net/bridge/br_multicast.c

   1 /*
   2  * Bridge multicast support.
   3  *
   4  * Copyright (c) 2010 Herbert Xu <herbert@gondor.apana.org.au>
   5  *
   6  * This program is free software; you can redistribute it and/or modify it
   7  * under the terms of the GNU General Public License as published by the Free
   8  * Software Foundation; either version 2 of the License, or (at your option)
   9  * any later version.
  10  *
  11  */
  12
  13 #include <linux/err.h>
  14 #include <linux/if_ether.h>
  15 #include <linux/igmp.h>
  16 #include <linux/jhash.h>
  17 #include <linux/kernel.h>
  18 #include <linux/log2.h>
  19 #include <linux/netdevice.h>
  20 #include <linux/netfilter_bridge.h>
  21 #include <linux/random.h>
  22 #include <linux/rculist.h>
  23 #include <linux/skbuff.h>
  24 #include <linux/slab.h>
  25 #include <linux/timer.h>
  26 #include <net/ip.h>
  27 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
  28 #include <net/ipv6.h>
  29 #include <net/mld.h>
  30 #include <net/addrconf.h>
  31 #include <net/ip6_checksum.h>
  32 #endif
  33
  34 #include "br_private.h"
  35
  36 #define mlock_dereference(X, br) \
  37         rcu_dereference_protected(X, lockdep_is_held(&br->multicast_lock))
  38
  39 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
  40 static inline int ipv6_is_local_multicast(const struct in6_addr *addr)
  41 {
  42         if (ipv6_addr_is_multicast(addr) &&
  43             IPV6_ADDR_MC_SCOPE(addr) <= IPV6_ADDR_SCOPE_LINKLOCAL)
  44                 return 1;
  45         return 0;
  46 }
  47 #endif
  48
  49 static inline int br_ip_equal(const struct br_ip *a, const struct br_ip *b)
  50 {
  51         if (a->proto != b->proto)
  52                 return 0;
  53         switch (a->proto) {
  54         case htons(ETH_P_IP):
  55                 return a->u.ip4 == b->u.ip4;
  56 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
  57         case htons(ETH_P_IPV6):
  58                 return ipv6_addr_equal(&a->u.ip6, &b->u.ip6);
  59 #endif
  60         }
  61         return 0;
  62 }
  63
  64 static inline int __br_ip4_hash(struct net_bridge_mdb_htable *mdb, __be32 ip)
  65 {
  66         return jhash_1word(mdb->secret, (__force u32)ip) & (mdb->max - 1);
  67 }
  68
  69 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
  70 static inline int __br_ip6_hash(struct net_bridge_mdb_htable *mdb,
  71                                 const struct in6_addr *ip)
  72 {
  73         return jhash2((__force u32 *)ip->s6_addr32, 4, mdb->secret) & (mdb->max - 1);
  74 }
  75 #endif
  76
  77 static inline int br_ip_hash(struct net_bridge_mdb_htable *mdb,
  78                              struct br_ip *ip)
  79 {
  80         switch (ip->proto) {
  81         case htons(ETH_P_IP):
  82                 return __br_ip4_hash(mdb, ip->u.ip4);
  83 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
  84         case htons(ETH_P_IPV6):
  85                 return __br_ip6_hash(mdb, &ip->u.ip6);
  86 #endif
  87         }
  88         return 0;
  89 }
  90
  91 static struct net_bridge_mdb_entry *__br_mdb_ip_get(
  92         struct net_bridge_mdb_htable *mdb, struct br_ip *dst, int hash)
  93 {
  94         struct net_bridge_mdb_entry *mp;
  95         struct hlist_node *p;
  96
  97         hlist_for_each_entry_rcu(mp, p, &mdb->mhash[hash], hlist[mdb->ver]) {
  98                 if (br_ip_equal(&mp->addr, dst))
  99                         return mp;
 100         }
 101
 102         return NULL;
 103 }
 104
 105 static struct net_bridge_mdb_entry *br_mdb_ip_get(
 106         struct net_bridge_mdb_htable *mdb, struct br_ip *dst)
 107 {
 108         if (!mdb)
 109                 return NULL;
 110
 111         return __br_mdb_ip_get(mdb, dst, br_ip_hash(mdb, dst));
 112 }
 113
 114 static struct net_bridge_mdb_entry *br_mdb_ip4_get(
 115         struct net_bridge_mdb_htable *mdb, __be32 dst)
 116 {
 117         struct br_ip br_dst;
 118
 119         br_dst.u.ip4 = dst;
 120         br_dst.proto = htons(ETH_P_IP);
 121
 122         return br_mdb_ip_get(mdb, &br_dst);
 123 }
 124
 125 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
 126 static struct net_bridge_mdb_entry *br_mdb_ip6_get(
 127         struct net_bridge_mdb_htable *mdb, const struct in6_addr *dst)
 128 {
 129         struct br_ip br_dst;
 130
 131         ipv6_addr_copy(&br_dst.u.ip6, dst);
 132         br_dst.proto = htons(ETH_P_IPV6);
 133
 134         return br_mdb_ip_get(mdb, &br_dst);
 135 }
 136 #endif
 137
 138 struct net_bridge_mdb_entry *br_mdb_get(struct net_bridge *br,
 139                                         struct sk_buff *skb)
 140 {
 141         struct net_bridge_mdb_htable *mdb = rcu_dereference(br->mdb);
 142         struct br_ip ip;
 143
 144         if (br->multicast_disabled)
 145                 return NULL;
 146
 147         if (BR_INPUT_SKB_CB(skb)->igmp)
 148                 return NULL;
 149
 150         ip.proto = skb->protocol;
 151
 152         switch (skb->protocol) {
 153         case htons(ETH_P_IP):
 154                 ip.u.ip4 = ip_hdr(skb)->daddr;
 155                 break;
 156 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
 157         case htons(ETH_P_IPV6):
 158                 ipv6_addr_copy(&ip.u.ip6, &ipv6_hdr(skb)->daddr);
 159                 break;
 160 #endif
 161         default:
 162                 return NULL;
 163         }
 164
 165         return br_mdb_ip_get(mdb, &ip);
 166 }
 167
 168 static void br_mdb_free(struct rcu_head *head)
 169 {
 170         struct net_bridge_mdb_htable *mdb =
 171                 container_of(head, struct net_bridge_mdb_htable, rcu);
 172         struct net_bridge_mdb_htable *old = mdb->old;
 173
 174         mdb->old = NULL;
 175         kfree(old->mhash);
 176         kfree(old);
 177 }
 178
 179 static int br_mdb_copy(struct net_bridge_mdb_htable *new,
 180                        struct net_bridge_mdb_htable *old,
 181                        int elasticity)
 182 {
 183         struct net_bridge_mdb_entry *mp;
 184         struct hlist_node *p;
 185         int maxlen;
 186         int len;
 187         int i;
 188
 189         for (i = 0; i < old->max; i++)
 190                 hlist_for_each_entry(mp, p, &old->mhash[i], hlist[old->ver])
 191                         hlist_add_head(&mp->hlist[new->ver],
 192                                        &new->mhash[br_ip_hash(new, &mp->addr)]);
 193
 194         if (!elasticity)
 195                 return 0;
 196
 197         maxlen = 0;
 198         for (i = 0; i < new->max; i++) {
 199                 len = 0;
 200                 hlist_for_each_entry(mp, p, &new->mhash[i], hlist[new->ver])
 201                         len++;
 202                 if (len > maxlen)
 203                         maxlen = len;
 204         }
 205
 206         return maxlen > elasticity ? -EINVAL : 0;
 207 }
 208
 209 static void br_multicast_free_pg(struct rcu_head *head)
 210 {
 211         struct net_bridge_port_group *p =
 212                 container_of(head, struct net_bridge_port_group, rcu);
 213
 214         kfree(p);
 215 }
 216
 217 static void br_multicast_free_group(struct rcu_head *head)
 218 {
 219         struct net_bridge_mdb_entry *mp =
 220                 container_of(head, struct net_bridge_mdb_entry, rcu);
 221
 222         kfree(mp);
 223 }
 224
 225 static void br_multicast_group_expired(unsigned long data)
 226 {
 227         struct net_bridge_mdb_entry *mp = (void *)data;
 228         struct net_bridge *br = mp->br;
 229         struct net_bridge_mdb_htable *mdb;
 230
 231         spin_lock(&br->multicast_lock);
 232         if (!netif_running(br->dev) || timer_pending(&mp->timer))
 233                 goto out;
 234
 235         if (!hlist_unhashed(&mp->mglist))
 236                 hlist_del_init(&mp->mglist);
 237
 238         if (mp->ports)
 239                 goto out;
 240
 241         mdb = mlock_dereference(br->mdb, br);
 242
 243         hlist_del_rcu(&mp->hlist[mdb->ver]);
 244         mdb->size--;
 245
 246         del_timer(&mp->query_timer);
 247         call_rcu_bh(&mp->rcu, br_multicast_free_group);
 248
 249 out:
 250         spin_unlock(&br->multicast_lock);
 251 }
 252
 253 static void br_multicast_del_pg(struct net_bridge *br,
 254                                 struct net_bridge_port_group *pg)
 255 {
 256         struct net_bridge_mdb_htable *mdb;
 257         struct net_bridge_mdb_entry *mp;
 258         struct net_bridge_port_group *p;
 259         struct net_bridge_port_group __rcu **pp;
 260
 261         mdb = mlock_dereference(br->mdb, br);
 262
 263         mp = br_mdb_ip_get(mdb, &pg->addr);
 264         if (WARN_ON(!mp))
 265                 return;
 266
 267         for (pp = &mp->ports;
 268              (p = mlock_dereference(*pp, br)) != NULL;
 269              pp = &p->next) {
 270                 if (p != pg)
 271                         continue;
 272
 273                 rcu_assign_pointer(*pp, p->next);
 274                 hlist_del_init(&p->mglist);
 275                 del_timer(&p->timer);
 276                 del_timer(&p->query_timer);
 277                 call_rcu_bh(&p->rcu, br_multicast_free_pg);
 278
 279                 if (!mp->ports && hlist_unhashed(&mp->mglist) &&
 280                     netif_running(br->dev))
 281                         mod_timer(&mp->timer, jiffies);
 282
 283                 return;
 284         }
 285
 286         WARN_ON(1);
 287 }
 288
 289 static void br_multicast_port_group_expired(unsigned long data)
 290 {
 291         struct net_bridge_port_group *pg = (void *)data;
 292         struct net_bridge *br = pg->port->br;
 293
 294         spin_lock(&br->multicast_lock);
 295         if (!netif_running(br->dev) || timer_pending(&pg->timer) ||
 296             hlist_unhashed(&pg->mglist))
 297                 goto out;
 298
 299         br_multicast_del_pg(br, pg);
 300
 301 out:
 302         spin_unlock(&br->multicast_lock);
 303 }
 304
 305 static int br_mdb_rehash(struct net_bridge_mdb_htable __rcu **mdbp, int max,
 306                          int elasticity)
 307 {
 308         struct net_bridge_mdb_htable *old = rcu_dereference_protected(*mdbp, 1);
 309         struct net_bridge_mdb_htable *mdb;
 310         int err;
 311
 312         mdb = kmalloc(sizeof(*mdb), GFP_ATOMIC);
 313         if (!mdb)
 314                 return -ENOMEM;
 315
 316         mdb->max = max;
 317         mdb->old = old;
 318
 319         mdb->mhash = kzalloc(max * sizeof(*mdb->mhash), GFP_ATOMIC);
 320         if (!mdb->mhash) {
 321                 kfree(mdb);
 322                 return -ENOMEM;
 323         }
 324
 325         mdb->size = old ? old->size : 0;
 326         mdb->ver = old ? old->ver ^ 1 : 0;
 327
 328         if (!old || elasticity)
 329                 get_random_bytes(&mdb->secret, sizeof(mdb->secret));
 330         else
 331                 mdb->secret = old->secret;
 332
 333         if (!old)
 334                 goto out;
 335
 336         err = br_mdb_copy(mdb, old, elasticity);
 337         if (err) {
 338                 kfree(mdb->mhash);
 339                 kfree(mdb);
 340                 return err;
 341         }
 342
 343         call_rcu_bh(&mdb->rcu, br_mdb_free);
 344
 345 out:
 346         rcu_assign_pointer(*mdbp, mdb);
 347
 348         return 0;
 349 }
 350
 351 static struct sk_buff *br_ip4_multicast_alloc_query(struct net_bridge *br,
 352                                                     __be32 group)
 353 {
 354         struct sk_buff *skb;
 355         struct igmphdr *ih;
 356         struct ethhdr *eth;
 357         struct iphdr *iph;
 358
 359         skb = netdev_alloc_skb_ip_align(br->dev, sizeof(*eth) + sizeof(*iph) +
 360                                                  sizeof(*ih) + 4);
 361         if (!skb)
 362                 goto out;
 363
 364         skb->protocol = htons(ETH_P_IP);
 365
 366         skb_reset_mac_header(skb);
 367         eth = eth_hdr(skb);
 368
 369         memcpy(eth->h_source, br->dev->dev_addr, 6);
 370         eth->h_dest[0] = 1;
 371         eth->h_dest[1] = 0;
 372         eth->h_dest[2] = 0x5e;
 373         eth->h_dest[3] = 0;
 374         eth->h_dest[4] = 0;
 375         eth->h_dest[5] = 1;
 376         eth->h_proto = htons(ETH_P_IP);
 377         skb_put(skb, sizeof(*eth));
 378
 379         skb_set_network_header(skb, skb->len);
 380         iph = ip_hdr(skb);
 381
 382         iph->version = 4;
 383         iph->ihl = 6;
 384         iph->tos = 0xc0;
 385         iph->tot_len = htons(sizeof(*iph) + sizeof(*ih) + 4);
 386         iph->id = 0;
 387         iph->frag_off = htons(IP_DF);
 388         iph->ttl = 1;
 389         iph->protocol = IPPROTO_IGMP;
 390         iph->saddr = 0;
 391         iph->daddr = htonl(INADDR_ALLHOSTS_GROUP);
 392         ((u8 *)&iph[1])[0] = IPOPT_RA;
 393         ((u8 *)&iph[1])[1] = 4;
 394         ((u8 *)&iph[1])[2] = 0;
 395         ((u8 *)&iph[1])[3] = 0;
 396         ip_send_check(iph);
 397         skb_put(skb, 24);
 398
 399         skb_set_transport_header(skb, skb->len);
 400         ih = igmp_hdr(skb);
 401         ih->type = IGMP_HOST_MEMBERSHIP_QUERY;
 402         ih->code = (group ? br->multicast_last_member_interval :
 403                             br->multicast_query_response_interval) /
 404                    (HZ / IGMP_TIMER_SCALE);
 405         ih->group = group;
 406         ih->csum = 0;
 407         ih->csum = ip_compute_csum((void *)ih, sizeof(struct igmphdr));
 408         skb_put(skb, sizeof(*ih));
 409
 410         __skb_pull(skb, sizeof(*eth));
 411
 412 out:
 413         return skb;
 414 }
 415
 416 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
 417 static struct sk_buff *br_ip6_multicast_alloc_query(struct net_bridge *br,
 418                                                     struct in6_addr *group)
 419 {
 420         struct sk_buff *skb;
 421         struct ipv6hdr *ip6h;
 422         struct mld_msg *mldq;
 423         struct ethhdr *eth;
 424         u8 *hopopt;
 425         unsigned long interval;
 426
 427         skb = netdev_alloc_skb_ip_align(br->dev, sizeof(*eth) + sizeof(*ip6h) +
 428                                                  8 + sizeof(*mldq));
 429         if (!skb)
 430                 goto out;
 431
 432         skb->protocol = htons(ETH_P_IPV6);
 433
 434         /* Ethernet header */
 435         skb_reset_mac_header(skb);
 436         eth = eth_hdr(skb);
 437
 438         memcpy(eth->h_source, br->dev->dev_addr, 6);
 439         ipv6_eth_mc_map(group, eth->h_dest);
 440         eth->h_proto = htons(ETH_P_IPV6);
 441         skb_put(skb, sizeof(*eth));
 442
 443         /* IPv6 header + HbH option */
 444         skb_set_network_header(skb, skb->len);
 445         ip6h = ipv6_hdr(skb);
 446
 447         *(__force __be32 *)ip6h = htonl(0x60000000);
 448         ip6h->payload_len = htons(8 + sizeof(*mldq));
 449         ip6h->nexthdr = IPPROTO_HOPOPTS;
 450         ip6h->hop_limit = 1;
 451         ipv6_addr_set(&ip6h->saddr, 0, 0, 0, 0);
 452         ipv6_addr_set(&ip6h->daddr, htonl(0xff020000), 0, 0, htonl(1));
 453
 454         hopopt = (u8 *)(ip6h + 1);
 455         hopopt[0] = IPPROTO_ICMPV6;             /* next hdr */
 456         hopopt[1] = 0;                          /* length of HbH */
 457         hopopt[2] = IPV6_TLV_ROUTERALERT;       /* Router Alert */
 458         hopopt[3] = 2;                          /* Length of RA Option */
 459         hopopt[4] = 0;                          /* Type = 0x0000 (MLD) */
 460         hopopt[5] = 0;
 461         hopopt[6] = IPV6_TLV_PAD0;              /* Pad0 */
 462         hopopt[7] = IPV6_TLV_PAD0;              /* Pad0 */
 463
 464         skb_put(skb, sizeof(*ip6h) + 8);
 465
 466         /* ICMPv6 */
 467         skb_set_transport_header(skb, skb->len);
 468         mldq = (struct mld_msg *) icmp6_hdr(skb);
 469
 470         interval = ipv6_addr_any(group) ? br->multicast_last_member_interval :
 471                                           br->multicast_query_response_interval;
 472
 473         mldq->mld_type = ICMPV6_MGM_QUERY;
 474         mldq->mld_code = 0;
 475         mldq->mld_cksum = 0;
 476         mldq->mld_maxdelay = htons((u16)jiffies_to_msecs(interval));
 477         mldq->mld_reserved = 0;
 478         ipv6_addr_copy(&mldq->mld_mca, group);
 479
 480         /* checksum */
 481         mldq->mld_cksum = csum_ipv6_magic(&ip6h->saddr, &ip6h->daddr,
 482                                           sizeof(*mldq), IPPROTO_ICMPV6,
 483                                           csum_partial(mldq,
 484                                                        sizeof(*mldq), 0));
 485         skb_put(skb, sizeof(*mldq));
 486
 487         __skb_pull(skb, sizeof(*eth));
 488
 489 out:
 490         return skb;
 491 }
 492 #endif
 493
 494 static struct sk_buff *br_multicast_alloc_query(struct net_bridge *br,
 495                                                 struct br_ip *addr)
 496 {
 497         switch (addr->proto) {
 498         case htons(ETH_P_IP):
 499                 return br_ip4_multicast_alloc_query(br, addr->u.ip4);
 500 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
 501         case htons(ETH_P_IPV6):
 502                 return br_ip6_multicast_alloc_query(br, &addr->u.ip6);
 503 #endif
 504         }
 505         return NULL;
 506 }
 507
 508 static void br_multicast_send_group_query(struct net_bridge_mdb_entry *mp)
 509 {
 510         struct net_bridge *br = mp->br;
 511         struct sk_buff *skb;
 512
 513         skb = br_multicast_alloc_query(br, &mp->addr);
 514         if (!skb)
 515                 goto timer;
 516
 517         netif_rx(skb);
 518
 519 timer:
 520         if (++mp->queries_sent < br->multicast_last_member_count)
 521                 mod_timer(&mp->query_timer,
 522                           jiffies + br->multicast_last_member_interval);
 523 }
 524
 525 static void br_multicast_group_query_expired(unsigned long data)
 526 {
 527         struct net_bridge_mdb_entry *mp = (void *)data;
 528         struct net_bridge *br = mp->br;
 529
 530         spin_lock(&br->multicast_lock);
 531         if (!netif_running(br->dev) || hlist_unhashed(&mp->mglist) ||
 532             mp->queries_sent >= br->multicast_last_member_count)
 533                 goto out;
 534
 535         br_multicast_send_group_query(mp);
 536
 537 out:
 538         spin_unlock(&br->multicast_lock);
 539 }
 540
 541 static void br_multicast_send_port_group_query(struct net_bridge_port_group *pg)
 542 {
 543         struct net_bridge_port *port = pg->port;
 544         struct net_bridge *br = port->br;
 545         struct sk_buff *skb;
 546
 547         skb = br_multicast_alloc_query(br, &pg->addr);
 548         if (!skb)
 549                 goto timer;
 550
 551         br_deliver(port, skb);
 552
 553 timer:
 554         if (++pg->queries_sent < br->multicast_last_member_count)
 555                 mod_timer(&pg->query_timer,
 556                           jiffies + br->multicast_last_member_interval);
 557 }
 558
 559 static void br_multicast_port_group_query_expired(unsigned long data)
 560 {
 561         struct net_bridge_port_group *pg = (void *)data;
 562         struct net_bridge_port *port = pg->port;
 563         struct net_bridge *br = port->br;
 564
 565         spin_lock(&br->multicast_lock);
 566         if (!netif_running(br->dev) || hlist_unhashed(&pg->mglist) ||
 567             pg->queries_sent >= br->multicast_last_member_count)
 568                 goto out;
 569
 570         br_multicast_send_port_group_query(pg);
 571
 572 out:
 573         spin_unlock(&br->multicast_lock);
 574 }
 575
 576 static struct net_bridge_mdb_entry *br_multicast_get_group(
 577         struct net_bridge *br, struct net_bridge_port *port,
 578         struct br_ip *group, int hash)
 579 {
 580         struct net_bridge_mdb_htable *mdb;
 581         struct net_bridge_mdb_entry *mp;
 582         struct hlist_node *p;
 583         unsigned count = 0;
 584         unsigned max;
 585         int elasticity;
 586         int err;
 587
 588         mdb = rcu_dereference_protected(br->mdb, 1);
 589         hlist_for_each_entry(mp, p, &mdb->mhash[hash], hlist[mdb->ver]) {
 590                 count++;
 591                 if (unlikely(br_ip_equal(group, &mp->addr)))
 592                         return mp;
 593         }
 594
 595         elasticity = 0;
 596         max = mdb->max;
 597
 598         if (unlikely(count > br->hash_elasticity && count)) {
 599                 if (net_ratelimit())
 600                         br_info(br, "Multicast hash table "
 601                                 "chain limit reached: %s\n",
 602                                 port ? port->dev->name : br->dev->name);
 603
 604                 elasticity = br->hash_elasticity;
 605         }
 606
 607         if (mdb->size >= max) {
 608                 max *= 2;
 609                 if (unlikely(max >= br->hash_max)) {
 610                         br_warn(br, "Multicast hash table maximum "
 611                                 "reached, disabling snooping: %s, %d\n",
 612                                 port ? port->dev->name : br->dev->name, max);
 613                         err = -E2BIG;
 614 disable:
 615                         br->multicast_disabled = 1;
 616                         goto err;
 617                 }
 618         }
 619
 620         if (max > mdb->max || elasticity) {
 621                 if (mdb->old) {
 622                         if (net_ratelimit())
 623                                 br_info(br, "Multicast hash table "
 624                                         "on fire: %s\n",
 625                                         port ? port->dev->name : br->dev->name);
 626                         err = -EEXIST;
 627                         goto err;
 628                 }
 629
 630                 err = br_mdb_rehash(&br->mdb, max, elasticity);
 631                 if (err) {
 632                         br_warn(br, "Cannot rehash multicast "
 633                                 "hash table, disabling snooping: %s, %d, %d\n",
 634                                 port ? port->dev->name : br->dev->name,
 635                                 mdb->size, err);
 636                         goto disable;
 637                 }
 638
 639                 err = -EAGAIN;
 640                 goto err;
 641         }
 642
 643         return NULL;
 644
 645 err:
 646         mp = ERR_PTR(err);
 647         return mp;
 648 }
 649
 650 static struct net_bridge_mdb_entry *br_multicast_new_group(
 651         struct net_bridge *br, struct net_bridge_port *port,
 652         struct br_ip *group)
 653 {
 654         struct net_bridge_mdb_htable *mdb;
 655         struct net_bridge_mdb_entry *mp;
 656         int hash;
 657         int err;
 658
 659         mdb = rcu_dereference_protected(br->mdb, 1);
 660         if (!mdb) {
 661                 err = br_mdb_rehash(&br->mdb, BR_HASH_SIZE, 0);
 662                 if (err)
 663                         return ERR_PTR(err);
 664                 goto rehash;
 665         }
 666
 667         hash = br_ip_hash(mdb, group);
 668         mp = br_multicast_get_group(br, port, group, hash);
 669         switch (PTR_ERR(mp)) {
 670         case 0:
 671                 break;
 672
 673         case -EAGAIN:
 674 rehash:
 675                 mdb = rcu_dereference_protected(br->mdb, 1);
 676                 hash = br_ip_hash(mdb, group);
 677                 break;
 678
 679         default:
 680                 goto out;
 681         }
 682
 683         mp = kzalloc(sizeof(*mp), GFP_ATOMIC);
 684         if (unlikely(!mp))
 685                 return ERR_PTR(-ENOMEM);
 686
 687         mp->br = br;
 688         mp->addr = *group;
 689         setup_timer(&mp->timer, br_multicast_group_expired,
 690                     (unsigned long)mp);
 691         setup_timer(&mp->query_timer, br_multicast_group_query_expired,
 692                     (unsigned long)mp);
 693
 694         hlist_add_head_rcu(&mp->hlist[mdb->ver], &mdb->mhash[hash]);
 695         mdb->size++;
 696
 697 out:
 698         return mp;
 699 }
 700
 701 static int br_multicast_add_group(struct net_bridge *br,
 702                                   struct net_bridge_port *port,
 703                                   struct br_ip *group)
 704 {
 705         struct net_bridge_mdb_entry *mp;
 706         struct net_bridge_port_group *p;
 707         struct net_bridge_port_group __rcu **pp;
 708         unsigned long now = jiffies;
 709         int err;
 710
 711         spin_lock(&br->multicast_lock);
 712         if (!netif_running(br->dev) ||
 713             (port && port->state == BR_STATE_DISABLED))
 714                 goto out;
 715
 716         mp = br_multicast_new_group(br, port, group);
 717         err = PTR_ERR(mp);
 718         if (IS_ERR(mp))
 719                 goto err;
 720
 721         if (!port) {
 722                 hlist_add_head(&mp->mglist, &br->mglist);
 723                 mod_timer(&mp->timer, now + br->multicast_membership_interval);
 724                 goto out;
 725         }
 726
 727         for (pp = &mp->ports;
 728              (p = mlock_dereference(*pp, br)) != NULL;
 729              pp = &p->next) {
 730                 if (p->port == port)
 731                         goto found;
 732                 if ((unsigned long)p->port < (unsigned long)port)
 733                         break;
 734         }
 735
 736         p = kzalloc(sizeof(*p), GFP_ATOMIC);
 737         err = -ENOMEM;
 738         if (unlikely(!p))
 739                 goto err;
 740
 741         p->addr = *group;
 742         p->port = port;
 743         p->next = *pp;
 744         hlist_add_head(&p->mglist, &port->mglist);
 745         setup_timer(&p->timer, br_multicast_port_group_expired,
 746                     (unsigned long)p);
 747         setup_timer(&p->query_timer, br_multicast_port_group_query_expired,
 748                     (unsigned long)p);
 749
 750         rcu_assign_pointer(*pp, p);
 751
 752 found:
 753         mod_timer(&p->timer, now + br->multicast_membership_interval);
 754 out:
 755         err = 0;
 756
 757 err:
 758         spin_unlock(&br->multicast_lock);
 759         return err;
 760 }
 761
 762 static int br_ip4_multicast_add_group(struct net_bridge *br,
 763                                       struct net_bridge_port *port,
 764                                       __be32 group)
 765 {
 766         struct br_ip br_group;
 767
 768         if (ipv4_is_local_multicast(group))
 769                 return 0;
 770
 771         br_group.u.ip4 = group;
 772         br_group.proto = htons(ETH_P_IP);
 773
 774         return br_multicast_add_group(br, port, &br_group);
 775 }
 776
 777 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
 778 static int br_ip6_multicast_add_group(struct net_bridge *br,
 779                                       struct net_bridge_port *port,
 780                                       const struct in6_addr *group)
 781 {
 782         struct br_ip br_group;
 783
 784         if (ipv6_is_local_multicast(group))
 785                 return 0;
 786
 787         ipv6_addr_copy(&br_group.u.ip6, group);
 788         br_group.proto = htons(ETH_P_IP);
 789
 790         return br_multicast_add_group(br, port, &br_group);
 791 }
 792 #endif
 793
 794 static void br_multicast_router_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->multicast_router != 1 ||
 801             timer_pending(&port->multicast_router_timer) ||
 802             hlist_unhashed(&port->rlist))
 803                 goto out;
 804
 805         hlist_del_init_rcu(&port->rlist);
 806
 807 out:
 808         spin_unlock(&br->multicast_lock);
 809 }
 810
 811 static void br_multicast_local_router_expired(unsigned long data)
 812 {
 813 }
 814
 815 static void __br_multicast_send_query(struct net_bridge *br,
 816                                       struct net_bridge_port *port,
 817                                       struct br_ip *ip)
 818 {
 819         struct sk_buff *skb;
 820
 821         skb = br_multicast_alloc_query(br, ip);
 822         if (!skb)
 823                 return;
 824
 825         if (port) {
 826                 __skb_push(skb, sizeof(struct ethhdr));
 827                 skb->dev = port->dev;
 828                 NF_HOOK(NFPROTO_BRIDGE, NF_BR_LOCAL_OUT, skb, NULL, skb->dev,
 829                         dev_queue_xmit);
 830         } else
 831                 netif_rx(skb);
 832 }
 833
 834 static void br_multicast_send_query(struct net_bridge *br,
 835                                     struct net_bridge_port *port, u32 sent)
 836 {
 837         unsigned long time;
 838         struct br_ip br_group;
 839
 840         if (!netif_running(br->dev) || br->multicast_disabled ||
 841             timer_pending(&br->multicast_querier_timer))
 842                 return;
 843
 844         memset(&br_group.u, 0, sizeof(br_group.u));
 845
 846         br_group.proto = htons(ETH_P_IP);
 847         __br_multicast_send_query(br, port, &br_group);
 848
 849 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
 850         br_group.proto = htons(ETH_P_IPV6);
 851         __br_multicast_send_query(br, port, &br_group);
 852 #endif
 853
 854         time = jiffies;
 855         time += sent < br->multicast_startup_query_count ?
 856                 br->multicast_startup_query_interval :
 857                 br->multicast_query_interval;
 858         mod_timer(port ? &port->multicast_query_timer :
 859                          &br->multicast_query_timer, time);
 860 }
 861
 862 static void br_multicast_port_query_expired(unsigned long data)
 863 {
 864         struct net_bridge_port *port = (void *)data;
 865         struct net_bridge *br = port->br;
 866
 867         spin_lock(&br->multicast_lock);
 868         if (port->state == BR_STATE_DISABLED ||
 869             port->state == BR_STATE_BLOCKING)
 870                 goto out;
 871
 872         if (port->multicast_startup_queries_sent <
 873             br->multicast_startup_query_count)
 874                 port->multicast_startup_queries_sent++;
 875
 876         br_multicast_send_query(port->br, port,
 877                                 port->multicast_startup_queries_sent);
 878
 879 out:
 880         spin_unlock(&br->multicast_lock);
 881 }
 882
 883 void br_multicast_add_port(struct net_bridge_port *port)
 884 {
 885         port->multicast_router = 1;
 886
 887         setup_timer(&port->multicast_router_timer, br_multicast_router_expired,
 888                     (unsigned long)port);
 889         setup_timer(&port->multicast_query_timer,
 890                     br_multicast_port_query_expired, (unsigned long)port);
 891 }
 892
 893 void br_multicast_del_port(struct net_bridge_port *port)
 894 {
 895         del_timer_sync(&port->multicast_router_timer);
 896 }
 897
 898 static void __br_multicast_enable_port(struct net_bridge_port *port)
 899 {
 900         port->multicast_startup_queries_sent = 0;
 901
 902         if (try_to_del_timer_sync(&port->multicast_query_timer) >= 0 ||
 903             del_timer(&port->multicast_query_timer))
 904                 mod_timer(&port->multicast_query_timer, jiffies);
 905 }
 906
 907 void br_multicast_enable_port(struct net_bridge_port *port)
 908 {
 909         struct net_bridge *br = port->br;
 910
 911         spin_lock(&br->multicast_lock);
 912         if (br->multicast_disabled || !netif_running(br->dev))
 913                 goto out;
 914
 915         __br_multicast_enable_port(port);
 916
 917 out:
 918         spin_unlock(&br->multicast_lock);
 919 }
 920
 921 void br_multicast_disable_port(struct net_bridge_port *port)
 922 {
 923         struct net_bridge *br = port->br;
 924         struct net_bridge_port_group *pg;
 925         struct hlist_node *p, *n;
 926
 927         spin_lock(&br->multicast_lock);
 928         hlist_for_each_entry_safe(pg, p, n, &port->mglist, mglist)
 929                 br_multicast_del_pg(br, pg);
 930
 931         if (!hlist_unhashed(&port->rlist))
 932                 hlist_del_init_rcu(&port->rlist);
 933         del_timer(&port->multicast_router_timer);
 934         del_timer(&port->multicast_query_timer);
 935         spin_unlock(&br->multicast_lock);
 936 }
 937
 938 static int br_ip4_multicast_igmp3_report(struct net_bridge *br,
 939                                          struct net_bridge_port *port,
 940                                          struct sk_buff *skb)
 941 {
 942         struct igmpv3_report *ih;
 943         struct igmpv3_grec *grec;
 944         int i;
 945         int len;
 946         int num;
 947         int type;
 948         int err = 0;
 949         __be32 group;
 950
 951         if (!pskb_may_pull(skb, sizeof(*ih)))
 952                 return -EINVAL;
 953
 954         ih = igmpv3_report_hdr(skb);
 955         num = ntohs(ih->ngrec);
 956         len = sizeof(*ih);
 957
 958         for (i = 0; i < num; i++) {
 959                 len += sizeof(*grec);
 960                 if (!pskb_may_pull(skb, len))
 961                         return -EINVAL;
 962
 963                 grec = (void *)(skb->data + len - sizeof(*grec));
 964                 group = grec->grec_mca;
 965                 type = grec->grec_type;
 966
 967                 len += ntohs(grec->grec_nsrcs) * 4;
 968                 if (!pskb_may_pull(skb, len))
 969                         return -EINVAL;
 970
 971                 /* We treat this as an IGMPv2 report for now. */
 972                 switch (type) {
 973                 case IGMPV3_MODE_IS_INCLUDE:
 974                 case IGMPV3_MODE_IS_EXCLUDE:
 975                 case IGMPV3_CHANGE_TO_INCLUDE:
 976                 case IGMPV3_CHANGE_TO_EXCLUDE:
 977                 case IGMPV3_ALLOW_NEW_SOURCES:
 978                 case IGMPV3_BLOCK_OLD_SOURCES:
 979                         break;
 980
 981                 default:
 982                         continue;
 983                 }
 984
 985                 err = br_ip4_multicast_add_group(br, port, group);
 986                 if (err)
 987                         break;
 988         }
 989
 990         return err;
 991 }
 992
 993 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
 994 static int br_ip6_multicast_mld2_report(struct net_bridge *br,
 995                                         struct net_bridge_port *port,
 996                                         struct sk_buff *skb)
 997 {
 998         struct icmp6hdr *icmp6h;
 999         struct mld2_grec *grec;
1000         int i;
1001         int len;
1002         int num;
1003         int err = 0;
1004
1005         if (!pskb_may_pull(skb, sizeof(*icmp6h)))
1006                 return -EINVAL;
1007
1008         icmp6h = icmp6_hdr(skb);
1009         num = ntohs(icmp6h->icmp6_dataun.un_data16[1]);
1010         len = sizeof(*icmp6h);
1011
1012         for (i = 0; i < num; i++) {
1013                 __be16 *nsrcs, _nsrcs;
1014
1015                 nsrcs = skb_header_pointer(skb,
1016                                            len + offsetof(struct mld2_grec,
1017                                                           grec_mca),
1018                                            sizeof(_nsrcs), &_nsrcs);
1019                 if (!nsrcs)
1020                         return -EINVAL;
1021
1022                 if (!pskb_may_pull(skb,
1023                                    len + sizeof(*grec) +
1024                                    sizeof(struct in6_addr) * (*nsrcs)))
1025                         return -EINVAL;
1026
1027                 grec = (struct mld2_grec *)(skb->data + len);
1028                 len += sizeof(*grec) + sizeof(struct in6_addr) * (*nsrcs);
1029
1030                 /* We treat these as MLDv1 reports for now. */
1031                 switch (grec->grec_type) {
1032                 case MLD2_MODE_IS_INCLUDE:
1033                 case MLD2_MODE_IS_EXCLUDE:
1034                 case MLD2_CHANGE_TO_INCLUDE:
1035                 case MLD2_CHANGE_TO_EXCLUDE:
1036                 case MLD2_ALLOW_NEW_SOURCES:
1037                 case MLD2_BLOCK_OLD_SOURCES:
1038                         break;
1039
1040                 default:
1041                         continue;
1042                 }
1043
1044                 err = br_ip6_multicast_add_group(br, port, &grec->grec_mca);
1045                 if (!err)
1046                         break;
1047         }
1048
1049         return err;
1050 }
1051 #endif
1052
1053 /*
1054  * Add port to rotuer_list
1055  *  list is maintained ordered by pointer value
1056  *  and locked by br->multicast_lock and RCU
1057  */
1058 static void br_multicast_add_router(struct net_bridge *br,
1059                                     struct net_bridge_port *port)
1060 {
1061         struct net_bridge_port *p;
1062         struct hlist_node *n, *slot = NULL;
1063
1064         hlist_for_each_entry(p, n, &br->router_list, rlist) {
1065                 if ((unsigned long) port >= (unsigned long) p)
1066                         break;
1067                 slot = n;
1068         }
1069
1070         if (slot)
1071                 hlist_add_after_rcu(slot, &port->rlist);
1072         else
1073                 hlist_add_head_rcu(&port->rlist, &br->router_list);
1074 }
1075
1076 static void br_multicast_mark_router(struct net_bridge *br,
1077                                      struct net_bridge_port *port)
1078 {
1079         unsigned long now = jiffies;
1080
1081         if (!port) {
1082                 if (br->multicast_router == 1)
1083                         mod_timer(&br->multicast_router_timer,
1084                                   now + br->multicast_querier_interval);
1085                 return;
1086         }
1087
1088         if (port->multicast_router != 1)
1089                 return;
1090
1091         if (!hlist_unhashed(&port->rlist))
1092                 goto timer;
1093
1094         br_multicast_add_router(br, port);
1095
1096 timer:
1097         mod_timer(&port->multicast_router_timer,
1098                   now + br->multicast_querier_interval);
1099 }
1100
1101 static void br_multicast_query_received(struct net_bridge *br,
1102                                         struct net_bridge_port *port,
1103                                         int saddr)
1104 {
1105         if (saddr)
1106                 mod_timer(&br->multicast_querier_timer,
1107                           jiffies + br->multicast_querier_interval);
1108         else if (timer_pending(&br->multicast_querier_timer))
1109                 return;
1110
1111         br_multicast_mark_router(br, port);
1112 }
1113
1114 static int br_ip4_multicast_query(struct net_bridge *br,
1115                                   struct net_bridge_port *port,
1116                                   struct sk_buff *skb)
1117 {
1118         struct iphdr *iph = ip_hdr(skb);
1119         struct igmphdr *ih = igmp_hdr(skb);
1120         struct net_bridge_mdb_entry *mp;
1121         struct igmpv3_query *ih3;
1122         struct net_bridge_port_group *p;
1123         struct net_bridge_port_group __rcu **pp;
1124         unsigned long max_delay;
1125         unsigned long now = jiffies;
1126         __be32 group;
1127         int err = 0;
1128
1129         spin_lock(&br->multicast_lock);
1130         if (!netif_running(br->dev) ||
1131             (port && port->state == BR_STATE_DISABLED))
1132                 goto out;
1133
1134         br_multicast_query_received(br, port, !!iph->saddr);
1135
1136         group = ih->group;
1137
1138         if (skb->len == sizeof(*ih)) {
1139                 max_delay = ih->code * (HZ / IGMP_TIMER_SCALE);
1140
1141                 if (!max_delay) {
1142                         max_delay = 10 * HZ;
1143                         group = 0;
1144                 }
1145         } else {
1146                 if (!pskb_may_pull(skb, sizeof(struct igmpv3_query))) {
1147                         err = -EINVAL;
1148                         goto out;
1149                 }
1150
1151                 ih3 = igmpv3_query_hdr(skb);
1152                 if (ih3->nsrcs)
1153                         goto out;
1154
1155                 max_delay = ih3->code ?
1156                             IGMPV3_MRC(ih3->code) * (HZ / IGMP_TIMER_SCALE) : 1;
1157         }
1158
1159         if (!group)
1160                 goto out;
1161
1162         mp = br_mdb_ip4_get(mlock_dereference(br->mdb, br), group);
1163         if (!mp)
1164                 goto out;
1165
1166         max_delay *= br->multicast_last_member_count;
1167
1168         if (!hlist_unhashed(&mp->mglist) &&
1169             (timer_pending(&mp->timer) ?
1170              time_after(mp->timer.expires, now + max_delay) :
1171              try_to_del_timer_sync(&mp->timer) >= 0))
1172                 mod_timer(&mp->timer, now + max_delay);
1173
1174         for (pp = &mp->ports;
1175              (p = mlock_dereference(*pp, br)) != NULL;
1176              pp = &p->next) {
1177                 if (timer_pending(&p->timer) ?
1178                     time_after(p->timer.expires, now + max_delay) :
1179                     try_to_del_timer_sync(&p->timer) >= 0)
1180                         mod_timer(&mp->timer, now + max_delay);
1181         }
1182
1183 out:
1184         spin_unlock(&br->multicast_lock);
1185         return err;
1186 }
1187
1188 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
1189 static int br_ip6_multicast_query(struct net_bridge *br,
1190                                   struct net_bridge_port *port,
1191                                   struct sk_buff *skb)
1192 {
1193         struct ipv6hdr *ip6h = ipv6_hdr(skb);
1194         struct mld_msg *mld = (struct mld_msg *) icmp6_hdr(skb);
1195         struct net_bridge_mdb_entry *mp;
1196         struct mld2_query *mld2q;
1197         struct net_bridge_port_group *p;
1198         struct net_bridge_port_group __rcu **pp;
1199         unsigned long max_delay;
1200         unsigned long now = jiffies;
1201         struct in6_addr *group = NULL;
1202         int err = 0;
1203
1204         spin_lock(&br->multicast_lock);
1205         if (!netif_running(br->dev) ||
1206             (port && port->state == BR_STATE_DISABLED))
1207                 goto out;
1208
1209         br_multicast_query_received(br, port, !ipv6_addr_any(&ip6h->saddr));
1210
1211         if (skb->len == sizeof(*mld)) {
1212                 if (!pskb_may_pull(skb, sizeof(*mld))) {
1213                         err = -EINVAL;
1214                         goto out;
1215                 }
1216                 mld = (struct mld_msg *) icmp6_hdr(skb);
1217                 max_delay = msecs_to_jiffies(htons(mld->mld_maxdelay));
1218                 if (max_delay)
1219                         group = &mld->mld_mca;
1220         } else if (skb->len >= sizeof(*mld2q)) {
1221                 if (!pskb_may_pull(skb, sizeof(*mld2q))) {
1222                         err = -EINVAL;
1223                         goto out;
1224                 }
1225                 mld2q = (struct mld2_query *)icmp6_hdr(skb);
1226                 if (!mld2q->mld2q_nsrcs)
1227                         group = &mld2q->mld2q_mca;
1228                 max_delay = mld2q->mld2q_mrc ? MLDV2_MRC(mld2q->mld2q_mrc) : 1;
1229         }
1230
1231         if (!group)
1232                 goto out;
1233
1234         mp = br_mdb_ip6_get(mlock_dereference(br->mdb, br), group);
1235         if (!mp)
1236                 goto out;
1237
1238         max_delay *= br->multicast_last_member_count;
1239         if (!hlist_unhashed(&mp->mglist) &&
1240             (timer_pending(&mp->timer) ?
1241              time_after(mp->timer.expires, now + max_delay) :
1242              try_to_del_timer_sync(&mp->timer) >= 0))
1243                 mod_timer(&mp->timer, now + max_delay);
1244
1245         for (pp = &mp->ports;
1246              (p = mlock_dereference(*pp, br)) != NULL;
1247              pp = &p->next) {
1248                 if (timer_pending(&p->timer) ?
1249                     time_after(p->timer.expires, now + max_delay) :
1250                     try_to_del_timer_sync(&p->timer) >= 0)
1251                         mod_timer(&mp->timer, now + max_delay);
1252         }
1253
1254 out:
1255         spin_unlock(&br->multicast_lock);
1256         return err;
1257 }
1258 #endif
1259
1260 static void br_multicast_leave_group(struct net_bridge *br,
1261                                      struct net_bridge_port *port,
1262                                      struct br_ip *group)
1263 {
1264         struct net_bridge_mdb_htable *mdb;
1265         struct net_bridge_mdb_entry *mp;
1266         struct net_bridge_port_group *p;
1267         unsigned long now;
1268         unsigned long time;
1269
1270         spin_lock(&br->multicast_lock);
1271         if (!netif_running(br->dev) ||
1272             (port && port->state == BR_STATE_DISABLED) ||
1273             timer_pending(&br->multicast_querier_timer))
1274                 goto out;
1275
1276         mdb = mlock_dereference(br->mdb, br);
1277         mp = br_mdb_ip_get(mdb, group);
1278         if (!mp)
1279                 goto out;
1280
1281         now = jiffies;
1282         time = now + br->multicast_last_member_count *
1283                      br->multicast_last_member_interval;
1284
1285         if (!port) {
1286                 if (!hlist_unhashed(&mp->mglist) &&
1287                     (timer_pending(&mp->timer) ?
1288                      time_after(mp->timer.expires, time) :
1289                      try_to_del_timer_sync(&mp->timer) >= 0)) {
1290                         mod_timer(&mp->timer, time);
1291
1292                         mp->queries_sent = 0;
1293                         mod_timer(&mp->query_timer, now);
1294                 }
1295
1296                 goto out;
1297         }
1298
1299         for (p = mlock_dereference(mp->ports, br);
1300              p != NULL;
1301              p = mlock_dereference(p->next, br)) {
1302                 if (p->port != port)
1303                         continue;
1304
1305                 if (!hlist_unhashed(&p->mglist) &&
1306                     (timer_pending(&p->timer) ?
1307                      time_after(p->timer.expires, time) :
1308                      try_to_del_timer_sync(&p->timer) >= 0)) {
1309                         mod_timer(&p->timer, time);
1310
1311                         p->queries_sent = 0;
1312                         mod_timer(&p->query_timer, now);
1313                 }
1314
1315                 break;
1316         }
1317
1318 out:
1319         spin_unlock(&br->multicast_lock);
1320 }
1321
1322 static void br_ip4_multicast_leave_group(struct net_bridge *br,
1323                                          struct net_bridge_port *port,
1324                                          __be32 group)
1325 {
1326         struct br_ip br_group;
1327
1328         if (ipv4_is_local_multicast(group))
1329                 return;
1330
1331         br_group.u.ip4 = group;
1332         br_group.proto = htons(ETH_P_IP);
1333
1334         br_multicast_leave_group(br, port, &br_group);
1335 }
1336
1337 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
1338 static void br_ip6_multicast_leave_group(struct net_bridge *br,
1339                                          struct net_bridge_port *port,
1340                                          const struct in6_addr *group)
1341 {
1342         struct br_ip br_group;
1343
1344         if (ipv6_is_local_multicast(group))
1345                 return;
1346
1347         ipv6_addr_copy(&br_group.u.ip6, group);
1348         br_group.proto = htons(ETH_P_IPV6);
1349
1350         br_multicast_leave_group(br, port, &br_group);
1351 }
1352 #endif
1353
1354 static int br_multicast_ipv4_rcv(struct net_bridge *br,
1355                                  struct net_bridge_port *port,
1356                                  struct sk_buff *skb)
1357 {
1358         struct sk_buff *skb2 = skb;
1359         struct iphdr *iph;
1360         struct igmphdr *ih;
1361         unsigned len;
1362         unsigned offset;
1363         int err;
1364
1365         /* We treat OOM as packet loss for now. */
1366         if (!pskb_may_pull(skb, sizeof(*iph)))
1367                 return -EINVAL;
1368
1369         iph = ip_hdr(skb);
1370
1371         if (iph->ihl < 5 || iph->version != 4)
1372                 return -EINVAL;
1373
1374         if (!pskb_may_pull(skb, ip_hdrlen(skb)))
1375                 return -EINVAL;
1376
1377         iph = ip_hdr(skb);
1378
1379         if (unlikely(ip_fast_csum((u8 *)iph, iph->ihl)))
1380                 return -EINVAL;
1381
1382         if (iph->protocol != IPPROTO_IGMP)
1383                 return 0;
1384
1385         len = ntohs(iph->tot_len);
1386         if (skb->len < len || len < ip_hdrlen(skb))
1387                 return -EINVAL;
1388
1389         if (skb->len > len) {
1390                 skb2 = skb_clone(skb, GFP_ATOMIC);
1391                 if (!skb2)
1392                         return -ENOMEM;
1393
1394                 err = pskb_trim_rcsum(skb2, len);
1395                 if (err)
1396                         goto err_out;
1397         }
1398
1399         len -= ip_hdrlen(skb2);
1400         offset = skb_network_offset(skb2) + ip_hdrlen(skb2);
1401         __skb_pull(skb2, offset);
1402         skb_reset_transport_header(skb2);
1403
1404         err = -EINVAL;
1405         if (!pskb_may_pull(skb2, sizeof(*ih)))
1406                 goto out;
1407
1408         switch (skb2->ip_summed) {
1409         case CHECKSUM_COMPLETE:
1410                 if (!csum_fold(skb2->csum))
1411                         break;
1412                 /* fall through */
1413         case CHECKSUM_NONE:
1414                 skb2->csum = 0;
1415                 if (skb_checksum_complete(skb2))
1416                         goto out;
1417         }
1418
1419         err = 0;
1420
1421         BR_INPUT_SKB_CB(skb)->igmp = 1;
1422         ih = igmp_hdr(skb2);
1423
1424         switch (ih->type) {
1425         case IGMP_HOST_MEMBERSHIP_REPORT:
1426         case IGMPV2_HOST_MEMBERSHIP_REPORT:
1427                 BR_INPUT_SKB_CB(skb2)->mrouters_only = 1;
1428                 err = br_ip4_multicast_add_group(br, port, ih->group);
1429                 break;
1430         case IGMPV3_HOST_MEMBERSHIP_REPORT:
1431                 err = br_ip4_multicast_igmp3_report(br, port, skb2);
1432                 break;
1433         case IGMP_HOST_MEMBERSHIP_QUERY:
1434                 err = br_ip4_multicast_query(br, port, skb2);
1435                 break;
1436         case IGMP_HOST_LEAVE_MESSAGE:
1437                 br_ip4_multicast_leave_group(br, port, ih->group);
1438                 break;
1439         }
1440
1441 out:
1442         __skb_push(skb2, offset);
1443 err_out:
1444         if (skb2 != skb)
1445                 kfree_skb(skb2);
1446         return err;
1447 }
1448
1449 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
1450 static int br_multicast_ipv6_rcv(struct net_bridge *br,
1451                                  struct net_bridge_port *port,
1452                                  struct sk_buff *skb)
1453 {
1454         struct sk_buff *skb2;
1455         struct ipv6hdr *ip6h;
1456         struct icmp6hdr *icmp6h;
1457         u8 nexthdr;
1458         unsigned len;
1459         int offset;
1460         int err;
1461
1462         if (!pskb_may_pull(skb, sizeof(*ip6h)))
1463                 return -EINVAL;
1464
1465         ip6h = ipv6_hdr(skb);
1466
1467         /*
1468          * We're interested in MLD messages only.
1469          *  - Version is 6
1470          *  - MLD has always Router Alert hop-by-hop option
1471          *  - But we do not support jumbrograms.
1472          */
1473         if (ip6h->version != 6 ||
1474             ip6h->nexthdr != IPPROTO_HOPOPTS ||
1475             ip6h->payload_len == 0)
1476                 return 0;
1477
1478         len = ntohs(ip6h->payload_len);
1479         if (skb->len < len)
1480                 return -EINVAL;
1481
1482         nexthdr = ip6h->nexthdr;
1483         offset = ipv6_skip_exthdr(skb, sizeof(*ip6h), &nexthdr);
1484
1485         if (offset < 0 || nexthdr != IPPROTO_ICMPV6)
1486                 return 0;
1487
1488         /* Okay, we found ICMPv6 header */
1489         skb2 = skb_clone(skb, GFP_ATOMIC);
1490         if (!skb2)
1491                 return -ENOMEM;
1492
1493         err = -EINVAL;
1494         if (!pskb_may_pull(skb2, offset + sizeof(struct icmp6hdr)))
1495                 goto out;
1496
1497         len -= offset - skb_network_offset(skb2);
1498
1499         __skb_pull(skb2, offset);
1500         skb_reset_transport_header(skb2);
1501
1502         icmp6h = icmp6_hdr(skb2);
1503
1504         switch (icmp6h->icmp6_type) {
1505         case ICMPV6_MGM_QUERY:
1506         case ICMPV6_MGM_REPORT:
1507         case ICMPV6_MGM_REDUCTION:
1508         case ICMPV6_MLD2_REPORT:
1509                 break;
1510         default:
1511                 err = 0;
1512                 goto out;
1513         }
1514
1515         /* Okay, we found MLD message. Check further. */
1516         if (skb2->len > len) {
1517                 err = pskb_trim_rcsum(skb2, len);
1518                 if (err)
1519                         goto out;
1520         }
1521
1522         switch (skb2->ip_summed) {
1523         case CHECKSUM_COMPLETE:
1524                 if (!csum_fold(skb2->csum))
1525                         break;
1526                 /*FALLTHROUGH*/
1527         case CHECKSUM_NONE:
1528                 skb2->csum = 0;
1529                 if (skb_checksum_complete(skb2))
1530                         goto out;
1531         }
1532
1533         err = 0;
1534
1535         BR_INPUT_SKB_CB(skb)->igmp = 1;
1536
1537         switch (icmp6h->icmp6_type) {
1538         case ICMPV6_MGM_REPORT:
1539             {
1540                 struct mld_msg *mld;
1541                 if (!pskb_may_pull(skb2, sizeof(*mld))) {
1542                         err = -EINVAL;
1543                         goto out;
1544                 }
1545                 mld = (struct mld_msg *)skb_transport_header(skb2);
1546                 BR_INPUT_SKB_CB(skb2)->mrouters_only = 1;
1547                 err = br_ip6_multicast_add_group(br, port, &mld->mld_mca);
1548                 break;
1549             }
1550         case ICMPV6_MLD2_REPORT:
1551                 err = br_ip6_multicast_mld2_report(br, port, skb2);
1552                 break;
1553         case ICMPV6_MGM_QUERY:
1554                 err = br_ip6_multicast_query(br, port, skb2);
1555                 break;
1556         case ICMPV6_MGM_REDUCTION:
1557             {
1558                 struct mld_msg *mld;
1559                 if (!pskb_may_pull(skb2, sizeof(*mld))) {
1560                         err = -EINVAL;
1561                         goto out;
1562                 }
1563                 mld = (struct mld_msg *)skb_transport_header(skb2);
1564                 br_ip6_multicast_leave_group(br, port, &mld->mld_mca);
1565             }
1566         }
1567
1568 out:
1569         kfree_skb(skb2);
1570         return err;
1571 }
1572 #endif
1573
1574 int br_multicast_rcv(struct net_bridge *br, struct net_bridge_port *port,
1575                      struct sk_buff *skb)
1576 {
1577         BR_INPUT_SKB_CB(skb)->igmp = 0;
1578         BR_INPUT_SKB_CB(skb)->mrouters_only = 0;
1579
1580         if (br->multicast_disabled)
1581                 return 0;
1582
1583         switch (skb->protocol) {
1584         case htons(ETH_P_IP):
1585                 return br_multicast_ipv4_rcv(br, port, skb);
1586 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
1587         case htons(ETH_P_IPV6):
1588                 return br_multicast_ipv6_rcv(br, port, skb);
1589 #endif
1590         }
1591
1592         return 0;
1593 }
1594
1595 static void br_multicast_query_expired(unsigned long data)
1596 {
1597         struct net_bridge *br = (void *)data;
1598
1599         spin_lock(&br->multicast_lock);
1600         if (br->multicast_startup_queries_sent <
1601             br->multicast_startup_query_count)
1602                 br->multicast_startup_queries_sent++;
1603
1604         br_multicast_send_query(br, NULL, br->multicast_startup_queries_sent);
1605
1606         spin_unlock(&br->multicast_lock);
1607 }
1608
1609 void br_multicast_init(struct net_bridge *br)
1610 {
1611         br->hash_elasticity = 4;
1612         br->hash_max = 512;
1613
1614         br->multicast_router = 1;
1615         br->multicast_last_member_count = 2;
1616         br->multicast_startup_query_count = 2;
1617
1618         br->multicast_last_member_interval = HZ;
1619         br->multicast_query_response_interval = 10 * HZ;
1620         br->multicast_startup_query_interval = 125 * HZ / 4;
1621         br->multicast_query_interval = 125 * HZ;
1622         br->multicast_querier_interval = 255 * HZ;
1623         br->multicast_membership_interval = 260 * HZ;
1624
1625         spin_lock_init(&br->multicast_lock);
1626         setup_timer(&br->multicast_router_timer,
1627                     br_multicast_local_router_expired, 0);
1628         setup_timer(&br->multicast_querier_timer,
1629                     br_multicast_local_router_expired, 0);
1630         setup_timer(&br->multicast_query_timer, br_multicast_query_expired,
1631                     (unsigned long)br);
1632 }
1633
1634 void br_multicast_open(struct net_bridge *br)
1635 {
1636         br->multicast_startup_queries_sent = 0;
1637
1638         if (br->multicast_disabled)
1639                 return;
1640
1641         mod_timer(&br->multicast_query_timer, jiffies);
1642 }
1643
1644 void br_multicast_stop(struct net_bridge *br)
1645 {
1646         struct net_bridge_mdb_htable *mdb;
1647         struct net_bridge_mdb_entry *mp;
1648         struct hlist_node *p, *n;
1649         u32 ver;
1650         int i;
1651
1652         del_timer_sync(&br->multicast_router_timer);
1653         del_timer_sync(&br->multicast_querier_timer);
1654         del_timer_sync(&br->multicast_query_timer);
1655
1656         spin_lock_bh(&br->multicast_lock);
1657         mdb = mlock_dereference(br->mdb, br);
1658         if (!mdb)
1659                 goto out;
1660
1661         br->mdb = NULL;
1662
1663         ver = mdb->ver;
1664         for (i = 0; i < mdb->max; i++) {
1665                 hlist_for_each_entry_safe(mp, p, n, &mdb->mhash[i],
1666                                           hlist[ver]) {
1667                         del_timer(&mp->timer);
1668                         del_timer(&mp->query_timer);
1669                         call_rcu_bh(&mp->rcu, br_multicast_free_group);
1670                 }
1671         }
1672
1673         if (mdb->old) {
1674                 spin_unlock_bh(&br->multicast_lock);
1675                 rcu_barrier_bh();
1676                 spin_lock_bh(&br->multicast_lock);
1677                 WARN_ON(mdb->old);
1678         }
1679
1680         mdb->old = mdb;
1681         call_rcu_bh(&mdb->rcu, br_mdb_free);
1682
1683 out:
1684         spin_unlock_bh(&br->multicast_lock);
1685 }
1686
1687 int br_multicast_set_router(struct net_bridge *br, unsigned long val)
1688 {
1689         int err = -ENOENT;
1690
1691         spin_lock_bh(&br->multicast_lock);
1692         if (!netif_running(br->dev))
1693                 goto unlock;
1694
1695         switch (val) {
1696         case 0:
1697         case 2:
1698                 del_timer(&br->multicast_router_timer);
1699                 /* fall through */
1700         case 1:
1701                 br->multicast_router = val;
1702                 err = 0;
1703                 break;
1704
1705         default:
1706                 err = -EINVAL;
1707                 break;
1708         }
1709
1710 unlock:
1711         spin_unlock_bh(&br->multicast_lock);
1712
1713         return err;
1714 }
1715
1716 int br_multicast_set_port_router(struct net_bridge_port *p, unsigned long val)
1717 {
1718         struct net_bridge *br = p->br;
1719         int err = -ENOENT;
1720
1721         spin_lock(&br->multicast_lock);
1722         if (!netif_running(br->dev) || p->state == BR_STATE_DISABLED)
1723                 goto unlock;
1724
1725         switch (val) {
1726         case 0:
1727         case 1:
1728         case 2:
1729                 p->multicast_router = val;
1730                 err = 0;
1731
1732                 if (val < 2 && !hlist_unhashed(&p->rlist))
1733                         hlist_del_init_rcu(&p->rlist);
1734
1735                 if (val == 1)
1736                         break;
1737
1738                 del_timer(&p->multicast_router_timer);
1739
1740                 if (val == 0)
1741                         break;
1742
1743                 br_multicast_add_router(br, p);
1744                 break;
1745
1746         default:
1747                 err = -EINVAL;
1748                 break;
1749         }
1750
1751 unlock:
1752         spin_unlock(&br->multicast_lock);
1753
1754         return err;
1755 }
1756
1757 int br_multicast_toggle(struct net_bridge *br, unsigned long val)
1758 {
1759         struct net_bridge_port *port;
1760         int err = 0;
1761         struct net_bridge_mdb_htable *mdb;
1762
1763         spin_lock(&br->multicast_lock);
1764         if (br->multicast_disabled == !val)
1765                 goto unlock;
1766
1767         br->multicast_disabled = !val;
1768         if (br->multicast_disabled)
1769                 goto unlock;
1770
1771         if (!netif_running(br->dev))
1772                 goto unlock;
1773
1774         mdb = mlock_dereference(br->mdb, br);
1775         if (mdb) {
1776                 if (mdb->old) {
1777                         err = -EEXIST;
1778 rollback:
1779                         br->multicast_disabled = !!val;
1780                         goto unlock;
1781                 }
1782
1783                 err = br_mdb_rehash(&br->mdb, mdb->max,
1784                                     br->hash_elasticity);
1785                 if (err)
1786                         goto rollback;
1787         }
1788
1789         br_multicast_open(br);
1790         list_for_each_entry(port, &br->port_list, list) {
1791                 if (port->state == BR_STATE_DISABLED ||
1792                     port->state == BR_STATE_BLOCKING)
1793                         continue;
1794
1795                 __br_multicast_enable_port(port);
1796         }
1797
1798 unlock:
1799         spin_unlock(&br->multicast_lock);
1800
1801         return err;
1802 }
1803
1804 int br_multicast_set_hash_max(struct net_bridge *br, unsigned long val)
1805 {
1806         int err = -ENOENT;
1807         u32 old;
1808         struct net_bridge_mdb_htable *mdb;
1809
1810         spin_lock(&br->multicast_lock);
1811         if (!netif_running(br->dev))
1812                 goto unlock;
1813
1814         err = -EINVAL;
1815         if (!is_power_of_2(val))
1816                 goto unlock;
1817
1818         mdb = mlock_dereference(br->mdb, br);
1819         if (mdb && val < mdb->size)
1820                 goto unlock;
1821
1822         err = 0;
1823
1824         old = br->hash_max;
1825         br->hash_max = val;
1826
1827         if (mdb) {
1828                 if (mdb->old) {
1829                         err = -EEXIST;
1830 rollback:
1831                         br->hash_max = old;
1832                         goto unlock;
1833                 }
1834
1835                 err = br_mdb_rehash(&br->mdb, br->hash_max,
1836                                     br->hash_elasticity);
1837                 if (err)
1838                         goto rollback;
1839         }
1840
1841 unlock:
1842         spin_unlock(&br->multicast_lock);
1843
1844         return err;
1845 }