net/bridge/br_multicast.c

   1 /*
   2  * Bridge multicast support.
   3  *
   4  * Copyright (c) 2010 Herbert Xu <herbert@gondor.apana.org.au>
   5  *
   6  * This program is free software; you can redistribute it and/or modify it
   7  * under the terms of the GNU General Public License as published by the Free
   8  * Software Foundation; either version 2 of the License, or (at your option)
   9  * any later version.
  10  *
  11  */
  12
  13 #include <linux/err.h>
  14 #include <linux/if_ether.h>
  15 #include <linux/igmp.h>
  16 #include <linux/jhash.h>
  17 #include <linux/kernel.h>
  18 #include <linux/log2.h>
  19 #include <linux/netdevice.h>
  20 #include <linux/netfilter_bridge.h>
  21 #include <linux/random.h>
  22 #include <linux/rculist.h>
  23 #include <linux/skbuff.h>
  24 #include <linux/slab.h>
  25 #include <linux/timer.h>
  26 #include <net/ip.h>
  27 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
  28 #include <net/ipv6.h>
  29 #include <net/mld.h>
  30 #include <net/addrconf.h>
  31 #include <net/ip6_checksum.h>
  32 #endif
  33
  34 #include "br_private.h"
  35
  36 #define mlock_dereference(X, br) \
  37         rcu_dereference_protected(X, lockdep_is_held(&br->multicast_lock))
  38
  39 static void br_multicast_add_router(struct net_bridge *br,
  40                                     struct net_bridge_port *port);
  41
  42 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
  43 static inline int ipv6_is_transient_multicast(const struct in6_addr *addr)
  44 {
  45         if (ipv6_addr_is_multicast(addr) && IPV6_ADDR_MC_FLAG_TRANSIENT(addr))
  46                 return 1;
  47         return 0;
  48 }
  49 #endif
  50
  51 static inline int br_ip_equal(const struct br_ip *a, const struct br_ip *b)
  52 {
  53         if (a->proto != b->proto)
  54                 return 0;
  55         switch (a->proto) {
  56         case htons(ETH_P_IP):
  57                 return a->u.ip4 == b->u.ip4;
  58 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
  59         case htons(ETH_P_IPV6):
  60                 return ipv6_addr_equal(&a->u.ip6, &b->u.ip6);
  61 #endif
  62         }
  63         return 0;
  64 }
  65
  66 static inline int __br_ip4_hash(struct net_bridge_mdb_htable *mdb, __be32 ip)
  67 {
  68         return jhash_1word(mdb->secret, (__force u32)ip) & (mdb->max - 1);
  69 }
  70
  71 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
  72 static inline int __br_ip6_hash(struct net_bridge_mdb_htable *mdb,
  73                                 const struct in6_addr *ip)
  74 {
  75         return jhash2((__force u32 *)ip->s6_addr32, 4, mdb->secret) & (mdb->max - 1);
  76 }
  77 #endif
  78
  79 static inline int br_ip_hash(struct net_bridge_mdb_htable *mdb,
  80                              struct br_ip *ip)
  81 {
  82         switch (ip->proto) {
  83         case htons(ETH_P_IP):
  84                 return __br_ip4_hash(mdb, ip->u.ip4);
  85 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
  86         case htons(ETH_P_IPV6):
  87                 return __br_ip6_hash(mdb, &ip->u.ip6);
  88 #endif
  89         }
  90         return 0;
  91 }
  92
  93 static struct net_bridge_mdb_entry *__br_mdb_ip_get(
  94         struct net_bridge_mdb_htable *mdb, struct br_ip *dst, int hash)
  95 {
  96         struct net_bridge_mdb_entry *mp;
  97         struct hlist_node *p;
  98
  99         hlist_for_each_entry_rcu(mp, p, &mdb->mhash[hash], hlist[mdb->ver]) {
 100                 if (br_ip_equal(&mp->addr, dst))
 101                         return mp;
 102         }
 103
 104         return NULL;
 105 }
 106
 107 static struct net_bridge_mdb_entry *br_mdb_ip_get(
 108         struct net_bridge_mdb_htable *mdb, struct br_ip *dst)
 109 {
 110         if (!mdb)
 111                 return NULL;
 112
 113         return __br_mdb_ip_get(mdb, dst, br_ip_hash(mdb, dst));
 114 }
 115
 116 static struct net_bridge_mdb_entry *br_mdb_ip4_get(
 117         struct net_bridge_mdb_htable *mdb, __be32 dst)
 118 {
 119         struct br_ip br_dst;
 120
 121         br_dst.u.ip4 = dst;
 122         br_dst.proto = htons(ETH_P_IP);
 123
 124         return br_mdb_ip_get(mdb, &br_dst);
 125 }
 126
 127 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
 128 static struct net_bridge_mdb_entry *br_mdb_ip6_get(
 129         struct net_bridge_mdb_htable *mdb, const struct in6_addr *dst)
 130 {
 131         struct br_ip br_dst;
 132
 133         ipv6_addr_copy(&br_dst.u.ip6, dst);
 134         br_dst.proto = htons(ETH_P_IPV6);
 135
 136         return br_mdb_ip_get(mdb, &br_dst);
 137 }
 138 #endif
 139
 140 struct net_bridge_mdb_entry *br_mdb_get(struct net_bridge *br,
 141                                         struct sk_buff *skb)
 142 {
 143         struct net_bridge_mdb_htable *mdb = rcu_dereference(br->mdb);
 144         struct br_ip ip;
 145
 146         if (br->multicast_disabled)
 147                 return NULL;
 148
 149         if (BR_INPUT_SKB_CB(skb)->igmp)
 150                 return NULL;
 151
 152         ip.proto = skb->protocol;
 153
 154         switch (skb->protocol) {
 155         case htons(ETH_P_IP):
 156                 ip.u.ip4 = ip_hdr(skb)->daddr;
 157                 break;
 158 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
 159         case htons(ETH_P_IPV6):
 160                 ipv6_addr_copy(&ip.u.ip6, &ipv6_hdr(skb)->daddr);
 161                 break;
 162 #endif
 163         default:
 164                 return NULL;
 165         }
 166
 167         return br_mdb_ip_get(mdb, &ip);
 168 }
 169
 170 static void br_mdb_free(struct rcu_head *head)
 171 {
 172         struct net_bridge_mdb_htable *mdb =
 173                 container_of(head, struct net_bridge_mdb_htable, rcu);
 174         struct net_bridge_mdb_htable *old = mdb->old;
 175
 176         mdb->old = NULL;
 177         kfree(old->mhash);
 178         kfree(old);
 179 }
 180
 181 static int br_mdb_copy(struct net_bridge_mdb_htable *new,
 182                        struct net_bridge_mdb_htable *old,
 183                        int elasticity)
 184 {
 185         struct net_bridge_mdb_entry *mp;
 186         struct hlist_node *p;
 187         int maxlen;
 188         int len;
 189         int i;
 190
 191         for (i = 0; i < old->max; i++)
 192                 hlist_for_each_entry(mp, p, &old->mhash[i], hlist[old->ver])
 193                         hlist_add_head(&mp->hlist[new->ver],
 194                                        &new->mhash[br_ip_hash(new, &mp->addr)]);
 195
 196         if (!elasticity)
 197                 return 0;
 198
 199         maxlen = 0;
 200         for (i = 0; i < new->max; i++) {
 201                 len = 0;
 202                 hlist_for_each_entry(mp, p, &new->mhash[i], hlist[new->ver])
 203                         len++;
 204                 if (len > maxlen)
 205                         maxlen = len;
 206         }
 207
 208         return maxlen > elasticity ? -EINVAL : 0;
 209 }
 210
 211 static void br_multicast_free_pg(struct rcu_head *head)
 212 {
 213         struct net_bridge_port_group *p =
 214                 container_of(head, struct net_bridge_port_group, rcu);
 215
 216         kfree(p);
 217 }
 218
 219 static void br_multicast_free_group(struct rcu_head *head)
 220 {
 221         struct net_bridge_mdb_entry *mp =
 222                 container_of(head, struct net_bridge_mdb_entry, rcu);
 223
 224         kfree(mp);
 225 }
 226
 227 static void br_multicast_group_expired(unsigned long data)
 228 {
 229         struct net_bridge_mdb_entry *mp = (void *)data;
 230         struct net_bridge *br = mp->br;
 231         struct net_bridge_mdb_htable *mdb;
 232
 233         spin_lock(&br->multicast_lock);
 234         if (!netif_running(br->dev) || timer_pending(&mp->timer))
 235                 goto out;
 236
 237         mp->mglist = false;
 238
 239         if (mp->ports)
 240                 goto out;
 241
 242         mdb = mlock_dereference(br->mdb, br);
 243
 244         hlist_del_rcu(&mp->hlist[mdb->ver]);
 245         mdb->size--;
 246
 247         call_rcu_bh(&mp->rcu, br_multicast_free_group);
 248
 249 out:
 250         spin_unlock(&br->multicast_lock);
 251 }
 252
 253 static void br_multicast_del_pg(struct net_bridge *br,
 254                                 struct net_bridge_port_group *pg)
 255 {
 256         struct net_bridge_mdb_htable *mdb;
 257         struct net_bridge_mdb_entry *mp;
 258         struct net_bridge_port_group *p;
 259         struct net_bridge_port_group __rcu **pp;
 260
 261         mdb = mlock_dereference(br->mdb, br);
 262
 263         mp = br_mdb_ip_get(mdb, &pg->addr);
 264         if (WARN_ON(!mp))
 265                 return;
 266
 267         for (pp = &mp->ports;
 268              (p = mlock_dereference(*pp, br)) != NULL;
 269              pp = &p->next) {
 270                 if (p != pg)
 271                         continue;
 272
 273                 rcu_assign_pointer(*pp, p->next);
 274                 hlist_del_init(&p->mglist);
 275                 del_timer(&p->timer);
 276                 call_rcu_bh(&p->rcu, br_multicast_free_pg);
 277
 278                 if (!mp->ports && !mp->mglist &&
 279                     netif_running(br->dev))
 280                         mod_timer(&mp->timer, jiffies);
 281
 282                 return;
 283         }
 284
 285         WARN_ON(1);
 286 }
 287
 288 static void br_multicast_port_group_expired(unsigned long data)
 289 {
 290         struct net_bridge_port_group *pg = (void *)data;
 291         struct net_bridge *br = pg->port->br;
 292
 293         spin_lock(&br->multicast_lock);
 294         if (!netif_running(br->dev) || timer_pending(&pg->timer) ||
 295             hlist_unhashed(&pg->mglist))
 296                 goto out;
 297
 298         br_multicast_del_pg(br, pg);
 299
 300 out:
 301         spin_unlock(&br->multicast_lock);
 302 }
 303
 304 static int br_mdb_rehash(struct net_bridge_mdb_htable __rcu **mdbp, int max,
 305                          int elasticity)
 306 {
 307         struct net_bridge_mdb_htable *old = rcu_dereference_protected(*mdbp, 1);
 308         struct net_bridge_mdb_htable *mdb;
 309         int err;
 310
 311         mdb = kmalloc(sizeof(*mdb), GFP_ATOMIC);
 312         if (!mdb)
 313                 return -ENOMEM;
 314
 315         mdb->max = max;
 316         mdb->old = old;
 317
 318         mdb->mhash = kzalloc(max * sizeof(*mdb->mhash), GFP_ATOMIC);
 319         if (!mdb->mhash) {
 320                 kfree(mdb);
 321                 return -ENOMEM;
 322         }
 323
 324         mdb->size = old ? old->size : 0;
 325         mdb->ver = old ? old->ver ^ 1 : 0;
 326
 327         if (!old || elasticity)
 328                 get_random_bytes(&mdb->secret, sizeof(mdb->secret));
 329         else
 330                 mdb->secret = old->secret;
 331
 332         if (!old)
 333                 goto out;
 334
 335         err = br_mdb_copy(mdb, old, elasticity);
 336         if (err) {
 337                 kfree(mdb->mhash);
 338                 kfree(mdb);
 339                 return err;
 340         }
 341
 342         call_rcu_bh(&mdb->rcu, br_mdb_free);
 343
 344 out:
 345         rcu_assign_pointer(*mdbp, mdb);
 346
 347         return 0;
 348 }
 349
 350 static struct sk_buff *br_ip4_multicast_alloc_query(struct net_bridge *br,
 351                                                     __be32 group)
 352 {
 353         struct sk_buff *skb;
 354         struct igmphdr *ih;
 355         struct ethhdr *eth;
 356         struct iphdr *iph;
 357
 358         skb = netdev_alloc_skb_ip_align(br->dev, sizeof(*eth) + sizeof(*iph) +
 359                                                  sizeof(*ih) + 4);
 360         if (!skb)
 361                 goto out;
 362
 363         skb->protocol = htons(ETH_P_IP);
 364
 365         skb_reset_mac_header(skb);
 366         eth = eth_hdr(skb);
 367
 368         memcpy(eth->h_source, br->dev->dev_addr, 6);
 369         eth->h_dest[0] = 1;
 370         eth->h_dest[1] = 0;
 371         eth->h_dest[2] = 0x5e;
 372         eth->h_dest[3] = 0;
 373         eth->h_dest[4] = 0;
 374         eth->h_dest[5] = 1;
 375         eth->h_proto = htons(ETH_P_IP);
 376         skb_put(skb, sizeof(*eth));
 377
 378         skb_set_network_header(skb, skb->len);
 379         iph = ip_hdr(skb);
 380
 381         iph->version = 4;
 382         iph->ihl = 6;
 383         iph->tos = 0xc0;
 384         iph->tot_len = htons(sizeof(*iph) + sizeof(*ih) + 4);
 385         iph->id = 0;
 386         iph->frag_off = htons(IP_DF);
 387         iph->ttl = 1;
 388         iph->protocol = IPPROTO_IGMP;
 389         iph->saddr = 0;
 390         iph->daddr = htonl(INADDR_ALLHOSTS_GROUP);
 391         ((u8 *)&iph[1])[0] = IPOPT_RA;
 392         ((u8 *)&iph[1])[1] = 4;
 393         ((u8 *)&iph[1])[2] = 0;
 394         ((u8 *)&iph[1])[3] = 0;
 395         ip_send_check(iph);
 396         skb_put(skb, 24);
 397
 398         skb_set_transport_header(skb, skb->len);
 399         ih = igmp_hdr(skb);
 400         ih->type = IGMP_HOST_MEMBERSHIP_QUERY;
 401         ih->code = (group ? br->multicast_last_member_interval :
 402                             br->multicast_query_response_interval) /
 403                    (HZ / IGMP_TIMER_SCALE);
 404         ih->group = group;
 405         ih->csum = 0;
 406         ih->csum = ip_compute_csum((void *)ih, sizeof(struct igmphdr));
 407         skb_put(skb, sizeof(*ih));
 408
 409         __skb_pull(skb, sizeof(*eth));
 410
 411 out:
 412         return skb;
 413 }
 414
 415 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
 416 static struct sk_buff *br_ip6_multicast_alloc_query(struct net_bridge *br,
 417                                                     const struct in6_addr *group)
 418 {
 419         struct sk_buff *skb;
 420         struct ipv6hdr *ip6h;
 421         struct mld_msg *mldq;
 422         struct ethhdr *eth;
 423         u8 *hopopt;
 424         unsigned long interval;
 425
 426         skb = netdev_alloc_skb_ip_align(br->dev, sizeof(*eth) + sizeof(*ip6h) +
 427                                                  8 + sizeof(*mldq));
 428         if (!skb)
 429                 goto out;
 430
 431         skb->protocol = htons(ETH_P_IPV6);
 432
 433         /* Ethernet header */
 434         skb_reset_mac_header(skb);
 435         eth = eth_hdr(skb);
 436
 437         memcpy(eth->h_source, br->dev->dev_addr, 6);
 438         eth->h_proto = htons(ETH_P_IPV6);
 439         skb_put(skb, sizeof(*eth));
 440
 441         /* IPv6 header + HbH option */
 442         skb_set_network_header(skb, skb->len);
 443         ip6h = ipv6_hdr(skb);
 444
 445         *(__force __be32 *)ip6h = htonl(0x60000000);
 446         ip6h->payload_len = htons(8 + sizeof(*mldq));
 447         ip6h->nexthdr = IPPROTO_HOPOPTS;
 448         ip6h->hop_limit = 1;
 449         ipv6_addr_set(&ip6h->daddr, htonl(0xff020000), 0, 0, htonl(1));
 450         if (ipv6_dev_get_saddr(dev_net(br->dev), br->dev, &ip6h->daddr, 0,
 451                                &ip6h->saddr)) {
 452                 kfree_skb(skb);
 453                 return NULL;
 454         }
 455         ipv6_eth_mc_map(&ip6h->daddr, eth->h_dest);
 456
 457         hopopt = (u8 *)(ip6h + 1);
 458         hopopt[0] = IPPROTO_ICMPV6;             /* next hdr */
 459         hopopt[1] = 0;                          /* length of HbH */
 460         hopopt[2] = IPV6_TLV_ROUTERALERT;       /* Router Alert */
 461         hopopt[3] = 2;                          /* Length of RA Option */
 462         hopopt[4] = 0;                          /* Type = 0x0000 (MLD) */
 463         hopopt[5] = 0;
 464         hopopt[6] = IPV6_TLV_PAD0;              /* Pad0 */
 465         hopopt[7] = IPV6_TLV_PAD0;              /* Pad0 */
 466
 467         skb_put(skb, sizeof(*ip6h) + 8);
 468
 469         /* ICMPv6 */
 470         skb_set_transport_header(skb, skb->len);
 471         mldq = (struct mld_msg *) icmp6_hdr(skb);
 472
 473         interval = ipv6_addr_any(group) ?
 474                         br->multicast_query_response_interval :
 475                         br->multicast_last_member_interval;
 476
 477         mldq->mld_type = ICMPV6_MGM_QUERY;
 478         mldq->mld_code = 0;
 479         mldq->mld_cksum = 0;
 480         mldq->mld_maxdelay = htons((u16)jiffies_to_msecs(interval));
 481         mldq->mld_reserved = 0;
 482         ipv6_addr_copy(&mldq->mld_mca, group);
 483
 484         /* checksum */
 485         mldq->mld_cksum = csum_ipv6_magic(&ip6h->saddr, &ip6h->daddr,
 486                                           sizeof(*mldq), IPPROTO_ICMPV6,
 487                                           csum_partial(mldq,
 488                                                        sizeof(*mldq), 0));
 489         skb_put(skb, sizeof(*mldq));
 490
 491         __skb_pull(skb, sizeof(*eth));
 492
 493 out:
 494         return skb;
 495 }
 496 #endif
 497
 498 static struct sk_buff *br_multicast_alloc_query(struct net_bridge *br,
 499                                                 struct br_ip *addr)
 500 {
 501         switch (addr->proto) {
 502         case htons(ETH_P_IP):
 503                 return br_ip4_multicast_alloc_query(br, addr->u.ip4);
 504 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
 505         case htons(ETH_P_IPV6):
 506                 return br_ip6_multicast_alloc_query(br, &addr->u.ip6);
 507 #endif
 508         }
 509         return NULL;
 510 }
 511
 512 static struct net_bridge_mdb_entry *br_multicast_get_group(
 513         struct net_bridge *br, struct net_bridge_port *port,
 514         struct br_ip *group, int hash)
 515 {
 516         struct net_bridge_mdb_htable *mdb;
 517         struct net_bridge_mdb_entry *mp;
 518         struct hlist_node *p;
 519         unsigned count = 0;
 520         unsigned max;
 521         int elasticity;
 522         int err;
 523
 524         mdb = rcu_dereference_protected(br->mdb, 1);
 525         hlist_for_each_entry(mp, p, &mdb->mhash[hash], hlist[mdb->ver]) {
 526                 count++;
 527                 if (unlikely(br_ip_equal(group, &mp->addr)))
 528                         return mp;
 529         }
 530
 531         elasticity = 0;
 532         max = mdb->max;
 533
 534         if (unlikely(count > br->hash_elasticity && count)) {
 535                 if (net_ratelimit())
 536                         br_info(br, "Multicast hash table "
 537                                 "chain limit reached: %s\n",
 538                                 port ? port->dev->name : br->dev->name);
 539
 540                 elasticity = br->hash_elasticity;
 541         }
 542
 543         if (mdb->size >= max) {
 544                 max *= 2;
 545                 if (unlikely(max >= br->hash_max)) {
 546                         br_warn(br, "Multicast hash table maximum "
 547                                 "reached, disabling snooping: %s, %d\n",
 548                                 port ? port->dev->name : br->dev->name, max);
 549                         err = -E2BIG;
 550 disable:
 551                         br->multicast_disabled = 1;
 552                         goto err;
 553                 }
 554         }
 555
 556         if (max > mdb->max || elasticity) {
 557                 if (mdb->old) {
 558                         if (net_ratelimit())
 559                                 br_info(br, "Multicast hash table "
 560                                         "on fire: %s\n",
 561                                         port ? port->dev->name : br->dev->name);
 562                         err = -EEXIST;
 563                         goto err;
 564                 }
 565
 566                 err = br_mdb_rehash(&br->mdb, max, elasticity);
 567                 if (err) {
 568                         br_warn(br, "Cannot rehash multicast "
 569                                 "hash table, disabling snooping: %s, %d, %d\n",
 570                                 port ? port->dev->name : br->dev->name,
 571                                 mdb->size, err);
 572                         goto disable;
 573                 }
 574
 575                 err = -EAGAIN;
 576                 goto err;
 577         }
 578
 579         return NULL;
 580
 581 err:
 582         mp = ERR_PTR(err);
 583         return mp;
 584 }
 585
 586 static struct net_bridge_mdb_entry *br_multicast_new_group(
 587         struct net_bridge *br, struct net_bridge_port *port,
 588         struct br_ip *group)
 589 {
 590         struct net_bridge_mdb_htable *mdb;
 591         struct net_bridge_mdb_entry *mp;
 592         int hash;
 593         int err;
 594
 595         mdb = rcu_dereference_protected(br->mdb, 1);
 596         if (!mdb) {
 597                 err = br_mdb_rehash(&br->mdb, BR_HASH_SIZE, 0);
 598                 if (err)
 599                         return ERR_PTR(err);
 600                 goto rehash;
 601         }
 602
 603         hash = br_ip_hash(mdb, group);
 604         mp = br_multicast_get_group(br, port, group, hash);
 605         switch (PTR_ERR(mp)) {
 606         case 0:
 607                 break;
 608
 609         case -EAGAIN:
 610 rehash:
 611                 mdb = rcu_dereference_protected(br->mdb, 1);
 612                 hash = br_ip_hash(mdb, group);
 613                 break;
 614
 615         default:
 616                 goto out;
 617         }
 618
 619         mp = kzalloc(sizeof(*mp), GFP_ATOMIC);
 620         if (unlikely(!mp))
 621                 return ERR_PTR(-ENOMEM);
 622
 623         mp->br = br;
 624         mp->addr = *group;
 625         setup_timer(&mp->timer, br_multicast_group_expired,
 626                     (unsigned long)mp);
 627
 628         hlist_add_head_rcu(&mp->hlist[mdb->ver], &mdb->mhash[hash]);
 629         mdb->size++;
 630
 631 out:
 632         return mp;
 633 }
 634
 635 static int br_multicast_add_group(struct net_bridge *br,
 636                                   struct net_bridge_port *port,
 637                                   struct br_ip *group)
 638 {
 639         struct net_bridge_mdb_entry *mp;
 640         struct net_bridge_port_group *p;
 641         struct net_bridge_port_group __rcu **pp;
 642         unsigned long now = jiffies;
 643         int err;
 644
 645         spin_lock(&br->multicast_lock);
 646         if (!netif_running(br->dev) ||
 647             (port && port->state == BR_STATE_DISABLED))
 648                 goto out;
 649
 650         mp = br_multicast_new_group(br, port, group);
 651         err = PTR_ERR(mp);
 652         if (IS_ERR(mp))
 653                 goto err;
 654
 655         if (!port) {
 656                 mp->mglist = true;
 657                 mod_timer(&mp->timer, now + br->multicast_membership_interval);
 658                 goto out;
 659         }
 660
 661         for (pp = &mp->ports;
 662              (p = mlock_dereference(*pp, br)) != NULL;
 663              pp = &p->next) {
 664                 if (p->port == port)
 665                         goto found;
 666                 if ((unsigned long)p->port < (unsigned long)port)
 667                         break;
 668         }
 669
 670         p = kzalloc(sizeof(*p), GFP_ATOMIC);
 671         err = -ENOMEM;
 672         if (unlikely(!p))
 673                 goto err;
 674
 675         p->addr = *group;
 676         p->port = port;
 677         p->next = *pp;
 678         hlist_add_head(&p->mglist, &port->mglist);
 679         setup_timer(&p->timer, br_multicast_port_group_expired,
 680                     (unsigned long)p);
 681
 682         rcu_assign_pointer(*pp, p);
 683
 684 found:
 685         mod_timer(&p->timer, now + br->multicast_membership_interval);
 686 out:
 687         err = 0;
 688
 689 err:
 690         spin_unlock(&br->multicast_lock);
 691         return err;
 692 }
 693
 694 static int br_ip4_multicast_add_group(struct net_bridge *br,
 695                                       struct net_bridge_port *port,
 696                                       __be32 group)
 697 {
 698         struct br_ip br_group;
 699
 700         if (ipv4_is_local_multicast(group))
 701                 return 0;
 702
 703         br_group.u.ip4 = group;
 704         br_group.proto = htons(ETH_P_IP);
 705
 706         return br_multicast_add_group(br, port, &br_group);
 707 }
 708
 709 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
 710 static int br_ip6_multicast_add_group(struct net_bridge *br,
 711                                       struct net_bridge_port *port,
 712                                       const struct in6_addr *group)
 713 {
 714         struct br_ip br_group;
 715
 716         if (!ipv6_is_transient_multicast(group))
 717                 return 0;
 718
 719         ipv6_addr_copy(&br_group.u.ip6, group);
 720         br_group.proto = htons(ETH_P_IPV6);
 721
 722         return br_multicast_add_group(br, port, &br_group);
 723 }
 724 #endif
 725
 726 static void br_multicast_router_expired(unsigned long data)
 727 {
 728         struct net_bridge_port *port = (void *)data;
 729         struct net_bridge *br = port->br;
 730
 731         spin_lock(&br->multicast_lock);
 732         if (port->multicast_router != 1 ||
 733             timer_pending(&port->multicast_router_timer) ||
 734             hlist_unhashed(&port->rlist))
 735                 goto out;
 736
 737         hlist_del_init_rcu(&port->rlist);
 738
 739 out:
 740         spin_unlock(&br->multicast_lock);
 741 }
 742
 743 static void br_multicast_local_router_expired(unsigned long data)
 744 {
 745 }
 746
 747 static void __br_multicast_send_query(struct net_bridge *br,
 748                                       struct net_bridge_port *port,
 749                                       struct br_ip *ip)
 750 {
 751         struct sk_buff *skb;
 752
 753         skb = br_multicast_alloc_query(br, ip);
 754         if (!skb)
 755                 return;
 756
 757         if (port) {
 758                 __skb_push(skb, sizeof(struct ethhdr));
 759                 skb->dev = port->dev;
 760                 NF_HOOK(NFPROTO_BRIDGE, NF_BR_LOCAL_OUT, skb, NULL, skb->dev,
 761                         dev_queue_xmit);
 762         } else
 763                 netif_rx(skb);
 764 }
 765
 766 static void br_multicast_send_query(struct net_bridge *br,
 767                                     struct net_bridge_port *port, u32 sent)
 768 {
 769         unsigned long time;
 770         struct br_ip br_group;
 771
 772         if (!netif_running(br->dev) || br->multicast_disabled ||
 773             timer_pending(&br->multicast_querier_timer))
 774                 return;
 775
 776         memset(&br_group.u, 0, sizeof(br_group.u));
 777
 778         br_group.proto = htons(ETH_P_IP);
 779         __br_multicast_send_query(br, port, &br_group);
 780
 781 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
 782         br_group.proto = htons(ETH_P_IPV6);
 783         __br_multicast_send_query(br, port, &br_group);
 784 #endif
 785
 786         time = jiffies;
 787         time += sent < br->multicast_startup_query_count ?
 788                 br->multicast_startup_query_interval :
 789                 br->multicast_query_interval;
 790         mod_timer(port ? &port->multicast_query_timer :
 791                          &br->multicast_query_timer, time);
 792 }
 793
 794 static void br_multicast_port_query_expired(unsigned long data)
 795 {
 796         struct net_bridge_port *port = (void *)data;
 797         struct net_bridge *br = port->br;
 798
 799         spin_lock(&br->multicast_lock);
 800         if (port->state == BR_STATE_DISABLED ||
 801             port->state == BR_STATE_BLOCKING)
 802                 goto out;
 803
 804         if (port->multicast_startup_queries_sent <
 805             br->multicast_startup_query_count)
 806                 port->multicast_startup_queries_sent++;
 807
 808         br_multicast_send_query(port->br, port,
 809                                 port->multicast_startup_queries_sent);
 810
 811 out:
 812         spin_unlock(&br->multicast_lock);
 813 }
 814
 815 void br_multicast_add_port(struct net_bridge_port *port)
 816 {
 817         port->multicast_router = 1;
 818
 819         setup_timer(&port->multicast_router_timer, br_multicast_router_expired,
 820                     (unsigned long)port);
 821         setup_timer(&port->multicast_query_timer,
 822                     br_multicast_port_query_expired, (unsigned long)port);
 823 }
 824
 825 void br_multicast_del_port(struct net_bridge_port *port)
 826 {
 827         del_timer_sync(&port->multicast_router_timer);
 828 }
 829
 830 static void __br_multicast_enable_port(struct net_bridge_port *port)
 831 {
 832         port->multicast_startup_queries_sent = 0;
 833
 834         if (try_to_del_timer_sync(&port->multicast_query_timer) >= 0 ||
 835             del_timer(&port->multicast_query_timer))
 836                 mod_timer(&port->multicast_query_timer, jiffies);
 837 }
 838
 839 void br_multicast_enable_port(struct net_bridge_port *port)
 840 {
 841         struct net_bridge *br = port->br;
 842
 843         spin_lock(&br->multicast_lock);
 844         if (br->multicast_disabled || !netif_running(br->dev))
 845                 goto out;
 846
 847         __br_multicast_enable_port(port);
 848         if (port->multicast_router == 2 && hlist_unhashed(&port->rlist))
 849                 br_multicast_add_router(br, port);
 850
 851 out:
 852         spin_unlock(&br->multicast_lock);
 853 }
 854
 855 void br_multicast_disable_port(struct net_bridge_port *port)
 856 {
 857         struct net_bridge *br = port->br;
 858         struct net_bridge_port_group *pg;
 859         struct hlist_node *p, *n;
 860
 861         spin_lock(&br->multicast_lock);
 862         hlist_for_each_entry_safe(pg, p, n, &port->mglist, mglist)
 863                 br_multicast_del_pg(br, pg);
 864
 865         if (!hlist_unhashed(&port->rlist))
 866                 hlist_del_init_rcu(&port->rlist);
 867         del_timer(&port->multicast_router_timer);
 868         del_timer(&port->multicast_query_timer);
 869         spin_unlock(&br->multicast_lock);
 870 }
 871
 872 static int br_ip4_multicast_igmp3_report(struct net_bridge *br,
 873                                          struct net_bridge_port *port,
 874                                          struct sk_buff *skb)
 875 {
 876         struct igmpv3_report *ih;
 877         struct igmpv3_grec *grec;
 878         int i;
 879         int len;
 880         int num;
 881         int type;
 882         int err = 0;
 883         __be32 group;
 884
 885         if (!pskb_may_pull(skb, sizeof(*ih)))
 886                 return -EINVAL;
 887
 888         ih = igmpv3_report_hdr(skb);
 889         num = ntohs(ih->ngrec);
 890         len = sizeof(*ih);
 891
 892         for (i = 0; i < num; i++) {
 893                 len += sizeof(*grec);
 894                 if (!pskb_may_pull(skb, len))
 895                         return -EINVAL;
 896
 897                 grec = (void *)(skb->data + len - sizeof(*grec));
 898                 group = grec->grec_mca;
 899                 type = grec->grec_type;
 900
 901                 len += ntohs(grec->grec_nsrcs) * 4;
 902                 if (!pskb_may_pull(skb, len))
 903                         return -EINVAL;
 904
 905                 /* We treat this as an IGMPv2 report for now. */
 906                 switch (type) {
 907                 case IGMPV3_MODE_IS_INCLUDE:
 908                 case IGMPV3_MODE_IS_EXCLUDE:
 909                 case IGMPV3_CHANGE_TO_INCLUDE:
 910                 case IGMPV3_CHANGE_TO_EXCLUDE:
 911                 case IGMPV3_ALLOW_NEW_SOURCES:
 912                 case IGMPV3_BLOCK_OLD_SOURCES:
 913                         break;
 914
 915                 default:
 916                         continue;
 917                 }
 918
 919                 err = br_ip4_multicast_add_group(br, port, group);
 920                 if (err)
 921                         break;
 922         }
 923
 924         return err;
 925 }
 926
 927 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
 928 static int br_ip6_multicast_mld2_report(struct net_bridge *br,
 929                                         struct net_bridge_port *port,
 930                                         struct sk_buff *skb)
 931 {
 932         struct icmp6hdr *icmp6h;
 933         struct mld2_grec *grec;
 934         int i;
 935         int len;
 936         int num;
 937         int err = 0;
 938
 939         if (!pskb_may_pull(skb, sizeof(*icmp6h)))
 940                 return -EINVAL;
 941
 942         icmp6h = icmp6_hdr(skb);
 943         num = ntohs(icmp6h->icmp6_dataun.un_data16[1]);
 944         len = sizeof(*icmp6h);
 945
 946         for (i = 0; i < num; i++) {
 947                 __be16 *nsrcs, _nsrcs;
 948
 949                 nsrcs = skb_header_pointer(skb,
 950                                            len + offsetof(struct mld2_grec,
 951                                                           grec_nsrcs),
 952                                            sizeof(_nsrcs), &_nsrcs);
 953                 if (!nsrcs)
 954                         return -EINVAL;
 955
 956                 if (!pskb_may_pull(skb,
 957                                    len + sizeof(*grec) +
 958                                    sizeof(struct in6_addr) * ntohs(*nsrcs)))
 959                         return -EINVAL;
 960
 961                 grec = (struct mld2_grec *)(skb->data + len);
 962                 len += sizeof(*grec) +
 963                        sizeof(struct in6_addr) * ntohs(*nsrcs);
 964
 965                 /* We treat these as MLDv1 reports for now. */
 966                 switch (grec->grec_type) {
 967                 case MLD2_MODE_IS_INCLUDE:
 968                 case MLD2_MODE_IS_EXCLUDE:
 969                 case MLD2_CHANGE_TO_INCLUDE:
 970                 case MLD2_CHANGE_TO_EXCLUDE:
 971                 case MLD2_ALLOW_NEW_SOURCES:
 972                 case MLD2_BLOCK_OLD_SOURCES:
 973                         break;
 974
 975                 default:
 976                         continue;
 977                 }
 978
 979                 err = br_ip6_multicast_add_group(br, port, &grec->grec_mca);
 980                 if (err)
 981                         break;
 982         }
 983
 984         return err;
 985 }
 986 #endif
 987
 988 /*
 989  * Add port to rotuer_list
 990  *  list is maintained ordered by pointer value
 991  *  and locked by br->multicast_lock and RCU
 992  */
 993 static void br_multicast_add_router(struct net_bridge *br,
 994                                     struct net_bridge_port *port)
 995 {
 996         struct net_bridge_port *p;
 997         struct hlist_node *n, *slot = NULL;
 998
 999         if (!hlist_unhashed(&port->rlist))
1000                 return;
1001
1002         hlist_for_each_entry(p, n, &br->router_list, rlist) {
1003                 if ((unsigned long) port >= (unsigned long) p)
1004                         break;
1005                 slot = n;
1006         }
1007
1008         if (slot)
1009                 hlist_add_after_rcu(slot, &port->rlist);
1010         else
1011                 hlist_add_head_rcu(&port->rlist, &br->router_list);
1012 }
1013
1014 static void br_multicast_mark_router(struct net_bridge *br,
1015                                      struct net_bridge_port *port)
1016 {
1017         unsigned long now = jiffies;
1018
1019         if (!port) {
1020                 if (br->multicast_router == 1)
1021                         mod_timer(&br->multicast_router_timer,
1022                                   now + br->multicast_querier_interval);
1023                 return;
1024         }
1025
1026         if (port->multicast_router != 1)
1027                 return;
1028
1029         br_multicast_add_router(br, port);
1030
1031         mod_timer(&port->multicast_router_timer,
1032                   now + br->multicast_querier_interval);
1033 }
1034
1035 static void br_multicast_query_received(struct net_bridge *br,
1036                                         struct net_bridge_port *port,
1037                                         int saddr)
1038 {
1039         if (saddr)
1040                 mod_timer(&br->multicast_querier_timer,
1041                           jiffies + br->multicast_querier_interval);
1042         else if (timer_pending(&br->multicast_querier_timer))
1043                 return;
1044
1045         br_multicast_mark_router(br, port);
1046 }
1047
1048 static int br_ip4_multicast_query(struct net_bridge *br,
1049                                   struct net_bridge_port *port,
1050                                   struct sk_buff *skb)
1051 {
1052         const struct iphdr *iph = ip_hdr(skb);
1053         struct igmphdr *ih = igmp_hdr(skb);
1054         struct net_bridge_mdb_entry *mp;
1055         struct igmpv3_query *ih3;
1056         struct net_bridge_port_group *p;
1057         struct net_bridge_port_group __rcu **pp;
1058         unsigned long max_delay;
1059         unsigned long now = jiffies;
1060         __be32 group;
1061         int err = 0;
1062
1063         spin_lock(&br->multicast_lock);
1064         if (!netif_running(br->dev) ||
1065             (port && port->state == BR_STATE_DISABLED))
1066                 goto out;
1067
1068         br_multicast_query_received(br, port, !!iph->saddr);
1069
1070         group = ih->group;
1071
1072         if (skb->len == sizeof(*ih)) {
1073                 max_delay = ih->code * (HZ / IGMP_TIMER_SCALE);
1074
1075                 if (!max_delay) {
1076                         max_delay = 10 * HZ;
1077                         group = 0;
1078                 }
1079         } else {
1080                 if (!pskb_may_pull(skb, sizeof(struct igmpv3_query))) {
1081                         err = -EINVAL;
1082                         goto out;
1083                 }
1084
1085                 ih3 = igmpv3_query_hdr(skb);
1086                 if (ih3->nsrcs)
1087                         goto out;
1088
1089                 max_delay = ih3->code ?
1090                             IGMPV3_MRC(ih3->code) * (HZ / IGMP_TIMER_SCALE) : 1;
1091         }
1092
1093         if (!group)
1094                 goto out;
1095
1096         mp = br_mdb_ip4_get(mlock_dereference(br->mdb, br), group);
1097         if (!mp)
1098                 goto out;
1099
1100         max_delay *= br->multicast_last_member_count;
1101
1102         if (mp->mglist &&
1103             (timer_pending(&mp->timer) ?
1104              time_after(mp->timer.expires, now + max_delay) :
1105              try_to_del_timer_sync(&mp->timer) >= 0))
1106                 mod_timer(&mp->timer, now + max_delay);
1107
1108         for (pp = &mp->ports;
1109              (p = mlock_dereference(*pp, br)) != NULL;
1110              pp = &p->next) {
1111                 if (timer_pending(&p->timer) ?
1112                     time_after(p->timer.expires, now + max_delay) :
1113                     try_to_del_timer_sync(&p->timer) >= 0)
1114                         mod_timer(&p->timer, now + max_delay);
1115         }
1116
1117 out:
1118         spin_unlock(&br->multicast_lock);
1119         return err;
1120 }
1121
1122 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
1123 static int br_ip6_multicast_query(struct net_bridge *br,
1124                                   struct net_bridge_port *port,
1125                                   struct sk_buff *skb)
1126 {
1127         const struct ipv6hdr *ip6h = ipv6_hdr(skb);
1128         struct mld_msg *mld = (struct mld_msg *) icmp6_hdr(skb);
1129         struct net_bridge_mdb_entry *mp;
1130         struct mld2_query *mld2q;
1131         struct net_bridge_port_group *p;
1132         struct net_bridge_port_group __rcu **pp;
1133         unsigned long max_delay;
1134         unsigned long now = jiffies;
1135         const struct in6_addr *group = NULL;
1136         int err = 0;
1137
1138         spin_lock(&br->multicast_lock);
1139         if (!netif_running(br->dev) ||
1140             (port && port->state == BR_STATE_DISABLED))
1141                 goto out;
1142
1143         br_multicast_query_received(br, port, !ipv6_addr_any(&ip6h->saddr));
1144
1145         /* RFC2710+RFC3810 (MLDv1+MLDv2) require link-local source addresses */
1146         if (!(ipv6_addr_type(&ip6h->saddr) & IPV6_ADDR_LINKLOCAL)) {
1147                 err = -EINVAL;
1148                 goto out;
1149         }
1150
1151         if (skb->len == sizeof(*mld)) {
1152                 if (!pskb_may_pull(skb, sizeof(*mld))) {
1153                         err = -EINVAL;
1154                         goto out;
1155                 }
1156                 mld = (struct mld_msg *) icmp6_hdr(skb);
1157                 max_delay = msecs_to_jiffies(htons(mld->mld_maxdelay));
1158                 if (max_delay)
1159                         group = &mld->mld_mca;
1160         } else if (skb->len >= sizeof(*mld2q)) {
1161                 if (!pskb_may_pull(skb, sizeof(*mld2q))) {
1162                         err = -EINVAL;
1163                         goto out;
1164                 }
1165                 mld2q = (struct mld2_query *)icmp6_hdr(skb);
1166                 if (!mld2q->mld2q_nsrcs)
1167                         group = &mld2q->mld2q_mca;
1168
1169                 max_delay = max(msecs_to_jiffies(MLDV2_MRC(ntohs(mld2q->mld2q_mrc))), 1UL);
1170         }
1171
1172         if (!group)
1173                 goto out;
1174
1175         mp = br_mdb_ip6_get(mlock_dereference(br->mdb, br), group);
1176         if (!mp)
1177                 goto out;
1178
1179         max_delay *= br->multicast_last_member_count;
1180         if (mp->mglist &&
1181             (timer_pending(&mp->timer) ?
1182              time_after(mp->timer.expires, now + max_delay) :
1183              try_to_del_timer_sync(&mp->timer) >= 0))
1184                 mod_timer(&mp->timer, now + max_delay);
1185
1186         for (pp = &mp->ports;
1187              (p = mlock_dereference(*pp, br)) != NULL;
1188              pp = &p->next) {
1189                 if (timer_pending(&p->timer) ?
1190                     time_after(p->timer.expires, now + max_delay) :
1191                     try_to_del_timer_sync(&p->timer) >= 0)
1192                         mod_timer(&p->timer, now + max_delay);
1193         }
1194
1195 out:
1196         spin_unlock(&br->multicast_lock);
1197         return err;
1198 }
1199 #endif
1200
1201 static void br_multicast_leave_group(struct net_bridge *br,
1202                                      struct net_bridge_port *port,
1203                                      struct br_ip *group)
1204 {
1205         struct net_bridge_mdb_htable *mdb;
1206         struct net_bridge_mdb_entry *mp;
1207         struct net_bridge_port_group *p;
1208         unsigned long now;
1209         unsigned long time;
1210
1211         spin_lock(&br->multicast_lock);
1212         if (!netif_running(br->dev) ||
1213             (port && port->state == BR_STATE_DISABLED) ||
1214             timer_pending(&br->multicast_querier_timer))
1215                 goto out;
1216
1217         mdb = mlock_dereference(br->mdb, br);
1218         mp = br_mdb_ip_get(mdb, group);
1219         if (!mp)
1220                 goto out;
1221
1222         now = jiffies;
1223         time = now + br->multicast_last_member_count *
1224                      br->multicast_last_member_interval;
1225
1226         if (!port) {
1227                 if (mp->mglist &&
1228                     (timer_pending(&mp->timer) ?
1229                      time_after(mp->timer.expires, time) :
1230                      try_to_del_timer_sync(&mp->timer) >= 0)) {
1231                         mod_timer(&mp->timer, time);
1232                 }
1233
1234                 goto out;
1235         }
1236
1237         for (p = mlock_dereference(mp->ports, br);
1238              p != NULL;
1239              p = mlock_dereference(p->next, br)) {
1240                 if (p->port != port)
1241                         continue;
1242
1243                 if (!hlist_unhashed(&p->mglist) &&
1244                     (timer_pending(&p->timer) ?
1245                      time_after(p->timer.expires, time) :
1246                      try_to_del_timer_sync(&p->timer) >= 0)) {
1247                         mod_timer(&p->timer, time);
1248                 }
1249
1250                 break;
1251         }
1252
1253 out:
1254         spin_unlock(&br->multicast_lock);
1255 }
1256
1257 static void br_ip4_multicast_leave_group(struct net_bridge *br,
1258                                          struct net_bridge_port *port,
1259                                          __be32 group)
1260 {
1261         struct br_ip br_group;
1262
1263         if (ipv4_is_local_multicast(group))
1264                 return;
1265
1266         br_group.u.ip4 = group;
1267         br_group.proto = htons(ETH_P_IP);
1268
1269         br_multicast_leave_group(br, port, &br_group);
1270 }
1271
1272 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
1273 static void br_ip6_multicast_leave_group(struct net_bridge *br,
1274                                          struct net_bridge_port *port,
1275                                          const struct in6_addr *group)
1276 {
1277         struct br_ip br_group;
1278
1279         if (!ipv6_is_transient_multicast(group))
1280                 return;
1281
1282         ipv6_addr_copy(&br_group.u.ip6, group);
1283         br_group.proto = htons(ETH_P_IPV6);
1284
1285         br_multicast_leave_group(br, port, &br_group);
1286 }
1287 #endif
1288
1289 static int br_multicast_ipv4_rcv(struct net_bridge *br,
1290                                  struct net_bridge_port *port,
1291                                  struct sk_buff *skb)
1292 {
1293         struct sk_buff *skb2 = skb;
1294         const struct iphdr *iph;
1295         struct igmphdr *ih;
1296         unsigned len;
1297         unsigned offset;
1298         int err;
1299
1300         /* We treat OOM as packet loss for now. */
1301         if (!pskb_may_pull(skb, sizeof(*iph)))
1302                 return -EINVAL;
1303
1304         iph = ip_hdr(skb);
1305
1306         if (iph->ihl < 5 || iph->version != 4)
1307                 return -EINVAL;
1308
1309         if (!pskb_may_pull(skb, ip_hdrlen(skb)))
1310                 return -EINVAL;
1311
1312         iph = ip_hdr(skb);
1313
1314         if (unlikely(ip_fast_csum((u8 *)iph, iph->ihl)))
1315                 return -EINVAL;
1316
1317         if (iph->protocol != IPPROTO_IGMP) {
1318                 if ((iph->daddr & IGMP_LOCAL_GROUP_MASK) != IGMP_LOCAL_GROUP)
1319                         BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1320                 return 0;
1321         }
1322
1323         len = ntohs(iph->tot_len);
1324         if (skb->len < len || len < ip_hdrlen(skb))
1325                 return -EINVAL;
1326
1327         if (skb->len > len) {
1328                 skb2 = skb_clone(skb, GFP_ATOMIC);
1329                 if (!skb2)
1330                         return -ENOMEM;
1331
1332                 err = pskb_trim_rcsum(skb2, len);
1333                 if (err)
1334                         goto err_out;
1335         }
1336
1337         len -= ip_hdrlen(skb2);
1338         offset = skb_network_offset(skb2) + ip_hdrlen(skb2);
1339         __skb_pull(skb2, offset);
1340         skb_reset_transport_header(skb2);
1341
1342         err = -EINVAL;
1343         if (!pskb_may_pull(skb2, sizeof(*ih)))
1344                 goto out;
1345
1346         switch (skb2->ip_summed) {
1347         case CHECKSUM_COMPLETE:
1348                 if (!csum_fold(skb2->csum))
1349                         break;
1350                 /* fall through */
1351         case CHECKSUM_NONE:
1352                 skb2->csum = 0;
1353                 if (skb_checksum_complete(skb2))
1354                         goto out;
1355         }
1356
1357         err = 0;
1358
1359         BR_INPUT_SKB_CB(skb)->igmp = 1;
1360         ih = igmp_hdr(skb2);
1361
1362         switch (ih->type) {
1363         case IGMP_HOST_MEMBERSHIP_REPORT:
1364         case IGMPV2_HOST_MEMBERSHIP_REPORT:
1365                 BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1366                 err = br_ip4_multicast_add_group(br, port, ih->group);
1367                 break;
1368         case IGMPV3_HOST_MEMBERSHIP_REPORT:
1369                 err = br_ip4_multicast_igmp3_report(br, port, skb2);
1370                 break;
1371         case IGMP_HOST_MEMBERSHIP_QUERY:
1372                 err = br_ip4_multicast_query(br, port, skb2);
1373                 break;
1374         case IGMP_HOST_LEAVE_MESSAGE:
1375                 br_ip4_multicast_leave_group(br, port, ih->group);
1376                 break;
1377         }
1378
1379 out:
1380         __skb_push(skb2, offset);
1381 err_out:
1382         if (skb2 != skb)
1383                 kfree_skb(skb2);
1384         return err;
1385 }
1386
1387 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
1388 static int br_multicast_ipv6_rcv(struct net_bridge *br,
1389                                  struct net_bridge_port *port,
1390                                  struct sk_buff *skb)
1391 {
1392         struct sk_buff *skb2;
1393         const struct ipv6hdr *ip6h;
1394         u8 icmp6_type;
1395         u8 nexthdr;
1396         unsigned len;
1397         int offset;
1398         int err;
1399
1400         if (!pskb_may_pull(skb, sizeof(*ip6h)))
1401                 return -EINVAL;
1402
1403         ip6h = ipv6_hdr(skb);
1404
1405         /*
1406          * We're interested in MLD messages only.
1407          *  - Version is 6
1408          *  - MLD has always Router Alert hop-by-hop option
1409          *  - But we do not support jumbrograms.
1410          */
1411         if (ip6h->version != 6 ||
1412             ip6h->nexthdr != IPPROTO_HOPOPTS ||
1413             ip6h->payload_len == 0)
1414                 return 0;
1415
1416         len = ntohs(ip6h->payload_len) + sizeof(*ip6h);
1417         if (skb->len < len)
1418                 return -EINVAL;
1419
1420         nexthdr = ip6h->nexthdr;
1421         offset = ipv6_skip_exthdr(skb, sizeof(*ip6h), &nexthdr);
1422
1423         if (offset < 0 || nexthdr != IPPROTO_ICMPV6)
1424                 return 0;
1425
1426         /* Okay, we found ICMPv6 header */
1427         skb2 = skb_clone(skb, GFP_ATOMIC);
1428         if (!skb2)
1429                 return -ENOMEM;
1430
1431         err = -EINVAL;
1432         if (!pskb_may_pull(skb2, offset + sizeof(struct icmp6hdr)))
1433                 goto out;
1434
1435         len -= offset - skb_network_offset(skb2);
1436
1437         __skb_pull(skb2, offset);
1438         skb_reset_transport_header(skb2);
1439         skb_postpull_rcsum(skb2, skb_network_header(skb2),
1440                            skb_network_header_len(skb2));
1441
1442         icmp6_type = icmp6_hdr(skb2)->icmp6_type;
1443
1444         switch (icmp6_type) {
1445         case ICMPV6_MGM_QUERY:
1446         case ICMPV6_MGM_REPORT:
1447         case ICMPV6_MGM_REDUCTION:
1448         case ICMPV6_MLD2_REPORT:
1449                 break;
1450         default:
1451                 err = 0;
1452                 goto out;
1453         }
1454
1455         /* Okay, we found MLD message. Check further. */
1456         if (skb2->len > len) {
1457                 err = pskb_trim_rcsum(skb2, len);
1458                 if (err)
1459                         goto out;
1460                 err = -EINVAL;
1461         }
1462
1463         ip6h = ipv6_hdr(skb2);
1464
1465         switch (skb2->ip_summed) {
1466         case CHECKSUM_COMPLETE:
1467                 if (!csum_ipv6_magic(&ip6h->saddr, &ip6h->daddr, skb2->len,
1468                                         IPPROTO_ICMPV6, skb2->csum))
1469                         break;
1470                 /*FALLTHROUGH*/
1471         case CHECKSUM_NONE:
1472                 skb2->csum = ~csum_unfold(csum_ipv6_magic(&ip6h->saddr,
1473                                                         &ip6h->daddr,
1474                                                         skb2->len,
1475                                                         IPPROTO_ICMPV6, 0));
1476                 if (__skb_checksum_complete(skb2))
1477                         goto out;
1478         }
1479
1480         err = 0;
1481
1482         BR_INPUT_SKB_CB(skb)->igmp = 1;
1483
1484         switch (icmp6_type) {
1485         case ICMPV6_MGM_REPORT:
1486             {
1487                 struct mld_msg *mld;
1488                 if (!pskb_may_pull(skb2, sizeof(*mld))) {
1489                         err = -EINVAL;
1490                         goto out;
1491                 }
1492                 mld = (struct mld_msg *)skb_transport_header(skb2);
1493                 BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1494                 err = br_ip6_multicast_add_group(br, port, &mld->mld_mca);
1495                 break;
1496             }
1497         case ICMPV6_MLD2_REPORT:
1498                 err = br_ip6_multicast_mld2_report(br, port, skb2);
1499                 break;
1500         case ICMPV6_MGM_QUERY:
1501                 err = br_ip6_multicast_query(br, port, skb2);
1502                 break;
1503         case ICMPV6_MGM_REDUCTION:
1504             {
1505                 struct mld_msg *mld;
1506                 if (!pskb_may_pull(skb2, sizeof(*mld))) {
1507                         err = -EINVAL;
1508                         goto out;
1509                 }
1510                 mld = (struct mld_msg *)skb_transport_header(skb2);
1511                 br_ip6_multicast_leave_group(br, port, &mld->mld_mca);
1512             }
1513         }
1514
1515 out:
1516         kfree_skb(skb2);
1517         return err;
1518 }
1519 #endif
1520
1521 int br_multicast_rcv(struct net_bridge *br, struct net_bridge_port *port,
1522                      struct sk_buff *skb)
1523 {
1524         BR_INPUT_SKB_CB(skb)->igmp = 0;
1525         BR_INPUT_SKB_CB(skb)->mrouters_only = 0;
1526
1527         if (br->multicast_disabled)
1528                 return 0;
1529
1530         switch (skb->protocol) {
1531         case htons(ETH_P_IP):
1532                 return br_multicast_ipv4_rcv(br, port, skb);
1533 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
1534         case htons(ETH_P_IPV6):
1535                 return br_multicast_ipv6_rcv(br, port, skb);
1536 #endif
1537         }
1538
1539         return 0;
1540 }
1541
1542 static void br_multicast_query_expired(unsigned long data)
1543 {
1544         struct net_bridge *br = (void *)data;
1545
1546         spin_lock(&br->multicast_lock);
1547         if (br->multicast_startup_queries_sent <
1548             br->multicast_startup_query_count)
1549                 br->multicast_startup_queries_sent++;
1550
1551         br_multicast_send_query(br, NULL, br->multicast_startup_queries_sent);
1552
1553         spin_unlock(&br->multicast_lock);
1554 }
1555
1556 void br_multicast_init(struct net_bridge *br)
1557 {
1558         br->hash_elasticity = 4;
1559         br->hash_max = 512;
1560
1561         br->multicast_router = 1;
1562         br->multicast_last_member_count = 2;
1563         br->multicast_startup_query_count = 2;
1564
1565         br->multicast_last_member_interval = HZ;
1566         br->multicast_query_response_interval = 10 * HZ;
1567         br->multicast_startup_query_interval = 125 * HZ / 4;
1568         br->multicast_query_interval = 125 * HZ;
1569         br->multicast_querier_interval = 255 * HZ;
1570         br->multicast_membership_interval = 260 * HZ;
1571
1572         spin_lock_init(&br->multicast_lock);
1573         setup_timer(&br->multicast_router_timer,
1574                     br_multicast_local_router_expired, 0);
1575         setup_timer(&br->multicast_querier_timer,
1576                     br_multicast_local_router_expired, 0);
1577         setup_timer(&br->multicast_query_timer, br_multicast_query_expired,
1578                     (unsigned long)br);
1579 }
1580
1581 void br_multicast_open(struct net_bridge *br)
1582 {
1583         br->multicast_startup_queries_sent = 0;
1584
1585         if (br->multicast_disabled)
1586                 return;
1587
1588         mod_timer(&br->multicast_query_timer, jiffies);
1589 }
1590
1591 void br_multicast_stop(struct net_bridge *br)
1592 {
1593         struct net_bridge_mdb_htable *mdb;
1594         struct net_bridge_mdb_entry *mp;
1595         struct hlist_node *p, *n;
1596         u32 ver;
1597         int i;
1598
1599         del_timer_sync(&br->multicast_router_timer);
1600         del_timer_sync(&br->multicast_querier_timer);
1601         del_timer_sync(&br->multicast_query_timer);
1602
1603         spin_lock_bh(&br->multicast_lock);
1604         mdb = mlock_dereference(br->mdb, br);
1605         if (!mdb)
1606                 goto out;
1607
1608         br->mdb = NULL;
1609
1610         ver = mdb->ver;
1611         for (i = 0; i < mdb->max; i++) {
1612                 hlist_for_each_entry_safe(mp, p, n, &mdb->mhash[i],
1613                                           hlist[ver]) {
1614                         del_timer(&mp->timer);
1615                         call_rcu_bh(&mp->rcu, br_multicast_free_group);
1616                 }
1617         }
1618
1619         if (mdb->old) {
1620                 spin_unlock_bh(&br->multicast_lock);
1621                 rcu_barrier_bh();
1622                 spin_lock_bh(&br->multicast_lock);
1623                 WARN_ON(mdb->old);
1624         }
1625
1626         mdb->old = mdb;
1627         call_rcu_bh(&mdb->rcu, br_mdb_free);
1628
1629 out:
1630         spin_unlock_bh(&br->multicast_lock);
1631 }
1632
1633 int br_multicast_set_router(struct net_bridge *br, unsigned long val)
1634 {
1635         int err = -ENOENT;
1636
1637         spin_lock_bh(&br->multicast_lock);
1638         if (!netif_running(br->dev))
1639                 goto unlock;
1640
1641         switch (val) {
1642         case 0:
1643         case 2:
1644                 del_timer(&br->multicast_router_timer);
1645                 /* fall through */
1646         case 1:
1647                 br->multicast_router = val;
1648                 err = 0;
1649                 break;
1650
1651         default:
1652                 err = -EINVAL;
1653                 break;
1654         }
1655
1656 unlock:
1657         spin_unlock_bh(&br->multicast_lock);
1658
1659         return err;
1660 }
1661
1662 int br_multicast_set_port_router(struct net_bridge_port *p, unsigned long val)
1663 {
1664         struct net_bridge *br = p->br;
1665         int err = -ENOENT;
1666
1667         spin_lock(&br->multicast_lock);
1668         if (!netif_running(br->dev) || p->state == BR_STATE_DISABLED)
1669                 goto unlock;
1670
1671         switch (val) {
1672         case 0:
1673         case 1:
1674         case 2:
1675                 p->multicast_router = val;
1676                 err = 0;
1677
1678                 if (val < 2 && !hlist_unhashed(&p->rlist))
1679                         hlist_del_init_rcu(&p->rlist);
1680
1681                 if (val == 1)
1682                         break;
1683
1684                 del_timer(&p->multicast_router_timer);
1685
1686                 if (val == 0)
1687                         break;
1688
1689                 br_multicast_add_router(br, p);
1690                 break;
1691
1692         default:
1693                 err = -EINVAL;
1694                 break;
1695         }
1696
1697 unlock:
1698         spin_unlock(&br->multicast_lock);
1699
1700         return err;
1701 }
1702
1703 int br_multicast_toggle(struct net_bridge *br, unsigned long val)
1704 {
1705         struct net_bridge_port *port;
1706         int err = 0;
1707         struct net_bridge_mdb_htable *mdb;
1708
1709         spin_lock_bh(&br->multicast_lock);
1710         if (br->multicast_disabled == !val)
1711                 goto unlock;
1712
1713         br->multicast_disabled = !val;
1714         if (br->multicast_disabled)
1715                 goto unlock;
1716
1717         if (!netif_running(br->dev))
1718                 goto unlock;
1719
1720         mdb = mlock_dereference(br->mdb, br);
1721         if (mdb) {
1722                 if (mdb->old) {
1723                         err = -EEXIST;
1724 rollback:
1725                         br->multicast_disabled = !!val;
1726                         goto unlock;
1727                 }
1728
1729                 err = br_mdb_rehash(&br->mdb, mdb->max,
1730                                     br->hash_elasticity);
1731                 if (err)
1732                         goto rollback;
1733         }
1734
1735         br_multicast_open(br);
1736         list_for_each_entry(port, &br->port_list, list) {
1737                 if (port->state == BR_STATE_DISABLED ||
1738                     port->state == BR_STATE_BLOCKING)
1739                         continue;
1740
1741                 __br_multicast_enable_port(port);
1742         }
1743
1744 unlock:
1745         spin_unlock_bh(&br->multicast_lock);
1746
1747         return err;
1748 }
1749
1750 int br_multicast_set_hash_max(struct net_bridge *br, unsigned long val)
1751 {
1752         int err = -ENOENT;
1753         u32 old;
1754         struct net_bridge_mdb_htable *mdb;
1755
1756         spin_lock_bh(&br->multicast_lock);
1757         if (!netif_running(br->dev))
1758                 goto unlock;
1759
1760         err = -EINVAL;
1761         if (!is_power_of_2(val))
1762                 goto unlock;
1763
1764         mdb = mlock_dereference(br->mdb, br);
1765         if (mdb && val < mdb->size)
1766                 goto unlock;
1767
1768         err = 0;
1769
1770         old = br->hash_max;
1771         br->hash_max = val;
1772
1773         if (mdb) {
1774                 if (mdb->old) {
1775                         err = -EEXIST;
1776 rollback:
1777                         br->hash_max = old;
1778                         goto unlock;
1779                 }
1780
1781                 err = br_mdb_rehash(&br->mdb, br->hash_max,
1782                                     br->hash_elasticity);
1783                 if (err)
1784                         goto rollback;
1785         }
1786
1787 unlock:
1788         spin_unlock_bh(&br->multicast_lock);
1789
1790         return err;
1791 }