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 <linux/inetdevice.h>
  27 #include <net/ip.h>
  28 #if IS_ENABLED(CONFIG_IPV6)
  29 #include <net/ipv6.h>
  30 #include <net/mld.h>
  31 #include <net/ip6_checksum.h>
  32 #include <net/addrconf.h>
  33 #endif
  34
  35 #include "br_private.h"
  36
  37 static void br_multicast_start_querier(struct net_bridge *br);
  38 unsigned int br_mdb_rehash_seq;
  39
  40 static inline int br_ip_equal(const struct br_ip *a, const struct br_ip *b)
  41 {
  42         if (a->proto != b->proto)
  43                 return 0;
  44         if (a->vid != b->vid)
  45                 return 0;
  46         switch (a->proto) {
  47         case htons(ETH_P_IP):
  48                 return a->u.ip4 == b->u.ip4;
  49 #if IS_ENABLED(CONFIG_IPV6)
  50         case htons(ETH_P_IPV6):
  51                 return ipv6_addr_equal(&a->u.ip6, &b->u.ip6);
  52 #endif
  53         }
  54         return 0;
  55 }
  56
  57 static inline int __br_ip4_hash(struct net_bridge_mdb_htable *mdb, __be32 ip,
  58                                 __u16 vid)
  59 {
  60         return jhash_2words((__force u32)ip, vid, mdb->secret) & (mdb->max - 1);
  61 }
  62
  63 #if IS_ENABLED(CONFIG_IPV6)
  64 static inline int __br_ip6_hash(struct net_bridge_mdb_htable *mdb,
  65                                 const struct in6_addr *ip,
  66                                 __u16 vid)
  67 {
  68         return jhash_2words(ipv6_addr_hash(ip), vid,
  69                             mdb->secret) & (mdb->max - 1);
  70 }
  71 #endif
  72
  73 static inline int br_ip_hash(struct net_bridge_mdb_htable *mdb,
  74                              struct br_ip *ip)
  75 {
  76         switch (ip->proto) {
  77         case htons(ETH_P_IP):
  78                 return __br_ip4_hash(mdb, ip->u.ip4, ip->vid);
  79 #if IS_ENABLED(CONFIG_IPV6)
  80         case htons(ETH_P_IPV6):
  81                 return __br_ip6_hash(mdb, &ip->u.ip6, ip->vid);
  82 #endif
  83         }
  84         return 0;
  85 }
  86
  87 static struct net_bridge_mdb_entry *__br_mdb_ip_get(
  88         struct net_bridge_mdb_htable *mdb, struct br_ip *dst, int hash)
  89 {
  90         struct net_bridge_mdb_entry *mp;
  91
  92         hlist_for_each_entry_rcu(mp, &mdb->mhash[hash], hlist[mdb->ver]) {
  93                 if (br_ip_equal(&mp->addr, dst))
  94                         return mp;
  95         }
  96
  97         return NULL;
  98 }
  99
 100 struct net_bridge_mdb_entry *br_mdb_ip_get(struct net_bridge_mdb_htable *mdb,
 101                                            struct br_ip *dst)
 102 {
 103         if (!mdb)
 104                 return NULL;
 105
 106         return __br_mdb_ip_get(mdb, dst, br_ip_hash(mdb, dst));
 107 }
 108
 109 static struct net_bridge_mdb_entry *br_mdb_ip4_get(
 110         struct net_bridge_mdb_htable *mdb, __be32 dst, __u16 vid)
 111 {
 112         struct br_ip br_dst;
 113
 114         br_dst.u.ip4 = dst;
 115         br_dst.proto = htons(ETH_P_IP);
 116         br_dst.vid = vid;
 117
 118         return br_mdb_ip_get(mdb, &br_dst);
 119 }
 120
 121 #if IS_ENABLED(CONFIG_IPV6)
 122 static struct net_bridge_mdb_entry *br_mdb_ip6_get(
 123         struct net_bridge_mdb_htable *mdb, const struct in6_addr *dst,
 124         __u16 vid)
 125 {
 126         struct br_ip br_dst;
 127
 128         br_dst.u.ip6 = *dst;
 129         br_dst.proto = htons(ETH_P_IPV6);
 130         br_dst.vid = vid;
 131
 132         return br_mdb_ip_get(mdb, &br_dst);
 133 }
 134 #endif
 135
 136 struct net_bridge_mdb_entry *br_mdb_get(struct net_bridge *br,
 137                                         struct sk_buff *skb, u16 vid)
 138 {
 139         struct net_bridge_mdb_htable *mdb = rcu_dereference(br->mdb);
 140         struct br_ip ip;
 141
 142         if (br->multicast_disabled)
 143                 return NULL;
 144
 145         if (BR_INPUT_SKB_CB(skb)->igmp)
 146                 return NULL;
 147
 148         ip.proto = skb->protocol;
 149         ip.vid = vid;
 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         int maxlen;
 184         int len;
 185         int i;
 186
 187         for (i = 0; i < old->max; i++)
 188                 hlist_for_each_entry(mp, &old->mhash[i], hlist[old->ver])
 189                         hlist_add_head(&mp->hlist[new->ver],
 190                                        &new->mhash[br_ip_hash(new, &mp->addr)]);
 191
 192         if (!elasticity)
 193                 return 0;
 194
 195         maxlen = 0;
 196         for (i = 0; i < new->max; i++) {
 197                 len = 0;
 198                 hlist_for_each_entry(mp, &new->mhash[i], hlist[new->ver])
 199                         len++;
 200                 if (len > maxlen)
 201                         maxlen = len;
 202         }
 203
 204         return maxlen > elasticity ? -EINVAL : 0;
 205 }
 206
 207 void br_multicast_free_pg(struct rcu_head *head)
 208 {
 209         struct net_bridge_port_group *p =
 210                 container_of(head, struct net_bridge_port_group, rcu);
 211
 212         kfree(p);
 213 }
 214
 215 static void br_multicast_free_group(struct rcu_head *head)
 216 {
 217         struct net_bridge_mdb_entry *mp =
 218                 container_of(head, struct net_bridge_mdb_entry, rcu);
 219
 220         kfree(mp);
 221 }
 222
 223 static void br_multicast_group_expired(unsigned long data)
 224 {
 225         struct net_bridge_mdb_entry *mp = (void *)data;
 226         struct net_bridge *br = mp->br;
 227         struct net_bridge_mdb_htable *mdb;
 228
 229         spin_lock(&br->multicast_lock);
 230         if (!netif_running(br->dev) || timer_pending(&mp->timer))
 231                 goto out;
 232
 233         mp->mglist = false;
 234
 235         if (mp->ports)
 236                 goto out;
 237
 238         mdb = mlock_dereference(br->mdb, br);
 239
 240         hlist_del_rcu(&mp->hlist[mdb->ver]);
 241         mdb->size--;
 242
 243         call_rcu_bh(&mp->rcu, br_multicast_free_group);
 244
 245 out:
 246         spin_unlock(&br->multicast_lock);
 247 }
 248
 249 static void br_multicast_del_pg(struct net_bridge *br,
 250                                 struct net_bridge_port_group *pg)
 251 {
 252         struct net_bridge_mdb_htable *mdb;
 253         struct net_bridge_mdb_entry *mp;
 254         struct net_bridge_port_group *p;
 255         struct net_bridge_port_group __rcu **pp;
 256
 257         mdb = mlock_dereference(br->mdb, br);
 258
 259         mp = br_mdb_ip_get(mdb, &pg->addr);
 260         if (WARN_ON(!mp))
 261                 return;
 262
 263         for (pp = &mp->ports;
 264              (p = mlock_dereference(*pp, br)) != NULL;
 265              pp = &p->next) {
 266                 if (p != pg)
 267                         continue;
 268
 269                 rcu_assign_pointer(*pp, p->next);
 270                 hlist_del_init(&p->mglist);
 271                 del_timer(&p->timer);
 272                 call_rcu_bh(&p->rcu, br_multicast_free_pg);
 273
 274                 if (!mp->ports && !mp->mglist && mp->timer_armed &&
 275                     netif_running(br->dev))
 276                         mod_timer(&mp->timer, jiffies);
 277
 278                 return;
 279         }
 280
 281         WARN_ON(1);
 282 }
 283
 284 static void br_multicast_port_group_expired(unsigned long data)
 285 {
 286         struct net_bridge_port_group *pg = (void *)data;
 287         struct net_bridge *br = pg->port->br;
 288
 289         spin_lock(&br->multicast_lock);
 290         if (!netif_running(br->dev) || timer_pending(&pg->timer) ||
 291             hlist_unhashed(&pg->mglist) || pg->state & MDB_PERMANENT)
 292                 goto out;
 293
 294         br_multicast_del_pg(br, pg);
 295
 296 out:
 297         spin_unlock(&br->multicast_lock);
 298 }
 299
 300 static int br_mdb_rehash(struct net_bridge_mdb_htable __rcu **mdbp, int max,
 301                          int elasticity)
 302 {
 303         struct net_bridge_mdb_htable *old = rcu_dereference_protected(*mdbp, 1);
 304         struct net_bridge_mdb_htable *mdb;
 305         int err;
 306
 307         mdb = kmalloc(sizeof(*mdb), GFP_ATOMIC);
 308         if (!mdb)
 309                 return -ENOMEM;
 310
 311         mdb->max = max;
 312         mdb->old = old;
 313
 314         mdb->mhash = kzalloc(max * sizeof(*mdb->mhash), GFP_ATOMIC);
 315         if (!mdb->mhash) {
 316                 kfree(mdb);
 317                 return -ENOMEM;
 318         }
 319
 320         mdb->size = old ? old->size : 0;
 321         mdb->ver = old ? old->ver ^ 1 : 0;
 322
 323         if (!old || elasticity)
 324                 get_random_bytes(&mdb->secret, sizeof(mdb->secret));
 325         else
 326                 mdb->secret = old->secret;
 327
 328         if (!old)
 329                 goto out;
 330
 331         err = br_mdb_copy(mdb, old, elasticity);
 332         if (err) {
 333                 kfree(mdb->mhash);
 334                 kfree(mdb);
 335                 return err;
 336         }
 337
 338         br_mdb_rehash_seq++;
 339         call_rcu_bh(&mdb->rcu, br_mdb_free);
 340
 341 out:
 342         rcu_assign_pointer(*mdbp, mdb);
 343
 344         return 0;
 345 }
 346
 347 static struct sk_buff *br_ip4_multicast_alloc_query(struct net_bridge *br,
 348                                                     __be32 group)
 349 {
 350         struct sk_buff *skb;
 351         struct igmphdr *ih;
 352         struct ethhdr *eth;
 353         struct iphdr *iph;
 354
 355         skb = netdev_alloc_skb_ip_align(br->dev, sizeof(*eth) + sizeof(*iph) +
 356                                                  sizeof(*ih) + 4);
 357         if (!skb)
 358                 goto out;
 359
 360         skb->protocol = htons(ETH_P_IP);
 361
 362         skb_reset_mac_header(skb);
 363         eth = eth_hdr(skb);
 364
 365         memcpy(eth->h_source, br->dev->dev_addr, 6);
 366         eth->h_dest[0] = 1;
 367         eth->h_dest[1] = 0;
 368         eth->h_dest[2] = 0x5e;
 369         eth->h_dest[3] = 0;
 370         eth->h_dest[4] = 0;
 371         eth->h_dest[5] = 1;
 372         eth->h_proto = htons(ETH_P_IP);
 373         skb_put(skb, sizeof(*eth));
 374
 375         skb_set_network_header(skb, skb->len);
 376         iph = ip_hdr(skb);
 377
 378         iph->version = 4;
 379         iph->ihl = 6;
 380         iph->tos = 0xc0;
 381         iph->tot_len = htons(sizeof(*iph) + sizeof(*ih) + 4);
 382         iph->id = 0;
 383         iph->frag_off = htons(IP_DF);
 384         iph->ttl = 1;
 385         iph->protocol = IPPROTO_IGMP;
 386         iph->saddr = br->multicast_query_use_ifaddr ?
 387                      inet_select_addr(br->dev, 0, RT_SCOPE_LINK) : 0;
 388         iph->daddr = htonl(INADDR_ALLHOSTS_GROUP);
 389         ((u8 *)&iph[1])[0] = IPOPT_RA;
 390         ((u8 *)&iph[1])[1] = 4;
 391         ((u8 *)&iph[1])[2] = 0;
 392         ((u8 *)&iph[1])[3] = 0;
 393         ip_send_check(iph);
 394         skb_put(skb, 24);
 395
 396         skb_set_transport_header(skb, skb->len);
 397         ih = igmp_hdr(skb);
 398         ih->type = IGMP_HOST_MEMBERSHIP_QUERY;
 399         ih->code = (group ? br->multicast_last_member_interval :
 400                             br->multicast_query_response_interval) /
 401                    (HZ / IGMP_TIMER_SCALE);
 402         ih->group = group;
 403         ih->csum = 0;
 404         ih->csum = ip_compute_csum((void *)ih, sizeof(struct igmphdr));
 405         skb_put(skb, sizeof(*ih));
 406
 407         __skb_pull(skb, sizeof(*eth));
 408
 409 out:
 410         return skb;
 411 }
 412
 413 #if IS_ENABLED(CONFIG_IPV6)
 414 static struct sk_buff *br_ip6_multicast_alloc_query(struct net_bridge *br,
 415                                                     const struct in6_addr *group)
 416 {
 417         struct sk_buff *skb;
 418         struct ipv6hdr *ip6h;
 419         struct mld_msg *mldq;
 420         struct ethhdr *eth;
 421         u8 *hopopt;
 422         unsigned long interval;
 423
 424         skb = netdev_alloc_skb_ip_align(br->dev, sizeof(*eth) + sizeof(*ip6h) +
 425                                                  8 + sizeof(*mldq));
 426         if (!skb)
 427                 goto out;
 428
 429         skb->protocol = htons(ETH_P_IPV6);
 430
 431         /* Ethernet header */
 432         skb_reset_mac_header(skb);
 433         eth = eth_hdr(skb);
 434
 435         memcpy(eth->h_source, br->dev->dev_addr, 6);
 436         eth->h_proto = htons(ETH_P_IPV6);
 437         skb_put(skb, sizeof(*eth));
 438
 439         /* IPv6 header + HbH option */
 440         skb_set_network_header(skb, skb->len);
 441         ip6h = ipv6_hdr(skb);
 442
 443         *(__force __be32 *)ip6h = htonl(0x60000000);
 444         ip6h->payload_len = htons(8 + sizeof(*mldq));
 445         ip6h->nexthdr = IPPROTO_HOPOPTS;
 446         ip6h->hop_limit = 1;
 447         ipv6_addr_set(&ip6h->daddr, htonl(0xff020000), 0, 0, htonl(1));
 448         if (ipv6_dev_get_saddr(dev_net(br->dev), br->dev, &ip6h->daddr, 0,
 449                                &ip6h->saddr)) {
 450                 kfree_skb(skb);
 451                 return NULL;
 452         }
 453         ipv6_eth_mc_map(&ip6h->daddr, eth->h_dest);
 454
 455         hopopt = (u8 *)(ip6h + 1);
 456         hopopt[0] = IPPROTO_ICMPV6;             /* next hdr */
 457         hopopt[1] = 0;                          /* length of HbH */
 458         hopopt[2] = IPV6_TLV_ROUTERALERT;       /* Router Alert */
 459         hopopt[3] = 2;                          /* Length of RA Option */
 460         hopopt[4] = 0;                          /* Type = 0x0000 (MLD) */
 461         hopopt[5] = 0;
 462         hopopt[6] = IPV6_TLV_PAD1;              /* Pad1 */
 463         hopopt[7] = IPV6_TLV_PAD1;              /* Pad1 */
 464
 465         skb_put(skb, sizeof(*ip6h) + 8);
 466
 467         /* ICMPv6 */
 468         skb_set_transport_header(skb, skb->len);
 469         mldq = (struct mld_msg *) icmp6_hdr(skb);
 470
 471         interval = ipv6_addr_any(group) ?
 472                         br->multicast_query_response_interval :
 473                         br->multicast_last_member_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         unsigned int count = 0;
 517         unsigned int max;
 518         int elasticity;
 519         int err;
 520
 521         mdb = rcu_dereference_protected(br->mdb, 1);
 522         hlist_for_each_entry(mp, &mdb->mhash[hash], hlist[mdb->ver]) {
 523                 count++;
 524                 if (unlikely(br_ip_equal(group, &mp->addr)))
 525                         return mp;
 526         }
 527
 528         elasticity = 0;
 529         max = mdb->max;
 530
 531         if (unlikely(count > br->hash_elasticity && count)) {
 532                 if (net_ratelimit())
 533                         br_info(br, "Multicast hash table "
 534                                 "chain limit reached: %s\n",
 535                                 port ? port->dev->name : br->dev->name);
 536
 537                 elasticity = br->hash_elasticity;
 538         }
 539
 540         if (mdb->size >= max) {
 541                 max *= 2;
 542                 if (unlikely(max > br->hash_max)) {
 543                         br_warn(br, "Multicast hash table maximum of %d "
 544                                 "reached, disabling snooping: %s\n",
 545                                 br->hash_max,
 546                                 port ? port->dev->name : br->dev->name);
 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 struct net_bridge_mdb_entry *br_multicast_new_group(struct net_bridge *br,
 585         struct net_bridge_port *port, struct br_ip *group)
 586 {
 587         struct net_bridge_mdb_htable *mdb;
 588         struct net_bridge_mdb_entry *mp;
 589         int hash;
 590         int err;
 591
 592         mdb = rcu_dereference_protected(br->mdb, 1);
 593         if (!mdb) {
 594                 err = br_mdb_rehash(&br->mdb, BR_HASH_SIZE, 0);
 595                 if (err)
 596                         return ERR_PTR(err);
 597                 goto rehash;
 598         }
 599
 600         hash = br_ip_hash(mdb, group);
 601         mp = br_multicast_get_group(br, port, group, hash);
 602         switch (PTR_ERR(mp)) {
 603         case 0:
 604                 break;
 605
 606         case -EAGAIN:
 607 rehash:
 608                 mdb = rcu_dereference_protected(br->mdb, 1);
 609                 hash = br_ip_hash(mdb, group);
 610                 break;
 611
 612         default:
 613                 goto out;
 614         }
 615
 616         mp = kzalloc(sizeof(*mp), GFP_ATOMIC);
 617         if (unlikely(!mp))
 618                 return ERR_PTR(-ENOMEM);
 619
 620         mp->br = br;
 621         mp->addr = *group;
 622
 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 struct net_bridge_port_group *br_multicast_new_port_group(
 634                         struct net_bridge_port *port,
 635                         struct br_ip *group,
 636                         struct net_bridge_port_group __rcu *next,
 637                         unsigned char state)
 638 {
 639         struct net_bridge_port_group *p;
 640
 641         p = kzalloc(sizeof(*p), GFP_ATOMIC);
 642         if (unlikely(!p))
 643                 return NULL;
 644
 645         p->addr = *group;
 646         p->port = port;
 647         p->state = state;
 648         rcu_assign_pointer(p->next, next);
 649         hlist_add_head(&p->mglist, &port->mglist);
 650         setup_timer(&p->timer, br_multicast_port_group_expired,
 651                     (unsigned long)p);
 652         return p;
 653 }
 654
 655 static int br_multicast_add_group(struct net_bridge *br,
 656                                   struct net_bridge_port *port,
 657                                   struct br_ip *group)
 658 {
 659         struct net_bridge_mdb_entry *mp;
 660         struct net_bridge_port_group *p;
 661         struct net_bridge_port_group __rcu **pp;
 662         int err;
 663
 664         spin_lock(&br->multicast_lock);
 665         if (!netif_running(br->dev) ||
 666             (port && port->state == BR_STATE_DISABLED))
 667                 goto out;
 668
 669         mp = br_multicast_new_group(br, port, group);
 670         err = PTR_ERR(mp);
 671         if (IS_ERR(mp))
 672                 goto err;
 673
 674         if (!port) {
 675                 mp->mglist = true;
 676                 goto out;
 677         }
 678
 679         for (pp = &mp->ports;
 680              (p = mlock_dereference(*pp, br)) != NULL;
 681              pp = &p->next) {
 682                 if (p->port == port)
 683                         goto out;
 684                 if ((unsigned long)p->port < (unsigned long)port)
 685                         break;
 686         }
 687
 688         p = br_multicast_new_port_group(port, group, *pp, MDB_TEMPORARY);
 689         if (unlikely(!p))
 690                 goto err;
 691         rcu_assign_pointer(*pp, p);
 692         br_mdb_notify(br->dev, port, group, RTM_NEWMDB);
 693
 694 out:
 695         err = 0;
 696
 697 err:
 698         spin_unlock(&br->multicast_lock);
 699         return err;
 700 }
 701
 702 static int br_ip4_multicast_add_group(struct net_bridge *br,
 703                                       struct net_bridge_port *port,
 704                                       __be32 group,
 705                                       __u16 vid)
 706 {
 707         struct br_ip br_group;
 708
 709         if (ipv4_is_local_multicast(group))
 710                 return 0;
 711
 712         br_group.u.ip4 = group;
 713         br_group.proto = htons(ETH_P_IP);
 714         br_group.vid = vid;
 715
 716         return br_multicast_add_group(br, port, &br_group);
 717 }
 718
 719 #if IS_ENABLED(CONFIG_IPV6)
 720 static int br_ip6_multicast_add_group(struct net_bridge *br,
 721                                       struct net_bridge_port *port,
 722                                       const struct in6_addr *group,
 723                                       __u16 vid)
 724 {
 725         struct br_ip br_group;
 726
 727         if (ipv6_addr_is_ll_all_nodes(group))
 728                 return 0;
 729
 730         br_group.u.ip6 = *group;
 731         br_group.proto = htons(ETH_P_IPV6);
 732         br_group.vid = vid;
 733
 734         return br_multicast_add_group(br, port, &br_group);
 735 }
 736 #endif
 737
 738 static void br_multicast_router_expired(unsigned long data)
 739 {
 740         struct net_bridge_port *port = (void *)data;
 741         struct net_bridge *br = port->br;
 742
 743         spin_lock(&br->multicast_lock);
 744         if (port->multicast_router != 1 ||
 745             timer_pending(&port->multicast_router_timer) ||
 746             hlist_unhashed(&port->rlist))
 747                 goto out;
 748
 749         hlist_del_init_rcu(&port->rlist);
 750
 751 out:
 752         spin_unlock(&br->multicast_lock);
 753 }
 754
 755 static void br_multicast_local_router_expired(unsigned long data)
 756 {
 757 }
 758
 759 static void br_multicast_querier_expired(unsigned long data)
 760 {
 761         struct net_bridge *br = (void *)data;
 762
 763         spin_lock(&br->multicast_lock);
 764         if (!netif_running(br->dev) || br->multicast_disabled)
 765                 goto out;
 766
 767         br_multicast_start_querier(br);
 768
 769 out:
 770         spin_unlock(&br->multicast_lock);
 771 }
 772
 773 static void __br_multicast_send_query(struct net_bridge *br,
 774                                       struct net_bridge_port *port,
 775                                       struct br_ip *ip)
 776 {
 777         struct sk_buff *skb;
 778
 779         skb = br_multicast_alloc_query(br, ip);
 780         if (!skb)
 781                 return;
 782
 783         if (port) {
 784                 __skb_push(skb, sizeof(struct ethhdr));
 785                 skb->dev = port->dev;
 786                 NF_HOOK(NFPROTO_BRIDGE, NF_BR_LOCAL_OUT, skb, NULL, skb->dev,
 787                         dev_queue_xmit);
 788         } else
 789                 netif_rx(skb);
 790 }
 791
 792 static void br_multicast_send_query(struct net_bridge *br,
 793                                     struct net_bridge_port *port, u32 sent)
 794 {
 795         unsigned long time;
 796         struct br_ip br_group;
 797
 798         if (!netif_running(br->dev) || br->multicast_disabled ||
 799             !br->multicast_querier ||
 800             timer_pending(&br->multicast_querier_timer))
 801                 return;
 802
 803         memset(&br_group.u, 0, sizeof(br_group.u));
 804
 805         br_group.proto = htons(ETH_P_IP);
 806         __br_multicast_send_query(br, port, &br_group);
 807
 808 #if IS_ENABLED(CONFIG_IPV6)
 809         br_group.proto = htons(ETH_P_IPV6);
 810         __br_multicast_send_query(br, port, &br_group);
 811 #endif
 812
 813         time = jiffies;
 814         time += sent < br->multicast_startup_query_count ?
 815                 br->multicast_startup_query_interval :
 816                 br->multicast_query_interval;
 817         mod_timer(port ? &port->multicast_query_timer :
 818                          &br->multicast_query_timer, time);
 819 }
 820
 821 static void br_multicast_port_query_expired(unsigned long data)
 822 {
 823         struct net_bridge_port *port = (void *)data;
 824         struct net_bridge *br = port->br;
 825
 826         spin_lock(&br->multicast_lock);
 827         if (port->state == BR_STATE_DISABLED ||
 828             port->state == BR_STATE_BLOCKING)
 829                 goto out;
 830
 831         if (port->multicast_startup_queries_sent <
 832             br->multicast_startup_query_count)
 833                 port->multicast_startup_queries_sent++;
 834
 835         br_multicast_send_query(port->br, port,
 836                                 port->multicast_startup_queries_sent);
 837
 838 out:
 839         spin_unlock(&br->multicast_lock);
 840 }
 841
 842 void br_multicast_add_port(struct net_bridge_port *port)
 843 {
 844         port->multicast_router = 1;
 845
 846         setup_timer(&port->multicast_router_timer, br_multicast_router_expired,
 847                     (unsigned long)port);
 848         setup_timer(&port->multicast_query_timer,
 849                     br_multicast_port_query_expired, (unsigned long)port);
 850 }
 851
 852 void br_multicast_del_port(struct net_bridge_port *port)
 853 {
 854         del_timer_sync(&port->multicast_router_timer);
 855 }
 856
 857 static void __br_multicast_enable_port(struct net_bridge_port *port)
 858 {
 859         port->multicast_startup_queries_sent = 0;
 860
 861         if (try_to_del_timer_sync(&port->multicast_query_timer) >= 0 ||
 862             del_timer(&port->multicast_query_timer))
 863                 mod_timer(&port->multicast_query_timer, jiffies);
 864 }
 865
 866 void br_multicast_enable_port(struct net_bridge_port *port)
 867 {
 868         struct net_bridge *br = port->br;
 869
 870         spin_lock(&br->multicast_lock);
 871         if (br->multicast_disabled || !netif_running(br->dev))
 872                 goto out;
 873
 874         __br_multicast_enable_port(port);
 875
 876 out:
 877         spin_unlock(&br->multicast_lock);
 878 }
 879
 880 void br_multicast_disable_port(struct net_bridge_port *port)
 881 {
 882         struct net_bridge *br = port->br;
 883         struct net_bridge_port_group *pg;
 884         struct hlist_node *n;
 885
 886         spin_lock(&br->multicast_lock);
 887         hlist_for_each_entry_safe(pg, n, &port->mglist, mglist)
 888                 br_multicast_del_pg(br, pg);
 889
 890         if (!hlist_unhashed(&port->rlist))
 891                 hlist_del_init_rcu(&port->rlist);
 892         del_timer(&port->multicast_router_timer);
 893         del_timer(&port->multicast_query_timer);
 894         spin_unlock(&br->multicast_lock);
 895 }
 896
 897 static int br_ip4_multicast_igmp3_report(struct net_bridge *br,
 898                                          struct net_bridge_port *port,
 899                                          struct sk_buff *skb)
 900 {
 901         struct igmpv3_report *ih;
 902         struct igmpv3_grec *grec;
 903         int i;
 904         int len;
 905         int num;
 906         int type;
 907         int err = 0;
 908         __be32 group;
 909         u16 vid = 0;
 910
 911         if (!pskb_may_pull(skb, sizeof(*ih)))
 912                 return -EINVAL;
 913
 914         br_vlan_get_tag(skb, &vid);
 915         ih = igmpv3_report_hdr(skb);
 916         num = ntohs(ih->ngrec);
 917         len = sizeof(*ih);
 918
 919         for (i = 0; i < num; i++) {
 920                 len += sizeof(*grec);
 921                 if (!pskb_may_pull(skb, len))
 922                         return -EINVAL;
 923
 924                 grec = (void *)(skb->data + len - sizeof(*grec));
 925                 group = grec->grec_mca;
 926                 type = grec->grec_type;
 927
 928                 len += ntohs(grec->grec_nsrcs) * 4;
 929                 if (!pskb_may_pull(skb, len))
 930                         return -EINVAL;
 931
 932                 /* We treat this as an IGMPv2 report for now. */
 933                 switch (type) {
 934                 case IGMPV3_MODE_IS_INCLUDE:
 935                 case IGMPV3_MODE_IS_EXCLUDE:
 936                 case IGMPV3_CHANGE_TO_INCLUDE:
 937                 case IGMPV3_CHANGE_TO_EXCLUDE:
 938                 case IGMPV3_ALLOW_NEW_SOURCES:
 939                 case IGMPV3_BLOCK_OLD_SOURCES:
 940                         break;
 941
 942                 default:
 943                         continue;
 944                 }
 945
 946                 err = br_ip4_multicast_add_group(br, port, group, vid);
 947                 if (err)
 948                         break;
 949         }
 950
 951         return err;
 952 }
 953
 954 #if IS_ENABLED(CONFIG_IPV6)
 955 static int br_ip6_multicast_mld2_report(struct net_bridge *br,
 956                                         struct net_bridge_port *port,
 957                                         struct sk_buff *skb)
 958 {
 959         struct icmp6hdr *icmp6h;
 960         struct mld2_grec *grec;
 961         int i;
 962         int len;
 963         int num;
 964         int err = 0;
 965         u16 vid = 0;
 966
 967         if (!pskb_may_pull(skb, sizeof(*icmp6h)))
 968                 return -EINVAL;
 969
 970         br_vlan_get_tag(skb, &vid);
 971         icmp6h = icmp6_hdr(skb);
 972         num = ntohs(icmp6h->icmp6_dataun.un_data16[1]);
 973         len = sizeof(*icmp6h);
 974
 975         for (i = 0; i < num; i++) {
 976                 __be16 *nsrcs, _nsrcs;
 977
 978                 nsrcs = skb_header_pointer(skb,
 979                                            len + offsetof(struct mld2_grec,
 980                                                           grec_nsrcs),
 981                                            sizeof(_nsrcs), &_nsrcs);
 982                 if (!nsrcs)
 983                         return -EINVAL;
 984
 985                 if (!pskb_may_pull(skb,
 986                                    len + sizeof(*grec) +
 987                                    sizeof(struct in6_addr) * ntohs(*nsrcs)))
 988                         return -EINVAL;
 989
 990                 grec = (struct mld2_grec *)(skb->data + len);
 991                 len += sizeof(*grec) +
 992                        sizeof(struct in6_addr) * ntohs(*nsrcs);
 993
 994                 /* We treat these as MLDv1 reports for now. */
 995                 switch (grec->grec_type) {
 996                 case MLD2_MODE_IS_INCLUDE:
 997                 case MLD2_MODE_IS_EXCLUDE:
 998                 case MLD2_CHANGE_TO_INCLUDE:
 999                 case MLD2_CHANGE_TO_EXCLUDE:
1000                 case MLD2_ALLOW_NEW_SOURCES:
1001                 case MLD2_BLOCK_OLD_SOURCES:
1002                         break;
1003
1004                 default:
1005                         continue;
1006                 }
1007
1008                 err = br_ip6_multicast_add_group(br, port, &grec->grec_mca,
1009                                                  vid);
1010                 if (!err)
1011                         break;
1012         }
1013
1014         return err;
1015 }
1016 #endif
1017
1018 static void br_multicast_update_querier_timer(struct net_bridge *br,
1019                                               unsigned long max_delay)
1020 {
1021         if (!timer_pending(&br->multicast_querier_timer))
1022                 br->multicast_querier_delay_time = jiffies + max_delay;
1023
1024         mod_timer(&br->multicast_querier_timer,
1025                   jiffies + br->multicast_querier_interval);
1026 }
1027
1028 /*
1029  * Add port to router_list
1030  *  list is maintained ordered by pointer value
1031  *  and locked by br->multicast_lock and RCU
1032  */
1033 static void br_multicast_add_router(struct net_bridge *br,
1034                                     struct net_bridge_port *port)
1035 {
1036         struct net_bridge_port *p;
1037         struct hlist_node *slot = NULL;
1038
1039         hlist_for_each_entry(p, &br->router_list, rlist) {
1040                 if ((unsigned long) port >= (unsigned long) p)
1041                         break;
1042                 slot = &p->rlist;
1043         }
1044
1045         if (slot)
1046                 hlist_add_after_rcu(slot, &port->rlist);
1047         else
1048                 hlist_add_head_rcu(&port->rlist, &br->router_list);
1049 }
1050
1051 static void br_multicast_mark_router(struct net_bridge *br,
1052                                      struct net_bridge_port *port)
1053 {
1054         unsigned long now = jiffies;
1055
1056         if (!port) {
1057                 if (br->multicast_router == 1)
1058                         mod_timer(&br->multicast_router_timer,
1059                                   now + br->multicast_querier_interval);
1060                 return;
1061         }
1062
1063         if (port->multicast_router != 1)
1064                 return;
1065
1066         if (!hlist_unhashed(&port->rlist))
1067                 goto timer;
1068
1069         br_multicast_add_router(br, port);
1070
1071 timer:
1072         mod_timer(&port->multicast_router_timer,
1073                   now + br->multicast_querier_interval);
1074 }
1075
1076 static void br_multicast_query_received(struct net_bridge *br,
1077                                         struct net_bridge_port *port,
1078                                         int saddr,
1079                                         unsigned long max_delay)
1080 {
1081         if (saddr)
1082                 br_multicast_update_querier_timer(br, max_delay);
1083         else if (timer_pending(&br->multicast_querier_timer))
1084                 return;
1085
1086         br_multicast_mark_router(br, port);
1087 }
1088
1089 static int br_ip4_multicast_query(struct net_bridge *br,
1090                                   struct net_bridge_port *port,
1091                                   struct sk_buff *skb)
1092 {
1093         const struct iphdr *iph = ip_hdr(skb);
1094         struct igmphdr *ih = igmp_hdr(skb);
1095         struct net_bridge_mdb_entry *mp;
1096         struct igmpv3_query *ih3;
1097         struct net_bridge_port_group *p;
1098         struct net_bridge_port_group __rcu **pp;
1099         unsigned long max_delay;
1100         unsigned long now = jiffies;
1101         __be32 group;
1102         int err = 0;
1103         u16 vid = 0;
1104
1105         spin_lock(&br->multicast_lock);
1106         if (!netif_running(br->dev) ||
1107             (port && port->state == BR_STATE_DISABLED))
1108                 goto out;
1109
1110         group = ih->group;
1111
1112         if (skb->len == sizeof(*ih)) {
1113                 max_delay = ih->code * (HZ / IGMP_TIMER_SCALE);
1114
1115                 if (!max_delay) {
1116                         max_delay = 10 * HZ;
1117                         group = 0;
1118                 }
1119         } else {
1120                 if (!pskb_may_pull(skb, sizeof(struct igmpv3_query))) {
1121                         err = -EINVAL;
1122                         goto out;
1123                 }
1124
1125                 ih3 = igmpv3_query_hdr(skb);
1126                 if (ih3->nsrcs)
1127                         goto out;
1128
1129                 max_delay = ih3->code ?
1130                             IGMPV3_MRC(ih3->code) * (HZ / IGMP_TIMER_SCALE) : 1;
1131         }
1132
1133         br_multicast_query_received(br, port, !!iph->saddr, max_delay);
1134
1135         if (!group)
1136                 goto out;
1137
1138         br_vlan_get_tag(skb, &vid);
1139         mp = br_mdb_ip4_get(mlock_dereference(br->mdb, br), group, vid);
1140         if (!mp)
1141                 goto out;
1142
1143         mod_timer(&mp->timer, now + br->multicast_membership_interval);
1144         mp->timer_armed = true;
1145
1146         max_delay *= br->multicast_last_member_count;
1147
1148         if (mp->mglist &&
1149             (timer_pending(&mp->timer) ?
1150              time_after(mp->timer.expires, now + max_delay) :
1151              try_to_del_timer_sync(&mp->timer) >= 0))
1152                 mod_timer(&mp->timer, now + max_delay);
1153
1154         for (pp = &mp->ports;
1155              (p = mlock_dereference(*pp, br)) != NULL;
1156              pp = &p->next) {
1157                 if (timer_pending(&p->timer) ?
1158                     time_after(p->timer.expires, now + max_delay) :
1159                     try_to_del_timer_sync(&p->timer) >= 0)
1160                         mod_timer(&p->timer, now + max_delay);
1161         }
1162
1163 out:
1164         spin_unlock(&br->multicast_lock);
1165         return err;
1166 }
1167
1168 #if IS_ENABLED(CONFIG_IPV6)
1169 static int br_ip6_multicast_query(struct net_bridge *br,
1170                                   struct net_bridge_port *port,
1171                                   struct sk_buff *skb)
1172 {
1173         const struct ipv6hdr *ip6h = ipv6_hdr(skb);
1174         struct mld_msg *mld;
1175         struct net_bridge_mdb_entry *mp;
1176         struct mld2_query *mld2q;
1177         struct net_bridge_port_group *p;
1178         struct net_bridge_port_group __rcu **pp;
1179         unsigned long max_delay;
1180         unsigned long now = jiffies;
1181         const struct in6_addr *group = NULL;
1182         int err = 0;
1183         u16 vid = 0;
1184
1185         spin_lock(&br->multicast_lock);
1186         if (!netif_running(br->dev) ||
1187             (port && port->state == BR_STATE_DISABLED))
1188                 goto out;
1189
1190         if (skb->len == sizeof(*mld)) {
1191                 if (!pskb_may_pull(skb, sizeof(*mld))) {
1192                         err = -EINVAL;
1193                         goto out;
1194                 }
1195                 mld = (struct mld_msg *) icmp6_hdr(skb);
1196                 max_delay = msecs_to_jiffies(ntohs(mld->mld_maxdelay));
1197                 if (max_delay)
1198                         group = &mld->mld_mca;
1199         } else {
1200                 if (!pskb_may_pull(skb, sizeof(*mld2q))) {
1201                         err = -EINVAL;
1202                         goto out;
1203                 }
1204                 mld2q = (struct mld2_query *)icmp6_hdr(skb);
1205                 if (!mld2q->mld2q_nsrcs)
1206                         group = &mld2q->mld2q_mca;
1207
1208                 max_delay = max(msecs_to_jiffies(mldv2_mrc(mld2q)), 1UL);
1209         }
1210
1211         br_multicast_query_received(br, port, !ipv6_addr_any(&ip6h->saddr),
1212                                     max_delay);
1213
1214         if (!group)
1215                 goto out;
1216
1217         br_vlan_get_tag(skb, &vid);
1218         mp = br_mdb_ip6_get(mlock_dereference(br->mdb, br), group, vid);
1219         if (!mp)
1220                 goto out;
1221
1222         mod_timer(&mp->timer, now + br->multicast_membership_interval);
1223         mp->timer_armed = true;
1224
1225         max_delay *= br->multicast_last_member_count;
1226         if (mp->mglist &&
1227             (timer_pending(&mp->timer) ?
1228              time_after(mp->timer.expires, now + max_delay) :
1229              try_to_del_timer_sync(&mp->timer) >= 0))
1230                 mod_timer(&mp->timer, now + max_delay);
1231
1232         for (pp = &mp->ports;
1233              (p = mlock_dereference(*pp, br)) != NULL;
1234              pp = &p->next) {
1235                 if (timer_pending(&p->timer) ?
1236                     time_after(p->timer.expires, now + max_delay) :
1237                     try_to_del_timer_sync(&p->timer) >= 0)
1238                         mod_timer(&p->timer, now + max_delay);
1239         }
1240
1241 out:
1242         spin_unlock(&br->multicast_lock);
1243         return err;
1244 }
1245 #endif
1246
1247 static void br_multicast_leave_group(struct net_bridge *br,
1248                                      struct net_bridge_port *port,
1249                                      struct br_ip *group)
1250 {
1251         struct net_bridge_mdb_htable *mdb;
1252         struct net_bridge_mdb_entry *mp;
1253         struct net_bridge_port_group *p;
1254         unsigned long now;
1255         unsigned long time;
1256
1257         spin_lock(&br->multicast_lock);
1258         if (!netif_running(br->dev) ||
1259             (port && port->state == BR_STATE_DISABLED) ||
1260             timer_pending(&br->multicast_querier_timer))
1261                 goto out;
1262
1263         mdb = mlock_dereference(br->mdb, br);
1264         mp = br_mdb_ip_get(mdb, group);
1265         if (!mp)
1266                 goto out;
1267
1268         if (br->multicast_querier &&
1269             !timer_pending(&br->multicast_querier_timer)) {
1270                 __br_multicast_send_query(br, port, &mp->addr);
1271
1272                 time = jiffies + br->multicast_last_member_count *
1273                                  br->multicast_last_member_interval;
1274                 mod_timer(port ? &port->multicast_query_timer :
1275                                  &br->multicast_query_timer, time);
1276
1277                 for (p = mlock_dereference(mp->ports, br);
1278                      p != NULL;
1279                      p = mlock_dereference(p->next, br)) {
1280                         if (p->port != port)
1281                                 continue;
1282
1283                         if (!hlist_unhashed(&p->mglist) &&
1284                             (timer_pending(&p->timer) ?
1285                              time_after(p->timer.expires, time) :
1286                              try_to_del_timer_sync(&p->timer) >= 0)) {
1287                                 mod_timer(&p->timer, time);
1288                         }
1289
1290                         break;
1291                 }
1292         }
1293
1294         if (port && (port->flags & BR_MULTICAST_FAST_LEAVE)) {
1295                 struct net_bridge_port_group __rcu **pp;
1296
1297                 for (pp = &mp->ports;
1298                      (p = mlock_dereference(*pp, br)) != NULL;
1299                      pp = &p->next) {
1300                         if (p->port != port)
1301                                 continue;
1302
1303                         rcu_assign_pointer(*pp, p->next);
1304                         hlist_del_init(&p->mglist);
1305                         del_timer(&p->timer);
1306                         call_rcu_bh(&p->rcu, br_multicast_free_pg);
1307                         br_mdb_notify(br->dev, port, group, RTM_DELMDB);
1308
1309                         if (!mp->ports && !mp->mglist && mp->timer_armed &&
1310                             netif_running(br->dev))
1311                                 mod_timer(&mp->timer, jiffies);
1312                 }
1313                 goto out;
1314         }
1315
1316         now = jiffies;
1317         time = now + br->multicast_last_member_count *
1318                      br->multicast_last_member_interval;
1319
1320         if (!port) {
1321                 if (mp->mglist && mp->timer_armed &&
1322                     (timer_pending(&mp->timer) ?
1323                      time_after(mp->timer.expires, time) :
1324                      try_to_del_timer_sync(&mp->timer) >= 0)) {
1325                         mod_timer(&mp->timer, time);
1326                 }
1327         }
1328
1329 out:
1330         spin_unlock(&br->multicast_lock);
1331 }
1332
1333 static void br_ip4_multicast_leave_group(struct net_bridge *br,
1334                                          struct net_bridge_port *port,
1335                                          __be32 group,
1336                                          __u16 vid)
1337 {
1338         struct br_ip br_group;
1339
1340         if (ipv4_is_local_multicast(group))
1341                 return;
1342
1343         br_group.u.ip4 = group;
1344         br_group.proto = htons(ETH_P_IP);
1345         br_group.vid = vid;
1346
1347         br_multicast_leave_group(br, port, &br_group);
1348 }
1349
1350 #if IS_ENABLED(CONFIG_IPV6)
1351 static void br_ip6_multicast_leave_group(struct net_bridge *br,
1352                                          struct net_bridge_port *port,
1353                                          const struct in6_addr *group,
1354                                          __u16 vid)
1355 {
1356         struct br_ip br_group;
1357
1358         if (ipv6_addr_is_ll_all_nodes(group))
1359                 return;
1360
1361         br_group.u.ip6 = *group;
1362         br_group.proto = htons(ETH_P_IPV6);
1363         br_group.vid = vid;
1364
1365         br_multicast_leave_group(br, port, &br_group);
1366 }
1367 #endif
1368
1369 static int br_multicast_ipv4_rcv(struct net_bridge *br,
1370                                  struct net_bridge_port *port,
1371                                  struct sk_buff *skb)
1372 {
1373         struct sk_buff *skb2 = skb;
1374         const struct iphdr *iph;
1375         struct igmphdr *ih;
1376         unsigned int len;
1377         unsigned int offset;
1378         int err;
1379         u16 vid = 0;
1380
1381         /* We treat OOM as packet loss for now. */
1382         if (!pskb_may_pull(skb, sizeof(*iph)))
1383                 return -EINVAL;
1384
1385         iph = ip_hdr(skb);
1386
1387         if (iph->ihl < 5 || iph->version != 4)
1388                 return -EINVAL;
1389
1390         if (!pskb_may_pull(skb, ip_hdrlen(skb)))
1391                 return -EINVAL;
1392
1393         iph = ip_hdr(skb);
1394
1395         if (unlikely(ip_fast_csum((u8 *)iph, iph->ihl)))
1396                 return -EINVAL;
1397
1398         if (iph->protocol != IPPROTO_IGMP) {
1399                 if (!ipv4_is_local_multicast(iph->daddr))
1400                         BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1401                 return 0;
1402         }
1403
1404         len = ntohs(iph->tot_len);
1405         if (skb->len < len || len < ip_hdrlen(skb))
1406                 return -EINVAL;
1407
1408         if (skb->len > len) {
1409                 skb2 = skb_clone(skb, GFP_ATOMIC);
1410                 if (!skb2)
1411                         return -ENOMEM;
1412
1413                 err = pskb_trim_rcsum(skb2, len);
1414                 if (err)
1415                         goto err_out;
1416         }
1417
1418         len -= ip_hdrlen(skb2);
1419         offset = skb_network_offset(skb2) + ip_hdrlen(skb2);
1420         __skb_pull(skb2, offset);
1421         skb_reset_transport_header(skb2);
1422
1423         err = -EINVAL;
1424         if (!pskb_may_pull(skb2, sizeof(*ih)))
1425                 goto out;
1426
1427         switch (skb2->ip_summed) {
1428         case CHECKSUM_COMPLETE:
1429                 if (!csum_fold(skb2->csum))
1430                         break;
1431                 /* fall through */
1432         case CHECKSUM_NONE:
1433                 skb2->csum = 0;
1434                 if (skb_checksum_complete(skb2))
1435                         goto out;
1436         }
1437
1438         err = 0;
1439
1440         br_vlan_get_tag(skb2, &vid);
1441         BR_INPUT_SKB_CB(skb)->igmp = 1;
1442         ih = igmp_hdr(skb2);
1443
1444         switch (ih->type) {
1445         case IGMP_HOST_MEMBERSHIP_REPORT:
1446         case IGMPV2_HOST_MEMBERSHIP_REPORT:
1447                 BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1448                 err = br_ip4_multicast_add_group(br, port, ih->group, vid);
1449                 break;
1450         case IGMPV3_HOST_MEMBERSHIP_REPORT:
1451                 err = br_ip4_multicast_igmp3_report(br, port, skb2);
1452                 break;
1453         case IGMP_HOST_MEMBERSHIP_QUERY:
1454                 err = br_ip4_multicast_query(br, port, skb2);
1455                 break;
1456         case IGMP_HOST_LEAVE_MESSAGE:
1457                 br_ip4_multicast_leave_group(br, port, ih->group, vid);
1458                 break;
1459         }
1460
1461 out:
1462         __skb_push(skb2, offset);
1463 err_out:
1464         if (skb2 != skb)
1465                 kfree_skb(skb2);
1466         return err;
1467 }
1468
1469 #if IS_ENABLED(CONFIG_IPV6)
1470 static int br_multicast_ipv6_rcv(struct net_bridge *br,
1471                                  struct net_bridge_port *port,
1472                                  struct sk_buff *skb)
1473 {
1474         struct sk_buff *skb2;
1475         const struct ipv6hdr *ip6h;
1476         u8 icmp6_type;
1477         u8 nexthdr;
1478         __be16 frag_off;
1479         unsigned int len;
1480         int offset;
1481         int err;
1482         u16 vid = 0;
1483
1484         if (!pskb_may_pull(skb, sizeof(*ip6h)))
1485                 return -EINVAL;
1486
1487         ip6h = ipv6_hdr(skb);
1488
1489         /*
1490          * We're interested in MLD messages only.
1491          *  - Version is 6
1492          *  - MLD has always Router Alert hop-by-hop option
1493          *  - But we do not support jumbrograms.
1494          */
1495         if (ip6h->version != 6)
1496                 return 0;
1497
1498         /* Prevent flooding this packet if there is no listener present */
1499         if (!ipv6_addr_is_ll_all_nodes(&ip6h->daddr))
1500                 BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1501
1502         if (ip6h->nexthdr != IPPROTO_HOPOPTS ||
1503             ip6h->payload_len == 0)
1504                 return 0;
1505
1506         len = ntohs(ip6h->payload_len) + sizeof(*ip6h);
1507         if (skb->len < len)
1508                 return -EINVAL;
1509
1510         nexthdr = ip6h->nexthdr;
1511         offset = ipv6_skip_exthdr(skb, sizeof(*ip6h), &nexthdr, &frag_off);
1512
1513         if (offset < 0 || nexthdr != IPPROTO_ICMPV6)
1514                 return 0;
1515
1516         /* Okay, we found ICMPv6 header */
1517         skb2 = skb_clone(skb, GFP_ATOMIC);
1518         if (!skb2)
1519                 return -ENOMEM;
1520
1521         err = -EINVAL;
1522         if (!pskb_may_pull(skb2, offset + sizeof(struct icmp6hdr)))
1523                 goto out;
1524
1525         len -= offset - skb_network_offset(skb2);
1526
1527         __skb_pull(skb2, offset);
1528         skb_reset_transport_header(skb2);
1529         skb_postpull_rcsum(skb2, skb_network_header(skb2),
1530                            skb_network_header_len(skb2));
1531
1532         icmp6_type = icmp6_hdr(skb2)->icmp6_type;
1533
1534         switch (icmp6_type) {
1535         case ICMPV6_MGM_QUERY:
1536         case ICMPV6_MGM_REPORT:
1537         case ICMPV6_MGM_REDUCTION:
1538         case ICMPV6_MLD2_REPORT:
1539                 break;
1540         default:
1541                 err = 0;
1542                 goto out;
1543         }
1544
1545         /* Okay, we found MLD message. Check further. */
1546         if (skb2->len > len) {
1547                 err = pskb_trim_rcsum(skb2, len);
1548                 if (err)
1549                         goto out;
1550                 err = -EINVAL;
1551         }
1552
1553         ip6h = ipv6_hdr(skb2);
1554
1555         switch (skb2->ip_summed) {
1556         case CHECKSUM_COMPLETE:
1557                 if (!csum_ipv6_magic(&ip6h->saddr, &ip6h->daddr, skb2->len,
1558                                         IPPROTO_ICMPV6, skb2->csum))
1559                         break;
1560                 /*FALLTHROUGH*/
1561         case CHECKSUM_NONE:
1562                 skb2->csum = ~csum_unfold(csum_ipv6_magic(&ip6h->saddr,
1563                                                         &ip6h->daddr,
1564                                                         skb2->len,
1565                                                         IPPROTO_ICMPV6, 0));
1566                 if (__skb_checksum_complete(skb2))
1567                         goto out;
1568         }
1569
1570         err = 0;
1571
1572         br_vlan_get_tag(skb, &vid);
1573         BR_INPUT_SKB_CB(skb)->igmp = 1;
1574
1575         switch (icmp6_type) {
1576         case ICMPV6_MGM_REPORT:
1577             {
1578                 struct mld_msg *mld;
1579                 if (!pskb_may_pull(skb2, sizeof(*mld))) {
1580                         err = -EINVAL;
1581                         goto out;
1582                 }
1583                 mld = (struct mld_msg *)skb_transport_header(skb2);
1584                 BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1585                 err = br_ip6_multicast_add_group(br, port, &mld->mld_mca, vid);
1586                 break;
1587             }
1588         case ICMPV6_MLD2_REPORT:
1589                 err = br_ip6_multicast_mld2_report(br, port, skb2);
1590                 break;
1591         case ICMPV6_MGM_QUERY:
1592                 err = br_ip6_multicast_query(br, port, skb2);
1593                 break;
1594         case ICMPV6_MGM_REDUCTION:
1595             {
1596                 struct mld_msg *mld;
1597                 if (!pskb_may_pull(skb2, sizeof(*mld))) {
1598                         err = -EINVAL;
1599                         goto out;
1600                 }
1601                 mld = (struct mld_msg *)skb_transport_header(skb2);
1602                 br_ip6_multicast_leave_group(br, port, &mld->mld_mca, vid);
1603             }
1604         }
1605
1606 out:
1607         kfree_skb(skb2);
1608         return err;
1609 }
1610 #endif
1611
1612 int br_multicast_rcv(struct net_bridge *br, struct net_bridge_port *port,
1613                      struct sk_buff *skb)
1614 {
1615         BR_INPUT_SKB_CB(skb)->igmp = 0;
1616         BR_INPUT_SKB_CB(skb)->mrouters_only = 0;
1617
1618         if (br->multicast_disabled)
1619                 return 0;
1620
1621         switch (skb->protocol) {
1622         case htons(ETH_P_IP):
1623                 return br_multicast_ipv4_rcv(br, port, skb);
1624 #if IS_ENABLED(CONFIG_IPV6)
1625         case htons(ETH_P_IPV6):
1626                 return br_multicast_ipv6_rcv(br, port, skb);
1627 #endif
1628         }
1629
1630         return 0;
1631 }
1632
1633 static void br_multicast_query_expired(unsigned long data)
1634 {
1635         struct net_bridge *br = (void *)data;
1636
1637         spin_lock(&br->multicast_lock);
1638         if (br->multicast_startup_queries_sent <
1639             br->multicast_startup_query_count)
1640                 br->multicast_startup_queries_sent++;
1641
1642         br_multicast_send_query(br, NULL, br->multicast_startup_queries_sent);
1643
1644         spin_unlock(&br->multicast_lock);
1645 }
1646
1647 void br_multicast_init(struct net_bridge *br)
1648 {
1649         br->hash_elasticity = 4;
1650         br->hash_max = 512;
1651
1652         br->multicast_router = 1;
1653         br->multicast_querier = 0;
1654         br->multicast_query_use_ifaddr = 0;
1655         br->multicast_last_member_count = 2;
1656         br->multicast_startup_query_count = 2;
1657
1658         br->multicast_last_member_interval = HZ;
1659         br->multicast_query_response_interval = 10 * HZ;
1660         br->multicast_startup_query_interval = 125 * HZ / 4;
1661         br->multicast_query_interval = 125 * HZ;
1662         br->multicast_querier_interval = 255 * HZ;
1663         br->multicast_membership_interval = 260 * HZ;
1664
1665         br->multicast_querier_delay_time = 0;
1666
1667         spin_lock_init(&br->multicast_lock);
1668         setup_timer(&br->multicast_router_timer,
1669                     br_multicast_local_router_expired, 0);
1670         setup_timer(&br->multicast_querier_timer,
1671                     br_multicast_querier_expired, (unsigned long)br);
1672         setup_timer(&br->multicast_query_timer, br_multicast_query_expired,
1673                     (unsigned long)br);
1674 }
1675
1676 void br_multicast_open(struct net_bridge *br)
1677 {
1678         br->multicast_startup_queries_sent = 0;
1679
1680         if (br->multicast_disabled)
1681                 return;
1682
1683         mod_timer(&br->multicast_query_timer, jiffies);
1684 }
1685
1686 void br_multicast_stop(struct net_bridge *br)
1687 {
1688         struct net_bridge_mdb_htable *mdb;
1689         struct net_bridge_mdb_entry *mp;
1690         struct hlist_node *n;
1691         u32 ver;
1692         int i;
1693
1694         del_timer_sync(&br->multicast_router_timer);
1695         del_timer_sync(&br->multicast_querier_timer);
1696         del_timer_sync(&br->multicast_query_timer);
1697
1698         spin_lock_bh(&br->multicast_lock);
1699         mdb = mlock_dereference(br->mdb, br);
1700         if (!mdb)
1701                 goto out;
1702
1703         br->mdb = NULL;
1704
1705         ver = mdb->ver;
1706         for (i = 0; i < mdb->max; i++) {
1707                 hlist_for_each_entry_safe(mp, n, &mdb->mhash[i],
1708                                           hlist[ver]) {
1709                         del_timer(&mp->timer);
1710                         mp->timer_armed = false;
1711                         call_rcu_bh(&mp->rcu, br_multicast_free_group);
1712                 }
1713         }
1714
1715         if (mdb->old) {
1716                 spin_unlock_bh(&br->multicast_lock);
1717                 rcu_barrier_bh();
1718                 spin_lock_bh(&br->multicast_lock);
1719                 WARN_ON(mdb->old);
1720         }
1721
1722         mdb->old = mdb;
1723         call_rcu_bh(&mdb->rcu, br_mdb_free);
1724
1725 out:
1726         spin_unlock_bh(&br->multicast_lock);
1727 }
1728
1729 int br_multicast_set_router(struct net_bridge *br, unsigned long val)
1730 {
1731         int err = -ENOENT;
1732
1733         spin_lock_bh(&br->multicast_lock);
1734         if (!netif_running(br->dev))
1735                 goto unlock;
1736
1737         switch (val) {
1738         case 0:
1739         case 2:
1740                 del_timer(&br->multicast_router_timer);
1741                 /* fall through */
1742         case 1:
1743                 br->multicast_router = val;
1744                 err = 0;
1745                 break;
1746
1747         default:
1748                 err = -EINVAL;
1749                 break;
1750         }
1751
1752 unlock:
1753         spin_unlock_bh(&br->multicast_lock);
1754
1755         return err;
1756 }
1757
1758 int br_multicast_set_port_router(struct net_bridge_port *p, unsigned long val)
1759 {
1760         struct net_bridge *br = p->br;
1761         int err = -ENOENT;
1762
1763         spin_lock(&br->multicast_lock);
1764         if (!netif_running(br->dev) || p->state == BR_STATE_DISABLED)
1765                 goto unlock;
1766
1767         switch (val) {
1768         case 0:
1769         case 1:
1770         case 2:
1771                 p->multicast_router = val;
1772                 err = 0;
1773
1774                 if (val < 2 && !hlist_unhashed(&p->rlist))
1775                         hlist_del_init_rcu(&p->rlist);
1776
1777                 if (val == 1)
1778                         break;
1779
1780                 del_timer(&p->multicast_router_timer);
1781
1782                 if (val == 0)
1783                         break;
1784
1785                 br_multicast_add_router(br, p);
1786                 break;
1787
1788         default:
1789                 err = -EINVAL;
1790                 break;
1791         }
1792
1793 unlock:
1794         spin_unlock(&br->multicast_lock);
1795
1796         return err;
1797 }
1798
1799 static void br_multicast_start_querier(struct net_bridge *br)
1800 {
1801         struct net_bridge_port *port;
1802
1803         br_multicast_open(br);
1804
1805         list_for_each_entry(port, &br->port_list, list) {
1806                 if (port->state == BR_STATE_DISABLED ||
1807                     port->state == BR_STATE_BLOCKING)
1808                         continue;
1809
1810                 __br_multicast_enable_port(port);
1811         }
1812 }
1813
1814 int br_multicast_toggle(struct net_bridge *br, unsigned long val)
1815 {
1816         int err = 0;
1817         struct net_bridge_mdb_htable *mdb;
1818
1819         spin_lock_bh(&br->multicast_lock);
1820         if (br->multicast_disabled == !val)
1821                 goto unlock;
1822
1823         br->multicast_disabled = !val;
1824         if (br->multicast_disabled)
1825                 goto unlock;
1826
1827         if (!netif_running(br->dev))
1828                 goto unlock;
1829
1830         mdb = mlock_dereference(br->mdb, br);
1831         if (mdb) {
1832                 if (mdb->old) {
1833                         err = -EEXIST;
1834 rollback:
1835                         br->multicast_disabled = !!val;
1836                         goto unlock;
1837                 }
1838
1839                 err = br_mdb_rehash(&br->mdb, mdb->max,
1840                                     br->hash_elasticity);
1841                 if (err)
1842                         goto rollback;
1843         }
1844
1845         br_multicast_start_querier(br);
1846
1847 unlock:
1848         spin_unlock_bh(&br->multicast_lock);
1849
1850         return err;
1851 }
1852
1853 int br_multicast_set_querier(struct net_bridge *br, unsigned long val)
1854 {
1855         unsigned long max_delay;
1856
1857         val = !!val;
1858
1859         spin_lock_bh(&br->multicast_lock);
1860         if (br->multicast_querier == val)
1861                 goto unlock;
1862
1863         br->multicast_querier = val;
1864         if (!val)
1865                 goto unlock;
1866
1867         max_delay = br->multicast_query_response_interval;
1868         if (!timer_pending(&br->multicast_querier_timer))
1869                 br->multicast_querier_delay_time = jiffies + max_delay;
1870
1871         br_multicast_start_querier(br);
1872
1873 unlock:
1874         spin_unlock_bh(&br->multicast_lock);
1875
1876         return 0;
1877 }
1878
1879 int br_multicast_set_hash_max(struct net_bridge *br, unsigned long val)
1880 {
1881         int err = -ENOENT;
1882         u32 old;
1883         struct net_bridge_mdb_htable *mdb;
1884
1885         spin_lock(&br->multicast_lock);
1886         if (!netif_running(br->dev))
1887                 goto unlock;
1888
1889         err = -EINVAL;
1890         if (!is_power_of_2(val))
1891                 goto unlock;
1892
1893         mdb = mlock_dereference(br->mdb, br);
1894         if (mdb && val < mdb->size)
1895                 goto unlock;
1896
1897         err = 0;
1898
1899         old = br->hash_max;
1900         br->hash_max = val;
1901
1902         if (mdb) {
1903                 if (mdb->old) {
1904                         err = -EEXIST;
1905 rollback:
1906                         br->hash_max = old;
1907                         goto unlock;
1908                 }
1909
1910                 err = br_mdb_rehash(&br->mdb, br->hash_max,
1911                                     br->hash_elasticity);
1912                 if (err)
1913                         goto rollback;
1914         }
1915
1916 unlock:
1917         spin_unlock(&br->multicast_lock);
1918
1919         return err;
1920 }