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 #endif
  33
  34 #include "br_private.h"
  35
  36 static void br_multicast_start_querier(struct net_bridge *br);
  37 unsigned int br_mdb_rehash_seq;
  38
  39 static inline int br_ip_equal(const struct br_ip *a, const struct br_ip *b)
  40 {
  41         if (a->proto != b->proto)
  42                 return 0;
  43         if (a->vid != b->vid)
  44                 return 0;
  45         switch (a->proto) {
  46         case htons(ETH_P_IP):
  47                 return a->u.ip4 == b->u.ip4;
  48 #if IS_ENABLED(CONFIG_IPV6)
  49         case htons(ETH_P_IPV6):
  50                 return ipv6_addr_equal(&a->u.ip6, &b->u.ip6);
  51 #endif
  52         }
  53         return 0;
  54 }
  55
  56 static inline int __br_ip4_hash(struct net_bridge_mdb_htable *mdb, __be32 ip,
  57                                 __u16 vid)
  58 {
  59         return jhash_2words((__force u32)ip, vid, mdb->secret) & (mdb->max - 1);
  60 }
  61
  62 #if IS_ENABLED(CONFIG_IPV6)
  63 static inline int __br_ip6_hash(struct net_bridge_mdb_htable *mdb,
  64                                 const struct in6_addr *ip,
  65                                 __u16 vid)
  66 {
  67         return jhash_2words(ipv6_addr_hash(ip), vid,
  68                             mdb->secret) & (mdb->max - 1);
  69 }
  70 #endif
  71
  72 static inline int br_ip_hash(struct net_bridge_mdb_htable *mdb,
  73                              struct br_ip *ip)
  74 {
  75         switch (ip->proto) {
  76         case htons(ETH_P_IP):
  77                 return __br_ip4_hash(mdb, ip->u.ip4, ip->vid);
  78 #if IS_ENABLED(CONFIG_IPV6)
  79         case htons(ETH_P_IPV6):
  80                 return __br_ip6_hash(mdb, &ip->u.ip6, ip->vid);
  81 #endif
  82         }
  83         return 0;
  84 }
  85
  86 static struct net_bridge_mdb_entry *__br_mdb_ip_get(
  87         struct net_bridge_mdb_htable *mdb, struct br_ip *dst, int hash)
  88 {
  89         struct net_bridge_mdb_entry *mp;
  90
  91         hlist_for_each_entry_rcu(mp, &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, u16 vid)
 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         ip.vid = vid;
 149
 150         switch (skb->protocol) {
 151         case htons(ETH_P_IP):
 152                 ip.u.ip4 = ip_hdr(skb)->daddr;
 153                 break;
 154 #if IS_ENABLED(CONFIG_IPV6)
 155         case htons(ETH_P_IPV6):
 156                 ip.u.ip6 = ipv6_hdr(skb)->daddr;
 157                 break;
 158 #endif
 159         default:
 160                 return NULL;
 161         }
 162
 163         return br_mdb_ip_get(mdb, &ip);
 164 }
 165
 166 static void br_mdb_free(struct rcu_head *head)
 167 {
 168         struct net_bridge_mdb_htable *mdb =
 169                 container_of(head, struct net_bridge_mdb_htable, rcu);
 170         struct net_bridge_mdb_htable *old = mdb->old;
 171
 172         mdb->old = NULL;
 173         kfree(old->mhash);
 174         kfree(old);
 175 }
 176
 177 static int br_mdb_copy(struct net_bridge_mdb_htable *new,
 178                        struct net_bridge_mdb_htable *old,
 179                        int elasticity)
 180 {
 181         struct net_bridge_mdb_entry *mp;
 182         int maxlen;
 183         int len;
 184         int i;
 185
 186         for (i = 0; i < old->max; i++)
 187                 hlist_for_each_entry(mp, &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, &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 = br->multicast_query_use_ifaddr ?
 386                      inet_select_addr(br->dev, 0, RT_SCOPE_LINK) : 0;
 387         iph->daddr = htonl(INADDR_ALLHOSTS_GROUP);
 388         ((u8 *)&iph[1])[0] = IPOPT_RA;
 389         ((u8 *)&iph[1])[1] = 4;
 390         ((u8 *)&iph[1])[2] = 0;
 391         ((u8 *)&iph[1])[3] = 0;
 392         ip_send_check(iph);
 393         skb_put(skb, 24);
 394
 395         skb_set_transport_header(skb, skb->len);
 396         ih = igmp_hdr(skb);
 397         ih->type = IGMP_HOST_MEMBERSHIP_QUERY;
 398         ih->code = (group ? br->multicast_last_member_interval :
 399                             br->multicast_query_response_interval) /
 400                    (HZ / IGMP_TIMER_SCALE);
 401         ih->group = group;
 402         ih->csum = 0;
 403         ih->csum = ip_compute_csum((void *)ih, sizeof(struct igmphdr));
 404         skb_put(skb, sizeof(*ih));
 405
 406         __skb_pull(skb, sizeof(*eth));
 407
 408 out:
 409         return skb;
 410 }
 411
 412 #if IS_ENABLED(CONFIG_IPV6)
 413 static struct sk_buff *br_ip6_multicast_alloc_query(struct net_bridge *br,
 414                                                     const struct in6_addr *group)
 415 {
 416         struct sk_buff *skb;
 417         struct ipv6hdr *ip6h;
 418         struct mld_msg *mldq;
 419         struct ethhdr *eth;
 420         u8 *hopopt;
 421         unsigned long interval;
 422
 423         skb = netdev_alloc_skb_ip_align(br->dev, sizeof(*eth) + sizeof(*ip6h) +
 424                                                  8 + sizeof(*mldq));
 425         if (!skb)
 426                 goto out;
 427
 428         skb->protocol = htons(ETH_P_IPV6);
 429
 430         /* Ethernet header */
 431         skb_reset_mac_header(skb);
 432         eth = eth_hdr(skb);
 433
 434         memcpy(eth->h_source, br->dev->dev_addr, 6);
 435         eth->h_proto = htons(ETH_P_IPV6);
 436         skb_put(skb, sizeof(*eth));
 437
 438         /* IPv6 header + HbH option */
 439         skb_set_network_header(skb, skb->len);
 440         ip6h = ipv6_hdr(skb);
 441
 442         *(__force __be32 *)ip6h = htonl(0x60000000);
 443         ip6h->payload_len = htons(8 + sizeof(*mldq));
 444         ip6h->nexthdr = IPPROTO_HOPOPTS;
 445         ip6h->hop_limit = 1;
 446         ipv6_addr_set(&ip6h->daddr, htonl(0xff020000), 0, 0, htonl(1));
 447         if (ipv6_dev_get_saddr(dev_net(br->dev), br->dev, &ip6h->daddr, 0,
 448                                &ip6h->saddr)) {
 449                 kfree_skb(skb);
 450                 return NULL;
 451         }
 452         ipv6_eth_mc_map(&ip6h->daddr, eth->h_dest);
 453
 454         hopopt = (u8 *)(ip6h + 1);
 455         hopopt[0] = IPPROTO_ICMPV6;             /* next hdr */
 456         hopopt[1] = 0;                          /* length of HbH */
 457         hopopt[2] = IPV6_TLV_ROUTERALERT;       /* Router Alert */
 458         hopopt[3] = 2;                          /* Length of RA Option */
 459         hopopt[4] = 0;                          /* Type = 0x0000 (MLD) */
 460         hopopt[5] = 0;
 461         hopopt[6] = IPV6_TLV_PAD1;              /* Pad1 */
 462         hopopt[7] = IPV6_TLV_PAD1;              /* Pad1 */
 463
 464         skb_put(skb, sizeof(*ip6h) + 8);
 465
 466         /* ICMPv6 */
 467         skb_set_transport_header(skb, skb->len);
 468         mldq = (struct mld_msg *) icmp6_hdr(skb);
 469
 470         interval = ipv6_addr_any(group) ? br->multicast_last_member_interval :
 471                                           br->multicast_query_response_interval;
 472
 473         mldq->mld_type = ICMPV6_MGM_QUERY;
 474         mldq->mld_code = 0;
 475         mldq->mld_cksum = 0;
 476         mldq->mld_maxdelay = htons((u16)jiffies_to_msecs(interval));
 477         mldq->mld_reserved = 0;
 478         mldq->mld_mca = *group;
 479
 480         /* checksum */
 481         mldq->mld_cksum = csum_ipv6_magic(&ip6h->saddr, &ip6h->daddr,
 482                                           sizeof(*mldq), IPPROTO_ICMPV6,
 483                                           csum_partial(mldq,
 484                                                        sizeof(*mldq), 0));
 485         skb_put(skb, sizeof(*mldq));
 486
 487         __skb_pull(skb, sizeof(*eth));
 488
 489 out:
 490         return skb;
 491 }
 492 #endif
 493
 494 static struct sk_buff *br_multicast_alloc_query(struct net_bridge *br,
 495                                                 struct br_ip *addr)
 496 {
 497         switch (addr->proto) {
 498         case htons(ETH_P_IP):
 499                 return br_ip4_multicast_alloc_query(br, addr->u.ip4);
 500 #if IS_ENABLED(CONFIG_IPV6)
 501         case htons(ETH_P_IPV6):
 502                 return br_ip6_multicast_alloc_query(br, &addr->u.ip6);
 503 #endif
 504         }
 505         return NULL;
 506 }
 507
 508 static struct net_bridge_mdb_entry *br_multicast_get_group(
 509         struct net_bridge *br, struct net_bridge_port *port,
 510         struct br_ip *group, int hash)
 511 {
 512         struct net_bridge_mdb_htable *mdb;
 513         struct net_bridge_mdb_entry *mp;
 514         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, &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
 621         hlist_add_head_rcu(&mp->hlist[mdb->ver], &mdb->mhash[hash]);
 622         mdb->size++;
 623
 624 out:
 625         return mp;
 626 }
 627
 628 struct net_bridge_port_group *br_multicast_new_port_group(
 629                         struct net_bridge_port *port,
 630                         struct br_ip *group,
 631                         struct net_bridge_port_group __rcu *next,
 632                         unsigned char state)
 633 {
 634         struct net_bridge_port_group *p;
 635
 636         p = kzalloc(sizeof(*p), GFP_ATOMIC);
 637         if (unlikely(!p))
 638                 return NULL;
 639
 640         p->addr = *group;
 641         p->port = port;
 642         p->state = state;
 643         rcu_assign_pointer(p->next, next);
 644         hlist_add_head(&p->mglist, &port->mglist);
 645         setup_timer(&p->timer, br_multicast_port_group_expired,
 646                     (unsigned long)p);
 647         return p;
 648 }
 649
 650 static int br_multicast_add_group(struct net_bridge *br,
 651                                   struct net_bridge_port *port,
 652                                   struct br_ip *group)
 653 {
 654         struct net_bridge_mdb_entry *mp;
 655         struct net_bridge_port_group *p;
 656         struct net_bridge_port_group __rcu **pp;
 657         int err;
 658
 659         spin_lock(&br->multicast_lock);
 660         if (!netif_running(br->dev) ||
 661             (port && port->state == BR_STATE_DISABLED))
 662                 goto out;
 663
 664         mp = br_multicast_new_group(br, port, group);
 665         err = PTR_ERR(mp);
 666         if (IS_ERR(mp))
 667                 goto err;
 668
 669         if (!port) {
 670                 mp->mglist = true;
 671                 goto out;
 672         }
 673
 674         for (pp = &mp->ports;
 675              (p = mlock_dereference(*pp, br)) != NULL;
 676              pp = &p->next) {
 677                 if (p->port == port)
 678                         goto out;
 679                 if ((unsigned long)p->port < (unsigned long)port)
 680                         break;
 681         }
 682
 683         p = br_multicast_new_port_group(port, group, *pp, MDB_TEMPORARY);
 684         if (unlikely(!p))
 685                 goto err;
 686         rcu_assign_pointer(*pp, p);
 687         br_mdb_notify(br->dev, port, group, RTM_NEWMDB);
 688
 689 out:
 690         err = 0;
 691
 692 err:
 693         spin_unlock(&br->multicast_lock);
 694         return err;
 695 }
 696
 697 static int br_ip4_multicast_add_group(struct net_bridge *br,
 698                                       struct net_bridge_port *port,
 699                                       __be32 group,
 700                                       __u16 vid)
 701 {
 702         struct br_ip br_group;
 703
 704         if (ipv4_is_local_multicast(group))
 705                 return 0;
 706
 707         br_group.u.ip4 = group;
 708         br_group.proto = htons(ETH_P_IP);
 709         br_group.vid = vid;
 710
 711         return br_multicast_add_group(br, port, &br_group);
 712 }
 713
 714 #if IS_ENABLED(CONFIG_IPV6)
 715 static int br_ip6_multicast_add_group(struct net_bridge *br,
 716                                       struct net_bridge_port *port,
 717                                       const struct in6_addr *group,
 718                                       __u16 vid)
 719 {
 720         struct br_ip br_group;
 721
 722         if (!ipv6_is_transient_multicast(group))
 723                 return 0;
 724
 725         br_group.u.ip6 = *group;
 726         br_group.proto = htons(ETH_P_IPV6);
 727         br_group.vid = vid;
 728
 729         return br_multicast_add_group(br, port, &br_group);
 730 }
 731 #endif
 732
 733 static void br_multicast_router_expired(unsigned long data)
 734 {
 735         struct net_bridge_port *port = (void *)data;
 736         struct net_bridge *br = port->br;
 737
 738         spin_lock(&br->multicast_lock);
 739         if (port->multicast_router != 1 ||
 740             timer_pending(&port->multicast_router_timer) ||
 741             hlist_unhashed(&port->rlist))
 742                 goto out;
 743
 744         hlist_del_init_rcu(&port->rlist);
 745
 746 out:
 747         spin_unlock(&br->multicast_lock);
 748 }
 749
 750 static void br_multicast_local_router_expired(unsigned long data)
 751 {
 752 }
 753
 754 static void br_multicast_querier_expired(unsigned long data)
 755 {
 756         struct net_bridge *br = (void *)data;
 757
 758         spin_lock(&br->multicast_lock);
 759         if (!netif_running(br->dev) || br->multicast_disabled)
 760                 goto out;
 761
 762         br_multicast_start_querier(br);
 763
 764 out:
 765         spin_unlock(&br->multicast_lock);
 766 }
 767
 768 static void __br_multicast_send_query(struct net_bridge *br,
 769                                       struct net_bridge_port *port,
 770                                       struct br_ip *ip)
 771 {
 772         struct sk_buff *skb;
 773
 774         skb = br_multicast_alloc_query(br, ip);
 775         if (!skb)
 776                 return;
 777
 778         if (port) {
 779                 __skb_push(skb, sizeof(struct ethhdr));
 780                 skb->dev = port->dev;
 781                 NF_HOOK(NFPROTO_BRIDGE, NF_BR_LOCAL_OUT, skb, NULL, skb->dev,
 782                         dev_queue_xmit);
 783         } else
 784                 netif_rx(skb);
 785 }
 786
 787 static void br_multicast_send_query(struct net_bridge *br,
 788                                     struct net_bridge_port *port, u32 sent)
 789 {
 790         unsigned long time;
 791         struct br_ip br_group;
 792
 793         if (!netif_running(br->dev) || br->multicast_disabled ||
 794             !br->multicast_querier ||
 795             timer_pending(&br->multicast_querier_timer))
 796                 return;
 797
 798         memset(&br_group.u, 0, sizeof(br_group.u));
 799
 800         br_group.proto = htons(ETH_P_IP);
 801         __br_multicast_send_query(br, port, &br_group);
 802
 803 #if IS_ENABLED(CONFIG_IPV6)
 804         br_group.proto = htons(ETH_P_IPV6);
 805         __br_multicast_send_query(br, port, &br_group);
 806 #endif
 807
 808         time = jiffies;
 809         time += sent < br->multicast_startup_query_count ?
 810                 br->multicast_startup_query_interval :
 811                 br->multicast_query_interval;
 812         mod_timer(port ? &port->multicast_query_timer :
 813                          &br->multicast_query_timer, time);
 814 }
 815
 816 static void br_multicast_port_query_expired(unsigned long data)
 817 {
 818         struct net_bridge_port *port = (void *)data;
 819         struct net_bridge *br = port->br;
 820
 821         spin_lock(&br->multicast_lock);
 822         if (port->state == BR_STATE_DISABLED ||
 823             port->state == BR_STATE_BLOCKING)
 824                 goto out;
 825
 826         if (port->multicast_startup_queries_sent <
 827             br->multicast_startup_query_count)
 828                 port->multicast_startup_queries_sent++;
 829
 830         br_multicast_send_query(port->br, port,
 831                                 port->multicast_startup_queries_sent);
 832
 833 out:
 834         spin_unlock(&br->multicast_lock);
 835 }
 836
 837 void br_multicast_add_port(struct net_bridge_port *port)
 838 {
 839         port->multicast_router = 1;
 840
 841         setup_timer(&port->multicast_router_timer, br_multicast_router_expired,
 842                     (unsigned long)port);
 843         setup_timer(&port->multicast_query_timer,
 844                     br_multicast_port_query_expired, (unsigned long)port);
 845 }
 846
 847 void br_multicast_del_port(struct net_bridge_port *port)
 848 {
 849         del_timer_sync(&port->multicast_router_timer);
 850 }
 851
 852 static void __br_multicast_enable_port(struct net_bridge_port *port)
 853 {
 854         port->multicast_startup_queries_sent = 0;
 855
 856         if (try_to_del_timer_sync(&port->multicast_query_timer) >= 0 ||
 857             del_timer(&port->multicast_query_timer))
 858                 mod_timer(&port->multicast_query_timer, jiffies);
 859 }
 860
 861 void br_multicast_enable_port(struct net_bridge_port *port)
 862 {
 863         struct net_bridge *br = port->br;
 864
 865         spin_lock(&br->multicast_lock);
 866         if (br->multicast_disabled || !netif_running(br->dev))
 867                 goto out;
 868
 869         __br_multicast_enable_port(port);
 870
 871 out:
 872         spin_unlock(&br->multicast_lock);
 873 }
 874
 875 void br_multicast_disable_port(struct net_bridge_port *port)
 876 {
 877         struct net_bridge *br = port->br;
 878         struct net_bridge_port_group *pg;
 879         struct hlist_node *n;
 880
 881         spin_lock(&br->multicast_lock);
 882         hlist_for_each_entry_safe(pg, n, &port->mglist, mglist)
 883                 br_multicast_del_pg(br, pg);
 884
 885         if (!hlist_unhashed(&port->rlist))
 886                 hlist_del_init_rcu(&port->rlist);
 887         del_timer(&port->multicast_router_timer);
 888         del_timer(&port->multicast_query_timer);
 889         spin_unlock(&br->multicast_lock);
 890 }
 891
 892 static int br_ip4_multicast_igmp3_report(struct net_bridge *br,
 893                                          struct net_bridge_port *port,
 894                                          struct sk_buff *skb)
 895 {
 896         struct igmpv3_report *ih;
 897         struct igmpv3_grec *grec;
 898         int i;
 899         int len;
 900         int num;
 901         int type;
 902         int err = 0;
 903         __be32 group;
 904         u16 vid = 0;
 905
 906         if (!pskb_may_pull(skb, sizeof(*ih)))
 907                 return -EINVAL;
 908
 909         br_vlan_get_tag(skb, &vid);
 910         ih = igmpv3_report_hdr(skb);
 911         num = ntohs(ih->ngrec);
 912         len = sizeof(*ih);
 913
 914         for (i = 0; i < num; i++) {
 915                 len += sizeof(*grec);
 916                 if (!pskb_may_pull(skb, len))
 917                         return -EINVAL;
 918
 919                 grec = (void *)(skb->data + len - sizeof(*grec));
 920                 group = grec->grec_mca;
 921                 type = grec->grec_type;
 922
 923                 len += ntohs(grec->grec_nsrcs) * 4;
 924                 if (!pskb_may_pull(skb, len))
 925                         return -EINVAL;
 926
 927                 /* We treat this as an IGMPv2 report for now. */
 928                 switch (type) {
 929                 case IGMPV3_MODE_IS_INCLUDE:
 930                 case IGMPV3_MODE_IS_EXCLUDE:
 931                 case IGMPV3_CHANGE_TO_INCLUDE:
 932                 case IGMPV3_CHANGE_TO_EXCLUDE:
 933                 case IGMPV3_ALLOW_NEW_SOURCES:
 934                 case IGMPV3_BLOCK_OLD_SOURCES:
 935                         break;
 936
 937                 default:
 938                         continue;
 939                 }
 940
 941                 err = br_ip4_multicast_add_group(br, port, group, vid);
 942                 if (err)
 943                         break;
 944         }
 945
 946         return err;
 947 }
 948
 949 #if IS_ENABLED(CONFIG_IPV6)
 950 static int br_ip6_multicast_mld2_report(struct net_bridge *br,
 951                                         struct net_bridge_port *port,
 952                                         struct sk_buff *skb)
 953 {
 954         struct icmp6hdr *icmp6h;
 955         struct mld2_grec *grec;
 956         int i;
 957         int len;
 958         int num;
 959         int err = 0;
 960         u16 vid = 0;
 961
 962         if (!pskb_may_pull(skb, sizeof(*icmp6h)))
 963                 return -EINVAL;
 964
 965         br_vlan_get_tag(skb, &vid);
 966         icmp6h = icmp6_hdr(skb);
 967         num = ntohs(icmp6h->icmp6_dataun.un_data16[1]);
 968         len = sizeof(*icmp6h);
 969
 970         for (i = 0; i < num; i++) {
 971                 __be16 *nsrcs, _nsrcs;
 972
 973                 nsrcs = skb_header_pointer(skb,
 974                                            len + offsetof(struct mld2_grec,
 975                                                           grec_nsrcs),
 976                                            sizeof(_nsrcs), &_nsrcs);
 977                 if (!nsrcs)
 978                         return -EINVAL;
 979
 980                 if (!pskb_may_pull(skb,
 981                                    len + sizeof(*grec) +
 982                                    sizeof(struct in6_addr) * ntohs(*nsrcs)))
 983                         return -EINVAL;
 984
 985                 grec = (struct mld2_grec *)(skb->data + len);
 986                 len += sizeof(*grec) +
 987                        sizeof(struct in6_addr) * ntohs(*nsrcs);
 988
 989                 /* We treat these as MLDv1 reports for now. */
 990                 switch (grec->grec_type) {
 991                 case MLD2_MODE_IS_INCLUDE:
 992                 case MLD2_MODE_IS_EXCLUDE:
 993                 case MLD2_CHANGE_TO_INCLUDE:
 994                 case MLD2_CHANGE_TO_EXCLUDE:
 995                 case MLD2_ALLOW_NEW_SOURCES:
 996                 case MLD2_BLOCK_OLD_SOURCES:
 997                         break;
 998
 999                 default:
1000                         continue;
1001                 }
1002
1003                 err = br_ip6_multicast_add_group(br, port, &grec->grec_mca,
1004                                                  vid);
1005                 if (!err)
1006                         break;
1007         }
1008
1009         return err;
1010 }
1011 #endif
1012
1013 /*
1014  * Add port to rotuer_list
1015  *  list is maintained ordered by pointer value
1016  *  and locked by br->multicast_lock and RCU
1017  */
1018 static void br_multicast_add_router(struct net_bridge *br,
1019                                     struct net_bridge_port *port)
1020 {
1021         struct net_bridge_port *p;
1022         struct hlist_node *slot = NULL;
1023
1024         hlist_for_each_entry(p, &br->router_list, rlist) {
1025                 if ((unsigned long) port >= (unsigned long) p)
1026                         break;
1027                 slot = &p->rlist;
1028         }
1029
1030         if (slot)
1031                 hlist_add_after_rcu(slot, &port->rlist);
1032         else
1033                 hlist_add_head_rcu(&port->rlist, &br->router_list);
1034 }
1035
1036 static void br_multicast_mark_router(struct net_bridge *br,
1037                                      struct net_bridge_port *port)
1038 {
1039         unsigned long now = jiffies;
1040
1041         if (!port) {
1042                 if (br->multicast_router == 1)
1043                         mod_timer(&br->multicast_router_timer,
1044                                   now + br->multicast_querier_interval);
1045                 return;
1046         }
1047
1048         if (port->multicast_router != 1)
1049                 return;
1050
1051         if (!hlist_unhashed(&port->rlist))
1052                 goto timer;
1053
1054         br_multicast_add_router(br, port);
1055
1056 timer:
1057         mod_timer(&port->multicast_router_timer,
1058                   now + br->multicast_querier_interval);
1059 }
1060
1061 static void br_multicast_query_received(struct net_bridge *br,
1062                                         struct net_bridge_port *port,
1063                                         int saddr)
1064 {
1065         if (saddr)
1066                 mod_timer(&br->multicast_querier_timer,
1067                           jiffies + br->multicast_querier_interval);
1068         else if (timer_pending(&br->multicast_querier_timer))
1069                 return;
1070
1071         br_multicast_mark_router(br, port);
1072 }
1073
1074 static int br_ip4_multicast_query(struct net_bridge *br,
1075                                   struct net_bridge_port *port,
1076                                   struct sk_buff *skb)
1077 {
1078         const struct iphdr *iph = ip_hdr(skb);
1079         struct igmphdr *ih = igmp_hdr(skb);
1080         struct net_bridge_mdb_entry *mp;
1081         struct igmpv3_query *ih3;
1082         struct net_bridge_port_group *p;
1083         struct net_bridge_port_group __rcu **pp;
1084         unsigned long max_delay;
1085         unsigned long now = jiffies;
1086         __be32 group;
1087         int err = 0;
1088         u16 vid = 0;
1089
1090         spin_lock(&br->multicast_lock);
1091         if (!netif_running(br->dev) ||
1092             (port && port->state == BR_STATE_DISABLED))
1093                 goto out;
1094
1095         br_multicast_query_received(br, port, !!iph->saddr);
1096
1097         group = ih->group;
1098
1099         if (skb->len == sizeof(*ih)) {
1100                 max_delay = ih->code * (HZ / IGMP_TIMER_SCALE);
1101
1102                 if (!max_delay) {
1103                         max_delay = 10 * HZ;
1104                         group = 0;
1105                 }
1106         } else {
1107                 if (!pskb_may_pull(skb, sizeof(struct igmpv3_query))) {
1108                         err = -EINVAL;
1109                         goto out;
1110                 }
1111
1112                 ih3 = igmpv3_query_hdr(skb);
1113                 if (ih3->nsrcs)
1114                         goto out;
1115
1116                 max_delay = ih3->code ?
1117                             IGMPV3_MRC(ih3->code) * (HZ / IGMP_TIMER_SCALE) : 1;
1118         }
1119
1120         if (!group)
1121                 goto out;
1122
1123         br_vlan_get_tag(skb, &vid);
1124         mp = br_mdb_ip4_get(mlock_dereference(br->mdb, br), group, vid);
1125         if (!mp)
1126                 goto out;
1127
1128         setup_timer(&mp->timer, br_multicast_group_expired, (unsigned long)mp);
1129         mod_timer(&mp->timer, now + br->multicast_membership_interval);
1130         mp->timer_armed = true;
1131
1132         max_delay *= br->multicast_last_member_count;
1133
1134         if (mp->mglist &&
1135             (timer_pending(&mp->timer) ?
1136              time_after(mp->timer.expires, now + max_delay) :
1137              try_to_del_timer_sync(&mp->timer) >= 0))
1138                 mod_timer(&mp->timer, now + max_delay);
1139
1140         for (pp = &mp->ports;
1141              (p = mlock_dereference(*pp, br)) != NULL;
1142              pp = &p->next) {
1143                 if (timer_pending(&p->timer) ?
1144                     time_after(p->timer.expires, now + max_delay) :
1145                     try_to_del_timer_sync(&p->timer) >= 0)
1146                         mod_timer(&p->timer, now + max_delay);
1147         }
1148
1149 out:
1150         spin_unlock(&br->multicast_lock);
1151         return err;
1152 }
1153
1154 #if IS_ENABLED(CONFIG_IPV6)
1155 static int br_ip6_multicast_query(struct net_bridge *br,
1156                                   struct net_bridge_port *port,
1157                                   struct sk_buff *skb)
1158 {
1159         const struct ipv6hdr *ip6h = ipv6_hdr(skb);
1160         struct mld_msg *mld;
1161         struct net_bridge_mdb_entry *mp;
1162         struct mld2_query *mld2q;
1163         struct net_bridge_port_group *p;
1164         struct net_bridge_port_group __rcu **pp;
1165         unsigned long max_delay;
1166         unsigned long now = jiffies;
1167         const struct in6_addr *group = NULL;
1168         int err = 0;
1169         u16 vid = 0;
1170
1171         spin_lock(&br->multicast_lock);
1172         if (!netif_running(br->dev) ||
1173             (port && port->state == BR_STATE_DISABLED))
1174                 goto out;
1175
1176         br_multicast_query_received(br, port, !ipv6_addr_any(&ip6h->saddr));
1177
1178         if (skb->len == sizeof(*mld)) {
1179                 if (!pskb_may_pull(skb, sizeof(*mld))) {
1180                         err = -EINVAL;
1181                         goto out;
1182                 }
1183                 mld = (struct mld_msg *) icmp6_hdr(skb);
1184                 max_delay = msecs_to_jiffies(ntohs(mld->mld_maxdelay));
1185                 if (max_delay)
1186                         group = &mld->mld_mca;
1187         } else if (skb->len >= sizeof(*mld2q)) {
1188                 if (!pskb_may_pull(skb, sizeof(*mld2q))) {
1189                         err = -EINVAL;
1190                         goto out;
1191                 }
1192                 mld2q = (struct mld2_query *)icmp6_hdr(skb);
1193                 if (!mld2q->mld2q_nsrcs)
1194                         group = &mld2q->mld2q_mca;
1195                 max_delay = mld2q->mld2q_mrc ? MLDV2_MRC(ntohs(mld2q->mld2q_mrc)) : 1;
1196         }
1197
1198         if (!group)
1199                 goto out;
1200
1201         br_vlan_get_tag(skb, &vid);
1202         mp = br_mdb_ip6_get(mlock_dereference(br->mdb, br), group, vid);
1203         if (!mp)
1204                 goto out;
1205
1206         setup_timer(&mp->timer, br_multicast_group_expired, (unsigned long)mp);
1207         mod_timer(&mp->timer, now + br->multicast_membership_interval);
1208         mp->timer_armed = true;
1209
1210         max_delay *= br->multicast_last_member_count;
1211         if (mp->mglist &&
1212             (timer_pending(&mp->timer) ?
1213              time_after(mp->timer.expires, now + max_delay) :
1214              try_to_del_timer_sync(&mp->timer) >= 0))
1215                 mod_timer(&mp->timer, now + max_delay);
1216
1217         for (pp = &mp->ports;
1218              (p = mlock_dereference(*pp, br)) != NULL;
1219              pp = &p->next) {
1220                 if (timer_pending(&p->timer) ?
1221                     time_after(p->timer.expires, now + max_delay) :
1222                     try_to_del_timer_sync(&p->timer) >= 0)
1223                         mod_timer(&p->timer, now + max_delay);
1224         }
1225
1226 out:
1227         spin_unlock(&br->multicast_lock);
1228         return err;
1229 }
1230 #endif
1231
1232 static void br_multicast_leave_group(struct net_bridge *br,
1233                                      struct net_bridge_port *port,
1234                                      struct br_ip *group)
1235 {
1236         struct net_bridge_mdb_htable *mdb;
1237         struct net_bridge_mdb_entry *mp;
1238         struct net_bridge_port_group *p;
1239         unsigned long now;
1240         unsigned long time;
1241
1242         spin_lock(&br->multicast_lock);
1243         if (!netif_running(br->dev) ||
1244             (port && port->state == BR_STATE_DISABLED) ||
1245             timer_pending(&br->multicast_querier_timer))
1246                 goto out;
1247
1248         mdb = mlock_dereference(br->mdb, br);
1249         mp = br_mdb_ip_get(mdb, group);
1250         if (!mp)
1251                 goto out;
1252
1253         if (br->multicast_querier &&
1254             !timer_pending(&br->multicast_querier_timer)) {
1255                 __br_multicast_send_query(br, port, &mp->addr);
1256
1257                 time = jiffies + br->multicast_last_member_count *
1258                                  br->multicast_last_member_interval;
1259                 mod_timer(port ? &port->multicast_query_timer :
1260                                  &br->multicast_query_timer, time);
1261
1262                 for (p = mlock_dereference(mp->ports, br);
1263                      p != NULL;
1264                      p = mlock_dereference(p->next, br)) {
1265                         if (p->port != port)
1266                                 continue;
1267
1268                         if (!hlist_unhashed(&p->mglist) &&
1269                             (timer_pending(&p->timer) ?
1270                              time_after(p->timer.expires, time) :
1271                              try_to_del_timer_sync(&p->timer) >= 0)) {
1272                                 mod_timer(&p->timer, time);
1273                         }
1274
1275                         break;
1276                 }
1277         }
1278
1279         if (port && (port->flags & BR_MULTICAST_FAST_LEAVE)) {
1280                 struct net_bridge_port_group __rcu **pp;
1281
1282                 for (pp = &mp->ports;
1283                      (p = mlock_dereference(*pp, br)) != NULL;
1284                      pp = &p->next) {
1285                         if (p->port != port)
1286                                 continue;
1287
1288                         rcu_assign_pointer(*pp, p->next);
1289                         hlist_del_init(&p->mglist);
1290                         del_timer(&p->timer);
1291                         call_rcu_bh(&p->rcu, br_multicast_free_pg);
1292                         br_mdb_notify(br->dev, port, group, RTM_DELMDB);
1293
1294                         if (!mp->ports && !mp->mglist && mp->timer_armed &&
1295                             netif_running(br->dev))
1296                                 mod_timer(&mp->timer, jiffies);
1297                 }
1298                 goto out;
1299         }
1300
1301         now = jiffies;
1302         time = now + br->multicast_last_member_count *
1303                      br->multicast_last_member_interval;
1304
1305         if (!port) {
1306                 if (mp->mglist && mp->timer_armed &&
1307                     (timer_pending(&mp->timer) ?
1308                      time_after(mp->timer.expires, time) :
1309                      try_to_del_timer_sync(&mp->timer) >= 0)) {
1310                         mod_timer(&mp->timer, time);
1311                 }
1312         }
1313
1314 out:
1315         spin_unlock(&br->multicast_lock);
1316 }
1317
1318 static void br_ip4_multicast_leave_group(struct net_bridge *br,
1319                                          struct net_bridge_port *port,
1320                                          __be32 group,
1321                                          __u16 vid)
1322 {
1323         struct br_ip br_group;
1324
1325         if (ipv4_is_local_multicast(group))
1326                 return;
1327
1328         br_group.u.ip4 = group;
1329         br_group.proto = htons(ETH_P_IP);
1330         br_group.vid = vid;
1331
1332         br_multicast_leave_group(br, port, &br_group);
1333 }
1334
1335 #if IS_ENABLED(CONFIG_IPV6)
1336 static void br_ip6_multicast_leave_group(struct net_bridge *br,
1337                                          struct net_bridge_port *port,
1338                                          const struct in6_addr *group,
1339                                          __u16 vid)
1340 {
1341         struct br_ip br_group;
1342
1343         if (!ipv6_is_transient_multicast(group))
1344                 return;
1345
1346         br_group.u.ip6 = *group;
1347         br_group.proto = htons(ETH_P_IPV6);
1348         br_group.vid = vid;
1349
1350         br_multicast_leave_group(br, port, &br_group);
1351 }
1352 #endif
1353
1354 static int br_multicast_ipv4_rcv(struct net_bridge *br,
1355                                  struct net_bridge_port *port,
1356                                  struct sk_buff *skb)
1357 {
1358         struct sk_buff *skb2 = skb;
1359         const struct iphdr *iph;
1360         struct igmphdr *ih;
1361         unsigned int len;
1362         unsigned int offset;
1363         int err;
1364         u16 vid = 0;
1365
1366         /* We treat OOM as packet loss for now. */
1367         if (!pskb_may_pull(skb, sizeof(*iph)))
1368                 return -EINVAL;
1369
1370         iph = ip_hdr(skb);
1371
1372         if (iph->ihl < 5 || iph->version != 4)
1373                 return -EINVAL;
1374
1375         if (!pskb_may_pull(skb, ip_hdrlen(skb)))
1376                 return -EINVAL;
1377
1378         iph = ip_hdr(skb);
1379
1380         if (unlikely(ip_fast_csum((u8 *)iph, iph->ihl)))
1381                 return -EINVAL;
1382
1383         if (iph->protocol != IPPROTO_IGMP) {
1384                 if (!ipv4_is_local_multicast(iph->daddr))
1385                         BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1386                 return 0;
1387         }
1388
1389         len = ntohs(iph->tot_len);
1390         if (skb->len < len || len < ip_hdrlen(skb))
1391                 return -EINVAL;
1392
1393         if (skb->len > len) {
1394                 skb2 = skb_clone(skb, GFP_ATOMIC);
1395                 if (!skb2)
1396                         return -ENOMEM;
1397
1398                 err = pskb_trim_rcsum(skb2, len);
1399                 if (err)
1400                         goto err_out;
1401         }
1402
1403         len -= ip_hdrlen(skb2);
1404         offset = skb_network_offset(skb2) + ip_hdrlen(skb2);
1405         __skb_pull(skb2, offset);
1406         skb_reset_transport_header(skb2);
1407
1408         err = -EINVAL;
1409         if (!pskb_may_pull(skb2, sizeof(*ih)))
1410                 goto out;
1411
1412         switch (skb2->ip_summed) {
1413         case CHECKSUM_COMPLETE:
1414                 if (!csum_fold(skb2->csum))
1415                         break;
1416                 /* fall through */
1417         case CHECKSUM_NONE:
1418                 skb2->csum = 0;
1419                 if (skb_checksum_complete(skb2))
1420                         goto out;
1421         }
1422
1423         err = 0;
1424
1425         br_vlan_get_tag(skb2, &vid);
1426         BR_INPUT_SKB_CB(skb)->igmp = 1;
1427         ih = igmp_hdr(skb2);
1428
1429         switch (ih->type) {
1430         case IGMP_HOST_MEMBERSHIP_REPORT:
1431         case IGMPV2_HOST_MEMBERSHIP_REPORT:
1432                 BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1433                 err = br_ip4_multicast_add_group(br, port, ih->group, vid);
1434                 break;
1435         case IGMPV3_HOST_MEMBERSHIP_REPORT:
1436                 err = br_ip4_multicast_igmp3_report(br, port, skb2);
1437                 break;
1438         case IGMP_HOST_MEMBERSHIP_QUERY:
1439                 err = br_ip4_multicast_query(br, port, skb2);
1440                 break;
1441         case IGMP_HOST_LEAVE_MESSAGE:
1442                 br_ip4_multicast_leave_group(br, port, ih->group, vid);
1443                 break;
1444         }
1445
1446 out:
1447         __skb_push(skb2, offset);
1448 err_out:
1449         if (skb2 != skb)
1450                 kfree_skb(skb2);
1451         return err;
1452 }
1453
1454 #if IS_ENABLED(CONFIG_IPV6)
1455 static int br_multicast_ipv6_rcv(struct net_bridge *br,
1456                                  struct net_bridge_port *port,
1457                                  struct sk_buff *skb)
1458 {
1459         struct sk_buff *skb2;
1460         const struct ipv6hdr *ip6h;
1461         u8 icmp6_type;
1462         u8 nexthdr;
1463         __be16 frag_off;
1464         unsigned int len;
1465         int offset;
1466         int err;
1467         u16 vid = 0;
1468
1469         if (!pskb_may_pull(skb, sizeof(*ip6h)))
1470                 return -EINVAL;
1471
1472         ip6h = ipv6_hdr(skb);
1473
1474         /*
1475          * We're interested in MLD messages only.
1476          *  - Version is 6
1477          *  - MLD has always Router Alert hop-by-hop option
1478          *  - But we do not support jumbrograms.
1479          */
1480         if (ip6h->version != 6 ||
1481             ip6h->nexthdr != IPPROTO_HOPOPTS ||
1482             ip6h->payload_len == 0)
1483                 return 0;
1484
1485         len = ntohs(ip6h->payload_len) + sizeof(*ip6h);
1486         if (skb->len < len)
1487                 return -EINVAL;
1488
1489         nexthdr = ip6h->nexthdr;
1490         offset = ipv6_skip_exthdr(skb, sizeof(*ip6h), &nexthdr, &frag_off);
1491
1492         if (offset < 0 || nexthdr != IPPROTO_ICMPV6)
1493                 return 0;
1494
1495         /* Okay, we found ICMPv6 header */
1496         skb2 = skb_clone(skb, GFP_ATOMIC);
1497         if (!skb2)
1498                 return -ENOMEM;
1499
1500         err = -EINVAL;
1501         if (!pskb_may_pull(skb2, offset + sizeof(struct icmp6hdr)))
1502                 goto out;
1503
1504         len -= offset - skb_network_offset(skb2);
1505
1506         __skb_pull(skb2, offset);
1507         skb_reset_transport_header(skb2);
1508         skb_postpull_rcsum(skb2, skb_network_header(skb2),
1509                            skb_network_header_len(skb2));
1510
1511         icmp6_type = icmp6_hdr(skb2)->icmp6_type;
1512
1513         switch (icmp6_type) {
1514         case ICMPV6_MGM_QUERY:
1515         case ICMPV6_MGM_REPORT:
1516         case ICMPV6_MGM_REDUCTION:
1517         case ICMPV6_MLD2_REPORT:
1518                 break;
1519         default:
1520                 err = 0;
1521                 goto out;
1522         }
1523
1524         /* Okay, we found MLD message. Check further. */
1525         if (skb2->len > len) {
1526                 err = pskb_trim_rcsum(skb2, len);
1527                 if (err)
1528                         goto out;
1529                 err = -EINVAL;
1530         }
1531
1532         ip6h = ipv6_hdr(skb2);
1533
1534         switch (skb2->ip_summed) {
1535         case CHECKSUM_COMPLETE:
1536                 if (!csum_ipv6_magic(&ip6h->saddr, &ip6h->daddr, skb2->len,
1537                                         IPPROTO_ICMPV6, skb2->csum))
1538                         break;
1539                 /*FALLTHROUGH*/
1540         case CHECKSUM_NONE:
1541                 skb2->csum = ~csum_unfold(csum_ipv6_magic(&ip6h->saddr,
1542                                                         &ip6h->daddr,
1543                                                         skb2->len,
1544                                                         IPPROTO_ICMPV6, 0));
1545                 if (__skb_checksum_complete(skb2))
1546                         goto out;
1547         }
1548
1549         err = 0;
1550
1551         br_vlan_get_tag(skb, &vid);
1552         BR_INPUT_SKB_CB(skb)->igmp = 1;
1553
1554         switch (icmp6_type) {
1555         case ICMPV6_MGM_REPORT:
1556             {
1557                 struct mld_msg *mld;
1558                 if (!pskb_may_pull(skb2, sizeof(*mld))) {
1559                         err = -EINVAL;
1560                         goto out;
1561                 }
1562                 mld = (struct mld_msg *)skb_transport_header(skb2);
1563                 BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1564                 err = br_ip6_multicast_add_group(br, port, &mld->mld_mca, vid);
1565                 break;
1566             }
1567         case ICMPV6_MLD2_REPORT:
1568                 err = br_ip6_multicast_mld2_report(br, port, skb2);
1569                 break;
1570         case ICMPV6_MGM_QUERY:
1571                 err = br_ip6_multicast_query(br, port, skb2);
1572                 break;
1573         case ICMPV6_MGM_REDUCTION:
1574             {
1575                 struct mld_msg *mld;
1576                 if (!pskb_may_pull(skb2, sizeof(*mld))) {
1577                         err = -EINVAL;
1578                         goto out;
1579                 }
1580                 mld = (struct mld_msg *)skb_transport_header(skb2);
1581                 br_ip6_multicast_leave_group(br, port, &mld->mld_mca, vid);
1582             }
1583         }
1584
1585 out:
1586         kfree_skb(skb2);
1587         return err;
1588 }
1589 #endif
1590
1591 int br_multicast_rcv(struct net_bridge *br, struct net_bridge_port *port,
1592                      struct sk_buff *skb)
1593 {
1594         BR_INPUT_SKB_CB(skb)->igmp = 0;
1595         BR_INPUT_SKB_CB(skb)->mrouters_only = 0;
1596
1597         if (br->multicast_disabled)
1598                 return 0;
1599
1600         switch (skb->protocol) {
1601         case htons(ETH_P_IP):
1602                 return br_multicast_ipv4_rcv(br, port, skb);
1603 #if IS_ENABLED(CONFIG_IPV6)
1604         case htons(ETH_P_IPV6):
1605                 return br_multicast_ipv6_rcv(br, port, skb);
1606 #endif
1607         }
1608
1609         return 0;
1610 }
1611
1612 static void br_multicast_query_expired(unsigned long data)
1613 {
1614         struct net_bridge *br = (void *)data;
1615
1616         spin_lock(&br->multicast_lock);
1617         if (br->multicast_startup_queries_sent <
1618             br->multicast_startup_query_count)
1619                 br->multicast_startup_queries_sent++;
1620
1621         br_multicast_send_query(br, NULL, br->multicast_startup_queries_sent);
1622
1623         spin_unlock(&br->multicast_lock);
1624 }
1625
1626 void br_multicast_init(struct net_bridge *br)
1627 {
1628         br->hash_elasticity = 4;
1629         br->hash_max = 512;
1630
1631         br->multicast_router = 1;
1632         br->multicast_querier = 0;
1633         br->multicast_query_use_ifaddr = 0;
1634         br->multicast_last_member_count = 2;
1635         br->multicast_startup_query_count = 2;
1636
1637         br->multicast_last_member_interval = HZ;
1638         br->multicast_query_response_interval = 10 * HZ;
1639         br->multicast_startup_query_interval = 125 * HZ / 4;
1640         br->multicast_query_interval = 125 * HZ;
1641         br->multicast_querier_interval = 255 * HZ;
1642         br->multicast_membership_interval = 260 * HZ;
1643
1644         spin_lock_init(&br->multicast_lock);
1645         setup_timer(&br->multicast_router_timer,
1646                     br_multicast_local_router_expired, 0);
1647         setup_timer(&br->multicast_querier_timer,
1648                     br_multicast_querier_expired, (unsigned long)br);
1649         setup_timer(&br->multicast_query_timer, br_multicast_query_expired,
1650                     (unsigned long)br);
1651 }
1652
1653 void br_multicast_open(struct net_bridge *br)
1654 {
1655         br->multicast_startup_queries_sent = 0;
1656
1657         if (br->multicast_disabled)
1658                 return;
1659
1660         mod_timer(&br->multicast_query_timer, jiffies);
1661 }
1662
1663 void br_multicast_stop(struct net_bridge *br)
1664 {
1665         struct net_bridge_mdb_htable *mdb;
1666         struct net_bridge_mdb_entry *mp;
1667         struct hlist_node *n;
1668         u32 ver;
1669         int i;
1670
1671         del_timer_sync(&br->multicast_router_timer);
1672         del_timer_sync(&br->multicast_querier_timer);
1673         del_timer_sync(&br->multicast_query_timer);
1674
1675         spin_lock_bh(&br->multicast_lock);
1676         mdb = mlock_dereference(br->mdb, br);
1677         if (!mdb)
1678                 goto out;
1679
1680         br->mdb = NULL;
1681
1682         ver = mdb->ver;
1683         for (i = 0; i < mdb->max; i++) {
1684                 hlist_for_each_entry_safe(mp, n, &mdb->mhash[i],
1685                                           hlist[ver]) {
1686                         del_timer(&mp->timer);
1687                         mp->timer_armed = false;
1688                         call_rcu_bh(&mp->rcu, br_multicast_free_group);
1689                 }
1690         }
1691
1692         if (mdb->old) {
1693                 spin_unlock_bh(&br->multicast_lock);
1694                 rcu_barrier_bh();
1695                 spin_lock_bh(&br->multicast_lock);
1696                 WARN_ON(mdb->old);
1697         }
1698
1699         mdb->old = mdb;
1700         call_rcu_bh(&mdb->rcu, br_mdb_free);
1701
1702 out:
1703         spin_unlock_bh(&br->multicast_lock);
1704 }
1705
1706 int br_multicast_set_router(struct net_bridge *br, unsigned long val)
1707 {
1708         int err = -ENOENT;
1709
1710         spin_lock_bh(&br->multicast_lock);
1711         if (!netif_running(br->dev))
1712                 goto unlock;
1713
1714         switch (val) {
1715         case 0:
1716         case 2:
1717                 del_timer(&br->multicast_router_timer);
1718                 /* fall through */
1719         case 1:
1720                 br->multicast_router = val;
1721                 err = 0;
1722                 break;
1723
1724         default:
1725                 err = -EINVAL;
1726                 break;
1727         }
1728
1729 unlock:
1730         spin_unlock_bh(&br->multicast_lock);
1731
1732         return err;
1733 }
1734
1735 int br_multicast_set_port_router(struct net_bridge_port *p, unsigned long val)
1736 {
1737         struct net_bridge *br = p->br;
1738         int err = -ENOENT;
1739
1740         spin_lock(&br->multicast_lock);
1741         if (!netif_running(br->dev) || p->state == BR_STATE_DISABLED)
1742                 goto unlock;
1743
1744         switch (val) {
1745         case 0:
1746         case 1:
1747         case 2:
1748                 p->multicast_router = val;
1749                 err = 0;
1750
1751                 if (val < 2 && !hlist_unhashed(&p->rlist))
1752                         hlist_del_init_rcu(&p->rlist);
1753
1754                 if (val == 1)
1755                         break;
1756
1757                 del_timer(&p->multicast_router_timer);
1758
1759                 if (val == 0)
1760                         break;
1761
1762                 br_multicast_add_router(br, p);
1763                 break;
1764
1765         default:
1766                 err = -EINVAL;
1767                 break;
1768         }
1769
1770 unlock:
1771         spin_unlock(&br->multicast_lock);
1772
1773         return err;
1774 }
1775
1776 static void br_multicast_start_querier(struct net_bridge *br)
1777 {
1778         struct net_bridge_port *port;
1779
1780         br_multicast_open(br);
1781
1782         list_for_each_entry(port, &br->port_list, list) {
1783                 if (port->state == BR_STATE_DISABLED ||
1784                     port->state == BR_STATE_BLOCKING)
1785                         continue;
1786
1787                 __br_multicast_enable_port(port);
1788         }
1789 }
1790
1791 int br_multicast_toggle(struct net_bridge *br, unsigned long val)
1792 {
1793         int err = 0;
1794         struct net_bridge_mdb_htable *mdb;
1795
1796         spin_lock_bh(&br->multicast_lock);
1797         if (br->multicast_disabled == !val)
1798                 goto unlock;
1799
1800         br->multicast_disabled = !val;
1801         if (br->multicast_disabled)
1802                 goto unlock;
1803
1804         if (!netif_running(br->dev))
1805                 goto unlock;
1806
1807         mdb = mlock_dereference(br->mdb, br);
1808         if (mdb) {
1809                 if (mdb->old) {
1810                         err = -EEXIST;
1811 rollback:
1812                         br->multicast_disabled = !!val;
1813                         goto unlock;
1814                 }
1815
1816                 err = br_mdb_rehash(&br->mdb, mdb->max,
1817                                     br->hash_elasticity);
1818                 if (err)
1819                         goto rollback;
1820         }
1821
1822         br_multicast_start_querier(br);
1823
1824 unlock:
1825         spin_unlock_bh(&br->multicast_lock);
1826
1827         return err;
1828 }
1829
1830 int br_multicast_set_querier(struct net_bridge *br, unsigned long val)
1831 {
1832         val = !!val;
1833
1834         spin_lock_bh(&br->multicast_lock);
1835         if (br->multicast_querier == val)
1836                 goto unlock;
1837
1838         br->multicast_querier = val;
1839         if (val)
1840                 br_multicast_start_querier(br);
1841
1842 unlock:
1843         spin_unlock_bh(&br->multicast_lock);
1844
1845         return 0;
1846 }
1847
1848 int br_multicast_set_hash_max(struct net_bridge *br, unsigned long val)
1849 {
1850         int err = -ENOENT;
1851         u32 old;
1852         struct net_bridge_mdb_htable *mdb;
1853
1854         spin_lock(&br->multicast_lock);
1855         if (!netif_running(br->dev))
1856                 goto unlock;
1857
1858         err = -EINVAL;
1859         if (!is_power_of_2(val))
1860                 goto unlock;
1861
1862         mdb = mlock_dereference(br->mdb, br);
1863         if (mdb && val < mdb->size)
1864                 goto unlock;
1865
1866         err = 0;
1867
1868         old = br->hash_max;
1869         br->hash_max = val;
1870
1871         if (mdb) {
1872                 if (mdb->old) {
1873                         err = -EEXIST;
1874 rollback:
1875                         br->hash_max = old;
1876                         goto unlock;
1877                 }
1878
1879                 err = br_mdb_rehash(&br->mdb, br->hash_max,
1880                                     br->hash_elasticity);
1881                 if (err)
1882                         goto rollback;
1883         }
1884
1885 unlock:
1886         spin_unlock(&br->multicast_lock);
1887
1888         return err;
1889 }