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 #if IS_ENABLED(CONFIG_IPV6)
  40 static inline int ipv6_is_transient_multicast(const struct in6_addr *addr)
  41 {
  42         if (ipv6_addr_is_multicast(addr) && IPV6_ADDR_MC_FLAG_TRANSIENT(addr))
  43                 return 1;
  44         return 0;
  45 }
  46 #endif
  47
  48 static inline int br_ip_equal(const struct br_ip *a, const struct br_ip *b)
  49 {
  50         if (a->proto != b->proto)
  51                 return 0;
  52         switch (a->proto) {
  53         case htons(ETH_P_IP):
  54                 return a->u.ip4 == b->u.ip4;
  55 #if IS_ENABLED(CONFIG_IPV6)
  56         case htons(ETH_P_IPV6):
  57                 return ipv6_addr_equal(&a->u.ip6, &b->u.ip6);
  58 #endif
  59         }
  60         return 0;
  61 }
  62
  63 static inline int __br_ip4_hash(struct net_bridge_mdb_htable *mdb, __be32 ip)
  64 {
  65         return jhash_1word(mdb->secret, (__force u32)ip) & (mdb->max - 1);
  66 }
  67
  68 #if IS_ENABLED(CONFIG_IPV6)
  69 static inline int __br_ip6_hash(struct net_bridge_mdb_htable *mdb,
  70                                 const struct in6_addr *ip)
  71 {
  72         return jhash2((__force u32 *)ip->s6_addr32, 4, mdb->secret) & (mdb->max - 1);
  73 }
  74 #endif
  75
  76 static inline int br_ip_hash(struct net_bridge_mdb_htable *mdb,
  77                              struct br_ip *ip)
  78 {
  79         switch (ip->proto) {
  80         case htons(ETH_P_IP):
  81                 return __br_ip4_hash(mdb, ip->u.ip4);
  82 #if IS_ENABLED(CONFIG_IPV6)
  83         case htons(ETH_P_IPV6):
  84                 return __br_ip6_hash(mdb, &ip->u.ip6);
  85 #endif
  86         }
  87         return 0;
  88 }
  89
  90 static struct net_bridge_mdb_entry *__br_mdb_ip_get(
  91         struct net_bridge_mdb_htable *mdb, struct br_ip *dst, int hash)
  92 {
  93         struct net_bridge_mdb_entry *mp;
  94         struct hlist_node *p;
  95
  96         hlist_for_each_entry_rcu(mp, p, &mdb->mhash[hash], hlist[mdb->ver]) {
  97                 if (br_ip_equal(&mp->addr, dst))
  98                         return mp;
  99         }
 100
 101         return NULL;
 102 }
 103
 104 static struct net_bridge_mdb_entry *br_mdb_ip_get(
 105         struct net_bridge_mdb_htable *mdb, struct br_ip *dst)
 106 {
 107         if (!mdb)
 108                 return NULL;
 109
 110         return __br_mdb_ip_get(mdb, dst, br_ip_hash(mdb, dst));
 111 }
 112
 113 static struct net_bridge_mdb_entry *br_mdb_ip4_get(
 114         struct net_bridge_mdb_htable *mdb, __be32 dst)
 115 {
 116         struct br_ip br_dst;
 117
 118         br_dst.u.ip4 = dst;
 119         br_dst.proto = htons(ETH_P_IP);
 120
 121         return br_mdb_ip_get(mdb, &br_dst);
 122 }
 123
 124 #if IS_ENABLED(CONFIG_IPV6)
 125 static struct net_bridge_mdb_entry *br_mdb_ip6_get(
 126         struct net_bridge_mdb_htable *mdb, const struct in6_addr *dst)
 127 {
 128         struct br_ip br_dst;
 129
 130         br_dst.u.ip6 = *dst;
 131         br_dst.proto = htons(ETH_P_IPV6);
 132
 133         return br_mdb_ip_get(mdb, &br_dst);
 134 }
 135 #endif
 136
 137 struct net_bridge_mdb_entry *br_mdb_get(struct net_bridge *br,
 138                                         struct sk_buff *skb)
 139 {
 140         struct net_bridge_mdb_htable *mdb = rcu_dereference(br->mdb);
 141         struct br_ip ip;
 142
 143         if (br->multicast_disabled)
 144                 return NULL;
 145
 146         if (BR_INPUT_SKB_CB(skb)->igmp)
 147                 return NULL;
 148
 149         ip.proto = skb->protocol;
 150
 151         switch (skb->protocol) {
 152         case htons(ETH_P_IP):
 153                 ip.u.ip4 = ip_hdr(skb)->daddr;
 154                 break;
 155 #if IS_ENABLED(CONFIG_IPV6)
 156         case htons(ETH_P_IPV6):
 157                 ip.u.ip6 = ipv6_hdr(skb)->daddr;
 158                 break;
 159 #endif
 160         default:
 161                 return NULL;
 162         }
 163
 164         return br_mdb_ip_get(mdb, &ip);
 165 }
 166
 167 static void br_mdb_free(struct rcu_head *head)
 168 {
 169         struct net_bridge_mdb_htable *mdb =
 170                 container_of(head, struct net_bridge_mdb_htable, rcu);
 171         struct net_bridge_mdb_htable *old = mdb->old;
 172
 173         mdb->old = NULL;
 174         kfree(old->mhash);
 175         kfree(old);
 176 }
 177
 178 static int br_mdb_copy(struct net_bridge_mdb_htable *new,
 179                        struct net_bridge_mdb_htable *old,
 180                        int elasticity)
 181 {
 182         struct net_bridge_mdb_entry *mp;
 183         struct hlist_node *p;
 184         int maxlen;
 185         int len;
 186         int i;
 187
 188         for (i = 0; i < old->max; i++)
 189                 hlist_for_each_entry(mp, p, &old->mhash[i], hlist[old->ver])
 190                         hlist_add_head(&mp->hlist[new->ver],
 191                                        &new->mhash[br_ip_hash(new, &mp->addr)]);
 192
 193         if (!elasticity)
 194                 return 0;
 195
 196         maxlen = 0;
 197         for (i = 0; i < new->max; i++) {
 198                 len = 0;
 199                 hlist_for_each_entry(mp, p, &new->mhash[i], hlist[new->ver])
 200                         len++;
 201                 if (len > maxlen)
 202                         maxlen = len;
 203         }
 204
 205         return maxlen > elasticity ? -EINVAL : 0;
 206 }
 207
 208 static void br_multicast_free_pg(struct rcu_head *head)
 209 {
 210         struct net_bridge_port_group *p =
 211                 container_of(head, struct net_bridge_port_group, rcu);
 212
 213         kfree(p);
 214 }
 215
 216 static void br_multicast_free_group(struct rcu_head *head)
 217 {
 218         struct net_bridge_mdb_entry *mp =
 219                 container_of(head, struct net_bridge_mdb_entry, rcu);
 220
 221         kfree(mp);
 222 }
 223
 224 static void br_multicast_group_expired(unsigned long data)
 225 {
 226         struct net_bridge_mdb_entry *mp = (void *)data;
 227         struct net_bridge *br = mp->br;
 228         struct net_bridge_mdb_htable *mdb;
 229
 230         spin_lock(&br->multicast_lock);
 231         if (!netif_running(br->dev) || timer_pending(&mp->timer))
 232                 goto out;
 233
 234         mp->mglist = false;
 235
 236         if (mp->ports)
 237                 goto out;
 238
 239         mdb = mlock_dereference(br->mdb, br);
 240
 241         hlist_del_rcu(&mp->hlist[mdb->ver]);
 242         mdb->size--;
 243
 244         call_rcu_bh(&mp->rcu, br_multicast_free_group);
 245
 246 out:
 247         spin_unlock(&br->multicast_lock);
 248 }
 249
 250 static void br_multicast_del_pg(struct net_bridge *br,
 251                                 struct net_bridge_port_group *pg)
 252 {
 253         struct net_bridge_mdb_htable *mdb;
 254         struct net_bridge_mdb_entry *mp;
 255         struct net_bridge_port_group *p;
 256         struct net_bridge_port_group __rcu **pp;
 257
 258         mdb = mlock_dereference(br->mdb, br);
 259
 260         mp = br_mdb_ip_get(mdb, &pg->addr);
 261         if (WARN_ON(!mp))
 262                 return;
 263
 264         for (pp = &mp->ports;
 265              (p = mlock_dereference(*pp, br)) != NULL;
 266              pp = &p->next) {
 267                 if (p != pg)
 268                         continue;
 269
 270                 rcu_assign_pointer(*pp, p->next);
 271                 hlist_del_init(&p->mglist);
 272                 del_timer(&p->timer);
 273                 call_rcu_bh(&p->rcu, br_multicast_free_pg);
 274
 275                 if (!mp->ports && !mp->mglist &&
 276                     netif_running(br->dev))
 277                         mod_timer(&mp->timer, jiffies);
 278
 279                 return;
 280         }
 281
 282         WARN_ON(1);
 283 }
 284
 285 static void br_multicast_port_group_expired(unsigned long data)
 286 {
 287         struct net_bridge_port_group *pg = (void *)data;
 288         struct net_bridge *br = pg->port->br;
 289
 290         spin_lock(&br->multicast_lock);
 291         if (!netif_running(br->dev) || timer_pending(&pg->timer) ||
 292             hlist_unhashed(&pg->mglist))
 293                 goto out;
 294
 295         br_multicast_del_pg(br, pg);
 296
 297 out:
 298         spin_unlock(&br->multicast_lock);
 299 }
 300
 301 static int br_mdb_rehash(struct net_bridge_mdb_htable __rcu **mdbp, int max,
 302                          int elasticity)
 303 {
 304         struct net_bridge_mdb_htable *old = rcu_dereference_protected(*mdbp, 1);
 305         struct net_bridge_mdb_htable *mdb;
 306         int err;
 307
 308         mdb = kmalloc(sizeof(*mdb), GFP_ATOMIC);
 309         if (!mdb)
 310                 return -ENOMEM;
 311
 312         mdb->max = max;
 313         mdb->old = old;
 314
 315         mdb->mhash = kzalloc(max * sizeof(*mdb->mhash), GFP_ATOMIC);
 316         if (!mdb->mhash) {
 317                 kfree(mdb);
 318                 return -ENOMEM;
 319         }
 320
 321         mdb->size = old ? old->size : 0;
 322         mdb->ver = old ? old->ver ^ 1 : 0;
 323
 324         if (!old || elasticity)
 325                 get_random_bytes(&mdb->secret, sizeof(mdb->secret));
 326         else
 327                 mdb->secret = old->secret;
 328
 329         if (!old)
 330                 goto out;
 331
 332         err = br_mdb_copy(mdb, old, elasticity);
 333         if (err) {
 334                 kfree(mdb->mhash);
 335                 kfree(mdb);
 336                 return err;
 337         }
 338
 339         call_rcu_bh(&mdb->rcu, br_mdb_free);
 340
 341 out:
 342         rcu_assign_pointer(*mdbp, mdb);
 343
 344         return 0;
 345 }
 346
 347 static struct sk_buff *br_ip4_multicast_alloc_query(struct net_bridge *br,
 348                                                     __be32 group)
 349 {
 350         struct sk_buff *skb;
 351         struct igmphdr *ih;
 352         struct ethhdr *eth;
 353         struct iphdr *iph;
 354
 355         skb = netdev_alloc_skb_ip_align(br->dev, sizeof(*eth) + sizeof(*iph) +
 356                                                  sizeof(*ih) + 4);
 357         if (!skb)
 358                 goto out;
 359
 360         skb->protocol = htons(ETH_P_IP);
 361
 362         skb_reset_mac_header(skb);
 363         eth = eth_hdr(skb);
 364
 365         memcpy(eth->h_source, br->dev->dev_addr, 6);
 366         eth->h_dest[0] = 1;
 367         eth->h_dest[1] = 0;
 368         eth->h_dest[2] = 0x5e;
 369         eth->h_dest[3] = 0;
 370         eth->h_dest[4] = 0;
 371         eth->h_dest[5] = 1;
 372         eth->h_proto = htons(ETH_P_IP);
 373         skb_put(skb, sizeof(*eth));
 374
 375         skb_set_network_header(skb, skb->len);
 376         iph = ip_hdr(skb);
 377
 378         iph->version = 4;
 379         iph->ihl = 6;
 380         iph->tos = 0xc0;
 381         iph->tot_len = htons(sizeof(*iph) + sizeof(*ih) + 4);
 382         iph->id = 0;
 383         iph->frag_off = htons(IP_DF);
 384         iph->ttl = 1;
 385         iph->protocol = IPPROTO_IGMP;
 386         iph->saddr = 0;
 387         iph->daddr = htonl(INADDR_ALLHOSTS_GROUP);
 388         ((u8 *)&iph[1])[0] = IPOPT_RA;
 389         ((u8 *)&iph[1])[1] = 4;
 390         ((u8 *)&iph[1])[2] = 0;
 391         ((u8 *)&iph[1])[3] = 0;
 392         ip_send_check(iph);
 393         skb_put(skb, 24);
 394
 395         skb_set_transport_header(skb, skb->len);
 396         ih = igmp_hdr(skb);
 397         ih->type = IGMP_HOST_MEMBERSHIP_QUERY;
 398         ih->code = (group ? br->multicast_last_member_interval :
 399                             br->multicast_query_response_interval) /
 400                    (HZ / IGMP_TIMER_SCALE);
 401         ih->group = group;
 402         ih->csum = 0;
 403         ih->csum = ip_compute_csum((void *)ih, sizeof(struct igmphdr));
 404         skb_put(skb, sizeof(*ih));
 405
 406         __skb_pull(skb, sizeof(*eth));
 407
 408 out:
 409         return skb;
 410 }
 411
 412 #if IS_ENABLED(CONFIG_IPV6)
 413 static struct sk_buff *br_ip6_multicast_alloc_query(struct net_bridge *br,
 414                                                     const struct in6_addr *group)
 415 {
 416         struct sk_buff *skb;
 417         struct ipv6hdr *ip6h;
 418         struct mld_msg *mldq;
 419         struct ethhdr *eth;
 420         u8 *hopopt;
 421         unsigned long interval;
 422
 423         skb = netdev_alloc_skb_ip_align(br->dev, sizeof(*eth) + sizeof(*ip6h) +
 424                                                  8 + sizeof(*mldq));
 425         if (!skb)
 426                 goto out;
 427
 428         skb->protocol = htons(ETH_P_IPV6);
 429
 430         /* Ethernet header */
 431         skb_reset_mac_header(skb);
 432         eth = eth_hdr(skb);
 433
 434         memcpy(eth->h_source, br->dev->dev_addr, 6);
 435         eth->h_proto = htons(ETH_P_IPV6);
 436         skb_put(skb, sizeof(*eth));
 437
 438         /* IPv6 header + HbH option */
 439         skb_set_network_header(skb, skb->len);
 440         ip6h = ipv6_hdr(skb);
 441
 442         *(__force __be32 *)ip6h = htonl(0x60000000);
 443         ip6h->payload_len = htons(8 + sizeof(*mldq));
 444         ip6h->nexthdr = IPPROTO_HOPOPTS;
 445         ip6h->hop_limit = 1;
 446         ipv6_addr_set(&ip6h->daddr, htonl(0xff020000), 0, 0, htonl(1));
 447         if (ipv6_dev_get_saddr(dev_net(br->dev), br->dev, &ip6h->daddr, 0,
 448                                &ip6h->saddr)) {
 449                 kfree_skb(skb);
 450                 return NULL;
 451         }
 452         ipv6_eth_mc_map(&ip6h->daddr, eth->h_dest);
 453
 454         hopopt = (u8 *)(ip6h + 1);
 455         hopopt[0] = IPPROTO_ICMPV6;             /* next hdr */
 456         hopopt[1] = 0;                          /* length of HbH */
 457         hopopt[2] = IPV6_TLV_ROUTERALERT;       /* Router Alert */
 458         hopopt[3] = 2;                          /* Length of RA Option */
 459         hopopt[4] = 0;                          /* Type = 0x0000 (MLD) */
 460         hopopt[5] = 0;
 461         hopopt[6] = IPV6_TLV_PAD0;              /* Pad0 */
 462         hopopt[7] = IPV6_TLV_PAD0;              /* Pad0 */
 463
 464         skb_put(skb, sizeof(*ip6h) + 8);
 465
 466         /* ICMPv6 */
 467         skb_set_transport_header(skb, skb->len);
 468         mldq = (struct mld_msg *) icmp6_hdr(skb);
 469
 470         interval = ipv6_addr_any(group) ? br->multicast_last_member_interval :
 471                                           br->multicast_query_response_interval;
 472
 473         mldq->mld_type = ICMPV6_MGM_QUERY;
 474         mldq->mld_code = 0;
 475         mldq->mld_cksum = 0;
 476         mldq->mld_maxdelay = htons((u16)jiffies_to_msecs(interval));
 477         mldq->mld_reserved = 0;
 478         mldq->mld_mca = *group;
 479
 480         /* checksum */
 481         mldq->mld_cksum = csum_ipv6_magic(&ip6h->saddr, &ip6h->daddr,
 482                                           sizeof(*mldq), IPPROTO_ICMPV6,
 483                                           csum_partial(mldq,
 484                                                        sizeof(*mldq), 0));
 485         skb_put(skb, sizeof(*mldq));
 486
 487         __skb_pull(skb, sizeof(*eth));
 488
 489 out:
 490         return skb;
 491 }
 492 #endif
 493
 494 static struct sk_buff *br_multicast_alloc_query(struct net_bridge *br,
 495                                                 struct br_ip *addr)
 496 {
 497         switch (addr->proto) {
 498         case htons(ETH_P_IP):
 499                 return br_ip4_multicast_alloc_query(br, addr->u.ip4);
 500 #if IS_ENABLED(CONFIG_IPV6)
 501         case htons(ETH_P_IPV6):
 502                 return br_ip6_multicast_alloc_query(br, &addr->u.ip6);
 503 #endif
 504         }
 505         return NULL;
 506 }
 507
 508 static struct net_bridge_mdb_entry *br_multicast_get_group(
 509         struct net_bridge *br, struct net_bridge_port *port,
 510         struct br_ip *group, int hash)
 511 {
 512         struct net_bridge_mdb_htable *mdb;
 513         struct net_bridge_mdb_entry *mp;
 514         struct hlist_node *p;
 515         unsigned int count = 0;
 516         unsigned int max;
 517         int elasticity;
 518         int err;
 519
 520         mdb = rcu_dereference_protected(br->mdb, 1);
 521         hlist_for_each_entry(mp, p, &mdb->mhash[hash], hlist[mdb->ver]) {
 522                 count++;
 523                 if (unlikely(br_ip_equal(group, &mp->addr)))
 524                         return mp;
 525         }
 526
 527         elasticity = 0;
 528         max = mdb->max;
 529
 530         if (unlikely(count > br->hash_elasticity && count)) {
 531                 if (net_ratelimit())
 532                         br_info(br, "Multicast hash table "
 533                                 "chain limit reached: %s\n",
 534                                 port ? port->dev->name : br->dev->name);
 535
 536                 elasticity = br->hash_elasticity;
 537         }
 538
 539         if (mdb->size >= max) {
 540                 max *= 2;
 541                 if (unlikely(max >= br->hash_max)) {
 542                         br_warn(br, "Multicast hash table maximum "
 543                                 "reached, disabling snooping: %s, %d\n",
 544                                 port ? port->dev->name : br->dev->name, max);
 545                         err = -E2BIG;
 546 disable:
 547                         br->multicast_disabled = 1;
 548                         goto err;
 549                 }
 550         }
 551
 552         if (max > mdb->max || elasticity) {
 553                 if (mdb->old) {
 554                         if (net_ratelimit())
 555                                 br_info(br, "Multicast hash table "
 556                                         "on fire: %s\n",
 557                                         port ? port->dev->name : br->dev->name);
 558                         err = -EEXIST;
 559                         goto err;
 560                 }
 561
 562                 err = br_mdb_rehash(&br->mdb, max, elasticity);
 563                 if (err) {
 564                         br_warn(br, "Cannot rehash multicast "
 565                                 "hash table, disabling snooping: %s, %d, %d\n",
 566                                 port ? port->dev->name : br->dev->name,
 567                                 mdb->size, err);
 568                         goto disable;
 569                 }
 570
 571                 err = -EAGAIN;
 572                 goto err;
 573         }
 574
 575         return NULL;
 576
 577 err:
 578         mp = ERR_PTR(err);
 579         return mp;
 580 }
 581
 582 static struct net_bridge_mdb_entry *br_multicast_new_group(
 583         struct net_bridge *br, struct net_bridge_port *port,
 584         struct br_ip *group)
 585 {
 586         struct net_bridge_mdb_htable *mdb;
 587         struct net_bridge_mdb_entry *mp;
 588         int hash;
 589         int err;
 590
 591         mdb = rcu_dereference_protected(br->mdb, 1);
 592         if (!mdb) {
 593                 err = br_mdb_rehash(&br->mdb, BR_HASH_SIZE, 0);
 594                 if (err)
 595                         return ERR_PTR(err);
 596                 goto rehash;
 597         }
 598
 599         hash = br_ip_hash(mdb, group);
 600         mp = br_multicast_get_group(br, port, group, hash);
 601         switch (PTR_ERR(mp)) {
 602         case 0:
 603                 break;
 604
 605         case -EAGAIN:
 606 rehash:
 607                 mdb = rcu_dereference_protected(br->mdb, 1);
 608                 hash = br_ip_hash(mdb, group);
 609                 break;
 610
 611         default:
 612                 goto out;
 613         }
 614
 615         mp = kzalloc(sizeof(*mp), GFP_ATOMIC);
 616         if (unlikely(!mp))
 617                 return ERR_PTR(-ENOMEM);
 618
 619         mp->br = br;
 620         mp->addr = *group;
 621         setup_timer(&mp->timer, br_multicast_group_expired,
 622                     (unsigned long)mp);
 623
 624         hlist_add_head_rcu(&mp->hlist[mdb->ver], &mdb->mhash[hash]);
 625         mdb->size++;
 626
 627 out:
 628         return mp;
 629 }
 630
 631 static int br_multicast_add_group(struct net_bridge *br,
 632                                   struct net_bridge_port *port,
 633                                   struct br_ip *group)
 634 {
 635         struct net_bridge_mdb_entry *mp;
 636         struct net_bridge_port_group *p;
 637         struct net_bridge_port_group __rcu **pp;
 638         unsigned long now = jiffies;
 639         int err;
 640
 641         spin_lock(&br->multicast_lock);
 642         if (!netif_running(br->dev) ||
 643             (port && port->state == BR_STATE_DISABLED))
 644                 goto out;
 645
 646         mp = br_multicast_new_group(br, port, group);
 647         err = PTR_ERR(mp);
 648         if (IS_ERR(mp))
 649                 goto err;
 650
 651         if (!port) {
 652                 mp->mglist = true;
 653                 mod_timer(&mp->timer, now + br->multicast_membership_interval);
 654                 goto out;
 655         }
 656
 657         for (pp = &mp->ports;
 658              (p = mlock_dereference(*pp, br)) != NULL;
 659              pp = &p->next) {
 660                 if (p->port == port)
 661                         goto found;
 662                 if ((unsigned long)p->port < (unsigned long)port)
 663                         break;
 664         }
 665
 666         p = kzalloc(sizeof(*p), GFP_ATOMIC);
 667         err = -ENOMEM;
 668         if (unlikely(!p))
 669                 goto err;
 670
 671         p->addr = *group;
 672         p->port = port;
 673         p->next = *pp;
 674         hlist_add_head(&p->mglist, &port->mglist);
 675         setup_timer(&p->timer, br_multicast_port_group_expired,
 676                     (unsigned long)p);
 677
 678         rcu_assign_pointer(*pp, p);
 679
 680 found:
 681         mod_timer(&p->timer, now + br->multicast_membership_interval);
 682 out:
 683         err = 0;
 684
 685 err:
 686         spin_unlock(&br->multicast_lock);
 687         return err;
 688 }
 689
 690 static int br_ip4_multicast_add_group(struct net_bridge *br,
 691                                       struct net_bridge_port *port,
 692                                       __be32 group)
 693 {
 694         struct br_ip br_group;
 695
 696         if (ipv4_is_local_multicast(group))
 697                 return 0;
 698
 699         br_group.u.ip4 = group;
 700         br_group.proto = htons(ETH_P_IP);
 701
 702         return br_multicast_add_group(br, port, &br_group);
 703 }
 704
 705 #if IS_ENABLED(CONFIG_IPV6)
 706 static int br_ip6_multicast_add_group(struct net_bridge *br,
 707                                       struct net_bridge_port *port,
 708                                       const struct in6_addr *group)
 709 {
 710         struct br_ip br_group;
 711
 712         if (!ipv6_is_transient_multicast(group))
 713                 return 0;
 714
 715         br_group.u.ip6 = *group;
 716         br_group.proto = htons(ETH_P_IPV6);
 717
 718         return br_multicast_add_group(br, port, &br_group);
 719 }
 720 #endif
 721
 722 static void br_multicast_router_expired(unsigned long data)
 723 {
 724         struct net_bridge_port *port = (void *)data;
 725         struct net_bridge *br = port->br;
 726
 727         spin_lock(&br->multicast_lock);
 728         if (port->multicast_router != 1 ||
 729             timer_pending(&port->multicast_router_timer) ||
 730             hlist_unhashed(&port->rlist))
 731                 goto out;
 732
 733         hlist_del_init_rcu(&port->rlist);
 734
 735 out:
 736         spin_unlock(&br->multicast_lock);
 737 }
 738
 739 static void br_multicast_local_router_expired(unsigned long data)
 740 {
 741 }
 742
 743 static void __br_multicast_send_query(struct net_bridge *br,
 744                                       struct net_bridge_port *port,
 745                                       struct br_ip *ip)
 746 {
 747         struct sk_buff *skb;
 748
 749         skb = br_multicast_alloc_query(br, ip);
 750         if (!skb)
 751                 return;
 752
 753         if (port) {
 754                 __skb_push(skb, sizeof(struct ethhdr));
 755                 skb->dev = port->dev;
 756                 NF_HOOK(NFPROTO_BRIDGE, NF_BR_LOCAL_OUT, skb, NULL, skb->dev,
 757                         dev_queue_xmit);
 758         } else
 759                 netif_rx(skb);
 760 }
 761
 762 static void br_multicast_send_query(struct net_bridge *br,
 763                                     struct net_bridge_port *port, u32 sent)
 764 {
 765         unsigned long time;
 766         struct br_ip br_group;
 767
 768         if (!netif_running(br->dev) || br->multicast_disabled ||
 769             timer_pending(&br->multicast_querier_timer))
 770                 return;
 771
 772         memset(&br_group.u, 0, sizeof(br_group.u));
 773
 774         br_group.proto = htons(ETH_P_IP);
 775         __br_multicast_send_query(br, port, &br_group);
 776
 777 #if IS_ENABLED(CONFIG_IPV6)
 778         br_group.proto = htons(ETH_P_IPV6);
 779         __br_multicast_send_query(br, port, &br_group);
 780 #endif
 781
 782         time = jiffies;
 783         time += sent < br->multicast_startup_query_count ?
 784                 br->multicast_startup_query_interval :
 785                 br->multicast_query_interval;
 786         mod_timer(port ? &port->multicast_query_timer :
 787                          &br->multicast_query_timer, time);
 788 }
 789
 790 static void br_multicast_port_query_expired(unsigned long data)
 791 {
 792         struct net_bridge_port *port = (void *)data;
 793         struct net_bridge *br = port->br;
 794
 795         spin_lock(&br->multicast_lock);
 796         if (port->state == BR_STATE_DISABLED ||
 797             port->state == BR_STATE_BLOCKING)
 798                 goto out;
 799
 800         if (port->multicast_startup_queries_sent <
 801             br->multicast_startup_query_count)
 802                 port->multicast_startup_queries_sent++;
 803
 804         br_multicast_send_query(port->br, port,
 805                                 port->multicast_startup_queries_sent);
 806
 807 out:
 808         spin_unlock(&br->multicast_lock);
 809 }
 810
 811 void br_multicast_add_port(struct net_bridge_port *port)
 812 {
 813         port->multicast_router = 1;
 814
 815         setup_timer(&port->multicast_router_timer, br_multicast_router_expired,
 816                     (unsigned long)port);
 817         setup_timer(&port->multicast_query_timer,
 818                     br_multicast_port_query_expired, (unsigned long)port);
 819 }
 820
 821 void br_multicast_del_port(struct net_bridge_port *port)
 822 {
 823         del_timer_sync(&port->multicast_router_timer);
 824 }
 825
 826 static void __br_multicast_enable_port(struct net_bridge_port *port)
 827 {
 828         port->multicast_startup_queries_sent = 0;
 829
 830         if (try_to_del_timer_sync(&port->multicast_query_timer) >= 0 ||
 831             del_timer(&port->multicast_query_timer))
 832                 mod_timer(&port->multicast_query_timer, jiffies);
 833 }
 834
 835 void br_multicast_enable_port(struct net_bridge_port *port)
 836 {
 837         struct net_bridge *br = port->br;
 838
 839         spin_lock(&br->multicast_lock);
 840         if (br->multicast_disabled || !netif_running(br->dev))
 841                 goto out;
 842
 843         __br_multicast_enable_port(port);
 844
 845 out:
 846         spin_unlock(&br->multicast_lock);
 847 }
 848
 849 void br_multicast_disable_port(struct net_bridge_port *port)
 850 {
 851         struct net_bridge *br = port->br;
 852         struct net_bridge_port_group *pg;
 853         struct hlist_node *p, *n;
 854
 855         spin_lock(&br->multicast_lock);
 856         hlist_for_each_entry_safe(pg, p, n, &port->mglist, mglist)
 857                 br_multicast_del_pg(br, pg);
 858
 859         if (!hlist_unhashed(&port->rlist))
 860                 hlist_del_init_rcu(&port->rlist);
 861         del_timer(&port->multicast_router_timer);
 862         del_timer(&port->multicast_query_timer);
 863         spin_unlock(&br->multicast_lock);
 864 }
 865
 866 static int br_ip4_multicast_igmp3_report(struct net_bridge *br,
 867                                          struct net_bridge_port *port,
 868                                          struct sk_buff *skb)
 869 {
 870         struct igmpv3_report *ih;
 871         struct igmpv3_grec *grec;
 872         int i;
 873         int len;
 874         int num;
 875         int type;
 876         int err = 0;
 877         __be32 group;
 878
 879         if (!pskb_may_pull(skb, sizeof(*ih)))
 880                 return -EINVAL;
 881
 882         ih = igmpv3_report_hdr(skb);
 883         num = ntohs(ih->ngrec);
 884         len = sizeof(*ih);
 885
 886         for (i = 0; i < num; i++) {
 887                 len += sizeof(*grec);
 888                 if (!pskb_may_pull(skb, len))
 889                         return -EINVAL;
 890
 891                 grec = (void *)(skb->data + len - sizeof(*grec));
 892                 group = grec->grec_mca;
 893                 type = grec->grec_type;
 894
 895                 len += ntohs(grec->grec_nsrcs) * 4;
 896                 if (!pskb_may_pull(skb, len))
 897                         return -EINVAL;
 898
 899                 /* We treat this as an IGMPv2 report for now. */
 900                 switch (type) {
 901                 case IGMPV3_MODE_IS_INCLUDE:
 902                 case IGMPV3_MODE_IS_EXCLUDE:
 903                 case IGMPV3_CHANGE_TO_INCLUDE:
 904                 case IGMPV3_CHANGE_TO_EXCLUDE:
 905                 case IGMPV3_ALLOW_NEW_SOURCES:
 906                 case IGMPV3_BLOCK_OLD_SOURCES:
 907                         break;
 908
 909                 default:
 910                         continue;
 911                 }
 912
 913                 err = br_ip4_multicast_add_group(br, port, group);
 914                 if (err)
 915                         break;
 916         }
 917
 918         return err;
 919 }
 920
 921 #if IS_ENABLED(CONFIG_IPV6)
 922 static int br_ip6_multicast_mld2_report(struct net_bridge *br,
 923                                         struct net_bridge_port *port,
 924                                         struct sk_buff *skb)
 925 {
 926         struct icmp6hdr *icmp6h;
 927         struct mld2_grec *grec;
 928         int i;
 929         int len;
 930         int num;
 931         int err = 0;
 932
 933         if (!pskb_may_pull(skb, sizeof(*icmp6h)))
 934                 return -EINVAL;
 935
 936         icmp6h = icmp6_hdr(skb);
 937         num = ntohs(icmp6h->icmp6_dataun.un_data16[1]);
 938         len = sizeof(*icmp6h);
 939
 940         for (i = 0; i < num; i++) {
 941                 __be16 *nsrcs, _nsrcs;
 942
 943                 nsrcs = skb_header_pointer(skb,
 944                                            len + offsetof(struct mld2_grec,
 945                                                           grec_nsrcs),
 946                                            sizeof(_nsrcs), &_nsrcs);
 947                 if (!nsrcs)
 948                         return -EINVAL;
 949
 950                 if (!pskb_may_pull(skb,
 951                                    len + sizeof(*grec) +
 952                                    sizeof(struct in6_addr) * ntohs(*nsrcs)))
 953                         return -EINVAL;
 954
 955                 grec = (struct mld2_grec *)(skb->data + len);
 956                 len += sizeof(*grec) +
 957                        sizeof(struct in6_addr) * ntohs(*nsrcs);
 958
 959                 /* We treat these as MLDv1 reports for now. */
 960                 switch (grec->grec_type) {
 961                 case MLD2_MODE_IS_INCLUDE:
 962                 case MLD2_MODE_IS_EXCLUDE:
 963                 case MLD2_CHANGE_TO_INCLUDE:
 964                 case MLD2_CHANGE_TO_EXCLUDE:
 965                 case MLD2_ALLOW_NEW_SOURCES:
 966                 case MLD2_BLOCK_OLD_SOURCES:
 967                         break;
 968
 969                 default:
 970                         continue;
 971                 }
 972
 973                 err = br_ip6_multicast_add_group(br, port, &grec->grec_mca);
 974                 if (!err)
 975                         break;
 976         }
 977
 978         return err;
 979 }
 980 #endif
 981
 982 /*
 983  * Add port to rotuer_list
 984  *  list is maintained ordered by pointer value
 985  *  and locked by br->multicast_lock and RCU
 986  */
 987 static void br_multicast_add_router(struct net_bridge *br,
 988                                     struct net_bridge_port *port)
 989 {
 990         struct net_bridge_port *p;
 991         struct hlist_node *n, *slot = NULL;
 992
 993         hlist_for_each_entry(p, n, &br->router_list, rlist) {
 994                 if ((unsigned long) port >= (unsigned long) p)
 995                         break;
 996                 slot = n;
 997         }
 998
 999         if (slot)
1000                 hlist_add_after_rcu(slot, &port->rlist);
1001         else
1002                 hlist_add_head_rcu(&port->rlist, &br->router_list);
1003 }
1004
1005 static void br_multicast_mark_router(struct net_bridge *br,
1006                                      struct net_bridge_port *port)
1007 {
1008         unsigned long now = jiffies;
1009
1010         if (!port) {
1011                 if (br->multicast_router == 1)
1012                         mod_timer(&br->multicast_router_timer,
1013                                   now + br->multicast_querier_interval);
1014                 return;
1015         }
1016
1017         if (port->multicast_router != 1)
1018                 return;
1019
1020         if (!hlist_unhashed(&port->rlist))
1021                 goto timer;
1022
1023         br_multicast_add_router(br, port);
1024
1025 timer:
1026         mod_timer(&port->multicast_router_timer,
1027                   now + br->multicast_querier_interval);
1028 }
1029
1030 static void br_multicast_query_received(struct net_bridge *br,
1031                                         struct net_bridge_port *port,
1032                                         int saddr)
1033 {
1034         if (saddr)
1035                 mod_timer(&br->multicast_querier_timer,
1036                           jiffies + br->multicast_querier_interval);
1037         else if (timer_pending(&br->multicast_querier_timer))
1038                 return;
1039
1040         br_multicast_mark_router(br, port);
1041 }
1042
1043 static int br_ip4_multicast_query(struct net_bridge *br,
1044                                   struct net_bridge_port *port,
1045                                   struct sk_buff *skb)
1046 {
1047         const struct iphdr *iph = ip_hdr(skb);
1048         struct igmphdr *ih = igmp_hdr(skb);
1049         struct net_bridge_mdb_entry *mp;
1050         struct igmpv3_query *ih3;
1051         struct net_bridge_port_group *p;
1052         struct net_bridge_port_group __rcu **pp;
1053         unsigned long max_delay;
1054         unsigned long now = jiffies;
1055         __be32 group;
1056         int err = 0;
1057
1058         spin_lock(&br->multicast_lock);
1059         if (!netif_running(br->dev) ||
1060             (port && port->state == BR_STATE_DISABLED))
1061                 goto out;
1062
1063         br_multicast_query_received(br, port, !!iph->saddr);
1064
1065         group = ih->group;
1066
1067         if (skb->len == sizeof(*ih)) {
1068                 max_delay = ih->code * (HZ / IGMP_TIMER_SCALE);
1069
1070                 if (!max_delay) {
1071                         max_delay = 10 * HZ;
1072                         group = 0;
1073                 }
1074         } else {
1075                 if (!pskb_may_pull(skb, sizeof(struct igmpv3_query))) {
1076                         err = -EINVAL;
1077                         goto out;
1078                 }
1079
1080                 ih3 = igmpv3_query_hdr(skb);
1081                 if (ih3->nsrcs)
1082                         goto out;
1083
1084                 max_delay = ih3->code ?
1085                             IGMPV3_MRC(ih3->code) * (HZ / IGMP_TIMER_SCALE) : 1;
1086         }
1087
1088         if (!group)
1089                 goto out;
1090
1091         mp = br_mdb_ip4_get(mlock_dereference(br->mdb, br), group);
1092         if (!mp)
1093                 goto out;
1094
1095         max_delay *= br->multicast_last_member_count;
1096
1097         if (mp->mglist &&
1098             (timer_pending(&mp->timer) ?
1099              time_after(mp->timer.expires, now + max_delay) :
1100              try_to_del_timer_sync(&mp->timer) >= 0))
1101                 mod_timer(&mp->timer, now + max_delay);
1102
1103         for (pp = &mp->ports;
1104              (p = mlock_dereference(*pp, br)) != NULL;
1105              pp = &p->next) {
1106                 if (timer_pending(&p->timer) ?
1107                     time_after(p->timer.expires, now + max_delay) :
1108                     try_to_del_timer_sync(&p->timer) >= 0)
1109                         mod_timer(&p->timer, now + max_delay);
1110         }
1111
1112 out:
1113         spin_unlock(&br->multicast_lock);
1114         return err;
1115 }
1116
1117 #if IS_ENABLED(CONFIG_IPV6)
1118 static int br_ip6_multicast_query(struct net_bridge *br,
1119                                   struct net_bridge_port *port,
1120                                   struct sk_buff *skb)
1121 {
1122         const struct ipv6hdr *ip6h = ipv6_hdr(skb);
1123         struct mld_msg *mld = (struct mld_msg *) icmp6_hdr(skb);
1124         struct net_bridge_mdb_entry *mp;
1125         struct mld2_query *mld2q;
1126         struct net_bridge_port_group *p;
1127         struct net_bridge_port_group __rcu **pp;
1128         unsigned long max_delay;
1129         unsigned long now = jiffies;
1130         const struct in6_addr *group = NULL;
1131         int err = 0;
1132
1133         spin_lock(&br->multicast_lock);
1134         if (!netif_running(br->dev) ||
1135             (port && port->state == BR_STATE_DISABLED))
1136                 goto out;
1137
1138         br_multicast_query_received(br, port, !ipv6_addr_any(&ip6h->saddr));
1139
1140         if (skb->len == sizeof(*mld)) {
1141                 if (!pskb_may_pull(skb, sizeof(*mld))) {
1142                         err = -EINVAL;
1143                         goto out;
1144                 }
1145                 mld = (struct mld_msg *) icmp6_hdr(skb);
1146                 max_delay = msecs_to_jiffies(htons(mld->mld_maxdelay));
1147                 if (max_delay)
1148                         group = &mld->mld_mca;
1149         } else if (skb->len >= sizeof(*mld2q)) {
1150                 if (!pskb_may_pull(skb, sizeof(*mld2q))) {
1151                         err = -EINVAL;
1152                         goto out;
1153                 }
1154                 mld2q = (struct mld2_query *)icmp6_hdr(skb);
1155                 if (!mld2q->mld2q_nsrcs)
1156                         group = &mld2q->mld2q_mca;
1157                 max_delay = mld2q->mld2q_mrc ? MLDV2_MRC(mld2q->mld2q_mrc) : 1;
1158         }
1159
1160         if (!group)
1161                 goto out;
1162
1163         mp = br_mdb_ip6_get(mlock_dereference(br->mdb, br), group);
1164         if (!mp)
1165                 goto out;
1166
1167         max_delay *= br->multicast_last_member_count;
1168         if (mp->mglist &&
1169             (timer_pending(&mp->timer) ?
1170              time_after(mp->timer.expires, now + max_delay) :
1171              try_to_del_timer_sync(&mp->timer) >= 0))
1172                 mod_timer(&mp->timer, now + max_delay);
1173
1174         for (pp = &mp->ports;
1175              (p = mlock_dereference(*pp, br)) != NULL;
1176              pp = &p->next) {
1177                 if (timer_pending(&p->timer) ?
1178                     time_after(p->timer.expires, now + max_delay) :
1179                     try_to_del_timer_sync(&p->timer) >= 0)
1180                         mod_timer(&p->timer, now + max_delay);
1181         }
1182
1183 out:
1184         spin_unlock(&br->multicast_lock);
1185         return err;
1186 }
1187 #endif
1188
1189 static void br_multicast_leave_group(struct net_bridge *br,
1190                                      struct net_bridge_port *port,
1191                                      struct br_ip *group)
1192 {
1193         struct net_bridge_mdb_htable *mdb;
1194         struct net_bridge_mdb_entry *mp;
1195         struct net_bridge_port_group *p;
1196         unsigned long now;
1197         unsigned long time;
1198
1199         spin_lock(&br->multicast_lock);
1200         if (!netif_running(br->dev) ||
1201             (port && port->state == BR_STATE_DISABLED) ||
1202             timer_pending(&br->multicast_querier_timer))
1203                 goto out;
1204
1205         mdb = mlock_dereference(br->mdb, br);
1206         mp = br_mdb_ip_get(mdb, group);
1207         if (!mp)
1208                 goto out;
1209
1210         now = jiffies;
1211         time = now + br->multicast_last_member_count *
1212                      br->multicast_last_member_interval;
1213
1214         if (!port) {
1215                 if (mp->mglist &&
1216                     (timer_pending(&mp->timer) ?
1217                      time_after(mp->timer.expires, time) :
1218                      try_to_del_timer_sync(&mp->timer) >= 0)) {
1219                         mod_timer(&mp->timer, time);
1220                 }
1221
1222                 goto out;
1223         }
1224
1225         for (p = mlock_dereference(mp->ports, br);
1226              p != NULL;
1227              p = mlock_dereference(p->next, br)) {
1228                 if (p->port != port)
1229                         continue;
1230
1231                 if (!hlist_unhashed(&p->mglist) &&
1232                     (timer_pending(&p->timer) ?
1233                      time_after(p->timer.expires, time) :
1234                      try_to_del_timer_sync(&p->timer) >= 0)) {
1235                         mod_timer(&p->timer, time);
1236                 }
1237
1238                 break;
1239         }
1240
1241 out:
1242         spin_unlock(&br->multicast_lock);
1243 }
1244
1245 static void br_ip4_multicast_leave_group(struct net_bridge *br,
1246                                          struct net_bridge_port *port,
1247                                          __be32 group)
1248 {
1249         struct br_ip br_group;
1250
1251         if (ipv4_is_local_multicast(group))
1252                 return;
1253
1254         br_group.u.ip4 = group;
1255         br_group.proto = htons(ETH_P_IP);
1256
1257         br_multicast_leave_group(br, port, &br_group);
1258 }
1259
1260 #if IS_ENABLED(CONFIG_IPV6)
1261 static void br_ip6_multicast_leave_group(struct net_bridge *br,
1262                                          struct net_bridge_port *port,
1263                                          const struct in6_addr *group)
1264 {
1265         struct br_ip br_group;
1266
1267         if (!ipv6_is_transient_multicast(group))
1268                 return;
1269
1270         br_group.u.ip6 = *group;
1271         br_group.proto = htons(ETH_P_IPV6);
1272
1273         br_multicast_leave_group(br, port, &br_group);
1274 }
1275 #endif
1276
1277 static int br_multicast_ipv4_rcv(struct net_bridge *br,
1278                                  struct net_bridge_port *port,
1279                                  struct sk_buff *skb)
1280 {
1281         struct sk_buff *skb2 = skb;
1282         const struct iphdr *iph;
1283         struct igmphdr *ih;
1284         unsigned int len;
1285         unsigned int offset;
1286         int err;
1287
1288         /* We treat OOM as packet loss for now. */
1289         if (!pskb_may_pull(skb, sizeof(*iph)))
1290                 return -EINVAL;
1291
1292         iph = ip_hdr(skb);
1293
1294         if (iph->ihl < 5 || iph->version != 4)
1295                 return -EINVAL;
1296
1297         if (!pskb_may_pull(skb, ip_hdrlen(skb)))
1298                 return -EINVAL;
1299
1300         iph = ip_hdr(skb);
1301
1302         if (unlikely(ip_fast_csum((u8 *)iph, iph->ihl)))
1303                 return -EINVAL;
1304
1305         if (iph->protocol != IPPROTO_IGMP) {
1306                 if ((iph->daddr & IGMP_LOCAL_GROUP_MASK) != IGMP_LOCAL_GROUP)
1307                         BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1308                 return 0;
1309         }
1310
1311         len = ntohs(iph->tot_len);
1312         if (skb->len < len || len < ip_hdrlen(skb))
1313                 return -EINVAL;
1314
1315         if (skb->len > len) {
1316                 skb2 = skb_clone(skb, GFP_ATOMIC);
1317                 if (!skb2)
1318                         return -ENOMEM;
1319
1320                 err = pskb_trim_rcsum(skb2, len);
1321                 if (err)
1322                         goto err_out;
1323         }
1324
1325         len -= ip_hdrlen(skb2);
1326         offset = skb_network_offset(skb2) + ip_hdrlen(skb2);
1327         __skb_pull(skb2, offset);
1328         skb_reset_transport_header(skb2);
1329
1330         err = -EINVAL;
1331         if (!pskb_may_pull(skb2, sizeof(*ih)))
1332                 goto out;
1333
1334         switch (skb2->ip_summed) {
1335         case CHECKSUM_COMPLETE:
1336                 if (!csum_fold(skb2->csum))
1337                         break;
1338                 /* fall through */
1339         case CHECKSUM_NONE:
1340                 skb2->csum = 0;
1341                 if (skb_checksum_complete(skb2))
1342                         goto out;
1343         }
1344
1345         err = 0;
1346
1347         BR_INPUT_SKB_CB(skb)->igmp = 1;
1348         ih = igmp_hdr(skb2);
1349
1350         switch (ih->type) {
1351         case IGMP_HOST_MEMBERSHIP_REPORT:
1352         case IGMPV2_HOST_MEMBERSHIP_REPORT:
1353                 BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1354                 err = br_ip4_multicast_add_group(br, port, ih->group);
1355                 break;
1356         case IGMPV3_HOST_MEMBERSHIP_REPORT:
1357                 err = br_ip4_multicast_igmp3_report(br, port, skb2);
1358                 break;
1359         case IGMP_HOST_MEMBERSHIP_QUERY:
1360                 err = br_ip4_multicast_query(br, port, skb2);
1361                 break;
1362         case IGMP_HOST_LEAVE_MESSAGE:
1363                 br_ip4_multicast_leave_group(br, port, ih->group);
1364                 break;
1365         }
1366
1367 out:
1368         __skb_push(skb2, offset);
1369 err_out:
1370         if (skb2 != skb)
1371                 kfree_skb(skb2);
1372         return err;
1373 }
1374
1375 #if IS_ENABLED(CONFIG_IPV6)
1376 static int br_multicast_ipv6_rcv(struct net_bridge *br,
1377                                  struct net_bridge_port *port,
1378                                  struct sk_buff *skb)
1379 {
1380         struct sk_buff *skb2;
1381         const struct ipv6hdr *ip6h;
1382         u8 icmp6_type;
1383         u8 nexthdr;
1384         __be16 frag_off;
1385         unsigned int len;
1386         int offset;
1387         int err;
1388
1389         if (!pskb_may_pull(skb, sizeof(*ip6h)))
1390                 return -EINVAL;
1391
1392         ip6h = ipv6_hdr(skb);
1393
1394         /*
1395          * We're interested in MLD messages only.
1396          *  - Version is 6
1397          *  - MLD has always Router Alert hop-by-hop option
1398          *  - But we do not support jumbrograms.
1399          */
1400         if (ip6h->version != 6 ||
1401             ip6h->nexthdr != IPPROTO_HOPOPTS ||
1402             ip6h->payload_len == 0)
1403                 return 0;
1404
1405         len = ntohs(ip6h->payload_len) + sizeof(*ip6h);
1406         if (skb->len < len)
1407                 return -EINVAL;
1408
1409         nexthdr = ip6h->nexthdr;
1410         offset = ipv6_skip_exthdr(skb, sizeof(*ip6h), &nexthdr, &frag_off);
1411
1412         if (offset < 0 || nexthdr != IPPROTO_ICMPV6)
1413                 return 0;
1414
1415         /* Okay, we found ICMPv6 header */
1416         skb2 = skb_clone(skb, GFP_ATOMIC);
1417         if (!skb2)
1418                 return -ENOMEM;
1419
1420         err = -EINVAL;
1421         if (!pskb_may_pull(skb2, offset + sizeof(struct icmp6hdr)))
1422                 goto out;
1423
1424         len -= offset - skb_network_offset(skb2);
1425
1426         __skb_pull(skb2, offset);
1427         skb_reset_transport_header(skb2);
1428         skb_postpull_rcsum(skb2, skb_network_header(skb2),
1429                            skb_network_header_len(skb2));
1430
1431         icmp6_type = icmp6_hdr(skb2)->icmp6_type;
1432
1433         switch (icmp6_type) {
1434         case ICMPV6_MGM_QUERY:
1435         case ICMPV6_MGM_REPORT:
1436         case ICMPV6_MGM_REDUCTION:
1437         case ICMPV6_MLD2_REPORT:
1438                 break;
1439         default:
1440                 err = 0;
1441                 goto out;
1442         }
1443
1444         /* Okay, we found MLD message. Check further. */
1445         if (skb2->len > len) {
1446                 err = pskb_trim_rcsum(skb2, len);
1447                 if (err)
1448                         goto out;
1449                 err = -EINVAL;
1450         }
1451
1452         ip6h = ipv6_hdr(skb2);
1453
1454         switch (skb2->ip_summed) {
1455         case CHECKSUM_COMPLETE:
1456                 if (!csum_ipv6_magic(&ip6h->saddr, &ip6h->daddr, skb2->len,
1457                                         IPPROTO_ICMPV6, skb2->csum))
1458                         break;
1459                 /*FALLTHROUGH*/
1460         case CHECKSUM_NONE:
1461                 skb2->csum = ~csum_unfold(csum_ipv6_magic(&ip6h->saddr,
1462                                                         &ip6h->daddr,
1463                                                         skb2->len,
1464                                                         IPPROTO_ICMPV6, 0));
1465                 if (__skb_checksum_complete(skb2))
1466                         goto out;
1467         }
1468
1469         err = 0;
1470
1471         BR_INPUT_SKB_CB(skb)->igmp = 1;
1472
1473         switch (icmp6_type) {
1474         case ICMPV6_MGM_REPORT:
1475             {
1476                 struct mld_msg *mld;
1477                 if (!pskb_may_pull(skb2, sizeof(*mld))) {
1478                         err = -EINVAL;
1479                         goto out;
1480                 }
1481                 mld = (struct mld_msg *)skb_transport_header(skb2);
1482                 BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1483                 err = br_ip6_multicast_add_group(br, port, &mld->mld_mca);
1484                 break;
1485             }
1486         case ICMPV6_MLD2_REPORT:
1487                 err = br_ip6_multicast_mld2_report(br, port, skb2);
1488                 break;
1489         case ICMPV6_MGM_QUERY:
1490                 err = br_ip6_multicast_query(br, port, skb2);
1491                 break;
1492         case ICMPV6_MGM_REDUCTION:
1493             {
1494                 struct mld_msg *mld;
1495                 if (!pskb_may_pull(skb2, sizeof(*mld))) {
1496                         err = -EINVAL;
1497                         goto out;
1498                 }
1499                 mld = (struct mld_msg *)skb_transport_header(skb2);
1500                 br_ip6_multicast_leave_group(br, port, &mld->mld_mca);
1501             }
1502         }
1503
1504 out:
1505         kfree_skb(skb2);
1506         return err;
1507 }
1508 #endif
1509
1510 int br_multicast_rcv(struct net_bridge *br, struct net_bridge_port *port,
1511                      struct sk_buff *skb)
1512 {
1513         BR_INPUT_SKB_CB(skb)->igmp = 0;
1514         BR_INPUT_SKB_CB(skb)->mrouters_only = 0;
1515
1516         if (br->multicast_disabled)
1517                 return 0;
1518
1519         switch (skb->protocol) {
1520         case htons(ETH_P_IP):
1521                 return br_multicast_ipv4_rcv(br, port, skb);
1522 #if IS_ENABLED(CONFIG_IPV6)
1523         case htons(ETH_P_IPV6):
1524                 return br_multicast_ipv6_rcv(br, port, skb);
1525 #endif
1526         }
1527
1528         return 0;
1529 }
1530
1531 static void br_multicast_query_expired(unsigned long data)
1532 {
1533         struct net_bridge *br = (void *)data;
1534
1535         spin_lock(&br->multicast_lock);
1536         if (br->multicast_startup_queries_sent <
1537             br->multicast_startup_query_count)
1538                 br->multicast_startup_queries_sent++;
1539
1540         br_multicast_send_query(br, NULL, br->multicast_startup_queries_sent);
1541
1542         spin_unlock(&br->multicast_lock);
1543 }
1544
1545 void br_multicast_init(struct net_bridge *br)
1546 {
1547         br->hash_elasticity = 4;
1548         br->hash_max = 512;
1549
1550         br->multicast_router = 1;
1551         br->multicast_last_member_count = 2;
1552         br->multicast_startup_query_count = 2;
1553
1554         br->multicast_last_member_interval = HZ;
1555         br->multicast_query_response_interval = 10 * HZ;
1556         br->multicast_startup_query_interval = 125 * HZ / 4;
1557         br->multicast_query_interval = 125 * HZ;
1558         br->multicast_querier_interval = 255 * HZ;
1559         br->multicast_membership_interval = 260 * HZ;
1560
1561         spin_lock_init(&br->multicast_lock);
1562         setup_timer(&br->multicast_router_timer,
1563                     br_multicast_local_router_expired, 0);
1564         setup_timer(&br->multicast_querier_timer,
1565                     br_multicast_local_router_expired, 0);
1566         setup_timer(&br->multicast_query_timer, br_multicast_query_expired,
1567                     (unsigned long)br);
1568 }
1569
1570 void br_multicast_open(struct net_bridge *br)
1571 {
1572         br->multicast_startup_queries_sent = 0;
1573
1574         if (br->multicast_disabled)
1575                 return;
1576
1577         mod_timer(&br->multicast_query_timer, jiffies);
1578 }
1579
1580 void br_multicast_stop(struct net_bridge *br)
1581 {
1582         struct net_bridge_mdb_htable *mdb;
1583         struct net_bridge_mdb_entry *mp;
1584         struct hlist_node *p, *n;
1585         u32 ver;
1586         int i;
1587
1588         del_timer_sync(&br->multicast_router_timer);
1589         del_timer_sync(&br->multicast_querier_timer);
1590         del_timer_sync(&br->multicast_query_timer);
1591
1592         spin_lock_bh(&br->multicast_lock);
1593         mdb = mlock_dereference(br->mdb, br);
1594         if (!mdb)
1595                 goto out;
1596
1597         br->mdb = NULL;
1598
1599         ver = mdb->ver;
1600         for (i = 0; i < mdb->max; i++) {
1601                 hlist_for_each_entry_safe(mp, p, n, &mdb->mhash[i],
1602                                           hlist[ver]) {
1603                         del_timer(&mp->timer);
1604                         call_rcu_bh(&mp->rcu, br_multicast_free_group);
1605                 }
1606         }
1607
1608         if (mdb->old) {
1609                 spin_unlock_bh(&br->multicast_lock);
1610                 rcu_barrier_bh();
1611                 spin_lock_bh(&br->multicast_lock);
1612                 WARN_ON(mdb->old);
1613         }
1614
1615         mdb->old = mdb;
1616         call_rcu_bh(&mdb->rcu, br_mdb_free);
1617
1618 out:
1619         spin_unlock_bh(&br->multicast_lock);
1620 }
1621
1622 int br_multicast_set_router(struct net_bridge *br, unsigned long val)
1623 {
1624         int err = -ENOENT;
1625
1626         spin_lock_bh(&br->multicast_lock);
1627         if (!netif_running(br->dev))
1628                 goto unlock;
1629
1630         switch (val) {
1631         case 0:
1632         case 2:
1633                 del_timer(&br->multicast_router_timer);
1634                 /* fall through */
1635         case 1:
1636                 br->multicast_router = val;
1637                 err = 0;
1638                 break;
1639
1640         default:
1641                 err = -EINVAL;
1642                 break;
1643         }
1644
1645 unlock:
1646         spin_unlock_bh(&br->multicast_lock);
1647
1648         return err;
1649 }
1650
1651 int br_multicast_set_port_router(struct net_bridge_port *p, unsigned long val)
1652 {
1653         struct net_bridge *br = p->br;
1654         int err = -ENOENT;
1655
1656         spin_lock(&br->multicast_lock);
1657         if (!netif_running(br->dev) || p->state == BR_STATE_DISABLED)
1658                 goto unlock;
1659
1660         switch (val) {
1661         case 0:
1662         case 1:
1663         case 2:
1664                 p->multicast_router = val;
1665                 err = 0;
1666
1667                 if (val < 2 && !hlist_unhashed(&p->rlist))
1668                         hlist_del_init_rcu(&p->rlist);
1669
1670                 if (val == 1)
1671                         break;
1672
1673                 del_timer(&p->multicast_router_timer);
1674
1675                 if (val == 0)
1676                         break;
1677
1678                 br_multicast_add_router(br, p);
1679                 break;
1680
1681         default:
1682                 err = -EINVAL;
1683                 break;
1684         }
1685
1686 unlock:
1687         spin_unlock(&br->multicast_lock);
1688
1689         return err;
1690 }
1691
1692 static void br_multicast_start_querier(struct net_bridge *br)
1693 {
1694         struct net_bridge_port *port;
1695
1696         br_multicast_open(br);
1697
1698         list_for_each_entry(port, &br->port_list, list) {
1699                 if (port->state == BR_STATE_DISABLED ||
1700                     port->state == BR_STATE_BLOCKING)
1701                         continue;
1702
1703                 __br_multicast_enable_port(port);
1704         }
1705 }
1706
1707 int br_multicast_toggle(struct net_bridge *br, unsigned long val)
1708 {
1709         int err = 0;
1710         struct net_bridge_mdb_htable *mdb;
1711
1712         spin_lock_bh(&br->multicast_lock);
1713         if (br->multicast_disabled == !val)
1714                 goto unlock;
1715
1716         br->multicast_disabled = !val;
1717         if (br->multicast_disabled)
1718                 goto unlock;
1719
1720         if (!netif_running(br->dev))
1721                 goto unlock;
1722
1723         mdb = mlock_dereference(br->mdb, br);
1724         if (mdb) {
1725                 if (mdb->old) {
1726                         err = -EEXIST;
1727 rollback:
1728                         br->multicast_disabled = !!val;
1729                         goto unlock;
1730                 }
1731
1732                 err = br_mdb_rehash(&br->mdb, mdb->max,
1733                                     br->hash_elasticity);
1734                 if (err)
1735                         goto rollback;
1736         }
1737
1738         br_multicast_start_querier(br);
1739
1740 unlock:
1741         spin_unlock_bh(&br->multicast_lock);
1742
1743         return err;
1744 }
1745
1746 int br_multicast_set_hash_max(struct net_bridge *br, unsigned long val)
1747 {
1748         int err = -ENOENT;
1749         u32 old;
1750         struct net_bridge_mdb_htable *mdb;
1751
1752         spin_lock(&br->multicast_lock);
1753         if (!netif_running(br->dev))
1754                 goto unlock;
1755
1756         err = -EINVAL;
1757         if (!is_power_of_2(val))
1758                 goto unlock;
1759
1760         mdb = mlock_dereference(br->mdb, br);
1761         if (mdb && val < mdb->size)
1762                 goto unlock;
1763
1764         err = 0;
1765
1766         old = br->hash_max;
1767         br->hash_max = val;
1768
1769         if (mdb) {
1770                 if (mdb->old) {
1771                         err = -EEXIST;
1772 rollback:
1773                         br->hash_max = old;
1774                         goto unlock;
1775                 }
1776
1777                 err = br_mdb_rehash(&br->mdb, br->hash_max,
1778                                     br->hash_elasticity);
1779                 if (err)
1780                         goto rollback;
1781         }
1782
1783 unlock:
1784         spin_unlock(&br->multicast_lock);
1785
1786         return err;
1787 }