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         setup_timer(&mp->timer, br_multicast_group_expired,
 623                     (unsigned long)mp);
 624
 625         hlist_add_head_rcu(&mp->hlist[mdb->ver], &mdb->mhash[hash]);
 626         mdb->size++;
 627
 628 out:
 629         return mp;
 630 }
 631
 632 struct net_bridge_port_group *br_multicast_new_port_group(
 633                         struct net_bridge_port *port,
 634                         struct br_ip *group,
 635                         struct net_bridge_port_group __rcu *next,
 636                         unsigned char state)
 637 {
 638         struct net_bridge_port_group *p;
 639
 640         p = kzalloc(sizeof(*p), GFP_ATOMIC);
 641         if (unlikely(!p))
 642                 return NULL;
 643
 644         p->addr = *group;
 645         p->port = port;
 646         p->state = state;
 647         rcu_assign_pointer(p->next, next);
 648         hlist_add_head(&p->mglist, &port->mglist);
 649         setup_timer(&p->timer, br_multicast_port_group_expired,
 650                     (unsigned long)p);
 651         return p;
 652 }
 653
 654 static int br_multicast_add_group(struct net_bridge *br,
 655                                   struct net_bridge_port *port,
 656                                   struct br_ip *group)
 657 {
 658         struct net_bridge_mdb_entry *mp;
 659         struct net_bridge_port_group *p;
 660         struct net_bridge_port_group __rcu **pp;
 661         int err;
 662
 663         spin_lock(&br->multicast_lock);
 664         if (!netif_running(br->dev) ||
 665             (port && port->state == BR_STATE_DISABLED))
 666                 goto out;
 667
 668         mp = br_multicast_new_group(br, port, group);
 669         err = PTR_ERR(mp);
 670         if (IS_ERR(mp))
 671                 goto err;
 672
 673         if (!port) {
 674                 mp->mglist = true;
 675                 goto out;
 676         }
 677
 678         for (pp = &mp->ports;
 679              (p = mlock_dereference(*pp, br)) != NULL;
 680              pp = &p->next) {
 681                 if (p->port == port)
 682                         goto out;
 683                 if ((unsigned long)p->port < (unsigned long)port)
 684                         break;
 685         }
 686
 687         p = br_multicast_new_port_group(port, group, *pp, MDB_TEMPORARY);
 688         if (unlikely(!p))
 689                 goto err;
 690         rcu_assign_pointer(*pp, p);
 691         br_mdb_notify(br->dev, port, group, RTM_NEWMDB);
 692
 693 out:
 694         err = 0;
 695
 696 err:
 697         spin_unlock(&br->multicast_lock);
 698         return err;
 699 }
 700
 701 static int br_ip4_multicast_add_group(struct net_bridge *br,
 702                                       struct net_bridge_port *port,
 703                                       __be32 group,
 704                                       __u16 vid)
 705 {
 706         struct br_ip br_group;
 707
 708         if (ipv4_is_local_multicast(group))
 709                 return 0;
 710
 711         br_group.u.ip4 = group;
 712         br_group.proto = htons(ETH_P_IP);
 713         br_group.vid = vid;
 714
 715         return br_multicast_add_group(br, port, &br_group);
 716 }
 717
 718 #if IS_ENABLED(CONFIG_IPV6)
 719 static int br_ip6_multicast_add_group(struct net_bridge *br,
 720                                       struct net_bridge_port *port,
 721                                       const struct in6_addr *group,
 722                                       __u16 vid)
 723 {
 724         struct br_ip br_group;
 725
 726         if (!ipv6_is_transient_multicast(group))
 727                 return 0;
 728
 729         br_group.u.ip6 = *group;
 730         br_group.proto = htons(ETH_P_IPV6);
 731         br_group.vid = vid;
 732
 733         return br_multicast_add_group(br, port, &br_group);
 734 }
 735 #endif
 736
 737 static void br_multicast_router_expired(unsigned long data)
 738 {
 739         struct net_bridge_port *port = (void *)data;
 740         struct net_bridge *br = port->br;
 741
 742         spin_lock(&br->multicast_lock);
 743         if (port->multicast_router != 1 ||
 744             timer_pending(&port->multicast_router_timer) ||
 745             hlist_unhashed(&port->rlist))
 746                 goto out;
 747
 748         hlist_del_init_rcu(&port->rlist);
 749
 750 out:
 751         spin_unlock(&br->multicast_lock);
 752 }
 753
 754 static void br_multicast_local_router_expired(unsigned long data)
 755 {
 756 }
 757
 758 static void br_multicast_querier_expired(unsigned long data)
 759 {
 760         struct net_bridge *br = (void *)data;
 761
 762         spin_lock(&br->multicast_lock);
 763         if (!netif_running(br->dev) || br->multicast_disabled)
 764                 goto out;
 765
 766         br_multicast_start_querier(br);
 767
 768 out:
 769         spin_unlock(&br->multicast_lock);
 770 }
 771
 772 static void __br_multicast_send_query(struct net_bridge *br,
 773                                       struct net_bridge_port *port,
 774                                       struct br_ip *ip)
 775 {
 776         struct sk_buff *skb;
 777
 778         skb = br_multicast_alloc_query(br, ip);
 779         if (!skb)
 780                 return;
 781
 782         if (port) {
 783                 __skb_push(skb, sizeof(struct ethhdr));
 784                 skb->dev = port->dev;
 785                 NF_HOOK(NFPROTO_BRIDGE, NF_BR_LOCAL_OUT, skb, NULL, skb->dev,
 786                         dev_queue_xmit);
 787         } else
 788                 netif_rx(skb);
 789 }
 790
 791 static void br_multicast_send_query(struct net_bridge *br,
 792                                     struct net_bridge_port *port, u32 sent)
 793 {
 794         unsigned long time;
 795         struct br_ip br_group;
 796
 797         if (!netif_running(br->dev) || br->multicast_disabled ||
 798             !br->multicast_querier ||
 799             timer_pending(&br->multicast_querier_timer))
 800                 return;
 801
 802         memset(&br_group.u, 0, sizeof(br_group.u));
 803
 804         br_group.proto = htons(ETH_P_IP);
 805         __br_multicast_send_query(br, port, &br_group);
 806
 807 #if IS_ENABLED(CONFIG_IPV6)
 808         br_group.proto = htons(ETH_P_IPV6);
 809         __br_multicast_send_query(br, port, &br_group);
 810 #endif
 811
 812         time = jiffies;
 813         time += sent < br->multicast_startup_query_count ?
 814                 br->multicast_startup_query_interval :
 815                 br->multicast_query_interval;
 816         mod_timer(port ? &port->multicast_query_timer :
 817                          &br->multicast_query_timer, time);
 818 }
 819
 820 static void br_multicast_port_query_expired(unsigned long data)
 821 {
 822         struct net_bridge_port *port = (void *)data;
 823         struct net_bridge *br = port->br;
 824
 825         spin_lock(&br->multicast_lock);
 826         if (port->state == BR_STATE_DISABLED ||
 827             port->state == BR_STATE_BLOCKING)
 828                 goto out;
 829
 830         if (port->multicast_startup_queries_sent <
 831             br->multicast_startup_query_count)
 832                 port->multicast_startup_queries_sent++;
 833
 834         br_multicast_send_query(port->br, port,
 835                                 port->multicast_startup_queries_sent);
 836
 837 out:
 838         spin_unlock(&br->multicast_lock);
 839 }
 840
 841 void br_multicast_add_port(struct net_bridge_port *port)
 842 {
 843         port->multicast_router = 1;
 844
 845         setup_timer(&port->multicast_router_timer, br_multicast_router_expired,
 846                     (unsigned long)port);
 847         setup_timer(&port->multicast_query_timer,
 848                     br_multicast_port_query_expired, (unsigned long)port);
 849 }
 850
 851 void br_multicast_del_port(struct net_bridge_port *port)
 852 {
 853         del_timer_sync(&port->multicast_router_timer);
 854 }
 855
 856 static void __br_multicast_enable_port(struct net_bridge_port *port)
 857 {
 858         port->multicast_startup_queries_sent = 0;
 859
 860         if (try_to_del_timer_sync(&port->multicast_query_timer) >= 0 ||
 861             del_timer(&port->multicast_query_timer))
 862                 mod_timer(&port->multicast_query_timer, jiffies);
 863 }
 864
 865 void br_multicast_enable_port(struct net_bridge_port *port)
 866 {
 867         struct net_bridge *br = port->br;
 868
 869         spin_lock(&br->multicast_lock);
 870         if (br->multicast_disabled || !netif_running(br->dev))
 871                 goto out;
 872
 873         __br_multicast_enable_port(port);
 874
 875 out:
 876         spin_unlock(&br->multicast_lock);
 877 }
 878
 879 void br_multicast_disable_port(struct net_bridge_port *port)
 880 {
 881         struct net_bridge *br = port->br;
 882         struct net_bridge_port_group *pg;
 883         struct hlist_node *n;
 884
 885         spin_lock(&br->multicast_lock);
 886         hlist_for_each_entry_safe(pg, n, &port->mglist, mglist)
 887                 br_multicast_del_pg(br, pg);
 888
 889         if (!hlist_unhashed(&port->rlist))
 890                 hlist_del_init_rcu(&port->rlist);
 891         del_timer(&port->multicast_router_timer);
 892         del_timer(&port->multicast_query_timer);
 893         spin_unlock(&br->multicast_lock);
 894 }
 895
 896 static int br_ip4_multicast_igmp3_report(struct net_bridge *br,
 897                                          struct net_bridge_port *port,
 898                                          struct sk_buff *skb)
 899 {
 900         struct igmpv3_report *ih;
 901         struct igmpv3_grec *grec;
 902         int i;
 903         int len;
 904         int num;
 905         int type;
 906         int err = 0;
 907         __be32 group;
 908         u16 vid = 0;
 909
 910         if (!pskb_may_pull(skb, sizeof(*ih)))
 911                 return -EINVAL;
 912
 913         br_vlan_get_tag(skb, &vid);
 914         ih = igmpv3_report_hdr(skb);
 915         num = ntohs(ih->ngrec);
 916         len = sizeof(*ih);
 917
 918         for (i = 0; i < num; i++) {
 919                 len += sizeof(*grec);
 920                 if (!pskb_may_pull(skb, len))
 921                         return -EINVAL;
 922
 923                 grec = (void *)(skb->data + len - sizeof(*grec));
 924                 group = grec->grec_mca;
 925                 type = grec->grec_type;
 926
 927                 len += ntohs(grec->grec_nsrcs) * 4;
 928                 if (!pskb_may_pull(skb, len))
 929                         return -EINVAL;
 930
 931                 /* We treat this as an IGMPv2 report for now. */
 932                 switch (type) {
 933                 case IGMPV3_MODE_IS_INCLUDE:
 934                 case IGMPV3_MODE_IS_EXCLUDE:
 935                 case IGMPV3_CHANGE_TO_INCLUDE:
 936                 case IGMPV3_CHANGE_TO_EXCLUDE:
 937                 case IGMPV3_ALLOW_NEW_SOURCES:
 938                 case IGMPV3_BLOCK_OLD_SOURCES:
 939                         break;
 940
 941                 default:
 942                         continue;
 943                 }
 944
 945                 err = br_ip4_multicast_add_group(br, port, group, vid);
 946                 if (err)
 947                         break;
 948         }
 949
 950         return err;
 951 }
 952
 953 #if IS_ENABLED(CONFIG_IPV6)
 954 static int br_ip6_multicast_mld2_report(struct net_bridge *br,
 955                                         struct net_bridge_port *port,
 956                                         struct sk_buff *skb)
 957 {
 958         struct icmp6hdr *icmp6h;
 959         struct mld2_grec *grec;
 960         int i;
 961         int len;
 962         int num;
 963         int err = 0;
 964         u16 vid = 0;
 965
 966         if (!pskb_may_pull(skb, sizeof(*icmp6h)))
 967                 return -EINVAL;
 968
 969         br_vlan_get_tag(skb, &vid);
 970         icmp6h = icmp6_hdr(skb);
 971         num = ntohs(icmp6h->icmp6_dataun.un_data16[1]);
 972         len = sizeof(*icmp6h);
 973
 974         for (i = 0; i < num; i++) {
 975                 __be16 *nsrcs, _nsrcs;
 976
 977                 nsrcs = skb_header_pointer(skb,
 978                                            len + offsetof(struct mld2_grec,
 979                                                           grec_nsrcs),
 980                                            sizeof(_nsrcs), &_nsrcs);
 981                 if (!nsrcs)
 982                         return -EINVAL;
 983
 984                 if (!pskb_may_pull(skb,
 985                                    len + sizeof(*grec) +
 986                                    sizeof(struct in6_addr) * ntohs(*nsrcs)))
 987                         return -EINVAL;
 988
 989                 grec = (struct mld2_grec *)(skb->data + len);
 990                 len += sizeof(*grec) +
 991                        sizeof(struct in6_addr) * ntohs(*nsrcs);
 992
 993                 /* We treat these as MLDv1 reports for now. */
 994                 switch (grec->grec_type) {
 995                 case MLD2_MODE_IS_INCLUDE:
 996                 case MLD2_MODE_IS_EXCLUDE:
 997                 case MLD2_CHANGE_TO_INCLUDE:
 998                 case MLD2_CHANGE_TO_EXCLUDE:
 999                 case MLD2_ALLOW_NEW_SOURCES:
1000                 case MLD2_BLOCK_OLD_SOURCES:
1001                         break;
1002
1003                 default:
1004                         continue;
1005                 }
1006
1007                 err = br_ip6_multicast_add_group(br, port, &grec->grec_mca,
1008                                                  vid);
1009                 if (!err)
1010                         break;
1011         }
1012
1013         return err;
1014 }
1015 #endif
1016
1017 static void br_multicast_update_querier_timer(struct net_bridge *br,
1018                                               unsigned long max_delay)
1019 {
1020         if (!timer_pending(&br->multicast_querier_timer))
1021                 br->multicast_querier_delay_time = jiffies + max_delay;
1022
1023         mod_timer(&br->multicast_querier_timer,
1024                   jiffies + br->multicast_querier_interval);
1025 }
1026
1027 /*
1028  * Add port to router_list
1029  *  list is maintained ordered by pointer value
1030  *  and locked by br->multicast_lock and RCU
1031  */
1032 static void br_multicast_add_router(struct net_bridge *br,
1033                                     struct net_bridge_port *port)
1034 {
1035         struct net_bridge_port *p;
1036         struct hlist_node *slot = NULL;
1037
1038         hlist_for_each_entry(p, &br->router_list, rlist) {
1039                 if ((unsigned long) port >= (unsigned long) p)
1040                         break;
1041                 slot = &p->rlist;
1042         }
1043
1044         if (slot)
1045                 hlist_add_after_rcu(slot, &port->rlist);
1046         else
1047                 hlist_add_head_rcu(&port->rlist, &br->router_list);
1048 }
1049
1050 static void br_multicast_mark_router(struct net_bridge *br,
1051                                      struct net_bridge_port *port)
1052 {
1053         unsigned long now = jiffies;
1054
1055         if (!port) {
1056                 if (br->multicast_router == 1)
1057                         mod_timer(&br->multicast_router_timer,
1058                                   now + br->multicast_querier_interval);
1059                 return;
1060         }
1061
1062         if (port->multicast_router != 1)
1063                 return;
1064
1065         if (!hlist_unhashed(&port->rlist))
1066                 goto timer;
1067
1068         br_multicast_add_router(br, port);
1069
1070 timer:
1071         mod_timer(&port->multicast_router_timer,
1072                   now + br->multicast_querier_interval);
1073 }
1074
1075 static void br_multicast_query_received(struct net_bridge *br,
1076                                         struct net_bridge_port *port,
1077                                         int saddr,
1078                                         unsigned long max_delay)
1079 {
1080         if (saddr)
1081                 br_multicast_update_querier_timer(br, max_delay);
1082         else if (timer_pending(&br->multicast_querier_timer))
1083                 return;
1084
1085         br_multicast_mark_router(br, port);
1086 }
1087
1088 static int br_ip4_multicast_query(struct net_bridge *br,
1089                                   struct net_bridge_port *port,
1090                                   struct sk_buff *skb)
1091 {
1092         const struct iphdr *iph = ip_hdr(skb);
1093         struct igmphdr *ih = igmp_hdr(skb);
1094         struct net_bridge_mdb_entry *mp;
1095         struct igmpv3_query *ih3;
1096         struct net_bridge_port_group *p;
1097         struct net_bridge_port_group __rcu **pp;
1098         unsigned long max_delay;
1099         unsigned long now = jiffies;
1100         __be32 group;
1101         int err = 0;
1102         u16 vid = 0;
1103
1104         spin_lock(&br->multicast_lock);
1105         if (!netif_running(br->dev) ||
1106             (port && port->state == BR_STATE_DISABLED))
1107                 goto out;
1108
1109         group = ih->group;
1110
1111         if (skb->len == sizeof(*ih)) {
1112                 max_delay = ih->code * (HZ / IGMP_TIMER_SCALE);
1113
1114                 if (!max_delay) {
1115                         max_delay = 10 * HZ;
1116                         group = 0;
1117                 }
1118         } else {
1119                 if (!pskb_may_pull(skb, sizeof(struct igmpv3_query))) {
1120                         err = -EINVAL;
1121                         goto out;
1122                 }
1123
1124                 ih3 = igmpv3_query_hdr(skb);
1125                 if (ih3->nsrcs)
1126                         goto out;
1127
1128                 max_delay = ih3->code ?
1129                             IGMPV3_MRC(ih3->code) * (HZ / IGMP_TIMER_SCALE) : 1;
1130         }
1131
1132         br_multicast_query_received(br, port, !!iph->saddr, max_delay);
1133
1134         if (!group)
1135                 goto out;
1136
1137         br_vlan_get_tag(skb, &vid);
1138         mp = br_mdb_ip4_get(mlock_dereference(br->mdb, br), group, vid);
1139         if (!mp)
1140                 goto out;
1141
1142         mod_timer(&mp->timer, now + br->multicast_membership_interval);
1143         mp->timer_armed = true;
1144
1145         max_delay *= br->multicast_last_member_count;
1146
1147         if (mp->mglist &&
1148             (timer_pending(&mp->timer) ?
1149              time_after(mp->timer.expires, now + max_delay) :
1150              try_to_del_timer_sync(&mp->timer) >= 0))
1151                 mod_timer(&mp->timer, now + max_delay);
1152
1153         for (pp = &mp->ports;
1154              (p = mlock_dereference(*pp, br)) != NULL;
1155              pp = &p->next) {
1156                 if (timer_pending(&p->timer) ?
1157                     time_after(p->timer.expires, now + max_delay) :
1158                     try_to_del_timer_sync(&p->timer) >= 0)
1159                         mod_timer(&p->timer, now + max_delay);
1160         }
1161
1162 out:
1163         spin_unlock(&br->multicast_lock);
1164         return err;
1165 }
1166
1167 #if IS_ENABLED(CONFIG_IPV6)
1168 static int br_ip6_multicast_query(struct net_bridge *br,
1169                                   struct net_bridge_port *port,
1170                                   struct sk_buff *skb)
1171 {
1172         const struct ipv6hdr *ip6h = ipv6_hdr(skb);
1173         struct mld_msg *mld;
1174         struct net_bridge_mdb_entry *mp;
1175         struct mld2_query *mld2q;
1176         struct net_bridge_port_group *p;
1177         struct net_bridge_port_group __rcu **pp;
1178         unsigned long max_delay;
1179         unsigned long now = jiffies;
1180         const struct in6_addr *group = NULL;
1181         int err = 0;
1182         u16 vid = 0;
1183
1184         spin_lock(&br->multicast_lock);
1185         if (!netif_running(br->dev) ||
1186             (port && port->state == BR_STATE_DISABLED))
1187                 goto out;
1188
1189         if (skb->len == sizeof(*mld)) {
1190                 if (!pskb_may_pull(skb, sizeof(*mld))) {
1191                         err = -EINVAL;
1192                         goto out;
1193                 }
1194                 mld = (struct mld_msg *) icmp6_hdr(skb);
1195                 max_delay = msecs_to_jiffies(ntohs(mld->mld_maxdelay));
1196                 if (max_delay)
1197                         group = &mld->mld_mca;
1198         } else if (skb->len >= sizeof(*mld2q)) {
1199                 if (!pskb_may_pull(skb, sizeof(*mld2q))) {
1200                         err = -EINVAL;
1201                         goto out;
1202                 }
1203                 mld2q = (struct mld2_query *)icmp6_hdr(skb);
1204                 if (!mld2q->mld2q_nsrcs)
1205                         group = &mld2q->mld2q_mca;
1206                 max_delay = mld2q->mld2q_mrc ? MLDV2_MRC(ntohs(mld2q->mld2q_mrc)) : 1;
1207         }
1208
1209         br_multicast_query_received(br, port, !ipv6_addr_any(&ip6h->saddr),
1210                                     max_delay);
1211
1212         if (!group)
1213                 goto out;
1214
1215         br_vlan_get_tag(skb, &vid);
1216         mp = br_mdb_ip6_get(mlock_dereference(br->mdb, br), group, vid);
1217         if (!mp)
1218                 goto out;
1219
1220         mod_timer(&mp->timer, now + br->multicast_membership_interval);
1221         mp->timer_armed = true;
1222
1223         max_delay *= br->multicast_last_member_count;
1224         if (mp->mglist &&
1225             (timer_pending(&mp->timer) ?
1226              time_after(mp->timer.expires, now + max_delay) :
1227              try_to_del_timer_sync(&mp->timer) >= 0))
1228                 mod_timer(&mp->timer, now + max_delay);
1229
1230         for (pp = &mp->ports;
1231              (p = mlock_dereference(*pp, br)) != NULL;
1232              pp = &p->next) {
1233                 if (timer_pending(&p->timer) ?
1234                     time_after(p->timer.expires, now + max_delay) :
1235                     try_to_del_timer_sync(&p->timer) >= 0)
1236                         mod_timer(&p->timer, now + max_delay);
1237         }
1238
1239 out:
1240         spin_unlock(&br->multicast_lock);
1241         return err;
1242 }
1243 #endif
1244
1245 static void br_multicast_leave_group(struct net_bridge *br,
1246                                      struct net_bridge_port *port,
1247                                      struct br_ip *group)
1248 {
1249         struct net_bridge_mdb_htable *mdb;
1250         struct net_bridge_mdb_entry *mp;
1251         struct net_bridge_port_group *p;
1252         unsigned long now;
1253         unsigned long time;
1254
1255         spin_lock(&br->multicast_lock);
1256         if (!netif_running(br->dev) ||
1257             (port && port->state == BR_STATE_DISABLED) ||
1258             timer_pending(&br->multicast_querier_timer))
1259                 goto out;
1260
1261         mdb = mlock_dereference(br->mdb, br);
1262         mp = br_mdb_ip_get(mdb, group);
1263         if (!mp)
1264                 goto out;
1265
1266         if (br->multicast_querier &&
1267             !timer_pending(&br->multicast_querier_timer)) {
1268                 __br_multicast_send_query(br, port, &mp->addr);
1269
1270                 time = jiffies + br->multicast_last_member_count *
1271                                  br->multicast_last_member_interval;
1272                 mod_timer(port ? &port->multicast_query_timer :
1273                                  &br->multicast_query_timer, time);
1274
1275                 for (p = mlock_dereference(mp->ports, br);
1276                      p != NULL;
1277                      p = mlock_dereference(p->next, br)) {
1278                         if (p->port != port)
1279                                 continue;
1280
1281                         if (!hlist_unhashed(&p->mglist) &&
1282                             (timer_pending(&p->timer) ?
1283                              time_after(p->timer.expires, time) :
1284                              try_to_del_timer_sync(&p->timer) >= 0)) {
1285                                 mod_timer(&p->timer, time);
1286                         }
1287
1288                         break;
1289                 }
1290         }
1291
1292         if (port && (port->flags & BR_MULTICAST_FAST_LEAVE)) {
1293                 struct net_bridge_port_group __rcu **pp;
1294
1295                 for (pp = &mp->ports;
1296                      (p = mlock_dereference(*pp, br)) != NULL;
1297                      pp = &p->next) {
1298                         if (p->port != port)
1299                                 continue;
1300
1301                         rcu_assign_pointer(*pp, p->next);
1302                         hlist_del_init(&p->mglist);
1303                         del_timer(&p->timer);
1304                         call_rcu_bh(&p->rcu, br_multicast_free_pg);
1305                         br_mdb_notify(br->dev, port, group, RTM_DELMDB);
1306
1307                         if (!mp->ports && !mp->mglist && mp->timer_armed &&
1308                             netif_running(br->dev))
1309                                 mod_timer(&mp->timer, jiffies);
1310                 }
1311                 goto out;
1312         }
1313
1314         now = jiffies;
1315         time = now + br->multicast_last_member_count *
1316                      br->multicast_last_member_interval;
1317
1318         if (!port) {
1319                 if (mp->mglist && mp->timer_armed &&
1320                     (timer_pending(&mp->timer) ?
1321                      time_after(mp->timer.expires, time) :
1322                      try_to_del_timer_sync(&mp->timer) >= 0)) {
1323                         mod_timer(&mp->timer, time);
1324                 }
1325         }
1326
1327 out:
1328         spin_unlock(&br->multicast_lock);
1329 }
1330
1331 static void br_ip4_multicast_leave_group(struct net_bridge *br,
1332                                          struct net_bridge_port *port,
1333                                          __be32 group,
1334                                          __u16 vid)
1335 {
1336         struct br_ip br_group;
1337
1338         if (ipv4_is_local_multicast(group))
1339                 return;
1340
1341         br_group.u.ip4 = group;
1342         br_group.proto = htons(ETH_P_IP);
1343         br_group.vid = vid;
1344
1345         br_multicast_leave_group(br, port, &br_group);
1346 }
1347
1348 #if IS_ENABLED(CONFIG_IPV6)
1349 static void br_ip6_multicast_leave_group(struct net_bridge *br,
1350                                          struct net_bridge_port *port,
1351                                          const struct in6_addr *group,
1352                                          __u16 vid)
1353 {
1354         struct br_ip br_group;
1355
1356         if (!ipv6_is_transient_multicast(group))
1357                 return;
1358
1359         br_group.u.ip6 = *group;
1360         br_group.proto = htons(ETH_P_IPV6);
1361         br_group.vid = vid;
1362
1363         br_multicast_leave_group(br, port, &br_group);
1364 }
1365 #endif
1366
1367 static int br_multicast_ipv4_rcv(struct net_bridge *br,
1368                                  struct net_bridge_port *port,
1369                                  struct sk_buff *skb)
1370 {
1371         struct sk_buff *skb2 = skb;
1372         const struct iphdr *iph;
1373         struct igmphdr *ih;
1374         unsigned int len;
1375         unsigned int offset;
1376         int err;
1377         u16 vid = 0;
1378
1379         /* We treat OOM as packet loss for now. */
1380         if (!pskb_may_pull(skb, sizeof(*iph)))
1381                 return -EINVAL;
1382
1383         iph = ip_hdr(skb);
1384
1385         if (iph->ihl < 5 || iph->version != 4)
1386                 return -EINVAL;
1387
1388         if (!pskb_may_pull(skb, ip_hdrlen(skb)))
1389                 return -EINVAL;
1390
1391         iph = ip_hdr(skb);
1392
1393         if (unlikely(ip_fast_csum((u8 *)iph, iph->ihl)))
1394                 return -EINVAL;
1395
1396         if (iph->protocol != IPPROTO_IGMP) {
1397                 if (!ipv4_is_local_multicast(iph->daddr))
1398                         BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1399                 return 0;
1400         }
1401
1402         len = ntohs(iph->tot_len);
1403         if (skb->len < len || len < ip_hdrlen(skb))
1404                 return -EINVAL;
1405
1406         if (skb->len > len) {
1407                 skb2 = skb_clone(skb, GFP_ATOMIC);
1408                 if (!skb2)
1409                         return -ENOMEM;
1410
1411                 err = pskb_trim_rcsum(skb2, len);
1412                 if (err)
1413                         goto err_out;
1414         }
1415
1416         len -= ip_hdrlen(skb2);
1417         offset = skb_network_offset(skb2) + ip_hdrlen(skb2);
1418         __skb_pull(skb2, offset);
1419         skb_reset_transport_header(skb2);
1420
1421         err = -EINVAL;
1422         if (!pskb_may_pull(skb2, sizeof(*ih)))
1423                 goto out;
1424
1425         switch (skb2->ip_summed) {
1426         case CHECKSUM_COMPLETE:
1427                 if (!csum_fold(skb2->csum))
1428                         break;
1429                 /* fall through */
1430         case CHECKSUM_NONE:
1431                 skb2->csum = 0;
1432                 if (skb_checksum_complete(skb2))
1433                         goto out;
1434         }
1435
1436         err = 0;
1437
1438         br_vlan_get_tag(skb2, &vid);
1439         BR_INPUT_SKB_CB(skb)->igmp = 1;
1440         ih = igmp_hdr(skb2);
1441
1442         switch (ih->type) {
1443         case IGMP_HOST_MEMBERSHIP_REPORT:
1444         case IGMPV2_HOST_MEMBERSHIP_REPORT:
1445                 BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1446                 err = br_ip4_multicast_add_group(br, port, ih->group, vid);
1447                 break;
1448         case IGMPV3_HOST_MEMBERSHIP_REPORT:
1449                 err = br_ip4_multicast_igmp3_report(br, port, skb2);
1450                 break;
1451         case IGMP_HOST_MEMBERSHIP_QUERY:
1452                 err = br_ip4_multicast_query(br, port, skb2);
1453                 break;
1454         case IGMP_HOST_LEAVE_MESSAGE:
1455                 br_ip4_multicast_leave_group(br, port, ih->group, vid);
1456                 break;
1457         }
1458
1459 out:
1460         __skb_push(skb2, offset);
1461 err_out:
1462         if (skb2 != skb)
1463                 kfree_skb(skb2);
1464         return err;
1465 }
1466
1467 #if IS_ENABLED(CONFIG_IPV6)
1468 static int br_multicast_ipv6_rcv(struct net_bridge *br,
1469                                  struct net_bridge_port *port,
1470                                  struct sk_buff *skb)
1471 {
1472         struct sk_buff *skb2;
1473         const struct ipv6hdr *ip6h;
1474         u8 icmp6_type;
1475         u8 nexthdr;
1476         __be16 frag_off;
1477         unsigned int len;
1478         int offset;
1479         int err;
1480         u16 vid = 0;
1481
1482         if (!pskb_may_pull(skb, sizeof(*ip6h)))
1483                 return -EINVAL;
1484
1485         ip6h = ipv6_hdr(skb);
1486
1487         /*
1488          * We're interested in MLD messages only.
1489          *  - Version is 6
1490          *  - MLD has always Router Alert hop-by-hop option
1491          *  - But we do not support jumbrograms.
1492          */
1493         if (ip6h->version != 6 ||
1494             ip6h->nexthdr != IPPROTO_HOPOPTS ||
1495             ip6h->payload_len == 0)
1496                 return 0;
1497
1498         len = ntohs(ip6h->payload_len) + sizeof(*ip6h);
1499         if (skb->len < len)
1500                 return -EINVAL;
1501
1502         nexthdr = ip6h->nexthdr;
1503         offset = ipv6_skip_exthdr(skb, sizeof(*ip6h), &nexthdr, &frag_off);
1504
1505         if (offset < 0 || nexthdr != IPPROTO_ICMPV6)
1506                 return 0;
1507
1508         /* Okay, we found ICMPv6 header */
1509         skb2 = skb_clone(skb, GFP_ATOMIC);
1510         if (!skb2)
1511                 return -ENOMEM;
1512
1513         err = -EINVAL;
1514         if (!pskb_may_pull(skb2, offset + sizeof(struct icmp6hdr)))
1515                 goto out;
1516
1517         len -= offset - skb_network_offset(skb2);
1518
1519         __skb_pull(skb2, offset);
1520         skb_reset_transport_header(skb2);
1521         skb_postpull_rcsum(skb2, skb_network_header(skb2),
1522                            skb_network_header_len(skb2));
1523
1524         icmp6_type = icmp6_hdr(skb2)->icmp6_type;
1525
1526         switch (icmp6_type) {
1527         case ICMPV6_MGM_QUERY:
1528         case ICMPV6_MGM_REPORT:
1529         case ICMPV6_MGM_REDUCTION:
1530         case ICMPV6_MLD2_REPORT:
1531                 break;
1532         default:
1533                 err = 0;
1534                 goto out;
1535         }
1536
1537         /* Okay, we found MLD message. Check further. */
1538         if (skb2->len > len) {
1539                 err = pskb_trim_rcsum(skb2, len);
1540                 if (err)
1541                         goto out;
1542                 err = -EINVAL;
1543         }
1544
1545         ip6h = ipv6_hdr(skb2);
1546
1547         switch (skb2->ip_summed) {
1548         case CHECKSUM_COMPLETE:
1549                 if (!csum_ipv6_magic(&ip6h->saddr, &ip6h->daddr, skb2->len,
1550                                         IPPROTO_ICMPV6, skb2->csum))
1551                         break;
1552                 /*FALLTHROUGH*/
1553         case CHECKSUM_NONE:
1554                 skb2->csum = ~csum_unfold(csum_ipv6_magic(&ip6h->saddr,
1555                                                         &ip6h->daddr,
1556                                                         skb2->len,
1557                                                         IPPROTO_ICMPV6, 0));
1558                 if (__skb_checksum_complete(skb2))
1559                         goto out;
1560         }
1561
1562         err = 0;
1563
1564         br_vlan_get_tag(skb, &vid);
1565         BR_INPUT_SKB_CB(skb)->igmp = 1;
1566
1567         switch (icmp6_type) {
1568         case ICMPV6_MGM_REPORT:
1569             {
1570                 struct mld_msg *mld;
1571                 if (!pskb_may_pull(skb2, sizeof(*mld))) {
1572                         err = -EINVAL;
1573                         goto out;
1574                 }
1575                 mld = (struct mld_msg *)skb_transport_header(skb2);
1576                 BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1577                 err = br_ip6_multicast_add_group(br, port, &mld->mld_mca, vid);
1578                 break;
1579             }
1580         case ICMPV6_MLD2_REPORT:
1581                 err = br_ip6_multicast_mld2_report(br, port, skb2);
1582                 break;
1583         case ICMPV6_MGM_QUERY:
1584                 err = br_ip6_multicast_query(br, port, skb2);
1585                 break;
1586         case ICMPV6_MGM_REDUCTION:
1587             {
1588                 struct mld_msg *mld;
1589                 if (!pskb_may_pull(skb2, sizeof(*mld))) {
1590                         err = -EINVAL;
1591                         goto out;
1592                 }
1593                 mld = (struct mld_msg *)skb_transport_header(skb2);
1594                 br_ip6_multicast_leave_group(br, port, &mld->mld_mca, vid);
1595             }
1596         }
1597
1598 out:
1599         kfree_skb(skb2);
1600         return err;
1601 }
1602 #endif
1603
1604 int br_multicast_rcv(struct net_bridge *br, struct net_bridge_port *port,
1605                      struct sk_buff *skb)
1606 {
1607         BR_INPUT_SKB_CB(skb)->igmp = 0;
1608         BR_INPUT_SKB_CB(skb)->mrouters_only = 0;
1609
1610         if (br->multicast_disabled)
1611                 return 0;
1612
1613         switch (skb->protocol) {
1614         case htons(ETH_P_IP):
1615                 return br_multicast_ipv4_rcv(br, port, skb);
1616 #if IS_ENABLED(CONFIG_IPV6)
1617         case htons(ETH_P_IPV6):
1618                 return br_multicast_ipv6_rcv(br, port, skb);
1619 #endif
1620         }
1621
1622         return 0;
1623 }
1624
1625 static void br_multicast_query_expired(unsigned long data)
1626 {
1627         struct net_bridge *br = (void *)data;
1628
1629         spin_lock(&br->multicast_lock);
1630         if (br->multicast_startup_queries_sent <
1631             br->multicast_startup_query_count)
1632                 br->multicast_startup_queries_sent++;
1633
1634         br_multicast_send_query(br, NULL, br->multicast_startup_queries_sent);
1635
1636         spin_unlock(&br->multicast_lock);
1637 }
1638
1639 void br_multicast_init(struct net_bridge *br)
1640 {
1641         br->hash_elasticity = 4;
1642         br->hash_max = 512;
1643
1644         br->multicast_router = 1;
1645         br->multicast_querier = 0;
1646         br->multicast_query_use_ifaddr = 0;
1647         br->multicast_last_member_count = 2;
1648         br->multicast_startup_query_count = 2;
1649
1650         br->multicast_last_member_interval = HZ;
1651         br->multicast_query_response_interval = 10 * HZ;
1652         br->multicast_startup_query_interval = 125 * HZ / 4;
1653         br->multicast_query_interval = 125 * HZ;
1654         br->multicast_querier_interval = 255 * HZ;
1655         br->multicast_membership_interval = 260 * HZ;
1656
1657         br->multicast_querier_delay_time = 0;
1658
1659         spin_lock_init(&br->multicast_lock);
1660         setup_timer(&br->multicast_router_timer,
1661                     br_multicast_local_router_expired, 0);
1662         setup_timer(&br->multicast_querier_timer,
1663                     br_multicast_querier_expired, (unsigned long)br);
1664         setup_timer(&br->multicast_query_timer, br_multicast_query_expired,
1665                     (unsigned long)br);
1666 }
1667
1668 void br_multicast_open(struct net_bridge *br)
1669 {
1670         br->multicast_startup_queries_sent = 0;
1671
1672         if (br->multicast_disabled)
1673                 return;
1674
1675         mod_timer(&br->multicast_query_timer, jiffies);
1676 }
1677
1678 void br_multicast_stop(struct net_bridge *br)
1679 {
1680         struct net_bridge_mdb_htable *mdb;
1681         struct net_bridge_mdb_entry *mp;
1682         struct hlist_node *n;
1683         u32 ver;
1684         int i;
1685
1686         del_timer_sync(&br->multicast_router_timer);
1687         del_timer_sync(&br->multicast_querier_timer);
1688         del_timer_sync(&br->multicast_query_timer);
1689
1690         spin_lock_bh(&br->multicast_lock);
1691         mdb = mlock_dereference(br->mdb, br);
1692         if (!mdb)
1693                 goto out;
1694
1695         br->mdb = NULL;
1696
1697         ver = mdb->ver;
1698         for (i = 0; i < mdb->max; i++) {
1699                 hlist_for_each_entry_safe(mp, n, &mdb->mhash[i],
1700                                           hlist[ver]) {
1701                         del_timer(&mp->timer);
1702                         mp->timer_armed = false;
1703                         call_rcu_bh(&mp->rcu, br_multicast_free_group);
1704                 }
1705         }
1706
1707         if (mdb->old) {
1708                 spin_unlock_bh(&br->multicast_lock);
1709                 rcu_barrier_bh();
1710                 spin_lock_bh(&br->multicast_lock);
1711                 WARN_ON(mdb->old);
1712         }
1713
1714         mdb->old = mdb;
1715         call_rcu_bh(&mdb->rcu, br_mdb_free);
1716
1717 out:
1718         spin_unlock_bh(&br->multicast_lock);
1719 }
1720
1721 int br_multicast_set_router(struct net_bridge *br, unsigned long val)
1722 {
1723         int err = -ENOENT;
1724
1725         spin_lock_bh(&br->multicast_lock);
1726         if (!netif_running(br->dev))
1727                 goto unlock;
1728
1729         switch (val) {
1730         case 0:
1731         case 2:
1732                 del_timer(&br->multicast_router_timer);
1733                 /* fall through */
1734         case 1:
1735                 br->multicast_router = val;
1736                 err = 0;
1737                 break;
1738
1739         default:
1740                 err = -EINVAL;
1741                 break;
1742         }
1743
1744 unlock:
1745         spin_unlock_bh(&br->multicast_lock);
1746
1747         return err;
1748 }
1749
1750 int br_multicast_set_port_router(struct net_bridge_port *p, unsigned long val)
1751 {
1752         struct net_bridge *br = p->br;
1753         int err = -ENOENT;
1754
1755         spin_lock(&br->multicast_lock);
1756         if (!netif_running(br->dev) || p->state == BR_STATE_DISABLED)
1757                 goto unlock;
1758
1759         switch (val) {
1760         case 0:
1761         case 1:
1762         case 2:
1763                 p->multicast_router = val;
1764                 err = 0;
1765
1766                 if (val < 2 && !hlist_unhashed(&p->rlist))
1767                         hlist_del_init_rcu(&p->rlist);
1768
1769                 if (val == 1)
1770                         break;
1771
1772                 del_timer(&p->multicast_router_timer);
1773
1774                 if (val == 0)
1775                         break;
1776
1777                 br_multicast_add_router(br, p);
1778                 break;
1779
1780         default:
1781                 err = -EINVAL;
1782                 break;
1783         }
1784
1785 unlock:
1786         spin_unlock(&br->multicast_lock);
1787
1788         return err;
1789 }
1790
1791 static void br_multicast_start_querier(struct net_bridge *br)
1792 {
1793         struct net_bridge_port *port;
1794
1795         br_multicast_open(br);
1796
1797         list_for_each_entry(port, &br->port_list, list) {
1798                 if (port->state == BR_STATE_DISABLED ||
1799                     port->state == BR_STATE_BLOCKING)
1800                         continue;
1801
1802                 __br_multicast_enable_port(port);
1803         }
1804 }
1805
1806 int br_multicast_toggle(struct net_bridge *br, unsigned long val)
1807 {
1808         int err = 0;
1809         struct net_bridge_mdb_htable *mdb;
1810
1811         spin_lock_bh(&br->multicast_lock);
1812         if (br->multicast_disabled == !val)
1813                 goto unlock;
1814
1815         br->multicast_disabled = !val;
1816         if (br->multicast_disabled)
1817                 goto unlock;
1818
1819         if (!netif_running(br->dev))
1820                 goto unlock;
1821
1822         mdb = mlock_dereference(br->mdb, br);
1823         if (mdb) {
1824                 if (mdb->old) {
1825                         err = -EEXIST;
1826 rollback:
1827                         br->multicast_disabled = !!val;
1828                         goto unlock;
1829                 }
1830
1831                 err = br_mdb_rehash(&br->mdb, mdb->max,
1832                                     br->hash_elasticity);
1833                 if (err)
1834                         goto rollback;
1835         }
1836
1837         br_multicast_start_querier(br);
1838
1839 unlock:
1840         spin_unlock_bh(&br->multicast_lock);
1841
1842         return err;
1843 }
1844
1845 int br_multicast_set_querier(struct net_bridge *br, unsigned long val)
1846 {
1847         unsigned long max_delay;
1848
1849         val = !!val;
1850
1851         spin_lock_bh(&br->multicast_lock);
1852         if (br->multicast_querier == val)
1853                 goto unlock;
1854
1855         br->multicast_querier = val;
1856         if (!val)
1857                 goto unlock;
1858
1859         max_delay = br->multicast_query_response_interval;
1860         if (!timer_pending(&br->multicast_querier_timer))
1861                 br->multicast_querier_delay_time = jiffies + max_delay;
1862
1863         br_multicast_start_querier(br);
1864
1865 unlock:
1866         spin_unlock_bh(&br->multicast_lock);
1867
1868         return 0;
1869 }
1870
1871 int br_multicast_set_hash_max(struct net_bridge *br, unsigned long val)
1872 {
1873         int err = -ENOENT;
1874         u32 old;
1875         struct net_bridge_mdb_htable *mdb;
1876
1877         spin_lock(&br->multicast_lock);
1878         if (!netif_running(br->dev))
1879                 goto unlock;
1880
1881         err = -EINVAL;
1882         if (!is_power_of_2(val))
1883                 goto unlock;
1884
1885         mdb = mlock_dereference(br->mdb, br);
1886         if (mdb && val < mdb->size)
1887                 goto unlock;
1888
1889         err = 0;
1890
1891         old = br->hash_max;
1892         br->hash_max = val;
1893
1894         if (mdb) {
1895                 if (mdb->old) {
1896                         err = -EEXIST;
1897 rollback:
1898                         br->hash_max = old;
1899                         goto unlock;
1900                 }
1901
1902                 err = br_mdb_rehash(&br->mdb, br->hash_max,
1903                                     br->hash_elasticity);
1904                 if (err)
1905                         goto rollback;
1906         }
1907
1908 unlock:
1909         spin_unlock(&br->multicast_lock);
1910
1911         return err;
1912 }