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 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
  37 static inline int ipv6_is_local_multicast(const struct in6_addr *addr)
  38 {
  39         if (ipv6_addr_is_multicast(addr) &&
  40             IPV6_ADDR_MC_SCOPE(addr) <= IPV6_ADDR_SCOPE_LINKLOCAL)
  41                 return 1;
  42         return 0;
  43 }
  44 #endif
  45
  46 static inline int br_ip_equal(const struct br_ip *a, const struct br_ip *b)
  47 {
  48         if (a->proto != b->proto)
  49                 return 0;
  50         switch (a->proto) {
  51         case htons(ETH_P_IP):
  52                 return a->u.ip4 == b->u.ip4;
  53 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
  54         case htons(ETH_P_IPV6):
  55                 return ipv6_addr_equal(&a->u.ip6, &b->u.ip6);
  56 #endif
  57         }
  58         return 0;
  59 }
  60
  61 static inline int __br_ip4_hash(struct net_bridge_mdb_htable *mdb, __be32 ip)
  62 {
  63         return jhash_1word(mdb->secret, (__force u32)ip) & (mdb->max - 1);
  64 }
  65
  66 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
  67 static inline int __br_ip6_hash(struct net_bridge_mdb_htable *mdb,
  68                                 const struct in6_addr *ip)
  69 {
  70         return jhash2((__force u32 *)ip->s6_addr32, 4, mdb->secret) & (mdb->max - 1);
  71 }
  72 #endif
  73
  74 static inline int br_ip_hash(struct net_bridge_mdb_htable *mdb,
  75                              struct br_ip *ip)
  76 {
  77         switch (ip->proto) {
  78         case htons(ETH_P_IP):
  79                 return __br_ip4_hash(mdb, ip->u.ip4);
  80 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
  81         case htons(ETH_P_IPV6):
  82                 return __br_ip6_hash(mdb, &ip->u.ip6);
  83 #endif
  84         }
  85         return 0;
  86 }
  87
  88 static struct net_bridge_mdb_entry *__br_mdb_ip_get(
  89         struct net_bridge_mdb_htable *mdb, struct br_ip *dst, int hash)
  90 {
  91         struct net_bridge_mdb_entry *mp;
  92         struct hlist_node *p;
  93
  94         hlist_for_each_entry_rcu(mp, p, &mdb->mhash[hash], hlist[mdb->ver]) {
  95                 if (br_ip_equal(&mp->addr, dst))
  96                         return mp;
  97         }
  98
  99         return NULL;
 100 }
 101
 102 static struct net_bridge_mdb_entry *br_mdb_ip4_get(
 103         struct net_bridge_mdb_htable *mdb, __be32 dst)
 104 {
 105         struct br_ip br_dst;
 106
 107         br_dst.u.ip4 = dst;
 108         br_dst.proto = htons(ETH_P_IP);
 109
 110         return __br_mdb_ip_get(mdb, &br_dst, __br_ip4_hash(mdb, dst));
 111 }
 112
 113 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
 114 static struct net_bridge_mdb_entry *br_mdb_ip6_get(
 115         struct net_bridge_mdb_htable *mdb, const struct in6_addr *dst)
 116 {
 117         struct br_ip br_dst;
 118
 119         ipv6_addr_copy(&br_dst.u.ip6, dst);
 120         br_dst.proto = htons(ETH_P_IPV6);
 121
 122         return __br_mdb_ip_get(mdb, &br_dst, __br_ip6_hash(mdb, dst));
 123 }
 124 #endif
 125
 126 static struct net_bridge_mdb_entry *br_mdb_ip_get(
 127         struct net_bridge_mdb_htable *mdb, struct br_ip *dst)
 128 {
 129         return __br_mdb_ip_get(mdb, dst, br_ip_hash(mdb, dst));
 130 }
 131
 132 struct net_bridge_mdb_entry *br_mdb_get(struct net_bridge *br,
 133                                         struct sk_buff *skb)
 134 {
 135         struct net_bridge_mdb_htable *mdb = br->mdb;
 136         struct br_ip ip;
 137
 138         if (!mdb || br->multicast_disabled)
 139                 return NULL;
 140
 141         if (BR_INPUT_SKB_CB(skb)->igmp)
 142                 return NULL;
 143
 144         ip.proto = skb->protocol;
 145
 146         switch (skb->protocol) {
 147         case htons(ETH_P_IP):
 148                 ip.u.ip4 = ip_hdr(skb)->daddr;
 149                 break;
 150 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
 151         case htons(ETH_P_IPV6):
 152                 ipv6_addr_copy(&ip.u.ip6, &ipv6_hdr(skb)->daddr);
 153                 break;
 154 #endif
 155         default:
 156                 return NULL;
 157         }
 158
 159         return br_mdb_ip_get(mdb, &ip);
 160 }
 161
 162 static void br_mdb_free(struct rcu_head *head)
 163 {
 164         struct net_bridge_mdb_htable *mdb =
 165                 container_of(head, struct net_bridge_mdb_htable, rcu);
 166         struct net_bridge_mdb_htable *old = mdb->old;
 167
 168         mdb->old = NULL;
 169         kfree(old->mhash);
 170         kfree(old);
 171 }
 172
 173 static int br_mdb_copy(struct net_bridge_mdb_htable *new,
 174                        struct net_bridge_mdb_htable *old,
 175                        int elasticity)
 176 {
 177         struct net_bridge_mdb_entry *mp;
 178         struct hlist_node *p;
 179         int maxlen;
 180         int len;
 181         int i;
 182
 183         for (i = 0; i < old->max; i++)
 184                 hlist_for_each_entry(mp, p, &old->mhash[i], hlist[old->ver])
 185                         hlist_add_head(&mp->hlist[new->ver],
 186                                        &new->mhash[br_ip_hash(new, &mp->addr)]);
 187
 188         if (!elasticity)
 189                 return 0;
 190
 191         maxlen = 0;
 192         for (i = 0; i < new->max; i++) {
 193                 len = 0;
 194                 hlist_for_each_entry(mp, p, &new->mhash[i], hlist[new->ver])
 195                         len++;
 196                 if (len > maxlen)
 197                         maxlen = len;
 198         }
 199
 200         return maxlen > elasticity ? -EINVAL : 0;
 201 }
 202
 203 static void br_multicast_free_pg(struct rcu_head *head)
 204 {
 205         struct net_bridge_port_group *p =
 206                 container_of(head, struct net_bridge_port_group, rcu);
 207
 208         kfree(p);
 209 }
 210
 211 static void br_multicast_free_group(struct rcu_head *head)
 212 {
 213         struct net_bridge_mdb_entry *mp =
 214                 container_of(head, struct net_bridge_mdb_entry, rcu);
 215
 216         kfree(mp);
 217 }
 218
 219 static void br_multicast_group_expired(unsigned long data)
 220 {
 221         struct net_bridge_mdb_entry *mp = (void *)data;
 222         struct net_bridge *br = mp->br;
 223         struct net_bridge_mdb_htable *mdb;
 224
 225         spin_lock(&br->multicast_lock);
 226         if (!netif_running(br->dev) || timer_pending(&mp->timer))
 227                 goto out;
 228
 229         if (!hlist_unhashed(&mp->mglist))
 230                 hlist_del_init(&mp->mglist);
 231
 232         if (mp->ports)
 233                 goto out;
 234
 235         mdb = br->mdb;
 236         hlist_del_rcu(&mp->hlist[mdb->ver]);
 237         mdb->size--;
 238
 239         del_timer(&mp->query_timer);
 240         call_rcu_bh(&mp->rcu, br_multicast_free_group);
 241
 242 out:
 243         spin_unlock(&br->multicast_lock);
 244 }
 245
 246 static void br_multicast_del_pg(struct net_bridge *br,
 247                                 struct net_bridge_port_group *pg)
 248 {
 249         struct net_bridge_mdb_htable *mdb = br->mdb;
 250         struct net_bridge_mdb_entry *mp;
 251         struct net_bridge_port_group *p;
 252         struct net_bridge_port_group **pp;
 253
 254         mp = br_mdb_ip_get(mdb, &pg->addr);
 255         if (WARN_ON(!mp))
 256                 return;
 257
 258         for (pp = &mp->ports; (p = *pp); pp = &p->next) {
 259                 if (p != pg)
 260                         continue;
 261
 262                 rcu_assign_pointer(*pp, p->next);
 263                 hlist_del_init(&p->mglist);
 264                 del_timer(&p->timer);
 265                 del_timer(&p->query_timer);
 266                 call_rcu_bh(&p->rcu, br_multicast_free_pg);
 267
 268                 if (!mp->ports && hlist_unhashed(&mp->mglist) &&
 269                     netif_running(br->dev))
 270                         mod_timer(&mp->timer, jiffies);
 271
 272                 return;
 273         }
 274
 275         WARN_ON(1);
 276 }
 277
 278 static void br_multicast_port_group_expired(unsigned long data)
 279 {
 280         struct net_bridge_port_group *pg = (void *)data;
 281         struct net_bridge *br = pg->port->br;
 282
 283         spin_lock(&br->multicast_lock);
 284         if (!netif_running(br->dev) || timer_pending(&pg->timer) ||
 285             hlist_unhashed(&pg->mglist))
 286                 goto out;
 287
 288         br_multicast_del_pg(br, pg);
 289
 290 out:
 291         spin_unlock(&br->multicast_lock);
 292 }
 293
 294 static int br_mdb_rehash(struct net_bridge_mdb_htable **mdbp, int max,
 295                          int elasticity)
 296 {
 297         struct net_bridge_mdb_htable *old = *mdbp;
 298         struct net_bridge_mdb_htable *mdb;
 299         int err;
 300
 301         mdb = kmalloc(sizeof(*mdb), GFP_ATOMIC);
 302         if (!mdb)
 303                 return -ENOMEM;
 304
 305         mdb->max = max;
 306         mdb->old = old;
 307
 308         mdb->mhash = kzalloc(max * sizeof(*mdb->mhash), GFP_ATOMIC);
 309         if (!mdb->mhash) {
 310                 kfree(mdb);
 311                 return -ENOMEM;
 312         }
 313
 314         mdb->size = old ? old->size : 0;
 315         mdb->ver = old ? old->ver ^ 1 : 0;
 316
 317         if (!old || elasticity)
 318                 get_random_bytes(&mdb->secret, sizeof(mdb->secret));
 319         else
 320                 mdb->secret = old->secret;
 321
 322         if (!old)
 323                 goto out;
 324
 325         err = br_mdb_copy(mdb, old, elasticity);
 326         if (err) {
 327                 kfree(mdb->mhash);
 328                 kfree(mdb);
 329                 return err;
 330         }
 331
 332         call_rcu_bh(&mdb->rcu, br_mdb_free);
 333
 334 out:
 335         rcu_assign_pointer(*mdbp, mdb);
 336
 337         return 0;
 338 }
 339
 340 static struct sk_buff *br_ip4_multicast_alloc_query(struct net_bridge *br,
 341                                                     __be32 group)
 342 {
 343         struct sk_buff *skb;
 344         struct igmphdr *ih;
 345         struct ethhdr *eth;
 346         struct iphdr *iph;
 347
 348         skb = netdev_alloc_skb_ip_align(br->dev, sizeof(*eth) + sizeof(*iph) +
 349                                                  sizeof(*ih) + 4);
 350         if (!skb)
 351                 goto out;
 352
 353         skb->protocol = htons(ETH_P_IP);
 354
 355         skb_reset_mac_header(skb);
 356         eth = eth_hdr(skb);
 357
 358         memcpy(eth->h_source, br->dev->dev_addr, 6);
 359         eth->h_dest[0] = 1;
 360         eth->h_dest[1] = 0;
 361         eth->h_dest[2] = 0x5e;
 362         eth->h_dest[3] = 0;
 363         eth->h_dest[4] = 0;
 364         eth->h_dest[5] = 1;
 365         eth->h_proto = htons(ETH_P_IP);
 366         skb_put(skb, sizeof(*eth));
 367
 368         skb_set_network_header(skb, skb->len);
 369         iph = ip_hdr(skb);
 370
 371         iph->version = 4;
 372         iph->ihl = 6;
 373         iph->tos = 0xc0;
 374         iph->tot_len = htons(sizeof(*iph) + sizeof(*ih) + 4);
 375         iph->id = 0;
 376         iph->frag_off = htons(IP_DF);
 377         iph->ttl = 1;
 378         iph->protocol = IPPROTO_IGMP;
 379         iph->saddr = 0;
 380         iph->daddr = htonl(INADDR_ALLHOSTS_GROUP);
 381         ((u8 *)&iph[1])[0] = IPOPT_RA;
 382         ((u8 *)&iph[1])[1] = 4;
 383         ((u8 *)&iph[1])[2] = 0;
 384         ((u8 *)&iph[1])[3] = 0;
 385         ip_send_check(iph);
 386         skb_put(skb, 24);
 387
 388         skb_set_transport_header(skb, skb->len);
 389         ih = igmp_hdr(skb);
 390         ih->type = IGMP_HOST_MEMBERSHIP_QUERY;
 391         ih->code = (group ? br->multicast_last_member_interval :
 392                             br->multicast_query_response_interval) /
 393                    (HZ / IGMP_TIMER_SCALE);
 394         ih->group = group;
 395         ih->csum = 0;
 396         ih->csum = ip_compute_csum((void *)ih, sizeof(struct igmphdr));
 397         skb_put(skb, sizeof(*ih));
 398
 399         __skb_pull(skb, sizeof(*eth));
 400
 401 out:
 402         return skb;
 403 }
 404
 405 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
 406 static struct sk_buff *br_ip6_multicast_alloc_query(struct net_bridge *br,
 407                                                     struct in6_addr *group)
 408 {
 409         struct sk_buff *skb;
 410         struct ipv6hdr *ip6h;
 411         struct mld_msg *mldq;
 412         struct ethhdr *eth;
 413         u8 *hopopt;
 414         unsigned long interval;
 415
 416         skb = netdev_alloc_skb_ip_align(br->dev, sizeof(*eth) + sizeof(*ip6h) +
 417                                                  8 + sizeof(*mldq));
 418         if (!skb)
 419                 goto out;
 420
 421         skb->protocol = htons(ETH_P_IPV6);
 422
 423         /* Ethernet header */
 424         skb_reset_mac_header(skb);
 425         eth = eth_hdr(skb);
 426
 427         memcpy(eth->h_source, br->dev->dev_addr, 6);
 428         ipv6_eth_mc_map(group, eth->h_dest);
 429         eth->h_proto = htons(ETH_P_IPV6);
 430         skb_put(skb, sizeof(*eth));
 431
 432         /* IPv6 header + HbH option */
 433         skb_set_network_header(skb, skb->len);
 434         ip6h = ipv6_hdr(skb);
 435
 436         *(__force __be32 *)ip6h = htonl(0x60000000);
 437         ip6h->payload_len = 8 + sizeof(*mldq);
 438         ip6h->nexthdr = IPPROTO_HOPOPTS;
 439         ip6h->hop_limit = 1;
 440         ipv6_addr_set(&ip6h->saddr, 0, 0, 0, 0);
 441         ipv6_addr_set(&ip6h->daddr, htonl(0xff020000), 0, 0, htonl(1));
 442
 443         hopopt = (u8 *)(ip6h + 1);
 444         hopopt[0] = IPPROTO_ICMPV6;             /* next hdr */
 445         hopopt[1] = 0;                          /* length of HbH */
 446         hopopt[2] = IPV6_TLV_ROUTERALERT;       /* Router Alert */
 447         hopopt[3] = 2;                          /* Length of RA Option */
 448         hopopt[4] = 0;                          /* Type = 0x0000 (MLD) */
 449         hopopt[5] = 0;
 450         hopopt[6] = IPV6_TLV_PAD0;              /* Pad0 */
 451         hopopt[7] = IPV6_TLV_PAD0;              /* Pad0 */
 452
 453         skb_put(skb, sizeof(*ip6h) + 8);
 454
 455         /* ICMPv6 */
 456         skb_set_transport_header(skb, skb->len);
 457         mldq = (struct mld_msg *) icmp6_hdr(skb);
 458
 459         interval = ipv6_addr_any(group) ? br->multicast_last_member_interval :
 460                                           br->multicast_query_response_interval;
 461
 462         mldq->mld_type = ICMPV6_MGM_QUERY;
 463         mldq->mld_code = 0;
 464         mldq->mld_cksum = 0;
 465         mldq->mld_maxdelay = htons((u16)jiffies_to_msecs(interval));
 466         mldq->mld_reserved = 0;
 467         ipv6_addr_copy(&mldq->mld_mca, group);
 468
 469         /* checksum */
 470         mldq->mld_cksum = csum_ipv6_magic(&ip6h->saddr, &ip6h->daddr,
 471                                           sizeof(*mldq), IPPROTO_ICMPV6,
 472                                           csum_partial(mldq,
 473                                                        sizeof(*mldq), 0));
 474         skb_put(skb, sizeof(*mldq));
 475
 476         __skb_pull(skb, sizeof(*eth));
 477
 478 out:
 479         return skb;
 480 }
 481 #endif
 482
 483 static struct sk_buff *br_multicast_alloc_query(struct net_bridge *br,
 484                                                 struct br_ip *addr)
 485 {
 486         switch (addr->proto) {
 487         case htons(ETH_P_IP):
 488                 return br_ip4_multicast_alloc_query(br, addr->u.ip4);
 489 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
 490         case htons(ETH_P_IPV6):
 491                 return br_ip6_multicast_alloc_query(br, &addr->u.ip6);
 492 #endif
 493         }
 494         return NULL;
 495 }
 496
 497 static void br_multicast_send_group_query(struct net_bridge_mdb_entry *mp)
 498 {
 499         struct net_bridge *br = mp->br;
 500         struct sk_buff *skb;
 501
 502         skb = br_multicast_alloc_query(br, &mp->addr);
 503         if (!skb)
 504                 goto timer;
 505
 506         netif_rx(skb);
 507
 508 timer:
 509         if (++mp->queries_sent < br->multicast_last_member_count)
 510                 mod_timer(&mp->query_timer,
 511                           jiffies + br->multicast_last_member_interval);
 512 }
 513
 514 static void br_multicast_group_query_expired(unsigned long data)
 515 {
 516         struct net_bridge_mdb_entry *mp = (void *)data;
 517         struct net_bridge *br = mp->br;
 518
 519         spin_lock(&br->multicast_lock);
 520         if (!netif_running(br->dev) || hlist_unhashed(&mp->mglist) ||
 521             mp->queries_sent >= br->multicast_last_member_count)
 522                 goto out;
 523
 524         br_multicast_send_group_query(mp);
 525
 526 out:
 527         spin_unlock(&br->multicast_lock);
 528 }
 529
 530 static void br_multicast_send_port_group_query(struct net_bridge_port_group *pg)
 531 {
 532         struct net_bridge_port *port = pg->port;
 533         struct net_bridge *br = port->br;
 534         struct sk_buff *skb;
 535
 536         skb = br_multicast_alloc_query(br, &pg->addr);
 537         if (!skb)
 538                 goto timer;
 539
 540         br_deliver(port, skb);
 541
 542 timer:
 543         if (++pg->queries_sent < br->multicast_last_member_count)
 544                 mod_timer(&pg->query_timer,
 545                           jiffies + br->multicast_last_member_interval);
 546 }
 547
 548 static void br_multicast_port_group_query_expired(unsigned long data)
 549 {
 550         struct net_bridge_port_group *pg = (void *)data;
 551         struct net_bridge_port *port = pg->port;
 552         struct net_bridge *br = port->br;
 553
 554         spin_lock(&br->multicast_lock);
 555         if (!netif_running(br->dev) || hlist_unhashed(&pg->mglist) ||
 556             pg->queries_sent >= br->multicast_last_member_count)
 557                 goto out;
 558
 559         br_multicast_send_port_group_query(pg);
 560
 561 out:
 562         spin_unlock(&br->multicast_lock);
 563 }
 564
 565 static struct net_bridge_mdb_entry *br_multicast_get_group(
 566         struct net_bridge *br, struct net_bridge_port *port,
 567         struct br_ip *group, int hash)
 568 {
 569         struct net_bridge_mdb_htable *mdb = br->mdb;
 570         struct net_bridge_mdb_entry *mp;
 571         struct hlist_node *p;
 572         unsigned count = 0;
 573         unsigned max;
 574         int elasticity;
 575         int err;
 576
 577         hlist_for_each_entry(mp, p, &mdb->mhash[hash], hlist[mdb->ver]) {
 578                 count++;
 579                 if (unlikely(br_ip_equal(group, &mp->addr)))
 580                         return mp;
 581         }
 582
 583         elasticity = 0;
 584         max = mdb->max;
 585
 586         if (unlikely(count > br->hash_elasticity && count)) {
 587                 if (net_ratelimit())
 588                         printk(KERN_INFO "%s: Multicast hash table "
 589                                "chain limit reached: %s\n",
 590                                br->dev->name, port ? port->dev->name :
 591                                                      br->dev->name);
 592
 593                 elasticity = br->hash_elasticity;
 594         }
 595
 596         if (mdb->size >= max) {
 597                 max *= 2;
 598                 if (unlikely(max >= br->hash_max)) {
 599                         printk(KERN_WARNING "%s: Multicast hash table maximum "
 600                                "reached, disabling snooping: %s, %d\n",
 601                                br->dev->name, port ? port->dev->name :
 602                                                      br->dev->name,
 603                                max);
 604                         err = -E2BIG;
 605 disable:
 606                         br->multicast_disabled = 1;
 607                         goto err;
 608                 }
 609         }
 610
 611         if (max > mdb->max || elasticity) {
 612                 if (mdb->old) {
 613                         if (net_ratelimit())
 614                                 printk(KERN_INFO "%s: Multicast hash table "
 615                                        "on fire: %s\n",
 616                                        br->dev->name, port ? port->dev->name :
 617                                                              br->dev->name);
 618                         err = -EEXIST;
 619                         goto err;
 620                 }
 621
 622                 err = br_mdb_rehash(&br->mdb, max, elasticity);
 623                 if (err) {
 624                         printk(KERN_WARNING "%s: Cannot rehash multicast "
 625                                "hash table, disabling snooping: "
 626                                "%s, %d, %d\n",
 627                                br->dev->name, port ? port->dev->name :
 628                                                      br->dev->name,
 629                                mdb->size, err);
 630                         goto disable;
 631                 }
 632
 633                 err = -EAGAIN;
 634                 goto err;
 635         }
 636
 637         return NULL;
 638
 639 err:
 640         mp = ERR_PTR(err);
 641         return mp;
 642 }
 643
 644 static struct net_bridge_mdb_entry *br_multicast_new_group(
 645         struct net_bridge *br, struct net_bridge_port *port,
 646         struct br_ip *group)
 647 {
 648         struct net_bridge_mdb_htable *mdb = br->mdb;
 649         struct net_bridge_mdb_entry *mp;
 650         int hash;
 651
 652         if (!mdb) {
 653                 if (br_mdb_rehash(&br->mdb, BR_HASH_SIZE, 0))
 654                         return NULL;
 655                 goto rehash;
 656         }
 657
 658         hash = br_ip_hash(mdb, group);
 659         mp = br_multicast_get_group(br, port, group, hash);
 660         switch (PTR_ERR(mp)) {
 661         case 0:
 662                 break;
 663
 664         case -EAGAIN:
 665 rehash:
 666                 mdb = br->mdb;
 667                 hash = br_ip_hash(mdb, group);
 668                 break;
 669
 670         default:
 671                 goto out;
 672         }
 673
 674         mp = kzalloc(sizeof(*mp), GFP_ATOMIC);
 675         if (unlikely(!mp))
 676                 goto out;
 677
 678         mp->br = br;
 679         mp->addr = *group;
 680         setup_timer(&mp->timer, br_multicast_group_expired,
 681                     (unsigned long)mp);
 682         setup_timer(&mp->query_timer, br_multicast_group_query_expired,
 683                     (unsigned long)mp);
 684
 685         hlist_add_head_rcu(&mp->hlist[mdb->ver], &mdb->mhash[hash]);
 686         mdb->size++;
 687
 688 out:
 689         return mp;
 690 }
 691
 692 static int br_multicast_add_group(struct net_bridge *br,
 693                                   struct net_bridge_port *port,
 694                                   struct br_ip *group)
 695 {
 696         struct net_bridge_mdb_entry *mp;
 697         struct net_bridge_port_group *p;
 698         struct net_bridge_port_group **pp;
 699         unsigned long now = jiffies;
 700         int err;
 701
 702         spin_lock(&br->multicast_lock);
 703         if (!netif_running(br->dev) ||
 704             (port && port->state == BR_STATE_DISABLED))
 705                 goto out;
 706
 707         mp = br_multicast_new_group(br, port, group);
 708         err = PTR_ERR(mp);
 709         if (unlikely(IS_ERR(mp) || !mp))
 710                 goto err;
 711
 712         if (!port) {
 713                 hlist_add_head(&mp->mglist, &br->mglist);
 714                 mod_timer(&mp->timer, now + br->multicast_membership_interval);
 715                 goto out;
 716         }
 717
 718         for (pp = &mp->ports; (p = *pp); pp = &p->next) {
 719                 if (p->port == port)
 720                         goto found;
 721                 if ((unsigned long)p->port < (unsigned long)port)
 722                         break;
 723         }
 724
 725         p = kzalloc(sizeof(*p), GFP_ATOMIC);
 726         err = -ENOMEM;
 727         if (unlikely(!p))
 728                 goto err;
 729
 730         p->addr = *group;
 731         p->port = port;
 732         p->next = *pp;
 733         hlist_add_head(&p->mglist, &port->mglist);
 734         setup_timer(&p->timer, br_multicast_port_group_expired,
 735                     (unsigned long)p);
 736         setup_timer(&p->query_timer, br_multicast_port_group_query_expired,
 737                     (unsigned long)p);
 738
 739         rcu_assign_pointer(*pp, p);
 740
 741 found:
 742         mod_timer(&p->timer, now + br->multicast_membership_interval);
 743 out:
 744         err = 0;
 745
 746 err:
 747         spin_unlock(&br->multicast_lock);
 748         return err;
 749 }
 750
 751 static int br_ip4_multicast_add_group(struct net_bridge *br,
 752                                       struct net_bridge_port *port,
 753                                       __be32 group)
 754 {
 755         struct br_ip br_group;
 756
 757         if (ipv4_is_local_multicast(group))
 758                 return 0;
 759
 760         br_group.u.ip4 = group;
 761         br_group.proto = htons(ETH_P_IP);
 762
 763         return br_multicast_add_group(br, port, &br_group);
 764 }
 765
 766 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
 767 static int br_ip6_multicast_add_group(struct net_bridge *br,
 768                                       struct net_bridge_port *port,
 769                                       const struct in6_addr *group)
 770 {
 771         struct br_ip br_group;
 772
 773         if (ipv6_is_local_multicast(group))
 774                 return 0;
 775
 776         ipv6_addr_copy(&br_group.u.ip6, group);
 777         br_group.proto = htons(ETH_P_IP);
 778
 779         return br_multicast_add_group(br, port, &br_group);
 780 }
 781 #endif
 782
 783 static void br_multicast_router_expired(unsigned long data)
 784 {
 785         struct net_bridge_port *port = (void *)data;
 786         struct net_bridge *br = port->br;
 787
 788         spin_lock(&br->multicast_lock);
 789         if (port->multicast_router != 1 ||
 790             timer_pending(&port->multicast_router_timer) ||
 791             hlist_unhashed(&port->rlist))
 792                 goto out;
 793
 794         hlist_del_init_rcu(&port->rlist);
 795
 796 out:
 797         spin_unlock(&br->multicast_lock);
 798 }
 799
 800 static void br_multicast_local_router_expired(unsigned long data)
 801 {
 802 }
 803
 804 static void __br_multicast_send_query(struct net_bridge *br,
 805                                       struct net_bridge_port *port,
 806                                       struct br_ip *ip)
 807 {
 808         struct sk_buff *skb;
 809
 810         skb = br_multicast_alloc_query(br, ip);
 811         if (!skb)
 812                 return;
 813
 814         if (port) {
 815                 __skb_push(skb, sizeof(struct ethhdr));
 816                 skb->dev = port->dev;
 817                 NF_HOOK(NFPROTO_BRIDGE, NF_BR_LOCAL_OUT, skb, NULL, skb->dev,
 818                         dev_queue_xmit);
 819         } else
 820                 netif_rx(skb);
 821 }
 822
 823 static void br_multicast_send_query(struct net_bridge *br,
 824                                     struct net_bridge_port *port, u32 sent)
 825 {
 826         unsigned long time;
 827         struct br_ip br_group;
 828
 829         if (!netif_running(br->dev) || br->multicast_disabled ||
 830             timer_pending(&br->multicast_querier_timer))
 831                 return;
 832
 833         memset(&br_group.u, 0, sizeof(br_group.u));
 834
 835         br_group.proto = htons(ETH_P_IP);
 836         __br_multicast_send_query(br, port, &br_group);
 837
 838 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
 839         br_group.proto = htons(ETH_P_IPV6);
 840         __br_multicast_send_query(br, port, &br_group);
 841 #endif
 842
 843         time = jiffies;
 844         time += sent < br->multicast_startup_query_count ?
 845                 br->multicast_startup_query_interval :
 846                 br->multicast_query_interval;
 847         mod_timer(port ? &port->multicast_query_timer :
 848                          &br->multicast_query_timer, time);
 849 }
 850
 851 static void br_multicast_port_query_expired(unsigned long data)
 852 {
 853         struct net_bridge_port *port = (void *)data;
 854         struct net_bridge *br = port->br;
 855
 856         spin_lock(&br->multicast_lock);
 857         if (port->state == BR_STATE_DISABLED ||
 858             port->state == BR_STATE_BLOCKING)
 859                 goto out;
 860
 861         if (port->multicast_startup_queries_sent <
 862             br->multicast_startup_query_count)
 863                 port->multicast_startup_queries_sent++;
 864
 865         br_multicast_send_query(port->br, port,
 866                                 port->multicast_startup_queries_sent);
 867
 868 out:
 869         spin_unlock(&br->multicast_lock);
 870 }
 871
 872 void br_multicast_add_port(struct net_bridge_port *port)
 873 {
 874         port->multicast_router = 1;
 875
 876         setup_timer(&port->multicast_router_timer, br_multicast_router_expired,
 877                     (unsigned long)port);
 878         setup_timer(&port->multicast_query_timer,
 879                     br_multicast_port_query_expired, (unsigned long)port);
 880 }
 881
 882 void br_multicast_del_port(struct net_bridge_port *port)
 883 {
 884         del_timer_sync(&port->multicast_router_timer);
 885 }
 886
 887 static void __br_multicast_enable_port(struct net_bridge_port *port)
 888 {
 889         port->multicast_startup_queries_sent = 0;
 890
 891         if (try_to_del_timer_sync(&port->multicast_query_timer) >= 0 ||
 892             del_timer(&port->multicast_query_timer))
 893                 mod_timer(&port->multicast_query_timer, jiffies);
 894 }
 895
 896 void br_multicast_enable_port(struct net_bridge_port *port)
 897 {
 898         struct net_bridge *br = port->br;
 899
 900         spin_lock(&br->multicast_lock);
 901         if (br->multicast_disabled || !netif_running(br->dev))
 902                 goto out;
 903
 904         __br_multicast_enable_port(port);
 905
 906 out:
 907         spin_unlock(&br->multicast_lock);
 908 }
 909
 910 void br_multicast_disable_port(struct net_bridge_port *port)
 911 {
 912         struct net_bridge *br = port->br;
 913         struct net_bridge_port_group *pg;
 914         struct hlist_node *p, *n;
 915
 916         spin_lock(&br->multicast_lock);
 917         hlist_for_each_entry_safe(pg, p, n, &port->mglist, mglist)
 918                 br_multicast_del_pg(br, pg);
 919
 920         if (!hlist_unhashed(&port->rlist))
 921                 hlist_del_init_rcu(&port->rlist);
 922         del_timer(&port->multicast_router_timer);
 923         del_timer(&port->multicast_query_timer);
 924         spin_unlock(&br->multicast_lock);
 925 }
 926
 927 static int br_ip4_multicast_igmp3_report(struct net_bridge *br,
 928                                          struct net_bridge_port *port,
 929                                          struct sk_buff *skb)
 930 {
 931         struct igmpv3_report *ih;
 932         struct igmpv3_grec *grec;
 933         int i;
 934         int len;
 935         int num;
 936         int type;
 937         int err = 0;
 938         __be32 group;
 939
 940         if (!pskb_may_pull(skb, sizeof(*ih)))
 941                 return -EINVAL;
 942
 943         ih = igmpv3_report_hdr(skb);
 944         num = ntohs(ih->ngrec);
 945         len = sizeof(*ih);
 946
 947         for (i = 0; i < num; i++) {
 948                 len += sizeof(*grec);
 949                 if (!pskb_may_pull(skb, len))
 950                         return -EINVAL;
 951
 952                 grec = (void *)(skb->data + len - sizeof(*grec));
 953                 group = grec->grec_mca;
 954                 type = grec->grec_type;
 955
 956                 len += ntohs(grec->grec_nsrcs) * 4;
 957                 if (!pskb_may_pull(skb, len))
 958                         return -EINVAL;
 959
 960                 /* We treat this as an IGMPv2 report for now. */
 961                 switch (type) {
 962                 case IGMPV3_MODE_IS_INCLUDE:
 963                 case IGMPV3_MODE_IS_EXCLUDE:
 964                 case IGMPV3_CHANGE_TO_INCLUDE:
 965                 case IGMPV3_CHANGE_TO_EXCLUDE:
 966                 case IGMPV3_ALLOW_NEW_SOURCES:
 967                 case IGMPV3_BLOCK_OLD_SOURCES:
 968                         break;
 969
 970                 default:
 971                         continue;
 972                 }
 973
 974                 err = br_ip4_multicast_add_group(br, port, group);
 975                 if (err)
 976                         break;
 977         }
 978
 979         return err;
 980 }
 981
 982 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
 983 static int br_ip6_multicast_mld2_report(struct net_bridge *br,
 984                                         struct net_bridge_port *port,
 985                                         struct sk_buff *skb)
 986 {
 987         struct icmp6hdr *icmp6h;
 988         struct mld2_grec *grec;
 989         int i;
 990         int len;
 991         int num;
 992         int err = 0;
 993
 994         if (!pskb_may_pull(skb, sizeof(*icmp6h)))
 995                 return -EINVAL;
 996
 997         icmp6h = icmp6_hdr(skb);
 998         num = ntohs(icmp6h->icmp6_dataun.un_data16[1]);
 999         len = sizeof(*icmp6h);
1000
1001         for (i = 0; i < num; i++) {
1002                 __be16 *nsrcs, _nsrcs;
1003
1004                 nsrcs = skb_header_pointer(skb,
1005                                            len + offsetof(struct mld2_grec,
1006                                                           grec_mca),
1007                                            sizeof(_nsrcs), &_nsrcs);
1008                 if (!nsrcs)
1009                         return -EINVAL;
1010
1011                 if (!pskb_may_pull(skb,
1012                                    len + sizeof(*grec) +
1013                                    sizeof(struct in6_addr) * (*nsrcs)))
1014                         return -EINVAL;
1015
1016                 grec = (struct mld2_grec *)(skb->data + len);
1017                 len += sizeof(*grec) + sizeof(struct in6_addr) * (*nsrcs);
1018
1019                 /* We treat these as MLDv1 reports for now. */
1020                 switch (grec->grec_type) {
1021                 case MLD2_MODE_IS_INCLUDE:
1022                 case MLD2_MODE_IS_EXCLUDE:
1023                 case MLD2_CHANGE_TO_INCLUDE:
1024                 case MLD2_CHANGE_TO_EXCLUDE:
1025                 case MLD2_ALLOW_NEW_SOURCES:
1026                 case MLD2_BLOCK_OLD_SOURCES:
1027                         break;
1028
1029                 default:
1030                         continue;
1031                 }
1032
1033                 err = br_ip6_multicast_add_group(br, port, &grec->grec_mca);
1034                 if (!err)
1035                         break;
1036         }
1037
1038         return err;
1039 }
1040 #endif
1041
1042 /*
1043  * Add port to rotuer_list
1044  *  list is maintained ordered by pointer value
1045  *  and locked by br->multicast_lock and RCU
1046  */
1047 static void br_multicast_add_router(struct net_bridge *br,
1048                                     struct net_bridge_port *port)
1049 {
1050         struct net_bridge_port *p;
1051         struct hlist_node *n, *slot = NULL;
1052
1053         hlist_for_each_entry(p, n, &br->router_list, rlist) {
1054                 if ((unsigned long) port >= (unsigned long) p)
1055                         break;
1056                 slot = n;
1057         }
1058
1059         if (slot)
1060                 hlist_add_after_rcu(slot, &port->rlist);
1061         else
1062                 hlist_add_head_rcu(&port->rlist, &br->router_list);
1063 }
1064
1065 static void br_multicast_mark_router(struct net_bridge *br,
1066                                      struct net_bridge_port *port)
1067 {
1068         unsigned long now = jiffies;
1069
1070         if (!port) {
1071                 if (br->multicast_router == 1)
1072                         mod_timer(&br->multicast_router_timer,
1073                                   now + br->multicast_querier_interval);
1074                 return;
1075         }
1076
1077         if (port->multicast_router != 1)
1078                 return;
1079
1080         if (!hlist_unhashed(&port->rlist))
1081                 goto timer;
1082
1083         br_multicast_add_router(br, port);
1084
1085 timer:
1086         mod_timer(&port->multicast_router_timer,
1087                   now + br->multicast_querier_interval);
1088 }
1089
1090 static void br_multicast_query_received(struct net_bridge *br,
1091                                         struct net_bridge_port *port,
1092                                         int saddr)
1093 {
1094         if (saddr)
1095                 mod_timer(&br->multicast_querier_timer,
1096                           jiffies + br->multicast_querier_interval);
1097         else if (timer_pending(&br->multicast_querier_timer))
1098                 return;
1099
1100         br_multicast_mark_router(br, port);
1101 }
1102
1103 static int br_ip4_multicast_query(struct net_bridge *br,
1104                                   struct net_bridge_port *port,
1105                                   struct sk_buff *skb)
1106 {
1107         struct iphdr *iph = ip_hdr(skb);
1108         struct igmphdr *ih = igmp_hdr(skb);
1109         struct net_bridge_mdb_entry *mp;
1110         struct igmpv3_query *ih3;
1111         struct net_bridge_port_group *p;
1112         struct net_bridge_port_group **pp;
1113         unsigned long max_delay;
1114         unsigned long now = jiffies;
1115         __be32 group;
1116         int err = 0;
1117
1118         spin_lock(&br->multicast_lock);
1119         if (!netif_running(br->dev) ||
1120             (port && port->state == BR_STATE_DISABLED))
1121                 goto out;
1122
1123         br_multicast_query_received(br, port, !!iph->saddr);
1124
1125         group = ih->group;
1126
1127         if (skb->len == sizeof(*ih)) {
1128                 max_delay = ih->code * (HZ / IGMP_TIMER_SCALE);
1129
1130                 if (!max_delay) {
1131                         max_delay = 10 * HZ;
1132                         group = 0;
1133                 }
1134         } else {
1135                 if (!pskb_may_pull(skb, sizeof(struct igmpv3_query))) {
1136                         err = -EINVAL;
1137                         goto out;
1138                 }
1139
1140                 ih3 = igmpv3_query_hdr(skb);
1141                 if (ih3->nsrcs)
1142                         goto out;
1143
1144                 max_delay = ih3->code ?
1145                             IGMPV3_MRC(ih3->code) * (HZ / IGMP_TIMER_SCALE) : 1;
1146         }
1147
1148         if (!group)
1149                 goto out;
1150
1151         mp = br_mdb_ip4_get(br->mdb, group);
1152         if (!mp)
1153                 goto out;
1154
1155         max_delay *= br->multicast_last_member_count;
1156
1157         if (!hlist_unhashed(&mp->mglist) &&
1158             (timer_pending(&mp->timer) ?
1159              time_after(mp->timer.expires, now + max_delay) :
1160              try_to_del_timer_sync(&mp->timer) >= 0))
1161                 mod_timer(&mp->timer, now + max_delay);
1162
1163         for (pp = &mp->ports; (p = *pp); pp = &p->next) {
1164                 if (timer_pending(&p->timer) ?
1165                     time_after(p->timer.expires, now + max_delay) :
1166                     try_to_del_timer_sync(&p->timer) >= 0)
1167                         mod_timer(&mp->timer, now + max_delay);
1168         }
1169
1170 out:
1171         spin_unlock(&br->multicast_lock);
1172         return err;
1173 }
1174
1175 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
1176 static int br_ip6_multicast_query(struct net_bridge *br,
1177                                   struct net_bridge_port *port,
1178                                   struct sk_buff *skb)
1179 {
1180         struct ipv6hdr *ip6h = ipv6_hdr(skb);
1181         struct mld_msg *mld = (struct mld_msg *) icmp6_hdr(skb);
1182         struct net_bridge_mdb_entry *mp;
1183         struct mld2_query *mld2q;
1184         struct net_bridge_port_group *p, **pp;
1185         unsigned long max_delay;
1186         unsigned long now = jiffies;
1187         struct in6_addr *group = NULL;
1188         int err = 0;
1189
1190         spin_lock(&br->multicast_lock);
1191         if (!netif_running(br->dev) ||
1192             (port && port->state == BR_STATE_DISABLED))
1193                 goto out;
1194
1195         br_multicast_query_received(br, port, !ipv6_addr_any(&ip6h->saddr));
1196
1197         if (skb->len == sizeof(*mld)) {
1198                 if (!pskb_may_pull(skb, sizeof(*mld))) {
1199                         err = -EINVAL;
1200                         goto out;
1201                 }
1202                 mld = (struct mld_msg *) icmp6_hdr(skb);
1203                 max_delay = msecs_to_jiffies(htons(mld->mld_maxdelay));
1204                 if (max_delay)
1205                         group = &mld->mld_mca;
1206         } else if (skb->len >= sizeof(*mld2q)) {
1207                 if (!pskb_may_pull(skb, sizeof(*mld2q))) {
1208                         err = -EINVAL;
1209                         goto out;
1210                 }
1211                 mld2q = (struct mld2_query *)icmp6_hdr(skb);
1212                 if (!mld2q->mld2q_nsrcs)
1213                         group = &mld2q->mld2q_mca;
1214                 max_delay = mld2q->mld2q_mrc ? MLDV2_MRC(mld2q->mld2q_mrc) : 1;
1215         }
1216
1217         if (!group)
1218                 goto out;
1219
1220         mp = br_mdb_ip6_get(br->mdb, group);
1221         if (!mp)
1222                 goto out;
1223
1224         max_delay *= br->multicast_last_member_count;
1225         if (!hlist_unhashed(&mp->mglist) &&
1226             (timer_pending(&mp->timer) ?
1227              time_after(mp->timer.expires, now + max_delay) :
1228              try_to_del_timer_sync(&mp->timer) >= 0))
1229                 mod_timer(&mp->timer, now + max_delay);
1230
1231         for (pp = &mp->ports; (p = *pp); pp = &p->next) {
1232                 if (timer_pending(&p->timer) ?
1233                     time_after(p->timer.expires, now + max_delay) :
1234                     try_to_del_timer_sync(&p->timer) >= 0)
1235                         mod_timer(&mp->timer, now + max_delay);
1236         }
1237
1238 out:
1239         spin_unlock(&br->multicast_lock);
1240         return err;
1241 }
1242 #endif
1243
1244 static void br_multicast_leave_group(struct net_bridge *br,
1245                                      struct net_bridge_port *port,
1246                                      struct br_ip *group)
1247 {
1248         struct net_bridge_mdb_htable *mdb;
1249         struct net_bridge_mdb_entry *mp;
1250         struct net_bridge_port_group *p;
1251         unsigned long now;
1252         unsigned long time;
1253
1254         spin_lock(&br->multicast_lock);
1255         if (!netif_running(br->dev) ||
1256             (port && port->state == BR_STATE_DISABLED) ||
1257             timer_pending(&br->multicast_querier_timer))
1258                 goto out;
1259
1260         mdb = br->mdb;
1261         mp = br_mdb_ip_get(mdb, group);
1262         if (!mp)
1263                 goto out;
1264
1265         now = jiffies;
1266         time = now + br->multicast_last_member_count *
1267                      br->multicast_last_member_interval;
1268
1269         if (!port) {
1270                 if (!hlist_unhashed(&mp->mglist) &&
1271                     (timer_pending(&mp->timer) ?
1272                      time_after(mp->timer.expires, time) :
1273                      try_to_del_timer_sync(&mp->timer) >= 0)) {
1274                         mod_timer(&mp->timer, time);
1275
1276                         mp->queries_sent = 0;
1277                         mod_timer(&mp->query_timer, now);
1278                 }
1279
1280                 goto out;
1281         }
1282
1283         for (p = mp->ports; p; p = p->next) {
1284                 if (p->port != port)
1285                         continue;
1286
1287                 if (!hlist_unhashed(&p->mglist) &&
1288                     (timer_pending(&p->timer) ?
1289                      time_after(p->timer.expires, time) :
1290                      try_to_del_timer_sync(&p->timer) >= 0)) {
1291                         mod_timer(&p->timer, time);
1292
1293                         p->queries_sent = 0;
1294                         mod_timer(&p->query_timer, now);
1295                 }
1296
1297                 break;
1298         }
1299
1300 out:
1301         spin_unlock(&br->multicast_lock);
1302 }
1303
1304 static void br_ip4_multicast_leave_group(struct net_bridge *br,
1305                                          struct net_bridge_port *port,
1306                                          __be32 group)
1307 {
1308         struct br_ip br_group;
1309
1310         if (ipv4_is_local_multicast(group))
1311                 return;
1312
1313         br_group.u.ip4 = group;
1314         br_group.proto = htons(ETH_P_IP);
1315
1316         br_multicast_leave_group(br, port, &br_group);
1317 }
1318
1319 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
1320 static void br_ip6_multicast_leave_group(struct net_bridge *br,
1321                                          struct net_bridge_port *port,
1322                                          const struct in6_addr *group)
1323 {
1324         struct br_ip br_group;
1325
1326         if (ipv6_is_local_multicast(group))
1327                 return;
1328
1329         ipv6_addr_copy(&br_group.u.ip6, group);
1330         br_group.proto = htons(ETH_P_IPV6);
1331
1332         br_multicast_leave_group(br, port, &br_group);
1333 }
1334 #endif
1335
1336 static int br_multicast_ipv4_rcv(struct net_bridge *br,
1337                                  struct net_bridge_port *port,
1338                                  struct sk_buff *skb)
1339 {
1340         struct sk_buff *skb2 = skb;
1341         struct iphdr *iph;
1342         struct igmphdr *ih;
1343         unsigned len;
1344         unsigned offset;
1345         int err;
1346
1347         /* We treat OOM as packet loss for now. */
1348         if (!pskb_may_pull(skb, sizeof(*iph)))
1349                 return -EINVAL;
1350
1351         iph = ip_hdr(skb);
1352
1353         if (iph->ihl < 5 || iph->version != 4)
1354                 return -EINVAL;
1355
1356         if (!pskb_may_pull(skb, ip_hdrlen(skb)))
1357                 return -EINVAL;
1358
1359         iph = ip_hdr(skb);
1360
1361         if (unlikely(ip_fast_csum((u8 *)iph, iph->ihl)))
1362                 return -EINVAL;
1363
1364         if (iph->protocol != IPPROTO_IGMP)
1365                 return 0;
1366
1367         len = ntohs(iph->tot_len);
1368         if (skb->len < len || len < ip_hdrlen(skb))
1369                 return -EINVAL;
1370
1371         if (skb->len > len) {
1372                 skb2 = skb_clone(skb, GFP_ATOMIC);
1373                 if (!skb2)
1374                         return -ENOMEM;
1375
1376                 err = pskb_trim_rcsum(skb2, len);
1377                 if (err)
1378                         goto err_out;
1379         }
1380
1381         len -= ip_hdrlen(skb2);
1382         offset = skb_network_offset(skb2) + ip_hdrlen(skb2);
1383         __skb_pull(skb2, offset);
1384         skb_reset_transport_header(skb2);
1385
1386         err = -EINVAL;
1387         if (!pskb_may_pull(skb2, sizeof(*ih)))
1388                 goto out;
1389
1390         switch (skb2->ip_summed) {
1391         case CHECKSUM_COMPLETE:
1392                 if (!csum_fold(skb2->csum))
1393                         break;
1394                 /* fall through */
1395         case CHECKSUM_NONE:
1396                 skb2->csum = 0;
1397                 if (skb_checksum_complete(skb2))
1398                         goto out;
1399         }
1400
1401         err = 0;
1402
1403         BR_INPUT_SKB_CB(skb)->igmp = 1;
1404         ih = igmp_hdr(skb2);
1405
1406         switch (ih->type) {
1407         case IGMP_HOST_MEMBERSHIP_REPORT:
1408         case IGMPV2_HOST_MEMBERSHIP_REPORT:
1409                 BR_INPUT_SKB_CB(skb2)->mrouters_only = 1;
1410                 err = br_ip4_multicast_add_group(br, port, ih->group);
1411                 break;
1412         case IGMPV3_HOST_MEMBERSHIP_REPORT:
1413                 err = br_ip4_multicast_igmp3_report(br, port, skb2);
1414                 break;
1415         case IGMP_HOST_MEMBERSHIP_QUERY:
1416                 err = br_ip4_multicast_query(br, port, skb2);
1417                 break;
1418         case IGMP_HOST_LEAVE_MESSAGE:
1419                 br_ip4_multicast_leave_group(br, port, ih->group);
1420                 break;
1421         }
1422
1423 out:
1424         __skb_push(skb2, offset);
1425 err_out:
1426         if (skb2 != skb)
1427                 kfree_skb(skb2);
1428         return err;
1429 }
1430
1431 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
1432 static int br_multicast_ipv6_rcv(struct net_bridge *br,
1433                                  struct net_bridge_port *port,
1434                                  struct sk_buff *skb)
1435 {
1436         struct sk_buff *skb2 = skb;
1437         struct ipv6hdr *ip6h;
1438         struct icmp6hdr *icmp6h;
1439         u8 nexthdr;
1440         unsigned len;
1441         unsigned offset;
1442         int err;
1443
1444         if (!pskb_may_pull(skb, sizeof(*ip6h)))
1445                 return -EINVAL;
1446
1447         ip6h = ipv6_hdr(skb);
1448
1449         /*
1450          * We're interested in MLD messages only.
1451          *  - Version is 6
1452          *  - MLD has always Router Alert hop-by-hop option
1453          *  - But we do not support jumbrograms.
1454          */
1455         if (ip6h->version != 6 ||
1456             ip6h->nexthdr != IPPROTO_HOPOPTS ||
1457             ip6h->payload_len == 0)
1458                 return 0;
1459
1460         len = ntohs(ip6h->payload_len);
1461         if (skb->len < len)
1462                 return -EINVAL;
1463
1464         nexthdr = ip6h->nexthdr;
1465         offset = ipv6_skip_exthdr(skb, sizeof(*ip6h), &nexthdr);
1466
1467         if (offset < 0 || nexthdr != IPPROTO_ICMPV6)
1468                 return 0;
1469
1470         /* Okay, we found ICMPv6 header */
1471         skb2 = skb_clone(skb, GFP_ATOMIC);
1472         if (!skb2)
1473                 return -ENOMEM;
1474
1475         len -= offset - skb_network_offset(skb2);
1476
1477         __skb_pull(skb2, offset);
1478         skb_reset_transport_header(skb2);
1479
1480         err = -EINVAL;
1481         if (!pskb_may_pull(skb2, sizeof(*icmp6h)))
1482                 goto out;
1483
1484         icmp6h = icmp6_hdr(skb2);
1485
1486         switch (icmp6h->icmp6_type) {
1487         case ICMPV6_MGM_QUERY:
1488         case ICMPV6_MGM_REPORT:
1489         case ICMPV6_MGM_REDUCTION:
1490         case ICMPV6_MLD2_REPORT:
1491                 break;
1492         default:
1493                 err = 0;
1494                 goto out;
1495         }
1496
1497         /* Okay, we found MLD message. Check further. */
1498         if (skb2->len > len) {
1499                 err = pskb_trim_rcsum(skb2, len);
1500                 if (err)
1501                         goto out;
1502         }
1503
1504         switch (skb2->ip_summed) {
1505         case CHECKSUM_COMPLETE:
1506                 if (!csum_fold(skb2->csum))
1507                         break;
1508                 /*FALLTHROUGH*/
1509         case CHECKSUM_NONE:
1510                 skb2->csum = 0;
1511                 if (skb_checksum_complete(skb2))
1512                         goto out;
1513         }
1514
1515         err = 0;
1516
1517         BR_INPUT_SKB_CB(skb)->igmp = 1;
1518
1519         switch (icmp6h->icmp6_type) {
1520         case ICMPV6_MGM_REPORT:
1521             {
1522                 struct mld_msg *mld = (struct mld_msg *)icmp6h;
1523                 BR_INPUT_SKB_CB(skb2)->mrouters_only = 1;
1524                 err = br_ip6_multicast_add_group(br, port, &mld->mld_mca);
1525                 break;
1526             }
1527         case ICMPV6_MLD2_REPORT:
1528                 err = br_ip6_multicast_mld2_report(br, port, skb2);
1529                 break;
1530         case ICMPV6_MGM_QUERY:
1531                 err = br_ip6_multicast_query(br, port, skb2);
1532                 break;
1533         case ICMPV6_MGM_REDUCTION:
1534             {
1535                 struct mld_msg *mld = (struct mld_msg *)icmp6h;
1536                 br_ip6_multicast_leave_group(br, port, &mld->mld_mca);
1537             }
1538         }
1539
1540 out:
1541         __skb_push(skb2, offset);
1542         if (skb2 != skb)
1543                 kfree_skb(skb2);
1544         return err;
1545 }
1546 #endif
1547
1548 int br_multicast_rcv(struct net_bridge *br, struct net_bridge_port *port,
1549                      struct sk_buff *skb)
1550 {
1551         BR_INPUT_SKB_CB(skb)->igmp = 0;
1552         BR_INPUT_SKB_CB(skb)->mrouters_only = 0;
1553
1554         if (br->multicast_disabled)
1555                 return 0;
1556
1557         switch (skb->protocol) {
1558         case htons(ETH_P_IP):
1559                 return br_multicast_ipv4_rcv(br, port, skb);
1560 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
1561         case htons(ETH_P_IPV6):
1562                 return br_multicast_ipv6_rcv(br, port, skb);
1563 #endif
1564         }
1565
1566         return 0;
1567 }
1568
1569 static void br_multicast_query_expired(unsigned long data)
1570 {
1571         struct net_bridge *br = (void *)data;
1572
1573         spin_lock(&br->multicast_lock);
1574         if (br->multicast_startup_queries_sent <
1575             br->multicast_startup_query_count)
1576                 br->multicast_startup_queries_sent++;
1577
1578         br_multicast_send_query(br, NULL, br->multicast_startup_queries_sent);
1579
1580         spin_unlock(&br->multicast_lock);
1581 }
1582
1583 void br_multicast_init(struct net_bridge *br)
1584 {
1585         br->hash_elasticity = 4;
1586         br->hash_max = 512;
1587
1588         br->multicast_router = 1;
1589         br->multicast_last_member_count = 2;
1590         br->multicast_startup_query_count = 2;
1591
1592         br->multicast_last_member_interval = HZ;
1593         br->multicast_query_response_interval = 10 * HZ;
1594         br->multicast_startup_query_interval = 125 * HZ / 4;
1595         br->multicast_query_interval = 125 * HZ;
1596         br->multicast_querier_interval = 255 * HZ;
1597         br->multicast_membership_interval = 260 * HZ;
1598
1599         spin_lock_init(&br->multicast_lock);
1600         setup_timer(&br->multicast_router_timer,
1601                     br_multicast_local_router_expired, 0);
1602         setup_timer(&br->multicast_querier_timer,
1603                     br_multicast_local_router_expired, 0);
1604         setup_timer(&br->multicast_query_timer, br_multicast_query_expired,
1605                     (unsigned long)br);
1606 }
1607
1608 void br_multicast_open(struct net_bridge *br)
1609 {
1610         br->multicast_startup_queries_sent = 0;
1611
1612         if (br->multicast_disabled)
1613                 return;
1614
1615         mod_timer(&br->multicast_query_timer, jiffies);
1616 }
1617
1618 void br_multicast_stop(struct net_bridge *br)
1619 {
1620         struct net_bridge_mdb_htable *mdb;
1621         struct net_bridge_mdb_entry *mp;
1622         struct hlist_node *p, *n;
1623         u32 ver;
1624         int i;
1625
1626         del_timer_sync(&br->multicast_router_timer);
1627         del_timer_sync(&br->multicast_querier_timer);
1628         del_timer_sync(&br->multicast_query_timer);
1629
1630         spin_lock_bh(&br->multicast_lock);
1631         mdb = br->mdb;
1632         if (!mdb)
1633                 goto out;
1634
1635         br->mdb = NULL;
1636
1637         ver = mdb->ver;
1638         for (i = 0; i < mdb->max; i++) {
1639                 hlist_for_each_entry_safe(mp, p, n, &mdb->mhash[i],
1640                                           hlist[ver]) {
1641                         del_timer(&mp->timer);
1642                         del_timer(&mp->query_timer);
1643                         call_rcu_bh(&mp->rcu, br_multicast_free_group);
1644                 }
1645         }
1646
1647         if (mdb->old) {
1648                 spin_unlock_bh(&br->multicast_lock);
1649                 rcu_barrier_bh();
1650                 spin_lock_bh(&br->multicast_lock);
1651                 WARN_ON(mdb->old);
1652         }
1653
1654         mdb->old = mdb;
1655         call_rcu_bh(&mdb->rcu, br_mdb_free);
1656
1657 out:
1658         spin_unlock_bh(&br->multicast_lock);
1659 }
1660
1661 int br_multicast_set_router(struct net_bridge *br, unsigned long val)
1662 {
1663         int err = -ENOENT;
1664
1665         spin_lock_bh(&br->multicast_lock);
1666         if (!netif_running(br->dev))
1667                 goto unlock;
1668
1669         switch (val) {
1670         case 0:
1671         case 2:
1672                 del_timer(&br->multicast_router_timer);
1673                 /* fall through */
1674         case 1:
1675                 br->multicast_router = val;
1676                 err = 0;
1677                 break;
1678
1679         default:
1680                 err = -EINVAL;
1681                 break;
1682         }
1683
1684 unlock:
1685         spin_unlock_bh(&br->multicast_lock);
1686
1687         return err;
1688 }
1689
1690 int br_multicast_set_port_router(struct net_bridge_port *p, unsigned long val)
1691 {
1692         struct net_bridge *br = p->br;
1693         int err = -ENOENT;
1694
1695         spin_lock(&br->multicast_lock);
1696         if (!netif_running(br->dev) || p->state == BR_STATE_DISABLED)
1697                 goto unlock;
1698
1699         switch (val) {
1700         case 0:
1701         case 1:
1702         case 2:
1703                 p->multicast_router = val;
1704                 err = 0;
1705
1706                 if (val < 2 && !hlist_unhashed(&p->rlist))
1707                         hlist_del_init_rcu(&p->rlist);
1708
1709                 if (val == 1)
1710                         break;
1711
1712                 del_timer(&p->multicast_router_timer);
1713
1714                 if (val == 0)
1715                         break;
1716
1717                 br_multicast_add_router(br, p);
1718                 break;
1719
1720         default:
1721                 err = -EINVAL;
1722                 break;
1723         }
1724
1725 unlock:
1726         spin_unlock(&br->multicast_lock);
1727
1728         return err;
1729 }
1730
1731 int br_multicast_toggle(struct net_bridge *br, unsigned long val)
1732 {
1733         struct net_bridge_port *port;
1734         int err = -ENOENT;
1735
1736         spin_lock(&br->multicast_lock);
1737         if (!netif_running(br->dev))
1738                 goto unlock;
1739
1740         err = 0;
1741         if (br->multicast_disabled == !val)
1742                 goto unlock;
1743
1744         br->multicast_disabled = !val;
1745         if (br->multicast_disabled)
1746                 goto unlock;
1747
1748         if (br->mdb) {
1749                 if (br->mdb->old) {
1750                         err = -EEXIST;
1751 rollback:
1752                         br->multicast_disabled = !!val;
1753                         goto unlock;
1754                 }
1755
1756                 err = br_mdb_rehash(&br->mdb, br->mdb->max,
1757                                     br->hash_elasticity);
1758                 if (err)
1759                         goto rollback;
1760         }
1761
1762         br_multicast_open(br);
1763         list_for_each_entry(port, &br->port_list, list) {
1764                 if (port->state == BR_STATE_DISABLED ||
1765                     port->state == BR_STATE_BLOCKING)
1766                         continue;
1767
1768                 __br_multicast_enable_port(port);
1769         }
1770
1771 unlock:
1772         spin_unlock(&br->multicast_lock);
1773
1774         return err;
1775 }
1776
1777 int br_multicast_set_hash_max(struct net_bridge *br, unsigned long val)
1778 {
1779         int err = -ENOENT;
1780         u32 old;
1781
1782         spin_lock(&br->multicast_lock);
1783         if (!netif_running(br->dev))
1784                 goto unlock;
1785
1786         err = -EINVAL;
1787         if (!is_power_of_2(val))
1788                 goto unlock;
1789         if (br->mdb && val < br->mdb->size)
1790                 goto unlock;
1791
1792         err = 0;
1793
1794         old = br->hash_max;
1795         br->hash_max = val;
1796
1797         if (br->mdb) {
1798                 if (br->mdb->old) {
1799                         err = -EEXIST;
1800 rollback:
1801                         br->hash_max = old;
1802                         goto unlock;
1803                 }
1804
1805                 err = br_mdb_rehash(&br->mdb, br->hash_max,
1806                                     br->hash_elasticity);
1807                 if (err)
1808                         goto rollback;
1809         }
1810
1811 unlock:
1812         spin_unlock(&br->multicast_lock);
1813
1814         return err;
1815 }