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                 if (im->mca_sfmode == MCAST_INCLUDE) {
 775                         swap(im->mca_tomb, pmc->mca_tomb);
 776                         swap(im->mca_sources, pmc->mca_sources);
 777                         for (psf = im->mca_sources; psf; psf = psf->sf_next)
 778                                 psf->sf_crcount = im->mca_crcount;
 779                 }
 780                 in6_dev_put(pmc->idev);
 781                 ip6_mc_clear_src(pmc);
 782                 kfree(pmc);
 783         }
 784         spin_unlock_bh(&im->mca_lock);
 785 }
 786
 787 static void mld_clear_delrec(struct inet6_dev *idev)
 788 {
 789         struct ifmcaddr6 *pmc, *nextpmc;
 790
 791         spin_lock_bh(&idev->mc_lock);
 792         pmc = idev->mc_tomb;
 793         idev->mc_tomb = NULL;
 794         spin_unlock_bh(&idev->mc_lock);
 795
 796         for (; pmc; pmc = nextpmc) {
 797                 nextpmc = pmc->next;
 798                 ip6_mc_clear_src(pmc);
 799                 in6_dev_put(pmc->idev);
 800                 kfree(pmc);
 801         }
 802
 803         /* clear dead sources, too */
 804         read_lock_bh(&idev->lock);
 805         for (pmc = idev->mc_list; pmc; pmc = pmc->next) {
 806                 struct ip6_sf_list *psf, *psf_next;
 807
 808                 spin_lock_bh(&pmc->mca_lock);
 809                 psf = pmc->mca_tomb;
 810                 pmc->mca_tomb = NULL;
 811                 spin_unlock_bh(&pmc->mca_lock);
 812                 for (; psf; psf = psf_next) {
 813                         psf_next = psf->sf_next;
 814                         kfree(psf);
 815                 }
 816         }
 817         read_unlock_bh(&idev->lock);
 818 }
 819
 820 static void mca_get(struct ifmcaddr6 *mc)
 821 {
 822         atomic_inc(&mc->mca_refcnt);
 823 }
 824
 825 static void ma_put(struct ifmcaddr6 *mc)
 826 {
 827         if (atomic_dec_and_test(&mc->mca_refcnt)) {
 828                 in6_dev_put(mc->idev);
 829                 kfree(mc);
 830         }
 831 }
 832
 833 static struct ifmcaddr6 *mca_alloc(struct inet6_dev *idev,
 834                                    const struct in6_addr *addr)
 835 {
 836         struct ifmcaddr6 *mc;
 837
 838         mc = kzalloc(sizeof(*mc), GFP_ATOMIC);
 839         if (!mc)
 840                 return NULL;
 841
 842         setup_timer(&mc->mca_timer, igmp6_timer_handler, (unsigned long)mc);
 843
 844         mc->mca_addr = *addr;
 845         mc->idev = idev; /* reference taken by caller */
 846         mc->mca_users = 1;
 847         /* mca_stamp should be updated upon changes */
 848         mc->mca_cstamp = mc->mca_tstamp = jiffies;
 849         atomic_set(&mc->mca_refcnt, 1);
 850         spin_lock_init(&mc->mca_lock);
 851
 852         /* initial mode is (EX, empty) */
 853         mc->mca_sfmode = MCAST_EXCLUDE;
 854         mc->mca_sfcount[MCAST_EXCLUDE] = 1;
 855
 856         if (ipv6_addr_is_ll_all_nodes(&mc->mca_addr) ||
 857             IPV6_ADDR_MC_SCOPE(&mc->mca_addr) < IPV6_ADDR_SCOPE_LINKLOCAL)
 858                 mc->mca_flags |= MAF_NOREPORT;
 859
 860         return mc;
 861 }
 862
 863 /*
 864  *      device multicast group inc (add if not found)
 865  */
 866 int ipv6_dev_mc_inc(struct net_device *dev, const struct in6_addr *addr)
 867 {
 868         struct ifmcaddr6 *mc;
 869         struct inet6_dev *idev;
 870
 871         ASSERT_RTNL();
 872
 873         /* we need to take a reference on idev */
 874         idev = in6_dev_get(dev);
 875
 876         if (!idev)
 877                 return -EINVAL;
 878
 879         write_lock_bh(&idev->lock);
 880         if (idev->dead) {
 881                 write_unlock_bh(&idev->lock);
 882                 in6_dev_put(idev);
 883                 return -ENODEV;
 884         }
 885
 886         for (mc = idev->mc_list; mc; mc = mc->next) {
 887                 if (ipv6_addr_equal(&mc->mca_addr, addr)) {
 888                         mc->mca_users++;
 889                         write_unlock_bh(&idev->lock);
 890                         ip6_mc_add_src(idev, &mc->mca_addr, MCAST_EXCLUDE, 0,
 891                                 NULL, 0);
 892                         in6_dev_put(idev);
 893                         return 0;
 894                 }
 895         }
 896
 897         mc = mca_alloc(idev, addr);
 898         if (!mc) {
 899                 write_unlock_bh(&idev->lock);
 900                 in6_dev_put(idev);
 901                 return -ENOMEM;
 902         }
 903
 904         mc->next = idev->mc_list;
 905         idev->mc_list = mc;
 906
 907         /* Hold this for the code below before we unlock,
 908          * it is already exposed via idev->mc_list.
 909          */
 910         mca_get(mc);
 911         write_unlock_bh(&idev->lock);
 912
 913         mld_del_delrec(idev, mc);
 914         igmp6_group_added(mc);
 915         ma_put(mc);
 916         return 0;
 917 }
 918
 919 /*
 920  *      device multicast group del
 921  */
 922 int __ipv6_dev_mc_dec(struct inet6_dev *idev, const struct in6_addr *addr)
 923 {
 924         struct ifmcaddr6 *ma, **map;
 925
 926         ASSERT_RTNL();
 927
 928         write_lock_bh(&idev->lock);
 929         for (map = &idev->mc_list; (ma = *map) != NULL; map = &ma->next) {
 930                 if (ipv6_addr_equal(&ma->mca_addr, addr)) {
 931                         if (--ma->mca_users == 0) {
 932                                 *map = ma->next;
 933                                 write_unlock_bh(&idev->lock);
 934
 935                                 igmp6_group_dropped(ma);
 936                                 ip6_mc_clear_src(ma);
 937
 938                                 ma_put(ma);
 939                                 return 0;
 940                         }
 941                         write_unlock_bh(&idev->lock);
 942                         return 0;
 943                 }
 944         }
 945         write_unlock_bh(&idev->lock);
 946
 947         return -ENOENT;
 948 }
 949
 950 int ipv6_dev_mc_dec(struct net_device *dev, const struct in6_addr *addr)
 951 {
 952         struct inet6_dev *idev;
 953         int err;
 954
 955         ASSERT_RTNL();
 956
 957         idev = __in6_dev_get(dev);
 958         if (!idev)
 959                 err = -ENODEV;
 960         else
 961                 err = __ipv6_dev_mc_dec(idev, addr);
 962
 963         return err;
 964 }
 965
 966 /*
 967  *      check if the interface/address pair is valid
 968  */
 969 bool ipv6_chk_mcast_addr(struct net_device *dev, const struct in6_addr *group,
 970                          const struct in6_addr *src_addr)
 971 {
 972         struct inet6_dev *idev;
 973         struct ifmcaddr6 *mc;
 974         bool rv = false;
 975
 976         rcu_read_lock();
 977         idev = __in6_dev_get(dev);
 978         if (idev) {
 979                 read_lock_bh(&idev->lock);
 980                 for (mc = idev->mc_list; mc; mc = mc->next) {
 981                         if (ipv6_addr_equal(&mc->mca_addr, group))
 982                                 break;
 983                 }
 984                 if (mc) {
 985                         if (src_addr && !ipv6_addr_any(src_addr)) {
 986                                 struct ip6_sf_list *psf;
 987
 988                                 spin_lock_bh(&mc->mca_lock);
 989                                 for (psf = mc->mca_sources; psf; psf = psf->sf_next) {
 990                                         if (ipv6_addr_equal(&psf->sf_addr, src_addr))
 991                                                 break;
 992                                 }
 993                                 if (psf)
 994                                         rv = psf->sf_count[MCAST_INCLUDE] ||
 995                                                 psf->sf_count[MCAST_EXCLUDE] !=
 996                                                 mc->mca_sfcount[MCAST_EXCLUDE];
 997                                 else
 998                                         rv = mc->mca_sfcount[MCAST_EXCLUDE] != 0;
 999                                 spin_unlock_bh(&mc->mca_lock);
1000                         } else
1001                                 rv = true; /* don't filter unspecified source */
1002                 }
1003                 read_unlock_bh(&idev->lock);
1004         }
1005         rcu_read_unlock();
1006         return rv;
1007 }
1008
1009 static void mld_gq_start_timer(struct inet6_dev *idev)
1010 {
1011         unsigned long tv = prandom_u32() % idev->mc_maxdelay;
1012
1013         idev->mc_gq_running = 1;
1014         if (!mod_timer(&idev->mc_gq_timer, jiffies+tv+2))
1015                 in6_dev_hold(idev);
1016 }
1017
1018 static void mld_gq_stop_timer(struct inet6_dev *idev)
1019 {
1020         idev->mc_gq_running = 0;
1021         if (del_timer(&idev->mc_gq_timer))
1022                 __in6_dev_put(idev);
1023 }
1024
1025 static void mld_ifc_start_timer(struct inet6_dev *idev, unsigned long delay)
1026 {
1027         unsigned long tv = prandom_u32() % delay;
1028
1029         if (!mod_timer(&idev->mc_ifc_timer, jiffies+tv+2))
1030                 in6_dev_hold(idev);
1031 }
1032
1033 static void mld_ifc_stop_timer(struct inet6_dev *idev)
1034 {
1035         idev->mc_ifc_count = 0;
1036         if (del_timer(&idev->mc_ifc_timer))
1037                 __in6_dev_put(idev);
1038 }
1039
1040 static void mld_dad_start_timer(struct inet6_dev *idev, unsigned long delay)
1041 {
1042         unsigned long tv = prandom_u32() % delay;
1043
1044         if (!mod_timer(&idev->mc_dad_timer, jiffies+tv+2))
1045                 in6_dev_hold(idev);
1046 }
1047
1048 static void mld_dad_stop_timer(struct inet6_dev *idev)
1049 {
1050         if (del_timer(&idev->mc_dad_timer))
1051                 __in6_dev_put(idev);
1052 }
1053
1054 /*
1055  *      IGMP handling (alias multicast ICMPv6 messages)
1056  */
1057
1058 static void igmp6_group_queried(struct ifmcaddr6 *ma, unsigned long resptime)
1059 {
1060         unsigned long delay = resptime;
1061
1062         /* Do not start timer for these addresses */
1063         if (ipv6_addr_is_ll_all_nodes(&ma->mca_addr) ||
1064             IPV6_ADDR_MC_SCOPE(&ma->mca_addr) < IPV6_ADDR_SCOPE_LINKLOCAL)
1065                 return;
1066
1067         if (del_timer(&ma->mca_timer)) {
1068                 atomic_dec(&ma->mca_refcnt);
1069                 delay = ma->mca_timer.expires - jiffies;
1070         }
1071
1072         if (delay >= resptime)
1073                 delay = prandom_u32() % resptime;
1074
1075         ma->mca_timer.expires = jiffies + delay;
1076         if (!mod_timer(&ma->mca_timer, jiffies + delay))
1077                 atomic_inc(&ma->mca_refcnt);
1078         ma->mca_flags |= MAF_TIMER_RUNNING;
1079 }
1080
1081 /* mark EXCLUDE-mode sources */
1082 static bool mld_xmarksources(struct ifmcaddr6 *pmc, int nsrcs,
1083                              const struct in6_addr *srcs)
1084 {
1085         struct ip6_sf_list *psf;
1086         int i, scount;
1087
1088         scount = 0;
1089         for (psf = pmc->mca_sources; psf; psf = psf->sf_next) {
1090                 if (scount == nsrcs)
1091                         break;
1092                 for (i = 0; i < nsrcs; i++) {
1093                         /* skip inactive filters */
1094                         if (psf->sf_count[MCAST_INCLUDE] ||
1095                             pmc->mca_sfcount[MCAST_EXCLUDE] !=
1096                             psf->sf_count[MCAST_EXCLUDE])
1097                                 break;
1098                         if (ipv6_addr_equal(&srcs[i], &psf->sf_addr)) {
1099                                 scount++;
1100                                 break;
1101                         }
1102                 }
1103         }
1104         pmc->mca_flags &= ~MAF_GSQUERY;
1105         if (scount == nsrcs)    /* all sources excluded */
1106                 return false;
1107         return true;
1108 }
1109
1110 static bool mld_marksources(struct ifmcaddr6 *pmc, int nsrcs,
1111                             const struct in6_addr *srcs)
1112 {
1113         struct ip6_sf_list *psf;
1114         int i, scount;
1115
1116         if (pmc->mca_sfmode == MCAST_EXCLUDE)
1117                 return mld_xmarksources(pmc, nsrcs, srcs);
1118
1119         /* mark INCLUDE-mode sources */
1120
1121         scount = 0;
1122         for (psf = pmc->mca_sources; psf; psf = psf->sf_next) {
1123                 if (scount == nsrcs)
1124                         break;
1125                 for (i = 0; i < nsrcs; i++) {
1126                         if (ipv6_addr_equal(&srcs[i], &psf->sf_addr)) {
1127                                 psf->sf_gsresp = 1;
1128                                 scount++;
1129                                 break;
1130                         }
1131                 }
1132         }
1133         if (!scount) {
1134                 pmc->mca_flags &= ~MAF_GSQUERY;
1135                 return false;
1136         }
1137         pmc->mca_flags |= MAF_GSQUERY;
1138         return true;
1139 }
1140
1141 static int mld_force_mld_version(const struct inet6_dev *idev)
1142 {
1143         /* Normally, both are 0 here. If enforcement to a particular is
1144          * being used, individual device enforcement will have a lower
1145          * precedence over 'all' device (.../conf/all/force_mld_version).
1146          */
1147
1148         if (dev_net(idev->dev)->ipv6.devconf_all->force_mld_version != 0)
1149                 return dev_net(idev->dev)->ipv6.devconf_all->force_mld_version;
1150         else
1151                 return idev->cnf.force_mld_version;
1152 }
1153
1154 static bool mld_in_v2_mode_only(const struct inet6_dev *idev)
1155 {
1156         return mld_force_mld_version(idev) == 2;
1157 }
1158
1159 static bool mld_in_v1_mode_only(const struct inet6_dev *idev)
1160 {
1161         return mld_force_mld_version(idev) == 1;
1162 }
1163
1164 static bool mld_in_v1_mode(const struct inet6_dev *idev)
1165 {
1166         if (mld_in_v2_mode_only(idev))
1167                 return false;
1168         if (mld_in_v1_mode_only(idev))
1169                 return true;
1170         if (idev->mc_v1_seen && time_before(jiffies, idev->mc_v1_seen))
1171                 return true;
1172
1173         return false;
1174 }
1175
1176 static void mld_set_v1_mode(struct inet6_dev *idev)
1177 {
1178         /* RFC3810, relevant sections:
1179          *  - 9.1. Robustness Variable
1180          *  - 9.2. Query Interval
1181          *  - 9.3. Query Response Interval
1182          *  - 9.12. Older Version Querier Present Timeout
1183          */
1184         unsigned long switchback;
1185
1186         switchback = (idev->mc_qrv * idev->mc_qi) + idev->mc_qri;
1187
1188         idev->mc_v1_seen = jiffies + switchback;
1189 }
1190
1191 static void mld_update_qrv(struct inet6_dev *idev,
1192                            const struct mld2_query *mlh2)
1193 {
1194         /* RFC3810, relevant sections:
1195          *  - 5.1.8. QRV (Querier's Robustness Variable)
1196          *  - 9.1. Robustness Variable
1197          */
1198
1199         /* The value of the Robustness Variable MUST NOT be zero,
1200          * and SHOULD NOT be one. Catch this here if we ever run
1201          * into such a case in future.
1202          */
1203         const int min_qrv = min(MLD_QRV_DEFAULT, sysctl_mld_qrv);
1204         WARN_ON(idev->mc_qrv == 0);
1205
1206         if (mlh2->mld2q_qrv > 0)
1207                 idev->mc_qrv = mlh2->mld2q_qrv;
1208
1209         if (unlikely(idev->mc_qrv < min_qrv)) {
1210                 net_warn_ratelimited("IPv6: MLD: clamping QRV from %u to %u!\n",
1211                                      idev->mc_qrv, min_qrv);
1212                 idev->mc_qrv = min_qrv;
1213         }
1214 }
1215
1216 static void mld_update_qi(struct inet6_dev *idev,
1217                           const struct mld2_query *mlh2)
1218 {
1219         /* RFC3810, relevant sections:
1220          *  - 5.1.9. QQIC (Querier's Query Interval Code)
1221          *  - 9.2. Query Interval
1222          *  - 9.12. Older Version Querier Present Timeout
1223          *    (the [Query Interval] in the last Query received)
1224          */
1225         unsigned long mc_qqi;
1226
1227         if (mlh2->mld2q_qqic < 128) {
1228                 mc_qqi = mlh2->mld2q_qqic;
1229         } else {
1230                 unsigned long mc_man, mc_exp;
1231
1232                 mc_exp = MLDV2_QQIC_EXP(mlh2->mld2q_qqic);
1233                 mc_man = MLDV2_QQIC_MAN(mlh2->mld2q_qqic);
1234
1235                 mc_qqi = (mc_man | 0x10) << (mc_exp + 3);
1236         }
1237
1238         idev->mc_qi = mc_qqi * HZ;
1239 }
1240
1241 static void mld_update_qri(struct inet6_dev *idev,
1242                            const struct mld2_query *mlh2)
1243 {
1244         /* RFC3810, relevant sections:
1245          *  - 5.1.3. Maximum Response Code
1246          *  - 9.3. Query Response Interval
1247          */
1248         idev->mc_qri = msecs_to_jiffies(mldv2_mrc(mlh2));
1249 }
1250
1251 static int mld_process_v1(struct inet6_dev *idev, struct mld_msg *mld,
1252                           unsigned long *max_delay, bool v1_query)
1253 {
1254         unsigned long mldv1_md;
1255
1256         /* Ignore v1 queries */
1257         if (mld_in_v2_mode_only(idev))
1258                 return -EINVAL;
1259
1260         mldv1_md = ntohs(mld->mld_maxdelay);
1261
1262         /* When in MLDv1 fallback and a MLDv2 router start-up being
1263          * unaware of current MLDv1 operation, the MRC == MRD mapping
1264          * only works when the exponential algorithm is not being
1265          * used (as MLDv1 is unaware of such things).
1266          *
1267          * According to the RFC author, the MLDv2 implementations
1268          * he's aware of all use a MRC < 32768 on start up queries.
1269          *
1270          * Thus, should we *ever* encounter something else larger
1271          * than that, just assume the maximum possible within our
1272          * reach.
1273          */
1274         if (!v1_query)
1275                 mldv1_md = min(mldv1_md, MLDV1_MRD_MAX_COMPAT);
1276
1277         *max_delay = max(msecs_to_jiffies(mldv1_md), 1UL);
1278
1279         /* MLDv1 router present: we need to go into v1 mode *only*
1280          * when an MLDv1 query is received as per section 9.12. of
1281          * RFC3810! And we know from RFC2710 section 3.7 that MLDv1
1282          * queries MUST be of exactly 24 octets.
1283          */
1284         if (v1_query)
1285                 mld_set_v1_mode(idev);
1286
1287         /* cancel MLDv2 report timer */
1288         mld_gq_stop_timer(idev);
1289         /* cancel the interface change timer */
1290         mld_ifc_stop_timer(idev);
1291         /* clear deleted report items */
1292         mld_clear_delrec(idev);
1293
1294         return 0;
1295 }
1296
1297 static int mld_process_v2(struct inet6_dev *idev, struct mld2_query *mld,
1298                           unsigned long *max_delay)
1299 {
1300         *max_delay = max(msecs_to_jiffies(mldv2_mrc(mld)), 1UL);
1301
1302         mld_update_qrv(idev, mld);
1303         mld_update_qi(idev, mld);
1304         mld_update_qri(idev, mld);
1305
1306         idev->mc_maxdelay = *max_delay;
1307
1308         return 0;
1309 }
1310
1311 /* called with rcu_read_lock() */
1312 int igmp6_event_query(struct sk_buff *skb)
1313 {
1314         struct mld2_query *mlh2 = NULL;
1315         struct ifmcaddr6 *ma;
1316         const struct in6_addr *group;
1317         unsigned long max_delay;
1318         struct inet6_dev *idev;
1319         struct mld_msg *mld;
1320         int group_type;
1321         int mark = 0;
1322         int len, err;
1323
1324         if (!pskb_may_pull(skb, sizeof(struct in6_addr)))
1325                 return -EINVAL;
1326
1327         /* compute payload length excluding extension headers */
1328         len = ntohs(ipv6_hdr(skb)->payload_len) + sizeof(struct ipv6hdr);
1329         len -= skb_network_header_len(skb);
1330
1331         /* RFC3810 6.2
1332          * Upon reception of an MLD message that contains a Query, the node
1333          * checks if the source address of the message is a valid link-local
1334          * address, if the Hop Limit is set to 1, and if the Router Alert
1335          * option is present in the Hop-By-Hop Options header of the IPv6
1336          * packet.  If any of these checks fails, the packet is dropped.
1337          */
1338         if (!(ipv6_addr_type(&ipv6_hdr(skb)->saddr) & IPV6_ADDR_LINKLOCAL) ||
1339             ipv6_hdr(skb)->hop_limit != 1 ||
1340             !(IP6CB(skb)->flags & IP6SKB_ROUTERALERT) ||
1341             IP6CB(skb)->ra != htons(IPV6_OPT_ROUTERALERT_MLD))
1342                 return -EINVAL;
1343
1344         idev = __in6_dev_get(skb->dev);
1345         if (!idev)
1346                 return 0;
1347
1348         mld = (struct mld_msg *)icmp6_hdr(skb);
1349         group = &mld->mld_mca;
1350         group_type = ipv6_addr_type(group);
1351
1352         if (group_type != IPV6_ADDR_ANY &&
1353             !(group_type&IPV6_ADDR_MULTICAST))
1354                 return -EINVAL;
1355
1356         if (len < MLD_V1_QUERY_LEN) {
1357                 return -EINVAL;
1358         } else if (len == MLD_V1_QUERY_LEN || mld_in_v1_mode(idev)) {
1359                 err = mld_process_v1(idev, mld, &max_delay,
1360                                      len == MLD_V1_QUERY_LEN);
1361                 if (err < 0)
1362                         return err;
1363         } else if (len >= MLD_V2_QUERY_LEN_MIN) {
1364                 int srcs_offset = sizeof(struct mld2_query) -
1365                                   sizeof(struct icmp6hdr);
1366
1367                 if (!pskb_may_pull(skb, srcs_offset))
1368                         return -EINVAL;
1369
1370                 mlh2 = (struct mld2_query *)skb_transport_header(skb);
1371
1372                 err = mld_process_v2(idev, mlh2, &max_delay);
1373                 if (err < 0)
1374                         return err;
1375
1376                 if (group_type == IPV6_ADDR_ANY) { /* general query */
1377                         if (mlh2->mld2q_nsrcs)
1378                                 return -EINVAL; /* no sources allowed */
1379
1380                         mld_gq_start_timer(idev);
1381                         return 0;
1382                 }
1383                 /* mark sources to include, if group & source-specific */
1384                 if (mlh2->mld2q_nsrcs != 0) {
1385                         if (!pskb_may_pull(skb, srcs_offset +
1386                             ntohs(mlh2->mld2q_nsrcs) * sizeof(struct in6_addr)))
1387                                 return -EINVAL;
1388
1389                         mlh2 = (struct mld2_query *)skb_transport_header(skb);
1390                         mark = 1;
1391                 }
1392         } else {
1393                 return -EINVAL;
1394         }
1395
1396         read_lock_bh(&idev->lock);
1397         if (group_type == IPV6_ADDR_ANY) {
1398                 for (ma = idev->mc_list; ma; ma = ma->next) {
1399                         spin_lock_bh(&ma->mca_lock);
1400                         igmp6_group_queried(ma, max_delay);
1401                         spin_unlock_bh(&ma->mca_lock);
1402                 }
1403         } else {
1404                 for (ma = idev->mc_list; ma; ma = ma->next) {
1405                         if (!ipv6_addr_equal(group, &ma->mca_addr))
1406                                 continue;
1407                         spin_lock_bh(&ma->mca_lock);
1408                         if (ma->mca_flags & MAF_TIMER_RUNNING) {
1409                                 /* gsquery <- gsquery && mark */
1410                                 if (!mark)
1411                                         ma->mca_flags &= ~MAF_GSQUERY;
1412                         } else {
1413                                 /* gsquery <- mark */
1414                                 if (mark)
1415                                         ma->mca_flags |= MAF_GSQUERY;
1416                                 else
1417                                         ma->mca_flags &= ~MAF_GSQUERY;
1418                         }
1419                         if (!(ma->mca_flags & MAF_GSQUERY) ||
1420                             mld_marksources(ma, ntohs(mlh2->mld2q_nsrcs), mlh2->mld2q_srcs))
1421                                 igmp6_group_queried(ma, max_delay);
1422                         spin_unlock_bh(&ma->mca_lock);
1423                         break;
1424                 }
1425         }
1426         read_unlock_bh(&idev->lock);
1427
1428         return 0;
1429 }
1430
1431 /* called with rcu_read_lock() */
1432 int igmp6_event_report(struct sk_buff *skb)
1433 {
1434         struct ifmcaddr6 *ma;
1435         struct inet6_dev *idev;
1436         struct mld_msg *mld;
1437         int addr_type;
1438
1439         /* Our own report looped back. Ignore it. */
1440         if (skb->pkt_type == PACKET_LOOPBACK)
1441                 return 0;
1442
1443         /* send our report if the MC router may not have heard this report */
1444         if (skb->pkt_type != PACKET_MULTICAST &&
1445             skb->pkt_type != PACKET_BROADCAST)
1446                 return 0;
1447
1448         if (!pskb_may_pull(skb, sizeof(*mld) - sizeof(struct icmp6hdr)))
1449                 return -EINVAL;
1450
1451         mld = (struct mld_msg *)icmp6_hdr(skb);
1452
1453         /* Drop reports with not link local source */
1454         addr_type = ipv6_addr_type(&ipv6_hdr(skb)->saddr);
1455         if (addr_type != IPV6_ADDR_ANY &&
1456             !(addr_type&IPV6_ADDR_LINKLOCAL))
1457                 return -EINVAL;
1458
1459         idev = __in6_dev_get(skb->dev);
1460         if (!idev)
1461                 return -ENODEV;
1462
1463         /*
1464          *      Cancel the timer for this group
1465          */
1466
1467         read_lock_bh(&idev->lock);
1468         for (ma = idev->mc_list; ma; ma = ma->next) {
1469                 if (ipv6_addr_equal(&ma->mca_addr, &mld->mld_mca)) {
1470                         spin_lock(&ma->mca_lock);
1471                         if (del_timer(&ma->mca_timer))
1472                                 atomic_dec(&ma->mca_refcnt);
1473                         ma->mca_flags &= ~(MAF_LAST_REPORTER|MAF_TIMER_RUNNING);
1474                         spin_unlock(&ma->mca_lock);
1475                         break;
1476                 }
1477         }
1478         read_unlock_bh(&idev->lock);
1479         return 0;
1480 }
1481
1482 static bool is_in(struct ifmcaddr6 *pmc, struct ip6_sf_list *psf, int type,
1483                   int gdeleted, int sdeleted)
1484 {
1485         switch (type) {
1486         case MLD2_MODE_IS_INCLUDE:
1487         case MLD2_MODE_IS_EXCLUDE:
1488                 if (gdeleted || sdeleted)
1489                         return false;
1490                 if (!((pmc->mca_flags & MAF_GSQUERY) && !psf->sf_gsresp)) {
1491                         if (pmc->mca_sfmode == MCAST_INCLUDE)
1492                                 return true;
1493                         /* don't include if this source is excluded
1494                          * in all filters
1495                          */
1496                         if (psf->sf_count[MCAST_INCLUDE])
1497                                 return type == MLD2_MODE_IS_INCLUDE;
1498                         return pmc->mca_sfcount[MCAST_EXCLUDE] ==
1499                                 psf->sf_count[MCAST_EXCLUDE];
1500                 }
1501                 return false;
1502         case MLD2_CHANGE_TO_INCLUDE:
1503                 if (gdeleted || sdeleted)
1504                         return false;
1505                 return psf->sf_count[MCAST_INCLUDE] != 0;
1506         case MLD2_CHANGE_TO_EXCLUDE:
1507                 if (gdeleted || sdeleted)
1508                         return false;
1509                 if (pmc->mca_sfcount[MCAST_EXCLUDE] == 0 ||
1510                     psf->sf_count[MCAST_INCLUDE])
1511                         return false;
1512                 return pmc->mca_sfcount[MCAST_EXCLUDE] ==
1513                         psf->sf_count[MCAST_EXCLUDE];
1514         case MLD2_ALLOW_NEW_SOURCES:
1515                 if (gdeleted || !psf->sf_crcount)
1516                         return false;
1517                 return (pmc->mca_sfmode == MCAST_INCLUDE) ^ sdeleted;
1518         case MLD2_BLOCK_OLD_SOURCES:
1519                 if (pmc->mca_sfmode == MCAST_INCLUDE)
1520                         return gdeleted || (psf->sf_crcount && sdeleted);
1521                 return psf->sf_crcount && !gdeleted && !sdeleted;
1522         }
1523         return false;
1524 }
1525
1526 static int
1527 mld_scount(struct ifmcaddr6 *pmc, int type, int gdeleted, int sdeleted)
1528 {
1529         struct ip6_sf_list *psf;
1530         int scount = 0;
1531
1532         for (psf = pmc->mca_sources; psf; psf = psf->sf_next) {
1533                 if (!is_in(pmc, psf, type, gdeleted, sdeleted))
1534                         continue;
1535                 scount++;
1536         }
1537         return scount;
1538 }
1539
1540 static void ip6_mc_hdr(struct sock *sk, struct sk_buff *skb,
1541                        struct net_device *dev,
1542                        const struct in6_addr *saddr,
1543                        const struct in6_addr *daddr,
1544                        int proto, int len)
1545 {
1546         struct ipv6hdr *hdr;
1547
1548         skb->protocol = htons(ETH_P_IPV6);
1549         skb->dev = dev;
1550
1551         skb_reset_network_header(skb);
1552         skb_put(skb, sizeof(struct ipv6hdr));
1553         hdr = ipv6_hdr(skb);
1554
1555         ip6_flow_hdr(hdr, 0, 0);
1556
1557         hdr->payload_len = htons(len);
1558         hdr->nexthdr = proto;
1559         hdr->hop_limit = inet6_sk(sk)->hop_limit;
1560
1561         hdr->saddr = *saddr;
1562         hdr->daddr = *daddr;
1563 }
1564
1565 static struct sk_buff *mld_newpack(struct inet6_dev *idev, unsigned int mtu)
1566 {
1567         struct net_device *dev = idev->dev;
1568         struct net *net = dev_net(dev);
1569         struct sock *sk = net->ipv6.igmp_sk;
1570         struct sk_buff *skb;
1571         struct mld2_report *pmr;
1572         struct in6_addr addr_buf;
1573         const struct in6_addr *saddr;
1574         int hlen = LL_RESERVED_SPACE(dev);
1575         int tlen = dev->needed_tailroom;
1576         unsigned int size = mtu + hlen + tlen;
1577         int err;
1578         u8 ra[8] = { IPPROTO_ICMPV6, 0,
1579                      IPV6_TLV_ROUTERALERT, 2, 0, 0,
1580                      IPV6_TLV_PADN, 0 };
1581
1582         /* we assume size > sizeof(ra) here */
1583         /* limit our allocations to order-0 page */
1584         size = min_t(int, size, SKB_MAX_ORDER(0, 0));
1585         skb = sock_alloc_send_skb(sk, size, 1, &err);
1586
1587         if (!skb)
1588                 return NULL;
1589
1590         skb->priority = TC_PRIO_CONTROL;
1591         skb_reserve(skb, hlen);
1592         skb_tailroom_reserve(skb, mtu, tlen);
1593
1594         if (__ipv6_get_lladdr(idev, &addr_buf, IFA_F_TENTATIVE)) {
1595                 /* <draft-ietf-magma-mld-source-05.txt>:
1596                  * use unspecified address as the source address
1597                  * when a valid link-local address is not available.
1598                  */
1599                 saddr = &in6addr_any;
1600         } else
1601                 saddr = &addr_buf;
1602
1603         ip6_mc_hdr(sk, skb, dev, saddr, &mld2_all_mcr, NEXTHDR_HOP, 0);
1604
1605         memcpy(skb_put(skb, sizeof(ra)), ra, sizeof(ra));
1606
1607         skb_set_transport_header(skb, skb_tail_pointer(skb) - skb->data);
1608         skb_put(skb, sizeof(*pmr));
1609         pmr = (struct mld2_report *)skb_transport_header(skb);
1610         pmr->mld2r_type = ICMPV6_MLD2_REPORT;
1611         pmr->mld2r_resv1 = 0;
1612         pmr->mld2r_cksum = 0;
1613         pmr->mld2r_resv2 = 0;
1614         pmr->mld2r_ngrec = 0;
1615         return skb;
1616 }
1617
1618 static void mld_sendpack(struct sk_buff *skb)
1619 {
1620         struct ipv6hdr *pip6 = ipv6_hdr(skb);
1621         struct mld2_report *pmr =
1622                               (struct mld2_report *)skb_transport_header(skb);
1623         int payload_len, mldlen;
1624         struct inet6_dev *idev;
1625         struct net *net = dev_net(skb->dev);
1626         int err;
1627         struct flowi6 fl6;
1628         struct dst_entry *dst;
1629
1630         rcu_read_lock();
1631         idev = __in6_dev_get(skb->dev);
1632         IP6_UPD_PO_STATS(net, idev, IPSTATS_MIB_OUT, skb->len);
1633
1634         payload_len = (skb_tail_pointer(skb) - skb_network_header(skb)) -
1635                 sizeof(*pip6);
1636         mldlen = skb_tail_pointer(skb) - skb_transport_header(skb);
1637         pip6->payload_len = htons(payload_len);
1638
1639         pmr->mld2r_cksum = csum_ipv6_magic(&pip6->saddr, &pip6->daddr, mldlen,
1640                                            IPPROTO_ICMPV6,
1641                                            csum_partial(skb_transport_header(skb),
1642                                                         mldlen, 0));
1643
1644         icmpv6_flow_init(net->ipv6.igmp_sk, &fl6, ICMPV6_MLD2_REPORT,
1645                          &ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr,
1646                          skb->dev->ifindex);
1647         dst = icmp6_dst_alloc(skb->dev, &fl6);
1648
1649         err = 0;
1650         if (IS_ERR(dst)) {
1651                 err = PTR_ERR(dst);
1652                 dst = NULL;
1653         }
1654         skb_dst_set(skb, dst);
1655         if (err)
1656                 goto err_out;
1657
1658         payload_len = skb->len;
1659
1660         err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT,
1661                       net, net->ipv6.igmp_sk, skb, NULL, skb->dev,
1662                       dst_output);
1663 out:
1664         if (!err) {
1665                 ICMP6MSGOUT_INC_STATS(net, idev, ICMPV6_MLD2_REPORT);
1666                 ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTMSGS);
1667         } else {
1668                 IP6_INC_STATS(net, idev, IPSTATS_MIB_OUTDISCARDS);
1669         }
1670
1671         rcu_read_unlock();
1672         return;
1673
1674 err_out:
1675         kfree_skb(skb);
1676         goto out;
1677 }
1678
1679 static int grec_size(struct ifmcaddr6 *pmc, int type, int gdel, int sdel)
1680 {
1681         return sizeof(struct mld2_grec) + 16 * mld_scount(pmc,type,gdel,sdel);
1682 }
1683
1684 static struct sk_buff *add_grhead(struct sk_buff *skb, struct ifmcaddr6 *pmc,
1685         int type, struct mld2_grec **ppgr, unsigned int mtu)
1686 {
1687         struct mld2_report *pmr;
1688         struct mld2_grec *pgr;
1689
1690         if (!skb) {
1691                 skb = mld_newpack(pmc->idev, mtu);
1692                 if (!skb)
1693                         return NULL;
1694         }
1695         pgr = (struct mld2_grec *)skb_put(skb, sizeof(struct mld2_grec));
1696         pgr->grec_type = type;
1697         pgr->grec_auxwords = 0;
1698         pgr->grec_nsrcs = 0;
1699         pgr->grec_mca = pmc->mca_addr;  /* structure copy */
1700         pmr = (struct mld2_report *)skb_transport_header(skb);
1701         pmr->mld2r_ngrec = htons(ntohs(pmr->mld2r_ngrec)+1);
1702         *ppgr = pgr;
1703         return skb;
1704 }
1705
1706 #define AVAILABLE(skb)  ((skb) ? skb_availroom(skb) : 0)
1707
1708 static struct sk_buff *add_grec(struct sk_buff *skb, struct ifmcaddr6 *pmc,
1709         int type, int gdeleted, int sdeleted, int crsend)
1710 {
1711         struct inet6_dev *idev = pmc->idev;
1712         struct net_device *dev = idev->dev;
1713         struct mld2_report *pmr;
1714         struct mld2_grec *pgr = NULL;
1715         struct ip6_sf_list *psf, *psf_next, *psf_prev, **psf_list;
1716         int scount, stotal, first, isquery, truncate;
1717         unsigned int mtu;
1718
1719         if (pmc->mca_flags & MAF_NOREPORT)
1720                 return skb;
1721
1722         mtu = READ_ONCE(dev->mtu);
1723         if (mtu < IPV6_MIN_MTU)
1724                 return skb;
1725
1726         isquery = type == MLD2_MODE_IS_INCLUDE ||
1727                   type == MLD2_MODE_IS_EXCLUDE;
1728         truncate = type == MLD2_MODE_IS_EXCLUDE ||
1729                     type == MLD2_CHANGE_TO_EXCLUDE;
1730
1731         stotal = scount = 0;
1732
1733         psf_list = sdeleted ? &pmc->mca_tomb : &pmc->mca_sources;
1734
1735         if (!*psf_list)
1736                 goto empty_source;
1737
1738         pmr = skb ? (struct mld2_report *)skb_transport_header(skb) : NULL;
1739
1740         /* EX and TO_EX get a fresh packet, if needed */
1741         if (truncate) {
1742                 if (pmr && pmr->mld2r_ngrec &&
1743                     AVAILABLE(skb) < grec_size(pmc, type, gdeleted, sdeleted)) {
1744                         if (skb)
1745                                 mld_sendpack(skb);
1746                         skb = mld_newpack(idev, mtu);
1747                 }
1748         }
1749         first = 1;
1750         psf_prev = NULL;
1751         for (psf = *psf_list; psf; psf = psf_next) {
1752                 struct in6_addr *psrc;
1753
1754                 psf_next = psf->sf_next;
1755
1756                 if (!is_in(pmc, psf, type, gdeleted, sdeleted)) {
1757                         psf_prev = psf;
1758                         continue;
1759                 }
1760
1761                 /* Based on RFC3810 6.1. Should not send source-list change
1762                  * records when there is a filter mode change.
1763                  */
1764                 if (((gdeleted && pmc->mca_sfmode == MCAST_EXCLUDE) ||
1765                      (!gdeleted && pmc->mca_crcount)) &&
1766                     (type == MLD2_ALLOW_NEW_SOURCES ||
1767                      type == MLD2_BLOCK_OLD_SOURCES) && psf->sf_crcount)
1768                         goto decrease_sf_crcount;
1769
1770                 /* clear marks on query responses */
1771                 if (isquery)
1772                         psf->sf_gsresp = 0;
1773
1774                 if (AVAILABLE(skb) < sizeof(*psrc) +
1775                     first*sizeof(struct mld2_grec)) {
1776                         if (truncate && !first)
1777                                 break;   /* truncate these */
1778                         if (pgr)
1779                                 pgr->grec_nsrcs = htons(scount);
1780                         if (skb)
1781                                 mld_sendpack(skb);
1782                         skb = mld_newpack(idev, mtu);
1783                         first = 1;
1784                         scount = 0;
1785                 }
1786                 if (first) {
1787                         skb = add_grhead(skb, pmc, type, &pgr, mtu);
1788                         first = 0;
1789                 }
1790                 if (!skb)
1791                         return NULL;
1792                 psrc = (struct in6_addr *)skb_put(skb, sizeof(*psrc));
1793                 *psrc = psf->sf_addr;
1794                 scount++; stotal++;
1795                 if ((type == MLD2_ALLOW_NEW_SOURCES ||
1796                      type == MLD2_BLOCK_OLD_SOURCES) && psf->sf_crcount) {
1797 decrease_sf_crcount:
1798                         psf->sf_crcount--;
1799                         if ((sdeleted || gdeleted) && psf->sf_crcount == 0) {
1800                                 if (psf_prev)
1801                                         psf_prev->sf_next = psf->sf_next;
1802                                 else
1803                                         *psf_list = psf->sf_next;
1804                                 kfree(psf);
1805                                 continue;
1806                         }
1807                 }
1808                 psf_prev = psf;
1809         }
1810
1811 empty_source:
1812         if (!stotal) {
1813                 if (type == MLD2_ALLOW_NEW_SOURCES ||
1814                     type == MLD2_BLOCK_OLD_SOURCES)
1815                         return skb;
1816                 if (pmc->mca_crcount || isquery || crsend) {
1817                         /* make sure we have room for group header */
1818                         if (skb && AVAILABLE(skb) < sizeof(struct mld2_grec)) {
1819                                 mld_sendpack(skb);
1820                                 skb = NULL; /* add_grhead will get a new one */
1821                         }
1822                         skb = add_grhead(skb, pmc, type, &pgr, mtu);
1823                 }
1824         }
1825         if (pgr)
1826                 pgr->grec_nsrcs = htons(scount);
1827
1828         if (isquery)
1829                 pmc->mca_flags &= ~MAF_GSQUERY; /* clear query state */
1830         return skb;
1831 }
1832
1833 static void mld_send_report(struct inet6_dev *idev, struct ifmcaddr6 *pmc)
1834 {
1835         struct sk_buff *skb = NULL;
1836         int type;
1837
1838         read_lock_bh(&idev->lock);
1839         if (!pmc) {
1840                 for (pmc = idev->mc_list; pmc; pmc = pmc->next) {
1841                         if (pmc->mca_flags & MAF_NOREPORT)
1842                                 continue;
1843                         spin_lock_bh(&pmc->mca_lock);
1844                         if (pmc->mca_sfcount[MCAST_EXCLUDE])
1845                                 type = MLD2_MODE_IS_EXCLUDE;
1846                         else
1847                                 type = MLD2_MODE_IS_INCLUDE;
1848                         skb = add_grec(skb, pmc, type, 0, 0, 0);
1849                         spin_unlock_bh(&pmc->mca_lock);
1850                 }
1851         } else {
1852                 spin_lock_bh(&pmc->mca_lock);
1853                 if (pmc->mca_sfcount[MCAST_EXCLUDE])
1854                         type = MLD2_MODE_IS_EXCLUDE;
1855                 else
1856                         type = MLD2_MODE_IS_INCLUDE;
1857                 skb = add_grec(skb, pmc, type, 0, 0, 0);
1858                 spin_unlock_bh(&pmc->mca_lock);
1859         }
1860         read_unlock_bh(&idev->lock);
1861         if (skb)
1862                 mld_sendpack(skb);
1863 }
1864
1865 /*
1866  * remove zero-count source records from a source filter list
1867  */
1868 static void mld_clear_zeros(struct ip6_sf_list **ppsf)
1869 {
1870         struct ip6_sf_list *psf_prev, *psf_next, *psf;
1871
1872         psf_prev = NULL;
1873         for (psf = *ppsf; psf; psf = psf_next) {
1874                 psf_next = psf->sf_next;
1875                 if (psf->sf_crcount == 0) {
1876                         if (psf_prev)
1877                                 psf_prev->sf_next = psf->sf_next;
1878                         else
1879                                 *ppsf = psf->sf_next;
1880                         kfree(psf);
1881                 } else
1882                         psf_prev = psf;
1883         }
1884 }
1885
1886 static void mld_send_cr(struct inet6_dev *idev)
1887 {
1888         struct ifmcaddr6 *pmc, *pmc_prev, *pmc_next;
1889         struct sk_buff *skb = NULL;
1890         int type, dtype;
1891
1892         read_lock_bh(&idev->lock);
1893         spin_lock(&idev->mc_lock);
1894
1895         /* deleted MCA's */
1896         pmc_prev = NULL;
1897         for (pmc = idev->mc_tomb; pmc; pmc = pmc_next) {
1898                 pmc_next = pmc->next;
1899                 if (pmc->mca_sfmode == MCAST_INCLUDE) {
1900                         type = MLD2_BLOCK_OLD_SOURCES;
1901                         dtype = MLD2_BLOCK_OLD_SOURCES;
1902                         skb = add_grec(skb, pmc, type, 1, 0, 0);
1903                         skb = add_grec(skb, pmc, dtype, 1, 1, 0);
1904                 }
1905                 if (pmc->mca_crcount) {
1906                         if (pmc->mca_sfmode == MCAST_EXCLUDE) {
1907                                 type = MLD2_CHANGE_TO_INCLUDE;
1908                                 skb = add_grec(skb, pmc, type, 1, 0, 0);
1909                         }
1910                         pmc->mca_crcount--;
1911                         if (pmc->mca_crcount == 0) {
1912                                 mld_clear_zeros(&pmc->mca_tomb);
1913                                 mld_clear_zeros(&pmc->mca_sources);
1914                         }
1915                 }
1916                 if (pmc->mca_crcount == 0 && !pmc->mca_tomb &&
1917                     !pmc->mca_sources) {
1918                         if (pmc_prev)
1919                                 pmc_prev->next = pmc_next;
1920                         else
1921                                 idev->mc_tomb = pmc_next;
1922                         in6_dev_put(pmc->idev);
1923                         kfree(pmc);
1924                 } else
1925                         pmc_prev = pmc;
1926         }
1927         spin_unlock(&idev->mc_lock);
1928
1929         /* change recs */
1930         for (pmc = idev->mc_list; pmc; pmc = pmc->next) {
1931                 spin_lock_bh(&pmc->mca_lock);
1932                 if (pmc->mca_sfcount[MCAST_EXCLUDE]) {
1933                         type = MLD2_BLOCK_OLD_SOURCES;
1934                         dtype = MLD2_ALLOW_NEW_SOURCES;
1935                 } else {
1936                         type = MLD2_ALLOW_NEW_SOURCES;
1937                         dtype = MLD2_BLOCK_OLD_SOURCES;
1938                 }
1939                 skb = add_grec(skb, pmc, type, 0, 0, 0);
1940                 skb = add_grec(skb, pmc, dtype, 0, 1, 0);       /* deleted sources */
1941
1942                 /* filter mode changes */
1943                 if (pmc->mca_crcount) {
1944                         if (pmc->mca_sfmode == MCAST_EXCLUDE)
1945                                 type = MLD2_CHANGE_TO_EXCLUDE;
1946                         else
1947                                 type = MLD2_CHANGE_TO_INCLUDE;
1948                         skb = add_grec(skb, pmc, type, 0, 0, 0);
1949                         pmc->mca_crcount--;
1950                 }
1951                 spin_unlock_bh(&pmc->mca_lock);
1952         }
1953         read_unlock_bh(&idev->lock);
1954         if (!skb)
1955                 return;
1956         (void) mld_sendpack(skb);
1957 }
1958
1959 static void igmp6_send(struct in6_addr *addr, struct net_device *dev, int type)
1960 {
1961         struct net *net = dev_net(dev);
1962         struct sock *sk = net->ipv6.igmp_sk;
1963         struct inet6_dev *idev;
1964         struct sk_buff *skb;
1965         struct mld_msg *hdr;
1966         const struct in6_addr *snd_addr, *saddr;
1967         struct in6_addr addr_buf;
1968         int hlen = LL_RESERVED_SPACE(dev);
1969         int tlen = dev->needed_tailroom;
1970         int err, len, payload_len, full_len;
1971         u8 ra[8] = { IPPROTO_ICMPV6, 0,
1972                      IPV6_TLV_ROUTERALERT, 2, 0, 0,
1973                      IPV6_TLV_PADN, 0 };
1974         struct flowi6 fl6;
1975         struct dst_entry *dst;
1976
1977         if (type == ICMPV6_MGM_REDUCTION)
1978                 snd_addr = &in6addr_linklocal_allrouters;
1979         else
1980                 snd_addr = addr;
1981
1982         len = sizeof(struct icmp6hdr) + sizeof(struct in6_addr);
1983         payload_len = len + sizeof(ra);
1984         full_len = sizeof(struct ipv6hdr) + payload_len;
1985
1986         rcu_read_lock();
1987         IP6_UPD_PO_STATS(net, __in6_dev_get(dev),
1988                       IPSTATS_MIB_OUT, full_len);
1989         rcu_read_unlock();
1990
1991         skb = sock_alloc_send_skb(sk, hlen + tlen + full_len, 1, &err);
1992
1993         if (!skb) {
1994                 rcu_read_lock();
1995                 IP6_INC_STATS(net, __in6_dev_get(dev),
1996                               IPSTATS_MIB_OUTDISCARDS);
1997                 rcu_read_unlock();
1998                 return;
1999         }
2000         skb->priority = TC_PRIO_CONTROL;
2001         skb_reserve(skb, hlen);
2002
2003         if (ipv6_get_lladdr(dev, &addr_buf, IFA_F_TENTATIVE)) {
2004                 /* <draft-ietf-magma-mld-source-05.txt>:
2005                  * use unspecified address as the source address
2006                  * when a valid link-local address is not available.
2007                  */
2008                 saddr = &in6addr_any;
2009         } else
2010                 saddr = &addr_buf;
2011
2012         ip6_mc_hdr(sk, skb, dev, saddr, snd_addr, NEXTHDR_HOP, payload_len);
2013
2014         memcpy(skb_put(skb, sizeof(ra)), ra, sizeof(ra));
2015
2016         hdr = (struct mld_msg *) skb_put(skb, sizeof(struct mld_msg));
2017         memset(hdr, 0, sizeof(struct mld_msg));
2018         hdr->mld_type = type;
2019         hdr->mld_mca = *addr;
2020
2021         hdr->mld_cksum = csum_ipv6_magic(saddr, snd_addr, len,
2022                                          IPPROTO_ICMPV6,
2023                                          csum_partial(hdr, len, 0));
2024
2025         rcu_read_lock();
2026         idev = __in6_dev_get(skb->dev);
2027
2028         icmpv6_flow_init(sk, &fl6, type,
2029                          &ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr,
2030                          skb->dev->ifindex);
2031         dst = icmp6_dst_alloc(skb->dev, &fl6);
2032         if (IS_ERR(dst)) {
2033                 err = PTR_ERR(dst);
2034                 goto err_out;
2035         }
2036
2037         skb_dst_set(skb, dst);
2038         err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT,
2039                       net, sk, skb, NULL, skb->dev,
2040                       dst_output);
2041 out:
2042         if (!err) {
2043                 ICMP6MSGOUT_INC_STATS(net, idev, type);
2044                 ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTMSGS);
2045         } else
2046                 IP6_INC_STATS(net, idev, IPSTATS_MIB_OUTDISCARDS);
2047
2048         rcu_read_unlock();
2049         return;
2050
2051 err_out:
2052         kfree_skb(skb);
2053         goto out;
2054 }
2055
2056 static void mld_send_initial_cr(struct inet6_dev *idev)
2057 {
2058         struct sk_buff *skb;
2059         struct ifmcaddr6 *pmc;
2060         int type;
2061
2062         if (mld_in_v1_mode(idev))
2063                 return;
2064
2065         skb = NULL;
2066         read_lock_bh(&idev->lock);
2067         for (pmc = idev->mc_list; pmc; pmc = pmc->next) {
2068                 spin_lock_bh(&pmc->mca_lock);
2069                 if (pmc->mca_sfcount[MCAST_EXCLUDE])
2070                         type = MLD2_CHANGE_TO_EXCLUDE;
2071                 else
2072                         type = MLD2_CHANGE_TO_INCLUDE;
2073                 skb = add_grec(skb, pmc, type, 0, 0, 1);
2074                 spin_unlock_bh(&pmc->mca_lock);
2075         }
2076         read_unlock_bh(&idev->lock);
2077         if (skb)
2078                 mld_sendpack(skb);
2079 }
2080
2081 void ipv6_mc_dad_complete(struct inet6_dev *idev)
2082 {
2083         idev->mc_dad_count = idev->mc_qrv;
2084         if (idev->mc_dad_count) {
2085                 mld_send_initial_cr(idev);
2086                 idev->mc_dad_count--;
2087                 if (idev->mc_dad_count)
2088                         mld_dad_start_timer(idev,
2089                                             unsolicited_report_interval(idev));
2090         }
2091 }
2092
2093 static void mld_dad_timer_expire(unsigned long data)
2094 {
2095         struct inet6_dev *idev = (struct inet6_dev *)data;
2096
2097         mld_send_initial_cr(idev);
2098         if (idev->mc_dad_count) {
2099                 idev->mc_dad_count--;
2100                 if (idev->mc_dad_count)
2101                         mld_dad_start_timer(idev,
2102                                             unsolicited_report_interval(idev));
2103         }
2104         in6_dev_put(idev);
2105 }
2106
2107 static int ip6_mc_del1_src(struct ifmcaddr6 *pmc, int sfmode,
2108         const struct in6_addr *psfsrc)
2109 {
2110         struct ip6_sf_list *psf, *psf_prev;
2111         int rv = 0;
2112
2113         psf_prev = NULL;
2114         for (psf = pmc->mca_sources; psf; psf = psf->sf_next) {
2115                 if (ipv6_addr_equal(&psf->sf_addr, psfsrc))
2116                         break;
2117                 psf_prev = psf;
2118         }
2119         if (!psf || psf->sf_count[sfmode] == 0) {
2120                 /* source filter not found, or count wrong =>  bug */
2121                 return -ESRCH;
2122         }
2123         psf->sf_count[sfmode]--;
2124         if (!psf->sf_count[MCAST_INCLUDE] && !psf->sf_count[MCAST_EXCLUDE]) {
2125                 struct inet6_dev *idev = pmc->idev;
2126
2127                 /* no more filters for this source */
2128                 if (psf_prev)
2129                         psf_prev->sf_next = psf->sf_next;
2130                 else
2131                         pmc->mca_sources = psf->sf_next;
2132                 if (psf->sf_oldin && !(pmc->mca_flags & MAF_NOREPORT) &&
2133                     !mld_in_v1_mode(idev)) {
2134                         psf->sf_crcount = idev->mc_qrv;
2135                         psf->sf_next = pmc->mca_tomb;
2136                         pmc->mca_tomb = psf;
2137                         rv = 1;
2138                 } else
2139                         kfree(psf);
2140         }
2141         return rv;
2142 }
2143
2144 static int ip6_mc_del_src(struct inet6_dev *idev, const struct in6_addr *pmca,
2145                           int sfmode, int sfcount, const struct in6_addr *psfsrc,
2146                           int delta)
2147 {
2148         struct ifmcaddr6 *pmc;
2149         int     changerec = 0;
2150         int     i, err;
2151
2152         if (!idev)
2153                 return -ENODEV;
2154         read_lock_bh(&idev->lock);
2155         for (pmc = idev->mc_list; pmc; pmc = pmc->next) {
2156                 if (ipv6_addr_equal(pmca, &pmc->mca_addr))
2157                         break;
2158         }
2159         if (!pmc) {
2160                 /* MCA not found?? bug */
2161                 read_unlock_bh(&idev->lock);
2162                 return -ESRCH;
2163         }
2164         spin_lock_bh(&pmc->mca_lock);
2165         sf_markstate(pmc);
2166         if (!delta) {
2167                 if (!pmc->mca_sfcount[sfmode]) {
2168                         spin_unlock_bh(&pmc->mca_lock);
2169                         read_unlock_bh(&idev->lock);
2170                         return -EINVAL;
2171                 }
2172                 pmc->mca_sfcount[sfmode]--;
2173         }
2174         err = 0;
2175         for (i = 0; i < sfcount; i++) {
2176                 int rv = ip6_mc_del1_src(pmc, sfmode, &psfsrc[i]);
2177
2178                 changerec |= rv > 0;
2179                 if (!err && rv < 0)
2180                         err = rv;
2181         }
2182         if (pmc->mca_sfmode == MCAST_EXCLUDE &&
2183             pmc->mca_sfcount[MCAST_EXCLUDE] == 0 &&
2184             pmc->mca_sfcount[MCAST_INCLUDE]) {
2185                 struct ip6_sf_list *psf;
2186
2187                 /* filter mode change */
2188                 pmc->mca_sfmode = MCAST_INCLUDE;
2189                 pmc->mca_crcount = idev->mc_qrv;
2190                 idev->mc_ifc_count = pmc->mca_crcount;
2191                 for (psf = pmc->mca_sources; psf; psf = psf->sf_next)
2192                         psf->sf_crcount = 0;
2193                 mld_ifc_event(pmc->idev);
2194         } else if (sf_setstate(pmc) || changerec)
2195                 mld_ifc_event(pmc->idev);
2196         spin_unlock_bh(&pmc->mca_lock);
2197         read_unlock_bh(&idev->lock);
2198         return err;
2199 }
2200
2201 /*
2202  * Add multicast single-source filter to the interface list
2203  */
2204 static int ip6_mc_add1_src(struct ifmcaddr6 *pmc, int sfmode,
2205         const struct in6_addr *psfsrc)
2206 {
2207         struct ip6_sf_list *psf, *psf_prev;
2208
2209         psf_prev = NULL;
2210         for (psf = pmc->mca_sources; psf; psf = psf->sf_next) {
2211                 if (ipv6_addr_equal(&psf->sf_addr, psfsrc))
2212                         break;
2213                 psf_prev = psf;
2214         }
2215         if (!psf) {
2216                 psf = kzalloc(sizeof(*psf), GFP_ATOMIC);
2217                 if (!psf)
2218                         return -ENOBUFS;
2219
2220                 psf->sf_addr = *psfsrc;
2221                 if (psf_prev) {
2222                         psf_prev->sf_next = psf;
2223                 } else
2224                         pmc->mca_sources = psf;
2225         }
2226         psf->sf_count[sfmode]++;
2227         return 0;
2228 }
2229
2230 static void sf_markstate(struct ifmcaddr6 *pmc)
2231 {
2232         struct ip6_sf_list *psf;
2233         int mca_xcount = pmc->mca_sfcount[MCAST_EXCLUDE];
2234
2235         for (psf = pmc->mca_sources; psf; psf = psf->sf_next)
2236                 if (pmc->mca_sfcount[MCAST_EXCLUDE]) {
2237                         psf->sf_oldin = mca_xcount ==
2238                                 psf->sf_count[MCAST_EXCLUDE] &&
2239                                 !psf->sf_count[MCAST_INCLUDE];
2240                 } else
2241                         psf->sf_oldin = psf->sf_count[MCAST_INCLUDE] != 0;
2242 }
2243
2244 static int sf_setstate(struct ifmcaddr6 *pmc)
2245 {
2246         struct ip6_sf_list *psf, *dpsf;
2247         int mca_xcount = pmc->mca_sfcount[MCAST_EXCLUDE];
2248         int qrv = pmc->idev->mc_qrv;
2249         int new_in, rv;
2250
2251         rv = 0;
2252         for (psf = pmc->mca_sources; psf; psf = psf->sf_next) {
2253                 if (pmc->mca_sfcount[MCAST_EXCLUDE]) {
2254                         new_in = mca_xcount == psf->sf_count[MCAST_EXCLUDE] &&
2255                                 !psf->sf_count[MCAST_INCLUDE];
2256                 } else
2257                         new_in = psf->sf_count[MCAST_INCLUDE] != 0;
2258                 if (new_in) {
2259                         if (!psf->sf_oldin) {
2260                                 struct ip6_sf_list *prev = NULL;
2261
2262                                 for (dpsf = pmc->mca_tomb; dpsf;
2263                                      dpsf = dpsf->sf_next) {
2264                                         if (ipv6_addr_equal(&dpsf->sf_addr,
2265                                             &psf->sf_addr))
2266                                                 break;
2267                                         prev = dpsf;
2268                                 }
2269                                 if (dpsf) {
2270                                         if (prev)
2271                                                 prev->sf_next = dpsf->sf_next;
2272                                         else
2273                                                 pmc->mca_tomb = dpsf->sf_next;
2274                                         kfree(dpsf);
2275                                 }
2276                                 psf->sf_crcount = qrv;
2277                                 rv++;
2278                         }
2279                 } else if (psf->sf_oldin) {
2280                         psf->sf_crcount = 0;
2281                         /*
2282                          * add or update "delete" records if an active filter
2283                          * is now inactive
2284                          */
2285                         for (dpsf = pmc->mca_tomb; dpsf; dpsf = dpsf->sf_next)
2286                                 if (ipv6_addr_equal(&dpsf->sf_addr,
2287                                     &psf->sf_addr))
2288                                         break;
2289                         if (!dpsf) {
2290                                 dpsf = kmalloc(sizeof(*dpsf), GFP_ATOMIC);
2291                                 if (!dpsf)
2292                                         continue;
2293                                 *dpsf = *psf;
2294                                 /* pmc->mca_lock held by callers */
2295                                 dpsf->sf_next = pmc->mca_tomb;
2296                                 pmc->mca_tomb = dpsf;
2297                         }
2298                         dpsf->sf_crcount = qrv;
2299                         rv++;
2300                 }
2301         }
2302         return rv;
2303 }
2304
2305 /*
2306  * Add multicast source filter list to the interface list
2307  */
2308 static int ip6_mc_add_src(struct inet6_dev *idev, const struct in6_addr *pmca,
2309                           int sfmode, int sfcount, const struct in6_addr *psfsrc,
2310                           int delta)
2311 {
2312         struct ifmcaddr6 *pmc;
2313         int     isexclude;
2314         int     i, err;
2315
2316         if (!idev)
2317                 return -ENODEV;
2318         read_lock_bh(&idev->lock);
2319         for (pmc = idev->mc_list; pmc; pmc = pmc->next) {
2320                 if (ipv6_addr_equal(pmca, &pmc->mca_addr))
2321                         break;
2322         }
2323         if (!pmc) {
2324                 /* MCA not found?? bug */
2325                 read_unlock_bh(&idev->lock);
2326                 return -ESRCH;
2327         }
2328         spin_lock_bh(&pmc->mca_lock);
2329
2330         sf_markstate(pmc);
2331         isexclude = pmc->mca_sfmode == MCAST_EXCLUDE;
2332         if (!delta)
2333                 pmc->mca_sfcount[sfmode]++;
2334         err = 0;
2335         for (i = 0; i < sfcount; i++) {
2336                 err = ip6_mc_add1_src(pmc, sfmode, &psfsrc[i]);
2337                 if (err)
2338                         break;
2339         }
2340         if (err) {
2341                 int j;
2342
2343                 if (!delta)
2344                         pmc->mca_sfcount[sfmode]--;
2345                 for (j = 0; j < i; j++)
2346                         ip6_mc_del1_src(pmc, sfmode, &psfsrc[j]);
2347         } else if (isexclude != (pmc->mca_sfcount[MCAST_EXCLUDE] != 0)) {
2348                 struct ip6_sf_list *psf;
2349
2350                 /* filter mode change */
2351                 if (pmc->mca_sfcount[MCAST_EXCLUDE])
2352                         pmc->mca_sfmode = MCAST_EXCLUDE;
2353                 else if (pmc->mca_sfcount[MCAST_INCLUDE])
2354                         pmc->mca_sfmode = MCAST_INCLUDE;
2355                 /* else no filters; keep old mode for reports */
2356
2357                 pmc->mca_crcount = idev->mc_qrv;
2358                 idev->mc_ifc_count = pmc->mca_crcount;
2359                 for (psf = pmc->mca_sources; psf; psf = psf->sf_next)
2360                         psf->sf_crcount = 0;
2361                 mld_ifc_event(idev);
2362         } else if (sf_setstate(pmc))
2363                 mld_ifc_event(idev);
2364         spin_unlock_bh(&pmc->mca_lock);
2365         read_unlock_bh(&idev->lock);
2366         return err;
2367 }
2368
2369 static void ip6_mc_clear_src(struct ifmcaddr6 *pmc)
2370 {
2371         struct ip6_sf_list *psf, *nextpsf;
2372
2373         for (psf = pmc->mca_tomb; psf; psf = nextpsf) {
2374                 nextpsf = psf->sf_next;
2375                 kfree(psf);
2376         }
2377         pmc->mca_tomb = NULL;
2378         for (psf = pmc->mca_sources; psf; psf = nextpsf) {
2379                 nextpsf = psf->sf_next;
2380                 kfree(psf);
2381         }
2382         pmc->mca_sources = NULL;
2383         pmc->mca_sfmode = MCAST_EXCLUDE;
2384         pmc->mca_sfcount[MCAST_INCLUDE] = 0;
2385         pmc->mca_sfcount[MCAST_EXCLUDE] = 1;
2386 }
2387
2388
2389 static void igmp6_join_group(struct ifmcaddr6 *ma)
2390 {
2391         unsigned long delay;
2392
2393         if (ma->mca_flags & MAF_NOREPORT)
2394                 return;
2395
2396         igmp6_send(&ma->mca_addr, ma->idev->dev, ICMPV6_MGM_REPORT);
2397
2398         delay = prandom_u32() % unsolicited_report_interval(ma->idev);
2399
2400         spin_lock_bh(&ma->mca_lock);
2401         if (del_timer(&ma->mca_timer)) {
2402                 atomic_dec(&ma->mca_refcnt);
2403                 delay = ma->mca_timer.expires - jiffies;
2404         }
2405
2406         if (!mod_timer(&ma->mca_timer, jiffies + delay))
2407                 atomic_inc(&ma->mca_refcnt);
2408         ma->mca_flags |= MAF_TIMER_RUNNING | MAF_LAST_REPORTER;
2409         spin_unlock_bh(&ma->mca_lock);
2410 }
2411
2412 static int ip6_mc_leave_src(struct sock *sk, struct ipv6_mc_socklist *iml,
2413                             struct inet6_dev *idev)
2414 {
2415         int err;
2416
2417         write_lock_bh(&iml->sflock);
2418         if (!iml->sflist) {
2419                 /* any-source empty exclude case */
2420                 err = ip6_mc_del_src(idev, &iml->addr, iml->sfmode, 0, NULL, 0);
2421         } else {
2422                 err = ip6_mc_del_src(idev, &iml->addr, iml->sfmode,
2423                                 iml->sflist->sl_count, iml->sflist->sl_addr, 0);
2424                 sock_kfree_s(sk, iml->sflist, IP6_SFLSIZE(iml->sflist->sl_max));
2425                 iml->sflist = NULL;
2426         }
2427         write_unlock_bh(&iml->sflock);
2428         return err;
2429 }
2430
2431 static void igmp6_leave_group(struct ifmcaddr6 *ma)
2432 {
2433         if (mld_in_v1_mode(ma->idev)) {
2434                 if (ma->mca_flags & MAF_LAST_REPORTER)
2435                         igmp6_send(&ma->mca_addr, ma->idev->dev,
2436                                 ICMPV6_MGM_REDUCTION);
2437         } else {
2438                 mld_add_delrec(ma->idev, ma);
2439                 mld_ifc_event(ma->idev);
2440         }
2441 }
2442
2443 static void mld_gq_timer_expire(unsigned long data)
2444 {
2445         struct inet6_dev *idev = (struct inet6_dev *)data;
2446
2447         idev->mc_gq_running = 0;
2448         mld_send_report(idev, NULL);
2449         in6_dev_put(idev);
2450 }
2451
2452 static void mld_ifc_timer_expire(unsigned long data)
2453 {
2454         struct inet6_dev *idev = (struct inet6_dev *)data;
2455
2456         mld_send_cr(idev);
2457         if (idev->mc_ifc_count) {
2458                 idev->mc_ifc_count--;
2459                 if (idev->mc_ifc_count)
2460                         mld_ifc_start_timer(idev,
2461                                             unsolicited_report_interval(idev));
2462         }
2463         in6_dev_put(idev);
2464 }
2465
2466 static void mld_ifc_event(struct inet6_dev *idev)
2467 {
2468         if (mld_in_v1_mode(idev))
2469                 return;
2470         idev->mc_ifc_count = idev->mc_qrv;
2471         mld_ifc_start_timer(idev, 1);
2472 }
2473
2474
2475 static void igmp6_timer_handler(unsigned long data)
2476 {
2477         struct ifmcaddr6 *ma = (struct ifmcaddr6 *) data;
2478
2479         if (mld_in_v1_mode(ma->idev))
2480                 igmp6_send(&ma->mca_addr, ma->idev->dev, ICMPV6_MGM_REPORT);
2481         else
2482                 mld_send_report(ma->idev, ma);
2483
2484         spin_lock(&ma->mca_lock);
2485         ma->mca_flags |=  MAF_LAST_REPORTER;
2486         ma->mca_flags &= ~MAF_TIMER_RUNNING;
2487         spin_unlock(&ma->mca_lock);
2488         ma_put(ma);
2489 }
2490
2491 /* Device changing type */
2492
2493 void ipv6_mc_unmap(struct inet6_dev *idev)
2494 {
2495         struct ifmcaddr6 *i;
2496
2497         /* Install multicast list, except for all-nodes (already installed) */
2498
2499         read_lock_bh(&idev->lock);
2500         for (i = idev->mc_list; i; i = i->next)
2501                 igmp6_group_dropped(i);
2502         read_unlock_bh(&idev->lock);
2503 }
2504
2505 void ipv6_mc_remap(struct inet6_dev *idev)
2506 {
2507         ipv6_mc_up(idev);
2508 }
2509
2510 /* Device going down */
2511
2512 void ipv6_mc_down(struct inet6_dev *idev)
2513 {
2514         struct ifmcaddr6 *i;
2515
2516         /* Withdraw multicast list */
2517
2518         read_lock_bh(&idev->lock);
2519
2520         for (i = idev->mc_list; i; i = i->next)
2521                 igmp6_group_dropped(i);
2522
2523         /* Should stop timer after group drop. or we will
2524          * start timer again in mld_ifc_event()
2525          */
2526         mld_ifc_stop_timer(idev);
2527         mld_gq_stop_timer(idev);
2528         mld_dad_stop_timer(idev);
2529         read_unlock_bh(&idev->lock);
2530 }
2531
2532 static void ipv6_mc_reset(struct inet6_dev *idev)
2533 {
2534         idev->mc_qrv = sysctl_mld_qrv;
2535         idev->mc_qi = MLD_QI_DEFAULT;
2536         idev->mc_qri = MLD_QRI_DEFAULT;
2537         idev->mc_v1_seen = 0;
2538         idev->mc_maxdelay = unsolicited_report_interval(idev);
2539 }
2540
2541 /* Device going up */
2542
2543 void ipv6_mc_up(struct inet6_dev *idev)
2544 {
2545         struct ifmcaddr6 *i;
2546
2547         /* Install multicast list, except for all-nodes (already installed) */
2548
2549         read_lock_bh(&idev->lock);
2550         ipv6_mc_reset(idev);
2551         for (i = idev->mc_list; i; i = i->next) {
2552                 mld_del_delrec(idev, i);
2553                 igmp6_group_added(i);
2554         }
2555         read_unlock_bh(&idev->lock);
2556 }
2557
2558 /* IPv6 device initialization. */
2559
2560 void ipv6_mc_init_dev(struct inet6_dev *idev)
2561 {
2562         write_lock_bh(&idev->lock);
2563         spin_lock_init(&idev->mc_lock);
2564         idev->mc_gq_running = 0;
2565         setup_timer(&idev->mc_gq_timer, mld_gq_timer_expire,
2566                         (unsigned long)idev);
2567         idev->mc_tomb = NULL;
2568         idev->mc_ifc_count = 0;
2569         setup_timer(&idev->mc_ifc_timer, mld_ifc_timer_expire,
2570                         (unsigned long)idev);
2571         setup_timer(&idev->mc_dad_timer, mld_dad_timer_expire,
2572                     (unsigned long)idev);
2573         ipv6_mc_reset(idev);
2574         write_unlock_bh(&idev->lock);
2575 }
2576
2577 /*
2578  *      Device is about to be destroyed: clean up.
2579  */
2580
2581 void ipv6_mc_destroy_dev(struct inet6_dev *idev)
2582 {
2583         struct ifmcaddr6 *i;
2584
2585         /* Deactivate timers */
2586         ipv6_mc_down(idev);
2587         mld_clear_delrec(idev);
2588
2589         /* Delete all-nodes address. */
2590         /* We cannot call ipv6_dev_mc_dec() directly, our caller in
2591          * addrconf.c has NULL'd out dev->ip6_ptr so in6_dev_get() will
2592          * fail.
2593          */
2594         __ipv6_dev_mc_dec(idev, &in6addr_linklocal_allnodes);
2595
2596         if (idev->cnf.forwarding)
2597                 __ipv6_dev_mc_dec(idev, &in6addr_linklocal_allrouters);
2598
2599         write_lock_bh(&idev->lock);
2600         while ((i = idev->mc_list) != NULL) {
2601                 idev->mc_list = i->next;
2602
2603                 write_unlock_bh(&idev->lock);
2604                 ma_put(i);
2605                 write_lock_bh(&idev->lock);
2606         }
2607         write_unlock_bh(&idev->lock);
2608 }
2609
2610 #ifdef CONFIG_PROC_FS
2611 struct igmp6_mc_iter_state {
2612         struct seq_net_private p;
2613         struct net_device *dev;
2614         struct inet6_dev *idev;
2615 };
2616
2617 #define igmp6_mc_seq_private(seq)       ((struct igmp6_mc_iter_state *)(seq)->private)
2618
2619 static inline struct ifmcaddr6 *igmp6_mc_get_first(struct seq_file *seq)
2620 {
2621         struct ifmcaddr6 *im = NULL;
2622         struct igmp6_mc_iter_state *state = igmp6_mc_seq_private(seq);
2623         struct net *net = seq_file_net(seq);
2624
2625         state->idev = NULL;
2626         for_each_netdev_rcu(net, state->dev) {
2627                 struct inet6_dev *idev;
2628                 idev = __in6_dev_get(state->dev);
2629                 if (!idev)
2630                         continue;
2631                 read_lock_bh(&idev->lock);
2632                 im = idev->mc_list;
2633                 if (im) {
2634                         state->idev = idev;
2635                         break;
2636                 }
2637                 read_unlock_bh(&idev->lock);
2638         }
2639         return im;
2640 }
2641
2642 static struct ifmcaddr6 *igmp6_mc_get_next(struct seq_file *seq, struct ifmcaddr6 *im)
2643 {
2644         struct igmp6_mc_iter_state *state = igmp6_mc_seq_private(seq);
2645
2646         im = im->next;
2647         while (!im) {
2648                 if (likely(state->idev))
2649                         read_unlock_bh(&state->idev->lock);
2650
2651                 state->dev = next_net_device_rcu(state->dev);
2652                 if (!state->dev) {
2653                         state->idev = NULL;
2654                         break;
2655                 }
2656                 state->idev = __in6_dev_get(state->dev);
2657                 if (!state->idev)
2658                         continue;
2659                 read_lock_bh(&state->idev->lock);
2660                 im = state->idev->mc_list;
2661         }
2662         return im;
2663 }
2664
2665 static struct ifmcaddr6 *igmp6_mc_get_idx(struct seq_file *seq, loff_t pos)
2666 {
2667         struct ifmcaddr6 *im = igmp6_mc_get_first(seq);
2668         if (im)
2669                 while (pos && (im = igmp6_mc_get_next(seq, im)) != NULL)
2670                         --pos;
2671         return pos ? NULL : im;
2672 }
2673
2674 static void *igmp6_mc_seq_start(struct seq_file *seq, loff_t *pos)
2675         __acquires(RCU)
2676 {
2677         rcu_read_lock();
2678         return igmp6_mc_get_idx(seq, *pos);
2679 }
2680
2681 static void *igmp6_mc_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2682 {
2683         struct ifmcaddr6 *im = igmp6_mc_get_next(seq, v);
2684
2685         ++*pos;
2686         return im;
2687 }
2688
2689 static void igmp6_mc_seq_stop(struct seq_file *seq, void *v)
2690         __releases(RCU)
2691 {
2692         struct igmp6_mc_iter_state *state = igmp6_mc_seq_private(seq);
2693
2694         if (likely(state->idev)) {
2695                 read_unlock_bh(&state->idev->lock);
2696                 state->idev = NULL;
2697         }
2698         state->dev = NULL;
2699         rcu_read_unlock();
2700 }
2701
2702 static int igmp6_mc_seq_show(struct seq_file *seq, void *v)
2703 {
2704         struct ifmcaddr6 *im = (struct ifmcaddr6 *)v;
2705         struct igmp6_mc_iter_state *state = igmp6_mc_seq_private(seq);
2706
2707         seq_printf(seq,
2708                    "%-4d %-15s %pi6 %5d %08X %ld\n",
2709                    state->dev->ifindex, state->dev->name,
2710                    &im->mca_addr,
2711                    im->mca_users, im->mca_flags,
2712                    (im->mca_flags&MAF_TIMER_RUNNING) ?
2713                    jiffies_to_clock_t(im->mca_timer.expires-jiffies) : 0);
2714         return 0;
2715 }
2716
2717 static const struct seq_operations igmp6_mc_seq_ops = {
2718         .start  =       igmp6_mc_seq_start,
2719         .next   =       igmp6_mc_seq_next,
2720         .stop   =       igmp6_mc_seq_stop,
2721         .show   =       igmp6_mc_seq_show,
2722 };
2723
2724 static int igmp6_mc_seq_open(struct inode *inode, struct file *file)
2725 {
2726         return seq_open_net(inode, file, &igmp6_mc_seq_ops,
2727                             sizeof(struct igmp6_mc_iter_state));
2728 }
2729
2730 static const struct file_operations igmp6_mc_seq_fops = {
2731         .owner          =       THIS_MODULE,
2732         .open           =       igmp6_mc_seq_open,
2733         .read           =       seq_read,
2734         .llseek         =       seq_lseek,
2735         .release        =       seq_release_net,
2736 };
2737
2738 struct igmp6_mcf_iter_state {
2739         struct seq_net_private p;
2740         struct net_device *dev;
2741         struct inet6_dev *idev;
2742         struct ifmcaddr6 *im;
2743 };
2744
2745 #define igmp6_mcf_seq_private(seq)      ((struct igmp6_mcf_iter_state *)(seq)->private)
2746
2747 static inline struct ip6_sf_list *igmp6_mcf_get_first(struct seq_file *seq)
2748 {
2749         struct ip6_sf_list *psf = NULL;
2750         struct ifmcaddr6 *im = NULL;
2751         struct igmp6_mcf_iter_state *state = igmp6_mcf_seq_private(seq);
2752         struct net *net = seq_file_net(seq);
2753
2754         state->idev = NULL;
2755         state->im = NULL;
2756         for_each_netdev_rcu(net, state->dev) {
2757                 struct inet6_dev *idev;
2758                 idev = __in6_dev_get(state->dev);
2759                 if (unlikely(idev == NULL))
2760                         continue;
2761                 read_lock_bh(&idev->lock);
2762                 im = idev->mc_list;
2763                 if (likely(im)) {
2764                         spin_lock_bh(&im->mca_lock);
2765                         psf = im->mca_sources;
2766                         if (likely(psf)) {
2767                                 state->im = im;
2768                                 state->idev = idev;
2769                                 break;
2770                         }
2771                         spin_unlock_bh(&im->mca_lock);
2772                 }
2773                 read_unlock_bh(&idev->lock);
2774         }
2775         return psf;
2776 }
2777
2778 static struct ip6_sf_list *igmp6_mcf_get_next(struct seq_file *seq, struct ip6_sf_list *psf)
2779 {
2780         struct igmp6_mcf_iter_state *state = igmp6_mcf_seq_private(seq);
2781
2782         psf = psf->sf_next;
2783         while (!psf) {
2784                 spin_unlock_bh(&state->im->mca_lock);
2785                 state->im = state->im->next;
2786                 while (!state->im) {
2787                         if (likely(state->idev))
2788                                 read_unlock_bh(&state->idev->lock);
2789
2790                         state->dev = next_net_device_rcu(state->dev);
2791                         if (!state->dev) {
2792                                 state->idev = NULL;
2793                                 goto out;
2794                         }
2795                         state->idev = __in6_dev_get(state->dev);
2796                         if (!state->idev)
2797                                 continue;
2798                         read_lock_bh(&state->idev->lock);
2799                         state->im = state->idev->mc_list;
2800                 }
2801                 if (!state->im)
2802                         break;
2803                 spin_lock_bh(&state->im->mca_lock);
2804                 psf = state->im->mca_sources;
2805         }
2806 out:
2807         return psf;
2808 }
2809
2810 static struct ip6_sf_list *igmp6_mcf_get_idx(struct seq_file *seq, loff_t pos)
2811 {
2812         struct ip6_sf_list *psf = igmp6_mcf_get_first(seq);
2813         if (psf)
2814                 while (pos && (psf = igmp6_mcf_get_next(seq, psf)) != NULL)
2815                         --pos;
2816         return pos ? NULL : psf;
2817 }
2818
2819 static void *igmp6_mcf_seq_start(struct seq_file *seq, loff_t *pos)
2820         __acquires(RCU)
2821 {
2822         rcu_read_lock();
2823         return *pos ? igmp6_mcf_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
2824 }
2825
2826 static void *igmp6_mcf_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2827 {
2828         struct ip6_sf_list *psf;
2829         if (v == SEQ_START_TOKEN)
2830                 psf = igmp6_mcf_get_first(seq);
2831         else
2832                 psf = igmp6_mcf_get_next(seq, v);
2833         ++*pos;
2834         return psf;
2835 }
2836
2837 static void igmp6_mcf_seq_stop(struct seq_file *seq, void *v)
2838         __releases(RCU)
2839 {
2840         struct igmp6_mcf_iter_state *state = igmp6_mcf_seq_private(seq);
2841         if (likely(state->im)) {
2842                 spin_unlock_bh(&state->im->mca_lock);
2843                 state->im = NULL;
2844         }
2845         if (likely(state->idev)) {
2846                 read_unlock_bh(&state->idev->lock);
2847                 state->idev = NULL;
2848         }
2849         state->dev = NULL;
2850         rcu_read_unlock();
2851 }
2852
2853 static int igmp6_mcf_seq_show(struct seq_file *seq, void *v)
2854 {
2855         struct ip6_sf_list *psf = (struct ip6_sf_list *)v;
2856         struct igmp6_mcf_iter_state *state = igmp6_mcf_seq_private(seq);
2857
2858         if (v == SEQ_START_TOKEN) {
2859                 seq_puts(seq, "Idx Device                Multicast Address                   Source Address    INC    EXC\n");
2860         } else {
2861                 seq_printf(seq,
2862                            "%3d %6.6s %pi6 %pi6 %6lu %6lu\n",
2863                            state->dev->ifindex, state->dev->name,
2864                            &state->im->mca_addr,
2865                            &psf->sf_addr,
2866                            psf->sf_count[MCAST_INCLUDE],
2867                            psf->sf_count[MCAST_EXCLUDE]);
2868         }
2869         return 0;
2870 }
2871
2872 static const struct seq_operations igmp6_mcf_seq_ops = {
2873         .start  =       igmp6_mcf_seq_start,
2874         .next   =       igmp6_mcf_seq_next,
2875         .stop   =       igmp6_mcf_seq_stop,
2876         .show   =       igmp6_mcf_seq_show,
2877 };
2878
2879 static int igmp6_mcf_seq_open(struct inode *inode, struct file *file)
2880 {
2881         return seq_open_net(inode, file, &igmp6_mcf_seq_ops,
2882                             sizeof(struct igmp6_mcf_iter_state));
2883 }
2884
2885 static const struct file_operations igmp6_mcf_seq_fops = {
2886         .owner          =       THIS_MODULE,
2887         .open           =       igmp6_mcf_seq_open,
2888         .read           =       seq_read,
2889         .llseek         =       seq_lseek,
2890         .release        =       seq_release_net,
2891 };
2892
2893 static int __net_init igmp6_proc_init(struct net *net)
2894 {
2895         int err;
2896
2897         err = -ENOMEM;
2898         if (!proc_create("igmp6", S_IRUGO, net->proc_net, &igmp6_mc_seq_fops))
2899                 goto out;
2900         if (!proc_create("mcfilter6", S_IRUGO, net->proc_net,
2901                          &igmp6_mcf_seq_fops))
2902                 goto out_proc_net_igmp6;
2903
2904         err = 0;
2905 out:
2906         return err;
2907
2908 out_proc_net_igmp6:
2909         remove_proc_entry("igmp6", net->proc_net);
2910         goto out;
2911 }
2912
2913 static void __net_exit igmp6_proc_exit(struct net *net)
2914 {
2915         remove_proc_entry("mcfilter6", net->proc_net);
2916         remove_proc_entry("igmp6", net->proc_net);
2917 }
2918 #else
2919 static inline int igmp6_proc_init(struct net *net)
2920 {
2921         return 0;
2922 }
2923 static inline void igmp6_proc_exit(struct net *net)
2924 {
2925 }
2926 #endif
2927
2928 static int __net_init igmp6_net_init(struct net *net)
2929 {
2930         int err;
2931
2932         err = inet_ctl_sock_create(&net->ipv6.igmp_sk, PF_INET6,
2933                                    SOCK_RAW, IPPROTO_ICMPV6, net);
2934         if (err < 0) {
2935                 pr_err("Failed to initialize the IGMP6 control socket (err %d)\n",
2936                        err);
2937                 goto out;
2938         }
2939
2940         inet6_sk(net->ipv6.igmp_sk)->hop_limit = 1;
2941
2942         err = inet_ctl_sock_create(&net->ipv6.mc_autojoin_sk, PF_INET6,
2943                                    SOCK_RAW, IPPROTO_ICMPV6, net);
2944         if (err < 0) {
2945                 pr_err("Failed to initialize the IGMP6 autojoin socket (err %d)\n",
2946                        err);
2947                 goto out_sock_create;
2948         }
2949
2950         err = igmp6_proc_init(net);
2951         if (err)
2952                 goto out_sock_create_autojoin;
2953
2954         return 0;
2955
2956 out_sock_create_autojoin:
2957         inet_ctl_sock_destroy(net->ipv6.mc_autojoin_sk);
2958 out_sock_create:
2959         inet_ctl_sock_destroy(net->ipv6.igmp_sk);
2960 out:
2961         return err;
2962 }
2963
2964 static void __net_exit igmp6_net_exit(struct net *net)
2965 {
2966         inet_ctl_sock_destroy(net->ipv6.igmp_sk);
2967         inet_ctl_sock_destroy(net->ipv6.mc_autojoin_sk);
2968         igmp6_proc_exit(net);
2969 }
2970
2971 static struct pernet_operations igmp6_net_ops = {
2972         .init = igmp6_net_init,
2973         .exit = igmp6_net_exit,
2974 };
2975
2976 int __init igmp6_init(void)
2977 {
2978         return register_pernet_subsys(&igmp6_net_ops);
2979 }
2980
2981 void igmp6_cleanup(void)
2982 {
2983         unregister_pernet_subsys(&igmp6_net_ops);
2984 }