net/bridge/br_multicast.c

   1 /*
   2  * Bridge multicast support.
   3  *
   4  * Copyright (c) 2010 Herbert Xu <herbert@gondor.apana.org.au>
   5  *
   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)
   9  * any later version.
  10  *
  11  */
  12
  13 #include <linux/err.h>
  14 #include <linux/if_ether.h>
  15 #include <linux/igmp.h>
  16 #include <linux/jhash.h>
  17 #include <linux/kernel.h>
  18 #include <linux/log2.h>
  19 #include <linux/netdevice.h>
  20 #include <linux/netfilter_bridge.h>
  21 #include <linux/random.h>
  22 #include <linux/rculist.h>
  23 #include <linux/skbuff.h>
  24 #include <linux/slab.h>
  25 #include <linux/timer.h>
  26 #include <net/ip.h>
  27 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
  28 #include <net/ipv6.h>
  29 #include <net/mld.h>
  30 #include <net/addrconf.h>
  31 #include <net/ip6_checksum.h>
  32 #endif
  33
  34 #include "br_private.h"
  35
  36 #define mlock_dereference(X, br) \
  37         rcu_dereference_protected(X, lockdep_is_held(&br->multicast_lock))
  38
  39 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
  40 static inline int ipv6_is_local_multicast(const struct in6_addr *addr)
  41 {
  42         if (ipv6_addr_is_multicast(addr) &&
  43             IPV6_ADDR_MC_SCOPE(addr) <= IPV6_ADDR_SCOPE_LINKLOCAL)
  44                 return 1;
  45         return 0;
  46 }
  47 #endif
  48
  49 static inline int br_ip_equal(const struct br_ip *a, const struct br_ip *b)
  50 {
  51         if (a->proto != b->proto)
  52                 return 0;
  53         switch (a->proto) {
  54         case htons(ETH_P_IP):
  55                 return a->u.ip4 == b->u.ip4;
  56 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
  57         case htons(ETH_P_IPV6):
  58                 return ipv6_addr_equal(&a->u.ip6, &b->u.ip6);
  59 #endif
  60         }
  61         return 0;
  62 }
  63
  64 static inline int __br_ip4_hash(struct net_bridge_mdb_htable *mdb, __be32 ip)
  65 {
  66         return jhash_1word(mdb->secret, (__force u32)ip) & (mdb->max - 1);
  67 }
  68
  69 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
  70 static inline int __br_ip6_hash(struct net_bridge_mdb_htable *mdb,
  71                                 const struct in6_addr *ip)
  72 {
  73         return jhash2((__force u32 *)ip->s6_addr32, 4, mdb->secret) & (mdb->max - 1);
  74 }
  75 #endif
  76
  77 static inline int br_ip_hash(struct net_bridge_mdb_htable *mdb,
  78                              struct br_ip *ip)
  79 {
  80         switch (ip->proto) {
  81         case htons(ETH_P_IP):
  82                 return __br_ip4_hash(mdb, ip->u.ip4);
  83 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
  84         case htons(ETH_P_IPV6):
  85                 return __br_ip6_hash(mdb, &ip->u.ip6);
  86 #endif
  87         }
  88         return 0;
  89 }
  90
  91 static struct net_bridge_mdb_entry *__br_mdb_ip_get(
  92         struct net_bridge_mdb_htable *mdb, struct br_ip *dst, int hash)
  93 {
  94         struct net_bridge_mdb_entry *mp;
  95         struct hlist_node *p;
  96
  97         hlist_for_each_entry_rcu(mp, p, &mdb->mhash[hash], hlist[mdb->ver]) {
  98                 if (br_ip_equal(&mp->addr, dst))
  99                         return mp;
 100         }
 101
 102         return NULL;
 103 }
 104
 105 static struct net_bridge_mdb_entry *br_mdb_ip_get(
 106         struct net_bridge_mdb_htable *mdb, struct br_ip *dst)
 107 {
 108         if (!mdb)
 109                 return NULL;
 110
 111         return __br_mdb_ip_get(mdb, dst, br_ip_hash(mdb, dst));
 112 }
 113
 114 static struct net_bridge_mdb_entry *br_mdb_ip4_get(
 115         struct net_bridge_mdb_htable *mdb, __be32 dst)
 116 {
 117         struct br_ip br_dst;
 118
 119         br_dst.u.ip4 = dst;
 120         br_dst.proto = htons(ETH_P_IP);
 121
 122         return br_mdb_ip_get(mdb, &br_dst);
 123 }
 124
 125 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
 126 static struct net_bridge_mdb_entry *br_mdb_ip6_get(
 127         struct net_bridge_mdb_htable *mdb, const struct in6_addr *dst)
 128 {
 129         struct br_ip br_dst;
 130
 131         ipv6_addr_copy(&br_dst.u.ip6, dst);
 132         br_dst.proto = htons(ETH_P_IPV6);
 133
 134         return br_mdb_ip_get(mdb, &br_dst);
 135 }
 136 #endif
 137
 138 struct net_bridge_mdb_entry *br_mdb_get(struct net_bridge *br,
 139                                         struct sk_buff *skb)
 140 {
 141         struct net_bridge_mdb_htable *mdb = rcu_dereference(br->mdb);
 142         struct br_ip ip;
 143
 144         if (br->multicast_disabled)
 145                 return NULL;
 146
 147         if (BR_INPUT_SKB_CB(skb)->igmp)
 148                 return NULL;
 149
 150         ip.proto = skb->protocol;
 151
 152         switch (skb->protocol) {
 153         case htons(ETH_P_IP):
 154                 ip.u.ip4 = ip_hdr(skb)->daddr;
 155                 break;
 156 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
 157         case htons(ETH_P_IPV6):
 158                 ipv6_addr_copy(&ip.u.ip6, &ipv6_hdr(skb)->daddr);
 159                 break;
 160 #endif
 161         default:
 162                 return NULL;
 163         }
 164
 165         return br_mdb_ip_get(mdb, &ip);
 166 }
 167
 168 static void br_mdb_free(struct rcu_head *head)
 169 {
 170         struct net_bridge_mdb_htable *mdb =
 171                 container_of(head, struct net_bridge_mdb_htable, rcu);
 172         struct net_bridge_mdb_htable *old = mdb->old;
 173
 174         mdb->old = NULL;
 175         kfree(old->mhash);
 176         kfree(old);
 177 }
 178
 179 static int br_mdb_copy(struct net_bridge_mdb_htable *new,
 180                        struct net_bridge_mdb_htable *old,
 181                        int elasticity)
 182 {
 183         struct net_bridge_mdb_entry *mp;
 184         struct hlist_node *p;
 185         int maxlen;
 186         int len;
 187         int i;
 188
 189         for (i = 0; i < old->max; i++)
 190                 hlist_for_each_entry(mp, p, &old->mhash[i], hlist[old->ver])
 191                         hlist_add_head(&mp->hlist[new->ver],
 192                                        &new->mhash[br_ip_hash(new, &mp->addr)]);
 193
 194         if (!elasticity)
 195                 return 0;
 196
 197         maxlen = 0;
 198         for (i = 0; i < new->max; i++) {
 199                 len = 0;
 200                 hlist_for_each_entry(mp, p, &new->mhash[i], hlist[new->ver])
 201                         len++;
 202                 if (len > maxlen)
 203                         maxlen = len;
 204         }
 205
 206         return maxlen > elasticity ? -EINVAL : 0;
 207 }
 208
 209 static void br_multicast_free_pg(struct rcu_head *head)
 210 {
 211         struct net_bridge_port_group *p =
 212                 container_of(head, struct net_bridge_port_group, rcu);
 213
 214         kfree(p);
 215 }
 216
 217 static void br_multicast_free_group(struct rcu_head *head)
 218 {
 219         struct net_bridge_mdb_entry *mp =
 220                 container_of(head, struct net_bridge_mdb_entry, rcu);
 221
 222         kfree(mp);
 223 }
 224
 225 static void br_multicast_group_expired(unsigned long data)
 226 {
 227         struct net_bridge_mdb_entry *mp = (void *)data;
 228         struct net_bridge *br = mp->br;
 229         struct net_bridge_mdb_htable *mdb;
 230
 231         spin_lock(&br->multicast_lock);
 232         if (!netif_running(br->dev) || timer_pending(&mp->timer))
 233                 goto out;
 234
 235         if (!hlist_unhashed(&mp->mglist))
 236                 hlist_del_init(&mp->mglist);
 237
 238         if (mp->ports)
 239                 goto out;
 240
 241         mdb = mlock_dereference(br->mdb, br);
 242
 243         hlist_del_rcu(&mp->hlist[mdb->ver]);
 244         mdb->size--;
 245
 246         del_timer(&mp->query_timer);
 247         call_rcu_bh(&mp->rcu, br_multicast_free_group);
 248
 249 out:
 250         spin_unlock(&br->multicast_lock);
 251 }
 252
 253 static void br_multicast_del_pg(struct net_bridge *br,
 254                                 struct net_bridge_port_group *pg)
 255 {
 256         struct net_bridge_mdb_htable *mdb;
 257         struct net_bridge_mdb_entry *mp;
 258         struct net_bridge_port_group *p;
 259         struct net_bridge_port_group __rcu **pp;
 260
 261         mdb = mlock_dereference(br->mdb, br);
 262
 263         mp = br_mdb_ip_get(mdb, &pg->addr);
 264         if (WARN_ON(!mp))
 265                 return;
 266
 267         for (pp = &mp->ports;
 268              (p = mlock_dereference(*pp, br)) != NULL;
 269              pp = &p->next) {
 270                 if (p != pg)
 271                         continue;
 272
 273                 rcu_assign_pointer(*pp, p->next);
 274                 hlist_del_init(&p->mglist);
 275                 del_timer(&p->timer);
 276                 del_timer(&p->query_timer);
 277                 call_rcu_bh(&p->rcu, br_multicast_free_pg);
 278
 279                 if (!mp->ports && hlist_unhashed(&mp->mglist) &&
 280                     netif_running(br->dev))
 281                         mod_timer(&mp->timer, jiffies);
 282
 283                 return;
 284         }
 285
 286         WARN_ON(1);
 287 }
 288
 289 static void br_multicast_port_group_expired(unsigned long data)
 290 {
 291         struct net_bridge_port_group *pg = (void *)data;
 292         struct net_bridge *br = pg->port->br;
 293
 294         spin_lock(&br->multicast_lock);
 295         if (!netif_running(br->dev) || timer_pending(&pg->timer) ||
 296             hlist_unhashed(&pg->mglist))
 297                 goto out;
 298
 299         br_multicast_del_pg(br, pg);
 300
 301 out:
 302         spin_unlock(&br->multicast_lock);
 303 }
 304
 305 static int br_mdb_rehash(struct net_bridge_mdb_htable __rcu **mdbp, int max,
 306                          int elasticity)
 307 {
 308         struct net_bridge_mdb_htable *old = rcu_dereference_protected(*mdbp, 1);
 309         struct net_bridge_mdb_htable *mdb;
 310         int err;
 311
 312         mdb = kmalloc(sizeof(*mdb), GFP_ATOMIC);
 313         if (!mdb)
 314                 return -ENOMEM;
 315
 316         mdb->max = max;
 317         mdb->old = old;
 318
 319         mdb->mhash = kzalloc(max * sizeof(*mdb->mhash), GFP_ATOMIC);
 320         if (!mdb->mhash) {
 321                 kfree(mdb);
 322                 return -ENOMEM;
 323         }
 324
 325         mdb->size = old ? old->size : 0;
 326         mdb->ver = old ? old->ver ^ 1 : 0;
 327
 328         if (!old || elasticity)
 329                 get_random_bytes(&mdb->secret, sizeof(mdb->secret));
 330         else
 331                 mdb->secret = old->secret;
 332
 333         if (!old)
 334                 goto out;
 335
 336         err = br_mdb_copy(mdb, old, elasticity);
 337         if (err) {
 338                 kfree(mdb->mhash);
 339                 kfree(mdb);
 340                 return err;
 341         }
 342
 343         call_rcu_bh(&mdb->rcu, br_mdb_free);
 344
 345 out:
 346         rcu_assign_pointer(*mdbp, mdb);
 347
 348         return 0;
 349 }
 350
 351 static struct sk_buff *br_ip4_multicast_alloc_query(struct net_bridge *br,
 352                                                     __be32 group)
 353 {
 354         struct sk_buff *skb;
 355         struct igmphdr *ih;
 356         struct ethhdr *eth;
 357         struct iphdr *iph;
 358
 359         skb = netdev_alloc_skb_ip_align(br->dev, sizeof(*eth) + sizeof(*iph) +
 360                                                  sizeof(*ih) + 4);
 361         if (!skb)
 362                 goto out;
 363
 364         skb->protocol = htons(ETH_P_IP);
 365
 366         skb_reset_mac_header(skb);
 367         eth = eth_hdr(skb);
 368
 369         memcpy(eth->h_source, br->dev->dev_addr, 6);
 370         eth->h_dest[0] = 1;
 371         eth->h_dest[1] = 0;
 372         eth->h_dest[2] = 0x5e;
 373         eth->h_dest[3] = 0;
 374         eth->h_dest[4] = 0;
 375         eth->h_dest[5] = 1;
 376         eth->h_proto = htons(ETH_P_IP);
 377         skb_put(skb, sizeof(*eth));
 378
 379         skb_set_network_header(skb, skb->len);
 380         iph = ip_hdr(skb);
 381
 382         iph->version = 4;
 383         iph->ihl = 6;
 384         iph->tos = 0xc0;
 385         iph->tot_len = htons(sizeof(*iph) + sizeof(*ih) + 4);
 386         iph->id = 0;
 387         iph->frag_off = htons(IP_DF);
 388         iph->ttl = 1;
 389         iph->protocol = IPPROTO_IGMP;
 390         iph->saddr = 0;
 391         iph->daddr = htonl(INADDR_ALLHOSTS_GROUP);
 392         ((u8 *)&iph[1])[0] = IPOPT_RA;
 393         ((u8 *)&iph[1])[1] = 4;
 394         ((u8 *)&iph[1])[2] = 0;
 395         ((u8 *)&iph[1])[3] = 0;
 396         ip_send_check(iph);
 397         skb_put(skb, 24);
 398
 399         skb_set_transport_header(skb, skb->len);
 400         ih = igmp_hdr(skb);
 401         ih->type = IGMP_HOST_MEMBERSHIP_QUERY;
 402         ih->code = (group ? br->multicast_last_member_interval :
 403                             br->multicast_query_response_interval) /
 404                    (HZ / IGMP_TIMER_SCALE);
 405         ih->group = group;
 406         ih->csum = 0;
 407         ih->csum = ip_compute_csum((void *)ih, sizeof(struct igmphdr));
 408         skb_put(skb, sizeof(*ih));
 409
 410         __skb_pull(skb, sizeof(*eth));
 411
 412 out:
 413         return skb;
 414 }
 415
 416 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
 417 static struct sk_buff *br_ip6_multicast_alloc_query(struct net_bridge *br,
 418                                                     struct in6_addr *group)
 419 {
 420         struct sk_buff *skb;
 421         struct ipv6hdr *ip6h;
 422         struct mld_msg *mldq;
 423         struct ethhdr *eth;
 424         u8 *hopopt;
 425         unsigned long interval;
 426
 427         skb = netdev_alloc_skb_ip_align(br->dev, sizeof(*eth) + sizeof(*ip6h) +
 428                                                  8 + sizeof(*mldq));
 429         if (!skb)
 430                 goto out;
 431
 432         skb->protocol = htons(ETH_P_IPV6);
 433
 434         /* Ethernet header */
 435         skb_reset_mac_header(skb);
 436         eth = eth_hdr(skb);
 437
 438         memcpy(eth->h_source, br->dev->dev_addr, 6);
 439         ipv6_eth_mc_map(group, eth->h_dest);
 440         eth->h_proto = htons(ETH_P_IPV6);
 441         skb_put(skb, sizeof(*eth));
 442
 443         /* IPv6 header + HbH option */
 444         skb_set_network_header(skb, skb->len);
 445         ip6h = ipv6_hdr(skb);
 446
 447         *(__force __be32 *)ip6h = htonl(0x60000000);
 448         ip6h->payload_len = 8 + sizeof(*mldq);
 449         ip6h->nexthdr = IPPROTO_HOPOPTS;
 450         ip6h->hop_limit = 1;
 451         ipv6_addr_set(&ip6h->saddr, 0, 0, 0, 0);
 452         ipv6_addr_set(&ip6h->daddr, htonl(0xff020000), 0, 0, htonl(1));
 453
 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) */
 460         hopopt[5] = 0;
 461         hopopt[6] = IPV6_TLV_PAD0;              /* Pad0 */
 462         hopopt[7] = IPV6_TLV_PAD0;              /* Pad0 */
 463
 464         skb_put(skb, sizeof(*ip6h) + 8);
 465
 466         /* ICMPv6 */
 467         skb_set_transport_header(skb, skb->len);
 468         mldq = (struct mld_msg *) icmp6_hdr(skb);
 469
 470         interval = ipv6_addr_any(group) ? br->multicast_last_member_interval :
 471                                           br->multicast_query_response_interval;
 472
 473         mldq->mld_type = ICMPV6_MGM_QUERY;
 474         mldq->mld_code = 0;
 475         mldq->mld_cksum = 0;
 476         mldq->mld_maxdelay = htons((u16)jiffies_to_msecs(interval));
 477         mldq->mld_reserved = 0;
 478         ipv6_addr_copy(&mldq->mld_mca, group);
 479
 480         /* checksum */
 481         mldq->mld_cksum = csum_ipv6_magic(&ip6h->saddr, &ip6h->daddr,
 482                                           sizeof(*mldq), IPPROTO_ICMPV6,
 483                                           csum_partial(mldq,
 484                                                        sizeof(*mldq), 0));
 485         skb_put(skb, sizeof(*mldq));
 486
 487         __skb_pull(skb, sizeof(*eth));
 488
 489 out:
 490         return skb;
 491 }
 492 #endif
 493
 494 static struct sk_buff *br_multicast_alloc_query(struct net_bridge *br,
 495                                                 struct br_ip *addr)
 496 {
 497         switch (addr->proto) {
 498         case htons(ETH_P_IP):
 499                 return br_ip4_multicast_alloc_query(br, addr->u.ip4);
 500 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
 501         case htons(ETH_P_IPV6):
 502                 return br_ip6_multicast_alloc_query(br, &addr->u.ip6);
 503 #endif
 504         }
 505         return NULL;
 506 }
 507
 508 static void br_multicast_send_group_query(struct net_bridge_mdb_entry *mp)
 509 {
 510         struct net_bridge *br = mp->br;
 511         struct sk_buff *skb;
 512
 513         skb = br_multicast_alloc_query(br, &mp->addr);
 514         if (!skb)
 515                 goto timer;
 516
 517         netif_rx(skb);
 518
 519 timer:
 520         if (++mp->queries_sent < br->multicast_last_member_count)
 521                 mod_timer(&mp->query_timer,
 522                           jiffies + br->multicast_last_member_interval);
 523 }
 524
 525 static void br_multicast_group_query_expired(unsigned long data)
 526 {
 527         struct net_bridge_mdb_entry *mp = (void *)data;
 528         struct net_bridge *br = mp->br;
 529
 530         spin_lock(&br->multicast_lock);
 531         if (!netif_running(br->dev) || hlist_unhashed(&mp->mglist) ||
 532             mp->queries_sent >= br->multicast_last_member_count)
 533                 goto out;
 534
 535         br_multicast_send_group_query(mp);
 536
 537 out:
 538         spin_unlock(&br->multicast_lock);
 539 }
 540
 541 static void br_multicast_send_port_group_query(struct net_bridge_port_group *pg)
 542 {
 543         struct net_bridge_port *port = pg->port;
 544         struct net_bridge *br = port->br;
 545         struct sk_buff *skb;
 546
 547         skb = br_multicast_alloc_query(br, &pg->addr);
 548         if (!skb)
 549                 goto timer;
 550
 551         br_deliver(port, skb);
 552
 553 timer:
 554         if (++pg->queries_sent < br->multicast_last_member_count)
 555                 mod_timer(&pg->query_timer,
 556                           jiffies + br->multicast_last_member_interval);
 557 }
 558
 559 static void br_multicast_port_group_query_expired(unsigned long data)
 560 {
 561         struct net_bridge_port_group *pg = (void *)data;
 562         struct net_bridge_port *port = pg->port;
 563         struct net_bridge *br = port->br;
 564
 565         spin_lock(&br->multicast_lock);
 566         if (!netif_running(br->dev) || hlist_unhashed(&pg->mglist) ||
 567             pg->queries_sent >= br->multicast_last_member_count)
 568                 goto out;
 569
 570         br_multicast_send_port_group_query(pg);
 571
 572 out:
 573         spin_unlock(&br->multicast_lock);
 574 }
 575
 576 static struct net_bridge_mdb_entry *br_multicast_get_group(
 577         struct net_bridge *br, struct net_bridge_port *port,
 578         struct br_ip *group, int hash)
 579 {
 580         struct net_bridge_mdb_htable *mdb;
 581         struct net_bridge_mdb_entry *mp;
 582         struct hlist_node *p;
 583         unsigned count = 0;
 584         unsigned max;
 585         int elasticity;
 586         int err;
 587
 588         mdb = rcu_dereference_protected(br->mdb, 1);
 589         hlist_for_each_entry(mp, p, &mdb->mhash[hash], hlist[mdb->ver]) {
 590                 count++;
 591                 if (unlikely(br_ip_equal(group, &mp->addr)))
 592                         return mp;
 593         }
 594
 595         elasticity = 0;
 596         max = mdb->max;
 597
 598         if (unlikely(count > br->hash_elasticity && count)) {
 599                 if (net_ratelimit())
 600                         br_info(br, "Multicast hash table "
 601                                 "chain limit reached: %s\n",
 602                                 port ? port->dev->name : br->dev->name);
 603
 604                 elasticity = br->hash_elasticity;
 605         }
 606
 607         if (mdb->size >= max) {
 608                 max *= 2;
 609                 if (unlikely(max >= br->hash_max)) {
 610                         br_warn(br, "Multicast hash table maximum "
 611                                 "reached, disabling snooping: %s, %d\n",
 612                                 port ? port->dev->name : br->dev->name, max);
 613                         err = -E2BIG;
 614 disable:
 615                         br->multicast_disabled = 1;
 616                         goto err;
 617                 }
 618         }
 619
 620         if (max > mdb->max || elasticity) {
 621                 if (mdb->old) {
 622                         if (net_ratelimit())
 623                                 br_info(br, "Multicast hash table "
 624                                         "on fire: %s\n",
 625                                         port ? port->dev->name : br->dev->name);
 626                         err = -EEXIST;
 627                         goto err;
 628                 }
 629
 630                 err = br_mdb_rehash(&br->mdb, max, elasticity);
 631                 if (err) {
 632                         br_warn(br, "Cannot rehash multicast "
 633                                 "hash table, disabling snooping: %s, %d, %d\n",
 634                                 port ? port->dev->name : br->dev->name,
 635                                 mdb->size, err);
 636                         goto disable;
 637                 }
 638
 639                 err = -EAGAIN;
 640                 goto err;
 641         }
 642
 643         return NULL;
 644
 645 err:
 646         mp = ERR_PTR(err);
 647         return mp;
 648 }
 649
 650 static struct net_bridge_mdb_entry *br_multicast_new_group(
 651         struct net_bridge *br, struct net_bridge_port *port,
 652         struct br_ip *group)
 653 {
 654         struct net_bridge_mdb_htable *mdb;
 655         struct net_bridge_mdb_entry *mp;
 656         int hash;
 657
 658         mdb = rcu_dereference_protected(br->mdb, 1);
 659         if (!mdb) {
 660                 if (br_mdb_rehash(&br->mdb, BR_HASH_SIZE, 0))
 661                         return NULL;
 662                 goto rehash;
 663         }
 664
 665         hash = br_ip_hash(mdb, group);
 666         mp = br_multicast_get_group(br, port, group, hash);
 667         switch (PTR_ERR(mp)) {
 668         case 0:
 669                 break;
 670
 671         case -EAGAIN:
 672 rehash:
 673                 mdb = rcu_dereference_protected(br->mdb, 1);
 674                 hash = br_ip_hash(mdb, group);
 675                 break;
 676
 677         default:
 678                 goto out;
 679         }
 680
 681         mp = kzalloc(sizeof(*mp), GFP_ATOMIC);
 682         if (unlikely(!mp))
 683                 goto out;
 684
 685         mp->br = br;
 686         mp->addr = *group;
 687         setup_timer(&mp->timer, br_multicast_group_expired,
 688                     (unsigned long)mp);
 689         setup_timer(&mp->query_timer, br_multicast_group_query_expired,
 690                     (unsigned long)mp);
 691
 692         hlist_add_head_rcu(&mp->hlist[mdb->ver], &mdb->mhash[hash]);
 693         mdb->size++;
 694
 695 out:
 696         return mp;
 697 }
 698
 699 static int br_multicast_add_group(struct net_bridge *br,
 700                                   struct net_bridge_port *port,
 701                                   struct br_ip *group)
 702 {
 703         struct net_bridge_mdb_entry *mp;
 704         struct net_bridge_port_group *p;
 705         struct net_bridge_port_group __rcu **pp;
 706         unsigned long now = jiffies;
 707         int err;
 708
 709         spin_lock(&br->multicast_lock);
 710         if (!netif_running(br->dev) ||
 711             (port && port->state == BR_STATE_DISABLED))
 712                 goto out;
 713
 714         mp = br_multicast_new_group(br, port, group);
 715         err = PTR_ERR(mp);
 716         if (unlikely(IS_ERR(mp) || !mp))
 717                 goto err;
 718
 719         if (!port) {
 720                 hlist_add_head(&mp->mglist, &br->mglist);
 721                 mod_timer(&mp->timer, now + br->multicast_membership_interval);
 722                 goto out;
 723         }
 724
 725         for (pp = &mp->ports;
 726              (p = mlock_dereference(*pp, br)) != NULL;
 727              pp = &p->next) {
 728                 if (p->port == port)
 729                         goto found;
 730                 if ((unsigned long)p->port < (unsigned long)port)
 731                         break;
 732         }
 733
 734         p = kzalloc(sizeof(*p), GFP_ATOMIC);
 735         err = -ENOMEM;
 736         if (unlikely(!p))
 737                 goto err;
 738
 739         p->addr = *group;
 740         p->port = port;
 741         p->next = *pp;
 742         hlist_add_head(&p->mglist, &port->mglist);
 743         setup_timer(&p->timer, br_multicast_port_group_expired,
 744                     (unsigned long)p);
 745         setup_timer(&p->query_timer, br_multicast_port_group_query_expired,
 746                     (unsigned long)p);
 747
 748         rcu_assign_pointer(*pp, p);
 749
 750 found:
 751         mod_timer(&p->timer, now + br->multicast_membership_interval);
 752 out:
 753         err = 0;
 754
 755 err:
 756         spin_unlock(&br->multicast_lock);
 757         return err;
 758 }
 759
 760 static int br_ip4_multicast_add_group(struct net_bridge *br,
 761                                       struct net_bridge_port *port,
 762                                       __be32 group)
 763 {
 764         struct br_ip br_group;
 765
 766         if (ipv4_is_local_multicast(group))
 767                 return 0;
 768
 769         br_group.u.ip4 = group;
 770         br_group.proto = htons(ETH_P_IP);
 771
 772         return br_multicast_add_group(br, port, &br_group);
 773 }
 774
 775 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
 776 static int br_ip6_multicast_add_group(struct net_bridge *br,
 777                                       struct net_bridge_port *port,
 778                                       const struct in6_addr *group)
 779 {
 780         struct br_ip br_group;
 781
 782         if (ipv6_is_local_multicast(group))
 783                 return 0;
 784
 785         ipv6_addr_copy(&br_group.u.ip6, group);
 786         br_group.proto = htons(ETH_P_IP);
 787
 788         return br_multicast_add_group(br, port, &br_group);
 789 }
 790 #endif
 791
 792 static void br_multicast_router_expired(unsigned long data)
 793 {
 794         struct net_bridge_port *port = (void *)data;
 795         struct net_bridge *br = port->br;
 796
 797         spin_lock(&br->multicast_lock);
 798         if (port->multicast_router != 1 ||
 799             timer_pending(&port->multicast_router_timer) ||
 800             hlist_unhashed(&port->rlist))
 801                 goto out;
 802
 803         hlist_del_init_rcu(&port->rlist);
 804
 805 out:
 806         spin_unlock(&br->multicast_lock);
 807 }
 808
 809 static void br_multicast_local_router_expired(unsigned long data)
 810 {
 811 }
 812
 813 static void __br_multicast_send_query(struct net_bridge *br,
 814                                       struct net_bridge_port *port,
 815                                       struct br_ip *ip)
 816 {
 817         struct sk_buff *skb;
 818
 819         skb = br_multicast_alloc_query(br, ip);
 820         if (!skb)
 821                 return;
 822
 823         if (port) {
 824                 __skb_push(skb, sizeof(struct ethhdr));
 825                 skb->dev = port->dev;
 826                 NF_HOOK(NFPROTO_BRIDGE, NF_BR_LOCAL_OUT, skb, NULL, skb->dev,
 827                         dev_queue_xmit);
 828         } else
 829                 netif_rx(skb);
 830 }
 831
 832 static void br_multicast_send_query(struct net_bridge *br,
 833                                     struct net_bridge_port *port, u32 sent)
 834 {
 835         unsigned long time;
 836         struct br_ip br_group;
 837
 838         if (!netif_running(br->dev) || br->multicast_disabled ||
 839             timer_pending(&br->multicast_querier_timer))
 840                 return;
 841
 842         memset(&br_group.u, 0, sizeof(br_group.u));
 843
 844         br_group.proto = htons(ETH_P_IP);
 845         __br_multicast_send_query(br, port, &br_group);
 846
 847 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
 848         br_group.proto = htons(ETH_P_IPV6);
 849         __br_multicast_send_query(br, port, &br_group);
 850 #endif
 851
 852         time = jiffies;
 853         time += sent < br->multicast_startup_query_count ?
 854                 br->multicast_startup_query_interval :
 855                 br->multicast_query_interval;
 856         mod_timer(port ? &port->multicast_query_timer :
 857                          &br->multicast_query_timer, time);
 858 }
 859
 860 static void br_multicast_port_query_expired(unsigned long data)
 861 {
 862         struct net_bridge_port *port = (void *)data;
 863         struct net_bridge *br = port->br;
 864
 865         spin_lock(&br->multicast_lock);
 866         if (port->state == BR_STATE_DISABLED ||
 867             port->state == BR_STATE_BLOCKING)
 868                 goto out;
 869
 870         if (port->multicast_startup_queries_sent <
 871             br->multicast_startup_query_count)
 872                 port->multicast_startup_queries_sent++;
 873
 874         br_multicast_send_query(port->br, port,
 875                                 port->multicast_startup_queries_sent);
 876
 877 out:
 878         spin_unlock(&br->multicast_lock);
 879 }
 880
 881 void br_multicast_add_port(struct net_bridge_port *port)
 882 {
 883         port->multicast_router = 1;
 884
 885         setup_timer(&port->multicast_router_timer, br_multicast_router_expired,
 886                     (unsigned long)port);
 887         setup_timer(&port->multicast_query_timer,
 888                     br_multicast_port_query_expired, (unsigned long)port);
 889 }
 890
 891 void br_multicast_del_port(struct net_bridge_port *port)
 892 {
 893         del_timer_sync(&port->multicast_router_timer);
 894 }
 895
 896 static void __br_multicast_enable_port(struct net_bridge_port *port)
 897 {
 898         port->multicast_startup_queries_sent = 0;
 899
 900         if (try_to_del_timer_sync(&port->multicast_query_timer) >= 0 ||
 901             del_timer(&port->multicast_query_timer))
 902                 mod_timer(&port->multicast_query_timer, jiffies);
 903 }
 904
 905 void br_multicast_enable_port(struct net_bridge_port *port)
 906 {
 907         struct net_bridge *br = port->br;
 908
 909         spin_lock(&br->multicast_lock);
 910         if (br->multicast_disabled || !netif_running(br->dev))
 911                 goto out;
 912
 913         __br_multicast_enable_port(port);
 914
 915 out:
 916         spin_unlock(&br->multicast_lock);
 917 }
 918
 919 void br_multicast_disable_port(struct net_bridge_port *port)
 920 {
 921         struct net_bridge *br = port->br;
 922         struct net_bridge_port_group *pg;
 923         struct hlist_node *p, *n;
 924
 925         spin_lock(&br->multicast_lock);
 926         hlist_for_each_entry_safe(pg, p, n, &port->mglist, mglist)
 927                 br_multicast_del_pg(br, pg);
 928
 929         if (!hlist_unhashed(&port->rlist))
 930                 hlist_del_init_rcu(&port->rlist);
 931         del_timer(&port->multicast_router_timer);
 932         del_timer(&port->multicast_query_timer);
 933         spin_unlock(&br->multicast_lock);
 934 }
 935
 936 static int br_ip4_multicast_igmp3_report(struct net_bridge *br,
 937                                          struct net_bridge_port *port,
 938                                          struct sk_buff *skb)
 939 {
 940         struct igmpv3_report *ih;
 941         struct igmpv3_grec *grec;
 942         int i;
 943         int len;
 944         int num;
 945         int type;
 946         int err = 0;
 947         __be32 group;
 948
 949         if (!pskb_may_pull(skb, sizeof(*ih)))
 950                 return -EINVAL;
 951
 952         ih = igmpv3_report_hdr(skb);
 953         num = ntohs(ih->ngrec);
 954         len = sizeof(*ih);
 955
 956         for (i = 0; i < num; i++) {
 957                 len += sizeof(*grec);
 958                 if (!pskb_may_pull(skb, len))
 959                         return -EINVAL;
 960
 961                 grec = (void *)(skb->data + len - sizeof(*grec));
 962                 group = grec->grec_mca;
 963                 type = grec->grec_type;
 964
 965                 len += ntohs(grec->grec_nsrcs) * 4;
 966                 if (!pskb_may_pull(skb, len))
 967                         return -EINVAL;
 968
 969                 /* We treat this as an IGMPv2 report for now. */
 970                 switch (type) {
 971                 case IGMPV3_MODE_IS_INCLUDE:
 972                 case IGMPV3_MODE_IS_EXCLUDE:
 973                 case IGMPV3_CHANGE_TO_INCLUDE:
 974                 case IGMPV3_CHANGE_TO_EXCLUDE:
 975                 case IGMPV3_ALLOW_NEW_SOURCES:
 976                 case IGMPV3_BLOCK_OLD_SOURCES:
 977                         break;
 978
 979                 default:
 980                         continue;
 981                 }
 982
 983                 err = br_ip4_multicast_add_group(br, port, group);
 984                 if (err)
 985                         break;
 986         }
 987
 988         return err;
 989 }
 990
 991 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
 992 static int br_ip6_multicast_mld2_report(struct net_bridge *br,
 993                                         struct net_bridge_port *port,
 994                                         struct sk_buff *skb)
 995 {
 996         struct icmp6hdr *icmp6h;
 997         struct mld2_grec *grec;
 998         int i;
 999         int len;
1000         int num;
1001         int err = 0;
1002
1003         if (!pskb_may_pull(skb, sizeof(*icmp6h)))
1004                 return -EINVAL;
1005
1006         icmp6h = icmp6_hdr(skb);
1007         num = ntohs(icmp6h->icmp6_dataun.un_data16[1]);
1008         len = sizeof(*icmp6h);
1009
1010         for (i = 0; i < num; i++) {
1011                 __be16 *nsrcs, _nsrcs;
1012
1013                 nsrcs = skb_header_pointer(skb,
1014                                            len + offsetof(struct mld2_grec,
1015                                                           grec_mca),
1016                                            sizeof(_nsrcs), &_nsrcs);
1017                 if (!nsrcs)
1018                         return -EINVAL;
1019
1020                 if (!pskb_may_pull(skb,
1021                                    len + sizeof(*grec) +
1022                                    sizeof(struct in6_addr) * (*nsrcs)))
1023                         return -EINVAL;
1024
1025                 grec = (struct mld2_grec *)(skb->data + len);
1026                 len += sizeof(*grec) + sizeof(struct in6_addr) * (*nsrcs);
1027
1028                 /* We treat these as MLDv1 reports for now. */
1029                 switch (grec->grec_type) {
1030                 case MLD2_MODE_IS_INCLUDE:
1031                 case MLD2_MODE_IS_EXCLUDE:
1032                 case MLD2_CHANGE_TO_INCLUDE:
1033                 case MLD2_CHANGE_TO_EXCLUDE:
1034                 case MLD2_ALLOW_NEW_SOURCES:
1035                 case MLD2_BLOCK_OLD_SOURCES:
1036                         break;
1037
1038                 default:
1039                         continue;
1040                 }
1041
1042                 err = br_ip6_multicast_add_group(br, port, &grec->grec_mca);
1043                 if (!err)
1044                         break;
1045         }
1046
1047         return err;
1048 }
1049 #endif
1050
1051 /*
1052  * Add port to rotuer_list
1053  *  list is maintained ordered by pointer value
1054  *  and locked by br->multicast_lock and RCU
1055  */
1056 static void br_multicast_add_router(struct net_bridge *br,
1057                                     struct net_bridge_port *port)
1058 {
1059         struct net_bridge_port *p;
1060         struct hlist_node *n, *slot = NULL;
1061
1062         hlist_for_each_entry(p, n, &br->router_list, rlist) {
1063                 if ((unsigned long) port >= (unsigned long) p)
1064                         break;
1065                 slot = n;
1066         }
1067
1068         if (slot)
1069                 hlist_add_after_rcu(slot, &port->rlist);
1070         else
1071                 hlist_add_head_rcu(&port->rlist, &br->router_list);
1072 }
1073
1074 static void br_multicast_mark_router(struct net_bridge *br,
1075                                      struct net_bridge_port *port)
1076 {
1077         unsigned long now = jiffies;
1078
1079         if (!port) {
1080                 if (br->multicast_router == 1)
1081                         mod_timer(&br->multicast_router_timer,
1082                                   now + br->multicast_querier_interval);
1083                 return;
1084         }
1085
1086         if (port->multicast_router != 1)
1087                 return;
1088
1089         if (!hlist_unhashed(&port->rlist))
1090                 goto timer;
1091
1092         br_multicast_add_router(br, port);
1093
1094 timer:
1095         mod_timer(&port->multicast_router_timer,
1096                   now + br->multicast_querier_interval);
1097 }
1098
1099 static void br_multicast_query_received(struct net_bridge *br,
1100                                         struct net_bridge_port *port,
1101                                         int saddr)
1102 {
1103         if (saddr)
1104                 mod_timer(&br->multicast_querier_timer,
1105                           jiffies + br->multicast_querier_interval);
1106         else if (timer_pending(&br->multicast_querier_timer))
1107                 return;
1108
1109         br_multicast_mark_router(br, port);
1110 }
1111
1112 static int br_ip4_multicast_query(struct net_bridge *br,
1113                                   struct net_bridge_port *port,
1114                                   struct sk_buff *skb)
1115 {
1116         struct iphdr *iph = ip_hdr(skb);
1117         struct igmphdr *ih = igmp_hdr(skb);
1118         struct net_bridge_mdb_entry *mp;
1119         struct igmpv3_query *ih3;
1120         struct net_bridge_port_group *p;
1121         struct net_bridge_port_group __rcu **pp;
1122         unsigned long max_delay;
1123         unsigned long now = jiffies;
1124         __be32 group;
1125         int err = 0;
1126
1127         spin_lock(&br->multicast_lock);
1128         if (!netif_running(br->dev) ||
1129             (port && port->state == BR_STATE_DISABLED))
1130                 goto out;
1131
1132         br_multicast_query_received(br, port, !!iph->saddr);
1133
1134         group = ih->group;
1135
1136         if (skb->len == sizeof(*ih)) {
1137                 max_delay = ih->code * (HZ / IGMP_TIMER_SCALE);
1138
1139                 if (!max_delay) {
1140                         max_delay = 10 * HZ;
1141                         group = 0;
1142                 }
1143         } else {
1144                 if (!pskb_may_pull(skb, sizeof(struct igmpv3_query))) {
1145                         err = -EINVAL;
1146                         goto out;
1147                 }
1148
1149                 ih3 = igmpv3_query_hdr(skb);
1150                 if (ih3->nsrcs)
1151                         goto out;
1152
1153                 max_delay = ih3->code ?
1154                             IGMPV3_MRC(ih3->code) * (HZ / IGMP_TIMER_SCALE) : 1;
1155         }
1156
1157         if (!group)
1158                 goto out;
1159
1160         mp = br_mdb_ip4_get(mlock_dereference(br->mdb, br), group);
1161         if (!mp)
1162                 goto out;
1163
1164         max_delay *= br->multicast_last_member_count;
1165
1166         if (!hlist_unhashed(&mp->mglist) &&
1167             (timer_pending(&mp->timer) ?
1168              time_after(mp->timer.expires, now + max_delay) :
1169              try_to_del_timer_sync(&mp->timer) >= 0))
1170                 mod_timer(&mp->timer, now + max_delay);
1171
1172         for (pp = &mp->ports;
1173              (p = mlock_dereference(*pp, br)) != NULL;
1174              pp = &p->next) {
1175                 if (timer_pending(&p->timer) ?
1176                     time_after(p->timer.expires, now + max_delay) :
1177                     try_to_del_timer_sync(&p->timer) >= 0)
1178                         mod_timer(&mp->timer, now + max_delay);
1179         }
1180
1181 out:
1182         spin_unlock(&br->multicast_lock);
1183         return err;
1184 }
1185
1186 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
1187 static int br_ip6_multicast_query(struct net_bridge *br,
1188                                   struct net_bridge_port *port,
1189                                   struct sk_buff *skb)
1190 {
1191         struct ipv6hdr *ip6h = ipv6_hdr(skb);
1192         struct mld_msg *mld = (struct mld_msg *) icmp6_hdr(skb);
1193         struct net_bridge_mdb_entry *mp;
1194         struct mld2_query *mld2q;
1195         struct net_bridge_port_group *p;
1196         struct net_bridge_port_group __rcu **pp;
1197         unsigned long max_delay;
1198         unsigned long now = jiffies;
1199         struct in6_addr *group = NULL;
1200         int err = 0;
1201
1202         spin_lock(&br->multicast_lock);
1203         if (!netif_running(br->dev) ||
1204             (port && port->state == BR_STATE_DISABLED))
1205                 goto out;
1206
1207         br_multicast_query_received(br, port, !ipv6_addr_any(&ip6h->saddr));
1208
1209         if (skb->len == sizeof(*mld)) {
1210                 if (!pskb_may_pull(skb, sizeof(*mld))) {
1211                         err = -EINVAL;
1212                         goto out;
1213                 }
1214                 mld = (struct mld_msg *) icmp6_hdr(skb);
1215                 max_delay = msecs_to_jiffies(htons(mld->mld_maxdelay));
1216                 if (max_delay)
1217                         group = &mld->mld_mca;
1218         } else if (skb->len >= sizeof(*mld2q)) {
1219                 if (!pskb_may_pull(skb, sizeof(*mld2q))) {
1220                         err = -EINVAL;
1221                         goto out;
1222                 }
1223                 mld2q = (struct mld2_query *)icmp6_hdr(skb);
1224                 if (!mld2q->mld2q_nsrcs)
1225                         group = &mld2q->mld2q_mca;
1226                 max_delay = mld2q->mld2q_mrc ? MLDV2_MRC(mld2q->mld2q_mrc) : 1;
1227         }
1228
1229         if (!group)
1230                 goto out;
1231
1232         mp = br_mdb_ip6_get(mlock_dereference(br->mdb, br), group);
1233         if (!mp)
1234                 goto out;
1235
1236         max_delay *= br->multicast_last_member_count;
1237         if (!hlist_unhashed(&mp->mglist) &&
1238             (timer_pending(&mp->timer) ?
1239              time_after(mp->timer.expires, now + max_delay) :
1240              try_to_del_timer_sync(&mp->timer) >= 0))
1241                 mod_timer(&mp->timer, now + max_delay);
1242
1243         for (pp = &mp->ports;
1244              (p = mlock_dereference(*pp, br)) != NULL;
1245              pp = &p->next) {
1246                 if (timer_pending(&p->timer) ?
1247                     time_after(p->timer.expires, now + max_delay) :
1248                     try_to_del_timer_sync(&p->timer) >= 0)
1249                         mod_timer(&mp->timer, now + max_delay);
1250         }
1251
1252 out:
1253         spin_unlock(&br->multicast_lock);
1254         return err;
1255 }
1256 #endif
1257
1258 static void br_multicast_leave_group(struct net_bridge *br,
1259                                      struct net_bridge_port *port,
1260                                      struct br_ip *group)
1261 {
1262         struct net_bridge_mdb_htable *mdb;
1263         struct net_bridge_mdb_entry *mp;
1264         struct net_bridge_port_group *p;
1265         unsigned long now;
1266         unsigned long time;
1267
1268         spin_lock(&br->multicast_lock);
1269         if (!netif_running(br->dev) ||
1270             (port && port->state == BR_STATE_DISABLED) ||
1271             timer_pending(&br->multicast_querier_timer))
1272                 goto out;
1273
1274         mdb = mlock_dereference(br->mdb, br);
1275         mp = br_mdb_ip_get(mdb, group);
1276         if (!mp)
1277                 goto out;
1278
1279         now = jiffies;
1280         time = now + br->multicast_last_member_count *
1281                      br->multicast_last_member_interval;
1282
1283         if (!port) {
1284                 if (!hlist_unhashed(&mp->mglist) &&
1285                     (timer_pending(&mp->timer) ?
1286                      time_after(mp->timer.expires, time) :
1287                      try_to_del_timer_sync(&mp->timer) >= 0)) {
1288                         mod_timer(&mp->timer, time);
1289
1290                         mp->queries_sent = 0;
1291                         mod_timer(&mp->query_timer, now);
1292                 }
1293
1294                 goto out;
1295         }
1296
1297         for (p = mlock_dereference(mp->ports, br);
1298              p != NULL;
1299              p = mlock_dereference(p->next, br)) {
1300                 if (p->port != port)
1301                         continue;
1302
1303                 if (!hlist_unhashed(&p->mglist) &&
1304                     (timer_pending(&p->timer) ?
1305                      time_after(p->timer.expires, time) :
1306                      try_to_del_timer_sync(&p->timer) >= 0)) {
1307                         mod_timer(&p->timer, time);
1308
1309                         p->queries_sent = 0;
1310                         mod_timer(&p->query_timer, now);
1311                 }
1312
1313                 break;
1314         }
1315
1316 out:
1317         spin_unlock(&br->multicast_lock);
1318 }
1319
1320 static void br_ip4_multicast_leave_group(struct net_bridge *br,
1321                                          struct net_bridge_port *port,
1322                                          __be32 group)
1323 {
1324         struct br_ip br_group;
1325
1326         if (ipv4_is_local_multicast(group))
1327                 return;
1328
1329         br_group.u.ip4 = group;
1330         br_group.proto = htons(ETH_P_IP);
1331
1332         br_multicast_leave_group(br, port, &br_group);
1333 }
1334
1335 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
1336 static void br_ip6_multicast_leave_group(struct net_bridge *br,
1337                                          struct net_bridge_port *port,
1338                                          const struct in6_addr *group)
1339 {
1340         struct br_ip br_group;
1341
1342         if (ipv6_is_local_multicast(group))
1343                 return;
1344
1345         ipv6_addr_copy(&br_group.u.ip6, group);
1346         br_group.proto = htons(ETH_P_IPV6);
1347
1348         br_multicast_leave_group(br, port, &br_group);
1349 }
1350 #endif
1351
1352 static int br_multicast_ipv4_rcv(struct net_bridge *br,
1353                                  struct net_bridge_port *port,
1354                                  struct sk_buff *skb)
1355 {
1356         struct sk_buff *skb2 = skb;
1357         struct iphdr *iph;
1358         struct igmphdr *ih;
1359         unsigned len;
1360         unsigned offset;
1361         int err;
1362
1363         /* We treat OOM as packet loss for now. */
1364         if (!pskb_may_pull(skb, sizeof(*iph)))
1365                 return -EINVAL;
1366
1367         iph = ip_hdr(skb);
1368
1369         if (iph->ihl < 5 || iph->version != 4)
1370                 return -EINVAL;
1371
1372         if (!pskb_may_pull(skb, ip_hdrlen(skb)))
1373                 return -EINVAL;
1374
1375         iph = ip_hdr(skb);
1376
1377         if (unlikely(ip_fast_csum((u8 *)iph, iph->ihl)))
1378                 return -EINVAL;
1379
1380         if (iph->protocol != IPPROTO_IGMP)
1381                 return 0;
1382
1383         len = ntohs(iph->tot_len);
1384         if (skb->len < len || len < ip_hdrlen(skb))
1385                 return -EINVAL;
1386
1387         if (skb->len > len) {
1388                 skb2 = skb_clone(skb, GFP_ATOMIC);
1389                 if (!skb2)
1390                         return -ENOMEM;
1391
1392                 err = pskb_trim_rcsum(skb2, len);
1393                 if (err)
1394                         goto err_out;
1395         }
1396
1397         len -= ip_hdrlen(skb2);
1398         offset = skb_network_offset(skb2) + ip_hdrlen(skb2);
1399         __skb_pull(skb2, offset);
1400         skb_reset_transport_header(skb2);
1401
1402         err = -EINVAL;
1403         if (!pskb_may_pull(skb2, sizeof(*ih)))
1404                 goto out;
1405
1406         switch (skb2->ip_summed) {
1407         case CHECKSUM_COMPLETE:
1408                 if (!csum_fold(skb2->csum))
1409                         break;
1410                 /* fall through */
1411         case CHECKSUM_NONE:
1412                 skb2->csum = 0;
1413                 if (skb_checksum_complete(skb2))
1414                         goto out;
1415         }
1416
1417         err = 0;
1418
1419         BR_INPUT_SKB_CB(skb)->igmp = 1;
1420         ih = igmp_hdr(skb2);
1421
1422         switch (ih->type) {
1423         case IGMP_HOST_MEMBERSHIP_REPORT:
1424         case IGMPV2_HOST_MEMBERSHIP_REPORT:
1425                 BR_INPUT_SKB_CB(skb2)->mrouters_only = 1;
1426                 err = br_ip4_multicast_add_group(br, port, ih->group);
1427                 break;
1428         case IGMPV3_HOST_MEMBERSHIP_REPORT:
1429                 err = br_ip4_multicast_igmp3_report(br, port, skb2);
1430                 break;
1431         case IGMP_HOST_MEMBERSHIP_QUERY:
1432                 err = br_ip4_multicast_query(br, port, skb2);
1433                 break;
1434         case IGMP_HOST_LEAVE_MESSAGE:
1435                 br_ip4_multicast_leave_group(br, port, ih->group);
1436                 break;
1437         }
1438
1439 out:
1440         __skb_push(skb2, offset);
1441 err_out:
1442         if (skb2 != skb)
1443                 kfree_skb(skb2);
1444         return err;
1445 }
1446
1447 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
1448 static int br_multicast_ipv6_rcv(struct net_bridge *br,
1449                                  struct net_bridge_port *port,
1450                                  struct sk_buff *skb)
1451 {
1452         struct sk_buff *skb2 = skb;
1453         struct ipv6hdr *ip6h;
1454         struct icmp6hdr *icmp6h;
1455         u8 nexthdr;
1456         unsigned len;
1457         int offset;
1458         int err;
1459
1460         if (!pskb_may_pull(skb, sizeof(*ip6h)))
1461                 return -EINVAL;
1462
1463         ip6h = ipv6_hdr(skb);
1464
1465         /*
1466          * We're interested in MLD messages only.
1467          *  - Version is 6
1468          *  - MLD has always Router Alert hop-by-hop option
1469          *  - But we do not support jumbrograms.
1470          */
1471         if (ip6h->version != 6 ||
1472             ip6h->nexthdr != IPPROTO_HOPOPTS ||
1473             ip6h->payload_len == 0)
1474                 return 0;
1475
1476         len = ntohs(ip6h->payload_len);
1477         if (skb->len < len)
1478                 return -EINVAL;
1479
1480         nexthdr = ip6h->nexthdr;
1481         offset = ipv6_skip_exthdr(skb, sizeof(*ip6h), &nexthdr);
1482
1483         if (offset < 0 || nexthdr != IPPROTO_ICMPV6)
1484                 return 0;
1485
1486         /* Okay, we found ICMPv6 header */
1487         skb2 = skb_clone(skb, GFP_ATOMIC);
1488         if (!skb2)
1489                 return -ENOMEM;
1490
1491         len -= offset - skb_network_offset(skb2);
1492
1493         __skb_pull(skb2, offset);
1494         skb_reset_transport_header(skb2);
1495
1496         err = -EINVAL;
1497         if (!pskb_may_pull(skb2, sizeof(*icmp6h)))
1498                 goto out;
1499
1500         icmp6h = icmp6_hdr(skb2);
1501
1502         switch (icmp6h->icmp6_type) {
1503         case ICMPV6_MGM_QUERY:
1504         case ICMPV6_MGM_REPORT:
1505         case ICMPV6_MGM_REDUCTION:
1506         case ICMPV6_MLD2_REPORT:
1507                 break;
1508         default:
1509                 err = 0;
1510                 goto out;
1511         }
1512
1513         /* Okay, we found MLD message. Check further. */
1514         if (skb2->len > len) {
1515                 err = pskb_trim_rcsum(skb2, len);
1516                 if (err)
1517                         goto out;
1518         }
1519
1520         switch (skb2->ip_summed) {
1521         case CHECKSUM_COMPLETE:
1522                 if (!csum_fold(skb2->csum))
1523                         break;
1524                 /*FALLTHROUGH*/
1525         case CHECKSUM_NONE:
1526                 skb2->csum = 0;
1527                 if (skb_checksum_complete(skb2))
1528                         goto out;
1529         }
1530
1531         err = 0;
1532
1533         BR_INPUT_SKB_CB(skb)->igmp = 1;
1534
1535         switch (icmp6h->icmp6_type) {
1536         case ICMPV6_MGM_REPORT:
1537             {
1538                 struct mld_msg *mld = (struct mld_msg *)icmp6h;
1539                 BR_INPUT_SKB_CB(skb2)->mrouters_only = 1;
1540                 err = br_ip6_multicast_add_group(br, port, &mld->mld_mca);
1541                 break;
1542             }
1543         case ICMPV6_MLD2_REPORT:
1544                 err = br_ip6_multicast_mld2_report(br, port, skb2);
1545                 break;
1546         case ICMPV6_MGM_QUERY:
1547                 err = br_ip6_multicast_query(br, port, skb2);
1548                 break;
1549         case ICMPV6_MGM_REDUCTION:
1550             {
1551                 struct mld_msg *mld = (struct mld_msg *)icmp6h;
1552                 br_ip6_multicast_leave_group(br, port, &mld->mld_mca);
1553             }
1554         }
1555
1556 out:
1557         __skb_push(skb2, offset);
1558         if (skb2 != skb)
1559                 kfree_skb(skb2);
1560         return err;
1561 }
1562 #endif
1563
1564 int br_multicast_rcv(struct net_bridge *br, struct net_bridge_port *port,
1565                      struct sk_buff *skb)
1566 {
1567         BR_INPUT_SKB_CB(skb)->igmp = 0;
1568         BR_INPUT_SKB_CB(skb)->mrouters_only = 0;
1569
1570         if (br->multicast_disabled)
1571                 return 0;
1572
1573         switch (skb->protocol) {
1574         case htons(ETH_P_IP):
1575                 return br_multicast_ipv4_rcv(br, port, skb);
1576 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
1577         case htons(ETH_P_IPV6):
1578                 return br_multicast_ipv6_rcv(br, port, skb);
1579 #endif
1580         }
1581
1582         return 0;
1583 }
1584
1585 static void br_multicast_query_expired(unsigned long data)
1586 {
1587         struct net_bridge *br = (void *)data;
1588
1589         spin_lock(&br->multicast_lock);
1590         if (br->multicast_startup_queries_sent <
1591             br->multicast_startup_query_count)
1592                 br->multicast_startup_queries_sent++;
1593
1594         br_multicast_send_query(br, NULL, br->multicast_startup_queries_sent);
1595
1596         spin_unlock(&br->multicast_lock);
1597 }
1598
1599 void br_multicast_init(struct net_bridge *br)
1600 {
1601         br->hash_elasticity = 4;
1602         br->hash_max = 512;
1603
1604         br->multicast_router = 1;
1605         br->multicast_last_member_count = 2;
1606         br->multicast_startup_query_count = 2;
1607
1608         br->multicast_last_member_interval = HZ;
1609         br->multicast_query_response_interval = 10 * HZ;
1610         br->multicast_startup_query_interval = 125 * HZ / 4;
1611         br->multicast_query_interval = 125 * HZ;
1612         br->multicast_querier_interval = 255 * HZ;
1613         br->multicast_membership_interval = 260 * HZ;
1614
1615         spin_lock_init(&br->multicast_lock);
1616         setup_timer(&br->multicast_router_timer,
1617                     br_multicast_local_router_expired, 0);
1618         setup_timer(&br->multicast_querier_timer,
1619                     br_multicast_local_router_expired, 0);
1620         setup_timer(&br->multicast_query_timer, br_multicast_query_expired,
1621                     (unsigned long)br);
1622 }
1623
1624 void br_multicast_open(struct net_bridge *br)
1625 {
1626         br->multicast_startup_queries_sent = 0;
1627
1628         if (br->multicast_disabled)
1629                 return;
1630
1631         mod_timer(&br->multicast_query_timer, jiffies);
1632 }
1633
1634 void br_multicast_stop(struct net_bridge *br)
1635 {
1636         struct net_bridge_mdb_htable *mdb;
1637         struct net_bridge_mdb_entry *mp;
1638         struct hlist_node *p, *n;
1639         u32 ver;
1640         int i;
1641
1642         del_timer_sync(&br->multicast_router_timer);
1643         del_timer_sync(&br->multicast_querier_timer);
1644         del_timer_sync(&br->multicast_query_timer);
1645
1646         spin_lock_bh(&br->multicast_lock);
1647         mdb = mlock_dereference(br->mdb, br);
1648         if (!mdb)
1649                 goto out;
1650
1651         br->mdb = NULL;
1652
1653         ver = mdb->ver;
1654         for (i = 0; i < mdb->max; i++) {
1655                 hlist_for_each_entry_safe(mp, p, n, &mdb->mhash[i],
1656                                           hlist[ver]) {
1657                         del_timer(&mp->timer);
1658                         del_timer(&mp->query_timer);
1659                         call_rcu_bh(&mp->rcu, br_multicast_free_group);
1660                 }
1661         }
1662
1663         if (mdb->old) {
1664                 spin_unlock_bh(&br->multicast_lock);
1665                 rcu_barrier_bh();
1666                 spin_lock_bh(&br->multicast_lock);
1667                 WARN_ON(mdb->old);
1668         }
1669
1670         mdb->old = mdb;
1671         call_rcu_bh(&mdb->rcu, br_mdb_free);
1672
1673 out:
1674         spin_unlock_bh(&br->multicast_lock);
1675 }
1676
1677 int br_multicast_set_router(struct net_bridge *br, unsigned long val)
1678 {
1679         int err = -ENOENT;
1680
1681         spin_lock_bh(&br->multicast_lock);
1682         if (!netif_running(br->dev))
1683                 goto unlock;
1684
1685         switch (val) {
1686         case 0:
1687         case 2:
1688                 del_timer(&br->multicast_router_timer);
1689                 /* fall through */
1690         case 1:
1691                 br->multicast_router = val;
1692                 err = 0;
1693                 break;
1694
1695         default:
1696                 err = -EINVAL;
1697                 break;
1698         }
1699
1700 unlock:
1701         spin_unlock_bh(&br->multicast_lock);
1702
1703         return err;
1704 }
1705
1706 int br_multicast_set_port_router(struct net_bridge_port *p, unsigned long val)
1707 {
1708         struct net_bridge *br = p->br;
1709         int err = -ENOENT;
1710
1711         spin_lock(&br->multicast_lock);
1712         if (!netif_running(br->dev) || p->state == BR_STATE_DISABLED)
1713                 goto unlock;
1714
1715         switch (val) {
1716         case 0:
1717         case 1:
1718         case 2:
1719                 p->multicast_router = val;
1720                 err = 0;
1721
1722                 if (val < 2 && !hlist_unhashed(&p->rlist))
1723                         hlist_del_init_rcu(&p->rlist);
1724
1725                 if (val == 1)
1726                         break;
1727
1728                 del_timer(&p->multicast_router_timer);
1729
1730                 if (val == 0)
1731                         break;
1732
1733                 br_multicast_add_router(br, p);
1734                 break;
1735
1736         default:
1737                 err = -EINVAL;
1738                 break;
1739         }
1740
1741 unlock:
1742         spin_unlock(&br->multicast_lock);
1743
1744         return err;
1745 }
1746
1747 int br_multicast_toggle(struct net_bridge *br, unsigned long val)
1748 {
1749         struct net_bridge_port *port;
1750         int err = 0;
1751         struct net_bridge_mdb_htable *mdb;
1752
1753         spin_lock(&br->multicast_lock);
1754         if (br->multicast_disabled == !val)
1755                 goto unlock;
1756
1757         br->multicast_disabled = !val;
1758         if (br->multicast_disabled)
1759                 goto unlock;
1760
1761         if (!netif_running(br->dev))
1762                 goto unlock;
1763
1764         mdb = mlock_dereference(br->mdb, br);
1765         if (mdb) {
1766                 if (mdb->old) {
1767                         err = -EEXIST;
1768 rollback:
1769                         br->multicast_disabled = !!val;
1770                         goto unlock;
1771                 }
1772
1773                 err = br_mdb_rehash(&br->mdb, mdb->max,
1774                                     br->hash_elasticity);
1775                 if (err)
1776                         goto rollback;
1777         }
1778
1779         br_multicast_open(br);
1780         list_for_each_entry(port, &br->port_list, list) {
1781                 if (port->state == BR_STATE_DISABLED ||
1782                     port->state == BR_STATE_BLOCKING)
1783                         continue;
1784
1785                 __br_multicast_enable_port(port);
1786         }
1787
1788 unlock:
1789         spin_unlock(&br->multicast_lock);
1790
1791         return err;
1792 }
1793
1794 int br_multicast_set_hash_max(struct net_bridge *br, unsigned long val)
1795 {
1796         int err = -ENOENT;
1797         u32 old;
1798         struct net_bridge_mdb_htable *mdb;
1799
1800         spin_lock(&br->multicast_lock);
1801         if (!netif_running(br->dev))
1802                 goto unlock;
1803
1804         err = -EINVAL;
1805         if (!is_power_of_2(val))
1806                 goto unlock;
1807
1808         mdb = mlock_dereference(br->mdb, br);
1809         if (mdb && val < mdb->size)
1810                 goto unlock;
1811
1812         err = 0;
1813
1814         old = br->hash_max;
1815         br->hash_max = val;
1816
1817         if (mdb) {
1818                 if (mdb->old) {
1819                         err = -EEXIST;
1820 rollback:
1821                         br->hash_max = old;
1822                         goto unlock;
1823                 }
1824
1825                 err = br_mdb_rehash(&br->mdb, br->hash_max,
1826                                     br->hash_elasticity);
1827                 if (err)
1828                         goto rollback;
1829         }
1830
1831 unlock:
1832         spin_unlock(&br->multicast_lock);
1833
1834         return err;
1835 }