xref: /openbmc/linux/net/bridge/br_multicast.c (revision 48c926cd)
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 <linux/mroute.h>
29 #include <net/ip.h>
30 #include <net/switchdev.h>
31 #if IS_ENABLED(CONFIG_IPV6)
32 #include <net/ipv6.h>
33 #include <net/mld.h>
34 #include <net/ip6_checksum.h>
35 #include <net/addrconf.h>
36 #endif
37 
38 #include "br_private.h"
39 
40 static void br_multicast_start_querier(struct net_bridge *br,
41 				       struct bridge_mcast_own_query *query);
42 static void br_multicast_add_router(struct net_bridge *br,
43 				    struct net_bridge_port *port);
44 static void br_ip4_multicast_leave_group(struct net_bridge *br,
45 					 struct net_bridge_port *port,
46 					 __be32 group,
47 					 __u16 vid,
48 					 const unsigned char *src);
49 
50 static void __del_port_router(struct net_bridge_port *p);
51 #if IS_ENABLED(CONFIG_IPV6)
52 static void br_ip6_multicast_leave_group(struct net_bridge *br,
53 					 struct net_bridge_port *port,
54 					 const struct in6_addr *group,
55 					 __u16 vid, const unsigned char *src);
56 #endif
57 unsigned int br_mdb_rehash_seq;
58 
59 static inline int br_ip_equal(const struct br_ip *a, const struct br_ip *b)
60 {
61 	if (a->proto != b->proto)
62 		return 0;
63 	if (a->vid != b->vid)
64 		return 0;
65 	switch (a->proto) {
66 	case htons(ETH_P_IP):
67 		return a->u.ip4 == b->u.ip4;
68 #if IS_ENABLED(CONFIG_IPV6)
69 	case htons(ETH_P_IPV6):
70 		return ipv6_addr_equal(&a->u.ip6, &b->u.ip6);
71 #endif
72 	}
73 	return 0;
74 }
75 
76 static inline int __br_ip4_hash(struct net_bridge_mdb_htable *mdb, __be32 ip,
77 				__u16 vid)
78 {
79 	return jhash_2words((__force u32)ip, vid, mdb->secret) & (mdb->max - 1);
80 }
81 
82 #if IS_ENABLED(CONFIG_IPV6)
83 static inline int __br_ip6_hash(struct net_bridge_mdb_htable *mdb,
84 				const struct in6_addr *ip,
85 				__u16 vid)
86 {
87 	return jhash_2words(ipv6_addr_hash(ip), vid,
88 			    mdb->secret) & (mdb->max - 1);
89 }
90 #endif
91 
92 static inline int br_ip_hash(struct net_bridge_mdb_htable *mdb,
93 			     struct br_ip *ip)
94 {
95 	switch (ip->proto) {
96 	case htons(ETH_P_IP):
97 		return __br_ip4_hash(mdb, ip->u.ip4, ip->vid);
98 #if IS_ENABLED(CONFIG_IPV6)
99 	case htons(ETH_P_IPV6):
100 		return __br_ip6_hash(mdb, &ip->u.ip6, ip->vid);
101 #endif
102 	}
103 	return 0;
104 }
105 
106 static struct net_bridge_mdb_entry *__br_mdb_ip_get(
107 	struct net_bridge_mdb_htable *mdb, struct br_ip *dst, int hash)
108 {
109 	struct net_bridge_mdb_entry *mp;
110 
111 	hlist_for_each_entry_rcu(mp, &mdb->mhash[hash], hlist[mdb->ver]) {
112 		if (br_ip_equal(&mp->addr, dst))
113 			return mp;
114 	}
115 
116 	return NULL;
117 }
118 
119 struct net_bridge_mdb_entry *br_mdb_ip_get(struct net_bridge_mdb_htable *mdb,
120 					   struct br_ip *dst)
121 {
122 	if (!mdb)
123 		return NULL;
124 
125 	return __br_mdb_ip_get(mdb, dst, br_ip_hash(mdb, dst));
126 }
127 
128 static struct net_bridge_mdb_entry *br_mdb_ip4_get(
129 	struct net_bridge_mdb_htable *mdb, __be32 dst, __u16 vid)
130 {
131 	struct br_ip br_dst;
132 
133 	br_dst.u.ip4 = dst;
134 	br_dst.proto = htons(ETH_P_IP);
135 	br_dst.vid = vid;
136 
137 	return br_mdb_ip_get(mdb, &br_dst);
138 }
139 
140 #if IS_ENABLED(CONFIG_IPV6)
141 static struct net_bridge_mdb_entry *br_mdb_ip6_get(
142 	struct net_bridge_mdb_htable *mdb, const struct in6_addr *dst,
143 	__u16 vid)
144 {
145 	struct br_ip br_dst;
146 
147 	br_dst.u.ip6 = *dst;
148 	br_dst.proto = htons(ETH_P_IPV6);
149 	br_dst.vid = vid;
150 
151 	return br_mdb_ip_get(mdb, &br_dst);
152 }
153 #endif
154 
155 struct net_bridge_mdb_entry *br_mdb_get(struct net_bridge *br,
156 					struct sk_buff *skb, u16 vid)
157 {
158 	struct net_bridge_mdb_htable *mdb = rcu_dereference(br->mdb);
159 	struct br_ip ip;
160 
161 	if (br->multicast_disabled)
162 		return NULL;
163 
164 	if (BR_INPUT_SKB_CB(skb)->igmp)
165 		return NULL;
166 
167 	ip.proto = skb->protocol;
168 	ip.vid = vid;
169 
170 	switch (skb->protocol) {
171 	case htons(ETH_P_IP):
172 		ip.u.ip4 = ip_hdr(skb)->daddr;
173 		break;
174 #if IS_ENABLED(CONFIG_IPV6)
175 	case htons(ETH_P_IPV6):
176 		ip.u.ip6 = ipv6_hdr(skb)->daddr;
177 		break;
178 #endif
179 	default:
180 		return NULL;
181 	}
182 
183 	return br_mdb_ip_get(mdb, &ip);
184 }
185 
186 static void br_mdb_free(struct rcu_head *head)
187 {
188 	struct net_bridge_mdb_htable *mdb =
189 		container_of(head, struct net_bridge_mdb_htable, rcu);
190 	struct net_bridge_mdb_htable *old = mdb->old;
191 
192 	mdb->old = NULL;
193 	kfree(old->mhash);
194 	kfree(old);
195 }
196 
197 static int br_mdb_copy(struct net_bridge_mdb_htable *new,
198 		       struct net_bridge_mdb_htable *old,
199 		       int elasticity)
200 {
201 	struct net_bridge_mdb_entry *mp;
202 	int maxlen;
203 	int len;
204 	int i;
205 
206 	for (i = 0; i < old->max; i++)
207 		hlist_for_each_entry(mp, &old->mhash[i], hlist[old->ver])
208 			hlist_add_head(&mp->hlist[new->ver],
209 				       &new->mhash[br_ip_hash(new, &mp->addr)]);
210 
211 	if (!elasticity)
212 		return 0;
213 
214 	maxlen = 0;
215 	for (i = 0; i < new->max; i++) {
216 		len = 0;
217 		hlist_for_each_entry(mp, &new->mhash[i], hlist[new->ver])
218 			len++;
219 		if (len > maxlen)
220 			maxlen = len;
221 	}
222 
223 	return maxlen > elasticity ? -EINVAL : 0;
224 }
225 
226 void br_multicast_free_pg(struct rcu_head *head)
227 {
228 	struct net_bridge_port_group *p =
229 		container_of(head, struct net_bridge_port_group, rcu);
230 
231 	kfree(p);
232 }
233 
234 static void br_multicast_free_group(struct rcu_head *head)
235 {
236 	struct net_bridge_mdb_entry *mp =
237 		container_of(head, struct net_bridge_mdb_entry, rcu);
238 
239 	kfree(mp);
240 }
241 
242 static void br_multicast_group_expired(unsigned long data)
243 {
244 	struct net_bridge_mdb_entry *mp = (void *)data;
245 	struct net_bridge *br = mp->br;
246 	struct net_bridge_mdb_htable *mdb;
247 
248 	spin_lock(&br->multicast_lock);
249 	if (!netif_running(br->dev) || timer_pending(&mp->timer))
250 		goto out;
251 
252 	mp->mglist = false;
253 
254 	if (mp->ports)
255 		goto out;
256 
257 	mdb = mlock_dereference(br->mdb, br);
258 
259 	hlist_del_rcu(&mp->hlist[mdb->ver]);
260 	mdb->size--;
261 
262 	call_rcu_bh(&mp->rcu, br_multicast_free_group);
263 
264 out:
265 	spin_unlock(&br->multicast_lock);
266 }
267 
268 static void br_multicast_del_pg(struct net_bridge *br,
269 				struct net_bridge_port_group *pg)
270 {
271 	struct net_bridge_mdb_htable *mdb;
272 	struct net_bridge_mdb_entry *mp;
273 	struct net_bridge_port_group *p;
274 	struct net_bridge_port_group __rcu **pp;
275 
276 	mdb = mlock_dereference(br->mdb, br);
277 
278 	mp = br_mdb_ip_get(mdb, &pg->addr);
279 	if (WARN_ON(!mp))
280 		return;
281 
282 	for (pp = &mp->ports;
283 	     (p = mlock_dereference(*pp, br)) != NULL;
284 	     pp = &p->next) {
285 		if (p != pg)
286 			continue;
287 
288 		rcu_assign_pointer(*pp, p->next);
289 		hlist_del_init(&p->mglist);
290 		del_timer(&p->timer);
291 		br_mdb_notify(br->dev, p->port, &pg->addr, RTM_DELMDB,
292 			      p->flags);
293 		call_rcu_bh(&p->rcu, br_multicast_free_pg);
294 
295 		if (!mp->ports && !mp->mglist &&
296 		    netif_running(br->dev))
297 			mod_timer(&mp->timer, jiffies);
298 
299 		return;
300 	}
301 
302 	WARN_ON(1);
303 }
304 
305 static void br_multicast_port_group_expired(unsigned long data)
306 {
307 	struct net_bridge_port_group *pg = (void *)data;
308 	struct net_bridge *br = pg->port->br;
309 
310 	spin_lock(&br->multicast_lock);
311 	if (!netif_running(br->dev) || timer_pending(&pg->timer) ||
312 	    hlist_unhashed(&pg->mglist) || pg->flags & MDB_PG_FLAGS_PERMANENT)
313 		goto out;
314 
315 	br_multicast_del_pg(br, pg);
316 
317 out:
318 	spin_unlock(&br->multicast_lock);
319 }
320 
321 static int br_mdb_rehash(struct net_bridge_mdb_htable __rcu **mdbp, int max,
322 			 int elasticity)
323 {
324 	struct net_bridge_mdb_htable *old = rcu_dereference_protected(*mdbp, 1);
325 	struct net_bridge_mdb_htable *mdb;
326 	int err;
327 
328 	mdb = kmalloc(sizeof(*mdb), GFP_ATOMIC);
329 	if (!mdb)
330 		return -ENOMEM;
331 
332 	mdb->max = max;
333 	mdb->old = old;
334 
335 	mdb->mhash = kzalloc(max * sizeof(*mdb->mhash), GFP_ATOMIC);
336 	if (!mdb->mhash) {
337 		kfree(mdb);
338 		return -ENOMEM;
339 	}
340 
341 	mdb->size = old ? old->size : 0;
342 	mdb->ver = old ? old->ver ^ 1 : 0;
343 
344 	if (!old || elasticity)
345 		get_random_bytes(&mdb->secret, sizeof(mdb->secret));
346 	else
347 		mdb->secret = old->secret;
348 
349 	if (!old)
350 		goto out;
351 
352 	err = br_mdb_copy(mdb, old, elasticity);
353 	if (err) {
354 		kfree(mdb->mhash);
355 		kfree(mdb);
356 		return err;
357 	}
358 
359 	br_mdb_rehash_seq++;
360 	call_rcu_bh(&mdb->rcu, br_mdb_free);
361 
362 out:
363 	rcu_assign_pointer(*mdbp, mdb);
364 
365 	return 0;
366 }
367 
368 static struct sk_buff *br_ip4_multicast_alloc_query(struct net_bridge *br,
369 						    __be32 group,
370 						    u8 *igmp_type)
371 {
372 	struct igmpv3_query *ihv3;
373 	size_t igmp_hdr_size;
374 	struct sk_buff *skb;
375 	struct igmphdr *ih;
376 	struct ethhdr *eth;
377 	struct iphdr *iph;
378 
379 	igmp_hdr_size = sizeof(*ih);
380 	if (br->multicast_igmp_version == 3)
381 		igmp_hdr_size = sizeof(*ihv3);
382 	skb = netdev_alloc_skb_ip_align(br->dev, sizeof(*eth) + sizeof(*iph) +
383 						 igmp_hdr_size + 4);
384 	if (!skb)
385 		goto out;
386 
387 	skb->protocol = htons(ETH_P_IP);
388 
389 	skb_reset_mac_header(skb);
390 	eth = eth_hdr(skb);
391 
392 	ether_addr_copy(eth->h_source, br->dev->dev_addr);
393 	eth->h_dest[0] = 1;
394 	eth->h_dest[1] = 0;
395 	eth->h_dest[2] = 0x5e;
396 	eth->h_dest[3] = 0;
397 	eth->h_dest[4] = 0;
398 	eth->h_dest[5] = 1;
399 	eth->h_proto = htons(ETH_P_IP);
400 	skb_put(skb, sizeof(*eth));
401 
402 	skb_set_network_header(skb, skb->len);
403 	iph = ip_hdr(skb);
404 
405 	iph->version = 4;
406 	iph->ihl = 6;
407 	iph->tos = 0xc0;
408 	iph->tot_len = htons(sizeof(*iph) + igmp_hdr_size + 4);
409 	iph->id = 0;
410 	iph->frag_off = htons(IP_DF);
411 	iph->ttl = 1;
412 	iph->protocol = IPPROTO_IGMP;
413 	iph->saddr = br->multicast_query_use_ifaddr ?
414 		     inet_select_addr(br->dev, 0, RT_SCOPE_LINK) : 0;
415 	iph->daddr = htonl(INADDR_ALLHOSTS_GROUP);
416 	((u8 *)&iph[1])[0] = IPOPT_RA;
417 	((u8 *)&iph[1])[1] = 4;
418 	((u8 *)&iph[1])[2] = 0;
419 	((u8 *)&iph[1])[3] = 0;
420 	ip_send_check(iph);
421 	skb_put(skb, 24);
422 
423 	skb_set_transport_header(skb, skb->len);
424 	*igmp_type = IGMP_HOST_MEMBERSHIP_QUERY;
425 
426 	switch (br->multicast_igmp_version) {
427 	case 2:
428 		ih = igmp_hdr(skb);
429 		ih->type = IGMP_HOST_MEMBERSHIP_QUERY;
430 		ih->code = (group ? br->multicast_last_member_interval :
431 				    br->multicast_query_response_interval) /
432 			   (HZ / IGMP_TIMER_SCALE);
433 		ih->group = group;
434 		ih->csum = 0;
435 		ih->csum = ip_compute_csum((void *)ih, sizeof(*ih));
436 		break;
437 	case 3:
438 		ihv3 = igmpv3_query_hdr(skb);
439 		ihv3->type = IGMP_HOST_MEMBERSHIP_QUERY;
440 		ihv3->code = (group ? br->multicast_last_member_interval :
441 				      br->multicast_query_response_interval) /
442 			     (HZ / IGMP_TIMER_SCALE);
443 		ihv3->group = group;
444 		ihv3->qqic = br->multicast_query_interval / HZ;
445 		ihv3->nsrcs = 0;
446 		ihv3->resv = 0;
447 		ihv3->suppress = 0;
448 		ihv3->qrv = 2;
449 		ihv3->csum = 0;
450 		ihv3->csum = ip_compute_csum((void *)ihv3, sizeof(*ihv3));
451 		break;
452 	}
453 
454 	skb_put(skb, igmp_hdr_size);
455 	__skb_pull(skb, sizeof(*eth));
456 
457 out:
458 	return skb;
459 }
460 
461 #if IS_ENABLED(CONFIG_IPV6)
462 static struct sk_buff *br_ip6_multicast_alloc_query(struct net_bridge *br,
463 						    const struct in6_addr *grp,
464 						    u8 *igmp_type)
465 {
466 	struct mld2_query *mld2q;
467 	unsigned long interval;
468 	struct ipv6hdr *ip6h;
469 	struct mld_msg *mldq;
470 	size_t mld_hdr_size;
471 	struct sk_buff *skb;
472 	struct ethhdr *eth;
473 	u8 *hopopt;
474 
475 	mld_hdr_size = sizeof(*mldq);
476 	if (br->multicast_mld_version == 2)
477 		mld_hdr_size = sizeof(*mld2q);
478 	skb = netdev_alloc_skb_ip_align(br->dev, sizeof(*eth) + sizeof(*ip6h) +
479 						 8 + mld_hdr_size);
480 	if (!skb)
481 		goto out;
482 
483 	skb->protocol = htons(ETH_P_IPV6);
484 
485 	/* Ethernet header */
486 	skb_reset_mac_header(skb);
487 	eth = eth_hdr(skb);
488 
489 	ether_addr_copy(eth->h_source, br->dev->dev_addr);
490 	eth->h_proto = htons(ETH_P_IPV6);
491 	skb_put(skb, sizeof(*eth));
492 
493 	/* IPv6 header + HbH option */
494 	skb_set_network_header(skb, skb->len);
495 	ip6h = ipv6_hdr(skb);
496 
497 	*(__force __be32 *)ip6h = htonl(0x60000000);
498 	ip6h->payload_len = htons(8 + mld_hdr_size);
499 	ip6h->nexthdr = IPPROTO_HOPOPTS;
500 	ip6h->hop_limit = 1;
501 	ipv6_addr_set(&ip6h->daddr, htonl(0xff020000), 0, 0, htonl(1));
502 	if (ipv6_dev_get_saddr(dev_net(br->dev), br->dev, &ip6h->daddr, 0,
503 			       &ip6h->saddr)) {
504 		kfree_skb(skb);
505 		br->has_ipv6_addr = 0;
506 		return NULL;
507 	}
508 
509 	br->has_ipv6_addr = 1;
510 	ipv6_eth_mc_map(&ip6h->daddr, eth->h_dest);
511 
512 	hopopt = (u8 *)(ip6h + 1);
513 	hopopt[0] = IPPROTO_ICMPV6;		/* next hdr */
514 	hopopt[1] = 0;				/* length of HbH */
515 	hopopt[2] = IPV6_TLV_ROUTERALERT;	/* Router Alert */
516 	hopopt[3] = 2;				/* Length of RA Option */
517 	hopopt[4] = 0;				/* Type = 0x0000 (MLD) */
518 	hopopt[5] = 0;
519 	hopopt[6] = IPV6_TLV_PAD1;		/* Pad1 */
520 	hopopt[7] = IPV6_TLV_PAD1;		/* Pad1 */
521 
522 	skb_put(skb, sizeof(*ip6h) + 8);
523 
524 	/* ICMPv6 */
525 	skb_set_transport_header(skb, skb->len);
526 	interval = ipv6_addr_any(grp) ?
527 			br->multicast_query_response_interval :
528 			br->multicast_last_member_interval;
529 	*igmp_type = ICMPV6_MGM_QUERY;
530 	switch (br->multicast_mld_version) {
531 	case 1:
532 		mldq = (struct mld_msg *)icmp6_hdr(skb);
533 		mldq->mld_type = ICMPV6_MGM_QUERY;
534 		mldq->mld_code = 0;
535 		mldq->mld_cksum = 0;
536 		mldq->mld_maxdelay = htons((u16)jiffies_to_msecs(interval));
537 		mldq->mld_reserved = 0;
538 		mldq->mld_mca = *grp;
539 		mldq->mld_cksum = csum_ipv6_magic(&ip6h->saddr, &ip6h->daddr,
540 						  sizeof(*mldq), IPPROTO_ICMPV6,
541 						  csum_partial(mldq,
542 							       sizeof(*mldq),
543 							       0));
544 		break;
545 	case 2:
546 		mld2q = (struct mld2_query *)icmp6_hdr(skb);
547 		mld2q->mld2q_mrc = htons((u16)jiffies_to_msecs(interval));
548 		mld2q->mld2q_type = ICMPV6_MGM_QUERY;
549 		mld2q->mld2q_code = 0;
550 		mld2q->mld2q_cksum = 0;
551 		mld2q->mld2q_resv1 = 0;
552 		mld2q->mld2q_resv2 = 0;
553 		mld2q->mld2q_suppress = 0;
554 		mld2q->mld2q_qrv = 2;
555 		mld2q->mld2q_nsrcs = 0;
556 		mld2q->mld2q_qqic = br->multicast_query_interval / HZ;
557 		mld2q->mld2q_mca = *grp;
558 		mld2q->mld2q_cksum = csum_ipv6_magic(&ip6h->saddr, &ip6h->daddr,
559 						     sizeof(*mld2q),
560 						     IPPROTO_ICMPV6,
561 						     csum_partial(mld2q,
562 								  sizeof(*mld2q),
563 								  0));
564 		break;
565 	}
566 	skb_put(skb, mld_hdr_size);
567 
568 	__skb_pull(skb, sizeof(*eth));
569 
570 out:
571 	return skb;
572 }
573 #endif
574 
575 static struct sk_buff *br_multicast_alloc_query(struct net_bridge *br,
576 						struct br_ip *addr,
577 						u8 *igmp_type)
578 {
579 	switch (addr->proto) {
580 	case htons(ETH_P_IP):
581 		return br_ip4_multicast_alloc_query(br, addr->u.ip4, igmp_type);
582 #if IS_ENABLED(CONFIG_IPV6)
583 	case htons(ETH_P_IPV6):
584 		return br_ip6_multicast_alloc_query(br, &addr->u.ip6,
585 						    igmp_type);
586 #endif
587 	}
588 	return NULL;
589 }
590 
591 static struct net_bridge_mdb_entry *br_multicast_get_group(
592 	struct net_bridge *br, struct net_bridge_port *port,
593 	struct br_ip *group, int hash)
594 {
595 	struct net_bridge_mdb_htable *mdb;
596 	struct net_bridge_mdb_entry *mp;
597 	unsigned int count = 0;
598 	unsigned int max;
599 	int elasticity;
600 	int err;
601 
602 	mdb = rcu_dereference_protected(br->mdb, 1);
603 	hlist_for_each_entry(mp, &mdb->mhash[hash], hlist[mdb->ver]) {
604 		count++;
605 		if (unlikely(br_ip_equal(group, &mp->addr)))
606 			return mp;
607 	}
608 
609 	elasticity = 0;
610 	max = mdb->max;
611 
612 	if (unlikely(count > br->hash_elasticity && count)) {
613 		if (net_ratelimit())
614 			br_info(br, "Multicast hash table "
615 				"chain limit reached: %s\n",
616 				port ? port->dev->name : br->dev->name);
617 
618 		elasticity = br->hash_elasticity;
619 	}
620 
621 	if (mdb->size >= max) {
622 		max *= 2;
623 		if (unlikely(max > br->hash_max)) {
624 			br_warn(br, "Multicast hash table maximum of %d "
625 				"reached, disabling snooping: %s\n",
626 				br->hash_max,
627 				port ? port->dev->name : br->dev->name);
628 			err = -E2BIG;
629 disable:
630 			br->multicast_disabled = 1;
631 			goto err;
632 		}
633 	}
634 
635 	if (max > mdb->max || elasticity) {
636 		if (mdb->old) {
637 			if (net_ratelimit())
638 				br_info(br, "Multicast hash table "
639 					"on fire: %s\n",
640 					port ? port->dev->name : br->dev->name);
641 			err = -EEXIST;
642 			goto err;
643 		}
644 
645 		err = br_mdb_rehash(&br->mdb, max, elasticity);
646 		if (err) {
647 			br_warn(br, "Cannot rehash multicast "
648 				"hash table, disabling snooping: %s, %d, %d\n",
649 				port ? port->dev->name : br->dev->name,
650 				mdb->size, err);
651 			goto disable;
652 		}
653 
654 		err = -EAGAIN;
655 		goto err;
656 	}
657 
658 	return NULL;
659 
660 err:
661 	mp = ERR_PTR(err);
662 	return mp;
663 }
664 
665 struct net_bridge_mdb_entry *br_multicast_new_group(struct net_bridge *br,
666 						    struct net_bridge_port *p,
667 						    struct br_ip *group)
668 {
669 	struct net_bridge_mdb_htable *mdb;
670 	struct net_bridge_mdb_entry *mp;
671 	int hash;
672 	int err;
673 
674 	mdb = rcu_dereference_protected(br->mdb, 1);
675 	if (!mdb) {
676 		err = br_mdb_rehash(&br->mdb, BR_HASH_SIZE, 0);
677 		if (err)
678 			return ERR_PTR(err);
679 		goto rehash;
680 	}
681 
682 	hash = br_ip_hash(mdb, group);
683 	mp = br_multicast_get_group(br, p, group, hash);
684 	switch (PTR_ERR(mp)) {
685 	case 0:
686 		break;
687 
688 	case -EAGAIN:
689 rehash:
690 		mdb = rcu_dereference_protected(br->mdb, 1);
691 		hash = br_ip_hash(mdb, group);
692 		break;
693 
694 	default:
695 		goto out;
696 	}
697 
698 	mp = kzalloc(sizeof(*mp), GFP_ATOMIC);
699 	if (unlikely(!mp))
700 		return ERR_PTR(-ENOMEM);
701 
702 	mp->br = br;
703 	mp->addr = *group;
704 	setup_timer(&mp->timer, br_multicast_group_expired,
705 		    (unsigned long)mp);
706 
707 	hlist_add_head_rcu(&mp->hlist[mdb->ver], &mdb->mhash[hash]);
708 	mdb->size++;
709 
710 out:
711 	return mp;
712 }
713 
714 struct net_bridge_port_group *br_multicast_new_port_group(
715 			struct net_bridge_port *port,
716 			struct br_ip *group,
717 			struct net_bridge_port_group __rcu *next,
718 			unsigned char flags,
719 			const unsigned char *src)
720 {
721 	struct net_bridge_port_group *p;
722 
723 	p = kzalloc(sizeof(*p), GFP_ATOMIC);
724 	if (unlikely(!p))
725 		return NULL;
726 
727 	p->addr = *group;
728 	p->port = port;
729 	p->flags = flags;
730 	rcu_assign_pointer(p->next, next);
731 	hlist_add_head(&p->mglist, &port->mglist);
732 	setup_timer(&p->timer, br_multicast_port_group_expired,
733 		    (unsigned long)p);
734 
735 	if (src)
736 		memcpy(p->eth_addr, src, ETH_ALEN);
737 	else
738 		memset(p->eth_addr, 0xff, ETH_ALEN);
739 
740 	return p;
741 }
742 
743 static bool br_port_group_equal(struct net_bridge_port_group *p,
744 				struct net_bridge_port *port,
745 				const unsigned char *src)
746 {
747 	if (p->port != port)
748 		return false;
749 
750 	if (!(port->flags & BR_MULTICAST_TO_UNICAST))
751 		return true;
752 
753 	return ether_addr_equal(src, p->eth_addr);
754 }
755 
756 static int br_multicast_add_group(struct net_bridge *br,
757 				  struct net_bridge_port *port,
758 				  struct br_ip *group,
759 				  const unsigned char *src)
760 {
761 	struct net_bridge_port_group __rcu **pp;
762 	struct net_bridge_port_group *p;
763 	struct net_bridge_mdb_entry *mp;
764 	unsigned long now = jiffies;
765 	int err;
766 
767 	spin_lock(&br->multicast_lock);
768 	if (!netif_running(br->dev) ||
769 	    (port && port->state == BR_STATE_DISABLED))
770 		goto out;
771 
772 	mp = br_multicast_new_group(br, port, group);
773 	err = PTR_ERR(mp);
774 	if (IS_ERR(mp))
775 		goto err;
776 
777 	if (!port) {
778 		mp->mglist = true;
779 		mod_timer(&mp->timer, now + br->multicast_membership_interval);
780 		goto out;
781 	}
782 
783 	for (pp = &mp->ports;
784 	     (p = mlock_dereference(*pp, br)) != NULL;
785 	     pp = &p->next) {
786 		if (br_port_group_equal(p, port, src))
787 			goto found;
788 		if ((unsigned long)p->port < (unsigned long)port)
789 			break;
790 	}
791 
792 	p = br_multicast_new_port_group(port, group, *pp, 0, src);
793 	if (unlikely(!p))
794 		goto err;
795 	rcu_assign_pointer(*pp, p);
796 	br_mdb_notify(br->dev, port, group, RTM_NEWMDB, 0);
797 
798 found:
799 	mod_timer(&p->timer, now + br->multicast_membership_interval);
800 out:
801 	err = 0;
802 
803 err:
804 	spin_unlock(&br->multicast_lock);
805 	return err;
806 }
807 
808 static int br_ip4_multicast_add_group(struct net_bridge *br,
809 				      struct net_bridge_port *port,
810 				      __be32 group,
811 				      __u16 vid,
812 				      const unsigned char *src)
813 {
814 	struct br_ip br_group;
815 
816 	if (ipv4_is_local_multicast(group))
817 		return 0;
818 
819 	br_group.u.ip4 = group;
820 	br_group.proto = htons(ETH_P_IP);
821 	br_group.vid = vid;
822 
823 	return br_multicast_add_group(br, port, &br_group, src);
824 }
825 
826 #if IS_ENABLED(CONFIG_IPV6)
827 static int br_ip6_multicast_add_group(struct net_bridge *br,
828 				      struct net_bridge_port *port,
829 				      const struct in6_addr *group,
830 				      __u16 vid,
831 				      const unsigned char *src)
832 {
833 	struct br_ip br_group;
834 
835 	if (ipv6_addr_is_ll_all_nodes(group))
836 		return 0;
837 
838 	br_group.u.ip6 = *group;
839 	br_group.proto = htons(ETH_P_IPV6);
840 	br_group.vid = vid;
841 
842 	return br_multicast_add_group(br, port, &br_group, src);
843 }
844 #endif
845 
846 static void br_multicast_router_expired(unsigned long data)
847 {
848 	struct net_bridge_port *port = (void *)data;
849 	struct net_bridge *br = port->br;
850 
851 	spin_lock(&br->multicast_lock);
852 	if (port->multicast_router == MDB_RTR_TYPE_DISABLED ||
853 	    port->multicast_router == MDB_RTR_TYPE_PERM ||
854 	    timer_pending(&port->multicast_router_timer))
855 		goto out;
856 
857 	__del_port_router(port);
858 out:
859 	spin_unlock(&br->multicast_lock);
860 }
861 
862 static void br_multicast_local_router_expired(unsigned long data)
863 {
864 }
865 
866 static void br_multicast_querier_expired(struct net_bridge *br,
867 					 struct bridge_mcast_own_query *query)
868 {
869 	spin_lock(&br->multicast_lock);
870 	if (!netif_running(br->dev) || br->multicast_disabled)
871 		goto out;
872 
873 	br_multicast_start_querier(br, query);
874 
875 out:
876 	spin_unlock(&br->multicast_lock);
877 }
878 
879 static void br_ip4_multicast_querier_expired(unsigned long data)
880 {
881 	struct net_bridge *br = (void *)data;
882 
883 	br_multicast_querier_expired(br, &br->ip4_own_query);
884 }
885 
886 #if IS_ENABLED(CONFIG_IPV6)
887 static void br_ip6_multicast_querier_expired(unsigned long data)
888 {
889 	struct net_bridge *br = (void *)data;
890 
891 	br_multicast_querier_expired(br, &br->ip6_own_query);
892 }
893 #endif
894 
895 static void br_multicast_select_own_querier(struct net_bridge *br,
896 					    struct br_ip *ip,
897 					    struct sk_buff *skb)
898 {
899 	if (ip->proto == htons(ETH_P_IP))
900 		br->ip4_querier.addr.u.ip4 = ip_hdr(skb)->saddr;
901 #if IS_ENABLED(CONFIG_IPV6)
902 	else
903 		br->ip6_querier.addr.u.ip6 = ipv6_hdr(skb)->saddr;
904 #endif
905 }
906 
907 static void __br_multicast_send_query(struct net_bridge *br,
908 				      struct net_bridge_port *port,
909 				      struct br_ip *ip)
910 {
911 	struct sk_buff *skb;
912 	u8 igmp_type;
913 
914 	skb = br_multicast_alloc_query(br, ip, &igmp_type);
915 	if (!skb)
916 		return;
917 
918 	if (port) {
919 		skb->dev = port->dev;
920 		br_multicast_count(br, port, skb, igmp_type,
921 				   BR_MCAST_DIR_TX);
922 		NF_HOOK(NFPROTO_BRIDGE, NF_BR_LOCAL_OUT,
923 			dev_net(port->dev), NULL, skb, NULL, skb->dev,
924 			br_dev_queue_push_xmit);
925 	} else {
926 		br_multicast_select_own_querier(br, ip, skb);
927 		br_multicast_count(br, port, skb, igmp_type,
928 				   BR_MCAST_DIR_RX);
929 		netif_rx(skb);
930 	}
931 }
932 
933 static void br_multicast_send_query(struct net_bridge *br,
934 				    struct net_bridge_port *port,
935 				    struct bridge_mcast_own_query *own_query)
936 {
937 	struct bridge_mcast_other_query *other_query = NULL;
938 	struct br_ip br_group;
939 	unsigned long time;
940 
941 	if (!netif_running(br->dev) || br->multicast_disabled ||
942 	    !br->multicast_querier)
943 		return;
944 
945 	memset(&br_group.u, 0, sizeof(br_group.u));
946 
947 	if (port ? (own_query == &port->ip4_own_query) :
948 		   (own_query == &br->ip4_own_query)) {
949 		other_query = &br->ip4_other_query;
950 		br_group.proto = htons(ETH_P_IP);
951 #if IS_ENABLED(CONFIG_IPV6)
952 	} else {
953 		other_query = &br->ip6_other_query;
954 		br_group.proto = htons(ETH_P_IPV6);
955 #endif
956 	}
957 
958 	if (!other_query || timer_pending(&other_query->timer))
959 		return;
960 
961 	__br_multicast_send_query(br, port, &br_group);
962 
963 	time = jiffies;
964 	time += own_query->startup_sent < br->multicast_startup_query_count ?
965 		br->multicast_startup_query_interval :
966 		br->multicast_query_interval;
967 	mod_timer(&own_query->timer, time);
968 }
969 
970 static void
971 br_multicast_port_query_expired(struct net_bridge_port *port,
972 				struct bridge_mcast_own_query *query)
973 {
974 	struct net_bridge *br = port->br;
975 
976 	spin_lock(&br->multicast_lock);
977 	if (port->state == BR_STATE_DISABLED ||
978 	    port->state == BR_STATE_BLOCKING)
979 		goto out;
980 
981 	if (query->startup_sent < br->multicast_startup_query_count)
982 		query->startup_sent++;
983 
984 	br_multicast_send_query(port->br, port, query);
985 
986 out:
987 	spin_unlock(&br->multicast_lock);
988 }
989 
990 static void br_ip4_multicast_port_query_expired(unsigned long data)
991 {
992 	struct net_bridge_port *port = (void *)data;
993 
994 	br_multicast_port_query_expired(port, &port->ip4_own_query);
995 }
996 
997 #if IS_ENABLED(CONFIG_IPV6)
998 static void br_ip6_multicast_port_query_expired(unsigned long data)
999 {
1000 	struct net_bridge_port *port = (void *)data;
1001 
1002 	br_multicast_port_query_expired(port, &port->ip6_own_query);
1003 }
1004 #endif
1005 
1006 static void br_mc_disabled_update(struct net_device *dev, bool value)
1007 {
1008 	struct switchdev_attr attr = {
1009 		.orig_dev = dev,
1010 		.id = SWITCHDEV_ATTR_ID_BRIDGE_MC_DISABLED,
1011 		.flags = SWITCHDEV_F_DEFER,
1012 		.u.mc_disabled = value,
1013 	};
1014 
1015 	switchdev_port_attr_set(dev, &attr);
1016 }
1017 
1018 int br_multicast_add_port(struct net_bridge_port *port)
1019 {
1020 	port->multicast_router = MDB_RTR_TYPE_TEMP_QUERY;
1021 
1022 	setup_timer(&port->multicast_router_timer, br_multicast_router_expired,
1023 		    (unsigned long)port);
1024 	setup_timer(&port->ip4_own_query.timer,
1025 		    br_ip4_multicast_port_query_expired, (unsigned long)port);
1026 #if IS_ENABLED(CONFIG_IPV6)
1027 	setup_timer(&port->ip6_own_query.timer,
1028 		    br_ip6_multicast_port_query_expired, (unsigned long)port);
1029 #endif
1030 	br_mc_disabled_update(port->dev, port->br->multicast_disabled);
1031 
1032 	port->mcast_stats = netdev_alloc_pcpu_stats(struct bridge_mcast_stats);
1033 	if (!port->mcast_stats)
1034 		return -ENOMEM;
1035 
1036 	return 0;
1037 }
1038 
1039 void br_multicast_del_port(struct net_bridge_port *port)
1040 {
1041 	struct net_bridge *br = port->br;
1042 	struct net_bridge_port_group *pg;
1043 	struct hlist_node *n;
1044 
1045 	/* Take care of the remaining groups, only perm ones should be left */
1046 	spin_lock_bh(&br->multicast_lock);
1047 	hlist_for_each_entry_safe(pg, n, &port->mglist, mglist)
1048 		br_multicast_del_pg(br, pg);
1049 	spin_unlock_bh(&br->multicast_lock);
1050 	del_timer_sync(&port->multicast_router_timer);
1051 	free_percpu(port->mcast_stats);
1052 }
1053 
1054 static void br_multicast_enable(struct bridge_mcast_own_query *query)
1055 {
1056 	query->startup_sent = 0;
1057 
1058 	if (try_to_del_timer_sync(&query->timer) >= 0 ||
1059 	    del_timer(&query->timer))
1060 		mod_timer(&query->timer, jiffies);
1061 }
1062 
1063 static void __br_multicast_enable_port(struct net_bridge_port *port)
1064 {
1065 	struct net_bridge *br = port->br;
1066 
1067 	if (br->multicast_disabled || !netif_running(br->dev))
1068 		return;
1069 
1070 	br_multicast_enable(&port->ip4_own_query);
1071 #if IS_ENABLED(CONFIG_IPV6)
1072 	br_multicast_enable(&port->ip6_own_query);
1073 #endif
1074 	if (port->multicast_router == MDB_RTR_TYPE_PERM &&
1075 	    hlist_unhashed(&port->rlist))
1076 		br_multicast_add_router(br, port);
1077 }
1078 
1079 void br_multicast_enable_port(struct net_bridge_port *port)
1080 {
1081 	struct net_bridge *br = port->br;
1082 
1083 	spin_lock(&br->multicast_lock);
1084 	__br_multicast_enable_port(port);
1085 	spin_unlock(&br->multicast_lock);
1086 }
1087 
1088 void br_multicast_disable_port(struct net_bridge_port *port)
1089 {
1090 	struct net_bridge *br = port->br;
1091 	struct net_bridge_port_group *pg;
1092 	struct hlist_node *n;
1093 
1094 	spin_lock(&br->multicast_lock);
1095 	hlist_for_each_entry_safe(pg, n, &port->mglist, mglist)
1096 		if (!(pg->flags & MDB_PG_FLAGS_PERMANENT))
1097 			br_multicast_del_pg(br, pg);
1098 
1099 	__del_port_router(port);
1100 
1101 	del_timer(&port->multicast_router_timer);
1102 	del_timer(&port->ip4_own_query.timer);
1103 #if IS_ENABLED(CONFIG_IPV6)
1104 	del_timer(&port->ip6_own_query.timer);
1105 #endif
1106 	spin_unlock(&br->multicast_lock);
1107 }
1108 
1109 static int br_ip4_multicast_igmp3_report(struct net_bridge *br,
1110 					 struct net_bridge_port *port,
1111 					 struct sk_buff *skb,
1112 					 u16 vid)
1113 {
1114 	const unsigned char *src;
1115 	struct igmpv3_report *ih;
1116 	struct igmpv3_grec *grec;
1117 	int i;
1118 	int len;
1119 	int num;
1120 	int type;
1121 	int err = 0;
1122 	__be32 group;
1123 
1124 	ih = igmpv3_report_hdr(skb);
1125 	num = ntohs(ih->ngrec);
1126 	len = skb_transport_offset(skb) + sizeof(*ih);
1127 
1128 	for (i = 0; i < num; i++) {
1129 		len += sizeof(*grec);
1130 		if (!pskb_may_pull(skb, len))
1131 			return -EINVAL;
1132 
1133 		grec = (void *)(skb->data + len - sizeof(*grec));
1134 		group = grec->grec_mca;
1135 		type = grec->grec_type;
1136 
1137 		len += ntohs(grec->grec_nsrcs) * 4;
1138 		if (!pskb_may_pull(skb, len))
1139 			return -EINVAL;
1140 
1141 		/* We treat this as an IGMPv2 report for now. */
1142 		switch (type) {
1143 		case IGMPV3_MODE_IS_INCLUDE:
1144 		case IGMPV3_MODE_IS_EXCLUDE:
1145 		case IGMPV3_CHANGE_TO_INCLUDE:
1146 		case IGMPV3_CHANGE_TO_EXCLUDE:
1147 		case IGMPV3_ALLOW_NEW_SOURCES:
1148 		case IGMPV3_BLOCK_OLD_SOURCES:
1149 			break;
1150 
1151 		default:
1152 			continue;
1153 		}
1154 
1155 		src = eth_hdr(skb)->h_source;
1156 		if ((type == IGMPV3_CHANGE_TO_INCLUDE ||
1157 		     type == IGMPV3_MODE_IS_INCLUDE) &&
1158 		    ntohs(grec->grec_nsrcs) == 0) {
1159 			br_ip4_multicast_leave_group(br, port, group, vid, src);
1160 		} else {
1161 			err = br_ip4_multicast_add_group(br, port, group, vid,
1162 							 src);
1163 			if (err)
1164 				break;
1165 		}
1166 	}
1167 
1168 	return err;
1169 }
1170 
1171 #if IS_ENABLED(CONFIG_IPV6)
1172 static int br_ip6_multicast_mld2_report(struct net_bridge *br,
1173 					struct net_bridge_port *port,
1174 					struct sk_buff *skb,
1175 					u16 vid)
1176 {
1177 	const unsigned char *src;
1178 	struct icmp6hdr *icmp6h;
1179 	struct mld2_grec *grec;
1180 	int i;
1181 	int len;
1182 	int num;
1183 	int err = 0;
1184 
1185 	if (!pskb_may_pull(skb, sizeof(*icmp6h)))
1186 		return -EINVAL;
1187 
1188 	icmp6h = icmp6_hdr(skb);
1189 	num = ntohs(icmp6h->icmp6_dataun.un_data16[1]);
1190 	len = skb_transport_offset(skb) + sizeof(*icmp6h);
1191 
1192 	for (i = 0; i < num; i++) {
1193 		__be16 *nsrcs, _nsrcs;
1194 
1195 		nsrcs = skb_header_pointer(skb,
1196 					   len + offsetof(struct mld2_grec,
1197 							  grec_nsrcs),
1198 					   sizeof(_nsrcs), &_nsrcs);
1199 		if (!nsrcs)
1200 			return -EINVAL;
1201 
1202 		if (!pskb_may_pull(skb,
1203 				   len + sizeof(*grec) +
1204 				   sizeof(struct in6_addr) * ntohs(*nsrcs)))
1205 			return -EINVAL;
1206 
1207 		grec = (struct mld2_grec *)(skb->data + len);
1208 		len += sizeof(*grec) +
1209 		       sizeof(struct in6_addr) * ntohs(*nsrcs);
1210 
1211 		/* We treat these as MLDv1 reports for now. */
1212 		switch (grec->grec_type) {
1213 		case MLD2_MODE_IS_INCLUDE:
1214 		case MLD2_MODE_IS_EXCLUDE:
1215 		case MLD2_CHANGE_TO_INCLUDE:
1216 		case MLD2_CHANGE_TO_EXCLUDE:
1217 		case MLD2_ALLOW_NEW_SOURCES:
1218 		case MLD2_BLOCK_OLD_SOURCES:
1219 			break;
1220 
1221 		default:
1222 			continue;
1223 		}
1224 
1225 		src = eth_hdr(skb)->h_source;
1226 		if ((grec->grec_type == MLD2_CHANGE_TO_INCLUDE ||
1227 		     grec->grec_type == MLD2_MODE_IS_INCLUDE) &&
1228 		    ntohs(*nsrcs) == 0) {
1229 			br_ip6_multicast_leave_group(br, port, &grec->grec_mca,
1230 						     vid, src);
1231 		} else {
1232 			err = br_ip6_multicast_add_group(br, port,
1233 							 &grec->grec_mca, vid,
1234 							 src);
1235 			if (err)
1236 				break;
1237 		}
1238 	}
1239 
1240 	return err;
1241 }
1242 #endif
1243 
1244 static bool br_ip4_multicast_select_querier(struct net_bridge *br,
1245 					    struct net_bridge_port *port,
1246 					    __be32 saddr)
1247 {
1248 	if (!timer_pending(&br->ip4_own_query.timer) &&
1249 	    !timer_pending(&br->ip4_other_query.timer))
1250 		goto update;
1251 
1252 	if (!br->ip4_querier.addr.u.ip4)
1253 		goto update;
1254 
1255 	if (ntohl(saddr) <= ntohl(br->ip4_querier.addr.u.ip4))
1256 		goto update;
1257 
1258 	return false;
1259 
1260 update:
1261 	br->ip4_querier.addr.u.ip4 = saddr;
1262 
1263 	/* update protected by general multicast_lock by caller */
1264 	rcu_assign_pointer(br->ip4_querier.port, port);
1265 
1266 	return true;
1267 }
1268 
1269 #if IS_ENABLED(CONFIG_IPV6)
1270 static bool br_ip6_multicast_select_querier(struct net_bridge *br,
1271 					    struct net_bridge_port *port,
1272 					    struct in6_addr *saddr)
1273 {
1274 	if (!timer_pending(&br->ip6_own_query.timer) &&
1275 	    !timer_pending(&br->ip6_other_query.timer))
1276 		goto update;
1277 
1278 	if (ipv6_addr_cmp(saddr, &br->ip6_querier.addr.u.ip6) <= 0)
1279 		goto update;
1280 
1281 	return false;
1282 
1283 update:
1284 	br->ip6_querier.addr.u.ip6 = *saddr;
1285 
1286 	/* update protected by general multicast_lock by caller */
1287 	rcu_assign_pointer(br->ip6_querier.port, port);
1288 
1289 	return true;
1290 }
1291 #endif
1292 
1293 static bool br_multicast_select_querier(struct net_bridge *br,
1294 					struct net_bridge_port *port,
1295 					struct br_ip *saddr)
1296 {
1297 	switch (saddr->proto) {
1298 	case htons(ETH_P_IP):
1299 		return br_ip4_multicast_select_querier(br, port, saddr->u.ip4);
1300 #if IS_ENABLED(CONFIG_IPV6)
1301 	case htons(ETH_P_IPV6):
1302 		return br_ip6_multicast_select_querier(br, port, &saddr->u.ip6);
1303 #endif
1304 	}
1305 
1306 	return false;
1307 }
1308 
1309 static void
1310 br_multicast_update_query_timer(struct net_bridge *br,
1311 				struct bridge_mcast_other_query *query,
1312 				unsigned long max_delay)
1313 {
1314 	if (!timer_pending(&query->timer))
1315 		query->delay_time = jiffies + max_delay;
1316 
1317 	mod_timer(&query->timer, jiffies + br->multicast_querier_interval);
1318 }
1319 
1320 static void br_port_mc_router_state_change(struct net_bridge_port *p,
1321 					   bool is_mc_router)
1322 {
1323 	struct switchdev_attr attr = {
1324 		.orig_dev = p->dev,
1325 		.id = SWITCHDEV_ATTR_ID_PORT_MROUTER,
1326 		.flags = SWITCHDEV_F_DEFER,
1327 		.u.mrouter = is_mc_router,
1328 	};
1329 
1330 	switchdev_port_attr_set(p->dev, &attr);
1331 }
1332 
1333 /*
1334  * Add port to router_list
1335  *  list is maintained ordered by pointer value
1336  *  and locked by br->multicast_lock and RCU
1337  */
1338 static void br_multicast_add_router(struct net_bridge *br,
1339 				    struct net_bridge_port *port)
1340 {
1341 	struct net_bridge_port *p;
1342 	struct hlist_node *slot = NULL;
1343 
1344 	if (!hlist_unhashed(&port->rlist))
1345 		return;
1346 
1347 	hlist_for_each_entry(p, &br->router_list, rlist) {
1348 		if ((unsigned long) port >= (unsigned long) p)
1349 			break;
1350 		slot = &p->rlist;
1351 	}
1352 
1353 	if (slot)
1354 		hlist_add_behind_rcu(&port->rlist, slot);
1355 	else
1356 		hlist_add_head_rcu(&port->rlist, &br->router_list);
1357 	br_rtr_notify(br->dev, port, RTM_NEWMDB);
1358 	br_port_mc_router_state_change(port, true);
1359 }
1360 
1361 static void br_multicast_mark_router(struct net_bridge *br,
1362 				     struct net_bridge_port *port)
1363 {
1364 	unsigned long now = jiffies;
1365 
1366 	if (!port) {
1367 		if (br->multicast_router == MDB_RTR_TYPE_TEMP_QUERY)
1368 			mod_timer(&br->multicast_router_timer,
1369 				  now + br->multicast_querier_interval);
1370 		return;
1371 	}
1372 
1373 	if (port->multicast_router == MDB_RTR_TYPE_DISABLED ||
1374 	    port->multicast_router == MDB_RTR_TYPE_PERM)
1375 		return;
1376 
1377 	br_multicast_add_router(br, port);
1378 
1379 	mod_timer(&port->multicast_router_timer,
1380 		  now + br->multicast_querier_interval);
1381 }
1382 
1383 static void br_multicast_query_received(struct net_bridge *br,
1384 					struct net_bridge_port *port,
1385 					struct bridge_mcast_other_query *query,
1386 					struct br_ip *saddr,
1387 					unsigned long max_delay)
1388 {
1389 	if (!br_multicast_select_querier(br, port, saddr))
1390 		return;
1391 
1392 	br_multicast_update_query_timer(br, query, max_delay);
1393 	br_multicast_mark_router(br, port);
1394 }
1395 
1396 static int br_ip4_multicast_query(struct net_bridge *br,
1397 				  struct net_bridge_port *port,
1398 				  struct sk_buff *skb,
1399 				  u16 vid)
1400 {
1401 	const struct iphdr *iph = ip_hdr(skb);
1402 	struct igmphdr *ih = igmp_hdr(skb);
1403 	struct net_bridge_mdb_entry *mp;
1404 	struct igmpv3_query *ih3;
1405 	struct net_bridge_port_group *p;
1406 	struct net_bridge_port_group __rcu **pp;
1407 	struct br_ip saddr;
1408 	unsigned long max_delay;
1409 	unsigned long now = jiffies;
1410 	unsigned int offset = skb_transport_offset(skb);
1411 	__be32 group;
1412 	int err = 0;
1413 
1414 	spin_lock(&br->multicast_lock);
1415 	if (!netif_running(br->dev) ||
1416 	    (port && port->state == BR_STATE_DISABLED))
1417 		goto out;
1418 
1419 	group = ih->group;
1420 
1421 	if (skb->len == offset + sizeof(*ih)) {
1422 		max_delay = ih->code * (HZ / IGMP_TIMER_SCALE);
1423 
1424 		if (!max_delay) {
1425 			max_delay = 10 * HZ;
1426 			group = 0;
1427 		}
1428 	} else if (skb->len >= offset + sizeof(*ih3)) {
1429 		ih3 = igmpv3_query_hdr(skb);
1430 		if (ih3->nsrcs)
1431 			goto out;
1432 
1433 		max_delay = ih3->code ?
1434 			    IGMPV3_MRC(ih3->code) * (HZ / IGMP_TIMER_SCALE) : 1;
1435 	} else {
1436 		goto out;
1437 	}
1438 
1439 	if (!group) {
1440 		saddr.proto = htons(ETH_P_IP);
1441 		saddr.u.ip4 = iph->saddr;
1442 
1443 		br_multicast_query_received(br, port, &br->ip4_other_query,
1444 					    &saddr, max_delay);
1445 		goto out;
1446 	}
1447 
1448 	mp = br_mdb_ip4_get(mlock_dereference(br->mdb, br), group, vid);
1449 	if (!mp)
1450 		goto out;
1451 
1452 	max_delay *= br->multicast_last_member_count;
1453 
1454 	if (mp->mglist &&
1455 	    (timer_pending(&mp->timer) ?
1456 	     time_after(mp->timer.expires, now + max_delay) :
1457 	     try_to_del_timer_sync(&mp->timer) >= 0))
1458 		mod_timer(&mp->timer, now + max_delay);
1459 
1460 	for (pp = &mp->ports;
1461 	     (p = mlock_dereference(*pp, br)) != NULL;
1462 	     pp = &p->next) {
1463 		if (timer_pending(&p->timer) ?
1464 		    time_after(p->timer.expires, now + max_delay) :
1465 		    try_to_del_timer_sync(&p->timer) >= 0)
1466 			mod_timer(&p->timer, now + max_delay);
1467 	}
1468 
1469 out:
1470 	spin_unlock(&br->multicast_lock);
1471 	return err;
1472 }
1473 
1474 #if IS_ENABLED(CONFIG_IPV6)
1475 static int br_ip6_multicast_query(struct net_bridge *br,
1476 				  struct net_bridge_port *port,
1477 				  struct sk_buff *skb,
1478 				  u16 vid)
1479 {
1480 	const struct ipv6hdr *ip6h = ipv6_hdr(skb);
1481 	struct mld_msg *mld;
1482 	struct net_bridge_mdb_entry *mp;
1483 	struct mld2_query *mld2q;
1484 	struct net_bridge_port_group *p;
1485 	struct net_bridge_port_group __rcu **pp;
1486 	struct br_ip saddr;
1487 	unsigned long max_delay;
1488 	unsigned long now = jiffies;
1489 	unsigned int offset = skb_transport_offset(skb);
1490 	const struct in6_addr *group = NULL;
1491 	bool is_general_query;
1492 	int err = 0;
1493 
1494 	spin_lock(&br->multicast_lock);
1495 	if (!netif_running(br->dev) ||
1496 	    (port && port->state == BR_STATE_DISABLED))
1497 		goto out;
1498 
1499 	if (skb->len == offset + sizeof(*mld)) {
1500 		if (!pskb_may_pull(skb, offset + sizeof(*mld))) {
1501 			err = -EINVAL;
1502 			goto out;
1503 		}
1504 		mld = (struct mld_msg *) icmp6_hdr(skb);
1505 		max_delay = msecs_to_jiffies(ntohs(mld->mld_maxdelay));
1506 		if (max_delay)
1507 			group = &mld->mld_mca;
1508 	} else {
1509 		if (!pskb_may_pull(skb, offset + sizeof(*mld2q))) {
1510 			err = -EINVAL;
1511 			goto out;
1512 		}
1513 		mld2q = (struct mld2_query *)icmp6_hdr(skb);
1514 		if (!mld2q->mld2q_nsrcs)
1515 			group = &mld2q->mld2q_mca;
1516 
1517 		max_delay = max(msecs_to_jiffies(mldv2_mrc(mld2q)), 1UL);
1518 	}
1519 
1520 	is_general_query = group && ipv6_addr_any(group);
1521 
1522 	if (is_general_query) {
1523 		saddr.proto = htons(ETH_P_IPV6);
1524 		saddr.u.ip6 = ip6h->saddr;
1525 
1526 		br_multicast_query_received(br, port, &br->ip6_other_query,
1527 					    &saddr, max_delay);
1528 		goto out;
1529 	} else if (!group) {
1530 		goto out;
1531 	}
1532 
1533 	mp = br_mdb_ip6_get(mlock_dereference(br->mdb, br), group, vid);
1534 	if (!mp)
1535 		goto out;
1536 
1537 	max_delay *= br->multicast_last_member_count;
1538 	if (mp->mglist &&
1539 	    (timer_pending(&mp->timer) ?
1540 	     time_after(mp->timer.expires, now + max_delay) :
1541 	     try_to_del_timer_sync(&mp->timer) >= 0))
1542 		mod_timer(&mp->timer, now + max_delay);
1543 
1544 	for (pp = &mp->ports;
1545 	     (p = mlock_dereference(*pp, br)) != NULL;
1546 	     pp = &p->next) {
1547 		if (timer_pending(&p->timer) ?
1548 		    time_after(p->timer.expires, now + max_delay) :
1549 		    try_to_del_timer_sync(&p->timer) >= 0)
1550 			mod_timer(&p->timer, now + max_delay);
1551 	}
1552 
1553 out:
1554 	spin_unlock(&br->multicast_lock);
1555 	return err;
1556 }
1557 #endif
1558 
1559 static void
1560 br_multicast_leave_group(struct net_bridge *br,
1561 			 struct net_bridge_port *port,
1562 			 struct br_ip *group,
1563 			 struct bridge_mcast_other_query *other_query,
1564 			 struct bridge_mcast_own_query *own_query,
1565 			 const unsigned char *src)
1566 {
1567 	struct net_bridge_mdb_htable *mdb;
1568 	struct net_bridge_mdb_entry *mp;
1569 	struct net_bridge_port_group *p;
1570 	unsigned long now;
1571 	unsigned long time;
1572 
1573 	spin_lock(&br->multicast_lock);
1574 	if (!netif_running(br->dev) ||
1575 	    (port && port->state == BR_STATE_DISABLED))
1576 		goto out;
1577 
1578 	mdb = mlock_dereference(br->mdb, br);
1579 	mp = br_mdb_ip_get(mdb, group);
1580 	if (!mp)
1581 		goto out;
1582 
1583 	if (port && (port->flags & BR_MULTICAST_FAST_LEAVE)) {
1584 		struct net_bridge_port_group __rcu **pp;
1585 
1586 		for (pp = &mp->ports;
1587 		     (p = mlock_dereference(*pp, br)) != NULL;
1588 		     pp = &p->next) {
1589 			if (!br_port_group_equal(p, port, src))
1590 				continue;
1591 
1592 			rcu_assign_pointer(*pp, p->next);
1593 			hlist_del_init(&p->mglist);
1594 			del_timer(&p->timer);
1595 			call_rcu_bh(&p->rcu, br_multicast_free_pg);
1596 			br_mdb_notify(br->dev, port, group, RTM_DELMDB,
1597 				      p->flags);
1598 
1599 			if (!mp->ports && !mp->mglist &&
1600 			    netif_running(br->dev))
1601 				mod_timer(&mp->timer, jiffies);
1602 		}
1603 		goto out;
1604 	}
1605 
1606 	if (timer_pending(&other_query->timer))
1607 		goto out;
1608 
1609 	if (br->multicast_querier) {
1610 		__br_multicast_send_query(br, port, &mp->addr);
1611 
1612 		time = jiffies + br->multicast_last_member_count *
1613 				 br->multicast_last_member_interval;
1614 
1615 		mod_timer(&own_query->timer, time);
1616 
1617 		for (p = mlock_dereference(mp->ports, br);
1618 		     p != NULL;
1619 		     p = mlock_dereference(p->next, br)) {
1620 			if (!br_port_group_equal(p, port, src))
1621 				continue;
1622 
1623 			if (!hlist_unhashed(&p->mglist) &&
1624 			    (timer_pending(&p->timer) ?
1625 			     time_after(p->timer.expires, time) :
1626 			     try_to_del_timer_sync(&p->timer) >= 0)) {
1627 				mod_timer(&p->timer, time);
1628 			}
1629 
1630 			break;
1631 		}
1632 	}
1633 
1634 	now = jiffies;
1635 	time = now + br->multicast_last_member_count *
1636 		     br->multicast_last_member_interval;
1637 
1638 	if (!port) {
1639 		if (mp->mglist &&
1640 		    (timer_pending(&mp->timer) ?
1641 		     time_after(mp->timer.expires, time) :
1642 		     try_to_del_timer_sync(&mp->timer) >= 0)) {
1643 			mod_timer(&mp->timer, time);
1644 		}
1645 
1646 		goto out;
1647 	}
1648 
1649 	for (p = mlock_dereference(mp->ports, br);
1650 	     p != NULL;
1651 	     p = mlock_dereference(p->next, br)) {
1652 		if (p->port != port)
1653 			continue;
1654 
1655 		if (!hlist_unhashed(&p->mglist) &&
1656 		    (timer_pending(&p->timer) ?
1657 		     time_after(p->timer.expires, time) :
1658 		     try_to_del_timer_sync(&p->timer) >= 0)) {
1659 			mod_timer(&p->timer, time);
1660 		}
1661 
1662 		break;
1663 	}
1664 out:
1665 	spin_unlock(&br->multicast_lock);
1666 }
1667 
1668 static void br_ip4_multicast_leave_group(struct net_bridge *br,
1669 					 struct net_bridge_port *port,
1670 					 __be32 group,
1671 					 __u16 vid,
1672 					 const unsigned char *src)
1673 {
1674 	struct br_ip br_group;
1675 	struct bridge_mcast_own_query *own_query;
1676 
1677 	if (ipv4_is_local_multicast(group))
1678 		return;
1679 
1680 	own_query = port ? &port->ip4_own_query : &br->ip4_own_query;
1681 
1682 	br_group.u.ip4 = group;
1683 	br_group.proto = htons(ETH_P_IP);
1684 	br_group.vid = vid;
1685 
1686 	br_multicast_leave_group(br, port, &br_group, &br->ip4_other_query,
1687 				 own_query, src);
1688 }
1689 
1690 #if IS_ENABLED(CONFIG_IPV6)
1691 static void br_ip6_multicast_leave_group(struct net_bridge *br,
1692 					 struct net_bridge_port *port,
1693 					 const struct in6_addr *group,
1694 					 __u16 vid,
1695 					 const unsigned char *src)
1696 {
1697 	struct br_ip br_group;
1698 	struct bridge_mcast_own_query *own_query;
1699 
1700 	if (ipv6_addr_is_ll_all_nodes(group))
1701 		return;
1702 
1703 	own_query = port ? &port->ip6_own_query : &br->ip6_own_query;
1704 
1705 	br_group.u.ip6 = *group;
1706 	br_group.proto = htons(ETH_P_IPV6);
1707 	br_group.vid = vid;
1708 
1709 	br_multicast_leave_group(br, port, &br_group, &br->ip6_other_query,
1710 				 own_query, src);
1711 }
1712 #endif
1713 
1714 static void br_multicast_err_count(const struct net_bridge *br,
1715 				   const struct net_bridge_port *p,
1716 				   __be16 proto)
1717 {
1718 	struct bridge_mcast_stats __percpu *stats;
1719 	struct bridge_mcast_stats *pstats;
1720 
1721 	if (!br->multicast_stats_enabled)
1722 		return;
1723 
1724 	if (p)
1725 		stats = p->mcast_stats;
1726 	else
1727 		stats = br->mcast_stats;
1728 	if (WARN_ON(!stats))
1729 		return;
1730 
1731 	pstats = this_cpu_ptr(stats);
1732 
1733 	u64_stats_update_begin(&pstats->syncp);
1734 	switch (proto) {
1735 	case htons(ETH_P_IP):
1736 		pstats->mstats.igmp_parse_errors++;
1737 		break;
1738 #if IS_ENABLED(CONFIG_IPV6)
1739 	case htons(ETH_P_IPV6):
1740 		pstats->mstats.mld_parse_errors++;
1741 		break;
1742 #endif
1743 	}
1744 	u64_stats_update_end(&pstats->syncp);
1745 }
1746 
1747 static void br_multicast_pim(struct net_bridge *br,
1748 			     struct net_bridge_port *port,
1749 			     const struct sk_buff *skb)
1750 {
1751 	unsigned int offset = skb_transport_offset(skb);
1752 	struct pimhdr *pimhdr, _pimhdr;
1753 
1754 	pimhdr = skb_header_pointer(skb, offset, sizeof(_pimhdr), &_pimhdr);
1755 	if (!pimhdr || pim_hdr_version(pimhdr) != PIM_VERSION ||
1756 	    pim_hdr_type(pimhdr) != PIM_TYPE_HELLO)
1757 		return;
1758 
1759 	br_multicast_mark_router(br, port);
1760 }
1761 
1762 static int br_multicast_ipv4_rcv(struct net_bridge *br,
1763 				 struct net_bridge_port *port,
1764 				 struct sk_buff *skb,
1765 				 u16 vid)
1766 {
1767 	struct sk_buff *skb_trimmed = NULL;
1768 	const unsigned char *src;
1769 	struct igmphdr *ih;
1770 	int err;
1771 
1772 	err = ip_mc_check_igmp(skb, &skb_trimmed);
1773 
1774 	if (err == -ENOMSG) {
1775 		if (!ipv4_is_local_multicast(ip_hdr(skb)->daddr)) {
1776 			BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1777 		} else if (pim_ipv4_all_pim_routers(ip_hdr(skb)->daddr)) {
1778 			if (ip_hdr(skb)->protocol == IPPROTO_PIM)
1779 				br_multicast_pim(br, port, skb);
1780 		}
1781 		return 0;
1782 	} else if (err < 0) {
1783 		br_multicast_err_count(br, port, skb->protocol);
1784 		return err;
1785 	}
1786 
1787 	ih = igmp_hdr(skb);
1788 	src = eth_hdr(skb)->h_source;
1789 	BR_INPUT_SKB_CB(skb)->igmp = ih->type;
1790 
1791 	switch (ih->type) {
1792 	case IGMP_HOST_MEMBERSHIP_REPORT:
1793 	case IGMPV2_HOST_MEMBERSHIP_REPORT:
1794 		BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1795 		err = br_ip4_multicast_add_group(br, port, ih->group, vid, src);
1796 		break;
1797 	case IGMPV3_HOST_MEMBERSHIP_REPORT:
1798 		err = br_ip4_multicast_igmp3_report(br, port, skb_trimmed, vid);
1799 		break;
1800 	case IGMP_HOST_MEMBERSHIP_QUERY:
1801 		err = br_ip4_multicast_query(br, port, skb_trimmed, vid);
1802 		break;
1803 	case IGMP_HOST_LEAVE_MESSAGE:
1804 		br_ip4_multicast_leave_group(br, port, ih->group, vid, src);
1805 		break;
1806 	}
1807 
1808 	if (skb_trimmed && skb_trimmed != skb)
1809 		kfree_skb(skb_trimmed);
1810 
1811 	br_multicast_count(br, port, skb, BR_INPUT_SKB_CB(skb)->igmp,
1812 			   BR_MCAST_DIR_RX);
1813 
1814 	return err;
1815 }
1816 
1817 #if IS_ENABLED(CONFIG_IPV6)
1818 static int br_multicast_ipv6_rcv(struct net_bridge *br,
1819 				 struct net_bridge_port *port,
1820 				 struct sk_buff *skb,
1821 				 u16 vid)
1822 {
1823 	struct sk_buff *skb_trimmed = NULL;
1824 	const unsigned char *src;
1825 	struct mld_msg *mld;
1826 	int err;
1827 
1828 	err = ipv6_mc_check_mld(skb, &skb_trimmed);
1829 
1830 	if (err == -ENOMSG) {
1831 		if (!ipv6_addr_is_ll_all_nodes(&ipv6_hdr(skb)->daddr))
1832 			BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1833 		return 0;
1834 	} else if (err < 0) {
1835 		br_multicast_err_count(br, port, skb->protocol);
1836 		return err;
1837 	}
1838 
1839 	mld = (struct mld_msg *)skb_transport_header(skb);
1840 	BR_INPUT_SKB_CB(skb)->igmp = mld->mld_type;
1841 
1842 	switch (mld->mld_type) {
1843 	case ICMPV6_MGM_REPORT:
1844 		src = eth_hdr(skb)->h_source;
1845 		BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1846 		err = br_ip6_multicast_add_group(br, port, &mld->mld_mca, vid,
1847 						 src);
1848 		break;
1849 	case ICMPV6_MLD2_REPORT:
1850 		err = br_ip6_multicast_mld2_report(br, port, skb_trimmed, vid);
1851 		break;
1852 	case ICMPV6_MGM_QUERY:
1853 		err = br_ip6_multicast_query(br, port, skb_trimmed, vid);
1854 		break;
1855 	case ICMPV6_MGM_REDUCTION:
1856 		src = eth_hdr(skb)->h_source;
1857 		br_ip6_multicast_leave_group(br, port, &mld->mld_mca, vid, src);
1858 		break;
1859 	}
1860 
1861 	if (skb_trimmed && skb_trimmed != skb)
1862 		kfree_skb(skb_trimmed);
1863 
1864 	br_multicast_count(br, port, skb, BR_INPUT_SKB_CB(skb)->igmp,
1865 			   BR_MCAST_DIR_RX);
1866 
1867 	return err;
1868 }
1869 #endif
1870 
1871 int br_multicast_rcv(struct net_bridge *br, struct net_bridge_port *port,
1872 		     struct sk_buff *skb, u16 vid)
1873 {
1874 	int ret = 0;
1875 
1876 	BR_INPUT_SKB_CB(skb)->igmp = 0;
1877 	BR_INPUT_SKB_CB(skb)->mrouters_only = 0;
1878 
1879 	if (br->multicast_disabled)
1880 		return 0;
1881 
1882 	switch (skb->protocol) {
1883 	case htons(ETH_P_IP):
1884 		ret = br_multicast_ipv4_rcv(br, port, skb, vid);
1885 		break;
1886 #if IS_ENABLED(CONFIG_IPV6)
1887 	case htons(ETH_P_IPV6):
1888 		ret = br_multicast_ipv6_rcv(br, port, skb, vid);
1889 		break;
1890 #endif
1891 	}
1892 
1893 	return ret;
1894 }
1895 
1896 static void br_multicast_query_expired(struct net_bridge *br,
1897 				       struct bridge_mcast_own_query *query,
1898 				       struct bridge_mcast_querier *querier)
1899 {
1900 	spin_lock(&br->multicast_lock);
1901 	if (query->startup_sent < br->multicast_startup_query_count)
1902 		query->startup_sent++;
1903 
1904 	RCU_INIT_POINTER(querier->port, NULL);
1905 	br_multicast_send_query(br, NULL, query);
1906 	spin_unlock(&br->multicast_lock);
1907 }
1908 
1909 static void br_ip4_multicast_query_expired(unsigned long data)
1910 {
1911 	struct net_bridge *br = (void *)data;
1912 
1913 	br_multicast_query_expired(br, &br->ip4_own_query, &br->ip4_querier);
1914 }
1915 
1916 #if IS_ENABLED(CONFIG_IPV6)
1917 static void br_ip6_multicast_query_expired(unsigned long data)
1918 {
1919 	struct net_bridge *br = (void *)data;
1920 
1921 	br_multicast_query_expired(br, &br->ip6_own_query, &br->ip6_querier);
1922 }
1923 #endif
1924 
1925 void br_multicast_init(struct net_bridge *br)
1926 {
1927 	br->hash_elasticity = 4;
1928 	br->hash_max = 512;
1929 
1930 	br->multicast_router = MDB_RTR_TYPE_TEMP_QUERY;
1931 	br->multicast_querier = 0;
1932 	br->multicast_query_use_ifaddr = 0;
1933 	br->multicast_last_member_count = 2;
1934 	br->multicast_startup_query_count = 2;
1935 
1936 	br->multicast_last_member_interval = HZ;
1937 	br->multicast_query_response_interval = 10 * HZ;
1938 	br->multicast_startup_query_interval = 125 * HZ / 4;
1939 	br->multicast_query_interval = 125 * HZ;
1940 	br->multicast_querier_interval = 255 * HZ;
1941 	br->multicast_membership_interval = 260 * HZ;
1942 
1943 	br->ip4_other_query.delay_time = 0;
1944 	br->ip4_querier.port = NULL;
1945 	br->multicast_igmp_version = 2;
1946 #if IS_ENABLED(CONFIG_IPV6)
1947 	br->multicast_mld_version = 1;
1948 	br->ip6_other_query.delay_time = 0;
1949 	br->ip6_querier.port = NULL;
1950 #endif
1951 	br->has_ipv6_addr = 1;
1952 
1953 	spin_lock_init(&br->multicast_lock);
1954 	setup_timer(&br->multicast_router_timer,
1955 		    br_multicast_local_router_expired, 0);
1956 	setup_timer(&br->ip4_other_query.timer,
1957 		    br_ip4_multicast_querier_expired, (unsigned long)br);
1958 	setup_timer(&br->ip4_own_query.timer, br_ip4_multicast_query_expired,
1959 		    (unsigned long)br);
1960 #if IS_ENABLED(CONFIG_IPV6)
1961 	setup_timer(&br->ip6_other_query.timer,
1962 		    br_ip6_multicast_querier_expired, (unsigned long)br);
1963 	setup_timer(&br->ip6_own_query.timer, br_ip6_multicast_query_expired,
1964 		    (unsigned long)br);
1965 #endif
1966 }
1967 
1968 static void __br_multicast_open(struct net_bridge *br,
1969 				struct bridge_mcast_own_query *query)
1970 {
1971 	query->startup_sent = 0;
1972 
1973 	if (br->multicast_disabled)
1974 		return;
1975 
1976 	mod_timer(&query->timer, jiffies);
1977 }
1978 
1979 void br_multicast_open(struct net_bridge *br)
1980 {
1981 	__br_multicast_open(br, &br->ip4_own_query);
1982 #if IS_ENABLED(CONFIG_IPV6)
1983 	__br_multicast_open(br, &br->ip6_own_query);
1984 #endif
1985 }
1986 
1987 void br_multicast_stop(struct net_bridge *br)
1988 {
1989 	del_timer_sync(&br->multicast_router_timer);
1990 	del_timer_sync(&br->ip4_other_query.timer);
1991 	del_timer_sync(&br->ip4_own_query.timer);
1992 #if IS_ENABLED(CONFIG_IPV6)
1993 	del_timer_sync(&br->ip6_other_query.timer);
1994 	del_timer_sync(&br->ip6_own_query.timer);
1995 #endif
1996 }
1997 
1998 void br_multicast_dev_del(struct net_bridge *br)
1999 {
2000 	struct net_bridge_mdb_htable *mdb;
2001 	struct net_bridge_mdb_entry *mp;
2002 	struct hlist_node *n;
2003 	u32 ver;
2004 	int i;
2005 
2006 	spin_lock_bh(&br->multicast_lock);
2007 	mdb = mlock_dereference(br->mdb, br);
2008 	if (!mdb)
2009 		goto out;
2010 
2011 	br->mdb = NULL;
2012 
2013 	ver = mdb->ver;
2014 	for (i = 0; i < mdb->max; i++) {
2015 		hlist_for_each_entry_safe(mp, n, &mdb->mhash[i],
2016 					  hlist[ver]) {
2017 			del_timer(&mp->timer);
2018 			call_rcu_bh(&mp->rcu, br_multicast_free_group);
2019 		}
2020 	}
2021 
2022 	if (mdb->old) {
2023 		spin_unlock_bh(&br->multicast_lock);
2024 		rcu_barrier_bh();
2025 		spin_lock_bh(&br->multicast_lock);
2026 		WARN_ON(mdb->old);
2027 	}
2028 
2029 	mdb->old = mdb;
2030 	call_rcu_bh(&mdb->rcu, br_mdb_free);
2031 
2032 out:
2033 	spin_unlock_bh(&br->multicast_lock);
2034 }
2035 
2036 int br_multicast_set_router(struct net_bridge *br, unsigned long val)
2037 {
2038 	int err = -EINVAL;
2039 
2040 	spin_lock_bh(&br->multicast_lock);
2041 
2042 	switch (val) {
2043 	case MDB_RTR_TYPE_DISABLED:
2044 	case MDB_RTR_TYPE_PERM:
2045 		del_timer(&br->multicast_router_timer);
2046 		/* fall through */
2047 	case MDB_RTR_TYPE_TEMP_QUERY:
2048 		br->multicast_router = val;
2049 		err = 0;
2050 		break;
2051 	}
2052 
2053 	spin_unlock_bh(&br->multicast_lock);
2054 
2055 	return err;
2056 }
2057 
2058 static void __del_port_router(struct net_bridge_port *p)
2059 {
2060 	if (hlist_unhashed(&p->rlist))
2061 		return;
2062 	hlist_del_init_rcu(&p->rlist);
2063 	br_rtr_notify(p->br->dev, p, RTM_DELMDB);
2064 	br_port_mc_router_state_change(p, false);
2065 
2066 	/* don't allow timer refresh */
2067 	if (p->multicast_router == MDB_RTR_TYPE_TEMP)
2068 		p->multicast_router = MDB_RTR_TYPE_TEMP_QUERY;
2069 }
2070 
2071 int br_multicast_set_port_router(struct net_bridge_port *p, unsigned long val)
2072 {
2073 	struct net_bridge *br = p->br;
2074 	unsigned long now = jiffies;
2075 	int err = -EINVAL;
2076 
2077 	spin_lock(&br->multicast_lock);
2078 	if (p->multicast_router == val) {
2079 		/* Refresh the temp router port timer */
2080 		if (p->multicast_router == MDB_RTR_TYPE_TEMP)
2081 			mod_timer(&p->multicast_router_timer,
2082 				  now + br->multicast_querier_interval);
2083 		err = 0;
2084 		goto unlock;
2085 	}
2086 	switch (val) {
2087 	case MDB_RTR_TYPE_DISABLED:
2088 		p->multicast_router = MDB_RTR_TYPE_DISABLED;
2089 		__del_port_router(p);
2090 		del_timer(&p->multicast_router_timer);
2091 		break;
2092 	case MDB_RTR_TYPE_TEMP_QUERY:
2093 		p->multicast_router = MDB_RTR_TYPE_TEMP_QUERY;
2094 		__del_port_router(p);
2095 		break;
2096 	case MDB_RTR_TYPE_PERM:
2097 		p->multicast_router = MDB_RTR_TYPE_PERM;
2098 		del_timer(&p->multicast_router_timer);
2099 		br_multicast_add_router(br, p);
2100 		break;
2101 	case MDB_RTR_TYPE_TEMP:
2102 		p->multicast_router = MDB_RTR_TYPE_TEMP;
2103 		br_multicast_mark_router(br, p);
2104 		break;
2105 	default:
2106 		goto unlock;
2107 	}
2108 	err = 0;
2109 unlock:
2110 	spin_unlock(&br->multicast_lock);
2111 
2112 	return err;
2113 }
2114 
2115 static void br_multicast_start_querier(struct net_bridge *br,
2116 				       struct bridge_mcast_own_query *query)
2117 {
2118 	struct net_bridge_port *port;
2119 
2120 	__br_multicast_open(br, query);
2121 
2122 	list_for_each_entry(port, &br->port_list, list) {
2123 		if (port->state == BR_STATE_DISABLED ||
2124 		    port->state == BR_STATE_BLOCKING)
2125 			continue;
2126 
2127 		if (query == &br->ip4_own_query)
2128 			br_multicast_enable(&port->ip4_own_query);
2129 #if IS_ENABLED(CONFIG_IPV6)
2130 		else
2131 			br_multicast_enable(&port->ip6_own_query);
2132 #endif
2133 	}
2134 }
2135 
2136 int br_multicast_toggle(struct net_bridge *br, unsigned long val)
2137 {
2138 	struct net_bridge_mdb_htable *mdb;
2139 	struct net_bridge_port *port;
2140 	int err = 0;
2141 
2142 	spin_lock_bh(&br->multicast_lock);
2143 	if (br->multicast_disabled == !val)
2144 		goto unlock;
2145 
2146 	br_mc_disabled_update(br->dev, !val);
2147 	br->multicast_disabled = !val;
2148 	if (br->multicast_disabled)
2149 		goto unlock;
2150 
2151 	if (!netif_running(br->dev))
2152 		goto unlock;
2153 
2154 	mdb = mlock_dereference(br->mdb, br);
2155 	if (mdb) {
2156 		if (mdb->old) {
2157 			err = -EEXIST;
2158 rollback:
2159 			br->multicast_disabled = !!val;
2160 			goto unlock;
2161 		}
2162 
2163 		err = br_mdb_rehash(&br->mdb, mdb->max,
2164 				    br->hash_elasticity);
2165 		if (err)
2166 			goto rollback;
2167 	}
2168 
2169 	br_multicast_open(br);
2170 	list_for_each_entry(port, &br->port_list, list)
2171 		__br_multicast_enable_port(port);
2172 
2173 unlock:
2174 	spin_unlock_bh(&br->multicast_lock);
2175 
2176 	return err;
2177 }
2178 
2179 bool br_multicast_enabled(const struct net_device *dev)
2180 {
2181 	struct net_bridge *br = netdev_priv(dev);
2182 
2183 	return !br->multicast_disabled;
2184 }
2185 EXPORT_SYMBOL_GPL(br_multicast_enabled);
2186 
2187 int br_multicast_set_querier(struct net_bridge *br, unsigned long val)
2188 {
2189 	unsigned long max_delay;
2190 
2191 	val = !!val;
2192 
2193 	spin_lock_bh(&br->multicast_lock);
2194 	if (br->multicast_querier == val)
2195 		goto unlock;
2196 
2197 	br->multicast_querier = val;
2198 	if (!val)
2199 		goto unlock;
2200 
2201 	max_delay = br->multicast_query_response_interval;
2202 
2203 	if (!timer_pending(&br->ip4_other_query.timer))
2204 		br->ip4_other_query.delay_time = jiffies + max_delay;
2205 
2206 	br_multicast_start_querier(br, &br->ip4_own_query);
2207 
2208 #if IS_ENABLED(CONFIG_IPV6)
2209 	if (!timer_pending(&br->ip6_other_query.timer))
2210 		br->ip6_other_query.delay_time = jiffies + max_delay;
2211 
2212 	br_multicast_start_querier(br, &br->ip6_own_query);
2213 #endif
2214 
2215 unlock:
2216 	spin_unlock_bh(&br->multicast_lock);
2217 
2218 	return 0;
2219 }
2220 
2221 int br_multicast_set_hash_max(struct net_bridge *br, unsigned long val)
2222 {
2223 	int err = -EINVAL;
2224 	u32 old;
2225 	struct net_bridge_mdb_htable *mdb;
2226 
2227 	spin_lock_bh(&br->multicast_lock);
2228 	if (!is_power_of_2(val))
2229 		goto unlock;
2230 
2231 	mdb = mlock_dereference(br->mdb, br);
2232 	if (mdb && val < mdb->size)
2233 		goto unlock;
2234 
2235 	err = 0;
2236 
2237 	old = br->hash_max;
2238 	br->hash_max = val;
2239 
2240 	if (mdb) {
2241 		if (mdb->old) {
2242 			err = -EEXIST;
2243 rollback:
2244 			br->hash_max = old;
2245 			goto unlock;
2246 		}
2247 
2248 		err = br_mdb_rehash(&br->mdb, br->hash_max,
2249 				    br->hash_elasticity);
2250 		if (err)
2251 			goto rollback;
2252 	}
2253 
2254 unlock:
2255 	spin_unlock_bh(&br->multicast_lock);
2256 
2257 	return err;
2258 }
2259 
2260 int br_multicast_set_igmp_version(struct net_bridge *br, unsigned long val)
2261 {
2262 	/* Currently we support only version 2 and 3 */
2263 	switch (val) {
2264 	case 2:
2265 	case 3:
2266 		break;
2267 	default:
2268 		return -EINVAL;
2269 	}
2270 
2271 	spin_lock_bh(&br->multicast_lock);
2272 	br->multicast_igmp_version = val;
2273 	spin_unlock_bh(&br->multicast_lock);
2274 
2275 	return 0;
2276 }
2277 
2278 #if IS_ENABLED(CONFIG_IPV6)
2279 int br_multicast_set_mld_version(struct net_bridge *br, unsigned long val)
2280 {
2281 	/* Currently we support version 1 and 2 */
2282 	switch (val) {
2283 	case 1:
2284 	case 2:
2285 		break;
2286 	default:
2287 		return -EINVAL;
2288 	}
2289 
2290 	spin_lock_bh(&br->multicast_lock);
2291 	br->multicast_mld_version = val;
2292 	spin_unlock_bh(&br->multicast_lock);
2293 
2294 	return 0;
2295 }
2296 #endif
2297 
2298 /**
2299  * br_multicast_list_adjacent - Returns snooped multicast addresses
2300  * @dev:	The bridge port adjacent to which to retrieve addresses
2301  * @br_ip_list:	The list to store found, snooped multicast IP addresses in
2302  *
2303  * Creates a list of IP addresses (struct br_ip_list) sensed by the multicast
2304  * snooping feature on all bridge ports of dev's bridge device, excluding
2305  * the addresses from dev itself.
2306  *
2307  * Returns the number of items added to br_ip_list.
2308  *
2309  * Notes:
2310  * - br_ip_list needs to be initialized by caller
2311  * - br_ip_list might contain duplicates in the end
2312  *   (needs to be taken care of by caller)
2313  * - br_ip_list needs to be freed by caller
2314  */
2315 int br_multicast_list_adjacent(struct net_device *dev,
2316 			       struct list_head *br_ip_list)
2317 {
2318 	struct net_bridge *br;
2319 	struct net_bridge_port *port;
2320 	struct net_bridge_port_group *group;
2321 	struct br_ip_list *entry;
2322 	int count = 0;
2323 
2324 	rcu_read_lock();
2325 	if (!br_ip_list || !br_port_exists(dev))
2326 		goto unlock;
2327 
2328 	port = br_port_get_rcu(dev);
2329 	if (!port || !port->br)
2330 		goto unlock;
2331 
2332 	br = port->br;
2333 
2334 	list_for_each_entry_rcu(port, &br->port_list, list) {
2335 		if (!port->dev || port->dev == dev)
2336 			continue;
2337 
2338 		hlist_for_each_entry_rcu(group, &port->mglist, mglist) {
2339 			entry = kmalloc(sizeof(*entry), GFP_ATOMIC);
2340 			if (!entry)
2341 				goto unlock;
2342 
2343 			entry->addr = group->addr;
2344 			list_add(&entry->list, br_ip_list);
2345 			count++;
2346 		}
2347 	}
2348 
2349 unlock:
2350 	rcu_read_unlock();
2351 	return count;
2352 }
2353 EXPORT_SYMBOL_GPL(br_multicast_list_adjacent);
2354 
2355 /**
2356  * br_multicast_has_querier_anywhere - Checks for a querier on a bridge
2357  * @dev: The bridge port providing the bridge on which to check for a querier
2358  * @proto: The protocol family to check for: IGMP -> ETH_P_IP, MLD -> ETH_P_IPV6
2359  *
2360  * Checks whether the given interface has a bridge on top and if so returns
2361  * true if a valid querier exists anywhere on the bridged link layer.
2362  * Otherwise returns false.
2363  */
2364 bool br_multicast_has_querier_anywhere(struct net_device *dev, int proto)
2365 {
2366 	struct net_bridge *br;
2367 	struct net_bridge_port *port;
2368 	struct ethhdr eth;
2369 	bool ret = false;
2370 
2371 	rcu_read_lock();
2372 	if (!br_port_exists(dev))
2373 		goto unlock;
2374 
2375 	port = br_port_get_rcu(dev);
2376 	if (!port || !port->br)
2377 		goto unlock;
2378 
2379 	br = port->br;
2380 
2381 	memset(&eth, 0, sizeof(eth));
2382 	eth.h_proto = htons(proto);
2383 
2384 	ret = br_multicast_querier_exists(br, &eth);
2385 
2386 unlock:
2387 	rcu_read_unlock();
2388 	return ret;
2389 }
2390 EXPORT_SYMBOL_GPL(br_multicast_has_querier_anywhere);
2391 
2392 /**
2393  * br_multicast_has_querier_adjacent - Checks for a querier behind a bridge port
2394  * @dev: The bridge port adjacent to which to check for a querier
2395  * @proto: The protocol family to check for: IGMP -> ETH_P_IP, MLD -> ETH_P_IPV6
2396  *
2397  * Checks whether the given interface has a bridge on top and if so returns
2398  * true if a selected querier is behind one of the other ports of this
2399  * bridge. Otherwise returns false.
2400  */
2401 bool br_multicast_has_querier_adjacent(struct net_device *dev, int proto)
2402 {
2403 	struct net_bridge *br;
2404 	struct net_bridge_port *port;
2405 	bool ret = false;
2406 
2407 	rcu_read_lock();
2408 	if (!br_port_exists(dev))
2409 		goto unlock;
2410 
2411 	port = br_port_get_rcu(dev);
2412 	if (!port || !port->br)
2413 		goto unlock;
2414 
2415 	br = port->br;
2416 
2417 	switch (proto) {
2418 	case ETH_P_IP:
2419 		if (!timer_pending(&br->ip4_other_query.timer) ||
2420 		    rcu_dereference(br->ip4_querier.port) == port)
2421 			goto unlock;
2422 		break;
2423 #if IS_ENABLED(CONFIG_IPV6)
2424 	case ETH_P_IPV6:
2425 		if (!timer_pending(&br->ip6_other_query.timer) ||
2426 		    rcu_dereference(br->ip6_querier.port) == port)
2427 			goto unlock;
2428 		break;
2429 #endif
2430 	default:
2431 		goto unlock;
2432 	}
2433 
2434 	ret = true;
2435 unlock:
2436 	rcu_read_unlock();
2437 	return ret;
2438 }
2439 EXPORT_SYMBOL_GPL(br_multicast_has_querier_adjacent);
2440 
2441 static void br_mcast_stats_add(struct bridge_mcast_stats __percpu *stats,
2442 			       const struct sk_buff *skb, u8 type, u8 dir)
2443 {
2444 	struct bridge_mcast_stats *pstats = this_cpu_ptr(stats);
2445 	__be16 proto = skb->protocol;
2446 	unsigned int t_len;
2447 
2448 	u64_stats_update_begin(&pstats->syncp);
2449 	switch (proto) {
2450 	case htons(ETH_P_IP):
2451 		t_len = ntohs(ip_hdr(skb)->tot_len) - ip_hdrlen(skb);
2452 		switch (type) {
2453 		case IGMP_HOST_MEMBERSHIP_REPORT:
2454 			pstats->mstats.igmp_v1reports[dir]++;
2455 			break;
2456 		case IGMPV2_HOST_MEMBERSHIP_REPORT:
2457 			pstats->mstats.igmp_v2reports[dir]++;
2458 			break;
2459 		case IGMPV3_HOST_MEMBERSHIP_REPORT:
2460 			pstats->mstats.igmp_v3reports[dir]++;
2461 			break;
2462 		case IGMP_HOST_MEMBERSHIP_QUERY:
2463 			if (t_len != sizeof(struct igmphdr)) {
2464 				pstats->mstats.igmp_v3queries[dir]++;
2465 			} else {
2466 				unsigned int offset = skb_transport_offset(skb);
2467 				struct igmphdr *ih, _ihdr;
2468 
2469 				ih = skb_header_pointer(skb, offset,
2470 							sizeof(_ihdr), &_ihdr);
2471 				if (!ih)
2472 					break;
2473 				if (!ih->code)
2474 					pstats->mstats.igmp_v1queries[dir]++;
2475 				else
2476 					pstats->mstats.igmp_v2queries[dir]++;
2477 			}
2478 			break;
2479 		case IGMP_HOST_LEAVE_MESSAGE:
2480 			pstats->mstats.igmp_leaves[dir]++;
2481 			break;
2482 		}
2483 		break;
2484 #if IS_ENABLED(CONFIG_IPV6)
2485 	case htons(ETH_P_IPV6):
2486 		t_len = ntohs(ipv6_hdr(skb)->payload_len) +
2487 			sizeof(struct ipv6hdr);
2488 		t_len -= skb_network_header_len(skb);
2489 		switch (type) {
2490 		case ICMPV6_MGM_REPORT:
2491 			pstats->mstats.mld_v1reports[dir]++;
2492 			break;
2493 		case ICMPV6_MLD2_REPORT:
2494 			pstats->mstats.mld_v2reports[dir]++;
2495 			break;
2496 		case ICMPV6_MGM_QUERY:
2497 			if (t_len != sizeof(struct mld_msg))
2498 				pstats->mstats.mld_v2queries[dir]++;
2499 			else
2500 				pstats->mstats.mld_v1queries[dir]++;
2501 			break;
2502 		case ICMPV6_MGM_REDUCTION:
2503 			pstats->mstats.mld_leaves[dir]++;
2504 			break;
2505 		}
2506 		break;
2507 #endif /* CONFIG_IPV6 */
2508 	}
2509 	u64_stats_update_end(&pstats->syncp);
2510 }
2511 
2512 void br_multicast_count(struct net_bridge *br, const struct net_bridge_port *p,
2513 			const struct sk_buff *skb, u8 type, u8 dir)
2514 {
2515 	struct bridge_mcast_stats __percpu *stats;
2516 
2517 	/* if multicast_disabled is true then igmp type can't be set */
2518 	if (!type || !br->multicast_stats_enabled)
2519 		return;
2520 
2521 	if (p)
2522 		stats = p->mcast_stats;
2523 	else
2524 		stats = br->mcast_stats;
2525 	if (WARN_ON(!stats))
2526 		return;
2527 
2528 	br_mcast_stats_add(stats, skb, type, dir);
2529 }
2530 
2531 int br_multicast_init_stats(struct net_bridge *br)
2532 {
2533 	br->mcast_stats = netdev_alloc_pcpu_stats(struct bridge_mcast_stats);
2534 	if (!br->mcast_stats)
2535 		return -ENOMEM;
2536 
2537 	return 0;
2538 }
2539 
2540 void br_multicast_uninit_stats(struct net_bridge *br)
2541 {
2542 	free_percpu(br->mcast_stats);
2543 }
2544 
2545 static void mcast_stats_add_dir(u64 *dst, u64 *src)
2546 {
2547 	dst[BR_MCAST_DIR_RX] += src[BR_MCAST_DIR_RX];
2548 	dst[BR_MCAST_DIR_TX] += src[BR_MCAST_DIR_TX];
2549 }
2550 
2551 void br_multicast_get_stats(const struct net_bridge *br,
2552 			    const struct net_bridge_port *p,
2553 			    struct br_mcast_stats *dest)
2554 {
2555 	struct bridge_mcast_stats __percpu *stats;
2556 	struct br_mcast_stats tdst;
2557 	int i;
2558 
2559 	memset(dest, 0, sizeof(*dest));
2560 	if (p)
2561 		stats = p->mcast_stats;
2562 	else
2563 		stats = br->mcast_stats;
2564 	if (WARN_ON(!stats))
2565 		return;
2566 
2567 	memset(&tdst, 0, sizeof(tdst));
2568 	for_each_possible_cpu(i) {
2569 		struct bridge_mcast_stats *cpu_stats = per_cpu_ptr(stats, i);
2570 		struct br_mcast_stats temp;
2571 		unsigned int start;
2572 
2573 		do {
2574 			start = u64_stats_fetch_begin_irq(&cpu_stats->syncp);
2575 			memcpy(&temp, &cpu_stats->mstats, sizeof(temp));
2576 		} while (u64_stats_fetch_retry_irq(&cpu_stats->syncp, start));
2577 
2578 		mcast_stats_add_dir(tdst.igmp_v1queries, temp.igmp_v1queries);
2579 		mcast_stats_add_dir(tdst.igmp_v2queries, temp.igmp_v2queries);
2580 		mcast_stats_add_dir(tdst.igmp_v3queries, temp.igmp_v3queries);
2581 		mcast_stats_add_dir(tdst.igmp_leaves, temp.igmp_leaves);
2582 		mcast_stats_add_dir(tdst.igmp_v1reports, temp.igmp_v1reports);
2583 		mcast_stats_add_dir(tdst.igmp_v2reports, temp.igmp_v2reports);
2584 		mcast_stats_add_dir(tdst.igmp_v3reports, temp.igmp_v3reports);
2585 		tdst.igmp_parse_errors += temp.igmp_parse_errors;
2586 
2587 		mcast_stats_add_dir(tdst.mld_v1queries, temp.mld_v1queries);
2588 		mcast_stats_add_dir(tdst.mld_v2queries, temp.mld_v2queries);
2589 		mcast_stats_add_dir(tdst.mld_leaves, temp.mld_leaves);
2590 		mcast_stats_add_dir(tdst.mld_v1reports, temp.mld_v1reports);
2591 		mcast_stats_add_dir(tdst.mld_v2reports, temp.mld_v2reports);
2592 		tdst.mld_parse_errors += temp.mld_parse_errors;
2593 	}
2594 	memcpy(dest, &tdst, sizeof(*dest));
2595 }
2596