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                                        struct bridge_mcast_query *query);
  38 unsigned int br_mdb_rehash_seq;
  39
  40 static inline int br_ip_equal(const struct br_ip *a, const struct br_ip *b)
  41 {
  42         if (a->proto != b->proto)
  43                 return 0;
  44         if (a->vid != b->vid)
  45                 return 0;
  46         switch (a->proto) {
  47         case htons(ETH_P_IP):
  48                 return a->u.ip4 == b->u.ip4;
  49 #if IS_ENABLED(CONFIG_IPV6)
  50         case htons(ETH_P_IPV6):
  51                 return ipv6_addr_equal(&a->u.ip6, &b->u.ip6);
  52 #endif
  53         }
  54         return 0;
  55 }
  56
  57 static inline int __br_ip4_hash(struct net_bridge_mdb_htable *mdb, __be32 ip,
  58                                 __u16 vid)
  59 {
  60         return jhash_2words((__force u32)ip, vid, mdb->secret) & (mdb->max - 1);
  61 }
  62
  63 #if IS_ENABLED(CONFIG_IPV6)
  64 static inline int __br_ip6_hash(struct net_bridge_mdb_htable *mdb,
  65                                 const struct in6_addr *ip,
  66                                 __u16 vid)
  67 {
  68         return jhash_2words(ipv6_addr_hash(ip), vid,
  69                             mdb->secret) & (mdb->max - 1);
  70 }
  71 #endif
  72
  73 static inline int br_ip_hash(struct net_bridge_mdb_htable *mdb,
  74                              struct br_ip *ip)
  75 {
  76         switch (ip->proto) {
  77         case htons(ETH_P_IP):
  78                 return __br_ip4_hash(mdb, ip->u.ip4, ip->vid);
  79 #if IS_ENABLED(CONFIG_IPV6)
  80         case htons(ETH_P_IPV6):
  81                 return __br_ip6_hash(mdb, &ip->u.ip6, ip->vid);
  82 #endif
  83         }
  84         return 0;
  85 }
  86
  87 static struct net_bridge_mdb_entry *__br_mdb_ip_get(
  88         struct net_bridge_mdb_htable *mdb, struct br_ip *dst, int hash)
  89 {
  90         struct net_bridge_mdb_entry *mp;
  91
  92         hlist_for_each_entry_rcu(mp, &mdb->mhash[hash], hlist[mdb->ver]) {
  93                 if (br_ip_equal(&mp->addr, dst))
  94                         return mp;
  95         }
  96
  97         return NULL;
  98 }
  99
 100 struct net_bridge_mdb_entry *br_mdb_ip_get(struct net_bridge_mdb_htable *mdb,
 101                                            struct br_ip *dst)
 102 {
 103         if (!mdb)
 104                 return NULL;
 105
 106         return __br_mdb_ip_get(mdb, dst, br_ip_hash(mdb, dst));
 107 }
 108
 109 static struct net_bridge_mdb_entry *br_mdb_ip4_get(
 110         struct net_bridge_mdb_htable *mdb, __be32 dst, __u16 vid)
 111 {
 112         struct br_ip br_dst;
 113
 114         br_dst.u.ip4 = dst;
 115         br_dst.proto = htons(ETH_P_IP);
 116         br_dst.vid = vid;
 117
 118         return br_mdb_ip_get(mdb, &br_dst);
 119 }
 120
 121 #if IS_ENABLED(CONFIG_IPV6)
 122 static struct net_bridge_mdb_entry *br_mdb_ip6_get(
 123         struct net_bridge_mdb_htable *mdb, const struct in6_addr *dst,
 124         __u16 vid)
 125 {
 126         struct br_ip br_dst;
 127
 128         br_dst.u.ip6 = *dst;
 129         br_dst.proto = htons(ETH_P_IPV6);
 130         br_dst.vid = vid;
 131
 132         return br_mdb_ip_get(mdb, &br_dst);
 133 }
 134 #endif
 135
 136 struct net_bridge_mdb_entry *br_mdb_get(struct net_bridge *br,
 137                                         struct sk_buff *skb, u16 vid)
 138 {
 139         struct net_bridge_mdb_htable *mdb = rcu_dereference(br->mdb);
 140         struct br_ip ip;
 141
 142         if (br->multicast_disabled)
 143                 return NULL;
 144
 145         if (BR_INPUT_SKB_CB(skb)->igmp)
 146                 return NULL;
 147
 148         ip.proto = skb->protocol;
 149         ip.vid = vid;
 150
 151         switch (skb->protocol) {
 152         case htons(ETH_P_IP):
 153                 ip.u.ip4 = ip_hdr(skb)->daddr;
 154                 break;
 155 #if IS_ENABLED(CONFIG_IPV6)
 156         case htons(ETH_P_IPV6):
 157                 ip.u.ip6 = ipv6_hdr(skb)->daddr;
 158                 break;
 159 #endif
 160         default:
 161                 return NULL;
 162         }
 163
 164         return br_mdb_ip_get(mdb, &ip);
 165 }
 166
 167 static void br_mdb_free(struct rcu_head *head)
 168 {
 169         struct net_bridge_mdb_htable *mdb =
 170                 container_of(head, struct net_bridge_mdb_htable, rcu);
 171         struct net_bridge_mdb_htable *old = mdb->old;
 172
 173         mdb->old = NULL;
 174         kfree(old->mhash);
 175         kfree(old);
 176 }
 177
 178 static int br_mdb_copy(struct net_bridge_mdb_htable *new,
 179                        struct net_bridge_mdb_htable *old,
 180                        int elasticity)
 181 {
 182         struct net_bridge_mdb_entry *mp;
 183         int maxlen;
 184         int len;
 185         int i;
 186
 187         for (i = 0; i < old->max; i++)
 188                 hlist_for_each_entry(mp, &old->mhash[i], hlist[old->ver])
 189                         hlist_add_head(&mp->hlist[new->ver],
 190                                        &new->mhash[br_ip_hash(new, &mp->addr)]);
 191
 192         if (!elasticity)
 193                 return 0;
 194
 195         maxlen = 0;
 196         for (i = 0; i < new->max; i++) {
 197                 len = 0;
 198                 hlist_for_each_entry(mp, &new->mhash[i], hlist[new->ver])
 199                         len++;
 200                 if (len > maxlen)
 201                         maxlen = len;
 202         }
 203
 204         return maxlen > elasticity ? -EINVAL : 0;
 205 }
 206
 207 void br_multicast_free_pg(struct rcu_head *head)
 208 {
 209         struct net_bridge_port_group *p =
 210                 container_of(head, struct net_bridge_port_group, rcu);
 211
 212         kfree(p);
 213 }
 214
 215 static void br_multicast_free_group(struct rcu_head *head)
 216 {
 217         struct net_bridge_mdb_entry *mp =
 218                 container_of(head, struct net_bridge_mdb_entry, rcu);
 219
 220         kfree(mp);
 221 }
 222
 223 static void br_multicast_group_expired(unsigned long data)
 224 {
 225         struct net_bridge_mdb_entry *mp = (void *)data;
 226         struct net_bridge *br = mp->br;
 227         struct net_bridge_mdb_htable *mdb;
 228
 229         spin_lock(&br->multicast_lock);
 230         if (!netif_running(br->dev) || timer_pending(&mp->timer))
 231                 goto out;
 232
 233         mp->mglist = false;
 234
 235         if (mp->ports)
 236                 goto out;
 237
 238         mdb = mlock_dereference(br->mdb, br);
 239
 240         hlist_del_rcu(&mp->hlist[mdb->ver]);
 241         mdb->size--;
 242
 243         call_rcu_bh(&mp->rcu, br_multicast_free_group);
 244
 245 out:
 246         spin_unlock(&br->multicast_lock);
 247 }
 248
 249 static void br_multicast_del_pg(struct net_bridge *br,
 250                                 struct net_bridge_port_group *pg)
 251 {
 252         struct net_bridge_mdb_htable *mdb;
 253         struct net_bridge_mdb_entry *mp;
 254         struct net_bridge_port_group *p;
 255         struct net_bridge_port_group __rcu **pp;
 256
 257         mdb = mlock_dereference(br->mdb, br);
 258
 259         mp = br_mdb_ip_get(mdb, &pg->addr);
 260         if (WARN_ON(!mp))
 261                 return;
 262
 263         for (pp = &mp->ports;
 264              (p = mlock_dereference(*pp, br)) != NULL;
 265              pp = &p->next) {
 266                 if (p != pg)
 267                         continue;
 268
 269                 rcu_assign_pointer(*pp, p->next);
 270                 hlist_del_init(&p->mglist);
 271                 del_timer(&p->timer);
 272                 call_rcu_bh(&p->rcu, br_multicast_free_pg);
 273
 274                 if (!mp->ports && !mp->mglist && mp->timer_armed &&
 275                     netif_running(br->dev))
 276                         mod_timer(&mp->timer, jiffies);
 277
 278                 return;
 279         }
 280
 281         WARN_ON(1);
 282 }
 283
 284 static void br_multicast_port_group_expired(unsigned long data)
 285 {
 286         struct net_bridge_port_group *pg = (void *)data;
 287         struct net_bridge *br = pg->port->br;
 288
 289         spin_lock(&br->multicast_lock);
 290         if (!netif_running(br->dev) || timer_pending(&pg->timer) ||
 291             hlist_unhashed(&pg->mglist) || pg->state & MDB_PERMANENT)
 292                 goto out;
 293
 294         br_multicast_del_pg(br, pg);
 295
 296 out:
 297         spin_unlock(&br->multicast_lock);
 298 }
 299
 300 static int br_mdb_rehash(struct net_bridge_mdb_htable __rcu **mdbp, int max,
 301                          int elasticity)
 302 {
 303         struct net_bridge_mdb_htable *old = rcu_dereference_protected(*mdbp, 1);
 304         struct net_bridge_mdb_htable *mdb;
 305         int err;
 306
 307         mdb = kmalloc(sizeof(*mdb), GFP_ATOMIC);
 308         if (!mdb)
 309                 return -ENOMEM;
 310
 311         mdb->max = max;
 312         mdb->old = old;
 313
 314         mdb->mhash = kzalloc(max * sizeof(*mdb->mhash), GFP_ATOMIC);
 315         if (!mdb->mhash) {
 316                 kfree(mdb);
 317                 return -ENOMEM;
 318         }
 319
 320         mdb->size = old ? old->size : 0;
 321         mdb->ver = old ? old->ver ^ 1 : 0;
 322
 323         if (!old || elasticity)
 324                 get_random_bytes(&mdb->secret, sizeof(mdb->secret));
 325         else
 326                 mdb->secret = old->secret;
 327
 328         if (!old)
 329                 goto out;
 330
 331         err = br_mdb_copy(mdb, old, elasticity);
 332         if (err) {
 333                 kfree(mdb->mhash);
 334                 kfree(mdb);
 335                 return err;
 336         }
 337
 338         br_mdb_rehash_seq++;
 339         call_rcu_bh(&mdb->rcu, br_mdb_free);
 340
 341 out:
 342         rcu_assign_pointer(*mdbp, mdb);
 343
 344         return 0;
 345 }
 346
 347 static struct sk_buff *br_ip4_multicast_alloc_query(struct net_bridge *br,
 348                                                     __be32 group)
 349 {
 350         struct sk_buff *skb;
 351         struct igmphdr *ih;
 352         struct ethhdr *eth;
 353         struct iphdr *iph;
 354
 355         skb = netdev_alloc_skb_ip_align(br->dev, sizeof(*eth) + sizeof(*iph) +
 356                                                  sizeof(*ih) + 4);
 357         if (!skb)
 358                 goto out;
 359
 360         skb->protocol = htons(ETH_P_IP);
 361
 362         skb_reset_mac_header(skb);
 363         eth = eth_hdr(skb);
 364
 365         memcpy(eth->h_source, br->dev->dev_addr, 6);
 366         eth->h_dest[0] = 1;
 367         eth->h_dest[1] = 0;
 368         eth->h_dest[2] = 0x5e;
 369         eth->h_dest[3] = 0;
 370         eth->h_dest[4] = 0;
 371         eth->h_dest[5] = 1;
 372         eth->h_proto = htons(ETH_P_IP);
 373         skb_put(skb, sizeof(*eth));
 374
 375         skb_set_network_header(skb, skb->len);
 376         iph = ip_hdr(skb);
 377
 378         iph->version = 4;
 379         iph->ihl = 6;
 380         iph->tos = 0xc0;
 381         iph->tot_len = htons(sizeof(*iph) + sizeof(*ih) + 4);
 382         iph->id = 0;
 383         iph->frag_off = htons(IP_DF);
 384         iph->ttl = 1;
 385         iph->protocol = IPPROTO_IGMP;
 386         iph->saddr = br->multicast_query_use_ifaddr ?
 387                      inet_select_addr(br->dev, 0, RT_SCOPE_LINK) : 0;
 388         iph->daddr = htonl(INADDR_ALLHOSTS_GROUP);
 389         ((u8 *)&iph[1])[0] = IPOPT_RA;
 390         ((u8 *)&iph[1])[1] = 4;
 391         ((u8 *)&iph[1])[2] = 0;
 392         ((u8 *)&iph[1])[3] = 0;
 393         ip_send_check(iph);
 394         skb_put(skb, 24);
 395
 396         skb_set_transport_header(skb, skb->len);
 397         ih = igmp_hdr(skb);
 398         ih->type = IGMP_HOST_MEMBERSHIP_QUERY;
 399         ih->code = (group ? br->multicast_last_member_interval :
 400                             br->multicast_query_response_interval) /
 401                    (HZ / IGMP_TIMER_SCALE);
 402         ih->group = group;
 403         ih->csum = 0;
 404         ih->csum = ip_compute_csum((void *)ih, sizeof(struct igmphdr));
 405         skb_put(skb, sizeof(*ih));
 406
 407         __skb_pull(skb, sizeof(*eth));
 408
 409 out:
 410         return skb;
 411 }
 412
 413 #if IS_ENABLED(CONFIG_IPV6)
 414 static struct sk_buff *br_ip6_multicast_alloc_query(struct net_bridge *br,
 415                                                     const struct in6_addr *group)
 416 {
 417         struct sk_buff *skb;
 418         struct ipv6hdr *ip6h;
 419         struct mld_msg *mldq;
 420         struct ethhdr *eth;
 421         u8 *hopopt;
 422         unsigned long interval;
 423
 424         skb = netdev_alloc_skb_ip_align(br->dev, sizeof(*eth) + sizeof(*ip6h) +
 425                                                  8 + sizeof(*mldq));
 426         if (!skb)
 427                 goto out;
 428
 429         skb->protocol = htons(ETH_P_IPV6);
 430
 431         /* Ethernet header */
 432         skb_reset_mac_header(skb);
 433         eth = eth_hdr(skb);
 434
 435         memcpy(eth->h_source, br->dev->dev_addr, 6);
 436         eth->h_proto = htons(ETH_P_IPV6);
 437         skb_put(skb, sizeof(*eth));
 438
 439         /* IPv6 header + HbH option */
 440         skb_set_network_header(skb, skb->len);
 441         ip6h = ipv6_hdr(skb);
 442
 443         *(__force __be32 *)ip6h = htonl(0x60000000);
 444         ip6h->payload_len = htons(8 + sizeof(*mldq));
 445         ip6h->nexthdr = IPPROTO_HOPOPTS;
 446         ip6h->hop_limit = 1;
 447         ipv6_addr_set(&ip6h->daddr, htonl(0xff020000), 0, 0, htonl(1));
 448         if (ipv6_dev_get_saddr(dev_net(br->dev), br->dev, &ip6h->daddr, 0,
 449                                &ip6h->saddr)) {
 450                 kfree_skb(skb);
 451                 return NULL;
 452         }
 453         ipv6_eth_mc_map(&ip6h->daddr, eth->h_dest);
 454
 455         hopopt = (u8 *)(ip6h + 1);
 456         hopopt[0] = IPPROTO_ICMPV6;             /* next hdr */
 457         hopopt[1] = 0;                          /* length of HbH */
 458         hopopt[2] = IPV6_TLV_ROUTERALERT;       /* Router Alert */
 459         hopopt[3] = 2;                          /* Length of RA Option */
 460         hopopt[4] = 0;                          /* Type = 0x0000 (MLD) */
 461         hopopt[5] = 0;
 462         hopopt[6] = IPV6_TLV_PAD1;              /* Pad1 */
 463         hopopt[7] = IPV6_TLV_PAD1;              /* Pad1 */
 464
 465         skb_put(skb, sizeof(*ip6h) + 8);
 466
 467         /* ICMPv6 */
 468         skb_set_transport_header(skb, skb->len);
 469         mldq = (struct mld_msg *) icmp6_hdr(skb);
 470
 471         interval = ipv6_addr_any(group) ?
 472                         br->multicast_query_response_interval :
 473                         br->multicast_last_member_interval;
 474
 475         mldq->mld_type = ICMPV6_MGM_QUERY;
 476         mldq->mld_code = 0;
 477         mldq->mld_cksum = 0;
 478         mldq->mld_maxdelay = htons((u16)jiffies_to_msecs(interval));
 479         mldq->mld_reserved = 0;
 480         mldq->mld_mca = *group;
 481
 482         /* checksum */
 483         mldq->mld_cksum = csum_ipv6_magic(&ip6h->saddr, &ip6h->daddr,
 484                                           sizeof(*mldq), IPPROTO_ICMPV6,
 485                                           csum_partial(mldq,
 486                                                        sizeof(*mldq), 0));
 487         skb_put(skb, sizeof(*mldq));
 488
 489         __skb_pull(skb, sizeof(*eth));
 490
 491 out:
 492         return skb;
 493 }
 494 #endif
 495
 496 static struct sk_buff *br_multicast_alloc_query(struct net_bridge *br,
 497                                                 struct br_ip *addr)
 498 {
 499         switch (addr->proto) {
 500         case htons(ETH_P_IP):
 501                 return br_ip4_multicast_alloc_query(br, addr->u.ip4);
 502 #if IS_ENABLED(CONFIG_IPV6)
 503         case htons(ETH_P_IPV6):
 504                 return br_ip6_multicast_alloc_query(br, &addr->u.ip6);
 505 #endif
 506         }
 507         return NULL;
 508 }
 509
 510 static struct net_bridge_mdb_entry *br_multicast_get_group(
 511         struct net_bridge *br, struct net_bridge_port *port,
 512         struct br_ip *group, int hash)
 513 {
 514         struct net_bridge_mdb_htable *mdb;
 515         struct net_bridge_mdb_entry *mp;
 516         unsigned int count = 0;
 517         unsigned int max;
 518         int elasticity;
 519         int err;
 520
 521         mdb = rcu_dereference_protected(br->mdb, 1);
 522         hlist_for_each_entry(mp, &mdb->mhash[hash], hlist[mdb->ver]) {
 523                 count++;
 524                 if (unlikely(br_ip_equal(group, &mp->addr)))
 525                         return mp;
 526         }
 527
 528         elasticity = 0;
 529         max = mdb->max;
 530
 531         if (unlikely(count > br->hash_elasticity && count)) {
 532                 if (net_ratelimit())
 533                         br_info(br, "Multicast hash table "
 534                                 "chain limit reached: %s\n",
 535                                 port ? port->dev->name : br->dev->name);
 536
 537                 elasticity = br->hash_elasticity;
 538         }
 539
 540         if (mdb->size >= max) {
 541                 max *= 2;
 542                 if (unlikely(max > br->hash_max)) {
 543                         br_warn(br, "Multicast hash table maximum of %d "
 544                                 "reached, disabling snooping: %s\n",
 545                                 br->hash_max,
 546                                 port ? port->dev->name : br->dev->name);
 547                         err = -E2BIG;
 548 disable:
 549                         br->multicast_disabled = 1;
 550                         goto err;
 551                 }
 552         }
 553
 554         if (max > mdb->max || elasticity) {
 555                 if (mdb->old) {
 556                         if (net_ratelimit())
 557                                 br_info(br, "Multicast hash table "
 558                                         "on fire: %s\n",
 559                                         port ? port->dev->name : br->dev->name);
 560                         err = -EEXIST;
 561                         goto err;
 562                 }
 563
 564                 err = br_mdb_rehash(&br->mdb, max, elasticity);
 565                 if (err) {
 566                         br_warn(br, "Cannot rehash multicast "
 567                                 "hash table, disabling snooping: %s, %d, %d\n",
 568                                 port ? port->dev->name : br->dev->name,
 569                                 mdb->size, err);
 570                         goto disable;
 571                 }
 572
 573                 err = -EAGAIN;
 574                 goto err;
 575         }
 576
 577         return NULL;
 578
 579 err:
 580         mp = ERR_PTR(err);
 581         return mp;
 582 }
 583
 584 struct net_bridge_mdb_entry *br_multicast_new_group(struct net_bridge *br,
 585         struct net_bridge_port *port, struct br_ip *group)
 586 {
 587         struct net_bridge_mdb_htable *mdb;
 588         struct net_bridge_mdb_entry *mp;
 589         int hash;
 590         int err;
 591
 592         mdb = rcu_dereference_protected(br->mdb, 1);
 593         if (!mdb) {
 594                 err = br_mdb_rehash(&br->mdb, BR_HASH_SIZE, 0);
 595                 if (err)
 596                         return ERR_PTR(err);
 597                 goto rehash;
 598         }
 599
 600         hash = br_ip_hash(mdb, group);
 601         mp = br_multicast_get_group(br, port, group, hash);
 602         switch (PTR_ERR(mp)) {
 603         case 0:
 604                 break;
 605
 606         case -EAGAIN:
 607 rehash:
 608                 mdb = rcu_dereference_protected(br->mdb, 1);
 609                 hash = br_ip_hash(mdb, group);
 610                 break;
 611
 612         default:
 613                 goto out;
 614         }
 615
 616         mp = kzalloc(sizeof(*mp), GFP_ATOMIC);
 617         if (unlikely(!mp))
 618                 return ERR_PTR(-ENOMEM);
 619
 620         mp->br = br;
 621         mp->addr = *group;
 622
 623         setup_timer(&mp->timer, br_multicast_group_expired,
 624                     (unsigned long)mp);
 625
 626         hlist_add_head_rcu(&mp->hlist[mdb->ver], &mdb->mhash[hash]);
 627         mdb->size++;
 628
 629 out:
 630         return mp;
 631 }
 632
 633 struct net_bridge_port_group *br_multicast_new_port_group(
 634                         struct net_bridge_port *port,
 635                         struct br_ip *group,
 636                         struct net_bridge_port_group __rcu *next,
 637                         unsigned char state)
 638 {
 639         struct net_bridge_port_group *p;
 640
 641         p = kzalloc(sizeof(*p), GFP_ATOMIC);
 642         if (unlikely(!p))
 643                 return NULL;
 644
 645         p->addr = *group;
 646         p->port = port;
 647         p->state = state;
 648         rcu_assign_pointer(p->next, next);
 649         hlist_add_head(&p->mglist, &port->mglist);
 650         setup_timer(&p->timer, br_multicast_port_group_expired,
 651                     (unsigned long)p);
 652         return p;
 653 }
 654
 655 static int br_multicast_add_group(struct net_bridge *br,
 656                                   struct net_bridge_port *port,
 657                                   struct br_ip *group)
 658 {
 659         struct net_bridge_mdb_entry *mp;
 660         struct net_bridge_port_group *p;
 661         struct net_bridge_port_group __rcu **pp;
 662         int err;
 663
 664         spin_lock(&br->multicast_lock);
 665         if (!netif_running(br->dev) ||
 666             (port && port->state == BR_STATE_DISABLED))
 667                 goto out;
 668
 669         mp = br_multicast_new_group(br, port, group);
 670         err = PTR_ERR(mp);
 671         if (IS_ERR(mp))
 672                 goto err;
 673
 674         if (!port) {
 675                 mp->mglist = true;
 676                 goto out;
 677         }
 678
 679         for (pp = &mp->ports;
 680              (p = mlock_dereference(*pp, br)) != NULL;
 681              pp = &p->next) {
 682                 if (p->port == port)
 683                         goto out;
 684                 if ((unsigned long)p->port < (unsigned long)port)
 685                         break;
 686         }
 687
 688         p = br_multicast_new_port_group(port, group, *pp, MDB_TEMPORARY);
 689         if (unlikely(!p))
 690                 goto err;
 691         rcu_assign_pointer(*pp, p);
 692         br_mdb_notify(br->dev, port, group, RTM_NEWMDB);
 693
 694 out:
 695         err = 0;
 696
 697 err:
 698         spin_unlock(&br->multicast_lock);
 699         return err;
 700 }
 701
 702 static int br_ip4_multicast_add_group(struct net_bridge *br,
 703                                       struct net_bridge_port *port,
 704                                       __be32 group,
 705                                       __u16 vid)
 706 {
 707         struct br_ip br_group;
 708
 709         if (ipv4_is_local_multicast(group))
 710                 return 0;
 711
 712         br_group.u.ip4 = group;
 713         br_group.proto = htons(ETH_P_IP);
 714         br_group.vid = vid;
 715
 716         return br_multicast_add_group(br, port, &br_group);
 717 }
 718
 719 #if IS_ENABLED(CONFIG_IPV6)
 720 static int br_ip6_multicast_add_group(struct net_bridge *br,
 721                                       struct net_bridge_port *port,
 722                                       const struct in6_addr *group,
 723                                       __u16 vid)
 724 {
 725         struct br_ip br_group;
 726
 727         if (!ipv6_is_transient_multicast(group))
 728                 return 0;
 729
 730         br_group.u.ip6 = *group;
 731         br_group.proto = htons(ETH_P_IPV6);
 732         br_group.vid = vid;
 733
 734         return br_multicast_add_group(br, port, &br_group);
 735 }
 736 #endif
 737
 738 static void br_multicast_router_expired(unsigned long data)
 739 {
 740         struct net_bridge_port *port = (void *)data;
 741         struct net_bridge *br = port->br;
 742
 743         spin_lock(&br->multicast_lock);
 744         if (port->multicast_router != 1 ||
 745             timer_pending(&port->multicast_router_timer) ||
 746             hlist_unhashed(&port->rlist))
 747                 goto out;
 748
 749         hlist_del_init_rcu(&port->rlist);
 750
 751 out:
 752         spin_unlock(&br->multicast_lock);
 753 }
 754
 755 static void br_multicast_local_router_expired(unsigned long data)
 756 {
 757 }
 758
 759 static void br_multicast_querier_expired(struct net_bridge *br,
 760                                          struct bridge_mcast_query *query)
 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, query);
 767
 768 out:
 769         spin_unlock(&br->multicast_lock);
 770 }
 771
 772 static void br_ip4_multicast_querier_expired(unsigned long data)
 773 {
 774         struct net_bridge *br = (void *)data;
 775
 776         br_multicast_querier_expired(br, &br->ip4_query);
 777 }
 778
 779 #if IS_ENABLED(CONFIG_IPV6)
 780 static void br_ip6_multicast_querier_expired(unsigned long data)
 781 {
 782         struct net_bridge *br = (void *)data;
 783
 784         br_multicast_querier_expired(br, &br->ip6_query);
 785 }
 786 #endif
 787
 788 static void __br_multicast_send_query(struct net_bridge *br,
 789                                       struct net_bridge_port *port,
 790                                       struct br_ip *ip)
 791 {
 792         struct sk_buff *skb;
 793
 794         skb = br_multicast_alloc_query(br, ip);
 795         if (!skb)
 796                 return;
 797
 798         if (port) {
 799                 __skb_push(skb, sizeof(struct ethhdr));
 800                 skb->dev = port->dev;
 801                 NF_HOOK(NFPROTO_BRIDGE, NF_BR_LOCAL_OUT, skb, NULL, skb->dev,
 802                         dev_queue_xmit);
 803         } else
 804                 netif_rx(skb);
 805 }
 806
 807 static void br_multicast_send_query(struct net_bridge *br,
 808                                     struct net_bridge_port *port,
 809                                     struct bridge_mcast_query *query)
 810 {
 811         unsigned long time;
 812         struct br_ip br_group;
 813         struct bridge_mcast_querier *querier = NULL;
 814
 815         if (!netif_running(br->dev) || br->multicast_disabled ||
 816             !br->multicast_querier)
 817                 return;
 818
 819         memset(&br_group.u, 0, sizeof(br_group.u));
 820
 821         if (port ? (query == &port->ip4_query) :
 822                    (query == &br->ip4_query)) {
 823                 querier = &br->ip4_querier;
 824                 br_group.proto = htons(ETH_P_IP);
 825 #if IS_ENABLED(CONFIG_IPV6)
 826         } else {
 827                 querier = &br->ip6_querier;
 828                 br_group.proto = htons(ETH_P_IPV6);
 829 #endif
 830         }
 831
 832         if (!querier || timer_pending(&querier->timer))
 833                 return;
 834
 835         __br_multicast_send_query(br, port, &br_group);
 836
 837         time = jiffies;
 838         time += query->startup_sent < br->multicast_startup_query_count ?
 839                 br->multicast_startup_query_interval :
 840                 br->multicast_query_interval;
 841         mod_timer(&query->timer, time);
 842 }
 843
 844 static void br_multicast_port_query_expired(struct net_bridge_port *port,
 845                                             struct bridge_mcast_query *query)
 846 {
 847         struct net_bridge *br = port->br;
 848
 849         spin_lock(&br->multicast_lock);
 850         if (port->state == BR_STATE_DISABLED ||
 851             port->state == BR_STATE_BLOCKING)
 852                 goto out;
 853
 854         if (query->startup_sent < br->multicast_startup_query_count)
 855                 query->startup_sent++;
 856
 857         br_multicast_send_query(port->br, port, query);
 858
 859 out:
 860         spin_unlock(&br->multicast_lock);
 861 }
 862
 863 static void br_ip4_multicast_port_query_expired(unsigned long data)
 864 {
 865         struct net_bridge_port *port = (void *)data;
 866
 867         br_multicast_port_query_expired(port, &port->ip4_query);
 868 }
 869
 870 #if IS_ENABLED(CONFIG_IPV6)
 871 static void br_ip6_multicast_port_query_expired(unsigned long data)
 872 {
 873         struct net_bridge_port *port = (void *)data;
 874
 875         br_multicast_port_query_expired(port, &port->ip6_query);
 876 }
 877 #endif
 878
 879 void br_multicast_add_port(struct net_bridge_port *port)
 880 {
 881         port->multicast_router = 1;
 882
 883         setup_timer(&port->multicast_router_timer, br_multicast_router_expired,
 884                     (unsigned long)port);
 885         setup_timer(&port->ip4_query.timer, br_ip4_multicast_port_query_expired,
 886                     (unsigned long)port);
 887 #if IS_ENABLED(CONFIG_IPV6)
 888         setup_timer(&port->ip6_query.timer, br_ip6_multicast_port_query_expired,
 889                     (unsigned long)port);
 890 #endif
 891 }
 892
 893 void br_multicast_del_port(struct net_bridge_port *port)
 894 {
 895         del_timer_sync(&port->multicast_router_timer);
 896 }
 897
 898 static void br_multicast_enable(struct bridge_mcast_query *query)
 899 {
 900         query->startup_sent = 0;
 901
 902         if (try_to_del_timer_sync(&query->timer) >= 0 ||
 903             del_timer(&query->timer))
 904                 mod_timer(&query->timer, jiffies);
 905 }
 906
 907 void br_multicast_enable_port(struct net_bridge_port *port)
 908 {
 909         struct net_bridge *br = port->br;
 910
 911         spin_lock(&br->multicast_lock);
 912         if (br->multicast_disabled || !netif_running(br->dev))
 913                 goto out;
 914
 915         br_multicast_enable(&port->ip4_query);
 916 #if IS_ENABLED(CONFIG_IPV6)
 917         br_multicast_enable(&port->ip6_query);
 918 #endif
 919
 920 out:
 921         spin_unlock(&br->multicast_lock);
 922 }
 923
 924 void br_multicast_disable_port(struct net_bridge_port *port)
 925 {
 926         struct net_bridge *br = port->br;
 927         struct net_bridge_port_group *pg;
 928         struct hlist_node *n;
 929
 930         spin_lock(&br->multicast_lock);
 931         hlist_for_each_entry_safe(pg, n, &port->mglist, mglist)
 932                 br_multicast_del_pg(br, pg);
 933
 934         if (!hlist_unhashed(&port->rlist))
 935                 hlist_del_init_rcu(&port->rlist);
 936         del_timer(&port->multicast_router_timer);
 937         del_timer(&port->ip4_query.timer);
 938 #if IS_ENABLED(CONFIG_IPV6)
 939         del_timer(&port->ip6_query.timer);
 940 #endif
 941         spin_unlock(&br->multicast_lock);
 942 }
 943
 944 static int br_ip4_multicast_igmp3_report(struct net_bridge *br,
 945                                          struct net_bridge_port *port,
 946                                          struct sk_buff *skb)
 947 {
 948         struct igmpv3_report *ih;
 949         struct igmpv3_grec *grec;
 950         int i;
 951         int len;
 952         int num;
 953         int type;
 954         int err = 0;
 955         __be32 group;
 956         u16 vid = 0;
 957
 958         if (!pskb_may_pull(skb, sizeof(*ih)))
 959                 return -EINVAL;
 960
 961         br_vlan_get_tag(skb, &vid);
 962         ih = igmpv3_report_hdr(skb);
 963         num = ntohs(ih->ngrec);
 964         len = sizeof(*ih);
 965
 966         for (i = 0; i < num; i++) {
 967                 len += sizeof(*grec);
 968                 if (!pskb_may_pull(skb, len))
 969                         return -EINVAL;
 970
 971                 grec = (void *)(skb->data + len - sizeof(*grec));
 972                 group = grec->grec_mca;
 973                 type = grec->grec_type;
 974
 975                 len += ntohs(grec->grec_nsrcs) * 4;
 976                 if (!pskb_may_pull(skb, len))
 977                         return -EINVAL;
 978
 979                 /* We treat this as an IGMPv2 report for now. */
 980                 switch (type) {
 981                 case IGMPV3_MODE_IS_INCLUDE:
 982                 case IGMPV3_MODE_IS_EXCLUDE:
 983                 case IGMPV3_CHANGE_TO_INCLUDE:
 984                 case IGMPV3_CHANGE_TO_EXCLUDE:
 985                 case IGMPV3_ALLOW_NEW_SOURCES:
 986                 case IGMPV3_BLOCK_OLD_SOURCES:
 987                         break;
 988
 989                 default:
 990                         continue;
 991                 }
 992
 993                 err = br_ip4_multicast_add_group(br, port, group, vid);
 994                 if (err)
 995                         break;
 996         }
 997
 998         return err;
 999 }
1000
1001 #if IS_ENABLED(CONFIG_IPV6)
1002 static int br_ip6_multicast_mld2_report(struct net_bridge *br,
1003                                         struct net_bridge_port *port,
1004                                         struct sk_buff *skb)
1005 {
1006         struct icmp6hdr *icmp6h;
1007         struct mld2_grec *grec;
1008         int i;
1009         int len;
1010         int num;
1011         int err = 0;
1012         u16 vid = 0;
1013
1014         if (!pskb_may_pull(skb, sizeof(*icmp6h)))
1015                 return -EINVAL;
1016
1017         br_vlan_get_tag(skb, &vid);
1018         icmp6h = icmp6_hdr(skb);
1019         num = ntohs(icmp6h->icmp6_dataun.un_data16[1]);
1020         len = sizeof(*icmp6h);
1021
1022         for (i = 0; i < num; i++) {
1023                 __be16 *nsrcs, _nsrcs;
1024
1025                 nsrcs = skb_header_pointer(skb,
1026                                            len + offsetof(struct mld2_grec,
1027                                                           grec_nsrcs),
1028                                            sizeof(_nsrcs), &_nsrcs);
1029                 if (!nsrcs)
1030                         return -EINVAL;
1031
1032                 if (!pskb_may_pull(skb,
1033                                    len + sizeof(*grec) +
1034                                    sizeof(struct in6_addr) * ntohs(*nsrcs)))
1035                         return -EINVAL;
1036
1037                 grec = (struct mld2_grec *)(skb->data + len);
1038                 len += sizeof(*grec) +
1039                        sizeof(struct in6_addr) * ntohs(*nsrcs);
1040
1041                 /* We treat these as MLDv1 reports for now. */
1042                 switch (grec->grec_type) {
1043                 case MLD2_MODE_IS_INCLUDE:
1044                 case MLD2_MODE_IS_EXCLUDE:
1045                 case MLD2_CHANGE_TO_INCLUDE:
1046                 case MLD2_CHANGE_TO_EXCLUDE:
1047                 case MLD2_ALLOW_NEW_SOURCES:
1048                 case MLD2_BLOCK_OLD_SOURCES:
1049                         break;
1050
1051                 default:
1052                         continue;
1053                 }
1054
1055                 err = br_ip6_multicast_add_group(br, port, &grec->grec_mca,
1056                                                  vid);
1057                 if (!err)
1058                         break;
1059         }
1060
1061         return err;
1062 }
1063 #endif
1064
1065 static void
1066 br_multicast_update_querier_timer(struct net_bridge *br,
1067                                   struct bridge_mcast_querier *querier,
1068                                   unsigned long max_delay)
1069 {
1070         if (!timer_pending(&querier->timer))
1071                 querier->delay_time = jiffies + max_delay;
1072
1073         mod_timer(&querier->timer, jiffies + br->multicast_querier_interval);
1074 }
1075
1076 /*
1077  * Add port to router_list
1078  *  list is maintained ordered by pointer value
1079  *  and locked by br->multicast_lock and RCU
1080  */
1081 static void br_multicast_add_router(struct net_bridge *br,
1082                                     struct net_bridge_port *port)
1083 {
1084         struct net_bridge_port *p;
1085         struct hlist_node *slot = NULL;
1086
1087         hlist_for_each_entry(p, &br->router_list, rlist) {
1088                 if ((unsigned long) port >= (unsigned long) p)
1089                         break;
1090                 slot = &p->rlist;
1091         }
1092
1093         if (slot)
1094                 hlist_add_after_rcu(slot, &port->rlist);
1095         else
1096                 hlist_add_head_rcu(&port->rlist, &br->router_list);
1097 }
1098
1099 static void br_multicast_mark_router(struct net_bridge *br,
1100                                      struct net_bridge_port *port)
1101 {
1102         unsigned long now = jiffies;
1103
1104         if (!port) {
1105                 if (br->multicast_router == 1)
1106                         mod_timer(&br->multicast_router_timer,
1107                                   now + br->multicast_querier_interval);
1108                 return;
1109         }
1110
1111         if (port->multicast_router != 1)
1112                 return;
1113
1114         if (!hlist_unhashed(&port->rlist))
1115                 goto timer;
1116
1117         br_multicast_add_router(br, port);
1118
1119 timer:
1120         mod_timer(&port->multicast_router_timer,
1121                   now + br->multicast_querier_interval);
1122 }
1123
1124 static void br_multicast_query_received(struct net_bridge *br,
1125                                         struct net_bridge_port *port,
1126                                         struct bridge_mcast_querier *querier,
1127                                         int saddr,
1128                                         unsigned long max_delay)
1129 {
1130         if (saddr)
1131                 br_multicast_update_querier_timer(br, querier, max_delay);
1132         else if (timer_pending(&querier->timer))
1133                 return;
1134
1135         br_multicast_mark_router(br, port);
1136 }
1137
1138 static int br_ip4_multicast_query(struct net_bridge *br,
1139                                   struct net_bridge_port *port,
1140                                   struct sk_buff *skb)
1141 {
1142         const struct iphdr *iph = ip_hdr(skb);
1143         struct igmphdr *ih = igmp_hdr(skb);
1144         struct net_bridge_mdb_entry *mp;
1145         struct igmpv3_query *ih3;
1146         struct net_bridge_port_group *p;
1147         struct net_bridge_port_group __rcu **pp;
1148         unsigned long max_delay;
1149         unsigned long now = jiffies;
1150         __be32 group;
1151         int err = 0;
1152         u16 vid = 0;
1153
1154         spin_lock(&br->multicast_lock);
1155         if (!netif_running(br->dev) ||
1156             (port && port->state == BR_STATE_DISABLED))
1157                 goto out;
1158
1159         group = ih->group;
1160
1161         if (skb->len == sizeof(*ih)) {
1162                 max_delay = ih->code * (HZ / IGMP_TIMER_SCALE);
1163
1164                 if (!max_delay) {
1165                         max_delay = 10 * HZ;
1166                         group = 0;
1167                 }
1168         } else {
1169                 if (!pskb_may_pull(skb, sizeof(struct igmpv3_query))) {
1170                         err = -EINVAL;
1171                         goto out;
1172                 }
1173
1174                 ih3 = igmpv3_query_hdr(skb);
1175                 if (ih3->nsrcs)
1176                         goto out;
1177
1178                 max_delay = ih3->code ?
1179                             IGMPV3_MRC(ih3->code) * (HZ / IGMP_TIMER_SCALE) : 1;
1180         }
1181
1182         br_multicast_query_received(br, port, &br->ip4_querier, !!iph->saddr,
1183                                     max_delay);
1184
1185         if (!group)
1186                 goto out;
1187
1188         br_vlan_get_tag(skb, &vid);
1189         mp = br_mdb_ip4_get(mlock_dereference(br->mdb, br), group, vid);
1190         if (!mp)
1191                 goto out;
1192
1193         mod_timer(&mp->timer, now + br->multicast_membership_interval);
1194         mp->timer_armed = true;
1195
1196         max_delay *= br->multicast_last_member_count;
1197
1198         if (mp->mglist &&
1199             (timer_pending(&mp->timer) ?
1200              time_after(mp->timer.expires, now + max_delay) :
1201              try_to_del_timer_sync(&mp->timer) >= 0))
1202                 mod_timer(&mp->timer, now + max_delay);
1203
1204         for (pp = &mp->ports;
1205              (p = mlock_dereference(*pp, br)) != NULL;
1206              pp = &p->next) {
1207                 if (timer_pending(&p->timer) ?
1208                     time_after(p->timer.expires, now + max_delay) :
1209                     try_to_del_timer_sync(&p->timer) >= 0)
1210                         mod_timer(&p->timer, now + max_delay);
1211         }
1212
1213 out:
1214         spin_unlock(&br->multicast_lock);
1215         return err;
1216 }
1217
1218 #if IS_ENABLED(CONFIG_IPV6)
1219 static int br_ip6_multicast_query(struct net_bridge *br,
1220                                   struct net_bridge_port *port,
1221                                   struct sk_buff *skb)
1222 {
1223         const struct ipv6hdr *ip6h = ipv6_hdr(skb);
1224         struct mld_msg *mld;
1225         struct net_bridge_mdb_entry *mp;
1226         struct mld2_query *mld2q;
1227         struct net_bridge_port_group *p;
1228         struct net_bridge_port_group __rcu **pp;
1229         unsigned long max_delay;
1230         unsigned long now = jiffies;
1231         const struct in6_addr *group = NULL;
1232         int err = 0;
1233         u16 vid = 0;
1234
1235         spin_lock(&br->multicast_lock);
1236         if (!netif_running(br->dev) ||
1237             (port && port->state == BR_STATE_DISABLED))
1238                 goto out;
1239
1240         if (skb->len == sizeof(*mld)) {
1241                 if (!pskb_may_pull(skb, sizeof(*mld))) {
1242                         err = -EINVAL;
1243                         goto out;
1244                 }
1245                 mld = (struct mld_msg *) icmp6_hdr(skb);
1246                 max_delay = msecs_to_jiffies(ntohs(mld->mld_maxdelay));
1247                 if (max_delay)
1248                         group = &mld->mld_mca;
1249         } else {
1250                 if (!pskb_may_pull(skb, sizeof(*mld2q))) {
1251                         err = -EINVAL;
1252                         goto out;
1253                 }
1254                 mld2q = (struct mld2_query *)icmp6_hdr(skb);
1255                 if (!mld2q->mld2q_nsrcs)
1256                         group = &mld2q->mld2q_mca;
1257
1258                 max_delay = max(msecs_to_jiffies(MLDV2_MRC(ntohs(mld2q->mld2q_mrc))), 1UL);
1259         }
1260
1261         br_multicast_query_received(br, port, &br->ip6_querier,
1262                                     !ipv6_addr_any(&ip6h->saddr), max_delay);
1263
1264         if (!group)
1265                 goto out;
1266
1267         br_vlan_get_tag(skb, &vid);
1268         mp = br_mdb_ip6_get(mlock_dereference(br->mdb, br), group, vid);
1269         if (!mp)
1270                 goto out;
1271
1272         mod_timer(&mp->timer, now + br->multicast_membership_interval);
1273         mp->timer_armed = true;
1274
1275         max_delay *= br->multicast_last_member_count;
1276         if (mp->mglist &&
1277             (timer_pending(&mp->timer) ?
1278              time_after(mp->timer.expires, now + max_delay) :
1279              try_to_del_timer_sync(&mp->timer) >= 0))
1280                 mod_timer(&mp->timer, now + max_delay);
1281
1282         for (pp = &mp->ports;
1283              (p = mlock_dereference(*pp, br)) != NULL;
1284              pp = &p->next) {
1285                 if (timer_pending(&p->timer) ?
1286                     time_after(p->timer.expires, now + max_delay) :
1287                     try_to_del_timer_sync(&p->timer) >= 0)
1288                         mod_timer(&p->timer, now + max_delay);
1289         }
1290
1291 out:
1292         spin_unlock(&br->multicast_lock);
1293         return err;
1294 }
1295 #endif
1296
1297 static void br_multicast_leave_group(struct net_bridge *br,
1298                                      struct net_bridge_port *port,
1299                                      struct br_ip *group,
1300                                      struct bridge_mcast_querier *querier,
1301                                      struct bridge_mcast_query *query)
1302 {
1303         struct net_bridge_mdb_htable *mdb;
1304         struct net_bridge_mdb_entry *mp;
1305         struct net_bridge_port_group *p;
1306         unsigned long now;
1307         unsigned long time;
1308
1309         spin_lock(&br->multicast_lock);
1310         if (!netif_running(br->dev) ||
1311             (port && port->state == BR_STATE_DISABLED) ||
1312             timer_pending(&querier->timer))
1313                 goto out;
1314
1315         mdb = mlock_dereference(br->mdb, br);
1316         mp = br_mdb_ip_get(mdb, group);
1317         if (!mp)
1318                 goto out;
1319
1320         if (br->multicast_querier) {
1321                 __br_multicast_send_query(br, port, &mp->addr);
1322
1323                 time = jiffies + br->multicast_last_member_count *
1324                                  br->multicast_last_member_interval;
1325
1326                 mod_timer(&query->timer, time);
1327
1328                 for (p = mlock_dereference(mp->ports, br);
1329                      p != NULL;
1330                      p = mlock_dereference(p->next, br)) {
1331                         if (p->port != port)
1332                                 continue;
1333
1334                         if (!hlist_unhashed(&p->mglist) &&
1335                             (timer_pending(&p->timer) ?
1336                              time_after(p->timer.expires, time) :
1337                              try_to_del_timer_sync(&p->timer) >= 0)) {
1338                                 mod_timer(&p->timer, time);
1339                         }
1340
1341                         break;
1342                 }
1343         }
1344
1345         if (port && (port->flags & BR_MULTICAST_FAST_LEAVE)) {
1346                 struct net_bridge_port_group __rcu **pp;
1347
1348                 for (pp = &mp->ports;
1349                      (p = mlock_dereference(*pp, br)) != NULL;
1350                      pp = &p->next) {
1351                         if (p->port != port)
1352                                 continue;
1353
1354                         rcu_assign_pointer(*pp, p->next);
1355                         hlist_del_init(&p->mglist);
1356                         del_timer(&p->timer);
1357                         call_rcu_bh(&p->rcu, br_multicast_free_pg);
1358                         br_mdb_notify(br->dev, port, group, RTM_DELMDB);
1359
1360                         if (!mp->ports && !mp->mglist && mp->timer_armed &&
1361                             netif_running(br->dev))
1362                                 mod_timer(&mp->timer, jiffies);
1363                 }
1364                 goto out;
1365         }
1366
1367         now = jiffies;
1368         time = now + br->multicast_last_member_count *
1369                      br->multicast_last_member_interval;
1370
1371         if (!port) {
1372                 if (mp->mglist && mp->timer_armed &&
1373                     (timer_pending(&mp->timer) ?
1374                      time_after(mp->timer.expires, time) :
1375                      try_to_del_timer_sync(&mp->timer) >= 0)) {
1376                         mod_timer(&mp->timer, time);
1377                 }
1378         }
1379 out:
1380         spin_unlock(&br->multicast_lock);
1381 }
1382
1383 static void br_ip4_multicast_leave_group(struct net_bridge *br,
1384                                          struct net_bridge_port *port,
1385                                          __be32 group,
1386                                          __u16 vid)
1387 {
1388         struct br_ip br_group;
1389         struct bridge_mcast_query *query = port ? &port->ip4_query :
1390                                                   &br->ip4_query;
1391
1392         if (ipv4_is_local_multicast(group))
1393                 return;
1394
1395         br_group.u.ip4 = group;
1396         br_group.proto = htons(ETH_P_IP);
1397         br_group.vid = vid;
1398
1399         br_multicast_leave_group(br, port, &br_group, &br->ip4_querier, query);
1400 }
1401
1402 #if IS_ENABLED(CONFIG_IPV6)
1403 static void br_ip6_multicast_leave_group(struct net_bridge *br,
1404                                          struct net_bridge_port *port,
1405                                          const struct in6_addr *group,
1406                                          __u16 vid)
1407 {
1408         struct br_ip br_group;
1409         struct bridge_mcast_query *query = port ? &port->ip6_query :
1410                                                   &br->ip6_query;
1411
1412
1413         if (!ipv6_is_transient_multicast(group))
1414                 return;
1415
1416         br_group.u.ip6 = *group;
1417         br_group.proto = htons(ETH_P_IPV6);
1418         br_group.vid = vid;
1419
1420         br_multicast_leave_group(br, port, &br_group, &br->ip6_querier, query);
1421 }
1422 #endif
1423
1424 static int br_multicast_ipv4_rcv(struct net_bridge *br,
1425                                  struct net_bridge_port *port,
1426                                  struct sk_buff *skb)
1427 {
1428         struct sk_buff *skb2 = skb;
1429         const struct iphdr *iph;
1430         struct igmphdr *ih;
1431         unsigned int len;
1432         unsigned int offset;
1433         int err;
1434         u16 vid = 0;
1435
1436         /* We treat OOM as packet loss for now. */
1437         if (!pskb_may_pull(skb, sizeof(*iph)))
1438                 return -EINVAL;
1439
1440         iph = ip_hdr(skb);
1441
1442         if (iph->ihl < 5 || iph->version != 4)
1443                 return -EINVAL;
1444
1445         if (!pskb_may_pull(skb, ip_hdrlen(skb)))
1446                 return -EINVAL;
1447
1448         iph = ip_hdr(skb);
1449
1450         if (unlikely(ip_fast_csum((u8 *)iph, iph->ihl)))
1451                 return -EINVAL;
1452
1453         if (iph->protocol != IPPROTO_IGMP) {
1454                 if (!ipv4_is_local_multicast(iph->daddr))
1455                         BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1456                 return 0;
1457         }
1458
1459         len = ntohs(iph->tot_len);
1460         if (skb->len < len || len < ip_hdrlen(skb))
1461                 return -EINVAL;
1462
1463         if (skb->len > len) {
1464                 skb2 = skb_clone(skb, GFP_ATOMIC);
1465                 if (!skb2)
1466                         return -ENOMEM;
1467
1468                 err = pskb_trim_rcsum(skb2, len);
1469                 if (err)
1470                         goto err_out;
1471         }
1472
1473         len -= ip_hdrlen(skb2);
1474         offset = skb_network_offset(skb2) + ip_hdrlen(skb2);
1475         __skb_pull(skb2, offset);
1476         skb_reset_transport_header(skb2);
1477
1478         err = -EINVAL;
1479         if (!pskb_may_pull(skb2, sizeof(*ih)))
1480                 goto out;
1481
1482         switch (skb2->ip_summed) {
1483         case CHECKSUM_COMPLETE:
1484                 if (!csum_fold(skb2->csum))
1485                         break;
1486                 /* fall through */
1487         case CHECKSUM_NONE:
1488                 skb2->csum = 0;
1489                 if (skb_checksum_complete(skb2))
1490                         goto out;
1491         }
1492
1493         err = 0;
1494
1495         br_vlan_get_tag(skb2, &vid);
1496         BR_INPUT_SKB_CB(skb)->igmp = 1;
1497         ih = igmp_hdr(skb2);
1498
1499         switch (ih->type) {
1500         case IGMP_HOST_MEMBERSHIP_REPORT:
1501         case IGMPV2_HOST_MEMBERSHIP_REPORT:
1502                 BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1503                 err = br_ip4_multicast_add_group(br, port, ih->group, vid);
1504                 break;
1505         case IGMPV3_HOST_MEMBERSHIP_REPORT:
1506                 err = br_ip4_multicast_igmp3_report(br, port, skb2);
1507                 break;
1508         case IGMP_HOST_MEMBERSHIP_QUERY:
1509                 err = br_ip4_multicast_query(br, port, skb2);
1510                 break;
1511         case IGMP_HOST_LEAVE_MESSAGE:
1512                 br_ip4_multicast_leave_group(br, port, ih->group, vid);
1513                 break;
1514         }
1515
1516 out:
1517         __skb_push(skb2, offset);
1518 err_out:
1519         if (skb2 != skb)
1520                 kfree_skb(skb2);
1521         return err;
1522 }
1523
1524 #if IS_ENABLED(CONFIG_IPV6)
1525 static int br_multicast_ipv6_rcv(struct net_bridge *br,
1526                                  struct net_bridge_port *port,
1527                                  struct sk_buff *skb)
1528 {
1529         struct sk_buff *skb2;
1530         const struct ipv6hdr *ip6h;
1531         u8 icmp6_type;
1532         u8 nexthdr;
1533         __be16 frag_off;
1534         unsigned int len;
1535         int offset;
1536         int err;
1537         u16 vid = 0;
1538
1539         if (!pskb_may_pull(skb, sizeof(*ip6h)))
1540                 return -EINVAL;
1541
1542         ip6h = ipv6_hdr(skb);
1543
1544         /*
1545          * We're interested in MLD messages only.
1546          *  - Version is 6
1547          *  - MLD has always Router Alert hop-by-hop option
1548          *  - But we do not support jumbrograms.
1549          */
1550         if (ip6h->version != 6 ||
1551             ip6h->nexthdr != IPPROTO_HOPOPTS ||
1552             ip6h->payload_len == 0)
1553                 return 0;
1554
1555         len = ntohs(ip6h->payload_len) + sizeof(*ip6h);
1556         if (skb->len < len)
1557                 return -EINVAL;
1558
1559         nexthdr = ip6h->nexthdr;
1560         offset = ipv6_skip_exthdr(skb, sizeof(*ip6h), &nexthdr, &frag_off);
1561
1562         if (offset < 0 || nexthdr != IPPROTO_ICMPV6)
1563                 return 0;
1564
1565         /* Okay, we found ICMPv6 header */
1566         skb2 = skb_clone(skb, GFP_ATOMIC);
1567         if (!skb2)
1568                 return -ENOMEM;
1569
1570         err = -EINVAL;
1571         if (!pskb_may_pull(skb2, offset + sizeof(struct icmp6hdr)))
1572                 goto out;
1573
1574         len -= offset - skb_network_offset(skb2);
1575
1576         __skb_pull(skb2, offset);
1577         skb_reset_transport_header(skb2);
1578         skb_postpull_rcsum(skb2, skb_network_header(skb2),
1579                            skb_network_header_len(skb2));
1580
1581         icmp6_type = icmp6_hdr(skb2)->icmp6_type;
1582
1583         switch (icmp6_type) {
1584         case ICMPV6_MGM_QUERY:
1585         case ICMPV6_MGM_REPORT:
1586         case ICMPV6_MGM_REDUCTION:
1587         case ICMPV6_MLD2_REPORT:
1588                 break;
1589         default:
1590                 err = 0;
1591                 goto out;
1592         }
1593
1594         /* Okay, we found MLD message. Check further. */
1595         if (skb2->len > len) {
1596                 err = pskb_trim_rcsum(skb2, len);
1597                 if (err)
1598                         goto out;
1599                 err = -EINVAL;
1600         }
1601
1602         ip6h = ipv6_hdr(skb2);
1603
1604         switch (skb2->ip_summed) {
1605         case CHECKSUM_COMPLETE:
1606                 if (!csum_ipv6_magic(&ip6h->saddr, &ip6h->daddr, skb2->len,
1607                                         IPPROTO_ICMPV6, skb2->csum))
1608                         break;
1609                 /*FALLTHROUGH*/
1610         case CHECKSUM_NONE:
1611                 skb2->csum = ~csum_unfold(csum_ipv6_magic(&ip6h->saddr,
1612                                                         &ip6h->daddr,
1613                                                         skb2->len,
1614                                                         IPPROTO_ICMPV6, 0));
1615                 if (__skb_checksum_complete(skb2))
1616                         goto out;
1617         }
1618
1619         err = 0;
1620
1621         br_vlan_get_tag(skb, &vid);
1622         BR_INPUT_SKB_CB(skb)->igmp = 1;
1623
1624         switch (icmp6_type) {
1625         case ICMPV6_MGM_REPORT:
1626             {
1627                 struct mld_msg *mld;
1628                 if (!pskb_may_pull(skb2, sizeof(*mld))) {
1629                         err = -EINVAL;
1630                         goto out;
1631                 }
1632                 mld = (struct mld_msg *)skb_transport_header(skb2);
1633                 BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1634                 err = br_ip6_multicast_add_group(br, port, &mld->mld_mca, vid);
1635                 break;
1636             }
1637         case ICMPV6_MLD2_REPORT:
1638                 err = br_ip6_multicast_mld2_report(br, port, skb2);
1639                 break;
1640         case ICMPV6_MGM_QUERY:
1641                 err = br_ip6_multicast_query(br, port, skb2);
1642                 break;
1643         case ICMPV6_MGM_REDUCTION:
1644             {
1645                 struct mld_msg *mld;
1646                 if (!pskb_may_pull(skb2, sizeof(*mld))) {
1647                         err = -EINVAL;
1648                         goto out;
1649                 }
1650                 mld = (struct mld_msg *)skb_transport_header(skb2);
1651                 br_ip6_multicast_leave_group(br, port, &mld->mld_mca, vid);
1652             }
1653         }
1654
1655 out:
1656         kfree_skb(skb2);
1657         return err;
1658 }
1659 #endif
1660
1661 int br_multicast_rcv(struct net_bridge *br, struct net_bridge_port *port,
1662                      struct sk_buff *skb)
1663 {
1664         BR_INPUT_SKB_CB(skb)->igmp = 0;
1665         BR_INPUT_SKB_CB(skb)->mrouters_only = 0;
1666
1667         if (br->multicast_disabled)
1668                 return 0;
1669
1670         switch (skb->protocol) {
1671         case htons(ETH_P_IP):
1672                 return br_multicast_ipv4_rcv(br, port, skb);
1673 #if IS_ENABLED(CONFIG_IPV6)
1674         case htons(ETH_P_IPV6):
1675                 return br_multicast_ipv6_rcv(br, port, skb);
1676 #endif
1677         }
1678
1679         return 0;
1680 }
1681
1682 static void br_multicast_query_expired(struct net_bridge *br,
1683                                        struct bridge_mcast_query *query)
1684 {
1685         spin_lock(&br->multicast_lock);
1686         if (query->startup_sent < br->multicast_startup_query_count)
1687                 query->startup_sent++;
1688
1689         br_multicast_send_query(br, NULL, query);
1690         spin_unlock(&br->multicast_lock);
1691 }
1692
1693 static void br_ip4_multicast_query_expired(unsigned long data)
1694 {
1695         struct net_bridge *br = (void *)data;
1696
1697         br_multicast_query_expired(br, &br->ip4_query);
1698 }
1699
1700 #if IS_ENABLED(CONFIG_IPV6)
1701 static void br_ip6_multicast_query_expired(unsigned long data)
1702 {
1703         struct net_bridge *br = (void *)data;
1704
1705         br_multicast_query_expired(br, &br->ip6_query);
1706 }
1707 #endif
1708
1709 void br_multicast_init(struct net_bridge *br)
1710 {
1711         br->hash_elasticity = 4;
1712         br->hash_max = 512;
1713
1714         br->multicast_router = 1;
1715         br->multicast_querier = 0;
1716         br->multicast_query_use_ifaddr = 0;
1717         br->multicast_last_member_count = 2;
1718         br->multicast_startup_query_count = 2;
1719
1720         br->multicast_last_member_interval = HZ;
1721         br->multicast_query_response_interval = 10 * HZ;
1722         br->multicast_startup_query_interval = 125 * HZ / 4;
1723         br->multicast_query_interval = 125 * HZ;
1724         br->multicast_querier_interval = 255 * HZ;
1725         br->multicast_membership_interval = 260 * HZ;
1726
1727         br->ip4_querier.delay_time = 0;
1728 #if IS_ENABLED(CONFIG_IPV6)
1729         br->ip6_querier.delay_time = 0;
1730 #endif
1731
1732         spin_lock_init(&br->multicast_lock);
1733         setup_timer(&br->multicast_router_timer,
1734                     br_multicast_local_router_expired, 0);
1735         setup_timer(&br->ip4_querier.timer, br_ip4_multicast_querier_expired,
1736                     (unsigned long)br);
1737         setup_timer(&br->ip4_query.timer, br_ip4_multicast_query_expired,
1738                     (unsigned long)br);
1739 #if IS_ENABLED(CONFIG_IPV6)
1740         setup_timer(&br->ip6_querier.timer, br_ip6_multicast_querier_expired,
1741                     (unsigned long)br);
1742         setup_timer(&br->ip6_query.timer, br_ip6_multicast_query_expired,
1743                     (unsigned long)br);
1744 #endif
1745 }
1746
1747 static void __br_multicast_open(struct net_bridge *br,
1748                                 struct bridge_mcast_query *query)
1749 {
1750         query->startup_sent = 0;
1751
1752         if (br->multicast_disabled)
1753                 return;
1754
1755         mod_timer(&query->timer, jiffies);
1756 }
1757
1758 void br_multicast_open(struct net_bridge *br)
1759 {
1760         __br_multicast_open(br, &br->ip4_query);
1761 #if IS_ENABLED(CONFIG_IPV6)
1762         __br_multicast_open(br, &br->ip6_query);
1763 #endif
1764 }
1765
1766 void br_multicast_stop(struct net_bridge *br)
1767 {
1768         struct net_bridge_mdb_htable *mdb;
1769         struct net_bridge_mdb_entry *mp;
1770         struct hlist_node *n;
1771         u32 ver;
1772         int i;
1773
1774         del_timer_sync(&br->multicast_router_timer);
1775         del_timer_sync(&br->ip4_querier.timer);
1776         del_timer_sync(&br->ip4_query.timer);
1777 #if IS_ENABLED(CONFIG_IPV6)
1778         del_timer_sync(&br->ip6_querier.timer);
1779         del_timer_sync(&br->ip6_query.timer);
1780 #endif
1781
1782         spin_lock_bh(&br->multicast_lock);
1783         mdb = mlock_dereference(br->mdb, br);
1784         if (!mdb)
1785                 goto out;
1786
1787         br->mdb = NULL;
1788
1789         ver = mdb->ver;
1790         for (i = 0; i < mdb->max; i++) {
1791                 hlist_for_each_entry_safe(mp, n, &mdb->mhash[i],
1792                                           hlist[ver]) {
1793                         del_timer(&mp->timer);
1794                         mp->timer_armed = false;
1795                         call_rcu_bh(&mp->rcu, br_multicast_free_group);
1796                 }
1797         }
1798
1799         if (mdb->old) {
1800                 spin_unlock_bh(&br->multicast_lock);
1801                 rcu_barrier_bh();
1802                 spin_lock_bh(&br->multicast_lock);
1803                 WARN_ON(mdb->old);
1804         }
1805
1806         mdb->old = mdb;
1807         call_rcu_bh(&mdb->rcu, br_mdb_free);
1808
1809 out:
1810         spin_unlock_bh(&br->multicast_lock);
1811 }
1812
1813 int br_multicast_set_router(struct net_bridge *br, unsigned long val)
1814 {
1815         int err = -ENOENT;
1816
1817         spin_lock_bh(&br->multicast_lock);
1818         if (!netif_running(br->dev))
1819                 goto unlock;
1820
1821         switch (val) {
1822         case 0:
1823         case 2:
1824                 del_timer(&br->multicast_router_timer);
1825                 /* fall through */
1826         case 1:
1827                 br->multicast_router = val;
1828                 err = 0;
1829                 break;
1830
1831         default:
1832                 err = -EINVAL;
1833                 break;
1834         }
1835
1836 unlock:
1837         spin_unlock_bh(&br->multicast_lock);
1838
1839         return err;
1840 }
1841
1842 int br_multicast_set_port_router(struct net_bridge_port *p, unsigned long val)
1843 {
1844         struct net_bridge *br = p->br;
1845         int err = -ENOENT;
1846
1847         spin_lock(&br->multicast_lock);
1848         if (!netif_running(br->dev) || p->state == BR_STATE_DISABLED)
1849                 goto unlock;
1850
1851         switch (val) {
1852         case 0:
1853         case 1:
1854         case 2:
1855                 p->multicast_router = val;
1856                 err = 0;
1857
1858                 if (val < 2 && !hlist_unhashed(&p->rlist))
1859                         hlist_del_init_rcu(&p->rlist);
1860
1861                 if (val == 1)
1862                         break;
1863
1864                 del_timer(&p->multicast_router_timer);
1865
1866                 if (val == 0)
1867                         break;
1868
1869                 br_multicast_add_router(br, p);
1870                 break;
1871
1872         default:
1873                 err = -EINVAL;
1874                 break;
1875         }
1876
1877 unlock:
1878         spin_unlock(&br->multicast_lock);
1879
1880         return err;
1881 }
1882
1883 static void br_multicast_start_querier(struct net_bridge *br,
1884                                        struct bridge_mcast_query *query)
1885 {
1886         struct net_bridge_port *port;
1887
1888         __br_multicast_open(br, query);
1889
1890         list_for_each_entry(port, &br->port_list, list) {
1891                 if (port->state == BR_STATE_DISABLED ||
1892                     port->state == BR_STATE_BLOCKING)
1893                         continue;
1894
1895                 if (query == &br->ip4_query)
1896                         br_multicast_enable(&port->ip4_query);
1897 #if IS_ENABLED(CONFIG_IPV6)
1898                 else
1899                         br_multicast_enable(&port->ip6_query);
1900 #endif
1901         }
1902 }
1903
1904 int br_multicast_toggle(struct net_bridge *br, unsigned long val)
1905 {
1906         int err = 0;
1907         struct net_bridge_mdb_htable *mdb;
1908
1909         spin_lock_bh(&br->multicast_lock);
1910         if (br->multicast_disabled == !val)
1911                 goto unlock;
1912
1913         br->multicast_disabled = !val;
1914         if (br->multicast_disabled)
1915                 goto unlock;
1916
1917         if (!netif_running(br->dev))
1918                 goto unlock;
1919
1920         mdb = mlock_dereference(br->mdb, br);
1921         if (mdb) {
1922                 if (mdb->old) {
1923                         err = -EEXIST;
1924 rollback:
1925                         br->multicast_disabled = !!val;
1926                         goto unlock;
1927                 }
1928
1929                 err = br_mdb_rehash(&br->mdb, mdb->max,
1930                                     br->hash_elasticity);
1931                 if (err)
1932                         goto rollback;
1933         }
1934
1935         br_multicast_start_querier(br, &br->ip4_query);
1936 #if IS_ENABLED(CONFIG_IPV6)
1937         br_multicast_start_querier(br, &br->ip6_query);
1938 #endif
1939
1940 unlock:
1941         spin_unlock_bh(&br->multicast_lock);
1942
1943         return err;
1944 }
1945
1946 int br_multicast_set_querier(struct net_bridge *br, unsigned long val)
1947 {
1948         unsigned long max_delay;
1949
1950         val = !!val;
1951
1952         spin_lock_bh(&br->multicast_lock);
1953         if (br->multicast_querier == val)
1954                 goto unlock;
1955
1956         br->multicast_querier = val;
1957         if (!val)
1958                 goto unlock;
1959
1960         max_delay = br->multicast_query_response_interval;
1961
1962         if (!timer_pending(&br->ip4_querier.timer))
1963                 br->ip4_querier.delay_time = jiffies + max_delay;
1964
1965         br_multicast_start_querier(br, &br->ip4_query);
1966
1967 #if IS_ENABLED(CONFIG_IPV6)
1968         if (!timer_pending(&br->ip6_querier.timer))
1969                 br->ip6_querier.delay_time = jiffies + max_delay;
1970
1971         br_multicast_start_querier(br, &br->ip6_query);
1972 #endif
1973
1974 unlock:
1975         spin_unlock_bh(&br->multicast_lock);
1976
1977         return 0;
1978 }
1979
1980 int br_multicast_set_hash_max(struct net_bridge *br, unsigned long val)
1981 {
1982         int err = -ENOENT;
1983         u32 old;
1984         struct net_bridge_mdb_htable *mdb;
1985
1986         spin_lock(&br->multicast_lock);
1987         if (!netif_running(br->dev))
1988                 goto unlock;
1989
1990         err = -EINVAL;
1991         if (!is_power_of_2(val))
1992                 goto unlock;
1993
1994         mdb = mlock_dereference(br->mdb, br);
1995         if (mdb && val < mdb->size)
1996                 goto unlock;
1997
1998         err = 0;
1999
2000         old = br->hash_max;
2001         br->hash_max = val;
2002
2003         if (mdb) {
2004                 if (mdb->old) {
2005                         err = -EEXIST;
2006 rollback:
2007                         br->hash_max = old;
2008                         goto unlock;
2009                 }
2010
2011                 err = br_mdb_rehash(&br->mdb, br->hash_max,
2012                                     br->hash_elasticity);
2013                 if (err)
2014                         goto rollback;
2015         }
2016
2017 unlock:
2018         spin_unlock(&br->multicast_lock);
2019
2020         return err;
2021 }