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,
1017 static void br_multicast_update_querier_timer(struct net_bridge *br,
1018 unsigned long max_delay)
1020 if (!timer_pending(&br->multicast_querier_timer))
1021 br->multicast_querier_delay_time = jiffies + max_delay;
1023 mod_timer(&br->multicast_querier_timer,
1024 jiffies + br->multicast_querier_interval);
1028 * Add port to router_list
1029 * list is maintained ordered by pointer value
1030 * and locked by br->multicast_lock and RCU
1032 static void br_multicast_add_router(struct net_bridge *br,
1033 struct net_bridge_port *port)
1035 struct net_bridge_port *p;
1036 struct hlist_node *slot = NULL;
1038 hlist_for_each_entry(p, &br->router_list, rlist) {
1039 if ((unsigned long) port >= (unsigned long) p)
1045 hlist_add_after_rcu(slot, &port->rlist);
1047 hlist_add_head_rcu(&port->rlist, &br->router_list);
1050 static void br_multicast_mark_router(struct net_bridge *br,
1051 struct net_bridge_port *port)
1053 unsigned long now = jiffies;
1056 if (br->multicast_router == 1)
1057 mod_timer(&br->multicast_router_timer,
1058 now + br->multicast_querier_interval);
1062 if (port->multicast_router != 1)
1065 if (!hlist_unhashed(&port->rlist))
1068 br_multicast_add_router(br, port);
1071 mod_timer(&port->multicast_router_timer,
1072 now + br->multicast_querier_interval);
1075 static void br_multicast_query_received(struct net_bridge *br,
1076 struct net_bridge_port *port,
1078 unsigned long max_delay)
1081 br_multicast_update_querier_timer(br, max_delay);
1082 else if (timer_pending(&br->multicast_querier_timer))
1085 br_multicast_mark_router(br, port);
1088 static int br_ip4_multicast_query(struct net_bridge *br,
1089 struct net_bridge_port *port,
1090 struct sk_buff *skb)
1092 const struct iphdr *iph = ip_hdr(skb);
1093 struct igmphdr *ih = igmp_hdr(skb);
1094 struct net_bridge_mdb_entry *mp;
1095 struct igmpv3_query *ih3;
1096 struct net_bridge_port_group *p;
1097 struct net_bridge_port_group __rcu **pp;
1098 unsigned long max_delay;
1099 unsigned long now = jiffies;
1104 spin_lock(&br->multicast_lock);
1105 if (!netif_running(br->dev) ||
1106 (port && port->state == BR_STATE_DISABLED))
1111 if (skb->len == sizeof(*ih)) {
1112 max_delay = ih->code * (HZ / IGMP_TIMER_SCALE);
1115 max_delay = 10 * HZ;
1119 if (!pskb_may_pull(skb, sizeof(struct igmpv3_query))) {
1124 ih3 = igmpv3_query_hdr(skb);
1128 max_delay = ih3->code ?
1129 IGMPV3_MRC(ih3->code) * (HZ / IGMP_TIMER_SCALE) : 1;
1132 br_multicast_query_received(br, port, !!iph->saddr, max_delay);
1137 br_vlan_get_tag(skb, &vid);
1138 mp = br_mdb_ip4_get(mlock_dereference(br->mdb, br), group, vid);
1142 mod_timer(&mp->timer, now + br->multicast_membership_interval);
1143 mp->timer_armed = true;
1145 max_delay *= br->multicast_last_member_count;
1148 (timer_pending(&mp->timer) ?
1149 time_after(mp->timer.expires, now + max_delay) :
1150 try_to_del_timer_sync(&mp->timer) >= 0))
1151 mod_timer(&mp->timer, now + max_delay);
1153 for (pp = &mp->ports;
1154 (p = mlock_dereference(*pp, br)) != NULL;
1156 if (timer_pending(&p->timer) ?
1157 time_after(p->timer.expires, now + max_delay) :
1158 try_to_del_timer_sync(&p->timer) >= 0)
1159 mod_timer(&p->timer, now + max_delay);
1163 spin_unlock(&br->multicast_lock);
1167 #if IS_ENABLED(CONFIG_IPV6)
1168 static int br_ip6_multicast_query(struct net_bridge *br,
1169 struct net_bridge_port *port,
1170 struct sk_buff *skb)
1172 const struct ipv6hdr *ip6h = ipv6_hdr(skb);
1173 struct mld_msg *mld;
1174 struct net_bridge_mdb_entry *mp;
1175 struct mld2_query *mld2q;
1176 struct net_bridge_port_group *p;
1177 struct net_bridge_port_group __rcu **pp;
1178 unsigned long max_delay;
1179 unsigned long now = jiffies;
1180 const struct in6_addr *group = NULL;
1184 spin_lock(&br->multicast_lock);
1185 if (!netif_running(br->dev) ||
1186 (port && port->state == BR_STATE_DISABLED))
1189 if (skb->len == sizeof(*mld)) {
1190 if (!pskb_may_pull(skb, sizeof(*mld))) {
1194 mld = (struct mld_msg *) icmp6_hdr(skb);
1195 max_delay = msecs_to_jiffies(ntohs(mld->mld_maxdelay));
1197 group = &mld->mld_mca;
1199 if (!pskb_may_pull(skb, sizeof(*mld2q))) {
1203 mld2q = (struct mld2_query *)icmp6_hdr(skb);
1204 if (!mld2q->mld2q_nsrcs)
1205 group = &mld2q->mld2q_mca;
1206 max_delay = mld2q->mld2q_mrc ? MLDV2_MRC(ntohs(mld2q->mld2q_mrc)) : 1;
1209 br_multicast_query_received(br, port, !ipv6_addr_any(&ip6h->saddr),
1215 br_vlan_get_tag(skb, &vid);
1216 mp = br_mdb_ip6_get(mlock_dereference(br->mdb, br), group, vid);
1220 mod_timer(&mp->timer, now + br->multicast_membership_interval);
1221 mp->timer_armed = true;
1223 max_delay *= br->multicast_last_member_count;
1225 (timer_pending(&mp->timer) ?
1226 time_after(mp->timer.expires, now + max_delay) :
1227 try_to_del_timer_sync(&mp->timer) >= 0))
1228 mod_timer(&mp->timer, now + max_delay);
1230 for (pp = &mp->ports;
1231 (p = mlock_dereference(*pp, br)) != NULL;
1233 if (timer_pending(&p->timer) ?
1234 time_after(p->timer.expires, now + max_delay) :
1235 try_to_del_timer_sync(&p->timer) >= 0)
1236 mod_timer(&p->timer, now + max_delay);
1240 spin_unlock(&br->multicast_lock);
1245 static void br_multicast_leave_group(struct net_bridge *br,
1246 struct net_bridge_port *port,
1247 struct br_ip *group)
1249 struct net_bridge_mdb_htable *mdb;
1250 struct net_bridge_mdb_entry *mp;
1251 struct net_bridge_port_group *p;
1255 spin_lock(&br->multicast_lock);
1256 if (!netif_running(br->dev) ||
1257 (port && port->state == BR_STATE_DISABLED) ||
1258 timer_pending(&br->multicast_querier_timer))
1261 mdb = mlock_dereference(br->mdb, br);
1262 mp = br_mdb_ip_get(mdb, group);
1266 if (br->multicast_querier &&
1267 !timer_pending(&br->multicast_querier_timer)) {
1268 __br_multicast_send_query(br, port, &mp->addr);
1270 time = jiffies + br->multicast_last_member_count *
1271 br->multicast_last_member_interval;
1272 mod_timer(port ? &port->multicast_query_timer :
1273 &br->multicast_query_timer, time);
1275 for (p = mlock_dereference(mp->ports, br);
1277 p = mlock_dereference(p->next, br)) {
1278 if (p->port != port)
1281 if (!hlist_unhashed(&p->mglist) &&
1282 (timer_pending(&p->timer) ?
1283 time_after(p->timer.expires, time) :
1284 try_to_del_timer_sync(&p->timer) >= 0)) {
1285 mod_timer(&p->timer, time);
1292 if (port && (port->flags & BR_MULTICAST_FAST_LEAVE)) {
1293 struct net_bridge_port_group __rcu **pp;
1295 for (pp = &mp->ports;
1296 (p = mlock_dereference(*pp, br)) != NULL;
1298 if (p->port != port)
1301 rcu_assign_pointer(*pp, p->next);
1302 hlist_del_init(&p->mglist);
1303 del_timer(&p->timer);
1304 call_rcu_bh(&p->rcu, br_multicast_free_pg);
1305 br_mdb_notify(br->dev, port, group, RTM_DELMDB);
1307 if (!mp->ports && !mp->mglist && mp->timer_armed &&
1308 netif_running(br->dev))
1309 mod_timer(&mp->timer, jiffies);
1315 time = now + br->multicast_last_member_count *
1316 br->multicast_last_member_interval;
1319 if (mp->mglist && mp->timer_armed &&
1320 (timer_pending(&mp->timer) ?
1321 time_after(mp->timer.expires, time) :
1322 try_to_del_timer_sync(&mp->timer) >= 0)) {
1323 mod_timer(&mp->timer, time);
1328 spin_unlock(&br->multicast_lock);
1331 static void br_ip4_multicast_leave_group(struct net_bridge *br,
1332 struct net_bridge_port *port,
1336 struct br_ip br_group;
1338 if (ipv4_is_local_multicast(group))
1341 br_group.u.ip4 = group;
1342 br_group.proto = htons(ETH_P_IP);
1345 br_multicast_leave_group(br, port, &br_group);
1348 #if IS_ENABLED(CONFIG_IPV6)
1349 static void br_ip6_multicast_leave_group(struct net_bridge *br,
1350 struct net_bridge_port *port,
1351 const struct in6_addr *group,
1354 struct br_ip br_group;
1356 if (!ipv6_is_transient_multicast(group))
1359 br_group.u.ip6 = *group;
1360 br_group.proto = htons(ETH_P_IPV6);
1363 br_multicast_leave_group(br, port, &br_group);
1367 static int br_multicast_ipv4_rcv(struct net_bridge *br,
1368 struct net_bridge_port *port,
1369 struct sk_buff *skb)
1371 struct sk_buff *skb2 = skb;
1372 const struct iphdr *iph;
1375 unsigned int offset;
1379 /* We treat OOM as packet loss for now. */
1380 if (!pskb_may_pull(skb, sizeof(*iph)))
1385 if (iph->ihl < 5 || iph->version != 4)
1388 if (!pskb_may_pull(skb, ip_hdrlen(skb)))
1393 if (unlikely(ip_fast_csum((u8 *)iph, iph->ihl)))
1396 if (iph->protocol != IPPROTO_IGMP) {
1397 if (!ipv4_is_local_multicast(iph->daddr))
1398 BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1402 len = ntohs(iph->tot_len);
1403 if (skb->len < len || len < ip_hdrlen(skb))
1406 if (skb->len > len) {
1407 skb2 = skb_clone(skb, GFP_ATOMIC);
1411 err = pskb_trim_rcsum(skb2, len);
1416 len -= ip_hdrlen(skb2);
1417 offset = skb_network_offset(skb2) + ip_hdrlen(skb2);
1418 __skb_pull(skb2, offset);
1419 skb_reset_transport_header(skb2);
1422 if (!pskb_may_pull(skb2, sizeof(*ih)))
1425 switch (skb2->ip_summed) {
1426 case CHECKSUM_COMPLETE:
1427 if (!csum_fold(skb2->csum))
1432 if (skb_checksum_complete(skb2))
1438 br_vlan_get_tag(skb2, &vid);
1439 BR_INPUT_SKB_CB(skb)->igmp = 1;
1440 ih = igmp_hdr(skb2);
1443 case IGMP_HOST_MEMBERSHIP_REPORT:
1444 case IGMPV2_HOST_MEMBERSHIP_REPORT:
1445 BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1446 err = br_ip4_multicast_add_group(br, port, ih->group, vid);
1448 case IGMPV3_HOST_MEMBERSHIP_REPORT:
1449 err = br_ip4_multicast_igmp3_report(br, port, skb2);
1451 case IGMP_HOST_MEMBERSHIP_QUERY:
1452 err = br_ip4_multicast_query(br, port, skb2);
1454 case IGMP_HOST_LEAVE_MESSAGE:
1455 br_ip4_multicast_leave_group(br, port, ih->group, vid);
1460 __skb_push(skb2, offset);
1467 #if IS_ENABLED(CONFIG_IPV6)
1468 static int br_multicast_ipv6_rcv(struct net_bridge *br,
1469 struct net_bridge_port *port,
1470 struct sk_buff *skb)
1472 struct sk_buff *skb2;
1473 const struct ipv6hdr *ip6h;
1482 if (!pskb_may_pull(skb, sizeof(*ip6h)))
1485 ip6h = ipv6_hdr(skb);
1488 * We're interested in MLD messages only.
1490 * - MLD has always Router Alert hop-by-hop option
1491 * - But we do not support jumbrograms.
1493 if (ip6h->version != 6 ||
1494 ip6h->nexthdr != IPPROTO_HOPOPTS ||
1495 ip6h->payload_len == 0)
1498 len = ntohs(ip6h->payload_len) + sizeof(*ip6h);
1502 nexthdr = ip6h->nexthdr;
1503 offset = ipv6_skip_exthdr(skb, sizeof(*ip6h), &nexthdr, &frag_off);
1505 if (offset < 0 || nexthdr != IPPROTO_ICMPV6)
1508 /* Okay, we found ICMPv6 header */
1509 skb2 = skb_clone(skb, GFP_ATOMIC);
1514 if (!pskb_may_pull(skb2, offset + sizeof(struct icmp6hdr)))
1517 len -= offset - skb_network_offset(skb2);
1519 __skb_pull(skb2, offset);
1520 skb_reset_transport_header(skb2);
1521 skb_postpull_rcsum(skb2, skb_network_header(skb2),
1522 skb_network_header_len(skb2));
1524 icmp6_type = icmp6_hdr(skb2)->icmp6_type;
1526 switch (icmp6_type) {
1527 case ICMPV6_MGM_QUERY:
1528 case ICMPV6_MGM_REPORT:
1529 case ICMPV6_MGM_REDUCTION:
1530 case ICMPV6_MLD2_REPORT:
1537 /* Okay, we found MLD message. Check further. */
1538 if (skb2->len > len) {
1539 err = pskb_trim_rcsum(skb2, len);
1545 ip6h = ipv6_hdr(skb2);
1547 switch (skb2->ip_summed) {
1548 case CHECKSUM_COMPLETE:
1549 if (!csum_ipv6_magic(&ip6h->saddr, &ip6h->daddr, skb2->len,
1550 IPPROTO_ICMPV6, skb2->csum))
1554 skb2->csum = ~csum_unfold(csum_ipv6_magic(&ip6h->saddr,
1557 IPPROTO_ICMPV6, 0));
1558 if (__skb_checksum_complete(skb2))
1564 br_vlan_get_tag(skb, &vid);
1565 BR_INPUT_SKB_CB(skb)->igmp = 1;
1567 switch (icmp6_type) {
1568 case ICMPV6_MGM_REPORT:
1570 struct mld_msg *mld;
1571 if (!pskb_may_pull(skb2, sizeof(*mld))) {
1575 mld = (struct mld_msg *)skb_transport_header(skb2);
1576 BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1577 err = br_ip6_multicast_add_group(br, port, &mld->mld_mca, vid);
1580 case ICMPV6_MLD2_REPORT:
1581 err = br_ip6_multicast_mld2_report(br, port, skb2);
1583 case ICMPV6_MGM_QUERY:
1584 err = br_ip6_multicast_query(br, port, skb2);
1586 case ICMPV6_MGM_REDUCTION:
1588 struct mld_msg *mld;
1589 if (!pskb_may_pull(skb2, sizeof(*mld))) {
1593 mld = (struct mld_msg *)skb_transport_header(skb2);
1594 br_ip6_multicast_leave_group(br, port, &mld->mld_mca, vid);
1604 int br_multicast_rcv(struct net_bridge *br, struct net_bridge_port *port,
1605 struct sk_buff *skb)
1607 BR_INPUT_SKB_CB(skb)->igmp = 0;
1608 BR_INPUT_SKB_CB(skb)->mrouters_only = 0;
1610 if (br->multicast_disabled)
1613 switch (skb->protocol) {
1614 case htons(ETH_P_IP):
1615 return br_multicast_ipv4_rcv(br, port, skb);
1616 #if IS_ENABLED(CONFIG_IPV6)
1617 case htons(ETH_P_IPV6):
1618 return br_multicast_ipv6_rcv(br, port, skb);
1625 static void br_multicast_query_expired(unsigned long data)
1627 struct net_bridge *br = (void *)data;
1629 spin_lock(&br->multicast_lock);
1630 if (br->multicast_startup_queries_sent <
1631 br->multicast_startup_query_count)
1632 br->multicast_startup_queries_sent++;
1634 br_multicast_send_query(br, NULL, br->multicast_startup_queries_sent);
1636 spin_unlock(&br->multicast_lock);
1639 void br_multicast_init(struct net_bridge *br)
1641 br->hash_elasticity = 4;
1644 br->multicast_router = 1;
1645 br->multicast_querier = 0;
1646 br->multicast_query_use_ifaddr = 0;
1647 br->multicast_last_member_count = 2;
1648 br->multicast_startup_query_count = 2;
1650 br->multicast_last_member_interval = HZ;
1651 br->multicast_query_response_interval = 10 * HZ;
1652 br->multicast_startup_query_interval = 125 * HZ / 4;
1653 br->multicast_query_interval = 125 * HZ;
1654 br->multicast_querier_interval = 255 * HZ;
1655 br->multicast_membership_interval = 260 * HZ;
1657 br->multicast_querier_delay_time = 0;
1659 spin_lock_init(&br->multicast_lock);
1660 setup_timer(&br->multicast_router_timer,
1661 br_multicast_local_router_expired, 0);
1662 setup_timer(&br->multicast_querier_timer,
1663 br_multicast_querier_expired, (unsigned long)br);
1664 setup_timer(&br->multicast_query_timer, br_multicast_query_expired,
1668 void br_multicast_open(struct net_bridge *br)
1670 br->multicast_startup_queries_sent = 0;
1672 if (br->multicast_disabled)
1675 mod_timer(&br->multicast_query_timer, jiffies);
1678 void br_multicast_stop(struct net_bridge *br)
1680 struct net_bridge_mdb_htable *mdb;
1681 struct net_bridge_mdb_entry *mp;
1682 struct hlist_node *n;
1686 del_timer_sync(&br->multicast_router_timer);
1687 del_timer_sync(&br->multicast_querier_timer);
1688 del_timer_sync(&br->multicast_query_timer);
1690 spin_lock_bh(&br->multicast_lock);
1691 mdb = mlock_dereference(br->mdb, br);
1698 for (i = 0; i < mdb->max; i++) {
1699 hlist_for_each_entry_safe(mp, n, &mdb->mhash[i],
1701 del_timer(&mp->timer);
1702 mp->timer_armed = false;
1703 call_rcu_bh(&mp->rcu, br_multicast_free_group);
1708 spin_unlock_bh(&br->multicast_lock);
1710 spin_lock_bh(&br->multicast_lock);
1715 call_rcu_bh(&mdb->rcu, br_mdb_free);
1718 spin_unlock_bh(&br->multicast_lock);
1721 int br_multicast_set_router(struct net_bridge *br, unsigned long val)
1725 spin_lock_bh(&br->multicast_lock);
1726 if (!netif_running(br->dev))
1732 del_timer(&br->multicast_router_timer);
1735 br->multicast_router = val;
1745 spin_unlock_bh(&br->multicast_lock);
1750 int br_multicast_set_port_router(struct net_bridge_port *p, unsigned long val)
1752 struct net_bridge *br = p->br;
1755 spin_lock(&br->multicast_lock);
1756 if (!netif_running(br->dev) || p->state == BR_STATE_DISABLED)
1763 p->multicast_router = val;
1766 if (val < 2 && !hlist_unhashed(&p->rlist))
1767 hlist_del_init_rcu(&p->rlist);
1772 del_timer(&p->multicast_router_timer);
1777 br_multicast_add_router(br, p);
1786 spin_unlock(&br->multicast_lock);
1791 static void br_multicast_start_querier(struct net_bridge *br)
1793 struct net_bridge_port *port;
1795 br_multicast_open(br);
1797 list_for_each_entry(port, &br->port_list, list) {
1798 if (port->state == BR_STATE_DISABLED ||
1799 port->state == BR_STATE_BLOCKING)
1802 __br_multicast_enable_port(port);
1806 int br_multicast_toggle(struct net_bridge *br, unsigned long val)
1809 struct net_bridge_mdb_htable *mdb;
1811 spin_lock_bh(&br->multicast_lock);
1812 if (br->multicast_disabled == !val)
1815 br->multicast_disabled = !val;
1816 if (br->multicast_disabled)
1819 if (!netif_running(br->dev))
1822 mdb = mlock_dereference(br->mdb, br);
1827 br->multicast_disabled = !!val;
1831 err = br_mdb_rehash(&br->mdb, mdb->max,
1832 br->hash_elasticity);
1837 br_multicast_start_querier(br);
1840 spin_unlock_bh(&br->multicast_lock);
1845 int br_multicast_set_querier(struct net_bridge *br, unsigned long val)
1847 unsigned long max_delay;
1851 spin_lock_bh(&br->multicast_lock);
1852 if (br->multicast_querier == val)
1855 br->multicast_querier = val;
1859 max_delay = br->multicast_query_response_interval;
1860 if (!timer_pending(&br->multicast_querier_timer))
1861 br->multicast_querier_delay_time = jiffies + max_delay;
1863 br_multicast_start_querier(br);
1866 spin_unlock_bh(&br->multicast_lock);
1871 int br_multicast_set_hash_max(struct net_bridge *br, unsigned long val)
1875 struct net_bridge_mdb_htable *mdb;
1877 spin_lock(&br->multicast_lock);
1878 if (!netif_running(br->dev))
1882 if (!is_power_of_2(val))
1885 mdb = mlock_dereference(br->mdb, br);
1886 if (mdb && val < mdb->size)
1902 err = br_mdb_rehash(&br->mdb, br->hash_max,
1903 br->hash_elasticity);
1909 spin_unlock(&br->multicast_lock);