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_equal(const struct br_ip *a, const struct br_ip *b)
  31 {
  32         if (a->proto != b->proto)
  33                 return 0;
  34         switch (a->proto) {
  35         case htons(ETH_P_IP):
  36                 return a->u.ip4 == b->u.ip4;
  37         }
  38         return 0;
  39 }
  40
  41 static inline int __br_ip4_hash(struct net_bridge_mdb_htable *mdb, __be32 ip)
  42 {
  43         return jhash_1word(mdb->secret, (__force u32)ip) & (mdb->max - 1);
  44 }
  45
  46 static inline int br_ip_hash(struct net_bridge_mdb_htable *mdb,
  47                              struct br_ip *ip)
  48 {
  49         switch (ip->proto) {
  50         case htons(ETH_P_IP):
  51                 return __br_ip4_hash(mdb, ip->u.ip4);
  52         }
  53         return 0;
  54 }
  55
  56 static struct net_bridge_mdb_entry *__br_mdb_ip_get(
  57         struct net_bridge_mdb_htable *mdb, struct br_ip *dst, int hash)
  58 {
  59         struct net_bridge_mdb_entry *mp;
  60         struct hlist_node *p;
  61
  62         hlist_for_each_entry_rcu(mp, p, &mdb->mhash[hash], hlist[mdb->ver]) {
  63                 if (br_ip_equal(&mp->addr, dst))
  64                         return mp;
  65         }
  66
  67         return NULL;
  68 }
  69
  70 static struct net_bridge_mdb_entry *br_mdb_ip4_get(
  71         struct net_bridge_mdb_htable *mdb, __be32 dst)
  72 {
  73         struct br_ip br_dst;
  74
  75         br_dst.u.ip4 = dst;
  76         br_dst.proto = htons(ETH_P_IP);
  77
  78         return __br_mdb_ip_get(mdb, &br_dst, __br_ip4_hash(mdb, dst));
  79 }
  80
  81 static struct net_bridge_mdb_entry *br_mdb_ip_get(
  82         struct net_bridge_mdb_htable *mdb, struct br_ip *dst)
  83 {
  84         return __br_mdb_ip_get(mdb, dst, br_ip_hash(mdb, dst));
  85 }
  86
  87 struct net_bridge_mdb_entry *br_mdb_get(struct net_bridge *br,
  88                                         struct sk_buff *skb)
  89 {
  90         struct net_bridge_mdb_htable *mdb = br->mdb;
  91         struct br_ip ip;
  92
  93         if (!mdb || br->multicast_disabled)
  94                 return NULL;
  95
  96         if (BR_INPUT_SKB_CB(skb)->igmp)
  97                 return NULL;
  98
  99         ip.proto = skb->protocol;
 100
 101         switch (skb->protocol) {
 102         case htons(ETH_P_IP):
 103                 ip.u.ip4 = ip_hdr(skb)->daddr;
 104                 break;
 105         default:
 106                 return NULL;
 107         }
 108
 109         return br_mdb_ip_get(mdb, &ip);
 110 }
 111
 112 static void br_mdb_free(struct rcu_head *head)
 113 {
 114         struct net_bridge_mdb_htable *mdb =
 115                 container_of(head, struct net_bridge_mdb_htable, rcu);
 116         struct net_bridge_mdb_htable *old = mdb->old;
 117
 118         mdb->old = NULL;
 119         kfree(old->mhash);
 120         kfree(old);
 121 }
 122
 123 static int br_mdb_copy(struct net_bridge_mdb_htable *new,
 124                        struct net_bridge_mdb_htable *old,
 125                        int elasticity)
 126 {
 127         struct net_bridge_mdb_entry *mp;
 128         struct hlist_node *p;
 129         int maxlen;
 130         int len;
 131         int i;
 132
 133         for (i = 0; i < old->max; i++)
 134                 hlist_for_each_entry(mp, p, &old->mhash[i], hlist[old->ver])
 135                         hlist_add_head(&mp->hlist[new->ver],
 136                                        &new->mhash[br_ip_hash(new, &mp->addr)]);
 137
 138         if (!elasticity)
 139                 return 0;
 140
 141         maxlen = 0;
 142         for (i = 0; i < new->max; i++) {
 143                 len = 0;
 144                 hlist_for_each_entry(mp, p, &new->mhash[i], hlist[new->ver])
 145                         len++;
 146                 if (len > maxlen)
 147                         maxlen = len;
 148         }
 149
 150         return maxlen > elasticity ? -EINVAL : 0;
 151 }
 152
 153 static void br_multicast_free_pg(struct rcu_head *head)
 154 {
 155         struct net_bridge_port_group *p =
 156                 container_of(head, struct net_bridge_port_group, rcu);
 157
 158         kfree(p);
 159 }
 160
 161 static void br_multicast_free_group(struct rcu_head *head)
 162 {
 163         struct net_bridge_mdb_entry *mp =
 164                 container_of(head, struct net_bridge_mdb_entry, rcu);
 165
 166         kfree(mp);
 167 }
 168
 169 static void br_multicast_group_expired(unsigned long data)
 170 {
 171         struct net_bridge_mdb_entry *mp = (void *)data;
 172         struct net_bridge *br = mp->br;
 173         struct net_bridge_mdb_htable *mdb;
 174
 175         spin_lock(&br->multicast_lock);
 176         if (!netif_running(br->dev) || timer_pending(&mp->timer))
 177                 goto out;
 178
 179         if (!hlist_unhashed(&mp->mglist))
 180                 hlist_del_init(&mp->mglist);
 181
 182         if (mp->ports)
 183                 goto out;
 184
 185         mdb = br->mdb;
 186         hlist_del_rcu(&mp->hlist[mdb->ver]);
 187         mdb->size--;
 188
 189         del_timer(&mp->query_timer);
 190         call_rcu_bh(&mp->rcu, br_multicast_free_group);
 191
 192 out:
 193         spin_unlock(&br->multicast_lock);
 194 }
 195
 196 static void br_multicast_del_pg(struct net_bridge *br,
 197                                 struct net_bridge_port_group *pg)
 198 {
 199         struct net_bridge_mdb_htable *mdb = br->mdb;
 200         struct net_bridge_mdb_entry *mp;
 201         struct net_bridge_port_group *p;
 202         struct net_bridge_port_group **pp;
 203
 204         mp = br_mdb_ip_get(mdb, &pg->addr);
 205         if (WARN_ON(!mp))
 206                 return;
 207
 208         for (pp = &mp->ports; (p = *pp); pp = &p->next) {
 209                 if (p != pg)
 210                         continue;
 211
 212                 *pp = p->next;
 213                 hlist_del_init(&p->mglist);
 214                 del_timer(&p->timer);
 215                 del_timer(&p->query_timer);
 216                 call_rcu_bh(&p->rcu, br_multicast_free_pg);
 217
 218                 if (!mp->ports && hlist_unhashed(&mp->mglist) &&
 219                     netif_running(br->dev))
 220                         mod_timer(&mp->timer, jiffies);
 221
 222                 return;
 223         }
 224
 225         WARN_ON(1);
 226 }
 227
 228 static void br_multicast_port_group_expired(unsigned long data)
 229 {
 230         struct net_bridge_port_group *pg = (void *)data;
 231         struct net_bridge *br = pg->port->br;
 232
 233         spin_lock(&br->multicast_lock);
 234         if (!netif_running(br->dev) || timer_pending(&pg->timer) ||
 235             hlist_unhashed(&pg->mglist))
 236                 goto out;
 237
 238         br_multicast_del_pg(br, pg);
 239
 240 out:
 241         spin_unlock(&br->multicast_lock);
 242 }
 243
 244 static int br_mdb_rehash(struct net_bridge_mdb_htable **mdbp, int max,
 245                          int elasticity)
 246 {
 247         struct net_bridge_mdb_htable *old = *mdbp;
 248         struct net_bridge_mdb_htable *mdb;
 249         int err;
 250
 251         mdb = kmalloc(sizeof(*mdb), GFP_ATOMIC);
 252         if (!mdb)
 253                 return -ENOMEM;
 254
 255         mdb->max = max;
 256         mdb->old = old;
 257
 258         mdb->mhash = kzalloc(max * sizeof(*mdb->mhash), GFP_ATOMIC);
 259         if (!mdb->mhash) {
 260                 kfree(mdb);
 261                 return -ENOMEM;
 262         }
 263
 264         mdb->size = old ? old->size : 0;
 265         mdb->ver = old ? old->ver ^ 1 : 0;
 266
 267         if (!old || elasticity)
 268                 get_random_bytes(&mdb->secret, sizeof(mdb->secret));
 269         else
 270                 mdb->secret = old->secret;
 271
 272         if (!old)
 273                 goto out;
 274
 275         err = br_mdb_copy(mdb, old, elasticity);
 276         if (err) {
 277                 kfree(mdb->mhash);
 278                 kfree(mdb);
 279                 return err;
 280         }
 281
 282         call_rcu_bh(&mdb->rcu, br_mdb_free);
 283
 284 out:
 285         rcu_assign_pointer(*mdbp, mdb);
 286
 287         return 0;
 288 }
 289
 290 static struct sk_buff *br_ip4_multicast_alloc_query(struct net_bridge *br,
 291                                                     __be32 group)
 292 {
 293         struct sk_buff *skb;
 294         struct igmphdr *ih;
 295         struct ethhdr *eth;
 296         struct iphdr *iph;
 297
 298         skb = netdev_alloc_skb_ip_align(br->dev, sizeof(*eth) + sizeof(*iph) +
 299                                                  sizeof(*ih) + 4);
 300         if (!skb)
 301                 goto out;
 302
 303         skb->protocol = htons(ETH_P_IP);
 304
 305         skb_reset_mac_header(skb);
 306         eth = eth_hdr(skb);
 307
 308         memcpy(eth->h_source, br->dev->dev_addr, 6);
 309         eth->h_dest[0] = 1;
 310         eth->h_dest[1] = 0;
 311         eth->h_dest[2] = 0x5e;
 312         eth->h_dest[3] = 0;
 313         eth->h_dest[4] = 0;
 314         eth->h_dest[5] = 1;
 315         eth->h_proto = htons(ETH_P_IP);
 316         skb_put(skb, sizeof(*eth));
 317
 318         skb_set_network_header(skb, skb->len);
 319         iph = ip_hdr(skb);
 320
 321         iph->version = 4;
 322         iph->ihl = 6;
 323         iph->tos = 0xc0;
 324         iph->tot_len = htons(sizeof(*iph) + sizeof(*ih) + 4);
 325         iph->id = 0;
 326         iph->frag_off = htons(IP_DF);
 327         iph->ttl = 1;
 328         iph->protocol = IPPROTO_IGMP;
 329         iph->saddr = 0;
 330         iph->daddr = htonl(INADDR_ALLHOSTS_GROUP);
 331         ((u8 *)&iph[1])[0] = IPOPT_RA;
 332         ((u8 *)&iph[1])[1] = 4;
 333         ((u8 *)&iph[1])[2] = 0;
 334         ((u8 *)&iph[1])[3] = 0;
 335         ip_send_check(iph);
 336         skb_put(skb, 24);
 337
 338         skb_set_transport_header(skb, skb->len);
 339         ih = igmp_hdr(skb);
 340         ih->type = IGMP_HOST_MEMBERSHIP_QUERY;
 341         ih->code = (group ? br->multicast_last_member_interval :
 342                             br->multicast_query_response_interval) /
 343                    (HZ / IGMP_TIMER_SCALE);
 344         ih->group = group;
 345         ih->csum = 0;
 346         ih->csum = ip_compute_csum((void *)ih, sizeof(struct igmphdr));
 347         skb_put(skb, sizeof(*ih));
 348
 349         __skb_pull(skb, sizeof(*eth));
 350
 351 out:
 352         return skb;
 353 }
 354
 355 static struct sk_buff *br_multicast_alloc_query(struct net_bridge *br,
 356                                                 struct br_ip *addr)
 357 {
 358         switch (addr->proto) {
 359         case htons(ETH_P_IP):
 360                 return br_ip4_multicast_alloc_query(br, addr->u.ip4);
 361         }
 362         return NULL;
 363 }
 364
 365 static void br_multicast_send_group_query(struct net_bridge_mdb_entry *mp)
 366 {
 367         struct net_bridge *br = mp->br;
 368         struct sk_buff *skb;
 369
 370         skb = br_multicast_alloc_query(br, &mp->addr);
 371         if (!skb)
 372                 goto timer;
 373
 374         netif_rx(skb);
 375
 376 timer:
 377         if (++mp->queries_sent < br->multicast_last_member_count)
 378                 mod_timer(&mp->query_timer,
 379                           jiffies + br->multicast_last_member_interval);
 380 }
 381
 382 static void br_multicast_group_query_expired(unsigned long data)
 383 {
 384         struct net_bridge_mdb_entry *mp = (void *)data;
 385         struct net_bridge *br = mp->br;
 386
 387         spin_lock(&br->multicast_lock);
 388         if (!netif_running(br->dev) || hlist_unhashed(&mp->mglist) ||
 389             mp->queries_sent >= br->multicast_last_member_count)
 390                 goto out;
 391
 392         br_multicast_send_group_query(mp);
 393
 394 out:
 395         spin_unlock(&br->multicast_lock);
 396 }
 397
 398 static void br_multicast_send_port_group_query(struct net_bridge_port_group *pg)
 399 {
 400         struct net_bridge_port *port = pg->port;
 401         struct net_bridge *br = port->br;
 402         struct sk_buff *skb;
 403
 404         skb = br_multicast_alloc_query(br, &pg->addr);
 405         if (!skb)
 406                 goto timer;
 407
 408         br_deliver(port, skb);
 409
 410 timer:
 411         if (++pg->queries_sent < br->multicast_last_member_count)
 412                 mod_timer(&pg->query_timer,
 413                           jiffies + br->multicast_last_member_interval);
 414 }
 415
 416 static void br_multicast_port_group_query_expired(unsigned long data)
 417 {
 418         struct net_bridge_port_group *pg = (void *)data;
 419         struct net_bridge_port *port = pg->port;
 420         struct net_bridge *br = port->br;
 421
 422         spin_lock(&br->multicast_lock);
 423         if (!netif_running(br->dev) || hlist_unhashed(&pg->mglist) ||
 424             pg->queries_sent >= br->multicast_last_member_count)
 425                 goto out;
 426
 427         br_multicast_send_port_group_query(pg);
 428
 429 out:
 430         spin_unlock(&br->multicast_lock);
 431 }
 432
 433 static struct net_bridge_mdb_entry *br_multicast_get_group(
 434         struct net_bridge *br, struct net_bridge_port *port,
 435         struct br_ip *group, int hash)
 436 {
 437         struct net_bridge_mdb_htable *mdb = br->mdb;
 438         struct net_bridge_mdb_entry *mp;
 439         struct hlist_node *p;
 440         unsigned count = 0;
 441         unsigned max;
 442         int elasticity;
 443         int err;
 444
 445         hlist_for_each_entry(mp, p, &mdb->mhash[hash], hlist[mdb->ver]) {
 446                 count++;
 447                 if (unlikely(br_ip_equal(group, &mp->addr)))
 448                         return mp;
 449         }
 450
 451         elasticity = 0;
 452         max = mdb->max;
 453
 454         if (unlikely(count > br->hash_elasticity && count)) {
 455                 if (net_ratelimit())
 456                         printk(KERN_INFO "%s: Multicast hash table "
 457                                "chain limit reached: %s\n",
 458                                br->dev->name, port ? port->dev->name :
 459                                                      br->dev->name);
 460
 461                 elasticity = br->hash_elasticity;
 462         }
 463
 464         if (mdb->size >= max) {
 465                 max *= 2;
 466                 if (unlikely(max >= br->hash_max)) {
 467                         printk(KERN_WARNING "%s: Multicast hash table maximum "
 468                                "reached, disabling snooping: %s, %d\n",
 469                                br->dev->name, port ? port->dev->name :
 470                                                      br->dev->name,
 471                                max);
 472                         err = -E2BIG;
 473 disable:
 474                         br->multicast_disabled = 1;
 475                         goto err;
 476                 }
 477         }
 478
 479         if (max > mdb->max || elasticity) {
 480                 if (mdb->old) {
 481                         if (net_ratelimit())
 482                                 printk(KERN_INFO "%s: Multicast hash table "
 483                                        "on fire: %s\n",
 484                                        br->dev->name, port ? port->dev->name :
 485                                                              br->dev->name);
 486                         err = -EEXIST;
 487                         goto err;
 488                 }
 489
 490                 err = br_mdb_rehash(&br->mdb, max, elasticity);
 491                 if (err) {
 492                         printk(KERN_WARNING "%s: Cannot rehash multicast "
 493                                "hash table, disabling snooping: "
 494                                "%s, %d, %d\n",
 495                                br->dev->name, port ? port->dev->name :
 496                                                      br->dev->name,
 497                                mdb->size, err);
 498                         goto disable;
 499                 }
 500
 501                 err = -EAGAIN;
 502                 goto err;
 503         }
 504
 505         return NULL;
 506
 507 err:
 508         mp = ERR_PTR(err);
 509         return mp;
 510 }
 511
 512 static struct net_bridge_mdb_entry *br_multicast_new_group(
 513         struct net_bridge *br, struct net_bridge_port *port,
 514         struct br_ip *group)
 515 {
 516         struct net_bridge_mdb_htable *mdb = br->mdb;
 517         struct net_bridge_mdb_entry *mp;
 518         int hash;
 519
 520         if (!mdb) {
 521                 if (br_mdb_rehash(&br->mdb, BR_HASH_SIZE, 0))
 522                         return NULL;
 523                 goto rehash;
 524         }
 525
 526         hash = br_ip_hash(mdb, group);
 527         mp = br_multicast_get_group(br, port, group, hash);
 528         switch (PTR_ERR(mp)) {
 529         case 0:
 530                 break;
 531
 532         case -EAGAIN:
 533 rehash:
 534                 mdb = br->mdb;
 535                 hash = br_ip_hash(mdb, group);
 536                 break;
 537
 538         default:
 539                 goto out;
 540         }
 541
 542         mp = kzalloc(sizeof(*mp), GFP_ATOMIC);
 543         if (unlikely(!mp))
 544                 goto out;
 545
 546         mp->br = br;
 547         mp->addr = *group;
 548         setup_timer(&mp->timer, br_multicast_group_expired,
 549                     (unsigned long)mp);
 550         setup_timer(&mp->query_timer, br_multicast_group_query_expired,
 551                     (unsigned long)mp);
 552
 553         hlist_add_head_rcu(&mp->hlist[mdb->ver], &mdb->mhash[hash]);
 554         mdb->size++;
 555
 556 out:
 557         return mp;
 558 }
 559
 560 static int br_multicast_add_group(struct net_bridge *br,
 561                                   struct net_bridge_port *port,
 562                                   struct br_ip *group)
 563 {
 564         struct net_bridge_mdb_entry *mp;
 565         struct net_bridge_port_group *p;
 566         struct net_bridge_port_group **pp;
 567         unsigned long now = jiffies;
 568         int err;
 569
 570         spin_lock(&br->multicast_lock);
 571         if (!netif_running(br->dev) ||
 572             (port && port->state == BR_STATE_DISABLED))
 573                 goto out;
 574
 575         mp = br_multicast_new_group(br, port, group);
 576         err = PTR_ERR(mp);
 577         if (unlikely(IS_ERR(mp) || !mp))
 578                 goto err;
 579
 580         if (!port) {
 581                 hlist_add_head(&mp->mglist, &br->mglist);
 582                 mod_timer(&mp->timer, now + br->multicast_membership_interval);
 583                 goto out;
 584         }
 585
 586         for (pp = &mp->ports; (p = *pp); pp = &p->next) {
 587                 if (p->port == port)
 588                         goto found;
 589                 if ((unsigned long)p->port < (unsigned long)port)
 590                         break;
 591         }
 592
 593         p = kzalloc(sizeof(*p), GFP_ATOMIC);
 594         err = -ENOMEM;
 595         if (unlikely(!p))
 596                 goto err;
 597
 598         p->addr = *group;
 599         p->port = port;
 600         p->next = *pp;
 601         hlist_add_head(&p->mglist, &port->mglist);
 602         setup_timer(&p->timer, br_multicast_port_group_expired,
 603                     (unsigned long)p);
 604         setup_timer(&p->query_timer, br_multicast_port_group_query_expired,
 605                     (unsigned long)p);
 606
 607         rcu_assign_pointer(*pp, p);
 608
 609 found:
 610         mod_timer(&p->timer, now + br->multicast_membership_interval);
 611 out:
 612         err = 0;
 613
 614 err:
 615         spin_unlock(&br->multicast_lock);
 616         return err;
 617 }
 618
 619 static int br_ip4_multicast_add_group(struct net_bridge *br,
 620                                       struct net_bridge_port *port,
 621                                       __be32 group)
 622 {
 623         struct br_ip br_group;
 624
 625         if (ipv4_is_local_multicast(group))
 626                 return 0;
 627
 628         br_group.u.ip4 = group;
 629         br_group.proto = htons(ETH_P_IP);
 630
 631         return br_multicast_add_group(br, port, &br_group);
 632 }
 633
 634 static void br_multicast_router_expired(unsigned long data)
 635 {
 636         struct net_bridge_port *port = (void *)data;
 637         struct net_bridge *br = port->br;
 638
 639         spin_lock(&br->multicast_lock);
 640         if (port->multicast_router != 1 ||
 641             timer_pending(&port->multicast_router_timer) ||
 642             hlist_unhashed(&port->rlist))
 643                 goto out;
 644
 645         hlist_del_init_rcu(&port->rlist);
 646
 647 out:
 648         spin_unlock(&br->multicast_lock);
 649 }
 650
 651 static void br_multicast_local_router_expired(unsigned long data)
 652 {
 653 }
 654
 655 static void __br_multicast_send_query(struct net_bridge *br,
 656                                       struct net_bridge_port *port,
 657                                       struct br_ip *ip)
 658 {
 659         struct sk_buff *skb;
 660
 661         skb = br_multicast_alloc_query(br, ip);
 662         if (!skb)
 663                 return;
 664
 665         if (port) {
 666                 __skb_push(skb, sizeof(struct ethhdr));
 667                 skb->dev = port->dev;
 668                 NF_HOOK(PF_BRIDGE, NF_BR_LOCAL_OUT, skb, NULL, skb->dev,
 669                         dev_queue_xmit);
 670         } else
 671                 netif_rx(skb);
 672 }
 673
 674 static void br_multicast_send_query(struct net_bridge *br,
 675                                     struct net_bridge_port *port, u32 sent)
 676 {
 677         unsigned long time;
 678         struct br_ip br_group;
 679
 680         if (!netif_running(br->dev) || br->multicast_disabled ||
 681             timer_pending(&br->multicast_querier_timer))
 682                 return;
 683
 684         br_group.u.ip4 = 0;
 685         br_group.proto = htons(ETH_P_IP);
 686
 687         __br_multicast_send_query(br, port, &br_group);
 688
 689         time = jiffies;
 690         time += sent < br->multicast_startup_query_count ?
 691                 br->multicast_startup_query_interval :
 692                 br->multicast_query_interval;
 693         mod_timer(port ? &port->multicast_query_timer :
 694                          &br->multicast_query_timer, time);
 695 }
 696
 697 static void br_multicast_port_query_expired(unsigned long data)
 698 {
 699         struct net_bridge_port *port = (void *)data;
 700         struct net_bridge *br = port->br;
 701
 702         spin_lock(&br->multicast_lock);
 703         if (port->state == BR_STATE_DISABLED ||
 704             port->state == BR_STATE_BLOCKING)
 705                 goto out;
 706
 707         if (port->multicast_startup_queries_sent <
 708             br->multicast_startup_query_count)
 709                 port->multicast_startup_queries_sent++;
 710
 711         br_multicast_send_query(port->br, port,
 712                                 port->multicast_startup_queries_sent);
 713
 714 out:
 715         spin_unlock(&br->multicast_lock);
 716 }
 717
 718 void br_multicast_add_port(struct net_bridge_port *port)
 719 {
 720         port->multicast_router = 1;
 721
 722         setup_timer(&port->multicast_router_timer, br_multicast_router_expired,
 723                     (unsigned long)port);
 724         setup_timer(&port->multicast_query_timer,
 725                     br_multicast_port_query_expired, (unsigned long)port);
 726 }
 727
 728 void br_multicast_del_port(struct net_bridge_port *port)
 729 {
 730         del_timer_sync(&port->multicast_router_timer);
 731 }
 732
 733 static void __br_multicast_enable_port(struct net_bridge_port *port)
 734 {
 735         port->multicast_startup_queries_sent = 0;
 736
 737         if (try_to_del_timer_sync(&port->multicast_query_timer) >= 0 ||
 738             del_timer(&port->multicast_query_timer))
 739                 mod_timer(&port->multicast_query_timer, jiffies);
 740 }
 741
 742 void br_multicast_enable_port(struct net_bridge_port *port)
 743 {
 744         struct net_bridge *br = port->br;
 745
 746         spin_lock(&br->multicast_lock);
 747         if (br->multicast_disabled || !netif_running(br->dev))
 748                 goto out;
 749
 750         __br_multicast_enable_port(port);
 751
 752 out:
 753         spin_unlock(&br->multicast_lock);
 754 }
 755
 756 void br_multicast_disable_port(struct net_bridge_port *port)
 757 {
 758         struct net_bridge *br = port->br;
 759         struct net_bridge_port_group *pg;
 760         struct hlist_node *p, *n;
 761
 762         spin_lock(&br->multicast_lock);
 763         hlist_for_each_entry_safe(pg, p, n, &port->mglist, mglist)
 764                 br_multicast_del_pg(br, pg);
 765
 766         if (!hlist_unhashed(&port->rlist))
 767                 hlist_del_init_rcu(&port->rlist);
 768         del_timer(&port->multicast_router_timer);
 769         del_timer(&port->multicast_query_timer);
 770         spin_unlock(&br->multicast_lock);
 771 }
 772
 773 static int br_ip4_multicast_igmp3_report(struct net_bridge *br,
 774                                          struct net_bridge_port *port,
 775                                          struct sk_buff *skb)
 776 {
 777         struct igmpv3_report *ih;
 778         struct igmpv3_grec *grec;
 779         int i;
 780         int len;
 781         int num;
 782         int type;
 783         int err = 0;
 784         __be32 group;
 785
 786         if (!pskb_may_pull(skb, sizeof(*ih)))
 787                 return -EINVAL;
 788
 789         ih = igmpv3_report_hdr(skb);
 790         num = ntohs(ih->ngrec);
 791         len = sizeof(*ih);
 792
 793         for (i = 0; i < num; i++) {
 794                 len += sizeof(*grec);
 795                 if (!pskb_may_pull(skb, len))
 796                         return -EINVAL;
 797
 798                 grec = (void *)(skb->data + len - sizeof(*grec));
 799                 group = grec->grec_mca;
 800                 type = grec->grec_type;
 801
 802                 len += ntohs(grec->grec_nsrcs) * 4;
 803                 if (!pskb_may_pull(skb, len))
 804                         return -EINVAL;
 805
 806                 /* We treat this as an IGMPv2 report for now. */
 807                 switch (type) {
 808                 case IGMPV3_MODE_IS_INCLUDE:
 809                 case IGMPV3_MODE_IS_EXCLUDE:
 810                 case IGMPV3_CHANGE_TO_INCLUDE:
 811                 case IGMPV3_CHANGE_TO_EXCLUDE:
 812                 case IGMPV3_ALLOW_NEW_SOURCES:
 813                 case IGMPV3_BLOCK_OLD_SOURCES:
 814                         break;
 815
 816                 default:
 817                         continue;
 818                 }
 819
 820                 err = br_ip4_multicast_add_group(br, port, group);
 821                 if (err)
 822                         break;
 823         }
 824
 825         return err;
 826 }
 827
 828 static void br_multicast_add_router(struct net_bridge *br,
 829                                     struct net_bridge_port *port)
 830 {
 831         struct hlist_node *p;
 832         struct hlist_node **h;
 833
 834         for (h = &br->router_list.first;
 835              (p = *h) &&
 836              (unsigned long)container_of(p, struct net_bridge_port, rlist) >
 837              (unsigned long)port;
 838              h = &p->next)
 839                 ;
 840
 841         port->rlist.pprev = h;
 842         port->rlist.next = p;
 843         rcu_assign_pointer(*h, &port->rlist);
 844         if (p)
 845                 p->pprev = &port->rlist.next;
 846 }
 847
 848 static void br_multicast_mark_router(struct net_bridge *br,
 849                                      struct net_bridge_port *port)
 850 {
 851         unsigned long now = jiffies;
 852
 853         if (!port) {
 854                 if (br->multicast_router == 1)
 855                         mod_timer(&br->multicast_router_timer,
 856                                   now + br->multicast_querier_interval);
 857                 return;
 858         }
 859
 860         if (port->multicast_router != 1)
 861                 return;
 862
 863         if (!hlist_unhashed(&port->rlist))
 864                 goto timer;
 865
 866         br_multicast_add_router(br, port);
 867
 868 timer:
 869         mod_timer(&port->multicast_router_timer,
 870                   now + br->multicast_querier_interval);
 871 }
 872
 873 static void br_multicast_query_received(struct net_bridge *br,
 874                                         struct net_bridge_port *port,
 875                                         int saddr)
 876 {
 877         if (saddr)
 878                 mod_timer(&br->multicast_querier_timer,
 879                           jiffies + br->multicast_querier_interval);
 880         else if (timer_pending(&br->multicast_querier_timer))
 881                 return;
 882
 883         br_multicast_mark_router(br, port);
 884 }
 885
 886 static int br_ip4_multicast_query(struct net_bridge *br,
 887                                   struct net_bridge_port *port,
 888                                   struct sk_buff *skb)
 889 {
 890         struct iphdr *iph = ip_hdr(skb);
 891         struct igmphdr *ih = igmp_hdr(skb);
 892         struct net_bridge_mdb_entry *mp;
 893         struct igmpv3_query *ih3;
 894         struct net_bridge_port_group *p;
 895         struct net_bridge_port_group **pp;
 896         unsigned long max_delay;
 897         unsigned long now = jiffies;
 898         __be32 group;
 899         int err = 0;
 900
 901         spin_lock(&br->multicast_lock);
 902         if (!netif_running(br->dev) ||
 903             (port && port->state == BR_STATE_DISABLED))
 904                 goto out;
 905
 906         br_multicast_query_received(br, port, !!iph->saddr);
 907
 908         group = ih->group;
 909
 910         if (skb->len == sizeof(*ih)) {
 911                 max_delay = ih->code * (HZ / IGMP_TIMER_SCALE);
 912
 913                 if (!max_delay) {
 914                         max_delay = 10 * HZ;
 915                         group = 0;
 916                 }
 917         } else {
 918                 if (!pskb_may_pull(skb, sizeof(struct igmpv3_query))) {
 919                         err = -EINVAL;
 920                         goto out;
 921                 }
 922
 923                 ih3 = igmpv3_query_hdr(skb);
 924                 if (ih3->nsrcs)
 925                         goto out;
 926
 927                 max_delay = ih3->code ?
 928                             IGMPV3_MRC(ih3->code) * (HZ / IGMP_TIMER_SCALE) : 1;
 929         }
 930
 931         if (!group)
 932                 goto out;
 933
 934         mp = br_mdb_ip4_get(br->mdb, group);
 935         if (!mp)
 936                 goto out;
 937
 938         max_delay *= br->multicast_last_member_count;
 939
 940         if (!hlist_unhashed(&mp->mglist) &&
 941             (timer_pending(&mp->timer) ?
 942              time_after(mp->timer.expires, now + max_delay) :
 943              try_to_del_timer_sync(&mp->timer) >= 0))
 944                 mod_timer(&mp->timer, now + max_delay);
 945
 946         for (pp = &mp->ports; (p = *pp); pp = &p->next) {
 947                 if (timer_pending(&p->timer) ?
 948                     time_after(p->timer.expires, now + max_delay) :
 949                     try_to_del_timer_sync(&p->timer) >= 0)
 950                         mod_timer(&mp->timer, now + max_delay);
 951         }
 952
 953 out:
 954         spin_unlock(&br->multicast_lock);
 955         return err;
 956 }
 957
 958 static void br_multicast_leave_group(struct net_bridge *br,
 959                                      struct net_bridge_port *port,
 960                                      struct br_ip *group)
 961 {
 962         struct net_bridge_mdb_htable *mdb;
 963         struct net_bridge_mdb_entry *mp;
 964         struct net_bridge_port_group *p;
 965         unsigned long now;
 966         unsigned long time;
 967
 968         spin_lock(&br->multicast_lock);
 969         if (!netif_running(br->dev) ||
 970             (port && port->state == BR_STATE_DISABLED) ||
 971             timer_pending(&br->multicast_querier_timer))
 972                 goto out;
 973
 974         mdb = br->mdb;
 975         mp = br_mdb_ip_get(mdb, group);
 976         if (!mp)
 977                 goto out;
 978
 979         now = jiffies;
 980         time = now + br->multicast_last_member_count *
 981                      br->multicast_last_member_interval;
 982
 983         if (!port) {
 984                 if (!hlist_unhashed(&mp->mglist) &&
 985                     (timer_pending(&mp->timer) ?
 986                      time_after(mp->timer.expires, time) :
 987                      try_to_del_timer_sync(&mp->timer) >= 0)) {
 988                         mod_timer(&mp->timer, time);
 989
 990                         mp->queries_sent = 0;
 991                         mod_timer(&mp->query_timer, now);
 992                 }
 993
 994                 goto out;
 995         }
 996
 997         for (p = mp->ports; p; p = p->next) {
 998                 if (p->port != port)
 999                         continue;
1000
1001                 if (!hlist_unhashed(&p->mglist) &&
1002                     (timer_pending(&p->timer) ?
1003                      time_after(p->timer.expires, time) :
1004                      try_to_del_timer_sync(&p->timer) >= 0)) {
1005                         mod_timer(&p->timer, time);
1006
1007                         p->queries_sent = 0;
1008                         mod_timer(&p->query_timer, now);
1009                 }
1010
1011                 break;
1012         }
1013
1014 out:
1015         spin_unlock(&br->multicast_lock);
1016 }
1017
1018 static void br_ip4_multicast_leave_group(struct net_bridge *br,
1019                                          struct net_bridge_port *port,
1020                                          __be32 group)
1021 {
1022         struct br_ip br_group;
1023
1024         if (ipv4_is_local_multicast(group))
1025                 return;
1026
1027         br_group.u.ip4 = group;
1028         br_group.proto = htons(ETH_P_IP);
1029
1030         br_multicast_leave_group(br, port, &br_group);
1031 }
1032
1033
1034 static int br_multicast_ipv4_rcv(struct net_bridge *br,
1035                                  struct net_bridge_port *port,
1036                                  struct sk_buff *skb)
1037 {
1038         struct sk_buff *skb2 = skb;
1039         struct iphdr *iph;
1040         struct igmphdr *ih;
1041         unsigned len;
1042         unsigned offset;
1043         int err;
1044
1045         BR_INPUT_SKB_CB(skb)->igmp = 0;
1046         BR_INPUT_SKB_CB(skb)->mrouters_only = 0;
1047
1048         /* We treat OOM as packet loss for now. */
1049         if (!pskb_may_pull(skb, sizeof(*iph)))
1050                 return -EINVAL;
1051
1052         iph = ip_hdr(skb);
1053
1054         if (iph->ihl < 5 || iph->version != 4)
1055                 return -EINVAL;
1056
1057         if (!pskb_may_pull(skb, ip_hdrlen(skb)))
1058                 return -EINVAL;
1059
1060         iph = ip_hdr(skb);
1061
1062         if (unlikely(ip_fast_csum((u8 *)iph, iph->ihl)))
1063                 return -EINVAL;
1064
1065         if (iph->protocol != IPPROTO_IGMP)
1066                 return 0;
1067
1068         len = ntohs(iph->tot_len);
1069         if (skb->len < len || len < ip_hdrlen(skb))
1070                 return -EINVAL;
1071
1072         if (skb->len > len) {
1073                 skb2 = skb_clone(skb, GFP_ATOMIC);
1074                 if (!skb2)
1075                         return -ENOMEM;
1076
1077                 err = pskb_trim_rcsum(skb2, len);
1078                 if (err)
1079                         goto err_out;
1080         }
1081
1082         len -= ip_hdrlen(skb2);
1083         offset = skb_network_offset(skb2) + ip_hdrlen(skb2);
1084         __skb_pull(skb2, offset);
1085         skb_reset_transport_header(skb2);
1086
1087         err = -EINVAL;
1088         if (!pskb_may_pull(skb2, sizeof(*ih)))
1089                 goto out;
1090
1091         switch (skb2->ip_summed) {
1092         case CHECKSUM_COMPLETE:
1093                 if (!csum_fold(skb2->csum))
1094                         break;
1095                 /* fall through */
1096         case CHECKSUM_NONE:
1097                 skb2->csum = 0;
1098                 if (skb_checksum_complete(skb2))
1099                         goto out;
1100         }
1101
1102         err = 0;
1103
1104         BR_INPUT_SKB_CB(skb)->igmp = 1;
1105         ih = igmp_hdr(skb2);
1106
1107         switch (ih->type) {
1108         case IGMP_HOST_MEMBERSHIP_REPORT:
1109         case IGMPV2_HOST_MEMBERSHIP_REPORT:
1110                 BR_INPUT_SKB_CB(skb2)->mrouters_only = 1;
1111                 err = br_ip4_multicast_add_group(br, port, ih->group);
1112                 break;
1113         case IGMPV3_HOST_MEMBERSHIP_REPORT:
1114                 err = br_ip4_multicast_igmp3_report(br, port, skb2);
1115                 break;
1116         case IGMP_HOST_MEMBERSHIP_QUERY:
1117                 err = br_ip4_multicast_query(br, port, skb2);
1118                 break;
1119         case IGMP_HOST_LEAVE_MESSAGE:
1120                 br_ip4_multicast_leave_group(br, port, ih->group);
1121                 break;
1122         }
1123
1124 out:
1125         __skb_push(skb2, offset);
1126 err_out:
1127         if (skb2 != skb)
1128                 kfree_skb(skb2);
1129         return err;
1130 }
1131
1132 int br_multicast_rcv(struct net_bridge *br, struct net_bridge_port *port,
1133                      struct sk_buff *skb)
1134 {
1135         if (br->multicast_disabled)
1136                 return 0;
1137
1138         switch (skb->protocol) {
1139         case htons(ETH_P_IP):
1140                 return br_multicast_ipv4_rcv(br, port, skb);
1141         }
1142
1143         return 0;
1144 }
1145
1146 static void br_multicast_query_expired(unsigned long data)
1147 {
1148         struct net_bridge *br = (void *)data;
1149
1150         spin_lock(&br->multicast_lock);
1151         if (br->multicast_startup_queries_sent <
1152             br->multicast_startup_query_count)
1153                 br->multicast_startup_queries_sent++;
1154
1155         br_multicast_send_query(br, NULL, br->multicast_startup_queries_sent);
1156
1157         spin_unlock(&br->multicast_lock);
1158 }
1159
1160 void br_multicast_init(struct net_bridge *br)
1161 {
1162         br->hash_elasticity = 4;
1163         br->hash_max = 512;
1164
1165         br->multicast_router = 1;
1166         br->multicast_last_member_count = 2;
1167         br->multicast_startup_query_count = 2;
1168
1169         br->multicast_last_member_interval = HZ;
1170         br->multicast_query_response_interval = 10 * HZ;
1171         br->multicast_startup_query_interval = 125 * HZ / 4;
1172         br->multicast_query_interval = 125 * HZ;
1173         br->multicast_querier_interval = 255 * HZ;
1174         br->multicast_membership_interval = 260 * HZ;
1175
1176         spin_lock_init(&br->multicast_lock);
1177         setup_timer(&br->multicast_router_timer,
1178                     br_multicast_local_router_expired, 0);
1179         setup_timer(&br->multicast_querier_timer,
1180                     br_multicast_local_router_expired, 0);
1181         setup_timer(&br->multicast_query_timer, br_multicast_query_expired,
1182                     (unsigned long)br);
1183 }
1184
1185 void br_multicast_open(struct net_bridge *br)
1186 {
1187         br->multicast_startup_queries_sent = 0;
1188
1189         if (br->multicast_disabled)
1190                 return;
1191
1192         mod_timer(&br->multicast_query_timer, jiffies);
1193 }
1194
1195 void br_multicast_stop(struct net_bridge *br)
1196 {
1197         struct net_bridge_mdb_htable *mdb;
1198         struct net_bridge_mdb_entry *mp;
1199         struct hlist_node *p, *n;
1200         u32 ver;
1201         int i;
1202
1203         del_timer_sync(&br->multicast_router_timer);
1204         del_timer_sync(&br->multicast_querier_timer);
1205         del_timer_sync(&br->multicast_query_timer);
1206
1207         spin_lock_bh(&br->multicast_lock);
1208         mdb = br->mdb;
1209         if (!mdb)
1210                 goto out;
1211
1212         br->mdb = NULL;
1213
1214         ver = mdb->ver;
1215         for (i = 0; i < mdb->max; i++) {
1216                 hlist_for_each_entry_safe(mp, p, n, &mdb->mhash[i],
1217                                           hlist[ver]) {
1218                         del_timer(&mp->timer);
1219                         del_timer(&mp->query_timer);
1220                         call_rcu_bh(&mp->rcu, br_multicast_free_group);
1221                 }
1222         }
1223
1224         if (mdb->old) {
1225                 spin_unlock_bh(&br->multicast_lock);
1226                 rcu_barrier_bh();
1227                 spin_lock_bh(&br->multicast_lock);
1228                 WARN_ON(mdb->old);
1229         }
1230
1231         mdb->old = mdb;
1232         call_rcu_bh(&mdb->rcu, br_mdb_free);
1233
1234 out:
1235         spin_unlock_bh(&br->multicast_lock);
1236 }
1237
1238 int br_multicast_set_router(struct net_bridge *br, unsigned long val)
1239 {
1240         int err = -ENOENT;
1241
1242         spin_lock_bh(&br->multicast_lock);
1243         if (!netif_running(br->dev))
1244                 goto unlock;
1245
1246         switch (val) {
1247         case 0:
1248         case 2:
1249                 del_timer(&br->multicast_router_timer);
1250                 /* fall through */
1251         case 1:
1252                 br->multicast_router = val;
1253                 err = 0;
1254                 break;
1255
1256         default:
1257                 err = -EINVAL;
1258                 break;
1259         }
1260
1261 unlock:
1262         spin_unlock_bh(&br->multicast_lock);
1263
1264         return err;
1265 }
1266
1267 int br_multicast_set_port_router(struct net_bridge_port *p, unsigned long val)
1268 {
1269         struct net_bridge *br = p->br;
1270         int err = -ENOENT;
1271
1272         spin_lock(&br->multicast_lock);
1273         if (!netif_running(br->dev) || p->state == BR_STATE_DISABLED)
1274                 goto unlock;
1275
1276         switch (val) {
1277         case 0:
1278         case 1:
1279         case 2:
1280                 p->multicast_router = val;
1281                 err = 0;
1282
1283                 if (val < 2 && !hlist_unhashed(&p->rlist))
1284                         hlist_del_init_rcu(&p->rlist);
1285
1286                 if (val == 1)
1287                         break;
1288
1289                 del_timer(&p->multicast_router_timer);
1290
1291                 if (val == 0)
1292                         break;
1293
1294                 br_multicast_add_router(br, p);
1295                 break;
1296
1297         default:
1298                 err = -EINVAL;
1299                 break;
1300         }
1301
1302 unlock:
1303         spin_unlock(&br->multicast_lock);
1304
1305         return err;
1306 }
1307
1308 int br_multicast_toggle(struct net_bridge *br, unsigned long val)
1309 {
1310         struct net_bridge_port *port;
1311         int err = -ENOENT;
1312
1313         spin_lock(&br->multicast_lock);
1314         if (!netif_running(br->dev))
1315                 goto unlock;
1316
1317         err = 0;
1318         if (br->multicast_disabled == !val)
1319                 goto unlock;
1320
1321         br->multicast_disabled = !val;
1322         if (br->multicast_disabled)
1323                 goto unlock;
1324
1325         if (br->mdb) {
1326                 if (br->mdb->old) {
1327                         err = -EEXIST;
1328 rollback:
1329                         br->multicast_disabled = !!val;
1330                         goto unlock;
1331                 }
1332
1333                 err = br_mdb_rehash(&br->mdb, br->mdb->max,
1334                                     br->hash_elasticity);
1335                 if (err)
1336                         goto rollback;
1337         }
1338
1339         br_multicast_open(br);
1340         list_for_each_entry(port, &br->port_list, list) {
1341                 if (port->state == BR_STATE_DISABLED ||
1342                     port->state == BR_STATE_BLOCKING)
1343                         continue;
1344
1345                 __br_multicast_enable_port(port);
1346         }
1347
1348 unlock:
1349         spin_unlock(&br->multicast_lock);
1350
1351         return err;
1352 }
1353
1354 int br_multicast_set_hash_max(struct net_bridge *br, unsigned long val)
1355 {
1356         int err = -ENOENT;
1357         u32 old;
1358
1359         spin_lock(&br->multicast_lock);
1360         if (!netif_running(br->dev))
1361                 goto unlock;
1362
1363         err = -EINVAL;
1364         if (!is_power_of_2(val))
1365                 goto unlock;
1366         if (br->mdb && val < br->mdb->size)
1367                 goto unlock;
1368
1369         err = 0;
1370
1371         old = br->hash_max;
1372         br->hash_max = val;
1373
1374         if (br->mdb) {
1375                 if (br->mdb->old) {
1376                         err = -EEXIST;
1377 rollback:
1378                         br->hash_max = old;
1379                         goto unlock;
1380                 }
1381
1382                 err = br_mdb_rehash(&br->mdb, br->hash_max,
1383                                     br->hash_elasticity);
1384                 if (err)
1385                         goto rollback;
1386         }
1387
1388 unlock:
1389         spin_unlock(&br->multicast_lock);
1390
1391         return err;
1392 }