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                         br_info(br, "Multicast hash table "
 589                                 "chain limit reached: %s\n",
 590                                 port ? port->dev->name : br->dev->name);
 591
 592                 elasticity = br->hash_elasticity;
 593         }
 594
 595         if (mdb->size >= max) {
 596                 max *= 2;
 597                 if (unlikely(max >= br->hash_max)) {
 598                         br_warn(br, "Multicast hash table maximum "
 599                                 "reached, disabling snooping: %s, %d\n",
 600                                 port ? port->dev->name : br->dev->name, max);
 601                         err = -E2BIG;
 602 disable:
 603                         br->multicast_disabled = 1;
 604                         goto err;
 605                 }
 606         }
 607
 608         if (max > mdb->max || elasticity) {
 609                 if (mdb->old) {
 610                         if (net_ratelimit())
 611                                 br_info(br, "Multicast hash table "
 612                                         "on fire: %s\n",
 613                                         port ? port->dev->name : br->dev->name);
 614                         err = -EEXIST;
 615                         goto err;
 616                 }
 617
 618                 err = br_mdb_rehash(&br->mdb, max, elasticity);
 619                 if (err) {
 620                         br_warn(br, "Cannot rehash multicast "
 621                                 "hash table, disabling snooping: %s, %d, %d\n",
 622                                 port ? port->dev->name : br->dev->name,
 623                                 mdb->size, err);
 624                         goto disable;
 625                 }
 626
 627                 err = -EAGAIN;
 628                 goto err;
 629         }
 630
 631         return NULL;
 632
 633 err:
 634         mp = ERR_PTR(err);
 635         return mp;
 636 }
 637
 638 static struct net_bridge_mdb_entry *br_multicast_new_group(
 639         struct net_bridge *br, struct net_bridge_port *port,
 640         struct br_ip *group)
 641 {
 642         struct net_bridge_mdb_htable *mdb = br->mdb;
 643         struct net_bridge_mdb_entry *mp;
 644         int hash;
 645
 646         if (!mdb) {
 647                 if (br_mdb_rehash(&br->mdb, BR_HASH_SIZE, 0))
 648                         return NULL;
 649                 goto rehash;
 650         }
 651
 652         hash = br_ip_hash(mdb, group);
 653         mp = br_multicast_get_group(br, port, group, hash);
 654         switch (PTR_ERR(mp)) {
 655         case 0:
 656                 break;
 657
 658         case -EAGAIN:
 659 rehash:
 660                 mdb = br->mdb;
 661                 hash = br_ip_hash(mdb, group);
 662                 break;
 663
 664         default:
 665                 goto out;
 666         }
 667
 668         mp = kzalloc(sizeof(*mp), GFP_ATOMIC);
 669         if (unlikely(!mp))
 670                 goto out;
 671
 672         mp->br = br;
 673         mp->addr = *group;
 674         setup_timer(&mp->timer, br_multicast_group_expired,
 675                     (unsigned long)mp);
 676         setup_timer(&mp->query_timer, br_multicast_group_query_expired,
 677                     (unsigned long)mp);
 678
 679         hlist_add_head_rcu(&mp->hlist[mdb->ver], &mdb->mhash[hash]);
 680         mdb->size++;
 681
 682 out:
 683         return mp;
 684 }
 685
 686 static int br_multicast_add_group(struct net_bridge *br,
 687                                   struct net_bridge_port *port,
 688                                   struct br_ip *group)
 689 {
 690         struct net_bridge_mdb_entry *mp;
 691         struct net_bridge_port_group *p;
 692         struct net_bridge_port_group **pp;
 693         unsigned long now = jiffies;
 694         int err;
 695
 696         spin_lock(&br->multicast_lock);
 697         if (!netif_running(br->dev) ||
 698             (port && port->state == BR_STATE_DISABLED))
 699                 goto out;
 700
 701         mp = br_multicast_new_group(br, port, group);
 702         err = PTR_ERR(mp);
 703         if (unlikely(IS_ERR(mp) || !mp))
 704                 goto err;
 705
 706         if (!port) {
 707                 hlist_add_head(&mp->mglist, &br->mglist);
 708                 mod_timer(&mp->timer, now + br->multicast_membership_interval);
 709                 goto out;
 710         }
 711
 712         for (pp = &mp->ports; (p = *pp); pp = &p->next) {
 713                 if (p->port == port)
 714                         goto found;
 715                 if ((unsigned long)p->port < (unsigned long)port)
 716                         break;
 717         }
 718
 719         p = kzalloc(sizeof(*p), GFP_ATOMIC);
 720         err = -ENOMEM;
 721         if (unlikely(!p))
 722                 goto err;
 723
 724         p->addr = *group;
 725         p->port = port;
 726         p->next = *pp;
 727         hlist_add_head(&p->mglist, &port->mglist);
 728         setup_timer(&p->timer, br_multicast_port_group_expired,
 729                     (unsigned long)p);
 730         setup_timer(&p->query_timer, br_multicast_port_group_query_expired,
 731                     (unsigned long)p);
 732
 733         rcu_assign_pointer(*pp, p);
 734
 735 found:
 736         mod_timer(&p->timer, now + br->multicast_membership_interval);
 737 out:
 738         err = 0;
 739
 740 err:
 741         spin_unlock(&br->multicast_lock);
 742         return err;
 743 }
 744
 745 static int br_ip4_multicast_add_group(struct net_bridge *br,
 746                                       struct net_bridge_port *port,
 747                                       __be32 group)
 748 {
 749         struct br_ip br_group;
 750
 751         if (ipv4_is_local_multicast(group))
 752                 return 0;
 753
 754         br_group.u.ip4 = group;
 755         br_group.proto = htons(ETH_P_IP);
 756
 757         return br_multicast_add_group(br, port, &br_group);
 758 }
 759
 760 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
 761 static int br_ip6_multicast_add_group(struct net_bridge *br,
 762                                       struct net_bridge_port *port,
 763                                       const struct in6_addr *group)
 764 {
 765         struct br_ip br_group;
 766
 767         if (ipv6_is_local_multicast(group))
 768                 return 0;
 769
 770         ipv6_addr_copy(&br_group.u.ip6, group);
 771         br_group.proto = htons(ETH_P_IP);
 772
 773         return br_multicast_add_group(br, port, &br_group);
 774 }
 775 #endif
 776
 777 static void br_multicast_router_expired(unsigned long data)
 778 {
 779         struct net_bridge_port *port = (void *)data;
 780         struct net_bridge *br = port->br;
 781
 782         spin_lock(&br->multicast_lock);
 783         if (port->multicast_router != 1 ||
 784             timer_pending(&port->multicast_router_timer) ||
 785             hlist_unhashed(&port->rlist))
 786                 goto out;
 787
 788         hlist_del_init_rcu(&port->rlist);
 789
 790 out:
 791         spin_unlock(&br->multicast_lock);
 792 }
 793
 794 static void br_multicast_local_router_expired(unsigned long data)
 795 {
 796 }
 797
 798 static void __br_multicast_send_query(struct net_bridge *br,
 799                                       struct net_bridge_port *port,
 800                                       struct br_ip *ip)
 801 {
 802         struct sk_buff *skb;
 803
 804         skb = br_multicast_alloc_query(br, ip);
 805         if (!skb)
 806                 return;
 807
 808         if (port) {
 809                 __skb_push(skb, sizeof(struct ethhdr));
 810                 skb->dev = port->dev;
 811                 NF_HOOK(NFPROTO_BRIDGE, NF_BR_LOCAL_OUT, skb, NULL, skb->dev,
 812                         dev_queue_xmit);
 813         } else
 814                 netif_rx(skb);
 815 }
 816
 817 static void br_multicast_send_query(struct net_bridge *br,
 818                                     struct net_bridge_port *port, u32 sent)
 819 {
 820         unsigned long time;
 821         struct br_ip br_group;
 822
 823         if (!netif_running(br->dev) || br->multicast_disabled ||
 824             timer_pending(&br->multicast_querier_timer))
 825                 return;
 826
 827         memset(&br_group.u, 0, sizeof(br_group.u));
 828
 829         br_group.proto = htons(ETH_P_IP);
 830         __br_multicast_send_query(br, port, &br_group);
 831
 832 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
 833         br_group.proto = htons(ETH_P_IPV6);
 834         __br_multicast_send_query(br, port, &br_group);
 835 #endif
 836
 837         time = jiffies;
 838         time += sent < br->multicast_startup_query_count ?
 839                 br->multicast_startup_query_interval :
 840                 br->multicast_query_interval;
 841         mod_timer(port ? &port->multicast_query_timer :
 842                          &br->multicast_query_timer, time);
 843 }
 844
 845 static void br_multicast_port_query_expired(unsigned long data)
 846 {
 847         struct net_bridge_port *port = (void *)data;
 848         struct net_bridge *br = port->br;
 849
 850         spin_lock(&br->multicast_lock);
 851         if (port->state == BR_STATE_DISABLED ||
 852             port->state == BR_STATE_BLOCKING)
 853                 goto out;
 854
 855         if (port->multicast_startup_queries_sent <
 856             br->multicast_startup_query_count)
 857                 port->multicast_startup_queries_sent++;
 858
 859         br_multicast_send_query(port->br, port,
 860                                 port->multicast_startup_queries_sent);
 861
 862 out:
 863         spin_unlock(&br->multicast_lock);
 864 }
 865
 866 void br_multicast_add_port(struct net_bridge_port *port)
 867 {
 868         port->multicast_router = 1;
 869
 870         setup_timer(&port->multicast_router_timer, br_multicast_router_expired,
 871                     (unsigned long)port);
 872         setup_timer(&port->multicast_query_timer,
 873                     br_multicast_port_query_expired, (unsigned long)port);
 874 }
 875
 876 void br_multicast_del_port(struct net_bridge_port *port)
 877 {
 878         del_timer_sync(&port->multicast_router_timer);
 879 }
 880
 881 static void __br_multicast_enable_port(struct net_bridge_port *port)
 882 {
 883         port->multicast_startup_queries_sent = 0;
 884
 885         if (try_to_del_timer_sync(&port->multicast_query_timer) >= 0 ||
 886             del_timer(&port->multicast_query_timer))
 887                 mod_timer(&port->multicast_query_timer, jiffies);
 888 }
 889
 890 void br_multicast_enable_port(struct net_bridge_port *port)
 891 {
 892         struct net_bridge *br = port->br;
 893
 894         spin_lock(&br->multicast_lock);
 895         if (br->multicast_disabled || !netif_running(br->dev))
 896                 goto out;
 897
 898         __br_multicast_enable_port(port);
 899
 900 out:
 901         spin_unlock(&br->multicast_lock);
 902 }
 903
 904 void br_multicast_disable_port(struct net_bridge_port *port)
 905 {
 906         struct net_bridge *br = port->br;
 907         struct net_bridge_port_group *pg;
 908         struct hlist_node *p, *n;
 909
 910         spin_lock(&br->multicast_lock);
 911         hlist_for_each_entry_safe(pg, p, n, &port->mglist, mglist)
 912                 br_multicast_del_pg(br, pg);
 913
 914         if (!hlist_unhashed(&port->rlist))
 915                 hlist_del_init_rcu(&port->rlist);
 916         del_timer(&port->multicast_router_timer);
 917         del_timer(&port->multicast_query_timer);
 918         spin_unlock(&br->multicast_lock);
 919 }
 920
 921 static int br_ip4_multicast_igmp3_report(struct net_bridge *br,
 922                                          struct net_bridge_port *port,
 923                                          struct sk_buff *skb)
 924 {
 925         struct igmpv3_report *ih;
 926         struct igmpv3_grec *grec;
 927         int i;
 928         int len;
 929         int num;
 930         int type;
 931         int err = 0;
 932         __be32 group;
 933
 934         if (!pskb_may_pull(skb, sizeof(*ih)))
 935                 return -EINVAL;
 936
 937         ih = igmpv3_report_hdr(skb);
 938         num = ntohs(ih->ngrec);
 939         len = sizeof(*ih);
 940
 941         for (i = 0; i < num; i++) {
 942                 len += sizeof(*grec);
 943                 if (!pskb_may_pull(skb, len))
 944                         return -EINVAL;
 945
 946                 grec = (void *)(skb->data + len - sizeof(*grec));
 947                 group = grec->grec_mca;
 948                 type = grec->grec_type;
 949
 950                 len += ntohs(grec->grec_nsrcs) * 4;
 951                 if (!pskb_may_pull(skb, len))
 952                         return -EINVAL;
 953
 954                 /* We treat this as an IGMPv2 report for now. */
 955                 switch (type) {
 956                 case IGMPV3_MODE_IS_INCLUDE:
 957                 case IGMPV3_MODE_IS_EXCLUDE:
 958                 case IGMPV3_CHANGE_TO_INCLUDE:
 959                 case IGMPV3_CHANGE_TO_EXCLUDE:
 960                 case IGMPV3_ALLOW_NEW_SOURCES:
 961                 case IGMPV3_BLOCK_OLD_SOURCES:
 962                         break;
 963
 964                 default:
 965                         continue;
 966                 }
 967
 968                 err = br_ip4_multicast_add_group(br, port, group);
 969                 if (err)
 970                         break;
 971         }
 972
 973         return err;
 974 }
 975
 976 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
 977 static int br_ip6_multicast_mld2_report(struct net_bridge *br,
 978                                         struct net_bridge_port *port,
 979                                         struct sk_buff *skb)
 980 {
 981         struct icmp6hdr *icmp6h;
 982         struct mld2_grec *grec;
 983         int i;
 984         int len;
 985         int num;
 986         int err = 0;
 987
 988         if (!pskb_may_pull(skb, sizeof(*icmp6h)))
 989                 return -EINVAL;
 990
 991         icmp6h = icmp6_hdr(skb);
 992         num = ntohs(icmp6h->icmp6_dataun.un_data16[1]);
 993         len = sizeof(*icmp6h);
 994
 995         for (i = 0; i < num; i++) {
 996                 __be16 *nsrcs, _nsrcs;
 997
 998                 nsrcs = skb_header_pointer(skb,
 999                                            len + offsetof(struct mld2_grec,
1000                                                           grec_mca),
1001                                            sizeof(_nsrcs), &_nsrcs);
1002                 if (!nsrcs)
1003                         return -EINVAL;
1004
1005                 if (!pskb_may_pull(skb,
1006                                    len + sizeof(*grec) +
1007                                    sizeof(struct in6_addr) * (*nsrcs)))
1008                         return -EINVAL;
1009
1010                 grec = (struct mld2_grec *)(skb->data + len);
1011                 len += sizeof(*grec) + sizeof(struct in6_addr) * (*nsrcs);
1012
1013                 /* We treat these as MLDv1 reports for now. */
1014                 switch (grec->grec_type) {
1015                 case MLD2_MODE_IS_INCLUDE:
1016                 case MLD2_MODE_IS_EXCLUDE:
1017                 case MLD2_CHANGE_TO_INCLUDE:
1018                 case MLD2_CHANGE_TO_EXCLUDE:
1019                 case MLD2_ALLOW_NEW_SOURCES:
1020                 case MLD2_BLOCK_OLD_SOURCES:
1021                         break;
1022
1023                 default:
1024                         continue;
1025                 }
1026
1027                 err = br_ip6_multicast_add_group(br, port, &grec->grec_mca);
1028                 if (!err)
1029                         break;
1030         }
1031
1032         return err;
1033 }
1034 #endif
1035
1036 /*
1037  * Add port to rotuer_list
1038  *  list is maintained ordered by pointer value
1039  *  and locked by br->multicast_lock and RCU
1040  */
1041 static void br_multicast_add_router(struct net_bridge *br,
1042                                     struct net_bridge_port *port)
1043 {
1044         struct net_bridge_port *p;
1045         struct hlist_node *n, *slot = NULL;
1046
1047         hlist_for_each_entry(p, n, &br->router_list, rlist) {
1048                 if ((unsigned long) port >= (unsigned long) p)
1049                         break;
1050                 slot = n;
1051         }
1052
1053         if (slot)
1054                 hlist_add_after_rcu(slot, &port->rlist);
1055         else
1056                 hlist_add_head_rcu(&port->rlist, &br->router_list);
1057 }
1058
1059 static void br_multicast_mark_router(struct net_bridge *br,
1060                                      struct net_bridge_port *port)
1061 {
1062         unsigned long now = jiffies;
1063
1064         if (!port) {
1065                 if (br->multicast_router == 1)
1066                         mod_timer(&br->multicast_router_timer,
1067                                   now + br->multicast_querier_interval);
1068                 return;
1069         }
1070
1071         if (port->multicast_router != 1)
1072                 return;
1073
1074         if (!hlist_unhashed(&port->rlist))
1075                 goto timer;
1076
1077         br_multicast_add_router(br, port);
1078
1079 timer:
1080         mod_timer(&port->multicast_router_timer,
1081                   now + br->multicast_querier_interval);
1082 }
1083
1084 static void br_multicast_query_received(struct net_bridge *br,
1085                                         struct net_bridge_port *port,
1086                                         int saddr)
1087 {
1088         if (saddr)
1089                 mod_timer(&br->multicast_querier_timer,
1090                           jiffies + br->multicast_querier_interval);
1091         else if (timer_pending(&br->multicast_querier_timer))
1092                 return;
1093
1094         br_multicast_mark_router(br, port);
1095 }
1096
1097 static int br_ip4_multicast_query(struct net_bridge *br,
1098                                   struct net_bridge_port *port,
1099                                   struct sk_buff *skb)
1100 {
1101         struct iphdr *iph = ip_hdr(skb);
1102         struct igmphdr *ih = igmp_hdr(skb);
1103         struct net_bridge_mdb_entry *mp;
1104         struct igmpv3_query *ih3;
1105         struct net_bridge_port_group *p;
1106         struct net_bridge_port_group **pp;
1107         unsigned long max_delay;
1108         unsigned long now = jiffies;
1109         __be32 group;
1110         int err = 0;
1111
1112         spin_lock(&br->multicast_lock);
1113         if (!netif_running(br->dev) ||
1114             (port && port->state == BR_STATE_DISABLED))
1115                 goto out;
1116
1117         br_multicast_query_received(br, port, !!iph->saddr);
1118
1119         group = ih->group;
1120
1121         if (skb->len == sizeof(*ih)) {
1122                 max_delay = ih->code * (HZ / IGMP_TIMER_SCALE);
1123
1124                 if (!max_delay) {
1125                         max_delay = 10 * HZ;
1126                         group = 0;
1127                 }
1128         } else {
1129                 if (!pskb_may_pull(skb, sizeof(struct igmpv3_query))) {
1130                         err = -EINVAL;
1131                         goto out;
1132                 }
1133
1134                 ih3 = igmpv3_query_hdr(skb);
1135                 if (ih3->nsrcs)
1136                         goto out;
1137
1138                 max_delay = ih3->code ?
1139                             IGMPV3_MRC(ih3->code) * (HZ / IGMP_TIMER_SCALE) : 1;
1140         }
1141
1142         if (!group)
1143                 goto out;
1144
1145         mp = br_mdb_ip4_get(br->mdb, group);
1146         if (!mp)
1147                 goto out;
1148
1149         max_delay *= br->multicast_last_member_count;
1150
1151         if (!hlist_unhashed(&mp->mglist) &&
1152             (timer_pending(&mp->timer) ?
1153              time_after(mp->timer.expires, now + max_delay) :
1154              try_to_del_timer_sync(&mp->timer) >= 0))
1155                 mod_timer(&mp->timer, now + max_delay);
1156
1157         for (pp = &mp->ports; (p = *pp); pp = &p->next) {
1158                 if (timer_pending(&p->timer) ?
1159                     time_after(p->timer.expires, now + max_delay) :
1160                     try_to_del_timer_sync(&p->timer) >= 0)
1161                         mod_timer(&mp->timer, now + max_delay);
1162         }
1163
1164 out:
1165         spin_unlock(&br->multicast_lock);
1166         return err;
1167 }
1168
1169 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
1170 static int br_ip6_multicast_query(struct net_bridge *br,
1171                                   struct net_bridge_port *port,
1172                                   struct sk_buff *skb)
1173 {
1174         struct ipv6hdr *ip6h = ipv6_hdr(skb);
1175         struct mld_msg *mld = (struct mld_msg *) icmp6_hdr(skb);
1176         struct net_bridge_mdb_entry *mp;
1177         struct mld2_query *mld2q;
1178         struct net_bridge_port_group *p, **pp;
1179         unsigned long max_delay;
1180         unsigned long now = jiffies;
1181         struct in6_addr *group = NULL;
1182         int err = 0;
1183
1184         spin_lock(&br->multicast_lock);
1185         if (!netif_running(br->dev) ||
1186             (port && port->state == BR_STATE_DISABLED))
1187                 goto out;
1188
1189         br_multicast_query_received(br, port, !ipv6_addr_any(&ip6h->saddr));
1190
1191         if (skb->len == sizeof(*mld)) {
1192                 if (!pskb_may_pull(skb, sizeof(*mld))) {
1193                         err = -EINVAL;
1194                         goto out;
1195                 }
1196                 mld = (struct mld_msg *) icmp6_hdr(skb);
1197                 max_delay = msecs_to_jiffies(htons(mld->mld_maxdelay));
1198                 if (max_delay)
1199                         group = &mld->mld_mca;
1200         } else if (skb->len >= sizeof(*mld2q)) {
1201                 if (!pskb_may_pull(skb, sizeof(*mld2q))) {
1202                         err = -EINVAL;
1203                         goto out;
1204                 }
1205                 mld2q = (struct mld2_query *)icmp6_hdr(skb);
1206                 if (!mld2q->mld2q_nsrcs)
1207                         group = &mld2q->mld2q_mca;
1208                 max_delay = mld2q->mld2q_mrc ? MLDV2_MRC(mld2q->mld2q_mrc) : 1;
1209         }
1210
1211         if (!group)
1212                 goto out;
1213
1214         mp = br_mdb_ip6_get(br->mdb, group);
1215         if (!mp)
1216                 goto out;
1217
1218         max_delay *= br->multicast_last_member_count;
1219         if (!hlist_unhashed(&mp->mglist) &&
1220             (timer_pending(&mp->timer) ?
1221              time_after(mp->timer.expires, now + max_delay) :
1222              try_to_del_timer_sync(&mp->timer) >= 0))
1223                 mod_timer(&mp->timer, now + max_delay);
1224
1225         for (pp = &mp->ports; (p = *pp); pp = &p->next) {
1226                 if (timer_pending(&p->timer) ?
1227                     time_after(p->timer.expires, now + max_delay) :
1228                     try_to_del_timer_sync(&p->timer) >= 0)
1229                         mod_timer(&mp->timer, now + max_delay);
1230         }
1231
1232 out:
1233         spin_unlock(&br->multicast_lock);
1234         return err;
1235 }
1236 #endif
1237
1238 static void br_multicast_leave_group(struct net_bridge *br,
1239                                      struct net_bridge_port *port,
1240                                      struct br_ip *group)
1241 {
1242         struct net_bridge_mdb_htable *mdb;
1243         struct net_bridge_mdb_entry *mp;
1244         struct net_bridge_port_group *p;
1245         unsigned long now;
1246         unsigned long time;
1247
1248         spin_lock(&br->multicast_lock);
1249         if (!netif_running(br->dev) ||
1250             (port && port->state == BR_STATE_DISABLED) ||
1251             timer_pending(&br->multicast_querier_timer))
1252                 goto out;
1253
1254         mdb = br->mdb;
1255         mp = br_mdb_ip_get(mdb, group);
1256         if (!mp)
1257                 goto out;
1258
1259         now = jiffies;
1260         time = now + br->multicast_last_member_count *
1261                      br->multicast_last_member_interval;
1262
1263         if (!port) {
1264                 if (!hlist_unhashed(&mp->mglist) &&
1265                     (timer_pending(&mp->timer) ?
1266                      time_after(mp->timer.expires, time) :
1267                      try_to_del_timer_sync(&mp->timer) >= 0)) {
1268                         mod_timer(&mp->timer, time);
1269
1270                         mp->queries_sent = 0;
1271                         mod_timer(&mp->query_timer, now);
1272                 }
1273
1274                 goto out;
1275         }
1276
1277         for (p = mp->ports; p; p = p->next) {
1278                 if (p->port != port)
1279                         continue;
1280
1281                 if (!hlist_unhashed(&p->mglist) &&
1282                     (timer_pending(&p->timer) ?
1283                      time_after(p->timer.expires, time) :
1284                      try_to_del_timer_sync(&p->timer) >= 0)) {
1285                         mod_timer(&p->timer, time);
1286
1287                         p->queries_sent = 0;
1288                         mod_timer(&p->query_timer, now);
1289                 }
1290
1291                 break;
1292         }
1293
1294 out:
1295         spin_unlock(&br->multicast_lock);
1296 }
1297
1298 static void br_ip4_multicast_leave_group(struct net_bridge *br,
1299                                          struct net_bridge_port *port,
1300                                          __be32 group)
1301 {
1302         struct br_ip br_group;
1303
1304         if (ipv4_is_local_multicast(group))
1305                 return;
1306
1307         br_group.u.ip4 = group;
1308         br_group.proto = htons(ETH_P_IP);
1309
1310         br_multicast_leave_group(br, port, &br_group);
1311 }
1312
1313 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
1314 static void br_ip6_multicast_leave_group(struct net_bridge *br,
1315                                          struct net_bridge_port *port,
1316                                          const struct in6_addr *group)
1317 {
1318         struct br_ip br_group;
1319
1320         if (ipv6_is_local_multicast(group))
1321                 return;
1322
1323         ipv6_addr_copy(&br_group.u.ip6, group);
1324         br_group.proto = htons(ETH_P_IPV6);
1325
1326         br_multicast_leave_group(br, port, &br_group);
1327 }
1328 #endif
1329
1330 static int br_multicast_ipv4_rcv(struct net_bridge *br,
1331                                  struct net_bridge_port *port,
1332                                  struct sk_buff *skb)
1333 {
1334         struct sk_buff *skb2 = skb;
1335         struct iphdr *iph;
1336         struct igmphdr *ih;
1337         unsigned len;
1338         unsigned offset;
1339         int err;
1340
1341         /* We treat OOM as packet loss for now. */
1342         if (!pskb_may_pull(skb, sizeof(*iph)))
1343                 return -EINVAL;
1344
1345         iph = ip_hdr(skb);
1346
1347         if (iph->ihl < 5 || iph->version != 4)
1348                 return -EINVAL;
1349
1350         if (!pskb_may_pull(skb, ip_hdrlen(skb)))
1351                 return -EINVAL;
1352
1353         iph = ip_hdr(skb);
1354
1355         if (unlikely(ip_fast_csum((u8 *)iph, iph->ihl)))
1356                 return -EINVAL;
1357
1358         if (iph->protocol != IPPROTO_IGMP)
1359                 return 0;
1360
1361         len = ntohs(iph->tot_len);
1362         if (skb->len < len || len < ip_hdrlen(skb))
1363                 return -EINVAL;
1364
1365         if (skb->len > len) {
1366                 skb2 = skb_clone(skb, GFP_ATOMIC);
1367                 if (!skb2)
1368                         return -ENOMEM;
1369
1370                 err = pskb_trim_rcsum(skb2, len);
1371                 if (err)
1372                         goto err_out;
1373         }
1374
1375         len -= ip_hdrlen(skb2);
1376         offset = skb_network_offset(skb2) + ip_hdrlen(skb2);
1377         __skb_pull(skb2, offset);
1378         skb_reset_transport_header(skb2);
1379
1380         err = -EINVAL;
1381         if (!pskb_may_pull(skb2, sizeof(*ih)))
1382                 goto out;
1383
1384         switch (skb2->ip_summed) {
1385         case CHECKSUM_COMPLETE:
1386                 if (!csum_fold(skb2->csum))
1387                         break;
1388                 /* fall through */
1389         case CHECKSUM_NONE:
1390                 skb2->csum = 0;
1391                 if (skb_checksum_complete(skb2))
1392                         goto out;
1393         }
1394
1395         err = 0;
1396
1397         BR_INPUT_SKB_CB(skb)->igmp = 1;
1398         ih = igmp_hdr(skb2);
1399
1400         switch (ih->type) {
1401         case IGMP_HOST_MEMBERSHIP_REPORT:
1402         case IGMPV2_HOST_MEMBERSHIP_REPORT:
1403                 BR_INPUT_SKB_CB(skb2)->mrouters_only = 1;
1404                 err = br_ip4_multicast_add_group(br, port, ih->group);
1405                 break;
1406         case IGMPV3_HOST_MEMBERSHIP_REPORT:
1407                 err = br_ip4_multicast_igmp3_report(br, port, skb2);
1408                 break;
1409         case IGMP_HOST_MEMBERSHIP_QUERY:
1410                 err = br_ip4_multicast_query(br, port, skb2);
1411                 break;
1412         case IGMP_HOST_LEAVE_MESSAGE:
1413                 br_ip4_multicast_leave_group(br, port, ih->group);
1414                 break;
1415         }
1416
1417 out:
1418         __skb_push(skb2, offset);
1419 err_out:
1420         if (skb2 != skb)
1421                 kfree_skb(skb2);
1422         return err;
1423 }
1424
1425 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
1426 static int br_multicast_ipv6_rcv(struct net_bridge *br,
1427                                  struct net_bridge_port *port,
1428                                  struct sk_buff *skb)
1429 {
1430         struct sk_buff *skb2 = skb;
1431         struct ipv6hdr *ip6h;
1432         struct icmp6hdr *icmp6h;
1433         u8 nexthdr;
1434         unsigned len;
1435         unsigned offset;
1436         int err;
1437
1438         if (!pskb_may_pull(skb, sizeof(*ip6h)))
1439                 return -EINVAL;
1440
1441         ip6h = ipv6_hdr(skb);
1442
1443         /*
1444          * We're interested in MLD messages only.
1445          *  - Version is 6
1446          *  - MLD has always Router Alert hop-by-hop option
1447          *  - But we do not support jumbrograms.
1448          */
1449         if (ip6h->version != 6 ||
1450             ip6h->nexthdr != IPPROTO_HOPOPTS ||
1451             ip6h->payload_len == 0)
1452                 return 0;
1453
1454         len = ntohs(ip6h->payload_len);
1455         if (skb->len < len)
1456                 return -EINVAL;
1457
1458         nexthdr = ip6h->nexthdr;
1459         offset = ipv6_skip_exthdr(skb, sizeof(*ip6h), &nexthdr);
1460
1461         if (offset < 0 || nexthdr != IPPROTO_ICMPV6)
1462                 return 0;
1463
1464         /* Okay, we found ICMPv6 header */
1465         skb2 = skb_clone(skb, GFP_ATOMIC);
1466         if (!skb2)
1467                 return -ENOMEM;
1468
1469         len -= offset - skb_network_offset(skb2);
1470
1471         __skb_pull(skb2, offset);
1472         skb_reset_transport_header(skb2);
1473
1474         err = -EINVAL;
1475         if (!pskb_may_pull(skb2, sizeof(*icmp6h)))
1476                 goto out;
1477
1478         icmp6h = icmp6_hdr(skb2);
1479
1480         switch (icmp6h->icmp6_type) {
1481         case ICMPV6_MGM_QUERY:
1482         case ICMPV6_MGM_REPORT:
1483         case ICMPV6_MGM_REDUCTION:
1484         case ICMPV6_MLD2_REPORT:
1485                 break;
1486         default:
1487                 err = 0;
1488                 goto out;
1489         }
1490
1491         /* Okay, we found MLD message. Check further. */
1492         if (skb2->len > len) {
1493                 err = pskb_trim_rcsum(skb2, len);
1494                 if (err)
1495                         goto out;
1496         }
1497
1498         switch (skb2->ip_summed) {
1499         case CHECKSUM_COMPLETE:
1500                 if (!csum_fold(skb2->csum))
1501                         break;
1502                 /*FALLTHROUGH*/
1503         case CHECKSUM_NONE:
1504                 skb2->csum = 0;
1505                 if (skb_checksum_complete(skb2))
1506                         goto out;
1507         }
1508
1509         err = 0;
1510
1511         BR_INPUT_SKB_CB(skb)->igmp = 1;
1512
1513         switch (icmp6h->icmp6_type) {
1514         case ICMPV6_MGM_REPORT:
1515             {
1516                 struct mld_msg *mld = (struct mld_msg *)icmp6h;
1517                 BR_INPUT_SKB_CB(skb2)->mrouters_only = 1;
1518                 err = br_ip6_multicast_add_group(br, port, &mld->mld_mca);
1519                 break;
1520             }
1521         case ICMPV6_MLD2_REPORT:
1522                 err = br_ip6_multicast_mld2_report(br, port, skb2);
1523                 break;
1524         case ICMPV6_MGM_QUERY:
1525                 err = br_ip6_multicast_query(br, port, skb2);
1526                 break;
1527         case ICMPV6_MGM_REDUCTION:
1528             {
1529                 struct mld_msg *mld = (struct mld_msg *)icmp6h;
1530                 br_ip6_multicast_leave_group(br, port, &mld->mld_mca);
1531             }
1532         }
1533
1534 out:
1535         __skb_push(skb2, offset);
1536         if (skb2 != skb)
1537                 kfree_skb(skb2);
1538         return err;
1539 }
1540 #endif
1541
1542 int br_multicast_rcv(struct net_bridge *br, struct net_bridge_port *port,
1543                      struct sk_buff *skb)
1544 {
1545         BR_INPUT_SKB_CB(skb)->igmp = 0;
1546         BR_INPUT_SKB_CB(skb)->mrouters_only = 0;
1547
1548         if (br->multicast_disabled)
1549                 return 0;
1550
1551         switch (skb->protocol) {
1552         case htons(ETH_P_IP):
1553                 return br_multicast_ipv4_rcv(br, port, skb);
1554 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
1555         case htons(ETH_P_IPV6):
1556                 return br_multicast_ipv6_rcv(br, port, skb);
1557 #endif
1558         }
1559
1560         return 0;
1561 }
1562
1563 static void br_multicast_query_expired(unsigned long data)
1564 {
1565         struct net_bridge *br = (void *)data;
1566
1567         spin_lock(&br->multicast_lock);
1568         if (br->multicast_startup_queries_sent <
1569             br->multicast_startup_query_count)
1570                 br->multicast_startup_queries_sent++;
1571
1572         br_multicast_send_query(br, NULL, br->multicast_startup_queries_sent);
1573
1574         spin_unlock(&br->multicast_lock);
1575 }
1576
1577 void br_multicast_init(struct net_bridge *br)
1578 {
1579         br->hash_elasticity = 4;
1580         br->hash_max = 512;
1581
1582         br->multicast_router = 1;
1583         br->multicast_last_member_count = 2;
1584         br->multicast_startup_query_count = 2;
1585
1586         br->multicast_last_member_interval = HZ;
1587         br->multicast_query_response_interval = 10 * HZ;
1588         br->multicast_startup_query_interval = 125 * HZ / 4;
1589         br->multicast_query_interval = 125 * HZ;
1590         br->multicast_querier_interval = 255 * HZ;
1591         br->multicast_membership_interval = 260 * HZ;
1592
1593         spin_lock_init(&br->multicast_lock);
1594         setup_timer(&br->multicast_router_timer,
1595                     br_multicast_local_router_expired, 0);
1596         setup_timer(&br->multicast_querier_timer,
1597                     br_multicast_local_router_expired, 0);
1598         setup_timer(&br->multicast_query_timer, br_multicast_query_expired,
1599                     (unsigned long)br);
1600 }
1601
1602 void br_multicast_open(struct net_bridge *br)
1603 {
1604         br->multicast_startup_queries_sent = 0;
1605
1606         if (br->multicast_disabled)
1607                 return;
1608
1609         mod_timer(&br->multicast_query_timer, jiffies);
1610 }
1611
1612 void br_multicast_stop(struct net_bridge *br)
1613 {
1614         struct net_bridge_mdb_htable *mdb;
1615         struct net_bridge_mdb_entry *mp;
1616         struct hlist_node *p, *n;
1617         u32 ver;
1618         int i;
1619
1620         del_timer_sync(&br->multicast_router_timer);
1621         del_timer_sync(&br->multicast_querier_timer);
1622         del_timer_sync(&br->multicast_query_timer);
1623
1624         spin_lock_bh(&br->multicast_lock);
1625         mdb = br->mdb;
1626         if (!mdb)
1627                 goto out;
1628
1629         br->mdb = NULL;
1630
1631         ver = mdb->ver;
1632         for (i = 0; i < mdb->max; i++) {
1633                 hlist_for_each_entry_safe(mp, p, n, &mdb->mhash[i],
1634                                           hlist[ver]) {
1635                         del_timer(&mp->timer);
1636                         del_timer(&mp->query_timer);
1637                         call_rcu_bh(&mp->rcu, br_multicast_free_group);
1638                 }
1639         }
1640
1641         if (mdb->old) {
1642                 spin_unlock_bh(&br->multicast_lock);
1643                 rcu_barrier_bh();
1644                 spin_lock_bh(&br->multicast_lock);
1645                 WARN_ON(mdb->old);
1646         }
1647
1648         mdb->old = mdb;
1649         call_rcu_bh(&mdb->rcu, br_mdb_free);
1650
1651 out:
1652         spin_unlock_bh(&br->multicast_lock);
1653 }
1654
1655 int br_multicast_set_router(struct net_bridge *br, unsigned long val)
1656 {
1657         int err = -ENOENT;
1658
1659         spin_lock_bh(&br->multicast_lock);
1660         if (!netif_running(br->dev))
1661                 goto unlock;
1662
1663         switch (val) {
1664         case 0:
1665         case 2:
1666                 del_timer(&br->multicast_router_timer);
1667                 /* fall through */
1668         case 1:
1669                 br->multicast_router = val;
1670                 err = 0;
1671                 break;
1672
1673         default:
1674                 err = -EINVAL;
1675                 break;
1676         }
1677
1678 unlock:
1679         spin_unlock_bh(&br->multicast_lock);
1680
1681         return err;
1682 }
1683
1684 int br_multicast_set_port_router(struct net_bridge_port *p, unsigned long val)
1685 {
1686         struct net_bridge *br = p->br;
1687         int err = -ENOENT;
1688
1689         spin_lock(&br->multicast_lock);
1690         if (!netif_running(br->dev) || p->state == BR_STATE_DISABLED)
1691                 goto unlock;
1692
1693         switch (val) {
1694         case 0:
1695         case 1:
1696         case 2:
1697                 p->multicast_router = val;
1698                 err = 0;
1699
1700                 if (val < 2 && !hlist_unhashed(&p->rlist))
1701                         hlist_del_init_rcu(&p->rlist);
1702
1703                 if (val == 1)
1704                         break;
1705
1706                 del_timer(&p->multicast_router_timer);
1707
1708                 if (val == 0)
1709                         break;
1710
1711                 br_multicast_add_router(br, p);
1712                 break;
1713
1714         default:
1715                 err = -EINVAL;
1716                 break;
1717         }
1718
1719 unlock:
1720         spin_unlock(&br->multicast_lock);
1721
1722         return err;
1723 }
1724
1725 int br_multicast_toggle(struct net_bridge *br, unsigned long val)
1726 {
1727         struct net_bridge_port *port;
1728         int err = -ENOENT;
1729
1730         spin_lock(&br->multicast_lock);
1731         if (!netif_running(br->dev))
1732                 goto unlock;
1733
1734         err = 0;
1735         if (br->multicast_disabled == !val)
1736                 goto unlock;
1737
1738         br->multicast_disabled = !val;
1739         if (br->multicast_disabled)
1740                 goto unlock;
1741
1742         if (br->mdb) {
1743                 if (br->mdb->old) {
1744                         err = -EEXIST;
1745 rollback:
1746                         br->multicast_disabled = !!val;
1747                         goto unlock;
1748                 }
1749
1750                 err = br_mdb_rehash(&br->mdb, br->mdb->max,
1751                                     br->hash_elasticity);
1752                 if (err)
1753                         goto rollback;
1754         }
1755
1756         br_multicast_open(br);
1757         list_for_each_entry(port, &br->port_list, list) {
1758                 if (port->state == BR_STATE_DISABLED ||
1759                     port->state == BR_STATE_BLOCKING)
1760                         continue;
1761
1762                 __br_multicast_enable_port(port);
1763         }
1764
1765 unlock:
1766         spin_unlock(&br->multicast_lock);
1767
1768         return err;
1769 }
1770
1771 int br_multicast_set_hash_max(struct net_bridge *br, unsigned long val)
1772 {
1773         int err = -ENOENT;
1774         u32 old;
1775
1776         spin_lock(&br->multicast_lock);
1777         if (!netif_running(br->dev))
1778                 goto unlock;
1779
1780         err = -EINVAL;
1781         if (!is_power_of_2(val))
1782                 goto unlock;
1783         if (br->mdb && val < br->mdb->size)
1784                 goto unlock;
1785
1786         err = 0;
1787
1788         old = br->hash_max;
1789         br->hash_max = val;
1790
1791         if (br->mdb) {
1792                 if (br->mdb->old) {
1793                         err = -EEXIST;
1794 rollback:
1795                         br->hash_max = old;
1796                         goto unlock;
1797                 }
1798
1799                 err = br_mdb_rehash(&br->mdb, br->hash_max,
1800                                     br->hash_elasticity);
1801                 if (err)
1802                         goto rollback;
1803         }
1804
1805 unlock:
1806         spin_unlock(&br->multicast_lock);
1807
1808         return err;
1809 }