xref: /openbmc/linux/net/bridge/br_multicast.c (revision 1fafc7a9353ef68e1b8d4bb130cb6402cf7dfd5a)
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 defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
28 #include <net/ipv6.h>
29 #include <net/mld.h>
30 #include <net/addrconf.h>
31 #endif
32 
33 #include "br_private.h"
34 
35 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
36 static inline int ipv6_is_local_multicast(const struct in6_addr *addr)
37 {
38 	if (ipv6_addr_is_multicast(addr) &&
39 	    IPV6_ADDR_MC_SCOPE(addr) <= IPV6_ADDR_SCOPE_LINKLOCAL)
40 		return 1;
41 	return 0;
42 }
43 #endif
44 
45 static inline int br_ip_equal(const struct br_ip *a, const struct br_ip *b)
46 {
47 	if (a->proto != b->proto)
48 		return 0;
49 	switch (a->proto) {
50 	case htons(ETH_P_IP):
51 		return a->u.ip4 == b->u.ip4;
52 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
53 	case htons(ETH_P_IPV6):
54 		return ipv6_addr_equal(&a->u.ip6, &b->u.ip6);
55 #endif
56 	}
57 	return 0;
58 }
59 
60 static inline int __br_ip4_hash(struct net_bridge_mdb_htable *mdb, __be32 ip)
61 {
62 	return jhash_1word(mdb->secret, (__force u32)ip) & (mdb->max - 1);
63 }
64 
65 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
66 static inline int __br_ip6_hash(struct net_bridge_mdb_htable *mdb,
67 				const struct in6_addr *ip)
68 {
69 	return jhash2((__force u32 *)ip->s6_addr32, 4, mdb->secret) & (mdb->max - 1);
70 }
71 #endif
72 
73 static inline int br_ip_hash(struct net_bridge_mdb_htable *mdb,
74 			     struct br_ip *ip)
75 {
76 	switch (ip->proto) {
77 	case htons(ETH_P_IP):
78 		return __br_ip4_hash(mdb, ip->u.ip4);
79 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
80 	case htons(ETH_P_IPV6):
81 		return __br_ip6_hash(mdb, &ip->u.ip6);
82 #endif
83 	}
84 	return 0;
85 }
86 
87 static struct net_bridge_mdb_entry *__br_mdb_ip_get(
88 	struct net_bridge_mdb_htable *mdb, struct br_ip *dst, int hash)
89 {
90 	struct net_bridge_mdb_entry *mp;
91 	struct hlist_node *p;
92 
93 	hlist_for_each_entry_rcu(mp, p, &mdb->mhash[hash], hlist[mdb->ver]) {
94 		if (br_ip_equal(&mp->addr, dst))
95 			return mp;
96 	}
97 
98 	return NULL;
99 }
100 
101 static struct net_bridge_mdb_entry *br_mdb_ip4_get(
102 	struct net_bridge_mdb_htable *mdb, __be32 dst)
103 {
104 	struct br_ip br_dst;
105 
106 	br_dst.u.ip4 = dst;
107 	br_dst.proto = htons(ETH_P_IP);
108 
109 	return __br_mdb_ip_get(mdb, &br_dst, __br_ip4_hash(mdb, dst));
110 }
111 
112 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
113 static struct net_bridge_mdb_entry *br_mdb_ip6_get(
114 	struct net_bridge_mdb_htable *mdb, const struct in6_addr *dst)
115 {
116 	struct br_ip br_dst;
117 
118 	ipv6_addr_copy(&br_dst.u.ip6, dst);
119 	br_dst.proto = htons(ETH_P_IPV6);
120 
121 	return __br_mdb_ip_get(mdb, &br_dst, __br_ip6_hash(mdb, dst));
122 }
123 #endif
124 
125 static struct net_bridge_mdb_entry *br_mdb_ip_get(
126 	struct net_bridge_mdb_htable *mdb, struct br_ip *dst)
127 {
128 	return __br_mdb_ip_get(mdb, dst, br_ip_hash(mdb, dst));
129 }
130 
131 struct net_bridge_mdb_entry *br_mdb_get(struct net_bridge *br,
132 					struct sk_buff *skb)
133 {
134 	struct net_bridge_mdb_htable *mdb = br->mdb;
135 	struct br_ip ip;
136 
137 	if (!mdb || br->multicast_disabled)
138 		return NULL;
139 
140 	if (BR_INPUT_SKB_CB(skb)->igmp)
141 		return NULL;
142 
143 	ip.proto = skb->protocol;
144 
145 	switch (skb->protocol) {
146 	case htons(ETH_P_IP):
147 		ip.u.ip4 = ip_hdr(skb)->daddr;
148 		break;
149 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
150 	case htons(ETH_P_IPV6):
151 		ipv6_addr_copy(&ip.u.ip6, &ipv6_hdr(skb)->daddr);
152 		break;
153 #endif
154 	default:
155 		return NULL;
156 	}
157 
158 	return br_mdb_ip_get(mdb, &ip);
159 }
160 
161 static void br_mdb_free(struct rcu_head *head)
162 {
163 	struct net_bridge_mdb_htable *mdb =
164 		container_of(head, struct net_bridge_mdb_htable, rcu);
165 	struct net_bridge_mdb_htable *old = mdb->old;
166 
167 	mdb->old = NULL;
168 	kfree(old->mhash);
169 	kfree(old);
170 }
171 
172 static int br_mdb_copy(struct net_bridge_mdb_htable *new,
173 		       struct net_bridge_mdb_htable *old,
174 		       int elasticity)
175 {
176 	struct net_bridge_mdb_entry *mp;
177 	struct hlist_node *p;
178 	int maxlen;
179 	int len;
180 	int i;
181 
182 	for (i = 0; i < old->max; i++)
183 		hlist_for_each_entry(mp, p, &old->mhash[i], hlist[old->ver])
184 			hlist_add_head(&mp->hlist[new->ver],
185 				       &new->mhash[br_ip_hash(new, &mp->addr)]);
186 
187 	if (!elasticity)
188 		return 0;
189 
190 	maxlen = 0;
191 	for (i = 0; i < new->max; i++) {
192 		len = 0;
193 		hlist_for_each_entry(mp, p, &new->mhash[i], hlist[new->ver])
194 			len++;
195 		if (len > maxlen)
196 			maxlen = len;
197 	}
198 
199 	return maxlen > elasticity ? -EINVAL : 0;
200 }
201 
202 static void br_multicast_free_pg(struct rcu_head *head)
203 {
204 	struct net_bridge_port_group *p =
205 		container_of(head, struct net_bridge_port_group, rcu);
206 
207 	kfree(p);
208 }
209 
210 static void br_multicast_free_group(struct rcu_head *head)
211 {
212 	struct net_bridge_mdb_entry *mp =
213 		container_of(head, struct net_bridge_mdb_entry, rcu);
214 
215 	kfree(mp);
216 }
217 
218 static void br_multicast_group_expired(unsigned long data)
219 {
220 	struct net_bridge_mdb_entry *mp = (void *)data;
221 	struct net_bridge *br = mp->br;
222 	struct net_bridge_mdb_htable *mdb;
223 
224 	spin_lock(&br->multicast_lock);
225 	if (!netif_running(br->dev) || timer_pending(&mp->timer))
226 		goto out;
227 
228 	if (!hlist_unhashed(&mp->mglist))
229 		hlist_del_init(&mp->mglist);
230 
231 	if (mp->ports)
232 		goto out;
233 
234 	mdb = br->mdb;
235 	hlist_del_rcu(&mp->hlist[mdb->ver]);
236 	mdb->size--;
237 
238 	del_timer(&mp->query_timer);
239 	call_rcu_bh(&mp->rcu, br_multicast_free_group);
240 
241 out:
242 	spin_unlock(&br->multicast_lock);
243 }
244 
245 static void br_multicast_del_pg(struct net_bridge *br,
246 				struct net_bridge_port_group *pg)
247 {
248 	struct net_bridge_mdb_htable *mdb = br->mdb;
249 	struct net_bridge_mdb_entry *mp;
250 	struct net_bridge_port_group *p;
251 	struct net_bridge_port_group **pp;
252 
253 	mp = br_mdb_ip_get(mdb, &pg->addr);
254 	if (WARN_ON(!mp))
255 		return;
256 
257 	for (pp = &mp->ports; (p = *pp); pp = &p->next) {
258 		if (p != pg)
259 			continue;
260 
261 		*pp = p->next;
262 		hlist_del_init(&p->mglist);
263 		del_timer(&p->timer);
264 		del_timer(&p->query_timer);
265 		call_rcu_bh(&p->rcu, br_multicast_free_pg);
266 
267 		if (!mp->ports && hlist_unhashed(&mp->mglist) &&
268 		    netif_running(br->dev))
269 			mod_timer(&mp->timer, jiffies);
270 
271 		return;
272 	}
273 
274 	WARN_ON(1);
275 }
276 
277 static void br_multicast_port_group_expired(unsigned long data)
278 {
279 	struct net_bridge_port_group *pg = (void *)data;
280 	struct net_bridge *br = pg->port->br;
281 
282 	spin_lock(&br->multicast_lock);
283 	if (!netif_running(br->dev) || timer_pending(&pg->timer) ||
284 	    hlist_unhashed(&pg->mglist))
285 		goto out;
286 
287 	br_multicast_del_pg(br, pg);
288 
289 out:
290 	spin_unlock(&br->multicast_lock);
291 }
292 
293 static int br_mdb_rehash(struct net_bridge_mdb_htable **mdbp, int max,
294 			 int elasticity)
295 {
296 	struct net_bridge_mdb_htable *old = *mdbp;
297 	struct net_bridge_mdb_htable *mdb;
298 	int err;
299 
300 	mdb = kmalloc(sizeof(*mdb), GFP_ATOMIC);
301 	if (!mdb)
302 		return -ENOMEM;
303 
304 	mdb->max = max;
305 	mdb->old = old;
306 
307 	mdb->mhash = kzalloc(max * sizeof(*mdb->mhash), GFP_ATOMIC);
308 	if (!mdb->mhash) {
309 		kfree(mdb);
310 		return -ENOMEM;
311 	}
312 
313 	mdb->size = old ? old->size : 0;
314 	mdb->ver = old ? old->ver ^ 1 : 0;
315 
316 	if (!old || elasticity)
317 		get_random_bytes(&mdb->secret, sizeof(mdb->secret));
318 	else
319 		mdb->secret = old->secret;
320 
321 	if (!old)
322 		goto out;
323 
324 	err = br_mdb_copy(mdb, old, elasticity);
325 	if (err) {
326 		kfree(mdb->mhash);
327 		kfree(mdb);
328 		return err;
329 	}
330 
331 	call_rcu_bh(&mdb->rcu, br_mdb_free);
332 
333 out:
334 	rcu_assign_pointer(*mdbp, mdb);
335 
336 	return 0;
337 }
338 
339 static struct sk_buff *br_ip4_multicast_alloc_query(struct net_bridge *br,
340 						    __be32 group)
341 {
342 	struct sk_buff *skb;
343 	struct igmphdr *ih;
344 	struct ethhdr *eth;
345 	struct iphdr *iph;
346 
347 	skb = netdev_alloc_skb_ip_align(br->dev, sizeof(*eth) + sizeof(*iph) +
348 						 sizeof(*ih) + 4);
349 	if (!skb)
350 		goto out;
351 
352 	skb->protocol = htons(ETH_P_IP);
353 
354 	skb_reset_mac_header(skb);
355 	eth = eth_hdr(skb);
356 
357 	memcpy(eth->h_source, br->dev->dev_addr, 6);
358 	eth->h_dest[0] = 1;
359 	eth->h_dest[1] = 0;
360 	eth->h_dest[2] = 0x5e;
361 	eth->h_dest[3] = 0;
362 	eth->h_dest[4] = 0;
363 	eth->h_dest[5] = 1;
364 	eth->h_proto = htons(ETH_P_IP);
365 	skb_put(skb, sizeof(*eth));
366 
367 	skb_set_network_header(skb, skb->len);
368 	iph = ip_hdr(skb);
369 
370 	iph->version = 4;
371 	iph->ihl = 6;
372 	iph->tos = 0xc0;
373 	iph->tot_len = htons(sizeof(*iph) + sizeof(*ih) + 4);
374 	iph->id = 0;
375 	iph->frag_off = htons(IP_DF);
376 	iph->ttl = 1;
377 	iph->protocol = IPPROTO_IGMP;
378 	iph->saddr = 0;
379 	iph->daddr = htonl(INADDR_ALLHOSTS_GROUP);
380 	((u8 *)&iph[1])[0] = IPOPT_RA;
381 	((u8 *)&iph[1])[1] = 4;
382 	((u8 *)&iph[1])[2] = 0;
383 	((u8 *)&iph[1])[3] = 0;
384 	ip_send_check(iph);
385 	skb_put(skb, 24);
386 
387 	skb_set_transport_header(skb, skb->len);
388 	ih = igmp_hdr(skb);
389 	ih->type = IGMP_HOST_MEMBERSHIP_QUERY;
390 	ih->code = (group ? br->multicast_last_member_interval :
391 			    br->multicast_query_response_interval) /
392 		   (HZ / IGMP_TIMER_SCALE);
393 	ih->group = group;
394 	ih->csum = 0;
395 	ih->csum = ip_compute_csum((void *)ih, sizeof(struct igmphdr));
396 	skb_put(skb, sizeof(*ih));
397 
398 	__skb_pull(skb, sizeof(*eth));
399 
400 out:
401 	return skb;
402 }
403 
404 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
405 static struct sk_buff *br_ip6_multicast_alloc_query(struct net_bridge *br,
406 						    struct in6_addr *group)
407 {
408 	struct sk_buff *skb;
409 	struct ipv6hdr *ip6h;
410 	struct mld_msg *mldq;
411 	struct ethhdr *eth;
412 	u8 *hopopt;
413 	unsigned long interval;
414 
415 	skb = netdev_alloc_skb_ip_align(br->dev, sizeof(*eth) + sizeof(*ip6h) +
416 						 8 + sizeof(*mldq));
417 	if (!skb)
418 		goto out;
419 
420 	skb->protocol = htons(ETH_P_IPV6);
421 
422 	/* Ethernet header */
423 	skb_reset_mac_header(skb);
424 	eth = eth_hdr(skb);
425 
426 	memcpy(eth->h_source, br->dev->dev_addr, 6);
427 	ipv6_eth_mc_map(group, eth->h_dest);
428 	eth->h_proto = htons(ETH_P_IPV6);
429 	skb_put(skb, sizeof(*eth));
430 
431 	/* IPv6 header + HbH option */
432 	skb_set_network_header(skb, skb->len);
433 	ip6h = ipv6_hdr(skb);
434 
435 	*(__force __be32 *)ip6h = htonl(0x60000000);
436 	ip6h->payload_len = 8 + sizeof(*mldq);
437 	ip6h->nexthdr = IPPROTO_HOPOPTS;
438 	ip6h->hop_limit = 1;
439 	ipv6_addr_set(&ip6h->saddr, 0, 0, 0, 0);
440 	ipv6_addr_set(&ip6h->daddr, htonl(0xff020000), 0, 0, htonl(1));
441 
442 	hopopt = (u8 *)(ip6h + 1);
443 	hopopt[0] = IPPROTO_ICMPV6;		/* next hdr */
444 	hopopt[1] = 0;				/* length of HbH */
445 	hopopt[2] = IPV6_TLV_ROUTERALERT;	/* Router Alert */
446 	hopopt[3] = 2;				/* Length of RA Option */
447 	hopopt[4] = 0;				/* Type = 0x0000 (MLD) */
448 	hopopt[5] = 0;
449 	hopopt[6] = IPV6_TLV_PAD0;		/* Pad0 */
450 	hopopt[7] = IPV6_TLV_PAD0;		/* Pad0 */
451 
452 	skb_put(skb, sizeof(*ip6h) + 8);
453 
454 	/* ICMPv6 */
455 	skb_set_transport_header(skb, skb->len);
456 	mldq = (struct mld_msg *) icmp6_hdr(skb);
457 
458 	interval = ipv6_addr_any(group) ? br->multicast_last_member_interval :
459 					  br->multicast_query_response_interval;
460 
461 	mldq->mld_type = ICMPV6_MGM_QUERY;
462 	mldq->mld_code = 0;
463 	mldq->mld_cksum = 0;
464 	mldq->mld_maxdelay = htons((u16)jiffies_to_msecs(interval));
465 	mldq->mld_reserved = 0;
466 	ipv6_addr_copy(&mldq->mld_mca, group);
467 
468 	/* checksum */
469 	mldq->mld_cksum = csum_ipv6_magic(&ip6h->saddr, &ip6h->daddr,
470 					  sizeof(*mldq), IPPROTO_ICMPV6,
471 					  csum_partial(mldq,
472 						       sizeof(*mldq), 0));
473 	skb_put(skb, sizeof(*mldq));
474 
475 	__skb_pull(skb, sizeof(*eth));
476 
477 out:
478 	return skb;
479 }
480 #endif
481 
482 static struct sk_buff *br_multicast_alloc_query(struct net_bridge *br,
483 						struct br_ip *addr)
484 {
485 	switch (addr->proto) {
486 	case htons(ETH_P_IP):
487 		return br_ip4_multicast_alloc_query(br, addr->u.ip4);
488 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
489 	case htons(ETH_P_IPV6):
490 		return br_ip6_multicast_alloc_query(br, &addr->u.ip6);
491 #endif
492 	}
493 	return NULL;
494 }
495 
496 static void br_multicast_send_group_query(struct net_bridge_mdb_entry *mp)
497 {
498 	struct net_bridge *br = mp->br;
499 	struct sk_buff *skb;
500 
501 	skb = br_multicast_alloc_query(br, &mp->addr);
502 	if (!skb)
503 		goto timer;
504 
505 	netif_rx(skb);
506 
507 timer:
508 	if (++mp->queries_sent < br->multicast_last_member_count)
509 		mod_timer(&mp->query_timer,
510 			  jiffies + br->multicast_last_member_interval);
511 }
512 
513 static void br_multicast_group_query_expired(unsigned long data)
514 {
515 	struct net_bridge_mdb_entry *mp = (void *)data;
516 	struct net_bridge *br = mp->br;
517 
518 	spin_lock(&br->multicast_lock);
519 	if (!netif_running(br->dev) || hlist_unhashed(&mp->mglist) ||
520 	    mp->queries_sent >= br->multicast_last_member_count)
521 		goto out;
522 
523 	br_multicast_send_group_query(mp);
524 
525 out:
526 	spin_unlock(&br->multicast_lock);
527 }
528 
529 static void br_multicast_send_port_group_query(struct net_bridge_port_group *pg)
530 {
531 	struct net_bridge_port *port = pg->port;
532 	struct net_bridge *br = port->br;
533 	struct sk_buff *skb;
534 
535 	skb = br_multicast_alloc_query(br, &pg->addr);
536 	if (!skb)
537 		goto timer;
538 
539 	br_deliver(port, skb);
540 
541 timer:
542 	if (++pg->queries_sent < br->multicast_last_member_count)
543 		mod_timer(&pg->query_timer,
544 			  jiffies + br->multicast_last_member_interval);
545 }
546 
547 static void br_multicast_port_group_query_expired(unsigned long data)
548 {
549 	struct net_bridge_port_group *pg = (void *)data;
550 	struct net_bridge_port *port = pg->port;
551 	struct net_bridge *br = port->br;
552 
553 	spin_lock(&br->multicast_lock);
554 	if (!netif_running(br->dev) || hlist_unhashed(&pg->mglist) ||
555 	    pg->queries_sent >= br->multicast_last_member_count)
556 		goto out;
557 
558 	br_multicast_send_port_group_query(pg);
559 
560 out:
561 	spin_unlock(&br->multicast_lock);
562 }
563 
564 static struct net_bridge_mdb_entry *br_multicast_get_group(
565 	struct net_bridge *br, struct net_bridge_port *port,
566 	struct br_ip *group, int hash)
567 {
568 	struct net_bridge_mdb_htable *mdb = br->mdb;
569 	struct net_bridge_mdb_entry *mp;
570 	struct hlist_node *p;
571 	unsigned count = 0;
572 	unsigned max;
573 	int elasticity;
574 	int err;
575 
576 	hlist_for_each_entry(mp, p, &mdb->mhash[hash], hlist[mdb->ver]) {
577 		count++;
578 		if (unlikely(br_ip_equal(group, &mp->addr)))
579 			return mp;
580 	}
581 
582 	elasticity = 0;
583 	max = mdb->max;
584 
585 	if (unlikely(count > br->hash_elasticity && count)) {
586 		if (net_ratelimit())
587 			printk(KERN_INFO "%s: Multicast hash table "
588 			       "chain limit reached: %s\n",
589 			       br->dev->name, port ? port->dev->name :
590 						     br->dev->name);
591 
592 		elasticity = br->hash_elasticity;
593 	}
594 
595 	if (mdb->size >= max) {
596 		max *= 2;
597 		if (unlikely(max >= br->hash_max)) {
598 			printk(KERN_WARNING "%s: Multicast hash table maximum "
599 			       "reached, disabling snooping: %s, %d\n",
600 			       br->dev->name, port ? port->dev->name :
601 						     br->dev->name,
602 			       max);
603 			err = -E2BIG;
604 disable:
605 			br->multicast_disabled = 1;
606 			goto err;
607 		}
608 	}
609 
610 	if (max > mdb->max || elasticity) {
611 		if (mdb->old) {
612 			if (net_ratelimit())
613 				printk(KERN_INFO "%s: Multicast hash table "
614 				       "on fire: %s\n",
615 				       br->dev->name, port ? port->dev->name :
616 							     br->dev->name);
617 			err = -EEXIST;
618 			goto err;
619 		}
620 
621 		err = br_mdb_rehash(&br->mdb, max, elasticity);
622 		if (err) {
623 			printk(KERN_WARNING "%s: Cannot rehash multicast "
624 			       "hash table, disabling snooping: "
625 			       "%s, %d, %d\n",
626 			       br->dev->name, port ? port->dev->name :
627 						     br->dev->name,
628 			       mdb->size, err);
629 			goto disable;
630 		}
631 
632 		err = -EAGAIN;
633 		goto err;
634 	}
635 
636 	return NULL;
637 
638 err:
639 	mp = ERR_PTR(err);
640 	return mp;
641 }
642 
643 static struct net_bridge_mdb_entry *br_multicast_new_group(
644 	struct net_bridge *br, struct net_bridge_port *port,
645 	struct br_ip *group)
646 {
647 	struct net_bridge_mdb_htable *mdb = br->mdb;
648 	struct net_bridge_mdb_entry *mp;
649 	int hash;
650 
651 	if (!mdb) {
652 		if (br_mdb_rehash(&br->mdb, BR_HASH_SIZE, 0))
653 			return NULL;
654 		goto rehash;
655 	}
656 
657 	hash = br_ip_hash(mdb, group);
658 	mp = br_multicast_get_group(br, port, group, hash);
659 	switch (PTR_ERR(mp)) {
660 	case 0:
661 		break;
662 
663 	case -EAGAIN:
664 rehash:
665 		mdb = br->mdb;
666 		hash = br_ip_hash(mdb, group);
667 		break;
668 
669 	default:
670 		goto out;
671 	}
672 
673 	mp = kzalloc(sizeof(*mp), GFP_ATOMIC);
674 	if (unlikely(!mp))
675 		goto out;
676 
677 	mp->br = br;
678 	mp->addr = *group;
679 	setup_timer(&mp->timer, br_multicast_group_expired,
680 		    (unsigned long)mp);
681 	setup_timer(&mp->query_timer, br_multicast_group_query_expired,
682 		    (unsigned long)mp);
683 
684 	hlist_add_head_rcu(&mp->hlist[mdb->ver], &mdb->mhash[hash]);
685 	mdb->size++;
686 
687 out:
688 	return mp;
689 }
690 
691 static int br_multicast_add_group(struct net_bridge *br,
692 				  struct net_bridge_port *port,
693 				  struct br_ip *group)
694 {
695 	struct net_bridge_mdb_entry *mp;
696 	struct net_bridge_port_group *p;
697 	struct net_bridge_port_group **pp;
698 	unsigned long now = jiffies;
699 	int err;
700 
701 	spin_lock(&br->multicast_lock);
702 	if (!netif_running(br->dev) ||
703 	    (port && port->state == BR_STATE_DISABLED))
704 		goto out;
705 
706 	mp = br_multicast_new_group(br, port, group);
707 	err = PTR_ERR(mp);
708 	if (unlikely(IS_ERR(mp) || !mp))
709 		goto err;
710 
711 	if (!port) {
712 		hlist_add_head(&mp->mglist, &br->mglist);
713 		mod_timer(&mp->timer, now + br->multicast_membership_interval);
714 		goto out;
715 	}
716 
717 	for (pp = &mp->ports; (p = *pp); pp = &p->next) {
718 		if (p->port == port)
719 			goto found;
720 		if ((unsigned long)p->port < (unsigned long)port)
721 			break;
722 	}
723 
724 	p = kzalloc(sizeof(*p), GFP_ATOMIC);
725 	err = -ENOMEM;
726 	if (unlikely(!p))
727 		goto err;
728 
729 	p->addr = *group;
730 	p->port = port;
731 	p->next = *pp;
732 	hlist_add_head(&p->mglist, &port->mglist);
733 	setup_timer(&p->timer, br_multicast_port_group_expired,
734 		    (unsigned long)p);
735 	setup_timer(&p->query_timer, br_multicast_port_group_query_expired,
736 		    (unsigned long)p);
737 
738 	rcu_assign_pointer(*pp, p);
739 
740 found:
741 	mod_timer(&p->timer, now + br->multicast_membership_interval);
742 out:
743 	err = 0;
744 
745 err:
746 	spin_unlock(&br->multicast_lock);
747 	return err;
748 }
749 
750 static int br_ip4_multicast_add_group(struct net_bridge *br,
751 				      struct net_bridge_port *port,
752 				      __be32 group)
753 {
754 	struct br_ip br_group;
755 
756 	if (ipv4_is_local_multicast(group))
757 		return 0;
758 
759 	br_group.u.ip4 = group;
760 	br_group.proto = htons(ETH_P_IP);
761 
762 	return br_multicast_add_group(br, port, &br_group);
763 }
764 
765 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
766 static int br_ip6_multicast_add_group(struct net_bridge *br,
767 				      struct net_bridge_port *port,
768 				      const struct in6_addr *group)
769 {
770 	struct br_ip br_group;
771 
772 	if (ipv6_is_local_multicast(group))
773 		return 0;
774 
775 	ipv6_addr_copy(&br_group.u.ip6, group);
776 	br_group.proto = htons(ETH_P_IP);
777 
778 	return br_multicast_add_group(br, port, &br_group);
779 }
780 #endif
781 
782 static void br_multicast_router_expired(unsigned long data)
783 {
784 	struct net_bridge_port *port = (void *)data;
785 	struct net_bridge *br = port->br;
786 
787 	spin_lock(&br->multicast_lock);
788 	if (port->multicast_router != 1 ||
789 	    timer_pending(&port->multicast_router_timer) ||
790 	    hlist_unhashed(&port->rlist))
791 		goto out;
792 
793 	hlist_del_init_rcu(&port->rlist);
794 
795 out:
796 	spin_unlock(&br->multicast_lock);
797 }
798 
799 static void br_multicast_local_router_expired(unsigned long data)
800 {
801 }
802 
803 static void __br_multicast_send_query(struct net_bridge *br,
804 				      struct net_bridge_port *port,
805 				      struct br_ip *ip)
806 {
807 	struct sk_buff *skb;
808 
809 	skb = br_multicast_alloc_query(br, ip);
810 	if (!skb)
811 		return;
812 
813 	if (port) {
814 		__skb_push(skb, sizeof(struct ethhdr));
815 		skb->dev = port->dev;
816 		NF_HOOK(PF_BRIDGE, NF_BR_LOCAL_OUT, skb, NULL, skb->dev,
817 			dev_queue_xmit);
818 	} else
819 		netif_rx(skb);
820 }
821 
822 static void br_multicast_send_query(struct net_bridge *br,
823 				    struct net_bridge_port *port, u32 sent)
824 {
825 	unsigned long time;
826 	struct br_ip br_group;
827 
828 	if (!netif_running(br->dev) || br->multicast_disabled ||
829 	    timer_pending(&br->multicast_querier_timer))
830 		return;
831 
832 	memset(&br_group.u, 0, sizeof(br_group.u));
833 
834 	br_group.proto = htons(ETH_P_IP);
835 	__br_multicast_send_query(br, port, &br_group);
836 
837 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
838 	br_group.proto = htons(ETH_P_IPV6);
839 	__br_multicast_send_query(br, port, &br_group);
840 #endif
841 
842 	time = jiffies;
843 	time += sent < br->multicast_startup_query_count ?
844 		br->multicast_startup_query_interval :
845 		br->multicast_query_interval;
846 	mod_timer(port ? &port->multicast_query_timer :
847 			 &br->multicast_query_timer, time);
848 }
849 
850 static void br_multicast_port_query_expired(unsigned long data)
851 {
852 	struct net_bridge_port *port = (void *)data;
853 	struct net_bridge *br = port->br;
854 
855 	spin_lock(&br->multicast_lock);
856 	if (port->state == BR_STATE_DISABLED ||
857 	    port->state == BR_STATE_BLOCKING)
858 		goto out;
859 
860 	if (port->multicast_startup_queries_sent <
861 	    br->multicast_startup_query_count)
862 		port->multicast_startup_queries_sent++;
863 
864 	br_multicast_send_query(port->br, port,
865 				port->multicast_startup_queries_sent);
866 
867 out:
868 	spin_unlock(&br->multicast_lock);
869 }
870 
871 void br_multicast_add_port(struct net_bridge_port *port)
872 {
873 	port->multicast_router = 1;
874 
875 	setup_timer(&port->multicast_router_timer, br_multicast_router_expired,
876 		    (unsigned long)port);
877 	setup_timer(&port->multicast_query_timer,
878 		    br_multicast_port_query_expired, (unsigned long)port);
879 }
880 
881 void br_multicast_del_port(struct net_bridge_port *port)
882 {
883 	del_timer_sync(&port->multicast_router_timer);
884 }
885 
886 static void __br_multicast_enable_port(struct net_bridge_port *port)
887 {
888 	port->multicast_startup_queries_sent = 0;
889 
890 	if (try_to_del_timer_sync(&port->multicast_query_timer) >= 0 ||
891 	    del_timer(&port->multicast_query_timer))
892 		mod_timer(&port->multicast_query_timer, jiffies);
893 }
894 
895 void br_multicast_enable_port(struct net_bridge_port *port)
896 {
897 	struct net_bridge *br = port->br;
898 
899 	spin_lock(&br->multicast_lock);
900 	if (br->multicast_disabled || !netif_running(br->dev))
901 		goto out;
902 
903 	__br_multicast_enable_port(port);
904 
905 out:
906 	spin_unlock(&br->multicast_lock);
907 }
908 
909 void br_multicast_disable_port(struct net_bridge_port *port)
910 {
911 	struct net_bridge *br = port->br;
912 	struct net_bridge_port_group *pg;
913 	struct hlist_node *p, *n;
914 
915 	spin_lock(&br->multicast_lock);
916 	hlist_for_each_entry_safe(pg, p, n, &port->mglist, mglist)
917 		br_multicast_del_pg(br, pg);
918 
919 	if (!hlist_unhashed(&port->rlist))
920 		hlist_del_init_rcu(&port->rlist);
921 	del_timer(&port->multicast_router_timer);
922 	del_timer(&port->multicast_query_timer);
923 	spin_unlock(&br->multicast_lock);
924 }
925 
926 static int br_ip4_multicast_igmp3_report(struct net_bridge *br,
927 					 struct net_bridge_port *port,
928 					 struct sk_buff *skb)
929 {
930 	struct igmpv3_report *ih;
931 	struct igmpv3_grec *grec;
932 	int i;
933 	int len;
934 	int num;
935 	int type;
936 	int err = 0;
937 	__be32 group;
938 
939 	if (!pskb_may_pull(skb, sizeof(*ih)))
940 		return -EINVAL;
941 
942 	ih = igmpv3_report_hdr(skb);
943 	num = ntohs(ih->ngrec);
944 	len = sizeof(*ih);
945 
946 	for (i = 0; i < num; i++) {
947 		len += sizeof(*grec);
948 		if (!pskb_may_pull(skb, len))
949 			return -EINVAL;
950 
951 		grec = (void *)(skb->data + len - sizeof(*grec));
952 		group = grec->grec_mca;
953 		type = grec->grec_type;
954 
955 		len += ntohs(grec->grec_nsrcs) * 4;
956 		if (!pskb_may_pull(skb, len))
957 			return -EINVAL;
958 
959 		/* We treat this as an IGMPv2 report for now. */
960 		switch (type) {
961 		case IGMPV3_MODE_IS_INCLUDE:
962 		case IGMPV3_MODE_IS_EXCLUDE:
963 		case IGMPV3_CHANGE_TO_INCLUDE:
964 		case IGMPV3_CHANGE_TO_EXCLUDE:
965 		case IGMPV3_ALLOW_NEW_SOURCES:
966 		case IGMPV3_BLOCK_OLD_SOURCES:
967 			break;
968 
969 		default:
970 			continue;
971 		}
972 
973 		err = br_ip4_multicast_add_group(br, port, group);
974 		if (err)
975 			break;
976 	}
977 
978 	return err;
979 }
980 
981 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
982 static int br_ip6_multicast_mld2_report(struct net_bridge *br,
983 					struct net_bridge_port *port,
984 					struct sk_buff *skb)
985 {
986 	struct icmp6hdr *icmp6h;
987 	struct mld2_grec *grec;
988 	int i;
989 	int len;
990 	int num;
991 	int err = 0;
992 
993 	if (!pskb_may_pull(skb, sizeof(*icmp6h)))
994 		return -EINVAL;
995 
996 	icmp6h = icmp6_hdr(skb);
997 	num = ntohs(icmp6h->icmp6_dataun.un_data16[1]);
998 	len = sizeof(*icmp6h);
999 
1000 	for (i = 0; i < num; i++) {
1001 		__be16 *nsrcs, _nsrcs;
1002 
1003 		nsrcs = skb_header_pointer(skb,
1004 					   len + offsetof(struct mld2_grec,
1005 							  grec_mca),
1006 					   sizeof(_nsrcs), &_nsrcs);
1007 		if (!nsrcs)
1008 			return -EINVAL;
1009 
1010 		if (!pskb_may_pull(skb,
1011 				   len + sizeof(*grec) +
1012 				   sizeof(struct in6_addr) * (*nsrcs)))
1013 			return -EINVAL;
1014 
1015 		grec = (struct mld2_grec *)(skb->data + len);
1016 		len += sizeof(*grec) + sizeof(struct in6_addr) * (*nsrcs);
1017 
1018 		/* We treat these as MLDv1 reports for now. */
1019 		switch (grec->grec_type) {
1020 		case MLD2_MODE_IS_INCLUDE:
1021 		case MLD2_MODE_IS_EXCLUDE:
1022 		case MLD2_CHANGE_TO_INCLUDE:
1023 		case MLD2_CHANGE_TO_EXCLUDE:
1024 		case MLD2_ALLOW_NEW_SOURCES:
1025 		case MLD2_BLOCK_OLD_SOURCES:
1026 			break;
1027 
1028 		default:
1029 			continue;
1030 		}
1031 
1032 		err = br_ip6_multicast_add_group(br, port, &grec->grec_mca);
1033 		if (!err)
1034 			break;
1035 	}
1036 
1037 	return err;
1038 }
1039 #endif
1040 
1041 static void br_multicast_add_router(struct net_bridge *br,
1042 				    struct net_bridge_port *port)
1043 {
1044 	struct hlist_node *p;
1045 	struct hlist_node **h;
1046 
1047 	for (h = &br->router_list.first;
1048 	     (p = *h) &&
1049 	     (unsigned long)container_of(p, struct net_bridge_port, rlist) >
1050 	     (unsigned long)port;
1051 	     h = &p->next)
1052 		;
1053 
1054 	port->rlist.pprev = h;
1055 	port->rlist.next = p;
1056 	rcu_assign_pointer(*h, &port->rlist);
1057 	if (p)
1058 		p->pprev = &port->rlist.next;
1059 }
1060 
1061 static void br_multicast_mark_router(struct net_bridge *br,
1062 				     struct net_bridge_port *port)
1063 {
1064 	unsigned long now = jiffies;
1065 
1066 	if (!port) {
1067 		if (br->multicast_router == 1)
1068 			mod_timer(&br->multicast_router_timer,
1069 				  now + br->multicast_querier_interval);
1070 		return;
1071 	}
1072 
1073 	if (port->multicast_router != 1)
1074 		return;
1075 
1076 	if (!hlist_unhashed(&port->rlist))
1077 		goto timer;
1078 
1079 	br_multicast_add_router(br, port);
1080 
1081 timer:
1082 	mod_timer(&port->multicast_router_timer,
1083 		  now + br->multicast_querier_interval);
1084 }
1085 
1086 static void br_multicast_query_received(struct net_bridge *br,
1087 					struct net_bridge_port *port,
1088 					int saddr)
1089 {
1090 	if (saddr)
1091 		mod_timer(&br->multicast_querier_timer,
1092 			  jiffies + br->multicast_querier_interval);
1093 	else if (timer_pending(&br->multicast_querier_timer))
1094 		return;
1095 
1096 	br_multicast_mark_router(br, port);
1097 }
1098 
1099 static int br_ip4_multicast_query(struct net_bridge *br,
1100 				  struct net_bridge_port *port,
1101 				  struct sk_buff *skb)
1102 {
1103 	struct iphdr *iph = ip_hdr(skb);
1104 	struct igmphdr *ih = igmp_hdr(skb);
1105 	struct net_bridge_mdb_entry *mp;
1106 	struct igmpv3_query *ih3;
1107 	struct net_bridge_port_group *p;
1108 	struct net_bridge_port_group **pp;
1109 	unsigned long max_delay;
1110 	unsigned long now = jiffies;
1111 	__be32 group;
1112 	int err = 0;
1113 
1114 	spin_lock(&br->multicast_lock);
1115 	if (!netif_running(br->dev) ||
1116 	    (port && port->state == BR_STATE_DISABLED))
1117 		goto out;
1118 
1119 	br_multicast_query_received(br, port, !!iph->saddr);
1120 
1121 	group = ih->group;
1122 
1123 	if (skb->len == sizeof(*ih)) {
1124 		max_delay = ih->code * (HZ / IGMP_TIMER_SCALE);
1125 
1126 		if (!max_delay) {
1127 			max_delay = 10 * HZ;
1128 			group = 0;
1129 		}
1130 	} else {
1131 		if (!pskb_may_pull(skb, sizeof(struct igmpv3_query))) {
1132 			err = -EINVAL;
1133 			goto out;
1134 		}
1135 
1136 		ih3 = igmpv3_query_hdr(skb);
1137 		if (ih3->nsrcs)
1138 			goto out;
1139 
1140 		max_delay = ih3->code ?
1141 			    IGMPV3_MRC(ih3->code) * (HZ / IGMP_TIMER_SCALE) : 1;
1142 	}
1143 
1144 	if (!group)
1145 		goto out;
1146 
1147 	mp = br_mdb_ip4_get(br->mdb, group);
1148 	if (!mp)
1149 		goto out;
1150 
1151 	max_delay *= br->multicast_last_member_count;
1152 
1153 	if (!hlist_unhashed(&mp->mglist) &&
1154 	    (timer_pending(&mp->timer) ?
1155 	     time_after(mp->timer.expires, now + max_delay) :
1156 	     try_to_del_timer_sync(&mp->timer) >= 0))
1157 		mod_timer(&mp->timer, now + max_delay);
1158 
1159 	for (pp = &mp->ports; (p = *pp); pp = &p->next) {
1160 		if (timer_pending(&p->timer) ?
1161 		    time_after(p->timer.expires, now + max_delay) :
1162 		    try_to_del_timer_sync(&p->timer) >= 0)
1163 			mod_timer(&mp->timer, now + max_delay);
1164 	}
1165 
1166 out:
1167 	spin_unlock(&br->multicast_lock);
1168 	return err;
1169 }
1170 
1171 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
1172 static int br_ip6_multicast_query(struct net_bridge *br,
1173 				  struct net_bridge_port *port,
1174 				  struct sk_buff *skb)
1175 {
1176 	struct ipv6hdr *ip6h = ipv6_hdr(skb);
1177 	struct mld_msg *mld = (struct mld_msg *) icmp6_hdr(skb);
1178 	struct net_bridge_mdb_entry *mp;
1179 	struct mld2_query *mld2q;
1180 	struct net_bridge_port_group *p, **pp;
1181 	unsigned long max_delay;
1182 	unsigned long now = jiffies;
1183 	struct in6_addr *group = NULL;
1184 	int err = 0;
1185 
1186 	spin_lock(&br->multicast_lock);
1187 	if (!netif_running(br->dev) ||
1188 	    (port && port->state == BR_STATE_DISABLED))
1189 		goto out;
1190 
1191 	br_multicast_query_received(br, port, !ipv6_addr_any(&ip6h->saddr));
1192 
1193 	if (skb->len == sizeof(*mld)) {
1194 		if (!pskb_may_pull(skb, sizeof(*mld))) {
1195 			err = -EINVAL;
1196 			goto out;
1197 		}
1198 		mld = (struct mld_msg *) icmp6_hdr(skb);
1199 		max_delay = msecs_to_jiffies(htons(mld->mld_maxdelay));
1200 		if (max_delay)
1201 			group = &mld->mld_mca;
1202 	} else if (skb->len >= sizeof(*mld2q)) {
1203 		if (!pskb_may_pull(skb, sizeof(*mld2q))) {
1204 			err = -EINVAL;
1205 			goto out;
1206 		}
1207 		mld2q = (struct mld2_query *)icmp6_hdr(skb);
1208 		if (!mld2q->mld2q_nsrcs)
1209 			group = &mld2q->mld2q_mca;
1210 		max_delay = mld2q->mld2q_mrc ? MLDV2_MRC(mld2q->mld2q_mrc) : 1;
1211 	}
1212 
1213 	if (!group)
1214 		goto out;
1215 
1216 	mp = br_mdb_ip6_get(br->mdb, group);
1217 	if (!mp)
1218 		goto out;
1219 
1220 	max_delay *= br->multicast_last_member_count;
1221 	if (!hlist_unhashed(&mp->mglist) &&
1222 	    (timer_pending(&mp->timer) ?
1223 	     time_after(mp->timer.expires, now + max_delay) :
1224 	     try_to_del_timer_sync(&mp->timer) >= 0))
1225 		mod_timer(&mp->timer, now + max_delay);
1226 
1227 	for (pp = &mp->ports; (p = *pp); pp = &p->next) {
1228 		if (timer_pending(&p->timer) ?
1229 		    time_after(p->timer.expires, now + max_delay) :
1230 		    try_to_del_timer_sync(&p->timer) >= 0)
1231 			mod_timer(&mp->timer, now + max_delay);
1232 	}
1233 
1234 out:
1235 	spin_unlock(&br->multicast_lock);
1236 	return err;
1237 }
1238 #endif
1239 
1240 static void br_multicast_leave_group(struct net_bridge *br,
1241 				     struct net_bridge_port *port,
1242 				     struct br_ip *group)
1243 {
1244 	struct net_bridge_mdb_htable *mdb;
1245 	struct net_bridge_mdb_entry *mp;
1246 	struct net_bridge_port_group *p;
1247 	unsigned long now;
1248 	unsigned long time;
1249 
1250 	spin_lock(&br->multicast_lock);
1251 	if (!netif_running(br->dev) ||
1252 	    (port && port->state == BR_STATE_DISABLED) ||
1253 	    timer_pending(&br->multicast_querier_timer))
1254 		goto out;
1255 
1256 	mdb = br->mdb;
1257 	mp = br_mdb_ip_get(mdb, group);
1258 	if (!mp)
1259 		goto out;
1260 
1261 	now = jiffies;
1262 	time = now + br->multicast_last_member_count *
1263 		     br->multicast_last_member_interval;
1264 
1265 	if (!port) {
1266 		if (!hlist_unhashed(&mp->mglist) &&
1267 		    (timer_pending(&mp->timer) ?
1268 		     time_after(mp->timer.expires, time) :
1269 		     try_to_del_timer_sync(&mp->timer) >= 0)) {
1270 			mod_timer(&mp->timer, time);
1271 
1272 			mp->queries_sent = 0;
1273 			mod_timer(&mp->query_timer, now);
1274 		}
1275 
1276 		goto out;
1277 	}
1278 
1279 	for (p = mp->ports; p; p = p->next) {
1280 		if (p->port != port)
1281 			continue;
1282 
1283 		if (!hlist_unhashed(&p->mglist) &&
1284 		    (timer_pending(&p->timer) ?
1285 		     time_after(p->timer.expires, time) :
1286 		     try_to_del_timer_sync(&p->timer) >= 0)) {
1287 			mod_timer(&p->timer, time);
1288 
1289 			p->queries_sent = 0;
1290 			mod_timer(&p->query_timer, now);
1291 		}
1292 
1293 		break;
1294 	}
1295 
1296 out:
1297 	spin_unlock(&br->multicast_lock);
1298 }
1299 
1300 static void br_ip4_multicast_leave_group(struct net_bridge *br,
1301 					 struct net_bridge_port *port,
1302 					 __be32 group)
1303 {
1304 	struct br_ip br_group;
1305 
1306 	if (ipv4_is_local_multicast(group))
1307 		return;
1308 
1309 	br_group.u.ip4 = group;
1310 	br_group.proto = htons(ETH_P_IP);
1311 
1312 	br_multicast_leave_group(br, port, &br_group);
1313 }
1314 
1315 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
1316 static void br_ip6_multicast_leave_group(struct net_bridge *br,
1317 					 struct net_bridge_port *port,
1318 					 const struct in6_addr *group)
1319 {
1320 	struct br_ip br_group;
1321 
1322 	if (ipv6_is_local_multicast(group))
1323 		return;
1324 
1325 	ipv6_addr_copy(&br_group.u.ip6, group);
1326 	br_group.proto = htons(ETH_P_IPV6);
1327 
1328 	br_multicast_leave_group(br, port, &br_group);
1329 }
1330 #endif
1331 
1332 static int br_multicast_ipv4_rcv(struct net_bridge *br,
1333 				 struct net_bridge_port *port,
1334 				 struct sk_buff *skb)
1335 {
1336 	struct sk_buff *skb2 = skb;
1337 	struct iphdr *iph;
1338 	struct igmphdr *ih;
1339 	unsigned len;
1340 	unsigned offset;
1341 	int err;
1342 
1343 	/* We treat OOM as packet loss for now. */
1344 	if (!pskb_may_pull(skb, sizeof(*iph)))
1345 		return -EINVAL;
1346 
1347 	iph = ip_hdr(skb);
1348 
1349 	if (iph->ihl < 5 || iph->version != 4)
1350 		return -EINVAL;
1351 
1352 	if (!pskb_may_pull(skb, ip_hdrlen(skb)))
1353 		return -EINVAL;
1354 
1355 	iph = ip_hdr(skb);
1356 
1357 	if (unlikely(ip_fast_csum((u8 *)iph, iph->ihl)))
1358 		return -EINVAL;
1359 
1360 	if (iph->protocol != IPPROTO_IGMP)
1361 		return 0;
1362 
1363 	len = ntohs(iph->tot_len);
1364 	if (skb->len < len || len < ip_hdrlen(skb))
1365 		return -EINVAL;
1366 
1367 	if (skb->len > len) {
1368 		skb2 = skb_clone(skb, GFP_ATOMIC);
1369 		if (!skb2)
1370 			return -ENOMEM;
1371 
1372 		err = pskb_trim_rcsum(skb2, len);
1373 		if (err)
1374 			goto err_out;
1375 	}
1376 
1377 	len -= ip_hdrlen(skb2);
1378 	offset = skb_network_offset(skb2) + ip_hdrlen(skb2);
1379 	__skb_pull(skb2, offset);
1380 	skb_reset_transport_header(skb2);
1381 
1382 	err = -EINVAL;
1383 	if (!pskb_may_pull(skb2, sizeof(*ih)))
1384 		goto out;
1385 
1386 	switch (skb2->ip_summed) {
1387 	case CHECKSUM_COMPLETE:
1388 		if (!csum_fold(skb2->csum))
1389 			break;
1390 		/* fall through */
1391 	case CHECKSUM_NONE:
1392 		skb2->csum = 0;
1393 		if (skb_checksum_complete(skb2))
1394 			goto out;
1395 	}
1396 
1397 	err = 0;
1398 
1399 	BR_INPUT_SKB_CB(skb)->igmp = 1;
1400 	ih = igmp_hdr(skb2);
1401 
1402 	switch (ih->type) {
1403 	case IGMP_HOST_MEMBERSHIP_REPORT:
1404 	case IGMPV2_HOST_MEMBERSHIP_REPORT:
1405 		BR_INPUT_SKB_CB(skb2)->mrouters_only = 1;
1406 		err = br_ip4_multicast_add_group(br, port, ih->group);
1407 		break;
1408 	case IGMPV3_HOST_MEMBERSHIP_REPORT:
1409 		err = br_ip4_multicast_igmp3_report(br, port, skb2);
1410 		break;
1411 	case IGMP_HOST_MEMBERSHIP_QUERY:
1412 		err = br_ip4_multicast_query(br, port, skb2);
1413 		break;
1414 	case IGMP_HOST_LEAVE_MESSAGE:
1415 		br_ip4_multicast_leave_group(br, port, ih->group);
1416 		break;
1417 	}
1418 
1419 out:
1420 	__skb_push(skb2, offset);
1421 err_out:
1422 	if (skb2 != skb)
1423 		kfree_skb(skb2);
1424 	return err;
1425 }
1426 
1427 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
1428 static int br_multicast_ipv6_rcv(struct net_bridge *br,
1429 				 struct net_bridge_port *port,
1430 				 struct sk_buff *skb)
1431 {
1432 	struct sk_buff *skb2 = skb;
1433 	struct ipv6hdr *ip6h;
1434 	struct icmp6hdr *icmp6h;
1435 	u8 nexthdr;
1436 	unsigned len;
1437 	unsigned offset;
1438 	int err;
1439 
1440 	if (!pskb_may_pull(skb, sizeof(*ip6h)))
1441 		return -EINVAL;
1442 
1443 	ip6h = ipv6_hdr(skb);
1444 
1445 	/*
1446 	 * We're interested in MLD messages only.
1447 	 *  - Version is 6
1448 	 *  - MLD has always Router Alert hop-by-hop option
1449 	 *  - But we do not support jumbrograms.
1450 	 */
1451 	if (ip6h->version != 6 ||
1452 	    ip6h->nexthdr != IPPROTO_HOPOPTS ||
1453 	    ip6h->payload_len == 0)
1454 		return 0;
1455 
1456 	len = ntohs(ip6h->payload_len);
1457 	if (skb->len < len)
1458 		return -EINVAL;
1459 
1460 	nexthdr = ip6h->nexthdr;
1461 	offset = ipv6_skip_exthdr(skb, sizeof(*ip6h), &nexthdr);
1462 
1463 	if (offset < 0 || nexthdr != IPPROTO_ICMPV6)
1464 		return 0;
1465 
1466 	/* Okay, we found ICMPv6 header */
1467 	skb2 = skb_clone(skb, GFP_ATOMIC);
1468 	if (!skb2)
1469 		return -ENOMEM;
1470 
1471 	len -= offset - skb_network_offset(skb2);
1472 
1473 	__skb_pull(skb2, offset);
1474 	skb_reset_transport_header(skb2);
1475 
1476 	err = -EINVAL;
1477 	if (!pskb_may_pull(skb2, sizeof(*icmp6h)))
1478 		goto out;
1479 
1480 	icmp6h = icmp6_hdr(skb2);
1481 
1482 	switch (icmp6h->icmp6_type) {
1483 	case ICMPV6_MGM_QUERY:
1484 	case ICMPV6_MGM_REPORT:
1485 	case ICMPV6_MGM_REDUCTION:
1486 	case ICMPV6_MLD2_REPORT:
1487 		break;
1488 	default:
1489 		err = 0;
1490 		goto out;
1491 	}
1492 
1493 	/* Okay, we found MLD message. Check further. */
1494 	if (skb2->len > len) {
1495 		err = pskb_trim_rcsum(skb2, len);
1496 		if (err)
1497 			goto out;
1498 	}
1499 
1500 	switch (skb2->ip_summed) {
1501 	case CHECKSUM_COMPLETE:
1502 		if (!csum_fold(skb2->csum))
1503 			break;
1504 		/*FALLTHROUGH*/
1505 	case CHECKSUM_NONE:
1506 		skb2->csum = 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 (icmp6h->icmp6_type) {
1516 	case ICMPV6_MGM_REPORT:
1517 	    {
1518 		struct mld_msg *mld = (struct mld_msg *)icmp6h;
1519 		BR_INPUT_SKB_CB(skb2)->mrouters_only = 1;
1520 		err = br_ip6_multicast_add_group(br, port, &mld->mld_mca);
1521 		break;
1522 	    }
1523 	case ICMPV6_MLD2_REPORT:
1524 		err = br_ip6_multicast_mld2_report(br, port, skb2);
1525 		break;
1526 	case ICMPV6_MGM_QUERY:
1527 		err = br_ip6_multicast_query(br, port, skb2);
1528 		break;
1529 	case ICMPV6_MGM_REDUCTION:
1530 	    {
1531 		struct mld_msg *mld = (struct mld_msg *)icmp6h;
1532 		br_ip6_multicast_leave_group(br, port, &mld->mld_mca);
1533 	    }
1534 	}
1535 
1536 out:
1537 	__skb_push(skb2, offset);
1538 	if (skb2 != skb)
1539 		kfree_skb(skb2);
1540 	return err;
1541 }
1542 #endif
1543 
1544 int br_multicast_rcv(struct net_bridge *br, struct net_bridge_port *port,
1545 		     struct sk_buff *skb)
1546 {
1547 	BR_INPUT_SKB_CB(skb)->igmp = 0;
1548 	BR_INPUT_SKB_CB(skb)->mrouters_only = 0;
1549 
1550 	if (br->multicast_disabled)
1551 		return 0;
1552 
1553 	switch (skb->protocol) {
1554 	case htons(ETH_P_IP):
1555 		return br_multicast_ipv4_rcv(br, port, skb);
1556 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
1557 	case htons(ETH_P_IPV6):
1558 		return br_multicast_ipv6_rcv(br, port, skb);
1559 #endif
1560 	}
1561 
1562 	return 0;
1563 }
1564 
1565 static void br_multicast_query_expired(unsigned long data)
1566 {
1567 	struct net_bridge *br = (void *)data;
1568 
1569 	spin_lock(&br->multicast_lock);
1570 	if (br->multicast_startup_queries_sent <
1571 	    br->multicast_startup_query_count)
1572 		br->multicast_startup_queries_sent++;
1573 
1574 	br_multicast_send_query(br, NULL, br->multicast_startup_queries_sent);
1575 
1576 	spin_unlock(&br->multicast_lock);
1577 }
1578 
1579 void br_multicast_init(struct net_bridge *br)
1580 {
1581 	br->hash_elasticity = 4;
1582 	br->hash_max = 512;
1583 
1584 	br->multicast_router = 1;
1585 	br->multicast_last_member_count = 2;
1586 	br->multicast_startup_query_count = 2;
1587 
1588 	br->multicast_last_member_interval = HZ;
1589 	br->multicast_query_response_interval = 10 * HZ;
1590 	br->multicast_startup_query_interval = 125 * HZ / 4;
1591 	br->multicast_query_interval = 125 * HZ;
1592 	br->multicast_querier_interval = 255 * HZ;
1593 	br->multicast_membership_interval = 260 * HZ;
1594 
1595 	spin_lock_init(&br->multicast_lock);
1596 	setup_timer(&br->multicast_router_timer,
1597 		    br_multicast_local_router_expired, 0);
1598 	setup_timer(&br->multicast_querier_timer,
1599 		    br_multicast_local_router_expired, 0);
1600 	setup_timer(&br->multicast_query_timer, br_multicast_query_expired,
1601 		    (unsigned long)br);
1602 }
1603 
1604 void br_multicast_open(struct net_bridge *br)
1605 {
1606 	br->multicast_startup_queries_sent = 0;
1607 
1608 	if (br->multicast_disabled)
1609 		return;
1610 
1611 	mod_timer(&br->multicast_query_timer, jiffies);
1612 }
1613 
1614 void br_multicast_stop(struct net_bridge *br)
1615 {
1616 	struct net_bridge_mdb_htable *mdb;
1617 	struct net_bridge_mdb_entry *mp;
1618 	struct hlist_node *p, *n;
1619 	u32 ver;
1620 	int i;
1621 
1622 	del_timer_sync(&br->multicast_router_timer);
1623 	del_timer_sync(&br->multicast_querier_timer);
1624 	del_timer_sync(&br->multicast_query_timer);
1625 
1626 	spin_lock_bh(&br->multicast_lock);
1627 	mdb = br->mdb;
1628 	if (!mdb)
1629 		goto out;
1630 
1631 	br->mdb = NULL;
1632 
1633 	ver = mdb->ver;
1634 	for (i = 0; i < mdb->max; i++) {
1635 		hlist_for_each_entry_safe(mp, p, n, &mdb->mhash[i],
1636 					  hlist[ver]) {
1637 			del_timer(&mp->timer);
1638 			del_timer(&mp->query_timer);
1639 			call_rcu_bh(&mp->rcu, br_multicast_free_group);
1640 		}
1641 	}
1642 
1643 	if (mdb->old) {
1644 		spin_unlock_bh(&br->multicast_lock);
1645 		rcu_barrier_bh();
1646 		spin_lock_bh(&br->multicast_lock);
1647 		WARN_ON(mdb->old);
1648 	}
1649 
1650 	mdb->old = mdb;
1651 	call_rcu_bh(&mdb->rcu, br_mdb_free);
1652 
1653 out:
1654 	spin_unlock_bh(&br->multicast_lock);
1655 }
1656 
1657 int br_multicast_set_router(struct net_bridge *br, unsigned long val)
1658 {
1659 	int err = -ENOENT;
1660 
1661 	spin_lock_bh(&br->multicast_lock);
1662 	if (!netif_running(br->dev))
1663 		goto unlock;
1664 
1665 	switch (val) {
1666 	case 0:
1667 	case 2:
1668 		del_timer(&br->multicast_router_timer);
1669 		/* fall through */
1670 	case 1:
1671 		br->multicast_router = val;
1672 		err = 0;
1673 		break;
1674 
1675 	default:
1676 		err = -EINVAL;
1677 		break;
1678 	}
1679 
1680 unlock:
1681 	spin_unlock_bh(&br->multicast_lock);
1682 
1683 	return err;
1684 }
1685 
1686 int br_multicast_set_port_router(struct net_bridge_port *p, unsigned long val)
1687 {
1688 	struct net_bridge *br = p->br;
1689 	int err = -ENOENT;
1690 
1691 	spin_lock(&br->multicast_lock);
1692 	if (!netif_running(br->dev) || p->state == BR_STATE_DISABLED)
1693 		goto unlock;
1694 
1695 	switch (val) {
1696 	case 0:
1697 	case 1:
1698 	case 2:
1699 		p->multicast_router = val;
1700 		err = 0;
1701 
1702 		if (val < 2 && !hlist_unhashed(&p->rlist))
1703 			hlist_del_init_rcu(&p->rlist);
1704 
1705 		if (val == 1)
1706 			break;
1707 
1708 		del_timer(&p->multicast_router_timer);
1709 
1710 		if (val == 0)
1711 			break;
1712 
1713 		br_multicast_add_router(br, p);
1714 		break;
1715 
1716 	default:
1717 		err = -EINVAL;
1718 		break;
1719 	}
1720 
1721 unlock:
1722 	spin_unlock(&br->multicast_lock);
1723 
1724 	return err;
1725 }
1726 
1727 int br_multicast_toggle(struct net_bridge *br, unsigned long val)
1728 {
1729 	struct net_bridge_port *port;
1730 	int err = -ENOENT;
1731 
1732 	spin_lock(&br->multicast_lock);
1733 	if (!netif_running(br->dev))
1734 		goto unlock;
1735 
1736 	err = 0;
1737 	if (br->multicast_disabled == !val)
1738 		goto unlock;
1739 
1740 	br->multicast_disabled = !val;
1741 	if (br->multicast_disabled)
1742 		goto unlock;
1743 
1744 	if (br->mdb) {
1745 		if (br->mdb->old) {
1746 			err = -EEXIST;
1747 rollback:
1748 			br->multicast_disabled = !!val;
1749 			goto unlock;
1750 		}
1751 
1752 		err = br_mdb_rehash(&br->mdb, br->mdb->max,
1753 				    br->hash_elasticity);
1754 		if (err)
1755 			goto rollback;
1756 	}
1757 
1758 	br_multicast_open(br);
1759 	list_for_each_entry(port, &br->port_list, list) {
1760 		if (port->state == BR_STATE_DISABLED ||
1761 		    port->state == BR_STATE_BLOCKING)
1762 			continue;
1763 
1764 		__br_multicast_enable_port(port);
1765 	}
1766 
1767 unlock:
1768 	spin_unlock(&br->multicast_lock);
1769 
1770 	return err;
1771 }
1772 
1773 int br_multicast_set_hash_max(struct net_bridge *br, unsigned long val)
1774 {
1775 	int err = -ENOENT;
1776 	u32 old;
1777 
1778 	spin_lock(&br->multicast_lock);
1779 	if (!netif_running(br->dev))
1780 		goto unlock;
1781 
1782 	err = -EINVAL;
1783 	if (!is_power_of_2(val))
1784 		goto unlock;
1785 	if (br->mdb && val < br->mdb->size)
1786 		goto unlock;
1787 
1788 	err = 0;
1789 
1790 	old = br->hash_max;
1791 	br->hash_max = val;
1792 
1793 	if (br->mdb) {
1794 		if (br->mdb->old) {
1795 			err = -EEXIST;
1796 rollback:
1797 			br->hash_max = old;
1798 			goto unlock;
1799 		}
1800 
1801 		err = br_mdb_rehash(&br->mdb, br->hash_max,
1802 				    br->hash_elasticity);
1803 		if (err)
1804 			goto rollback;
1805 	}
1806 
1807 unlock:
1808 	spin_unlock(&br->multicast_lock);
1809 
1810 	return err;
1811 }
1812