Drivers: hv: vmbus: Fix a bug in rescind processing in vmbus_close_internal()
[linux/fpc-iii.git] / net / bridge / br_multicast.c
blobc465876c7861814ba545cf83783c7ba11bbd91eb
1 /*
2 * Bridge multicast support.
4 * Copyright (c) 2010 Herbert Xu <herbert@gondor.apana.org.au>
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.
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
36 #include "br_private.h"
38 static void br_multicast_start_querier(struct net_bridge *br,
39 struct bridge_mcast_own_query *query);
40 unsigned int br_mdb_rehash_seq;
42 static inline int br_ip_equal(const struct br_ip *a, const struct br_ip *b)
44 if (a->proto != b->proto)
45 return 0;
46 if (a->vid != b->vid)
47 return 0;
48 switch (a->proto) {
49 case htons(ETH_P_IP):
50 return a->u.ip4 == b->u.ip4;
51 #if IS_ENABLED(CONFIG_IPV6)
52 case htons(ETH_P_IPV6):
53 return ipv6_addr_equal(&a->u.ip6, &b->u.ip6);
54 #endif
56 return 0;
59 static inline int __br_ip4_hash(struct net_bridge_mdb_htable *mdb, __be32 ip,
60 __u16 vid)
62 return jhash_2words((__force u32)ip, vid, mdb->secret) & (mdb->max - 1);
65 #if IS_ENABLED(CONFIG_IPV6)
66 static inline int __br_ip6_hash(struct net_bridge_mdb_htable *mdb,
67 const struct in6_addr *ip,
68 __u16 vid)
70 return jhash_2words(ipv6_addr_hash(ip), vid,
71 mdb->secret) & (mdb->max - 1);
73 #endif
75 static inline int br_ip_hash(struct net_bridge_mdb_htable *mdb,
76 struct br_ip *ip)
78 switch (ip->proto) {
79 case htons(ETH_P_IP):
80 return __br_ip4_hash(mdb, ip->u.ip4, ip->vid);
81 #if IS_ENABLED(CONFIG_IPV6)
82 case htons(ETH_P_IPV6):
83 return __br_ip6_hash(mdb, &ip->u.ip6, ip->vid);
84 #endif
86 return 0;
89 static struct net_bridge_mdb_entry *__br_mdb_ip_get(
90 struct net_bridge_mdb_htable *mdb, struct br_ip *dst, int hash)
92 struct net_bridge_mdb_entry *mp;
94 hlist_for_each_entry_rcu(mp, &mdb->mhash[hash], hlist[mdb->ver]) {
95 if (br_ip_equal(&mp->addr, dst))
96 return mp;
99 return NULL;
102 struct net_bridge_mdb_entry *br_mdb_ip_get(struct net_bridge_mdb_htable *mdb,
103 struct br_ip *dst)
105 if (!mdb)
106 return NULL;
108 return __br_mdb_ip_get(mdb, dst, br_ip_hash(mdb, dst));
111 static struct net_bridge_mdb_entry *br_mdb_ip4_get(
112 struct net_bridge_mdb_htable *mdb, __be32 dst, __u16 vid)
114 struct br_ip br_dst;
116 br_dst.u.ip4 = dst;
117 br_dst.proto = htons(ETH_P_IP);
118 br_dst.vid = vid;
120 return br_mdb_ip_get(mdb, &br_dst);
123 #if IS_ENABLED(CONFIG_IPV6)
124 static struct net_bridge_mdb_entry *br_mdb_ip6_get(
125 struct net_bridge_mdb_htable *mdb, const struct in6_addr *dst,
126 __u16 vid)
128 struct br_ip br_dst;
130 br_dst.u.ip6 = *dst;
131 br_dst.proto = htons(ETH_P_IPV6);
132 br_dst.vid = vid;
134 return br_mdb_ip_get(mdb, &br_dst);
136 #endif
138 struct net_bridge_mdb_entry *br_mdb_get(struct net_bridge *br,
139 struct sk_buff *skb, u16 vid)
141 struct net_bridge_mdb_htable *mdb = rcu_dereference(br->mdb);
142 struct br_ip ip;
144 if (br->multicast_disabled)
145 return NULL;
147 if (BR_INPUT_SKB_CB(skb)->igmp)
148 return NULL;
150 ip.proto = skb->protocol;
151 ip.vid = vid;
153 switch (skb->protocol) {
154 case htons(ETH_P_IP):
155 ip.u.ip4 = ip_hdr(skb)->daddr;
156 break;
157 #if IS_ENABLED(CONFIG_IPV6)
158 case htons(ETH_P_IPV6):
159 ip.u.ip6 = ipv6_hdr(skb)->daddr;
160 break;
161 #endif
162 default:
163 return NULL;
166 return br_mdb_ip_get(mdb, &ip);
169 static void br_mdb_free(struct rcu_head *head)
171 struct net_bridge_mdb_htable *mdb =
172 container_of(head, struct net_bridge_mdb_htable, rcu);
173 struct net_bridge_mdb_htable *old = mdb->old;
175 mdb->old = NULL;
176 kfree(old->mhash);
177 kfree(old);
180 static int br_mdb_copy(struct net_bridge_mdb_htable *new,
181 struct net_bridge_mdb_htable *old,
182 int elasticity)
184 struct net_bridge_mdb_entry *mp;
185 int maxlen;
186 int len;
187 int i;
189 for (i = 0; i < old->max; i++)
190 hlist_for_each_entry(mp, &old->mhash[i], hlist[old->ver])
191 hlist_add_head(&mp->hlist[new->ver],
192 &new->mhash[br_ip_hash(new, &mp->addr)]);
194 if (!elasticity)
195 return 0;
197 maxlen = 0;
198 for (i = 0; i < new->max; i++) {
199 len = 0;
200 hlist_for_each_entry(mp, &new->mhash[i], hlist[new->ver])
201 len++;
202 if (len > maxlen)
203 maxlen = len;
206 return maxlen > elasticity ? -EINVAL : 0;
209 void br_multicast_free_pg(struct rcu_head *head)
211 struct net_bridge_port_group *p =
212 container_of(head, struct net_bridge_port_group, rcu);
214 kfree(p);
217 static void br_multicast_free_group(struct rcu_head *head)
219 struct net_bridge_mdb_entry *mp =
220 container_of(head, struct net_bridge_mdb_entry, rcu);
222 kfree(mp);
225 static void br_multicast_group_expired(unsigned long data)
227 struct net_bridge_mdb_entry *mp = (void *)data;
228 struct net_bridge *br = mp->br;
229 struct net_bridge_mdb_htable *mdb;
231 spin_lock(&br->multicast_lock);
232 if (!netif_running(br->dev) || timer_pending(&mp->timer))
233 goto out;
235 mp->mglist = false;
237 if (mp->ports)
238 goto out;
240 mdb = mlock_dereference(br->mdb, br);
242 hlist_del_rcu(&mp->hlist[mdb->ver]);
243 mdb->size--;
245 call_rcu_bh(&mp->rcu, br_multicast_free_group);
247 out:
248 spin_unlock(&br->multicast_lock);
251 static void br_multicast_del_pg(struct net_bridge *br,
252 struct net_bridge_port_group *pg)
254 struct net_bridge_mdb_htable *mdb;
255 struct net_bridge_mdb_entry *mp;
256 struct net_bridge_port_group *p;
257 struct net_bridge_port_group __rcu **pp;
259 mdb = mlock_dereference(br->mdb, br);
261 mp = br_mdb_ip_get(mdb, &pg->addr);
262 if (WARN_ON(!mp))
263 return;
265 for (pp = &mp->ports;
266 (p = mlock_dereference(*pp, br)) != NULL;
267 pp = &p->next) {
268 if (p != pg)
269 continue;
271 rcu_assign_pointer(*pp, p->next);
272 hlist_del_init(&p->mglist);
273 del_timer(&p->timer);
274 call_rcu_bh(&p->rcu, br_multicast_free_pg);
276 if (!mp->ports && !mp->mglist &&
277 netif_running(br->dev))
278 mod_timer(&mp->timer, jiffies);
280 return;
283 WARN_ON(1);
286 static void br_multicast_port_group_expired(unsigned long data)
288 struct net_bridge_port_group *pg = (void *)data;
289 struct net_bridge *br = pg->port->br;
291 spin_lock(&br->multicast_lock);
292 if (!netif_running(br->dev) || timer_pending(&pg->timer) ||
293 hlist_unhashed(&pg->mglist) || pg->state & MDB_PERMANENT)
294 goto out;
296 br_multicast_del_pg(br, pg);
298 out:
299 spin_unlock(&br->multicast_lock);
302 static int br_mdb_rehash(struct net_bridge_mdb_htable __rcu **mdbp, int max,
303 int elasticity)
305 struct net_bridge_mdb_htable *old = rcu_dereference_protected(*mdbp, 1);
306 struct net_bridge_mdb_htable *mdb;
307 int err;
309 mdb = kmalloc(sizeof(*mdb), GFP_ATOMIC);
310 if (!mdb)
311 return -ENOMEM;
313 mdb->max = max;
314 mdb->old = old;
316 mdb->mhash = kzalloc(max * sizeof(*mdb->mhash), GFP_ATOMIC);
317 if (!mdb->mhash) {
318 kfree(mdb);
319 return -ENOMEM;
322 mdb->size = old ? old->size : 0;
323 mdb->ver = old ? old->ver ^ 1 : 0;
325 if (!old || elasticity)
326 get_random_bytes(&mdb->secret, sizeof(mdb->secret));
327 else
328 mdb->secret = old->secret;
330 if (!old)
331 goto out;
333 err = br_mdb_copy(mdb, old, elasticity);
334 if (err) {
335 kfree(mdb->mhash);
336 kfree(mdb);
337 return err;
340 br_mdb_rehash_seq++;
341 call_rcu_bh(&mdb->rcu, br_mdb_free);
343 out:
344 rcu_assign_pointer(*mdbp, mdb);
346 return 0;
349 static struct sk_buff *br_ip4_multicast_alloc_query(struct net_bridge *br,
350 __be32 group)
352 struct sk_buff *skb;
353 struct igmphdr *ih;
354 struct ethhdr *eth;
355 struct iphdr *iph;
357 skb = netdev_alloc_skb_ip_align(br->dev, sizeof(*eth) + sizeof(*iph) +
358 sizeof(*ih) + 4);
359 if (!skb)
360 goto out;
362 skb->protocol = htons(ETH_P_IP);
364 skb_reset_mac_header(skb);
365 eth = eth_hdr(skb);
367 ether_addr_copy(eth->h_source, br->dev->dev_addr);
368 eth->h_dest[0] = 1;
369 eth->h_dest[1] = 0;
370 eth->h_dest[2] = 0x5e;
371 eth->h_dest[3] = 0;
372 eth->h_dest[4] = 0;
373 eth->h_dest[5] = 1;
374 eth->h_proto = htons(ETH_P_IP);
375 skb_put(skb, sizeof(*eth));
377 skb_set_network_header(skb, skb->len);
378 iph = ip_hdr(skb);
380 iph->version = 4;
381 iph->ihl = 6;
382 iph->tos = 0xc0;
383 iph->tot_len = htons(sizeof(*iph) + sizeof(*ih) + 4);
384 iph->id = 0;
385 iph->frag_off = htons(IP_DF);
386 iph->ttl = 1;
387 iph->protocol = IPPROTO_IGMP;
388 iph->saddr = br->multicast_query_use_ifaddr ?
389 inet_select_addr(br->dev, 0, RT_SCOPE_LINK) : 0;
390 iph->daddr = htonl(INADDR_ALLHOSTS_GROUP);
391 ((u8 *)&iph[1])[0] = IPOPT_RA;
392 ((u8 *)&iph[1])[1] = 4;
393 ((u8 *)&iph[1])[2] = 0;
394 ((u8 *)&iph[1])[3] = 0;
395 ip_send_check(iph);
396 skb_put(skb, 24);
398 skb_set_transport_header(skb, skb->len);
399 ih = igmp_hdr(skb);
400 ih->type = IGMP_HOST_MEMBERSHIP_QUERY;
401 ih->code = (group ? br->multicast_last_member_interval :
402 br->multicast_query_response_interval) /
403 (HZ / IGMP_TIMER_SCALE);
404 ih->group = group;
405 ih->csum = 0;
406 ih->csum = ip_compute_csum((void *)ih, sizeof(struct igmphdr));
407 skb_put(skb, sizeof(*ih));
409 __skb_pull(skb, sizeof(*eth));
411 out:
412 return skb;
415 #if IS_ENABLED(CONFIG_IPV6)
416 static struct sk_buff *br_ip6_multicast_alloc_query(struct net_bridge *br,
417 const struct in6_addr *group)
419 struct sk_buff *skb;
420 struct ipv6hdr *ip6h;
421 struct mld_msg *mldq;
422 struct ethhdr *eth;
423 u8 *hopopt;
424 unsigned long interval;
426 skb = netdev_alloc_skb_ip_align(br->dev, sizeof(*eth) + sizeof(*ip6h) +
427 8 + sizeof(*mldq));
428 if (!skb)
429 goto out;
431 skb->protocol = htons(ETH_P_IPV6);
433 /* Ethernet header */
434 skb_reset_mac_header(skb);
435 eth = eth_hdr(skb);
437 ether_addr_copy(eth->h_source, br->dev->dev_addr);
438 eth->h_proto = htons(ETH_P_IPV6);
439 skb_put(skb, sizeof(*eth));
441 /* IPv6 header + HbH option */
442 skb_set_network_header(skb, skb->len);
443 ip6h = ipv6_hdr(skb);
445 *(__force __be32 *)ip6h = htonl(0x60000000);
446 ip6h->payload_len = htons(8 + sizeof(*mldq));
447 ip6h->nexthdr = IPPROTO_HOPOPTS;
448 ip6h->hop_limit = 1;
449 ipv6_addr_set(&ip6h->daddr, htonl(0xff020000), 0, 0, htonl(1));
450 if (ipv6_dev_get_saddr(dev_net(br->dev), br->dev, &ip6h->daddr, 0,
451 &ip6h->saddr)) {
452 kfree_skb(skb);
453 return NULL;
455 ipv6_eth_mc_map(&ip6h->daddr, eth->h_dest);
457 hopopt = (u8 *)(ip6h + 1);
458 hopopt[0] = IPPROTO_ICMPV6; /* next hdr */
459 hopopt[1] = 0; /* length of HbH */
460 hopopt[2] = IPV6_TLV_ROUTERALERT; /* Router Alert */
461 hopopt[3] = 2; /* Length of RA Option */
462 hopopt[4] = 0; /* Type = 0x0000 (MLD) */
463 hopopt[5] = 0;
464 hopopt[6] = IPV6_TLV_PAD1; /* Pad1 */
465 hopopt[7] = IPV6_TLV_PAD1; /* Pad1 */
467 skb_put(skb, sizeof(*ip6h) + 8);
469 /* ICMPv6 */
470 skb_set_transport_header(skb, skb->len);
471 mldq = (struct mld_msg *) icmp6_hdr(skb);
473 interval = ipv6_addr_any(group) ?
474 br->multicast_query_response_interval :
475 br->multicast_last_member_interval;
477 mldq->mld_type = ICMPV6_MGM_QUERY;
478 mldq->mld_code = 0;
479 mldq->mld_cksum = 0;
480 mldq->mld_maxdelay = htons((u16)jiffies_to_msecs(interval));
481 mldq->mld_reserved = 0;
482 mldq->mld_mca = *group;
484 /* checksum */
485 mldq->mld_cksum = csum_ipv6_magic(&ip6h->saddr, &ip6h->daddr,
486 sizeof(*mldq), IPPROTO_ICMPV6,
487 csum_partial(mldq,
488 sizeof(*mldq), 0));
489 skb_put(skb, sizeof(*mldq));
491 __skb_pull(skb, sizeof(*eth));
493 out:
494 return skb;
496 #endif
498 static struct sk_buff *br_multicast_alloc_query(struct net_bridge *br,
499 struct br_ip *addr)
501 switch (addr->proto) {
502 case htons(ETH_P_IP):
503 return br_ip4_multicast_alloc_query(br, addr->u.ip4);
504 #if IS_ENABLED(CONFIG_IPV6)
505 case htons(ETH_P_IPV6):
506 return br_ip6_multicast_alloc_query(br, &addr->u.ip6);
507 #endif
509 return NULL;
512 static struct net_bridge_mdb_entry *br_multicast_get_group(
513 struct net_bridge *br, struct net_bridge_port *port,
514 struct br_ip *group, int hash)
516 struct net_bridge_mdb_htable *mdb;
517 struct net_bridge_mdb_entry *mp;
518 unsigned int count = 0;
519 unsigned int max;
520 int elasticity;
521 int err;
523 mdb = rcu_dereference_protected(br->mdb, 1);
524 hlist_for_each_entry(mp, &mdb->mhash[hash], hlist[mdb->ver]) {
525 count++;
526 if (unlikely(br_ip_equal(group, &mp->addr)))
527 return mp;
530 elasticity = 0;
531 max = mdb->max;
533 if (unlikely(count > br->hash_elasticity && count)) {
534 if (net_ratelimit())
535 br_info(br, "Multicast hash table "
536 "chain limit reached: %s\n",
537 port ? port->dev->name : br->dev->name);
539 elasticity = br->hash_elasticity;
542 if (mdb->size >= max) {
543 max *= 2;
544 if (unlikely(max > br->hash_max)) {
545 br_warn(br, "Multicast hash table maximum of %d "
546 "reached, disabling snooping: %s\n",
547 br->hash_max,
548 port ? port->dev->name : br->dev->name);
549 err = -E2BIG;
550 disable:
551 br->multicast_disabled = 1;
552 goto err;
556 if (max > mdb->max || elasticity) {
557 if (mdb->old) {
558 if (net_ratelimit())
559 br_info(br, "Multicast hash table "
560 "on fire: %s\n",
561 port ? port->dev->name : br->dev->name);
562 err = -EEXIST;
563 goto err;
566 err = br_mdb_rehash(&br->mdb, max, elasticity);
567 if (err) {
568 br_warn(br, "Cannot rehash multicast "
569 "hash table, disabling snooping: %s, %d, %d\n",
570 port ? port->dev->name : br->dev->name,
571 mdb->size, err);
572 goto disable;
575 err = -EAGAIN;
576 goto err;
579 return NULL;
581 err:
582 mp = ERR_PTR(err);
583 return mp;
586 struct net_bridge_mdb_entry *br_multicast_new_group(struct net_bridge *br,
587 struct net_bridge_port *port, struct br_ip *group)
589 struct net_bridge_mdb_htable *mdb;
590 struct net_bridge_mdb_entry *mp;
591 int hash;
592 int err;
594 mdb = rcu_dereference_protected(br->mdb, 1);
595 if (!mdb) {
596 err = br_mdb_rehash(&br->mdb, BR_HASH_SIZE, 0);
597 if (err)
598 return ERR_PTR(err);
599 goto rehash;
602 hash = br_ip_hash(mdb, group);
603 mp = br_multicast_get_group(br, port, group, hash);
604 switch (PTR_ERR(mp)) {
605 case 0:
606 break;
608 case -EAGAIN:
609 rehash:
610 mdb = rcu_dereference_protected(br->mdb, 1);
611 hash = br_ip_hash(mdb, group);
612 break;
614 default:
615 goto out;
618 mp = kzalloc(sizeof(*mp), GFP_ATOMIC);
619 if (unlikely(!mp))
620 return ERR_PTR(-ENOMEM);
622 mp->br = br;
623 mp->addr = *group;
624 setup_timer(&mp->timer, br_multicast_group_expired,
625 (unsigned long)mp);
627 hlist_add_head_rcu(&mp->hlist[mdb->ver], &mdb->mhash[hash]);
628 mdb->size++;
630 out:
631 return mp;
634 struct net_bridge_port_group *br_multicast_new_port_group(
635 struct net_bridge_port *port,
636 struct br_ip *group,
637 struct net_bridge_port_group __rcu *next,
638 unsigned char state)
640 struct net_bridge_port_group *p;
642 p = kzalloc(sizeof(*p), GFP_ATOMIC);
643 if (unlikely(!p))
644 return NULL;
646 p->addr = *group;
647 p->port = port;
648 p->state = state;
649 rcu_assign_pointer(p->next, next);
650 hlist_add_head(&p->mglist, &port->mglist);
651 setup_timer(&p->timer, br_multicast_port_group_expired,
652 (unsigned long)p);
653 return p;
656 static int br_multicast_add_group(struct net_bridge *br,
657 struct net_bridge_port *port,
658 struct br_ip *group)
660 struct net_bridge_mdb_entry *mp;
661 struct net_bridge_port_group *p;
662 struct net_bridge_port_group __rcu **pp;
663 unsigned long now = jiffies;
664 int err;
666 spin_lock(&br->multicast_lock);
667 if (!netif_running(br->dev) ||
668 (port && port->state == BR_STATE_DISABLED))
669 goto out;
671 mp = br_multicast_new_group(br, port, group);
672 err = PTR_ERR(mp);
673 if (IS_ERR(mp))
674 goto err;
676 if (!port) {
677 mp->mglist = true;
678 mod_timer(&mp->timer, now + br->multicast_membership_interval);
679 goto out;
682 for (pp = &mp->ports;
683 (p = mlock_dereference(*pp, br)) != NULL;
684 pp = &p->next) {
685 if (p->port == port)
686 goto found;
687 if ((unsigned long)p->port < (unsigned long)port)
688 break;
691 p = br_multicast_new_port_group(port, group, *pp, MDB_TEMPORARY);
692 if (unlikely(!p))
693 goto err;
694 rcu_assign_pointer(*pp, p);
695 br_mdb_notify(br->dev, port, group, RTM_NEWMDB);
697 found:
698 mod_timer(&p->timer, now + br->multicast_membership_interval);
699 out:
700 err = 0;
702 err:
703 spin_unlock(&br->multicast_lock);
704 return err;
707 static int br_ip4_multicast_add_group(struct net_bridge *br,
708 struct net_bridge_port *port,
709 __be32 group,
710 __u16 vid)
712 struct br_ip br_group;
714 if (ipv4_is_local_multicast(group))
715 return 0;
717 br_group.u.ip4 = group;
718 br_group.proto = htons(ETH_P_IP);
719 br_group.vid = vid;
721 return br_multicast_add_group(br, port, &br_group);
724 #if IS_ENABLED(CONFIG_IPV6)
725 static int br_ip6_multicast_add_group(struct net_bridge *br,
726 struct net_bridge_port *port,
727 const struct in6_addr *group,
728 __u16 vid)
730 struct br_ip br_group;
732 if (ipv6_addr_is_ll_all_nodes(group))
733 return 0;
735 br_group.u.ip6 = *group;
736 br_group.proto = htons(ETH_P_IPV6);
737 br_group.vid = vid;
739 return br_multicast_add_group(br, port, &br_group);
741 #endif
743 static void br_multicast_router_expired(unsigned long data)
745 struct net_bridge_port *port = (void *)data;
746 struct net_bridge *br = port->br;
748 spin_lock(&br->multicast_lock);
749 if (port->multicast_router != 1 ||
750 timer_pending(&port->multicast_router_timer) ||
751 hlist_unhashed(&port->rlist))
752 goto out;
754 hlist_del_init_rcu(&port->rlist);
756 out:
757 spin_unlock(&br->multicast_lock);
760 static void br_multicast_local_router_expired(unsigned long data)
764 static void br_multicast_querier_expired(struct net_bridge *br,
765 struct bridge_mcast_own_query *query)
767 spin_lock(&br->multicast_lock);
768 if (!netif_running(br->dev) || br->multicast_disabled)
769 goto out;
771 br_multicast_start_querier(br, query);
773 out:
774 spin_unlock(&br->multicast_lock);
777 static void br_ip4_multicast_querier_expired(unsigned long data)
779 struct net_bridge *br = (void *)data;
781 br_multicast_querier_expired(br, &br->ip4_own_query);
784 #if IS_ENABLED(CONFIG_IPV6)
785 static void br_ip6_multicast_querier_expired(unsigned long data)
787 struct net_bridge *br = (void *)data;
789 br_multicast_querier_expired(br, &br->ip6_own_query);
791 #endif
793 static void br_multicast_select_own_querier(struct net_bridge *br,
794 struct br_ip *ip,
795 struct sk_buff *skb)
797 if (ip->proto == htons(ETH_P_IP))
798 br->ip4_querier.addr.u.ip4 = ip_hdr(skb)->saddr;
799 #if IS_ENABLED(CONFIG_IPV6)
800 else
801 br->ip6_querier.addr.u.ip6 = ipv6_hdr(skb)->saddr;
802 #endif
805 static void __br_multicast_send_query(struct net_bridge *br,
806 struct net_bridge_port *port,
807 struct br_ip *ip)
809 struct sk_buff *skb;
811 skb = br_multicast_alloc_query(br, ip);
812 if (!skb)
813 return;
815 if (port) {
816 skb->dev = port->dev;
817 NF_HOOK(NFPROTO_BRIDGE, NF_BR_LOCAL_OUT, skb, NULL, skb->dev,
818 br_dev_queue_push_xmit);
819 } else {
820 br_multicast_select_own_querier(br, ip, skb);
821 netif_rx(skb);
825 static void br_multicast_send_query(struct net_bridge *br,
826 struct net_bridge_port *port,
827 struct bridge_mcast_own_query *own_query)
829 unsigned long time;
830 struct br_ip br_group;
831 struct bridge_mcast_other_query *other_query = NULL;
833 if (!netif_running(br->dev) || br->multicast_disabled ||
834 !br->multicast_querier)
835 return;
837 memset(&br_group.u, 0, sizeof(br_group.u));
839 if (port ? (own_query == &port->ip4_own_query) :
840 (own_query == &br->ip4_own_query)) {
841 other_query = &br->ip4_other_query;
842 br_group.proto = htons(ETH_P_IP);
843 #if IS_ENABLED(CONFIG_IPV6)
844 } else {
845 other_query = &br->ip6_other_query;
846 br_group.proto = htons(ETH_P_IPV6);
847 #endif
850 if (!other_query || timer_pending(&other_query->timer))
851 return;
853 __br_multicast_send_query(br, port, &br_group);
855 time = jiffies;
856 time += own_query->startup_sent < br->multicast_startup_query_count ?
857 br->multicast_startup_query_interval :
858 br->multicast_query_interval;
859 mod_timer(&own_query->timer, time);
862 static void
863 br_multicast_port_query_expired(struct net_bridge_port *port,
864 struct bridge_mcast_own_query *query)
866 struct net_bridge *br = port->br;
868 spin_lock(&br->multicast_lock);
869 if (port->state == BR_STATE_DISABLED ||
870 port->state == BR_STATE_BLOCKING)
871 goto out;
873 if (query->startup_sent < br->multicast_startup_query_count)
874 query->startup_sent++;
876 br_multicast_send_query(port->br, port, query);
878 out:
879 spin_unlock(&br->multicast_lock);
882 static void br_ip4_multicast_port_query_expired(unsigned long data)
884 struct net_bridge_port *port = (void *)data;
886 br_multicast_port_query_expired(port, &port->ip4_own_query);
889 #if IS_ENABLED(CONFIG_IPV6)
890 static void br_ip6_multicast_port_query_expired(unsigned long data)
892 struct net_bridge_port *port = (void *)data;
894 br_multicast_port_query_expired(port, &port->ip6_own_query);
896 #endif
898 void br_multicast_add_port(struct net_bridge_port *port)
900 port->multicast_router = 1;
902 setup_timer(&port->multicast_router_timer, br_multicast_router_expired,
903 (unsigned long)port);
904 setup_timer(&port->ip4_own_query.timer,
905 br_ip4_multicast_port_query_expired, (unsigned long)port);
906 #if IS_ENABLED(CONFIG_IPV6)
907 setup_timer(&port->ip6_own_query.timer,
908 br_ip6_multicast_port_query_expired, (unsigned long)port);
909 #endif
912 void br_multicast_del_port(struct net_bridge_port *port)
914 del_timer_sync(&port->multicast_router_timer);
917 static void br_multicast_enable(struct bridge_mcast_own_query *query)
919 query->startup_sent = 0;
921 if (try_to_del_timer_sync(&query->timer) >= 0 ||
922 del_timer(&query->timer))
923 mod_timer(&query->timer, jiffies);
926 void br_multicast_enable_port(struct net_bridge_port *port)
928 struct net_bridge *br = port->br;
930 spin_lock(&br->multicast_lock);
931 if (br->multicast_disabled || !netif_running(br->dev))
932 goto out;
934 br_multicast_enable(&port->ip4_own_query);
935 #if IS_ENABLED(CONFIG_IPV6)
936 br_multicast_enable(&port->ip6_own_query);
937 #endif
939 out:
940 spin_unlock(&br->multicast_lock);
943 void br_multicast_disable_port(struct net_bridge_port *port)
945 struct net_bridge *br = port->br;
946 struct net_bridge_port_group *pg;
947 struct hlist_node *n;
949 spin_lock(&br->multicast_lock);
950 hlist_for_each_entry_safe(pg, n, &port->mglist, mglist)
951 br_multicast_del_pg(br, pg);
953 if (!hlist_unhashed(&port->rlist))
954 hlist_del_init_rcu(&port->rlist);
955 del_timer(&port->multicast_router_timer);
956 del_timer(&port->ip4_own_query.timer);
957 #if IS_ENABLED(CONFIG_IPV6)
958 del_timer(&port->ip6_own_query.timer);
959 #endif
960 spin_unlock(&br->multicast_lock);
963 static int br_ip4_multicast_igmp3_report(struct net_bridge *br,
964 struct net_bridge_port *port,
965 struct sk_buff *skb,
966 u16 vid)
968 struct igmpv3_report *ih;
969 struct igmpv3_grec *grec;
970 int i;
971 int len;
972 int num;
973 int type;
974 int err = 0;
975 __be32 group;
977 if (!pskb_may_pull(skb, sizeof(*ih)))
978 return -EINVAL;
980 ih = igmpv3_report_hdr(skb);
981 num = ntohs(ih->ngrec);
982 len = sizeof(*ih);
984 for (i = 0; i < num; i++) {
985 len += sizeof(*grec);
986 if (!pskb_may_pull(skb, len))
987 return -EINVAL;
989 grec = (void *)(skb->data + len - sizeof(*grec));
990 group = grec->grec_mca;
991 type = grec->grec_type;
993 len += ntohs(grec->grec_nsrcs) * 4;
994 if (!pskb_may_pull(skb, len))
995 return -EINVAL;
997 /* We treat this as an IGMPv2 report for now. */
998 switch (type) {
999 case IGMPV3_MODE_IS_INCLUDE:
1000 case IGMPV3_MODE_IS_EXCLUDE:
1001 case IGMPV3_CHANGE_TO_INCLUDE:
1002 case IGMPV3_CHANGE_TO_EXCLUDE:
1003 case IGMPV3_ALLOW_NEW_SOURCES:
1004 case IGMPV3_BLOCK_OLD_SOURCES:
1005 break;
1007 default:
1008 continue;
1011 err = br_ip4_multicast_add_group(br, port, group, vid);
1012 if (err)
1013 break;
1016 return err;
1019 #if IS_ENABLED(CONFIG_IPV6)
1020 static int br_ip6_multicast_mld2_report(struct net_bridge *br,
1021 struct net_bridge_port *port,
1022 struct sk_buff *skb,
1023 u16 vid)
1025 struct icmp6hdr *icmp6h;
1026 struct mld2_grec *grec;
1027 int i;
1028 int len;
1029 int num;
1030 int err = 0;
1032 if (!pskb_may_pull(skb, sizeof(*icmp6h)))
1033 return -EINVAL;
1035 icmp6h = icmp6_hdr(skb);
1036 num = ntohs(icmp6h->icmp6_dataun.un_data16[1]);
1037 len = sizeof(*icmp6h);
1039 for (i = 0; i < num; i++) {
1040 __be16 *nsrcs, _nsrcs;
1042 nsrcs = skb_header_pointer(skb,
1043 len + offsetof(struct mld2_grec,
1044 grec_nsrcs),
1045 sizeof(_nsrcs), &_nsrcs);
1046 if (!nsrcs)
1047 return -EINVAL;
1049 if (!pskb_may_pull(skb,
1050 len + sizeof(*grec) +
1051 sizeof(struct in6_addr) * ntohs(*nsrcs)))
1052 return -EINVAL;
1054 grec = (struct mld2_grec *)(skb->data + len);
1055 len += sizeof(*grec) +
1056 sizeof(struct in6_addr) * ntohs(*nsrcs);
1058 /* We treat these as MLDv1 reports for now. */
1059 switch (grec->grec_type) {
1060 case MLD2_MODE_IS_INCLUDE:
1061 case MLD2_MODE_IS_EXCLUDE:
1062 case MLD2_CHANGE_TO_INCLUDE:
1063 case MLD2_CHANGE_TO_EXCLUDE:
1064 case MLD2_ALLOW_NEW_SOURCES:
1065 case MLD2_BLOCK_OLD_SOURCES:
1066 break;
1068 default:
1069 continue;
1072 err = br_ip6_multicast_add_group(br, port, &grec->grec_mca,
1073 vid);
1074 if (!err)
1075 break;
1078 return err;
1080 #endif
1082 static bool br_ip4_multicast_select_querier(struct net_bridge *br,
1083 struct net_bridge_port *port,
1084 __be32 saddr)
1086 if (!timer_pending(&br->ip4_own_query.timer) &&
1087 !timer_pending(&br->ip4_other_query.timer))
1088 goto update;
1090 if (!br->ip4_querier.addr.u.ip4)
1091 goto update;
1093 if (ntohl(saddr) <= ntohl(br->ip4_querier.addr.u.ip4))
1094 goto update;
1096 return false;
1098 update:
1099 br->ip4_querier.addr.u.ip4 = saddr;
1101 /* update protected by general multicast_lock by caller */
1102 rcu_assign_pointer(br->ip4_querier.port, port);
1104 return true;
1107 #if IS_ENABLED(CONFIG_IPV6)
1108 static bool br_ip6_multicast_select_querier(struct net_bridge *br,
1109 struct net_bridge_port *port,
1110 struct in6_addr *saddr)
1112 if (!timer_pending(&br->ip6_own_query.timer) &&
1113 !timer_pending(&br->ip6_other_query.timer))
1114 goto update;
1116 if (ipv6_addr_cmp(saddr, &br->ip6_querier.addr.u.ip6) <= 0)
1117 goto update;
1119 return false;
1121 update:
1122 br->ip6_querier.addr.u.ip6 = *saddr;
1124 /* update protected by general multicast_lock by caller */
1125 rcu_assign_pointer(br->ip6_querier.port, port);
1127 return true;
1129 #endif
1131 static bool br_multicast_select_querier(struct net_bridge *br,
1132 struct net_bridge_port *port,
1133 struct br_ip *saddr)
1135 switch (saddr->proto) {
1136 case htons(ETH_P_IP):
1137 return br_ip4_multicast_select_querier(br, port, saddr->u.ip4);
1138 #if IS_ENABLED(CONFIG_IPV6)
1139 case htons(ETH_P_IPV6):
1140 return br_ip6_multicast_select_querier(br, port, &saddr->u.ip6);
1141 #endif
1144 return false;
1147 static void
1148 br_multicast_update_query_timer(struct net_bridge *br,
1149 struct bridge_mcast_other_query *query,
1150 unsigned long max_delay)
1152 if (!timer_pending(&query->timer))
1153 query->delay_time = jiffies + max_delay;
1155 mod_timer(&query->timer, jiffies + br->multicast_querier_interval);
1159 * Add port to router_list
1160 * list is maintained ordered by pointer value
1161 * and locked by br->multicast_lock and RCU
1163 static void br_multicast_add_router(struct net_bridge *br,
1164 struct net_bridge_port *port)
1166 struct net_bridge_port *p;
1167 struct hlist_node *slot = NULL;
1169 hlist_for_each_entry(p, &br->router_list, rlist) {
1170 if ((unsigned long) port >= (unsigned long) p)
1171 break;
1172 slot = &p->rlist;
1175 if (slot)
1176 hlist_add_behind_rcu(&port->rlist, slot);
1177 else
1178 hlist_add_head_rcu(&port->rlist, &br->router_list);
1181 static void br_multicast_mark_router(struct net_bridge *br,
1182 struct net_bridge_port *port)
1184 unsigned long now = jiffies;
1186 if (!port) {
1187 if (br->multicast_router == 1)
1188 mod_timer(&br->multicast_router_timer,
1189 now + br->multicast_querier_interval);
1190 return;
1193 if (port->multicast_router != 1)
1194 return;
1196 if (!hlist_unhashed(&port->rlist))
1197 goto timer;
1199 br_multicast_add_router(br, port);
1201 timer:
1202 mod_timer(&port->multicast_router_timer,
1203 now + br->multicast_querier_interval);
1206 static void br_multicast_query_received(struct net_bridge *br,
1207 struct net_bridge_port *port,
1208 struct bridge_mcast_other_query *query,
1209 struct br_ip *saddr,
1210 unsigned long max_delay)
1212 if (!br_multicast_select_querier(br, port, saddr))
1213 return;
1215 br_multicast_update_query_timer(br, query, max_delay);
1216 br_multicast_mark_router(br, port);
1219 static int br_ip4_multicast_query(struct net_bridge *br,
1220 struct net_bridge_port *port,
1221 struct sk_buff *skb,
1222 u16 vid)
1224 const struct iphdr *iph = ip_hdr(skb);
1225 struct igmphdr *ih = igmp_hdr(skb);
1226 struct net_bridge_mdb_entry *mp;
1227 struct igmpv3_query *ih3;
1228 struct net_bridge_port_group *p;
1229 struct net_bridge_port_group __rcu **pp;
1230 struct br_ip saddr;
1231 unsigned long max_delay;
1232 unsigned long now = jiffies;
1233 __be32 group;
1234 int err = 0;
1236 spin_lock(&br->multicast_lock);
1237 if (!netif_running(br->dev) ||
1238 (port && port->state == BR_STATE_DISABLED))
1239 goto out;
1241 group = ih->group;
1243 if (skb->len == sizeof(*ih)) {
1244 max_delay = ih->code * (HZ / IGMP_TIMER_SCALE);
1246 if (!max_delay) {
1247 max_delay = 10 * HZ;
1248 group = 0;
1250 } else {
1251 if (!pskb_may_pull(skb, sizeof(struct igmpv3_query))) {
1252 err = -EINVAL;
1253 goto out;
1256 ih3 = igmpv3_query_hdr(skb);
1257 if (ih3->nsrcs)
1258 goto out;
1260 max_delay = ih3->code ?
1261 IGMPV3_MRC(ih3->code) * (HZ / IGMP_TIMER_SCALE) : 1;
1264 /* RFC2236+RFC3376 (IGMPv2+IGMPv3) require the multicast link layer
1265 * all-systems destination addresses (224.0.0.1) for general queries
1267 if (!group && iph->daddr != htonl(INADDR_ALLHOSTS_GROUP)) {
1268 err = -EINVAL;
1269 goto out;
1272 if (!group) {
1273 saddr.proto = htons(ETH_P_IP);
1274 saddr.u.ip4 = iph->saddr;
1276 br_multicast_query_received(br, port, &br->ip4_other_query,
1277 &saddr, max_delay);
1278 goto out;
1281 mp = br_mdb_ip4_get(mlock_dereference(br->mdb, br), group, vid);
1282 if (!mp)
1283 goto out;
1285 max_delay *= br->multicast_last_member_count;
1287 if (mp->mglist &&
1288 (timer_pending(&mp->timer) ?
1289 time_after(mp->timer.expires, now + max_delay) :
1290 try_to_del_timer_sync(&mp->timer) >= 0))
1291 mod_timer(&mp->timer, now + max_delay);
1293 for (pp = &mp->ports;
1294 (p = mlock_dereference(*pp, br)) != NULL;
1295 pp = &p->next) {
1296 if (timer_pending(&p->timer) ?
1297 time_after(p->timer.expires, now + max_delay) :
1298 try_to_del_timer_sync(&p->timer) >= 0)
1299 mod_timer(&p->timer, now + max_delay);
1302 out:
1303 spin_unlock(&br->multicast_lock);
1304 return err;
1307 #if IS_ENABLED(CONFIG_IPV6)
1308 static int br_ip6_multicast_query(struct net_bridge *br,
1309 struct net_bridge_port *port,
1310 struct sk_buff *skb,
1311 u16 vid)
1313 const struct ipv6hdr *ip6h = ipv6_hdr(skb);
1314 struct mld_msg *mld;
1315 struct net_bridge_mdb_entry *mp;
1316 struct mld2_query *mld2q;
1317 struct net_bridge_port_group *p;
1318 struct net_bridge_port_group __rcu **pp;
1319 struct br_ip saddr;
1320 unsigned long max_delay;
1321 unsigned long now = jiffies;
1322 const struct in6_addr *group = NULL;
1323 bool is_general_query;
1324 int err = 0;
1326 spin_lock(&br->multicast_lock);
1327 if (!netif_running(br->dev) ||
1328 (port && port->state == BR_STATE_DISABLED))
1329 goto out;
1331 /* RFC2710+RFC3810 (MLDv1+MLDv2) require link-local source addresses */
1332 if (!(ipv6_addr_type(&ip6h->saddr) & IPV6_ADDR_LINKLOCAL)) {
1333 err = -EINVAL;
1334 goto out;
1337 if (skb->len == sizeof(*mld)) {
1338 if (!pskb_may_pull(skb, sizeof(*mld))) {
1339 err = -EINVAL;
1340 goto out;
1342 mld = (struct mld_msg *) icmp6_hdr(skb);
1343 max_delay = msecs_to_jiffies(ntohs(mld->mld_maxdelay));
1344 if (max_delay)
1345 group = &mld->mld_mca;
1346 } else {
1347 if (!pskb_may_pull(skb, sizeof(*mld2q))) {
1348 err = -EINVAL;
1349 goto out;
1351 mld2q = (struct mld2_query *)icmp6_hdr(skb);
1352 if (!mld2q->mld2q_nsrcs)
1353 group = &mld2q->mld2q_mca;
1355 max_delay = max(msecs_to_jiffies(mldv2_mrc(mld2q)), 1UL);
1358 is_general_query = group && ipv6_addr_any(group);
1360 /* RFC2710+RFC3810 (MLDv1+MLDv2) require the multicast link layer
1361 * all-nodes destination address (ff02::1) for general queries
1363 if (is_general_query && !ipv6_addr_is_ll_all_nodes(&ip6h->daddr)) {
1364 err = -EINVAL;
1365 goto out;
1368 if (is_general_query) {
1369 saddr.proto = htons(ETH_P_IPV6);
1370 saddr.u.ip6 = ip6h->saddr;
1372 br_multicast_query_received(br, port, &br->ip6_other_query,
1373 &saddr, max_delay);
1374 goto out;
1375 } else if (!group) {
1376 goto out;
1379 mp = br_mdb_ip6_get(mlock_dereference(br->mdb, br), group, vid);
1380 if (!mp)
1381 goto out;
1383 max_delay *= br->multicast_last_member_count;
1384 if (mp->mglist &&
1385 (timer_pending(&mp->timer) ?
1386 time_after(mp->timer.expires, now + max_delay) :
1387 try_to_del_timer_sync(&mp->timer) >= 0))
1388 mod_timer(&mp->timer, now + max_delay);
1390 for (pp = &mp->ports;
1391 (p = mlock_dereference(*pp, br)) != NULL;
1392 pp = &p->next) {
1393 if (timer_pending(&p->timer) ?
1394 time_after(p->timer.expires, now + max_delay) :
1395 try_to_del_timer_sync(&p->timer) >= 0)
1396 mod_timer(&p->timer, now + max_delay);
1399 out:
1400 spin_unlock(&br->multicast_lock);
1401 return err;
1403 #endif
1405 static void
1406 br_multicast_leave_group(struct net_bridge *br,
1407 struct net_bridge_port *port,
1408 struct br_ip *group,
1409 struct bridge_mcast_other_query *other_query,
1410 struct bridge_mcast_own_query *own_query)
1412 struct net_bridge_mdb_htable *mdb;
1413 struct net_bridge_mdb_entry *mp;
1414 struct net_bridge_port_group *p;
1415 unsigned long now;
1416 unsigned long time;
1418 spin_lock(&br->multicast_lock);
1419 if (!netif_running(br->dev) ||
1420 (port && port->state == BR_STATE_DISABLED) ||
1421 timer_pending(&other_query->timer))
1422 goto out;
1424 mdb = mlock_dereference(br->mdb, br);
1425 mp = br_mdb_ip_get(mdb, group);
1426 if (!mp)
1427 goto out;
1429 if (br->multicast_querier) {
1430 __br_multicast_send_query(br, port, &mp->addr);
1432 time = jiffies + br->multicast_last_member_count *
1433 br->multicast_last_member_interval;
1435 mod_timer(&own_query->timer, time);
1437 for (p = mlock_dereference(mp->ports, br);
1438 p != NULL;
1439 p = mlock_dereference(p->next, br)) {
1440 if (p->port != port)
1441 continue;
1443 if (!hlist_unhashed(&p->mglist) &&
1444 (timer_pending(&p->timer) ?
1445 time_after(p->timer.expires, time) :
1446 try_to_del_timer_sync(&p->timer) >= 0)) {
1447 mod_timer(&p->timer, time);
1450 break;
1454 if (port && (port->flags & BR_MULTICAST_FAST_LEAVE)) {
1455 struct net_bridge_port_group __rcu **pp;
1457 for (pp = &mp->ports;
1458 (p = mlock_dereference(*pp, br)) != NULL;
1459 pp = &p->next) {
1460 if (p->port != port)
1461 continue;
1463 rcu_assign_pointer(*pp, p->next);
1464 hlist_del_init(&p->mglist);
1465 del_timer(&p->timer);
1466 call_rcu_bh(&p->rcu, br_multicast_free_pg);
1467 br_mdb_notify(br->dev, port, group, RTM_DELMDB);
1469 if (!mp->ports && !mp->mglist &&
1470 netif_running(br->dev))
1471 mod_timer(&mp->timer, jiffies);
1473 goto out;
1476 now = jiffies;
1477 time = now + br->multicast_last_member_count *
1478 br->multicast_last_member_interval;
1480 if (!port) {
1481 if (mp->mglist &&
1482 (timer_pending(&mp->timer) ?
1483 time_after(mp->timer.expires, time) :
1484 try_to_del_timer_sync(&mp->timer) >= 0)) {
1485 mod_timer(&mp->timer, time);
1488 goto out;
1491 for (p = mlock_dereference(mp->ports, br);
1492 p != NULL;
1493 p = mlock_dereference(p->next, br)) {
1494 if (p->port != port)
1495 continue;
1497 if (!hlist_unhashed(&p->mglist) &&
1498 (timer_pending(&p->timer) ?
1499 time_after(p->timer.expires, time) :
1500 try_to_del_timer_sync(&p->timer) >= 0)) {
1501 mod_timer(&p->timer, time);
1504 break;
1506 out:
1507 spin_unlock(&br->multicast_lock);
1510 static void br_ip4_multicast_leave_group(struct net_bridge *br,
1511 struct net_bridge_port *port,
1512 __be32 group,
1513 __u16 vid)
1515 struct br_ip br_group;
1516 struct bridge_mcast_own_query *own_query;
1518 if (ipv4_is_local_multicast(group))
1519 return;
1521 own_query = port ? &port->ip4_own_query : &br->ip4_own_query;
1523 br_group.u.ip4 = group;
1524 br_group.proto = htons(ETH_P_IP);
1525 br_group.vid = vid;
1527 br_multicast_leave_group(br, port, &br_group, &br->ip4_other_query,
1528 own_query);
1531 #if IS_ENABLED(CONFIG_IPV6)
1532 static void br_ip6_multicast_leave_group(struct net_bridge *br,
1533 struct net_bridge_port *port,
1534 const struct in6_addr *group,
1535 __u16 vid)
1537 struct br_ip br_group;
1538 struct bridge_mcast_own_query *own_query;
1540 if (ipv6_addr_is_ll_all_nodes(group))
1541 return;
1543 own_query = port ? &port->ip6_own_query : &br->ip6_own_query;
1545 br_group.u.ip6 = *group;
1546 br_group.proto = htons(ETH_P_IPV6);
1547 br_group.vid = vid;
1549 br_multicast_leave_group(br, port, &br_group, &br->ip6_other_query,
1550 own_query);
1552 #endif
1554 static int br_multicast_ipv4_rcv(struct net_bridge *br,
1555 struct net_bridge_port *port,
1556 struct sk_buff *skb,
1557 u16 vid)
1559 struct sk_buff *skb2 = skb;
1560 const struct iphdr *iph;
1561 struct igmphdr *ih;
1562 unsigned int len;
1563 unsigned int offset;
1564 int err;
1566 /* We treat OOM as packet loss for now. */
1567 if (!pskb_may_pull(skb, sizeof(*iph)))
1568 return -EINVAL;
1570 iph = ip_hdr(skb);
1572 if (iph->ihl < 5 || iph->version != 4)
1573 return -EINVAL;
1575 if (!pskb_may_pull(skb, ip_hdrlen(skb)))
1576 return -EINVAL;
1578 iph = ip_hdr(skb);
1580 if (unlikely(ip_fast_csum((u8 *)iph, iph->ihl)))
1581 return -EINVAL;
1583 if (iph->protocol != IPPROTO_IGMP) {
1584 if (!ipv4_is_local_multicast(iph->daddr))
1585 BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1586 return 0;
1589 len = ntohs(iph->tot_len);
1590 if (skb->len < len || len < ip_hdrlen(skb))
1591 return -EINVAL;
1593 if (skb->len > len) {
1594 skb2 = skb_clone(skb, GFP_ATOMIC);
1595 if (!skb2)
1596 return -ENOMEM;
1598 err = pskb_trim_rcsum(skb2, len);
1599 if (err)
1600 goto err_out;
1603 len -= ip_hdrlen(skb2);
1604 offset = skb_network_offset(skb2) + ip_hdrlen(skb2);
1605 __skb_pull(skb2, offset);
1606 skb_reset_transport_header(skb2);
1608 err = -EINVAL;
1609 if (!pskb_may_pull(skb2, sizeof(*ih)))
1610 goto out;
1612 switch (skb2->ip_summed) {
1613 case CHECKSUM_COMPLETE:
1614 if (!csum_fold(skb2->csum))
1615 break;
1616 /* fall through */
1617 case CHECKSUM_NONE:
1618 skb2->csum = 0;
1619 if (skb_checksum_complete(skb2))
1620 goto out;
1623 err = 0;
1625 BR_INPUT_SKB_CB(skb)->igmp = 1;
1626 ih = igmp_hdr(skb2);
1628 switch (ih->type) {
1629 case IGMP_HOST_MEMBERSHIP_REPORT:
1630 case IGMPV2_HOST_MEMBERSHIP_REPORT:
1631 BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1632 err = br_ip4_multicast_add_group(br, port, ih->group, vid);
1633 break;
1634 case IGMPV3_HOST_MEMBERSHIP_REPORT:
1635 err = br_ip4_multicast_igmp3_report(br, port, skb2, vid);
1636 break;
1637 case IGMP_HOST_MEMBERSHIP_QUERY:
1638 err = br_ip4_multicast_query(br, port, skb2, vid);
1639 break;
1640 case IGMP_HOST_LEAVE_MESSAGE:
1641 br_ip4_multicast_leave_group(br, port, ih->group, vid);
1642 break;
1645 out:
1646 __skb_push(skb2, offset);
1647 err_out:
1648 if (skb2 != skb)
1649 kfree_skb(skb2);
1650 return err;
1653 #if IS_ENABLED(CONFIG_IPV6)
1654 static int br_multicast_ipv6_rcv(struct net_bridge *br,
1655 struct net_bridge_port *port,
1656 struct sk_buff *skb,
1657 u16 vid)
1659 struct sk_buff *skb2;
1660 const struct ipv6hdr *ip6h;
1661 u8 icmp6_type;
1662 u8 nexthdr;
1663 __be16 frag_off;
1664 unsigned int len;
1665 int offset;
1666 int err;
1668 if (!pskb_may_pull(skb, sizeof(*ip6h)))
1669 return -EINVAL;
1671 ip6h = ipv6_hdr(skb);
1674 * We're interested in MLD messages only.
1675 * - Version is 6
1676 * - MLD has always Router Alert hop-by-hop option
1677 * - But we do not support jumbrograms.
1679 if (ip6h->version != 6)
1680 return 0;
1682 /* Prevent flooding this packet if there is no listener present */
1683 if (!ipv6_addr_is_ll_all_nodes(&ip6h->daddr))
1684 BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1686 if (ip6h->nexthdr != IPPROTO_HOPOPTS ||
1687 ip6h->payload_len == 0)
1688 return 0;
1690 len = ntohs(ip6h->payload_len) + sizeof(*ip6h);
1691 if (skb->len < len)
1692 return -EINVAL;
1694 nexthdr = ip6h->nexthdr;
1695 offset = ipv6_skip_exthdr(skb, sizeof(*ip6h), &nexthdr, &frag_off);
1697 if (offset < 0 || nexthdr != IPPROTO_ICMPV6)
1698 return 0;
1700 /* Okay, we found ICMPv6 header */
1701 skb2 = skb_clone(skb, GFP_ATOMIC);
1702 if (!skb2)
1703 return -ENOMEM;
1705 err = -EINVAL;
1706 if (!pskb_may_pull(skb2, offset + sizeof(struct icmp6hdr)))
1707 goto out;
1709 len -= offset - skb_network_offset(skb2);
1711 __skb_pull(skb2, offset);
1712 skb_reset_transport_header(skb2);
1713 skb_postpull_rcsum(skb2, skb_network_header(skb2),
1714 skb_network_header_len(skb2));
1716 icmp6_type = icmp6_hdr(skb2)->icmp6_type;
1718 switch (icmp6_type) {
1719 case ICMPV6_MGM_QUERY:
1720 case ICMPV6_MGM_REPORT:
1721 case ICMPV6_MGM_REDUCTION:
1722 case ICMPV6_MLD2_REPORT:
1723 break;
1724 default:
1725 err = 0;
1726 goto out;
1729 /* Okay, we found MLD message. Check further. */
1730 if (skb2->len > len) {
1731 err = pskb_trim_rcsum(skb2, len);
1732 if (err)
1733 goto out;
1734 err = -EINVAL;
1737 ip6h = ipv6_hdr(skb2);
1739 switch (skb2->ip_summed) {
1740 case CHECKSUM_COMPLETE:
1741 if (!csum_ipv6_magic(&ip6h->saddr, &ip6h->daddr, skb2->len,
1742 IPPROTO_ICMPV6, skb2->csum))
1743 break;
1744 /*FALLTHROUGH*/
1745 case CHECKSUM_NONE:
1746 skb2->csum = ~csum_unfold(csum_ipv6_magic(&ip6h->saddr,
1747 &ip6h->daddr,
1748 skb2->len,
1749 IPPROTO_ICMPV6, 0));
1750 if (__skb_checksum_complete(skb2))
1751 goto out;
1754 err = 0;
1756 BR_INPUT_SKB_CB(skb)->igmp = 1;
1758 switch (icmp6_type) {
1759 case ICMPV6_MGM_REPORT:
1761 struct mld_msg *mld;
1762 if (!pskb_may_pull(skb2, sizeof(*mld))) {
1763 err = -EINVAL;
1764 goto out;
1766 mld = (struct mld_msg *)skb_transport_header(skb2);
1767 BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1768 err = br_ip6_multicast_add_group(br, port, &mld->mld_mca, vid);
1769 break;
1771 case ICMPV6_MLD2_REPORT:
1772 err = br_ip6_multicast_mld2_report(br, port, skb2, vid);
1773 break;
1774 case ICMPV6_MGM_QUERY:
1775 err = br_ip6_multicast_query(br, port, skb2, vid);
1776 break;
1777 case ICMPV6_MGM_REDUCTION:
1779 struct mld_msg *mld;
1780 if (!pskb_may_pull(skb2, sizeof(*mld))) {
1781 err = -EINVAL;
1782 goto out;
1784 mld = (struct mld_msg *)skb_transport_header(skb2);
1785 br_ip6_multicast_leave_group(br, port, &mld->mld_mca, vid);
1789 out:
1790 kfree_skb(skb2);
1791 return err;
1793 #endif
1795 int br_multicast_rcv(struct net_bridge *br, struct net_bridge_port *port,
1796 struct sk_buff *skb, u16 vid)
1798 BR_INPUT_SKB_CB(skb)->igmp = 0;
1799 BR_INPUT_SKB_CB(skb)->mrouters_only = 0;
1801 if (br->multicast_disabled)
1802 return 0;
1804 switch (skb->protocol) {
1805 case htons(ETH_P_IP):
1806 return br_multicast_ipv4_rcv(br, port, skb, vid);
1807 #if IS_ENABLED(CONFIG_IPV6)
1808 case htons(ETH_P_IPV6):
1809 return br_multicast_ipv6_rcv(br, port, skb, vid);
1810 #endif
1813 return 0;
1816 static void br_multicast_query_expired(struct net_bridge *br,
1817 struct bridge_mcast_own_query *query,
1818 struct bridge_mcast_querier *querier)
1820 spin_lock(&br->multicast_lock);
1821 if (query->startup_sent < br->multicast_startup_query_count)
1822 query->startup_sent++;
1824 RCU_INIT_POINTER(querier, NULL);
1825 br_multicast_send_query(br, NULL, query);
1826 spin_unlock(&br->multicast_lock);
1829 static void br_ip4_multicast_query_expired(unsigned long data)
1831 struct net_bridge *br = (void *)data;
1833 br_multicast_query_expired(br, &br->ip4_own_query, &br->ip4_querier);
1836 #if IS_ENABLED(CONFIG_IPV6)
1837 static void br_ip6_multicast_query_expired(unsigned long data)
1839 struct net_bridge *br = (void *)data;
1841 br_multicast_query_expired(br, &br->ip6_own_query, &br->ip6_querier);
1843 #endif
1845 void br_multicast_init(struct net_bridge *br)
1847 br->hash_elasticity = 4;
1848 br->hash_max = 512;
1850 br->multicast_router = 1;
1851 br->multicast_querier = 0;
1852 br->multicast_query_use_ifaddr = 0;
1853 br->multicast_last_member_count = 2;
1854 br->multicast_startup_query_count = 2;
1856 br->multicast_last_member_interval = HZ;
1857 br->multicast_query_response_interval = 10 * HZ;
1858 br->multicast_startup_query_interval = 125 * HZ / 4;
1859 br->multicast_query_interval = 125 * HZ;
1860 br->multicast_querier_interval = 255 * HZ;
1861 br->multicast_membership_interval = 260 * HZ;
1863 br->ip4_other_query.delay_time = 0;
1864 br->ip4_querier.port = NULL;
1865 #if IS_ENABLED(CONFIG_IPV6)
1866 br->ip6_other_query.delay_time = 0;
1867 br->ip6_querier.port = NULL;
1868 #endif
1870 spin_lock_init(&br->multicast_lock);
1871 setup_timer(&br->multicast_router_timer,
1872 br_multicast_local_router_expired, 0);
1873 setup_timer(&br->ip4_other_query.timer,
1874 br_ip4_multicast_querier_expired, (unsigned long)br);
1875 setup_timer(&br->ip4_own_query.timer, br_ip4_multicast_query_expired,
1876 (unsigned long)br);
1877 #if IS_ENABLED(CONFIG_IPV6)
1878 setup_timer(&br->ip6_other_query.timer,
1879 br_ip6_multicast_querier_expired, (unsigned long)br);
1880 setup_timer(&br->ip6_own_query.timer, br_ip6_multicast_query_expired,
1881 (unsigned long)br);
1882 #endif
1885 static void __br_multicast_open(struct net_bridge *br,
1886 struct bridge_mcast_own_query *query)
1888 query->startup_sent = 0;
1890 if (br->multicast_disabled)
1891 return;
1893 mod_timer(&query->timer, jiffies);
1896 void br_multicast_open(struct net_bridge *br)
1898 __br_multicast_open(br, &br->ip4_own_query);
1899 #if IS_ENABLED(CONFIG_IPV6)
1900 __br_multicast_open(br, &br->ip6_own_query);
1901 #endif
1904 void br_multicast_stop(struct net_bridge *br)
1906 struct net_bridge_mdb_htable *mdb;
1907 struct net_bridge_mdb_entry *mp;
1908 struct hlist_node *n;
1909 u32 ver;
1910 int i;
1912 del_timer_sync(&br->multicast_router_timer);
1913 del_timer_sync(&br->ip4_other_query.timer);
1914 del_timer_sync(&br->ip4_own_query.timer);
1915 #if IS_ENABLED(CONFIG_IPV6)
1916 del_timer_sync(&br->ip6_other_query.timer);
1917 del_timer_sync(&br->ip6_own_query.timer);
1918 #endif
1920 spin_lock_bh(&br->multicast_lock);
1921 mdb = mlock_dereference(br->mdb, br);
1922 if (!mdb)
1923 goto out;
1925 br->mdb = NULL;
1927 ver = mdb->ver;
1928 for (i = 0; i < mdb->max; i++) {
1929 hlist_for_each_entry_safe(mp, n, &mdb->mhash[i],
1930 hlist[ver]) {
1931 del_timer(&mp->timer);
1932 call_rcu_bh(&mp->rcu, br_multicast_free_group);
1936 if (mdb->old) {
1937 spin_unlock_bh(&br->multicast_lock);
1938 rcu_barrier_bh();
1939 spin_lock_bh(&br->multicast_lock);
1940 WARN_ON(mdb->old);
1943 mdb->old = mdb;
1944 call_rcu_bh(&mdb->rcu, br_mdb_free);
1946 out:
1947 spin_unlock_bh(&br->multicast_lock);
1950 int br_multicast_set_router(struct net_bridge *br, unsigned long val)
1952 int err = -ENOENT;
1954 spin_lock_bh(&br->multicast_lock);
1955 if (!netif_running(br->dev))
1956 goto unlock;
1958 switch (val) {
1959 case 0:
1960 case 2:
1961 del_timer(&br->multicast_router_timer);
1962 /* fall through */
1963 case 1:
1964 br->multicast_router = val;
1965 err = 0;
1966 break;
1968 default:
1969 err = -EINVAL;
1970 break;
1973 unlock:
1974 spin_unlock_bh(&br->multicast_lock);
1976 return err;
1979 int br_multicast_set_port_router(struct net_bridge_port *p, unsigned long val)
1981 struct net_bridge *br = p->br;
1982 int err = -ENOENT;
1984 spin_lock(&br->multicast_lock);
1985 if (!netif_running(br->dev) || p->state == BR_STATE_DISABLED)
1986 goto unlock;
1988 switch (val) {
1989 case 0:
1990 case 1:
1991 case 2:
1992 p->multicast_router = val;
1993 err = 0;
1995 if (val < 2 && !hlist_unhashed(&p->rlist))
1996 hlist_del_init_rcu(&p->rlist);
1998 if (val == 1)
1999 break;
2001 del_timer(&p->multicast_router_timer);
2003 if (val == 0)
2004 break;
2006 br_multicast_add_router(br, p);
2007 break;
2009 default:
2010 err = -EINVAL;
2011 break;
2014 unlock:
2015 spin_unlock(&br->multicast_lock);
2017 return err;
2020 static void br_multicast_start_querier(struct net_bridge *br,
2021 struct bridge_mcast_own_query *query)
2023 struct net_bridge_port *port;
2025 __br_multicast_open(br, query);
2027 list_for_each_entry(port, &br->port_list, list) {
2028 if (port->state == BR_STATE_DISABLED ||
2029 port->state == BR_STATE_BLOCKING)
2030 continue;
2032 if (query == &br->ip4_own_query)
2033 br_multicast_enable(&port->ip4_own_query);
2034 #if IS_ENABLED(CONFIG_IPV6)
2035 else
2036 br_multicast_enable(&port->ip6_own_query);
2037 #endif
2041 int br_multicast_toggle(struct net_bridge *br, unsigned long val)
2043 int err = 0;
2044 struct net_bridge_mdb_htable *mdb;
2046 spin_lock_bh(&br->multicast_lock);
2047 if (br->multicast_disabled == !val)
2048 goto unlock;
2050 br->multicast_disabled = !val;
2051 if (br->multicast_disabled)
2052 goto unlock;
2054 if (!netif_running(br->dev))
2055 goto unlock;
2057 mdb = mlock_dereference(br->mdb, br);
2058 if (mdb) {
2059 if (mdb->old) {
2060 err = -EEXIST;
2061 rollback:
2062 br->multicast_disabled = !!val;
2063 goto unlock;
2066 err = br_mdb_rehash(&br->mdb, mdb->max,
2067 br->hash_elasticity);
2068 if (err)
2069 goto rollback;
2072 br_multicast_start_querier(br, &br->ip4_own_query);
2073 #if IS_ENABLED(CONFIG_IPV6)
2074 br_multicast_start_querier(br, &br->ip6_own_query);
2075 #endif
2077 unlock:
2078 spin_unlock_bh(&br->multicast_lock);
2080 return err;
2083 int br_multicast_set_querier(struct net_bridge *br, unsigned long val)
2085 unsigned long max_delay;
2087 val = !!val;
2089 spin_lock_bh(&br->multicast_lock);
2090 if (br->multicast_querier == val)
2091 goto unlock;
2093 br->multicast_querier = val;
2094 if (!val)
2095 goto unlock;
2097 max_delay = br->multicast_query_response_interval;
2099 if (!timer_pending(&br->ip4_other_query.timer))
2100 br->ip4_other_query.delay_time = jiffies + max_delay;
2102 br_multicast_start_querier(br, &br->ip4_own_query);
2104 #if IS_ENABLED(CONFIG_IPV6)
2105 if (!timer_pending(&br->ip6_other_query.timer))
2106 br->ip6_other_query.delay_time = jiffies + max_delay;
2108 br_multicast_start_querier(br, &br->ip6_own_query);
2109 #endif
2111 unlock:
2112 spin_unlock_bh(&br->multicast_lock);
2114 return 0;
2117 int br_multicast_set_hash_max(struct net_bridge *br, unsigned long val)
2119 int err = -ENOENT;
2120 u32 old;
2121 struct net_bridge_mdb_htable *mdb;
2123 spin_lock_bh(&br->multicast_lock);
2124 if (!netif_running(br->dev))
2125 goto unlock;
2127 err = -EINVAL;
2128 if (!is_power_of_2(val))
2129 goto unlock;
2131 mdb = mlock_dereference(br->mdb, br);
2132 if (mdb && val < mdb->size)
2133 goto unlock;
2135 err = 0;
2137 old = br->hash_max;
2138 br->hash_max = val;
2140 if (mdb) {
2141 if (mdb->old) {
2142 err = -EEXIST;
2143 rollback:
2144 br->hash_max = old;
2145 goto unlock;
2148 err = br_mdb_rehash(&br->mdb, br->hash_max,
2149 br->hash_elasticity);
2150 if (err)
2151 goto rollback;
2154 unlock:
2155 spin_unlock_bh(&br->multicast_lock);
2157 return err;
2161 * br_multicast_list_adjacent - Returns snooped multicast addresses
2162 * @dev: The bridge port adjacent to which to retrieve addresses
2163 * @br_ip_list: The list to store found, snooped multicast IP addresses in
2165 * Creates a list of IP addresses (struct br_ip_list) sensed by the multicast
2166 * snooping feature on all bridge ports of dev's bridge device, excluding
2167 * the addresses from dev itself.
2169 * Returns the number of items added to br_ip_list.
2171 * Notes:
2172 * - br_ip_list needs to be initialized by caller
2173 * - br_ip_list might contain duplicates in the end
2174 * (needs to be taken care of by caller)
2175 * - br_ip_list needs to be freed by caller
2177 int br_multicast_list_adjacent(struct net_device *dev,
2178 struct list_head *br_ip_list)
2180 struct net_bridge *br;
2181 struct net_bridge_port *port;
2182 struct net_bridge_port_group *group;
2183 struct br_ip_list *entry;
2184 int count = 0;
2186 rcu_read_lock();
2187 if (!br_ip_list || !br_port_exists(dev))
2188 goto unlock;
2190 port = br_port_get_rcu(dev);
2191 if (!port || !port->br)
2192 goto unlock;
2194 br = port->br;
2196 list_for_each_entry_rcu(port, &br->port_list, list) {
2197 if (!port->dev || port->dev == dev)
2198 continue;
2200 hlist_for_each_entry_rcu(group, &port->mglist, mglist) {
2201 entry = kmalloc(sizeof(*entry), GFP_ATOMIC);
2202 if (!entry)
2203 goto unlock;
2205 entry->addr = group->addr;
2206 list_add(&entry->list, br_ip_list);
2207 count++;
2211 unlock:
2212 rcu_read_unlock();
2213 return count;
2215 EXPORT_SYMBOL_GPL(br_multicast_list_adjacent);
2218 * br_multicast_has_querier_anywhere - Checks for a querier on a bridge
2219 * @dev: The bridge port providing the bridge on which to check for a querier
2220 * @proto: The protocol family to check for: IGMP -> ETH_P_IP, MLD -> ETH_P_IPV6
2222 * Checks whether the given interface has a bridge on top and if so returns
2223 * true if a valid querier exists anywhere on the bridged link layer.
2224 * Otherwise returns false.
2226 bool br_multicast_has_querier_anywhere(struct net_device *dev, int proto)
2228 struct net_bridge *br;
2229 struct net_bridge_port *port;
2230 struct ethhdr eth;
2231 bool ret = false;
2233 rcu_read_lock();
2234 if (!br_port_exists(dev))
2235 goto unlock;
2237 port = br_port_get_rcu(dev);
2238 if (!port || !port->br)
2239 goto unlock;
2241 br = port->br;
2243 memset(&eth, 0, sizeof(eth));
2244 eth.h_proto = htons(proto);
2246 ret = br_multicast_querier_exists(br, &eth);
2248 unlock:
2249 rcu_read_unlock();
2250 return ret;
2252 EXPORT_SYMBOL_GPL(br_multicast_has_querier_anywhere);
2255 * br_multicast_has_querier_adjacent - Checks for a querier behind a bridge port
2256 * @dev: The bridge port adjacent to which to check for a querier
2257 * @proto: The protocol family to check for: IGMP -> ETH_P_IP, MLD -> ETH_P_IPV6
2259 * Checks whether the given interface has a bridge on top and if so returns
2260 * true if a selected querier is behind one of the other ports of this
2261 * bridge. Otherwise returns false.
2263 bool br_multicast_has_querier_adjacent(struct net_device *dev, int proto)
2265 struct net_bridge *br;
2266 struct net_bridge_port *port;
2267 bool ret = false;
2269 rcu_read_lock();
2270 if (!br_port_exists(dev))
2271 goto unlock;
2273 port = br_port_get_rcu(dev);
2274 if (!port || !port->br)
2275 goto unlock;
2277 br = port->br;
2279 switch (proto) {
2280 case ETH_P_IP:
2281 if (!timer_pending(&br->ip4_other_query.timer) ||
2282 rcu_dereference(br->ip4_querier.port) == port)
2283 goto unlock;
2284 break;
2285 #if IS_ENABLED(CONFIG_IPV6)
2286 case ETH_P_IPV6:
2287 if (!timer_pending(&br->ip6_other_query.timer) ||
2288 rcu_dereference(br->ip6_querier.port) == port)
2289 goto unlock;
2290 break;
2291 #endif
2292 default:
2293 goto unlock;
2296 ret = true;
2297 unlock:
2298 rcu_read_unlock();
2299 return ret;
2301 EXPORT_SYMBOL_GPL(br_multicast_has_querier_adjacent);