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                 if (hlist_unhashed(&mp->mglist))
 723                         hlist_add_head(&mp->mglist, &br->mglist);
 724                 mod_timer(&mp->timer, now + br->multicast_membership_interval);
 725                 goto out;
 726         }
 727
 728         for (pp = &mp->ports;
 729              (p = mlock_dereference(*pp, br)) != NULL;
 730              pp = &p->next) {
 731                 if (p->port == port)
 732                         goto found;
 733                 if ((unsigned long)p->port < (unsigned long)port)
 734                         break;
 735         }
 736
 737         p = kzalloc(sizeof(*p), GFP_ATOMIC);
 738         err = -ENOMEM;
 739         if (unlikely(!p))
 740                 goto err;
 741
 742         p->addr = *group;
 743         p->port = port;
 744         p->next = *pp;
 745         hlist_add_head(&p->mglist, &port->mglist);
 746         setup_timer(&p->timer, br_multicast_port_group_expired,
 747                     (unsigned long)p);
 748         setup_timer(&p->query_timer, br_multicast_port_group_query_expired,
 749                     (unsigned long)p);
 750
 751         rcu_assign_pointer(*pp, p);
 752
 753 found:
 754         mod_timer(&p->timer, now + br->multicast_membership_interval);
 755 out:
 756         err = 0;
 757
 758 err:
 759         spin_unlock(&br->multicast_lock);
 760         return err;
 761 }
 762
 763 static int br_ip4_multicast_add_group(struct net_bridge *br,
 764                                       struct net_bridge_port *port,
 765                                       __be32 group)
 766 {
 767         struct br_ip br_group;
 768
 769         if (ipv4_is_local_multicast(group))
 770                 return 0;
 771
 772         br_group.u.ip4 = group;
 773         br_group.proto = htons(ETH_P_IP);
 774
 775         return br_multicast_add_group(br, port, &br_group);
 776 }
 777
 778 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
 779 static int br_ip6_multicast_add_group(struct net_bridge *br,
 780                                       struct net_bridge_port *port,
 781                                       const struct in6_addr *group)
 782 {
 783         struct br_ip br_group;
 784
 785         if (ipv6_is_local_multicast(group))
 786                 return 0;
 787
 788         ipv6_addr_copy(&br_group.u.ip6, group);
 789         br_group.proto = htons(ETH_P_IP);
 790
 791         return br_multicast_add_group(br, port, &br_group);
 792 }
 793 #endif
 794
 795 static void br_multicast_router_expired(unsigned long data)
 796 {
 797         struct net_bridge_port *port = (void *)data;
 798         struct net_bridge *br = port->br;
 799
 800         spin_lock(&br->multicast_lock);
 801         if (port->multicast_router != 1 ||
 802             timer_pending(&port->multicast_router_timer) ||
 803             hlist_unhashed(&port->rlist))
 804                 goto out;
 805
 806         hlist_del_init_rcu(&port->rlist);
 807
 808 out:
 809         spin_unlock(&br->multicast_lock);
 810 }
 811
 812 static void br_multicast_local_router_expired(unsigned long data)
 813 {
 814 }
 815
 816 static void __br_multicast_send_query(struct net_bridge *br,
 817                                       struct net_bridge_port *port,
 818                                       struct br_ip *ip)
 819 {
 820         struct sk_buff *skb;
 821
 822         skb = br_multicast_alloc_query(br, ip);
 823         if (!skb)
 824                 return;
 825
 826         if (port) {
 827                 __skb_push(skb, sizeof(struct ethhdr));
 828                 skb->dev = port->dev;
 829                 NF_HOOK(NFPROTO_BRIDGE, NF_BR_LOCAL_OUT, skb, NULL, skb->dev,
 830                         dev_queue_xmit);
 831         } else
 832                 netif_rx(skb);
 833 }
 834
 835 static void br_multicast_send_query(struct net_bridge *br,
 836                                     struct net_bridge_port *port, u32 sent)
 837 {
 838         unsigned long time;
 839         struct br_ip br_group;
 840
 841         if (!netif_running(br->dev) || br->multicast_disabled ||
 842             timer_pending(&br->multicast_querier_timer))
 843                 return;
 844
 845         memset(&br_group.u, 0, sizeof(br_group.u));
 846
 847         br_group.proto = htons(ETH_P_IP);
 848         __br_multicast_send_query(br, port, &br_group);
 849
 850 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
 851         br_group.proto = htons(ETH_P_IPV6);
 852         __br_multicast_send_query(br, port, &br_group);
 853 #endif
 854
 855         time = jiffies;
 856         time += sent < br->multicast_startup_query_count ?
 857                 br->multicast_startup_query_interval :
 858                 br->multicast_query_interval;
 859         mod_timer(port ? &port->multicast_query_timer :
 860                          &br->multicast_query_timer, time);
 861 }
 862
 863 static void br_multicast_port_query_expired(unsigned long data)
 864 {
 865         struct net_bridge_port *port = (void *)data;
 866         struct net_bridge *br = port->br;
 867
 868         spin_lock(&br->multicast_lock);
 869         if (port->state == BR_STATE_DISABLED ||
 870             port->state == BR_STATE_BLOCKING)
 871                 goto out;
 872
 873         if (port->multicast_startup_queries_sent <
 874             br->multicast_startup_query_count)
 875                 port->multicast_startup_queries_sent++;
 876
 877         br_multicast_send_query(port->br, port,
 878                                 port->multicast_startup_queries_sent);
 879
 880 out:
 881         spin_unlock(&br->multicast_lock);
 882 }
 883
 884 void br_multicast_add_port(struct net_bridge_port *port)
 885 {
 886         port->multicast_router = 1;
 887
 888         setup_timer(&port->multicast_router_timer, br_multicast_router_expired,
 889                     (unsigned long)port);
 890         setup_timer(&port->multicast_query_timer,
 891                     br_multicast_port_query_expired, (unsigned long)port);
 892 }
 893
 894 void br_multicast_del_port(struct net_bridge_port *port)
 895 {
 896         del_timer_sync(&port->multicast_router_timer);
 897 }
 898
 899 static void __br_multicast_enable_port(struct net_bridge_port *port)
 900 {
 901         port->multicast_startup_queries_sent = 0;
 902
 903         if (try_to_del_timer_sync(&port->multicast_query_timer) >= 0 ||
 904             del_timer(&port->multicast_query_timer))
 905                 mod_timer(&port->multicast_query_timer, jiffies);
 906 }
 907
 908 void br_multicast_enable_port(struct net_bridge_port *port)
 909 {
 910         struct net_bridge *br = port->br;
 911
 912         spin_lock(&br->multicast_lock);
 913         if (br->multicast_disabled || !netif_running(br->dev))
 914                 goto out;
 915
 916         __br_multicast_enable_port(port);
 917
 918 out:
 919         spin_unlock(&br->multicast_lock);
 920 }
 921
 922 void br_multicast_disable_port(struct net_bridge_port *port)
 923 {
 924         struct net_bridge *br = port->br;
 925         struct net_bridge_port_group *pg;
 926         struct hlist_node *p, *n;
 927
 928         spin_lock(&br->multicast_lock);
 929         hlist_for_each_entry_safe(pg, p, n, &port->mglist, mglist)
 930                 br_multicast_del_pg(br, pg);
 931
 932         if (!hlist_unhashed(&port->rlist))
 933                 hlist_del_init_rcu(&port->rlist);
 934         del_timer(&port->multicast_router_timer);
 935         del_timer(&port->multicast_query_timer);
 936         spin_unlock(&br->multicast_lock);
 937 }
 938
 939 static int br_ip4_multicast_igmp3_report(struct net_bridge *br,
 940                                          struct net_bridge_port *port,
 941                                          struct sk_buff *skb)
 942 {
 943         struct igmpv3_report *ih;
 944         struct igmpv3_grec *grec;
 945         int i;
 946         int len;
 947         int num;
 948         int type;
 949         int err = 0;
 950         __be32 group;
 951
 952         if (!pskb_may_pull(skb, sizeof(*ih)))
 953                 return -EINVAL;
 954
 955         ih = igmpv3_report_hdr(skb);
 956         num = ntohs(ih->ngrec);
 957         len = sizeof(*ih);
 958
 959         for (i = 0; i < num; i++) {
 960                 len += sizeof(*grec);
 961                 if (!pskb_may_pull(skb, len))
 962                         return -EINVAL;
 963
 964                 grec = (void *)(skb->data + len - sizeof(*grec));
 965                 group = grec->grec_mca;
 966                 type = grec->grec_type;
 967
 968                 len += ntohs(grec->grec_nsrcs) * 4;
 969                 if (!pskb_may_pull(skb, len))
 970                         return -EINVAL;
 971
 972                 /* We treat this as an IGMPv2 report for now. */
 973                 switch (type) {
 974                 case IGMPV3_MODE_IS_INCLUDE:
 975                 case IGMPV3_MODE_IS_EXCLUDE:
 976                 case IGMPV3_CHANGE_TO_INCLUDE:
 977                 case IGMPV3_CHANGE_TO_EXCLUDE:
 978                 case IGMPV3_ALLOW_NEW_SOURCES:
 979                 case IGMPV3_BLOCK_OLD_SOURCES:
 980                         break;
 981
 982                 default:
 983                         continue;
 984                 }
 985
 986                 err = br_ip4_multicast_add_group(br, port, group);
 987                 if (err)
 988                         break;
 989         }
 990
 991         return err;
 992 }
 993
 994 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
 995 static int br_ip6_multicast_mld2_report(struct net_bridge *br,
 996                                         struct net_bridge_port *port,
 997                                         struct sk_buff *skb)
 998 {
 999         struct icmp6hdr *icmp6h;
1000         struct mld2_grec *grec;
1001         int i;
1002         int len;
1003         int num;
1004         int err = 0;
1005
1006         if (!pskb_may_pull(skb, sizeof(*icmp6h)))
1007                 return -EINVAL;
1008
1009         icmp6h = icmp6_hdr(skb);
1010         num = ntohs(icmp6h->icmp6_dataun.un_data16[1]);
1011         len = sizeof(*icmp6h);
1012
1013         for (i = 0; i < num; i++) {
1014                 __be16 *nsrcs, _nsrcs;
1015
1016                 nsrcs = skb_header_pointer(skb,
1017                                            len + offsetof(struct mld2_grec,
1018                                                           grec_mca),
1019                                            sizeof(_nsrcs), &_nsrcs);
1020                 if (!nsrcs)
1021                         return -EINVAL;
1022
1023                 if (!pskb_may_pull(skb,
1024                                    len + sizeof(*grec) +
1025                                    sizeof(struct in6_addr) * (*nsrcs)))
1026                         return -EINVAL;
1027
1028                 grec = (struct mld2_grec *)(skb->data + len);
1029                 len += sizeof(*grec) + sizeof(struct in6_addr) * (*nsrcs);
1030
1031                 /* We treat these as MLDv1 reports for now. */
1032                 switch (grec->grec_type) {
1033                 case MLD2_MODE_IS_INCLUDE:
1034                 case MLD2_MODE_IS_EXCLUDE:
1035                 case MLD2_CHANGE_TO_INCLUDE:
1036                 case MLD2_CHANGE_TO_EXCLUDE:
1037                 case MLD2_ALLOW_NEW_SOURCES:
1038                 case MLD2_BLOCK_OLD_SOURCES:
1039                         break;
1040
1041                 default:
1042                         continue;
1043                 }
1044
1045                 err = br_ip6_multicast_add_group(br, port, &grec->grec_mca);
1046                 if (!err)
1047                         break;
1048         }
1049
1050         return err;
1051 }
1052 #endif
1053
1054 /*
1055  * Add port to rotuer_list
1056  *  list is maintained ordered by pointer value
1057  *  and locked by br->multicast_lock and RCU
1058  */
1059 static void br_multicast_add_router(struct net_bridge *br,
1060                                     struct net_bridge_port *port)
1061 {
1062         struct net_bridge_port *p;
1063         struct hlist_node *n, *slot = NULL;
1064
1065         hlist_for_each_entry(p, n, &br->router_list, rlist) {
1066                 if ((unsigned long) port >= (unsigned long) p)
1067                         break;
1068                 slot = n;
1069         }
1070
1071         if (slot)
1072                 hlist_add_after_rcu(slot, &port->rlist);
1073         else
1074                 hlist_add_head_rcu(&port->rlist, &br->router_list);
1075 }
1076
1077 static void br_multicast_mark_router(struct net_bridge *br,
1078                                      struct net_bridge_port *port)
1079 {
1080         unsigned long now = jiffies;
1081
1082         if (!port) {
1083                 if (br->multicast_router == 1)
1084                         mod_timer(&br->multicast_router_timer,
1085                                   now + br->multicast_querier_interval);
1086                 return;
1087         }
1088
1089         if (port->multicast_router != 1)
1090                 return;
1091
1092         if (!hlist_unhashed(&port->rlist))
1093                 goto timer;
1094
1095         br_multicast_add_router(br, port);
1096
1097 timer:
1098         mod_timer(&port->multicast_router_timer,
1099                   now + br->multicast_querier_interval);
1100 }
1101
1102 static void br_multicast_query_received(struct net_bridge *br,
1103                                         struct net_bridge_port *port,
1104                                         int saddr)
1105 {
1106         if (saddr)
1107                 mod_timer(&br->multicast_querier_timer,
1108                           jiffies + br->multicast_querier_interval);
1109         else if (timer_pending(&br->multicast_querier_timer))
1110                 return;
1111
1112         br_multicast_mark_router(br, port);
1113 }
1114
1115 static int br_ip4_multicast_query(struct net_bridge *br,
1116                                   struct net_bridge_port *port,
1117                                   struct sk_buff *skb)
1118 {
1119         struct iphdr *iph = ip_hdr(skb);
1120         struct igmphdr *ih = igmp_hdr(skb);
1121         struct net_bridge_mdb_entry *mp;
1122         struct igmpv3_query *ih3;
1123         struct net_bridge_port_group *p;
1124         struct net_bridge_port_group __rcu **pp;
1125         unsigned long max_delay;
1126         unsigned long now = jiffies;
1127         __be32 group;
1128         int err = 0;
1129
1130         spin_lock(&br->multicast_lock);
1131         if (!netif_running(br->dev) ||
1132             (port && port->state == BR_STATE_DISABLED))
1133                 goto out;
1134
1135         br_multicast_query_received(br, port, !!iph->saddr);
1136
1137         group = ih->group;
1138
1139         if (skb->len == sizeof(*ih)) {
1140                 max_delay = ih->code * (HZ / IGMP_TIMER_SCALE);
1141
1142                 if (!max_delay) {
1143                         max_delay = 10 * HZ;
1144                         group = 0;
1145                 }
1146         } else {
1147                 if (!pskb_may_pull(skb, sizeof(struct igmpv3_query))) {
1148                         err = -EINVAL;
1149                         goto out;
1150                 }
1151
1152                 ih3 = igmpv3_query_hdr(skb);
1153                 if (ih3->nsrcs)
1154                         goto out;
1155
1156                 max_delay = ih3->code ?
1157                             IGMPV3_MRC(ih3->code) * (HZ / IGMP_TIMER_SCALE) : 1;
1158         }
1159
1160         if (!group)
1161                 goto out;
1162
1163         mp = br_mdb_ip4_get(mlock_dereference(br->mdb, br), group);
1164         if (!mp)
1165                 goto out;
1166
1167         max_delay *= br->multicast_last_member_count;
1168
1169         if (!hlist_unhashed(&mp->mglist) &&
1170             (timer_pending(&mp->timer) ?
1171              time_after(mp->timer.expires, now + max_delay) :
1172              try_to_del_timer_sync(&mp->timer) >= 0))
1173                 mod_timer(&mp->timer, now + max_delay);
1174
1175         for (pp = &mp->ports;
1176              (p = mlock_dereference(*pp, br)) != NULL;
1177              pp = &p->next) {
1178                 if (timer_pending(&p->timer) ?
1179                     time_after(p->timer.expires, now + max_delay) :
1180                     try_to_del_timer_sync(&p->timer) >= 0)
1181                         mod_timer(&p->timer, now + max_delay);
1182         }
1183
1184 out:
1185         spin_unlock(&br->multicast_lock);
1186         return err;
1187 }
1188
1189 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
1190 static int br_ip6_multicast_query(struct net_bridge *br,
1191                                   struct net_bridge_port *port,
1192                                   struct sk_buff *skb)
1193 {
1194         struct ipv6hdr *ip6h = ipv6_hdr(skb);
1195         struct mld_msg *mld = (struct mld_msg *) icmp6_hdr(skb);
1196         struct net_bridge_mdb_entry *mp;
1197         struct mld2_query *mld2q;
1198         struct net_bridge_port_group *p;
1199         struct net_bridge_port_group __rcu **pp;
1200         unsigned long max_delay;
1201         unsigned long now = jiffies;
1202         struct in6_addr *group = NULL;
1203         int err = 0;
1204
1205         spin_lock(&br->multicast_lock);
1206         if (!netif_running(br->dev) ||
1207             (port && port->state == BR_STATE_DISABLED))
1208                 goto out;
1209
1210         br_multicast_query_received(br, port, !ipv6_addr_any(&ip6h->saddr));
1211
1212         if (skb->len == sizeof(*mld)) {
1213                 if (!pskb_may_pull(skb, sizeof(*mld))) {
1214                         err = -EINVAL;
1215                         goto out;
1216                 }
1217                 mld = (struct mld_msg *) icmp6_hdr(skb);
1218                 max_delay = msecs_to_jiffies(htons(mld->mld_maxdelay));
1219                 if (max_delay)
1220                         group = &mld->mld_mca;
1221         } else if (skb->len >= sizeof(*mld2q)) {
1222                 if (!pskb_may_pull(skb, sizeof(*mld2q))) {
1223                         err = -EINVAL;
1224                         goto out;
1225                 }
1226                 mld2q = (struct mld2_query *)icmp6_hdr(skb);
1227                 if (!mld2q->mld2q_nsrcs)
1228                         group = &mld2q->mld2q_mca;
1229                 max_delay = mld2q->mld2q_mrc ? MLDV2_MRC(mld2q->mld2q_mrc) : 1;
1230         }
1231
1232         if (!group)
1233                 goto out;
1234
1235         mp = br_mdb_ip6_get(mlock_dereference(br->mdb, br), group);
1236         if (!mp)
1237                 goto out;
1238
1239         max_delay *= br->multicast_last_member_count;
1240         if (!hlist_unhashed(&mp->mglist) &&
1241             (timer_pending(&mp->timer) ?
1242              time_after(mp->timer.expires, now + max_delay) :
1243              try_to_del_timer_sync(&mp->timer) >= 0))
1244                 mod_timer(&mp->timer, now + max_delay);
1245
1246         for (pp = &mp->ports;
1247              (p = mlock_dereference(*pp, br)) != NULL;
1248              pp = &p->next) {
1249                 if (timer_pending(&p->timer) ?
1250                     time_after(p->timer.expires, now + max_delay) :
1251                     try_to_del_timer_sync(&p->timer) >= 0)
1252                         mod_timer(&p->timer, now + max_delay);
1253         }
1254
1255 out:
1256         spin_unlock(&br->multicast_lock);
1257         return err;
1258 }
1259 #endif
1260
1261 static void br_multicast_leave_group(struct net_bridge *br,
1262                                      struct net_bridge_port *port,
1263                                      struct br_ip *group)
1264 {
1265         struct net_bridge_mdb_htable *mdb;
1266         struct net_bridge_mdb_entry *mp;
1267         struct net_bridge_port_group *p;
1268         unsigned long now;
1269         unsigned long time;
1270
1271         spin_lock(&br->multicast_lock);
1272         if (!netif_running(br->dev) ||
1273             (port && port->state == BR_STATE_DISABLED) ||
1274             timer_pending(&br->multicast_querier_timer))
1275                 goto out;
1276
1277         mdb = mlock_dereference(br->mdb, br);
1278         mp = br_mdb_ip_get(mdb, group);
1279         if (!mp)
1280                 goto out;
1281
1282         now = jiffies;
1283         time = now + br->multicast_last_member_count *
1284                      br->multicast_last_member_interval;
1285
1286         if (!port) {
1287                 if (!hlist_unhashed(&mp->mglist) &&
1288                     (timer_pending(&mp->timer) ?
1289                      time_after(mp->timer.expires, time) :
1290                      try_to_del_timer_sync(&mp->timer) >= 0)) {
1291                         mod_timer(&mp->timer, time);
1292
1293                         mp->queries_sent = 0;
1294                         mod_timer(&mp->query_timer, now);
1295                 }
1296
1297                 goto out;
1298         }
1299
1300         for (p = mlock_dereference(mp->ports, br);
1301              p != NULL;
1302              p = mlock_dereference(p->next, br)) {
1303                 if (p->port != port)
1304                         continue;
1305
1306                 if (!hlist_unhashed(&p->mglist) &&
1307                     (timer_pending(&p->timer) ?
1308                      time_after(p->timer.expires, time) :
1309                      try_to_del_timer_sync(&p->timer) >= 0)) {
1310                         mod_timer(&p->timer, time);
1311
1312                         p->queries_sent = 0;
1313                         mod_timer(&p->query_timer, now);
1314                 }
1315
1316                 break;
1317         }
1318
1319 out:
1320         spin_unlock(&br->multicast_lock);
1321 }
1322
1323 static void br_ip4_multicast_leave_group(struct net_bridge *br,
1324                                          struct net_bridge_port *port,
1325                                          __be32 group)
1326 {
1327         struct br_ip br_group;
1328
1329         if (ipv4_is_local_multicast(group))
1330                 return;
1331
1332         br_group.u.ip4 = group;
1333         br_group.proto = htons(ETH_P_IP);
1334
1335         br_multicast_leave_group(br, port, &br_group);
1336 }
1337
1338 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
1339 static void br_ip6_multicast_leave_group(struct net_bridge *br,
1340                                          struct net_bridge_port *port,
1341                                          const struct in6_addr *group)
1342 {
1343         struct br_ip br_group;
1344
1345         if (ipv6_is_local_multicast(group))
1346                 return;
1347
1348         ipv6_addr_copy(&br_group.u.ip6, group);
1349         br_group.proto = htons(ETH_P_IPV6);
1350
1351         br_multicast_leave_group(br, port, &br_group);
1352 }
1353 #endif
1354
1355 static int br_multicast_ipv4_rcv(struct net_bridge *br,
1356                                  struct net_bridge_port *port,
1357                                  struct sk_buff *skb)
1358 {
1359         struct sk_buff *skb2 = skb;
1360         struct iphdr *iph;
1361         struct igmphdr *ih;
1362         unsigned len;
1363         unsigned offset;
1364         int err;
1365
1366         /* We treat OOM as packet loss for now. */
1367         if (!pskb_may_pull(skb, sizeof(*iph)))
1368                 return -EINVAL;
1369
1370         iph = ip_hdr(skb);
1371
1372         if (iph->ihl < 5 || iph->version != 4)
1373                 return -EINVAL;
1374
1375         if (!pskb_may_pull(skb, ip_hdrlen(skb)))
1376                 return -EINVAL;
1377
1378         iph = ip_hdr(skb);
1379
1380         if (unlikely(ip_fast_csum((u8 *)iph, iph->ihl)))
1381                 return -EINVAL;
1382
1383         if (iph->protocol != IPPROTO_IGMP)
1384                 return 0;
1385
1386         len = ntohs(iph->tot_len);
1387         if (skb->len < len || len < ip_hdrlen(skb))
1388                 return -EINVAL;
1389
1390         if (skb->len > len) {
1391                 skb2 = skb_clone(skb, GFP_ATOMIC);
1392                 if (!skb2)
1393                         return -ENOMEM;
1394
1395                 err = pskb_trim_rcsum(skb2, len);
1396                 if (err)
1397                         goto err_out;
1398         }
1399
1400         len -= ip_hdrlen(skb2);
1401         offset = skb_network_offset(skb2) + ip_hdrlen(skb2);
1402         __skb_pull(skb2, offset);
1403         skb_reset_transport_header(skb2);
1404
1405         err = -EINVAL;
1406         if (!pskb_may_pull(skb2, sizeof(*ih)))
1407                 goto out;
1408
1409         switch (skb2->ip_summed) {
1410         case CHECKSUM_COMPLETE:
1411                 if (!csum_fold(skb2->csum))
1412                         break;
1413                 /* fall through */
1414         case CHECKSUM_NONE:
1415                 skb2->csum = 0;
1416                 if (skb_checksum_complete(skb2))
1417                         goto out;
1418         }
1419
1420         err = 0;
1421
1422         BR_INPUT_SKB_CB(skb)->igmp = 1;
1423         ih = igmp_hdr(skb2);
1424
1425         switch (ih->type) {
1426         case IGMP_HOST_MEMBERSHIP_REPORT:
1427         case IGMPV2_HOST_MEMBERSHIP_REPORT:
1428                 BR_INPUT_SKB_CB(skb2)->mrouters_only = 1;
1429                 err = br_ip4_multicast_add_group(br, port, ih->group);
1430                 break;
1431         case IGMPV3_HOST_MEMBERSHIP_REPORT:
1432                 err = br_ip4_multicast_igmp3_report(br, port, skb2);
1433                 break;
1434         case IGMP_HOST_MEMBERSHIP_QUERY:
1435                 err = br_ip4_multicast_query(br, port, skb2);
1436                 break;
1437         case IGMP_HOST_LEAVE_MESSAGE:
1438                 br_ip4_multicast_leave_group(br, port, ih->group);
1439                 break;
1440         }
1441
1442 out:
1443         __skb_push(skb2, offset);
1444 err_out:
1445         if (skb2 != skb)
1446                 kfree_skb(skb2);
1447         return err;
1448 }
1449
1450 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
1451 static int br_multicast_ipv6_rcv(struct net_bridge *br,
1452                                  struct net_bridge_port *port,
1453                                  struct sk_buff *skb)
1454 {
1455         struct sk_buff *skb2;
1456         struct ipv6hdr *ip6h;
1457         struct icmp6hdr *icmp6h;
1458         u8 nexthdr;
1459         unsigned len;
1460         int offset;
1461         int err;
1462
1463         if (!pskb_may_pull(skb, sizeof(*ip6h)))
1464                 return -EINVAL;
1465
1466         ip6h = ipv6_hdr(skb);
1467
1468         /*
1469          * We're interested in MLD messages only.
1470          *  - Version is 6
1471          *  - MLD has always Router Alert hop-by-hop option
1472          *  - But we do not support jumbrograms.
1473          */
1474         if (ip6h->version != 6 ||
1475             ip6h->nexthdr != IPPROTO_HOPOPTS ||
1476             ip6h->payload_len == 0)
1477                 return 0;
1478
1479         len = ntohs(ip6h->payload_len);
1480         if (skb->len < len)
1481                 return -EINVAL;
1482
1483         nexthdr = ip6h->nexthdr;
1484         offset = ipv6_skip_exthdr(skb, sizeof(*ip6h), &nexthdr);
1485
1486         if (offset < 0 || nexthdr != IPPROTO_ICMPV6)
1487                 return 0;
1488
1489         /* Okay, we found ICMPv6 header */
1490         skb2 = skb_clone(skb, GFP_ATOMIC);
1491         if (!skb2)
1492                 return -ENOMEM;
1493
1494         err = -EINVAL;
1495         if (!pskb_may_pull(skb2, offset + sizeof(struct icmp6hdr)))
1496                 goto out;
1497
1498         len -= offset - skb_network_offset(skb2);
1499
1500         __skb_pull(skb2, offset);
1501         skb_reset_transport_header(skb2);
1502
1503         icmp6h = icmp6_hdr(skb2);
1504
1505         switch (icmp6h->icmp6_type) {
1506         case ICMPV6_MGM_QUERY:
1507         case ICMPV6_MGM_REPORT:
1508         case ICMPV6_MGM_REDUCTION:
1509         case ICMPV6_MLD2_REPORT:
1510                 break;
1511         default:
1512                 err = 0;
1513                 goto out;
1514         }
1515
1516         /* Okay, we found MLD message. Check further. */
1517         if (skb2->len > len) {
1518                 err = pskb_trim_rcsum(skb2, len);
1519                 if (err)
1520                         goto out;
1521         }
1522
1523         switch (skb2->ip_summed) {
1524         case CHECKSUM_COMPLETE:
1525                 if (!csum_fold(skb2->csum))
1526                         break;
1527                 /*FALLTHROUGH*/
1528         case CHECKSUM_NONE:
1529                 skb2->csum = 0;
1530                 if (skb_checksum_complete(skb2))
1531                         goto out;
1532         }
1533
1534         err = 0;
1535
1536         BR_INPUT_SKB_CB(skb)->igmp = 1;
1537
1538         switch (icmp6h->icmp6_type) {
1539         case ICMPV6_MGM_REPORT:
1540             {
1541                 struct mld_msg *mld;
1542                 if (!pskb_may_pull(skb2, sizeof(*mld))) {
1543                         err = -EINVAL;
1544                         goto out;
1545                 }
1546                 mld = (struct mld_msg *)skb_transport_header(skb2);
1547                 BR_INPUT_SKB_CB(skb2)->mrouters_only = 1;
1548                 err = br_ip6_multicast_add_group(br, port, &mld->mld_mca);
1549                 break;
1550             }
1551         case ICMPV6_MLD2_REPORT:
1552                 err = br_ip6_multicast_mld2_report(br, port, skb2);
1553                 break;
1554         case ICMPV6_MGM_QUERY:
1555                 err = br_ip6_multicast_query(br, port, skb2);
1556                 break;
1557         case ICMPV6_MGM_REDUCTION:
1558             {
1559                 struct mld_msg *mld;
1560                 if (!pskb_may_pull(skb2, sizeof(*mld))) {
1561                         err = -EINVAL;
1562                         goto out;
1563                 }
1564                 mld = (struct mld_msg *)skb_transport_header(skb2);
1565                 br_ip6_multicast_leave_group(br, port, &mld->mld_mca);
1566             }
1567         }
1568
1569 out:
1570         kfree_skb(skb2);
1571         return err;
1572 }
1573 #endif
1574
1575 int br_multicast_rcv(struct net_bridge *br, struct net_bridge_port *port,
1576                      struct sk_buff *skb)
1577 {
1578         BR_INPUT_SKB_CB(skb)->igmp = 0;
1579         BR_INPUT_SKB_CB(skb)->mrouters_only = 0;
1580
1581         if (br->multicast_disabled)
1582                 return 0;
1583
1584         switch (skb->protocol) {
1585         case htons(ETH_P_IP):
1586                 return br_multicast_ipv4_rcv(br, port, skb);
1587 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
1588         case htons(ETH_P_IPV6):
1589                 return br_multicast_ipv6_rcv(br, port, skb);
1590 #endif
1591         }
1592
1593         return 0;
1594 }
1595
1596 static void br_multicast_query_expired(unsigned long data)
1597 {
1598         struct net_bridge *br = (void *)data;
1599
1600         spin_lock(&br->multicast_lock);
1601         if (br->multicast_startup_queries_sent <
1602             br->multicast_startup_query_count)
1603                 br->multicast_startup_queries_sent++;
1604
1605         br_multicast_send_query(br, NULL, br->multicast_startup_queries_sent);
1606
1607         spin_unlock(&br->multicast_lock);
1608 }
1609
1610 void br_multicast_init(struct net_bridge *br)
1611 {
1612         br->hash_elasticity = 4;
1613         br->hash_max = 512;
1614
1615         br->multicast_router = 1;
1616         br->multicast_last_member_count = 2;
1617         br->multicast_startup_query_count = 2;
1618
1619         br->multicast_last_member_interval = HZ;
1620         br->multicast_query_response_interval = 10 * HZ;
1621         br->multicast_startup_query_interval = 125 * HZ / 4;
1622         br->multicast_query_interval = 125 * HZ;
1623         br->multicast_querier_interval = 255 * HZ;
1624         br->multicast_membership_interval = 260 * HZ;
1625
1626         spin_lock_init(&br->multicast_lock);
1627         setup_timer(&br->multicast_router_timer,
1628                     br_multicast_local_router_expired, 0);
1629         setup_timer(&br->multicast_querier_timer,
1630                     br_multicast_local_router_expired, 0);
1631         setup_timer(&br->multicast_query_timer, br_multicast_query_expired,
1632                     (unsigned long)br);
1633 }
1634
1635 void br_multicast_open(struct net_bridge *br)
1636 {
1637         br->multicast_startup_queries_sent = 0;
1638
1639         if (br->multicast_disabled)
1640                 return;
1641
1642         mod_timer(&br->multicast_query_timer, jiffies);
1643 }
1644
1645 void br_multicast_stop(struct net_bridge *br)
1646 {
1647         struct net_bridge_mdb_htable *mdb;
1648         struct net_bridge_mdb_entry *mp;
1649         struct hlist_node *p, *n;
1650         u32 ver;
1651         int i;
1652
1653         del_timer_sync(&br->multicast_router_timer);
1654         del_timer_sync(&br->multicast_querier_timer);
1655         del_timer_sync(&br->multicast_query_timer);
1656
1657         spin_lock_bh(&br->multicast_lock);
1658         mdb = mlock_dereference(br->mdb, br);
1659         if (!mdb)
1660                 goto out;
1661
1662         br->mdb = NULL;
1663
1664         ver = mdb->ver;
1665         for (i = 0; i < mdb->max; i++) {
1666                 hlist_for_each_entry_safe(mp, p, n, &mdb->mhash[i],
1667                                           hlist[ver]) {
1668                         del_timer(&mp->timer);
1669                         del_timer(&mp->query_timer);
1670                         call_rcu_bh(&mp->rcu, br_multicast_free_group);
1671                 }
1672         }
1673
1674         if (mdb->old) {
1675                 spin_unlock_bh(&br->multicast_lock);
1676                 rcu_barrier_bh();
1677                 spin_lock_bh(&br->multicast_lock);
1678                 WARN_ON(mdb->old);
1679         }
1680
1681         mdb->old = mdb;
1682         call_rcu_bh(&mdb->rcu, br_mdb_free);
1683
1684 out:
1685         spin_unlock_bh(&br->multicast_lock);
1686 }
1687
1688 int br_multicast_set_router(struct net_bridge *br, unsigned long val)
1689 {
1690         int err = -ENOENT;
1691
1692         spin_lock_bh(&br->multicast_lock);
1693         if (!netif_running(br->dev))
1694                 goto unlock;
1695
1696         switch (val) {
1697         case 0:
1698         case 2:
1699                 del_timer(&br->multicast_router_timer);
1700                 /* fall through */
1701         case 1:
1702                 br->multicast_router = val;
1703                 err = 0;
1704                 break;
1705
1706         default:
1707                 err = -EINVAL;
1708                 break;
1709         }
1710
1711 unlock:
1712         spin_unlock_bh(&br->multicast_lock);
1713
1714         return err;
1715 }
1716
1717 int br_multicast_set_port_router(struct net_bridge_port *p, unsigned long val)
1718 {
1719         struct net_bridge *br = p->br;
1720         int err = -ENOENT;
1721
1722         spin_lock(&br->multicast_lock);
1723         if (!netif_running(br->dev) || p->state == BR_STATE_DISABLED)
1724                 goto unlock;
1725
1726         switch (val) {
1727         case 0:
1728         case 1:
1729         case 2:
1730                 p->multicast_router = val;
1731                 err = 0;
1732
1733                 if (val < 2 && !hlist_unhashed(&p->rlist))
1734                         hlist_del_init_rcu(&p->rlist);
1735
1736                 if (val == 1)
1737                         break;
1738
1739                 del_timer(&p->multicast_router_timer);
1740
1741                 if (val == 0)
1742                         break;
1743
1744                 br_multicast_add_router(br, p);
1745                 break;
1746
1747         default:
1748                 err = -EINVAL;
1749                 break;
1750         }
1751
1752 unlock:
1753         spin_unlock(&br->multicast_lock);
1754
1755         return err;
1756 }
1757
1758 int br_multicast_toggle(struct net_bridge *br, unsigned long val)
1759 {
1760         struct net_bridge_port *port;
1761         int err = 0;
1762         struct net_bridge_mdb_htable *mdb;
1763
1764         spin_lock(&br->multicast_lock);
1765         if (br->multicast_disabled == !val)
1766                 goto unlock;
1767
1768         br->multicast_disabled = !val;
1769         if (br->multicast_disabled)
1770                 goto unlock;
1771
1772         if (!netif_running(br->dev))
1773                 goto unlock;
1774
1775         mdb = mlock_dereference(br->mdb, br);
1776         if (mdb) {
1777                 if (mdb->old) {
1778                         err = -EEXIST;
1779 rollback:
1780                         br->multicast_disabled = !!val;
1781                         goto unlock;
1782                 }
1783
1784                 err = br_mdb_rehash(&br->mdb, mdb->max,
1785                                     br->hash_elasticity);
1786                 if (err)
1787                         goto rollback;
1788         }
1789
1790         br_multicast_open(br);
1791         list_for_each_entry(port, &br->port_list, list) {
1792                 if (port->state == BR_STATE_DISABLED ||
1793                     port->state == BR_STATE_BLOCKING)
1794                         continue;
1795
1796                 __br_multicast_enable_port(port);
1797         }
1798
1799 unlock:
1800         spin_unlock(&br->multicast_lock);
1801
1802         return err;
1803 }
1804
1805 int br_multicast_set_hash_max(struct net_bridge *br, unsigned long val)
1806 {
1807         int err = -ENOENT;
1808         u32 old;
1809         struct net_bridge_mdb_htable *mdb;
1810
1811         spin_lock(&br->multicast_lock);
1812         if (!netif_running(br->dev))
1813                 goto unlock;
1814
1815         err = -EINVAL;
1816         if (!is_power_of_2(val))
1817                 goto unlock;
1818
1819         mdb = mlock_dereference(br->mdb, br);
1820         if (mdb && val < mdb->size)
1821                 goto unlock;
1822
1823         err = 0;
1824
1825         old = br->hash_max;
1826         br->hash_max = val;
1827
1828         if (mdb) {
1829                 if (mdb->old) {
1830                         err = -EEXIST;
1831 rollback:
1832                         br->hash_max = old;
1833                         goto unlock;
1834                 }
1835
1836                 err = br_mdb_rehash(&br->mdb, br->hash_max,
1837                                     br->hash_elasticity);
1838                 if (err)
1839                         goto rollback;
1840         }
1841
1842 unlock:
1843         spin_unlock(&br->multicast_lock);
1844
1845         return err;
1846 }