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