2 * Bridge multicast support.
4 * Copyright (c) 2010 Herbert Xu <herbert@gondor.apana.org.au>
6 * This program is free software; you can redistribute it and/or modify it
7 * under the terms of the GNU General Public License as published by the Free
8 * Software Foundation; either version 2 of the License, or (at your option)
13 #include <linux/err.h>
14 #include <linux/if_ether.h>
15 #include <linux/igmp.h>
16 #include <linux/jhash.h>
17 #include <linux/kernel.h>
18 #include <linux/log2.h>
19 #include <linux/netdevice.h>
20 #include <linux/netfilter_bridge.h>
21 #include <linux/random.h>
22 #include <linux/rculist.h>
23 #include <linux/skbuff.h>
24 #include <linux/slab.h>
25 #include <linux/timer.h>
27 #if IS_ENABLED(CONFIG_IPV6)
30 #include <net/ip6_checksum.h>
33 #include "br_private.h"
35 static void br_multicast_start_querier(struct net_bridge *br);
36 unsigned int br_mdb_rehash_seq;
38 static inline int br_ip_equal(const struct br_ip *a, const struct br_ip *b)
40 if (a->proto != b->proto)
46 return a->u.ip4 == b->u.ip4;
47 #if IS_ENABLED(CONFIG_IPV6)
48 case htons(ETH_P_IPV6):
49 return ipv6_addr_equal(&a->u.ip6, &b->u.ip6);
55 static inline int __br_ip4_hash(struct net_bridge_mdb_htable *mdb, __be32 ip,
58 return jhash_2words((__force u32)ip, vid, mdb->secret) & (mdb->max - 1);
61 #if IS_ENABLED(CONFIG_IPV6)
62 static inline int __br_ip6_hash(struct net_bridge_mdb_htable *mdb,
63 const struct in6_addr *ip,
66 return jhash_2words(ipv6_addr_hash(ip), vid,
67 mdb->secret) & (mdb->max - 1);
71 static inline int br_ip_hash(struct net_bridge_mdb_htable *mdb,
76 return __br_ip4_hash(mdb, ip->u.ip4, ip->vid);
77 #if IS_ENABLED(CONFIG_IPV6)
78 case htons(ETH_P_IPV6):
79 return __br_ip6_hash(mdb, &ip->u.ip6, ip->vid);
85 static struct net_bridge_mdb_entry *__br_mdb_ip_get(
86 struct net_bridge_mdb_htable *mdb, struct br_ip *dst, int hash)
88 struct net_bridge_mdb_entry *mp;
91 hlist_for_each_entry_rcu(mp, p, &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,
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;
149 switch (skb->protocol) {
150 case htons(ETH_P_IP):
151 ip.u.ip4 = ip_hdr(skb)->daddr;
153 #if IS_ENABLED(CONFIG_IPV6)
154 case htons(ETH_P_IPV6):
155 ip.u.ip6 = ipv6_hdr(skb)->daddr;
162 return br_mdb_ip_get(mdb, &ip);
165 static void br_mdb_free(struct rcu_head *head)
167 struct net_bridge_mdb_htable *mdb =
168 container_of(head, struct net_bridge_mdb_htable, rcu);
169 struct net_bridge_mdb_htable *old = mdb->old;
176 static int br_mdb_copy(struct net_bridge_mdb_htable *new,
177 struct net_bridge_mdb_htable *old,
180 struct net_bridge_mdb_entry *mp;
181 struct hlist_node *p;
186 for (i = 0; i < old->max; i++)
187 hlist_for_each_entry(mp, p, &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, p, &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 &&
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;
386 iph->daddr = htonl(INADDR_ALLHOSTS_GROUP);
387 ((u8 *)&iph[1])[0] = IPOPT_RA;
388 ((u8 *)&iph[1])[1] = 4;
389 ((u8 *)&iph[1])[2] = 0;
390 ((u8 *)&iph[1])[3] = 0;
394 skb_set_transport_header(skb, skb->len);
396 ih->type = IGMP_HOST_MEMBERSHIP_QUERY;
397 ih->code = (group ? br->multicast_last_member_interval :
398 br->multicast_query_response_interval) /
399 (HZ / IGMP_TIMER_SCALE);
402 ih->csum = ip_compute_csum((void *)ih, sizeof(struct igmphdr));
403 skb_put(skb, sizeof(*ih));
405 __skb_pull(skb, sizeof(*eth));
411 #if IS_ENABLED(CONFIG_IPV6)
412 static struct sk_buff *br_ip6_multicast_alloc_query(struct net_bridge *br,
413 const struct in6_addr *group)
416 struct ipv6hdr *ip6h;
417 struct mld_msg *mldq;
420 unsigned long interval;
422 skb = netdev_alloc_skb_ip_align(br->dev, sizeof(*eth) + sizeof(*ip6h) +
427 skb->protocol = htons(ETH_P_IPV6);
429 /* Ethernet header */
430 skb_reset_mac_header(skb);
433 memcpy(eth->h_source, br->dev->dev_addr, 6);
434 eth->h_proto = htons(ETH_P_IPV6);
435 skb_put(skb, sizeof(*eth));
437 /* IPv6 header + HbH option */
438 skb_set_network_header(skb, skb->len);
439 ip6h = ipv6_hdr(skb);
441 *(__force __be32 *)ip6h = htonl(0x60000000);
442 ip6h->payload_len = htons(8 + sizeof(*mldq));
443 ip6h->nexthdr = IPPROTO_HOPOPTS;
445 ipv6_addr_set(&ip6h->daddr, htonl(0xff020000), 0, 0, htonl(1));
446 if (ipv6_dev_get_saddr(dev_net(br->dev), br->dev, &ip6h->daddr, 0,
451 ipv6_eth_mc_map(&ip6h->daddr, eth->h_dest);
453 hopopt = (u8 *)(ip6h + 1);
454 hopopt[0] = IPPROTO_ICMPV6; /* next hdr */
455 hopopt[1] = 0; /* length of HbH */
456 hopopt[2] = IPV6_TLV_ROUTERALERT; /* Router Alert */
457 hopopt[3] = 2; /* Length of RA Option */
458 hopopt[4] = 0; /* Type = 0x0000 (MLD) */
460 hopopt[6] = IPV6_TLV_PAD1; /* Pad1 */
461 hopopt[7] = IPV6_TLV_PAD1; /* Pad1 */
463 skb_put(skb, sizeof(*ip6h) + 8);
466 skb_set_transport_header(skb, skb->len);
467 mldq = (struct mld_msg *) icmp6_hdr(skb);
469 interval = ipv6_addr_any(group) ? br->multicast_last_member_interval :
470 br->multicast_query_response_interval;
472 mldq->mld_type = ICMPV6_MGM_QUERY;
475 mldq->mld_maxdelay = htons((u16)jiffies_to_msecs(interval));
476 mldq->mld_reserved = 0;
477 mldq->mld_mca = *group;
480 mldq->mld_cksum = csum_ipv6_magic(&ip6h->saddr, &ip6h->daddr,
481 sizeof(*mldq), IPPROTO_ICMPV6,
484 skb_put(skb, sizeof(*mldq));
486 __skb_pull(skb, sizeof(*eth));
493 static struct sk_buff *br_multicast_alloc_query(struct net_bridge *br,
496 switch (addr->proto) {
497 case htons(ETH_P_IP):
498 return br_ip4_multicast_alloc_query(br, addr->u.ip4);
499 #if IS_ENABLED(CONFIG_IPV6)
500 case htons(ETH_P_IPV6):
501 return br_ip6_multicast_alloc_query(br, &addr->u.ip6);
507 static struct net_bridge_mdb_entry *br_multicast_get_group(
508 struct net_bridge *br, struct net_bridge_port *port,
509 struct br_ip *group, int hash)
511 struct net_bridge_mdb_htable *mdb;
512 struct net_bridge_mdb_entry *mp;
513 struct hlist_node *p;
514 unsigned int count = 0;
519 mdb = rcu_dereference_protected(br->mdb, 1);
520 hlist_for_each_entry(mp, p, &mdb->mhash[hash], hlist[mdb->ver]) {
522 if (unlikely(br_ip_equal(group, &mp->addr)))
529 if (unlikely(count > br->hash_elasticity && count)) {
531 br_info(br, "Multicast hash table "
532 "chain limit reached: %s\n",
533 port ? port->dev->name : br->dev->name);
535 elasticity = br->hash_elasticity;
538 if (mdb->size >= max) {
540 if (unlikely(max > br->hash_max)) {
541 br_warn(br, "Multicast hash table maximum of %d "
542 "reached, disabling snooping: %s\n",
544 port ? port->dev->name : br->dev->name);
547 br->multicast_disabled = 1;
552 if (max > mdb->max || elasticity) {
555 br_info(br, "Multicast hash table "
557 port ? port->dev->name : br->dev->name);
562 err = br_mdb_rehash(&br->mdb, max, elasticity);
564 br_warn(br, "Cannot rehash multicast "
565 "hash table, disabling snooping: %s, %d, %d\n",
566 port ? port->dev->name : br->dev->name,
582 struct net_bridge_mdb_entry *br_multicast_new_group(struct net_bridge *br,
583 struct net_bridge_port *port, struct br_ip *group)
585 struct net_bridge_mdb_htable *mdb;
586 struct net_bridge_mdb_entry *mp;
590 mdb = rcu_dereference_protected(br->mdb, 1);
592 err = br_mdb_rehash(&br->mdb, BR_HASH_SIZE, 0);
598 hash = br_ip_hash(mdb, group);
599 mp = br_multicast_get_group(br, port, group, hash);
600 switch (PTR_ERR(mp)) {
606 mdb = rcu_dereference_protected(br->mdb, 1);
607 hash = br_ip_hash(mdb, group);
614 mp = kzalloc(sizeof(*mp), GFP_ATOMIC);
616 return ERR_PTR(-ENOMEM);
620 setup_timer(&mp->timer, br_multicast_group_expired,
623 hlist_add_head_rcu(&mp->hlist[mdb->ver], &mdb->mhash[hash]);
630 struct net_bridge_port_group *br_multicast_new_port_group(
631 struct net_bridge_port *port,
633 struct net_bridge_port_group __rcu *next,
636 struct net_bridge_port_group *p;
638 p = kzalloc(sizeof(*p), GFP_ATOMIC);
645 rcu_assign_pointer(p->next, next);
646 hlist_add_head(&p->mglist, &port->mglist);
647 setup_timer(&p->timer, br_multicast_port_group_expired,
652 static int br_multicast_add_group(struct net_bridge *br,
653 struct net_bridge_port *port,
656 struct net_bridge_mdb_entry *mp;
657 struct net_bridge_port_group *p;
658 struct net_bridge_port_group __rcu **pp;
659 unsigned long now = jiffies;
662 spin_lock(&br->multicast_lock);
663 if (!netif_running(br->dev) ||
664 (port && port->state == BR_STATE_DISABLED))
667 mp = br_multicast_new_group(br, port, group);
674 mod_timer(&mp->timer, now + br->multicast_membership_interval);
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);
694 mod_timer(&p->timer, now + br->multicast_membership_interval);
699 spin_unlock(&br->multicast_lock);
703 static int br_ip4_multicast_add_group(struct net_bridge *br,
704 struct net_bridge_port *port,
708 struct br_ip br_group;
710 if (ipv4_is_local_multicast(group))
713 br_group.u.ip4 = group;
714 br_group.proto = htons(ETH_P_IP);
717 return br_multicast_add_group(br, port, &br_group);
720 #if IS_ENABLED(CONFIG_IPV6)
721 static int br_ip6_multicast_add_group(struct net_bridge *br,
722 struct net_bridge_port *port,
723 const struct in6_addr *group,
726 struct br_ip br_group;
728 if (!ipv6_is_transient_multicast(group))
731 br_group.u.ip6 = *group;
732 br_group.proto = htons(ETH_P_IPV6);
735 return br_multicast_add_group(br, port, &br_group);
739 static void br_multicast_router_expired(unsigned long data)
741 struct net_bridge_port *port = (void *)data;
742 struct net_bridge *br = port->br;
744 spin_lock(&br->multicast_lock);
745 if (port->multicast_router != 1 ||
746 timer_pending(&port->multicast_router_timer) ||
747 hlist_unhashed(&port->rlist))
750 hlist_del_init_rcu(&port->rlist);
753 spin_unlock(&br->multicast_lock);
756 static void br_multicast_local_router_expired(unsigned long data)
760 static void br_multicast_querier_expired(unsigned long data)
762 struct net_bridge *br = (void *)data;
764 spin_lock(&br->multicast_lock);
765 if (!netif_running(br->dev) || br->multicast_disabled)
768 br_multicast_start_querier(br);
771 spin_unlock(&br->multicast_lock);
774 static void __br_multicast_send_query(struct net_bridge *br,
775 struct net_bridge_port *port,
780 skb = br_multicast_alloc_query(br, ip);
785 __skb_push(skb, sizeof(struct ethhdr));
786 skb->dev = port->dev;
787 NF_HOOK(NFPROTO_BRIDGE, NF_BR_LOCAL_OUT, skb, NULL, skb->dev,
793 static void br_multicast_send_query(struct net_bridge *br,
794 struct net_bridge_port *port, u32 sent)
797 struct br_ip br_group;
799 if (!netif_running(br->dev) || br->multicast_disabled ||
800 !br->multicast_querier ||
801 timer_pending(&br->multicast_querier_timer))
804 memset(&br_group.u, 0, sizeof(br_group.u));
806 br_group.proto = htons(ETH_P_IP);
807 __br_multicast_send_query(br, port, &br_group);
809 #if IS_ENABLED(CONFIG_IPV6)
810 br_group.proto = htons(ETH_P_IPV6);
811 __br_multicast_send_query(br, port, &br_group);
815 time += sent < br->multicast_startup_query_count ?
816 br->multicast_startup_query_interval :
817 br->multicast_query_interval;
818 mod_timer(port ? &port->multicast_query_timer :
819 &br->multicast_query_timer, time);
822 static void br_multicast_port_query_expired(unsigned long data)
824 struct net_bridge_port *port = (void *)data;
825 struct net_bridge *br = port->br;
827 spin_lock(&br->multicast_lock);
828 if (port->state == BR_STATE_DISABLED ||
829 port->state == BR_STATE_BLOCKING)
832 if (port->multicast_startup_queries_sent <
833 br->multicast_startup_query_count)
834 port->multicast_startup_queries_sent++;
836 br_multicast_send_query(port->br, port,
837 port->multicast_startup_queries_sent);
840 spin_unlock(&br->multicast_lock);
843 void br_multicast_add_port(struct net_bridge_port *port)
845 port->multicast_router = 1;
847 setup_timer(&port->multicast_router_timer, br_multicast_router_expired,
848 (unsigned long)port);
849 setup_timer(&port->multicast_query_timer,
850 br_multicast_port_query_expired, (unsigned long)port);
853 void br_multicast_del_port(struct net_bridge_port *port)
855 del_timer_sync(&port->multicast_router_timer);
858 static void __br_multicast_enable_port(struct net_bridge_port *port)
860 port->multicast_startup_queries_sent = 0;
862 if (try_to_del_timer_sync(&port->multicast_query_timer) >= 0 ||
863 del_timer(&port->multicast_query_timer))
864 mod_timer(&port->multicast_query_timer, jiffies);
867 void br_multicast_enable_port(struct net_bridge_port *port)
869 struct net_bridge *br = port->br;
871 spin_lock(&br->multicast_lock);
872 if (br->multicast_disabled || !netif_running(br->dev))
875 __br_multicast_enable_port(port);
878 spin_unlock(&br->multicast_lock);
881 void br_multicast_disable_port(struct net_bridge_port *port)
883 struct net_bridge *br = port->br;
884 struct net_bridge_port_group *pg;
885 struct hlist_node *p, *n;
887 spin_lock(&br->multicast_lock);
888 hlist_for_each_entry_safe(pg, p, n, &port->mglist, mglist)
889 br_multicast_del_pg(br, pg);
891 if (!hlist_unhashed(&port->rlist))
892 hlist_del_init_rcu(&port->rlist);
893 del_timer(&port->multicast_router_timer);
894 del_timer(&port->multicast_query_timer);
895 spin_unlock(&br->multicast_lock);
898 static int br_ip4_multicast_igmp3_report(struct net_bridge *br,
899 struct net_bridge_port *port,
902 struct igmpv3_report *ih;
903 struct igmpv3_grec *grec;
912 if (!pskb_may_pull(skb, sizeof(*ih)))
915 br_vlan_get_tag(skb, &vid);
916 ih = igmpv3_report_hdr(skb);
917 num = ntohs(ih->ngrec);
920 for (i = 0; i < num; i++) {
921 len += sizeof(*grec);
922 if (!pskb_may_pull(skb, len))
925 grec = (void *)(skb->data + len - sizeof(*grec));
926 group = grec->grec_mca;
927 type = grec->grec_type;
929 len += ntohs(grec->grec_nsrcs) * 4;
930 if (!pskb_may_pull(skb, len))
933 /* We treat this as an IGMPv2 report for now. */
935 case IGMPV3_MODE_IS_INCLUDE:
936 case IGMPV3_MODE_IS_EXCLUDE:
937 case IGMPV3_CHANGE_TO_INCLUDE:
938 case IGMPV3_CHANGE_TO_EXCLUDE:
939 case IGMPV3_ALLOW_NEW_SOURCES:
940 case IGMPV3_BLOCK_OLD_SOURCES:
947 err = br_ip4_multicast_add_group(br, port, group, vid);
955 #if IS_ENABLED(CONFIG_IPV6)
956 static int br_ip6_multicast_mld2_report(struct net_bridge *br,
957 struct net_bridge_port *port,
960 struct icmp6hdr *icmp6h;
961 struct mld2_grec *grec;
968 if (!pskb_may_pull(skb, sizeof(*icmp6h)))
971 br_vlan_get_tag(skb, &vid);
972 icmp6h = icmp6_hdr(skb);
973 num = ntohs(icmp6h->icmp6_dataun.un_data16[1]);
974 len = sizeof(*icmp6h);
976 for (i = 0; i < num; i++) {
977 __be16 *nsrcs, _nsrcs;
979 nsrcs = skb_header_pointer(skb,
980 len + offsetof(struct mld2_grec,
982 sizeof(_nsrcs), &_nsrcs);
986 if (!pskb_may_pull(skb,
987 len + sizeof(*grec) +
988 sizeof(struct in6_addr) * ntohs(*nsrcs)))
991 grec = (struct mld2_grec *)(skb->data + len);
992 len += sizeof(*grec) +
993 sizeof(struct in6_addr) * ntohs(*nsrcs);
995 /* We treat these as MLDv1 reports for now. */
996 switch (grec->grec_type) {
997 case MLD2_MODE_IS_INCLUDE:
998 case MLD2_MODE_IS_EXCLUDE:
999 case MLD2_CHANGE_TO_INCLUDE:
1000 case MLD2_CHANGE_TO_EXCLUDE:
1001 case MLD2_ALLOW_NEW_SOURCES:
1002 case MLD2_BLOCK_OLD_SOURCES:
1009 err = br_ip6_multicast_add_group(br, port, &grec->grec_mca,
1020 * Add port to rotuer_list
1021 * list is maintained ordered by pointer value
1022 * and locked by br->multicast_lock and RCU
1024 static void br_multicast_add_router(struct net_bridge *br,
1025 struct net_bridge_port *port)
1027 struct net_bridge_port *p;
1028 struct hlist_node *n, *slot = NULL;
1030 hlist_for_each_entry(p, n, &br->router_list, rlist) {
1031 if ((unsigned long) port >= (unsigned long) p)
1037 hlist_add_after_rcu(slot, &port->rlist);
1039 hlist_add_head_rcu(&port->rlist, &br->router_list);
1042 static void br_multicast_mark_router(struct net_bridge *br,
1043 struct net_bridge_port *port)
1045 unsigned long now = jiffies;
1048 if (br->multicast_router == 1)
1049 mod_timer(&br->multicast_router_timer,
1050 now + br->multicast_querier_interval);
1054 if (port->multicast_router != 1)
1057 if (!hlist_unhashed(&port->rlist))
1060 br_multicast_add_router(br, port);
1063 mod_timer(&port->multicast_router_timer,
1064 now + br->multicast_querier_interval);
1067 static void br_multicast_query_received(struct net_bridge *br,
1068 struct net_bridge_port *port,
1072 mod_timer(&br->multicast_querier_timer,
1073 jiffies + br->multicast_querier_interval);
1074 else if (timer_pending(&br->multicast_querier_timer))
1077 br_multicast_mark_router(br, port);
1080 static int br_ip4_multicast_query(struct net_bridge *br,
1081 struct net_bridge_port *port,
1082 struct sk_buff *skb)
1084 const struct iphdr *iph = ip_hdr(skb);
1085 struct igmphdr *ih = igmp_hdr(skb);
1086 struct net_bridge_mdb_entry *mp;
1087 struct igmpv3_query *ih3;
1088 struct net_bridge_port_group *p;
1089 struct net_bridge_port_group __rcu **pp;
1090 unsigned long max_delay;
1091 unsigned long now = jiffies;
1096 spin_lock(&br->multicast_lock);
1097 if (!netif_running(br->dev) ||
1098 (port && port->state == BR_STATE_DISABLED))
1101 br_multicast_query_received(br, port, !!iph->saddr);
1105 if (skb->len == sizeof(*ih)) {
1106 max_delay = ih->code * (HZ / IGMP_TIMER_SCALE);
1109 max_delay = 10 * HZ;
1113 if (!pskb_may_pull(skb, sizeof(struct igmpv3_query))) {
1118 ih3 = igmpv3_query_hdr(skb);
1122 max_delay = ih3->code ?
1123 IGMPV3_MRC(ih3->code) * (HZ / IGMP_TIMER_SCALE) : 1;
1129 br_vlan_get_tag(skb, &vid);
1130 mp = br_mdb_ip4_get(mlock_dereference(br->mdb, br), group, vid);
1134 max_delay *= br->multicast_last_member_count;
1137 (timer_pending(&mp->timer) ?
1138 time_after(mp->timer.expires, now + max_delay) :
1139 try_to_del_timer_sync(&mp->timer) >= 0))
1140 mod_timer(&mp->timer, now + max_delay);
1142 for (pp = &mp->ports;
1143 (p = mlock_dereference(*pp, br)) != NULL;
1145 if (timer_pending(&p->timer) ?
1146 time_after(p->timer.expires, now + max_delay) :
1147 try_to_del_timer_sync(&p->timer) >= 0)
1148 mod_timer(&p->timer, now + max_delay);
1152 spin_unlock(&br->multicast_lock);
1156 #if IS_ENABLED(CONFIG_IPV6)
1157 static int br_ip6_multicast_query(struct net_bridge *br,
1158 struct net_bridge_port *port,
1159 struct sk_buff *skb)
1161 const struct ipv6hdr *ip6h = ipv6_hdr(skb);
1162 struct mld_msg *mld;
1163 struct net_bridge_mdb_entry *mp;
1164 struct mld2_query *mld2q;
1165 struct net_bridge_port_group *p;
1166 struct net_bridge_port_group __rcu **pp;
1167 unsigned long max_delay;
1168 unsigned long now = jiffies;
1169 const struct in6_addr *group = NULL;
1173 spin_lock(&br->multicast_lock);
1174 if (!netif_running(br->dev) ||
1175 (port && port->state == BR_STATE_DISABLED))
1178 br_multicast_query_received(br, port, !ipv6_addr_any(&ip6h->saddr));
1180 if (skb->len == sizeof(*mld)) {
1181 if (!pskb_may_pull(skb, sizeof(*mld))) {
1185 mld = (struct mld_msg *) icmp6_hdr(skb);
1186 max_delay = msecs_to_jiffies(ntohs(mld->mld_maxdelay));
1188 group = &mld->mld_mca;
1189 } else if (skb->len >= sizeof(*mld2q)) {
1190 if (!pskb_may_pull(skb, sizeof(*mld2q))) {
1194 mld2q = (struct mld2_query *)icmp6_hdr(skb);
1195 if (!mld2q->mld2q_nsrcs)
1196 group = &mld2q->mld2q_mca;
1197 max_delay = mld2q->mld2q_mrc ? MLDV2_MRC(ntohs(mld2q->mld2q_mrc)) : 1;
1203 br_vlan_get_tag(skb, &vid);
1204 mp = br_mdb_ip6_get(mlock_dereference(br->mdb, br), group, vid);
1208 max_delay *= br->multicast_last_member_count;
1210 (timer_pending(&mp->timer) ?
1211 time_after(mp->timer.expires, now + max_delay) :
1212 try_to_del_timer_sync(&mp->timer) >= 0))
1213 mod_timer(&mp->timer, now + max_delay);
1215 for (pp = &mp->ports;
1216 (p = mlock_dereference(*pp, br)) != NULL;
1218 if (timer_pending(&p->timer) ?
1219 time_after(p->timer.expires, now + max_delay) :
1220 try_to_del_timer_sync(&p->timer) >= 0)
1221 mod_timer(&p->timer, now + max_delay);
1225 spin_unlock(&br->multicast_lock);
1230 static void br_multicast_leave_group(struct net_bridge *br,
1231 struct net_bridge_port *port,
1232 struct br_ip *group)
1234 struct net_bridge_mdb_htable *mdb;
1235 struct net_bridge_mdb_entry *mp;
1236 struct net_bridge_port_group *p;
1240 spin_lock(&br->multicast_lock);
1241 if (!netif_running(br->dev) ||
1242 (port && port->state == BR_STATE_DISABLED) ||
1243 timer_pending(&br->multicast_querier_timer))
1246 mdb = mlock_dereference(br->mdb, br);
1247 mp = br_mdb_ip_get(mdb, group);
1251 if (port && (port->flags & BR_MULTICAST_FAST_LEAVE)) {
1252 struct net_bridge_port_group __rcu **pp;
1254 for (pp = &mp->ports;
1255 (p = mlock_dereference(*pp, br)) != NULL;
1257 if (p->port != port)
1260 rcu_assign_pointer(*pp, p->next);
1261 hlist_del_init(&p->mglist);
1262 del_timer(&p->timer);
1263 call_rcu_bh(&p->rcu, br_multicast_free_pg);
1264 br_mdb_notify(br->dev, port, group, RTM_DELMDB);
1266 if (!mp->ports && !mp->mglist &&
1267 netif_running(br->dev))
1268 mod_timer(&mp->timer, jiffies);
1274 time = now + br->multicast_last_member_count *
1275 br->multicast_last_member_interval;
1279 (timer_pending(&mp->timer) ?
1280 time_after(mp->timer.expires, time) :
1281 try_to_del_timer_sync(&mp->timer) >= 0)) {
1282 mod_timer(&mp->timer, time);
1288 for (p = mlock_dereference(mp->ports, br);
1290 p = mlock_dereference(p->next, br)) {
1291 if (p->port != port)
1294 if (!hlist_unhashed(&p->mglist) &&
1295 (timer_pending(&p->timer) ?
1296 time_after(p->timer.expires, time) :
1297 try_to_del_timer_sync(&p->timer) >= 0)) {
1298 mod_timer(&p->timer, time);
1305 spin_unlock(&br->multicast_lock);
1308 static void br_ip4_multicast_leave_group(struct net_bridge *br,
1309 struct net_bridge_port *port,
1313 struct br_ip br_group;
1315 if (ipv4_is_local_multicast(group))
1318 br_group.u.ip4 = group;
1319 br_group.proto = htons(ETH_P_IP);
1322 br_multicast_leave_group(br, port, &br_group);
1325 #if IS_ENABLED(CONFIG_IPV6)
1326 static void br_ip6_multicast_leave_group(struct net_bridge *br,
1327 struct net_bridge_port *port,
1328 const struct in6_addr *group,
1331 struct br_ip br_group;
1333 if (!ipv6_is_transient_multicast(group))
1336 br_group.u.ip6 = *group;
1337 br_group.proto = htons(ETH_P_IPV6);
1340 br_multicast_leave_group(br, port, &br_group);
1344 static int br_multicast_ipv4_rcv(struct net_bridge *br,
1345 struct net_bridge_port *port,
1346 struct sk_buff *skb)
1348 struct sk_buff *skb2 = skb;
1349 const struct iphdr *iph;
1352 unsigned int offset;
1356 /* We treat OOM as packet loss for now. */
1357 if (!pskb_may_pull(skb, sizeof(*iph)))
1362 if (iph->ihl < 5 || iph->version != 4)
1365 if (!pskb_may_pull(skb, ip_hdrlen(skb)))
1370 if (unlikely(ip_fast_csum((u8 *)iph, iph->ihl)))
1373 if (iph->protocol != IPPROTO_IGMP) {
1374 if ((iph->daddr & IGMP_LOCAL_GROUP_MASK) != IGMP_LOCAL_GROUP)
1375 BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1379 len = ntohs(iph->tot_len);
1380 if (skb->len < len || len < ip_hdrlen(skb))
1383 if (skb->len > len) {
1384 skb2 = skb_clone(skb, GFP_ATOMIC);
1388 err = pskb_trim_rcsum(skb2, len);
1393 len -= ip_hdrlen(skb2);
1394 offset = skb_network_offset(skb2) + ip_hdrlen(skb2);
1395 __skb_pull(skb2, offset);
1396 skb_reset_transport_header(skb2);
1399 if (!pskb_may_pull(skb2, sizeof(*ih)))
1402 switch (skb2->ip_summed) {
1403 case CHECKSUM_COMPLETE:
1404 if (!csum_fold(skb2->csum))
1409 if (skb_checksum_complete(skb2))
1415 br_vlan_get_tag(skb2, &vid);
1416 BR_INPUT_SKB_CB(skb)->igmp = 1;
1417 ih = igmp_hdr(skb2);
1420 case IGMP_HOST_MEMBERSHIP_REPORT:
1421 case IGMPV2_HOST_MEMBERSHIP_REPORT:
1422 BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1423 err = br_ip4_multicast_add_group(br, port, ih->group, vid);
1425 case IGMPV3_HOST_MEMBERSHIP_REPORT:
1426 err = br_ip4_multicast_igmp3_report(br, port, skb2);
1428 case IGMP_HOST_MEMBERSHIP_QUERY:
1429 err = br_ip4_multicast_query(br, port, skb2);
1431 case IGMP_HOST_LEAVE_MESSAGE:
1432 br_ip4_multicast_leave_group(br, port, ih->group, vid);
1437 __skb_push(skb2, offset);
1444 #if IS_ENABLED(CONFIG_IPV6)
1445 static int br_multicast_ipv6_rcv(struct net_bridge *br,
1446 struct net_bridge_port *port,
1447 struct sk_buff *skb)
1449 struct sk_buff *skb2;
1450 const struct ipv6hdr *ip6h;
1459 if (!pskb_may_pull(skb, sizeof(*ip6h)))
1462 ip6h = ipv6_hdr(skb);
1465 * We're interested in MLD messages only.
1467 * - MLD has always Router Alert hop-by-hop option
1468 * - But we do not support jumbrograms.
1470 if (ip6h->version != 6 ||
1471 ip6h->nexthdr != IPPROTO_HOPOPTS ||
1472 ip6h->payload_len == 0)
1475 len = ntohs(ip6h->payload_len) + sizeof(*ip6h);
1479 nexthdr = ip6h->nexthdr;
1480 offset = ipv6_skip_exthdr(skb, sizeof(*ip6h), &nexthdr, &frag_off);
1482 if (offset < 0 || nexthdr != IPPROTO_ICMPV6)
1485 /* Okay, we found ICMPv6 header */
1486 skb2 = skb_clone(skb, GFP_ATOMIC);
1491 if (!pskb_may_pull(skb2, offset + sizeof(struct icmp6hdr)))
1494 len -= offset - skb_network_offset(skb2);
1496 __skb_pull(skb2, offset);
1497 skb_reset_transport_header(skb2);
1498 skb_postpull_rcsum(skb2, skb_network_header(skb2),
1499 skb_network_header_len(skb2));
1501 icmp6_type = icmp6_hdr(skb2)->icmp6_type;
1503 switch (icmp6_type) {
1504 case ICMPV6_MGM_QUERY:
1505 case ICMPV6_MGM_REPORT:
1506 case ICMPV6_MGM_REDUCTION:
1507 case ICMPV6_MLD2_REPORT:
1514 /* Okay, we found MLD message. Check further. */
1515 if (skb2->len > len) {
1516 err = pskb_trim_rcsum(skb2, len);
1522 ip6h = ipv6_hdr(skb2);
1524 switch (skb2->ip_summed) {
1525 case CHECKSUM_COMPLETE:
1526 if (!csum_ipv6_magic(&ip6h->saddr, &ip6h->daddr, skb2->len,
1527 IPPROTO_ICMPV6, skb2->csum))
1531 skb2->csum = ~csum_unfold(csum_ipv6_magic(&ip6h->saddr,
1534 IPPROTO_ICMPV6, 0));
1535 if (__skb_checksum_complete(skb2))
1541 br_vlan_get_tag(skb, &vid);
1542 BR_INPUT_SKB_CB(skb)->igmp = 1;
1544 switch (icmp6_type) {
1545 case ICMPV6_MGM_REPORT:
1547 struct mld_msg *mld;
1548 if (!pskb_may_pull(skb2, sizeof(*mld))) {
1552 mld = (struct mld_msg *)skb_transport_header(skb2);
1553 BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1554 err = br_ip6_multicast_add_group(br, port, &mld->mld_mca, vid);
1557 case ICMPV6_MLD2_REPORT:
1558 err = br_ip6_multicast_mld2_report(br, port, skb2);
1560 case ICMPV6_MGM_QUERY:
1561 err = br_ip6_multicast_query(br, port, skb2);
1563 case ICMPV6_MGM_REDUCTION:
1565 struct mld_msg *mld;
1566 if (!pskb_may_pull(skb2, sizeof(*mld))) {
1570 mld = (struct mld_msg *)skb_transport_header(skb2);
1571 br_ip6_multicast_leave_group(br, port, &mld->mld_mca, vid);
1581 int br_multicast_rcv(struct net_bridge *br, struct net_bridge_port *port,
1582 struct sk_buff *skb)
1584 BR_INPUT_SKB_CB(skb)->igmp = 0;
1585 BR_INPUT_SKB_CB(skb)->mrouters_only = 0;
1587 if (br->multicast_disabled)
1590 switch (skb->protocol) {
1591 case htons(ETH_P_IP):
1592 return br_multicast_ipv4_rcv(br, port, skb);
1593 #if IS_ENABLED(CONFIG_IPV6)
1594 case htons(ETH_P_IPV6):
1595 return br_multicast_ipv6_rcv(br, port, skb);
1602 static void br_multicast_query_expired(unsigned long data)
1604 struct net_bridge *br = (void *)data;
1606 spin_lock(&br->multicast_lock);
1607 if (br->multicast_startup_queries_sent <
1608 br->multicast_startup_query_count)
1609 br->multicast_startup_queries_sent++;
1611 br_multicast_send_query(br, NULL, br->multicast_startup_queries_sent);
1613 spin_unlock(&br->multicast_lock);
1616 void br_multicast_init(struct net_bridge *br)
1618 br->hash_elasticity = 4;
1621 br->multicast_router = 1;
1622 br->multicast_querier = 0;
1623 br->multicast_last_member_count = 2;
1624 br->multicast_startup_query_count = 2;
1626 br->multicast_last_member_interval = HZ;
1627 br->multicast_query_response_interval = 10 * HZ;
1628 br->multicast_startup_query_interval = 125 * HZ / 4;
1629 br->multicast_query_interval = 125 * HZ;
1630 br->multicast_querier_interval = 255 * HZ;
1631 br->multicast_membership_interval = 260 * HZ;
1633 spin_lock_init(&br->multicast_lock);
1634 setup_timer(&br->multicast_router_timer,
1635 br_multicast_local_router_expired, 0);
1636 setup_timer(&br->multicast_querier_timer,
1637 br_multicast_querier_expired, (unsigned long)br);
1638 setup_timer(&br->multicast_query_timer, br_multicast_query_expired,
1642 void br_multicast_open(struct net_bridge *br)
1644 br->multicast_startup_queries_sent = 0;
1646 if (br->multicast_disabled)
1649 mod_timer(&br->multicast_query_timer, jiffies);
1652 void br_multicast_stop(struct net_bridge *br)
1654 struct net_bridge_mdb_htable *mdb;
1655 struct net_bridge_mdb_entry *mp;
1656 struct hlist_node *p, *n;
1660 del_timer_sync(&br->multicast_router_timer);
1661 del_timer_sync(&br->multicast_querier_timer);
1662 del_timer_sync(&br->multicast_query_timer);
1664 spin_lock_bh(&br->multicast_lock);
1665 mdb = mlock_dereference(br->mdb, br);
1672 for (i = 0; i < mdb->max; i++) {
1673 hlist_for_each_entry_safe(mp, p, n, &mdb->mhash[i],
1675 del_timer(&mp->timer);
1676 call_rcu_bh(&mp->rcu, br_multicast_free_group);
1681 spin_unlock_bh(&br->multicast_lock);
1683 spin_lock_bh(&br->multicast_lock);
1688 call_rcu_bh(&mdb->rcu, br_mdb_free);
1691 spin_unlock_bh(&br->multicast_lock);
1694 int br_multicast_set_router(struct net_bridge *br, unsigned long val)
1698 spin_lock_bh(&br->multicast_lock);
1699 if (!netif_running(br->dev))
1705 del_timer(&br->multicast_router_timer);
1708 br->multicast_router = val;
1718 spin_unlock_bh(&br->multicast_lock);
1723 int br_multicast_set_port_router(struct net_bridge_port *p, unsigned long val)
1725 struct net_bridge *br = p->br;
1728 spin_lock(&br->multicast_lock);
1729 if (!netif_running(br->dev) || p->state == BR_STATE_DISABLED)
1736 p->multicast_router = val;
1739 if (val < 2 && !hlist_unhashed(&p->rlist))
1740 hlist_del_init_rcu(&p->rlist);
1745 del_timer(&p->multicast_router_timer);
1750 br_multicast_add_router(br, p);
1759 spin_unlock(&br->multicast_lock);
1764 static void br_multicast_start_querier(struct net_bridge *br)
1766 struct net_bridge_port *port;
1768 br_multicast_open(br);
1770 list_for_each_entry(port, &br->port_list, list) {
1771 if (port->state == BR_STATE_DISABLED ||
1772 port->state == BR_STATE_BLOCKING)
1775 __br_multicast_enable_port(port);
1779 int br_multicast_toggle(struct net_bridge *br, unsigned long val)
1782 struct net_bridge_mdb_htable *mdb;
1784 spin_lock_bh(&br->multicast_lock);
1785 if (br->multicast_disabled == !val)
1788 br->multicast_disabled = !val;
1789 if (br->multicast_disabled)
1792 if (!netif_running(br->dev))
1795 mdb = mlock_dereference(br->mdb, br);
1800 br->multicast_disabled = !!val;
1804 err = br_mdb_rehash(&br->mdb, mdb->max,
1805 br->hash_elasticity);
1810 br_multicast_start_querier(br);
1813 spin_unlock_bh(&br->multicast_lock);
1818 int br_multicast_set_querier(struct net_bridge *br, unsigned long val)
1822 spin_lock_bh(&br->multicast_lock);
1823 if (br->multicast_querier == val)
1826 br->multicast_querier = val;
1828 br_multicast_start_querier(br);
1831 spin_unlock_bh(&br->multicast_lock);
1836 int br_multicast_set_hash_max(struct net_bridge *br, unsigned long val)
1840 struct net_bridge_mdb_htable *mdb;
1842 spin_lock(&br->multicast_lock);
1843 if (!netif_running(br->dev))
1847 if (!is_power_of_2(val))
1850 mdb = mlock_dereference(br->mdb, br);
1851 if (mdb && val < mdb->size)
1867 err = br_mdb_rehash(&br->mdb, br->hash_max,
1868 br->hash_elasticity);
1874 spin_unlock(&br->multicast_lock);