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>
26 #include <linux/inetdevice.h>
28 #if IS_ENABLED(CONFIG_IPV6)
31 #include <net/ip6_checksum.h>
34 #include "br_private.h"
36 static void br_multicast_start_querier(struct net_bridge *br);
37 unsigned int br_mdb_rehash_seq;
39 static inline int br_ip_equal(const struct br_ip *a, const struct br_ip *b)
41 if (a->proto != b->proto)
47 return a->u.ip4 == b->u.ip4;
48 #if IS_ENABLED(CONFIG_IPV6)
49 case htons(ETH_P_IPV6):
50 return ipv6_addr_equal(&a->u.ip6, &b->u.ip6);
56 static inline int __br_ip4_hash(struct net_bridge_mdb_htable *mdb, __be32 ip,
59 return jhash_2words((__force u32)ip, vid, mdb->secret) & (mdb->max - 1);
62 #if IS_ENABLED(CONFIG_IPV6)
63 static inline int __br_ip6_hash(struct net_bridge_mdb_htable *mdb,
64 const struct in6_addr *ip,
67 return jhash_2words(ipv6_addr_hash(ip), vid,
68 mdb->secret) & (mdb->max - 1);
72 static inline int br_ip_hash(struct net_bridge_mdb_htable *mdb,
77 return __br_ip4_hash(mdb, ip->u.ip4, ip->vid);
78 #if IS_ENABLED(CONFIG_IPV6)
79 case htons(ETH_P_IPV6):
80 return __br_ip6_hash(mdb, &ip->u.ip6, ip->vid);
86 static struct net_bridge_mdb_entry *__br_mdb_ip_get(
87 struct net_bridge_mdb_htable *mdb, struct br_ip *dst, int hash)
89 struct net_bridge_mdb_entry *mp;
91 hlist_for_each_entry_rcu(mp, &mdb->mhash[hash], hlist[mdb->ver]) {
92 if (br_ip_equal(&mp->addr, dst))
99 struct net_bridge_mdb_entry *br_mdb_ip_get(struct net_bridge_mdb_htable *mdb,
105 return __br_mdb_ip_get(mdb, dst, br_ip_hash(mdb, dst));
108 static struct net_bridge_mdb_entry *br_mdb_ip4_get(
109 struct net_bridge_mdb_htable *mdb, __be32 dst, __u16 vid)
114 br_dst.proto = htons(ETH_P_IP);
117 return br_mdb_ip_get(mdb, &br_dst);
120 #if IS_ENABLED(CONFIG_IPV6)
121 static struct net_bridge_mdb_entry *br_mdb_ip6_get(
122 struct net_bridge_mdb_htable *mdb, const struct in6_addr *dst,
128 br_dst.proto = htons(ETH_P_IPV6);
131 return br_mdb_ip_get(mdb, &br_dst);
135 struct net_bridge_mdb_entry *br_mdb_get(struct net_bridge *br,
136 struct sk_buff *skb, u16 vid)
138 struct net_bridge_mdb_htable *mdb = rcu_dereference(br->mdb);
141 if (br->multicast_disabled)
144 if (BR_INPUT_SKB_CB(skb)->igmp)
147 ip.proto = skb->protocol;
150 switch (skb->protocol) {
151 case htons(ETH_P_IP):
152 ip.u.ip4 = ip_hdr(skb)->daddr;
154 #if IS_ENABLED(CONFIG_IPV6)
155 case htons(ETH_P_IPV6):
156 ip.u.ip6 = ipv6_hdr(skb)->daddr;
163 return br_mdb_ip_get(mdb, &ip);
166 static void br_mdb_free(struct rcu_head *head)
168 struct net_bridge_mdb_htable *mdb =
169 container_of(head, struct net_bridge_mdb_htable, rcu);
170 struct net_bridge_mdb_htable *old = mdb->old;
177 static int br_mdb_copy(struct net_bridge_mdb_htable *new,
178 struct net_bridge_mdb_htable *old,
181 struct net_bridge_mdb_entry *mp;
186 for (i = 0; i < old->max; i++)
187 hlist_for_each_entry(mp, &old->mhash[i], hlist[old->ver])
188 hlist_add_head(&mp->hlist[new->ver],
189 &new->mhash[br_ip_hash(new, &mp->addr)]);
195 for (i = 0; i < new->max; i++) {
197 hlist_for_each_entry(mp, &new->mhash[i], hlist[new->ver])
203 return maxlen > elasticity ? -EINVAL : 0;
206 void br_multicast_free_pg(struct rcu_head *head)
208 struct net_bridge_port_group *p =
209 container_of(head, struct net_bridge_port_group, rcu);
214 static void br_multicast_free_group(struct rcu_head *head)
216 struct net_bridge_mdb_entry *mp =
217 container_of(head, struct net_bridge_mdb_entry, rcu);
222 static void br_multicast_group_expired(unsigned long data)
224 struct net_bridge_mdb_entry *mp = (void *)data;
225 struct net_bridge *br = mp->br;
226 struct net_bridge_mdb_htable *mdb;
228 spin_lock(&br->multicast_lock);
229 if (!netif_running(br->dev) || timer_pending(&mp->timer))
237 mdb = mlock_dereference(br->mdb, br);
239 hlist_del_rcu(&mp->hlist[mdb->ver]);
242 call_rcu_bh(&mp->rcu, br_multicast_free_group);
245 spin_unlock(&br->multicast_lock);
248 static void br_multicast_del_pg(struct net_bridge *br,
249 struct net_bridge_port_group *pg)
251 struct net_bridge_mdb_htable *mdb;
252 struct net_bridge_mdb_entry *mp;
253 struct net_bridge_port_group *p;
254 struct net_bridge_port_group __rcu **pp;
256 mdb = mlock_dereference(br->mdb, br);
258 mp = br_mdb_ip_get(mdb, &pg->addr);
262 for (pp = &mp->ports;
263 (p = mlock_dereference(*pp, br)) != NULL;
268 rcu_assign_pointer(*pp, p->next);
269 hlist_del_init(&p->mglist);
270 del_timer(&p->timer);
271 call_rcu_bh(&p->rcu, br_multicast_free_pg);
273 if (!mp->ports && !mp->mglist && mp->timer_armed &&
274 netif_running(br->dev))
275 mod_timer(&mp->timer, jiffies);
283 static void br_multicast_port_group_expired(unsigned long data)
285 struct net_bridge_port_group *pg = (void *)data;
286 struct net_bridge *br = pg->port->br;
288 spin_lock(&br->multicast_lock);
289 if (!netif_running(br->dev) || timer_pending(&pg->timer) ||
290 hlist_unhashed(&pg->mglist) || pg->state & MDB_PERMANENT)
293 br_multicast_del_pg(br, pg);
296 spin_unlock(&br->multicast_lock);
299 static int br_mdb_rehash(struct net_bridge_mdb_htable __rcu **mdbp, int max,
302 struct net_bridge_mdb_htable *old = rcu_dereference_protected(*mdbp, 1);
303 struct net_bridge_mdb_htable *mdb;
306 mdb = kmalloc(sizeof(*mdb), GFP_ATOMIC);
313 mdb->mhash = kzalloc(max * sizeof(*mdb->mhash), GFP_ATOMIC);
319 mdb->size = old ? old->size : 0;
320 mdb->ver = old ? old->ver ^ 1 : 0;
322 if (!old || elasticity)
323 get_random_bytes(&mdb->secret, sizeof(mdb->secret));
325 mdb->secret = old->secret;
330 err = br_mdb_copy(mdb, old, elasticity);
338 call_rcu_bh(&mdb->rcu, br_mdb_free);
341 rcu_assign_pointer(*mdbp, mdb);
346 static struct sk_buff *br_ip4_multicast_alloc_query(struct net_bridge *br,
354 skb = netdev_alloc_skb_ip_align(br->dev, sizeof(*eth) + sizeof(*iph) +
359 skb->protocol = htons(ETH_P_IP);
361 skb_reset_mac_header(skb);
364 memcpy(eth->h_source, br->dev->dev_addr, 6);
367 eth->h_dest[2] = 0x5e;
371 eth->h_proto = htons(ETH_P_IP);
372 skb_put(skb, sizeof(*eth));
374 skb_set_network_header(skb, skb->len);
380 iph->tot_len = htons(sizeof(*iph) + sizeof(*ih) + 4);
382 iph->frag_off = htons(IP_DF);
384 iph->protocol = IPPROTO_IGMP;
385 iph->saddr = br->multicast_query_use_ifaddr ?
386 inet_select_addr(br->dev, 0, RT_SCOPE_LINK) : 0;
387 iph->daddr = htonl(INADDR_ALLHOSTS_GROUP);
388 ((u8 *)&iph[1])[0] = IPOPT_RA;
389 ((u8 *)&iph[1])[1] = 4;
390 ((u8 *)&iph[1])[2] = 0;
391 ((u8 *)&iph[1])[3] = 0;
395 skb_set_transport_header(skb, skb->len);
397 ih->type = IGMP_HOST_MEMBERSHIP_QUERY;
398 ih->code = (group ? br->multicast_last_member_interval :
399 br->multicast_query_response_interval) /
400 (HZ / IGMP_TIMER_SCALE);
403 ih->csum = ip_compute_csum((void *)ih, sizeof(struct igmphdr));
404 skb_put(skb, sizeof(*ih));
406 __skb_pull(skb, sizeof(*eth));
412 #if IS_ENABLED(CONFIG_IPV6)
413 static struct sk_buff *br_ip6_multicast_alloc_query(struct net_bridge *br,
414 const struct in6_addr *group)
417 struct ipv6hdr *ip6h;
418 struct mld_msg *mldq;
421 unsigned long interval;
423 skb = netdev_alloc_skb_ip_align(br->dev, sizeof(*eth) + sizeof(*ip6h) +
428 skb->protocol = htons(ETH_P_IPV6);
430 /* Ethernet header */
431 skb_reset_mac_header(skb);
434 memcpy(eth->h_source, br->dev->dev_addr, 6);
435 eth->h_proto = htons(ETH_P_IPV6);
436 skb_put(skb, sizeof(*eth));
438 /* IPv6 header + HbH option */
439 skb_set_network_header(skb, skb->len);
440 ip6h = ipv6_hdr(skb);
442 *(__force __be32 *)ip6h = htonl(0x60000000);
443 ip6h->payload_len = htons(8 + sizeof(*mldq));
444 ip6h->nexthdr = IPPROTO_HOPOPTS;
446 ipv6_addr_set(&ip6h->daddr, htonl(0xff020000), 0, 0, htonl(1));
447 if (ipv6_dev_get_saddr(dev_net(br->dev), br->dev, &ip6h->daddr, 0,
452 ipv6_eth_mc_map(&ip6h->daddr, eth->h_dest);
454 hopopt = (u8 *)(ip6h + 1);
455 hopopt[0] = IPPROTO_ICMPV6; /* next hdr */
456 hopopt[1] = 0; /* length of HbH */
457 hopopt[2] = IPV6_TLV_ROUTERALERT; /* Router Alert */
458 hopopt[3] = 2; /* Length of RA Option */
459 hopopt[4] = 0; /* Type = 0x0000 (MLD) */
461 hopopt[6] = IPV6_TLV_PAD1; /* Pad1 */
462 hopopt[7] = IPV6_TLV_PAD1; /* Pad1 */
464 skb_put(skb, sizeof(*ip6h) + 8);
467 skb_set_transport_header(skb, skb->len);
468 mldq = (struct mld_msg *) icmp6_hdr(skb);
470 interval = ipv6_addr_any(group) ?
471 br->multicast_query_response_interval :
472 br->multicast_last_member_interval;
474 mldq->mld_type = ICMPV6_MGM_QUERY;
477 mldq->mld_maxdelay = htons((u16)jiffies_to_msecs(interval));
478 mldq->mld_reserved = 0;
479 mldq->mld_mca = *group;
482 mldq->mld_cksum = csum_ipv6_magic(&ip6h->saddr, &ip6h->daddr,
483 sizeof(*mldq), IPPROTO_ICMPV6,
486 skb_put(skb, sizeof(*mldq));
488 __skb_pull(skb, sizeof(*eth));
495 static struct sk_buff *br_multicast_alloc_query(struct net_bridge *br,
498 switch (addr->proto) {
499 case htons(ETH_P_IP):
500 return br_ip4_multicast_alloc_query(br, addr->u.ip4);
501 #if IS_ENABLED(CONFIG_IPV6)
502 case htons(ETH_P_IPV6):
503 return br_ip6_multicast_alloc_query(br, &addr->u.ip6);
509 static struct net_bridge_mdb_entry *br_multicast_get_group(
510 struct net_bridge *br, struct net_bridge_port *port,
511 struct br_ip *group, int hash)
513 struct net_bridge_mdb_htable *mdb;
514 struct net_bridge_mdb_entry *mp;
515 unsigned int count = 0;
520 mdb = rcu_dereference_protected(br->mdb, 1);
521 hlist_for_each_entry(mp, &mdb->mhash[hash], hlist[mdb->ver]) {
523 if (unlikely(br_ip_equal(group, &mp->addr)))
530 if (unlikely(count > br->hash_elasticity && count)) {
532 br_info(br, "Multicast hash table "
533 "chain limit reached: %s\n",
534 port ? port->dev->name : br->dev->name);
536 elasticity = br->hash_elasticity;
539 if (mdb->size >= max) {
541 if (unlikely(max > br->hash_max)) {
542 br_warn(br, "Multicast hash table maximum of %d "
543 "reached, disabling snooping: %s\n",
545 port ? port->dev->name : br->dev->name);
548 br->multicast_disabled = 1;
553 if (max > mdb->max || elasticity) {
556 br_info(br, "Multicast hash table "
558 port ? port->dev->name : br->dev->name);
563 err = br_mdb_rehash(&br->mdb, max, elasticity);
565 br_warn(br, "Cannot rehash multicast "
566 "hash table, disabling snooping: %s, %d, %d\n",
567 port ? port->dev->name : br->dev->name,
583 struct net_bridge_mdb_entry *br_multicast_new_group(struct net_bridge *br,
584 struct net_bridge_port *port, struct br_ip *group)
586 struct net_bridge_mdb_htable *mdb;
587 struct net_bridge_mdb_entry *mp;
591 mdb = rcu_dereference_protected(br->mdb, 1);
593 err = br_mdb_rehash(&br->mdb, BR_HASH_SIZE, 0);
599 hash = br_ip_hash(mdb, group);
600 mp = br_multicast_get_group(br, port, group, hash);
601 switch (PTR_ERR(mp)) {
607 mdb = rcu_dereference_protected(br->mdb, 1);
608 hash = br_ip_hash(mdb, group);
615 mp = kzalloc(sizeof(*mp), GFP_ATOMIC);
617 return ERR_PTR(-ENOMEM);
622 setup_timer(&mp->timer, br_multicast_group_expired,
625 hlist_add_head_rcu(&mp->hlist[mdb->ver], &mdb->mhash[hash]);
632 struct net_bridge_port_group *br_multicast_new_port_group(
633 struct net_bridge_port *port,
635 struct net_bridge_port_group __rcu *next,
638 struct net_bridge_port_group *p;
640 p = kzalloc(sizeof(*p), GFP_ATOMIC);
647 rcu_assign_pointer(p->next, next);
648 hlist_add_head(&p->mglist, &port->mglist);
649 setup_timer(&p->timer, br_multicast_port_group_expired,
654 static int br_multicast_add_group(struct net_bridge *br,
655 struct net_bridge_port *port,
658 struct net_bridge_mdb_entry *mp;
659 struct net_bridge_port_group *p;
660 struct net_bridge_port_group __rcu **pp;
663 spin_lock(&br->multicast_lock);
664 if (!netif_running(br->dev) ||
665 (port && port->state == BR_STATE_DISABLED))
668 mp = br_multicast_new_group(br, port, group);
678 for (pp = &mp->ports;
679 (p = mlock_dereference(*pp, br)) != NULL;
683 if ((unsigned long)p->port < (unsigned long)port)
687 p = br_multicast_new_port_group(port, group, *pp, MDB_TEMPORARY);
690 rcu_assign_pointer(*pp, p);
691 br_mdb_notify(br->dev, port, group, RTM_NEWMDB);
697 spin_unlock(&br->multicast_lock);
701 static int br_ip4_multicast_add_group(struct net_bridge *br,
702 struct net_bridge_port *port,
706 struct br_ip br_group;
708 if (ipv4_is_local_multicast(group))
711 br_group.u.ip4 = group;
712 br_group.proto = htons(ETH_P_IP);
715 return br_multicast_add_group(br, port, &br_group);
718 #if IS_ENABLED(CONFIG_IPV6)
719 static int br_ip6_multicast_add_group(struct net_bridge *br,
720 struct net_bridge_port *port,
721 const struct in6_addr *group,
724 struct br_ip br_group;
726 if (!ipv6_is_transient_multicast(group))
729 br_group.u.ip6 = *group;
730 br_group.proto = htons(ETH_P_IPV6);
733 return br_multicast_add_group(br, port, &br_group);
737 static void br_multicast_router_expired(unsigned long data)
739 struct net_bridge_port *port = (void *)data;
740 struct net_bridge *br = port->br;
742 spin_lock(&br->multicast_lock);
743 if (port->multicast_router != 1 ||
744 timer_pending(&port->multicast_router_timer) ||
745 hlist_unhashed(&port->rlist))
748 hlist_del_init_rcu(&port->rlist);
751 spin_unlock(&br->multicast_lock);
754 static void br_multicast_local_router_expired(unsigned long data)
758 static void br_multicast_querier_expired(unsigned long data)
760 struct net_bridge *br = (void *)data;
762 spin_lock(&br->multicast_lock);
763 if (!netif_running(br->dev) || br->multicast_disabled)
766 br_multicast_start_querier(br);
769 spin_unlock(&br->multicast_lock);
772 static void __br_multicast_send_query(struct net_bridge *br,
773 struct net_bridge_port *port,
778 skb = br_multicast_alloc_query(br, ip);
783 __skb_push(skb, sizeof(struct ethhdr));
784 skb->dev = port->dev;
785 NF_HOOK(NFPROTO_BRIDGE, NF_BR_LOCAL_OUT, skb, NULL, skb->dev,
791 static void br_multicast_send_query(struct net_bridge *br,
792 struct net_bridge_port *port, u32 sent)
795 struct br_ip br_group;
797 if (!netif_running(br->dev) || br->multicast_disabled ||
798 !br->multicast_querier ||
799 timer_pending(&br->multicast_querier_timer))
802 memset(&br_group.u, 0, sizeof(br_group.u));
804 br_group.proto = htons(ETH_P_IP);
805 __br_multicast_send_query(br, port, &br_group);
807 #if IS_ENABLED(CONFIG_IPV6)
808 br_group.proto = htons(ETH_P_IPV6);
809 __br_multicast_send_query(br, port, &br_group);
813 time += sent < br->multicast_startup_query_count ?
814 br->multicast_startup_query_interval :
815 br->multicast_query_interval;
816 mod_timer(port ? &port->multicast_query_timer :
817 &br->multicast_query_timer, time);
820 static void br_multicast_port_query_expired(unsigned long data)
822 struct net_bridge_port *port = (void *)data;
823 struct net_bridge *br = port->br;
825 spin_lock(&br->multicast_lock);
826 if (port->state == BR_STATE_DISABLED ||
827 port->state == BR_STATE_BLOCKING)
830 if (port->multicast_startup_queries_sent <
831 br->multicast_startup_query_count)
832 port->multicast_startup_queries_sent++;
834 br_multicast_send_query(port->br, port,
835 port->multicast_startup_queries_sent);
838 spin_unlock(&br->multicast_lock);
841 void br_multicast_add_port(struct net_bridge_port *port)
843 port->multicast_router = 1;
845 setup_timer(&port->multicast_router_timer, br_multicast_router_expired,
846 (unsigned long)port);
847 setup_timer(&port->multicast_query_timer,
848 br_multicast_port_query_expired, (unsigned long)port);
851 void br_multicast_del_port(struct net_bridge_port *port)
853 del_timer_sync(&port->multicast_router_timer);
856 static void __br_multicast_enable_port(struct net_bridge_port *port)
858 port->multicast_startup_queries_sent = 0;
860 if (try_to_del_timer_sync(&port->multicast_query_timer) >= 0 ||
861 del_timer(&port->multicast_query_timer))
862 mod_timer(&port->multicast_query_timer, jiffies);
865 void br_multicast_enable_port(struct net_bridge_port *port)
867 struct net_bridge *br = port->br;
869 spin_lock(&br->multicast_lock);
870 if (br->multicast_disabled || !netif_running(br->dev))
873 __br_multicast_enable_port(port);
876 spin_unlock(&br->multicast_lock);
879 void br_multicast_disable_port(struct net_bridge_port *port)
881 struct net_bridge *br = port->br;
882 struct net_bridge_port_group *pg;
883 struct hlist_node *n;
885 spin_lock(&br->multicast_lock);
886 hlist_for_each_entry_safe(pg, n, &port->mglist, mglist)
887 br_multicast_del_pg(br, pg);
889 if (!hlist_unhashed(&port->rlist))
890 hlist_del_init_rcu(&port->rlist);
891 del_timer(&port->multicast_router_timer);
892 del_timer(&port->multicast_query_timer);
893 spin_unlock(&br->multicast_lock);
896 static int br_ip4_multicast_igmp3_report(struct net_bridge *br,
897 struct net_bridge_port *port,
900 struct igmpv3_report *ih;
901 struct igmpv3_grec *grec;
910 if (!pskb_may_pull(skb, sizeof(*ih)))
913 br_vlan_get_tag(skb, &vid);
914 ih = igmpv3_report_hdr(skb);
915 num = ntohs(ih->ngrec);
918 for (i = 0; i < num; i++) {
919 len += sizeof(*grec);
920 if (!pskb_may_pull(skb, len))
923 grec = (void *)(skb->data + len - sizeof(*grec));
924 group = grec->grec_mca;
925 type = grec->grec_type;
927 len += ntohs(grec->grec_nsrcs) * 4;
928 if (!pskb_may_pull(skb, len))
931 /* We treat this as an IGMPv2 report for now. */
933 case IGMPV3_MODE_IS_INCLUDE:
934 case IGMPV3_MODE_IS_EXCLUDE:
935 case IGMPV3_CHANGE_TO_INCLUDE:
936 case IGMPV3_CHANGE_TO_EXCLUDE:
937 case IGMPV3_ALLOW_NEW_SOURCES:
938 case IGMPV3_BLOCK_OLD_SOURCES:
945 err = br_ip4_multicast_add_group(br, port, group, vid);
953 #if IS_ENABLED(CONFIG_IPV6)
954 static int br_ip6_multicast_mld2_report(struct net_bridge *br,
955 struct net_bridge_port *port,
958 struct icmp6hdr *icmp6h;
959 struct mld2_grec *grec;
966 if (!pskb_may_pull(skb, sizeof(*icmp6h)))
969 br_vlan_get_tag(skb, &vid);
970 icmp6h = icmp6_hdr(skb);
971 num = ntohs(icmp6h->icmp6_dataun.un_data16[1]);
972 len = sizeof(*icmp6h);
974 for (i = 0; i < num; i++) {
975 __be16 *nsrcs, _nsrcs;
977 nsrcs = skb_header_pointer(skb,
978 len + offsetof(struct mld2_grec,
980 sizeof(_nsrcs), &_nsrcs);
984 if (!pskb_may_pull(skb,
985 len + sizeof(*grec) +
986 sizeof(struct in6_addr) * ntohs(*nsrcs)))
989 grec = (struct mld2_grec *)(skb->data + len);
990 len += sizeof(*grec) +
991 sizeof(struct in6_addr) * ntohs(*nsrcs);
993 /* We treat these as MLDv1 reports for now. */
994 switch (grec->grec_type) {
995 case MLD2_MODE_IS_INCLUDE:
996 case MLD2_MODE_IS_EXCLUDE:
997 case MLD2_CHANGE_TO_INCLUDE:
998 case MLD2_CHANGE_TO_EXCLUDE:
999 case MLD2_ALLOW_NEW_SOURCES:
1000 case MLD2_BLOCK_OLD_SOURCES:
1007 err = br_ip6_multicast_add_group(br, port, &grec->grec_mca,
1018 * Add port to router_list
1019 * list is maintained ordered by pointer value
1020 * and locked by br->multicast_lock and RCU
1022 static void br_multicast_add_router(struct net_bridge *br,
1023 struct net_bridge_port *port)
1025 struct net_bridge_port *p;
1026 struct hlist_node *slot = NULL;
1028 hlist_for_each_entry(p, &br->router_list, rlist) {
1029 if ((unsigned long) port >= (unsigned long) p)
1035 hlist_add_after_rcu(slot, &port->rlist);
1037 hlist_add_head_rcu(&port->rlist, &br->router_list);
1040 static void br_multicast_mark_router(struct net_bridge *br,
1041 struct net_bridge_port *port)
1043 unsigned long now = jiffies;
1046 if (br->multicast_router == 1)
1047 mod_timer(&br->multicast_router_timer,
1048 now + br->multicast_querier_interval);
1052 if (port->multicast_router != 1)
1055 if (!hlist_unhashed(&port->rlist))
1058 br_multicast_add_router(br, port);
1061 mod_timer(&port->multicast_router_timer,
1062 now + br->multicast_querier_interval);
1065 static void br_multicast_query_received(struct net_bridge *br,
1066 struct net_bridge_port *port,
1070 mod_timer(&br->multicast_querier_timer,
1071 jiffies + br->multicast_querier_interval);
1072 else if (timer_pending(&br->multicast_querier_timer))
1075 br_multicast_mark_router(br, port);
1078 static int br_ip4_multicast_query(struct net_bridge *br,
1079 struct net_bridge_port *port,
1080 struct sk_buff *skb)
1082 const struct iphdr *iph = ip_hdr(skb);
1083 struct igmphdr *ih = igmp_hdr(skb);
1084 struct net_bridge_mdb_entry *mp;
1085 struct igmpv3_query *ih3;
1086 struct net_bridge_port_group *p;
1087 struct net_bridge_port_group __rcu **pp;
1088 unsigned long max_delay;
1089 unsigned long now = jiffies;
1094 spin_lock(&br->multicast_lock);
1095 if (!netif_running(br->dev) ||
1096 (port && port->state == BR_STATE_DISABLED))
1099 br_multicast_query_received(br, port, !!iph->saddr);
1103 if (skb->len == sizeof(*ih)) {
1104 max_delay = ih->code * (HZ / IGMP_TIMER_SCALE);
1107 max_delay = 10 * HZ;
1111 if (!pskb_may_pull(skb, sizeof(struct igmpv3_query))) {
1116 ih3 = igmpv3_query_hdr(skb);
1120 max_delay = ih3->code ?
1121 IGMPV3_MRC(ih3->code) * (HZ / IGMP_TIMER_SCALE) : 1;
1127 br_vlan_get_tag(skb, &vid);
1128 mp = br_mdb_ip4_get(mlock_dereference(br->mdb, br), group, vid);
1132 mod_timer(&mp->timer, now + br->multicast_membership_interval);
1133 mp->timer_armed = true;
1135 max_delay *= br->multicast_last_member_count;
1138 (timer_pending(&mp->timer) ?
1139 time_after(mp->timer.expires, now + max_delay) :
1140 try_to_del_timer_sync(&mp->timer) >= 0))
1141 mod_timer(&mp->timer, now + max_delay);
1143 for (pp = &mp->ports;
1144 (p = mlock_dereference(*pp, br)) != NULL;
1146 if (timer_pending(&p->timer) ?
1147 time_after(p->timer.expires, now + max_delay) :
1148 try_to_del_timer_sync(&p->timer) >= 0)
1149 mod_timer(&p->timer, now + max_delay);
1153 spin_unlock(&br->multicast_lock);
1157 #if IS_ENABLED(CONFIG_IPV6)
1158 static int br_ip6_multicast_query(struct net_bridge *br,
1159 struct net_bridge_port *port,
1160 struct sk_buff *skb)
1162 const struct ipv6hdr *ip6h = ipv6_hdr(skb);
1163 struct mld_msg *mld;
1164 struct net_bridge_mdb_entry *mp;
1165 struct mld2_query *mld2q;
1166 struct net_bridge_port_group *p;
1167 struct net_bridge_port_group __rcu **pp;
1168 unsigned long max_delay;
1169 unsigned long now = jiffies;
1170 const struct in6_addr *group = NULL;
1174 spin_lock(&br->multicast_lock);
1175 if (!netif_running(br->dev) ||
1176 (port && port->state == BR_STATE_DISABLED))
1179 br_multicast_query_received(br, port, !ipv6_addr_any(&ip6h->saddr));
1181 if (skb->len == sizeof(*mld)) {
1182 if (!pskb_may_pull(skb, sizeof(*mld))) {
1186 mld = (struct mld_msg *) icmp6_hdr(skb);
1187 max_delay = msecs_to_jiffies(ntohs(mld->mld_maxdelay));
1189 group = &mld->mld_mca;
1190 } else if (skb->len >= sizeof(*mld2q)) {
1191 if (!pskb_may_pull(skb, sizeof(*mld2q))) {
1195 mld2q = (struct mld2_query *)icmp6_hdr(skb);
1196 if (!mld2q->mld2q_nsrcs)
1197 group = &mld2q->mld2q_mca;
1198 max_delay = mld2q->mld2q_mrc ? MLDV2_MRC(ntohs(mld2q->mld2q_mrc)) : 1;
1204 br_vlan_get_tag(skb, &vid);
1205 mp = br_mdb_ip6_get(mlock_dereference(br->mdb, br), group, vid);
1209 mod_timer(&mp->timer, now + br->multicast_membership_interval);
1210 mp->timer_armed = true;
1212 max_delay *= br->multicast_last_member_count;
1214 (timer_pending(&mp->timer) ?
1215 time_after(mp->timer.expires, now + max_delay) :
1216 try_to_del_timer_sync(&mp->timer) >= 0))
1217 mod_timer(&mp->timer, now + max_delay);
1219 for (pp = &mp->ports;
1220 (p = mlock_dereference(*pp, br)) != NULL;
1222 if (timer_pending(&p->timer) ?
1223 time_after(p->timer.expires, now + max_delay) :
1224 try_to_del_timer_sync(&p->timer) >= 0)
1225 mod_timer(&p->timer, now + max_delay);
1229 spin_unlock(&br->multicast_lock);
1234 static void br_multicast_leave_group(struct net_bridge *br,
1235 struct net_bridge_port *port,
1236 struct br_ip *group)
1238 struct net_bridge_mdb_htable *mdb;
1239 struct net_bridge_mdb_entry *mp;
1240 struct net_bridge_port_group *p;
1244 spin_lock(&br->multicast_lock);
1245 if (!netif_running(br->dev) ||
1246 (port && port->state == BR_STATE_DISABLED) ||
1247 timer_pending(&br->multicast_querier_timer))
1250 mdb = mlock_dereference(br->mdb, br);
1251 mp = br_mdb_ip_get(mdb, group);
1255 if (br->multicast_querier &&
1256 !timer_pending(&br->multicast_querier_timer)) {
1257 __br_multicast_send_query(br, port, &mp->addr);
1259 time = jiffies + br->multicast_last_member_count *
1260 br->multicast_last_member_interval;
1261 mod_timer(port ? &port->multicast_query_timer :
1262 &br->multicast_query_timer, time);
1264 for (p = mlock_dereference(mp->ports, br);
1266 p = mlock_dereference(p->next, br)) {
1267 if (p->port != port)
1270 if (!hlist_unhashed(&p->mglist) &&
1271 (timer_pending(&p->timer) ?
1272 time_after(p->timer.expires, time) :
1273 try_to_del_timer_sync(&p->timer) >= 0)) {
1274 mod_timer(&p->timer, time);
1281 if (port && (port->flags & BR_MULTICAST_FAST_LEAVE)) {
1282 struct net_bridge_port_group __rcu **pp;
1284 for (pp = &mp->ports;
1285 (p = mlock_dereference(*pp, br)) != NULL;
1287 if (p->port != port)
1290 rcu_assign_pointer(*pp, p->next);
1291 hlist_del_init(&p->mglist);
1292 del_timer(&p->timer);
1293 call_rcu_bh(&p->rcu, br_multicast_free_pg);
1294 br_mdb_notify(br->dev, port, group, RTM_DELMDB);
1296 if (!mp->ports && !mp->mglist && mp->timer_armed &&
1297 netif_running(br->dev))
1298 mod_timer(&mp->timer, jiffies);
1304 time = now + br->multicast_last_member_count *
1305 br->multicast_last_member_interval;
1308 if (mp->mglist && mp->timer_armed &&
1309 (timer_pending(&mp->timer) ?
1310 time_after(mp->timer.expires, time) :
1311 try_to_del_timer_sync(&mp->timer) >= 0)) {
1312 mod_timer(&mp->timer, time);
1317 spin_unlock(&br->multicast_lock);
1320 static void br_ip4_multicast_leave_group(struct net_bridge *br,
1321 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);
1334 br_multicast_leave_group(br, port, &br_group);
1337 #if IS_ENABLED(CONFIG_IPV6)
1338 static void br_ip6_multicast_leave_group(struct net_bridge *br,
1339 struct net_bridge_port *port,
1340 const struct in6_addr *group,
1343 struct br_ip br_group;
1345 if (!ipv6_is_transient_multicast(group))
1348 br_group.u.ip6 = *group;
1349 br_group.proto = htons(ETH_P_IPV6);
1352 br_multicast_leave_group(br, port, &br_group);
1356 static int br_multicast_ipv4_rcv(struct net_bridge *br,
1357 struct net_bridge_port *port,
1358 struct sk_buff *skb)
1360 struct sk_buff *skb2 = skb;
1361 const struct iphdr *iph;
1364 unsigned int offset;
1368 /* We treat OOM as packet loss for now. */
1369 if (!pskb_may_pull(skb, sizeof(*iph)))
1374 if (iph->ihl < 5 || iph->version != 4)
1377 if (!pskb_may_pull(skb, ip_hdrlen(skb)))
1382 if (unlikely(ip_fast_csum((u8 *)iph, iph->ihl)))
1385 if (iph->protocol != IPPROTO_IGMP) {
1386 if (!ipv4_is_local_multicast(iph->daddr))
1387 BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1391 len = ntohs(iph->tot_len);
1392 if (skb->len < len || len < ip_hdrlen(skb))
1395 if (skb->len > len) {
1396 skb2 = skb_clone(skb, GFP_ATOMIC);
1400 err = pskb_trim_rcsum(skb2, len);
1405 len -= ip_hdrlen(skb2);
1406 offset = skb_network_offset(skb2) + ip_hdrlen(skb2);
1407 __skb_pull(skb2, offset);
1408 skb_reset_transport_header(skb2);
1411 if (!pskb_may_pull(skb2, sizeof(*ih)))
1414 switch (skb2->ip_summed) {
1415 case CHECKSUM_COMPLETE:
1416 if (!csum_fold(skb2->csum))
1421 if (skb_checksum_complete(skb2))
1427 br_vlan_get_tag(skb2, &vid);
1428 BR_INPUT_SKB_CB(skb)->igmp = 1;
1429 ih = igmp_hdr(skb2);
1432 case IGMP_HOST_MEMBERSHIP_REPORT:
1433 case IGMPV2_HOST_MEMBERSHIP_REPORT:
1434 BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1435 err = br_ip4_multicast_add_group(br, port, ih->group, vid);
1437 case IGMPV3_HOST_MEMBERSHIP_REPORT:
1438 err = br_ip4_multicast_igmp3_report(br, port, skb2);
1440 case IGMP_HOST_MEMBERSHIP_QUERY:
1441 err = br_ip4_multicast_query(br, port, skb2);
1443 case IGMP_HOST_LEAVE_MESSAGE:
1444 br_ip4_multicast_leave_group(br, port, ih->group, vid);
1449 __skb_push(skb2, offset);
1456 #if IS_ENABLED(CONFIG_IPV6)
1457 static int br_multicast_ipv6_rcv(struct net_bridge *br,
1458 struct net_bridge_port *port,
1459 struct sk_buff *skb)
1461 struct sk_buff *skb2;
1462 const struct ipv6hdr *ip6h;
1471 if (!pskb_may_pull(skb, sizeof(*ip6h)))
1474 ip6h = ipv6_hdr(skb);
1477 * We're interested in MLD messages only.
1479 * - MLD has always Router Alert hop-by-hop option
1480 * - But we do not support jumbrograms.
1482 if (ip6h->version != 6 ||
1483 ip6h->nexthdr != IPPROTO_HOPOPTS ||
1484 ip6h->payload_len == 0)
1487 len = ntohs(ip6h->payload_len) + sizeof(*ip6h);
1491 nexthdr = ip6h->nexthdr;
1492 offset = ipv6_skip_exthdr(skb, sizeof(*ip6h), &nexthdr, &frag_off);
1494 if (offset < 0 || nexthdr != IPPROTO_ICMPV6)
1497 /* Okay, we found ICMPv6 header */
1498 skb2 = skb_clone(skb, GFP_ATOMIC);
1503 if (!pskb_may_pull(skb2, offset + sizeof(struct icmp6hdr)))
1506 len -= offset - skb_network_offset(skb2);
1508 __skb_pull(skb2, offset);
1509 skb_reset_transport_header(skb2);
1510 skb_postpull_rcsum(skb2, skb_network_header(skb2),
1511 skb_network_header_len(skb2));
1513 icmp6_type = icmp6_hdr(skb2)->icmp6_type;
1515 switch (icmp6_type) {
1516 case ICMPV6_MGM_QUERY:
1517 case ICMPV6_MGM_REPORT:
1518 case ICMPV6_MGM_REDUCTION:
1519 case ICMPV6_MLD2_REPORT:
1526 /* Okay, we found MLD message. Check further. */
1527 if (skb2->len > len) {
1528 err = pskb_trim_rcsum(skb2, len);
1534 ip6h = ipv6_hdr(skb2);
1536 switch (skb2->ip_summed) {
1537 case CHECKSUM_COMPLETE:
1538 if (!csum_ipv6_magic(&ip6h->saddr, &ip6h->daddr, skb2->len,
1539 IPPROTO_ICMPV6, skb2->csum))
1543 skb2->csum = ~csum_unfold(csum_ipv6_magic(&ip6h->saddr,
1546 IPPROTO_ICMPV6, 0));
1547 if (__skb_checksum_complete(skb2))
1553 br_vlan_get_tag(skb, &vid);
1554 BR_INPUT_SKB_CB(skb)->igmp = 1;
1556 switch (icmp6_type) {
1557 case ICMPV6_MGM_REPORT:
1559 struct mld_msg *mld;
1560 if (!pskb_may_pull(skb2, sizeof(*mld))) {
1564 mld = (struct mld_msg *)skb_transport_header(skb2);
1565 BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1566 err = br_ip6_multicast_add_group(br, port, &mld->mld_mca, vid);
1569 case ICMPV6_MLD2_REPORT:
1570 err = br_ip6_multicast_mld2_report(br, port, skb2);
1572 case ICMPV6_MGM_QUERY:
1573 err = br_ip6_multicast_query(br, port, skb2);
1575 case ICMPV6_MGM_REDUCTION:
1577 struct mld_msg *mld;
1578 if (!pskb_may_pull(skb2, sizeof(*mld))) {
1582 mld = (struct mld_msg *)skb_transport_header(skb2);
1583 br_ip6_multicast_leave_group(br, port, &mld->mld_mca, vid);
1593 int br_multicast_rcv(struct net_bridge *br, struct net_bridge_port *port,
1594 struct sk_buff *skb)
1596 BR_INPUT_SKB_CB(skb)->igmp = 0;
1597 BR_INPUT_SKB_CB(skb)->mrouters_only = 0;
1599 if (br->multicast_disabled)
1602 switch (skb->protocol) {
1603 case htons(ETH_P_IP):
1604 return br_multicast_ipv4_rcv(br, port, skb);
1605 #if IS_ENABLED(CONFIG_IPV6)
1606 case htons(ETH_P_IPV6):
1607 return br_multicast_ipv6_rcv(br, port, skb);
1614 static void br_multicast_query_expired(unsigned long data)
1616 struct net_bridge *br = (void *)data;
1618 spin_lock(&br->multicast_lock);
1619 if (br->multicast_startup_queries_sent <
1620 br->multicast_startup_query_count)
1621 br->multicast_startup_queries_sent++;
1623 br_multicast_send_query(br, NULL, br->multicast_startup_queries_sent);
1625 spin_unlock(&br->multicast_lock);
1628 void br_multicast_init(struct net_bridge *br)
1630 br->hash_elasticity = 4;
1633 br->multicast_router = 1;
1634 br->multicast_querier = 0;
1635 br->multicast_query_use_ifaddr = 0;
1636 br->multicast_last_member_count = 2;
1637 br->multicast_startup_query_count = 2;
1639 br->multicast_last_member_interval = HZ;
1640 br->multicast_query_response_interval = 10 * HZ;
1641 br->multicast_startup_query_interval = 125 * HZ / 4;
1642 br->multicast_query_interval = 125 * HZ;
1643 br->multicast_querier_interval = 255 * HZ;
1644 br->multicast_membership_interval = 260 * HZ;
1646 spin_lock_init(&br->multicast_lock);
1647 setup_timer(&br->multicast_router_timer,
1648 br_multicast_local_router_expired, 0);
1649 setup_timer(&br->multicast_querier_timer,
1650 br_multicast_querier_expired, (unsigned long)br);
1651 setup_timer(&br->multicast_query_timer, br_multicast_query_expired,
1655 void br_multicast_open(struct net_bridge *br)
1657 br->multicast_startup_queries_sent = 0;
1659 if (br->multicast_disabled)
1662 mod_timer(&br->multicast_query_timer, jiffies);
1665 void br_multicast_stop(struct net_bridge *br)
1667 struct net_bridge_mdb_htable *mdb;
1668 struct net_bridge_mdb_entry *mp;
1669 struct hlist_node *n;
1673 del_timer_sync(&br->multicast_router_timer);
1674 del_timer_sync(&br->multicast_querier_timer);
1675 del_timer_sync(&br->multicast_query_timer);
1677 spin_lock_bh(&br->multicast_lock);
1678 mdb = mlock_dereference(br->mdb, br);
1685 for (i = 0; i < mdb->max; i++) {
1686 hlist_for_each_entry_safe(mp, n, &mdb->mhash[i],
1688 del_timer(&mp->timer);
1689 mp->timer_armed = false;
1690 call_rcu_bh(&mp->rcu, br_multicast_free_group);
1695 spin_unlock_bh(&br->multicast_lock);
1697 spin_lock_bh(&br->multicast_lock);
1702 call_rcu_bh(&mdb->rcu, br_mdb_free);
1705 spin_unlock_bh(&br->multicast_lock);
1708 int br_multicast_set_router(struct net_bridge *br, unsigned long val)
1712 spin_lock_bh(&br->multicast_lock);
1713 if (!netif_running(br->dev))
1719 del_timer(&br->multicast_router_timer);
1722 br->multicast_router = val;
1732 spin_unlock_bh(&br->multicast_lock);
1737 int br_multicast_set_port_router(struct net_bridge_port *p, unsigned long val)
1739 struct net_bridge *br = p->br;
1742 spin_lock(&br->multicast_lock);
1743 if (!netif_running(br->dev) || p->state == BR_STATE_DISABLED)
1750 p->multicast_router = val;
1753 if (val < 2 && !hlist_unhashed(&p->rlist))
1754 hlist_del_init_rcu(&p->rlist);
1759 del_timer(&p->multicast_router_timer);
1764 br_multicast_add_router(br, p);
1773 spin_unlock(&br->multicast_lock);
1778 static void br_multicast_start_querier(struct net_bridge *br)
1780 struct net_bridge_port *port;
1782 br_multicast_open(br);
1784 list_for_each_entry(port, &br->port_list, list) {
1785 if (port->state == BR_STATE_DISABLED ||
1786 port->state == BR_STATE_BLOCKING)
1789 __br_multicast_enable_port(port);
1793 int br_multicast_toggle(struct net_bridge *br, unsigned long val)
1796 struct net_bridge_mdb_htable *mdb;
1798 spin_lock_bh(&br->multicast_lock);
1799 if (br->multicast_disabled == !val)
1802 br->multicast_disabled = !val;
1803 if (br->multicast_disabled)
1806 if (!netif_running(br->dev))
1809 mdb = mlock_dereference(br->mdb, br);
1814 br->multicast_disabled = !!val;
1818 err = br_mdb_rehash(&br->mdb, mdb->max,
1819 br->hash_elasticity);
1824 br_multicast_start_querier(br);
1827 spin_unlock_bh(&br->multicast_lock);
1832 int br_multicast_set_querier(struct net_bridge *br, unsigned long val)
1836 spin_lock_bh(&br->multicast_lock);
1837 if (br->multicast_querier == val)
1840 br->multicast_querier = val;
1842 br_multicast_start_querier(br);
1845 spin_unlock_bh(&br->multicast_lock);
1850 int br_multicast_set_hash_max(struct net_bridge *br, unsigned long val)
1854 struct net_bridge_mdb_htable *mdb;
1856 spin_lock(&br->multicast_lock);
1857 if (!netif_running(br->dev))
1861 if (!is_power_of_2(val))
1864 mdb = mlock_dereference(br->mdb, br);
1865 if (mdb && val < mdb->size)
1881 err = br_mdb_rehash(&br->mdb, br->hash_max,
1882 br->hash_elasticity);
1888 spin_unlock(&br->multicast_lock);