Pileus Git - ~andy/linux/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/log2.h>
  19 #include <linux/netdevice.h>
  20 #include <linux/netfilter_bridge.h>
  21 #include <linux/random.h>
  22 #include <linux/rculist.h>
  23 #include <linux/skbuff.h>
  24 #include <linux/slab.h>
  25 #include <linux/timer.h>
  26 #include <net/ip.h>
  27 #if IS_ENABLED(CONFIG_IPV6)
  28 #include <net/ipv6.h>
  29 #include <net/mld.h>
  30 #include <net/ip6_checksum.h>
  31 #endif
  32
  33 #include "br_private.h"
  34
  35 static void br_multicast_start_querier(struct net_bridge *br);
  36 unsigned int br_mdb_rehash_seq;
  37
  38 static inline int br_ip_equal(const struct br_ip *a, const struct br_ip *b)
  39 {
  40         if (a->proto != b->proto)
  41                 return 0;
  42         switch (a->proto) {
  43         case htons(ETH_P_IP):
  44                 return a->u.ip4 == b->u.ip4;
  45 #if IS_ENABLED(CONFIG_IPV6)
  46         case htons(ETH_P_IPV6):
  47                 return ipv6_addr_equal(&a->u.ip6, &b->u.ip6);
  48 #endif
  49         }
  50         return 0;
  51 }
  52
  53 static inline int __br_ip4_hash(struct net_bridge_mdb_htable *mdb, __be32 ip)
  54 {
  55         return jhash_1word(mdb->secret, (__force u32)ip) & (mdb->max - 1);
  56 }
  57
  58 #if IS_ENABLED(CONFIG_IPV6)
  59 static inline int __br_ip6_hash(struct net_bridge_mdb_htable *mdb,
  60                                 const struct in6_addr *ip)
  61 {
  62         return jhash2((__force u32 *)ip->s6_addr32, 4, mdb->secret) & (mdb->max - 1);
  63 }
  64 #endif
  65
  66 static inline int br_ip_hash(struct net_bridge_mdb_htable *mdb,
  67                              struct br_ip *ip)
  68 {
  69         switch (ip->proto) {
  70         case htons(ETH_P_IP):
  71                 return __br_ip4_hash(mdb, ip->u.ip4);
  72 #if IS_ENABLED(CONFIG_IPV6)
  73         case htons(ETH_P_IPV6):
  74                 return __br_ip6_hash(mdb, &ip->u.ip6);
  75 #endif
  76         }
  77         return 0;
  78 }
  79
  80 static struct net_bridge_mdb_entry *__br_mdb_ip_get(
  81         struct net_bridge_mdb_htable *mdb, struct br_ip *dst, int hash)
  82 {
  83         struct net_bridge_mdb_entry *mp;
  84         struct hlist_node *p;
  85
  86         hlist_for_each_entry_rcu(mp, p, &mdb->mhash[hash], hlist[mdb->ver]) {
  87                 if (br_ip_equal(&mp->addr, dst))
  88                         return mp;
  89         }
  90
  91         return NULL;
  92 }
  93
  94 struct net_bridge_mdb_entry *br_mdb_ip_get(struct net_bridge_mdb_htable *mdb,
  95                                            struct br_ip *dst)
  96 {
  97         if (!mdb)
  98                 return NULL;
  99
 100         return __br_mdb_ip_get(mdb, dst, br_ip_hash(mdb, dst));
 101 }
 102
 103 static struct net_bridge_mdb_entry *br_mdb_ip4_get(
 104         struct net_bridge_mdb_htable *mdb, __be32 dst)
 105 {
 106         struct br_ip br_dst;
 107
 108         br_dst.u.ip4 = dst;
 109         br_dst.proto = htons(ETH_P_IP);
 110
 111         return br_mdb_ip_get(mdb, &br_dst);
 112 }
 113
 114 #if IS_ENABLED(CONFIG_IPV6)
 115 static struct net_bridge_mdb_entry *br_mdb_ip6_get(
 116         struct net_bridge_mdb_htable *mdb, const struct in6_addr *dst)
 117 {
 118         struct br_ip br_dst;
 119
 120         br_dst.u.ip6 = *dst;
 121         br_dst.proto = htons(ETH_P_IPV6);
 122
 123         return br_mdb_ip_get(mdb, &br_dst);
 124 }
 125 #endif
 126
 127 struct net_bridge_mdb_entry *br_mdb_get(struct net_bridge *br,
 128                                         struct sk_buff *skb)
 129 {
 130         struct net_bridge_mdb_htable *mdb = rcu_dereference(br->mdb);
 131         struct br_ip ip;
 132
 133         if (br->multicast_disabled)
 134                 return NULL;
 135
 136         if (BR_INPUT_SKB_CB(skb)->igmp)
 137                 return NULL;
 138
 139         ip.proto = skb->protocol;
 140
 141         switch (skb->protocol) {
 142         case htons(ETH_P_IP):
 143                 ip.u.ip4 = ip_hdr(skb)->daddr;
 144                 break;
 145 #if IS_ENABLED(CONFIG_IPV6)
 146         case htons(ETH_P_IPV6):
 147                 ip.u.ip6 = ipv6_hdr(skb)->daddr;
 148                 break;
 149 #endif
 150         default:
 151                 return NULL;
 152         }
 153
 154         return br_mdb_ip_get(mdb, &ip);
 155 }
 156
 157 static void br_mdb_free(struct rcu_head *head)
 158 {
 159         struct net_bridge_mdb_htable *mdb =
 160                 container_of(head, struct net_bridge_mdb_htable, rcu);
 161         struct net_bridge_mdb_htable *old = mdb->old;
 162
 163         mdb->old = NULL;
 164         kfree(old->mhash);
 165         kfree(old);
 166 }
 167
 168 static int br_mdb_copy(struct net_bridge_mdb_htable *new,
 169                        struct net_bridge_mdb_htable *old,
 170                        int elasticity)
 171 {
 172         struct net_bridge_mdb_entry *mp;
 173         struct hlist_node *p;
 174         int maxlen;
 175         int len;
 176         int i;
 177
 178         for (i = 0; i < old->max; i++)
 179                 hlist_for_each_entry(mp, p, &old->mhash[i], hlist[old->ver])
 180                         hlist_add_head(&mp->hlist[new->ver],
 181                                        &new->mhash[br_ip_hash(new, &mp->addr)]);
 182
 183         if (!elasticity)
 184                 return 0;
 185
 186         maxlen = 0;
 187         for (i = 0; i < new->max; i++) {
 188                 len = 0;
 189                 hlist_for_each_entry(mp, p, &new->mhash[i], hlist[new->ver])
 190                         len++;
 191                 if (len > maxlen)
 192                         maxlen = len;
 193         }
 194
 195         return maxlen > elasticity ? -EINVAL : 0;
 196 }
 197
 198 void br_multicast_free_pg(struct rcu_head *head)
 199 {
 200         struct net_bridge_port_group *p =
 201                 container_of(head, struct net_bridge_port_group, rcu);
 202
 203         kfree(p);
 204 }
 205
 206 static void br_multicast_free_group(struct rcu_head *head)
 207 {
 208         struct net_bridge_mdb_entry *mp =
 209                 container_of(head, struct net_bridge_mdb_entry, rcu);
 210
 211         kfree(mp);
 212 }
 213
 214 static void br_multicast_group_expired(unsigned long data)
 215 {
 216         struct net_bridge_mdb_entry *mp = (void *)data;
 217         struct net_bridge *br = mp->br;
 218         struct net_bridge_mdb_htable *mdb;
 219
 220         spin_lock(&br->multicast_lock);
 221         if (!netif_running(br->dev) || timer_pending(&mp->timer))
 222                 goto out;
 223
 224         mp->mglist = false;
 225
 226         if (mp->ports)
 227                 goto out;
 228
 229         mdb = mlock_dereference(br->mdb, br);
 230
 231         hlist_del_rcu(&mp->hlist[mdb->ver]);
 232         mdb->size--;
 233
 234         call_rcu_bh(&mp->rcu, br_multicast_free_group);
 235
 236 out:
 237         spin_unlock(&br->multicast_lock);
 238 }
 239
 240 static void br_multicast_del_pg(struct net_bridge *br,
 241                                 struct net_bridge_port_group *pg)
 242 {
 243         struct net_bridge_mdb_htable *mdb;
 244         struct net_bridge_mdb_entry *mp;
 245         struct net_bridge_port_group *p;
 246         struct net_bridge_port_group __rcu **pp;
 247
 248         mdb = mlock_dereference(br->mdb, br);
 249
 250         mp = br_mdb_ip_get(mdb, &pg->addr);
 251         if (WARN_ON(!mp))
 252                 return;
 253
 254         for (pp = &mp->ports;
 255              (p = mlock_dereference(*pp, br)) != NULL;
 256              pp = &p->next) {
 257                 if (p != pg)
 258                         continue;
 259
 260                 rcu_assign_pointer(*pp, p->next);
 261                 hlist_del_init(&p->mglist);
 262                 del_timer(&p->timer);
 263                 call_rcu_bh(&p->rcu, br_multicast_free_pg);
 264
 265                 if (!mp->ports && !mp->mglist &&
 266                     netif_running(br->dev))
 267                         mod_timer(&mp->timer, jiffies);
 268
 269                 return;
 270         }
 271
 272         WARN_ON(1);
 273 }
 274
 275 static void br_multicast_port_group_expired(unsigned long data)
 276 {
 277         struct net_bridge_port_group *pg = (void *)data;
 278         struct net_bridge *br = pg->port->br;
 279
 280         spin_lock(&br->multicast_lock);
 281         if (!netif_running(br->dev) || timer_pending(&pg->timer) ||
 282             hlist_unhashed(&pg->mglist))
 283                 goto out;
 284
 285         br_multicast_del_pg(br, pg);
 286
 287 out:
 288         spin_unlock(&br->multicast_lock);
 289 }
 290
 291 static int br_mdb_rehash(struct net_bridge_mdb_htable __rcu **mdbp, int max,
 292                          int elasticity)
 293 {
 294         struct net_bridge_mdb_htable *old = rcu_dereference_protected(*mdbp, 1);
 295         struct net_bridge_mdb_htable *mdb;
 296         int err;
 297
 298         mdb = kmalloc(sizeof(*mdb), GFP_ATOMIC);
 299         if (!mdb)
 300                 return -ENOMEM;
 301
 302         mdb->max = max;
 303         mdb->old = old;
 304
 305         mdb->mhash = kzalloc(max * sizeof(*mdb->mhash), GFP_ATOMIC);
 306         if (!mdb->mhash) {
 307                 kfree(mdb);
 308                 return -ENOMEM;
 309         }
 310
 311         mdb->size = old ? old->size : 0;
 312         mdb->ver = old ? old->ver ^ 1 : 0;
 313
 314         if (!old || elasticity)
 315                 get_random_bytes(&mdb->secret, sizeof(mdb->secret));
 316         else
 317                 mdb->secret = old->secret;
 318
 319         if (!old)
 320                 goto out;
 321
 322         err = br_mdb_copy(mdb, old, elasticity);
 323         if (err) {
 324                 kfree(mdb->mhash);
 325                 kfree(mdb);
 326                 return err;
 327         }
 328
 329         br_mdb_rehash_seq++;
 330         call_rcu_bh(&mdb->rcu, br_mdb_free);
 331
 332 out:
 333         rcu_assign_pointer(*mdbp, mdb);
 334
 335         return 0;
 336 }
 337
 338 static struct sk_buff *br_ip4_multicast_alloc_query(struct net_bridge *br,
 339                                                     __be32 group)
 340 {
 341         struct sk_buff *skb;
 342         struct igmphdr *ih;
 343         struct ethhdr *eth;
 344         struct iphdr *iph;
 345
 346         skb = netdev_alloc_skb_ip_align(br->dev, sizeof(*eth) + sizeof(*iph) +
 347                                                  sizeof(*ih) + 4);
 348         if (!skb)
 349                 goto out;
 350
 351         skb->protocol = htons(ETH_P_IP);
 352
 353         skb_reset_mac_header(skb);
 354         eth = eth_hdr(skb);
 355
 356         memcpy(eth->h_source, br->dev->dev_addr, 6);
 357         eth->h_dest[0] = 1;
 358         eth->h_dest[1] = 0;
 359         eth->h_dest[2] = 0x5e;
 360         eth->h_dest[3] = 0;
 361         eth->h_dest[4] = 0;
 362         eth->h_dest[5] = 1;
 363         eth->h_proto = htons(ETH_P_IP);
 364         skb_put(skb, sizeof(*eth));
 365
 366         skb_set_network_header(skb, skb->len);
 367         iph = ip_hdr(skb);
 368
 369         iph->version = 4;
 370         iph->ihl = 6;
 371         iph->tos = 0xc0;
 372         iph->tot_len = htons(sizeof(*iph) + sizeof(*ih) + 4);
 373         iph->id = 0;
 374         iph->frag_off = htons(IP_DF);
 375         iph->ttl = 1;
 376         iph->protocol = IPPROTO_IGMP;
 377         iph->saddr = 0;
 378         iph->daddr = htonl(INADDR_ALLHOSTS_GROUP);
 379         ((u8 *)&iph[1])[0] = IPOPT_RA;
 380         ((u8 *)&iph[1])[1] = 4;
 381         ((u8 *)&iph[1])[2] = 0;
 382         ((u8 *)&iph[1])[3] = 0;
 383         ip_send_check(iph);
 384         skb_put(skb, 24);
 385
 386         skb_set_transport_header(skb, skb->len);
 387         ih = igmp_hdr(skb);
 388         ih->type = IGMP_HOST_MEMBERSHIP_QUERY;
 389         ih->code = (group ? br->multicast_last_member_interval :
 390                             br->multicast_query_response_interval) /
 391                    (HZ / IGMP_TIMER_SCALE);
 392         ih->group = group;
 393         ih->csum = 0;
 394         ih->csum = ip_compute_csum((void *)ih, sizeof(struct igmphdr));
 395         skb_put(skb, sizeof(*ih));
 396
 397         __skb_pull(skb, sizeof(*eth));
 398
 399 out:
 400         return skb;
 401 }
 402
 403 #if IS_ENABLED(CONFIG_IPV6)
 404 static struct sk_buff *br_ip6_multicast_alloc_query(struct net_bridge *br,
 405                                                     const struct in6_addr *group)
 406 {
 407         struct sk_buff *skb;
 408         struct ipv6hdr *ip6h;
 409         struct mld_msg *mldq;
 410         struct ethhdr *eth;
 411         u8 *hopopt;
 412         unsigned long interval;
 413
 414         skb = netdev_alloc_skb_ip_align(br->dev, sizeof(*eth) + sizeof(*ip6h) +
 415                                                  8 + sizeof(*mldq));
 416         if (!skb)
 417                 goto out;
 418
 419         skb->protocol = htons(ETH_P_IPV6);
 420
 421         /* Ethernet header */
 422         skb_reset_mac_header(skb);
 423         eth = eth_hdr(skb);
 424
 425         memcpy(eth->h_source, br->dev->dev_addr, 6);
 426         eth->h_proto = htons(ETH_P_IPV6);
 427         skb_put(skb, sizeof(*eth));
 428
 429         /* IPv6 header + HbH option */
 430         skb_set_network_header(skb, skb->len);
 431         ip6h = ipv6_hdr(skb);
 432
 433         *(__force __be32 *)ip6h = htonl(0x60000000);
 434         ip6h->payload_len = htons(8 + sizeof(*mldq));
 435         ip6h->nexthdr = IPPROTO_HOPOPTS;
 436         ip6h->hop_limit = 1;
 437         ipv6_addr_set(&ip6h->daddr, htonl(0xff020000), 0, 0, htonl(1));
 438         if (ipv6_dev_get_saddr(dev_net(br->dev), br->dev, &ip6h->daddr, 0,
 439                                &ip6h->saddr)) {
 440                 kfree_skb(skb);
 441                 return NULL;
 442         }
 443         ipv6_eth_mc_map(&ip6h->daddr, eth->h_dest);
 444
 445         hopopt = (u8 *)(ip6h + 1);
 446         hopopt[0] = IPPROTO_ICMPV6;             /* next hdr */
 447         hopopt[1] = 0;                          /* length of HbH */
 448         hopopt[2] = IPV6_TLV_ROUTERALERT;       /* Router Alert */
 449         hopopt[3] = 2;                          /* Length of RA Option */
 450         hopopt[4] = 0;                          /* Type = 0x0000 (MLD) */
 451         hopopt[5] = 0;
 452         hopopt[6] = IPV6_TLV_PAD1;              /* Pad1 */
 453         hopopt[7] = IPV6_TLV_PAD1;              /* Pad1 */
 454
 455         skb_put(skb, sizeof(*ip6h) + 8);
 456
 457         /* ICMPv6 */
 458         skb_set_transport_header(skb, skb->len);
 459         mldq = (struct mld_msg *) icmp6_hdr(skb);
 460
 461         interval = ipv6_addr_any(group) ? br->multicast_last_member_interval :
 462                                           br->multicast_query_response_interval;
 463
 464         mldq->mld_type = ICMPV6_MGM_QUERY;
 465         mldq->mld_code = 0;
 466         mldq->mld_cksum = 0;
 467         mldq->mld_maxdelay = htons((u16)jiffies_to_msecs(interval));
 468         mldq->mld_reserved = 0;
 469         mldq->mld_mca = *group;
 470
 471         /* checksum */
 472         mldq->mld_cksum = csum_ipv6_magic(&ip6h->saddr, &ip6h->daddr,
 473                                           sizeof(*mldq), IPPROTO_ICMPV6,
 474                                           csum_partial(mldq,
 475                                                        sizeof(*mldq), 0));
 476         skb_put(skb, sizeof(*mldq));
 477
 478         __skb_pull(skb, sizeof(*eth));
 479
 480 out:
 481         return skb;
 482 }
 483 #endif
 484
 485 static struct sk_buff *br_multicast_alloc_query(struct net_bridge *br,
 486                                                 struct br_ip *addr)
 487 {
 488         switch (addr->proto) {
 489         case htons(ETH_P_IP):
 490                 return br_ip4_multicast_alloc_query(br, addr->u.ip4);
 491 #if IS_ENABLED(CONFIG_IPV6)
 492         case htons(ETH_P_IPV6):
 493                 return br_ip6_multicast_alloc_query(br, &addr->u.ip6);
 494 #endif
 495         }
 496         return NULL;
 497 }
 498
 499 static struct net_bridge_mdb_entry *br_multicast_get_group(
 500         struct net_bridge *br, struct net_bridge_port *port,
 501         struct br_ip *group, int hash)
 502 {
 503         struct net_bridge_mdb_htable *mdb;
 504         struct net_bridge_mdb_entry *mp;
 505         struct hlist_node *p;
 506         unsigned int count = 0;
 507         unsigned int max;
 508         int elasticity;
 509         int err;
 510
 511         mdb = rcu_dereference_protected(br->mdb, 1);
 512         hlist_for_each_entry(mp, p, &mdb->mhash[hash], hlist[mdb->ver]) {
 513                 count++;
 514                 if (unlikely(br_ip_equal(group, &mp->addr)))
 515                         return mp;
 516         }
 517
 518         elasticity = 0;
 519         max = mdb->max;
 520
 521         if (unlikely(count > br->hash_elasticity && count)) {
 522                 if (net_ratelimit())
 523                         br_info(br, "Multicast hash table "
 524                                 "chain limit reached: %s\n",
 525                                 port ? port->dev->name : br->dev->name);
 526
 527                 elasticity = br->hash_elasticity;
 528         }
 529
 530         if (mdb->size >= max) {
 531                 max *= 2;
 532                 if (unlikely(max > br->hash_max)) {
 533                         br_warn(br, "Multicast hash table maximum of %d "
 534                                 "reached, disabling snooping: %s\n",
 535                                 br->hash_max,
 536                                 port ? port->dev->name : br->dev->name);
 537                         err = -E2BIG;
 538 disable:
 539                         br->multicast_disabled = 1;
 540                         goto err;
 541                 }
 542         }
 543
 544         if (max > mdb->max || elasticity) {
 545                 if (mdb->old) {
 546                         if (net_ratelimit())
 547                                 br_info(br, "Multicast hash table "
 548                                         "on fire: %s\n",
 549                                         port ? port->dev->name : br->dev->name);
 550                         err = -EEXIST;
 551                         goto err;
 552                 }
 553
 554                 err = br_mdb_rehash(&br->mdb, max, elasticity);
 555                 if (err) {
 556                         br_warn(br, "Cannot rehash multicast "
 557                                 "hash table, disabling snooping: %s, %d, %d\n",
 558                                 port ? port->dev->name : br->dev->name,
 559                                 mdb->size, err);
 560                         goto disable;
 561                 }
 562
 563                 err = -EAGAIN;
 564                 goto err;
 565         }
 566
 567         return NULL;
 568
 569 err:
 570         mp = ERR_PTR(err);
 571         return mp;
 572 }
 573
 574 struct net_bridge_mdb_entry *br_multicast_new_group(struct net_bridge *br,
 575         struct net_bridge_port *port, struct br_ip *group)
 576 {
 577         struct net_bridge_mdb_htable *mdb;
 578         struct net_bridge_mdb_entry *mp;
 579         int hash;
 580         int err;
 581
 582         mdb = rcu_dereference_protected(br->mdb, 1);
 583         if (!mdb) {
 584                 err = br_mdb_rehash(&br->mdb, BR_HASH_SIZE, 0);
 585                 if (err)
 586                         return ERR_PTR(err);
 587                 goto rehash;
 588         }
 589
 590         hash = br_ip_hash(mdb, group);
 591         mp = br_multicast_get_group(br, port, group, hash);
 592         switch (PTR_ERR(mp)) {
 593         case 0:
 594                 break;
 595
 596         case -EAGAIN:
 597 rehash:
 598                 mdb = rcu_dereference_protected(br->mdb, 1);
 599                 hash = br_ip_hash(mdb, group);
 600                 break;
 601
 602         default:
 603                 goto out;
 604         }
 605
 606         mp = kzalloc(sizeof(*mp), GFP_ATOMIC);
 607         if (unlikely(!mp))
 608                 return ERR_PTR(-ENOMEM);
 609
 610         mp->br = br;
 611         mp->addr = *group;
 612         setup_timer(&mp->timer, br_multicast_group_expired,
 613                     (unsigned long)mp);
 614
 615         hlist_add_head_rcu(&mp->hlist[mdb->ver], &mdb->mhash[hash]);
 616         mdb->size++;
 617
 618 out:
 619         return mp;
 620 }
 621
 622 struct net_bridge_port_group *br_multicast_new_port_group(
 623                         struct net_bridge_port *port,
 624                         struct br_ip *group,
 625                         struct net_bridge_port_group *next)
 626 {
 627         struct net_bridge_port_group *p;
 628
 629         p = kzalloc(sizeof(*p), GFP_ATOMIC);
 630         if (unlikely(!p))
 631                 return NULL;
 632
 633         p->addr = *group;
 634         p->port = port;
 635         p->next = next;
 636         hlist_add_head(&p->mglist, &port->mglist);
 637         setup_timer(&p->timer, br_multicast_port_group_expired,
 638                     (unsigned long)p);
 639         return p;
 640 }
 641
 642 static int br_multicast_add_group(struct net_bridge *br,
 643                                   struct net_bridge_port *port,
 644                                   struct br_ip *group)
 645 {
 646         struct net_bridge_mdb_entry *mp;
 647         struct net_bridge_port_group *p;
 648         struct net_bridge_port_group __rcu **pp;
 649         unsigned long now = jiffies;
 650         int err;
 651
 652         spin_lock(&br->multicast_lock);
 653         if (!netif_running(br->dev) ||
 654             (port && port->state == BR_STATE_DISABLED))
 655                 goto out;
 656
 657         mp = br_multicast_new_group(br, port, group);
 658         err = PTR_ERR(mp);
 659         if (IS_ERR(mp))
 660                 goto err;
 661
 662         if (!port) {
 663                 mp->mglist = true;
 664                 mod_timer(&mp->timer, now + br->multicast_membership_interval);
 665                 goto out;
 666         }
 667
 668         for (pp = &mp->ports;
 669              (p = mlock_dereference(*pp, br)) != NULL;
 670              pp = &p->next) {
 671                 if (p->port == port)
 672                         goto found;
 673                 if ((unsigned long)p->port < (unsigned long)port)
 674                         break;
 675         }
 676
 677         p = br_multicast_new_port_group(port, group, *pp);
 678         if (unlikely(!p))
 679                 goto err;
 680         rcu_assign_pointer(*pp, p);
 681         br_mdb_notify(br->dev, port, group, RTM_NEWMDB);
 682
 683 found:
 684         mod_timer(&p->timer, now + br->multicast_membership_interval);
 685 out:
 686         err = 0;
 687
 688 err:
 689         spin_unlock(&br->multicast_lock);
 690         return err;
 691 }
 692
 693 static int br_ip4_multicast_add_group(struct net_bridge *br,
 694                                       struct net_bridge_port *port,
 695                                       __be32 group)
 696 {
 697         struct br_ip br_group;
 698
 699         if (ipv4_is_local_multicast(group))
 700                 return 0;
 701
 702         br_group.u.ip4 = group;
 703         br_group.proto = htons(ETH_P_IP);
 704
 705         return br_multicast_add_group(br, port, &br_group);
 706 }
 707
 708 #if IS_ENABLED(CONFIG_IPV6)
 709 static int br_ip6_multicast_add_group(struct net_bridge *br,
 710                                       struct net_bridge_port *port,
 711                                       const struct in6_addr *group)
 712 {
 713         struct br_ip br_group;
 714
 715         if (!ipv6_is_transient_multicast(group))
 716                 return 0;
 717
 718         br_group.u.ip6 = *group;
 719         br_group.proto = htons(ETH_P_IPV6);
 720
 721         return br_multicast_add_group(br, port, &br_group);
 722 }
 723 #endif
 724
 725 static void br_multicast_router_expired(unsigned long data)
 726 {
 727         struct net_bridge_port *port = (void *)data;
 728         struct net_bridge *br = port->br;
 729
 730         spin_lock(&br->multicast_lock);
 731         if (port->multicast_router != 1 ||
 732             timer_pending(&port->multicast_router_timer) ||
 733             hlist_unhashed(&port->rlist))
 734                 goto out;
 735
 736         hlist_del_init_rcu(&port->rlist);
 737
 738 out:
 739         spin_unlock(&br->multicast_lock);
 740 }
 741
 742 static void br_multicast_local_router_expired(unsigned long data)
 743 {
 744 }
 745
 746 static void br_multicast_querier_expired(unsigned long data)
 747 {
 748         struct net_bridge *br = (void *)data;
 749
 750         spin_lock(&br->multicast_lock);
 751         if (!netif_running(br->dev) || br->multicast_disabled)
 752                 goto out;
 753
 754         br_multicast_start_querier(br);
 755
 756 out:
 757         spin_unlock(&br->multicast_lock);
 758 }
 759
 760 static void __br_multicast_send_query(struct net_bridge *br,
 761                                       struct net_bridge_port *port,
 762                                       struct br_ip *ip)
 763 {
 764         struct sk_buff *skb;
 765
 766         skb = br_multicast_alloc_query(br, ip);
 767         if (!skb)
 768                 return;
 769
 770         if (port) {
 771                 __skb_push(skb, sizeof(struct ethhdr));
 772                 skb->dev = port->dev;
 773                 NF_HOOK(NFPROTO_BRIDGE, NF_BR_LOCAL_OUT, skb, NULL, skb->dev,
 774                         dev_queue_xmit);
 775         } else
 776                 netif_rx(skb);
 777 }
 778
 779 static void br_multicast_send_query(struct net_bridge *br,
 780                                     struct net_bridge_port *port, u32 sent)
 781 {
 782         unsigned long time;
 783         struct br_ip br_group;
 784
 785         if (!netif_running(br->dev) || br->multicast_disabled ||
 786             !br->multicast_querier ||
 787             timer_pending(&br->multicast_querier_timer))
 788                 return;
 789
 790         memset(&br_group.u, 0, sizeof(br_group.u));
 791
 792         br_group.proto = htons(ETH_P_IP);
 793         __br_multicast_send_query(br, port, &br_group);
 794
 795 #if IS_ENABLED(CONFIG_IPV6)
 796         br_group.proto = htons(ETH_P_IPV6);
 797         __br_multicast_send_query(br, port, &br_group);
 798 #endif
 799
 800         time = jiffies;
 801         time += sent < br->multicast_startup_query_count ?
 802                 br->multicast_startup_query_interval :
 803                 br->multicast_query_interval;
 804         mod_timer(port ? &port->multicast_query_timer :
 805                          &br->multicast_query_timer, time);
 806 }
 807
 808 static void br_multicast_port_query_expired(unsigned long data)
 809 {
 810         struct net_bridge_port *port = (void *)data;
 811         struct net_bridge *br = port->br;
 812
 813         spin_lock(&br->multicast_lock);
 814         if (port->state == BR_STATE_DISABLED ||
 815             port->state == BR_STATE_BLOCKING)
 816                 goto out;
 817
 818         if (port->multicast_startup_queries_sent <
 819             br->multicast_startup_query_count)
 820                 port->multicast_startup_queries_sent++;
 821
 822         br_multicast_send_query(port->br, port,
 823                                 port->multicast_startup_queries_sent);
 824
 825 out:
 826         spin_unlock(&br->multicast_lock);
 827 }
 828
 829 void br_multicast_add_port(struct net_bridge_port *port)
 830 {
 831         port->multicast_router = 1;
 832
 833         setup_timer(&port->multicast_router_timer, br_multicast_router_expired,
 834                     (unsigned long)port);
 835         setup_timer(&port->multicast_query_timer,
 836                     br_multicast_port_query_expired, (unsigned long)port);
 837 }
 838
 839 void br_multicast_del_port(struct net_bridge_port *port)
 840 {
 841         del_timer_sync(&port->multicast_router_timer);
 842 }
 843
 844 static void __br_multicast_enable_port(struct net_bridge_port *port)
 845 {
 846         port->multicast_startup_queries_sent = 0;
 847
 848         if (try_to_del_timer_sync(&port->multicast_query_timer) >= 0 ||
 849             del_timer(&port->multicast_query_timer))
 850                 mod_timer(&port->multicast_query_timer, jiffies);
 851 }
 852
 853 void br_multicast_enable_port(struct net_bridge_port *port)
 854 {
 855         struct net_bridge *br = port->br;
 856
 857         spin_lock(&br->multicast_lock);
 858         if (br->multicast_disabled || !netif_running(br->dev))
 859                 goto out;
 860
 861         __br_multicast_enable_port(port);
 862
 863 out:
 864         spin_unlock(&br->multicast_lock);
 865 }
 866
 867 void br_multicast_disable_port(struct net_bridge_port *port)
 868 {
 869         struct net_bridge *br = port->br;
 870         struct net_bridge_port_group *pg;
 871         struct hlist_node *p, *n;
 872
 873         spin_lock(&br->multicast_lock);
 874         hlist_for_each_entry_safe(pg, p, n, &port->mglist, mglist)
 875                 br_multicast_del_pg(br, pg);
 876
 877         if (!hlist_unhashed(&port->rlist))
 878                 hlist_del_init_rcu(&port->rlist);
 879         del_timer(&port->multicast_router_timer);
 880         del_timer(&port->multicast_query_timer);
 881         spin_unlock(&br->multicast_lock);
 882 }
 883
 884 static int br_ip4_multicast_igmp3_report(struct net_bridge *br,
 885                                          struct net_bridge_port *port,
 886                                          struct sk_buff *skb)
 887 {
 888         struct igmpv3_report *ih;
 889         struct igmpv3_grec *grec;
 890         int i;
 891         int len;
 892         int num;
 893         int type;
 894         int err = 0;
 895         __be32 group;
 896
 897         if (!pskb_may_pull(skb, sizeof(*ih)))
 898                 return -EINVAL;
 899
 900         ih = igmpv3_report_hdr(skb);
 901         num = ntohs(ih->ngrec);
 902         len = sizeof(*ih);
 903
 904         for (i = 0; i < num; i++) {
 905                 len += sizeof(*grec);
 906                 if (!pskb_may_pull(skb, len))
 907                         return -EINVAL;
 908
 909                 grec = (void *)(skb->data + len - sizeof(*grec));
 910                 group = grec->grec_mca;
 911                 type = grec->grec_type;
 912
 913                 len += ntohs(grec->grec_nsrcs) * 4;
 914                 if (!pskb_may_pull(skb, len))
 915                         return -EINVAL;
 916
 917                 /* We treat this as an IGMPv2 report for now. */
 918                 switch (type) {
 919                 case IGMPV3_MODE_IS_INCLUDE:
 920                 case IGMPV3_MODE_IS_EXCLUDE:
 921                 case IGMPV3_CHANGE_TO_INCLUDE:
 922                 case IGMPV3_CHANGE_TO_EXCLUDE:
 923                 case IGMPV3_ALLOW_NEW_SOURCES:
 924                 case IGMPV3_BLOCK_OLD_SOURCES:
 925                         break;
 926
 927                 default:
 928                         continue;
 929                 }
 930
 931                 err = br_ip4_multicast_add_group(br, port, group);
 932                 if (err)
 933                         break;
 934         }
 935
 936         return err;
 937 }
 938
 939 #if IS_ENABLED(CONFIG_IPV6)
 940 static int br_ip6_multicast_mld2_report(struct net_bridge *br,
 941                                         struct net_bridge_port *port,
 942                                         struct sk_buff *skb)
 943 {
 944         struct icmp6hdr *icmp6h;
 945         struct mld2_grec *grec;
 946         int i;
 947         int len;
 948         int num;
 949         int err = 0;
 950
 951         if (!pskb_may_pull(skb, sizeof(*icmp6h)))
 952                 return -EINVAL;
 953
 954         icmp6h = icmp6_hdr(skb);
 955         num = ntohs(icmp6h->icmp6_dataun.un_data16[1]);
 956         len = sizeof(*icmp6h);
 957
 958         for (i = 0; i < num; i++) {
 959                 __be16 *nsrcs, _nsrcs;
 960
 961                 nsrcs = skb_header_pointer(skb,
 962                                            len + offsetof(struct mld2_grec,
 963                                                           grec_nsrcs),
 964                                            sizeof(_nsrcs), &_nsrcs);
 965                 if (!nsrcs)
 966                         return -EINVAL;
 967
 968                 if (!pskb_may_pull(skb,
 969                                    len + sizeof(*grec) +
 970                                    sizeof(struct in6_addr) * ntohs(*nsrcs)))
 971                         return -EINVAL;
 972
 973                 grec = (struct mld2_grec *)(skb->data + len);
 974                 len += sizeof(*grec) +
 975                        sizeof(struct in6_addr) * ntohs(*nsrcs);
 976
 977                 /* We treat these as MLDv1 reports for now. */
 978                 switch (grec->grec_type) {
 979                 case MLD2_MODE_IS_INCLUDE:
 980                 case MLD2_MODE_IS_EXCLUDE:
 981                 case MLD2_CHANGE_TO_INCLUDE:
 982                 case MLD2_CHANGE_TO_EXCLUDE:
 983                 case MLD2_ALLOW_NEW_SOURCES:
 984                 case MLD2_BLOCK_OLD_SOURCES:
 985                         break;
 986
 987                 default:
 988                         continue;
 989                 }
 990
 991                 err = br_ip6_multicast_add_group(br, port, &grec->grec_mca);
 992                 if (!err)
 993                         break;
 994         }
 995
 996         return err;
 997 }
 998 #endif
 999
1000 /*
1001  * Add port to rotuer_list
1002  *  list is maintained ordered by pointer value
1003  *  and locked by br->multicast_lock and RCU
1004  */
1005 static void br_multicast_add_router(struct net_bridge *br,
1006                                     struct net_bridge_port *port)
1007 {
1008         struct net_bridge_port *p;
1009         struct hlist_node *n, *slot = NULL;
1010
1011         hlist_for_each_entry(p, n, &br->router_list, rlist) {
1012                 if ((unsigned long) port >= (unsigned long) p)
1013                         break;
1014                 slot = n;
1015         }
1016
1017         if (slot)
1018                 hlist_add_after_rcu(slot, &port->rlist);
1019         else
1020                 hlist_add_head_rcu(&port->rlist, &br->router_list);
1021 }
1022
1023 static void br_multicast_mark_router(struct net_bridge *br,
1024                                      struct net_bridge_port *port)
1025 {
1026         unsigned long now = jiffies;
1027
1028         if (!port) {
1029                 if (br->multicast_router == 1)
1030                         mod_timer(&br->multicast_router_timer,
1031                                   now + br->multicast_querier_interval);
1032                 return;
1033         }
1034
1035         if (port->multicast_router != 1)
1036                 return;
1037
1038         if (!hlist_unhashed(&port->rlist))
1039                 goto timer;
1040
1041         br_multicast_add_router(br, port);
1042
1043 timer:
1044         mod_timer(&port->multicast_router_timer,
1045                   now + br->multicast_querier_interval);
1046 }
1047
1048 static void br_multicast_query_received(struct net_bridge *br,
1049                                         struct net_bridge_port *port,
1050                                         int saddr)
1051 {
1052         if (saddr)
1053                 mod_timer(&br->multicast_querier_timer,
1054                           jiffies + br->multicast_querier_interval);
1055         else if (timer_pending(&br->multicast_querier_timer))
1056                 return;
1057
1058         br_multicast_mark_router(br, port);
1059 }
1060
1061 static int br_ip4_multicast_query(struct net_bridge *br,
1062                                   struct net_bridge_port *port,
1063                                   struct sk_buff *skb)
1064 {
1065         const struct iphdr *iph = ip_hdr(skb);
1066         struct igmphdr *ih = igmp_hdr(skb);
1067         struct net_bridge_mdb_entry *mp;
1068         struct igmpv3_query *ih3;
1069         struct net_bridge_port_group *p;
1070         struct net_bridge_port_group __rcu **pp;
1071         unsigned long max_delay;
1072         unsigned long now = jiffies;
1073         __be32 group;
1074         int err = 0;
1075
1076         spin_lock(&br->multicast_lock);
1077         if (!netif_running(br->dev) ||
1078             (port && port->state == BR_STATE_DISABLED))
1079                 goto out;
1080
1081         br_multicast_query_received(br, port, !!iph->saddr);
1082
1083         group = ih->group;
1084
1085         if (skb->len == sizeof(*ih)) {
1086                 max_delay = ih->code * (HZ / IGMP_TIMER_SCALE);
1087
1088                 if (!max_delay) {
1089                         max_delay = 10 * HZ;
1090                         group = 0;
1091                 }
1092         } else {
1093                 if (!pskb_may_pull(skb, sizeof(struct igmpv3_query))) {
1094                         err = -EINVAL;
1095                         goto out;
1096                 }
1097
1098                 ih3 = igmpv3_query_hdr(skb);
1099                 if (ih3->nsrcs)
1100                         goto out;
1101
1102                 max_delay = ih3->code ?
1103                             IGMPV3_MRC(ih3->code) * (HZ / IGMP_TIMER_SCALE) : 1;
1104         }
1105
1106         if (!group)
1107                 goto out;
1108
1109         mp = br_mdb_ip4_get(mlock_dereference(br->mdb, br), group);
1110         if (!mp)
1111                 goto out;
1112
1113         max_delay *= br->multicast_last_member_count;
1114
1115         if (mp->mglist &&
1116             (timer_pending(&mp->timer) ?
1117              time_after(mp->timer.expires, now + max_delay) :
1118              try_to_del_timer_sync(&mp->timer) >= 0))
1119                 mod_timer(&mp->timer, now + max_delay);
1120
1121         for (pp = &mp->ports;
1122              (p = mlock_dereference(*pp, br)) != NULL;
1123              pp = &p->next) {
1124                 if (timer_pending(&p->timer) ?
1125                     time_after(p->timer.expires, now + max_delay) :
1126                     try_to_del_timer_sync(&p->timer) >= 0)
1127                         mod_timer(&p->timer, now + max_delay);
1128         }
1129
1130 out:
1131         spin_unlock(&br->multicast_lock);
1132         return err;
1133 }
1134
1135 #if IS_ENABLED(CONFIG_IPV6)
1136 static int br_ip6_multicast_query(struct net_bridge *br,
1137                                   struct net_bridge_port *port,
1138                                   struct sk_buff *skb)
1139 {
1140         const struct ipv6hdr *ip6h = ipv6_hdr(skb);
1141         struct mld_msg *mld = (struct mld_msg *) icmp6_hdr(skb);
1142         struct net_bridge_mdb_entry *mp;
1143         struct mld2_query *mld2q;
1144         struct net_bridge_port_group *p;
1145         struct net_bridge_port_group __rcu **pp;
1146         unsigned long max_delay;
1147         unsigned long now = jiffies;
1148         const struct in6_addr *group = NULL;
1149         int err = 0;
1150
1151         spin_lock(&br->multicast_lock);
1152         if (!netif_running(br->dev) ||
1153             (port && port->state == BR_STATE_DISABLED))
1154                 goto out;
1155
1156         br_multicast_query_received(br, port, !ipv6_addr_any(&ip6h->saddr));
1157
1158         if (skb->len == sizeof(*mld)) {
1159                 if (!pskb_may_pull(skb, sizeof(*mld))) {
1160                         err = -EINVAL;
1161                         goto out;
1162                 }
1163                 mld = (struct mld_msg *) icmp6_hdr(skb);
1164                 max_delay = msecs_to_jiffies(ntohs(mld->mld_maxdelay));
1165                 if (max_delay)
1166                         group = &mld->mld_mca;
1167         } else if (skb->len >= sizeof(*mld2q)) {
1168                 if (!pskb_may_pull(skb, sizeof(*mld2q))) {
1169                         err = -EINVAL;
1170                         goto out;
1171                 }
1172                 mld2q = (struct mld2_query *)icmp6_hdr(skb);
1173                 if (!mld2q->mld2q_nsrcs)
1174                         group = &mld2q->mld2q_mca;
1175                 max_delay = mld2q->mld2q_mrc ? MLDV2_MRC(mld2q->mld2q_mrc) : 1;
1176         }
1177
1178         if (!group)
1179                 goto out;
1180
1181         mp = br_mdb_ip6_get(mlock_dereference(br->mdb, br), group);
1182         if (!mp)
1183                 goto out;
1184
1185         max_delay *= br->multicast_last_member_count;
1186         if (mp->mglist &&
1187             (timer_pending(&mp->timer) ?
1188              time_after(mp->timer.expires, now + max_delay) :
1189              try_to_del_timer_sync(&mp->timer) >= 0))
1190                 mod_timer(&mp->timer, now + max_delay);
1191
1192         for (pp = &mp->ports;
1193              (p = mlock_dereference(*pp, br)) != NULL;
1194              pp = &p->next) {
1195                 if (timer_pending(&p->timer) ?
1196                     time_after(p->timer.expires, now + max_delay) :
1197                     try_to_del_timer_sync(&p->timer) >= 0)
1198                         mod_timer(&p->timer, now + max_delay);
1199         }
1200
1201 out:
1202         spin_unlock(&br->multicast_lock);
1203         return err;
1204 }
1205 #endif
1206
1207 static void br_multicast_leave_group(struct net_bridge *br,
1208                                      struct net_bridge_port *port,
1209                                      struct br_ip *group)
1210 {
1211         struct net_bridge_mdb_htable *mdb;
1212         struct net_bridge_mdb_entry *mp;
1213         struct net_bridge_port_group *p;
1214         unsigned long now;
1215         unsigned long time;
1216
1217         spin_lock(&br->multicast_lock);
1218         if (!netif_running(br->dev) ||
1219             (port && port->state == BR_STATE_DISABLED) ||
1220             timer_pending(&br->multicast_querier_timer))
1221                 goto out;
1222
1223         mdb = mlock_dereference(br->mdb, br);
1224         mp = br_mdb_ip_get(mdb, group);
1225         if (!mp)
1226                 goto out;
1227
1228         if (port && (port->flags & BR_MULTICAST_FAST_LEAVE)) {
1229                 struct net_bridge_port_group __rcu **pp;
1230
1231                 for (pp = &mp->ports;
1232                      (p = mlock_dereference(*pp, br)) != NULL;
1233                      pp = &p->next) {
1234                         if (p->port != port)
1235                                 continue;
1236
1237                         rcu_assign_pointer(*pp, p->next);
1238                         hlist_del_init(&p->mglist);
1239                         del_timer(&p->timer);
1240                         call_rcu_bh(&p->rcu, br_multicast_free_pg);
1241                         br_mdb_notify(br->dev, port, group, RTM_DELMDB);
1242
1243                         if (!mp->ports && !mp->mglist &&
1244                             netif_running(br->dev))
1245                                 mod_timer(&mp->timer, jiffies);
1246                 }
1247                 goto out;
1248         }
1249
1250         now = jiffies;
1251         time = now + br->multicast_last_member_count *
1252                      br->multicast_last_member_interval;
1253
1254         if (!port) {
1255                 if (mp->mglist &&
1256                     (timer_pending(&mp->timer) ?
1257                      time_after(mp->timer.expires, time) :
1258                      try_to_del_timer_sync(&mp->timer) >= 0)) {
1259                         mod_timer(&mp->timer, time);
1260                 }
1261
1262                 goto out;
1263         }
1264
1265         for (p = mlock_dereference(mp->ports, br);
1266              p != NULL;
1267              p = mlock_dereference(p->next, br)) {
1268                 if (p->port != port)
1269                         continue;
1270
1271                 if (!hlist_unhashed(&p->mglist) &&
1272                     (timer_pending(&p->timer) ?
1273                      time_after(p->timer.expires, time) :
1274                      try_to_del_timer_sync(&p->timer) >= 0)) {
1275                         mod_timer(&p->timer, time);
1276                 }
1277
1278                 break;
1279         }
1280
1281 out:
1282         spin_unlock(&br->multicast_lock);
1283 }
1284
1285 static void br_ip4_multicast_leave_group(struct net_bridge *br,
1286                                          struct net_bridge_port *port,
1287                                          __be32 group)
1288 {
1289         struct br_ip br_group;
1290
1291         if (ipv4_is_local_multicast(group))
1292                 return;
1293
1294         br_group.u.ip4 = group;
1295         br_group.proto = htons(ETH_P_IP);
1296
1297         br_multicast_leave_group(br, port, &br_group);
1298 }
1299
1300 #if IS_ENABLED(CONFIG_IPV6)
1301 static void br_ip6_multicast_leave_group(struct net_bridge *br,
1302                                          struct net_bridge_port *port,
1303                                          const struct in6_addr *group)
1304 {
1305         struct br_ip br_group;
1306
1307         if (!ipv6_is_transient_multicast(group))
1308                 return;
1309
1310         br_group.u.ip6 = *group;
1311         br_group.proto = htons(ETH_P_IPV6);
1312
1313         br_multicast_leave_group(br, port, &br_group);
1314 }
1315 #endif
1316
1317 static int br_multicast_ipv4_rcv(struct net_bridge *br,
1318                                  struct net_bridge_port *port,
1319                                  struct sk_buff *skb)
1320 {
1321         struct sk_buff *skb2 = skb;
1322         const struct iphdr *iph;
1323         struct igmphdr *ih;
1324         unsigned int len;
1325         unsigned int offset;
1326         int err;
1327
1328         /* We treat OOM as packet loss for now. */
1329         if (!pskb_may_pull(skb, sizeof(*iph)))
1330                 return -EINVAL;
1331
1332         iph = ip_hdr(skb);
1333
1334         if (iph->ihl < 5 || iph->version != 4)
1335                 return -EINVAL;
1336
1337         if (!pskb_may_pull(skb, ip_hdrlen(skb)))
1338                 return -EINVAL;
1339
1340         iph = ip_hdr(skb);
1341
1342         if (unlikely(ip_fast_csum((u8 *)iph, iph->ihl)))
1343                 return -EINVAL;
1344
1345         if (iph->protocol != IPPROTO_IGMP) {
1346                 if ((iph->daddr & IGMP_LOCAL_GROUP_MASK) != IGMP_LOCAL_GROUP)
1347                         BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1348                 return 0;
1349         }
1350
1351         len = ntohs(iph->tot_len);
1352         if (skb->len < len || len < ip_hdrlen(skb))
1353                 return -EINVAL;
1354
1355         if (skb->len > len) {
1356                 skb2 = skb_clone(skb, GFP_ATOMIC);
1357                 if (!skb2)
1358                         return -ENOMEM;
1359
1360                 err = pskb_trim_rcsum(skb2, len);
1361                 if (err)
1362                         goto err_out;
1363         }
1364
1365         len -= ip_hdrlen(skb2);
1366         offset = skb_network_offset(skb2) + ip_hdrlen(skb2);
1367         __skb_pull(skb2, offset);
1368         skb_reset_transport_header(skb2);
1369
1370         err = -EINVAL;
1371         if (!pskb_may_pull(skb2, sizeof(*ih)))
1372                 goto out;
1373
1374         switch (skb2->ip_summed) {
1375         case CHECKSUM_COMPLETE:
1376                 if (!csum_fold(skb2->csum))
1377                         break;
1378                 /* fall through */
1379         case CHECKSUM_NONE:
1380                 skb2->csum = 0;
1381                 if (skb_checksum_complete(skb2))
1382                         goto out;
1383         }
1384
1385         err = 0;
1386
1387         BR_INPUT_SKB_CB(skb)->igmp = 1;
1388         ih = igmp_hdr(skb2);
1389
1390         switch (ih->type) {
1391         case IGMP_HOST_MEMBERSHIP_REPORT:
1392         case IGMPV2_HOST_MEMBERSHIP_REPORT:
1393                 BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1394                 err = br_ip4_multicast_add_group(br, port, ih->group);
1395                 break;
1396         case IGMPV3_HOST_MEMBERSHIP_REPORT:
1397                 err = br_ip4_multicast_igmp3_report(br, port, skb2);
1398                 break;
1399         case IGMP_HOST_MEMBERSHIP_QUERY:
1400                 err = br_ip4_multicast_query(br, port, skb2);
1401                 break;
1402         case IGMP_HOST_LEAVE_MESSAGE:
1403                 br_ip4_multicast_leave_group(br, port, ih->group);
1404                 break;
1405         }
1406
1407 out:
1408         __skb_push(skb2, offset);
1409 err_out:
1410         if (skb2 != skb)
1411                 kfree_skb(skb2);
1412         return err;
1413 }
1414
1415 #if IS_ENABLED(CONFIG_IPV6)
1416 static int br_multicast_ipv6_rcv(struct net_bridge *br,
1417                                  struct net_bridge_port *port,
1418                                  struct sk_buff *skb)
1419 {
1420         struct sk_buff *skb2;
1421         const struct ipv6hdr *ip6h;
1422         u8 icmp6_type;
1423         u8 nexthdr;
1424         __be16 frag_off;
1425         unsigned int len;
1426         int offset;
1427         int err;
1428
1429         if (!pskb_may_pull(skb, sizeof(*ip6h)))
1430                 return -EINVAL;
1431
1432         ip6h = ipv6_hdr(skb);
1433
1434         /*
1435          * We're interested in MLD messages only.
1436          *  - Version is 6
1437          *  - MLD has always Router Alert hop-by-hop option
1438          *  - But we do not support jumbrograms.
1439          */
1440         if (ip6h->version != 6 ||
1441             ip6h->nexthdr != IPPROTO_HOPOPTS ||
1442             ip6h->payload_len == 0)
1443                 return 0;
1444
1445         len = ntohs(ip6h->payload_len) + sizeof(*ip6h);
1446         if (skb->len < len)
1447                 return -EINVAL;
1448
1449         nexthdr = ip6h->nexthdr;
1450         offset = ipv6_skip_exthdr(skb, sizeof(*ip6h), &nexthdr, &frag_off);
1451
1452         if (offset < 0 || nexthdr != IPPROTO_ICMPV6)
1453                 return 0;
1454
1455         /* Okay, we found ICMPv6 header */
1456         skb2 = skb_clone(skb, GFP_ATOMIC);
1457         if (!skb2)
1458                 return -ENOMEM;
1459
1460         err = -EINVAL;
1461         if (!pskb_may_pull(skb2, offset + sizeof(struct icmp6hdr)))
1462                 goto out;
1463
1464         len -= offset - skb_network_offset(skb2);
1465
1466         __skb_pull(skb2, offset);
1467         skb_reset_transport_header(skb2);
1468         skb_postpull_rcsum(skb2, skb_network_header(skb2),
1469                            skb_network_header_len(skb2));
1470
1471         icmp6_type = icmp6_hdr(skb2)->icmp6_type;
1472
1473         switch (icmp6_type) {
1474         case ICMPV6_MGM_QUERY:
1475         case ICMPV6_MGM_REPORT:
1476         case ICMPV6_MGM_REDUCTION:
1477         case ICMPV6_MLD2_REPORT:
1478                 break;
1479         default:
1480                 err = 0;
1481                 goto out;
1482         }
1483
1484         /* Okay, we found MLD message. Check further. */
1485         if (skb2->len > len) {
1486                 err = pskb_trim_rcsum(skb2, len);
1487                 if (err)
1488                         goto out;
1489                 err = -EINVAL;
1490         }
1491
1492         ip6h = ipv6_hdr(skb2);
1493
1494         switch (skb2->ip_summed) {
1495         case CHECKSUM_COMPLETE:
1496                 if (!csum_ipv6_magic(&ip6h->saddr, &ip6h->daddr, skb2->len,
1497                                         IPPROTO_ICMPV6, skb2->csum))
1498                         break;
1499                 /*FALLTHROUGH*/
1500         case CHECKSUM_NONE:
1501                 skb2->csum = ~csum_unfold(csum_ipv6_magic(&ip6h->saddr,
1502                                                         &ip6h->daddr,
1503                                                         skb2->len,
1504                                                         IPPROTO_ICMPV6, 0));
1505                 if (__skb_checksum_complete(skb2))
1506                         goto out;
1507         }
1508
1509         err = 0;
1510
1511         BR_INPUT_SKB_CB(skb)->igmp = 1;
1512
1513         switch (icmp6_type) {
1514         case ICMPV6_MGM_REPORT:
1515             {
1516                 struct mld_msg *mld;
1517                 if (!pskb_may_pull(skb2, sizeof(*mld))) {
1518                         err = -EINVAL;
1519                         goto out;
1520                 }
1521                 mld = (struct mld_msg *)skb_transport_header(skb2);
1522                 BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1523                 err = br_ip6_multicast_add_group(br, port, &mld->mld_mca);
1524                 break;
1525             }
1526         case ICMPV6_MLD2_REPORT:
1527                 err = br_ip6_multicast_mld2_report(br, port, skb2);
1528                 break;
1529         case ICMPV6_MGM_QUERY:
1530                 err = br_ip6_multicast_query(br, port, skb2);
1531                 break;
1532         case ICMPV6_MGM_REDUCTION:
1533             {
1534                 struct mld_msg *mld;
1535                 if (!pskb_may_pull(skb2, sizeof(*mld))) {
1536                         err = -EINVAL;
1537                         goto out;
1538                 }
1539                 mld = (struct mld_msg *)skb_transport_header(skb2);
1540                 br_ip6_multicast_leave_group(br, port, &mld->mld_mca);
1541             }
1542         }
1543
1544 out:
1545         kfree_skb(skb2);
1546         return err;
1547 }
1548 #endif
1549
1550 int br_multicast_rcv(struct net_bridge *br, struct net_bridge_port *port,
1551                      struct sk_buff *skb)
1552 {
1553         BR_INPUT_SKB_CB(skb)->igmp = 0;
1554         BR_INPUT_SKB_CB(skb)->mrouters_only = 0;
1555
1556         if (br->multicast_disabled)
1557                 return 0;
1558
1559         switch (skb->protocol) {
1560         case htons(ETH_P_IP):
1561                 return br_multicast_ipv4_rcv(br, port, skb);
1562 #if IS_ENABLED(CONFIG_IPV6)
1563         case htons(ETH_P_IPV6):
1564                 return br_multicast_ipv6_rcv(br, port, skb);
1565 #endif
1566         }
1567
1568         return 0;
1569 }
1570
1571 static void br_multicast_query_expired(unsigned long data)
1572 {
1573         struct net_bridge *br = (void *)data;
1574
1575         spin_lock(&br->multicast_lock);
1576         if (br->multicast_startup_queries_sent <
1577             br->multicast_startup_query_count)
1578                 br->multicast_startup_queries_sent++;
1579
1580         br_multicast_send_query(br, NULL, br->multicast_startup_queries_sent);
1581
1582         spin_unlock(&br->multicast_lock);
1583 }
1584
1585 void br_multicast_init(struct net_bridge *br)
1586 {
1587         br->hash_elasticity = 4;
1588         br->hash_max = 512;
1589
1590         br->multicast_router = 1;
1591         br->multicast_querier = 0;
1592         br->multicast_last_member_count = 2;
1593         br->multicast_startup_query_count = 2;
1594
1595         br->multicast_last_member_interval = HZ;
1596         br->multicast_query_response_interval = 10 * HZ;
1597         br->multicast_startup_query_interval = 125 * HZ / 4;
1598         br->multicast_query_interval = 125 * HZ;
1599         br->multicast_querier_interval = 255 * HZ;
1600         br->multicast_membership_interval = 260 * HZ;
1601
1602         spin_lock_init(&br->multicast_lock);
1603         setup_timer(&br->multicast_router_timer,
1604                     br_multicast_local_router_expired, 0);
1605         setup_timer(&br->multicast_querier_timer,
1606                     br_multicast_querier_expired, (unsigned long)br);
1607         setup_timer(&br->multicast_query_timer, br_multicast_query_expired,
1608                     (unsigned long)br);
1609         br_mdb_init();
1610 }
1611
1612 void br_multicast_open(struct net_bridge *br)
1613 {
1614         br->multicast_startup_queries_sent = 0;
1615
1616         if (br->multicast_disabled)
1617                 return;
1618
1619         mod_timer(&br->multicast_query_timer, jiffies);
1620 }
1621
1622 void br_multicast_stop(struct net_bridge *br)
1623 {
1624         struct net_bridge_mdb_htable *mdb;
1625         struct net_bridge_mdb_entry *mp;
1626         struct hlist_node *p, *n;
1627         u32 ver;
1628         int i;
1629
1630         del_timer_sync(&br->multicast_router_timer);
1631         del_timer_sync(&br->multicast_querier_timer);
1632         del_timer_sync(&br->multicast_query_timer);
1633
1634         spin_lock_bh(&br->multicast_lock);
1635         mdb = mlock_dereference(br->mdb, br);
1636         if (!mdb)
1637                 goto out;
1638
1639         br->mdb = NULL;
1640
1641         ver = mdb->ver;
1642         for (i = 0; i < mdb->max; i++) {
1643                 hlist_for_each_entry_safe(mp, p, n, &mdb->mhash[i],
1644                                           hlist[ver]) {
1645                         del_timer(&mp->timer);
1646                         call_rcu_bh(&mp->rcu, br_multicast_free_group);
1647                 }
1648         }
1649
1650         if (mdb->old) {
1651                 spin_unlock_bh(&br->multicast_lock);
1652                 rcu_barrier_bh();
1653                 spin_lock_bh(&br->multicast_lock);
1654                 WARN_ON(mdb->old);
1655         }
1656
1657         mdb->old = mdb;
1658         call_rcu_bh(&mdb->rcu, br_mdb_free);
1659
1660 out:
1661         spin_unlock_bh(&br->multicast_lock);
1662 }
1663
1664 int br_multicast_set_router(struct net_bridge *br, unsigned long val)
1665 {
1666         int err = -ENOENT;
1667
1668         spin_lock_bh(&br->multicast_lock);
1669         if (!netif_running(br->dev))
1670                 goto unlock;
1671
1672         switch (val) {
1673         case 0:
1674         case 2:
1675                 del_timer(&br->multicast_router_timer);
1676                 /* fall through */
1677         case 1:
1678                 br->multicast_router = val;
1679                 err = 0;
1680                 break;
1681
1682         default:
1683                 err = -EINVAL;
1684                 break;
1685         }
1686
1687 unlock:
1688         spin_unlock_bh(&br->multicast_lock);
1689
1690         return err;
1691 }
1692
1693 int br_multicast_set_port_router(struct net_bridge_port *p, unsigned long val)
1694 {
1695         struct net_bridge *br = p->br;
1696         int err = -ENOENT;
1697
1698         spin_lock(&br->multicast_lock);
1699         if (!netif_running(br->dev) || p->state == BR_STATE_DISABLED)
1700                 goto unlock;
1701
1702         switch (val) {
1703         case 0:
1704         case 1:
1705         case 2:
1706                 p->multicast_router = val;
1707                 err = 0;
1708
1709                 if (val < 2 && !hlist_unhashed(&p->rlist))
1710                         hlist_del_init_rcu(&p->rlist);
1711
1712                 if (val == 1)
1713                         break;
1714
1715                 del_timer(&p->multicast_router_timer);
1716
1717                 if (val == 0)
1718                         break;
1719
1720                 br_multicast_add_router(br, p);
1721                 break;
1722
1723         default:
1724                 err = -EINVAL;
1725                 break;
1726         }
1727
1728 unlock:
1729         spin_unlock(&br->multicast_lock);
1730
1731         return err;
1732 }
1733
1734 static void br_multicast_start_querier(struct net_bridge *br)
1735 {
1736         struct net_bridge_port *port;
1737
1738         br_multicast_open(br);
1739
1740         list_for_each_entry(port, &br->port_list, list) {
1741                 if (port->state == BR_STATE_DISABLED ||
1742                     port->state == BR_STATE_BLOCKING)
1743                         continue;
1744
1745                 __br_multicast_enable_port(port);
1746         }
1747 }
1748
1749 int br_multicast_toggle(struct net_bridge *br, unsigned long val)
1750 {
1751         int err = 0;
1752         struct net_bridge_mdb_htable *mdb;
1753
1754         spin_lock_bh(&br->multicast_lock);
1755         if (br->multicast_disabled == !val)
1756                 goto unlock;
1757
1758         br->multicast_disabled = !val;
1759         if (br->multicast_disabled)
1760                 goto unlock;
1761
1762         if (!netif_running(br->dev))
1763                 goto unlock;
1764
1765         mdb = mlock_dereference(br->mdb, br);
1766         if (mdb) {
1767                 if (mdb->old) {
1768                         err = -EEXIST;
1769 rollback:
1770                         br->multicast_disabled = !!val;
1771                         goto unlock;
1772                 }
1773
1774                 err = br_mdb_rehash(&br->mdb, mdb->max,
1775                                     br->hash_elasticity);
1776                 if (err)
1777                         goto rollback;
1778         }
1779
1780         br_multicast_start_querier(br);
1781
1782 unlock:
1783         spin_unlock_bh(&br->multicast_lock);
1784
1785         return err;
1786 }
1787
1788 int br_multicast_set_querier(struct net_bridge *br, unsigned long val)
1789 {
1790         val = !!val;
1791
1792         spin_lock_bh(&br->multicast_lock);
1793         if (br->multicast_querier == val)
1794                 goto unlock;
1795
1796         br->multicast_querier = val;
1797         if (val)
1798                 br_multicast_start_querier(br);
1799
1800 unlock:
1801         spin_unlock_bh(&br->multicast_lock);
1802
1803         return 0;
1804 }
1805
1806 int br_multicast_set_hash_max(struct net_bridge *br, unsigned long val)
1807 {
1808         int err = -ENOENT;
1809         u32 old;
1810         struct net_bridge_mdb_htable *mdb;
1811
1812         spin_lock(&br->multicast_lock);
1813         if (!netif_running(br->dev))
1814                 goto unlock;
1815
1816         err = -EINVAL;
1817         if (!is_power_of_2(val))
1818                 goto unlock;
1819
1820         mdb = mlock_dereference(br->mdb, br);
1821         if (mdb && val < mdb->size)
1822                 goto unlock;
1823
1824         err = 0;
1825
1826         old = br->hash_max;
1827         br->hash_max = val;
1828
1829         if (mdb) {
1830                 if (mdb->old) {
1831                         err = -EEXIST;
1832 rollback:
1833                         br->hash_max = old;
1834                         goto unlock;
1835                 }
1836
1837                 err = br_mdb_rehash(&br->mdb, br->hash_max,
1838                                     br->hash_elasticity);
1839                 if (err)
1840                         goto rollback;
1841         }
1842
1843 unlock:
1844         spin_unlock(&br->multicast_lock);
1845
1846         return err;
1847 }