net/bridge/br_multicast.c

   1 /*
   2  * Bridge multicast support.
   3  *
   4  * Copyright (c) 2010 Herbert Xu <herbert@gondor.apana.org.au>
   5  *
   6  * This program is free software; you can redistribute it and/or modify it
   7  * under the terms of the GNU General Public License as published by the Free
   8  * Software Foundation; either version 2 of the License, or (at your option)
   9  * any later version.
  10  *
  11  */
  12
  13 #include <linux/err.h>
  14 #include <linux/if_ether.h>
  15 #include <linux/igmp.h>
  16 #include <linux/jhash.h>
  17 #include <linux/kernel.h>
  18 #include <linux/log2.h>
  19 #include <linux/netdevice.h>
  20 #include <linux/netfilter_bridge.h>
  21 #include <linux/random.h>
  22 #include <linux/rculist.h>
  23 #include <linux/skbuff.h>
  24 #include <linux/slab.h>
  25 #include <linux/timer.h>
  26 #include <net/ip.h>
  27 #if IS_ENABLED(CONFIG_IPV6)
  28 #include <net/ipv6.h>
  29 #include <net/mld.h>
  30 #include <net/ip6_checksum.h>
  31 #endif
  32
  33 #include "br_private.h"
  34
  35 static void br_multicast_start_querier(struct net_bridge *br);
  36 unsigned int br_mdb_rehash_seq;
  37
  38 static inline int br_ip_equal(const struct br_ip *a, const struct br_ip *b)
  39 {
  40         if (a->proto != b->proto)
  41                 return 0;
  42         if (a->vid != b->vid)
  43                 return 0;
  44         switch (a->proto) {
  45         case htons(ETH_P_IP):
  46                 return a->u.ip4 == b->u.ip4;
  47 #if IS_ENABLED(CONFIG_IPV6)
  48         case htons(ETH_P_IPV6):
  49                 return ipv6_addr_equal(&a->u.ip6, &b->u.ip6);
  50 #endif
  51         }
  52         return 0;
  53 }
  54
  55 static inline int __br_ip4_hash(struct net_bridge_mdb_htable *mdb, __be32 ip,
  56                                 __u16 vid)
  57 {
  58         return jhash_2words((__force u32)ip, vid, mdb->secret) & (mdb->max - 1);
  59 }
  60
  61 #if IS_ENABLED(CONFIG_IPV6)
  62 static inline int __br_ip6_hash(struct net_bridge_mdb_htable *mdb,
  63                                 const struct in6_addr *ip,
  64                                 __u16 vid)
  65 {
  66         return jhash_2words(ipv6_addr_hash(ip), vid,
  67                             mdb->secret) & (mdb->max - 1);
  68 }
  69 #endif
  70
  71 static inline int br_ip_hash(struct net_bridge_mdb_htable *mdb,
  72                              struct br_ip *ip)
  73 {
  74         switch (ip->proto) {
  75         case htons(ETH_P_IP):
  76                 return __br_ip4_hash(mdb, ip->u.ip4, ip->vid);
  77 #if IS_ENABLED(CONFIG_IPV6)
  78         case htons(ETH_P_IPV6):
  79                 return __br_ip6_hash(mdb, &ip->u.ip6, ip->vid);
  80 #endif
  81         }
  82         return 0;
  83 }
  84
  85 static struct net_bridge_mdb_entry *__br_mdb_ip_get(
  86         struct net_bridge_mdb_htable *mdb, struct br_ip *dst, int hash)
  87 {
  88         struct net_bridge_mdb_entry *mp;
  89         struct hlist_node *p;
  90
  91         hlist_for_each_entry_rcu(mp, p, &mdb->mhash[hash], hlist[mdb->ver]) {
  92                 if (br_ip_equal(&mp->addr, dst))
  93                         return mp;
  94         }
  95
  96         return NULL;
  97 }
  98
  99 struct net_bridge_mdb_entry *br_mdb_ip_get(struct net_bridge_mdb_htable *mdb,
 100                                            struct br_ip *dst)
 101 {
 102         if (!mdb)
 103                 return NULL;
 104
 105         return __br_mdb_ip_get(mdb, dst, br_ip_hash(mdb, dst));
 106 }
 107
 108 static struct net_bridge_mdb_entry *br_mdb_ip4_get(
 109         struct net_bridge_mdb_htable *mdb, __be32 dst, __u16 vid)
 110 {
 111         struct br_ip br_dst;
 112
 113         br_dst.u.ip4 = dst;
 114         br_dst.proto = htons(ETH_P_IP);
 115         br_dst.vid = vid;
 116
 117         return br_mdb_ip_get(mdb, &br_dst);
 118 }
 119
 120 #if IS_ENABLED(CONFIG_IPV6)
 121 static struct net_bridge_mdb_entry *br_mdb_ip6_get(
 122         struct net_bridge_mdb_htable *mdb, const struct in6_addr *dst,
 123         __u16 vid)
 124 {
 125         struct br_ip br_dst;
 126
 127         br_dst.u.ip6 = *dst;
 128         br_dst.proto = htons(ETH_P_IPV6);
 129         br_dst.vid = vid;
 130
 131         return br_mdb_ip_get(mdb, &br_dst);
 132 }
 133 #endif
 134
 135 struct net_bridge_mdb_entry *br_mdb_get(struct net_bridge *br,
 136                                         struct sk_buff *skb)
 137 {
 138         struct net_bridge_mdb_htable *mdb = rcu_dereference(br->mdb);
 139         struct br_ip ip;
 140
 141         if (br->multicast_disabled)
 142                 return NULL;
 143
 144         if (BR_INPUT_SKB_CB(skb)->igmp)
 145                 return NULL;
 146
 147         ip.proto = skb->protocol;
 148
 149         switch (skb->protocol) {
 150         case htons(ETH_P_IP):
 151                 ip.u.ip4 = ip_hdr(skb)->daddr;
 152                 break;
 153 #if IS_ENABLED(CONFIG_IPV6)
 154         case htons(ETH_P_IPV6):
 155                 ip.u.ip6 = ipv6_hdr(skb)->daddr;
 156                 break;
 157 #endif
 158         default:
 159                 return NULL;
 160         }
 161
 162         return br_mdb_ip_get(mdb, &ip);
 163 }
 164
 165 static void br_mdb_free(struct rcu_head *head)
 166 {
 167         struct net_bridge_mdb_htable *mdb =
 168                 container_of(head, struct net_bridge_mdb_htable, rcu);
 169         struct net_bridge_mdb_htable *old = mdb->old;
 170
 171         mdb->old = NULL;
 172         kfree(old->mhash);
 173         kfree(old);
 174 }
 175
 176 static int br_mdb_copy(struct net_bridge_mdb_htable *new,
 177                        struct net_bridge_mdb_htable *old,
 178                        int elasticity)
 179 {
 180         struct net_bridge_mdb_entry *mp;
 181         struct hlist_node *p;
 182         int maxlen;
 183         int len;
 184         int i;
 185
 186         for (i = 0; i < old->max; i++)
 187                 hlist_for_each_entry(mp, p, &old->mhash[i], hlist[old->ver])
 188                         hlist_add_head(&mp->hlist[new->ver],
 189                                        &new->mhash[br_ip_hash(new, &mp->addr)]);
 190
 191         if (!elasticity)
 192                 return 0;
 193
 194         maxlen = 0;
 195         for (i = 0; i < new->max; i++) {
 196                 len = 0;
 197                 hlist_for_each_entry(mp, p, &new->mhash[i], hlist[new->ver])
 198                         len++;
 199                 if (len > maxlen)
 200                         maxlen = len;
 201         }
 202
 203         return maxlen > elasticity ? -EINVAL : 0;
 204 }
 205
 206 void br_multicast_free_pg(struct rcu_head *head)
 207 {
 208         struct net_bridge_port_group *p =
 209                 container_of(head, struct net_bridge_port_group, rcu);
 210
 211         kfree(p);
 212 }
 213
 214 static void br_multicast_free_group(struct rcu_head *head)
 215 {
 216         struct net_bridge_mdb_entry *mp =
 217                 container_of(head, struct net_bridge_mdb_entry, rcu);
 218
 219         kfree(mp);
 220 }
 221
 222 static void br_multicast_group_expired(unsigned long data)
 223 {
 224         struct net_bridge_mdb_entry *mp = (void *)data;
 225         struct net_bridge *br = mp->br;
 226         struct net_bridge_mdb_htable *mdb;
 227
 228         spin_lock(&br->multicast_lock);
 229         if (!netif_running(br->dev) || timer_pending(&mp->timer))
 230                 goto out;
 231
 232         mp->mglist = false;
 233
 234         if (mp->ports)
 235                 goto out;
 236
 237         mdb = mlock_dereference(br->mdb, br);
 238
 239         hlist_del_rcu(&mp->hlist[mdb->ver]);
 240         mdb->size--;
 241
 242         call_rcu_bh(&mp->rcu, br_multicast_free_group);
 243
 244 out:
 245         spin_unlock(&br->multicast_lock);
 246 }
 247
 248 static void br_multicast_del_pg(struct net_bridge *br,
 249                                 struct net_bridge_port_group *pg)
 250 {
 251         struct net_bridge_mdb_htable *mdb;
 252         struct net_bridge_mdb_entry *mp;
 253         struct net_bridge_port_group *p;
 254         struct net_bridge_port_group __rcu **pp;
 255
 256         mdb = mlock_dereference(br->mdb, br);
 257
 258         mp = br_mdb_ip_get(mdb, &pg->addr);
 259         if (WARN_ON(!mp))
 260                 return;
 261
 262         for (pp = &mp->ports;
 263              (p = mlock_dereference(*pp, br)) != NULL;
 264              pp = &p->next) {
 265                 if (p != pg)
 266                         continue;
 267
 268                 rcu_assign_pointer(*pp, p->next);
 269                 hlist_del_init(&p->mglist);
 270                 del_timer(&p->timer);
 271                 call_rcu_bh(&p->rcu, br_multicast_free_pg);
 272
 273                 if (!mp->ports && !mp->mglist &&
 274                     netif_running(br->dev))
 275                         mod_timer(&mp->timer, jiffies);
 276
 277                 return;
 278         }
 279
 280         WARN_ON(1);
 281 }
 282
 283 static void br_multicast_port_group_expired(unsigned long data)
 284 {
 285         struct net_bridge_port_group *pg = (void *)data;
 286         struct net_bridge *br = pg->port->br;
 287
 288         spin_lock(&br->multicast_lock);
 289         if (!netif_running(br->dev) || timer_pending(&pg->timer) ||
 290             hlist_unhashed(&pg->mglist) || pg->state & MDB_PERMANENT)
 291                 goto out;
 292
 293         br_multicast_del_pg(br, pg);
 294
 295 out:
 296         spin_unlock(&br->multicast_lock);
 297 }
 298
 299 static int br_mdb_rehash(struct net_bridge_mdb_htable __rcu **mdbp, int max,
 300                          int elasticity)
 301 {
 302         struct net_bridge_mdb_htable *old = rcu_dereference_protected(*mdbp, 1);
 303         struct net_bridge_mdb_htable *mdb;
 304         int err;
 305
 306         mdb = kmalloc(sizeof(*mdb), GFP_ATOMIC);
 307         if (!mdb)
 308                 return -ENOMEM;
 309
 310         mdb->max = max;
 311         mdb->old = old;
 312
 313         mdb->mhash = kzalloc(max * sizeof(*mdb->mhash), GFP_ATOMIC);
 314         if (!mdb->mhash) {
 315                 kfree(mdb);
 316                 return -ENOMEM;
 317         }
 318
 319         mdb->size = old ? old->size : 0;
 320         mdb->ver = old ? old->ver ^ 1 : 0;
 321
 322         if (!old || elasticity)
 323                 get_random_bytes(&mdb->secret, sizeof(mdb->secret));
 324         else
 325                 mdb->secret = old->secret;
 326
 327         if (!old)
 328                 goto out;
 329
 330         err = br_mdb_copy(mdb, old, elasticity);
 331         if (err) {
 332                 kfree(mdb->mhash);
 333                 kfree(mdb);
 334                 return err;
 335         }
 336
 337         br_mdb_rehash_seq++;
 338         call_rcu_bh(&mdb->rcu, br_mdb_free);
 339
 340 out:
 341         rcu_assign_pointer(*mdbp, mdb);
 342
 343         return 0;
 344 }
 345
 346 static struct sk_buff *br_ip4_multicast_alloc_query(struct net_bridge *br,
 347                                                     __be32 group)
 348 {
 349         struct sk_buff *skb;
 350         struct igmphdr *ih;
 351         struct ethhdr *eth;
 352         struct iphdr *iph;
 353
 354         skb = netdev_alloc_skb_ip_align(br->dev, sizeof(*eth) + sizeof(*iph) +
 355                                                  sizeof(*ih) + 4);
 356         if (!skb)
 357                 goto out;
 358
 359         skb->protocol = htons(ETH_P_IP);
 360
 361         skb_reset_mac_header(skb);
 362         eth = eth_hdr(skb);
 363
 364         memcpy(eth->h_source, br->dev->dev_addr, 6);
 365         eth->h_dest[0] = 1;
 366         eth->h_dest[1] = 0;
 367         eth->h_dest[2] = 0x5e;
 368         eth->h_dest[3] = 0;
 369         eth->h_dest[4] = 0;
 370         eth->h_dest[5] = 1;
 371         eth->h_proto = htons(ETH_P_IP);
 372         skb_put(skb, sizeof(*eth));
 373
 374         skb_set_network_header(skb, skb->len);
 375         iph = ip_hdr(skb);
 376
 377         iph->version = 4;
 378         iph->ihl = 6;
 379         iph->tos = 0xc0;
 380         iph->tot_len = htons(sizeof(*iph) + sizeof(*ih) + 4);
 381         iph->id = 0;
 382         iph->frag_off = htons(IP_DF);
 383         iph->ttl = 1;
 384         iph->protocol = IPPROTO_IGMP;
 385         iph->saddr = 0;
 386         iph->daddr = htonl(INADDR_ALLHOSTS_GROUP);
 387         ((u8 *)&iph[1])[0] = IPOPT_RA;
 388         ((u8 *)&iph[1])[1] = 4;
 389         ((u8 *)&iph[1])[2] = 0;
 390         ((u8 *)&iph[1])[3] = 0;
 391         ip_send_check(iph);
 392         skb_put(skb, 24);
 393
 394         skb_set_transport_header(skb, skb->len);
 395         ih = igmp_hdr(skb);
 396         ih->type = IGMP_HOST_MEMBERSHIP_QUERY;
 397         ih->code = (group ? br->multicast_last_member_interval :
 398                             br->multicast_query_response_interval) /
 399                    (HZ / IGMP_TIMER_SCALE);
 400         ih->group = group;
 401         ih->csum = 0;
 402         ih->csum = ip_compute_csum((void *)ih, sizeof(struct igmphdr));
 403         skb_put(skb, sizeof(*ih));
 404
 405         __skb_pull(skb, sizeof(*eth));
 406
 407 out:
 408         return skb;
 409 }
 410
 411 #if IS_ENABLED(CONFIG_IPV6)
 412 static struct sk_buff *br_ip6_multicast_alloc_query(struct net_bridge *br,
 413                                                     const struct in6_addr *group)
 414 {
 415         struct sk_buff *skb;
 416         struct ipv6hdr *ip6h;
 417         struct mld_msg *mldq;
 418         struct ethhdr *eth;
 419         u8 *hopopt;
 420         unsigned long interval;
 421
 422         skb = netdev_alloc_skb_ip_align(br->dev, sizeof(*eth) + sizeof(*ip6h) +
 423                                                  8 + sizeof(*mldq));
 424         if (!skb)
 425                 goto out;
 426
 427         skb->protocol = htons(ETH_P_IPV6);
 428
 429         /* Ethernet header */
 430         skb_reset_mac_header(skb);
 431         eth = eth_hdr(skb);
 432
 433         memcpy(eth->h_source, br->dev->dev_addr, 6);
 434         eth->h_proto = htons(ETH_P_IPV6);
 435         skb_put(skb, sizeof(*eth));
 436
 437         /* IPv6 header + HbH option */
 438         skb_set_network_header(skb, skb->len);
 439         ip6h = ipv6_hdr(skb);
 440
 441         *(__force __be32 *)ip6h = htonl(0x60000000);
 442         ip6h->payload_len = htons(8 + sizeof(*mldq));
 443         ip6h->nexthdr = IPPROTO_HOPOPTS;
 444         ip6h->hop_limit = 1;
 445         ipv6_addr_set(&ip6h->daddr, htonl(0xff020000), 0, 0, htonl(1));
 446         if (ipv6_dev_get_saddr(dev_net(br->dev), br->dev, &ip6h->daddr, 0,
 447                                &ip6h->saddr)) {
 448                 kfree_skb(skb);
 449                 return NULL;
 450         }
 451         ipv6_eth_mc_map(&ip6h->daddr, eth->h_dest);
 452
 453         hopopt = (u8 *)(ip6h + 1);
 454         hopopt[0] = IPPROTO_ICMPV6;             /* next hdr */
 455         hopopt[1] = 0;                          /* length of HbH */
 456         hopopt[2] = IPV6_TLV_ROUTERALERT;       /* Router Alert */
 457         hopopt[3] = 2;                          /* Length of RA Option */
 458         hopopt[4] = 0;                          /* Type = 0x0000 (MLD) */
 459         hopopt[5] = 0;
 460         hopopt[6] = IPV6_TLV_PAD1;              /* Pad1 */
 461         hopopt[7] = IPV6_TLV_PAD1;              /* Pad1 */
 462
 463         skb_put(skb, sizeof(*ip6h) + 8);
 464
 465         /* ICMPv6 */
 466         skb_set_transport_header(skb, skb->len);
 467         mldq = (struct mld_msg *) icmp6_hdr(skb);
 468
 469         interval = ipv6_addr_any(group) ? br->multicast_last_member_interval :
 470                                           br->multicast_query_response_interval;
 471
 472         mldq->mld_type = ICMPV6_MGM_QUERY;
 473         mldq->mld_code = 0;
 474         mldq->mld_cksum = 0;
 475         mldq->mld_maxdelay = htons((u16)jiffies_to_msecs(interval));
 476         mldq->mld_reserved = 0;
 477         mldq->mld_mca = *group;
 478
 479         /* checksum */
 480         mldq->mld_cksum = csum_ipv6_magic(&ip6h->saddr, &ip6h->daddr,
 481                                           sizeof(*mldq), IPPROTO_ICMPV6,
 482                                           csum_partial(mldq,
 483                                                        sizeof(*mldq), 0));
 484         skb_put(skb, sizeof(*mldq));
 485
 486         __skb_pull(skb, sizeof(*eth));
 487
 488 out:
 489         return skb;
 490 }
 491 #endif
 492
 493 static struct sk_buff *br_multicast_alloc_query(struct net_bridge *br,
 494                                                 struct br_ip *addr)
 495 {
 496         switch (addr->proto) {
 497         case htons(ETH_P_IP):
 498                 return br_ip4_multicast_alloc_query(br, addr->u.ip4);
 499 #if IS_ENABLED(CONFIG_IPV6)
 500         case htons(ETH_P_IPV6):
 501                 return br_ip6_multicast_alloc_query(br, &addr->u.ip6);
 502 #endif
 503         }
 504         return NULL;
 505 }
 506
 507 static struct net_bridge_mdb_entry *br_multicast_get_group(
 508         struct net_bridge *br, struct net_bridge_port *port,
 509         struct br_ip *group, int hash)
 510 {
 511         struct net_bridge_mdb_htable *mdb;
 512         struct net_bridge_mdb_entry *mp;
 513         struct hlist_node *p;
 514         unsigned int count = 0;
 515         unsigned int max;
 516         int elasticity;
 517         int err;
 518
 519         mdb = rcu_dereference_protected(br->mdb, 1);
 520         hlist_for_each_entry(mp, p, &mdb->mhash[hash], hlist[mdb->ver]) {
 521                 count++;
 522                 if (unlikely(br_ip_equal(group, &mp->addr)))
 523                         return mp;
 524         }
 525
 526         elasticity = 0;
 527         max = mdb->max;
 528
 529         if (unlikely(count > br->hash_elasticity && count)) {
 530                 if (net_ratelimit())
 531                         br_info(br, "Multicast hash table "
 532                                 "chain limit reached: %s\n",
 533                                 port ? port->dev->name : br->dev->name);
 534
 535                 elasticity = br->hash_elasticity;
 536         }
 537
 538         if (mdb->size >= max) {
 539                 max *= 2;
 540                 if (unlikely(max > br->hash_max)) {
 541                         br_warn(br, "Multicast hash table maximum of %d "
 542                                 "reached, disabling snooping: %s\n",
 543                                 br->hash_max,
 544                                 port ? port->dev->name : br->dev->name);
 545                         err = -E2BIG;
 546 disable:
 547                         br->multicast_disabled = 1;
 548                         goto err;
 549                 }
 550         }
 551
 552         if (max > mdb->max || elasticity) {
 553                 if (mdb->old) {
 554                         if (net_ratelimit())
 555                                 br_info(br, "Multicast hash table "
 556                                         "on fire: %s\n",
 557                                         port ? port->dev->name : br->dev->name);
 558                         err = -EEXIST;
 559                         goto err;
 560                 }
 561
 562                 err = br_mdb_rehash(&br->mdb, max, elasticity);
 563                 if (err) {
 564                         br_warn(br, "Cannot rehash multicast "
 565                                 "hash table, disabling snooping: %s, %d, %d\n",
 566                                 port ? port->dev->name : br->dev->name,
 567                                 mdb->size, err);
 568                         goto disable;
 569                 }
 570
 571                 err = -EAGAIN;
 572                 goto err;
 573         }
 574
 575         return NULL;
 576
 577 err:
 578         mp = ERR_PTR(err);
 579         return mp;
 580 }
 581
 582 struct net_bridge_mdb_entry *br_multicast_new_group(struct net_bridge *br,
 583         struct net_bridge_port *port, struct br_ip *group)
 584 {
 585         struct net_bridge_mdb_htable *mdb;
 586         struct net_bridge_mdb_entry *mp;
 587         int hash;
 588         int err;
 589
 590         mdb = rcu_dereference_protected(br->mdb, 1);
 591         if (!mdb) {
 592                 err = br_mdb_rehash(&br->mdb, BR_HASH_SIZE, 0);
 593                 if (err)
 594                         return ERR_PTR(err);
 595                 goto rehash;
 596         }
 597
 598         hash = br_ip_hash(mdb, group);
 599         mp = br_multicast_get_group(br, port, group, hash);
 600         switch (PTR_ERR(mp)) {
 601         case 0:
 602                 break;
 603
 604         case -EAGAIN:
 605 rehash:
 606                 mdb = rcu_dereference_protected(br->mdb, 1);
 607                 hash = br_ip_hash(mdb, group);
 608                 break;
 609
 610         default:
 611                 goto out;
 612         }
 613
 614         mp = kzalloc(sizeof(*mp), GFP_ATOMIC);
 615         if (unlikely(!mp))
 616                 return ERR_PTR(-ENOMEM);
 617
 618         mp->br = br;
 619         mp->addr = *group;
 620         setup_timer(&mp->timer, br_multicast_group_expired,
 621                     (unsigned long)mp);
 622
 623         hlist_add_head_rcu(&mp->hlist[mdb->ver], &mdb->mhash[hash]);
 624         mdb->size++;
 625
 626 out:
 627         return mp;
 628 }
 629
 630 struct net_bridge_port_group *br_multicast_new_port_group(
 631                         struct net_bridge_port *port,
 632                         struct br_ip *group,
 633                         struct net_bridge_port_group __rcu *next,
 634                         unsigned char state)
 635 {
 636         struct net_bridge_port_group *p;
 637
 638         p = kzalloc(sizeof(*p), GFP_ATOMIC);
 639         if (unlikely(!p))
 640                 return NULL;
 641
 642         p->addr = *group;
 643         p->port = port;
 644         p->state = state;
 645         rcu_assign_pointer(p->next, next);
 646         hlist_add_head(&p->mglist, &port->mglist);
 647         setup_timer(&p->timer, br_multicast_port_group_expired,
 648                     (unsigned long)p);
 649         return p;
 650 }
 651
 652 static int br_multicast_add_group(struct net_bridge *br,
 653                                   struct net_bridge_port *port,
 654                                   struct br_ip *group)
 655 {
 656         struct net_bridge_mdb_entry *mp;
 657         struct net_bridge_port_group *p;
 658         struct net_bridge_port_group __rcu **pp;
 659         unsigned long now = jiffies;
 660         int err;
 661
 662         spin_lock(&br->multicast_lock);
 663         if (!netif_running(br->dev) ||
 664             (port && port->state == BR_STATE_DISABLED))
 665                 goto out;
 666
 667         mp = br_multicast_new_group(br, port, group);
 668         err = PTR_ERR(mp);
 669         if (IS_ERR(mp))
 670                 goto err;
 671
 672         if (!port) {
 673                 mp->mglist = true;
 674                 mod_timer(&mp->timer, now + br->multicast_membership_interval);
 675                 goto out;
 676         }
 677
 678         for (pp = &mp->ports;
 679              (p = mlock_dereference(*pp, br)) != NULL;
 680              pp = &p->next) {
 681                 if (p->port == port)
 682                         goto found;
 683                 if ((unsigned long)p->port < (unsigned long)port)
 684                         break;
 685         }
 686
 687         p = br_multicast_new_port_group(port, group, *pp, MDB_TEMPORARY);
 688         if (unlikely(!p))
 689                 goto err;
 690         rcu_assign_pointer(*pp, p);
 691         br_mdb_notify(br->dev, port, group, RTM_NEWMDB);
 692
 693 found:
 694         mod_timer(&p->timer, now + br->multicast_membership_interval);
 695 out:
 696         err = 0;
 697
 698 err:
 699         spin_unlock(&br->multicast_lock);
 700         return err;
 701 }
 702
 703 static int br_ip4_multicast_add_group(struct net_bridge *br,
 704                                       struct net_bridge_port *port,
 705                                       __be32 group,
 706                                       __u16 vid)
 707 {
 708         struct br_ip br_group;
 709
 710         if (ipv4_is_local_multicast(group))
 711                 return 0;
 712
 713         br_group.u.ip4 = group;
 714         br_group.proto = htons(ETH_P_IP);
 715         br_group.vid = vid;
 716
 717         return br_multicast_add_group(br, port, &br_group);
 718 }
 719
 720 #if IS_ENABLED(CONFIG_IPV6)
 721 static int br_ip6_multicast_add_group(struct net_bridge *br,
 722                                       struct net_bridge_port *port,
 723                                       const struct in6_addr *group,
 724                                       __u16 vid)
 725 {
 726         struct br_ip br_group;
 727
 728         if (!ipv6_is_transient_multicast(group))
 729                 return 0;
 730
 731         br_group.u.ip6 = *group;
 732         br_group.proto = htons(ETH_P_IPV6);
 733         br_group.vid = vid;
 734
 735         return br_multicast_add_group(br, port, &br_group);
 736 }
 737 #endif
 738
 739 static void br_multicast_router_expired(unsigned long data)
 740 {
 741         struct net_bridge_port *port = (void *)data;
 742         struct net_bridge *br = port->br;
 743
 744         spin_lock(&br->multicast_lock);
 745         if (port->multicast_router != 1 ||
 746             timer_pending(&port->multicast_router_timer) ||
 747             hlist_unhashed(&port->rlist))
 748                 goto out;
 749
 750         hlist_del_init_rcu(&port->rlist);
 751
 752 out:
 753         spin_unlock(&br->multicast_lock);
 754 }
 755
 756 static void br_multicast_local_router_expired(unsigned long data)
 757 {
 758 }
 759
 760 static void br_multicast_querier_expired(unsigned long data)
 761 {
 762         struct net_bridge *br = (void *)data;
 763
 764         spin_lock(&br->multicast_lock);
 765         if (!netif_running(br->dev) || br->multicast_disabled)
 766                 goto out;
 767
 768         br_multicast_start_querier(br);
 769
 770 out:
 771         spin_unlock(&br->multicast_lock);
 772 }
 773
 774 static void __br_multicast_send_query(struct net_bridge *br,
 775                                       struct net_bridge_port *port,
 776                                       struct br_ip *ip)
 777 {
 778         struct sk_buff *skb;
 779
 780         skb = br_multicast_alloc_query(br, ip);
 781         if (!skb)
 782                 return;
 783
 784         if (port) {
 785                 __skb_push(skb, sizeof(struct ethhdr));
 786                 skb->dev = port->dev;
 787                 NF_HOOK(NFPROTO_BRIDGE, NF_BR_LOCAL_OUT, skb, NULL, skb->dev,
 788                         dev_queue_xmit);
 789         } else
 790                 netif_rx(skb);
 791 }
 792
 793 static void br_multicast_send_query(struct net_bridge *br,
 794                                     struct net_bridge_port *port, u32 sent)
 795 {
 796         unsigned long time;
 797         struct br_ip br_group;
 798
 799         if (!netif_running(br->dev) || br->multicast_disabled ||
 800             !br->multicast_querier ||
 801             timer_pending(&br->multicast_querier_timer))
 802                 return;
 803
 804         memset(&br_group.u, 0, sizeof(br_group.u));
 805
 806         br_group.proto = htons(ETH_P_IP);
 807         __br_multicast_send_query(br, port, &br_group);
 808
 809 #if IS_ENABLED(CONFIG_IPV6)
 810         br_group.proto = htons(ETH_P_IPV6);
 811         __br_multicast_send_query(br, port, &br_group);
 812 #endif
 813
 814         time = jiffies;
 815         time += sent < br->multicast_startup_query_count ?
 816                 br->multicast_startup_query_interval :
 817                 br->multicast_query_interval;
 818         mod_timer(port ? &port->multicast_query_timer :
 819                          &br->multicast_query_timer, time);
 820 }
 821
 822 static void br_multicast_port_query_expired(unsigned long data)
 823 {
 824         struct net_bridge_port *port = (void *)data;
 825         struct net_bridge *br = port->br;
 826
 827         spin_lock(&br->multicast_lock);
 828         if (port->state == BR_STATE_DISABLED ||
 829             port->state == BR_STATE_BLOCKING)
 830                 goto out;
 831
 832         if (port->multicast_startup_queries_sent <
 833             br->multicast_startup_query_count)
 834                 port->multicast_startup_queries_sent++;
 835
 836         br_multicast_send_query(port->br, port,
 837                                 port->multicast_startup_queries_sent);
 838
 839 out:
 840         spin_unlock(&br->multicast_lock);
 841 }
 842
 843 void br_multicast_add_port(struct net_bridge_port *port)
 844 {
 845         port->multicast_router = 1;
 846
 847         setup_timer(&port->multicast_router_timer, br_multicast_router_expired,
 848                     (unsigned long)port);
 849         setup_timer(&port->multicast_query_timer,
 850                     br_multicast_port_query_expired, (unsigned long)port);
 851 }
 852
 853 void br_multicast_del_port(struct net_bridge_port *port)
 854 {
 855         del_timer_sync(&port->multicast_router_timer);
 856 }
 857
 858 static void __br_multicast_enable_port(struct net_bridge_port *port)
 859 {
 860         port->multicast_startup_queries_sent = 0;
 861
 862         if (try_to_del_timer_sync(&port->multicast_query_timer) >= 0 ||
 863             del_timer(&port->multicast_query_timer))
 864                 mod_timer(&port->multicast_query_timer, jiffies);
 865 }
 866
 867 void br_multicast_enable_port(struct net_bridge_port *port)
 868 {
 869         struct net_bridge *br = port->br;
 870
 871         spin_lock(&br->multicast_lock);
 872         if (br->multicast_disabled || !netif_running(br->dev))
 873                 goto out;
 874
 875         __br_multicast_enable_port(port);
 876
 877 out:
 878         spin_unlock(&br->multicast_lock);
 879 }
 880
 881 void br_multicast_disable_port(struct net_bridge_port *port)
 882 {
 883         struct net_bridge *br = port->br;
 884         struct net_bridge_port_group *pg;
 885         struct hlist_node *p, *n;
 886
 887         spin_lock(&br->multicast_lock);
 888         hlist_for_each_entry_safe(pg, p, n, &port->mglist, mglist)
 889                 br_multicast_del_pg(br, pg);
 890
 891         if (!hlist_unhashed(&port->rlist))
 892                 hlist_del_init_rcu(&port->rlist);
 893         del_timer(&port->multicast_router_timer);
 894         del_timer(&port->multicast_query_timer);
 895         spin_unlock(&br->multicast_lock);
 896 }
 897
 898 static int br_ip4_multicast_igmp3_report(struct net_bridge *br,
 899                                          struct net_bridge_port *port,
 900                                          struct sk_buff *skb)
 901 {
 902         struct igmpv3_report *ih;
 903         struct igmpv3_grec *grec;
 904         int i;
 905         int len;
 906         int num;
 907         int type;
 908         int err = 0;
 909         __be32 group;
 910         u16 vid = 0;
 911
 912         if (!pskb_may_pull(skb, sizeof(*ih)))
 913                 return -EINVAL;
 914
 915         br_vlan_get_tag(skb, &vid);
 916         ih = igmpv3_report_hdr(skb);
 917         num = ntohs(ih->ngrec);
 918         len = sizeof(*ih);
 919
 920         for (i = 0; i < num; i++) {
 921                 len += sizeof(*grec);
 922                 if (!pskb_may_pull(skb, len))
 923                         return -EINVAL;
 924
 925                 grec = (void *)(skb->data + len - sizeof(*grec));
 926                 group = grec->grec_mca;
 927                 type = grec->grec_type;
 928
 929                 len += ntohs(grec->grec_nsrcs) * 4;
 930                 if (!pskb_may_pull(skb, len))
 931                         return -EINVAL;
 932
 933                 /* We treat this as an IGMPv2 report for now. */
 934                 switch (type) {
 935                 case IGMPV3_MODE_IS_INCLUDE:
 936                 case IGMPV3_MODE_IS_EXCLUDE:
 937                 case IGMPV3_CHANGE_TO_INCLUDE:
 938                 case IGMPV3_CHANGE_TO_EXCLUDE:
 939                 case IGMPV3_ALLOW_NEW_SOURCES:
 940                 case IGMPV3_BLOCK_OLD_SOURCES:
 941                         break;
 942
 943                 default:
 944                         continue;
 945                 }
 946
 947                 err = br_ip4_multicast_add_group(br, port, group, vid);
 948                 if (err)
 949                         break;
 950         }
 951
 952         return err;
 953 }
 954
 955 #if IS_ENABLED(CONFIG_IPV6)
 956 static int br_ip6_multicast_mld2_report(struct net_bridge *br,
 957                                         struct net_bridge_port *port,
 958                                         struct sk_buff *skb)
 959 {
 960         struct icmp6hdr *icmp6h;
 961         struct mld2_grec *grec;
 962         int i;
 963         int len;
 964         int num;
 965         int err = 0;
 966         u16 vid = 0;
 967
 968         if (!pskb_may_pull(skb, sizeof(*icmp6h)))
 969                 return -EINVAL;
 970
 971         br_vlan_get_tag(skb, &vid);
 972         icmp6h = icmp6_hdr(skb);
 973         num = ntohs(icmp6h->icmp6_dataun.un_data16[1]);
 974         len = sizeof(*icmp6h);
 975
 976         for (i = 0; i < num; i++) {
 977                 __be16 *nsrcs, _nsrcs;
 978
 979                 nsrcs = skb_header_pointer(skb,
 980                                            len + offsetof(struct mld2_grec,
 981                                                           grec_nsrcs),
 982                                            sizeof(_nsrcs), &_nsrcs);
 983                 if (!nsrcs)
 984                         return -EINVAL;
 985
 986                 if (!pskb_may_pull(skb,
 987                                    len + sizeof(*grec) +
 988                                    sizeof(struct in6_addr) * ntohs(*nsrcs)))
 989                         return -EINVAL;
 990
 991                 grec = (struct mld2_grec *)(skb->data + len);
 992                 len += sizeof(*grec) +
 993                        sizeof(struct in6_addr) * ntohs(*nsrcs);
 994
 995                 /* We treat these as MLDv1 reports for now. */
 996                 switch (grec->grec_type) {
 997                 case MLD2_MODE_IS_INCLUDE:
 998                 case MLD2_MODE_IS_EXCLUDE:
 999                 case MLD2_CHANGE_TO_INCLUDE:
1000                 case MLD2_CHANGE_TO_EXCLUDE:
1001                 case MLD2_ALLOW_NEW_SOURCES:
1002                 case MLD2_BLOCK_OLD_SOURCES:
1003                         break;
1004
1005                 default:
1006                         continue;
1007                 }
1008
1009                 err = br_ip6_multicast_add_group(br, port, &grec->grec_mca,
1010                                                  vid);
1011                 if (!err)
1012                         break;
1013         }
1014
1015         return err;
1016 }
1017 #endif
1018
1019 /*
1020  * Add port to rotuer_list
1021  *  list is maintained ordered by pointer value
1022  *  and locked by br->multicast_lock and RCU
1023  */
1024 static void br_multicast_add_router(struct net_bridge *br,
1025                                     struct net_bridge_port *port)
1026 {
1027         struct net_bridge_port *p;
1028         struct hlist_node *n, *slot = NULL;
1029
1030         hlist_for_each_entry(p, n, &br->router_list, rlist) {
1031                 if ((unsigned long) port >= (unsigned long) p)
1032                         break;
1033                 slot = n;
1034         }
1035
1036         if (slot)
1037                 hlist_add_after_rcu(slot, &port->rlist);
1038         else
1039                 hlist_add_head_rcu(&port->rlist, &br->router_list);
1040 }
1041
1042 static void br_multicast_mark_router(struct net_bridge *br,
1043                                      struct net_bridge_port *port)
1044 {
1045         unsigned long now = jiffies;
1046
1047         if (!port) {
1048                 if (br->multicast_router == 1)
1049                         mod_timer(&br->multicast_router_timer,
1050                                   now + br->multicast_querier_interval);
1051                 return;
1052         }
1053
1054         if (port->multicast_router != 1)
1055                 return;
1056
1057         if (!hlist_unhashed(&port->rlist))
1058                 goto timer;
1059
1060         br_multicast_add_router(br, port);
1061
1062 timer:
1063         mod_timer(&port->multicast_router_timer,
1064                   now + br->multicast_querier_interval);
1065 }
1066
1067 static void br_multicast_query_received(struct net_bridge *br,
1068                                         struct net_bridge_port *port,
1069                                         int saddr)
1070 {
1071         if (saddr)
1072                 mod_timer(&br->multicast_querier_timer,
1073                           jiffies + br->multicast_querier_interval);
1074         else if (timer_pending(&br->multicast_querier_timer))
1075                 return;
1076
1077         br_multicast_mark_router(br, port);
1078 }
1079
1080 static int br_ip4_multicast_query(struct net_bridge *br,
1081                                   struct net_bridge_port *port,
1082                                   struct sk_buff *skb)
1083 {
1084         const struct iphdr *iph = ip_hdr(skb);
1085         struct igmphdr *ih = igmp_hdr(skb);
1086         struct net_bridge_mdb_entry *mp;
1087         struct igmpv3_query *ih3;
1088         struct net_bridge_port_group *p;
1089         struct net_bridge_port_group __rcu **pp;
1090         unsigned long max_delay;
1091         unsigned long now = jiffies;
1092         __be32 group;
1093         int err = 0;
1094         u16 vid = 0;
1095
1096         spin_lock(&br->multicast_lock);
1097         if (!netif_running(br->dev) ||
1098             (port && port->state == BR_STATE_DISABLED))
1099                 goto out;
1100
1101         br_multicast_query_received(br, port, !!iph->saddr);
1102
1103         group = ih->group;
1104
1105         if (skb->len == sizeof(*ih)) {
1106                 max_delay = ih->code * (HZ / IGMP_TIMER_SCALE);
1107
1108                 if (!max_delay) {
1109                         max_delay = 10 * HZ;
1110                         group = 0;
1111                 }
1112         } else {
1113                 if (!pskb_may_pull(skb, sizeof(struct igmpv3_query))) {
1114                         err = -EINVAL;
1115                         goto out;
1116                 }
1117
1118                 ih3 = igmpv3_query_hdr(skb);
1119                 if (ih3->nsrcs)
1120                         goto out;
1121
1122                 max_delay = ih3->code ?
1123                             IGMPV3_MRC(ih3->code) * (HZ / IGMP_TIMER_SCALE) : 1;
1124         }
1125
1126         if (!group)
1127                 goto out;
1128
1129         br_vlan_get_tag(skb, &vid);
1130         mp = br_mdb_ip4_get(mlock_dereference(br->mdb, br), group, vid);
1131         if (!mp)
1132                 goto out;
1133
1134         max_delay *= br->multicast_last_member_count;
1135
1136         if (mp->mglist &&
1137             (timer_pending(&mp->timer) ?
1138              time_after(mp->timer.expires, now + max_delay) :
1139              try_to_del_timer_sync(&mp->timer) >= 0))
1140                 mod_timer(&mp->timer, now + max_delay);
1141
1142         for (pp = &mp->ports;
1143              (p = mlock_dereference(*pp, br)) != NULL;
1144              pp = &p->next) {
1145                 if (timer_pending(&p->timer) ?
1146                     time_after(p->timer.expires, now + max_delay) :
1147                     try_to_del_timer_sync(&p->timer) >= 0)
1148                         mod_timer(&p->timer, now + max_delay);
1149         }
1150
1151 out:
1152         spin_unlock(&br->multicast_lock);
1153         return err;
1154 }
1155
1156 #if IS_ENABLED(CONFIG_IPV6)
1157 static int br_ip6_multicast_query(struct net_bridge *br,
1158                                   struct net_bridge_port *port,
1159                                   struct sk_buff *skb)
1160 {
1161         const struct ipv6hdr *ip6h = ipv6_hdr(skb);
1162         struct mld_msg *mld;
1163         struct net_bridge_mdb_entry *mp;
1164         struct mld2_query *mld2q;
1165         struct net_bridge_port_group *p;
1166         struct net_bridge_port_group __rcu **pp;
1167         unsigned long max_delay;
1168         unsigned long now = jiffies;
1169         const struct in6_addr *group = NULL;
1170         int err = 0;
1171         u16 vid = 0;
1172
1173         spin_lock(&br->multicast_lock);
1174         if (!netif_running(br->dev) ||
1175             (port && port->state == BR_STATE_DISABLED))
1176                 goto out;
1177
1178         br_multicast_query_received(br, port, !ipv6_addr_any(&ip6h->saddr));
1179
1180         if (skb->len == sizeof(*mld)) {
1181                 if (!pskb_may_pull(skb, sizeof(*mld))) {
1182                         err = -EINVAL;
1183                         goto out;
1184                 }
1185                 mld = (struct mld_msg *) icmp6_hdr(skb);
1186                 max_delay = msecs_to_jiffies(ntohs(mld->mld_maxdelay));
1187                 if (max_delay)
1188                         group = &mld->mld_mca;
1189         } else if (skb->len >= sizeof(*mld2q)) {
1190                 if (!pskb_may_pull(skb, sizeof(*mld2q))) {
1191                         err = -EINVAL;
1192                         goto out;
1193                 }
1194                 mld2q = (struct mld2_query *)icmp6_hdr(skb);
1195                 if (!mld2q->mld2q_nsrcs)
1196                         group = &mld2q->mld2q_mca;
1197                 max_delay = mld2q->mld2q_mrc ? MLDV2_MRC(ntohs(mld2q->mld2q_mrc)) : 1;
1198         }
1199
1200         if (!group)
1201                 goto out;
1202
1203         br_vlan_get_tag(skb, &vid);
1204         mp = br_mdb_ip6_get(mlock_dereference(br->mdb, br), group, vid);
1205         if (!mp)
1206                 goto out;
1207
1208         max_delay *= br->multicast_last_member_count;
1209         if (mp->mglist &&
1210             (timer_pending(&mp->timer) ?
1211              time_after(mp->timer.expires, now + max_delay) :
1212              try_to_del_timer_sync(&mp->timer) >= 0))
1213                 mod_timer(&mp->timer, now + max_delay);
1214
1215         for (pp = &mp->ports;
1216              (p = mlock_dereference(*pp, br)) != NULL;
1217              pp = &p->next) {
1218                 if (timer_pending(&p->timer) ?
1219                     time_after(p->timer.expires, now + max_delay) :
1220                     try_to_del_timer_sync(&p->timer) >= 0)
1221                         mod_timer(&p->timer, now + max_delay);
1222         }
1223
1224 out:
1225         spin_unlock(&br->multicast_lock);
1226         return err;
1227 }
1228 #endif
1229
1230 static void br_multicast_leave_group(struct net_bridge *br,
1231                                      struct net_bridge_port *port,
1232                                      struct br_ip *group)
1233 {
1234         struct net_bridge_mdb_htable *mdb;
1235         struct net_bridge_mdb_entry *mp;
1236         struct net_bridge_port_group *p;
1237         unsigned long now;
1238         unsigned long time;
1239
1240         spin_lock(&br->multicast_lock);
1241         if (!netif_running(br->dev) ||
1242             (port && port->state == BR_STATE_DISABLED) ||
1243             timer_pending(&br->multicast_querier_timer))
1244                 goto out;
1245
1246         mdb = mlock_dereference(br->mdb, br);
1247         mp = br_mdb_ip_get(mdb, group);
1248         if (!mp)
1249                 goto out;
1250
1251         if (port && (port->flags & BR_MULTICAST_FAST_LEAVE)) {
1252                 struct net_bridge_port_group __rcu **pp;
1253
1254                 for (pp = &mp->ports;
1255                      (p = mlock_dereference(*pp, br)) != NULL;
1256                      pp = &p->next) {
1257                         if (p->port != port)
1258                                 continue;
1259
1260                         rcu_assign_pointer(*pp, p->next);
1261                         hlist_del_init(&p->mglist);
1262                         del_timer(&p->timer);
1263                         call_rcu_bh(&p->rcu, br_multicast_free_pg);
1264                         br_mdb_notify(br->dev, port, group, RTM_DELMDB);
1265
1266                         if (!mp->ports && !mp->mglist &&
1267                             netif_running(br->dev))
1268                                 mod_timer(&mp->timer, jiffies);
1269                 }
1270                 goto out;
1271         }
1272
1273         now = jiffies;
1274         time = now + br->multicast_last_member_count *
1275                      br->multicast_last_member_interval;
1276
1277         if (!port) {
1278                 if (mp->mglist &&
1279                     (timer_pending(&mp->timer) ?
1280                      time_after(mp->timer.expires, time) :
1281                      try_to_del_timer_sync(&mp->timer) >= 0)) {
1282                         mod_timer(&mp->timer, time);
1283                 }
1284
1285                 goto out;
1286         }
1287
1288         for (p = mlock_dereference(mp->ports, br);
1289              p != NULL;
1290              p = mlock_dereference(p->next, br)) {
1291                 if (p->port != port)
1292                         continue;
1293
1294                 if (!hlist_unhashed(&p->mglist) &&
1295                     (timer_pending(&p->timer) ?
1296                      time_after(p->timer.expires, time) :
1297                      try_to_del_timer_sync(&p->timer) >= 0)) {
1298                         mod_timer(&p->timer, time);
1299                 }
1300
1301                 break;
1302         }
1303
1304 out:
1305         spin_unlock(&br->multicast_lock);
1306 }
1307
1308 static void br_ip4_multicast_leave_group(struct net_bridge *br,
1309                                          struct net_bridge_port *port,
1310                                          __be32 group,
1311                                          __u16 vid)
1312 {
1313         struct br_ip br_group;
1314
1315         if (ipv4_is_local_multicast(group))
1316                 return;
1317
1318         br_group.u.ip4 = group;
1319         br_group.proto = htons(ETH_P_IP);
1320         br_group.vid = vid;
1321
1322         br_multicast_leave_group(br, port, &br_group);
1323 }
1324
1325 #if IS_ENABLED(CONFIG_IPV6)
1326 static void br_ip6_multicast_leave_group(struct net_bridge *br,
1327                                          struct net_bridge_port *port,
1328                                          const struct in6_addr *group,
1329                                          __u16 vid)
1330 {
1331         struct br_ip br_group;
1332
1333         if (!ipv6_is_transient_multicast(group))
1334                 return;
1335
1336         br_group.u.ip6 = *group;
1337         br_group.proto = htons(ETH_P_IPV6);
1338         br_group.vid = vid;
1339
1340         br_multicast_leave_group(br, port, &br_group);
1341 }
1342 #endif
1343
1344 static int br_multicast_ipv4_rcv(struct net_bridge *br,
1345                                  struct net_bridge_port *port,
1346                                  struct sk_buff *skb)
1347 {
1348         struct sk_buff *skb2 = skb;
1349         const struct iphdr *iph;
1350         struct igmphdr *ih;
1351         unsigned int len;
1352         unsigned int offset;
1353         int err;
1354         u16 vid = 0;
1355
1356         /* We treat OOM as packet loss for now. */
1357         if (!pskb_may_pull(skb, sizeof(*iph)))
1358                 return -EINVAL;
1359
1360         iph = ip_hdr(skb);
1361
1362         if (iph->ihl < 5 || iph->version != 4)
1363                 return -EINVAL;
1364
1365         if (!pskb_may_pull(skb, ip_hdrlen(skb)))
1366                 return -EINVAL;
1367
1368         iph = ip_hdr(skb);
1369
1370         if (unlikely(ip_fast_csum((u8 *)iph, iph->ihl)))
1371                 return -EINVAL;
1372
1373         if (iph->protocol != IPPROTO_IGMP) {
1374                 if ((iph->daddr & IGMP_LOCAL_GROUP_MASK) != IGMP_LOCAL_GROUP)
1375                         BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1376                 return 0;
1377         }
1378
1379         len = ntohs(iph->tot_len);
1380         if (skb->len < len || len < ip_hdrlen(skb))
1381                 return -EINVAL;
1382
1383         if (skb->len > len) {
1384                 skb2 = skb_clone(skb, GFP_ATOMIC);
1385                 if (!skb2)
1386                         return -ENOMEM;
1387
1388                 err = pskb_trim_rcsum(skb2, len);
1389                 if (err)
1390                         goto err_out;
1391         }
1392
1393         len -= ip_hdrlen(skb2);
1394         offset = skb_network_offset(skb2) + ip_hdrlen(skb2);
1395         __skb_pull(skb2, offset);
1396         skb_reset_transport_header(skb2);
1397
1398         err = -EINVAL;
1399         if (!pskb_may_pull(skb2, sizeof(*ih)))
1400                 goto out;
1401
1402         switch (skb2->ip_summed) {
1403         case CHECKSUM_COMPLETE:
1404                 if (!csum_fold(skb2->csum))
1405                         break;
1406                 /* fall through */
1407         case CHECKSUM_NONE:
1408                 skb2->csum = 0;
1409                 if (skb_checksum_complete(skb2))
1410                         goto out;
1411         }
1412
1413         err = 0;
1414
1415         br_vlan_get_tag(skb2, &vid);
1416         BR_INPUT_SKB_CB(skb)->igmp = 1;
1417         ih = igmp_hdr(skb2);
1418
1419         switch (ih->type) {
1420         case IGMP_HOST_MEMBERSHIP_REPORT:
1421         case IGMPV2_HOST_MEMBERSHIP_REPORT:
1422                 BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1423                 err = br_ip4_multicast_add_group(br, port, ih->group, vid);
1424                 break;
1425         case IGMPV3_HOST_MEMBERSHIP_REPORT:
1426                 err = br_ip4_multicast_igmp3_report(br, port, skb2);
1427                 break;
1428         case IGMP_HOST_MEMBERSHIP_QUERY:
1429                 err = br_ip4_multicast_query(br, port, skb2);
1430                 break;
1431         case IGMP_HOST_LEAVE_MESSAGE:
1432                 br_ip4_multicast_leave_group(br, port, ih->group, vid);
1433                 break;
1434         }
1435
1436 out:
1437         __skb_push(skb2, offset);
1438 err_out:
1439         if (skb2 != skb)
1440                 kfree_skb(skb2);
1441         return err;
1442 }
1443
1444 #if IS_ENABLED(CONFIG_IPV6)
1445 static int br_multicast_ipv6_rcv(struct net_bridge *br,
1446                                  struct net_bridge_port *port,
1447                                  struct sk_buff *skb)
1448 {
1449         struct sk_buff *skb2;
1450         const struct ipv6hdr *ip6h;
1451         u8 icmp6_type;
1452         u8 nexthdr;
1453         __be16 frag_off;
1454         unsigned int len;
1455         int offset;
1456         int err;
1457         u16 vid = 0;
1458
1459         if (!pskb_may_pull(skb, sizeof(*ip6h)))
1460                 return -EINVAL;
1461
1462         ip6h = ipv6_hdr(skb);
1463
1464         /*
1465          * We're interested in MLD messages only.
1466          *  - Version is 6
1467          *  - MLD has always Router Alert hop-by-hop option
1468          *  - But we do not support jumbrograms.
1469          */
1470         if (ip6h->version != 6 ||
1471             ip6h->nexthdr != IPPROTO_HOPOPTS ||
1472             ip6h->payload_len == 0)
1473                 return 0;
1474
1475         len = ntohs(ip6h->payload_len) + sizeof(*ip6h);
1476         if (skb->len < len)
1477                 return -EINVAL;
1478
1479         nexthdr = ip6h->nexthdr;
1480         offset = ipv6_skip_exthdr(skb, sizeof(*ip6h), &nexthdr, &frag_off);
1481
1482         if (offset < 0 || nexthdr != IPPROTO_ICMPV6)
1483                 return 0;
1484
1485         /* Okay, we found ICMPv6 header */
1486         skb2 = skb_clone(skb, GFP_ATOMIC);
1487         if (!skb2)
1488                 return -ENOMEM;
1489
1490         err = -EINVAL;
1491         if (!pskb_may_pull(skb2, offset + sizeof(struct icmp6hdr)))
1492                 goto out;
1493
1494         len -= offset - skb_network_offset(skb2);
1495
1496         __skb_pull(skb2, offset);
1497         skb_reset_transport_header(skb2);
1498         skb_postpull_rcsum(skb2, skb_network_header(skb2),
1499                            skb_network_header_len(skb2));
1500
1501         icmp6_type = icmp6_hdr(skb2)->icmp6_type;
1502
1503         switch (icmp6_type) {
1504         case ICMPV6_MGM_QUERY:
1505         case ICMPV6_MGM_REPORT:
1506         case ICMPV6_MGM_REDUCTION:
1507         case ICMPV6_MLD2_REPORT:
1508                 break;
1509         default:
1510                 err = 0;
1511                 goto out;
1512         }
1513
1514         /* Okay, we found MLD message. Check further. */
1515         if (skb2->len > len) {
1516                 err = pskb_trim_rcsum(skb2, len);
1517                 if (err)
1518                         goto out;
1519                 err = -EINVAL;
1520         }
1521
1522         ip6h = ipv6_hdr(skb2);
1523
1524         switch (skb2->ip_summed) {
1525         case CHECKSUM_COMPLETE:
1526                 if (!csum_ipv6_magic(&ip6h->saddr, &ip6h->daddr, skb2->len,
1527                                         IPPROTO_ICMPV6, skb2->csum))
1528                         break;
1529                 /*FALLTHROUGH*/
1530         case CHECKSUM_NONE:
1531                 skb2->csum = ~csum_unfold(csum_ipv6_magic(&ip6h->saddr,
1532                                                         &ip6h->daddr,
1533                                                         skb2->len,
1534                                                         IPPROTO_ICMPV6, 0));
1535                 if (__skb_checksum_complete(skb2))
1536                         goto out;
1537         }
1538
1539         err = 0;
1540
1541         br_vlan_get_tag(skb, &vid);
1542         BR_INPUT_SKB_CB(skb)->igmp = 1;
1543
1544         switch (icmp6_type) {
1545         case ICMPV6_MGM_REPORT:
1546             {
1547                 struct mld_msg *mld;
1548                 if (!pskb_may_pull(skb2, sizeof(*mld))) {
1549                         err = -EINVAL;
1550                         goto out;
1551                 }
1552                 mld = (struct mld_msg *)skb_transport_header(skb2);
1553                 BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1554                 err = br_ip6_multicast_add_group(br, port, &mld->mld_mca, vid);
1555                 break;
1556             }
1557         case ICMPV6_MLD2_REPORT:
1558                 err = br_ip6_multicast_mld2_report(br, port, skb2);
1559                 break;
1560         case ICMPV6_MGM_QUERY:
1561                 err = br_ip6_multicast_query(br, port, skb2);
1562                 break;
1563         case ICMPV6_MGM_REDUCTION:
1564             {
1565                 struct mld_msg *mld;
1566                 if (!pskb_may_pull(skb2, sizeof(*mld))) {
1567                         err = -EINVAL;
1568                         goto out;
1569                 }
1570                 mld = (struct mld_msg *)skb_transport_header(skb2);
1571                 br_ip6_multicast_leave_group(br, port, &mld->mld_mca, vid);
1572             }
1573         }
1574
1575 out:
1576         kfree_skb(skb2);
1577         return err;
1578 }
1579 #endif
1580
1581 int br_multicast_rcv(struct net_bridge *br, struct net_bridge_port *port,
1582                      struct sk_buff *skb)
1583 {
1584         BR_INPUT_SKB_CB(skb)->igmp = 0;
1585         BR_INPUT_SKB_CB(skb)->mrouters_only = 0;
1586
1587         if (br->multicast_disabled)
1588                 return 0;
1589
1590         switch (skb->protocol) {
1591         case htons(ETH_P_IP):
1592                 return br_multicast_ipv4_rcv(br, port, skb);
1593 #if IS_ENABLED(CONFIG_IPV6)
1594         case htons(ETH_P_IPV6):
1595                 return br_multicast_ipv6_rcv(br, port, skb);
1596 #endif
1597         }
1598
1599         return 0;
1600 }
1601
1602 static void br_multicast_query_expired(unsigned long data)
1603 {
1604         struct net_bridge *br = (void *)data;
1605
1606         spin_lock(&br->multicast_lock);
1607         if (br->multicast_startup_queries_sent <
1608             br->multicast_startup_query_count)
1609                 br->multicast_startup_queries_sent++;
1610
1611         br_multicast_send_query(br, NULL, br->multicast_startup_queries_sent);
1612
1613         spin_unlock(&br->multicast_lock);
1614 }
1615
1616 void br_multicast_init(struct net_bridge *br)
1617 {
1618         br->hash_elasticity = 4;
1619         br->hash_max = 512;
1620
1621         br->multicast_router = 1;
1622         br->multicast_querier = 0;
1623         br->multicast_last_member_count = 2;
1624         br->multicast_startup_query_count = 2;
1625
1626         br->multicast_last_member_interval = HZ;
1627         br->multicast_query_response_interval = 10 * HZ;
1628         br->multicast_startup_query_interval = 125 * HZ / 4;
1629         br->multicast_query_interval = 125 * HZ;
1630         br->multicast_querier_interval = 255 * HZ;
1631         br->multicast_membership_interval = 260 * HZ;
1632
1633         spin_lock_init(&br->multicast_lock);
1634         setup_timer(&br->multicast_router_timer,
1635                     br_multicast_local_router_expired, 0);
1636         setup_timer(&br->multicast_querier_timer,
1637                     br_multicast_querier_expired, (unsigned long)br);
1638         setup_timer(&br->multicast_query_timer, br_multicast_query_expired,
1639                     (unsigned long)br);
1640 }
1641
1642 void br_multicast_open(struct net_bridge *br)
1643 {
1644         br->multicast_startup_queries_sent = 0;
1645
1646         if (br->multicast_disabled)
1647                 return;
1648
1649         mod_timer(&br->multicast_query_timer, jiffies);
1650 }
1651
1652 void br_multicast_stop(struct net_bridge *br)
1653 {
1654         struct net_bridge_mdb_htable *mdb;
1655         struct net_bridge_mdb_entry *mp;
1656         struct hlist_node *p, *n;
1657         u32 ver;
1658         int i;
1659
1660         del_timer_sync(&br->multicast_router_timer);
1661         del_timer_sync(&br->multicast_querier_timer);
1662         del_timer_sync(&br->multicast_query_timer);
1663
1664         spin_lock_bh(&br->multicast_lock);
1665         mdb = mlock_dereference(br->mdb, br);
1666         if (!mdb)
1667                 goto out;
1668
1669         br->mdb = NULL;
1670
1671         ver = mdb->ver;
1672         for (i = 0; i < mdb->max; i++) {
1673                 hlist_for_each_entry_safe(mp, p, n, &mdb->mhash[i],
1674                                           hlist[ver]) {
1675                         del_timer(&mp->timer);
1676                         call_rcu_bh(&mp->rcu, br_multicast_free_group);
1677                 }
1678         }
1679
1680         if (mdb->old) {
1681                 spin_unlock_bh(&br->multicast_lock);
1682                 rcu_barrier_bh();
1683                 spin_lock_bh(&br->multicast_lock);
1684                 WARN_ON(mdb->old);
1685         }
1686
1687         mdb->old = mdb;
1688         call_rcu_bh(&mdb->rcu, br_mdb_free);
1689
1690 out:
1691         spin_unlock_bh(&br->multicast_lock);
1692 }
1693
1694 int br_multicast_set_router(struct net_bridge *br, unsigned long val)
1695 {
1696         int err = -ENOENT;
1697
1698         spin_lock_bh(&br->multicast_lock);
1699         if (!netif_running(br->dev))
1700                 goto unlock;
1701
1702         switch (val) {
1703         case 0:
1704         case 2:
1705                 del_timer(&br->multicast_router_timer);
1706                 /* fall through */
1707         case 1:
1708                 br->multicast_router = val;
1709                 err = 0;
1710                 break;
1711
1712         default:
1713                 err = -EINVAL;
1714                 break;
1715         }
1716
1717 unlock:
1718         spin_unlock_bh(&br->multicast_lock);
1719
1720         return err;
1721 }
1722
1723 int br_multicast_set_port_router(struct net_bridge_port *p, unsigned long val)
1724 {
1725         struct net_bridge *br = p->br;
1726         int err = -ENOENT;
1727
1728         spin_lock(&br->multicast_lock);
1729         if (!netif_running(br->dev) || p->state == BR_STATE_DISABLED)
1730                 goto unlock;
1731
1732         switch (val) {
1733         case 0:
1734         case 1:
1735         case 2:
1736                 p->multicast_router = val;
1737                 err = 0;
1738
1739                 if (val < 2 && !hlist_unhashed(&p->rlist))
1740                         hlist_del_init_rcu(&p->rlist);
1741
1742                 if (val == 1)
1743                         break;
1744
1745                 del_timer(&p->multicast_router_timer);
1746
1747                 if (val == 0)
1748                         break;
1749
1750                 br_multicast_add_router(br, p);
1751                 break;
1752
1753         default:
1754                 err = -EINVAL;
1755                 break;
1756         }
1757
1758 unlock:
1759         spin_unlock(&br->multicast_lock);
1760
1761         return err;
1762 }
1763
1764 static void br_multicast_start_querier(struct net_bridge *br)
1765 {
1766         struct net_bridge_port *port;
1767
1768         br_multicast_open(br);
1769
1770         list_for_each_entry(port, &br->port_list, list) {
1771                 if (port->state == BR_STATE_DISABLED ||
1772                     port->state == BR_STATE_BLOCKING)
1773                         continue;
1774
1775                 __br_multicast_enable_port(port);
1776         }
1777 }
1778
1779 int br_multicast_toggle(struct net_bridge *br, unsigned long val)
1780 {
1781         int err = 0;
1782         struct net_bridge_mdb_htable *mdb;
1783
1784         spin_lock_bh(&br->multicast_lock);
1785         if (br->multicast_disabled == !val)
1786                 goto unlock;
1787
1788         br->multicast_disabled = !val;
1789         if (br->multicast_disabled)
1790                 goto unlock;
1791
1792         if (!netif_running(br->dev))
1793                 goto unlock;
1794
1795         mdb = mlock_dereference(br->mdb, br);
1796         if (mdb) {
1797                 if (mdb->old) {
1798                         err = -EEXIST;
1799 rollback:
1800                         br->multicast_disabled = !!val;
1801                         goto unlock;
1802                 }
1803
1804                 err = br_mdb_rehash(&br->mdb, mdb->max,
1805                                     br->hash_elasticity);
1806                 if (err)
1807                         goto rollback;
1808         }
1809
1810         br_multicast_start_querier(br);
1811
1812 unlock:
1813         spin_unlock_bh(&br->multicast_lock);
1814
1815         return err;
1816 }
1817
1818 int br_multicast_set_querier(struct net_bridge *br, unsigned long val)
1819 {
1820         val = !!val;
1821
1822         spin_lock_bh(&br->multicast_lock);
1823         if (br->multicast_querier == val)
1824                 goto unlock;
1825
1826         br->multicast_querier = val;
1827         if (val)
1828                 br_multicast_start_querier(br);
1829
1830 unlock:
1831         spin_unlock_bh(&br->multicast_lock);
1832
1833         return 0;
1834 }
1835
1836 int br_multicast_set_hash_max(struct net_bridge *br, unsigned long val)
1837 {
1838         int err = -ENOENT;
1839         u32 old;
1840         struct net_bridge_mdb_htable *mdb;
1841
1842         spin_lock(&br->multicast_lock);
1843         if (!netif_running(br->dev))
1844                 goto unlock;
1845
1846         err = -EINVAL;
1847         if (!is_power_of_2(val))
1848                 goto unlock;
1849
1850         mdb = mlock_dereference(br->mdb, br);
1851         if (mdb && val < mdb->size)
1852                 goto unlock;
1853
1854         err = 0;
1855
1856         old = br->hash_max;
1857         br->hash_max = val;
1858
1859         if (mdb) {
1860                 if (mdb->old) {
1861                         err = -EEXIST;
1862 rollback:
1863                         br->hash_max = old;
1864                         goto unlock;
1865                 }
1866
1867                 err = br_mdb_rehash(&br->mdb, br->hash_max,
1868                                     br->hash_elasticity);
1869                 if (err)
1870                         goto rollback;
1871         }
1872
1873 unlock:
1874         spin_unlock(&br->multicast_lock);
1875
1876         return err;
1877 }