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/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>
30 #if IS_ENABLED(CONFIG_IPV6)
33 #include <net/ip6_checksum.h>
34 #include <net/addrconf.h>
37 #include "br_private.h"
39 static void br_multicast_start_querier(struct net_bridge *br,
40 struct bridge_mcast_own_query *query);
41 static void br_multicast_add_router(struct net_bridge *br,
42 struct net_bridge_port *port);
43 static void br_ip4_multicast_leave_group(struct net_bridge *br,
44 struct net_bridge_port *port,
47 #if IS_ENABLED(CONFIG_IPV6)
48 static void br_ip6_multicast_leave_group(struct net_bridge *br,
49 struct net_bridge_port *port,
50 const struct in6_addr *group,
53 unsigned int br_mdb_rehash_seq;
55 static inline int br_ip_equal(const struct br_ip *a, const struct br_ip *b)
57 if (a->proto != b->proto)
63 return a->u.ip4 == b->u.ip4;
64 #if IS_ENABLED(CONFIG_IPV6)
65 case htons(ETH_P_IPV6):
66 return ipv6_addr_equal(&a->u.ip6, &b->u.ip6);
72 static inline int __br_ip4_hash(struct net_bridge_mdb_htable *mdb, __be32 ip,
75 return jhash_2words((__force u32)ip, vid, mdb->secret) & (mdb->max - 1);
78 #if IS_ENABLED(CONFIG_IPV6)
79 static inline int __br_ip6_hash(struct net_bridge_mdb_htable *mdb,
80 const struct in6_addr *ip,
83 return jhash_2words(ipv6_addr_hash(ip), vid,
84 mdb->secret) & (mdb->max - 1);
88 static inline int br_ip_hash(struct net_bridge_mdb_htable *mdb,
93 return __br_ip4_hash(mdb, ip->u.ip4, ip->vid);
94 #if IS_ENABLED(CONFIG_IPV6)
95 case htons(ETH_P_IPV6):
96 return __br_ip6_hash(mdb, &ip->u.ip6, ip->vid);
102 static struct net_bridge_mdb_entry *__br_mdb_ip_get(
103 struct net_bridge_mdb_htable *mdb, struct br_ip *dst, int hash)
105 struct net_bridge_mdb_entry *mp;
107 hlist_for_each_entry_rcu(mp, &mdb->mhash[hash], hlist[mdb->ver]) {
108 if (br_ip_equal(&mp->addr, dst))
115 struct net_bridge_mdb_entry *br_mdb_ip_get(struct net_bridge_mdb_htable *mdb,
121 return __br_mdb_ip_get(mdb, dst, br_ip_hash(mdb, dst));
124 static struct net_bridge_mdb_entry *br_mdb_ip4_get(
125 struct net_bridge_mdb_htable *mdb, __be32 dst, __u16 vid)
130 br_dst.proto = htons(ETH_P_IP);
133 return br_mdb_ip_get(mdb, &br_dst);
136 #if IS_ENABLED(CONFIG_IPV6)
137 static struct net_bridge_mdb_entry *br_mdb_ip6_get(
138 struct net_bridge_mdb_htable *mdb, const struct in6_addr *dst,
144 br_dst.proto = htons(ETH_P_IPV6);
147 return br_mdb_ip_get(mdb, &br_dst);
151 struct net_bridge_mdb_entry *br_mdb_get(struct net_bridge *br,
152 struct sk_buff *skb, u16 vid)
154 struct net_bridge_mdb_htable *mdb = rcu_dereference(br->mdb);
157 if (br->multicast_disabled)
160 if (BR_INPUT_SKB_CB(skb)->igmp)
163 ip.proto = skb->protocol;
166 switch (skb->protocol) {
167 case htons(ETH_P_IP):
168 ip.u.ip4 = ip_hdr(skb)->daddr;
170 #if IS_ENABLED(CONFIG_IPV6)
171 case htons(ETH_P_IPV6):
172 ip.u.ip6 = ipv6_hdr(skb)->daddr;
179 return br_mdb_ip_get(mdb, &ip);
182 static void br_mdb_free(struct rcu_head *head)
184 struct net_bridge_mdb_htable *mdb =
185 container_of(head, struct net_bridge_mdb_htable, rcu);
186 struct net_bridge_mdb_htable *old = mdb->old;
193 static int br_mdb_copy(struct net_bridge_mdb_htable *new,
194 struct net_bridge_mdb_htable *old,
197 struct net_bridge_mdb_entry *mp;
202 for (i = 0; i < old->max; i++)
203 hlist_for_each_entry(mp, &old->mhash[i], hlist[old->ver])
204 hlist_add_head(&mp->hlist[new->ver],
205 &new->mhash[br_ip_hash(new, &mp->addr)]);
211 for (i = 0; i < new->max; i++) {
213 hlist_for_each_entry(mp, &new->mhash[i], hlist[new->ver])
219 return maxlen > elasticity ? -EINVAL : 0;
222 void br_multicast_free_pg(struct rcu_head *head)
224 struct net_bridge_port_group *p =
225 container_of(head, struct net_bridge_port_group, rcu);
230 static void br_multicast_free_group(struct rcu_head *head)
232 struct net_bridge_mdb_entry *mp =
233 container_of(head, struct net_bridge_mdb_entry, rcu);
238 static void br_multicast_group_expired(unsigned long data)
240 struct net_bridge_mdb_entry *mp = (void *)data;
241 struct net_bridge *br = mp->br;
242 struct net_bridge_mdb_htable *mdb;
244 spin_lock(&br->multicast_lock);
245 if (!netif_running(br->dev) || timer_pending(&mp->timer))
253 mdb = mlock_dereference(br->mdb, br);
255 hlist_del_rcu(&mp->hlist[mdb->ver]);
258 call_rcu_bh(&mp->rcu, br_multicast_free_group);
261 spin_unlock(&br->multicast_lock);
264 static void br_multicast_del_pg(struct net_bridge *br,
265 struct net_bridge_port_group *pg)
267 struct net_bridge_mdb_htable *mdb;
268 struct net_bridge_mdb_entry *mp;
269 struct net_bridge_port_group *p;
270 struct net_bridge_port_group __rcu **pp;
272 mdb = mlock_dereference(br->mdb, br);
274 mp = br_mdb_ip_get(mdb, &pg->addr);
278 for (pp = &mp->ports;
279 (p = mlock_dereference(*pp, br)) != NULL;
284 rcu_assign_pointer(*pp, p->next);
285 hlist_del_init(&p->mglist);
286 del_timer(&p->timer);
287 br_mdb_notify(br->dev, p->port, &pg->addr, RTM_DELMDB,
289 call_rcu_bh(&p->rcu, br_multicast_free_pg);
291 if (!mp->ports && !mp->mglist &&
292 netif_running(br->dev))
293 mod_timer(&mp->timer, jiffies);
301 static void br_multicast_port_group_expired(unsigned long data)
303 struct net_bridge_port_group *pg = (void *)data;
304 struct net_bridge *br = pg->port->br;
306 spin_lock(&br->multicast_lock);
307 if (!netif_running(br->dev) || timer_pending(&pg->timer) ||
308 hlist_unhashed(&pg->mglist) || pg->flags & MDB_PG_FLAGS_PERMANENT)
311 br_multicast_del_pg(br, pg);
314 spin_unlock(&br->multicast_lock);
317 static int br_mdb_rehash(struct net_bridge_mdb_htable __rcu **mdbp, int max,
320 struct net_bridge_mdb_htable *old = rcu_dereference_protected(*mdbp, 1);
321 struct net_bridge_mdb_htable *mdb;
324 mdb = kmalloc(sizeof(*mdb), GFP_ATOMIC);
331 mdb->mhash = kzalloc(max * sizeof(*mdb->mhash), GFP_ATOMIC);
337 mdb->size = old ? old->size : 0;
338 mdb->ver = old ? old->ver ^ 1 : 0;
340 if (!old || elasticity)
341 get_random_bytes(&mdb->secret, sizeof(mdb->secret));
343 mdb->secret = old->secret;
348 err = br_mdb_copy(mdb, old, elasticity);
356 call_rcu_bh(&mdb->rcu, br_mdb_free);
359 rcu_assign_pointer(*mdbp, mdb);
364 static struct sk_buff *br_ip4_multicast_alloc_query(struct net_bridge *br,
368 struct igmpv3_query *ihv3;
369 size_t igmp_hdr_size;
375 igmp_hdr_size = sizeof(*ih);
376 if (br->multicast_igmp_version == 3)
377 igmp_hdr_size = sizeof(*ihv3);
378 skb = netdev_alloc_skb_ip_align(br->dev, sizeof(*eth) + sizeof(*iph) +
383 skb->protocol = htons(ETH_P_IP);
385 skb_reset_mac_header(skb);
388 ether_addr_copy(eth->h_source, br->dev->dev_addr);
391 eth->h_dest[2] = 0x5e;
395 eth->h_proto = htons(ETH_P_IP);
396 skb_put(skb, sizeof(*eth));
398 skb_set_network_header(skb, skb->len);
404 iph->tot_len = htons(sizeof(*iph) + igmp_hdr_size + 4);
406 iph->frag_off = htons(IP_DF);
408 iph->protocol = IPPROTO_IGMP;
409 iph->saddr = br->multicast_query_use_ifaddr ?
410 inet_select_addr(br->dev, 0, RT_SCOPE_LINK) : 0;
411 iph->daddr = htonl(INADDR_ALLHOSTS_GROUP);
412 ((u8 *)&iph[1])[0] = IPOPT_RA;
413 ((u8 *)&iph[1])[1] = 4;
414 ((u8 *)&iph[1])[2] = 0;
415 ((u8 *)&iph[1])[3] = 0;
419 skb_set_transport_header(skb, skb->len);
420 *igmp_type = IGMP_HOST_MEMBERSHIP_QUERY;
422 switch (br->multicast_igmp_version) {
425 ih->type = IGMP_HOST_MEMBERSHIP_QUERY;
426 ih->code = (group ? br->multicast_last_member_interval :
427 br->multicast_query_response_interval) /
428 (HZ / IGMP_TIMER_SCALE);
431 ih->csum = ip_compute_csum((void *)ih, sizeof(*ih));
434 ihv3 = igmpv3_query_hdr(skb);
435 ihv3->type = IGMP_HOST_MEMBERSHIP_QUERY;
436 ihv3->code = (group ? br->multicast_last_member_interval :
437 br->multicast_query_response_interval) /
438 (HZ / IGMP_TIMER_SCALE);
440 ihv3->qqic = br->multicast_query_interval / HZ;
446 ihv3->csum = ip_compute_csum((void *)ihv3, sizeof(*ihv3));
450 skb_put(skb, igmp_hdr_size);
451 __skb_pull(skb, sizeof(*eth));
457 #if IS_ENABLED(CONFIG_IPV6)
458 static struct sk_buff *br_ip6_multicast_alloc_query(struct net_bridge *br,
459 const struct in6_addr *grp,
462 struct mld2_query *mld2q;
463 unsigned long interval;
464 struct ipv6hdr *ip6h;
465 struct mld_msg *mldq;
471 mld_hdr_size = sizeof(*mldq);
472 if (br->multicast_mld_version == 2)
473 mld_hdr_size = sizeof(*mld2q);
474 skb = netdev_alloc_skb_ip_align(br->dev, sizeof(*eth) + sizeof(*ip6h) +
479 skb->protocol = htons(ETH_P_IPV6);
481 /* Ethernet header */
482 skb_reset_mac_header(skb);
485 ether_addr_copy(eth->h_source, br->dev->dev_addr);
486 eth->h_proto = htons(ETH_P_IPV6);
487 skb_put(skb, sizeof(*eth));
489 /* IPv6 header + HbH option */
490 skb_set_network_header(skb, skb->len);
491 ip6h = ipv6_hdr(skb);
493 *(__force __be32 *)ip6h = htonl(0x60000000);
494 ip6h->payload_len = htons(8 + mld_hdr_size);
495 ip6h->nexthdr = IPPROTO_HOPOPTS;
497 ipv6_addr_set(&ip6h->daddr, htonl(0xff020000), 0, 0, htonl(1));
498 if (ipv6_dev_get_saddr(dev_net(br->dev), br->dev, &ip6h->daddr, 0,
501 br->has_ipv6_addr = 0;
505 br->has_ipv6_addr = 1;
506 ipv6_eth_mc_map(&ip6h->daddr, eth->h_dest);
508 hopopt = (u8 *)(ip6h + 1);
509 hopopt[0] = IPPROTO_ICMPV6; /* next hdr */
510 hopopt[1] = 0; /* length of HbH */
511 hopopt[2] = IPV6_TLV_ROUTERALERT; /* Router Alert */
512 hopopt[3] = 2; /* Length of RA Option */
513 hopopt[4] = 0; /* Type = 0x0000 (MLD) */
515 hopopt[6] = IPV6_TLV_PAD1; /* Pad1 */
516 hopopt[7] = IPV6_TLV_PAD1; /* Pad1 */
518 skb_put(skb, sizeof(*ip6h) + 8);
521 skb_set_transport_header(skb, skb->len);
522 interval = ipv6_addr_any(grp) ?
523 br->multicast_query_response_interval :
524 br->multicast_last_member_interval;
525 *igmp_type = ICMPV6_MGM_QUERY;
526 switch (br->multicast_mld_version) {
528 mldq = (struct mld_msg *)icmp6_hdr(skb);
529 mldq->mld_type = ICMPV6_MGM_QUERY;
532 mldq->mld_maxdelay = htons((u16)jiffies_to_msecs(interval));
533 mldq->mld_reserved = 0;
534 mldq->mld_mca = *grp;
535 mldq->mld_cksum = csum_ipv6_magic(&ip6h->saddr, &ip6h->daddr,
536 sizeof(*mldq), IPPROTO_ICMPV6,
542 mld2q = (struct mld2_query *)icmp6_hdr(skb);
543 mld2q->mld2q_mrc = ntohs((u16)jiffies_to_msecs(interval));
544 mld2q->mld2q_type = ICMPV6_MGM_QUERY;
545 mld2q->mld2q_code = 0;
546 mld2q->mld2q_cksum = 0;
547 mld2q->mld2q_resv1 = 0;
548 mld2q->mld2q_resv2 = 0;
549 mld2q->mld2q_suppress = 0;
550 mld2q->mld2q_qrv = 2;
551 mld2q->mld2q_nsrcs = 0;
552 mld2q->mld2q_qqic = br->multicast_query_interval / HZ;
553 mld2q->mld2q_mca = *grp;
554 mld2q->mld2q_cksum = csum_ipv6_magic(&ip6h->saddr, &ip6h->daddr,
562 skb_put(skb, mld_hdr_size);
564 __skb_pull(skb, sizeof(*eth));
571 static struct sk_buff *br_multicast_alloc_query(struct net_bridge *br,
575 switch (addr->proto) {
576 case htons(ETH_P_IP):
577 return br_ip4_multicast_alloc_query(br, addr->u.ip4, igmp_type);
578 #if IS_ENABLED(CONFIG_IPV6)
579 case htons(ETH_P_IPV6):
580 return br_ip6_multicast_alloc_query(br, &addr->u.ip6,
587 static struct net_bridge_mdb_entry *br_multicast_get_group(
588 struct net_bridge *br, struct net_bridge_port *port,
589 struct br_ip *group, int hash)
591 struct net_bridge_mdb_htable *mdb;
592 struct net_bridge_mdb_entry *mp;
593 unsigned int count = 0;
598 mdb = rcu_dereference_protected(br->mdb, 1);
599 hlist_for_each_entry(mp, &mdb->mhash[hash], hlist[mdb->ver]) {
601 if (unlikely(br_ip_equal(group, &mp->addr)))
608 if (unlikely(count > br->hash_elasticity && count)) {
610 br_info(br, "Multicast hash table "
611 "chain limit reached: %s\n",
612 port ? port->dev->name : br->dev->name);
614 elasticity = br->hash_elasticity;
617 if (mdb->size >= max) {
619 if (unlikely(max > br->hash_max)) {
620 br_warn(br, "Multicast hash table maximum of %d "
621 "reached, disabling snooping: %s\n",
623 port ? port->dev->name : br->dev->name);
626 br->multicast_disabled = 1;
631 if (max > mdb->max || elasticity) {
634 br_info(br, "Multicast hash table "
636 port ? port->dev->name : br->dev->name);
641 err = br_mdb_rehash(&br->mdb, max, elasticity);
643 br_warn(br, "Cannot rehash multicast "
644 "hash table, disabling snooping: %s, %d, %d\n",
645 port ? port->dev->name : br->dev->name,
661 struct net_bridge_mdb_entry *br_multicast_new_group(struct net_bridge *br,
662 struct net_bridge_port *p,
665 struct net_bridge_mdb_htable *mdb;
666 struct net_bridge_mdb_entry *mp;
670 mdb = rcu_dereference_protected(br->mdb, 1);
672 err = br_mdb_rehash(&br->mdb, BR_HASH_SIZE, 0);
678 hash = br_ip_hash(mdb, group);
679 mp = br_multicast_get_group(br, p, group, hash);
680 switch (PTR_ERR(mp)) {
686 mdb = rcu_dereference_protected(br->mdb, 1);
687 hash = br_ip_hash(mdb, group);
694 mp = kzalloc(sizeof(*mp), GFP_ATOMIC);
696 return ERR_PTR(-ENOMEM);
700 setup_timer(&mp->timer, br_multicast_group_expired,
703 hlist_add_head_rcu(&mp->hlist[mdb->ver], &mdb->mhash[hash]);
710 struct net_bridge_port_group *br_multicast_new_port_group(
711 struct net_bridge_port *port,
713 struct net_bridge_port_group __rcu *next,
716 struct net_bridge_port_group *p;
718 p = kzalloc(sizeof(*p), GFP_ATOMIC);
725 rcu_assign_pointer(p->next, next);
726 hlist_add_head(&p->mglist, &port->mglist);
727 setup_timer(&p->timer, br_multicast_port_group_expired,
732 static int br_multicast_add_group(struct net_bridge *br,
733 struct net_bridge_port *port,
736 struct net_bridge_port_group __rcu **pp;
737 struct net_bridge_port_group *p;
738 struct net_bridge_mdb_entry *mp;
739 unsigned long now = jiffies;
742 spin_lock(&br->multicast_lock);
743 if (!netif_running(br->dev) ||
744 (port && port->state == BR_STATE_DISABLED))
747 mp = br_multicast_new_group(br, port, group);
754 mod_timer(&mp->timer, now + br->multicast_membership_interval);
758 for (pp = &mp->ports;
759 (p = mlock_dereference(*pp, br)) != NULL;
763 if ((unsigned long)p->port < (unsigned long)port)
767 p = br_multicast_new_port_group(port, group, *pp, 0);
770 rcu_assign_pointer(*pp, p);
771 br_mdb_notify(br->dev, port, group, RTM_NEWMDB, 0);
774 mod_timer(&p->timer, now + br->multicast_membership_interval);
779 spin_unlock(&br->multicast_lock);
783 static int br_ip4_multicast_add_group(struct net_bridge *br,
784 struct net_bridge_port *port,
788 struct br_ip br_group;
790 if (ipv4_is_local_multicast(group))
793 br_group.u.ip4 = group;
794 br_group.proto = htons(ETH_P_IP);
797 return br_multicast_add_group(br, port, &br_group);
800 #if IS_ENABLED(CONFIG_IPV6)
801 static int br_ip6_multicast_add_group(struct net_bridge *br,
802 struct net_bridge_port *port,
803 const struct in6_addr *group,
806 struct br_ip br_group;
808 if (ipv6_addr_is_ll_all_nodes(group))
811 br_group.u.ip6 = *group;
812 br_group.proto = htons(ETH_P_IPV6);
815 return br_multicast_add_group(br, port, &br_group);
819 static void br_multicast_router_expired(unsigned long data)
821 struct net_bridge_port *port = (void *)data;
822 struct net_bridge *br = port->br;
824 spin_lock(&br->multicast_lock);
825 if (port->multicast_router == MDB_RTR_TYPE_DISABLED ||
826 port->multicast_router == MDB_RTR_TYPE_PERM ||
827 timer_pending(&port->multicast_router_timer) ||
828 hlist_unhashed(&port->rlist))
831 hlist_del_init_rcu(&port->rlist);
832 br_rtr_notify(br->dev, port, RTM_DELMDB);
833 /* Don't allow timer refresh if the router expired */
834 if (port->multicast_router == MDB_RTR_TYPE_TEMP)
835 port->multicast_router = MDB_RTR_TYPE_TEMP_QUERY;
838 spin_unlock(&br->multicast_lock);
841 static void br_multicast_local_router_expired(unsigned long data)
845 static void br_multicast_querier_expired(struct net_bridge *br,
846 struct bridge_mcast_own_query *query)
848 spin_lock(&br->multicast_lock);
849 if (!netif_running(br->dev) || br->multicast_disabled)
852 br_multicast_start_querier(br, query);
855 spin_unlock(&br->multicast_lock);
858 static void br_ip4_multicast_querier_expired(unsigned long data)
860 struct net_bridge *br = (void *)data;
862 br_multicast_querier_expired(br, &br->ip4_own_query);
865 #if IS_ENABLED(CONFIG_IPV6)
866 static void br_ip6_multicast_querier_expired(unsigned long data)
868 struct net_bridge *br = (void *)data;
870 br_multicast_querier_expired(br, &br->ip6_own_query);
874 static void br_multicast_select_own_querier(struct net_bridge *br,
878 if (ip->proto == htons(ETH_P_IP))
879 br->ip4_querier.addr.u.ip4 = ip_hdr(skb)->saddr;
880 #if IS_ENABLED(CONFIG_IPV6)
882 br->ip6_querier.addr.u.ip6 = ipv6_hdr(skb)->saddr;
886 static void __br_multicast_send_query(struct net_bridge *br,
887 struct net_bridge_port *port,
893 skb = br_multicast_alloc_query(br, ip, &igmp_type);
898 skb->dev = port->dev;
899 br_multicast_count(br, port, skb, igmp_type,
901 NF_HOOK(NFPROTO_BRIDGE, NF_BR_LOCAL_OUT,
902 dev_net(port->dev), NULL, skb, NULL, skb->dev,
903 br_dev_queue_push_xmit);
905 br_multicast_select_own_querier(br, ip, skb);
906 br_multicast_count(br, port, skb, igmp_type,
912 static void br_multicast_send_query(struct net_bridge *br,
913 struct net_bridge_port *port,
914 struct bridge_mcast_own_query *own_query)
916 struct bridge_mcast_other_query *other_query = NULL;
917 struct br_ip br_group;
920 if (!netif_running(br->dev) || br->multicast_disabled ||
921 !br->multicast_querier)
924 memset(&br_group.u, 0, sizeof(br_group.u));
926 if (port ? (own_query == &port->ip4_own_query) :
927 (own_query == &br->ip4_own_query)) {
928 other_query = &br->ip4_other_query;
929 br_group.proto = htons(ETH_P_IP);
930 #if IS_ENABLED(CONFIG_IPV6)
932 other_query = &br->ip6_other_query;
933 br_group.proto = htons(ETH_P_IPV6);
937 if (!other_query || timer_pending(&other_query->timer))
940 __br_multicast_send_query(br, port, &br_group);
943 time += own_query->startup_sent < br->multicast_startup_query_count ?
944 br->multicast_startup_query_interval :
945 br->multicast_query_interval;
946 mod_timer(&own_query->timer, time);
950 br_multicast_port_query_expired(struct net_bridge_port *port,
951 struct bridge_mcast_own_query *query)
953 struct net_bridge *br = port->br;
955 spin_lock(&br->multicast_lock);
956 if (port->state == BR_STATE_DISABLED ||
957 port->state == BR_STATE_BLOCKING)
960 if (query->startup_sent < br->multicast_startup_query_count)
961 query->startup_sent++;
963 br_multicast_send_query(port->br, port, query);
966 spin_unlock(&br->multicast_lock);
969 static void br_ip4_multicast_port_query_expired(unsigned long data)
971 struct net_bridge_port *port = (void *)data;
973 br_multicast_port_query_expired(port, &port->ip4_own_query);
976 #if IS_ENABLED(CONFIG_IPV6)
977 static void br_ip6_multicast_port_query_expired(unsigned long data)
979 struct net_bridge_port *port = (void *)data;
981 br_multicast_port_query_expired(port, &port->ip6_own_query);
985 int br_multicast_add_port(struct net_bridge_port *port)
987 port->multicast_router = MDB_RTR_TYPE_TEMP_QUERY;
989 setup_timer(&port->multicast_router_timer, br_multicast_router_expired,
990 (unsigned long)port);
991 setup_timer(&port->ip4_own_query.timer,
992 br_ip4_multicast_port_query_expired, (unsigned long)port);
993 #if IS_ENABLED(CONFIG_IPV6)
994 setup_timer(&port->ip6_own_query.timer,
995 br_ip6_multicast_port_query_expired, (unsigned long)port);
997 port->mcast_stats = netdev_alloc_pcpu_stats(struct bridge_mcast_stats);
998 if (!port->mcast_stats)
1004 void br_multicast_del_port(struct net_bridge_port *port)
1006 struct net_bridge *br = port->br;
1007 struct net_bridge_port_group *pg;
1008 struct hlist_node *n;
1010 /* Take care of the remaining groups, only perm ones should be left */
1011 spin_lock_bh(&br->multicast_lock);
1012 hlist_for_each_entry_safe(pg, n, &port->mglist, mglist)
1013 br_multicast_del_pg(br, pg);
1014 spin_unlock_bh(&br->multicast_lock);
1015 del_timer_sync(&port->multicast_router_timer);
1016 free_percpu(port->mcast_stats);
1019 static void br_multicast_enable(struct bridge_mcast_own_query *query)
1021 query->startup_sent = 0;
1023 if (try_to_del_timer_sync(&query->timer) >= 0 ||
1024 del_timer(&query->timer))
1025 mod_timer(&query->timer, jiffies);
1028 static void __br_multicast_enable_port(struct net_bridge_port *port)
1030 struct net_bridge *br = port->br;
1032 if (br->multicast_disabled || !netif_running(br->dev))
1035 br_multicast_enable(&port->ip4_own_query);
1036 #if IS_ENABLED(CONFIG_IPV6)
1037 br_multicast_enable(&port->ip6_own_query);
1039 if (port->multicast_router == MDB_RTR_TYPE_PERM &&
1040 hlist_unhashed(&port->rlist))
1041 br_multicast_add_router(br, port);
1044 void br_multicast_enable_port(struct net_bridge_port *port)
1046 struct net_bridge *br = port->br;
1048 spin_lock(&br->multicast_lock);
1049 __br_multicast_enable_port(port);
1050 spin_unlock(&br->multicast_lock);
1053 void br_multicast_disable_port(struct net_bridge_port *port)
1055 struct net_bridge *br = port->br;
1056 struct net_bridge_port_group *pg;
1057 struct hlist_node *n;
1059 spin_lock(&br->multicast_lock);
1060 hlist_for_each_entry_safe(pg, n, &port->mglist, mglist)
1061 if (!(pg->flags & MDB_PG_FLAGS_PERMANENT))
1062 br_multicast_del_pg(br, pg);
1064 if (!hlist_unhashed(&port->rlist)) {
1065 hlist_del_init_rcu(&port->rlist);
1066 br_rtr_notify(br->dev, port, RTM_DELMDB);
1067 /* Don't allow timer refresh if disabling */
1068 if (port->multicast_router == MDB_RTR_TYPE_TEMP)
1069 port->multicast_router = MDB_RTR_TYPE_TEMP_QUERY;
1071 del_timer(&port->multicast_router_timer);
1072 del_timer(&port->ip4_own_query.timer);
1073 #if IS_ENABLED(CONFIG_IPV6)
1074 del_timer(&port->ip6_own_query.timer);
1076 spin_unlock(&br->multicast_lock);
1079 static int br_ip4_multicast_igmp3_report(struct net_bridge *br,
1080 struct net_bridge_port *port,
1081 struct sk_buff *skb,
1084 struct igmpv3_report *ih;
1085 struct igmpv3_grec *grec;
1093 ih = igmpv3_report_hdr(skb);
1094 num = ntohs(ih->ngrec);
1095 len = skb_transport_offset(skb) + sizeof(*ih);
1097 for (i = 0; i < num; i++) {
1098 len += sizeof(*grec);
1099 if (!pskb_may_pull(skb, len))
1102 grec = (void *)(skb->data + len - sizeof(*grec));
1103 group = grec->grec_mca;
1104 type = grec->grec_type;
1106 len += ntohs(grec->grec_nsrcs) * 4;
1107 if (!pskb_may_pull(skb, len))
1110 /* We treat this as an IGMPv2 report for now. */
1112 case IGMPV3_MODE_IS_INCLUDE:
1113 case IGMPV3_MODE_IS_EXCLUDE:
1114 case IGMPV3_CHANGE_TO_INCLUDE:
1115 case IGMPV3_CHANGE_TO_EXCLUDE:
1116 case IGMPV3_ALLOW_NEW_SOURCES:
1117 case IGMPV3_BLOCK_OLD_SOURCES:
1124 if ((type == IGMPV3_CHANGE_TO_INCLUDE ||
1125 type == IGMPV3_MODE_IS_INCLUDE) &&
1126 ntohs(grec->grec_nsrcs) == 0) {
1127 br_ip4_multicast_leave_group(br, port, group, vid);
1129 err = br_ip4_multicast_add_group(br, port, group, vid);
1138 #if IS_ENABLED(CONFIG_IPV6)
1139 static int br_ip6_multicast_mld2_report(struct net_bridge *br,
1140 struct net_bridge_port *port,
1141 struct sk_buff *skb,
1144 struct icmp6hdr *icmp6h;
1145 struct mld2_grec *grec;
1151 if (!pskb_may_pull(skb, sizeof(*icmp6h)))
1154 icmp6h = icmp6_hdr(skb);
1155 num = ntohs(icmp6h->icmp6_dataun.un_data16[1]);
1156 len = skb_transport_offset(skb) + sizeof(*icmp6h);
1158 for (i = 0; i < num; i++) {
1159 __be16 *nsrcs, _nsrcs;
1161 nsrcs = skb_header_pointer(skb,
1162 len + offsetof(struct mld2_grec,
1164 sizeof(_nsrcs), &_nsrcs);
1168 if (!pskb_may_pull(skb,
1169 len + sizeof(*grec) +
1170 sizeof(struct in6_addr) * ntohs(*nsrcs)))
1173 grec = (struct mld2_grec *)(skb->data + len);
1174 len += sizeof(*grec) +
1175 sizeof(struct in6_addr) * ntohs(*nsrcs);
1177 /* We treat these as MLDv1 reports for now. */
1178 switch (grec->grec_type) {
1179 case MLD2_MODE_IS_INCLUDE:
1180 case MLD2_MODE_IS_EXCLUDE:
1181 case MLD2_CHANGE_TO_INCLUDE:
1182 case MLD2_CHANGE_TO_EXCLUDE:
1183 case MLD2_ALLOW_NEW_SOURCES:
1184 case MLD2_BLOCK_OLD_SOURCES:
1191 if ((grec->grec_type == MLD2_CHANGE_TO_INCLUDE ||
1192 grec->grec_type == MLD2_MODE_IS_INCLUDE) &&
1193 ntohs(*nsrcs) == 0) {
1194 br_ip6_multicast_leave_group(br, port, &grec->grec_mca,
1197 err = br_ip6_multicast_add_group(br, port,
1198 &grec->grec_mca, vid);
1208 static bool br_ip4_multicast_select_querier(struct net_bridge *br,
1209 struct net_bridge_port *port,
1212 if (!timer_pending(&br->ip4_own_query.timer) &&
1213 !timer_pending(&br->ip4_other_query.timer))
1216 if (!br->ip4_querier.addr.u.ip4)
1219 if (ntohl(saddr) <= ntohl(br->ip4_querier.addr.u.ip4))
1225 br->ip4_querier.addr.u.ip4 = saddr;
1227 /* update protected by general multicast_lock by caller */
1228 rcu_assign_pointer(br->ip4_querier.port, port);
1233 #if IS_ENABLED(CONFIG_IPV6)
1234 static bool br_ip6_multicast_select_querier(struct net_bridge *br,
1235 struct net_bridge_port *port,
1236 struct in6_addr *saddr)
1238 if (!timer_pending(&br->ip6_own_query.timer) &&
1239 !timer_pending(&br->ip6_other_query.timer))
1242 if (ipv6_addr_cmp(saddr, &br->ip6_querier.addr.u.ip6) <= 0)
1248 br->ip6_querier.addr.u.ip6 = *saddr;
1250 /* update protected by general multicast_lock by caller */
1251 rcu_assign_pointer(br->ip6_querier.port, port);
1257 static bool br_multicast_select_querier(struct net_bridge *br,
1258 struct net_bridge_port *port,
1259 struct br_ip *saddr)
1261 switch (saddr->proto) {
1262 case htons(ETH_P_IP):
1263 return br_ip4_multicast_select_querier(br, port, saddr->u.ip4);
1264 #if IS_ENABLED(CONFIG_IPV6)
1265 case htons(ETH_P_IPV6):
1266 return br_ip6_multicast_select_querier(br, port, &saddr->u.ip6);
1274 br_multicast_update_query_timer(struct net_bridge *br,
1275 struct bridge_mcast_other_query *query,
1276 unsigned long max_delay)
1278 if (!timer_pending(&query->timer))
1279 query->delay_time = jiffies + max_delay;
1281 mod_timer(&query->timer, jiffies + br->multicast_querier_interval);
1285 * Add port to router_list
1286 * list is maintained ordered by pointer value
1287 * and locked by br->multicast_lock and RCU
1289 static void br_multicast_add_router(struct net_bridge *br,
1290 struct net_bridge_port *port)
1292 struct net_bridge_port *p;
1293 struct hlist_node *slot = NULL;
1295 if (!hlist_unhashed(&port->rlist))
1298 hlist_for_each_entry(p, &br->router_list, rlist) {
1299 if ((unsigned long) port >= (unsigned long) p)
1305 hlist_add_behind_rcu(&port->rlist, slot);
1307 hlist_add_head_rcu(&port->rlist, &br->router_list);
1308 br_rtr_notify(br->dev, port, RTM_NEWMDB);
1311 static void br_multicast_mark_router(struct net_bridge *br,
1312 struct net_bridge_port *port)
1314 unsigned long now = jiffies;
1317 if (br->multicast_router == MDB_RTR_TYPE_TEMP_QUERY)
1318 mod_timer(&br->multicast_router_timer,
1319 now + br->multicast_querier_interval);
1323 if (port->multicast_router == MDB_RTR_TYPE_DISABLED ||
1324 port->multicast_router == MDB_RTR_TYPE_PERM)
1327 br_multicast_add_router(br, port);
1329 mod_timer(&port->multicast_router_timer,
1330 now + br->multicast_querier_interval);
1333 static void br_multicast_query_received(struct net_bridge *br,
1334 struct net_bridge_port *port,
1335 struct bridge_mcast_other_query *query,
1336 struct br_ip *saddr,
1337 unsigned long max_delay)
1339 if (!br_multicast_select_querier(br, port, saddr))
1342 br_multicast_update_query_timer(br, query, max_delay);
1343 br_multicast_mark_router(br, port);
1346 static int br_ip4_multicast_query(struct net_bridge *br,
1347 struct net_bridge_port *port,
1348 struct sk_buff *skb,
1351 const struct iphdr *iph = ip_hdr(skb);
1352 struct igmphdr *ih = igmp_hdr(skb);
1353 struct net_bridge_mdb_entry *mp;
1354 struct igmpv3_query *ih3;
1355 struct net_bridge_port_group *p;
1356 struct net_bridge_port_group __rcu **pp;
1358 unsigned long max_delay;
1359 unsigned long now = jiffies;
1360 unsigned int offset = skb_transport_offset(skb);
1364 spin_lock(&br->multicast_lock);
1365 if (!netif_running(br->dev) ||
1366 (port && port->state == BR_STATE_DISABLED))
1371 if (skb->len == offset + sizeof(*ih)) {
1372 max_delay = ih->code * (HZ / IGMP_TIMER_SCALE);
1375 max_delay = 10 * HZ;
1378 } else if (skb->len >= offset + sizeof(*ih3)) {
1379 ih3 = igmpv3_query_hdr(skb);
1383 max_delay = ih3->code ?
1384 IGMPV3_MRC(ih3->code) * (HZ / IGMP_TIMER_SCALE) : 1;
1390 saddr.proto = htons(ETH_P_IP);
1391 saddr.u.ip4 = iph->saddr;
1393 br_multicast_query_received(br, port, &br->ip4_other_query,
1398 mp = br_mdb_ip4_get(mlock_dereference(br->mdb, br), group, vid);
1402 max_delay *= br->multicast_last_member_count;
1405 (timer_pending(&mp->timer) ?
1406 time_after(mp->timer.expires, now + max_delay) :
1407 try_to_del_timer_sync(&mp->timer) >= 0))
1408 mod_timer(&mp->timer, now + max_delay);
1410 for (pp = &mp->ports;
1411 (p = mlock_dereference(*pp, br)) != NULL;
1413 if (timer_pending(&p->timer) ?
1414 time_after(p->timer.expires, now + max_delay) :
1415 try_to_del_timer_sync(&p->timer) >= 0)
1416 mod_timer(&p->timer, now + max_delay);
1420 spin_unlock(&br->multicast_lock);
1424 #if IS_ENABLED(CONFIG_IPV6)
1425 static int br_ip6_multicast_query(struct net_bridge *br,
1426 struct net_bridge_port *port,
1427 struct sk_buff *skb,
1430 const struct ipv6hdr *ip6h = ipv6_hdr(skb);
1431 struct mld_msg *mld;
1432 struct net_bridge_mdb_entry *mp;
1433 struct mld2_query *mld2q;
1434 struct net_bridge_port_group *p;
1435 struct net_bridge_port_group __rcu **pp;
1437 unsigned long max_delay;
1438 unsigned long now = jiffies;
1439 unsigned int offset = skb_transport_offset(skb);
1440 const struct in6_addr *group = NULL;
1441 bool is_general_query;
1444 spin_lock(&br->multicast_lock);
1445 if (!netif_running(br->dev) ||
1446 (port && port->state == BR_STATE_DISABLED))
1449 if (skb->len == offset + sizeof(*mld)) {
1450 if (!pskb_may_pull(skb, offset + sizeof(*mld))) {
1454 mld = (struct mld_msg *) icmp6_hdr(skb);
1455 max_delay = msecs_to_jiffies(ntohs(mld->mld_maxdelay));
1457 group = &mld->mld_mca;
1459 if (!pskb_may_pull(skb, offset + sizeof(*mld2q))) {
1463 mld2q = (struct mld2_query *)icmp6_hdr(skb);
1464 if (!mld2q->mld2q_nsrcs)
1465 group = &mld2q->mld2q_mca;
1467 max_delay = max(msecs_to_jiffies(mldv2_mrc(mld2q)), 1UL);
1470 is_general_query = group && ipv6_addr_any(group);
1472 if (is_general_query) {
1473 saddr.proto = htons(ETH_P_IPV6);
1474 saddr.u.ip6 = ip6h->saddr;
1476 br_multicast_query_received(br, port, &br->ip6_other_query,
1479 } else if (!group) {
1483 mp = br_mdb_ip6_get(mlock_dereference(br->mdb, br), group, vid);
1487 max_delay *= br->multicast_last_member_count;
1489 (timer_pending(&mp->timer) ?
1490 time_after(mp->timer.expires, now + max_delay) :
1491 try_to_del_timer_sync(&mp->timer) >= 0))
1492 mod_timer(&mp->timer, now + max_delay);
1494 for (pp = &mp->ports;
1495 (p = mlock_dereference(*pp, br)) != NULL;
1497 if (timer_pending(&p->timer) ?
1498 time_after(p->timer.expires, now + max_delay) :
1499 try_to_del_timer_sync(&p->timer) >= 0)
1500 mod_timer(&p->timer, now + max_delay);
1504 spin_unlock(&br->multicast_lock);
1510 br_multicast_leave_group(struct net_bridge *br,
1511 struct net_bridge_port *port,
1512 struct br_ip *group,
1513 struct bridge_mcast_other_query *other_query,
1514 struct bridge_mcast_own_query *own_query)
1516 struct net_bridge_mdb_htable *mdb;
1517 struct net_bridge_mdb_entry *mp;
1518 struct net_bridge_port_group *p;
1522 spin_lock(&br->multicast_lock);
1523 if (!netif_running(br->dev) ||
1524 (port && port->state == BR_STATE_DISABLED))
1527 mdb = mlock_dereference(br->mdb, br);
1528 mp = br_mdb_ip_get(mdb, group);
1532 if (port && (port->flags & BR_MULTICAST_FAST_LEAVE)) {
1533 struct net_bridge_port_group __rcu **pp;
1535 for (pp = &mp->ports;
1536 (p = mlock_dereference(*pp, br)) != NULL;
1538 if (p->port != port)
1541 rcu_assign_pointer(*pp, p->next);
1542 hlist_del_init(&p->mglist);
1543 del_timer(&p->timer);
1544 call_rcu_bh(&p->rcu, br_multicast_free_pg);
1545 br_mdb_notify(br->dev, port, group, RTM_DELMDB,
1548 if (!mp->ports && !mp->mglist &&
1549 netif_running(br->dev))
1550 mod_timer(&mp->timer, jiffies);
1555 if (timer_pending(&other_query->timer))
1558 if (br->multicast_querier) {
1559 __br_multicast_send_query(br, port, &mp->addr);
1561 time = jiffies + br->multicast_last_member_count *
1562 br->multicast_last_member_interval;
1564 mod_timer(&own_query->timer, time);
1566 for (p = mlock_dereference(mp->ports, br);
1568 p = mlock_dereference(p->next, br)) {
1569 if (p->port != port)
1572 if (!hlist_unhashed(&p->mglist) &&
1573 (timer_pending(&p->timer) ?
1574 time_after(p->timer.expires, time) :
1575 try_to_del_timer_sync(&p->timer) >= 0)) {
1576 mod_timer(&p->timer, time);
1584 time = now + br->multicast_last_member_count *
1585 br->multicast_last_member_interval;
1589 (timer_pending(&mp->timer) ?
1590 time_after(mp->timer.expires, time) :
1591 try_to_del_timer_sync(&mp->timer) >= 0)) {
1592 mod_timer(&mp->timer, time);
1598 for (p = mlock_dereference(mp->ports, br);
1600 p = mlock_dereference(p->next, br)) {
1601 if (p->port != port)
1604 if (!hlist_unhashed(&p->mglist) &&
1605 (timer_pending(&p->timer) ?
1606 time_after(p->timer.expires, time) :
1607 try_to_del_timer_sync(&p->timer) >= 0)) {
1608 mod_timer(&p->timer, time);
1614 spin_unlock(&br->multicast_lock);
1617 static void br_ip4_multicast_leave_group(struct net_bridge *br,
1618 struct net_bridge_port *port,
1622 struct br_ip br_group;
1623 struct bridge_mcast_own_query *own_query;
1625 if (ipv4_is_local_multicast(group))
1628 own_query = port ? &port->ip4_own_query : &br->ip4_own_query;
1630 br_group.u.ip4 = group;
1631 br_group.proto = htons(ETH_P_IP);
1634 br_multicast_leave_group(br, port, &br_group, &br->ip4_other_query,
1638 #if IS_ENABLED(CONFIG_IPV6)
1639 static void br_ip6_multicast_leave_group(struct net_bridge *br,
1640 struct net_bridge_port *port,
1641 const struct in6_addr *group,
1644 struct br_ip br_group;
1645 struct bridge_mcast_own_query *own_query;
1647 if (ipv6_addr_is_ll_all_nodes(group))
1650 own_query = port ? &port->ip6_own_query : &br->ip6_own_query;
1652 br_group.u.ip6 = *group;
1653 br_group.proto = htons(ETH_P_IPV6);
1656 br_multicast_leave_group(br, port, &br_group, &br->ip6_other_query,
1661 static void br_multicast_err_count(const struct net_bridge *br,
1662 const struct net_bridge_port *p,
1665 struct bridge_mcast_stats __percpu *stats;
1666 struct bridge_mcast_stats *pstats;
1668 if (!br->multicast_stats_enabled)
1672 stats = p->mcast_stats;
1674 stats = br->mcast_stats;
1675 if (WARN_ON(!stats))
1678 pstats = this_cpu_ptr(stats);
1680 u64_stats_update_begin(&pstats->syncp);
1682 case htons(ETH_P_IP):
1683 pstats->mstats.igmp_parse_errors++;
1685 #if IS_ENABLED(CONFIG_IPV6)
1686 case htons(ETH_P_IPV6):
1687 pstats->mstats.mld_parse_errors++;
1691 u64_stats_update_end(&pstats->syncp);
1694 static void br_multicast_pim(struct net_bridge *br,
1695 struct net_bridge_port *port,
1696 const struct sk_buff *skb)
1698 unsigned int offset = skb_transport_offset(skb);
1699 struct pimhdr *pimhdr, _pimhdr;
1701 pimhdr = skb_header_pointer(skb, offset, sizeof(_pimhdr), &_pimhdr);
1702 if (!pimhdr || pim_hdr_version(pimhdr) != PIM_VERSION ||
1703 pim_hdr_type(pimhdr) != PIM_TYPE_HELLO)
1706 br_multicast_mark_router(br, port);
1709 static int br_multicast_ipv4_rcv(struct net_bridge *br,
1710 struct net_bridge_port *port,
1711 struct sk_buff *skb,
1714 struct sk_buff *skb_trimmed = NULL;
1718 err = ip_mc_check_igmp(skb, &skb_trimmed);
1720 if (err == -ENOMSG) {
1721 if (!ipv4_is_local_multicast(ip_hdr(skb)->daddr)) {
1722 BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1723 } else if (pim_ipv4_all_pim_routers(ip_hdr(skb)->daddr)) {
1724 if (ip_hdr(skb)->protocol == IPPROTO_PIM)
1725 br_multicast_pim(br, port, skb);
1728 } else if (err < 0) {
1729 br_multicast_err_count(br, port, skb->protocol);
1734 BR_INPUT_SKB_CB(skb)->igmp = ih->type;
1737 case IGMP_HOST_MEMBERSHIP_REPORT:
1738 case IGMPV2_HOST_MEMBERSHIP_REPORT:
1739 BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1740 err = br_ip4_multicast_add_group(br, port, ih->group, vid);
1742 case IGMPV3_HOST_MEMBERSHIP_REPORT:
1743 err = br_ip4_multicast_igmp3_report(br, port, skb_trimmed, vid);
1745 case IGMP_HOST_MEMBERSHIP_QUERY:
1746 err = br_ip4_multicast_query(br, port, skb_trimmed, vid);
1748 case IGMP_HOST_LEAVE_MESSAGE:
1749 br_ip4_multicast_leave_group(br, port, ih->group, vid);
1753 if (skb_trimmed && skb_trimmed != skb)
1754 kfree_skb(skb_trimmed);
1756 br_multicast_count(br, port, skb, BR_INPUT_SKB_CB(skb)->igmp,
1762 #if IS_ENABLED(CONFIG_IPV6)
1763 static int br_multicast_ipv6_rcv(struct net_bridge *br,
1764 struct net_bridge_port *port,
1765 struct sk_buff *skb,
1768 struct sk_buff *skb_trimmed = NULL;
1769 struct mld_msg *mld;
1772 err = ipv6_mc_check_mld(skb, &skb_trimmed);
1774 if (err == -ENOMSG) {
1775 if (!ipv6_addr_is_ll_all_nodes(&ipv6_hdr(skb)->daddr))
1776 BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1778 } else if (err < 0) {
1779 br_multicast_err_count(br, port, skb->protocol);
1783 mld = (struct mld_msg *)skb_transport_header(skb);
1784 BR_INPUT_SKB_CB(skb)->igmp = mld->mld_type;
1786 switch (mld->mld_type) {
1787 case ICMPV6_MGM_REPORT:
1788 BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1789 err = br_ip6_multicast_add_group(br, port, &mld->mld_mca, vid);
1791 case ICMPV6_MLD2_REPORT:
1792 err = br_ip6_multicast_mld2_report(br, port, skb_trimmed, vid);
1794 case ICMPV6_MGM_QUERY:
1795 err = br_ip6_multicast_query(br, port, skb_trimmed, vid);
1797 case ICMPV6_MGM_REDUCTION:
1798 br_ip6_multicast_leave_group(br, port, &mld->mld_mca, vid);
1802 if (skb_trimmed && skb_trimmed != skb)
1803 kfree_skb(skb_trimmed);
1805 br_multicast_count(br, port, skb, BR_INPUT_SKB_CB(skb)->igmp,
1812 int br_multicast_rcv(struct net_bridge *br, struct net_bridge_port *port,
1813 struct sk_buff *skb, u16 vid)
1817 BR_INPUT_SKB_CB(skb)->igmp = 0;
1818 BR_INPUT_SKB_CB(skb)->mrouters_only = 0;
1820 if (br->multicast_disabled)
1823 switch (skb->protocol) {
1824 case htons(ETH_P_IP):
1825 ret = br_multicast_ipv4_rcv(br, port, skb, vid);
1827 #if IS_ENABLED(CONFIG_IPV6)
1828 case htons(ETH_P_IPV6):
1829 ret = br_multicast_ipv6_rcv(br, port, skb, vid);
1837 static void br_multicast_query_expired(struct net_bridge *br,
1838 struct bridge_mcast_own_query *query,
1839 struct bridge_mcast_querier *querier)
1841 spin_lock(&br->multicast_lock);
1842 if (query->startup_sent < br->multicast_startup_query_count)
1843 query->startup_sent++;
1845 RCU_INIT_POINTER(querier->port, NULL);
1846 br_multicast_send_query(br, NULL, query);
1847 spin_unlock(&br->multicast_lock);
1850 static void br_ip4_multicast_query_expired(unsigned long data)
1852 struct net_bridge *br = (void *)data;
1854 br_multicast_query_expired(br, &br->ip4_own_query, &br->ip4_querier);
1857 #if IS_ENABLED(CONFIG_IPV6)
1858 static void br_ip6_multicast_query_expired(unsigned long data)
1860 struct net_bridge *br = (void *)data;
1862 br_multicast_query_expired(br, &br->ip6_own_query, &br->ip6_querier);
1866 void br_multicast_init(struct net_bridge *br)
1868 br->hash_elasticity = 4;
1871 br->multicast_router = MDB_RTR_TYPE_TEMP_QUERY;
1872 br->multicast_querier = 0;
1873 br->multicast_query_use_ifaddr = 0;
1874 br->multicast_last_member_count = 2;
1875 br->multicast_startup_query_count = 2;
1877 br->multicast_last_member_interval = HZ;
1878 br->multicast_query_response_interval = 10 * HZ;
1879 br->multicast_startup_query_interval = 125 * HZ / 4;
1880 br->multicast_query_interval = 125 * HZ;
1881 br->multicast_querier_interval = 255 * HZ;
1882 br->multicast_membership_interval = 260 * HZ;
1884 br->ip4_other_query.delay_time = 0;
1885 br->ip4_querier.port = NULL;
1886 br->multicast_igmp_version = 2;
1887 #if IS_ENABLED(CONFIG_IPV6)
1888 br->multicast_mld_version = 1;
1889 br->ip6_other_query.delay_time = 0;
1890 br->ip6_querier.port = NULL;
1892 br->has_ipv6_addr = 1;
1894 spin_lock_init(&br->multicast_lock);
1895 setup_timer(&br->multicast_router_timer,
1896 br_multicast_local_router_expired, 0);
1897 setup_timer(&br->ip4_other_query.timer,
1898 br_ip4_multicast_querier_expired, (unsigned long)br);
1899 setup_timer(&br->ip4_own_query.timer, br_ip4_multicast_query_expired,
1901 #if IS_ENABLED(CONFIG_IPV6)
1902 setup_timer(&br->ip6_other_query.timer,
1903 br_ip6_multicast_querier_expired, (unsigned long)br);
1904 setup_timer(&br->ip6_own_query.timer, br_ip6_multicast_query_expired,
1909 static void __br_multicast_open(struct net_bridge *br,
1910 struct bridge_mcast_own_query *query)
1912 query->startup_sent = 0;
1914 if (br->multicast_disabled)
1917 mod_timer(&query->timer, jiffies);
1920 void br_multicast_open(struct net_bridge *br)
1922 __br_multicast_open(br, &br->ip4_own_query);
1923 #if IS_ENABLED(CONFIG_IPV6)
1924 __br_multicast_open(br, &br->ip6_own_query);
1928 void br_multicast_stop(struct net_bridge *br)
1930 del_timer_sync(&br->multicast_router_timer);
1931 del_timer_sync(&br->ip4_other_query.timer);
1932 del_timer_sync(&br->ip4_own_query.timer);
1933 #if IS_ENABLED(CONFIG_IPV6)
1934 del_timer_sync(&br->ip6_other_query.timer);
1935 del_timer_sync(&br->ip6_own_query.timer);
1939 void br_multicast_dev_del(struct net_bridge *br)
1941 struct net_bridge_mdb_htable *mdb;
1942 struct net_bridge_mdb_entry *mp;
1943 struct hlist_node *n;
1947 spin_lock_bh(&br->multicast_lock);
1948 mdb = mlock_dereference(br->mdb, br);
1955 for (i = 0; i < mdb->max; i++) {
1956 hlist_for_each_entry_safe(mp, n, &mdb->mhash[i],
1958 del_timer(&mp->timer);
1959 call_rcu_bh(&mp->rcu, br_multicast_free_group);
1964 spin_unlock_bh(&br->multicast_lock);
1966 spin_lock_bh(&br->multicast_lock);
1971 call_rcu_bh(&mdb->rcu, br_mdb_free);
1974 spin_unlock_bh(&br->multicast_lock);
1976 free_percpu(br->mcast_stats);
1979 int br_multicast_set_router(struct net_bridge *br, unsigned long val)
1983 spin_lock_bh(&br->multicast_lock);
1986 case MDB_RTR_TYPE_DISABLED:
1987 case MDB_RTR_TYPE_PERM:
1988 del_timer(&br->multicast_router_timer);
1990 case MDB_RTR_TYPE_TEMP_QUERY:
1991 br->multicast_router = val;
1996 spin_unlock_bh(&br->multicast_lock);
2001 static void __del_port_router(struct net_bridge_port *p)
2003 if (hlist_unhashed(&p->rlist))
2005 hlist_del_init_rcu(&p->rlist);
2006 br_rtr_notify(p->br->dev, p, RTM_DELMDB);
2009 int br_multicast_set_port_router(struct net_bridge_port *p, unsigned long val)
2011 struct net_bridge *br = p->br;
2012 unsigned long now = jiffies;
2015 spin_lock(&br->multicast_lock);
2016 if (p->multicast_router == val) {
2017 /* Refresh the temp router port timer */
2018 if (p->multicast_router == MDB_RTR_TYPE_TEMP)
2019 mod_timer(&p->multicast_router_timer,
2020 now + br->multicast_querier_interval);
2025 case MDB_RTR_TYPE_DISABLED:
2026 p->multicast_router = MDB_RTR_TYPE_DISABLED;
2027 __del_port_router(p);
2028 del_timer(&p->multicast_router_timer);
2030 case MDB_RTR_TYPE_TEMP_QUERY:
2031 p->multicast_router = MDB_RTR_TYPE_TEMP_QUERY;
2032 __del_port_router(p);
2034 case MDB_RTR_TYPE_PERM:
2035 p->multicast_router = MDB_RTR_TYPE_PERM;
2036 del_timer(&p->multicast_router_timer);
2037 br_multicast_add_router(br, p);
2039 case MDB_RTR_TYPE_TEMP:
2040 p->multicast_router = MDB_RTR_TYPE_TEMP;
2041 br_multicast_mark_router(br, p);
2048 spin_unlock(&br->multicast_lock);
2053 static void br_multicast_start_querier(struct net_bridge *br,
2054 struct bridge_mcast_own_query *query)
2056 struct net_bridge_port *port;
2058 __br_multicast_open(br, query);
2060 list_for_each_entry(port, &br->port_list, list) {
2061 if (port->state == BR_STATE_DISABLED ||
2062 port->state == BR_STATE_BLOCKING)
2065 if (query == &br->ip4_own_query)
2066 br_multicast_enable(&port->ip4_own_query);
2067 #if IS_ENABLED(CONFIG_IPV6)
2069 br_multicast_enable(&port->ip6_own_query);
2074 int br_multicast_toggle(struct net_bridge *br, unsigned long val)
2076 struct net_bridge_mdb_htable *mdb;
2077 struct net_bridge_port *port;
2080 spin_lock_bh(&br->multicast_lock);
2081 if (br->multicast_disabled == !val)
2084 br->multicast_disabled = !val;
2085 if (br->multicast_disabled)
2088 if (!netif_running(br->dev))
2091 mdb = mlock_dereference(br->mdb, br);
2096 br->multicast_disabled = !!val;
2100 err = br_mdb_rehash(&br->mdb, mdb->max,
2101 br->hash_elasticity);
2106 br_multicast_open(br);
2107 list_for_each_entry(port, &br->port_list, list)
2108 __br_multicast_enable_port(port);
2111 spin_unlock_bh(&br->multicast_lock);
2116 int br_multicast_set_querier(struct net_bridge *br, unsigned long val)
2118 unsigned long max_delay;
2122 spin_lock_bh(&br->multicast_lock);
2123 if (br->multicast_querier == val)
2126 br->multicast_querier = val;
2130 max_delay = br->multicast_query_response_interval;
2132 if (!timer_pending(&br->ip4_other_query.timer))
2133 br->ip4_other_query.delay_time = jiffies + max_delay;
2135 br_multicast_start_querier(br, &br->ip4_own_query);
2137 #if IS_ENABLED(CONFIG_IPV6)
2138 if (!timer_pending(&br->ip6_other_query.timer))
2139 br->ip6_other_query.delay_time = jiffies + max_delay;
2141 br_multicast_start_querier(br, &br->ip6_own_query);
2145 spin_unlock_bh(&br->multicast_lock);
2150 int br_multicast_set_hash_max(struct net_bridge *br, unsigned long val)
2154 struct net_bridge_mdb_htable *mdb;
2156 spin_lock_bh(&br->multicast_lock);
2157 if (!is_power_of_2(val))
2160 mdb = mlock_dereference(br->mdb, br);
2161 if (mdb && val < mdb->size)
2177 err = br_mdb_rehash(&br->mdb, br->hash_max,
2178 br->hash_elasticity);
2184 spin_unlock_bh(&br->multicast_lock);
2189 int br_multicast_set_igmp_version(struct net_bridge *br, unsigned long val)
2191 /* Currently we support only version 2 and 3 */
2200 spin_lock_bh(&br->multicast_lock);
2201 br->multicast_igmp_version = val;
2202 spin_unlock_bh(&br->multicast_lock);
2207 #if IS_ENABLED(CONFIG_IPV6)
2208 int br_multicast_set_mld_version(struct net_bridge *br, unsigned long val)
2210 /* Currently we support version 1 and 2 */
2219 spin_lock_bh(&br->multicast_lock);
2220 br->multicast_mld_version = val;
2221 spin_unlock_bh(&br->multicast_lock);
2228 * br_multicast_list_adjacent - Returns snooped multicast addresses
2229 * @dev: The bridge port adjacent to which to retrieve addresses
2230 * @br_ip_list: The list to store found, snooped multicast IP addresses in
2232 * Creates a list of IP addresses (struct br_ip_list) sensed by the multicast
2233 * snooping feature on all bridge ports of dev's bridge device, excluding
2234 * the addresses from dev itself.
2236 * Returns the number of items added to br_ip_list.
2239 * - br_ip_list needs to be initialized by caller
2240 * - br_ip_list might contain duplicates in the end
2241 * (needs to be taken care of by caller)
2242 * - br_ip_list needs to be freed by caller
2244 int br_multicast_list_adjacent(struct net_device *dev,
2245 struct list_head *br_ip_list)
2247 struct net_bridge *br;
2248 struct net_bridge_port *port;
2249 struct net_bridge_port_group *group;
2250 struct br_ip_list *entry;
2254 if (!br_ip_list || !br_port_exists(dev))
2257 port = br_port_get_rcu(dev);
2258 if (!port || !port->br)
2263 list_for_each_entry_rcu(port, &br->port_list, list) {
2264 if (!port->dev || port->dev == dev)
2267 hlist_for_each_entry_rcu(group, &port->mglist, mglist) {
2268 entry = kmalloc(sizeof(*entry), GFP_ATOMIC);
2272 entry->addr = group->addr;
2273 list_add(&entry->list, br_ip_list);
2282 EXPORT_SYMBOL_GPL(br_multicast_list_adjacent);
2285 * br_multicast_has_querier_anywhere - Checks for a querier on a bridge
2286 * @dev: The bridge port providing the bridge on which to check for a querier
2287 * @proto: The protocol family to check for: IGMP -> ETH_P_IP, MLD -> ETH_P_IPV6
2289 * Checks whether the given interface has a bridge on top and if so returns
2290 * true if a valid querier exists anywhere on the bridged link layer.
2291 * Otherwise returns false.
2293 bool br_multicast_has_querier_anywhere(struct net_device *dev, int proto)
2295 struct net_bridge *br;
2296 struct net_bridge_port *port;
2301 if (!br_port_exists(dev))
2304 port = br_port_get_rcu(dev);
2305 if (!port || !port->br)
2310 memset(ð, 0, sizeof(eth));
2311 eth.h_proto = htons(proto);
2313 ret = br_multicast_querier_exists(br, ð);
2319 EXPORT_SYMBOL_GPL(br_multicast_has_querier_anywhere);
2322 * br_multicast_has_querier_adjacent - Checks for a querier behind a bridge port
2323 * @dev: The bridge port adjacent to which to check for a querier
2324 * @proto: The protocol family to check for: IGMP -> ETH_P_IP, MLD -> ETH_P_IPV6
2326 * Checks whether the given interface has a bridge on top and if so returns
2327 * true if a selected querier is behind one of the other ports of this
2328 * bridge. Otherwise returns false.
2330 bool br_multicast_has_querier_adjacent(struct net_device *dev, int proto)
2332 struct net_bridge *br;
2333 struct net_bridge_port *port;
2337 if (!br_port_exists(dev))
2340 port = br_port_get_rcu(dev);
2341 if (!port || !port->br)
2348 if (!timer_pending(&br->ip4_other_query.timer) ||
2349 rcu_dereference(br->ip4_querier.port) == port)
2352 #if IS_ENABLED(CONFIG_IPV6)
2354 if (!timer_pending(&br->ip6_other_query.timer) ||
2355 rcu_dereference(br->ip6_querier.port) == port)
2368 EXPORT_SYMBOL_GPL(br_multicast_has_querier_adjacent);
2370 static void br_mcast_stats_add(struct bridge_mcast_stats __percpu *stats,
2371 const struct sk_buff *skb, u8 type, u8 dir)
2373 struct bridge_mcast_stats *pstats = this_cpu_ptr(stats);
2374 __be16 proto = skb->protocol;
2377 u64_stats_update_begin(&pstats->syncp);
2379 case htons(ETH_P_IP):
2380 t_len = ntohs(ip_hdr(skb)->tot_len) - ip_hdrlen(skb);
2382 case IGMP_HOST_MEMBERSHIP_REPORT:
2383 pstats->mstats.igmp_v1reports[dir]++;
2385 case IGMPV2_HOST_MEMBERSHIP_REPORT:
2386 pstats->mstats.igmp_v2reports[dir]++;
2388 case IGMPV3_HOST_MEMBERSHIP_REPORT:
2389 pstats->mstats.igmp_v3reports[dir]++;
2391 case IGMP_HOST_MEMBERSHIP_QUERY:
2392 if (t_len != sizeof(struct igmphdr)) {
2393 pstats->mstats.igmp_v3queries[dir]++;
2395 unsigned int offset = skb_transport_offset(skb);
2396 struct igmphdr *ih, _ihdr;
2398 ih = skb_header_pointer(skb, offset,
2399 sizeof(_ihdr), &_ihdr);
2403 pstats->mstats.igmp_v1queries[dir]++;
2405 pstats->mstats.igmp_v2queries[dir]++;
2408 case IGMP_HOST_LEAVE_MESSAGE:
2409 pstats->mstats.igmp_leaves[dir]++;
2413 #if IS_ENABLED(CONFIG_IPV6)
2414 case htons(ETH_P_IPV6):
2415 t_len = ntohs(ipv6_hdr(skb)->payload_len) +
2416 sizeof(struct ipv6hdr);
2417 t_len -= skb_network_header_len(skb);
2419 case ICMPV6_MGM_REPORT:
2420 pstats->mstats.mld_v1reports[dir]++;
2422 case ICMPV6_MLD2_REPORT:
2423 pstats->mstats.mld_v2reports[dir]++;
2425 case ICMPV6_MGM_QUERY:
2426 if (t_len != sizeof(struct mld_msg))
2427 pstats->mstats.mld_v2queries[dir]++;
2429 pstats->mstats.mld_v1queries[dir]++;
2431 case ICMPV6_MGM_REDUCTION:
2432 pstats->mstats.mld_leaves[dir]++;
2436 #endif /* CONFIG_IPV6 */
2438 u64_stats_update_end(&pstats->syncp);
2441 void br_multicast_count(struct net_bridge *br, const struct net_bridge_port *p,
2442 const struct sk_buff *skb, u8 type, u8 dir)
2444 struct bridge_mcast_stats __percpu *stats;
2446 /* if multicast_disabled is true then igmp type can't be set */
2447 if (!type || !br->multicast_stats_enabled)
2451 stats = p->mcast_stats;
2453 stats = br->mcast_stats;
2454 if (WARN_ON(!stats))
2457 br_mcast_stats_add(stats, skb, type, dir);
2460 int br_multicast_init_stats(struct net_bridge *br)
2462 br->mcast_stats = netdev_alloc_pcpu_stats(struct bridge_mcast_stats);
2463 if (!br->mcast_stats)
2469 static void mcast_stats_add_dir(u64 *dst, u64 *src)
2471 dst[BR_MCAST_DIR_RX] += src[BR_MCAST_DIR_RX];
2472 dst[BR_MCAST_DIR_TX] += src[BR_MCAST_DIR_TX];
2475 void br_multicast_get_stats(const struct net_bridge *br,
2476 const struct net_bridge_port *p,
2477 struct br_mcast_stats *dest)
2479 struct bridge_mcast_stats __percpu *stats;
2480 struct br_mcast_stats tdst;
2483 memset(dest, 0, sizeof(*dest));
2485 stats = p->mcast_stats;
2487 stats = br->mcast_stats;
2488 if (WARN_ON(!stats))
2491 memset(&tdst, 0, sizeof(tdst));
2492 for_each_possible_cpu(i) {
2493 struct bridge_mcast_stats *cpu_stats = per_cpu_ptr(stats, i);
2494 struct br_mcast_stats temp;
2498 start = u64_stats_fetch_begin_irq(&cpu_stats->syncp);
2499 memcpy(&temp, &cpu_stats->mstats, sizeof(temp));
2500 } while (u64_stats_fetch_retry_irq(&cpu_stats->syncp, start));
2502 mcast_stats_add_dir(tdst.igmp_v1queries, temp.igmp_v1queries);
2503 mcast_stats_add_dir(tdst.igmp_v2queries, temp.igmp_v2queries);
2504 mcast_stats_add_dir(tdst.igmp_v3queries, temp.igmp_v3queries);
2505 mcast_stats_add_dir(tdst.igmp_leaves, temp.igmp_leaves);
2506 mcast_stats_add_dir(tdst.igmp_v1reports, temp.igmp_v1reports);
2507 mcast_stats_add_dir(tdst.igmp_v2reports, temp.igmp_v2reports);
2508 mcast_stats_add_dir(tdst.igmp_v3reports, temp.igmp_v3reports);
2509 tdst.igmp_parse_errors += temp.igmp_parse_errors;
2511 mcast_stats_add_dir(tdst.mld_v1queries, temp.mld_v1queries);
2512 mcast_stats_add_dir(tdst.mld_v2queries, temp.mld_v2queries);
2513 mcast_stats_add_dir(tdst.mld_leaves, temp.mld_leaves);
2514 mcast_stats_add_dir(tdst.mld_v1reports, temp.mld_v1reports);
2515 mcast_stats_add_dir(tdst.mld_v2reports, temp.mld_v2reports);
2516 tdst.mld_parse_errors += temp.mld_parse_errors;
2518 memcpy(dest, &tdst, sizeof(*dest));