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/addrconf.h>
  31 #include <net/ip6_checksum.h>
  32 #endif
  33
  34 #include "br_private.h"
  35
  36 #define mlock_dereference(X, br) \
  37         rcu_dereference_protected(X, lockdep_is_held(&br->multicast_lock))
  38
  39 static void br_multicast_start_querier(struct net_bridge *br);
  40 unsigned int br_mdb_rehash_seq;
  41
  42 #if IS_ENABLED(CONFIG_IPV6)
  43 static inline int ipv6_is_transient_multicast(const struct in6_addr *addr)
  44 {
  45         if (ipv6_addr_is_multicast(addr) && IPV6_ADDR_MC_FLAG_TRANSIENT(addr))
  46                 return 1;
  47         return 0;
  48 }
  49 #endif
  50
  51 static inline int br_ip_equal(const struct br_ip *a, const struct br_ip *b)
  52 {
  53         if (a->proto != b->proto)
  54                 return 0;
  55         switch (a->proto) {
  56         case htons(ETH_P_IP):
  57                 return a->u.ip4 == b->u.ip4;
  58 #if IS_ENABLED(CONFIG_IPV6)
  59         case htons(ETH_P_IPV6):
  60                 return ipv6_addr_equal(&a->u.ip6, &b->u.ip6);
  61 #endif
  62         }
  63         return 0;
  64 }
  65
  66 static inline int __br_ip4_hash(struct net_bridge_mdb_htable *mdb, __be32 ip)
  67 {
  68         return jhash_1word(mdb->secret, (__force u32)ip) & (mdb->max - 1);
  69 }
  70
  71 #if IS_ENABLED(CONFIG_IPV6)
  72 static inline int __br_ip6_hash(struct net_bridge_mdb_htable *mdb,
  73                                 const struct in6_addr *ip)
  74 {
  75         return jhash2((__force u32 *)ip->s6_addr32, 4, mdb->secret) & (mdb->max - 1);
  76 }
  77 #endif
  78
  79 static inline int br_ip_hash(struct net_bridge_mdb_htable *mdb,
  80                              struct br_ip *ip)
  81 {
  82         switch (ip->proto) {
  83         case htons(ETH_P_IP):
  84                 return __br_ip4_hash(mdb, ip->u.ip4);
  85 #if IS_ENABLED(CONFIG_IPV6)
  86         case htons(ETH_P_IPV6):
  87                 return __br_ip6_hash(mdb, &ip->u.ip6);
  88 #endif
  89         }
  90         return 0;
  91 }
  92
  93 static struct net_bridge_mdb_entry *__br_mdb_ip_get(
  94         struct net_bridge_mdb_htable *mdb, struct br_ip *dst, int hash)
  95 {
  96         struct net_bridge_mdb_entry *mp;
  97         struct hlist_node *p;
  98
  99         hlist_for_each_entry_rcu(mp, p, &mdb->mhash[hash], hlist[mdb->ver]) {
 100                 if (br_ip_equal(&mp->addr, dst))
 101                         return mp;
 102         }
 103
 104         return NULL;
 105 }
 106
 107 static struct net_bridge_mdb_entry *br_mdb_ip_get(
 108         struct net_bridge_mdb_htable *mdb, struct br_ip *dst)
 109 {
 110         if (!mdb)
 111                 return NULL;
 112
 113         return __br_mdb_ip_get(mdb, dst, br_ip_hash(mdb, dst));
 114 }
 115
 116 static struct net_bridge_mdb_entry *br_mdb_ip4_get(
 117         struct net_bridge_mdb_htable *mdb, __be32 dst)
 118 {
 119         struct br_ip br_dst;
 120
 121         br_dst.u.ip4 = dst;
 122         br_dst.proto = htons(ETH_P_IP);
 123
 124         return br_mdb_ip_get(mdb, &br_dst);
 125 }
 126
 127 #if IS_ENABLED(CONFIG_IPV6)
 128 static struct net_bridge_mdb_entry *br_mdb_ip6_get(
 129         struct net_bridge_mdb_htable *mdb, const struct in6_addr *dst)
 130 {
 131         struct br_ip br_dst;
 132
 133         br_dst.u.ip6 = *dst;
 134         br_dst.proto = htons(ETH_P_IPV6);
 135
 136         return br_mdb_ip_get(mdb, &br_dst);
 137 }
 138 #endif
 139
 140 struct net_bridge_mdb_entry *br_mdb_get(struct net_bridge *br,
 141                                         struct sk_buff *skb)
 142 {
 143         struct net_bridge_mdb_htable *mdb = rcu_dereference(br->mdb);
 144         struct br_ip ip;
 145
 146         if (br->multicast_disabled)
 147                 return NULL;
 148
 149         if (BR_INPUT_SKB_CB(skb)->igmp)
 150                 return NULL;
 151
 152         ip.proto = skb->protocol;
 153
 154         switch (skb->protocol) {
 155         case htons(ETH_P_IP):
 156                 ip.u.ip4 = ip_hdr(skb)->daddr;
 157                 break;
 158 #if IS_ENABLED(CONFIG_IPV6)
 159         case htons(ETH_P_IPV6):
 160                 ip.u.ip6 = ipv6_hdr(skb)->daddr;
 161                 break;
 162 #endif
 163         default:
 164                 return NULL;
 165         }
 166
 167         return br_mdb_ip_get(mdb, &ip);
 168 }
 169
 170 static void br_mdb_free(struct rcu_head *head)
 171 {
 172         struct net_bridge_mdb_htable *mdb =
 173                 container_of(head, struct net_bridge_mdb_htable, rcu);
 174         struct net_bridge_mdb_htable *old = mdb->old;
 175
 176         mdb->old = NULL;
 177         kfree(old->mhash);
 178         kfree(old);
 179 }
 180
 181 static int br_mdb_copy(struct net_bridge_mdb_htable *new,
 182                        struct net_bridge_mdb_htable *old,
 183                        int elasticity)
 184 {
 185         struct net_bridge_mdb_entry *mp;
 186         struct hlist_node *p;
 187         int maxlen;
 188         int len;
 189         int i;
 190
 191         for (i = 0; i < old->max; i++)
 192                 hlist_for_each_entry(mp, p, &old->mhash[i], hlist[old->ver])
 193                         hlist_add_head(&mp->hlist[new->ver],
 194                                        &new->mhash[br_ip_hash(new, &mp->addr)]);
 195
 196         if (!elasticity)
 197                 return 0;
 198
 199         maxlen = 0;
 200         for (i = 0; i < new->max; i++) {
 201                 len = 0;
 202                 hlist_for_each_entry(mp, p, &new->mhash[i], hlist[new->ver])
 203                         len++;
 204                 if (len > maxlen)
 205                         maxlen = len;
 206         }
 207
 208         return maxlen > elasticity ? -EINVAL : 0;
 209 }
 210
 211 static void br_multicast_free_pg(struct rcu_head *head)
 212 {
 213         struct net_bridge_port_group *p =
 214                 container_of(head, struct net_bridge_port_group, rcu);
 215
 216         kfree(p);
 217 }
 218
 219 static void br_multicast_free_group(struct rcu_head *head)
 220 {
 221         struct net_bridge_mdb_entry *mp =
 222                 container_of(head, struct net_bridge_mdb_entry, rcu);
 223
 224         kfree(mp);
 225 }
 226
 227 static void br_multicast_group_expired(unsigned long data)
 228 {
 229         struct net_bridge_mdb_entry *mp = (void *)data;
 230         struct net_bridge *br = mp->br;
 231         struct net_bridge_mdb_htable *mdb;
 232
 233         spin_lock(&br->multicast_lock);
 234         if (!netif_running(br->dev) || timer_pending(&mp->timer))
 235                 goto out;
 236
 237         mp->mglist = false;
 238
 239         if (mp->ports)
 240                 goto out;
 241
 242         mdb = mlock_dereference(br->mdb, br);
 243
 244         hlist_del_rcu(&mp->hlist[mdb->ver]);
 245         mdb->size--;
 246
 247         call_rcu_bh(&mp->rcu, br_multicast_free_group);
 248
 249 out:
 250         spin_unlock(&br->multicast_lock);
 251 }
 252
 253 static void br_multicast_del_pg(struct net_bridge *br,
 254                                 struct net_bridge_port_group *pg)
 255 {
 256         struct net_bridge_mdb_htable *mdb;
 257         struct net_bridge_mdb_entry *mp;
 258         struct net_bridge_port_group *p;
 259         struct net_bridge_port_group __rcu **pp;
 260
 261         mdb = mlock_dereference(br->mdb, br);
 262
 263         mp = br_mdb_ip_get(mdb, &pg->addr);
 264         if (WARN_ON(!mp))
 265                 return;
 266
 267         for (pp = &mp->ports;
 268              (p = mlock_dereference(*pp, br)) != NULL;
 269              pp = &p->next) {
 270                 if (p != pg)
 271                         continue;
 272
 273                 rcu_assign_pointer(*pp, p->next);
 274                 hlist_del_init(&p->mglist);
 275                 del_timer(&p->timer);
 276                 call_rcu_bh(&p->rcu, br_multicast_free_pg);
 277
 278                 if (!mp->ports && !mp->mglist &&
 279                     netif_running(br->dev))
 280                         mod_timer(&mp->timer, jiffies);
 281
 282                 return;
 283         }
 284
 285         WARN_ON(1);
 286 }
 287
 288 static void br_multicast_port_group_expired(unsigned long data)
 289 {
 290         struct net_bridge_port_group *pg = (void *)data;
 291         struct net_bridge *br = pg->port->br;
 292
 293         spin_lock(&br->multicast_lock);
 294         if (!netif_running(br->dev) || timer_pending(&pg->timer) ||
 295             hlist_unhashed(&pg->mglist))
 296                 goto out;
 297
 298         br_multicast_del_pg(br, pg);
 299
 300 out:
 301         spin_unlock(&br->multicast_lock);
 302 }
 303
 304 static int br_mdb_rehash(struct net_bridge_mdb_htable __rcu **mdbp, int max,
 305                          int elasticity)
 306 {
 307         struct net_bridge_mdb_htable *old = rcu_dereference_protected(*mdbp, 1);
 308         struct net_bridge_mdb_htable *mdb;
 309         int err;
 310
 311         mdb = kmalloc(sizeof(*mdb), GFP_ATOMIC);
 312         if (!mdb)
 313                 return -ENOMEM;
 314
 315         mdb->max = max;
 316         mdb->old = old;
 317
 318         mdb->mhash = kzalloc(max * sizeof(*mdb->mhash), GFP_ATOMIC);
 319         if (!mdb->mhash) {
 320                 kfree(mdb);
 321                 return -ENOMEM;
 322         }
 323
 324         mdb->size = old ? old->size : 0;
 325         mdb->ver = old ? old->ver ^ 1 : 0;
 326
 327         if (!old || elasticity)
 328                 get_random_bytes(&mdb->secret, sizeof(mdb->secret));
 329         else
 330                 mdb->secret = old->secret;
 331
 332         if (!old)
 333                 goto out;
 334
 335         err = br_mdb_copy(mdb, old, elasticity);
 336         if (err) {
 337                 kfree(mdb->mhash);
 338                 kfree(mdb);
 339                 return err;
 340         }
 341
 342         br_mdb_rehash_seq++;
 343         call_rcu_bh(&mdb->rcu, br_mdb_free);
 344
 345 out:
 346         rcu_assign_pointer(*mdbp, mdb);
 347
 348         return 0;
 349 }
 350
 351 static struct sk_buff *br_ip4_multicast_alloc_query(struct net_bridge *br,
 352                                                     __be32 group)
 353 {
 354         struct sk_buff *skb;
 355         struct igmphdr *ih;
 356         struct ethhdr *eth;
 357         struct iphdr *iph;
 358
 359         skb = netdev_alloc_skb_ip_align(br->dev, sizeof(*eth) + sizeof(*iph) +
 360                                                  sizeof(*ih) + 4);
 361         if (!skb)
 362                 goto out;
 363
 364         skb->protocol = htons(ETH_P_IP);
 365
 366         skb_reset_mac_header(skb);
 367         eth = eth_hdr(skb);
 368
 369         memcpy(eth->h_source, br->dev->dev_addr, 6);
 370         eth->h_dest[0] = 1;
 371         eth->h_dest[1] = 0;
 372         eth->h_dest[2] = 0x5e;
 373         eth->h_dest[3] = 0;
 374         eth->h_dest[4] = 0;
 375         eth->h_dest[5] = 1;
 376         eth->h_proto = htons(ETH_P_IP);
 377         skb_put(skb, sizeof(*eth));
 378
 379         skb_set_network_header(skb, skb->len);
 380         iph = ip_hdr(skb);
 381
 382         iph->version = 4;
 383         iph->ihl = 6;
 384         iph->tos = 0xc0;
 385         iph->tot_len = htons(sizeof(*iph) + sizeof(*ih) + 4);
 386         iph->id = 0;
 387         iph->frag_off = htons(IP_DF);
 388         iph->ttl = 1;
 389         iph->protocol = IPPROTO_IGMP;
 390         iph->saddr = 0;
 391         iph->daddr = htonl(INADDR_ALLHOSTS_GROUP);
 392         ((u8 *)&iph[1])[0] = IPOPT_RA;
 393         ((u8 *)&iph[1])[1] = 4;
 394         ((u8 *)&iph[1])[2] = 0;
 395         ((u8 *)&iph[1])[3] = 0;
 396         ip_send_check(iph);
 397         skb_put(skb, 24);
 398
 399         skb_set_transport_header(skb, skb->len);
 400         ih = igmp_hdr(skb);
 401         ih->type = IGMP_HOST_MEMBERSHIP_QUERY;
 402         ih->code = (group ? br->multicast_last_member_interval :
 403                             br->multicast_query_response_interval) /
 404                    (HZ / IGMP_TIMER_SCALE);
 405         ih->group = group;
 406         ih->csum = 0;
 407         ih->csum = ip_compute_csum((void *)ih, sizeof(struct igmphdr));
 408         skb_put(skb, sizeof(*ih));
 409
 410         __skb_pull(skb, sizeof(*eth));
 411
 412 out:
 413         return skb;
 414 }
 415
 416 #if IS_ENABLED(CONFIG_IPV6)
 417 static struct sk_buff *br_ip6_multicast_alloc_query(struct net_bridge *br,
 418                                                     const struct in6_addr *group)
 419 {
 420         struct sk_buff *skb;
 421         struct ipv6hdr *ip6h;
 422         struct mld_msg *mldq;
 423         struct ethhdr *eth;
 424         u8 *hopopt;
 425         unsigned long interval;
 426
 427         skb = netdev_alloc_skb_ip_align(br->dev, sizeof(*eth) + sizeof(*ip6h) +
 428                                                  8 + sizeof(*mldq));
 429         if (!skb)
 430                 goto out;
 431
 432         skb->protocol = htons(ETH_P_IPV6);
 433
 434         /* Ethernet header */
 435         skb_reset_mac_header(skb);
 436         eth = eth_hdr(skb);
 437
 438         memcpy(eth->h_source, br->dev->dev_addr, 6);
 439         eth->h_proto = htons(ETH_P_IPV6);
 440         skb_put(skb, sizeof(*eth));
 441
 442         /* IPv6 header + HbH option */
 443         skb_set_network_header(skb, skb->len);
 444         ip6h = ipv6_hdr(skb);
 445
 446         *(__force __be32 *)ip6h = htonl(0x60000000);
 447         ip6h->payload_len = htons(8 + sizeof(*mldq));
 448         ip6h->nexthdr = IPPROTO_HOPOPTS;
 449         ip6h->hop_limit = 1;
 450         ipv6_addr_set(&ip6h->daddr, htonl(0xff020000), 0, 0, htonl(1));
 451         if (ipv6_dev_get_saddr(dev_net(br->dev), br->dev, &ip6h->daddr, 0,
 452                                &ip6h->saddr)) {
 453                 kfree_skb(skb);
 454                 return NULL;
 455         }
 456         ipv6_eth_mc_map(&ip6h->daddr, eth->h_dest);
 457
 458         hopopt = (u8 *)(ip6h + 1);
 459         hopopt[0] = IPPROTO_ICMPV6;             /* next hdr */
 460         hopopt[1] = 0;                          /* length of HbH */
 461         hopopt[2] = IPV6_TLV_ROUTERALERT;       /* Router Alert */
 462         hopopt[3] = 2;                          /* Length of RA Option */
 463         hopopt[4] = 0;                          /* Type = 0x0000 (MLD) */
 464         hopopt[5] = 0;
 465         hopopt[6] = IPV6_TLV_PAD1;              /* Pad1 */
 466         hopopt[7] = IPV6_TLV_PAD1;              /* Pad1 */
 467
 468         skb_put(skb, sizeof(*ip6h) + 8);
 469
 470         /* ICMPv6 */
 471         skb_set_transport_header(skb, skb->len);
 472         mldq = (struct mld_msg *) icmp6_hdr(skb);
 473
 474         interval = ipv6_addr_any(group) ? br->multicast_last_member_interval :
 475                                           br->multicast_query_response_interval;
 476
 477         mldq->mld_type = ICMPV6_MGM_QUERY;
 478         mldq->mld_code = 0;
 479         mldq->mld_cksum = 0;
 480         mldq->mld_maxdelay = htons((u16)jiffies_to_msecs(interval));
 481         mldq->mld_reserved = 0;
 482         mldq->mld_mca = *group;
 483
 484         /* checksum */
 485         mldq->mld_cksum = csum_ipv6_magic(&ip6h->saddr, &ip6h->daddr,
 486                                           sizeof(*mldq), IPPROTO_ICMPV6,
 487                                           csum_partial(mldq,
 488                                                        sizeof(*mldq), 0));
 489         skb_put(skb, sizeof(*mldq));
 490
 491         __skb_pull(skb, sizeof(*eth));
 492
 493 out:
 494         return skb;
 495 }
 496 #endif
 497
 498 static struct sk_buff *br_multicast_alloc_query(struct net_bridge *br,
 499                                                 struct br_ip *addr)
 500 {
 501         switch (addr->proto) {
 502         case htons(ETH_P_IP):
 503                 return br_ip4_multicast_alloc_query(br, addr->u.ip4);
 504 #if IS_ENABLED(CONFIG_IPV6)
 505         case htons(ETH_P_IPV6):
 506                 return br_ip6_multicast_alloc_query(br, &addr->u.ip6);
 507 #endif
 508         }
 509         return NULL;
 510 }
 511
 512 static struct net_bridge_mdb_entry *br_multicast_get_group(
 513         struct net_bridge *br, struct net_bridge_port *port,
 514         struct br_ip *group, int hash)
 515 {
 516         struct net_bridge_mdb_htable *mdb;
 517         struct net_bridge_mdb_entry *mp;
 518         struct hlist_node *p;
 519         unsigned int count = 0;
 520         unsigned int max;
 521         int elasticity;
 522         int err;
 523
 524         mdb = rcu_dereference_protected(br->mdb, 1);
 525         hlist_for_each_entry(mp, p, &mdb->mhash[hash], hlist[mdb->ver]) {
 526                 count++;
 527                 if (unlikely(br_ip_equal(group, &mp->addr)))
 528                         return mp;
 529         }
 530
 531         elasticity = 0;
 532         max = mdb->max;
 533
 534         if (unlikely(count > br->hash_elasticity && count)) {
 535                 if (net_ratelimit())
 536                         br_info(br, "Multicast hash table "
 537                                 "chain limit reached: %s\n",
 538                                 port ? port->dev->name : br->dev->name);
 539
 540                 elasticity = br->hash_elasticity;
 541         }
 542
 543         if (mdb->size >= max) {
 544                 max *= 2;
 545                 if (unlikely(max > br->hash_max)) {
 546                         br_warn(br, "Multicast hash table maximum of %d "
 547                                 "reached, disabling snooping: %s\n",
 548                                 br->hash_max,
 549                                 port ? port->dev->name : br->dev->name);
 550                         err = -E2BIG;
 551 disable:
 552                         br->multicast_disabled = 1;
 553                         goto err;
 554                 }
 555         }
 556
 557         if (max > mdb->max || elasticity) {
 558                 if (mdb->old) {
 559                         if (net_ratelimit())
 560                                 br_info(br, "Multicast hash table "
 561                                         "on fire: %s\n",
 562                                         port ? port->dev->name : br->dev->name);
 563                         err = -EEXIST;
 564                         goto err;
 565                 }
 566
 567                 err = br_mdb_rehash(&br->mdb, max, elasticity);
 568                 if (err) {
 569                         br_warn(br, "Cannot rehash multicast "
 570                                 "hash table, disabling snooping: %s, %d, %d\n",
 571                                 port ? port->dev->name : br->dev->name,
 572                                 mdb->size, err);
 573                         goto disable;
 574                 }
 575
 576                 err = -EAGAIN;
 577                 goto err;
 578         }
 579
 580         return NULL;
 581
 582 err:
 583         mp = ERR_PTR(err);
 584         return mp;
 585 }
 586
 587 static struct net_bridge_mdb_entry *br_multicast_new_group(
 588         struct net_bridge *br, struct net_bridge_port *port,
 589         struct br_ip *group)
 590 {
 591         struct net_bridge_mdb_htable *mdb;
 592         struct net_bridge_mdb_entry *mp;
 593         int hash;
 594         int err;
 595
 596         mdb = rcu_dereference_protected(br->mdb, 1);
 597         if (!mdb) {
 598                 err = br_mdb_rehash(&br->mdb, BR_HASH_SIZE, 0);
 599                 if (err)
 600                         return ERR_PTR(err);
 601                 goto rehash;
 602         }
 603
 604         hash = br_ip_hash(mdb, group);
 605         mp = br_multicast_get_group(br, port, group, hash);
 606         switch (PTR_ERR(mp)) {
 607         case 0:
 608                 break;
 609
 610         case -EAGAIN:
 611 rehash:
 612                 mdb = rcu_dereference_protected(br->mdb, 1);
 613                 hash = br_ip_hash(mdb, group);
 614                 break;
 615
 616         default:
 617                 goto out;
 618         }
 619
 620         mp = kzalloc(sizeof(*mp), GFP_ATOMIC);
 621         if (unlikely(!mp))
 622                 return ERR_PTR(-ENOMEM);
 623
 624         mp->br = br;
 625         mp->addr = *group;
 626         setup_timer(&mp->timer, br_multicast_group_expired,
 627                     (unsigned long)mp);
 628
 629         hlist_add_head_rcu(&mp->hlist[mdb->ver], &mdb->mhash[hash]);
 630         mdb->size++;
 631
 632 out:
 633         return mp;
 634 }
 635
 636 static int br_multicast_add_group(struct net_bridge *br,
 637                                   struct net_bridge_port *port,
 638                                   struct br_ip *group)
 639 {
 640         struct net_bridge_mdb_entry *mp;
 641         struct net_bridge_port_group *p;
 642         struct net_bridge_port_group __rcu **pp;
 643         unsigned long now = jiffies;
 644         int err;
 645
 646         spin_lock(&br->multicast_lock);
 647         if (!netif_running(br->dev) ||
 648             (port && port->state == BR_STATE_DISABLED))
 649                 goto out;
 650
 651         mp = br_multicast_new_group(br, port, group);
 652         err = PTR_ERR(mp);
 653         if (IS_ERR(mp))
 654                 goto err;
 655
 656         if (!port) {
 657                 mp->mglist = true;
 658                 mod_timer(&mp->timer, now + br->multicast_membership_interval);
 659                 goto out;
 660         }
 661
 662         for (pp = &mp->ports;
 663              (p = mlock_dereference(*pp, br)) != NULL;
 664              pp = &p->next) {
 665                 if (p->port == port)
 666                         goto found;
 667                 if ((unsigned long)p->port < (unsigned long)port)
 668                         break;
 669         }
 670
 671         p = kzalloc(sizeof(*p), GFP_ATOMIC);
 672         err = -ENOMEM;
 673         if (unlikely(!p))
 674                 goto err;
 675
 676         p->addr = *group;
 677         p->port = port;
 678         p->next = *pp;
 679         hlist_add_head(&p->mglist, &port->mglist);
 680         setup_timer(&p->timer, br_multicast_port_group_expired,
 681                     (unsigned long)p);
 682
 683         rcu_assign_pointer(*pp, p);
 684         br_mdb_notify(br->dev, port, group, RTM_NEWMDB);
 685
 686 found:
 687         mod_timer(&p->timer, now + br->multicast_membership_interval);
 688 out:
 689         err = 0;
 690
 691 err:
 692         spin_unlock(&br->multicast_lock);
 693         return err;
 694 }
 695
 696 static int br_ip4_multicast_add_group(struct net_bridge *br,
 697                                       struct net_bridge_port *port,
 698                                       __be32 group)
 699 {
 700         struct br_ip br_group;
 701
 702         if (ipv4_is_local_multicast(group))
 703                 return 0;
 704
 705         br_group.u.ip4 = group;
 706         br_group.proto = htons(ETH_P_IP);
 707
 708         return br_multicast_add_group(br, port, &br_group);
 709 }
 710
 711 #if IS_ENABLED(CONFIG_IPV6)
 712 static int br_ip6_multicast_add_group(struct net_bridge *br,
 713                                       struct net_bridge_port *port,
 714                                       const struct in6_addr *group)
 715 {
 716         struct br_ip br_group;
 717
 718         if (!ipv6_is_transient_multicast(group))
 719                 return 0;
 720
 721         br_group.u.ip6 = *group;
 722         br_group.proto = htons(ETH_P_IPV6);
 723
 724         return br_multicast_add_group(br, port, &br_group);
 725 }
 726 #endif
 727
 728 static void br_multicast_router_expired(unsigned long data)
 729 {
 730         struct net_bridge_port *port = (void *)data;
 731         struct net_bridge *br = port->br;
 732
 733         spin_lock(&br->multicast_lock);
 734         if (port->multicast_router != 1 ||
 735             timer_pending(&port->multicast_router_timer) ||
 736             hlist_unhashed(&port->rlist))
 737                 goto out;
 738
 739         hlist_del_init_rcu(&port->rlist);
 740
 741 out:
 742         spin_unlock(&br->multicast_lock);
 743 }
 744
 745 static void br_multicast_local_router_expired(unsigned long data)
 746 {
 747 }
 748
 749 static void br_multicast_querier_expired(unsigned long data)
 750 {
 751         struct net_bridge *br = (void *)data;
 752
 753         spin_lock(&br->multicast_lock);
 754         if (!netif_running(br->dev) || br->multicast_disabled)
 755                 goto out;
 756
 757         br_multicast_start_querier(br);
 758
 759 out:
 760         spin_unlock(&br->multicast_lock);
 761 }
 762
 763 static void __br_multicast_send_query(struct net_bridge *br,
 764                                       struct net_bridge_port *port,
 765                                       struct br_ip *ip)
 766 {
 767         struct sk_buff *skb;
 768
 769         skb = br_multicast_alloc_query(br, ip);
 770         if (!skb)
 771                 return;
 772
 773         if (port) {
 774                 __skb_push(skb, sizeof(struct ethhdr));
 775                 skb->dev = port->dev;
 776                 NF_HOOK(NFPROTO_BRIDGE, NF_BR_LOCAL_OUT, skb, NULL, skb->dev,
 777                         dev_queue_xmit);
 778         } else
 779                 netif_rx(skb);
 780 }
 781
 782 static void br_multicast_send_query(struct net_bridge *br,
 783                                     struct net_bridge_port *port, u32 sent)
 784 {
 785         unsigned long time;
 786         struct br_ip br_group;
 787
 788         if (!netif_running(br->dev) || br->multicast_disabled ||
 789             !br->multicast_querier ||
 790             timer_pending(&br->multicast_querier_timer))
 791                 return;
 792
 793         memset(&br_group.u, 0, sizeof(br_group.u));
 794
 795         br_group.proto = htons(ETH_P_IP);
 796         __br_multicast_send_query(br, port, &br_group);
 797
 798 #if IS_ENABLED(CONFIG_IPV6)
 799         br_group.proto = htons(ETH_P_IPV6);
 800         __br_multicast_send_query(br, port, &br_group);
 801 #endif
 802
 803         time = jiffies;
 804         time += sent < br->multicast_startup_query_count ?
 805                 br->multicast_startup_query_interval :
 806                 br->multicast_query_interval;
 807         mod_timer(port ? &port->multicast_query_timer :
 808                          &br->multicast_query_timer, time);
 809 }
 810
 811 static void br_multicast_port_query_expired(unsigned long data)
 812 {
 813         struct net_bridge_port *port = (void *)data;
 814         struct net_bridge *br = port->br;
 815
 816         spin_lock(&br->multicast_lock);
 817         if (port->state == BR_STATE_DISABLED ||
 818             port->state == BR_STATE_BLOCKING)
 819                 goto out;
 820
 821         if (port->multicast_startup_queries_sent <
 822             br->multicast_startup_query_count)
 823                 port->multicast_startup_queries_sent++;
 824
 825         br_multicast_send_query(port->br, port,
 826                                 port->multicast_startup_queries_sent);
 827
 828 out:
 829         spin_unlock(&br->multicast_lock);
 830 }
 831
 832 void br_multicast_add_port(struct net_bridge_port *port)
 833 {
 834         port->multicast_router = 1;
 835
 836         setup_timer(&port->multicast_router_timer, br_multicast_router_expired,
 837                     (unsigned long)port);
 838         setup_timer(&port->multicast_query_timer,
 839                     br_multicast_port_query_expired, (unsigned long)port);
 840 }
 841
 842 void br_multicast_del_port(struct net_bridge_port *port)
 843 {
 844         del_timer_sync(&port->multicast_router_timer);
 845 }
 846
 847 static void __br_multicast_enable_port(struct net_bridge_port *port)
 848 {
 849         port->multicast_startup_queries_sent = 0;
 850
 851         if (try_to_del_timer_sync(&port->multicast_query_timer) >= 0 ||
 852             del_timer(&port->multicast_query_timer))
 853                 mod_timer(&port->multicast_query_timer, jiffies);
 854 }
 855
 856 void br_multicast_enable_port(struct net_bridge_port *port)
 857 {
 858         struct net_bridge *br = port->br;
 859
 860         spin_lock(&br->multicast_lock);
 861         if (br->multicast_disabled || !netif_running(br->dev))
 862                 goto out;
 863
 864         __br_multicast_enable_port(port);
 865
 866 out:
 867         spin_unlock(&br->multicast_lock);
 868 }
 869
 870 void br_multicast_disable_port(struct net_bridge_port *port)
 871 {
 872         struct net_bridge *br = port->br;
 873         struct net_bridge_port_group *pg;
 874         struct hlist_node *p, *n;
 875
 876         spin_lock(&br->multicast_lock);
 877         hlist_for_each_entry_safe(pg, p, n, &port->mglist, mglist)
 878                 br_multicast_del_pg(br, pg);
 879
 880         if (!hlist_unhashed(&port->rlist))
 881                 hlist_del_init_rcu(&port->rlist);
 882         del_timer(&port->multicast_router_timer);
 883         del_timer(&port->multicast_query_timer);
 884         spin_unlock(&br->multicast_lock);
 885 }
 886
 887 static int br_ip4_multicast_igmp3_report(struct net_bridge *br,
 888                                          struct net_bridge_port *port,
 889                                          struct sk_buff *skb)
 890 {
 891         struct igmpv3_report *ih;
 892         struct igmpv3_grec *grec;
 893         int i;
 894         int len;
 895         int num;
 896         int type;
 897         int err = 0;
 898         __be32 group;
 899
 900         if (!pskb_may_pull(skb, sizeof(*ih)))
 901                 return -EINVAL;
 902
 903         ih = igmpv3_report_hdr(skb);
 904         num = ntohs(ih->ngrec);
 905         len = sizeof(*ih);
 906
 907         for (i = 0; i < num; i++) {
 908                 len += sizeof(*grec);
 909                 if (!pskb_may_pull(skb, len))
 910                         return -EINVAL;
 911
 912                 grec = (void *)(skb->data + len - sizeof(*grec));
 913                 group = grec->grec_mca;
 914                 type = grec->grec_type;
 915
 916                 len += ntohs(grec->grec_nsrcs) * 4;
 917                 if (!pskb_may_pull(skb, len))
 918                         return -EINVAL;
 919
 920                 /* We treat this as an IGMPv2 report for now. */
 921                 switch (type) {
 922                 case IGMPV3_MODE_IS_INCLUDE:
 923                 case IGMPV3_MODE_IS_EXCLUDE:
 924                 case IGMPV3_CHANGE_TO_INCLUDE:
 925                 case IGMPV3_CHANGE_TO_EXCLUDE:
 926                 case IGMPV3_ALLOW_NEW_SOURCES:
 927                 case IGMPV3_BLOCK_OLD_SOURCES:
 928                         break;
 929
 930                 default:
 931                         continue;
 932                 }
 933
 934                 err = br_ip4_multicast_add_group(br, port, group);
 935                 if (err)
 936                         break;
 937         }
 938
 939         return err;
 940 }
 941
 942 #if IS_ENABLED(CONFIG_IPV6)
 943 static int br_ip6_multicast_mld2_report(struct net_bridge *br,
 944                                         struct net_bridge_port *port,
 945                                         struct sk_buff *skb)
 946 {
 947         struct icmp6hdr *icmp6h;
 948         struct mld2_grec *grec;
 949         int i;
 950         int len;
 951         int num;
 952         int err = 0;
 953
 954         if (!pskb_may_pull(skb, sizeof(*icmp6h)))
 955                 return -EINVAL;
 956
 957         icmp6h = icmp6_hdr(skb);
 958         num = ntohs(icmp6h->icmp6_dataun.un_data16[1]);
 959         len = sizeof(*icmp6h);
 960
 961         for (i = 0; i < num; i++) {
 962                 __be16 *nsrcs, _nsrcs;
 963
 964                 nsrcs = skb_header_pointer(skb,
 965                                            len + offsetof(struct mld2_grec,
 966                                                           grec_nsrcs),
 967                                            sizeof(_nsrcs), &_nsrcs);
 968                 if (!nsrcs)
 969                         return -EINVAL;
 970
 971                 if (!pskb_may_pull(skb,
 972                                    len + sizeof(*grec) +
 973                                    sizeof(struct in6_addr) * ntohs(*nsrcs)))
 974                         return -EINVAL;
 975
 976                 grec = (struct mld2_grec *)(skb->data + len);
 977                 len += sizeof(*grec) +
 978                        sizeof(struct in6_addr) * ntohs(*nsrcs);
 979
 980                 /* We treat these as MLDv1 reports for now. */
 981                 switch (grec->grec_type) {
 982                 case MLD2_MODE_IS_INCLUDE:
 983                 case MLD2_MODE_IS_EXCLUDE:
 984                 case MLD2_CHANGE_TO_INCLUDE:
 985                 case MLD2_CHANGE_TO_EXCLUDE:
 986                 case MLD2_ALLOW_NEW_SOURCES:
 987                 case MLD2_BLOCK_OLD_SOURCES:
 988                         break;
 989
 990                 default:
 991                         continue;
 992                 }
 993
 994                 err = br_ip6_multicast_add_group(br, port, &grec->grec_mca);
 995                 if (!err)
 996                         break;
 997         }
 998
 999         return err;
1000 }
1001 #endif
1002
1003 /*
1004  * Add port to rotuer_list
1005  *  list is maintained ordered by pointer value
1006  *  and locked by br->multicast_lock and RCU
1007  */
1008 static void br_multicast_add_router(struct net_bridge *br,
1009                                     struct net_bridge_port *port)
1010 {
1011         struct net_bridge_port *p;
1012         struct hlist_node *n, *slot = NULL;
1013
1014         hlist_for_each_entry(p, n, &br->router_list, rlist) {
1015                 if ((unsigned long) port >= (unsigned long) p)
1016                         break;
1017                 slot = n;
1018         }
1019
1020         if (slot)
1021                 hlist_add_after_rcu(slot, &port->rlist);
1022         else
1023                 hlist_add_head_rcu(&port->rlist, &br->router_list);
1024 }
1025
1026 static void br_multicast_mark_router(struct net_bridge *br,
1027                                      struct net_bridge_port *port)
1028 {
1029         unsigned long now = jiffies;
1030
1031         if (!port) {
1032                 if (br->multicast_router == 1)
1033                         mod_timer(&br->multicast_router_timer,
1034                                   now + br->multicast_querier_interval);
1035                 return;
1036         }
1037
1038         if (port->multicast_router != 1)
1039                 return;
1040
1041         if (!hlist_unhashed(&port->rlist))
1042                 goto timer;
1043
1044         br_multicast_add_router(br, port);
1045
1046 timer:
1047         mod_timer(&port->multicast_router_timer,
1048                   now + br->multicast_querier_interval);
1049 }
1050
1051 static void br_multicast_query_received(struct net_bridge *br,
1052                                         struct net_bridge_port *port,
1053                                         int saddr)
1054 {
1055         if (saddr)
1056                 mod_timer(&br->multicast_querier_timer,
1057                           jiffies + br->multicast_querier_interval);
1058         else if (timer_pending(&br->multicast_querier_timer))
1059                 return;
1060
1061         br_multicast_mark_router(br, port);
1062 }
1063
1064 static int br_ip4_multicast_query(struct net_bridge *br,
1065                                   struct net_bridge_port *port,
1066                                   struct sk_buff *skb)
1067 {
1068         const struct iphdr *iph = ip_hdr(skb);
1069         struct igmphdr *ih = igmp_hdr(skb);
1070         struct net_bridge_mdb_entry *mp;
1071         struct igmpv3_query *ih3;
1072         struct net_bridge_port_group *p;
1073         struct net_bridge_port_group __rcu **pp;
1074         unsigned long max_delay;
1075         unsigned long now = jiffies;
1076         __be32 group;
1077         int err = 0;
1078
1079         spin_lock(&br->multicast_lock);
1080         if (!netif_running(br->dev) ||
1081             (port && port->state == BR_STATE_DISABLED))
1082                 goto out;
1083
1084         br_multicast_query_received(br, port, !!iph->saddr);
1085
1086         group = ih->group;
1087
1088         if (skb->len == sizeof(*ih)) {
1089                 max_delay = ih->code * (HZ / IGMP_TIMER_SCALE);
1090
1091                 if (!max_delay) {
1092                         max_delay = 10 * HZ;
1093                         group = 0;
1094                 }
1095         } else {
1096                 if (!pskb_may_pull(skb, sizeof(struct igmpv3_query))) {
1097                         err = -EINVAL;
1098                         goto out;
1099                 }
1100
1101                 ih3 = igmpv3_query_hdr(skb);
1102                 if (ih3->nsrcs)
1103                         goto out;
1104
1105                 max_delay = ih3->code ?
1106                             IGMPV3_MRC(ih3->code) * (HZ / IGMP_TIMER_SCALE) : 1;
1107         }
1108
1109         if (!group)
1110                 goto out;
1111
1112         mp = br_mdb_ip4_get(mlock_dereference(br->mdb, br), group);
1113         if (!mp)
1114                 goto out;
1115
1116         max_delay *= br->multicast_last_member_count;
1117
1118         if (mp->mglist &&
1119             (timer_pending(&mp->timer) ?
1120              time_after(mp->timer.expires, now + max_delay) :
1121              try_to_del_timer_sync(&mp->timer) >= 0))
1122                 mod_timer(&mp->timer, now + max_delay);
1123
1124         for (pp = &mp->ports;
1125              (p = mlock_dereference(*pp, br)) != NULL;
1126              pp = &p->next) {
1127                 if (timer_pending(&p->timer) ?
1128                     time_after(p->timer.expires, now + max_delay) :
1129                     try_to_del_timer_sync(&p->timer) >= 0)
1130                         mod_timer(&p->timer, now + max_delay);
1131         }
1132
1133 out:
1134         spin_unlock(&br->multicast_lock);
1135         return err;
1136 }
1137
1138 #if IS_ENABLED(CONFIG_IPV6)
1139 static int br_ip6_multicast_query(struct net_bridge *br,
1140                                   struct net_bridge_port *port,
1141                                   struct sk_buff *skb)
1142 {
1143         const struct ipv6hdr *ip6h = ipv6_hdr(skb);
1144         struct mld_msg *mld = (struct mld_msg *) icmp6_hdr(skb);
1145         struct net_bridge_mdb_entry *mp;
1146         struct mld2_query *mld2q;
1147         struct net_bridge_port_group *p;
1148         struct net_bridge_port_group __rcu **pp;
1149         unsigned long max_delay;
1150         unsigned long now = jiffies;
1151         const struct in6_addr *group = NULL;
1152         int err = 0;
1153
1154         spin_lock(&br->multicast_lock);
1155         if (!netif_running(br->dev) ||
1156             (port && port->state == BR_STATE_DISABLED))
1157                 goto out;
1158
1159         br_multicast_query_received(br, port, !ipv6_addr_any(&ip6h->saddr));
1160
1161         if (skb->len == sizeof(*mld)) {
1162                 if (!pskb_may_pull(skb, sizeof(*mld))) {
1163                         err = -EINVAL;
1164                         goto out;
1165                 }
1166                 mld = (struct mld_msg *) icmp6_hdr(skb);
1167                 max_delay = msecs_to_jiffies(ntohs(mld->mld_maxdelay));
1168                 if (max_delay)
1169                         group = &mld->mld_mca;
1170         } else if (skb->len >= sizeof(*mld2q)) {
1171                 if (!pskb_may_pull(skb, sizeof(*mld2q))) {
1172                         err = -EINVAL;
1173                         goto out;
1174                 }
1175                 mld2q = (struct mld2_query *)icmp6_hdr(skb);
1176                 if (!mld2q->mld2q_nsrcs)
1177                         group = &mld2q->mld2q_mca;
1178                 max_delay = mld2q->mld2q_mrc ? MLDV2_MRC(mld2q->mld2q_mrc) : 1;
1179         }
1180
1181         if (!group)
1182                 goto out;
1183
1184         mp = br_mdb_ip6_get(mlock_dereference(br->mdb, br), group);
1185         if (!mp)
1186                 goto out;
1187
1188         max_delay *= br->multicast_last_member_count;
1189         if (mp->mglist &&
1190             (timer_pending(&mp->timer) ?
1191              time_after(mp->timer.expires, now + max_delay) :
1192              try_to_del_timer_sync(&mp->timer) >= 0))
1193                 mod_timer(&mp->timer, now + max_delay);
1194
1195         for (pp = &mp->ports;
1196              (p = mlock_dereference(*pp, br)) != NULL;
1197              pp = &p->next) {
1198                 if (timer_pending(&p->timer) ?
1199                     time_after(p->timer.expires, now + max_delay) :
1200                     try_to_del_timer_sync(&p->timer) >= 0)
1201                         mod_timer(&p->timer, now + max_delay);
1202         }
1203
1204 out:
1205         spin_unlock(&br->multicast_lock);
1206         return err;
1207 }
1208 #endif
1209
1210 static void br_multicast_leave_group(struct net_bridge *br,
1211                                      struct net_bridge_port *port,
1212                                      struct br_ip *group)
1213 {
1214         struct net_bridge_mdb_htable *mdb;
1215         struct net_bridge_mdb_entry *mp;
1216         struct net_bridge_port_group *p;
1217         unsigned long now;
1218         unsigned long time;
1219
1220         spin_lock(&br->multicast_lock);
1221         if (!netif_running(br->dev) ||
1222             (port && port->state == BR_STATE_DISABLED) ||
1223             timer_pending(&br->multicast_querier_timer))
1224                 goto out;
1225
1226         mdb = mlock_dereference(br->mdb, br);
1227         mp = br_mdb_ip_get(mdb, group);
1228         if (!mp)
1229                 goto out;
1230
1231         if (port && (port->flags & BR_MULTICAST_FAST_LEAVE)) {
1232                 struct net_bridge_port_group __rcu **pp;
1233
1234                 for (pp = &mp->ports;
1235                      (p = mlock_dereference(*pp, br)) != NULL;
1236                      pp = &p->next) {
1237                         if (p->port != port)
1238                                 continue;
1239
1240                         rcu_assign_pointer(*pp, p->next);
1241                         hlist_del_init(&p->mglist);
1242                         del_timer(&p->timer);
1243                         call_rcu_bh(&p->rcu, br_multicast_free_pg);
1244                         br_mdb_notify(br->dev, port, group, RTM_DELMDB);
1245
1246                         if (!mp->ports && !mp->mglist &&
1247                             netif_running(br->dev))
1248                                 mod_timer(&mp->timer, jiffies);
1249                 }
1250                 goto out;
1251         }
1252
1253         now = jiffies;
1254         time = now + br->multicast_last_member_count *
1255                      br->multicast_last_member_interval;
1256
1257         if (!port) {
1258                 if (mp->mglist &&
1259                     (timer_pending(&mp->timer) ?
1260                      time_after(mp->timer.expires, time) :
1261                      try_to_del_timer_sync(&mp->timer) >= 0)) {
1262                         mod_timer(&mp->timer, time);
1263                 }
1264
1265                 goto out;
1266         }
1267
1268         for (p = mlock_dereference(mp->ports, br);
1269              p != NULL;
1270              p = mlock_dereference(p->next, br)) {
1271                 if (p->port != port)
1272                         continue;
1273
1274                 if (!hlist_unhashed(&p->mglist) &&
1275                     (timer_pending(&p->timer) ?
1276                      time_after(p->timer.expires, time) :
1277                      try_to_del_timer_sync(&p->timer) >= 0)) {
1278                         mod_timer(&p->timer, time);
1279                 }
1280
1281                 break;
1282         }
1283
1284 out:
1285         spin_unlock(&br->multicast_lock);
1286 }
1287
1288 static void br_ip4_multicast_leave_group(struct net_bridge *br,
1289                                          struct net_bridge_port *port,
1290                                          __be32 group)
1291 {
1292         struct br_ip br_group;
1293
1294         if (ipv4_is_local_multicast(group))
1295                 return;
1296
1297         br_group.u.ip4 = group;
1298         br_group.proto = htons(ETH_P_IP);
1299
1300         br_multicast_leave_group(br, port, &br_group);
1301 }
1302
1303 #if IS_ENABLED(CONFIG_IPV6)
1304 static void br_ip6_multicast_leave_group(struct net_bridge *br,
1305                                          struct net_bridge_port *port,
1306                                          const struct in6_addr *group)
1307 {
1308         struct br_ip br_group;
1309
1310         if (!ipv6_is_transient_multicast(group))
1311                 return;
1312
1313         br_group.u.ip6 = *group;
1314         br_group.proto = htons(ETH_P_IPV6);
1315
1316         br_multicast_leave_group(br, port, &br_group);
1317 }
1318 #endif
1319
1320 static int br_multicast_ipv4_rcv(struct net_bridge *br,
1321                                  struct net_bridge_port *port,
1322                                  struct sk_buff *skb)
1323 {
1324         struct sk_buff *skb2 = skb;
1325         const struct iphdr *iph;
1326         struct igmphdr *ih;
1327         unsigned int len;
1328         unsigned int offset;
1329         int err;
1330
1331         /* We treat OOM as packet loss for now. */
1332         if (!pskb_may_pull(skb, sizeof(*iph)))
1333                 return -EINVAL;
1334
1335         iph = ip_hdr(skb);
1336
1337         if (iph->ihl < 5 || iph->version != 4)
1338                 return -EINVAL;
1339
1340         if (!pskb_may_pull(skb, ip_hdrlen(skb)))
1341                 return -EINVAL;
1342
1343         iph = ip_hdr(skb);
1344
1345         if (unlikely(ip_fast_csum((u8 *)iph, iph->ihl)))
1346                 return -EINVAL;
1347
1348         if (iph->protocol != IPPROTO_IGMP) {
1349                 if ((iph->daddr & IGMP_LOCAL_GROUP_MASK) != IGMP_LOCAL_GROUP)
1350                         BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1351                 return 0;
1352         }
1353
1354         len = ntohs(iph->tot_len);
1355         if (skb->len < len || len < ip_hdrlen(skb))
1356                 return -EINVAL;
1357
1358         if (skb->len > len) {
1359                 skb2 = skb_clone(skb, GFP_ATOMIC);
1360                 if (!skb2)
1361                         return -ENOMEM;
1362
1363                 err = pskb_trim_rcsum(skb2, len);
1364                 if (err)
1365                         goto err_out;
1366         }
1367
1368         len -= ip_hdrlen(skb2);
1369         offset = skb_network_offset(skb2) + ip_hdrlen(skb2);
1370         __skb_pull(skb2, offset);
1371         skb_reset_transport_header(skb2);
1372
1373         err = -EINVAL;
1374         if (!pskb_may_pull(skb2, sizeof(*ih)))
1375                 goto out;
1376
1377         switch (skb2->ip_summed) {
1378         case CHECKSUM_COMPLETE:
1379                 if (!csum_fold(skb2->csum))
1380                         break;
1381                 /* fall through */
1382         case CHECKSUM_NONE:
1383                 skb2->csum = 0;
1384                 if (skb_checksum_complete(skb2))
1385                         goto out;
1386         }
1387
1388         err = 0;
1389
1390         BR_INPUT_SKB_CB(skb)->igmp = 1;
1391         ih = igmp_hdr(skb2);
1392
1393         switch (ih->type) {
1394         case IGMP_HOST_MEMBERSHIP_REPORT:
1395         case IGMPV2_HOST_MEMBERSHIP_REPORT:
1396                 BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1397                 err = br_ip4_multicast_add_group(br, port, ih->group);
1398                 break;
1399         case IGMPV3_HOST_MEMBERSHIP_REPORT:
1400                 err = br_ip4_multicast_igmp3_report(br, port, skb2);
1401                 break;
1402         case IGMP_HOST_MEMBERSHIP_QUERY:
1403                 err = br_ip4_multicast_query(br, port, skb2);
1404                 break;
1405         case IGMP_HOST_LEAVE_MESSAGE:
1406                 br_ip4_multicast_leave_group(br, port, ih->group);
1407                 break;
1408         }
1409
1410 out:
1411         __skb_push(skb2, offset);
1412 err_out:
1413         if (skb2 != skb)
1414                 kfree_skb(skb2);
1415         return err;
1416 }
1417
1418 #if IS_ENABLED(CONFIG_IPV6)
1419 static int br_multicast_ipv6_rcv(struct net_bridge *br,
1420                                  struct net_bridge_port *port,
1421                                  struct sk_buff *skb)
1422 {
1423         struct sk_buff *skb2;
1424         const struct ipv6hdr *ip6h;
1425         u8 icmp6_type;
1426         u8 nexthdr;
1427         __be16 frag_off;
1428         unsigned int len;
1429         int offset;
1430         int err;
1431
1432         if (!pskb_may_pull(skb, sizeof(*ip6h)))
1433                 return -EINVAL;
1434
1435         ip6h = ipv6_hdr(skb);
1436
1437         /*
1438          * We're interested in MLD messages only.
1439          *  - Version is 6
1440          *  - MLD has always Router Alert hop-by-hop option
1441          *  - But we do not support jumbrograms.
1442          */
1443         if (ip6h->version != 6 ||
1444             ip6h->nexthdr != IPPROTO_HOPOPTS ||
1445             ip6h->payload_len == 0)
1446                 return 0;
1447
1448         len = ntohs(ip6h->payload_len) + sizeof(*ip6h);
1449         if (skb->len < len)
1450                 return -EINVAL;
1451
1452         nexthdr = ip6h->nexthdr;
1453         offset = ipv6_skip_exthdr(skb, sizeof(*ip6h), &nexthdr, &frag_off);
1454
1455         if (offset < 0 || nexthdr != IPPROTO_ICMPV6)
1456                 return 0;
1457
1458         /* Okay, we found ICMPv6 header */
1459         skb2 = skb_clone(skb, GFP_ATOMIC);
1460         if (!skb2)
1461                 return -ENOMEM;
1462
1463         err = -EINVAL;
1464         if (!pskb_may_pull(skb2, offset + sizeof(struct icmp6hdr)))
1465                 goto out;
1466
1467         len -= offset - skb_network_offset(skb2);
1468
1469         __skb_pull(skb2, offset);
1470         skb_reset_transport_header(skb2);
1471         skb_postpull_rcsum(skb2, skb_network_header(skb2),
1472                            skb_network_header_len(skb2));
1473
1474         icmp6_type = icmp6_hdr(skb2)->icmp6_type;
1475
1476         switch (icmp6_type) {
1477         case ICMPV6_MGM_QUERY:
1478         case ICMPV6_MGM_REPORT:
1479         case ICMPV6_MGM_REDUCTION:
1480         case ICMPV6_MLD2_REPORT:
1481                 break;
1482         default:
1483                 err = 0;
1484                 goto out;
1485         }
1486
1487         /* Okay, we found MLD message. Check further. */
1488         if (skb2->len > len) {
1489                 err = pskb_trim_rcsum(skb2, len);
1490                 if (err)
1491                         goto out;
1492                 err = -EINVAL;
1493         }
1494
1495         ip6h = ipv6_hdr(skb2);
1496
1497         switch (skb2->ip_summed) {
1498         case CHECKSUM_COMPLETE:
1499                 if (!csum_ipv6_magic(&ip6h->saddr, &ip6h->daddr, skb2->len,
1500                                         IPPROTO_ICMPV6, skb2->csum))
1501                         break;
1502                 /*FALLTHROUGH*/
1503         case CHECKSUM_NONE:
1504                 skb2->csum = ~csum_unfold(csum_ipv6_magic(&ip6h->saddr,
1505                                                         &ip6h->daddr,
1506                                                         skb2->len,
1507                                                         IPPROTO_ICMPV6, 0));
1508                 if (__skb_checksum_complete(skb2))
1509                         goto out;
1510         }
1511
1512         err = 0;
1513
1514         BR_INPUT_SKB_CB(skb)->igmp = 1;
1515
1516         switch (icmp6_type) {
1517         case ICMPV6_MGM_REPORT:
1518             {
1519                 struct mld_msg *mld;
1520                 if (!pskb_may_pull(skb2, sizeof(*mld))) {
1521                         err = -EINVAL;
1522                         goto out;
1523                 }
1524                 mld = (struct mld_msg *)skb_transport_header(skb2);
1525                 BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1526                 err = br_ip6_multicast_add_group(br, port, &mld->mld_mca);
1527                 break;
1528             }
1529         case ICMPV6_MLD2_REPORT:
1530                 err = br_ip6_multicast_mld2_report(br, port, skb2);
1531                 break;
1532         case ICMPV6_MGM_QUERY:
1533                 err = br_ip6_multicast_query(br, port, skb2);
1534                 break;
1535         case ICMPV6_MGM_REDUCTION:
1536             {
1537                 struct mld_msg *mld;
1538                 if (!pskb_may_pull(skb2, sizeof(*mld))) {
1539                         err = -EINVAL;
1540                         goto out;
1541                 }
1542                 mld = (struct mld_msg *)skb_transport_header(skb2);
1543                 br_ip6_multicast_leave_group(br, port, &mld->mld_mca);
1544             }
1545         }
1546
1547 out:
1548         kfree_skb(skb2);
1549         return err;
1550 }
1551 #endif
1552
1553 int br_multicast_rcv(struct net_bridge *br, struct net_bridge_port *port,
1554                      struct sk_buff *skb)
1555 {
1556         BR_INPUT_SKB_CB(skb)->igmp = 0;
1557         BR_INPUT_SKB_CB(skb)->mrouters_only = 0;
1558
1559         if (br->multicast_disabled)
1560                 return 0;
1561
1562         switch (skb->protocol) {
1563         case htons(ETH_P_IP):
1564                 return br_multicast_ipv4_rcv(br, port, skb);
1565 #if IS_ENABLED(CONFIG_IPV6)
1566         case htons(ETH_P_IPV6):
1567                 return br_multicast_ipv6_rcv(br, port, skb);
1568 #endif
1569         }
1570
1571         return 0;
1572 }
1573
1574 static void br_multicast_query_expired(unsigned long data)
1575 {
1576         struct net_bridge *br = (void *)data;
1577
1578         spin_lock(&br->multicast_lock);
1579         if (br->multicast_startup_queries_sent <
1580             br->multicast_startup_query_count)
1581                 br->multicast_startup_queries_sent++;
1582
1583         br_multicast_send_query(br, NULL, br->multicast_startup_queries_sent);
1584
1585         spin_unlock(&br->multicast_lock);
1586 }
1587
1588 void br_multicast_init(struct net_bridge *br)
1589 {
1590         br->hash_elasticity = 4;
1591         br->hash_max = 512;
1592
1593         br->multicast_router = 1;
1594         br->multicast_querier = 0;
1595         br->multicast_last_member_count = 2;
1596         br->multicast_startup_query_count = 2;
1597
1598         br->multicast_last_member_interval = HZ;
1599         br->multicast_query_response_interval = 10 * HZ;
1600         br->multicast_startup_query_interval = 125 * HZ / 4;
1601         br->multicast_query_interval = 125 * HZ;
1602         br->multicast_querier_interval = 255 * HZ;
1603         br->multicast_membership_interval = 260 * HZ;
1604
1605         spin_lock_init(&br->multicast_lock);
1606         setup_timer(&br->multicast_router_timer,
1607                     br_multicast_local_router_expired, 0);
1608         setup_timer(&br->multicast_querier_timer,
1609                     br_multicast_querier_expired, (unsigned long)br);
1610         setup_timer(&br->multicast_query_timer, br_multicast_query_expired,
1611                     (unsigned long)br);
1612         br_mdb_init();
1613 }
1614
1615 void br_multicast_open(struct net_bridge *br)
1616 {
1617         br->multicast_startup_queries_sent = 0;
1618
1619         if (br->multicast_disabled)
1620                 return;
1621
1622         mod_timer(&br->multicast_query_timer, jiffies);
1623 }
1624
1625 void br_multicast_stop(struct net_bridge *br)
1626 {
1627         struct net_bridge_mdb_htable *mdb;
1628         struct net_bridge_mdb_entry *mp;
1629         struct hlist_node *p, *n;
1630         u32 ver;
1631         int i;
1632
1633         del_timer_sync(&br->multicast_router_timer);
1634         del_timer_sync(&br->multicast_querier_timer);
1635         del_timer_sync(&br->multicast_query_timer);
1636
1637         spin_lock_bh(&br->multicast_lock);
1638         mdb = mlock_dereference(br->mdb, br);
1639         if (!mdb)
1640                 goto out;
1641
1642         br->mdb = NULL;
1643
1644         ver = mdb->ver;
1645         for (i = 0; i < mdb->max; i++) {
1646                 hlist_for_each_entry_safe(mp, p, n, &mdb->mhash[i],
1647                                           hlist[ver]) {
1648                         del_timer(&mp->timer);
1649                         call_rcu_bh(&mp->rcu, br_multicast_free_group);
1650                 }
1651         }
1652
1653         if (mdb->old) {
1654                 spin_unlock_bh(&br->multicast_lock);
1655                 rcu_barrier_bh();
1656                 spin_lock_bh(&br->multicast_lock);
1657                 WARN_ON(mdb->old);
1658         }
1659
1660         mdb->old = mdb;
1661         call_rcu_bh(&mdb->rcu, br_mdb_free);
1662
1663 out:
1664         spin_unlock_bh(&br->multicast_lock);
1665 }
1666
1667 int br_multicast_set_router(struct net_bridge *br, unsigned long val)
1668 {
1669         int err = -ENOENT;
1670
1671         spin_lock_bh(&br->multicast_lock);
1672         if (!netif_running(br->dev))
1673                 goto unlock;
1674
1675         switch (val) {
1676         case 0:
1677         case 2:
1678                 del_timer(&br->multicast_router_timer);
1679                 /* fall through */
1680         case 1:
1681                 br->multicast_router = val;
1682                 err = 0;
1683                 break;
1684
1685         default:
1686                 err = -EINVAL;
1687                 break;
1688         }
1689
1690 unlock:
1691         spin_unlock_bh(&br->multicast_lock);
1692
1693         return err;
1694 }
1695
1696 int br_multicast_set_port_router(struct net_bridge_port *p, unsigned long val)
1697 {
1698         struct net_bridge *br = p->br;
1699         int err = -ENOENT;
1700
1701         spin_lock(&br->multicast_lock);
1702         if (!netif_running(br->dev) || p->state == BR_STATE_DISABLED)
1703                 goto unlock;
1704
1705         switch (val) {
1706         case 0:
1707         case 1:
1708         case 2:
1709                 p->multicast_router = val;
1710                 err = 0;
1711
1712                 if (val < 2 && !hlist_unhashed(&p->rlist))
1713                         hlist_del_init_rcu(&p->rlist);
1714
1715                 if (val == 1)
1716                         break;
1717
1718                 del_timer(&p->multicast_router_timer);
1719
1720                 if (val == 0)
1721                         break;
1722
1723                 br_multicast_add_router(br, p);
1724                 break;
1725
1726         default:
1727                 err = -EINVAL;
1728                 break;
1729         }
1730
1731 unlock:
1732         spin_unlock(&br->multicast_lock);
1733
1734         return err;
1735 }
1736
1737 static void br_multicast_start_querier(struct net_bridge *br)
1738 {
1739         struct net_bridge_port *port;
1740
1741         br_multicast_open(br);
1742
1743         list_for_each_entry(port, &br->port_list, list) {
1744                 if (port->state == BR_STATE_DISABLED ||
1745                     port->state == BR_STATE_BLOCKING)
1746                         continue;
1747
1748                 __br_multicast_enable_port(port);
1749         }
1750 }
1751
1752 int br_multicast_toggle(struct net_bridge *br, unsigned long val)
1753 {
1754         int err = 0;
1755         struct net_bridge_mdb_htable *mdb;
1756
1757         spin_lock_bh(&br->multicast_lock);
1758         if (br->multicast_disabled == !val)
1759                 goto unlock;
1760
1761         br->multicast_disabled = !val;
1762         if (br->multicast_disabled)
1763                 goto unlock;
1764
1765         if (!netif_running(br->dev))
1766                 goto unlock;
1767
1768         mdb = mlock_dereference(br->mdb, br);
1769         if (mdb) {
1770                 if (mdb->old) {
1771                         err = -EEXIST;
1772 rollback:
1773                         br->multicast_disabled = !!val;
1774                         goto unlock;
1775                 }
1776
1777                 err = br_mdb_rehash(&br->mdb, mdb->max,
1778                                     br->hash_elasticity);
1779                 if (err)
1780                         goto rollback;
1781         }
1782
1783         br_multicast_start_querier(br);
1784
1785 unlock:
1786         spin_unlock_bh(&br->multicast_lock);
1787
1788         return err;
1789 }
1790
1791 int br_multicast_set_querier(struct net_bridge *br, unsigned long val)
1792 {
1793         val = !!val;
1794
1795         spin_lock_bh(&br->multicast_lock);
1796         if (br->multicast_querier == val)
1797                 goto unlock;
1798
1799         br->multicast_querier = val;
1800         if (val)
1801                 br_multicast_start_querier(br);
1802
1803 unlock:
1804         spin_unlock_bh(&br->multicast_lock);
1805
1806         return 0;
1807 }
1808
1809 int br_multicast_set_hash_max(struct net_bridge *br, unsigned long val)
1810 {
1811         int err = -ENOENT;
1812         u32 old;
1813         struct net_bridge_mdb_htable *mdb;
1814
1815         spin_lock(&br->multicast_lock);
1816         if (!netif_running(br->dev))
1817                 goto unlock;
1818
1819         err = -EINVAL;
1820         if (!is_power_of_2(val))
1821                 goto unlock;
1822
1823         mdb = mlock_dereference(br->mdb, br);
1824         if (mdb && val < mdb->size)
1825                 goto unlock;
1826
1827         err = 0;
1828
1829         old = br->hash_max;
1830         br->hash_max = val;
1831
1832         if (mdb) {
1833                 if (mdb->old) {
1834                         err = -EEXIST;
1835 rollback:
1836                         br->hash_max = old;
1837                         goto unlock;
1838                 }
1839
1840                 err = br_mdb_rehash(&br->mdb, br->hash_max,
1841                                     br->hash_elasticity);
1842                 if (err)
1843                         goto rollback;
1844         }
1845
1846 unlock:
1847         spin_unlock(&br->multicast_lock);
1848
1849         return err;
1850 }