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
  28 #include "br_private.h"
  29
  30 static inline int br_ip_hash(struct net_bridge_mdb_htable *mdb, __be32 ip)
  31 {
  32         return jhash_1word(mdb->secret, (u32)ip) & (mdb->max - 1);
  33 }
  34
  35 static struct net_bridge_mdb_entry *__br_mdb_ip_get(
  36         struct net_bridge_mdb_htable *mdb, __be32 dst, int hash)
  37 {
  38         struct net_bridge_mdb_entry *mp;
  39         struct hlist_node *p;
  40
  41         hlist_for_each_entry_rcu(mp, p, &mdb->mhash[hash], hlist[mdb->ver]) {
  42                 if (dst == mp->addr)
  43                         return mp;
  44         }
  45
  46         return NULL;
  47 }
  48
  49 static struct net_bridge_mdb_entry *br_mdb_ip_get(
  50         struct net_bridge_mdb_htable *mdb, __be32 dst)
  51 {
  52         if (!mdb)
  53                 return NULL;
  54
  55         return __br_mdb_ip_get(mdb, dst, br_ip_hash(mdb, dst));
  56 }
  57
  58 struct net_bridge_mdb_entry *br_mdb_get(struct net_bridge *br,
  59                                         struct sk_buff *skb)
  60 {
  61         if (br->multicast_disabled)
  62                 return NULL;
  63
  64         switch (skb->protocol) {
  65         case htons(ETH_P_IP):
  66                 if (BR_INPUT_SKB_CB(skb)->igmp)
  67                         break;
  68                 return br_mdb_ip_get(br->mdb, ip_hdr(skb)->daddr);
  69         }
  70
  71         return NULL;
  72 }
  73
  74 static void br_mdb_free(struct rcu_head *head)
  75 {
  76         struct net_bridge_mdb_htable *mdb =
  77                 container_of(head, struct net_bridge_mdb_htable, rcu);
  78         struct net_bridge_mdb_htable *old = mdb->old;
  79
  80         mdb->old = NULL;
  81         kfree(old->mhash);
  82         kfree(old);
  83 }
  84
  85 static int br_mdb_copy(struct net_bridge_mdb_htable *new,
  86                        struct net_bridge_mdb_htable *old,
  87                        int elasticity)
  88 {
  89         struct net_bridge_mdb_entry *mp;
  90         struct hlist_node *p;
  91         int maxlen;
  92         int len;
  93         int i;
  94
  95         for (i = 0; i < old->max; i++)
  96                 hlist_for_each_entry(mp, p, &old->mhash[i], hlist[old->ver])
  97                         hlist_add_head(&mp->hlist[new->ver],
  98                                        &new->mhash[br_ip_hash(new, mp->addr)]);
  99
 100         if (!elasticity)
 101                 return 0;
 102
 103         maxlen = 0;
 104         for (i = 0; i < new->max; i++) {
 105                 len = 0;
 106                 hlist_for_each_entry(mp, p, &new->mhash[i], hlist[new->ver])
 107                         len++;
 108                 if (len > maxlen)
 109                         maxlen = len;
 110         }
 111
 112         return maxlen > elasticity ? -EINVAL : 0;
 113 }
 114
 115 static void br_multicast_free_pg(struct rcu_head *head)
 116 {
 117         struct net_bridge_port_group *p =
 118                 container_of(head, struct net_bridge_port_group, rcu);
 119
 120         kfree(p);
 121 }
 122
 123 static void br_multicast_free_group(struct rcu_head *head)
 124 {
 125         struct net_bridge_mdb_entry *mp =
 126                 container_of(head, struct net_bridge_mdb_entry, rcu);
 127
 128         kfree(mp);
 129 }
 130
 131 static void br_multicast_group_expired(unsigned long data)
 132 {
 133         struct net_bridge_mdb_entry *mp = (void *)data;
 134         struct net_bridge *br = mp->br;
 135         struct net_bridge_mdb_htable *mdb;
 136
 137         spin_lock(&br->multicast_lock);
 138         if (!netif_running(br->dev) || timer_pending(&mp->timer))
 139                 goto out;
 140
 141         if (!hlist_unhashed(&mp->mglist))
 142                 hlist_del_init(&mp->mglist);
 143
 144         if (mp->ports)
 145                 goto out;
 146
 147         mdb = br->mdb;
 148         hlist_del_rcu(&mp->hlist[mdb->ver]);
 149         mdb->size--;
 150
 151         del_timer(&mp->query_timer);
 152         call_rcu_bh(&mp->rcu, br_multicast_free_group);
 153
 154 out:
 155         spin_unlock(&br->multicast_lock);
 156 }
 157
 158 static void br_multicast_del_pg(struct net_bridge *br,
 159                                 struct net_bridge_port_group *pg)
 160 {
 161         struct net_bridge_mdb_htable *mdb = br->mdb;
 162         struct net_bridge_mdb_entry *mp;
 163         struct net_bridge_port_group *p;
 164         struct net_bridge_port_group **pp;
 165
 166         mp = br_mdb_ip_get(mdb, pg->addr);
 167         if (WARN_ON(!mp))
 168                 return;
 169
 170         for (pp = &mp->ports; (p = *pp); pp = &p->next) {
 171                 if (p != pg)
 172                         continue;
 173
 174                 *pp = p->next;
 175                 hlist_del_init(&p->mglist);
 176                 del_timer(&p->timer);
 177                 del_timer(&p->query_timer);
 178                 call_rcu_bh(&p->rcu, br_multicast_free_pg);
 179
 180                 if (!mp->ports && hlist_unhashed(&mp->mglist) &&
 181                     netif_running(br->dev))
 182                         mod_timer(&mp->timer, jiffies);
 183
 184                 return;
 185         }
 186
 187         WARN_ON(1);
 188 }
 189
 190 static void br_multicast_port_group_expired(unsigned long data)
 191 {
 192         struct net_bridge_port_group *pg = (void *)data;
 193         struct net_bridge *br = pg->port->br;
 194
 195         spin_lock(&br->multicast_lock);
 196         if (!netif_running(br->dev) || timer_pending(&pg->timer) ||
 197             hlist_unhashed(&pg->mglist))
 198                 goto out;
 199
 200         br_multicast_del_pg(br, pg);
 201
 202 out:
 203         spin_unlock(&br->multicast_lock);
 204 }
 205
 206 static int br_mdb_rehash(struct net_bridge_mdb_htable **mdbp, int max,
 207                          int elasticity)
 208 {
 209         struct net_bridge_mdb_htable *old = *mdbp;
 210         struct net_bridge_mdb_htable *mdb;
 211         int err;
 212
 213         mdb = kmalloc(sizeof(*mdb), GFP_ATOMIC);
 214         if (!mdb)
 215                 return -ENOMEM;
 216
 217         mdb->max = max;
 218         mdb->old = old;
 219
 220         mdb->mhash = kzalloc(max * sizeof(*mdb->mhash), GFP_ATOMIC);
 221         if (!mdb->mhash) {
 222                 kfree(mdb);
 223                 return -ENOMEM;
 224         }
 225
 226         mdb->size = old ? old->size : 0;
 227         mdb->ver = old ? old->ver ^ 1 : 0;
 228
 229         if (!old || elasticity)
 230                 get_random_bytes(&mdb->secret, sizeof(mdb->secret));
 231         else
 232                 mdb->secret = old->secret;
 233
 234         if (!old)
 235                 goto out;
 236
 237         err = br_mdb_copy(mdb, old, elasticity);
 238         if (err) {
 239                 kfree(mdb->mhash);
 240                 kfree(mdb);
 241                 return err;
 242         }
 243
 244         call_rcu_bh(&mdb->rcu, br_mdb_free);
 245
 246 out:
 247         rcu_assign_pointer(*mdbp, mdb);
 248
 249         return 0;
 250 }
 251
 252 static struct sk_buff *br_multicast_alloc_query(struct net_bridge *br,
 253                                                 __be32 group)
 254 {
 255         struct sk_buff *skb;
 256         struct igmphdr *ih;
 257         struct ethhdr *eth;
 258         struct iphdr *iph;
 259
 260         skb = netdev_alloc_skb_ip_align(br->dev, sizeof(*eth) + sizeof(*iph) +
 261                                                  sizeof(*ih) + 4);
 262         if (!skb)
 263                 goto out;
 264
 265         skb->protocol = htons(ETH_P_IP);
 266
 267         skb_reset_mac_header(skb);
 268         eth = eth_hdr(skb);
 269
 270         memcpy(eth->h_source, br->dev->dev_addr, 6);
 271         eth->h_dest[0] = 1;
 272         eth->h_dest[1] = 0;
 273         eth->h_dest[2] = 0x5e;
 274         eth->h_dest[3] = 0;
 275         eth->h_dest[4] = 0;
 276         eth->h_dest[5] = 1;
 277         eth->h_proto = htons(ETH_P_IP);
 278         skb_put(skb, sizeof(*eth));
 279
 280         skb_set_network_header(skb, skb->len);
 281         iph = ip_hdr(skb);
 282
 283         iph->version = 4;
 284         iph->ihl = 6;
 285         iph->tos = 0xc0;
 286         iph->tot_len = htons(sizeof(*iph) + sizeof(*ih) + 4);
 287         iph->id = 0;
 288         iph->frag_off = htons(IP_DF);
 289         iph->ttl = 1;
 290         iph->protocol = IPPROTO_IGMP;
 291         iph->saddr = 0;
 292         iph->daddr = htonl(INADDR_ALLHOSTS_GROUP);
 293         ((u8 *)&iph[1])[0] = IPOPT_RA;
 294         ((u8 *)&iph[1])[1] = 4;
 295         ((u8 *)&iph[1])[2] = 0;
 296         ((u8 *)&iph[1])[3] = 0;
 297         ip_send_check(iph);
 298         skb_put(skb, 24);
 299
 300         skb_set_transport_header(skb, skb->len);
 301         ih = igmp_hdr(skb);
 302         ih->type = IGMP_HOST_MEMBERSHIP_QUERY;
 303         ih->code = (group ? br->multicast_last_member_interval :
 304                             br->multicast_query_response_interval) /
 305                    (HZ / IGMP_TIMER_SCALE);
 306         ih->group = group;
 307         ih->csum = 0;
 308         ih->csum = ip_compute_csum((void *)ih, sizeof(struct igmphdr));
 309         skb_put(skb, sizeof(*ih));
 310
 311         __skb_pull(skb, sizeof(*eth));
 312
 313 out:
 314         return skb;
 315 }
 316
 317 static void br_multicast_send_group_query(struct net_bridge_mdb_entry *mp)
 318 {
 319         struct net_bridge *br = mp->br;
 320         struct sk_buff *skb;
 321
 322         skb = br_multicast_alloc_query(br, mp->addr);
 323         if (!skb)
 324                 goto timer;
 325
 326         netif_rx(skb);
 327
 328 timer:
 329         if (++mp->queries_sent < br->multicast_last_member_count)
 330                 mod_timer(&mp->query_timer,
 331                           jiffies + br->multicast_last_member_interval);
 332 }
 333
 334 static void br_multicast_group_query_expired(unsigned long data)
 335 {
 336         struct net_bridge_mdb_entry *mp = (void *)data;
 337         struct net_bridge *br = mp->br;
 338
 339         spin_lock(&br->multicast_lock);
 340         if (!netif_running(br->dev) || hlist_unhashed(&mp->mglist) ||
 341             mp->queries_sent >= br->multicast_last_member_count)
 342                 goto out;
 343
 344         br_multicast_send_group_query(mp);
 345
 346 out:
 347         spin_unlock(&br->multicast_lock);
 348 }
 349
 350 static void br_multicast_send_port_group_query(struct net_bridge_port_group *pg)
 351 {
 352         struct net_bridge_port *port = pg->port;
 353         struct net_bridge *br = port->br;
 354         struct sk_buff *skb;
 355
 356         skb = br_multicast_alloc_query(br, pg->addr);
 357         if (!skb)
 358                 goto timer;
 359
 360         br_deliver(port, skb);
 361
 362 timer:
 363         if (++pg->queries_sent < br->multicast_last_member_count)
 364                 mod_timer(&pg->query_timer,
 365                           jiffies + br->multicast_last_member_interval);
 366 }
 367
 368 static void br_multicast_port_group_query_expired(unsigned long data)
 369 {
 370         struct net_bridge_port_group *pg = (void *)data;
 371         struct net_bridge_port *port = pg->port;
 372         struct net_bridge *br = port->br;
 373
 374         spin_lock(&br->multicast_lock);
 375         if (!netif_running(br->dev) || hlist_unhashed(&pg->mglist) ||
 376             pg->queries_sent >= br->multicast_last_member_count)
 377                 goto out;
 378
 379         br_multicast_send_port_group_query(pg);
 380
 381 out:
 382         spin_unlock(&br->multicast_lock);
 383 }
 384
 385 static struct net_bridge_mdb_entry *br_multicast_get_group(
 386         struct net_bridge *br, struct net_bridge_port *port, __be32 group,
 387         int hash)
 388 {
 389         struct net_bridge_mdb_htable *mdb = br->mdb;
 390         struct net_bridge_mdb_entry *mp;
 391         struct hlist_node *p;
 392         unsigned count = 0;
 393         unsigned max;
 394         int elasticity;
 395         int err;
 396
 397         hlist_for_each_entry(mp, p, &mdb->mhash[hash], hlist[mdb->ver]) {
 398                 count++;
 399                 if (unlikely(group == mp->addr)) {
 400                         return mp;
 401                 }
 402         }
 403
 404         elasticity = 0;
 405         max = mdb->max;
 406
 407         if (unlikely(count > br->hash_elasticity && count)) {
 408                 if (net_ratelimit())
 409                         printk(KERN_INFO "%s: Multicast hash table "
 410                                "chain limit reached: %s\n",
 411                                br->dev->name, port ? port->dev->name :
 412                                                      br->dev->name);
 413
 414                 elasticity = br->hash_elasticity;
 415         }
 416
 417         if (mdb->size >= max) {
 418                 max *= 2;
 419                 if (unlikely(max >= br->hash_max)) {
 420                         printk(KERN_WARNING "%s: Multicast hash table maximum "
 421                                "reached, disabling snooping: %s, %d\n",
 422                                br->dev->name, port ? port->dev->name :
 423                                                      br->dev->name,
 424                                max);
 425                         err = -E2BIG;
 426 disable:
 427                         br->multicast_disabled = 1;
 428                         goto err;
 429                 }
 430         }
 431
 432         if (max > mdb->max || elasticity) {
 433                 if (mdb->old) {
 434                         if (net_ratelimit())
 435                                 printk(KERN_INFO "%s: Multicast hash table "
 436                                        "on fire: %s\n",
 437                                        br->dev->name, port ? port->dev->name :
 438                                                              br->dev->name);
 439                         err = -EEXIST;
 440                         goto err;
 441                 }
 442
 443                 err = br_mdb_rehash(&br->mdb, max, elasticity);
 444                 if (err) {
 445                         printk(KERN_WARNING "%s: Cannot rehash multicast "
 446                                "hash table, disabling snooping: "
 447                                "%s, %d, %d\n",
 448                                br->dev->name, port ? port->dev->name :
 449                                                      br->dev->name,
 450                                mdb->size, err);
 451                         goto disable;
 452                 }
 453
 454                 err = -EAGAIN;
 455                 goto err;
 456         }
 457
 458         return NULL;
 459
 460 err:
 461         mp = ERR_PTR(err);
 462         return mp;
 463 }
 464
 465 static struct net_bridge_mdb_entry *br_multicast_new_group(
 466         struct net_bridge *br, struct net_bridge_port *port, __be32 group)
 467 {
 468         struct net_bridge_mdb_htable *mdb = br->mdb;
 469         struct net_bridge_mdb_entry *mp;
 470         int hash;
 471
 472         if (!mdb) {
 473                 if (br_mdb_rehash(&br->mdb, BR_HASH_SIZE, 0))
 474                         return NULL;
 475                 goto rehash;
 476         }
 477
 478         hash = br_ip_hash(mdb, group);
 479         mp = br_multicast_get_group(br, port, group, hash);
 480         switch (PTR_ERR(mp)) {
 481         case 0:
 482                 break;
 483
 484         case -EAGAIN:
 485 rehash:
 486                 mdb = br->mdb;
 487                 hash = br_ip_hash(mdb, group);
 488                 break;
 489
 490         default:
 491                 goto out;
 492         }
 493
 494         mp = kzalloc(sizeof(*mp), GFP_ATOMIC);
 495         if (unlikely(!mp))
 496                 goto out;
 497
 498         mp->br = br;
 499         mp->addr = group;
 500         setup_timer(&mp->timer, br_multicast_group_expired,
 501                     (unsigned long)mp);
 502         setup_timer(&mp->query_timer, br_multicast_group_query_expired,
 503                     (unsigned long)mp);
 504
 505         hlist_add_head_rcu(&mp->hlist[mdb->ver], &mdb->mhash[hash]);
 506         mdb->size++;
 507
 508 out:
 509         return mp;
 510 }
 511
 512 static int br_multicast_add_group(struct net_bridge *br,
 513                                   struct net_bridge_port *port, __be32 group)
 514 {
 515         struct net_bridge_mdb_entry *mp;
 516         struct net_bridge_port_group *p;
 517         struct net_bridge_port_group **pp;
 518         unsigned long now = jiffies;
 519         int err;
 520
 521         if (ipv4_is_local_multicast(group))
 522                 return 0;
 523
 524         spin_lock(&br->multicast_lock);
 525         if (!netif_running(br->dev) ||
 526             (port && port->state == BR_STATE_DISABLED))
 527                 goto out;
 528
 529         mp = br_multicast_new_group(br, port, group);
 530         err = PTR_ERR(mp);
 531         if (unlikely(IS_ERR(mp) || !mp))
 532                 goto err;
 533
 534         if (!port) {
 535                 hlist_add_head(&mp->mglist, &br->mglist);
 536                 mod_timer(&mp->timer, now + br->multicast_membership_interval);
 537                 goto out;
 538         }
 539
 540         for (pp = &mp->ports; (p = *pp); pp = &p->next) {
 541                 if (p->port == port)
 542                         goto found;
 543                 if ((unsigned long)p->port < (unsigned long)port)
 544                         break;
 545         }
 546
 547         p = kzalloc(sizeof(*p), GFP_ATOMIC);
 548         err = -ENOMEM;
 549         if (unlikely(!p))
 550                 goto err;
 551
 552         p->addr = group;
 553         p->port = port;
 554         p->next = *pp;
 555         hlist_add_head(&p->mglist, &port->mglist);
 556         setup_timer(&p->timer, br_multicast_port_group_expired,
 557                     (unsigned long)p);
 558         setup_timer(&p->query_timer, br_multicast_port_group_query_expired,
 559                     (unsigned long)p);
 560
 561         rcu_assign_pointer(*pp, p);
 562
 563 found:
 564         mod_timer(&p->timer, now + br->multicast_membership_interval);
 565 out:
 566         err = 0;
 567
 568 err:
 569         spin_unlock(&br->multicast_lock);
 570         return err;
 571 }
 572
 573 static void br_multicast_router_expired(unsigned long data)
 574 {
 575         struct net_bridge_port *port = (void *)data;
 576         struct net_bridge *br = port->br;
 577
 578         spin_lock(&br->multicast_lock);
 579         if (port->multicast_router != 1 ||
 580             timer_pending(&port->multicast_router_timer) ||
 581             hlist_unhashed(&port->rlist))
 582                 goto out;
 583
 584         hlist_del_init_rcu(&port->rlist);
 585
 586 out:
 587         spin_unlock(&br->multicast_lock);
 588 }
 589
 590 static void br_multicast_local_router_expired(unsigned long data)
 591 {
 592 }
 593
 594 static void br_multicast_send_query(struct net_bridge *br,
 595                                     struct net_bridge_port *port, u32 sent)
 596 {
 597         unsigned long time;
 598         struct sk_buff *skb;
 599
 600         if (!netif_running(br->dev) || br->multicast_disabled ||
 601             timer_pending(&br->multicast_querier_timer))
 602                 return;
 603
 604         skb = br_multicast_alloc_query(br, 0);
 605         if (!skb)
 606                 goto timer;
 607
 608         if (port) {
 609                 __skb_push(skb, sizeof(struct ethhdr));
 610                 skb->dev = port->dev;
 611                 NF_HOOK(PF_BRIDGE, NF_BR_LOCAL_OUT, skb, NULL, skb->dev,
 612                         dev_queue_xmit);
 613         } else
 614                 netif_rx(skb);
 615
 616 timer:
 617         time = jiffies;
 618         time += sent < br->multicast_startup_query_count ?
 619                 br->multicast_startup_query_interval :
 620                 br->multicast_query_interval;
 621         mod_timer(port ? &port->multicast_query_timer :
 622                          &br->multicast_query_timer, time);
 623 }
 624
 625 static void br_multicast_port_query_expired(unsigned long data)
 626 {
 627         struct net_bridge_port *port = (void *)data;
 628         struct net_bridge *br = port->br;
 629
 630         spin_lock(&br->multicast_lock);
 631         if (port->state == BR_STATE_DISABLED ||
 632             port->state == BR_STATE_BLOCKING)
 633                 goto out;
 634
 635         if (port->multicast_startup_queries_sent <
 636             br->multicast_startup_query_count)
 637                 port->multicast_startup_queries_sent++;
 638
 639         br_multicast_send_query(port->br, port,
 640                                 port->multicast_startup_queries_sent);
 641
 642 out:
 643         spin_unlock(&br->multicast_lock);
 644 }
 645
 646 void br_multicast_add_port(struct net_bridge_port *port)
 647 {
 648         port->multicast_router = 1;
 649
 650         setup_timer(&port->multicast_router_timer, br_multicast_router_expired,
 651                     (unsigned long)port);
 652         setup_timer(&port->multicast_query_timer,
 653                     br_multicast_port_query_expired, (unsigned long)port);
 654 }
 655
 656 void br_multicast_del_port(struct net_bridge_port *port)
 657 {
 658         del_timer_sync(&port->multicast_router_timer);
 659 }
 660
 661 static void __br_multicast_enable_port(struct net_bridge_port *port)
 662 {
 663         port->multicast_startup_queries_sent = 0;
 664
 665         if (try_to_del_timer_sync(&port->multicast_query_timer) >= 0 ||
 666             del_timer(&port->multicast_query_timer))
 667                 mod_timer(&port->multicast_query_timer, jiffies);
 668 }
 669
 670 void br_multicast_enable_port(struct net_bridge_port *port)
 671 {
 672         struct net_bridge *br = port->br;
 673
 674         spin_lock(&br->multicast_lock);
 675         if (br->multicast_disabled || !netif_running(br->dev))
 676                 goto out;
 677
 678         __br_multicast_enable_port(port);
 679
 680 out:
 681         spin_unlock(&br->multicast_lock);
 682 }
 683
 684 void br_multicast_disable_port(struct net_bridge_port *port)
 685 {
 686         struct net_bridge *br = port->br;
 687         struct net_bridge_port_group *pg;
 688         struct hlist_node *p, *n;
 689
 690         spin_lock(&br->multicast_lock);
 691         hlist_for_each_entry_safe(pg, p, n, &port->mglist, mglist)
 692                 br_multicast_del_pg(br, pg);
 693
 694         if (!hlist_unhashed(&port->rlist))
 695                 hlist_del_init_rcu(&port->rlist);
 696         del_timer(&port->multicast_router_timer);
 697         del_timer(&port->multicast_query_timer);
 698         spin_unlock(&br->multicast_lock);
 699 }
 700
 701 static int br_multicast_igmp3_report(struct net_bridge *br,
 702                                      struct net_bridge_port *port,
 703                                      struct sk_buff *skb)
 704 {
 705         struct igmpv3_report *ih;
 706         struct igmpv3_grec *grec;
 707         int i;
 708         int len;
 709         int num;
 710         int type;
 711         int err = 0;
 712         __be32 group;
 713
 714         if (!pskb_may_pull(skb, sizeof(*ih)))
 715                 return -EINVAL;
 716
 717         ih = igmpv3_report_hdr(skb);
 718         num = ntohs(ih->ngrec);
 719         len = sizeof(*ih);
 720
 721         for (i = 0; i < num; i++) {
 722                 len += sizeof(*grec);
 723                 if (!pskb_may_pull(skb, len))
 724                         return -EINVAL;
 725
 726                 grec = (void *)(skb->data + len - sizeof(*grec));
 727                 group = grec->grec_mca;
 728                 type = grec->grec_type;
 729
 730                 len += ntohs(grec->grec_nsrcs) * 4;
 731                 if (!pskb_may_pull(skb, len))
 732                         return -EINVAL;
 733
 734                 /* We treat this as an IGMPv2 report for now. */
 735                 switch (type) {
 736                 case IGMPV3_MODE_IS_INCLUDE:
 737                 case IGMPV3_MODE_IS_EXCLUDE:
 738                 case IGMPV3_CHANGE_TO_INCLUDE:
 739                 case IGMPV3_CHANGE_TO_EXCLUDE:
 740                 case IGMPV3_ALLOW_NEW_SOURCES:
 741                 case IGMPV3_BLOCK_OLD_SOURCES:
 742                         break;
 743
 744                 default:
 745                         continue;
 746                 }
 747
 748                 err = br_multicast_add_group(br, port, group);
 749                 if (err)
 750                         break;
 751         }
 752
 753         return err;
 754 }
 755
 756 static void br_multicast_add_router(struct net_bridge *br,
 757                                     struct net_bridge_port *port)
 758 {
 759         struct hlist_node *p;
 760         struct hlist_node **h;
 761
 762         for (h = &br->router_list.first;
 763              (p = *h) &&
 764              (unsigned long)container_of(p, struct net_bridge_port, rlist) >
 765              (unsigned long)port;
 766              h = &p->next)
 767                 ;
 768
 769         port->rlist.pprev = h;
 770         port->rlist.next = p;
 771         rcu_assign_pointer(*h, &port->rlist);
 772         if (p)
 773                 p->pprev = &port->rlist.next;
 774 }
 775
 776 static void br_multicast_mark_router(struct net_bridge *br,
 777                                      struct net_bridge_port *port)
 778 {
 779         unsigned long now = jiffies;
 780
 781         if (!port) {
 782                 if (br->multicast_router == 1)
 783                         mod_timer(&br->multicast_router_timer,
 784                                   now + br->multicast_querier_interval);
 785                 return;
 786         }
 787
 788         if (port->multicast_router != 1)
 789                 return;
 790
 791         if (!hlist_unhashed(&port->rlist))
 792                 goto timer;
 793
 794         br_multicast_add_router(br, port);
 795
 796 timer:
 797         mod_timer(&port->multicast_router_timer,
 798                   now + br->multicast_querier_interval);
 799 }
 800
 801 static void br_multicast_query_received(struct net_bridge *br,
 802                                         struct net_bridge_port *port,
 803                                         __be32 saddr)
 804 {
 805         if (saddr)
 806                 mod_timer(&br->multicast_querier_timer,
 807                           jiffies + br->multicast_querier_interval);
 808         else if (timer_pending(&br->multicast_querier_timer))
 809                 return;
 810
 811         br_multicast_mark_router(br, port);
 812 }
 813
 814 static int br_multicast_query(struct net_bridge *br,
 815                               struct net_bridge_port *port,
 816                               struct sk_buff *skb)
 817 {
 818         struct iphdr *iph = ip_hdr(skb);
 819         struct igmphdr *ih = igmp_hdr(skb);
 820         struct net_bridge_mdb_entry *mp;
 821         struct igmpv3_query *ih3;
 822         struct net_bridge_port_group *p;
 823         struct net_bridge_port_group **pp;
 824         unsigned long max_delay;
 825         unsigned long now = jiffies;
 826         __be32 group;
 827         int err = 0;
 828
 829         spin_lock(&br->multicast_lock);
 830         if (!netif_running(br->dev) ||
 831             (port && port->state == BR_STATE_DISABLED))
 832                 goto out;
 833
 834         br_multicast_query_received(br, port, iph->saddr);
 835
 836         group = ih->group;
 837
 838         if (skb->len == sizeof(*ih)) {
 839                 max_delay = ih->code * (HZ / IGMP_TIMER_SCALE);
 840
 841                 if (!max_delay) {
 842                         max_delay = 10 * HZ;
 843                         group = 0;
 844                 }
 845         } else {
 846                 if (!pskb_may_pull(skb, sizeof(struct igmpv3_query))) {
 847                         err = -EINVAL;
 848                         goto out;
 849                 }
 850
 851                 ih3 = igmpv3_query_hdr(skb);
 852                 if (ih3->nsrcs)
 853                         goto out;
 854
 855                 max_delay = ih3->code ?
 856                             IGMPV3_MRC(ih3->code) * (HZ / IGMP_TIMER_SCALE) : 1;
 857         }
 858
 859         if (!group)
 860                 goto out;
 861
 862         mp = br_mdb_ip_get(br->mdb, group);
 863         if (!mp)
 864                 goto out;
 865
 866         max_delay *= br->multicast_last_member_count;
 867
 868         if (!hlist_unhashed(&mp->mglist) &&
 869             (timer_pending(&mp->timer) ?
 870              time_after(mp->timer.expires, now + max_delay) :
 871              try_to_del_timer_sync(&mp->timer) >= 0))
 872                 mod_timer(&mp->timer, now + max_delay);
 873
 874         for (pp = &mp->ports; (p = *pp); pp = &p->next) {
 875                 if (timer_pending(&p->timer) ?
 876                     time_after(p->timer.expires, now + max_delay) :
 877                     try_to_del_timer_sync(&p->timer) >= 0)
 878                         mod_timer(&mp->timer, now + max_delay);
 879         }
 880
 881 out:
 882         spin_unlock(&br->multicast_lock);
 883         return err;
 884 }
 885
 886 static void br_multicast_leave_group(struct net_bridge *br,
 887                                      struct net_bridge_port *port,
 888                                      __be32 group)
 889 {
 890         struct net_bridge_mdb_htable *mdb;
 891         struct net_bridge_mdb_entry *mp;
 892         struct net_bridge_port_group *p;
 893         unsigned long now;
 894         unsigned long time;
 895
 896         if (ipv4_is_local_multicast(group))
 897                 return;
 898
 899         spin_lock(&br->multicast_lock);
 900         if (!netif_running(br->dev) ||
 901             (port && port->state == BR_STATE_DISABLED) ||
 902             timer_pending(&br->multicast_querier_timer))
 903                 goto out;
 904
 905         mdb = br->mdb;
 906         mp = br_mdb_ip_get(mdb, group);
 907         if (!mp)
 908                 goto out;
 909
 910         now = jiffies;
 911         time = now + br->multicast_last_member_count *
 912                      br->multicast_last_member_interval;
 913
 914         if (!port) {
 915                 if (!hlist_unhashed(&mp->mglist) &&
 916                     (timer_pending(&mp->timer) ?
 917                      time_after(mp->timer.expires, time) :
 918                      try_to_del_timer_sync(&mp->timer) >= 0)) {
 919                         mod_timer(&mp->timer, time);
 920
 921                         mp->queries_sent = 0;
 922                         mod_timer(&mp->query_timer, now);
 923                 }
 924
 925                 goto out;
 926         }
 927
 928         for (p = mp->ports; p; p = p->next) {
 929                 if (p->port != port)
 930                         continue;
 931
 932                 if (!hlist_unhashed(&p->mglist) &&
 933                     (timer_pending(&p->timer) ?
 934                      time_after(p->timer.expires, time) :
 935                      try_to_del_timer_sync(&p->timer) >= 0)) {
 936                         mod_timer(&p->timer, time);
 937
 938                         p->queries_sent = 0;
 939                         mod_timer(&p->query_timer, now);
 940                 }
 941
 942                 break;
 943         }
 944
 945 out:
 946         spin_unlock(&br->multicast_lock);
 947 }
 948
 949 static int br_multicast_ipv4_rcv(struct net_bridge *br,
 950                                  struct net_bridge_port *port,
 951                                  struct sk_buff *skb)
 952 {
 953         struct sk_buff *skb2 = skb;
 954         struct iphdr *iph;
 955         struct igmphdr *ih;
 956         unsigned len;
 957         unsigned offset;
 958         int err;
 959
 960         /* We treat OOM as packet loss for now. */
 961         if (!pskb_may_pull(skb, sizeof(*iph)))
 962                 return -EINVAL;
 963
 964         iph = ip_hdr(skb);
 965
 966         if (iph->ihl < 5 || iph->version != 4)
 967                 return -EINVAL;
 968
 969         if (!pskb_may_pull(skb, ip_hdrlen(skb)))
 970                 return -EINVAL;
 971
 972         iph = ip_hdr(skb);
 973
 974         if (unlikely(ip_fast_csum((u8 *)iph, iph->ihl)))
 975                 return -EINVAL;
 976
 977         if (iph->protocol != IPPROTO_IGMP)
 978                 return 0;
 979
 980         len = ntohs(iph->tot_len);
 981         if (skb->len < len || len < ip_hdrlen(skb))
 982                 return -EINVAL;
 983
 984         if (skb->len > len) {
 985                 skb2 = skb_clone(skb, GFP_ATOMIC);
 986                 if (!skb2)
 987                         return -ENOMEM;
 988
 989                 err = pskb_trim_rcsum(skb2, len);
 990                 if (err)
 991                         goto err_out;
 992         }
 993
 994         len -= ip_hdrlen(skb2);
 995         offset = skb_network_offset(skb2) + ip_hdrlen(skb2);
 996         __skb_pull(skb2, offset);
 997         skb_reset_transport_header(skb2);
 998
 999         err = -EINVAL;
1000         if (!pskb_may_pull(skb2, sizeof(*ih)))
1001                 goto out;
1002
1003         iph = ip_hdr(skb2);
1004
1005         switch (skb2->ip_summed) {
1006         case CHECKSUM_COMPLETE:
1007                 if (!csum_fold(skb2->csum))
1008                         break;
1009                 /* fall through */
1010         case CHECKSUM_NONE:
1011                 skb2->csum = 0;
1012                 if (skb_checksum_complete(skb2))
1013                         goto out;
1014         }
1015
1016         err = 0;
1017
1018         BR_INPUT_SKB_CB(skb)->igmp = 1;
1019         ih = igmp_hdr(skb2);
1020
1021         switch (ih->type) {
1022         case IGMP_HOST_MEMBERSHIP_REPORT:
1023         case IGMPV2_HOST_MEMBERSHIP_REPORT:
1024                 BR_INPUT_SKB_CB(skb2)->mrouters_only = 1;
1025                 err = br_multicast_add_group(br, port, ih->group);
1026                 break;
1027         case IGMPV3_HOST_MEMBERSHIP_REPORT:
1028                 err = br_multicast_igmp3_report(br, port, skb2);
1029                 break;
1030         case IGMP_HOST_MEMBERSHIP_QUERY:
1031                 err = br_multicast_query(br, port, skb2);
1032                 break;
1033         case IGMP_HOST_LEAVE_MESSAGE:
1034                 br_multicast_leave_group(br, port, ih->group);
1035                 break;
1036         }
1037
1038 out:
1039         __skb_push(skb2, offset);
1040 err_out:
1041         if (skb2 != skb)
1042                 kfree_skb(skb2);
1043         return err;
1044 }
1045
1046 int br_multicast_rcv(struct net_bridge *br, struct net_bridge_port *port,
1047                      struct sk_buff *skb)
1048 {
1049         BR_INPUT_SKB_CB(skb)->igmp = 0;
1050         BR_INPUT_SKB_CB(skb)->mrouters_only = 0;
1051
1052         if (br->multicast_disabled)
1053                 return 0;
1054
1055         switch (skb->protocol) {
1056         case htons(ETH_P_IP):
1057                 return br_multicast_ipv4_rcv(br, port, skb);
1058         }
1059
1060         return 0;
1061 }
1062
1063 static void br_multicast_query_expired(unsigned long data)
1064 {
1065         struct net_bridge *br = (void *)data;
1066
1067         spin_lock(&br->multicast_lock);
1068         if (br->multicast_startup_queries_sent <
1069             br->multicast_startup_query_count)
1070                 br->multicast_startup_queries_sent++;
1071
1072         br_multicast_send_query(br, NULL, br->multicast_startup_queries_sent);
1073
1074         spin_unlock(&br->multicast_lock);
1075 }
1076
1077 void br_multicast_init(struct net_bridge *br)
1078 {
1079         br->hash_elasticity = 4;
1080         br->hash_max = 512;
1081
1082         br->multicast_router = 1;
1083         br->multicast_last_member_count = 2;
1084         br->multicast_startup_query_count = 2;
1085
1086         br->multicast_last_member_interval = HZ;
1087         br->multicast_query_response_interval = 10 * HZ;
1088         br->multicast_startup_query_interval = 125 * HZ / 4;
1089         br->multicast_query_interval = 125 * HZ;
1090         br->multicast_querier_interval = 255 * HZ;
1091         br->multicast_membership_interval = 260 * HZ;
1092
1093         spin_lock_init(&br->multicast_lock);
1094         setup_timer(&br->multicast_router_timer,
1095                     br_multicast_local_router_expired, 0);
1096         setup_timer(&br->multicast_querier_timer,
1097                     br_multicast_local_router_expired, 0);
1098         setup_timer(&br->multicast_query_timer, br_multicast_query_expired,
1099                     (unsigned long)br);
1100 }
1101
1102 void br_multicast_open(struct net_bridge *br)
1103 {
1104         br->multicast_startup_queries_sent = 0;
1105
1106         if (br->multicast_disabled)
1107                 return;
1108
1109         mod_timer(&br->multicast_query_timer, jiffies);
1110 }
1111
1112 void br_multicast_stop(struct net_bridge *br)
1113 {
1114         struct net_bridge_mdb_htable *mdb;
1115         struct net_bridge_mdb_entry *mp;
1116         struct hlist_node *p, *n;
1117         u32 ver;
1118         int i;
1119
1120         del_timer_sync(&br->multicast_router_timer);
1121         del_timer_sync(&br->multicast_querier_timer);
1122         del_timer_sync(&br->multicast_query_timer);
1123
1124         spin_lock_bh(&br->multicast_lock);
1125         mdb = br->mdb;
1126         if (!mdb)
1127                 goto out;
1128
1129         br->mdb = NULL;
1130
1131         ver = mdb->ver;
1132         for (i = 0; i < mdb->max; i++) {
1133                 hlist_for_each_entry_safe(mp, p, n, &mdb->mhash[i],
1134                                           hlist[ver]) {
1135                         del_timer(&mp->timer);
1136                         del_timer(&mp->query_timer);
1137                         call_rcu_bh(&mp->rcu, br_multicast_free_group);
1138                 }
1139         }
1140
1141         if (mdb->old) {
1142                 spin_unlock_bh(&br->multicast_lock);
1143                 rcu_barrier_bh();
1144                 spin_lock_bh(&br->multicast_lock);
1145                 WARN_ON(mdb->old);
1146         }
1147
1148         mdb->old = mdb;
1149         call_rcu_bh(&mdb->rcu, br_mdb_free);
1150
1151 out:
1152         spin_unlock_bh(&br->multicast_lock);
1153 }
1154
1155 int br_multicast_set_router(struct net_bridge *br, unsigned long val)
1156 {
1157         int err = -ENOENT;
1158
1159         spin_lock_bh(&br->multicast_lock);
1160         if (!netif_running(br->dev))
1161                 goto unlock;
1162
1163         switch (val) {
1164         case 0:
1165         case 2:
1166                 del_timer(&br->multicast_router_timer);
1167                 /* fall through */
1168         case 1:
1169                 br->multicast_router = val;
1170                 err = 0;
1171                 break;
1172
1173         default:
1174                 err = -EINVAL;
1175                 break;
1176         }
1177
1178 unlock:
1179         spin_unlock_bh(&br->multicast_lock);
1180
1181         return err;
1182 }
1183
1184 int br_multicast_set_port_router(struct net_bridge_port *p, unsigned long val)
1185 {
1186         struct net_bridge *br = p->br;
1187         int err = -ENOENT;
1188
1189         spin_lock(&br->multicast_lock);
1190         if (!netif_running(br->dev) || p->state == BR_STATE_DISABLED)
1191                 goto unlock;
1192
1193         switch (val) {
1194         case 0:
1195         case 1:
1196         case 2:
1197                 p->multicast_router = val;
1198                 err = 0;
1199
1200                 if (val < 2 && !hlist_unhashed(&p->rlist))
1201                         hlist_del_init_rcu(&p->rlist);
1202
1203                 if (val == 1)
1204                         break;
1205
1206                 del_timer(&p->multicast_router_timer);
1207
1208                 if (val == 0)
1209                         break;
1210
1211                 br_multicast_add_router(br, p);
1212                 break;
1213
1214         default:
1215                 err = -EINVAL;
1216                 break;
1217         }
1218
1219 unlock:
1220         spin_unlock(&br->multicast_lock);
1221
1222         return err;
1223 }
1224
1225 int br_multicast_toggle(struct net_bridge *br, unsigned long val)
1226 {
1227         struct net_bridge_port *port;
1228         int err = -ENOENT;
1229
1230         spin_lock(&br->multicast_lock);
1231         if (!netif_running(br->dev))
1232                 goto unlock;
1233
1234         err = 0;
1235         if (br->multicast_disabled == !val)
1236                 goto unlock;
1237
1238         br->multicast_disabled = !val;
1239         if (br->multicast_disabled)
1240                 goto unlock;
1241
1242         if (br->mdb) {
1243                 if (br->mdb->old) {
1244                         err = -EEXIST;
1245 rollback:
1246                         br->multicast_disabled = !!val;
1247                         goto unlock;
1248                 }
1249
1250                 err = br_mdb_rehash(&br->mdb, br->mdb->max,
1251                                     br->hash_elasticity);
1252                 if (err)
1253                         goto rollback;
1254         }
1255
1256         br_multicast_open(br);
1257         list_for_each_entry(port, &br->port_list, list) {
1258                 if (port->state == BR_STATE_DISABLED ||
1259                     port->state == BR_STATE_BLOCKING)
1260                         continue;
1261
1262                 __br_multicast_enable_port(port);
1263         }
1264
1265 unlock:
1266         spin_unlock(&br->multicast_lock);
1267
1268         return err;
1269 }
1270
1271 int br_multicast_set_hash_max(struct net_bridge *br, unsigned long val)
1272 {
1273         int err = -ENOENT;
1274         u32 old;
1275
1276         spin_lock(&br->multicast_lock);
1277         if (!netif_running(br->dev))
1278                 goto unlock;
1279
1280         err = -EINVAL;
1281         if (!is_power_of_2(val))
1282                 goto unlock;
1283         if (br->mdb && val < br->mdb->size)
1284                 goto unlock;
1285
1286         err = 0;
1287
1288         old = br->hash_max;
1289         br->hash_max = val;
1290
1291         if (br->mdb) {
1292                 if (br->mdb->old) {
1293                         err = -EEXIST;
1294 rollback:
1295                         br->hash_max = old;
1296                         goto unlock;
1297                 }
1298
1299                 err = br_mdb_rehash(&br->mdb, br->hash_max,
1300                                     br->hash_elasticity);
1301                 if (err)
1302                         goto rollback;
1303         }
1304
1305 unlock:
1306         spin_unlock(&br->multicast_lock);
1307
1308         return err;
1309 }