net/ipv6/mcast.c

   1 /*
   2  *      Multicast support for IPv6
   3  *      Linux INET6 implementation
   4  *
   5  *      Authors:
   6  *      Pedro Roque             <roque@di.fc.ul.pt>
   7  *
   8  *      Based on linux/ipv4/igmp.c and linux/ipv4/ip_sockglue.c
   9  *
  10  *      This program is free software; you can redistribute it and/or
  11  *      modify it under the terms of the GNU General Public License
  12  *      as published by the Free Software Foundation; either version
  13  *      2 of the License, or (at your option) any later version.
  14  */
  15
  16 /* Changes:
  17  *
  18  *      yoshfuji        : fix format of router-alert option
  19  *      YOSHIFUJI Hideaki @USAGI:
  20  *              Fixed source address for MLD message based on
  21  *              <draft-ietf-magma-mld-source-05.txt>.
  22  *      YOSHIFUJI Hideaki @USAGI:
  23  *              - Ignore Queries for invalid addresses.
  24  *              - MLD for link-local addresses.
  25  *      David L Stevens <dlstevens@us.ibm.com>:
  26  *              - MLDv2 support
  27  */
  28
  29 #include <linux/module.h>
  30 #include <linux/errno.h>
  31 #include <linux/types.h>
  32 #include <linux/string.h>
  33 #include <linux/socket.h>
  34 #include <linux/sockios.h>
  35 #include <linux/jiffies.h>
  36 #include <linux/times.h>
  37 #include <linux/net.h>
  38 #include <linux/in.h>
  39 #include <linux/in6.h>
  40 #include <linux/netdevice.h>
  41 #include <linux/if_arp.h>
  42 #include <linux/route.h>
  43 #include <linux/init.h>
  44 #include <linux/proc_fs.h>
  45 #include <linux/seq_file.h>
  46 #include <linux/slab.h>
  47 #include <linux/pkt_sched.h>
  48 #include <net/mld.h>
  49
  50 #include <linux/netfilter.h>
  51 #include <linux/netfilter_ipv6.h>
  52
  53 #include <net/net_namespace.h>
  54 #include <net/sock.h>
  55 #include <net/snmp.h>
  56
  57 #include <net/ipv6.h>
  58 #include <net/protocol.h>
  59 #include <net/if_inet6.h>
  60 #include <net/ndisc.h>
  61 #include <net/addrconf.h>
  62 #include <net/ip6_route.h>
  63 #include <net/inet_common.h>
  64
  65 #include <net/ip6_checksum.h>
  66
  67 /* Ensure that we have struct in6_addr aligned on 32bit word. */
  68 static void *__mld2_query_bugs[] __attribute__((__unused__)) = {
  69         BUILD_BUG_ON_NULL(offsetof(struct mld2_query, mld2q_srcs) % 4),
  70         BUILD_BUG_ON_NULL(offsetof(struct mld2_report, mld2r_grec) % 4),
  71         BUILD_BUG_ON_NULL(offsetof(struct mld2_grec, grec_mca) % 4)
  72 };
  73
  74 static struct in6_addr mld2_all_mcr = MLD2_ALL_MCR_INIT;
  75
  76 static void igmp6_join_group(struct ifmcaddr6 *ma);
  77 static void igmp6_leave_group(struct ifmcaddr6 *ma);
  78 static void igmp6_timer_handler(unsigned long data);
  79
  80 static void mld_gq_timer_expire(unsigned long data);
  81 static void mld_ifc_timer_expire(unsigned long data);
  82 static void mld_ifc_event(struct inet6_dev *idev);
  83 static void mld_add_delrec(struct inet6_dev *idev, struct ifmcaddr6 *pmc);
  84 static void mld_del_delrec(struct inet6_dev *idev, const struct in6_addr *addr);
  85 static void mld_clear_delrec(struct inet6_dev *idev);
  86 static bool mld_in_v1_mode(const struct inet6_dev *idev);
  87 static int sf_setstate(struct ifmcaddr6 *pmc);
  88 static void sf_markstate(struct ifmcaddr6 *pmc);
  89 static void ip6_mc_clear_src(struct ifmcaddr6 *pmc);
  90 static int ip6_mc_del_src(struct inet6_dev *idev, const struct in6_addr *pmca,
  91                           int sfmode, int sfcount, const struct in6_addr *psfsrc,
  92                           int delta);
  93 static int ip6_mc_add_src(struct inet6_dev *idev, const struct in6_addr *pmca,
  94                           int sfmode, int sfcount, const struct in6_addr *psfsrc,
  95                           int delta);
  96 static int ip6_mc_leave_src(struct sock *sk, struct ipv6_mc_socklist *iml,
  97                             struct inet6_dev *idev);
  98
  99 #define MLD_QRV_DEFAULT         2
 100 /* RFC3810, 9.2. Query Interval */
 101 #define MLD_QI_DEFAULT          (125 * HZ)
 102 /* RFC3810, 9.3. Query Response Interval */
 103 #define MLD_QRI_DEFAULT         (10 * HZ)
 104
 105 /* RFC3810, 8.1 Query Version Distinctions */
 106 #define MLD_V1_QUERY_LEN        24
 107 #define MLD_V2_QUERY_LEN_MIN    28
 108
 109 #define IPV6_MLD_MAX_MSF        64
 110
 111 int sysctl_mld_max_msf __read_mostly = IPV6_MLD_MAX_MSF;
 112 int sysctl_mld_qrv __read_mostly = MLD_QRV_DEFAULT;
 113
 114 /*
 115  *      socket join on multicast group
 116  */
 117
 118 #define for_each_pmc_rcu(np, pmc)                               \
 119         for (pmc = rcu_dereference(np->ipv6_mc_list);           \
 120              pmc != NULL;                                       \
 121              pmc = rcu_dereference(pmc->next))
 122
 123 static int unsolicited_report_interval(struct inet6_dev *idev)
 124 {
 125         int iv;
 126
 127         if (mld_in_v1_mode(idev))
 128                 iv = idev->cnf.mldv1_unsolicited_report_interval;
 129         else
 130                 iv = idev->cnf.mldv2_unsolicited_report_interval;
 131
 132         return iv > 0 ? iv : 1;
 133 }
 134
 135 int ipv6_sock_mc_join(struct sock *sk, int ifindex, const struct in6_addr *addr)
 136 {
 137         struct net_device *dev = NULL;
 138         struct ipv6_mc_socklist *mc_lst;
 139         struct ipv6_pinfo *np = inet6_sk(sk);
 140         struct net *net = sock_net(sk);
 141         int err;
 142
 143         ASSERT_RTNL();
 144
 145         if (!ipv6_addr_is_multicast(addr))
 146                 return -EINVAL;
 147
 148         rcu_read_lock();
 149         for_each_pmc_rcu(np, mc_lst) {
 150                 if ((ifindex == 0 || mc_lst->ifindex == ifindex) &&
 151                     ipv6_addr_equal(&mc_lst->addr, addr)) {
 152                         rcu_read_unlock();
 153                         return -EADDRINUSE;
 154                 }
 155         }
 156         rcu_read_unlock();
 157
 158         mc_lst = sock_kmalloc(sk, sizeof(struct ipv6_mc_socklist), GFP_KERNEL);
 159
 160         if (!mc_lst)
 161                 return -ENOMEM;
 162
 163         mc_lst->next = NULL;
 164         mc_lst->addr = *addr;
 165
 166         if (ifindex == 0) {
 167                 struct rt6_info *rt;
 168                 rt = rt6_lookup(net, addr, NULL, 0, 0);
 169                 if (rt) {
 170                         dev = rt->dst.dev;
 171                         ip6_rt_put(rt);
 172                 }
 173         } else
 174                 dev = __dev_get_by_index(net, ifindex);
 175
 176         if (!dev) {
 177                 sock_kfree_s(sk, mc_lst, sizeof(*mc_lst));
 178                 return -ENODEV;
 179         }
 180
 181         mc_lst->ifindex = dev->ifindex;
 182         mc_lst->sfmode = MCAST_EXCLUDE;
 183         rwlock_init(&mc_lst->sflock);
 184         mc_lst->sflist = NULL;
 185
 186         /*
 187          *      now add/increase the group membership on the device
 188          */
 189
 190         err = ipv6_dev_mc_inc(dev, addr);
 191
 192         if (err) {
 193                 sock_kfree_s(sk, mc_lst, sizeof(*mc_lst));
 194                 return err;
 195         }
 196
 197         mc_lst->next = np->ipv6_mc_list;
 198         rcu_assign_pointer(np->ipv6_mc_list, mc_lst);
 199
 200         return 0;
 201 }
 202 EXPORT_SYMBOL(ipv6_sock_mc_join);
 203
 204 /*
 205  *      socket leave on multicast group
 206  */
 207 int ipv6_sock_mc_drop(struct sock *sk, int ifindex, const struct in6_addr *addr)
 208 {
 209         struct ipv6_pinfo *np = inet6_sk(sk);
 210         struct ipv6_mc_socklist *mc_lst;
 211         struct ipv6_mc_socklist __rcu **lnk;
 212         struct net *net = sock_net(sk);
 213
 214         ASSERT_RTNL();
 215
 216         if (!ipv6_addr_is_multicast(addr))
 217                 return -EINVAL;
 218
 219         for (lnk = &np->ipv6_mc_list;
 220              (mc_lst = rtnl_dereference(*lnk)) != NULL;
 221               lnk = &mc_lst->next) {
 222                 if ((ifindex == 0 || mc_lst->ifindex == ifindex) &&
 223                     ipv6_addr_equal(&mc_lst->addr, addr)) {
 224                         struct net_device *dev;
 225
 226                         *lnk = mc_lst->next;
 227
 228                         dev = __dev_get_by_index(net, mc_lst->ifindex);
 229                         if (dev) {
 230                                 struct inet6_dev *idev = __in6_dev_get(dev);
 231
 232                                 (void) ip6_mc_leave_src(sk, mc_lst, idev);
 233                                 if (idev)
 234                                         __ipv6_dev_mc_dec(idev, &mc_lst->addr);
 235                         } else
 236                                 (void) ip6_mc_leave_src(sk, mc_lst, NULL);
 237
 238                         atomic_sub(sizeof(*mc_lst), &sk->sk_omem_alloc);
 239                         kfree_rcu(mc_lst, rcu);
 240                         return 0;
 241                 }
 242         }
 243
 244         return -EADDRNOTAVAIL;
 245 }
 246 EXPORT_SYMBOL(ipv6_sock_mc_drop);
 247
 248 /* called with rcu_read_lock() */
 249 static struct inet6_dev *ip6_mc_find_dev_rcu(struct net *net,
 250                                              const struct in6_addr *group,
 251                                              int ifindex)
 252 {
 253         struct net_device *dev = NULL;
 254         struct inet6_dev *idev = NULL;
 255
 256         if (ifindex == 0) {
 257                 struct rt6_info *rt = rt6_lookup(net, group, NULL, 0, 0);
 258
 259                 if (rt) {
 260                         dev = rt->dst.dev;
 261                         ip6_rt_put(rt);
 262                 }
 263         } else
 264                 dev = dev_get_by_index_rcu(net, ifindex);
 265
 266         if (!dev)
 267                 return NULL;
 268         idev = __in6_dev_get(dev);
 269         if (!idev)
 270                 return NULL;
 271         read_lock_bh(&idev->lock);
 272         if (idev->dead) {
 273                 read_unlock_bh(&idev->lock);
 274                 return NULL;
 275         }
 276         return idev;
 277 }
 278
 279 void __ipv6_sock_mc_close(struct sock *sk)
 280 {
 281         struct ipv6_pinfo *np = inet6_sk(sk);
 282         struct ipv6_mc_socklist *mc_lst;
 283         struct net *net = sock_net(sk);
 284
 285         ASSERT_RTNL();
 286
 287         while ((mc_lst = rtnl_dereference(np->ipv6_mc_list)) != NULL) {
 288                 struct net_device *dev;
 289
 290                 np->ipv6_mc_list = mc_lst->next;
 291
 292                 dev = __dev_get_by_index(net, mc_lst->ifindex);
 293                 if (dev) {
 294                         struct inet6_dev *idev = __in6_dev_get(dev);
 295
 296                         (void) ip6_mc_leave_src(sk, mc_lst, idev);
 297                         if (idev)
 298                                 __ipv6_dev_mc_dec(idev, &mc_lst->addr);
 299                 } else
 300                         (void) ip6_mc_leave_src(sk, mc_lst, NULL);
 301
 302                 atomic_sub(sizeof(*mc_lst), &sk->sk_omem_alloc);
 303                 kfree_rcu(mc_lst, rcu);
 304         }
 305 }
 306
 307 void ipv6_sock_mc_close(struct sock *sk)
 308 {
 309         struct ipv6_pinfo *np = inet6_sk(sk);
 310
 311         if (!rcu_access_pointer(np->ipv6_mc_list))
 312                 return;
 313         rtnl_lock();
 314         __ipv6_sock_mc_close(sk);
 315         rtnl_unlock();
 316 }
 317
 318 int ip6_mc_source(int add, int omode, struct sock *sk,
 319         struct group_source_req *pgsr)
 320 {
 321         struct in6_addr *source, *group;
 322         struct ipv6_mc_socklist *pmc;
 323         struct inet6_dev *idev;
 324         struct ipv6_pinfo *inet6 = inet6_sk(sk);
 325         struct ip6_sf_socklist *psl;
 326         struct net *net = sock_net(sk);
 327         int i, j, rv;
 328         int leavegroup = 0;
 329         int pmclocked = 0;
 330         int err;
 331
 332         source = &((struct sockaddr_in6 *)&pgsr->gsr_source)->sin6_addr;
 333         group = &((struct sockaddr_in6 *)&pgsr->gsr_group)->sin6_addr;
 334
 335         if (!ipv6_addr_is_multicast(group))
 336                 return -EINVAL;
 337
 338         rcu_read_lock();
 339         idev = ip6_mc_find_dev_rcu(net, group, pgsr->gsr_interface);
 340         if (!idev) {
 341                 rcu_read_unlock();
 342                 return -ENODEV;
 343         }
 344
 345         err = -EADDRNOTAVAIL;
 346
 347         for_each_pmc_rcu(inet6, pmc) {
 348                 if (pgsr->gsr_interface && pmc->ifindex != pgsr->gsr_interface)
 349                         continue;
 350                 if (ipv6_addr_equal(&pmc->addr, group))
 351                         break;
 352         }
 353         if (!pmc) {             /* must have a prior join */
 354                 err = -EINVAL;
 355                 goto done;
 356         }
 357         /* if a source filter was set, must be the same mode as before */
 358         if (pmc->sflist) {
 359                 if (pmc->sfmode != omode) {
 360                         err = -EINVAL;
 361                         goto done;
 362                 }
 363         } else if (pmc->sfmode != omode) {
 364                 /* allow mode switches for empty-set filters */
 365                 ip6_mc_add_src(idev, group, omode, 0, NULL, 0);
 366                 ip6_mc_del_src(idev, group, pmc->sfmode, 0, NULL, 0);
 367                 pmc->sfmode = omode;
 368         }
 369
 370         write_lock(&pmc->sflock);
 371         pmclocked = 1;
 372
 373         psl = pmc->sflist;
 374         if (!add) {
 375                 if (!psl)
 376                         goto done;      /* err = -EADDRNOTAVAIL */
 377                 rv = !0;
 378                 for (i = 0; i < psl->sl_count; i++) {
 379                         rv = !ipv6_addr_equal(&psl->sl_addr[i], source);
 380                         if (rv == 0)
 381                                 break;
 382                 }
 383                 if (rv)         /* source not found */
 384                         goto done;      /* err = -EADDRNOTAVAIL */
 385
 386                 /* special case - (INCLUDE, empty) == LEAVE_GROUP */
 387                 if (psl->sl_count == 1 && omode == MCAST_INCLUDE) {
 388                         leavegroup = 1;
 389                         goto done;
 390                 }
 391
 392                 /* update the interface filter */
 393                 ip6_mc_del_src(idev, group, omode, 1, source, 1);
 394
 395                 for (j = i+1; j < psl->sl_count; j++)
 396                         psl->sl_addr[j-1] = psl->sl_addr[j];
 397                 psl->sl_count--;
 398                 err = 0;
 399                 goto done;
 400         }
 401         /* else, add a new source to the filter */
 402
 403         if (psl && psl->sl_count >= sysctl_mld_max_msf) {
 404                 err = -ENOBUFS;
 405                 goto done;
 406         }
 407         if (!psl || psl->sl_count == psl->sl_max) {
 408                 struct ip6_sf_socklist *newpsl;
 409                 int count = IP6_SFBLOCK;
 410
 411                 if (psl)
 412                         count += psl->sl_max;
 413                 newpsl = sock_kmalloc(sk, IP6_SFLSIZE(count), GFP_ATOMIC);
 414                 if (!newpsl) {
 415                         err = -ENOBUFS;
 416                         goto done;
 417                 }
 418                 newpsl->sl_max = count;
 419                 newpsl->sl_count = count - IP6_SFBLOCK;
 420                 if (psl) {
 421                         for (i = 0; i < psl->sl_count; i++)
 422                                 newpsl->sl_addr[i] = psl->sl_addr[i];
 423                         sock_kfree_s(sk, psl, IP6_SFLSIZE(psl->sl_max));
 424                 }
 425                 pmc->sflist = psl = newpsl;
 426         }
 427         rv = 1; /* > 0 for insert logic below if sl_count is 0 */
 428         for (i = 0; i < psl->sl_count; i++) {
 429                 rv = !ipv6_addr_equal(&psl->sl_addr[i], source);
 430                 if (rv == 0) /* There is an error in the address. */
 431                         goto done;
 432         }
 433         for (j = psl->sl_count-1; j >= i; j--)
 434                 psl->sl_addr[j+1] = psl->sl_addr[j];
 435         psl->sl_addr[i] = *source;
 436         psl->sl_count++;
 437         err = 0;
 438         /* update the interface list */
 439         ip6_mc_add_src(idev, group, omode, 1, source, 1);
 440 done:
 441         if (pmclocked)
 442                 write_unlock(&pmc->sflock);
 443         read_unlock_bh(&idev->lock);
 444         rcu_read_unlock();
 445         if (leavegroup)
 446                 err = ipv6_sock_mc_drop(sk, pgsr->gsr_interface, group);
 447         return err;
 448 }
 449
 450 int ip6_mc_msfilter(struct sock *sk, struct group_filter *gsf)
 451 {
 452         const struct in6_addr *group;
 453         struct ipv6_mc_socklist *pmc;
 454         struct inet6_dev *idev;
 455         struct ipv6_pinfo *inet6 = inet6_sk(sk);
 456         struct ip6_sf_socklist *newpsl, *psl;
 457         struct net *net = sock_net(sk);
 458         int leavegroup = 0;
 459         int i, err;
 460
 461         group = &((struct sockaddr_in6 *)&gsf->gf_group)->sin6_addr;
 462
 463         if (!ipv6_addr_is_multicast(group))
 464                 return -EINVAL;
 465         if (gsf->gf_fmode != MCAST_INCLUDE &&
 466             gsf->gf_fmode != MCAST_EXCLUDE)
 467                 return -EINVAL;
 468
 469         rcu_read_lock();
 470         idev = ip6_mc_find_dev_rcu(net, group, gsf->gf_interface);
 471
 472         if (!idev) {
 473                 rcu_read_unlock();
 474                 return -ENODEV;
 475         }
 476
 477         err = 0;
 478
 479         if (gsf->gf_fmode == MCAST_INCLUDE && gsf->gf_numsrc == 0) {
 480                 leavegroup = 1;
 481                 goto done;
 482         }
 483
 484         for_each_pmc_rcu(inet6, pmc) {
 485                 if (pmc->ifindex != gsf->gf_interface)
 486                         continue;
 487                 if (ipv6_addr_equal(&pmc->addr, group))
 488                         break;
 489         }
 490         if (!pmc) {             /* must have a prior join */
 491                 err = -EINVAL;
 492                 goto done;
 493         }
 494         if (gsf->gf_numsrc) {
 495                 newpsl = sock_kmalloc(sk, IP6_SFLSIZE(gsf->gf_numsrc),
 496                                                           GFP_ATOMIC);
 497                 if (!newpsl) {
 498                         err = -ENOBUFS;
 499                         goto done;
 500                 }
 501                 newpsl->sl_max = newpsl->sl_count = gsf->gf_numsrc;
 502                 for (i = 0; i < newpsl->sl_count; ++i) {
 503                         struct sockaddr_in6 *psin6;
 504
 505                         psin6 = (struct sockaddr_in6 *)&gsf->gf_slist[i];
 506                         newpsl->sl_addr[i] = psin6->sin6_addr;
 507                 }
 508                 err = ip6_mc_add_src(idev, group, gsf->gf_fmode,
 509                         newpsl->sl_count, newpsl->sl_addr, 0);
 510                 if (err) {
 511                         sock_kfree_s(sk, newpsl, IP6_SFLSIZE(newpsl->sl_max));
 512                         goto done;
 513                 }
 514         } else {
 515                 newpsl = NULL;
 516                 (void) ip6_mc_add_src(idev, group, gsf->gf_fmode, 0, NULL, 0);
 517         }
 518
 519         write_lock(&pmc->sflock);
 520         psl = pmc->sflist;
 521         if (psl) {
 522                 (void) ip6_mc_del_src(idev, group, pmc->sfmode,
 523                         psl->sl_count, psl->sl_addr, 0);
 524                 sock_kfree_s(sk, psl, IP6_SFLSIZE(psl->sl_max));
 525         } else
 526                 (void) ip6_mc_del_src(idev, group, pmc->sfmode, 0, NULL, 0);
 527         pmc->sflist = newpsl;
 528         pmc->sfmode = gsf->gf_fmode;
 529         write_unlock(&pmc->sflock);
 530         err = 0;
 531 done:
 532         read_unlock_bh(&idev->lock);
 533         rcu_read_unlock();
 534         if (leavegroup)
 535                 err = ipv6_sock_mc_drop(sk, gsf->gf_interface, group);
 536         return err;
 537 }
 538
 539 int ip6_mc_msfget(struct sock *sk, struct group_filter *gsf,
 540         struct group_filter __user *optval, int __user *optlen)
 541 {
 542         int err, i, count, copycount;
 543         const struct in6_addr *group;
 544         struct ipv6_mc_socklist *pmc;
 545         struct inet6_dev *idev;
 546         struct ipv6_pinfo *inet6 = inet6_sk(sk);
 547         struct ip6_sf_socklist *psl;
 548         struct net *net = sock_net(sk);
 549
 550         group = &((struct sockaddr_in6 *)&gsf->gf_group)->sin6_addr;
 551
 552         if (!ipv6_addr_is_multicast(group))
 553                 return -EINVAL;
 554
 555         rcu_read_lock();
 556         idev = ip6_mc_find_dev_rcu(net, group, gsf->gf_interface);
 557
 558         if (!idev) {
 559                 rcu_read_unlock();
 560                 return -ENODEV;
 561         }
 562
 563         err = -EADDRNOTAVAIL;
 564         /* changes to the ipv6_mc_list require the socket lock and
 565          * rtnl lock. We have the socket lock and rcu read lock,
 566          * so reading the list is safe.
 567          */
 568
 569         for_each_pmc_rcu(inet6, pmc) {
 570                 if (pmc->ifindex != gsf->gf_interface)
 571                         continue;
 572                 if (ipv6_addr_equal(group, &pmc->addr))
 573                         break;
 574         }
 575         if (!pmc)               /* must have a prior join */
 576                 goto done;
 577         gsf->gf_fmode = pmc->sfmode;
 578         psl = pmc->sflist;
 579         count = psl ? psl->sl_count : 0;
 580         read_unlock_bh(&idev->lock);
 581         rcu_read_unlock();
 582
 583         copycount = count < gsf->gf_numsrc ? count : gsf->gf_numsrc;
 584         gsf->gf_numsrc = count;
 585         if (put_user(GROUP_FILTER_SIZE(copycount), optlen) ||
 586             copy_to_user(optval, gsf, GROUP_FILTER_SIZE(0))) {
 587                 return -EFAULT;
 588         }
 589         /* changes to psl require the socket lock, and a write lock
 590          * on pmc->sflock. We have the socket lock so reading here is safe.
 591          */
 592         for (i = 0; i < copycount; i++) {
 593                 struct sockaddr_in6 *psin6;
 594                 struct sockaddr_storage ss;
 595
 596                 psin6 = (struct sockaddr_in6 *)&ss;
 597                 memset(&ss, 0, sizeof(ss));
 598                 psin6->sin6_family = AF_INET6;
 599                 psin6->sin6_addr = psl->sl_addr[i];
 600                 if (copy_to_user(&optval->gf_slist[i], &ss, sizeof(ss)))
 601                         return -EFAULT;
 602         }
 603         return 0;
 604 done:
 605         read_unlock_bh(&idev->lock);
 606         rcu_read_unlock();
 607         return err;
 608 }
 609
 610 bool inet6_mc_check(struct sock *sk, const struct in6_addr *mc_addr,
 611                     const struct in6_addr *src_addr)
 612 {
 613         struct ipv6_pinfo *np = inet6_sk(sk);
 614         struct ipv6_mc_socklist *mc;
 615         struct ip6_sf_socklist *psl;
 616         bool rv = true;
 617
 618         rcu_read_lock();
 619         for_each_pmc_rcu(np, mc) {
 620                 if (ipv6_addr_equal(&mc->addr, mc_addr))
 621                         break;
 622         }
 623         if (!mc) {
 624                 rcu_read_unlock();
 625                 return true;
 626         }
 627         read_lock(&mc->sflock);
 628         psl = mc->sflist;
 629         if (!psl) {
 630                 rv = mc->sfmode == MCAST_EXCLUDE;
 631         } else {
 632                 int i;
 633
 634                 for (i = 0; i < psl->sl_count; i++) {
 635                         if (ipv6_addr_equal(&psl->sl_addr[i], src_addr))
 636                                 break;
 637                 }
 638                 if (mc->sfmode == MCAST_INCLUDE && i >= psl->sl_count)
 639                         rv = false;
 640                 if (mc->sfmode == MCAST_EXCLUDE && i < psl->sl_count)
 641                         rv = false;
 642         }
 643         read_unlock(&mc->sflock);
 644         rcu_read_unlock();
 645
 646         return rv;
 647 }
 648
 649 static void igmp6_group_added(struct ifmcaddr6 *mc)
 650 {
 651         struct net_device *dev = mc->idev->dev;
 652         char buf[MAX_ADDR_LEN];
 653
 654         if (IPV6_ADDR_MC_SCOPE(&mc->mca_addr) <
 655             IPV6_ADDR_SCOPE_LINKLOCAL)
 656                 return;
 657
 658         spin_lock_bh(&mc->mca_lock);
 659         if (!(mc->mca_flags&MAF_LOADED)) {
 660                 mc->mca_flags |= MAF_LOADED;
 661                 if (ndisc_mc_map(&mc->mca_addr, buf, dev, 0) == 0)
 662                         dev_mc_add(dev, buf);
 663         }
 664         spin_unlock_bh(&mc->mca_lock);
 665
 666         if (!(dev->flags & IFF_UP) || (mc->mca_flags & MAF_NOREPORT))
 667                 return;
 668
 669         if (mld_in_v1_mode(mc->idev)) {
 670                 igmp6_join_group(mc);
 671                 return;
 672         }
 673         /* else v2 */
 674
 675         mc->mca_crcount = mc->idev->mc_qrv;
 676         mld_ifc_event(mc->idev);
 677 }
 678
 679 static void igmp6_group_dropped(struct ifmcaddr6 *mc)
 680 {
 681         struct net_device *dev = mc->idev->dev;
 682         char buf[MAX_ADDR_LEN];
 683
 684         if (IPV6_ADDR_MC_SCOPE(&mc->mca_addr) <
 685             IPV6_ADDR_SCOPE_LINKLOCAL)
 686                 return;
 687
 688         spin_lock_bh(&mc->mca_lock);
 689         if (mc->mca_flags&MAF_LOADED) {
 690                 mc->mca_flags &= ~MAF_LOADED;
 691                 if (ndisc_mc_map(&mc->mca_addr, buf, dev, 0) == 0)
 692                         dev_mc_del(dev, buf);
 693         }
 694
 695         if (mc->mca_flags & MAF_NOREPORT)
 696                 goto done;
 697         spin_unlock_bh(&mc->mca_lock);
 698
 699         if (!mc->idev->dead)
 700                 igmp6_leave_group(mc);
 701
 702         spin_lock_bh(&mc->mca_lock);
 703         if (del_timer(&mc->mca_timer))
 704                 atomic_dec(&mc->mca_refcnt);
 705 done:
 706         ip6_mc_clear_src(mc);
 707         spin_unlock_bh(&mc->mca_lock);
 708 }
 709
 710 /*
 711  * deleted ifmcaddr6 manipulation
 712  */
 713 static void mld_add_delrec(struct inet6_dev *idev, struct ifmcaddr6 *im)
 714 {
 715         struct ifmcaddr6 *pmc;
 716
 717         /* this is an "ifmcaddr6" for convenience; only the fields below
 718          * are actually used. In particular, the refcnt and users are not
 719          * used for management of the delete list. Using the same structure
 720          * for deleted items allows change reports to use common code with
 721          * non-deleted or query-response MCA's.
 722          */
 723         pmc = kzalloc(sizeof(*pmc), GFP_ATOMIC);
 724         if (!pmc)
 725                 return;
 726
 727         spin_lock_bh(&im->mca_lock);
 728         spin_lock_init(&pmc->mca_lock);
 729         pmc->idev = im->idev;
 730         in6_dev_hold(idev);
 731         pmc->mca_addr = im->mca_addr;
 732         pmc->mca_crcount = idev->mc_qrv;
 733         pmc->mca_sfmode = im->mca_sfmode;
 734         if (pmc->mca_sfmode == MCAST_INCLUDE) {
 735                 struct ip6_sf_list *psf;
 736
 737                 pmc->mca_tomb = im->mca_tomb;
 738                 pmc->mca_sources = im->mca_sources;
 739                 im->mca_tomb = im->mca_sources = NULL;
 740                 for (psf = pmc->mca_sources; psf; psf = psf->sf_next)
 741                         psf->sf_crcount = pmc->mca_crcount;
 742         }
 743         spin_unlock_bh(&im->mca_lock);
 744
 745         spin_lock_bh(&idev->mc_lock);
 746         pmc->next = idev->mc_tomb;
 747         idev->mc_tomb = pmc;
 748         spin_unlock_bh(&idev->mc_lock);
 749 }
 750
 751 static void mld_del_delrec(struct inet6_dev *idev, const struct in6_addr *pmca)
 752 {
 753         struct ifmcaddr6 *pmc, *pmc_prev;
 754         struct ip6_sf_list *psf, *psf_next;
 755
 756         spin_lock_bh(&idev->mc_lock);
 757         pmc_prev = NULL;
 758         for (pmc = idev->mc_tomb; pmc; pmc = pmc->next) {
 759                 if (ipv6_addr_equal(&pmc->mca_addr, pmca))
 760                         break;
 761                 pmc_prev = pmc;
 762         }
 763         if (pmc) {
 764                 if (pmc_prev)
 765                         pmc_prev->next = pmc->next;
 766                 else
 767                         idev->mc_tomb = pmc->next;
 768         }
 769         spin_unlock_bh(&idev->mc_lock);
 770
 771         if (pmc) {
 772                 for (psf = pmc->mca_tomb; psf; psf = psf_next) {
 773                         psf_next = psf->sf_next;
 774                         kfree(psf);
 775                 }
 776                 in6_dev_put(pmc->idev);
 777                 kfree(pmc);
 778         }
 779 }
 780
 781 static void mld_clear_delrec(struct inet6_dev *idev)
 782 {
 783         struct ifmcaddr6 *pmc, *nextpmc;
 784
 785         spin_lock_bh(&idev->mc_lock);
 786         pmc = idev->mc_tomb;
 787         idev->mc_tomb = NULL;
 788         spin_unlock_bh(&idev->mc_lock);
 789
 790         for (; pmc; pmc = nextpmc) {
 791                 nextpmc = pmc->next;
 792                 ip6_mc_clear_src(pmc);
 793                 in6_dev_put(pmc->idev);
 794                 kfree(pmc);
 795         }
 796
 797         /* clear dead sources, too */
 798         read_lock_bh(&idev->lock);
 799         for (pmc = idev->mc_list; pmc; pmc = pmc->next) {
 800                 struct ip6_sf_list *psf, *psf_next;
 801
 802                 spin_lock_bh(&pmc->mca_lock);
 803                 psf = pmc->mca_tomb;
 804                 pmc->mca_tomb = NULL;
 805                 spin_unlock_bh(&pmc->mca_lock);
 806                 for (; psf; psf = psf_next) {
 807                         psf_next = psf->sf_next;
 808                         kfree(psf);
 809                 }
 810         }
 811         read_unlock_bh(&idev->lock);
 812 }
 813
 814 static void mca_get(struct ifmcaddr6 *mc)
 815 {
 816         atomic_inc(&mc->mca_refcnt);
 817 }
 818
 819 static void ma_put(struct ifmcaddr6 *mc)
 820 {
 821         if (atomic_dec_and_test(&mc->mca_refcnt)) {
 822                 in6_dev_put(mc->idev);
 823                 kfree(mc);
 824         }
 825 }
 826
 827 static struct ifmcaddr6 *mca_alloc(struct inet6_dev *idev,
 828                                    const struct in6_addr *addr)
 829 {
 830         struct ifmcaddr6 *mc;
 831
 832         mc = kzalloc(sizeof(*mc), GFP_ATOMIC);
 833         if (!mc)
 834                 return NULL;
 835
 836         setup_timer(&mc->mca_timer, igmp6_timer_handler, (unsigned long)mc);
 837
 838         mc->mca_addr = *addr;
 839         mc->idev = idev; /* reference taken by caller */
 840         mc->mca_users = 1;
 841         /* mca_stamp should be updated upon changes */
 842         mc->mca_cstamp = mc->mca_tstamp = jiffies;
 843         atomic_set(&mc->mca_refcnt, 1);
 844         spin_lock_init(&mc->mca_lock);
 845
 846         /* initial mode is (EX, empty) */
 847         mc->mca_sfmode = MCAST_EXCLUDE;
 848         mc->mca_sfcount[MCAST_EXCLUDE] = 1;
 849
 850         if (ipv6_addr_is_ll_all_nodes(&mc->mca_addr) ||
 851             IPV6_ADDR_MC_SCOPE(&mc->mca_addr) < IPV6_ADDR_SCOPE_LINKLOCAL)
 852                 mc->mca_flags |= MAF_NOREPORT;
 853
 854         return mc;
 855 }
 856
 857 /*
 858  *      device multicast group inc (add if not found)
 859  */
 860 int ipv6_dev_mc_inc(struct net_device *dev, const struct in6_addr *addr)
 861 {
 862         struct ifmcaddr6 *mc;
 863         struct inet6_dev *idev;
 864
 865         ASSERT_RTNL();
 866
 867         /* we need to take a reference on idev */
 868         idev = in6_dev_get(dev);
 869
 870         if (!idev)
 871                 return -EINVAL;
 872
 873         write_lock_bh(&idev->lock);
 874         if (idev->dead) {
 875                 write_unlock_bh(&idev->lock);
 876                 in6_dev_put(idev);
 877                 return -ENODEV;
 878         }
 879
 880         for (mc = idev->mc_list; mc; mc = mc->next) {
 881                 if (ipv6_addr_equal(&mc->mca_addr, addr)) {
 882                         mc->mca_users++;
 883                         write_unlock_bh(&idev->lock);
 884                         ip6_mc_add_src(idev, &mc->mca_addr, MCAST_EXCLUDE, 0,
 885                                 NULL, 0);
 886                         in6_dev_put(idev);
 887                         return 0;
 888                 }
 889         }
 890
 891         mc = mca_alloc(idev, addr);
 892         if (!mc) {
 893                 write_unlock_bh(&idev->lock);
 894                 in6_dev_put(idev);
 895                 return -ENOMEM;
 896         }
 897
 898         mc->next = idev->mc_list;
 899         idev->mc_list = mc;
 900
 901         /* Hold this for the code below before we unlock,
 902          * it is already exposed via idev->mc_list.
 903          */
 904         mca_get(mc);
 905         write_unlock_bh(&idev->lock);
 906
 907         mld_del_delrec(idev, &mc->mca_addr);
 908         igmp6_group_added(mc);
 909         ma_put(mc);
 910         return 0;
 911 }
 912
 913 /*
 914  *      device multicast group del
 915  */
 916 int __ipv6_dev_mc_dec(struct inet6_dev *idev, const struct in6_addr *addr)
 917 {
 918         struct ifmcaddr6 *ma, **map;
 919
 920         ASSERT_RTNL();
 921
 922         write_lock_bh(&idev->lock);
 923         for (map = &idev->mc_list; (ma = *map) != NULL; map = &ma->next) {
 924                 if (ipv6_addr_equal(&ma->mca_addr, addr)) {
 925                         if (--ma->mca_users == 0) {
 926                                 *map = ma->next;
 927                                 write_unlock_bh(&idev->lock);
 928
 929                                 igmp6_group_dropped(ma);
 930
 931                                 ma_put(ma);
 932                                 return 0;
 933                         }
 934                         write_unlock_bh(&idev->lock);
 935                         return 0;
 936                 }
 937         }
 938         write_unlock_bh(&idev->lock);
 939
 940         return -ENOENT;
 941 }
 942
 943 int ipv6_dev_mc_dec(struct net_device *dev, const struct in6_addr *addr)
 944 {
 945         struct inet6_dev *idev;
 946         int err;
 947
 948         ASSERT_RTNL();
 949
 950         idev = __in6_dev_get(dev);
 951         if (!idev)
 952                 err = -ENODEV;
 953         else
 954                 err = __ipv6_dev_mc_dec(idev, addr);
 955
 956         return err;
 957 }
 958
 959 /*
 960  *      check if the interface/address pair is valid
 961  */
 962 bool ipv6_chk_mcast_addr(struct net_device *dev, const struct in6_addr *group,
 963                          const struct in6_addr *src_addr)
 964 {
 965         struct inet6_dev *idev;
 966         struct ifmcaddr6 *mc;
 967         bool rv = false;
 968
 969         rcu_read_lock();
 970         idev = __in6_dev_get(dev);
 971         if (idev) {
 972                 read_lock_bh(&idev->lock);
 973                 for (mc = idev->mc_list; mc; mc = mc->next) {
 974                         if (ipv6_addr_equal(&mc->mca_addr, group))
 975                                 break;
 976                 }
 977                 if (mc) {
 978                         if (src_addr && !ipv6_addr_any(src_addr)) {
 979                                 struct ip6_sf_list *psf;
 980
 981                                 spin_lock_bh(&mc->mca_lock);
 982                                 for (psf = mc->mca_sources; psf; psf = psf->sf_next) {
 983                                         if (ipv6_addr_equal(&psf->sf_addr, src_addr))
 984                                                 break;
 985                                 }
 986                                 if (psf)
 987                                         rv = psf->sf_count[MCAST_INCLUDE] ||
 988                                                 psf->sf_count[MCAST_EXCLUDE] !=
 989                                                 mc->mca_sfcount[MCAST_EXCLUDE];
 990                                 else
 991                                         rv = mc->mca_sfcount[MCAST_EXCLUDE] != 0;
 992                                 spin_unlock_bh(&mc->mca_lock);
 993                         } else
 994                                 rv = true; /* don't filter unspecified source */
 995                 }
 996                 read_unlock_bh(&idev->lock);
 997         }
 998         rcu_read_unlock();
 999         return rv;
1000 }
1001
1002 static void mld_gq_start_timer(struct inet6_dev *idev)
1003 {
1004         unsigned long tv = prandom_u32() % idev->mc_maxdelay;
1005
1006         idev->mc_gq_running = 1;
1007         if (!mod_timer(&idev->mc_gq_timer, jiffies+tv+2))
1008                 in6_dev_hold(idev);
1009 }
1010
1011 static void mld_gq_stop_timer(struct inet6_dev *idev)
1012 {
1013         idev->mc_gq_running = 0;
1014         if (del_timer(&idev->mc_gq_timer))
1015                 __in6_dev_put(idev);
1016 }
1017
1018 static void mld_ifc_start_timer(struct inet6_dev *idev, unsigned long delay)
1019 {
1020         unsigned long tv = prandom_u32() % delay;
1021
1022         if (!mod_timer(&idev->mc_ifc_timer, jiffies+tv+2))
1023                 in6_dev_hold(idev);
1024 }
1025
1026 static void mld_ifc_stop_timer(struct inet6_dev *idev)
1027 {
1028         idev->mc_ifc_count = 0;
1029         if (del_timer(&idev->mc_ifc_timer))
1030                 __in6_dev_put(idev);
1031 }
1032
1033 static void mld_dad_start_timer(struct inet6_dev *idev, unsigned long delay)
1034 {
1035         unsigned long tv = prandom_u32() % delay;
1036
1037         if (!mod_timer(&idev->mc_dad_timer, jiffies+tv+2))
1038                 in6_dev_hold(idev);
1039 }
1040
1041 static void mld_dad_stop_timer(struct inet6_dev *idev)
1042 {
1043         if (del_timer(&idev->mc_dad_timer))
1044                 __in6_dev_put(idev);
1045 }
1046
1047 /*
1048  *      IGMP handling (alias multicast ICMPv6 messages)
1049  */
1050
1051 static void igmp6_group_queried(struct ifmcaddr6 *ma, unsigned long resptime)
1052 {
1053         unsigned long delay = resptime;
1054
1055         /* Do not start timer for these addresses */
1056         if (ipv6_addr_is_ll_all_nodes(&ma->mca_addr) ||
1057             IPV6_ADDR_MC_SCOPE(&ma->mca_addr) < IPV6_ADDR_SCOPE_LINKLOCAL)
1058                 return;
1059
1060         if (del_timer(&ma->mca_timer)) {
1061                 atomic_dec(&ma->mca_refcnt);
1062                 delay = ma->mca_timer.expires - jiffies;
1063         }
1064
1065         if (delay >= resptime)
1066                 delay = prandom_u32() % resptime;
1067
1068         ma->mca_timer.expires = jiffies + delay;
1069         if (!mod_timer(&ma->mca_timer, jiffies + delay))
1070                 atomic_inc(&ma->mca_refcnt);
1071         ma->mca_flags |= MAF_TIMER_RUNNING;
1072 }
1073
1074 /* mark EXCLUDE-mode sources */
1075 static bool mld_xmarksources(struct ifmcaddr6 *pmc, int nsrcs,
1076                              const struct in6_addr *srcs)
1077 {
1078         struct ip6_sf_list *psf;
1079         int i, scount;
1080
1081         scount = 0;
1082         for (psf = pmc->mca_sources; psf; psf = psf->sf_next) {
1083                 if (scount == nsrcs)
1084                         break;
1085                 for (i = 0; i < nsrcs; i++) {
1086                         /* skip inactive filters */
1087                         if (psf->sf_count[MCAST_INCLUDE] ||
1088                             pmc->mca_sfcount[MCAST_EXCLUDE] !=
1089                             psf->sf_count[MCAST_EXCLUDE])
1090                                 break;
1091                         if (ipv6_addr_equal(&srcs[i], &psf->sf_addr)) {
1092                                 scount++;
1093                                 break;
1094                         }
1095                 }
1096         }
1097         pmc->mca_flags &= ~MAF_GSQUERY;
1098         if (scount == nsrcs)    /* all sources excluded */
1099                 return false;
1100         return true;
1101 }
1102
1103 static bool mld_marksources(struct ifmcaddr6 *pmc, int nsrcs,
1104                             const struct in6_addr *srcs)
1105 {
1106         struct ip6_sf_list *psf;
1107         int i, scount;
1108
1109         if (pmc->mca_sfmode == MCAST_EXCLUDE)
1110                 return mld_xmarksources(pmc, nsrcs, srcs);
1111
1112         /* mark INCLUDE-mode sources */
1113
1114         scount = 0;
1115         for (psf = pmc->mca_sources; psf; psf = psf->sf_next) {
1116                 if (scount == nsrcs)
1117                         break;
1118                 for (i = 0; i < nsrcs; i++) {
1119                         if (ipv6_addr_equal(&srcs[i], &psf->sf_addr)) {
1120                                 psf->sf_gsresp = 1;
1121                                 scount++;
1122                                 break;
1123                         }
1124                 }
1125         }
1126         if (!scount) {
1127                 pmc->mca_flags &= ~MAF_GSQUERY;
1128                 return false;
1129         }
1130         pmc->mca_flags |= MAF_GSQUERY;
1131         return true;
1132 }
1133
1134 static int mld_force_mld_version(const struct inet6_dev *idev)
1135 {
1136         /* Normally, both are 0 here. If enforcement to a particular is
1137          * being used, individual device enforcement will have a lower
1138          * precedence over 'all' device (.../conf/all/force_mld_version).
1139          */
1140
1141         if (dev_net(idev->dev)->ipv6.devconf_all->force_mld_version != 0)
1142                 return dev_net(idev->dev)->ipv6.devconf_all->force_mld_version;
1143         else
1144                 return idev->cnf.force_mld_version;
1145 }
1146
1147 static bool mld_in_v2_mode_only(const struct inet6_dev *idev)
1148 {
1149         return mld_force_mld_version(idev) == 2;
1150 }
1151
1152 static bool mld_in_v1_mode_only(const struct inet6_dev *idev)
1153 {
1154         return mld_force_mld_version(idev) == 1;
1155 }
1156
1157 static bool mld_in_v1_mode(const struct inet6_dev *idev)
1158 {
1159         if (mld_in_v2_mode_only(idev))
1160                 return false;
1161         if (mld_in_v1_mode_only(idev))
1162                 return true;
1163         if (idev->mc_v1_seen && time_before(jiffies, idev->mc_v1_seen))
1164                 return true;
1165
1166         return false;
1167 }
1168
1169 static void mld_set_v1_mode(struct inet6_dev *idev)
1170 {
1171         /* RFC3810, relevant sections:
1172          *  - 9.1. Robustness Variable
1173          *  - 9.2. Query Interval
1174          *  - 9.3. Query Response Interval
1175          *  - 9.12. Older Version Querier Present Timeout
1176          */
1177         unsigned long switchback;
1178
1179         switchback = (idev->mc_qrv * idev->mc_qi) + idev->mc_qri;
1180
1181         idev->mc_v1_seen = jiffies + switchback;
1182 }
1183
1184 static void mld_update_qrv(struct inet6_dev *idev,
1185                            const struct mld2_query *mlh2)
1186 {
1187         /* RFC3810, relevant sections:
1188          *  - 5.1.8. QRV (Querier's Robustness Variable)
1189          *  - 9.1. Robustness Variable
1190          */
1191
1192         /* The value of the Robustness Variable MUST NOT be zero,
1193          * and SHOULD NOT be one. Catch this here if we ever run
1194          * into such a case in future.
1195          */
1196         const int min_qrv = min(MLD_QRV_DEFAULT, sysctl_mld_qrv);
1197         WARN_ON(idev->mc_qrv == 0);
1198
1199         if (mlh2->mld2q_qrv > 0)
1200                 idev->mc_qrv = mlh2->mld2q_qrv;
1201
1202         if (unlikely(idev->mc_qrv < min_qrv)) {
1203                 net_warn_ratelimited("IPv6: MLD: clamping QRV from %u to %u!\n",
1204                                      idev->mc_qrv, min_qrv);
1205                 idev->mc_qrv = min_qrv;
1206         }
1207 }
1208
1209 static void mld_update_qi(struct inet6_dev *idev,
1210                           const struct mld2_query *mlh2)
1211 {
1212         /* RFC3810, relevant sections:
1213          *  - 5.1.9. QQIC (Querier's Query Interval Code)
1214          *  - 9.2. Query Interval
1215          *  - 9.12. Older Version Querier Present Timeout
1216          *    (the [Query Interval] in the last Query received)
1217          */
1218         unsigned long mc_qqi;
1219
1220         if (mlh2->mld2q_qqic < 128) {
1221                 mc_qqi = mlh2->mld2q_qqic;
1222         } else {
1223                 unsigned long mc_man, mc_exp;
1224
1225                 mc_exp = MLDV2_QQIC_EXP(mlh2->mld2q_qqic);
1226                 mc_man = MLDV2_QQIC_MAN(mlh2->mld2q_qqic);
1227
1228                 mc_qqi = (mc_man | 0x10) << (mc_exp + 3);
1229         }
1230
1231         idev->mc_qi = mc_qqi * HZ;
1232 }
1233
1234 static void mld_update_qri(struct inet6_dev *idev,
1235                            const struct mld2_query *mlh2)
1236 {
1237         /* RFC3810, relevant sections:
1238          *  - 5.1.3. Maximum Response Code
1239          *  - 9.3. Query Response Interval
1240          */
1241         idev->mc_qri = msecs_to_jiffies(mldv2_mrc(mlh2));
1242 }
1243
1244 static int mld_process_v1(struct inet6_dev *idev, struct mld_msg *mld,
1245                           unsigned long *max_delay, bool v1_query)
1246 {
1247         unsigned long mldv1_md;
1248
1249         /* Ignore v1 queries */
1250         if (mld_in_v2_mode_only(idev))
1251                 return -EINVAL;
1252
1253         mldv1_md = ntohs(mld->mld_maxdelay);
1254
1255         /* When in MLDv1 fallback and a MLDv2 router start-up being
1256          * unaware of current MLDv1 operation, the MRC == MRD mapping
1257          * only works when the exponential algorithm is not being
1258          * used (as MLDv1 is unaware of such things).
1259          *
1260          * According to the RFC author, the MLDv2 implementations
1261          * he's aware of all use a MRC < 32768 on start up queries.
1262          *
1263          * Thus, should we *ever* encounter something else larger
1264          * than that, just assume the maximum possible within our
1265          * reach.
1266          */
1267         if (!v1_query)
1268                 mldv1_md = min(mldv1_md, MLDV1_MRD_MAX_COMPAT);
1269
1270         *max_delay = max(msecs_to_jiffies(mldv1_md), 1UL);
1271
1272         /* MLDv1 router present: we need to go into v1 mode *only*
1273          * when an MLDv1 query is received as per section 9.12. of
1274          * RFC3810! And we know from RFC2710 section 3.7 that MLDv1
1275          * queries MUST be of exactly 24 octets.
1276          */
1277         if (v1_query)
1278                 mld_set_v1_mode(idev);
1279
1280         /* cancel MLDv2 report timer */
1281         mld_gq_stop_timer(idev);
1282         /* cancel the interface change timer */
1283         mld_ifc_stop_timer(idev);
1284         /* clear deleted report items */
1285         mld_clear_delrec(idev);
1286
1287         return 0;
1288 }
1289
1290 static int mld_process_v2(struct inet6_dev *idev, struct mld2_query *mld,
1291                           unsigned long *max_delay)
1292 {
1293         *max_delay = max(msecs_to_jiffies(mldv2_mrc(mld)), 1UL);
1294
1295         mld_update_qrv(idev, mld);
1296         mld_update_qi(idev, mld);
1297         mld_update_qri(idev, mld);
1298
1299         idev->mc_maxdelay = *max_delay;
1300
1301         return 0;
1302 }
1303
1304 /* called with rcu_read_lock() */
1305 int igmp6_event_query(struct sk_buff *skb)
1306 {
1307         struct mld2_query *mlh2 = NULL;
1308         struct ifmcaddr6 *ma;
1309         const struct in6_addr *group;
1310         unsigned long max_delay;
1311         struct inet6_dev *idev;
1312         struct mld_msg *mld;
1313         int group_type;
1314         int mark = 0;
1315         int len, err;
1316
1317         if (!pskb_may_pull(skb, sizeof(struct in6_addr)))
1318                 return -EINVAL;
1319
1320         /* compute payload length excluding extension headers */
1321         len = ntohs(ipv6_hdr(skb)->payload_len) + sizeof(struct ipv6hdr);
1322         len -= skb_network_header_len(skb);
1323
1324         /* RFC3810 6.2
1325          * Upon reception of an MLD message that contains a Query, the node
1326          * checks if the source address of the message is a valid link-local
1327          * address, if the Hop Limit is set to 1, and if the Router Alert
1328          * option is present in the Hop-By-Hop Options header of the IPv6
1329          * packet.  If any of these checks fails, the packet is dropped.
1330          */
1331         if (!(ipv6_addr_type(&ipv6_hdr(skb)->saddr) & IPV6_ADDR_LINKLOCAL) ||
1332             ipv6_hdr(skb)->hop_limit != 1 ||
1333             !(IP6CB(skb)->flags & IP6SKB_ROUTERALERT) ||
1334             IP6CB(skb)->ra != htons(IPV6_OPT_ROUTERALERT_MLD))
1335                 return -EINVAL;
1336
1337         idev = __in6_dev_get(skb->dev);
1338         if (!idev)
1339                 return 0;
1340
1341         mld = (struct mld_msg *)icmp6_hdr(skb);
1342         group = &mld->mld_mca;
1343         group_type = ipv6_addr_type(group);
1344
1345         if (group_type != IPV6_ADDR_ANY &&
1346             !(group_type&IPV6_ADDR_MULTICAST))
1347                 return -EINVAL;
1348
1349         if (len < MLD_V1_QUERY_LEN) {
1350                 return -EINVAL;
1351         } else if (len == MLD_V1_QUERY_LEN || mld_in_v1_mode(idev)) {
1352                 err = mld_process_v1(idev, mld, &max_delay,
1353                                      len == MLD_V1_QUERY_LEN);
1354                 if (err < 0)
1355                         return err;
1356         } else if (len >= MLD_V2_QUERY_LEN_MIN) {
1357                 int srcs_offset = sizeof(struct mld2_query) -
1358                                   sizeof(struct icmp6hdr);
1359
1360                 if (!pskb_may_pull(skb, srcs_offset))
1361                         return -EINVAL;
1362
1363                 mlh2 = (struct mld2_query *)skb_transport_header(skb);
1364
1365                 err = mld_process_v2(idev, mlh2, &max_delay);
1366                 if (err < 0)
1367                         return err;
1368
1369                 if (group_type == IPV6_ADDR_ANY) { /* general query */
1370                         if (mlh2->mld2q_nsrcs)
1371                                 return -EINVAL; /* no sources allowed */
1372
1373                         mld_gq_start_timer(idev);
1374                         return 0;
1375                 }
1376                 /* mark sources to include, if group & source-specific */
1377                 if (mlh2->mld2q_nsrcs != 0) {
1378                         if (!pskb_may_pull(skb, srcs_offset +
1379                             ntohs(mlh2->mld2q_nsrcs) * sizeof(struct in6_addr)))
1380                                 return -EINVAL;
1381
1382                         mlh2 = (struct mld2_query *)skb_transport_header(skb);
1383                         mark = 1;
1384                 }
1385         } else {
1386                 return -EINVAL;
1387         }
1388
1389         read_lock_bh(&idev->lock);
1390         if (group_type == IPV6_ADDR_ANY) {
1391                 for (ma = idev->mc_list; ma; ma = ma->next) {
1392                         spin_lock_bh(&ma->mca_lock);
1393                         igmp6_group_queried(ma, max_delay);
1394                         spin_unlock_bh(&ma->mca_lock);
1395                 }
1396         } else {
1397                 for (ma = idev->mc_list; ma; ma = ma->next) {
1398                         if (!ipv6_addr_equal(group, &ma->mca_addr))
1399                                 continue;
1400                         spin_lock_bh(&ma->mca_lock);
1401                         if (ma->mca_flags & MAF_TIMER_RUNNING) {
1402                                 /* gsquery <- gsquery && mark */
1403                                 if (!mark)
1404                                         ma->mca_flags &= ~MAF_GSQUERY;
1405                         } else {
1406                                 /* gsquery <- mark */
1407                                 if (mark)
1408                                         ma->mca_flags |= MAF_GSQUERY;
1409                                 else
1410                                         ma->mca_flags &= ~MAF_GSQUERY;
1411                         }
1412                         if (!(ma->mca_flags & MAF_GSQUERY) ||
1413                             mld_marksources(ma, ntohs(mlh2->mld2q_nsrcs), mlh2->mld2q_srcs))
1414                                 igmp6_group_queried(ma, max_delay);
1415                         spin_unlock_bh(&ma->mca_lock);
1416                         break;
1417                 }
1418         }
1419         read_unlock_bh(&idev->lock);
1420
1421         return 0;
1422 }
1423
1424 /* called with rcu_read_lock() */
1425 int igmp6_event_report(struct sk_buff *skb)
1426 {
1427         struct ifmcaddr6 *ma;
1428         struct inet6_dev *idev;
1429         struct mld_msg *mld;
1430         int addr_type;
1431
1432         /* Our own report looped back. Ignore it. */
1433         if (skb->pkt_type == PACKET_LOOPBACK)
1434                 return 0;
1435
1436         /* send our report if the MC router may not have heard this report */
1437         if (skb->pkt_type != PACKET_MULTICAST &&
1438             skb->pkt_type != PACKET_BROADCAST)
1439                 return 0;
1440
1441         if (!pskb_may_pull(skb, sizeof(*mld) - sizeof(struct icmp6hdr)))
1442                 return -EINVAL;
1443
1444         mld = (struct mld_msg *)icmp6_hdr(skb);
1445
1446         /* Drop reports with not link local source */
1447         addr_type = ipv6_addr_type(&ipv6_hdr(skb)->saddr);
1448         if (addr_type != IPV6_ADDR_ANY &&
1449             !(addr_type&IPV6_ADDR_LINKLOCAL))
1450                 return -EINVAL;
1451
1452         idev = __in6_dev_get(skb->dev);
1453         if (!idev)
1454                 return -ENODEV;
1455
1456         /*
1457          *      Cancel the timer for this group
1458          */
1459
1460         read_lock_bh(&idev->lock);
1461         for (ma = idev->mc_list; ma; ma = ma->next) {
1462                 if (ipv6_addr_equal(&ma->mca_addr, &mld->mld_mca)) {
1463                         spin_lock(&ma->mca_lock);
1464                         if (del_timer(&ma->mca_timer))
1465                                 atomic_dec(&ma->mca_refcnt);
1466                         ma->mca_flags &= ~(MAF_LAST_REPORTER|MAF_TIMER_RUNNING);
1467                         spin_unlock(&ma->mca_lock);
1468                         break;
1469                 }
1470         }
1471         read_unlock_bh(&idev->lock);
1472         return 0;
1473 }
1474
1475 static bool is_in(struct ifmcaddr6 *pmc, struct ip6_sf_list *psf, int type,
1476                   int gdeleted, int sdeleted)
1477 {
1478         switch (type) {
1479         case MLD2_MODE_IS_INCLUDE:
1480         case MLD2_MODE_IS_EXCLUDE:
1481                 if (gdeleted || sdeleted)
1482                         return false;
1483                 if (!((pmc->mca_flags & MAF_GSQUERY) && !psf->sf_gsresp)) {
1484                         if (pmc->mca_sfmode == MCAST_INCLUDE)
1485                                 return true;
1486                         /* don't include if this source is excluded
1487                          * in all filters
1488                          */
1489                         if (psf->sf_count[MCAST_INCLUDE])
1490                                 return type == MLD2_MODE_IS_INCLUDE;
1491                         return pmc->mca_sfcount[MCAST_EXCLUDE] ==
1492                                 psf->sf_count[MCAST_EXCLUDE];
1493                 }
1494                 return false;
1495         case MLD2_CHANGE_TO_INCLUDE:
1496                 if (gdeleted || sdeleted)
1497                         return false;
1498                 return psf->sf_count[MCAST_INCLUDE] != 0;
1499         case MLD2_CHANGE_TO_EXCLUDE:
1500                 if (gdeleted || sdeleted)
1501                         return false;
1502                 if (pmc->mca_sfcount[MCAST_EXCLUDE] == 0 ||
1503                     psf->sf_count[MCAST_INCLUDE])
1504                         return false;
1505                 return pmc->mca_sfcount[MCAST_EXCLUDE] ==
1506                         psf->sf_count[MCAST_EXCLUDE];
1507         case MLD2_ALLOW_NEW_SOURCES:
1508                 if (gdeleted || !psf->sf_crcount)
1509                         return false;
1510                 return (pmc->mca_sfmode == MCAST_INCLUDE) ^ sdeleted;
1511         case MLD2_BLOCK_OLD_SOURCES:
1512                 if (pmc->mca_sfmode == MCAST_INCLUDE)
1513                         return gdeleted || (psf->sf_crcount && sdeleted);
1514                 return psf->sf_crcount && !gdeleted && !sdeleted;
1515         }
1516         return false;
1517 }
1518
1519 static int
1520 mld_scount(struct ifmcaddr6 *pmc, int type, int gdeleted, int sdeleted)
1521 {
1522         struct ip6_sf_list *psf;
1523         int scount = 0;
1524
1525         for (psf = pmc->mca_sources; psf; psf = psf->sf_next) {
1526                 if (!is_in(pmc, psf, type, gdeleted, sdeleted))
1527                         continue;
1528                 scount++;
1529         }
1530         return scount;
1531 }
1532
1533 static void ip6_mc_hdr(struct sock *sk, struct sk_buff *skb,
1534                        struct net_device *dev,
1535                        const struct in6_addr *saddr,
1536                        const struct in6_addr *daddr,
1537                        int proto, int len)
1538 {
1539         struct ipv6hdr *hdr;
1540
1541         skb->protocol = htons(ETH_P_IPV6);
1542         skb->dev = dev;
1543
1544         skb_reset_network_header(skb);
1545         skb_put(skb, sizeof(struct ipv6hdr));
1546         hdr = ipv6_hdr(skb);
1547
1548         ip6_flow_hdr(hdr, 0, 0);
1549
1550         hdr->payload_len = htons(len);
1551         hdr->nexthdr = proto;
1552         hdr->hop_limit = inet6_sk(sk)->hop_limit;
1553
1554         hdr->saddr = *saddr;
1555         hdr->daddr = *daddr;
1556 }
1557
1558 static struct sk_buff *mld_newpack(struct inet6_dev *idev, unsigned int mtu)
1559 {
1560         struct net_device *dev = idev->dev;
1561         struct net *net = dev_net(dev);
1562         struct sock *sk = net->ipv6.igmp_sk;
1563         struct sk_buff *skb;
1564         struct mld2_report *pmr;
1565         struct in6_addr addr_buf;
1566         const struct in6_addr *saddr;
1567         int hlen = LL_RESERVED_SPACE(dev);
1568         int tlen = dev->needed_tailroom;
1569         unsigned int size = mtu + hlen + tlen;
1570         int err;
1571         u8 ra[8] = { IPPROTO_ICMPV6, 0,
1572                      IPV6_TLV_ROUTERALERT, 2, 0, 0,
1573                      IPV6_TLV_PADN, 0 };
1574
1575         /* we assume size > sizeof(ra) here */
1576         /* limit our allocations to order-0 page */
1577         size = min_t(int, size, SKB_MAX_ORDER(0, 0));
1578         skb = sock_alloc_send_skb(sk, size, 1, &err);
1579
1580         if (!skb)
1581                 return NULL;
1582
1583         skb->priority = TC_PRIO_CONTROL;
1584         skb_reserve(skb, hlen);
1585         skb_tailroom_reserve(skb, mtu, tlen);
1586
1587         if (__ipv6_get_lladdr(idev, &addr_buf, IFA_F_TENTATIVE)) {
1588                 /* <draft-ietf-magma-mld-source-05.txt>:
1589                  * use unspecified address as the source address
1590                  * when a valid link-local address is not available.
1591                  */
1592                 saddr = &in6addr_any;
1593         } else
1594                 saddr = &addr_buf;
1595
1596         ip6_mc_hdr(sk, skb, dev, saddr, &mld2_all_mcr, NEXTHDR_HOP, 0);
1597
1598         memcpy(skb_put(skb, sizeof(ra)), ra, sizeof(ra));
1599
1600         skb_set_transport_header(skb, skb_tail_pointer(skb) - skb->data);
1601         skb_put(skb, sizeof(*pmr));
1602         pmr = (struct mld2_report *)skb_transport_header(skb);
1603         pmr->mld2r_type = ICMPV6_MLD2_REPORT;
1604         pmr->mld2r_resv1 = 0;
1605         pmr->mld2r_cksum = 0;
1606         pmr->mld2r_resv2 = 0;
1607         pmr->mld2r_ngrec = 0;
1608         return skb;
1609 }
1610
1611 static void mld_sendpack(struct sk_buff *skb)
1612 {
1613         struct ipv6hdr *pip6 = ipv6_hdr(skb);
1614         struct mld2_report *pmr =
1615                               (struct mld2_report *)skb_transport_header(skb);
1616         int payload_len, mldlen;
1617         struct inet6_dev *idev;
1618         struct net *net = dev_net(skb->dev);
1619         int err;
1620         struct flowi6 fl6;
1621         struct dst_entry *dst;
1622
1623         rcu_read_lock();
1624         idev = __in6_dev_get(skb->dev);
1625         IP6_UPD_PO_STATS(net, idev, IPSTATS_MIB_OUT, skb->len);
1626
1627         payload_len = (skb_tail_pointer(skb) - skb_network_header(skb)) -
1628                 sizeof(*pip6);
1629         mldlen = skb_tail_pointer(skb) - skb_transport_header(skb);
1630         pip6->payload_len = htons(payload_len);
1631
1632         pmr->mld2r_cksum = csum_ipv6_magic(&pip6->saddr, &pip6->daddr, mldlen,
1633                                            IPPROTO_ICMPV6,
1634                                            csum_partial(skb_transport_header(skb),
1635                                                         mldlen, 0));
1636
1637         icmpv6_flow_init(net->ipv6.igmp_sk, &fl6, ICMPV6_MLD2_REPORT,
1638                          &ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr,
1639                          skb->dev->ifindex);
1640         dst = icmp6_dst_alloc(skb->dev, &fl6);
1641
1642         err = 0;
1643         if (IS_ERR(dst)) {
1644                 err = PTR_ERR(dst);
1645                 dst = NULL;
1646         }
1647         skb_dst_set(skb, dst);
1648         if (err)
1649                 goto err_out;
1650
1651         payload_len = skb->len;
1652
1653         err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT,
1654                       net, net->ipv6.igmp_sk, skb, NULL, skb->dev,
1655                       dst_output);
1656 out:
1657         if (!err) {
1658                 ICMP6MSGOUT_INC_STATS(net, idev, ICMPV6_MLD2_REPORT);
1659                 ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTMSGS);
1660         } else {
1661                 IP6_INC_STATS(net, idev, IPSTATS_MIB_OUTDISCARDS);
1662         }
1663
1664         rcu_read_unlock();
1665         return;
1666
1667 err_out:
1668         kfree_skb(skb);
1669         goto out;
1670 }
1671
1672 static int grec_size(struct ifmcaddr6 *pmc, int type, int gdel, int sdel)
1673 {
1674         return sizeof(struct mld2_grec) + 16 * mld_scount(pmc,type,gdel,sdel);
1675 }
1676
1677 static struct sk_buff *add_grhead(struct sk_buff *skb, struct ifmcaddr6 *pmc,
1678         int type, struct mld2_grec **ppgr)
1679 {
1680         struct net_device *dev = pmc->idev->dev;
1681         struct mld2_report *pmr;
1682         struct mld2_grec *pgr;
1683
1684         if (!skb)
1685                 skb = mld_newpack(pmc->idev, dev->mtu);
1686         if (!skb)
1687                 return NULL;
1688         pgr = (struct mld2_grec *)skb_put(skb, sizeof(struct mld2_grec));
1689         pgr->grec_type = type;
1690         pgr->grec_auxwords = 0;
1691         pgr->grec_nsrcs = 0;
1692         pgr->grec_mca = pmc->mca_addr;  /* structure copy */
1693         pmr = (struct mld2_report *)skb_transport_header(skb);
1694         pmr->mld2r_ngrec = htons(ntohs(pmr->mld2r_ngrec)+1);
1695         *ppgr = pgr;
1696         return skb;
1697 }
1698
1699 #define AVAILABLE(skb)  ((skb) ? skb_availroom(skb) : 0)
1700
1701 static struct sk_buff *add_grec(struct sk_buff *skb, struct ifmcaddr6 *pmc,
1702         int type, int gdeleted, int sdeleted, int crsend)
1703 {
1704         struct inet6_dev *idev = pmc->idev;
1705         struct net_device *dev = idev->dev;
1706         struct mld2_report *pmr;
1707         struct mld2_grec *pgr = NULL;
1708         struct ip6_sf_list *psf, *psf_next, *psf_prev, **psf_list;
1709         int scount, stotal, first, isquery, truncate;
1710
1711         if (pmc->mca_flags & MAF_NOREPORT)
1712                 return skb;
1713
1714         isquery = type == MLD2_MODE_IS_INCLUDE ||
1715                   type == MLD2_MODE_IS_EXCLUDE;
1716         truncate = type == MLD2_MODE_IS_EXCLUDE ||
1717                     type == MLD2_CHANGE_TO_EXCLUDE;
1718
1719         stotal = scount = 0;
1720
1721         psf_list = sdeleted ? &pmc->mca_tomb : &pmc->mca_sources;
1722
1723         if (!*psf_list)
1724                 goto empty_source;
1725
1726         pmr = skb ? (struct mld2_report *)skb_transport_header(skb) : NULL;
1727
1728         /* EX and TO_EX get a fresh packet, if needed */
1729         if (truncate) {
1730                 if (pmr && pmr->mld2r_ngrec &&
1731                     AVAILABLE(skb) < grec_size(pmc, type, gdeleted, sdeleted)) {
1732                         if (skb)
1733                                 mld_sendpack(skb);
1734                         skb = mld_newpack(idev, dev->mtu);
1735                 }
1736         }
1737         first = 1;
1738         psf_prev = NULL;
1739         for (psf = *psf_list; psf; psf = psf_next) {
1740                 struct in6_addr *psrc;
1741
1742                 psf_next = psf->sf_next;
1743
1744                 if (!is_in(pmc, psf, type, gdeleted, sdeleted)) {
1745                         psf_prev = psf;
1746                         continue;
1747                 }
1748
1749                 /* Based on RFC3810 6.1. Should not send source-list change
1750                  * records when there is a filter mode change.
1751                  */
1752                 if (((gdeleted && pmc->mca_sfmode == MCAST_EXCLUDE) ||
1753                      (!gdeleted && pmc->mca_crcount)) &&
1754                     (type == MLD2_ALLOW_NEW_SOURCES ||
1755                      type == MLD2_BLOCK_OLD_SOURCES) && psf->sf_crcount)
1756                         goto decrease_sf_crcount;
1757
1758                 /* clear marks on query responses */
1759                 if (isquery)
1760                         psf->sf_gsresp = 0;
1761
1762                 if (AVAILABLE(skb) < sizeof(*psrc) +
1763                     first*sizeof(struct mld2_grec)) {
1764                         if (truncate && !first)
1765                                 break;   /* truncate these */
1766                         if (pgr)
1767                                 pgr->grec_nsrcs = htons(scount);
1768                         if (skb)
1769                                 mld_sendpack(skb);
1770                         skb = mld_newpack(idev, dev->mtu);
1771                         first = 1;
1772                         scount = 0;
1773                 }
1774                 if (first) {
1775                         skb = add_grhead(skb, pmc, type, &pgr);
1776                         first = 0;
1777                 }
1778                 if (!skb)
1779                         return NULL;
1780                 psrc = (struct in6_addr *)skb_put(skb, sizeof(*psrc));
1781                 *psrc = psf->sf_addr;
1782                 scount++; stotal++;
1783                 if ((type == MLD2_ALLOW_NEW_SOURCES ||
1784                      type == MLD2_BLOCK_OLD_SOURCES) && psf->sf_crcount) {
1785 decrease_sf_crcount:
1786                         psf->sf_crcount--;
1787                         if ((sdeleted || gdeleted) && psf->sf_crcount == 0) {
1788                                 if (psf_prev)
1789                                         psf_prev->sf_next = psf->sf_next;
1790                                 else
1791                                         *psf_list = psf->sf_next;
1792                                 kfree(psf);
1793                                 continue;
1794                         }
1795                 }
1796                 psf_prev = psf;
1797         }
1798
1799 empty_source:
1800         if (!stotal) {
1801                 if (type == MLD2_ALLOW_NEW_SOURCES ||
1802                     type == MLD2_BLOCK_OLD_SOURCES)
1803                         return skb;
1804                 if (pmc->mca_crcount || isquery || crsend) {
1805                         /* make sure we have room for group header */
1806                         if (skb && AVAILABLE(skb) < sizeof(struct mld2_grec)) {
1807                                 mld_sendpack(skb);
1808                                 skb = NULL; /* add_grhead will get a new one */
1809                         }
1810                         skb = add_grhead(skb, pmc, type, &pgr);
1811                 }
1812         }
1813         if (pgr)
1814                 pgr->grec_nsrcs = htons(scount);
1815
1816         if (isquery)
1817                 pmc->mca_flags &= ~MAF_GSQUERY; /* clear query state */
1818         return skb;
1819 }
1820
1821 static void mld_send_report(struct inet6_dev *idev, struct ifmcaddr6 *pmc)
1822 {
1823         struct sk_buff *skb = NULL;
1824         int type;
1825
1826         read_lock_bh(&idev->lock);
1827         if (!pmc) {
1828                 for (pmc = idev->mc_list; pmc; pmc = pmc->next) {
1829                         if (pmc->mca_flags & MAF_NOREPORT)
1830                                 continue;
1831                         spin_lock_bh(&pmc->mca_lock);
1832                         if (pmc->mca_sfcount[MCAST_EXCLUDE])
1833                                 type = MLD2_MODE_IS_EXCLUDE;
1834                         else
1835                                 type = MLD2_MODE_IS_INCLUDE;
1836                         skb = add_grec(skb, pmc, type, 0, 0, 0);
1837                         spin_unlock_bh(&pmc->mca_lock);
1838                 }
1839         } else {
1840                 spin_lock_bh(&pmc->mca_lock);
1841                 if (pmc->mca_sfcount[MCAST_EXCLUDE])
1842                         type = MLD2_MODE_IS_EXCLUDE;
1843                 else
1844                         type = MLD2_MODE_IS_INCLUDE;
1845                 skb = add_grec(skb, pmc, type, 0, 0, 0);
1846                 spin_unlock_bh(&pmc->mca_lock);
1847         }
1848         read_unlock_bh(&idev->lock);
1849         if (skb)
1850                 mld_sendpack(skb);
1851 }
1852
1853 /*
1854  * remove zero-count source records from a source filter list
1855  */
1856 static void mld_clear_zeros(struct ip6_sf_list **ppsf)
1857 {
1858         struct ip6_sf_list *psf_prev, *psf_next, *psf;
1859
1860         psf_prev = NULL;
1861         for (psf = *ppsf; psf; psf = psf_next) {
1862                 psf_next = psf->sf_next;
1863                 if (psf->sf_crcount == 0) {
1864                         if (psf_prev)
1865                                 psf_prev->sf_next = psf->sf_next;
1866                         else
1867                                 *ppsf = psf->sf_next;
1868                         kfree(psf);
1869                 } else
1870                         psf_prev = psf;
1871         }
1872 }
1873
1874 static void mld_send_cr(struct inet6_dev *idev)
1875 {
1876         struct ifmcaddr6 *pmc, *pmc_prev, *pmc_next;
1877         struct sk_buff *skb = NULL;
1878         int type, dtype;
1879
1880         read_lock_bh(&idev->lock);
1881         spin_lock(&idev->mc_lock);
1882
1883         /* deleted MCA's */
1884         pmc_prev = NULL;
1885         for (pmc = idev->mc_tomb; pmc; pmc = pmc_next) {
1886                 pmc_next = pmc->next;
1887                 if (pmc->mca_sfmode == MCAST_INCLUDE) {
1888                         type = MLD2_BLOCK_OLD_SOURCES;
1889                         dtype = MLD2_BLOCK_OLD_SOURCES;
1890                         skb = add_grec(skb, pmc, type, 1, 0, 0);
1891                         skb = add_grec(skb, pmc, dtype, 1, 1, 0);
1892                 }
1893                 if (pmc->mca_crcount) {
1894                         if (pmc->mca_sfmode == MCAST_EXCLUDE) {
1895                                 type = MLD2_CHANGE_TO_INCLUDE;
1896                                 skb = add_grec(skb, pmc, type, 1, 0, 0);
1897                         }
1898                         pmc->mca_crcount--;
1899                         if (pmc->mca_crcount == 0) {
1900                                 mld_clear_zeros(&pmc->mca_tomb);
1901                                 mld_clear_zeros(&pmc->mca_sources);
1902                         }
1903                 }
1904                 if (pmc->mca_crcount == 0 && !pmc->mca_tomb &&
1905                     !pmc->mca_sources) {
1906                         if (pmc_prev)
1907                                 pmc_prev->next = pmc_next;
1908                         else
1909                                 idev->mc_tomb = pmc_next;
1910                         in6_dev_put(pmc->idev);
1911                         kfree(pmc);
1912                 } else
1913                         pmc_prev = pmc;
1914         }
1915         spin_unlock(&idev->mc_lock);
1916
1917         /* change recs */
1918         for (pmc = idev->mc_list; pmc; pmc = pmc->next) {
1919                 spin_lock_bh(&pmc->mca_lock);
1920                 if (pmc->mca_sfcount[MCAST_EXCLUDE]) {
1921                         type = MLD2_BLOCK_OLD_SOURCES;
1922                         dtype = MLD2_ALLOW_NEW_SOURCES;
1923                 } else {
1924                         type = MLD2_ALLOW_NEW_SOURCES;
1925                         dtype = MLD2_BLOCK_OLD_SOURCES;
1926                 }
1927                 skb = add_grec(skb, pmc, type, 0, 0, 0);
1928                 skb = add_grec(skb, pmc, dtype, 0, 1, 0);       /* deleted sources */
1929
1930                 /* filter mode changes */
1931                 if (pmc->mca_crcount) {
1932                         if (pmc->mca_sfmode == MCAST_EXCLUDE)
1933                                 type = MLD2_CHANGE_TO_EXCLUDE;
1934                         else
1935                                 type = MLD2_CHANGE_TO_INCLUDE;
1936                         skb = add_grec(skb, pmc, type, 0, 0, 0);
1937                         pmc->mca_crcount--;
1938                 }
1939                 spin_unlock_bh(&pmc->mca_lock);
1940         }
1941         read_unlock_bh(&idev->lock);
1942         if (!skb)
1943                 return;
1944         (void) mld_sendpack(skb);
1945 }
1946
1947 static void igmp6_send(struct in6_addr *addr, struct net_device *dev, int type)
1948 {
1949         struct net *net = dev_net(dev);
1950         struct sock *sk = net->ipv6.igmp_sk;
1951         struct inet6_dev *idev;
1952         struct sk_buff *skb;
1953         struct mld_msg *hdr;
1954         const struct in6_addr *snd_addr, *saddr;
1955         struct in6_addr addr_buf;
1956         int hlen = LL_RESERVED_SPACE(dev);
1957         int tlen = dev->needed_tailroom;
1958         int err, len, payload_len, full_len;
1959         u8 ra[8] = { IPPROTO_ICMPV6, 0,
1960                      IPV6_TLV_ROUTERALERT, 2, 0, 0,
1961                      IPV6_TLV_PADN, 0 };
1962         struct flowi6 fl6;
1963         struct dst_entry *dst;
1964
1965         if (type == ICMPV6_MGM_REDUCTION)
1966                 snd_addr = &in6addr_linklocal_allrouters;
1967         else
1968                 snd_addr = addr;
1969
1970         len = sizeof(struct icmp6hdr) + sizeof(struct in6_addr);
1971         payload_len = len + sizeof(ra);
1972         full_len = sizeof(struct ipv6hdr) + payload_len;
1973
1974         rcu_read_lock();
1975         IP6_UPD_PO_STATS(net, __in6_dev_get(dev),
1976                       IPSTATS_MIB_OUT, full_len);
1977         rcu_read_unlock();
1978
1979         skb = sock_alloc_send_skb(sk, hlen + tlen + full_len, 1, &err);
1980
1981         if (!skb) {
1982                 rcu_read_lock();
1983                 IP6_INC_STATS(net, __in6_dev_get(dev),
1984                               IPSTATS_MIB_OUTDISCARDS);
1985                 rcu_read_unlock();
1986                 return;
1987         }
1988         skb->priority = TC_PRIO_CONTROL;
1989         skb_reserve(skb, hlen);
1990
1991         if (ipv6_get_lladdr(dev, &addr_buf, IFA_F_TENTATIVE)) {
1992                 /* <draft-ietf-magma-mld-source-05.txt>:
1993                  * use unspecified address as the source address
1994                  * when a valid link-local address is not available.
1995                  */
1996                 saddr = &in6addr_any;
1997         } else
1998                 saddr = &addr_buf;
1999
2000         ip6_mc_hdr(sk, skb, dev, saddr, snd_addr, NEXTHDR_HOP, payload_len);
2001
2002         memcpy(skb_put(skb, sizeof(ra)), ra, sizeof(ra));
2003
2004         hdr = (struct mld_msg *) skb_put(skb, sizeof(struct mld_msg));
2005         memset(hdr, 0, sizeof(struct mld_msg));
2006         hdr->mld_type = type;
2007         hdr->mld_mca = *addr;
2008
2009         hdr->mld_cksum = csum_ipv6_magic(saddr, snd_addr, len,
2010                                          IPPROTO_ICMPV6,
2011                                          csum_partial(hdr, len, 0));
2012
2013         rcu_read_lock();
2014         idev = __in6_dev_get(skb->dev);
2015
2016         icmpv6_flow_init(sk, &fl6, type,
2017                          &ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr,
2018                          skb->dev->ifindex);
2019         dst = icmp6_dst_alloc(skb->dev, &fl6);
2020         if (IS_ERR(dst)) {
2021                 err = PTR_ERR(dst);
2022                 goto err_out;
2023         }
2024
2025         skb_dst_set(skb, dst);
2026         err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT,
2027                       net, sk, skb, NULL, skb->dev,
2028                       dst_output);
2029 out:
2030         if (!err) {
2031                 ICMP6MSGOUT_INC_STATS(net, idev, type);
2032                 ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTMSGS);
2033         } else
2034                 IP6_INC_STATS(net, idev, IPSTATS_MIB_OUTDISCARDS);
2035
2036         rcu_read_unlock();
2037         return;
2038
2039 err_out:
2040         kfree_skb(skb);
2041         goto out;
2042 }
2043
2044 static void mld_send_initial_cr(struct inet6_dev *idev)
2045 {
2046         struct sk_buff *skb;
2047         struct ifmcaddr6 *pmc;
2048         int type;
2049
2050         if (mld_in_v1_mode(idev))
2051                 return;
2052
2053         skb = NULL;
2054         read_lock_bh(&idev->lock);
2055         for (pmc = idev->mc_list; pmc; pmc = pmc->next) {
2056                 spin_lock_bh(&pmc->mca_lock);
2057                 if (pmc->mca_sfcount[MCAST_EXCLUDE])
2058                         type = MLD2_CHANGE_TO_EXCLUDE;
2059                 else
2060                         type = MLD2_CHANGE_TO_INCLUDE;
2061                 skb = add_grec(skb, pmc, type, 0, 0, 1);
2062                 spin_unlock_bh(&pmc->mca_lock);
2063         }
2064         read_unlock_bh(&idev->lock);
2065         if (skb)
2066                 mld_sendpack(skb);
2067 }
2068
2069 void ipv6_mc_dad_complete(struct inet6_dev *idev)
2070 {
2071         idev->mc_dad_count = idev->mc_qrv;
2072         if (idev->mc_dad_count) {
2073                 mld_send_initial_cr(idev);
2074                 idev->mc_dad_count--;
2075                 if (idev->mc_dad_count)
2076                         mld_dad_start_timer(idev, idev->mc_maxdelay);
2077         }
2078 }
2079
2080 static void mld_dad_timer_expire(unsigned long data)
2081 {
2082         struct inet6_dev *idev = (struct inet6_dev *)data;
2083
2084         mld_send_initial_cr(idev);
2085         if (idev->mc_dad_count) {
2086                 idev->mc_dad_count--;
2087                 if (idev->mc_dad_count)
2088                         mld_dad_start_timer(idev, idev->mc_maxdelay);
2089         }
2090         in6_dev_put(idev);
2091 }
2092
2093 static int ip6_mc_del1_src(struct ifmcaddr6 *pmc, int sfmode,
2094         const struct in6_addr *psfsrc)
2095 {
2096         struct ip6_sf_list *psf, *psf_prev;
2097         int rv = 0;
2098
2099         psf_prev = NULL;
2100         for (psf = pmc->mca_sources; psf; psf = psf->sf_next) {
2101                 if (ipv6_addr_equal(&psf->sf_addr, psfsrc))
2102                         break;
2103                 psf_prev = psf;
2104         }
2105         if (!psf || psf->sf_count[sfmode] == 0) {
2106                 /* source filter not found, or count wrong =>  bug */
2107                 return -ESRCH;
2108         }
2109         psf->sf_count[sfmode]--;
2110         if (!psf->sf_count[MCAST_INCLUDE] && !psf->sf_count[MCAST_EXCLUDE]) {
2111                 struct inet6_dev *idev = pmc->idev;
2112
2113                 /* no more filters for this source */
2114                 if (psf_prev)
2115                         psf_prev->sf_next = psf->sf_next;
2116                 else
2117                         pmc->mca_sources = psf->sf_next;
2118                 if (psf->sf_oldin && !(pmc->mca_flags & MAF_NOREPORT) &&
2119                     !mld_in_v1_mode(idev)) {
2120                         psf->sf_crcount = idev->mc_qrv;
2121                         psf->sf_next = pmc->mca_tomb;
2122                         pmc->mca_tomb = psf;
2123                         rv = 1;
2124                 } else
2125                         kfree(psf);
2126         }
2127         return rv;
2128 }
2129
2130 static int ip6_mc_del_src(struct inet6_dev *idev, const struct in6_addr *pmca,
2131                           int sfmode, int sfcount, const struct in6_addr *psfsrc,
2132                           int delta)
2133 {
2134         struct ifmcaddr6 *pmc;
2135         int     changerec = 0;
2136         int     i, err;
2137
2138         if (!idev)
2139                 return -ENODEV;
2140         read_lock_bh(&idev->lock);
2141         for (pmc = idev->mc_list; pmc; pmc = pmc->next) {
2142                 if (ipv6_addr_equal(pmca, &pmc->mca_addr))
2143                         break;
2144         }
2145         if (!pmc) {
2146                 /* MCA not found?? bug */
2147                 read_unlock_bh(&idev->lock);
2148                 return -ESRCH;
2149         }
2150         spin_lock_bh(&pmc->mca_lock);
2151         sf_markstate(pmc);
2152         if (!delta) {
2153                 if (!pmc->mca_sfcount[sfmode]) {
2154                         spin_unlock_bh(&pmc->mca_lock);
2155                         read_unlock_bh(&idev->lock);
2156                         return -EINVAL;
2157                 }
2158                 pmc->mca_sfcount[sfmode]--;
2159         }
2160         err = 0;
2161         for (i = 0; i < sfcount; i++) {
2162                 int rv = ip6_mc_del1_src(pmc, sfmode, &psfsrc[i]);
2163
2164                 changerec |= rv > 0;
2165                 if (!err && rv < 0)
2166                         err = rv;
2167         }
2168         if (pmc->mca_sfmode == MCAST_EXCLUDE &&
2169             pmc->mca_sfcount[MCAST_EXCLUDE] == 0 &&
2170             pmc->mca_sfcount[MCAST_INCLUDE]) {
2171                 struct ip6_sf_list *psf;
2172
2173                 /* filter mode change */
2174                 pmc->mca_sfmode = MCAST_INCLUDE;
2175                 pmc->mca_crcount = idev->mc_qrv;
2176                 idev->mc_ifc_count = pmc->mca_crcount;
2177                 for (psf = pmc->mca_sources; psf; psf = psf->sf_next)
2178                         psf->sf_crcount = 0;
2179                 mld_ifc_event(pmc->idev);
2180         } else if (sf_setstate(pmc) || changerec)
2181                 mld_ifc_event(pmc->idev);
2182         spin_unlock_bh(&pmc->mca_lock);
2183         read_unlock_bh(&idev->lock);
2184         return err;
2185 }
2186
2187 /*
2188  * Add multicast single-source filter to the interface list
2189  */
2190 static int ip6_mc_add1_src(struct ifmcaddr6 *pmc, int sfmode,
2191         const struct in6_addr *psfsrc)
2192 {
2193         struct ip6_sf_list *psf, *psf_prev;
2194
2195         psf_prev = NULL;
2196         for (psf = pmc->mca_sources; psf; psf = psf->sf_next) {
2197                 if (ipv6_addr_equal(&psf->sf_addr, psfsrc))
2198                         break;
2199                 psf_prev = psf;
2200         }
2201         if (!psf) {
2202                 psf = kzalloc(sizeof(*psf), GFP_ATOMIC);
2203                 if (!psf)
2204                         return -ENOBUFS;
2205
2206                 psf->sf_addr = *psfsrc;
2207                 if (psf_prev) {
2208                         psf_prev->sf_next = psf;
2209                 } else
2210                         pmc->mca_sources = psf;
2211         }
2212         psf->sf_count[sfmode]++;
2213         return 0;
2214 }
2215
2216 static void sf_markstate(struct ifmcaddr6 *pmc)
2217 {
2218         struct ip6_sf_list *psf;
2219         int mca_xcount = pmc->mca_sfcount[MCAST_EXCLUDE];
2220
2221         for (psf = pmc->mca_sources; psf; psf = psf->sf_next)
2222                 if (pmc->mca_sfcount[MCAST_EXCLUDE]) {
2223                         psf->sf_oldin = mca_xcount ==
2224                                 psf->sf_count[MCAST_EXCLUDE] &&
2225                                 !psf->sf_count[MCAST_INCLUDE];
2226                 } else
2227                         psf->sf_oldin = psf->sf_count[MCAST_INCLUDE] != 0;
2228 }
2229
2230 static int sf_setstate(struct ifmcaddr6 *pmc)
2231 {
2232         struct ip6_sf_list *psf, *dpsf;
2233         int mca_xcount = pmc->mca_sfcount[MCAST_EXCLUDE];
2234         int qrv = pmc->idev->mc_qrv;
2235         int new_in, rv;
2236
2237         rv = 0;
2238         for (psf = pmc->mca_sources; psf; psf = psf->sf_next) {
2239                 if (pmc->mca_sfcount[MCAST_EXCLUDE]) {
2240                         new_in = mca_xcount == psf->sf_count[MCAST_EXCLUDE] &&
2241                                 !psf->sf_count[MCAST_INCLUDE];
2242                 } else
2243                         new_in = psf->sf_count[MCAST_INCLUDE] != 0;
2244                 if (new_in) {
2245                         if (!psf->sf_oldin) {
2246                                 struct ip6_sf_list *prev = NULL;
2247
2248                                 for (dpsf = pmc->mca_tomb; dpsf;
2249                                      dpsf = dpsf->sf_next) {
2250                                         if (ipv6_addr_equal(&dpsf->sf_addr,
2251                                             &psf->sf_addr))
2252                                                 break;
2253                                         prev = dpsf;
2254                                 }
2255                                 if (dpsf) {
2256                                         if (prev)
2257                                                 prev->sf_next = dpsf->sf_next;
2258                                         else
2259                                                 pmc->mca_tomb = dpsf->sf_next;
2260                                         kfree(dpsf);
2261                                 }
2262                                 psf->sf_crcount = qrv;
2263                                 rv++;
2264                         }
2265                 } else if (psf->sf_oldin) {
2266                         psf->sf_crcount = 0;
2267                         /*
2268                          * add or update "delete" records if an active filter
2269                          * is now inactive
2270                          */
2271                         for (dpsf = pmc->mca_tomb; dpsf; dpsf = dpsf->sf_next)
2272                                 if (ipv6_addr_equal(&dpsf->sf_addr,
2273                                     &psf->sf_addr))
2274                                         break;
2275                         if (!dpsf) {
2276                                 dpsf = kmalloc(sizeof(*dpsf), GFP_ATOMIC);
2277                                 if (!dpsf)
2278                                         continue;
2279                                 *dpsf = *psf;
2280                                 /* pmc->mca_lock held by callers */
2281                                 dpsf->sf_next = pmc->mca_tomb;
2282                                 pmc->mca_tomb = dpsf;
2283                         }
2284                         dpsf->sf_crcount = qrv;
2285                         rv++;
2286                 }
2287         }
2288         return rv;
2289 }
2290
2291 /*
2292  * Add multicast source filter list to the interface list
2293  */
2294 static int ip6_mc_add_src(struct inet6_dev *idev, const struct in6_addr *pmca,
2295                           int sfmode, int sfcount, const struct in6_addr *psfsrc,
2296                           int delta)
2297 {
2298         struct ifmcaddr6 *pmc;
2299         int     isexclude;
2300         int     i, err;
2301
2302         if (!idev)
2303                 return -ENODEV;
2304         read_lock_bh(&idev->lock);
2305         for (pmc = idev->mc_list; pmc; pmc = pmc->next) {
2306                 if (ipv6_addr_equal(pmca, &pmc->mca_addr))
2307                         break;
2308         }
2309         if (!pmc) {
2310                 /* MCA not found?? bug */
2311                 read_unlock_bh(&idev->lock);
2312                 return -ESRCH;
2313         }
2314         spin_lock_bh(&pmc->mca_lock);
2315
2316         sf_markstate(pmc);
2317         isexclude = pmc->mca_sfmode == MCAST_EXCLUDE;
2318         if (!delta)
2319                 pmc->mca_sfcount[sfmode]++;
2320         err = 0;
2321         for (i = 0; i < sfcount; i++) {
2322                 err = ip6_mc_add1_src(pmc, sfmode, &psfsrc[i]);
2323                 if (err)
2324                         break;
2325         }
2326         if (err) {
2327                 int j;
2328
2329                 if (!delta)
2330                         pmc->mca_sfcount[sfmode]--;
2331                 for (j = 0; j < i; j++)
2332                         ip6_mc_del1_src(pmc, sfmode, &psfsrc[j]);
2333         } else if (isexclude != (pmc->mca_sfcount[MCAST_EXCLUDE] != 0)) {
2334                 struct ip6_sf_list *psf;
2335
2336                 /* filter mode change */
2337                 if (pmc->mca_sfcount[MCAST_EXCLUDE])
2338                         pmc->mca_sfmode = MCAST_EXCLUDE;
2339                 else if (pmc->mca_sfcount[MCAST_INCLUDE])
2340                         pmc->mca_sfmode = MCAST_INCLUDE;
2341                 /* else no filters; keep old mode for reports */
2342
2343                 pmc->mca_crcount = idev->mc_qrv;
2344                 idev->mc_ifc_count = pmc->mca_crcount;
2345                 for (psf = pmc->mca_sources; psf; psf = psf->sf_next)
2346                         psf->sf_crcount = 0;
2347                 mld_ifc_event(idev);
2348         } else if (sf_setstate(pmc))
2349                 mld_ifc_event(idev);
2350         spin_unlock_bh(&pmc->mca_lock);
2351         read_unlock_bh(&idev->lock);
2352         return err;
2353 }
2354
2355 static void ip6_mc_clear_src(struct ifmcaddr6 *pmc)
2356 {
2357         struct ip6_sf_list *psf, *nextpsf;
2358
2359         for (psf = pmc->mca_tomb; psf; psf = nextpsf) {
2360                 nextpsf = psf->sf_next;
2361                 kfree(psf);
2362         }
2363         pmc->mca_tomb = NULL;
2364         for (psf = pmc->mca_sources; psf; psf = nextpsf) {
2365                 nextpsf = psf->sf_next;
2366                 kfree(psf);
2367         }
2368         pmc->mca_sources = NULL;
2369         pmc->mca_sfmode = MCAST_EXCLUDE;
2370         pmc->mca_sfcount[MCAST_INCLUDE] = 0;
2371         pmc->mca_sfcount[MCAST_EXCLUDE] = 1;
2372 }
2373
2374
2375 static void igmp6_join_group(struct ifmcaddr6 *ma)
2376 {
2377         unsigned long delay;
2378
2379         if (ma->mca_flags & MAF_NOREPORT)
2380                 return;
2381
2382         igmp6_send(&ma->mca_addr, ma->idev->dev, ICMPV6_MGM_REPORT);
2383
2384         delay = prandom_u32() % unsolicited_report_interval(ma->idev);
2385
2386         spin_lock_bh(&ma->mca_lock);
2387         if (del_timer(&ma->mca_timer)) {
2388                 atomic_dec(&ma->mca_refcnt);
2389                 delay = ma->mca_timer.expires - jiffies;
2390         }
2391
2392         if (!mod_timer(&ma->mca_timer, jiffies + delay))
2393                 atomic_inc(&ma->mca_refcnt);
2394         ma->mca_flags |= MAF_TIMER_RUNNING | MAF_LAST_REPORTER;
2395         spin_unlock_bh(&ma->mca_lock);
2396 }
2397
2398 static int ip6_mc_leave_src(struct sock *sk, struct ipv6_mc_socklist *iml,
2399                             struct inet6_dev *idev)
2400 {
2401         int err;
2402
2403         /* callers have the socket lock and rtnl lock
2404          * so no other readers or writers of iml or its sflist
2405          */
2406         if (!iml->sflist) {
2407                 /* any-source empty exclude case */
2408                 return ip6_mc_del_src(idev, &iml->addr, iml->sfmode, 0, NULL, 0);
2409         }
2410         err = ip6_mc_del_src(idev, &iml->addr, iml->sfmode,
2411                 iml->sflist->sl_count, iml->sflist->sl_addr, 0);
2412         sock_kfree_s(sk, iml->sflist, IP6_SFLSIZE(iml->sflist->sl_max));
2413         iml->sflist = NULL;
2414         return err;
2415 }
2416
2417 static void igmp6_leave_group(struct ifmcaddr6 *ma)
2418 {
2419         if (mld_in_v1_mode(ma->idev)) {
2420                 if (ma->mca_flags & MAF_LAST_REPORTER)
2421                         igmp6_send(&ma->mca_addr, ma->idev->dev,
2422                                 ICMPV6_MGM_REDUCTION);
2423         } else {
2424                 mld_add_delrec(ma->idev, ma);
2425                 mld_ifc_event(ma->idev);
2426         }
2427 }
2428
2429 static void mld_gq_timer_expire(unsigned long data)
2430 {
2431         struct inet6_dev *idev = (struct inet6_dev *)data;
2432
2433         idev->mc_gq_running = 0;
2434         mld_send_report(idev, NULL);
2435         in6_dev_put(idev);
2436 }
2437
2438 static void mld_ifc_timer_expire(unsigned long data)
2439 {
2440         struct inet6_dev *idev = (struct inet6_dev *)data;
2441
2442         mld_send_cr(idev);
2443         if (idev->mc_ifc_count) {
2444                 idev->mc_ifc_count--;
2445                 if (idev->mc_ifc_count)
2446                         mld_ifc_start_timer(idev, idev->mc_maxdelay);
2447         }
2448         in6_dev_put(idev);
2449 }
2450
2451 static void mld_ifc_event(struct inet6_dev *idev)
2452 {
2453         if (mld_in_v1_mode(idev))
2454                 return;
2455         idev->mc_ifc_count = idev->mc_qrv;
2456         mld_ifc_start_timer(idev, 1);
2457 }
2458
2459
2460 static void igmp6_timer_handler(unsigned long data)
2461 {
2462         struct ifmcaddr6 *ma = (struct ifmcaddr6 *) data;
2463
2464         if (mld_in_v1_mode(ma->idev))
2465                 igmp6_send(&ma->mca_addr, ma->idev->dev, ICMPV6_MGM_REPORT);
2466         else
2467                 mld_send_report(ma->idev, ma);
2468
2469         spin_lock(&ma->mca_lock);
2470         ma->mca_flags |=  MAF_LAST_REPORTER;
2471         ma->mca_flags &= ~MAF_TIMER_RUNNING;
2472         spin_unlock(&ma->mca_lock);
2473         ma_put(ma);
2474 }
2475
2476 /* Device changing type */
2477
2478 void ipv6_mc_unmap(struct inet6_dev *idev)
2479 {
2480         struct ifmcaddr6 *i;
2481
2482         /* Install multicast list, except for all-nodes (already installed) */
2483
2484         read_lock_bh(&idev->lock);
2485         for (i = idev->mc_list; i; i = i->next)
2486                 igmp6_group_dropped(i);
2487         read_unlock_bh(&idev->lock);
2488 }
2489
2490 void ipv6_mc_remap(struct inet6_dev *idev)
2491 {
2492         ipv6_mc_up(idev);
2493 }
2494
2495 /* Device going down */
2496
2497 void ipv6_mc_down(struct inet6_dev *idev)
2498 {
2499         struct ifmcaddr6 *i;
2500
2501         /* Withdraw multicast list */
2502
2503         read_lock_bh(&idev->lock);
2504         mld_ifc_stop_timer(idev);
2505         mld_gq_stop_timer(idev);
2506         mld_dad_stop_timer(idev);
2507
2508         for (i = idev->mc_list; i; i = i->next)
2509                 igmp6_group_dropped(i);
2510         read_unlock_bh(&idev->lock);
2511
2512         mld_clear_delrec(idev);
2513 }
2514
2515 static void ipv6_mc_reset(struct inet6_dev *idev)
2516 {
2517         idev->mc_qrv = sysctl_mld_qrv;
2518         idev->mc_qi = MLD_QI_DEFAULT;
2519         idev->mc_qri = MLD_QRI_DEFAULT;
2520         idev->mc_v1_seen = 0;
2521         idev->mc_maxdelay = unsolicited_report_interval(idev);
2522 }
2523
2524 /* Device going up */
2525
2526 void ipv6_mc_up(struct inet6_dev *idev)
2527 {
2528         struct ifmcaddr6 *i;
2529
2530         /* Install multicast list, except for all-nodes (already installed) */
2531
2532         read_lock_bh(&idev->lock);
2533         ipv6_mc_reset(idev);
2534         for (i = idev->mc_list; i; i = i->next)
2535                 igmp6_group_added(i);
2536         read_unlock_bh(&idev->lock);
2537 }
2538
2539 /* IPv6 device initialization. */
2540
2541 void ipv6_mc_init_dev(struct inet6_dev *idev)
2542 {
2543         write_lock_bh(&idev->lock);
2544         spin_lock_init(&idev->mc_lock);
2545         idev->mc_gq_running = 0;
2546         setup_timer(&idev->mc_gq_timer, mld_gq_timer_expire,
2547                         (unsigned long)idev);
2548         idev->mc_tomb = NULL;
2549         idev->mc_ifc_count = 0;
2550         setup_timer(&idev->mc_ifc_timer, mld_ifc_timer_expire,
2551                         (unsigned long)idev);
2552         setup_timer(&idev->mc_dad_timer, mld_dad_timer_expire,
2553                     (unsigned long)idev);
2554         ipv6_mc_reset(idev);
2555         write_unlock_bh(&idev->lock);
2556 }
2557
2558 /*
2559  *      Device is about to be destroyed: clean up.
2560  */
2561
2562 void ipv6_mc_destroy_dev(struct inet6_dev *idev)
2563 {
2564         struct ifmcaddr6 *i;
2565
2566         /* Deactivate timers */
2567         ipv6_mc_down(idev);
2568
2569         /* Delete all-nodes address. */
2570         /* We cannot call ipv6_dev_mc_dec() directly, our caller in
2571          * addrconf.c has NULL'd out dev->ip6_ptr so in6_dev_get() will
2572          * fail.
2573          */
2574         __ipv6_dev_mc_dec(idev, &in6addr_linklocal_allnodes);
2575
2576         if (idev->cnf.forwarding)
2577                 __ipv6_dev_mc_dec(idev, &in6addr_linklocal_allrouters);
2578
2579         write_lock_bh(&idev->lock);
2580         while ((i = idev->mc_list) != NULL) {
2581                 idev->mc_list = i->next;
2582                 write_unlock_bh(&idev->lock);
2583
2584                 igmp6_group_dropped(i);
2585                 ma_put(i);
2586
2587                 write_lock_bh(&idev->lock);
2588         }
2589         write_unlock_bh(&idev->lock);
2590 }
2591
2592 #ifdef CONFIG_PROC_FS
2593 struct igmp6_mc_iter_state {
2594         struct seq_net_private p;
2595         struct net_device *dev;
2596         struct inet6_dev *idev;
2597 };
2598
2599 #define igmp6_mc_seq_private(seq)       ((struct igmp6_mc_iter_state *)(seq)->private)
2600
2601 static inline struct ifmcaddr6 *igmp6_mc_get_first(struct seq_file *seq)
2602 {
2603         struct ifmcaddr6 *im = NULL;
2604         struct igmp6_mc_iter_state *state = igmp6_mc_seq_private(seq);
2605         struct net *net = seq_file_net(seq);
2606
2607         state->idev = NULL;
2608         for_each_netdev_rcu(net, state->dev) {
2609                 struct inet6_dev *idev;
2610                 idev = __in6_dev_get(state->dev);
2611                 if (!idev)
2612                         continue;
2613                 read_lock_bh(&idev->lock);
2614                 im = idev->mc_list;
2615                 if (im) {
2616                         state->idev = idev;
2617                         break;
2618                 }
2619                 read_unlock_bh(&idev->lock);
2620         }
2621         return im;
2622 }
2623
2624 static struct ifmcaddr6 *igmp6_mc_get_next(struct seq_file *seq, struct ifmcaddr6 *im)
2625 {
2626         struct igmp6_mc_iter_state *state = igmp6_mc_seq_private(seq);
2627
2628         im = im->next;
2629         while (!im) {
2630                 if (likely(state->idev))
2631                         read_unlock_bh(&state->idev->lock);
2632
2633                 state->dev = next_net_device_rcu(state->dev);
2634                 if (!state->dev) {
2635                         state->idev = NULL;
2636                         break;
2637                 }
2638                 state->idev = __in6_dev_get(state->dev);
2639                 if (!state->idev)
2640                         continue;
2641                 read_lock_bh(&state->idev->lock);
2642                 im = state->idev->mc_list;
2643         }
2644         return im;
2645 }
2646
2647 static struct ifmcaddr6 *igmp6_mc_get_idx(struct seq_file *seq, loff_t pos)
2648 {
2649         struct ifmcaddr6 *im = igmp6_mc_get_first(seq);
2650         if (im)
2651                 while (pos && (im = igmp6_mc_get_next(seq, im)) != NULL)
2652                         --pos;
2653         return pos ? NULL : im;
2654 }
2655
2656 static void *igmp6_mc_seq_start(struct seq_file *seq, loff_t *pos)
2657         __acquires(RCU)
2658 {
2659         rcu_read_lock();
2660         return igmp6_mc_get_idx(seq, *pos);
2661 }
2662
2663 static void *igmp6_mc_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2664 {
2665         struct ifmcaddr6 *im = igmp6_mc_get_next(seq, v);
2666
2667         ++*pos;
2668         return im;
2669 }
2670
2671 static void igmp6_mc_seq_stop(struct seq_file *seq, void *v)
2672         __releases(RCU)
2673 {
2674         struct igmp6_mc_iter_state *state = igmp6_mc_seq_private(seq);
2675
2676         if (likely(state->idev)) {
2677                 read_unlock_bh(&state->idev->lock);
2678                 state->idev = NULL;
2679         }
2680         state->dev = NULL;
2681         rcu_read_unlock();
2682 }
2683
2684 static int igmp6_mc_seq_show(struct seq_file *seq, void *v)
2685 {
2686         struct ifmcaddr6 *im = (struct ifmcaddr6 *)v;
2687         struct igmp6_mc_iter_state *state = igmp6_mc_seq_private(seq);
2688
2689         seq_printf(seq,
2690                    "%-4d %-15s %pi6 %5d %08X %ld\n",
2691                    state->dev->ifindex, state->dev->name,
2692                    &im->mca_addr,
2693                    im->mca_users, im->mca_flags,
2694                    (im->mca_flags&MAF_TIMER_RUNNING) ?
2695                    jiffies_to_clock_t(im->mca_timer.expires-jiffies) : 0);
2696         return 0;
2697 }
2698
2699 static const struct seq_operations igmp6_mc_seq_ops = {
2700         .start  =       igmp6_mc_seq_start,
2701         .next   =       igmp6_mc_seq_next,
2702         .stop   =       igmp6_mc_seq_stop,
2703         .show   =       igmp6_mc_seq_show,
2704 };
2705
2706 static int igmp6_mc_seq_open(struct inode *inode, struct file *file)
2707 {
2708         return seq_open_net(inode, file, &igmp6_mc_seq_ops,
2709                             sizeof(struct igmp6_mc_iter_state));
2710 }
2711
2712 static const struct file_operations igmp6_mc_seq_fops = {
2713         .owner          =       THIS_MODULE,
2714         .open           =       igmp6_mc_seq_open,
2715         .read           =       seq_read,
2716         .llseek         =       seq_lseek,
2717         .release        =       seq_release_net,
2718 };
2719
2720 struct igmp6_mcf_iter_state {
2721         struct seq_net_private p;
2722         struct net_device *dev;
2723         struct inet6_dev *idev;
2724         struct ifmcaddr6 *im;
2725 };
2726
2727 #define igmp6_mcf_seq_private(seq)      ((struct igmp6_mcf_iter_state *)(seq)->private)
2728
2729 static inline struct ip6_sf_list *igmp6_mcf_get_first(struct seq_file *seq)
2730 {
2731         struct ip6_sf_list *psf = NULL;
2732         struct ifmcaddr6 *im = NULL;
2733         struct igmp6_mcf_iter_state *state = igmp6_mcf_seq_private(seq);
2734         struct net *net = seq_file_net(seq);
2735
2736         state->idev = NULL;
2737         state->im = NULL;
2738         for_each_netdev_rcu(net, state->dev) {
2739                 struct inet6_dev *idev;
2740                 idev = __in6_dev_get(state->dev);
2741                 if (unlikely(idev == NULL))
2742                         continue;
2743                 read_lock_bh(&idev->lock);
2744                 im = idev->mc_list;
2745                 if (likely(im)) {
2746                         spin_lock_bh(&im->mca_lock);
2747                         psf = im->mca_sources;
2748                         if (likely(psf)) {
2749                                 state->im = im;
2750                                 state->idev = idev;
2751                                 break;
2752                         }
2753                         spin_unlock_bh(&im->mca_lock);
2754                 }
2755                 read_unlock_bh(&idev->lock);
2756         }
2757         return psf;
2758 }
2759
2760 static struct ip6_sf_list *igmp6_mcf_get_next(struct seq_file *seq, struct ip6_sf_list *psf)
2761 {
2762         struct igmp6_mcf_iter_state *state = igmp6_mcf_seq_private(seq);
2763
2764         psf = psf->sf_next;
2765         while (!psf) {
2766                 spin_unlock_bh(&state->im->mca_lock);
2767                 state->im = state->im->next;
2768                 while (!state->im) {
2769                         if (likely(state->idev))
2770                                 read_unlock_bh(&state->idev->lock);
2771
2772                         state->dev = next_net_device_rcu(state->dev);
2773                         if (!state->dev) {
2774                                 state->idev = NULL;
2775                                 goto out;
2776                         }
2777                         state->idev = __in6_dev_get(state->dev);
2778                         if (!state->idev)
2779                                 continue;
2780                         read_lock_bh(&state->idev->lock);
2781                         state->im = state->idev->mc_list;
2782                 }
2783                 if (!state->im)
2784                         break;
2785                 spin_lock_bh(&state->im->mca_lock);
2786                 psf = state->im->mca_sources;
2787         }
2788 out:
2789         return psf;
2790 }
2791
2792 static struct ip6_sf_list *igmp6_mcf_get_idx(struct seq_file *seq, loff_t pos)
2793 {
2794         struct ip6_sf_list *psf = igmp6_mcf_get_first(seq);
2795         if (psf)
2796                 while (pos && (psf = igmp6_mcf_get_next(seq, psf)) != NULL)
2797                         --pos;
2798         return pos ? NULL : psf;
2799 }
2800
2801 static void *igmp6_mcf_seq_start(struct seq_file *seq, loff_t *pos)
2802         __acquires(RCU)
2803 {
2804         rcu_read_lock();
2805         return *pos ? igmp6_mcf_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
2806 }
2807
2808 static void *igmp6_mcf_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2809 {
2810         struct ip6_sf_list *psf;
2811         if (v == SEQ_START_TOKEN)
2812                 psf = igmp6_mcf_get_first(seq);
2813         else
2814                 psf = igmp6_mcf_get_next(seq, v);
2815         ++*pos;
2816         return psf;
2817 }
2818
2819 static void igmp6_mcf_seq_stop(struct seq_file *seq, void *v)
2820         __releases(RCU)
2821 {
2822         struct igmp6_mcf_iter_state *state = igmp6_mcf_seq_private(seq);
2823         if (likely(state->im)) {
2824                 spin_unlock_bh(&state->im->mca_lock);
2825                 state->im = NULL;
2826         }
2827         if (likely(state->idev)) {
2828                 read_unlock_bh(&state->idev->lock);
2829                 state->idev = NULL;
2830         }
2831         state->dev = NULL;
2832         rcu_read_unlock();
2833 }
2834
2835 static int igmp6_mcf_seq_show(struct seq_file *seq, void *v)
2836 {
2837         struct ip6_sf_list *psf = (struct ip6_sf_list *)v;
2838         struct igmp6_mcf_iter_state *state = igmp6_mcf_seq_private(seq);
2839
2840         if (v == SEQ_START_TOKEN) {
2841                 seq_puts(seq, "Idx Device                Multicast Address                   Source Address    INC    EXC\n");
2842         } else {
2843                 seq_printf(seq,
2844                            "%3d %6.6s %pi6 %pi6 %6lu %6lu\n",
2845                            state->dev->ifindex, state->dev->name,
2846                            &state->im->mca_addr,
2847                            &psf->sf_addr,
2848                            psf->sf_count[MCAST_INCLUDE],
2849                            psf->sf_count[MCAST_EXCLUDE]);
2850         }
2851         return 0;
2852 }
2853
2854 static const struct seq_operations igmp6_mcf_seq_ops = {
2855         .start  =       igmp6_mcf_seq_start,
2856         .next   =       igmp6_mcf_seq_next,
2857         .stop   =       igmp6_mcf_seq_stop,
2858         .show   =       igmp6_mcf_seq_show,
2859 };
2860
2861 static int igmp6_mcf_seq_open(struct inode *inode, struct file *file)
2862 {
2863         return seq_open_net(inode, file, &igmp6_mcf_seq_ops,
2864                             sizeof(struct igmp6_mcf_iter_state));
2865 }
2866
2867 static const struct file_operations igmp6_mcf_seq_fops = {
2868         .owner          =       THIS_MODULE,
2869         .open           =       igmp6_mcf_seq_open,
2870         .read           =       seq_read,
2871         .llseek         =       seq_lseek,
2872         .release        =       seq_release_net,
2873 };
2874
2875 static int __net_init igmp6_proc_init(struct net *net)
2876 {
2877         int err;
2878
2879         err = -ENOMEM;
2880         if (!proc_create("igmp6", S_IRUGO, net->proc_net, &igmp6_mc_seq_fops))
2881                 goto out;
2882         if (!proc_create("mcfilter6", S_IRUGO, net->proc_net,
2883                          &igmp6_mcf_seq_fops))
2884                 goto out_proc_net_igmp6;
2885
2886         err = 0;
2887 out:
2888         return err;
2889
2890 out_proc_net_igmp6:
2891         remove_proc_entry("igmp6", net->proc_net);
2892         goto out;
2893 }
2894
2895 static void __net_exit igmp6_proc_exit(struct net *net)
2896 {
2897         remove_proc_entry("mcfilter6", net->proc_net);
2898         remove_proc_entry("igmp6", net->proc_net);
2899 }
2900 #else
2901 static inline int igmp6_proc_init(struct net *net)
2902 {
2903         return 0;
2904 }
2905 static inline void igmp6_proc_exit(struct net *net)
2906 {
2907 }
2908 #endif
2909
2910 static int __net_init igmp6_net_init(struct net *net)
2911 {
2912         int err;
2913
2914         err = inet_ctl_sock_create(&net->ipv6.igmp_sk, PF_INET6,
2915                                    SOCK_RAW, IPPROTO_ICMPV6, net);
2916         if (err < 0) {
2917                 pr_err("Failed to initialize the IGMP6 control socket (err %d)\n",
2918                        err);
2919                 goto out;
2920         }
2921
2922         inet6_sk(net->ipv6.igmp_sk)->hop_limit = 1;
2923
2924         err = inet_ctl_sock_create(&net->ipv6.mc_autojoin_sk, PF_INET6,
2925                                    SOCK_RAW, IPPROTO_ICMPV6, net);
2926         if (err < 0) {
2927                 pr_err("Failed to initialize the IGMP6 autojoin socket (err %d)\n",
2928                        err);
2929                 goto out_sock_create;
2930         }
2931
2932         err = igmp6_proc_init(net);
2933         if (err)
2934                 goto out_sock_create_autojoin;
2935
2936         return 0;
2937
2938 out_sock_create_autojoin:
2939         inet_ctl_sock_destroy(net->ipv6.mc_autojoin_sk);
2940 out_sock_create:
2941         inet_ctl_sock_destroy(net->ipv6.igmp_sk);
2942 out:
2943         return err;
2944 }
2945
2946 static void __net_exit igmp6_net_exit(struct net *net)
2947 {
2948         inet_ctl_sock_destroy(net->ipv6.igmp_sk);
2949         inet_ctl_sock_destroy(net->ipv6.mc_autojoin_sk);
2950         igmp6_proc_exit(net);
2951 }
2952
2953 static struct pernet_operations igmp6_net_ops = {
2954         .init = igmp6_net_init,
2955         .exit = igmp6_net_exit,
2956 };
2957
2958 int __init igmp6_init(void)
2959 {
2960         return register_pernet_subsys(&igmp6_net_ops);
2961 }
2962
2963 void igmp6_cleanup(void)
2964 {
2965         unregister_pernet_subsys(&igmp6_net_ops);
2966 }