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)
13 #include <linux/err.h>
14 #include <linux/if_ether.h>
15 #include <linux/igmp.h>
16 #include <linux/jhash.h>
17 #include <linux/kernel.h>
18 #include <linux/log2.h>
19 #include <linux/netdevice.h>
20 #include <linux/netfilter_bridge.h>
21 #include <linux/random.h>
22 #include <linux/rculist.h>
23 #include <linux/skbuff.h>
24 #include <linux/slab.h>
25 #include <linux/timer.h>
27 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
30 #include <net/addrconf.h>
31 #include <net/ip6_checksum.h>
34 #include "br_private.h"
36 #define mlock_dereference(X, br) \
37 rcu_dereference_protected(X, lockdep_is_held(&br->multicast_lock))
39 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
40 static inline int ipv6_is_transient_multicast(const struct in6_addr *addr)
42 if (ipv6_addr_is_multicast(addr) && IPV6_ADDR_MC_FLAG_TRANSIENT(addr))
48 static inline int br_ip_equal(const struct br_ip *a, const struct br_ip *b)
50 if (a->proto != b->proto)
54 return a->u.ip4 == b->u.ip4;
55 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
56 case htons(ETH_P_IPV6):
57 return ipv6_addr_equal(&a->u.ip6, &b->u.ip6);
63 static inline int __br_ip4_hash(struct net_bridge_mdb_htable *mdb, __be32 ip)
65 return jhash_1word(mdb->secret, (__force u32)ip) & (mdb->max - 1);
68 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
69 static inline int __br_ip6_hash(struct net_bridge_mdb_htable *mdb,
70 const struct in6_addr *ip)
72 return jhash2((__force u32 *)ip->s6_addr32, 4, mdb->secret) & (mdb->max - 1);
76 static inline int br_ip_hash(struct net_bridge_mdb_htable *mdb,
81 return __br_ip4_hash(mdb, ip->u.ip4);
82 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
83 case htons(ETH_P_IPV6):
84 return __br_ip6_hash(mdb, &ip->u.ip6);
90 static struct net_bridge_mdb_entry *__br_mdb_ip_get(
91 struct net_bridge_mdb_htable *mdb, struct br_ip *dst, int hash)
93 struct net_bridge_mdb_entry *mp;
96 hlist_for_each_entry_rcu(mp, p, &mdb->mhash[hash], hlist[mdb->ver]) {
97 if (br_ip_equal(&mp->addr, dst))
104 static struct net_bridge_mdb_entry *br_mdb_ip_get(
105 struct net_bridge_mdb_htable *mdb, struct br_ip *dst)
110 return __br_mdb_ip_get(mdb, dst, br_ip_hash(mdb, dst));
113 static struct net_bridge_mdb_entry *br_mdb_ip4_get(
114 struct net_bridge_mdb_htable *mdb, __be32 dst)
119 br_dst.proto = htons(ETH_P_IP);
121 return br_mdb_ip_get(mdb, &br_dst);
124 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
125 static struct net_bridge_mdb_entry *br_mdb_ip6_get(
126 struct net_bridge_mdb_htable *mdb, const struct in6_addr *dst)
130 ipv6_addr_copy(&br_dst.u.ip6, dst);
131 br_dst.proto = htons(ETH_P_IPV6);
133 return br_mdb_ip_get(mdb, &br_dst);
137 struct net_bridge_mdb_entry *br_mdb_get(struct net_bridge *br,
140 struct net_bridge_mdb_htable *mdb = rcu_dereference(br->mdb);
143 if (br->multicast_disabled)
146 if (BR_INPUT_SKB_CB(skb)->igmp)
149 ip.proto = skb->protocol;
151 switch (skb->protocol) {
152 case htons(ETH_P_IP):
153 ip.u.ip4 = ip_hdr(skb)->daddr;
155 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
156 case htons(ETH_P_IPV6):
157 ipv6_addr_copy(&ip.u.ip6, &ipv6_hdr(skb)->daddr);
164 return br_mdb_ip_get(mdb, &ip);
167 static void br_mdb_free(struct rcu_head *head)
169 struct net_bridge_mdb_htable *mdb =
170 container_of(head, struct net_bridge_mdb_htable, rcu);
171 struct net_bridge_mdb_htable *old = mdb->old;
178 static int br_mdb_copy(struct net_bridge_mdb_htable *new,
179 struct net_bridge_mdb_htable *old,
182 struct net_bridge_mdb_entry *mp;
183 struct hlist_node *p;
188 for (i = 0; i < old->max; i++)
189 hlist_for_each_entry(mp, p, &old->mhash[i], hlist[old->ver])
190 hlist_add_head(&mp->hlist[new->ver],
191 &new->mhash[br_ip_hash(new, &mp->addr)]);
197 for (i = 0; i < new->max; i++) {
199 hlist_for_each_entry(mp, p, &new->mhash[i], hlist[new->ver])
205 return maxlen > elasticity ? -EINVAL : 0;
208 static void br_multicast_free_pg(struct rcu_head *head)
210 struct net_bridge_port_group *p =
211 container_of(head, struct net_bridge_port_group, rcu);
216 static void br_multicast_free_group(struct rcu_head *head)
218 struct net_bridge_mdb_entry *mp =
219 container_of(head, struct net_bridge_mdb_entry, rcu);
224 static void br_multicast_group_expired(unsigned long data)
226 struct net_bridge_mdb_entry *mp = (void *)data;
227 struct net_bridge *br = mp->br;
228 struct net_bridge_mdb_htable *mdb;
230 spin_lock(&br->multicast_lock);
231 if (!netif_running(br->dev) || timer_pending(&mp->timer))
239 mdb = mlock_dereference(br->mdb, br);
241 hlist_del_rcu(&mp->hlist[mdb->ver]);
244 del_timer(&mp->query_timer);
245 call_rcu_bh(&mp->rcu, br_multicast_free_group);
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);
265 for (pp = &mp->ports;
266 (p = mlock_dereference(*pp, br)) != NULL;
271 rcu_assign_pointer(*pp, p->next);
272 hlist_del_init(&p->mglist);
273 del_timer(&p->timer);
274 del_timer(&p->query_timer);
275 call_rcu_bh(&p->rcu, br_multicast_free_pg);
277 if (!mp->ports && !mp->mglist &&
278 netif_running(br->dev))
279 mod_timer(&mp->timer, jiffies);
287 static void br_multicast_port_group_expired(unsigned long data)
289 struct net_bridge_port_group *pg = (void *)data;
290 struct net_bridge *br = pg->port->br;
292 spin_lock(&br->multicast_lock);
293 if (!netif_running(br->dev) || timer_pending(&pg->timer) ||
294 hlist_unhashed(&pg->mglist))
297 br_multicast_del_pg(br, pg);
300 spin_unlock(&br->multicast_lock);
303 static int br_mdb_rehash(struct net_bridge_mdb_htable __rcu **mdbp, int max,
306 struct net_bridge_mdb_htable *old = rcu_dereference_protected(*mdbp, 1);
307 struct net_bridge_mdb_htable *mdb;
310 mdb = kmalloc(sizeof(*mdb), GFP_ATOMIC);
317 mdb->mhash = kzalloc(max * sizeof(*mdb->mhash), GFP_ATOMIC);
323 mdb->size = old ? old->size : 0;
324 mdb->ver = old ? old->ver ^ 1 : 0;
326 if (!old || elasticity)
327 get_random_bytes(&mdb->secret, sizeof(mdb->secret));
329 mdb->secret = old->secret;
334 err = br_mdb_copy(mdb, old, elasticity);
341 call_rcu_bh(&mdb->rcu, br_mdb_free);
344 rcu_assign_pointer(*mdbp, mdb);
349 static struct sk_buff *br_ip4_multicast_alloc_query(struct net_bridge *br,
357 skb = netdev_alloc_skb_ip_align(br->dev, sizeof(*eth) + sizeof(*iph) +
362 skb->protocol = htons(ETH_P_IP);
364 skb_reset_mac_header(skb);
367 memcpy(eth->h_source, br->dev->dev_addr, 6);
370 eth->h_dest[2] = 0x5e;
374 eth->h_proto = htons(ETH_P_IP);
375 skb_put(skb, sizeof(*eth));
377 skb_set_network_header(skb, skb->len);
383 iph->tot_len = htons(sizeof(*iph) + sizeof(*ih) + 4);
385 iph->frag_off = htons(IP_DF);
387 iph->protocol = IPPROTO_IGMP;
389 iph->daddr = htonl(INADDR_ALLHOSTS_GROUP);
390 ((u8 *)&iph[1])[0] = IPOPT_RA;
391 ((u8 *)&iph[1])[1] = 4;
392 ((u8 *)&iph[1])[2] = 0;
393 ((u8 *)&iph[1])[3] = 0;
397 skb_set_transport_header(skb, skb->len);
399 ih->type = IGMP_HOST_MEMBERSHIP_QUERY;
400 ih->code = (group ? br->multicast_last_member_interval :
401 br->multicast_query_response_interval) /
402 (HZ / IGMP_TIMER_SCALE);
405 ih->csum = ip_compute_csum((void *)ih, sizeof(struct igmphdr));
406 skb_put(skb, sizeof(*ih));
408 __skb_pull(skb, sizeof(*eth));
414 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
415 static struct sk_buff *br_ip6_multicast_alloc_query(struct net_bridge *br,
416 struct in6_addr *group)
419 struct ipv6hdr *ip6h;
420 struct mld_msg *mldq;
423 unsigned long interval;
425 skb = netdev_alloc_skb_ip_align(br->dev, sizeof(*eth) + sizeof(*ip6h) +
430 skb->protocol = htons(ETH_P_IPV6);
432 /* Ethernet header */
433 skb_reset_mac_header(skb);
436 memcpy(eth->h_source, br->dev->dev_addr, 6);
437 eth->h_proto = htons(ETH_P_IPV6);
438 skb_put(skb, sizeof(*eth));
440 /* IPv6 header + HbH option */
441 skb_set_network_header(skb, skb->len);
442 ip6h = ipv6_hdr(skb);
444 *(__force __be32 *)ip6h = htonl(0x60000000);
445 ip6h->payload_len = htons(8 + sizeof(*mldq));
446 ip6h->nexthdr = IPPROTO_HOPOPTS;
448 ipv6_addr_set(&ip6h->saddr, 0, 0, 0, 0);
449 ipv6_addr_set(&ip6h->daddr, htonl(0xff020000), 0, 0, htonl(1));
450 ipv6_eth_mc_map(&ip6h->daddr, eth->h_dest);
452 hopopt = (u8 *)(ip6h + 1);
453 hopopt[0] = IPPROTO_ICMPV6; /* next hdr */
454 hopopt[1] = 0; /* length of HbH */
455 hopopt[2] = IPV6_TLV_ROUTERALERT; /* Router Alert */
456 hopopt[3] = 2; /* Length of RA Option */
457 hopopt[4] = 0; /* Type = 0x0000 (MLD) */
459 hopopt[6] = IPV6_TLV_PAD0; /* Pad0 */
460 hopopt[7] = IPV6_TLV_PAD0; /* Pad0 */
462 skb_put(skb, sizeof(*ip6h) + 8);
465 skb_set_transport_header(skb, skb->len);
466 mldq = (struct mld_msg *) icmp6_hdr(skb);
468 interval = ipv6_addr_any(group) ? br->multicast_last_member_interval :
469 br->multicast_query_response_interval;
471 mldq->mld_type = ICMPV6_MGM_QUERY;
474 mldq->mld_maxdelay = htons((u16)jiffies_to_msecs(interval));
475 mldq->mld_reserved = 0;
476 ipv6_addr_copy(&mldq->mld_mca, group);
479 mldq->mld_cksum = csum_ipv6_magic(&ip6h->saddr, &ip6h->daddr,
480 sizeof(*mldq), IPPROTO_ICMPV6,
483 skb_put(skb, sizeof(*mldq));
485 __skb_pull(skb, sizeof(*eth));
492 static struct sk_buff *br_multicast_alloc_query(struct net_bridge *br,
495 switch (addr->proto) {
496 case htons(ETH_P_IP):
497 return br_ip4_multicast_alloc_query(br, addr->u.ip4);
498 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
499 case htons(ETH_P_IPV6):
500 return br_ip6_multicast_alloc_query(br, &addr->u.ip6);
506 static void br_multicast_send_group_query(struct net_bridge_mdb_entry *mp)
508 struct net_bridge *br = mp->br;
511 skb = br_multicast_alloc_query(br, &mp->addr);
518 if (++mp->queries_sent < br->multicast_last_member_count)
519 mod_timer(&mp->query_timer,
520 jiffies + br->multicast_last_member_interval);
523 static void br_multicast_group_query_expired(unsigned long data)
525 struct net_bridge_mdb_entry *mp = (void *)data;
526 struct net_bridge *br = mp->br;
528 spin_lock(&br->multicast_lock);
529 if (!netif_running(br->dev) || !mp->mglist ||
530 mp->queries_sent >= br->multicast_last_member_count)
533 br_multicast_send_group_query(mp);
536 spin_unlock(&br->multicast_lock);
539 static void br_multicast_send_port_group_query(struct net_bridge_port_group *pg)
541 struct net_bridge_port *port = pg->port;
542 struct net_bridge *br = port->br;
545 skb = br_multicast_alloc_query(br, &pg->addr);
549 br_deliver(port, skb);
552 if (++pg->queries_sent < br->multicast_last_member_count)
553 mod_timer(&pg->query_timer,
554 jiffies + br->multicast_last_member_interval);
557 static void br_multicast_port_group_query_expired(unsigned long data)
559 struct net_bridge_port_group *pg = (void *)data;
560 struct net_bridge_port *port = pg->port;
561 struct net_bridge *br = port->br;
563 spin_lock(&br->multicast_lock);
564 if (!netif_running(br->dev) || hlist_unhashed(&pg->mglist) ||
565 pg->queries_sent >= br->multicast_last_member_count)
568 br_multicast_send_port_group_query(pg);
571 spin_unlock(&br->multicast_lock);
574 static struct net_bridge_mdb_entry *br_multicast_get_group(
575 struct net_bridge *br, struct net_bridge_port *port,
576 struct br_ip *group, int hash)
578 struct net_bridge_mdb_htable *mdb;
579 struct net_bridge_mdb_entry *mp;
580 struct hlist_node *p;
586 mdb = rcu_dereference_protected(br->mdb, 1);
587 hlist_for_each_entry(mp, p, &mdb->mhash[hash], hlist[mdb->ver]) {
589 if (unlikely(br_ip_equal(group, &mp->addr)))
596 if (unlikely(count > br->hash_elasticity && count)) {
598 br_info(br, "Multicast hash table "
599 "chain limit reached: %s\n",
600 port ? port->dev->name : br->dev->name);
602 elasticity = br->hash_elasticity;
605 if (mdb->size >= max) {
607 if (unlikely(max >= br->hash_max)) {
608 br_warn(br, "Multicast hash table maximum "
609 "reached, disabling snooping: %s, %d\n",
610 port ? port->dev->name : br->dev->name, max);
613 br->multicast_disabled = 1;
618 if (max > mdb->max || elasticity) {
621 br_info(br, "Multicast hash table "
623 port ? port->dev->name : br->dev->name);
628 err = br_mdb_rehash(&br->mdb, max, elasticity);
630 br_warn(br, "Cannot rehash multicast "
631 "hash table, disabling snooping: %s, %d, %d\n",
632 port ? port->dev->name : br->dev->name,
648 static struct net_bridge_mdb_entry *br_multicast_new_group(
649 struct net_bridge *br, struct net_bridge_port *port,
652 struct net_bridge_mdb_htable *mdb;
653 struct net_bridge_mdb_entry *mp;
657 mdb = rcu_dereference_protected(br->mdb, 1);
659 err = br_mdb_rehash(&br->mdb, BR_HASH_SIZE, 0);
665 hash = br_ip_hash(mdb, group);
666 mp = br_multicast_get_group(br, port, group, hash);
667 switch (PTR_ERR(mp)) {
673 mdb = rcu_dereference_protected(br->mdb, 1);
674 hash = br_ip_hash(mdb, group);
681 mp = kzalloc(sizeof(*mp), GFP_ATOMIC);
683 return ERR_PTR(-ENOMEM);
687 setup_timer(&mp->timer, br_multicast_group_expired,
689 setup_timer(&mp->query_timer, br_multicast_group_query_expired,
692 hlist_add_head_rcu(&mp->hlist[mdb->ver], &mdb->mhash[hash]);
699 static int br_multicast_add_group(struct net_bridge *br,
700 struct net_bridge_port *port,
703 struct net_bridge_mdb_entry *mp;
704 struct net_bridge_port_group *p;
705 struct net_bridge_port_group __rcu **pp;
706 unsigned long now = jiffies;
709 spin_lock(&br->multicast_lock);
710 if (!netif_running(br->dev) ||
711 (port && port->state == BR_STATE_DISABLED))
714 mp = br_multicast_new_group(br, port, group);
721 mod_timer(&mp->timer, now + br->multicast_membership_interval);
725 for (pp = &mp->ports;
726 (p = mlock_dereference(*pp, br)) != NULL;
730 if ((unsigned long)p->port < (unsigned long)port)
734 p = kzalloc(sizeof(*p), GFP_ATOMIC);
742 hlist_add_head(&p->mglist, &port->mglist);
743 setup_timer(&p->timer, br_multicast_port_group_expired,
745 setup_timer(&p->query_timer, br_multicast_port_group_query_expired,
748 rcu_assign_pointer(*pp, p);
751 mod_timer(&p->timer, now + br->multicast_membership_interval);
756 spin_unlock(&br->multicast_lock);
760 static int br_ip4_multicast_add_group(struct net_bridge *br,
761 struct net_bridge_port *port,
764 struct br_ip br_group;
766 if (ipv4_is_local_multicast(group))
769 br_group.u.ip4 = group;
770 br_group.proto = htons(ETH_P_IP);
772 return br_multicast_add_group(br, port, &br_group);
775 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
776 static int br_ip6_multicast_add_group(struct net_bridge *br,
777 struct net_bridge_port *port,
778 const struct in6_addr *group)
780 struct br_ip br_group;
782 if (!ipv6_is_transient_multicast(group))
785 ipv6_addr_copy(&br_group.u.ip6, group);
786 br_group.proto = htons(ETH_P_IPV6);
788 return br_multicast_add_group(br, port, &br_group);
792 static void br_multicast_router_expired(unsigned long data)
794 struct net_bridge_port *port = (void *)data;
795 struct net_bridge *br = port->br;
797 spin_lock(&br->multicast_lock);
798 if (port->multicast_router != 1 ||
799 timer_pending(&port->multicast_router_timer) ||
800 hlist_unhashed(&port->rlist))
803 hlist_del_init_rcu(&port->rlist);
806 spin_unlock(&br->multicast_lock);
809 static void br_multicast_local_router_expired(unsigned long data)
813 static void __br_multicast_send_query(struct net_bridge *br,
814 struct net_bridge_port *port,
819 skb = br_multicast_alloc_query(br, ip);
824 __skb_push(skb, sizeof(struct ethhdr));
825 skb->dev = port->dev;
826 NF_HOOK(NFPROTO_BRIDGE, NF_BR_LOCAL_OUT, skb, NULL, skb->dev,
832 static void br_multicast_send_query(struct net_bridge *br,
833 struct net_bridge_port *port, u32 sent)
836 struct br_ip br_group;
838 if (!netif_running(br->dev) || br->multicast_disabled ||
839 timer_pending(&br->multicast_querier_timer))
842 memset(&br_group.u, 0, sizeof(br_group.u));
844 br_group.proto = htons(ETH_P_IP);
845 __br_multicast_send_query(br, port, &br_group);
847 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
848 br_group.proto = htons(ETH_P_IPV6);
849 __br_multicast_send_query(br, port, &br_group);
853 time += sent < br->multicast_startup_query_count ?
854 br->multicast_startup_query_interval :
855 br->multicast_query_interval;
856 mod_timer(port ? &port->multicast_query_timer :
857 &br->multicast_query_timer, time);
860 static void br_multicast_port_query_expired(unsigned long data)
862 struct net_bridge_port *port = (void *)data;
863 struct net_bridge *br = port->br;
865 spin_lock(&br->multicast_lock);
866 if (port->state == BR_STATE_DISABLED ||
867 port->state == BR_STATE_BLOCKING)
870 if (port->multicast_startup_queries_sent <
871 br->multicast_startup_query_count)
872 port->multicast_startup_queries_sent++;
874 br_multicast_send_query(port->br, port,
875 port->multicast_startup_queries_sent);
878 spin_unlock(&br->multicast_lock);
881 void br_multicast_add_port(struct net_bridge_port *port)
883 port->multicast_router = 1;
885 setup_timer(&port->multicast_router_timer, br_multicast_router_expired,
886 (unsigned long)port);
887 setup_timer(&port->multicast_query_timer,
888 br_multicast_port_query_expired, (unsigned long)port);
891 void br_multicast_del_port(struct net_bridge_port *port)
893 del_timer_sync(&port->multicast_router_timer);
896 static void __br_multicast_enable_port(struct net_bridge_port *port)
898 port->multicast_startup_queries_sent = 0;
900 if (try_to_del_timer_sync(&port->multicast_query_timer) >= 0 ||
901 del_timer(&port->multicast_query_timer))
902 mod_timer(&port->multicast_query_timer, jiffies);
905 void br_multicast_enable_port(struct net_bridge_port *port)
907 struct net_bridge *br = port->br;
909 spin_lock(&br->multicast_lock);
910 if (br->multicast_disabled || !netif_running(br->dev))
913 __br_multicast_enable_port(port);
916 spin_unlock(&br->multicast_lock);
919 void br_multicast_disable_port(struct net_bridge_port *port)
921 struct net_bridge *br = port->br;
922 struct net_bridge_port_group *pg;
923 struct hlist_node *p, *n;
925 spin_lock(&br->multicast_lock);
926 hlist_for_each_entry_safe(pg, p, n, &port->mglist, mglist)
927 br_multicast_del_pg(br, pg);
929 if (!hlist_unhashed(&port->rlist))
930 hlist_del_init_rcu(&port->rlist);
931 del_timer(&port->multicast_router_timer);
932 del_timer(&port->multicast_query_timer);
933 spin_unlock(&br->multicast_lock);
936 static int br_ip4_multicast_igmp3_report(struct net_bridge *br,
937 struct net_bridge_port *port,
940 struct igmpv3_report *ih;
941 struct igmpv3_grec *grec;
949 if (!pskb_may_pull(skb, sizeof(*ih)))
952 ih = igmpv3_report_hdr(skb);
953 num = ntohs(ih->ngrec);
956 for (i = 0; i < num; i++) {
957 len += sizeof(*grec);
958 if (!pskb_may_pull(skb, len))
961 grec = (void *)(skb->data + len - sizeof(*grec));
962 group = grec->grec_mca;
963 type = grec->grec_type;
965 len += ntohs(grec->grec_nsrcs) * 4;
966 if (!pskb_may_pull(skb, len))
969 /* We treat this as an IGMPv2 report for now. */
971 case IGMPV3_MODE_IS_INCLUDE:
972 case IGMPV3_MODE_IS_EXCLUDE:
973 case IGMPV3_CHANGE_TO_INCLUDE:
974 case IGMPV3_CHANGE_TO_EXCLUDE:
975 case IGMPV3_ALLOW_NEW_SOURCES:
976 case IGMPV3_BLOCK_OLD_SOURCES:
983 err = br_ip4_multicast_add_group(br, port, group);
991 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
992 static int br_ip6_multicast_mld2_report(struct net_bridge *br,
993 struct net_bridge_port *port,
996 struct icmp6hdr *icmp6h;
997 struct mld2_grec *grec;
1003 if (!pskb_may_pull(skb, sizeof(*icmp6h)))
1006 icmp6h = icmp6_hdr(skb);
1007 num = ntohs(icmp6h->icmp6_dataun.un_data16[1]);
1008 len = sizeof(*icmp6h);
1010 for (i = 0; i < num; i++) {
1011 __be16 *nsrcs, _nsrcs;
1013 nsrcs = skb_header_pointer(skb,
1014 len + offsetof(struct mld2_grec,
1016 sizeof(_nsrcs), &_nsrcs);
1020 if (!pskb_may_pull(skb,
1021 len + sizeof(*grec) +
1022 sizeof(struct in6_addr) * ntohs(*nsrcs)))
1025 grec = (struct mld2_grec *)(skb->data + len);
1026 len += sizeof(*grec) +
1027 sizeof(struct in6_addr) * ntohs(*nsrcs);
1029 /* We treat these as MLDv1 reports for now. */
1030 switch (grec->grec_type) {
1031 case MLD2_MODE_IS_INCLUDE:
1032 case MLD2_MODE_IS_EXCLUDE:
1033 case MLD2_CHANGE_TO_INCLUDE:
1034 case MLD2_CHANGE_TO_EXCLUDE:
1035 case MLD2_ALLOW_NEW_SOURCES:
1036 case MLD2_BLOCK_OLD_SOURCES:
1043 err = br_ip6_multicast_add_group(br, port, &grec->grec_mca);
1053 * Add port to rotuer_list
1054 * list is maintained ordered by pointer value
1055 * and locked by br->multicast_lock and RCU
1057 static void br_multicast_add_router(struct net_bridge *br,
1058 struct net_bridge_port *port)
1060 struct net_bridge_port *p;
1061 struct hlist_node *n, *slot = NULL;
1063 hlist_for_each_entry(p, n, &br->router_list, rlist) {
1064 if ((unsigned long) port >= (unsigned long) p)
1070 hlist_add_after_rcu(slot, &port->rlist);
1072 hlist_add_head_rcu(&port->rlist, &br->router_list);
1075 static void br_multicast_mark_router(struct net_bridge *br,
1076 struct net_bridge_port *port)
1078 unsigned long now = jiffies;
1081 if (br->multicast_router == 1)
1082 mod_timer(&br->multicast_router_timer,
1083 now + br->multicast_querier_interval);
1087 if (port->multicast_router != 1)
1090 if (!hlist_unhashed(&port->rlist))
1093 br_multicast_add_router(br, port);
1096 mod_timer(&port->multicast_router_timer,
1097 now + br->multicast_querier_interval);
1100 static void br_multicast_query_received(struct net_bridge *br,
1101 struct net_bridge_port *port,
1105 mod_timer(&br->multicast_querier_timer,
1106 jiffies + br->multicast_querier_interval);
1107 else if (timer_pending(&br->multicast_querier_timer))
1110 br_multicast_mark_router(br, port);
1113 static int br_ip4_multicast_query(struct net_bridge *br,
1114 struct net_bridge_port *port,
1115 struct sk_buff *skb)
1117 struct iphdr *iph = ip_hdr(skb);
1118 struct igmphdr *ih = igmp_hdr(skb);
1119 struct net_bridge_mdb_entry *mp;
1120 struct igmpv3_query *ih3;
1121 struct net_bridge_port_group *p;
1122 struct net_bridge_port_group __rcu **pp;
1123 unsigned long max_delay;
1124 unsigned long now = jiffies;
1128 spin_lock(&br->multicast_lock);
1129 if (!netif_running(br->dev) ||
1130 (port && port->state == BR_STATE_DISABLED))
1133 br_multicast_query_received(br, port, !!iph->saddr);
1137 if (skb->len == sizeof(*ih)) {
1138 max_delay = ih->code * (HZ / IGMP_TIMER_SCALE);
1141 max_delay = 10 * HZ;
1145 if (!pskb_may_pull(skb, sizeof(struct igmpv3_query))) {
1150 ih3 = igmpv3_query_hdr(skb);
1154 max_delay = ih3->code ?
1155 IGMPV3_MRC(ih3->code) * (HZ / IGMP_TIMER_SCALE) : 1;
1161 mp = br_mdb_ip4_get(mlock_dereference(br->mdb, br), group);
1165 max_delay *= br->multicast_last_member_count;
1168 (timer_pending(&mp->timer) ?
1169 time_after(mp->timer.expires, now + max_delay) :
1170 try_to_del_timer_sync(&mp->timer) >= 0))
1171 mod_timer(&mp->timer, now + max_delay);
1173 for (pp = &mp->ports;
1174 (p = mlock_dereference(*pp, br)) != NULL;
1176 if (timer_pending(&p->timer) ?
1177 time_after(p->timer.expires, now + max_delay) :
1178 try_to_del_timer_sync(&p->timer) >= 0)
1179 mod_timer(&p->timer, now + max_delay);
1183 spin_unlock(&br->multicast_lock);
1187 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
1188 static int br_ip6_multicast_query(struct net_bridge *br,
1189 struct net_bridge_port *port,
1190 struct sk_buff *skb)
1192 struct ipv6hdr *ip6h = ipv6_hdr(skb);
1193 struct mld_msg *mld = (struct mld_msg *) icmp6_hdr(skb);
1194 struct net_bridge_mdb_entry *mp;
1195 struct mld2_query *mld2q;
1196 struct net_bridge_port_group *p;
1197 struct net_bridge_port_group __rcu **pp;
1198 unsigned long max_delay;
1199 unsigned long now = jiffies;
1200 struct in6_addr *group = NULL;
1203 spin_lock(&br->multicast_lock);
1204 if (!netif_running(br->dev) ||
1205 (port && port->state == BR_STATE_DISABLED))
1208 br_multicast_query_received(br, port, !ipv6_addr_any(&ip6h->saddr));
1210 if (skb->len == sizeof(*mld)) {
1211 if (!pskb_may_pull(skb, sizeof(*mld))) {
1215 mld = (struct mld_msg *) icmp6_hdr(skb);
1216 max_delay = msecs_to_jiffies(htons(mld->mld_maxdelay));
1218 group = &mld->mld_mca;
1219 } else if (skb->len >= sizeof(*mld2q)) {
1220 if (!pskb_may_pull(skb, sizeof(*mld2q))) {
1224 mld2q = (struct mld2_query *)icmp6_hdr(skb);
1225 if (!mld2q->mld2q_nsrcs)
1226 group = &mld2q->mld2q_mca;
1227 max_delay = mld2q->mld2q_mrc ? MLDV2_MRC(mld2q->mld2q_mrc) : 1;
1233 mp = br_mdb_ip6_get(mlock_dereference(br->mdb, br), group);
1237 max_delay *= br->multicast_last_member_count;
1239 (timer_pending(&mp->timer) ?
1240 time_after(mp->timer.expires, now + max_delay) :
1241 try_to_del_timer_sync(&mp->timer) >= 0))
1242 mod_timer(&mp->timer, now + max_delay);
1244 for (pp = &mp->ports;
1245 (p = mlock_dereference(*pp, br)) != NULL;
1247 if (timer_pending(&p->timer) ?
1248 time_after(p->timer.expires, now + max_delay) :
1249 try_to_del_timer_sync(&p->timer) >= 0)
1250 mod_timer(&p->timer, now + max_delay);
1254 spin_unlock(&br->multicast_lock);
1259 static void br_multicast_leave_group(struct net_bridge *br,
1260 struct net_bridge_port *port,
1261 struct br_ip *group)
1263 struct net_bridge_mdb_htable *mdb;
1264 struct net_bridge_mdb_entry *mp;
1265 struct net_bridge_port_group *p;
1269 spin_lock(&br->multicast_lock);
1270 if (!netif_running(br->dev) ||
1271 (port && port->state == BR_STATE_DISABLED) ||
1272 timer_pending(&br->multicast_querier_timer))
1275 mdb = mlock_dereference(br->mdb, br);
1276 mp = br_mdb_ip_get(mdb, group);
1281 time = now + br->multicast_last_member_count *
1282 br->multicast_last_member_interval;
1286 (timer_pending(&mp->timer) ?
1287 time_after(mp->timer.expires, time) :
1288 try_to_del_timer_sync(&mp->timer) >= 0)) {
1289 mod_timer(&mp->timer, time);
1291 mp->queries_sent = 0;
1292 mod_timer(&mp->query_timer, now);
1298 for (p = mlock_dereference(mp->ports, br);
1300 p = mlock_dereference(p->next, br)) {
1301 if (p->port != port)
1304 if (!hlist_unhashed(&p->mglist) &&
1305 (timer_pending(&p->timer) ?
1306 time_after(p->timer.expires, time) :
1307 try_to_del_timer_sync(&p->timer) >= 0)) {
1308 mod_timer(&p->timer, time);
1310 p->queries_sent = 0;
1311 mod_timer(&p->query_timer, now);
1318 spin_unlock(&br->multicast_lock);
1321 static void br_ip4_multicast_leave_group(struct net_bridge *br,
1322 struct net_bridge_port *port,
1325 struct br_ip br_group;
1327 if (ipv4_is_local_multicast(group))
1330 br_group.u.ip4 = group;
1331 br_group.proto = htons(ETH_P_IP);
1333 br_multicast_leave_group(br, port, &br_group);
1336 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
1337 static void br_ip6_multicast_leave_group(struct net_bridge *br,
1338 struct net_bridge_port *port,
1339 const struct in6_addr *group)
1341 struct br_ip br_group;
1343 if (!ipv6_is_transient_multicast(group))
1346 ipv6_addr_copy(&br_group.u.ip6, group);
1347 br_group.proto = htons(ETH_P_IPV6);
1349 br_multicast_leave_group(br, port, &br_group);
1353 static int br_multicast_ipv4_rcv(struct net_bridge *br,
1354 struct net_bridge_port *port,
1355 struct sk_buff *skb)
1357 struct sk_buff *skb2 = skb;
1364 /* We treat OOM as packet loss for now. */
1365 if (!pskb_may_pull(skb, sizeof(*iph)))
1370 if (iph->ihl < 5 || iph->version != 4)
1373 if (!pskb_may_pull(skb, ip_hdrlen(skb)))
1378 if (unlikely(ip_fast_csum((u8 *)iph, iph->ihl)))
1381 if (iph->protocol != IPPROTO_IGMP)
1384 len = ntohs(iph->tot_len);
1385 if (skb->len < len || len < ip_hdrlen(skb))
1388 if (skb->len > len) {
1389 skb2 = skb_clone(skb, GFP_ATOMIC);
1393 err = pskb_trim_rcsum(skb2, len);
1398 len -= ip_hdrlen(skb2);
1399 offset = skb_network_offset(skb2) + ip_hdrlen(skb2);
1400 __skb_pull(skb2, offset);
1401 skb_reset_transport_header(skb2);
1404 if (!pskb_may_pull(skb2, sizeof(*ih)))
1407 switch (skb2->ip_summed) {
1408 case CHECKSUM_COMPLETE:
1409 if (!csum_fold(skb2->csum))
1414 if (skb_checksum_complete(skb2))
1420 BR_INPUT_SKB_CB(skb)->igmp = 1;
1421 ih = igmp_hdr(skb2);
1424 case IGMP_HOST_MEMBERSHIP_REPORT:
1425 case IGMPV2_HOST_MEMBERSHIP_REPORT:
1426 BR_INPUT_SKB_CB(skb2)->mrouters_only = 1;
1427 err = br_ip4_multicast_add_group(br, port, ih->group);
1429 case IGMPV3_HOST_MEMBERSHIP_REPORT:
1430 err = br_ip4_multicast_igmp3_report(br, port, skb2);
1432 case IGMP_HOST_MEMBERSHIP_QUERY:
1433 err = br_ip4_multicast_query(br, port, skb2);
1435 case IGMP_HOST_LEAVE_MESSAGE:
1436 br_ip4_multicast_leave_group(br, port, ih->group);
1441 __skb_push(skb2, offset);
1448 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
1449 static int br_multicast_ipv6_rcv(struct net_bridge *br,
1450 struct net_bridge_port *port,
1451 struct sk_buff *skb)
1453 struct sk_buff *skb2;
1454 struct ipv6hdr *ip6h;
1455 struct icmp6hdr *icmp6h;
1461 if (!pskb_may_pull(skb, sizeof(*ip6h)))
1464 ip6h = ipv6_hdr(skb);
1467 * We're interested in MLD messages only.
1469 * - MLD has always Router Alert hop-by-hop option
1470 * - But we do not support jumbrograms.
1472 if (ip6h->version != 6 ||
1473 ip6h->nexthdr != IPPROTO_HOPOPTS ||
1474 ip6h->payload_len == 0)
1477 len = ntohs(ip6h->payload_len);
1481 nexthdr = ip6h->nexthdr;
1482 offset = ipv6_skip_exthdr(skb, sizeof(*ip6h), &nexthdr);
1484 if (offset < 0 || nexthdr != IPPROTO_ICMPV6)
1487 /* Okay, we found ICMPv6 header */
1488 skb2 = skb_clone(skb, GFP_ATOMIC);
1493 if (!pskb_may_pull(skb2, offset + sizeof(struct icmp6hdr)))
1496 len -= offset - skb_network_offset(skb2);
1498 __skb_pull(skb2, offset);
1499 skb_reset_transport_header(skb2);
1501 icmp6h = icmp6_hdr(skb2);
1503 switch (icmp6h->icmp6_type) {
1504 case ICMPV6_MGM_QUERY:
1505 case ICMPV6_MGM_REPORT:
1506 case ICMPV6_MGM_REDUCTION:
1507 case ICMPV6_MLD2_REPORT:
1514 /* Okay, we found MLD message. Check further. */
1515 if (skb2->len > len) {
1516 err = pskb_trim_rcsum(skb2, len);
1521 switch (skb2->ip_summed) {
1522 case CHECKSUM_COMPLETE:
1523 if (!csum_fold(skb2->csum))
1528 if (skb_checksum_complete(skb2))
1534 BR_INPUT_SKB_CB(skb)->igmp = 1;
1536 switch (icmp6h->icmp6_type) {
1537 case ICMPV6_MGM_REPORT:
1539 struct mld_msg *mld;
1540 if (!pskb_may_pull(skb2, sizeof(*mld))) {
1544 mld = (struct mld_msg *)skb_transport_header(skb2);
1545 BR_INPUT_SKB_CB(skb2)->mrouters_only = 1;
1546 err = br_ip6_multicast_add_group(br, port, &mld->mld_mca);
1549 case ICMPV6_MLD2_REPORT:
1550 err = br_ip6_multicast_mld2_report(br, port, skb2);
1552 case ICMPV6_MGM_QUERY:
1553 err = br_ip6_multicast_query(br, port, skb2);
1555 case ICMPV6_MGM_REDUCTION:
1557 struct mld_msg *mld;
1558 if (!pskb_may_pull(skb2, sizeof(*mld))) {
1562 mld = (struct mld_msg *)skb_transport_header(skb2);
1563 br_ip6_multicast_leave_group(br, port, &mld->mld_mca);
1573 int br_multicast_rcv(struct net_bridge *br, struct net_bridge_port *port,
1574 struct sk_buff *skb)
1576 BR_INPUT_SKB_CB(skb)->igmp = 0;
1577 BR_INPUT_SKB_CB(skb)->mrouters_only = 0;
1579 if (br->multicast_disabled)
1582 switch (skb->protocol) {
1583 case htons(ETH_P_IP):
1584 return br_multicast_ipv4_rcv(br, port, skb);
1585 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
1586 case htons(ETH_P_IPV6):
1587 return br_multicast_ipv6_rcv(br, port, skb);
1594 static void br_multicast_query_expired(unsigned long data)
1596 struct net_bridge *br = (void *)data;
1598 spin_lock(&br->multicast_lock);
1599 if (br->multicast_startup_queries_sent <
1600 br->multicast_startup_query_count)
1601 br->multicast_startup_queries_sent++;
1603 br_multicast_send_query(br, NULL, br->multicast_startup_queries_sent);
1605 spin_unlock(&br->multicast_lock);
1608 void br_multicast_init(struct net_bridge *br)
1610 br->hash_elasticity = 4;
1613 br->multicast_router = 1;
1614 br->multicast_last_member_count = 2;
1615 br->multicast_startup_query_count = 2;
1617 br->multicast_last_member_interval = HZ;
1618 br->multicast_query_response_interval = 10 * HZ;
1619 br->multicast_startup_query_interval = 125 * HZ / 4;
1620 br->multicast_query_interval = 125 * HZ;
1621 br->multicast_querier_interval = 255 * HZ;
1622 br->multicast_membership_interval = 260 * HZ;
1624 spin_lock_init(&br->multicast_lock);
1625 setup_timer(&br->multicast_router_timer,
1626 br_multicast_local_router_expired, 0);
1627 setup_timer(&br->multicast_querier_timer,
1628 br_multicast_local_router_expired, 0);
1629 setup_timer(&br->multicast_query_timer, br_multicast_query_expired,
1633 void br_multicast_open(struct net_bridge *br)
1635 br->multicast_startup_queries_sent = 0;
1637 if (br->multicast_disabled)
1640 mod_timer(&br->multicast_query_timer, jiffies);
1643 void br_multicast_stop(struct net_bridge *br)
1645 struct net_bridge_mdb_htable *mdb;
1646 struct net_bridge_mdb_entry *mp;
1647 struct hlist_node *p, *n;
1651 del_timer_sync(&br->multicast_router_timer);
1652 del_timer_sync(&br->multicast_querier_timer);
1653 del_timer_sync(&br->multicast_query_timer);
1655 spin_lock_bh(&br->multicast_lock);
1656 mdb = mlock_dereference(br->mdb, br);
1663 for (i = 0; i < mdb->max; i++) {
1664 hlist_for_each_entry_safe(mp, p, n, &mdb->mhash[i],
1666 del_timer(&mp->timer);
1667 del_timer(&mp->query_timer);
1668 call_rcu_bh(&mp->rcu, br_multicast_free_group);
1673 spin_unlock_bh(&br->multicast_lock);
1675 spin_lock_bh(&br->multicast_lock);
1680 call_rcu_bh(&mdb->rcu, br_mdb_free);
1683 spin_unlock_bh(&br->multicast_lock);
1686 int br_multicast_set_router(struct net_bridge *br, unsigned long val)
1690 spin_lock_bh(&br->multicast_lock);
1691 if (!netif_running(br->dev))
1697 del_timer(&br->multicast_router_timer);
1700 br->multicast_router = val;
1710 spin_unlock_bh(&br->multicast_lock);
1715 int br_multicast_set_port_router(struct net_bridge_port *p, unsigned long val)
1717 struct net_bridge *br = p->br;
1720 spin_lock(&br->multicast_lock);
1721 if (!netif_running(br->dev) || p->state == BR_STATE_DISABLED)
1728 p->multicast_router = val;
1731 if (val < 2 && !hlist_unhashed(&p->rlist))
1732 hlist_del_init_rcu(&p->rlist);
1737 del_timer(&p->multicast_router_timer);
1742 br_multicast_add_router(br, p);
1751 spin_unlock(&br->multicast_lock);
1756 int br_multicast_toggle(struct net_bridge *br, unsigned long val)
1758 struct net_bridge_port *port;
1760 struct net_bridge_mdb_htable *mdb;
1762 spin_lock(&br->multicast_lock);
1763 if (br->multicast_disabled == !val)
1766 br->multicast_disabled = !val;
1767 if (br->multicast_disabled)
1770 if (!netif_running(br->dev))
1773 mdb = mlock_dereference(br->mdb, br);
1778 br->multicast_disabled = !!val;
1782 err = br_mdb_rehash(&br->mdb, mdb->max,
1783 br->hash_elasticity);
1788 br_multicast_open(br);
1789 list_for_each_entry(port, &br->port_list, list) {
1790 if (port->state == BR_STATE_DISABLED ||
1791 port->state == BR_STATE_BLOCKING)
1794 __br_multicast_enable_port(port);
1798 spin_unlock(&br->multicast_lock);
1803 int br_multicast_set_hash_max(struct net_bridge *br, unsigned long val)
1807 struct net_bridge_mdb_htable *mdb;
1809 spin_lock(&br->multicast_lock);
1810 if (!netif_running(br->dev))
1814 if (!is_power_of_2(val))
1817 mdb = mlock_dereference(br->mdb, br);
1818 if (mdb && val < mdb->size)
1834 err = br_mdb_rehash(&br->mdb, br->hash_max,
1835 br->hash_elasticity);
1841 spin_unlock(&br->multicast_lock);