net/bridge/br_multicast.c

   1 /*
   2  * Bridge multicast support.
   3  *
   4  * Copyright (c) 2010 Herbert Xu <herbert@gondor.apana.org.au>
   5  *
   6  * This program is free software; you can redistribute it and/or modify it
   7  * under the terms of the GNU General Public License as published by the Free
   8  * Software Foundation; either version 2 of the License, or (at your option)
   9  * any later version.
  10  *
  11  */
  12
  13 #include <linux/err.h>
  14 #include <linux/if_ether.h>
  15 #include <linux/igmp.h>
  16 #include <linux/jhash.h>
  17 #include <linux/kernel.h>
  18 #include <linux/log2.h>
  19 #include <linux/netdevice.h>
  20 #include <linux/netfilter_bridge.h>
  21 #include <linux/random.h>
  22 #include <linux/rculist.h>
  23 #include <linux/skbuff.h>
  24 #include <linux/slab.h>
  25 #include <linux/timer.h>
  26 #include <net/ip.h>
  27 #if IS_ENABLED(CONFIG_IPV6)
  28 #include <net/ipv6.h>
  29 #include <net/mld.h>
  30 #include <net/addrconf.h>
  31 #include <net/ip6_checksum.h>
  32 #endif
  33
  34 #include "br_private.h"
  35
  36 #define mlock_dereference(X, br) \
  37         rcu_dereference_protected(X, lockdep_is_held(&br->multicast_lock))
  38
  39 static void br_multicast_start_querier(struct net_bridge *br);
  40
  41 #if IS_ENABLED(CONFIG_IPV6)
  42 static inline int ipv6_is_transient_multicast(const struct in6_addr *addr)
  43 {
  44         if (ipv6_addr_is_multicast(addr) && IPV6_ADDR_MC_FLAG_TRANSIENT(addr))
  45                 return 1;
  46         return 0;
  47 }
  48 #endif
  49
  50 static inline int br_ip_equal(const struct br_ip *a, const struct br_ip *b)
  51 {
  52         if (a->proto != b->proto)
  53                 return 0;
  54         switch (a->proto) {
  55         case htons(ETH_P_IP):
  56                 return a->u.ip4 == b->u.ip4;
  57 #if IS_ENABLED(CONFIG_IPV6)
  58         case htons(ETH_P_IPV6):
  59                 return ipv6_addr_equal(&a->u.ip6, &b->u.ip6);
  60 #endif
  61         }
  62         return 0;
  63 }
  64
  65 static inline int __br_ip4_hash(struct net_bridge_mdb_htable *mdb, __be32 ip)
  66 {
  67         return jhash_1word(mdb->secret, (__force u32)ip) & (mdb->max - 1);
  68 }
  69
  70 #if IS_ENABLED(CONFIG_IPV6)
  71 static inline int __br_ip6_hash(struct net_bridge_mdb_htable *mdb,
  72                                 const struct in6_addr *ip)
  73 {
  74         return jhash2((__force u32 *)ip->s6_addr32, 4, mdb->secret) & (mdb->max - 1);
  75 }
  76 #endif
  77
  78 static inline int br_ip_hash(struct net_bridge_mdb_htable *mdb,
  79                              struct br_ip *ip)
  80 {
  81         switch (ip->proto) {
  82         case htons(ETH_P_IP):
  83                 return __br_ip4_hash(mdb, ip->u.ip4);
  84 #if IS_ENABLED(CONFIG_IPV6)
  85         case htons(ETH_P_IPV6):
  86                 return __br_ip6_hash(mdb, &ip->u.ip6);
  87 #endif
  88         }
  89         return 0;
  90 }
  91
  92 static struct net_bridge_mdb_entry *__br_mdb_ip_get(
  93         struct net_bridge_mdb_htable *mdb, struct br_ip *dst, int hash)
  94 {
  95         struct net_bridge_mdb_entry *mp;
  96         struct hlist_node *p;
  97
  98         hlist_for_each_entry_rcu(mp, p, &mdb->mhash[hash], hlist[mdb->ver]) {
  99                 if (br_ip_equal(&mp->addr, dst))
 100                         return mp;
 101         }
 102
 103         return NULL;
 104 }
 105
 106 static struct net_bridge_mdb_entry *br_mdb_ip_get(
 107         struct net_bridge_mdb_htable *mdb, struct br_ip *dst)
 108 {
 109         if (!mdb)
 110                 return NULL;
 111
 112         return __br_mdb_ip_get(mdb, dst, br_ip_hash(mdb, dst));
 113 }
 114
 115 static struct net_bridge_mdb_entry *br_mdb_ip4_get(
 116         struct net_bridge_mdb_htable *mdb, __be32 dst)
 117 {
 118         struct br_ip br_dst;
 119
 120         br_dst.u.ip4 = dst;
 121         br_dst.proto = htons(ETH_P_IP);
 122
 123         return br_mdb_ip_get(mdb, &br_dst);
 124 }
 125
 126 #if IS_ENABLED(CONFIG_IPV6)
 127 static struct net_bridge_mdb_entry *br_mdb_ip6_get(
 128         struct net_bridge_mdb_htable *mdb, const struct in6_addr *dst)
 129 {
 130         struct br_ip br_dst;
 131
 132         br_dst.u.ip6 = *dst;
 133         br_dst.proto = htons(ETH_P_IPV6);
 134
 135         return br_mdb_ip_get(mdb, &br_dst);
 136 }
 137 #endif
 138
 139 struct net_bridge_mdb_entry *br_mdb_get(struct net_bridge *br,
 140                                         struct sk_buff *skb)
 141 {
 142         struct net_bridge_mdb_htable *mdb = rcu_dereference(br->mdb);
 143         struct br_ip ip;
 144
 145         if (br->multicast_disabled)
 146                 return NULL;
 147
 148         if (BR_INPUT_SKB_CB(skb)->igmp)
 149                 return NULL;
 150
 151         ip.proto = skb->protocol;
 152
 153         switch (skb->protocol) {
 154         case htons(ETH_P_IP):
 155                 ip.u.ip4 = ip_hdr(skb)->daddr;
 156                 break;
 157 #if IS_ENABLED(CONFIG_IPV6)
 158         case htons(ETH_P_IPV6):
 159                 ip.u.ip6 = ipv6_hdr(skb)->daddr;
 160                 break;
 161 #endif
 162         default:
 163                 return NULL;
 164         }
 165
 166         return br_mdb_ip_get(mdb, &ip);
 167 }
 168
 169 static void br_mdb_free(struct rcu_head *head)
 170 {
 171         struct net_bridge_mdb_htable *mdb =
 172                 container_of(head, struct net_bridge_mdb_htable, rcu);
 173         struct net_bridge_mdb_htable *old = mdb->old;
 174
 175         mdb->old = NULL;
 176         kfree(old->mhash);
 177         kfree(old);
 178 }
 179
 180 static int br_mdb_copy(struct net_bridge_mdb_htable *new,
 181                        struct net_bridge_mdb_htable *old,
 182                        int elasticity)
 183 {
 184         struct net_bridge_mdb_entry *mp;
 185         struct hlist_node *p;
 186         int maxlen;
 187         int len;
 188         int i;
 189
 190         for (i = 0; i < old->max; i++)
 191                 hlist_for_each_entry(mp, p, &old->mhash[i], hlist[old->ver])
 192                         hlist_add_head(&mp->hlist[new->ver],
 193                                        &new->mhash[br_ip_hash(new, &mp->addr)]);
 194
 195         if (!elasticity)
 196                 return 0;
 197
 198         maxlen = 0;
 199         for (i = 0; i < new->max; i++) {
 200                 len = 0;
 201                 hlist_for_each_entry(mp, p, &new->mhash[i], hlist[new->ver])
 202                         len++;
 203                 if (len > maxlen)
 204                         maxlen = len;
 205         }
 206
 207         return maxlen > elasticity ? -EINVAL : 0;
 208 }
 209
 210 static void br_multicast_free_pg(struct rcu_head *head)
 211 {
 212         struct net_bridge_port_group *p =
 213                 container_of(head, struct net_bridge_port_group, rcu);
 214
 215         kfree(p);
 216 }
 217
 218 static void br_multicast_free_group(struct rcu_head *head)
 219 {
 220         struct net_bridge_mdb_entry *mp =
 221                 container_of(head, struct net_bridge_mdb_entry, rcu);
 222
 223         kfree(mp);
 224 }
 225
 226 static void br_multicast_group_expired(unsigned long data)
 227 {
 228         struct net_bridge_mdb_entry *mp = (void *)data;
 229         struct net_bridge *br = mp->br;
 230         struct net_bridge_mdb_htable *mdb;
 231
 232         spin_lock(&br->multicast_lock);
 233         if (!netif_running(br->dev) || timer_pending(&mp->timer))
 234                 goto out;
 235
 236         mp->mglist = false;
 237
 238         if (mp->ports)
 239                 goto out;
 240
 241         mdb = mlock_dereference(br->mdb, br);
 242
 243         hlist_del_rcu(&mp->hlist[mdb->ver]);
 244         mdb->size--;
 245
 246         call_rcu_bh(&mp->rcu, br_multicast_free_group);
 247
 248 out:
 249         spin_unlock(&br->multicast_lock);
 250 }
 251
 252 static void br_multicast_del_pg(struct net_bridge *br,
 253                                 struct net_bridge_port_group *pg)
 254 {
 255         struct net_bridge_mdb_htable *mdb;
 256         struct net_bridge_mdb_entry *mp;
 257         struct net_bridge_port_group *p;
 258         struct net_bridge_port_group __rcu **pp;
 259
 260         mdb = mlock_dereference(br->mdb, br);
 261
 262         mp = br_mdb_ip_get(mdb, &pg->addr);
 263         if (WARN_ON(!mp))
 264                 return;
 265
 266         for (pp = &mp->ports;
 267              (p = mlock_dereference(*pp, br)) != NULL;
 268              pp = &p->next) {
 269                 if (p != pg)
 270                         continue;
 271
 272                 rcu_assign_pointer(*pp, p->next);
 273                 hlist_del_init(&p->mglist);
 274                 del_timer(&p->timer);
 275                 call_rcu_bh(&p->rcu, br_multicast_free_pg);
 276
 277                 if (!mp->ports && !mp->mglist &&
 278                     netif_running(br->dev))
 279                         mod_timer(&mp->timer, jiffies);
 280
 281                 return;
 282         }
 283
 284         WARN_ON(1);
 285 }
 286
 287 static void br_multicast_port_group_expired(unsigned long data)
 288 {
 289         struct net_bridge_port_group *pg = (void *)data;
 290         struct net_bridge *br = pg->port->br;
 291
 292         spin_lock(&br->multicast_lock);
 293         if (!netif_running(br->dev) || timer_pending(&pg->timer) ||
 294             hlist_unhashed(&pg->mglist))
 295                 goto out;
 296
 297         br_multicast_del_pg(br, pg);
 298
 299 out:
 300         spin_unlock(&br->multicast_lock);
 301 }
 302
 303 static int br_mdb_rehash(struct net_bridge_mdb_htable __rcu **mdbp, int max,
 304                          int elasticity)
 305 {
 306         struct net_bridge_mdb_htable *old = rcu_dereference_protected(*mdbp, 1);
 307         struct net_bridge_mdb_htable *mdb;
 308         int err;
 309
 310         mdb = kmalloc(sizeof(*mdb), GFP_ATOMIC);
 311         if (!mdb)
 312                 return -ENOMEM;
 313
 314         mdb->max = max;
 315         mdb->old = old;
 316
 317         mdb->mhash = kzalloc(max * sizeof(*mdb->mhash), GFP_ATOMIC);
 318         if (!mdb->mhash) {
 319                 kfree(mdb);
 320                 return -ENOMEM;
 321         }
 322
 323         mdb->size = old ? old->size : 0;
 324         mdb->ver = old ? old->ver ^ 1 : 0;
 325
 326         if (!old || elasticity)
 327                 get_random_bytes(&mdb->secret, sizeof(mdb->secret));
 328         else
 329                 mdb->secret = old->secret;
 330
 331         if (!old)
 332                 goto out;
 333
 334         err = br_mdb_copy(mdb, old, elasticity);
 335         if (err) {
 336                 kfree(mdb->mhash);
 337                 kfree(mdb);
 338                 return err;
 339         }
 340
 341         call_rcu_bh(&mdb->rcu, br_mdb_free);
 342
 343 out:
 344         rcu_assign_pointer(*mdbp, mdb);
 345
 346         return 0;
 347 }
 348
 349 static struct sk_buff *br_ip4_multicast_alloc_query(struct net_bridge *br,
 350                                                     __be32 group)
 351 {
 352         struct sk_buff *skb;
 353         struct igmphdr *ih;
 354         struct ethhdr *eth;
 355         struct iphdr *iph;
 356
 357         skb = netdev_alloc_skb_ip_align(br->dev, sizeof(*eth) + sizeof(*iph) +
 358                                                  sizeof(*ih) + 4);
 359         if (!skb)
 360                 goto out;
 361
 362         skb->protocol = htons(ETH_P_IP);
 363
 364         skb_reset_mac_header(skb);
 365         eth = eth_hdr(skb);
 366
 367         memcpy(eth->h_source, br->dev->dev_addr, 6);
 368         eth->h_dest[0] = 1;
 369         eth->h_dest[1] = 0;
 370         eth->h_dest[2] = 0x5e;
 371         eth->h_dest[3] = 0;
 372         eth->h_dest[4] = 0;
 373         eth->h_dest[5] = 1;
 374         eth->h_proto = htons(ETH_P_IP);
 375         skb_put(skb, sizeof(*eth));
 376
 377         skb_set_network_header(skb, skb->len);
 378         iph = ip_hdr(skb);
 379
 380         iph->version = 4;
 381         iph->ihl = 6;
 382         iph->tos = 0xc0;
 383         iph->tot_len = htons(sizeof(*iph) + sizeof(*ih) + 4);
 384         iph->id = 0;
 385         iph->frag_off = htons(IP_DF);
 386         iph->ttl = 1;
 387         iph->protocol = IPPROTO_IGMP;
 388         iph->saddr = 0;
 389         iph->daddr = htonl(INADDR_ALLHOSTS_GROUP);
 390         ((u8 *)&iph[1])[0] = IPOPT_RA;
 391         ((u8 *)&iph[1])[1] = 4;
 392         ((u8 *)&iph[1])[2] = 0;
 393         ((u8 *)&iph[1])[3] = 0;
 394         ip_send_check(iph);
 395         skb_put(skb, 24);
 396
 397         skb_set_transport_header(skb, skb->len);
 398         ih = igmp_hdr(skb);
 399         ih->type = IGMP_HOST_MEMBERSHIP_QUERY;
 400         ih->code = (group ? br->multicast_last_member_interval :
 401                             br->multicast_query_response_interval) /
 402                    (HZ / IGMP_TIMER_SCALE);
 403         ih->group = group;
 404         ih->csum = 0;
 405         ih->csum = ip_compute_csum((void *)ih, sizeof(struct igmphdr));
 406         skb_put(skb, sizeof(*ih));
 407
 408         __skb_pull(skb, sizeof(*eth));
 409
 410 out:
 411         return skb;
 412 }
 413
 414 #if IS_ENABLED(CONFIG_IPV6)
 415 static struct sk_buff *br_ip6_multicast_alloc_query(struct net_bridge *br,
 416                                                     const struct in6_addr *group)
 417 {
 418         struct sk_buff *skb;
 419         struct ipv6hdr *ip6h;
 420         struct mld_msg *mldq;
 421         struct ethhdr *eth;
 422         u8 *hopopt;
 423         unsigned long interval;
 424
 425         skb = netdev_alloc_skb_ip_align(br->dev, sizeof(*eth) + sizeof(*ip6h) +
 426                                                  8 + sizeof(*mldq));
 427         if (!skb)
 428                 goto out;
 429
 430         skb->protocol = htons(ETH_P_IPV6);
 431
 432         /* Ethernet header */
 433         skb_reset_mac_header(skb);
 434         eth = eth_hdr(skb);
 435
 436         memcpy(eth->h_source, br->dev->dev_addr, 6);
 437         eth->h_proto = htons(ETH_P_IPV6);
 438         skb_put(skb, sizeof(*eth));
 439
 440         /* IPv6 header + HbH option */
 441         skb_set_network_header(skb, skb->len);
 442         ip6h = ipv6_hdr(skb);
 443
 444         *(__force __be32 *)ip6h = htonl(0x60000000);
 445         ip6h->payload_len = htons(8 + sizeof(*mldq));
 446         ip6h->nexthdr = IPPROTO_HOPOPTS;
 447         ip6h->hop_limit = 1;
 448         ipv6_addr_set(&ip6h->daddr, htonl(0xff020000), 0, 0, htonl(1));
 449         if (ipv6_dev_get_saddr(dev_net(br->dev), br->dev, &ip6h->daddr, 0,
 450                                &ip6h->saddr)) {
 451                 kfree_skb(skb);
 452                 return NULL;
 453         }
 454         ipv6_eth_mc_map(&ip6h->daddr, eth->h_dest);
 455
 456         hopopt = (u8 *)(ip6h + 1);
 457         hopopt[0] = IPPROTO_ICMPV6;             /* next hdr */
 458         hopopt[1] = 0;                          /* length of HbH */
 459         hopopt[2] = IPV6_TLV_ROUTERALERT;       /* Router Alert */
 460         hopopt[3] = 2;                          /* Length of RA Option */
 461         hopopt[4] = 0;                          /* Type = 0x0000 (MLD) */
 462         hopopt[5] = 0;
 463         hopopt[6] = IPV6_TLV_PAD1;              /* Pad1 */
 464         hopopt[7] = IPV6_TLV_PAD1;              /* Pad1 */
 465
 466         skb_put(skb, sizeof(*ip6h) + 8);
 467
 468         /* ICMPv6 */
 469         skb_set_transport_header(skb, skb->len);
 470         mldq = (struct mld_msg *) icmp6_hdr(skb);
 471
 472         interval = ipv6_addr_any(group) ? br->multicast_last_member_interval :
 473                                           br->multicast_query_response_interval;
 474
 475         mldq->mld_type = ICMPV6_MGM_QUERY;
 476         mldq->mld_code = 0;
 477         mldq->mld_cksum = 0;
 478         mldq->mld_maxdelay = htons((u16)jiffies_to_msecs(interval));
 479         mldq->mld_reserved = 0;
 480         mldq->mld_mca = *group;
 481
 482         /* checksum */
 483         mldq->mld_cksum = csum_ipv6_magic(&ip6h->saddr, &ip6h->daddr,
 484                                           sizeof(*mldq), IPPROTO_ICMPV6,
 485                                           csum_partial(mldq,
 486                                                        sizeof(*mldq), 0));
 487         skb_put(skb, sizeof(*mldq));
 488
 489         __skb_pull(skb, sizeof(*eth));
 490
 491 out:
 492         return skb;
 493 }
 494 #endif
 495
 496 static struct sk_buff *br_multicast_alloc_query(struct net_bridge *br,
 497                                                 struct br_ip *addr)
 498 {
 499         switch (addr->proto) {
 500         case htons(ETH_P_IP):
 501                 return br_ip4_multicast_alloc_query(br, addr->u.ip4);
 502 #if IS_ENABLED(CONFIG_IPV6)
 503         case htons(ETH_P_IPV6):
 504                 return br_ip6_multicast_alloc_query(br, &addr->u.ip6);
 505 #endif
 506         }
 507         return NULL;
 508 }
 509
 510 static struct net_bridge_mdb_entry *br_multicast_get_group(
 511         struct net_bridge *br, struct net_bridge_port *port,
 512         struct br_ip *group, int hash)
 513 {
 514         struct net_bridge_mdb_htable *mdb;
 515         struct net_bridge_mdb_entry *mp;
 516         struct hlist_node *p;
 517         unsigned int count = 0;
 518         unsigned int max;
 519         int elasticity;
 520         int err;
 521
 522         mdb = rcu_dereference_protected(br->mdb, 1);
 523         hlist_for_each_entry(mp, p, &mdb->mhash[hash], hlist[mdb->ver]) {
 524                 count++;
 525                 if (unlikely(br_ip_equal(group, &mp->addr)))
 526                         return mp;
 527         }
 528
 529         elasticity = 0;
 530         max = mdb->max;
 531
 532         if (unlikely(count > br->hash_elasticity && count)) {
 533                 if (net_ratelimit())
 534                         br_info(br, "Multicast hash table "
 535                                 "chain limit reached: %s\n",
 536                                 port ? port->dev->name : br->dev->name);
 537
 538                 elasticity = br->hash_elasticity;
 539         }
 540
 541         if (mdb->size >= max) {
 542                 max *= 2;
 543                 if (unlikely(max >= br->hash_max)) {
 544                         br_warn(br, "Multicast hash table maximum "
 545                                 "reached, disabling snooping: %s, %d\n",
 546                                 port ? port->dev->name : br->dev->name, max);
 547                         err = -E2BIG;
 548 disable:
 549                         br->multicast_disabled = 1;
 550                         goto err;
 551                 }
 552         }
 553
 554         if (max > mdb->max || elasticity) {
 555                 if (mdb->old) {
 556                         if (net_ratelimit())
 557                                 br_info(br, "Multicast hash table "
 558                                         "on fire: %s\n",
 559                                         port ? port->dev->name : br->dev->name);
 560                         err = -EEXIST;
 561                         goto err;
 562                 }
 563
 564                 err = br_mdb_rehash(&br->mdb, max, elasticity);
 565                 if (err) {
 566                         br_warn(br, "Cannot rehash multicast "
 567                                 "hash table, disabling snooping: %s, %d, %d\n",
 568                                 port ? port->dev->name : br->dev->name,
 569                                 mdb->size, err);
 570                         goto disable;
 571                 }
 572
 573                 err = -EAGAIN;
 574                 goto err;
 575         }
 576
 577         return NULL;
 578
 579 err:
 580         mp = ERR_PTR(err);
 581         return mp;
 582 }
 583
 584 static struct net_bridge_mdb_entry *br_multicast_new_group(
 585         struct net_bridge *br, struct net_bridge_port *port,
 586         struct br_ip *group)
 587 {
 588         struct net_bridge_mdb_htable *mdb;
 589         struct net_bridge_mdb_entry *mp;
 590         int hash;
 591         int err;
 592
 593         mdb = rcu_dereference_protected(br->mdb, 1);
 594         if (!mdb) {
 595                 err = br_mdb_rehash(&br->mdb, BR_HASH_SIZE, 0);
 596                 if (err)
 597                         return ERR_PTR(err);
 598                 goto rehash;
 599         }
 600
 601         hash = br_ip_hash(mdb, group);
 602         mp = br_multicast_get_group(br, port, group, hash);
 603         switch (PTR_ERR(mp)) {
 604         case 0:
 605                 break;
 606
 607         case -EAGAIN:
 608 rehash:
 609                 mdb = rcu_dereference_protected(br->mdb, 1);
 610                 hash = br_ip_hash(mdb, group);
 611                 break;
 612
 613         default:
 614                 goto out;
 615         }
 616
 617         mp = kzalloc(sizeof(*mp), GFP_ATOMIC);
 618         if (unlikely(!mp))
 619                 return ERR_PTR(-ENOMEM);
 620
 621         mp->br = br;
 622         mp->addr = *group;
 623         setup_timer(&mp->timer, br_multicast_group_expired,
 624                     (unsigned long)mp);
 625
 626         hlist_add_head_rcu(&mp->hlist[mdb->ver], &mdb->mhash[hash]);
 627         mdb->size++;
 628
 629 out:
 630         return mp;
 631 }
 632
 633 static int br_multicast_add_group(struct net_bridge *br,
 634                                   struct net_bridge_port *port,
 635                                   struct br_ip *group)
 636 {
 637         struct net_bridge_mdb_entry *mp;
 638         struct net_bridge_port_group *p;
 639         struct net_bridge_port_group __rcu **pp;
 640         unsigned long now = jiffies;
 641         int err;
 642
 643         spin_lock(&br->multicast_lock);
 644         if (!netif_running(br->dev) ||
 645             (port && port->state == BR_STATE_DISABLED))
 646                 goto out;
 647
 648         mp = br_multicast_new_group(br, port, group);
 649         err = PTR_ERR(mp);
 650         if (IS_ERR(mp))
 651                 goto err;
 652
 653         if (!port) {
 654                 mp->mglist = true;
 655                 mod_timer(&mp->timer, now + br->multicast_membership_interval);
 656                 goto out;
 657         }
 658
 659         for (pp = &mp->ports;
 660              (p = mlock_dereference(*pp, br)) != NULL;
 661              pp = &p->next) {
 662                 if (p->port == port)
 663                         goto found;
 664                 if ((unsigned long)p->port < (unsigned long)port)
 665                         break;
 666         }
 667
 668         p = kzalloc(sizeof(*p), GFP_ATOMIC);
 669         err = -ENOMEM;
 670         if (unlikely(!p))
 671                 goto err;
 672
 673         p->addr = *group;
 674         p->port = port;
 675         p->next = *pp;
 676         hlist_add_head(&p->mglist, &port->mglist);
 677         setup_timer(&p->timer, br_multicast_port_group_expired,
 678                     (unsigned long)p);
 679
 680         rcu_assign_pointer(*pp, p);
 681
 682 found:
 683         mod_timer(&p->timer, now + br->multicast_membership_interval);
 684 out:
 685         err = 0;
 686
 687 err:
 688         spin_unlock(&br->multicast_lock);
 689         return err;
 690 }
 691
 692 static int br_ip4_multicast_add_group(struct net_bridge *br,
 693                                       struct net_bridge_port *port,
 694                                       __be32 group)
 695 {
 696         struct br_ip br_group;
 697
 698         if (ipv4_is_local_multicast(group))
 699                 return 0;
 700
 701         br_group.u.ip4 = group;
 702         br_group.proto = htons(ETH_P_IP);
 703
 704         return br_multicast_add_group(br, port, &br_group);
 705 }
 706
 707 #if IS_ENABLED(CONFIG_IPV6)
 708 static int br_ip6_multicast_add_group(struct net_bridge *br,
 709                                       struct net_bridge_port *port,
 710                                       const struct in6_addr *group)
 711 {
 712         struct br_ip br_group;
 713
 714         if (!ipv6_is_transient_multicast(group))
 715                 return 0;
 716
 717         br_group.u.ip6 = *group;
 718         br_group.proto = htons(ETH_P_IPV6);
 719
 720         return br_multicast_add_group(br, port, &br_group);
 721 }
 722 #endif
 723
 724 static void br_multicast_router_expired(unsigned long data)
 725 {
 726         struct net_bridge_port *port = (void *)data;
 727         struct net_bridge *br = port->br;
 728
 729         spin_lock(&br->multicast_lock);
 730         if (port->multicast_router != 1 ||
 731             timer_pending(&port->multicast_router_timer) ||
 732             hlist_unhashed(&port->rlist))
 733                 goto out;
 734
 735         hlist_del_init_rcu(&port->rlist);
 736
 737 out:
 738         spin_unlock(&br->multicast_lock);
 739 }
 740
 741 static void br_multicast_local_router_expired(unsigned long data)
 742 {
 743 }
 744
 745 static void br_multicast_querier_expired(unsigned long data)
 746 {
 747         struct net_bridge *br = (void *)data;
 748
 749         spin_lock(&br->multicast_lock);
 750         if (!netif_running(br->dev) || br->multicast_disabled)
 751                 goto out;
 752
 753         br_multicast_start_querier(br);
 754
 755 out:
 756         spin_unlock(&br->multicast_lock);
 757 }
 758
 759 static void __br_multicast_send_query(struct net_bridge *br,
 760                                       struct net_bridge_port *port,
 761                                       struct br_ip *ip)
 762 {
 763         struct sk_buff *skb;
 764
 765         skb = br_multicast_alloc_query(br, ip);
 766         if (!skb)
 767                 return;
 768
 769         if (port) {
 770                 __skb_push(skb, sizeof(struct ethhdr));
 771                 skb->dev = port->dev;
 772                 NF_HOOK(NFPROTO_BRIDGE, NF_BR_LOCAL_OUT, skb, NULL, skb->dev,
 773                         dev_queue_xmit);
 774         } else
 775                 netif_rx(skb);
 776 }
 777
 778 static void br_multicast_send_query(struct net_bridge *br,
 779                                     struct net_bridge_port *port, u32 sent)
 780 {
 781         unsigned long time;
 782         struct br_ip br_group;
 783
 784         if (!netif_running(br->dev) || br->multicast_disabled ||
 785             !br->multicast_querier ||
 786             timer_pending(&br->multicast_querier_timer))
 787                 return;
 788
 789         memset(&br_group.u, 0, sizeof(br_group.u));
 790
 791         br_group.proto = htons(ETH_P_IP);
 792         __br_multicast_send_query(br, port, &br_group);
 793
 794 #if IS_ENABLED(CONFIG_IPV6)
 795         br_group.proto = htons(ETH_P_IPV6);
 796         __br_multicast_send_query(br, port, &br_group);
 797 #endif
 798
 799         time = jiffies;
 800         time += sent < br->multicast_startup_query_count ?
 801                 br->multicast_startup_query_interval :
 802                 br->multicast_query_interval;
 803         mod_timer(port ? &port->multicast_query_timer :
 804                          &br->multicast_query_timer, time);
 805 }
 806
 807 static void br_multicast_port_query_expired(unsigned long data)
 808 {
 809         struct net_bridge_port *port = (void *)data;
 810         struct net_bridge *br = port->br;
 811
 812         spin_lock(&br->multicast_lock);
 813         if (port->state == BR_STATE_DISABLED ||
 814             port->state == BR_STATE_BLOCKING)
 815                 goto out;
 816
 817         if (port->multicast_startup_queries_sent <
 818             br->multicast_startup_query_count)
 819                 port->multicast_startup_queries_sent++;
 820
 821         br_multicast_send_query(port->br, port,
 822                                 port->multicast_startup_queries_sent);
 823
 824 out:
 825         spin_unlock(&br->multicast_lock);
 826 }
 827
 828 void br_multicast_add_port(struct net_bridge_port *port)
 829 {
 830         port->multicast_router = 1;
 831
 832         setup_timer(&port->multicast_router_timer, br_multicast_router_expired,
 833                     (unsigned long)port);
 834         setup_timer(&port->multicast_query_timer,
 835                     br_multicast_port_query_expired, (unsigned long)port);
 836 }
 837
 838 void br_multicast_del_port(struct net_bridge_port *port)
 839 {
 840         del_timer_sync(&port->multicast_router_timer);
 841 }
 842
 843 static void __br_multicast_enable_port(struct net_bridge_port *port)
 844 {
 845         port->multicast_startup_queries_sent = 0;
 846
 847         if (try_to_del_timer_sync(&port->multicast_query_timer) >= 0 ||
 848             del_timer(&port->multicast_query_timer))
 849                 mod_timer(&port->multicast_query_timer, jiffies);
 850 }
 851
 852 void br_multicast_enable_port(struct net_bridge_port *port)
 853 {
 854         struct net_bridge *br = port->br;
 855
 856         spin_lock(&br->multicast_lock);
 857         if (br->multicast_disabled || !netif_running(br->dev))
 858                 goto out;
 859
 860         __br_multicast_enable_port(port);
 861
 862 out:
 863         spin_unlock(&br->multicast_lock);
 864 }
 865
 866 void br_multicast_disable_port(struct net_bridge_port *port)
 867 {
 868         struct net_bridge *br = port->br;
 869         struct net_bridge_port_group *pg;
 870         struct hlist_node *p, *n;
 871
 872         spin_lock(&br->multicast_lock);
 873         hlist_for_each_entry_safe(pg, p, n, &port->mglist, mglist)
 874                 br_multicast_del_pg(br, pg);
 875
 876         if (!hlist_unhashed(&port->rlist))
 877                 hlist_del_init_rcu(&port->rlist);
 878         del_timer(&port->multicast_router_timer);
 879         del_timer(&port->multicast_query_timer);
 880         spin_unlock(&br->multicast_lock);
 881 }
 882
 883 static int br_ip4_multicast_igmp3_report(struct net_bridge *br,
 884                                          struct net_bridge_port *port,
 885                                          struct sk_buff *skb)
 886 {
 887         struct igmpv3_report *ih;
 888         struct igmpv3_grec *grec;
 889         int i;
 890         int len;
 891         int num;
 892         int type;
 893         int err = 0;
 894         __be32 group;
 895
 896         if (!pskb_may_pull(skb, sizeof(*ih)))
 897                 return -EINVAL;
 898
 899         ih = igmpv3_report_hdr(skb);
 900         num = ntohs(ih->ngrec);
 901         len = sizeof(*ih);
 902
 903         for (i = 0; i < num; i++) {
 904                 len += sizeof(*grec);
 905                 if (!pskb_may_pull(skb, len))
 906                         return -EINVAL;
 907
 908                 grec = (void *)(skb->data + len - sizeof(*grec));
 909                 group = grec->grec_mca;
 910                 type = grec->grec_type;
 911
 912                 len += ntohs(grec->grec_nsrcs) * 4;
 913                 if (!pskb_may_pull(skb, len))
 914                         return -EINVAL;
 915
 916                 /* We treat this as an IGMPv2 report for now. */
 917                 switch (type) {
 918                 case IGMPV3_MODE_IS_INCLUDE:
 919                 case IGMPV3_MODE_IS_EXCLUDE:
 920                 case IGMPV3_CHANGE_TO_INCLUDE:
 921                 case IGMPV3_CHANGE_TO_EXCLUDE:
 922                 case IGMPV3_ALLOW_NEW_SOURCES:
 923                 case IGMPV3_BLOCK_OLD_SOURCES:
 924                         break;
 925
 926                 default:
 927                         continue;
 928                 }
 929
 930                 err = br_ip4_multicast_add_group(br, port, group);
 931                 if (err)
 932                         break;
 933         }
 934
 935         return err;
 936 }
 937
 938 #if IS_ENABLED(CONFIG_IPV6)
 939 static int br_ip6_multicast_mld2_report(struct net_bridge *br,
 940                                         struct net_bridge_port *port,
 941                                         struct sk_buff *skb)
 942 {
 943         struct icmp6hdr *icmp6h;
 944         struct mld2_grec *grec;
 945         int i;
 946         int len;
 947         int num;
 948         int err = 0;
 949
 950         if (!pskb_may_pull(skb, sizeof(*icmp6h)))
 951                 return -EINVAL;
 952
 953         icmp6h = icmp6_hdr(skb);
 954         num = ntohs(icmp6h->icmp6_dataun.un_data16[1]);
 955         len = sizeof(*icmp6h);
 956
 957         for (i = 0; i < num; i++) {
 958                 __be16 *nsrcs, _nsrcs;
 959
 960                 nsrcs = skb_header_pointer(skb,
 961                                            len + offsetof(struct mld2_grec,
 962                                                           grec_nsrcs),
 963                                            sizeof(_nsrcs), &_nsrcs);
 964                 if (!nsrcs)
 965                         return -EINVAL;
 966
 967                 if (!pskb_may_pull(skb,
 968                                    len + sizeof(*grec) +
 969                                    sizeof(struct in6_addr) * ntohs(*nsrcs)))
 970                         return -EINVAL;
 971
 972                 grec = (struct mld2_grec *)(skb->data + len);
 973                 len += sizeof(*grec) +
 974                        sizeof(struct in6_addr) * ntohs(*nsrcs);
 975
 976                 /* We treat these as MLDv1 reports for now. */
 977                 switch (grec->grec_type) {
 978                 case MLD2_MODE_IS_INCLUDE:
 979                 case MLD2_MODE_IS_EXCLUDE:
 980                 case MLD2_CHANGE_TO_INCLUDE:
 981                 case MLD2_CHANGE_TO_EXCLUDE:
 982                 case MLD2_ALLOW_NEW_SOURCES:
 983                 case MLD2_BLOCK_OLD_SOURCES:
 984                         break;
 985
 986                 default:
 987                         continue;
 988                 }
 989
 990                 err = br_ip6_multicast_add_group(br, port, &grec->grec_mca);
 991                 if (!err)
 992                         break;
 993         }
 994
 995         return err;
 996 }
 997 #endif
 998
 999 /*
1000  * Add port to rotuer_list
1001  *  list is maintained ordered by pointer value
1002  *  and locked by br->multicast_lock and RCU
1003  */
1004 static void br_multicast_add_router(struct net_bridge *br,
1005                                     struct net_bridge_port *port)
1006 {
1007         struct net_bridge_port *p;
1008         struct hlist_node *n, *slot = NULL;
1009
1010         hlist_for_each_entry(p, n, &br->router_list, rlist) {
1011                 if ((unsigned long) port >= (unsigned long) p)
1012                         break;
1013                 slot = n;
1014         }
1015
1016         if (slot)
1017                 hlist_add_after_rcu(slot, &port->rlist);
1018         else
1019                 hlist_add_head_rcu(&port->rlist, &br->router_list);
1020 }
1021
1022 static void br_multicast_mark_router(struct net_bridge *br,
1023                                      struct net_bridge_port *port)
1024 {
1025         unsigned long now = jiffies;
1026
1027         if (!port) {
1028                 if (br->multicast_router == 1)
1029                         mod_timer(&br->multicast_router_timer,
1030                                   now + br->multicast_querier_interval);
1031                 return;
1032         }
1033
1034         if (port->multicast_router != 1)
1035                 return;
1036
1037         if (!hlist_unhashed(&port->rlist))
1038                 goto timer;
1039
1040         br_multicast_add_router(br, port);
1041
1042 timer:
1043         mod_timer(&port->multicast_router_timer,
1044                   now + br->multicast_querier_interval);
1045 }
1046
1047 static void br_multicast_query_received(struct net_bridge *br,
1048                                         struct net_bridge_port *port,
1049                                         int saddr)
1050 {
1051         if (saddr)
1052                 mod_timer(&br->multicast_querier_timer,
1053                           jiffies + br->multicast_querier_interval);
1054         else if (timer_pending(&br->multicast_querier_timer))
1055                 return;
1056
1057         br_multicast_mark_router(br, port);
1058 }
1059
1060 static int br_ip4_multicast_query(struct net_bridge *br,
1061                                   struct net_bridge_port *port,
1062                                   struct sk_buff *skb)
1063 {
1064         const struct iphdr *iph = ip_hdr(skb);
1065         struct igmphdr *ih = igmp_hdr(skb);
1066         struct net_bridge_mdb_entry *mp;
1067         struct igmpv3_query *ih3;
1068         struct net_bridge_port_group *p;
1069         struct net_bridge_port_group __rcu **pp;
1070         unsigned long max_delay;
1071         unsigned long now = jiffies;
1072         __be32 group;
1073         int err = 0;
1074
1075         spin_lock(&br->multicast_lock);
1076         if (!netif_running(br->dev) ||
1077             (port && port->state == BR_STATE_DISABLED))
1078                 goto out;
1079
1080         br_multicast_query_received(br, port, !!iph->saddr);
1081
1082         group = ih->group;
1083
1084         if (skb->len == sizeof(*ih)) {
1085                 max_delay = ih->code * (HZ / IGMP_TIMER_SCALE);
1086
1087                 if (!max_delay) {
1088                         max_delay = 10 * HZ;
1089                         group = 0;
1090                 }
1091         } else {
1092                 if (!pskb_may_pull(skb, sizeof(struct igmpv3_query))) {
1093                         err = -EINVAL;
1094                         goto out;
1095                 }
1096
1097                 ih3 = igmpv3_query_hdr(skb);
1098                 if (ih3->nsrcs)
1099                         goto out;
1100
1101                 max_delay = ih3->code ?
1102                             IGMPV3_MRC(ih3->code) * (HZ / IGMP_TIMER_SCALE) : 1;
1103         }
1104
1105         if (!group)
1106                 goto out;
1107
1108         mp = br_mdb_ip4_get(mlock_dereference(br->mdb, br), group);
1109         if (!mp)
1110                 goto out;
1111
1112         max_delay *= br->multicast_last_member_count;
1113
1114         if (mp->mglist &&
1115             (timer_pending(&mp->timer) ?
1116              time_after(mp->timer.expires, now + max_delay) :
1117              try_to_del_timer_sync(&mp->timer) >= 0))
1118                 mod_timer(&mp->timer, now + max_delay);
1119
1120         for (pp = &mp->ports;
1121              (p = mlock_dereference(*pp, br)) != NULL;
1122              pp = &p->next) {
1123                 if (timer_pending(&p->timer) ?
1124                     time_after(p->timer.expires, now + max_delay) :
1125                     try_to_del_timer_sync(&p->timer) >= 0)
1126                         mod_timer(&p->timer, now + max_delay);
1127         }
1128
1129 out:
1130         spin_unlock(&br->multicast_lock);
1131         return err;
1132 }
1133
1134 #if IS_ENABLED(CONFIG_IPV6)
1135 static int br_ip6_multicast_query(struct net_bridge *br,
1136                                   struct net_bridge_port *port,
1137                                   struct sk_buff *skb)
1138 {
1139         const struct ipv6hdr *ip6h = ipv6_hdr(skb);
1140         struct mld_msg *mld = (struct mld_msg *) icmp6_hdr(skb);
1141         struct net_bridge_mdb_entry *mp;
1142         struct mld2_query *mld2q;
1143         struct net_bridge_port_group *p;
1144         struct net_bridge_port_group __rcu **pp;
1145         unsigned long max_delay;
1146         unsigned long now = jiffies;
1147         const struct in6_addr *group = NULL;
1148         int err = 0;
1149
1150         spin_lock(&br->multicast_lock);
1151         if (!netif_running(br->dev) ||
1152             (port && port->state == BR_STATE_DISABLED))
1153                 goto out;
1154
1155         br_multicast_query_received(br, port, !ipv6_addr_any(&ip6h->saddr));
1156
1157         if (skb->len == sizeof(*mld)) {
1158                 if (!pskb_may_pull(skb, sizeof(*mld))) {
1159                         err = -EINVAL;
1160                         goto out;
1161                 }
1162                 mld = (struct mld_msg *) icmp6_hdr(skb);
1163                 max_delay = msecs_to_jiffies(htons(mld->mld_maxdelay));
1164                 if (max_delay)
1165                         group = &mld->mld_mca;
1166         } else if (skb->len >= sizeof(*mld2q)) {
1167                 if (!pskb_may_pull(skb, sizeof(*mld2q))) {
1168                         err = -EINVAL;
1169                         goto out;
1170                 }
1171                 mld2q = (struct mld2_query *)icmp6_hdr(skb);
1172                 if (!mld2q->mld2q_nsrcs)
1173                         group = &mld2q->mld2q_mca;
1174                 max_delay = mld2q->mld2q_mrc ? MLDV2_MRC(mld2q->mld2q_mrc) : 1;
1175         }
1176
1177         if (!group)
1178                 goto out;
1179
1180         mp = br_mdb_ip6_get(mlock_dereference(br->mdb, br), group);
1181         if (!mp)
1182                 goto out;
1183
1184         max_delay *= br->multicast_last_member_count;
1185         if (mp->mglist &&
1186             (timer_pending(&mp->timer) ?
1187              time_after(mp->timer.expires, now + max_delay) :
1188              try_to_del_timer_sync(&mp->timer) >= 0))
1189                 mod_timer(&mp->timer, now + max_delay);
1190
1191         for (pp = &mp->ports;
1192              (p = mlock_dereference(*pp, br)) != NULL;
1193              pp = &p->next) {
1194                 if (timer_pending(&p->timer) ?
1195                     time_after(p->timer.expires, now + max_delay) :
1196                     try_to_del_timer_sync(&p->timer) >= 0)
1197                         mod_timer(&p->timer, now + max_delay);
1198         }
1199
1200 out:
1201         spin_unlock(&br->multicast_lock);
1202         return err;
1203 }
1204 #endif
1205
1206 static void br_multicast_leave_group(struct net_bridge *br,
1207                                      struct net_bridge_port *port,
1208                                      struct br_ip *group)
1209 {
1210         struct net_bridge_mdb_htable *mdb;
1211         struct net_bridge_mdb_entry *mp;
1212         struct net_bridge_port_group *p;
1213         unsigned long now;
1214         unsigned long time;
1215
1216         spin_lock(&br->multicast_lock);
1217         if (!netif_running(br->dev) ||
1218             (port && port->state == BR_STATE_DISABLED) ||
1219             timer_pending(&br->multicast_querier_timer))
1220                 goto out;
1221
1222         mdb = mlock_dereference(br->mdb, br);
1223         mp = br_mdb_ip_get(mdb, group);
1224         if (!mp)
1225                 goto out;
1226
1227         now = jiffies;
1228         time = now + br->multicast_last_member_count *
1229                      br->multicast_last_member_interval;
1230
1231         if (!port) {
1232                 if (mp->mglist &&
1233                     (timer_pending(&mp->timer) ?
1234                      time_after(mp->timer.expires, time) :
1235                      try_to_del_timer_sync(&mp->timer) >= 0)) {
1236                         mod_timer(&mp->timer, time);
1237                 }
1238
1239                 goto out;
1240         }
1241
1242         for (p = mlock_dereference(mp->ports, br);
1243              p != NULL;
1244              p = mlock_dereference(p->next, br)) {
1245                 if (p->port != port)
1246                         continue;
1247
1248                 if (!hlist_unhashed(&p->mglist) &&
1249                     (timer_pending(&p->timer) ?
1250                      time_after(p->timer.expires, time) :
1251                      try_to_del_timer_sync(&p->timer) >= 0)) {
1252                         mod_timer(&p->timer, time);
1253                 }
1254
1255                 break;
1256         }
1257
1258 out:
1259         spin_unlock(&br->multicast_lock);
1260 }
1261
1262 static void br_ip4_multicast_leave_group(struct net_bridge *br,
1263                                          struct net_bridge_port *port,
1264                                          __be32 group)
1265 {
1266         struct br_ip br_group;
1267
1268         if (ipv4_is_local_multicast(group))
1269                 return;
1270
1271         br_group.u.ip4 = group;
1272         br_group.proto = htons(ETH_P_IP);
1273
1274         br_multicast_leave_group(br, port, &br_group);
1275 }
1276
1277 #if IS_ENABLED(CONFIG_IPV6)
1278 static void br_ip6_multicast_leave_group(struct net_bridge *br,
1279                                          struct net_bridge_port *port,
1280                                          const struct in6_addr *group)
1281 {
1282         struct br_ip br_group;
1283
1284         if (!ipv6_is_transient_multicast(group))
1285                 return;
1286
1287         br_group.u.ip6 = *group;
1288         br_group.proto = htons(ETH_P_IPV6);
1289
1290         br_multicast_leave_group(br, port, &br_group);
1291 }
1292 #endif
1293
1294 static int br_multicast_ipv4_rcv(struct net_bridge *br,
1295                                  struct net_bridge_port *port,
1296                                  struct sk_buff *skb)
1297 {
1298         struct sk_buff *skb2 = skb;
1299         const struct iphdr *iph;
1300         struct igmphdr *ih;
1301         unsigned int len;
1302         unsigned int offset;
1303         int err;
1304
1305         /* We treat OOM as packet loss for now. */
1306         if (!pskb_may_pull(skb, sizeof(*iph)))
1307                 return -EINVAL;
1308
1309         iph = ip_hdr(skb);
1310
1311         if (iph->ihl < 5 || iph->version != 4)
1312                 return -EINVAL;
1313
1314         if (!pskb_may_pull(skb, ip_hdrlen(skb)))
1315                 return -EINVAL;
1316
1317         iph = ip_hdr(skb);
1318
1319         if (unlikely(ip_fast_csum((u8 *)iph, iph->ihl)))
1320                 return -EINVAL;
1321
1322         if (iph->protocol != IPPROTO_IGMP) {
1323                 if ((iph->daddr & IGMP_LOCAL_GROUP_MASK) != IGMP_LOCAL_GROUP)
1324                         BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1325                 return 0;
1326         }
1327
1328         len = ntohs(iph->tot_len);
1329         if (skb->len < len || len < ip_hdrlen(skb))
1330                 return -EINVAL;
1331
1332         if (skb->len > len) {
1333                 skb2 = skb_clone(skb, GFP_ATOMIC);
1334                 if (!skb2)
1335                         return -ENOMEM;
1336
1337                 err = pskb_trim_rcsum(skb2, len);
1338                 if (err)
1339                         goto err_out;
1340         }
1341
1342         len -= ip_hdrlen(skb2);
1343         offset = skb_network_offset(skb2) + ip_hdrlen(skb2);
1344         __skb_pull(skb2, offset);
1345         skb_reset_transport_header(skb2);
1346
1347         err = -EINVAL;
1348         if (!pskb_may_pull(skb2, sizeof(*ih)))
1349                 goto out;
1350
1351         switch (skb2->ip_summed) {
1352         case CHECKSUM_COMPLETE:
1353                 if (!csum_fold(skb2->csum))
1354                         break;
1355                 /* fall through */
1356         case CHECKSUM_NONE:
1357                 skb2->csum = 0;
1358                 if (skb_checksum_complete(skb2))
1359                         goto out;
1360         }
1361
1362         err = 0;
1363
1364         BR_INPUT_SKB_CB(skb)->igmp = 1;
1365         ih = igmp_hdr(skb2);
1366
1367         switch (ih->type) {
1368         case IGMP_HOST_MEMBERSHIP_REPORT:
1369         case IGMPV2_HOST_MEMBERSHIP_REPORT:
1370                 BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1371                 err = br_ip4_multicast_add_group(br, port, ih->group);
1372                 break;
1373         case IGMPV3_HOST_MEMBERSHIP_REPORT:
1374                 err = br_ip4_multicast_igmp3_report(br, port, skb2);
1375                 break;
1376         case IGMP_HOST_MEMBERSHIP_QUERY:
1377                 err = br_ip4_multicast_query(br, port, skb2);
1378                 break;
1379         case IGMP_HOST_LEAVE_MESSAGE:
1380                 br_ip4_multicast_leave_group(br, port, ih->group);
1381                 break;
1382         }
1383
1384 out:
1385         __skb_push(skb2, offset);
1386 err_out:
1387         if (skb2 != skb)
1388                 kfree_skb(skb2);
1389         return err;
1390 }
1391
1392 #if IS_ENABLED(CONFIG_IPV6)
1393 static int br_multicast_ipv6_rcv(struct net_bridge *br,
1394                                  struct net_bridge_port *port,
1395                                  struct sk_buff *skb)
1396 {
1397         struct sk_buff *skb2;
1398         const struct ipv6hdr *ip6h;
1399         u8 icmp6_type;
1400         u8 nexthdr;
1401         __be16 frag_off;
1402         unsigned int len;
1403         int offset;
1404         int err;
1405
1406         if (!pskb_may_pull(skb, sizeof(*ip6h)))
1407                 return -EINVAL;
1408
1409         ip6h = ipv6_hdr(skb);
1410
1411         /*
1412          * We're interested in MLD messages only.
1413          *  - Version is 6
1414          *  - MLD has always Router Alert hop-by-hop option
1415          *  - But we do not support jumbrograms.
1416          */
1417         if (ip6h->version != 6 ||
1418             ip6h->nexthdr != IPPROTO_HOPOPTS ||
1419             ip6h->payload_len == 0)
1420                 return 0;
1421
1422         len = ntohs(ip6h->payload_len) + sizeof(*ip6h);
1423         if (skb->len < len)
1424                 return -EINVAL;
1425
1426         nexthdr = ip6h->nexthdr;
1427         offset = ipv6_skip_exthdr(skb, sizeof(*ip6h), &nexthdr, &frag_off);
1428
1429         if (offset < 0 || nexthdr != IPPROTO_ICMPV6)
1430                 return 0;
1431
1432         /* Okay, we found ICMPv6 header */
1433         skb2 = skb_clone(skb, GFP_ATOMIC);
1434         if (!skb2)
1435                 return -ENOMEM;
1436
1437         err = -EINVAL;
1438         if (!pskb_may_pull(skb2, offset + sizeof(struct icmp6hdr)))
1439                 goto out;
1440
1441         len -= offset - skb_network_offset(skb2);
1442
1443         __skb_pull(skb2, offset);
1444         skb_reset_transport_header(skb2);
1445         skb_postpull_rcsum(skb2, skb_network_header(skb2),
1446                            skb_network_header_len(skb2));
1447
1448         icmp6_type = icmp6_hdr(skb2)->icmp6_type;
1449
1450         switch (icmp6_type) {
1451         case ICMPV6_MGM_QUERY:
1452         case ICMPV6_MGM_REPORT:
1453         case ICMPV6_MGM_REDUCTION:
1454         case ICMPV6_MLD2_REPORT:
1455                 break;
1456         default:
1457                 err = 0;
1458                 goto out;
1459         }
1460
1461         /* Okay, we found MLD message. Check further. */
1462         if (skb2->len > len) {
1463                 err = pskb_trim_rcsum(skb2, len);
1464                 if (err)
1465                         goto out;
1466                 err = -EINVAL;
1467         }
1468
1469         ip6h = ipv6_hdr(skb2);
1470
1471         switch (skb2->ip_summed) {
1472         case CHECKSUM_COMPLETE:
1473                 if (!csum_ipv6_magic(&ip6h->saddr, &ip6h->daddr, skb2->len,
1474                                         IPPROTO_ICMPV6, skb2->csum))
1475                         break;
1476                 /*FALLTHROUGH*/
1477         case CHECKSUM_NONE:
1478                 skb2->csum = ~csum_unfold(csum_ipv6_magic(&ip6h->saddr,
1479                                                         &ip6h->daddr,
1480                                                         skb2->len,
1481                                                         IPPROTO_ICMPV6, 0));
1482                 if (__skb_checksum_complete(skb2))
1483                         goto out;
1484         }
1485
1486         err = 0;
1487
1488         BR_INPUT_SKB_CB(skb)->igmp = 1;
1489
1490         switch (icmp6_type) {
1491         case ICMPV6_MGM_REPORT:
1492             {
1493                 struct mld_msg *mld;
1494                 if (!pskb_may_pull(skb2, sizeof(*mld))) {
1495                         err = -EINVAL;
1496                         goto out;
1497                 }
1498                 mld = (struct mld_msg *)skb_transport_header(skb2);
1499                 BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1500                 err = br_ip6_multicast_add_group(br, port, &mld->mld_mca);
1501                 break;
1502             }
1503         case ICMPV6_MLD2_REPORT:
1504                 err = br_ip6_multicast_mld2_report(br, port, skb2);
1505                 break;
1506         case ICMPV6_MGM_QUERY:
1507                 err = br_ip6_multicast_query(br, port, skb2);
1508                 break;
1509         case ICMPV6_MGM_REDUCTION:
1510             {
1511                 struct mld_msg *mld;
1512                 if (!pskb_may_pull(skb2, sizeof(*mld))) {
1513                         err = -EINVAL;
1514                         goto out;
1515                 }
1516                 mld = (struct mld_msg *)skb_transport_header(skb2);
1517                 br_ip6_multicast_leave_group(br, port, &mld->mld_mca);
1518             }
1519         }
1520
1521 out:
1522         kfree_skb(skb2);
1523         return err;
1524 }
1525 #endif
1526
1527 int br_multicast_rcv(struct net_bridge *br, struct net_bridge_port *port,
1528                      struct sk_buff *skb)
1529 {
1530         BR_INPUT_SKB_CB(skb)->igmp = 0;
1531         BR_INPUT_SKB_CB(skb)->mrouters_only = 0;
1532
1533         if (br->multicast_disabled)
1534                 return 0;
1535
1536         switch (skb->protocol) {
1537         case htons(ETH_P_IP):
1538                 return br_multicast_ipv4_rcv(br, port, skb);
1539 #if IS_ENABLED(CONFIG_IPV6)
1540         case htons(ETH_P_IPV6):
1541                 return br_multicast_ipv6_rcv(br, port, skb);
1542 #endif
1543         }
1544
1545         return 0;
1546 }
1547
1548 static void br_multicast_query_expired(unsigned long data)
1549 {
1550         struct net_bridge *br = (void *)data;
1551
1552         spin_lock(&br->multicast_lock);
1553         if (br->multicast_startup_queries_sent <
1554             br->multicast_startup_query_count)
1555                 br->multicast_startup_queries_sent++;
1556
1557         br_multicast_send_query(br, NULL, br->multicast_startup_queries_sent);
1558
1559         spin_unlock(&br->multicast_lock);
1560 }
1561
1562 void br_multicast_init(struct net_bridge *br)
1563 {
1564         br->hash_elasticity = 4;
1565         br->hash_max = 512;
1566
1567         br->multicast_router = 1;
1568         br->multicast_querier = 0;
1569         br->multicast_last_member_count = 2;
1570         br->multicast_startup_query_count = 2;
1571
1572         br->multicast_last_member_interval = HZ;
1573         br->multicast_query_response_interval = 10 * HZ;
1574         br->multicast_startup_query_interval = 125 * HZ / 4;
1575         br->multicast_query_interval = 125 * HZ;
1576         br->multicast_querier_interval = 255 * HZ;
1577         br->multicast_membership_interval = 260 * HZ;
1578
1579         spin_lock_init(&br->multicast_lock);
1580         setup_timer(&br->multicast_router_timer,
1581                     br_multicast_local_router_expired, 0);
1582         setup_timer(&br->multicast_querier_timer,
1583                     br_multicast_querier_expired, (unsigned long)br);
1584         setup_timer(&br->multicast_query_timer, br_multicast_query_expired,
1585                     (unsigned long)br);
1586 }
1587
1588 void br_multicast_open(struct net_bridge *br)
1589 {
1590         br->multicast_startup_queries_sent = 0;
1591
1592         if (br->multicast_disabled)
1593                 return;
1594
1595         mod_timer(&br->multicast_query_timer, jiffies);
1596 }
1597
1598 void br_multicast_stop(struct net_bridge *br)
1599 {
1600         struct net_bridge_mdb_htable *mdb;
1601         struct net_bridge_mdb_entry *mp;
1602         struct hlist_node *p, *n;
1603         u32 ver;
1604         int i;
1605
1606         del_timer_sync(&br->multicast_router_timer);
1607         del_timer_sync(&br->multicast_querier_timer);
1608         del_timer_sync(&br->multicast_query_timer);
1609
1610         spin_lock_bh(&br->multicast_lock);
1611         mdb = mlock_dereference(br->mdb, br);
1612         if (!mdb)
1613                 goto out;
1614
1615         br->mdb = NULL;
1616
1617         ver = mdb->ver;
1618         for (i = 0; i < mdb->max; i++) {
1619                 hlist_for_each_entry_safe(mp, p, n, &mdb->mhash[i],
1620                                           hlist[ver]) {
1621                         del_timer(&mp->timer);
1622                         call_rcu_bh(&mp->rcu, br_multicast_free_group);
1623                 }
1624         }
1625
1626         if (mdb->old) {
1627                 spin_unlock_bh(&br->multicast_lock);
1628                 rcu_barrier_bh();
1629                 spin_lock_bh(&br->multicast_lock);
1630                 WARN_ON(mdb->old);
1631         }
1632
1633         mdb->old = mdb;
1634         call_rcu_bh(&mdb->rcu, br_mdb_free);
1635
1636 out:
1637         spin_unlock_bh(&br->multicast_lock);
1638 }
1639
1640 int br_multicast_set_router(struct net_bridge *br, unsigned long val)
1641 {
1642         int err = -ENOENT;
1643
1644         spin_lock_bh(&br->multicast_lock);
1645         if (!netif_running(br->dev))
1646                 goto unlock;
1647
1648         switch (val) {
1649         case 0:
1650         case 2:
1651                 del_timer(&br->multicast_router_timer);
1652                 /* fall through */
1653         case 1:
1654                 br->multicast_router = val;
1655                 err = 0;
1656                 break;
1657
1658         default:
1659                 err = -EINVAL;
1660                 break;
1661         }
1662
1663 unlock:
1664         spin_unlock_bh(&br->multicast_lock);
1665
1666         return err;
1667 }
1668
1669 int br_multicast_set_port_router(struct net_bridge_port *p, unsigned long val)
1670 {
1671         struct net_bridge *br = p->br;
1672         int err = -ENOENT;
1673
1674         spin_lock(&br->multicast_lock);
1675         if (!netif_running(br->dev) || p->state == BR_STATE_DISABLED)
1676                 goto unlock;
1677
1678         switch (val) {
1679         case 0:
1680         case 1:
1681         case 2:
1682                 p->multicast_router = val;
1683                 err = 0;
1684
1685                 if (val < 2 && !hlist_unhashed(&p->rlist))
1686                         hlist_del_init_rcu(&p->rlist);
1687
1688                 if (val == 1)
1689                         break;
1690
1691                 del_timer(&p->multicast_router_timer);
1692
1693                 if (val == 0)
1694                         break;
1695
1696                 br_multicast_add_router(br, p);
1697                 break;
1698
1699         default:
1700                 err = -EINVAL;
1701                 break;
1702         }
1703
1704 unlock:
1705         spin_unlock(&br->multicast_lock);
1706
1707         return err;
1708 }
1709
1710 static void br_multicast_start_querier(struct net_bridge *br)
1711 {
1712         struct net_bridge_port *port;
1713
1714         br_multicast_open(br);
1715
1716         list_for_each_entry(port, &br->port_list, list) {
1717                 if (port->state == BR_STATE_DISABLED ||
1718                     port->state == BR_STATE_BLOCKING)
1719                         continue;
1720
1721                 __br_multicast_enable_port(port);
1722         }
1723 }
1724
1725 int br_multicast_toggle(struct net_bridge *br, unsigned long val)
1726 {
1727         int err = 0;
1728         struct net_bridge_mdb_htable *mdb;
1729
1730         spin_lock_bh(&br->multicast_lock);
1731         if (br->multicast_disabled == !val)
1732                 goto unlock;
1733
1734         br->multicast_disabled = !val;
1735         if (br->multicast_disabled)
1736                 goto unlock;
1737
1738         if (!netif_running(br->dev))
1739                 goto unlock;
1740
1741         mdb = mlock_dereference(br->mdb, br);
1742         if (mdb) {
1743                 if (mdb->old) {
1744                         err = -EEXIST;
1745 rollback:
1746                         br->multicast_disabled = !!val;
1747                         goto unlock;
1748                 }
1749
1750                 err = br_mdb_rehash(&br->mdb, mdb->max,
1751                                     br->hash_elasticity);
1752                 if (err)
1753                         goto rollback;
1754         }
1755
1756         br_multicast_start_querier(br);
1757
1758 unlock:
1759         spin_unlock_bh(&br->multicast_lock);
1760
1761         return err;
1762 }
1763
1764 int br_multicast_set_querier(struct net_bridge *br, unsigned long val)
1765 {
1766         val = !!val;
1767
1768         spin_lock_bh(&br->multicast_lock);
1769         if (br->multicast_querier == val)
1770                 goto unlock;
1771
1772         br->multicast_querier = val;
1773         if (val)
1774                 br_multicast_start_querier(br);
1775
1776 unlock:
1777         spin_unlock_bh(&br->multicast_lock);
1778
1779         return 0;
1780 }
1781
1782 int br_multicast_set_hash_max(struct net_bridge *br, unsigned long val)
1783 {
1784         int err = -ENOENT;
1785         u32 old;
1786         struct net_bridge_mdb_htable *mdb;
1787
1788         spin_lock(&br->multicast_lock);
1789         if (!netif_running(br->dev))
1790                 goto unlock;
1791
1792         err = -EINVAL;
1793         if (!is_power_of_2(val))
1794                 goto unlock;
1795
1796         mdb = mlock_dereference(br->mdb, br);
1797         if (mdb && val < mdb->size)
1798                 goto unlock;
1799
1800         err = 0;
1801
1802         old = br->hash_max;
1803         br->hash_max = val;
1804
1805         if (mdb) {
1806                 if (mdb->old) {
1807                         err = -EEXIST;
1808 rollback:
1809                         br->hash_max = old;
1810                         goto unlock;
1811                 }
1812
1813                 err = br_mdb_rehash(&br->mdb, br->hash_max,
1814                                     br->hash_elasticity);
1815                 if (err)
1816                         goto rollback;
1817         }
1818
1819 unlock:
1820         spin_unlock(&br->multicast_lock);
1821
1822         return err;
1823 }