xref: /openbmc/linux/net/bridge/br_mdb.c (revision 871808fd)
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.state = MDB_PG_FLAGS_PERMANENT;
263 	}
264 	e.addr.proto = mp->addr.proto;
265 	nest_ent = nla_nest_start_noflag(skb,
266 					 MDBA_MDB_ENTRY_INFO);
267 	if (!nest_ent)
268 		return -EMSGSIZE;
269 
270 	if (nla_put_nohdr(skb, sizeof(e), &e) ||
271 	    nla_put_u32(skb,
272 			MDBA_MDB_EATTR_TIMER,
273 			br_timer_value(mtimer)))
274 		goto nest_err;
275 
276 	switch (mp->addr.proto) {
277 	case htons(ETH_P_IP):
278 		dump_srcs_mode = !!(mp->br->multicast_ctx.multicast_igmp_version == 3);
279 		if (mp->addr.src.ip4) {
280 			if (nla_put_in_addr(skb, MDBA_MDB_EATTR_SOURCE,
281 					    mp->addr.src.ip4))
282 				goto nest_err;
283 			break;
284 		}
285 		break;
286 #if IS_ENABLED(CONFIG_IPV6)
287 	case htons(ETH_P_IPV6):
288 		dump_srcs_mode = !!(mp->br->multicast_ctx.multicast_mld_version == 2);
289 		if (!ipv6_addr_any(&mp->addr.src.ip6)) {
290 			if (nla_put_in6_addr(skb, MDBA_MDB_EATTR_SOURCE,
291 					     &mp->addr.src.ip6))
292 				goto nest_err;
293 			break;
294 		}
295 		break;
296 #endif
297 	default:
298 		ether_addr_copy(e.addr.u.mac_addr, mp->addr.dst.mac_addr);
299 	}
300 	if (p) {
301 		if (nla_put_u8(skb, MDBA_MDB_EATTR_RTPROT, p->rt_protocol))
302 			goto nest_err;
303 		if (dump_srcs_mode &&
304 		    (__mdb_fill_srcs(skb, p) ||
305 		     nla_put_u8(skb, MDBA_MDB_EATTR_GROUP_MODE,
306 				p->filter_mode)))
307 			goto nest_err;
308 	}
309 	nla_nest_end(skb, nest_ent);
310 
311 	return 0;
312 
313 nest_err:
314 	nla_nest_cancel(skb, nest_ent);
315 	return -EMSGSIZE;
316 }
317 
318 static int br_mdb_fill_info(struct sk_buff *skb, struct netlink_callback *cb,
319 			    struct net_device *dev)
320 {
321 	int idx = 0, s_idx = cb->args[1], err = 0, pidx = 0, s_pidx = cb->args[2];
322 	struct net_bridge *br = netdev_priv(dev);
323 	struct net_bridge_mdb_entry *mp;
324 	struct nlattr *nest, *nest2;
325 
326 	if (!br_opt_get(br, BROPT_MULTICAST_ENABLED))
327 		return 0;
328 
329 	nest = nla_nest_start_noflag(skb, MDBA_MDB);
330 	if (nest == NULL)
331 		return -EMSGSIZE;
332 
333 	hlist_for_each_entry_rcu(mp, &br->mdb_list, mdb_node) {
334 		struct net_bridge_port_group *p;
335 		struct net_bridge_port_group __rcu **pp;
336 
337 		if (idx < s_idx)
338 			goto skip;
339 
340 		nest2 = nla_nest_start_noflag(skb, MDBA_MDB_ENTRY);
341 		if (!nest2) {
342 			err = -EMSGSIZE;
343 			break;
344 		}
345 
346 		if (!s_pidx && mp->host_joined) {
347 			err = __mdb_fill_info(skb, mp, NULL);
348 			if (err) {
349 				nla_nest_cancel(skb, nest2);
350 				break;
351 			}
352 		}
353 
354 		for (pp = &mp->ports; (p = rcu_dereference(*pp)) != NULL;
355 		      pp = &p->next) {
356 			if (!p->key.port)
357 				continue;
358 			if (pidx < s_pidx)
359 				goto skip_pg;
360 
361 			err = __mdb_fill_info(skb, mp, p);
362 			if (err) {
363 				nla_nest_end(skb, nest2);
364 				goto out;
365 			}
366 skip_pg:
367 			pidx++;
368 		}
369 		pidx = 0;
370 		s_pidx = 0;
371 		nla_nest_end(skb, nest2);
372 skip:
373 		idx++;
374 	}
375 
376 out:
377 	cb->args[1] = idx;
378 	cb->args[2] = pidx;
379 	nla_nest_end(skb, nest);
380 	return err;
381 }
382 
383 static int br_mdb_valid_dump_req(const struct nlmsghdr *nlh,
384 				 struct netlink_ext_ack *extack)
385 {
386 	struct br_port_msg *bpm;
387 
388 	if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*bpm))) {
389 		NL_SET_ERR_MSG_MOD(extack, "Invalid header for mdb dump request");
390 		return -EINVAL;
391 	}
392 
393 	bpm = nlmsg_data(nlh);
394 	if (bpm->ifindex) {
395 		NL_SET_ERR_MSG_MOD(extack, "Filtering by device index is not supported for mdb dump request");
396 		return -EINVAL;
397 	}
398 	if (nlmsg_attrlen(nlh, sizeof(*bpm))) {
399 		NL_SET_ERR_MSG(extack, "Invalid data after header in mdb dump request");
400 		return -EINVAL;
401 	}
402 
403 	return 0;
404 }
405 
406 static int br_mdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
407 {
408 	struct net_device *dev;
409 	struct net *net = sock_net(skb->sk);
410 	struct nlmsghdr *nlh = NULL;
411 	int idx = 0, s_idx;
412 
413 	if (cb->strict_check) {
414 		int err = br_mdb_valid_dump_req(cb->nlh, cb->extack);
415 
416 		if (err < 0)
417 			return err;
418 	}
419 
420 	s_idx = cb->args[0];
421 
422 	rcu_read_lock();
423 
424 	cb->seq = net->dev_base_seq;
425 
426 	for_each_netdev_rcu(net, dev) {
427 		if (netif_is_bridge_master(dev)) {
428 			struct net_bridge *br = netdev_priv(dev);
429 			struct br_port_msg *bpm;
430 
431 			if (idx < s_idx)
432 				goto skip;
433 
434 			nlh = nlmsg_put(skb, NETLINK_CB(cb->skb).portid,
435 					cb->nlh->nlmsg_seq, RTM_GETMDB,
436 					sizeof(*bpm), NLM_F_MULTI);
437 			if (nlh == NULL)
438 				break;
439 
440 			bpm = nlmsg_data(nlh);
441 			memset(bpm, 0, sizeof(*bpm));
442 			bpm->ifindex = dev->ifindex;
443 			if (br_mdb_fill_info(skb, cb, dev) < 0)
444 				goto out;
445 			if (br_rports_fill_info(skb, &br->multicast_ctx) < 0)
446 				goto out;
447 
448 			cb->args[1] = 0;
449 			nlmsg_end(skb, nlh);
450 		skip:
451 			idx++;
452 		}
453 	}
454 
455 out:
456 	if (nlh)
457 		nlmsg_end(skb, nlh);
458 	rcu_read_unlock();
459 	cb->args[0] = idx;
460 	return skb->len;
461 }
462 
463 static int nlmsg_populate_mdb_fill(struct sk_buff *skb,
464 				   struct net_device *dev,
465 				   struct net_bridge_mdb_entry *mp,
466 				   struct net_bridge_port_group *pg,
467 				   int type)
468 {
469 	struct nlmsghdr *nlh;
470 	struct br_port_msg *bpm;
471 	struct nlattr *nest, *nest2;
472 
473 	nlh = nlmsg_put(skb, 0, 0, type, sizeof(*bpm), 0);
474 	if (!nlh)
475 		return -EMSGSIZE;
476 
477 	bpm = nlmsg_data(nlh);
478 	memset(bpm, 0, sizeof(*bpm));
479 	bpm->family  = AF_BRIDGE;
480 	bpm->ifindex = dev->ifindex;
481 	nest = nla_nest_start_noflag(skb, MDBA_MDB);
482 	if (nest == NULL)
483 		goto cancel;
484 	nest2 = nla_nest_start_noflag(skb, MDBA_MDB_ENTRY);
485 	if (nest2 == NULL)
486 		goto end;
487 
488 	if (__mdb_fill_info(skb, mp, pg))
489 		goto end;
490 
491 	nla_nest_end(skb, nest2);
492 	nla_nest_end(skb, nest);
493 	nlmsg_end(skb, nlh);
494 	return 0;
495 
496 end:
497 	nla_nest_end(skb, nest);
498 cancel:
499 	nlmsg_cancel(skb, nlh);
500 	return -EMSGSIZE;
501 }
502 
503 static size_t rtnl_mdb_nlmsg_size(struct net_bridge_port_group *pg)
504 {
505 	size_t nlmsg_size = NLMSG_ALIGN(sizeof(struct br_port_msg)) +
506 			    nla_total_size(sizeof(struct br_mdb_entry)) +
507 			    nla_total_size(sizeof(u32));
508 	struct net_bridge_group_src *ent;
509 	size_t addr_size = 0;
510 
511 	if (!pg)
512 		goto out;
513 
514 	/* MDBA_MDB_EATTR_RTPROT */
515 	nlmsg_size += nla_total_size(sizeof(u8));
516 
517 	switch (pg->key.addr.proto) {
518 	case htons(ETH_P_IP):
519 		/* MDBA_MDB_EATTR_SOURCE */
520 		if (pg->key.addr.src.ip4)
521 			nlmsg_size += nla_total_size(sizeof(__be32));
522 		if (pg->key.port->br->multicast_ctx.multicast_igmp_version == 2)
523 			goto out;
524 		addr_size = sizeof(__be32);
525 		break;
526 #if IS_ENABLED(CONFIG_IPV6)
527 	case htons(ETH_P_IPV6):
528 		/* MDBA_MDB_EATTR_SOURCE */
529 		if (!ipv6_addr_any(&pg->key.addr.src.ip6))
530 			nlmsg_size += nla_total_size(sizeof(struct in6_addr));
531 		if (pg->key.port->br->multicast_ctx.multicast_mld_version == 1)
532 			goto out;
533 		addr_size = sizeof(struct in6_addr);
534 		break;
535 #endif
536 	}
537 
538 	/* MDBA_MDB_EATTR_GROUP_MODE */
539 	nlmsg_size += nla_total_size(sizeof(u8));
540 
541 	/* MDBA_MDB_EATTR_SRC_LIST nested attr */
542 	if (!hlist_empty(&pg->src_list))
543 		nlmsg_size += nla_total_size(0);
544 
545 	hlist_for_each_entry(ent, &pg->src_list, node) {
546 		/* MDBA_MDB_SRCLIST_ENTRY nested attr +
547 		 * MDBA_MDB_SRCATTR_ADDRESS + MDBA_MDB_SRCATTR_TIMER
548 		 */
549 		nlmsg_size += nla_total_size(0) +
550 			      nla_total_size(addr_size) +
551 			      nla_total_size(sizeof(u32));
552 	}
553 out:
554 	return nlmsg_size;
555 }
556 
557 void br_mdb_notify(struct net_device *dev,
558 		   struct net_bridge_mdb_entry *mp,
559 		   struct net_bridge_port_group *pg,
560 		   int type)
561 {
562 	struct net *net = dev_net(dev);
563 	struct sk_buff *skb;
564 	int err = -ENOBUFS;
565 
566 	br_switchdev_mdb_notify(dev, mp, pg, type);
567 
568 	skb = nlmsg_new(rtnl_mdb_nlmsg_size(pg), GFP_ATOMIC);
569 	if (!skb)
570 		goto errout;
571 
572 	err = nlmsg_populate_mdb_fill(skb, dev, mp, pg, type);
573 	if (err < 0) {
574 		kfree_skb(skb);
575 		goto errout;
576 	}
577 
578 	rtnl_notify(skb, net, 0, RTNLGRP_MDB, NULL, GFP_ATOMIC);
579 	return;
580 errout:
581 	rtnl_set_sk_err(net, RTNLGRP_MDB, err);
582 }
583 
584 static int nlmsg_populate_rtr_fill(struct sk_buff *skb,
585 				   struct net_device *dev,
586 				   int ifindex, u16 vid, u32 pid,
587 				   u32 seq, int type, unsigned int flags)
588 {
589 	struct nlattr *nest, *port_nest;
590 	struct br_port_msg *bpm;
591 	struct nlmsghdr *nlh;
592 
593 	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*bpm), 0);
594 	if (!nlh)
595 		return -EMSGSIZE;
596 
597 	bpm = nlmsg_data(nlh);
598 	memset(bpm, 0, sizeof(*bpm));
599 	bpm->family = AF_BRIDGE;
600 	bpm->ifindex = dev->ifindex;
601 	nest = nla_nest_start_noflag(skb, MDBA_ROUTER);
602 	if (!nest)
603 		goto cancel;
604 
605 	port_nest = nla_nest_start_noflag(skb, MDBA_ROUTER_PORT);
606 	if (!port_nest)
607 		goto end;
608 	if (nla_put_nohdr(skb, sizeof(u32), &ifindex)) {
609 		nla_nest_cancel(skb, port_nest);
610 		goto end;
611 	}
612 	if (vid && nla_put_u16(skb, MDBA_ROUTER_PATTR_VID, vid)) {
613 		nla_nest_cancel(skb, port_nest);
614 		goto end;
615 	}
616 	nla_nest_end(skb, port_nest);
617 
618 	nla_nest_end(skb, nest);
619 	nlmsg_end(skb, nlh);
620 	return 0;
621 
622 end:
623 	nla_nest_end(skb, nest);
624 cancel:
625 	nlmsg_cancel(skb, nlh);
626 	return -EMSGSIZE;
627 }
628 
629 static inline size_t rtnl_rtr_nlmsg_size(void)
630 {
631 	return NLMSG_ALIGN(sizeof(struct br_port_msg))
632 		+ nla_total_size(sizeof(__u32))
633 		+ nla_total_size(sizeof(u16));
634 }
635 
636 void br_rtr_notify(struct net_device *dev, struct net_bridge_mcast_port *pmctx,
637 		   int type)
638 {
639 	struct net *net = dev_net(dev);
640 	struct sk_buff *skb;
641 	int err = -ENOBUFS;
642 	int ifindex;
643 	u16 vid;
644 
645 	ifindex = pmctx ? pmctx->port->dev->ifindex : 0;
646 	vid = pmctx && br_multicast_port_ctx_is_vlan(pmctx) ? pmctx->vlan->vid :
647 							      0;
648 	skb = nlmsg_new(rtnl_rtr_nlmsg_size(), GFP_ATOMIC);
649 	if (!skb)
650 		goto errout;
651 
652 	err = nlmsg_populate_rtr_fill(skb, dev, ifindex, vid, 0, 0, type,
653 				      NTF_SELF);
654 	if (err < 0) {
655 		kfree_skb(skb);
656 		goto errout;
657 	}
658 
659 	rtnl_notify(skb, net, 0, RTNLGRP_MDB, NULL, GFP_ATOMIC);
660 	return;
661 
662 errout:
663 	rtnl_set_sk_err(net, RTNLGRP_MDB, err);
664 }
665 
666 static bool is_valid_mdb_entry(struct br_mdb_entry *entry,
667 			       struct netlink_ext_ack *extack)
668 {
669 	if (entry->ifindex == 0) {
670 		NL_SET_ERR_MSG_MOD(extack, "Zero entry ifindex is not allowed");
671 		return false;
672 	}
673 
674 	if (entry->addr.proto == htons(ETH_P_IP)) {
675 		if (!ipv4_is_multicast(entry->addr.u.ip4)) {
676 			NL_SET_ERR_MSG_MOD(extack, "IPv4 entry group address is not multicast");
677 			return false;
678 		}
679 		if (ipv4_is_local_multicast(entry->addr.u.ip4)) {
680 			NL_SET_ERR_MSG_MOD(extack, "IPv4 entry group address is local multicast");
681 			return false;
682 		}
683 #if IS_ENABLED(CONFIG_IPV6)
684 	} else if (entry->addr.proto == htons(ETH_P_IPV6)) {
685 		if (ipv6_addr_is_ll_all_nodes(&entry->addr.u.ip6)) {
686 			NL_SET_ERR_MSG_MOD(extack, "IPv6 entry group address is link-local all nodes");
687 			return false;
688 		}
689 #endif
690 	} else if (entry->addr.proto == 0) {
691 		/* L2 mdb */
692 		if (!is_multicast_ether_addr(entry->addr.u.mac_addr)) {
693 			NL_SET_ERR_MSG_MOD(extack, "L2 entry group is not multicast");
694 			return false;
695 		}
696 	} else {
697 		NL_SET_ERR_MSG_MOD(extack, "Unknown entry protocol");
698 		return false;
699 	}
700 
701 	if (entry->state != MDB_PERMANENT && entry->state != MDB_TEMPORARY) {
702 		NL_SET_ERR_MSG_MOD(extack, "Unknown entry state");
703 		return false;
704 	}
705 	if (entry->vid >= VLAN_VID_MASK) {
706 		NL_SET_ERR_MSG_MOD(extack, "Invalid entry VLAN id");
707 		return false;
708 	}
709 
710 	return true;
711 }
712 
713 static bool is_valid_mdb_source(struct nlattr *attr, __be16 proto,
714 				struct netlink_ext_ack *extack)
715 {
716 	switch (proto) {
717 	case htons(ETH_P_IP):
718 		if (nla_len(attr) != sizeof(struct in_addr)) {
719 			NL_SET_ERR_MSG_MOD(extack, "IPv4 invalid source address length");
720 			return false;
721 		}
722 		if (ipv4_is_multicast(nla_get_in_addr(attr))) {
723 			NL_SET_ERR_MSG_MOD(extack, "IPv4 multicast source address is not allowed");
724 			return false;
725 		}
726 		break;
727 #if IS_ENABLED(CONFIG_IPV6)
728 	case htons(ETH_P_IPV6): {
729 		struct in6_addr src;
730 
731 		if (nla_len(attr) != sizeof(struct in6_addr)) {
732 			NL_SET_ERR_MSG_MOD(extack, "IPv6 invalid source address length");
733 			return false;
734 		}
735 		src = nla_get_in6_addr(attr);
736 		if (ipv6_addr_is_multicast(&src)) {
737 			NL_SET_ERR_MSG_MOD(extack, "IPv6 multicast source address is not allowed");
738 			return false;
739 		}
740 		break;
741 	}
742 #endif
743 	default:
744 		NL_SET_ERR_MSG_MOD(extack, "Invalid protocol used with source address");
745 		return false;
746 	}
747 
748 	return true;
749 }
750 
751 static const struct nla_policy br_mdbe_attrs_pol[MDBE_ATTR_MAX + 1] = {
752 	[MDBE_ATTR_SOURCE] = NLA_POLICY_RANGE(NLA_BINARY,
753 					      sizeof(struct in_addr),
754 					      sizeof(struct in6_addr)),
755 };
756 
757 static int br_mdb_parse(struct sk_buff *skb, struct nlmsghdr *nlh,
758 			struct net_device **pdev, struct br_mdb_entry **pentry,
759 			struct nlattr **mdb_attrs, struct netlink_ext_ack *extack)
760 {
761 	struct net *net = sock_net(skb->sk);
762 	struct br_mdb_entry *entry;
763 	struct br_port_msg *bpm;
764 	struct nlattr *tb[MDBA_SET_ENTRY_MAX+1];
765 	struct net_device *dev;
766 	int err;
767 
768 	err = nlmsg_parse_deprecated(nlh, sizeof(*bpm), tb,
769 				     MDBA_SET_ENTRY_MAX, NULL, NULL);
770 	if (err < 0)
771 		return err;
772 
773 	bpm = nlmsg_data(nlh);
774 	if (bpm->ifindex == 0) {
775 		NL_SET_ERR_MSG_MOD(extack, "Invalid bridge ifindex");
776 		return -EINVAL;
777 	}
778 
779 	dev = __dev_get_by_index(net, bpm->ifindex);
780 	if (dev == NULL) {
781 		NL_SET_ERR_MSG_MOD(extack, "Bridge device doesn't exist");
782 		return -ENODEV;
783 	}
784 
785 	if (!netif_is_bridge_master(dev)) {
786 		NL_SET_ERR_MSG_MOD(extack, "Device is not a bridge");
787 		return -EOPNOTSUPP;
788 	}
789 
790 	*pdev = dev;
791 
792 	if (!tb[MDBA_SET_ENTRY]) {
793 		NL_SET_ERR_MSG_MOD(extack, "Missing MDBA_SET_ENTRY attribute");
794 		return -EINVAL;
795 	}
796 	if (nla_len(tb[MDBA_SET_ENTRY]) != sizeof(struct br_mdb_entry)) {
797 		NL_SET_ERR_MSG_MOD(extack, "Invalid MDBA_SET_ENTRY attribute length");
798 		return -EINVAL;
799 	}
800 
801 	entry = nla_data(tb[MDBA_SET_ENTRY]);
802 	if (!is_valid_mdb_entry(entry, extack))
803 		return -EINVAL;
804 	*pentry = entry;
805 
806 	if (tb[MDBA_SET_ENTRY_ATTRS]) {
807 		err = nla_parse_nested(mdb_attrs, MDBE_ATTR_MAX,
808 				       tb[MDBA_SET_ENTRY_ATTRS],
809 				       br_mdbe_attrs_pol, extack);
810 		if (err)
811 			return err;
812 		if (mdb_attrs[MDBE_ATTR_SOURCE] &&
813 		    !is_valid_mdb_source(mdb_attrs[MDBE_ATTR_SOURCE],
814 					 entry->addr.proto, extack))
815 			return -EINVAL;
816 	} else {
817 		memset(mdb_attrs, 0,
818 		       sizeof(struct nlattr *) * (MDBE_ATTR_MAX + 1));
819 	}
820 
821 	return 0;
822 }
823 
824 static struct net_bridge_mcast *
825 __br_mdb_choose_context(struct net_bridge *br,
826 			const struct br_mdb_entry *entry,
827 			struct netlink_ext_ack *extack)
828 {
829 	struct net_bridge_mcast *brmctx = NULL;
830 	struct net_bridge_vlan *v;
831 
832 	if (!br_opt_get(br, BROPT_MCAST_VLAN_SNOOPING_ENABLED)) {
833 		brmctx = &br->multicast_ctx;
834 		goto out;
835 	}
836 
837 	if (!entry->vid) {
838 		NL_SET_ERR_MSG_MOD(extack, "Cannot add an entry without a vlan when vlan snooping is enabled");
839 		goto out;
840 	}
841 
842 	v = br_vlan_find(br_vlan_group(br), entry->vid);
843 	if (!v) {
844 		NL_SET_ERR_MSG_MOD(extack, "Vlan is not configured");
845 		goto out;
846 	}
847 	if (br_multicast_ctx_vlan_global_disabled(&v->br_mcast_ctx)) {
848 		NL_SET_ERR_MSG_MOD(extack, "Vlan's multicast processing is disabled");
849 		goto out;
850 	}
851 	brmctx = &v->br_mcast_ctx;
852 out:
853 	return brmctx;
854 }
855 
856 static int br_mdb_add_group(struct net_bridge *br, struct net_bridge_port *port,
857 			    struct br_mdb_entry *entry,
858 			    struct nlattr **mdb_attrs,
859 			    struct netlink_ext_ack *extack)
860 {
861 	struct net_bridge_mdb_entry *mp, *star_mp;
862 	struct net_bridge_port_group __rcu **pp;
863 	struct net_bridge_port_group *p;
864 	struct net_bridge_mcast *brmctx;
865 	struct br_ip group, star_group;
866 	unsigned long now = jiffies;
867 	unsigned char flags = 0;
868 	u8 filter_mode;
869 	int err;
870 
871 	__mdb_entry_to_br_ip(entry, &group, mdb_attrs);
872 
873 	brmctx = __br_mdb_choose_context(br, entry, extack);
874 	if (!brmctx)
875 		return -EINVAL;
876 
877 	/* host join errors which can happen before creating the group */
878 	if (!port && !br_group_is_l2(&group)) {
879 		/* don't allow any flags for host-joined IP groups */
880 		if (entry->state) {
881 			NL_SET_ERR_MSG_MOD(extack, "Flags are not allowed for host groups");
882 			return -EINVAL;
883 		}
884 		if (!br_multicast_is_star_g(&group)) {
885 			NL_SET_ERR_MSG_MOD(extack, "Groups with sources cannot be manually host joined");
886 			return -EINVAL;
887 		}
888 	}
889 
890 	if (br_group_is_l2(&group) && entry->state != MDB_PERMANENT) {
891 		NL_SET_ERR_MSG_MOD(extack, "Only permanent L2 entries allowed");
892 		return -EINVAL;
893 	}
894 
895 	mp = br_mdb_ip_get(br, &group);
896 	if (!mp) {
897 		mp = br_multicast_new_group(br, &group);
898 		err = PTR_ERR_OR_ZERO(mp);
899 		if (err)
900 			return err;
901 	}
902 
903 	/* host join */
904 	if (!port) {
905 		if (mp->host_joined) {
906 			NL_SET_ERR_MSG_MOD(extack, "Group is already joined by host");
907 			return -EEXIST;
908 		}
909 
910 		br_multicast_host_join(brmctx, mp, false);
911 		br_mdb_notify(br->dev, mp, NULL, RTM_NEWMDB);
912 
913 		return 0;
914 	}
915 
916 	for (pp = &mp->ports;
917 	     (p = mlock_dereference(*pp, br)) != NULL;
918 	     pp = &p->next) {
919 		if (p->key.port == port) {
920 			NL_SET_ERR_MSG_MOD(extack, "Group is already joined by port");
921 			return -EEXIST;
922 		}
923 		if ((unsigned long)p->key.port < (unsigned long)port)
924 			break;
925 	}
926 
927 	filter_mode = br_multicast_is_star_g(&group) ? MCAST_EXCLUDE :
928 						       MCAST_INCLUDE;
929 
930 	if (entry->state == MDB_PERMANENT)
931 		flags |= MDB_PG_FLAGS_PERMANENT;
932 
933 	p = br_multicast_new_port_group(port, &group, *pp, flags, NULL,
934 					filter_mode, RTPROT_STATIC);
935 	if (unlikely(!p)) {
936 		NL_SET_ERR_MSG_MOD(extack, "Couldn't allocate new port group");
937 		return -ENOMEM;
938 	}
939 	rcu_assign_pointer(*pp, p);
940 	if (entry->state == MDB_TEMPORARY)
941 		mod_timer(&p->timer,
942 			  now + brmctx->multicast_membership_interval);
943 	br_mdb_notify(br->dev, mp, p, RTM_NEWMDB);
944 	/* if we are adding a new EXCLUDE port group (*,G) it needs to be also
945 	 * added to all S,G entries for proper replication, if we are adding
946 	 * a new INCLUDE port (S,G) then all of *,G EXCLUDE ports need to be
947 	 * added to it for proper replication
948 	 */
949 	if (br_multicast_should_handle_mode(brmctx, group.proto)) {
950 		switch (filter_mode) {
951 		case MCAST_EXCLUDE:
952 			br_multicast_star_g_handle_mode(p, MCAST_EXCLUDE);
953 			break;
954 		case MCAST_INCLUDE:
955 			star_group = p->key.addr;
956 			memset(&star_group.src, 0, sizeof(star_group.src));
957 			star_mp = br_mdb_ip_get(br, &star_group);
958 			if (star_mp)
959 				br_multicast_sg_add_exclude_ports(star_mp, p);
960 			break;
961 		}
962 	}
963 
964 	return 0;
965 }
966 
967 static int __br_mdb_add(struct net *net, struct net_bridge *br,
968 			struct net_bridge_port *p,
969 			struct br_mdb_entry *entry,
970 			struct nlattr **mdb_attrs,
971 			struct netlink_ext_ack *extack)
972 {
973 	int ret;
974 
975 	spin_lock_bh(&br->multicast_lock);
976 	ret = br_mdb_add_group(br, p, entry, mdb_attrs, extack);
977 	spin_unlock_bh(&br->multicast_lock);
978 
979 	return ret;
980 }
981 
982 static int br_mdb_add(struct sk_buff *skb, struct nlmsghdr *nlh,
983 		      struct netlink_ext_ack *extack)
984 {
985 	struct nlattr *mdb_attrs[MDBE_ATTR_MAX + 1];
986 	struct net *net = sock_net(skb->sk);
987 	struct net_bridge_vlan_group *vg;
988 	struct net_bridge_port *p = NULL;
989 	struct net_device *dev, *pdev;
990 	struct br_mdb_entry *entry;
991 	struct net_bridge_vlan *v;
992 	struct net_bridge *br;
993 	int err;
994 
995 	err = br_mdb_parse(skb, nlh, &dev, &entry, mdb_attrs, extack);
996 	if (err < 0)
997 		return err;
998 
999 	br = netdev_priv(dev);
1000 
1001 	if (!netif_running(br->dev)) {
1002 		NL_SET_ERR_MSG_MOD(extack, "Bridge device is not running");
1003 		return -EINVAL;
1004 	}
1005 
1006 	if (!br_opt_get(br, BROPT_MULTICAST_ENABLED)) {
1007 		NL_SET_ERR_MSG_MOD(extack, "Bridge's multicast processing is disabled");
1008 		return -EINVAL;
1009 	}
1010 
1011 	if (entry->ifindex != br->dev->ifindex) {
1012 		pdev = __dev_get_by_index(net, entry->ifindex);
1013 		if (!pdev) {
1014 			NL_SET_ERR_MSG_MOD(extack, "Port net device doesn't exist");
1015 			return -ENODEV;
1016 		}
1017 
1018 		p = br_port_get_rtnl(pdev);
1019 		if (!p) {
1020 			NL_SET_ERR_MSG_MOD(extack, "Net device is not a bridge port");
1021 			return -EINVAL;
1022 		}
1023 
1024 		if (p->br != br) {
1025 			NL_SET_ERR_MSG_MOD(extack, "Port belongs to a different bridge device");
1026 			return -EINVAL;
1027 		}
1028 		if (p->state == BR_STATE_DISABLED) {
1029 			NL_SET_ERR_MSG_MOD(extack, "Port is in disabled state");
1030 			return -EINVAL;
1031 		}
1032 		vg = nbp_vlan_group(p);
1033 	} else {
1034 		vg = br_vlan_group(br);
1035 	}
1036 
1037 	/* If vlan filtering is enabled and VLAN is not specified
1038 	 * install mdb entry on all vlans configured on the port.
1039 	 */
1040 	if (br_vlan_enabled(br->dev) && vg && entry->vid == 0) {
1041 		list_for_each_entry(v, &vg->vlan_list, vlist) {
1042 			entry->vid = v->vid;
1043 			err = __br_mdb_add(net, br, p, entry, mdb_attrs, extack);
1044 			if (err)
1045 				break;
1046 		}
1047 	} else {
1048 		err = __br_mdb_add(net, br, p, entry, mdb_attrs, extack);
1049 	}
1050 
1051 	return err;
1052 }
1053 
1054 static int __br_mdb_del(struct net_bridge *br, struct br_mdb_entry *entry,
1055 			struct nlattr **mdb_attrs)
1056 {
1057 	struct net_bridge_mdb_entry *mp;
1058 	struct net_bridge_port_group *p;
1059 	struct net_bridge_port_group __rcu **pp;
1060 	struct br_ip ip;
1061 	int err = -EINVAL;
1062 
1063 	if (!netif_running(br->dev) || !br_opt_get(br, BROPT_MULTICAST_ENABLED))
1064 		return -EINVAL;
1065 
1066 	__mdb_entry_to_br_ip(entry, &ip, mdb_attrs);
1067 
1068 	spin_lock_bh(&br->multicast_lock);
1069 	mp = br_mdb_ip_get(br, &ip);
1070 	if (!mp)
1071 		goto unlock;
1072 
1073 	/* host leave */
1074 	if (entry->ifindex == mp->br->dev->ifindex && mp->host_joined) {
1075 		br_multicast_host_leave(mp, false);
1076 		err = 0;
1077 		br_mdb_notify(br->dev, mp, NULL, RTM_DELMDB);
1078 		if (!mp->ports && netif_running(br->dev))
1079 			mod_timer(&mp->timer, jiffies);
1080 		goto unlock;
1081 	}
1082 
1083 	for (pp = &mp->ports;
1084 	     (p = mlock_dereference(*pp, br)) != NULL;
1085 	     pp = &p->next) {
1086 		if (!p->key.port || p->key.port->dev->ifindex != entry->ifindex)
1087 			continue;
1088 
1089 		if (p->key.port->state == BR_STATE_DISABLED)
1090 			goto unlock;
1091 
1092 		br_multicast_del_pg(mp, p, pp);
1093 		err = 0;
1094 		break;
1095 	}
1096 
1097 unlock:
1098 	spin_unlock_bh(&br->multicast_lock);
1099 	return err;
1100 }
1101 
1102 static int br_mdb_del(struct sk_buff *skb, struct nlmsghdr *nlh,
1103 		      struct netlink_ext_ack *extack)
1104 {
1105 	struct nlattr *mdb_attrs[MDBE_ATTR_MAX + 1];
1106 	struct net *net = sock_net(skb->sk);
1107 	struct net_bridge_vlan_group *vg;
1108 	struct net_bridge_port *p = NULL;
1109 	struct net_device *dev, *pdev;
1110 	struct br_mdb_entry *entry;
1111 	struct net_bridge_vlan *v;
1112 	struct net_bridge *br;
1113 	int err;
1114 
1115 	err = br_mdb_parse(skb, nlh, &dev, &entry, mdb_attrs, extack);
1116 	if (err < 0)
1117 		return err;
1118 
1119 	br = netdev_priv(dev);
1120 
1121 	if (entry->ifindex != br->dev->ifindex) {
1122 		pdev = __dev_get_by_index(net, entry->ifindex);
1123 		if (!pdev)
1124 			return -ENODEV;
1125 
1126 		p = br_port_get_rtnl(pdev);
1127 		if (!p || p->br != br || p->state == BR_STATE_DISABLED)
1128 			return -EINVAL;
1129 		vg = nbp_vlan_group(p);
1130 	} else {
1131 		vg = br_vlan_group(br);
1132 	}
1133 
1134 	/* If vlan filtering is enabled and VLAN is not specified
1135 	 * delete mdb entry on all vlans configured on the port.
1136 	 */
1137 	if (br_vlan_enabled(br->dev) && vg && entry->vid == 0) {
1138 		list_for_each_entry(v, &vg->vlan_list, vlist) {
1139 			entry->vid = v->vid;
1140 			err = __br_mdb_del(br, entry, mdb_attrs);
1141 		}
1142 	} else {
1143 		err = __br_mdb_del(br, entry, mdb_attrs);
1144 	}
1145 
1146 	return err;
1147 }
1148 
1149 void br_mdb_init(void)
1150 {
1151 	rtnl_register_module(THIS_MODULE, PF_BRIDGE, RTM_GETMDB, NULL, br_mdb_dump, 0);
1152 	rtnl_register_module(THIS_MODULE, PF_BRIDGE, RTM_NEWMDB, br_mdb_add, NULL, 0);
1153 	rtnl_register_module(THIS_MODULE, PF_BRIDGE, RTM_DELMDB, br_mdb_del, NULL, 0);
1154 }
1155 
1156 void br_mdb_uninit(void)
1157 {
1158 	rtnl_unregister(PF_BRIDGE, RTM_GETMDB);
1159 	rtnl_unregister(PF_BRIDGE, RTM_NEWMDB);
1160 	rtnl_unregister(PF_BRIDGE, RTM_DELMDB);
1161 }
1162