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_transient_multicast(const struct in6_addr *addr)
  41 {
  42         if (ipv6_addr_is_multicast(addr) && IPV6_ADDR_MC_FLAG_TRANSIENT(addr))
  43                 return 1;
  44         return 0;
  45 }
  46 #endif
  47
  48 static inline int br_ip_equal(const struct br_ip *a, const struct br_ip *b)
  49 {
  50         if (a->proto != b->proto)
  51                 return 0;
  52         switch (a->proto) {
  53         case htons(ETH_P_IP):
  54                 return a->u.ip4 == b->u.ip4;
  55 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
  56         case htons(ETH_P_IPV6):
  57                 return ipv6_addr_equal(&a->u.ip6, &b->u.ip6);
  58 #endif
  59         }
  60         return 0;
  61 }
  62
  63 static inline int __br_ip4_hash(struct net_bridge_mdb_htable *mdb, __be32 ip)
  64 {
  65         return jhash_1word(mdb->secret, (__force u32)ip) & (mdb->max - 1);
  66 }
  67
  68 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
  69 static inline int __br_ip6_hash(struct net_bridge_mdb_htable *mdb,
  70                                 const struct in6_addr *ip)
  71 {
  72         return jhash2((__force u32 *)ip->s6_addr32, 4, mdb->secret) & (mdb->max - 1);
  73 }
  74 #endif
  75
  76 static inline int br_ip_hash(struct net_bridge_mdb_htable *mdb,
  77                              struct br_ip *ip)
  78 {
  79         switch (ip->proto) {
  80         case htons(ETH_P_IP):
  81                 return __br_ip4_hash(mdb, ip->u.ip4);
  82 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
  83         case htons(ETH_P_IPV6):
  84                 return __br_ip6_hash(mdb, &ip->u.ip6);
  85 #endif
  86         }
  87         return 0;
  88 }
  89
  90 static struct net_bridge_mdb_entry *__br_mdb_ip_get(
  91         struct net_bridge_mdb_htable *mdb, struct br_ip *dst, int hash)
  92 {
  93         struct net_bridge_mdb_entry *mp;
  94         struct hlist_node *p;
  95
  96         hlist_for_each_entry_rcu(mp, p, &mdb->mhash[hash], hlist[mdb->ver]) {
  97                 if (br_ip_equal(&mp->addr, dst))
  98                         return mp;
  99         }
 100
 101         return NULL;
 102 }
 103
 104 static struct net_bridge_mdb_entry *br_mdb_ip_get(
 105         struct net_bridge_mdb_htable *mdb, struct br_ip *dst)
 106 {
 107         if (!mdb)
 108                 return NULL;
 109
 110         return __br_mdb_ip_get(mdb, dst, br_ip_hash(mdb, dst));
 111 }
 112
 113 static struct net_bridge_mdb_entry *br_mdb_ip4_get(
 114         struct net_bridge_mdb_htable *mdb, __be32 dst)
 115 {
 116         struct br_ip br_dst;
 117
 118         br_dst.u.ip4 = dst;
 119         br_dst.proto = htons(ETH_P_IP);
 120
 121         return br_mdb_ip_get(mdb, &br_dst);
 122 }
 123
 124 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
 125 static struct net_bridge_mdb_entry *br_mdb_ip6_get(
 126         struct net_bridge_mdb_htable *mdb, const struct in6_addr *dst)
 127 {
 128         struct br_ip br_dst;
 129
 130         ipv6_addr_copy(&br_dst.u.ip6, dst);
 131         br_dst.proto = htons(ETH_P_IPV6);
 132
 133         return br_mdb_ip_get(mdb, &br_dst);
 134 }
 135 #endif
 136
 137 struct net_bridge_mdb_entry *br_mdb_get(struct net_bridge *br,
 138                                         struct sk_buff *skb)
 139 {
 140         struct net_bridge_mdb_htable *mdb = rcu_dereference(br->mdb);
 141         struct br_ip ip;
 142
 143         if (br->multicast_disabled)
 144                 return NULL;
 145
 146         if (BR_INPUT_SKB_CB(skb)->igmp)
 147                 return NULL;
 148
 149         ip.proto = skb->protocol;
 150
 151         switch (skb->protocol) {
 152         case htons(ETH_P_IP):
 153                 ip.u.ip4 = ip_hdr(skb)->daddr;
 154                 break;
 155 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
 156         case htons(ETH_P_IPV6):
 157                 ipv6_addr_copy(&ip.u.ip6, &ipv6_hdr(skb)->daddr);
 158                 break;
 159 #endif
 160         default:
 161                 return NULL;
 162         }
 163
 164         return br_mdb_ip_get(mdb, &ip);
 165 }
 166
 167 static void br_mdb_free(struct rcu_head *head)
 168 {
 169         struct net_bridge_mdb_htable *mdb =
 170                 container_of(head, struct net_bridge_mdb_htable, rcu);
 171         struct net_bridge_mdb_htable *old = mdb->old;
 172
 173         mdb->old = NULL;
 174         kfree(old->mhash);
 175         kfree(old);
 176 }
 177
 178 static int br_mdb_copy(struct net_bridge_mdb_htable *new,
 179                        struct net_bridge_mdb_htable *old,
 180                        int elasticity)
 181 {
 182         struct net_bridge_mdb_entry *mp;
 183         struct hlist_node *p;
 184         int maxlen;
 185         int len;
 186         int i;
 187
 188         for (i = 0; i < old->max; i++)
 189                 hlist_for_each_entry(mp, p, &old->mhash[i], hlist[old->ver])
 190                         hlist_add_head(&mp->hlist[new->ver],
 191                                        &new->mhash[br_ip_hash(new, &mp->addr)]);
 192
 193         if (!elasticity)
 194                 return 0;
 195
 196         maxlen = 0;
 197         for (i = 0; i < new->max; i++) {
 198                 len = 0;
 199                 hlist_for_each_entry(mp, p, &new->mhash[i], hlist[new->ver])
 200                         len++;
 201                 if (len > maxlen)
 202                         maxlen = len;
 203         }
 204
 205         return maxlen > elasticity ? -EINVAL : 0;
 206 }
 207
 208 static void br_multicast_free_pg(struct rcu_head *head)
 209 {
 210         struct net_bridge_port_group *p =
 211                 container_of(head, struct net_bridge_port_group, rcu);
 212
 213         kfree(p);
 214 }
 215
 216 static void br_multicast_free_group(struct rcu_head *head)
 217 {
 218         struct net_bridge_mdb_entry *mp =
 219                 container_of(head, struct net_bridge_mdb_entry, rcu);
 220
 221         kfree(mp);
 222 }
 223
 224 static void br_multicast_group_expired(unsigned long data)
 225 {
 226         struct net_bridge_mdb_entry *mp = (void *)data;
 227         struct net_bridge *br = mp->br;
 228         struct net_bridge_mdb_htable *mdb;
 229
 230         spin_lock(&br->multicast_lock);
 231         if (!netif_running(br->dev) || timer_pending(&mp->timer))
 232                 goto out;
 233
 234         mp->mglist = false;
 235
 236         if (mp->ports)
 237                 goto out;
 238
 239         mdb = mlock_dereference(br->mdb, br);
 240
 241         hlist_del_rcu(&mp->hlist[mdb->ver]);
 242         mdb->size--;
 243
 244         del_timer(&mp->query_timer);
 245         call_rcu_bh(&mp->rcu, br_multicast_free_group);
 246
 247 out:
 248         spin_unlock(&br->multicast_lock);
 249 }
 250
 251 static void br_multicast_del_pg(struct net_bridge *br,
 252                                 struct net_bridge_port_group *pg)
 253 {
 254         struct net_bridge_mdb_htable *mdb;
 255         struct net_bridge_mdb_entry *mp;
 256         struct net_bridge_port_group *p;
 257         struct net_bridge_port_group __rcu **pp;
 258
 259         mdb = mlock_dereference(br->mdb, br);
 260
 261         mp = br_mdb_ip_get(mdb, &pg->addr);
 262         if (WARN_ON(!mp))
 263                 return;
 264
 265         for (pp = &mp->ports;
 266              (p = mlock_dereference(*pp, br)) != NULL;
 267              pp = &p->next) {
 268                 if (p != pg)
 269                         continue;
 270
 271                 rcu_assign_pointer(*pp, p->next);
 272                 hlist_del_init(&p->mglist);
 273                 del_timer(&p->timer);
 274                 del_timer(&p->query_timer);
 275                 call_rcu_bh(&p->rcu, br_multicast_free_pg);
 276
 277                 if (!mp->ports && !mp->mglist &&
 278                     netif_running(br->dev))
 279                         mod_timer(&mp->timer, jiffies);
 280
 281                 return;
 282         }
 283
 284         WARN_ON(1);
 285 }
 286
 287 static void br_multicast_port_group_expired(unsigned long data)
 288 {
 289         struct net_bridge_port_group *pg = (void *)data;
 290         struct net_bridge *br = pg->port->br;
 291
 292         spin_lock(&br->multicast_lock);
 293         if (!netif_running(br->dev) || timer_pending(&pg->timer) ||
 294             hlist_unhashed(&pg->mglist))
 295                 goto out;
 296
 297         br_multicast_del_pg(br, pg);
 298
 299 out:
 300         spin_unlock(&br->multicast_lock);
 301 }
 302
 303 static int br_mdb_rehash(struct net_bridge_mdb_htable __rcu **mdbp, int max,
 304                          int elasticity)
 305 {
 306         struct net_bridge_mdb_htable *old = rcu_dereference_protected(*mdbp, 1);
 307         struct net_bridge_mdb_htable *mdb;
 308         int err;
 309
 310         mdb = kmalloc(sizeof(*mdb), GFP_ATOMIC);
 311         if (!mdb)
 312                 return -ENOMEM;
 313
 314         mdb->max = max;
 315         mdb->old = old;
 316
 317         mdb->mhash = kzalloc(max * sizeof(*mdb->mhash), GFP_ATOMIC);
 318         if (!mdb->mhash) {
 319                 kfree(mdb);
 320                 return -ENOMEM;
 321         }
 322
 323         mdb->size = old ? old->size : 0;
 324         mdb->ver = old ? old->ver ^ 1 : 0;
 325
 326         if (!old || elasticity)
 327                 get_random_bytes(&mdb->secret, sizeof(mdb->secret));
 328         else
 329                 mdb->secret = old->secret;
 330
 331         if (!old)
 332                 goto out;
 333
 334         err = br_mdb_copy(mdb, old, elasticity);
 335         if (err) {
 336                 kfree(mdb->mhash);
 337                 kfree(mdb);
 338                 return err;
 339         }
 340
 341         call_rcu_bh(&mdb->rcu, br_mdb_free);
 342
 343 out:
 344         rcu_assign_pointer(*mdbp, mdb);
 345
 346         return 0;
 347 }
 348
 349 static struct sk_buff *br_ip4_multicast_alloc_query(struct net_bridge *br,
 350                                                     __be32 group)
 351 {
 352         struct sk_buff *skb;
 353         struct igmphdr *ih;
 354         struct ethhdr *eth;
 355         struct iphdr *iph;
 356
 357         skb = netdev_alloc_skb_ip_align(br->dev, sizeof(*eth) + sizeof(*iph) +
 358                                                  sizeof(*ih) + 4);
 359         if (!skb)
 360                 goto out;
 361
 362         skb->protocol = htons(ETH_P_IP);
 363
 364         skb_reset_mac_header(skb);
 365         eth = eth_hdr(skb);
 366
 367         memcpy(eth->h_source, br->dev->dev_addr, 6);
 368         eth->h_dest[0] = 1;
 369         eth->h_dest[1] = 0;
 370         eth->h_dest[2] = 0x5e;
 371         eth->h_dest[3] = 0;
 372         eth->h_dest[4] = 0;
 373         eth->h_dest[5] = 1;
 374         eth->h_proto = htons(ETH_P_IP);
 375         skb_put(skb, sizeof(*eth));
 376
 377         skb_set_network_header(skb, skb->len);
 378         iph = ip_hdr(skb);
 379
 380         iph->version = 4;
 381         iph->ihl = 6;
 382         iph->tos = 0xc0;
 383         iph->tot_len = htons(sizeof(*iph) + sizeof(*ih) + 4);
 384         iph->id = 0;
 385         iph->frag_off = htons(IP_DF);
 386         iph->ttl = 1;
 387         iph->protocol = IPPROTO_IGMP;
 388         iph->saddr = 0;
 389         iph->daddr = htonl(INADDR_ALLHOSTS_GROUP);
 390         ((u8 *)&iph[1])[0] = IPOPT_RA;
 391         ((u8 *)&iph[1])[1] = 4;
 392         ((u8 *)&iph[1])[2] = 0;
 393         ((u8 *)&iph[1])[3] = 0;
 394         ip_send_check(iph);
 395         skb_put(skb, 24);
 396
 397         skb_set_transport_header(skb, skb->len);
 398         ih = igmp_hdr(skb);
 399         ih->type = IGMP_HOST_MEMBERSHIP_QUERY;
 400         ih->code = (group ? br->multicast_last_member_interval :
 401                             br->multicast_query_response_interval) /
 402                    (HZ / IGMP_TIMER_SCALE);
 403         ih->group = group;
 404         ih->csum = 0;
 405         ih->csum = ip_compute_csum((void *)ih, sizeof(struct igmphdr));
 406         skb_put(skb, sizeof(*ih));
 407
 408         __skb_pull(skb, sizeof(*eth));
 409
 410 out:
 411         return skb;
 412 }
 413
 414 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
 415 static struct sk_buff *br_ip6_multicast_alloc_query(struct net_bridge *br,
 416                                                     struct in6_addr *group)
 417 {
 418         struct sk_buff *skb;
 419         struct ipv6hdr *ip6h;
 420         struct mld_msg *mldq;
 421         struct ethhdr *eth;
 422         u8 *hopopt;
 423         unsigned long interval;
 424
 425         skb = netdev_alloc_skb_ip_align(br->dev, sizeof(*eth) + sizeof(*ip6h) +
 426                                                  8 + sizeof(*mldq));
 427         if (!skb)
 428                 goto out;
 429
 430         skb->protocol = htons(ETH_P_IPV6);
 431
 432         /* Ethernet header */
 433         skb_reset_mac_header(skb);
 434         eth = eth_hdr(skb);
 435
 436         memcpy(eth->h_source, br->dev->dev_addr, 6);
 437         eth->h_proto = htons(ETH_P_IPV6);
 438         skb_put(skb, sizeof(*eth));
 439
 440         /* IPv6 header + HbH option */
 441         skb_set_network_header(skb, skb->len);
 442         ip6h = ipv6_hdr(skb);
 443
 444         *(__force __be32 *)ip6h = htonl(0x60000000);
 445         ip6h->payload_len = htons(8 + sizeof(*mldq));
 446         ip6h->nexthdr = IPPROTO_HOPOPTS;
 447         ip6h->hop_limit = 1;
 448         ipv6_addr_set(&ip6h->saddr, 0, 0, 0, 0);
 449         ipv6_addr_set(&ip6h->daddr, htonl(0xff020000), 0, 0, htonl(1));
 450         ipv6_eth_mc_map(&ip6h->daddr, eth->h_dest);
 451
 452         hopopt = (u8 *)(ip6h + 1);
 453         hopopt[0] = IPPROTO_ICMPV6;             /* next hdr */
 454         hopopt[1] = 0;                          /* length of HbH */
 455         hopopt[2] = IPV6_TLV_ROUTERALERT;       /* Router Alert */
 456         hopopt[3] = 2;                          /* Length of RA Option */
 457         hopopt[4] = 0;                          /* Type = 0x0000 (MLD) */
 458         hopopt[5] = 0;
 459         hopopt[6] = IPV6_TLV_PAD0;              /* Pad0 */
 460         hopopt[7] = IPV6_TLV_PAD0;              /* Pad0 */
 461
 462         skb_put(skb, sizeof(*ip6h) + 8);
 463
 464         /* ICMPv6 */
 465         skb_set_transport_header(skb, skb->len);
 466         mldq = (struct mld_msg *) icmp6_hdr(skb);
 467
 468         interval = ipv6_addr_any(group) ? br->multicast_last_member_interval :
 469                                           br->multicast_query_response_interval;
 470
 471         mldq->mld_type = ICMPV6_MGM_QUERY;
 472         mldq->mld_code = 0;
 473         mldq->mld_cksum = 0;
 474         mldq->mld_maxdelay = htons((u16)jiffies_to_msecs(interval));
 475         mldq->mld_reserved = 0;
 476         ipv6_addr_copy(&mldq->mld_mca, group);
 477
 478         /* checksum */
 479         mldq->mld_cksum = csum_ipv6_magic(&ip6h->saddr, &ip6h->daddr,
 480                                           sizeof(*mldq), IPPROTO_ICMPV6,
 481                                           csum_partial(mldq,
 482                                                        sizeof(*mldq), 0));
 483         skb_put(skb, sizeof(*mldq));
 484
 485         __skb_pull(skb, sizeof(*eth));
 486
 487 out:
 488         return skb;
 489 }
 490 #endif
 491
 492 static struct sk_buff *br_multicast_alloc_query(struct net_bridge *br,
 493                                                 struct br_ip *addr)
 494 {
 495         switch (addr->proto) {
 496         case htons(ETH_P_IP):
 497                 return br_ip4_multicast_alloc_query(br, addr->u.ip4);
 498 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
 499         case htons(ETH_P_IPV6):
 500                 return br_ip6_multicast_alloc_query(br, &addr->u.ip6);
 501 #endif
 502         }
 503         return NULL;
 504 }
 505
 506 static void br_multicast_send_group_query(struct net_bridge_mdb_entry *mp)
 507 {
 508         struct net_bridge *br = mp->br;
 509         struct sk_buff *skb;
 510
 511         skb = br_multicast_alloc_query(br, &mp->addr);
 512         if (!skb)
 513                 goto timer;
 514
 515         netif_rx(skb);
 516
 517 timer:
 518         if (++mp->queries_sent < br->multicast_last_member_count)
 519                 mod_timer(&mp->query_timer,
 520                           jiffies + br->multicast_last_member_interval);
 521 }
 522
 523 static void br_multicast_group_query_expired(unsigned long data)
 524 {
 525         struct net_bridge_mdb_entry *mp = (void *)data;
 526         struct net_bridge *br = mp->br;
 527
 528         spin_lock(&br->multicast_lock);
 529         if (!netif_running(br->dev) || !mp->mglist ||
 530             mp->queries_sent >= br->multicast_last_member_count)
 531                 goto out;
 532
 533         br_multicast_send_group_query(mp);
 534
 535 out:
 536         spin_unlock(&br->multicast_lock);
 537 }
 538
 539 static void br_multicast_send_port_group_query(struct net_bridge_port_group *pg)
 540 {
 541         struct net_bridge_port *port = pg->port;
 542         struct net_bridge *br = port->br;
 543         struct sk_buff *skb;
 544
 545         skb = br_multicast_alloc_query(br, &pg->addr);
 546         if (!skb)
 547                 goto timer;
 548
 549         br_deliver(port, skb);
 550
 551 timer:
 552         if (++pg->queries_sent < br->multicast_last_member_count)
 553                 mod_timer(&pg->query_timer,
 554                           jiffies + br->multicast_last_member_interval);
 555 }
 556
 557 static void br_multicast_port_group_query_expired(unsigned long data)
 558 {
 559         struct net_bridge_port_group *pg = (void *)data;
 560         struct net_bridge_port *port = pg->port;
 561         struct net_bridge *br = port->br;
 562
 563         spin_lock(&br->multicast_lock);
 564         if (!netif_running(br->dev) || hlist_unhashed(&pg->mglist) ||
 565             pg->queries_sent >= br->multicast_last_member_count)
 566                 goto out;
 567
 568         br_multicast_send_port_group_query(pg);
 569
 570 out:
 571         spin_unlock(&br->multicast_lock);
 572 }
 573
 574 static struct net_bridge_mdb_entry *br_multicast_get_group(
 575         struct net_bridge *br, struct net_bridge_port *port,
 576         struct br_ip *group, int hash)
 577 {
 578         struct net_bridge_mdb_htable *mdb;
 579         struct net_bridge_mdb_entry *mp;
 580         struct hlist_node *p;
 581         unsigned count = 0;
 582         unsigned max;
 583         int elasticity;
 584         int err;
 585
 586         mdb = rcu_dereference_protected(br->mdb, 1);
 587         hlist_for_each_entry(mp, p, &mdb->mhash[hash], hlist[mdb->ver]) {
 588                 count++;
 589                 if (unlikely(br_ip_equal(group, &mp->addr)))
 590                         return mp;
 591         }
 592
 593         elasticity = 0;
 594         max = mdb->max;
 595
 596         if (unlikely(count > br->hash_elasticity && count)) {
 597                 if (net_ratelimit())
 598                         br_info(br, "Multicast hash table "
 599                                 "chain limit reached: %s\n",
 600                                 port ? port->dev->name : br->dev->name);
 601
 602                 elasticity = br->hash_elasticity;
 603         }
 604
 605         if (mdb->size >= max) {
 606                 max *= 2;
 607                 if (unlikely(max >= br->hash_max)) {
 608                         br_warn(br, "Multicast hash table maximum "
 609                                 "reached, disabling snooping: %s, %d\n",
 610                                 port ? port->dev->name : br->dev->name, max);
 611                         err = -E2BIG;
 612 disable:
 613                         br->multicast_disabled = 1;
 614                         goto err;
 615                 }
 616         }
 617
 618         if (max > mdb->max || elasticity) {
 619                 if (mdb->old) {
 620                         if (net_ratelimit())
 621                                 br_info(br, "Multicast hash table "
 622                                         "on fire: %s\n",
 623                                         port ? port->dev->name : br->dev->name);
 624                         err = -EEXIST;
 625                         goto err;
 626                 }
 627
 628                 err = br_mdb_rehash(&br->mdb, max, elasticity);
 629                 if (err) {
 630                         br_warn(br, "Cannot rehash multicast "
 631                                 "hash table, disabling snooping: %s, %d, %d\n",
 632                                 port ? port->dev->name : br->dev->name,
 633                                 mdb->size, err);
 634                         goto disable;
 635                 }
 636
 637                 err = -EAGAIN;
 638                 goto err;
 639         }
 640
 641         return NULL;
 642
 643 err:
 644         mp = ERR_PTR(err);
 645         return mp;
 646 }
 647
 648 static struct net_bridge_mdb_entry *br_multicast_new_group(
 649         struct net_bridge *br, struct net_bridge_port *port,
 650         struct br_ip *group)
 651 {
 652         struct net_bridge_mdb_htable *mdb;
 653         struct net_bridge_mdb_entry *mp;
 654         int hash;
 655         int err;
 656
 657         mdb = rcu_dereference_protected(br->mdb, 1);
 658         if (!mdb) {
 659                 err = br_mdb_rehash(&br->mdb, BR_HASH_SIZE, 0);
 660                 if (err)
 661                         return ERR_PTR(err);
 662                 goto rehash;
 663         }
 664
 665         hash = br_ip_hash(mdb, group);
 666         mp = br_multicast_get_group(br, port, group, hash);
 667         switch (PTR_ERR(mp)) {
 668         case 0:
 669                 break;
 670
 671         case -EAGAIN:
 672 rehash:
 673                 mdb = rcu_dereference_protected(br->mdb, 1);
 674                 hash = br_ip_hash(mdb, group);
 675                 break;
 676
 677         default:
 678                 goto out;
 679         }
 680
 681         mp = kzalloc(sizeof(*mp), GFP_ATOMIC);
 682         if (unlikely(!mp))
 683                 return ERR_PTR(-ENOMEM);
 684
 685         mp->br = br;
 686         mp->addr = *group;
 687         setup_timer(&mp->timer, br_multicast_group_expired,
 688                     (unsigned long)mp);
 689         setup_timer(&mp->query_timer, br_multicast_group_query_expired,
 690                     (unsigned long)mp);
 691
 692         hlist_add_head_rcu(&mp->hlist[mdb->ver], &mdb->mhash[hash]);
 693         mdb->size++;
 694
 695 out:
 696         return mp;
 697 }
 698
 699 static int br_multicast_add_group(struct net_bridge *br,
 700                                   struct net_bridge_port *port,
 701                                   struct br_ip *group)
 702 {
 703         struct net_bridge_mdb_entry *mp;
 704         struct net_bridge_port_group *p;
 705         struct net_bridge_port_group __rcu **pp;
 706         unsigned long now = jiffies;
 707         int err;
 708
 709         spin_lock(&br->multicast_lock);
 710         if (!netif_running(br->dev) ||
 711             (port && port->state == BR_STATE_DISABLED))
 712                 goto out;
 713
 714         mp = br_multicast_new_group(br, port, group);
 715         err = PTR_ERR(mp);
 716         if (IS_ERR(mp))
 717                 goto err;
 718
 719         if (!port) {
 720                 mp->mglist = true;
 721                 mod_timer(&mp->timer, now + br->multicast_membership_interval);
 722                 goto out;
 723         }
 724
 725         for (pp = &mp->ports;
 726              (p = mlock_dereference(*pp, br)) != NULL;
 727              pp = &p->next) {
 728                 if (p->port == port)
 729                         goto found;
 730                 if ((unsigned long)p->port < (unsigned long)port)
 731                         break;
 732         }
 733
 734         p = kzalloc(sizeof(*p), GFP_ATOMIC);
 735         err = -ENOMEM;
 736         if (unlikely(!p))
 737                 goto err;
 738
 739         p->addr = *group;
 740         p->port = port;
 741         p->next = *pp;
 742         hlist_add_head(&p->mglist, &port->mglist);
 743         setup_timer(&p->timer, br_multicast_port_group_expired,
 744                     (unsigned long)p);
 745         setup_timer(&p->query_timer, br_multicast_port_group_query_expired,
 746                     (unsigned long)p);
 747
 748         rcu_assign_pointer(*pp, p);
 749
 750 found:
 751         mod_timer(&p->timer, now + br->multicast_membership_interval);
 752 out:
 753         err = 0;
 754
 755 err:
 756         spin_unlock(&br->multicast_lock);
 757         return err;
 758 }
 759
 760 static int br_ip4_multicast_add_group(struct net_bridge *br,
 761                                       struct net_bridge_port *port,
 762                                       __be32 group)
 763 {
 764         struct br_ip br_group;
 765
 766         if (ipv4_is_local_multicast(group))
 767                 return 0;
 768
 769         br_group.u.ip4 = group;
 770         br_group.proto = htons(ETH_P_IP);
 771
 772         return br_multicast_add_group(br, port, &br_group);
 773 }
 774
 775 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
 776 static int br_ip6_multicast_add_group(struct net_bridge *br,
 777                                       struct net_bridge_port *port,
 778                                       const struct in6_addr *group)
 779 {
 780         struct br_ip br_group;
 781
 782         if (!ipv6_is_transient_multicast(group))
 783                 return 0;
 784
 785         ipv6_addr_copy(&br_group.u.ip6, group);
 786         br_group.proto = htons(ETH_P_IPV6);
 787
 788         return br_multicast_add_group(br, port, &br_group);
 789 }
 790 #endif
 791
 792 static void br_multicast_router_expired(unsigned long data)
 793 {
 794         struct net_bridge_port *port = (void *)data;
 795         struct net_bridge *br = port->br;
 796
 797         spin_lock(&br->multicast_lock);
 798         if (port->multicast_router != 1 ||
 799             timer_pending(&port->multicast_router_timer) ||
 800             hlist_unhashed(&port->rlist))
 801                 goto out;
 802
 803         hlist_del_init_rcu(&port->rlist);
 804
 805 out:
 806         spin_unlock(&br->multicast_lock);
 807 }
 808
 809 static void br_multicast_local_router_expired(unsigned long data)
 810 {
 811 }
 812
 813 static void __br_multicast_send_query(struct net_bridge *br,
 814                                       struct net_bridge_port *port,
 815                                       struct br_ip *ip)
 816 {
 817         struct sk_buff *skb;
 818
 819         skb = br_multicast_alloc_query(br, ip);
 820         if (!skb)
 821                 return;
 822
 823         if (port) {
 824                 __skb_push(skb, sizeof(struct ethhdr));
 825                 skb->dev = port->dev;
 826                 NF_HOOK(NFPROTO_BRIDGE, NF_BR_LOCAL_OUT, skb, NULL, skb->dev,
 827                         dev_queue_xmit);
 828         } else
 829                 netif_rx(skb);
 830 }
 831
 832 static void br_multicast_send_query(struct net_bridge *br,
 833                                     struct net_bridge_port *port, u32 sent)
 834 {
 835         unsigned long time;
 836         struct br_ip br_group;
 837
 838         if (!netif_running(br->dev) || br->multicast_disabled ||
 839             timer_pending(&br->multicast_querier_timer))
 840                 return;
 841
 842         memset(&br_group.u, 0, sizeof(br_group.u));
 843
 844         br_group.proto = htons(ETH_P_IP);
 845         __br_multicast_send_query(br, port, &br_group);
 846
 847 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
 848         br_group.proto = htons(ETH_P_IPV6);
 849         __br_multicast_send_query(br, port, &br_group);
 850 #endif
 851
 852         time = jiffies;
 853         time += sent < br->multicast_startup_query_count ?
 854                 br->multicast_startup_query_interval :
 855                 br->multicast_query_interval;
 856         mod_timer(port ? &port->multicast_query_timer :
 857                          &br->multicast_query_timer, time);
 858 }
 859
 860 static void br_multicast_port_query_expired(unsigned long data)
 861 {
 862         struct net_bridge_port *port = (void *)data;
 863         struct net_bridge *br = port->br;
 864
 865         spin_lock(&br->multicast_lock);
 866         if (port->state == BR_STATE_DISABLED ||
 867             port->state == BR_STATE_BLOCKING)
 868                 goto out;
 869
 870         if (port->multicast_startup_queries_sent <
 871             br->multicast_startup_query_count)
 872                 port->multicast_startup_queries_sent++;
 873
 874         br_multicast_send_query(port->br, port,
 875                                 port->multicast_startup_queries_sent);
 876
 877 out:
 878         spin_unlock(&br->multicast_lock);
 879 }
 880
 881 void br_multicast_add_port(struct net_bridge_port *port)
 882 {
 883         port->multicast_router = 1;
 884
 885         setup_timer(&port->multicast_router_timer, br_multicast_router_expired,
 886                     (unsigned long)port);
 887         setup_timer(&port->multicast_query_timer,
 888                     br_multicast_port_query_expired, (unsigned long)port);
 889 }
 890
 891 void br_multicast_del_port(struct net_bridge_port *port)
 892 {
 893         del_timer_sync(&port->multicast_router_timer);
 894 }
 895
 896 static void __br_multicast_enable_port(struct net_bridge_port *port)
 897 {
 898         port->multicast_startup_queries_sent = 0;
 899
 900         if (try_to_del_timer_sync(&port->multicast_query_timer) >= 0 ||
 901             del_timer(&port->multicast_query_timer))
 902                 mod_timer(&port->multicast_query_timer, jiffies);
 903 }
 904
 905 void br_multicast_enable_port(struct net_bridge_port *port)
 906 {
 907         struct net_bridge *br = port->br;
 908
 909         spin_lock(&br->multicast_lock);
 910         if (br->multicast_disabled || !netif_running(br->dev))
 911                 goto out;
 912
 913         __br_multicast_enable_port(port);
 914
 915 out:
 916         spin_unlock(&br->multicast_lock);
 917 }
 918
 919 void br_multicast_disable_port(struct net_bridge_port *port)
 920 {
 921         struct net_bridge *br = port->br;
 922         struct net_bridge_port_group *pg;
 923         struct hlist_node *p, *n;
 924
 925         spin_lock(&br->multicast_lock);
 926         hlist_for_each_entry_safe(pg, p, n, &port->mglist, mglist)
 927                 br_multicast_del_pg(br, pg);
 928
 929         if (!hlist_unhashed(&port->rlist))
 930                 hlist_del_init_rcu(&port->rlist);
 931         del_timer(&port->multicast_router_timer);
 932         del_timer(&port->multicast_query_timer);
 933         spin_unlock(&br->multicast_lock);
 934 }
 935
 936 static int br_ip4_multicast_igmp3_report(struct net_bridge *br,
 937                                          struct net_bridge_port *port,
 938                                          struct sk_buff *skb)
 939 {
 940         struct igmpv3_report *ih;
 941         struct igmpv3_grec *grec;
 942         int i;
 943         int len;
 944         int num;
 945         int type;
 946         int err = 0;
 947         __be32 group;
 948
 949         if (!pskb_may_pull(skb, sizeof(*ih)))
 950                 return -EINVAL;
 951
 952         ih = igmpv3_report_hdr(skb);
 953         num = ntohs(ih->ngrec);
 954         len = sizeof(*ih);
 955
 956         for (i = 0; i < num; i++) {
 957                 len += sizeof(*grec);
 958                 if (!pskb_may_pull(skb, len))
 959                         return -EINVAL;
 960
 961                 grec = (void *)(skb->data + len - sizeof(*grec));
 962                 group = grec->grec_mca;
 963                 type = grec->grec_type;
 964
 965                 len += ntohs(grec->grec_nsrcs) * 4;
 966                 if (!pskb_may_pull(skb, len))
 967                         return -EINVAL;
 968
 969                 /* We treat this as an IGMPv2 report for now. */
 970                 switch (type) {
 971                 case IGMPV3_MODE_IS_INCLUDE:
 972                 case IGMPV3_MODE_IS_EXCLUDE:
 973                 case IGMPV3_CHANGE_TO_INCLUDE:
 974                 case IGMPV3_CHANGE_TO_EXCLUDE:
 975                 case IGMPV3_ALLOW_NEW_SOURCES:
 976                 case IGMPV3_BLOCK_OLD_SOURCES:
 977                         break;
 978
 979                 default:
 980                         continue;
 981                 }
 982
 983                 err = br_ip4_multicast_add_group(br, port, group);
 984                 if (err)
 985                         break;
 986         }
 987
 988         return err;
 989 }
 990
 991 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
 992 static int br_ip6_multicast_mld2_report(struct net_bridge *br,
 993                                         struct net_bridge_port *port,
 994                                         struct sk_buff *skb)
 995 {
 996         struct icmp6hdr *icmp6h;
 997         struct mld2_grec *grec;
 998         int i;
 999         int len;
1000         int num;
1001         int err = 0;
1002
1003         if (!pskb_may_pull(skb, sizeof(*icmp6h)))
1004                 return -EINVAL;
1005
1006         icmp6h = icmp6_hdr(skb);
1007         num = ntohs(icmp6h->icmp6_dataun.un_data16[1]);
1008         len = sizeof(*icmp6h);
1009
1010         for (i = 0; i < num; i++) {
1011                 __be16 *nsrcs, _nsrcs;
1012
1013                 nsrcs = skb_header_pointer(skb,
1014                                            len + offsetof(struct mld2_grec,
1015                                                           grec_nsrcs),
1016                                            sizeof(_nsrcs), &_nsrcs);
1017                 if (!nsrcs)
1018                         return -EINVAL;
1019
1020                 if (!pskb_may_pull(skb,
1021                                    len + sizeof(*grec) +
1022                                    sizeof(struct in6_addr) * ntohs(*nsrcs)))
1023                         return -EINVAL;
1024
1025                 grec = (struct mld2_grec *)(skb->data + len);
1026                 len += sizeof(*grec) +
1027                        sizeof(struct in6_addr) * ntohs(*nsrcs);
1028
1029                 /* We treat these as MLDv1 reports for now. */
1030                 switch (grec->grec_type) {
1031                 case MLD2_MODE_IS_INCLUDE:
1032                 case MLD2_MODE_IS_EXCLUDE:
1033                 case MLD2_CHANGE_TO_INCLUDE:
1034                 case MLD2_CHANGE_TO_EXCLUDE:
1035                 case MLD2_ALLOW_NEW_SOURCES:
1036                 case MLD2_BLOCK_OLD_SOURCES:
1037                         break;
1038
1039                 default:
1040                         continue;
1041                 }
1042
1043                 err = br_ip6_multicast_add_group(br, port, &grec->grec_mca);
1044                 if (!err)
1045                         break;
1046         }
1047
1048         return err;
1049 }
1050 #endif
1051
1052 /*
1053  * Add port to rotuer_list
1054  *  list is maintained ordered by pointer value
1055  *  and locked by br->multicast_lock and RCU
1056  */
1057 static void br_multicast_add_router(struct net_bridge *br,
1058                                     struct net_bridge_port *port)
1059 {
1060         struct net_bridge_port *p;
1061         struct hlist_node *n, *slot = NULL;
1062
1063         hlist_for_each_entry(p, n, &br->router_list, rlist) {
1064                 if ((unsigned long) port >= (unsigned long) p)
1065                         break;
1066                 slot = n;
1067         }
1068
1069         if (slot)
1070                 hlist_add_after_rcu(slot, &port->rlist);
1071         else
1072                 hlist_add_head_rcu(&port->rlist, &br->router_list);
1073 }
1074
1075 static void br_multicast_mark_router(struct net_bridge *br,
1076                                      struct net_bridge_port *port)
1077 {
1078         unsigned long now = jiffies;
1079
1080         if (!port) {
1081                 if (br->multicast_router == 1)
1082                         mod_timer(&br->multicast_router_timer,
1083                                   now + br->multicast_querier_interval);
1084                 return;
1085         }
1086
1087         if (port->multicast_router != 1)
1088                 return;
1089
1090         if (!hlist_unhashed(&port->rlist))
1091                 goto timer;
1092
1093         br_multicast_add_router(br, port);
1094
1095 timer:
1096         mod_timer(&port->multicast_router_timer,
1097                   now + br->multicast_querier_interval);
1098 }
1099
1100 static void br_multicast_query_received(struct net_bridge *br,
1101                                         struct net_bridge_port *port,
1102                                         int saddr)
1103 {
1104         if (saddr)
1105                 mod_timer(&br->multicast_querier_timer,
1106                           jiffies + br->multicast_querier_interval);
1107         else if (timer_pending(&br->multicast_querier_timer))
1108                 return;
1109
1110         br_multicast_mark_router(br, port);
1111 }
1112
1113 static int br_ip4_multicast_query(struct net_bridge *br,
1114                                   struct net_bridge_port *port,
1115                                   struct sk_buff *skb)
1116 {
1117         struct iphdr *iph = ip_hdr(skb);
1118         struct igmphdr *ih = igmp_hdr(skb);
1119         struct net_bridge_mdb_entry *mp;
1120         struct igmpv3_query *ih3;
1121         struct net_bridge_port_group *p;
1122         struct net_bridge_port_group __rcu **pp;
1123         unsigned long max_delay;
1124         unsigned long now = jiffies;
1125         __be32 group;
1126         int err = 0;
1127
1128         spin_lock(&br->multicast_lock);
1129         if (!netif_running(br->dev) ||
1130             (port && port->state == BR_STATE_DISABLED))
1131                 goto out;
1132
1133         br_multicast_query_received(br, port, !!iph->saddr);
1134
1135         group = ih->group;
1136
1137         if (skb->len == sizeof(*ih)) {
1138                 max_delay = ih->code * (HZ / IGMP_TIMER_SCALE);
1139
1140                 if (!max_delay) {
1141                         max_delay = 10 * HZ;
1142                         group = 0;
1143                 }
1144         } else {
1145                 if (!pskb_may_pull(skb, sizeof(struct igmpv3_query))) {
1146                         err = -EINVAL;
1147                         goto out;
1148                 }
1149
1150                 ih3 = igmpv3_query_hdr(skb);
1151                 if (ih3->nsrcs)
1152                         goto out;
1153
1154                 max_delay = ih3->code ?
1155                             IGMPV3_MRC(ih3->code) * (HZ / IGMP_TIMER_SCALE) : 1;
1156         }
1157
1158         if (!group)
1159                 goto out;
1160
1161         mp = br_mdb_ip4_get(mlock_dereference(br->mdb, br), group);
1162         if (!mp)
1163                 goto out;
1164
1165         max_delay *= br->multicast_last_member_count;
1166
1167         if (mp->mglist &&
1168             (timer_pending(&mp->timer) ?
1169              time_after(mp->timer.expires, now + max_delay) :
1170              try_to_del_timer_sync(&mp->timer) >= 0))
1171                 mod_timer(&mp->timer, now + max_delay);
1172
1173         for (pp = &mp->ports;
1174              (p = mlock_dereference(*pp, br)) != NULL;
1175              pp = &p->next) {
1176                 if (timer_pending(&p->timer) ?
1177                     time_after(p->timer.expires, now + max_delay) :
1178                     try_to_del_timer_sync(&p->timer) >= 0)
1179                         mod_timer(&p->timer, now + max_delay);
1180         }
1181
1182 out:
1183         spin_unlock(&br->multicast_lock);
1184         return err;
1185 }
1186
1187 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
1188 static int br_ip6_multicast_query(struct net_bridge *br,
1189                                   struct net_bridge_port *port,
1190                                   struct sk_buff *skb)
1191 {
1192         struct ipv6hdr *ip6h = ipv6_hdr(skb);
1193         struct mld_msg *mld = (struct mld_msg *) icmp6_hdr(skb);
1194         struct net_bridge_mdb_entry *mp;
1195         struct mld2_query *mld2q;
1196         struct net_bridge_port_group *p;
1197         struct net_bridge_port_group __rcu **pp;
1198         unsigned long max_delay;
1199         unsigned long now = jiffies;
1200         struct in6_addr *group = NULL;
1201         int err = 0;
1202
1203         spin_lock(&br->multicast_lock);
1204         if (!netif_running(br->dev) ||
1205             (port && port->state == BR_STATE_DISABLED))
1206                 goto out;
1207
1208         br_multicast_query_received(br, port, !ipv6_addr_any(&ip6h->saddr));
1209
1210         if (skb->len == sizeof(*mld)) {
1211                 if (!pskb_may_pull(skb, sizeof(*mld))) {
1212                         err = -EINVAL;
1213                         goto out;
1214                 }
1215                 mld = (struct mld_msg *) icmp6_hdr(skb);
1216                 max_delay = msecs_to_jiffies(htons(mld->mld_maxdelay));
1217                 if (max_delay)
1218                         group = &mld->mld_mca;
1219         } else if (skb->len >= sizeof(*mld2q)) {
1220                 if (!pskb_may_pull(skb, sizeof(*mld2q))) {
1221                         err = -EINVAL;
1222                         goto out;
1223                 }
1224                 mld2q = (struct mld2_query *)icmp6_hdr(skb);
1225                 if (!mld2q->mld2q_nsrcs)
1226                         group = &mld2q->mld2q_mca;
1227                 max_delay = mld2q->mld2q_mrc ? MLDV2_MRC(mld2q->mld2q_mrc) : 1;
1228         }
1229
1230         if (!group)
1231                 goto out;
1232
1233         mp = br_mdb_ip6_get(mlock_dereference(br->mdb, br), group);
1234         if (!mp)
1235                 goto out;
1236
1237         max_delay *= br->multicast_last_member_count;
1238         if (mp->mglist &&
1239             (timer_pending(&mp->timer) ?
1240              time_after(mp->timer.expires, now + max_delay) :
1241              try_to_del_timer_sync(&mp->timer) >= 0))
1242                 mod_timer(&mp->timer, now + max_delay);
1243
1244         for (pp = &mp->ports;
1245              (p = mlock_dereference(*pp, br)) != NULL;
1246              pp = &p->next) {
1247                 if (timer_pending(&p->timer) ?
1248                     time_after(p->timer.expires, now + max_delay) :
1249                     try_to_del_timer_sync(&p->timer) >= 0)
1250                         mod_timer(&p->timer, now + max_delay);
1251         }
1252
1253 out:
1254         spin_unlock(&br->multicast_lock);
1255         return err;
1256 }
1257 #endif
1258
1259 static void br_multicast_leave_group(struct net_bridge *br,
1260                                      struct net_bridge_port *port,
1261                                      struct br_ip *group)
1262 {
1263         struct net_bridge_mdb_htable *mdb;
1264         struct net_bridge_mdb_entry *mp;
1265         struct net_bridge_port_group *p;
1266         unsigned long now;
1267         unsigned long time;
1268
1269         spin_lock(&br->multicast_lock);
1270         if (!netif_running(br->dev) ||
1271             (port && port->state == BR_STATE_DISABLED) ||
1272             timer_pending(&br->multicast_querier_timer))
1273                 goto out;
1274
1275         mdb = mlock_dereference(br->mdb, br);
1276         mp = br_mdb_ip_get(mdb, group);
1277         if (!mp)
1278                 goto out;
1279
1280         now = jiffies;
1281         time = now + br->multicast_last_member_count *
1282                      br->multicast_last_member_interval;
1283
1284         if (!port) {
1285                 if (mp->mglist &&
1286                     (timer_pending(&mp->timer) ?
1287                      time_after(mp->timer.expires, time) :
1288                      try_to_del_timer_sync(&mp->timer) >= 0)) {
1289                         mod_timer(&mp->timer, time);
1290
1291                         mp->queries_sent = 0;
1292                         mod_timer(&mp->query_timer, now);
1293                 }
1294
1295                 goto out;
1296         }
1297
1298         for (p = mlock_dereference(mp->ports, br);
1299              p != NULL;
1300              p = mlock_dereference(p->next, br)) {
1301                 if (p->port != port)
1302                         continue;
1303
1304                 if (!hlist_unhashed(&p->mglist) &&
1305                     (timer_pending(&p->timer) ?
1306                      time_after(p->timer.expires, time) :
1307                      try_to_del_timer_sync(&p->timer) >= 0)) {
1308                         mod_timer(&p->timer, time);
1309
1310                         p->queries_sent = 0;
1311                         mod_timer(&p->query_timer, now);
1312                 }
1313
1314                 break;
1315         }
1316
1317 out:
1318         spin_unlock(&br->multicast_lock);
1319 }
1320
1321 static void br_ip4_multicast_leave_group(struct net_bridge *br,
1322                                          struct net_bridge_port *port,
1323                                          __be32 group)
1324 {
1325         struct br_ip br_group;
1326
1327         if (ipv4_is_local_multicast(group))
1328                 return;
1329
1330         br_group.u.ip4 = group;
1331         br_group.proto = htons(ETH_P_IP);
1332
1333         br_multicast_leave_group(br, port, &br_group);
1334 }
1335
1336 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
1337 static void br_ip6_multicast_leave_group(struct net_bridge *br,
1338                                          struct net_bridge_port *port,
1339                                          const struct in6_addr *group)
1340 {
1341         struct br_ip br_group;
1342
1343         if (!ipv6_is_transient_multicast(group))
1344                 return;
1345
1346         ipv6_addr_copy(&br_group.u.ip6, group);
1347         br_group.proto = htons(ETH_P_IPV6);
1348
1349         br_multicast_leave_group(br, port, &br_group);
1350 }
1351 #endif
1352
1353 static int br_multicast_ipv4_rcv(struct net_bridge *br,
1354                                  struct net_bridge_port *port,
1355                                  struct sk_buff *skb)
1356 {
1357         struct sk_buff *skb2 = skb;
1358         struct iphdr *iph;
1359         struct igmphdr *ih;
1360         unsigned len;
1361         unsigned offset;
1362         int err;
1363
1364         /* We treat OOM as packet loss for now. */
1365         if (!pskb_may_pull(skb, sizeof(*iph)))
1366                 return -EINVAL;
1367
1368         iph = ip_hdr(skb);
1369
1370         if (iph->ihl < 5 || iph->version != 4)
1371                 return -EINVAL;
1372
1373         if (!pskb_may_pull(skb, ip_hdrlen(skb)))
1374                 return -EINVAL;
1375
1376         iph = ip_hdr(skb);
1377
1378         if (unlikely(ip_fast_csum((u8 *)iph, iph->ihl)))
1379                 return -EINVAL;
1380
1381         if (iph->protocol != IPPROTO_IGMP)
1382                 return 0;
1383
1384         len = ntohs(iph->tot_len);
1385         if (skb->len < len || len < ip_hdrlen(skb))
1386                 return -EINVAL;
1387
1388         if (skb->len > len) {
1389                 skb2 = skb_clone(skb, GFP_ATOMIC);
1390                 if (!skb2)
1391                         return -ENOMEM;
1392
1393                 err = pskb_trim_rcsum(skb2, len);
1394                 if (err)
1395                         goto err_out;
1396         }
1397
1398         len -= ip_hdrlen(skb2);
1399         offset = skb_network_offset(skb2) + ip_hdrlen(skb2);
1400         __skb_pull(skb2, offset);
1401         skb_reset_transport_header(skb2);
1402
1403         err = -EINVAL;
1404         if (!pskb_may_pull(skb2, sizeof(*ih)))
1405                 goto out;
1406
1407         switch (skb2->ip_summed) {
1408         case CHECKSUM_COMPLETE:
1409                 if (!csum_fold(skb2->csum))
1410                         break;
1411                 /* fall through */
1412         case CHECKSUM_NONE:
1413                 skb2->csum = 0;
1414                 if (skb_checksum_complete(skb2))
1415                         goto out;
1416         }
1417
1418         err = 0;
1419
1420         BR_INPUT_SKB_CB(skb)->igmp = 1;
1421         ih = igmp_hdr(skb2);
1422
1423         switch (ih->type) {
1424         case IGMP_HOST_MEMBERSHIP_REPORT:
1425         case IGMPV2_HOST_MEMBERSHIP_REPORT:
1426                 BR_INPUT_SKB_CB(skb2)->mrouters_only = 1;
1427                 err = br_ip4_multicast_add_group(br, port, ih->group);
1428                 break;
1429         case IGMPV3_HOST_MEMBERSHIP_REPORT:
1430                 err = br_ip4_multicast_igmp3_report(br, port, skb2);
1431                 break;
1432         case IGMP_HOST_MEMBERSHIP_QUERY:
1433                 err = br_ip4_multicast_query(br, port, skb2);
1434                 break;
1435         case IGMP_HOST_LEAVE_MESSAGE:
1436                 br_ip4_multicast_leave_group(br, port, ih->group);
1437                 break;
1438         }
1439
1440 out:
1441         __skb_push(skb2, offset);
1442 err_out:
1443         if (skb2 != skb)
1444                 kfree_skb(skb2);
1445         return err;
1446 }
1447
1448 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
1449 static int br_multicast_ipv6_rcv(struct net_bridge *br,
1450                                  struct net_bridge_port *port,
1451                                  struct sk_buff *skb)
1452 {
1453         struct sk_buff *skb2;
1454         struct ipv6hdr *ip6h;
1455         struct icmp6hdr *icmp6h;
1456         u8 nexthdr;
1457         unsigned len;
1458         int offset;
1459         int err;
1460
1461         if (!pskb_may_pull(skb, sizeof(*ip6h)))
1462                 return -EINVAL;
1463
1464         ip6h = ipv6_hdr(skb);
1465
1466         /*
1467          * We're interested in MLD messages only.
1468          *  - Version is 6
1469          *  - MLD has always Router Alert hop-by-hop option
1470          *  - But we do not support jumbrograms.
1471          */
1472         if (ip6h->version != 6 ||
1473             ip6h->nexthdr != IPPROTO_HOPOPTS ||
1474             ip6h->payload_len == 0)
1475                 return 0;
1476
1477         len = ntohs(ip6h->payload_len);
1478         if (skb->len < len)
1479                 return -EINVAL;
1480
1481         nexthdr = ip6h->nexthdr;
1482         offset = ipv6_skip_exthdr(skb, sizeof(*ip6h), &nexthdr);
1483
1484         if (offset < 0 || nexthdr != IPPROTO_ICMPV6)
1485                 return 0;
1486
1487         /* Okay, we found ICMPv6 header */
1488         skb2 = skb_clone(skb, GFP_ATOMIC);
1489         if (!skb2)
1490                 return -ENOMEM;
1491
1492         err = -EINVAL;
1493         if (!pskb_may_pull(skb2, offset + sizeof(struct icmp6hdr)))
1494                 goto out;
1495
1496         len -= offset - skb_network_offset(skb2);
1497
1498         __skb_pull(skb2, offset);
1499         skb_reset_transport_header(skb2);
1500
1501         icmp6h = icmp6_hdr(skb2);
1502
1503         switch (icmp6h->icmp6_type) {
1504         case ICMPV6_MGM_QUERY:
1505         case ICMPV6_MGM_REPORT:
1506         case ICMPV6_MGM_REDUCTION:
1507         case ICMPV6_MLD2_REPORT:
1508                 break;
1509         default:
1510                 err = 0;
1511                 goto out;
1512         }
1513
1514         /* Okay, we found MLD message. Check further. */
1515         if (skb2->len > len) {
1516                 err = pskb_trim_rcsum(skb2, len);
1517                 if (err)
1518                         goto out;
1519         }
1520
1521         switch (skb2->ip_summed) {
1522         case CHECKSUM_COMPLETE:
1523                 if (!csum_fold(skb2->csum))
1524                         break;
1525                 /*FALLTHROUGH*/
1526         case CHECKSUM_NONE:
1527                 skb2->csum = 0;
1528                 if (skb_checksum_complete(skb2))
1529                         goto out;
1530         }
1531
1532         err = 0;
1533
1534         BR_INPUT_SKB_CB(skb)->igmp = 1;
1535
1536         switch (icmp6h->icmp6_type) {
1537         case ICMPV6_MGM_REPORT:
1538             {
1539                 struct mld_msg *mld;
1540                 if (!pskb_may_pull(skb2, sizeof(*mld))) {
1541                         err = -EINVAL;
1542                         goto out;
1543                 }
1544                 mld = (struct mld_msg *)skb_transport_header(skb2);
1545                 BR_INPUT_SKB_CB(skb2)->mrouters_only = 1;
1546                 err = br_ip6_multicast_add_group(br, port, &mld->mld_mca);
1547                 break;
1548             }
1549         case ICMPV6_MLD2_REPORT:
1550                 err = br_ip6_multicast_mld2_report(br, port, skb2);
1551                 break;
1552         case ICMPV6_MGM_QUERY:
1553                 err = br_ip6_multicast_query(br, port, skb2);
1554                 break;
1555         case ICMPV6_MGM_REDUCTION:
1556             {
1557                 struct mld_msg *mld;
1558                 if (!pskb_may_pull(skb2, sizeof(*mld))) {
1559                         err = -EINVAL;
1560                         goto out;
1561                 }
1562                 mld = (struct mld_msg *)skb_transport_header(skb2);
1563                 br_ip6_multicast_leave_group(br, port, &mld->mld_mca);
1564             }
1565         }
1566
1567 out:
1568         kfree_skb(skb2);
1569         return err;
1570 }
1571 #endif
1572
1573 int br_multicast_rcv(struct net_bridge *br, struct net_bridge_port *port,
1574                      struct sk_buff *skb)
1575 {
1576         BR_INPUT_SKB_CB(skb)->igmp = 0;
1577         BR_INPUT_SKB_CB(skb)->mrouters_only = 0;
1578
1579         if (br->multicast_disabled)
1580                 return 0;
1581
1582         switch (skb->protocol) {
1583         case htons(ETH_P_IP):
1584                 return br_multicast_ipv4_rcv(br, port, skb);
1585 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
1586         case htons(ETH_P_IPV6):
1587                 return br_multicast_ipv6_rcv(br, port, skb);
1588 #endif
1589         }
1590
1591         return 0;
1592 }
1593
1594 static void br_multicast_query_expired(unsigned long data)
1595 {
1596         struct net_bridge *br = (void *)data;
1597
1598         spin_lock(&br->multicast_lock);
1599         if (br->multicast_startup_queries_sent <
1600             br->multicast_startup_query_count)
1601                 br->multicast_startup_queries_sent++;
1602
1603         br_multicast_send_query(br, NULL, br->multicast_startup_queries_sent);
1604
1605         spin_unlock(&br->multicast_lock);
1606 }
1607
1608 void br_multicast_init(struct net_bridge *br)
1609 {
1610         br->hash_elasticity = 4;
1611         br->hash_max = 512;
1612
1613         br->multicast_router = 1;
1614         br->multicast_last_member_count = 2;
1615         br->multicast_startup_query_count = 2;
1616
1617         br->multicast_last_member_interval = HZ;
1618         br->multicast_query_response_interval = 10 * HZ;
1619         br->multicast_startup_query_interval = 125 * HZ / 4;
1620         br->multicast_query_interval = 125 * HZ;
1621         br->multicast_querier_interval = 255 * HZ;
1622         br->multicast_membership_interval = 260 * HZ;
1623
1624         spin_lock_init(&br->multicast_lock);
1625         setup_timer(&br->multicast_router_timer,
1626                     br_multicast_local_router_expired, 0);
1627         setup_timer(&br->multicast_querier_timer,
1628                     br_multicast_local_router_expired, 0);
1629         setup_timer(&br->multicast_query_timer, br_multicast_query_expired,
1630                     (unsigned long)br);
1631 }
1632
1633 void br_multicast_open(struct net_bridge *br)
1634 {
1635         br->multicast_startup_queries_sent = 0;
1636
1637         if (br->multicast_disabled)
1638                 return;
1639
1640         mod_timer(&br->multicast_query_timer, jiffies);
1641 }
1642
1643 void br_multicast_stop(struct net_bridge *br)
1644 {
1645         struct net_bridge_mdb_htable *mdb;
1646         struct net_bridge_mdb_entry *mp;
1647         struct hlist_node *p, *n;
1648         u32 ver;
1649         int i;
1650
1651         del_timer_sync(&br->multicast_router_timer);
1652         del_timer_sync(&br->multicast_querier_timer);
1653         del_timer_sync(&br->multicast_query_timer);
1654
1655         spin_lock_bh(&br->multicast_lock);
1656         mdb = mlock_dereference(br->mdb, br);
1657         if (!mdb)
1658                 goto out;
1659
1660         br->mdb = NULL;
1661
1662         ver = mdb->ver;
1663         for (i = 0; i < mdb->max; i++) {
1664                 hlist_for_each_entry_safe(mp, p, n, &mdb->mhash[i],
1665                                           hlist[ver]) {
1666                         del_timer(&mp->timer);
1667                         del_timer(&mp->query_timer);
1668                         call_rcu_bh(&mp->rcu, br_multicast_free_group);
1669                 }
1670         }
1671
1672         if (mdb->old) {
1673                 spin_unlock_bh(&br->multicast_lock);
1674                 rcu_barrier_bh();
1675                 spin_lock_bh(&br->multicast_lock);
1676                 WARN_ON(mdb->old);
1677         }
1678
1679         mdb->old = mdb;
1680         call_rcu_bh(&mdb->rcu, br_mdb_free);
1681
1682 out:
1683         spin_unlock_bh(&br->multicast_lock);
1684 }
1685
1686 int br_multicast_set_router(struct net_bridge *br, unsigned long val)
1687 {
1688         int err = -ENOENT;
1689
1690         spin_lock_bh(&br->multicast_lock);
1691         if (!netif_running(br->dev))
1692                 goto unlock;
1693
1694         switch (val) {
1695         case 0:
1696         case 2:
1697                 del_timer(&br->multicast_router_timer);
1698                 /* fall through */
1699         case 1:
1700                 br->multicast_router = val;
1701                 err = 0;
1702                 break;
1703
1704         default:
1705                 err = -EINVAL;
1706                 break;
1707         }
1708
1709 unlock:
1710         spin_unlock_bh(&br->multicast_lock);
1711
1712         return err;
1713 }
1714
1715 int br_multicast_set_port_router(struct net_bridge_port *p, unsigned long val)
1716 {
1717         struct net_bridge *br = p->br;
1718         int err = -ENOENT;
1719
1720         spin_lock(&br->multicast_lock);
1721         if (!netif_running(br->dev) || p->state == BR_STATE_DISABLED)
1722                 goto unlock;
1723
1724         switch (val) {
1725         case 0:
1726         case 1:
1727         case 2:
1728                 p->multicast_router = val;
1729                 err = 0;
1730
1731                 if (val < 2 && !hlist_unhashed(&p->rlist))
1732                         hlist_del_init_rcu(&p->rlist);
1733
1734                 if (val == 1)
1735                         break;
1736
1737                 del_timer(&p->multicast_router_timer);
1738
1739                 if (val == 0)
1740                         break;
1741
1742                 br_multicast_add_router(br, p);
1743                 break;
1744
1745         default:
1746                 err = -EINVAL;
1747                 break;
1748         }
1749
1750 unlock:
1751         spin_unlock(&br->multicast_lock);
1752
1753         return err;
1754 }
1755
1756 int br_multicast_toggle(struct net_bridge *br, unsigned long val)
1757 {
1758         struct net_bridge_port *port;
1759         int err = 0;
1760         struct net_bridge_mdb_htable *mdb;
1761
1762         spin_lock(&br->multicast_lock);
1763         if (br->multicast_disabled == !val)
1764                 goto unlock;
1765
1766         br->multicast_disabled = !val;
1767         if (br->multicast_disabled)
1768                 goto unlock;
1769
1770         if (!netif_running(br->dev))
1771                 goto unlock;
1772
1773         mdb = mlock_dereference(br->mdb, br);
1774         if (mdb) {
1775                 if (mdb->old) {
1776                         err = -EEXIST;
1777 rollback:
1778                         br->multicast_disabled = !!val;
1779                         goto unlock;
1780                 }
1781
1782                 err = br_mdb_rehash(&br->mdb, mdb->max,
1783                                     br->hash_elasticity);
1784                 if (err)
1785                         goto rollback;
1786         }
1787
1788         br_multicast_open(br);
1789         list_for_each_entry(port, &br->port_list, list) {
1790                 if (port->state == BR_STATE_DISABLED ||
1791                     port->state == BR_STATE_BLOCKING)
1792                         continue;
1793
1794                 __br_multicast_enable_port(port);
1795         }
1796
1797 unlock:
1798         spin_unlock(&br->multicast_lock);
1799
1800         return err;
1801 }
1802
1803 int br_multicast_set_hash_max(struct net_bridge *br, unsigned long val)
1804 {
1805         int err = -ENOENT;
1806         u32 old;
1807         struct net_bridge_mdb_htable *mdb;
1808
1809         spin_lock(&br->multicast_lock);
1810         if (!netif_running(br->dev))
1811                 goto unlock;
1812
1813         err = -EINVAL;
1814         if (!is_power_of_2(val))
1815                 goto unlock;
1816
1817         mdb = mlock_dereference(br->mdb, br);
1818         if (mdb && val < mdb->size)
1819                 goto unlock;
1820
1821         err = 0;
1822
1823         old = br->hash_max;
1824         br->hash_max = val;
1825
1826         if (mdb) {
1827                 if (mdb->old) {
1828                         err = -EEXIST;
1829 rollback:
1830                         br->hash_max = old;
1831                         goto unlock;
1832                 }
1833
1834                 err = br_mdb_rehash(&br->mdb, br->hash_max,
1835                                     br->hash_elasticity);
1836                 if (err)
1837                         goto rollback;
1838         }
1839
1840 unlock:
1841         spin_unlock(&br->multicast_lock);
1842
1843         return err;
1844 }