return 0;
1265 }
1266 
mlx5_eswitch_set_vport_vlan(struct mlx5_eswitch * esw,int vport,u16 vlan,u8 qos)1267 int mlx5_eswitch_set_vport_vlan(struct mlx5_eswitch *esw,
1268 				int vport, u16 vlan, u8 qos)
1269 {
1270 	struct mlx5_vport *evport;
1271 	int err = 0;
1272 	int set = 0;
1273 
1274 	if (!ESW_ALLOWED(esw))
1275 		return -EPERM;
1276 	if (!LEGAL_VPORT(esw, vport) || (vlan > 4095) || (qos > 7))
1277 		return -EINVAL;
1278 
1279 	if (vlan || qos)
1280 		set = 1;
1281 
1282 	evport = &esw->vports[vport];
1283 
1284 	err = modify_esw_vport_cvlan(esw->dev, vport, vlan, qos, set);
1285 	if (err)
1286 		return err;
1287 
1288 	mutex_lock(&evport->state_lock);
1289 	evport->vlan = vlan;
1290 	evport->qos = qos;
1291 	if (evport->enabled) {
1292 		esw_vport_ingress_config(esw, evport);
1293 		esw_vport_egress_config(esw, evport);
1294 	}
1295 	mutex_unlock(&evport->state_lock);
1296 	return err;
1297 }
1298 
1299