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 && mp->timer_armed &&
 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, 6);
 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, 6);
 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                 /* If we have an existing entry, update it's expire timer */
 615                 mod_timer(&mp->timer,
 616                           jiffies + br->multicast_membership_interval);
 617                 goto out;
 618         }
 619
 620         mp = kzalloc(sizeof(*mp), GFP_ATOMIC);
 621         if (unlikely(!mp))
 622                 return ERR_PTR(-ENOMEM);
 623
 624         mp->br = br;
 625         mp->addr = *group;
 626
 627         setup_timer(&mp->timer, br_multicast_group_expired,
 628                     (unsigned long)mp);
 629
 630         hlist_add_head_rcu(&mp->hlist[mdb->ver], &mdb->mhash[hash]);
 631         mdb->size++;
 632
 633 out:
 634         return mp;
 635 }
 636
 637 struct net_bridge_port_group *br_multicast_new_port_group(
 638                         struct net_bridge_port *port,
 639                         struct br_ip *group,
 640                         struct net_bridge_port_group __rcu *next,
 641                         unsigned char state)
 642 {
 643         struct net_bridge_port_group *p;
 644
 645         p = kzalloc(sizeof(*p), GFP_ATOMIC);
 646         if (unlikely(!p))
 647                 return NULL;
 648
 649         p->addr = *group;
 650         p->port = port;
 651         p->state = state;
 652         rcu_assign_pointer(p->next, next);
 653         hlist_add_head(&p->mglist, &port->mglist);
 654         setup_timer(&p->timer, br_multicast_port_group_expired,
 655                     (unsigned long)p);
 656         return p;
 657 }
 658
 659 static int br_multicast_add_group(struct net_bridge *br,
 660                                   struct net_bridge_port *port,
 661                                   struct br_ip *group)
 662 {
 663         struct net_bridge_mdb_entry *mp;
 664         struct net_bridge_port_group *p;
 665         struct net_bridge_port_group __rcu **pp;
 666         int err;
 667
 668         spin_lock(&br->multicast_lock);
 669         if (!netif_running(br->dev) ||
 670             (port && port->state == BR_STATE_DISABLED))
 671                 goto out;
 672
 673         mp = br_multicast_new_group(br, port, group);
 674         err = PTR_ERR(mp);
 675         if (IS_ERR(mp))
 676                 goto err;
 677
 678         if (!port) {
 679                 mp->mglist = true;
 680                 goto out;
 681         }
 682
 683         for (pp = &mp->ports;
 684              (p = mlock_dereference(*pp, br)) != NULL;
 685              pp = &p->next) {
 686                 if (p->port == port) {
 687                         /* We already have a portgroup, update the timer.  */
 688                         mod_timer(&p->timer,
 689                                   jiffies + br->multicast_membership_interval);
 690                         goto out;
 691                 }
 692                 if ((unsigned long)p->port < (unsigned long)port)
 693                         break;
 694         }
 695
 696         p = br_multicast_new_port_group(port, group, *pp, MDB_TEMPORARY);
 697         if (unlikely(!p))
 698                 goto err;
 699         rcu_assign_pointer(*pp, p);
 700         br_mdb_notify(br->dev, port, group, RTM_NEWMDB);
 701
 702 out:
 703         err = 0;
 704
 705 err:
 706         spin_unlock(&br->multicast_lock);
 707         return err;
 708 }
 709
 710 static int br_ip4_multicast_add_group(struct net_bridge *br,
 711                                       struct net_bridge_port *port,
 712                                       __be32 group,
 713                                       __u16 vid)
 714 {
 715         struct br_ip br_group;
 716
 717         if (ipv4_is_local_multicast(group))
 718                 return 0;
 719
 720         br_group.u.ip4 = group;
 721         br_group.proto = htons(ETH_P_IP);
 722         br_group.vid = vid;
 723
 724         return br_multicast_add_group(br, port, &br_group);
 725 }
 726
 727 #if IS_ENABLED(CONFIG_IPV6)
 728 static int br_ip6_multicast_add_group(struct net_bridge *br,
 729                                       struct net_bridge_port *port,
 730                                       const struct in6_addr *group,
 731                                       __u16 vid)
 732 {
 733         struct br_ip br_group;
 734
 735         if (ipv6_addr_is_ll_all_nodes(group))
 736                 return 0;
 737
 738         br_group.u.ip6 = *group;
 739         br_group.proto = htons(ETH_P_IPV6);
 740         br_group.vid = vid;
 741
 742         return br_multicast_add_group(br, port, &br_group);
 743 }
 744 #endif
 745
 746 static void br_multicast_router_expired(unsigned long data)
 747 {
 748         struct net_bridge_port *port = (void *)data;
 749         struct net_bridge *br = port->br;
 750
 751         spin_lock(&br->multicast_lock);
 752         if (port->multicast_router != 1 ||
 753             timer_pending(&port->multicast_router_timer) ||
 754             hlist_unhashed(&port->rlist))
 755                 goto out;
 756
 757         hlist_del_init_rcu(&port->rlist);
 758
 759 out:
 760         spin_unlock(&br->multicast_lock);
 761 }
 762
 763 static void br_multicast_local_router_expired(unsigned long data)
 764 {
 765 }
 766
 767 static void br_multicast_querier_expired(struct net_bridge *br,
 768                                          struct bridge_mcast_query *query)
 769 {
 770         spin_lock(&br->multicast_lock);
 771         if (!netif_running(br->dev) || br->multicast_disabled)
 772                 goto out;
 773
 774         br_multicast_start_querier(br, query);
 775
 776 out:
 777         spin_unlock(&br->multicast_lock);
 778 }
 779
 780 static void br_ip4_multicast_querier_expired(unsigned long data)
 781 {
 782         struct net_bridge *br = (void *)data;
 783
 784         br_multicast_querier_expired(br, &br->ip4_query);
 785 }
 786
 787 #if IS_ENABLED(CONFIG_IPV6)
 788 static void br_ip6_multicast_querier_expired(unsigned long data)
 789 {
 790         struct net_bridge *br = (void *)data;
 791
 792         br_multicast_querier_expired(br, &br->ip6_query);
 793 }
 794 #endif
 795
 796 static void __br_multicast_send_query(struct net_bridge *br,
 797                                       struct net_bridge_port *port,
 798                                       struct br_ip *ip)
 799 {
 800         struct sk_buff *skb;
 801
 802         skb = br_multicast_alloc_query(br, ip);
 803         if (!skb)
 804                 return;
 805
 806         if (port) {
 807                 __skb_push(skb, sizeof(struct ethhdr));
 808                 skb->dev = port->dev;
 809                 NF_HOOK(NFPROTO_BRIDGE, NF_BR_LOCAL_OUT, skb, NULL, skb->dev,
 810                         dev_queue_xmit);
 811         } else
 812                 netif_rx(skb);
 813 }
 814
 815 static void br_multicast_send_query(struct net_bridge *br,
 816                                     struct net_bridge_port *port,
 817                                     struct bridge_mcast_query *query)
 818 {
 819         unsigned long time;
 820         struct br_ip br_group;
 821         struct bridge_mcast_querier *querier = NULL;
 822
 823         if (!netif_running(br->dev) || br->multicast_disabled ||
 824             !br->multicast_querier)
 825                 return;
 826
 827         memset(&br_group.u, 0, sizeof(br_group.u));
 828
 829         if (port ? (query == &port->ip4_query) :
 830                    (query == &br->ip4_query)) {
 831                 querier = &br->ip4_querier;
 832                 br_group.proto = htons(ETH_P_IP);
 833 #if IS_ENABLED(CONFIG_IPV6)
 834         } else {
 835                 querier = &br->ip6_querier;
 836                 br_group.proto = htons(ETH_P_IPV6);
 837 #endif
 838         }
 839
 840         if (!querier || timer_pending(&querier->timer))
 841                 return;
 842
 843         __br_multicast_send_query(br, port, &br_group);
 844
 845         time = jiffies;
 846         time += query->startup_sent < br->multicast_startup_query_count ?
 847                 br->multicast_startup_query_interval :
 848                 br->multicast_query_interval;
 849         mod_timer(&query->timer, time);
 850 }
 851
 852 static void br_multicast_port_query_expired(struct net_bridge_port *port,
 853                                             struct bridge_mcast_query *query)
 854 {
 855         struct net_bridge *br = port->br;
 856
 857         spin_lock(&br->multicast_lock);
 858         if (port->state == BR_STATE_DISABLED ||
 859             port->state == BR_STATE_BLOCKING)
 860                 goto out;
 861
 862         if (query->startup_sent < br->multicast_startup_query_count)
 863                 query->startup_sent++;
 864
 865         br_multicast_send_query(port->br, port, query);
 866
 867 out:
 868         spin_unlock(&br->multicast_lock);
 869 }
 870
 871 static void br_ip4_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->ip4_query);
 876 }
 877
 878 #if IS_ENABLED(CONFIG_IPV6)
 879 static void br_ip6_multicast_port_query_expired(unsigned long data)
 880 {
 881         struct net_bridge_port *port = (void *)data;
 882
 883         br_multicast_port_query_expired(port, &port->ip6_query);
 884 }
 885 #endif
 886
 887 void br_multicast_add_port(struct net_bridge_port *port)
 888 {
 889         port->multicast_router = 1;
 890
 891         setup_timer(&port->multicast_router_timer, br_multicast_router_expired,
 892                     (unsigned long)port);
 893         setup_timer(&port->ip4_query.timer, br_ip4_multicast_port_query_expired,
 894                     (unsigned long)port);
 895 #if IS_ENABLED(CONFIG_IPV6)
 896         setup_timer(&port->ip6_query.timer, br_ip6_multicast_port_query_expired,
 897                     (unsigned long)port);
 898 #endif
 899 }
 900
 901 void br_multicast_del_port(struct net_bridge_port *port)
 902 {
 903         del_timer_sync(&port->multicast_router_timer);
 904 }
 905
 906 static void br_multicast_enable(struct bridge_mcast_query *query)
 907 {
 908         query->startup_sent = 0;
 909
 910         if (try_to_del_timer_sync(&query->timer) >= 0 ||
 911             del_timer(&query->timer))
 912                 mod_timer(&query->timer, jiffies);
 913 }
 914
 915 void br_multicast_enable_port(struct net_bridge_port *port)
 916 {
 917         struct net_bridge *br = port->br;
 918
 919         spin_lock(&br->multicast_lock);
 920         if (br->multicast_disabled || !netif_running(br->dev))
 921                 goto out;
 922
 923         br_multicast_enable(&port->ip4_query);
 924 #if IS_ENABLED(CONFIG_IPV6)
 925         br_multicast_enable(&port->ip6_query);
 926 #endif
 927
 928 out:
 929         spin_unlock(&br->multicast_lock);
 930 }
 931
 932 void br_multicast_disable_port(struct net_bridge_port *port)
 933 {
 934         struct net_bridge *br = port->br;
 935         struct net_bridge_port_group *pg;
 936         struct hlist_node *n;
 937
 938         spin_lock(&br->multicast_lock);
 939         hlist_for_each_entry_safe(pg, n, &port->mglist, mglist)
 940                 br_multicast_del_pg(br, pg);
 941
 942         if (!hlist_unhashed(&port->rlist))
 943                 hlist_del_init_rcu(&port->rlist);
 944         del_timer(&port->multicast_router_timer);
 945         del_timer(&port->ip4_query.timer);
 946 #if IS_ENABLED(CONFIG_IPV6)
 947         del_timer(&port->ip6_query.timer);
 948 #endif
 949         spin_unlock(&br->multicast_lock);
 950 }
 951
 952 static int br_ip4_multicast_igmp3_report(struct net_bridge *br,
 953                                          struct net_bridge_port *port,
 954                                          struct sk_buff *skb)
 955 {
 956         struct igmpv3_report *ih;
 957         struct igmpv3_grec *grec;
 958         int i;
 959         int len;
 960         int num;
 961         int type;
 962         int err = 0;
 963         __be32 group;
 964         u16 vid = 0;
 965
 966         if (!pskb_may_pull(skb, sizeof(*ih)))
 967                 return -EINVAL;
 968
 969         br_vlan_get_tag(skb, &vid);
 970         ih = igmpv3_report_hdr(skb);
 971         num = ntohs(ih->ngrec);
 972         len = sizeof(*ih);
 973
 974         for (i = 0; i < num; i++) {
 975                 len += sizeof(*grec);
 976                 if (!pskb_may_pull(skb, len))
 977                         return -EINVAL;
 978
 979                 grec = (void *)(skb->data + len - sizeof(*grec));
 980                 group = grec->grec_mca;
 981                 type = grec->grec_type;
 982
 983                 len += ntohs(grec->grec_nsrcs) * 4;
 984                 if (!pskb_may_pull(skb, len))
 985                         return -EINVAL;
 986
 987                 /* We treat this as an IGMPv2 report for now. */
 988                 switch (type) {
 989                 case IGMPV3_MODE_IS_INCLUDE:
 990                 case IGMPV3_MODE_IS_EXCLUDE:
 991                 case IGMPV3_CHANGE_TO_INCLUDE:
 992                 case IGMPV3_CHANGE_TO_EXCLUDE:
 993                 case IGMPV3_ALLOW_NEW_SOURCES:
 994                 case IGMPV3_BLOCK_OLD_SOURCES:
 995                         break;
 996
 997                 default:
 998                         continue;
 999                 }
1000
1001                 err = br_ip4_multicast_add_group(br, port, group, vid);
1002                 if (err)
1003                         break;
1004         }
1005
1006         return err;
1007 }
1008
1009 #if IS_ENABLED(CONFIG_IPV6)
1010 static int br_ip6_multicast_mld2_report(struct net_bridge *br,
1011                                         struct net_bridge_port *port,
1012                                         struct sk_buff *skb)
1013 {
1014         struct icmp6hdr *icmp6h;
1015         struct mld2_grec *grec;
1016         int i;
1017         int len;
1018         int num;
1019         int err = 0;
1020         u16 vid = 0;
1021
1022         if (!pskb_may_pull(skb, sizeof(*icmp6h)))
1023                 return -EINVAL;
1024
1025         br_vlan_get_tag(skb, &vid);
1026         icmp6h = icmp6_hdr(skb);
1027         num = ntohs(icmp6h->icmp6_dataun.un_data16[1]);
1028         len = sizeof(*icmp6h);
1029
1030         for (i = 0; i < num; i++) {
1031                 __be16 *nsrcs, _nsrcs;
1032
1033                 nsrcs = skb_header_pointer(skb,
1034                                            len + offsetof(struct mld2_grec,
1035                                                           grec_nsrcs),
1036                                            sizeof(_nsrcs), &_nsrcs);
1037                 if (!nsrcs)
1038                         return -EINVAL;
1039
1040                 if (!pskb_may_pull(skb,
1041                                    len + sizeof(*grec) +
1042                                    sizeof(struct in6_addr) * ntohs(*nsrcs)))
1043                         return -EINVAL;
1044
1045                 grec = (struct mld2_grec *)(skb->data + len);
1046                 len += sizeof(*grec) +
1047                        sizeof(struct in6_addr) * ntohs(*nsrcs);
1048
1049                 /* We treat these as MLDv1 reports for now. */
1050                 switch (grec->grec_type) {
1051                 case MLD2_MODE_IS_INCLUDE:
1052                 case MLD2_MODE_IS_EXCLUDE:
1053                 case MLD2_CHANGE_TO_INCLUDE:
1054                 case MLD2_CHANGE_TO_EXCLUDE:
1055                 case MLD2_ALLOW_NEW_SOURCES:
1056                 case MLD2_BLOCK_OLD_SOURCES:
1057                         break;
1058
1059                 default:
1060                         continue;
1061                 }
1062
1063                 err = br_ip6_multicast_add_group(br, port, &grec->grec_mca,
1064                                                  vid);
1065                 if (!err)
1066                         break;
1067         }
1068
1069         return err;
1070 }
1071 #endif
1072
1073 static void
1074 br_multicast_update_querier_timer(struct net_bridge *br,
1075                                   struct bridge_mcast_querier *querier,
1076                                   unsigned long max_delay)
1077 {
1078         if (!timer_pending(&querier->timer))
1079                 querier->delay_time = jiffies + max_delay;
1080
1081         mod_timer(&querier->timer, jiffies + br->multicast_querier_interval);
1082 }
1083
1084 /*
1085  * Add port to router_list
1086  *  list is maintained ordered by pointer value
1087  *  and locked by br->multicast_lock and RCU
1088  */
1089 static void br_multicast_add_router(struct net_bridge *br,
1090                                     struct net_bridge_port *port)
1091 {
1092         struct net_bridge_port *p;
1093         struct hlist_node *slot = NULL;
1094
1095         hlist_for_each_entry(p, &br->router_list, rlist) {
1096                 if ((unsigned long) port >= (unsigned long) p)
1097                         break;
1098                 slot = &p->rlist;
1099         }
1100
1101         if (slot)
1102                 hlist_add_after_rcu(slot, &port->rlist);
1103         else
1104                 hlist_add_head_rcu(&port->rlist, &br->router_list);
1105 }
1106
1107 static void br_multicast_mark_router(struct net_bridge *br,
1108                                      struct net_bridge_port *port)
1109 {
1110         unsigned long now = jiffies;
1111
1112         if (!port) {
1113                 if (br->multicast_router == 1)
1114                         mod_timer(&br->multicast_router_timer,
1115                                   now + br->multicast_querier_interval);
1116                 return;
1117         }
1118
1119         if (port->multicast_router != 1)
1120                 return;
1121
1122         if (!hlist_unhashed(&port->rlist))
1123                 goto timer;
1124
1125         br_multicast_add_router(br, port);
1126
1127 timer:
1128         mod_timer(&port->multicast_router_timer,
1129                   now + br->multicast_querier_interval);
1130 }
1131
1132 static void br_multicast_query_received(struct net_bridge *br,
1133                                         struct net_bridge_port *port,
1134                                         struct bridge_mcast_querier *querier,
1135                                         int saddr,
1136                                         unsigned long max_delay)
1137 {
1138         if (saddr)
1139                 br_multicast_update_querier_timer(br, querier, max_delay);
1140         else if (timer_pending(&querier->timer))
1141                 return;
1142
1143         br_multicast_mark_router(br, port);
1144 }
1145
1146 static int br_ip4_multicast_query(struct net_bridge *br,
1147                                   struct net_bridge_port *port,
1148                                   struct sk_buff *skb)
1149 {
1150         const struct iphdr *iph = ip_hdr(skb);
1151         struct igmphdr *ih = igmp_hdr(skb);
1152         struct net_bridge_mdb_entry *mp;
1153         struct igmpv3_query *ih3;
1154         struct net_bridge_port_group *p;
1155         struct net_bridge_port_group __rcu **pp;
1156         unsigned long max_delay;
1157         unsigned long now = jiffies;
1158         __be32 group;
1159         int err = 0;
1160         u16 vid = 0;
1161
1162         spin_lock(&br->multicast_lock);
1163         if (!netif_running(br->dev) ||
1164             (port && port->state == BR_STATE_DISABLED))
1165                 goto out;
1166
1167         group = ih->group;
1168
1169         if (skb->len == sizeof(*ih)) {
1170                 max_delay = ih->code * (HZ / IGMP_TIMER_SCALE);
1171
1172                 if (!max_delay) {
1173                         max_delay = 10 * HZ;
1174                         group = 0;
1175                 }
1176         } else {
1177                 if (!pskb_may_pull(skb, sizeof(struct igmpv3_query))) {
1178                         err = -EINVAL;
1179                         goto out;
1180                 }
1181
1182                 ih3 = igmpv3_query_hdr(skb);
1183                 if (ih3->nsrcs)
1184                         goto out;
1185
1186                 max_delay = ih3->code ?
1187                             IGMPV3_MRC(ih3->code) * (HZ / IGMP_TIMER_SCALE) : 1;
1188         }
1189
1190         br_multicast_query_received(br, port, &br->ip4_querier, !!iph->saddr,
1191                                     max_delay);
1192
1193         if (!group)
1194                 goto out;
1195
1196         br_vlan_get_tag(skb, &vid);
1197         mp = br_mdb_ip4_get(mlock_dereference(br->mdb, br), group, vid);
1198         if (!mp)
1199                 goto out;
1200
1201         mod_timer(&mp->timer, now + br->multicast_membership_interval);
1202         mp->timer_armed = true;
1203
1204         max_delay *= br->multicast_last_member_count;
1205
1206         if (mp->mglist &&
1207             (timer_pending(&mp->timer) ?
1208              time_after(mp->timer.expires, now + max_delay) :
1209              try_to_del_timer_sync(&mp->timer) >= 0))
1210                 mod_timer(&mp->timer, now + max_delay);
1211
1212         for (pp = &mp->ports;
1213              (p = mlock_dereference(*pp, br)) != NULL;
1214              pp = &p->next) {
1215                 if (timer_pending(&p->timer) ?
1216                     time_after(p->timer.expires, now + max_delay) :
1217                     try_to_del_timer_sync(&p->timer) >= 0)
1218                         mod_timer(&p->timer, now + max_delay);
1219         }
1220
1221 out:
1222         spin_unlock(&br->multicast_lock);
1223         return err;
1224 }
1225
1226 #if IS_ENABLED(CONFIG_IPV6)
1227 static int br_ip6_multicast_query(struct net_bridge *br,
1228                                   struct net_bridge_port *port,
1229                                   struct sk_buff *skb)
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         int err = 0;
1241         u16 vid = 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         if (skb->len == sizeof(*mld)) {
1249                 if (!pskb_may_pull(skb, sizeof(*mld))) {
1250                         err = -EINVAL;
1251                         goto out;
1252                 }
1253                 mld = (struct mld_msg *) icmp6_hdr(skb);
1254                 max_delay = msecs_to_jiffies(ntohs(mld->mld_maxdelay));
1255                 if (max_delay)
1256                         group = &mld->mld_mca;
1257         } else {
1258                 if (!pskb_may_pull(skb, sizeof(*mld2q))) {
1259                         err = -EINVAL;
1260                         goto out;
1261                 }
1262                 mld2q = (struct mld2_query *)icmp6_hdr(skb);
1263                 if (!mld2q->mld2q_nsrcs)
1264                         group = &mld2q->mld2q_mca;
1265
1266                 max_delay = max(msecs_to_jiffies(mldv2_mrc(mld2q)), 1UL);
1267         }
1268
1269         br_multicast_query_received(br, port, &br->ip6_querier,
1270                                     !ipv6_addr_any(&ip6h->saddr), max_delay);
1271
1272         if (!group)
1273                 goto out;
1274
1275         br_vlan_get_tag(skb, &vid);
1276         mp = br_mdb_ip6_get(mlock_dereference(br->mdb, br), group, vid);
1277         if (!mp)
1278                 goto out;
1279
1280         mod_timer(&mp->timer, now + br->multicast_membership_interval);
1281         mp->timer_armed = true;
1282
1283         max_delay *= br->multicast_last_member_count;
1284         if (mp->mglist &&
1285             (timer_pending(&mp->timer) ?
1286              time_after(mp->timer.expires, now + max_delay) :
1287              try_to_del_timer_sync(&mp->timer) >= 0))
1288                 mod_timer(&mp->timer, now + max_delay);
1289
1290         for (pp = &mp->ports;
1291              (p = mlock_dereference(*pp, br)) != NULL;
1292              pp = &p->next) {
1293                 if (timer_pending(&p->timer) ?
1294                     time_after(p->timer.expires, now + max_delay) :
1295                     try_to_del_timer_sync(&p->timer) >= 0)
1296                         mod_timer(&p->timer, now + max_delay);
1297         }
1298
1299 out:
1300         spin_unlock(&br->multicast_lock);
1301         return err;
1302 }
1303 #endif
1304
1305 static void br_multicast_leave_group(struct net_bridge *br,
1306                                      struct net_bridge_port *port,
1307                                      struct br_ip *group,
1308                                      struct bridge_mcast_querier *querier,
1309                                      struct bridge_mcast_query *query)
1310 {
1311         struct net_bridge_mdb_htable *mdb;
1312         struct net_bridge_mdb_entry *mp;
1313         struct net_bridge_port_group *p;
1314         unsigned long now;
1315         unsigned long time;
1316
1317         spin_lock(&br->multicast_lock);
1318         if (!netif_running(br->dev) ||
1319             (port && port->state == BR_STATE_DISABLED) ||
1320             timer_pending(&querier->timer))
1321                 goto out;
1322
1323         mdb = mlock_dereference(br->mdb, br);
1324         mp = br_mdb_ip_get(mdb, group);
1325         if (!mp)
1326                 goto out;
1327
1328         if (br->multicast_querier) {
1329                 __br_multicast_send_query(br, port, &mp->addr);
1330
1331                 time = jiffies + br->multicast_last_member_count *
1332                                  br->multicast_last_member_interval;
1333
1334                 mod_timer(&query->timer, time);
1335
1336                 for (p = mlock_dereference(mp->ports, br);
1337                      p != NULL;
1338                      p = mlock_dereference(p->next, br)) {
1339                         if (p->port != port)
1340                                 continue;
1341
1342                         if (!hlist_unhashed(&p->mglist) &&
1343                             (timer_pending(&p->timer) ?
1344                              time_after(p->timer.expires, time) :
1345                              try_to_del_timer_sync(&p->timer) >= 0)) {
1346                                 mod_timer(&p->timer, time);
1347                         }
1348
1349                         break;
1350                 }
1351         }
1352
1353         if (port && (port->flags & BR_MULTICAST_FAST_LEAVE)) {
1354                 struct net_bridge_port_group __rcu **pp;
1355
1356                 for (pp = &mp->ports;
1357                      (p = mlock_dereference(*pp, br)) != NULL;
1358                      pp = &p->next) {
1359                         if (p->port != port)
1360                                 continue;
1361
1362                         rcu_assign_pointer(*pp, p->next);
1363                         hlist_del_init(&p->mglist);
1364                         del_timer(&p->timer);
1365                         call_rcu_bh(&p->rcu, br_multicast_free_pg);
1366                         br_mdb_notify(br->dev, port, group, RTM_DELMDB);
1367
1368                         if (!mp->ports && !mp->mglist && mp->timer_armed &&
1369                             netif_running(br->dev))
1370                                 mod_timer(&mp->timer, jiffies);
1371                 }
1372                 goto out;
1373         }
1374
1375         now = jiffies;
1376         time = now + br->multicast_last_member_count *
1377                      br->multicast_last_member_interval;
1378
1379         if (!port) {
1380                 if (mp->mglist && mp->timer_armed &&
1381                     (timer_pending(&mp->timer) ?
1382                      time_after(mp->timer.expires, time) :
1383                      try_to_del_timer_sync(&mp->timer) >= 0)) {
1384                         mod_timer(&mp->timer, time);
1385                 }
1386         }
1387 out:
1388         spin_unlock(&br->multicast_lock);
1389 }
1390
1391 static void br_ip4_multicast_leave_group(struct net_bridge *br,
1392                                          struct net_bridge_port *port,
1393                                          __be32 group,
1394                                          __u16 vid)
1395 {
1396         struct br_ip br_group;
1397         struct bridge_mcast_query *query = port ? &port->ip4_query :
1398                                                   &br->ip4_query;
1399
1400         if (ipv4_is_local_multicast(group))
1401                 return;
1402
1403         br_group.u.ip4 = group;
1404         br_group.proto = htons(ETH_P_IP);
1405         br_group.vid = vid;
1406
1407         br_multicast_leave_group(br, port, &br_group, &br->ip4_querier, query);
1408 }
1409
1410 #if IS_ENABLED(CONFIG_IPV6)
1411 static void br_ip6_multicast_leave_group(struct net_bridge *br,
1412                                          struct net_bridge_port *port,
1413                                          const struct in6_addr *group,
1414                                          __u16 vid)
1415 {
1416         struct br_ip br_group;
1417         struct bridge_mcast_query *query = port ? &port->ip6_query :
1418                                                   &br->ip6_query;
1419
1420
1421         if (ipv6_addr_is_ll_all_nodes(group))
1422                 return;
1423
1424         br_group.u.ip6 = *group;
1425         br_group.proto = htons(ETH_P_IPV6);
1426         br_group.vid = vid;
1427
1428         br_multicast_leave_group(br, port, &br_group, &br->ip6_querier, query);
1429 }
1430 #endif
1431
1432 static int br_multicast_ipv4_rcv(struct net_bridge *br,
1433                                  struct net_bridge_port *port,
1434                                  struct sk_buff *skb)
1435 {
1436         struct sk_buff *skb2 = skb;
1437         const struct iphdr *iph;
1438         struct igmphdr *ih;
1439         unsigned int len;
1440         unsigned int offset;
1441         int err;
1442         u16 vid = 0;
1443
1444         /* We treat OOM as packet loss for now. */
1445         if (!pskb_may_pull(skb, sizeof(*iph)))
1446                 return -EINVAL;
1447
1448         iph = ip_hdr(skb);
1449
1450         if (iph->ihl < 5 || iph->version != 4)
1451                 return -EINVAL;
1452
1453         if (!pskb_may_pull(skb, ip_hdrlen(skb)))
1454                 return -EINVAL;
1455
1456         iph = ip_hdr(skb);
1457
1458         if (unlikely(ip_fast_csum((u8 *)iph, iph->ihl)))
1459                 return -EINVAL;
1460
1461         if (iph->protocol != IPPROTO_IGMP) {
1462                 if (!ipv4_is_local_multicast(iph->daddr))
1463                         BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1464                 return 0;
1465         }
1466
1467         len = ntohs(iph->tot_len);
1468         if (skb->len < len || len < ip_hdrlen(skb))
1469                 return -EINVAL;
1470
1471         if (skb->len > len) {
1472                 skb2 = skb_clone(skb, GFP_ATOMIC);
1473                 if (!skb2)
1474                         return -ENOMEM;
1475
1476                 err = pskb_trim_rcsum(skb2, len);
1477                 if (err)
1478                         goto err_out;
1479         }
1480
1481         len -= ip_hdrlen(skb2);
1482         offset = skb_network_offset(skb2) + ip_hdrlen(skb2);
1483         __skb_pull(skb2, offset);
1484         skb_reset_transport_header(skb2);
1485
1486         err = -EINVAL;
1487         if (!pskb_may_pull(skb2, sizeof(*ih)))
1488                 goto out;
1489
1490         switch (skb2->ip_summed) {
1491         case CHECKSUM_COMPLETE:
1492                 if (!csum_fold(skb2->csum))
1493                         break;
1494                 /* fall through */
1495         case CHECKSUM_NONE:
1496                 skb2->csum = 0;
1497                 if (skb_checksum_complete(skb2))
1498                         goto out;
1499         }
1500
1501         err = 0;
1502
1503         br_vlan_get_tag(skb2, &vid);
1504         BR_INPUT_SKB_CB(skb)->igmp = 1;
1505         ih = igmp_hdr(skb2);
1506
1507         switch (ih->type) {
1508         case IGMP_HOST_MEMBERSHIP_REPORT:
1509         case IGMPV2_HOST_MEMBERSHIP_REPORT:
1510                 BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1511                 err = br_ip4_multicast_add_group(br, port, ih->group, vid);
1512                 break;
1513         case IGMPV3_HOST_MEMBERSHIP_REPORT:
1514                 err = br_ip4_multicast_igmp3_report(br, port, skb2);
1515                 break;
1516         case IGMP_HOST_MEMBERSHIP_QUERY:
1517                 err = br_ip4_multicast_query(br, port, skb2);
1518                 break;
1519         case IGMP_HOST_LEAVE_MESSAGE:
1520                 br_ip4_multicast_leave_group(br, port, ih->group, vid);
1521                 break;
1522         }
1523
1524 out:
1525         __skb_push(skb2, offset);
1526 err_out:
1527         if (skb2 != skb)
1528                 kfree_skb(skb2);
1529         return err;
1530 }
1531
1532 #if IS_ENABLED(CONFIG_IPV6)
1533 static int br_multicast_ipv6_rcv(struct net_bridge *br,
1534                                  struct net_bridge_port *port,
1535                                  struct sk_buff *skb)
1536 {
1537         struct sk_buff *skb2;
1538         const struct ipv6hdr *ip6h;
1539         u8 icmp6_type;
1540         u8 nexthdr;
1541         __be16 frag_off;
1542         unsigned int len;
1543         int offset;
1544         int err;
1545         u16 vid = 0;
1546
1547         if (!pskb_may_pull(skb, sizeof(*ip6h)))
1548                 return -EINVAL;
1549
1550         ip6h = ipv6_hdr(skb);
1551
1552         /*
1553          * We're interested in MLD messages only.
1554          *  - Version is 6
1555          *  - MLD has always Router Alert hop-by-hop option
1556          *  - But we do not support jumbrograms.
1557          */
1558         if (ip6h->version != 6)
1559                 return 0;
1560
1561         /* Prevent flooding this packet if there is no listener present */
1562         if (!ipv6_addr_is_ll_all_nodes(&ip6h->daddr))
1563                 BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1564
1565         if (ip6h->nexthdr != IPPROTO_HOPOPTS ||
1566             ip6h->payload_len == 0)
1567                 return 0;
1568
1569         len = ntohs(ip6h->payload_len) + sizeof(*ip6h);
1570         if (skb->len < len)
1571                 return -EINVAL;
1572
1573         nexthdr = ip6h->nexthdr;
1574         offset = ipv6_skip_exthdr(skb, sizeof(*ip6h), &nexthdr, &frag_off);
1575
1576         if (offset < 0 || nexthdr != IPPROTO_ICMPV6)
1577                 return 0;
1578
1579         /* Okay, we found ICMPv6 header */
1580         skb2 = skb_clone(skb, GFP_ATOMIC);
1581         if (!skb2)
1582                 return -ENOMEM;
1583
1584         err = -EINVAL;
1585         if (!pskb_may_pull(skb2, offset + sizeof(struct icmp6hdr)))
1586                 goto out;
1587
1588         len -= offset - skb_network_offset(skb2);
1589
1590         __skb_pull(skb2, offset);
1591         skb_reset_transport_header(skb2);
1592         skb_postpull_rcsum(skb2, skb_network_header(skb2),
1593                            skb_network_header_len(skb2));
1594
1595         icmp6_type = icmp6_hdr(skb2)->icmp6_type;
1596
1597         switch (icmp6_type) {
1598         case ICMPV6_MGM_QUERY:
1599         case ICMPV6_MGM_REPORT:
1600         case ICMPV6_MGM_REDUCTION:
1601         case ICMPV6_MLD2_REPORT:
1602                 break;
1603         default:
1604                 err = 0;
1605                 goto out;
1606         }
1607
1608         /* Okay, we found MLD message. Check further. */
1609         if (skb2->len > len) {
1610                 err = pskb_trim_rcsum(skb2, len);
1611                 if (err)
1612                         goto out;
1613                 err = -EINVAL;
1614         }
1615
1616         ip6h = ipv6_hdr(skb2);
1617
1618         switch (skb2->ip_summed) {
1619         case CHECKSUM_COMPLETE:
1620                 if (!csum_ipv6_magic(&ip6h->saddr, &ip6h->daddr, skb2->len,
1621                                         IPPROTO_ICMPV6, skb2->csum))
1622                         break;
1623                 /*FALLTHROUGH*/
1624         case CHECKSUM_NONE:
1625                 skb2->csum = ~csum_unfold(csum_ipv6_magic(&ip6h->saddr,
1626                                                         &ip6h->daddr,
1627                                                         skb2->len,
1628                                                         IPPROTO_ICMPV6, 0));
1629                 if (__skb_checksum_complete(skb2))
1630                         goto out;
1631         }
1632
1633         err = 0;
1634
1635         br_vlan_get_tag(skb, &vid);
1636         BR_INPUT_SKB_CB(skb)->igmp = 1;
1637
1638         switch (icmp6_type) {
1639         case ICMPV6_MGM_REPORT:
1640             {
1641                 struct mld_msg *mld;
1642                 if (!pskb_may_pull(skb2, sizeof(*mld))) {
1643                         err = -EINVAL;
1644                         goto out;
1645                 }
1646                 mld = (struct mld_msg *)skb_transport_header(skb2);
1647                 BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1648                 err = br_ip6_multicast_add_group(br, port, &mld->mld_mca, vid);
1649                 break;
1650             }
1651         case ICMPV6_MLD2_REPORT:
1652                 err = br_ip6_multicast_mld2_report(br, port, skb2);
1653                 break;
1654         case ICMPV6_MGM_QUERY:
1655                 err = br_ip6_multicast_query(br, port, skb2);
1656                 break;
1657         case ICMPV6_MGM_REDUCTION:
1658             {
1659                 struct mld_msg *mld;
1660                 if (!pskb_may_pull(skb2, sizeof(*mld))) {
1661                         err = -EINVAL;
1662                         goto out;
1663                 }
1664                 mld = (struct mld_msg *)skb_transport_header(skb2);
1665                 br_ip6_multicast_leave_group(br, port, &mld->mld_mca, vid);
1666             }
1667         }
1668
1669 out:
1670         kfree_skb(skb2);
1671         return err;
1672 }
1673 #endif
1674
1675 int br_multicast_rcv(struct net_bridge *br, struct net_bridge_port *port,
1676                      struct sk_buff *skb)
1677 {
1678         BR_INPUT_SKB_CB(skb)->igmp = 0;
1679         BR_INPUT_SKB_CB(skb)->mrouters_only = 0;
1680
1681         if (br->multicast_disabled)
1682                 return 0;
1683
1684         switch (skb->protocol) {
1685         case htons(ETH_P_IP):
1686                 return br_multicast_ipv4_rcv(br, port, skb);
1687 #if IS_ENABLED(CONFIG_IPV6)
1688         case htons(ETH_P_IPV6):
1689                 return br_multicast_ipv6_rcv(br, port, skb);
1690 #endif
1691         }
1692
1693         return 0;
1694 }
1695
1696 static void br_multicast_query_expired(struct net_bridge *br,
1697                                        struct bridge_mcast_query *query)
1698 {
1699         spin_lock(&br->multicast_lock);
1700         if (query->startup_sent < br->multicast_startup_query_count)
1701                 query->startup_sent++;
1702
1703         br_multicast_send_query(br, NULL, query);
1704         spin_unlock(&br->multicast_lock);
1705 }
1706
1707 static void br_ip4_multicast_query_expired(unsigned long data)
1708 {
1709         struct net_bridge *br = (void *)data;
1710
1711         br_multicast_query_expired(br, &br->ip4_query);
1712 }
1713
1714 #if IS_ENABLED(CONFIG_IPV6)
1715 static void br_ip6_multicast_query_expired(unsigned long data)
1716 {
1717         struct net_bridge *br = (void *)data;
1718
1719         br_multicast_query_expired(br, &br->ip6_query);
1720 }
1721 #endif
1722
1723 void br_multicast_init(struct net_bridge *br)
1724 {
1725         br->hash_elasticity = 4;
1726         br->hash_max = 512;
1727
1728         br->multicast_router = 1;
1729         br->multicast_querier = 0;
1730         br->multicast_query_use_ifaddr = 0;
1731         br->multicast_last_member_count = 2;
1732         br->multicast_startup_query_count = 2;
1733
1734         br->multicast_last_member_interval = HZ;
1735         br->multicast_query_response_interval = 10 * HZ;
1736         br->multicast_startup_query_interval = 125 * HZ / 4;
1737         br->multicast_query_interval = 125 * HZ;
1738         br->multicast_querier_interval = 255 * HZ;
1739         br->multicast_membership_interval = 260 * HZ;
1740
1741         br->ip4_querier.delay_time = 0;
1742 #if IS_ENABLED(CONFIG_IPV6)
1743         br->ip6_querier.delay_time = 0;
1744 #endif
1745
1746         spin_lock_init(&br->multicast_lock);
1747         setup_timer(&br->multicast_router_timer,
1748                     br_multicast_local_router_expired, 0);
1749         setup_timer(&br->ip4_querier.timer, br_ip4_multicast_querier_expired,
1750                     (unsigned long)br);
1751         setup_timer(&br->ip4_query.timer, br_ip4_multicast_query_expired,
1752                     (unsigned long)br);
1753 #if IS_ENABLED(CONFIG_IPV6)
1754         setup_timer(&br->ip6_querier.timer, br_ip6_multicast_querier_expired,
1755                     (unsigned long)br);
1756         setup_timer(&br->ip6_query.timer, br_ip6_multicast_query_expired,
1757                     (unsigned long)br);
1758 #endif
1759 }
1760
1761 static void __br_multicast_open(struct net_bridge *br,
1762                                 struct bridge_mcast_query *query)
1763 {
1764         query->startup_sent = 0;
1765
1766         if (br->multicast_disabled)
1767                 return;
1768
1769         mod_timer(&query->timer, jiffies);
1770 }
1771
1772 void br_multicast_open(struct net_bridge *br)
1773 {
1774         __br_multicast_open(br, &br->ip4_query);
1775 #if IS_ENABLED(CONFIG_IPV6)
1776         __br_multicast_open(br, &br->ip6_query);
1777 #endif
1778 }
1779
1780 void br_multicast_stop(struct net_bridge *br)
1781 {
1782         struct net_bridge_mdb_htable *mdb;
1783         struct net_bridge_mdb_entry *mp;
1784         struct hlist_node *n;
1785         u32 ver;
1786         int i;
1787
1788         del_timer_sync(&br->multicast_router_timer);
1789         del_timer_sync(&br->ip4_querier.timer);
1790         del_timer_sync(&br->ip4_query.timer);
1791 #if IS_ENABLED(CONFIG_IPV6)
1792         del_timer_sync(&br->ip6_querier.timer);
1793         del_timer_sync(&br->ip6_query.timer);
1794 #endif
1795
1796         spin_lock_bh(&br->multicast_lock);
1797         mdb = mlock_dereference(br->mdb, br);
1798         if (!mdb)
1799                 goto out;
1800
1801         br->mdb = NULL;
1802
1803         ver = mdb->ver;
1804         for (i = 0; i < mdb->max; i++) {
1805                 hlist_for_each_entry_safe(mp, n, &mdb->mhash[i],
1806                                           hlist[ver]) {
1807                         del_timer(&mp->timer);
1808                         mp->timer_armed = false;
1809                         call_rcu_bh(&mp->rcu, br_multicast_free_group);
1810                 }
1811         }
1812
1813         if (mdb->old) {
1814                 spin_unlock_bh(&br->multicast_lock);
1815                 rcu_barrier_bh();
1816                 spin_lock_bh(&br->multicast_lock);
1817                 WARN_ON(mdb->old);
1818         }
1819
1820         mdb->old = mdb;
1821         call_rcu_bh(&mdb->rcu, br_mdb_free);
1822
1823 out:
1824         spin_unlock_bh(&br->multicast_lock);
1825 }
1826
1827 int br_multicast_set_router(struct net_bridge *br, unsigned long val)
1828 {
1829         int err = -ENOENT;
1830
1831         spin_lock_bh(&br->multicast_lock);
1832         if (!netif_running(br->dev))
1833                 goto unlock;
1834
1835         switch (val) {
1836         case 0:
1837         case 2:
1838                 del_timer(&br->multicast_router_timer);
1839                 /* fall through */
1840         case 1:
1841                 br->multicast_router = val;
1842                 err = 0;
1843                 break;
1844
1845         default:
1846                 err = -EINVAL;
1847                 break;
1848         }
1849
1850 unlock:
1851         spin_unlock_bh(&br->multicast_lock);
1852
1853         return err;
1854 }
1855
1856 int br_multicast_set_port_router(struct net_bridge_port *p, unsigned long val)
1857 {
1858         struct net_bridge *br = p->br;
1859         int err = -ENOENT;
1860
1861         spin_lock(&br->multicast_lock);
1862         if (!netif_running(br->dev) || p->state == BR_STATE_DISABLED)
1863                 goto unlock;
1864
1865         switch (val) {
1866         case 0:
1867         case 1:
1868         case 2:
1869                 p->multicast_router = val;
1870                 err = 0;
1871
1872                 if (val < 2 && !hlist_unhashed(&p->rlist))
1873                         hlist_del_init_rcu(&p->rlist);
1874
1875                 if (val == 1)
1876                         break;
1877
1878                 del_timer(&p->multicast_router_timer);
1879
1880                 if (val == 0)
1881                         break;
1882
1883                 br_multicast_add_router(br, p);
1884                 break;
1885
1886         default:
1887                 err = -EINVAL;
1888                 break;
1889         }
1890
1891 unlock:
1892         spin_unlock(&br->multicast_lock);
1893
1894         return err;
1895 }
1896
1897 static void br_multicast_start_querier(struct net_bridge *br,
1898                                        struct bridge_mcast_query *query)
1899 {
1900         struct net_bridge_port *port;
1901
1902         __br_multicast_open(br, query);
1903
1904         list_for_each_entry(port, &br->port_list, list) {
1905                 if (port->state == BR_STATE_DISABLED ||
1906                     port->state == BR_STATE_BLOCKING)
1907                         continue;
1908
1909                 if (query == &br->ip4_query)
1910                         br_multicast_enable(&port->ip4_query);
1911 #if IS_ENABLED(CONFIG_IPV6)
1912                 else
1913                         br_multicast_enable(&port->ip6_query);
1914 #endif
1915         }
1916 }
1917
1918 int br_multicast_toggle(struct net_bridge *br, unsigned long val)
1919 {
1920         int err = 0;
1921         struct net_bridge_mdb_htable *mdb;
1922
1923         spin_lock_bh(&br->multicast_lock);
1924         if (br->multicast_disabled == !val)
1925                 goto unlock;
1926
1927         br->multicast_disabled = !val;
1928         if (br->multicast_disabled)
1929                 goto unlock;
1930
1931         if (!netif_running(br->dev))
1932                 goto unlock;
1933
1934         mdb = mlock_dereference(br->mdb, br);
1935         if (mdb) {
1936                 if (mdb->old) {
1937                         err = -EEXIST;
1938 rollback:
1939                         br->multicast_disabled = !!val;
1940                         goto unlock;
1941                 }
1942
1943                 err = br_mdb_rehash(&br->mdb, mdb->max,
1944                                     br->hash_elasticity);
1945                 if (err)
1946                         goto rollback;
1947         }
1948
1949         br_multicast_start_querier(br, &br->ip4_query);
1950 #if IS_ENABLED(CONFIG_IPV6)
1951         br_multicast_start_querier(br, &br->ip6_query);
1952 #endif
1953
1954 unlock:
1955         spin_unlock_bh(&br->multicast_lock);
1956
1957         return err;
1958 }
1959
1960 int br_multicast_set_querier(struct net_bridge *br, unsigned long val)
1961 {
1962         unsigned long max_delay;
1963
1964         val = !!val;
1965
1966         spin_lock_bh(&br->multicast_lock);
1967         if (br->multicast_querier == val)
1968                 goto unlock;
1969
1970         br->multicast_querier = val;
1971         if (!val)
1972                 goto unlock;
1973
1974         max_delay = br->multicast_query_response_interval;
1975
1976         if (!timer_pending(&br->ip4_querier.timer))
1977                 br->ip4_querier.delay_time = jiffies + max_delay;
1978
1979         br_multicast_start_querier(br, &br->ip4_query);
1980
1981 #if IS_ENABLED(CONFIG_IPV6)
1982         if (!timer_pending(&br->ip6_querier.timer))
1983                 br->ip6_querier.delay_time = jiffies + max_delay;
1984
1985         br_multicast_start_querier(br, &br->ip6_query);
1986 #endif
1987
1988 unlock:
1989         spin_unlock_bh(&br->multicast_lock);
1990
1991         return 0;
1992 }
1993
1994 int br_multicast_set_hash_max(struct net_bridge *br, unsigned long val)
1995 {
1996         int err = -ENOENT;
1997         u32 old;
1998         struct net_bridge_mdb_htable *mdb;
1999
2000         spin_lock(&br->multicast_lock);
2001         if (!netif_running(br->dev))
2002                 goto unlock;
2003
2004         err = -EINVAL;
2005         if (!is_power_of_2(val))
2006                 goto unlock;
2007
2008         mdb = mlock_dereference(br->mdb, br);
2009         if (mdb && val < mdb->size)
2010                 goto unlock;
2011
2012         err = 0;
2013
2014         old = br->hash_max;
2015         br->hash_max = val;
2016
2017         if (mdb) {
2018                 if (mdb->old) {
2019                         err = -EEXIST;
2020 rollback:
2021                         br->hash_max = old;
2022                         goto unlock;
2023                 }
2024
2025                 err = br_mdb_rehash(&br->mdb, br->hash_max,
2026                                     br->hash_elasticity);
2027                 if (err)
2028                         goto rollback;
2029         }
2030
2031 unlock:
2032         spin_unlock(&br->multicast_lock);
2033
2034         return err;
2035 }