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)
44 return a->u.ip4 == b->u.ip4;
45 #if IS_ENABLED(CONFIG_IPV6)
46 case htons(ETH_P_IPV6):
47 return ipv6_addr_equal(&a->u.ip6, &b->u.ip6);
53 static inline int __br_ip4_hash(struct net_bridge_mdb_htable *mdb, __be32 ip)
55 return jhash_1word(mdb->secret, (__force u32)ip) & (mdb->max - 1);
58 #if IS_ENABLED(CONFIG_IPV6)
59 static inline int __br_ip6_hash(struct net_bridge_mdb_htable *mdb,
60 const struct in6_addr *ip)
62 return jhash2((__force u32 *)ip->s6_addr32, 4, mdb->secret) & (mdb->max - 1);
66 static inline int br_ip_hash(struct net_bridge_mdb_htable *mdb,
71 return __br_ip4_hash(mdb, ip->u.ip4);
72 #if IS_ENABLED(CONFIG_IPV6)
73 case htons(ETH_P_IPV6):
74 return __br_ip6_hash(mdb, &ip->u.ip6);
80 static struct net_bridge_mdb_entry *__br_mdb_ip_get(
81 struct net_bridge_mdb_htable *mdb, struct br_ip *dst, int hash)
83 struct net_bridge_mdb_entry *mp;
86 hlist_for_each_entry_rcu(mp, p, &mdb->mhash[hash], hlist[mdb->ver]) {
87 if (br_ip_equal(&mp->addr, dst))
94 struct net_bridge_mdb_entry *br_mdb_ip_get(struct net_bridge_mdb_htable *mdb,
100 return __br_mdb_ip_get(mdb, dst, br_ip_hash(mdb, dst));
103 static struct net_bridge_mdb_entry *br_mdb_ip4_get(
104 struct net_bridge_mdb_htable *mdb, __be32 dst)
109 br_dst.proto = htons(ETH_P_IP);
111 return br_mdb_ip_get(mdb, &br_dst);
114 #if IS_ENABLED(CONFIG_IPV6)
115 static struct net_bridge_mdb_entry *br_mdb_ip6_get(
116 struct net_bridge_mdb_htable *mdb, const struct in6_addr *dst)
121 br_dst.proto = htons(ETH_P_IPV6);
123 return br_mdb_ip_get(mdb, &br_dst);
127 struct net_bridge_mdb_entry *br_mdb_get(struct net_bridge *br,
130 struct net_bridge_mdb_htable *mdb = rcu_dereference(br->mdb);
133 if (br->multicast_disabled)
136 if (BR_INPUT_SKB_CB(skb)->igmp)
139 ip.proto = skb->protocol;
141 switch (skb->protocol) {
142 case htons(ETH_P_IP):
143 ip.u.ip4 = ip_hdr(skb)->daddr;
145 #if IS_ENABLED(CONFIG_IPV6)
146 case htons(ETH_P_IPV6):
147 ip.u.ip6 = ipv6_hdr(skb)->daddr;
154 return br_mdb_ip_get(mdb, &ip);
157 static void br_mdb_free(struct rcu_head *head)
159 struct net_bridge_mdb_htable *mdb =
160 container_of(head, struct net_bridge_mdb_htable, rcu);
161 struct net_bridge_mdb_htable *old = mdb->old;
168 static int br_mdb_copy(struct net_bridge_mdb_htable *new,
169 struct net_bridge_mdb_htable *old,
172 struct net_bridge_mdb_entry *mp;
173 struct hlist_node *p;
178 for (i = 0; i < old->max; i++)
179 hlist_for_each_entry(mp, p, &old->mhash[i], hlist[old->ver])
180 hlist_add_head(&mp->hlist[new->ver],
181 &new->mhash[br_ip_hash(new, &mp->addr)]);
187 for (i = 0; i < new->max; i++) {
189 hlist_for_each_entry(mp, p, &new->mhash[i], hlist[new->ver])
195 return maxlen > elasticity ? -EINVAL : 0;
198 void br_multicast_free_pg(struct rcu_head *head)
200 struct net_bridge_port_group *p =
201 container_of(head, struct net_bridge_port_group, rcu);
206 static void br_multicast_free_group(struct rcu_head *head)
208 struct net_bridge_mdb_entry *mp =
209 container_of(head, struct net_bridge_mdb_entry, rcu);
214 static void br_multicast_group_expired(unsigned long data)
216 struct net_bridge_mdb_entry *mp = (void *)data;
217 struct net_bridge *br = mp->br;
218 struct net_bridge_mdb_htable *mdb;
220 spin_lock(&br->multicast_lock);
221 if (!netif_running(br->dev) || timer_pending(&mp->timer))
229 mdb = mlock_dereference(br->mdb, br);
231 hlist_del_rcu(&mp->hlist[mdb->ver]);
234 call_rcu_bh(&mp->rcu, br_multicast_free_group);
237 spin_unlock(&br->multicast_lock);
240 static void br_multicast_del_pg(struct net_bridge *br,
241 struct net_bridge_port_group *pg)
243 struct net_bridge_mdb_htable *mdb;
244 struct net_bridge_mdb_entry *mp;
245 struct net_bridge_port_group *p;
246 struct net_bridge_port_group __rcu **pp;
248 mdb = mlock_dereference(br->mdb, br);
250 mp = br_mdb_ip_get(mdb, &pg->addr);
254 for (pp = &mp->ports;
255 (p = mlock_dereference(*pp, br)) != NULL;
260 rcu_assign_pointer(*pp, p->next);
261 hlist_del_init(&p->mglist);
262 del_timer(&p->timer);
263 call_rcu_bh(&p->rcu, br_multicast_free_pg);
265 if (!mp->ports && !mp->mglist &&
266 netif_running(br->dev))
267 mod_timer(&mp->timer, jiffies);
275 static void br_multicast_port_group_expired(unsigned long data)
277 struct net_bridge_port_group *pg = (void *)data;
278 struct net_bridge *br = pg->port->br;
280 spin_lock(&br->multicast_lock);
281 if (!netif_running(br->dev) || timer_pending(&pg->timer) ||
282 hlist_unhashed(&pg->mglist) || pg->state & MDB_PERMANENT)
285 br_multicast_del_pg(br, pg);
288 spin_unlock(&br->multicast_lock);
291 static int br_mdb_rehash(struct net_bridge_mdb_htable __rcu **mdbp, int max,
294 struct net_bridge_mdb_htable *old = rcu_dereference_protected(*mdbp, 1);
295 struct net_bridge_mdb_htable *mdb;
298 mdb = kmalloc(sizeof(*mdb), GFP_ATOMIC);
305 mdb->mhash = kzalloc(max * sizeof(*mdb->mhash), GFP_ATOMIC);
311 mdb->size = old ? old->size : 0;
312 mdb->ver = old ? old->ver ^ 1 : 0;
314 if (!old || elasticity)
315 get_random_bytes(&mdb->secret, sizeof(mdb->secret));
317 mdb->secret = old->secret;
322 err = br_mdb_copy(mdb, old, elasticity);
330 call_rcu_bh(&mdb->rcu, br_mdb_free);
333 rcu_assign_pointer(*mdbp, mdb);
338 static struct sk_buff *br_ip4_multicast_alloc_query(struct net_bridge *br,
346 skb = netdev_alloc_skb_ip_align(br->dev, sizeof(*eth) + sizeof(*iph) +
351 skb->protocol = htons(ETH_P_IP);
353 skb_reset_mac_header(skb);
356 memcpy(eth->h_source, br->dev->dev_addr, 6);
359 eth->h_dest[2] = 0x5e;
363 eth->h_proto = htons(ETH_P_IP);
364 skb_put(skb, sizeof(*eth));
366 skb_set_network_header(skb, skb->len);
372 iph->tot_len = htons(sizeof(*iph) + sizeof(*ih) + 4);
374 iph->frag_off = htons(IP_DF);
376 iph->protocol = IPPROTO_IGMP;
378 iph->daddr = htonl(INADDR_ALLHOSTS_GROUP);
379 ((u8 *)&iph[1])[0] = IPOPT_RA;
380 ((u8 *)&iph[1])[1] = 4;
381 ((u8 *)&iph[1])[2] = 0;
382 ((u8 *)&iph[1])[3] = 0;
386 skb_set_transport_header(skb, skb->len);
388 ih->type = IGMP_HOST_MEMBERSHIP_QUERY;
389 ih->code = (group ? br->multicast_last_member_interval :
390 br->multicast_query_response_interval) /
391 (HZ / IGMP_TIMER_SCALE);
394 ih->csum = ip_compute_csum((void *)ih, sizeof(struct igmphdr));
395 skb_put(skb, sizeof(*ih));
397 __skb_pull(skb, sizeof(*eth));
403 #if IS_ENABLED(CONFIG_IPV6)
404 static struct sk_buff *br_ip6_multicast_alloc_query(struct net_bridge *br,
405 const struct in6_addr *group)
408 struct ipv6hdr *ip6h;
409 struct mld_msg *mldq;
412 unsigned long interval;
414 skb = netdev_alloc_skb_ip_align(br->dev, sizeof(*eth) + sizeof(*ip6h) +
419 skb->protocol = htons(ETH_P_IPV6);
421 /* Ethernet header */
422 skb_reset_mac_header(skb);
425 memcpy(eth->h_source, br->dev->dev_addr, 6);
426 eth->h_proto = htons(ETH_P_IPV6);
427 skb_put(skb, sizeof(*eth));
429 /* IPv6 header + HbH option */
430 skb_set_network_header(skb, skb->len);
431 ip6h = ipv6_hdr(skb);
433 *(__force __be32 *)ip6h = htonl(0x60000000);
434 ip6h->payload_len = htons(8 + sizeof(*mldq));
435 ip6h->nexthdr = IPPROTO_HOPOPTS;
437 ipv6_addr_set(&ip6h->daddr, htonl(0xff020000), 0, 0, htonl(1));
438 if (ipv6_dev_get_saddr(dev_net(br->dev), br->dev, &ip6h->daddr, 0,
443 ipv6_eth_mc_map(&ip6h->daddr, eth->h_dest);
445 hopopt = (u8 *)(ip6h + 1);
446 hopopt[0] = IPPROTO_ICMPV6; /* next hdr */
447 hopopt[1] = 0; /* length of HbH */
448 hopopt[2] = IPV6_TLV_ROUTERALERT; /* Router Alert */
449 hopopt[3] = 2; /* Length of RA Option */
450 hopopt[4] = 0; /* Type = 0x0000 (MLD) */
452 hopopt[6] = IPV6_TLV_PAD1; /* Pad1 */
453 hopopt[7] = IPV6_TLV_PAD1; /* Pad1 */
455 skb_put(skb, sizeof(*ip6h) + 8);
458 skb_set_transport_header(skb, skb->len);
459 mldq = (struct mld_msg *) icmp6_hdr(skb);
461 interval = ipv6_addr_any(group) ? br->multicast_last_member_interval :
462 br->multicast_query_response_interval;
464 mldq->mld_type = ICMPV6_MGM_QUERY;
467 mldq->mld_maxdelay = htons((u16)jiffies_to_msecs(interval));
468 mldq->mld_reserved = 0;
469 mldq->mld_mca = *group;
472 mldq->mld_cksum = csum_ipv6_magic(&ip6h->saddr, &ip6h->daddr,
473 sizeof(*mldq), IPPROTO_ICMPV6,
476 skb_put(skb, sizeof(*mldq));
478 __skb_pull(skb, sizeof(*eth));
485 static struct sk_buff *br_multicast_alloc_query(struct net_bridge *br,
488 switch (addr->proto) {
489 case htons(ETH_P_IP):
490 return br_ip4_multicast_alloc_query(br, addr->u.ip4);
491 #if IS_ENABLED(CONFIG_IPV6)
492 case htons(ETH_P_IPV6):
493 return br_ip6_multicast_alloc_query(br, &addr->u.ip6);
499 static struct net_bridge_mdb_entry *br_multicast_get_group(
500 struct net_bridge *br, struct net_bridge_port *port,
501 struct br_ip *group, int hash)
503 struct net_bridge_mdb_htable *mdb;
504 struct net_bridge_mdb_entry *mp;
505 struct hlist_node *p;
506 unsigned int count = 0;
511 mdb = rcu_dereference_protected(br->mdb, 1);
512 hlist_for_each_entry(mp, p, &mdb->mhash[hash], hlist[mdb->ver]) {
514 if (unlikely(br_ip_equal(group, &mp->addr)))
521 if (unlikely(count > br->hash_elasticity && count)) {
523 br_info(br, "Multicast hash table "
524 "chain limit reached: %s\n",
525 port ? port->dev->name : br->dev->name);
527 elasticity = br->hash_elasticity;
530 if (mdb->size >= max) {
532 if (unlikely(max > br->hash_max)) {
533 br_warn(br, "Multicast hash table maximum of %d "
534 "reached, disabling snooping: %s\n",
536 port ? port->dev->name : br->dev->name);
539 br->multicast_disabled = 1;
544 if (max > mdb->max || elasticity) {
547 br_info(br, "Multicast hash table "
549 port ? port->dev->name : br->dev->name);
554 err = br_mdb_rehash(&br->mdb, max, elasticity);
556 br_warn(br, "Cannot rehash multicast "
557 "hash table, disabling snooping: %s, %d, %d\n",
558 port ? port->dev->name : br->dev->name,
574 struct net_bridge_mdb_entry *br_multicast_new_group(struct net_bridge *br,
575 struct net_bridge_port *port, struct br_ip *group)
577 struct net_bridge_mdb_htable *mdb;
578 struct net_bridge_mdb_entry *mp;
582 mdb = rcu_dereference_protected(br->mdb, 1);
584 err = br_mdb_rehash(&br->mdb, BR_HASH_SIZE, 0);
590 hash = br_ip_hash(mdb, group);
591 mp = br_multicast_get_group(br, port, group, hash);
592 switch (PTR_ERR(mp)) {
598 mdb = rcu_dereference_protected(br->mdb, 1);
599 hash = br_ip_hash(mdb, group);
606 mp = kzalloc(sizeof(*mp), GFP_ATOMIC);
608 return ERR_PTR(-ENOMEM);
612 setup_timer(&mp->timer, br_multicast_group_expired,
615 hlist_add_head_rcu(&mp->hlist[mdb->ver], &mdb->mhash[hash]);
622 struct net_bridge_port_group *br_multicast_new_port_group(
623 struct net_bridge_port *port,
625 struct net_bridge_port_group __rcu *next,
628 struct net_bridge_port_group *p;
630 p = kzalloc(sizeof(*p), GFP_ATOMIC);
637 rcu_assign_pointer(p->next, next);
638 hlist_add_head(&p->mglist, &port->mglist);
639 setup_timer(&p->timer, br_multicast_port_group_expired,
644 static int br_multicast_add_group(struct net_bridge *br,
645 struct net_bridge_port *port,
648 struct net_bridge_mdb_entry *mp;
649 struct net_bridge_port_group *p;
650 struct net_bridge_port_group __rcu **pp;
651 unsigned long now = jiffies;
654 spin_lock(&br->multicast_lock);
655 if (!netif_running(br->dev) ||
656 (port && port->state == BR_STATE_DISABLED))
659 mp = br_multicast_new_group(br, port, group);
666 mod_timer(&mp->timer, now + br->multicast_membership_interval);
670 for (pp = &mp->ports;
671 (p = mlock_dereference(*pp, br)) != NULL;
675 if ((unsigned long)p->port < (unsigned long)port)
679 p = br_multicast_new_port_group(port, group, *pp, MDB_TEMPORARY);
682 rcu_assign_pointer(*pp, p);
683 br_mdb_notify(br->dev, port, group, RTM_NEWMDB);
686 mod_timer(&p->timer, now + br->multicast_membership_interval);
691 spin_unlock(&br->multicast_lock);
695 static int br_ip4_multicast_add_group(struct net_bridge *br,
696 struct net_bridge_port *port,
699 struct br_ip br_group;
701 if (ipv4_is_local_multicast(group))
704 br_group.u.ip4 = group;
705 br_group.proto = htons(ETH_P_IP);
707 return br_multicast_add_group(br, port, &br_group);
710 #if IS_ENABLED(CONFIG_IPV6)
711 static int br_ip6_multicast_add_group(struct net_bridge *br,
712 struct net_bridge_port *port,
713 const struct in6_addr *group)
715 struct br_ip br_group;
717 if (!ipv6_is_transient_multicast(group))
720 br_group.u.ip6 = *group;
721 br_group.proto = htons(ETH_P_IPV6);
723 return br_multicast_add_group(br, port, &br_group);
727 static void br_multicast_router_expired(unsigned long data)
729 struct net_bridge_port *port = (void *)data;
730 struct net_bridge *br = port->br;
732 spin_lock(&br->multicast_lock);
733 if (port->multicast_router != 1 ||
734 timer_pending(&port->multicast_router_timer) ||
735 hlist_unhashed(&port->rlist))
738 hlist_del_init_rcu(&port->rlist);
741 spin_unlock(&br->multicast_lock);
744 static void br_multicast_local_router_expired(unsigned long data)
748 static void br_multicast_querier_expired(unsigned long data)
750 struct net_bridge *br = (void *)data;
752 spin_lock(&br->multicast_lock);
753 if (!netif_running(br->dev) || br->multicast_disabled)
756 br_multicast_start_querier(br);
759 spin_unlock(&br->multicast_lock);
762 static void __br_multicast_send_query(struct net_bridge *br,
763 struct net_bridge_port *port,
768 skb = br_multicast_alloc_query(br, ip);
773 __skb_push(skb, sizeof(struct ethhdr));
774 skb->dev = port->dev;
775 NF_HOOK(NFPROTO_BRIDGE, NF_BR_LOCAL_OUT, skb, NULL, skb->dev,
781 static void br_multicast_send_query(struct net_bridge *br,
782 struct net_bridge_port *port, u32 sent)
785 struct br_ip br_group;
787 if (!netif_running(br->dev) || br->multicast_disabled ||
788 !br->multicast_querier ||
789 timer_pending(&br->multicast_querier_timer))
792 memset(&br_group.u, 0, sizeof(br_group.u));
794 br_group.proto = htons(ETH_P_IP);
795 __br_multicast_send_query(br, port, &br_group);
797 #if IS_ENABLED(CONFIG_IPV6)
798 br_group.proto = htons(ETH_P_IPV6);
799 __br_multicast_send_query(br, port, &br_group);
803 time += sent < br->multicast_startup_query_count ?
804 br->multicast_startup_query_interval :
805 br->multicast_query_interval;
806 mod_timer(port ? &port->multicast_query_timer :
807 &br->multicast_query_timer, time);
810 static void br_multicast_port_query_expired(unsigned long data)
812 struct net_bridge_port *port = (void *)data;
813 struct net_bridge *br = port->br;
815 spin_lock(&br->multicast_lock);
816 if (port->state == BR_STATE_DISABLED ||
817 port->state == BR_STATE_BLOCKING)
820 if (port->multicast_startup_queries_sent <
821 br->multicast_startup_query_count)
822 port->multicast_startup_queries_sent++;
824 br_multicast_send_query(port->br, port,
825 port->multicast_startup_queries_sent);
828 spin_unlock(&br->multicast_lock);
831 void br_multicast_add_port(struct net_bridge_port *port)
833 port->multicast_router = 1;
835 setup_timer(&port->multicast_router_timer, br_multicast_router_expired,
836 (unsigned long)port);
837 setup_timer(&port->multicast_query_timer,
838 br_multicast_port_query_expired, (unsigned long)port);
841 void br_multicast_del_port(struct net_bridge_port *port)
843 del_timer_sync(&port->multicast_router_timer);
846 static void __br_multicast_enable_port(struct net_bridge_port *port)
848 port->multicast_startup_queries_sent = 0;
850 if (try_to_del_timer_sync(&port->multicast_query_timer) >= 0 ||
851 del_timer(&port->multicast_query_timer))
852 mod_timer(&port->multicast_query_timer, jiffies);
855 void br_multicast_enable_port(struct net_bridge_port *port)
857 struct net_bridge *br = port->br;
859 spin_lock(&br->multicast_lock);
860 if (br->multicast_disabled || !netif_running(br->dev))
863 __br_multicast_enable_port(port);
866 spin_unlock(&br->multicast_lock);
869 void br_multicast_disable_port(struct net_bridge_port *port)
871 struct net_bridge *br = port->br;
872 struct net_bridge_port_group *pg;
873 struct hlist_node *p, *n;
875 spin_lock(&br->multicast_lock);
876 hlist_for_each_entry_safe(pg, p, n, &port->mglist, mglist)
877 br_multicast_del_pg(br, pg);
879 if (!hlist_unhashed(&port->rlist))
880 hlist_del_init_rcu(&port->rlist);
881 del_timer(&port->multicast_router_timer);
882 del_timer(&port->multicast_query_timer);
883 spin_unlock(&br->multicast_lock);
886 static int br_ip4_multicast_igmp3_report(struct net_bridge *br,
887 struct net_bridge_port *port,
890 struct igmpv3_report *ih;
891 struct igmpv3_grec *grec;
899 if (!pskb_may_pull(skb, sizeof(*ih)))
902 ih = igmpv3_report_hdr(skb);
903 num = ntohs(ih->ngrec);
906 for (i = 0; i < num; i++) {
907 len += sizeof(*grec);
908 if (!pskb_may_pull(skb, len))
911 grec = (void *)(skb->data + len - sizeof(*grec));
912 group = grec->grec_mca;
913 type = grec->grec_type;
915 len += ntohs(grec->grec_nsrcs) * 4;
916 if (!pskb_may_pull(skb, len))
919 /* We treat this as an IGMPv2 report for now. */
921 case IGMPV3_MODE_IS_INCLUDE:
922 case IGMPV3_MODE_IS_EXCLUDE:
923 case IGMPV3_CHANGE_TO_INCLUDE:
924 case IGMPV3_CHANGE_TO_EXCLUDE:
925 case IGMPV3_ALLOW_NEW_SOURCES:
926 case IGMPV3_BLOCK_OLD_SOURCES:
933 err = br_ip4_multicast_add_group(br, port, group);
941 #if IS_ENABLED(CONFIG_IPV6)
942 static int br_ip6_multicast_mld2_report(struct net_bridge *br,
943 struct net_bridge_port *port,
946 struct icmp6hdr *icmp6h;
947 struct mld2_grec *grec;
953 if (!pskb_may_pull(skb, sizeof(*icmp6h)))
956 icmp6h = icmp6_hdr(skb);
957 num = ntohs(icmp6h->icmp6_dataun.un_data16[1]);
958 len = sizeof(*icmp6h);
960 for (i = 0; i < num; i++) {
961 __be16 *nsrcs, _nsrcs;
963 nsrcs = skb_header_pointer(skb,
964 len + offsetof(struct mld2_grec,
966 sizeof(_nsrcs), &_nsrcs);
970 if (!pskb_may_pull(skb,
971 len + sizeof(*grec) +
972 sizeof(struct in6_addr) * ntohs(*nsrcs)))
975 grec = (struct mld2_grec *)(skb->data + len);
976 len += sizeof(*grec) +
977 sizeof(struct in6_addr) * ntohs(*nsrcs);
979 /* We treat these as MLDv1 reports for now. */
980 switch (grec->grec_type) {
981 case MLD2_MODE_IS_INCLUDE:
982 case MLD2_MODE_IS_EXCLUDE:
983 case MLD2_CHANGE_TO_INCLUDE:
984 case MLD2_CHANGE_TO_EXCLUDE:
985 case MLD2_ALLOW_NEW_SOURCES:
986 case MLD2_BLOCK_OLD_SOURCES:
993 err = br_ip6_multicast_add_group(br, port, &grec->grec_mca);
1003 * Add port to rotuer_list
1004 * list is maintained ordered by pointer value
1005 * and locked by br->multicast_lock and RCU
1007 static void br_multicast_add_router(struct net_bridge *br,
1008 struct net_bridge_port *port)
1010 struct net_bridge_port *p;
1011 struct hlist_node *n, *slot = NULL;
1013 hlist_for_each_entry(p, n, &br->router_list, rlist) {
1014 if ((unsigned long) port >= (unsigned long) p)
1020 hlist_add_after_rcu(slot, &port->rlist);
1022 hlist_add_head_rcu(&port->rlist, &br->router_list);
1025 static void br_multicast_mark_router(struct net_bridge *br,
1026 struct net_bridge_port *port)
1028 unsigned long now = jiffies;
1031 if (br->multicast_router == 1)
1032 mod_timer(&br->multicast_router_timer,
1033 now + br->multicast_querier_interval);
1037 if (port->multicast_router != 1)
1040 if (!hlist_unhashed(&port->rlist))
1043 br_multicast_add_router(br, port);
1046 mod_timer(&port->multicast_router_timer,
1047 now + br->multicast_querier_interval);
1050 static void br_multicast_query_received(struct net_bridge *br,
1051 struct net_bridge_port *port,
1055 mod_timer(&br->multicast_querier_timer,
1056 jiffies + br->multicast_querier_interval);
1057 else if (timer_pending(&br->multicast_querier_timer))
1060 br_multicast_mark_router(br, port);
1063 static int br_ip4_multicast_query(struct net_bridge *br,
1064 struct net_bridge_port *port,
1065 struct sk_buff *skb)
1067 const struct iphdr *iph = ip_hdr(skb);
1068 struct igmphdr *ih = igmp_hdr(skb);
1069 struct net_bridge_mdb_entry *mp;
1070 struct igmpv3_query *ih3;
1071 struct net_bridge_port_group *p;
1072 struct net_bridge_port_group __rcu **pp;
1073 unsigned long max_delay;
1074 unsigned long now = jiffies;
1078 spin_lock(&br->multicast_lock);
1079 if (!netif_running(br->dev) ||
1080 (port && port->state == BR_STATE_DISABLED))
1083 br_multicast_query_received(br, port, !!iph->saddr);
1087 if (skb->len == sizeof(*ih)) {
1088 max_delay = ih->code * (HZ / IGMP_TIMER_SCALE);
1091 max_delay = 10 * HZ;
1095 if (!pskb_may_pull(skb, sizeof(struct igmpv3_query))) {
1100 ih3 = igmpv3_query_hdr(skb);
1104 max_delay = ih3->code ?
1105 IGMPV3_MRC(ih3->code) * (HZ / IGMP_TIMER_SCALE) : 1;
1111 mp = br_mdb_ip4_get(mlock_dereference(br->mdb, br), group);
1115 max_delay *= br->multicast_last_member_count;
1118 (timer_pending(&mp->timer) ?
1119 time_after(mp->timer.expires, now + max_delay) :
1120 try_to_del_timer_sync(&mp->timer) >= 0))
1121 mod_timer(&mp->timer, now + max_delay);
1123 for (pp = &mp->ports;
1124 (p = mlock_dereference(*pp, br)) != NULL;
1126 if (timer_pending(&p->timer) ?
1127 time_after(p->timer.expires, now + max_delay) :
1128 try_to_del_timer_sync(&p->timer) >= 0)
1129 mod_timer(&p->timer, now + max_delay);
1133 spin_unlock(&br->multicast_lock);
1137 #if IS_ENABLED(CONFIG_IPV6)
1138 static int br_ip6_multicast_query(struct net_bridge *br,
1139 struct net_bridge_port *port,
1140 struct sk_buff *skb)
1142 const struct ipv6hdr *ip6h = ipv6_hdr(skb);
1143 struct mld_msg *mld;
1144 struct net_bridge_mdb_entry *mp;
1145 struct mld2_query *mld2q;
1146 struct net_bridge_port_group *p;
1147 struct net_bridge_port_group __rcu **pp;
1148 unsigned long max_delay;
1149 unsigned long now = jiffies;
1150 const struct in6_addr *group = NULL;
1153 spin_lock(&br->multicast_lock);
1154 if (!netif_running(br->dev) ||
1155 (port && port->state == BR_STATE_DISABLED))
1158 br_multicast_query_received(br, port, !ipv6_addr_any(&ip6h->saddr));
1160 if (skb->len == sizeof(*mld)) {
1161 if (!pskb_may_pull(skb, sizeof(*mld))) {
1165 mld = (struct mld_msg *) icmp6_hdr(skb);
1166 max_delay = msecs_to_jiffies(ntohs(mld->mld_maxdelay));
1168 group = &mld->mld_mca;
1169 } else if (skb->len >= sizeof(*mld2q)) {
1170 if (!pskb_may_pull(skb, sizeof(*mld2q))) {
1174 mld2q = (struct mld2_query *)icmp6_hdr(skb);
1175 if (!mld2q->mld2q_nsrcs)
1176 group = &mld2q->mld2q_mca;
1177 max_delay = mld2q->mld2q_mrc ? MLDV2_MRC(ntohs(mld2q->mld2q_mrc)) : 1;
1183 mp = br_mdb_ip6_get(mlock_dereference(br->mdb, br), group);
1187 max_delay *= br->multicast_last_member_count;
1189 (timer_pending(&mp->timer) ?
1190 time_after(mp->timer.expires, now + max_delay) :
1191 try_to_del_timer_sync(&mp->timer) >= 0))
1192 mod_timer(&mp->timer, now + max_delay);
1194 for (pp = &mp->ports;
1195 (p = mlock_dereference(*pp, br)) != NULL;
1197 if (timer_pending(&p->timer) ?
1198 time_after(p->timer.expires, now + max_delay) :
1199 try_to_del_timer_sync(&p->timer) >= 0)
1200 mod_timer(&p->timer, now + max_delay);
1204 spin_unlock(&br->multicast_lock);
1209 static void br_multicast_leave_group(struct net_bridge *br,
1210 struct net_bridge_port *port,
1211 struct br_ip *group)
1213 struct net_bridge_mdb_htable *mdb;
1214 struct net_bridge_mdb_entry *mp;
1215 struct net_bridge_port_group *p;
1219 spin_lock(&br->multicast_lock);
1220 if (!netif_running(br->dev) ||
1221 (port && port->state == BR_STATE_DISABLED) ||
1222 timer_pending(&br->multicast_querier_timer))
1225 mdb = mlock_dereference(br->mdb, br);
1226 mp = br_mdb_ip_get(mdb, group);
1230 if (port && (port->flags & BR_MULTICAST_FAST_LEAVE)) {
1231 struct net_bridge_port_group __rcu **pp;
1233 for (pp = &mp->ports;
1234 (p = mlock_dereference(*pp, br)) != NULL;
1236 if (p->port != port)
1239 rcu_assign_pointer(*pp, p->next);
1240 hlist_del_init(&p->mglist);
1241 del_timer(&p->timer);
1242 call_rcu_bh(&p->rcu, br_multicast_free_pg);
1243 br_mdb_notify(br->dev, port, group, RTM_DELMDB);
1245 if (!mp->ports && !mp->mglist &&
1246 netif_running(br->dev))
1247 mod_timer(&mp->timer, jiffies);
1253 time = now + br->multicast_last_member_count *
1254 br->multicast_last_member_interval;
1258 (timer_pending(&mp->timer) ?
1259 time_after(mp->timer.expires, time) :
1260 try_to_del_timer_sync(&mp->timer) >= 0)) {
1261 mod_timer(&mp->timer, time);
1267 for (p = mlock_dereference(mp->ports, br);
1269 p = mlock_dereference(p->next, br)) {
1270 if (p->port != port)
1273 if (!hlist_unhashed(&p->mglist) &&
1274 (timer_pending(&p->timer) ?
1275 time_after(p->timer.expires, time) :
1276 try_to_del_timer_sync(&p->timer) >= 0)) {
1277 mod_timer(&p->timer, time);
1284 spin_unlock(&br->multicast_lock);
1287 static void br_ip4_multicast_leave_group(struct net_bridge *br,
1288 struct net_bridge_port *port,
1291 struct br_ip br_group;
1293 if (ipv4_is_local_multicast(group))
1296 br_group.u.ip4 = group;
1297 br_group.proto = htons(ETH_P_IP);
1299 br_multicast_leave_group(br, port, &br_group);
1302 #if IS_ENABLED(CONFIG_IPV6)
1303 static void br_ip6_multicast_leave_group(struct net_bridge *br,
1304 struct net_bridge_port *port,
1305 const struct in6_addr *group)
1307 struct br_ip br_group;
1309 if (!ipv6_is_transient_multicast(group))
1312 br_group.u.ip6 = *group;
1313 br_group.proto = htons(ETH_P_IPV6);
1315 br_multicast_leave_group(br, port, &br_group);
1319 static int br_multicast_ipv4_rcv(struct net_bridge *br,
1320 struct net_bridge_port *port,
1321 struct sk_buff *skb)
1323 struct sk_buff *skb2 = skb;
1324 const struct iphdr *iph;
1327 unsigned int offset;
1330 /* We treat OOM as packet loss for now. */
1331 if (!pskb_may_pull(skb, sizeof(*iph)))
1336 if (iph->ihl < 5 || iph->version != 4)
1339 if (!pskb_may_pull(skb, ip_hdrlen(skb)))
1344 if (unlikely(ip_fast_csum((u8 *)iph, iph->ihl)))
1347 if (iph->protocol != IPPROTO_IGMP) {
1348 if ((iph->daddr & IGMP_LOCAL_GROUP_MASK) != IGMP_LOCAL_GROUP)
1349 BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1353 len = ntohs(iph->tot_len);
1354 if (skb->len < len || len < ip_hdrlen(skb))
1357 if (skb->len > len) {
1358 skb2 = skb_clone(skb, GFP_ATOMIC);
1362 err = pskb_trim_rcsum(skb2, len);
1367 len -= ip_hdrlen(skb2);
1368 offset = skb_network_offset(skb2) + ip_hdrlen(skb2);
1369 __skb_pull(skb2, offset);
1370 skb_reset_transport_header(skb2);
1373 if (!pskb_may_pull(skb2, sizeof(*ih)))
1376 switch (skb2->ip_summed) {
1377 case CHECKSUM_COMPLETE:
1378 if (!csum_fold(skb2->csum))
1383 if (skb_checksum_complete(skb2))
1389 BR_INPUT_SKB_CB(skb)->igmp = 1;
1390 ih = igmp_hdr(skb2);
1393 case IGMP_HOST_MEMBERSHIP_REPORT:
1394 case IGMPV2_HOST_MEMBERSHIP_REPORT:
1395 BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1396 err = br_ip4_multicast_add_group(br, port, ih->group);
1398 case IGMPV3_HOST_MEMBERSHIP_REPORT:
1399 err = br_ip4_multicast_igmp3_report(br, port, skb2);
1401 case IGMP_HOST_MEMBERSHIP_QUERY:
1402 err = br_ip4_multicast_query(br, port, skb2);
1404 case IGMP_HOST_LEAVE_MESSAGE:
1405 br_ip4_multicast_leave_group(br, port, ih->group);
1410 __skb_push(skb2, offset);
1417 #if IS_ENABLED(CONFIG_IPV6)
1418 static int br_multicast_ipv6_rcv(struct net_bridge *br,
1419 struct net_bridge_port *port,
1420 struct sk_buff *skb)
1422 struct sk_buff *skb2;
1423 const struct ipv6hdr *ip6h;
1431 if (!pskb_may_pull(skb, sizeof(*ip6h)))
1434 ip6h = ipv6_hdr(skb);
1437 * We're interested in MLD messages only.
1439 * - MLD has always Router Alert hop-by-hop option
1440 * - But we do not support jumbrograms.
1442 if (ip6h->version != 6 ||
1443 ip6h->nexthdr != IPPROTO_HOPOPTS ||
1444 ip6h->payload_len == 0)
1447 len = ntohs(ip6h->payload_len) + sizeof(*ip6h);
1451 nexthdr = ip6h->nexthdr;
1452 offset = ipv6_skip_exthdr(skb, sizeof(*ip6h), &nexthdr, &frag_off);
1454 if (offset < 0 || nexthdr != IPPROTO_ICMPV6)
1457 /* Okay, we found ICMPv6 header */
1458 skb2 = skb_clone(skb, GFP_ATOMIC);
1463 if (!pskb_may_pull(skb2, offset + sizeof(struct icmp6hdr)))
1466 len -= offset - skb_network_offset(skb2);
1468 __skb_pull(skb2, offset);
1469 skb_reset_transport_header(skb2);
1470 skb_postpull_rcsum(skb2, skb_network_header(skb2),
1471 skb_network_header_len(skb2));
1473 icmp6_type = icmp6_hdr(skb2)->icmp6_type;
1475 switch (icmp6_type) {
1476 case ICMPV6_MGM_QUERY:
1477 case ICMPV6_MGM_REPORT:
1478 case ICMPV6_MGM_REDUCTION:
1479 case ICMPV6_MLD2_REPORT:
1486 /* Okay, we found MLD message. Check further. */
1487 if (skb2->len > len) {
1488 err = pskb_trim_rcsum(skb2, len);
1494 ip6h = ipv6_hdr(skb2);
1496 switch (skb2->ip_summed) {
1497 case CHECKSUM_COMPLETE:
1498 if (!csum_ipv6_magic(&ip6h->saddr, &ip6h->daddr, skb2->len,
1499 IPPROTO_ICMPV6, skb2->csum))
1503 skb2->csum = ~csum_unfold(csum_ipv6_magic(&ip6h->saddr,
1506 IPPROTO_ICMPV6, 0));
1507 if (__skb_checksum_complete(skb2))
1513 BR_INPUT_SKB_CB(skb)->igmp = 1;
1515 switch (icmp6_type) {
1516 case ICMPV6_MGM_REPORT:
1518 struct mld_msg *mld;
1519 if (!pskb_may_pull(skb2, sizeof(*mld))) {
1523 mld = (struct mld_msg *)skb_transport_header(skb2);
1524 BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1525 err = br_ip6_multicast_add_group(br, port, &mld->mld_mca);
1528 case ICMPV6_MLD2_REPORT:
1529 err = br_ip6_multicast_mld2_report(br, port, skb2);
1531 case ICMPV6_MGM_QUERY:
1532 err = br_ip6_multicast_query(br, port, skb2);
1534 case ICMPV6_MGM_REDUCTION:
1536 struct mld_msg *mld;
1537 if (!pskb_may_pull(skb2, sizeof(*mld))) {
1541 mld = (struct mld_msg *)skb_transport_header(skb2);
1542 br_ip6_multicast_leave_group(br, port, &mld->mld_mca);
1552 int br_multicast_rcv(struct net_bridge *br, struct net_bridge_port *port,
1553 struct sk_buff *skb)
1555 BR_INPUT_SKB_CB(skb)->igmp = 0;
1556 BR_INPUT_SKB_CB(skb)->mrouters_only = 0;
1558 if (br->multicast_disabled)
1561 switch (skb->protocol) {
1562 case htons(ETH_P_IP):
1563 return br_multicast_ipv4_rcv(br, port, skb);
1564 #if IS_ENABLED(CONFIG_IPV6)
1565 case htons(ETH_P_IPV6):
1566 return br_multicast_ipv6_rcv(br, port, skb);
1573 static void br_multicast_query_expired(unsigned long data)
1575 struct net_bridge *br = (void *)data;
1577 spin_lock(&br->multicast_lock);
1578 if (br->multicast_startup_queries_sent <
1579 br->multicast_startup_query_count)
1580 br->multicast_startup_queries_sent++;
1582 br_multicast_send_query(br, NULL, br->multicast_startup_queries_sent);
1584 spin_unlock(&br->multicast_lock);
1587 void br_multicast_init(struct net_bridge *br)
1589 br->hash_elasticity = 4;
1592 br->multicast_router = 1;
1593 br->multicast_querier = 0;
1594 br->multicast_last_member_count = 2;
1595 br->multicast_startup_query_count = 2;
1597 br->multicast_last_member_interval = HZ;
1598 br->multicast_query_response_interval = 10 * HZ;
1599 br->multicast_startup_query_interval = 125 * HZ / 4;
1600 br->multicast_query_interval = 125 * HZ;
1601 br->multicast_querier_interval = 255 * HZ;
1602 br->multicast_membership_interval = 260 * HZ;
1604 spin_lock_init(&br->multicast_lock);
1605 setup_timer(&br->multicast_router_timer,
1606 br_multicast_local_router_expired, 0);
1607 setup_timer(&br->multicast_querier_timer,
1608 br_multicast_querier_expired, (unsigned long)br);
1609 setup_timer(&br->multicast_query_timer, br_multicast_query_expired,
1613 void br_multicast_open(struct net_bridge *br)
1615 br->multicast_startup_queries_sent = 0;
1617 if (br->multicast_disabled)
1620 mod_timer(&br->multicast_query_timer, jiffies);
1623 void br_multicast_stop(struct net_bridge *br)
1625 struct net_bridge_mdb_htable *mdb;
1626 struct net_bridge_mdb_entry *mp;
1627 struct hlist_node *p, *n;
1631 del_timer_sync(&br->multicast_router_timer);
1632 del_timer_sync(&br->multicast_querier_timer);
1633 del_timer_sync(&br->multicast_query_timer);
1635 spin_lock_bh(&br->multicast_lock);
1636 mdb = mlock_dereference(br->mdb, br);
1643 for (i = 0; i < mdb->max; i++) {
1644 hlist_for_each_entry_safe(mp, p, n, &mdb->mhash[i],
1646 del_timer(&mp->timer);
1647 call_rcu_bh(&mp->rcu, br_multicast_free_group);
1652 spin_unlock_bh(&br->multicast_lock);
1654 spin_lock_bh(&br->multicast_lock);
1659 call_rcu_bh(&mdb->rcu, br_mdb_free);
1662 spin_unlock_bh(&br->multicast_lock);
1665 int br_multicast_set_router(struct net_bridge *br, unsigned long val)
1669 spin_lock_bh(&br->multicast_lock);
1670 if (!netif_running(br->dev))
1676 del_timer(&br->multicast_router_timer);
1679 br->multicast_router = val;
1689 spin_unlock_bh(&br->multicast_lock);
1694 int br_multicast_set_port_router(struct net_bridge_port *p, unsigned long val)
1696 struct net_bridge *br = p->br;
1699 spin_lock(&br->multicast_lock);
1700 if (!netif_running(br->dev) || p->state == BR_STATE_DISABLED)
1707 p->multicast_router = val;
1710 if (val < 2 && !hlist_unhashed(&p->rlist))
1711 hlist_del_init_rcu(&p->rlist);
1716 del_timer(&p->multicast_router_timer);
1721 br_multicast_add_router(br, p);
1730 spin_unlock(&br->multicast_lock);
1735 static void br_multicast_start_querier(struct net_bridge *br)
1737 struct net_bridge_port *port;
1739 br_multicast_open(br);
1741 list_for_each_entry(port, &br->port_list, list) {
1742 if (port->state == BR_STATE_DISABLED ||
1743 port->state == BR_STATE_BLOCKING)
1746 __br_multicast_enable_port(port);
1750 int br_multicast_toggle(struct net_bridge *br, unsigned long val)
1753 struct net_bridge_mdb_htable *mdb;
1755 spin_lock_bh(&br->multicast_lock);
1756 if (br->multicast_disabled == !val)
1759 br->multicast_disabled = !val;
1760 if (br->multicast_disabled)
1763 if (!netif_running(br->dev))
1766 mdb = mlock_dereference(br->mdb, br);
1771 br->multicast_disabled = !!val;
1775 err = br_mdb_rehash(&br->mdb, mdb->max,
1776 br->hash_elasticity);
1781 br_multicast_start_querier(br);
1784 spin_unlock_bh(&br->multicast_lock);
1789 int br_multicast_set_querier(struct net_bridge *br, unsigned long val)
1793 spin_lock_bh(&br->multicast_lock);
1794 if (br->multicast_querier == val)
1797 br->multicast_querier = val;
1799 br_multicast_start_querier(br);
1802 spin_unlock_bh(&br->multicast_lock);
1807 int br_multicast_set_hash_max(struct net_bridge *br, unsigned long val)
1811 struct net_bridge_mdb_htable *mdb;
1813 spin_lock(&br->multicast_lock);
1814 if (!netif_running(br->dev))
1818 if (!is_power_of_2(val))
1821 mdb = mlock_dereference(br->mdb, br);
1822 if (mdb && val < mdb->size)
1838 err = br_mdb_rehash(&br->mdb, br->hash_max,
1839 br->hash_elasticity);
1845 spin_unlock(&br->multicast_lock);