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/addrconf.h>
31 #include <net/ip6_checksum.h>
34 #include "br_private.h"
36 #define mlock_dereference(X, br) \
37 rcu_dereference_protected(X, lockdep_is_held(&br->multicast_lock))
39 static void br_multicast_start_querier(struct net_bridge *br);
41 #if IS_ENABLED(CONFIG_IPV6)
42 static inline int ipv6_is_transient_multicast(const struct in6_addr *addr)
44 if (ipv6_addr_is_multicast(addr) && IPV6_ADDR_MC_FLAG_TRANSIENT(addr))
50 static inline int br_ip_equal(const struct br_ip *a, const struct br_ip *b)
52 if (a->proto != b->proto)
56 return a->u.ip4 == b->u.ip4;
57 #if IS_ENABLED(CONFIG_IPV6)
58 case htons(ETH_P_IPV6):
59 return ipv6_addr_equal(&a->u.ip6, &b->u.ip6);
65 static inline int __br_ip4_hash(struct net_bridge_mdb_htable *mdb, __be32 ip)
67 return jhash_1word(mdb->secret, (__force u32)ip) & (mdb->max - 1);
70 #if IS_ENABLED(CONFIG_IPV6)
71 static inline int __br_ip6_hash(struct net_bridge_mdb_htable *mdb,
72 const struct in6_addr *ip)
74 return jhash2((__force u32 *)ip->s6_addr32, 4, mdb->secret) & (mdb->max - 1);
78 static inline int br_ip_hash(struct net_bridge_mdb_htable *mdb,
83 return __br_ip4_hash(mdb, ip->u.ip4);
84 #if IS_ENABLED(CONFIG_IPV6)
85 case htons(ETH_P_IPV6):
86 return __br_ip6_hash(mdb, &ip->u.ip6);
92 static struct net_bridge_mdb_entry *__br_mdb_ip_get(
93 struct net_bridge_mdb_htable *mdb, struct br_ip *dst, int hash)
95 struct net_bridge_mdb_entry *mp;
98 hlist_for_each_entry_rcu(mp, p, &mdb->mhash[hash], hlist[mdb->ver]) {
99 if (br_ip_equal(&mp->addr, dst))
106 static struct net_bridge_mdb_entry *br_mdb_ip_get(
107 struct net_bridge_mdb_htable *mdb, struct br_ip *dst)
112 return __br_mdb_ip_get(mdb, dst, br_ip_hash(mdb, dst));
115 static struct net_bridge_mdb_entry *br_mdb_ip4_get(
116 struct net_bridge_mdb_htable *mdb, __be32 dst)
121 br_dst.proto = htons(ETH_P_IP);
123 return br_mdb_ip_get(mdb, &br_dst);
126 #if IS_ENABLED(CONFIG_IPV6)
127 static struct net_bridge_mdb_entry *br_mdb_ip6_get(
128 struct net_bridge_mdb_htable *mdb, const struct in6_addr *dst)
133 br_dst.proto = htons(ETH_P_IPV6);
135 return br_mdb_ip_get(mdb, &br_dst);
139 struct net_bridge_mdb_entry *br_mdb_get(struct net_bridge *br,
142 struct net_bridge_mdb_htable *mdb = rcu_dereference(br->mdb);
145 if (br->multicast_disabled)
148 if (BR_INPUT_SKB_CB(skb)->igmp)
151 ip.proto = skb->protocol;
153 switch (skb->protocol) {
154 case htons(ETH_P_IP):
155 ip.u.ip4 = ip_hdr(skb)->daddr;
157 #if IS_ENABLED(CONFIG_IPV6)
158 case htons(ETH_P_IPV6):
159 ip.u.ip6 = ipv6_hdr(skb)->daddr;
166 return br_mdb_ip_get(mdb, &ip);
169 static void br_mdb_free(struct rcu_head *head)
171 struct net_bridge_mdb_htable *mdb =
172 container_of(head, struct net_bridge_mdb_htable, rcu);
173 struct net_bridge_mdb_htable *old = mdb->old;
180 static int br_mdb_copy(struct net_bridge_mdb_htable *new,
181 struct net_bridge_mdb_htable *old,
184 struct net_bridge_mdb_entry *mp;
185 struct hlist_node *p;
190 for (i = 0; i < old->max; i++)
191 hlist_for_each_entry(mp, p, &old->mhash[i], hlist[old->ver])
192 hlist_add_head(&mp->hlist[new->ver],
193 &new->mhash[br_ip_hash(new, &mp->addr)]);
199 for (i = 0; i < new->max; i++) {
201 hlist_for_each_entry(mp, p, &new->mhash[i], hlist[new->ver])
207 return maxlen > elasticity ? -EINVAL : 0;
210 static void br_multicast_free_pg(struct rcu_head *head)
212 struct net_bridge_port_group *p =
213 container_of(head, struct net_bridge_port_group, rcu);
218 static void br_multicast_free_group(struct rcu_head *head)
220 struct net_bridge_mdb_entry *mp =
221 container_of(head, struct net_bridge_mdb_entry, rcu);
226 static void br_multicast_group_expired(unsigned long data)
228 struct net_bridge_mdb_entry *mp = (void *)data;
229 struct net_bridge *br = mp->br;
230 struct net_bridge_mdb_htable *mdb;
232 spin_lock(&br->multicast_lock);
233 if (!netif_running(br->dev) || timer_pending(&mp->timer))
241 mdb = mlock_dereference(br->mdb, br);
243 hlist_del_rcu(&mp->hlist[mdb->ver]);
246 call_rcu_bh(&mp->rcu, br_multicast_free_group);
249 spin_unlock(&br->multicast_lock);
252 static void br_multicast_del_pg(struct net_bridge *br,
253 struct net_bridge_port_group *pg)
255 struct net_bridge_mdb_htable *mdb;
256 struct net_bridge_mdb_entry *mp;
257 struct net_bridge_port_group *p;
258 struct net_bridge_port_group __rcu **pp;
260 mdb = mlock_dereference(br->mdb, br);
262 mp = br_mdb_ip_get(mdb, &pg->addr);
266 for (pp = &mp->ports;
267 (p = mlock_dereference(*pp, br)) != NULL;
272 rcu_assign_pointer(*pp, p->next);
273 hlist_del_init(&p->mglist);
274 del_timer(&p->timer);
275 call_rcu_bh(&p->rcu, br_multicast_free_pg);
277 if (!mp->ports && !mp->mglist &&
278 netif_running(br->dev))
279 mod_timer(&mp->timer, jiffies);
287 static void br_multicast_port_group_expired(unsigned long data)
289 struct net_bridge_port_group *pg = (void *)data;
290 struct net_bridge *br = pg->port->br;
292 spin_lock(&br->multicast_lock);
293 if (!netif_running(br->dev) || timer_pending(&pg->timer) ||
294 hlist_unhashed(&pg->mglist))
297 br_multicast_del_pg(br, pg);
300 spin_unlock(&br->multicast_lock);
303 static int br_mdb_rehash(struct net_bridge_mdb_htable __rcu **mdbp, int max,
306 struct net_bridge_mdb_htable *old = rcu_dereference_protected(*mdbp, 1);
307 struct net_bridge_mdb_htable *mdb;
310 mdb = kmalloc(sizeof(*mdb), GFP_ATOMIC);
317 mdb->mhash = kzalloc(max * sizeof(*mdb->mhash), GFP_ATOMIC);
323 mdb->size = old ? old->size : 0;
324 mdb->ver = old ? old->ver ^ 1 : 0;
326 if (!old || elasticity)
327 get_random_bytes(&mdb->secret, sizeof(mdb->secret));
329 mdb->secret = old->secret;
334 err = br_mdb_copy(mdb, old, elasticity);
341 call_rcu_bh(&mdb->rcu, br_mdb_free);
344 rcu_assign_pointer(*mdbp, mdb);
349 static struct sk_buff *br_ip4_multicast_alloc_query(struct net_bridge *br,
357 skb = netdev_alloc_skb_ip_align(br->dev, sizeof(*eth) + sizeof(*iph) +
362 skb->protocol = htons(ETH_P_IP);
364 skb_reset_mac_header(skb);
367 memcpy(eth->h_source, br->dev->dev_addr, 6);
370 eth->h_dest[2] = 0x5e;
374 eth->h_proto = htons(ETH_P_IP);
375 skb_put(skb, sizeof(*eth));
377 skb_set_network_header(skb, skb->len);
383 iph->tot_len = htons(sizeof(*iph) + sizeof(*ih) + 4);
385 iph->frag_off = htons(IP_DF);
387 iph->protocol = IPPROTO_IGMP;
389 iph->daddr = htonl(INADDR_ALLHOSTS_GROUP);
390 ((u8 *)&iph[1])[0] = IPOPT_RA;
391 ((u8 *)&iph[1])[1] = 4;
392 ((u8 *)&iph[1])[2] = 0;
393 ((u8 *)&iph[1])[3] = 0;
397 skb_set_transport_header(skb, skb->len);
399 ih->type = IGMP_HOST_MEMBERSHIP_QUERY;
400 ih->code = (group ? br->multicast_last_member_interval :
401 br->multicast_query_response_interval) /
402 (HZ / IGMP_TIMER_SCALE);
405 ih->csum = ip_compute_csum((void *)ih, sizeof(struct igmphdr));
406 skb_put(skb, sizeof(*ih));
408 __skb_pull(skb, sizeof(*eth));
414 #if IS_ENABLED(CONFIG_IPV6)
415 static struct sk_buff *br_ip6_multicast_alloc_query(struct net_bridge *br,
416 const struct in6_addr *group)
419 struct ipv6hdr *ip6h;
420 struct mld_msg *mldq;
423 unsigned long interval;
425 skb = netdev_alloc_skb_ip_align(br->dev, sizeof(*eth) + sizeof(*ip6h) +
430 skb->protocol = htons(ETH_P_IPV6);
432 /* Ethernet header */
433 skb_reset_mac_header(skb);
436 memcpy(eth->h_source, br->dev->dev_addr, 6);
437 eth->h_proto = htons(ETH_P_IPV6);
438 skb_put(skb, sizeof(*eth));
440 /* IPv6 header + HbH option */
441 skb_set_network_header(skb, skb->len);
442 ip6h = ipv6_hdr(skb);
444 *(__force __be32 *)ip6h = htonl(0x60000000);
445 ip6h->payload_len = htons(8 + sizeof(*mldq));
446 ip6h->nexthdr = IPPROTO_HOPOPTS;
448 ipv6_addr_set(&ip6h->daddr, htonl(0xff020000), 0, 0, htonl(1));
449 if (ipv6_dev_get_saddr(dev_net(br->dev), br->dev, &ip6h->daddr, 0,
454 ipv6_eth_mc_map(&ip6h->daddr, eth->h_dest);
456 hopopt = (u8 *)(ip6h + 1);
457 hopopt[0] = IPPROTO_ICMPV6; /* next hdr */
458 hopopt[1] = 0; /* length of HbH */
459 hopopt[2] = IPV6_TLV_ROUTERALERT; /* Router Alert */
460 hopopt[3] = 2; /* Length of RA Option */
461 hopopt[4] = 0; /* Type = 0x0000 (MLD) */
463 hopopt[6] = IPV6_TLV_PAD1; /* Pad1 */
464 hopopt[7] = IPV6_TLV_PAD1; /* Pad1 */
466 skb_put(skb, sizeof(*ip6h) + 8);
469 skb_set_transport_header(skb, skb->len);
470 mldq = (struct mld_msg *) icmp6_hdr(skb);
472 interval = ipv6_addr_any(group) ? br->multicast_last_member_interval :
473 br->multicast_query_response_interval;
475 mldq->mld_type = ICMPV6_MGM_QUERY;
478 mldq->mld_maxdelay = htons((u16)jiffies_to_msecs(interval));
479 mldq->mld_reserved = 0;
480 mldq->mld_mca = *group;
483 mldq->mld_cksum = csum_ipv6_magic(&ip6h->saddr, &ip6h->daddr,
484 sizeof(*mldq), IPPROTO_ICMPV6,
487 skb_put(skb, sizeof(*mldq));
489 __skb_pull(skb, sizeof(*eth));
496 static struct sk_buff *br_multicast_alloc_query(struct net_bridge *br,
499 switch (addr->proto) {
500 case htons(ETH_P_IP):
501 return br_ip4_multicast_alloc_query(br, addr->u.ip4);
502 #if IS_ENABLED(CONFIG_IPV6)
503 case htons(ETH_P_IPV6):
504 return br_ip6_multicast_alloc_query(br, &addr->u.ip6);
510 static struct net_bridge_mdb_entry *br_multicast_get_group(
511 struct net_bridge *br, struct net_bridge_port *port,
512 struct br_ip *group, int hash)
514 struct net_bridge_mdb_htable *mdb;
515 struct net_bridge_mdb_entry *mp;
516 struct hlist_node *p;
517 unsigned int count = 0;
522 mdb = rcu_dereference_protected(br->mdb, 1);
523 hlist_for_each_entry(mp, p, &mdb->mhash[hash], hlist[mdb->ver]) {
525 if (unlikely(br_ip_equal(group, &mp->addr)))
532 if (unlikely(count > br->hash_elasticity && count)) {
534 br_info(br, "Multicast hash table "
535 "chain limit reached: %s\n",
536 port ? port->dev->name : br->dev->name);
538 elasticity = br->hash_elasticity;
541 if (mdb->size >= max) {
543 if (unlikely(max >= br->hash_max)) {
544 br_warn(br, "Multicast hash table maximum "
545 "reached, disabling snooping: %s, %d\n",
546 port ? port->dev->name : br->dev->name, max);
549 br->multicast_disabled = 1;
554 if (max > mdb->max || elasticity) {
557 br_info(br, "Multicast hash table "
559 port ? port->dev->name : br->dev->name);
564 err = br_mdb_rehash(&br->mdb, max, elasticity);
566 br_warn(br, "Cannot rehash multicast "
567 "hash table, disabling snooping: %s, %d, %d\n",
568 port ? port->dev->name : br->dev->name,
584 static struct net_bridge_mdb_entry *br_multicast_new_group(
585 struct net_bridge *br, struct net_bridge_port *port,
588 struct net_bridge_mdb_htable *mdb;
589 struct net_bridge_mdb_entry *mp;
593 mdb = rcu_dereference_protected(br->mdb, 1);
595 err = br_mdb_rehash(&br->mdb, BR_HASH_SIZE, 0);
601 hash = br_ip_hash(mdb, group);
602 mp = br_multicast_get_group(br, port, group, hash);
603 switch (PTR_ERR(mp)) {
609 mdb = rcu_dereference_protected(br->mdb, 1);
610 hash = br_ip_hash(mdb, group);
617 mp = kzalloc(sizeof(*mp), GFP_ATOMIC);
619 return ERR_PTR(-ENOMEM);
623 setup_timer(&mp->timer, br_multicast_group_expired,
626 hlist_add_head_rcu(&mp->hlist[mdb->ver], &mdb->mhash[hash]);
633 static int br_multicast_add_group(struct net_bridge *br,
634 struct net_bridge_port *port,
637 struct net_bridge_mdb_entry *mp;
638 struct net_bridge_port_group *p;
639 struct net_bridge_port_group __rcu **pp;
640 unsigned long now = jiffies;
643 spin_lock(&br->multicast_lock);
644 if (!netif_running(br->dev) ||
645 (port && port->state == BR_STATE_DISABLED))
648 mp = br_multicast_new_group(br, port, group);
655 mod_timer(&mp->timer, now + br->multicast_membership_interval);
659 for (pp = &mp->ports;
660 (p = mlock_dereference(*pp, br)) != NULL;
664 if ((unsigned long)p->port < (unsigned long)port)
668 p = kzalloc(sizeof(*p), GFP_ATOMIC);
676 hlist_add_head(&p->mglist, &port->mglist);
677 setup_timer(&p->timer, br_multicast_port_group_expired,
680 rcu_assign_pointer(*pp, p);
683 mod_timer(&p->timer, now + br->multicast_membership_interval);
688 spin_unlock(&br->multicast_lock);
692 static int br_ip4_multicast_add_group(struct net_bridge *br,
693 struct net_bridge_port *port,
696 struct br_ip br_group;
698 if (ipv4_is_local_multicast(group))
701 br_group.u.ip4 = group;
702 br_group.proto = htons(ETH_P_IP);
704 return br_multicast_add_group(br, port, &br_group);
707 #if IS_ENABLED(CONFIG_IPV6)
708 static int br_ip6_multicast_add_group(struct net_bridge *br,
709 struct net_bridge_port *port,
710 const struct in6_addr *group)
712 struct br_ip br_group;
714 if (!ipv6_is_transient_multicast(group))
717 br_group.u.ip6 = *group;
718 br_group.proto = htons(ETH_P_IPV6);
720 return br_multicast_add_group(br, port, &br_group);
724 static void br_multicast_router_expired(unsigned long data)
726 struct net_bridge_port *port = (void *)data;
727 struct net_bridge *br = port->br;
729 spin_lock(&br->multicast_lock);
730 if (port->multicast_router != 1 ||
731 timer_pending(&port->multicast_router_timer) ||
732 hlist_unhashed(&port->rlist))
735 hlist_del_init_rcu(&port->rlist);
738 spin_unlock(&br->multicast_lock);
741 static void br_multicast_local_router_expired(unsigned long data)
745 static void br_multicast_querier_expired(unsigned long data)
747 struct net_bridge *br = (void *)data;
749 spin_lock(&br->multicast_lock);
750 if (!netif_running(br->dev) || br->multicast_disabled)
753 br_multicast_start_querier(br);
756 spin_unlock(&br->multicast_lock);
759 static void __br_multicast_send_query(struct net_bridge *br,
760 struct net_bridge_port *port,
765 skb = br_multicast_alloc_query(br, ip);
770 __skb_push(skb, sizeof(struct ethhdr));
771 skb->dev = port->dev;
772 NF_HOOK(NFPROTO_BRIDGE, NF_BR_LOCAL_OUT, skb, NULL, skb->dev,
778 static void br_multicast_send_query(struct net_bridge *br,
779 struct net_bridge_port *port, u32 sent)
782 struct br_ip br_group;
784 if (!netif_running(br->dev) || br->multicast_disabled ||
785 !br->multicast_querier ||
786 timer_pending(&br->multicast_querier_timer))
789 memset(&br_group.u, 0, sizeof(br_group.u));
791 br_group.proto = htons(ETH_P_IP);
792 __br_multicast_send_query(br, port, &br_group);
794 #if IS_ENABLED(CONFIG_IPV6)
795 br_group.proto = htons(ETH_P_IPV6);
796 __br_multicast_send_query(br, port, &br_group);
800 time += sent < br->multicast_startup_query_count ?
801 br->multicast_startup_query_interval :
802 br->multicast_query_interval;
803 mod_timer(port ? &port->multicast_query_timer :
804 &br->multicast_query_timer, time);
807 static void br_multicast_port_query_expired(unsigned long data)
809 struct net_bridge_port *port = (void *)data;
810 struct net_bridge *br = port->br;
812 spin_lock(&br->multicast_lock);
813 if (port->state == BR_STATE_DISABLED ||
814 port->state == BR_STATE_BLOCKING)
817 if (port->multicast_startup_queries_sent <
818 br->multicast_startup_query_count)
819 port->multicast_startup_queries_sent++;
821 br_multicast_send_query(port->br, port,
822 port->multicast_startup_queries_sent);
825 spin_unlock(&br->multicast_lock);
828 void br_multicast_add_port(struct net_bridge_port *port)
830 port->multicast_router = 1;
832 setup_timer(&port->multicast_router_timer, br_multicast_router_expired,
833 (unsigned long)port);
834 setup_timer(&port->multicast_query_timer,
835 br_multicast_port_query_expired, (unsigned long)port);
838 void br_multicast_del_port(struct net_bridge_port *port)
840 del_timer_sync(&port->multicast_router_timer);
843 static void __br_multicast_enable_port(struct net_bridge_port *port)
845 port->multicast_startup_queries_sent = 0;
847 if (try_to_del_timer_sync(&port->multicast_query_timer) >= 0 ||
848 del_timer(&port->multicast_query_timer))
849 mod_timer(&port->multicast_query_timer, jiffies);
852 void br_multicast_enable_port(struct net_bridge_port *port)
854 struct net_bridge *br = port->br;
856 spin_lock(&br->multicast_lock);
857 if (br->multicast_disabled || !netif_running(br->dev))
860 __br_multicast_enable_port(port);
863 spin_unlock(&br->multicast_lock);
866 void br_multicast_disable_port(struct net_bridge_port *port)
868 struct net_bridge *br = port->br;
869 struct net_bridge_port_group *pg;
870 struct hlist_node *p, *n;
872 spin_lock(&br->multicast_lock);
873 hlist_for_each_entry_safe(pg, p, n, &port->mglist, mglist)
874 br_multicast_del_pg(br, pg);
876 if (!hlist_unhashed(&port->rlist))
877 hlist_del_init_rcu(&port->rlist);
878 del_timer(&port->multicast_router_timer);
879 del_timer(&port->multicast_query_timer);
880 spin_unlock(&br->multicast_lock);
883 static int br_ip4_multicast_igmp3_report(struct net_bridge *br,
884 struct net_bridge_port *port,
887 struct igmpv3_report *ih;
888 struct igmpv3_grec *grec;
896 if (!pskb_may_pull(skb, sizeof(*ih)))
899 ih = igmpv3_report_hdr(skb);
900 num = ntohs(ih->ngrec);
903 for (i = 0; i < num; i++) {
904 len += sizeof(*grec);
905 if (!pskb_may_pull(skb, len))
908 grec = (void *)(skb->data + len - sizeof(*grec));
909 group = grec->grec_mca;
910 type = grec->grec_type;
912 len += ntohs(grec->grec_nsrcs) * 4;
913 if (!pskb_may_pull(skb, len))
916 /* We treat this as an IGMPv2 report for now. */
918 case IGMPV3_MODE_IS_INCLUDE:
919 case IGMPV3_MODE_IS_EXCLUDE:
920 case IGMPV3_CHANGE_TO_INCLUDE:
921 case IGMPV3_CHANGE_TO_EXCLUDE:
922 case IGMPV3_ALLOW_NEW_SOURCES:
923 case IGMPV3_BLOCK_OLD_SOURCES:
930 err = br_ip4_multicast_add_group(br, port, group);
938 #if IS_ENABLED(CONFIG_IPV6)
939 static int br_ip6_multicast_mld2_report(struct net_bridge *br,
940 struct net_bridge_port *port,
943 struct icmp6hdr *icmp6h;
944 struct mld2_grec *grec;
950 if (!pskb_may_pull(skb, sizeof(*icmp6h)))
953 icmp6h = icmp6_hdr(skb);
954 num = ntohs(icmp6h->icmp6_dataun.un_data16[1]);
955 len = sizeof(*icmp6h);
957 for (i = 0; i < num; i++) {
958 __be16 *nsrcs, _nsrcs;
960 nsrcs = skb_header_pointer(skb,
961 len + offsetof(struct mld2_grec,
963 sizeof(_nsrcs), &_nsrcs);
967 if (!pskb_may_pull(skb,
968 len + sizeof(*grec) +
969 sizeof(struct in6_addr) * ntohs(*nsrcs)))
972 grec = (struct mld2_grec *)(skb->data + len);
973 len += sizeof(*grec) +
974 sizeof(struct in6_addr) * ntohs(*nsrcs);
976 /* We treat these as MLDv1 reports for now. */
977 switch (grec->grec_type) {
978 case MLD2_MODE_IS_INCLUDE:
979 case MLD2_MODE_IS_EXCLUDE:
980 case MLD2_CHANGE_TO_INCLUDE:
981 case MLD2_CHANGE_TO_EXCLUDE:
982 case MLD2_ALLOW_NEW_SOURCES:
983 case MLD2_BLOCK_OLD_SOURCES:
990 err = br_ip6_multicast_add_group(br, port, &grec->grec_mca);
1000 * Add port to rotuer_list
1001 * list is maintained ordered by pointer value
1002 * and locked by br->multicast_lock and RCU
1004 static void br_multicast_add_router(struct net_bridge *br,
1005 struct net_bridge_port *port)
1007 struct net_bridge_port *p;
1008 struct hlist_node *n, *slot = NULL;
1010 hlist_for_each_entry(p, n, &br->router_list, rlist) {
1011 if ((unsigned long) port >= (unsigned long) p)
1017 hlist_add_after_rcu(slot, &port->rlist);
1019 hlist_add_head_rcu(&port->rlist, &br->router_list);
1022 static void br_multicast_mark_router(struct net_bridge *br,
1023 struct net_bridge_port *port)
1025 unsigned long now = jiffies;
1028 if (br->multicast_router == 1)
1029 mod_timer(&br->multicast_router_timer,
1030 now + br->multicast_querier_interval);
1034 if (port->multicast_router != 1)
1037 if (!hlist_unhashed(&port->rlist))
1040 br_multicast_add_router(br, port);
1043 mod_timer(&port->multicast_router_timer,
1044 now + br->multicast_querier_interval);
1047 static void br_multicast_query_received(struct net_bridge *br,
1048 struct net_bridge_port *port,
1052 mod_timer(&br->multicast_querier_timer,
1053 jiffies + br->multicast_querier_interval);
1054 else if (timer_pending(&br->multicast_querier_timer))
1057 br_multicast_mark_router(br, port);
1060 static int br_ip4_multicast_query(struct net_bridge *br,
1061 struct net_bridge_port *port,
1062 struct sk_buff *skb)
1064 const struct iphdr *iph = ip_hdr(skb);
1065 struct igmphdr *ih = igmp_hdr(skb);
1066 struct net_bridge_mdb_entry *mp;
1067 struct igmpv3_query *ih3;
1068 struct net_bridge_port_group *p;
1069 struct net_bridge_port_group __rcu **pp;
1070 unsigned long max_delay;
1071 unsigned long now = jiffies;
1075 spin_lock(&br->multicast_lock);
1076 if (!netif_running(br->dev) ||
1077 (port && port->state == BR_STATE_DISABLED))
1080 br_multicast_query_received(br, port, !!iph->saddr);
1084 if (skb->len == sizeof(*ih)) {
1085 max_delay = ih->code * (HZ / IGMP_TIMER_SCALE);
1088 max_delay = 10 * HZ;
1092 if (!pskb_may_pull(skb, sizeof(struct igmpv3_query))) {
1097 ih3 = igmpv3_query_hdr(skb);
1101 max_delay = ih3->code ?
1102 IGMPV3_MRC(ih3->code) * (HZ / IGMP_TIMER_SCALE) : 1;
1108 mp = br_mdb_ip4_get(mlock_dereference(br->mdb, br), group);
1112 max_delay *= br->multicast_last_member_count;
1115 (timer_pending(&mp->timer) ?
1116 time_after(mp->timer.expires, now + max_delay) :
1117 try_to_del_timer_sync(&mp->timer) >= 0))
1118 mod_timer(&mp->timer, now + max_delay);
1120 for (pp = &mp->ports;
1121 (p = mlock_dereference(*pp, br)) != NULL;
1123 if (timer_pending(&p->timer) ?
1124 time_after(p->timer.expires, now + max_delay) :
1125 try_to_del_timer_sync(&p->timer) >= 0)
1126 mod_timer(&p->timer, now + max_delay);
1130 spin_unlock(&br->multicast_lock);
1134 #if IS_ENABLED(CONFIG_IPV6)
1135 static int br_ip6_multicast_query(struct net_bridge *br,
1136 struct net_bridge_port *port,
1137 struct sk_buff *skb)
1139 const struct ipv6hdr *ip6h = ipv6_hdr(skb);
1140 struct mld_msg *mld = (struct mld_msg *) icmp6_hdr(skb);
1141 struct net_bridge_mdb_entry *mp;
1142 struct mld2_query *mld2q;
1143 struct net_bridge_port_group *p;
1144 struct net_bridge_port_group __rcu **pp;
1145 unsigned long max_delay;
1146 unsigned long now = jiffies;
1147 const struct in6_addr *group = NULL;
1150 spin_lock(&br->multicast_lock);
1151 if (!netif_running(br->dev) ||
1152 (port && port->state == BR_STATE_DISABLED))
1155 br_multicast_query_received(br, port, !ipv6_addr_any(&ip6h->saddr));
1157 if (skb->len == sizeof(*mld)) {
1158 if (!pskb_may_pull(skb, sizeof(*mld))) {
1162 mld = (struct mld_msg *) icmp6_hdr(skb);
1163 max_delay = msecs_to_jiffies(ntohs(mld->mld_maxdelay));
1165 group = &mld->mld_mca;
1166 } else if (skb->len >= sizeof(*mld2q)) {
1167 if (!pskb_may_pull(skb, sizeof(*mld2q))) {
1171 mld2q = (struct mld2_query *)icmp6_hdr(skb);
1172 if (!mld2q->mld2q_nsrcs)
1173 group = &mld2q->mld2q_mca;
1174 max_delay = mld2q->mld2q_mrc ? MLDV2_MRC(mld2q->mld2q_mrc) : 1;
1180 mp = br_mdb_ip6_get(mlock_dereference(br->mdb, br), group);
1184 max_delay *= br->multicast_last_member_count;
1186 (timer_pending(&mp->timer) ?
1187 time_after(mp->timer.expires, now + max_delay) :
1188 try_to_del_timer_sync(&mp->timer) >= 0))
1189 mod_timer(&mp->timer, now + max_delay);
1191 for (pp = &mp->ports;
1192 (p = mlock_dereference(*pp, br)) != NULL;
1194 if (timer_pending(&p->timer) ?
1195 time_after(p->timer.expires, now + max_delay) :
1196 try_to_del_timer_sync(&p->timer) >= 0)
1197 mod_timer(&p->timer, now + max_delay);
1201 spin_unlock(&br->multicast_lock);
1206 static void br_multicast_leave_group(struct net_bridge *br,
1207 struct net_bridge_port *port,
1208 struct br_ip *group)
1210 struct net_bridge_mdb_htable *mdb;
1211 struct net_bridge_mdb_entry *mp;
1212 struct net_bridge_port_group *p;
1216 spin_lock(&br->multicast_lock);
1217 if (!netif_running(br->dev) ||
1218 (port && port->state == BR_STATE_DISABLED) ||
1219 timer_pending(&br->multicast_querier_timer))
1222 mdb = mlock_dereference(br->mdb, br);
1223 mp = br_mdb_ip_get(mdb, group);
1228 time = now + br->multicast_last_member_count *
1229 br->multicast_last_member_interval;
1233 (timer_pending(&mp->timer) ?
1234 time_after(mp->timer.expires, time) :
1235 try_to_del_timer_sync(&mp->timer) >= 0)) {
1236 mod_timer(&mp->timer, time);
1242 for (p = mlock_dereference(mp->ports, br);
1244 p = mlock_dereference(p->next, br)) {
1245 if (p->port != port)
1248 if (!hlist_unhashed(&p->mglist) &&
1249 (timer_pending(&p->timer) ?
1250 time_after(p->timer.expires, time) :
1251 try_to_del_timer_sync(&p->timer) >= 0)) {
1252 mod_timer(&p->timer, time);
1259 spin_unlock(&br->multicast_lock);
1262 static void br_ip4_multicast_leave_group(struct net_bridge *br,
1263 struct net_bridge_port *port,
1266 struct br_ip br_group;
1268 if (ipv4_is_local_multicast(group))
1271 br_group.u.ip4 = group;
1272 br_group.proto = htons(ETH_P_IP);
1274 br_multicast_leave_group(br, port, &br_group);
1277 #if IS_ENABLED(CONFIG_IPV6)
1278 static void br_ip6_multicast_leave_group(struct net_bridge *br,
1279 struct net_bridge_port *port,
1280 const struct in6_addr *group)
1282 struct br_ip br_group;
1284 if (!ipv6_is_transient_multicast(group))
1287 br_group.u.ip6 = *group;
1288 br_group.proto = htons(ETH_P_IPV6);
1290 br_multicast_leave_group(br, port, &br_group);
1294 static int br_multicast_ipv4_rcv(struct net_bridge *br,
1295 struct net_bridge_port *port,
1296 struct sk_buff *skb)
1298 struct sk_buff *skb2 = skb;
1299 const struct iphdr *iph;
1302 unsigned int offset;
1305 /* We treat OOM as packet loss for now. */
1306 if (!pskb_may_pull(skb, sizeof(*iph)))
1311 if (iph->ihl < 5 || iph->version != 4)
1314 if (!pskb_may_pull(skb, ip_hdrlen(skb)))
1319 if (unlikely(ip_fast_csum((u8 *)iph, iph->ihl)))
1322 if (iph->protocol != IPPROTO_IGMP) {
1323 if ((iph->daddr & IGMP_LOCAL_GROUP_MASK) != IGMP_LOCAL_GROUP)
1324 BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1328 len = ntohs(iph->tot_len);
1329 if (skb->len < len || len < ip_hdrlen(skb))
1332 if (skb->len > len) {
1333 skb2 = skb_clone(skb, GFP_ATOMIC);
1337 err = pskb_trim_rcsum(skb2, len);
1342 len -= ip_hdrlen(skb2);
1343 offset = skb_network_offset(skb2) + ip_hdrlen(skb2);
1344 __skb_pull(skb2, offset);
1345 skb_reset_transport_header(skb2);
1348 if (!pskb_may_pull(skb2, sizeof(*ih)))
1351 switch (skb2->ip_summed) {
1352 case CHECKSUM_COMPLETE:
1353 if (!csum_fold(skb2->csum))
1358 if (skb_checksum_complete(skb2))
1364 BR_INPUT_SKB_CB(skb)->igmp = 1;
1365 ih = igmp_hdr(skb2);
1368 case IGMP_HOST_MEMBERSHIP_REPORT:
1369 case IGMPV2_HOST_MEMBERSHIP_REPORT:
1370 BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1371 err = br_ip4_multicast_add_group(br, port, ih->group);
1373 case IGMPV3_HOST_MEMBERSHIP_REPORT:
1374 err = br_ip4_multicast_igmp3_report(br, port, skb2);
1376 case IGMP_HOST_MEMBERSHIP_QUERY:
1377 err = br_ip4_multicast_query(br, port, skb2);
1379 case IGMP_HOST_LEAVE_MESSAGE:
1380 br_ip4_multicast_leave_group(br, port, ih->group);
1385 __skb_push(skb2, offset);
1392 #if IS_ENABLED(CONFIG_IPV6)
1393 static int br_multicast_ipv6_rcv(struct net_bridge *br,
1394 struct net_bridge_port *port,
1395 struct sk_buff *skb)
1397 struct sk_buff *skb2;
1398 const struct ipv6hdr *ip6h;
1406 if (!pskb_may_pull(skb, sizeof(*ip6h)))
1409 ip6h = ipv6_hdr(skb);
1412 * We're interested in MLD messages only.
1414 * - MLD has always Router Alert hop-by-hop option
1415 * - But we do not support jumbrograms.
1417 if (ip6h->version != 6 ||
1418 ip6h->nexthdr != IPPROTO_HOPOPTS ||
1419 ip6h->payload_len == 0)
1422 len = ntohs(ip6h->payload_len) + sizeof(*ip6h);
1426 nexthdr = ip6h->nexthdr;
1427 offset = ipv6_skip_exthdr(skb, sizeof(*ip6h), &nexthdr, &frag_off);
1429 if (offset < 0 || nexthdr != IPPROTO_ICMPV6)
1432 /* Okay, we found ICMPv6 header */
1433 skb2 = skb_clone(skb, GFP_ATOMIC);
1438 if (!pskb_may_pull(skb2, offset + sizeof(struct icmp6hdr)))
1441 len -= offset - skb_network_offset(skb2);
1443 __skb_pull(skb2, offset);
1444 skb_reset_transport_header(skb2);
1445 skb_postpull_rcsum(skb2, skb_network_header(skb2),
1446 skb_network_header_len(skb2));
1448 icmp6_type = icmp6_hdr(skb2)->icmp6_type;
1450 switch (icmp6_type) {
1451 case ICMPV6_MGM_QUERY:
1452 case ICMPV6_MGM_REPORT:
1453 case ICMPV6_MGM_REDUCTION:
1454 case ICMPV6_MLD2_REPORT:
1461 /* Okay, we found MLD message. Check further. */
1462 if (skb2->len > len) {
1463 err = pskb_trim_rcsum(skb2, len);
1469 ip6h = ipv6_hdr(skb2);
1471 switch (skb2->ip_summed) {
1472 case CHECKSUM_COMPLETE:
1473 if (!csum_ipv6_magic(&ip6h->saddr, &ip6h->daddr, skb2->len,
1474 IPPROTO_ICMPV6, skb2->csum))
1478 skb2->csum = ~csum_unfold(csum_ipv6_magic(&ip6h->saddr,
1481 IPPROTO_ICMPV6, 0));
1482 if (__skb_checksum_complete(skb2))
1488 BR_INPUT_SKB_CB(skb)->igmp = 1;
1490 switch (icmp6_type) {
1491 case ICMPV6_MGM_REPORT:
1493 struct mld_msg *mld;
1494 if (!pskb_may_pull(skb2, sizeof(*mld))) {
1498 mld = (struct mld_msg *)skb_transport_header(skb2);
1499 BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1500 err = br_ip6_multicast_add_group(br, port, &mld->mld_mca);
1503 case ICMPV6_MLD2_REPORT:
1504 err = br_ip6_multicast_mld2_report(br, port, skb2);
1506 case ICMPV6_MGM_QUERY:
1507 err = br_ip6_multicast_query(br, port, skb2);
1509 case ICMPV6_MGM_REDUCTION:
1511 struct mld_msg *mld;
1512 if (!pskb_may_pull(skb2, sizeof(*mld))) {
1516 mld = (struct mld_msg *)skb_transport_header(skb2);
1517 br_ip6_multicast_leave_group(br, port, &mld->mld_mca);
1527 int br_multicast_rcv(struct net_bridge *br, struct net_bridge_port *port,
1528 struct sk_buff *skb)
1530 BR_INPUT_SKB_CB(skb)->igmp = 0;
1531 BR_INPUT_SKB_CB(skb)->mrouters_only = 0;
1533 if (br->multicast_disabled)
1536 switch (skb->protocol) {
1537 case htons(ETH_P_IP):
1538 return br_multicast_ipv4_rcv(br, port, skb);
1539 #if IS_ENABLED(CONFIG_IPV6)
1540 case htons(ETH_P_IPV6):
1541 return br_multicast_ipv6_rcv(br, port, skb);
1548 static void br_multicast_query_expired(unsigned long data)
1550 struct net_bridge *br = (void *)data;
1552 spin_lock(&br->multicast_lock);
1553 if (br->multicast_startup_queries_sent <
1554 br->multicast_startup_query_count)
1555 br->multicast_startup_queries_sent++;
1557 br_multicast_send_query(br, NULL, br->multicast_startup_queries_sent);
1559 spin_unlock(&br->multicast_lock);
1562 void br_multicast_init(struct net_bridge *br)
1564 br->hash_elasticity = 4;
1567 br->multicast_router = 1;
1568 br->multicast_querier = 0;
1569 br->multicast_last_member_count = 2;
1570 br->multicast_startup_query_count = 2;
1572 br->multicast_last_member_interval = HZ;
1573 br->multicast_query_response_interval = 10 * HZ;
1574 br->multicast_startup_query_interval = 125 * HZ / 4;
1575 br->multicast_query_interval = 125 * HZ;
1576 br->multicast_querier_interval = 255 * HZ;
1577 br->multicast_membership_interval = 260 * HZ;
1579 spin_lock_init(&br->multicast_lock);
1580 setup_timer(&br->multicast_router_timer,
1581 br_multicast_local_router_expired, 0);
1582 setup_timer(&br->multicast_querier_timer,
1583 br_multicast_querier_expired, (unsigned long)br);
1584 setup_timer(&br->multicast_query_timer, br_multicast_query_expired,
1588 void br_multicast_open(struct net_bridge *br)
1590 br->multicast_startup_queries_sent = 0;
1592 if (br->multicast_disabled)
1595 mod_timer(&br->multicast_query_timer, jiffies);
1598 void br_multicast_stop(struct net_bridge *br)
1600 struct net_bridge_mdb_htable *mdb;
1601 struct net_bridge_mdb_entry *mp;
1602 struct hlist_node *p, *n;
1606 del_timer_sync(&br->multicast_router_timer);
1607 del_timer_sync(&br->multicast_querier_timer);
1608 del_timer_sync(&br->multicast_query_timer);
1610 spin_lock_bh(&br->multicast_lock);
1611 mdb = mlock_dereference(br->mdb, br);
1618 for (i = 0; i < mdb->max; i++) {
1619 hlist_for_each_entry_safe(mp, p, n, &mdb->mhash[i],
1621 del_timer(&mp->timer);
1622 call_rcu_bh(&mp->rcu, br_multicast_free_group);
1627 spin_unlock_bh(&br->multicast_lock);
1629 spin_lock_bh(&br->multicast_lock);
1634 call_rcu_bh(&mdb->rcu, br_mdb_free);
1637 spin_unlock_bh(&br->multicast_lock);
1640 int br_multicast_set_router(struct net_bridge *br, unsigned long val)
1644 spin_lock_bh(&br->multicast_lock);
1645 if (!netif_running(br->dev))
1651 del_timer(&br->multicast_router_timer);
1654 br->multicast_router = val;
1664 spin_unlock_bh(&br->multicast_lock);
1669 int br_multicast_set_port_router(struct net_bridge_port *p, unsigned long val)
1671 struct net_bridge *br = p->br;
1674 spin_lock(&br->multicast_lock);
1675 if (!netif_running(br->dev) || p->state == BR_STATE_DISABLED)
1682 p->multicast_router = val;
1685 if (val < 2 && !hlist_unhashed(&p->rlist))
1686 hlist_del_init_rcu(&p->rlist);
1691 del_timer(&p->multicast_router_timer);
1696 br_multicast_add_router(br, p);
1705 spin_unlock(&br->multicast_lock);
1710 static void br_multicast_start_querier(struct net_bridge *br)
1712 struct net_bridge_port *port;
1714 br_multicast_open(br);
1716 list_for_each_entry(port, &br->port_list, list) {
1717 if (port->state == BR_STATE_DISABLED ||
1718 port->state == BR_STATE_BLOCKING)
1721 __br_multicast_enable_port(port);
1725 int br_multicast_toggle(struct net_bridge *br, unsigned long val)
1728 struct net_bridge_mdb_htable *mdb;
1730 spin_lock_bh(&br->multicast_lock);
1731 if (br->multicast_disabled == !val)
1734 br->multicast_disabled = !val;
1735 if (br->multicast_disabled)
1738 if (!netif_running(br->dev))
1741 mdb = mlock_dereference(br->mdb, br);
1746 br->multicast_disabled = !!val;
1750 err = br_mdb_rehash(&br->mdb, mdb->max,
1751 br->hash_elasticity);
1756 br_multicast_start_querier(br);
1759 spin_unlock_bh(&br->multicast_lock);
1764 int br_multicast_set_querier(struct net_bridge *br, unsigned long val)
1768 spin_lock_bh(&br->multicast_lock);
1769 if (br->multicast_querier == val)
1772 br->multicast_querier = val;
1774 br_multicast_start_querier(br);
1777 spin_unlock_bh(&br->multicast_lock);
1782 int br_multicast_set_hash_max(struct net_bridge *br, unsigned long val)
1786 struct net_bridge_mdb_htable *mdb;
1788 spin_lock(&br->multicast_lock);
1789 if (!netif_running(br->dev))
1793 if (!is_power_of_2(val))
1796 mdb = mlock_dereference(br->mdb, br);
1797 if (mdb && val < mdb->size)
1813 err = br_mdb_rehash(&br->mdb, br->hash_max,
1814 br->hash_elasticity);
1820 spin_unlock(&br->multicast_lock);