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_transient_multicast(const struct in6_addr *addr)
  41 {
  42         if (ipv6_addr_is_multicast(addr) && IPV6_ADDR_MC_FLAG_TRANSIENT(addr))
  43                 return 1;
  44         return 0;
  45 }
  46 #endif
  47
  48 static inline int br_ip_equal(const struct br_ip *a, const struct br_ip *b)
  49 {
  50         if (a->proto != b->proto)
  51                 return 0;
  52         switch (a->proto) {
  53         case htons(ETH_P_IP):
  54                 return a->u.ip4 == b->u.ip4;
  55 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
  56         case htons(ETH_P_IPV6):
  57                 return ipv6_addr_equal(&a->u.ip6, &b->u.ip6);
  58 #endif
  59         }
  60         return 0;
  61 }
  62
  63 static inline int __br_ip4_hash(struct net_bridge_mdb_htable *mdb, __be32 ip)
  64 {
  65         return jhash_1word(mdb->secret, (__force u32)ip) & (mdb->max - 1);
  66 }
  67
  68 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
  69 static inline int __br_ip6_hash(struct net_bridge_mdb_htable *mdb,
  70                                 const struct in6_addr *ip)
  71 {
  72         return jhash2((__force u32 *)ip->s6_addr32, 4, mdb->secret) & (mdb->max - 1);
  73 }
  74 #endif
  75
  76 static inline int br_ip_hash(struct net_bridge_mdb_htable *mdb,
  77                              struct br_ip *ip)
  78 {
  79         switch (ip->proto) {
  80         case htons(ETH_P_IP):
  81                 return __br_ip4_hash(mdb, ip->u.ip4);
  82 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
  83         case htons(ETH_P_IPV6):
  84                 return __br_ip6_hash(mdb, &ip->u.ip6);
  85 #endif
  86         }
  87         return 0;
  88 }
  89
  90 static struct net_bridge_mdb_entry *__br_mdb_ip_get(
  91         struct net_bridge_mdb_htable *mdb, struct br_ip *dst, int hash)
  92 {
  93         struct net_bridge_mdb_entry *mp;
  94         struct hlist_node *p;
  95
  96         hlist_for_each_entry_rcu(mp, p, &mdb->mhash[hash], hlist[mdb->ver]) {
  97                 if (br_ip_equal(&mp->addr, dst))
  98                         return mp;
  99         }
 100
 101         return NULL;
 102 }
 103
 104 static struct net_bridge_mdb_entry *br_mdb_ip_get(
 105         struct net_bridge_mdb_htable *mdb, struct br_ip *dst)
 106 {
 107         if (!mdb)
 108                 return NULL;
 109
 110         return __br_mdb_ip_get(mdb, dst, br_ip_hash(mdb, dst));
 111 }
 112
 113 static struct net_bridge_mdb_entry *br_mdb_ip4_get(
 114         struct net_bridge_mdb_htable *mdb, __be32 dst)
 115 {
 116         struct br_ip br_dst;
 117
 118         br_dst.u.ip4 = dst;
 119         br_dst.proto = htons(ETH_P_IP);
 120
 121         return br_mdb_ip_get(mdb, &br_dst);
 122 }
 123
 124 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
 125 static struct net_bridge_mdb_entry *br_mdb_ip6_get(
 126         struct net_bridge_mdb_htable *mdb, const struct in6_addr *dst)
 127 {
 128         struct br_ip br_dst;
 129
 130         ipv6_addr_copy(&br_dst.u.ip6, dst);
 131         br_dst.proto = htons(ETH_P_IPV6);
 132
 133         return br_mdb_ip_get(mdb, &br_dst);
 134 }
 135 #endif
 136
 137 struct net_bridge_mdb_entry *br_mdb_get(struct net_bridge *br,
 138                                         struct sk_buff *skb)
 139 {
 140         struct net_bridge_mdb_htable *mdb = rcu_dereference(br->mdb);
 141         struct br_ip ip;
 142
 143         if (br->multicast_disabled)
 144                 return NULL;
 145
 146         if (BR_INPUT_SKB_CB(skb)->igmp)
 147                 return NULL;
 148
 149         ip.proto = skb->protocol;
 150
 151         switch (skb->protocol) {
 152         case htons(ETH_P_IP):
 153                 ip.u.ip4 = ip_hdr(skb)->daddr;
 154                 break;
 155 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
 156         case htons(ETH_P_IPV6):
 157                 ipv6_addr_copy(&ip.u.ip6, &ipv6_hdr(skb)->daddr);
 158                 break;
 159 #endif
 160         default:
 161                 return NULL;
 162         }
 163
 164         return br_mdb_ip_get(mdb, &ip);
 165 }
 166
 167 static void br_mdb_free(struct rcu_head *head)
 168 {
 169         struct net_bridge_mdb_htable *mdb =
 170                 container_of(head, struct net_bridge_mdb_htable, rcu);
 171         struct net_bridge_mdb_htable *old = mdb->old;
 172
 173         mdb->old = NULL;
 174         kfree(old->mhash);
 175         kfree(old);
 176 }
 177
 178 static int br_mdb_copy(struct net_bridge_mdb_htable *new,
 179                        struct net_bridge_mdb_htable *old,
 180                        int elasticity)
 181 {
 182         struct net_bridge_mdb_entry *mp;
 183         struct hlist_node *p;
 184         int maxlen;
 185         int len;
 186         int i;
 187
 188         for (i = 0; i < old->max; i++)
 189                 hlist_for_each_entry(mp, p, &old->mhash[i], hlist[old->ver])
 190                         hlist_add_head(&mp->hlist[new->ver],
 191                                        &new->mhash[br_ip_hash(new, &mp->addr)]);
 192
 193         if (!elasticity)
 194                 return 0;
 195
 196         maxlen = 0;
 197         for (i = 0; i < new->max; i++) {
 198                 len = 0;
 199                 hlist_for_each_entry(mp, p, &new->mhash[i], hlist[new->ver])
 200                         len++;
 201                 if (len > maxlen)
 202                         maxlen = len;
 203         }
 204
 205         return maxlen > elasticity ? -EINVAL : 0;
 206 }
 207
 208 static void br_multicast_free_pg(struct rcu_head *head)
 209 {
 210         struct net_bridge_port_group *p =
 211                 container_of(head, struct net_bridge_port_group, rcu);
 212
 213         kfree(p);
 214 }
 215
 216 static void br_multicast_free_group(struct rcu_head *head)
 217 {
 218         struct net_bridge_mdb_entry *mp =
 219                 container_of(head, struct net_bridge_mdb_entry, rcu);
 220
 221         kfree(mp);
 222 }
 223
 224 static void br_multicast_group_expired(unsigned long data)
 225 {
 226         struct net_bridge_mdb_entry *mp = (void *)data;
 227         struct net_bridge *br = mp->br;
 228         struct net_bridge_mdb_htable *mdb;
 229
 230         spin_lock(&br->multicast_lock);
 231         if (!netif_running(br->dev) || timer_pending(&mp->timer))
 232                 goto out;
 233
 234         mp->mglist = false;
 235
 236         if (mp->ports)
 237                 goto out;
 238
 239         mdb = mlock_dereference(br->mdb, br);
 240
 241         hlist_del_rcu(&mp->hlist[mdb->ver]);
 242         mdb->size--;
 243
 244         del_timer(&mp->query_timer);
 245         call_rcu_bh(&mp->rcu, br_multicast_free_group);
 246
 247 out:
 248         spin_unlock(&br->multicast_lock);
 249 }
 250
 251 static void br_multicast_del_pg(struct net_bridge *br,
 252                                 struct net_bridge_port_group *pg)
 253 {
 254         struct net_bridge_mdb_htable *mdb;
 255         struct net_bridge_mdb_entry *mp;
 256         struct net_bridge_port_group *p;
 257         struct net_bridge_port_group __rcu **pp;
 258
 259         mdb = mlock_dereference(br->mdb, br);
 260
 261         mp = br_mdb_ip_get(mdb, &pg->addr);
 262         if (WARN_ON(!mp))
 263                 return;
 264
 265         for (pp = &mp->ports;
 266              (p = mlock_dereference(*pp, br)) != NULL;
 267              pp = &p->next) {
 268                 if (p != pg)
 269                         continue;
 270
 271                 rcu_assign_pointer(*pp, p->next);
 272                 hlist_del_init(&p->mglist);
 273                 del_timer(&p->timer);
 274                 del_timer(&p->query_timer);
 275                 call_rcu_bh(&p->rcu, br_multicast_free_pg);
 276
 277                 if (!mp->ports && !mp->mglist &&
 278                     netif_running(br->dev))
 279                         mod_timer(&mp->timer, jiffies);
 280
 281                 return;
 282         }
 283
 284         WARN_ON(1);
 285 }
 286
 287 static void br_multicast_port_group_expired(unsigned long data)
 288 {
 289         struct net_bridge_port_group *pg = (void *)data;
 290         struct net_bridge *br = pg->port->br;
 291
 292         spin_lock(&br->multicast_lock);
 293         if (!netif_running(br->dev) || timer_pending(&pg->timer) ||
 294             hlist_unhashed(&pg->mglist))
 295                 goto out;
 296
 297         br_multicast_del_pg(br, pg);
 298
 299 out:
 300         spin_unlock(&br->multicast_lock);
 301 }
 302
 303 static int br_mdb_rehash(struct net_bridge_mdb_htable __rcu **mdbp, int max,
 304                          int elasticity)
 305 {
 306         struct net_bridge_mdb_htable *old = rcu_dereference_protected(*mdbp, 1);
 307         struct net_bridge_mdb_htable *mdb;
 308         int err;
 309
 310         mdb = kmalloc(sizeof(*mdb), GFP_ATOMIC);
 311         if (!mdb)
 312                 return -ENOMEM;
 313
 314         mdb->max = max;
 315         mdb->old = old;
 316
 317         mdb->mhash = kzalloc(max * sizeof(*mdb->mhash), GFP_ATOMIC);
 318         if (!mdb->mhash) {
 319                 kfree(mdb);
 320                 return -ENOMEM;
 321         }
 322
 323         mdb->size = old ? old->size : 0;
 324         mdb->ver = old ? old->ver ^ 1 : 0;
 325
 326         if (!old || elasticity)
 327                 get_random_bytes(&mdb->secret, sizeof(mdb->secret));
 328         else
 329                 mdb->secret = old->secret;
 330
 331         if (!old)
 332                 goto out;
 333
 334         err = br_mdb_copy(mdb, old, elasticity);
 335         if (err) {
 336                 kfree(mdb->mhash);
 337                 kfree(mdb);
 338                 return err;
 339         }
 340
 341         call_rcu_bh(&mdb->rcu, br_mdb_free);
 342
 343 out:
 344         rcu_assign_pointer(*mdbp, mdb);
 345
 346         return 0;
 347 }
 348
 349 static struct sk_buff *br_ip4_multicast_alloc_query(struct net_bridge *br,
 350                                                     __be32 group)
 351 {
 352         struct sk_buff *skb;
 353         struct igmphdr *ih;
 354         struct ethhdr *eth;
 355         struct iphdr *iph;
 356
 357         skb = netdev_alloc_skb_ip_align(br->dev, sizeof(*eth) + sizeof(*iph) +
 358                                                  sizeof(*ih) + 4);
 359         if (!skb)
 360                 goto out;
 361
 362         skb->protocol = htons(ETH_P_IP);
 363
 364         skb_reset_mac_header(skb);
 365         eth = eth_hdr(skb);
 366
 367         memcpy(eth->h_source, br->dev->dev_addr, 6);
 368         eth->h_dest[0] = 1;
 369         eth->h_dest[1] = 0;
 370         eth->h_dest[2] = 0x5e;
 371         eth->h_dest[3] = 0;
 372         eth->h_dest[4] = 0;
 373         eth->h_dest[5] = 1;
 374         eth->h_proto = htons(ETH_P_IP);
 375         skb_put(skb, sizeof(*eth));
 376
 377         skb_set_network_header(skb, skb->len);
 378         iph = ip_hdr(skb);
 379
 380         iph->version = 4;
 381         iph->ihl = 6;
 382         iph->tos = 0xc0;
 383         iph->tot_len = htons(sizeof(*iph) + sizeof(*ih) + 4);
 384         iph->id = 0;
 385         iph->frag_off = htons(IP_DF);
 386         iph->ttl = 1;
 387         iph->protocol = IPPROTO_IGMP;
 388         iph->saddr = 0;
 389         iph->daddr = htonl(INADDR_ALLHOSTS_GROUP);
 390         ((u8 *)&iph[1])[0] = IPOPT_RA;
 391         ((u8 *)&iph[1])[1] = 4;
 392         ((u8 *)&iph[1])[2] = 0;
 393         ((u8 *)&iph[1])[3] = 0;
 394         ip_send_check(iph);
 395         skb_put(skb, 24);
 396
 397         skb_set_transport_header(skb, skb->len);
 398         ih = igmp_hdr(skb);
 399         ih->type = IGMP_HOST_MEMBERSHIP_QUERY;
 400         ih->code = (group ? br->multicast_last_member_interval :
 401                             br->multicast_query_response_interval) /
 402                    (HZ / IGMP_TIMER_SCALE);
 403         ih->group = group;
 404         ih->csum = 0;
 405         ih->csum = ip_compute_csum((void *)ih, sizeof(struct igmphdr));
 406         skb_put(skb, sizeof(*ih));
 407
 408         __skb_pull(skb, sizeof(*eth));
 409
 410 out:
 411         return skb;
 412 }
 413
 414 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
 415 static struct sk_buff *br_ip6_multicast_alloc_query(struct net_bridge *br,
 416                                                     const struct in6_addr *group)
 417 {
 418         struct sk_buff *skb;
 419         struct ipv6hdr *ip6h;
 420         struct mld_msg *mldq;
 421         struct ethhdr *eth;
 422         u8 *hopopt;
 423         unsigned long interval;
 424
 425         skb = netdev_alloc_skb_ip_align(br->dev, sizeof(*eth) + sizeof(*ip6h) +
 426                                                  8 + sizeof(*mldq));
 427         if (!skb)
 428                 goto out;
 429
 430         skb->protocol = htons(ETH_P_IPV6);
 431
 432         /* Ethernet header */
 433         skb_reset_mac_header(skb);
 434         eth = eth_hdr(skb);
 435
 436         memcpy(eth->h_source, br->dev->dev_addr, 6);
 437         eth->h_proto = htons(ETH_P_IPV6);
 438         skb_put(skb, sizeof(*eth));
 439
 440         /* IPv6 header + HbH option */
 441         skb_set_network_header(skb, skb->len);
 442         ip6h = ipv6_hdr(skb);
 443
 444         *(__force __be32 *)ip6h = htonl(0x60000000);
 445         ip6h->payload_len = htons(8 + sizeof(*mldq));
 446         ip6h->nexthdr = IPPROTO_HOPOPTS;
 447         ip6h->hop_limit = 1;
 448         ipv6_addr_set(&ip6h->daddr, htonl(0xff020000), 0, 0, htonl(1));
 449         ipv6_dev_get_saddr(dev_net(br->dev), br->dev, &ip6h->daddr, 0,
 450                            &ip6h->saddr);
 451         ipv6_eth_mc_map(&ip6h->daddr, eth->h_dest);
 452
 453         hopopt = (u8 *)(ip6h + 1);
 454         hopopt[0] = IPPROTO_ICMPV6;             /* next hdr */
 455         hopopt[1] = 0;                          /* length of HbH */
 456         hopopt[2] = IPV6_TLV_ROUTERALERT;       /* Router Alert */
 457         hopopt[3] = 2;                          /* Length of RA Option */
 458         hopopt[4] = 0;                          /* Type = 0x0000 (MLD) */
 459         hopopt[5] = 0;
 460         hopopt[6] = IPV6_TLV_PAD0;              /* Pad0 */
 461         hopopt[7] = IPV6_TLV_PAD0;              /* Pad0 */
 462
 463         skb_put(skb, sizeof(*ip6h) + 8);
 464
 465         /* ICMPv6 */
 466         skb_set_transport_header(skb, skb->len);
 467         mldq = (struct mld_msg *) icmp6_hdr(skb);
 468
 469         interval = ipv6_addr_any(group) ? br->multicast_last_member_interval :
 470                                           br->multicast_query_response_interval;
 471
 472         mldq->mld_type = ICMPV6_MGM_QUERY;
 473         mldq->mld_code = 0;
 474         mldq->mld_cksum = 0;
 475         mldq->mld_maxdelay = htons((u16)jiffies_to_msecs(interval));
 476         mldq->mld_reserved = 0;
 477         ipv6_addr_copy(&mldq->mld_mca, group);
 478
 479         /* checksum */
 480         mldq->mld_cksum = csum_ipv6_magic(&ip6h->saddr, &ip6h->daddr,
 481                                           sizeof(*mldq), IPPROTO_ICMPV6,
 482                                           csum_partial(mldq,
 483                                                        sizeof(*mldq), 0));
 484         skb_put(skb, sizeof(*mldq));
 485
 486         __skb_pull(skb, sizeof(*eth));
 487
 488 out:
 489         return skb;
 490 }
 491 #endif
 492
 493 static struct sk_buff *br_multicast_alloc_query(struct net_bridge *br,
 494                                                 struct br_ip *addr)
 495 {
 496         switch (addr->proto) {
 497         case htons(ETH_P_IP):
 498                 return br_ip4_multicast_alloc_query(br, addr->u.ip4);
 499 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
 500         case htons(ETH_P_IPV6):
 501                 return br_ip6_multicast_alloc_query(br, &addr->u.ip6);
 502 #endif
 503         }
 504         return NULL;
 505 }
 506
 507 static void br_multicast_send_group_query(struct net_bridge_mdb_entry *mp)
 508 {
 509         struct net_bridge *br = mp->br;
 510         struct sk_buff *skb;
 511
 512         skb = br_multicast_alloc_query(br, &mp->addr);
 513         if (!skb)
 514                 goto timer;
 515
 516         netif_rx(skb);
 517
 518 timer:
 519         if (++mp->queries_sent < br->multicast_last_member_count)
 520                 mod_timer(&mp->query_timer,
 521                           jiffies + br->multicast_last_member_interval);
 522 }
 523
 524 static void br_multicast_group_query_expired(unsigned long data)
 525 {
 526         struct net_bridge_mdb_entry *mp = (void *)data;
 527         struct net_bridge *br = mp->br;
 528
 529         spin_lock(&br->multicast_lock);
 530         if (!netif_running(br->dev) || !mp->mglist ||
 531             mp->queries_sent >= br->multicast_last_member_count)
 532                 goto out;
 533
 534         br_multicast_send_group_query(mp);
 535
 536 out:
 537         spin_unlock(&br->multicast_lock);
 538 }
 539
 540 static void br_multicast_send_port_group_query(struct net_bridge_port_group *pg)
 541 {
 542         struct net_bridge_port *port = pg->port;
 543         struct net_bridge *br = port->br;
 544         struct sk_buff *skb;
 545
 546         skb = br_multicast_alloc_query(br, &pg->addr);
 547         if (!skb)
 548                 goto timer;
 549
 550         br_deliver(port, skb);
 551
 552 timer:
 553         if (++pg->queries_sent < br->multicast_last_member_count)
 554                 mod_timer(&pg->query_timer,
 555                           jiffies + br->multicast_last_member_interval);
 556 }
 557
 558 static void br_multicast_port_group_query_expired(unsigned long data)
 559 {
 560         struct net_bridge_port_group *pg = (void *)data;
 561         struct net_bridge_port *port = pg->port;
 562         struct net_bridge *br = port->br;
 563
 564         spin_lock(&br->multicast_lock);
 565         if (!netif_running(br->dev) || hlist_unhashed(&pg->mglist) ||
 566             pg->queries_sent >= br->multicast_last_member_count)
 567                 goto out;
 568
 569         br_multicast_send_port_group_query(pg);
 570
 571 out:
 572         spin_unlock(&br->multicast_lock);
 573 }
 574
 575 static struct net_bridge_mdb_entry *br_multicast_get_group(
 576         struct net_bridge *br, struct net_bridge_port *port,
 577         struct br_ip *group, int hash)
 578 {
 579         struct net_bridge_mdb_htable *mdb;
 580         struct net_bridge_mdb_entry *mp;
 581         struct hlist_node *p;
 582         unsigned count = 0;
 583         unsigned max;
 584         int elasticity;
 585         int err;
 586
 587         mdb = rcu_dereference_protected(br->mdb, 1);
 588         hlist_for_each_entry(mp, p, &mdb->mhash[hash], hlist[mdb->ver]) {
 589                 count++;
 590                 if (unlikely(br_ip_equal(group, &mp->addr)))
 591                         return mp;
 592         }
 593
 594         elasticity = 0;
 595         max = mdb->max;
 596
 597         if (unlikely(count > br->hash_elasticity && count)) {
 598                 if (net_ratelimit())
 599                         br_info(br, "Multicast hash table "
 600                                 "chain limit reached: %s\n",
 601                                 port ? port->dev->name : br->dev->name);
 602
 603                 elasticity = br->hash_elasticity;
 604         }
 605
 606         if (mdb->size >= max) {
 607                 max *= 2;
 608                 if (unlikely(max >= br->hash_max)) {
 609                         br_warn(br, "Multicast hash table maximum "
 610                                 "reached, disabling snooping: %s, %d\n",
 611                                 port ? port->dev->name : br->dev->name, max);
 612                         err = -E2BIG;
 613 disable:
 614                         br->multicast_disabled = 1;
 615                         goto err;
 616                 }
 617         }
 618
 619         if (max > mdb->max || elasticity) {
 620                 if (mdb->old) {
 621                         if (net_ratelimit())
 622                                 br_info(br, "Multicast hash table "
 623                                         "on fire: %s\n",
 624                                         port ? port->dev->name : br->dev->name);
 625                         err = -EEXIST;
 626                         goto err;
 627                 }
 628
 629                 err = br_mdb_rehash(&br->mdb, max, elasticity);
 630                 if (err) {
 631                         br_warn(br, "Cannot rehash multicast "
 632                                 "hash table, disabling snooping: %s, %d, %d\n",
 633                                 port ? port->dev->name : br->dev->name,
 634                                 mdb->size, err);
 635                         goto disable;
 636                 }
 637
 638                 err = -EAGAIN;
 639                 goto err;
 640         }
 641
 642         return NULL;
 643
 644 err:
 645         mp = ERR_PTR(err);
 646         return mp;
 647 }
 648
 649 static struct net_bridge_mdb_entry *br_multicast_new_group(
 650         struct net_bridge *br, struct net_bridge_port *port,
 651         struct br_ip *group)
 652 {
 653         struct net_bridge_mdb_htable *mdb;
 654         struct net_bridge_mdb_entry *mp;
 655         int hash;
 656         int err;
 657
 658         mdb = rcu_dereference_protected(br->mdb, 1);
 659         if (!mdb) {
 660                 err = br_mdb_rehash(&br->mdb, BR_HASH_SIZE, 0);
 661                 if (err)
 662                         return ERR_PTR(err);
 663                 goto rehash;
 664         }
 665
 666         hash = br_ip_hash(mdb, group);
 667         mp = br_multicast_get_group(br, port, group, hash);
 668         switch (PTR_ERR(mp)) {
 669         case 0:
 670                 break;
 671
 672         case -EAGAIN:
 673 rehash:
 674                 mdb = rcu_dereference_protected(br->mdb, 1);
 675                 hash = br_ip_hash(mdb, group);
 676                 break;
 677
 678         default:
 679                 goto out;
 680         }
 681
 682         mp = kzalloc(sizeof(*mp), GFP_ATOMIC);
 683         if (unlikely(!mp))
 684                 return ERR_PTR(-ENOMEM);
 685
 686         mp->br = br;
 687         mp->addr = *group;
 688         setup_timer(&mp->timer, br_multicast_group_expired,
 689                     (unsigned long)mp);
 690         setup_timer(&mp->query_timer, br_multicast_group_query_expired,
 691                     (unsigned long)mp);
 692
 693         hlist_add_head_rcu(&mp->hlist[mdb->ver], &mdb->mhash[hash]);
 694         mdb->size++;
 695
 696 out:
 697         return mp;
 698 }
 699
 700 static int br_multicast_add_group(struct net_bridge *br,
 701                                   struct net_bridge_port *port,
 702                                   struct br_ip *group)
 703 {
 704         struct net_bridge_mdb_entry *mp;
 705         struct net_bridge_port_group *p;
 706         struct net_bridge_port_group __rcu **pp;
 707         unsigned long now = jiffies;
 708         int err;
 709
 710         spin_lock(&br->multicast_lock);
 711         if (!netif_running(br->dev) ||
 712             (port && port->state == BR_STATE_DISABLED))
 713                 goto out;
 714
 715         mp = br_multicast_new_group(br, port, group);
 716         err = PTR_ERR(mp);
 717         if (IS_ERR(mp))
 718                 goto err;
 719
 720         if (!port) {
 721                 mp->mglist = true;
 722                 mod_timer(&mp->timer, now + br->multicast_membership_interval);
 723                 goto out;
 724         }
 725
 726         for (pp = &mp->ports;
 727              (p = mlock_dereference(*pp, br)) != NULL;
 728              pp = &p->next) {
 729                 if (p->port == port)
 730                         goto found;
 731                 if ((unsigned long)p->port < (unsigned long)port)
 732                         break;
 733         }
 734
 735         p = kzalloc(sizeof(*p), GFP_ATOMIC);
 736         err = -ENOMEM;
 737         if (unlikely(!p))
 738                 goto err;
 739
 740         p->addr = *group;
 741         p->port = port;
 742         p->next = *pp;
 743         hlist_add_head(&p->mglist, &port->mglist);
 744         setup_timer(&p->timer, br_multicast_port_group_expired,
 745                     (unsigned long)p);
 746         setup_timer(&p->query_timer, br_multicast_port_group_query_expired,
 747                     (unsigned long)p);
 748
 749         rcu_assign_pointer(*pp, p);
 750
 751 found:
 752         mod_timer(&p->timer, now + br->multicast_membership_interval);
 753 out:
 754         err = 0;
 755
 756 err:
 757         spin_unlock(&br->multicast_lock);
 758         return err;
 759 }
 760
 761 static int br_ip4_multicast_add_group(struct net_bridge *br,
 762                                       struct net_bridge_port *port,
 763                                       __be32 group)
 764 {
 765         struct br_ip br_group;
 766
 767         if (ipv4_is_local_multicast(group))
 768                 return 0;
 769
 770         br_group.u.ip4 = group;
 771         br_group.proto = htons(ETH_P_IP);
 772
 773         return br_multicast_add_group(br, port, &br_group);
 774 }
 775
 776 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
 777 static int br_ip6_multicast_add_group(struct net_bridge *br,
 778                                       struct net_bridge_port *port,
 779                                       const struct in6_addr *group)
 780 {
 781         struct br_ip br_group;
 782
 783         if (!ipv6_is_transient_multicast(group))
 784                 return 0;
 785
 786         ipv6_addr_copy(&br_group.u.ip6, group);
 787         br_group.proto = htons(ETH_P_IPV6);
 788
 789         return br_multicast_add_group(br, port, &br_group);
 790 }
 791 #endif
 792
 793 static void br_multicast_router_expired(unsigned long data)
 794 {
 795         struct net_bridge_port *port = (void *)data;
 796         struct net_bridge *br = port->br;
 797
 798         spin_lock(&br->multicast_lock);
 799         if (port->multicast_router != 1 ||
 800             timer_pending(&port->multicast_router_timer) ||
 801             hlist_unhashed(&port->rlist))
 802                 goto out;
 803
 804         hlist_del_init_rcu(&port->rlist);
 805
 806 out:
 807         spin_unlock(&br->multicast_lock);
 808 }
 809
 810 static void br_multicast_local_router_expired(unsigned long data)
 811 {
 812 }
 813
 814 static void __br_multicast_send_query(struct net_bridge *br,
 815                                       struct net_bridge_port *port,
 816                                       struct br_ip *ip)
 817 {
 818         struct sk_buff *skb;
 819
 820         skb = br_multicast_alloc_query(br, ip);
 821         if (!skb)
 822                 return;
 823
 824         if (port) {
 825                 __skb_push(skb, sizeof(struct ethhdr));
 826                 skb->dev = port->dev;
 827                 NF_HOOK(NFPROTO_BRIDGE, NF_BR_LOCAL_OUT, skb, NULL, skb->dev,
 828                         dev_queue_xmit);
 829         } else
 830                 netif_rx(skb);
 831 }
 832
 833 static void br_multicast_send_query(struct net_bridge *br,
 834                                     struct net_bridge_port *port, u32 sent)
 835 {
 836         unsigned long time;
 837         struct br_ip br_group;
 838
 839         if (!netif_running(br->dev) || br->multicast_disabled ||
 840             timer_pending(&br->multicast_querier_timer))
 841                 return;
 842
 843         memset(&br_group.u, 0, sizeof(br_group.u));
 844
 845         br_group.proto = htons(ETH_P_IP);
 846         __br_multicast_send_query(br, port, &br_group);
 847
 848 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
 849         br_group.proto = htons(ETH_P_IPV6);
 850         __br_multicast_send_query(br, port, &br_group);
 851 #endif
 852
 853         time = jiffies;
 854         time += sent < br->multicast_startup_query_count ?
 855                 br->multicast_startup_query_interval :
 856                 br->multicast_query_interval;
 857         mod_timer(port ? &port->multicast_query_timer :
 858                          &br->multicast_query_timer, time);
 859 }
 860
 861 static void br_multicast_port_query_expired(unsigned long data)
 862 {
 863         struct net_bridge_port *port = (void *)data;
 864         struct net_bridge *br = port->br;
 865
 866         spin_lock(&br->multicast_lock);
 867         if (port->state == BR_STATE_DISABLED ||
 868             port->state == BR_STATE_BLOCKING)
 869                 goto out;
 870
 871         if (port->multicast_startup_queries_sent <
 872             br->multicast_startup_query_count)
 873                 port->multicast_startup_queries_sent++;
 874
 875         br_multicast_send_query(port->br, port,
 876                                 port->multicast_startup_queries_sent);
 877
 878 out:
 879         spin_unlock(&br->multicast_lock);
 880 }
 881
 882 void br_multicast_add_port(struct net_bridge_port *port)
 883 {
 884         port->multicast_router = 1;
 885
 886         setup_timer(&port->multicast_router_timer, br_multicast_router_expired,
 887                     (unsigned long)port);
 888         setup_timer(&port->multicast_query_timer,
 889                     br_multicast_port_query_expired, (unsigned long)port);
 890 }
 891
 892 void br_multicast_del_port(struct net_bridge_port *port)
 893 {
 894         del_timer_sync(&port->multicast_router_timer);
 895 }
 896
 897 static void __br_multicast_enable_port(struct net_bridge_port *port)
 898 {
 899         port->multicast_startup_queries_sent = 0;
 900
 901         if (try_to_del_timer_sync(&port->multicast_query_timer) >= 0 ||
 902             del_timer(&port->multicast_query_timer))
 903                 mod_timer(&port->multicast_query_timer, jiffies);
 904 }
 905
 906 void br_multicast_enable_port(struct net_bridge_port *port)
 907 {
 908         struct net_bridge *br = port->br;
 909
 910         spin_lock(&br->multicast_lock);
 911         if (br->multicast_disabled || !netif_running(br->dev))
 912                 goto out;
 913
 914         __br_multicast_enable_port(port);
 915
 916 out:
 917         spin_unlock(&br->multicast_lock);
 918 }
 919
 920 void br_multicast_disable_port(struct net_bridge_port *port)
 921 {
 922         struct net_bridge *br = port->br;
 923         struct net_bridge_port_group *pg;
 924         struct hlist_node *p, *n;
 925
 926         spin_lock(&br->multicast_lock);
 927         hlist_for_each_entry_safe(pg, p, n, &port->mglist, mglist)
 928                 br_multicast_del_pg(br, pg);
 929
 930         if (!hlist_unhashed(&port->rlist))
 931                 hlist_del_init_rcu(&port->rlist);
 932         del_timer(&port->multicast_router_timer);
 933         del_timer(&port->multicast_query_timer);
 934         spin_unlock(&br->multicast_lock);
 935 }
 936
 937 static int br_ip4_multicast_igmp3_report(struct net_bridge *br,
 938                                          struct net_bridge_port *port,
 939                                          struct sk_buff *skb)
 940 {
 941         struct igmpv3_report *ih;
 942         struct igmpv3_grec *grec;
 943         int i;
 944         int len;
 945         int num;
 946         int type;
 947         int err = 0;
 948         __be32 group;
 949
 950         if (!pskb_may_pull(skb, sizeof(*ih)))
 951                 return -EINVAL;
 952
 953         ih = igmpv3_report_hdr(skb);
 954         num = ntohs(ih->ngrec);
 955         len = sizeof(*ih);
 956
 957         for (i = 0; i < num; i++) {
 958                 len += sizeof(*grec);
 959                 if (!pskb_may_pull(skb, len))
 960                         return -EINVAL;
 961
 962                 grec = (void *)(skb->data + len - sizeof(*grec));
 963                 group = grec->grec_mca;
 964                 type = grec->grec_type;
 965
 966                 len += ntohs(grec->grec_nsrcs) * 4;
 967                 if (!pskb_may_pull(skb, len))
 968                         return -EINVAL;
 969
 970                 /* We treat this as an IGMPv2 report for now. */
 971                 switch (type) {
 972                 case IGMPV3_MODE_IS_INCLUDE:
 973                 case IGMPV3_MODE_IS_EXCLUDE:
 974                 case IGMPV3_CHANGE_TO_INCLUDE:
 975                 case IGMPV3_CHANGE_TO_EXCLUDE:
 976                 case IGMPV3_ALLOW_NEW_SOURCES:
 977                 case IGMPV3_BLOCK_OLD_SOURCES:
 978                         break;
 979
 980                 default:
 981                         continue;
 982                 }
 983
 984                 err = br_ip4_multicast_add_group(br, port, group);
 985                 if (err)
 986                         break;
 987         }
 988
 989         return err;
 990 }
 991
 992 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
 993 static int br_ip6_multicast_mld2_report(struct net_bridge *br,
 994                                         struct net_bridge_port *port,
 995                                         struct sk_buff *skb)
 996 {
 997         struct icmp6hdr *icmp6h;
 998         struct mld2_grec *grec;
 999         int i;
1000         int len;
1001         int num;
1002         int err = 0;
1003
1004         if (!pskb_may_pull(skb, sizeof(*icmp6h)))
1005                 return -EINVAL;
1006
1007         icmp6h = icmp6_hdr(skb);
1008         num = ntohs(icmp6h->icmp6_dataun.un_data16[1]);
1009         len = sizeof(*icmp6h);
1010
1011         for (i = 0; i < num; i++) {
1012                 __be16 *nsrcs, _nsrcs;
1013
1014                 nsrcs = skb_header_pointer(skb,
1015                                            len + offsetof(struct mld2_grec,
1016                                                           grec_nsrcs),
1017                                            sizeof(_nsrcs), &_nsrcs);
1018                 if (!nsrcs)
1019                         return -EINVAL;
1020
1021                 if (!pskb_may_pull(skb,
1022                                    len + sizeof(*grec) +
1023                                    sizeof(struct in6_addr) * ntohs(*nsrcs)))
1024                         return -EINVAL;
1025
1026                 grec = (struct mld2_grec *)(skb->data + len);
1027                 len += sizeof(*grec) +
1028                        sizeof(struct in6_addr) * ntohs(*nsrcs);
1029
1030                 /* We treat these as MLDv1 reports for now. */
1031                 switch (grec->grec_type) {
1032                 case MLD2_MODE_IS_INCLUDE:
1033                 case MLD2_MODE_IS_EXCLUDE:
1034                 case MLD2_CHANGE_TO_INCLUDE:
1035                 case MLD2_CHANGE_TO_EXCLUDE:
1036                 case MLD2_ALLOW_NEW_SOURCES:
1037                 case MLD2_BLOCK_OLD_SOURCES:
1038                         break;
1039
1040                 default:
1041                         continue;
1042                 }
1043
1044                 err = br_ip6_multicast_add_group(br, port, &grec->grec_mca);
1045                 if (!err)
1046                         break;
1047         }
1048
1049         return err;
1050 }
1051 #endif
1052
1053 /*
1054  * Add port to rotuer_list
1055  *  list is maintained ordered by pointer value
1056  *  and locked by br->multicast_lock and RCU
1057  */
1058 static void br_multicast_add_router(struct net_bridge *br,
1059                                     struct net_bridge_port *port)
1060 {
1061         struct net_bridge_port *p;
1062         struct hlist_node *n, *slot = NULL;
1063
1064         hlist_for_each_entry(p, n, &br->router_list, rlist) {
1065                 if ((unsigned long) port >= (unsigned long) p)
1066                         break;
1067                 slot = n;
1068         }
1069
1070         if (slot)
1071                 hlist_add_after_rcu(slot, &port->rlist);
1072         else
1073                 hlist_add_head_rcu(&port->rlist, &br->router_list);
1074 }
1075
1076 static void br_multicast_mark_router(struct net_bridge *br,
1077                                      struct net_bridge_port *port)
1078 {
1079         unsigned long now = jiffies;
1080
1081         if (!port) {
1082                 if (br->multicast_router == 1)
1083                         mod_timer(&br->multicast_router_timer,
1084                                   now + br->multicast_querier_interval);
1085                 return;
1086         }
1087
1088         if (port->multicast_router != 1)
1089                 return;
1090
1091         if (!hlist_unhashed(&port->rlist))
1092                 goto timer;
1093
1094         br_multicast_add_router(br, port);
1095
1096 timer:
1097         mod_timer(&port->multicast_router_timer,
1098                   now + br->multicast_querier_interval);
1099 }
1100
1101 static void br_multicast_query_received(struct net_bridge *br,
1102                                         struct net_bridge_port *port,
1103                                         int saddr)
1104 {
1105         if (saddr)
1106                 mod_timer(&br->multicast_querier_timer,
1107                           jiffies + br->multicast_querier_interval);
1108         else if (timer_pending(&br->multicast_querier_timer))
1109                 return;
1110
1111         br_multicast_mark_router(br, port);
1112 }
1113
1114 static int br_ip4_multicast_query(struct net_bridge *br,
1115                                   struct net_bridge_port *port,
1116                                   struct sk_buff *skb)
1117 {
1118         const struct iphdr *iph = ip_hdr(skb);
1119         struct igmphdr *ih = igmp_hdr(skb);
1120         struct net_bridge_mdb_entry *mp;
1121         struct igmpv3_query *ih3;
1122         struct net_bridge_port_group *p;
1123         struct net_bridge_port_group __rcu **pp;
1124         unsigned long max_delay;
1125         unsigned long now = jiffies;
1126         __be32 group;
1127         int err = 0;
1128
1129         spin_lock(&br->multicast_lock);
1130         if (!netif_running(br->dev) ||
1131             (port && port->state == BR_STATE_DISABLED))
1132                 goto out;
1133
1134         br_multicast_query_received(br, port, !!iph->saddr);
1135
1136         group = ih->group;
1137
1138         if (skb->len == sizeof(*ih)) {
1139                 max_delay = ih->code * (HZ / IGMP_TIMER_SCALE);
1140
1141                 if (!max_delay) {
1142                         max_delay = 10 * HZ;
1143                         group = 0;
1144                 }
1145         } else {
1146                 if (!pskb_may_pull(skb, sizeof(struct igmpv3_query))) {
1147                         err = -EINVAL;
1148                         goto out;
1149                 }
1150
1151                 ih3 = igmpv3_query_hdr(skb);
1152                 if (ih3->nsrcs)
1153                         goto out;
1154
1155                 max_delay = ih3->code ?
1156                             IGMPV3_MRC(ih3->code) * (HZ / IGMP_TIMER_SCALE) : 1;
1157         }
1158
1159         if (!group)
1160                 goto out;
1161
1162         mp = br_mdb_ip4_get(mlock_dereference(br->mdb, br), group);
1163         if (!mp)
1164                 goto out;
1165
1166         max_delay *= br->multicast_last_member_count;
1167
1168         if (mp->mglist &&
1169             (timer_pending(&mp->timer) ?
1170              time_after(mp->timer.expires, now + max_delay) :
1171              try_to_del_timer_sync(&mp->timer) >= 0))
1172                 mod_timer(&mp->timer, now + max_delay);
1173
1174         for (pp = &mp->ports;
1175              (p = mlock_dereference(*pp, br)) != NULL;
1176              pp = &p->next) {
1177                 if (timer_pending(&p->timer) ?
1178                     time_after(p->timer.expires, now + max_delay) :
1179                     try_to_del_timer_sync(&p->timer) >= 0)
1180                         mod_timer(&p->timer, now + max_delay);
1181         }
1182
1183 out:
1184         spin_unlock(&br->multicast_lock);
1185         return err;
1186 }
1187
1188 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
1189 static int br_ip6_multicast_query(struct net_bridge *br,
1190                                   struct net_bridge_port *port,
1191                                   struct sk_buff *skb)
1192 {
1193         const struct ipv6hdr *ip6h = ipv6_hdr(skb);
1194         struct mld_msg *mld = (struct mld_msg *) icmp6_hdr(skb);
1195         struct net_bridge_mdb_entry *mp;
1196         struct mld2_query *mld2q;
1197         struct net_bridge_port_group *p;
1198         struct net_bridge_port_group __rcu **pp;
1199         unsigned long max_delay;
1200         unsigned long now = jiffies;
1201         const struct in6_addr *group = NULL;
1202         int err = 0;
1203
1204         spin_lock(&br->multicast_lock);
1205         if (!netif_running(br->dev) ||
1206             (port && port->state == BR_STATE_DISABLED))
1207                 goto out;
1208
1209         br_multicast_query_received(br, port, !ipv6_addr_any(&ip6h->saddr));
1210
1211         if (skb->len == sizeof(*mld)) {
1212                 if (!pskb_may_pull(skb, sizeof(*mld))) {
1213                         err = -EINVAL;
1214                         goto out;
1215                 }
1216                 mld = (struct mld_msg *) icmp6_hdr(skb);
1217                 max_delay = msecs_to_jiffies(htons(mld->mld_maxdelay));
1218                 if (max_delay)
1219                         group = &mld->mld_mca;
1220         } else if (skb->len >= sizeof(*mld2q)) {
1221                 if (!pskb_may_pull(skb, sizeof(*mld2q))) {
1222                         err = -EINVAL;
1223                         goto out;
1224                 }
1225                 mld2q = (struct mld2_query *)icmp6_hdr(skb);
1226                 if (!mld2q->mld2q_nsrcs)
1227                         group = &mld2q->mld2q_mca;
1228                 max_delay = mld2q->mld2q_mrc ? MLDV2_MRC(mld2q->mld2q_mrc) : 1;
1229         }
1230
1231         if (!group)
1232                 goto out;
1233
1234         mp = br_mdb_ip6_get(mlock_dereference(br->mdb, br), group);
1235         if (!mp)
1236                 goto out;
1237
1238         max_delay *= br->multicast_last_member_count;
1239         if (mp->mglist &&
1240             (timer_pending(&mp->timer) ?
1241              time_after(mp->timer.expires, now + max_delay) :
1242              try_to_del_timer_sync(&mp->timer) >= 0))
1243                 mod_timer(&mp->timer, now + max_delay);
1244
1245         for (pp = &mp->ports;
1246              (p = mlock_dereference(*pp, br)) != NULL;
1247              pp = &p->next) {
1248                 if (timer_pending(&p->timer) ?
1249                     time_after(p->timer.expires, now + max_delay) :
1250                     try_to_del_timer_sync(&p->timer) >= 0)
1251                         mod_timer(&p->timer, now + max_delay);
1252         }
1253
1254 out:
1255         spin_unlock(&br->multicast_lock);
1256         return err;
1257 }
1258 #endif
1259
1260 static void br_multicast_leave_group(struct net_bridge *br,
1261                                      struct net_bridge_port *port,
1262                                      struct br_ip *group)
1263 {
1264         struct net_bridge_mdb_htable *mdb;
1265         struct net_bridge_mdb_entry *mp;
1266         struct net_bridge_port_group *p;
1267         unsigned long now;
1268         unsigned long time;
1269
1270         spin_lock(&br->multicast_lock);
1271         if (!netif_running(br->dev) ||
1272             (port && port->state == BR_STATE_DISABLED) ||
1273             timer_pending(&br->multicast_querier_timer))
1274                 goto out;
1275
1276         mdb = mlock_dereference(br->mdb, br);
1277         mp = br_mdb_ip_get(mdb, group);
1278         if (!mp)
1279                 goto out;
1280
1281         now = jiffies;
1282         time = now + br->multicast_last_member_count *
1283                      br->multicast_last_member_interval;
1284
1285         if (!port) {
1286                 if (mp->mglist &&
1287                     (timer_pending(&mp->timer) ?
1288                      time_after(mp->timer.expires, time) :
1289                      try_to_del_timer_sync(&mp->timer) >= 0)) {
1290                         mod_timer(&mp->timer, time);
1291
1292                         mp->queries_sent = 0;
1293                         mod_timer(&mp->query_timer, now);
1294                 }
1295
1296                 goto out;
1297         }
1298
1299         for (p = mlock_dereference(mp->ports, br);
1300              p != NULL;
1301              p = mlock_dereference(p->next, br)) {
1302                 if (p->port != port)
1303                         continue;
1304
1305                 if (!hlist_unhashed(&p->mglist) &&
1306                     (timer_pending(&p->timer) ?
1307                      time_after(p->timer.expires, time) :
1308                      try_to_del_timer_sync(&p->timer) >= 0)) {
1309                         mod_timer(&p->timer, time);
1310
1311                         p->queries_sent = 0;
1312                         mod_timer(&p->query_timer, now);
1313                 }
1314
1315                 break;
1316         }
1317
1318 out:
1319         spin_unlock(&br->multicast_lock);
1320 }
1321
1322 static void br_ip4_multicast_leave_group(struct net_bridge *br,
1323                                          struct net_bridge_port *port,
1324                                          __be32 group)
1325 {
1326         struct br_ip br_group;
1327
1328         if (ipv4_is_local_multicast(group))
1329                 return;
1330
1331         br_group.u.ip4 = group;
1332         br_group.proto = htons(ETH_P_IP);
1333
1334         br_multicast_leave_group(br, port, &br_group);
1335 }
1336
1337 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
1338 static void br_ip6_multicast_leave_group(struct net_bridge *br,
1339                                          struct net_bridge_port *port,
1340                                          const struct in6_addr *group)
1341 {
1342         struct br_ip br_group;
1343
1344         if (!ipv6_is_transient_multicast(group))
1345                 return;
1346
1347         ipv6_addr_copy(&br_group.u.ip6, group);
1348         br_group.proto = htons(ETH_P_IPV6);
1349
1350         br_multicast_leave_group(br, port, &br_group);
1351 }
1352 #endif
1353
1354 static int br_multicast_ipv4_rcv(struct net_bridge *br,
1355                                  struct net_bridge_port *port,
1356                                  struct sk_buff *skb)
1357 {
1358         struct sk_buff *skb2 = skb;
1359         const struct iphdr *iph;
1360         struct igmphdr *ih;
1361         unsigned len;
1362         unsigned offset;
1363         int err;
1364
1365         /* We treat OOM as packet loss for now. */
1366         if (!pskb_may_pull(skb, sizeof(*iph)))
1367                 return -EINVAL;
1368
1369         iph = ip_hdr(skb);
1370
1371         if (iph->ihl < 5 || iph->version != 4)
1372                 return -EINVAL;
1373
1374         if (!pskb_may_pull(skb, ip_hdrlen(skb)))
1375                 return -EINVAL;
1376
1377         iph = ip_hdr(skb);
1378
1379         if (unlikely(ip_fast_csum((u8 *)iph, iph->ihl)))
1380                 return -EINVAL;
1381
1382         if (iph->protocol != IPPROTO_IGMP) {
1383                 if ((iph->daddr & IGMP_LOCAL_GROUP_MASK) != IGMP_LOCAL_GROUP)
1384                         BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1385                 return 0;
1386         }
1387
1388         len = ntohs(iph->tot_len);
1389         if (skb->len < len || len < ip_hdrlen(skb))
1390                 return -EINVAL;
1391
1392         if (skb->len > len) {
1393                 skb2 = skb_clone(skb, GFP_ATOMIC);
1394                 if (!skb2)
1395                         return -ENOMEM;
1396
1397                 err = pskb_trim_rcsum(skb2, len);
1398                 if (err)
1399                         goto err_out;
1400         }
1401
1402         len -= ip_hdrlen(skb2);
1403         offset = skb_network_offset(skb2) + ip_hdrlen(skb2);
1404         __skb_pull(skb2, offset);
1405         skb_reset_transport_header(skb2);
1406
1407         err = -EINVAL;
1408         if (!pskb_may_pull(skb2, sizeof(*ih)))
1409                 goto out;
1410
1411         switch (skb2->ip_summed) {
1412         case CHECKSUM_COMPLETE:
1413                 if (!csum_fold(skb2->csum))
1414                         break;
1415                 /* fall through */
1416         case CHECKSUM_NONE:
1417                 skb2->csum = 0;
1418                 if (skb_checksum_complete(skb2))
1419                         goto out;
1420         }
1421
1422         err = 0;
1423
1424         BR_INPUT_SKB_CB(skb)->igmp = 1;
1425         ih = igmp_hdr(skb2);
1426
1427         switch (ih->type) {
1428         case IGMP_HOST_MEMBERSHIP_REPORT:
1429         case IGMPV2_HOST_MEMBERSHIP_REPORT:
1430                 BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1431                 err = br_ip4_multicast_add_group(br, port, ih->group);
1432                 break;
1433         case IGMPV3_HOST_MEMBERSHIP_REPORT:
1434                 err = br_ip4_multicast_igmp3_report(br, port, skb2);
1435                 break;
1436         case IGMP_HOST_MEMBERSHIP_QUERY:
1437                 err = br_ip4_multicast_query(br, port, skb2);
1438                 break;
1439         case IGMP_HOST_LEAVE_MESSAGE:
1440                 br_ip4_multicast_leave_group(br, port, ih->group);
1441                 break;
1442         }
1443
1444 out:
1445         __skb_push(skb2, offset);
1446 err_out:
1447         if (skb2 != skb)
1448                 kfree_skb(skb2);
1449         return err;
1450 }
1451
1452 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
1453 static int br_multicast_ipv6_rcv(struct net_bridge *br,
1454                                  struct net_bridge_port *port,
1455                                  struct sk_buff *skb)
1456 {
1457         struct sk_buff *skb2;
1458         const struct ipv6hdr *ip6h;
1459         struct icmp6hdr *icmp6h;
1460         u8 nexthdr;
1461         unsigned len;
1462         int offset;
1463         int err;
1464
1465         if (!pskb_may_pull(skb, sizeof(*ip6h)))
1466                 return -EINVAL;
1467
1468         ip6h = ipv6_hdr(skb);
1469
1470         /*
1471          * We're interested in MLD messages only.
1472          *  - Version is 6
1473          *  - MLD has always Router Alert hop-by-hop option
1474          *  - But we do not support jumbrograms.
1475          */
1476         if (ip6h->version != 6 ||
1477             ip6h->nexthdr != IPPROTO_HOPOPTS ||
1478             ip6h->payload_len == 0)
1479                 return 0;
1480
1481         len = ntohs(ip6h->payload_len) + sizeof(*ip6h);
1482         if (skb->len < len)
1483                 return -EINVAL;
1484
1485         nexthdr = ip6h->nexthdr;
1486         offset = ipv6_skip_exthdr(skb, sizeof(*ip6h), &nexthdr);
1487
1488         if (offset < 0 || nexthdr != IPPROTO_ICMPV6)
1489                 return 0;
1490
1491         /* Okay, we found ICMPv6 header */
1492         skb2 = skb_clone(skb, GFP_ATOMIC);
1493         if (!skb2)
1494                 return -ENOMEM;
1495
1496         err = -EINVAL;
1497         if (!pskb_may_pull(skb2, offset + sizeof(struct icmp6hdr)))
1498                 goto out;
1499
1500         len -= offset - skb_network_offset(skb2);
1501
1502         __skb_pull(skb2, offset);
1503         skb_reset_transport_header(skb2);
1504
1505         icmp6h = icmp6_hdr(skb2);
1506
1507         switch (icmp6h->icmp6_type) {
1508         case ICMPV6_MGM_QUERY:
1509         case ICMPV6_MGM_REPORT:
1510         case ICMPV6_MGM_REDUCTION:
1511         case ICMPV6_MLD2_REPORT:
1512                 break;
1513         default:
1514                 err = 0;
1515                 goto out;
1516         }
1517
1518         /* Okay, we found MLD message. Check further. */
1519         if (skb2->len > len) {
1520                 err = pskb_trim_rcsum(skb2, len);
1521                 if (err)
1522                         goto out;
1523         }
1524
1525         switch (skb2->ip_summed) {
1526         case CHECKSUM_COMPLETE:
1527                 if (!csum_fold(skb2->csum))
1528                         break;
1529                 /*FALLTHROUGH*/
1530         case CHECKSUM_NONE:
1531                 skb2->csum = 0;
1532                 if (skb_checksum_complete(skb2))
1533                         goto out;
1534         }
1535
1536         err = 0;
1537
1538         BR_INPUT_SKB_CB(skb)->igmp = 1;
1539
1540         switch (icmp6h->icmp6_type) {
1541         case ICMPV6_MGM_REPORT:
1542             {
1543                 struct mld_msg *mld;
1544                 if (!pskb_may_pull(skb2, sizeof(*mld))) {
1545                         err = -EINVAL;
1546                         goto out;
1547                 }
1548                 mld = (struct mld_msg *)skb_transport_header(skb2);
1549                 BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1550                 err = br_ip6_multicast_add_group(br, port, &mld->mld_mca);
1551                 break;
1552             }
1553         case ICMPV6_MLD2_REPORT:
1554                 err = br_ip6_multicast_mld2_report(br, port, skb2);
1555                 break;
1556         case ICMPV6_MGM_QUERY:
1557                 err = br_ip6_multicast_query(br, port, skb2);
1558                 break;
1559         case ICMPV6_MGM_REDUCTION:
1560             {
1561                 struct mld_msg *mld;
1562                 if (!pskb_may_pull(skb2, sizeof(*mld))) {
1563                         err = -EINVAL;
1564                         goto out;
1565                 }
1566                 mld = (struct mld_msg *)skb_transport_header(skb2);
1567                 br_ip6_multicast_leave_group(br, port, &mld->mld_mca);
1568             }
1569         }
1570
1571 out:
1572         kfree_skb(skb2);
1573         return err;
1574 }
1575 #endif
1576
1577 int br_multicast_rcv(struct net_bridge *br, struct net_bridge_port *port,
1578                      struct sk_buff *skb)
1579 {
1580         BR_INPUT_SKB_CB(skb)->igmp = 0;
1581         BR_INPUT_SKB_CB(skb)->mrouters_only = 0;
1582
1583         if (br->multicast_disabled)
1584                 return 0;
1585
1586         switch (skb->protocol) {
1587         case htons(ETH_P_IP):
1588                 return br_multicast_ipv4_rcv(br, port, skb);
1589 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
1590         case htons(ETH_P_IPV6):
1591                 return br_multicast_ipv6_rcv(br, port, skb);
1592 #endif
1593         }
1594
1595         return 0;
1596 }
1597
1598 static void br_multicast_query_expired(unsigned long data)
1599 {
1600         struct net_bridge *br = (void *)data;
1601
1602         spin_lock(&br->multicast_lock);
1603         if (br->multicast_startup_queries_sent <
1604             br->multicast_startup_query_count)
1605                 br->multicast_startup_queries_sent++;
1606
1607         br_multicast_send_query(br, NULL, br->multicast_startup_queries_sent);
1608
1609         spin_unlock(&br->multicast_lock);
1610 }
1611
1612 void br_multicast_init(struct net_bridge *br)
1613 {
1614         br->hash_elasticity = 4;
1615         br->hash_max = 512;
1616
1617         br->multicast_router = 1;
1618         br->multicast_last_member_count = 2;
1619         br->multicast_startup_query_count = 2;
1620
1621         br->multicast_last_member_interval = HZ;
1622         br->multicast_query_response_interval = 10 * HZ;
1623         br->multicast_startup_query_interval = 125 * HZ / 4;
1624         br->multicast_query_interval = 125 * HZ;
1625         br->multicast_querier_interval = 255 * HZ;
1626         br->multicast_membership_interval = 260 * HZ;
1627
1628         spin_lock_init(&br->multicast_lock);
1629         setup_timer(&br->multicast_router_timer,
1630                     br_multicast_local_router_expired, 0);
1631         setup_timer(&br->multicast_querier_timer,
1632                     br_multicast_local_router_expired, 0);
1633         setup_timer(&br->multicast_query_timer, br_multicast_query_expired,
1634                     (unsigned long)br);
1635 }
1636
1637 void br_multicast_open(struct net_bridge *br)
1638 {
1639         br->multicast_startup_queries_sent = 0;
1640
1641         if (br->multicast_disabled)
1642                 return;
1643
1644         mod_timer(&br->multicast_query_timer, jiffies);
1645 }
1646
1647 void br_multicast_stop(struct net_bridge *br)
1648 {
1649         struct net_bridge_mdb_htable *mdb;
1650         struct net_bridge_mdb_entry *mp;
1651         struct hlist_node *p, *n;
1652         u32 ver;
1653         int i;
1654
1655         del_timer_sync(&br->multicast_router_timer);
1656         del_timer_sync(&br->multicast_querier_timer);
1657         del_timer_sync(&br->multicast_query_timer);
1658
1659         spin_lock_bh(&br->multicast_lock);
1660         mdb = mlock_dereference(br->mdb, br);
1661         if (!mdb)
1662                 goto out;
1663
1664         br->mdb = NULL;
1665
1666         ver = mdb->ver;
1667         for (i = 0; i < mdb->max; i++) {
1668                 hlist_for_each_entry_safe(mp, p, n, &mdb->mhash[i],
1669                                           hlist[ver]) {
1670                         del_timer(&mp->timer);
1671                         del_timer(&mp->query_timer);
1672                         call_rcu_bh(&mp->rcu, br_multicast_free_group);
1673                 }
1674         }
1675
1676         if (mdb->old) {
1677                 spin_unlock_bh(&br->multicast_lock);
1678                 rcu_barrier_bh();
1679                 spin_lock_bh(&br->multicast_lock);
1680                 WARN_ON(mdb->old);
1681         }
1682
1683         mdb->old = mdb;
1684         call_rcu_bh(&mdb->rcu, br_mdb_free);
1685
1686 out:
1687         spin_unlock_bh(&br->multicast_lock);
1688 }
1689
1690 int br_multicast_set_router(struct net_bridge *br, unsigned long val)
1691 {
1692         int err = -ENOENT;
1693
1694         spin_lock_bh(&br->multicast_lock);
1695         if (!netif_running(br->dev))
1696                 goto unlock;
1697
1698         switch (val) {
1699         case 0:
1700         case 2:
1701                 del_timer(&br->multicast_router_timer);
1702                 /* fall through */
1703         case 1:
1704                 br->multicast_router = val;
1705                 err = 0;
1706                 break;
1707
1708         default:
1709                 err = -EINVAL;
1710                 break;
1711         }
1712
1713 unlock:
1714         spin_unlock_bh(&br->multicast_lock);
1715
1716         return err;
1717 }
1718
1719 int br_multicast_set_port_router(struct net_bridge_port *p, unsigned long val)
1720 {
1721         struct net_bridge *br = p->br;
1722         int err = -ENOENT;
1723
1724         spin_lock(&br->multicast_lock);
1725         if (!netif_running(br->dev) || p->state == BR_STATE_DISABLED)
1726                 goto unlock;
1727
1728         switch (val) {
1729         case 0:
1730         case 1:
1731         case 2:
1732                 p->multicast_router = val;
1733                 err = 0;
1734
1735                 if (val < 2 && !hlist_unhashed(&p->rlist))
1736                         hlist_del_init_rcu(&p->rlist);
1737
1738                 if (val == 1)
1739                         break;
1740
1741                 del_timer(&p->multicast_router_timer);
1742
1743                 if (val == 0)
1744                         break;
1745
1746                 br_multicast_add_router(br, p);
1747                 break;
1748
1749         default:
1750                 err = -EINVAL;
1751                 break;
1752         }
1753
1754 unlock:
1755         spin_unlock(&br->multicast_lock);
1756
1757         return err;
1758 }
1759
1760 int br_multicast_toggle(struct net_bridge *br, unsigned long val)
1761 {
1762         struct net_bridge_port *port;
1763         int err = 0;
1764         struct net_bridge_mdb_htable *mdb;
1765
1766         spin_lock(&br->multicast_lock);
1767         if (br->multicast_disabled == !val)
1768                 goto unlock;
1769
1770         br->multicast_disabled = !val;
1771         if (br->multicast_disabled)
1772                 goto unlock;
1773
1774         if (!netif_running(br->dev))
1775                 goto unlock;
1776
1777         mdb = mlock_dereference(br->mdb, br);
1778         if (mdb) {
1779                 if (mdb->old) {
1780                         err = -EEXIST;
1781 rollback:
1782                         br->multicast_disabled = !!val;
1783                         goto unlock;
1784                 }
1785
1786                 err = br_mdb_rehash(&br->mdb, mdb->max,
1787                                     br->hash_elasticity);
1788                 if (err)
1789                         goto rollback;
1790         }
1791
1792         br_multicast_open(br);
1793         list_for_each_entry(port, &br->port_list, list) {
1794                 if (port->state == BR_STATE_DISABLED ||
1795                     port->state == BR_STATE_BLOCKING)
1796                         continue;
1797
1798                 __br_multicast_enable_port(port);
1799         }
1800
1801 unlock:
1802         spin_unlock(&br->multicast_lock);
1803
1804         return err;
1805 }
1806
1807 int br_multicast_set_hash_max(struct net_bridge *br, unsigned long val)
1808 {
1809         int err = -ENOENT;
1810         u32 old;
1811         struct net_bridge_mdb_htable *mdb;
1812
1813         spin_lock(&br->multicast_lock);
1814         if (!netif_running(br->dev))
1815                 goto unlock;
1816
1817         err = -EINVAL;
1818         if (!is_power_of_2(val))
1819                 goto unlock;
1820
1821         mdb = mlock_dereference(br->mdb, br);
1822         if (mdb && val < mdb->size)
1823                 goto unlock;
1824
1825         err = 0;
1826
1827         old = br->hash_max;
1828         br->hash_max = val;
1829
1830         if (mdb) {
1831                 if (mdb->old) {
1832                         err = -EEXIST;
1833 rollback:
1834                         br->hash_max = old;
1835                         goto unlock;
1836                 }
1837
1838                 err = br_mdb_rehash(&br->mdb, br->hash_max,
1839                                     br->hash_elasticity);
1840                 if (err)
1841                         goto rollback;
1842         }
1843
1844 unlock:
1845         spin_unlock(&br->multicast_lock);
1846
1847         return err;
1848 }