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 && mp->timer_armed &&
 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) ?
 471                         br->multicast_query_response_interval :
 472                         br->multicast_last_member_interval;
 473
 474         mldq->mld_type = ICMPV6_MGM_QUERY;
 475         mldq->mld_code = 0;
 476         mldq->mld_cksum = 0;
 477         mldq->mld_maxdelay = htons((u16)jiffies_to_msecs(interval));
 478         mldq->mld_reserved = 0;
 479         mldq->mld_mca = *group;
 480
 481         /* checksum */
 482         mldq->mld_cksum = csum_ipv6_magic(&ip6h->saddr, &ip6h->daddr,
 483                                           sizeof(*mldq), IPPROTO_ICMPV6,
 484                                           csum_partial(mldq,
 485                                                        sizeof(*mldq), 0));
 486         skb_put(skb, sizeof(*mldq));
 487
 488         __skb_pull(skb, sizeof(*eth));
 489
 490 out:
 491         return skb;
 492 }
 493 #endif
 494
 495 static struct sk_buff *br_multicast_alloc_query(struct net_bridge *br,
 496                                                 struct br_ip *addr)
 497 {
 498         switch (addr->proto) {
 499         case htons(ETH_P_IP):
 500                 return br_ip4_multicast_alloc_query(br, addr->u.ip4);
 501 #if IS_ENABLED(CONFIG_IPV6)
 502         case htons(ETH_P_IPV6):
 503                 return br_ip6_multicast_alloc_query(br, &addr->u.ip6);
 504 #endif
 505         }
 506         return NULL;
 507 }
 508
 509 static struct net_bridge_mdb_entry *br_multicast_get_group(
 510         struct net_bridge *br, struct net_bridge_port *port,
 511         struct br_ip *group, int hash)
 512 {
 513         struct net_bridge_mdb_htable *mdb;
 514         struct net_bridge_mdb_entry *mp;
 515         unsigned int count = 0;
 516         unsigned int max;
 517         int elasticity;
 518         int err;
 519
 520         mdb = rcu_dereference_protected(br->mdb, 1);
 521         hlist_for_each_entry(mp, &mdb->mhash[hash], hlist[mdb->ver]) {
 522                 count++;
 523                 if (unlikely(br_ip_equal(group, &mp->addr)))
 524                         return mp;
 525         }
 526
 527         elasticity = 0;
 528         max = mdb->max;
 529
 530         if (unlikely(count > br->hash_elasticity && count)) {
 531                 if (net_ratelimit())
 532                         br_info(br, "Multicast hash table "
 533                                 "chain limit reached: %s\n",
 534                                 port ? port->dev->name : br->dev->name);
 535
 536                 elasticity = br->hash_elasticity;
 537         }
 538
 539         if (mdb->size >= max) {
 540                 max *= 2;
 541                 if (unlikely(max > br->hash_max)) {
 542                         br_warn(br, "Multicast hash table maximum of %d "
 543                                 "reached, disabling snooping: %s\n",
 544                                 br->hash_max,
 545                                 port ? port->dev->name : br->dev->name);
 546                         err = -E2BIG;
 547 disable:
 548                         br->multicast_disabled = 1;
 549                         goto err;
 550                 }
 551         }
 552
 553         if (max > mdb->max || elasticity) {
 554                 if (mdb->old) {
 555                         if (net_ratelimit())
 556                                 br_info(br, "Multicast hash table "
 557                                         "on fire: %s\n",
 558                                         port ? port->dev->name : br->dev->name);
 559                         err = -EEXIST;
 560                         goto err;
 561                 }
 562
 563                 err = br_mdb_rehash(&br->mdb, max, elasticity);
 564                 if (err) {
 565                         br_warn(br, "Cannot rehash multicast "
 566                                 "hash table, disabling snooping: %s, %d, %d\n",
 567                                 port ? port->dev->name : br->dev->name,
 568                                 mdb->size, err);
 569                         goto disable;
 570                 }
 571
 572                 err = -EAGAIN;
 573                 goto err;
 574         }
 575
 576         return NULL;
 577
 578 err:
 579         mp = ERR_PTR(err);
 580         return mp;
 581 }
 582
 583 struct net_bridge_mdb_entry *br_multicast_new_group(struct net_bridge *br,
 584         struct net_bridge_port *port, struct br_ip *group)
 585 {
 586         struct net_bridge_mdb_htable *mdb;
 587         struct net_bridge_mdb_entry *mp;
 588         int hash;
 589         int err;
 590
 591         mdb = rcu_dereference_protected(br->mdb, 1);
 592         if (!mdb) {
 593                 err = br_mdb_rehash(&br->mdb, BR_HASH_SIZE, 0);
 594                 if (err)
 595                         return ERR_PTR(err);
 596                 goto rehash;
 597         }
 598
 599         hash = br_ip_hash(mdb, group);
 600         mp = br_multicast_get_group(br, port, group, hash);
 601         switch (PTR_ERR(mp)) {
 602         case 0:
 603                 break;
 604
 605         case -EAGAIN:
 606 rehash:
 607                 mdb = rcu_dereference_protected(br->mdb, 1);
 608                 hash = br_ip_hash(mdb, group);
 609                 break;
 610
 611         default:
 612                 goto out;
 613         }
 614
 615         mp = kzalloc(sizeof(*mp), GFP_ATOMIC);
 616         if (unlikely(!mp))
 617                 return ERR_PTR(-ENOMEM);
 618
 619         mp->br = br;
 620         mp->addr = *group;
 621
 622         hlist_add_head_rcu(&mp->hlist[mdb->ver], &mdb->mhash[hash]);
 623         mdb->size++;
 624
 625 out:
 626         return mp;
 627 }
 628
 629 struct net_bridge_port_group *br_multicast_new_port_group(
 630                         struct net_bridge_port *port,
 631                         struct br_ip *group,
 632                         struct net_bridge_port_group __rcu *next,
 633                         unsigned char state)
 634 {
 635         struct net_bridge_port_group *p;
 636
 637         p = kzalloc(sizeof(*p), GFP_ATOMIC);
 638         if (unlikely(!p))
 639                 return NULL;
 640
 641         p->addr = *group;
 642         p->port = port;
 643         p->state = state;
 644         rcu_assign_pointer(p->next, next);
 645         hlist_add_head(&p->mglist, &port->mglist);
 646         setup_timer(&p->timer, br_multicast_port_group_expired,
 647                     (unsigned long)p);
 648         return p;
 649 }
 650
 651 static int br_multicast_add_group(struct net_bridge *br,
 652                                   struct net_bridge_port *port,
 653                                   struct br_ip *group)
 654 {
 655         struct net_bridge_mdb_entry *mp;
 656         struct net_bridge_port_group *p;
 657         struct net_bridge_port_group __rcu **pp;
 658         int err;
 659
 660         spin_lock(&br->multicast_lock);
 661         if (!netif_running(br->dev) ||
 662             (port && port->state == BR_STATE_DISABLED))
 663                 goto out;
 664
 665         mp = br_multicast_new_group(br, port, group);
 666         err = PTR_ERR(mp);
 667         if (IS_ERR(mp))
 668                 goto err;
 669
 670         if (!port) {
 671                 mp->mglist = true;
 672                 goto out;
 673         }
 674
 675         for (pp = &mp->ports;
 676              (p = mlock_dereference(*pp, br)) != NULL;
 677              pp = &p->next) {
 678                 if (p->port == port)
 679                         goto out;
 680                 if ((unsigned long)p->port < (unsigned long)port)
 681                         break;
 682         }
 683
 684         p = br_multicast_new_port_group(port, group, *pp, MDB_TEMPORARY);
 685         if (unlikely(!p))
 686                 goto err;
 687         rcu_assign_pointer(*pp, p);
 688         br_mdb_notify(br->dev, port, group, RTM_NEWMDB);
 689
 690 out:
 691         err = 0;
 692
 693 err:
 694         spin_unlock(&br->multicast_lock);
 695         return err;
 696 }
 697
 698 static int br_ip4_multicast_add_group(struct net_bridge *br,
 699                                       struct net_bridge_port *port,
 700                                       __be32 group,
 701                                       __u16 vid)
 702 {
 703         struct br_ip br_group;
 704
 705         if (ipv4_is_local_multicast(group))
 706                 return 0;
 707
 708         br_group.u.ip4 = group;
 709         br_group.proto = htons(ETH_P_IP);
 710         br_group.vid = vid;
 711
 712         return br_multicast_add_group(br, port, &br_group);
 713 }
 714
 715 #if IS_ENABLED(CONFIG_IPV6)
 716 static int br_ip6_multicast_add_group(struct net_bridge *br,
 717                                       struct net_bridge_port *port,
 718                                       const struct in6_addr *group,
 719                                       __u16 vid)
 720 {
 721         struct br_ip br_group;
 722
 723         if (!ipv6_is_transient_multicast(group))
 724                 return 0;
 725
 726         br_group.u.ip6 = *group;
 727         br_group.proto = htons(ETH_P_IPV6);
 728         br_group.vid = vid;
 729
 730         return br_multicast_add_group(br, port, &br_group);
 731 }
 732 #endif
 733
 734 static void br_multicast_router_expired(unsigned long data)
 735 {
 736         struct net_bridge_port *port = (void *)data;
 737         struct net_bridge *br = port->br;
 738
 739         spin_lock(&br->multicast_lock);
 740         if (port->multicast_router != 1 ||
 741             timer_pending(&port->multicast_router_timer) ||
 742             hlist_unhashed(&port->rlist))
 743                 goto out;
 744
 745         hlist_del_init_rcu(&port->rlist);
 746
 747 out:
 748         spin_unlock(&br->multicast_lock);
 749 }
 750
 751 static void br_multicast_local_router_expired(unsigned long data)
 752 {
 753 }
 754
 755 static void br_multicast_querier_expired(unsigned long data)
 756 {
 757         struct net_bridge *br = (void *)data;
 758
 759         spin_lock(&br->multicast_lock);
 760         if (!netif_running(br->dev) || br->multicast_disabled)
 761                 goto out;
 762
 763         br_multicast_start_querier(br);
 764
 765 out:
 766         spin_unlock(&br->multicast_lock);
 767 }
 768
 769 static void __br_multicast_send_query(struct net_bridge *br,
 770                                       struct net_bridge_port *port,
 771                                       struct br_ip *ip)
 772 {
 773         struct sk_buff *skb;
 774
 775         skb = br_multicast_alloc_query(br, ip);
 776         if (!skb)
 777                 return;
 778
 779         if (port) {
 780                 __skb_push(skb, sizeof(struct ethhdr));
 781                 skb->dev = port->dev;
 782                 NF_HOOK(NFPROTO_BRIDGE, NF_BR_LOCAL_OUT, skb, NULL, skb->dev,
 783                         dev_queue_xmit);
 784         } else
 785                 netif_rx(skb);
 786 }
 787
 788 static void br_multicast_send_query(struct net_bridge *br,
 789                                     struct net_bridge_port *port, u32 sent)
 790 {
 791         unsigned long time;
 792         struct br_ip br_group;
 793
 794         if (!netif_running(br->dev) || br->multicast_disabled ||
 795             !br->multicast_querier ||
 796             timer_pending(&br->multicast_querier_timer))
 797                 return;
 798
 799         memset(&br_group.u, 0, sizeof(br_group.u));
 800
 801         br_group.proto = htons(ETH_P_IP);
 802         __br_multicast_send_query(br, port, &br_group);
 803
 804 #if IS_ENABLED(CONFIG_IPV6)
 805         br_group.proto = htons(ETH_P_IPV6);
 806         __br_multicast_send_query(br, port, &br_group);
 807 #endif
 808
 809         time = jiffies;
 810         time += sent < br->multicast_startup_query_count ?
 811                 br->multicast_startup_query_interval :
 812                 br->multicast_query_interval;
 813         mod_timer(port ? &port->multicast_query_timer :
 814                          &br->multicast_query_timer, time);
 815 }
 816
 817 static void br_multicast_port_query_expired(unsigned long data)
 818 {
 819         struct net_bridge_port *port = (void *)data;
 820         struct net_bridge *br = port->br;
 821
 822         spin_lock(&br->multicast_lock);
 823         if (port->state == BR_STATE_DISABLED ||
 824             port->state == BR_STATE_BLOCKING)
 825                 goto out;
 826
 827         if (port->multicast_startup_queries_sent <
 828             br->multicast_startup_query_count)
 829                 port->multicast_startup_queries_sent++;
 830
 831         br_multicast_send_query(port->br, port,
 832                                 port->multicast_startup_queries_sent);
 833
 834 out:
 835         spin_unlock(&br->multicast_lock);
 836 }
 837
 838 void br_multicast_add_port(struct net_bridge_port *port)
 839 {
 840         port->multicast_router = 1;
 841
 842         setup_timer(&port->multicast_router_timer, br_multicast_router_expired,
 843                     (unsigned long)port);
 844         setup_timer(&port->multicast_query_timer,
 845                     br_multicast_port_query_expired, (unsigned long)port);
 846 }
 847
 848 void br_multicast_del_port(struct net_bridge_port *port)
 849 {
 850         del_timer_sync(&port->multicast_router_timer);
 851 }
 852
 853 static void __br_multicast_enable_port(struct net_bridge_port *port)
 854 {
 855         port->multicast_startup_queries_sent = 0;
 856
 857         if (try_to_del_timer_sync(&port->multicast_query_timer) >= 0 ||
 858             del_timer(&port->multicast_query_timer))
 859                 mod_timer(&port->multicast_query_timer, jiffies);
 860 }
 861
 862 void br_multicast_enable_port(struct net_bridge_port *port)
 863 {
 864         struct net_bridge *br = port->br;
 865
 866         spin_lock(&br->multicast_lock);
 867         if (br->multicast_disabled || !netif_running(br->dev))
 868                 goto out;
 869
 870         __br_multicast_enable_port(port);
 871
 872 out:
 873         spin_unlock(&br->multicast_lock);
 874 }
 875
 876 void br_multicast_disable_port(struct net_bridge_port *port)
 877 {
 878         struct net_bridge *br = port->br;
 879         struct net_bridge_port_group *pg;
 880         struct hlist_node *n;
 881
 882         spin_lock(&br->multicast_lock);
 883         hlist_for_each_entry_safe(pg, n, &port->mglist, mglist)
 884                 br_multicast_del_pg(br, pg);
 885
 886         if (!hlist_unhashed(&port->rlist))
 887                 hlist_del_init_rcu(&port->rlist);
 888         del_timer(&port->multicast_router_timer);
 889         del_timer(&port->multicast_query_timer);
 890         spin_unlock(&br->multicast_lock);
 891 }
 892
 893 static int br_ip4_multicast_igmp3_report(struct net_bridge *br,
 894                                          struct net_bridge_port *port,
 895                                          struct sk_buff *skb)
 896 {
 897         struct igmpv3_report *ih;
 898         struct igmpv3_grec *grec;
 899         int i;
 900         int len;
 901         int num;
 902         int type;
 903         int err = 0;
 904         __be32 group;
 905         u16 vid = 0;
 906
 907         if (!pskb_may_pull(skb, sizeof(*ih)))
 908                 return -EINVAL;
 909
 910         br_vlan_get_tag(skb, &vid);
 911         ih = igmpv3_report_hdr(skb);
 912         num = ntohs(ih->ngrec);
 913         len = sizeof(*ih);
 914
 915         for (i = 0; i < num; i++) {
 916                 len += sizeof(*grec);
 917                 if (!pskb_may_pull(skb, len))
 918                         return -EINVAL;
 919
 920                 grec = (void *)(skb->data + len - sizeof(*grec));
 921                 group = grec->grec_mca;
 922                 type = grec->grec_type;
 923
 924                 len += ntohs(grec->grec_nsrcs) * 4;
 925                 if (!pskb_may_pull(skb, len))
 926                         return -EINVAL;
 927
 928                 /* We treat this as an IGMPv2 report for now. */
 929                 switch (type) {
 930                 case IGMPV3_MODE_IS_INCLUDE:
 931                 case IGMPV3_MODE_IS_EXCLUDE:
 932                 case IGMPV3_CHANGE_TO_INCLUDE:
 933                 case IGMPV3_CHANGE_TO_EXCLUDE:
 934                 case IGMPV3_ALLOW_NEW_SOURCES:
 935                 case IGMPV3_BLOCK_OLD_SOURCES:
 936                         break;
 937
 938                 default:
 939                         continue;
 940                 }
 941
 942                 err = br_ip4_multicast_add_group(br, port, group, vid);
 943                 if (err)
 944                         break;
 945         }
 946
 947         return err;
 948 }
 949
 950 #if IS_ENABLED(CONFIG_IPV6)
 951 static int br_ip6_multicast_mld2_report(struct net_bridge *br,
 952                                         struct net_bridge_port *port,
 953                                         struct sk_buff *skb)
 954 {
 955         struct icmp6hdr *icmp6h;
 956         struct mld2_grec *grec;
 957         int i;
 958         int len;
 959         int num;
 960         int err = 0;
 961         u16 vid = 0;
 962
 963         if (!pskb_may_pull(skb, sizeof(*icmp6h)))
 964                 return -EINVAL;
 965
 966         br_vlan_get_tag(skb, &vid);
 967         icmp6h = icmp6_hdr(skb);
 968         num = ntohs(icmp6h->icmp6_dataun.un_data16[1]);
 969         len = sizeof(*icmp6h);
 970
 971         for (i = 0; i < num; i++) {
 972                 __be16 *nsrcs, _nsrcs;
 973
 974                 nsrcs = skb_header_pointer(skb,
 975                                            len + offsetof(struct mld2_grec,
 976                                                           grec_nsrcs),
 977                                            sizeof(_nsrcs), &_nsrcs);
 978                 if (!nsrcs)
 979                         return -EINVAL;
 980
 981                 if (!pskb_may_pull(skb,
 982                                    len + sizeof(*grec) +
 983                                    sizeof(struct in6_addr) * ntohs(*nsrcs)))
 984                         return -EINVAL;
 985
 986                 grec = (struct mld2_grec *)(skb->data + len);
 987                 len += sizeof(*grec) +
 988                        sizeof(struct in6_addr) * ntohs(*nsrcs);
 989
 990                 /* We treat these as MLDv1 reports for now. */
 991                 switch (grec->grec_type) {
 992                 case MLD2_MODE_IS_INCLUDE:
 993                 case MLD2_MODE_IS_EXCLUDE:
 994                 case MLD2_CHANGE_TO_INCLUDE:
 995                 case MLD2_CHANGE_TO_EXCLUDE:
 996                 case MLD2_ALLOW_NEW_SOURCES:
 997                 case MLD2_BLOCK_OLD_SOURCES:
 998                         break;
 999
1000                 default:
1001                         continue;
1002                 }
1003
1004                 err = br_ip6_multicast_add_group(br, port, &grec->grec_mca,
1005                                                  vid);
1006                 if (!err)
1007                         break;
1008         }
1009
1010         return err;
1011 }
1012 #endif
1013
1014 /*
1015  * Add port to router_list
1016  *  list is maintained ordered by pointer value
1017  *  and locked by br->multicast_lock and RCU
1018  */
1019 static void br_multicast_add_router(struct net_bridge *br,
1020                                     struct net_bridge_port *port)
1021 {
1022         struct net_bridge_port *p;
1023         struct hlist_node *slot = NULL;
1024
1025         hlist_for_each_entry(p, &br->router_list, rlist) {
1026                 if ((unsigned long) port >= (unsigned long) p)
1027                         break;
1028                 slot = &p->rlist;
1029         }
1030
1031         if (slot)
1032                 hlist_add_after_rcu(slot, &port->rlist);
1033         else
1034                 hlist_add_head_rcu(&port->rlist, &br->router_list);
1035 }
1036
1037 static void br_multicast_mark_router(struct net_bridge *br,
1038                                      struct net_bridge_port *port)
1039 {
1040         unsigned long now = jiffies;
1041
1042         if (!port) {
1043                 if (br->multicast_router == 1)
1044                         mod_timer(&br->multicast_router_timer,
1045                                   now + br->multicast_querier_interval);
1046                 return;
1047         }
1048
1049         if (port->multicast_router != 1)
1050                 return;
1051
1052         if (!hlist_unhashed(&port->rlist))
1053                 goto timer;
1054
1055         br_multicast_add_router(br, port);
1056
1057 timer:
1058         mod_timer(&port->multicast_router_timer,
1059                   now + br->multicast_querier_interval);
1060 }
1061
1062 static void br_multicast_query_received(struct net_bridge *br,
1063                                         struct net_bridge_port *port,
1064                                         int saddr)
1065 {
1066         if (saddr)
1067                 mod_timer(&br->multicast_querier_timer,
1068                           jiffies + br->multicast_querier_interval);
1069         else if (timer_pending(&br->multicast_querier_timer))
1070                 return;
1071
1072         br_multicast_mark_router(br, port);
1073 }
1074
1075 static int br_ip4_multicast_query(struct net_bridge *br,
1076                                   struct net_bridge_port *port,
1077                                   struct sk_buff *skb)
1078 {
1079         const struct iphdr *iph = ip_hdr(skb);
1080         struct igmphdr *ih = igmp_hdr(skb);
1081         struct net_bridge_mdb_entry *mp;
1082         struct igmpv3_query *ih3;
1083         struct net_bridge_port_group *p;
1084         struct net_bridge_port_group __rcu **pp;
1085         unsigned long max_delay;
1086         unsigned long now = jiffies;
1087         __be32 group;
1088         int err = 0;
1089         u16 vid = 0;
1090
1091         spin_lock(&br->multicast_lock);
1092         if (!netif_running(br->dev) ||
1093             (port && port->state == BR_STATE_DISABLED))
1094                 goto out;
1095
1096         br_multicast_query_received(br, port, !!iph->saddr);
1097
1098         group = ih->group;
1099
1100         if (skb->len == sizeof(*ih)) {
1101                 max_delay = ih->code * (HZ / IGMP_TIMER_SCALE);
1102
1103                 if (!max_delay) {
1104                         max_delay = 10 * HZ;
1105                         group = 0;
1106                 }
1107         } else {
1108                 if (!pskb_may_pull(skb, sizeof(struct igmpv3_query))) {
1109                         err = -EINVAL;
1110                         goto out;
1111                 }
1112
1113                 ih3 = igmpv3_query_hdr(skb);
1114                 if (ih3->nsrcs)
1115                         goto out;
1116
1117                 max_delay = ih3->code ?
1118                             IGMPV3_MRC(ih3->code) * (HZ / IGMP_TIMER_SCALE) : 1;
1119         }
1120
1121         if (!group)
1122                 goto out;
1123
1124         br_vlan_get_tag(skb, &vid);
1125         mp = br_mdb_ip4_get(mlock_dereference(br->mdb, br), group, vid);
1126         if (!mp)
1127                 goto out;
1128
1129         setup_timer(&mp->timer, br_multicast_group_expired, (unsigned long)mp);
1130         mod_timer(&mp->timer, now + br->multicast_membership_interval);
1131         mp->timer_armed = true;
1132
1133         max_delay *= br->multicast_last_member_count;
1134
1135         if (mp->mglist &&
1136             (timer_pending(&mp->timer) ?
1137              time_after(mp->timer.expires, now + max_delay) :
1138              try_to_del_timer_sync(&mp->timer) >= 0))
1139                 mod_timer(&mp->timer, now + max_delay);
1140
1141         for (pp = &mp->ports;
1142              (p = mlock_dereference(*pp, br)) != NULL;
1143              pp = &p->next) {
1144                 if (timer_pending(&p->timer) ?
1145                     time_after(p->timer.expires, now + max_delay) :
1146                     try_to_del_timer_sync(&p->timer) >= 0)
1147                         mod_timer(&p->timer, now + max_delay);
1148         }
1149
1150 out:
1151         spin_unlock(&br->multicast_lock);
1152         return err;
1153 }
1154
1155 #if IS_ENABLED(CONFIG_IPV6)
1156 static int br_ip6_multicast_query(struct net_bridge *br,
1157                                   struct net_bridge_port *port,
1158                                   struct sk_buff *skb)
1159 {
1160         const struct ipv6hdr *ip6h = ipv6_hdr(skb);
1161         struct mld_msg *mld;
1162         struct net_bridge_mdb_entry *mp;
1163         struct mld2_query *mld2q;
1164         struct net_bridge_port_group *p;
1165         struct net_bridge_port_group __rcu **pp;
1166         unsigned long max_delay;
1167         unsigned long now = jiffies;
1168         const struct in6_addr *group = NULL;
1169         int err = 0;
1170         u16 vid = 0;
1171
1172         spin_lock(&br->multicast_lock);
1173         if (!netif_running(br->dev) ||
1174             (port && port->state == BR_STATE_DISABLED))
1175                 goto out;
1176
1177         br_multicast_query_received(br, port, !ipv6_addr_any(&ip6h->saddr));
1178
1179         if (skb->len == sizeof(*mld)) {
1180                 if (!pskb_may_pull(skb, sizeof(*mld))) {
1181                         err = -EINVAL;
1182                         goto out;
1183                 }
1184                 mld = (struct mld_msg *) icmp6_hdr(skb);
1185                 max_delay = msecs_to_jiffies(ntohs(mld->mld_maxdelay));
1186                 if (max_delay)
1187                         group = &mld->mld_mca;
1188         } else if (skb->len >= sizeof(*mld2q)) {
1189                 if (!pskb_may_pull(skb, sizeof(*mld2q))) {
1190                         err = -EINVAL;
1191                         goto out;
1192                 }
1193                 mld2q = (struct mld2_query *)icmp6_hdr(skb);
1194                 if (!mld2q->mld2q_nsrcs)
1195                         group = &mld2q->mld2q_mca;
1196                 max_delay = mld2q->mld2q_mrc ? MLDV2_MRC(ntohs(mld2q->mld2q_mrc)) : 1;
1197         }
1198
1199         if (!group)
1200                 goto out;
1201
1202         br_vlan_get_tag(skb, &vid);
1203         mp = br_mdb_ip6_get(mlock_dereference(br->mdb, br), group, vid);
1204         if (!mp)
1205                 goto out;
1206
1207         setup_timer(&mp->timer, br_multicast_group_expired, (unsigned long)mp);
1208         mod_timer(&mp->timer, now + br->multicast_membership_interval);
1209         mp->timer_armed = true;
1210
1211         max_delay *= br->multicast_last_member_count;
1212         if (mp->mglist &&
1213             (timer_pending(&mp->timer) ?
1214              time_after(mp->timer.expires, now + max_delay) :
1215              try_to_del_timer_sync(&mp->timer) >= 0))
1216                 mod_timer(&mp->timer, now + max_delay);
1217
1218         for (pp = &mp->ports;
1219              (p = mlock_dereference(*pp, br)) != NULL;
1220              pp = &p->next) {
1221                 if (timer_pending(&p->timer) ?
1222                     time_after(p->timer.expires, now + max_delay) :
1223                     try_to_del_timer_sync(&p->timer) >= 0)
1224                         mod_timer(&p->timer, now + max_delay);
1225         }
1226
1227 out:
1228         spin_unlock(&br->multicast_lock);
1229         return err;
1230 }
1231 #endif
1232
1233 static void br_multicast_leave_group(struct net_bridge *br,
1234                                      struct net_bridge_port *port,
1235                                      struct br_ip *group)
1236 {
1237         struct net_bridge_mdb_htable *mdb;
1238         struct net_bridge_mdb_entry *mp;
1239         struct net_bridge_port_group *p;
1240         unsigned long now;
1241         unsigned long time;
1242
1243         spin_lock(&br->multicast_lock);
1244         if (!netif_running(br->dev) ||
1245             (port && port->state == BR_STATE_DISABLED) ||
1246             timer_pending(&br->multicast_querier_timer))
1247                 goto out;
1248
1249         mdb = mlock_dereference(br->mdb, br);
1250         mp = br_mdb_ip_get(mdb, group);
1251         if (!mp)
1252                 goto out;
1253
1254         if (br->multicast_querier &&
1255             !timer_pending(&br->multicast_querier_timer)) {
1256                 __br_multicast_send_query(br, port, &mp->addr);
1257
1258                 time = jiffies + br->multicast_last_member_count *
1259                                  br->multicast_last_member_interval;
1260                 mod_timer(port ? &port->multicast_query_timer :
1261                                  &br->multicast_query_timer, time);
1262
1263                 for (p = mlock_dereference(mp->ports, br);
1264                      p != NULL;
1265                      p = mlock_dereference(p->next, br)) {
1266                         if (p->port != port)
1267                                 continue;
1268
1269                         if (!hlist_unhashed(&p->mglist) &&
1270                             (timer_pending(&p->timer) ?
1271                              time_after(p->timer.expires, time) :
1272                              try_to_del_timer_sync(&p->timer) >= 0)) {
1273                                 mod_timer(&p->timer, time);
1274                         }
1275
1276                         break;
1277                 }
1278         }
1279
1280         if (port && (port->flags & BR_MULTICAST_FAST_LEAVE)) {
1281                 struct net_bridge_port_group __rcu **pp;
1282
1283                 for (pp = &mp->ports;
1284                      (p = mlock_dereference(*pp, br)) != NULL;
1285                      pp = &p->next) {
1286                         if (p->port != port)
1287                                 continue;
1288
1289                         rcu_assign_pointer(*pp, p->next);
1290                         hlist_del_init(&p->mglist);
1291                         del_timer(&p->timer);
1292                         call_rcu_bh(&p->rcu, br_multicast_free_pg);
1293                         br_mdb_notify(br->dev, port, group, RTM_DELMDB);
1294
1295                         if (!mp->ports && !mp->mglist && mp->timer_armed &&
1296                             netif_running(br->dev))
1297                                 mod_timer(&mp->timer, jiffies);
1298                 }
1299                 goto out;
1300         }
1301
1302         now = jiffies;
1303         time = now + br->multicast_last_member_count *
1304                      br->multicast_last_member_interval;
1305
1306         if (!port) {
1307                 if (mp->mglist && mp->timer_armed &&
1308                     (timer_pending(&mp->timer) ?
1309                      time_after(mp->timer.expires, time) :
1310                      try_to_del_timer_sync(&mp->timer) >= 0)) {
1311                         mod_timer(&mp->timer, time);
1312                 }
1313         }
1314
1315 out:
1316         spin_unlock(&br->multicast_lock);
1317 }
1318
1319 static void br_ip4_multicast_leave_group(struct net_bridge *br,
1320                                          struct net_bridge_port *port,
1321                                          __be32 group,
1322                                          __u16 vid)
1323 {
1324         struct br_ip br_group;
1325
1326         if (ipv4_is_local_multicast(group))
1327                 return;
1328
1329         br_group.u.ip4 = group;
1330         br_group.proto = htons(ETH_P_IP);
1331         br_group.vid = vid;
1332
1333         br_multicast_leave_group(br, port, &br_group);
1334 }
1335
1336 #if IS_ENABLED(CONFIG_IPV6)
1337 static void br_ip6_multicast_leave_group(struct net_bridge *br,
1338                                          struct net_bridge_port *port,
1339                                          const struct in6_addr *group,
1340                                          __u16 vid)
1341 {
1342         struct br_ip br_group;
1343
1344         if (!ipv6_is_transient_multicast(group))
1345                 return;
1346
1347         br_group.u.ip6 = *group;
1348         br_group.proto = htons(ETH_P_IPV6);
1349         br_group.vid = vid;
1350
1351         br_multicast_leave_group(br, port, &br_group);
1352 }
1353 #endif
1354
1355 static int br_multicast_ipv4_rcv(struct net_bridge *br,
1356                                  struct net_bridge_port *port,
1357                                  struct sk_buff *skb)
1358 {
1359         struct sk_buff *skb2 = skb;
1360         const struct iphdr *iph;
1361         struct igmphdr *ih;
1362         unsigned int len;
1363         unsigned int offset;
1364         int err;
1365         u16 vid = 0;
1366
1367         /* We treat OOM as packet loss for now. */
1368         if (!pskb_may_pull(skb, sizeof(*iph)))
1369                 return -EINVAL;
1370
1371         iph = ip_hdr(skb);
1372
1373         if (iph->ihl < 5 || iph->version != 4)
1374                 return -EINVAL;
1375
1376         if (!pskb_may_pull(skb, ip_hdrlen(skb)))
1377                 return -EINVAL;
1378
1379         iph = ip_hdr(skb);
1380
1381         if (unlikely(ip_fast_csum((u8 *)iph, iph->ihl)))
1382                 return -EINVAL;
1383
1384         if (iph->protocol != IPPROTO_IGMP) {
1385                 if (!ipv4_is_local_multicast(iph->daddr))
1386                         BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1387                 return 0;
1388         }
1389
1390         len = ntohs(iph->tot_len);
1391         if (skb->len < len || len < ip_hdrlen(skb))
1392                 return -EINVAL;
1393
1394         if (skb->len > len) {
1395                 skb2 = skb_clone(skb, GFP_ATOMIC);
1396                 if (!skb2)
1397                         return -ENOMEM;
1398
1399                 err = pskb_trim_rcsum(skb2, len);
1400                 if (err)
1401                         goto err_out;
1402         }
1403
1404         len -= ip_hdrlen(skb2);
1405         offset = skb_network_offset(skb2) + ip_hdrlen(skb2);
1406         __skb_pull(skb2, offset);
1407         skb_reset_transport_header(skb2);
1408
1409         err = -EINVAL;
1410         if (!pskb_may_pull(skb2, sizeof(*ih)))
1411                 goto out;
1412
1413         switch (skb2->ip_summed) {
1414         case CHECKSUM_COMPLETE:
1415                 if (!csum_fold(skb2->csum))
1416                         break;
1417                 /* fall through */
1418         case CHECKSUM_NONE:
1419                 skb2->csum = 0;
1420                 if (skb_checksum_complete(skb2))
1421                         goto out;
1422         }
1423
1424         err = 0;
1425
1426         br_vlan_get_tag(skb2, &vid);
1427         BR_INPUT_SKB_CB(skb)->igmp = 1;
1428         ih = igmp_hdr(skb2);
1429
1430         switch (ih->type) {
1431         case IGMP_HOST_MEMBERSHIP_REPORT:
1432         case IGMPV2_HOST_MEMBERSHIP_REPORT:
1433                 BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1434                 err = br_ip4_multicast_add_group(br, port, ih->group, vid);
1435                 break;
1436         case IGMPV3_HOST_MEMBERSHIP_REPORT:
1437                 err = br_ip4_multicast_igmp3_report(br, port, skb2);
1438                 break;
1439         case IGMP_HOST_MEMBERSHIP_QUERY:
1440                 err = br_ip4_multicast_query(br, port, skb2);
1441                 break;
1442         case IGMP_HOST_LEAVE_MESSAGE:
1443                 br_ip4_multicast_leave_group(br, port, ih->group, vid);
1444                 break;
1445         }
1446
1447 out:
1448         __skb_push(skb2, offset);
1449 err_out:
1450         if (skb2 != skb)
1451                 kfree_skb(skb2);
1452         return err;
1453 }
1454
1455 #if IS_ENABLED(CONFIG_IPV6)
1456 static int br_multicast_ipv6_rcv(struct net_bridge *br,
1457                                  struct net_bridge_port *port,
1458                                  struct sk_buff *skb)
1459 {
1460         struct sk_buff *skb2;
1461         const struct ipv6hdr *ip6h;
1462         u8 icmp6_type;
1463         u8 nexthdr;
1464         __be16 frag_off;
1465         unsigned int len;
1466         int offset;
1467         int err;
1468         u16 vid = 0;
1469
1470         if (!pskb_may_pull(skb, sizeof(*ip6h)))
1471                 return -EINVAL;
1472
1473         ip6h = ipv6_hdr(skb);
1474
1475         /*
1476          * We're interested in MLD messages only.
1477          *  - Version is 6
1478          *  - MLD has always Router Alert hop-by-hop option
1479          *  - But we do not support jumbrograms.
1480          */
1481         if (ip6h->version != 6 ||
1482             ip6h->nexthdr != IPPROTO_HOPOPTS ||
1483             ip6h->payload_len == 0)
1484                 return 0;
1485
1486         len = ntohs(ip6h->payload_len) + sizeof(*ip6h);
1487         if (skb->len < len)
1488                 return -EINVAL;
1489
1490         nexthdr = ip6h->nexthdr;
1491         offset = ipv6_skip_exthdr(skb, sizeof(*ip6h), &nexthdr, &frag_off);
1492
1493         if (offset < 0 || nexthdr != IPPROTO_ICMPV6)
1494                 return 0;
1495
1496         /* Okay, we found ICMPv6 header */
1497         skb2 = skb_clone(skb, GFP_ATOMIC);
1498         if (!skb2)
1499                 return -ENOMEM;
1500
1501         err = -EINVAL;
1502         if (!pskb_may_pull(skb2, offset + sizeof(struct icmp6hdr)))
1503                 goto out;
1504
1505         len -= offset - skb_network_offset(skb2);
1506
1507         __skb_pull(skb2, offset);
1508         skb_reset_transport_header(skb2);
1509         skb_postpull_rcsum(skb2, skb_network_header(skb2),
1510                            skb_network_header_len(skb2));
1511
1512         icmp6_type = icmp6_hdr(skb2)->icmp6_type;
1513
1514         switch (icmp6_type) {
1515         case ICMPV6_MGM_QUERY:
1516         case ICMPV6_MGM_REPORT:
1517         case ICMPV6_MGM_REDUCTION:
1518         case ICMPV6_MLD2_REPORT:
1519                 break;
1520         default:
1521                 err = 0;
1522                 goto out;
1523         }
1524
1525         /* Okay, we found MLD message. Check further. */
1526         if (skb2->len > len) {
1527                 err = pskb_trim_rcsum(skb2, len);
1528                 if (err)
1529                         goto out;
1530                 err = -EINVAL;
1531         }
1532
1533         ip6h = ipv6_hdr(skb2);
1534
1535         switch (skb2->ip_summed) {
1536         case CHECKSUM_COMPLETE:
1537                 if (!csum_ipv6_magic(&ip6h->saddr, &ip6h->daddr, skb2->len,
1538                                         IPPROTO_ICMPV6, skb2->csum))
1539                         break;
1540                 /*FALLTHROUGH*/
1541         case CHECKSUM_NONE:
1542                 skb2->csum = ~csum_unfold(csum_ipv6_magic(&ip6h->saddr,
1543                                                         &ip6h->daddr,
1544                                                         skb2->len,
1545                                                         IPPROTO_ICMPV6, 0));
1546                 if (__skb_checksum_complete(skb2))
1547                         goto out;
1548         }
1549
1550         err = 0;
1551
1552         br_vlan_get_tag(skb, &vid);
1553         BR_INPUT_SKB_CB(skb)->igmp = 1;
1554
1555         switch (icmp6_type) {
1556         case ICMPV6_MGM_REPORT:
1557             {
1558                 struct mld_msg *mld;
1559                 if (!pskb_may_pull(skb2, sizeof(*mld))) {
1560                         err = -EINVAL;
1561                         goto out;
1562                 }
1563                 mld = (struct mld_msg *)skb_transport_header(skb2);
1564                 BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1565                 err = br_ip6_multicast_add_group(br, port, &mld->mld_mca, vid);
1566                 break;
1567             }
1568         case ICMPV6_MLD2_REPORT:
1569                 err = br_ip6_multicast_mld2_report(br, port, skb2);
1570                 break;
1571         case ICMPV6_MGM_QUERY:
1572                 err = br_ip6_multicast_query(br, port, skb2);
1573                 break;
1574         case ICMPV6_MGM_REDUCTION:
1575             {
1576                 struct mld_msg *mld;
1577                 if (!pskb_may_pull(skb2, sizeof(*mld))) {
1578                         err = -EINVAL;
1579                         goto out;
1580                 }
1581                 mld = (struct mld_msg *)skb_transport_header(skb2);
1582                 br_ip6_multicast_leave_group(br, port, &mld->mld_mca, vid);
1583             }
1584         }
1585
1586 out:
1587         kfree_skb(skb2);
1588         return err;
1589 }
1590 #endif
1591
1592 int br_multicast_rcv(struct net_bridge *br, struct net_bridge_port *port,
1593                      struct sk_buff *skb)
1594 {
1595         BR_INPUT_SKB_CB(skb)->igmp = 0;
1596         BR_INPUT_SKB_CB(skb)->mrouters_only = 0;
1597
1598         if (br->multicast_disabled)
1599                 return 0;
1600
1601         switch (skb->protocol) {
1602         case htons(ETH_P_IP):
1603                 return br_multicast_ipv4_rcv(br, port, skb);
1604 #if IS_ENABLED(CONFIG_IPV6)
1605         case htons(ETH_P_IPV6):
1606                 return br_multicast_ipv6_rcv(br, port, skb);
1607 #endif
1608         }
1609
1610         return 0;
1611 }
1612
1613 static void br_multicast_query_expired(unsigned long data)
1614 {
1615         struct net_bridge *br = (void *)data;
1616
1617         spin_lock(&br->multicast_lock);
1618         if (br->multicast_startup_queries_sent <
1619             br->multicast_startup_query_count)
1620                 br->multicast_startup_queries_sent++;
1621
1622         br_multicast_send_query(br, NULL, br->multicast_startup_queries_sent);
1623
1624         spin_unlock(&br->multicast_lock);
1625 }
1626
1627 void br_multicast_init(struct net_bridge *br)
1628 {
1629         br->hash_elasticity = 4;
1630         br->hash_max = 512;
1631
1632         br->multicast_router = 1;
1633         br->multicast_querier = 0;
1634         br->multicast_query_use_ifaddr = 0;
1635         br->multicast_last_member_count = 2;
1636         br->multicast_startup_query_count = 2;
1637
1638         br->multicast_last_member_interval = HZ;
1639         br->multicast_query_response_interval = 10 * HZ;
1640         br->multicast_startup_query_interval = 125 * HZ / 4;
1641         br->multicast_query_interval = 125 * HZ;
1642         br->multicast_querier_interval = 255 * HZ;
1643         br->multicast_membership_interval = 260 * HZ;
1644
1645         spin_lock_init(&br->multicast_lock);
1646         setup_timer(&br->multicast_router_timer,
1647                     br_multicast_local_router_expired, 0);
1648         setup_timer(&br->multicast_querier_timer,
1649                     br_multicast_querier_expired, (unsigned long)br);
1650         setup_timer(&br->multicast_query_timer, br_multicast_query_expired,
1651                     (unsigned long)br);
1652 }
1653
1654 void br_multicast_open(struct net_bridge *br)
1655 {
1656         br->multicast_startup_queries_sent = 0;
1657
1658         if (br->multicast_disabled)
1659                 return;
1660
1661         mod_timer(&br->multicast_query_timer, jiffies);
1662 }
1663
1664 void br_multicast_stop(struct net_bridge *br)
1665 {
1666         struct net_bridge_mdb_htable *mdb;
1667         struct net_bridge_mdb_entry *mp;
1668         struct hlist_node *n;
1669         u32 ver;
1670         int i;
1671
1672         del_timer_sync(&br->multicast_router_timer);
1673         del_timer_sync(&br->multicast_querier_timer);
1674         del_timer_sync(&br->multicast_query_timer);
1675
1676         spin_lock_bh(&br->multicast_lock);
1677         mdb = mlock_dereference(br->mdb, br);
1678         if (!mdb)
1679                 goto out;
1680
1681         br->mdb = NULL;
1682
1683         ver = mdb->ver;
1684         for (i = 0; i < mdb->max; i++) {
1685                 hlist_for_each_entry_safe(mp, n, &mdb->mhash[i],
1686                                           hlist[ver]) {
1687                         del_timer(&mp->timer);
1688                         mp->timer_armed = false;
1689                         call_rcu_bh(&mp->rcu, br_multicast_free_group);
1690                 }
1691         }
1692
1693         if (mdb->old) {
1694                 spin_unlock_bh(&br->multicast_lock);
1695                 rcu_barrier_bh();
1696                 spin_lock_bh(&br->multicast_lock);
1697                 WARN_ON(mdb->old);
1698         }
1699
1700         mdb->old = mdb;
1701         call_rcu_bh(&mdb->rcu, br_mdb_free);
1702
1703 out:
1704         spin_unlock_bh(&br->multicast_lock);
1705 }
1706
1707 int br_multicast_set_router(struct net_bridge *br, unsigned long val)
1708 {
1709         int err = -ENOENT;
1710
1711         spin_lock_bh(&br->multicast_lock);
1712         if (!netif_running(br->dev))
1713                 goto unlock;
1714
1715         switch (val) {
1716         case 0:
1717         case 2:
1718                 del_timer(&br->multicast_router_timer);
1719                 /* fall through */
1720         case 1:
1721                 br->multicast_router = val;
1722                 err = 0;
1723                 break;
1724
1725         default:
1726                 err = -EINVAL;
1727                 break;
1728         }
1729
1730 unlock:
1731         spin_unlock_bh(&br->multicast_lock);
1732
1733         return err;
1734 }
1735
1736 int br_multicast_set_port_router(struct net_bridge_port *p, unsigned long val)
1737 {
1738         struct net_bridge *br = p->br;
1739         int err = -ENOENT;
1740
1741         spin_lock(&br->multicast_lock);
1742         if (!netif_running(br->dev) || p->state == BR_STATE_DISABLED)
1743                 goto unlock;
1744
1745         switch (val) {
1746         case 0:
1747         case 1:
1748         case 2:
1749                 p->multicast_router = val;
1750                 err = 0;
1751
1752                 if (val < 2 && !hlist_unhashed(&p->rlist))
1753                         hlist_del_init_rcu(&p->rlist);
1754
1755                 if (val == 1)
1756                         break;
1757
1758                 del_timer(&p->multicast_router_timer);
1759
1760                 if (val == 0)
1761                         break;
1762
1763                 br_multicast_add_router(br, p);
1764                 break;
1765
1766         default:
1767                 err = -EINVAL;
1768                 break;
1769         }
1770
1771 unlock:
1772         spin_unlock(&br->multicast_lock);
1773
1774         return err;
1775 }
1776
1777 static void br_multicast_start_querier(struct net_bridge *br)
1778 {
1779         struct net_bridge_port *port;
1780
1781         br_multicast_open(br);
1782
1783         list_for_each_entry(port, &br->port_list, list) {
1784                 if (port->state == BR_STATE_DISABLED ||
1785                     port->state == BR_STATE_BLOCKING)
1786                         continue;
1787
1788                 __br_multicast_enable_port(port);
1789         }
1790 }
1791
1792 int br_multicast_toggle(struct net_bridge *br, unsigned long val)
1793 {
1794         int err = 0;
1795         struct net_bridge_mdb_htable *mdb;
1796
1797         spin_lock_bh(&br->multicast_lock);
1798         if (br->multicast_disabled == !val)
1799                 goto unlock;
1800
1801         br->multicast_disabled = !val;
1802         if (br->multicast_disabled)
1803                 goto unlock;
1804
1805         if (!netif_running(br->dev))
1806                 goto unlock;
1807
1808         mdb = mlock_dereference(br->mdb, br);
1809         if (mdb) {
1810                 if (mdb->old) {
1811                         err = -EEXIST;
1812 rollback:
1813                         br->multicast_disabled = !!val;
1814                         goto unlock;
1815                 }
1816
1817                 err = br_mdb_rehash(&br->mdb, mdb->max,
1818                                     br->hash_elasticity);
1819                 if (err)
1820                         goto rollback;
1821         }
1822
1823         br_multicast_start_querier(br);
1824
1825 unlock:
1826         spin_unlock_bh(&br->multicast_lock);
1827
1828         return err;
1829 }
1830
1831 int br_multicast_set_querier(struct net_bridge *br, unsigned long val)
1832 {
1833         val = !!val;
1834
1835         spin_lock_bh(&br->multicast_lock);
1836         if (br->multicast_querier == val)
1837                 goto unlock;
1838
1839         br->multicast_querier = val;
1840         if (val)
1841                 br_multicast_start_querier(br);
1842
1843 unlock:
1844         spin_unlock_bh(&br->multicast_lock);
1845
1846         return 0;
1847 }
1848
1849 int br_multicast_set_hash_max(struct net_bridge *br, unsigned long val)
1850 {
1851         int err = -ENOENT;
1852         u32 old;
1853         struct net_bridge_mdb_htable *mdb;
1854
1855         spin_lock(&br->multicast_lock);
1856         if (!netif_running(br->dev))
1857                 goto unlock;
1858
1859         err = -EINVAL;
1860         if (!is_power_of_2(val))
1861                 goto unlock;
1862
1863         mdb = mlock_dereference(br->mdb, br);
1864         if (mdb && val < mdb->size)
1865                 goto unlock;
1866
1867         err = 0;
1868
1869         old = br->hash_max;
1870         br->hash_max = val;
1871
1872         if (mdb) {
1873                 if (mdb->old) {
1874                         err = -EEXIST;
1875 rollback:
1876                         br->hash_max = old;
1877                         goto unlock;
1878                 }
1879
1880                 err = br_mdb_rehash(&br->mdb, br->hash_max,
1881                                     br->hash_elasticity);
1882                 if (err)
1883                         goto rollback;
1884         }
1885
1886 unlock:
1887         spin_unlock(&br->multicast_lock);
1888
1889         return err;
1890 }