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 defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
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 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
40 static inline int ipv6_is_transient_multicast(const struct in6_addr *addr)
42 if (ipv6_addr_is_multicast(addr) && IPV6_ADDR_MC_FLAG_TRANSIENT(addr))
48 static inline int br_ip_equal(const struct br_ip *a, const struct br_ip *b)
50 if (a->proto != b->proto)
54 return a->u.ip4 == b->u.ip4;
55 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
56 case htons(ETH_P_IPV6):
57 return ipv6_addr_equal(&a->u.ip6, &b->u.ip6);
63 static inline int __br_ip4_hash(struct net_bridge_mdb_htable *mdb, __be32 ip)
65 return jhash_1word(mdb->secret, (__force u32)ip) & (mdb->max - 1);
68 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
69 static inline int __br_ip6_hash(struct net_bridge_mdb_htable *mdb,
70 const struct in6_addr *ip)
72 return jhash2((__force u32 *)ip->s6_addr32, 4, mdb->secret) & (mdb->max - 1);
76 static inline int br_ip_hash(struct net_bridge_mdb_htable *mdb,
81 return __br_ip4_hash(mdb, ip->u.ip4);
82 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
83 case htons(ETH_P_IPV6):
84 return __br_ip6_hash(mdb, &ip->u.ip6);
90 static struct net_bridge_mdb_entry *__br_mdb_ip_get(
91 struct net_bridge_mdb_htable *mdb, struct br_ip *dst, int hash)
93 struct net_bridge_mdb_entry *mp;
96 hlist_for_each_entry_rcu(mp, p, &mdb->mhash[hash], hlist[mdb->ver]) {
97 if (br_ip_equal(&mp->addr, dst))
104 static struct net_bridge_mdb_entry *br_mdb_ip_get(
105 struct net_bridge_mdb_htable *mdb, struct br_ip *dst)
110 return __br_mdb_ip_get(mdb, dst, br_ip_hash(mdb, dst));
113 static struct net_bridge_mdb_entry *br_mdb_ip4_get(
114 struct net_bridge_mdb_htable *mdb, __be32 dst)
119 br_dst.proto = htons(ETH_P_IP);
121 return br_mdb_ip_get(mdb, &br_dst);
124 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
125 static struct net_bridge_mdb_entry *br_mdb_ip6_get(
126 struct net_bridge_mdb_htable *mdb, const struct in6_addr *dst)
130 ipv6_addr_copy(&br_dst.u.ip6, dst);
131 br_dst.proto = htons(ETH_P_IPV6);
133 return br_mdb_ip_get(mdb, &br_dst);
137 struct net_bridge_mdb_entry *br_mdb_get(struct net_bridge *br,
140 struct net_bridge_mdb_htable *mdb = rcu_dereference(br->mdb);
143 if (br->multicast_disabled)
146 if (BR_INPUT_SKB_CB(skb)->igmp)
149 ip.proto = skb->protocol;
151 switch (skb->protocol) {
152 case htons(ETH_P_IP):
153 ip.u.ip4 = ip_hdr(skb)->daddr;
155 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
156 case htons(ETH_P_IPV6):
157 ipv6_addr_copy(&ip.u.ip6, &ipv6_hdr(skb)->daddr);
164 return br_mdb_ip_get(mdb, &ip);
167 static void br_mdb_free(struct rcu_head *head)
169 struct net_bridge_mdb_htable *mdb =
170 container_of(head, struct net_bridge_mdb_htable, rcu);
171 struct net_bridge_mdb_htable *old = mdb->old;
178 static int br_mdb_copy(struct net_bridge_mdb_htable *new,
179 struct net_bridge_mdb_htable *old,
182 struct net_bridge_mdb_entry *mp;
183 struct hlist_node *p;
188 for (i = 0; i < old->max; i++)
189 hlist_for_each_entry(mp, p, &old->mhash[i], hlist[old->ver])
190 hlist_add_head(&mp->hlist[new->ver],
191 &new->mhash[br_ip_hash(new, &mp->addr)]);
197 for (i = 0; i < new->max; i++) {
199 hlist_for_each_entry(mp, p, &new->mhash[i], hlist[new->ver])
205 return maxlen > elasticity ? -EINVAL : 0;
208 static void br_multicast_free_pg(struct rcu_head *head)
210 struct net_bridge_port_group *p =
211 container_of(head, struct net_bridge_port_group, rcu);
216 static void br_multicast_free_group(struct rcu_head *head)
218 struct net_bridge_mdb_entry *mp =
219 container_of(head, struct net_bridge_mdb_entry, rcu);
224 static void br_multicast_group_expired(unsigned long data)
226 struct net_bridge_mdb_entry *mp = (void *)data;
227 struct net_bridge *br = mp->br;
228 struct net_bridge_mdb_htable *mdb;
230 spin_lock(&br->multicast_lock);
231 if (!netif_running(br->dev) || timer_pending(&mp->timer))
239 mdb = mlock_dereference(br->mdb, br);
241 hlist_del_rcu(&mp->hlist[mdb->ver]);
244 call_rcu_bh(&mp->rcu, br_multicast_free_group);
247 spin_unlock(&br->multicast_lock);
250 static void br_multicast_del_pg(struct net_bridge *br,
251 struct net_bridge_port_group *pg)
253 struct net_bridge_mdb_htable *mdb;
254 struct net_bridge_mdb_entry *mp;
255 struct net_bridge_port_group *p;
256 struct net_bridge_port_group __rcu **pp;
258 mdb = mlock_dereference(br->mdb, br);
260 mp = br_mdb_ip_get(mdb, &pg->addr);
264 for (pp = &mp->ports;
265 (p = mlock_dereference(*pp, br)) != NULL;
270 rcu_assign_pointer(*pp, p->next);
271 hlist_del_init(&p->mglist);
272 del_timer(&p->timer);
273 call_rcu_bh(&p->rcu, br_multicast_free_pg);
275 if (!mp->ports && !mp->mglist &&
276 netif_running(br->dev))
277 mod_timer(&mp->timer, jiffies);
285 static void br_multicast_port_group_expired(unsigned long data)
287 struct net_bridge_port_group *pg = (void *)data;
288 struct net_bridge *br = pg->port->br;
290 spin_lock(&br->multicast_lock);
291 if (!netif_running(br->dev) || timer_pending(&pg->timer) ||
292 hlist_unhashed(&pg->mglist))
295 br_multicast_del_pg(br, pg);
298 spin_unlock(&br->multicast_lock);
301 static int br_mdb_rehash(struct net_bridge_mdb_htable __rcu **mdbp, int max,
304 struct net_bridge_mdb_htable *old = rcu_dereference_protected(*mdbp, 1);
305 struct net_bridge_mdb_htable *mdb;
308 mdb = kmalloc(sizeof(*mdb), GFP_ATOMIC);
315 mdb->mhash = kzalloc(max * sizeof(*mdb->mhash), GFP_ATOMIC);
321 mdb->size = old ? old->size : 0;
322 mdb->ver = old ? old->ver ^ 1 : 0;
324 if (!old || elasticity)
325 get_random_bytes(&mdb->secret, sizeof(mdb->secret));
327 mdb->secret = old->secret;
332 err = br_mdb_copy(mdb, old, elasticity);
339 call_rcu_bh(&mdb->rcu, br_mdb_free);
342 rcu_assign_pointer(*mdbp, mdb);
347 static struct sk_buff *br_ip4_multicast_alloc_query(struct net_bridge *br,
355 skb = netdev_alloc_skb_ip_align(br->dev, sizeof(*eth) + sizeof(*iph) +
360 skb->protocol = htons(ETH_P_IP);
362 skb_reset_mac_header(skb);
365 memcpy(eth->h_source, br->dev->dev_addr, 6);
368 eth->h_dest[2] = 0x5e;
372 eth->h_proto = htons(ETH_P_IP);
373 skb_put(skb, sizeof(*eth));
375 skb_set_network_header(skb, skb->len);
381 iph->tot_len = htons(sizeof(*iph) + sizeof(*ih) + 4);
383 iph->frag_off = htons(IP_DF);
385 iph->protocol = IPPROTO_IGMP;
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 defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
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_PAD0; /* Pad0 */
462 hopopt[7] = IPV6_TLV_PAD0; /* Pad0 */
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 ipv6_addr_copy(&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 defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
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 struct hlist_node *p;
521 mdb = rcu_dereference_protected(br->mdb, 1);
522 hlist_for_each_entry(mp, p, &mdb->mhash[hash], hlist[mdb->ver]) {
524 if (unlikely(br_ip_equal(group, &mp->addr)))
531 if (unlikely(count > br->hash_elasticity && count)) {
533 br_info(br, "Multicast hash table "
534 "chain limit reached: %s\n",
535 port ? port->dev->name : br->dev->name);
537 elasticity = br->hash_elasticity;
540 if (mdb->size >= max) {
542 if (unlikely(max >= br->hash_max)) {
543 br_warn(br, "Multicast hash table maximum "
544 "reached, disabling snooping: %s, %d\n",
545 port ? port->dev->name : br->dev->name, max);
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 static struct net_bridge_mdb_entry *br_multicast_new_group(
584 struct net_bridge *br, struct net_bridge_port *port,
587 struct net_bridge_mdb_htable *mdb;
588 struct net_bridge_mdb_entry *mp;
592 mdb = rcu_dereference_protected(br->mdb, 1);
594 err = br_mdb_rehash(&br->mdb, BR_HASH_SIZE, 0);
600 hash = br_ip_hash(mdb, group);
601 mp = br_multicast_get_group(br, port, group, hash);
602 switch (PTR_ERR(mp)) {
608 mdb = rcu_dereference_protected(br->mdb, 1);
609 hash = br_ip_hash(mdb, group);
616 mp = kzalloc(sizeof(*mp), GFP_ATOMIC);
618 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 static int br_multicast_add_group(struct net_bridge *br,
633 struct net_bridge_port *port,
636 struct net_bridge_mdb_entry *mp;
637 struct net_bridge_port_group *p;
638 struct net_bridge_port_group __rcu **pp;
639 unsigned long now = jiffies;
642 spin_lock(&br->multicast_lock);
643 if (!netif_running(br->dev) ||
644 (port && port->state == BR_STATE_DISABLED))
647 mp = br_multicast_new_group(br, port, group);
654 mod_timer(&mp->timer, now + br->multicast_membership_interval);
658 for (pp = &mp->ports;
659 (p = mlock_dereference(*pp, br)) != NULL;
663 if ((unsigned long)p->port < (unsigned long)port)
667 p = kzalloc(sizeof(*p), GFP_ATOMIC);
675 hlist_add_head(&p->mglist, &port->mglist);
676 setup_timer(&p->timer, br_multicast_port_group_expired,
679 rcu_assign_pointer(*pp, p);
682 mod_timer(&p->timer, now + br->multicast_membership_interval);
687 spin_unlock(&br->multicast_lock);
691 static int br_ip4_multicast_add_group(struct net_bridge *br,
692 struct net_bridge_port *port,
695 struct br_ip br_group;
697 if (ipv4_is_local_multicast(group))
700 br_group.u.ip4 = group;
701 br_group.proto = htons(ETH_P_IP);
703 return br_multicast_add_group(br, port, &br_group);
706 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
707 static int br_ip6_multicast_add_group(struct net_bridge *br,
708 struct net_bridge_port *port,
709 const struct in6_addr *group)
711 struct br_ip br_group;
713 if (!ipv6_is_transient_multicast(group))
716 ipv6_addr_copy(&br_group.u.ip6, group);
717 br_group.proto = htons(ETH_P_IPV6);
719 return br_multicast_add_group(br, port, &br_group);
723 static void br_multicast_router_expired(unsigned long data)
725 struct net_bridge_port *port = (void *)data;
726 struct net_bridge *br = port->br;
728 spin_lock(&br->multicast_lock);
729 if (port->multicast_router != 1 ||
730 timer_pending(&port->multicast_router_timer) ||
731 hlist_unhashed(&port->rlist))
734 hlist_del_init_rcu(&port->rlist);
737 spin_unlock(&br->multicast_lock);
740 static void br_multicast_local_router_expired(unsigned long data)
744 static void __br_multicast_send_query(struct net_bridge *br,
745 struct net_bridge_port *port,
750 skb = br_multicast_alloc_query(br, ip);
755 __skb_push(skb, sizeof(struct ethhdr));
756 skb->dev = port->dev;
757 NF_HOOK(NFPROTO_BRIDGE, NF_BR_LOCAL_OUT, skb, NULL, skb->dev,
763 static void br_multicast_send_query(struct net_bridge *br,
764 struct net_bridge_port *port, u32 sent)
767 struct br_ip br_group;
769 if (!netif_running(br->dev) || br->multicast_disabled ||
770 timer_pending(&br->multicast_querier_timer))
773 memset(&br_group.u, 0, sizeof(br_group.u));
775 br_group.proto = htons(ETH_P_IP);
776 __br_multicast_send_query(br, port, &br_group);
778 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
779 br_group.proto = htons(ETH_P_IPV6);
780 __br_multicast_send_query(br, port, &br_group);
784 time += sent < br->multicast_startup_query_count ?
785 br->multicast_startup_query_interval :
786 br->multicast_query_interval;
787 mod_timer(port ? &port->multicast_query_timer :
788 &br->multicast_query_timer, time);
791 static void br_multicast_port_query_expired(unsigned long data)
793 struct net_bridge_port *port = (void *)data;
794 struct net_bridge *br = port->br;
796 spin_lock(&br->multicast_lock);
797 if (port->state == BR_STATE_DISABLED ||
798 port->state == BR_STATE_BLOCKING)
801 if (port->multicast_startup_queries_sent <
802 br->multicast_startup_query_count)
803 port->multicast_startup_queries_sent++;
805 br_multicast_send_query(port->br, port,
806 port->multicast_startup_queries_sent);
809 spin_unlock(&br->multicast_lock);
812 void br_multicast_add_port(struct net_bridge_port *port)
814 port->multicast_router = 1;
816 setup_timer(&port->multicast_router_timer, br_multicast_router_expired,
817 (unsigned long)port);
818 setup_timer(&port->multicast_query_timer,
819 br_multicast_port_query_expired, (unsigned long)port);
822 void br_multicast_del_port(struct net_bridge_port *port)
824 del_timer_sync(&port->multicast_router_timer);
827 static void __br_multicast_enable_port(struct net_bridge_port *port)
829 port->multicast_startup_queries_sent = 0;
831 if (try_to_del_timer_sync(&port->multicast_query_timer) >= 0 ||
832 del_timer(&port->multicast_query_timer))
833 mod_timer(&port->multicast_query_timer, jiffies);
836 void br_multicast_enable_port(struct net_bridge_port *port)
838 struct net_bridge *br = port->br;
840 spin_lock(&br->multicast_lock);
841 if (br->multicast_disabled || !netif_running(br->dev))
844 __br_multicast_enable_port(port);
847 spin_unlock(&br->multicast_lock);
850 void br_multicast_disable_port(struct net_bridge_port *port)
852 struct net_bridge *br = port->br;
853 struct net_bridge_port_group *pg;
854 struct hlist_node *p, *n;
856 spin_lock(&br->multicast_lock);
857 hlist_for_each_entry_safe(pg, p, n, &port->mglist, mglist)
858 br_multicast_del_pg(br, pg);
860 if (!hlist_unhashed(&port->rlist))
861 hlist_del_init_rcu(&port->rlist);
862 del_timer(&port->multicast_router_timer);
863 del_timer(&port->multicast_query_timer);
864 spin_unlock(&br->multicast_lock);
867 static int br_ip4_multicast_igmp3_report(struct net_bridge *br,
868 struct net_bridge_port *port,
871 struct igmpv3_report *ih;
872 struct igmpv3_grec *grec;
880 if (!pskb_may_pull(skb, sizeof(*ih)))
883 ih = igmpv3_report_hdr(skb);
884 num = ntohs(ih->ngrec);
887 for (i = 0; i < num; i++) {
888 len += sizeof(*grec);
889 if (!pskb_may_pull(skb, len))
892 grec = (void *)(skb->data + len - sizeof(*grec));
893 group = grec->grec_mca;
894 type = grec->grec_type;
896 len += ntohs(grec->grec_nsrcs) * 4;
897 if (!pskb_may_pull(skb, len))
900 /* We treat this as an IGMPv2 report for now. */
902 case IGMPV3_MODE_IS_INCLUDE:
903 case IGMPV3_MODE_IS_EXCLUDE:
904 case IGMPV3_CHANGE_TO_INCLUDE:
905 case IGMPV3_CHANGE_TO_EXCLUDE:
906 case IGMPV3_ALLOW_NEW_SOURCES:
907 case IGMPV3_BLOCK_OLD_SOURCES:
914 err = br_ip4_multicast_add_group(br, port, group);
922 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
923 static int br_ip6_multicast_mld2_report(struct net_bridge *br,
924 struct net_bridge_port *port,
927 struct icmp6hdr *icmp6h;
928 struct mld2_grec *grec;
934 if (!pskb_may_pull(skb, sizeof(*icmp6h)))
937 icmp6h = icmp6_hdr(skb);
938 num = ntohs(icmp6h->icmp6_dataun.un_data16[1]);
939 len = sizeof(*icmp6h);
941 for (i = 0; i < num; i++) {
942 __be16 *nsrcs, _nsrcs;
944 nsrcs = skb_header_pointer(skb,
945 len + offsetof(struct mld2_grec,
947 sizeof(_nsrcs), &_nsrcs);
951 if (!pskb_may_pull(skb,
952 len + sizeof(*grec) +
953 sizeof(struct in6_addr) * ntohs(*nsrcs)))
956 grec = (struct mld2_grec *)(skb->data + len);
957 len += sizeof(*grec) +
958 sizeof(struct in6_addr) * ntohs(*nsrcs);
960 /* We treat these as MLDv1 reports for now. */
961 switch (grec->grec_type) {
962 case MLD2_MODE_IS_INCLUDE:
963 case MLD2_MODE_IS_EXCLUDE:
964 case MLD2_CHANGE_TO_INCLUDE:
965 case MLD2_CHANGE_TO_EXCLUDE:
966 case MLD2_ALLOW_NEW_SOURCES:
967 case MLD2_BLOCK_OLD_SOURCES:
974 err = br_ip6_multicast_add_group(br, port, &grec->grec_mca);
984 * Add port to rotuer_list
985 * list is maintained ordered by pointer value
986 * and locked by br->multicast_lock and RCU
988 static void br_multicast_add_router(struct net_bridge *br,
989 struct net_bridge_port *port)
991 struct net_bridge_port *p;
992 struct hlist_node *n, *slot = NULL;
994 hlist_for_each_entry(p, n, &br->router_list, rlist) {
995 if ((unsigned long) port >= (unsigned long) p)
1001 hlist_add_after_rcu(slot, &port->rlist);
1003 hlist_add_head_rcu(&port->rlist, &br->router_list);
1006 static void br_multicast_mark_router(struct net_bridge *br,
1007 struct net_bridge_port *port)
1009 unsigned long now = jiffies;
1012 if (br->multicast_router == 1)
1013 mod_timer(&br->multicast_router_timer,
1014 now + br->multicast_querier_interval);
1018 if (port->multicast_router != 1)
1021 if (!hlist_unhashed(&port->rlist))
1024 br_multicast_add_router(br, port);
1027 mod_timer(&port->multicast_router_timer,
1028 now + br->multicast_querier_interval);
1031 static void br_multicast_query_received(struct net_bridge *br,
1032 struct net_bridge_port *port,
1036 mod_timer(&br->multicast_querier_timer,
1037 jiffies + br->multicast_querier_interval);
1038 else if (timer_pending(&br->multicast_querier_timer))
1041 br_multicast_mark_router(br, port);
1044 static int br_ip4_multicast_query(struct net_bridge *br,
1045 struct net_bridge_port *port,
1046 struct sk_buff *skb)
1048 const struct iphdr *iph = ip_hdr(skb);
1049 struct igmphdr *ih = igmp_hdr(skb);
1050 struct net_bridge_mdb_entry *mp;
1051 struct igmpv3_query *ih3;
1052 struct net_bridge_port_group *p;
1053 struct net_bridge_port_group __rcu **pp;
1054 unsigned long max_delay;
1055 unsigned long now = jiffies;
1059 spin_lock(&br->multicast_lock);
1060 if (!netif_running(br->dev) ||
1061 (port && port->state == BR_STATE_DISABLED))
1064 br_multicast_query_received(br, port, !!iph->saddr);
1068 if (skb->len == sizeof(*ih)) {
1069 max_delay = ih->code * (HZ / IGMP_TIMER_SCALE);
1072 max_delay = 10 * HZ;
1076 if (!pskb_may_pull(skb, sizeof(struct igmpv3_query))) {
1081 ih3 = igmpv3_query_hdr(skb);
1085 max_delay = ih3->code ?
1086 IGMPV3_MRC(ih3->code) * (HZ / IGMP_TIMER_SCALE) : 1;
1092 mp = br_mdb_ip4_get(mlock_dereference(br->mdb, br), group);
1096 max_delay *= br->multicast_last_member_count;
1099 (timer_pending(&mp->timer) ?
1100 time_after(mp->timer.expires, now + max_delay) :
1101 try_to_del_timer_sync(&mp->timer) >= 0))
1102 mod_timer(&mp->timer, now + max_delay);
1104 for (pp = &mp->ports;
1105 (p = mlock_dereference(*pp, br)) != NULL;
1107 if (timer_pending(&p->timer) ?
1108 time_after(p->timer.expires, now + max_delay) :
1109 try_to_del_timer_sync(&p->timer) >= 0)
1110 mod_timer(&p->timer, now + max_delay);
1114 spin_unlock(&br->multicast_lock);
1118 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
1119 static int br_ip6_multicast_query(struct net_bridge *br,
1120 struct net_bridge_port *port,
1121 struct sk_buff *skb)
1123 const struct ipv6hdr *ip6h = ipv6_hdr(skb);
1124 struct mld_msg *mld = (struct mld_msg *) icmp6_hdr(skb);
1125 struct net_bridge_mdb_entry *mp;
1126 struct mld2_query *mld2q;
1127 struct net_bridge_port_group *p;
1128 struct net_bridge_port_group __rcu **pp;
1129 unsigned long max_delay;
1130 unsigned long now = jiffies;
1131 const struct in6_addr *group = NULL;
1134 spin_lock(&br->multicast_lock);
1135 if (!netif_running(br->dev) ||
1136 (port && port->state == BR_STATE_DISABLED))
1139 br_multicast_query_received(br, port, !ipv6_addr_any(&ip6h->saddr));
1141 if (skb->len == sizeof(*mld)) {
1142 if (!pskb_may_pull(skb, sizeof(*mld))) {
1146 mld = (struct mld_msg *) icmp6_hdr(skb);
1147 max_delay = msecs_to_jiffies(htons(mld->mld_maxdelay));
1149 group = &mld->mld_mca;
1150 } else if (skb->len >= sizeof(*mld2q)) {
1151 if (!pskb_may_pull(skb, sizeof(*mld2q))) {
1155 mld2q = (struct mld2_query *)icmp6_hdr(skb);
1156 if (!mld2q->mld2q_nsrcs)
1157 group = &mld2q->mld2q_mca;
1159 max_delay = max(msecs_to_jiffies(MLDV2_MRC(ntohs(mld2q->mld2q_mrc))), 1UL);
1165 mp = br_mdb_ip6_get(mlock_dereference(br->mdb, br), group);
1169 max_delay *= br->multicast_last_member_count;
1171 (timer_pending(&mp->timer) ?
1172 time_after(mp->timer.expires, now + max_delay) :
1173 try_to_del_timer_sync(&mp->timer) >= 0))
1174 mod_timer(&mp->timer, now + max_delay);
1176 for (pp = &mp->ports;
1177 (p = mlock_dereference(*pp, br)) != NULL;
1179 if (timer_pending(&p->timer) ?
1180 time_after(p->timer.expires, now + max_delay) :
1181 try_to_del_timer_sync(&p->timer) >= 0)
1182 mod_timer(&p->timer, now + max_delay);
1186 spin_unlock(&br->multicast_lock);
1191 static void br_multicast_leave_group(struct net_bridge *br,
1192 struct net_bridge_port *port,
1193 struct br_ip *group)
1195 struct net_bridge_mdb_htable *mdb;
1196 struct net_bridge_mdb_entry *mp;
1197 struct net_bridge_port_group *p;
1201 spin_lock(&br->multicast_lock);
1202 if (!netif_running(br->dev) ||
1203 (port && port->state == BR_STATE_DISABLED) ||
1204 timer_pending(&br->multicast_querier_timer))
1207 mdb = mlock_dereference(br->mdb, br);
1208 mp = br_mdb_ip_get(mdb, group);
1213 time = now + br->multicast_last_member_count *
1214 br->multicast_last_member_interval;
1218 (timer_pending(&mp->timer) ?
1219 time_after(mp->timer.expires, time) :
1220 try_to_del_timer_sync(&mp->timer) >= 0)) {
1221 mod_timer(&mp->timer, time);
1227 for (p = mlock_dereference(mp->ports, br);
1229 p = mlock_dereference(p->next, br)) {
1230 if (p->port != port)
1233 if (!hlist_unhashed(&p->mglist) &&
1234 (timer_pending(&p->timer) ?
1235 time_after(p->timer.expires, time) :
1236 try_to_del_timer_sync(&p->timer) >= 0)) {
1237 mod_timer(&p->timer, time);
1244 spin_unlock(&br->multicast_lock);
1247 static void br_ip4_multicast_leave_group(struct net_bridge *br,
1248 struct net_bridge_port *port,
1251 struct br_ip br_group;
1253 if (ipv4_is_local_multicast(group))
1256 br_group.u.ip4 = group;
1257 br_group.proto = htons(ETH_P_IP);
1259 br_multicast_leave_group(br, port, &br_group);
1262 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
1263 static void br_ip6_multicast_leave_group(struct net_bridge *br,
1264 struct net_bridge_port *port,
1265 const struct in6_addr *group)
1267 struct br_ip br_group;
1269 if (!ipv6_is_transient_multicast(group))
1272 ipv6_addr_copy(&br_group.u.ip6, group);
1273 br_group.proto = htons(ETH_P_IPV6);
1275 br_multicast_leave_group(br, port, &br_group);
1279 static int br_multicast_ipv4_rcv(struct net_bridge *br,
1280 struct net_bridge_port *port,
1281 struct sk_buff *skb)
1283 struct sk_buff *skb2 = skb;
1284 const struct iphdr *iph;
1290 /* We treat OOM as packet loss for now. */
1291 if (!pskb_may_pull(skb, sizeof(*iph)))
1296 if (iph->ihl < 5 || iph->version != 4)
1299 if (!pskb_may_pull(skb, ip_hdrlen(skb)))
1304 if (unlikely(ip_fast_csum((u8 *)iph, iph->ihl)))
1307 if (iph->protocol != IPPROTO_IGMP) {
1308 if ((iph->daddr & IGMP_LOCAL_GROUP_MASK) != IGMP_LOCAL_GROUP)
1309 BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1313 len = ntohs(iph->tot_len);
1314 if (skb->len < len || len < ip_hdrlen(skb))
1317 if (skb->len > len) {
1318 skb2 = skb_clone(skb, GFP_ATOMIC);
1322 err = pskb_trim_rcsum(skb2, len);
1327 len -= ip_hdrlen(skb2);
1328 offset = skb_network_offset(skb2) + ip_hdrlen(skb2);
1329 __skb_pull(skb2, offset);
1330 skb_reset_transport_header(skb2);
1333 if (!pskb_may_pull(skb2, sizeof(*ih)))
1336 switch (skb2->ip_summed) {
1337 case CHECKSUM_COMPLETE:
1338 if (!csum_fold(skb2->csum))
1343 if (skb_checksum_complete(skb2))
1349 BR_INPUT_SKB_CB(skb)->igmp = 1;
1350 ih = igmp_hdr(skb2);
1353 case IGMP_HOST_MEMBERSHIP_REPORT:
1354 case IGMPV2_HOST_MEMBERSHIP_REPORT:
1355 BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1356 err = br_ip4_multicast_add_group(br, port, ih->group);
1358 case IGMPV3_HOST_MEMBERSHIP_REPORT:
1359 err = br_ip4_multicast_igmp3_report(br, port, skb2);
1361 case IGMP_HOST_MEMBERSHIP_QUERY:
1362 err = br_ip4_multicast_query(br, port, skb2);
1364 case IGMP_HOST_LEAVE_MESSAGE:
1365 br_ip4_multicast_leave_group(br, port, ih->group);
1370 __skb_push(skb2, offset);
1377 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
1378 static int br_multicast_ipv6_rcv(struct net_bridge *br,
1379 struct net_bridge_port *port,
1380 struct sk_buff *skb)
1382 struct sk_buff *skb2;
1383 const struct ipv6hdr *ip6h;
1390 if (!pskb_may_pull(skb, sizeof(*ip6h)))
1393 ip6h = ipv6_hdr(skb);
1396 * We're interested in MLD messages only.
1398 * - MLD has always Router Alert hop-by-hop option
1399 * - But we do not support jumbrograms.
1401 if (ip6h->version != 6 ||
1402 ip6h->nexthdr != IPPROTO_HOPOPTS ||
1403 ip6h->payload_len == 0)
1406 len = ntohs(ip6h->payload_len) + sizeof(*ip6h);
1410 nexthdr = ip6h->nexthdr;
1411 offset = ipv6_skip_exthdr(skb, sizeof(*ip6h), &nexthdr);
1413 if (offset < 0 || nexthdr != IPPROTO_ICMPV6)
1416 /* Okay, we found ICMPv6 header */
1417 skb2 = skb_clone(skb, GFP_ATOMIC);
1422 if (!pskb_may_pull(skb2, offset + sizeof(struct icmp6hdr)))
1425 len -= offset - skb_network_offset(skb2);
1427 __skb_pull(skb2, offset);
1428 skb_reset_transport_header(skb2);
1429 skb_postpull_rcsum(skb2, skb_network_header(skb2),
1430 skb_network_header_len(skb2));
1432 icmp6_type = icmp6_hdr(skb2)->icmp6_type;
1434 switch (icmp6_type) {
1435 case ICMPV6_MGM_QUERY:
1436 case ICMPV6_MGM_REPORT:
1437 case ICMPV6_MGM_REDUCTION:
1438 case ICMPV6_MLD2_REPORT:
1445 /* Okay, we found MLD message. Check further. */
1446 if (skb2->len > len) {
1447 err = pskb_trim_rcsum(skb2, len);
1453 ip6h = ipv6_hdr(skb2);
1455 switch (skb2->ip_summed) {
1456 case CHECKSUM_COMPLETE:
1457 if (!csum_ipv6_magic(&ip6h->saddr, &ip6h->daddr, skb2->len,
1458 IPPROTO_ICMPV6, skb2->csum))
1462 skb2->csum = ~csum_unfold(csum_ipv6_magic(&ip6h->saddr,
1465 IPPROTO_ICMPV6, 0));
1466 if (__skb_checksum_complete(skb2))
1472 BR_INPUT_SKB_CB(skb)->igmp = 1;
1474 switch (icmp6_type) {
1475 case ICMPV6_MGM_REPORT:
1477 struct mld_msg *mld;
1478 if (!pskb_may_pull(skb2, sizeof(*mld))) {
1482 mld = (struct mld_msg *)skb_transport_header(skb2);
1483 BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1484 err = br_ip6_multicast_add_group(br, port, &mld->mld_mca);
1487 case ICMPV6_MLD2_REPORT:
1488 err = br_ip6_multicast_mld2_report(br, port, skb2);
1490 case ICMPV6_MGM_QUERY:
1491 err = br_ip6_multicast_query(br, port, skb2);
1493 case ICMPV6_MGM_REDUCTION:
1495 struct mld_msg *mld;
1496 if (!pskb_may_pull(skb2, sizeof(*mld))) {
1500 mld = (struct mld_msg *)skb_transport_header(skb2);
1501 br_ip6_multicast_leave_group(br, port, &mld->mld_mca);
1511 int br_multicast_rcv(struct net_bridge *br, struct net_bridge_port *port,
1512 struct sk_buff *skb)
1514 BR_INPUT_SKB_CB(skb)->igmp = 0;
1515 BR_INPUT_SKB_CB(skb)->mrouters_only = 0;
1517 if (br->multicast_disabled)
1520 switch (skb->protocol) {
1521 case htons(ETH_P_IP):
1522 return br_multicast_ipv4_rcv(br, port, skb);
1523 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
1524 case htons(ETH_P_IPV6):
1525 return br_multicast_ipv6_rcv(br, port, skb);
1532 static void br_multicast_query_expired(unsigned long data)
1534 struct net_bridge *br = (void *)data;
1536 spin_lock(&br->multicast_lock);
1537 if (br->multicast_startup_queries_sent <
1538 br->multicast_startup_query_count)
1539 br->multicast_startup_queries_sent++;
1541 br_multicast_send_query(br, NULL, br->multicast_startup_queries_sent);
1543 spin_unlock(&br->multicast_lock);
1546 void br_multicast_init(struct net_bridge *br)
1548 br->hash_elasticity = 4;
1551 br->multicast_router = 1;
1552 br->multicast_last_member_count = 2;
1553 br->multicast_startup_query_count = 2;
1555 br->multicast_last_member_interval = HZ;
1556 br->multicast_query_response_interval = 10 * HZ;
1557 br->multicast_startup_query_interval = 125 * HZ / 4;
1558 br->multicast_query_interval = 125 * HZ;
1559 br->multicast_querier_interval = 255 * HZ;
1560 br->multicast_membership_interval = 260 * HZ;
1562 spin_lock_init(&br->multicast_lock);
1563 setup_timer(&br->multicast_router_timer,
1564 br_multicast_local_router_expired, 0);
1565 setup_timer(&br->multicast_querier_timer,
1566 br_multicast_local_router_expired, 0);
1567 setup_timer(&br->multicast_query_timer, br_multicast_query_expired,
1571 void br_multicast_open(struct net_bridge *br)
1573 br->multicast_startup_queries_sent = 0;
1575 if (br->multicast_disabled)
1578 mod_timer(&br->multicast_query_timer, jiffies);
1581 void br_multicast_stop(struct net_bridge *br)
1583 struct net_bridge_mdb_htable *mdb;
1584 struct net_bridge_mdb_entry *mp;
1585 struct hlist_node *p, *n;
1589 del_timer_sync(&br->multicast_router_timer);
1590 del_timer_sync(&br->multicast_querier_timer);
1591 del_timer_sync(&br->multicast_query_timer);
1593 spin_lock_bh(&br->multicast_lock);
1594 mdb = mlock_dereference(br->mdb, br);
1601 for (i = 0; i < mdb->max; i++) {
1602 hlist_for_each_entry_safe(mp, p, n, &mdb->mhash[i],
1604 del_timer(&mp->timer);
1605 call_rcu_bh(&mp->rcu, br_multicast_free_group);
1610 spin_unlock_bh(&br->multicast_lock);
1612 spin_lock_bh(&br->multicast_lock);
1617 call_rcu_bh(&mdb->rcu, br_mdb_free);
1620 spin_unlock_bh(&br->multicast_lock);
1623 int br_multicast_set_router(struct net_bridge *br, unsigned long val)
1627 spin_lock_bh(&br->multicast_lock);
1628 if (!netif_running(br->dev))
1634 del_timer(&br->multicast_router_timer);
1637 br->multicast_router = val;
1647 spin_unlock_bh(&br->multicast_lock);
1652 int br_multicast_set_port_router(struct net_bridge_port *p, unsigned long val)
1654 struct net_bridge *br = p->br;
1657 spin_lock(&br->multicast_lock);
1658 if (!netif_running(br->dev) || p->state == BR_STATE_DISABLED)
1665 p->multicast_router = val;
1668 if (val < 2 && !hlist_unhashed(&p->rlist))
1669 hlist_del_init_rcu(&p->rlist);
1674 del_timer(&p->multicast_router_timer);
1679 br_multicast_add_router(br, p);
1688 spin_unlock(&br->multicast_lock);
1693 int br_multicast_toggle(struct net_bridge *br, unsigned long val)
1695 struct net_bridge_port *port;
1697 struct net_bridge_mdb_htable *mdb;
1699 spin_lock_bh(&br->multicast_lock);
1700 if (br->multicast_disabled == !val)
1703 br->multicast_disabled = !val;
1704 if (br->multicast_disabled)
1707 if (!netif_running(br->dev))
1710 mdb = mlock_dereference(br->mdb, br);
1715 br->multicast_disabled = !!val;
1719 err = br_mdb_rehash(&br->mdb, mdb->max,
1720 br->hash_elasticity);
1725 br_multicast_open(br);
1726 list_for_each_entry(port, &br->port_list, list) {
1727 if (port->state == BR_STATE_DISABLED ||
1728 port->state == BR_STATE_BLOCKING)
1731 __br_multicast_enable_port(port);
1735 spin_unlock_bh(&br->multicast_lock);
1740 int br_multicast_set_hash_max(struct net_bridge *br, unsigned long val)
1744 struct net_bridge_mdb_htable *mdb;
1746 spin_lock(&br->multicast_lock);
1747 if (!netif_running(br->dev))
1751 if (!is_power_of_2(val))
1754 mdb = mlock_dereference(br->mdb, br);
1755 if (mdb && val < mdb->size)
1771 err = br_mdb_rehash(&br->mdb, br->hash_max,
1772 br->hash_elasticity);
1778 spin_unlock(&br->multicast_lock);