net/ipv6/mcast.c

   1 /*
   2  *      Multicast support for IPv6
   3  *      Linux INET6 implementation
   4  *
   5  *      Authors:
   6  *      Pedro Roque             <roque@di.fc.ul.pt>
   7  *
   8  *      Based on linux/ipv4/igmp.c and linux/ipv4/ip_sockglue.c
   9  *
  10  *      This program is free software; you can redistribute it and/or
  11  *      modify it under the terms of the GNU General Public License
  12  *      as published by the Free Software Foundation; either version
  13  *      2 of the License, or (at your option) any later version.
  14  */
  15
  16 /* Changes:
  17  *
  18  *      yoshfuji        : fix format of router-alert option
  19  *      YOSHIFUJI Hideaki @USAGI:
  20  *              Fixed source address for MLD message based on
  21  *              <draft-ietf-magma-mld-source-05.txt>.
  22  *      YOSHIFUJI Hideaki @USAGI:
  23  *              - Ignore Queries for invalid addresses.
  24  *              - MLD for link-local addresses.
  25  *      David L Stevens <dlstevens@us.ibm.com>:
  26  *              - MLDv2 support
  27  */
  28
  29 #include <linux/module.h>
  30 #include <linux/errno.h>
  31 #include <linux/types.h>
  32 #include <linux/string.h>
  33 #include <linux/socket.h>
  34 #include <linux/sockios.h>
  35 #include <linux/jiffies.h>
  36 #include <linux/times.h>
  37 #include <linux/net.h>
  38 #include <linux/in.h>
  39 #include <linux/in6.h>
  40 #include <linux/netdevice.h>
  41 #include <linux/if_arp.h>
  42 #include <linux/route.h>
  43 #include <linux/init.h>
  44 #include <linux/proc_fs.h>
  45 #include <linux/seq_file.h>
  46 #include <linux/slab.h>
  47 #include <linux/pkt_sched.h>
  48 #include <net/mld.h>
  49
  50 #include <linux/netfilter.h>
  51 #include <linux/netfilter_ipv6.h>
  52
  53 #include <net/net_namespace.h>
  54 #include <net/sock.h>
  55 #include <net/snmp.h>
  56
  57 #include <net/ipv6.h>
  58 #include <net/protocol.h>
  59 #include <net/if_inet6.h>
  60 #include <net/ndisc.h>
  61 #include <net/addrconf.h>
  62 #include <net/ip6_route.h>
  63 #include <net/inet_common.h>
  64
  65 #include <net/ip6_checksum.h>
  66
  67 /* Ensure that we have struct in6_addr aligned on 32bit word. */
  68 static void *__mld2_query_bugs[] __attribute__((__unused__)) = {
  69         BUILD_BUG_ON_NULL(offsetof(struct mld2_query, mld2q_srcs) % 4),
  70         BUILD_BUG_ON_NULL(offsetof(struct mld2_report, mld2r_grec) % 4),
  71         BUILD_BUG_ON_NULL(offsetof(struct mld2_grec, grec_mca) % 4)
  72 };
  73
  74 static struct in6_addr mld2_all_mcr = MLD2_ALL_MCR_INIT;
  75
  76 static void igmp6_join_group(struct ifmcaddr6 *ma);
  77 static void igmp6_leave_group(struct ifmcaddr6 *ma);
  78 static void igmp6_timer_handler(unsigned long data);
  79
  80 static void mld_gq_timer_expire(unsigned long data);
  81 static void mld_ifc_timer_expire(unsigned long data);
  82 static void mld_ifc_event(struct inet6_dev *idev);
  83 static void mld_add_delrec(struct inet6_dev *idev, struct ifmcaddr6 *pmc);
  84 static void mld_del_delrec(struct inet6_dev *idev, struct ifmcaddr6 *pmc);
  85 static void mld_clear_delrec(struct inet6_dev *idev);
  86 static bool mld_in_v1_mode(const struct inet6_dev *idev);
  87 static int sf_setstate(struct ifmcaddr6 *pmc);
  88 static void sf_markstate(struct ifmcaddr6 *pmc);
  89 static void ip6_mc_clear_src(struct ifmcaddr6 *pmc);
  90 static int ip6_mc_del_src(struct inet6_dev *idev, const struct in6_addr *pmca,
  91                           int sfmode, int sfcount, const struct in6_addr *psfsrc,
  92                           int delta);
  93 static int ip6_mc_add_src(struct inet6_dev *idev, const struct in6_addr *pmca,
  94                           int sfmode, int sfcount, const struct in6_addr *psfsrc,
  95                           int delta);
  96 static int ip6_mc_leave_src(struct sock *sk, struct ipv6_mc_socklist *iml,
  97                             struct inet6_dev *idev);
  98
  99 #define MLD_QRV_DEFAULT         2
 100 /* RFC3810, 9.2. Query Interval */
 101 #define MLD_QI_DEFAULT          (125 * HZ)
 102 /* RFC3810, 9.3. Query Response Interval */
 103 #define MLD_QRI_DEFAULT         (10 * HZ)
 104
 105 /* RFC3810, 8.1 Query Version Distinctions */
 106 #define MLD_V1_QUERY_LEN        24
 107 #define MLD_V2_QUERY_LEN_MIN    28
 108
 109 #define IPV6_MLD_MAX_MSF        64
 110
 111 int sysctl_mld_max_msf __read_mostly = IPV6_MLD_MAX_MSF;
 112 int sysctl_mld_qrv __read_mostly = MLD_QRV_DEFAULT;
 113
 114 /*
 115  *      socket join on multicast group
 116  */
 117
 118 #define for_each_pmc_rcu(np, pmc)                               \
 119         for (pmc = rcu_dereference(np->ipv6_mc_list);           \
 120              pmc != NULL;                                       \
 121              pmc = rcu_dereference(pmc->next))
 122
 123 static int unsolicited_report_interval(struct inet6_dev *idev)
 124 {
 125         int iv;
 126
 127         if (mld_in_v1_mode(idev))
 128                 iv = idev->cnf.mldv1_unsolicited_report_interval;
 129         else
 130                 iv = idev->cnf.mldv2_unsolicited_report_interval;
 131
 132         return iv > 0 ? iv : 1;
 133 }
 134
 135 int ipv6_sock_mc_join(struct sock *sk, int ifindex, const struct in6_addr *addr)
 136 {
 137         struct net_device *dev = NULL;
 138         struct ipv6_mc_socklist *mc_lst;
 139         struct ipv6_pinfo *np = inet6_sk(sk);
 140         struct net *net = sock_net(sk);
 141         int err;
 142
 143         ASSERT_RTNL();
 144
 145         if (!ipv6_addr_is_multicast(addr))
 146                 return -EINVAL;
 147
 148         rcu_read_lock();
 149         for_each_pmc_rcu(np, mc_lst) {
 150                 if ((ifindex == 0 || mc_lst->ifindex == ifindex) &&
 151                     ipv6_addr_equal(&mc_lst->addr, addr)) {
 152                         rcu_read_unlock();
 153                         return -EADDRINUSE;
 154                 }
 155         }
 156         rcu_read_unlock();
 157
 158         mc_lst = sock_kmalloc(sk, sizeof(struct ipv6_mc_socklist), GFP_KERNEL);
 159
 160         if (!mc_lst)
 161                 return -ENOMEM;
 162
 163         mc_lst->next = NULL;
 164         mc_lst->addr = *addr;
 165
 166         if (ifindex == 0) {
 167                 struct rt6_info *rt;
 168                 rt = rt6_lookup(net, addr, NULL, 0, 0);
 169                 if (rt) {
 170                         dev = rt->dst.dev;
 171                         ip6_rt_put(rt);
 172                 }
 173         } else
 174                 dev = __dev_get_by_index(net, ifindex);
 175
 176         if (!dev) {
 177                 sock_kfree_s(sk, mc_lst, sizeof(*mc_lst));
 178                 return -ENODEV;
 179         }
 180
 181         mc_lst->ifindex = dev->ifindex;
 182         mc_lst->sfmode = MCAST_EXCLUDE;
 183         rwlock_init(&mc_lst->sflock);
 184         mc_lst->sflist = NULL;
 185
 186         /*
 187          *      now add/increase the group membership on the device
 188          */
 189
 190         err = ipv6_dev_mc_inc(dev, addr);
 191
 192         if (err) {
 193                 sock_kfree_s(sk, mc_lst, sizeof(*mc_lst));
 194                 return err;
 195         }
 196
 197         mc_lst->next = np->ipv6_mc_list;
 198         rcu_assign_pointer(np->ipv6_mc_list, mc_lst);
 199
 200         return 0;
 201 }
 202 EXPORT_SYMBOL(ipv6_sock_mc_join);
 203
 204 /*
 205  *      socket leave on multicast group
 206  */
 207 int ipv6_sock_mc_drop(struct sock *sk, int ifindex, const struct in6_addr *addr)
 208 {
 209         struct ipv6_pinfo *np = inet6_sk(sk);
 210         struct ipv6_mc_socklist *mc_lst;
 211         struct ipv6_mc_socklist __rcu **lnk;
 212         struct net *net = sock_net(sk);
 213
 214         ASSERT_RTNL();
 215
 216         if (!ipv6_addr_is_multicast(addr))
 217                 return -EINVAL;
 218
 219         for (lnk = &np->ipv6_mc_list;
 220              (mc_lst = rtnl_dereference(*lnk)) != NULL;
 221               lnk = &mc_lst->next) {
 222                 if ((ifindex == 0 || mc_lst->ifindex == ifindex) &&
 223                     ipv6_addr_equal(&mc_lst->addr, addr)) {
 224                         struct net_device *dev;
 225
 226                         *lnk = mc_lst->next;
 227
 228                         dev = __dev_get_by_index(net, mc_lst->ifindex);
 229                         if (dev) {
 230                                 struct inet6_dev *idev = __in6_dev_get(dev);
 231
 232                                 (void) ip6_mc_leave_src(sk, mc_lst, idev);
 233                                 if (idev)
 234                                         __ipv6_dev_mc_dec(idev, &mc_lst->addr);
 235                         } else
 236                                 (void) ip6_mc_leave_src(sk, mc_lst, NULL);
 237
 238                         atomic_sub(sizeof(*mc_lst), &sk->sk_omem_alloc);
 239                         kfree_rcu(mc_lst, rcu);
 240                         return 0;
 241                 }
 242         }
 243
 244         return -EADDRNOTAVAIL;
 245 }
 246 EXPORT_SYMBOL(ipv6_sock_mc_drop);
 247
 248 /* called with rcu_read_lock() */
 249 static struct inet6_dev *ip6_mc_find_dev_rcu(struct net *net,
 250                                              const struct in6_addr *group,
 251                                              int ifindex)
 252 {
 253         struct net_device *dev = NULL;
 254         struct inet6_dev *idev = NULL;
 255
 256         if (ifindex == 0) {
 257                 struct rt6_info *rt = rt6_lookup(net, group, NULL, 0, 0);
 258
 259                 if (rt) {
 260                         dev = rt->dst.dev;
 261                         ip6_rt_put(rt);
 262                 }
 263         } else
 264                 dev = dev_get_by_index_rcu(net, ifindex);
 265
 266         if (!dev)
 267                 return NULL;
 268         idev = __in6_dev_get(dev);
 269         if (!idev)
 270                 return NULL;
 271         read_lock_bh(&idev->lock);
 272         if (idev->dead) {
 273                 read_unlock_bh(&idev->lock);
 274                 return NULL;
 275         }
 276         return idev;
 277 }
 278
 279 void __ipv6_sock_mc_close(struct sock *sk)
 280 {
 281         struct ipv6_pinfo *np = inet6_sk(sk);
 282         struct ipv6_mc_socklist *mc_lst;
 283         struct net *net = sock_net(sk);
 284
 285         ASSERT_RTNL();
 286
 287         while ((mc_lst = rtnl_dereference(np->ipv6_mc_list)) != NULL) {
 288                 struct net_device *dev;
 289
 290                 np->ipv6_mc_list = mc_lst->next;
 291
 292                 dev = __dev_get_by_index(net, mc_lst->ifindex);
 293                 if (dev) {
 294                         struct inet6_dev *idev = __in6_dev_get(dev);
 295
 296                         (void) ip6_mc_leave_src(sk, mc_lst, idev);
 297                         if (idev)
 298                                 __ipv6_dev_mc_dec(idev, &mc_lst->addr);
 299                 } else
 300                         (void) ip6_mc_leave_src(sk, mc_lst, NULL);
 301
 302                 atomic_sub(sizeof(*mc_lst), &sk->sk_omem_alloc);
 303                 kfree_rcu(mc_lst, rcu);
 304         }
 305 }
 306
 307 void ipv6_sock_mc_close(struct sock *sk)
 308 {
 309         struct ipv6_pinfo *np = inet6_sk(sk);
 310
 311         if (!rcu_access_pointer(np->ipv6_mc_list))
 312                 return;
 313         rtnl_lock();
 314         __ipv6_sock_mc_close(sk);
 315         rtnl_unlock();
 316 }
 317
 318 int ip6_mc_source(int add, int omode, struct sock *sk,
 319         struct group_source_req *pgsr)
 320 {
 321         struct in6_addr *source, *group;
 322         struct ipv6_mc_socklist *pmc;
 323         struct inet6_dev *idev;
 324         struct ipv6_pinfo *inet6 = inet6_sk(sk);
 325         struct ip6_sf_socklist *psl;
 326         struct net *net = sock_net(sk);
 327         int i, j, rv;
 328         int leavegroup = 0;
 329         int pmclocked = 0;
 330         int err;
 331
 332         source = &((struct sockaddr_in6 *)&pgsr->gsr_source)->sin6_addr;
 333         group = &((struct sockaddr_in6 *)&pgsr->gsr_group)->sin6_addr;
 334
 335         if (!ipv6_addr_is_multicast(group))
 336                 return -EINVAL;
 337
 338         rcu_read_lock();
 339         idev = ip6_mc_find_dev_rcu(net, group, pgsr->gsr_interface);
 340         if (!idev) {
 341                 rcu_read_unlock();
 342                 return -ENODEV;
 343         }
 344
 345         err = -EADDRNOTAVAIL;
 346
 347         for_each_pmc_rcu(inet6, pmc) {
 348                 if (pgsr->gsr_interface && pmc->ifindex != pgsr->gsr_interface)
 349                         continue;
 350                 if (ipv6_addr_equal(&pmc->addr, group))
 351                         break;
 352         }
 353         if (!pmc) {             /* must have a prior join */
 354                 err = -EINVAL;
 355                 goto done;
 356         }
 357         /* if a source filter was set, must be the same mode as before */
 358         if (pmc->sflist) {
 359                 if (pmc->sfmode != omode) {
 360                         err = -EINVAL;
 361                         goto done;
 362                 }
 363         } else if (pmc->sfmode != omode) {
 364                 /* allow mode switches for empty-set filters */
 365                 ip6_mc_add_src(idev, group, omode, 0, NULL, 0);
 366                 ip6_mc_del_src(idev, group, pmc->sfmode, 0, NULL, 0);
 367                 pmc->sfmode = omode;
 368         }
 369
 370         write_lock(&pmc->sflock);
 371         pmclocked = 1;
 372
 373         psl = pmc->sflist;
 374         if (!add) {
 375                 if (!psl)
 376                         goto done;      /* err = -EADDRNOTAVAIL */
 377                 rv = !0;
 378                 for (i = 0; i < psl->sl_count; i++) {
 379                         rv = !ipv6_addr_equal(&psl->sl_addr[i], source);
 380                         if (rv == 0)
 381                                 break;
 382                 }
 383                 if (rv)         /* source not found */
 384                         goto done;      /* err = -EADDRNOTAVAIL */
 385
 386                 /* special case - (INCLUDE, empty) == LEAVE_GROUP */
 387                 if (psl->sl_count == 1 && omode == MCAST_INCLUDE) {
 388                         leavegroup = 1;
 389                         goto done;
 390                 }
 391
 392                 /* update the interface filter */
 393                 ip6_mc_del_src(idev, group, omode, 1, source, 1);
 394
 395                 for (j = i+1; j < psl->sl_count; j++)
 396                         psl->sl_addr[j-1] = psl->sl_addr[j];
 397                 psl->sl_count--;
 398                 err = 0;
 399                 goto done;
 400         }
 401         /* else, add a new source to the filter */
 402
 403         if (psl && psl->sl_count >= sysctl_mld_max_msf) {
 404                 err = -ENOBUFS;
 405                 goto done;
 406         }
 407         if (!psl || psl->sl_count == psl->sl_max) {
 408                 struct ip6_sf_socklist *newpsl;
 409                 int count = IP6_SFBLOCK;
 410
 411                 if (psl)
 412                         count += psl->sl_max;
 413                 newpsl = sock_kmalloc(sk, IP6_SFLSIZE(count), GFP_ATOMIC);
 414                 if (!newpsl) {
 415                         err = -ENOBUFS;
 416                         goto done;
 417                 }
 418                 newpsl->sl_max = count;
 419                 newpsl->sl_count = count - IP6_SFBLOCK;
 420                 if (psl) {
 421                         for (i = 0; i < psl->sl_count; i++)
 422                                 newpsl->sl_addr[i] = psl->sl_addr[i];
 423                         sock_kfree_s(sk, psl, IP6_SFLSIZE(psl->sl_max));
 424                 }
 425                 pmc->sflist = psl = newpsl;
 426         }
 427         rv = 1; /* > 0 for insert logic below if sl_count is 0 */
 428         for (i = 0; i < psl->sl_count; i++) {
 429                 rv = !ipv6_addr_equal(&psl->sl_addr[i], source);
 430                 if (rv == 0) /* There is an error in the address. */
 431                         goto done;
 432         }
 433         for (j = psl->sl_count-1; j >= i; j--)
 434                 psl->sl_addr[j+1] = psl->sl_addr[j];
 435         psl->sl_addr[i] = *source;
 436         psl->sl_count++;
 437         err = 0;
 438         /* update the interface list */
 439         ip6_mc_add_src(idev, group, omode, 1, source, 1);
 440 done:
 441         if (pmclocked)
 442                 write_unlock(&pmc->sflock);
 443         read_unlock_bh(&idev->lock);
 444         rcu_read_unlock();
 445         if (leavegroup)
 446                 err = ipv6_sock_mc_drop(sk, pgsr->gsr_interface, group);
 447         return err;
 448 }
 449
 450 int ip6_mc_msfilter(struct sock *sk, struct group_filter *gsf)
 451 {
 452         const struct in6_addr *group;
 453         struct ipv6_mc_socklist *pmc;
 454         struct inet6_dev *idev;
 455         struct ipv6_pinfo *inet6 = inet6_sk(sk);
 456         struct ip6_sf_socklist *newpsl, *psl;
 457         struct net *net = sock_net(sk);
 458         int leavegroup = 0;
 459         int i, err;
 460
 461         group = &((struct sockaddr_in6 *)&gsf->gf_group)->sin6_addr;
 462
 463         if (!ipv6_addr_is_multicast(group))
 464                 return -EINVAL;
 465         if (gsf->gf_fmode != MCAST_INCLUDE &&
 466             gsf->gf_fmode != MCAST_EXCLUDE)
 467                 return -EINVAL;
 468
 469         rcu_read_lock();
 470         idev = ip6_mc_find_dev_rcu(net, group, gsf->gf_interface);
 471
 472         if (!idev) {
 473                 rcu_read_unlock();
 474                 return -ENODEV;
 475         }
 476
 477         err = 0;
 478
 479         if (gsf->gf_fmode == MCAST_INCLUDE && gsf->gf_numsrc == 0) {
 480                 leavegroup = 1;
 481                 goto done;
 482         }
 483
 484         for_each_pmc_rcu(inet6, pmc) {
 485                 if (pmc->ifindex != gsf->gf_interface)
 486                         continue;
 487                 if (ipv6_addr_equal(&pmc->addr, group))
 488                         break;
 489         }
 490         if (!pmc) {             /* must have a prior join */
 491                 err = -EINVAL;
 492                 goto done;
 493         }
 494         if (gsf->gf_numsrc) {
 495                 newpsl = sock_kmalloc(sk, IP6_SFLSIZE(gsf->gf_numsrc),
 496                                                           GFP_ATOMIC);
 497                 if (!newpsl) {
 498                         err = -ENOBUFS;
 499                         goto done;
 500                 }
 501                 newpsl->sl_max = newpsl->sl_count = gsf->gf_numsrc;
 502                 for (i = 0; i < newpsl->sl_count; ++i) {
 503                         struct sockaddr_in6 *psin6;
 504
 505                         psin6 = (struct sockaddr_in6 *)&gsf->gf_slist[i];
 506                         newpsl->sl_addr[i] = psin6->sin6_addr;
 507                 }
 508                 err = ip6_mc_add_src(idev, group, gsf->gf_fmode,
 509                         newpsl->sl_count, newpsl->sl_addr, 0);
 510                 if (err) {
 511                         sock_kfree_s(sk, newpsl, IP6_SFLSIZE(newpsl->sl_max));
 512                         goto done;
 513                 }
 514         } else {
 515                 newpsl = NULL;
 516                 (void) ip6_mc_add_src(idev, group, gsf->gf_fmode, 0, NULL, 0);
 517         }
 518
 519         write_lock(&pmc->sflock);
 520         psl = pmc->sflist;
 521         if (psl) {
 522                 (void) ip6_mc_del_src(idev, group, pmc->sfmode,
 523                         psl->sl_count, psl->sl_addr, 0);
 524                 sock_kfree_s(sk, psl, IP6_SFLSIZE(psl->sl_max));
 525         } else
 526                 (void) ip6_mc_del_src(idev, group, pmc->sfmode, 0, NULL, 0);
 527         pmc->sflist = newpsl;
 528         pmc->sfmode = gsf->gf_fmode;
 529         write_unlock(&pmc->sflock);
 530         err = 0;
 531 done:
 532         read_unlock_bh(&idev->lock);
 533         rcu_read_unlock();
 534         if (leavegroup)
 535                 err = ipv6_sock_mc_drop(sk, gsf->gf_interface, group);
 536         return err;
 537 }
 538
 539 int ip6_mc_msfget(struct sock *sk, struct group_filter *gsf,
 540         struct group_filter __user *optval, int __user *optlen)
 541 {
 542         int err, i, count, copycount;
 543         const struct in6_addr *group;
 544         struct ipv6_mc_socklist *pmc;
 545         struct inet6_dev *idev;
 546         struct ipv6_pinfo *inet6 = inet6_sk(sk);
 547         struct ip6_sf_socklist *psl;
 548         struct net *net = sock_net(sk);
 549
 550         group = &((struct sockaddr_in6 *)&gsf->gf_group)->sin6_addr;
 551
 552         if (!ipv6_addr_is_multicast(group))
 553                 return -EINVAL;
 554
 555         rcu_read_lock();
 556         idev = ip6_mc_find_dev_rcu(net, group, gsf->gf_interface);
 557
 558         if (!idev) {
 559                 rcu_read_unlock();
 560                 return -ENODEV;
 561         }
 562
 563         err = -EADDRNOTAVAIL;
 564         /* changes to the ipv6_mc_list require the socket lock and
 565          * rtnl lock. We have the socket lock and rcu read lock,
 566          * so reading the list is safe.
 567          */
 568
 569         for_each_pmc_rcu(inet6, pmc) {
 570                 if (pmc->ifindex != gsf->gf_interface)
 571                         continue;
 572                 if (ipv6_addr_equal(group, &pmc->addr))
 573                         break;
 574         }
 575         if (!pmc)               /* must have a prior join */
 576                 goto done;
 577         gsf->gf_fmode = pmc->sfmode;
 578         psl = pmc->sflist;
 579         count = psl ? psl->sl_count : 0;
 580         read_unlock_bh(&idev->lock);
 581         rcu_read_unlock();
 582
 583         copycount = count < gsf->gf_numsrc ? count : gsf->gf_numsrc;
 584         gsf->gf_numsrc = count;
 585         if (put_user(GROUP_FILTER_SIZE(copycount), optlen) ||
 586             copy_to_user(optval, gsf, GROUP_FILTER_SIZE(0))) {
 587                 return -EFAULT;
 588         }
 589         /* changes to psl require the socket lock, and a write lock
 590          * on pmc->sflock. We have the socket lock so reading here is safe.
 591          */
 592         for (i = 0; i < copycount; i++) {
 593                 struct sockaddr_in6 *psin6;
 594                 struct sockaddr_storage ss;
 595
 596                 psin6 = (struct sockaddr_in6 *)&ss;
 597                 memset(&ss, 0, sizeof(ss));
 598                 psin6->sin6_family = AF_INET6;
 599                 psin6->sin6_addr = psl->sl_addr[i];
 600                 if (copy_to_user(&optval->gf_slist[i], &ss, sizeof(ss)))
 601                         return -EFAULT;
 602         }
 603         return 0;
 604 done:
 605         read_unlock_bh(&idev->lock);
 606         rcu_read_unlock();
 607         return err;
 608 }
 609
 610 bool inet6_mc_check(struct sock *sk, const struct in6_addr *mc_addr,
 611                     const struct in6_addr *src_addr)
 612 {
 613         struct ipv6_pinfo *np = inet6_sk(sk);
 614         struct ipv6_mc_socklist *mc;
 615         struct ip6_sf_socklist *psl;
 616         bool rv = true;
 617
 618         rcu_read_lock();
 619         for_each_pmc_rcu(np, mc) {
 620                 if (ipv6_addr_equal(&mc->addr, mc_addr))
 621                         break;
 622         }
 623         if (!mc) {
 624                 rcu_read_unlock();
 625                 return true;
 626         }
 627         read_lock(&mc->sflock);
 628         psl = mc->sflist;
 629         if (!psl) {
 630                 rv = mc->sfmode == MCAST_EXCLUDE;
 631         } else {
 632                 int i;
 633
 634                 for (i = 0; i < psl->sl_count; i++) {
 635                         if (ipv6_addr_equal(&psl->sl_addr[i], src_addr))
 636                                 break;
 637                 }
 638                 if (mc->sfmode == MCAST_INCLUDE && i >= psl->sl_count)
 639                         rv = false;
 640                 if (mc->sfmode == MCAST_EXCLUDE && i < psl->sl_count)
 641                         rv = false;
 642         }
 643         read_unlock(&mc->sflock);
 644         rcu_read_unlock();
 645
 646         return rv;
 647 }
 648
 649 static void igmp6_group_added(struct ifmcaddr6 *mc)
 650 {
 651         struct net_device *dev = mc->idev->dev;
 652         char buf[MAX_ADDR_LEN];
 653
 654         if (IPV6_ADDR_MC_SCOPE(&mc->mca_addr) <
 655             IPV6_ADDR_SCOPE_LINKLOCAL)
 656                 return;
 657
 658         spin_lock_bh(&mc->mca_lock);
 659         if (!(mc->mca_flags&MAF_LOADED)) {
 660                 mc->mca_flags |= MAF_LOADED;
 661                 if (ndisc_mc_map(&mc->mca_addr, buf, dev, 0) == 0)
 662                         dev_mc_add(dev, buf);
 663         }
 664         spin_unlock_bh(&mc->mca_lock);
 665
 666         if (!(dev->flags & IFF_UP) || (mc->mca_flags & MAF_NOREPORT))
 667                 return;
 668
 669         if (mld_in_v1_mode(mc->idev)) {
 670                 igmp6_join_group(mc);
 671                 return;
 672         }
 673         /* else v2 */
 674
 675         mc->mca_crcount = mc->idev->mc_qrv;
 676         mld_ifc_event(mc->idev);
 677 }
 678
 679 static void igmp6_group_dropped(struct ifmcaddr6 *mc)
 680 {
 681         struct net_device *dev = mc->idev->dev;
 682         char buf[MAX_ADDR_LEN];
 683
 684         if (IPV6_ADDR_MC_SCOPE(&mc->mca_addr) <
 685             IPV6_ADDR_SCOPE_LINKLOCAL)
 686                 return;
 687
 688         spin_lock_bh(&mc->mca_lock);
 689         if (mc->mca_flags&MAF_LOADED) {
 690                 mc->mca_flags &= ~MAF_LOADED;
 691                 if (ndisc_mc_map(&mc->mca_addr, buf, dev, 0) == 0)
 692                         dev_mc_del(dev, buf);
 693         }
 694
 695         spin_unlock_bh(&mc->mca_lock);
 696         if (mc->mca_flags & MAF_NOREPORT)
 697                 return;
 698
 699         if (!mc->idev->dead)
 700                 igmp6_leave_group(mc);
 701
 702         spin_lock_bh(&mc->mca_lock);
 703         if (del_timer(&mc->mca_timer))
 704                 atomic_dec(&mc->mca_refcnt);
 705         spin_unlock_bh(&mc->mca_lock);
 706 }
 707
 708 /*
 709  * deleted ifmcaddr6 manipulation
 710  */
 711 static void mld_add_delrec(struct inet6_dev *idev, struct ifmcaddr6 *im)
 712 {
 713         struct ifmcaddr6 *pmc;
 714
 715         /* this is an "ifmcaddr6" for convenience; only the fields below
 716          * are actually used. In particular, the refcnt and users are not
 717          * used for management of the delete list. Using the same structure
 718          * for deleted items allows change reports to use common code with
 719          * non-deleted or query-response MCA's.
 720          */
 721         pmc = kzalloc(sizeof(*pmc), GFP_ATOMIC);
 722         if (!pmc)
 723                 return;
 724
 725         spin_lock_bh(&im->mca_lock);
 726         spin_lock_init(&pmc->mca_lock);
 727         pmc->idev = im->idev;
 728         in6_dev_hold(idev);
 729         pmc->mca_addr = im->mca_addr;
 730         pmc->mca_crcount = idev->mc_qrv;
 731         pmc->mca_sfmode = im->mca_sfmode;
 732         if (pmc->mca_sfmode == MCAST_INCLUDE) {
 733                 struct ip6_sf_list *psf;
 734
 735                 pmc->mca_tomb = im->mca_tomb;
 736                 pmc->mca_sources = im->mca_sources;
 737                 im->mca_tomb = im->mca_sources = NULL;
 738                 for (psf = pmc->mca_sources; psf; psf = psf->sf_next)
 739                         psf->sf_crcount = pmc->mca_crcount;
 740         }
 741         spin_unlock_bh(&im->mca_lock);
 742
 743         spin_lock_bh(&idev->mc_lock);
 744         pmc->next = idev->mc_tomb;
 745         idev->mc_tomb = pmc;
 746         spin_unlock_bh(&idev->mc_lock);
 747 }
 748
 749 static void mld_del_delrec(struct inet6_dev *idev, struct ifmcaddr6 *im)
 750 {
 751         struct ifmcaddr6 *pmc, *pmc_prev;
 752         struct ip6_sf_list *psf;
 753         struct in6_addr *pmca = &im->mca_addr;
 754
 755         spin_lock_bh(&idev->mc_lock);
 756         pmc_prev = NULL;
 757         for (pmc = idev->mc_tomb; pmc; pmc = pmc->next) {
 758                 if (ipv6_addr_equal(&pmc->mca_addr, pmca))
 759                         break;
 760                 pmc_prev = pmc;
 761         }
 762         if (pmc) {
 763                 if (pmc_prev)
 764                         pmc_prev->next = pmc->next;
 765                 else
 766                         idev->mc_tomb = pmc->next;
 767         }
 768         spin_unlock_bh(&idev->mc_lock);
 769
 770         spin_lock_bh(&im->mca_lock);
 771         if (pmc) {
 772                 im->idev = pmc->idev;
 773                 im->mca_crcount = idev->mc_qrv;
 774                 im->mca_sfmode = pmc->mca_sfmode;
 775                 if (pmc->mca_sfmode == MCAST_INCLUDE) {
 776                         im->mca_tomb = pmc->mca_tomb;
 777                         im->mca_sources = pmc->mca_sources;
 778                         for (psf = im->mca_sources; psf; psf = psf->sf_next)
 779                                 psf->sf_crcount = im->mca_crcount;
 780                 }
 781                 in6_dev_put(pmc->idev);
 782         }
 783         spin_unlock_bh(&im->mca_lock);
 784 }
 785
 786 static void mld_clear_delrec(struct inet6_dev *idev)
 787 {
 788         struct ifmcaddr6 *pmc, *nextpmc;
 789
 790         spin_lock_bh(&idev->mc_lock);
 791         pmc = idev->mc_tomb;
 792         idev->mc_tomb = NULL;
 793         spin_unlock_bh(&idev->mc_lock);
 794
 795         for (; pmc; pmc = nextpmc) {
 796                 nextpmc = pmc->next;
 797                 ip6_mc_clear_src(pmc);
 798                 in6_dev_put(pmc->idev);
 799                 kfree(pmc);
 800         }
 801
 802         /* clear dead sources, too */
 803         read_lock_bh(&idev->lock);
 804         for (pmc = idev->mc_list; pmc; pmc = pmc->next) {
 805                 struct ip6_sf_list *psf, *psf_next;
 806
 807                 spin_lock_bh(&pmc->mca_lock);
 808                 psf = pmc->mca_tomb;
 809                 pmc->mca_tomb = NULL;
 810                 spin_unlock_bh(&pmc->mca_lock);
 811                 for (; psf; psf = psf_next) {
 812                         psf_next = psf->sf_next;
 813                         kfree(psf);
 814                 }
 815         }
 816         read_unlock_bh(&idev->lock);
 817 }
 818
 819 static void mca_get(struct ifmcaddr6 *mc)
 820 {
 821         atomic_inc(&mc->mca_refcnt);
 822 }
 823
 824 static void ma_put(struct ifmcaddr6 *mc)
 825 {
 826         if (atomic_dec_and_test(&mc->mca_refcnt)) {
 827                 in6_dev_put(mc->idev);
 828                 kfree(mc);
 829         }
 830 }
 831
 832 static struct ifmcaddr6 *mca_alloc(struct inet6_dev *idev,
 833                                    const struct in6_addr *addr)
 834 {
 835         struct ifmcaddr6 *mc;
 836
 837         mc = kzalloc(sizeof(*mc), GFP_ATOMIC);
 838         if (!mc)
 839                 return NULL;
 840
 841         setup_timer(&mc->mca_timer, igmp6_timer_handler, (unsigned long)mc);
 842
 843         mc->mca_addr = *addr;
 844         mc->idev = idev; /* reference taken by caller */
 845         mc->mca_users = 1;
 846         /* mca_stamp should be updated upon changes */
 847         mc->mca_cstamp = mc->mca_tstamp = jiffies;
 848         atomic_set(&mc->mca_refcnt, 1);
 849         spin_lock_init(&mc->mca_lock);
 850
 851         /* initial mode is (EX, empty) */
 852         mc->mca_sfmode = MCAST_EXCLUDE;
 853         mc->mca_sfcount[MCAST_EXCLUDE] = 1;
 854
 855         if (ipv6_addr_is_ll_all_nodes(&mc->mca_addr) ||
 856             IPV6_ADDR_MC_SCOPE(&mc->mca_addr) < IPV6_ADDR_SCOPE_LINKLOCAL)
 857                 mc->mca_flags |= MAF_NOREPORT;
 858
 859         return mc;
 860 }
 861
 862 /*
 863  *      device multicast group inc (add if not found)
 864  */
 865 int ipv6_dev_mc_inc(struct net_device *dev, const struct in6_addr *addr)
 866 {
 867         struct ifmcaddr6 *mc;
 868         struct inet6_dev *idev;
 869
 870         ASSERT_RTNL();
 871
 872         /* we need to take a reference on idev */
 873         idev = in6_dev_get(dev);
 874
 875         if (!idev)
 876                 return -EINVAL;
 877
 878         write_lock_bh(&idev->lock);
 879         if (idev->dead) {
 880                 write_unlock_bh(&idev->lock);
 881                 in6_dev_put(idev);
 882                 return -ENODEV;
 883         }
 884
 885         for (mc = idev->mc_list; mc; mc = mc->next) {
 886                 if (ipv6_addr_equal(&mc->mca_addr, addr)) {
 887                         mc->mca_users++;
 888                         write_unlock_bh(&idev->lock);
 889                         ip6_mc_add_src(idev, &mc->mca_addr, MCAST_EXCLUDE, 0,
 890                                 NULL, 0);
 891                         in6_dev_put(idev);
 892                         return 0;
 893                 }
 894         }
 895
 896         mc = mca_alloc(idev, addr);
 897         if (!mc) {
 898                 write_unlock_bh(&idev->lock);
 899                 in6_dev_put(idev);
 900                 return -ENOMEM;
 901         }
 902
 903         mc->next = idev->mc_list;
 904         idev->mc_list = mc;
 905
 906         /* Hold this for the code below before we unlock,
 907          * it is already exposed via idev->mc_list.
 908          */
 909         mca_get(mc);
 910         write_unlock_bh(&idev->lock);
 911
 912         mld_del_delrec(idev, mc);
 913         igmp6_group_added(mc);
 914         ma_put(mc);
 915         return 0;
 916 }
 917
 918 /*
 919  *      device multicast group del
 920  */
 921 int __ipv6_dev_mc_dec(struct inet6_dev *idev, const struct in6_addr *addr)
 922 {
 923         struct ifmcaddr6 *ma, **map;
 924
 925         ASSERT_RTNL();
 926
 927         write_lock_bh(&idev->lock);
 928         for (map = &idev->mc_list; (ma = *map) != NULL; map = &ma->next) {
 929                 if (ipv6_addr_equal(&ma->mca_addr, addr)) {
 930                         if (--ma->mca_users == 0) {
 931                                 *map = ma->next;
 932                                 write_unlock_bh(&idev->lock);
 933
 934                                 igmp6_group_dropped(ma);
 935                                 ip6_mc_clear_src(ma);
 936
 937                                 ma_put(ma);
 938                                 return 0;
 939                         }
 940                         write_unlock_bh(&idev->lock);
 941                         return 0;
 942                 }
 943         }
 944         write_unlock_bh(&idev->lock);
 945
 946         return -ENOENT;
 947 }
 948
 949 int ipv6_dev_mc_dec(struct net_device *dev, const struct in6_addr *addr)
 950 {
 951         struct inet6_dev *idev;
 952         int err;
 953
 954         ASSERT_RTNL();
 955
 956         idev = __in6_dev_get(dev);
 957         if (!idev)
 958                 err = -ENODEV;
 959         else
 960                 err = __ipv6_dev_mc_dec(idev, addr);
 961
 962         return err;
 963 }
 964
 965 /*
 966  *      check if the interface/address pair is valid
 967  */
 968 bool ipv6_chk_mcast_addr(struct net_device *dev, const struct in6_addr *group,
 969                          const struct in6_addr *src_addr)
 970 {
 971         struct inet6_dev *idev;
 972         struct ifmcaddr6 *mc;
 973         bool rv = false;
 974
 975         rcu_read_lock();
 976         idev = __in6_dev_get(dev);
 977         if (idev) {
 978                 read_lock_bh(&idev->lock);
 979                 for (mc = idev->mc_list; mc; mc = mc->next) {
 980                         if (ipv6_addr_equal(&mc->mca_addr, group))
 981                                 break;
 982                 }
 983                 if (mc) {
 984                         if (src_addr && !ipv6_addr_any(src_addr)) {
 985                                 struct ip6_sf_list *psf;
 986
 987                                 spin_lock_bh(&mc->mca_lock);
 988                                 for (psf = mc->mca_sources; psf; psf = psf->sf_next) {
 989                                         if (ipv6_addr_equal(&psf->sf_addr, src_addr))
 990                                                 break;
 991                                 }
 992                                 if (psf)
 993                                         rv = psf->sf_count[MCAST_INCLUDE] ||
 994                                                 psf->sf_count[MCAST_EXCLUDE] !=
 995                                                 mc->mca_sfcount[MCAST_EXCLUDE];
 996                                 else
 997                                         rv = mc->mca_sfcount[MCAST_EXCLUDE] != 0;
 998                                 spin_unlock_bh(&mc->mca_lock);
 999                         } else
1000                                 rv = true; /* don't filter unspecified source */
1001                 }
1002                 read_unlock_bh(&idev->lock);
1003         }
1004         rcu_read_unlock();
1005         return rv;
1006 }
1007
1008 static void mld_gq_start_timer(struct inet6_dev *idev)
1009 {
1010         unsigned long tv = prandom_u32() % idev->mc_maxdelay;
1011
1012         idev->mc_gq_running = 1;
1013         if (!mod_timer(&idev->mc_gq_timer, jiffies+tv+2))
1014                 in6_dev_hold(idev);
1015 }
1016
1017 static void mld_gq_stop_timer(struct inet6_dev *idev)
1018 {
1019         idev->mc_gq_running = 0;
1020         if (del_timer(&idev->mc_gq_timer))
1021                 __in6_dev_put(idev);
1022 }
1023
1024 static void mld_ifc_start_timer(struct inet6_dev *idev, unsigned long delay)
1025 {
1026         unsigned long tv = prandom_u32() % delay;
1027
1028         if (!mod_timer(&idev->mc_ifc_timer, jiffies+tv+2))
1029                 in6_dev_hold(idev);
1030 }
1031
1032 static void mld_ifc_stop_timer(struct inet6_dev *idev)
1033 {
1034         idev->mc_ifc_count = 0;
1035         if (del_timer(&idev->mc_ifc_timer))
1036                 __in6_dev_put(idev);
1037 }
1038
1039 static void mld_dad_start_timer(struct inet6_dev *idev, unsigned long delay)
1040 {
1041         unsigned long tv = prandom_u32() % delay;
1042
1043         if (!mod_timer(&idev->mc_dad_timer, jiffies+tv+2))
1044                 in6_dev_hold(idev);
1045 }
1046
1047 static void mld_dad_stop_timer(struct inet6_dev *idev)
1048 {
1049         if (del_timer(&idev->mc_dad_timer))
1050                 __in6_dev_put(idev);
1051 }
1052
1053 /*
1054  *      IGMP handling (alias multicast ICMPv6 messages)
1055  */
1056
1057 static void igmp6_group_queried(struct ifmcaddr6 *ma, unsigned long resptime)
1058 {
1059         unsigned long delay = resptime;
1060
1061         /* Do not start timer for these addresses */
1062         if (ipv6_addr_is_ll_all_nodes(&ma->mca_addr) ||
1063             IPV6_ADDR_MC_SCOPE(&ma->mca_addr) < IPV6_ADDR_SCOPE_LINKLOCAL)
1064                 return;
1065
1066         if (del_timer(&ma->mca_timer)) {
1067                 atomic_dec(&ma->mca_refcnt);
1068                 delay = ma->mca_timer.expires - jiffies;
1069         }
1070
1071         if (delay >= resptime)
1072                 delay = prandom_u32() % resptime;
1073
1074         ma->mca_timer.expires = jiffies + delay;
1075         if (!mod_timer(&ma->mca_timer, jiffies + delay))
1076                 atomic_inc(&ma->mca_refcnt);
1077         ma->mca_flags |= MAF_TIMER_RUNNING;
1078 }
1079
1080 /* mark EXCLUDE-mode sources */
1081 static bool mld_xmarksources(struct ifmcaddr6 *pmc, int nsrcs,
1082                              const struct in6_addr *srcs)
1083 {
1084         struct ip6_sf_list *psf;
1085         int i, scount;
1086
1087         scount = 0;
1088         for (psf = pmc->mca_sources; psf; psf = psf->sf_next) {
1089                 if (scount == nsrcs)
1090                         break;
1091                 for (i = 0; i < nsrcs; i++) {
1092                         /* skip inactive filters */
1093                         if (psf->sf_count[MCAST_INCLUDE] ||
1094                             pmc->mca_sfcount[MCAST_EXCLUDE] !=
1095                             psf->sf_count[MCAST_EXCLUDE])
1096                                 break;
1097                         if (ipv6_addr_equal(&srcs[i], &psf->sf_addr)) {
1098                                 scount++;
1099                                 break;
1100                         }
1101                 }
1102         }
1103         pmc->mca_flags &= ~MAF_GSQUERY;
1104         if (scount == nsrcs)    /* all sources excluded */
1105                 return false;
1106         return true;
1107 }
1108
1109 static bool mld_marksources(struct ifmcaddr6 *pmc, int nsrcs,
1110                             const struct in6_addr *srcs)
1111 {
1112         struct ip6_sf_list *psf;
1113         int i, scount;
1114
1115         if (pmc->mca_sfmode == MCAST_EXCLUDE)
1116                 return mld_xmarksources(pmc, nsrcs, srcs);
1117
1118         /* mark INCLUDE-mode sources */
1119
1120         scount = 0;
1121         for (psf = pmc->mca_sources; psf; psf = psf->sf_next) {
1122                 if (scount == nsrcs)
1123                         break;
1124                 for (i = 0; i < nsrcs; i++) {
1125                         if (ipv6_addr_equal(&srcs[i], &psf->sf_addr)) {
1126                                 psf->sf_gsresp = 1;
1127                                 scount++;
1128                                 break;
1129                         }
1130                 }
1131         }
1132         if (!scount) {
1133                 pmc->mca_flags &= ~MAF_GSQUERY;
1134                 return false;
1135         }
1136         pmc->mca_flags |= MAF_GSQUERY;
1137         return true;
1138 }
1139
1140 static int mld_force_mld_version(const struct inet6_dev *idev)
1141 {
1142         /* Normally, both are 0 here. If enforcement to a particular is
1143          * being used, individual device enforcement will have a lower
1144          * precedence over 'all' device (.../conf/all/force_mld_version).
1145          */
1146
1147         if (dev_net(idev->dev)->ipv6.devconf_all->force_mld_version != 0)
1148                 return dev_net(idev->dev)->ipv6.devconf_all->force_mld_version;
1149         else
1150                 return idev->cnf.force_mld_version;
1151 }
1152
1153 static bool mld_in_v2_mode_only(const struct inet6_dev *idev)
1154 {
1155         return mld_force_mld_version(idev) == 2;
1156 }
1157
1158 static bool mld_in_v1_mode_only(const struct inet6_dev *idev)
1159 {
1160         return mld_force_mld_version(idev) == 1;
1161 }
1162
1163 static bool mld_in_v1_mode(const struct inet6_dev *idev)
1164 {
1165         if (mld_in_v2_mode_only(idev))
1166                 return false;
1167         if (mld_in_v1_mode_only(idev))
1168                 return true;
1169         if (idev->mc_v1_seen && time_before(jiffies, idev->mc_v1_seen))
1170                 return true;
1171
1172         return false;
1173 }
1174
1175 static void mld_set_v1_mode(struct inet6_dev *idev)
1176 {
1177         /* RFC3810, relevant sections:
1178          *  - 9.1. Robustness Variable
1179          *  - 9.2. Query Interval
1180          *  - 9.3. Query Response Interval
1181          *  - 9.12. Older Version Querier Present Timeout
1182          */
1183         unsigned long switchback;
1184
1185         switchback = (idev->mc_qrv * idev->mc_qi) + idev->mc_qri;
1186
1187         idev->mc_v1_seen = jiffies + switchback;
1188 }
1189
1190 static void mld_update_qrv(struct inet6_dev *idev,
1191                            const struct mld2_query *mlh2)
1192 {
1193         /* RFC3810, relevant sections:
1194          *  - 5.1.8. QRV (Querier's Robustness Variable)
1195          *  - 9.1. Robustness Variable
1196          */
1197
1198         /* The value of the Robustness Variable MUST NOT be zero,
1199          * and SHOULD NOT be one. Catch this here if we ever run
1200          * into such a case in future.
1201          */
1202         const int min_qrv = min(MLD_QRV_DEFAULT, sysctl_mld_qrv);
1203         WARN_ON(idev->mc_qrv == 0);
1204
1205         if (mlh2->mld2q_qrv > 0)
1206                 idev->mc_qrv = mlh2->mld2q_qrv;
1207
1208         if (unlikely(idev->mc_qrv < min_qrv)) {
1209                 net_warn_ratelimited("IPv6: MLD: clamping QRV from %u to %u!\n",
1210                                      idev->mc_qrv, min_qrv);
1211                 idev->mc_qrv = min_qrv;
1212         }
1213 }
1214
1215 static void mld_update_qi(struct inet6_dev *idev,
1216                           const struct mld2_query *mlh2)
1217 {
1218         /* RFC3810, relevant sections:
1219          *  - 5.1.9. QQIC (Querier's Query Interval Code)
1220          *  - 9.2. Query Interval
1221          *  - 9.12. Older Version Querier Present Timeout
1222          *    (the [Query Interval] in the last Query received)
1223          */
1224         unsigned long mc_qqi;
1225
1226         if (mlh2->mld2q_qqic < 128) {
1227                 mc_qqi = mlh2->mld2q_qqic;
1228         } else {
1229                 unsigned long mc_man, mc_exp;
1230
1231                 mc_exp = MLDV2_QQIC_EXP(mlh2->mld2q_qqic);
1232                 mc_man = MLDV2_QQIC_MAN(mlh2->mld2q_qqic);
1233
1234                 mc_qqi = (mc_man | 0x10) << (mc_exp + 3);
1235         }
1236
1237         idev->mc_qi = mc_qqi * HZ;
1238 }
1239
1240 static void mld_update_qri(struct inet6_dev *idev,
1241                            const struct mld2_query *mlh2)
1242 {
1243         /* RFC3810, relevant sections:
1244          *  - 5.1.3. Maximum Response Code
1245          *  - 9.3. Query Response Interval
1246          */
1247         idev->mc_qri = msecs_to_jiffies(mldv2_mrc(mlh2));
1248 }
1249
1250 static int mld_process_v1(struct inet6_dev *idev, struct mld_msg *mld,
1251                           unsigned long *max_delay, bool v1_query)
1252 {
1253         unsigned long mldv1_md;
1254
1255         /* Ignore v1 queries */
1256         if (mld_in_v2_mode_only(idev))
1257                 return -EINVAL;
1258
1259         mldv1_md = ntohs(mld->mld_maxdelay);
1260
1261         /* When in MLDv1 fallback and a MLDv2 router start-up being
1262          * unaware of current MLDv1 operation, the MRC == MRD mapping
1263          * only works when the exponential algorithm is not being
1264          * used (as MLDv1 is unaware of such things).
1265          *
1266          * According to the RFC author, the MLDv2 implementations
1267          * he's aware of all use a MRC < 32768 on start up queries.
1268          *
1269          * Thus, should we *ever* encounter something else larger
1270          * than that, just assume the maximum possible within our
1271          * reach.
1272          */
1273         if (!v1_query)
1274                 mldv1_md = min(mldv1_md, MLDV1_MRD_MAX_COMPAT);
1275
1276         *max_delay = max(msecs_to_jiffies(mldv1_md), 1UL);
1277
1278         /* MLDv1 router present: we need to go into v1 mode *only*
1279          * when an MLDv1 query is received as per section 9.12. of
1280          * RFC3810! And we know from RFC2710 section 3.7 that MLDv1
1281          * queries MUST be of exactly 24 octets.
1282          */
1283         if (v1_query)
1284                 mld_set_v1_mode(idev);
1285
1286         /* cancel MLDv2 report timer */
1287         mld_gq_stop_timer(idev);
1288         /* cancel the interface change timer */
1289         mld_ifc_stop_timer(idev);
1290         /* clear deleted report items */
1291         mld_clear_delrec(idev);
1292
1293         return 0;
1294 }
1295
1296 static int mld_process_v2(struct inet6_dev *idev, struct mld2_query *mld,
1297                           unsigned long *max_delay)
1298 {
1299         *max_delay = max(msecs_to_jiffies(mldv2_mrc(mld)), 1UL);
1300
1301         mld_update_qrv(idev, mld);
1302         mld_update_qi(idev, mld);
1303         mld_update_qri(idev, mld);
1304
1305         idev->mc_maxdelay = *max_delay;
1306
1307         return 0;
1308 }
1309
1310 /* called with rcu_read_lock() */
1311 int igmp6_event_query(struct sk_buff *skb)
1312 {
1313         struct mld2_query *mlh2 = NULL;
1314         struct ifmcaddr6 *ma;
1315         const struct in6_addr *group;
1316         unsigned long max_delay;
1317         struct inet6_dev *idev;
1318         struct mld_msg *mld;
1319         int group_type;
1320         int mark = 0;
1321         int len, err;
1322
1323         if (!pskb_may_pull(skb, sizeof(struct in6_addr)))
1324                 return -EINVAL;
1325
1326         /* compute payload length excluding extension headers */
1327         len = ntohs(ipv6_hdr(skb)->payload_len) + sizeof(struct ipv6hdr);
1328         len -= skb_network_header_len(skb);
1329
1330         /* RFC3810 6.2
1331          * Upon reception of an MLD message that contains a Query, the node
1332          * checks if the source address of the message is a valid link-local
1333          * address, if the Hop Limit is set to 1, and if the Router Alert
1334          * option is present in the Hop-By-Hop Options header of the IPv6
1335          * packet.  If any of these checks fails, the packet is dropped.
1336          */
1337         if (!(ipv6_addr_type(&ipv6_hdr(skb)->saddr) & IPV6_ADDR_LINKLOCAL) ||
1338             ipv6_hdr(skb)->hop_limit != 1 ||
1339             !(IP6CB(skb)->flags & IP6SKB_ROUTERALERT) ||
1340             IP6CB(skb)->ra != htons(IPV6_OPT_ROUTERALERT_MLD))
1341                 return -EINVAL;
1342
1343         idev = __in6_dev_get(skb->dev);
1344         if (!idev)
1345                 return 0;
1346
1347         mld = (struct mld_msg *)icmp6_hdr(skb);
1348         group = &mld->mld_mca;
1349         group_type = ipv6_addr_type(group);
1350
1351         if (group_type != IPV6_ADDR_ANY &&
1352             !(group_type&IPV6_ADDR_MULTICAST))
1353                 return -EINVAL;
1354
1355         if (len < MLD_V1_QUERY_LEN) {
1356                 return -EINVAL;
1357         } else if (len == MLD_V1_QUERY_LEN || mld_in_v1_mode(idev)) {
1358                 err = mld_process_v1(idev, mld, &max_delay,
1359                                      len == MLD_V1_QUERY_LEN);
1360                 if (err < 0)
1361                         return err;
1362         } else if (len >= MLD_V2_QUERY_LEN_MIN) {
1363                 int srcs_offset = sizeof(struct mld2_query) -
1364                                   sizeof(struct icmp6hdr);
1365
1366                 if (!pskb_may_pull(skb, srcs_offset))
1367                         return -EINVAL;
1368
1369                 mlh2 = (struct mld2_query *)skb_transport_header(skb);
1370
1371                 err = mld_process_v2(idev, mlh2, &max_delay);
1372                 if (err < 0)
1373                         return err;
1374
1375                 if (group_type == IPV6_ADDR_ANY) { /* general query */
1376                         if (mlh2->mld2q_nsrcs)
1377                                 return -EINVAL; /* no sources allowed */
1378
1379                         mld_gq_start_timer(idev);
1380                         return 0;
1381                 }
1382                 /* mark sources to include, if group & source-specific */
1383                 if (mlh2->mld2q_nsrcs != 0) {
1384                         if (!pskb_may_pull(skb, srcs_offset +
1385                             ntohs(mlh2->mld2q_nsrcs) * sizeof(struct in6_addr)))
1386                                 return -EINVAL;
1387
1388                         mlh2 = (struct mld2_query *)skb_transport_header(skb);
1389                         mark = 1;
1390                 }
1391         } else {
1392                 return -EINVAL;
1393         }
1394
1395         read_lock_bh(&idev->lock);
1396         if (group_type == IPV6_ADDR_ANY) {
1397                 for (ma = idev->mc_list; ma; ma = ma->next) {
1398                         spin_lock_bh(&ma->mca_lock);
1399                         igmp6_group_queried(ma, max_delay);
1400                         spin_unlock_bh(&ma->mca_lock);
1401                 }
1402         } else {
1403                 for (ma = idev->mc_list; ma; ma = ma->next) {
1404                         if (!ipv6_addr_equal(group, &ma->mca_addr))
1405                                 continue;
1406                         spin_lock_bh(&ma->mca_lock);
1407                         if (ma->mca_flags & MAF_TIMER_RUNNING) {
1408                                 /* gsquery <- gsquery && mark */
1409                                 if (!mark)
1410                                         ma->mca_flags &= ~MAF_GSQUERY;
1411                         } else {
1412                                 /* gsquery <- mark */
1413                                 if (mark)
1414                                         ma->mca_flags |= MAF_GSQUERY;
1415                                 else
1416                                         ma->mca_flags &= ~MAF_GSQUERY;
1417                         }
1418                         if (!(ma->mca_flags & MAF_GSQUERY) ||
1419                             mld_marksources(ma, ntohs(mlh2->mld2q_nsrcs), mlh2->mld2q_srcs))
1420                                 igmp6_group_queried(ma, max_delay);
1421                         spin_unlock_bh(&ma->mca_lock);
1422                         break;
1423                 }
1424         }
1425         read_unlock_bh(&idev->lock);
1426
1427         return 0;
1428 }
1429
1430 /* called with rcu_read_lock() */
1431 int igmp6_event_report(struct sk_buff *skb)
1432 {
1433         struct ifmcaddr6 *ma;
1434         struct inet6_dev *idev;
1435         struct mld_msg *mld;
1436         int addr_type;
1437
1438         /* Our own report looped back. Ignore it. */
1439         if (skb->pkt_type == PACKET_LOOPBACK)
1440                 return 0;
1441
1442         /* send our report if the MC router may not have heard this report */
1443         if (skb->pkt_type != PACKET_MULTICAST &&
1444             skb->pkt_type != PACKET_BROADCAST)
1445                 return 0;
1446
1447         if (!pskb_may_pull(skb, sizeof(*mld) - sizeof(struct icmp6hdr)))
1448                 return -EINVAL;
1449
1450         mld = (struct mld_msg *)icmp6_hdr(skb);
1451
1452         /* Drop reports with not link local source */
1453         addr_type = ipv6_addr_type(&ipv6_hdr(skb)->saddr);
1454         if (addr_type != IPV6_ADDR_ANY &&
1455             !(addr_type&IPV6_ADDR_LINKLOCAL))
1456                 return -EINVAL;
1457
1458         idev = __in6_dev_get(skb->dev);
1459         if (!idev)
1460                 return -ENODEV;
1461
1462         /*
1463          *      Cancel the timer for this group
1464          */
1465
1466         read_lock_bh(&idev->lock);
1467         for (ma = idev->mc_list; ma; ma = ma->next) {
1468                 if (ipv6_addr_equal(&ma->mca_addr, &mld->mld_mca)) {
1469                         spin_lock(&ma->mca_lock);
1470                         if (del_timer(&ma->mca_timer))
1471                                 atomic_dec(&ma->mca_refcnt);
1472                         ma->mca_flags &= ~(MAF_LAST_REPORTER|MAF_TIMER_RUNNING);
1473                         spin_unlock(&ma->mca_lock);
1474                         break;
1475                 }
1476         }
1477         read_unlock_bh(&idev->lock);
1478         return 0;
1479 }
1480
1481 static bool is_in(struct ifmcaddr6 *pmc, struct ip6_sf_list *psf, int type,
1482                   int gdeleted, int sdeleted)
1483 {
1484         switch (type) {
1485         case MLD2_MODE_IS_INCLUDE:
1486         case MLD2_MODE_IS_EXCLUDE:
1487                 if (gdeleted || sdeleted)
1488                         return false;
1489                 if (!((pmc->mca_flags & MAF_GSQUERY) && !psf->sf_gsresp)) {
1490                         if (pmc->mca_sfmode == MCAST_INCLUDE)
1491                                 return true;
1492                         /* don't include if this source is excluded
1493                          * in all filters
1494                          */
1495                         if (psf->sf_count[MCAST_INCLUDE])
1496                                 return type == MLD2_MODE_IS_INCLUDE;
1497                         return pmc->mca_sfcount[MCAST_EXCLUDE] ==
1498                                 psf->sf_count[MCAST_EXCLUDE];
1499                 }
1500                 return false;
1501         case MLD2_CHANGE_TO_INCLUDE:
1502                 if (gdeleted || sdeleted)
1503                         return false;
1504                 return psf->sf_count[MCAST_INCLUDE] != 0;
1505         case MLD2_CHANGE_TO_EXCLUDE:
1506                 if (gdeleted || sdeleted)
1507                         return false;
1508                 if (pmc->mca_sfcount[MCAST_EXCLUDE] == 0 ||
1509                     psf->sf_count[MCAST_INCLUDE])
1510                         return false;
1511                 return pmc->mca_sfcount[MCAST_EXCLUDE] ==
1512                         psf->sf_count[MCAST_EXCLUDE];
1513         case MLD2_ALLOW_NEW_SOURCES:
1514                 if (gdeleted || !psf->sf_crcount)
1515                         return false;
1516                 return (pmc->mca_sfmode == MCAST_INCLUDE) ^ sdeleted;
1517         case MLD2_BLOCK_OLD_SOURCES:
1518                 if (pmc->mca_sfmode == MCAST_INCLUDE)
1519                         return gdeleted || (psf->sf_crcount && sdeleted);
1520                 return psf->sf_crcount && !gdeleted && !sdeleted;
1521         }
1522         return false;
1523 }
1524
1525 static int
1526 mld_scount(struct ifmcaddr6 *pmc, int type, int gdeleted, int sdeleted)
1527 {
1528         struct ip6_sf_list *psf;
1529         int scount = 0;
1530
1531         for (psf = pmc->mca_sources; psf; psf = psf->sf_next) {
1532                 if (!is_in(pmc, psf, type, gdeleted, sdeleted))
1533                         continue;
1534                 scount++;
1535         }
1536         return scount;
1537 }
1538
1539 static void ip6_mc_hdr(struct sock *sk, struct sk_buff *skb,
1540                        struct net_device *dev,
1541                        const struct in6_addr *saddr,
1542                        const struct in6_addr *daddr,
1543                        int proto, int len)
1544 {
1545         struct ipv6hdr *hdr;
1546
1547         skb->protocol = htons(ETH_P_IPV6);
1548         skb->dev = dev;
1549
1550         skb_reset_network_header(skb);
1551         skb_put(skb, sizeof(struct ipv6hdr));
1552         hdr = ipv6_hdr(skb);
1553
1554         ip6_flow_hdr(hdr, 0, 0);
1555
1556         hdr->payload_len = htons(len);
1557         hdr->nexthdr = proto;
1558         hdr->hop_limit = inet6_sk(sk)->hop_limit;
1559
1560         hdr->saddr = *saddr;
1561         hdr->daddr = *daddr;
1562 }
1563
1564 static struct sk_buff *mld_newpack(struct inet6_dev *idev, unsigned int mtu)
1565 {
1566         struct net_device *dev = idev->dev;
1567         struct net *net = dev_net(dev);
1568         struct sock *sk = net->ipv6.igmp_sk;
1569         struct sk_buff *skb;
1570         struct mld2_report *pmr;
1571         struct in6_addr addr_buf;
1572         const struct in6_addr *saddr;
1573         int hlen = LL_RESERVED_SPACE(dev);
1574         int tlen = dev->needed_tailroom;
1575         unsigned int size = mtu + hlen + tlen;
1576         int err;
1577         u8 ra[8] = { IPPROTO_ICMPV6, 0,
1578                      IPV6_TLV_ROUTERALERT, 2, 0, 0,
1579                      IPV6_TLV_PADN, 0 };
1580
1581         /* we assume size > sizeof(ra) here */
1582         /* limit our allocations to order-0 page */
1583         size = min_t(int, size, SKB_MAX_ORDER(0, 0));
1584         skb = sock_alloc_send_skb(sk, size, 1, &err);
1585
1586         if (!skb)
1587                 return NULL;
1588
1589         skb->priority = TC_PRIO_CONTROL;
1590         skb_reserve(skb, hlen);
1591         skb_tailroom_reserve(skb, mtu, tlen);
1592
1593         if (__ipv6_get_lladdr(idev, &addr_buf, IFA_F_TENTATIVE)) {
1594                 /* <draft-ietf-magma-mld-source-05.txt>:
1595                  * use unspecified address as the source address
1596                  * when a valid link-local address is not available.
1597                  */
1598                 saddr = &in6addr_any;
1599         } else
1600                 saddr = &addr_buf;
1601
1602         ip6_mc_hdr(sk, skb, dev, saddr, &mld2_all_mcr, NEXTHDR_HOP, 0);
1603
1604         memcpy(skb_put(skb, sizeof(ra)), ra, sizeof(ra));
1605
1606         skb_set_transport_header(skb, skb_tail_pointer(skb) - skb->data);
1607         skb_put(skb, sizeof(*pmr));
1608         pmr = (struct mld2_report *)skb_transport_header(skb);
1609         pmr->mld2r_type = ICMPV6_MLD2_REPORT;
1610         pmr->mld2r_resv1 = 0;
1611         pmr->mld2r_cksum = 0;
1612         pmr->mld2r_resv2 = 0;
1613         pmr->mld2r_ngrec = 0;
1614         return skb;
1615 }
1616
1617 static void mld_sendpack(struct sk_buff *skb)
1618 {
1619         struct ipv6hdr *pip6 = ipv6_hdr(skb);
1620         struct mld2_report *pmr =
1621                               (struct mld2_report *)skb_transport_header(skb);
1622         int payload_len, mldlen;
1623         struct inet6_dev *idev;
1624         struct net *net = dev_net(skb->dev);
1625         int err;
1626         struct flowi6 fl6;
1627         struct dst_entry *dst;
1628
1629         rcu_read_lock();
1630         idev = __in6_dev_get(skb->dev);
1631         IP6_UPD_PO_STATS(net, idev, IPSTATS_MIB_OUT, skb->len);
1632
1633         payload_len = (skb_tail_pointer(skb) - skb_network_header(skb)) -
1634                 sizeof(*pip6);
1635         mldlen = skb_tail_pointer(skb) - skb_transport_header(skb);
1636         pip6->payload_len = htons(payload_len);
1637
1638         pmr->mld2r_cksum = csum_ipv6_magic(&pip6->saddr, &pip6->daddr, mldlen,
1639                                            IPPROTO_ICMPV6,
1640                                            csum_partial(skb_transport_header(skb),
1641                                                         mldlen, 0));
1642
1643         icmpv6_flow_init(net->ipv6.igmp_sk, &fl6, ICMPV6_MLD2_REPORT,
1644                          &ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr,
1645                          skb->dev->ifindex);
1646         dst = icmp6_dst_alloc(skb->dev, &fl6);
1647
1648         err = 0;
1649         if (IS_ERR(dst)) {
1650                 err = PTR_ERR(dst);
1651                 dst = NULL;
1652         }
1653         skb_dst_set(skb, dst);
1654         if (err)
1655                 goto err_out;
1656
1657         payload_len = skb->len;
1658
1659         err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT,
1660                       net, net->ipv6.igmp_sk, skb, NULL, skb->dev,
1661                       dst_output);
1662 out:
1663         if (!err) {
1664                 ICMP6MSGOUT_INC_STATS(net, idev, ICMPV6_MLD2_REPORT);
1665                 ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTMSGS);
1666         } else {
1667                 IP6_INC_STATS(net, idev, IPSTATS_MIB_OUTDISCARDS);
1668         }
1669
1670         rcu_read_unlock();
1671         return;
1672
1673 err_out:
1674         kfree_skb(skb);
1675         goto out;
1676 }
1677
1678 static int grec_size(struct ifmcaddr6 *pmc, int type, int gdel, int sdel)
1679 {
1680         return sizeof(struct mld2_grec) + 16 * mld_scount(pmc,type,gdel,sdel);
1681 }
1682
1683 static struct sk_buff *add_grhead(struct sk_buff *skb, struct ifmcaddr6 *pmc,
1684         int type, struct mld2_grec **ppgr)
1685 {
1686         struct net_device *dev = pmc->idev->dev;
1687         struct mld2_report *pmr;
1688         struct mld2_grec *pgr;
1689
1690         if (!skb)
1691                 skb = mld_newpack(pmc->idev, dev->mtu);
1692         if (!skb)
1693                 return NULL;
1694         pgr = (struct mld2_grec *)skb_put(skb, sizeof(struct mld2_grec));
1695         pgr->grec_type = type;
1696         pgr->grec_auxwords = 0;
1697         pgr->grec_nsrcs = 0;
1698         pgr->grec_mca = pmc->mca_addr;  /* structure copy */
1699         pmr = (struct mld2_report *)skb_transport_header(skb);
1700         pmr->mld2r_ngrec = htons(ntohs(pmr->mld2r_ngrec)+1);
1701         *ppgr = pgr;
1702         return skb;
1703 }
1704
1705 #define AVAILABLE(skb)  ((skb) ? skb_availroom(skb) : 0)
1706
1707 static struct sk_buff *add_grec(struct sk_buff *skb, struct ifmcaddr6 *pmc,
1708         int type, int gdeleted, int sdeleted, int crsend)
1709 {
1710         struct inet6_dev *idev = pmc->idev;
1711         struct net_device *dev = idev->dev;
1712         struct mld2_report *pmr;
1713         struct mld2_grec *pgr = NULL;
1714         struct ip6_sf_list *psf, *psf_next, *psf_prev, **psf_list;
1715         int scount, stotal, first, isquery, truncate;
1716
1717         if (pmc->mca_flags & MAF_NOREPORT)
1718                 return skb;
1719
1720         isquery = type == MLD2_MODE_IS_INCLUDE ||
1721                   type == MLD2_MODE_IS_EXCLUDE;
1722         truncate = type == MLD2_MODE_IS_EXCLUDE ||
1723                     type == MLD2_CHANGE_TO_EXCLUDE;
1724
1725         stotal = scount = 0;
1726
1727         psf_list = sdeleted ? &pmc->mca_tomb : &pmc->mca_sources;
1728
1729         if (!*psf_list)
1730                 goto empty_source;
1731
1732         pmr = skb ? (struct mld2_report *)skb_transport_header(skb) : NULL;
1733
1734         /* EX and TO_EX get a fresh packet, if needed */
1735         if (truncate) {
1736                 if (pmr && pmr->mld2r_ngrec &&
1737                     AVAILABLE(skb) < grec_size(pmc, type, gdeleted, sdeleted)) {
1738                         if (skb)
1739                                 mld_sendpack(skb);
1740                         skb = mld_newpack(idev, dev->mtu);
1741                 }
1742         }
1743         first = 1;
1744         psf_prev = NULL;
1745         for (psf = *psf_list; psf; psf = psf_next) {
1746                 struct in6_addr *psrc;
1747
1748                 psf_next = psf->sf_next;
1749
1750                 if (!is_in(pmc, psf, type, gdeleted, sdeleted)) {
1751                         psf_prev = psf;
1752                         continue;
1753                 }
1754
1755                 /* Based on RFC3810 6.1. Should not send source-list change
1756                  * records when there is a filter mode change.
1757                  */
1758                 if (((gdeleted && pmc->mca_sfmode == MCAST_EXCLUDE) ||
1759                      (!gdeleted && pmc->mca_crcount)) &&
1760                     (type == MLD2_ALLOW_NEW_SOURCES ||
1761                      type == MLD2_BLOCK_OLD_SOURCES) && psf->sf_crcount)
1762                         goto decrease_sf_crcount;
1763
1764                 /* clear marks on query responses */
1765                 if (isquery)
1766                         psf->sf_gsresp = 0;
1767
1768                 if (AVAILABLE(skb) < sizeof(*psrc) +
1769                     first*sizeof(struct mld2_grec)) {
1770                         if (truncate && !first)
1771                                 break;   /* truncate these */
1772                         if (pgr)
1773                                 pgr->grec_nsrcs = htons(scount);
1774                         if (skb)
1775                                 mld_sendpack(skb);
1776                         skb = mld_newpack(idev, dev->mtu);
1777                         first = 1;
1778                         scount = 0;
1779                 }
1780                 if (first) {
1781                         skb = add_grhead(skb, pmc, type, &pgr);
1782                         first = 0;
1783                 }
1784                 if (!skb)
1785                         return NULL;
1786                 psrc = (struct in6_addr *)skb_put(skb, sizeof(*psrc));
1787                 *psrc = psf->sf_addr;
1788                 scount++; stotal++;
1789                 if ((type == MLD2_ALLOW_NEW_SOURCES ||
1790                      type == MLD2_BLOCK_OLD_SOURCES) && psf->sf_crcount) {
1791 decrease_sf_crcount:
1792                         psf->sf_crcount--;
1793                         if ((sdeleted || gdeleted) && psf->sf_crcount == 0) {
1794                                 if (psf_prev)
1795                                         psf_prev->sf_next = psf->sf_next;
1796                                 else
1797                                         *psf_list = psf->sf_next;
1798                                 kfree(psf);
1799                                 continue;
1800                         }
1801                 }
1802                 psf_prev = psf;
1803         }
1804
1805 empty_source:
1806         if (!stotal) {
1807                 if (type == MLD2_ALLOW_NEW_SOURCES ||
1808                     type == MLD2_BLOCK_OLD_SOURCES)
1809                         return skb;
1810                 if (pmc->mca_crcount || isquery || crsend) {
1811                         /* make sure we have room for group header */
1812                         if (skb && AVAILABLE(skb) < sizeof(struct mld2_grec)) {
1813                                 mld_sendpack(skb);
1814                                 skb = NULL; /* add_grhead will get a new one */
1815                         }
1816                         skb = add_grhead(skb, pmc, type, &pgr);
1817                 }
1818         }
1819         if (pgr)
1820                 pgr->grec_nsrcs = htons(scount);
1821
1822         if (isquery)
1823                 pmc->mca_flags &= ~MAF_GSQUERY; /* clear query state */
1824         return skb;
1825 }
1826
1827 static void mld_send_report(struct inet6_dev *idev, struct ifmcaddr6 *pmc)
1828 {
1829         struct sk_buff *skb = NULL;
1830         int type;
1831
1832         read_lock_bh(&idev->lock);
1833         if (!pmc) {
1834                 for (pmc = idev->mc_list; pmc; pmc = pmc->next) {
1835                         if (pmc->mca_flags & MAF_NOREPORT)
1836                                 continue;
1837                         spin_lock_bh(&pmc->mca_lock);
1838                         if (pmc->mca_sfcount[MCAST_EXCLUDE])
1839                                 type = MLD2_MODE_IS_EXCLUDE;
1840                         else
1841                                 type = MLD2_MODE_IS_INCLUDE;
1842                         skb = add_grec(skb, pmc, type, 0, 0, 0);
1843                         spin_unlock_bh(&pmc->mca_lock);
1844                 }
1845         } else {
1846                 spin_lock_bh(&pmc->mca_lock);
1847                 if (pmc->mca_sfcount[MCAST_EXCLUDE])
1848                         type = MLD2_MODE_IS_EXCLUDE;
1849                 else
1850                         type = MLD2_MODE_IS_INCLUDE;
1851                 skb = add_grec(skb, pmc, type, 0, 0, 0);
1852                 spin_unlock_bh(&pmc->mca_lock);
1853         }
1854         read_unlock_bh(&idev->lock);
1855         if (skb)
1856                 mld_sendpack(skb);
1857 }
1858
1859 /*
1860  * remove zero-count source records from a source filter list
1861  */
1862 static void mld_clear_zeros(struct ip6_sf_list **ppsf)
1863 {
1864         struct ip6_sf_list *psf_prev, *psf_next, *psf;
1865
1866         psf_prev = NULL;
1867         for (psf = *ppsf; psf; psf = psf_next) {
1868                 psf_next = psf->sf_next;
1869                 if (psf->sf_crcount == 0) {
1870                         if (psf_prev)
1871                                 psf_prev->sf_next = psf->sf_next;
1872                         else
1873                                 *ppsf = psf->sf_next;
1874                         kfree(psf);
1875                 } else
1876                         psf_prev = psf;
1877         }
1878 }
1879
1880 static void mld_send_cr(struct inet6_dev *idev)
1881 {
1882         struct ifmcaddr6 *pmc, *pmc_prev, *pmc_next;
1883         struct sk_buff *skb = NULL;
1884         int type, dtype;
1885
1886         read_lock_bh(&idev->lock);
1887         spin_lock(&idev->mc_lock);
1888
1889         /* deleted MCA's */
1890         pmc_prev = NULL;
1891         for (pmc = idev->mc_tomb; pmc; pmc = pmc_next) {
1892                 pmc_next = pmc->next;
1893                 if (pmc->mca_sfmode == MCAST_INCLUDE) {
1894                         type = MLD2_BLOCK_OLD_SOURCES;
1895                         dtype = MLD2_BLOCK_OLD_SOURCES;
1896                         skb = add_grec(skb, pmc, type, 1, 0, 0);
1897                         skb = add_grec(skb, pmc, dtype, 1, 1, 0);
1898                 }
1899                 if (pmc->mca_crcount) {
1900                         if (pmc->mca_sfmode == MCAST_EXCLUDE) {
1901                                 type = MLD2_CHANGE_TO_INCLUDE;
1902                                 skb = add_grec(skb, pmc, type, 1, 0, 0);
1903                         }
1904                         pmc->mca_crcount--;
1905                         if (pmc->mca_crcount == 0) {
1906                                 mld_clear_zeros(&pmc->mca_tomb);
1907                                 mld_clear_zeros(&pmc->mca_sources);
1908                         }
1909                 }
1910                 if (pmc->mca_crcount == 0 && !pmc->mca_tomb &&
1911                     !pmc->mca_sources) {
1912                         if (pmc_prev)
1913                                 pmc_prev->next = pmc_next;
1914                         else
1915                                 idev->mc_tomb = pmc_next;
1916                         in6_dev_put(pmc->idev);
1917                         kfree(pmc);
1918                 } else
1919                         pmc_prev = pmc;
1920         }
1921         spin_unlock(&idev->mc_lock);
1922
1923         /* change recs */
1924         for (pmc = idev->mc_list; pmc; pmc = pmc->next) {
1925                 spin_lock_bh(&pmc->mca_lock);
1926                 if (pmc->mca_sfcount[MCAST_EXCLUDE]) {
1927                         type = MLD2_BLOCK_OLD_SOURCES;
1928                         dtype = MLD2_ALLOW_NEW_SOURCES;
1929                 } else {
1930                         type = MLD2_ALLOW_NEW_SOURCES;
1931                         dtype = MLD2_BLOCK_OLD_SOURCES;
1932                 }
1933                 skb = add_grec(skb, pmc, type, 0, 0, 0);
1934                 skb = add_grec(skb, pmc, dtype, 0, 1, 0);       /* deleted sources */
1935
1936                 /* filter mode changes */
1937                 if (pmc->mca_crcount) {
1938                         if (pmc->mca_sfmode == MCAST_EXCLUDE)
1939                                 type = MLD2_CHANGE_TO_EXCLUDE;
1940                         else
1941                                 type = MLD2_CHANGE_TO_INCLUDE;
1942                         skb = add_grec(skb, pmc, type, 0, 0, 0);
1943                         pmc->mca_crcount--;
1944                 }
1945                 spin_unlock_bh(&pmc->mca_lock);
1946         }
1947         read_unlock_bh(&idev->lock);
1948         if (!skb)
1949                 return;
1950         (void) mld_sendpack(skb);
1951 }
1952
1953 static void igmp6_send(struct in6_addr *addr, struct net_device *dev, int type)
1954 {
1955         struct net *net = dev_net(dev);
1956         struct sock *sk = net->ipv6.igmp_sk;
1957         struct inet6_dev *idev;
1958         struct sk_buff *skb;
1959         struct mld_msg *hdr;
1960         const struct in6_addr *snd_addr, *saddr;
1961         struct in6_addr addr_buf;
1962         int hlen = LL_RESERVED_SPACE(dev);
1963         int tlen = dev->needed_tailroom;
1964         int err, len, payload_len, full_len;
1965         u8 ra[8] = { IPPROTO_ICMPV6, 0,
1966                      IPV6_TLV_ROUTERALERT, 2, 0, 0,
1967                      IPV6_TLV_PADN, 0 };
1968         struct flowi6 fl6;
1969         struct dst_entry *dst;
1970
1971         if (type == ICMPV6_MGM_REDUCTION)
1972                 snd_addr = &in6addr_linklocal_allrouters;
1973         else
1974                 snd_addr = addr;
1975
1976         len = sizeof(struct icmp6hdr) + sizeof(struct in6_addr);
1977         payload_len = len + sizeof(ra);
1978         full_len = sizeof(struct ipv6hdr) + payload_len;
1979
1980         rcu_read_lock();
1981         IP6_UPD_PO_STATS(net, __in6_dev_get(dev),
1982                       IPSTATS_MIB_OUT, full_len);
1983         rcu_read_unlock();
1984
1985         skb = sock_alloc_send_skb(sk, hlen + tlen + full_len, 1, &err);
1986
1987         if (!skb) {
1988                 rcu_read_lock();
1989                 IP6_INC_STATS(net, __in6_dev_get(dev),
1990                               IPSTATS_MIB_OUTDISCARDS);
1991                 rcu_read_unlock();
1992                 return;
1993         }
1994         skb->priority = TC_PRIO_CONTROL;
1995         skb_reserve(skb, hlen);
1996
1997         if (ipv6_get_lladdr(dev, &addr_buf, IFA_F_TENTATIVE)) {
1998                 /* <draft-ietf-magma-mld-source-05.txt>:
1999                  * use unspecified address as the source address
2000                  * when a valid link-local address is not available.
2001                  */
2002                 saddr = &in6addr_any;
2003         } else
2004                 saddr = &addr_buf;
2005
2006         ip6_mc_hdr(sk, skb, dev, saddr, snd_addr, NEXTHDR_HOP, payload_len);
2007
2008         memcpy(skb_put(skb, sizeof(ra)), ra, sizeof(ra));
2009
2010         hdr = (struct mld_msg *) skb_put(skb, sizeof(struct mld_msg));
2011         memset(hdr, 0, sizeof(struct mld_msg));
2012         hdr->mld_type = type;
2013         hdr->mld_mca = *addr;
2014
2015         hdr->mld_cksum = csum_ipv6_magic(saddr, snd_addr, len,
2016                                          IPPROTO_ICMPV6,
2017                                          csum_partial(hdr, len, 0));
2018
2019         rcu_read_lock();
2020         idev = __in6_dev_get(skb->dev);
2021
2022         icmpv6_flow_init(sk, &fl6, type,
2023                          &ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr,
2024                          skb->dev->ifindex);
2025         dst = icmp6_dst_alloc(skb->dev, &fl6);
2026         if (IS_ERR(dst)) {
2027                 err = PTR_ERR(dst);
2028                 goto err_out;
2029         }
2030
2031         skb_dst_set(skb, dst);
2032         err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT,
2033                       net, sk, skb, NULL, skb->dev,
2034                       dst_output);
2035 out:
2036         if (!err) {
2037                 ICMP6MSGOUT_INC_STATS(net, idev, type);
2038                 ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTMSGS);
2039         } else
2040                 IP6_INC_STATS(net, idev, IPSTATS_MIB_OUTDISCARDS);
2041
2042         rcu_read_unlock();
2043         return;
2044
2045 err_out:
2046         kfree_skb(skb);
2047         goto out;
2048 }
2049
2050 static void mld_send_initial_cr(struct inet6_dev *idev)
2051 {
2052         struct sk_buff *skb;
2053         struct ifmcaddr6 *pmc;
2054         int type;
2055
2056         if (mld_in_v1_mode(idev))
2057                 return;
2058
2059         skb = NULL;
2060         read_lock_bh(&idev->lock);
2061         for (pmc = idev->mc_list; pmc; pmc = pmc->next) {
2062                 spin_lock_bh(&pmc->mca_lock);
2063                 if (pmc->mca_sfcount[MCAST_EXCLUDE])
2064                         type = MLD2_CHANGE_TO_EXCLUDE;
2065                 else
2066                         type = MLD2_CHANGE_TO_INCLUDE;
2067                 skb = add_grec(skb, pmc, type, 0, 0, 1);
2068                 spin_unlock_bh(&pmc->mca_lock);
2069         }
2070         read_unlock_bh(&idev->lock);
2071         if (skb)
2072                 mld_sendpack(skb);
2073 }
2074
2075 void ipv6_mc_dad_complete(struct inet6_dev *idev)
2076 {
2077         idev->mc_dad_count = idev->mc_qrv;
2078         if (idev->mc_dad_count) {
2079                 mld_send_initial_cr(idev);
2080                 idev->mc_dad_count--;
2081                 if (idev->mc_dad_count)
2082                         mld_dad_start_timer(idev, idev->mc_maxdelay);
2083         }
2084 }
2085
2086 static void mld_dad_timer_expire(unsigned long data)
2087 {
2088         struct inet6_dev *idev = (struct inet6_dev *)data;
2089
2090         mld_send_initial_cr(idev);
2091         if (idev->mc_dad_count) {
2092                 idev->mc_dad_count--;
2093                 if (idev->mc_dad_count)
2094                         mld_dad_start_timer(idev, idev->mc_maxdelay);
2095         }
2096         in6_dev_put(idev);
2097 }
2098
2099 static int ip6_mc_del1_src(struct ifmcaddr6 *pmc, int sfmode,
2100         const struct in6_addr *psfsrc)
2101 {
2102         struct ip6_sf_list *psf, *psf_prev;
2103         int rv = 0;
2104
2105         psf_prev = NULL;
2106         for (psf = pmc->mca_sources; psf; psf = psf->sf_next) {
2107                 if (ipv6_addr_equal(&psf->sf_addr, psfsrc))
2108                         break;
2109                 psf_prev = psf;
2110         }
2111         if (!psf || psf->sf_count[sfmode] == 0) {
2112                 /* source filter not found, or count wrong =>  bug */
2113                 return -ESRCH;
2114         }
2115         psf->sf_count[sfmode]--;
2116         if (!psf->sf_count[MCAST_INCLUDE] && !psf->sf_count[MCAST_EXCLUDE]) {
2117                 struct inet6_dev *idev = pmc->idev;
2118
2119                 /* no more filters for this source */
2120                 if (psf_prev)
2121                         psf_prev->sf_next = psf->sf_next;
2122                 else
2123                         pmc->mca_sources = psf->sf_next;
2124                 if (psf->sf_oldin && !(pmc->mca_flags & MAF_NOREPORT) &&
2125                     !mld_in_v1_mode(idev)) {
2126                         psf->sf_crcount = idev->mc_qrv;
2127                         psf->sf_next = pmc->mca_tomb;
2128                         pmc->mca_tomb = psf;
2129                         rv = 1;
2130                 } else
2131                         kfree(psf);
2132         }
2133         return rv;
2134 }
2135
2136 static int ip6_mc_del_src(struct inet6_dev *idev, const struct in6_addr *pmca,
2137                           int sfmode, int sfcount, const struct in6_addr *psfsrc,
2138                           int delta)
2139 {
2140         struct ifmcaddr6 *pmc;
2141         int     changerec = 0;
2142         int     i, err;
2143
2144         if (!idev)
2145                 return -ENODEV;
2146         read_lock_bh(&idev->lock);
2147         for (pmc = idev->mc_list; pmc; pmc = pmc->next) {
2148                 if (ipv6_addr_equal(pmca, &pmc->mca_addr))
2149                         break;
2150         }
2151         if (!pmc) {
2152                 /* MCA not found?? bug */
2153                 read_unlock_bh(&idev->lock);
2154                 return -ESRCH;
2155         }
2156         spin_lock_bh(&pmc->mca_lock);
2157         sf_markstate(pmc);
2158         if (!delta) {
2159                 if (!pmc->mca_sfcount[sfmode]) {
2160                         spin_unlock_bh(&pmc->mca_lock);
2161                         read_unlock_bh(&idev->lock);
2162                         return -EINVAL;
2163                 }
2164                 pmc->mca_sfcount[sfmode]--;
2165         }
2166         err = 0;
2167         for (i = 0; i < sfcount; i++) {
2168                 int rv = ip6_mc_del1_src(pmc, sfmode, &psfsrc[i]);
2169
2170                 changerec |= rv > 0;
2171                 if (!err && rv < 0)
2172                         err = rv;
2173         }
2174         if (pmc->mca_sfmode == MCAST_EXCLUDE &&
2175             pmc->mca_sfcount[MCAST_EXCLUDE] == 0 &&
2176             pmc->mca_sfcount[MCAST_INCLUDE]) {
2177                 struct ip6_sf_list *psf;
2178
2179                 /* filter mode change */
2180                 pmc->mca_sfmode = MCAST_INCLUDE;
2181                 pmc->mca_crcount = idev->mc_qrv;
2182                 idev->mc_ifc_count = pmc->mca_crcount;
2183                 for (psf = pmc->mca_sources; psf; psf = psf->sf_next)
2184                         psf->sf_crcount = 0;
2185                 mld_ifc_event(pmc->idev);
2186         } else if (sf_setstate(pmc) || changerec)
2187                 mld_ifc_event(pmc->idev);
2188         spin_unlock_bh(&pmc->mca_lock);
2189         read_unlock_bh(&idev->lock);
2190         return err;
2191 }
2192
2193 /*
2194  * Add multicast single-source filter to the interface list
2195  */
2196 static int ip6_mc_add1_src(struct ifmcaddr6 *pmc, int sfmode,
2197         const struct in6_addr *psfsrc)
2198 {
2199         struct ip6_sf_list *psf, *psf_prev;
2200
2201         psf_prev = NULL;
2202         for (psf = pmc->mca_sources; psf; psf = psf->sf_next) {
2203                 if (ipv6_addr_equal(&psf->sf_addr, psfsrc))
2204                         break;
2205                 psf_prev = psf;
2206         }
2207         if (!psf) {
2208                 psf = kzalloc(sizeof(*psf), GFP_ATOMIC);
2209                 if (!psf)
2210                         return -ENOBUFS;
2211
2212                 psf->sf_addr = *psfsrc;
2213                 if (psf_prev) {
2214                         psf_prev->sf_next = psf;
2215                 } else
2216                         pmc->mca_sources = psf;
2217         }
2218         psf->sf_count[sfmode]++;
2219         return 0;
2220 }
2221
2222 static void sf_markstate(struct ifmcaddr6 *pmc)
2223 {
2224         struct ip6_sf_list *psf;
2225         int mca_xcount = pmc->mca_sfcount[MCAST_EXCLUDE];
2226
2227         for (psf = pmc->mca_sources; psf; psf = psf->sf_next)
2228                 if (pmc->mca_sfcount[MCAST_EXCLUDE]) {
2229                         psf->sf_oldin = mca_xcount ==
2230                                 psf->sf_count[MCAST_EXCLUDE] &&
2231                                 !psf->sf_count[MCAST_INCLUDE];
2232                 } else
2233                         psf->sf_oldin = psf->sf_count[MCAST_INCLUDE] != 0;
2234 }
2235
2236 static int sf_setstate(struct ifmcaddr6 *pmc)
2237 {
2238         struct ip6_sf_list *psf, *dpsf;
2239         int mca_xcount = pmc->mca_sfcount[MCAST_EXCLUDE];
2240         int qrv = pmc->idev->mc_qrv;
2241         int new_in, rv;
2242
2243         rv = 0;
2244         for (psf = pmc->mca_sources; psf; psf = psf->sf_next) {
2245                 if (pmc->mca_sfcount[MCAST_EXCLUDE]) {
2246                         new_in = mca_xcount == psf->sf_count[MCAST_EXCLUDE] &&
2247                                 !psf->sf_count[MCAST_INCLUDE];
2248                 } else
2249                         new_in = psf->sf_count[MCAST_INCLUDE] != 0;
2250                 if (new_in) {
2251                         if (!psf->sf_oldin) {
2252                                 struct ip6_sf_list *prev = NULL;
2253
2254                                 for (dpsf = pmc->mca_tomb; dpsf;
2255                                      dpsf = dpsf->sf_next) {
2256                                         if (ipv6_addr_equal(&dpsf->sf_addr,
2257                                             &psf->sf_addr))
2258                                                 break;
2259                                         prev = dpsf;
2260                                 }
2261                                 if (dpsf) {
2262                                         if (prev)
2263                                                 prev->sf_next = dpsf->sf_next;
2264                                         else
2265                                                 pmc->mca_tomb = dpsf->sf_next;
2266                                         kfree(dpsf);
2267                                 }
2268                                 psf->sf_crcount = qrv;
2269                                 rv++;
2270                         }
2271                 } else if (psf->sf_oldin) {
2272                         psf->sf_crcount = 0;
2273                         /*
2274                          * add or update "delete" records if an active filter
2275                          * is now inactive
2276                          */
2277                         for (dpsf = pmc->mca_tomb; dpsf; dpsf = dpsf->sf_next)
2278                                 if (ipv6_addr_equal(&dpsf->sf_addr,
2279                                     &psf->sf_addr))
2280                                         break;
2281                         if (!dpsf) {
2282                                 dpsf = kmalloc(sizeof(*dpsf), GFP_ATOMIC);
2283                                 if (!dpsf)
2284                                         continue;
2285                                 *dpsf = *psf;
2286                                 /* pmc->mca_lock held by callers */
2287                                 dpsf->sf_next = pmc->mca_tomb;
2288                                 pmc->mca_tomb = dpsf;
2289                         }
2290                         dpsf->sf_crcount = qrv;
2291                         rv++;
2292                 }
2293         }
2294         return rv;
2295 }
2296
2297 /*
2298  * Add multicast source filter list to the interface list
2299  */
2300 static int ip6_mc_add_src(struct inet6_dev *idev, const struct in6_addr *pmca,
2301                           int sfmode, int sfcount, const struct in6_addr *psfsrc,
2302                           int delta)
2303 {
2304         struct ifmcaddr6 *pmc;
2305         int     isexclude;
2306         int     i, err;
2307
2308         if (!idev)
2309                 return -ENODEV;
2310         read_lock_bh(&idev->lock);
2311         for (pmc = idev->mc_list; pmc; pmc = pmc->next) {
2312                 if (ipv6_addr_equal(pmca, &pmc->mca_addr))
2313                         break;
2314         }
2315         if (!pmc) {
2316                 /* MCA not found?? bug */
2317                 read_unlock_bh(&idev->lock);
2318                 return -ESRCH;
2319         }
2320         spin_lock_bh(&pmc->mca_lock);
2321
2322         sf_markstate(pmc);
2323         isexclude = pmc->mca_sfmode == MCAST_EXCLUDE;
2324         if (!delta)
2325                 pmc->mca_sfcount[sfmode]++;
2326         err = 0;
2327         for (i = 0; i < sfcount; i++) {
2328                 err = ip6_mc_add1_src(pmc, sfmode, &psfsrc[i]);
2329                 if (err)
2330                         break;
2331         }
2332         if (err) {
2333                 int j;
2334
2335                 if (!delta)
2336                         pmc->mca_sfcount[sfmode]--;
2337                 for (j = 0; j < i; j++)
2338                         ip6_mc_del1_src(pmc, sfmode, &psfsrc[j]);
2339         } else if (isexclude != (pmc->mca_sfcount[MCAST_EXCLUDE] != 0)) {
2340                 struct ip6_sf_list *psf;
2341
2342                 /* filter mode change */
2343                 if (pmc->mca_sfcount[MCAST_EXCLUDE])
2344                         pmc->mca_sfmode = MCAST_EXCLUDE;
2345                 else if (pmc->mca_sfcount[MCAST_INCLUDE])
2346                         pmc->mca_sfmode = MCAST_INCLUDE;
2347                 /* else no filters; keep old mode for reports */
2348
2349                 pmc->mca_crcount = idev->mc_qrv;
2350                 idev->mc_ifc_count = pmc->mca_crcount;
2351                 for (psf = pmc->mca_sources; psf; psf = psf->sf_next)
2352                         psf->sf_crcount = 0;
2353                 mld_ifc_event(idev);
2354         } else if (sf_setstate(pmc))
2355                 mld_ifc_event(idev);
2356         spin_unlock_bh(&pmc->mca_lock);
2357         read_unlock_bh(&idev->lock);
2358         return err;
2359 }
2360
2361 static void ip6_mc_clear_src(struct ifmcaddr6 *pmc)
2362 {
2363         struct ip6_sf_list *psf, *nextpsf;
2364
2365         for (psf = pmc->mca_tomb; psf; psf = nextpsf) {
2366                 nextpsf = psf->sf_next;
2367                 kfree(psf);
2368         }
2369         pmc->mca_tomb = NULL;
2370         for (psf = pmc->mca_sources; psf; psf = nextpsf) {
2371                 nextpsf = psf->sf_next;
2372                 kfree(psf);
2373         }
2374         pmc->mca_sources = NULL;
2375         pmc->mca_sfmode = MCAST_EXCLUDE;
2376         pmc->mca_sfcount[MCAST_INCLUDE] = 0;
2377         pmc->mca_sfcount[MCAST_EXCLUDE] = 1;
2378 }
2379
2380
2381 static void igmp6_join_group(struct ifmcaddr6 *ma)
2382 {
2383         unsigned long delay;
2384
2385         if (ma->mca_flags & MAF_NOREPORT)
2386                 return;
2387
2388         igmp6_send(&ma->mca_addr, ma->idev->dev, ICMPV6_MGM_REPORT);
2389
2390         delay = prandom_u32() % unsolicited_report_interval(ma->idev);
2391
2392         spin_lock_bh(&ma->mca_lock);
2393         if (del_timer(&ma->mca_timer)) {
2394                 atomic_dec(&ma->mca_refcnt);
2395                 delay = ma->mca_timer.expires - jiffies;
2396         }
2397
2398         if (!mod_timer(&ma->mca_timer, jiffies + delay))
2399                 atomic_inc(&ma->mca_refcnt);
2400         ma->mca_flags |= MAF_TIMER_RUNNING | MAF_LAST_REPORTER;
2401         spin_unlock_bh(&ma->mca_lock);
2402 }
2403
2404 static int ip6_mc_leave_src(struct sock *sk, struct ipv6_mc_socklist *iml,
2405                             struct inet6_dev *idev)
2406 {
2407         int err;
2408
2409         /* callers have the socket lock and rtnl lock
2410          * so no other readers or writers of iml or its sflist
2411          */
2412         if (!iml->sflist) {
2413                 /* any-source empty exclude case */
2414                 return ip6_mc_del_src(idev, &iml->addr, iml->sfmode, 0, NULL, 0);
2415         }
2416         err = ip6_mc_del_src(idev, &iml->addr, iml->sfmode,
2417                 iml->sflist->sl_count, iml->sflist->sl_addr, 0);
2418         sock_kfree_s(sk, iml->sflist, IP6_SFLSIZE(iml->sflist->sl_max));
2419         iml->sflist = NULL;
2420         return err;
2421 }
2422
2423 static void igmp6_leave_group(struct ifmcaddr6 *ma)
2424 {
2425         if (mld_in_v1_mode(ma->idev)) {
2426                 if (ma->mca_flags & MAF_LAST_REPORTER)
2427                         igmp6_send(&ma->mca_addr, ma->idev->dev,
2428                                 ICMPV6_MGM_REDUCTION);
2429         } else {
2430                 mld_add_delrec(ma->idev, ma);
2431                 mld_ifc_event(ma->idev);
2432         }
2433 }
2434
2435 static void mld_gq_timer_expire(unsigned long data)
2436 {
2437         struct inet6_dev *idev = (struct inet6_dev *)data;
2438
2439         idev->mc_gq_running = 0;
2440         mld_send_report(idev, NULL);
2441         in6_dev_put(idev);
2442 }
2443
2444 static void mld_ifc_timer_expire(unsigned long data)
2445 {
2446         struct inet6_dev *idev = (struct inet6_dev *)data;
2447
2448         mld_send_cr(idev);
2449         if (idev->mc_ifc_count) {
2450                 idev->mc_ifc_count--;
2451                 if (idev->mc_ifc_count)
2452                         mld_ifc_start_timer(idev, idev->mc_maxdelay);
2453         }
2454         in6_dev_put(idev);
2455 }
2456
2457 static void mld_ifc_event(struct inet6_dev *idev)
2458 {
2459         if (mld_in_v1_mode(idev))
2460                 return;
2461         idev->mc_ifc_count = idev->mc_qrv;
2462         mld_ifc_start_timer(idev, 1);
2463 }
2464
2465
2466 static void igmp6_timer_handler(unsigned long data)
2467 {
2468         struct ifmcaddr6 *ma = (struct ifmcaddr6 *) data;
2469
2470         if (mld_in_v1_mode(ma->idev))
2471                 igmp6_send(&ma->mca_addr, ma->idev->dev, ICMPV6_MGM_REPORT);
2472         else
2473                 mld_send_report(ma->idev, ma);
2474
2475         spin_lock(&ma->mca_lock);
2476         ma->mca_flags |=  MAF_LAST_REPORTER;
2477         ma->mca_flags &= ~MAF_TIMER_RUNNING;
2478         spin_unlock(&ma->mca_lock);
2479         ma_put(ma);
2480 }
2481
2482 /* Device changing type */
2483
2484 void ipv6_mc_unmap(struct inet6_dev *idev)
2485 {
2486         struct ifmcaddr6 *i;
2487
2488         /* Install multicast list, except for all-nodes (already installed) */
2489
2490         read_lock_bh(&idev->lock);
2491         for (i = idev->mc_list; i; i = i->next)
2492                 igmp6_group_dropped(i);
2493         read_unlock_bh(&idev->lock);
2494 }
2495
2496 void ipv6_mc_remap(struct inet6_dev *idev)
2497 {
2498         ipv6_mc_up(idev);
2499 }
2500
2501 /* Device going down */
2502
2503 void ipv6_mc_down(struct inet6_dev *idev)
2504 {
2505         struct ifmcaddr6 *i;
2506
2507         /* Withdraw multicast list */
2508
2509         read_lock_bh(&idev->lock);
2510
2511         for (i = idev->mc_list; i; i = i->next)
2512                 igmp6_group_dropped(i);
2513
2514         /* Should stop timer after group drop. or we will
2515          * start timer again in mld_ifc_event()
2516          */
2517         mld_ifc_stop_timer(idev);
2518         mld_gq_stop_timer(idev);
2519         mld_dad_stop_timer(idev);
2520         read_unlock_bh(&idev->lock);
2521 }
2522
2523 static void ipv6_mc_reset(struct inet6_dev *idev)
2524 {
2525         idev->mc_qrv = sysctl_mld_qrv;
2526         idev->mc_qi = MLD_QI_DEFAULT;
2527         idev->mc_qri = MLD_QRI_DEFAULT;
2528         idev->mc_v1_seen = 0;
2529         idev->mc_maxdelay = unsolicited_report_interval(idev);
2530 }
2531
2532 /* Device going up */
2533
2534 void ipv6_mc_up(struct inet6_dev *idev)
2535 {
2536         struct ifmcaddr6 *i;
2537
2538         /* Install multicast list, except for all-nodes (already installed) */
2539
2540         read_lock_bh(&idev->lock);
2541         ipv6_mc_reset(idev);
2542         for (i = idev->mc_list; i; i = i->next) {
2543                 mld_del_delrec(idev, i);
2544                 igmp6_group_added(i);
2545         }
2546         read_unlock_bh(&idev->lock);
2547 }
2548
2549 /* IPv6 device initialization. */
2550
2551 void ipv6_mc_init_dev(struct inet6_dev *idev)
2552 {
2553         write_lock_bh(&idev->lock);
2554         spin_lock_init(&idev->mc_lock);
2555         idev->mc_gq_running = 0;
2556         setup_timer(&idev->mc_gq_timer, mld_gq_timer_expire,
2557                         (unsigned long)idev);
2558         idev->mc_tomb = NULL;
2559         idev->mc_ifc_count = 0;
2560         setup_timer(&idev->mc_ifc_timer, mld_ifc_timer_expire,
2561                         (unsigned long)idev);
2562         setup_timer(&idev->mc_dad_timer, mld_dad_timer_expire,
2563                     (unsigned long)idev);
2564         ipv6_mc_reset(idev);
2565         write_unlock_bh(&idev->lock);
2566 }
2567
2568 /*
2569  *      Device is about to be destroyed: clean up.
2570  */
2571
2572 void ipv6_mc_destroy_dev(struct inet6_dev *idev)
2573 {
2574         struct ifmcaddr6 *i;
2575
2576         /* Deactivate timers */
2577         ipv6_mc_down(idev);
2578         mld_clear_delrec(idev);
2579
2580         /* Delete all-nodes address. */
2581         /* We cannot call ipv6_dev_mc_dec() directly, our caller in
2582          * addrconf.c has NULL'd out dev->ip6_ptr so in6_dev_get() will
2583          * fail.
2584          */
2585         __ipv6_dev_mc_dec(idev, &in6addr_linklocal_allnodes);
2586
2587         if (idev->cnf.forwarding)
2588                 __ipv6_dev_mc_dec(idev, &in6addr_linklocal_allrouters);
2589
2590         write_lock_bh(&idev->lock);
2591         while ((i = idev->mc_list) != NULL) {
2592                 idev->mc_list = i->next;
2593
2594                 write_unlock_bh(&idev->lock);
2595                 ma_put(i);
2596                 write_lock_bh(&idev->lock);
2597         }
2598         write_unlock_bh(&idev->lock);
2599 }
2600
2601 #ifdef CONFIG_PROC_FS
2602 struct igmp6_mc_iter_state {
2603         struct seq_net_private p;
2604         struct net_device *dev;
2605         struct inet6_dev *idev;
2606 };
2607
2608 #define igmp6_mc_seq_private(seq)       ((struct igmp6_mc_iter_state *)(seq)->private)
2609
2610 static inline struct ifmcaddr6 *igmp6_mc_get_first(struct seq_file *seq)
2611 {
2612         struct ifmcaddr6 *im = NULL;
2613         struct igmp6_mc_iter_state *state = igmp6_mc_seq_private(seq);
2614         struct net *net = seq_file_net(seq);
2615
2616         state->idev = NULL;
2617         for_each_netdev_rcu(net, state->dev) {
2618                 struct inet6_dev *idev;
2619                 idev = __in6_dev_get(state->dev);
2620                 if (!idev)
2621                         continue;
2622                 read_lock_bh(&idev->lock);
2623                 im = idev->mc_list;
2624                 if (im) {
2625                         state->idev = idev;
2626                         break;
2627                 }
2628                 read_unlock_bh(&idev->lock);
2629         }
2630         return im;
2631 }
2632
2633 static struct ifmcaddr6 *igmp6_mc_get_next(struct seq_file *seq, struct ifmcaddr6 *im)
2634 {
2635         struct igmp6_mc_iter_state *state = igmp6_mc_seq_private(seq);
2636
2637         im = im->next;
2638         while (!im) {
2639                 if (likely(state->idev))
2640                         read_unlock_bh(&state->idev->lock);
2641
2642                 state->dev = next_net_device_rcu(state->dev);
2643                 if (!state->dev) {
2644                         state->idev = NULL;
2645                         break;
2646                 }
2647                 state->idev = __in6_dev_get(state->dev);
2648                 if (!state->idev)
2649                         continue;
2650                 read_lock_bh(&state->idev->lock);
2651                 im = state->idev->mc_list;
2652         }
2653         return im;
2654 }
2655
2656 static struct ifmcaddr6 *igmp6_mc_get_idx(struct seq_file *seq, loff_t pos)
2657 {
2658         struct ifmcaddr6 *im = igmp6_mc_get_first(seq);
2659         if (im)
2660                 while (pos && (im = igmp6_mc_get_next(seq, im)) != NULL)
2661                         --pos;
2662         return pos ? NULL : im;
2663 }
2664
2665 static void *igmp6_mc_seq_start(struct seq_file *seq, loff_t *pos)
2666         __acquires(RCU)
2667 {
2668         rcu_read_lock();
2669         return igmp6_mc_get_idx(seq, *pos);
2670 }
2671
2672 static void *igmp6_mc_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2673 {
2674         struct ifmcaddr6 *im = igmp6_mc_get_next(seq, v);
2675
2676         ++*pos;
2677         return im;
2678 }
2679
2680 static void igmp6_mc_seq_stop(struct seq_file *seq, void *v)
2681         __releases(RCU)
2682 {
2683         struct igmp6_mc_iter_state *state = igmp6_mc_seq_private(seq);
2684
2685         if (likely(state->idev)) {
2686                 read_unlock_bh(&state->idev->lock);
2687                 state->idev = NULL;
2688         }
2689         state->dev = NULL;
2690         rcu_read_unlock();
2691 }
2692
2693 static int igmp6_mc_seq_show(struct seq_file *seq, void *v)
2694 {
2695         struct ifmcaddr6 *im = (struct ifmcaddr6 *)v;
2696         struct igmp6_mc_iter_state *state = igmp6_mc_seq_private(seq);
2697
2698         seq_printf(seq,
2699                    "%-4d %-15s %pi6 %5d %08X %ld\n",
2700                    state->dev->ifindex, state->dev->name,
2701                    &im->mca_addr,
2702                    im->mca_users, im->mca_flags,
2703                    (im->mca_flags&MAF_TIMER_RUNNING) ?
2704                    jiffies_to_clock_t(im->mca_timer.expires-jiffies) : 0);
2705         return 0;
2706 }
2707
2708 static const struct seq_operations igmp6_mc_seq_ops = {
2709         .start  =       igmp6_mc_seq_start,
2710         .next   =       igmp6_mc_seq_next,
2711         .stop   =       igmp6_mc_seq_stop,
2712         .show   =       igmp6_mc_seq_show,
2713 };
2714
2715 static int igmp6_mc_seq_open(struct inode *inode, struct file *file)
2716 {
2717         return seq_open_net(inode, file, &igmp6_mc_seq_ops,
2718                             sizeof(struct igmp6_mc_iter_state));
2719 }
2720
2721 static const struct file_operations igmp6_mc_seq_fops = {
2722         .owner          =       THIS_MODULE,
2723         .open           =       igmp6_mc_seq_open,
2724         .read           =       seq_read,
2725         .llseek         =       seq_lseek,
2726         .release        =       seq_release_net,
2727 };
2728
2729 struct igmp6_mcf_iter_state {
2730         struct seq_net_private p;
2731         struct net_device *dev;
2732         struct inet6_dev *idev;
2733         struct ifmcaddr6 *im;
2734 };
2735
2736 #define igmp6_mcf_seq_private(seq)      ((struct igmp6_mcf_iter_state *)(seq)->private)
2737
2738 static inline struct ip6_sf_list *igmp6_mcf_get_first(struct seq_file *seq)
2739 {
2740         struct ip6_sf_list *psf = NULL;
2741         struct ifmcaddr6 *im = NULL;
2742         struct igmp6_mcf_iter_state *state = igmp6_mcf_seq_private(seq);
2743         struct net *net = seq_file_net(seq);
2744
2745         state->idev = NULL;
2746         state->im = NULL;
2747         for_each_netdev_rcu(net, state->dev) {
2748                 struct inet6_dev *idev;
2749                 idev = __in6_dev_get(state->dev);
2750                 if (unlikely(idev == NULL))
2751                         continue;
2752                 read_lock_bh(&idev->lock);
2753                 im = idev->mc_list;
2754                 if (likely(im)) {
2755                         spin_lock_bh(&im->mca_lock);
2756                         psf = im->mca_sources;
2757                         if (likely(psf)) {
2758                                 state->im = im;
2759                                 state->idev = idev;
2760                                 break;
2761                         }
2762                         spin_unlock_bh(&im->mca_lock);
2763                 }
2764                 read_unlock_bh(&idev->lock);
2765         }
2766         return psf;
2767 }
2768
2769 static struct ip6_sf_list *igmp6_mcf_get_next(struct seq_file *seq, struct ip6_sf_list *psf)
2770 {
2771         struct igmp6_mcf_iter_state *state = igmp6_mcf_seq_private(seq);
2772
2773         psf = psf->sf_next;
2774         while (!psf) {
2775                 spin_unlock_bh(&state->im->mca_lock);
2776                 state->im = state->im->next;
2777                 while (!state->im) {
2778                         if (likely(state->idev))
2779                                 read_unlock_bh(&state->idev->lock);
2780
2781                         state->dev = next_net_device_rcu(state->dev);
2782                         if (!state->dev) {
2783                                 state->idev = NULL;
2784                                 goto out;
2785                         }
2786                         state->idev = __in6_dev_get(state->dev);
2787                         if (!state->idev)
2788                                 continue;
2789                         read_lock_bh(&state->idev->lock);
2790                         state->im = state->idev->mc_list;
2791                 }
2792                 if (!state->im)
2793                         break;
2794                 spin_lock_bh(&state->im->mca_lock);
2795                 psf = state->im->mca_sources;
2796         }
2797 out:
2798         return psf;
2799 }
2800
2801 static struct ip6_sf_list *igmp6_mcf_get_idx(struct seq_file *seq, loff_t pos)
2802 {
2803         struct ip6_sf_list *psf = igmp6_mcf_get_first(seq);
2804         if (psf)
2805                 while (pos && (psf = igmp6_mcf_get_next(seq, psf)) != NULL)
2806                         --pos;
2807         return pos ? NULL : psf;
2808 }
2809
2810 static void *igmp6_mcf_seq_start(struct seq_file *seq, loff_t *pos)
2811         __acquires(RCU)
2812 {
2813         rcu_read_lock();
2814         return *pos ? igmp6_mcf_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
2815 }
2816
2817 static void *igmp6_mcf_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2818 {
2819         struct ip6_sf_list *psf;
2820         if (v == SEQ_START_TOKEN)
2821                 psf = igmp6_mcf_get_first(seq);
2822         else
2823                 psf = igmp6_mcf_get_next(seq, v);
2824         ++*pos;
2825         return psf;
2826 }
2827
2828 static void igmp6_mcf_seq_stop(struct seq_file *seq, void *v)
2829         __releases(RCU)
2830 {
2831         struct igmp6_mcf_iter_state *state = igmp6_mcf_seq_private(seq);
2832         if (likely(state->im)) {
2833                 spin_unlock_bh(&state->im->mca_lock);
2834                 state->im = NULL;
2835         }
2836         if (likely(state->idev)) {
2837                 read_unlock_bh(&state->idev->lock);
2838                 state->idev = NULL;
2839         }
2840         state->dev = NULL;
2841         rcu_read_unlock();
2842 }
2843
2844 static int igmp6_mcf_seq_show(struct seq_file *seq, void *v)
2845 {
2846         struct ip6_sf_list *psf = (struct ip6_sf_list *)v;
2847         struct igmp6_mcf_iter_state *state = igmp6_mcf_seq_private(seq);
2848
2849         if (v == SEQ_START_TOKEN) {
2850                 seq_puts(seq, "Idx Device                Multicast Address                   Source Address    INC    EXC\n");
2851         } else {
2852                 seq_printf(seq,
2853                            "%3d %6.6s %pi6 %pi6 %6lu %6lu\n",
2854                            state->dev->ifindex, state->dev->name,
2855                            &state->im->mca_addr,
2856                            &psf->sf_addr,
2857                            psf->sf_count[MCAST_INCLUDE],
2858                            psf->sf_count[MCAST_EXCLUDE]);
2859         }
2860         return 0;
2861 }
2862
2863 static const struct seq_operations igmp6_mcf_seq_ops = {
2864         .start  =       igmp6_mcf_seq_start,
2865         .next   =       igmp6_mcf_seq_next,
2866         .stop   =       igmp6_mcf_seq_stop,
2867         .show   =       igmp6_mcf_seq_show,
2868 };
2869
2870 static int igmp6_mcf_seq_open(struct inode *inode, struct file *file)
2871 {
2872         return seq_open_net(inode, file, &igmp6_mcf_seq_ops,
2873                             sizeof(struct igmp6_mcf_iter_state));
2874 }
2875
2876 static const struct file_operations igmp6_mcf_seq_fops = {
2877         .owner          =       THIS_MODULE,
2878         .open           =       igmp6_mcf_seq_open,
2879         .read           =       seq_read,
2880         .llseek         =       seq_lseek,
2881         .release        =       seq_release_net,
2882 };
2883
2884 static int __net_init igmp6_proc_init(struct net *net)
2885 {
2886         int err;
2887
2888         err = -ENOMEM;
2889         if (!proc_create("igmp6", S_IRUGO, net->proc_net, &igmp6_mc_seq_fops))
2890                 goto out;
2891         if (!proc_create("mcfilter6", S_IRUGO, net->proc_net,
2892                          &igmp6_mcf_seq_fops))
2893                 goto out_proc_net_igmp6;
2894
2895         err = 0;
2896 out:
2897         return err;
2898
2899 out_proc_net_igmp6:
2900         remove_proc_entry("igmp6", net->proc_net);
2901         goto out;
2902 }
2903
2904 static void __net_exit igmp6_proc_exit(struct net *net)
2905 {
2906         remove_proc_entry("mcfilter6", net->proc_net);
2907         remove_proc_entry("igmp6", net->proc_net);
2908 }
2909 #else
2910 static inline int igmp6_proc_init(struct net *net)
2911 {
2912         return 0;
2913 }
2914 static inline void igmp6_proc_exit(struct net *net)
2915 {
2916 }
2917 #endif
2918
2919 static int __net_init igmp6_net_init(struct net *net)
2920 {
2921         int err;
2922
2923         err = inet_ctl_sock_create(&net->ipv6.igmp_sk, PF_INET6,
2924                                    SOCK_RAW, IPPROTO_ICMPV6, net);
2925         if (err < 0) {
2926                 pr_err("Failed to initialize the IGMP6 control socket (err %d)\n",
2927                        err);
2928                 goto out;
2929         }
2930
2931         inet6_sk(net->ipv6.igmp_sk)->hop_limit = 1;
2932
2933         err = inet_ctl_sock_create(&net->ipv6.mc_autojoin_sk, PF_INET6,
2934                                    SOCK_RAW, IPPROTO_ICMPV6, net);
2935         if (err < 0) {
2936                 pr_err("Failed to initialize the IGMP6 autojoin socket (err %d)\n",
2937                        err);
2938                 goto out_sock_create;
2939         }
2940
2941         err = igmp6_proc_init(net);
2942         if (err)
2943                 goto out_sock_create_autojoin;
2944
2945         return 0;
2946
2947 out_sock_create_autojoin:
2948         inet_ctl_sock_destroy(net->ipv6.mc_autojoin_sk);
2949 out_sock_create:
2950         inet_ctl_sock_destroy(net->ipv6.igmp_sk);
2951 out:
2952         return err;
2953 }
2954
2955 static void __net_exit igmp6_net_exit(struct net *net)
2956 {
2957         inet_ctl_sock_destroy(net->ipv6.igmp_sk);
2958         inet_ctl_sock_destroy(net->ipv6.mc_autojoin_sk);
2959         igmp6_proc_exit(net);
2960 }
2961
2962 static struct pernet_operations igmp6_net_ops = {
2963         .init = igmp6_net_init,
2964         .exit = igmp6_net_exit,
2965 };
2966
2967 int __init igmp6_init(void)
2968 {
2969         return register_pernet_subsys(&igmp6_net_ops);
2970 }
2971
2972 void igmp6_cleanup(void)
2973 {
2974         unregister_pernet_subsys(&igmp6_net_ops);
2975 }