xref: /linux/net/bridge/br_mdb.c (revision 83869019c74cc2d01c96a3be2463a4eebe362224)
1 // SPDX-License-Identifier: GPL-2.0
2 #include <linux/err.h>
3 #include <linux/igmp.h>
4 #include <linux/kernel.h>
5 #include <linux/netdevice.h>
6 #include <linux/rculist.h>
7 #include <linux/skbuff.h>
8 #include <linux/if_ether.h>
9 #include <net/ip.h>
10 #include <net/netlink.h>
11 #include <net/switchdev.h>
12 #if IS_ENABLED(CONFIG_IPV6)
13 #include <net/ipv6.h>
14 #include <net/addrconf.h>
15 #endif
16 
17 #include "br_private.h"
18 
19 static bool
20 br_ip4_rports_get_timer(struct net_bridge_mcast_port *pmctx,
21 			unsigned long *timer)
22 {
23 	*timer = br_timer_value(&pmctx->ip4_mc_router_timer);
24 	return !hlist_unhashed(&pmctx->ip4_rlist);
25 }
26 
27 static bool
28 br_ip6_rports_get_timer(struct net_bridge_mcast_port *pmctx,
29 			unsigned long *timer)
30 {
31 #if IS_ENABLED(CONFIG_IPV6)
32 	*timer = br_timer_value(&pmctx->ip6_mc_router_timer);
33 	return !hlist_unhashed(&pmctx->ip6_rlist);
34 #else
35 	*timer = 0;
36 	return false;
37 #endif
38 }
39 
40 static size_t __br_rports_one_size(void)
41 {
42 	return nla_total_size(sizeof(u32)) + /* MDBA_ROUTER_PORT */
43 	       nla_total_size(sizeof(u32)) + /* MDBA_ROUTER_PATTR_TIMER */
44 	       nla_total_size(sizeof(u8)) +  /* MDBA_ROUTER_PATTR_TYPE */
45 	       nla_total_size(sizeof(u32)) + /* MDBA_ROUTER_PATTR_INET_TIMER */
46 	       nla_total_size(sizeof(u32)) + /* MDBA_ROUTER_PATTR_INET6_TIMER */
47 	       nla_total_size(sizeof(u32));  /* MDBA_ROUTER_PATTR_VID */
48 }
49 
50 size_t br_rports_size(const struct net_bridge_mcast *brmctx)
51 {
52 	struct net_bridge_mcast_port *pmctx;
53 	size_t size = nla_total_size(0); /* MDBA_ROUTER */
54 
55 	rcu_read_lock();
56 	hlist_for_each_entry_rcu(pmctx, &brmctx->ip4_mc_router_list,
57 				 ip4_rlist)
58 		size += __br_rports_one_size();
59 
60 #if IS_ENABLED(CONFIG_IPV6)
61 	hlist_for_each_entry_rcu(pmctx, &brmctx->ip6_mc_router_list,
62 				 ip6_rlist)
63 		size += __br_rports_one_size();
64 #endif
65 	rcu_read_unlock();
66 
67 	return size;
68 }
69 
70 int br_rports_fill_info(struct sk_buff *skb,
71 			const struct net_bridge_mcast *brmctx)
72 {
73 	u16 vid = brmctx->vlan ? brmctx->vlan->vid : 0;
74 	bool have_ip4_mc_rtr, have_ip6_mc_rtr;
75 	unsigned long ip4_timer, ip6_timer;
76 	struct nlattr *nest, *port_nest;
77 	struct net_bridge_port *p;
78 
79 	if (!brmctx->multicast_router || !br_rports_have_mc_router(brmctx))
80 		return 0;
81 
82 	nest = nla_nest_start_noflag(skb, MDBA_ROUTER);
83 	if (nest == NULL)
84 		return -EMSGSIZE;
85 
86 	list_for_each_entry_rcu(p, &brmctx->br->port_list, list) {
87 		struct net_bridge_mcast_port *pmctx;
88 
89 		if (vid) {
90 			struct net_bridge_vlan *v;
91 
92 			v = br_vlan_find(nbp_vlan_group(p), vid);
93 			if (!v)
94 				continue;
95 			pmctx = &v->port_mcast_ctx;
96 		} else {
97 			pmctx = &p->multicast_ctx;
98 		}
99 
100 		have_ip4_mc_rtr = br_ip4_rports_get_timer(pmctx, &ip4_timer);
101 		have_ip6_mc_rtr = br_ip6_rports_get_timer(pmctx, &ip6_timer);
102 
103 		if (!have_ip4_mc_rtr && !have_ip6_mc_rtr)
104 			continue;
105 
106 		port_nest = nla_nest_start_noflag(skb, MDBA_ROUTER_PORT);
107 		if (!port_nest)
108 			goto fail;
109 
110 		if (nla_put_nohdr(skb, sizeof(u32), &p->dev->ifindex) ||
111 		    nla_put_u32(skb, MDBA_ROUTER_PATTR_TIMER,
112 				max(ip4_timer, ip6_timer)) ||
113 		    nla_put_u8(skb, MDBA_ROUTER_PATTR_TYPE,
114 			       p->multicast_ctx.multicast_router) ||
115 		    (have_ip4_mc_rtr &&
116 		     nla_put_u32(skb, MDBA_ROUTER_PATTR_INET_TIMER,
117 				 ip4_timer)) ||
118 		    (have_ip6_mc_rtr &&
119 		     nla_put_u32(skb, MDBA_ROUTER_PATTR_INET6_TIMER,
120 				 ip6_timer)) ||
121 		    (vid && nla_put_u16(skb, MDBA_ROUTER_PATTR_VID, vid))) {
122 			nla_nest_cancel(skb, port_nest);
123 			goto fail;
124 		}
125 		nla_nest_end(skb, port_nest);
126 	}
127 
128 	nla_nest_end(skb, nest);
129 	return 0;
130 fail:
131 	nla_nest_cancel(skb, nest);
132 	return -EMSGSIZE;
133 }
134 
135 static void __mdb_entry_fill_flags(struct br_mdb_entry *e, unsigned char flags)
136 {
137 	e->state = flags & MDB_PG_FLAGS_PERMANENT;
138 	e->flags = 0;
139 	if (flags & MDB_PG_FLAGS_OFFLOAD)
140 		e->flags |= MDB_FLAGS_OFFLOAD;
141 	if (flags & MDB_PG_FLAGS_FAST_LEAVE)
142 		e->flags |= MDB_FLAGS_FAST_LEAVE;
143 	if (flags & MDB_PG_FLAGS_STAR_EXCL)
144 		e->flags |= MDB_FLAGS_STAR_EXCL;
145 	if (flags & MDB_PG_FLAGS_BLOCKED)
146 		e->flags |= MDB_FLAGS_BLOCKED;
147 }
148 
149 static void __mdb_entry_to_br_ip(struct br_mdb_entry *entry, struct br_ip *ip,
150 				 struct nlattr **mdb_attrs)
151 {
152 	memset(ip, 0, sizeof(struct br_ip));
153 	ip->vid = entry->vid;
154 	ip->proto = entry->addr.proto;
155 	switch (ip->proto) {
156 	case htons(ETH_P_IP):
157 		ip->dst.ip4 = entry->addr.u.ip4;
158 		if (mdb_attrs && mdb_attrs[MDBE_ATTR_SOURCE])
159 			ip->src.ip4 = nla_get_in_addr(mdb_attrs[MDBE_ATTR_SOURCE]);
160 		break;
161 #if IS_ENABLED(CONFIG_IPV6)
162 	case htons(ETH_P_IPV6):
163 		ip->dst.ip6 = entry->addr.u.ip6;
164 		if (mdb_attrs && mdb_attrs[MDBE_ATTR_SOURCE])
165 			ip->src.ip6 = nla_get_in6_addr(mdb_attrs[MDBE_ATTR_SOURCE]);
166 		break;
167 #endif
168 	default:
169 		ether_addr_copy(ip->dst.mac_addr, entry->addr.u.mac_addr);
170 	}
171 
172 }
173 
174 static int __mdb_fill_srcs(struct sk_buff *skb,
175 			   struct net_bridge_port_group *p)
176 {
177 	struct net_bridge_group_src *ent;
178 	struct nlattr *nest, *nest_ent;
179 
180 	if (hlist_empty(&p->src_list))
181 		return 0;
182 
183 	nest = nla_nest_start(skb, MDBA_MDB_EATTR_SRC_LIST);
184 	if (!nest)
185 		return -EMSGSIZE;
186 
187 	hlist_for_each_entry_rcu(ent, &p->src_list, node,
188 				 lockdep_is_held(&p->key.port->br->multicast_lock)) {
189 		nest_ent = nla_nest_start(skb, MDBA_MDB_SRCLIST_ENTRY);
190 		if (!nest_ent)
191 			goto out_cancel_err;
192 		switch (ent->addr.proto) {
193 		case htons(ETH_P_IP):
194 			if (nla_put_in_addr(skb, MDBA_MDB_SRCATTR_ADDRESS,
195 					    ent->addr.src.ip4)) {
196 				nla_nest_cancel(skb, nest_ent);
197 				goto out_cancel_err;
198 			}
199 			break;
200 #if IS_ENABLED(CONFIG_IPV6)
201 		case htons(ETH_P_IPV6):
202 			if (nla_put_in6_addr(skb, MDBA_MDB_SRCATTR_ADDRESS,
203 					     &ent->addr.src.ip6)) {
204 				nla_nest_cancel(skb, nest_ent);
205 				goto out_cancel_err;
206 			}
207 			break;
208 #endif
209 		default:
210 			nla_nest_cancel(skb, nest_ent);
211 			continue;
212 		}
213 		if (nla_put_u32(skb, MDBA_MDB_SRCATTR_TIMER,
214 				br_timer_value(&ent->timer))) {
215 			nla_nest_cancel(skb, nest_ent);
216 			goto out_cancel_err;
217 		}
218 		nla_nest_end(skb, nest_ent);
219 	}
220 
221 	nla_nest_end(skb, nest);
222 
223 	return 0;
224 
225 out_cancel_err:
226 	nla_nest_cancel(skb, nest);
227 	return -EMSGSIZE;
228 }
229 
230 static int __mdb_fill_info(struct sk_buff *skb,
231 			   struct net_bridge_mdb_entry *mp,
232 			   struct net_bridge_port_group *p)
233 {
234 	bool dump_srcs_mode = false;
235 	struct timer_list *mtimer;
236 	struct nlattr *nest_ent;
237 	struct br_mdb_entry e;
238 	u8 flags = 0;
239 	int ifindex;
240 
241 	memset(&e, 0, sizeof(e));
242 	if (p) {
243 		ifindex = p->key.port->dev->ifindex;
244 		mtimer = &p->timer;
245 		flags = p->flags;
246 	} else {
247 		ifindex = mp->br->dev->ifindex;
248 		mtimer = &mp->timer;
249 	}
250 
251 	__mdb_entry_fill_flags(&e, flags);
252 	e.ifindex = ifindex;
253 	e.vid = mp->addr.vid;
254 	if (mp->addr.proto == htons(ETH_P_IP))
255 		e.addr.u.ip4 = mp->addr.dst.ip4;
256 #if IS_ENABLED(CONFIG_IPV6)
257 	else if (mp->addr.proto == htons(ETH_P_IPV6))
258 		e.addr.u.ip6 = mp->addr.dst.ip6;
259 #endif
260 	else
261 		ether_addr_copy(e.addr.u.mac_addr, mp->addr.dst.mac_addr);
262 	e.addr.proto = mp->addr.proto;
263 	nest_ent = nla_nest_start_noflag(skb,
264 					 MDBA_MDB_ENTRY_INFO);
265 	if (!nest_ent)
266 		return -EMSGSIZE;
267 
268 	if (nla_put_nohdr(skb, sizeof(e), &e) ||
269 	    nla_put_u32(skb,
270 			MDBA_MDB_EATTR_TIMER,
271 			br_timer_value(mtimer)))
272 		goto nest_err;
273 
274 	switch (mp->addr.proto) {
275 	case htons(ETH_P_IP):
276 		dump_srcs_mode = !!(mp->br->multicast_ctx.multicast_igmp_version == 3);
277 		if (mp->addr.src.ip4) {
278 			if (nla_put_in_addr(skb, MDBA_MDB_EATTR_SOURCE,
279 					    mp->addr.src.ip4))
280 				goto nest_err;
281 			break;
282 		}
283 		break;
284 #if IS_ENABLED(CONFIG_IPV6)
285 	case htons(ETH_P_IPV6):
286 		dump_srcs_mode = !!(mp->br->multicast_ctx.multicast_mld_version == 2);
287 		if (!ipv6_addr_any(&mp->addr.src.ip6)) {
288 			if (nla_put_in6_addr(skb, MDBA_MDB_EATTR_SOURCE,
289 					     &mp->addr.src.ip6))
290 				goto nest_err;
291 			break;
292 		}
293 		break;
294 #endif
295 	default:
296 		ether_addr_copy(e.addr.u.mac_addr, mp->addr.dst.mac_addr);
297 	}
298 	if (p) {
299 		if (nla_put_u8(skb, MDBA_MDB_EATTR_RTPROT, p->rt_protocol))
300 			goto nest_err;
301 		if (dump_srcs_mode &&
302 		    (__mdb_fill_srcs(skb, p) ||
303 		     nla_put_u8(skb, MDBA_MDB_EATTR_GROUP_MODE,
304 				p->filter_mode)))
305 			goto nest_err;
306 	}
307 	nla_nest_end(skb, nest_ent);
308 
309 	return 0;
310 
311 nest_err:
312 	nla_nest_cancel(skb, nest_ent);
313 	return -EMSGSIZE;
314 }
315 
316 static int br_mdb_fill_info(struct sk_buff *skb, struct netlink_callback *cb,
317 			    struct net_device *dev)
318 {
319 	int idx = 0, s_idx = cb->args[1], err = 0, pidx = 0, s_pidx = cb->args[2];
320 	struct net_bridge *br = netdev_priv(dev);
321 	struct net_bridge_mdb_entry *mp;
322 	struct nlattr *nest, *nest2;
323 
324 	if (!br_opt_get(br, BROPT_MULTICAST_ENABLED))
325 		return 0;
326 
327 	nest = nla_nest_start_noflag(skb, MDBA_MDB);
328 	if (nest == NULL)
329 		return -EMSGSIZE;
330 
331 	hlist_for_each_entry_rcu(mp, &br->mdb_list, mdb_node) {
332 		struct net_bridge_port_group *p;
333 		struct net_bridge_port_group __rcu **pp;
334 
335 		if (idx < s_idx)
336 			goto skip;
337 
338 		nest2 = nla_nest_start_noflag(skb, MDBA_MDB_ENTRY);
339 		if (!nest2) {
340 			err = -EMSGSIZE;
341 			break;
342 		}
343 
344 		if (!s_pidx && mp->host_joined) {
345 			err = __mdb_fill_info(skb, mp, NULL);
346 			if (err) {
347 				nla_nest_cancel(skb, nest2);
348 				break;
349 			}
350 		}
351 
352 		for (pp = &mp->ports; (p = rcu_dereference(*pp)) != NULL;
353 		      pp = &p->next) {
354 			if (!p->key.port)
355 				continue;
356 			if (pidx < s_pidx)
357 				goto skip_pg;
358 
359 			err = __mdb_fill_info(skb, mp, p);
360 			if (err) {
361 				nla_nest_end(skb, nest2);
362 				goto out;
363 			}
364 skip_pg:
365 			pidx++;
366 		}
367 		pidx = 0;
368 		s_pidx = 0;
369 		nla_nest_end(skb, nest2);
370 skip:
371 		idx++;
372 	}
373 
374 out:
375 	cb->args[1] = idx;
376 	cb->args[2] = pidx;
377 	nla_nest_end(skb, nest);
378 	return err;
379 }
380 
381 static int br_mdb_valid_dump_req(const struct nlmsghdr *nlh,
382 				 struct netlink_ext_ack *extack)
383 {
384 	struct br_port_msg *bpm;
385 
386 	if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*bpm))) {
387 		NL_SET_ERR_MSG_MOD(extack, "Invalid header for mdb dump request");
388 		return -EINVAL;
389 	}
390 
391 	bpm = nlmsg_data(nlh);
392 	if (bpm->ifindex) {
393 		NL_SET_ERR_MSG_MOD(extack, "Filtering by device index is not supported for mdb dump request");
394 		return -EINVAL;
395 	}
396 	if (nlmsg_attrlen(nlh, sizeof(*bpm))) {
397 		NL_SET_ERR_MSG(extack, "Invalid data after header in mdb dump request");
398 		return -EINVAL;
399 	}
400 
401 	return 0;
402 }
403 
404 static int br_mdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
405 {
406 	struct net_device *dev;
407 	struct net *net = sock_net(skb->sk);
408 	struct nlmsghdr *nlh = NULL;
409 	int idx = 0, s_idx;
410 
411 	if (cb->strict_check) {
412 		int err = br_mdb_valid_dump_req(cb->nlh, cb->extack);
413 
414 		if (err < 0)
415 			return err;
416 	}
417 
418 	s_idx = cb->args[0];
419 
420 	rcu_read_lock();
421 
422 	cb->seq = net->dev_base_seq;
423 
424 	for_each_netdev_rcu(net, dev) {
425 		if (netif_is_bridge_master(dev)) {
426 			struct net_bridge *br = netdev_priv(dev);
427 			struct br_port_msg *bpm;
428 
429 			if (idx < s_idx)
430 				goto skip;
431 
432 			nlh = nlmsg_put(skb, NETLINK_CB(cb->skb).portid,
433 					cb->nlh->nlmsg_seq, RTM_GETMDB,
434 					sizeof(*bpm), NLM_F_MULTI);
435 			if (nlh == NULL)
436 				break;
437 
438 			bpm = nlmsg_data(nlh);
439 			memset(bpm, 0, sizeof(*bpm));
440 			bpm->ifindex = dev->ifindex;
441 			if (br_mdb_fill_info(skb, cb, dev) < 0)
442 				goto out;
443 			if (br_rports_fill_info(skb, &br->multicast_ctx) < 0)
444 				goto out;
445 
446 			cb->args[1] = 0;
447 			nlmsg_end(skb, nlh);
448 		skip:
449 			idx++;
450 		}
451 	}
452 
453 out:
454 	if (nlh)
455 		nlmsg_end(skb, nlh);
456 	rcu_read_unlock();
457 	cb->args[0] = idx;
458 	return skb->len;
459 }
460 
461 static int nlmsg_populate_mdb_fill(struct sk_buff *skb,
462 				   struct net_device *dev,
463 				   struct net_bridge_mdb_entry *mp,
464 				   struct net_bridge_port_group *pg,
465 				   int type)
466 {
467 	struct nlmsghdr *nlh;
468 	struct br_port_msg *bpm;
469 	struct nlattr *nest, *nest2;
470 
471 	nlh = nlmsg_put(skb, 0, 0, type, sizeof(*bpm), 0);
472 	if (!nlh)
473 		return -EMSGSIZE;
474 
475 	bpm = nlmsg_data(nlh);
476 	memset(bpm, 0, sizeof(*bpm));
477 	bpm->family  = AF_BRIDGE;
478 	bpm->ifindex = dev->ifindex;
479 	nest = nla_nest_start_noflag(skb, MDBA_MDB);
480 	if (nest == NULL)
481 		goto cancel;
482 	nest2 = nla_nest_start_noflag(skb, MDBA_MDB_ENTRY);
483 	if (nest2 == NULL)
484 		goto end;
485 
486 	if (__mdb_fill_info(skb, mp, pg))
487 		goto end;
488 
489 	nla_nest_end(skb, nest2);
490 	nla_nest_end(skb, nest);
491 	nlmsg_end(skb, nlh);
492 	return 0;
493 
494 end:
495 	nla_nest_end(skb, nest);
496 cancel:
497 	nlmsg_cancel(skb, nlh);
498 	return -EMSGSIZE;
499 }
500 
501 static size_t rtnl_mdb_nlmsg_size(struct net_bridge_port_group *pg)
502 {
503 	size_t nlmsg_size = NLMSG_ALIGN(sizeof(struct br_port_msg)) +
504 			    nla_total_size(sizeof(struct br_mdb_entry)) +
505 			    nla_total_size(sizeof(u32));
506 	struct net_bridge_group_src *ent;
507 	size_t addr_size = 0;
508 
509 	if (!pg)
510 		goto out;
511 
512 	/* MDBA_MDB_EATTR_RTPROT */
513 	nlmsg_size += nla_total_size(sizeof(u8));
514 
515 	switch (pg->key.addr.proto) {
516 	case htons(ETH_P_IP):
517 		/* MDBA_MDB_EATTR_SOURCE */
518 		if (pg->key.addr.src.ip4)
519 			nlmsg_size += nla_total_size(sizeof(__be32));
520 		if (pg->key.port->br->multicast_ctx.multicast_igmp_version == 2)
521 			goto out;
522 		addr_size = sizeof(__be32);
523 		break;
524 #if IS_ENABLED(CONFIG_IPV6)
525 	case htons(ETH_P_IPV6):
526 		/* MDBA_MDB_EATTR_SOURCE */
527 		if (!ipv6_addr_any(&pg->key.addr.src.ip6))
528 			nlmsg_size += nla_total_size(sizeof(struct in6_addr));
529 		if (pg->key.port->br->multicast_ctx.multicast_mld_version == 1)
530 			goto out;
531 		addr_size = sizeof(struct in6_addr);
532 		break;
533 #endif
534 	}
535 
536 	/* MDBA_MDB_EATTR_GROUP_MODE */
537 	nlmsg_size += nla_total_size(sizeof(u8));
538 
539 	/* MDBA_MDB_EATTR_SRC_LIST nested attr */
540 	if (!hlist_empty(&pg->src_list))
541 		nlmsg_size += nla_total_size(0);
542 
543 	hlist_for_each_entry(ent, &pg->src_list, node) {
544 		/* MDBA_MDB_SRCLIST_ENTRY nested attr +
545 		 * MDBA_MDB_SRCATTR_ADDRESS + MDBA_MDB_SRCATTR_TIMER
546 		 */
547 		nlmsg_size += nla_total_size(0) +
548 			      nla_total_size(addr_size) +
549 			      nla_total_size(sizeof(u32));
550 	}
551 out:
552 	return nlmsg_size;
553 }
554 
555 void br_mdb_notify(struct net_device *dev,
556 		   struct net_bridge_mdb_entry *mp,
557 		   struct net_bridge_port_group *pg,
558 		   int type)
559 {
560 	struct net *net = dev_net(dev);
561 	struct sk_buff *skb;
562 	int err = -ENOBUFS;
563 
564 	br_switchdev_mdb_notify(dev, mp, pg, type);
565 
566 	skb = nlmsg_new(rtnl_mdb_nlmsg_size(pg), GFP_ATOMIC);
567 	if (!skb)
568 		goto errout;
569 
570 	err = nlmsg_populate_mdb_fill(skb, dev, mp, pg, type);
571 	if (err < 0) {
572 		kfree_skb(skb);
573 		goto errout;
574 	}
575 
576 	rtnl_notify(skb, net, 0, RTNLGRP_MDB, NULL, GFP_ATOMIC);
577 	return;
578 errout:
579 	rtnl_set_sk_err(net, RTNLGRP_MDB, err);
580 }
581 
582 static int nlmsg_populate_rtr_fill(struct sk_buff *skb,
583 				   struct net_device *dev,
584 				   int ifindex, u16 vid, u32 pid,
585 				   u32 seq, int type, unsigned int flags)
586 {
587 	struct nlattr *nest, *port_nest;
588 	struct br_port_msg *bpm;
589 	struct nlmsghdr *nlh;
590 
591 	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*bpm), 0);
592 	if (!nlh)
593 		return -EMSGSIZE;
594 
595 	bpm = nlmsg_data(nlh);
596 	memset(bpm, 0, sizeof(*bpm));
597 	bpm->family = AF_BRIDGE;
598 	bpm->ifindex = dev->ifindex;
599 	nest = nla_nest_start_noflag(skb, MDBA_ROUTER);
600 	if (!nest)
601 		goto cancel;
602 
603 	port_nest = nla_nest_start_noflag(skb, MDBA_ROUTER_PORT);
604 	if (!port_nest)
605 		goto end;
606 	if (nla_put_nohdr(skb, sizeof(u32), &ifindex)) {
607 		nla_nest_cancel(skb, port_nest);
608 		goto end;
609 	}
610 	if (vid && nla_put_u16(skb, MDBA_ROUTER_PATTR_VID, vid)) {
611 		nla_nest_cancel(skb, port_nest);
612 		goto end;
613 	}
614 	nla_nest_end(skb, port_nest);
615 
616 	nla_nest_end(skb, nest);
617 	nlmsg_end(skb, nlh);
618 	return 0;
619 
620 end:
621 	nla_nest_end(skb, nest);
622 cancel:
623 	nlmsg_cancel(skb, nlh);
624 	return -EMSGSIZE;
625 }
626 
627 static inline size_t rtnl_rtr_nlmsg_size(void)
628 {
629 	return NLMSG_ALIGN(sizeof(struct br_port_msg))
630 		+ nla_total_size(sizeof(__u32))
631 		+ nla_total_size(sizeof(u16));
632 }
633 
634 void br_rtr_notify(struct net_device *dev, struct net_bridge_mcast_port *pmctx,
635 		   int type)
636 {
637 	struct net *net = dev_net(dev);
638 	struct sk_buff *skb;
639 	int err = -ENOBUFS;
640 	int ifindex;
641 	u16 vid;
642 
643 	ifindex = pmctx ? pmctx->port->dev->ifindex : 0;
644 	vid = pmctx && br_multicast_port_ctx_is_vlan(pmctx) ? pmctx->vlan->vid :
645 							      0;
646 	skb = nlmsg_new(rtnl_rtr_nlmsg_size(), GFP_ATOMIC);
647 	if (!skb)
648 		goto errout;
649 
650 	err = nlmsg_populate_rtr_fill(skb, dev, ifindex, vid, 0, 0, type,
651 				      NTF_SELF);
652 	if (err < 0) {
653 		kfree_skb(skb);
654 		goto errout;
655 	}
656 
657 	rtnl_notify(skb, net, 0, RTNLGRP_MDB, NULL, GFP_ATOMIC);
658 	return;
659 
660 errout:
661 	rtnl_set_sk_err(net, RTNLGRP_MDB, err);
662 }
663 
664 static bool is_valid_mdb_entry(struct br_mdb_entry *entry,
665 			       struct netlink_ext_ack *extack)
666 {
667 	if (entry->ifindex == 0) {
668 		NL_SET_ERR_MSG_MOD(extack, "Zero entry ifindex is not allowed");
669 		return false;
670 	}
671 
672 	if (entry->addr.proto == htons(ETH_P_IP)) {
673 		if (!ipv4_is_multicast(entry->addr.u.ip4)) {
674 			NL_SET_ERR_MSG_MOD(extack, "IPv4 entry group address is not multicast");
675 			return false;
676 		}
677 		if (ipv4_is_local_multicast(entry->addr.u.ip4)) {
678 			NL_SET_ERR_MSG_MOD(extack, "IPv4 entry group address is local multicast");
679 			return false;
680 		}
681 #if IS_ENABLED(CONFIG_IPV6)
682 	} else if (entry->addr.proto == htons(ETH_P_IPV6)) {
683 		if (ipv6_addr_is_ll_all_nodes(&entry->addr.u.ip6)) {
684 			NL_SET_ERR_MSG_MOD(extack, "IPv6 entry group address is link-local all nodes");
685 			return false;
686 		}
687 #endif
688 	} else if (entry->addr.proto == 0) {
689 		/* L2 mdb */
690 		if (!is_multicast_ether_addr(entry->addr.u.mac_addr)) {
691 			NL_SET_ERR_MSG_MOD(extack, "L2 entry group is not multicast");
692 			return false;
693 		}
694 	} else {
695 		NL_SET_ERR_MSG_MOD(extack, "Unknown entry protocol");
696 		return false;
697 	}
698 
699 	if (entry->state != MDB_PERMANENT && entry->state != MDB_TEMPORARY) {
700 		NL_SET_ERR_MSG_MOD(extack, "Unknown entry state");
701 		return false;
702 	}
703 	if (entry->vid >= VLAN_VID_MASK) {
704 		NL_SET_ERR_MSG_MOD(extack, "Invalid entry VLAN id");
705 		return false;
706 	}
707 
708 	return true;
709 }
710 
711 static bool is_valid_mdb_source(struct nlattr *attr, __be16 proto,
712 				struct netlink_ext_ack *extack)
713 {
714 	switch (proto) {
715 	case htons(ETH_P_IP):
716 		if (nla_len(attr) != sizeof(struct in_addr)) {
717 			NL_SET_ERR_MSG_MOD(extack, "IPv4 invalid source address length");
718 			return false;
719 		}
720 		if (ipv4_is_multicast(nla_get_in_addr(attr))) {
721 			NL_SET_ERR_MSG_MOD(extack, "IPv4 multicast source address is not allowed");
722 			return false;
723 		}
724 		break;
725 #if IS_ENABLED(CONFIG_IPV6)
726 	case htons(ETH_P_IPV6): {
727 		struct in6_addr src;
728 
729 		if (nla_len(attr) != sizeof(struct in6_addr)) {
730 			NL_SET_ERR_MSG_MOD(extack, "IPv6 invalid source address length");
731 			return false;
732 		}
733 		src = nla_get_in6_addr(attr);
734 		if (ipv6_addr_is_multicast(&src)) {
735 			NL_SET_ERR_MSG_MOD(extack, "IPv6 multicast source address is not allowed");
736 			return false;
737 		}
738 		break;
739 	}
740 #endif
741 	default:
742 		NL_SET_ERR_MSG_MOD(extack, "Invalid protocol used with source address");
743 		return false;
744 	}
745 
746 	return true;
747 }
748 
749 static const struct nla_policy br_mdbe_attrs_pol[MDBE_ATTR_MAX + 1] = {
750 	[MDBE_ATTR_SOURCE] = NLA_POLICY_RANGE(NLA_BINARY,
751 					      sizeof(struct in_addr),
752 					      sizeof(struct in6_addr)),
753 };
754 
755 static int br_mdb_parse(struct sk_buff *skb, struct nlmsghdr *nlh,
756 			struct net_device **pdev, struct br_mdb_entry **pentry,
757 			struct nlattr **mdb_attrs, struct netlink_ext_ack *extack)
758 {
759 	struct net *net = sock_net(skb->sk);
760 	struct br_mdb_entry *entry;
761 	struct br_port_msg *bpm;
762 	struct nlattr *tb[MDBA_SET_ENTRY_MAX+1];
763 	struct net_device *dev;
764 	int err;
765 
766 	err = nlmsg_parse_deprecated(nlh, sizeof(*bpm), tb,
767 				     MDBA_SET_ENTRY_MAX, NULL, NULL);
768 	if (err < 0)
769 		return err;
770 
771 	bpm = nlmsg_data(nlh);
772 	if (bpm->ifindex == 0) {
773 		NL_SET_ERR_MSG_MOD(extack, "Invalid bridge ifindex");
774 		return -EINVAL;
775 	}
776 
777 	dev = __dev_get_by_index(net, bpm->ifindex);
778 	if (dev == NULL) {
779 		NL_SET_ERR_MSG_MOD(extack, "Bridge device doesn't exist");
780 		return -ENODEV;
781 	}
782 
783 	if (!netif_is_bridge_master(dev)) {
784 		NL_SET_ERR_MSG_MOD(extack, "Device is not a bridge");
785 		return -EOPNOTSUPP;
786 	}
787 
788 	*pdev = dev;
789 
790 	if (!tb[MDBA_SET_ENTRY]) {
791 		NL_SET_ERR_MSG_MOD(extack, "Missing MDBA_SET_ENTRY attribute");
792 		return -EINVAL;
793 	}
794 	if (nla_len(tb[MDBA_SET_ENTRY]) != sizeof(struct br_mdb_entry)) {
795 		NL_SET_ERR_MSG_MOD(extack, "Invalid MDBA_SET_ENTRY attribute length");
796 		return -EINVAL;
797 	}
798 
799 	entry = nla_data(tb[MDBA_SET_ENTRY]);
800 	if (!is_valid_mdb_entry(entry, extack))
801 		return -EINVAL;
802 	*pentry = entry;
803 
804 	if (tb[MDBA_SET_ENTRY_ATTRS]) {
805 		err = nla_parse_nested(mdb_attrs, MDBE_ATTR_MAX,
806 				       tb[MDBA_SET_ENTRY_ATTRS],
807 				       br_mdbe_attrs_pol, extack);
808 		if (err)
809 			return err;
810 		if (mdb_attrs[MDBE_ATTR_SOURCE] &&
811 		    !is_valid_mdb_source(mdb_attrs[MDBE_ATTR_SOURCE],
812 					 entry->addr.proto, extack))
813 			return -EINVAL;
814 	} else {
815 		memset(mdb_attrs, 0,
816 		       sizeof(struct nlattr *) * (MDBE_ATTR_MAX + 1));
817 	}
818 
819 	return 0;
820 }
821 
822 static struct net_bridge_mcast *
823 __br_mdb_choose_context(struct net_bridge *br,
824 			const struct br_mdb_entry *entry,
825 			struct netlink_ext_ack *extack)
826 {
827 	struct net_bridge_mcast *brmctx = NULL;
828 	struct net_bridge_vlan *v;
829 
830 	if (!br_opt_get(br, BROPT_MCAST_VLAN_SNOOPING_ENABLED)) {
831 		brmctx = &br->multicast_ctx;
832 		goto out;
833 	}
834 
835 	if (!entry->vid) {
836 		NL_SET_ERR_MSG_MOD(extack, "Cannot add an entry without a vlan when vlan snooping is enabled");
837 		goto out;
838 	}
839 
840 	v = br_vlan_find(br_vlan_group(br), entry->vid);
841 	if (!v) {
842 		NL_SET_ERR_MSG_MOD(extack, "Vlan is not configured");
843 		goto out;
844 	}
845 	if (br_multicast_ctx_vlan_global_disabled(&v->br_mcast_ctx)) {
846 		NL_SET_ERR_MSG_MOD(extack, "Vlan's multicast processing is disabled");
847 		goto out;
848 	}
849 	brmctx = &v->br_mcast_ctx;
850 out:
851 	return brmctx;
852 }
853 
854 static int br_mdb_add_group(struct net_bridge *br, struct net_bridge_port *port,
855 			    struct br_mdb_entry *entry,
856 			    struct nlattr **mdb_attrs,
857 			    struct netlink_ext_ack *extack)
858 {
859 	struct net_bridge_mdb_entry *mp, *star_mp;
860 	struct net_bridge_port_group __rcu **pp;
861 	struct net_bridge_port_group *p;
862 	struct net_bridge_mcast *brmctx;
863 	struct br_ip group, star_group;
864 	unsigned long now = jiffies;
865 	unsigned char flags = 0;
866 	u8 filter_mode;
867 	int err;
868 
869 	__mdb_entry_to_br_ip(entry, &group, mdb_attrs);
870 
871 	brmctx = __br_mdb_choose_context(br, entry, extack);
872 	if (!brmctx)
873 		return -EINVAL;
874 
875 	/* host join errors which can happen before creating the group */
876 	if (!port) {
877 		/* don't allow any flags for host-joined groups */
878 		if (entry->state) {
879 			NL_SET_ERR_MSG_MOD(extack, "Flags are not allowed for host groups");
880 			return -EINVAL;
881 		}
882 		if (!br_multicast_is_star_g(&group)) {
883 			NL_SET_ERR_MSG_MOD(extack, "Groups with sources cannot be manually host joined");
884 			return -EINVAL;
885 		}
886 	}
887 
888 	if (br_group_is_l2(&group) && entry->state != MDB_PERMANENT) {
889 		NL_SET_ERR_MSG_MOD(extack, "Only permanent L2 entries allowed");
890 		return -EINVAL;
891 	}
892 
893 	mp = br_mdb_ip_get(br, &group);
894 	if (!mp) {
895 		mp = br_multicast_new_group(br, &group);
896 		err = PTR_ERR_OR_ZERO(mp);
897 		if (err)
898 			return err;
899 	}
900 
901 	/* host join */
902 	if (!port) {
903 		if (mp->host_joined) {
904 			NL_SET_ERR_MSG_MOD(extack, "Group is already joined by host");
905 			return -EEXIST;
906 		}
907 
908 		br_multicast_host_join(brmctx, mp, false);
909 		br_mdb_notify(br->dev, mp, NULL, RTM_NEWMDB);
910 
911 		return 0;
912 	}
913 
914 	for (pp = &mp->ports;
915 	     (p = mlock_dereference(*pp, br)) != NULL;
916 	     pp = &p->next) {
917 		if (p->key.port == port) {
918 			NL_SET_ERR_MSG_MOD(extack, "Group is already joined by port");
919 			return -EEXIST;
920 		}
921 		if ((unsigned long)p->key.port < (unsigned long)port)
922 			break;
923 	}
924 
925 	filter_mode = br_multicast_is_star_g(&group) ? MCAST_EXCLUDE :
926 						       MCAST_INCLUDE;
927 
928 	if (entry->state == MDB_PERMANENT)
929 		flags |= MDB_PG_FLAGS_PERMANENT;
930 
931 	p = br_multicast_new_port_group(port, &group, *pp, flags, NULL,
932 					filter_mode, RTPROT_STATIC);
933 	if (unlikely(!p)) {
934 		NL_SET_ERR_MSG_MOD(extack, "Couldn't allocate new port group");
935 		return -ENOMEM;
936 	}
937 	rcu_assign_pointer(*pp, p);
938 	if (entry->state == MDB_TEMPORARY)
939 		mod_timer(&p->timer,
940 			  now + brmctx->multicast_membership_interval);
941 	br_mdb_notify(br->dev, mp, p, RTM_NEWMDB);
942 	/* if we are adding a new EXCLUDE port group (*,G) it needs to be also
943 	 * added to all S,G entries for proper replication, if we are adding
944 	 * a new INCLUDE port (S,G) then all of *,G EXCLUDE ports need to be
945 	 * added to it for proper replication
946 	 */
947 	if (br_multicast_should_handle_mode(brmctx, group.proto)) {
948 		switch (filter_mode) {
949 		case MCAST_EXCLUDE:
950 			br_multicast_star_g_handle_mode(p, MCAST_EXCLUDE);
951 			break;
952 		case MCAST_INCLUDE:
953 			star_group = p->key.addr;
954 			memset(&star_group.src, 0, sizeof(star_group.src));
955 			star_mp = br_mdb_ip_get(br, &star_group);
956 			if (star_mp)
957 				br_multicast_sg_add_exclude_ports(star_mp, p);
958 			break;
959 		}
960 	}
961 
962 	return 0;
963 }
964 
965 static int __br_mdb_add(struct net *net, struct net_bridge *br,
966 			struct net_bridge_port *p,
967 			struct br_mdb_entry *entry,
968 			struct nlattr **mdb_attrs,
969 			struct netlink_ext_ack *extack)
970 {
971 	int ret;
972 
973 	spin_lock_bh(&br->multicast_lock);
974 	ret = br_mdb_add_group(br, p, entry, mdb_attrs, extack);
975 	spin_unlock_bh(&br->multicast_lock);
976 
977 	return ret;
978 }
979 
980 static int br_mdb_add(struct sk_buff *skb, struct nlmsghdr *nlh,
981 		      struct netlink_ext_ack *extack)
982 {
983 	struct nlattr *mdb_attrs[MDBE_ATTR_MAX + 1];
984 	struct net *net = sock_net(skb->sk);
985 	struct net_bridge_vlan_group *vg;
986 	struct net_bridge_port *p = NULL;
987 	struct net_device *dev, *pdev;
988 	struct br_mdb_entry *entry;
989 	struct net_bridge_vlan *v;
990 	struct net_bridge *br;
991 	int err;
992 
993 	err = br_mdb_parse(skb, nlh, &dev, &entry, mdb_attrs, extack);
994 	if (err < 0)
995 		return err;
996 
997 	br = netdev_priv(dev);
998 
999 	if (!netif_running(br->dev)) {
1000 		NL_SET_ERR_MSG_MOD(extack, "Bridge device is not running");
1001 		return -EINVAL;
1002 	}
1003 
1004 	if (!br_opt_get(br, BROPT_MULTICAST_ENABLED)) {
1005 		NL_SET_ERR_MSG_MOD(extack, "Bridge's multicast processing is disabled");
1006 		return -EINVAL;
1007 	}
1008 
1009 	if (entry->ifindex != br->dev->ifindex) {
1010 		pdev = __dev_get_by_index(net, entry->ifindex);
1011 		if (!pdev) {
1012 			NL_SET_ERR_MSG_MOD(extack, "Port net device doesn't exist");
1013 			return -ENODEV;
1014 		}
1015 
1016 		p = br_port_get_rtnl(pdev);
1017 		if (!p) {
1018 			NL_SET_ERR_MSG_MOD(extack, "Net device is not a bridge port");
1019 			return -EINVAL;
1020 		}
1021 
1022 		if (p->br != br) {
1023 			NL_SET_ERR_MSG_MOD(extack, "Port belongs to a different bridge device");
1024 			return -EINVAL;
1025 		}
1026 		if (p->state == BR_STATE_DISABLED) {
1027 			NL_SET_ERR_MSG_MOD(extack, "Port is in disabled state");
1028 			return -EINVAL;
1029 		}
1030 		vg = nbp_vlan_group(p);
1031 	} else {
1032 		vg = br_vlan_group(br);
1033 	}
1034 
1035 	/* If vlan filtering is enabled and VLAN is not specified
1036 	 * install mdb entry on all vlans configured on the port.
1037 	 */
1038 	if (br_vlan_enabled(br->dev) && vg && entry->vid == 0) {
1039 		list_for_each_entry(v, &vg->vlan_list, vlist) {
1040 			entry->vid = v->vid;
1041 			err = __br_mdb_add(net, br, p, entry, mdb_attrs, extack);
1042 			if (err)
1043 				break;
1044 		}
1045 	} else {
1046 		err = __br_mdb_add(net, br, p, entry, mdb_attrs, extack);
1047 	}
1048 
1049 	return err;
1050 }
1051 
1052 static int __br_mdb_del(struct net_bridge *br, struct br_mdb_entry *entry,
1053 			struct nlattr **mdb_attrs)
1054 {
1055 	struct net_bridge_mdb_entry *mp;
1056 	struct net_bridge_port_group *p;
1057 	struct net_bridge_port_group __rcu **pp;
1058 	struct br_ip ip;
1059 	int err = -EINVAL;
1060 
1061 	if (!netif_running(br->dev) || !br_opt_get(br, BROPT_MULTICAST_ENABLED))
1062 		return -EINVAL;
1063 
1064 	__mdb_entry_to_br_ip(entry, &ip, mdb_attrs);
1065 
1066 	spin_lock_bh(&br->multicast_lock);
1067 	mp = br_mdb_ip_get(br, &ip);
1068 	if (!mp)
1069 		goto unlock;
1070 
1071 	/* host leave */
1072 	if (entry->ifindex == mp->br->dev->ifindex && mp->host_joined) {
1073 		br_multicast_host_leave(mp, false);
1074 		err = 0;
1075 		br_mdb_notify(br->dev, mp, NULL, RTM_DELMDB);
1076 		if (!mp->ports && netif_running(br->dev))
1077 			mod_timer(&mp->timer, jiffies);
1078 		goto unlock;
1079 	}
1080 
1081 	for (pp = &mp->ports;
1082 	     (p = mlock_dereference(*pp, br)) != NULL;
1083 	     pp = &p->next) {
1084 		if (!p->key.port || p->key.port->dev->ifindex != entry->ifindex)
1085 			continue;
1086 
1087 		if (p->key.port->state == BR_STATE_DISABLED)
1088 			goto unlock;
1089 
1090 		br_multicast_del_pg(mp, p, pp);
1091 		err = 0;
1092 		break;
1093 	}
1094 
1095 unlock:
1096 	spin_unlock_bh(&br->multicast_lock);
1097 	return err;
1098 }
1099 
1100 static int br_mdb_del(struct sk_buff *skb, struct nlmsghdr *nlh,
1101 		      struct netlink_ext_ack *extack)
1102 {
1103 	struct nlattr *mdb_attrs[MDBE_ATTR_MAX + 1];
1104 	struct net *net = sock_net(skb->sk);
1105 	struct net_bridge_vlan_group *vg;
1106 	struct net_bridge_port *p = NULL;
1107 	struct net_device *dev, *pdev;
1108 	struct br_mdb_entry *entry;
1109 	struct net_bridge_vlan *v;
1110 	struct net_bridge *br;
1111 	int err;
1112 
1113 	err = br_mdb_parse(skb, nlh, &dev, &entry, mdb_attrs, extack);
1114 	if (err < 0)
1115 		return err;
1116 
1117 	br = netdev_priv(dev);
1118 
1119 	if (entry->ifindex != br->dev->ifindex) {
1120 		pdev = __dev_get_by_index(net, entry->ifindex);
1121 		if (!pdev)
1122 			return -ENODEV;
1123 
1124 		p = br_port_get_rtnl(pdev);
1125 		if (!p || p->br != br || p->state == BR_STATE_DISABLED)
1126 			return -EINVAL;
1127 		vg = nbp_vlan_group(p);
1128 	} else {
1129 		vg = br_vlan_group(br);
1130 	}
1131 
1132 	/* If vlan filtering is enabled and VLAN is not specified
1133 	 * delete mdb entry on all vlans configured on the port.
1134 	 */
1135 	if (br_vlan_enabled(br->dev) && vg && entry->vid == 0) {
1136 		list_for_each_entry(v, &vg->vlan_list, vlist) {
1137 			entry->vid = v->vid;
1138 			err = __br_mdb_del(br, entry, mdb_attrs);
1139 		}
1140 	} else {
1141 		err = __br_mdb_del(br, entry, mdb_attrs);
1142 	}
1143 
1144 	return err;
1145 }
1146 
1147 void br_mdb_init(void)
1148 {
1149 	rtnl_register_module(THIS_MODULE, PF_BRIDGE, RTM_GETMDB, NULL, br_mdb_dump, 0);
1150 	rtnl_register_module(THIS_MODULE, PF_BRIDGE, RTM_NEWMDB, br_mdb_add, NULL, 0);
1151 	rtnl_register_module(THIS_MODULE, PF_BRIDGE, RTM_DELMDB, br_mdb_del, NULL, 0);
1152 }
1153 
1154 void br_mdb_uninit(void)
1155 {
1156 	rtnl_unregister(PF_BRIDGE, RTM_GETMDB);
1157 	rtnl_unregister(PF_BRIDGE, RTM_NEWMDB);
1158 	rtnl_unregister(PF_BRIDGE, RTM_DELMDB);
1159 }
1160