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                 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
 624         setup_timer(&mp->timer, br_multicast_group_expired,
 625                     (unsigned long)mp);
 626
 627         hlist_add_head_rcu(&mp->hlist[mdb->ver], &mdb->mhash[hash]);
 628         mdb->size++;
 629
 630 out:
 631         return mp;
 632 }
 633
 634 struct net_bridge_port_group *br_multicast_new_port_group(
 635                         struct net_bridge_port *port,
 636                         struct br_ip *group,
 637                         struct net_bridge_port_group __rcu *next,
 638                         unsigned char state)
 639 {
 640         struct net_bridge_port_group *p;
 641
 642         p = kzalloc(sizeof(*p), GFP_ATOMIC);
 643         if (unlikely(!p))
 644                 return NULL;
 645
 646         p->addr = *group;
 647         p->port = port;
 648         p->state = state;
 649         rcu_assign_pointer(p->next, next);
 650         hlist_add_head(&p->mglist, &port->mglist);
 651         setup_timer(&p->timer, br_multicast_port_group_expired,
 652                     (unsigned long)p);
 653         return p;
 654 }
 655
 656 static int br_multicast_add_group(struct net_bridge *br,
 657                                   struct net_bridge_port *port,
 658                                   struct br_ip *group)
 659 {
 660         struct net_bridge_mdb_entry *mp;
 661         struct net_bridge_port_group *p;
 662         struct net_bridge_port_group __rcu **pp;
 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                 goto out;
 678         }
 679
 680         for (pp = &mp->ports;
 681              (p = mlock_dereference(*pp, br)) != NULL;
 682              pp = &p->next) {
 683                 if (p->port == port)
 684                         goto out;
 685                 if ((unsigned long)p->port < (unsigned long)port)
 686                         break;
 687         }
 688
 689         p = br_multicast_new_port_group(port, group, *pp, MDB_TEMPORARY);
 690         if (unlikely(!p))
 691                 goto err;
 692         rcu_assign_pointer(*pp, p);
 693         br_mdb_notify(br->dev, port, group, RTM_NEWMDB);
 694
 695 out:
 696         err = 0;
 697
 698 err:
 699         spin_unlock(&br->multicast_lock);
 700         return err;
 701 }
 702
 703 static int br_ip4_multicast_add_group(struct net_bridge *br,
 704                                       struct net_bridge_port *port,
 705                                       __be32 group,
 706                                       __u16 vid)
 707 {
 708         struct br_ip br_group;
 709
 710         if (ipv4_is_local_multicast(group))
 711                 return 0;
 712
 713         br_group.u.ip4 = group;
 714         br_group.proto = htons(ETH_P_IP);
 715         br_group.vid = vid;
 716
 717         return br_multicast_add_group(br, port, &br_group);
 718 }
 719
 720 #if IS_ENABLED(CONFIG_IPV6)
 721 static int br_ip6_multicast_add_group(struct net_bridge *br,
 722                                       struct net_bridge_port *port,
 723                                       const struct in6_addr *group,
 724                                       __u16 vid)
 725 {
 726         struct br_ip br_group;
 727
 728         if (ipv6_addr_is_ll_all_nodes(group))
 729                 return 0;
 730
 731         br_group.u.ip6 = *group;
 732         br_group.proto = htons(ETH_P_IPV6);
 733         br_group.vid = vid;
 734
 735         return br_multicast_add_group(br, port, &br_group);
 736 }
 737 #endif
 738
 739 static void br_multicast_router_expired(unsigned long data)
 740 {
 741         struct net_bridge_port *port = (void *)data;
 742         struct net_bridge *br = port->br;
 743
 744         spin_lock(&br->multicast_lock);
 745         if (port->multicast_router != 1 ||
 746             timer_pending(&port->multicast_router_timer) ||
 747             hlist_unhashed(&port->rlist))
 748                 goto out;
 749
 750         hlist_del_init_rcu(&port->rlist);
 751
 752 out:
 753         spin_unlock(&br->multicast_lock);
 754 }
 755
 756 static void br_multicast_local_router_expired(unsigned long data)
 757 {
 758 }
 759
 760 static void br_multicast_querier_expired(struct net_bridge *br,
 761                                          struct bridge_mcast_query *query)
 762 {
 763         spin_lock(&br->multicast_lock);
 764         if (!netif_running(br->dev) || br->multicast_disabled)
 765                 goto out;
 766
 767         br_multicast_start_querier(br, query);
 768
 769 out:
 770         spin_unlock(&br->multicast_lock);
 771 }
 772
 773 static void br_ip4_multicast_querier_expired(unsigned long data)
 774 {
 775         struct net_bridge *br = (void *)data;
 776
 777         br_multicast_querier_expired(br, &br->ip4_query);
 778 }
 779
 780 #if IS_ENABLED(CONFIG_IPV6)
 781 static void br_ip6_multicast_querier_expired(unsigned long data)
 782 {
 783         struct net_bridge *br = (void *)data;
 784
 785         br_multicast_querier_expired(br, &br->ip6_query);
 786 }
 787 #endif
 788
 789 static void __br_multicast_send_query(struct net_bridge *br,
 790                                       struct net_bridge_port *port,
 791                                       struct br_ip *ip)
 792 {
 793         struct sk_buff *skb;
 794
 795         skb = br_multicast_alloc_query(br, ip);
 796         if (!skb)
 797                 return;
 798
 799         if (port) {
 800                 __skb_push(skb, sizeof(struct ethhdr));
 801                 skb->dev = port->dev;
 802                 NF_HOOK(NFPROTO_BRIDGE, NF_BR_LOCAL_OUT, skb, NULL, skb->dev,
 803                         dev_queue_xmit);
 804         } else
 805                 netif_rx(skb);
 806 }
 807
 808 static void br_multicast_send_query(struct net_bridge *br,
 809                                     struct net_bridge_port *port,
 810                                     struct bridge_mcast_query *query)
 811 {
 812         unsigned long time;
 813         struct br_ip br_group;
 814         struct bridge_mcast_querier *querier = NULL;
 815
 816         if (!netif_running(br->dev) || br->multicast_disabled ||
 817             !br->multicast_querier)
 818                 return;
 819
 820         memset(&br_group.u, 0, sizeof(br_group.u));
 821
 822         if (port ? (query == &port->ip4_query) :
 823                    (query == &br->ip4_query)) {
 824                 querier = &br->ip4_querier;
 825                 br_group.proto = htons(ETH_P_IP);
 826 #if IS_ENABLED(CONFIG_IPV6)
 827         } else {
 828                 querier = &br->ip6_querier;
 829                 br_group.proto = htons(ETH_P_IPV6);
 830 #endif
 831         }
 832
 833         if (!querier || timer_pending(&querier->timer))
 834                 return;
 835
 836         __br_multicast_send_query(br, port, &br_group);
 837
 838         time = jiffies;
 839         time += query->startup_sent < br->multicast_startup_query_count ?
 840                 br->multicast_startup_query_interval :
 841                 br->multicast_query_interval;
 842         mod_timer(&query->timer, time);
 843 }
 844
 845 static void br_multicast_port_query_expired(struct net_bridge_port *port,
 846                                             struct bridge_mcast_query *query)
 847 {
 848         struct net_bridge *br = port->br;
 849
 850         spin_lock(&br->multicast_lock);
 851         if (port->state == BR_STATE_DISABLED ||
 852             port->state == BR_STATE_BLOCKING)
 853                 goto out;
 854
 855         if (query->startup_sent < br->multicast_startup_query_count)
 856                 query->startup_sent++;
 857
 858         br_multicast_send_query(port->br, port, query);
 859
 860 out:
 861         spin_unlock(&br->multicast_lock);
 862 }
 863
 864 static void br_ip4_multicast_port_query_expired(unsigned long data)
 865 {
 866         struct net_bridge_port *port = (void *)data;
 867
 868         br_multicast_port_query_expired(port, &port->ip4_query);
 869 }
 870
 871 #if IS_ENABLED(CONFIG_IPV6)
 872 static void br_ip6_multicast_port_query_expired(unsigned long data)
 873 {
 874         struct net_bridge_port *port = (void *)data;
 875
 876         br_multicast_port_query_expired(port, &port->ip6_query);
 877 }
 878 #endif
 879
 880 void br_multicast_add_port(struct net_bridge_port *port)
 881 {
 882         port->multicast_router = 1;
 883
 884         setup_timer(&port->multicast_router_timer, br_multicast_router_expired,
 885                     (unsigned long)port);
 886         setup_timer(&port->ip4_query.timer, br_ip4_multicast_port_query_expired,
 887                     (unsigned long)port);
 888 #if IS_ENABLED(CONFIG_IPV6)
 889         setup_timer(&port->ip6_query.timer, br_ip6_multicast_port_query_expired,
 890                     (unsigned long)port);
 891 #endif
 892 }
 893
 894 void br_multicast_del_port(struct net_bridge_port *port)
 895 {
 896         del_timer_sync(&port->multicast_router_timer);
 897 }
 898
 899 static void br_multicast_enable(struct bridge_mcast_query *query)
 900 {
 901         query->startup_sent = 0;
 902
 903         if (try_to_del_timer_sync(&query->timer) >= 0 ||
 904             del_timer(&query->timer))
 905                 mod_timer(&query->timer, jiffies);
 906 }
 907
 908 void br_multicast_enable_port(struct net_bridge_port *port)
 909 {
 910         struct net_bridge *br = port->br;
 911
 912         spin_lock(&br->multicast_lock);
 913         if (br->multicast_disabled || !netif_running(br->dev))
 914                 goto out;
 915
 916         br_multicast_enable(&port->ip4_query);
 917 #if IS_ENABLED(CONFIG_IPV6)
 918         br_multicast_enable(&port->ip6_query);
 919 #endif
 920
 921 out:
 922         spin_unlock(&br->multicast_lock);
 923 }
 924
 925 void br_multicast_disable_port(struct net_bridge_port *port)
 926 {
 927         struct net_bridge *br = port->br;
 928         struct net_bridge_port_group *pg;
 929         struct hlist_node *n;
 930
 931         spin_lock(&br->multicast_lock);
 932         hlist_for_each_entry_safe(pg, n, &port->mglist, mglist)
 933                 br_multicast_del_pg(br, pg);
 934
 935         if (!hlist_unhashed(&port->rlist))
 936                 hlist_del_init_rcu(&port->rlist);
 937         del_timer(&port->multicast_router_timer);
 938         del_timer(&port->ip4_query.timer);
 939 #if IS_ENABLED(CONFIG_IPV6)
 940         del_timer(&port->ip6_query.timer);
 941 #endif
 942         spin_unlock(&br->multicast_lock);
 943 }
 944
 945 static int br_ip4_multicast_igmp3_report(struct net_bridge *br,
 946                                          struct net_bridge_port *port,
 947                                          struct sk_buff *skb)
 948 {
 949         struct igmpv3_report *ih;
 950         struct igmpv3_grec *grec;
 951         int i;
 952         int len;
 953         int num;
 954         int type;
 955         int err = 0;
 956         __be32 group;
 957         u16 vid = 0;
 958
 959         if (!pskb_may_pull(skb, sizeof(*ih)))
 960                 return -EINVAL;
 961
 962         br_vlan_get_tag(skb, &vid);
 963         ih = igmpv3_report_hdr(skb);
 964         num = ntohs(ih->ngrec);
 965         len = sizeof(*ih);
 966
 967         for (i = 0; i < num; i++) {
 968                 len += sizeof(*grec);
 969                 if (!pskb_may_pull(skb, len))
 970                         return -EINVAL;
 971
 972                 grec = (void *)(skb->data + len - sizeof(*grec));
 973                 group = grec->grec_mca;
 974                 type = grec->grec_type;
 975
 976                 len += ntohs(grec->grec_nsrcs) * 4;
 977                 if (!pskb_may_pull(skb, len))
 978                         return -EINVAL;
 979
 980                 /* We treat this as an IGMPv2 report for now. */
 981                 switch (type) {
 982                 case IGMPV3_MODE_IS_INCLUDE:
 983                 case IGMPV3_MODE_IS_EXCLUDE:
 984                 case IGMPV3_CHANGE_TO_INCLUDE:
 985                 case IGMPV3_CHANGE_TO_EXCLUDE:
 986                 case IGMPV3_ALLOW_NEW_SOURCES:
 987                 case IGMPV3_BLOCK_OLD_SOURCES:
 988                         break;
 989
 990                 default:
 991                         continue;
 992                 }
 993
 994                 err = br_ip4_multicast_add_group(br, port, group, vid);
 995                 if (err)
 996                         break;
 997         }
 998
 999         return err;
1000 }
1001
1002 #if IS_ENABLED(CONFIG_IPV6)
1003 static int br_ip6_multicast_mld2_report(struct net_bridge *br,
1004                                         struct net_bridge_port *port,
1005                                         struct sk_buff *skb)
1006 {
1007         struct icmp6hdr *icmp6h;
1008         struct mld2_grec *grec;
1009         int i;
1010         int len;
1011         int num;
1012         int err = 0;
1013         u16 vid = 0;
1014
1015         if (!pskb_may_pull(skb, sizeof(*icmp6h)))
1016                 return -EINVAL;
1017
1018         br_vlan_get_tag(skb, &vid);
1019         icmp6h = icmp6_hdr(skb);
1020         num = ntohs(icmp6h->icmp6_dataun.un_data16[1]);
1021         len = sizeof(*icmp6h);
1022
1023         for (i = 0; i < num; i++) {
1024                 __be16 *nsrcs, _nsrcs;
1025
1026                 nsrcs = skb_header_pointer(skb,
1027                                            len + offsetof(struct mld2_grec,
1028                                                           grec_nsrcs),
1029                                            sizeof(_nsrcs), &_nsrcs);
1030                 if (!nsrcs)
1031                         return -EINVAL;
1032
1033                 if (!pskb_may_pull(skb,
1034                                    len + sizeof(*grec) +
1035                                    sizeof(struct in6_addr) * ntohs(*nsrcs)))
1036                         return -EINVAL;
1037
1038                 grec = (struct mld2_grec *)(skb->data + len);
1039                 len += sizeof(*grec) +
1040                        sizeof(struct in6_addr) * ntohs(*nsrcs);
1041
1042                 /* We treat these as MLDv1 reports for now. */
1043                 switch (grec->grec_type) {
1044                 case MLD2_MODE_IS_INCLUDE:
1045                 case MLD2_MODE_IS_EXCLUDE:
1046                 case MLD2_CHANGE_TO_INCLUDE:
1047                 case MLD2_CHANGE_TO_EXCLUDE:
1048                 case MLD2_ALLOW_NEW_SOURCES:
1049                 case MLD2_BLOCK_OLD_SOURCES:
1050                         break;
1051
1052                 default:
1053                         continue;
1054                 }
1055
1056                 err = br_ip6_multicast_add_group(br, port, &grec->grec_mca,
1057                                                  vid);
1058                 if (!err)
1059                         break;
1060         }
1061
1062         return err;
1063 }
1064 #endif
1065
1066 static void
1067 br_multicast_update_querier_timer(struct net_bridge *br,
1068                                   struct bridge_mcast_querier *querier,
1069                                   unsigned long max_delay)
1070 {
1071         if (!timer_pending(&querier->timer))
1072                 querier->delay_time = jiffies + max_delay;
1073
1074         mod_timer(&querier->timer, jiffies + br->multicast_querier_interval);
1075 }
1076
1077 /*
1078  * Add port to router_list
1079  *  list is maintained ordered by pointer value
1080  *  and locked by br->multicast_lock and RCU
1081  */
1082 static void br_multicast_add_router(struct net_bridge *br,
1083                                     struct net_bridge_port *port)
1084 {
1085         struct net_bridge_port *p;
1086         struct hlist_node *slot = NULL;
1087
1088         hlist_for_each_entry(p, &br->router_list, rlist) {
1089                 if ((unsigned long) port >= (unsigned long) p)
1090                         break;
1091                 slot = &p->rlist;
1092         }
1093
1094         if (slot)
1095                 hlist_add_after_rcu(slot, &port->rlist);
1096         else
1097                 hlist_add_head_rcu(&port->rlist, &br->router_list);
1098 }
1099
1100 static void br_multicast_mark_router(struct net_bridge *br,
1101                                      struct net_bridge_port *port)
1102 {
1103         unsigned long now = jiffies;
1104
1105         if (!port) {
1106                 if (br->multicast_router == 1)
1107                         mod_timer(&br->multicast_router_timer,
1108                                   now + br->multicast_querier_interval);
1109                 return;
1110         }
1111
1112         if (port->multicast_router != 1)
1113                 return;
1114
1115         if (!hlist_unhashed(&port->rlist))
1116                 goto timer;
1117
1118         br_multicast_add_router(br, port);
1119
1120 timer:
1121         mod_timer(&port->multicast_router_timer,
1122                   now + br->multicast_querier_interval);
1123 }
1124
1125 static void br_multicast_query_received(struct net_bridge *br,
1126                                         struct net_bridge_port *port,
1127                                         struct bridge_mcast_querier *querier,
1128                                         int saddr,
1129                                         unsigned long max_delay)
1130 {
1131         if (saddr)
1132                 br_multicast_update_querier_timer(br, querier, max_delay);
1133         else if (timer_pending(&querier->timer))
1134                 return;
1135
1136         br_multicast_mark_router(br, port);
1137 }
1138
1139 static int br_ip4_multicast_query(struct net_bridge *br,
1140                                   struct net_bridge_port *port,
1141                                   struct sk_buff *skb)
1142 {
1143         const struct iphdr *iph = ip_hdr(skb);
1144         struct igmphdr *ih = igmp_hdr(skb);
1145         struct net_bridge_mdb_entry *mp;
1146         struct igmpv3_query *ih3;
1147         struct net_bridge_port_group *p;
1148         struct net_bridge_port_group __rcu **pp;
1149         unsigned long max_delay;
1150         unsigned long now = jiffies;
1151         __be32 group;
1152         int err = 0;
1153         u16 vid = 0;
1154
1155         spin_lock(&br->multicast_lock);
1156         if (!netif_running(br->dev) ||
1157             (port && port->state == BR_STATE_DISABLED))
1158                 goto out;
1159
1160         group = ih->group;
1161
1162         if (skb->len == sizeof(*ih)) {
1163                 max_delay = ih->code * (HZ / IGMP_TIMER_SCALE);
1164
1165                 if (!max_delay) {
1166                         max_delay = 10 * HZ;
1167                         group = 0;
1168                 }
1169         } else {
1170                 if (!pskb_may_pull(skb, sizeof(struct igmpv3_query))) {
1171                         err = -EINVAL;
1172                         goto out;
1173                 }
1174
1175                 ih3 = igmpv3_query_hdr(skb);
1176                 if (ih3->nsrcs)
1177                         goto out;
1178
1179                 max_delay = ih3->code ?
1180                             IGMPV3_MRC(ih3->code) * (HZ / IGMP_TIMER_SCALE) : 1;
1181         }
1182
1183         br_multicast_query_received(br, port, &br->ip4_querier, !!iph->saddr,
1184                                     max_delay);
1185
1186         if (!group)
1187                 goto out;
1188
1189         br_vlan_get_tag(skb, &vid);
1190         mp = br_mdb_ip4_get(mlock_dereference(br->mdb, br), group, vid);
1191         if (!mp)
1192                 goto out;
1193
1194         mod_timer(&mp->timer, now + br->multicast_membership_interval);
1195         mp->timer_armed = true;
1196
1197         max_delay *= br->multicast_last_member_count;
1198
1199         if (mp->mglist &&
1200             (timer_pending(&mp->timer) ?
1201              time_after(mp->timer.expires, now + max_delay) :
1202              try_to_del_timer_sync(&mp->timer) >= 0))
1203                 mod_timer(&mp->timer, now + max_delay);
1204
1205         for (pp = &mp->ports;
1206              (p = mlock_dereference(*pp, br)) != NULL;
1207              pp = &p->next) {
1208                 if (timer_pending(&p->timer) ?
1209                     time_after(p->timer.expires, now + max_delay) :
1210                     try_to_del_timer_sync(&p->timer) >= 0)
1211                         mod_timer(&p->timer, now + max_delay);
1212         }
1213
1214 out:
1215         spin_unlock(&br->multicast_lock);
1216         return err;
1217 }
1218
1219 #if IS_ENABLED(CONFIG_IPV6)
1220 static int br_ip6_multicast_query(struct net_bridge *br,
1221                                   struct net_bridge_port *port,
1222                                   struct sk_buff *skb)
1223 {
1224         const struct ipv6hdr *ip6h = ipv6_hdr(skb);
1225         struct mld_msg *mld;
1226         struct net_bridge_mdb_entry *mp;
1227         struct mld2_query *mld2q;
1228         struct net_bridge_port_group *p;
1229         struct net_bridge_port_group __rcu **pp;
1230         unsigned long max_delay;
1231         unsigned long now = jiffies;
1232         const struct in6_addr *group = NULL;
1233         int err = 0;
1234         u16 vid = 0;
1235
1236         spin_lock(&br->multicast_lock);
1237         if (!netif_running(br->dev) ||
1238             (port && port->state == BR_STATE_DISABLED))
1239                 goto out;
1240
1241         if (skb->len == sizeof(*mld)) {
1242                 if (!pskb_may_pull(skb, sizeof(*mld))) {
1243                         err = -EINVAL;
1244                         goto out;
1245                 }
1246                 mld = (struct mld_msg *) icmp6_hdr(skb);
1247                 max_delay = msecs_to_jiffies(ntohs(mld->mld_maxdelay));
1248                 if (max_delay)
1249                         group = &mld->mld_mca;
1250         } else {
1251                 if (!pskb_may_pull(skb, sizeof(*mld2q))) {
1252                         err = -EINVAL;
1253                         goto out;
1254                 }
1255                 mld2q = (struct mld2_query *)icmp6_hdr(skb);
1256                 if (!mld2q->mld2q_nsrcs)
1257                         group = &mld2q->mld2q_mca;
1258
1259                 max_delay = max(msecs_to_jiffies(mldv2_mrc(mld2q)), 1UL);
1260         }
1261
1262         br_multicast_query_received(br, port, &br->ip6_querier,
1263                                     !ipv6_addr_any(&ip6h->saddr), max_delay);
1264
1265         if (!group)
1266                 goto out;
1267
1268         br_vlan_get_tag(skb, &vid);
1269         mp = br_mdb_ip6_get(mlock_dereference(br->mdb, br), group, vid);
1270         if (!mp)
1271                 goto out;
1272
1273         mod_timer(&mp->timer, now + br->multicast_membership_interval);
1274         mp->timer_armed = true;
1275
1276         max_delay *= br->multicast_last_member_count;
1277         if (mp->mglist &&
1278             (timer_pending(&mp->timer) ?
1279              time_after(mp->timer.expires, now + max_delay) :
1280              try_to_del_timer_sync(&mp->timer) >= 0))
1281                 mod_timer(&mp->timer, now + max_delay);
1282
1283         for (pp = &mp->ports;
1284              (p = mlock_dereference(*pp, br)) != NULL;
1285              pp = &p->next) {
1286                 if (timer_pending(&p->timer) ?
1287                     time_after(p->timer.expires, now + max_delay) :
1288                     try_to_del_timer_sync(&p->timer) >= 0)
1289                         mod_timer(&p->timer, now + max_delay);
1290         }
1291
1292 out:
1293         spin_unlock(&br->multicast_lock);
1294         return err;
1295 }
1296 #endif
1297
1298 static void br_multicast_leave_group(struct net_bridge *br,
1299                                      struct net_bridge_port *port,
1300                                      struct br_ip *group,
1301                                      struct bridge_mcast_querier *querier,
1302                                      struct bridge_mcast_query *query)
1303 {
1304         struct net_bridge_mdb_htable *mdb;
1305         struct net_bridge_mdb_entry *mp;
1306         struct net_bridge_port_group *p;
1307         unsigned long now;
1308         unsigned long time;
1309
1310         spin_lock(&br->multicast_lock);
1311         if (!netif_running(br->dev) ||
1312             (port && port->state == BR_STATE_DISABLED) ||
1313             timer_pending(&querier->timer))
1314                 goto out;
1315
1316         mdb = mlock_dereference(br->mdb, br);
1317         mp = br_mdb_ip_get(mdb, group);
1318         if (!mp)
1319                 goto out;
1320
1321         if (br->multicast_querier) {
1322                 __br_multicast_send_query(br, port, &mp->addr);
1323
1324                 time = jiffies + br->multicast_last_member_count *
1325                                  br->multicast_last_member_interval;
1326
1327                 mod_timer(&query->timer, time);
1328
1329                 for (p = mlock_dereference(mp->ports, br);
1330                      p != NULL;
1331                      p = mlock_dereference(p->next, br)) {
1332                         if (p->port != port)
1333                                 continue;
1334
1335                         if (!hlist_unhashed(&p->mglist) &&
1336                             (timer_pending(&p->timer) ?
1337                              time_after(p->timer.expires, time) :
1338                              try_to_del_timer_sync(&p->timer) >= 0)) {
1339                                 mod_timer(&p->timer, time);
1340                         }
1341
1342                         break;
1343                 }
1344         }
1345
1346         if (port && (port->flags & BR_MULTICAST_FAST_LEAVE)) {
1347                 struct net_bridge_port_group __rcu **pp;
1348
1349                 for (pp = &mp->ports;
1350                      (p = mlock_dereference(*pp, br)) != NULL;
1351                      pp = &p->next) {
1352                         if (p->port != port)
1353                                 continue;
1354
1355                         rcu_assign_pointer(*pp, p->next);
1356                         hlist_del_init(&p->mglist);
1357                         del_timer(&p->timer);
1358                         call_rcu_bh(&p->rcu, br_multicast_free_pg);
1359                         br_mdb_notify(br->dev, port, group, RTM_DELMDB);
1360
1361                         if (!mp->ports && !mp->mglist && mp->timer_armed &&
1362                             netif_running(br->dev))
1363                                 mod_timer(&mp->timer, jiffies);
1364                 }
1365                 goto out;
1366         }
1367
1368         now = jiffies;
1369         time = now + br->multicast_last_member_count *
1370                      br->multicast_last_member_interval;
1371
1372         if (!port) {
1373                 if (mp->mglist && mp->timer_armed &&
1374                     (timer_pending(&mp->timer) ?
1375                      time_after(mp->timer.expires, time) :
1376                      try_to_del_timer_sync(&mp->timer) >= 0)) {
1377                         mod_timer(&mp->timer, time);
1378                 }
1379         }
1380 out:
1381         spin_unlock(&br->multicast_lock);
1382 }
1383
1384 static void br_ip4_multicast_leave_group(struct net_bridge *br,
1385                                          struct net_bridge_port *port,
1386                                          __be32 group,
1387                                          __u16 vid)
1388 {
1389         struct br_ip br_group;
1390         struct bridge_mcast_query *query = port ? &port->ip4_query :
1391                                                   &br->ip4_query;
1392
1393         if (ipv4_is_local_multicast(group))
1394                 return;
1395
1396         br_group.u.ip4 = group;
1397         br_group.proto = htons(ETH_P_IP);
1398         br_group.vid = vid;
1399
1400         br_multicast_leave_group(br, port, &br_group, &br->ip4_querier, query);
1401 }
1402
1403 #if IS_ENABLED(CONFIG_IPV6)
1404 static void br_ip6_multicast_leave_group(struct net_bridge *br,
1405                                          struct net_bridge_port *port,
1406                                          const struct in6_addr *group,
1407                                          __u16 vid)
1408 {
1409         struct br_ip br_group;
1410         struct bridge_mcast_query *query = port ? &port->ip6_query :
1411                                                   &br->ip6_query;
1412
1413
1414         if (ipv6_addr_is_ll_all_nodes(group))
1415                 return;
1416
1417         br_group.u.ip6 = *group;
1418         br_group.proto = htons(ETH_P_IPV6);
1419         br_group.vid = vid;
1420
1421         br_multicast_leave_group(br, port, &br_group, &br->ip6_querier, query);
1422 }
1423 #endif
1424
1425 static int br_multicast_ipv4_rcv(struct net_bridge *br,
1426                                  struct net_bridge_port *port,
1427                                  struct sk_buff *skb)
1428 {
1429         struct sk_buff *skb2 = skb;
1430         const struct iphdr *iph;
1431         struct igmphdr *ih;
1432         unsigned int len;
1433         unsigned int offset;
1434         int err;
1435         u16 vid = 0;
1436
1437         /* We treat OOM as packet loss for now. */
1438         if (!pskb_may_pull(skb, sizeof(*iph)))
1439                 return -EINVAL;
1440
1441         iph = ip_hdr(skb);
1442
1443         if (iph->ihl < 5 || iph->version != 4)
1444                 return -EINVAL;
1445
1446         if (!pskb_may_pull(skb, ip_hdrlen(skb)))
1447                 return -EINVAL;
1448
1449         iph = ip_hdr(skb);
1450
1451         if (unlikely(ip_fast_csum((u8 *)iph, iph->ihl)))
1452                 return -EINVAL;
1453
1454         if (iph->protocol != IPPROTO_IGMP) {
1455                 if (!ipv4_is_local_multicast(iph->daddr))
1456                         BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1457                 return 0;
1458         }
1459
1460         len = ntohs(iph->tot_len);
1461         if (skb->len < len || len < ip_hdrlen(skb))
1462                 return -EINVAL;
1463
1464         if (skb->len > len) {
1465                 skb2 = skb_clone(skb, GFP_ATOMIC);
1466                 if (!skb2)
1467                         return -ENOMEM;
1468
1469                 err = pskb_trim_rcsum(skb2, len);
1470                 if (err)
1471                         goto err_out;
1472         }
1473
1474         len -= ip_hdrlen(skb2);
1475         offset = skb_network_offset(skb2) + ip_hdrlen(skb2);
1476         __skb_pull(skb2, offset);
1477         skb_reset_transport_header(skb2);
1478
1479         err = -EINVAL;
1480         if (!pskb_may_pull(skb2, sizeof(*ih)))
1481                 goto out;
1482
1483         switch (skb2->ip_summed) {
1484         case CHECKSUM_COMPLETE:
1485                 if (!csum_fold(skb2->csum))
1486                         break;
1487                 /* fall through */
1488         case CHECKSUM_NONE:
1489                 skb2->csum = 0;
1490                 if (skb_checksum_complete(skb2))
1491                         goto out;
1492         }
1493
1494         err = 0;
1495
1496         br_vlan_get_tag(skb2, &vid);
1497         BR_INPUT_SKB_CB(skb)->igmp = 1;
1498         ih = igmp_hdr(skb2);
1499
1500         switch (ih->type) {
1501         case IGMP_HOST_MEMBERSHIP_REPORT:
1502         case IGMPV2_HOST_MEMBERSHIP_REPORT:
1503                 BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1504                 err = br_ip4_multicast_add_group(br, port, ih->group, vid);
1505                 break;
1506         case IGMPV3_HOST_MEMBERSHIP_REPORT:
1507                 err = br_ip4_multicast_igmp3_report(br, port, skb2);
1508                 break;
1509         case IGMP_HOST_MEMBERSHIP_QUERY:
1510                 err = br_ip4_multicast_query(br, port, skb2);
1511                 break;
1512         case IGMP_HOST_LEAVE_MESSAGE:
1513                 br_ip4_multicast_leave_group(br, port, ih->group, vid);
1514                 break;
1515         }
1516
1517 out:
1518         __skb_push(skb2, offset);
1519 err_out:
1520         if (skb2 != skb)
1521                 kfree_skb(skb2);
1522         return err;
1523 }
1524
1525 #if IS_ENABLED(CONFIG_IPV6)
1526 static int br_multicast_ipv6_rcv(struct net_bridge *br,
1527                                  struct net_bridge_port *port,
1528                                  struct sk_buff *skb)
1529 {
1530         struct sk_buff *skb2;
1531         const struct ipv6hdr *ip6h;
1532         u8 icmp6_type;
1533         u8 nexthdr;
1534         __be16 frag_off;
1535         unsigned int len;
1536         int offset;
1537         int err;
1538         u16 vid = 0;
1539
1540         if (!pskb_may_pull(skb, sizeof(*ip6h)))
1541                 return -EINVAL;
1542
1543         ip6h = ipv6_hdr(skb);
1544
1545         /*
1546          * We're interested in MLD messages only.
1547          *  - Version is 6
1548          *  - MLD has always Router Alert hop-by-hop option
1549          *  - But we do not support jumbrograms.
1550          */
1551         if (ip6h->version != 6)
1552                 return 0;
1553
1554         /* Prevent flooding this packet if there is no listener present */
1555         if (!ipv6_addr_is_ll_all_nodes(&ip6h->daddr))
1556                 BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1557
1558         if (ip6h->nexthdr != IPPROTO_HOPOPTS ||
1559             ip6h->payload_len == 0)
1560                 return 0;
1561
1562         len = ntohs(ip6h->payload_len) + sizeof(*ip6h);
1563         if (skb->len < len)
1564                 return -EINVAL;
1565
1566         nexthdr = ip6h->nexthdr;
1567         offset = ipv6_skip_exthdr(skb, sizeof(*ip6h), &nexthdr, &frag_off);
1568
1569         if (offset < 0 || nexthdr != IPPROTO_ICMPV6)
1570                 return 0;
1571
1572         /* Okay, we found ICMPv6 header */
1573         skb2 = skb_clone(skb, GFP_ATOMIC);
1574         if (!skb2)
1575                 return -ENOMEM;
1576
1577         err = -EINVAL;
1578         if (!pskb_may_pull(skb2, offset + sizeof(struct icmp6hdr)))
1579                 goto out;
1580
1581         len -= offset - skb_network_offset(skb2);
1582
1583         __skb_pull(skb2, offset);
1584         skb_reset_transport_header(skb2);
1585         skb_postpull_rcsum(skb2, skb_network_header(skb2),
1586                            skb_network_header_len(skb2));
1587
1588         icmp6_type = icmp6_hdr(skb2)->icmp6_type;
1589
1590         switch (icmp6_type) {
1591         case ICMPV6_MGM_QUERY:
1592         case ICMPV6_MGM_REPORT:
1593         case ICMPV6_MGM_REDUCTION:
1594         case ICMPV6_MLD2_REPORT:
1595                 break;
1596         default:
1597                 err = 0;
1598                 goto out;
1599         }
1600
1601         /* Okay, we found MLD message. Check further. */
1602         if (skb2->len > len) {
1603                 err = pskb_trim_rcsum(skb2, len);
1604                 if (err)
1605                         goto out;
1606                 err = -EINVAL;
1607         }
1608
1609         ip6h = ipv6_hdr(skb2);
1610
1611         switch (skb2->ip_summed) {
1612         case CHECKSUM_COMPLETE:
1613                 if (!csum_ipv6_magic(&ip6h->saddr, &ip6h->daddr, skb2->len,
1614                                         IPPROTO_ICMPV6, skb2->csum))
1615                         break;
1616                 /*FALLTHROUGH*/
1617         case CHECKSUM_NONE:
1618                 skb2->csum = ~csum_unfold(csum_ipv6_magic(&ip6h->saddr,
1619                                                         &ip6h->daddr,
1620                                                         skb2->len,
1621                                                         IPPROTO_ICMPV6, 0));
1622                 if (__skb_checksum_complete(skb2))
1623                         goto out;
1624         }
1625
1626         err = 0;
1627
1628         br_vlan_get_tag(skb, &vid);
1629         BR_INPUT_SKB_CB(skb)->igmp = 1;
1630
1631         switch (icmp6_type) {
1632         case ICMPV6_MGM_REPORT:
1633             {
1634                 struct mld_msg *mld;
1635                 if (!pskb_may_pull(skb2, sizeof(*mld))) {
1636                         err = -EINVAL;
1637                         goto out;
1638                 }
1639                 mld = (struct mld_msg *)skb_transport_header(skb2);
1640                 BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1641                 err = br_ip6_multicast_add_group(br, port, &mld->mld_mca, vid);
1642                 break;
1643             }
1644         case ICMPV6_MLD2_REPORT:
1645                 err = br_ip6_multicast_mld2_report(br, port, skb2);
1646                 break;
1647         case ICMPV6_MGM_QUERY:
1648                 err = br_ip6_multicast_query(br, port, skb2);
1649                 break;
1650         case ICMPV6_MGM_REDUCTION:
1651             {
1652                 struct mld_msg *mld;
1653                 if (!pskb_may_pull(skb2, sizeof(*mld))) {
1654                         err = -EINVAL;
1655                         goto out;
1656                 }
1657                 mld = (struct mld_msg *)skb_transport_header(skb2);
1658                 br_ip6_multicast_leave_group(br, port, &mld->mld_mca, vid);
1659             }
1660         }
1661
1662 out:
1663         kfree_skb(skb2);
1664         return err;
1665 }
1666 #endif
1667
1668 int br_multicast_rcv(struct net_bridge *br, struct net_bridge_port *port,
1669                      struct sk_buff *skb)
1670 {
1671         BR_INPUT_SKB_CB(skb)->igmp = 0;
1672         BR_INPUT_SKB_CB(skb)->mrouters_only = 0;
1673
1674         if (br->multicast_disabled)
1675                 return 0;
1676
1677         switch (skb->protocol) {
1678         case htons(ETH_P_IP):
1679                 return br_multicast_ipv4_rcv(br, port, skb);
1680 #if IS_ENABLED(CONFIG_IPV6)
1681         case htons(ETH_P_IPV6):
1682                 return br_multicast_ipv6_rcv(br, port, skb);
1683 #endif
1684         }
1685
1686         return 0;
1687 }
1688
1689 static void br_multicast_query_expired(struct net_bridge *br,
1690                                        struct bridge_mcast_query *query)
1691 {
1692         spin_lock(&br->multicast_lock);
1693         if (query->startup_sent < br->multicast_startup_query_count)
1694                 query->startup_sent++;
1695
1696         br_multicast_send_query(br, NULL, query);
1697         spin_unlock(&br->multicast_lock);
1698 }
1699
1700 static void br_ip4_multicast_query_expired(unsigned long data)
1701 {
1702         struct net_bridge *br = (void *)data;
1703
1704         br_multicast_query_expired(br, &br->ip4_query);
1705 }
1706
1707 #if IS_ENABLED(CONFIG_IPV6)
1708 static void br_ip6_multicast_query_expired(unsigned long data)
1709 {
1710         struct net_bridge *br = (void *)data;
1711
1712         br_multicast_query_expired(br, &br->ip6_query);
1713 }
1714 #endif
1715
1716 void br_multicast_init(struct net_bridge *br)
1717 {
1718         br->hash_elasticity = 4;
1719         br->hash_max = 512;
1720
1721         br->multicast_router = 1;
1722         br->multicast_querier = 0;
1723         br->multicast_query_use_ifaddr = 0;
1724         br->multicast_last_member_count = 2;
1725         br->multicast_startup_query_count = 2;
1726
1727         br->multicast_last_member_interval = HZ;
1728         br->multicast_query_response_interval = 10 * HZ;
1729         br->multicast_startup_query_interval = 125 * HZ / 4;
1730         br->multicast_query_interval = 125 * HZ;
1731         br->multicast_querier_interval = 255 * HZ;
1732         br->multicast_membership_interval = 260 * HZ;
1733
1734         br->ip4_querier.delay_time = 0;
1735 #if IS_ENABLED(CONFIG_IPV6)
1736         br->ip6_querier.delay_time = 0;
1737 #endif
1738
1739         spin_lock_init(&br->multicast_lock);
1740         setup_timer(&br->multicast_router_timer,
1741                     br_multicast_local_router_expired, 0);
1742         setup_timer(&br->ip4_querier.timer, br_ip4_multicast_querier_expired,
1743                     (unsigned long)br);
1744         setup_timer(&br->ip4_query.timer, br_ip4_multicast_query_expired,
1745                     (unsigned long)br);
1746 #if IS_ENABLED(CONFIG_IPV6)
1747         setup_timer(&br->ip6_querier.timer, br_ip6_multicast_querier_expired,
1748                     (unsigned long)br);
1749         setup_timer(&br->ip6_query.timer, br_ip6_multicast_query_expired,
1750                     (unsigned long)br);
1751 #endif
1752 }
1753
1754 static void __br_multicast_open(struct net_bridge *br,
1755                                 struct bridge_mcast_query *query)
1756 {
1757         query->startup_sent = 0;
1758
1759         if (br->multicast_disabled)
1760                 return;
1761
1762         mod_timer(&query->timer, jiffies);
1763 }
1764
1765 void br_multicast_open(struct net_bridge *br)
1766 {
1767         __br_multicast_open(br, &br->ip4_query);
1768 #if IS_ENABLED(CONFIG_IPV6)
1769         __br_multicast_open(br, &br->ip6_query);
1770 #endif
1771 }
1772
1773 void br_multicast_stop(struct net_bridge *br)
1774 {
1775         struct net_bridge_mdb_htable *mdb;
1776         struct net_bridge_mdb_entry *mp;
1777         struct hlist_node *n;
1778         u32 ver;
1779         int i;
1780
1781         del_timer_sync(&br->multicast_router_timer);
1782         del_timer_sync(&br->ip4_querier.timer);
1783         del_timer_sync(&br->ip4_query.timer);
1784 #if IS_ENABLED(CONFIG_IPV6)
1785         del_timer_sync(&br->ip6_querier.timer);
1786         del_timer_sync(&br->ip6_query.timer);
1787 #endif
1788
1789         spin_lock_bh(&br->multicast_lock);
1790         mdb = mlock_dereference(br->mdb, br);
1791         if (!mdb)
1792                 goto out;
1793
1794         br->mdb = NULL;
1795
1796         ver = mdb->ver;
1797         for (i = 0; i < mdb->max; i++) {
1798                 hlist_for_each_entry_safe(mp, n, &mdb->mhash[i],
1799                                           hlist[ver]) {
1800                         del_timer(&mp->timer);
1801                         mp->timer_armed = false;
1802                         call_rcu_bh(&mp->rcu, br_multicast_free_group);
1803                 }
1804         }
1805
1806         if (mdb->old) {
1807                 spin_unlock_bh(&br->multicast_lock);
1808                 rcu_barrier_bh();
1809                 spin_lock_bh(&br->multicast_lock);
1810                 WARN_ON(mdb->old);
1811         }
1812
1813         mdb->old = mdb;
1814         call_rcu_bh(&mdb->rcu, br_mdb_free);
1815
1816 out:
1817         spin_unlock_bh(&br->multicast_lock);
1818 }
1819
1820 int br_multicast_set_router(struct net_bridge *br, unsigned long val)
1821 {
1822         int err = -ENOENT;
1823
1824         spin_lock_bh(&br->multicast_lock);
1825         if (!netif_running(br->dev))
1826                 goto unlock;
1827
1828         switch (val) {
1829         case 0:
1830         case 2:
1831                 del_timer(&br->multicast_router_timer);
1832                 /* fall through */
1833         case 1:
1834                 br->multicast_router = val;
1835                 err = 0;
1836                 break;
1837
1838         default:
1839                 err = -EINVAL;
1840                 break;
1841         }
1842
1843 unlock:
1844         spin_unlock_bh(&br->multicast_lock);
1845
1846         return err;
1847 }
1848
1849 int br_multicast_set_port_router(struct net_bridge_port *p, unsigned long val)
1850 {
1851         struct net_bridge *br = p->br;
1852         int err = -ENOENT;
1853
1854         spin_lock(&br->multicast_lock);
1855         if (!netif_running(br->dev) || p->state == BR_STATE_DISABLED)
1856                 goto unlock;
1857
1858         switch (val) {
1859         case 0:
1860         case 1:
1861         case 2:
1862                 p->multicast_router = val;
1863                 err = 0;
1864
1865                 if (val < 2 && !hlist_unhashed(&p->rlist))
1866                         hlist_del_init_rcu(&p->rlist);
1867
1868                 if (val == 1)
1869                         break;
1870
1871                 del_timer(&p->multicast_router_timer);
1872
1873                 if (val == 0)
1874                         break;
1875
1876                 br_multicast_add_router(br, p);
1877                 break;
1878
1879         default:
1880                 err = -EINVAL;
1881                 break;
1882         }
1883
1884 unlock:
1885         spin_unlock(&br->multicast_lock);
1886
1887         return err;
1888 }
1889
1890 static void br_multicast_start_querier(struct net_bridge *br,
1891                                        struct bridge_mcast_query *query)
1892 {
1893         struct net_bridge_port *port;
1894
1895         __br_multicast_open(br, query);
1896
1897         list_for_each_entry(port, &br->port_list, list) {
1898                 if (port->state == BR_STATE_DISABLED ||
1899                     port->state == BR_STATE_BLOCKING)
1900                         continue;
1901
1902                 if (query == &br->ip4_query)
1903                         br_multicast_enable(&port->ip4_query);
1904 #if IS_ENABLED(CONFIG_IPV6)
1905                 else
1906                         br_multicast_enable(&port->ip6_query);
1907 #endif
1908         }
1909 }
1910
1911 int br_multicast_toggle(struct net_bridge *br, unsigned long val)
1912 {
1913         int err = 0;
1914         struct net_bridge_mdb_htable *mdb;
1915
1916         spin_lock_bh(&br->multicast_lock);
1917         if (br->multicast_disabled == !val)
1918                 goto unlock;
1919
1920         br->multicast_disabled = !val;
1921         if (br->multicast_disabled)
1922                 goto unlock;
1923
1924         if (!netif_running(br->dev))
1925                 goto unlock;
1926
1927         mdb = mlock_dereference(br->mdb, br);
1928         if (mdb) {
1929                 if (mdb->old) {
1930                         err = -EEXIST;
1931 rollback:
1932                         br->multicast_disabled = !!val;
1933                         goto unlock;
1934                 }
1935
1936                 err = br_mdb_rehash(&br->mdb, mdb->max,
1937                                     br->hash_elasticity);
1938                 if (err)
1939                         goto rollback;
1940         }
1941
1942         br_multicast_start_querier(br, &br->ip4_query);
1943 #if IS_ENABLED(CONFIG_IPV6)
1944         br_multicast_start_querier(br, &br->ip6_query);
1945 #endif
1946
1947 unlock:
1948         spin_unlock_bh(&br->multicast_lock);
1949
1950         return err;
1951 }
1952
1953 int br_multicast_set_querier(struct net_bridge *br, unsigned long val)
1954 {
1955         unsigned long max_delay;
1956
1957         val = !!val;
1958
1959         spin_lock_bh(&br->multicast_lock);
1960         if (br->multicast_querier == val)
1961                 goto unlock;
1962
1963         br->multicast_querier = val;
1964         if (!val)
1965                 goto unlock;
1966
1967         max_delay = br->multicast_query_response_interval;
1968
1969         if (!timer_pending(&br->ip4_querier.timer))
1970                 br->ip4_querier.delay_time = jiffies + max_delay;
1971
1972         br_multicast_start_querier(br, &br->ip4_query);
1973
1974 #if IS_ENABLED(CONFIG_IPV6)
1975         if (!timer_pending(&br->ip6_querier.timer))
1976                 br->ip6_querier.delay_time = jiffies + max_delay;
1977
1978         br_multicast_start_querier(br, &br->ip6_query);
1979 #endif
1980
1981 unlock:
1982         spin_unlock_bh(&br->multicast_lock);
1983
1984         return 0;
1985 }
1986
1987 int br_multicast_set_hash_max(struct net_bridge *br, unsigned long val)
1988 {
1989         int err = -ENOENT;
1990         u32 old;
1991         struct net_bridge_mdb_htable *mdb;
1992
1993         spin_lock(&br->multicast_lock);
1994         if (!netif_running(br->dev))
1995                 goto unlock;
1996
1997         err = -EINVAL;
1998         if (!is_power_of_2(val))
1999                 goto unlock;
2000
2001         mdb = mlock_dereference(br->mdb, br);
2002         if (mdb && val < mdb->size)
2003                 goto unlock;
2004
2005         err = 0;
2006
2007         old = br->hash_max;
2008         br->hash_max = val;
2009
2010         if (mdb) {
2011                 if (mdb->old) {
2012                         err = -EEXIST;
2013 rollback:
2014                         br->hash_max = old;
2015                         goto unlock;
2016                 }
2017
2018                 err = br_mdb_rehash(&br->mdb, br->hash_max,
2019                                     br->hash_elasticity);
2020                 if (err)
2021                         goto rollback;
2022         }
2023
2024 unlock:
2025         spin_unlock(&br->multicast_lock);
2026
2027         return err;
2028 }