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         if (a->vid != b->vid)
  43                 return 0;
  44         switch (a->proto) {
  45         case htons(ETH_P_IP):
  46                 return a->u.ip4 == b->u.ip4;
  47 #if IS_ENABLED(CONFIG_IPV6)
  48         case htons(ETH_P_IPV6):
  49                 return ipv6_addr_equal(&a->u.ip6, &b->u.ip6);
  50 #endif
  51         }
  52         return 0;
  53 }
  54
  55 static inline int __br_ip4_hash(struct net_bridge_mdb_htable *mdb, __be32 ip,
  56                                 __u16 vid)
  57 {
  58         return jhash_2words((__force u32)ip, vid, mdb->secret) & (mdb->max - 1);
  59 }
  60
  61 #if IS_ENABLED(CONFIG_IPV6)
  62 static inline int __br_ip6_hash(struct net_bridge_mdb_htable *mdb,
  63                                 const struct in6_addr *ip,
  64                                 __u16 vid)
  65 {
  66         return jhash_2words(ipv6_addr_hash(ip), vid,
  67                             mdb->secret) & (mdb->max - 1);
  68 }
  69 #endif
  70
  71 static inline int br_ip_hash(struct net_bridge_mdb_htable *mdb,
  72                              struct br_ip *ip)
  73 {
  74         switch (ip->proto) {
  75         case htons(ETH_P_IP):
  76                 return __br_ip4_hash(mdb, ip->u.ip4, ip->vid);
  77 #if IS_ENABLED(CONFIG_IPV6)
  78         case htons(ETH_P_IPV6):
  79                 return __br_ip6_hash(mdb, &ip->u.ip6, ip->vid);
  80 #endif
  81         }
  82         return 0;
  83 }
  84
  85 static struct net_bridge_mdb_entry *__br_mdb_ip_get(
  86         struct net_bridge_mdb_htable *mdb, struct br_ip *dst, int hash)
  87 {
  88         struct net_bridge_mdb_entry *mp;
  89
  90         hlist_for_each_entry_rcu(mp, &mdb->mhash[hash], hlist[mdb->ver]) {
  91                 if (br_ip_equal(&mp->addr, dst))
  92                         return mp;
  93         }
  94
  95         return NULL;
  96 }
  97
  98 struct net_bridge_mdb_entry *br_mdb_ip_get(struct net_bridge_mdb_htable *mdb,
  99                                            struct br_ip *dst)
 100 {
 101         if (!mdb)
 102                 return NULL;
 103
 104         return __br_mdb_ip_get(mdb, dst, br_ip_hash(mdb, dst));
 105 }
 106
 107 static struct net_bridge_mdb_entry *br_mdb_ip4_get(
 108         struct net_bridge_mdb_htable *mdb, __be32 dst, __u16 vid)
 109 {
 110         struct br_ip br_dst;
 111
 112         br_dst.u.ip4 = dst;
 113         br_dst.proto = htons(ETH_P_IP);
 114         br_dst.vid = vid;
 115
 116         return br_mdb_ip_get(mdb, &br_dst);
 117 }
 118
 119 #if IS_ENABLED(CONFIG_IPV6)
 120 static struct net_bridge_mdb_entry *br_mdb_ip6_get(
 121         struct net_bridge_mdb_htable *mdb, const struct in6_addr *dst,
 122         __u16 vid)
 123 {
 124         struct br_ip br_dst;
 125
 126         br_dst.u.ip6 = *dst;
 127         br_dst.proto = htons(ETH_P_IPV6);
 128         br_dst.vid = vid;
 129
 130         return br_mdb_ip_get(mdb, &br_dst);
 131 }
 132 #endif
 133
 134 struct net_bridge_mdb_entry *br_mdb_get(struct net_bridge *br,
 135                                         struct sk_buff *skb, u16 vid)
 136 {
 137         struct net_bridge_mdb_htable *mdb = rcu_dereference(br->mdb);
 138         struct br_ip ip;
 139
 140         if (br->multicast_disabled)
 141                 return NULL;
 142
 143         if (BR_INPUT_SKB_CB(skb)->igmp)
 144                 return NULL;
 145
 146         ip.proto = skb->protocol;
 147         ip.vid = vid;
 148
 149         switch (skb->protocol) {
 150         case htons(ETH_P_IP):
 151                 ip.u.ip4 = ip_hdr(skb)->daddr;
 152                 break;
 153 #if IS_ENABLED(CONFIG_IPV6)
 154         case htons(ETH_P_IPV6):
 155                 ip.u.ip6 = ipv6_hdr(skb)->daddr;
 156                 break;
 157 #endif
 158         default:
 159                 return NULL;
 160         }
 161
 162         return br_mdb_ip_get(mdb, &ip);
 163 }
 164
 165 static void br_mdb_free(struct rcu_head *head)
 166 {
 167         struct net_bridge_mdb_htable *mdb =
 168                 container_of(head, struct net_bridge_mdb_htable, rcu);
 169         struct net_bridge_mdb_htable *old = mdb->old;
 170
 171         mdb->old = NULL;
 172         kfree(old->mhash);
 173         kfree(old);
 174 }
 175
 176 static int br_mdb_copy(struct net_bridge_mdb_htable *new,
 177                        struct net_bridge_mdb_htable *old,
 178                        int elasticity)
 179 {
 180         struct net_bridge_mdb_entry *mp;
 181         int maxlen;
 182         int len;
 183         int i;
 184
 185         for (i = 0; i < old->max; i++)
 186                 hlist_for_each_entry(mp, &old->mhash[i], hlist[old->ver])
 187                         hlist_add_head(&mp->hlist[new->ver],
 188                                        &new->mhash[br_ip_hash(new, &mp->addr)]);
 189
 190         if (!elasticity)
 191                 return 0;
 192
 193         maxlen = 0;
 194         for (i = 0; i < new->max; i++) {
 195                 len = 0;
 196                 hlist_for_each_entry(mp, &new->mhash[i], hlist[new->ver])
 197                         len++;
 198                 if (len > maxlen)
 199                         maxlen = len;
 200         }
 201
 202         return maxlen > elasticity ? -EINVAL : 0;
 203 }
 204
 205 void br_multicast_free_pg(struct rcu_head *head)
 206 {
 207         struct net_bridge_port_group *p =
 208                 container_of(head, struct net_bridge_port_group, rcu);
 209
 210         kfree(p);
 211 }
 212
 213 static void br_multicast_free_group(struct rcu_head *head)
 214 {
 215         struct net_bridge_mdb_entry *mp =
 216                 container_of(head, struct net_bridge_mdb_entry, rcu);
 217
 218         kfree(mp);
 219 }
 220
 221 static void br_multicast_group_expired(unsigned long data)
 222 {
 223         struct net_bridge_mdb_entry *mp = (void *)data;
 224         struct net_bridge *br = mp->br;
 225         struct net_bridge_mdb_htable *mdb;
 226
 227         spin_lock(&br->multicast_lock);
 228         if (!netif_running(br->dev) || timer_pending(&mp->timer))
 229                 goto out;
 230
 231         mp->mglist = false;
 232
 233         if (mp->ports)
 234                 goto out;
 235
 236         mdb = mlock_dereference(br->mdb, br);
 237
 238         hlist_del_rcu(&mp->hlist[mdb->ver]);
 239         mdb->size--;
 240
 241         call_rcu_bh(&mp->rcu, br_multicast_free_group);
 242
 243 out:
 244         spin_unlock(&br->multicast_lock);
 245 }
 246
 247 static void br_multicast_del_pg(struct net_bridge *br,
 248                                 struct net_bridge_port_group *pg)
 249 {
 250         struct net_bridge_mdb_htable *mdb;
 251         struct net_bridge_mdb_entry *mp;
 252         struct net_bridge_port_group *p;
 253         struct net_bridge_port_group __rcu **pp;
 254
 255         mdb = mlock_dereference(br->mdb, br);
 256
 257         mp = br_mdb_ip_get(mdb, &pg->addr);
 258         if (WARN_ON(!mp))
 259                 return;
 260
 261         for (pp = &mp->ports;
 262              (p = mlock_dereference(*pp, br)) != NULL;
 263              pp = &p->next) {
 264                 if (p != pg)
 265                         continue;
 266
 267                 rcu_assign_pointer(*pp, p->next);
 268                 hlist_del_init(&p->mglist);
 269                 del_timer(&p->timer);
 270                 call_rcu_bh(&p->rcu, br_multicast_free_pg);
 271
 272                 if (!mp->ports && !mp->mglist &&
 273                     netif_running(br->dev))
 274                         mod_timer(&mp->timer, jiffies);
 275
 276                 return;
 277         }
 278
 279         WARN_ON(1);
 280 }
 281
 282 static void br_multicast_port_group_expired(unsigned long data)
 283 {
 284         struct net_bridge_port_group *pg = (void *)data;
 285         struct net_bridge *br = pg->port->br;
 286
 287         spin_lock(&br->multicast_lock);
 288         if (!netif_running(br->dev) || timer_pending(&pg->timer) ||
 289             hlist_unhashed(&pg->mglist) || pg->state & MDB_PERMANENT)
 290                 goto out;
 291
 292         br_multicast_del_pg(br, pg);
 293
 294 out:
 295         spin_unlock(&br->multicast_lock);
 296 }
 297
 298 static int br_mdb_rehash(struct net_bridge_mdb_htable __rcu **mdbp, int max,
 299                          int elasticity)
 300 {
 301         struct net_bridge_mdb_htable *old = rcu_dereference_protected(*mdbp, 1);
 302         struct net_bridge_mdb_htable *mdb;
 303         int err;
 304
 305         mdb = kmalloc(sizeof(*mdb), GFP_ATOMIC);
 306         if (!mdb)
 307                 return -ENOMEM;
 308
 309         mdb->max = max;
 310         mdb->old = old;
 311
 312         mdb->mhash = kzalloc(max * sizeof(*mdb->mhash), GFP_ATOMIC);
 313         if (!mdb->mhash) {
 314                 kfree(mdb);
 315                 return -ENOMEM;
 316         }
 317
 318         mdb->size = old ? old->size : 0;
 319         mdb->ver = old ? old->ver ^ 1 : 0;
 320
 321         if (!old || elasticity)
 322                 get_random_bytes(&mdb->secret, sizeof(mdb->secret));
 323         else
 324                 mdb->secret = old->secret;
 325
 326         if (!old)
 327                 goto out;
 328
 329         err = br_mdb_copy(mdb, old, elasticity);
 330         if (err) {
 331                 kfree(mdb->mhash);
 332                 kfree(mdb);
 333                 return err;
 334         }
 335
 336         br_mdb_rehash_seq++;
 337         call_rcu_bh(&mdb->rcu, br_mdb_free);
 338
 339 out:
 340         rcu_assign_pointer(*mdbp, mdb);
 341
 342         return 0;
 343 }
 344
 345 static struct sk_buff *br_ip4_multicast_alloc_query(struct net_bridge *br,
 346                                                     __be32 group)
 347 {
 348         struct sk_buff *skb;
 349         struct igmphdr *ih;
 350         struct ethhdr *eth;
 351         struct iphdr *iph;
 352
 353         skb = netdev_alloc_skb_ip_align(br->dev, sizeof(*eth) + sizeof(*iph) +
 354                                                  sizeof(*ih) + 4);
 355         if (!skb)
 356                 goto out;
 357
 358         skb->protocol = htons(ETH_P_IP);
 359
 360         skb_reset_mac_header(skb);
 361         eth = eth_hdr(skb);
 362
 363         memcpy(eth->h_source, br->dev->dev_addr, 6);
 364         eth->h_dest[0] = 1;
 365         eth->h_dest[1] = 0;
 366         eth->h_dest[2] = 0x5e;
 367         eth->h_dest[3] = 0;
 368         eth->h_dest[4] = 0;
 369         eth->h_dest[5] = 1;
 370         eth->h_proto = htons(ETH_P_IP);
 371         skb_put(skb, sizeof(*eth));
 372
 373         skb_set_network_header(skb, skb->len);
 374         iph = ip_hdr(skb);
 375
 376         iph->version = 4;
 377         iph->ihl = 6;
 378         iph->tos = 0xc0;
 379         iph->tot_len = htons(sizeof(*iph) + sizeof(*ih) + 4);
 380         iph->id = 0;
 381         iph->frag_off = htons(IP_DF);
 382         iph->ttl = 1;
 383         iph->protocol = IPPROTO_IGMP;
 384         iph->saddr = 0;
 385         iph->daddr = htonl(INADDR_ALLHOSTS_GROUP);
 386         ((u8 *)&iph[1])[0] = IPOPT_RA;
 387         ((u8 *)&iph[1])[1] = 4;
 388         ((u8 *)&iph[1])[2] = 0;
 389         ((u8 *)&iph[1])[3] = 0;
 390         ip_send_check(iph);
 391         skb_put(skb, 24);
 392
 393         skb_set_transport_header(skb, skb->len);
 394         ih = igmp_hdr(skb);
 395         ih->type = IGMP_HOST_MEMBERSHIP_QUERY;
 396         ih->code = (group ? br->multicast_last_member_interval :
 397                             br->multicast_query_response_interval) /
 398                    (HZ / IGMP_TIMER_SCALE);
 399         ih->group = group;
 400         ih->csum = 0;
 401         ih->csum = ip_compute_csum((void *)ih, sizeof(struct igmphdr));
 402         skb_put(skb, sizeof(*ih));
 403
 404         __skb_pull(skb, sizeof(*eth));
 405
 406 out:
 407         return skb;
 408 }
 409
 410 #if IS_ENABLED(CONFIG_IPV6)
 411 static struct sk_buff *br_ip6_multicast_alloc_query(struct net_bridge *br,
 412                                                     const struct in6_addr *group)
 413 {
 414         struct sk_buff *skb;
 415         struct ipv6hdr *ip6h;
 416         struct mld_msg *mldq;
 417         struct ethhdr *eth;
 418         u8 *hopopt;
 419         unsigned long interval;
 420
 421         skb = netdev_alloc_skb_ip_align(br->dev, sizeof(*eth) + sizeof(*ip6h) +
 422                                                  8 + sizeof(*mldq));
 423         if (!skb)
 424                 goto out;
 425
 426         skb->protocol = htons(ETH_P_IPV6);
 427
 428         /* Ethernet header */
 429         skb_reset_mac_header(skb);
 430         eth = eth_hdr(skb);
 431
 432         memcpy(eth->h_source, br->dev->dev_addr, 6);
 433         eth->h_proto = htons(ETH_P_IPV6);
 434         skb_put(skb, sizeof(*eth));
 435
 436         /* IPv6 header + HbH option */
 437         skb_set_network_header(skb, skb->len);
 438         ip6h = ipv6_hdr(skb);
 439
 440         *(__force __be32 *)ip6h = htonl(0x60000000);
 441         ip6h->payload_len = htons(8 + sizeof(*mldq));
 442         ip6h->nexthdr = IPPROTO_HOPOPTS;
 443         ip6h->hop_limit = 1;
 444         ipv6_addr_set(&ip6h->daddr, htonl(0xff020000), 0, 0, htonl(1));
 445         if (ipv6_dev_get_saddr(dev_net(br->dev), br->dev, &ip6h->daddr, 0,
 446                                &ip6h->saddr)) {
 447                 kfree_skb(skb);
 448                 return NULL;
 449         }
 450         ipv6_eth_mc_map(&ip6h->daddr, eth->h_dest);
 451
 452         hopopt = (u8 *)(ip6h + 1);
 453         hopopt[0] = IPPROTO_ICMPV6;             /* next hdr */
 454         hopopt[1] = 0;                          /* length of HbH */
 455         hopopt[2] = IPV6_TLV_ROUTERALERT;       /* Router Alert */
 456         hopopt[3] = 2;                          /* Length of RA Option */
 457         hopopt[4] = 0;                          /* Type = 0x0000 (MLD) */
 458         hopopt[5] = 0;
 459         hopopt[6] = IPV6_TLV_PAD1;              /* Pad1 */
 460         hopopt[7] = IPV6_TLV_PAD1;              /* Pad1 */
 461
 462         skb_put(skb, sizeof(*ip6h) + 8);
 463
 464         /* ICMPv6 */
 465         skb_set_transport_header(skb, skb->len);
 466         mldq = (struct mld_msg *) icmp6_hdr(skb);
 467
 468         interval = ipv6_addr_any(group) ? br->multicast_last_member_interval :
 469                                           br->multicast_query_response_interval;
 470
 471         mldq->mld_type = ICMPV6_MGM_QUERY;
 472         mldq->mld_code = 0;
 473         mldq->mld_cksum = 0;
 474         mldq->mld_maxdelay = htons((u16)jiffies_to_msecs(interval));
 475         mldq->mld_reserved = 0;
 476         mldq->mld_mca = *group;
 477
 478         /* checksum */
 479         mldq->mld_cksum = csum_ipv6_magic(&ip6h->saddr, &ip6h->daddr,
 480                                           sizeof(*mldq), IPPROTO_ICMPV6,
 481                                           csum_partial(mldq,
 482                                                        sizeof(*mldq), 0));
 483         skb_put(skb, sizeof(*mldq));
 484
 485         __skb_pull(skb, sizeof(*eth));
 486
 487 out:
 488         return skb;
 489 }
 490 #endif
 491
 492 static struct sk_buff *br_multicast_alloc_query(struct net_bridge *br,
 493                                                 struct br_ip *addr)
 494 {
 495         switch (addr->proto) {
 496         case htons(ETH_P_IP):
 497                 return br_ip4_multicast_alloc_query(br, addr->u.ip4);
 498 #if IS_ENABLED(CONFIG_IPV6)
 499         case htons(ETH_P_IPV6):
 500                 return br_ip6_multicast_alloc_query(br, &addr->u.ip6);
 501 #endif
 502         }
 503         return NULL;
 504 }
 505
 506 static struct net_bridge_mdb_entry *br_multicast_get_group(
 507         struct net_bridge *br, struct net_bridge_port *port,
 508         struct br_ip *group, int hash)
 509 {
 510         struct net_bridge_mdb_htable *mdb;
 511         struct net_bridge_mdb_entry *mp;
 512         unsigned int count = 0;
 513         unsigned int max;
 514         int elasticity;
 515         int err;
 516
 517         mdb = rcu_dereference_protected(br->mdb, 1);
 518         hlist_for_each_entry(mp, &mdb->mhash[hash], hlist[mdb->ver]) {
 519                 count++;
 520                 if (unlikely(br_ip_equal(group, &mp->addr)))
 521                         return mp;
 522         }
 523
 524         elasticity = 0;
 525         max = mdb->max;
 526
 527         if (unlikely(count > br->hash_elasticity && count)) {
 528                 if (net_ratelimit())
 529                         br_info(br, "Multicast hash table "
 530                                 "chain limit reached: %s\n",
 531                                 port ? port->dev->name : br->dev->name);
 532
 533                 elasticity = br->hash_elasticity;
 534         }
 535
 536         if (mdb->size >= max) {
 537                 max *= 2;
 538                 if (unlikely(max > br->hash_max)) {
 539                         br_warn(br, "Multicast hash table maximum of %d "
 540                                 "reached, disabling snooping: %s\n",
 541                                 br->hash_max,
 542                                 port ? port->dev->name : br->dev->name);
 543                         err = -E2BIG;
 544 disable:
 545                         br->multicast_disabled = 1;
 546                         goto err;
 547                 }
 548         }
 549
 550         if (max > mdb->max || elasticity) {
 551                 if (mdb->old) {
 552                         if (net_ratelimit())
 553                                 br_info(br, "Multicast hash table "
 554                                         "on fire: %s\n",
 555                                         port ? port->dev->name : br->dev->name);
 556                         err = -EEXIST;
 557                         goto err;
 558                 }
 559
 560                 err = br_mdb_rehash(&br->mdb, max, elasticity);
 561                 if (err) {
 562                         br_warn(br, "Cannot rehash multicast "
 563                                 "hash table, disabling snooping: %s, %d, %d\n",
 564                                 port ? port->dev->name : br->dev->name,
 565                                 mdb->size, err);
 566                         goto disable;
 567                 }
 568
 569                 err = -EAGAIN;
 570                 goto err;
 571         }
 572
 573         return NULL;
 574
 575 err:
 576         mp = ERR_PTR(err);
 577         return mp;
 578 }
 579
 580 struct net_bridge_mdb_entry *br_multicast_new_group(struct net_bridge *br,
 581         struct net_bridge_port *port, struct br_ip *group)
 582 {
 583         struct net_bridge_mdb_htable *mdb;
 584         struct net_bridge_mdb_entry *mp;
 585         int hash;
 586         int err;
 587
 588         mdb = rcu_dereference_protected(br->mdb, 1);
 589         if (!mdb) {
 590                 err = br_mdb_rehash(&br->mdb, BR_HASH_SIZE, 0);
 591                 if (err)
 592                         return ERR_PTR(err);
 593                 goto rehash;
 594         }
 595
 596         hash = br_ip_hash(mdb, group);
 597         mp = br_multicast_get_group(br, port, group, hash);
 598         switch (PTR_ERR(mp)) {
 599         case 0:
 600                 break;
 601
 602         case -EAGAIN:
 603 rehash:
 604                 mdb = rcu_dereference_protected(br->mdb, 1);
 605                 hash = br_ip_hash(mdb, group);
 606                 break;
 607
 608         default:
 609                 goto out;
 610         }
 611
 612         mp = kzalloc(sizeof(*mp), GFP_ATOMIC);
 613         if (unlikely(!mp))
 614                 return ERR_PTR(-ENOMEM);
 615
 616         mp->br = br;
 617         mp->addr = *group;
 618         setup_timer(&mp->timer, br_multicast_group_expired,
 619                     (unsigned long)mp);
 620
 621         hlist_add_head_rcu(&mp->hlist[mdb->ver], &mdb->mhash[hash]);
 622         mdb->size++;
 623
 624 out:
 625         return mp;
 626 }
 627
 628 struct net_bridge_port_group *br_multicast_new_port_group(
 629                         struct net_bridge_port *port,
 630                         struct br_ip *group,
 631                         struct net_bridge_port_group __rcu *next,
 632                         unsigned char state)
 633 {
 634         struct net_bridge_port_group *p;
 635
 636         p = kzalloc(sizeof(*p), GFP_ATOMIC);
 637         if (unlikely(!p))
 638                 return NULL;
 639
 640         p->addr = *group;
 641         p->port = port;
 642         p->state = state;
 643         rcu_assign_pointer(p->next, next);
 644         hlist_add_head(&p->mglist, &port->mglist);
 645         setup_timer(&p->timer, br_multicast_port_group_expired,
 646                     (unsigned long)p);
 647         return p;
 648 }
 649
 650 static int br_multicast_add_group(struct net_bridge *br,
 651                                   struct net_bridge_port *port,
 652                                   struct br_ip *group)
 653 {
 654         struct net_bridge_mdb_entry *mp;
 655         struct net_bridge_port_group *p;
 656         struct net_bridge_port_group __rcu **pp;
 657         unsigned long now = jiffies;
 658         int err;
 659
 660         spin_lock(&br->multicast_lock);
 661         if (!netif_running(br->dev) ||
 662             (port && port->state == BR_STATE_DISABLED))
 663                 goto out;
 664
 665         mp = br_multicast_new_group(br, port, group);
 666         err = PTR_ERR(mp);
 667         if (IS_ERR(mp))
 668                 goto err;
 669
 670         if (!port) {
 671                 mp->mglist = true;
 672                 mod_timer(&mp->timer, now + br->multicast_membership_interval);
 673                 goto out;
 674         }
 675
 676         for (pp = &mp->ports;
 677              (p = mlock_dereference(*pp, br)) != NULL;
 678              pp = &p->next) {
 679                 if (p->port == port)
 680                         goto found;
 681                 if ((unsigned long)p->port < (unsigned long)port)
 682                         break;
 683         }
 684
 685         p = br_multicast_new_port_group(port, group, *pp, MDB_TEMPORARY);
 686         if (unlikely(!p))
 687                 goto err;
 688         rcu_assign_pointer(*pp, p);
 689         br_mdb_notify(br->dev, port, group, RTM_NEWMDB);
 690
 691 found:
 692         mod_timer(&p->timer, now + br->multicast_membership_interval);
 693 out:
 694         err = 0;
 695
 696 err:
 697         spin_unlock(&br->multicast_lock);
 698         return err;
 699 }
 700
 701 static int br_ip4_multicast_add_group(struct net_bridge *br,
 702                                       struct net_bridge_port *port,
 703                                       __be32 group,
 704                                       __u16 vid)
 705 {
 706         struct br_ip br_group;
 707
 708         if (ipv4_is_local_multicast(group))
 709                 return 0;
 710
 711         br_group.u.ip4 = group;
 712         br_group.proto = htons(ETH_P_IP);
 713         br_group.vid = vid;
 714
 715         return br_multicast_add_group(br, port, &br_group);
 716 }
 717
 718 #if IS_ENABLED(CONFIG_IPV6)
 719 static int br_ip6_multicast_add_group(struct net_bridge *br,
 720                                       struct net_bridge_port *port,
 721                                       const struct in6_addr *group,
 722                                       __u16 vid)
 723 {
 724         struct br_ip br_group;
 725
 726         if (!ipv6_is_transient_multicast(group))
 727                 return 0;
 728
 729         br_group.u.ip6 = *group;
 730         br_group.proto = htons(ETH_P_IPV6);
 731         br_group.vid = vid;
 732
 733         return br_multicast_add_group(br, port, &br_group);
 734 }
 735 #endif
 736
 737 static void br_multicast_router_expired(unsigned long data)
 738 {
 739         struct net_bridge_port *port = (void *)data;
 740         struct net_bridge *br = port->br;
 741
 742         spin_lock(&br->multicast_lock);
 743         if (port->multicast_router != 1 ||
 744             timer_pending(&port->multicast_router_timer) ||
 745             hlist_unhashed(&port->rlist))
 746                 goto out;
 747
 748         hlist_del_init_rcu(&port->rlist);
 749
 750 out:
 751         spin_unlock(&br->multicast_lock);
 752 }
 753
 754 static void br_multicast_local_router_expired(unsigned long data)
 755 {
 756 }
 757
 758 static void br_multicast_querier_expired(unsigned long data)
 759 {
 760         struct net_bridge *br = (void *)data;
 761
 762         spin_lock(&br->multicast_lock);
 763         if (!netif_running(br->dev) || br->multicast_disabled)
 764                 goto out;
 765
 766         br_multicast_start_querier(br);
 767
 768 out:
 769         spin_unlock(&br->multicast_lock);
 770 }
 771
 772 static void __br_multicast_send_query(struct net_bridge *br,
 773                                       struct net_bridge_port *port,
 774                                       struct br_ip *ip)
 775 {
 776         struct sk_buff *skb;
 777
 778         skb = br_multicast_alloc_query(br, ip);
 779         if (!skb)
 780                 return;
 781
 782         if (port) {
 783                 __skb_push(skb, sizeof(struct ethhdr));
 784                 skb->dev = port->dev;
 785                 NF_HOOK(NFPROTO_BRIDGE, NF_BR_LOCAL_OUT, skb, NULL, skb->dev,
 786                         dev_queue_xmit);
 787         } else
 788                 netif_rx(skb);
 789 }
 790
 791 static void br_multicast_send_query(struct net_bridge *br,
 792                                     struct net_bridge_port *port, u32 sent)
 793 {
 794         unsigned long time;
 795         struct br_ip br_group;
 796
 797         if (!netif_running(br->dev) || br->multicast_disabled ||
 798             !br->multicast_querier ||
 799             timer_pending(&br->multicast_querier_timer))
 800                 return;
 801
 802         memset(&br_group.u, 0, sizeof(br_group.u));
 803
 804         br_group.proto = htons(ETH_P_IP);
 805         __br_multicast_send_query(br, port, &br_group);
 806
 807 #if IS_ENABLED(CONFIG_IPV6)
 808         br_group.proto = htons(ETH_P_IPV6);
 809         __br_multicast_send_query(br, port, &br_group);
 810 #endif
 811
 812         time = jiffies;
 813         time += sent < br->multicast_startup_query_count ?
 814                 br->multicast_startup_query_interval :
 815                 br->multicast_query_interval;
 816         mod_timer(port ? &port->multicast_query_timer :
 817                          &br->multicast_query_timer, time);
 818 }
 819
 820 static void br_multicast_port_query_expired(unsigned long data)
 821 {
 822         struct net_bridge_port *port = (void *)data;
 823         struct net_bridge *br = port->br;
 824
 825         spin_lock(&br->multicast_lock);
 826         if (port->state == BR_STATE_DISABLED ||
 827             port->state == BR_STATE_BLOCKING)
 828                 goto out;
 829
 830         if (port->multicast_startup_queries_sent <
 831             br->multicast_startup_query_count)
 832                 port->multicast_startup_queries_sent++;
 833
 834         br_multicast_send_query(port->br, port,
 835                                 port->multicast_startup_queries_sent);
 836
 837 out:
 838         spin_unlock(&br->multicast_lock);
 839 }
 840
 841 void br_multicast_add_port(struct net_bridge_port *port)
 842 {
 843         port->multicast_router = 1;
 844
 845         setup_timer(&port->multicast_router_timer, br_multicast_router_expired,
 846                     (unsigned long)port);
 847         setup_timer(&port->multicast_query_timer,
 848                     br_multicast_port_query_expired, (unsigned long)port);
 849 }
 850
 851 void br_multicast_del_port(struct net_bridge_port *port)
 852 {
 853         del_timer_sync(&port->multicast_router_timer);
 854 }
 855
 856 static void __br_multicast_enable_port(struct net_bridge_port *port)
 857 {
 858         port->multicast_startup_queries_sent = 0;
 859
 860         if (try_to_del_timer_sync(&port->multicast_query_timer) >= 0 ||
 861             del_timer(&port->multicast_query_timer))
 862                 mod_timer(&port->multicast_query_timer, jiffies);
 863 }
 864
 865 void br_multicast_enable_port(struct net_bridge_port *port)
 866 {
 867         struct net_bridge *br = port->br;
 868
 869         spin_lock(&br->multicast_lock);
 870         if (br->multicast_disabled || !netif_running(br->dev))
 871                 goto out;
 872
 873         __br_multicast_enable_port(port);
 874
 875 out:
 876         spin_unlock(&br->multicast_lock);
 877 }
 878
 879 void br_multicast_disable_port(struct net_bridge_port *port)
 880 {
 881         struct net_bridge *br = port->br;
 882         struct net_bridge_port_group *pg;
 883         struct hlist_node *n;
 884
 885         spin_lock(&br->multicast_lock);
 886         hlist_for_each_entry_safe(pg, n, &port->mglist, mglist)
 887                 br_multicast_del_pg(br, pg);
 888
 889         if (!hlist_unhashed(&port->rlist))
 890                 hlist_del_init_rcu(&port->rlist);
 891         del_timer(&port->multicast_router_timer);
 892         del_timer(&port->multicast_query_timer);
 893         spin_unlock(&br->multicast_lock);
 894 }
 895
 896 static int br_ip4_multicast_igmp3_report(struct net_bridge *br,
 897                                          struct net_bridge_port *port,
 898                                          struct sk_buff *skb)
 899 {
 900         struct igmpv3_report *ih;
 901         struct igmpv3_grec *grec;
 902         int i;
 903         int len;
 904         int num;
 905         int type;
 906         int err = 0;
 907         __be32 group;
 908         u16 vid = 0;
 909
 910         if (!pskb_may_pull(skb, sizeof(*ih)))
 911                 return -EINVAL;
 912
 913         br_vlan_get_tag(skb, &vid);
 914         ih = igmpv3_report_hdr(skb);
 915         num = ntohs(ih->ngrec);
 916         len = sizeof(*ih);
 917
 918         for (i = 0; i < num; i++) {
 919                 len += sizeof(*grec);
 920                 if (!pskb_may_pull(skb, len))
 921                         return -EINVAL;
 922
 923                 grec = (void *)(skb->data + len - sizeof(*grec));
 924                 group = grec->grec_mca;
 925                 type = grec->grec_type;
 926
 927                 len += ntohs(grec->grec_nsrcs) * 4;
 928                 if (!pskb_may_pull(skb, len))
 929                         return -EINVAL;
 930
 931                 /* We treat this as an IGMPv2 report for now. */
 932                 switch (type) {
 933                 case IGMPV3_MODE_IS_INCLUDE:
 934                 case IGMPV3_MODE_IS_EXCLUDE:
 935                 case IGMPV3_CHANGE_TO_INCLUDE:
 936                 case IGMPV3_CHANGE_TO_EXCLUDE:
 937                 case IGMPV3_ALLOW_NEW_SOURCES:
 938                 case IGMPV3_BLOCK_OLD_SOURCES:
 939                         break;
 940
 941                 default:
 942                         continue;
 943                 }
 944
 945                 err = br_ip4_multicast_add_group(br, port, group, vid);
 946                 if (err)
 947                         break;
 948         }
 949
 950         return err;
 951 }
 952
 953 #if IS_ENABLED(CONFIG_IPV6)
 954 static int br_ip6_multicast_mld2_report(struct net_bridge *br,
 955                                         struct net_bridge_port *port,
 956                                         struct sk_buff *skb)
 957 {
 958         struct icmp6hdr *icmp6h;
 959         struct mld2_grec *grec;
 960         int i;
 961         int len;
 962         int num;
 963         int err = 0;
 964         u16 vid = 0;
 965
 966         if (!pskb_may_pull(skb, sizeof(*icmp6h)))
 967                 return -EINVAL;
 968
 969         br_vlan_get_tag(skb, &vid);
 970         icmp6h = icmp6_hdr(skb);
 971         num = ntohs(icmp6h->icmp6_dataun.un_data16[1]);
 972         len = sizeof(*icmp6h);
 973
 974         for (i = 0; i < num; i++) {
 975                 __be16 *nsrcs, _nsrcs;
 976
 977                 nsrcs = skb_header_pointer(skb,
 978                                            len + offsetof(struct mld2_grec,
 979                                                           grec_nsrcs),
 980                                            sizeof(_nsrcs), &_nsrcs);
 981                 if (!nsrcs)
 982                         return -EINVAL;
 983
 984                 if (!pskb_may_pull(skb,
 985                                    len + sizeof(*grec) +
 986                                    sizeof(struct in6_addr) * ntohs(*nsrcs)))
 987                         return -EINVAL;
 988
 989                 grec = (struct mld2_grec *)(skb->data + len);
 990                 len += sizeof(*grec) +
 991                        sizeof(struct in6_addr) * ntohs(*nsrcs);
 992
 993                 /* We treat these as MLDv1 reports for now. */
 994                 switch (grec->grec_type) {
 995                 case MLD2_MODE_IS_INCLUDE:
 996                 case MLD2_MODE_IS_EXCLUDE:
 997                 case MLD2_CHANGE_TO_INCLUDE:
 998                 case MLD2_CHANGE_TO_EXCLUDE:
 999                 case MLD2_ALLOW_NEW_SOURCES:
1000                 case MLD2_BLOCK_OLD_SOURCES:
1001                         break;
1002
1003                 default:
1004                         continue;
1005                 }
1006
1007                 err = br_ip6_multicast_add_group(br, port, &grec->grec_mca,
1008                                                  vid);
1009                 if (!err)
1010                         break;
1011         }
1012
1013         return err;
1014 }
1015 #endif
1016
1017 /*
1018  * Add port to rotuer_list
1019  *  list is maintained ordered by pointer value
1020  *  and locked by br->multicast_lock and RCU
1021  */
1022 static void br_multicast_add_router(struct net_bridge *br,
1023                                     struct net_bridge_port *port)
1024 {
1025         struct net_bridge_port *p;
1026         struct hlist_node *slot = NULL;
1027
1028         hlist_for_each_entry(p, &br->router_list, rlist) {
1029                 if ((unsigned long) port >= (unsigned long) p)
1030                         break;
1031                 slot = &p->rlist;
1032         }
1033
1034         if (slot)
1035                 hlist_add_after_rcu(slot, &port->rlist);
1036         else
1037                 hlist_add_head_rcu(&port->rlist, &br->router_list);
1038 }
1039
1040 static void br_multicast_mark_router(struct net_bridge *br,
1041                                      struct net_bridge_port *port)
1042 {
1043         unsigned long now = jiffies;
1044
1045         if (!port) {
1046                 if (br->multicast_router == 1)
1047                         mod_timer(&br->multicast_router_timer,
1048                                   now + br->multicast_querier_interval);
1049                 return;
1050         }
1051
1052         if (port->multicast_router != 1)
1053                 return;
1054
1055         if (!hlist_unhashed(&port->rlist))
1056                 goto timer;
1057
1058         br_multicast_add_router(br, port);
1059
1060 timer:
1061         mod_timer(&port->multicast_router_timer,
1062                   now + br->multicast_querier_interval);
1063 }
1064
1065 static void br_multicast_query_received(struct net_bridge *br,
1066                                         struct net_bridge_port *port,
1067                                         int saddr)
1068 {
1069         if (saddr)
1070                 mod_timer(&br->multicast_querier_timer,
1071                           jiffies + br->multicast_querier_interval);
1072         else if (timer_pending(&br->multicast_querier_timer))
1073                 return;
1074
1075         br_multicast_mark_router(br, port);
1076 }
1077
1078 static int br_ip4_multicast_query(struct net_bridge *br,
1079                                   struct net_bridge_port *port,
1080                                   struct sk_buff *skb)
1081 {
1082         const struct iphdr *iph = ip_hdr(skb);
1083         struct igmphdr *ih = igmp_hdr(skb);
1084         struct net_bridge_mdb_entry *mp;
1085         struct igmpv3_query *ih3;
1086         struct net_bridge_port_group *p;
1087         struct net_bridge_port_group __rcu **pp;
1088         unsigned long max_delay;
1089         unsigned long now = jiffies;
1090         __be32 group;
1091         int err = 0;
1092         u16 vid = 0;
1093
1094         spin_lock(&br->multicast_lock);
1095         if (!netif_running(br->dev) ||
1096             (port && port->state == BR_STATE_DISABLED))
1097                 goto out;
1098
1099         br_multicast_query_received(br, port, !!iph->saddr);
1100
1101         group = ih->group;
1102
1103         if (skb->len == sizeof(*ih)) {
1104                 max_delay = ih->code * (HZ / IGMP_TIMER_SCALE);
1105
1106                 if (!max_delay) {
1107                         max_delay = 10 * HZ;
1108                         group = 0;
1109                 }
1110         } else {
1111                 if (!pskb_may_pull(skb, sizeof(struct igmpv3_query))) {
1112                         err = -EINVAL;
1113                         goto out;
1114                 }
1115
1116                 ih3 = igmpv3_query_hdr(skb);
1117                 if (ih3->nsrcs)
1118                         goto out;
1119
1120                 max_delay = ih3->code ?
1121                             IGMPV3_MRC(ih3->code) * (HZ / IGMP_TIMER_SCALE) : 1;
1122         }
1123
1124         if (!group)
1125                 goto out;
1126
1127         br_vlan_get_tag(skb, &vid);
1128         mp = br_mdb_ip4_get(mlock_dereference(br->mdb, br), group, vid);
1129         if (!mp)
1130                 goto out;
1131
1132         max_delay *= br->multicast_last_member_count;
1133
1134         if (mp->mglist &&
1135             (timer_pending(&mp->timer) ?
1136              time_after(mp->timer.expires, now + max_delay) :
1137              try_to_del_timer_sync(&mp->timer) >= 0))
1138                 mod_timer(&mp->timer, now + max_delay);
1139
1140         for (pp = &mp->ports;
1141              (p = mlock_dereference(*pp, br)) != NULL;
1142              pp = &p->next) {
1143                 if (timer_pending(&p->timer) ?
1144                     time_after(p->timer.expires, now + max_delay) :
1145                     try_to_del_timer_sync(&p->timer) >= 0)
1146                         mod_timer(&p->timer, now + max_delay);
1147         }
1148
1149 out:
1150         spin_unlock(&br->multicast_lock);
1151         return err;
1152 }
1153
1154 #if IS_ENABLED(CONFIG_IPV6)
1155 static int br_ip6_multicast_query(struct net_bridge *br,
1156                                   struct net_bridge_port *port,
1157                                   struct sk_buff *skb)
1158 {
1159         const struct ipv6hdr *ip6h = ipv6_hdr(skb);
1160         struct mld_msg *mld;
1161         struct net_bridge_mdb_entry *mp;
1162         struct mld2_query *mld2q;
1163         struct net_bridge_port_group *p;
1164         struct net_bridge_port_group __rcu **pp;
1165         unsigned long max_delay;
1166         unsigned long now = jiffies;
1167         const struct in6_addr *group = NULL;
1168         int err = 0;
1169         u16 vid = 0;
1170
1171         spin_lock(&br->multicast_lock);
1172         if (!netif_running(br->dev) ||
1173             (port && port->state == BR_STATE_DISABLED))
1174                 goto out;
1175
1176         br_multicast_query_received(br, port, !ipv6_addr_any(&ip6h->saddr));
1177
1178         if (skb->len == sizeof(*mld)) {
1179                 if (!pskb_may_pull(skb, sizeof(*mld))) {
1180                         err = -EINVAL;
1181                         goto out;
1182                 }
1183                 mld = (struct mld_msg *) icmp6_hdr(skb);
1184                 max_delay = msecs_to_jiffies(ntohs(mld->mld_maxdelay));
1185                 if (max_delay)
1186                         group = &mld->mld_mca;
1187         } else if (skb->len >= sizeof(*mld2q)) {
1188                 if (!pskb_may_pull(skb, sizeof(*mld2q))) {
1189                         err = -EINVAL;
1190                         goto out;
1191                 }
1192                 mld2q = (struct mld2_query *)icmp6_hdr(skb);
1193                 if (!mld2q->mld2q_nsrcs)
1194                         group = &mld2q->mld2q_mca;
1195                 max_delay = mld2q->mld2q_mrc ? MLDV2_MRC(ntohs(mld2q->mld2q_mrc)) : 1;
1196         }
1197
1198         if (!group)
1199                 goto out;
1200
1201         br_vlan_get_tag(skb, &vid);
1202         mp = br_mdb_ip6_get(mlock_dereference(br->mdb, br), group, vid);
1203         if (!mp)
1204                 goto out;
1205
1206         max_delay *= br->multicast_last_member_count;
1207         if (mp->mglist &&
1208             (timer_pending(&mp->timer) ?
1209              time_after(mp->timer.expires, now + max_delay) :
1210              try_to_del_timer_sync(&mp->timer) >= 0))
1211                 mod_timer(&mp->timer, now + max_delay);
1212
1213         for (pp = &mp->ports;
1214              (p = mlock_dereference(*pp, br)) != NULL;
1215              pp = &p->next) {
1216                 if (timer_pending(&p->timer) ?
1217                     time_after(p->timer.expires, now + max_delay) :
1218                     try_to_del_timer_sync(&p->timer) >= 0)
1219                         mod_timer(&p->timer, now + max_delay);
1220         }
1221
1222 out:
1223         spin_unlock(&br->multicast_lock);
1224         return err;
1225 }
1226 #endif
1227
1228 static void br_multicast_leave_group(struct net_bridge *br,
1229                                      struct net_bridge_port *port,
1230                                      struct br_ip *group)
1231 {
1232         struct net_bridge_mdb_htable *mdb;
1233         struct net_bridge_mdb_entry *mp;
1234         struct net_bridge_port_group *p;
1235         unsigned long now;
1236         unsigned long time;
1237
1238         spin_lock(&br->multicast_lock);
1239         if (!netif_running(br->dev) ||
1240             (port && port->state == BR_STATE_DISABLED) ||
1241             timer_pending(&br->multicast_querier_timer))
1242                 goto out;
1243
1244         mdb = mlock_dereference(br->mdb, br);
1245         mp = br_mdb_ip_get(mdb, group);
1246         if (!mp)
1247                 goto out;
1248
1249         if (port && (port->flags & BR_MULTICAST_FAST_LEAVE)) {
1250                 struct net_bridge_port_group __rcu **pp;
1251
1252                 for (pp = &mp->ports;
1253                      (p = mlock_dereference(*pp, br)) != NULL;
1254                      pp = &p->next) {
1255                         if (p->port != port)
1256                                 continue;
1257
1258                         rcu_assign_pointer(*pp, p->next);
1259                         hlist_del_init(&p->mglist);
1260                         del_timer(&p->timer);
1261                         call_rcu_bh(&p->rcu, br_multicast_free_pg);
1262                         br_mdb_notify(br->dev, port, group, RTM_DELMDB);
1263
1264                         if (!mp->ports && !mp->mglist &&
1265                             netif_running(br->dev))
1266                                 mod_timer(&mp->timer, jiffies);
1267                 }
1268                 goto out;
1269         }
1270
1271         now = jiffies;
1272         time = now + br->multicast_last_member_count *
1273                      br->multicast_last_member_interval;
1274
1275         if (!port) {
1276                 if (mp->mglist &&
1277                     (timer_pending(&mp->timer) ?
1278                      time_after(mp->timer.expires, time) :
1279                      try_to_del_timer_sync(&mp->timer) >= 0)) {
1280                         mod_timer(&mp->timer, time);
1281                 }
1282
1283                 goto out;
1284         }
1285
1286         for (p = mlock_dereference(mp->ports, br);
1287              p != NULL;
1288              p = mlock_dereference(p->next, br)) {
1289                 if (p->port != port)
1290                         continue;
1291
1292                 if (!hlist_unhashed(&p->mglist) &&
1293                     (timer_pending(&p->timer) ?
1294                      time_after(p->timer.expires, time) :
1295                      try_to_del_timer_sync(&p->timer) >= 0)) {
1296                         mod_timer(&p->timer, time);
1297                 }
1298
1299                 break;
1300         }
1301
1302 out:
1303         spin_unlock(&br->multicast_lock);
1304 }
1305
1306 static void br_ip4_multicast_leave_group(struct net_bridge *br,
1307                                          struct net_bridge_port *port,
1308                                          __be32 group,
1309                                          __u16 vid)
1310 {
1311         struct br_ip br_group;
1312
1313         if (ipv4_is_local_multicast(group))
1314                 return;
1315
1316         br_group.u.ip4 = group;
1317         br_group.proto = htons(ETH_P_IP);
1318         br_group.vid = vid;
1319
1320         br_multicast_leave_group(br, port, &br_group);
1321 }
1322
1323 #if IS_ENABLED(CONFIG_IPV6)
1324 static void br_ip6_multicast_leave_group(struct net_bridge *br,
1325                                          struct net_bridge_port *port,
1326                                          const struct in6_addr *group,
1327                                          __u16 vid)
1328 {
1329         struct br_ip br_group;
1330
1331         if (!ipv6_is_transient_multicast(group))
1332                 return;
1333
1334         br_group.u.ip6 = *group;
1335         br_group.proto = htons(ETH_P_IPV6);
1336         br_group.vid = vid;
1337
1338         br_multicast_leave_group(br, port, &br_group);
1339 }
1340 #endif
1341
1342 static int br_multicast_ipv4_rcv(struct net_bridge *br,
1343                                  struct net_bridge_port *port,
1344                                  struct sk_buff *skb)
1345 {
1346         struct sk_buff *skb2 = skb;
1347         const struct iphdr *iph;
1348         struct igmphdr *ih;
1349         unsigned int len;
1350         unsigned int offset;
1351         int err;
1352         u16 vid = 0;
1353
1354         /* We treat OOM as packet loss for now. */
1355         if (!pskb_may_pull(skb, sizeof(*iph)))
1356                 return -EINVAL;
1357
1358         iph = ip_hdr(skb);
1359
1360         if (iph->ihl < 5 || iph->version != 4)
1361                 return -EINVAL;
1362
1363         if (!pskb_may_pull(skb, ip_hdrlen(skb)))
1364                 return -EINVAL;
1365
1366         iph = ip_hdr(skb);
1367
1368         if (unlikely(ip_fast_csum((u8 *)iph, iph->ihl)))
1369                 return -EINVAL;
1370
1371         if (iph->protocol != IPPROTO_IGMP) {
1372                 if ((iph->daddr & IGMP_LOCAL_GROUP_MASK) != IGMP_LOCAL_GROUP)
1373                         BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1374                 return 0;
1375         }
1376
1377         len = ntohs(iph->tot_len);
1378         if (skb->len < len || len < ip_hdrlen(skb))
1379                 return -EINVAL;
1380
1381         if (skb->len > len) {
1382                 skb2 = skb_clone(skb, GFP_ATOMIC);
1383                 if (!skb2)
1384                         return -ENOMEM;
1385
1386                 err = pskb_trim_rcsum(skb2, len);
1387                 if (err)
1388                         goto err_out;
1389         }
1390
1391         len -= ip_hdrlen(skb2);
1392         offset = skb_network_offset(skb2) + ip_hdrlen(skb2);
1393         __skb_pull(skb2, offset);
1394         skb_reset_transport_header(skb2);
1395
1396         err = -EINVAL;
1397         if (!pskb_may_pull(skb2, sizeof(*ih)))
1398                 goto out;
1399
1400         switch (skb2->ip_summed) {
1401         case CHECKSUM_COMPLETE:
1402                 if (!csum_fold(skb2->csum))
1403                         break;
1404                 /* fall through */
1405         case CHECKSUM_NONE:
1406                 skb2->csum = 0;
1407                 if (skb_checksum_complete(skb2))
1408                         goto out;
1409         }
1410
1411         err = 0;
1412
1413         br_vlan_get_tag(skb2, &vid);
1414         BR_INPUT_SKB_CB(skb)->igmp = 1;
1415         ih = igmp_hdr(skb2);
1416
1417         switch (ih->type) {
1418         case IGMP_HOST_MEMBERSHIP_REPORT:
1419         case IGMPV2_HOST_MEMBERSHIP_REPORT:
1420                 BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1421                 err = br_ip4_multicast_add_group(br, port, ih->group, vid);
1422                 break;
1423         case IGMPV3_HOST_MEMBERSHIP_REPORT:
1424                 err = br_ip4_multicast_igmp3_report(br, port, skb2);
1425                 break;
1426         case IGMP_HOST_MEMBERSHIP_QUERY:
1427                 err = br_ip4_multicast_query(br, port, skb2);
1428                 break;
1429         case IGMP_HOST_LEAVE_MESSAGE:
1430                 br_ip4_multicast_leave_group(br, port, ih->group, vid);
1431                 break;
1432         }
1433
1434 out:
1435         __skb_push(skb2, offset);
1436 err_out:
1437         if (skb2 != skb)
1438                 kfree_skb(skb2);
1439         return err;
1440 }
1441
1442 #if IS_ENABLED(CONFIG_IPV6)
1443 static int br_multicast_ipv6_rcv(struct net_bridge *br,
1444                                  struct net_bridge_port *port,
1445                                  struct sk_buff *skb)
1446 {
1447         struct sk_buff *skb2;
1448         const struct ipv6hdr *ip6h;
1449         u8 icmp6_type;
1450         u8 nexthdr;
1451         __be16 frag_off;
1452         unsigned int len;
1453         int offset;
1454         int err;
1455         u16 vid = 0;
1456
1457         if (!pskb_may_pull(skb, sizeof(*ip6h)))
1458                 return -EINVAL;
1459
1460         ip6h = ipv6_hdr(skb);
1461
1462         /*
1463          * We're interested in MLD messages only.
1464          *  - Version is 6
1465          *  - MLD has always Router Alert hop-by-hop option
1466          *  - But we do not support jumbrograms.
1467          */
1468         if (ip6h->version != 6 ||
1469             ip6h->nexthdr != IPPROTO_HOPOPTS ||
1470             ip6h->payload_len == 0)
1471                 return 0;
1472
1473         len = ntohs(ip6h->payload_len) + sizeof(*ip6h);
1474         if (skb->len < len)
1475                 return -EINVAL;
1476
1477         nexthdr = ip6h->nexthdr;
1478         offset = ipv6_skip_exthdr(skb, sizeof(*ip6h), &nexthdr, &frag_off);
1479
1480         if (offset < 0 || nexthdr != IPPROTO_ICMPV6)
1481                 return 0;
1482
1483         /* Okay, we found ICMPv6 header */
1484         skb2 = skb_clone(skb, GFP_ATOMIC);
1485         if (!skb2)
1486                 return -ENOMEM;
1487
1488         err = -EINVAL;
1489         if (!pskb_may_pull(skb2, offset + sizeof(struct icmp6hdr)))
1490                 goto out;
1491
1492         len -= offset - skb_network_offset(skb2);
1493
1494         __skb_pull(skb2, offset);
1495         skb_reset_transport_header(skb2);
1496         skb_postpull_rcsum(skb2, skb_network_header(skb2),
1497                            skb_network_header_len(skb2));
1498
1499         icmp6_type = icmp6_hdr(skb2)->icmp6_type;
1500
1501         switch (icmp6_type) {
1502         case ICMPV6_MGM_QUERY:
1503         case ICMPV6_MGM_REPORT:
1504         case ICMPV6_MGM_REDUCTION:
1505         case ICMPV6_MLD2_REPORT:
1506                 break;
1507         default:
1508                 err = 0;
1509                 goto out;
1510         }
1511
1512         /* Okay, we found MLD message. Check further. */
1513         if (skb2->len > len) {
1514                 err = pskb_trim_rcsum(skb2, len);
1515                 if (err)
1516                         goto out;
1517                 err = -EINVAL;
1518         }
1519
1520         ip6h = ipv6_hdr(skb2);
1521
1522         switch (skb2->ip_summed) {
1523         case CHECKSUM_COMPLETE:
1524                 if (!csum_ipv6_magic(&ip6h->saddr, &ip6h->daddr, skb2->len,
1525                                         IPPROTO_ICMPV6, skb2->csum))
1526                         break;
1527                 /*FALLTHROUGH*/
1528         case CHECKSUM_NONE:
1529                 skb2->csum = ~csum_unfold(csum_ipv6_magic(&ip6h->saddr,
1530                                                         &ip6h->daddr,
1531                                                         skb2->len,
1532                                                         IPPROTO_ICMPV6, 0));
1533                 if (__skb_checksum_complete(skb2))
1534                         goto out;
1535         }
1536
1537         err = 0;
1538
1539         br_vlan_get_tag(skb, &vid);
1540         BR_INPUT_SKB_CB(skb)->igmp = 1;
1541
1542         switch (icmp6_type) {
1543         case ICMPV6_MGM_REPORT:
1544             {
1545                 struct mld_msg *mld;
1546                 if (!pskb_may_pull(skb2, sizeof(*mld))) {
1547                         err = -EINVAL;
1548                         goto out;
1549                 }
1550                 mld = (struct mld_msg *)skb_transport_header(skb2);
1551                 BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1552                 err = br_ip6_multicast_add_group(br, port, &mld->mld_mca, vid);
1553                 break;
1554             }
1555         case ICMPV6_MLD2_REPORT:
1556                 err = br_ip6_multicast_mld2_report(br, port, skb2);
1557                 break;
1558         case ICMPV6_MGM_QUERY:
1559                 err = br_ip6_multicast_query(br, port, skb2);
1560                 break;
1561         case ICMPV6_MGM_REDUCTION:
1562             {
1563                 struct mld_msg *mld;
1564                 if (!pskb_may_pull(skb2, sizeof(*mld))) {
1565                         err = -EINVAL;
1566                         goto out;
1567                 }
1568                 mld = (struct mld_msg *)skb_transport_header(skb2);
1569                 br_ip6_multicast_leave_group(br, port, &mld->mld_mca, vid);
1570             }
1571         }
1572
1573 out:
1574         kfree_skb(skb2);
1575         return err;
1576 }
1577 #endif
1578
1579 int br_multicast_rcv(struct net_bridge *br, struct net_bridge_port *port,
1580                      struct sk_buff *skb)
1581 {
1582         BR_INPUT_SKB_CB(skb)->igmp = 0;
1583         BR_INPUT_SKB_CB(skb)->mrouters_only = 0;
1584
1585         if (br->multicast_disabled)
1586                 return 0;
1587
1588         switch (skb->protocol) {
1589         case htons(ETH_P_IP):
1590                 return br_multicast_ipv4_rcv(br, port, skb);
1591 #if IS_ENABLED(CONFIG_IPV6)
1592         case htons(ETH_P_IPV6):
1593                 return br_multicast_ipv6_rcv(br, port, skb);
1594 #endif
1595         }
1596
1597         return 0;
1598 }
1599
1600 static void br_multicast_query_expired(unsigned long data)
1601 {
1602         struct net_bridge *br = (void *)data;
1603
1604         spin_lock(&br->multicast_lock);
1605         if (br->multicast_startup_queries_sent <
1606             br->multicast_startup_query_count)
1607                 br->multicast_startup_queries_sent++;
1608
1609         br_multicast_send_query(br, NULL, br->multicast_startup_queries_sent);
1610
1611         spin_unlock(&br->multicast_lock);
1612 }
1613
1614 void br_multicast_init(struct net_bridge *br)
1615 {
1616         br->hash_elasticity = 4;
1617         br->hash_max = 512;
1618
1619         br->multicast_router = 1;
1620         br->multicast_querier = 0;
1621         br->multicast_last_member_count = 2;
1622         br->multicast_startup_query_count = 2;
1623
1624         br->multicast_last_member_interval = HZ;
1625         br->multicast_query_response_interval = 10 * HZ;
1626         br->multicast_startup_query_interval = 125 * HZ / 4;
1627         br->multicast_query_interval = 125 * HZ;
1628         br->multicast_querier_interval = 255 * HZ;
1629         br->multicast_membership_interval = 260 * HZ;
1630
1631         spin_lock_init(&br->multicast_lock);
1632         setup_timer(&br->multicast_router_timer,
1633                     br_multicast_local_router_expired, 0);
1634         setup_timer(&br->multicast_querier_timer,
1635                     br_multicast_querier_expired, (unsigned long)br);
1636         setup_timer(&br->multicast_query_timer, br_multicast_query_expired,
1637                     (unsigned long)br);
1638 }
1639
1640 void br_multicast_open(struct net_bridge *br)
1641 {
1642         br->multicast_startup_queries_sent = 0;
1643
1644         if (br->multicast_disabled)
1645                 return;
1646
1647         mod_timer(&br->multicast_query_timer, jiffies);
1648 }
1649
1650 void br_multicast_stop(struct net_bridge *br)
1651 {
1652         struct net_bridge_mdb_htable *mdb;
1653         struct net_bridge_mdb_entry *mp;
1654         struct hlist_node *n;
1655         u32 ver;
1656         int i;
1657
1658         del_timer_sync(&br->multicast_router_timer);
1659         del_timer_sync(&br->multicast_querier_timer);
1660         del_timer_sync(&br->multicast_query_timer);
1661
1662         spin_lock_bh(&br->multicast_lock);
1663         mdb = mlock_dereference(br->mdb, br);
1664         if (!mdb)
1665                 goto out;
1666
1667         br->mdb = NULL;
1668
1669         ver = mdb->ver;
1670         for (i = 0; i < mdb->max; i++) {
1671                 hlist_for_each_entry_safe(mp, n, &mdb->mhash[i],
1672                                           hlist[ver]) {
1673                         del_timer(&mp->timer);
1674                         call_rcu_bh(&mp->rcu, br_multicast_free_group);
1675                 }
1676         }
1677
1678         if (mdb->old) {
1679                 spin_unlock_bh(&br->multicast_lock);
1680                 rcu_barrier_bh();
1681                 spin_lock_bh(&br->multicast_lock);
1682                 WARN_ON(mdb->old);
1683         }
1684
1685         mdb->old = mdb;
1686         call_rcu_bh(&mdb->rcu, br_mdb_free);
1687
1688 out:
1689         spin_unlock_bh(&br->multicast_lock);
1690 }
1691
1692 int br_multicast_set_router(struct net_bridge *br, unsigned long val)
1693 {
1694         int err = -ENOENT;
1695
1696         spin_lock_bh(&br->multicast_lock);
1697         if (!netif_running(br->dev))
1698                 goto unlock;
1699
1700         switch (val) {
1701         case 0:
1702         case 2:
1703                 del_timer(&br->multicast_router_timer);
1704                 /* fall through */
1705         case 1:
1706                 br->multicast_router = val;
1707                 err = 0;
1708                 break;
1709
1710         default:
1711                 err = -EINVAL;
1712                 break;
1713         }
1714
1715 unlock:
1716         spin_unlock_bh(&br->multicast_lock);
1717
1718         return err;
1719 }
1720
1721 int br_multicast_set_port_router(struct net_bridge_port *p, unsigned long val)
1722 {
1723         struct net_bridge *br = p->br;
1724         int err = -ENOENT;
1725
1726         spin_lock(&br->multicast_lock);
1727         if (!netif_running(br->dev) || p->state == BR_STATE_DISABLED)
1728                 goto unlock;
1729
1730         switch (val) {
1731         case 0:
1732         case 1:
1733         case 2:
1734                 p->multicast_router = val;
1735                 err = 0;
1736
1737                 if (val < 2 && !hlist_unhashed(&p->rlist))
1738                         hlist_del_init_rcu(&p->rlist);
1739
1740                 if (val == 1)
1741                         break;
1742
1743                 del_timer(&p->multicast_router_timer);
1744
1745                 if (val == 0)
1746                         break;
1747
1748                 br_multicast_add_router(br, p);
1749                 break;
1750
1751         default:
1752                 err = -EINVAL;
1753                 break;
1754         }
1755
1756 unlock:
1757         spin_unlock(&br->multicast_lock);
1758
1759         return err;
1760 }
1761
1762 static void br_multicast_start_querier(struct net_bridge *br)
1763 {
1764         struct net_bridge_port *port;
1765
1766         br_multicast_open(br);
1767
1768         list_for_each_entry(port, &br->port_list, list) {
1769                 if (port->state == BR_STATE_DISABLED ||
1770                     port->state == BR_STATE_BLOCKING)
1771                         continue;
1772
1773                 __br_multicast_enable_port(port);
1774         }
1775 }
1776
1777 int br_multicast_toggle(struct net_bridge *br, unsigned long val)
1778 {
1779         int err = 0;
1780         struct net_bridge_mdb_htable *mdb;
1781
1782         spin_lock_bh(&br->multicast_lock);
1783         if (br->multicast_disabled == !val)
1784                 goto unlock;
1785
1786         br->multicast_disabled = !val;
1787         if (br->multicast_disabled)
1788                 goto unlock;
1789
1790         if (!netif_running(br->dev))
1791                 goto unlock;
1792
1793         mdb = mlock_dereference(br->mdb, br);
1794         if (mdb) {
1795                 if (mdb->old) {
1796                         err = -EEXIST;
1797 rollback:
1798                         br->multicast_disabled = !!val;
1799                         goto unlock;
1800                 }
1801
1802                 err = br_mdb_rehash(&br->mdb, mdb->max,
1803                                     br->hash_elasticity);
1804                 if (err)
1805                         goto rollback;
1806         }
1807
1808         br_multicast_start_querier(br);
1809
1810 unlock:
1811         spin_unlock_bh(&br->multicast_lock);
1812
1813         return err;
1814 }
1815
1816 int br_multicast_set_querier(struct net_bridge *br, unsigned long val)
1817 {
1818         val = !!val;
1819
1820         spin_lock_bh(&br->multicast_lock);
1821         if (br->multicast_querier == val)
1822                 goto unlock;
1823
1824         br->multicast_querier = val;
1825         if (val)
1826                 br_multicast_start_querier(br);
1827
1828 unlock:
1829         spin_unlock_bh(&br->multicast_lock);
1830
1831         return 0;
1832 }
1833
1834 int br_multicast_set_hash_max(struct net_bridge *br, unsigned long val)
1835 {
1836         int err = -ENOENT;
1837         u32 old;
1838         struct net_bridge_mdb_htable *mdb;
1839
1840         spin_lock(&br->multicast_lock);
1841         if (!netif_running(br->dev))
1842                 goto unlock;
1843
1844         err = -EINVAL;
1845         if (!is_power_of_2(val))
1846                 goto unlock;
1847
1848         mdb = mlock_dereference(br->mdb, br);
1849         if (mdb && val < mdb->size)
1850                 goto unlock;
1851
1852         err = 0;
1853
1854         old = br->hash_max;
1855         br->hash_max = val;
1856
1857         if (mdb) {
1858                 if (mdb->old) {
1859                         err = -EEXIST;
1860 rollback:
1861                         br->hash_max = old;
1862                         goto unlock;
1863                 }
1864
1865                 err = br_mdb_rehash(&br->mdb, br->hash_max,
1866                                     br->hash_elasticity);
1867                 if (err)
1868                         goto rollback;
1869         }
1870
1871 unlock:
1872         spin_unlock(&br->multicast_lock);
1873
1874         return err;
1875 }