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/export.h>
  15 #include <linux/if_ether.h>
  16 #include <linux/igmp.h>
  17 #include <linux/jhash.h>
  18 #include <linux/kernel.h>
  19 #include <linux/log2.h>
  20 #include <linux/netdevice.h>
  21 #include <linux/netfilter_bridge.h>
  22 #include <linux/random.h>
  23 #include <linux/rculist.h>
  24 #include <linux/skbuff.h>
  25 #include <linux/slab.h>
  26 #include <linux/timer.h>
  27 #include <linux/inetdevice.h>
  28 #include <linux/mroute.h>
  29 #include <net/ip.h>
  30 #if IS_ENABLED(CONFIG_IPV6)
  31 #include <net/ipv6.h>
  32 #include <net/mld.h>
  33 #include <net/ip6_checksum.h>
  34 #include <net/addrconf.h>
  35 #endif
  36
  37 #include "br_private.h"
  38
  39 static void br_multicast_start_querier(struct net_bridge *br,
  40                                        struct bridge_mcast_own_query *query);
  41 static void br_multicast_add_router(struct net_bridge *br,
  42                                     struct net_bridge_port *port);
  43 static void br_ip4_multicast_leave_group(struct net_bridge *br,
  44                                          struct net_bridge_port *port,
  45                                          __be32 group,
  46                                          __u16 vid);
  47 #if IS_ENABLED(CONFIG_IPV6)
  48 static void br_ip6_multicast_leave_group(struct net_bridge *br,
  49                                          struct net_bridge_port *port,
  50                                          const struct in6_addr *group,
  51                                          __u16 vid);
  52 #endif
  53 unsigned int br_mdb_rehash_seq;
  54
  55 static inline int br_ip_equal(const struct br_ip *a, const struct br_ip *b)
  56 {
  57         if (a->proto != b->proto)
  58                 return 0;
  59         if (a->vid != b->vid)
  60                 return 0;
  61         switch (a->proto) {
  62         case htons(ETH_P_IP):
  63                 return a->u.ip4 == b->u.ip4;
  64 #if IS_ENABLED(CONFIG_IPV6)
  65         case htons(ETH_P_IPV6):
  66                 return ipv6_addr_equal(&a->u.ip6, &b->u.ip6);
  67 #endif
  68         }
  69         return 0;
  70 }
  71
  72 static inline int __br_ip4_hash(struct net_bridge_mdb_htable *mdb, __be32 ip,
  73                                 __u16 vid)
  74 {
  75         return jhash_2words((__force u32)ip, vid, mdb->secret) & (mdb->max - 1);
  76 }
  77
  78 #if IS_ENABLED(CONFIG_IPV6)
  79 static inline int __br_ip6_hash(struct net_bridge_mdb_htable *mdb,
  80                                 const struct in6_addr *ip,
  81                                 __u16 vid)
  82 {
  83         return jhash_2words(ipv6_addr_hash(ip), vid,
  84                             mdb->secret) & (mdb->max - 1);
  85 }
  86 #endif
  87
  88 static inline int br_ip_hash(struct net_bridge_mdb_htable *mdb,
  89                              struct br_ip *ip)
  90 {
  91         switch (ip->proto) {
  92         case htons(ETH_P_IP):
  93                 return __br_ip4_hash(mdb, ip->u.ip4, ip->vid);
  94 #if IS_ENABLED(CONFIG_IPV6)
  95         case htons(ETH_P_IPV6):
  96                 return __br_ip6_hash(mdb, &ip->u.ip6, ip->vid);
  97 #endif
  98         }
  99         return 0;
 100 }
 101
 102 static struct net_bridge_mdb_entry *__br_mdb_ip_get(
 103         struct net_bridge_mdb_htable *mdb, struct br_ip *dst, int hash)
 104 {
 105         struct net_bridge_mdb_entry *mp;
 106
 107         hlist_for_each_entry_rcu(mp, &mdb->mhash[hash], hlist[mdb->ver]) {
 108                 if (br_ip_equal(&mp->addr, dst))
 109                         return mp;
 110         }
 111
 112         return NULL;
 113 }
 114
 115 struct net_bridge_mdb_entry *br_mdb_ip_get(struct net_bridge_mdb_htable *mdb,
 116                                            struct br_ip *dst)
 117 {
 118         if (!mdb)
 119                 return NULL;
 120
 121         return __br_mdb_ip_get(mdb, dst, br_ip_hash(mdb, dst));
 122 }
 123
 124 static struct net_bridge_mdb_entry *br_mdb_ip4_get(
 125         struct net_bridge_mdb_htable *mdb, __be32 dst, __u16 vid)
 126 {
 127         struct br_ip br_dst;
 128
 129         br_dst.u.ip4 = dst;
 130         br_dst.proto = htons(ETH_P_IP);
 131         br_dst.vid = vid;
 132
 133         return br_mdb_ip_get(mdb, &br_dst);
 134 }
 135
 136 #if IS_ENABLED(CONFIG_IPV6)
 137 static struct net_bridge_mdb_entry *br_mdb_ip6_get(
 138         struct net_bridge_mdb_htable *mdb, const struct in6_addr *dst,
 139         __u16 vid)
 140 {
 141         struct br_ip br_dst;
 142
 143         br_dst.u.ip6 = *dst;
 144         br_dst.proto = htons(ETH_P_IPV6);
 145         br_dst.vid = vid;
 146
 147         return br_mdb_ip_get(mdb, &br_dst);
 148 }
 149 #endif
 150
 151 struct net_bridge_mdb_entry *br_mdb_get(struct net_bridge *br,
 152                                         struct sk_buff *skb, u16 vid)
 153 {
 154         struct net_bridge_mdb_htable *mdb = rcu_dereference(br->mdb);
 155         struct br_ip ip;
 156
 157         if (br->multicast_disabled)
 158                 return NULL;
 159
 160         if (BR_INPUT_SKB_CB(skb)->igmp)
 161                 return NULL;
 162
 163         ip.proto = skb->protocol;
 164         ip.vid = vid;
 165
 166         switch (skb->protocol) {
 167         case htons(ETH_P_IP):
 168                 ip.u.ip4 = ip_hdr(skb)->daddr;
 169                 break;
 170 #if IS_ENABLED(CONFIG_IPV6)
 171         case htons(ETH_P_IPV6):
 172                 ip.u.ip6 = ipv6_hdr(skb)->daddr;
 173                 break;
 174 #endif
 175         default:
 176                 return NULL;
 177         }
 178
 179         return br_mdb_ip_get(mdb, &ip);
 180 }
 181
 182 static void br_mdb_free(struct rcu_head *head)
 183 {
 184         struct net_bridge_mdb_htable *mdb =
 185                 container_of(head, struct net_bridge_mdb_htable, rcu);
 186         struct net_bridge_mdb_htable *old = mdb->old;
 187
 188         mdb->old = NULL;
 189         kfree(old->mhash);
 190         kfree(old);
 191 }
 192
 193 static int br_mdb_copy(struct net_bridge_mdb_htable *new,
 194                        struct net_bridge_mdb_htable *old,
 195                        int elasticity)
 196 {
 197         struct net_bridge_mdb_entry *mp;
 198         int maxlen;
 199         int len;
 200         int i;
 201
 202         for (i = 0; i < old->max; i++)
 203                 hlist_for_each_entry(mp, &old->mhash[i], hlist[old->ver])
 204                         hlist_add_head(&mp->hlist[new->ver],
 205                                        &new->mhash[br_ip_hash(new, &mp->addr)]);
 206
 207         if (!elasticity)
 208                 return 0;
 209
 210         maxlen = 0;
 211         for (i = 0; i < new->max; i++) {
 212                 len = 0;
 213                 hlist_for_each_entry(mp, &new->mhash[i], hlist[new->ver])
 214                         len++;
 215                 if (len > maxlen)
 216                         maxlen = len;
 217         }
 218
 219         return maxlen > elasticity ? -EINVAL : 0;
 220 }
 221
 222 void br_multicast_free_pg(struct rcu_head *head)
 223 {
 224         struct net_bridge_port_group *p =
 225                 container_of(head, struct net_bridge_port_group, rcu);
 226
 227         kfree(p);
 228 }
 229
 230 static void br_multicast_free_group(struct rcu_head *head)
 231 {
 232         struct net_bridge_mdb_entry *mp =
 233                 container_of(head, struct net_bridge_mdb_entry, rcu);
 234
 235         kfree(mp);
 236 }
 237
 238 static void br_multicast_group_expired(unsigned long data)
 239 {
 240         struct net_bridge_mdb_entry *mp = (void *)data;
 241         struct net_bridge *br = mp->br;
 242         struct net_bridge_mdb_htable *mdb;
 243
 244         spin_lock(&br->multicast_lock);
 245         if (!netif_running(br->dev) || timer_pending(&mp->timer))
 246                 goto out;
 247
 248         mp->mglist = false;
 249
 250         if (mp->ports)
 251                 goto out;
 252
 253         mdb = mlock_dereference(br->mdb, br);
 254
 255         hlist_del_rcu(&mp->hlist[mdb->ver]);
 256         mdb->size--;
 257
 258         call_rcu_bh(&mp->rcu, br_multicast_free_group);
 259
 260 out:
 261         spin_unlock(&br->multicast_lock);
 262 }
 263
 264 static void br_multicast_del_pg(struct net_bridge *br,
 265                                 struct net_bridge_port_group *pg)
 266 {
 267         struct net_bridge_mdb_htable *mdb;
 268         struct net_bridge_mdb_entry *mp;
 269         struct net_bridge_port_group *p;
 270         struct net_bridge_port_group __rcu **pp;
 271
 272         mdb = mlock_dereference(br->mdb, br);
 273
 274         mp = br_mdb_ip_get(mdb, &pg->addr);
 275         if (WARN_ON(!mp))
 276                 return;
 277
 278         for (pp = &mp->ports;
 279              (p = mlock_dereference(*pp, br)) != NULL;
 280              pp = &p->next) {
 281                 if (p != pg)
 282                         continue;
 283
 284                 rcu_assign_pointer(*pp, p->next);
 285                 hlist_del_init(&p->mglist);
 286                 del_timer(&p->timer);
 287                 br_mdb_notify(br->dev, p->port, &pg->addr, RTM_DELMDB,
 288                               p->flags);
 289                 call_rcu_bh(&p->rcu, br_multicast_free_pg);
 290
 291                 if (!mp->ports && !mp->mglist &&
 292                     netif_running(br->dev))
 293                         mod_timer(&mp->timer, jiffies);
 294
 295                 return;
 296         }
 297
 298         WARN_ON(1);
 299 }
 300
 301 static void br_multicast_port_group_expired(unsigned long data)
 302 {
 303         struct net_bridge_port_group *pg = (void *)data;
 304         struct net_bridge *br = pg->port->br;
 305
 306         spin_lock(&br->multicast_lock);
 307         if (!netif_running(br->dev) || timer_pending(&pg->timer) ||
 308             hlist_unhashed(&pg->mglist) || pg->flags & MDB_PG_FLAGS_PERMANENT)
 309                 goto out;
 310
 311         br_multicast_del_pg(br, pg);
 312
 313 out:
 314         spin_unlock(&br->multicast_lock);
 315 }
 316
 317 static int br_mdb_rehash(struct net_bridge_mdb_htable __rcu **mdbp, int max,
 318                          int elasticity)
 319 {
 320         struct net_bridge_mdb_htable *old = rcu_dereference_protected(*mdbp, 1);
 321         struct net_bridge_mdb_htable *mdb;
 322         int err;
 323
 324         mdb = kmalloc(sizeof(*mdb), GFP_ATOMIC);
 325         if (!mdb)
 326                 return -ENOMEM;
 327
 328         mdb->max = max;
 329         mdb->old = old;
 330
 331         mdb->mhash = kzalloc(max * sizeof(*mdb->mhash), GFP_ATOMIC);
 332         if (!mdb->mhash) {
 333                 kfree(mdb);
 334                 return -ENOMEM;
 335         }
 336
 337         mdb->size = old ? old->size : 0;
 338         mdb->ver = old ? old->ver ^ 1 : 0;
 339
 340         if (!old || elasticity)
 341                 get_random_bytes(&mdb->secret, sizeof(mdb->secret));
 342         else
 343                 mdb->secret = old->secret;
 344
 345         if (!old)
 346                 goto out;
 347
 348         err = br_mdb_copy(mdb, old, elasticity);
 349         if (err) {
 350                 kfree(mdb->mhash);
 351                 kfree(mdb);
 352                 return err;
 353         }
 354
 355         br_mdb_rehash_seq++;
 356         call_rcu_bh(&mdb->rcu, br_mdb_free);
 357
 358 out:
 359         rcu_assign_pointer(*mdbp, mdb);
 360
 361         return 0;
 362 }
 363
 364 static struct sk_buff *br_ip4_multicast_alloc_query(struct net_bridge *br,
 365                                                     __be32 group,
 366                                                     u8 *igmp_type)
 367 {
 368         struct igmpv3_query *ihv3;
 369         size_t igmp_hdr_size;
 370         struct sk_buff *skb;
 371         struct igmphdr *ih;
 372         struct ethhdr *eth;
 373         struct iphdr *iph;
 374
 375         igmp_hdr_size = sizeof(*ih);
 376         if (br->multicast_igmp_version == 3)
 377                 igmp_hdr_size = sizeof(*ihv3);
 378         skb = netdev_alloc_skb_ip_align(br->dev, sizeof(*eth) + sizeof(*iph) +
 379                                                  igmp_hdr_size + 4);
 380         if (!skb)
 381                 goto out;
 382
 383         skb->protocol = htons(ETH_P_IP);
 384
 385         skb_reset_mac_header(skb);
 386         eth = eth_hdr(skb);
 387
 388         ether_addr_copy(eth->h_source, br->dev->dev_addr);
 389         eth->h_dest[0] = 1;
 390         eth->h_dest[1] = 0;
 391         eth->h_dest[2] = 0x5e;
 392         eth->h_dest[3] = 0;
 393         eth->h_dest[4] = 0;
 394         eth->h_dest[5] = 1;
 395         eth->h_proto = htons(ETH_P_IP);
 396         skb_put(skb, sizeof(*eth));
 397
 398         skb_set_network_header(skb, skb->len);
 399         iph = ip_hdr(skb);
 400
 401         iph->version = 4;
 402         iph->ihl = 6;
 403         iph->tos = 0xc0;
 404         iph->tot_len = htons(sizeof(*iph) + igmp_hdr_size + 4);
 405         iph->id = 0;
 406         iph->frag_off = htons(IP_DF);
 407         iph->ttl = 1;
 408         iph->protocol = IPPROTO_IGMP;
 409         iph->saddr = br->multicast_query_use_ifaddr ?
 410                      inet_select_addr(br->dev, 0, RT_SCOPE_LINK) : 0;
 411         iph->daddr = htonl(INADDR_ALLHOSTS_GROUP);
 412         ((u8 *)&iph[1])[0] = IPOPT_RA;
 413         ((u8 *)&iph[1])[1] = 4;
 414         ((u8 *)&iph[1])[2] = 0;
 415         ((u8 *)&iph[1])[3] = 0;
 416         ip_send_check(iph);
 417         skb_put(skb, 24);
 418
 419         skb_set_transport_header(skb, skb->len);
 420         *igmp_type = IGMP_HOST_MEMBERSHIP_QUERY;
 421
 422         switch (br->multicast_igmp_version) {
 423         case 2:
 424                 ih = igmp_hdr(skb);
 425                 ih->type = IGMP_HOST_MEMBERSHIP_QUERY;
 426                 ih->code = (group ? br->multicast_last_member_interval :
 427                                     br->multicast_query_response_interval) /
 428                            (HZ / IGMP_TIMER_SCALE);
 429                 ih->group = group;
 430                 ih->csum = 0;
 431                 ih->csum = ip_compute_csum((void *)ih, sizeof(*ih));
 432                 break;
 433         case 3:
 434                 ihv3 = igmpv3_query_hdr(skb);
 435                 ihv3->type = IGMP_HOST_MEMBERSHIP_QUERY;
 436                 ihv3->code = (group ? br->multicast_last_member_interval :
 437                                       br->multicast_query_response_interval) /
 438                              (HZ / IGMP_TIMER_SCALE);
 439                 ihv3->group = group;
 440                 ihv3->qqic = br->multicast_query_interval / HZ;
 441                 ihv3->nsrcs = 0;
 442                 ihv3->resv = 0;
 443                 ihv3->suppress = 0;
 444                 ihv3->qrv = 2;
 445                 ihv3->csum = 0;
 446                 ihv3->csum = ip_compute_csum((void *)ihv3, sizeof(*ihv3));
 447                 break;
 448         }
 449
 450         skb_put(skb, igmp_hdr_size);
 451         __skb_pull(skb, sizeof(*eth));
 452
 453 out:
 454         return skb;
 455 }
 456
 457 #if IS_ENABLED(CONFIG_IPV6)
 458 static struct sk_buff *br_ip6_multicast_alloc_query(struct net_bridge *br,
 459                                                     const struct in6_addr *grp,
 460                                                     u8 *igmp_type)
 461 {
 462         struct mld2_query *mld2q;
 463         unsigned long interval;
 464         struct ipv6hdr *ip6h;
 465         struct mld_msg *mldq;
 466         size_t mld_hdr_size;
 467         struct sk_buff *skb;
 468         struct ethhdr *eth;
 469         u8 *hopopt;
 470
 471         mld_hdr_size = sizeof(*mldq);
 472         if (br->multicast_mld_version == 2)
 473                 mld_hdr_size = sizeof(*mld2q);
 474         skb = netdev_alloc_skb_ip_align(br->dev, sizeof(*eth) + sizeof(*ip6h) +
 475                                                  8 + mld_hdr_size);
 476         if (!skb)
 477                 goto out;
 478
 479         skb->protocol = htons(ETH_P_IPV6);
 480
 481         /* Ethernet header */
 482         skb_reset_mac_header(skb);
 483         eth = eth_hdr(skb);
 484
 485         ether_addr_copy(eth->h_source, br->dev->dev_addr);
 486         eth->h_proto = htons(ETH_P_IPV6);
 487         skb_put(skb, sizeof(*eth));
 488
 489         /* IPv6 header + HbH option */
 490         skb_set_network_header(skb, skb->len);
 491         ip6h = ipv6_hdr(skb);
 492
 493         *(__force __be32 *)ip6h = htonl(0x60000000);
 494         ip6h->payload_len = htons(8 + mld_hdr_size);
 495         ip6h->nexthdr = IPPROTO_HOPOPTS;
 496         ip6h->hop_limit = 1;
 497         ipv6_addr_set(&ip6h->daddr, htonl(0xff020000), 0, 0, htonl(1));
 498         if (ipv6_dev_get_saddr(dev_net(br->dev), br->dev, &ip6h->daddr, 0,
 499                                &ip6h->saddr)) {
 500                 kfree_skb(skb);
 501                 br->has_ipv6_addr = 0;
 502                 return NULL;
 503         }
 504
 505         br->has_ipv6_addr = 1;
 506         ipv6_eth_mc_map(&ip6h->daddr, eth->h_dest);
 507
 508         hopopt = (u8 *)(ip6h + 1);
 509         hopopt[0] = IPPROTO_ICMPV6;             /* next hdr */
 510         hopopt[1] = 0;                          /* length of HbH */
 511         hopopt[2] = IPV6_TLV_ROUTERALERT;       /* Router Alert */
 512         hopopt[3] = 2;                          /* Length of RA Option */
 513         hopopt[4] = 0;                          /* Type = 0x0000 (MLD) */
 514         hopopt[5] = 0;
 515         hopopt[6] = IPV6_TLV_PAD1;              /* Pad1 */
 516         hopopt[7] = IPV6_TLV_PAD1;              /* Pad1 */
 517
 518         skb_put(skb, sizeof(*ip6h) + 8);
 519
 520         /* ICMPv6 */
 521         skb_set_transport_header(skb, skb->len);
 522         interval = ipv6_addr_any(grp) ?
 523                         br->multicast_query_response_interval :
 524                         br->multicast_last_member_interval;
 525         *igmp_type = ICMPV6_MGM_QUERY;
 526         switch (br->multicast_mld_version) {
 527         case 1:
 528                 mldq = (struct mld_msg *)icmp6_hdr(skb);
 529                 mldq->mld_type = ICMPV6_MGM_QUERY;
 530                 mldq->mld_code = 0;
 531                 mldq->mld_cksum = 0;
 532                 mldq->mld_maxdelay = htons((u16)jiffies_to_msecs(interval));
 533                 mldq->mld_reserved = 0;
 534                 mldq->mld_mca = *grp;
 535                 mldq->mld_cksum = csum_ipv6_magic(&ip6h->saddr, &ip6h->daddr,
 536                                                   sizeof(*mldq), IPPROTO_ICMPV6,
 537                                                   csum_partial(mldq,
 538                                                                sizeof(*mldq),
 539                                                                0));
 540                 break;
 541         case 2:
 542                 mld2q = (struct mld2_query *)icmp6_hdr(skb);
 543                 mld2q->mld2q_mrc = ntohs((u16)jiffies_to_msecs(interval));
 544                 mld2q->mld2q_type = ICMPV6_MGM_QUERY;
 545                 mld2q->mld2q_code = 0;
 546                 mld2q->mld2q_cksum = 0;
 547                 mld2q->mld2q_resv1 = 0;
 548                 mld2q->mld2q_resv2 = 0;
 549                 mld2q->mld2q_suppress = 0;
 550                 mld2q->mld2q_qrv = 2;
 551                 mld2q->mld2q_nsrcs = 0;
 552                 mld2q->mld2q_qqic = br->multicast_query_interval / HZ;
 553                 mld2q->mld2q_mca = *grp;
 554                 mld2q->mld2q_cksum = csum_ipv6_magic(&ip6h->saddr, &ip6h->daddr,
 555                                                      sizeof(*mld2q),
 556                                                      IPPROTO_ICMPV6,
 557                                                      csum_partial(mld2q,
 558                                                                   sizeof(*mld2q),
 559                                                                   0));
 560                 break;
 561         }
 562         skb_put(skb, mld_hdr_size);
 563
 564         __skb_pull(skb, sizeof(*eth));
 565
 566 out:
 567         return skb;
 568 }
 569 #endif
 570
 571 static struct sk_buff *br_multicast_alloc_query(struct net_bridge *br,
 572                                                 struct br_ip *addr,
 573                                                 u8 *igmp_type)
 574 {
 575         switch (addr->proto) {
 576         case htons(ETH_P_IP):
 577                 return br_ip4_multicast_alloc_query(br, addr->u.ip4, igmp_type);
 578 #if IS_ENABLED(CONFIG_IPV6)
 579         case htons(ETH_P_IPV6):
 580                 return br_ip6_multicast_alloc_query(br, &addr->u.ip6,
 581                                                     igmp_type);
 582 #endif
 583         }
 584         return NULL;
 585 }
 586
 587 static struct net_bridge_mdb_entry *br_multicast_get_group(
 588         struct net_bridge *br, struct net_bridge_port *port,
 589         struct br_ip *group, int hash)
 590 {
 591         struct net_bridge_mdb_htable *mdb;
 592         struct net_bridge_mdb_entry *mp;
 593         unsigned int count = 0;
 594         unsigned int max;
 595         int elasticity;
 596         int err;
 597
 598         mdb = rcu_dereference_protected(br->mdb, 1);
 599         hlist_for_each_entry(mp, &mdb->mhash[hash], hlist[mdb->ver]) {
 600                 count++;
 601                 if (unlikely(br_ip_equal(group, &mp->addr)))
 602                         return mp;
 603         }
 604
 605         elasticity = 0;
 606         max = mdb->max;
 607
 608         if (unlikely(count > br->hash_elasticity && count)) {
 609                 if (net_ratelimit())
 610                         br_info(br, "Multicast hash table "
 611                                 "chain limit reached: %s\n",
 612                                 port ? port->dev->name : br->dev->name);
 613
 614                 elasticity = br->hash_elasticity;
 615         }
 616
 617         if (mdb->size >= max) {
 618                 max *= 2;
 619                 if (unlikely(max > br->hash_max)) {
 620                         br_warn(br, "Multicast hash table maximum of %d "
 621                                 "reached, disabling snooping: %s\n",
 622                                 br->hash_max,
 623                                 port ? port->dev->name : br->dev->name);
 624                         err = -E2BIG;
 625 disable:
 626                         br->multicast_disabled = 1;
 627                         goto err;
 628                 }
 629         }
 630
 631         if (max > mdb->max || elasticity) {
 632                 if (mdb->old) {
 633                         if (net_ratelimit())
 634                                 br_info(br, "Multicast hash table "
 635                                         "on fire: %s\n",
 636                                         port ? port->dev->name : br->dev->name);
 637                         err = -EEXIST;
 638                         goto err;
 639                 }
 640
 641                 err = br_mdb_rehash(&br->mdb, max, elasticity);
 642                 if (err) {
 643                         br_warn(br, "Cannot rehash multicast "
 644                                 "hash table, disabling snooping: %s, %d, %d\n",
 645                                 port ? port->dev->name : br->dev->name,
 646                                 mdb->size, err);
 647                         goto disable;
 648                 }
 649
 650                 err = -EAGAIN;
 651                 goto err;
 652         }
 653
 654         return NULL;
 655
 656 err:
 657         mp = ERR_PTR(err);
 658         return mp;
 659 }
 660
 661 struct net_bridge_mdb_entry *br_multicast_new_group(struct net_bridge *br,
 662                                                     struct net_bridge_port *p,
 663                                                     struct br_ip *group)
 664 {
 665         struct net_bridge_mdb_htable *mdb;
 666         struct net_bridge_mdb_entry *mp;
 667         int hash;
 668         int err;
 669
 670         mdb = rcu_dereference_protected(br->mdb, 1);
 671         if (!mdb) {
 672                 err = br_mdb_rehash(&br->mdb, BR_HASH_SIZE, 0);
 673                 if (err)
 674                         return ERR_PTR(err);
 675                 goto rehash;
 676         }
 677
 678         hash = br_ip_hash(mdb, group);
 679         mp = br_multicast_get_group(br, p, group, hash);
 680         switch (PTR_ERR(mp)) {
 681         case 0:
 682                 break;
 683
 684         case -EAGAIN:
 685 rehash:
 686                 mdb = rcu_dereference_protected(br->mdb, 1);
 687                 hash = br_ip_hash(mdb, group);
 688                 break;
 689
 690         default:
 691                 goto out;
 692         }
 693
 694         mp = kzalloc(sizeof(*mp), GFP_ATOMIC);
 695         if (unlikely(!mp))
 696                 return ERR_PTR(-ENOMEM);
 697
 698         mp->br = br;
 699         mp->addr = *group;
 700         setup_timer(&mp->timer, br_multicast_group_expired,
 701                     (unsigned long)mp);
 702
 703         hlist_add_head_rcu(&mp->hlist[mdb->ver], &mdb->mhash[hash]);
 704         mdb->size++;
 705
 706 out:
 707         return mp;
 708 }
 709
 710 struct net_bridge_port_group *br_multicast_new_port_group(
 711                         struct net_bridge_port *port,
 712                         struct br_ip *group,
 713                         struct net_bridge_port_group __rcu *next,
 714                         unsigned char flags)
 715 {
 716         struct net_bridge_port_group *p;
 717
 718         p = kzalloc(sizeof(*p), GFP_ATOMIC);
 719         if (unlikely(!p))
 720                 return NULL;
 721
 722         p->addr = *group;
 723         p->port = port;
 724         p->flags = flags;
 725         rcu_assign_pointer(p->next, next);
 726         hlist_add_head(&p->mglist, &port->mglist);
 727         setup_timer(&p->timer, br_multicast_port_group_expired,
 728                     (unsigned long)p);
 729         return p;
 730 }
 731
 732 static int br_multicast_add_group(struct net_bridge *br,
 733                                   struct net_bridge_port *port,
 734                                   struct br_ip *group)
 735 {
 736         struct net_bridge_port_group __rcu **pp;
 737         struct net_bridge_port_group *p;
 738         struct net_bridge_mdb_entry *mp;
 739         unsigned long now = jiffies;
 740         int err;
 741
 742         spin_lock(&br->multicast_lock);
 743         if (!netif_running(br->dev) ||
 744             (port && port->state == BR_STATE_DISABLED))
 745                 goto out;
 746
 747         mp = br_multicast_new_group(br, port, group);
 748         err = PTR_ERR(mp);
 749         if (IS_ERR(mp))
 750                 goto err;
 751
 752         if (!port) {
 753                 mp->mglist = true;
 754                 mod_timer(&mp->timer, now + br->multicast_membership_interval);
 755                 goto out;
 756         }
 757
 758         for (pp = &mp->ports;
 759              (p = mlock_dereference(*pp, br)) != NULL;
 760              pp = &p->next) {
 761                 if (p->port == port)
 762                         goto found;
 763                 if ((unsigned long)p->port < (unsigned long)port)
 764                         break;
 765         }
 766
 767         p = br_multicast_new_port_group(port, group, *pp, 0);
 768         if (unlikely(!p))
 769                 goto err;
 770         rcu_assign_pointer(*pp, p);
 771         br_mdb_notify(br->dev, port, group, RTM_NEWMDB, 0);
 772
 773 found:
 774         mod_timer(&p->timer, now + br->multicast_membership_interval);
 775 out:
 776         err = 0;
 777
 778 err:
 779         spin_unlock(&br->multicast_lock);
 780         return err;
 781 }
 782
 783 static int br_ip4_multicast_add_group(struct net_bridge *br,
 784                                       struct net_bridge_port *port,
 785                                       __be32 group,
 786                                       __u16 vid)
 787 {
 788         struct br_ip br_group;
 789
 790         if (ipv4_is_local_multicast(group))
 791                 return 0;
 792
 793         br_group.u.ip4 = group;
 794         br_group.proto = htons(ETH_P_IP);
 795         br_group.vid = vid;
 796
 797         return br_multicast_add_group(br, port, &br_group);
 798 }
 799
 800 #if IS_ENABLED(CONFIG_IPV6)
 801 static int br_ip6_multicast_add_group(struct net_bridge *br,
 802                                       struct net_bridge_port *port,
 803                                       const struct in6_addr *group,
 804                                       __u16 vid)
 805 {
 806         struct br_ip br_group;
 807
 808         if (ipv6_addr_is_ll_all_nodes(group))
 809                 return 0;
 810
 811         br_group.u.ip6 = *group;
 812         br_group.proto = htons(ETH_P_IPV6);
 813         br_group.vid = vid;
 814
 815         return br_multicast_add_group(br, port, &br_group);
 816 }
 817 #endif
 818
 819 static void br_multicast_router_expired(unsigned long data)
 820 {
 821         struct net_bridge_port *port = (void *)data;
 822         struct net_bridge *br = port->br;
 823
 824         spin_lock(&br->multicast_lock);
 825         if (port->multicast_router == MDB_RTR_TYPE_DISABLED ||
 826             port->multicast_router == MDB_RTR_TYPE_PERM ||
 827             timer_pending(&port->multicast_router_timer) ||
 828             hlist_unhashed(&port->rlist))
 829                 goto out;
 830
 831         hlist_del_init_rcu(&port->rlist);
 832         br_rtr_notify(br->dev, port, RTM_DELMDB);
 833         /* Don't allow timer refresh if the router expired */
 834         if (port->multicast_router == MDB_RTR_TYPE_TEMP)
 835                 port->multicast_router = MDB_RTR_TYPE_TEMP_QUERY;
 836
 837 out:
 838         spin_unlock(&br->multicast_lock);
 839 }
 840
 841 static void br_multicast_local_router_expired(unsigned long data)
 842 {
 843 }
 844
 845 static void br_multicast_querier_expired(struct net_bridge *br,
 846                                          struct bridge_mcast_own_query *query)
 847 {
 848         spin_lock(&br->multicast_lock);
 849         if (!netif_running(br->dev) || br->multicast_disabled)
 850                 goto out;
 851
 852         br_multicast_start_querier(br, query);
 853
 854 out:
 855         spin_unlock(&br->multicast_lock);
 856 }
 857
 858 static void br_ip4_multicast_querier_expired(unsigned long data)
 859 {
 860         struct net_bridge *br = (void *)data;
 861
 862         br_multicast_querier_expired(br, &br->ip4_own_query);
 863 }
 864
 865 #if IS_ENABLED(CONFIG_IPV6)
 866 static void br_ip6_multicast_querier_expired(unsigned long data)
 867 {
 868         struct net_bridge *br = (void *)data;
 869
 870         br_multicast_querier_expired(br, &br->ip6_own_query);
 871 }
 872 #endif
 873
 874 static void br_multicast_select_own_querier(struct net_bridge *br,
 875                                             struct br_ip *ip,
 876                                             struct sk_buff *skb)
 877 {
 878         if (ip->proto == htons(ETH_P_IP))
 879                 br->ip4_querier.addr.u.ip4 = ip_hdr(skb)->saddr;
 880 #if IS_ENABLED(CONFIG_IPV6)
 881         else
 882                 br->ip6_querier.addr.u.ip6 = ipv6_hdr(skb)->saddr;
 883 #endif
 884 }
 885
 886 static void __br_multicast_send_query(struct net_bridge *br,
 887                                       struct net_bridge_port *port,
 888                                       struct br_ip *ip)
 889 {
 890         struct sk_buff *skb;
 891         u8 igmp_type;
 892
 893         skb = br_multicast_alloc_query(br, ip, &igmp_type);
 894         if (!skb)
 895                 return;
 896
 897         if (port) {
 898                 skb->dev = port->dev;
 899                 br_multicast_count(br, port, skb, igmp_type,
 900                                    BR_MCAST_DIR_TX);
 901                 NF_HOOK(NFPROTO_BRIDGE, NF_BR_LOCAL_OUT,
 902                         dev_net(port->dev), NULL, skb, NULL, skb->dev,
 903                         br_dev_queue_push_xmit);
 904         } else {
 905                 br_multicast_select_own_querier(br, ip, skb);
 906                 br_multicast_count(br, port, skb, igmp_type,
 907                                    BR_MCAST_DIR_RX);
 908                 netif_rx(skb);
 909         }
 910 }
 911
 912 static void br_multicast_send_query(struct net_bridge *br,
 913                                     struct net_bridge_port *port,
 914                                     struct bridge_mcast_own_query *own_query)
 915 {
 916         struct bridge_mcast_other_query *other_query = NULL;
 917         struct br_ip br_group;
 918         unsigned long time;
 919
 920         if (!netif_running(br->dev) || br->multicast_disabled ||
 921             !br->multicast_querier)
 922                 return;
 923
 924         memset(&br_group.u, 0, sizeof(br_group.u));
 925
 926         if (port ? (own_query == &port->ip4_own_query) :
 927                    (own_query == &br->ip4_own_query)) {
 928                 other_query = &br->ip4_other_query;
 929                 br_group.proto = htons(ETH_P_IP);
 930 #if IS_ENABLED(CONFIG_IPV6)
 931         } else {
 932                 other_query = &br->ip6_other_query;
 933                 br_group.proto = htons(ETH_P_IPV6);
 934 #endif
 935         }
 936
 937         if (!other_query || timer_pending(&other_query->timer))
 938                 return;
 939
 940         __br_multicast_send_query(br, port, &br_group);
 941
 942         time = jiffies;
 943         time += own_query->startup_sent < br->multicast_startup_query_count ?
 944                 br->multicast_startup_query_interval :
 945                 br->multicast_query_interval;
 946         mod_timer(&own_query->timer, time);
 947 }
 948
 949 static void
 950 br_multicast_port_query_expired(struct net_bridge_port *port,
 951                                 struct bridge_mcast_own_query *query)
 952 {
 953         struct net_bridge *br = port->br;
 954
 955         spin_lock(&br->multicast_lock);
 956         if (port->state == BR_STATE_DISABLED ||
 957             port->state == BR_STATE_BLOCKING)
 958                 goto out;
 959
 960         if (query->startup_sent < br->multicast_startup_query_count)
 961                 query->startup_sent++;
 962
 963         br_multicast_send_query(port->br, port, query);
 964
 965 out:
 966         spin_unlock(&br->multicast_lock);
 967 }
 968
 969 static void br_ip4_multicast_port_query_expired(unsigned long data)
 970 {
 971         struct net_bridge_port *port = (void *)data;
 972
 973         br_multicast_port_query_expired(port, &port->ip4_own_query);
 974 }
 975
 976 #if IS_ENABLED(CONFIG_IPV6)
 977 static void br_ip6_multicast_port_query_expired(unsigned long data)
 978 {
 979         struct net_bridge_port *port = (void *)data;
 980
 981         br_multicast_port_query_expired(port, &port->ip6_own_query);
 982 }
 983 #endif
 984
 985 int br_multicast_add_port(struct net_bridge_port *port)
 986 {
 987         port->multicast_router = MDB_RTR_TYPE_TEMP_QUERY;
 988
 989         setup_timer(&port->multicast_router_timer, br_multicast_router_expired,
 990                     (unsigned long)port);
 991         setup_timer(&port->ip4_own_query.timer,
 992                     br_ip4_multicast_port_query_expired, (unsigned long)port);
 993 #if IS_ENABLED(CONFIG_IPV6)
 994         setup_timer(&port->ip6_own_query.timer,
 995                     br_ip6_multicast_port_query_expired, (unsigned long)port);
 996 #endif
 997         port->mcast_stats = netdev_alloc_pcpu_stats(struct bridge_mcast_stats);
 998         if (!port->mcast_stats)
 999                 return -ENOMEM;
1000
1001         return 0;
1002 }
1003
1004 void br_multicast_del_port(struct net_bridge_port *port)
1005 {
1006         struct net_bridge *br = port->br;
1007         struct net_bridge_port_group *pg;
1008         struct hlist_node *n;
1009
1010         /* Take care of the remaining groups, only perm ones should be left */
1011         spin_lock_bh(&br->multicast_lock);
1012         hlist_for_each_entry_safe(pg, n, &port->mglist, mglist)
1013                 br_multicast_del_pg(br, pg);
1014         spin_unlock_bh(&br->multicast_lock);
1015         del_timer_sync(&port->multicast_router_timer);
1016         free_percpu(port->mcast_stats);
1017 }
1018
1019 static void br_multicast_enable(struct bridge_mcast_own_query *query)
1020 {
1021         query->startup_sent = 0;
1022
1023         if (try_to_del_timer_sync(&query->timer) >= 0 ||
1024             del_timer(&query->timer))
1025                 mod_timer(&query->timer, jiffies);
1026 }
1027
1028 static void __br_multicast_enable_port(struct net_bridge_port *port)
1029 {
1030         struct net_bridge *br = port->br;
1031
1032         if (br->multicast_disabled || !netif_running(br->dev))
1033                 return;
1034
1035         br_multicast_enable(&port->ip4_own_query);
1036 #if IS_ENABLED(CONFIG_IPV6)
1037         br_multicast_enable(&port->ip6_own_query);
1038 #endif
1039         if (port->multicast_router == MDB_RTR_TYPE_PERM &&
1040             hlist_unhashed(&port->rlist))
1041                 br_multicast_add_router(br, port);
1042 }
1043
1044 void br_multicast_enable_port(struct net_bridge_port *port)
1045 {
1046         struct net_bridge *br = port->br;
1047
1048         spin_lock(&br->multicast_lock);
1049         __br_multicast_enable_port(port);
1050         spin_unlock(&br->multicast_lock);
1051 }
1052
1053 void br_multicast_disable_port(struct net_bridge_port *port)
1054 {
1055         struct net_bridge *br = port->br;
1056         struct net_bridge_port_group *pg;
1057         struct hlist_node *n;
1058
1059         spin_lock(&br->multicast_lock);
1060         hlist_for_each_entry_safe(pg, n, &port->mglist, mglist)
1061                 if (!(pg->flags & MDB_PG_FLAGS_PERMANENT))
1062                         br_multicast_del_pg(br, pg);
1063
1064         if (!hlist_unhashed(&port->rlist)) {
1065                 hlist_del_init_rcu(&port->rlist);
1066                 br_rtr_notify(br->dev, port, RTM_DELMDB);
1067                 /* Don't allow timer refresh if disabling */
1068                 if (port->multicast_router == MDB_RTR_TYPE_TEMP)
1069                         port->multicast_router = MDB_RTR_TYPE_TEMP_QUERY;
1070         }
1071         del_timer(&port->multicast_router_timer);
1072         del_timer(&port->ip4_own_query.timer);
1073 #if IS_ENABLED(CONFIG_IPV6)
1074         del_timer(&port->ip6_own_query.timer);
1075 #endif
1076         spin_unlock(&br->multicast_lock);
1077 }
1078
1079 static int br_ip4_multicast_igmp3_report(struct net_bridge *br,
1080                                          struct net_bridge_port *port,
1081                                          struct sk_buff *skb,
1082                                          u16 vid)
1083 {
1084         struct igmpv3_report *ih;
1085         struct igmpv3_grec *grec;
1086         int i;
1087         int len;
1088         int num;
1089         int type;
1090         int err = 0;
1091         __be32 group;
1092
1093         ih = igmpv3_report_hdr(skb);
1094         num = ntohs(ih->ngrec);
1095         len = skb_transport_offset(skb) + sizeof(*ih);
1096
1097         for (i = 0; i < num; i++) {
1098                 len += sizeof(*grec);
1099                 if (!pskb_may_pull(skb, len))
1100                         return -EINVAL;
1101
1102                 grec = (void *)(skb->data + len - sizeof(*grec));
1103                 group = grec->grec_mca;
1104                 type = grec->grec_type;
1105
1106                 len += ntohs(grec->grec_nsrcs) * 4;
1107                 if (!pskb_may_pull(skb, len))
1108                         return -EINVAL;
1109
1110                 /* We treat this as an IGMPv2 report for now. */
1111                 switch (type) {
1112                 case IGMPV3_MODE_IS_INCLUDE:
1113                 case IGMPV3_MODE_IS_EXCLUDE:
1114                 case IGMPV3_CHANGE_TO_INCLUDE:
1115                 case IGMPV3_CHANGE_TO_EXCLUDE:
1116                 case IGMPV3_ALLOW_NEW_SOURCES:
1117                 case IGMPV3_BLOCK_OLD_SOURCES:
1118                         break;
1119
1120                 default:
1121                         continue;
1122                 }
1123
1124                 if ((type == IGMPV3_CHANGE_TO_INCLUDE ||
1125                      type == IGMPV3_MODE_IS_INCLUDE) &&
1126                     ntohs(grec->grec_nsrcs) == 0) {
1127                         br_ip4_multicast_leave_group(br, port, group, vid);
1128                 } else {
1129                         err = br_ip4_multicast_add_group(br, port, group, vid);
1130                         if (err)
1131                                 break;
1132                 }
1133         }
1134
1135         return err;
1136 }
1137
1138 #if IS_ENABLED(CONFIG_IPV6)
1139 static int br_ip6_multicast_mld2_report(struct net_bridge *br,
1140                                         struct net_bridge_port *port,
1141                                         struct sk_buff *skb,
1142                                         u16 vid)
1143 {
1144         struct icmp6hdr *icmp6h;
1145         struct mld2_grec *grec;
1146         int i;
1147         int len;
1148         int num;
1149         int err = 0;
1150
1151         if (!pskb_may_pull(skb, sizeof(*icmp6h)))
1152                 return -EINVAL;
1153
1154         icmp6h = icmp6_hdr(skb);
1155         num = ntohs(icmp6h->icmp6_dataun.un_data16[1]);
1156         len = skb_transport_offset(skb) + sizeof(*icmp6h);
1157
1158         for (i = 0; i < num; i++) {
1159                 __be16 *nsrcs, _nsrcs;
1160
1161                 nsrcs = skb_header_pointer(skb,
1162                                            len + offsetof(struct mld2_grec,
1163                                                           grec_nsrcs),
1164                                            sizeof(_nsrcs), &_nsrcs);
1165                 if (!nsrcs)
1166                         return -EINVAL;
1167
1168                 if (!pskb_may_pull(skb,
1169                                    len + sizeof(*grec) +
1170                                    sizeof(struct in6_addr) * ntohs(*nsrcs)))
1171                         return -EINVAL;
1172
1173                 grec = (struct mld2_grec *)(skb->data + len);
1174                 len += sizeof(*grec) +
1175                        sizeof(struct in6_addr) * ntohs(*nsrcs);
1176
1177                 /* We treat these as MLDv1 reports for now. */
1178                 switch (grec->grec_type) {
1179                 case MLD2_MODE_IS_INCLUDE:
1180                 case MLD2_MODE_IS_EXCLUDE:
1181                 case MLD2_CHANGE_TO_INCLUDE:
1182                 case MLD2_CHANGE_TO_EXCLUDE:
1183                 case MLD2_ALLOW_NEW_SOURCES:
1184                 case MLD2_BLOCK_OLD_SOURCES:
1185                         break;
1186
1187                 default:
1188                         continue;
1189                 }
1190
1191                 if ((grec->grec_type == MLD2_CHANGE_TO_INCLUDE ||
1192                      grec->grec_type == MLD2_MODE_IS_INCLUDE) &&
1193                     ntohs(*nsrcs) == 0) {
1194                         br_ip6_multicast_leave_group(br, port, &grec->grec_mca,
1195                                                      vid);
1196                 } else {
1197                         err = br_ip6_multicast_add_group(br, port,
1198                                                          &grec->grec_mca, vid);
1199                         if (err)
1200                                 break;
1201                 }
1202         }
1203
1204         return err;
1205 }
1206 #endif
1207
1208 static bool br_ip4_multicast_select_querier(struct net_bridge *br,
1209                                             struct net_bridge_port *port,
1210                                             __be32 saddr)
1211 {
1212         if (!timer_pending(&br->ip4_own_query.timer) &&
1213             !timer_pending(&br->ip4_other_query.timer))
1214                 goto update;
1215
1216         if (!br->ip4_querier.addr.u.ip4)
1217                 goto update;
1218
1219         if (ntohl(saddr) <= ntohl(br->ip4_querier.addr.u.ip4))
1220                 goto update;
1221
1222         return false;
1223
1224 update:
1225         br->ip4_querier.addr.u.ip4 = saddr;
1226
1227         /* update protected by general multicast_lock by caller */
1228         rcu_assign_pointer(br->ip4_querier.port, port);
1229
1230         return true;
1231 }
1232
1233 #if IS_ENABLED(CONFIG_IPV6)
1234 static bool br_ip6_multicast_select_querier(struct net_bridge *br,
1235                                             struct net_bridge_port *port,
1236                                             struct in6_addr *saddr)
1237 {
1238         if (!timer_pending(&br->ip6_own_query.timer) &&
1239             !timer_pending(&br->ip6_other_query.timer))
1240                 goto update;
1241
1242         if (ipv6_addr_cmp(saddr, &br->ip6_querier.addr.u.ip6) <= 0)
1243                 goto update;
1244
1245         return false;
1246
1247 update:
1248         br->ip6_querier.addr.u.ip6 = *saddr;
1249
1250         /* update protected by general multicast_lock by caller */
1251         rcu_assign_pointer(br->ip6_querier.port, port);
1252
1253         return true;
1254 }
1255 #endif
1256
1257 static bool br_multicast_select_querier(struct net_bridge *br,
1258                                         struct net_bridge_port *port,
1259                                         struct br_ip *saddr)
1260 {
1261         switch (saddr->proto) {
1262         case htons(ETH_P_IP):
1263                 return br_ip4_multicast_select_querier(br, port, saddr->u.ip4);
1264 #if IS_ENABLED(CONFIG_IPV6)
1265         case htons(ETH_P_IPV6):
1266                 return br_ip6_multicast_select_querier(br, port, &saddr->u.ip6);
1267 #endif
1268         }
1269
1270         return false;
1271 }
1272
1273 static void
1274 br_multicast_update_query_timer(struct net_bridge *br,
1275                                 struct bridge_mcast_other_query *query,
1276                                 unsigned long max_delay)
1277 {
1278         if (!timer_pending(&query->timer))
1279                 query->delay_time = jiffies + max_delay;
1280
1281         mod_timer(&query->timer, jiffies + br->multicast_querier_interval);
1282 }
1283
1284 /*
1285  * Add port to router_list
1286  *  list is maintained ordered by pointer value
1287  *  and locked by br->multicast_lock and RCU
1288  */
1289 static void br_multicast_add_router(struct net_bridge *br,
1290                                     struct net_bridge_port *port)
1291 {
1292         struct net_bridge_port *p;
1293         struct hlist_node *slot = NULL;
1294
1295         if (!hlist_unhashed(&port->rlist))
1296                 return;
1297
1298         hlist_for_each_entry(p, &br->router_list, rlist) {
1299                 if ((unsigned long) port >= (unsigned long) p)
1300                         break;
1301                 slot = &p->rlist;
1302         }
1303
1304         if (slot)
1305                 hlist_add_behind_rcu(&port->rlist, slot);
1306         else
1307                 hlist_add_head_rcu(&port->rlist, &br->router_list);
1308         br_rtr_notify(br->dev, port, RTM_NEWMDB);
1309 }
1310
1311 static void br_multicast_mark_router(struct net_bridge *br,
1312                                      struct net_bridge_port *port)
1313 {
1314         unsigned long now = jiffies;
1315
1316         if (!port) {
1317                 if (br->multicast_router == MDB_RTR_TYPE_TEMP_QUERY)
1318                         mod_timer(&br->multicast_router_timer,
1319                                   now + br->multicast_querier_interval);
1320                 return;
1321         }
1322
1323         if (port->multicast_router == MDB_RTR_TYPE_DISABLED ||
1324             port->multicast_router == MDB_RTR_TYPE_PERM)
1325                 return;
1326
1327         br_multicast_add_router(br, port);
1328
1329         mod_timer(&port->multicast_router_timer,
1330                   now + br->multicast_querier_interval);
1331 }
1332
1333 static void br_multicast_query_received(struct net_bridge *br,
1334                                         struct net_bridge_port *port,
1335                                         struct bridge_mcast_other_query *query,
1336                                         struct br_ip *saddr,
1337                                         unsigned long max_delay)
1338 {
1339         if (!br_multicast_select_querier(br, port, saddr))
1340                 return;
1341
1342         br_multicast_update_query_timer(br, query, max_delay);
1343         br_multicast_mark_router(br, port);
1344 }
1345
1346 static int br_ip4_multicast_query(struct net_bridge *br,
1347                                   struct net_bridge_port *port,
1348                                   struct sk_buff *skb,
1349                                   u16 vid)
1350 {
1351         const struct iphdr *iph = ip_hdr(skb);
1352         struct igmphdr *ih = igmp_hdr(skb);
1353         struct net_bridge_mdb_entry *mp;
1354         struct igmpv3_query *ih3;
1355         struct net_bridge_port_group *p;
1356         struct net_bridge_port_group __rcu **pp;
1357         struct br_ip saddr;
1358         unsigned long max_delay;
1359         unsigned long now = jiffies;
1360         unsigned int offset = skb_transport_offset(skb);
1361         __be32 group;
1362         int err = 0;
1363
1364         spin_lock(&br->multicast_lock);
1365         if (!netif_running(br->dev) ||
1366             (port && port->state == BR_STATE_DISABLED))
1367                 goto out;
1368
1369         group = ih->group;
1370
1371         if (skb->len == offset + sizeof(*ih)) {
1372                 max_delay = ih->code * (HZ / IGMP_TIMER_SCALE);
1373
1374                 if (!max_delay) {
1375                         max_delay = 10 * HZ;
1376                         group = 0;
1377                 }
1378         } else if (skb->len >= offset + sizeof(*ih3)) {
1379                 ih3 = igmpv3_query_hdr(skb);
1380                 if (ih3->nsrcs)
1381                         goto out;
1382
1383                 max_delay = ih3->code ?
1384                             IGMPV3_MRC(ih3->code) * (HZ / IGMP_TIMER_SCALE) : 1;
1385         } else {
1386                 goto out;
1387         }
1388
1389         if (!group) {
1390                 saddr.proto = htons(ETH_P_IP);
1391                 saddr.u.ip4 = iph->saddr;
1392
1393                 br_multicast_query_received(br, port, &br->ip4_other_query,
1394                                             &saddr, max_delay);
1395                 goto out;
1396         }
1397
1398         mp = br_mdb_ip4_get(mlock_dereference(br->mdb, br), group, vid);
1399         if (!mp)
1400                 goto out;
1401
1402         max_delay *= br->multicast_last_member_count;
1403
1404         if (mp->mglist &&
1405             (timer_pending(&mp->timer) ?
1406              time_after(mp->timer.expires, now + max_delay) :
1407              try_to_del_timer_sync(&mp->timer) >= 0))
1408                 mod_timer(&mp->timer, now + max_delay);
1409
1410         for (pp = &mp->ports;
1411              (p = mlock_dereference(*pp, br)) != NULL;
1412              pp = &p->next) {
1413                 if (timer_pending(&p->timer) ?
1414                     time_after(p->timer.expires, now + max_delay) :
1415                     try_to_del_timer_sync(&p->timer) >= 0)
1416                         mod_timer(&p->timer, now + max_delay);
1417         }
1418
1419 out:
1420         spin_unlock(&br->multicast_lock);
1421         return err;
1422 }
1423
1424 #if IS_ENABLED(CONFIG_IPV6)
1425 static int br_ip6_multicast_query(struct net_bridge *br,
1426                                   struct net_bridge_port *port,
1427                                   struct sk_buff *skb,
1428                                   u16 vid)
1429 {
1430         const struct ipv6hdr *ip6h = ipv6_hdr(skb);
1431         struct mld_msg *mld;
1432         struct net_bridge_mdb_entry *mp;
1433         struct mld2_query *mld2q;
1434         struct net_bridge_port_group *p;
1435         struct net_bridge_port_group __rcu **pp;
1436         struct br_ip saddr;
1437         unsigned long max_delay;
1438         unsigned long now = jiffies;
1439         unsigned int offset = skb_transport_offset(skb);
1440         const struct in6_addr *group = NULL;
1441         bool is_general_query;
1442         int err = 0;
1443
1444         spin_lock(&br->multicast_lock);
1445         if (!netif_running(br->dev) ||
1446             (port && port->state == BR_STATE_DISABLED))
1447                 goto out;
1448
1449         if (skb->len == offset + sizeof(*mld)) {
1450                 if (!pskb_may_pull(skb, offset + sizeof(*mld))) {
1451                         err = -EINVAL;
1452                         goto out;
1453                 }
1454                 mld = (struct mld_msg *) icmp6_hdr(skb);
1455                 max_delay = msecs_to_jiffies(ntohs(mld->mld_maxdelay));
1456                 if (max_delay)
1457                         group = &mld->mld_mca;
1458         } else {
1459                 if (!pskb_may_pull(skb, offset + sizeof(*mld2q))) {
1460                         err = -EINVAL;
1461                         goto out;
1462                 }
1463                 mld2q = (struct mld2_query *)icmp6_hdr(skb);
1464                 if (!mld2q->mld2q_nsrcs)
1465                         group = &mld2q->mld2q_mca;
1466
1467                 max_delay = max(msecs_to_jiffies(mldv2_mrc(mld2q)), 1UL);
1468         }
1469
1470         is_general_query = group && ipv6_addr_any(group);
1471
1472         if (is_general_query) {
1473                 saddr.proto = htons(ETH_P_IPV6);
1474                 saddr.u.ip6 = ip6h->saddr;
1475
1476                 br_multicast_query_received(br, port, &br->ip6_other_query,
1477                                             &saddr, max_delay);
1478                 goto out;
1479         } else if (!group) {
1480                 goto out;
1481         }
1482
1483         mp = br_mdb_ip6_get(mlock_dereference(br->mdb, br), group, vid);
1484         if (!mp)
1485                 goto out;
1486
1487         max_delay *= br->multicast_last_member_count;
1488         if (mp->mglist &&
1489             (timer_pending(&mp->timer) ?
1490              time_after(mp->timer.expires, now + max_delay) :
1491              try_to_del_timer_sync(&mp->timer) >= 0))
1492                 mod_timer(&mp->timer, now + max_delay);
1493
1494         for (pp = &mp->ports;
1495              (p = mlock_dereference(*pp, br)) != NULL;
1496              pp = &p->next) {
1497                 if (timer_pending(&p->timer) ?
1498                     time_after(p->timer.expires, now + max_delay) :
1499                     try_to_del_timer_sync(&p->timer) >= 0)
1500                         mod_timer(&p->timer, now + max_delay);
1501         }
1502
1503 out:
1504         spin_unlock(&br->multicast_lock);
1505         return err;
1506 }
1507 #endif
1508
1509 static void
1510 br_multicast_leave_group(struct net_bridge *br,
1511                          struct net_bridge_port *port,
1512                          struct br_ip *group,
1513                          struct bridge_mcast_other_query *other_query,
1514                          struct bridge_mcast_own_query *own_query)
1515 {
1516         struct net_bridge_mdb_htable *mdb;
1517         struct net_bridge_mdb_entry *mp;
1518         struct net_bridge_port_group *p;
1519         unsigned long now;
1520         unsigned long time;
1521
1522         spin_lock(&br->multicast_lock);
1523         if (!netif_running(br->dev) ||
1524             (port && port->state == BR_STATE_DISABLED))
1525                 goto out;
1526
1527         mdb = mlock_dereference(br->mdb, br);
1528         mp = br_mdb_ip_get(mdb, group);
1529         if (!mp)
1530                 goto out;
1531
1532         if (port && (port->flags & BR_MULTICAST_FAST_LEAVE)) {
1533                 struct net_bridge_port_group __rcu **pp;
1534
1535                 for (pp = &mp->ports;
1536                      (p = mlock_dereference(*pp, br)) != NULL;
1537                      pp = &p->next) {
1538                         if (p->port != port)
1539                                 continue;
1540
1541                         rcu_assign_pointer(*pp, p->next);
1542                         hlist_del_init(&p->mglist);
1543                         del_timer(&p->timer);
1544                         call_rcu_bh(&p->rcu, br_multicast_free_pg);
1545                         br_mdb_notify(br->dev, port, group, RTM_DELMDB,
1546                                       p->flags);
1547
1548                         if (!mp->ports && !mp->mglist &&
1549                             netif_running(br->dev))
1550                                 mod_timer(&mp->timer, jiffies);
1551                 }
1552                 goto out;
1553         }
1554
1555         if (timer_pending(&other_query->timer))
1556                 goto out;
1557
1558         if (br->multicast_querier) {
1559                 __br_multicast_send_query(br, port, &mp->addr);
1560
1561                 time = jiffies + br->multicast_last_member_count *
1562                                  br->multicast_last_member_interval;
1563
1564                 mod_timer(&own_query->timer, time);
1565
1566                 for (p = mlock_dereference(mp->ports, br);
1567                      p != NULL;
1568                      p = mlock_dereference(p->next, br)) {
1569                         if (p->port != port)
1570                                 continue;
1571
1572                         if (!hlist_unhashed(&p->mglist) &&
1573                             (timer_pending(&p->timer) ?
1574                              time_after(p->timer.expires, time) :
1575                              try_to_del_timer_sync(&p->timer) >= 0)) {
1576                                 mod_timer(&p->timer, time);
1577                         }
1578
1579                         break;
1580                 }
1581         }
1582
1583         now = jiffies;
1584         time = now + br->multicast_last_member_count *
1585                      br->multicast_last_member_interval;
1586
1587         if (!port) {
1588                 if (mp->mglist &&
1589                     (timer_pending(&mp->timer) ?
1590                      time_after(mp->timer.expires, time) :
1591                      try_to_del_timer_sync(&mp->timer) >= 0)) {
1592                         mod_timer(&mp->timer, time);
1593                 }
1594
1595                 goto out;
1596         }
1597
1598         for (p = mlock_dereference(mp->ports, br);
1599              p != NULL;
1600              p = mlock_dereference(p->next, br)) {
1601                 if (p->port != port)
1602                         continue;
1603
1604                 if (!hlist_unhashed(&p->mglist) &&
1605                     (timer_pending(&p->timer) ?
1606                      time_after(p->timer.expires, time) :
1607                      try_to_del_timer_sync(&p->timer) >= 0)) {
1608                         mod_timer(&p->timer, time);
1609                 }
1610
1611                 break;
1612         }
1613 out:
1614         spin_unlock(&br->multicast_lock);
1615 }
1616
1617 static void br_ip4_multicast_leave_group(struct net_bridge *br,
1618                                          struct net_bridge_port *port,
1619                                          __be32 group,
1620                                          __u16 vid)
1621 {
1622         struct br_ip br_group;
1623         struct bridge_mcast_own_query *own_query;
1624
1625         if (ipv4_is_local_multicast(group))
1626                 return;
1627
1628         own_query = port ? &port->ip4_own_query : &br->ip4_own_query;
1629
1630         br_group.u.ip4 = group;
1631         br_group.proto = htons(ETH_P_IP);
1632         br_group.vid = vid;
1633
1634         br_multicast_leave_group(br, port, &br_group, &br->ip4_other_query,
1635                                  own_query);
1636 }
1637
1638 #if IS_ENABLED(CONFIG_IPV6)
1639 static void br_ip6_multicast_leave_group(struct net_bridge *br,
1640                                          struct net_bridge_port *port,
1641                                          const struct in6_addr *group,
1642                                          __u16 vid)
1643 {
1644         struct br_ip br_group;
1645         struct bridge_mcast_own_query *own_query;
1646
1647         if (ipv6_addr_is_ll_all_nodes(group))
1648                 return;
1649
1650         own_query = port ? &port->ip6_own_query : &br->ip6_own_query;
1651
1652         br_group.u.ip6 = *group;
1653         br_group.proto = htons(ETH_P_IPV6);
1654         br_group.vid = vid;
1655
1656         br_multicast_leave_group(br, port, &br_group, &br->ip6_other_query,
1657                                  own_query);
1658 }
1659 #endif
1660
1661 static void br_multicast_err_count(const struct net_bridge *br,
1662                                    const struct net_bridge_port *p,
1663                                    __be16 proto)
1664 {
1665         struct bridge_mcast_stats __percpu *stats;
1666         struct bridge_mcast_stats *pstats;
1667
1668         if (!br->multicast_stats_enabled)
1669                 return;
1670
1671         if (p)
1672                 stats = p->mcast_stats;
1673         else
1674                 stats = br->mcast_stats;
1675         if (WARN_ON(!stats))
1676                 return;
1677
1678         pstats = this_cpu_ptr(stats);
1679
1680         u64_stats_update_begin(&pstats->syncp);
1681         switch (proto) {
1682         case htons(ETH_P_IP):
1683                 pstats->mstats.igmp_parse_errors++;
1684                 break;
1685 #if IS_ENABLED(CONFIG_IPV6)
1686         case htons(ETH_P_IPV6):
1687                 pstats->mstats.mld_parse_errors++;
1688                 break;
1689 #endif
1690         }
1691         u64_stats_update_end(&pstats->syncp);
1692 }
1693
1694 static void br_multicast_pim(struct net_bridge *br,
1695                              struct net_bridge_port *port,
1696                              const struct sk_buff *skb)
1697 {
1698         unsigned int offset = skb_transport_offset(skb);
1699         struct pimhdr *pimhdr, _pimhdr;
1700
1701         pimhdr = skb_header_pointer(skb, offset, sizeof(_pimhdr), &_pimhdr);
1702         if (!pimhdr || pim_hdr_version(pimhdr) != PIM_VERSION ||
1703             pim_hdr_type(pimhdr) != PIM_TYPE_HELLO)
1704                 return;
1705
1706         br_multicast_mark_router(br, port);
1707 }
1708
1709 static int br_multicast_ipv4_rcv(struct net_bridge *br,
1710                                  struct net_bridge_port *port,
1711                                  struct sk_buff *skb,
1712                                  u16 vid)
1713 {
1714         struct sk_buff *skb_trimmed = NULL;
1715         struct igmphdr *ih;
1716         int err;
1717
1718         err = ip_mc_check_igmp(skb, &skb_trimmed);
1719
1720         if (err == -ENOMSG) {
1721                 if (!ipv4_is_local_multicast(ip_hdr(skb)->daddr)) {
1722                         BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1723                 } else if (pim_ipv4_all_pim_routers(ip_hdr(skb)->daddr)) {
1724                         if (ip_hdr(skb)->protocol == IPPROTO_PIM)
1725                                 br_multicast_pim(br, port, skb);
1726                 }
1727                 return 0;
1728         } else if (err < 0) {
1729                 br_multicast_err_count(br, port, skb->protocol);
1730                 return err;
1731         }
1732
1733         ih = igmp_hdr(skb);
1734         BR_INPUT_SKB_CB(skb)->igmp = ih->type;
1735
1736         switch (ih->type) {
1737         case IGMP_HOST_MEMBERSHIP_REPORT:
1738         case IGMPV2_HOST_MEMBERSHIP_REPORT:
1739                 BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1740                 err = br_ip4_multicast_add_group(br, port, ih->group, vid);
1741                 break;
1742         case IGMPV3_HOST_MEMBERSHIP_REPORT:
1743                 err = br_ip4_multicast_igmp3_report(br, port, skb_trimmed, vid);
1744                 break;
1745         case IGMP_HOST_MEMBERSHIP_QUERY:
1746                 err = br_ip4_multicast_query(br, port, skb_trimmed, vid);
1747                 break;
1748         case IGMP_HOST_LEAVE_MESSAGE:
1749                 br_ip4_multicast_leave_group(br, port, ih->group, vid);
1750                 break;
1751         }
1752
1753         if (skb_trimmed && skb_trimmed != skb)
1754                 kfree_skb(skb_trimmed);
1755
1756         br_multicast_count(br, port, skb, BR_INPUT_SKB_CB(skb)->igmp,
1757                            BR_MCAST_DIR_RX);
1758
1759         return err;
1760 }
1761
1762 #if IS_ENABLED(CONFIG_IPV6)
1763 static int br_multicast_ipv6_rcv(struct net_bridge *br,
1764                                  struct net_bridge_port *port,
1765                                  struct sk_buff *skb,
1766                                  u16 vid)
1767 {
1768         struct sk_buff *skb_trimmed = NULL;
1769         struct mld_msg *mld;
1770         int err;
1771
1772         err = ipv6_mc_check_mld(skb, &skb_trimmed);
1773
1774         if (err == -ENOMSG) {
1775                 if (!ipv6_addr_is_ll_all_nodes(&ipv6_hdr(skb)->daddr))
1776                         BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1777                 return 0;
1778         } else if (err < 0) {
1779                 br_multicast_err_count(br, port, skb->protocol);
1780                 return err;
1781         }
1782
1783         mld = (struct mld_msg *)skb_transport_header(skb);
1784         BR_INPUT_SKB_CB(skb)->igmp = mld->mld_type;
1785
1786         switch (mld->mld_type) {
1787         case ICMPV6_MGM_REPORT:
1788                 BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1789                 err = br_ip6_multicast_add_group(br, port, &mld->mld_mca, vid);
1790                 break;
1791         case ICMPV6_MLD2_REPORT:
1792                 err = br_ip6_multicast_mld2_report(br, port, skb_trimmed, vid);
1793                 break;
1794         case ICMPV6_MGM_QUERY:
1795                 err = br_ip6_multicast_query(br, port, skb_trimmed, vid);
1796                 break;
1797         case ICMPV6_MGM_REDUCTION:
1798                 br_ip6_multicast_leave_group(br, port, &mld->mld_mca, vid);
1799                 break;
1800         }
1801
1802         if (skb_trimmed && skb_trimmed != skb)
1803                 kfree_skb(skb_trimmed);
1804
1805         br_multicast_count(br, port, skb, BR_INPUT_SKB_CB(skb)->igmp,
1806                            BR_MCAST_DIR_RX);
1807
1808         return err;
1809 }
1810 #endif
1811
1812 int br_multicast_rcv(struct net_bridge *br, struct net_bridge_port *port,
1813                      struct sk_buff *skb, u16 vid)
1814 {
1815         int ret = 0;
1816
1817         BR_INPUT_SKB_CB(skb)->igmp = 0;
1818         BR_INPUT_SKB_CB(skb)->mrouters_only = 0;
1819
1820         if (br->multicast_disabled)
1821                 return 0;
1822
1823         switch (skb->protocol) {
1824         case htons(ETH_P_IP):
1825                 ret = br_multicast_ipv4_rcv(br, port, skb, vid);
1826                 break;
1827 #if IS_ENABLED(CONFIG_IPV6)
1828         case htons(ETH_P_IPV6):
1829                 ret = br_multicast_ipv6_rcv(br, port, skb, vid);
1830                 break;
1831 #endif
1832         }
1833
1834         return ret;
1835 }
1836
1837 static void br_multicast_query_expired(struct net_bridge *br,
1838                                        struct bridge_mcast_own_query *query,
1839                                        struct bridge_mcast_querier *querier)
1840 {
1841         spin_lock(&br->multicast_lock);
1842         if (query->startup_sent < br->multicast_startup_query_count)
1843                 query->startup_sent++;
1844
1845         RCU_INIT_POINTER(querier->port, NULL);
1846         br_multicast_send_query(br, NULL, query);
1847         spin_unlock(&br->multicast_lock);
1848 }
1849
1850 static void br_ip4_multicast_query_expired(unsigned long data)
1851 {
1852         struct net_bridge *br = (void *)data;
1853
1854         br_multicast_query_expired(br, &br->ip4_own_query, &br->ip4_querier);
1855 }
1856
1857 #if IS_ENABLED(CONFIG_IPV6)
1858 static void br_ip6_multicast_query_expired(unsigned long data)
1859 {
1860         struct net_bridge *br = (void *)data;
1861
1862         br_multicast_query_expired(br, &br->ip6_own_query, &br->ip6_querier);
1863 }
1864 #endif
1865
1866 void br_multicast_init(struct net_bridge *br)
1867 {
1868         br->hash_elasticity = 4;
1869         br->hash_max = 512;
1870
1871         br->multicast_router = MDB_RTR_TYPE_TEMP_QUERY;
1872         br->multicast_querier = 0;
1873         br->multicast_query_use_ifaddr = 0;
1874         br->multicast_last_member_count = 2;
1875         br->multicast_startup_query_count = 2;
1876
1877         br->multicast_last_member_interval = HZ;
1878         br->multicast_query_response_interval = 10 * HZ;
1879         br->multicast_startup_query_interval = 125 * HZ / 4;
1880         br->multicast_query_interval = 125 * HZ;
1881         br->multicast_querier_interval = 255 * HZ;
1882         br->multicast_membership_interval = 260 * HZ;
1883
1884         br->ip4_other_query.delay_time = 0;
1885         br->ip4_querier.port = NULL;
1886         br->multicast_igmp_version = 2;
1887 #if IS_ENABLED(CONFIG_IPV6)
1888         br->multicast_mld_version = 1;
1889         br->ip6_other_query.delay_time = 0;
1890         br->ip6_querier.port = NULL;
1891 #endif
1892         br->has_ipv6_addr = 1;
1893
1894         spin_lock_init(&br->multicast_lock);
1895         setup_timer(&br->multicast_router_timer,
1896                     br_multicast_local_router_expired, 0);
1897         setup_timer(&br->ip4_other_query.timer,
1898                     br_ip4_multicast_querier_expired, (unsigned long)br);
1899         setup_timer(&br->ip4_own_query.timer, br_ip4_multicast_query_expired,
1900                     (unsigned long)br);
1901 #if IS_ENABLED(CONFIG_IPV6)
1902         setup_timer(&br->ip6_other_query.timer,
1903                     br_ip6_multicast_querier_expired, (unsigned long)br);
1904         setup_timer(&br->ip6_own_query.timer, br_ip6_multicast_query_expired,
1905                     (unsigned long)br);
1906 #endif
1907 }
1908
1909 static void __br_multicast_open(struct net_bridge *br,
1910                                 struct bridge_mcast_own_query *query)
1911 {
1912         query->startup_sent = 0;
1913
1914         if (br->multicast_disabled)
1915                 return;
1916
1917         mod_timer(&query->timer, jiffies);
1918 }
1919
1920 void br_multicast_open(struct net_bridge *br)
1921 {
1922         __br_multicast_open(br, &br->ip4_own_query);
1923 #if IS_ENABLED(CONFIG_IPV6)
1924         __br_multicast_open(br, &br->ip6_own_query);
1925 #endif
1926 }
1927
1928 void br_multicast_stop(struct net_bridge *br)
1929 {
1930         del_timer_sync(&br->multicast_router_timer);
1931         del_timer_sync(&br->ip4_other_query.timer);
1932         del_timer_sync(&br->ip4_own_query.timer);
1933 #if IS_ENABLED(CONFIG_IPV6)
1934         del_timer_sync(&br->ip6_other_query.timer);
1935         del_timer_sync(&br->ip6_own_query.timer);
1936 #endif
1937 }
1938
1939 void br_multicast_dev_del(struct net_bridge *br)
1940 {
1941         struct net_bridge_mdb_htable *mdb;
1942         struct net_bridge_mdb_entry *mp;
1943         struct hlist_node *n;
1944         u32 ver;
1945         int i;
1946
1947         spin_lock_bh(&br->multicast_lock);
1948         mdb = mlock_dereference(br->mdb, br);
1949         if (!mdb)
1950                 goto out;
1951
1952         br->mdb = NULL;
1953
1954         ver = mdb->ver;
1955         for (i = 0; i < mdb->max; i++) {
1956                 hlist_for_each_entry_safe(mp, n, &mdb->mhash[i],
1957                                           hlist[ver]) {
1958                         del_timer(&mp->timer);
1959                         call_rcu_bh(&mp->rcu, br_multicast_free_group);
1960                 }
1961         }
1962
1963         if (mdb->old) {
1964                 spin_unlock_bh(&br->multicast_lock);
1965                 rcu_barrier_bh();
1966                 spin_lock_bh(&br->multicast_lock);
1967                 WARN_ON(mdb->old);
1968         }
1969
1970         mdb->old = mdb;
1971         call_rcu_bh(&mdb->rcu, br_mdb_free);
1972
1973 out:
1974         spin_unlock_bh(&br->multicast_lock);
1975
1976         free_percpu(br->mcast_stats);
1977 }
1978
1979 int br_multicast_set_router(struct net_bridge *br, unsigned long val)
1980 {
1981         int err = -EINVAL;
1982
1983         spin_lock_bh(&br->multicast_lock);
1984
1985         switch (val) {
1986         case MDB_RTR_TYPE_DISABLED:
1987         case MDB_RTR_TYPE_PERM:
1988                 del_timer(&br->multicast_router_timer);
1989                 /* fall through */
1990         case MDB_RTR_TYPE_TEMP_QUERY:
1991                 br->multicast_router = val;
1992                 err = 0;
1993                 break;
1994         }
1995
1996         spin_unlock_bh(&br->multicast_lock);
1997
1998         return err;
1999 }
2000
2001 static void __del_port_router(struct net_bridge_port *p)
2002 {
2003         if (hlist_unhashed(&p->rlist))
2004                 return;
2005         hlist_del_init_rcu(&p->rlist);
2006         br_rtr_notify(p->br->dev, p, RTM_DELMDB);
2007 }
2008
2009 int br_multicast_set_port_router(struct net_bridge_port *p, unsigned long val)
2010 {
2011         struct net_bridge *br = p->br;
2012         unsigned long now = jiffies;
2013         int err = -EINVAL;
2014
2015         spin_lock(&br->multicast_lock);
2016         if (p->multicast_router == val) {
2017                 /* Refresh the temp router port timer */
2018                 if (p->multicast_router == MDB_RTR_TYPE_TEMP)
2019                         mod_timer(&p->multicast_router_timer,
2020                                   now + br->multicast_querier_interval);
2021                 err = 0;
2022                 goto unlock;
2023         }
2024         switch (val) {
2025         case MDB_RTR_TYPE_DISABLED:
2026                 p->multicast_router = MDB_RTR_TYPE_DISABLED;
2027                 __del_port_router(p);
2028                 del_timer(&p->multicast_router_timer);
2029                 break;
2030         case MDB_RTR_TYPE_TEMP_QUERY:
2031                 p->multicast_router = MDB_RTR_TYPE_TEMP_QUERY;
2032                 __del_port_router(p);
2033                 break;
2034         case MDB_RTR_TYPE_PERM:
2035                 p->multicast_router = MDB_RTR_TYPE_PERM;
2036                 del_timer(&p->multicast_router_timer);
2037                 br_multicast_add_router(br, p);
2038                 break;
2039         case MDB_RTR_TYPE_TEMP:
2040                 p->multicast_router = MDB_RTR_TYPE_TEMP;
2041                 br_multicast_mark_router(br, p);
2042                 break;
2043         default:
2044                 goto unlock;
2045         }
2046         err = 0;
2047 unlock:
2048         spin_unlock(&br->multicast_lock);
2049
2050         return err;
2051 }
2052
2053 static void br_multicast_start_querier(struct net_bridge *br,
2054                                        struct bridge_mcast_own_query *query)
2055 {
2056         struct net_bridge_port *port;
2057
2058         __br_multicast_open(br, query);
2059
2060         list_for_each_entry(port, &br->port_list, list) {
2061                 if (port->state == BR_STATE_DISABLED ||
2062                     port->state == BR_STATE_BLOCKING)
2063                         continue;
2064
2065                 if (query == &br->ip4_own_query)
2066                         br_multicast_enable(&port->ip4_own_query);
2067 #if IS_ENABLED(CONFIG_IPV6)
2068                 else
2069                         br_multicast_enable(&port->ip6_own_query);
2070 #endif
2071         }
2072 }
2073
2074 int br_multicast_toggle(struct net_bridge *br, unsigned long val)
2075 {
2076         struct net_bridge_mdb_htable *mdb;
2077         struct net_bridge_port *port;
2078         int err = 0;
2079
2080         spin_lock_bh(&br->multicast_lock);
2081         if (br->multicast_disabled == !val)
2082                 goto unlock;
2083
2084         br->multicast_disabled = !val;
2085         if (br->multicast_disabled)
2086                 goto unlock;
2087
2088         if (!netif_running(br->dev))
2089                 goto unlock;
2090
2091         mdb = mlock_dereference(br->mdb, br);
2092         if (mdb) {
2093                 if (mdb->old) {
2094                         err = -EEXIST;
2095 rollback:
2096                         br->multicast_disabled = !!val;
2097                         goto unlock;
2098                 }
2099
2100                 err = br_mdb_rehash(&br->mdb, mdb->max,
2101                                     br->hash_elasticity);
2102                 if (err)
2103                         goto rollback;
2104         }
2105
2106         br_multicast_open(br);
2107         list_for_each_entry(port, &br->port_list, list)
2108                 __br_multicast_enable_port(port);
2109
2110 unlock:
2111         spin_unlock_bh(&br->multicast_lock);
2112
2113         return err;
2114 }
2115
2116 int br_multicast_set_querier(struct net_bridge *br, unsigned long val)
2117 {
2118         unsigned long max_delay;
2119
2120         val = !!val;
2121
2122         spin_lock_bh(&br->multicast_lock);
2123         if (br->multicast_querier == val)
2124                 goto unlock;
2125
2126         br->multicast_querier = val;
2127         if (!val)
2128                 goto unlock;
2129
2130         max_delay = br->multicast_query_response_interval;
2131
2132         if (!timer_pending(&br->ip4_other_query.timer))
2133                 br->ip4_other_query.delay_time = jiffies + max_delay;
2134
2135         br_multicast_start_querier(br, &br->ip4_own_query);
2136
2137 #if IS_ENABLED(CONFIG_IPV6)
2138         if (!timer_pending(&br->ip6_other_query.timer))
2139                 br->ip6_other_query.delay_time = jiffies + max_delay;
2140
2141         br_multicast_start_querier(br, &br->ip6_own_query);
2142 #endif
2143
2144 unlock:
2145         spin_unlock_bh(&br->multicast_lock);
2146
2147         return 0;
2148 }
2149
2150 int br_multicast_set_hash_max(struct net_bridge *br, unsigned long val)
2151 {
2152         int err = -EINVAL;
2153         u32 old;
2154         struct net_bridge_mdb_htable *mdb;
2155
2156         spin_lock_bh(&br->multicast_lock);
2157         if (!is_power_of_2(val))
2158                 goto unlock;
2159
2160         mdb = mlock_dereference(br->mdb, br);
2161         if (mdb && val < mdb->size)
2162                 goto unlock;
2163
2164         err = 0;
2165
2166         old = br->hash_max;
2167         br->hash_max = val;
2168
2169         if (mdb) {
2170                 if (mdb->old) {
2171                         err = -EEXIST;
2172 rollback:
2173                         br->hash_max = old;
2174                         goto unlock;
2175                 }
2176
2177                 err = br_mdb_rehash(&br->mdb, br->hash_max,
2178                                     br->hash_elasticity);
2179                 if (err)
2180                         goto rollback;
2181         }
2182
2183 unlock:
2184         spin_unlock_bh(&br->multicast_lock);
2185
2186         return err;
2187 }
2188
2189 int br_multicast_set_igmp_version(struct net_bridge *br, unsigned long val)
2190 {
2191         /* Currently we support only version 2 and 3 */
2192         switch (val) {
2193         case 2:
2194         case 3:
2195                 break;
2196         default:
2197                 return -EINVAL;
2198         }
2199
2200         spin_lock_bh(&br->multicast_lock);
2201         br->multicast_igmp_version = val;
2202         spin_unlock_bh(&br->multicast_lock);
2203
2204         return 0;
2205 }
2206
2207 #if IS_ENABLED(CONFIG_IPV6)
2208 int br_multicast_set_mld_version(struct net_bridge *br, unsigned long val)
2209 {
2210         /* Currently we support version 1 and 2 */
2211         switch (val) {
2212         case 1:
2213         case 2:
2214                 break;
2215         default:
2216                 return -EINVAL;
2217         }
2218
2219         spin_lock_bh(&br->multicast_lock);
2220         br->multicast_mld_version = val;
2221         spin_unlock_bh(&br->multicast_lock);
2222
2223         return 0;
2224 }
2225 #endif
2226
2227 /**
2228  * br_multicast_list_adjacent - Returns snooped multicast addresses
2229  * @dev:        The bridge port adjacent to which to retrieve addresses
2230  * @br_ip_list: The list to store found, snooped multicast IP addresses in
2231  *
2232  * Creates a list of IP addresses (struct br_ip_list) sensed by the multicast
2233  * snooping feature on all bridge ports of dev's bridge device, excluding
2234  * the addresses from dev itself.
2235  *
2236  * Returns the number of items added to br_ip_list.
2237  *
2238  * Notes:
2239  * - br_ip_list needs to be initialized by caller
2240  * - br_ip_list might contain duplicates in the end
2241  *   (needs to be taken care of by caller)
2242  * - br_ip_list needs to be freed by caller
2243  */
2244 int br_multicast_list_adjacent(struct net_device *dev,
2245                                struct list_head *br_ip_list)
2246 {
2247         struct net_bridge *br;
2248         struct net_bridge_port *port;
2249         struct net_bridge_port_group *group;
2250         struct br_ip_list *entry;
2251         int count = 0;
2252
2253         rcu_read_lock();
2254         if (!br_ip_list || !br_port_exists(dev))
2255                 goto unlock;
2256
2257         port = br_port_get_rcu(dev);
2258         if (!port || !port->br)
2259                 goto unlock;
2260
2261         br = port->br;
2262
2263         list_for_each_entry_rcu(port, &br->port_list, list) {
2264                 if (!port->dev || port->dev == dev)
2265                         continue;
2266
2267                 hlist_for_each_entry_rcu(group, &port->mglist, mglist) {
2268                         entry = kmalloc(sizeof(*entry), GFP_ATOMIC);
2269                         if (!entry)
2270                                 goto unlock;
2271
2272                         entry->addr = group->addr;
2273                         list_add(&entry->list, br_ip_list);
2274                         count++;
2275                 }
2276         }
2277
2278 unlock:
2279         rcu_read_unlock();
2280         return count;
2281 }
2282 EXPORT_SYMBOL_GPL(br_multicast_list_adjacent);
2283
2284 /**
2285  * br_multicast_has_querier_anywhere - Checks for a querier on a bridge
2286  * @dev: The bridge port providing the bridge on which to check for a querier
2287  * @proto: The protocol family to check for: IGMP -> ETH_P_IP, MLD -> ETH_P_IPV6
2288  *
2289  * Checks whether the given interface has a bridge on top and if so returns
2290  * true if a valid querier exists anywhere on the bridged link layer.
2291  * Otherwise returns false.
2292  */
2293 bool br_multicast_has_querier_anywhere(struct net_device *dev, int proto)
2294 {
2295         struct net_bridge *br;
2296         struct net_bridge_port *port;
2297         struct ethhdr eth;
2298         bool ret = false;
2299
2300         rcu_read_lock();
2301         if (!br_port_exists(dev))
2302                 goto unlock;
2303
2304         port = br_port_get_rcu(dev);
2305         if (!port || !port->br)
2306                 goto unlock;
2307
2308         br = port->br;
2309
2310         memset(&eth, 0, sizeof(eth));
2311         eth.h_proto = htons(proto);
2312
2313         ret = br_multicast_querier_exists(br, &eth);
2314
2315 unlock:
2316         rcu_read_unlock();
2317         return ret;
2318 }
2319 EXPORT_SYMBOL_GPL(br_multicast_has_querier_anywhere);
2320
2321 /**
2322  * br_multicast_has_querier_adjacent - Checks for a querier behind a bridge port
2323  * @dev: The bridge port adjacent to which to check for a querier
2324  * @proto: The protocol family to check for: IGMP -> ETH_P_IP, MLD -> ETH_P_IPV6
2325  *
2326  * Checks whether the given interface has a bridge on top and if so returns
2327  * true if a selected querier is behind one of the other ports of this
2328  * bridge. Otherwise returns false.
2329  */
2330 bool br_multicast_has_querier_adjacent(struct net_device *dev, int proto)
2331 {
2332         struct net_bridge *br;
2333         struct net_bridge_port *port;
2334         bool ret = false;
2335
2336         rcu_read_lock();
2337         if (!br_port_exists(dev))
2338                 goto unlock;
2339
2340         port = br_port_get_rcu(dev);
2341         if (!port || !port->br)
2342                 goto unlock;
2343
2344         br = port->br;
2345
2346         switch (proto) {
2347         case ETH_P_IP:
2348                 if (!timer_pending(&br->ip4_other_query.timer) ||
2349                     rcu_dereference(br->ip4_querier.port) == port)
2350                         goto unlock;
2351                 break;
2352 #if IS_ENABLED(CONFIG_IPV6)
2353         case ETH_P_IPV6:
2354                 if (!timer_pending(&br->ip6_other_query.timer) ||
2355                     rcu_dereference(br->ip6_querier.port) == port)
2356                         goto unlock;
2357                 break;
2358 #endif
2359         default:
2360                 goto unlock;
2361         }
2362
2363         ret = true;
2364 unlock:
2365         rcu_read_unlock();
2366         return ret;
2367 }
2368 EXPORT_SYMBOL_GPL(br_multicast_has_querier_adjacent);
2369
2370 static void br_mcast_stats_add(struct bridge_mcast_stats __percpu *stats,
2371                                const struct sk_buff *skb, u8 type, u8 dir)
2372 {
2373         struct bridge_mcast_stats *pstats = this_cpu_ptr(stats);
2374         __be16 proto = skb->protocol;
2375         unsigned int t_len;
2376
2377         u64_stats_update_begin(&pstats->syncp);
2378         switch (proto) {
2379         case htons(ETH_P_IP):
2380                 t_len = ntohs(ip_hdr(skb)->tot_len) - ip_hdrlen(skb);
2381                 switch (type) {
2382                 case IGMP_HOST_MEMBERSHIP_REPORT:
2383                         pstats->mstats.igmp_v1reports[dir]++;
2384                         break;
2385                 case IGMPV2_HOST_MEMBERSHIP_REPORT:
2386                         pstats->mstats.igmp_v2reports[dir]++;
2387                         break;
2388                 case IGMPV3_HOST_MEMBERSHIP_REPORT:
2389                         pstats->mstats.igmp_v3reports[dir]++;
2390                         break;
2391                 case IGMP_HOST_MEMBERSHIP_QUERY:
2392                         if (t_len != sizeof(struct igmphdr)) {
2393                                 pstats->mstats.igmp_v3queries[dir]++;
2394                         } else {
2395                                 unsigned int offset = skb_transport_offset(skb);
2396                                 struct igmphdr *ih, _ihdr;
2397
2398                                 ih = skb_header_pointer(skb, offset,
2399                                                         sizeof(_ihdr), &_ihdr);
2400                                 if (!ih)
2401                                         break;
2402                                 if (!ih->code)
2403                                         pstats->mstats.igmp_v1queries[dir]++;
2404                                 else
2405                                         pstats->mstats.igmp_v2queries[dir]++;
2406                         }
2407                         break;
2408                 case IGMP_HOST_LEAVE_MESSAGE:
2409                         pstats->mstats.igmp_leaves[dir]++;
2410                         break;
2411                 }
2412                 break;
2413 #if IS_ENABLED(CONFIG_IPV6)
2414         case htons(ETH_P_IPV6):
2415                 t_len = ntohs(ipv6_hdr(skb)->payload_len) +
2416                         sizeof(struct ipv6hdr);
2417                 t_len -= skb_network_header_len(skb);
2418                 switch (type) {
2419                 case ICMPV6_MGM_REPORT:
2420                         pstats->mstats.mld_v1reports[dir]++;
2421                         break;
2422                 case ICMPV6_MLD2_REPORT:
2423                         pstats->mstats.mld_v2reports[dir]++;
2424                         break;
2425                 case ICMPV6_MGM_QUERY:
2426                         if (t_len != sizeof(struct mld_msg))
2427                                 pstats->mstats.mld_v2queries[dir]++;
2428                         else
2429                                 pstats->mstats.mld_v1queries[dir]++;
2430                         break;
2431                 case ICMPV6_MGM_REDUCTION:
2432                         pstats->mstats.mld_leaves[dir]++;
2433                         break;
2434                 }
2435                 break;
2436 #endif /* CONFIG_IPV6 */
2437         }
2438         u64_stats_update_end(&pstats->syncp);
2439 }
2440
2441 void br_multicast_count(struct net_bridge *br, const struct net_bridge_port *p,
2442                         const struct sk_buff *skb, u8 type, u8 dir)
2443 {
2444         struct bridge_mcast_stats __percpu *stats;
2445
2446         /* if multicast_disabled is true then igmp type can't be set */
2447         if (!type || !br->multicast_stats_enabled)
2448                 return;
2449
2450         if (p)
2451                 stats = p->mcast_stats;
2452         else
2453                 stats = br->mcast_stats;
2454         if (WARN_ON(!stats))
2455                 return;
2456
2457         br_mcast_stats_add(stats, skb, type, dir);
2458 }
2459
2460 int br_multicast_init_stats(struct net_bridge *br)
2461 {
2462         br->mcast_stats = netdev_alloc_pcpu_stats(struct bridge_mcast_stats);
2463         if (!br->mcast_stats)
2464                 return -ENOMEM;
2465
2466         return 0;
2467 }
2468
2469 static void mcast_stats_add_dir(u64 *dst, u64 *src)
2470 {
2471         dst[BR_MCAST_DIR_RX] += src[BR_MCAST_DIR_RX];
2472         dst[BR_MCAST_DIR_TX] += src[BR_MCAST_DIR_TX];
2473 }
2474
2475 void br_multicast_get_stats(const struct net_bridge *br,
2476                             const struct net_bridge_port *p,
2477                             struct br_mcast_stats *dest)
2478 {
2479         struct bridge_mcast_stats __percpu *stats;
2480         struct br_mcast_stats tdst;
2481         int i;
2482
2483         memset(dest, 0, sizeof(*dest));
2484         if (p)
2485                 stats = p->mcast_stats;
2486         else
2487                 stats = br->mcast_stats;
2488         if (WARN_ON(!stats))
2489                 return;
2490
2491         memset(&tdst, 0, sizeof(tdst));
2492         for_each_possible_cpu(i) {
2493                 struct bridge_mcast_stats *cpu_stats = per_cpu_ptr(stats, i);
2494                 struct br_mcast_stats temp;
2495                 unsigned int start;
2496
2497                 do {
2498                         start = u64_stats_fetch_begin_irq(&cpu_stats->syncp);
2499                         memcpy(&temp, &cpu_stats->mstats, sizeof(temp));
2500                 } while (u64_stats_fetch_retry_irq(&cpu_stats->syncp, start));
2501
2502                 mcast_stats_add_dir(tdst.igmp_v1queries, temp.igmp_v1queries);
2503                 mcast_stats_add_dir(tdst.igmp_v2queries, temp.igmp_v2queries);
2504                 mcast_stats_add_dir(tdst.igmp_v3queries, temp.igmp_v3queries);
2505                 mcast_stats_add_dir(tdst.igmp_leaves, temp.igmp_leaves);
2506                 mcast_stats_add_dir(tdst.igmp_v1reports, temp.igmp_v1reports);
2507                 mcast_stats_add_dir(tdst.igmp_v2reports, temp.igmp_v2reports);
2508                 mcast_stats_add_dir(tdst.igmp_v3reports, temp.igmp_v3reports);
2509                 tdst.igmp_parse_errors += temp.igmp_parse_errors;
2510
2511                 mcast_stats_add_dir(tdst.mld_v1queries, temp.mld_v1queries);
2512                 mcast_stats_add_dir(tdst.mld_v2queries, temp.mld_v2queries);
2513                 mcast_stats_add_dir(tdst.mld_leaves, temp.mld_leaves);
2514                 mcast_stats_add_dir(tdst.mld_v1reports, temp.mld_v1reports);
2515                 mcast_stats_add_dir(tdst.mld_v2reports, temp.mld_v2reports);
2516                 tdst.mld_parse_errors += temp.mld_parse_errors;
2517         }
2518         memcpy(dest, &tdst, sizeof(*dest));
2519 }