xref: /openbmc/linux/net/bridge/br_multicast.c (revision e1e0a9e6)
1 /*
2  * Bridge multicast support.
3  *
4  * Copyright (c) 2010 Herbert Xu <herbert@gondor.apana.org.au>
5  *
6  * This program is free software; you can redistribute it and/or modify it
7  * under the terms of the GNU General Public License as published by the Free
8  * Software Foundation; either version 2 of the License, or (at your option)
9  * any later version.
10  *
11  */
12 
13 #include <linux/err.h>
14 #include <linux/if_ether.h>
15 #include <linux/igmp.h>
16 #include <linux/jhash.h>
17 #include <linux/kernel.h>
18 #include <linux/log2.h>
19 #include <linux/netdevice.h>
20 #include <linux/netfilter_bridge.h>
21 #include <linux/random.h>
22 #include <linux/rculist.h>
23 #include <linux/skbuff.h>
24 #include <linux/slab.h>
25 #include <linux/timer.h>
26 #include <net/ip.h>
27 #if IS_ENABLED(CONFIG_IPV6)
28 #include <net/ipv6.h>
29 #include <net/mld.h>
30 #include <net/ip6_checksum.h>
31 #endif
32 
33 #include "br_private.h"
34 
35 static void br_multicast_start_querier(struct net_bridge *br);
36 unsigned int br_mdb_rehash_seq;
37 
38 static inline int br_ip_equal(const struct br_ip *a, const struct br_ip *b)
39 {
40 	if (a->proto != b->proto)
41 		return 0;
42 	switch (a->proto) {
43 	case htons(ETH_P_IP):
44 		return a->u.ip4 == b->u.ip4;
45 #if IS_ENABLED(CONFIG_IPV6)
46 	case htons(ETH_P_IPV6):
47 		return ipv6_addr_equal(&a->u.ip6, &b->u.ip6);
48 #endif
49 	}
50 	return 0;
51 }
52 
53 static inline int __br_ip4_hash(struct net_bridge_mdb_htable *mdb, __be32 ip)
54 {
55 	return jhash_1word(mdb->secret, (__force u32)ip) & (mdb->max - 1);
56 }
57 
58 #if IS_ENABLED(CONFIG_IPV6)
59 static inline int __br_ip6_hash(struct net_bridge_mdb_htable *mdb,
60 				const struct in6_addr *ip)
61 {
62 	return jhash2((__force u32 *)ip->s6_addr32, 4, mdb->secret) & (mdb->max - 1);
63 }
64 #endif
65 
66 static inline int br_ip_hash(struct net_bridge_mdb_htable *mdb,
67 			     struct br_ip *ip)
68 {
69 	switch (ip->proto) {
70 	case htons(ETH_P_IP):
71 		return __br_ip4_hash(mdb, ip->u.ip4);
72 #if IS_ENABLED(CONFIG_IPV6)
73 	case htons(ETH_P_IPV6):
74 		return __br_ip6_hash(mdb, &ip->u.ip6);
75 #endif
76 	}
77 	return 0;
78 }
79 
80 static struct net_bridge_mdb_entry *__br_mdb_ip_get(
81 	struct net_bridge_mdb_htable *mdb, struct br_ip *dst, int hash)
82 {
83 	struct net_bridge_mdb_entry *mp;
84 	struct hlist_node *p;
85 
86 	hlist_for_each_entry_rcu(mp, p, &mdb->mhash[hash], hlist[mdb->ver]) {
87 		if (br_ip_equal(&mp->addr, dst))
88 			return mp;
89 	}
90 
91 	return NULL;
92 }
93 
94 struct net_bridge_mdb_entry *br_mdb_ip_get(struct net_bridge_mdb_htable *mdb,
95 					   struct br_ip *dst)
96 {
97 	if (!mdb)
98 		return NULL;
99 
100 	return __br_mdb_ip_get(mdb, dst, br_ip_hash(mdb, dst));
101 }
102 
103 static struct net_bridge_mdb_entry *br_mdb_ip4_get(
104 	struct net_bridge_mdb_htable *mdb, __be32 dst)
105 {
106 	struct br_ip br_dst;
107 
108 	br_dst.u.ip4 = dst;
109 	br_dst.proto = htons(ETH_P_IP);
110 
111 	return br_mdb_ip_get(mdb, &br_dst);
112 }
113 
114 #if IS_ENABLED(CONFIG_IPV6)
115 static struct net_bridge_mdb_entry *br_mdb_ip6_get(
116 	struct net_bridge_mdb_htable *mdb, const struct in6_addr *dst)
117 {
118 	struct br_ip br_dst;
119 
120 	br_dst.u.ip6 = *dst;
121 	br_dst.proto = htons(ETH_P_IPV6);
122 
123 	return br_mdb_ip_get(mdb, &br_dst);
124 }
125 #endif
126 
127 struct net_bridge_mdb_entry *br_mdb_get(struct net_bridge *br,
128 					struct sk_buff *skb)
129 {
130 	struct net_bridge_mdb_htable *mdb = rcu_dereference(br->mdb);
131 	struct br_ip ip;
132 
133 	if (br->multicast_disabled)
134 		return NULL;
135 
136 	if (BR_INPUT_SKB_CB(skb)->igmp)
137 		return NULL;
138 
139 	ip.proto = skb->protocol;
140 
141 	switch (skb->protocol) {
142 	case htons(ETH_P_IP):
143 		ip.u.ip4 = ip_hdr(skb)->daddr;
144 		break;
145 #if IS_ENABLED(CONFIG_IPV6)
146 	case htons(ETH_P_IPV6):
147 		ip.u.ip6 = ipv6_hdr(skb)->daddr;
148 		break;
149 #endif
150 	default:
151 		return NULL;
152 	}
153 
154 	return br_mdb_ip_get(mdb, &ip);
155 }
156 
157 static void br_mdb_free(struct rcu_head *head)
158 {
159 	struct net_bridge_mdb_htable *mdb =
160 		container_of(head, struct net_bridge_mdb_htable, rcu);
161 	struct net_bridge_mdb_htable *old = mdb->old;
162 
163 	mdb->old = NULL;
164 	kfree(old->mhash);
165 	kfree(old);
166 }
167 
168 static int br_mdb_copy(struct net_bridge_mdb_htable *new,
169 		       struct net_bridge_mdb_htable *old,
170 		       int elasticity)
171 {
172 	struct net_bridge_mdb_entry *mp;
173 	struct hlist_node *p;
174 	int maxlen;
175 	int len;
176 	int i;
177 
178 	for (i = 0; i < old->max; i++)
179 		hlist_for_each_entry(mp, p, &old->mhash[i], hlist[old->ver])
180 			hlist_add_head(&mp->hlist[new->ver],
181 				       &new->mhash[br_ip_hash(new, &mp->addr)]);
182 
183 	if (!elasticity)
184 		return 0;
185 
186 	maxlen = 0;
187 	for (i = 0; i < new->max; i++) {
188 		len = 0;
189 		hlist_for_each_entry(mp, p, &new->mhash[i], hlist[new->ver])
190 			len++;
191 		if (len > maxlen)
192 			maxlen = len;
193 	}
194 
195 	return maxlen > elasticity ? -EINVAL : 0;
196 }
197 
198 void br_multicast_free_pg(struct rcu_head *head)
199 {
200 	struct net_bridge_port_group *p =
201 		container_of(head, struct net_bridge_port_group, rcu);
202 
203 	kfree(p);
204 }
205 
206 static void br_multicast_free_group(struct rcu_head *head)
207 {
208 	struct net_bridge_mdb_entry *mp =
209 		container_of(head, struct net_bridge_mdb_entry, rcu);
210 
211 	kfree(mp);
212 }
213 
214 static void br_multicast_group_expired(unsigned long data)
215 {
216 	struct net_bridge_mdb_entry *mp = (void *)data;
217 	struct net_bridge *br = mp->br;
218 	struct net_bridge_mdb_htable *mdb;
219 
220 	spin_lock(&br->multicast_lock);
221 	if (!netif_running(br->dev) || timer_pending(&mp->timer))
222 		goto out;
223 
224 	mp->mglist = false;
225 
226 	if (mp->ports)
227 		goto out;
228 
229 	mdb = mlock_dereference(br->mdb, br);
230 
231 	hlist_del_rcu(&mp->hlist[mdb->ver]);
232 	mdb->size--;
233 
234 	call_rcu_bh(&mp->rcu, br_multicast_free_group);
235 
236 out:
237 	spin_unlock(&br->multicast_lock);
238 }
239 
240 static void br_multicast_del_pg(struct net_bridge *br,
241 				struct net_bridge_port_group *pg)
242 {
243 	struct net_bridge_mdb_htable *mdb;
244 	struct net_bridge_mdb_entry *mp;
245 	struct net_bridge_port_group *p;
246 	struct net_bridge_port_group __rcu **pp;
247 
248 	mdb = mlock_dereference(br->mdb, br);
249 
250 	mp = br_mdb_ip_get(mdb, &pg->addr);
251 	if (WARN_ON(!mp))
252 		return;
253 
254 	for (pp = &mp->ports;
255 	     (p = mlock_dereference(*pp, br)) != NULL;
256 	     pp = &p->next) {
257 		if (p != pg)
258 			continue;
259 
260 		rcu_assign_pointer(*pp, p->next);
261 		hlist_del_init(&p->mglist);
262 		del_timer(&p->timer);
263 		call_rcu_bh(&p->rcu, br_multicast_free_pg);
264 
265 		if (!mp->ports && !mp->mglist &&
266 		    netif_running(br->dev))
267 			mod_timer(&mp->timer, jiffies);
268 
269 		return;
270 	}
271 
272 	WARN_ON(1);
273 }
274 
275 static void br_multicast_port_group_expired(unsigned long data)
276 {
277 	struct net_bridge_port_group *pg = (void *)data;
278 	struct net_bridge *br = pg->port->br;
279 
280 	spin_lock(&br->multicast_lock);
281 	if (!netif_running(br->dev) || timer_pending(&pg->timer) ||
282 	    hlist_unhashed(&pg->mglist) || pg->state & MDB_PERMANENT)
283 		goto out;
284 
285 	br_multicast_del_pg(br, pg);
286 
287 out:
288 	spin_unlock(&br->multicast_lock);
289 }
290 
291 static int br_mdb_rehash(struct net_bridge_mdb_htable __rcu **mdbp, int max,
292 			 int elasticity)
293 {
294 	struct net_bridge_mdb_htable *old = rcu_dereference_protected(*mdbp, 1);
295 	struct net_bridge_mdb_htable *mdb;
296 	int err;
297 
298 	mdb = kmalloc(sizeof(*mdb), GFP_ATOMIC);
299 	if (!mdb)
300 		return -ENOMEM;
301 
302 	mdb->max = max;
303 	mdb->old = old;
304 
305 	mdb->mhash = kzalloc(max * sizeof(*mdb->mhash), GFP_ATOMIC);
306 	if (!mdb->mhash) {
307 		kfree(mdb);
308 		return -ENOMEM;
309 	}
310 
311 	mdb->size = old ? old->size : 0;
312 	mdb->ver = old ? old->ver ^ 1 : 0;
313 
314 	if (!old || elasticity)
315 		get_random_bytes(&mdb->secret, sizeof(mdb->secret));
316 	else
317 		mdb->secret = old->secret;
318 
319 	if (!old)
320 		goto out;
321 
322 	err = br_mdb_copy(mdb, old, elasticity);
323 	if (err) {
324 		kfree(mdb->mhash);
325 		kfree(mdb);
326 		return err;
327 	}
328 
329 	br_mdb_rehash_seq++;
330 	call_rcu_bh(&mdb->rcu, br_mdb_free);
331 
332 out:
333 	rcu_assign_pointer(*mdbp, mdb);
334 
335 	return 0;
336 }
337 
338 static struct sk_buff *br_ip4_multicast_alloc_query(struct net_bridge *br,
339 						    __be32 group)
340 {
341 	struct sk_buff *skb;
342 	struct igmphdr *ih;
343 	struct ethhdr *eth;
344 	struct iphdr *iph;
345 
346 	skb = netdev_alloc_skb_ip_align(br->dev, sizeof(*eth) + sizeof(*iph) +
347 						 sizeof(*ih) + 4);
348 	if (!skb)
349 		goto out;
350 
351 	skb->protocol = htons(ETH_P_IP);
352 
353 	skb_reset_mac_header(skb);
354 	eth = eth_hdr(skb);
355 
356 	memcpy(eth->h_source, br->dev->dev_addr, 6);
357 	eth->h_dest[0] = 1;
358 	eth->h_dest[1] = 0;
359 	eth->h_dest[2] = 0x5e;
360 	eth->h_dest[3] = 0;
361 	eth->h_dest[4] = 0;
362 	eth->h_dest[5] = 1;
363 	eth->h_proto = htons(ETH_P_IP);
364 	skb_put(skb, sizeof(*eth));
365 
366 	skb_set_network_header(skb, skb->len);
367 	iph = ip_hdr(skb);
368 
369 	iph->version = 4;
370 	iph->ihl = 6;
371 	iph->tos = 0xc0;
372 	iph->tot_len = htons(sizeof(*iph) + sizeof(*ih) + 4);
373 	iph->id = 0;
374 	iph->frag_off = htons(IP_DF);
375 	iph->ttl = 1;
376 	iph->protocol = IPPROTO_IGMP;
377 	iph->saddr = 0;
378 	iph->daddr = htonl(INADDR_ALLHOSTS_GROUP);
379 	((u8 *)&iph[1])[0] = IPOPT_RA;
380 	((u8 *)&iph[1])[1] = 4;
381 	((u8 *)&iph[1])[2] = 0;
382 	((u8 *)&iph[1])[3] = 0;
383 	ip_send_check(iph);
384 	skb_put(skb, 24);
385 
386 	skb_set_transport_header(skb, skb->len);
387 	ih = igmp_hdr(skb);
388 	ih->type = IGMP_HOST_MEMBERSHIP_QUERY;
389 	ih->code = (group ? br->multicast_last_member_interval :
390 			    br->multicast_query_response_interval) /
391 		   (HZ / IGMP_TIMER_SCALE);
392 	ih->group = group;
393 	ih->csum = 0;
394 	ih->csum = ip_compute_csum((void *)ih, sizeof(struct igmphdr));
395 	skb_put(skb, sizeof(*ih));
396 
397 	__skb_pull(skb, sizeof(*eth));
398 
399 out:
400 	return skb;
401 }
402 
403 #if IS_ENABLED(CONFIG_IPV6)
404 static struct sk_buff *br_ip6_multicast_alloc_query(struct net_bridge *br,
405 						    const struct in6_addr *group)
406 {
407 	struct sk_buff *skb;
408 	struct ipv6hdr *ip6h;
409 	struct mld_msg *mldq;
410 	struct ethhdr *eth;
411 	u8 *hopopt;
412 	unsigned long interval;
413 
414 	skb = netdev_alloc_skb_ip_align(br->dev, sizeof(*eth) + sizeof(*ip6h) +
415 						 8 + sizeof(*mldq));
416 	if (!skb)
417 		goto out;
418 
419 	skb->protocol = htons(ETH_P_IPV6);
420 
421 	/* Ethernet header */
422 	skb_reset_mac_header(skb);
423 	eth = eth_hdr(skb);
424 
425 	memcpy(eth->h_source, br->dev->dev_addr, 6);
426 	eth->h_proto = htons(ETH_P_IPV6);
427 	skb_put(skb, sizeof(*eth));
428 
429 	/* IPv6 header + HbH option */
430 	skb_set_network_header(skb, skb->len);
431 	ip6h = ipv6_hdr(skb);
432 
433 	*(__force __be32 *)ip6h = htonl(0x60000000);
434 	ip6h->payload_len = htons(8 + sizeof(*mldq));
435 	ip6h->nexthdr = IPPROTO_HOPOPTS;
436 	ip6h->hop_limit = 1;
437 	ipv6_addr_set(&ip6h->daddr, htonl(0xff020000), 0, 0, htonl(1));
438 	if (ipv6_dev_get_saddr(dev_net(br->dev), br->dev, &ip6h->daddr, 0,
439 			       &ip6h->saddr)) {
440 		kfree_skb(skb);
441 		return NULL;
442 	}
443 	ipv6_eth_mc_map(&ip6h->daddr, eth->h_dest);
444 
445 	hopopt = (u8 *)(ip6h + 1);
446 	hopopt[0] = IPPROTO_ICMPV6;		/* next hdr */
447 	hopopt[1] = 0;				/* length of HbH */
448 	hopopt[2] = IPV6_TLV_ROUTERALERT;	/* Router Alert */
449 	hopopt[3] = 2;				/* Length of RA Option */
450 	hopopt[4] = 0;				/* Type = 0x0000 (MLD) */
451 	hopopt[5] = 0;
452 	hopopt[6] = IPV6_TLV_PAD1;		/* Pad1 */
453 	hopopt[7] = IPV6_TLV_PAD1;		/* Pad1 */
454 
455 	skb_put(skb, sizeof(*ip6h) + 8);
456 
457 	/* ICMPv6 */
458 	skb_set_transport_header(skb, skb->len);
459 	mldq = (struct mld_msg *) icmp6_hdr(skb);
460 
461 	interval = ipv6_addr_any(group) ? br->multicast_last_member_interval :
462 					  br->multicast_query_response_interval;
463 
464 	mldq->mld_type = ICMPV6_MGM_QUERY;
465 	mldq->mld_code = 0;
466 	mldq->mld_cksum = 0;
467 	mldq->mld_maxdelay = htons((u16)jiffies_to_msecs(interval));
468 	mldq->mld_reserved = 0;
469 	mldq->mld_mca = *group;
470 
471 	/* checksum */
472 	mldq->mld_cksum = csum_ipv6_magic(&ip6h->saddr, &ip6h->daddr,
473 					  sizeof(*mldq), IPPROTO_ICMPV6,
474 					  csum_partial(mldq,
475 						       sizeof(*mldq), 0));
476 	skb_put(skb, sizeof(*mldq));
477 
478 	__skb_pull(skb, sizeof(*eth));
479 
480 out:
481 	return skb;
482 }
483 #endif
484 
485 static struct sk_buff *br_multicast_alloc_query(struct net_bridge *br,
486 						struct br_ip *addr)
487 {
488 	switch (addr->proto) {
489 	case htons(ETH_P_IP):
490 		return br_ip4_multicast_alloc_query(br, addr->u.ip4);
491 #if IS_ENABLED(CONFIG_IPV6)
492 	case htons(ETH_P_IPV6):
493 		return br_ip6_multicast_alloc_query(br, &addr->u.ip6);
494 #endif
495 	}
496 	return NULL;
497 }
498 
499 static struct net_bridge_mdb_entry *br_multicast_get_group(
500 	struct net_bridge *br, struct net_bridge_port *port,
501 	struct br_ip *group, int hash)
502 {
503 	struct net_bridge_mdb_htable *mdb;
504 	struct net_bridge_mdb_entry *mp;
505 	struct hlist_node *p;
506 	unsigned int count = 0;
507 	unsigned int max;
508 	int elasticity;
509 	int err;
510 
511 	mdb = rcu_dereference_protected(br->mdb, 1);
512 	hlist_for_each_entry(mp, p, &mdb->mhash[hash], hlist[mdb->ver]) {
513 		count++;
514 		if (unlikely(br_ip_equal(group, &mp->addr)))
515 			return mp;
516 	}
517 
518 	elasticity = 0;
519 	max = mdb->max;
520 
521 	if (unlikely(count > br->hash_elasticity && count)) {
522 		if (net_ratelimit())
523 			br_info(br, "Multicast hash table "
524 				"chain limit reached: %s\n",
525 				port ? port->dev->name : br->dev->name);
526 
527 		elasticity = br->hash_elasticity;
528 	}
529 
530 	if (mdb->size >= max) {
531 		max *= 2;
532 		if (unlikely(max > br->hash_max)) {
533 			br_warn(br, "Multicast hash table maximum of %d "
534 				"reached, disabling snooping: %s\n",
535 				br->hash_max,
536 				port ? port->dev->name : br->dev->name);
537 			err = -E2BIG;
538 disable:
539 			br->multicast_disabled = 1;
540 			goto err;
541 		}
542 	}
543 
544 	if (max > mdb->max || elasticity) {
545 		if (mdb->old) {
546 			if (net_ratelimit())
547 				br_info(br, "Multicast hash table "
548 					"on fire: %s\n",
549 					port ? port->dev->name : br->dev->name);
550 			err = -EEXIST;
551 			goto err;
552 		}
553 
554 		err = br_mdb_rehash(&br->mdb, max, elasticity);
555 		if (err) {
556 			br_warn(br, "Cannot rehash multicast "
557 				"hash table, disabling snooping: %s, %d, %d\n",
558 				port ? port->dev->name : br->dev->name,
559 				mdb->size, err);
560 			goto disable;
561 		}
562 
563 		err = -EAGAIN;
564 		goto err;
565 	}
566 
567 	return NULL;
568 
569 err:
570 	mp = ERR_PTR(err);
571 	return mp;
572 }
573 
574 struct net_bridge_mdb_entry *br_multicast_new_group(struct net_bridge *br,
575 	struct net_bridge_port *port, struct br_ip *group)
576 {
577 	struct net_bridge_mdb_htable *mdb;
578 	struct net_bridge_mdb_entry *mp;
579 	int hash;
580 	int err;
581 
582 	mdb = rcu_dereference_protected(br->mdb, 1);
583 	if (!mdb) {
584 		err = br_mdb_rehash(&br->mdb, BR_HASH_SIZE, 0);
585 		if (err)
586 			return ERR_PTR(err);
587 		goto rehash;
588 	}
589 
590 	hash = br_ip_hash(mdb, group);
591 	mp = br_multicast_get_group(br, port, group, hash);
592 	switch (PTR_ERR(mp)) {
593 	case 0:
594 		break;
595 
596 	case -EAGAIN:
597 rehash:
598 		mdb = rcu_dereference_protected(br->mdb, 1);
599 		hash = br_ip_hash(mdb, group);
600 		break;
601 
602 	default:
603 		goto out;
604 	}
605 
606 	mp = kzalloc(sizeof(*mp), GFP_ATOMIC);
607 	if (unlikely(!mp))
608 		return ERR_PTR(-ENOMEM);
609 
610 	mp->br = br;
611 	mp->addr = *group;
612 	setup_timer(&mp->timer, br_multicast_group_expired,
613 		    (unsigned long)mp);
614 
615 	hlist_add_head_rcu(&mp->hlist[mdb->ver], &mdb->mhash[hash]);
616 	mdb->size++;
617 
618 out:
619 	return mp;
620 }
621 
622 struct net_bridge_port_group *br_multicast_new_port_group(
623 			struct net_bridge_port *port,
624 			struct br_ip *group,
625 			struct net_bridge_port_group __rcu *next,
626 			unsigned char state)
627 {
628 	struct net_bridge_port_group *p;
629 
630 	p = kzalloc(sizeof(*p), GFP_ATOMIC);
631 	if (unlikely(!p))
632 		return NULL;
633 
634 	p->addr = *group;
635 	p->port = port;
636 	p->state = state;
637 	rcu_assign_pointer(p->next, next);
638 	hlist_add_head(&p->mglist, &port->mglist);
639 	setup_timer(&p->timer, br_multicast_port_group_expired,
640 		    (unsigned long)p);
641 	return p;
642 }
643 
644 static int br_multicast_add_group(struct net_bridge *br,
645 				  struct net_bridge_port *port,
646 				  struct br_ip *group)
647 {
648 	struct net_bridge_mdb_entry *mp;
649 	struct net_bridge_port_group *p;
650 	struct net_bridge_port_group __rcu **pp;
651 	unsigned long now = jiffies;
652 	int err;
653 
654 	spin_lock(&br->multicast_lock);
655 	if (!netif_running(br->dev) ||
656 	    (port && port->state == BR_STATE_DISABLED))
657 		goto out;
658 
659 	mp = br_multicast_new_group(br, port, group);
660 	err = PTR_ERR(mp);
661 	if (IS_ERR(mp))
662 		goto err;
663 
664 	if (!port) {
665 		mp->mglist = true;
666 		mod_timer(&mp->timer, now + br->multicast_membership_interval);
667 		goto out;
668 	}
669 
670 	for (pp = &mp->ports;
671 	     (p = mlock_dereference(*pp, br)) != NULL;
672 	     pp = &p->next) {
673 		if (p->port == port)
674 			goto found;
675 		if ((unsigned long)p->port < (unsigned long)port)
676 			break;
677 	}
678 
679 	p = br_multicast_new_port_group(port, group, *pp, MDB_TEMPORARY);
680 	if (unlikely(!p))
681 		goto err;
682 	rcu_assign_pointer(*pp, p);
683 	br_mdb_notify(br->dev, port, group, RTM_NEWMDB);
684 
685 found:
686 	mod_timer(&p->timer, now + br->multicast_membership_interval);
687 out:
688 	err = 0;
689 
690 err:
691 	spin_unlock(&br->multicast_lock);
692 	return err;
693 }
694 
695 static int br_ip4_multicast_add_group(struct net_bridge *br,
696 				      struct net_bridge_port *port,
697 				      __be32 group)
698 {
699 	struct br_ip br_group;
700 
701 	if (ipv4_is_local_multicast(group))
702 		return 0;
703 
704 	br_group.u.ip4 = group;
705 	br_group.proto = htons(ETH_P_IP);
706 
707 	return br_multicast_add_group(br, port, &br_group);
708 }
709 
710 #if IS_ENABLED(CONFIG_IPV6)
711 static int br_ip6_multicast_add_group(struct net_bridge *br,
712 				      struct net_bridge_port *port,
713 				      const struct in6_addr *group)
714 {
715 	struct br_ip br_group;
716 
717 	if (!ipv6_is_transient_multicast(group))
718 		return 0;
719 
720 	br_group.u.ip6 = *group;
721 	br_group.proto = htons(ETH_P_IPV6);
722 
723 	return br_multicast_add_group(br, port, &br_group);
724 }
725 #endif
726 
727 static void br_multicast_router_expired(unsigned long data)
728 {
729 	struct net_bridge_port *port = (void *)data;
730 	struct net_bridge *br = port->br;
731 
732 	spin_lock(&br->multicast_lock);
733 	if (port->multicast_router != 1 ||
734 	    timer_pending(&port->multicast_router_timer) ||
735 	    hlist_unhashed(&port->rlist))
736 		goto out;
737 
738 	hlist_del_init_rcu(&port->rlist);
739 
740 out:
741 	spin_unlock(&br->multicast_lock);
742 }
743 
744 static void br_multicast_local_router_expired(unsigned long data)
745 {
746 }
747 
748 static void br_multicast_querier_expired(unsigned long data)
749 {
750 	struct net_bridge *br = (void *)data;
751 
752 	spin_lock(&br->multicast_lock);
753 	if (!netif_running(br->dev) || br->multicast_disabled)
754 		goto out;
755 
756 	br_multicast_start_querier(br);
757 
758 out:
759 	spin_unlock(&br->multicast_lock);
760 }
761 
762 static void __br_multicast_send_query(struct net_bridge *br,
763 				      struct net_bridge_port *port,
764 				      struct br_ip *ip)
765 {
766 	struct sk_buff *skb;
767 
768 	skb = br_multicast_alloc_query(br, ip);
769 	if (!skb)
770 		return;
771 
772 	if (port) {
773 		__skb_push(skb, sizeof(struct ethhdr));
774 		skb->dev = port->dev;
775 		NF_HOOK(NFPROTO_BRIDGE, NF_BR_LOCAL_OUT, skb, NULL, skb->dev,
776 			dev_queue_xmit);
777 	} else
778 		netif_rx(skb);
779 }
780 
781 static void br_multicast_send_query(struct net_bridge *br,
782 				    struct net_bridge_port *port, u32 sent)
783 {
784 	unsigned long time;
785 	struct br_ip br_group;
786 
787 	if (!netif_running(br->dev) || br->multicast_disabled ||
788 	    !br->multicast_querier ||
789 	    timer_pending(&br->multicast_querier_timer))
790 		return;
791 
792 	memset(&br_group.u, 0, sizeof(br_group.u));
793 
794 	br_group.proto = htons(ETH_P_IP);
795 	__br_multicast_send_query(br, port, &br_group);
796 
797 #if IS_ENABLED(CONFIG_IPV6)
798 	br_group.proto = htons(ETH_P_IPV6);
799 	__br_multicast_send_query(br, port, &br_group);
800 #endif
801 
802 	time = jiffies;
803 	time += sent < br->multicast_startup_query_count ?
804 		br->multicast_startup_query_interval :
805 		br->multicast_query_interval;
806 	mod_timer(port ? &port->multicast_query_timer :
807 			 &br->multicast_query_timer, time);
808 }
809 
810 static void br_multicast_port_query_expired(unsigned long data)
811 {
812 	struct net_bridge_port *port = (void *)data;
813 	struct net_bridge *br = port->br;
814 
815 	spin_lock(&br->multicast_lock);
816 	if (port->state == BR_STATE_DISABLED ||
817 	    port->state == BR_STATE_BLOCKING)
818 		goto out;
819 
820 	if (port->multicast_startup_queries_sent <
821 	    br->multicast_startup_query_count)
822 		port->multicast_startup_queries_sent++;
823 
824 	br_multicast_send_query(port->br, port,
825 				port->multicast_startup_queries_sent);
826 
827 out:
828 	spin_unlock(&br->multicast_lock);
829 }
830 
831 void br_multicast_add_port(struct net_bridge_port *port)
832 {
833 	port->multicast_router = 1;
834 
835 	setup_timer(&port->multicast_router_timer, br_multicast_router_expired,
836 		    (unsigned long)port);
837 	setup_timer(&port->multicast_query_timer,
838 		    br_multicast_port_query_expired, (unsigned long)port);
839 }
840 
841 void br_multicast_del_port(struct net_bridge_port *port)
842 {
843 	del_timer_sync(&port->multicast_router_timer);
844 }
845 
846 static void __br_multicast_enable_port(struct net_bridge_port *port)
847 {
848 	port->multicast_startup_queries_sent = 0;
849 
850 	if (try_to_del_timer_sync(&port->multicast_query_timer) >= 0 ||
851 	    del_timer(&port->multicast_query_timer))
852 		mod_timer(&port->multicast_query_timer, jiffies);
853 }
854 
855 void br_multicast_enable_port(struct net_bridge_port *port)
856 {
857 	struct net_bridge *br = port->br;
858 
859 	spin_lock(&br->multicast_lock);
860 	if (br->multicast_disabled || !netif_running(br->dev))
861 		goto out;
862 
863 	__br_multicast_enable_port(port);
864 
865 out:
866 	spin_unlock(&br->multicast_lock);
867 }
868 
869 void br_multicast_disable_port(struct net_bridge_port *port)
870 {
871 	struct net_bridge *br = port->br;
872 	struct net_bridge_port_group *pg;
873 	struct hlist_node *p, *n;
874 
875 	spin_lock(&br->multicast_lock);
876 	hlist_for_each_entry_safe(pg, p, n, &port->mglist, mglist)
877 		br_multicast_del_pg(br, pg);
878 
879 	if (!hlist_unhashed(&port->rlist))
880 		hlist_del_init_rcu(&port->rlist);
881 	del_timer(&port->multicast_router_timer);
882 	del_timer(&port->multicast_query_timer);
883 	spin_unlock(&br->multicast_lock);
884 }
885 
886 static int br_ip4_multicast_igmp3_report(struct net_bridge *br,
887 					 struct net_bridge_port *port,
888 					 struct sk_buff *skb)
889 {
890 	struct igmpv3_report *ih;
891 	struct igmpv3_grec *grec;
892 	int i;
893 	int len;
894 	int num;
895 	int type;
896 	int err = 0;
897 	__be32 group;
898 
899 	if (!pskb_may_pull(skb, sizeof(*ih)))
900 		return -EINVAL;
901 
902 	ih = igmpv3_report_hdr(skb);
903 	num = ntohs(ih->ngrec);
904 	len = sizeof(*ih);
905 
906 	for (i = 0; i < num; i++) {
907 		len += sizeof(*grec);
908 		if (!pskb_may_pull(skb, len))
909 			return -EINVAL;
910 
911 		grec = (void *)(skb->data + len - sizeof(*grec));
912 		group = grec->grec_mca;
913 		type = grec->grec_type;
914 
915 		len += ntohs(grec->grec_nsrcs) * 4;
916 		if (!pskb_may_pull(skb, len))
917 			return -EINVAL;
918 
919 		/* We treat this as an IGMPv2 report for now. */
920 		switch (type) {
921 		case IGMPV3_MODE_IS_INCLUDE:
922 		case IGMPV3_MODE_IS_EXCLUDE:
923 		case IGMPV3_CHANGE_TO_INCLUDE:
924 		case IGMPV3_CHANGE_TO_EXCLUDE:
925 		case IGMPV3_ALLOW_NEW_SOURCES:
926 		case IGMPV3_BLOCK_OLD_SOURCES:
927 			break;
928 
929 		default:
930 			continue;
931 		}
932 
933 		err = br_ip4_multicast_add_group(br, port, group);
934 		if (err)
935 			break;
936 	}
937 
938 	return err;
939 }
940 
941 #if IS_ENABLED(CONFIG_IPV6)
942 static int br_ip6_multicast_mld2_report(struct net_bridge *br,
943 					struct net_bridge_port *port,
944 					struct sk_buff *skb)
945 {
946 	struct icmp6hdr *icmp6h;
947 	struct mld2_grec *grec;
948 	int i;
949 	int len;
950 	int num;
951 	int err = 0;
952 
953 	if (!pskb_may_pull(skb, sizeof(*icmp6h)))
954 		return -EINVAL;
955 
956 	icmp6h = icmp6_hdr(skb);
957 	num = ntohs(icmp6h->icmp6_dataun.un_data16[1]);
958 	len = sizeof(*icmp6h);
959 
960 	for (i = 0; i < num; i++) {
961 		__be16 *nsrcs, _nsrcs;
962 
963 		nsrcs = skb_header_pointer(skb,
964 					   len + offsetof(struct mld2_grec,
965 							  grec_nsrcs),
966 					   sizeof(_nsrcs), &_nsrcs);
967 		if (!nsrcs)
968 			return -EINVAL;
969 
970 		if (!pskb_may_pull(skb,
971 				   len + sizeof(*grec) +
972 				   sizeof(struct in6_addr) * ntohs(*nsrcs)))
973 			return -EINVAL;
974 
975 		grec = (struct mld2_grec *)(skb->data + len);
976 		len += sizeof(*grec) +
977 		       sizeof(struct in6_addr) * ntohs(*nsrcs);
978 
979 		/* We treat these as MLDv1 reports for now. */
980 		switch (grec->grec_type) {
981 		case MLD2_MODE_IS_INCLUDE:
982 		case MLD2_MODE_IS_EXCLUDE:
983 		case MLD2_CHANGE_TO_INCLUDE:
984 		case MLD2_CHANGE_TO_EXCLUDE:
985 		case MLD2_ALLOW_NEW_SOURCES:
986 		case MLD2_BLOCK_OLD_SOURCES:
987 			break;
988 
989 		default:
990 			continue;
991 		}
992 
993 		err = br_ip6_multicast_add_group(br, port, &grec->grec_mca);
994 		if (!err)
995 			break;
996 	}
997 
998 	return err;
999 }
1000 #endif
1001 
1002 /*
1003  * Add port to rotuer_list
1004  *  list is maintained ordered by pointer value
1005  *  and locked by br->multicast_lock and RCU
1006  */
1007 static void br_multicast_add_router(struct net_bridge *br,
1008 				    struct net_bridge_port *port)
1009 {
1010 	struct net_bridge_port *p;
1011 	struct hlist_node *n, *slot = NULL;
1012 
1013 	hlist_for_each_entry(p, n, &br->router_list, rlist) {
1014 		if ((unsigned long) port >= (unsigned long) p)
1015 			break;
1016 		slot = n;
1017 	}
1018 
1019 	if (slot)
1020 		hlist_add_after_rcu(slot, &port->rlist);
1021 	else
1022 		hlist_add_head_rcu(&port->rlist, &br->router_list);
1023 }
1024 
1025 static void br_multicast_mark_router(struct net_bridge *br,
1026 				     struct net_bridge_port *port)
1027 {
1028 	unsigned long now = jiffies;
1029 
1030 	if (!port) {
1031 		if (br->multicast_router == 1)
1032 			mod_timer(&br->multicast_router_timer,
1033 				  now + br->multicast_querier_interval);
1034 		return;
1035 	}
1036 
1037 	if (port->multicast_router != 1)
1038 		return;
1039 
1040 	if (!hlist_unhashed(&port->rlist))
1041 		goto timer;
1042 
1043 	br_multicast_add_router(br, port);
1044 
1045 timer:
1046 	mod_timer(&port->multicast_router_timer,
1047 		  now + br->multicast_querier_interval);
1048 }
1049 
1050 static void br_multicast_query_received(struct net_bridge *br,
1051 					struct net_bridge_port *port,
1052 					int saddr)
1053 {
1054 	if (saddr)
1055 		mod_timer(&br->multicast_querier_timer,
1056 			  jiffies + br->multicast_querier_interval);
1057 	else if (timer_pending(&br->multicast_querier_timer))
1058 		return;
1059 
1060 	br_multicast_mark_router(br, port);
1061 }
1062 
1063 static int br_ip4_multicast_query(struct net_bridge *br,
1064 				  struct net_bridge_port *port,
1065 				  struct sk_buff *skb)
1066 {
1067 	const struct iphdr *iph = ip_hdr(skb);
1068 	struct igmphdr *ih = igmp_hdr(skb);
1069 	struct net_bridge_mdb_entry *mp;
1070 	struct igmpv3_query *ih3;
1071 	struct net_bridge_port_group *p;
1072 	struct net_bridge_port_group __rcu **pp;
1073 	unsigned long max_delay;
1074 	unsigned long now = jiffies;
1075 	__be32 group;
1076 	int err = 0;
1077 
1078 	spin_lock(&br->multicast_lock);
1079 	if (!netif_running(br->dev) ||
1080 	    (port && port->state == BR_STATE_DISABLED))
1081 		goto out;
1082 
1083 	br_multicast_query_received(br, port, !!iph->saddr);
1084 
1085 	group = ih->group;
1086 
1087 	if (skb->len == sizeof(*ih)) {
1088 		max_delay = ih->code * (HZ / IGMP_TIMER_SCALE);
1089 
1090 		if (!max_delay) {
1091 			max_delay = 10 * HZ;
1092 			group = 0;
1093 		}
1094 	} else {
1095 		if (!pskb_may_pull(skb, sizeof(struct igmpv3_query))) {
1096 			err = -EINVAL;
1097 			goto out;
1098 		}
1099 
1100 		ih3 = igmpv3_query_hdr(skb);
1101 		if (ih3->nsrcs)
1102 			goto out;
1103 
1104 		max_delay = ih3->code ?
1105 			    IGMPV3_MRC(ih3->code) * (HZ / IGMP_TIMER_SCALE) : 1;
1106 	}
1107 
1108 	if (!group)
1109 		goto out;
1110 
1111 	mp = br_mdb_ip4_get(mlock_dereference(br->mdb, br), group);
1112 	if (!mp)
1113 		goto out;
1114 
1115 	max_delay *= br->multicast_last_member_count;
1116 
1117 	if (mp->mglist &&
1118 	    (timer_pending(&mp->timer) ?
1119 	     time_after(mp->timer.expires, now + max_delay) :
1120 	     try_to_del_timer_sync(&mp->timer) >= 0))
1121 		mod_timer(&mp->timer, now + max_delay);
1122 
1123 	for (pp = &mp->ports;
1124 	     (p = mlock_dereference(*pp, br)) != NULL;
1125 	     pp = &p->next) {
1126 		if (timer_pending(&p->timer) ?
1127 		    time_after(p->timer.expires, now + max_delay) :
1128 		    try_to_del_timer_sync(&p->timer) >= 0)
1129 			mod_timer(&p->timer, now + max_delay);
1130 	}
1131 
1132 out:
1133 	spin_unlock(&br->multicast_lock);
1134 	return err;
1135 }
1136 
1137 #if IS_ENABLED(CONFIG_IPV6)
1138 static int br_ip6_multicast_query(struct net_bridge *br,
1139 				  struct net_bridge_port *port,
1140 				  struct sk_buff *skb)
1141 {
1142 	const struct ipv6hdr *ip6h = ipv6_hdr(skb);
1143 	struct mld_msg *mld;
1144 	struct net_bridge_mdb_entry *mp;
1145 	struct mld2_query *mld2q;
1146 	struct net_bridge_port_group *p;
1147 	struct net_bridge_port_group __rcu **pp;
1148 	unsigned long max_delay;
1149 	unsigned long now = jiffies;
1150 	const struct in6_addr *group = NULL;
1151 	int err = 0;
1152 
1153 	spin_lock(&br->multicast_lock);
1154 	if (!netif_running(br->dev) ||
1155 	    (port && port->state == BR_STATE_DISABLED))
1156 		goto out;
1157 
1158 	br_multicast_query_received(br, port, !ipv6_addr_any(&ip6h->saddr));
1159 
1160 	if (skb->len == sizeof(*mld)) {
1161 		if (!pskb_may_pull(skb, sizeof(*mld))) {
1162 			err = -EINVAL;
1163 			goto out;
1164 		}
1165 		mld = (struct mld_msg *) icmp6_hdr(skb);
1166 		max_delay = msecs_to_jiffies(ntohs(mld->mld_maxdelay));
1167 		if (max_delay)
1168 			group = &mld->mld_mca;
1169 	} else if (skb->len >= sizeof(*mld2q)) {
1170 		if (!pskb_may_pull(skb, sizeof(*mld2q))) {
1171 			err = -EINVAL;
1172 			goto out;
1173 		}
1174 		mld2q = (struct mld2_query *)icmp6_hdr(skb);
1175 		if (!mld2q->mld2q_nsrcs)
1176 			group = &mld2q->mld2q_mca;
1177 		max_delay = mld2q->mld2q_mrc ? MLDV2_MRC(ntohs(mld2q->mld2q_mrc)) : 1;
1178 	}
1179 
1180 	if (!group)
1181 		goto out;
1182 
1183 	mp = br_mdb_ip6_get(mlock_dereference(br->mdb, br), group);
1184 	if (!mp)
1185 		goto out;
1186 
1187 	max_delay *= br->multicast_last_member_count;
1188 	if (mp->mglist &&
1189 	    (timer_pending(&mp->timer) ?
1190 	     time_after(mp->timer.expires, now + max_delay) :
1191 	     try_to_del_timer_sync(&mp->timer) >= 0))
1192 		mod_timer(&mp->timer, now + max_delay);
1193 
1194 	for (pp = &mp->ports;
1195 	     (p = mlock_dereference(*pp, br)) != NULL;
1196 	     pp = &p->next) {
1197 		if (timer_pending(&p->timer) ?
1198 		    time_after(p->timer.expires, now + max_delay) :
1199 		    try_to_del_timer_sync(&p->timer) >= 0)
1200 			mod_timer(&p->timer, now + max_delay);
1201 	}
1202 
1203 out:
1204 	spin_unlock(&br->multicast_lock);
1205 	return err;
1206 }
1207 #endif
1208 
1209 static void br_multicast_leave_group(struct net_bridge *br,
1210 				     struct net_bridge_port *port,
1211 				     struct br_ip *group)
1212 {
1213 	struct net_bridge_mdb_htable *mdb;
1214 	struct net_bridge_mdb_entry *mp;
1215 	struct net_bridge_port_group *p;
1216 	unsigned long now;
1217 	unsigned long time;
1218 
1219 	spin_lock(&br->multicast_lock);
1220 	if (!netif_running(br->dev) ||
1221 	    (port && port->state == BR_STATE_DISABLED) ||
1222 	    timer_pending(&br->multicast_querier_timer))
1223 		goto out;
1224 
1225 	mdb = mlock_dereference(br->mdb, br);
1226 	mp = br_mdb_ip_get(mdb, group);
1227 	if (!mp)
1228 		goto out;
1229 
1230 	if (port && (port->flags & BR_MULTICAST_FAST_LEAVE)) {
1231 		struct net_bridge_port_group __rcu **pp;
1232 
1233 		for (pp = &mp->ports;
1234 		     (p = mlock_dereference(*pp, br)) != NULL;
1235 		     pp = &p->next) {
1236 			if (p->port != port)
1237 				continue;
1238 
1239 			rcu_assign_pointer(*pp, p->next);
1240 			hlist_del_init(&p->mglist);
1241 			del_timer(&p->timer);
1242 			call_rcu_bh(&p->rcu, br_multicast_free_pg);
1243 			br_mdb_notify(br->dev, port, group, RTM_DELMDB);
1244 
1245 			if (!mp->ports && !mp->mglist &&
1246 			    netif_running(br->dev))
1247 				mod_timer(&mp->timer, jiffies);
1248 		}
1249 		goto out;
1250 	}
1251 
1252 	now = jiffies;
1253 	time = now + br->multicast_last_member_count *
1254 		     br->multicast_last_member_interval;
1255 
1256 	if (!port) {
1257 		if (mp->mglist &&
1258 		    (timer_pending(&mp->timer) ?
1259 		     time_after(mp->timer.expires, time) :
1260 		     try_to_del_timer_sync(&mp->timer) >= 0)) {
1261 			mod_timer(&mp->timer, time);
1262 		}
1263 
1264 		goto out;
1265 	}
1266 
1267 	for (p = mlock_dereference(mp->ports, br);
1268 	     p != NULL;
1269 	     p = mlock_dereference(p->next, br)) {
1270 		if (p->port != port)
1271 			continue;
1272 
1273 		if (!hlist_unhashed(&p->mglist) &&
1274 		    (timer_pending(&p->timer) ?
1275 		     time_after(p->timer.expires, time) :
1276 		     try_to_del_timer_sync(&p->timer) >= 0)) {
1277 			mod_timer(&p->timer, time);
1278 		}
1279 
1280 		break;
1281 	}
1282 
1283 out:
1284 	spin_unlock(&br->multicast_lock);
1285 }
1286 
1287 static void br_ip4_multicast_leave_group(struct net_bridge *br,
1288 					 struct net_bridge_port *port,
1289 					 __be32 group)
1290 {
1291 	struct br_ip br_group;
1292 
1293 	if (ipv4_is_local_multicast(group))
1294 		return;
1295 
1296 	br_group.u.ip4 = group;
1297 	br_group.proto = htons(ETH_P_IP);
1298 
1299 	br_multicast_leave_group(br, port, &br_group);
1300 }
1301 
1302 #if IS_ENABLED(CONFIG_IPV6)
1303 static void br_ip6_multicast_leave_group(struct net_bridge *br,
1304 					 struct net_bridge_port *port,
1305 					 const struct in6_addr *group)
1306 {
1307 	struct br_ip br_group;
1308 
1309 	if (!ipv6_is_transient_multicast(group))
1310 		return;
1311 
1312 	br_group.u.ip6 = *group;
1313 	br_group.proto = htons(ETH_P_IPV6);
1314 
1315 	br_multicast_leave_group(br, port, &br_group);
1316 }
1317 #endif
1318 
1319 static int br_multicast_ipv4_rcv(struct net_bridge *br,
1320 				 struct net_bridge_port *port,
1321 				 struct sk_buff *skb)
1322 {
1323 	struct sk_buff *skb2 = skb;
1324 	const struct iphdr *iph;
1325 	struct igmphdr *ih;
1326 	unsigned int len;
1327 	unsigned int offset;
1328 	int err;
1329 
1330 	/* We treat OOM as packet loss for now. */
1331 	if (!pskb_may_pull(skb, sizeof(*iph)))
1332 		return -EINVAL;
1333 
1334 	iph = ip_hdr(skb);
1335 
1336 	if (iph->ihl < 5 || iph->version != 4)
1337 		return -EINVAL;
1338 
1339 	if (!pskb_may_pull(skb, ip_hdrlen(skb)))
1340 		return -EINVAL;
1341 
1342 	iph = ip_hdr(skb);
1343 
1344 	if (unlikely(ip_fast_csum((u8 *)iph, iph->ihl)))
1345 		return -EINVAL;
1346 
1347 	if (iph->protocol != IPPROTO_IGMP) {
1348 		if ((iph->daddr & IGMP_LOCAL_GROUP_MASK) != IGMP_LOCAL_GROUP)
1349 			BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1350 		return 0;
1351 	}
1352 
1353 	len = ntohs(iph->tot_len);
1354 	if (skb->len < len || len < ip_hdrlen(skb))
1355 		return -EINVAL;
1356 
1357 	if (skb->len > len) {
1358 		skb2 = skb_clone(skb, GFP_ATOMIC);
1359 		if (!skb2)
1360 			return -ENOMEM;
1361 
1362 		err = pskb_trim_rcsum(skb2, len);
1363 		if (err)
1364 			goto err_out;
1365 	}
1366 
1367 	len -= ip_hdrlen(skb2);
1368 	offset = skb_network_offset(skb2) + ip_hdrlen(skb2);
1369 	__skb_pull(skb2, offset);
1370 	skb_reset_transport_header(skb2);
1371 
1372 	err = -EINVAL;
1373 	if (!pskb_may_pull(skb2, sizeof(*ih)))
1374 		goto out;
1375 
1376 	switch (skb2->ip_summed) {
1377 	case CHECKSUM_COMPLETE:
1378 		if (!csum_fold(skb2->csum))
1379 			break;
1380 		/* fall through */
1381 	case CHECKSUM_NONE:
1382 		skb2->csum = 0;
1383 		if (skb_checksum_complete(skb2))
1384 			goto out;
1385 	}
1386 
1387 	err = 0;
1388 
1389 	BR_INPUT_SKB_CB(skb)->igmp = 1;
1390 	ih = igmp_hdr(skb2);
1391 
1392 	switch (ih->type) {
1393 	case IGMP_HOST_MEMBERSHIP_REPORT:
1394 	case IGMPV2_HOST_MEMBERSHIP_REPORT:
1395 		BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1396 		err = br_ip4_multicast_add_group(br, port, ih->group);
1397 		break;
1398 	case IGMPV3_HOST_MEMBERSHIP_REPORT:
1399 		err = br_ip4_multicast_igmp3_report(br, port, skb2);
1400 		break;
1401 	case IGMP_HOST_MEMBERSHIP_QUERY:
1402 		err = br_ip4_multicast_query(br, port, skb2);
1403 		break;
1404 	case IGMP_HOST_LEAVE_MESSAGE:
1405 		br_ip4_multicast_leave_group(br, port, ih->group);
1406 		break;
1407 	}
1408 
1409 out:
1410 	__skb_push(skb2, offset);
1411 err_out:
1412 	if (skb2 != skb)
1413 		kfree_skb(skb2);
1414 	return err;
1415 }
1416 
1417 #if IS_ENABLED(CONFIG_IPV6)
1418 static int br_multicast_ipv6_rcv(struct net_bridge *br,
1419 				 struct net_bridge_port *port,
1420 				 struct sk_buff *skb)
1421 {
1422 	struct sk_buff *skb2;
1423 	const struct ipv6hdr *ip6h;
1424 	u8 icmp6_type;
1425 	u8 nexthdr;
1426 	__be16 frag_off;
1427 	unsigned int len;
1428 	int offset;
1429 	int err;
1430 
1431 	if (!pskb_may_pull(skb, sizeof(*ip6h)))
1432 		return -EINVAL;
1433 
1434 	ip6h = ipv6_hdr(skb);
1435 
1436 	/*
1437 	 * We're interested in MLD messages only.
1438 	 *  - Version is 6
1439 	 *  - MLD has always Router Alert hop-by-hop option
1440 	 *  - But we do not support jumbrograms.
1441 	 */
1442 	if (ip6h->version != 6 ||
1443 	    ip6h->nexthdr != IPPROTO_HOPOPTS ||
1444 	    ip6h->payload_len == 0)
1445 		return 0;
1446 
1447 	len = ntohs(ip6h->payload_len) + sizeof(*ip6h);
1448 	if (skb->len < len)
1449 		return -EINVAL;
1450 
1451 	nexthdr = ip6h->nexthdr;
1452 	offset = ipv6_skip_exthdr(skb, sizeof(*ip6h), &nexthdr, &frag_off);
1453 
1454 	if (offset < 0 || nexthdr != IPPROTO_ICMPV6)
1455 		return 0;
1456 
1457 	/* Okay, we found ICMPv6 header */
1458 	skb2 = skb_clone(skb, GFP_ATOMIC);
1459 	if (!skb2)
1460 		return -ENOMEM;
1461 
1462 	err = -EINVAL;
1463 	if (!pskb_may_pull(skb2, offset + sizeof(struct icmp6hdr)))
1464 		goto out;
1465 
1466 	len -= offset - skb_network_offset(skb2);
1467 
1468 	__skb_pull(skb2, offset);
1469 	skb_reset_transport_header(skb2);
1470 	skb_postpull_rcsum(skb2, skb_network_header(skb2),
1471 			   skb_network_header_len(skb2));
1472 
1473 	icmp6_type = icmp6_hdr(skb2)->icmp6_type;
1474 
1475 	switch (icmp6_type) {
1476 	case ICMPV6_MGM_QUERY:
1477 	case ICMPV6_MGM_REPORT:
1478 	case ICMPV6_MGM_REDUCTION:
1479 	case ICMPV6_MLD2_REPORT:
1480 		break;
1481 	default:
1482 		err = 0;
1483 		goto out;
1484 	}
1485 
1486 	/* Okay, we found MLD message. Check further. */
1487 	if (skb2->len > len) {
1488 		err = pskb_trim_rcsum(skb2, len);
1489 		if (err)
1490 			goto out;
1491 		err = -EINVAL;
1492 	}
1493 
1494 	ip6h = ipv6_hdr(skb2);
1495 
1496 	switch (skb2->ip_summed) {
1497 	case CHECKSUM_COMPLETE:
1498 		if (!csum_ipv6_magic(&ip6h->saddr, &ip6h->daddr, skb2->len,
1499 					IPPROTO_ICMPV6, skb2->csum))
1500 			break;
1501 		/*FALLTHROUGH*/
1502 	case CHECKSUM_NONE:
1503 		skb2->csum = ~csum_unfold(csum_ipv6_magic(&ip6h->saddr,
1504 							&ip6h->daddr,
1505 							skb2->len,
1506 							IPPROTO_ICMPV6, 0));
1507 		if (__skb_checksum_complete(skb2))
1508 			goto out;
1509 	}
1510 
1511 	err = 0;
1512 
1513 	BR_INPUT_SKB_CB(skb)->igmp = 1;
1514 
1515 	switch (icmp6_type) {
1516 	case ICMPV6_MGM_REPORT:
1517 	    {
1518 		struct mld_msg *mld;
1519 		if (!pskb_may_pull(skb2, sizeof(*mld))) {
1520 			err = -EINVAL;
1521 			goto out;
1522 		}
1523 		mld = (struct mld_msg *)skb_transport_header(skb2);
1524 		BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1525 		err = br_ip6_multicast_add_group(br, port, &mld->mld_mca);
1526 		break;
1527 	    }
1528 	case ICMPV6_MLD2_REPORT:
1529 		err = br_ip6_multicast_mld2_report(br, port, skb2);
1530 		break;
1531 	case ICMPV6_MGM_QUERY:
1532 		err = br_ip6_multicast_query(br, port, skb2);
1533 		break;
1534 	case ICMPV6_MGM_REDUCTION:
1535 	    {
1536 		struct mld_msg *mld;
1537 		if (!pskb_may_pull(skb2, sizeof(*mld))) {
1538 			err = -EINVAL;
1539 			goto out;
1540 		}
1541 		mld = (struct mld_msg *)skb_transport_header(skb2);
1542 		br_ip6_multicast_leave_group(br, port, &mld->mld_mca);
1543 	    }
1544 	}
1545 
1546 out:
1547 	kfree_skb(skb2);
1548 	return err;
1549 }
1550 #endif
1551 
1552 int br_multicast_rcv(struct net_bridge *br, struct net_bridge_port *port,
1553 		     struct sk_buff *skb)
1554 {
1555 	BR_INPUT_SKB_CB(skb)->igmp = 0;
1556 	BR_INPUT_SKB_CB(skb)->mrouters_only = 0;
1557 
1558 	if (br->multicast_disabled)
1559 		return 0;
1560 
1561 	switch (skb->protocol) {
1562 	case htons(ETH_P_IP):
1563 		return br_multicast_ipv4_rcv(br, port, skb);
1564 #if IS_ENABLED(CONFIG_IPV6)
1565 	case htons(ETH_P_IPV6):
1566 		return br_multicast_ipv6_rcv(br, port, skb);
1567 #endif
1568 	}
1569 
1570 	return 0;
1571 }
1572 
1573 static void br_multicast_query_expired(unsigned long data)
1574 {
1575 	struct net_bridge *br = (void *)data;
1576 
1577 	spin_lock(&br->multicast_lock);
1578 	if (br->multicast_startup_queries_sent <
1579 	    br->multicast_startup_query_count)
1580 		br->multicast_startup_queries_sent++;
1581 
1582 	br_multicast_send_query(br, NULL, br->multicast_startup_queries_sent);
1583 
1584 	spin_unlock(&br->multicast_lock);
1585 }
1586 
1587 void br_multicast_init(struct net_bridge *br)
1588 {
1589 	br->hash_elasticity = 4;
1590 	br->hash_max = 512;
1591 
1592 	br->multicast_router = 1;
1593 	br->multicast_querier = 0;
1594 	br->multicast_last_member_count = 2;
1595 	br->multicast_startup_query_count = 2;
1596 
1597 	br->multicast_last_member_interval = HZ;
1598 	br->multicast_query_response_interval = 10 * HZ;
1599 	br->multicast_startup_query_interval = 125 * HZ / 4;
1600 	br->multicast_query_interval = 125 * HZ;
1601 	br->multicast_querier_interval = 255 * HZ;
1602 	br->multicast_membership_interval = 260 * HZ;
1603 
1604 	spin_lock_init(&br->multicast_lock);
1605 	setup_timer(&br->multicast_router_timer,
1606 		    br_multicast_local_router_expired, 0);
1607 	setup_timer(&br->multicast_querier_timer,
1608 		    br_multicast_querier_expired, (unsigned long)br);
1609 	setup_timer(&br->multicast_query_timer, br_multicast_query_expired,
1610 		    (unsigned long)br);
1611 }
1612 
1613 void br_multicast_open(struct net_bridge *br)
1614 {
1615 	br->multicast_startup_queries_sent = 0;
1616 
1617 	if (br->multicast_disabled)
1618 		return;
1619 
1620 	mod_timer(&br->multicast_query_timer, jiffies);
1621 }
1622 
1623 void br_multicast_stop(struct net_bridge *br)
1624 {
1625 	struct net_bridge_mdb_htable *mdb;
1626 	struct net_bridge_mdb_entry *mp;
1627 	struct hlist_node *p, *n;
1628 	u32 ver;
1629 	int i;
1630 
1631 	del_timer_sync(&br->multicast_router_timer);
1632 	del_timer_sync(&br->multicast_querier_timer);
1633 	del_timer_sync(&br->multicast_query_timer);
1634 
1635 	spin_lock_bh(&br->multicast_lock);
1636 	mdb = mlock_dereference(br->mdb, br);
1637 	if (!mdb)
1638 		goto out;
1639 
1640 	br->mdb = NULL;
1641 
1642 	ver = mdb->ver;
1643 	for (i = 0; i < mdb->max; i++) {
1644 		hlist_for_each_entry_safe(mp, p, n, &mdb->mhash[i],
1645 					  hlist[ver]) {
1646 			del_timer(&mp->timer);
1647 			call_rcu_bh(&mp->rcu, br_multicast_free_group);
1648 		}
1649 	}
1650 
1651 	if (mdb->old) {
1652 		spin_unlock_bh(&br->multicast_lock);
1653 		rcu_barrier_bh();
1654 		spin_lock_bh(&br->multicast_lock);
1655 		WARN_ON(mdb->old);
1656 	}
1657 
1658 	mdb->old = mdb;
1659 	call_rcu_bh(&mdb->rcu, br_mdb_free);
1660 
1661 out:
1662 	spin_unlock_bh(&br->multicast_lock);
1663 }
1664 
1665 int br_multicast_set_router(struct net_bridge *br, unsigned long val)
1666 {
1667 	int err = -ENOENT;
1668 
1669 	spin_lock_bh(&br->multicast_lock);
1670 	if (!netif_running(br->dev))
1671 		goto unlock;
1672 
1673 	switch (val) {
1674 	case 0:
1675 	case 2:
1676 		del_timer(&br->multicast_router_timer);
1677 		/* fall through */
1678 	case 1:
1679 		br->multicast_router = val;
1680 		err = 0;
1681 		break;
1682 
1683 	default:
1684 		err = -EINVAL;
1685 		break;
1686 	}
1687 
1688 unlock:
1689 	spin_unlock_bh(&br->multicast_lock);
1690 
1691 	return err;
1692 }
1693 
1694 int br_multicast_set_port_router(struct net_bridge_port *p, unsigned long val)
1695 {
1696 	struct net_bridge *br = p->br;
1697 	int err = -ENOENT;
1698 
1699 	spin_lock(&br->multicast_lock);
1700 	if (!netif_running(br->dev) || p->state == BR_STATE_DISABLED)
1701 		goto unlock;
1702 
1703 	switch (val) {
1704 	case 0:
1705 	case 1:
1706 	case 2:
1707 		p->multicast_router = val;
1708 		err = 0;
1709 
1710 		if (val < 2 && !hlist_unhashed(&p->rlist))
1711 			hlist_del_init_rcu(&p->rlist);
1712 
1713 		if (val == 1)
1714 			break;
1715 
1716 		del_timer(&p->multicast_router_timer);
1717 
1718 		if (val == 0)
1719 			break;
1720 
1721 		br_multicast_add_router(br, p);
1722 		break;
1723 
1724 	default:
1725 		err = -EINVAL;
1726 		break;
1727 	}
1728 
1729 unlock:
1730 	spin_unlock(&br->multicast_lock);
1731 
1732 	return err;
1733 }
1734 
1735 static void br_multicast_start_querier(struct net_bridge *br)
1736 {
1737 	struct net_bridge_port *port;
1738 
1739 	br_multicast_open(br);
1740 
1741 	list_for_each_entry(port, &br->port_list, list) {
1742 		if (port->state == BR_STATE_DISABLED ||
1743 		    port->state == BR_STATE_BLOCKING)
1744 			continue;
1745 
1746 		__br_multicast_enable_port(port);
1747 	}
1748 }
1749 
1750 int br_multicast_toggle(struct net_bridge *br, unsigned long val)
1751 {
1752 	int err = 0;
1753 	struct net_bridge_mdb_htable *mdb;
1754 
1755 	spin_lock_bh(&br->multicast_lock);
1756 	if (br->multicast_disabled == !val)
1757 		goto unlock;
1758 
1759 	br->multicast_disabled = !val;
1760 	if (br->multicast_disabled)
1761 		goto unlock;
1762 
1763 	if (!netif_running(br->dev))
1764 		goto unlock;
1765 
1766 	mdb = mlock_dereference(br->mdb, br);
1767 	if (mdb) {
1768 		if (mdb->old) {
1769 			err = -EEXIST;
1770 rollback:
1771 			br->multicast_disabled = !!val;
1772 			goto unlock;
1773 		}
1774 
1775 		err = br_mdb_rehash(&br->mdb, mdb->max,
1776 				    br->hash_elasticity);
1777 		if (err)
1778 			goto rollback;
1779 	}
1780 
1781 	br_multicast_start_querier(br);
1782 
1783 unlock:
1784 	spin_unlock_bh(&br->multicast_lock);
1785 
1786 	return err;
1787 }
1788 
1789 int br_multicast_set_querier(struct net_bridge *br, unsigned long val)
1790 {
1791 	val = !!val;
1792 
1793 	spin_lock_bh(&br->multicast_lock);
1794 	if (br->multicast_querier == val)
1795 		goto unlock;
1796 
1797 	br->multicast_querier = val;
1798 	if (val)
1799 		br_multicast_start_querier(br);
1800 
1801 unlock:
1802 	spin_unlock_bh(&br->multicast_lock);
1803 
1804 	return 0;
1805 }
1806 
1807 int br_multicast_set_hash_max(struct net_bridge *br, unsigned long val)
1808 {
1809 	int err = -ENOENT;
1810 	u32 old;
1811 	struct net_bridge_mdb_htable *mdb;
1812 
1813 	spin_lock(&br->multicast_lock);
1814 	if (!netif_running(br->dev))
1815 		goto unlock;
1816 
1817 	err = -EINVAL;
1818 	if (!is_power_of_2(val))
1819 		goto unlock;
1820 
1821 	mdb = mlock_dereference(br->mdb, br);
1822 	if (mdb && val < mdb->size)
1823 		goto unlock;
1824 
1825 	err = 0;
1826 
1827 	old = br->hash_max;
1828 	br->hash_max = val;
1829 
1830 	if (mdb) {
1831 		if (mdb->old) {
1832 			err = -EEXIST;
1833 rollback:
1834 			br->hash_max = old;
1835 			goto unlock;
1836 		}
1837 
1838 		err = br_mdb_rehash(&br->mdb, br->hash_max,
1839 				    br->hash_elasticity);
1840 		if (err)
1841 			goto rollback;
1842 	}
1843 
1844 unlock:
1845 	spin_unlock(&br->multicast_lock);
1846 
1847 	return err;
1848 }
1849