Pileus Git - ~andy/linux/blob - 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 #include <net/addrconf.h>
  33 #endif
  34
  35 #include "br_private.h"
  36
  37 static void br_multicast_start_querier(struct net_bridge *br,
  38                                        struct bridge_mcast_query *query);
  39 unsigned int br_mdb_rehash_seq;
  40
  41 static inline int br_ip_equal(const struct br_ip *a, const struct br_ip *b)
  42 {
  43         if (a->proto != b->proto)
  44                 return 0;
  45         if (a->vid != b->vid)
  46                 return 0;
  47         switch (a->proto) {
  48         case htons(ETH_P_IP):
  49                 return a->u.ip4 == b->u.ip4;
  50 #if IS_ENABLED(CONFIG_IPV6)
  51         case htons(ETH_P_IPV6):
  52                 return ipv6_addr_equal(&a->u.ip6, &b->u.ip6);
  53 #endif
  54         }
  55         return 0;
  56 }
  57
  58 static inline int __br_ip4_hash(struct net_bridge_mdb_htable *mdb, __be32 ip,
  59                                 __u16 vid)
  60 {
  61         return jhash_2words((__force u32)ip, vid, mdb->secret) & (mdb->max - 1);
  62 }
  63
  64 #if IS_ENABLED(CONFIG_IPV6)
  65 static inline int __br_ip6_hash(struct net_bridge_mdb_htable *mdb,
  66                                 const struct in6_addr *ip,
  67                                 __u16 vid)
  68 {
  69         return jhash_2words(ipv6_addr_hash(ip), vid,
  70                             mdb->secret) & (mdb->max - 1);
  71 }
  72 #endif
  73
  74 static inline int br_ip_hash(struct net_bridge_mdb_htable *mdb,
  75                              struct br_ip *ip)
  76 {
  77         switch (ip->proto) {
  78         case htons(ETH_P_IP):
  79                 return __br_ip4_hash(mdb, ip->u.ip4, ip->vid);
  80 #if IS_ENABLED(CONFIG_IPV6)
  81         case htons(ETH_P_IPV6):
  82                 return __br_ip6_hash(mdb, &ip->u.ip6, ip->vid);
  83 #endif
  84         }
  85         return 0;
  86 }
  87
  88 static struct net_bridge_mdb_entry *__br_mdb_ip_get(
  89         struct net_bridge_mdb_htable *mdb, struct br_ip *dst, int hash)
  90 {
  91         struct net_bridge_mdb_entry *mp;
  92
  93         hlist_for_each_entry_rcu(mp, &mdb->mhash[hash], hlist[mdb->ver]) {
  94                 if (br_ip_equal(&mp->addr, dst))
  95                         return mp;
  96         }
  97
  98         return NULL;
  99 }
 100
 101 struct net_bridge_mdb_entry *br_mdb_ip_get(struct net_bridge_mdb_htable *mdb,
 102                                            struct br_ip *dst)
 103 {
 104         if (!mdb)
 105                 return NULL;
 106
 107         return __br_mdb_ip_get(mdb, dst, br_ip_hash(mdb, dst));
 108 }
 109
 110 static struct net_bridge_mdb_entry *br_mdb_ip4_get(
 111         struct net_bridge_mdb_htable *mdb, __be32 dst, __u16 vid)
 112 {
 113         struct br_ip br_dst;
 114
 115         br_dst.u.ip4 = dst;
 116         br_dst.proto = htons(ETH_P_IP);
 117         br_dst.vid = vid;
 118
 119         return br_mdb_ip_get(mdb, &br_dst);
 120 }
 121
 122 #if IS_ENABLED(CONFIG_IPV6)
 123 static struct net_bridge_mdb_entry *br_mdb_ip6_get(
 124         struct net_bridge_mdb_htable *mdb, const struct in6_addr *dst,
 125         __u16 vid)
 126 {
 127         struct br_ip br_dst;
 128
 129         br_dst.u.ip6 = *dst;
 130         br_dst.proto = htons(ETH_P_IPV6);
 131         br_dst.vid = vid;
 132
 133         return br_mdb_ip_get(mdb, &br_dst);
 134 }
 135 #endif
 136
 137 struct net_bridge_mdb_entry *br_mdb_get(struct net_bridge *br,
 138                                         struct sk_buff *skb, u16 vid)
 139 {
 140         struct net_bridge_mdb_htable *mdb = rcu_dereference(br->mdb);
 141         struct br_ip ip;
 142
 143         if (br->multicast_disabled)
 144                 return NULL;
 145
 146         if (BR_INPUT_SKB_CB(skb)->igmp)
 147                 return NULL;
 148
 149         ip.proto = skb->protocol;
 150         ip.vid = vid;
 151
 152         switch (skb->protocol) {
 153         case htons(ETH_P_IP):
 154                 ip.u.ip4 = ip_hdr(skb)->daddr;
 155                 break;
 156 #if IS_ENABLED(CONFIG_IPV6)
 157         case htons(ETH_P_IPV6):
 158                 ip.u.ip6 = ipv6_hdr(skb)->daddr;
 159                 break;
 160 #endif
 161         default:
 162                 return NULL;
 163         }
 164
 165         return br_mdb_ip_get(mdb, &ip);
 166 }
 167
 168 static void br_mdb_free(struct rcu_head *head)
 169 {
 170         struct net_bridge_mdb_htable *mdb =
 171                 container_of(head, struct net_bridge_mdb_htable, rcu);
 172         struct net_bridge_mdb_htable *old = mdb->old;
 173
 174         mdb->old = NULL;
 175         kfree(old->mhash);
 176         kfree(old);
 177 }
 178
 179 static int br_mdb_copy(struct net_bridge_mdb_htable *new,
 180                        struct net_bridge_mdb_htable *old,
 181                        int elasticity)
 182 {
 183         struct net_bridge_mdb_entry *mp;
 184         int maxlen;
 185         int len;
 186         int i;
 187
 188         for (i = 0; i < old->max; i++)
 189                 hlist_for_each_entry(mp, &old->mhash[i], hlist[old->ver])
 190                         hlist_add_head(&mp->hlist[new->ver],
 191                                        &new->mhash[br_ip_hash(new, &mp->addr)]);
 192
 193         if (!elasticity)
 194                 return 0;
 195
 196         maxlen = 0;
 197         for (i = 0; i < new->max; i++) {
 198                 len = 0;
 199                 hlist_for_each_entry(mp, &new->mhash[i], hlist[new->ver])
 200                         len++;
 201                 if (len > maxlen)
 202                         maxlen = len;
 203         }
 204
 205         return maxlen > elasticity ? -EINVAL : 0;
 206 }
 207
 208 void br_multicast_free_pg(struct rcu_head *head)
 209 {
 210         struct net_bridge_port_group *p =
 211                 container_of(head, struct net_bridge_port_group, rcu);
 212
 213         kfree(p);
 214 }
 215
 216 static void br_multicast_free_group(struct rcu_head *head)
 217 {
 218         struct net_bridge_mdb_entry *mp =
 219                 container_of(head, struct net_bridge_mdb_entry, rcu);
 220
 221         kfree(mp);
 222 }
 223
 224 static void br_multicast_group_expired(unsigned long data)
 225 {
 226         struct net_bridge_mdb_entry *mp = (void *)data;
 227         struct net_bridge *br = mp->br;
 228         struct net_bridge_mdb_htable *mdb;
 229
 230         spin_lock(&br->multicast_lock);
 231         if (!netif_running(br->dev) || timer_pending(&mp->timer))
 232                 goto out;
 233
 234         mp->mglist = false;
 235
 236         if (mp->ports)
 237                 goto out;
 238
 239         mdb = mlock_dereference(br->mdb, br);
 240
 241         hlist_del_rcu(&mp->hlist[mdb->ver]);
 242         mdb->size--;
 243
 244         call_rcu_bh(&mp->rcu, br_multicast_free_group);
 245
 246 out:
 247         spin_unlock(&br->multicast_lock);
 248 }
 249
 250 static void br_multicast_del_pg(struct net_bridge *br,
 251                                 struct net_bridge_port_group *pg)
 252 {
 253         struct net_bridge_mdb_htable *mdb;
 254         struct net_bridge_mdb_entry *mp;
 255         struct net_bridge_port_group *p;
 256         struct net_bridge_port_group __rcu **pp;
 257
 258         mdb = mlock_dereference(br->mdb, br);
 259
 260         mp = br_mdb_ip_get(mdb, &pg->addr);
 261         if (WARN_ON(!mp))
 262                 return;
 263
 264         for (pp = &mp->ports;
 265              (p = mlock_dereference(*pp, br)) != NULL;
 266              pp = &p->next) {
 267                 if (p != pg)
 268                         continue;
 269
 270                 rcu_assign_pointer(*pp, p->next);
 271                 hlist_del_init(&p->mglist);
 272                 del_timer(&p->timer);
 273                 call_rcu_bh(&p->rcu, br_multicast_free_pg);
 274
 275                 if (!mp->ports && !mp->mglist &&
 276                     netif_running(br->dev))
 277                         mod_timer(&mp->timer, jiffies);
 278
 279                 return;
 280         }
 281
 282         WARN_ON(1);
 283 }
 284
 285 static void br_multicast_port_group_expired(unsigned long data)
 286 {
 287         struct net_bridge_port_group *pg = (void *)data;
 288         struct net_bridge *br = pg->port->br;
 289
 290         spin_lock(&br->multicast_lock);
 291         if (!netif_running(br->dev) || timer_pending(&pg->timer) ||
 292             hlist_unhashed(&pg->mglist) || pg->state & MDB_PERMANENT)
 293                 goto out;
 294
 295         br_multicast_del_pg(br, pg);
 296
 297 out:
 298         spin_unlock(&br->multicast_lock);
 299 }
 300
 301 static int br_mdb_rehash(struct net_bridge_mdb_htable __rcu **mdbp, int max,
 302                          int elasticity)
 303 {
 304         struct net_bridge_mdb_htable *old = rcu_dereference_protected(*mdbp, 1);
 305         struct net_bridge_mdb_htable *mdb;
 306         int err;
 307
 308         mdb = kmalloc(sizeof(*mdb), GFP_ATOMIC);
 309         if (!mdb)
 310                 return -ENOMEM;
 311
 312         mdb->max = max;
 313         mdb->old = old;
 314
 315         mdb->mhash = kzalloc(max * sizeof(*mdb->mhash), GFP_ATOMIC);
 316         if (!mdb->mhash) {
 317                 kfree(mdb);
 318                 return -ENOMEM;
 319         }
 320
 321         mdb->size = old ? old->size : 0;
 322         mdb->ver = old ? old->ver ^ 1 : 0;
 323
 324         if (!old || elasticity)
 325                 get_random_bytes(&mdb->secret, sizeof(mdb->secret));
 326         else
 327                 mdb->secret = old->secret;
 328
 329         if (!old)
 330                 goto out;
 331
 332         err = br_mdb_copy(mdb, old, elasticity);
 333         if (err) {
 334                 kfree(mdb->mhash);
 335                 kfree(mdb);
 336                 return err;
 337         }
 338
 339         br_mdb_rehash_seq++;
 340         call_rcu_bh(&mdb->rcu, br_mdb_free);
 341
 342 out:
 343         rcu_assign_pointer(*mdbp, mdb);
 344
 345         return 0;
 346 }
 347
 348 static struct sk_buff *br_ip4_multicast_alloc_query(struct net_bridge *br,
 349                                                     __be32 group)
 350 {
 351         struct sk_buff *skb;
 352         struct igmphdr *ih;
 353         struct ethhdr *eth;
 354         struct iphdr *iph;
 355
 356         skb = netdev_alloc_skb_ip_align(br->dev, sizeof(*eth) + sizeof(*iph) +
 357                                                  sizeof(*ih) + 4);
 358         if (!skb)
 359                 goto out;
 360
 361         skb->protocol = htons(ETH_P_IP);
 362
 363         skb_reset_mac_header(skb);
 364         eth = eth_hdr(skb);
 365
 366         memcpy(eth->h_source, br->dev->dev_addr, ETH_ALEN);
 367         eth->h_dest[0] = 1;
 368         eth->h_dest[1] = 0;
 369         eth->h_dest[2] = 0x5e;
 370         eth->h_dest[3] = 0;
 371         eth->h_dest[4] = 0;
 372         eth->h_dest[5] = 1;
 373         eth->h_proto = htons(ETH_P_IP);
 374         skb_put(skb, sizeof(*eth));
 375
 376         skb_set_network_header(skb, skb->len);
 377         iph = ip_hdr(skb);
 378
 379         iph->version = 4;
 380         iph->ihl = 6;
 381         iph->tos = 0xc0;
 382         iph->tot_len = htons(sizeof(*iph) + sizeof(*ih) + 4);
 383         iph->id = 0;
 384         iph->frag_off = htons(IP_DF);
 385         iph->ttl = 1;
 386         iph->protocol = IPPROTO_IGMP;
 387         iph->saddr = br->multicast_query_use_ifaddr ?
 388                      inet_select_addr(br->dev, 0, RT_SCOPE_LINK) : 0;
 389         iph->daddr = htonl(INADDR_ALLHOSTS_GROUP);
 390         ((u8 *)&iph[1])[0] = IPOPT_RA;
 391         ((u8 *)&iph[1])[1] = 4;
 392         ((u8 *)&iph[1])[2] = 0;
 393         ((u8 *)&iph[1])[3] = 0;
 394         ip_send_check(iph);
 395         skb_put(skb, 24);
 396
 397         skb_set_transport_header(skb, skb->len);
 398         ih = igmp_hdr(skb);
 399         ih->type = IGMP_HOST_MEMBERSHIP_QUERY;
 400         ih->code = (group ? br->multicast_last_member_interval :
 401                             br->multicast_query_response_interval) /
 402                    (HZ / IGMP_TIMER_SCALE);
 403         ih->group = group;
 404         ih->csum = 0;
 405         ih->csum = ip_compute_csum((void *)ih, sizeof(struct igmphdr));
 406         skb_put(skb, sizeof(*ih));
 407
 408         __skb_pull(skb, sizeof(*eth));
 409
 410 out:
 411         return skb;
 412 }
 413
 414 #if IS_ENABLED(CONFIG_IPV6)
 415 static struct sk_buff *br_ip6_multicast_alloc_query(struct net_bridge *br,
 416                                                     const struct in6_addr *group)
 417 {
 418         struct sk_buff *skb;
 419         struct ipv6hdr *ip6h;
 420         struct mld_msg *mldq;
 421         struct ethhdr *eth;
 422         u8 *hopopt;
 423         unsigned long interval;
 424
 425         skb = netdev_alloc_skb_ip_align(br->dev, sizeof(*eth) + sizeof(*ip6h) +
 426                                                  8 + sizeof(*mldq));
 427         if (!skb)
 428                 goto out;
 429
 430         skb->protocol = htons(ETH_P_IPV6);
 431
 432         /* Ethernet header */
 433         skb_reset_mac_header(skb);
 434         eth = eth_hdr(skb);
 435
 436         memcpy(eth->h_source, br->dev->dev_addr, ETH_ALEN);
 437         eth->h_proto = htons(ETH_P_IPV6);
 438         skb_put(skb, sizeof(*eth));
 439
 440         /* IPv6 header + HbH option */
 441         skb_set_network_header(skb, skb->len);
 442         ip6h = ipv6_hdr(skb);
 443
 444         *(__force __be32 *)ip6h = htonl(0x60000000);
 445         ip6h->payload_len = htons(8 + sizeof(*mldq));
 446         ip6h->nexthdr = IPPROTO_HOPOPTS;
 447         ip6h->hop_limit = 1;
 448         ipv6_addr_set(&ip6h->daddr, htonl(0xff020000), 0, 0, htonl(1));
 449         if (ipv6_dev_get_saddr(dev_net(br->dev), br->dev, &ip6h->daddr, 0,
 450                                &ip6h->saddr)) {
 451                 kfree_skb(skb);
 452                 return NULL;
 453         }
 454         ipv6_eth_mc_map(&ip6h->daddr, eth->h_dest);
 455
 456         hopopt = (u8 *)(ip6h + 1);
 457         hopopt[0] = IPPROTO_ICMPV6;             /* next hdr */
 458         hopopt[1] = 0;                          /* length of HbH */
 459         hopopt[2] = IPV6_TLV_ROUTERALERT;       /* Router Alert */
 460         hopopt[3] = 2;                          /* Length of RA Option */
 461         hopopt[4] = 0;                          /* Type = 0x0000 (MLD) */
 462         hopopt[5] = 0;
 463         hopopt[6] = IPV6_TLV_PAD1;              /* Pad1 */
 464         hopopt[7] = IPV6_TLV_PAD1;              /* Pad1 */
 465
 466         skb_put(skb, sizeof(*ip6h) + 8);
 467
 468         /* ICMPv6 */
 469         skb_set_transport_header(skb, skb->len);
 470         mldq = (struct mld_msg *) icmp6_hdr(skb);
 471
 472         interval = ipv6_addr_any(group) ?
 473                         br->multicast_query_response_interval :
 474                         br->multicast_last_member_interval;
 475
 476         mldq->mld_type = ICMPV6_MGM_QUERY;
 477         mldq->mld_code = 0;
 478         mldq->mld_cksum = 0;
 479         mldq->mld_maxdelay = htons((u16)jiffies_to_msecs(interval));
 480         mldq->mld_reserved = 0;
 481         mldq->mld_mca = *group;
 482
 483         /* checksum */
 484         mldq->mld_cksum = csum_ipv6_magic(&ip6h->saddr, &ip6h->daddr,
 485                                           sizeof(*mldq), IPPROTO_ICMPV6,
 486                                           csum_partial(mldq,
 487                                                        sizeof(*mldq), 0));
 488         skb_put(skb, sizeof(*mldq));
 489
 490         __skb_pull(skb, sizeof(*eth));
 491
 492 out:
 493         return skb;
 494 }
 495 #endif
 496
 497 static struct sk_buff *br_multicast_alloc_query(struct net_bridge *br,
 498                                                 struct br_ip *addr)
 499 {
 500         switch (addr->proto) {
 501         case htons(ETH_P_IP):
 502                 return br_ip4_multicast_alloc_query(br, addr->u.ip4);
 503 #if IS_ENABLED(CONFIG_IPV6)
 504         case htons(ETH_P_IPV6):
 505                 return br_ip6_multicast_alloc_query(br, &addr->u.ip6);
 506 #endif
 507         }
 508         return NULL;
 509 }
 510
 511 static struct net_bridge_mdb_entry *br_multicast_get_group(
 512         struct net_bridge *br, struct net_bridge_port *port,
 513         struct br_ip *group, int hash)
 514 {
 515         struct net_bridge_mdb_htable *mdb;
 516         struct net_bridge_mdb_entry *mp;
 517         unsigned int count = 0;
 518         unsigned int max;
 519         int elasticity;
 520         int err;
 521
 522         mdb = rcu_dereference_protected(br->mdb, 1);
 523         hlist_for_each_entry(mp, &mdb->mhash[hash], hlist[mdb->ver]) {
 524                 count++;
 525                 if (unlikely(br_ip_equal(group, &mp->addr)))
 526                         return mp;
 527         }
 528
 529         elasticity = 0;
 530         max = mdb->max;
 531
 532         if (unlikely(count > br->hash_elasticity && count)) {
 533                 if (net_ratelimit())
 534                         br_info(br, "Multicast hash table "
 535                                 "chain limit reached: %s\n",
 536                                 port ? port->dev->name : br->dev->name);
 537
 538                 elasticity = br->hash_elasticity;
 539         }
 540
 541         if (mdb->size >= max) {
 542                 max *= 2;
 543                 if (unlikely(max > br->hash_max)) {
 544                         br_warn(br, "Multicast hash table maximum of %d "
 545                                 "reached, disabling snooping: %s\n",
 546                                 br->hash_max,
 547                                 port ? port->dev->name : br->dev->name);
 548                         err = -E2BIG;
 549 disable:
 550                         br->multicast_disabled = 1;
 551                         goto err;
 552                 }
 553         }
 554
 555         if (max > mdb->max || elasticity) {
 556                 if (mdb->old) {
 557                         if (net_ratelimit())
 558                                 br_info(br, "Multicast hash table "
 559                                         "on fire: %s\n",
 560                                         port ? port->dev->name : br->dev->name);
 561                         err = -EEXIST;
 562                         goto err;
 563                 }
 564
 565                 err = br_mdb_rehash(&br->mdb, max, elasticity);
 566                 if (err) {
 567                         br_warn(br, "Cannot rehash multicast "
 568                                 "hash table, disabling snooping: %s, %d, %d\n",
 569                                 port ? port->dev->name : br->dev->name,
 570                                 mdb->size, err);
 571                         goto disable;
 572                 }
 573
 574                 err = -EAGAIN;
 575                 goto err;
 576         }
 577
 578         return NULL;
 579
 580 err:
 581         mp = ERR_PTR(err);
 582         return mp;
 583 }
 584
 585 struct net_bridge_mdb_entry *br_multicast_new_group(struct net_bridge *br,
 586         struct net_bridge_port *port, struct br_ip *group)
 587 {
 588         struct net_bridge_mdb_htable *mdb;
 589         struct net_bridge_mdb_entry *mp;
 590         int hash;
 591         int err;
 592
 593         mdb = rcu_dereference_protected(br->mdb, 1);
 594         if (!mdb) {
 595                 err = br_mdb_rehash(&br->mdb, BR_HASH_SIZE, 0);
 596                 if (err)
 597                         return ERR_PTR(err);
 598                 goto rehash;
 599         }
 600
 601         hash = br_ip_hash(mdb, group);
 602         mp = br_multicast_get_group(br, port, group, hash);
 603         switch (PTR_ERR(mp)) {
 604         case 0:
 605                 break;
 606
 607         case -EAGAIN:
 608 rehash:
 609                 mdb = rcu_dereference_protected(br->mdb, 1);
 610                 hash = br_ip_hash(mdb, group);
 611                 break;
 612
 613         default:
 614                 goto out;
 615         }
 616
 617         mp = kzalloc(sizeof(*mp), GFP_ATOMIC);
 618         if (unlikely(!mp))
 619                 return ERR_PTR(-ENOMEM);
 620
 621         mp->br = br;
 622         mp->addr = *group;
 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         unsigned long now = jiffies;
 663         int err;
 664
 665         spin_lock(&br->multicast_lock);
 666         if (!netif_running(br->dev) ||
 667             (port && port->state == BR_STATE_DISABLED))
 668                 goto out;
 669
 670         mp = br_multicast_new_group(br, port, group);
 671         err = PTR_ERR(mp);
 672         if (IS_ERR(mp))
 673                 goto err;
 674
 675         if (!port) {
 676                 mp->mglist = true;
 677                 mod_timer(&mp->timer, now + br->multicast_membership_interval);
 678                 goto out;
 679         }
 680
 681         for (pp = &mp->ports;
 682              (p = mlock_dereference(*pp, br)) != NULL;
 683              pp = &p->next) {
 684                 if (p->port == port)
 685                         goto found;
 686                 if ((unsigned long)p->port < (unsigned long)port)
 687                         break;
 688         }
 689
 690         p = br_multicast_new_port_group(port, group, *pp, MDB_TEMPORARY);
 691         if (unlikely(!p))
 692                 goto err;
 693         rcu_assign_pointer(*pp, p);
 694         br_mdb_notify(br->dev, port, group, RTM_NEWMDB);
 695
 696 found:
 697         mod_timer(&p->timer, now + br->multicast_membership_interval);
 698 out:
 699         err = 0;
 700
 701 err:
 702         spin_unlock(&br->multicast_lock);
 703         return err;
 704 }
 705
 706 static int br_ip4_multicast_add_group(struct net_bridge *br,
 707                                       struct net_bridge_port *port,
 708                                       __be32 group,
 709                                       __u16 vid)
 710 {
 711         struct br_ip br_group;
 712
 713         if (ipv4_is_local_multicast(group))
 714                 return 0;
 715
 716         br_group.u.ip4 = group;
 717         br_group.proto = htons(ETH_P_IP);
 718         br_group.vid = vid;
 719
 720         return br_multicast_add_group(br, port, &br_group);
 721 }
 722
 723 #if IS_ENABLED(CONFIG_IPV6)
 724 static int br_ip6_multicast_add_group(struct net_bridge *br,
 725                                       struct net_bridge_port *port,
 726                                       const struct in6_addr *group,
 727                                       __u16 vid)
 728 {
 729         struct br_ip br_group;
 730
 731         if (ipv6_addr_is_ll_all_nodes(group))
 732                 return 0;
 733
 734         br_group.u.ip6 = *group;
 735         br_group.proto = htons(ETH_P_IPV6);
 736         br_group.vid = vid;
 737
 738         return br_multicast_add_group(br, port, &br_group);
 739 }
 740 #endif
 741
 742 static void br_multicast_router_expired(unsigned long data)
 743 {
 744         struct net_bridge_port *port = (void *)data;
 745         struct net_bridge *br = port->br;
 746
 747         spin_lock(&br->multicast_lock);
 748         if (port->multicast_router != 1 ||
 749             timer_pending(&port->multicast_router_timer) ||
 750             hlist_unhashed(&port->rlist))
 751                 goto out;
 752
 753         hlist_del_init_rcu(&port->rlist);
 754
 755 out:
 756         spin_unlock(&br->multicast_lock);
 757 }
 758
 759 static void br_multicast_local_router_expired(unsigned long data)
 760 {
 761 }
 762
 763 static void br_multicast_querier_expired(struct net_bridge *br,
 764                                          struct bridge_mcast_query *query)
 765 {
 766         spin_lock(&br->multicast_lock);
 767         if (!netif_running(br->dev) || br->multicast_disabled)
 768                 goto out;
 769
 770         br_multicast_start_querier(br, query);
 771
 772 out:
 773         spin_unlock(&br->multicast_lock);
 774 }
 775
 776 static void br_ip4_multicast_querier_expired(unsigned long data)
 777 {
 778         struct net_bridge *br = (void *)data;
 779
 780         br_multicast_querier_expired(br, &br->ip4_query);
 781 }
 782
 783 #if IS_ENABLED(CONFIG_IPV6)
 784 static void br_ip6_multicast_querier_expired(unsigned long data)
 785 {
 786         struct net_bridge *br = (void *)data;
 787
 788         br_multicast_querier_expired(br, &br->ip6_query);
 789 }
 790 #endif
 791
 792 static void __br_multicast_send_query(struct net_bridge *br,
 793                                       struct net_bridge_port *port,
 794                                       struct br_ip *ip)
 795 {
 796         struct sk_buff *skb;
 797
 798         skb = br_multicast_alloc_query(br, ip);
 799         if (!skb)
 800                 return;
 801
 802         if (port) {
 803                 __skb_push(skb, sizeof(struct ethhdr));
 804                 skb->dev = port->dev;
 805                 NF_HOOK(NFPROTO_BRIDGE, NF_BR_LOCAL_OUT, skb, NULL, skb->dev,
 806                         dev_queue_xmit);
 807         } else
 808                 netif_rx(skb);
 809 }
 810
 811 static void br_multicast_send_query(struct net_bridge *br,
 812                                     struct net_bridge_port *port,
 813                                     struct bridge_mcast_query *query)
 814 {
 815         unsigned long time;
 816         struct br_ip br_group;
 817         struct bridge_mcast_querier *querier = NULL;
 818
 819         if (!netif_running(br->dev) || br->multicast_disabled ||
 820             !br->multicast_querier)
 821                 return;
 822
 823         memset(&br_group.u, 0, sizeof(br_group.u));
 824
 825         if (port ? (query == &port->ip4_query) :
 826                    (query == &br->ip4_query)) {
 827                 querier = &br->ip4_querier;
 828                 br_group.proto = htons(ETH_P_IP);
 829 #if IS_ENABLED(CONFIG_IPV6)
 830         } else {
 831                 querier = &br->ip6_querier;
 832                 br_group.proto = htons(ETH_P_IPV6);
 833 #endif
 834         }
 835
 836         if (!querier || timer_pending(&querier->timer))
 837                 return;
 838
 839         __br_multicast_send_query(br, port, &br_group);
 840
 841         time = jiffies;
 842         time += query->startup_sent < br->multicast_startup_query_count ?
 843                 br->multicast_startup_query_interval :
 844                 br->multicast_query_interval;
 845         mod_timer(&query->timer, time);
 846 }
 847
 848 static void br_multicast_port_query_expired(struct net_bridge_port *port,
 849                                             struct bridge_mcast_query *query)
 850 {
 851         struct net_bridge *br = port->br;
 852
 853         spin_lock(&br->multicast_lock);
 854         if (port->state == BR_STATE_DISABLED ||
 855             port->state == BR_STATE_BLOCKING)
 856                 goto out;
 857
 858         if (query->startup_sent < br->multicast_startup_query_count)
 859                 query->startup_sent++;
 860
 861         br_multicast_send_query(port->br, port, query);
 862
 863 out:
 864         spin_unlock(&br->multicast_lock);
 865 }
 866
 867 static void br_ip4_multicast_port_query_expired(unsigned long data)
 868 {
 869         struct net_bridge_port *port = (void *)data;
 870
 871         br_multicast_port_query_expired(port, &port->ip4_query);
 872 }
 873
 874 #if IS_ENABLED(CONFIG_IPV6)
 875 static void br_ip6_multicast_port_query_expired(unsigned long data)
 876 {
 877         struct net_bridge_port *port = (void *)data;
 878
 879         br_multicast_port_query_expired(port, &port->ip6_query);
 880 }
 881 #endif
 882
 883 void br_multicast_add_port(struct net_bridge_port *port)
 884 {
 885         port->multicast_router = 1;
 886
 887         setup_timer(&port->multicast_router_timer, br_multicast_router_expired,
 888                     (unsigned long)port);
 889         setup_timer(&port->ip4_query.timer, br_ip4_multicast_port_query_expired,
 890                     (unsigned long)port);
 891 #if IS_ENABLED(CONFIG_IPV6)
 892         setup_timer(&port->ip6_query.timer, br_ip6_multicast_port_query_expired,
 893                     (unsigned long)port);
 894 #endif
 895 }
 896
 897 void br_multicast_del_port(struct net_bridge_port *port)
 898 {
 899         del_timer_sync(&port->multicast_router_timer);
 900 }
 901
 902 static void br_multicast_enable(struct bridge_mcast_query *query)
 903 {
 904         query->startup_sent = 0;
 905
 906         if (try_to_del_timer_sync(&query->timer) >= 0 ||
 907             del_timer(&query->timer))
 908                 mod_timer(&query->timer, jiffies);
 909 }
 910
 911 void br_multicast_enable_port(struct net_bridge_port *port)
 912 {
 913         struct net_bridge *br = port->br;
 914
 915         spin_lock(&br->multicast_lock);
 916         if (br->multicast_disabled || !netif_running(br->dev))
 917                 goto out;
 918
 919         br_multicast_enable(&port->ip4_query);
 920 #if IS_ENABLED(CONFIG_IPV6)
 921         br_multicast_enable(&port->ip6_query);
 922 #endif
 923
 924 out:
 925         spin_unlock(&br->multicast_lock);
 926 }
 927
 928 void br_multicast_disable_port(struct net_bridge_port *port)
 929 {
 930         struct net_bridge *br = port->br;
 931         struct net_bridge_port_group *pg;
 932         struct hlist_node *n;
 933
 934         spin_lock(&br->multicast_lock);
 935         hlist_for_each_entry_safe(pg, n, &port->mglist, mglist)
 936                 br_multicast_del_pg(br, pg);
 937
 938         if (!hlist_unhashed(&port->rlist))
 939                 hlist_del_init_rcu(&port->rlist);
 940         del_timer(&port->multicast_router_timer);
 941         del_timer(&port->ip4_query.timer);
 942 #if IS_ENABLED(CONFIG_IPV6)
 943         del_timer(&port->ip6_query.timer);
 944 #endif
 945         spin_unlock(&br->multicast_lock);
 946 }
 947
 948 static int br_ip4_multicast_igmp3_report(struct net_bridge *br,
 949                                          struct net_bridge_port *port,
 950                                          struct sk_buff *skb,
 951                                          u16 vid)
 952 {
 953         struct igmpv3_report *ih;
 954         struct igmpv3_grec *grec;
 955         int i;
 956         int len;
 957         int num;
 958         int type;
 959         int err = 0;
 960         __be32 group;
 961
 962         if (!pskb_may_pull(skb, sizeof(*ih)))
 963                 return -EINVAL;
 964
 965         ih = igmpv3_report_hdr(skb);
 966         num = ntohs(ih->ngrec);
 967         len = sizeof(*ih);
 968
 969         for (i = 0; i < num; i++) {
 970                 len += sizeof(*grec);
 971                 if (!pskb_may_pull(skb, len))
 972                         return -EINVAL;
 973
 974                 grec = (void *)(skb->data + len - sizeof(*grec));
 975                 group = grec->grec_mca;
 976                 type = grec->grec_type;
 977
 978                 len += ntohs(grec->grec_nsrcs) * 4;
 979                 if (!pskb_may_pull(skb, len))
 980                         return -EINVAL;
 981
 982                 /* We treat this as an IGMPv2 report for now. */
 983                 switch (type) {
 984                 case IGMPV3_MODE_IS_INCLUDE:
 985                 case IGMPV3_MODE_IS_EXCLUDE:
 986                 case IGMPV3_CHANGE_TO_INCLUDE:
 987                 case IGMPV3_CHANGE_TO_EXCLUDE:
 988                 case IGMPV3_ALLOW_NEW_SOURCES:
 989                 case IGMPV3_BLOCK_OLD_SOURCES:
 990                         break;
 991
 992                 default:
 993                         continue;
 994                 }
 995
 996                 err = br_ip4_multicast_add_group(br, port, group, vid);
 997                 if (err)
 998                         break;
 999         }
1000
1001         return err;
1002 }
1003
1004 #if IS_ENABLED(CONFIG_IPV6)
1005 static int br_ip6_multicast_mld2_report(struct net_bridge *br,
1006                                         struct net_bridge_port *port,
1007                                         struct sk_buff *skb,
1008                                         u16 vid)
1009 {
1010         struct icmp6hdr *icmp6h;
1011         struct mld2_grec *grec;
1012         int i;
1013         int len;
1014         int num;
1015         int err = 0;
1016
1017         if (!pskb_may_pull(skb, sizeof(*icmp6h)))
1018                 return -EINVAL;
1019
1020         icmp6h = icmp6_hdr(skb);
1021         num = ntohs(icmp6h->icmp6_dataun.un_data16[1]);
1022         len = sizeof(*icmp6h);
1023
1024         for (i = 0; i < num; i++) {
1025                 __be16 *nsrcs, _nsrcs;
1026
1027                 nsrcs = skb_header_pointer(skb,
1028                                            len + offsetof(struct mld2_grec,
1029                                                           grec_nsrcs),
1030                                            sizeof(_nsrcs), &_nsrcs);
1031                 if (!nsrcs)
1032                         return -EINVAL;
1033
1034                 if (!pskb_may_pull(skb,
1035                                    len + sizeof(*grec) +
1036                                    sizeof(struct in6_addr) * ntohs(*nsrcs)))
1037                         return -EINVAL;
1038
1039                 grec = (struct mld2_grec *)(skb->data + len);
1040                 len += sizeof(*grec) +
1041                        sizeof(struct in6_addr) * ntohs(*nsrcs);
1042
1043                 /* We treat these as MLDv1 reports for now. */
1044                 switch (grec->grec_type) {
1045                 case MLD2_MODE_IS_INCLUDE:
1046                 case MLD2_MODE_IS_EXCLUDE:
1047                 case MLD2_CHANGE_TO_INCLUDE:
1048                 case MLD2_CHANGE_TO_EXCLUDE:
1049                 case MLD2_ALLOW_NEW_SOURCES:
1050                 case MLD2_BLOCK_OLD_SOURCES:
1051                         break;
1052
1053                 default:
1054                         continue;
1055                 }
1056
1057                 err = br_ip6_multicast_add_group(br, port, &grec->grec_mca,
1058                                                  vid);
1059                 if (!err)
1060                         break;
1061         }
1062
1063         return err;
1064 }
1065 #endif
1066
1067 static void
1068 br_multicast_update_querier_timer(struct net_bridge *br,
1069                                   struct bridge_mcast_querier *querier,
1070                                   unsigned long max_delay)
1071 {
1072         if (!timer_pending(&querier->timer))
1073                 querier->delay_time = jiffies + max_delay;
1074
1075         mod_timer(&querier->timer, jiffies + br->multicast_querier_interval);
1076 }
1077
1078 /*
1079  * Add port to router_list
1080  *  list is maintained ordered by pointer value
1081  *  and locked by br->multicast_lock and RCU
1082  */
1083 static void br_multicast_add_router(struct net_bridge *br,
1084                                     struct net_bridge_port *port)
1085 {
1086         struct net_bridge_port *p;
1087         struct hlist_node *slot = NULL;
1088
1089         hlist_for_each_entry(p, &br->router_list, rlist) {
1090                 if ((unsigned long) port >= (unsigned long) p)
1091                         break;
1092                 slot = &p->rlist;
1093         }
1094
1095         if (slot)
1096                 hlist_add_after_rcu(slot, &port->rlist);
1097         else
1098                 hlist_add_head_rcu(&port->rlist, &br->router_list);
1099 }
1100
1101 static void br_multicast_mark_router(struct net_bridge *br,
1102                                      struct net_bridge_port *port)
1103 {
1104         unsigned long now = jiffies;
1105
1106         if (!port) {
1107                 if (br->multicast_router == 1)
1108                         mod_timer(&br->multicast_router_timer,
1109                                   now + br->multicast_querier_interval);
1110                 return;
1111         }
1112
1113         if (port->multicast_router != 1)
1114                 return;
1115
1116         if (!hlist_unhashed(&port->rlist))
1117                 goto timer;
1118
1119         br_multicast_add_router(br, port);
1120
1121 timer:
1122         mod_timer(&port->multicast_router_timer,
1123                   now + br->multicast_querier_interval);
1124 }
1125
1126 static void br_multicast_query_received(struct net_bridge *br,
1127                                         struct net_bridge_port *port,
1128                                         struct bridge_mcast_querier *querier,
1129                                         int saddr,
1130                                         bool is_general_query,
1131                                         unsigned long max_delay)
1132 {
1133         if (saddr && is_general_query)
1134                 br_multicast_update_querier_timer(br, querier, max_delay);
1135         else if (timer_pending(&querier->timer))
1136                 return;
1137
1138         br_multicast_mark_router(br, port);
1139 }
1140
1141 static int br_ip4_multicast_query(struct net_bridge *br,
1142                                   struct net_bridge_port *port,
1143                                   struct sk_buff *skb,
1144                                   u16 vid)
1145 {
1146         const struct iphdr *iph = ip_hdr(skb);
1147         struct igmphdr *ih = igmp_hdr(skb);
1148         struct net_bridge_mdb_entry *mp;
1149         struct igmpv3_query *ih3;
1150         struct net_bridge_port_group *p;
1151         struct net_bridge_port_group __rcu **pp;
1152         unsigned long max_delay;
1153         unsigned long now = jiffies;
1154         __be32 group;
1155         int err = 0;
1156
1157         spin_lock(&br->multicast_lock);
1158         if (!netif_running(br->dev) ||
1159             (port && port->state == BR_STATE_DISABLED))
1160                 goto out;
1161
1162         group = ih->group;
1163
1164         if (skb->len == sizeof(*ih)) {
1165                 max_delay = ih->code * (HZ / IGMP_TIMER_SCALE);
1166
1167                 if (!max_delay) {
1168                         max_delay = 10 * HZ;
1169                         group = 0;
1170                 }
1171         } else {
1172                 if (!pskb_may_pull(skb, sizeof(struct igmpv3_query))) {
1173                         err = -EINVAL;
1174                         goto out;
1175                 }
1176
1177                 ih3 = igmpv3_query_hdr(skb);
1178                 if (ih3->nsrcs)
1179                         goto out;
1180
1181                 max_delay = ih3->code ?
1182                             IGMPV3_MRC(ih3->code) * (HZ / IGMP_TIMER_SCALE) : 1;
1183         }
1184
1185         /* RFC2236+RFC3376 (IGMPv2+IGMPv3) require the multicast link layer
1186          * all-systems destination addresses (224.0.0.1) for general queries
1187          */
1188         if (!group && iph->daddr != htonl(INADDR_ALLHOSTS_GROUP)) {
1189                 err = -EINVAL;
1190                 goto out;
1191         }
1192
1193         br_multicast_query_received(br, port, &br->ip4_querier, !!iph->saddr,
1194                                     !group, max_delay);
1195
1196         if (!group)
1197                 goto out;
1198
1199         mp = br_mdb_ip4_get(mlock_dereference(br->mdb, br), group, vid);
1200         if (!mp)
1201                 goto out;
1202
1203         max_delay *= br->multicast_last_member_count;
1204
1205         if (mp->mglist &&
1206             (timer_pending(&mp->timer) ?
1207              time_after(mp->timer.expires, now + max_delay) :
1208              try_to_del_timer_sync(&mp->timer) >= 0))
1209                 mod_timer(&mp->timer, now + max_delay);
1210
1211         for (pp = &mp->ports;
1212              (p = mlock_dereference(*pp, br)) != NULL;
1213              pp = &p->next) {
1214                 if (timer_pending(&p->timer) ?
1215                     time_after(p->timer.expires, now + max_delay) :
1216                     try_to_del_timer_sync(&p->timer) >= 0)
1217                         mod_timer(&p->timer, now + max_delay);
1218         }
1219
1220 out:
1221         spin_unlock(&br->multicast_lock);
1222         return err;
1223 }
1224
1225 #if IS_ENABLED(CONFIG_IPV6)
1226 static int br_ip6_multicast_query(struct net_bridge *br,
1227                                   struct net_bridge_port *port,
1228                                   struct sk_buff *skb,
1229                                   u16 vid)
1230 {
1231         const struct ipv6hdr *ip6h = ipv6_hdr(skb);
1232         struct mld_msg *mld;
1233         struct net_bridge_mdb_entry *mp;
1234         struct mld2_query *mld2q;
1235         struct net_bridge_port_group *p;
1236         struct net_bridge_port_group __rcu **pp;
1237         unsigned long max_delay;
1238         unsigned long now = jiffies;
1239         const struct in6_addr *group = NULL;
1240         bool is_general_query;
1241         int err = 0;
1242
1243         spin_lock(&br->multicast_lock);
1244         if (!netif_running(br->dev) ||
1245             (port && port->state == BR_STATE_DISABLED))
1246                 goto out;
1247
1248         /* RFC2710+RFC3810 (MLDv1+MLDv2) require link-local source addresses */
1249         if (!(ipv6_addr_type(&ip6h->saddr) & IPV6_ADDR_LINKLOCAL)) {
1250                 err = -EINVAL;
1251                 goto out;
1252         }
1253
1254         if (skb->len == sizeof(*mld)) {
1255                 if (!pskb_may_pull(skb, sizeof(*mld))) {
1256                         err = -EINVAL;
1257                         goto out;
1258                 }
1259                 mld = (struct mld_msg *) icmp6_hdr(skb);
1260                 max_delay = msecs_to_jiffies(ntohs(mld->mld_maxdelay));
1261                 if (max_delay)
1262                         group = &mld->mld_mca;
1263         } else {
1264                 if (!pskb_may_pull(skb, sizeof(*mld2q))) {
1265                         err = -EINVAL;
1266                         goto out;
1267                 }
1268                 mld2q = (struct mld2_query *)icmp6_hdr(skb);
1269                 if (!mld2q->mld2q_nsrcs)
1270                         group = &mld2q->mld2q_mca;
1271
1272                 max_delay = max(msecs_to_jiffies(mldv2_mrc(mld2q)), 1UL);
1273         }
1274
1275         is_general_query = group && ipv6_addr_any(group);
1276
1277         /* RFC2710+RFC3810 (MLDv1+MLDv2) require the multicast link layer
1278          * all-nodes destination address (ff02::1) for general queries
1279          */
1280         if (is_general_query && !ipv6_addr_is_ll_all_nodes(&ip6h->daddr)) {
1281                 err = -EINVAL;
1282                 goto out;
1283         }
1284
1285         br_multicast_query_received(br, port, &br->ip6_querier,
1286                                     !ipv6_addr_any(&ip6h->saddr),
1287                                     is_general_query, max_delay);
1288
1289         if (!group)
1290                 goto out;
1291
1292         mp = br_mdb_ip6_get(mlock_dereference(br->mdb, br), group, vid);
1293         if (!mp)
1294                 goto out;
1295
1296         max_delay *= br->multicast_last_member_count;
1297         if (mp->mglist &&
1298             (timer_pending(&mp->timer) ?
1299              time_after(mp->timer.expires, now + max_delay) :
1300              try_to_del_timer_sync(&mp->timer) >= 0))
1301                 mod_timer(&mp->timer, now + max_delay);
1302
1303         for (pp = &mp->ports;
1304              (p = mlock_dereference(*pp, br)) != NULL;
1305              pp = &p->next) {
1306                 if (timer_pending(&p->timer) ?
1307                     time_after(p->timer.expires, now + max_delay) :
1308                     try_to_del_timer_sync(&p->timer) >= 0)
1309                         mod_timer(&p->timer, now + max_delay);
1310         }
1311
1312 out:
1313         spin_unlock(&br->multicast_lock);
1314         return err;
1315 }
1316 #endif
1317
1318 static void br_multicast_leave_group(struct net_bridge *br,
1319                                      struct net_bridge_port *port,
1320                                      struct br_ip *group,
1321                                      struct bridge_mcast_querier *querier,
1322                                      struct bridge_mcast_query *query)
1323 {
1324         struct net_bridge_mdb_htable *mdb;
1325         struct net_bridge_mdb_entry *mp;
1326         struct net_bridge_port_group *p;
1327         unsigned long now;
1328         unsigned long time;
1329
1330         spin_lock(&br->multicast_lock);
1331         if (!netif_running(br->dev) ||
1332             (port && port->state == BR_STATE_DISABLED) ||
1333             timer_pending(&querier->timer))
1334                 goto out;
1335
1336         mdb = mlock_dereference(br->mdb, br);
1337         mp = br_mdb_ip_get(mdb, group);
1338         if (!mp)
1339                 goto out;
1340
1341         if (br->multicast_querier) {
1342                 __br_multicast_send_query(br, port, &mp->addr);
1343
1344                 time = jiffies + br->multicast_last_member_count *
1345                                  br->multicast_last_member_interval;
1346
1347                 mod_timer(&query->timer, time);
1348
1349                 for (p = mlock_dereference(mp->ports, br);
1350                      p != NULL;
1351                      p = mlock_dereference(p->next, br)) {
1352                         if (p->port != port)
1353                                 continue;
1354
1355                         if (!hlist_unhashed(&p->mglist) &&
1356                             (timer_pending(&p->timer) ?
1357                              time_after(p->timer.expires, time) :
1358                              try_to_del_timer_sync(&p->timer) >= 0)) {
1359                                 mod_timer(&p->timer, time);
1360                         }
1361
1362                         break;
1363                 }
1364         }
1365
1366         if (port && (port->flags & BR_MULTICAST_FAST_LEAVE)) {
1367                 struct net_bridge_port_group __rcu **pp;
1368
1369                 for (pp = &mp->ports;
1370                      (p = mlock_dereference(*pp, br)) != NULL;
1371                      pp = &p->next) {
1372                         if (p->port != port)
1373                                 continue;
1374
1375                         rcu_assign_pointer(*pp, p->next);
1376                         hlist_del_init(&p->mglist);
1377                         del_timer(&p->timer);
1378                         call_rcu_bh(&p->rcu, br_multicast_free_pg);
1379                         br_mdb_notify(br->dev, port, group, RTM_DELMDB);
1380
1381                         if (!mp->ports && !mp->mglist &&
1382                             netif_running(br->dev))
1383                                 mod_timer(&mp->timer, jiffies);
1384                 }
1385                 goto out;
1386         }
1387
1388         now = jiffies;
1389         time = now + br->multicast_last_member_count *
1390                      br->multicast_last_member_interval;
1391
1392         if (!port) {
1393                 if (mp->mglist &&
1394                     (timer_pending(&mp->timer) ?
1395                      time_after(mp->timer.expires, time) :
1396                      try_to_del_timer_sync(&mp->timer) >= 0)) {
1397                         mod_timer(&mp->timer, time);
1398                 }
1399
1400                 goto out;
1401         }
1402
1403         for (p = mlock_dereference(mp->ports, br);
1404              p != NULL;
1405              p = mlock_dereference(p->next, br)) {
1406                 if (p->port != port)
1407                         continue;
1408
1409                 if (!hlist_unhashed(&p->mglist) &&
1410                     (timer_pending(&p->timer) ?
1411                      time_after(p->timer.expires, time) :
1412                      try_to_del_timer_sync(&p->timer) >= 0)) {
1413                         mod_timer(&p->timer, time);
1414                 }
1415
1416                 break;
1417         }
1418 out:
1419         spin_unlock(&br->multicast_lock);
1420 }
1421
1422 static void br_ip4_multicast_leave_group(struct net_bridge *br,
1423                                          struct net_bridge_port *port,
1424                                          __be32 group,
1425                                          __u16 vid)
1426 {
1427         struct br_ip br_group;
1428         struct bridge_mcast_query *query = port ? &port->ip4_query :
1429                                                   &br->ip4_query;
1430
1431         if (ipv4_is_local_multicast(group))
1432                 return;
1433
1434         br_group.u.ip4 = group;
1435         br_group.proto = htons(ETH_P_IP);
1436         br_group.vid = vid;
1437
1438         br_multicast_leave_group(br, port, &br_group, &br->ip4_querier, query);
1439 }
1440
1441 #if IS_ENABLED(CONFIG_IPV6)
1442 static void br_ip6_multicast_leave_group(struct net_bridge *br,
1443                                          struct net_bridge_port *port,
1444                                          const struct in6_addr *group,
1445                                          __u16 vid)
1446 {
1447         struct br_ip br_group;
1448         struct bridge_mcast_query *query = port ? &port->ip6_query :
1449                                                   &br->ip6_query;
1450
1451
1452         if (ipv6_addr_is_ll_all_nodes(group))
1453                 return;
1454
1455         br_group.u.ip6 = *group;
1456         br_group.proto = htons(ETH_P_IPV6);
1457         br_group.vid = vid;
1458
1459         br_multicast_leave_group(br, port, &br_group, &br->ip6_querier, query);
1460 }
1461 #endif
1462
1463 static int br_multicast_ipv4_rcv(struct net_bridge *br,
1464                                  struct net_bridge_port *port,
1465                                  struct sk_buff *skb,
1466                                  u16 vid)
1467 {
1468         struct sk_buff *skb2 = skb;
1469         const struct iphdr *iph;
1470         struct igmphdr *ih;
1471         unsigned int len;
1472         unsigned int offset;
1473         int err;
1474
1475         /* We treat OOM as packet loss for now. */
1476         if (!pskb_may_pull(skb, sizeof(*iph)))
1477                 return -EINVAL;
1478
1479         iph = ip_hdr(skb);
1480
1481         if (iph->ihl < 5 || iph->version != 4)
1482                 return -EINVAL;
1483
1484         if (!pskb_may_pull(skb, ip_hdrlen(skb)))
1485                 return -EINVAL;
1486
1487         iph = ip_hdr(skb);
1488
1489         if (unlikely(ip_fast_csum((u8 *)iph, iph->ihl)))
1490                 return -EINVAL;
1491
1492         if (iph->protocol != IPPROTO_IGMP) {
1493                 if (!ipv4_is_local_multicast(iph->daddr))
1494                         BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1495                 return 0;
1496         }
1497
1498         len = ntohs(iph->tot_len);
1499         if (skb->len < len || len < ip_hdrlen(skb))
1500                 return -EINVAL;
1501
1502         if (skb->len > len) {
1503                 skb2 = skb_clone(skb, GFP_ATOMIC);
1504                 if (!skb2)
1505                         return -ENOMEM;
1506
1507                 err = pskb_trim_rcsum(skb2, len);
1508                 if (err)
1509                         goto err_out;
1510         }
1511
1512         len -= ip_hdrlen(skb2);
1513         offset = skb_network_offset(skb2) + ip_hdrlen(skb2);
1514         __skb_pull(skb2, offset);
1515         skb_reset_transport_header(skb2);
1516
1517         err = -EINVAL;
1518         if (!pskb_may_pull(skb2, sizeof(*ih)))
1519                 goto out;
1520
1521         switch (skb2->ip_summed) {
1522         case CHECKSUM_COMPLETE:
1523                 if (!csum_fold(skb2->csum))
1524                         break;
1525                 /* fall through */
1526         case CHECKSUM_NONE:
1527                 skb2->csum = 0;
1528                 if (skb_checksum_complete(skb2))
1529                         goto out;
1530         }
1531
1532         err = 0;
1533
1534         BR_INPUT_SKB_CB(skb)->igmp = 1;
1535         ih = igmp_hdr(skb2);
1536
1537         switch (ih->type) {
1538         case IGMP_HOST_MEMBERSHIP_REPORT:
1539         case IGMPV2_HOST_MEMBERSHIP_REPORT:
1540                 BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1541                 err = br_ip4_multicast_add_group(br, port, ih->group, vid);
1542                 break;
1543         case IGMPV3_HOST_MEMBERSHIP_REPORT:
1544                 err = br_ip4_multicast_igmp3_report(br, port, skb2, vid);
1545                 break;
1546         case IGMP_HOST_MEMBERSHIP_QUERY:
1547                 err = br_ip4_multicast_query(br, port, skb2, vid);
1548                 break;
1549         case IGMP_HOST_LEAVE_MESSAGE:
1550                 br_ip4_multicast_leave_group(br, port, ih->group, vid);
1551                 break;
1552         }
1553
1554 out:
1555         __skb_push(skb2, offset);
1556 err_out:
1557         if (skb2 != skb)
1558                 kfree_skb(skb2);
1559         return err;
1560 }
1561
1562 #if IS_ENABLED(CONFIG_IPV6)
1563 static int br_multicast_ipv6_rcv(struct net_bridge *br,
1564                                  struct net_bridge_port *port,
1565                                  struct sk_buff *skb,
1566                                  u16 vid)
1567 {
1568         struct sk_buff *skb2;
1569         const struct ipv6hdr *ip6h;
1570         u8 icmp6_type;
1571         u8 nexthdr;
1572         __be16 frag_off;
1573         unsigned int len;
1574         int offset;
1575         int err;
1576
1577         if (!pskb_may_pull(skb, sizeof(*ip6h)))
1578                 return -EINVAL;
1579
1580         ip6h = ipv6_hdr(skb);
1581
1582         /*
1583          * We're interested in MLD messages only.
1584          *  - Version is 6
1585          *  - MLD has always Router Alert hop-by-hop option
1586          *  - But we do not support jumbrograms.
1587          */
1588         if (ip6h->version != 6)
1589                 return 0;
1590
1591         /* Prevent flooding this packet if there is no listener present */
1592         if (!ipv6_addr_is_ll_all_nodes(&ip6h->daddr))
1593                 BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1594
1595         if (ip6h->nexthdr != IPPROTO_HOPOPTS ||
1596             ip6h->payload_len == 0)
1597                 return 0;
1598
1599         len = ntohs(ip6h->payload_len) + sizeof(*ip6h);
1600         if (skb->len < len)
1601                 return -EINVAL;
1602
1603         nexthdr = ip6h->nexthdr;
1604         offset = ipv6_skip_exthdr(skb, sizeof(*ip6h), &nexthdr, &frag_off);
1605
1606         if (offset < 0 || nexthdr != IPPROTO_ICMPV6)
1607                 return 0;
1608
1609         /* Okay, we found ICMPv6 header */
1610         skb2 = skb_clone(skb, GFP_ATOMIC);
1611         if (!skb2)
1612                 return -ENOMEM;
1613
1614         err = -EINVAL;
1615         if (!pskb_may_pull(skb2, offset + sizeof(struct icmp6hdr)))
1616                 goto out;
1617
1618         len -= offset - skb_network_offset(skb2);
1619
1620         __skb_pull(skb2, offset);
1621         skb_reset_transport_header(skb2);
1622         skb_postpull_rcsum(skb2, skb_network_header(skb2),
1623                            skb_network_header_len(skb2));
1624
1625         icmp6_type = icmp6_hdr(skb2)->icmp6_type;
1626
1627         switch (icmp6_type) {
1628         case ICMPV6_MGM_QUERY:
1629         case ICMPV6_MGM_REPORT:
1630         case ICMPV6_MGM_REDUCTION:
1631         case ICMPV6_MLD2_REPORT:
1632                 break;
1633         default:
1634                 err = 0;
1635                 goto out;
1636         }
1637
1638         /* Okay, we found MLD message. Check further. */
1639         if (skb2->len > len) {
1640                 err = pskb_trim_rcsum(skb2, len);
1641                 if (err)
1642                         goto out;
1643                 err = -EINVAL;
1644         }
1645
1646         ip6h = ipv6_hdr(skb2);
1647
1648         switch (skb2->ip_summed) {
1649         case CHECKSUM_COMPLETE:
1650                 if (!csum_ipv6_magic(&ip6h->saddr, &ip6h->daddr, skb2->len,
1651                                         IPPROTO_ICMPV6, skb2->csum))
1652                         break;
1653                 /*FALLTHROUGH*/
1654         case CHECKSUM_NONE:
1655                 skb2->csum = ~csum_unfold(csum_ipv6_magic(&ip6h->saddr,
1656                                                         &ip6h->daddr,
1657                                                         skb2->len,
1658                                                         IPPROTO_ICMPV6, 0));
1659                 if (__skb_checksum_complete(skb2))
1660                         goto out;
1661         }
1662
1663         err = 0;
1664
1665         BR_INPUT_SKB_CB(skb)->igmp = 1;
1666
1667         switch (icmp6_type) {
1668         case ICMPV6_MGM_REPORT:
1669             {
1670                 struct mld_msg *mld;
1671                 if (!pskb_may_pull(skb2, sizeof(*mld))) {
1672                         err = -EINVAL;
1673                         goto out;
1674                 }
1675                 mld = (struct mld_msg *)skb_transport_header(skb2);
1676                 BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1677                 err = br_ip6_multicast_add_group(br, port, &mld->mld_mca, vid);
1678                 break;
1679             }
1680         case ICMPV6_MLD2_REPORT:
1681                 err = br_ip6_multicast_mld2_report(br, port, skb2, vid);
1682                 break;
1683         case ICMPV6_MGM_QUERY:
1684                 err = br_ip6_multicast_query(br, port, skb2, vid);
1685                 break;
1686         case ICMPV6_MGM_REDUCTION:
1687             {
1688                 struct mld_msg *mld;
1689                 if (!pskb_may_pull(skb2, sizeof(*mld))) {
1690                         err = -EINVAL;
1691                         goto out;
1692                 }
1693                 mld = (struct mld_msg *)skb_transport_header(skb2);
1694                 br_ip6_multicast_leave_group(br, port, &mld->mld_mca, vid);
1695             }
1696         }
1697
1698 out:
1699         kfree_skb(skb2);
1700         return err;
1701 }
1702 #endif
1703
1704 int br_multicast_rcv(struct net_bridge *br, struct net_bridge_port *port,
1705                      struct sk_buff *skb, u16 vid)
1706 {
1707         BR_INPUT_SKB_CB(skb)->igmp = 0;
1708         BR_INPUT_SKB_CB(skb)->mrouters_only = 0;
1709
1710         if (br->multicast_disabled)
1711                 return 0;
1712
1713         switch (skb->protocol) {
1714         case htons(ETH_P_IP):
1715                 return br_multicast_ipv4_rcv(br, port, skb, vid);
1716 #if IS_ENABLED(CONFIG_IPV6)
1717         case htons(ETH_P_IPV6):
1718                 return br_multicast_ipv6_rcv(br, port, skb, vid);
1719 #endif
1720         }
1721
1722         return 0;
1723 }
1724
1725 static void br_multicast_query_expired(struct net_bridge *br,
1726                                        struct bridge_mcast_query *query)
1727 {
1728         spin_lock(&br->multicast_lock);
1729         if (query->startup_sent < br->multicast_startup_query_count)
1730                 query->startup_sent++;
1731
1732         br_multicast_send_query(br, NULL, query);
1733         spin_unlock(&br->multicast_lock);
1734 }
1735
1736 static void br_ip4_multicast_query_expired(unsigned long data)
1737 {
1738         struct net_bridge *br = (void *)data;
1739
1740         br_multicast_query_expired(br, &br->ip4_query);
1741 }
1742
1743 #if IS_ENABLED(CONFIG_IPV6)
1744 static void br_ip6_multicast_query_expired(unsigned long data)
1745 {
1746         struct net_bridge *br = (void *)data;
1747
1748         br_multicast_query_expired(br, &br->ip6_query);
1749 }
1750 #endif
1751
1752 void br_multicast_init(struct net_bridge *br)
1753 {
1754         br->hash_elasticity = 4;
1755         br->hash_max = 512;
1756
1757         br->multicast_router = 1;
1758         br->multicast_querier = 0;
1759         br->multicast_query_use_ifaddr = 0;
1760         br->multicast_last_member_count = 2;
1761         br->multicast_startup_query_count = 2;
1762
1763         br->multicast_last_member_interval = HZ;
1764         br->multicast_query_response_interval = 10 * HZ;
1765         br->multicast_startup_query_interval = 125 * HZ / 4;
1766         br->multicast_query_interval = 125 * HZ;
1767         br->multicast_querier_interval = 255 * HZ;
1768         br->multicast_membership_interval = 260 * HZ;
1769
1770         br->ip4_querier.delay_time = 0;
1771 #if IS_ENABLED(CONFIG_IPV6)
1772         br->ip6_querier.delay_time = 0;
1773 #endif
1774
1775         spin_lock_init(&br->multicast_lock);
1776         setup_timer(&br->multicast_router_timer,
1777                     br_multicast_local_router_expired, 0);
1778         setup_timer(&br->ip4_querier.timer, br_ip4_multicast_querier_expired,
1779                     (unsigned long)br);
1780         setup_timer(&br->ip4_query.timer, br_ip4_multicast_query_expired,
1781                     (unsigned long)br);
1782 #if IS_ENABLED(CONFIG_IPV6)
1783         setup_timer(&br->ip6_querier.timer, br_ip6_multicast_querier_expired,
1784                     (unsigned long)br);
1785         setup_timer(&br->ip6_query.timer, br_ip6_multicast_query_expired,
1786                     (unsigned long)br);
1787 #endif
1788 }
1789
1790 static void __br_multicast_open(struct net_bridge *br,
1791                                 struct bridge_mcast_query *query)
1792 {
1793         query->startup_sent = 0;
1794
1795         if (br->multicast_disabled)
1796                 return;
1797
1798         mod_timer(&query->timer, jiffies);
1799 }
1800
1801 void br_multicast_open(struct net_bridge *br)
1802 {
1803         __br_multicast_open(br, &br->ip4_query);
1804 #if IS_ENABLED(CONFIG_IPV6)
1805         __br_multicast_open(br, &br->ip6_query);
1806 #endif
1807 }
1808
1809 void br_multicast_stop(struct net_bridge *br)
1810 {
1811         struct net_bridge_mdb_htable *mdb;
1812         struct net_bridge_mdb_entry *mp;
1813         struct hlist_node *n;
1814         u32 ver;
1815         int i;
1816
1817         del_timer_sync(&br->multicast_router_timer);
1818         del_timer_sync(&br->ip4_querier.timer);
1819         del_timer_sync(&br->ip4_query.timer);
1820 #if IS_ENABLED(CONFIG_IPV6)
1821         del_timer_sync(&br->ip6_querier.timer);
1822         del_timer_sync(&br->ip6_query.timer);
1823 #endif
1824
1825         spin_lock_bh(&br->multicast_lock);
1826         mdb = mlock_dereference(br->mdb, br);
1827         if (!mdb)
1828                 goto out;
1829
1830         br->mdb = NULL;
1831
1832         ver = mdb->ver;
1833         for (i = 0; i < mdb->max; i++) {
1834                 hlist_for_each_entry_safe(mp, n, &mdb->mhash[i],
1835                                           hlist[ver]) {
1836                         del_timer(&mp->timer);
1837                         call_rcu_bh(&mp->rcu, br_multicast_free_group);
1838                 }
1839         }
1840
1841         if (mdb->old) {
1842                 spin_unlock_bh(&br->multicast_lock);
1843                 rcu_barrier_bh();
1844                 spin_lock_bh(&br->multicast_lock);
1845                 WARN_ON(mdb->old);
1846         }
1847
1848         mdb->old = mdb;
1849         call_rcu_bh(&mdb->rcu, br_mdb_free);
1850
1851 out:
1852         spin_unlock_bh(&br->multicast_lock);
1853 }
1854
1855 int br_multicast_set_router(struct net_bridge *br, unsigned long val)
1856 {
1857         int err = -ENOENT;
1858
1859         spin_lock_bh(&br->multicast_lock);
1860         if (!netif_running(br->dev))
1861                 goto unlock;
1862
1863         switch (val) {
1864         case 0:
1865         case 2:
1866                 del_timer(&br->multicast_router_timer);
1867                 /* fall through */
1868         case 1:
1869                 br->multicast_router = val;
1870                 err = 0;
1871                 break;
1872
1873         default:
1874                 err = -EINVAL;
1875                 break;
1876         }
1877
1878 unlock:
1879         spin_unlock_bh(&br->multicast_lock);
1880
1881         return err;
1882 }
1883
1884 int br_multicast_set_port_router(struct net_bridge_port *p, unsigned long val)
1885 {
1886         struct net_bridge *br = p->br;
1887         int err = -ENOENT;
1888
1889         spin_lock(&br->multicast_lock);
1890         if (!netif_running(br->dev) || p->state == BR_STATE_DISABLED)
1891                 goto unlock;
1892
1893         switch (val) {
1894         case 0:
1895         case 1:
1896         case 2:
1897                 p->multicast_router = val;
1898                 err = 0;
1899
1900                 if (val < 2 && !hlist_unhashed(&p->rlist))
1901                         hlist_del_init_rcu(&p->rlist);
1902
1903                 if (val == 1)
1904                         break;
1905
1906                 del_timer(&p->multicast_router_timer);
1907
1908                 if (val == 0)
1909                         break;
1910
1911                 br_multicast_add_router(br, p);
1912                 break;
1913
1914         default:
1915                 err = -EINVAL;
1916                 break;
1917         }
1918
1919 unlock:
1920         spin_unlock(&br->multicast_lock);
1921
1922         return err;
1923 }
1924
1925 static void br_multicast_start_querier(struct net_bridge *br,
1926                                        struct bridge_mcast_query *query)
1927 {
1928         struct net_bridge_port *port;
1929
1930         __br_multicast_open(br, query);
1931
1932         list_for_each_entry(port, &br->port_list, list) {
1933                 if (port->state == BR_STATE_DISABLED ||
1934                     port->state == BR_STATE_BLOCKING)
1935                         continue;
1936
1937                 if (query == &br->ip4_query)
1938                         br_multicast_enable(&port->ip4_query);
1939 #if IS_ENABLED(CONFIG_IPV6)
1940                 else
1941                         br_multicast_enable(&port->ip6_query);
1942 #endif
1943         }
1944 }
1945
1946 int br_multicast_toggle(struct net_bridge *br, unsigned long val)
1947 {
1948         int err = 0;
1949         struct net_bridge_mdb_htable *mdb;
1950
1951         spin_lock_bh(&br->multicast_lock);
1952         if (br->multicast_disabled == !val)
1953                 goto unlock;
1954
1955         br->multicast_disabled = !val;
1956         if (br->multicast_disabled)
1957                 goto unlock;
1958
1959         if (!netif_running(br->dev))
1960                 goto unlock;
1961
1962         mdb = mlock_dereference(br->mdb, br);
1963         if (mdb) {
1964                 if (mdb->old) {
1965                         err = -EEXIST;
1966 rollback:
1967                         br->multicast_disabled = !!val;
1968                         goto unlock;
1969                 }
1970
1971                 err = br_mdb_rehash(&br->mdb, mdb->max,
1972                                     br->hash_elasticity);
1973                 if (err)
1974                         goto rollback;
1975         }
1976
1977         br_multicast_start_querier(br, &br->ip4_query);
1978 #if IS_ENABLED(CONFIG_IPV6)
1979         br_multicast_start_querier(br, &br->ip6_query);
1980 #endif
1981
1982 unlock:
1983         spin_unlock_bh(&br->multicast_lock);
1984
1985         return err;
1986 }
1987
1988 int br_multicast_set_querier(struct net_bridge *br, unsigned long val)
1989 {
1990         unsigned long max_delay;
1991
1992         val = !!val;
1993
1994         spin_lock_bh(&br->multicast_lock);
1995         if (br->multicast_querier == val)
1996                 goto unlock;
1997
1998         br->multicast_querier = val;
1999         if (!val)
2000                 goto unlock;
2001
2002         max_delay = br->multicast_query_response_interval;
2003
2004         if (!timer_pending(&br->ip4_querier.timer))
2005                 br->ip4_querier.delay_time = jiffies + max_delay;
2006
2007         br_multicast_start_querier(br, &br->ip4_query);
2008
2009 #if IS_ENABLED(CONFIG_IPV6)
2010         if (!timer_pending(&br->ip6_querier.timer))
2011                 br->ip6_querier.delay_time = jiffies + max_delay;
2012
2013         br_multicast_start_querier(br, &br->ip6_query);
2014 #endif
2015
2016 unlock:
2017         spin_unlock_bh(&br->multicast_lock);
2018
2019         return 0;
2020 }
2021
2022 int br_multicast_set_hash_max(struct net_bridge *br, unsigned long val)
2023 {
2024         int err = -ENOENT;
2025         u32 old;
2026         struct net_bridge_mdb_htable *mdb;
2027
2028         spin_lock_bh(&br->multicast_lock);
2029         if (!netif_running(br->dev))
2030                 goto unlock;
2031
2032         err = -EINVAL;
2033         if (!is_power_of_2(val))
2034                 goto unlock;
2035
2036         mdb = mlock_dereference(br->mdb, br);
2037         if (mdb && val < mdb->size)
2038                 goto unlock;
2039
2040         err = 0;
2041
2042         old = br->hash_max;
2043         br->hash_max = val;
2044
2045         if (mdb) {
2046                 if (mdb->old) {
2047                         err = -EEXIST;
2048 rollback:
2049                         br->hash_max = old;
2050                         goto unlock;
2051                 }
2052
2053                 err = br_mdb_rehash(&br->mdb, br->hash_max,
2054                                     br->hash_elasticity);
2055                 if (err)
2056                         goto rollback;
2057         }
2058
2059 unlock:
2060         spin_unlock_bh(&br->multicast_lock);
2061
2062         return err;
2063 }