net/ipv6/mcast.c

   1 // SPDX-License-Identifier: GPL-2.0-or-later
   2 /*
   3  *      Multicast support for IPv6
   4  *      Linux INET6 implementation
   5  *
   6  *      Authors:
   7  *      Pedro Roque             <[email protected]>
   8  *
   9  *      Based on linux/ipv4/igmp.c and linux/ipv4/ip_sockglue.c
  10  */
  11
  12 /* Changes:
  13  *
  14  *      yoshfuji        : fix format of router-alert option
  15  *      YOSHIFUJI Hideaki @USAGI:
  16  *              Fixed source address for MLD message based on
  17  *              <draft-ietf-magma-mld-source-05.txt>.
  18  *      YOSHIFUJI Hideaki @USAGI:
  19  *              - Ignore Queries for invalid addresses.
  20  *              - MLD for link-local addresses.
  21  *      David L Stevens <[email protected]>:
  22  *              - MLDv2 support
  23  */
  24
  25 #include <linux/module.h>
  26 #include <linux/errno.h>
  27 #include <linux/types.h>
  28 #include <linux/string.h>
  29 #include <linux/socket.h>
  30 #include <linux/sockios.h>
  31 #include <linux/jiffies.h>
  32 #include <linux/net.h>
  33 #include <linux/in.h>
  34 #include <linux/in6.h>
  35 #include <linux/netdevice.h>
  36 #include <linux/if_arp.h>
  37 #include <linux/route.h>
  38 #include <linux/init.h>
  39 #include <linux/proc_fs.h>
  40 #include <linux/seq_file.h>
  41 #include <linux/slab.h>
  42 #include <linux/pkt_sched.h>
  43 #include <net/mld.h>
  44 #include <linux/workqueue.h>
  45
  46 #include <linux/netfilter.h>
  47 #include <linux/netfilter_ipv6.h>
  48
  49 #include <net/net_namespace.h>
  50 #include <net/sock.h>
  51 #include <net/snmp.h>
  52
  53 #include <net/ipv6.h>
  54 #include <net/protocol.h>
  55 #include <net/if_inet6.h>
  56 #include <net/ndisc.h>
  57 #include <net/addrconf.h>
  58 #include <net/ip6_route.h>
  59 #include <net/inet_common.h>
  60
  61 #include <net/ip6_checksum.h>
  62
  63 /* Ensure that we have struct in6_addr aligned on 32bit word. */
  64 static int __mld2_query_bugs[] __attribute__((__unused__)) = {
  65         BUILD_BUG_ON_ZERO(offsetof(struct mld2_query, mld2q_srcs) % 4),
  66         BUILD_BUG_ON_ZERO(offsetof(struct mld2_report, mld2r_grec) % 4),
  67         BUILD_BUG_ON_ZERO(offsetof(struct mld2_grec, grec_mca) % 4)
  68 };
  69
  70 static struct workqueue_struct *mld_wq;
  71 static struct in6_addr mld2_all_mcr = MLD2_ALL_MCR_INIT;
  72
  73 static void igmp6_join_group(struct ifmcaddr6 *ma);
  74 static void igmp6_leave_group(struct ifmcaddr6 *ma);
  75 static void mld_mca_work(struct work_struct *work);
  76
  77 static void mld_ifc_event(struct inet6_dev *idev);
  78 static bool mld_in_v1_mode(const struct inet6_dev *idev);
  79 static int sf_setstate(struct ifmcaddr6 *pmc);
  80 static void sf_markstate(struct ifmcaddr6 *pmc);
  81 static void ip6_mc_clear_src(struct ifmcaddr6 *pmc);
  82 static int ip6_mc_del_src(struct inet6_dev *idev, const struct in6_addr *pmca,
  83                           int sfmode, int sfcount, const struct in6_addr *psfsrc,
  84                           int delta);
  85 static int ip6_mc_add_src(struct inet6_dev *idev, const struct in6_addr *pmca,
  86                           int sfmode, int sfcount, const struct in6_addr *psfsrc,
  87                           int delta);
  88 static int ip6_mc_leave_src(struct sock *sk, struct ipv6_mc_socklist *iml,
  89                             struct inet6_dev *idev);
  90 static int __ipv6_dev_mc_inc(struct net_device *dev,
  91                              const struct in6_addr *addr, unsigned int mode);
  92
  93 #define MLD_QRV_DEFAULT         2
  94 /* RFC3810, 9.2. Query Interval */
  95 #define MLD_QI_DEFAULT          (125 * HZ)
  96 /* RFC3810, 9.3. Query Response Interval */
  97 #define MLD_QRI_DEFAULT         (10 * HZ)
  98
  99 /* RFC3810, 8.1 Query Version Distinctions */
 100 #define MLD_V1_QUERY_LEN        24
 101 #define MLD_V2_QUERY_LEN_MIN    28
 102
 103 #define IPV6_MLD_MAX_MSF        64
 104
 105 int sysctl_mld_max_msf __read_mostly = IPV6_MLD_MAX_MSF;
 106 int sysctl_mld_qrv __read_mostly = MLD_QRV_DEFAULT;
 107
 108 /*
 109  *      socket join on multicast group
 110  */
 111 #define mc_dereference(e, idev) \
 112         rcu_dereference_protected(e, lockdep_is_held(&(idev)->mc_lock))
 113
 114 #define sock_dereference(e, sk) \
 115         rcu_dereference_protected(e, lockdep_sock_is_held(sk))
 116
 117 #define for_each_pmc_socklock(np, sk, pmc)                      \
 118         for (pmc = sock_dereference((np)->ipv6_mc_list, sk);    \
 119              pmc;                                               \
 120              pmc = sock_dereference(pmc->next, sk))
 121
 122 #define for_each_pmc_rcu(np, pmc)                               \
 123         for (pmc = rcu_dereference((np)->ipv6_mc_list);         \
 124              pmc;                                               \
 125              pmc = rcu_dereference(pmc->next))
 126
 127 #define for_each_psf_mclock(mc, psf)                            \
 128         for (psf = mc_dereference((mc)->mca_sources, mc->idev); \
 129              psf;                                               \
 130              psf = mc_dereference(psf->sf_next, mc->idev))
 131
 132 #define for_each_psf_rcu(mc, psf)                               \
 133         for (psf = rcu_dereference((mc)->mca_sources);          \
 134              psf;                                               \
 135              psf = rcu_dereference(psf->sf_next))
 136
 137 #define for_each_psf_tomb(mc, psf)                              \
 138         for (psf = mc_dereference((mc)->mca_tomb, mc->idev);    \
 139              psf;                                               \
 140              psf = mc_dereference(psf->sf_next, mc->idev))
 141
 142 #define for_each_mc_mclock(idev, mc)                            \
 143         for (mc = mc_dereference((idev)->mc_list, idev);        \
 144              mc;                                                \
 145              mc = mc_dereference(mc->next, idev))
 146
 147 #define for_each_mc_rcu(idev, mc)                               \
 148         for (mc = rcu_dereference((idev)->mc_list);             \
 149              mc;                                                \
 150              mc = rcu_dereference(mc->next))
 151
 152 #define for_each_mc_tomb(idev, mc)                              \
 153         for (mc = mc_dereference((idev)->mc_tomb, idev);        \
 154              mc;                                                \
 155              mc = mc_dereference(mc->next, idev))
 156
 157 static int unsolicited_report_interval(struct inet6_dev *idev)
 158 {
 159         int iv;
 160
 161         if (mld_in_v1_mode(idev))
 162                 iv = idev->cnf.mldv1_unsolicited_report_interval;
 163         else
 164                 iv = idev->cnf.mldv2_unsolicited_report_interval;
 165
 166         return iv > 0 ? iv : 1;
 167 }
 168
 169 static int __ipv6_sock_mc_join(struct sock *sk, int ifindex,
 170                                const struct in6_addr *addr, unsigned int mode)
 171 {
 172         struct net_device *dev = NULL;
 173         struct ipv6_mc_socklist *mc_lst;
 174         struct ipv6_pinfo *np = inet6_sk(sk);
 175         struct net *net = sock_net(sk);
 176         int err;
 177
 178         ASSERT_RTNL();
 179
 180         if (!ipv6_addr_is_multicast(addr))
 181                 return -EINVAL;
 182
 183         for_each_pmc_socklock(np, sk, mc_lst) {
 184                 if ((ifindex == 0 || mc_lst->ifindex == ifindex) &&
 185                     ipv6_addr_equal(&mc_lst->addr, addr))
 186                         return -EADDRINUSE;
 187         }
 188
 189         mc_lst = sock_kmalloc(sk, sizeof(struct ipv6_mc_socklist), GFP_KERNEL);
 190
 191         if (!mc_lst)
 192                 return -ENOMEM;
 193
 194         mc_lst->next = NULL;
 195         mc_lst->addr = *addr;
 196
 197         if (ifindex == 0) {
 198                 struct rt6_info *rt;
 199                 rt = rt6_lookup(net, addr, NULL, 0, NULL, 0);
 200                 if (rt) {
 201                         dev = rt->dst.dev;
 202                         ip6_rt_put(rt);
 203                 }
 204         } else
 205                 dev = __dev_get_by_index(net, ifindex);
 206
 207         if (!dev) {
 208                 sock_kfree_s(sk, mc_lst, sizeof(*mc_lst));
 209                 return -ENODEV;
 210         }
 211
 212         mc_lst->ifindex = dev->ifindex;
 213         mc_lst->sfmode = mode;
 214         RCU_INIT_POINTER(mc_lst->sflist, NULL);
 215
 216         /*
 217          *      now add/increase the group membership on the device
 218          */
 219
 220         err = __ipv6_dev_mc_inc(dev, addr, mode);
 221
 222         if (err) {
 223                 sock_kfree_s(sk, mc_lst, sizeof(*mc_lst));
 224                 return err;
 225         }
 226
 227         mc_lst->next = np->ipv6_mc_list;
 228         rcu_assign_pointer(np->ipv6_mc_list, mc_lst);
 229
 230         return 0;
 231 }
 232
 233 int ipv6_sock_mc_join(struct sock *sk, int ifindex, const struct in6_addr *addr)
 234 {
 235         return __ipv6_sock_mc_join(sk, ifindex, addr, MCAST_EXCLUDE);
 236 }
 237 EXPORT_SYMBOL(ipv6_sock_mc_join);
 238
 239 int ipv6_sock_mc_join_ssm(struct sock *sk, int ifindex,
 240                           const struct in6_addr *addr, unsigned int mode)
 241 {
 242         return __ipv6_sock_mc_join(sk, ifindex, addr, mode);
 243 }
 244
 245 /*
 246  *      socket leave on multicast group
 247  */
 248 int ipv6_sock_mc_drop(struct sock *sk, int ifindex, const struct in6_addr *addr)
 249 {
 250         struct ipv6_pinfo *np = inet6_sk(sk);
 251         struct ipv6_mc_socklist *mc_lst;
 252         struct ipv6_mc_socklist __rcu **lnk;
 253         struct net *net = sock_net(sk);
 254
 255         ASSERT_RTNL();
 256
 257         if (!ipv6_addr_is_multicast(addr))
 258                 return -EINVAL;
 259
 260         for (lnk = &np->ipv6_mc_list;
 261              (mc_lst = sock_dereference(*lnk, sk)) != NULL;
 262               lnk = &mc_lst->next) {
 263                 if ((ifindex == 0 || mc_lst->ifindex == ifindex) &&
 264                     ipv6_addr_equal(&mc_lst->addr, addr)) {
 265                         struct net_device *dev;
 266
 267                         *lnk = mc_lst->next;
 268
 269                         dev = __dev_get_by_index(net, mc_lst->ifindex);
 270                         if (dev) {
 271                                 struct inet6_dev *idev = __in6_dev_get(dev);
 272
 273                                 ip6_mc_leave_src(sk, mc_lst, idev);
 274                                 if (idev)
 275                                         __ipv6_dev_mc_dec(idev, &mc_lst->addr);
 276                         } else {
 277                                 ip6_mc_leave_src(sk, mc_lst, NULL);
 278                         }
 279
 280                         atomic_sub(sizeof(*mc_lst), &sk->sk_omem_alloc);
 281                         kfree_rcu(mc_lst, rcu);
 282                         return 0;
 283                 }
 284         }
 285
 286         return -EADDRNOTAVAIL;
 287 }
 288 EXPORT_SYMBOL(ipv6_sock_mc_drop);
 289
 290 static struct inet6_dev *ip6_mc_find_dev_rtnl(struct net *net,
 291                                               const struct in6_addr *group,
 292                                               int ifindex)
 293 {
 294         struct net_device *dev = NULL;
 295         struct inet6_dev *idev = NULL;
 296
 297         if (ifindex == 0) {
 298                 struct rt6_info *rt = rt6_lookup(net, group, NULL, 0, NULL, 0);
 299
 300                 if (rt) {
 301                         dev = rt->dst.dev;
 302                         ip6_rt_put(rt);
 303                 }
 304         } else {
 305                 dev = __dev_get_by_index(net, ifindex);
 306         }
 307
 308         if (!dev)
 309                 return NULL;
 310         idev = __in6_dev_get(dev);
 311         if (!idev)
 312                 return NULL;
 313         if (idev->dead)
 314                 return NULL;
 315         return idev;
 316 }
 317
 318 void __ipv6_sock_mc_close(struct sock *sk)
 319 {
 320         struct ipv6_pinfo *np = inet6_sk(sk);
 321         struct ipv6_mc_socklist *mc_lst;
 322         struct net *net = sock_net(sk);
 323
 324         ASSERT_RTNL();
 325
 326         while ((mc_lst = sock_dereference(np->ipv6_mc_list, sk)) != NULL) {
 327                 struct net_device *dev;
 328
 329                 np->ipv6_mc_list = mc_lst->next;
 330
 331                 dev = __dev_get_by_index(net, mc_lst->ifindex);
 332                 if (dev) {
 333                         struct inet6_dev *idev = __in6_dev_get(dev);
 334
 335                         ip6_mc_leave_src(sk, mc_lst, idev);
 336                         if (idev)
 337                                 __ipv6_dev_mc_dec(idev, &mc_lst->addr);
 338                 } else {
 339                         ip6_mc_leave_src(sk, mc_lst, NULL);
 340                 }
 341
 342                 atomic_sub(sizeof(*mc_lst), &sk->sk_omem_alloc);
 343                 kfree_rcu(mc_lst, rcu);
 344         }
 345 }
 346
 347 void ipv6_sock_mc_close(struct sock *sk)
 348 {
 349         struct ipv6_pinfo *np = inet6_sk(sk);
 350
 351         if (!rcu_access_pointer(np->ipv6_mc_list))
 352                 return;
 353
 354         rtnl_lock();
 355         lock_sock(sk);
 356         __ipv6_sock_mc_close(sk);
 357         release_sock(sk);
 358         rtnl_unlock();
 359 }
 360
 361 int ip6_mc_source(int add, int omode, struct sock *sk,
 362         struct group_source_req *pgsr)
 363 {
 364         struct in6_addr *source, *group;
 365         struct ipv6_mc_socklist *pmc;
 366         struct inet6_dev *idev;
 367         struct ipv6_pinfo *inet6 = inet6_sk(sk);
 368         struct ip6_sf_socklist *psl;
 369         struct net *net = sock_net(sk);
 370         int i, j, rv;
 371         int leavegroup = 0;
 372         int err;
 373
 374         source = &((struct sockaddr_in6 *)&pgsr->gsr_source)->sin6_addr;
 375         group = &((struct sockaddr_in6 *)&pgsr->gsr_group)->sin6_addr;
 376
 377         if (!ipv6_addr_is_multicast(group))
 378                 return -EINVAL;
 379
 380         idev = ip6_mc_find_dev_rtnl(net, group, pgsr->gsr_interface);
 381         if (!idev)
 382                 return -ENODEV;
 383
 384         err = -EADDRNOTAVAIL;
 385
 386         mutex_lock(&idev->mc_lock);
 387         for_each_pmc_socklock(inet6, sk, pmc) {
 388                 if (pgsr->gsr_interface && pmc->ifindex != pgsr->gsr_interface)
 389                         continue;
 390                 if (ipv6_addr_equal(&pmc->addr, group))
 391                         break;
 392         }
 393         if (!pmc) {             /* must have a prior join */
 394                 err = -EINVAL;
 395                 goto done;
 396         }
 397         /* if a source filter was set, must be the same mode as before */
 398         if (rcu_access_pointer(pmc->sflist)) {
 399                 if (pmc->sfmode != omode) {
 400                         err = -EINVAL;
 401                         goto done;
 402                 }
 403         } else if (pmc->sfmode != omode) {
 404                 /* allow mode switches for empty-set filters */
 405                 ip6_mc_add_src(idev, group, omode, 0, NULL, 0);
 406                 ip6_mc_del_src(idev, group, pmc->sfmode, 0, NULL, 0);
 407                 pmc->sfmode = omode;
 408         }
 409
 410         psl = sock_dereference(pmc->sflist, sk);
 411         if (!add) {
 412                 if (!psl)
 413                         goto done;      /* err = -EADDRNOTAVAIL */
 414                 rv = !0;
 415                 for (i = 0; i < psl->sl_count; i++) {
 416                         rv = !ipv6_addr_equal(&psl->sl_addr[i], source);
 417                         if (rv == 0)
 418                                 break;
 419                 }
 420                 if (rv)         /* source not found */
 421                         goto done;      /* err = -EADDRNOTAVAIL */
 422
 423                 /* special case - (INCLUDE, empty) == LEAVE_GROUP */
 424                 if (psl->sl_count == 1 && omode == MCAST_INCLUDE) {
 425                         leavegroup = 1;
 426                         goto done;
 427                 }
 428
 429                 /* update the interface filter */
 430                 ip6_mc_del_src(idev, group, omode, 1, source, 1);
 431
 432                 for (j = i+1; j < psl->sl_count; j++)
 433                         psl->sl_addr[j-1] = psl->sl_addr[j];
 434                 psl->sl_count--;
 435                 err = 0;
 436                 goto done;
 437         }
 438         /* else, add a new source to the filter */
 439
 440         if (psl && psl->sl_count >= sysctl_mld_max_msf) {
 441                 err = -ENOBUFS;
 442                 goto done;
 443         }
 444         if (!psl || psl->sl_count == psl->sl_max) {
 445                 struct ip6_sf_socklist *newpsl;
 446                 int count = IP6_SFBLOCK;
 447
 448                 if (psl)
 449                         count += psl->sl_max;
 450                 newpsl = sock_kmalloc(sk, IP6_SFLSIZE(count), GFP_KERNEL);
 451                 if (!newpsl) {
 452                         err = -ENOBUFS;
 453                         goto done;
 454                 }
 455                 newpsl->sl_max = count;
 456                 newpsl->sl_count = count - IP6_SFBLOCK;
 457                 if (psl) {
 458                         for (i = 0; i < psl->sl_count; i++)
 459                                 newpsl->sl_addr[i] = psl->sl_addr[i];
 460                         atomic_sub(IP6_SFLSIZE(psl->sl_max), &sk->sk_omem_alloc);
 461                         kfree_rcu(psl, rcu);
 462                 }
 463                 psl = newpsl;
 464                 rcu_assign_pointer(pmc->sflist, psl);
 465         }
 466         rv = 1; /* > 0 for insert logic below if sl_count is 0 */
 467         for (i = 0; i < psl->sl_count; i++) {
 468                 rv = !ipv6_addr_equal(&psl->sl_addr[i], source);
 469                 if (rv == 0) /* There is an error in the address. */
 470                         goto done;
 471         }
 472         for (j = psl->sl_count-1; j >= i; j--)
 473                 psl->sl_addr[j+1] = psl->sl_addr[j];
 474         psl->sl_addr[i] = *source;
 475         psl->sl_count++;
 476         err = 0;
 477         /* update the interface list */
 478         ip6_mc_add_src(idev, group, omode, 1, source, 1);
 479 done:
 480         mutex_unlock(&idev->mc_lock);
 481         if (leavegroup)
 482                 err = ipv6_sock_mc_drop(sk, pgsr->gsr_interface, group);
 483         return err;
 484 }
 485
 486 int ip6_mc_msfilter(struct sock *sk, struct group_filter *gsf,
 487                     struct sockaddr_storage *list)
 488 {
 489         const struct in6_addr *group;
 490         struct ipv6_mc_socklist *pmc;
 491         struct inet6_dev *idev;
 492         struct ipv6_pinfo *inet6 = inet6_sk(sk);
 493         struct ip6_sf_socklist *newpsl, *psl;
 494         struct net *net = sock_net(sk);
 495         int leavegroup = 0;
 496         int i, err;
 497
 498         group = &((struct sockaddr_in6 *)&gsf->gf_group)->sin6_addr;
 499
 500         if (!ipv6_addr_is_multicast(group))
 501                 return -EINVAL;
 502         if (gsf->gf_fmode != MCAST_INCLUDE &&
 503             gsf->gf_fmode != MCAST_EXCLUDE)
 504                 return -EINVAL;
 505
 506         idev = ip6_mc_find_dev_rtnl(net, group, gsf->gf_interface);
 507         if (!idev)
 508                 return -ENODEV;
 509
 510         err = 0;
 511
 512         if (gsf->gf_fmode == MCAST_INCLUDE && gsf->gf_numsrc == 0) {
 513                 leavegroup = 1;
 514                 goto done;
 515         }
 516
 517         for_each_pmc_socklock(inet6, sk, pmc) {
 518                 if (pmc->ifindex != gsf->gf_interface)
 519                         continue;
 520                 if (ipv6_addr_equal(&pmc->addr, group))
 521                         break;
 522         }
 523         if (!pmc) {             /* must have a prior join */
 524                 err = -EINVAL;
 525                 goto done;
 526         }
 527         if (gsf->gf_numsrc) {
 528                 newpsl = sock_kmalloc(sk, IP6_SFLSIZE(gsf->gf_numsrc),
 529                                                           GFP_KERNEL);
 530                 if (!newpsl) {
 531                         err = -ENOBUFS;
 532                         goto done;
 533                 }
 534                 newpsl->sl_max = newpsl->sl_count = gsf->gf_numsrc;
 535                 for (i = 0; i < newpsl->sl_count; ++i, ++list) {
 536                         struct sockaddr_in6 *psin6;
 537
 538                         psin6 = (struct sockaddr_in6 *)list;
 539                         newpsl->sl_addr[i] = psin6->sin6_addr;
 540                 }
 541                 mutex_lock(&idev->mc_lock);
 542                 err = ip6_mc_add_src(idev, group, gsf->gf_fmode,
 543                                      newpsl->sl_count, newpsl->sl_addr, 0);
 544                 if (err) {
 545                         mutex_unlock(&idev->mc_lock);
 546                         sock_kfree_s(sk, newpsl, IP6_SFLSIZE(newpsl->sl_max));
 547                         goto done;
 548                 }
 549                 mutex_unlock(&idev->mc_lock);
 550         } else {
 551                 newpsl = NULL;
 552                 mutex_lock(&idev->mc_lock);
 553                 ip6_mc_add_src(idev, group, gsf->gf_fmode, 0, NULL, 0);
 554                 mutex_unlock(&idev->mc_lock);
 555         }
 556
 557         mutex_lock(&idev->mc_lock);
 558         psl = sock_dereference(pmc->sflist, sk);
 559         if (psl) {
 560                 ip6_mc_del_src(idev, group, pmc->sfmode,
 561                                psl->sl_count, psl->sl_addr, 0);
 562                 atomic_sub(IP6_SFLSIZE(psl->sl_max), &sk->sk_omem_alloc);
 563                 kfree_rcu(psl, rcu);
 564         } else {
 565                 ip6_mc_del_src(idev, group, pmc->sfmode, 0, NULL, 0);
 566         }
 567         mutex_unlock(&idev->mc_lock);
 568         rcu_assign_pointer(pmc->sflist, newpsl);
 569         pmc->sfmode = gsf->gf_fmode;
 570         err = 0;
 571 done:
 572         if (leavegroup)
 573                 err = ipv6_sock_mc_drop(sk, gsf->gf_interface, group);
 574         return err;
 575 }
 576
 577 int ip6_mc_msfget(struct sock *sk, struct group_filter *gsf,
 578                   struct sockaddr_storage __user *p)
 579 {
 580         struct ipv6_pinfo *inet6 = inet6_sk(sk);
 581         const struct in6_addr *group;
 582         struct ipv6_mc_socklist *pmc;
 583         struct ip6_sf_socklist *psl;
 584         int i, count, copycount;
 585
 586         group = &((struct sockaddr_in6 *)&gsf->gf_group)->sin6_addr;
 587
 588         if (!ipv6_addr_is_multicast(group))
 589                 return -EINVAL;
 590
 591         /* changes to the ipv6_mc_list require the socket lock and
 592          * rtnl lock. We have the socket lock, so reading the list is safe.
 593          */
 594
 595         for_each_pmc_socklock(inet6, sk, pmc) {
 596                 if (pmc->ifindex != gsf->gf_interface)
 597                         continue;
 598                 if (ipv6_addr_equal(group, &pmc->addr))
 599                         break;
 600         }
 601         if (!pmc)               /* must have a prior join */
 602                 return -EADDRNOTAVAIL;
 603
 604         gsf->gf_fmode = pmc->sfmode;
 605         psl = sock_dereference(pmc->sflist, sk);
 606         count = psl ? psl->sl_count : 0;
 607
 608         copycount = count < gsf->gf_numsrc ? count : gsf->gf_numsrc;
 609         gsf->gf_numsrc = count;
 610
 611         for (i = 0; i < copycount; i++, p++) {
 612                 struct sockaddr_in6 *psin6;
 613                 struct sockaddr_storage ss;
 614
 615                 psin6 = (struct sockaddr_in6 *)&ss;
 616                 memset(&ss, 0, sizeof(ss));
 617                 psin6->sin6_family = AF_INET6;
 618                 psin6->sin6_addr = psl->sl_addr[i];
 619                 if (copy_to_user(p, &ss, sizeof(ss)))
 620                         return -EFAULT;
 621         }
 622         return 0;
 623 }
 624
 625 bool inet6_mc_check(struct sock *sk, const struct in6_addr *mc_addr,
 626                     const struct in6_addr *src_addr)
 627 {
 628         struct ipv6_pinfo *np = inet6_sk(sk);
 629         struct ipv6_mc_socklist *mc;
 630         struct ip6_sf_socklist *psl;
 631         bool rv = true;
 632
 633         rcu_read_lock();
 634         for_each_pmc_rcu(np, mc) {
 635                 if (ipv6_addr_equal(&mc->addr, mc_addr))
 636                         break;
 637         }
 638         if (!mc) {
 639                 rcu_read_unlock();
 640                 return np->mc_all;
 641         }
 642         psl = rcu_dereference(mc->sflist);
 643         if (!psl) {
 644                 rv = mc->sfmode == MCAST_EXCLUDE;
 645         } else {
 646                 int i;
 647
 648                 for (i = 0; i < psl->sl_count; i++) {
 649                         if (ipv6_addr_equal(&psl->sl_addr[i], src_addr))
 650                                 break;
 651                 }
 652                 if (mc->sfmode == MCAST_INCLUDE && i >= psl->sl_count)
 653                         rv = false;
 654                 if (mc->sfmode == MCAST_EXCLUDE && i < psl->sl_count)
 655                         rv = false;
 656         }
 657         rcu_read_unlock();
 658
 659         return rv;
 660 }
 661
 662 /* called with mc_lock */
 663 static void igmp6_group_added(struct ifmcaddr6 *mc)
 664 {
 665         struct net_device *dev = mc->idev->dev;
 666         char buf[MAX_ADDR_LEN];
 667
 668         if (IPV6_ADDR_MC_SCOPE(&mc->mca_addr) <
 669             IPV6_ADDR_SCOPE_LINKLOCAL)
 670                 return;
 671
 672         if (!(mc->mca_flags&MAF_LOADED)) {
 673                 mc->mca_flags |= MAF_LOADED;
 674                 if (ndisc_mc_map(&mc->mca_addr, buf, dev, 0) == 0)
 675                         dev_mc_add(dev, buf);
 676         }
 677
 678         if (!(dev->flags & IFF_UP) || (mc->mca_flags & MAF_NOREPORT))
 679                 return;
 680
 681         if (mld_in_v1_mode(mc->idev)) {
 682                 igmp6_join_group(mc);
 683                 return;
 684         }
 685         /* else v2 */
 686
 687         /* Based on RFC3810 6.1, for newly added INCLUDE SSM, we
 688          * should not send filter-mode change record as the mode
 689          * should be from IN() to IN(A).
 690          */
 691         if (mc->mca_sfmode == MCAST_EXCLUDE)
 692                 mc->mca_crcount = mc->idev->mc_qrv;
 693
 694         mld_ifc_event(mc->idev);
 695 }
 696
 697 /* called with mc_lock */
 698 static void igmp6_group_dropped(struct ifmcaddr6 *mc)
 699 {
 700         struct net_device *dev = mc->idev->dev;
 701         char buf[MAX_ADDR_LEN];
 702
 703         if (IPV6_ADDR_MC_SCOPE(&mc->mca_addr) <
 704             IPV6_ADDR_SCOPE_LINKLOCAL)
 705                 return;
 706
 707         if (mc->mca_flags&MAF_LOADED) {
 708                 mc->mca_flags &= ~MAF_LOADED;
 709                 if (ndisc_mc_map(&mc->mca_addr, buf, dev, 0) == 0)
 710                         dev_mc_del(dev, buf);
 711         }
 712
 713         if (mc->mca_flags & MAF_NOREPORT)
 714                 return;
 715
 716         if (!mc->idev->dead)
 717                 igmp6_leave_group(mc);
 718
 719         if (cancel_delayed_work(&mc->mca_work))
 720                 refcount_dec(&mc->mca_refcnt);
 721 }
 722
 723 /*
 724  * deleted ifmcaddr6 manipulation
 725  * called with mc_lock
 726  */
 727 static void mld_add_delrec(struct inet6_dev *idev, struct ifmcaddr6 *im)
 728 {
 729         struct ifmcaddr6 *pmc;
 730
 731         /* this is an "ifmcaddr6" for convenience; only the fields below
 732          * are actually used. In particular, the refcnt and users are not
 733          * used for management of the delete list. Using the same structure
 734          * for deleted items allows change reports to use common code with
 735          * non-deleted or query-response MCA's.
 736          */
 737         pmc = kzalloc(sizeof(*pmc), GFP_KERNEL);
 738         if (!pmc)
 739                 return;
 740
 741         pmc->idev = im->idev;
 742         in6_dev_hold(idev);
 743         pmc->mca_addr = im->mca_addr;
 744         pmc->mca_crcount = idev->mc_qrv;
 745         pmc->mca_sfmode = im->mca_sfmode;
 746         if (pmc->mca_sfmode == MCAST_INCLUDE) {
 747                 struct ip6_sf_list *psf;
 748
 749                 rcu_assign_pointer(pmc->mca_tomb,
 750                                    mc_dereference(im->mca_tomb, idev));
 751                 rcu_assign_pointer(pmc->mca_sources,
 752                                    mc_dereference(im->mca_sources, idev));
 753                 RCU_INIT_POINTER(im->mca_tomb, NULL);
 754                 RCU_INIT_POINTER(im->mca_sources, NULL);
 755
 756                 for_each_psf_mclock(pmc, psf)
 757                         psf->sf_crcount = pmc->mca_crcount;
 758         }
 759
 760         rcu_assign_pointer(pmc->next, idev->mc_tomb);
 761         rcu_assign_pointer(idev->mc_tomb, pmc);
 762 }
 763
 764 /* called with mc_lock */
 765 static void mld_del_delrec(struct inet6_dev *idev, struct ifmcaddr6 *im)
 766 {
 767         struct ip6_sf_list *psf, *sources, *tomb;
 768         struct in6_addr *pmca = &im->mca_addr;
 769         struct ifmcaddr6 *pmc, *pmc_prev;
 770
 771         pmc_prev = NULL;
 772         for_each_mc_tomb(idev, pmc) {
 773                 if (ipv6_addr_equal(&pmc->mca_addr, pmca))
 774                         break;
 775                 pmc_prev = pmc;
 776         }
 777         if (pmc) {
 778                 if (pmc_prev)
 779                         rcu_assign_pointer(pmc_prev->next, pmc->next);
 780                 else
 781                         rcu_assign_pointer(idev->mc_tomb, pmc->next);
 782         }
 783
 784         if (pmc) {
 785                 im->idev = pmc->idev;
 786                 if (im->mca_sfmode == MCAST_INCLUDE) {
 787                         tomb = rcu_replace_pointer(im->mca_tomb,
 788                                                    mc_dereference(pmc->mca_tomb, pmc->idev),
 789                                                    lockdep_is_held(&im->idev->mc_lock));
 790                         rcu_assign_pointer(pmc->mca_tomb, tomb);
 791
 792                         sources = rcu_replace_pointer(im->mca_sources,
 793                                                       mc_dereference(pmc->mca_sources, pmc->idev),
 794                                                       lockdep_is_held(&im->idev->mc_lock));
 795                         rcu_assign_pointer(pmc->mca_sources, sources);
 796                         for_each_psf_mclock(im, psf)
 797                                 psf->sf_crcount = idev->mc_qrv;
 798                 } else {
 799                         im->mca_crcount = idev->mc_qrv;
 800                 }
 801                 in6_dev_put(pmc->idev);
 802                 ip6_mc_clear_src(pmc);
 803                 kfree_rcu(pmc, rcu);
 804         }
 805 }
 806
 807 /* called with mc_lock */
 808 static void mld_clear_delrec(struct inet6_dev *idev)
 809 {
 810         struct ifmcaddr6 *pmc, *nextpmc;
 811
 812         pmc = mc_dereference(idev->mc_tomb, idev);
 813         RCU_INIT_POINTER(idev->mc_tomb, NULL);
 814
 815         for (; pmc; pmc = nextpmc) {
 816                 nextpmc = mc_dereference(pmc->next, idev);
 817                 ip6_mc_clear_src(pmc);
 818                 in6_dev_put(pmc->idev);
 819                 kfree_rcu(pmc, rcu);
 820         }
 821
 822         /* clear dead sources, too */
 823         for_each_mc_mclock(idev, pmc) {
 824                 struct ip6_sf_list *psf, *psf_next;
 825
 826                 psf = mc_dereference(pmc->mca_tomb, idev);
 827                 RCU_INIT_POINTER(pmc->mca_tomb, NULL);
 828                 for (; psf; psf = psf_next) {
 829                         psf_next = mc_dereference(psf->sf_next, idev);
 830                         kfree_rcu(psf, rcu);
 831                 }
 832         }
 833 }
 834
 835 static void mld_clear_query(struct inet6_dev *idev)
 836 {
 837         struct sk_buff *skb;
 838
 839         spin_lock_bh(&idev->mc_query_lock);
 840         while ((skb = __skb_dequeue(&idev->mc_query_queue)))
 841                 kfree_skb(skb);
 842         spin_unlock_bh(&idev->mc_query_lock);
 843 }
 844
 845 static void mld_clear_report(struct inet6_dev *idev)
 846 {
 847         struct sk_buff *skb;
 848
 849         spin_lock_bh(&idev->mc_report_lock);
 850         while ((skb = __skb_dequeue(&idev->mc_report_queue)))
 851                 kfree_skb(skb);
 852         spin_unlock_bh(&idev->mc_report_lock);
 853 }
 854
 855 static void mca_get(struct ifmcaddr6 *mc)
 856 {
 857         refcount_inc(&mc->mca_refcnt);
 858 }
 859
 860 static void ma_put(struct ifmcaddr6 *mc)
 861 {
 862         if (refcount_dec_and_test(&mc->mca_refcnt)) {
 863                 in6_dev_put(mc->idev);
 864                 kfree_rcu(mc, rcu);
 865         }
 866 }
 867
 868 /* called with mc_lock */
 869 static struct ifmcaddr6 *mca_alloc(struct inet6_dev *idev,
 870                                    const struct in6_addr *addr,
 871                                    unsigned int mode)
 872 {
 873         struct ifmcaddr6 *mc;
 874
 875         mc = kzalloc(sizeof(*mc), GFP_KERNEL);
 876         if (!mc)
 877                 return NULL;
 878
 879         INIT_DELAYED_WORK(&mc->mca_work, mld_mca_work);
 880
 881         mc->mca_addr = *addr;
 882         mc->idev = idev; /* reference taken by caller */
 883         mc->mca_users = 1;
 884         /* mca_stamp should be updated upon changes */
 885         mc->mca_cstamp = mc->mca_tstamp = jiffies;
 886         refcount_set(&mc->mca_refcnt, 1);
 887
 888         mc->mca_sfmode = mode;
 889         mc->mca_sfcount[mode] = 1;
 890
 891         if (ipv6_addr_is_ll_all_nodes(&mc->mca_addr) ||
 892             IPV6_ADDR_MC_SCOPE(&mc->mca_addr) < IPV6_ADDR_SCOPE_LINKLOCAL)
 893                 mc->mca_flags |= MAF_NOREPORT;
 894
 895         return mc;
 896 }
 897
 898 /*
 899  *      device multicast group inc (add if not found)
 900  */
 901 static int __ipv6_dev_mc_inc(struct net_device *dev,
 902                              const struct in6_addr *addr, unsigned int mode)
 903 {
 904         struct ifmcaddr6 *mc;
 905         struct inet6_dev *idev;
 906
 907         ASSERT_RTNL();
 908
 909         /* we need to take a reference on idev */
 910         idev = in6_dev_get(dev);
 911
 912         if (!idev)
 913                 return -EINVAL;
 914
 915         if (idev->dead) {
 916                 in6_dev_put(idev);
 917                 return -ENODEV;
 918         }
 919
 920         mutex_lock(&idev->mc_lock);
 921         for_each_mc_mclock(idev, mc) {
 922                 if (ipv6_addr_equal(&mc->mca_addr, addr)) {
 923                         mc->mca_users++;
 924                         ip6_mc_add_src(idev, &mc->mca_addr, mode, 0, NULL, 0);
 925                         mutex_unlock(&idev->mc_lock);
 926                         in6_dev_put(idev);
 927                         return 0;
 928                 }
 929         }
 930
 931         mc = mca_alloc(idev, addr, mode);
 932         if (!mc) {
 933                 mutex_unlock(&idev->mc_lock);
 934                 in6_dev_put(idev);
 935                 return -ENOMEM;
 936         }
 937
 938         rcu_assign_pointer(mc->next, idev->mc_list);
 939         rcu_assign_pointer(idev->mc_list, mc);
 940
 941         mca_get(mc);
 942
 943         mld_del_delrec(idev, mc);
 944         igmp6_group_added(mc);
 945         mutex_unlock(&idev->mc_lock);
 946         ma_put(mc);
 947         return 0;
 948 }
 949
 950 int ipv6_dev_mc_inc(struct net_device *dev, const struct in6_addr *addr)
 951 {
 952         return __ipv6_dev_mc_inc(dev, addr, MCAST_EXCLUDE);
 953 }
 954 EXPORT_SYMBOL(ipv6_dev_mc_inc);
 955
 956 /*
 957  * device multicast group del
 958  */
 959 int __ipv6_dev_mc_dec(struct inet6_dev *idev, const struct in6_addr *addr)
 960 {
 961         struct ifmcaddr6 *ma, __rcu **map;
 962
 963         ASSERT_RTNL();
 964
 965         mutex_lock(&idev->mc_lock);
 966         for (map = &idev->mc_list;
 967              (ma = mc_dereference(*map, idev));
 968              map = &ma->next) {
 969                 if (ipv6_addr_equal(&ma->mca_addr, addr)) {
 970                         if (--ma->mca_users == 0) {
 971                                 *map = ma->next;
 972
 973                                 igmp6_group_dropped(ma);
 974                                 ip6_mc_clear_src(ma);
 975                                 mutex_unlock(&idev->mc_lock);
 976
 977                                 ma_put(ma);
 978                                 return 0;
 979                         }
 980                         mutex_unlock(&idev->mc_lock);
 981                         return 0;
 982                 }
 983         }
 984
 985         mutex_unlock(&idev->mc_lock);
 986         return -ENOENT;
 987 }
 988
 989 int ipv6_dev_mc_dec(struct net_device *dev, const struct in6_addr *addr)
 990 {
 991         struct inet6_dev *idev;
 992         int err;
 993
 994         ASSERT_RTNL();
 995
 996         idev = __in6_dev_get(dev);
 997         if (!idev)
 998                 err = -ENODEV;
 999         else
1000                 err = __ipv6_dev_mc_dec(idev, addr);
1001
1002         return err;
1003 }
1004 EXPORT_SYMBOL(ipv6_dev_mc_dec);
1005
1006 /*
1007  *      check if the interface/address pair is valid
1008  */
1009 bool ipv6_chk_mcast_addr(struct net_device *dev, const struct in6_addr *group,
1010                          const struct in6_addr *src_addr)
1011 {
1012         struct inet6_dev *idev;
1013         struct ifmcaddr6 *mc;
1014         bool rv = false;
1015
1016         rcu_read_lock();
1017         idev = __in6_dev_get(dev);
1018         if (idev) {
1019                 for_each_mc_rcu(idev, mc) {
1020                         if (ipv6_addr_equal(&mc->mca_addr, group))
1021                                 break;
1022                 }
1023                 if (mc) {
1024                         if (src_addr && !ipv6_addr_any(src_addr)) {
1025                                 struct ip6_sf_list *psf;
1026
1027                                 for_each_psf_rcu(mc, psf) {
1028                                         if (ipv6_addr_equal(&psf->sf_addr, src_addr))
1029                                                 break;
1030                                 }
1031                                 if (psf)
1032                                         rv = psf->sf_count[MCAST_INCLUDE] ||
1033                                                 psf->sf_count[MCAST_EXCLUDE] !=
1034                                                 mc->mca_sfcount[MCAST_EXCLUDE];
1035                                 else
1036                                         rv = mc->mca_sfcount[MCAST_EXCLUDE] != 0;
1037                         } else
1038                                 rv = true; /* don't filter unspecified source */
1039                 }
1040         }
1041         rcu_read_unlock();
1042         return rv;
1043 }
1044
1045 /* called with mc_lock */
1046 static void mld_gq_start_work(struct inet6_dev *idev)
1047 {
1048         unsigned long tv = prandom_u32() % idev->mc_maxdelay;
1049
1050         idev->mc_gq_running = 1;
1051         if (!mod_delayed_work(mld_wq, &idev->mc_gq_work, tv + 2))
1052                 in6_dev_hold(idev);
1053 }
1054
1055 /* called with mc_lock */
1056 static void mld_gq_stop_work(struct inet6_dev *idev)
1057 {
1058         idev->mc_gq_running = 0;
1059         if (cancel_delayed_work(&idev->mc_gq_work))
1060                 __in6_dev_put(idev);
1061 }
1062
1063 /* called with mc_lock */
1064 static void mld_ifc_start_work(struct inet6_dev *idev, unsigned long delay)
1065 {
1066         unsigned long tv = prandom_u32() % delay;
1067
1068         if (!mod_delayed_work(mld_wq, &idev->mc_ifc_work, tv + 2))
1069                 in6_dev_hold(idev);
1070 }
1071
1072 /* called with mc_lock */
1073 static void mld_ifc_stop_work(struct inet6_dev *idev)
1074 {
1075         idev->mc_ifc_count = 0;
1076         if (cancel_delayed_work(&idev->mc_ifc_work))
1077                 __in6_dev_put(idev);
1078 }
1079
1080 /* called with mc_lock */
1081 static void mld_dad_start_work(struct inet6_dev *idev, unsigned long delay)
1082 {
1083         unsigned long tv = prandom_u32() % delay;
1084
1085         if (!mod_delayed_work(mld_wq, &idev->mc_dad_work, tv + 2))
1086                 in6_dev_hold(idev);
1087 }
1088
1089 static void mld_dad_stop_work(struct inet6_dev *idev)
1090 {
1091         if (cancel_delayed_work(&idev->mc_dad_work))
1092                 __in6_dev_put(idev);
1093 }
1094
1095 static void mld_query_stop_work(struct inet6_dev *idev)
1096 {
1097         spin_lock_bh(&idev->mc_query_lock);
1098         if (cancel_delayed_work(&idev->mc_query_work))
1099                 __in6_dev_put(idev);
1100         spin_unlock_bh(&idev->mc_query_lock);
1101 }
1102
1103 static void mld_report_stop_work(struct inet6_dev *idev)
1104 {
1105         if (cancel_delayed_work_sync(&idev->mc_report_work))
1106                 __in6_dev_put(idev);
1107 }
1108
1109 /*
1110  * IGMP handling (alias multicast ICMPv6 messages)
1111  * called with mc_lock
1112  */
1113 static void igmp6_group_queried(struct ifmcaddr6 *ma, unsigned long resptime)
1114 {
1115         unsigned long delay = resptime;
1116
1117         /* Do not start work for these addresses */
1118         if (ipv6_addr_is_ll_all_nodes(&ma->mca_addr) ||
1119             IPV6_ADDR_MC_SCOPE(&ma->mca_addr) < IPV6_ADDR_SCOPE_LINKLOCAL)
1120                 return;
1121
1122         if (cancel_delayed_work(&ma->mca_work)) {
1123                 refcount_dec(&ma->mca_refcnt);
1124                 delay = ma->mca_work.timer.expires - jiffies;
1125         }
1126
1127         if (delay >= resptime)
1128                 delay = prandom_u32() % resptime;
1129
1130         if (!mod_delayed_work(mld_wq, &ma->mca_work, delay))
1131                 refcount_inc(&ma->mca_refcnt);
1132         ma->mca_flags |= MAF_TIMER_RUNNING;
1133 }
1134
1135 /* mark EXCLUDE-mode sources
1136  * called with mc_lock
1137  */
1138 static bool mld_xmarksources(struct ifmcaddr6 *pmc, int nsrcs,
1139                              const struct in6_addr *srcs)
1140 {
1141         struct ip6_sf_list *psf;
1142         int i, scount;
1143
1144         scount = 0;
1145         for_each_psf_mclock(pmc, psf) {
1146                 if (scount == nsrcs)
1147                         break;
1148                 for (i = 0; i < nsrcs; i++) {
1149                         /* skip inactive filters */
1150                         if (psf->sf_count[MCAST_INCLUDE] ||
1151                             pmc->mca_sfcount[MCAST_EXCLUDE] !=
1152                             psf->sf_count[MCAST_EXCLUDE])
1153                                 break;
1154                         if (ipv6_addr_equal(&srcs[i], &psf->sf_addr)) {
1155                                 scount++;
1156                                 break;
1157                         }
1158                 }
1159         }
1160         pmc->mca_flags &= ~MAF_GSQUERY;
1161         if (scount == nsrcs)    /* all sources excluded */
1162                 return false;
1163         return true;
1164 }
1165
1166 /* called with mc_lock */
1167 static bool mld_marksources(struct ifmcaddr6 *pmc, int nsrcs,
1168                             const struct in6_addr *srcs)
1169 {
1170         struct ip6_sf_list *psf;
1171         int i, scount;
1172
1173         if (pmc->mca_sfmode == MCAST_EXCLUDE)
1174                 return mld_xmarksources(pmc, nsrcs, srcs);
1175
1176         /* mark INCLUDE-mode sources */
1177
1178         scount = 0;
1179         for_each_psf_mclock(pmc, psf) {
1180                 if (scount == nsrcs)
1181                         break;
1182                 for (i = 0; i < nsrcs; i++) {
1183                         if (ipv6_addr_equal(&srcs[i], &psf->sf_addr)) {
1184                                 psf->sf_gsresp = 1;
1185                                 scount++;
1186                                 break;
1187                         }
1188                 }
1189         }
1190         if (!scount) {
1191                 pmc->mca_flags &= ~MAF_GSQUERY;
1192                 return false;
1193         }
1194         pmc->mca_flags |= MAF_GSQUERY;
1195         return true;
1196 }
1197
1198 static int mld_force_mld_version(const struct inet6_dev *idev)
1199 {
1200         /* Normally, both are 0 here. If enforcement to a particular is
1201          * being used, individual device enforcement will have a lower
1202          * precedence over 'all' device (.../conf/all/force_mld_version).
1203          */
1204
1205         if (dev_net(idev->dev)->ipv6.devconf_all->force_mld_version != 0)
1206                 return dev_net(idev->dev)->ipv6.devconf_all->force_mld_version;
1207         else
1208                 return idev->cnf.force_mld_version;
1209 }
1210
1211 static bool mld_in_v2_mode_only(const struct inet6_dev *idev)
1212 {
1213         return mld_force_mld_version(idev) == 2;
1214 }
1215
1216 static bool mld_in_v1_mode_only(const struct inet6_dev *idev)
1217 {
1218         return mld_force_mld_version(idev) == 1;
1219 }
1220
1221 static bool mld_in_v1_mode(const struct inet6_dev *idev)
1222 {
1223         if (mld_in_v2_mode_only(idev))
1224                 return false;
1225         if (mld_in_v1_mode_only(idev))
1226                 return true;
1227         if (idev->mc_v1_seen && time_before(jiffies, idev->mc_v1_seen))
1228                 return true;
1229
1230         return false;
1231 }
1232
1233 static void mld_set_v1_mode(struct inet6_dev *idev)
1234 {
1235         /* RFC3810, relevant sections:
1236          *  - 9.1. Robustness Variable
1237          *  - 9.2. Query Interval
1238          *  - 9.3. Query Response Interval
1239          *  - 9.12. Older Version Querier Present Timeout
1240          */
1241         unsigned long switchback;
1242
1243         switchback = (idev->mc_qrv * idev->mc_qi) + idev->mc_qri;
1244
1245         idev->mc_v1_seen = jiffies + switchback;
1246 }
1247
1248 static void mld_update_qrv(struct inet6_dev *idev,
1249                            const struct mld2_query *mlh2)
1250 {
1251         /* RFC3810, relevant sections:
1252          *  - 5.1.8. QRV (Querier's Robustness Variable)
1253          *  - 9.1. Robustness Variable
1254          */
1255
1256         /* The value of the Robustness Variable MUST NOT be zero,
1257          * and SHOULD NOT be one. Catch this here if we ever run
1258          * into such a case in future.
1259          */
1260         const int min_qrv = min(MLD_QRV_DEFAULT, sysctl_mld_qrv);
1261         WARN_ON(idev->mc_qrv == 0);
1262
1263         if (mlh2->mld2q_qrv > 0)
1264                 idev->mc_qrv = mlh2->mld2q_qrv;
1265
1266         if (unlikely(idev->mc_qrv < min_qrv)) {
1267                 net_warn_ratelimited("IPv6: MLD: clamping QRV from %u to %u!\n",
1268                                      idev->mc_qrv, min_qrv);
1269                 idev->mc_qrv = min_qrv;
1270         }
1271 }
1272
1273 static void mld_update_qi(struct inet6_dev *idev,
1274                           const struct mld2_query *mlh2)
1275 {
1276         /* RFC3810, relevant sections:
1277          *  - 5.1.9. QQIC (Querier's Query Interval Code)
1278          *  - 9.2. Query Interval
1279          *  - 9.12. Older Version Querier Present Timeout
1280          *    (the [Query Interval] in the last Query received)
1281          */
1282         unsigned long mc_qqi;
1283
1284         if (mlh2->mld2q_qqic < 128) {
1285                 mc_qqi = mlh2->mld2q_qqic;
1286         } else {
1287                 unsigned long mc_man, mc_exp;
1288
1289                 mc_exp = MLDV2_QQIC_EXP(mlh2->mld2q_qqic);
1290                 mc_man = MLDV2_QQIC_MAN(mlh2->mld2q_qqic);
1291
1292                 mc_qqi = (mc_man | 0x10) << (mc_exp + 3);
1293         }
1294
1295         idev->mc_qi = mc_qqi * HZ;
1296 }
1297
1298 static void mld_update_qri(struct inet6_dev *idev,
1299                            const struct mld2_query *mlh2)
1300 {
1301         /* RFC3810, relevant sections:
1302          *  - 5.1.3. Maximum Response Code
1303          *  - 9.3. Query Response Interval
1304          */
1305         idev->mc_qri = msecs_to_jiffies(mldv2_mrc(mlh2));
1306 }
1307
1308 static int mld_process_v1(struct inet6_dev *idev, struct mld_msg *mld,
1309                           unsigned long *max_delay, bool v1_query)
1310 {
1311         unsigned long mldv1_md;
1312
1313         /* Ignore v1 queries */
1314         if (mld_in_v2_mode_only(idev))
1315                 return -EINVAL;
1316
1317         mldv1_md = ntohs(mld->mld_maxdelay);
1318
1319         /* When in MLDv1 fallback and a MLDv2 router start-up being
1320          * unaware of current MLDv1 operation, the MRC == MRD mapping
1321          * only works when the exponential algorithm is not being
1322          * used (as MLDv1 is unaware of such things).
1323          *
1324          * According to the RFC author, the MLDv2 implementations
1325          * he's aware of all use a MRC < 32768 on start up queries.
1326          *
1327          * Thus, should we *ever* encounter something else larger
1328          * than that, just assume the maximum possible within our
1329          * reach.
1330          */
1331         if (!v1_query)
1332                 mldv1_md = min(mldv1_md, MLDV1_MRD_MAX_COMPAT);
1333
1334         *max_delay = max(msecs_to_jiffies(mldv1_md), 1UL);
1335
1336         /* MLDv1 router present: we need to go into v1 mode *only*
1337          * when an MLDv1 query is received as per section 9.12. of
1338          * RFC3810! And we know from RFC2710 section 3.7 that MLDv1
1339          * queries MUST be of exactly 24 octets.
1340          */
1341         if (v1_query)
1342                 mld_set_v1_mode(idev);
1343
1344         /* cancel MLDv2 report work */
1345         mld_gq_stop_work(idev);
1346         /* cancel the interface change work */
1347         mld_ifc_stop_work(idev);
1348         /* clear deleted report items */
1349         mld_clear_delrec(idev);
1350
1351         return 0;
1352 }
1353
1354 static int mld_process_v2(struct inet6_dev *idev, struct mld2_query *mld,
1355                           unsigned long *max_delay)
1356 {
1357         *max_delay = max(msecs_to_jiffies(mldv2_mrc(mld)), 1UL);
1358
1359         mld_update_qrv(idev, mld);
1360         mld_update_qi(idev, mld);
1361         mld_update_qri(idev, mld);
1362
1363         idev->mc_maxdelay = *max_delay;
1364
1365         return 0;
1366 }
1367
1368 /* called with rcu_read_lock() */
1369 int igmp6_event_query(struct sk_buff *skb)
1370 {
1371         struct inet6_dev *idev = __in6_dev_get(skb->dev);
1372
1373         if (!idev)
1374                 return -EINVAL;
1375
1376         if (idev->dead) {
1377                 kfree_skb(skb);
1378                 return -ENODEV;
1379         }
1380
1381         spin_lock_bh(&idev->mc_query_lock);
1382         if (skb_queue_len(&idev->mc_query_queue) < MLD_MAX_SKBS) {
1383                 __skb_queue_tail(&idev->mc_query_queue, skb);
1384                 if (!mod_delayed_work(mld_wq, &idev->mc_query_work, 0))
1385                         in6_dev_hold(idev);
1386         }
1387         spin_unlock_bh(&idev->mc_query_lock);
1388
1389         return 0;
1390 }
1391
1392 static void __mld_query_work(struct sk_buff *skb)
1393 {
1394         struct mld2_query *mlh2 = NULL;
1395         const struct in6_addr *group;
1396         unsigned long max_delay;
1397         struct inet6_dev *idev;
1398         struct ifmcaddr6 *ma;
1399         struct mld_msg *mld;
1400         int group_type;
1401         int mark = 0;
1402         int len, err;
1403
1404         if (!pskb_may_pull(skb, sizeof(struct in6_addr)))
1405                 goto kfree_skb;
1406
1407         /* compute payload length excluding extension headers */
1408         len = ntohs(ipv6_hdr(skb)->payload_len) + sizeof(struct ipv6hdr);
1409         len -= skb_network_header_len(skb);
1410
1411         /* RFC3810 6.2
1412          * Upon reception of an MLD message that contains a Query, the node
1413          * checks if the source address of the message is a valid link-local
1414          * address, if the Hop Limit is set to 1, and if the Router Alert
1415          * option is present in the Hop-By-Hop Options header of the IPv6
1416          * packet.  If any of these checks fails, the packet is dropped.
1417          */
1418         if (!(ipv6_addr_type(&ipv6_hdr(skb)->saddr) & IPV6_ADDR_LINKLOCAL) ||
1419             ipv6_hdr(skb)->hop_limit != 1 ||
1420             !(IP6CB(skb)->flags & IP6SKB_ROUTERALERT) ||
1421             IP6CB(skb)->ra != htons(IPV6_OPT_ROUTERALERT_MLD))
1422                 goto kfree_skb;
1423
1424         idev = in6_dev_get(skb->dev);
1425         if (!idev)
1426                 goto kfree_skb;
1427
1428         mld = (struct mld_msg *)icmp6_hdr(skb);
1429         group = &mld->mld_mca;
1430         group_type = ipv6_addr_type(group);
1431
1432         if (group_type != IPV6_ADDR_ANY &&
1433             !(group_type&IPV6_ADDR_MULTICAST))
1434                 goto out;
1435
1436         if (len < MLD_V1_QUERY_LEN) {
1437                 goto out;
1438         } else if (len == MLD_V1_QUERY_LEN || mld_in_v1_mode(idev)) {
1439                 err = mld_process_v1(idev, mld, &max_delay,
1440                                      len == MLD_V1_QUERY_LEN);
1441                 if (err < 0)
1442                         goto out;
1443         } else if (len >= MLD_V2_QUERY_LEN_MIN) {
1444                 int srcs_offset = sizeof(struct mld2_query) -
1445                                   sizeof(struct icmp6hdr);
1446
1447                 if (!pskb_may_pull(skb, srcs_offset))
1448                         goto out;
1449
1450                 mlh2 = (struct mld2_query *)skb_transport_header(skb);
1451
1452                 err = mld_process_v2(idev, mlh2, &max_delay);
1453                 if (err < 0)
1454                         goto out;
1455
1456                 if (group_type == IPV6_ADDR_ANY) { /* general query */
1457                         if (mlh2->mld2q_nsrcs)
1458                                 goto out; /* no sources allowed */
1459
1460                         mld_gq_start_work(idev);
1461                         goto out;
1462                 }
1463                 /* mark sources to include, if group & source-specific */
1464                 if (mlh2->mld2q_nsrcs != 0) {
1465                         if (!pskb_may_pull(skb, srcs_offset +
1466                             ntohs(mlh2->mld2q_nsrcs) * sizeof(struct in6_addr)))
1467                                 goto out;
1468
1469                         mlh2 = (struct mld2_query *)skb_transport_header(skb);
1470                         mark = 1;
1471                 }
1472         } else {
1473                 goto out;
1474         }
1475
1476         if (group_type == IPV6_ADDR_ANY) {
1477                 for_each_mc_mclock(idev, ma) {
1478                         igmp6_group_queried(ma, max_delay);
1479                 }
1480         } else {
1481                 for_each_mc_mclock(idev, ma) {
1482                         if (!ipv6_addr_equal(group, &ma->mca_addr))
1483                                 continue;
1484                         if (ma->mca_flags & MAF_TIMER_RUNNING) {
1485                                 /* gsquery <- gsquery && mark */
1486                                 if (!mark)
1487                                         ma->mca_flags &= ~MAF_GSQUERY;
1488                         } else {
1489                                 /* gsquery <- mark */
1490                                 if (mark)
1491                                         ma->mca_flags |= MAF_GSQUERY;
1492                                 else
1493                                         ma->mca_flags &= ~MAF_GSQUERY;
1494                         }
1495                         if (!(ma->mca_flags & MAF_GSQUERY) ||
1496                             mld_marksources(ma, ntohs(mlh2->mld2q_nsrcs), mlh2->mld2q_srcs))
1497                                 igmp6_group_queried(ma, max_delay);
1498                         break;
1499                 }
1500         }
1501
1502 out:
1503         in6_dev_put(idev);
1504 kfree_skb:
1505         consume_skb(skb);
1506 }
1507
1508 static void mld_query_work(struct work_struct *work)
1509 {
1510         struct inet6_dev *idev = container_of(to_delayed_work(work),
1511                                               struct inet6_dev,
1512                                               mc_query_work);
1513         struct sk_buff_head q;
1514         struct sk_buff *skb;
1515         bool rework = false;
1516         int cnt = 0;
1517
1518         skb_queue_head_init(&q);
1519
1520         spin_lock_bh(&idev->mc_query_lock);
1521         while ((skb = __skb_dequeue(&idev->mc_query_queue))) {
1522                 __skb_queue_tail(&q, skb);
1523
1524                 if (++cnt >= MLD_MAX_QUEUE) {
1525                         rework = true;
1526                         schedule_delayed_work(&idev->mc_query_work, 0);
1527                         break;
1528                 }
1529         }
1530         spin_unlock_bh(&idev->mc_query_lock);
1531
1532         mutex_lock(&idev->mc_lock);
1533         while ((skb = __skb_dequeue(&q)))
1534                 __mld_query_work(skb);
1535         mutex_unlock(&idev->mc_lock);
1536
1537         if (!rework)
1538                 in6_dev_put(idev);
1539 }
1540
1541 /* called with rcu_read_lock() */
1542 int igmp6_event_report(struct sk_buff *skb)
1543 {
1544         struct inet6_dev *idev = __in6_dev_get(skb->dev);
1545
1546         if (!idev)
1547                 return -EINVAL;
1548
1549         if (idev->dead) {
1550                 kfree_skb(skb);
1551                 return -ENODEV;
1552         }
1553
1554         spin_lock_bh(&idev->mc_report_lock);
1555         if (skb_queue_len(&idev->mc_report_queue) < MLD_MAX_SKBS) {
1556                 __skb_queue_tail(&idev->mc_report_queue, skb);
1557                 if (!mod_delayed_work(mld_wq, &idev->mc_report_work, 0))
1558                         in6_dev_hold(idev);
1559         }
1560         spin_unlock_bh(&idev->mc_report_lock);
1561
1562         return 0;
1563 }
1564
1565 static void __mld_report_work(struct sk_buff *skb)
1566 {
1567         struct inet6_dev *idev;
1568         struct ifmcaddr6 *ma;
1569         struct mld_msg *mld;
1570         int addr_type;
1571
1572         /* Our own report looped back. Ignore it. */
1573         if (skb->pkt_type == PACKET_LOOPBACK)
1574                 goto kfree_skb;
1575
1576         /* send our report if the MC router may not have heard this report */
1577         if (skb->pkt_type != PACKET_MULTICAST &&
1578             skb->pkt_type != PACKET_BROADCAST)
1579                 goto kfree_skb;
1580
1581         if (!pskb_may_pull(skb, sizeof(*mld) - sizeof(struct icmp6hdr)))
1582                 goto kfree_skb;
1583
1584         mld = (struct mld_msg *)icmp6_hdr(skb);
1585
1586         /* Drop reports with not link local source */
1587         addr_type = ipv6_addr_type(&ipv6_hdr(skb)->saddr);
1588         if (addr_type != IPV6_ADDR_ANY &&
1589             !(addr_type&IPV6_ADDR_LINKLOCAL))
1590                 goto kfree_skb;
1591
1592         idev = in6_dev_get(skb->dev);
1593         if (!idev)
1594                 goto kfree_skb;
1595
1596         /*
1597          *      Cancel the work for this group
1598          */
1599
1600         for_each_mc_mclock(idev, ma) {
1601                 if (ipv6_addr_equal(&ma->mca_addr, &mld->mld_mca)) {
1602                         if (cancel_delayed_work(&ma->mca_work))
1603                                 refcount_dec(&ma->mca_refcnt);
1604                         ma->mca_flags &= ~(MAF_LAST_REPORTER |
1605                                            MAF_TIMER_RUNNING);
1606                         break;
1607                 }
1608         }
1609
1610         in6_dev_put(idev);
1611 kfree_skb:
1612         consume_skb(skb);
1613 }
1614
1615 static void mld_report_work(struct work_struct *work)
1616 {
1617         struct inet6_dev *idev = container_of(to_delayed_work(work),
1618                                               struct inet6_dev,
1619                                               mc_report_work);
1620         struct sk_buff_head q;
1621         struct sk_buff *skb;
1622         bool rework = false;
1623         int cnt = 0;
1624
1625         skb_queue_head_init(&q);
1626         spin_lock_bh(&idev->mc_report_lock);
1627         while ((skb = __skb_dequeue(&idev->mc_report_queue))) {
1628                 __skb_queue_tail(&q, skb);
1629
1630                 if (++cnt >= MLD_MAX_QUEUE) {
1631                         rework = true;
1632                         schedule_delayed_work(&idev->mc_report_work, 0);
1633                         break;
1634                 }
1635         }
1636         spin_unlock_bh(&idev->mc_report_lock);
1637
1638         mutex_lock(&idev->mc_lock);
1639         while ((skb = __skb_dequeue(&q)))
1640                 __mld_report_work(skb);
1641         mutex_unlock(&idev->mc_lock);
1642
1643         if (!rework)
1644                 in6_dev_put(idev);
1645 }
1646
1647 static bool is_in(struct ifmcaddr6 *pmc, struct ip6_sf_list *psf, int type,
1648                   int gdeleted, int sdeleted)
1649 {
1650         switch (type) {
1651         case MLD2_MODE_IS_INCLUDE:
1652         case MLD2_MODE_IS_EXCLUDE:
1653                 if (gdeleted || sdeleted)
1654                         return false;
1655                 if (!((pmc->mca_flags & MAF_GSQUERY) && !psf->sf_gsresp)) {
1656                         if (pmc->mca_sfmode == MCAST_INCLUDE)
1657                                 return true;
1658                         /* don't include if this source is excluded
1659                          * in all filters
1660                          */
1661                         if (psf->sf_count[MCAST_INCLUDE])
1662                                 return type == MLD2_MODE_IS_INCLUDE;
1663                         return pmc->mca_sfcount[MCAST_EXCLUDE] ==
1664                                 psf->sf_count[MCAST_EXCLUDE];
1665                 }
1666                 return false;
1667         case MLD2_CHANGE_TO_INCLUDE:
1668                 if (gdeleted || sdeleted)
1669                         return false;
1670                 return psf->sf_count[MCAST_INCLUDE] != 0;
1671         case MLD2_CHANGE_TO_EXCLUDE:
1672                 if (gdeleted || sdeleted)
1673                         return false;
1674                 if (pmc->mca_sfcount[MCAST_EXCLUDE] == 0 ||
1675                     psf->sf_count[MCAST_INCLUDE])
1676                         return false;
1677                 return pmc->mca_sfcount[MCAST_EXCLUDE] ==
1678                         psf->sf_count[MCAST_EXCLUDE];
1679         case MLD2_ALLOW_NEW_SOURCES:
1680                 if (gdeleted || !psf->sf_crcount)
1681                         return false;
1682                 return (pmc->mca_sfmode == MCAST_INCLUDE) ^ sdeleted;
1683         case MLD2_BLOCK_OLD_SOURCES:
1684                 if (pmc->mca_sfmode == MCAST_INCLUDE)
1685                         return gdeleted || (psf->sf_crcount && sdeleted);
1686                 return psf->sf_crcount && !gdeleted && !sdeleted;
1687         }
1688         return false;
1689 }
1690
1691 static int
1692 mld_scount(struct ifmcaddr6 *pmc, int type, int gdeleted, int sdeleted)
1693 {
1694         struct ip6_sf_list *psf;
1695         int scount = 0;
1696
1697         for_each_psf_mclock(pmc, psf) {
1698                 if (!is_in(pmc, psf, type, gdeleted, sdeleted))
1699                         continue;
1700                 scount++;
1701         }
1702         return scount;
1703 }
1704
1705 static void ip6_mc_hdr(struct sock *sk, struct sk_buff *skb,
1706                        struct net_device *dev,
1707                        const struct in6_addr *saddr,
1708                        const struct in6_addr *daddr,
1709                        int proto, int len)
1710 {
1711         struct ipv6hdr *hdr;
1712
1713         skb->protocol = htons(ETH_P_IPV6);
1714         skb->dev = dev;
1715
1716         skb_reset_network_header(skb);
1717         skb_put(skb, sizeof(struct ipv6hdr));
1718         hdr = ipv6_hdr(skb);
1719
1720         ip6_flow_hdr(hdr, 0, 0);
1721
1722         hdr->payload_len = htons(len);
1723         hdr->nexthdr = proto;
1724         hdr->hop_limit = inet6_sk(sk)->hop_limit;
1725
1726         hdr->saddr = *saddr;
1727         hdr->daddr = *daddr;
1728 }
1729
1730 static struct sk_buff *mld_newpack(struct inet6_dev *idev, unsigned int mtu)
1731 {
1732         u8 ra[8] = { IPPROTO_ICMPV6, 0, IPV6_TLV_ROUTERALERT,
1733                      2, 0, 0, IPV6_TLV_PADN, 0 };
1734         struct net_device *dev = idev->dev;
1735         int hlen = LL_RESERVED_SPACE(dev);
1736         int tlen = dev->needed_tailroom;
1737         struct net *net = dev_net(dev);
1738         const struct in6_addr *saddr;
1739         struct in6_addr addr_buf;
1740         struct mld2_report *pmr;
1741         struct sk_buff *skb;
1742         unsigned int size;
1743         struct sock *sk;
1744         int err;
1745
1746         sk = net->ipv6.igmp_sk;
1747         /* we assume size > sizeof(ra) here
1748          * Also try to not allocate high-order pages for big MTU
1749          */
1750         size = min_t(int, mtu, PAGE_SIZE / 2) + hlen + tlen;
1751         skb = sock_alloc_send_skb(sk, size, 1, &err);
1752         if (!skb)
1753                 return NULL;
1754
1755         skb->priority = TC_PRIO_CONTROL;
1756         skb_reserve(skb, hlen);
1757         skb_tailroom_reserve(skb, mtu, tlen);
1758
1759         if (__ipv6_get_lladdr(idev, &addr_buf, IFA_F_TENTATIVE)) {
1760                 /* <draft-ietf-magma-mld-source-05.txt>:
1761                  * use unspecified address as the source address
1762                  * when a valid link-local address is not available.
1763                  */
1764                 saddr = &in6addr_any;
1765         } else
1766                 saddr = &addr_buf;
1767
1768         ip6_mc_hdr(sk, skb, dev, saddr, &mld2_all_mcr, NEXTHDR_HOP, 0);
1769
1770         skb_put_data(skb, ra, sizeof(ra));
1771
1772         skb_set_transport_header(skb, skb_tail_pointer(skb) - skb->data);
1773         skb_put(skb, sizeof(*pmr));
1774         pmr = (struct mld2_report *)skb_transport_header(skb);
1775         pmr->mld2r_type = ICMPV6_MLD2_REPORT;
1776         pmr->mld2r_resv1 = 0;
1777         pmr->mld2r_cksum = 0;
1778         pmr->mld2r_resv2 = 0;
1779         pmr->mld2r_ngrec = 0;
1780         return skb;
1781 }
1782
1783 static void mld_sendpack(struct sk_buff *skb)
1784 {
1785         struct ipv6hdr *pip6 = ipv6_hdr(skb);
1786         struct mld2_report *pmr =
1787                               (struct mld2_report *)skb_transport_header(skb);
1788         int payload_len, mldlen;
1789         struct inet6_dev *idev;
1790         struct net *net = dev_net(skb->dev);
1791         int err;
1792         struct flowi6 fl6;
1793         struct dst_entry *dst;
1794
1795         rcu_read_lock();
1796         idev = __in6_dev_get(skb->dev);
1797         IP6_UPD_PO_STATS(net, idev, IPSTATS_MIB_OUT, skb->len);
1798
1799         payload_len = (skb_tail_pointer(skb) - skb_network_header(skb)) -
1800                 sizeof(*pip6);
1801         mldlen = skb_tail_pointer(skb) - skb_transport_header(skb);
1802         pip6->payload_len = htons(payload_len);
1803
1804         pmr->mld2r_cksum = csum_ipv6_magic(&pip6->saddr, &pip6->daddr, mldlen,
1805                                            IPPROTO_ICMPV6,
1806                                            csum_partial(skb_transport_header(skb),
1807                                                         mldlen, 0));
1808
1809         icmpv6_flow_init(net->ipv6.igmp_sk, &fl6, ICMPV6_MLD2_REPORT,
1810                          &ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr,
1811                          skb->dev->ifindex);
1812         dst = icmp6_dst_alloc(skb->dev, &fl6);
1813
1814         err = 0;
1815         if (IS_ERR(dst)) {
1816                 err = PTR_ERR(dst);
1817                 dst = NULL;
1818         }
1819         skb_dst_set(skb, dst);
1820         if (err)
1821                 goto err_out;
1822
1823         err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT,
1824                       net, net->ipv6.igmp_sk, skb, NULL, skb->dev,
1825                       dst_output);
1826 out:
1827         if (!err) {
1828                 ICMP6MSGOUT_INC_STATS(net, idev, ICMPV6_MLD2_REPORT);
1829                 ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTMSGS);
1830         } else {
1831                 IP6_INC_STATS(net, idev, IPSTATS_MIB_OUTDISCARDS);
1832         }
1833
1834         rcu_read_unlock();
1835         return;
1836
1837 err_out:
1838         kfree_skb(skb);
1839         goto out;
1840 }
1841
1842 static int grec_size(struct ifmcaddr6 *pmc, int type, int gdel, int sdel)
1843 {
1844         return sizeof(struct mld2_grec) + 16 * mld_scount(pmc,type,gdel,sdel);
1845 }
1846
1847 static struct sk_buff *add_grhead(struct sk_buff *skb, struct ifmcaddr6 *pmc,
1848         int type, struct mld2_grec **ppgr, unsigned int mtu)
1849 {
1850         struct mld2_report *pmr;
1851         struct mld2_grec *pgr;
1852
1853         if (!skb) {
1854                 skb = mld_newpack(pmc->idev, mtu);
1855                 if (!skb)
1856                         return NULL;
1857         }
1858         pgr = skb_put(skb, sizeof(struct mld2_grec));
1859         pgr->grec_type = type;
1860         pgr->grec_auxwords = 0;
1861         pgr->grec_nsrcs = 0;
1862         pgr->grec_mca = pmc->mca_addr;  /* structure copy */
1863         pmr = (struct mld2_report *)skb_transport_header(skb);
1864         pmr->mld2r_ngrec = htons(ntohs(pmr->mld2r_ngrec)+1);
1865         *ppgr = pgr;
1866         return skb;
1867 }
1868
1869 #define AVAILABLE(skb)  ((skb) ? skb_availroom(skb) : 0)
1870
1871 /* called with mc_lock */
1872 static struct sk_buff *add_grec(struct sk_buff *skb, struct ifmcaddr6 *pmc,
1873                                 int type, int gdeleted, int sdeleted,
1874                                 int crsend)
1875 {
1876         struct ip6_sf_list *psf, *psf_prev, *psf_next;
1877         int scount, stotal, first, isquery, truncate;
1878         struct ip6_sf_list __rcu **psf_list;
1879         struct inet6_dev *idev = pmc->idev;
1880         struct net_device *dev = idev->dev;
1881         struct mld2_grec *pgr = NULL;
1882         struct mld2_report *pmr;
1883         unsigned int mtu;
1884
1885         if (pmc->mca_flags & MAF_NOREPORT)
1886                 return skb;
1887
1888         mtu = READ_ONCE(dev->mtu);
1889         if (mtu < IPV6_MIN_MTU)
1890                 return skb;
1891
1892         isquery = type == MLD2_MODE_IS_INCLUDE ||
1893                   type == MLD2_MODE_IS_EXCLUDE;
1894         truncate = type == MLD2_MODE_IS_EXCLUDE ||
1895                     type == MLD2_CHANGE_TO_EXCLUDE;
1896
1897         stotal = scount = 0;
1898
1899         psf_list = sdeleted ? &pmc->mca_tomb : &pmc->mca_sources;
1900
1901         if (!rcu_access_pointer(*psf_list))
1902                 goto empty_source;
1903
1904         pmr = skb ? (struct mld2_report *)skb_transport_header(skb) : NULL;
1905
1906         /* EX and TO_EX get a fresh packet, if needed */
1907         if (truncate) {
1908                 if (pmr && pmr->mld2r_ngrec &&
1909                     AVAILABLE(skb) < grec_size(pmc, type, gdeleted, sdeleted)) {
1910                         if (skb)
1911                                 mld_sendpack(skb);
1912                         skb = mld_newpack(idev, mtu);
1913                 }
1914         }
1915         first = 1;
1916         psf_prev = NULL;
1917         for (psf = mc_dereference(*psf_list, idev);
1918              psf;
1919              psf = psf_next) {
1920                 struct in6_addr *psrc;
1921
1922                 psf_next = mc_dereference(psf->sf_next, idev);
1923
1924                 if (!is_in(pmc, psf, type, gdeleted, sdeleted) && !crsend) {
1925                         psf_prev = psf;
1926                         continue;
1927                 }
1928
1929                 /* Based on RFC3810 6.1. Should not send source-list change
1930                  * records when there is a filter mode change.
1931                  */
1932                 if (((gdeleted && pmc->mca_sfmode == MCAST_EXCLUDE) ||
1933                      (!gdeleted && pmc->mca_crcount)) &&
1934                     (type == MLD2_ALLOW_NEW_SOURCES ||
1935                      type == MLD2_BLOCK_OLD_SOURCES) && psf->sf_crcount)
1936                         goto decrease_sf_crcount;
1937
1938                 /* clear marks on query responses */
1939                 if (isquery)
1940                         psf->sf_gsresp = 0;
1941
1942                 if (AVAILABLE(skb) < sizeof(*psrc) +
1943                     first*sizeof(struct mld2_grec)) {
1944                         if (truncate && !first)
1945                                 break;   /* truncate these */
1946                         if (pgr)
1947                                 pgr->grec_nsrcs = htons(scount);
1948                         if (skb)
1949                                 mld_sendpack(skb);
1950                         skb = mld_newpack(idev, mtu);
1951                         first = 1;
1952                         scount = 0;
1953                 }
1954                 if (first) {
1955                         skb = add_grhead(skb, pmc, type, &pgr, mtu);
1956                         first = 0;
1957                 }
1958                 if (!skb)
1959                         return NULL;
1960                 psrc = skb_put(skb, sizeof(*psrc));
1961                 *psrc = psf->sf_addr;
1962                 scount++; stotal++;
1963                 if ((type == MLD2_ALLOW_NEW_SOURCES ||
1964                      type == MLD2_BLOCK_OLD_SOURCES) && psf->sf_crcount) {
1965 decrease_sf_crcount:
1966                         psf->sf_crcount--;
1967                         if ((sdeleted || gdeleted) && psf->sf_crcount == 0) {
1968                                 if (psf_prev)
1969                                         rcu_assign_pointer(psf_prev->sf_next,
1970                                                            mc_dereference(psf->sf_next, idev));
1971                                 else
1972                                         rcu_assign_pointer(*psf_list,
1973                                                            mc_dereference(psf->sf_next, idev));
1974                                 kfree_rcu(psf, rcu);
1975                                 continue;
1976                         }
1977                 }
1978                 psf_prev = psf;
1979         }
1980
1981 empty_source:
1982         if (!stotal) {
1983                 if (type == MLD2_ALLOW_NEW_SOURCES ||
1984                     type == MLD2_BLOCK_OLD_SOURCES)
1985                         return skb;
1986                 if (pmc->mca_crcount || isquery || crsend) {
1987                         /* make sure we have room for group header */
1988                         if (skb && AVAILABLE(skb) < sizeof(struct mld2_grec)) {
1989                                 mld_sendpack(skb);
1990                                 skb = NULL; /* add_grhead will get a new one */
1991                         }
1992                         skb = add_grhead(skb, pmc, type, &pgr, mtu);
1993                 }
1994         }
1995         if (pgr)
1996                 pgr->grec_nsrcs = htons(scount);
1997
1998         if (isquery)
1999                 pmc->mca_flags &= ~MAF_GSQUERY; /* clear query state */
2000         return skb;
2001 }
2002
2003 /* called with mc_lock */
2004 static void mld_send_report(struct inet6_dev *idev, struct ifmcaddr6 *pmc)
2005 {
2006         struct sk_buff *skb = NULL;
2007         int type;
2008
2009         if (!pmc) {
2010                 for_each_mc_mclock(idev, pmc) {
2011                         if (pmc->mca_flags & MAF_NOREPORT)
2012                                 continue;
2013                         if (pmc->mca_sfcount[MCAST_EXCLUDE])
2014                                 type = MLD2_MODE_IS_EXCLUDE;
2015                         else
2016                                 type = MLD2_MODE_IS_INCLUDE;
2017                         skb = add_grec(skb, pmc, type, 0, 0, 0);
2018                 }
2019         } else {
2020                 if (pmc->mca_sfcount[MCAST_EXCLUDE])
2021                         type = MLD2_MODE_IS_EXCLUDE;
2022                 else
2023                         type = MLD2_MODE_IS_INCLUDE;
2024                 skb = add_grec(skb, pmc, type, 0, 0, 0);
2025         }
2026         if (skb)
2027                 mld_sendpack(skb);
2028 }
2029
2030 /*
2031  * remove zero-count source records from a source filter list
2032  * called with mc_lock
2033  */
2034 static void mld_clear_zeros(struct ip6_sf_list __rcu **ppsf, struct inet6_dev *idev)
2035 {
2036         struct ip6_sf_list *psf_prev, *psf_next, *psf;
2037
2038         psf_prev = NULL;
2039         for (psf = mc_dereference(*ppsf, idev);
2040              psf;
2041              psf = psf_next) {
2042                 psf_next = mc_dereference(psf->sf_next, idev);
2043                 if (psf->sf_crcount == 0) {
2044                         if (psf_prev)
2045                                 rcu_assign_pointer(psf_prev->sf_next,
2046                                                    mc_dereference(psf->sf_next, idev));
2047                         else
2048                                 rcu_assign_pointer(*ppsf,
2049                                                    mc_dereference(psf->sf_next, idev));
2050                         kfree_rcu(psf, rcu);
2051                 } else {
2052                         psf_prev = psf;
2053                 }
2054         }
2055 }
2056
2057 /* called with mc_lock */
2058 static void mld_send_cr(struct inet6_dev *idev)
2059 {
2060         struct ifmcaddr6 *pmc, *pmc_prev, *pmc_next;
2061         struct sk_buff *skb = NULL;
2062         int type, dtype;
2063
2064         /* deleted MCA's */
2065         pmc_prev = NULL;
2066         for (pmc = mc_dereference(idev->mc_tomb, idev);
2067              pmc;
2068              pmc = pmc_next) {
2069                 pmc_next = mc_dereference(pmc->next, idev);
2070                 if (pmc->mca_sfmode == MCAST_INCLUDE) {
2071                         type = MLD2_BLOCK_OLD_SOURCES;
2072                         dtype = MLD2_BLOCK_OLD_SOURCES;
2073                         skb = add_grec(skb, pmc, type, 1, 0, 0);
2074                         skb = add_grec(skb, pmc, dtype, 1, 1, 0);
2075                 }
2076                 if (pmc->mca_crcount) {
2077                         if (pmc->mca_sfmode == MCAST_EXCLUDE) {
2078                                 type = MLD2_CHANGE_TO_INCLUDE;
2079                                 skb = add_grec(skb, pmc, type, 1, 0, 0);
2080                         }
2081                         pmc->mca_crcount--;
2082                         if (pmc->mca_crcount == 0) {
2083                                 mld_clear_zeros(&pmc->mca_tomb, idev);
2084                                 mld_clear_zeros(&pmc->mca_sources, idev);
2085                         }
2086                 }
2087                 if (pmc->mca_crcount == 0 &&
2088                     !rcu_access_pointer(pmc->mca_tomb) &&
2089                     !rcu_access_pointer(pmc->mca_sources)) {
2090                         if (pmc_prev)
2091                                 rcu_assign_pointer(pmc_prev->next, pmc_next);
2092                         else
2093                                 rcu_assign_pointer(idev->mc_tomb, pmc_next);
2094                         in6_dev_put(pmc->idev);
2095                         kfree_rcu(pmc, rcu);
2096                 } else
2097                         pmc_prev = pmc;
2098         }
2099
2100         /* change recs */
2101         for_each_mc_mclock(idev, pmc) {
2102                 if (pmc->mca_sfcount[MCAST_EXCLUDE]) {
2103                         type = MLD2_BLOCK_OLD_SOURCES;
2104                         dtype = MLD2_ALLOW_NEW_SOURCES;
2105                 } else {
2106                         type = MLD2_ALLOW_NEW_SOURCES;
2107                         dtype = MLD2_BLOCK_OLD_SOURCES;
2108                 }
2109                 skb = add_grec(skb, pmc, type, 0, 0, 0);
2110                 skb = add_grec(skb, pmc, dtype, 0, 1, 0);       /* deleted sources */
2111
2112                 /* filter mode changes */
2113                 if (pmc->mca_crcount) {
2114                         if (pmc->mca_sfmode == MCAST_EXCLUDE)
2115                                 type = MLD2_CHANGE_TO_EXCLUDE;
2116                         else
2117                                 type = MLD2_CHANGE_TO_INCLUDE;
2118                         skb = add_grec(skb, pmc, type, 0, 0, 0);
2119                         pmc->mca_crcount--;
2120                 }
2121         }
2122         if (!skb)
2123                 return;
2124         (void) mld_sendpack(skb);
2125 }
2126
2127 static void igmp6_send(struct in6_addr *addr, struct net_device *dev, int type)
2128 {
2129         struct net *net = dev_net(dev);
2130         struct sock *sk = net->ipv6.igmp_sk;
2131         struct inet6_dev *idev;
2132         struct sk_buff *skb;
2133         struct mld_msg *hdr;
2134         const struct in6_addr *snd_addr, *saddr;
2135         struct in6_addr addr_buf;
2136         int hlen = LL_RESERVED_SPACE(dev);
2137         int tlen = dev->needed_tailroom;
2138         int err, len, payload_len, full_len;
2139         u8 ra[8] = { IPPROTO_ICMPV6, 0,
2140                      IPV6_TLV_ROUTERALERT, 2, 0, 0,
2141                      IPV6_TLV_PADN, 0 };
2142         struct flowi6 fl6;
2143         struct dst_entry *dst;
2144
2145         if (type == ICMPV6_MGM_REDUCTION)
2146                 snd_addr = &in6addr_linklocal_allrouters;
2147         else
2148                 snd_addr = addr;
2149
2150         len = sizeof(struct icmp6hdr) + sizeof(struct in6_addr);
2151         payload_len = len + sizeof(ra);
2152         full_len = sizeof(struct ipv6hdr) + payload_len;
2153
2154         rcu_read_lock();
2155         IP6_UPD_PO_STATS(net, __in6_dev_get(dev),
2156                       IPSTATS_MIB_OUT, full_len);
2157         rcu_read_unlock();
2158
2159         skb = sock_alloc_send_skb(sk, hlen + tlen + full_len, 1, &err);
2160
2161         if (!skb) {
2162                 rcu_read_lock();
2163                 IP6_INC_STATS(net, __in6_dev_get(dev),
2164                               IPSTATS_MIB_OUTDISCARDS);
2165                 rcu_read_unlock();
2166                 return;
2167         }
2168         skb->priority = TC_PRIO_CONTROL;
2169         skb_reserve(skb, hlen);
2170
2171         if (ipv6_get_lladdr(dev, &addr_buf, IFA_F_TENTATIVE)) {
2172                 /* <draft-ietf-magma-mld-source-05.txt>:
2173                  * use unspecified address as the source address
2174                  * when a valid link-local address is not available.
2175                  */
2176                 saddr = &in6addr_any;
2177         } else
2178                 saddr = &addr_buf;
2179
2180         ip6_mc_hdr(sk, skb, dev, saddr, snd_addr, NEXTHDR_HOP, payload_len);
2181
2182         skb_put_data(skb, ra, sizeof(ra));
2183
2184         hdr = skb_put_zero(skb, sizeof(struct mld_msg));
2185         hdr->mld_type = type;
2186         hdr->mld_mca = *addr;
2187
2188         hdr->mld_cksum = csum_ipv6_magic(saddr, snd_addr, len,
2189                                          IPPROTO_ICMPV6,
2190                                          csum_partial(hdr, len, 0));
2191
2192         rcu_read_lock();
2193         idev = __in6_dev_get(skb->dev);
2194
2195         icmpv6_flow_init(sk, &fl6, type,
2196                          &ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr,
2197                          skb->dev->ifindex);
2198         dst = icmp6_dst_alloc(skb->dev, &fl6);
2199         if (IS_ERR(dst)) {
2200                 err = PTR_ERR(dst);
2201                 goto err_out;
2202         }
2203
2204         skb_dst_set(skb, dst);
2205         err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT,
2206                       net, sk, skb, NULL, skb->dev,
2207                       dst_output);
2208 out:
2209         if (!err) {
2210                 ICMP6MSGOUT_INC_STATS(net, idev, type);
2211                 ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTMSGS);
2212         } else
2213                 IP6_INC_STATS(net, idev, IPSTATS_MIB_OUTDISCARDS);
2214
2215         rcu_read_unlock();
2216         return;
2217
2218 err_out:
2219         kfree_skb(skb);
2220         goto out;
2221 }
2222
2223 /* called with mc_lock */
2224 static void mld_send_initial_cr(struct inet6_dev *idev)
2225 {
2226         struct sk_buff *skb;
2227         struct ifmcaddr6 *pmc;
2228         int type;
2229
2230         if (mld_in_v1_mode(idev))
2231                 return;
2232
2233         skb = NULL;
2234         for_each_mc_mclock(idev, pmc) {
2235                 if (pmc->mca_sfcount[MCAST_EXCLUDE])
2236                         type = MLD2_CHANGE_TO_EXCLUDE;
2237                 else
2238                         type = MLD2_ALLOW_NEW_SOURCES;
2239                 skb = add_grec(skb, pmc, type, 0, 0, 1);
2240         }
2241         if (skb)
2242                 mld_sendpack(skb);
2243 }
2244
2245 void ipv6_mc_dad_complete(struct inet6_dev *idev)
2246 {
2247         mutex_lock(&idev->mc_lock);
2248         idev->mc_dad_count = idev->mc_qrv;
2249         if (idev->mc_dad_count) {
2250                 mld_send_initial_cr(idev);
2251                 idev->mc_dad_count--;
2252                 if (idev->mc_dad_count)
2253                         mld_dad_start_work(idev,
2254                                            unsolicited_report_interval(idev));
2255         }
2256         mutex_unlock(&idev->mc_lock);
2257 }
2258
2259 static void mld_dad_work(struct work_struct *work)
2260 {
2261         struct inet6_dev *idev = container_of(to_delayed_work(work),
2262                                               struct inet6_dev,
2263                                               mc_dad_work);
2264         mutex_lock(&idev->mc_lock);
2265         mld_send_initial_cr(idev);
2266         if (idev->mc_dad_count) {
2267                 idev->mc_dad_count--;
2268                 if (idev->mc_dad_count)
2269                         mld_dad_start_work(idev,
2270                                            unsolicited_report_interval(idev));
2271         }
2272         mutex_unlock(&idev->mc_lock);
2273         in6_dev_put(idev);
2274 }
2275
2276 /* called with mc_lock */
2277 static int ip6_mc_del1_src(struct ifmcaddr6 *pmc, int sfmode,
2278         const struct in6_addr *psfsrc)
2279 {
2280         struct ip6_sf_list *psf, *psf_prev;
2281         int rv = 0;
2282
2283         psf_prev = NULL;
2284         for_each_psf_mclock(pmc, psf) {
2285                 if (ipv6_addr_equal(&psf->sf_addr, psfsrc))
2286                         break;
2287                 psf_prev = psf;
2288         }
2289         if (!psf || psf->sf_count[sfmode] == 0) {
2290                 /* source filter not found, or count wrong =>  bug */
2291                 return -ESRCH;
2292         }
2293         psf->sf_count[sfmode]--;
2294         if (!psf->sf_count[MCAST_INCLUDE] && !psf->sf_count[MCAST_EXCLUDE]) {
2295                 struct inet6_dev *idev = pmc->idev;
2296
2297                 /* no more filters for this source */
2298                 if (psf_prev)
2299                         rcu_assign_pointer(psf_prev->sf_next,
2300                                            mc_dereference(psf->sf_next, idev));
2301                 else
2302                         rcu_assign_pointer(pmc->mca_sources,
2303                                            mc_dereference(psf->sf_next, idev));
2304
2305                 if (psf->sf_oldin && !(pmc->mca_flags & MAF_NOREPORT) &&
2306                     !mld_in_v1_mode(idev)) {
2307                         psf->sf_crcount = idev->mc_qrv;
2308                         rcu_assign_pointer(psf->sf_next,
2309                                            mc_dereference(pmc->mca_tomb, idev));
2310                         rcu_assign_pointer(pmc->mca_tomb, psf);
2311                         rv = 1;
2312                 } else {
2313                         kfree_rcu(psf, rcu);
2314                 }
2315         }
2316         return rv;
2317 }
2318
2319 /* called with mc_lock */
2320 static int ip6_mc_del_src(struct inet6_dev *idev, const struct in6_addr *pmca,
2321                           int sfmode, int sfcount, const struct in6_addr *psfsrc,
2322                           int delta)
2323 {
2324         struct ifmcaddr6 *pmc;
2325         int     changerec = 0;
2326         int     i, err;
2327
2328         if (!idev)
2329                 return -ENODEV;
2330
2331         for_each_mc_mclock(idev, pmc) {
2332                 if (ipv6_addr_equal(pmca, &pmc->mca_addr))
2333                         break;
2334         }
2335         if (!pmc)
2336                 return -ESRCH;
2337
2338         sf_markstate(pmc);
2339         if (!delta) {
2340                 if (!pmc->mca_sfcount[sfmode])
2341                         return -EINVAL;
2342
2343                 pmc->mca_sfcount[sfmode]--;
2344         }
2345         err = 0;
2346         for (i = 0; i < sfcount; i++) {
2347                 int rv = ip6_mc_del1_src(pmc, sfmode, &psfsrc[i]);
2348
2349                 changerec |= rv > 0;
2350                 if (!err && rv < 0)
2351                         err = rv;
2352         }
2353         if (pmc->mca_sfmode == MCAST_EXCLUDE &&
2354             pmc->mca_sfcount[MCAST_EXCLUDE] == 0 &&
2355             pmc->mca_sfcount[MCAST_INCLUDE]) {
2356                 struct ip6_sf_list *psf;
2357
2358                 /* filter mode change */
2359                 pmc->mca_sfmode = MCAST_INCLUDE;
2360                 pmc->mca_crcount = idev->mc_qrv;
2361                 idev->mc_ifc_count = pmc->mca_crcount;
2362                 for_each_psf_mclock(pmc, psf)
2363                         psf->sf_crcount = 0;
2364                 mld_ifc_event(pmc->idev);
2365         } else if (sf_setstate(pmc) || changerec) {
2366                 mld_ifc_event(pmc->idev);
2367         }
2368
2369         return err;
2370 }
2371
2372 /*
2373  * Add multicast single-source filter to the interface list
2374  * called with mc_lock
2375  */
2376 static int ip6_mc_add1_src(struct ifmcaddr6 *pmc, int sfmode,
2377         const struct in6_addr *psfsrc)
2378 {
2379         struct ip6_sf_list *psf, *psf_prev;
2380
2381         psf_prev = NULL;
2382         for_each_psf_mclock(pmc, psf) {
2383                 if (ipv6_addr_equal(&psf->sf_addr, psfsrc))
2384                         break;
2385                 psf_prev = psf;
2386         }
2387         if (!psf) {
2388                 psf = kzalloc(sizeof(*psf), GFP_KERNEL);
2389                 if (!psf)
2390                         return -ENOBUFS;
2391
2392                 psf->sf_addr = *psfsrc;
2393                 if (psf_prev) {
2394                         rcu_assign_pointer(psf_prev->sf_next, psf);
2395                 } else {
2396                         rcu_assign_pointer(pmc->mca_sources, psf);
2397                 }
2398         }
2399         psf->sf_count[sfmode]++;
2400         return 0;
2401 }
2402
2403 /* called with mc_lock */
2404 static void sf_markstate(struct ifmcaddr6 *pmc)
2405 {
2406         struct ip6_sf_list *psf;
2407         int mca_xcount = pmc->mca_sfcount[MCAST_EXCLUDE];
2408
2409         for_each_psf_mclock(pmc, psf) {
2410                 if (pmc->mca_sfcount[MCAST_EXCLUDE]) {
2411                         psf->sf_oldin = mca_xcount ==
2412                                 psf->sf_count[MCAST_EXCLUDE] &&
2413                                 !psf->sf_count[MCAST_INCLUDE];
2414                 } else {
2415                         psf->sf_oldin = psf->sf_count[MCAST_INCLUDE] != 0;
2416                 }
2417         }
2418 }
2419
2420 /* called with mc_lock */
2421 static int sf_setstate(struct ifmcaddr6 *pmc)
2422 {
2423         struct ip6_sf_list *psf, *dpsf;
2424         int mca_xcount = pmc->mca_sfcount[MCAST_EXCLUDE];
2425         int qrv = pmc->idev->mc_qrv;
2426         int new_in, rv;
2427
2428         rv = 0;
2429         for_each_psf_mclock(pmc, psf) {
2430                 if (pmc->mca_sfcount[MCAST_EXCLUDE]) {
2431                         new_in = mca_xcount == psf->sf_count[MCAST_EXCLUDE] &&
2432                                 !psf->sf_count[MCAST_INCLUDE];
2433                 } else
2434                         new_in = psf->sf_count[MCAST_INCLUDE] != 0;
2435                 if (new_in) {
2436                         if (!psf->sf_oldin) {
2437                                 struct ip6_sf_list *prev = NULL;
2438
2439                                 for_each_psf_tomb(pmc, dpsf) {
2440                                         if (ipv6_addr_equal(&dpsf->sf_addr,
2441                                             &psf->sf_addr))
2442                                                 break;
2443                                         prev = dpsf;
2444                                 }
2445                                 if (dpsf) {
2446                                         if (prev)
2447                                                 rcu_assign_pointer(prev->sf_next,
2448                                                                    mc_dereference(dpsf->sf_next,
2449                                                                                   pmc->idev));
2450                                         else
2451                                                 rcu_assign_pointer(pmc->mca_tomb,
2452                                                                    mc_dereference(dpsf->sf_next,
2453                                                                                   pmc->idev));
2454                                         kfree_rcu(dpsf, rcu);
2455                                 }
2456                                 psf->sf_crcount = qrv;
2457                                 rv++;
2458                         }
2459                 } else if (psf->sf_oldin) {
2460                         psf->sf_crcount = 0;
2461                         /*
2462                          * add or update "delete" records if an active filter
2463                          * is now inactive
2464                          */
2465
2466                         for_each_psf_tomb(pmc, dpsf)
2467                                 if (ipv6_addr_equal(&dpsf->sf_addr,
2468                                     &psf->sf_addr))
2469                                         break;
2470                         if (!dpsf) {
2471                                 dpsf = kmalloc(sizeof(*dpsf), GFP_KERNEL);
2472                                 if (!dpsf)
2473                                         continue;
2474                                 *dpsf = *psf;
2475                                 rcu_assign_pointer(dpsf->sf_next,
2476                                                    mc_dereference(pmc->mca_tomb, pmc->idev));
2477                                 rcu_assign_pointer(pmc->mca_tomb, dpsf);
2478                         }
2479                         dpsf->sf_crcount = qrv;
2480                         rv++;
2481                 }
2482         }
2483         return rv;
2484 }
2485
2486 /*
2487  * Add multicast source filter list to the interface list
2488  * called with mc_lock
2489  */
2490 static int ip6_mc_add_src(struct inet6_dev *idev, const struct in6_addr *pmca,
2491                           int sfmode, int sfcount, const struct in6_addr *psfsrc,
2492                           int delta)
2493 {
2494         struct ifmcaddr6 *pmc;
2495         int     isexclude;
2496         int     i, err;
2497
2498         if (!idev)
2499                 return -ENODEV;
2500
2501         for_each_mc_mclock(idev, pmc) {
2502                 if (ipv6_addr_equal(pmca, &pmc->mca_addr))
2503                         break;
2504         }
2505         if (!pmc)
2506                 return -ESRCH;
2507
2508         sf_markstate(pmc);
2509         isexclude = pmc->mca_sfmode == MCAST_EXCLUDE;
2510         if (!delta)
2511                 pmc->mca_sfcount[sfmode]++;
2512         err = 0;
2513         for (i = 0; i < sfcount; i++) {
2514                 err = ip6_mc_add1_src(pmc, sfmode, &psfsrc[i]);
2515                 if (err)
2516                         break;
2517         }
2518         if (err) {
2519                 int j;
2520
2521                 if (!delta)
2522                         pmc->mca_sfcount[sfmode]--;
2523                 for (j = 0; j < i; j++)
2524                         ip6_mc_del1_src(pmc, sfmode, &psfsrc[j]);
2525         } else if (isexclude != (pmc->mca_sfcount[MCAST_EXCLUDE] != 0)) {
2526                 struct ip6_sf_list *psf;
2527
2528                 /* filter mode change */
2529                 if (pmc->mca_sfcount[MCAST_EXCLUDE])
2530                         pmc->mca_sfmode = MCAST_EXCLUDE;
2531                 else if (pmc->mca_sfcount[MCAST_INCLUDE])
2532                         pmc->mca_sfmode = MCAST_INCLUDE;
2533                 /* else no filters; keep old mode for reports */
2534
2535                 pmc->mca_crcount = idev->mc_qrv;
2536                 idev->mc_ifc_count = pmc->mca_crcount;
2537                 for_each_psf_mclock(pmc, psf)
2538                         psf->sf_crcount = 0;
2539                 mld_ifc_event(idev);
2540         } else if (sf_setstate(pmc)) {
2541                 mld_ifc_event(idev);
2542         }
2543         return err;
2544 }
2545
2546 /* called with mc_lock */
2547 static void ip6_mc_clear_src(struct ifmcaddr6 *pmc)
2548 {
2549         struct ip6_sf_list *psf, *nextpsf;
2550
2551         for (psf = mc_dereference(pmc->mca_tomb, pmc->idev);
2552              psf;
2553              psf = nextpsf) {
2554                 nextpsf = mc_dereference(psf->sf_next, pmc->idev);
2555                 kfree_rcu(psf, rcu);
2556         }
2557         RCU_INIT_POINTER(pmc->mca_tomb, NULL);
2558         for (psf = mc_dereference(pmc->mca_sources, pmc->idev);
2559              psf;
2560              psf = nextpsf) {
2561                 nextpsf = mc_dereference(psf->sf_next, pmc->idev);
2562                 kfree_rcu(psf, rcu);
2563         }
2564         RCU_INIT_POINTER(pmc->mca_sources, NULL);
2565         pmc->mca_sfmode = MCAST_EXCLUDE;
2566         pmc->mca_sfcount[MCAST_INCLUDE] = 0;
2567         pmc->mca_sfcount[MCAST_EXCLUDE] = 1;
2568 }
2569
2570 /* called with mc_lock */
2571 static void igmp6_join_group(struct ifmcaddr6 *ma)
2572 {
2573         unsigned long delay;
2574
2575         if (ma->mca_flags & MAF_NOREPORT)
2576                 return;
2577
2578         igmp6_send(&ma->mca_addr, ma->idev->dev, ICMPV6_MGM_REPORT);
2579
2580         delay = prandom_u32() % unsolicited_report_interval(ma->idev);
2581
2582         if (cancel_delayed_work(&ma->mca_work)) {
2583                 refcount_dec(&ma->mca_refcnt);
2584                 delay = ma->mca_work.timer.expires - jiffies;
2585         }
2586
2587         if (!mod_delayed_work(mld_wq, &ma->mca_work, delay))
2588                 refcount_inc(&ma->mca_refcnt);
2589         ma->mca_flags |= MAF_TIMER_RUNNING | MAF_LAST_REPORTER;
2590 }
2591
2592 static int ip6_mc_leave_src(struct sock *sk, struct ipv6_mc_socklist *iml,
2593                             struct inet6_dev *idev)
2594 {
2595         struct ip6_sf_socklist *psl;
2596         int err;
2597
2598         psl = sock_dereference(iml->sflist, sk);
2599
2600         if (idev)
2601                 mutex_lock(&idev->mc_lock);
2602
2603         if (!psl) {
2604                 /* any-source empty exclude case */
2605                 err = ip6_mc_del_src(idev, &iml->addr, iml->sfmode, 0, NULL, 0);
2606         } else {
2607                 err = ip6_mc_del_src(idev, &iml->addr, iml->sfmode,
2608                                      psl->sl_count, psl->sl_addr, 0);
2609                 RCU_INIT_POINTER(iml->sflist, NULL);
2610                 atomic_sub(IP6_SFLSIZE(psl->sl_max), &sk->sk_omem_alloc);
2611                 kfree_rcu(psl, rcu);
2612         }
2613
2614         if (idev)
2615                 mutex_unlock(&idev->mc_lock);
2616
2617         return err;
2618 }
2619
2620 /* called with mc_lock */
2621 static void igmp6_leave_group(struct ifmcaddr6 *ma)
2622 {
2623         if (mld_in_v1_mode(ma->idev)) {
2624                 if (ma->mca_flags & MAF_LAST_REPORTER) {
2625                         igmp6_send(&ma->mca_addr, ma->idev->dev,
2626                                 ICMPV6_MGM_REDUCTION);
2627                 }
2628         } else {
2629                 mld_add_delrec(ma->idev, ma);
2630                 mld_ifc_event(ma->idev);
2631         }
2632 }
2633
2634 static void mld_gq_work(struct work_struct *work)
2635 {
2636         struct inet6_dev *idev = container_of(to_delayed_work(work),
2637                                               struct inet6_dev,
2638                                               mc_gq_work);
2639
2640         mutex_lock(&idev->mc_lock);
2641         mld_send_report(idev, NULL);
2642         idev->mc_gq_running = 0;
2643         mutex_unlock(&idev->mc_lock);
2644
2645         in6_dev_put(idev);
2646 }
2647
2648 static void mld_ifc_work(struct work_struct *work)
2649 {
2650         struct inet6_dev *idev = container_of(to_delayed_work(work),
2651                                               struct inet6_dev,
2652                                               mc_ifc_work);
2653
2654         mutex_lock(&idev->mc_lock);
2655         mld_send_cr(idev);
2656
2657         if (idev->mc_ifc_count) {
2658                 idev->mc_ifc_count--;
2659                 if (idev->mc_ifc_count)
2660                         mld_ifc_start_work(idev,
2661                                            unsolicited_report_interval(idev));
2662         }
2663         mutex_unlock(&idev->mc_lock);
2664         in6_dev_put(idev);
2665 }
2666
2667 /* called with mc_lock */
2668 static void mld_ifc_event(struct inet6_dev *idev)
2669 {
2670         if (mld_in_v1_mode(idev))
2671                 return;
2672
2673         idev->mc_ifc_count = idev->mc_qrv;
2674         mld_ifc_start_work(idev, 1);
2675 }
2676
2677 static void mld_mca_work(struct work_struct *work)
2678 {
2679         struct ifmcaddr6 *ma = container_of(to_delayed_work(work),
2680                                             struct ifmcaddr6, mca_work);
2681
2682         mutex_lock(&ma->idev->mc_lock);
2683         if (mld_in_v1_mode(ma->idev))
2684                 igmp6_send(&ma->mca_addr, ma->idev->dev, ICMPV6_MGM_REPORT);
2685         else
2686                 mld_send_report(ma->idev, ma);
2687         ma->mca_flags |=  MAF_LAST_REPORTER;
2688         ma->mca_flags &= ~MAF_TIMER_RUNNING;
2689         mutex_unlock(&ma->idev->mc_lock);
2690
2691         ma_put(ma);
2692 }
2693
2694 /* Device changing type */
2695
2696 void ipv6_mc_unmap(struct inet6_dev *idev)
2697 {
2698         struct ifmcaddr6 *i;
2699
2700         /* Install multicast list, except for all-nodes (already installed) */
2701
2702         mutex_lock(&idev->mc_lock);
2703         for_each_mc_mclock(idev, i)
2704                 igmp6_group_dropped(i);
2705         mutex_unlock(&idev->mc_lock);
2706 }
2707
2708 void ipv6_mc_remap(struct inet6_dev *idev)
2709 {
2710         ipv6_mc_up(idev);
2711 }
2712
2713 /* Device going down */
2714 void ipv6_mc_down(struct inet6_dev *idev)
2715 {
2716         struct ifmcaddr6 *i;
2717
2718         mutex_lock(&idev->mc_lock);
2719         /* Withdraw multicast list */
2720         for_each_mc_mclock(idev, i)
2721                 igmp6_group_dropped(i);
2722         mutex_unlock(&idev->mc_lock);
2723
2724         /* Should stop work after group drop. or we will
2725          * start work again in mld_ifc_event()
2726          */
2727         synchronize_net();
2728         mld_query_stop_work(idev);
2729         mld_report_stop_work(idev);
2730         mld_ifc_stop_work(idev);
2731         mld_gq_stop_work(idev);
2732         mld_dad_stop_work(idev);
2733 }
2734
2735 static void ipv6_mc_reset(struct inet6_dev *idev)
2736 {
2737         idev->mc_qrv = sysctl_mld_qrv;
2738         idev->mc_qi = MLD_QI_DEFAULT;
2739         idev->mc_qri = MLD_QRI_DEFAULT;
2740         idev->mc_v1_seen = 0;
2741         idev->mc_maxdelay = unsolicited_report_interval(idev);
2742 }
2743
2744 /* Device going up */
2745
2746 void ipv6_mc_up(struct inet6_dev *idev)
2747 {
2748         struct ifmcaddr6 *i;
2749
2750         /* Install multicast list, except for all-nodes (already installed) */
2751
2752         ipv6_mc_reset(idev);
2753         mutex_lock(&idev->mc_lock);
2754         for_each_mc_mclock(idev, i) {
2755                 mld_del_delrec(idev, i);
2756                 igmp6_group_added(i);
2757         }
2758         mutex_unlock(&idev->mc_lock);
2759 }
2760
2761 /* IPv6 device initialization. */
2762
2763 void ipv6_mc_init_dev(struct inet6_dev *idev)
2764 {
2765         idev->mc_gq_running = 0;
2766         INIT_DELAYED_WORK(&idev->mc_gq_work, mld_gq_work);
2767         RCU_INIT_POINTER(idev->mc_tomb, NULL);
2768         idev->mc_ifc_count = 0;
2769         INIT_DELAYED_WORK(&idev->mc_ifc_work, mld_ifc_work);
2770         INIT_DELAYED_WORK(&idev->mc_dad_work, mld_dad_work);
2771         INIT_DELAYED_WORK(&idev->mc_query_work, mld_query_work);
2772         INIT_DELAYED_WORK(&idev->mc_report_work, mld_report_work);
2773         skb_queue_head_init(&idev->mc_query_queue);
2774         skb_queue_head_init(&idev->mc_report_queue);
2775         spin_lock_init(&idev->mc_query_lock);
2776         spin_lock_init(&idev->mc_report_lock);
2777         mutex_init(&idev->mc_lock);
2778         ipv6_mc_reset(idev);
2779 }
2780
2781 /*
2782  *      Device is about to be destroyed: clean up.
2783  */
2784
2785 void ipv6_mc_destroy_dev(struct inet6_dev *idev)
2786 {
2787         struct ifmcaddr6 *i;
2788
2789         /* Deactivate works */
2790         ipv6_mc_down(idev);
2791         mutex_lock(&idev->mc_lock);
2792         mld_clear_delrec(idev);
2793         mutex_unlock(&idev->mc_lock);
2794         mld_clear_query(idev);
2795         mld_clear_report(idev);
2796
2797         /* Delete all-nodes address. */
2798         /* We cannot call ipv6_dev_mc_dec() directly, our caller in
2799          * addrconf.c has NULL'd out dev->ip6_ptr so in6_dev_get() will
2800          * fail.
2801          */
2802         __ipv6_dev_mc_dec(idev, &in6addr_linklocal_allnodes);
2803
2804         if (idev->cnf.forwarding)
2805                 __ipv6_dev_mc_dec(idev, &in6addr_linklocal_allrouters);
2806
2807         mutex_lock(&idev->mc_lock);
2808         while ((i = mc_dereference(idev->mc_list, idev))) {
2809                 rcu_assign_pointer(idev->mc_list, mc_dereference(i->next, idev));
2810
2811                 ip6_mc_clear_src(i);
2812                 ma_put(i);
2813         }
2814         mutex_unlock(&idev->mc_lock);
2815 }
2816
2817 static void ipv6_mc_rejoin_groups(struct inet6_dev *idev)
2818 {
2819         struct ifmcaddr6 *pmc;
2820
2821         ASSERT_RTNL();
2822
2823         mutex_lock(&idev->mc_lock);
2824         if (mld_in_v1_mode(idev)) {
2825                 for_each_mc_mclock(idev, pmc)
2826                         igmp6_join_group(pmc);
2827         } else {
2828                 mld_send_report(idev, NULL);
2829         }
2830         mutex_unlock(&idev->mc_lock);
2831 }
2832
2833 static int ipv6_mc_netdev_event(struct notifier_block *this,
2834                                 unsigned long event,
2835                                 void *ptr)
2836 {
2837         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2838         struct inet6_dev *idev = __in6_dev_get(dev);
2839
2840         switch (event) {
2841         case NETDEV_RESEND_IGMP:
2842                 if (idev)
2843                         ipv6_mc_rejoin_groups(idev);
2844                 break;
2845         default:
2846                 break;
2847         }
2848
2849         return NOTIFY_DONE;
2850 }
2851
2852 static struct notifier_block igmp6_netdev_notifier = {
2853         .notifier_call = ipv6_mc_netdev_event,
2854 };
2855
2856 #ifdef CONFIG_PROC_FS
2857 struct igmp6_mc_iter_state {
2858         struct seq_net_private p;
2859         struct net_device *dev;
2860         struct inet6_dev *idev;
2861 };
2862
2863 #define igmp6_mc_seq_private(seq)       ((struct igmp6_mc_iter_state *)(seq)->private)
2864
2865 static inline struct ifmcaddr6 *igmp6_mc_get_first(struct seq_file *seq)
2866 {
2867         struct ifmcaddr6 *im = NULL;
2868         struct igmp6_mc_iter_state *state = igmp6_mc_seq_private(seq);
2869         struct net *net = seq_file_net(seq);
2870
2871         state->idev = NULL;
2872         for_each_netdev_rcu(net, state->dev) {
2873                 struct inet6_dev *idev;
2874                 idev = __in6_dev_get(state->dev);
2875                 if (!idev)
2876                         continue;
2877
2878                 im = rcu_dereference(idev->mc_list);
2879                 if (im) {
2880                         state->idev = idev;
2881                         break;
2882                 }
2883         }
2884         return im;
2885 }
2886
2887 static struct ifmcaddr6 *igmp6_mc_get_next(struct seq_file *seq, struct ifmcaddr6 *im)
2888 {
2889         struct igmp6_mc_iter_state *state = igmp6_mc_seq_private(seq);
2890
2891         im = rcu_dereference(im->next);
2892         while (!im) {
2893                 state->dev = next_net_device_rcu(state->dev);
2894                 if (!state->dev) {
2895                         state->idev = NULL;
2896                         break;
2897                 }
2898                 state->idev = __in6_dev_get(state->dev);
2899                 if (!state->idev)
2900                         continue;
2901                 im = rcu_dereference(state->idev->mc_list);
2902         }
2903         return im;
2904 }
2905
2906 static struct ifmcaddr6 *igmp6_mc_get_idx(struct seq_file *seq, loff_t pos)
2907 {
2908         struct ifmcaddr6 *im = igmp6_mc_get_first(seq);
2909         if (im)
2910                 while (pos && (im = igmp6_mc_get_next(seq, im)) != NULL)
2911                         --pos;
2912         return pos ? NULL : im;
2913 }
2914
2915 static void *igmp6_mc_seq_start(struct seq_file *seq, loff_t *pos)
2916         __acquires(RCU)
2917 {
2918         rcu_read_lock();
2919         return igmp6_mc_get_idx(seq, *pos);
2920 }
2921
2922 static void *igmp6_mc_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2923 {
2924         struct ifmcaddr6 *im = igmp6_mc_get_next(seq, v);
2925
2926         ++*pos;
2927         return im;
2928 }
2929
2930 static void igmp6_mc_seq_stop(struct seq_file *seq, void *v)
2931         __releases(RCU)
2932 {
2933         struct igmp6_mc_iter_state *state = igmp6_mc_seq_private(seq);
2934
2935         if (likely(state->idev))
2936                 state->idev = NULL;
2937         state->dev = NULL;
2938         rcu_read_unlock();
2939 }
2940
2941 static int igmp6_mc_seq_show(struct seq_file *seq, void *v)
2942 {
2943         struct ifmcaddr6 *im = (struct ifmcaddr6 *)v;
2944         struct igmp6_mc_iter_state *state = igmp6_mc_seq_private(seq);
2945
2946         seq_printf(seq,
2947                    "%-4d %-15s %pi6 %5d %08X %ld\n",
2948                    state->dev->ifindex, state->dev->name,
2949                    &im->mca_addr,
2950                    im->mca_users, im->mca_flags,
2951                    (im->mca_flags & MAF_TIMER_RUNNING) ?
2952                    jiffies_to_clock_t(im->mca_work.timer.expires - jiffies) : 0);
2953         return 0;
2954 }
2955
2956 static const struct seq_operations igmp6_mc_seq_ops = {
2957         .start  =       igmp6_mc_seq_start,
2958         .next   =       igmp6_mc_seq_next,
2959         .stop   =       igmp6_mc_seq_stop,
2960         .show   =       igmp6_mc_seq_show,
2961 };
2962
2963 struct igmp6_mcf_iter_state {
2964         struct seq_net_private p;
2965         struct net_device *dev;
2966         struct inet6_dev *idev;
2967         struct ifmcaddr6 *im;
2968 };
2969
2970 #define igmp6_mcf_seq_private(seq)      ((struct igmp6_mcf_iter_state *)(seq)->private)
2971
2972 static inline struct ip6_sf_list *igmp6_mcf_get_first(struct seq_file *seq)
2973 {
2974         struct ip6_sf_list *psf = NULL;
2975         struct ifmcaddr6 *im = NULL;
2976         struct igmp6_mcf_iter_state *state = igmp6_mcf_seq_private(seq);
2977         struct net *net = seq_file_net(seq);
2978
2979         state->idev = NULL;
2980         state->im = NULL;
2981         for_each_netdev_rcu(net, state->dev) {
2982                 struct inet6_dev *idev;
2983                 idev = __in6_dev_get(state->dev);
2984                 if (unlikely(idev == NULL))
2985                         continue;
2986
2987                 im = rcu_dereference(idev->mc_list);
2988                 if (likely(im)) {
2989                         psf = rcu_dereference(im->mca_sources);
2990                         if (likely(psf)) {
2991                                 state->im = im;
2992                                 state->idev = idev;
2993                                 break;
2994                         }
2995                 }
2996         }
2997         return psf;
2998 }
2999
3000 static struct ip6_sf_list *igmp6_mcf_get_next(struct seq_file *seq, struct ip6_sf_list *psf)
3001 {
3002         struct igmp6_mcf_iter_state *state = igmp6_mcf_seq_private(seq);
3003
3004         psf = rcu_dereference(psf->sf_next);
3005         while (!psf) {
3006                 state->im = rcu_dereference(state->im->next);
3007                 while (!state->im) {
3008                         state->dev = next_net_device_rcu(state->dev);
3009                         if (!state->dev) {
3010                                 state->idev = NULL;
3011                                 goto out;
3012                         }
3013                         state->idev = __in6_dev_get(state->dev);
3014                         if (!state->idev)
3015                                 continue;
3016                         state->im = rcu_dereference(state->idev->mc_list);
3017                 }
3018                 if (!state->im)
3019                         break;
3020                 psf = rcu_dereference(state->im->mca_sources);
3021         }
3022 out:
3023         return psf;
3024 }
3025
3026 static struct ip6_sf_list *igmp6_mcf_get_idx(struct seq_file *seq, loff_t pos)
3027 {
3028         struct ip6_sf_list *psf = igmp6_mcf_get_first(seq);
3029         if (psf)
3030                 while (pos && (psf = igmp6_mcf_get_next(seq, psf)) != NULL)
3031                         --pos;
3032         return pos ? NULL : psf;
3033 }
3034
3035 static void *igmp6_mcf_seq_start(struct seq_file *seq, loff_t *pos)
3036         __acquires(RCU)
3037 {
3038         rcu_read_lock();
3039         return *pos ? igmp6_mcf_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
3040 }
3041
3042 static void *igmp6_mcf_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3043 {
3044         struct ip6_sf_list *psf;
3045         if (v == SEQ_START_TOKEN)
3046                 psf = igmp6_mcf_get_first(seq);
3047         else
3048                 psf = igmp6_mcf_get_next(seq, v);
3049         ++*pos;
3050         return psf;
3051 }
3052
3053 static void igmp6_mcf_seq_stop(struct seq_file *seq, void *v)
3054         __releases(RCU)
3055 {
3056         struct igmp6_mcf_iter_state *state = igmp6_mcf_seq_private(seq);
3057
3058         if (likely(state->im))
3059                 state->im = NULL;
3060         if (likely(state->idev))
3061                 state->idev = NULL;
3062
3063         state->dev = NULL;
3064         rcu_read_unlock();
3065 }
3066
3067 static int igmp6_mcf_seq_show(struct seq_file *seq, void *v)
3068 {
3069         struct ip6_sf_list *psf = (struct ip6_sf_list *)v;
3070         struct igmp6_mcf_iter_state *state = igmp6_mcf_seq_private(seq);
3071
3072         if (v == SEQ_START_TOKEN) {
3073                 seq_puts(seq, "Idx Device                Multicast Address                   Source Address    INC    EXC\n");
3074         } else {
3075                 seq_printf(seq,
3076                            "%3d %6.6s %pi6 %pi6 %6lu %6lu\n",
3077                            state->dev->ifindex, state->dev->name,
3078                            &state->im->mca_addr,
3079                            &psf->sf_addr,
3080                            psf->sf_count[MCAST_INCLUDE],
3081                            psf->sf_count[MCAST_EXCLUDE]);
3082         }
3083         return 0;
3084 }
3085
3086 static const struct seq_operations igmp6_mcf_seq_ops = {
3087         .start  =       igmp6_mcf_seq_start,
3088         .next   =       igmp6_mcf_seq_next,
3089         .stop   =       igmp6_mcf_seq_stop,
3090         .show   =       igmp6_mcf_seq_show,
3091 };
3092
3093 static int __net_init igmp6_proc_init(struct net *net)
3094 {
3095         int err;
3096
3097         err = -ENOMEM;
3098         if (!proc_create_net("igmp6", 0444, net->proc_net, &igmp6_mc_seq_ops,
3099                         sizeof(struct igmp6_mc_iter_state)))
3100                 goto out;
3101         if (!proc_create_net("mcfilter6", 0444, net->proc_net,
3102                         &igmp6_mcf_seq_ops,
3103                         sizeof(struct igmp6_mcf_iter_state)))
3104                 goto out_proc_net_igmp6;
3105
3106         err = 0;
3107 out:
3108         return err;
3109
3110 out_proc_net_igmp6:
3111         remove_proc_entry("igmp6", net->proc_net);
3112         goto out;
3113 }
3114
3115 static void __net_exit igmp6_proc_exit(struct net *net)
3116 {
3117         remove_proc_entry("mcfilter6", net->proc_net);
3118         remove_proc_entry("igmp6", net->proc_net);
3119 }
3120 #else
3121 static inline int igmp6_proc_init(struct net *net)
3122 {
3123         return 0;
3124 }
3125 static inline void igmp6_proc_exit(struct net *net)
3126 {
3127 }
3128 #endif
3129
3130 static int __net_init igmp6_net_init(struct net *net)
3131 {
3132         int err;
3133
3134         err = inet_ctl_sock_create(&net->ipv6.igmp_sk, PF_INET6,
3135                                    SOCK_RAW, IPPROTO_ICMPV6, net);
3136         if (err < 0) {
3137                 pr_err("Failed to initialize the IGMP6 control socket (err %d)\n",
3138                        err);
3139                 goto out;
3140         }
3141
3142         inet6_sk(net->ipv6.igmp_sk)->hop_limit = 1;
3143         net->ipv6.igmp_sk->sk_allocation = GFP_KERNEL;
3144
3145         err = inet_ctl_sock_create(&net->ipv6.mc_autojoin_sk, PF_INET6,
3146                                    SOCK_RAW, IPPROTO_ICMPV6, net);
3147         if (err < 0) {
3148                 pr_err("Failed to initialize the IGMP6 autojoin socket (err %d)\n",
3149                        err);
3150                 goto out_sock_create;
3151         }
3152
3153         err = igmp6_proc_init(net);
3154         if (err)
3155                 goto out_sock_create_autojoin;
3156
3157         return 0;
3158
3159 out_sock_create_autojoin:
3160         inet_ctl_sock_destroy(net->ipv6.mc_autojoin_sk);
3161 out_sock_create:
3162         inet_ctl_sock_destroy(net->ipv6.igmp_sk);
3163 out:
3164         return err;
3165 }
3166
3167 static void __net_exit igmp6_net_exit(struct net *net)
3168 {
3169         inet_ctl_sock_destroy(net->ipv6.igmp_sk);
3170         inet_ctl_sock_destroy(net->ipv6.mc_autojoin_sk);
3171         igmp6_proc_exit(net);
3172 }
3173
3174 static struct pernet_operations igmp6_net_ops = {
3175         .init = igmp6_net_init,
3176         .exit = igmp6_net_exit,
3177 };
3178
3179 int __init igmp6_init(void)
3180 {
3181         int err;
3182
3183         err = register_pernet_subsys(&igmp6_net_ops);
3184         if (err)
3185                 return err;
3186
3187         mld_wq = create_workqueue("mld");
3188         if (!mld_wq) {
3189                 unregister_pernet_subsys(&igmp6_net_ops);
3190                 return -ENOMEM;
3191         }
3192
3193         return err;
3194 }
3195
3196 int __init igmp6_late_init(void)
3197 {
3198         return register_netdevice_notifier(&igmp6_netdev_notifier);
3199 }
3200
3201 void igmp6_cleanup(void)
3202 {
3203         unregister_pernet_subsys(&igmp6_net_ops);
3204         destroy_workqueue(mld_wq);
3205 }
3206
3207 void igmp6_late_cleanup(void)
3208 {
3209         unregister_netdevice_notifier(&igmp6_netdev_notifier);
3210 }