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/export.h>
15#include <linux/if_ether.h>
16#include <linux/igmp.h>
17#include <linux/jhash.h>
18#include <linux/kernel.h>
19#include <linux/log2.h>
20#include <linux/netdevice.h>
21#include <linux/netfilter_bridge.h>
22#include <linux/random.h>
23#include <linux/rculist.h>
24#include <linux/skbuff.h>
25#include <linux/slab.h>
26#include <linux/timer.h>
27#include <linux/inetdevice.h>
28#include <net/ip.h>
29#if IS_ENABLED(CONFIG_IPV6)
30#include <net/ipv6.h>
31#include <net/mld.h>
32#include <net/ip6_checksum.h>
33#include <net/addrconf.h>
34#endif
35
36#include "br_private.h"
37
38static void br_multicast_start_querier(struct net_bridge *br,
39				       struct bridge_mcast_own_query *query);
40static void br_multicast_add_router(struct net_bridge *br,
41				    struct net_bridge_port *port);
42static void br_ip4_multicast_leave_group(struct net_bridge *br,
43					 struct net_bridge_port *port,
44					 __be32 group,
45					 __u16 vid);
46#if IS_ENABLED(CONFIG_IPV6)
47static void br_ip6_multicast_leave_group(struct net_bridge *br,
48					 struct net_bridge_port *port,
49					 const struct in6_addr *group,
50					 __u16 vid);
51#endif
52unsigned int br_mdb_rehash_seq;
53
54static inline int br_ip_equal(const struct br_ip *a, const struct br_ip *b)
55{
56	if (a->proto != b->proto)
57		return 0;
58	if (a->vid != b->vid)
59		return 0;
60	switch (a->proto) {
61	case htons(ETH_P_IP):
62		return a->u.ip4 == b->u.ip4;
63#if IS_ENABLED(CONFIG_IPV6)
64	case htons(ETH_P_IPV6):
65		return ipv6_addr_equal(&a->u.ip6, &b->u.ip6);
66#endif
67	}
68	return 0;
69}
70
71static inline int __br_ip4_hash(struct net_bridge_mdb_htable *mdb, __be32 ip,
72				__u16 vid)
73{
74	return jhash_2words((__force u32)ip, vid, mdb->secret) & (mdb->max - 1);
75}
76
77#if IS_ENABLED(CONFIG_IPV6)
78static inline int __br_ip6_hash(struct net_bridge_mdb_htable *mdb,
79				const struct in6_addr *ip,
80				__u16 vid)
81{
82	return jhash_2words(ipv6_addr_hash(ip), vid,
83			    mdb->secret) & (mdb->max - 1);
84}
85#endif
86
87static inline int br_ip_hash(struct net_bridge_mdb_htable *mdb,
88			     struct br_ip *ip)
89{
90	switch (ip->proto) {
91	case htons(ETH_P_IP):
92		return __br_ip4_hash(mdb, ip->u.ip4, ip->vid);
93#if IS_ENABLED(CONFIG_IPV6)
94	case htons(ETH_P_IPV6):
95		return __br_ip6_hash(mdb, &ip->u.ip6, ip->vid);
96#endif
97	}
98	return 0;
99}
100
101static struct net_bridge_mdb_entry *__br_mdb_ip_get(
102	struct net_bridge_mdb_htable *mdb, struct br_ip *dst, int hash)
103{
104	struct net_bridge_mdb_entry *mp;
105
106	hlist_for_each_entry_rcu(mp, &mdb->mhash[hash], hlist[mdb->ver]) {
107		if (br_ip_equal(&mp->addr, dst))
108			return mp;
109	}
110
111	return NULL;
112}
113
114struct net_bridge_mdb_entry *br_mdb_ip_get(struct net_bridge_mdb_htable *mdb,
115					   struct br_ip *dst)
116{
117	if (!mdb)
118		return NULL;
119
120	return __br_mdb_ip_get(mdb, dst, br_ip_hash(mdb, dst));
121}
122
123static struct net_bridge_mdb_entry *br_mdb_ip4_get(
124	struct net_bridge_mdb_htable *mdb, __be32 dst, __u16 vid)
125{
126	struct br_ip br_dst;
127
128	br_dst.u.ip4 = dst;
129	br_dst.proto = htons(ETH_P_IP);
130	br_dst.vid = vid;
131
132	return br_mdb_ip_get(mdb, &br_dst);
133}
134
135#if IS_ENABLED(CONFIG_IPV6)
136static struct net_bridge_mdb_entry *br_mdb_ip6_get(
137	struct net_bridge_mdb_htable *mdb, const struct in6_addr *dst,
138	__u16 vid)
139{
140	struct br_ip br_dst;
141
142	br_dst.u.ip6 = *dst;
143	br_dst.proto = htons(ETH_P_IPV6);
144	br_dst.vid = vid;
145
146	return br_mdb_ip_get(mdb, &br_dst);
147}
148#endif
149
150struct net_bridge_mdb_entry *br_mdb_get(struct net_bridge *br,
151					struct sk_buff *skb, u16 vid)
152{
153	struct net_bridge_mdb_htable *mdb = rcu_dereference(br->mdb);
154	struct br_ip ip;
155
156	if (br->multicast_disabled)
157		return NULL;
158
159	if (BR_INPUT_SKB_CB(skb)->igmp)
160		return NULL;
161
162	ip.proto = skb->protocol;
163	ip.vid = vid;
164
165	switch (skb->protocol) {
166	case htons(ETH_P_IP):
167		ip.u.ip4 = ip_hdr(skb)->daddr;
168		break;
169#if IS_ENABLED(CONFIG_IPV6)
170	case htons(ETH_P_IPV6):
171		ip.u.ip6 = ipv6_hdr(skb)->daddr;
172		break;
173#endif
174	default:
175		return NULL;
176	}
177
178	return br_mdb_ip_get(mdb, &ip);
179}
180
181static void br_mdb_free(struct rcu_head *head)
182{
183	struct net_bridge_mdb_htable *mdb =
184		container_of(head, struct net_bridge_mdb_htable, rcu);
185	struct net_bridge_mdb_htable *old = mdb->old;
186
187	mdb->old = NULL;
188	kfree(old->mhash);
189	kfree(old);
190}
191
192static int br_mdb_copy(struct net_bridge_mdb_htable *new,
193		       struct net_bridge_mdb_htable *old,
194		       int elasticity)
195{
196	struct net_bridge_mdb_entry *mp;
197	int maxlen;
198	int len;
199	int i;
200
201	for (i = 0; i < old->max; i++)
202		hlist_for_each_entry(mp, &old->mhash[i], hlist[old->ver])
203			hlist_add_head(&mp->hlist[new->ver],
204				       &new->mhash[br_ip_hash(new, &mp->addr)]);
205
206	if (!elasticity)
207		return 0;
208
209	maxlen = 0;
210	for (i = 0; i < new->max; i++) {
211		len = 0;
212		hlist_for_each_entry(mp, &new->mhash[i], hlist[new->ver])
213			len++;
214		if (len > maxlen)
215			maxlen = len;
216	}
217
218	return maxlen > elasticity ? -EINVAL : 0;
219}
220
221void br_multicast_free_pg(struct rcu_head *head)
222{
223	struct net_bridge_port_group *p =
224		container_of(head, struct net_bridge_port_group, rcu);
225
226	kfree(p);
227}
228
229static void br_multicast_free_group(struct rcu_head *head)
230{
231	struct net_bridge_mdb_entry *mp =
232		container_of(head, struct net_bridge_mdb_entry, rcu);
233
234	kfree(mp);
235}
236
237static void br_multicast_group_expired(unsigned long data)
238{
239	struct net_bridge_mdb_entry *mp = (void *)data;
240	struct net_bridge *br = mp->br;
241	struct net_bridge_mdb_htable *mdb;
242
243	spin_lock(&br->multicast_lock);
244	if (!netif_running(br->dev) || timer_pending(&mp->timer))
245		goto out;
246
247	mp->mglist = false;
248
249	if (mp->ports)
250		goto out;
251
252	mdb = mlock_dereference(br->mdb, br);
253
254	hlist_del_rcu(&mp->hlist[mdb->ver]);
255	mdb->size--;
256
257	call_rcu_bh(&mp->rcu, br_multicast_free_group);
258
259out:
260	spin_unlock(&br->multicast_lock);
261}
262
263static void br_multicast_del_pg(struct net_bridge *br,
264				struct net_bridge_port_group *pg)
265{
266	struct net_bridge_mdb_htable *mdb;
267	struct net_bridge_mdb_entry *mp;
268	struct net_bridge_port_group *p;
269	struct net_bridge_port_group __rcu **pp;
270
271	mdb = mlock_dereference(br->mdb, br);
272
273	mp = br_mdb_ip_get(mdb, &pg->addr);
274	if (WARN_ON(!mp))
275		return;
276
277	for (pp = &mp->ports;
278	     (p = mlock_dereference(*pp, br)) != NULL;
279	     pp = &p->next) {
280		if (p != pg)
281			continue;
282
283		rcu_assign_pointer(*pp, p->next);
284		hlist_del_init(&p->mglist);
285		del_timer(&p->timer);
286		br_mdb_notify(br->dev, p->port, &pg->addr, RTM_DELMDB,
287			      p->state);
288		call_rcu_bh(&p->rcu, br_multicast_free_pg);
289
290		if (!mp->ports && !mp->mglist &&
291		    netif_running(br->dev))
292			mod_timer(&mp->timer, jiffies);
293
294		return;
295	}
296
297	WARN_ON(1);
298}
299
300static void br_multicast_port_group_expired(unsigned long data)
301{
302	struct net_bridge_port_group *pg = (void *)data;
303	struct net_bridge *br = pg->port->br;
304
305	spin_lock(&br->multicast_lock);
306	if (!netif_running(br->dev) || timer_pending(&pg->timer) ||
307	    hlist_unhashed(&pg->mglist) || pg->state & MDB_PERMANENT)
308		goto out;
309
310	br_multicast_del_pg(br, pg);
311
312out:
313	spin_unlock(&br->multicast_lock);
314}
315
316static int br_mdb_rehash(struct net_bridge_mdb_htable __rcu **mdbp, int max,
317			 int elasticity)
318{
319	struct net_bridge_mdb_htable *old = rcu_dereference_protected(*mdbp, 1);
320	struct net_bridge_mdb_htable *mdb;
321	int err;
322
323	mdb = kmalloc(sizeof(*mdb), GFP_ATOMIC);
324	if (!mdb)
325		return -ENOMEM;
326
327	mdb->max = max;
328	mdb->old = old;
329
330	mdb->mhash = kzalloc(max * sizeof(*mdb->mhash), GFP_ATOMIC);
331	if (!mdb->mhash) {
332		kfree(mdb);
333		return -ENOMEM;
334	}
335
336	mdb->size = old ? old->size : 0;
337	mdb->ver = old ? old->ver ^ 1 : 0;
338
339	if (!old || elasticity)
340		get_random_bytes(&mdb->secret, sizeof(mdb->secret));
341	else
342		mdb->secret = old->secret;
343
344	if (!old)
345		goto out;
346
347	err = br_mdb_copy(mdb, old, elasticity);
348	if (err) {
349		kfree(mdb->mhash);
350		kfree(mdb);
351		return err;
352	}
353
354	br_mdb_rehash_seq++;
355	call_rcu_bh(&mdb->rcu, br_mdb_free);
356
357out:
358	rcu_assign_pointer(*mdbp, mdb);
359
360	return 0;
361}
362
363static struct sk_buff *br_ip4_multicast_alloc_query(struct net_bridge *br,
364						    __be32 group)
365{
366	struct sk_buff *skb;
367	struct igmphdr *ih;
368	struct ethhdr *eth;
369	struct iphdr *iph;
370
371	skb = netdev_alloc_skb_ip_align(br->dev, sizeof(*eth) + sizeof(*iph) +
372						 sizeof(*ih) + 4);
373	if (!skb)
374		goto out;
375
376	skb->protocol = htons(ETH_P_IP);
377
378	skb_reset_mac_header(skb);
379	eth = eth_hdr(skb);
380
381	ether_addr_copy(eth->h_source, br->dev->dev_addr);
382	eth->h_dest[0] = 1;
383	eth->h_dest[1] = 0;
384	eth->h_dest[2] = 0x5e;
385	eth->h_dest[3] = 0;
386	eth->h_dest[4] = 0;
387	eth->h_dest[5] = 1;
388	eth->h_proto = htons(ETH_P_IP);
389	skb_put(skb, sizeof(*eth));
390
391	skb_set_network_header(skb, skb->len);
392	iph = ip_hdr(skb);
393
394	iph->version = 4;
395	iph->ihl = 6;
396	iph->tos = 0xc0;
397	iph->tot_len = htons(sizeof(*iph) + sizeof(*ih) + 4);
398	iph->id = 0;
399	iph->frag_off = htons(IP_DF);
400	iph->ttl = 1;
401	iph->protocol = IPPROTO_IGMP;
402	iph->saddr = br->multicast_query_use_ifaddr ?
403		     inet_select_addr(br->dev, 0, RT_SCOPE_LINK) : 0;
404	iph->daddr = htonl(INADDR_ALLHOSTS_GROUP);
405	((u8 *)&iph[1])[0] = IPOPT_RA;
406	((u8 *)&iph[1])[1] = 4;
407	((u8 *)&iph[1])[2] = 0;
408	((u8 *)&iph[1])[3] = 0;
409	ip_send_check(iph);
410	skb_put(skb, 24);
411
412	skb_set_transport_header(skb, skb->len);
413	ih = igmp_hdr(skb);
414	ih->type = IGMP_HOST_MEMBERSHIP_QUERY;
415	ih->code = (group ? br->multicast_last_member_interval :
416			    br->multicast_query_response_interval) /
417		   (HZ / IGMP_TIMER_SCALE);
418	ih->group = group;
419	ih->csum = 0;
420	ih->csum = ip_compute_csum((void *)ih, sizeof(struct igmphdr));
421	skb_put(skb, sizeof(*ih));
422
423	__skb_pull(skb, sizeof(*eth));
424
425out:
426	return skb;
427}
428
429#if IS_ENABLED(CONFIG_IPV6)
430static struct sk_buff *br_ip6_multicast_alloc_query(struct net_bridge *br,
431						    const struct in6_addr *group)
432{
433	struct sk_buff *skb;
434	struct ipv6hdr *ip6h;
435	struct mld_msg *mldq;
436	struct ethhdr *eth;
437	u8 *hopopt;
438	unsigned long interval;
439
440	skb = netdev_alloc_skb_ip_align(br->dev, sizeof(*eth) + sizeof(*ip6h) +
441						 8 + sizeof(*mldq));
442	if (!skb)
443		goto out;
444
445	skb->protocol = htons(ETH_P_IPV6);
446
447	/* Ethernet header */
448	skb_reset_mac_header(skb);
449	eth = eth_hdr(skb);
450
451	ether_addr_copy(eth->h_source, br->dev->dev_addr);
452	eth->h_proto = htons(ETH_P_IPV6);
453	skb_put(skb, sizeof(*eth));
454
455	/* IPv6 header + HbH option */
456	skb_set_network_header(skb, skb->len);
457	ip6h = ipv6_hdr(skb);
458
459	*(__force __be32 *)ip6h = htonl(0x60000000);
460	ip6h->payload_len = htons(8 + sizeof(*mldq));
461	ip6h->nexthdr = IPPROTO_HOPOPTS;
462	ip6h->hop_limit = 1;
463	ipv6_addr_set(&ip6h->daddr, htonl(0xff020000), 0, 0, htonl(1));
464	if (ipv6_dev_get_saddr(dev_net(br->dev), br->dev, &ip6h->daddr, 0,
465			       &ip6h->saddr)) {
466		kfree_skb(skb);
467		return NULL;
468	}
469	ipv6_eth_mc_map(&ip6h->daddr, eth->h_dest);
470
471	hopopt = (u8 *)(ip6h + 1);
472	hopopt[0] = IPPROTO_ICMPV6;		/* next hdr */
473	hopopt[1] = 0;				/* length of HbH */
474	hopopt[2] = IPV6_TLV_ROUTERALERT;	/* Router Alert */
475	hopopt[3] = 2;				/* Length of RA Option */
476	hopopt[4] = 0;				/* Type = 0x0000 (MLD) */
477	hopopt[5] = 0;
478	hopopt[6] = IPV6_TLV_PAD1;		/* Pad1 */
479	hopopt[7] = IPV6_TLV_PAD1;		/* Pad1 */
480
481	skb_put(skb, sizeof(*ip6h) + 8);
482
483	/* ICMPv6 */
484	skb_set_transport_header(skb, skb->len);
485	mldq = (struct mld_msg *) icmp6_hdr(skb);
486
487	interval = ipv6_addr_any(group) ?
488			br->multicast_query_response_interval :
489			br->multicast_last_member_interval;
490
491	mldq->mld_type = ICMPV6_MGM_QUERY;
492	mldq->mld_code = 0;
493	mldq->mld_cksum = 0;
494	mldq->mld_maxdelay = htons((u16)jiffies_to_msecs(interval));
495	mldq->mld_reserved = 0;
496	mldq->mld_mca = *group;
497
498	/* checksum */
499	mldq->mld_cksum = csum_ipv6_magic(&ip6h->saddr, &ip6h->daddr,
500					  sizeof(*mldq), IPPROTO_ICMPV6,
501					  csum_partial(mldq,
502						       sizeof(*mldq), 0));
503	skb_put(skb, sizeof(*mldq));
504
505	__skb_pull(skb, sizeof(*eth));
506
507out:
508	return skb;
509}
510#endif
511
512static struct sk_buff *br_multicast_alloc_query(struct net_bridge *br,
513						struct br_ip *addr)
514{
515	switch (addr->proto) {
516	case htons(ETH_P_IP):
517		return br_ip4_multicast_alloc_query(br, addr->u.ip4);
518#if IS_ENABLED(CONFIG_IPV6)
519	case htons(ETH_P_IPV6):
520		return br_ip6_multicast_alloc_query(br, &addr->u.ip6);
521#endif
522	}
523	return NULL;
524}
525
526static struct net_bridge_mdb_entry *br_multicast_get_group(
527	struct net_bridge *br, struct net_bridge_port *port,
528	struct br_ip *group, int hash)
529{
530	struct net_bridge_mdb_htable *mdb;
531	struct net_bridge_mdb_entry *mp;
532	unsigned int count = 0;
533	unsigned int max;
534	int elasticity;
535	int err;
536
537	mdb = rcu_dereference_protected(br->mdb, 1);
538	hlist_for_each_entry(mp, &mdb->mhash[hash], hlist[mdb->ver]) {
539		count++;
540		if (unlikely(br_ip_equal(group, &mp->addr)))
541			return mp;
542	}
543
544	elasticity = 0;
545	max = mdb->max;
546
547	if (unlikely(count > br->hash_elasticity && count)) {
548		if (net_ratelimit())
549			br_info(br, "Multicast hash table "
550				"chain limit reached: %s\n",
551				port ? port->dev->name : br->dev->name);
552
553		elasticity = br->hash_elasticity;
554	}
555
556	if (mdb->size >= max) {
557		max *= 2;
558		if (unlikely(max > br->hash_max)) {
559			br_warn(br, "Multicast hash table maximum of %d "
560				"reached, disabling snooping: %s\n",
561				br->hash_max,
562				port ? port->dev->name : br->dev->name);
563			err = -E2BIG;
564disable:
565			br->multicast_disabled = 1;
566			goto err;
567		}
568	}
569
570	if (max > mdb->max || elasticity) {
571		if (mdb->old) {
572			if (net_ratelimit())
573				br_info(br, "Multicast hash table "
574					"on fire: %s\n",
575					port ? port->dev->name : br->dev->name);
576			err = -EEXIST;
577			goto err;
578		}
579
580		err = br_mdb_rehash(&br->mdb, max, elasticity);
581		if (err) {
582			br_warn(br, "Cannot rehash multicast "
583				"hash table, disabling snooping: %s, %d, %d\n",
584				port ? port->dev->name : br->dev->name,
585				mdb->size, err);
586			goto disable;
587		}
588
589		err = -EAGAIN;
590		goto err;
591	}
592
593	return NULL;
594
595err:
596	mp = ERR_PTR(err);
597	return mp;
598}
599
600struct net_bridge_mdb_entry *br_multicast_new_group(struct net_bridge *br,
601	struct net_bridge_port *port, struct br_ip *group)
602{
603	struct net_bridge_mdb_htable *mdb;
604	struct net_bridge_mdb_entry *mp;
605	int hash;
606	int err;
607
608	mdb = rcu_dereference_protected(br->mdb, 1);
609	if (!mdb) {
610		err = br_mdb_rehash(&br->mdb, BR_HASH_SIZE, 0);
611		if (err)
612			return ERR_PTR(err);
613		goto rehash;
614	}
615
616	hash = br_ip_hash(mdb, group);
617	mp = br_multicast_get_group(br, port, group, hash);
618	switch (PTR_ERR(mp)) {
619	case 0:
620		break;
621
622	case -EAGAIN:
623rehash:
624		mdb = rcu_dereference_protected(br->mdb, 1);
625		hash = br_ip_hash(mdb, group);
626		break;
627
628	default:
629		goto out;
630	}
631
632	mp = kzalloc(sizeof(*mp), GFP_ATOMIC);
633	if (unlikely(!mp))
634		return ERR_PTR(-ENOMEM);
635
636	mp->br = br;
637	mp->addr = *group;
638	setup_timer(&mp->timer, br_multicast_group_expired,
639		    (unsigned long)mp);
640
641	hlist_add_head_rcu(&mp->hlist[mdb->ver], &mdb->mhash[hash]);
642	mdb->size++;
643
644out:
645	return mp;
646}
647
648struct net_bridge_port_group *br_multicast_new_port_group(
649			struct net_bridge_port *port,
650			struct br_ip *group,
651			struct net_bridge_port_group __rcu *next,
652			unsigned char state)
653{
654	struct net_bridge_port_group *p;
655
656	p = kzalloc(sizeof(*p), GFP_ATOMIC);
657	if (unlikely(!p))
658		return NULL;
659
660	p->addr = *group;
661	p->port = port;
662	p->state = state;
663	rcu_assign_pointer(p->next, next);
664	hlist_add_head(&p->mglist, &port->mglist);
665	setup_timer(&p->timer, br_multicast_port_group_expired,
666		    (unsigned long)p);
667	return p;
668}
669
670static int br_multicast_add_group(struct net_bridge *br,
671				  struct net_bridge_port *port,
672				  struct br_ip *group)
673{
674	struct net_bridge_mdb_entry *mp;
675	struct net_bridge_port_group *p;
676	struct net_bridge_port_group __rcu **pp;
677	unsigned long now = jiffies;
678	int err;
679
680	spin_lock(&br->multicast_lock);
681	if (!netif_running(br->dev) ||
682	    (port && port->state == BR_STATE_DISABLED))
683		goto out;
684
685	mp = br_multicast_new_group(br, port, group);
686	err = PTR_ERR(mp);
687	if (IS_ERR(mp))
688		goto err;
689
690	if (!port) {
691		mp->mglist = true;
692		mod_timer(&mp->timer, now + br->multicast_membership_interval);
693		goto out;
694	}
695
696	for (pp = &mp->ports;
697	     (p = mlock_dereference(*pp, br)) != NULL;
698	     pp = &p->next) {
699		if (p->port == port)
700			goto found;
701		if ((unsigned long)p->port < (unsigned long)port)
702			break;
703	}
704
705	p = br_multicast_new_port_group(port, group, *pp, MDB_TEMPORARY);
706	if (unlikely(!p))
707		goto err;
708	rcu_assign_pointer(*pp, p);
709	br_mdb_notify(br->dev, port, group, RTM_NEWMDB, MDB_TEMPORARY);
710
711found:
712	mod_timer(&p->timer, now + br->multicast_membership_interval);
713out:
714	err = 0;
715
716err:
717	spin_unlock(&br->multicast_lock);
718	return err;
719}
720
721static int br_ip4_multicast_add_group(struct net_bridge *br,
722				      struct net_bridge_port *port,
723				      __be32 group,
724				      __u16 vid)
725{
726	struct br_ip br_group;
727
728	if (ipv4_is_local_multicast(group))
729		return 0;
730
731	br_group.u.ip4 = group;
732	br_group.proto = htons(ETH_P_IP);
733	br_group.vid = vid;
734
735	return br_multicast_add_group(br, port, &br_group);
736}
737
738#if IS_ENABLED(CONFIG_IPV6)
739static int br_ip6_multicast_add_group(struct net_bridge *br,
740				      struct net_bridge_port *port,
741				      const struct in6_addr *group,
742				      __u16 vid)
743{
744	struct br_ip br_group;
745
746	if (ipv6_addr_is_ll_all_nodes(group))
747		return 0;
748
749	br_group.u.ip6 = *group;
750	br_group.proto = htons(ETH_P_IPV6);
751	br_group.vid = vid;
752
753	return br_multicast_add_group(br, port, &br_group);
754}
755#endif
756
757static void br_multicast_router_expired(unsigned long data)
758{
759	struct net_bridge_port *port = (void *)data;
760	struct net_bridge *br = port->br;
761
762	spin_lock(&br->multicast_lock);
763	if (port->multicast_router != 1 ||
764	    timer_pending(&port->multicast_router_timer) ||
765	    hlist_unhashed(&port->rlist))
766		goto out;
767
768	hlist_del_init_rcu(&port->rlist);
769	br_rtr_notify(br->dev, port, RTM_DELMDB);
770
771out:
772	spin_unlock(&br->multicast_lock);
773}
774
775static void br_multicast_local_router_expired(unsigned long data)
776{
777}
778
779static void br_multicast_querier_expired(struct net_bridge *br,
780					 struct bridge_mcast_own_query *query)
781{
782	spin_lock(&br->multicast_lock);
783	if (!netif_running(br->dev) || br->multicast_disabled)
784		goto out;
785
786	br_multicast_start_querier(br, query);
787
788out:
789	spin_unlock(&br->multicast_lock);
790}
791
792static void br_ip4_multicast_querier_expired(unsigned long data)
793{
794	struct net_bridge *br = (void *)data;
795
796	br_multicast_querier_expired(br, &br->ip4_own_query);
797}
798
799#if IS_ENABLED(CONFIG_IPV6)
800static void br_ip6_multicast_querier_expired(unsigned long data)
801{
802	struct net_bridge *br = (void *)data;
803
804	br_multicast_querier_expired(br, &br->ip6_own_query);
805}
806#endif
807
808static void br_multicast_select_own_querier(struct net_bridge *br,
809					    struct br_ip *ip,
810					    struct sk_buff *skb)
811{
812	if (ip->proto == htons(ETH_P_IP))
813		br->ip4_querier.addr.u.ip4 = ip_hdr(skb)->saddr;
814#if IS_ENABLED(CONFIG_IPV6)
815	else
816		br->ip6_querier.addr.u.ip6 = ipv6_hdr(skb)->saddr;
817#endif
818}
819
820static void __br_multicast_send_query(struct net_bridge *br,
821				      struct net_bridge_port *port,
822				      struct br_ip *ip)
823{
824	struct sk_buff *skb;
825
826	skb = br_multicast_alloc_query(br, ip);
827	if (!skb)
828		return;
829
830	if (port) {
831		skb->dev = port->dev;
832		NF_HOOK(NFPROTO_BRIDGE, NF_BR_LOCAL_OUT,
833			dev_net(port->dev), NULL, skb, NULL, skb->dev,
834			br_dev_queue_push_xmit);
835	} else {
836		br_multicast_select_own_querier(br, ip, skb);
837		netif_rx(skb);
838	}
839}
840
841static void br_multicast_send_query(struct net_bridge *br,
842				    struct net_bridge_port *port,
843				    struct bridge_mcast_own_query *own_query)
844{
845	unsigned long time;
846	struct br_ip br_group;
847	struct bridge_mcast_other_query *other_query = NULL;
848
849	if (!netif_running(br->dev) || br->multicast_disabled ||
850	    !br->multicast_querier)
851		return;
852
853	memset(&br_group.u, 0, sizeof(br_group.u));
854
855	if (port ? (own_query == &port->ip4_own_query) :
856		   (own_query == &br->ip4_own_query)) {
857		other_query = &br->ip4_other_query;
858		br_group.proto = htons(ETH_P_IP);
859#if IS_ENABLED(CONFIG_IPV6)
860	} else {
861		other_query = &br->ip6_other_query;
862		br_group.proto = htons(ETH_P_IPV6);
863#endif
864	}
865
866	if (!other_query || timer_pending(&other_query->timer))
867		return;
868
869	__br_multicast_send_query(br, port, &br_group);
870
871	time = jiffies;
872	time += own_query->startup_sent < br->multicast_startup_query_count ?
873		br->multicast_startup_query_interval :
874		br->multicast_query_interval;
875	mod_timer(&own_query->timer, time);
876}
877
878static void
879br_multicast_port_query_expired(struct net_bridge_port *port,
880				struct bridge_mcast_own_query *query)
881{
882	struct net_bridge *br = port->br;
883
884	spin_lock(&br->multicast_lock);
885	if (port->state == BR_STATE_DISABLED ||
886	    port->state == BR_STATE_BLOCKING)
887		goto out;
888
889	if (query->startup_sent < br->multicast_startup_query_count)
890		query->startup_sent++;
891
892	br_multicast_send_query(port->br, port, query);
893
894out:
895	spin_unlock(&br->multicast_lock);
896}
897
898static void br_ip4_multicast_port_query_expired(unsigned long data)
899{
900	struct net_bridge_port *port = (void *)data;
901
902	br_multicast_port_query_expired(port, &port->ip4_own_query);
903}
904
905#if IS_ENABLED(CONFIG_IPV6)
906static void br_ip6_multicast_port_query_expired(unsigned long data)
907{
908	struct net_bridge_port *port = (void *)data;
909
910	br_multicast_port_query_expired(port, &port->ip6_own_query);
911}
912#endif
913
914void br_multicast_add_port(struct net_bridge_port *port)
915{
916	port->multicast_router = 1;
917
918	setup_timer(&port->multicast_router_timer, br_multicast_router_expired,
919		    (unsigned long)port);
920	setup_timer(&port->ip4_own_query.timer,
921		    br_ip4_multicast_port_query_expired, (unsigned long)port);
922#if IS_ENABLED(CONFIG_IPV6)
923	setup_timer(&port->ip6_own_query.timer,
924		    br_ip6_multicast_port_query_expired, (unsigned long)port);
925#endif
926}
927
928void br_multicast_del_port(struct net_bridge_port *port)
929{
930	struct net_bridge *br = port->br;
931	struct net_bridge_port_group *pg;
932	struct hlist_node *n;
933
934	/* Take care of the remaining groups, only perm ones should be left */
935	spin_lock_bh(&br->multicast_lock);
936	hlist_for_each_entry_safe(pg, n, &port->mglist, mglist)
937		br_multicast_del_pg(br, pg);
938	spin_unlock_bh(&br->multicast_lock);
939	del_timer_sync(&port->multicast_router_timer);
940}
941
942static void br_multicast_enable(struct bridge_mcast_own_query *query)
943{
944	query->startup_sent = 0;
945
946	if (try_to_del_timer_sync(&query->timer) >= 0 ||
947	    del_timer(&query->timer))
948		mod_timer(&query->timer, jiffies);
949}
950
951void br_multicast_enable_port(struct net_bridge_port *port)
952{
953	struct net_bridge *br = port->br;
954
955	spin_lock(&br->multicast_lock);
956	if (br->multicast_disabled || !netif_running(br->dev))
957		goto out;
958
959	br_multicast_enable(&port->ip4_own_query);
960#if IS_ENABLED(CONFIG_IPV6)
961	br_multicast_enable(&port->ip6_own_query);
962#endif
963	if (port->multicast_router == 2 && hlist_unhashed(&port->rlist))
964		br_multicast_add_router(br, port);
965
966out:
967	spin_unlock(&br->multicast_lock);
968}
969
970void br_multicast_disable_port(struct net_bridge_port *port)
971{
972	struct net_bridge *br = port->br;
973	struct net_bridge_port_group *pg;
974	struct hlist_node *n;
975
976	spin_lock(&br->multicast_lock);
977	hlist_for_each_entry_safe(pg, n, &port->mglist, mglist)
978		if (pg->state == MDB_TEMPORARY)
979			br_multicast_del_pg(br, pg);
980
981	if (!hlist_unhashed(&port->rlist)) {
982		hlist_del_init_rcu(&port->rlist);
983		br_rtr_notify(br->dev, port, RTM_DELMDB);
984	}
985	del_timer(&port->multicast_router_timer);
986	del_timer(&port->ip4_own_query.timer);
987#if IS_ENABLED(CONFIG_IPV6)
988	del_timer(&port->ip6_own_query.timer);
989#endif
990	spin_unlock(&br->multicast_lock);
991}
992
993static int br_ip4_multicast_igmp3_report(struct net_bridge *br,
994					 struct net_bridge_port *port,
995					 struct sk_buff *skb,
996					 u16 vid)
997{
998	struct igmpv3_report *ih;
999	struct igmpv3_grec *grec;
1000	int i;
1001	int len;
1002	int num;
1003	int type;
1004	int err = 0;
1005	__be32 group;
1006
1007	ih = igmpv3_report_hdr(skb);
1008	num = ntohs(ih->ngrec);
1009	len = skb_transport_offset(skb) + sizeof(*ih);
1010
1011	for (i = 0; i < num; i++) {
1012		len += sizeof(*grec);
1013		if (!pskb_may_pull(skb, len))
1014			return -EINVAL;
1015
1016		grec = (void *)(skb->data + len - sizeof(*grec));
1017		group = grec->grec_mca;
1018		type = grec->grec_type;
1019
1020		len += ntohs(grec->grec_nsrcs) * 4;
1021		if (!pskb_may_pull(skb, len))
1022			return -EINVAL;
1023
1024		/* We treat this as an IGMPv2 report for now. */
1025		switch (type) {
1026		case IGMPV3_MODE_IS_INCLUDE:
1027		case IGMPV3_MODE_IS_EXCLUDE:
1028		case IGMPV3_CHANGE_TO_INCLUDE:
1029		case IGMPV3_CHANGE_TO_EXCLUDE:
1030		case IGMPV3_ALLOW_NEW_SOURCES:
1031		case IGMPV3_BLOCK_OLD_SOURCES:
1032			break;
1033
1034		default:
1035			continue;
1036		}
1037
1038		if ((type == IGMPV3_CHANGE_TO_INCLUDE ||
1039		     type == IGMPV3_MODE_IS_INCLUDE) &&
1040		    ntohs(grec->grec_nsrcs) == 0) {
1041			br_ip4_multicast_leave_group(br, port, group, vid);
1042		} else {
1043			err = br_ip4_multicast_add_group(br, port, group, vid);
1044			if (err)
1045				break;
1046		}
1047	}
1048
1049	return err;
1050}
1051
1052#if IS_ENABLED(CONFIG_IPV6)
1053static int br_ip6_multicast_mld2_report(struct net_bridge *br,
1054					struct net_bridge_port *port,
1055					struct sk_buff *skb,
1056					u16 vid)
1057{
1058	struct icmp6hdr *icmp6h;
1059	struct mld2_grec *grec;
1060	int i;
1061	int len;
1062	int num;
1063	int err = 0;
1064
1065	if (!pskb_may_pull(skb, sizeof(*icmp6h)))
1066		return -EINVAL;
1067
1068	icmp6h = icmp6_hdr(skb);
1069	num = ntohs(icmp6h->icmp6_dataun.un_data16[1]);
1070	len = skb_transport_offset(skb) + sizeof(*icmp6h);
1071
1072	for (i = 0; i < num; i++) {
1073		__be16 *nsrcs, _nsrcs;
1074
1075		nsrcs = skb_header_pointer(skb,
1076					   len + offsetof(struct mld2_grec,
1077							  grec_nsrcs),
1078					   sizeof(_nsrcs), &_nsrcs);
1079		if (!nsrcs)
1080			return -EINVAL;
1081
1082		if (!pskb_may_pull(skb,
1083				   len + sizeof(*grec) +
1084				   sizeof(struct in6_addr) * ntohs(*nsrcs)))
1085			return -EINVAL;
1086
1087		grec = (struct mld2_grec *)(skb->data + len);
1088		len += sizeof(*grec) +
1089		       sizeof(struct in6_addr) * ntohs(*nsrcs);
1090
1091		/* We treat these as MLDv1 reports for now. */
1092		switch (grec->grec_type) {
1093		case MLD2_MODE_IS_INCLUDE:
1094		case MLD2_MODE_IS_EXCLUDE:
1095		case MLD2_CHANGE_TO_INCLUDE:
1096		case MLD2_CHANGE_TO_EXCLUDE:
1097		case MLD2_ALLOW_NEW_SOURCES:
1098		case MLD2_BLOCK_OLD_SOURCES:
1099			break;
1100
1101		default:
1102			continue;
1103		}
1104
1105		if ((grec->grec_type == MLD2_CHANGE_TO_INCLUDE ||
1106		     grec->grec_type == MLD2_MODE_IS_INCLUDE) &&
1107		    ntohs(*nsrcs) == 0) {
1108			br_ip6_multicast_leave_group(br, port, &grec->grec_mca,
1109						     vid);
1110		} else {
1111			err = br_ip6_multicast_add_group(br, port,
1112							 &grec->grec_mca, vid);
1113			if (!err)
1114				break;
1115		}
1116	}
1117
1118	return err;
1119}
1120#endif
1121
1122static bool br_ip4_multicast_select_querier(struct net_bridge *br,
1123					    struct net_bridge_port *port,
1124					    __be32 saddr)
1125{
1126	if (!timer_pending(&br->ip4_own_query.timer) &&
1127	    !timer_pending(&br->ip4_other_query.timer))
1128		goto update;
1129
1130	if (!br->ip4_querier.addr.u.ip4)
1131		goto update;
1132
1133	if (ntohl(saddr) <= ntohl(br->ip4_querier.addr.u.ip4))
1134		goto update;
1135
1136	return false;
1137
1138update:
1139	br->ip4_querier.addr.u.ip4 = saddr;
1140
1141	/* update protected by general multicast_lock by caller */
1142	rcu_assign_pointer(br->ip4_querier.port, port);
1143
1144	return true;
1145}
1146
1147#if IS_ENABLED(CONFIG_IPV6)
1148static bool br_ip6_multicast_select_querier(struct net_bridge *br,
1149					    struct net_bridge_port *port,
1150					    struct in6_addr *saddr)
1151{
1152	if (!timer_pending(&br->ip6_own_query.timer) &&
1153	    !timer_pending(&br->ip6_other_query.timer))
1154		goto update;
1155
1156	if (ipv6_addr_cmp(saddr, &br->ip6_querier.addr.u.ip6) <= 0)
1157		goto update;
1158
1159	return false;
1160
1161update:
1162	br->ip6_querier.addr.u.ip6 = *saddr;
1163
1164	/* update protected by general multicast_lock by caller */
1165	rcu_assign_pointer(br->ip6_querier.port, port);
1166
1167	return true;
1168}
1169#endif
1170
1171static bool br_multicast_select_querier(struct net_bridge *br,
1172					struct net_bridge_port *port,
1173					struct br_ip *saddr)
1174{
1175	switch (saddr->proto) {
1176	case htons(ETH_P_IP):
1177		return br_ip4_multicast_select_querier(br, port, saddr->u.ip4);
1178#if IS_ENABLED(CONFIG_IPV6)
1179	case htons(ETH_P_IPV6):
1180		return br_ip6_multicast_select_querier(br, port, &saddr->u.ip6);
1181#endif
1182	}
1183
1184	return false;
1185}
1186
1187static void
1188br_multicast_update_query_timer(struct net_bridge *br,
1189				struct bridge_mcast_other_query *query,
1190				unsigned long max_delay)
1191{
1192	if (!timer_pending(&query->timer))
1193		query->delay_time = jiffies + max_delay;
1194
1195	mod_timer(&query->timer, jiffies + br->multicast_querier_interval);
1196}
1197
1198/*
1199 * Add port to router_list
1200 *  list is maintained ordered by pointer value
1201 *  and locked by br->multicast_lock and RCU
1202 */
1203static void br_multicast_add_router(struct net_bridge *br,
1204				    struct net_bridge_port *port)
1205{
1206	struct net_bridge_port *p;
1207	struct hlist_node *slot = NULL;
1208
1209	if (!hlist_unhashed(&port->rlist))
1210		return;
1211
1212	hlist_for_each_entry(p, &br->router_list, rlist) {
1213		if ((unsigned long) port >= (unsigned long) p)
1214			break;
1215		slot = &p->rlist;
1216	}
1217
1218	if (slot)
1219		hlist_add_behind_rcu(&port->rlist, slot);
1220	else
1221		hlist_add_head_rcu(&port->rlist, &br->router_list);
1222	br_rtr_notify(br->dev, port, RTM_NEWMDB);
1223}
1224
1225static void br_multicast_mark_router(struct net_bridge *br,
1226				     struct net_bridge_port *port)
1227{
1228	unsigned long now = jiffies;
1229
1230	if (!port) {
1231		if (br->multicast_router == 1)
1232			mod_timer(&br->multicast_router_timer,
1233				  now + br->multicast_querier_interval);
1234		return;
1235	}
1236
1237	if (port->multicast_router != 1)
1238		return;
1239
1240	br_multicast_add_router(br, port);
1241
1242	mod_timer(&port->multicast_router_timer,
1243		  now + br->multicast_querier_interval);
1244}
1245
1246static void br_multicast_query_received(struct net_bridge *br,
1247					struct net_bridge_port *port,
1248					struct bridge_mcast_other_query *query,
1249					struct br_ip *saddr,
1250					unsigned long max_delay)
1251{
1252	if (!br_multicast_select_querier(br, port, saddr))
1253		return;
1254
1255	br_multicast_update_query_timer(br, query, max_delay);
1256	br_multicast_mark_router(br, port);
1257}
1258
1259static int br_ip4_multicast_query(struct net_bridge *br,
1260				  struct net_bridge_port *port,
1261				  struct sk_buff *skb,
1262				  u16 vid)
1263{
1264	const struct iphdr *iph = ip_hdr(skb);
1265	struct igmphdr *ih = igmp_hdr(skb);
1266	struct net_bridge_mdb_entry *mp;
1267	struct igmpv3_query *ih3;
1268	struct net_bridge_port_group *p;
1269	struct net_bridge_port_group __rcu **pp;
1270	struct br_ip saddr;
1271	unsigned long max_delay;
1272	unsigned long now = jiffies;
1273	unsigned int offset = skb_transport_offset(skb);
1274	__be32 group;
1275	int err = 0;
1276
1277	spin_lock(&br->multicast_lock);
1278	if (!netif_running(br->dev) ||
1279	    (port && port->state == BR_STATE_DISABLED))
1280		goto out;
1281
1282	group = ih->group;
1283
1284	if (skb->len == offset + sizeof(*ih)) {
1285		max_delay = ih->code * (HZ / IGMP_TIMER_SCALE);
1286
1287		if (!max_delay) {
1288			max_delay = 10 * HZ;
1289			group = 0;
1290		}
1291	} else if (skb->len >= offset + sizeof(*ih3)) {
1292		ih3 = igmpv3_query_hdr(skb);
1293		if (ih3->nsrcs)
1294			goto out;
1295
1296		max_delay = ih3->code ?
1297			    IGMPV3_MRC(ih3->code) * (HZ / IGMP_TIMER_SCALE) : 1;
1298	} else {
1299		goto out;
1300	}
1301
1302	if (!group) {
1303		saddr.proto = htons(ETH_P_IP);
1304		saddr.u.ip4 = iph->saddr;
1305
1306		br_multicast_query_received(br, port, &br->ip4_other_query,
1307					    &saddr, max_delay);
1308		goto out;
1309	}
1310
1311	mp = br_mdb_ip4_get(mlock_dereference(br->mdb, br), group, vid);
1312	if (!mp)
1313		goto out;
1314
1315	max_delay *= br->multicast_last_member_count;
1316
1317	if (mp->mglist &&
1318	    (timer_pending(&mp->timer) ?
1319	     time_after(mp->timer.expires, now + max_delay) :
1320	     try_to_del_timer_sync(&mp->timer) >= 0))
1321		mod_timer(&mp->timer, now + max_delay);
1322
1323	for (pp = &mp->ports;
1324	     (p = mlock_dereference(*pp, br)) != NULL;
1325	     pp = &p->next) {
1326		if (timer_pending(&p->timer) ?
1327		    time_after(p->timer.expires, now + max_delay) :
1328		    try_to_del_timer_sync(&p->timer) >= 0)
1329			mod_timer(&p->timer, now + max_delay);
1330	}
1331
1332out:
1333	spin_unlock(&br->multicast_lock);
1334	return err;
1335}
1336
1337#if IS_ENABLED(CONFIG_IPV6)
1338static int br_ip6_multicast_query(struct net_bridge *br,
1339				  struct net_bridge_port *port,
1340				  struct sk_buff *skb,
1341				  u16 vid)
1342{
1343	const struct ipv6hdr *ip6h = ipv6_hdr(skb);
1344	struct mld_msg *mld;
1345	struct net_bridge_mdb_entry *mp;
1346	struct mld2_query *mld2q;
1347	struct net_bridge_port_group *p;
1348	struct net_bridge_port_group __rcu **pp;
1349	struct br_ip saddr;
1350	unsigned long max_delay;
1351	unsigned long now = jiffies;
1352	unsigned int offset = skb_transport_offset(skb);
1353	const struct in6_addr *group = NULL;
1354	bool is_general_query;
1355	int err = 0;
1356
1357	spin_lock(&br->multicast_lock);
1358	if (!netif_running(br->dev) ||
1359	    (port && port->state == BR_STATE_DISABLED))
1360		goto out;
1361
1362	if (skb->len == offset + sizeof(*mld)) {
1363		if (!pskb_may_pull(skb, offset + sizeof(*mld))) {
1364			err = -EINVAL;
1365			goto out;
1366		}
1367		mld = (struct mld_msg *) icmp6_hdr(skb);
1368		max_delay = msecs_to_jiffies(ntohs(mld->mld_maxdelay));
1369		if (max_delay)
1370			group = &mld->mld_mca;
1371	} else {
1372		if (!pskb_may_pull(skb, offset + sizeof(*mld2q))) {
1373			err = -EINVAL;
1374			goto out;
1375		}
1376		mld2q = (struct mld2_query *)icmp6_hdr(skb);
1377		if (!mld2q->mld2q_nsrcs)
1378			group = &mld2q->mld2q_mca;
1379
1380		max_delay = max(msecs_to_jiffies(mldv2_mrc(mld2q)), 1UL);
1381	}
1382
1383	is_general_query = group && ipv6_addr_any(group);
1384
1385	if (is_general_query) {
1386		saddr.proto = htons(ETH_P_IPV6);
1387		saddr.u.ip6 = ip6h->saddr;
1388
1389		br_multicast_query_received(br, port, &br->ip6_other_query,
1390					    &saddr, max_delay);
1391		goto out;
1392	} else if (!group) {
1393		goto out;
1394	}
1395
1396	mp = br_mdb_ip6_get(mlock_dereference(br->mdb, br), group, vid);
1397	if (!mp)
1398		goto out;
1399
1400	max_delay *= br->multicast_last_member_count;
1401	if (mp->mglist &&
1402	    (timer_pending(&mp->timer) ?
1403	     time_after(mp->timer.expires, now + max_delay) :
1404	     try_to_del_timer_sync(&mp->timer) >= 0))
1405		mod_timer(&mp->timer, now + max_delay);
1406
1407	for (pp = &mp->ports;
1408	     (p = mlock_dereference(*pp, br)) != NULL;
1409	     pp = &p->next) {
1410		if (timer_pending(&p->timer) ?
1411		    time_after(p->timer.expires, now + max_delay) :
1412		    try_to_del_timer_sync(&p->timer) >= 0)
1413			mod_timer(&p->timer, now + max_delay);
1414	}
1415
1416out:
1417	spin_unlock(&br->multicast_lock);
1418	return err;
1419}
1420#endif
1421
1422static void
1423br_multicast_leave_group(struct net_bridge *br,
1424			 struct net_bridge_port *port,
1425			 struct br_ip *group,
1426			 struct bridge_mcast_other_query *other_query,
1427			 struct bridge_mcast_own_query *own_query)
1428{
1429	struct net_bridge_mdb_htable *mdb;
1430	struct net_bridge_mdb_entry *mp;
1431	struct net_bridge_port_group *p;
1432	unsigned long now;
1433	unsigned long time;
1434
1435	spin_lock(&br->multicast_lock);
1436	if (!netif_running(br->dev) ||
1437	    (port && port->state == BR_STATE_DISABLED))
1438		goto out;
1439
1440	mdb = mlock_dereference(br->mdb, br);
1441	mp = br_mdb_ip_get(mdb, group);
1442	if (!mp)
1443		goto out;
1444
1445	if (port && (port->flags & BR_MULTICAST_FAST_LEAVE)) {
1446		struct net_bridge_port_group __rcu **pp;
1447
1448		for (pp = &mp->ports;
1449		     (p = mlock_dereference(*pp, br)) != NULL;
1450		     pp = &p->next) {
1451			if (p->port != port)
1452				continue;
1453
1454			rcu_assign_pointer(*pp, p->next);
1455			hlist_del_init(&p->mglist);
1456			del_timer(&p->timer);
1457			call_rcu_bh(&p->rcu, br_multicast_free_pg);
1458			br_mdb_notify(br->dev, port, group, RTM_DELMDB,
1459				      p->state);
1460
1461			if (!mp->ports && !mp->mglist &&
1462			    netif_running(br->dev))
1463				mod_timer(&mp->timer, jiffies);
1464		}
1465		goto out;
1466	}
1467
1468	if (timer_pending(&other_query->timer))
1469		goto out;
1470
1471	if (br->multicast_querier) {
1472		__br_multicast_send_query(br, port, &mp->addr);
1473
1474		time = jiffies + br->multicast_last_member_count *
1475				 br->multicast_last_member_interval;
1476
1477		mod_timer(&own_query->timer, time);
1478
1479		for (p = mlock_dereference(mp->ports, br);
1480		     p != NULL;
1481		     p = mlock_dereference(p->next, br)) {
1482			if (p->port != port)
1483				continue;
1484
1485			if (!hlist_unhashed(&p->mglist) &&
1486			    (timer_pending(&p->timer) ?
1487			     time_after(p->timer.expires, time) :
1488			     try_to_del_timer_sync(&p->timer) >= 0)) {
1489				mod_timer(&p->timer, time);
1490			}
1491
1492			break;
1493		}
1494	}
1495
1496	now = jiffies;
1497	time = now + br->multicast_last_member_count *
1498		     br->multicast_last_member_interval;
1499
1500	if (!port) {
1501		if (mp->mglist &&
1502		    (timer_pending(&mp->timer) ?
1503		     time_after(mp->timer.expires, time) :
1504		     try_to_del_timer_sync(&mp->timer) >= 0)) {
1505			mod_timer(&mp->timer, time);
1506		}
1507
1508		goto out;
1509	}
1510
1511	for (p = mlock_dereference(mp->ports, br);
1512	     p != NULL;
1513	     p = mlock_dereference(p->next, br)) {
1514		if (p->port != port)
1515			continue;
1516
1517		if (!hlist_unhashed(&p->mglist) &&
1518		    (timer_pending(&p->timer) ?
1519		     time_after(p->timer.expires, time) :
1520		     try_to_del_timer_sync(&p->timer) >= 0)) {
1521			mod_timer(&p->timer, time);
1522		}
1523
1524		break;
1525	}
1526out:
1527	spin_unlock(&br->multicast_lock);
1528}
1529
1530static void br_ip4_multicast_leave_group(struct net_bridge *br,
1531					 struct net_bridge_port *port,
1532					 __be32 group,
1533					 __u16 vid)
1534{
1535	struct br_ip br_group;
1536	struct bridge_mcast_own_query *own_query;
1537
1538	if (ipv4_is_local_multicast(group))
1539		return;
1540
1541	own_query = port ? &port->ip4_own_query : &br->ip4_own_query;
1542
1543	br_group.u.ip4 = group;
1544	br_group.proto = htons(ETH_P_IP);
1545	br_group.vid = vid;
1546
1547	br_multicast_leave_group(br, port, &br_group, &br->ip4_other_query,
1548				 own_query);
1549}
1550
1551#if IS_ENABLED(CONFIG_IPV6)
1552static void br_ip6_multicast_leave_group(struct net_bridge *br,
1553					 struct net_bridge_port *port,
1554					 const struct in6_addr *group,
1555					 __u16 vid)
1556{
1557	struct br_ip br_group;
1558	struct bridge_mcast_own_query *own_query;
1559
1560	if (ipv6_addr_is_ll_all_nodes(group))
1561		return;
1562
1563	own_query = port ? &port->ip6_own_query : &br->ip6_own_query;
1564
1565	br_group.u.ip6 = *group;
1566	br_group.proto = htons(ETH_P_IPV6);
1567	br_group.vid = vid;
1568
1569	br_multicast_leave_group(br, port, &br_group, &br->ip6_other_query,
1570				 own_query);
1571}
1572#endif
1573
1574static int br_multicast_ipv4_rcv(struct net_bridge *br,
1575				 struct net_bridge_port *port,
1576				 struct sk_buff *skb,
1577				 u16 vid)
1578{
1579	struct sk_buff *skb_trimmed = NULL;
1580	struct igmphdr *ih;
1581	int err;
1582
1583	err = ip_mc_check_igmp(skb, &skb_trimmed);
1584
1585	if (err == -ENOMSG) {
1586		if (!ipv4_is_local_multicast(ip_hdr(skb)->daddr))
1587			BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1588		return 0;
1589	} else if (err < 0) {
1590		return err;
1591	}
1592
1593	BR_INPUT_SKB_CB(skb)->igmp = 1;
1594	ih = igmp_hdr(skb);
1595
1596	switch (ih->type) {
1597	case IGMP_HOST_MEMBERSHIP_REPORT:
1598	case IGMPV2_HOST_MEMBERSHIP_REPORT:
1599		BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1600		err = br_ip4_multicast_add_group(br, port, ih->group, vid);
1601		break;
1602	case IGMPV3_HOST_MEMBERSHIP_REPORT:
1603		err = br_ip4_multicast_igmp3_report(br, port, skb_trimmed, vid);
1604		break;
1605	case IGMP_HOST_MEMBERSHIP_QUERY:
1606		err = br_ip4_multicast_query(br, port, skb_trimmed, vid);
1607		break;
1608	case IGMP_HOST_LEAVE_MESSAGE:
1609		br_ip4_multicast_leave_group(br, port, ih->group, vid);
1610		break;
1611	}
1612
1613	if (skb_trimmed && skb_trimmed != skb)
1614		kfree_skb(skb_trimmed);
1615
1616	return err;
1617}
1618
1619#if IS_ENABLED(CONFIG_IPV6)
1620static int br_multicast_ipv6_rcv(struct net_bridge *br,
1621				 struct net_bridge_port *port,
1622				 struct sk_buff *skb,
1623				 u16 vid)
1624{
1625	struct sk_buff *skb_trimmed = NULL;
1626	struct mld_msg *mld;
1627	int err;
1628
1629	err = ipv6_mc_check_mld(skb, &skb_trimmed);
1630
1631	if (err == -ENOMSG) {
1632		if (!ipv6_addr_is_ll_all_nodes(&ipv6_hdr(skb)->daddr))
1633			BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1634		return 0;
1635	} else if (err < 0) {
1636		return err;
1637	}
1638
1639	BR_INPUT_SKB_CB(skb)->igmp = 1;
1640	mld = (struct mld_msg *)skb_transport_header(skb);
1641
1642	switch (mld->mld_type) {
1643	case ICMPV6_MGM_REPORT:
1644		BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1645		err = br_ip6_multicast_add_group(br, port, &mld->mld_mca, vid);
1646		break;
1647	case ICMPV6_MLD2_REPORT:
1648		err = br_ip6_multicast_mld2_report(br, port, skb_trimmed, vid);
1649		break;
1650	case ICMPV6_MGM_QUERY:
1651		err = br_ip6_multicast_query(br, port, skb_trimmed, vid);
1652		break;
1653	case ICMPV6_MGM_REDUCTION:
1654		br_ip6_multicast_leave_group(br, port, &mld->mld_mca, vid);
1655		break;
1656	}
1657
1658	if (skb_trimmed && skb_trimmed != skb)
1659		kfree_skb(skb_trimmed);
1660
1661	return err;
1662}
1663#endif
1664
1665int br_multicast_rcv(struct net_bridge *br, struct net_bridge_port *port,
1666		     struct sk_buff *skb, u16 vid)
1667{
1668	BR_INPUT_SKB_CB(skb)->igmp = 0;
1669	BR_INPUT_SKB_CB(skb)->mrouters_only = 0;
1670
1671	if (br->multicast_disabled)
1672		return 0;
1673
1674	switch (skb->protocol) {
1675	case htons(ETH_P_IP):
1676		return br_multicast_ipv4_rcv(br, port, skb, vid);
1677#if IS_ENABLED(CONFIG_IPV6)
1678	case htons(ETH_P_IPV6):
1679		return br_multicast_ipv6_rcv(br, port, skb, vid);
1680#endif
1681	}
1682
1683	return 0;
1684}
1685
1686static void br_multicast_query_expired(struct net_bridge *br,
1687				       struct bridge_mcast_own_query *query,
1688				       struct bridge_mcast_querier *querier)
1689{
1690	spin_lock(&br->multicast_lock);
1691	if (query->startup_sent < br->multicast_startup_query_count)
1692		query->startup_sent++;
1693
1694	RCU_INIT_POINTER(querier->port, NULL);
1695	br_multicast_send_query(br, NULL, query);
1696	spin_unlock(&br->multicast_lock);
1697}
1698
1699static void br_ip4_multicast_query_expired(unsigned long data)
1700{
1701	struct net_bridge *br = (void *)data;
1702
1703	br_multicast_query_expired(br, &br->ip4_own_query, &br->ip4_querier);
1704}
1705
1706#if IS_ENABLED(CONFIG_IPV6)
1707static void br_ip6_multicast_query_expired(unsigned long data)
1708{
1709	struct net_bridge *br = (void *)data;
1710
1711	br_multicast_query_expired(br, &br->ip6_own_query, &br->ip6_querier);
1712}
1713#endif
1714
1715void br_multicast_init(struct net_bridge *br)
1716{
1717	br->hash_elasticity = 4;
1718	br->hash_max = 512;
1719
1720	br->multicast_router = 1;
1721	br->multicast_querier = 0;
1722	br->multicast_query_use_ifaddr = 0;
1723	br->multicast_last_member_count = 2;
1724	br->multicast_startup_query_count = 2;
1725
1726	br->multicast_last_member_interval = HZ;
1727	br->multicast_query_response_interval = 10 * HZ;
1728	br->multicast_startup_query_interval = 125 * HZ / 4;
1729	br->multicast_query_interval = 125 * HZ;
1730	br->multicast_querier_interval = 255 * HZ;
1731	br->multicast_membership_interval = 260 * HZ;
1732
1733	br->ip4_other_query.delay_time = 0;
1734	br->ip4_querier.port = NULL;
1735#if IS_ENABLED(CONFIG_IPV6)
1736	br->ip6_other_query.delay_time = 0;
1737	br->ip6_querier.port = NULL;
1738#endif
1739
1740	spin_lock_init(&br->multicast_lock);
1741	setup_timer(&br->multicast_router_timer,
1742		    br_multicast_local_router_expired, 0);
1743	setup_timer(&br->ip4_other_query.timer,
1744		    br_ip4_multicast_querier_expired, (unsigned long)br);
1745	setup_timer(&br->ip4_own_query.timer, br_ip4_multicast_query_expired,
1746		    (unsigned long)br);
1747#if IS_ENABLED(CONFIG_IPV6)
1748	setup_timer(&br->ip6_other_query.timer,
1749		    br_ip6_multicast_querier_expired, (unsigned long)br);
1750	setup_timer(&br->ip6_own_query.timer, br_ip6_multicast_query_expired,
1751		    (unsigned long)br);
1752#endif
1753}
1754
1755static void __br_multicast_open(struct net_bridge *br,
1756				struct bridge_mcast_own_query *query)
1757{
1758	query->startup_sent = 0;
1759
1760	if (br->multicast_disabled)
1761		return;
1762
1763	mod_timer(&query->timer, jiffies);
1764}
1765
1766void br_multicast_open(struct net_bridge *br)
1767{
1768	__br_multicast_open(br, &br->ip4_own_query);
1769#if IS_ENABLED(CONFIG_IPV6)
1770	__br_multicast_open(br, &br->ip6_own_query);
1771#endif
1772}
1773
1774void br_multicast_stop(struct net_bridge *br)
1775{
1776	del_timer_sync(&br->multicast_router_timer);
1777	del_timer_sync(&br->ip4_other_query.timer);
1778	del_timer_sync(&br->ip4_own_query.timer);
1779#if IS_ENABLED(CONFIG_IPV6)
1780	del_timer_sync(&br->ip6_other_query.timer);
1781	del_timer_sync(&br->ip6_own_query.timer);
1782#endif
1783}
1784
1785void br_multicast_dev_del(struct net_bridge *br)
1786{
1787	struct net_bridge_mdb_htable *mdb;
1788	struct net_bridge_mdb_entry *mp;
1789	struct hlist_node *n;
1790	u32 ver;
1791	int i;
1792
1793	spin_lock_bh(&br->multicast_lock);
1794	mdb = mlock_dereference(br->mdb, br);
1795	if (!mdb)
1796		goto out;
1797
1798	br->mdb = NULL;
1799
1800	ver = mdb->ver;
1801	for (i = 0; i < mdb->max; i++) {
1802		hlist_for_each_entry_safe(mp, n, &mdb->mhash[i],
1803					  hlist[ver]) {
1804			del_timer(&mp->timer);
1805			call_rcu_bh(&mp->rcu, br_multicast_free_group);
1806		}
1807	}
1808
1809	if (mdb->old) {
1810		spin_unlock_bh(&br->multicast_lock);
1811		rcu_barrier_bh();
1812		spin_lock_bh(&br->multicast_lock);
1813		WARN_ON(mdb->old);
1814	}
1815
1816	mdb->old = mdb;
1817	call_rcu_bh(&mdb->rcu, br_mdb_free);
1818
1819out:
1820	spin_unlock_bh(&br->multicast_lock);
1821}
1822
1823int br_multicast_set_router(struct net_bridge *br, unsigned long val)
1824{
1825	int err = -EINVAL;
1826
1827	spin_lock_bh(&br->multicast_lock);
1828
1829	switch (val) {
1830	case 0:
1831	case 2:
1832		del_timer(&br->multicast_router_timer);
1833		/* fall through */
1834	case 1:
1835		br->multicast_router = val;
1836		err = 0;
1837		break;
1838	}
1839
1840	spin_unlock_bh(&br->multicast_lock);
1841
1842	return err;
1843}
1844
1845int br_multicast_set_port_router(struct net_bridge_port *p, unsigned long val)
1846{
1847	struct net_bridge *br = p->br;
1848	int err = -EINVAL;
1849
1850	spin_lock(&br->multicast_lock);
1851
1852	switch (val) {
1853	case 0:
1854	case 1:
1855	case 2:
1856		p->multicast_router = val;
1857		err = 0;
1858
1859		if (val < 2 && !hlist_unhashed(&p->rlist)) {
1860			hlist_del_init_rcu(&p->rlist);
1861			br_rtr_notify(br->dev, p, RTM_DELMDB);
1862		}
1863
1864		if (val == 1)
1865			break;
1866
1867		del_timer(&p->multicast_router_timer);
1868
1869		if (val == 0)
1870			break;
1871
1872		br_multicast_add_router(br, p);
1873		break;
1874	}
1875
1876	spin_unlock(&br->multicast_lock);
1877
1878	return err;
1879}
1880
1881static void br_multicast_start_querier(struct net_bridge *br,
1882				       struct bridge_mcast_own_query *query)
1883{
1884	struct net_bridge_port *port;
1885
1886	__br_multicast_open(br, query);
1887
1888	list_for_each_entry(port, &br->port_list, list) {
1889		if (port->state == BR_STATE_DISABLED ||
1890		    port->state == BR_STATE_BLOCKING)
1891			continue;
1892
1893		if (query == &br->ip4_own_query)
1894			br_multicast_enable(&port->ip4_own_query);
1895#if IS_ENABLED(CONFIG_IPV6)
1896		else
1897			br_multicast_enable(&port->ip6_own_query);
1898#endif
1899	}
1900}
1901
1902int br_multicast_toggle(struct net_bridge *br, unsigned long val)
1903{
1904	int err = 0;
1905	struct net_bridge_mdb_htable *mdb;
1906
1907	spin_lock_bh(&br->multicast_lock);
1908	if (br->multicast_disabled == !val)
1909		goto unlock;
1910
1911	br->multicast_disabled = !val;
1912	if (br->multicast_disabled)
1913		goto unlock;
1914
1915	if (!netif_running(br->dev))
1916		goto unlock;
1917
1918	mdb = mlock_dereference(br->mdb, br);
1919	if (mdb) {
1920		if (mdb->old) {
1921			err = -EEXIST;
1922rollback:
1923			br->multicast_disabled = !!val;
1924			goto unlock;
1925		}
1926
1927		err = br_mdb_rehash(&br->mdb, mdb->max,
1928				    br->hash_elasticity);
1929		if (err)
1930			goto rollback;
1931	}
1932
1933	br_multicast_start_querier(br, &br->ip4_own_query);
1934#if IS_ENABLED(CONFIG_IPV6)
1935	br_multicast_start_querier(br, &br->ip6_own_query);
1936#endif
1937
1938unlock:
1939	spin_unlock_bh(&br->multicast_lock);
1940
1941	return err;
1942}
1943
1944int br_multicast_set_querier(struct net_bridge *br, unsigned long val)
1945{
1946	unsigned long max_delay;
1947
1948	val = !!val;
1949
1950	spin_lock_bh(&br->multicast_lock);
1951	if (br->multicast_querier == val)
1952		goto unlock;
1953
1954	br->multicast_querier = val;
1955	if (!val)
1956		goto unlock;
1957
1958	max_delay = br->multicast_query_response_interval;
1959
1960	if (!timer_pending(&br->ip4_other_query.timer))
1961		br->ip4_other_query.delay_time = jiffies + max_delay;
1962
1963	br_multicast_start_querier(br, &br->ip4_own_query);
1964
1965#if IS_ENABLED(CONFIG_IPV6)
1966	if (!timer_pending(&br->ip6_other_query.timer))
1967		br->ip6_other_query.delay_time = jiffies + max_delay;
1968
1969	br_multicast_start_querier(br, &br->ip6_own_query);
1970#endif
1971
1972unlock:
1973	spin_unlock_bh(&br->multicast_lock);
1974
1975	return 0;
1976}
1977
1978int br_multicast_set_hash_max(struct net_bridge *br, unsigned long val)
1979{
1980	int err = -EINVAL;
1981	u32 old;
1982	struct net_bridge_mdb_htable *mdb;
1983
1984	spin_lock_bh(&br->multicast_lock);
1985	if (!is_power_of_2(val))
1986		goto unlock;
1987
1988	mdb = mlock_dereference(br->mdb, br);
1989	if (mdb && val < mdb->size)
1990		goto unlock;
1991
1992	err = 0;
1993
1994	old = br->hash_max;
1995	br->hash_max = val;
1996
1997	if (mdb) {
1998		if (mdb->old) {
1999			err = -EEXIST;
2000rollback:
2001			br->hash_max = old;
2002			goto unlock;
2003		}
2004
2005		err = br_mdb_rehash(&br->mdb, br->hash_max,
2006				    br->hash_elasticity);
2007		if (err)
2008			goto rollback;
2009	}
2010
2011unlock:
2012	spin_unlock_bh(&br->multicast_lock);
2013
2014	return err;
2015}
2016
2017/**
2018 * br_multicast_list_adjacent - Returns snooped multicast addresses
2019 * @dev:	The bridge port adjacent to which to retrieve addresses
2020 * @br_ip_list:	The list to store found, snooped multicast IP addresses in
2021 *
2022 * Creates a list of IP addresses (struct br_ip_list) sensed by the multicast
2023 * snooping feature on all bridge ports of dev's bridge device, excluding
2024 * the addresses from dev itself.
2025 *
2026 * Returns the number of items added to br_ip_list.
2027 *
2028 * Notes:
2029 * - br_ip_list needs to be initialized by caller
2030 * - br_ip_list might contain duplicates in the end
2031 *   (needs to be taken care of by caller)
2032 * - br_ip_list needs to be freed by caller
2033 */
2034int br_multicast_list_adjacent(struct net_device *dev,
2035			       struct list_head *br_ip_list)
2036{
2037	struct net_bridge *br;
2038	struct net_bridge_port *port;
2039	struct net_bridge_port_group *group;
2040	struct br_ip_list *entry;
2041	int count = 0;
2042
2043	rcu_read_lock();
2044	if (!br_ip_list || !br_port_exists(dev))
2045		goto unlock;
2046
2047	port = br_port_get_rcu(dev);
2048	if (!port || !port->br)
2049		goto unlock;
2050
2051	br = port->br;
2052
2053	list_for_each_entry_rcu(port, &br->port_list, list) {
2054		if (!port->dev || port->dev == dev)
2055			continue;
2056
2057		hlist_for_each_entry_rcu(group, &port->mglist, mglist) {
2058			entry = kmalloc(sizeof(*entry), GFP_ATOMIC);
2059			if (!entry)
2060				goto unlock;
2061
2062			entry->addr = group->addr;
2063			list_add(&entry->list, br_ip_list);
2064			count++;
2065		}
2066	}
2067
2068unlock:
2069	rcu_read_unlock();
2070	return count;
2071}
2072EXPORT_SYMBOL_GPL(br_multicast_list_adjacent);
2073
2074/**
2075 * br_multicast_has_querier_anywhere - Checks for a querier on a bridge
2076 * @dev: The bridge port providing the bridge on which to check for a querier
2077 * @proto: The protocol family to check for: IGMP -> ETH_P_IP, MLD -> ETH_P_IPV6
2078 *
2079 * Checks whether the given interface has a bridge on top and if so returns
2080 * true if a valid querier exists anywhere on the bridged link layer.
2081 * Otherwise returns false.
2082 */
2083bool br_multicast_has_querier_anywhere(struct net_device *dev, int proto)
2084{
2085	struct net_bridge *br;
2086	struct net_bridge_port *port;
2087	struct ethhdr eth;
2088	bool ret = false;
2089
2090	rcu_read_lock();
2091	if (!br_port_exists(dev))
2092		goto unlock;
2093
2094	port = br_port_get_rcu(dev);
2095	if (!port || !port->br)
2096		goto unlock;
2097
2098	br = port->br;
2099
2100	memset(&eth, 0, sizeof(eth));
2101	eth.h_proto = htons(proto);
2102
2103	ret = br_multicast_querier_exists(br, &eth);
2104
2105unlock:
2106	rcu_read_unlock();
2107	return ret;
2108}
2109EXPORT_SYMBOL_GPL(br_multicast_has_querier_anywhere);
2110
2111/**
2112 * br_multicast_has_querier_adjacent - Checks for a querier behind a bridge port
2113 * @dev: The bridge port adjacent to which to check for a querier
2114 * @proto: The protocol family to check for: IGMP -> ETH_P_IP, MLD -> ETH_P_IPV6
2115 *
2116 * Checks whether the given interface has a bridge on top and if so returns
2117 * true if a selected querier is behind one of the other ports of this
2118 * bridge. Otherwise returns false.
2119 */
2120bool br_multicast_has_querier_adjacent(struct net_device *dev, int proto)
2121{
2122	struct net_bridge *br;
2123	struct net_bridge_port *port;
2124	bool ret = false;
2125
2126	rcu_read_lock();
2127	if (!br_port_exists(dev))
2128		goto unlock;
2129
2130	port = br_port_get_rcu(dev);
2131	if (!port || !port->br)
2132		goto unlock;
2133
2134	br = port->br;
2135
2136	switch (proto) {
2137	case ETH_P_IP:
2138		if (!timer_pending(&br->ip4_other_query.timer) ||
2139		    rcu_dereference(br->ip4_querier.port) == port)
2140			goto unlock;
2141		break;
2142#if IS_ENABLED(CONFIG_IPV6)
2143	case ETH_P_IPV6:
2144		if (!timer_pending(&br->ip6_other_query.timer) ||
2145		    rcu_dereference(br->ip6_querier.port) == port)
2146			goto unlock;
2147		break;
2148#endif
2149	default:
2150		goto unlock;
2151	}
2152
2153	ret = true;
2154unlock:
2155	rcu_read_unlock();
2156	return ret;
2157}
2158EXPORT_SYMBOL_GPL(br_multicast_has_querier_adjacent);
2159