SAFE public projects git trees. - safe/jmp/linux-2.6/blob - 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/netdevice.h>
  19 #include <linux/netfilter_bridge.h>
  20 #include <linux/random.h>
  21 #include <linux/rculist.h>
  22 #include <linux/skbuff.h>
  23 #include <linux/slab.h>
  24 #include <linux/timer.h>
  25 #include <net/ip.h>
  26
  27 #include "br_private.h"
  28
  29 static inline int br_ip_hash(struct net_bridge_mdb_htable *mdb, __be32 ip)
  30 {
  31         return jhash_1word(mdb->secret, (u32)ip) & (mdb->max - 1);
  32 }
  33
  34 static struct net_bridge_mdb_entry *__br_mdb_ip_get(
  35         struct net_bridge_mdb_htable *mdb, __be32 dst, int hash)
  36 {
  37         struct net_bridge_mdb_entry *mp;
  38         struct hlist_node *p;
  39
  40         hlist_for_each_entry(mp, p, &mdb->mhash[hash], hlist[mdb->ver]) {
  41                 if (dst == mp->addr)
  42                         return mp;
  43         }
  44
  45         return NULL;
  46 }
  47
  48 static struct net_bridge_mdb_entry *br_mdb_ip_get(
  49         struct net_bridge_mdb_htable *mdb, __be32 dst)
  50 {
  51         return __br_mdb_ip_get(mdb, dst, br_ip_hash(mdb, dst));
  52 }
  53
  54 struct net_bridge_mdb_entry *br_mdb_get(struct net_bridge *br,
  55                                         struct sk_buff *skb)
  56 {
  57         struct net_bridge_mdb_htable *mdb = br->mdb;
  58
  59         if (!mdb || br->multicast_disabled)
  60                 return NULL;
  61
  62         switch (skb->protocol) {
  63         case htons(ETH_P_IP):
  64                 if (BR_INPUT_SKB_CB(skb)->igmp)
  65                         break;
  66                 return br_mdb_ip_get(mdb, ip_hdr(skb)->daddr);
  67         }
  68
  69         return NULL;
  70 }
  71
  72 static void br_mdb_free(struct rcu_head *head)
  73 {
  74         struct net_bridge_mdb_htable *mdb =
  75                 container_of(head, struct net_bridge_mdb_htable, rcu);
  76         struct net_bridge_mdb_htable *old = mdb->old;
  77
  78         mdb->old = NULL;
  79         kfree(old->mhash);
  80         kfree(old);
  81 }
  82
  83 static int br_mdb_copy(struct net_bridge_mdb_htable *new,
  84                        struct net_bridge_mdb_htable *old,
  85                        int elasticity)
  86 {
  87         struct net_bridge_mdb_entry *mp;
  88         struct hlist_node *p;
  89         int maxlen;
  90         int len;
  91         int i;
  92
  93         for (i = 0; i < old->max; i++)
  94                 hlist_for_each_entry(mp, p, &old->mhash[i], hlist[old->ver])
  95                         hlist_add_head(&mp->hlist[new->ver],
  96                                        &new->mhash[br_ip_hash(new, mp->addr)]);
  97
  98         if (!elasticity)
  99                 return 0;
 100
 101         maxlen = 0;
 102         for (i = 0; i < new->max; i++) {
 103                 len = 0;
 104                 hlist_for_each_entry(mp, p, &new->mhash[i], hlist[new->ver])
 105                         len++;
 106                 if (len > maxlen)
 107                         maxlen = len;
 108         }
 109
 110         return maxlen > elasticity ? -EINVAL : 0;
 111 }
 112
 113 static void br_multicast_free_pg(struct rcu_head *head)
 114 {
 115         struct net_bridge_port_group *p =
 116                 container_of(head, struct net_bridge_port_group, rcu);
 117
 118         kfree(p);
 119 }
 120
 121 static void br_multicast_free_group(struct rcu_head *head)
 122 {
 123         struct net_bridge_mdb_entry *mp =
 124                 container_of(head, struct net_bridge_mdb_entry, rcu);
 125
 126         kfree(mp);
 127 }
 128
 129 static void br_multicast_group_expired(unsigned long data)
 130 {
 131         struct net_bridge_mdb_entry *mp = (void *)data;
 132         struct net_bridge *br = mp->br;
 133         struct net_bridge_mdb_htable *mdb;
 134
 135         spin_lock(&br->multicast_lock);
 136         if (!netif_running(br->dev) || timer_pending(&mp->timer))
 137                 goto out;
 138
 139         if (!hlist_unhashed(&mp->mglist))
 140                 hlist_del_init(&mp->mglist);
 141
 142         if (mp->ports)
 143                 goto out;
 144
 145         mdb = br->mdb;
 146         hlist_del_rcu(&mp->hlist[mdb->ver]);
 147         mdb->size--;
 148
 149         del_timer(&mp->query_timer);
 150         call_rcu_bh(&mp->rcu, br_multicast_free_group);
 151
 152 out:
 153         spin_unlock(&br->multicast_lock);
 154 }
 155
 156 static void br_multicast_del_pg(struct net_bridge *br,
 157                                 struct net_bridge_port_group *pg)
 158 {
 159         struct net_bridge_mdb_htable *mdb = br->mdb;
 160         struct net_bridge_mdb_entry *mp;
 161         struct net_bridge_port_group *p;
 162         struct net_bridge_port_group **pp;
 163
 164         mp = br_mdb_ip_get(mdb, pg->addr);
 165         if (WARN_ON(!mp))
 166                 return;
 167
 168         for (pp = &mp->ports; (p = *pp); pp = &p->next) {
 169                 if (p != pg)
 170                         continue;
 171
 172                 *pp = p->next;
 173                 hlist_del_init(&p->mglist);
 174                 del_timer(&p->timer);
 175                 del_timer(&p->query_timer);
 176                 call_rcu_bh(&p->rcu, br_multicast_free_pg);
 177
 178                 if (!mp->ports && hlist_unhashed(&mp->mglist) &&
 179                     netif_running(br->dev))
 180                         mod_timer(&mp->timer, jiffies);
 181
 182                 return;
 183         }
 184
 185         WARN_ON(1);
 186 }
 187
 188 static void br_multicast_port_group_expired(unsigned long data)
 189 {
 190         struct net_bridge_port_group *pg = (void *)data;
 191         struct net_bridge *br = pg->port->br;
 192
 193         spin_lock(&br->multicast_lock);
 194         if (!netif_running(br->dev) || timer_pending(&pg->timer) ||
 195             hlist_unhashed(&pg->mglist))
 196                 goto out;
 197
 198         br_multicast_del_pg(br, pg);
 199
 200 out:
 201         spin_unlock(&br->multicast_lock);
 202 }
 203
 204 static int br_mdb_rehash(struct net_bridge_mdb_htable **mdbp, int max,
 205                          int elasticity)
 206 {
 207         struct net_bridge_mdb_htable *old = *mdbp;
 208         struct net_bridge_mdb_htable *mdb;
 209         int err;
 210
 211         mdb = kmalloc(sizeof(*mdb), GFP_ATOMIC);
 212         if (!mdb)
 213                 return -ENOMEM;
 214
 215         mdb->max = max;
 216         mdb->old = old;
 217
 218         mdb->mhash = kzalloc(max * sizeof(*mdb->mhash), GFP_ATOMIC);
 219         if (!mdb->mhash) {
 220                 kfree(mdb);
 221                 return -ENOMEM;
 222         }
 223
 224         mdb->size = old ? old->size : 0;
 225         mdb->ver = old ? old->ver ^ 1 : 0;
 226
 227         if (!old || elasticity)
 228                 get_random_bytes(&mdb->secret, sizeof(mdb->secret));
 229         else
 230                 mdb->secret = old->secret;
 231
 232         if (!old)
 233                 goto out;
 234
 235         err = br_mdb_copy(mdb, old, elasticity);
 236         if (err) {
 237                 kfree(mdb->mhash);
 238                 kfree(mdb);
 239                 return err;
 240         }
 241
 242         call_rcu_bh(&mdb->rcu, br_mdb_free);
 243
 244 out:
 245         rcu_assign_pointer(*mdbp, mdb);
 246
 247         return 0;
 248 }
 249
 250 static struct sk_buff *br_multicast_alloc_query(struct net_bridge *br,
 251                                                 __be32 group)
 252 {
 253         struct sk_buff *skb;
 254         struct igmphdr *ih;
 255         struct ethhdr *eth;
 256         struct iphdr *iph;
 257
 258         skb = netdev_alloc_skb_ip_align(br->dev, sizeof(*eth) + sizeof(*iph) +
 259                                                  sizeof(*ih) + 4);
 260         if (!skb)
 261                 goto out;
 262
 263         skb->protocol = htons(ETH_P_IP);
 264
 265         skb_reset_mac_header(skb);
 266         eth = eth_hdr(skb);
 267
 268         memcpy(eth->h_source, br->dev->dev_addr, 6);
 269         eth->h_dest[0] = 1;
 270         eth->h_dest[1] = 0;
 271         eth->h_dest[2] = 0x5e;
 272         eth->h_dest[3] = 0;
 273         eth->h_dest[4] = 0;
 274         eth->h_dest[5] = 1;
 275         eth->h_proto = htons(ETH_P_IP);
 276         skb_put(skb, sizeof(*eth));
 277
 278         skb_set_network_header(skb, skb->len);
 279         iph = ip_hdr(skb);
 280
 281         iph->version = 4;
 282         iph->ihl = 6;
 283         iph->tos = 0xc0;
 284         iph->tot_len = htons(sizeof(*iph) + sizeof(*ih) + 4);
 285         iph->id = 0;
 286         iph->frag_off = htons(IP_DF);
 287         iph->ttl = 1;
 288         iph->protocol = IPPROTO_IGMP;
 289         iph->saddr = 0;
 290         iph->daddr = htonl(INADDR_ALLHOSTS_GROUP);
 291         ((u8 *)&iph[1])[0] = IPOPT_RA;
 292         ((u8 *)&iph[1])[1] = 4;
 293         ((u8 *)&iph[1])[2] = 0;
 294         ((u8 *)&iph[1])[3] = 0;
 295         ip_send_check(iph);
 296         skb_put(skb, 24);
 297
 298         skb_set_transport_header(skb, skb->len);
 299         ih = igmp_hdr(skb);
 300         ih->type = IGMP_HOST_MEMBERSHIP_QUERY;
 301         ih->code = (group ? br->multicast_last_member_interval :
 302                             br->multicast_query_response_interval) /
 303                    (HZ / IGMP_TIMER_SCALE);
 304         ih->group = group;
 305         ih->csum = 0;
 306         ih->csum = ip_compute_csum((void *)ih, sizeof(struct igmphdr));
 307         skb_put(skb, sizeof(*ih));
 308
 309         __skb_pull(skb, sizeof(*eth));
 310
 311 out:
 312         return skb;
 313 }
 314
 315 static void br_multicast_send_group_query(struct net_bridge_mdb_entry *mp)
 316 {
 317         struct net_bridge *br = mp->br;
 318         struct sk_buff *skb;
 319
 320         skb = br_multicast_alloc_query(br, mp->addr);
 321         if (!skb)
 322                 goto timer;
 323
 324         netif_rx(skb);
 325
 326 timer:
 327         if (++mp->queries_sent < br->multicast_last_member_count)
 328                 mod_timer(&mp->query_timer,
 329                           jiffies + br->multicast_last_member_interval);
 330 }
 331
 332 static void br_multicast_group_query_expired(unsigned long data)
 333 {
 334         struct net_bridge_mdb_entry *mp = (void *)data;
 335         struct net_bridge *br = mp->br;
 336
 337         spin_lock(&br->multicast_lock);
 338         if (!netif_running(br->dev) || hlist_unhashed(&mp->mglist) ||
 339             mp->queries_sent >= br->multicast_last_member_count)
 340                 goto out;
 341
 342         br_multicast_send_group_query(mp);
 343
 344 out:
 345         spin_unlock(&br->multicast_lock);
 346 }
 347
 348 static void br_multicast_send_port_group_query(struct net_bridge_port_group *pg)
 349 {
 350         struct net_bridge_port *port = pg->port;
 351         struct net_bridge *br = port->br;
 352         struct sk_buff *skb;
 353
 354         skb = br_multicast_alloc_query(br, pg->addr);
 355         if (!skb)
 356                 goto timer;
 357
 358         br_deliver(port, skb);
 359
 360 timer:
 361         if (++pg->queries_sent < br->multicast_last_member_count)
 362                 mod_timer(&pg->query_timer,
 363                           jiffies + br->multicast_last_member_interval);
 364 }
 365
 366 static void br_multicast_port_group_query_expired(unsigned long data)
 367 {
 368         struct net_bridge_port_group *pg = (void *)data;
 369         struct net_bridge_port *port = pg->port;
 370         struct net_bridge *br = port->br;
 371
 372         spin_lock(&br->multicast_lock);
 373         if (!netif_running(br->dev) || hlist_unhashed(&pg->mglist) ||
 374             pg->queries_sent >= br->multicast_last_member_count)
 375                 goto out;
 376
 377         br_multicast_send_port_group_query(pg);
 378
 379 out:
 380         spin_unlock(&br->multicast_lock);
 381 }
 382
 383 static struct net_bridge_mdb_entry *br_multicast_get_group(
 384         struct net_bridge *br, struct net_bridge_port *port, __be32 group,
 385         int hash)
 386 {
 387         struct net_bridge_mdb_htable *mdb = br->mdb;
 388         struct net_bridge_mdb_entry *mp;
 389         struct hlist_node *p;
 390         unsigned count = 0;
 391         unsigned max;
 392         int elasticity;
 393         int err;
 394
 395         hlist_for_each_entry(mp, p, &mdb->mhash[hash], hlist[mdb->ver]) {
 396                 count++;
 397                 if (unlikely(group == mp->addr)) {
 398                         return mp;
 399                 }
 400         }
 401
 402         elasticity = 0;
 403         max = mdb->max;
 404
 405         if (unlikely(count > br->hash_elasticity && count)) {
 406                 if (net_ratelimit())
 407                         printk(KERN_INFO "%s: Multicast hash table "
 408                                "chain limit reached: %s\n",
 409                                br->dev->name, port ? port->dev->name :
 410                                                      br->dev->name);
 411
 412                 elasticity = br->hash_elasticity;
 413         }
 414
 415         if (mdb->size >= max) {
 416                 max *= 2;
 417                 if (unlikely(max >= br->hash_max)) {
 418                         printk(KERN_WARNING "%s: Multicast hash table maximum "
 419                                "reached, disabling snooping: %s, %d\n",
 420                                br->dev->name, port ? port->dev->name :
 421                                                      br->dev->name,
 422                                max);
 423                         err = -E2BIG;
 424 disable:
 425                         br->multicast_disabled = 1;
 426                         goto err;
 427                 }
 428         }
 429
 430         if (max > mdb->max || elasticity) {
 431                 if (mdb->old) {
 432                         if (net_ratelimit())
 433                                 printk(KERN_INFO "%s: Multicast hash table "
 434                                        "on fire: %s\n",
 435                                        br->dev->name, port ? port->dev->name :
 436                                                              br->dev->name);
 437                         err = -EEXIST;
 438                         goto err;
 439                 }
 440
 441                 err = br_mdb_rehash(&br->mdb, max, elasticity);
 442                 if (err) {
 443                         printk(KERN_WARNING "%s: Cannot rehash multicast "
 444                                "hash table, disabling snooping: "
 445                                "%s, %d, %d\n",
 446                                br->dev->name, port ? port->dev->name :
 447                                                      br->dev->name,
 448                                mdb->size, err);
 449                         goto disable;
 450                 }
 451
 452                 err = -EAGAIN;
 453                 goto err;
 454         }
 455
 456         return NULL;
 457
 458 err:
 459         mp = ERR_PTR(err);
 460         return mp;
 461 }
 462
 463 static struct net_bridge_mdb_entry *br_multicast_new_group(
 464         struct net_bridge *br, struct net_bridge_port *port, __be32 group)
 465 {
 466         struct net_bridge_mdb_htable *mdb = br->mdb;
 467         struct net_bridge_mdb_entry *mp;
 468         int hash;
 469
 470         if (!mdb) {
 471                 if (br_mdb_rehash(&br->mdb, BR_HASH_SIZE, 0))
 472                         return NULL;
 473                 goto rehash;
 474         }
 475
 476         hash = br_ip_hash(mdb, group);
 477         mp = br_multicast_get_group(br, port, group, hash);
 478         switch (PTR_ERR(mp)) {
 479         case 0:
 480                 break;
 481
 482         case -EAGAIN:
 483 rehash:
 484                 mdb = br->mdb;
 485                 hash = br_ip_hash(mdb, group);
 486                 break;
 487
 488         default:
 489                 goto out;
 490         }
 491
 492         mp = kzalloc(sizeof(*mp), GFP_ATOMIC);
 493         if (unlikely(!mp))
 494                 goto out;
 495
 496         mp->br = br;
 497         mp->addr = group;
 498         setup_timer(&mp->timer, br_multicast_group_expired,
 499                     (unsigned long)mp);
 500         setup_timer(&mp->query_timer, br_multicast_group_query_expired,
 501                     (unsigned long)mp);
 502
 503         hlist_add_head_rcu(&mp->hlist[mdb->ver], &mdb->mhash[hash]);
 504         mdb->size++;
 505
 506 out:
 507         return mp;
 508 }
 509
 510 static int br_multicast_add_group(struct net_bridge *br,
 511                                   struct net_bridge_port *port, __be32 group)
 512 {
 513         struct net_bridge_mdb_entry *mp;
 514         struct net_bridge_port_group *p;
 515         struct net_bridge_port_group **pp;
 516         unsigned long now = jiffies;
 517         int err;
 518
 519         if (ipv4_is_local_multicast(group))
 520                 return 0;
 521
 522         spin_lock(&br->multicast_lock);
 523         if (!netif_running(br->dev) ||
 524             (port && port->state == BR_STATE_DISABLED))
 525                 goto out;
 526
 527         mp = br_multicast_new_group(br, port, group);
 528         err = PTR_ERR(mp);
 529         if (unlikely(IS_ERR(mp) || !mp))
 530                 goto err;
 531
 532         if (!port) {
 533                 hlist_add_head(&mp->mglist, &br->mglist);
 534                 mod_timer(&mp->timer, now + br->multicast_membership_interval);
 535                 goto out;
 536         }
 537
 538         for (pp = &mp->ports; (p = *pp); pp = &p->next) {
 539                 if (p->port == port)
 540                         goto found;
 541                 if ((unsigned long)p->port < (unsigned long)port)
 542                         break;
 543         }
 544
 545         p = kzalloc(sizeof(*p), GFP_ATOMIC);
 546         err = -ENOMEM;
 547         if (unlikely(!p))
 548                 goto err;
 549
 550         p->addr = group;
 551         p->port = port;
 552         p->next = *pp;
 553         hlist_add_head(&p->mglist, &port->mglist);
 554         setup_timer(&p->timer, br_multicast_port_group_expired,
 555                     (unsigned long)p);
 556         setup_timer(&p->query_timer, br_multicast_port_group_query_expired,
 557                     (unsigned long)p);
 558
 559         rcu_assign_pointer(*pp, p);
 560
 561 found:
 562         mod_timer(&p->timer, now + br->multicast_membership_interval);
 563 out:
 564         err = 0;
 565
 566 err:
 567         spin_unlock(&br->multicast_lock);
 568         return err;
 569 }
 570
 571 static void br_multicast_router_expired(unsigned long data)
 572 {
 573         struct net_bridge_port *port = (void *)data;
 574         struct net_bridge *br = port->br;
 575
 576         spin_lock(&br->multicast_lock);
 577         if (port->multicast_router != 1 ||
 578             timer_pending(&port->multicast_router_timer) ||
 579             hlist_unhashed(&port->rlist))
 580                 goto out;
 581
 582         hlist_del_init_rcu(&port->rlist);
 583
 584 out:
 585         spin_unlock(&br->multicast_lock);
 586 }
 587
 588 static void br_multicast_local_router_expired(unsigned long data)
 589 {
 590 }
 591
 592 static void br_multicast_send_query(struct net_bridge *br,
 593                                     struct net_bridge_port *port, u32 sent)
 594 {
 595         unsigned long time;
 596         struct sk_buff *skb;
 597
 598         if (!netif_running(br->dev) || br->multicast_disabled ||
 599             timer_pending(&br->multicast_querier_timer))
 600                 return;
 601
 602         skb = br_multicast_alloc_query(br, 0);
 603         if (!skb)
 604                 goto timer;
 605
 606         if (port) {
 607                 __skb_push(skb, sizeof(struct ethhdr));
 608                 skb->dev = port->dev;
 609                 NF_HOOK(PF_BRIDGE, NF_BR_LOCAL_OUT, skb, NULL, skb->dev,
 610                         dev_queue_xmit);
 611         } else
 612                 netif_rx(skb);
 613
 614 timer:
 615         time = jiffies;
 616         time += sent < br->multicast_startup_query_count ?
 617                 br->multicast_startup_query_interval :
 618                 br->multicast_query_interval;
 619         mod_timer(port ? &port->multicast_query_timer :
 620                          &br->multicast_query_timer, time);
 621 }
 622
 623 static void br_multicast_port_query_expired(unsigned long data)
 624 {
 625         struct net_bridge_port *port = (void *)data;
 626         struct net_bridge *br = port->br;
 627
 628         spin_lock(&br->multicast_lock);
 629         if (port && (port->state == BR_STATE_DISABLED ||
 630                      port->state == BR_STATE_BLOCKING))
 631                 goto out;
 632
 633         if (port->multicast_startup_queries_sent <
 634             br->multicast_startup_query_count)
 635                 port->multicast_startup_queries_sent++;
 636
 637         br_multicast_send_query(port->br, port,
 638                                 port->multicast_startup_queries_sent);
 639
 640 out:
 641         spin_unlock(&br->multicast_lock);
 642 }
 643
 644 void br_multicast_add_port(struct net_bridge_port *port)
 645 {
 646         port->multicast_router = 1;
 647
 648         setup_timer(&port->multicast_router_timer, br_multicast_router_expired,
 649                     (unsigned long)port);
 650         setup_timer(&port->multicast_query_timer,
 651                     br_multicast_port_query_expired, (unsigned long)port);
 652 }
 653
 654 void br_multicast_del_port(struct net_bridge_port *port)
 655 {
 656         del_timer_sync(&port->multicast_router_timer);
 657 }
 658
 659 static void __br_multicast_enable_port(struct net_bridge_port *port)
 660 {
 661         port->multicast_startup_queries_sent = 0;
 662
 663         if (try_to_del_timer_sync(&port->multicast_query_timer) >= 0 ||
 664             del_timer(&port->multicast_query_timer))
 665                 mod_timer(&port->multicast_query_timer, jiffies);
 666 }
 667
 668 void br_multicast_enable_port(struct net_bridge_port *port)
 669 {
 670         struct net_bridge *br = port->br;
 671
 672         spin_lock(&br->multicast_lock);
 673         if (br->multicast_disabled || !netif_running(br->dev))
 674                 goto out;
 675
 676         __br_multicast_enable_port(port);
 677
 678 out:
 679         spin_unlock(&br->multicast_lock);
 680 }
 681
 682 void br_multicast_disable_port(struct net_bridge_port *port)
 683 {
 684         struct net_bridge *br = port->br;
 685         struct net_bridge_port_group *pg;
 686         struct hlist_node *p, *n;
 687
 688         spin_lock(&br->multicast_lock);
 689         hlist_for_each_entry_safe(pg, p, n, &port->mglist, mglist)
 690                 br_multicast_del_pg(br, pg);
 691
 692         if (!hlist_unhashed(&port->rlist))
 693                 hlist_del_init_rcu(&port->rlist);
 694         del_timer(&port->multicast_router_timer);
 695         del_timer(&port->multicast_query_timer);
 696         spin_unlock(&br->multicast_lock);
 697 }
 698
 699 static int br_multicast_igmp3_report(struct net_bridge *br,
 700                                      struct net_bridge_port *port,
 701                                      struct sk_buff *skb)
 702 {
 703         struct igmpv3_report *ih;
 704         struct igmpv3_grec *grec;
 705         int i;
 706         int len;
 707         int num;
 708         int type;
 709         int err = 0;
 710         __be32 group;
 711
 712         if (!pskb_may_pull(skb, sizeof(*ih)))
 713                 return -EINVAL;
 714
 715         ih = igmpv3_report_hdr(skb);
 716         num = ntohs(ih->ngrec);
 717         len = sizeof(*ih);
 718
 719         for (i = 0; i < num; i++) {
 720                 len += sizeof(*grec);
 721                 if (!pskb_may_pull(skb, len))
 722                         return -EINVAL;
 723
 724                 grec = (void *)(skb->data + len);
 725                 group = grec->grec_mca;
 726                 type = grec->grec_type;
 727
 728                 len += grec->grec_nsrcs * 4;
 729                 if (!pskb_may_pull(skb, len))
 730                         return -EINVAL;
 731
 732                 /* We treat this as an IGMPv2 report for now. */
 733                 switch (type) {
 734                 case IGMPV3_MODE_IS_INCLUDE:
 735                 case IGMPV3_MODE_IS_EXCLUDE:
 736                 case IGMPV3_CHANGE_TO_INCLUDE:
 737                 case IGMPV3_CHANGE_TO_EXCLUDE:
 738                 case IGMPV3_ALLOW_NEW_SOURCES:
 739                 case IGMPV3_BLOCK_OLD_SOURCES:
 740                         break;
 741
 742                 default:
 743                         continue;
 744                 }
 745
 746                 err = br_multicast_add_group(br, port, group);
 747                 if (err)
 748                         break;
 749         }
 750
 751         return err;
 752 }
 753
 754 static void br_multicast_add_router(struct net_bridge *br,
 755                                     struct net_bridge_port *port)
 756 {
 757         struct hlist_node *p;
 758         struct hlist_node **h;
 759
 760         for (h = &br->router_list.first;
 761              (p = *h) &&
 762              (unsigned long)container_of(p, struct net_bridge_port, rlist) >
 763              (unsigned long)port;
 764              h = &p->next)
 765                 ;
 766
 767         port->rlist.pprev = h;
 768         port->rlist.next = p;
 769         rcu_assign_pointer(*h, &port->rlist);
 770         if (p)
 771                 p->pprev = &port->rlist.next;
 772 }
 773
 774 static void br_multicast_mark_router(struct net_bridge *br,
 775                                      struct net_bridge_port *port)
 776 {
 777         unsigned long now = jiffies;
 778
 779         if (!port) {
 780                 if (br->multicast_router == 1)
 781                         mod_timer(&br->multicast_router_timer,
 782                                   now + br->multicast_querier_interval);
 783                 return;
 784         }
 785
 786         if (port->multicast_router != 1)
 787                 return;
 788
 789         if (!hlist_unhashed(&port->rlist))
 790                 goto timer;
 791
 792         br_multicast_add_router(br, port);
 793
 794 timer:
 795         mod_timer(&port->multicast_router_timer,
 796                   now + br->multicast_querier_interval);
 797 }
 798
 799 static void br_multicast_query_received(struct net_bridge *br,
 800                                         struct net_bridge_port *port,
 801                                         __be32 saddr)
 802 {
 803         if (saddr)
 804                 mod_timer(&br->multicast_querier_timer,
 805                           jiffies + br->multicast_querier_interval);
 806         else if (timer_pending(&br->multicast_querier_timer))
 807                 return;
 808
 809         br_multicast_mark_router(br, port);
 810 }
 811
 812 static int br_multicast_query(struct net_bridge *br,
 813                               struct net_bridge_port *port,
 814                               struct sk_buff *skb)
 815 {
 816         struct iphdr *iph = ip_hdr(skb);
 817         struct igmphdr *ih = igmp_hdr(skb);
 818         struct net_bridge_mdb_entry *mp;
 819         struct igmpv3_query *ih3;
 820         struct net_bridge_port_group *p;
 821         struct net_bridge_port_group **pp;
 822         unsigned long max_delay;
 823         unsigned long now = jiffies;
 824         __be32 group;
 825
 826         spin_lock(&br->multicast_lock);
 827         if (!netif_running(br->dev) ||
 828             (port && port->state == BR_STATE_DISABLED))
 829                 goto out;
 830
 831         br_multicast_query_received(br, port, iph->saddr);
 832
 833         group = ih->group;
 834
 835         if (skb->len == sizeof(*ih)) {
 836                 max_delay = ih->code * (HZ / IGMP_TIMER_SCALE);
 837
 838                 if (!max_delay) {
 839                         max_delay = 10 * HZ;
 840                         group = 0;
 841                 }
 842         } else {
 843                 if (!pskb_may_pull(skb, sizeof(struct igmpv3_query)))
 844                         return -EINVAL;
 845
 846                 ih3 = igmpv3_query_hdr(skb);
 847                 if (ih3->nsrcs)
 848                         return 0;
 849
 850                 max_delay = ih3->code ? 1 :
 851                             IGMPV3_MRC(ih3->code) * (HZ / IGMP_TIMER_SCALE);
 852         }
 853
 854         if (!group)
 855                 goto out;
 856
 857         mp = br_mdb_ip_get(br->mdb, group);
 858         if (!mp)
 859                 goto out;
 860
 861         max_delay *= br->multicast_last_member_count;
 862
 863         if (!hlist_unhashed(&mp->mglist) &&
 864             (timer_pending(&mp->timer) ?
 865              time_after(mp->timer.expires, now + max_delay) :
 866              try_to_del_timer_sync(&mp->timer) >= 0))
 867                 mod_timer(&mp->timer, now + max_delay);
 868
 869         for (pp = &mp->ports; (p = *pp); pp = &p->next) {
 870                 if (timer_pending(&p->timer) ?
 871                     time_after(p->timer.expires, now + max_delay) :
 872                     try_to_del_timer_sync(&p->timer) >= 0)
 873                         mod_timer(&mp->timer, now + max_delay);
 874         }
 875
 876 out:
 877         spin_unlock(&br->multicast_lock);
 878         return 0;
 879 }
 880
 881 static void br_multicast_leave_group(struct net_bridge *br,
 882                                      struct net_bridge_port *port,
 883                                      __be32 group)
 884 {
 885         struct net_bridge_mdb_htable *mdb;
 886         struct net_bridge_mdb_entry *mp;
 887         struct net_bridge_port_group *p;
 888         unsigned long now;
 889         unsigned long time;
 890
 891         if (ipv4_is_local_multicast(group))
 892                 return;
 893
 894         spin_lock(&br->multicast_lock);
 895         if (!netif_running(br->dev) ||
 896             (port && port->state == BR_STATE_DISABLED) ||
 897             timer_pending(&br->multicast_querier_timer))
 898                 goto out;
 899
 900         mdb = br->mdb;
 901         mp = br_mdb_ip_get(mdb, group);
 902         if (!mp)
 903                 goto out;
 904
 905         now = jiffies;
 906         time = now + br->multicast_last_member_count *
 907                      br->multicast_last_member_interval;
 908
 909         if (!port) {
 910                 if (!hlist_unhashed(&mp->mglist) &&
 911                     (timer_pending(&mp->timer) ?
 912                      time_after(mp->timer.expires, time) :
 913                      try_to_del_timer_sync(&mp->timer) >= 0)) {
 914                         mod_timer(&mp->timer, time);
 915
 916                         mp->queries_sent = 0;
 917                         mod_timer(&mp->query_timer, now);
 918                 }
 919
 920                 goto out;
 921         }
 922
 923         for (p = mp->ports; p; p = p->next) {
 924                 if (p->port != port)
 925                         continue;
 926
 927                 if (!hlist_unhashed(&p->mglist) &&
 928                     (timer_pending(&p->timer) ?
 929                      time_after(p->timer.expires, time) :
 930                      try_to_del_timer_sync(&p->timer) >= 0)) {
 931                         mod_timer(&p->timer, time);
 932
 933                         p->queries_sent = 0;
 934                         mod_timer(&p->query_timer, now);
 935                 }
 936
 937                 break;
 938         }
 939
 940 out:
 941         spin_unlock(&br->multicast_lock);
 942 }
 943
 944 static int br_multicast_ipv4_rcv(struct net_bridge *br,
 945                                  struct net_bridge_port *port,
 946                                  struct sk_buff *skb)
 947 {
 948         struct sk_buff *skb2 = skb;
 949         struct iphdr *iph;
 950         struct igmphdr *ih;
 951         unsigned len;
 952         unsigned offset;
 953         int err;
 954
 955         BR_INPUT_SKB_CB(skb)->igmp = 0;
 956         BR_INPUT_SKB_CB(skb)->mrouters_only = 0;
 957
 958         /* We treat OOM as packet loss for now. */
 959         if (!pskb_may_pull(skb, sizeof(*iph)))
 960                 return -EINVAL;
 961
 962         iph = ip_hdr(skb);
 963
 964         if (iph->ihl < 5 || iph->version != 4)
 965                 return -EINVAL;
 966
 967         if (!pskb_may_pull(skb, ip_hdrlen(skb)))
 968                 return -EINVAL;
 969
 970         iph = ip_hdr(skb);
 971
 972         if (unlikely(ip_fast_csum((u8 *)iph, iph->ihl)))
 973                 return -EINVAL;
 974
 975         if (iph->protocol != IPPROTO_IGMP)
 976                 return 0;
 977
 978         len = ntohs(iph->tot_len);
 979         if (skb->len < len || len < ip_hdrlen(skb))
 980                 return -EINVAL;
 981
 982         if (skb->len > len) {
 983                 skb2 = skb_clone(skb, GFP_ATOMIC);
 984                 if (!skb2)
 985                         return -ENOMEM;
 986
 987                 err = pskb_trim_rcsum(skb2, len);
 988                 if (err)
 989                         return err;
 990         }
 991
 992         len -= ip_hdrlen(skb2);
 993         offset = skb_network_offset(skb2) + ip_hdrlen(skb2);
 994         __skb_pull(skb2, offset);
 995         skb_reset_transport_header(skb2);
 996
 997         err = -EINVAL;
 998         if (!pskb_may_pull(skb2, sizeof(*ih)))
 999                 goto out;
1000
1001         iph = ip_hdr(skb2);
1002
1003         switch (skb2->ip_summed) {
1004         case CHECKSUM_COMPLETE:
1005                 if (!csum_fold(skb2->csum))
1006                         break;
1007                 /* fall through */
1008         case CHECKSUM_NONE:
1009                 skb2->csum = 0;
1010                 if (skb_checksum_complete(skb2))
1011                         return -EINVAL;
1012         }
1013
1014         err = 0;
1015
1016         BR_INPUT_SKB_CB(skb)->igmp = 1;
1017         ih = igmp_hdr(skb2);
1018
1019         switch (ih->type) {
1020         case IGMP_HOST_MEMBERSHIP_REPORT:
1021         case IGMPV2_HOST_MEMBERSHIP_REPORT:
1022                 BR_INPUT_SKB_CB(skb2)->mrouters_only = 1;
1023                 err = br_multicast_add_group(br, port, ih->group);
1024                 break;
1025         case IGMPV3_HOST_MEMBERSHIP_REPORT:
1026                 err = br_multicast_igmp3_report(br, port, skb2);
1027                 break;
1028         case IGMP_HOST_MEMBERSHIP_QUERY:
1029                 err = br_multicast_query(br, port, skb2);
1030                 break;
1031         case IGMP_HOST_LEAVE_MESSAGE:
1032                 br_multicast_leave_group(br, port, ih->group);
1033                 break;
1034         }
1035
1036 out:
1037         __skb_push(skb2, offset);
1038         if (skb2 != skb)
1039                 kfree_skb(skb2);
1040         return err;
1041 }
1042
1043 int br_multicast_rcv(struct net_bridge *br, struct net_bridge_port *port,
1044                      struct sk_buff *skb)
1045 {
1046         if (br->multicast_disabled)
1047                 return 0;
1048
1049         switch (skb->protocol) {
1050         case htons(ETH_P_IP):
1051                 return br_multicast_ipv4_rcv(br, port, skb);
1052         }
1053
1054         return 0;
1055 }
1056
1057 static void br_multicast_query_expired(unsigned long data)
1058 {
1059         struct net_bridge *br = (void *)data;
1060
1061         spin_lock(&br->multicast_lock);
1062         if (br->multicast_startup_queries_sent <
1063             br->multicast_startup_query_count)
1064                 br->multicast_startup_queries_sent++;
1065
1066         br_multicast_send_query(br, NULL, br->multicast_startup_queries_sent);
1067
1068         spin_unlock(&br->multicast_lock);
1069 }
1070
1071 void br_multicast_init(struct net_bridge *br)
1072 {
1073         br->hash_elasticity = 4;
1074         br->hash_max = 512;
1075
1076         br->multicast_router = 1;
1077         br->multicast_last_member_count = 2;
1078         br->multicast_startup_query_count = 2;
1079
1080         br->multicast_last_member_interval = HZ;
1081         br->multicast_query_response_interval = 10 * HZ;
1082         br->multicast_startup_query_interval = 125 * HZ / 4;
1083         br->multicast_query_interval = 125 * HZ;
1084         br->multicast_querier_interval = 255 * HZ;
1085         br->multicast_membership_interval = 260 * HZ;
1086
1087         spin_lock_init(&br->multicast_lock);
1088         setup_timer(&br->multicast_router_timer,
1089                     br_multicast_local_router_expired, 0);
1090         setup_timer(&br->multicast_querier_timer,
1091                     br_multicast_local_router_expired, 0);
1092         setup_timer(&br->multicast_query_timer, br_multicast_query_expired,
1093                     (unsigned long)br);
1094 }
1095
1096 void br_multicast_open(struct net_bridge *br)
1097 {
1098         br->multicast_startup_queries_sent = 0;
1099
1100         if (br->multicast_disabled)
1101                 return;
1102
1103         mod_timer(&br->multicast_query_timer, jiffies);
1104 }
1105
1106 void br_multicast_stop(struct net_bridge *br)
1107 {
1108         struct net_bridge_mdb_htable *mdb;
1109         struct net_bridge_mdb_entry *mp;
1110         struct hlist_node *p, *n;
1111         u32 ver;
1112         int i;
1113
1114         del_timer_sync(&br->multicast_router_timer);
1115         del_timer_sync(&br->multicast_querier_timer);
1116         del_timer_sync(&br->multicast_query_timer);
1117
1118         spin_lock_bh(&br->multicast_lock);
1119         mdb = br->mdb;
1120         if (!mdb)
1121                 goto out;
1122
1123         br->mdb = NULL;
1124
1125         ver = mdb->ver;
1126         for (i = 0; i < mdb->max; i++) {
1127                 hlist_for_each_entry_safe(mp, p, n, &mdb->mhash[i],
1128                                           hlist[ver]) {
1129                         del_timer(&mp->timer);
1130                         del_timer(&mp->query_timer);
1131                         call_rcu_bh(&mp->rcu, br_multicast_free_group);
1132                 }
1133         }
1134
1135         if (mdb->old) {
1136                 spin_unlock_bh(&br->multicast_lock);
1137                 synchronize_rcu_bh();
1138                 spin_lock_bh(&br->multicast_lock);
1139                 WARN_ON(mdb->old);
1140         }
1141
1142         mdb->old = mdb;
1143         call_rcu_bh(&mdb->rcu, br_mdb_free);
1144
1145 out:
1146         spin_unlock_bh(&br->multicast_lock);
1147 }
1148
1149 int br_multicast_set_router(struct net_bridge *br, unsigned long val)
1150 {
1151         int err = -ENOENT;
1152
1153         spin_lock_bh(&br->multicast_lock);
1154         if (!netif_running(br->dev))
1155                 goto unlock;
1156
1157         switch (val) {
1158         case 0:
1159         case 2:
1160                 del_timer(&br->multicast_router_timer);
1161                 /* fall through */
1162         case 1:
1163                 br->multicast_router = val;
1164                 err = 0;
1165                 break;
1166
1167         default:
1168                 err = -EINVAL;
1169                 break;
1170         }
1171
1172 unlock:
1173         spin_unlock_bh(&br->multicast_lock);
1174
1175         return err;
1176 }
1177
1178 int br_multicast_set_port_router(struct net_bridge_port *p, unsigned long val)
1179 {
1180         struct net_bridge *br = p->br;
1181         int err = -ENOENT;
1182
1183         spin_lock(&br->multicast_lock);
1184         if (!netif_running(br->dev) || p->state == BR_STATE_DISABLED)
1185                 goto unlock;
1186
1187         switch (val) {
1188         case 0:
1189         case 1:
1190         case 2:
1191                 p->multicast_router = val;
1192                 err = 0;
1193
1194                 if (val < 2 && !hlist_unhashed(&p->rlist))
1195                         hlist_del_init_rcu(&p->rlist);
1196
1197                 if (val == 1)
1198                         break;
1199
1200                 del_timer(&p->multicast_router_timer);
1201
1202                 if (val == 0)
1203                         break;
1204
1205                 br_multicast_add_router(br, p);
1206                 break;
1207
1208         default:
1209                 err = -EINVAL;
1210                 break;
1211         }
1212
1213 unlock:
1214         spin_unlock(&br->multicast_lock);
1215
1216         return err;
1217 }
1218
1219 int br_multicast_toggle(struct net_bridge *br, unsigned long val)
1220 {
1221         struct net_bridge_port *port;
1222         int err = -ENOENT;
1223
1224         spin_lock(&br->multicast_lock);
1225         if (!netif_running(br->dev))
1226                 goto unlock;
1227
1228         err = 0;
1229         if (br->multicast_disabled == !val)
1230                 goto unlock;
1231
1232         br->multicast_disabled = !val;
1233         if (br->multicast_disabled)
1234                 goto unlock;
1235
1236         if (br->mdb) {
1237                 if (br->mdb->old) {
1238                         err = -EEXIST;
1239 rollback:
1240                         br->multicast_disabled = !!val;
1241                         goto unlock;
1242                 }
1243
1244                 err = br_mdb_rehash(&br->mdb, br->mdb->max,
1245                                     br->hash_elasticity);
1246                 if (err)
1247                         goto rollback;
1248         }
1249
1250         br_multicast_open(br);
1251         list_for_each_entry(port, &br->port_list, list) {
1252                 if (port->state == BR_STATE_DISABLED ||
1253                     port->state == BR_STATE_BLOCKING)
1254                         continue;
1255
1256                 __br_multicast_enable_port(port);
1257         }
1258
1259 unlock:
1260         spin_unlock(&br->multicast_lock);
1261
1262         return err;
1263 }