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 #include <net/switchdev.h>
  31 #if IS_ENABLED(CONFIG_IPV6)
  32 #include <net/ipv6.h>
  33 #include <net/mld.h>
  34 #include <net/ip6_checksum.h>
  35 #include <net/addrconf.h>
  36 #endif
  37
  38 #include "br_private.h"
  39
  40 static void br_multicast_start_querier(struct net_bridge *br,
  41                                        struct bridge_mcast_own_query *query);
  42 static void br_multicast_add_router(struct net_bridge *br,
  43                                     struct net_bridge_port *port);
  44 static void br_ip4_multicast_leave_group(struct net_bridge *br,
  45                                          struct net_bridge_port *port,
  46                                          __be32 group,
  47                                          __u16 vid,
  48                                          const unsigned char *src);
  49
  50 static void __del_port_router(struct net_bridge_port *p);
  51 #if IS_ENABLED(CONFIG_IPV6)
  52 static void br_ip6_multicast_leave_group(struct net_bridge *br,
  53                                          struct net_bridge_port *port,
  54                                          const struct in6_addr *group,
  55                                          __u16 vid, const unsigned char *src);
  56 #endif
  57 unsigned int br_mdb_rehash_seq;
  58
  59 static inline int br_ip_equal(const struct br_ip *a, const struct br_ip *b)
  60 {
  61         if (a->proto != b->proto)
  62                 return 0;
  63         if (a->vid != b->vid)
  64                 return 0;
  65         switch (a->proto) {
  66         case htons(ETH_P_IP):
  67                 return a->u.ip4 == b->u.ip4;
  68 #if IS_ENABLED(CONFIG_IPV6)
  69         case htons(ETH_P_IPV6):
  70                 return ipv6_addr_equal(&a->u.ip6, &b->u.ip6);
  71 #endif
  72         }
  73         return 0;
  74 }
  75
  76 static inline int __br_ip4_hash(struct net_bridge_mdb_htable *mdb, __be32 ip,
  77                                 __u16 vid)
  78 {
  79         return jhash_2words((__force u32)ip, vid, mdb->secret) & (mdb->max - 1);
  80 }
  81
  82 #if IS_ENABLED(CONFIG_IPV6)
  83 static inline int __br_ip6_hash(struct net_bridge_mdb_htable *mdb,
  84                                 const struct in6_addr *ip,
  85                                 __u16 vid)
  86 {
  87         return jhash_2words(ipv6_addr_hash(ip), vid,
  88                             mdb->secret) & (mdb->max - 1);
  89 }
  90 #endif
  91
  92 static inline int br_ip_hash(struct net_bridge_mdb_htable *mdb,
  93                              struct br_ip *ip)
  94 {
  95         switch (ip->proto) {
  96         case htons(ETH_P_IP):
  97                 return __br_ip4_hash(mdb, ip->u.ip4, ip->vid);
  98 #if IS_ENABLED(CONFIG_IPV6)
  99         case htons(ETH_P_IPV6):
 100                 return __br_ip6_hash(mdb, &ip->u.ip6, ip->vid);
 101 #endif
 102         }
 103         return 0;
 104 }
 105
 106 static struct net_bridge_mdb_entry *__br_mdb_ip_get(
 107         struct net_bridge_mdb_htable *mdb, struct br_ip *dst, int hash)
 108 {
 109         struct net_bridge_mdb_entry *mp;
 110
 111         hlist_for_each_entry_rcu(mp, &mdb->mhash[hash], hlist[mdb->ver]) {
 112                 if (br_ip_equal(&mp->addr, dst))
 113                         return mp;
 114         }
 115
 116         return NULL;
 117 }
 118
 119 struct net_bridge_mdb_entry *br_mdb_ip_get(struct net_bridge_mdb_htable *mdb,
 120                                            struct br_ip *dst)
 121 {
 122         if (!mdb)
 123                 return NULL;
 124
 125         return __br_mdb_ip_get(mdb, dst, br_ip_hash(mdb, dst));
 126 }
 127
 128 static struct net_bridge_mdb_entry *br_mdb_ip4_get(
 129         struct net_bridge_mdb_htable *mdb, __be32 dst, __u16 vid)
 130 {
 131         struct br_ip br_dst;
 132
 133         br_dst.u.ip4 = dst;
 134         br_dst.proto = htons(ETH_P_IP);
 135         br_dst.vid = vid;
 136
 137         return br_mdb_ip_get(mdb, &br_dst);
 138 }
 139
 140 #if IS_ENABLED(CONFIG_IPV6)
 141 static struct net_bridge_mdb_entry *br_mdb_ip6_get(
 142         struct net_bridge_mdb_htable *mdb, const struct in6_addr *dst,
 143         __u16 vid)
 144 {
 145         struct br_ip br_dst;
 146
 147         br_dst.u.ip6 = *dst;
 148         br_dst.proto = htons(ETH_P_IPV6);
 149         br_dst.vid = vid;
 150
 151         return br_mdb_ip_get(mdb, &br_dst);
 152 }
 153 #endif
 154
 155 struct net_bridge_mdb_entry *br_mdb_get(struct net_bridge *br,
 156                                         struct sk_buff *skb, u16 vid)
 157 {
 158         struct net_bridge_mdb_htable *mdb = rcu_dereference(br->mdb);
 159         struct br_ip ip;
 160
 161         if (br->multicast_disabled)
 162                 return NULL;
 163
 164         if (BR_INPUT_SKB_CB(skb)->igmp)
 165                 return NULL;
 166
 167         ip.proto = skb->protocol;
 168         ip.vid = vid;
 169
 170         switch (skb->protocol) {
 171         case htons(ETH_P_IP):
 172                 ip.u.ip4 = ip_hdr(skb)->daddr;
 173                 break;
 174 #if IS_ENABLED(CONFIG_IPV6)
 175         case htons(ETH_P_IPV6):
 176                 ip.u.ip6 = ipv6_hdr(skb)->daddr;
 177                 break;
 178 #endif
 179         default:
 180                 return NULL;
 181         }
 182
 183         return br_mdb_ip_get(mdb, &ip);
 184 }
 185
 186 static void br_mdb_free(struct rcu_head *head)
 187 {
 188         struct net_bridge_mdb_htable *mdb =
 189                 container_of(head, struct net_bridge_mdb_htable, rcu);
 190         struct net_bridge_mdb_htable *old = mdb->old;
 191
 192         mdb->old = NULL;
 193         kfree(old->mhash);
 194         kfree(old);
 195 }
 196
 197 static int br_mdb_copy(struct net_bridge_mdb_htable *new,
 198                        struct net_bridge_mdb_htable *old,
 199                        int elasticity)
 200 {
 201         struct net_bridge_mdb_entry *mp;
 202         int maxlen;
 203         int len;
 204         int i;
 205
 206         for (i = 0; i < old->max; i++)
 207                 hlist_for_each_entry(mp, &old->mhash[i], hlist[old->ver])
 208                         hlist_add_head(&mp->hlist[new->ver],
 209                                        &new->mhash[br_ip_hash(new, &mp->addr)]);
 210
 211         if (!elasticity)
 212                 return 0;
 213
 214         maxlen = 0;
 215         for (i = 0; i < new->max; i++) {
 216                 len = 0;
 217                 hlist_for_each_entry(mp, &new->mhash[i], hlist[new->ver])
 218                         len++;
 219                 if (len > maxlen)
 220                         maxlen = len;
 221         }
 222
 223         return maxlen > elasticity ? -EINVAL : 0;
 224 }
 225
 226 void br_multicast_free_pg(struct rcu_head *head)
 227 {
 228         struct net_bridge_port_group *p =
 229                 container_of(head, struct net_bridge_port_group, rcu);
 230
 231         kfree(p);
 232 }
 233
 234 static void br_multicast_free_group(struct rcu_head *head)
 235 {
 236         struct net_bridge_mdb_entry *mp =
 237                 container_of(head, struct net_bridge_mdb_entry, rcu);
 238
 239         kfree(mp);
 240 }
 241
 242 static void br_multicast_group_expired(unsigned long data)
 243 {
 244         struct net_bridge_mdb_entry *mp = (void *)data;
 245         struct net_bridge *br = mp->br;
 246         struct net_bridge_mdb_htable *mdb;
 247
 248         spin_lock(&br->multicast_lock);
 249         if (!netif_running(br->dev) || timer_pending(&mp->timer))
 250                 goto out;
 251
 252         mp->mglist = false;
 253
 254         if (mp->ports)
 255                 goto out;
 256
 257         mdb = mlock_dereference(br->mdb, br);
 258
 259         hlist_del_rcu(&mp->hlist[mdb->ver]);
 260         mdb->size--;
 261
 262         call_rcu_bh(&mp->rcu, br_multicast_free_group);
 263
 264 out:
 265         spin_unlock(&br->multicast_lock);
 266 }
 267
 268 static void br_multicast_del_pg(struct net_bridge *br,
 269                                 struct net_bridge_port_group *pg)
 270 {
 271         struct net_bridge_mdb_htable *mdb;
 272         struct net_bridge_mdb_entry *mp;
 273         struct net_bridge_port_group *p;
 274         struct net_bridge_port_group __rcu **pp;
 275
 276         mdb = mlock_dereference(br->mdb, br);
 277
 278         mp = br_mdb_ip_get(mdb, &pg->addr);
 279         if (WARN_ON(!mp))
 280                 return;
 281
 282         for (pp = &mp->ports;
 283              (p = mlock_dereference(*pp, br)) != NULL;
 284              pp = &p->next) {
 285                 if (p != pg)
 286                         continue;
 287
 288                 rcu_assign_pointer(*pp, p->next);
 289                 hlist_del_init(&p->mglist);
 290                 del_timer(&p->timer);
 291                 br_mdb_notify(br->dev, p->port, &pg->addr, RTM_DELMDB,
 292                               p->flags);
 293                 call_rcu_bh(&p->rcu, br_multicast_free_pg);
 294
 295                 if (!mp->ports && !mp->mglist &&
 296                     netif_running(br->dev))
 297                         mod_timer(&mp->timer, jiffies);
 298
 299                 return;
 300         }
 301
 302         WARN_ON(1);
 303 }
 304
 305 static void br_multicast_port_group_expired(unsigned long data)
 306 {
 307         struct net_bridge_port_group *pg = (void *)data;
 308         struct net_bridge *br = pg->port->br;
 309
 310         spin_lock(&br->multicast_lock);
 311         if (!netif_running(br->dev) || timer_pending(&pg->timer) ||
 312             hlist_unhashed(&pg->mglist) || pg->flags & MDB_PG_FLAGS_PERMANENT)
 313                 goto out;
 314
 315         br_multicast_del_pg(br, pg);
 316
 317 out:
 318         spin_unlock(&br->multicast_lock);
 319 }
 320
 321 static int br_mdb_rehash(struct net_bridge_mdb_htable __rcu **mdbp, int max,
 322                          int elasticity)
 323 {
 324         struct net_bridge_mdb_htable *old = rcu_dereference_protected(*mdbp, 1);
 325         struct net_bridge_mdb_htable *mdb;
 326         int err;
 327
 328         mdb = kmalloc(sizeof(*mdb), GFP_ATOMIC);
 329         if (!mdb)
 330                 return -ENOMEM;
 331
 332         mdb->max = max;
 333         mdb->old = old;
 334
 335         mdb->mhash = kzalloc(max * sizeof(*mdb->mhash), GFP_ATOMIC);
 336         if (!mdb->mhash) {
 337                 kfree(mdb);
 338                 return -ENOMEM;
 339         }
 340
 341         mdb->size = old ? old->size : 0;
 342         mdb->ver = old ? old->ver ^ 1 : 0;
 343
 344         if (!old || elasticity)
 345                 get_random_bytes(&mdb->secret, sizeof(mdb->secret));
 346         else
 347                 mdb->secret = old->secret;
 348
 349         if (!old)
 350                 goto out;
 351
 352         err = br_mdb_copy(mdb, old, elasticity);
 353         if (err) {
 354                 kfree(mdb->mhash);
 355                 kfree(mdb);
 356                 return err;
 357         }
 358
 359         br_mdb_rehash_seq++;
 360         call_rcu_bh(&mdb->rcu, br_mdb_free);
 361
 362 out:
 363         rcu_assign_pointer(*mdbp, mdb);
 364
 365         return 0;
 366 }
 367
 368 static struct sk_buff *br_ip4_multicast_alloc_query(struct net_bridge *br,
 369                                                     __be32 group,
 370                                                     u8 *igmp_type)
 371 {
 372         struct igmpv3_query *ihv3;
 373         size_t igmp_hdr_size;
 374         struct sk_buff *skb;
 375         struct igmphdr *ih;
 376         struct ethhdr *eth;
 377         struct iphdr *iph;
 378
 379         igmp_hdr_size = sizeof(*ih);
 380         if (br->multicast_igmp_version == 3)
 381                 igmp_hdr_size = sizeof(*ihv3);
 382         skb = netdev_alloc_skb_ip_align(br->dev, sizeof(*eth) + sizeof(*iph) +
 383                                                  igmp_hdr_size + 4);
 384         if (!skb)
 385                 goto out;
 386
 387         skb->protocol = htons(ETH_P_IP);
 388
 389         skb_reset_mac_header(skb);
 390         eth = eth_hdr(skb);
 391
 392         ether_addr_copy(eth->h_source, br->dev->dev_addr);
 393         eth->h_dest[0] = 1;
 394         eth->h_dest[1] = 0;
 395         eth->h_dest[2] = 0x5e;
 396         eth->h_dest[3] = 0;
 397         eth->h_dest[4] = 0;
 398         eth->h_dest[5] = 1;
 399         eth->h_proto = htons(ETH_P_IP);
 400         skb_put(skb, sizeof(*eth));
 401
 402         skb_set_network_header(skb, skb->len);
 403         iph = ip_hdr(skb);
 404
 405         iph->version = 4;
 406         iph->ihl = 6;
 407         iph->tos = 0xc0;
 408         iph->tot_len = htons(sizeof(*iph) + igmp_hdr_size + 4);
 409         iph->id = 0;
 410         iph->frag_off = htons(IP_DF);
 411         iph->ttl = 1;
 412         iph->protocol = IPPROTO_IGMP;
 413         iph->saddr = br->multicast_query_use_ifaddr ?
 414                      inet_select_addr(br->dev, 0, RT_SCOPE_LINK) : 0;
 415         iph->daddr = htonl(INADDR_ALLHOSTS_GROUP);
 416         ((u8 *)&iph[1])[0] = IPOPT_RA;
 417         ((u8 *)&iph[1])[1] = 4;
 418         ((u8 *)&iph[1])[2] = 0;
 419         ((u8 *)&iph[1])[3] = 0;
 420         ip_send_check(iph);
 421         skb_put(skb, 24);
 422
 423         skb_set_transport_header(skb, skb->len);
 424         *igmp_type = IGMP_HOST_MEMBERSHIP_QUERY;
 425
 426         switch (br->multicast_igmp_version) {
 427         case 2:
 428                 ih = igmp_hdr(skb);
 429                 ih->type = IGMP_HOST_MEMBERSHIP_QUERY;
 430                 ih->code = (group ? br->multicast_last_member_interval :
 431                                     br->multicast_query_response_interval) /
 432                            (HZ / IGMP_TIMER_SCALE);
 433                 ih->group = group;
 434                 ih->csum = 0;
 435                 ih->csum = ip_compute_csum((void *)ih, sizeof(*ih));
 436                 break;
 437         case 3:
 438                 ihv3 = igmpv3_query_hdr(skb);
 439                 ihv3->type = IGMP_HOST_MEMBERSHIP_QUERY;
 440                 ihv3->code = (group ? br->multicast_last_member_interval :
 441                                       br->multicast_query_response_interval) /
 442                              (HZ / IGMP_TIMER_SCALE);
 443                 ihv3->group = group;
 444                 ihv3->qqic = br->multicast_query_interval / HZ;
 445                 ihv3->nsrcs = 0;
 446                 ihv3->resv = 0;
 447                 ihv3->suppress = 0;
 448                 ihv3->qrv = 2;
 449                 ihv3->csum = 0;
 450                 ihv3->csum = ip_compute_csum((void *)ihv3, sizeof(*ihv3));
 451                 break;
 452         }
 453
 454         skb_put(skb, igmp_hdr_size);
 455         __skb_pull(skb, sizeof(*eth));
 456
 457 out:
 458         return skb;
 459 }
 460
 461 #if IS_ENABLED(CONFIG_IPV6)
 462 static struct sk_buff *br_ip6_multicast_alloc_query(struct net_bridge *br,
 463                                                     const struct in6_addr *grp,
 464                                                     u8 *igmp_type)
 465 {
 466         struct mld2_query *mld2q;
 467         unsigned long interval;
 468         struct ipv6hdr *ip6h;
 469         struct mld_msg *mldq;
 470         size_t mld_hdr_size;
 471         struct sk_buff *skb;
 472         struct ethhdr *eth;
 473         u8 *hopopt;
 474
 475         mld_hdr_size = sizeof(*mldq);
 476         if (br->multicast_mld_version == 2)
 477                 mld_hdr_size = sizeof(*mld2q);
 478         skb = netdev_alloc_skb_ip_align(br->dev, sizeof(*eth) + sizeof(*ip6h) +
 479                                                  8 + mld_hdr_size);
 480         if (!skb)
 481                 goto out;
 482
 483         skb->protocol = htons(ETH_P_IPV6);
 484
 485         /* Ethernet header */
 486         skb_reset_mac_header(skb);
 487         eth = eth_hdr(skb);
 488
 489         ether_addr_copy(eth->h_source, br->dev->dev_addr);
 490         eth->h_proto = htons(ETH_P_IPV6);
 491         skb_put(skb, sizeof(*eth));
 492
 493         /* IPv6 header + HbH option */
 494         skb_set_network_header(skb, skb->len);
 495         ip6h = ipv6_hdr(skb);
 496
 497         *(__force __be32 *)ip6h = htonl(0x60000000);
 498         ip6h->payload_len = htons(8 + mld_hdr_size);
 499         ip6h->nexthdr = IPPROTO_HOPOPTS;
 500         ip6h->hop_limit = 1;
 501         ipv6_addr_set(&ip6h->daddr, htonl(0xff020000), 0, 0, htonl(1));
 502         if (ipv6_dev_get_saddr(dev_net(br->dev), br->dev, &ip6h->daddr, 0,
 503                                &ip6h->saddr)) {
 504                 kfree_skb(skb);
 505                 br->has_ipv6_addr = 0;
 506                 return NULL;
 507         }
 508
 509         br->has_ipv6_addr = 1;
 510         ipv6_eth_mc_map(&ip6h->daddr, eth->h_dest);
 511
 512         hopopt = (u8 *)(ip6h + 1);
 513         hopopt[0] = IPPROTO_ICMPV6;             /* next hdr */
 514         hopopt[1] = 0;                          /* length of HbH */
 515         hopopt[2] = IPV6_TLV_ROUTERALERT;       /* Router Alert */
 516         hopopt[3] = 2;                          /* Length of RA Option */
 517         hopopt[4] = 0;                          /* Type = 0x0000 (MLD) */
 518         hopopt[5] = 0;
 519         hopopt[6] = IPV6_TLV_PAD1;              /* Pad1 */
 520         hopopt[7] = IPV6_TLV_PAD1;              /* Pad1 */
 521
 522         skb_put(skb, sizeof(*ip6h) + 8);
 523
 524         /* ICMPv6 */
 525         skb_set_transport_header(skb, skb->len);
 526         interval = ipv6_addr_any(grp) ?
 527                         br->multicast_query_response_interval :
 528                         br->multicast_last_member_interval;
 529         *igmp_type = ICMPV6_MGM_QUERY;
 530         switch (br->multicast_mld_version) {
 531         case 1:
 532                 mldq = (struct mld_msg *)icmp6_hdr(skb);
 533                 mldq->mld_type = ICMPV6_MGM_QUERY;
 534                 mldq->mld_code = 0;
 535                 mldq->mld_cksum = 0;
 536                 mldq->mld_maxdelay = htons((u16)jiffies_to_msecs(interval));
 537                 mldq->mld_reserved = 0;
 538                 mldq->mld_mca = *grp;
 539                 mldq->mld_cksum = csum_ipv6_magic(&ip6h->saddr, &ip6h->daddr,
 540                                                   sizeof(*mldq), IPPROTO_ICMPV6,
 541                                                   csum_partial(mldq,
 542                                                                sizeof(*mldq),
 543                                                                0));
 544                 break;
 545         case 2:
 546                 mld2q = (struct mld2_query *)icmp6_hdr(skb);
 547                 mld2q->mld2q_mrc = htons((u16)jiffies_to_msecs(interval));
 548                 mld2q->mld2q_type = ICMPV6_MGM_QUERY;
 549                 mld2q->mld2q_code = 0;
 550                 mld2q->mld2q_cksum = 0;
 551                 mld2q->mld2q_resv1 = 0;
 552                 mld2q->mld2q_resv2 = 0;
 553                 mld2q->mld2q_suppress = 0;
 554                 mld2q->mld2q_qrv = 2;
 555                 mld2q->mld2q_nsrcs = 0;
 556                 mld2q->mld2q_qqic = br->multicast_query_interval / HZ;
 557                 mld2q->mld2q_mca = *grp;
 558                 mld2q->mld2q_cksum = csum_ipv6_magic(&ip6h->saddr, &ip6h->daddr,
 559                                                      sizeof(*mld2q),
 560                                                      IPPROTO_ICMPV6,
 561                                                      csum_partial(mld2q,
 562                                                                   sizeof(*mld2q),
 563                                                                   0));
 564                 break;
 565         }
 566         skb_put(skb, mld_hdr_size);
 567
 568         __skb_pull(skb, sizeof(*eth));
 569
 570 out:
 571         return skb;
 572 }
 573 #endif
 574
 575 static struct sk_buff *br_multicast_alloc_query(struct net_bridge *br,
 576                                                 struct br_ip *addr,
 577                                                 u8 *igmp_type)
 578 {
 579         switch (addr->proto) {
 580         case htons(ETH_P_IP):
 581                 return br_ip4_multicast_alloc_query(br, addr->u.ip4, igmp_type);
 582 #if IS_ENABLED(CONFIG_IPV6)
 583         case htons(ETH_P_IPV6):
 584                 return br_ip6_multicast_alloc_query(br, &addr->u.ip6,
 585                                                     igmp_type);
 586 #endif
 587         }
 588         return NULL;
 589 }
 590
 591 static struct net_bridge_mdb_entry *br_multicast_get_group(
 592         struct net_bridge *br, struct net_bridge_port *port,
 593         struct br_ip *group, int hash)
 594 {
 595         struct net_bridge_mdb_htable *mdb;
 596         struct net_bridge_mdb_entry *mp;
 597         unsigned int count = 0;
 598         unsigned int max;
 599         int elasticity;
 600         int err;
 601
 602         mdb = rcu_dereference_protected(br->mdb, 1);
 603         hlist_for_each_entry(mp, &mdb->mhash[hash], hlist[mdb->ver]) {
 604                 count++;
 605                 if (unlikely(br_ip_equal(group, &mp->addr)))
 606                         return mp;
 607         }
 608
 609         elasticity = 0;
 610         max = mdb->max;
 611
 612         if (unlikely(count > br->hash_elasticity && count)) {
 613                 if (net_ratelimit())
 614                         br_info(br, "Multicast hash table "
 615                                 "chain limit reached: %s\n",
 616                                 port ? port->dev->name : br->dev->name);
 617
 618                 elasticity = br->hash_elasticity;
 619         }
 620
 621         if (mdb->size >= max) {
 622                 max *= 2;
 623                 if (unlikely(max > br->hash_max)) {
 624                         br_warn(br, "Multicast hash table maximum of %d "
 625                                 "reached, disabling snooping: %s\n",
 626                                 br->hash_max,
 627                                 port ? port->dev->name : br->dev->name);
 628                         err = -E2BIG;
 629 disable:
 630                         br->multicast_disabled = 1;
 631                         goto err;
 632                 }
 633         }
 634
 635         if (max > mdb->max || elasticity) {
 636                 if (mdb->old) {
 637                         if (net_ratelimit())
 638                                 br_info(br, "Multicast hash table "
 639                                         "on fire: %s\n",
 640                                         port ? port->dev->name : br->dev->name);
 641                         err = -EEXIST;
 642                         goto err;
 643                 }
 644
 645                 err = br_mdb_rehash(&br->mdb, max, elasticity);
 646                 if (err) {
 647                         br_warn(br, "Cannot rehash multicast "
 648                                 "hash table, disabling snooping: %s, %d, %d\n",
 649                                 port ? port->dev->name : br->dev->name,
 650                                 mdb->size, err);
 651                         goto disable;
 652                 }
 653
 654                 err = -EAGAIN;
 655                 goto err;
 656         }
 657
 658         return NULL;
 659
 660 err:
 661         mp = ERR_PTR(err);
 662         return mp;
 663 }
 664
 665 struct net_bridge_mdb_entry *br_multicast_new_group(struct net_bridge *br,
 666                                                     struct net_bridge_port *p,
 667                                                     struct br_ip *group)
 668 {
 669         struct net_bridge_mdb_htable *mdb;
 670         struct net_bridge_mdb_entry *mp;
 671         int hash;
 672         int err;
 673
 674         mdb = rcu_dereference_protected(br->mdb, 1);
 675         if (!mdb) {
 676                 err = br_mdb_rehash(&br->mdb, BR_HASH_SIZE, 0);
 677                 if (err)
 678                         return ERR_PTR(err);
 679                 goto rehash;
 680         }
 681
 682         hash = br_ip_hash(mdb, group);
 683         mp = br_multicast_get_group(br, p, group, hash);
 684         switch (PTR_ERR(mp)) {
 685         case 0:
 686                 break;
 687
 688         case -EAGAIN:
 689 rehash:
 690                 mdb = rcu_dereference_protected(br->mdb, 1);
 691                 hash = br_ip_hash(mdb, group);
 692                 break;
 693
 694         default:
 695                 goto out;
 696         }
 697
 698         mp = kzalloc(sizeof(*mp), GFP_ATOMIC);
 699         if (unlikely(!mp))
 700                 return ERR_PTR(-ENOMEM);
 701
 702         mp->br = br;
 703         mp->addr = *group;
 704         setup_timer(&mp->timer, br_multicast_group_expired,
 705                     (unsigned long)mp);
 706
 707         hlist_add_head_rcu(&mp->hlist[mdb->ver], &mdb->mhash[hash]);
 708         mdb->size++;
 709
 710 out:
 711         return mp;
 712 }
 713
 714 struct net_bridge_port_group *br_multicast_new_port_group(
 715                         struct net_bridge_port *port,
 716                         struct br_ip *group,
 717                         struct net_bridge_port_group __rcu *next,
 718                         unsigned char flags,
 719                         const unsigned char *src)
 720 {
 721         struct net_bridge_port_group *p;
 722
 723         p = kzalloc(sizeof(*p), GFP_ATOMIC);
 724         if (unlikely(!p))
 725                 return NULL;
 726
 727         p->addr = *group;
 728         p->port = port;
 729         p->flags = flags;
 730         rcu_assign_pointer(p->next, next);
 731         hlist_add_head(&p->mglist, &port->mglist);
 732         setup_timer(&p->timer, br_multicast_port_group_expired,
 733                     (unsigned long)p);
 734
 735         if (src)
 736                 memcpy(p->eth_addr, src, ETH_ALEN);
 737         else
 738                 memset(p->eth_addr, 0xff, ETH_ALEN);
 739
 740         return p;
 741 }
 742
 743 static bool br_port_group_equal(struct net_bridge_port_group *p,
 744                                 struct net_bridge_port *port,
 745                                 const unsigned char *src)
 746 {
 747         if (p->port != port)
 748                 return false;
 749
 750         if (!(port->flags & BR_MULTICAST_TO_UNICAST))
 751                 return true;
 752
 753         return ether_addr_equal(src, p->eth_addr);
 754 }
 755
 756 static int br_multicast_add_group(struct net_bridge *br,
 757                                   struct net_bridge_port *port,
 758                                   struct br_ip *group,
 759                                   const unsigned char *src)
 760 {
 761         struct net_bridge_port_group __rcu **pp;
 762         struct net_bridge_port_group *p;
 763         struct net_bridge_mdb_entry *mp;
 764         unsigned long now = jiffies;
 765         int err;
 766
 767         spin_lock(&br->multicast_lock);
 768         if (!netif_running(br->dev) ||
 769             (port && port->state == BR_STATE_DISABLED))
 770                 goto out;
 771
 772         mp = br_multicast_new_group(br, port, group);
 773         err = PTR_ERR(mp);
 774         if (IS_ERR(mp))
 775                 goto err;
 776
 777         if (!port) {
 778                 mp->mglist = true;
 779                 mod_timer(&mp->timer, now + br->multicast_membership_interval);
 780                 goto out;
 781         }
 782
 783         for (pp = &mp->ports;
 784              (p = mlock_dereference(*pp, br)) != NULL;
 785              pp = &p->next) {
 786                 if (br_port_group_equal(p, port, src))
 787                         goto found;
 788                 if ((unsigned long)p->port < (unsigned long)port)
 789                         break;
 790         }
 791
 792         p = br_multicast_new_port_group(port, group, *pp, 0, src);
 793         if (unlikely(!p))
 794                 goto err;
 795         rcu_assign_pointer(*pp, p);
 796         br_mdb_notify(br->dev, port, group, RTM_NEWMDB, 0);
 797
 798 found:
 799         mod_timer(&p->timer, now + br->multicast_membership_interval);
 800 out:
 801         err = 0;
 802
 803 err:
 804         spin_unlock(&br->multicast_lock);
 805         return err;
 806 }
 807
 808 static int br_ip4_multicast_add_group(struct net_bridge *br,
 809                                       struct net_bridge_port *port,
 810                                       __be32 group,
 811                                       __u16 vid,
 812                                       const unsigned char *src)
 813 {
 814         struct br_ip br_group;
 815
 816         if (ipv4_is_local_multicast(group))
 817                 return 0;
 818
 819         br_group.u.ip4 = group;
 820         br_group.proto = htons(ETH_P_IP);
 821         br_group.vid = vid;
 822
 823         return br_multicast_add_group(br, port, &br_group, src);
 824 }
 825
 826 #if IS_ENABLED(CONFIG_IPV6)
 827 static int br_ip6_multicast_add_group(struct net_bridge *br,
 828                                       struct net_bridge_port *port,
 829                                       const struct in6_addr *group,
 830                                       __u16 vid,
 831                                       const unsigned char *src)
 832 {
 833         struct br_ip br_group;
 834
 835         if (ipv6_addr_is_ll_all_nodes(group))
 836                 return 0;
 837
 838         br_group.u.ip6 = *group;
 839         br_group.proto = htons(ETH_P_IPV6);
 840         br_group.vid = vid;
 841
 842         return br_multicast_add_group(br, port, &br_group, src);
 843 }
 844 #endif
 845
 846 static void br_multicast_router_expired(unsigned long data)
 847 {
 848         struct net_bridge_port *port = (void *)data;
 849         struct net_bridge *br = port->br;
 850
 851         spin_lock(&br->multicast_lock);
 852         if (port->multicast_router == MDB_RTR_TYPE_DISABLED ||
 853             port->multicast_router == MDB_RTR_TYPE_PERM ||
 854             timer_pending(&port->multicast_router_timer))
 855                 goto out;
 856
 857         __del_port_router(port);
 858 out:
 859         spin_unlock(&br->multicast_lock);
 860 }
 861
 862 static void br_multicast_local_router_expired(unsigned long data)
 863 {
 864 }
 865
 866 static void br_multicast_querier_expired(struct net_bridge *br,
 867                                          struct bridge_mcast_own_query *query)
 868 {
 869         spin_lock(&br->multicast_lock);
 870         if (!netif_running(br->dev) || br->multicast_disabled)
 871                 goto out;
 872
 873         br_multicast_start_querier(br, query);
 874
 875 out:
 876         spin_unlock(&br->multicast_lock);
 877 }
 878
 879 static void br_ip4_multicast_querier_expired(unsigned long data)
 880 {
 881         struct net_bridge *br = (void *)data;
 882
 883         br_multicast_querier_expired(br, &br->ip4_own_query);
 884 }
 885
 886 #if IS_ENABLED(CONFIG_IPV6)
 887 static void br_ip6_multicast_querier_expired(unsigned long data)
 888 {
 889         struct net_bridge *br = (void *)data;
 890
 891         br_multicast_querier_expired(br, &br->ip6_own_query);
 892 }
 893 #endif
 894
 895 static void br_multicast_select_own_querier(struct net_bridge *br,
 896                                             struct br_ip *ip,
 897                                             struct sk_buff *skb)
 898 {
 899         if (ip->proto == htons(ETH_P_IP))
 900                 br->ip4_querier.addr.u.ip4 = ip_hdr(skb)->saddr;
 901 #if IS_ENABLED(CONFIG_IPV6)
 902         else
 903                 br->ip6_querier.addr.u.ip6 = ipv6_hdr(skb)->saddr;
 904 #endif
 905 }
 906
 907 static void __br_multicast_send_query(struct net_bridge *br,
 908                                       struct net_bridge_port *port,
 909                                       struct br_ip *ip)
 910 {
 911         struct sk_buff *skb;
 912         u8 igmp_type;
 913
 914         skb = br_multicast_alloc_query(br, ip, &igmp_type);
 915         if (!skb)
 916                 return;
 917
 918         if (port) {
 919                 skb->dev = port->dev;
 920                 br_multicast_count(br, port, skb, igmp_type,
 921                                    BR_MCAST_DIR_TX);
 922                 NF_HOOK(NFPROTO_BRIDGE, NF_BR_LOCAL_OUT,
 923                         dev_net(port->dev), NULL, skb, NULL, skb->dev,
 924                         br_dev_queue_push_xmit);
 925         } else {
 926                 br_multicast_select_own_querier(br, ip, skb);
 927                 br_multicast_count(br, port, skb, igmp_type,
 928                                    BR_MCAST_DIR_RX);
 929                 netif_rx(skb);
 930         }
 931 }
 932
 933 static void br_multicast_send_query(struct net_bridge *br,
 934                                     struct net_bridge_port *port,
 935                                     struct bridge_mcast_own_query *own_query)
 936 {
 937         struct bridge_mcast_other_query *other_query = NULL;
 938         struct br_ip br_group;
 939         unsigned long time;
 940
 941         if (!netif_running(br->dev) || br->multicast_disabled ||
 942             !br->multicast_querier)
 943                 return;
 944
 945         memset(&br_group.u, 0, sizeof(br_group.u));
 946
 947         if (port ? (own_query == &port->ip4_own_query) :
 948                    (own_query == &br->ip4_own_query)) {
 949                 other_query = &br->ip4_other_query;
 950                 br_group.proto = htons(ETH_P_IP);
 951 #if IS_ENABLED(CONFIG_IPV6)
 952         } else {
 953                 other_query = &br->ip6_other_query;
 954                 br_group.proto = htons(ETH_P_IPV6);
 955 #endif
 956         }
 957
 958         if (!other_query || timer_pending(&other_query->timer))
 959                 return;
 960
 961         __br_multicast_send_query(br, port, &br_group);
 962
 963         time = jiffies;
 964         time += own_query->startup_sent < br->multicast_startup_query_count ?
 965                 br->multicast_startup_query_interval :
 966                 br->multicast_query_interval;
 967         mod_timer(&own_query->timer, time);
 968 }
 969
 970 static void
 971 br_multicast_port_query_expired(struct net_bridge_port *port,
 972                                 struct bridge_mcast_own_query *query)
 973 {
 974         struct net_bridge *br = port->br;
 975
 976         spin_lock(&br->multicast_lock);
 977         if (port->state == BR_STATE_DISABLED ||
 978             port->state == BR_STATE_BLOCKING)
 979                 goto out;
 980
 981         if (query->startup_sent < br->multicast_startup_query_count)
 982                 query->startup_sent++;
 983
 984         br_multicast_send_query(port->br, port, query);
 985
 986 out:
 987         spin_unlock(&br->multicast_lock);
 988 }
 989
 990 static void br_ip4_multicast_port_query_expired(unsigned long data)
 991 {
 992         struct net_bridge_port *port = (void *)data;
 993
 994         br_multicast_port_query_expired(port, &port->ip4_own_query);
 995 }
 996
 997 #if IS_ENABLED(CONFIG_IPV6)
 998 static void br_ip6_multicast_port_query_expired(unsigned long data)
 999 {
1000         struct net_bridge_port *port = (void *)data;
1001
1002         br_multicast_port_query_expired(port, &port->ip6_own_query);
1003 }
1004 #endif
1005
1006 static void br_mc_disabled_update(struct net_device *dev, bool value)
1007 {
1008         struct switchdev_attr attr = {
1009                 .orig_dev = dev,
1010                 .id = SWITCHDEV_ATTR_ID_BRIDGE_MC_DISABLED,
1011                 .flags = SWITCHDEV_F_DEFER,
1012                 .u.mc_disabled = value,
1013         };
1014
1015         switchdev_port_attr_set(dev, &attr);
1016 }
1017
1018 int br_multicast_add_port(struct net_bridge_port *port)
1019 {
1020         port->multicast_router = MDB_RTR_TYPE_TEMP_QUERY;
1021
1022         setup_timer(&port->multicast_router_timer, br_multicast_router_expired,
1023                     (unsigned long)port);
1024         setup_timer(&port->ip4_own_query.timer,
1025                     br_ip4_multicast_port_query_expired, (unsigned long)port);
1026 #if IS_ENABLED(CONFIG_IPV6)
1027         setup_timer(&port->ip6_own_query.timer,
1028                     br_ip6_multicast_port_query_expired, (unsigned long)port);
1029 #endif
1030         br_mc_disabled_update(port->dev, port->br->multicast_disabled);
1031
1032         port->mcast_stats = netdev_alloc_pcpu_stats(struct bridge_mcast_stats);
1033         if (!port->mcast_stats)
1034                 return -ENOMEM;
1035
1036         return 0;
1037 }
1038
1039 void br_multicast_del_port(struct net_bridge_port *port)
1040 {
1041         struct net_bridge *br = port->br;
1042         struct net_bridge_port_group *pg;
1043         struct hlist_node *n;
1044
1045         /* Take care of the remaining groups, only perm ones should be left */
1046         spin_lock_bh(&br->multicast_lock);
1047         hlist_for_each_entry_safe(pg, n, &port->mglist, mglist)
1048                 br_multicast_del_pg(br, pg);
1049         spin_unlock_bh(&br->multicast_lock);
1050         del_timer_sync(&port->multicast_router_timer);
1051         free_percpu(port->mcast_stats);
1052 }
1053
1054 static void br_multicast_enable(struct bridge_mcast_own_query *query)
1055 {
1056         query->startup_sent = 0;
1057
1058         if (try_to_del_timer_sync(&query->timer) >= 0 ||
1059             del_timer(&query->timer))
1060                 mod_timer(&query->timer, jiffies);
1061 }
1062
1063 static void __br_multicast_enable_port(struct net_bridge_port *port)
1064 {
1065         struct net_bridge *br = port->br;
1066
1067         if (br->multicast_disabled || !netif_running(br->dev))
1068                 return;
1069
1070         br_multicast_enable(&port->ip4_own_query);
1071 #if IS_ENABLED(CONFIG_IPV6)
1072         br_multicast_enable(&port->ip6_own_query);
1073 #endif
1074         if (port->multicast_router == MDB_RTR_TYPE_PERM &&
1075             hlist_unhashed(&port->rlist))
1076                 br_multicast_add_router(br, port);
1077 }
1078
1079 void br_multicast_enable_port(struct net_bridge_port *port)
1080 {
1081         struct net_bridge *br = port->br;
1082
1083         spin_lock(&br->multicast_lock);
1084         __br_multicast_enable_port(port);
1085         spin_unlock(&br->multicast_lock);
1086 }
1087
1088 void br_multicast_disable_port(struct net_bridge_port *port)
1089 {
1090         struct net_bridge *br = port->br;
1091         struct net_bridge_port_group *pg;
1092         struct hlist_node *n;
1093
1094         spin_lock(&br->multicast_lock);
1095         hlist_for_each_entry_safe(pg, n, &port->mglist, mglist)
1096                 if (!(pg->flags & MDB_PG_FLAGS_PERMANENT))
1097                         br_multicast_del_pg(br, pg);
1098
1099         __del_port_router(port);
1100
1101         del_timer(&port->multicast_router_timer);
1102         del_timer(&port->ip4_own_query.timer);
1103 #if IS_ENABLED(CONFIG_IPV6)
1104         del_timer(&port->ip6_own_query.timer);
1105 #endif
1106         spin_unlock(&br->multicast_lock);
1107 }
1108
1109 static int br_ip4_multicast_igmp3_report(struct net_bridge *br,
1110                                          struct net_bridge_port *port,
1111                                          struct sk_buff *skb,
1112                                          u16 vid)
1113 {
1114         const unsigned char *src;
1115         struct igmpv3_report *ih;
1116         struct igmpv3_grec *grec;
1117         int i;
1118         int len;
1119         int num;
1120         int type;
1121         int err = 0;
1122         __be32 group;
1123         u16 nsrcs;
1124
1125         ih = igmpv3_report_hdr(skb);
1126         num = ntohs(ih->ngrec);
1127         len = skb_transport_offset(skb) + sizeof(*ih);
1128
1129         for (i = 0; i < num; i++) {
1130                 len += sizeof(*grec);
1131                 if (!pskb_may_pull(skb, len))
1132                         return -EINVAL;
1133
1134                 grec = (void *)(skb->data + len - sizeof(*grec));
1135                 group = grec->grec_mca;
1136                 type = grec->grec_type;
1137                 nsrcs = ntohs(grec->grec_nsrcs);
1138
1139                 len += nsrcs * 4;
1140                 if (!pskb_may_pull(skb, len))
1141                         return -EINVAL;
1142
1143                 /* We treat this as an IGMPv2 report for now. */
1144                 switch (type) {
1145                 case IGMPV3_MODE_IS_INCLUDE:
1146                 case IGMPV3_MODE_IS_EXCLUDE:
1147                 case IGMPV3_CHANGE_TO_INCLUDE:
1148                 case IGMPV3_CHANGE_TO_EXCLUDE:
1149                 case IGMPV3_ALLOW_NEW_SOURCES:
1150                 case IGMPV3_BLOCK_OLD_SOURCES:
1151                         break;
1152
1153                 default:
1154                         continue;
1155                 }
1156
1157                 src = eth_hdr(skb)->h_source;
1158                 if ((type == IGMPV3_CHANGE_TO_INCLUDE ||
1159                      type == IGMPV3_MODE_IS_INCLUDE) &&
1160                     nsrcs == 0) {
1161                         br_ip4_multicast_leave_group(br, port, group, vid, src);
1162                 } else {
1163                         err = br_ip4_multicast_add_group(br, port, group, vid,
1164                                                          src);
1165                         if (err)
1166                                 break;
1167                 }
1168         }
1169
1170         return err;
1171 }
1172
1173 #if IS_ENABLED(CONFIG_IPV6)
1174 static int br_ip6_multicast_mld2_report(struct net_bridge *br,
1175                                         struct net_bridge_port *port,
1176                                         struct sk_buff *skb,
1177                                         u16 vid)
1178 {
1179         const unsigned char *src;
1180         struct icmp6hdr *icmp6h;
1181         struct mld2_grec *grec;
1182         int i;
1183         int len;
1184         int num;
1185         int err = 0;
1186
1187         if (!pskb_may_pull(skb, sizeof(*icmp6h)))
1188                 return -EINVAL;
1189
1190         icmp6h = icmp6_hdr(skb);
1191         num = ntohs(icmp6h->icmp6_dataun.un_data16[1]);
1192         len = skb_transport_offset(skb) + sizeof(*icmp6h);
1193
1194         for (i = 0; i < num; i++) {
1195                 __be16 *_nsrcs, __nsrcs;
1196                 u16 nsrcs;
1197
1198                 _nsrcs = skb_header_pointer(skb,
1199                                             len + offsetof(struct mld2_grec,
1200                                                            grec_nsrcs),
1201                                             sizeof(__nsrcs), &__nsrcs);
1202                 if (!_nsrcs)
1203                         return -EINVAL;
1204
1205                 nsrcs = ntohs(*_nsrcs);
1206
1207                 if (!pskb_may_pull(skb,
1208                                    len + sizeof(*grec) +
1209                                    sizeof(struct in6_addr) * nsrcs))
1210                         return -EINVAL;
1211
1212                 grec = (struct mld2_grec *)(skb->data + len);
1213                 len += sizeof(*grec) +
1214                        sizeof(struct in6_addr) * nsrcs;
1215
1216                 /* We treat these as MLDv1 reports for now. */
1217                 switch (grec->grec_type) {
1218                 case MLD2_MODE_IS_INCLUDE:
1219                 case MLD2_MODE_IS_EXCLUDE:
1220                 case MLD2_CHANGE_TO_INCLUDE:
1221                 case MLD2_CHANGE_TO_EXCLUDE:
1222                 case MLD2_ALLOW_NEW_SOURCES:
1223                 case MLD2_BLOCK_OLD_SOURCES:
1224                         break;
1225
1226                 default:
1227                         continue;
1228                 }
1229
1230                 src = eth_hdr(skb)->h_source;
1231                 if ((grec->grec_type == MLD2_CHANGE_TO_INCLUDE ||
1232                      grec->grec_type == MLD2_MODE_IS_INCLUDE) &&
1233                     nsrcs == 0) {
1234                         br_ip6_multicast_leave_group(br, port, &grec->grec_mca,
1235                                                      vid, src);
1236                 } else {
1237                         err = br_ip6_multicast_add_group(br, port,
1238                                                          &grec->grec_mca, vid,
1239                                                          src);
1240                         if (err)
1241                                 break;
1242                 }
1243         }
1244
1245         return err;
1246 }
1247 #endif
1248
1249 static bool br_ip4_multicast_select_querier(struct net_bridge *br,
1250                                             struct net_bridge_port *port,
1251                                             __be32 saddr)
1252 {
1253         if (!timer_pending(&br->ip4_own_query.timer) &&
1254             !timer_pending(&br->ip4_other_query.timer))
1255                 goto update;
1256
1257         if (!br->ip4_querier.addr.u.ip4)
1258                 goto update;
1259
1260         if (ntohl(saddr) <= ntohl(br->ip4_querier.addr.u.ip4))
1261                 goto update;
1262
1263         return false;
1264
1265 update:
1266         br->ip4_querier.addr.u.ip4 = saddr;
1267
1268         /* update protected by general multicast_lock by caller */
1269         rcu_assign_pointer(br->ip4_querier.port, port);
1270
1271         return true;
1272 }
1273
1274 #if IS_ENABLED(CONFIG_IPV6)
1275 static bool br_ip6_multicast_select_querier(struct net_bridge *br,
1276                                             struct net_bridge_port *port,
1277                                             struct in6_addr *saddr)
1278 {
1279         if (!timer_pending(&br->ip6_own_query.timer) &&
1280             !timer_pending(&br->ip6_other_query.timer))
1281                 goto update;
1282
1283         if (ipv6_addr_cmp(saddr, &br->ip6_querier.addr.u.ip6) <= 0)
1284                 goto update;
1285
1286         return false;
1287
1288 update:
1289         br->ip6_querier.addr.u.ip6 = *saddr;
1290
1291         /* update protected by general multicast_lock by caller */
1292         rcu_assign_pointer(br->ip6_querier.port, port);
1293
1294         return true;
1295 }
1296 #endif
1297
1298 static bool br_multicast_select_querier(struct net_bridge *br,
1299                                         struct net_bridge_port *port,
1300                                         struct br_ip *saddr)
1301 {
1302         switch (saddr->proto) {
1303         case htons(ETH_P_IP):
1304                 return br_ip4_multicast_select_querier(br, port, saddr->u.ip4);
1305 #if IS_ENABLED(CONFIG_IPV6)
1306         case htons(ETH_P_IPV6):
1307                 return br_ip6_multicast_select_querier(br, port, &saddr->u.ip6);
1308 #endif
1309         }
1310
1311         return false;
1312 }
1313
1314 static void
1315 br_multicast_update_query_timer(struct net_bridge *br,
1316                                 struct bridge_mcast_other_query *query,
1317                                 unsigned long max_delay)
1318 {
1319         if (!timer_pending(&query->timer))
1320                 query->delay_time = jiffies + max_delay;
1321
1322         mod_timer(&query->timer, jiffies + br->multicast_querier_interval);
1323 }
1324
1325 static void br_port_mc_router_state_change(struct net_bridge_port *p,
1326                                            bool is_mc_router)
1327 {
1328         struct switchdev_attr attr = {
1329                 .orig_dev = p->dev,
1330                 .id = SWITCHDEV_ATTR_ID_PORT_MROUTER,
1331                 .flags = SWITCHDEV_F_DEFER,
1332                 .u.mrouter = is_mc_router,
1333         };
1334
1335         switchdev_port_attr_set(p->dev, &attr);
1336 }
1337
1338 /*
1339  * Add port to router_list
1340  *  list is maintained ordered by pointer value
1341  *  and locked by br->multicast_lock and RCU
1342  */
1343 static void br_multicast_add_router(struct net_bridge *br,
1344                                     struct net_bridge_port *port)
1345 {
1346         struct net_bridge_port *p;
1347         struct hlist_node *slot = NULL;
1348
1349         if (!hlist_unhashed(&port->rlist))
1350                 return;
1351
1352         hlist_for_each_entry(p, &br->router_list, rlist) {
1353                 if ((unsigned long) port >= (unsigned long) p)
1354                         break;
1355                 slot = &p->rlist;
1356         }
1357
1358         if (slot)
1359                 hlist_add_behind_rcu(&port->rlist, slot);
1360         else
1361                 hlist_add_head_rcu(&port->rlist, &br->router_list);
1362         br_rtr_notify(br->dev, port, RTM_NEWMDB);
1363         br_port_mc_router_state_change(port, true);
1364 }
1365
1366 static void br_multicast_mark_router(struct net_bridge *br,
1367                                      struct net_bridge_port *port)
1368 {
1369         unsigned long now = jiffies;
1370
1371         if (!port) {
1372                 if (br->multicast_router == MDB_RTR_TYPE_TEMP_QUERY)
1373                         mod_timer(&br->multicast_router_timer,
1374                                   now + br->multicast_querier_interval);
1375                 return;
1376         }
1377
1378         if (port->multicast_router == MDB_RTR_TYPE_DISABLED ||
1379             port->multicast_router == MDB_RTR_TYPE_PERM)
1380                 return;
1381
1382         br_multicast_add_router(br, port);
1383
1384         mod_timer(&port->multicast_router_timer,
1385                   now + br->multicast_querier_interval);
1386 }
1387
1388 static void br_multicast_query_received(struct net_bridge *br,
1389                                         struct net_bridge_port *port,
1390                                         struct bridge_mcast_other_query *query,
1391                                         struct br_ip *saddr,
1392                                         unsigned long max_delay)
1393 {
1394         if (!br_multicast_select_querier(br, port, saddr))
1395                 return;
1396
1397         br_multicast_update_query_timer(br, query, max_delay);
1398         br_multicast_mark_router(br, port);
1399 }
1400
1401 static int br_ip4_multicast_query(struct net_bridge *br,
1402                                   struct net_bridge_port *port,
1403                                   struct sk_buff *skb,
1404                                   u16 vid)
1405 {
1406         const struct iphdr *iph = ip_hdr(skb);
1407         struct igmphdr *ih = igmp_hdr(skb);
1408         struct net_bridge_mdb_entry *mp;
1409         struct igmpv3_query *ih3;
1410         struct net_bridge_port_group *p;
1411         struct net_bridge_port_group __rcu **pp;
1412         struct br_ip saddr;
1413         unsigned long max_delay;
1414         unsigned long now = jiffies;
1415         unsigned int offset = skb_transport_offset(skb);
1416         __be32 group;
1417         int err = 0;
1418
1419         spin_lock(&br->multicast_lock);
1420         if (!netif_running(br->dev) ||
1421             (port && port->state == BR_STATE_DISABLED))
1422                 goto out;
1423
1424         group = ih->group;
1425
1426         if (skb->len == offset + sizeof(*ih)) {
1427                 max_delay = ih->code * (HZ / IGMP_TIMER_SCALE);
1428
1429                 if (!max_delay) {
1430                         max_delay = 10 * HZ;
1431                         group = 0;
1432                 }
1433         } else if (skb->len >= offset + sizeof(*ih3)) {
1434                 ih3 = igmpv3_query_hdr(skb);
1435                 if (ih3->nsrcs)
1436                         goto out;
1437
1438                 max_delay = ih3->code ?
1439                             IGMPV3_MRC(ih3->code) * (HZ / IGMP_TIMER_SCALE) : 1;
1440         } else {
1441                 goto out;
1442         }
1443
1444         if (!group) {
1445                 saddr.proto = htons(ETH_P_IP);
1446                 saddr.u.ip4 = iph->saddr;
1447
1448                 br_multicast_query_received(br, port, &br->ip4_other_query,
1449                                             &saddr, max_delay);
1450                 goto out;
1451         }
1452
1453         mp = br_mdb_ip4_get(mlock_dereference(br->mdb, br), group, vid);
1454         if (!mp)
1455                 goto out;
1456
1457         max_delay *= br->multicast_last_member_count;
1458
1459         if (mp->mglist &&
1460             (timer_pending(&mp->timer) ?
1461              time_after(mp->timer.expires, now + max_delay) :
1462              try_to_del_timer_sync(&mp->timer) >= 0))
1463                 mod_timer(&mp->timer, now + max_delay);
1464
1465         for (pp = &mp->ports;
1466              (p = mlock_dereference(*pp, br)) != NULL;
1467              pp = &p->next) {
1468                 if (timer_pending(&p->timer) ?
1469                     time_after(p->timer.expires, now + max_delay) :
1470                     try_to_del_timer_sync(&p->timer) >= 0)
1471                         mod_timer(&p->timer, now + max_delay);
1472         }
1473
1474 out:
1475         spin_unlock(&br->multicast_lock);
1476         return err;
1477 }
1478
1479 #if IS_ENABLED(CONFIG_IPV6)
1480 static int br_ip6_multicast_query(struct net_bridge *br,
1481                                   struct net_bridge_port *port,
1482                                   struct sk_buff *skb,
1483                                   u16 vid)
1484 {
1485         struct mld_msg *mld;
1486         struct net_bridge_mdb_entry *mp;
1487         struct mld2_query *mld2q;
1488         struct net_bridge_port_group *p;
1489         struct net_bridge_port_group __rcu **pp;
1490         struct br_ip saddr;
1491         unsigned long max_delay;
1492         unsigned long now = jiffies;
1493         unsigned int offset = skb_transport_offset(skb);
1494         const struct in6_addr *group = NULL;
1495         bool is_general_query;
1496         int err = 0;
1497
1498         spin_lock(&br->multicast_lock);
1499         if (!netif_running(br->dev) ||
1500             (port && port->state == BR_STATE_DISABLED))
1501                 goto out;
1502
1503         if (skb->len == offset + sizeof(*mld)) {
1504                 if (!pskb_may_pull(skb, offset + sizeof(*mld))) {
1505                         err = -EINVAL;
1506                         goto out;
1507                 }
1508                 mld = (struct mld_msg *) icmp6_hdr(skb);
1509                 max_delay = msecs_to_jiffies(ntohs(mld->mld_maxdelay));
1510                 if (max_delay)
1511                         group = &mld->mld_mca;
1512         } else {
1513                 if (!pskb_may_pull(skb, offset + sizeof(*mld2q))) {
1514                         err = -EINVAL;
1515                         goto out;
1516                 }
1517                 mld2q = (struct mld2_query *)icmp6_hdr(skb);
1518                 if (!mld2q->mld2q_nsrcs)
1519                         group = &mld2q->mld2q_mca;
1520
1521                 max_delay = max(msecs_to_jiffies(mldv2_mrc(mld2q)), 1UL);
1522         }
1523
1524         is_general_query = group && ipv6_addr_any(group);
1525
1526         if (is_general_query) {
1527                 saddr.proto = htons(ETH_P_IPV6);
1528                 saddr.u.ip6 = ipv6_hdr(skb)->saddr;
1529
1530                 br_multicast_query_received(br, port, &br->ip6_other_query,
1531                                             &saddr, max_delay);
1532                 goto out;
1533         } else if (!group) {
1534                 goto out;
1535         }
1536
1537         mp = br_mdb_ip6_get(mlock_dereference(br->mdb, br), group, vid);
1538         if (!mp)
1539                 goto out;
1540
1541         max_delay *= br->multicast_last_member_count;
1542         if (mp->mglist &&
1543             (timer_pending(&mp->timer) ?
1544              time_after(mp->timer.expires, now + max_delay) :
1545              try_to_del_timer_sync(&mp->timer) >= 0))
1546                 mod_timer(&mp->timer, now + max_delay);
1547
1548         for (pp = &mp->ports;
1549              (p = mlock_dereference(*pp, br)) != NULL;
1550              pp = &p->next) {
1551                 if (timer_pending(&p->timer) ?
1552                     time_after(p->timer.expires, now + max_delay) :
1553                     try_to_del_timer_sync(&p->timer) >= 0)
1554                         mod_timer(&p->timer, now + max_delay);
1555         }
1556
1557 out:
1558         spin_unlock(&br->multicast_lock);
1559         return err;
1560 }
1561 #endif
1562
1563 static void
1564 br_multicast_leave_group(struct net_bridge *br,
1565                          struct net_bridge_port *port,
1566                          struct br_ip *group,
1567                          struct bridge_mcast_other_query *other_query,
1568                          struct bridge_mcast_own_query *own_query,
1569                          const unsigned char *src)
1570 {
1571         struct net_bridge_mdb_htable *mdb;
1572         struct net_bridge_mdb_entry *mp;
1573         struct net_bridge_port_group *p;
1574         unsigned long now;
1575         unsigned long time;
1576
1577         spin_lock(&br->multicast_lock);
1578         if (!netif_running(br->dev) ||
1579             (port && port->state == BR_STATE_DISABLED))
1580                 goto out;
1581
1582         mdb = mlock_dereference(br->mdb, br);
1583         mp = br_mdb_ip_get(mdb, group);
1584         if (!mp)
1585                 goto out;
1586
1587         if (port && (port->flags & BR_MULTICAST_FAST_LEAVE)) {
1588                 struct net_bridge_port_group __rcu **pp;
1589
1590                 for (pp = &mp->ports;
1591                      (p = mlock_dereference(*pp, br)) != NULL;
1592                      pp = &p->next) {
1593                         if (!br_port_group_equal(p, port, src))
1594                                 continue;
1595
1596                         if (p->flags & MDB_PG_FLAGS_PERMANENT)
1597                                 break;
1598
1599                         rcu_assign_pointer(*pp, p->next);
1600                         hlist_del_init(&p->mglist);
1601                         del_timer(&p->timer);
1602                         call_rcu_bh(&p->rcu, br_multicast_free_pg);
1603                         br_mdb_notify(br->dev, port, group, RTM_DELMDB,
1604                                       p->flags);
1605
1606                         if (!mp->ports && !mp->mglist &&
1607                             netif_running(br->dev))
1608                                 mod_timer(&mp->timer, jiffies);
1609                 }
1610                 goto out;
1611         }
1612
1613         if (timer_pending(&other_query->timer))
1614                 goto out;
1615
1616         if (br->multicast_querier) {
1617                 __br_multicast_send_query(br, port, &mp->addr);
1618
1619                 time = jiffies + br->multicast_last_member_count *
1620                                  br->multicast_last_member_interval;
1621
1622                 mod_timer(&own_query->timer, time);
1623
1624                 for (p = mlock_dereference(mp->ports, br);
1625                      p != NULL;
1626                      p = mlock_dereference(p->next, br)) {
1627                         if (!br_port_group_equal(p, port, src))
1628                                 continue;
1629
1630                         if (!hlist_unhashed(&p->mglist) &&
1631                             (timer_pending(&p->timer) ?
1632                              time_after(p->timer.expires, time) :
1633                              try_to_del_timer_sync(&p->timer) >= 0)) {
1634                                 mod_timer(&p->timer, time);
1635                         }
1636
1637                         break;
1638                 }
1639         }
1640
1641         now = jiffies;
1642         time = now + br->multicast_last_member_count *
1643                      br->multicast_last_member_interval;
1644
1645         if (!port) {
1646                 if (mp->mglist &&
1647                     (timer_pending(&mp->timer) ?
1648                      time_after(mp->timer.expires, time) :
1649                      try_to_del_timer_sync(&mp->timer) >= 0)) {
1650                         mod_timer(&mp->timer, time);
1651                 }
1652
1653                 goto out;
1654         }
1655
1656         for (p = mlock_dereference(mp->ports, br);
1657              p != NULL;
1658              p = mlock_dereference(p->next, br)) {
1659                 if (p->port != port)
1660                         continue;
1661
1662                 if (!hlist_unhashed(&p->mglist) &&
1663                     (timer_pending(&p->timer) ?
1664                      time_after(p->timer.expires, time) :
1665                      try_to_del_timer_sync(&p->timer) >= 0)) {
1666                         mod_timer(&p->timer, time);
1667                 }
1668
1669                 break;
1670         }
1671 out:
1672         spin_unlock(&br->multicast_lock);
1673 }
1674
1675 static void br_ip4_multicast_leave_group(struct net_bridge *br,
1676                                          struct net_bridge_port *port,
1677                                          __be32 group,
1678                                          __u16 vid,
1679                                          const unsigned char *src)
1680 {
1681         struct br_ip br_group;
1682         struct bridge_mcast_own_query *own_query;
1683
1684         if (ipv4_is_local_multicast(group))
1685                 return;
1686
1687         own_query = port ? &port->ip4_own_query : &br->ip4_own_query;
1688
1689         br_group.u.ip4 = group;
1690         br_group.proto = htons(ETH_P_IP);
1691         br_group.vid = vid;
1692
1693         br_multicast_leave_group(br, port, &br_group, &br->ip4_other_query,
1694                                  own_query, src);
1695 }
1696
1697 #if IS_ENABLED(CONFIG_IPV6)
1698 static void br_ip6_multicast_leave_group(struct net_bridge *br,
1699                                          struct net_bridge_port *port,
1700                                          const struct in6_addr *group,
1701                                          __u16 vid,
1702                                          const unsigned char *src)
1703 {
1704         struct br_ip br_group;
1705         struct bridge_mcast_own_query *own_query;
1706
1707         if (ipv6_addr_is_ll_all_nodes(group))
1708                 return;
1709
1710         own_query = port ? &port->ip6_own_query : &br->ip6_own_query;
1711
1712         br_group.u.ip6 = *group;
1713         br_group.proto = htons(ETH_P_IPV6);
1714         br_group.vid = vid;
1715
1716         br_multicast_leave_group(br, port, &br_group, &br->ip6_other_query,
1717                                  own_query, src);
1718 }
1719 #endif
1720
1721 static void br_multicast_err_count(const struct net_bridge *br,
1722                                    const struct net_bridge_port *p,
1723                                    __be16 proto)
1724 {
1725         struct bridge_mcast_stats __percpu *stats;
1726         struct bridge_mcast_stats *pstats;
1727
1728         if (!br->multicast_stats_enabled)
1729                 return;
1730
1731         if (p)
1732                 stats = p->mcast_stats;
1733         else
1734                 stats = br->mcast_stats;
1735         if (WARN_ON(!stats))
1736                 return;
1737
1738         pstats = this_cpu_ptr(stats);
1739
1740         u64_stats_update_begin(&pstats->syncp);
1741         switch (proto) {
1742         case htons(ETH_P_IP):
1743                 pstats->mstats.igmp_parse_errors++;
1744                 break;
1745 #if IS_ENABLED(CONFIG_IPV6)
1746         case htons(ETH_P_IPV6):
1747                 pstats->mstats.mld_parse_errors++;
1748                 break;
1749 #endif
1750         }
1751         u64_stats_update_end(&pstats->syncp);
1752 }
1753
1754 static void br_multicast_pim(struct net_bridge *br,
1755                              struct net_bridge_port *port,
1756                              const struct sk_buff *skb)
1757 {
1758         unsigned int offset = skb_transport_offset(skb);
1759         struct pimhdr *pimhdr, _pimhdr;
1760
1761         pimhdr = skb_header_pointer(skb, offset, sizeof(_pimhdr), &_pimhdr);
1762         if (!pimhdr || pim_hdr_version(pimhdr) != PIM_VERSION ||
1763             pim_hdr_type(pimhdr) != PIM_TYPE_HELLO)
1764                 return;
1765
1766         spin_lock(&br->multicast_lock);
1767         br_multicast_mark_router(br, port);
1768         spin_unlock(&br->multicast_lock);
1769 }
1770
1771 static int br_multicast_ipv4_rcv(struct net_bridge *br,
1772                                  struct net_bridge_port *port,
1773                                  struct sk_buff *skb,
1774                                  u16 vid)
1775 {
1776         struct sk_buff *skb_trimmed = NULL;
1777         const unsigned char *src;
1778         struct igmphdr *ih;
1779         int err;
1780
1781         err = ip_mc_check_igmp(skb, &skb_trimmed);
1782
1783         if (err == -ENOMSG) {
1784                 if (!ipv4_is_local_multicast(ip_hdr(skb)->daddr)) {
1785                         BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1786                 } else if (pim_ipv4_all_pim_routers(ip_hdr(skb)->daddr)) {
1787                         if (ip_hdr(skb)->protocol == IPPROTO_PIM)
1788                                 br_multicast_pim(br, port, skb);
1789                 }
1790                 return 0;
1791         } else if (err < 0) {
1792                 br_multicast_err_count(br, port, skb->protocol);
1793                 return err;
1794         }
1795
1796         ih = igmp_hdr(skb);
1797         src = eth_hdr(skb)->h_source;
1798         BR_INPUT_SKB_CB(skb)->igmp = ih->type;
1799
1800         switch (ih->type) {
1801         case IGMP_HOST_MEMBERSHIP_REPORT:
1802         case IGMPV2_HOST_MEMBERSHIP_REPORT:
1803                 BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1804                 err = br_ip4_multicast_add_group(br, port, ih->group, vid, src);
1805                 break;
1806         case IGMPV3_HOST_MEMBERSHIP_REPORT:
1807                 err = br_ip4_multicast_igmp3_report(br, port, skb_trimmed, vid);
1808                 break;
1809         case IGMP_HOST_MEMBERSHIP_QUERY:
1810                 err = br_ip4_multicast_query(br, port, skb_trimmed, vid);
1811                 break;
1812         case IGMP_HOST_LEAVE_MESSAGE:
1813                 br_ip4_multicast_leave_group(br, port, ih->group, vid, src);
1814                 break;
1815         }
1816
1817         if (skb_trimmed && skb_trimmed != skb)
1818                 kfree_skb(skb_trimmed);
1819
1820         br_multicast_count(br, port, skb, BR_INPUT_SKB_CB(skb)->igmp,
1821                            BR_MCAST_DIR_RX);
1822
1823         return err;
1824 }
1825
1826 #if IS_ENABLED(CONFIG_IPV6)
1827 static int br_multicast_ipv6_rcv(struct net_bridge *br,
1828                                  struct net_bridge_port *port,
1829                                  struct sk_buff *skb,
1830                                  u16 vid)
1831 {
1832         struct sk_buff *skb_trimmed = NULL;
1833         const unsigned char *src;
1834         struct mld_msg *mld;
1835         int err;
1836
1837         err = ipv6_mc_check_mld(skb, &skb_trimmed);
1838
1839         if (err == -ENOMSG) {
1840                 if (!ipv6_addr_is_ll_all_nodes(&ipv6_hdr(skb)->daddr))
1841                         BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1842                 return 0;
1843         } else if (err < 0) {
1844                 br_multicast_err_count(br, port, skb->protocol);
1845                 return err;
1846         }
1847
1848         mld = (struct mld_msg *)skb_transport_header(skb);
1849         BR_INPUT_SKB_CB(skb)->igmp = mld->mld_type;
1850
1851         switch (mld->mld_type) {
1852         case ICMPV6_MGM_REPORT:
1853                 src = eth_hdr(skb)->h_source;
1854                 BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
1855                 err = br_ip6_multicast_add_group(br, port, &mld->mld_mca, vid,
1856                                                  src);
1857                 break;
1858         case ICMPV6_MLD2_REPORT:
1859                 err = br_ip6_multicast_mld2_report(br, port, skb_trimmed, vid);
1860                 break;
1861         case ICMPV6_MGM_QUERY:
1862                 err = br_ip6_multicast_query(br, port, skb_trimmed, vid);
1863                 break;
1864         case ICMPV6_MGM_REDUCTION:
1865                 src = eth_hdr(skb)->h_source;
1866                 br_ip6_multicast_leave_group(br, port, &mld->mld_mca, vid, src);
1867                 break;
1868         }
1869
1870         if (skb_trimmed && skb_trimmed != skb)
1871                 kfree_skb(skb_trimmed);
1872
1873         br_multicast_count(br, port, skb, BR_INPUT_SKB_CB(skb)->igmp,
1874                            BR_MCAST_DIR_RX);
1875
1876         return err;
1877 }
1878 #endif
1879
1880 int br_multicast_rcv(struct net_bridge *br, struct net_bridge_port *port,
1881                      struct sk_buff *skb, u16 vid)
1882 {
1883         int ret = 0;
1884
1885         BR_INPUT_SKB_CB(skb)->igmp = 0;
1886         BR_INPUT_SKB_CB(skb)->mrouters_only = 0;
1887
1888         if (br->multicast_disabled)
1889                 return 0;
1890
1891         switch (skb->protocol) {
1892         case htons(ETH_P_IP):
1893                 ret = br_multicast_ipv4_rcv(br, port, skb, vid);
1894                 break;
1895 #if IS_ENABLED(CONFIG_IPV6)
1896         case htons(ETH_P_IPV6):
1897                 ret = br_multicast_ipv6_rcv(br, port, skb, vid);
1898                 break;
1899 #endif
1900         }
1901
1902         return ret;
1903 }
1904
1905 static void br_multicast_query_expired(struct net_bridge *br,
1906                                        struct bridge_mcast_own_query *query,
1907                                        struct bridge_mcast_querier *querier)
1908 {
1909         spin_lock(&br->multicast_lock);
1910         if (query->startup_sent < br->multicast_startup_query_count)
1911                 query->startup_sent++;
1912
1913         RCU_INIT_POINTER(querier->port, NULL);
1914         br_multicast_send_query(br, NULL, query);
1915         spin_unlock(&br->multicast_lock);
1916 }
1917
1918 static void br_ip4_multicast_query_expired(unsigned long data)
1919 {
1920         struct net_bridge *br = (void *)data;
1921
1922         br_multicast_query_expired(br, &br->ip4_own_query, &br->ip4_querier);
1923 }
1924
1925 #if IS_ENABLED(CONFIG_IPV6)
1926 static void br_ip6_multicast_query_expired(unsigned long data)
1927 {
1928         struct net_bridge *br = (void *)data;
1929
1930         br_multicast_query_expired(br, &br->ip6_own_query, &br->ip6_querier);
1931 }
1932 #endif
1933
1934 void br_multicast_init(struct net_bridge *br)
1935 {
1936         br->hash_elasticity = 4;
1937         br->hash_max = 512;
1938
1939         br->multicast_router = MDB_RTR_TYPE_TEMP_QUERY;
1940         br->multicast_querier = 0;
1941         br->multicast_query_use_ifaddr = 0;
1942         br->multicast_last_member_count = 2;
1943         br->multicast_startup_query_count = 2;
1944
1945         br->multicast_last_member_interval = HZ;
1946         br->multicast_query_response_interval = 10 * HZ;
1947         br->multicast_startup_query_interval = 125 * HZ / 4;
1948         br->multicast_query_interval = 125 * HZ;
1949         br->multicast_querier_interval = 255 * HZ;
1950         br->multicast_membership_interval = 260 * HZ;
1951
1952         br->ip4_other_query.delay_time = 0;
1953         br->ip4_querier.port = NULL;
1954         br->multicast_igmp_version = 2;
1955 #if IS_ENABLED(CONFIG_IPV6)
1956         br->multicast_mld_version = 1;
1957         br->ip6_other_query.delay_time = 0;
1958         br->ip6_querier.port = NULL;
1959 #endif
1960         br->has_ipv6_addr = 1;
1961
1962         spin_lock_init(&br->multicast_lock);
1963         setup_timer(&br->multicast_router_timer,
1964                     br_multicast_local_router_expired, 0);
1965         setup_timer(&br->ip4_other_query.timer,
1966                     br_ip4_multicast_querier_expired, (unsigned long)br);
1967         setup_timer(&br->ip4_own_query.timer, br_ip4_multicast_query_expired,
1968                     (unsigned long)br);
1969 #if IS_ENABLED(CONFIG_IPV6)
1970         setup_timer(&br->ip6_other_query.timer,
1971                     br_ip6_multicast_querier_expired, (unsigned long)br);
1972         setup_timer(&br->ip6_own_query.timer, br_ip6_multicast_query_expired,
1973                     (unsigned long)br);
1974 #endif
1975 }
1976
1977 static void __br_multicast_open(struct net_bridge *br,
1978                                 struct bridge_mcast_own_query *query)
1979 {
1980         query->startup_sent = 0;
1981
1982         if (br->multicast_disabled)
1983                 return;
1984
1985         mod_timer(&query->timer, jiffies);
1986 }
1987
1988 void br_multicast_open(struct net_bridge *br)
1989 {
1990         __br_multicast_open(br, &br->ip4_own_query);
1991 #if IS_ENABLED(CONFIG_IPV6)
1992         __br_multicast_open(br, &br->ip6_own_query);
1993 #endif
1994 }
1995
1996 void br_multicast_stop(struct net_bridge *br)
1997 {
1998         del_timer_sync(&br->multicast_router_timer);
1999         del_timer_sync(&br->ip4_other_query.timer);
2000         del_timer_sync(&br->ip4_own_query.timer);
2001 #if IS_ENABLED(CONFIG_IPV6)
2002         del_timer_sync(&br->ip6_other_query.timer);
2003         del_timer_sync(&br->ip6_own_query.timer);
2004 #endif
2005 }
2006
2007 void br_multicast_dev_del(struct net_bridge *br)
2008 {
2009         struct net_bridge_mdb_htable *mdb;
2010         struct net_bridge_mdb_entry *mp;
2011         struct hlist_node *n;
2012         u32 ver;
2013         int i;
2014
2015         spin_lock_bh(&br->multicast_lock);
2016         mdb = mlock_dereference(br->mdb, br);
2017         if (!mdb)
2018                 goto out;
2019
2020         br->mdb = NULL;
2021
2022         ver = mdb->ver;
2023         for (i = 0; i < mdb->max; i++) {
2024                 hlist_for_each_entry_safe(mp, n, &mdb->mhash[i],
2025                                           hlist[ver]) {
2026                         del_timer(&mp->timer);
2027                         call_rcu_bh(&mp->rcu, br_multicast_free_group);
2028                 }
2029         }
2030
2031         if (mdb->old) {
2032                 spin_unlock_bh(&br->multicast_lock);
2033                 rcu_barrier_bh();
2034                 spin_lock_bh(&br->multicast_lock);
2035                 WARN_ON(mdb->old);
2036         }
2037
2038         mdb->old = mdb;
2039         call_rcu_bh(&mdb->rcu, br_mdb_free);
2040
2041 out:
2042         spin_unlock_bh(&br->multicast_lock);
2043 }
2044
2045 int br_multicast_set_router(struct net_bridge *br, unsigned long val)
2046 {
2047         int err = -EINVAL;
2048
2049         spin_lock_bh(&br->multicast_lock);
2050
2051         switch (val) {
2052         case MDB_RTR_TYPE_DISABLED:
2053         case MDB_RTR_TYPE_PERM:
2054                 del_timer(&br->multicast_router_timer);
2055                 /* fall through */
2056         case MDB_RTR_TYPE_TEMP_QUERY:
2057                 br->multicast_router = val;
2058                 err = 0;
2059                 break;
2060         }
2061
2062         spin_unlock_bh(&br->multicast_lock);
2063
2064         return err;
2065 }
2066
2067 static void __del_port_router(struct net_bridge_port *p)
2068 {
2069         if (hlist_unhashed(&p->rlist))
2070                 return;
2071         hlist_del_init_rcu(&p->rlist);
2072         br_rtr_notify(p->br->dev, p, RTM_DELMDB);
2073         br_port_mc_router_state_change(p, false);
2074
2075         /* don't allow timer refresh */
2076         if (p->multicast_router == MDB_RTR_TYPE_TEMP)
2077                 p->multicast_router = MDB_RTR_TYPE_TEMP_QUERY;
2078 }
2079
2080 int br_multicast_set_port_router(struct net_bridge_port *p, unsigned long val)
2081 {
2082         struct net_bridge *br = p->br;
2083         unsigned long now = jiffies;
2084         int err = -EINVAL;
2085
2086         spin_lock(&br->multicast_lock);
2087         if (p->multicast_router == val) {
2088                 /* Refresh the temp router port timer */
2089                 if (p->multicast_router == MDB_RTR_TYPE_TEMP)
2090                         mod_timer(&p->multicast_router_timer,
2091                                   now + br->multicast_querier_interval);
2092                 err = 0;
2093                 goto unlock;
2094         }
2095         switch (val) {
2096         case MDB_RTR_TYPE_DISABLED:
2097                 p->multicast_router = MDB_RTR_TYPE_DISABLED;
2098                 __del_port_router(p);
2099                 del_timer(&p->multicast_router_timer);
2100                 break;
2101         case MDB_RTR_TYPE_TEMP_QUERY:
2102                 p->multicast_router = MDB_RTR_TYPE_TEMP_QUERY;
2103                 __del_port_router(p);
2104                 break;
2105         case MDB_RTR_TYPE_PERM:
2106                 p->multicast_router = MDB_RTR_TYPE_PERM;
2107                 del_timer(&p->multicast_router_timer);
2108                 br_multicast_add_router(br, p);
2109                 break;
2110         case MDB_RTR_TYPE_TEMP:
2111                 p->multicast_router = MDB_RTR_TYPE_TEMP;
2112                 br_multicast_mark_router(br, p);
2113                 break;
2114         default:
2115                 goto unlock;
2116         }
2117         err = 0;
2118 unlock:
2119         spin_unlock(&br->multicast_lock);
2120
2121         return err;
2122 }
2123
2124 static void br_multicast_start_querier(struct net_bridge *br,
2125                                        struct bridge_mcast_own_query *query)
2126 {
2127         struct net_bridge_port *port;
2128
2129         __br_multicast_open(br, query);
2130
2131         rcu_read_lock();
2132         list_for_each_entry_rcu(port, &br->port_list, list) {
2133                 if (port->state == BR_STATE_DISABLED ||
2134                     port->state == BR_STATE_BLOCKING)
2135                         continue;
2136
2137                 if (query == &br->ip4_own_query)
2138                         br_multicast_enable(&port->ip4_own_query);
2139 #if IS_ENABLED(CONFIG_IPV6)
2140                 else
2141                         br_multicast_enable(&port->ip6_own_query);
2142 #endif
2143         }
2144         rcu_read_unlock();
2145 }
2146
2147 int br_multicast_toggle(struct net_bridge *br, unsigned long val)
2148 {
2149         struct net_bridge_mdb_htable *mdb;
2150         struct net_bridge_port *port;
2151         int err = 0;
2152
2153         spin_lock_bh(&br->multicast_lock);
2154         if (br->multicast_disabled == !val)
2155                 goto unlock;
2156
2157         br_mc_disabled_update(br->dev, !val);
2158         br->multicast_disabled = !val;
2159         if (br->multicast_disabled)
2160                 goto unlock;
2161
2162         if (!netif_running(br->dev))
2163                 goto unlock;
2164
2165         mdb = mlock_dereference(br->mdb, br);
2166         if (mdb) {
2167                 if (mdb->old) {
2168                         err = -EEXIST;
2169 rollback:
2170                         br->multicast_disabled = !!val;
2171                         goto unlock;
2172                 }
2173
2174                 err = br_mdb_rehash(&br->mdb, mdb->max,
2175                                     br->hash_elasticity);
2176                 if (err)
2177                         goto rollback;
2178         }
2179
2180         br_multicast_open(br);
2181         list_for_each_entry(port, &br->port_list, list)
2182                 __br_multicast_enable_port(port);
2183
2184 unlock:
2185         spin_unlock_bh(&br->multicast_lock);
2186
2187         return err;
2188 }
2189
2190 bool br_multicast_enabled(const struct net_device *dev)
2191 {
2192         struct net_bridge *br = netdev_priv(dev);
2193
2194         return !br->multicast_disabled;
2195 }
2196 EXPORT_SYMBOL_GPL(br_multicast_enabled);
2197
2198 int br_multicast_set_querier(struct net_bridge *br, unsigned long val)
2199 {
2200         unsigned long max_delay;
2201
2202         val = !!val;
2203
2204         spin_lock_bh(&br->multicast_lock);
2205         if (br->multicast_querier == val)
2206                 goto unlock;
2207
2208         br->multicast_querier = val;
2209         if (!val)
2210                 goto unlock;
2211
2212         max_delay = br->multicast_query_response_interval;
2213
2214         if (!timer_pending(&br->ip4_other_query.timer))
2215                 br->ip4_other_query.delay_time = jiffies + max_delay;
2216
2217         br_multicast_start_querier(br, &br->ip4_own_query);
2218
2219 #if IS_ENABLED(CONFIG_IPV6)
2220         if (!timer_pending(&br->ip6_other_query.timer))
2221                 br->ip6_other_query.delay_time = jiffies + max_delay;
2222
2223         br_multicast_start_querier(br, &br->ip6_own_query);
2224 #endif
2225
2226 unlock:
2227         spin_unlock_bh(&br->multicast_lock);
2228
2229         return 0;
2230 }
2231
2232 int br_multicast_set_hash_max(struct net_bridge *br, unsigned long val)
2233 {
2234         int err = -EINVAL;
2235         u32 old;
2236         struct net_bridge_mdb_htable *mdb;
2237
2238         spin_lock_bh(&br->multicast_lock);
2239         if (!is_power_of_2(val))
2240                 goto unlock;
2241
2242         mdb = mlock_dereference(br->mdb, br);
2243         if (mdb && val < mdb->size)
2244                 goto unlock;
2245
2246         err = 0;
2247
2248         old = br->hash_max;
2249         br->hash_max = val;
2250
2251         if (mdb) {
2252                 if (mdb->old) {
2253                         err = -EEXIST;
2254 rollback:
2255                         br->hash_max = old;
2256                         goto unlock;
2257                 }
2258
2259                 err = br_mdb_rehash(&br->mdb, br->hash_max,
2260                                     br->hash_elasticity);
2261                 if (err)
2262                         goto rollback;
2263         }
2264
2265 unlock:
2266         spin_unlock_bh(&br->multicast_lock);
2267
2268         return err;
2269 }
2270
2271 int br_multicast_set_igmp_version(struct net_bridge *br, unsigned long val)
2272 {
2273         /* Currently we support only version 2 and 3 */
2274         switch (val) {
2275         case 2:
2276         case 3:
2277                 break;
2278         default:
2279                 return -EINVAL;
2280         }
2281
2282         spin_lock_bh(&br->multicast_lock);
2283         br->multicast_igmp_version = val;
2284         spin_unlock_bh(&br->multicast_lock);
2285
2286         return 0;
2287 }
2288
2289 #if IS_ENABLED(CONFIG_IPV6)
2290 int br_multicast_set_mld_version(struct net_bridge *br, unsigned long val)
2291 {
2292         /* Currently we support version 1 and 2 */
2293         switch (val) {
2294         case 1:
2295         case 2:
2296                 break;
2297         default:
2298                 return -EINVAL;
2299         }
2300
2301         spin_lock_bh(&br->multicast_lock);
2302         br->multicast_mld_version = val;
2303         spin_unlock_bh(&br->multicast_lock);
2304
2305         return 0;
2306 }
2307 #endif
2308
2309 /**
2310  * br_multicast_list_adjacent - Returns snooped multicast addresses
2311  * @dev:        The bridge port adjacent to which to retrieve addresses
2312  * @br_ip_list: The list to store found, snooped multicast IP addresses in
2313  *
2314  * Creates a list of IP addresses (struct br_ip_list) sensed by the multicast
2315  * snooping feature on all bridge ports of dev's bridge device, excluding
2316  * the addresses from dev itself.
2317  *
2318  * Returns the number of items added to br_ip_list.
2319  *
2320  * Notes:
2321  * - br_ip_list needs to be initialized by caller
2322  * - br_ip_list might contain duplicates in the end
2323  *   (needs to be taken care of by caller)
2324  * - br_ip_list needs to be freed by caller
2325  */
2326 int br_multicast_list_adjacent(struct net_device *dev,
2327                                struct list_head *br_ip_list)
2328 {
2329         struct net_bridge *br;
2330         struct net_bridge_port *port;
2331         struct net_bridge_port_group *group;
2332         struct br_ip_list *entry;
2333         int count = 0;
2334
2335         rcu_read_lock();
2336         if (!br_ip_list || !br_port_exists(dev))
2337                 goto unlock;
2338
2339         port = br_port_get_rcu(dev);
2340         if (!port || !port->br)
2341                 goto unlock;
2342
2343         br = port->br;
2344
2345         list_for_each_entry_rcu(port, &br->port_list, list) {
2346                 if (!port->dev || port->dev == dev)
2347                         continue;
2348
2349                 hlist_for_each_entry_rcu(group, &port->mglist, mglist) {
2350                         entry = kmalloc(sizeof(*entry), GFP_ATOMIC);
2351                         if (!entry)
2352                                 goto unlock;
2353
2354                         entry->addr = group->addr;
2355                         list_add(&entry->list, br_ip_list);
2356                         count++;
2357                 }
2358         }
2359
2360 unlock:
2361         rcu_read_unlock();
2362         return count;
2363 }
2364 EXPORT_SYMBOL_GPL(br_multicast_list_adjacent);
2365
2366 /**
2367  * br_multicast_has_querier_anywhere - Checks for a querier on a bridge
2368  * @dev: The bridge port providing the bridge on which to check for a querier
2369  * @proto: The protocol family to check for: IGMP -> ETH_P_IP, MLD -> ETH_P_IPV6
2370  *
2371  * Checks whether the given interface has a bridge on top and if so returns
2372  * true if a valid querier exists anywhere on the bridged link layer.
2373  * Otherwise returns false.
2374  */
2375 bool br_multicast_has_querier_anywhere(struct net_device *dev, int proto)
2376 {
2377         struct net_bridge *br;
2378         struct net_bridge_port *port;
2379         struct ethhdr eth;
2380         bool ret = false;
2381
2382         rcu_read_lock();
2383         if (!br_port_exists(dev))
2384                 goto unlock;
2385
2386         port = br_port_get_rcu(dev);
2387         if (!port || !port->br)
2388                 goto unlock;
2389
2390         br = port->br;
2391
2392         memset(&eth, 0, sizeof(eth));
2393         eth.h_proto = htons(proto);
2394
2395         ret = br_multicast_querier_exists(br, &eth);
2396
2397 unlock:
2398         rcu_read_unlock();
2399         return ret;
2400 }
2401 EXPORT_SYMBOL_GPL(br_multicast_has_querier_anywhere);
2402
2403 /**
2404  * br_multicast_has_querier_adjacent - Checks for a querier behind a bridge port
2405  * @dev: The bridge port adjacent to which to check for a querier
2406  * @proto: The protocol family to check for: IGMP -> ETH_P_IP, MLD -> ETH_P_IPV6
2407  *
2408  * Checks whether the given interface has a bridge on top and if so returns
2409  * true if a selected querier is behind one of the other ports of this
2410  * bridge. Otherwise returns false.
2411  */
2412 bool br_multicast_has_querier_adjacent(struct net_device *dev, int proto)
2413 {
2414         struct net_bridge *br;
2415         struct net_bridge_port *port;
2416         bool ret = false;
2417
2418         rcu_read_lock();
2419         if (!br_port_exists(dev))
2420                 goto unlock;
2421
2422         port = br_port_get_rcu(dev);
2423         if (!port || !port->br)
2424                 goto unlock;
2425
2426         br = port->br;
2427
2428         switch (proto) {
2429         case ETH_P_IP:
2430                 if (!timer_pending(&br->ip4_other_query.timer) ||
2431                     rcu_dereference(br->ip4_querier.port) == port)
2432                         goto unlock;
2433                 break;
2434 #if IS_ENABLED(CONFIG_IPV6)
2435         case ETH_P_IPV6:
2436                 if (!timer_pending(&br->ip6_other_query.timer) ||
2437                     rcu_dereference(br->ip6_querier.port) == port)
2438                         goto unlock;
2439                 break;
2440 #endif
2441         default:
2442                 goto unlock;
2443         }
2444
2445         ret = true;
2446 unlock:
2447         rcu_read_unlock();
2448         return ret;
2449 }
2450 EXPORT_SYMBOL_GPL(br_multicast_has_querier_adjacent);
2451
2452 static void br_mcast_stats_add(struct bridge_mcast_stats __percpu *stats,
2453                                const struct sk_buff *skb, u8 type, u8 dir)
2454 {
2455         struct bridge_mcast_stats *pstats = this_cpu_ptr(stats);
2456         __be16 proto = skb->protocol;
2457         unsigned int t_len;
2458
2459         u64_stats_update_begin(&pstats->syncp);
2460         switch (proto) {
2461         case htons(ETH_P_IP):
2462                 t_len = ntohs(ip_hdr(skb)->tot_len) - ip_hdrlen(skb);
2463                 switch (type) {
2464                 case IGMP_HOST_MEMBERSHIP_REPORT:
2465                         pstats->mstats.igmp_v1reports[dir]++;
2466                         break;
2467                 case IGMPV2_HOST_MEMBERSHIP_REPORT:
2468                         pstats->mstats.igmp_v2reports[dir]++;
2469                         break;
2470                 case IGMPV3_HOST_MEMBERSHIP_REPORT:
2471                         pstats->mstats.igmp_v3reports[dir]++;
2472                         break;
2473                 case IGMP_HOST_MEMBERSHIP_QUERY:
2474                         if (t_len != sizeof(struct igmphdr)) {
2475                                 pstats->mstats.igmp_v3queries[dir]++;
2476                         } else {
2477                                 unsigned int offset = skb_transport_offset(skb);
2478                                 struct igmphdr *ih, _ihdr;
2479
2480                                 ih = skb_header_pointer(skb, offset,
2481                                                         sizeof(_ihdr), &_ihdr);
2482                                 if (!ih)
2483                                         break;
2484                                 if (!ih->code)
2485                                         pstats->mstats.igmp_v1queries[dir]++;
2486                                 else
2487                                         pstats->mstats.igmp_v2queries[dir]++;
2488                         }
2489                         break;
2490                 case IGMP_HOST_LEAVE_MESSAGE:
2491                         pstats->mstats.igmp_leaves[dir]++;
2492                         break;
2493                 }
2494                 break;
2495 #if IS_ENABLED(CONFIG_IPV6)
2496         case htons(ETH_P_IPV6):
2497                 t_len = ntohs(ipv6_hdr(skb)->payload_len) +
2498                         sizeof(struct ipv6hdr);
2499                 t_len -= skb_network_header_len(skb);
2500                 switch (type) {
2501                 case ICMPV6_MGM_REPORT:
2502                         pstats->mstats.mld_v1reports[dir]++;
2503                         break;
2504                 case ICMPV6_MLD2_REPORT:
2505                         pstats->mstats.mld_v2reports[dir]++;
2506                         break;
2507                 case ICMPV6_MGM_QUERY:
2508                         if (t_len != sizeof(struct mld_msg))
2509                                 pstats->mstats.mld_v2queries[dir]++;
2510                         else
2511                                 pstats->mstats.mld_v1queries[dir]++;
2512                         break;
2513                 case ICMPV6_MGM_REDUCTION:
2514                         pstats->mstats.mld_leaves[dir]++;
2515                         break;
2516                 }
2517                 break;
2518 #endif /* CONFIG_IPV6 */
2519         }
2520         u64_stats_update_end(&pstats->syncp);
2521 }
2522
2523 void br_multicast_count(struct net_bridge *br, const struct net_bridge_port *p,
2524                         const struct sk_buff *skb, u8 type, u8 dir)
2525 {
2526         struct bridge_mcast_stats __percpu *stats;
2527
2528         /* if multicast_disabled is true then igmp type can't be set */
2529         if (!type || !br->multicast_stats_enabled)
2530                 return;
2531
2532         if (p)
2533                 stats = p->mcast_stats;
2534         else
2535                 stats = br->mcast_stats;
2536         if (WARN_ON(!stats))
2537                 return;
2538
2539         br_mcast_stats_add(stats, skb, type, dir);
2540 }
2541
2542 int br_multicast_init_stats(struct net_bridge *br)
2543 {
2544         br->mcast_stats = netdev_alloc_pcpu_stats(struct bridge_mcast_stats);
2545         if (!br->mcast_stats)
2546                 return -ENOMEM;
2547
2548         return 0;
2549 }
2550
2551 void br_multicast_uninit_stats(struct net_bridge *br)
2552 {
2553         free_percpu(br->mcast_stats);
2554 }
2555
2556 static void mcast_stats_add_dir(u64 *dst, u64 *src)
2557 {
2558         dst[BR_MCAST_DIR_RX] += src[BR_MCAST_DIR_RX];
2559         dst[BR_MCAST_DIR_TX] += src[BR_MCAST_DIR_TX];
2560 }
2561
2562 void br_multicast_get_stats(const struct net_bridge *br,
2563                             const struct net_bridge_port *p,
2564                             struct br_mcast_stats *dest)
2565 {
2566         struct bridge_mcast_stats __percpu *stats;
2567         struct br_mcast_stats tdst;
2568         int i;
2569
2570         memset(dest, 0, sizeof(*dest));
2571         if (p)
2572                 stats = p->mcast_stats;
2573         else
2574                 stats = br->mcast_stats;
2575         if (WARN_ON(!stats))
2576                 return;
2577
2578         memset(&tdst, 0, sizeof(tdst));
2579         for_each_possible_cpu(i) {
2580                 struct bridge_mcast_stats *cpu_stats = per_cpu_ptr(stats, i);
2581                 struct br_mcast_stats temp;
2582                 unsigned int start;
2583
2584                 do {
2585                         start = u64_stats_fetch_begin_irq(&cpu_stats->syncp);
2586                         memcpy(&temp, &cpu_stats->mstats, sizeof(temp));
2587                 } while (u64_stats_fetch_retry_irq(&cpu_stats->syncp, start));
2588
2589                 mcast_stats_add_dir(tdst.igmp_v1queries, temp.igmp_v1queries);
2590                 mcast_stats_add_dir(tdst.igmp_v2queries, temp.igmp_v2queries);
2591                 mcast_stats_add_dir(tdst.igmp_v3queries, temp.igmp_v3queries);
2592                 mcast_stats_add_dir(tdst.igmp_leaves, temp.igmp_leaves);
2593                 mcast_stats_add_dir(tdst.igmp_v1reports, temp.igmp_v1reports);
2594                 mcast_stats_add_dir(tdst.igmp_v2reports, temp.igmp_v2reports);
2595                 mcast_stats_add_dir(tdst.igmp_v3reports, temp.igmp_v3reports);
2596                 tdst.igmp_parse_errors += temp.igmp_parse_errors;
2597
2598                 mcast_stats_add_dir(tdst.mld_v1queries, temp.mld_v1queries);
2599                 mcast_stats_add_dir(tdst.mld_v2queries, temp.mld_v2queries);
2600                 mcast_stats_add_dir(tdst.mld_leaves, temp.mld_leaves);
2601                 mcast_stats_add_dir(tdst.mld_v1reports, temp.mld_v1reports);
2602                 mcast_stats_add_dir(tdst.mld_v2reports, temp.mld_v2reports);
2603                 tdst.mld_parse_errors += temp.mld_parse_errors;
2604         }
2605         memcpy(dest, &tdst, sizeof(*dest));
2606 }