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         del_timer(&mp->query_timer);
 245         call_rcu_bh(&mp->rcu, br_multicast_free_group);
 246
 247 out:
 248         spin_unlock(&br->multicast_lock);
 249 }
 250
 251 static void br_multicast_del_pg(struct net_bridge *br,
 252                                 struct net_bridge_port_group *pg)
 253 {
 254         struct net_bridge_mdb_htable *mdb;
 255         struct net_bridge_mdb_entry *mp;
 256         struct net_bridge_port_group *p;
 257         struct net_bridge_port_group __rcu **pp;
 258
 259         mdb = mlock_dereference(br->mdb, br);
 260
 261         mp = br_mdb_ip_get(mdb, &pg->addr);
 262         if (WARN_ON(!mp))
 263                 return;
 264
 265         for (pp = &mp->ports;
 266              (p = mlock_dereference(*pp, br)) != NULL;
 267              pp = &p->next) {
 268                 if (p != pg)
 269                         continue;
 270
 271                 rcu_assign_pointer(*pp, p->next);
 272                 hlist_del_init(&p->mglist);
 273                 del_timer(&p->timer);
 274                 del_timer(&p->query_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_PAD0;              /* Pad0 */
 464         hopopt[7] = IPV6_TLV_PAD0;              /* Pad0 */
 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 void br_multicast_send_group_query(struct net_bridge_mdb_entry *mp)
 511 {
 512         struct net_bridge *br = mp->br;
 513         struct sk_buff *skb;
 514
 515         skb = br_multicast_alloc_query(br, &mp->addr);
 516         if (!skb)
 517                 goto timer;
 518
 519         netif_rx(skb);
 520
 521 timer:
 522         if (++mp->queries_sent < br->multicast_last_member_count)
 523                 mod_timer(&mp->query_timer,
 524                           jiffies + br->multicast_last_member_interval);
 525 }
 526
 527 static void br_multicast_group_query_expired(unsigned long data)
 528 {
 529         struct net_bridge_mdb_entry *mp = (void *)data;
 530         struct net_bridge *br = mp->br;
 531
 532         spin_lock(&br->multicast_lock);
 533         if (!netif_running(br->dev) || !mp->mglist ||
 534             mp->queries_sent >= br->multicast_last_member_count)
 535                 goto out;
 536
 537         br_multicast_send_group_query(mp);
 538
 539 out:
 540         spin_unlock(&br->multicast_lock);
 541 }
 542
 543 static void br_multicast_send_port_group_query(struct net_bridge_port_group *pg)
 544 {
 545         struct net_bridge_port *port = pg->port;
 546         struct net_bridge *br = port->br;
 547         struct sk_buff *skb;
 548
 549         skb = br_multicast_alloc_query(br, &pg->addr);
 550         if (!skb)
 551                 goto timer;
 552
 553         br_deliver(port, skb);
 554
 555 timer:
 556         if (++pg->queries_sent < br->multicast_last_member_count)
 557                 mod_timer(&pg->query_timer,
 558                           jiffies + br->multicast_last_member_interval);
 559 }
 560
 561 static void br_multicast_port_group_query_expired(unsigned long data)
 562 {
 563         struct net_bridge_port_group *pg = (void *)data;
 564         struct net_bridge_port *port = pg->port;
 565         struct net_bridge *br = port->br;
 566
 567         spin_lock(&br->multicast_lock);
 568         if (!netif_running(br->dev) || hlist_unhashed(&pg->mglist) ||
 569             pg->queries_sent >= br->multicast_last_member_count)
 570                 goto out;
 571
 572         br_multicast_send_port_group_query(pg);
 573
 574 out:
 575         spin_unlock(&br->multicast_lock);
 576 }
 577
 578 static struct net_bridge_mdb_entry *br_multicast_get_group(
 579         struct net_bridge *br, struct net_bridge_port *port,
 580         struct br_ip *group, int hash)
 581 {
 582         struct net_bridge_mdb_htable *mdb;
 583         struct net_bridge_mdb_entry *mp;
 584         struct hlist_node *p;
 585         unsigned count = 0;
 586         unsigned max;
 587         int elasticity;
 588         int err;
 589
 590         mdb = rcu_dereference_protected(br->mdb, 1);
 591         hlist_for_each_entry(mp, p, &mdb->mhash[hash], hlist[mdb->ver]) {
 592                 count++;
 593                 if (unlikely(br_ip_equal(group, &mp->addr)))
 594                         return mp;
 595         }
 596
 597         elasticity = 0;
 598         max = mdb->max;
 599
 600         if (unlikely(count > br->hash_elasticity && count)) {
 601                 if (net_ratelimit())
 602                         br_info(br, "Multicast hash table "
 603                                 "chain limit reached: %s\n",
 604                                 port ? port->dev->name : br->dev->name);
 605
 606                 elasticity = br->hash_elasticity;
 607         }
 608
 609         if (mdb->size >= max) {
 610                 max *= 2;
 611                 if (unlikely(max >= br->hash_max)) {
 612                         br_warn(br, "Multicast hash table maximum "
 613                                 "reached, disabling snooping: %s, %d\n",
 614                                 port ? port->dev->name : br->dev->name, max);
 615                         err = -E2BIG;
 616 disable:
 617                         br->multicast_disabled = 1;
 618                         goto err;
 619                 }
 620         }
 621
 622         if (max > mdb->max || elasticity) {
 623                 if (mdb->old) {
 624                         if (net_ratelimit())
 625                                 br_info(br, "Multicast hash table "
 626                                         "on fire: %s\n",
 627                                         port ? port->dev->name : br->dev->name);
 628                         err = -EEXIST;
 629                         goto err;
 630                 }
 631
 632                 err = br_mdb_rehash(&br->mdb, max, elasticity);
 633                 if (err) {
 634                         br_warn(br, "Cannot rehash multicast "
 635                                 "hash table, disabling snooping: %s, %d, %d\n",
 636                                 port ? port->dev->name : br->dev->name,
 637                                 mdb->size, err);
 638                         goto disable;
 639                 }
 640
 641                 err = -EAGAIN;
 642                 goto err;
 643         }
 644
 645         return NULL;
 646
 647 err:
 648         mp = ERR_PTR(err);
 649         return mp;
 650 }
 651
 652 static struct net_bridge_mdb_entry *br_multicast_new_group(
 653         struct net_bridge *br, struct net_bridge_port *port,
 654         struct br_ip *group)
 655 {
 656         struct net_bridge_mdb_htable *mdb;
 657         struct net_bridge_mdb_entry *mp;
 658         int hash;
 659         int err;
 660
 661         mdb = rcu_dereference_protected(br->mdb, 1);
 662         if (!mdb) {
 663                 err = br_mdb_rehash(&br->mdb, BR_HASH_SIZE, 0);
 664                 if (err)
 665                         return ERR_PTR(err);
 666                 goto rehash;
 667         }
 668
 669         hash = br_ip_hash(mdb, group);
 670         mp = br_multicast_get_group(br, port, group, hash);
 671         switch (PTR_ERR(mp)) {
 672         case 0:
 673                 break;
 674
 675         case -EAGAIN:
 676 rehash:
 677                 mdb = rcu_dereference_protected(br->mdb, 1);
 678                 hash = br_ip_hash(mdb, group);
 679                 break;
 680
 681         default:
 682                 goto out;
 683         }
 684
 685         mp = kzalloc(sizeof(*mp), GFP_ATOMIC);
 686         if (unlikely(!mp))
 687                 return ERR_PTR(-ENOMEM);
 688
 689         mp->br = br;
 690         mp->addr = *group;
 691         setup_timer(&mp->timer, br_multicast_group_expired,
 692                     (unsigned long)mp);
 693         setup_timer(&mp->query_timer, br_multicast_group_query_expired,
 694                     (unsigned long)mp);
 695
 696         hlist_add_head_rcu(&mp->hlist[mdb->ver], &mdb->mhash[hash]);
 697         mdb->size++;
 698
 699 out:
 700         return mp;
 701 }
 702
 703 static int br_multicast_add_group(struct net_bridge *br,
 704                                   struct net_bridge_port *port,
 705                                   struct br_ip *group)
 706 {
 707         struct net_bridge_mdb_entry *mp;
 708         struct net_bridge_port_group *p;
 709         struct net_bridge_port_group __rcu **pp;
 710         unsigned long now = jiffies;
 711         int err;
 712
 713         spin_lock(&br->multicast_lock);
 714         if (!netif_running(br->dev) ||
 715             (port && port->state == BR_STATE_DISABLED))
 716                 goto out;
 717
 718         mp = br_multicast_new_group(br, port, group);
 719         err = PTR_ERR(mp);
 720         if (IS_ERR(mp))
 721                 goto err;
 722
 723         if (!port) {
 724                 mp->mglist = true;
 725                 mod_timer(&mp->timer, now + br->multicast_membership_interval);
 726                 goto out;
 727         }
 728
 729         for (pp = &mp->ports;
 730              (p = mlock_dereference(*pp, br)) != NULL;
 731              pp = &p->next) {
 732                 if (p->port == port)
 733                         goto found;
 734                 if ((unsigned long)p->port < (unsigned long)port)
 735                         break;
 736         }
 737
 738         p = kzalloc(sizeof(*p), GFP_ATOMIC);
 739         err = -ENOMEM;
 740         if (unlikely(!p))
 741                 goto err;
 742
 743         p->addr = *group;
 744         p->port = port;
 745         p->next = *pp;
 746         hlist_add_head(&p->mglist, &port->mglist);
 747         setup_timer(&p->timer, br_multicast_port_group_expired,
 748                     (unsigned long)p);
 749         setup_timer(&p->query_timer, br_multicast_port_group_query_expired,
 750                     (unsigned long)p);
 751
 752         rcu_assign_pointer(*pp, p);
 753
 754 found:
 755         mod_timer(&p->timer, now + br->multicast_membership_interval);
 756 out:
 757         err = 0;
 758
 759 err:
 760         spin_unlock(&br->multicast_lock);
 761         return err;
 762 }
 763
 764 static int br_ip4_multicast_add_group(struct net_bridge *br,
 765                                       struct net_bridge_port *port,
 766                                       __be32 group)
 767 {
 768         struct br_ip br_group;
 769
 770         if (ipv4_is_local_multicast(group))
 771                 return 0;
 772
 773         br_group.u.ip4 = group;
 774         br_group.proto = htons(ETH_P_IP);
 775
 776         return br_multicast_add_group(br, port, &br_group);
 777 }
 778
 779 #if IS_ENABLED(CONFIG_IPV6)
 780 static int br_ip6_multicast_add_group(struct net_bridge *br,
 781                                       struct net_bridge_port *port,
 782                                       const struct in6_addr *group)
 783 {
 784         struct br_ip br_group;
 785
 786         if (!ipv6_is_transient_multicast(group))
 787                 return 0;
 788
 789         br_group.u.ip6 = *group;
 790         br_group.proto = htons(ETH_P_IPV6);
 791
 792         return br_multicast_add_group(br, port, &br_group);
 793 }
 794 #endif
 795
 796 static void br_multicast_router_expired(unsigned long data)
 797 {
 798         struct net_bridge_port *port = (void *)data;
 799         struct net_bridge *br = port->br;
 800
 801         spin_lock(&br->multicast_lock);
 802         if (port->multicast_router != 1 ||
 803             timer_pending(&port->multicast_router_timer) ||
 804             hlist_unhashed(&port->rlist))
 805                 goto out;
 806
 807         hlist_del_init_rcu(&port->rlist);
 808
 809 out:
 810         spin_unlock(&br->multicast_lock);
 811 }
 812
 813 static void br_multicast_local_router_expired(unsigned long data)
 814 {
 815 }
 816
 817 static void __br_multicast_send_query(struct net_bridge *br,
 818                                       struct net_bridge_port *port,
 819                                       struct br_ip *ip)
 820 {
 821         struct sk_buff *skb;
 822
 823         skb = br_multicast_alloc_query(br, ip);
 824         if (!skb)
 825                 return;
 826
 827         if (port) {
 828                 __skb_push(skb, sizeof(struct ethhdr));
 829                 skb->dev = port->dev;
 830                 NF_HOOK(NFPROTO_BRIDGE, NF_BR_LOCAL_OUT, skb, NULL, skb->dev,
 831                         dev_queue_xmit);
 832         } else
 833                 netif_rx(skb);
 834 }
 835
 836 static void br_multicast_send_query(struct net_bridge *br,
 837                                     struct net_bridge_port *port, u32 sent)
 838 {
 839         unsigned long time;
 840         struct br_ip br_group;
 841
 842         if (!netif_running(br->dev) || br->multicast_disabled ||
 843             timer_pending(&br->multicast_querier_timer))
 844                 return;
 845
 846         memset(&br_group.u, 0, sizeof(br_group.u));
 847
 848         br_group.proto = htons(ETH_P_IP);
 849         __br_multicast_send_query(br, port, &br_group);
 850
 851 #if IS_ENABLED(CONFIG_IPV6)
 852         br_group.proto = htons(ETH_P_IPV6);
 853         __br_multicast_send_query(br, port, &br_group);
 854 #endif
 855
 856         time = jiffies;
 857         time += sent < br->multicast_startup_query_count ?
 858                 br->multicast_startup_query_interval :
 859                 br->multicast_query_interval;
 860         mod_timer(port ? &port->multicast_query_timer :
 861                          &br->multicast_query_timer, time);
 862 }
 863
 864 static void br_multicast_port_query_expired(unsigned long data)
 865 {
 866         struct net_bridge_port *port = (void *)data;
 867         struct net_bridge *br = port->br;
 868
 869         spin_lock(&br->multicast_lock);
 870         if (port->state == BR_STATE_DISABLED ||
 871             port->state == BR_STATE_BLOCKING)
 872                 goto out;
 873
 874         if (port->multicast_startup_queries_sent <
 875             br->multicast_startup_query_count)
 876                 port->multicast_startup_queries_sent++;
 877
 878         br_multicast_send_query(port->br, port,
 879                                 port->multicast_startup_queries_sent);
 880
 881 out:
 882         spin_unlock(&br->multicast_lock);
 883 }
 884
 885 void br_multicast_add_port(struct net_bridge_port *port)
 886 {
 887         port->multicast_router = 1;
 888
 889         setup_timer(&port->multicast_router_timer, br_multicast_router_expired,
 890                     (unsigned long)port);
 891         setup_timer(&port->multicast_query_timer,
 892                     br_multicast_port_query_expired, (unsigned long)port);
 893 }
 894
 895 void br_multicast_del_port(struct net_bridge_port *port)
 896 {
 897         del_timer_sync(&port->multicast_router_timer);
 898 }
 899
 900 static void __br_multicast_enable_port(struct net_bridge_port *port)
 901 {
 902         port->multicast_startup_queries_sent = 0;
 903
 904         if (try_to_del_timer_sync(&port->multicast_query_timer) >= 0 ||
 905             del_timer(&port->multicast_query_timer))
 906                 mod_timer(&port->multicast_query_timer, jiffies);
 907 }
 908
 909 void br_multicast_enable_port(struct net_bridge_port *port)
 910 {
 911         struct net_bridge *br = port->br;
 912
 913         spin_lock(&br->multicast_lock);
 914         if (br->multicast_disabled || !netif_running(br->dev))
 915                 goto out;
 916
 917         __br_multicast_enable_port(port);
 918
 919 out:
 920         spin_unlock(&br->multicast_lock);
 921 }
 922
 923 void br_multicast_disable_port(struct net_bridge_port *port)
 924 {
 925         struct net_bridge *br = port->br;
 926         struct net_bridge_port_group *pg;
 927         struct hlist_node *p, *n;
 928
 929         spin_lock(&br->multicast_lock);
 930         hlist_for_each_entry_safe(pg, p, n, &port->mglist, mglist)
 931                 br_multicast_del_pg(br, pg);
 932
 933         if (!hlist_unhashed(&port->rlist))
 934                 hlist_del_init_rcu(&port->rlist);
 935         del_timer(&port->multicast_router_timer);
 936         del_timer(&port->multicast_query_timer);
 937         spin_unlock(&br->multicast_lock);
 938 }
 939
 940 static int br_ip4_multicast_igmp3_report(struct net_bridge *br,
 941                                          struct net_bridge_port *port,
 942                                          struct sk_buff *skb)
 943 {
 944         struct igmpv3_report *ih;
 945         struct igmpv3_grec *grec;
 946         int i;
 947         int len;
 948         int num;
 949         int type;
 950         int err = 0;
 951         __be32 group;
 952
 953         if (!pskb_may_pull(skb, sizeof(*ih)))
 954                 return -EINVAL;
 955
 956         ih = igmpv3_report_hdr(skb);
 957         num = ntohs(ih->ngrec);
 958         len = sizeof(*ih);
 959
 960         for (i = 0; i < num; i++) {
 961                 len += sizeof(*grec);
 962                 if (!pskb_may_pull(skb, len))
 963                         return -EINVAL;
 964
 965                 grec = (void *)(skb->data + len - sizeof(*grec));
 966                 group = grec->grec_mca;
 967                 type = grec->grec_type;
 968
 969                 len += ntohs(grec->grec_nsrcs) * 4;
 970                 if (!pskb_may_pull(skb, len))
 971                         return -EINVAL;
 972
 973                 /* We treat this as an IGMPv2 report for now. */
 974                 switch (type) {
 975                 case IGMPV3_MODE_IS_INCLUDE:
 976                 case IGMPV3_MODE_IS_EXCLUDE:
 977                 case IGMPV3_CHANGE_TO_INCLUDE:
 978                 case IGMPV3_CHANGE_TO_EXCLUDE:
 979                 case IGMPV3_ALLOW_NEW_SOURCES:
 980                 case IGMPV3_BLOCK_OLD_SOURCES:
 981                         break;
 982
 983                 default:
 984                         continue;
 985                 }
 986
 987                 err = br_ip4_multicast_add_group(br, port, group);
 988                 if (err)
 989                         break;
 990         }
 991
 992         return err;
 993 }
 994
 995 #if IS_ENABLED(CONFIG_IPV6)
 996 static int br_ip6_multicast_mld2_report(struct net_bridge *br,
 997                                         struct net_bridge_port *port,
 998                                         struct sk_buff *skb)
 999 {
1000         struct icmp6hdr *icmp6h;
1001         struct mld2_grec *grec;
1002         int i;
1003         int len;
1004         int num;
1005         int err = 0;
1006
1007         if (!pskb_may_pull(skb, sizeof(*icmp6h)))
1008                 return -EINVAL;
1009
1010         icmp6h = icmp6_hdr(skb);
1011         num = ntohs(icmp6h->icmp6_dataun.un_data16[1]);
1012         len = sizeof(*icmp6h);
1013
1014         for (i = 0; i < num; i++) {
1015                 __be16 *nsrcs, _nsrcs;
1016
1017                 nsrcs = skb_header_pointer(skb,
1018                                            len + offsetof(struct mld2_grec,
1019                                                           grec_nsrcs),
1020                                            sizeof(_nsrcs), &_nsrcs);
1021                 if (!nsrcs)
1022                         return -EINVAL;
1023
1024                 if (!pskb_may_pull(skb,
1025                                    len + sizeof(*grec) +
1026                                    sizeof(struct in6_addr) * ntohs(*nsrcs)))
1027                         return -EINVAL;
1028
1029                 grec = (struct mld2_grec *)(skb->data + len);
1030                 len += sizeof(*grec) +
1031                        sizeof(struct in6_addr) * ntohs(*nsrcs);
1032
1033                 /* We treat these as MLDv1 reports for now. */
1034                 switch (grec->grec_type) {
1035                 case MLD2_MODE_IS_INCLUDE:
1036                 case MLD2_MODE_IS_EXCLUDE:
1037                 case MLD2_CHANGE_TO_INCLUDE:
1038                 case MLD2_CHANGE_TO_EXCLUDE:
1039                 case MLD2_ALLOW_NEW_SOURCES:
1040                 case MLD2_BLOCK_OLD_SOURCES:
1041                         break;
1042
1043                 default:
1044                         continue;
1045                 }
1046
1047                 err = br_ip6_multicast_add_group(br, port, &grec->grec_mca);
1048                 if (!err)
1049                         break;
1050         }
1051
1052         return err;
1053 }
1054 #endif
1055
1056 /*
1057  * Add port to rotuer_list
1058  *  list is maintained ordered by pointer value
1059  *  and locked by br->multicast_lock and RCU
1060  */
1061 static void br_multicast_add_router(struct net_bridge *br,
1062                                     struct net_bridge_port *port)
1063 {
1064         struct net_bridge_port *p;
1065         struct hlist_node *n, *slot = NULL;
1066
1067         hlist_for_each_entry(p, n, &br->router_list, rlist) {
1068                 if ((unsigned long) port >= (unsigned long) p)
1069                         break;
1070                 slot = n;
1071         }
1072
1073         if (slot)
1074                 hlist_add_after_rcu(slot, &port->rlist);
1075         else
1076                 hlist_add_head_rcu(&port->rlist, &br->router_list);
1077 }
1078
1079 static void br_multicast_mark_router(struct net_bridge *br,
1080                                      struct net_bridge_port *port)
1081 {
1082         unsigned long now = jiffies;
1083
1084         if (!port) {
1085                 if (br->multicast_router == 1)
1086                         mod_timer(&br->multicast_router_timer,
1087                                   now + br->multicast_querier_interval);
1088                 return;
1089         }
1090
1091         if (port->multicast_router != 1)
1092                 return;
1093
1094         if (!hlist_unhashed(&port->rlist))
1095                 goto timer;
1096
1097         br_multicast_add_router(br, port);
1098
1099 timer:
1100         mod_timer(&port->multicast_router_timer,
1101                   now + br->multicast_querier_interval);
1102 }
1103
1104 static void br_multicast_query_received(struct net_bridge *br,
1105                                         struct net_bridge_port *port,
1106                                         int saddr)
1107 {
1108         if (saddr)
1109                 mod_timer(&br->multicast_querier_timer,
1110                           jiffies + br->multicast_querier_interval);
1111         else if (timer_pending(&br->multicast_querier_timer))
1112                 return;
1113
1114         br_multicast_mark_router(br, port);
1115 }
1116
1117 static int br_ip4_multicast_query(struct net_bridge *br,
1118                                   struct net_bridge_port *port,
1119                                   struct sk_buff *skb)
1120 {
1121         const struct iphdr *iph = ip_hdr(skb);
1122         struct igmphdr *ih = igmp_hdr(skb);
1123         struct net_bridge_mdb_entry *mp;
1124         struct igmpv3_query *ih3;
1125         struct net_bridge_port_group *p;
1126         struct net_bridge_port_group __rcu **pp;
1127         unsigned long max_delay;
1128         unsigned long now = jiffies;
1129         __be32 group;
1130         int err = 0;
1131
1132         spin_lock(&br->multicast_lock);
1133         if (!netif_running(br->dev) ||
1134             (port && port->state == BR_STATE_DISABLED))
1135                 goto out;
1136
1137         br_multicast_query_received(br, port, !!iph->saddr);
1138
1139         group = ih->group;
1140
1141         if (skb->len == sizeof(*ih)) {
1142                 max_delay = ih->code * (HZ / IGMP_TIMER_SCALE);
1143
1144                 if (!max_delay) {
1145                         max_delay = 10 * HZ;
1146                         group = 0;
1147                 }
1148         } else {
1149                 if (!pskb_may_pull(skb, sizeof(struct igmpv3_query))) {
1150                         err = -EINVAL;
1151                         goto out;
1152                 }
1153
1154                 ih3 = igmpv3_query_hdr(skb);
1155                 if (ih3->nsrcs)
1156                         goto out;
1157
1158                 max_delay = ih3->code ?
1159                             IGMPV3_MRC(ih3->code) * (HZ / IGMP_TIMER_SCALE) : 1;
1160         }
1161
1162         if (!group)
1163                 goto out;
1164
1165         mp = br_mdb_ip4_get(mlock_dereference(br->mdb, br), group);
1166         if (!mp)
1167                 goto out;
1168
1169         max_delay *= br->multicast_last_member_count;
1170
1171         if (mp->mglist &&
1172             (timer_pending(&mp->timer) ?
1173              time_after(mp->timer.expires, now + max_delay) :
1174              try_to_del_timer_sync(&mp->timer) >= 0))
1175                 mod_timer(&mp->timer, now + max_delay);
1176
1177         for (pp = &mp->ports;
1178              (p = mlock_dereference(*pp, br)) != NULL;
1179              pp = &p->next) {
1180                 if (timer_pending(&p->timer) ?
1181                     time_after(p->timer.expires, now + max_delay) :
1182                     try_to_del_timer_sync(&p->timer) >= 0)
1183                         mod_timer(&p->timer, now + max_delay);
1184         }
1185
1186 out:
1187         spin_unlock(&br->multicast_lock);
1188         return err;
1189 }
1190
1191 #if IS_ENABLED(CONFIG_IPV6)
1192 static int br_ip6_multicast_query(struct net_bridge *br,
1193                                   struct net_bridge_port *port,
1194                                   struct sk_buff *skb)
1195 {
1196         const struct ipv6hdr *ip6h = ipv6_hdr(skb);
1197         struct mld_msg *mld = (struct mld_msg *) icmp6_hdr(skb);
1198         struct net_bridge_mdb_entry *mp;
1199         struct mld2_query *mld2q;
1200         struct net_bridge_port_group *p;
1201         struct net_bridge_port_group __rcu **pp;
1202         unsigned long max_delay;
1203         unsigned long now = jiffies;
1204         const struct in6_addr *group = NULL;
1205         int err = 0;
1206
1207         spin_lock(&br->multicast_lock);
1208         if (!netif_running(br->dev) ||
1209             (port && port->state == BR_STATE_DISABLED))
1210                 goto out;
1211
1212         br_multicast_query_received(br, port, !ipv6_addr_any(&ip6h->saddr));
1213
1214         if (skb->len == sizeof(*mld)) {
1215                 if (!pskb_may_pull(skb, sizeof(*mld))) {
1216                         err = -EINVAL;
1217                         goto out;
1218                 }
1219                 mld = (struct mld_msg *) icmp6_hdr(skb);
1220                 max_delay = msecs_to_jiffies(htons(mld->mld_maxdelay));
1221                 if (max_delay)
1222                         group = &mld->mld_mca;
1223         } else if (skb->len >= sizeof(*mld2q)) {
1224                 if (!pskb_may_pull(skb, sizeof(*mld2q))) {
1225                         err = -EINVAL;
1226                         goto out;
1227                 }
1228                 mld2q = (struct mld2_query *)icmp6_hdr(skb);
1229                 if (!mld2q->mld2q_nsrcs)
1230                         group = &mld2q->mld2q_mca;
1231                 max_delay = mld2q->mld2q_mrc ? MLDV2_MRC(mld2q->mld2q_mrc) : 1;
1232         }
1233
1234         if (!group)
1235                 goto out;
1236
1237         mp = br_mdb_ip6_get(mlock_dereference(br->mdb, br), group);
1238         if (!mp)
1239                 goto out;
1240
1241         max_delay *= br->multicast_last_member_count;
1242         if (mp->mglist &&
1243             (timer_pending(&mp->timer) ?
1244              time_after(mp->timer.expires, now + max_delay) :
1245              try_to_del_timer_sync(&mp->timer) >= 0))
1246                 mod_timer(&mp->timer, now + max_delay);
1247
1248         for (pp = &mp->ports;
1249              (p = mlock_dereference(*pp, br)) != NULL;
1250              pp = &p->next) {
1251                 if (timer_pending(&p->timer) ?
1252                     time_after(p->timer.expires, now + max_delay) :
1253                     try_to_del_timer_sync(&p->timer) >= 0)
1254                         mod_timer(&p->timer, now + max_delay);
1255         }
1256
1257 out:
1258         spin_unlock(&br->multicast_lock);
1259         return err;
1260 }
1261 #endif
1262
1263 static void br_multicast_leave_group(struct net_bridge *br,
1264                                      struct net_bridge_port *port,
1265                                      struct br_ip *group)
1266 {
1267         struct net_bridge_mdb_htable *mdb;
1268         struct net_bridge_mdb_entry *mp;
1269         struct net_bridge_port_group *p;
1270         unsigned long now;
1271         unsigned long time;
1272
1273         spin_lock(&br->multicast_lock);
1274         if (!netif_running(br->dev) ||
1275             (port && port->state == BR_STATE_DISABLED) ||
1276             timer_pending(&br->multicast_querier_timer))
1277                 goto out;
1278
1279         mdb = mlock_dereference(br->mdb, br);
1280         mp = br_mdb_ip_get(mdb, group);
1281         if (!mp)
1282                 goto out;
1283
1284         now = jiffies;
1285         time = now + br->multicast_last_member_count *
1286                      br->multicast_last_member_interval;
1287
1288         if (!port) {
1289                 if (mp->mglist &&
1290                     (timer_pending(&mp->timer) ?
1291                      time_after(mp->timer.expires, time) :
1292                      try_to_del_timer_sync(&mp->timer) >= 0)) {
1293                         mod_timer(&mp->timer, time);
1294
1295                         mp->queries_sent = 0;
1296                         mod_timer(&mp->query_timer, now);
1297                 }
1298
1299                 goto out;
1300         }
1301
1302         for (p = mlock_dereference(mp->ports, br);
1303              p != NULL;
1304              p = mlock_dereference(p->next, br)) {
1305                 if (p->port != port)
1306                         continue;
1307
1308                 if (!hlist_unhashed(&p->mglist) &&
1309                     (timer_pending(&p->timer) ?
1310                      time_after(p->timer.expires, time) :
1311                      try_to_del_timer_sync(&p->timer) >= 0)) {
1312                         mod_timer(&p->timer, time);
1313
1314                         p->queries_sent = 0;
1315                         mod_timer(&p->query_timer, now);
1316                 }
1317
1318                 break;
1319         }
1320
1321 out:
1322         spin_unlock(&br->multicast_lock);
1323 }
1324
1325 static void br_ip4_multicast_leave_group(struct net_bridge *br,
1326                                          struct net_bridge_port *port,
1327                                          __be32 group)
1328 {
1329         struct br_ip br_group;
1330
1331         if (ipv4_is_local_multicast(group))
1332                 return;
1333
1334         br_group.u.ip4 = group;
1335         br_group.proto = htons(ETH_P_IP);
1336
1337         br_multicast_leave_group(br, port, &br_group);
1338 }
1339
1340 #if IS_ENABLED(CONFIG_IPV6)
1341 static void br_ip6_multicast_leave_group(struct net_bridge *br,
1342                                          struct net_bridge_port *port,
1343                                          const struct in6_addr *group)
1344 {
1345         struct br_ip br_group;
1346
1347         if (!ipv6_is_transient_multicast(group))
1348                 return;
1349
1350         br_group.u.ip6 = *group;
1351         br_group.proto = htons(ETH_P_IPV6);
1352
1353         br_multicast_leave_group(br, port, &br_group);
1354 }
1355 #endif
1356
1357 static int br_multicast_ipv4_rcv(struct net_bridge *br,
1358                                  struct net_bridge_port *port,
1359                                  struct sk_buff *skb)
1360 {
1361         struct sk_buff *skb2 = skb;
1362         const struct iphdr *iph;
1363         struct igmphdr *ih;
1364         unsigned len;
1365         unsigned offset;
1366         int err;
1367
1368         /* We treat OOM as packet loss for now. */
1369         if (!pskb_may_pull(skb, sizeof(*iph)))
1370                 return -EINVAL;
1371
1372         iph = ip_hdr(skb);
1373
1374         if (iph->ihl < 5 || iph->version != 4)
1375                 return -EINVAL;
1376
1377         if (!pskb_may_pull(skb, ip_hdrlen(skb)))
1378                 return -EINVAL;
1379
1380         iph = ip_hdr(skb);
1381
1382         if (unlikely(ip_fast_csum((u8 *)iph, iph->ihl)))
1383                 return -EINVAL;
1384
1385         if (iph->protocol != IPPROTO_IGMP) {
1386                 if ((iph->daddr & IGMP_LOCAL_GROUP_MASK) != IGMP_LOCAL_GROUP)
1387                         BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1388                 return 0;
1389         }
1390
1391         len = ntohs(iph->tot_len);
1392         if (skb->len < len || len < ip_hdrlen(skb))
1393                 return -EINVAL;
1394
1395         if (skb->len > len) {
1396                 skb2 = skb_clone(skb, GFP_ATOMIC);
1397                 if (!skb2)
1398                         return -ENOMEM;
1399
1400                 err = pskb_trim_rcsum(skb2, len);
1401                 if (err)
1402                         goto err_out;
1403         }
1404
1405         len -= ip_hdrlen(skb2);
1406         offset = skb_network_offset(skb2) + ip_hdrlen(skb2);
1407         __skb_pull(skb2, offset);
1408         skb_reset_transport_header(skb2);
1409
1410         err = -EINVAL;
1411         if (!pskb_may_pull(skb2, sizeof(*ih)))
1412                 goto out;
1413
1414         switch (skb2->ip_summed) {
1415         case CHECKSUM_COMPLETE:
1416                 if (!csum_fold(skb2->csum))
1417                         break;
1418                 /* fall through */
1419         case CHECKSUM_NONE:
1420                 skb2->csum = 0;
1421                 if (skb_checksum_complete(skb2))
1422                         goto out;
1423         }
1424
1425         err = 0;
1426
1427         BR_INPUT_SKB_CB(skb)->igmp = 1;
1428         ih = igmp_hdr(skb2);
1429
1430         switch (ih->type) {
1431         case IGMP_HOST_MEMBERSHIP_REPORT:
1432         case IGMPV2_HOST_MEMBERSHIP_REPORT:
1433                 BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1434                 err = br_ip4_multicast_add_group(br, port, ih->group);
1435                 break;
1436         case IGMPV3_HOST_MEMBERSHIP_REPORT:
1437                 err = br_ip4_multicast_igmp3_report(br, port, skb2);
1438                 break;
1439         case IGMP_HOST_MEMBERSHIP_QUERY:
1440                 err = br_ip4_multicast_query(br, port, skb2);
1441                 break;
1442         case IGMP_HOST_LEAVE_MESSAGE:
1443                 br_ip4_multicast_leave_group(br, port, ih->group);
1444                 break;
1445         }
1446
1447 out:
1448         __skb_push(skb2, offset);
1449 err_out:
1450         if (skb2 != skb)
1451                 kfree_skb(skb2);
1452         return err;
1453 }
1454
1455 #if IS_ENABLED(CONFIG_IPV6)
1456 static int br_multicast_ipv6_rcv(struct net_bridge *br,
1457                                  struct net_bridge_port *port,
1458                                  struct sk_buff *skb)
1459 {
1460         struct sk_buff *skb2;
1461         const struct ipv6hdr *ip6h;
1462         u8 icmp6_type;
1463         u8 nexthdr;
1464         __be16 frag_off;
1465         unsigned len;
1466         int offset;
1467         int err;
1468
1469         if (!pskb_may_pull(skb, sizeof(*ip6h)))
1470                 return -EINVAL;
1471
1472         ip6h = ipv6_hdr(skb);
1473
1474         /*
1475          * We're interested in MLD messages only.
1476          *  - Version is 6
1477          *  - MLD has always Router Alert hop-by-hop option
1478          *  - But we do not support jumbrograms.
1479          */
1480         if (ip6h->version != 6 ||
1481             ip6h->nexthdr != IPPROTO_HOPOPTS ||
1482             ip6h->payload_len == 0)
1483                 return 0;
1484
1485         len = ntohs(ip6h->payload_len) + sizeof(*ip6h);
1486         if (skb->len < len)
1487                 return -EINVAL;
1488
1489         nexthdr = ip6h->nexthdr;
1490         offset = ipv6_skip_exthdr(skb, sizeof(*ip6h), &nexthdr, &frag_off);
1491
1492         if (offset < 0 || nexthdr != IPPROTO_ICMPV6)
1493                 return 0;
1494
1495         /* Okay, we found ICMPv6 header */
1496         skb2 = skb_clone(skb, GFP_ATOMIC);
1497         if (!skb2)
1498                 return -ENOMEM;
1499
1500         err = -EINVAL;
1501         if (!pskb_may_pull(skb2, offset + sizeof(struct icmp6hdr)))
1502                 goto out;
1503
1504         len -= offset - skb_network_offset(skb2);
1505
1506         __skb_pull(skb2, offset);
1507         skb_reset_transport_header(skb2);
1508         skb_postpull_rcsum(skb2, skb_network_header(skb2),
1509                            skb_network_header_len(skb2));
1510
1511         icmp6_type = icmp6_hdr(skb2)->icmp6_type;
1512
1513         switch (icmp6_type) {
1514         case ICMPV6_MGM_QUERY:
1515         case ICMPV6_MGM_REPORT:
1516         case ICMPV6_MGM_REDUCTION:
1517         case ICMPV6_MLD2_REPORT:
1518                 break;
1519         default:
1520                 err = 0;
1521                 goto out;
1522         }
1523
1524         /* Okay, we found MLD message. Check further. */
1525         if (skb2->len > len) {
1526                 err = pskb_trim_rcsum(skb2, len);
1527                 if (err)
1528                         goto out;
1529                 err = -EINVAL;
1530         }
1531
1532         ip6h = ipv6_hdr(skb2);
1533
1534         switch (skb2->ip_summed) {
1535         case CHECKSUM_COMPLETE:
1536                 if (!csum_ipv6_magic(&ip6h->saddr, &ip6h->daddr, skb2->len,
1537                                         IPPROTO_ICMPV6, skb2->csum))
1538                         break;
1539                 /*FALLTHROUGH*/
1540         case CHECKSUM_NONE:
1541                 skb2->csum = ~csum_unfold(csum_ipv6_magic(&ip6h->saddr,
1542                                                         &ip6h->daddr,
1543                                                         skb2->len,
1544                                                         IPPROTO_ICMPV6, 0));
1545                 if (__skb_checksum_complete(skb2))
1546                         goto out;
1547         }
1548
1549         err = 0;
1550
1551         BR_INPUT_SKB_CB(skb)->igmp = 1;
1552
1553         switch (icmp6_type) {
1554         case ICMPV6_MGM_REPORT:
1555             {
1556                 struct mld_msg *mld;
1557                 if (!pskb_may_pull(skb2, sizeof(*mld))) {
1558                         err = -EINVAL;
1559                         goto out;
1560                 }
1561                 mld = (struct mld_msg *)skb_transport_header(skb2);
1562                 BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1563                 err = br_ip6_multicast_add_group(br, port, &mld->mld_mca);
1564                 break;
1565             }
1566         case ICMPV6_MLD2_REPORT:
1567                 err = br_ip6_multicast_mld2_report(br, port, skb2);
1568                 break;
1569         case ICMPV6_MGM_QUERY:
1570                 err = br_ip6_multicast_query(br, port, skb2);
1571                 break;
1572         case ICMPV6_MGM_REDUCTION:
1573             {
1574                 struct mld_msg *mld;
1575                 if (!pskb_may_pull(skb2, sizeof(*mld))) {
1576                         err = -EINVAL;
1577                         goto out;
1578                 }
1579                 mld = (struct mld_msg *)skb_transport_header(skb2);
1580                 br_ip6_multicast_leave_group(br, port, &mld->mld_mca);
1581             }
1582         }
1583
1584 out:
1585         kfree_skb(skb2);
1586         return err;
1587 }
1588 #endif
1589
1590 int br_multicast_rcv(struct net_bridge *br, struct net_bridge_port *port,
1591                      struct sk_buff *skb)
1592 {
1593         BR_INPUT_SKB_CB(skb)->igmp = 0;
1594         BR_INPUT_SKB_CB(skb)->mrouters_only = 0;
1595
1596         if (br->multicast_disabled)
1597                 return 0;
1598
1599         switch (skb->protocol) {
1600         case htons(ETH_P_IP):
1601                 return br_multicast_ipv4_rcv(br, port, skb);
1602 #if IS_ENABLED(CONFIG_IPV6)
1603         case htons(ETH_P_IPV6):
1604                 return br_multicast_ipv6_rcv(br, port, skb);
1605 #endif
1606         }
1607
1608         return 0;
1609 }
1610
1611 static void br_multicast_query_expired(unsigned long data)
1612 {
1613         struct net_bridge *br = (void *)data;
1614
1615         spin_lock(&br->multicast_lock);
1616         if (br->multicast_startup_queries_sent <
1617             br->multicast_startup_query_count)
1618                 br->multicast_startup_queries_sent++;
1619
1620         br_multicast_send_query(br, NULL, br->multicast_startup_queries_sent);
1621
1622         spin_unlock(&br->multicast_lock);
1623 }
1624
1625 void br_multicast_init(struct net_bridge *br)
1626 {
1627         br->hash_elasticity = 4;
1628         br->hash_max = 512;
1629
1630         br->multicast_router = 1;
1631         br->multicast_last_member_count = 2;
1632         br->multicast_startup_query_count = 2;
1633
1634         br->multicast_last_member_interval = HZ;
1635         br->multicast_query_response_interval = 10 * HZ;
1636         br->multicast_startup_query_interval = 125 * HZ / 4;
1637         br->multicast_query_interval = 125 * HZ;
1638         br->multicast_querier_interval = 255 * HZ;
1639         br->multicast_membership_interval = 260 * HZ;
1640
1641         spin_lock_init(&br->multicast_lock);
1642         setup_timer(&br->multicast_router_timer,
1643                     br_multicast_local_router_expired, 0);
1644         setup_timer(&br->multicast_querier_timer,
1645                     br_multicast_local_router_expired, 0);
1646         setup_timer(&br->multicast_query_timer, br_multicast_query_expired,
1647                     (unsigned long)br);
1648 }
1649
1650 void br_multicast_open(struct net_bridge *br)
1651 {
1652         br->multicast_startup_queries_sent = 0;
1653
1654         if (br->multicast_disabled)
1655                 return;
1656
1657         mod_timer(&br->multicast_query_timer, jiffies);
1658 }
1659
1660 void br_multicast_stop(struct net_bridge *br)
1661 {
1662         struct net_bridge_mdb_htable *mdb;
1663         struct net_bridge_mdb_entry *mp;
1664         struct hlist_node *p, *n;
1665         u32 ver;
1666         int i;
1667
1668         del_timer_sync(&br->multicast_router_timer);
1669         del_timer_sync(&br->multicast_querier_timer);
1670         del_timer_sync(&br->multicast_query_timer);
1671
1672         spin_lock_bh(&br->multicast_lock);
1673         mdb = mlock_dereference(br->mdb, br);
1674         if (!mdb)
1675                 goto out;
1676
1677         br->mdb = NULL;
1678
1679         ver = mdb->ver;
1680         for (i = 0; i < mdb->max; i++) {
1681                 hlist_for_each_entry_safe(mp, p, n, &mdb->mhash[i],
1682                                           hlist[ver]) {
1683                         del_timer(&mp->timer);
1684                         del_timer(&mp->query_timer);
1685                         call_rcu_bh(&mp->rcu, br_multicast_free_group);
1686                 }
1687         }
1688
1689         if (mdb->old) {
1690                 spin_unlock_bh(&br->multicast_lock);
1691                 rcu_barrier_bh();
1692                 spin_lock_bh(&br->multicast_lock);
1693                 WARN_ON(mdb->old);
1694         }
1695
1696         mdb->old = mdb;
1697         call_rcu_bh(&mdb->rcu, br_mdb_free);
1698
1699 out:
1700         spin_unlock_bh(&br->multicast_lock);
1701 }
1702
1703 int br_multicast_set_router(struct net_bridge *br, unsigned long val)
1704 {
1705         int err = -ENOENT;
1706
1707         spin_lock_bh(&br->multicast_lock);
1708         if (!netif_running(br->dev))
1709                 goto unlock;
1710
1711         switch (val) {
1712         case 0:
1713         case 2:
1714                 del_timer(&br->multicast_router_timer);
1715                 /* fall through */
1716         case 1:
1717                 br->multicast_router = val;
1718                 err = 0;
1719                 break;
1720
1721         default:
1722                 err = -EINVAL;
1723                 break;
1724         }
1725
1726 unlock:
1727         spin_unlock_bh(&br->multicast_lock);
1728
1729         return err;
1730 }
1731
1732 int br_multicast_set_port_router(struct net_bridge_port *p, unsigned long val)
1733 {
1734         struct net_bridge *br = p->br;
1735         int err = -ENOENT;
1736
1737         spin_lock(&br->multicast_lock);
1738         if (!netif_running(br->dev) || p->state == BR_STATE_DISABLED)
1739                 goto unlock;
1740
1741         switch (val) {
1742         case 0:
1743         case 1:
1744         case 2:
1745                 p->multicast_router = val;
1746                 err = 0;
1747
1748                 if (val < 2 && !hlist_unhashed(&p->rlist))
1749                         hlist_del_init_rcu(&p->rlist);
1750
1751                 if (val == 1)
1752                         break;
1753
1754                 del_timer(&p->multicast_router_timer);
1755
1756                 if (val == 0)
1757                         break;
1758
1759                 br_multicast_add_router(br, p);
1760                 break;
1761
1762         default:
1763                 err = -EINVAL;
1764                 break;
1765         }
1766
1767 unlock:
1768         spin_unlock(&br->multicast_lock);
1769
1770         return err;
1771 }
1772
1773 int br_multicast_toggle(struct net_bridge *br, unsigned long val)
1774 {
1775         struct net_bridge_port *port;
1776         int err = 0;
1777         struct net_bridge_mdb_htable *mdb;
1778
1779         spin_lock_bh(&br->multicast_lock);
1780         if (br->multicast_disabled == !val)
1781                 goto unlock;
1782
1783         br->multicast_disabled = !val;
1784         if (br->multicast_disabled)
1785                 goto unlock;
1786
1787         if (!netif_running(br->dev))
1788                 goto unlock;
1789
1790         mdb = mlock_dereference(br->mdb, br);
1791         if (mdb) {
1792                 if (mdb->old) {
1793                         err = -EEXIST;
1794 rollback:
1795                         br->multicast_disabled = !!val;
1796                         goto unlock;
1797                 }
1798
1799                 err = br_mdb_rehash(&br->mdb, mdb->max,
1800                                     br->hash_elasticity);
1801                 if (err)
1802                         goto rollback;
1803         }
1804
1805         br_multicast_open(br);
1806         list_for_each_entry(port, &br->port_list, list) {
1807                 if (port->state == BR_STATE_DISABLED ||
1808                     port->state == BR_STATE_BLOCKING)
1809                         continue;
1810
1811                 __br_multicast_enable_port(port);
1812         }
1813
1814 unlock:
1815         spin_unlock_bh(&br->multicast_lock);
1816
1817         return err;
1818 }
1819
1820 int br_multicast_set_hash_max(struct net_bridge *br, unsigned long val)
1821 {
1822         int err = -ENOENT;
1823         u32 old;
1824         struct net_bridge_mdb_htable *mdb;
1825
1826         spin_lock(&br->multicast_lock);
1827         if (!netif_running(br->dev))
1828                 goto unlock;
1829
1830         err = -EINVAL;
1831         if (!is_power_of_2(val))
1832                 goto unlock;
1833
1834         mdb = mlock_dereference(br->mdb, br);
1835         if (mdb && val < mdb->size)
1836                 goto unlock;
1837
1838         err = 0;
1839
1840         old = br->hash_max;
1841         br->hash_max = val;
1842
1843         if (mdb) {
1844                 if (mdb->old) {
1845                         err = -EEXIST;
1846 rollback:
1847                         br->hash_max = old;
1848                         goto unlock;
1849                 }
1850
1851                 err = br_mdb_rehash(&br->mdb, br->hash_max,
1852                                     br->hash_elasticity);
1853                 if (err)
1854                         goto rollback;
1855         }
1856
1857 unlock:
1858         spin_unlock(&br->multicast_lock);
1859
1860         return err;
1861 }