net/ipv6/mcast.c

   1 // SPDX-License-Identifier: GPL-2.0-or-later
   2 /*
   3  *      Multicast support for IPv6
   4  *      Linux INET6 implementation
   5  *
   6  *      Authors:
   7  *      Pedro Roque             <roque@di.fc.ul.pt>
   8  *
   9  *      Based on linux/ipv4/igmp.c and linux/ipv4/ip_sockglue.c
  10  */
  11
  12 /* Changes:
  13  *
  14  *      yoshfuji        : fix format of router-alert option
  15  *      YOSHIFUJI Hideaki @USAGI:
  16  *              Fixed source address for MLD message based on
  17  *              <draft-ietf-magma-mld-source-05.txt>.
  18  *      YOSHIFUJI Hideaki @USAGI:
  19  *              - Ignore Queries for invalid addresses.
  20  *              - MLD for link-local addresses.
  21  *      David L Stevens <dlstevens@us.ibm.com>:
  22  *              - MLDv2 support
  23  */
  24
  25 #include <linux/module.h>
  26 #include <linux/errno.h>
  27 #include <linux/types.h>
  28 #include <linux/string.h>
  29 #include <linux/socket.h>
  30 #include <linux/sockios.h>
  31 #include <linux/jiffies.h>
  32 #include <linux/times.h>
  33 #include <linux/net.h>
  34 #include <linux/in.h>
  35 #include <linux/in6.h>
  36 #include <linux/netdevice.h>
  37 #include <linux/if_arp.h>
  38 #include <linux/route.h>
  39 #include <linux/init.h>
  40 #include <linux/proc_fs.h>
  41 #include <linux/seq_file.h>
  42 #include <linux/slab.h>
  43 #include <linux/pkt_sched.h>
  44 #include <net/mld.h>
  45
  46 #include <linux/netfilter.h>
  47 #include <linux/netfilter_ipv6.h>
  48
  49 #include <net/net_namespace.h>
  50 #include <net/sock.h>
  51 #include <net/snmp.h>
  52
  53 #include <net/ipv6.h>
  54 #include <net/protocol.h>
  55 #include <net/if_inet6.h>
  56 #include <net/ndisc.h>
  57 #include <net/addrconf.h>
  58 #include <net/ip6_route.h>
  59 #include <net/inet_common.h>
  60
  61 #include <net/ip6_checksum.h>
  62
  63 /* Ensure that we have struct in6_addr aligned on 32bit word. */
  64 static int __mld2_query_bugs[] __attribute__((__unused__)) = {
  65         BUILD_BUG_ON_ZERO(offsetof(struct mld2_query, mld2q_srcs) % 4),
  66         BUILD_BUG_ON_ZERO(offsetof(struct mld2_report, mld2r_grec) % 4),
  67         BUILD_BUG_ON_ZERO(offsetof(struct mld2_grec, grec_mca) % 4)
  68 };
  69
  70 static struct in6_addr mld2_all_mcr = MLD2_ALL_MCR_INIT;
  71
  72 static void igmp6_join_group(struct ifmcaddr6 *ma);
  73 static void igmp6_leave_group(struct ifmcaddr6 *ma);
  74 static void igmp6_timer_handler(struct timer_list *t);
  75
  76 static void mld_gq_timer_expire(struct timer_list *t);
  77 static void mld_ifc_timer_expire(struct timer_list *t);
  78 static void mld_ifc_event(struct inet6_dev *idev);
  79 static void mld_add_delrec(struct inet6_dev *idev, struct ifmcaddr6 *pmc);
  80 static void mld_del_delrec(struct inet6_dev *idev, struct ifmcaddr6 *pmc);
  81 static void mld_clear_delrec(struct inet6_dev *idev);
  82 static bool mld_in_v1_mode(const struct inet6_dev *idev);
  83 static int sf_setstate(struct ifmcaddr6 *pmc);
  84 static void sf_markstate(struct ifmcaddr6 *pmc);
  85 static void ip6_mc_clear_src(struct ifmcaddr6 *pmc);
  86 static int ip6_mc_del_src(struct inet6_dev *idev, const struct in6_addr *pmca,
  87                           int sfmode, int sfcount, const struct in6_addr *psfsrc,
  88                           int delta);
  89 static int ip6_mc_add_src(struct inet6_dev *idev, const struct in6_addr *pmca,
  90                           int sfmode, int sfcount, const struct in6_addr *psfsrc,
  91                           int delta);
  92 static int ip6_mc_leave_src(struct sock *sk, struct ipv6_mc_socklist *iml,
  93                             struct inet6_dev *idev);
  94 static int __ipv6_dev_mc_inc(struct net_device *dev,
  95                              const struct in6_addr *addr, unsigned int mode);
  96
  97 #define MLD_QRV_DEFAULT         2
  98 /* RFC3810, 9.2. Query Interval */
  99 #define MLD_QI_DEFAULT          (125 * HZ)
 100 /* RFC3810, 9.3. Query Response Interval */
 101 #define MLD_QRI_DEFAULT         (10 * HZ)
 102
 103 /* RFC3810, 8.1 Query Version Distinctions */
 104 #define MLD_V1_QUERY_LEN        24
 105 #define MLD_V2_QUERY_LEN_MIN    28
 106
 107 #define IPV6_MLD_MAX_MSF        64
 108
 109 int sysctl_mld_max_msf __read_mostly = IPV6_MLD_MAX_MSF;
 110 int sysctl_mld_qrv __read_mostly = MLD_QRV_DEFAULT;
 111
 112 /*
 113  *      socket join on multicast group
 114  */
 115
 116 #define for_each_pmc_rcu(np, pmc)                               \
 117         for (pmc = rcu_dereference(np->ipv6_mc_list);           \
 118              pmc != NULL;                                       \
 119              pmc = rcu_dereference(pmc->next))
 120
 121 static int unsolicited_report_interval(struct inet6_dev *idev)
 122 {
 123         int iv;
 124
 125         if (mld_in_v1_mode(idev))
 126                 iv = idev->cnf.mldv1_unsolicited_report_interval;
 127         else
 128                 iv = idev->cnf.mldv2_unsolicited_report_interval;
 129
 130         return iv > 0 ? iv : 1;
 131 }
 132
 133 static int __ipv6_sock_mc_join(struct sock *sk, int ifindex,
 134                                const struct in6_addr *addr, unsigned int mode)
 135 {
 136         struct net_device *dev = NULL;
 137         struct ipv6_mc_socklist *mc_lst;
 138         struct ipv6_pinfo *np = inet6_sk(sk);
 139         struct net *net = sock_net(sk);
 140         int err;
 141
 142         ASSERT_RTNL();
 143
 144         if (!ipv6_addr_is_multicast(addr))
 145                 return -EINVAL;
 146
 147         rcu_read_lock();
 148         for_each_pmc_rcu(np, mc_lst) {
 149                 if ((ifindex == 0 || mc_lst->ifindex == ifindex) &&
 150                     ipv6_addr_equal(&mc_lst->addr, addr)) {
 151                         rcu_read_unlock();
 152                         return -EADDRINUSE;
 153                 }
 154         }
 155         rcu_read_unlock();
 156
 157         mc_lst = sock_kmalloc(sk, sizeof(struct ipv6_mc_socklist), GFP_KERNEL);
 158
 159         if (!mc_lst)
 160                 return -ENOMEM;
 161
 162         mc_lst->next = NULL;
 163         mc_lst->addr = *addr;
 164
 165         if (ifindex == 0) {
 166                 struct rt6_info *rt;
 167                 rt = rt6_lookup(net, addr, NULL, 0, NULL, 0);
 168                 if (rt) {
 169                         dev = rt->dst.dev;
 170                         ip6_rt_put(rt);
 171                 }
 172         } else
 173                 dev = __dev_get_by_index(net, ifindex);
 174
 175         if (!dev) {
 176                 sock_kfree_s(sk, mc_lst, sizeof(*mc_lst));
 177                 return -ENODEV;
 178         }
 179
 180         mc_lst->ifindex = dev->ifindex;
 181         mc_lst->sfmode = mode;
 182         rwlock_init(&mc_lst->sflock);
 183         mc_lst->sflist = NULL;
 184
 185         /*
 186          *      now add/increase the group membership on the device
 187          */
 188
 189         err = __ipv6_dev_mc_inc(dev, addr, mode);
 190
 191         if (err) {
 192                 sock_kfree_s(sk, mc_lst, sizeof(*mc_lst));
 193                 return err;
 194         }
 195
 196         mc_lst->next = np->ipv6_mc_list;
 197         rcu_assign_pointer(np->ipv6_mc_list, mc_lst);
 198
 199         return 0;
 200 }
 201
 202 int ipv6_sock_mc_join(struct sock *sk, int ifindex, const struct in6_addr *addr)
 203 {
 204         return __ipv6_sock_mc_join(sk, ifindex, addr, MCAST_EXCLUDE);
 205 }
 206 EXPORT_SYMBOL(ipv6_sock_mc_join);
 207
 208 int ipv6_sock_mc_join_ssm(struct sock *sk, int ifindex,
 209                           const struct in6_addr *addr, unsigned int mode)
 210 {
 211         return __ipv6_sock_mc_join(sk, ifindex, addr, mode);
 212 }
 213
 214 /*
 215  *      socket leave on multicast group
 216  */
 217 int ipv6_sock_mc_drop(struct sock *sk, int ifindex, const struct in6_addr *addr)
 218 {
 219         struct ipv6_pinfo *np = inet6_sk(sk);
 220         struct ipv6_mc_socklist *mc_lst;
 221         struct ipv6_mc_socklist __rcu **lnk;
 222         struct net *net = sock_net(sk);
 223
 224         ASSERT_RTNL();
 225
 226         if (!ipv6_addr_is_multicast(addr))
 227                 return -EINVAL;
 228
 229         for (lnk = &np->ipv6_mc_list;
 230              (mc_lst = rtnl_dereference(*lnk)) != NULL;
 231               lnk = &mc_lst->next) {
 232                 if ((ifindex == 0 || mc_lst->ifindex == ifindex) &&
 233                     ipv6_addr_equal(&mc_lst->addr, addr)) {
 234                         struct net_device *dev;
 235
 236                         *lnk = mc_lst->next;
 237
 238                         dev = __dev_get_by_index(net, mc_lst->ifindex);
 239                         if (dev) {
 240                                 struct inet6_dev *idev = __in6_dev_get(dev);
 241
 242                                 (void) ip6_mc_leave_src(sk, mc_lst, idev);
 243                                 if (idev)
 244                                         __ipv6_dev_mc_dec(idev, &mc_lst->addr);
 245                         } else
 246                                 (void) ip6_mc_leave_src(sk, mc_lst, NULL);
 247
 248                         atomic_sub(sizeof(*mc_lst), &sk->sk_omem_alloc);
 249                         kfree_rcu(mc_lst, rcu);
 250                         return 0;
 251                 }
 252         }
 253
 254         return -EADDRNOTAVAIL;
 255 }
 256 EXPORT_SYMBOL(ipv6_sock_mc_drop);
 257
 258 /* called with rcu_read_lock() */
 259 static struct inet6_dev *ip6_mc_find_dev_rcu(struct net *net,
 260                                              const struct in6_addr *group,
 261                                              int ifindex)
 262 {
 263         struct net_device *dev = NULL;
 264         struct inet6_dev *idev = NULL;
 265
 266         if (ifindex == 0) {
 267                 struct rt6_info *rt = rt6_lookup(net, group, NULL, 0, NULL, 0);
 268
 269                 if (rt) {
 270                         dev = rt->dst.dev;
 271                         ip6_rt_put(rt);
 272                 }
 273         } else
 274                 dev = dev_get_by_index_rcu(net, ifindex);
 275
 276         if (!dev)
 277                 return NULL;
 278         idev = __in6_dev_get(dev);
 279         if (!idev)
 280                 return NULL;
 281         read_lock_bh(&idev->lock);
 282         if (idev->dead) {
 283                 read_unlock_bh(&idev->lock);
 284                 return NULL;
 285         }
 286         return idev;
 287 }
 288
 289 void __ipv6_sock_mc_close(struct sock *sk)
 290 {
 291         struct ipv6_pinfo *np = inet6_sk(sk);
 292         struct ipv6_mc_socklist *mc_lst;
 293         struct net *net = sock_net(sk);
 294
 295         ASSERT_RTNL();
 296
 297         while ((mc_lst = rtnl_dereference(np->ipv6_mc_list)) != NULL) {
 298                 struct net_device *dev;
 299
 300                 np->ipv6_mc_list = mc_lst->next;
 301
 302                 dev = __dev_get_by_index(net, mc_lst->ifindex);
 303                 if (dev) {
 304                         struct inet6_dev *idev = __in6_dev_get(dev);
 305
 306                         (void) ip6_mc_leave_src(sk, mc_lst, idev);
 307                         if (idev)
 308                                 __ipv6_dev_mc_dec(idev, &mc_lst->addr);
 309                 } else
 310                         (void) ip6_mc_leave_src(sk, mc_lst, NULL);
 311
 312                 atomic_sub(sizeof(*mc_lst), &sk->sk_omem_alloc);
 313                 kfree_rcu(mc_lst, rcu);
 314         }
 315 }
 316
 317 void ipv6_sock_mc_close(struct sock *sk)
 318 {
 319         struct ipv6_pinfo *np = inet6_sk(sk);
 320
 321         if (!rcu_access_pointer(np->ipv6_mc_list))
 322                 return;
 323         rtnl_lock();
 324         __ipv6_sock_mc_close(sk);
 325         rtnl_unlock();
 326 }
 327
 328 int ip6_mc_source(int add, int omode, struct sock *sk,
 329         struct group_source_req *pgsr)
 330 {
 331         struct in6_addr *source, *group;
 332         struct ipv6_mc_socklist *pmc;
 333         struct inet6_dev *idev;
 334         struct ipv6_pinfo *inet6 = inet6_sk(sk);
 335         struct ip6_sf_socklist *psl;
 336         struct net *net = sock_net(sk);
 337         int i, j, rv;
 338         int leavegroup = 0;
 339         int pmclocked = 0;
 340         int err;
 341
 342         source = &((struct sockaddr_in6 *)&pgsr->gsr_source)->sin6_addr;
 343         group = &((struct sockaddr_in6 *)&pgsr->gsr_group)->sin6_addr;
 344
 345         if (!ipv6_addr_is_multicast(group))
 346                 return -EINVAL;
 347
 348         rcu_read_lock();
 349         idev = ip6_mc_find_dev_rcu(net, group, pgsr->gsr_interface);
 350         if (!idev) {
 351                 rcu_read_unlock();
 352                 return -ENODEV;
 353         }
 354
 355         err = -EADDRNOTAVAIL;
 356
 357         for_each_pmc_rcu(inet6, pmc) {
 358                 if (pgsr->gsr_interface && pmc->ifindex != pgsr->gsr_interface)
 359                         continue;
 360                 if (ipv6_addr_equal(&pmc->addr, group))
 361                         break;
 362         }
 363         if (!pmc) {             /* must have a prior join */
 364                 err = -EINVAL;
 365                 goto done;
 366         }
 367         /* if a source filter was set, must be the same mode as before */
 368         if (pmc->sflist) {
 369                 if (pmc->sfmode != omode) {
 370                         err = -EINVAL;
 371                         goto done;
 372                 }
 373         } else if (pmc->sfmode != omode) {
 374                 /* allow mode switches for empty-set filters */
 375                 ip6_mc_add_src(idev, group, omode, 0, NULL, 0);
 376                 ip6_mc_del_src(idev, group, pmc->sfmode, 0, NULL, 0);
 377                 pmc->sfmode = omode;
 378         }
 379
 380         write_lock(&pmc->sflock);
 381         pmclocked = 1;
 382
 383         psl = pmc->sflist;
 384         if (!add) {
 385                 if (!psl)
 386                         goto done;      /* err = -EADDRNOTAVAIL */
 387                 rv = !0;
 388                 for (i = 0; i < psl->sl_count; i++) {
 389                         rv = !ipv6_addr_equal(&psl->sl_addr[i], source);
 390                         if (rv == 0)
 391                                 break;
 392                 }
 393                 if (rv)         /* source not found */
 394                         goto done;      /* err = -EADDRNOTAVAIL */
 395
 396                 /* special case - (INCLUDE, empty) == LEAVE_GROUP */
 397                 if (psl->sl_count == 1 && omode == MCAST_INCLUDE) {
 398                         leavegroup = 1;
 399                         goto done;
 400                 }
 401
 402                 /* update the interface filter */
 403                 ip6_mc_del_src(idev, group, omode, 1, source, 1);
 404
 405                 for (j = i+1; j < psl->sl_count; j++)
 406                         psl->sl_addr[j-1] = psl->sl_addr[j];
 407                 psl->sl_count--;
 408                 err = 0;
 409                 goto done;
 410         }
 411         /* else, add a new source to the filter */
 412
 413         if (psl && psl->sl_count >= sysctl_mld_max_msf) {
 414                 err = -ENOBUFS;
 415                 goto done;
 416         }
 417         if (!psl || psl->sl_count == psl->sl_max) {
 418                 struct ip6_sf_socklist *newpsl;
 419                 int count = IP6_SFBLOCK;
 420
 421                 if (psl)
 422                         count += psl->sl_max;
 423                 newpsl = sock_kmalloc(sk, IP6_SFLSIZE(count), GFP_ATOMIC);
 424                 if (!newpsl) {
 425                         err = -ENOBUFS;
 426                         goto done;
 427                 }
 428                 newpsl->sl_max = count;
 429                 newpsl->sl_count = count - IP6_SFBLOCK;
 430                 if (psl) {
 431                         for (i = 0; i < psl->sl_count; i++)
 432                                 newpsl->sl_addr[i] = psl->sl_addr[i];
 433                         sock_kfree_s(sk, psl, IP6_SFLSIZE(psl->sl_max));
 434                 }
 435                 pmc->sflist = psl = newpsl;
 436         }
 437         rv = 1; /* > 0 for insert logic below if sl_count is 0 */
 438         for (i = 0; i < psl->sl_count; i++) {
 439                 rv = !ipv6_addr_equal(&psl->sl_addr[i], source);
 440                 if (rv == 0) /* There is an error in the address. */
 441                         goto done;
 442         }
 443         for (j = psl->sl_count-1; j >= i; j--)
 444                 psl->sl_addr[j+1] = psl->sl_addr[j];
 445         psl->sl_addr[i] = *source;
 446         psl->sl_count++;
 447         err = 0;
 448         /* update the interface list */
 449         ip6_mc_add_src(idev, group, omode, 1, source, 1);
 450 done:
 451         if (pmclocked)
 452                 write_unlock(&pmc->sflock);
 453         read_unlock_bh(&idev->lock);
 454         rcu_read_unlock();
 455         if (leavegroup)
 456                 err = ipv6_sock_mc_drop(sk, pgsr->gsr_interface, group);
 457         return err;
 458 }
 459
 460 int ip6_mc_msfilter(struct sock *sk, struct group_filter *gsf,
 461                     struct sockaddr_storage *list)
 462 {
 463         const struct in6_addr *group;
 464         struct ipv6_mc_socklist *pmc;
 465         struct inet6_dev *idev;
 466         struct ipv6_pinfo *inet6 = inet6_sk(sk);
 467         struct ip6_sf_socklist *newpsl, *psl;
 468         struct net *net = sock_net(sk);
 469         int leavegroup = 0;
 470         int i, err;
 471
 472         group = &((struct sockaddr_in6 *)&gsf->gf_group)->sin6_addr;
 473
 474         if (!ipv6_addr_is_multicast(group))
 475                 return -EINVAL;
 476         if (gsf->gf_fmode != MCAST_INCLUDE &&
 477             gsf->gf_fmode != MCAST_EXCLUDE)
 478                 return -EINVAL;
 479
 480         rcu_read_lock();
 481         idev = ip6_mc_find_dev_rcu(net, group, gsf->gf_interface);
 482
 483         if (!idev) {
 484                 rcu_read_unlock();
 485                 return -ENODEV;
 486         }
 487
 488         err = 0;
 489
 490         if (gsf->gf_fmode == MCAST_INCLUDE && gsf->gf_numsrc == 0) {
 491                 leavegroup = 1;
 492                 goto done;
 493         }
 494
 495         for_each_pmc_rcu(inet6, pmc) {
 496                 if (pmc->ifindex != gsf->gf_interface)
 497                         continue;
 498                 if (ipv6_addr_equal(&pmc->addr, group))
 499                         break;
 500         }
 501         if (!pmc) {             /* must have a prior join */
 502                 err = -EINVAL;
 503                 goto done;
 504         }
 505         if (gsf->gf_numsrc) {
 506                 newpsl = sock_kmalloc(sk, IP6_SFLSIZE(gsf->gf_numsrc),
 507                                                           GFP_ATOMIC);
 508                 if (!newpsl) {
 509                         err = -ENOBUFS;
 510                         goto done;
 511                 }
 512                 newpsl->sl_max = newpsl->sl_count = gsf->gf_numsrc;
 513                 for (i = 0; i < newpsl->sl_count; ++i, ++list) {
 514                         struct sockaddr_in6 *psin6;
 515
 516                         psin6 = (struct sockaddr_in6 *)list;
 517                         newpsl->sl_addr[i] = psin6->sin6_addr;
 518                 }
 519                 err = ip6_mc_add_src(idev, group, gsf->gf_fmode,
 520                         newpsl->sl_count, newpsl->sl_addr, 0);
 521                 if (err) {
 522                         sock_kfree_s(sk, newpsl, IP6_SFLSIZE(newpsl->sl_max));
 523                         goto done;
 524                 }
 525         } else {
 526                 newpsl = NULL;
 527                 (void) ip6_mc_add_src(idev, group, gsf->gf_fmode, 0, NULL, 0);
 528         }
 529
 530         write_lock(&pmc->sflock);
 531         psl = pmc->sflist;
 532         if (psl) {
 533                 (void) ip6_mc_del_src(idev, group, pmc->sfmode,
 534                         psl->sl_count, psl->sl_addr, 0);
 535                 sock_kfree_s(sk, psl, IP6_SFLSIZE(psl->sl_max));
 536         } else
 537                 (void) ip6_mc_del_src(idev, group, pmc->sfmode, 0, NULL, 0);
 538         pmc->sflist = newpsl;
 539         pmc->sfmode = gsf->gf_fmode;
 540         write_unlock(&pmc->sflock);
 541         err = 0;
 542 done:
 543         read_unlock_bh(&idev->lock);
 544         rcu_read_unlock();
 545         if (leavegroup)
 546                 err = ipv6_sock_mc_drop(sk, gsf->gf_interface, group);
 547         return err;
 548 }
 549
 550 int ip6_mc_msfget(struct sock *sk, struct group_filter *gsf,
 551         struct sockaddr_storage *p)
 552 {
 553         int err, i, count, copycount;
 554         const struct in6_addr *group;
 555         struct ipv6_mc_socklist *pmc;
 556         struct inet6_dev *idev;
 557         struct ipv6_pinfo *inet6 = inet6_sk(sk);
 558         struct ip6_sf_socklist *psl;
 559         struct net *net = sock_net(sk);
 560
 561         group = &((struct sockaddr_in6 *)&gsf->gf_group)->sin6_addr;
 562
 563         if (!ipv6_addr_is_multicast(group))
 564                 return -EINVAL;
 565
 566         rcu_read_lock();
 567         idev = ip6_mc_find_dev_rcu(net, group, gsf->gf_interface);
 568
 569         if (!idev) {
 570                 rcu_read_unlock();
 571                 return -ENODEV;
 572         }
 573
 574         err = -EADDRNOTAVAIL;
 575         /* changes to the ipv6_mc_list require the socket lock and
 576          * rtnl lock. We have the socket lock and rcu read lock,
 577          * so reading the list is safe.
 578          */
 579
 580         for_each_pmc_rcu(inet6, pmc) {
 581                 if (pmc->ifindex != gsf->gf_interface)
 582                         continue;
 583                 if (ipv6_addr_equal(group, &pmc->addr))
 584                         break;
 585         }
 586         if (!pmc)               /* must have a prior join */
 587                 goto done;
 588         gsf->gf_fmode = pmc->sfmode;
 589         psl = pmc->sflist;
 590         count = psl ? psl->sl_count : 0;
 591         read_unlock_bh(&idev->lock);
 592         rcu_read_unlock();
 593
 594         copycount = count < gsf->gf_numsrc ? count : gsf->gf_numsrc;
 595         gsf->gf_numsrc = count;
 596         /* changes to psl require the socket lock, and a write lock
 597          * on pmc->sflock. We have the socket lock so reading here is safe.
 598          */
 599         for (i = 0; i < copycount; i++, p++) {
 600                 struct sockaddr_in6 *psin6;
 601                 struct sockaddr_storage ss;
 602
 603                 psin6 = (struct sockaddr_in6 *)&ss;
 604                 memset(&ss, 0, sizeof(ss));
 605                 psin6->sin6_family = AF_INET6;
 606                 psin6->sin6_addr = psl->sl_addr[i];
 607                 if (copy_to_user(p, &ss, sizeof(ss)))
 608                         return -EFAULT;
 609         }
 610         return 0;
 611 done:
 612         read_unlock_bh(&idev->lock);
 613         rcu_read_unlock();
 614         return err;
 615 }
 616
 617 bool inet6_mc_check(struct sock *sk, const struct in6_addr *mc_addr,
 618                     const struct in6_addr *src_addr)
 619 {
 620         struct ipv6_pinfo *np = inet6_sk(sk);
 621         struct ipv6_mc_socklist *mc;
 622         struct ip6_sf_socklist *psl;
 623         bool rv = true;
 624
 625         rcu_read_lock();
 626         for_each_pmc_rcu(np, mc) {
 627                 if (ipv6_addr_equal(&mc->addr, mc_addr))
 628                         break;
 629         }
 630         if (!mc) {
 631                 rcu_read_unlock();
 632                 return np->mc_all;
 633         }
 634         read_lock(&mc->sflock);
 635         psl = mc->sflist;
 636         if (!psl) {
 637                 rv = mc->sfmode == MCAST_EXCLUDE;
 638         } else {
 639                 int i;
 640
 641                 for (i = 0; i < psl->sl_count; i++) {
 642                         if (ipv6_addr_equal(&psl->sl_addr[i], src_addr))
 643                                 break;
 644                 }
 645                 if (mc->sfmode == MCAST_INCLUDE && i >= psl->sl_count)
 646                         rv = false;
 647                 if (mc->sfmode == MCAST_EXCLUDE && i < psl->sl_count)
 648                         rv = false;
 649         }
 650         read_unlock(&mc->sflock);
 651         rcu_read_unlock();
 652
 653         return rv;
 654 }
 655
 656 static void igmp6_group_added(struct ifmcaddr6 *mc)
 657 {
 658         struct net_device *dev = mc->idev->dev;
 659         char buf[MAX_ADDR_LEN];
 660
 661         if (IPV6_ADDR_MC_SCOPE(&mc->mca_addr) <
 662             IPV6_ADDR_SCOPE_LINKLOCAL)
 663                 return;
 664
 665         spin_lock_bh(&mc->mca_lock);
 666         if (!(mc->mca_flags&MAF_LOADED)) {
 667                 mc->mca_flags |= MAF_LOADED;
 668                 if (ndisc_mc_map(&mc->mca_addr, buf, dev, 0) == 0)
 669                         dev_mc_add(dev, buf);
 670         }
 671         spin_unlock_bh(&mc->mca_lock);
 672
 673         if (!(dev->flags & IFF_UP) || (mc->mca_flags & MAF_NOREPORT))
 674                 return;
 675
 676         if (mld_in_v1_mode(mc->idev)) {
 677                 igmp6_join_group(mc);
 678                 return;
 679         }
 680         /* else v2 */
 681
 682         /* Based on RFC3810 6.1, for newly added INCLUDE SSM, we
 683          * should not send filter-mode change record as the mode
 684          * should be from IN() to IN(A).
 685          */
 686         if (mc->mca_sfmode == MCAST_EXCLUDE)
 687                 mc->mca_crcount = mc->idev->mc_qrv;
 688
 689         mld_ifc_event(mc->idev);
 690 }
 691
 692 static void igmp6_group_dropped(struct ifmcaddr6 *mc)
 693 {
 694         struct net_device *dev = mc->idev->dev;
 695         char buf[MAX_ADDR_LEN];
 696
 697         if (IPV6_ADDR_MC_SCOPE(&mc->mca_addr) <
 698             IPV6_ADDR_SCOPE_LINKLOCAL)
 699                 return;
 700
 701         spin_lock_bh(&mc->mca_lock);
 702         if (mc->mca_flags&MAF_LOADED) {
 703                 mc->mca_flags &= ~MAF_LOADED;
 704                 if (ndisc_mc_map(&mc->mca_addr, buf, dev, 0) == 0)
 705                         dev_mc_del(dev, buf);
 706         }
 707
 708         spin_unlock_bh(&mc->mca_lock);
 709         if (mc->mca_flags & MAF_NOREPORT)
 710                 return;
 711
 712         if (!mc->idev->dead)
 713                 igmp6_leave_group(mc);
 714
 715         spin_lock_bh(&mc->mca_lock);
 716         if (del_timer(&mc->mca_timer))
 717                 refcount_dec(&mc->mca_refcnt);
 718         spin_unlock_bh(&mc->mca_lock);
 719 }
 720
 721 /*
 722  * deleted ifmcaddr6 manipulation
 723  */
 724 static void mld_add_delrec(struct inet6_dev *idev, struct ifmcaddr6 *im)
 725 {
 726         struct ifmcaddr6 *pmc;
 727
 728         /* this is an "ifmcaddr6" for convenience; only the fields below
 729          * are actually used. In particular, the refcnt and users are not
 730          * used for management of the delete list. Using the same structure
 731          * for deleted items allows change reports to use common code with
 732          * non-deleted or query-response MCA's.
 733          */
 734         pmc = kzalloc(sizeof(*pmc), GFP_ATOMIC);
 735         if (!pmc)
 736                 return;
 737
 738         spin_lock_bh(&im->mca_lock);
 739         spin_lock_init(&pmc->mca_lock);
 740         pmc->idev = im->idev;
 741         in6_dev_hold(idev);
 742         pmc->mca_addr = im->mca_addr;
 743         pmc->mca_crcount = idev->mc_qrv;
 744         pmc->mca_sfmode = im->mca_sfmode;
 745         if (pmc->mca_sfmode == MCAST_INCLUDE) {
 746                 struct ip6_sf_list *psf;
 747
 748                 pmc->mca_tomb = im->mca_tomb;
 749                 pmc->mca_sources = im->mca_sources;
 750                 im->mca_tomb = im->mca_sources = NULL;
 751                 for (psf = pmc->mca_sources; psf; psf = psf->sf_next)
 752                         psf->sf_crcount = pmc->mca_crcount;
 753         }
 754         spin_unlock_bh(&im->mca_lock);
 755
 756         spin_lock_bh(&idev->mc_lock);
 757         pmc->next = idev->mc_tomb;
 758         idev->mc_tomb = pmc;
 759         spin_unlock_bh(&idev->mc_lock);
 760 }
 761
 762 static void mld_del_delrec(struct inet6_dev *idev, struct ifmcaddr6 *im)
 763 {
 764         struct ifmcaddr6 *pmc, *pmc_prev;
 765         struct ip6_sf_list *psf;
 766         struct in6_addr *pmca = &im->mca_addr;
 767
 768         spin_lock_bh(&idev->mc_lock);
 769         pmc_prev = NULL;
 770         for (pmc = idev->mc_tomb; pmc; pmc = pmc->next) {
 771                 if (ipv6_addr_equal(&pmc->mca_addr, pmca))
 772                         break;
 773                 pmc_prev = pmc;
 774         }
 775         if (pmc) {
 776                 if (pmc_prev)
 777                         pmc_prev->next = pmc->next;
 778                 else
 779                         idev->mc_tomb = pmc->next;
 780         }
 781         spin_unlock_bh(&idev->mc_lock);
 782
 783         spin_lock_bh(&im->mca_lock);
 784         if (pmc) {
 785                 im->idev = pmc->idev;
 786                 if (im->mca_sfmode == MCAST_INCLUDE) {
 787                         swap(im->mca_tomb, pmc->mca_tomb);
 788                         swap(im->mca_sources, pmc->mca_sources);
 789                         for (psf = im->mca_sources; psf; psf = psf->sf_next)
 790                                 psf->sf_crcount = idev->mc_qrv;
 791                 } else {
 792                         im->mca_crcount = idev->mc_qrv;
 793                 }
 794                 in6_dev_put(pmc->idev);
 795                 ip6_mc_clear_src(pmc);
 796                 kfree(pmc);
 797         }
 798         spin_unlock_bh(&im->mca_lock);
 799 }
 800
 801 static void mld_clear_delrec(struct inet6_dev *idev)
 802 {
 803         struct ifmcaddr6 *pmc, *nextpmc;
 804
 805         spin_lock_bh(&idev->mc_lock);
 806         pmc = idev->mc_tomb;
 807         idev->mc_tomb = NULL;
 808         spin_unlock_bh(&idev->mc_lock);
 809
 810         for (; pmc; pmc = nextpmc) {
 811                 nextpmc = pmc->next;
 812                 ip6_mc_clear_src(pmc);
 813                 in6_dev_put(pmc->idev);
 814                 kfree(pmc);
 815         }
 816
 817         /* clear dead sources, too */
 818         read_lock_bh(&idev->lock);
 819         for (pmc = idev->mc_list; pmc; pmc = pmc->next) {
 820                 struct ip6_sf_list *psf, *psf_next;
 821
 822                 spin_lock_bh(&pmc->mca_lock);
 823                 psf = pmc->mca_tomb;
 824                 pmc->mca_tomb = NULL;
 825                 spin_unlock_bh(&pmc->mca_lock);
 826                 for (; psf; psf = psf_next) {
 827                         psf_next = psf->sf_next;
 828                         kfree(psf);
 829                 }
 830         }
 831         read_unlock_bh(&idev->lock);
 832 }
 833
 834 static void mca_get(struct ifmcaddr6 *mc)
 835 {
 836         refcount_inc(&mc->mca_refcnt);
 837 }
 838
 839 static void ma_put(struct ifmcaddr6 *mc)
 840 {
 841         if (refcount_dec_and_test(&mc->mca_refcnt)) {
 842                 in6_dev_put(mc->idev);
 843                 kfree(mc);
 844         }
 845 }
 846
 847 static struct ifmcaddr6 *mca_alloc(struct inet6_dev *idev,
 848                                    const struct in6_addr *addr,
 849                                    unsigned int mode)
 850 {
 851         struct ifmcaddr6 *mc;
 852
 853         mc = kzalloc(sizeof(*mc), GFP_ATOMIC);
 854         if (!mc)
 855                 return NULL;
 856
 857         timer_setup(&mc->mca_timer, igmp6_timer_handler, 0);
 858
 859         mc->mca_addr = *addr;
 860         mc->idev = idev; /* reference taken by caller */
 861         mc->mca_users = 1;
 862         /* mca_stamp should be updated upon changes */
 863         mc->mca_cstamp = mc->mca_tstamp = jiffies;
 864         refcount_set(&mc->mca_refcnt, 1);
 865         spin_lock_init(&mc->mca_lock);
 866
 867         mc->mca_sfmode = mode;
 868         mc->mca_sfcount[mode] = 1;
 869
 870         if (ipv6_addr_is_ll_all_nodes(&mc->mca_addr) ||
 871             IPV6_ADDR_MC_SCOPE(&mc->mca_addr) < IPV6_ADDR_SCOPE_LINKLOCAL)
 872                 mc->mca_flags |= MAF_NOREPORT;
 873
 874         return mc;
 875 }
 876
 877 /*
 878  *      device multicast group inc (add if not found)
 879  */
 880 static int __ipv6_dev_mc_inc(struct net_device *dev,
 881                              const struct in6_addr *addr, unsigned int mode)
 882 {
 883         struct ifmcaddr6 *mc;
 884         struct inet6_dev *idev;
 885
 886         ASSERT_RTNL();
 887
 888         /* we need to take a reference on idev */
 889         idev = in6_dev_get(dev);
 890
 891         if (!idev)
 892                 return -EINVAL;
 893
 894         write_lock_bh(&idev->lock);
 895         if (idev->dead) {
 896                 write_unlock_bh(&idev->lock);
 897                 in6_dev_put(idev);
 898                 return -ENODEV;
 899         }
 900
 901         for (mc = idev->mc_list; mc; mc = mc->next) {
 902                 if (ipv6_addr_equal(&mc->mca_addr, addr)) {
 903                         mc->mca_users++;
 904                         write_unlock_bh(&idev->lock);
 905                         ip6_mc_add_src(idev, &mc->mca_addr, mode, 0, NULL, 0);
 906                         in6_dev_put(idev);
 907                         return 0;
 908                 }
 909         }
 910
 911         mc = mca_alloc(idev, addr, mode);
 912         if (!mc) {
 913                 write_unlock_bh(&idev->lock);
 914                 in6_dev_put(idev);
 915                 return -ENOMEM;
 916         }
 917
 918         mc->next = idev->mc_list;
 919         idev->mc_list = mc;
 920
 921         /* Hold this for the code below before we unlock,
 922          * it is already exposed via idev->mc_list.
 923          */
 924         mca_get(mc);
 925         write_unlock_bh(&idev->lock);
 926
 927         mld_del_delrec(idev, mc);
 928         igmp6_group_added(mc);
 929         ma_put(mc);
 930         return 0;
 931 }
 932
 933 int ipv6_dev_mc_inc(struct net_device *dev, const struct in6_addr *addr)
 934 {
 935         return __ipv6_dev_mc_inc(dev, addr, MCAST_EXCLUDE);
 936 }
 937 EXPORT_SYMBOL(ipv6_dev_mc_inc);
 938
 939 /*
 940  *      device multicast group del
 941  */
 942 int __ipv6_dev_mc_dec(struct inet6_dev *idev, const struct in6_addr *addr)
 943 {
 944         struct ifmcaddr6 *ma, **map;
 945
 946         ASSERT_RTNL();
 947
 948         write_lock_bh(&idev->lock);
 949         for (map = &idev->mc_list; (ma = *map) != NULL; map = &ma->next) {
 950                 if (ipv6_addr_equal(&ma->mca_addr, addr)) {
 951                         if (--ma->mca_users == 0) {
 952                                 *map = ma->next;
 953                                 write_unlock_bh(&idev->lock);
 954
 955                                 igmp6_group_dropped(ma);
 956                                 ip6_mc_clear_src(ma);
 957
 958                                 ma_put(ma);
 959                                 return 0;
 960                         }
 961                         write_unlock_bh(&idev->lock);
 962                         return 0;
 963                 }
 964         }
 965         write_unlock_bh(&idev->lock);
 966
 967         return -ENOENT;
 968 }
 969
 970 int ipv6_dev_mc_dec(struct net_device *dev, const struct in6_addr *addr)
 971 {
 972         struct inet6_dev *idev;
 973         int err;
 974
 975         ASSERT_RTNL();
 976
 977         idev = __in6_dev_get(dev);
 978         if (!idev)
 979                 err = -ENODEV;
 980         else
 981                 err = __ipv6_dev_mc_dec(idev, addr);
 982
 983         return err;
 984 }
 985 EXPORT_SYMBOL(ipv6_dev_mc_dec);
 986
 987 /*
 988  *      check if the interface/address pair is valid
 989  */
 990 bool ipv6_chk_mcast_addr(struct net_device *dev, const struct in6_addr *group,
 991                          const struct in6_addr *src_addr)
 992 {
 993         struct inet6_dev *idev;
 994         struct ifmcaddr6 *mc;
 995         bool rv = false;
 996
 997         rcu_read_lock();
 998         idev = __in6_dev_get(dev);
 999         if (idev) {
1000                 read_lock_bh(&idev->lock);
1001                 for (mc = idev->mc_list; mc; mc = mc->next) {
1002                         if (ipv6_addr_equal(&mc->mca_addr, group))
1003                                 break;
1004                 }
1005                 if (mc) {
1006                         if (src_addr && !ipv6_addr_any(src_addr)) {
1007                                 struct ip6_sf_list *psf;
1008
1009                                 spin_lock_bh(&mc->mca_lock);
1010                                 for (psf = mc->mca_sources; psf; psf = psf->sf_next) {
1011                                         if (ipv6_addr_equal(&psf->sf_addr, src_addr))
1012                                                 break;
1013                                 }
1014                                 if (psf)
1015                                         rv = psf->sf_count[MCAST_INCLUDE] ||
1016                                                 psf->sf_count[MCAST_EXCLUDE] !=
1017                                                 mc->mca_sfcount[MCAST_EXCLUDE];
1018                                 else
1019                                         rv = mc->mca_sfcount[MCAST_EXCLUDE] != 0;
1020                                 spin_unlock_bh(&mc->mca_lock);
1021                         } else
1022                                 rv = true; /* don't filter unspecified source */
1023                 }
1024                 read_unlock_bh(&idev->lock);
1025         }
1026         rcu_read_unlock();
1027         return rv;
1028 }
1029
1030 static void mld_gq_start_timer(struct inet6_dev *idev)
1031 {
1032         unsigned long tv = prandom_u32() % idev->mc_maxdelay;
1033
1034         idev->mc_gq_running = 1;
1035         if (!mod_timer(&idev->mc_gq_timer, jiffies+tv+2))
1036                 in6_dev_hold(idev);
1037 }
1038
1039 static void mld_gq_stop_timer(struct inet6_dev *idev)
1040 {
1041         idev->mc_gq_running = 0;
1042         if (del_timer(&idev->mc_gq_timer))
1043                 __in6_dev_put(idev);
1044 }
1045
1046 static void mld_ifc_start_timer(struct inet6_dev *idev, unsigned long delay)
1047 {
1048         unsigned long tv = prandom_u32() % delay;
1049
1050         if (!mod_timer(&idev->mc_ifc_timer, jiffies+tv+2))
1051                 in6_dev_hold(idev);
1052 }
1053
1054 static void mld_ifc_stop_timer(struct inet6_dev *idev)
1055 {
1056         idev->mc_ifc_count = 0;
1057         if (del_timer(&idev->mc_ifc_timer))
1058                 __in6_dev_put(idev);
1059 }
1060
1061 static void mld_dad_start_timer(struct inet6_dev *idev, unsigned long delay)
1062 {
1063         unsigned long tv = prandom_u32() % delay;
1064
1065         if (!mod_timer(&idev->mc_dad_timer, jiffies+tv+2))
1066                 in6_dev_hold(idev);
1067 }
1068
1069 static void mld_dad_stop_timer(struct inet6_dev *idev)
1070 {
1071         if (del_timer(&idev->mc_dad_timer))
1072                 __in6_dev_put(idev);
1073 }
1074
1075 /*
1076  *      IGMP handling (alias multicast ICMPv6 messages)
1077  */
1078
1079 static void igmp6_group_queried(struct ifmcaddr6 *ma, unsigned long resptime)
1080 {
1081         unsigned long delay = resptime;
1082
1083         /* Do not start timer for these addresses */
1084         if (ipv6_addr_is_ll_all_nodes(&ma->mca_addr) ||
1085             IPV6_ADDR_MC_SCOPE(&ma->mca_addr) < IPV6_ADDR_SCOPE_LINKLOCAL)
1086                 return;
1087
1088         if (del_timer(&ma->mca_timer)) {
1089                 refcount_dec(&ma->mca_refcnt);
1090                 delay = ma->mca_timer.expires - jiffies;
1091         }
1092
1093         if (delay >= resptime)
1094                 delay = prandom_u32() % resptime;
1095
1096         ma->mca_timer.expires = jiffies + delay;
1097         if (!mod_timer(&ma->mca_timer, jiffies + delay))
1098                 refcount_inc(&ma->mca_refcnt);
1099         ma->mca_flags |= MAF_TIMER_RUNNING;
1100 }
1101
1102 /* mark EXCLUDE-mode sources */
1103 static bool mld_xmarksources(struct ifmcaddr6 *pmc, int nsrcs,
1104                              const struct in6_addr *srcs)
1105 {
1106         struct ip6_sf_list *psf;
1107         int i, scount;
1108
1109         scount = 0;
1110         for (psf = pmc->mca_sources; psf; psf = psf->sf_next) {
1111                 if (scount == nsrcs)
1112                         break;
1113                 for (i = 0; i < nsrcs; i++) {
1114                         /* skip inactive filters */
1115                         if (psf->sf_count[MCAST_INCLUDE] ||
1116                             pmc->mca_sfcount[MCAST_EXCLUDE] !=
1117                             psf->sf_count[MCAST_EXCLUDE])
1118                                 break;
1119                         if (ipv6_addr_equal(&srcs[i], &psf->sf_addr)) {
1120                                 scount++;
1121                                 break;
1122                         }
1123                 }
1124         }
1125         pmc->mca_flags &= ~MAF_GSQUERY;
1126         if (scount == nsrcs)    /* all sources excluded */
1127                 return false;
1128         return true;
1129 }
1130
1131 static bool mld_marksources(struct ifmcaddr6 *pmc, int nsrcs,
1132                             const struct in6_addr *srcs)
1133 {
1134         struct ip6_sf_list *psf;
1135         int i, scount;
1136
1137         if (pmc->mca_sfmode == MCAST_EXCLUDE)
1138                 return mld_xmarksources(pmc, nsrcs, srcs);
1139
1140         /* mark INCLUDE-mode sources */
1141
1142         scount = 0;
1143         for (psf = pmc->mca_sources; psf; psf = psf->sf_next) {
1144                 if (scount == nsrcs)
1145                         break;
1146                 for (i = 0; i < nsrcs; i++) {
1147                         if (ipv6_addr_equal(&srcs[i], &psf->sf_addr)) {
1148                                 psf->sf_gsresp = 1;
1149                                 scount++;
1150                                 break;
1151                         }
1152                 }
1153         }
1154         if (!scount) {
1155                 pmc->mca_flags &= ~MAF_GSQUERY;
1156                 return false;
1157         }
1158         pmc->mca_flags |= MAF_GSQUERY;
1159         return true;
1160 }
1161
1162 static int mld_force_mld_version(const struct inet6_dev *idev)
1163 {
1164         /* Normally, both are 0 here. If enforcement to a particular is
1165          * being used, individual device enforcement will have a lower
1166          * precedence over 'all' device (.../conf/all/force_mld_version).
1167          */
1168
1169         if (dev_net(idev->dev)->ipv6.devconf_all->force_mld_version != 0)
1170                 return dev_net(idev->dev)->ipv6.devconf_all->force_mld_version;
1171         else
1172                 return idev->cnf.force_mld_version;
1173 }
1174
1175 static bool mld_in_v2_mode_only(const struct inet6_dev *idev)
1176 {
1177         return mld_force_mld_version(idev) == 2;
1178 }
1179
1180 static bool mld_in_v1_mode_only(const struct inet6_dev *idev)
1181 {
1182         return mld_force_mld_version(idev) == 1;
1183 }
1184
1185 static bool mld_in_v1_mode(const struct inet6_dev *idev)
1186 {
1187         if (mld_in_v2_mode_only(idev))
1188                 return false;
1189         if (mld_in_v1_mode_only(idev))
1190                 return true;
1191         if (idev->mc_v1_seen && time_before(jiffies, idev->mc_v1_seen))
1192                 return true;
1193
1194         return false;
1195 }
1196
1197 static void mld_set_v1_mode(struct inet6_dev *idev)
1198 {
1199         /* RFC3810, relevant sections:
1200          *  - 9.1. Robustness Variable
1201          *  - 9.2. Query Interval
1202          *  - 9.3. Query Response Interval
1203          *  - 9.12. Older Version Querier Present Timeout
1204          */
1205         unsigned long switchback;
1206
1207         switchback = (idev->mc_qrv * idev->mc_qi) + idev->mc_qri;
1208
1209         idev->mc_v1_seen = jiffies + switchback;
1210 }
1211
1212 static void mld_update_qrv(struct inet6_dev *idev,
1213                            const struct mld2_query *mlh2)
1214 {
1215         /* RFC3810, relevant sections:
1216          *  - 5.1.8. QRV (Querier's Robustness Variable)
1217          *  - 9.1. Robustness Variable
1218          */
1219
1220         /* The value of the Robustness Variable MUST NOT be zero,
1221          * and SHOULD NOT be one. Catch this here if we ever run
1222          * into such a case in future.
1223          */
1224         const int min_qrv = min(MLD_QRV_DEFAULT, sysctl_mld_qrv);
1225         WARN_ON(idev->mc_qrv == 0);
1226
1227         if (mlh2->mld2q_qrv > 0)
1228                 idev->mc_qrv = mlh2->mld2q_qrv;
1229
1230         if (unlikely(idev->mc_qrv < min_qrv)) {
1231                 net_warn_ratelimited("IPv6: MLD: clamping QRV from %u to %u!\n",
1232                                      idev->mc_qrv, min_qrv);
1233                 idev->mc_qrv = min_qrv;
1234         }
1235 }
1236
1237 static void mld_update_qi(struct inet6_dev *idev,
1238                           const struct mld2_query *mlh2)
1239 {
1240         /* RFC3810, relevant sections:
1241          *  - 5.1.9. QQIC (Querier's Query Interval Code)
1242          *  - 9.2. Query Interval
1243          *  - 9.12. Older Version Querier Present Timeout
1244          *    (the [Query Interval] in the last Query received)
1245          */
1246         unsigned long mc_qqi;
1247
1248         if (mlh2->mld2q_qqic < 128) {
1249                 mc_qqi = mlh2->mld2q_qqic;
1250         } else {
1251                 unsigned long mc_man, mc_exp;
1252
1253                 mc_exp = MLDV2_QQIC_EXP(mlh2->mld2q_qqic);
1254                 mc_man = MLDV2_QQIC_MAN(mlh2->mld2q_qqic);
1255
1256                 mc_qqi = (mc_man | 0x10) << (mc_exp + 3);
1257         }
1258
1259         idev->mc_qi = mc_qqi * HZ;
1260 }
1261
1262 static void mld_update_qri(struct inet6_dev *idev,
1263                            const struct mld2_query *mlh2)
1264 {
1265         /* RFC3810, relevant sections:
1266          *  - 5.1.3. Maximum Response Code
1267          *  - 9.3. Query Response Interval
1268          */
1269         idev->mc_qri = msecs_to_jiffies(mldv2_mrc(mlh2));
1270 }
1271
1272 static int mld_process_v1(struct inet6_dev *idev, struct mld_msg *mld,
1273                           unsigned long *max_delay, bool v1_query)
1274 {
1275         unsigned long mldv1_md;
1276
1277         /* Ignore v1 queries */
1278         if (mld_in_v2_mode_only(idev))
1279                 return -EINVAL;
1280
1281         mldv1_md = ntohs(mld->mld_maxdelay);
1282
1283         /* When in MLDv1 fallback and a MLDv2 router start-up being
1284          * unaware of current MLDv1 operation, the MRC == MRD mapping
1285          * only works when the exponential algorithm is not being
1286          * used (as MLDv1 is unaware of such things).
1287          *
1288          * According to the RFC author, the MLDv2 implementations
1289          * he's aware of all use a MRC < 32768 on start up queries.
1290          *
1291          * Thus, should we *ever* encounter something else larger
1292          * than that, just assume the maximum possible within our
1293          * reach.
1294          */
1295         if (!v1_query)
1296                 mldv1_md = min(mldv1_md, MLDV1_MRD_MAX_COMPAT);
1297
1298         *max_delay = max(msecs_to_jiffies(mldv1_md), 1UL);
1299
1300         /* MLDv1 router present: we need to go into v1 mode *only*
1301          * when an MLDv1 query is received as per section 9.12. of
1302          * RFC3810! And we know from RFC2710 section 3.7 that MLDv1
1303          * queries MUST be of exactly 24 octets.
1304          */
1305         if (v1_query)
1306                 mld_set_v1_mode(idev);
1307
1308         /* cancel MLDv2 report timer */
1309         mld_gq_stop_timer(idev);
1310         /* cancel the interface change timer */
1311         mld_ifc_stop_timer(idev);
1312         /* clear deleted report items */
1313         mld_clear_delrec(idev);
1314
1315         return 0;
1316 }
1317
1318 static int mld_process_v2(struct inet6_dev *idev, struct mld2_query *mld,
1319                           unsigned long *max_delay)
1320 {
1321         *max_delay = max(msecs_to_jiffies(mldv2_mrc(mld)), 1UL);
1322
1323         mld_update_qrv(idev, mld);
1324         mld_update_qi(idev, mld);
1325         mld_update_qri(idev, mld);
1326
1327         idev->mc_maxdelay = *max_delay;
1328
1329         return 0;
1330 }
1331
1332 /* called with rcu_read_lock() */
1333 int igmp6_event_query(struct sk_buff *skb)
1334 {
1335         struct mld2_query *mlh2 = NULL;
1336         struct ifmcaddr6 *ma;
1337         const struct in6_addr *group;
1338         unsigned long max_delay;
1339         struct inet6_dev *idev;
1340         struct mld_msg *mld;
1341         int group_type;
1342         int mark = 0;
1343         int len, err;
1344
1345         if (!pskb_may_pull(skb, sizeof(struct in6_addr)))
1346                 return -EINVAL;
1347
1348         /* compute payload length excluding extension headers */
1349         len = ntohs(ipv6_hdr(skb)->payload_len) + sizeof(struct ipv6hdr);
1350         len -= skb_network_header_len(skb);
1351
1352         /* RFC3810 6.2
1353          * Upon reception of an MLD message that contains a Query, the node
1354          * checks if the source address of the message is a valid link-local
1355          * address, if the Hop Limit is set to 1, and if the Router Alert
1356          * option is present in the Hop-By-Hop Options header of the IPv6
1357          * packet.  If any of these checks fails, the packet is dropped.
1358          */
1359         if (!(ipv6_addr_type(&ipv6_hdr(skb)->saddr) & IPV6_ADDR_LINKLOCAL) ||
1360             ipv6_hdr(skb)->hop_limit != 1 ||
1361             !(IP6CB(skb)->flags & IP6SKB_ROUTERALERT) ||
1362             IP6CB(skb)->ra != htons(IPV6_OPT_ROUTERALERT_MLD))
1363                 return -EINVAL;
1364
1365         idev = __in6_dev_get(skb->dev);
1366         if (!idev)
1367                 return 0;
1368
1369         mld = (struct mld_msg *)icmp6_hdr(skb);
1370         group = &mld->mld_mca;
1371         group_type = ipv6_addr_type(group);
1372
1373         if (group_type != IPV6_ADDR_ANY &&
1374             !(group_type&IPV6_ADDR_MULTICAST))
1375                 return -EINVAL;
1376
1377         if (len < MLD_V1_QUERY_LEN) {
1378                 return -EINVAL;
1379         } else if (len == MLD_V1_QUERY_LEN || mld_in_v1_mode(idev)) {
1380                 err = mld_process_v1(idev, mld, &max_delay,
1381                                      len == MLD_V1_QUERY_LEN);
1382                 if (err < 0)
1383                         return err;
1384         } else if (len >= MLD_V2_QUERY_LEN_MIN) {
1385                 int srcs_offset = sizeof(struct mld2_query) -
1386                                   sizeof(struct icmp6hdr);
1387
1388                 if (!pskb_may_pull(skb, srcs_offset))
1389                         return -EINVAL;
1390
1391                 mlh2 = (struct mld2_query *)skb_transport_header(skb);
1392
1393                 err = mld_process_v2(idev, mlh2, &max_delay);
1394                 if (err < 0)
1395                         return err;
1396
1397                 if (group_type == IPV6_ADDR_ANY) { /* general query */
1398                         if (mlh2->mld2q_nsrcs)
1399                                 return -EINVAL; /* no sources allowed */
1400
1401                         mld_gq_start_timer(idev);
1402                         return 0;
1403                 }
1404                 /* mark sources to include, if group & source-specific */
1405                 if (mlh2->mld2q_nsrcs != 0) {
1406                         if (!pskb_may_pull(skb, srcs_offset +
1407                             ntohs(mlh2->mld2q_nsrcs) * sizeof(struct in6_addr)))
1408                                 return -EINVAL;
1409
1410                         mlh2 = (struct mld2_query *)skb_transport_header(skb);
1411                         mark = 1;
1412                 }
1413         } else {
1414                 return -EINVAL;
1415         }
1416
1417         read_lock_bh(&idev->lock);
1418         if (group_type == IPV6_ADDR_ANY) {
1419                 for (ma = idev->mc_list; ma; ma = ma->next) {
1420                         spin_lock_bh(&ma->mca_lock);
1421                         igmp6_group_queried(ma, max_delay);
1422                         spin_unlock_bh(&ma->mca_lock);
1423                 }
1424         } else {
1425                 for (ma = idev->mc_list; ma; ma = ma->next) {
1426                         if (!ipv6_addr_equal(group, &ma->mca_addr))
1427                                 continue;
1428                         spin_lock_bh(&ma->mca_lock);
1429                         if (ma->mca_flags & MAF_TIMER_RUNNING) {
1430                                 /* gsquery <- gsquery && mark */
1431                                 if (!mark)
1432                                         ma->mca_flags &= ~MAF_GSQUERY;
1433                         } else {
1434                                 /* gsquery <- mark */
1435                                 if (mark)
1436                                         ma->mca_flags |= MAF_GSQUERY;
1437                                 else
1438                                         ma->mca_flags &= ~MAF_GSQUERY;
1439                         }
1440                         if (!(ma->mca_flags & MAF_GSQUERY) ||
1441                             mld_marksources(ma, ntohs(mlh2->mld2q_nsrcs), mlh2->mld2q_srcs))
1442                                 igmp6_group_queried(ma, max_delay);
1443                         spin_unlock_bh(&ma->mca_lock);
1444                         break;
1445                 }
1446         }
1447         read_unlock_bh(&idev->lock);
1448
1449         return 0;
1450 }
1451
1452 /* called with rcu_read_lock() */
1453 int igmp6_event_report(struct sk_buff *skb)
1454 {
1455         struct ifmcaddr6 *ma;
1456         struct inet6_dev *idev;
1457         struct mld_msg *mld;
1458         int addr_type;
1459
1460         /* Our own report looped back. Ignore it. */
1461         if (skb->pkt_type == PACKET_LOOPBACK)
1462                 return 0;
1463
1464         /* send our report if the MC router may not have heard this report */
1465         if (skb->pkt_type != PACKET_MULTICAST &&
1466             skb->pkt_type != PACKET_BROADCAST)
1467                 return 0;
1468
1469         if (!pskb_may_pull(skb, sizeof(*mld) - sizeof(struct icmp6hdr)))
1470                 return -EINVAL;
1471
1472         mld = (struct mld_msg *)icmp6_hdr(skb);
1473
1474         /* Drop reports with not link local source */
1475         addr_type = ipv6_addr_type(&ipv6_hdr(skb)->saddr);
1476         if (addr_type != IPV6_ADDR_ANY &&
1477             !(addr_type&IPV6_ADDR_LINKLOCAL))
1478                 return -EINVAL;
1479
1480         idev = __in6_dev_get(skb->dev);
1481         if (!idev)
1482                 return -ENODEV;
1483
1484         /*
1485          *      Cancel the timer for this group
1486          */
1487
1488         read_lock_bh(&idev->lock);
1489         for (ma = idev->mc_list; ma; ma = ma->next) {
1490                 if (ipv6_addr_equal(&ma->mca_addr, &mld->mld_mca)) {
1491                         spin_lock(&ma->mca_lock);
1492                         if (del_timer(&ma->mca_timer))
1493                                 refcount_dec(&ma->mca_refcnt);
1494                         ma->mca_flags &= ~(MAF_LAST_REPORTER|MAF_TIMER_RUNNING);
1495                         spin_unlock(&ma->mca_lock);
1496                         break;
1497                 }
1498         }
1499         read_unlock_bh(&idev->lock);
1500         return 0;
1501 }
1502
1503 static bool is_in(struct ifmcaddr6 *pmc, struct ip6_sf_list *psf, int type,
1504                   int gdeleted, int sdeleted)
1505 {
1506         switch (type) {
1507         case MLD2_MODE_IS_INCLUDE:
1508         case MLD2_MODE_IS_EXCLUDE:
1509                 if (gdeleted || sdeleted)
1510                         return false;
1511                 if (!((pmc->mca_flags & MAF_GSQUERY) && !psf->sf_gsresp)) {
1512                         if (pmc->mca_sfmode == MCAST_INCLUDE)
1513                                 return true;
1514                         /* don't include if this source is excluded
1515                          * in all filters
1516                          */
1517                         if (psf->sf_count[MCAST_INCLUDE])
1518                                 return type == MLD2_MODE_IS_INCLUDE;
1519                         return pmc->mca_sfcount[MCAST_EXCLUDE] ==
1520                                 psf->sf_count[MCAST_EXCLUDE];
1521                 }
1522                 return false;
1523         case MLD2_CHANGE_TO_INCLUDE:
1524                 if (gdeleted || sdeleted)
1525                         return false;
1526                 return psf->sf_count[MCAST_INCLUDE] != 0;
1527         case MLD2_CHANGE_TO_EXCLUDE:
1528                 if (gdeleted || sdeleted)
1529                         return false;
1530                 if (pmc->mca_sfcount[MCAST_EXCLUDE] == 0 ||
1531                     psf->sf_count[MCAST_INCLUDE])
1532                         return false;
1533                 return pmc->mca_sfcount[MCAST_EXCLUDE] ==
1534                         psf->sf_count[MCAST_EXCLUDE];
1535         case MLD2_ALLOW_NEW_SOURCES:
1536                 if (gdeleted || !psf->sf_crcount)
1537                         return false;
1538                 return (pmc->mca_sfmode == MCAST_INCLUDE) ^ sdeleted;
1539         case MLD2_BLOCK_OLD_SOURCES:
1540                 if (pmc->mca_sfmode == MCAST_INCLUDE)
1541                         return gdeleted || (psf->sf_crcount && sdeleted);
1542                 return psf->sf_crcount && !gdeleted && !sdeleted;
1543         }
1544         return false;
1545 }
1546
1547 static int
1548 mld_scount(struct ifmcaddr6 *pmc, int type, int gdeleted, int sdeleted)
1549 {
1550         struct ip6_sf_list *psf;
1551         int scount = 0;
1552
1553         for (psf = pmc->mca_sources; psf; psf = psf->sf_next) {
1554                 if (!is_in(pmc, psf, type, gdeleted, sdeleted))
1555                         continue;
1556                 scount++;
1557         }
1558         return scount;
1559 }
1560
1561 static void ip6_mc_hdr(struct sock *sk, struct sk_buff *skb,
1562                        struct net_device *dev,
1563                        const struct in6_addr *saddr,
1564                        const struct in6_addr *daddr,
1565                        int proto, int len)
1566 {
1567         struct ipv6hdr *hdr;
1568
1569         skb->protocol = htons(ETH_P_IPV6);
1570         skb->dev = dev;
1571
1572         skb_reset_network_header(skb);
1573         skb_put(skb, sizeof(struct ipv6hdr));
1574         hdr = ipv6_hdr(skb);
1575
1576         ip6_flow_hdr(hdr, 0, 0);
1577
1578         hdr->payload_len = htons(len);
1579         hdr->nexthdr = proto;
1580         hdr->hop_limit = inet6_sk(sk)->hop_limit;
1581
1582         hdr->saddr = *saddr;
1583         hdr->daddr = *daddr;
1584 }
1585
1586 static struct sk_buff *mld_newpack(struct inet6_dev *idev, unsigned int mtu)
1587 {
1588         struct net_device *dev = idev->dev;
1589         struct net *net = dev_net(dev);
1590         struct sock *sk = net->ipv6.igmp_sk;
1591         struct sk_buff *skb;
1592         struct mld2_report *pmr;
1593         struct in6_addr addr_buf;
1594         const struct in6_addr *saddr;
1595         int hlen = LL_RESERVED_SPACE(dev);
1596         int tlen = dev->needed_tailroom;
1597         unsigned int size = mtu + hlen + tlen;
1598         int err;
1599         u8 ra[8] = { IPPROTO_ICMPV6, 0,
1600                      IPV6_TLV_ROUTERALERT, 2, 0, 0,
1601                      IPV6_TLV_PADN, 0 };
1602
1603         /* we assume size > sizeof(ra) here */
1604         /* limit our allocations to order-0 page */
1605         size = min_t(int, size, SKB_MAX_ORDER(0, 0));
1606         skb = sock_alloc_send_skb(sk, size, 1, &err);
1607
1608         if (!skb)
1609                 return NULL;
1610
1611         skb->priority = TC_PRIO_CONTROL;
1612         skb_reserve(skb, hlen);
1613         skb_tailroom_reserve(skb, mtu, tlen);
1614
1615         if (__ipv6_get_lladdr(idev, &addr_buf, IFA_F_TENTATIVE)) {
1616                 /* <draft-ietf-magma-mld-source-05.txt>:
1617                  * use unspecified address as the source address
1618                  * when a valid link-local address is not available.
1619                  */
1620                 saddr = &in6addr_any;
1621         } else
1622                 saddr = &addr_buf;
1623
1624         ip6_mc_hdr(sk, skb, dev, saddr, &mld2_all_mcr, NEXTHDR_HOP, 0);
1625
1626         skb_put_data(skb, ra, sizeof(ra));
1627
1628         skb_set_transport_header(skb, skb_tail_pointer(skb) - skb->data);
1629         skb_put(skb, sizeof(*pmr));
1630         pmr = (struct mld2_report *)skb_transport_header(skb);
1631         pmr->mld2r_type = ICMPV6_MLD2_REPORT;
1632         pmr->mld2r_resv1 = 0;
1633         pmr->mld2r_cksum = 0;
1634         pmr->mld2r_resv2 = 0;
1635         pmr->mld2r_ngrec = 0;
1636         return skb;
1637 }
1638
1639 static void mld_sendpack(struct sk_buff *skb)
1640 {
1641         struct ipv6hdr *pip6 = ipv6_hdr(skb);
1642         struct mld2_report *pmr =
1643                               (struct mld2_report *)skb_transport_header(skb);
1644         int payload_len, mldlen;
1645         struct inet6_dev *idev;
1646         struct net *net = dev_net(skb->dev);
1647         int err;
1648         struct flowi6 fl6;
1649         struct dst_entry *dst;
1650
1651         rcu_read_lock();
1652         idev = __in6_dev_get(skb->dev);
1653         IP6_UPD_PO_STATS(net, idev, IPSTATS_MIB_OUT, skb->len);
1654
1655         payload_len = (skb_tail_pointer(skb) - skb_network_header(skb)) -
1656                 sizeof(*pip6);
1657         mldlen = skb_tail_pointer(skb) - skb_transport_header(skb);
1658         pip6->payload_len = htons(payload_len);
1659
1660         pmr->mld2r_cksum = csum_ipv6_magic(&pip6->saddr, &pip6->daddr, mldlen,
1661                                            IPPROTO_ICMPV6,
1662                                            csum_partial(skb_transport_header(skb),
1663                                                         mldlen, 0));
1664
1665         icmpv6_flow_init(net->ipv6.igmp_sk, &fl6, ICMPV6_MLD2_REPORT,
1666                          &ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr,
1667                          skb->dev->ifindex);
1668         dst = icmp6_dst_alloc(skb->dev, &fl6);
1669
1670         err = 0;
1671         if (IS_ERR(dst)) {
1672                 err = PTR_ERR(dst);
1673                 dst = NULL;
1674         }
1675         skb_dst_set(skb, dst);
1676         if (err)
1677                 goto err_out;
1678
1679         err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT,
1680                       net, net->ipv6.igmp_sk, skb, NULL, skb->dev,
1681                       dst_output);
1682 out:
1683         if (!err) {
1684                 ICMP6MSGOUT_INC_STATS(net, idev, ICMPV6_MLD2_REPORT);
1685                 ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTMSGS);
1686         } else {
1687                 IP6_INC_STATS(net, idev, IPSTATS_MIB_OUTDISCARDS);
1688         }
1689
1690         rcu_read_unlock();
1691         return;
1692
1693 err_out:
1694         kfree_skb(skb);
1695         goto out;
1696 }
1697
1698 static int grec_size(struct ifmcaddr6 *pmc, int type, int gdel, int sdel)
1699 {
1700         return sizeof(struct mld2_grec) + 16 * mld_scount(pmc,type,gdel,sdel);
1701 }
1702
1703 static struct sk_buff *add_grhead(struct sk_buff *skb, struct ifmcaddr6 *pmc,
1704         int type, struct mld2_grec **ppgr, unsigned int mtu)
1705 {
1706         struct mld2_report *pmr;
1707         struct mld2_grec *pgr;
1708
1709         if (!skb) {
1710                 skb = mld_newpack(pmc->idev, mtu);
1711                 if (!skb)
1712                         return NULL;
1713         }
1714         pgr = skb_put(skb, sizeof(struct mld2_grec));
1715         pgr->grec_type = type;
1716         pgr->grec_auxwords = 0;
1717         pgr->grec_nsrcs = 0;
1718         pgr->grec_mca = pmc->mca_addr;  /* structure copy */
1719         pmr = (struct mld2_report *)skb_transport_header(skb);
1720         pmr->mld2r_ngrec = htons(ntohs(pmr->mld2r_ngrec)+1);
1721         *ppgr = pgr;
1722         return skb;
1723 }
1724
1725 #define AVAILABLE(skb)  ((skb) ? skb_availroom(skb) : 0)
1726
1727 static struct sk_buff *add_grec(struct sk_buff *skb, struct ifmcaddr6 *pmc,
1728         int type, int gdeleted, int sdeleted, int crsend)
1729 {
1730         struct inet6_dev *idev = pmc->idev;
1731         struct net_device *dev = idev->dev;
1732         struct mld2_report *pmr;
1733         struct mld2_grec *pgr = NULL;
1734         struct ip6_sf_list *psf, *psf_next, *psf_prev, **psf_list;
1735         int scount, stotal, first, isquery, truncate;
1736         unsigned int mtu;
1737
1738         if (pmc->mca_flags & MAF_NOREPORT)
1739                 return skb;
1740
1741         mtu = READ_ONCE(dev->mtu);
1742         if (mtu < IPV6_MIN_MTU)
1743                 return skb;
1744
1745         isquery = type == MLD2_MODE_IS_INCLUDE ||
1746                   type == MLD2_MODE_IS_EXCLUDE;
1747         truncate = type == MLD2_MODE_IS_EXCLUDE ||
1748                     type == MLD2_CHANGE_TO_EXCLUDE;
1749
1750         stotal = scount = 0;
1751
1752         psf_list = sdeleted ? &pmc->mca_tomb : &pmc->mca_sources;
1753
1754         if (!*psf_list)
1755                 goto empty_source;
1756
1757         pmr = skb ? (struct mld2_report *)skb_transport_header(skb) : NULL;
1758
1759         /* EX and TO_EX get a fresh packet, if needed */
1760         if (truncate) {
1761                 if (pmr && pmr->mld2r_ngrec &&
1762                     AVAILABLE(skb) < grec_size(pmc, type, gdeleted, sdeleted)) {
1763                         if (skb)
1764                                 mld_sendpack(skb);
1765                         skb = mld_newpack(idev, mtu);
1766                 }
1767         }
1768         first = 1;
1769         psf_prev = NULL;
1770         for (psf = *psf_list; psf; psf = psf_next) {
1771                 struct in6_addr *psrc;
1772
1773                 psf_next = psf->sf_next;
1774
1775                 if (!is_in(pmc, psf, type, gdeleted, sdeleted) && !crsend) {
1776                         psf_prev = psf;
1777                         continue;
1778                 }
1779
1780                 /* Based on RFC3810 6.1. Should not send source-list change
1781                  * records when there is a filter mode change.
1782                  */
1783                 if (((gdeleted && pmc->mca_sfmode == MCAST_EXCLUDE) ||
1784                      (!gdeleted && pmc->mca_crcount)) &&
1785                     (type == MLD2_ALLOW_NEW_SOURCES ||
1786                      type == MLD2_BLOCK_OLD_SOURCES) && psf->sf_crcount)
1787                         goto decrease_sf_crcount;
1788
1789                 /* clear marks on query responses */
1790                 if (isquery)
1791                         psf->sf_gsresp = 0;
1792
1793                 if (AVAILABLE(skb) < sizeof(*psrc) +
1794                     first*sizeof(struct mld2_grec)) {
1795                         if (truncate && !first)
1796                                 break;   /* truncate these */
1797                         if (pgr)
1798                                 pgr->grec_nsrcs = htons(scount);
1799                         if (skb)
1800                                 mld_sendpack(skb);
1801                         skb = mld_newpack(idev, mtu);
1802                         first = 1;
1803                         scount = 0;
1804                 }
1805                 if (first) {
1806                         skb = add_grhead(skb, pmc, type, &pgr, mtu);
1807                         first = 0;
1808                 }
1809                 if (!skb)
1810                         return NULL;
1811                 psrc = skb_put(skb, sizeof(*psrc));
1812                 *psrc = psf->sf_addr;
1813                 scount++; stotal++;
1814                 if ((type == MLD2_ALLOW_NEW_SOURCES ||
1815                      type == MLD2_BLOCK_OLD_SOURCES) && psf->sf_crcount) {
1816 decrease_sf_crcount:
1817                         psf->sf_crcount--;
1818                         if ((sdeleted || gdeleted) && psf->sf_crcount == 0) {
1819                                 if (psf_prev)
1820                                         psf_prev->sf_next = psf->sf_next;
1821                                 else
1822                                         *psf_list = psf->sf_next;
1823                                 kfree(psf);
1824                                 continue;
1825                         }
1826                 }
1827                 psf_prev = psf;
1828         }
1829
1830 empty_source:
1831         if (!stotal) {
1832                 if (type == MLD2_ALLOW_NEW_SOURCES ||
1833                     type == MLD2_BLOCK_OLD_SOURCES)
1834                         return skb;
1835                 if (pmc->mca_crcount || isquery || crsend) {
1836                         /* make sure we have room for group header */
1837                         if (skb && AVAILABLE(skb) < sizeof(struct mld2_grec)) {
1838                                 mld_sendpack(skb);
1839                                 skb = NULL; /* add_grhead will get a new one */
1840                         }
1841                         skb = add_grhead(skb, pmc, type, &pgr, mtu);
1842                 }
1843         }
1844         if (pgr)
1845                 pgr->grec_nsrcs = htons(scount);
1846
1847         if (isquery)
1848                 pmc->mca_flags &= ~MAF_GSQUERY; /* clear query state */
1849         return skb;
1850 }
1851
1852 static void mld_send_report(struct inet6_dev *idev, struct ifmcaddr6 *pmc)
1853 {
1854         struct sk_buff *skb = NULL;
1855         int type;
1856
1857         read_lock_bh(&idev->lock);
1858         if (!pmc) {
1859                 for (pmc = idev->mc_list; pmc; pmc = pmc->next) {
1860                         if (pmc->mca_flags & MAF_NOREPORT)
1861                                 continue;
1862                         spin_lock_bh(&pmc->mca_lock);
1863                         if (pmc->mca_sfcount[MCAST_EXCLUDE])
1864                                 type = MLD2_MODE_IS_EXCLUDE;
1865                         else
1866                                 type = MLD2_MODE_IS_INCLUDE;
1867                         skb = add_grec(skb, pmc, type, 0, 0, 0);
1868                         spin_unlock_bh(&pmc->mca_lock);
1869                 }
1870         } else {
1871                 spin_lock_bh(&pmc->mca_lock);
1872                 if (pmc->mca_sfcount[MCAST_EXCLUDE])
1873                         type = MLD2_MODE_IS_EXCLUDE;
1874                 else
1875                         type = MLD2_MODE_IS_INCLUDE;
1876                 skb = add_grec(skb, pmc, type, 0, 0, 0);
1877                 spin_unlock_bh(&pmc->mca_lock);
1878         }
1879         read_unlock_bh(&idev->lock);
1880         if (skb)
1881                 mld_sendpack(skb);
1882 }
1883
1884 /*
1885  * remove zero-count source records from a source filter list
1886  */
1887 static void mld_clear_zeros(struct ip6_sf_list **ppsf)
1888 {
1889         struct ip6_sf_list *psf_prev, *psf_next, *psf;
1890
1891         psf_prev = NULL;
1892         for (psf = *ppsf; psf; psf = psf_next) {
1893                 psf_next = psf->sf_next;
1894                 if (psf->sf_crcount == 0) {
1895                         if (psf_prev)
1896                                 psf_prev->sf_next = psf->sf_next;
1897                         else
1898                                 *ppsf = psf->sf_next;
1899                         kfree(psf);
1900                 } else
1901                         psf_prev = psf;
1902         }
1903 }
1904
1905 static void mld_send_cr(struct inet6_dev *idev)
1906 {
1907         struct ifmcaddr6 *pmc, *pmc_prev, *pmc_next;
1908         struct sk_buff *skb = NULL;
1909         int type, dtype;
1910
1911         read_lock_bh(&idev->lock);
1912         spin_lock(&idev->mc_lock);
1913
1914         /* deleted MCA's */
1915         pmc_prev = NULL;
1916         for (pmc = idev->mc_tomb; pmc; pmc = pmc_next) {
1917                 pmc_next = pmc->next;
1918                 if (pmc->mca_sfmode == MCAST_INCLUDE) {
1919                         type = MLD2_BLOCK_OLD_SOURCES;
1920                         dtype = MLD2_BLOCK_OLD_SOURCES;
1921                         skb = add_grec(skb, pmc, type, 1, 0, 0);
1922                         skb = add_grec(skb, pmc, dtype, 1, 1, 0);
1923                 }
1924                 if (pmc->mca_crcount) {
1925                         if (pmc->mca_sfmode == MCAST_EXCLUDE) {
1926                                 type = MLD2_CHANGE_TO_INCLUDE;
1927                                 skb = add_grec(skb, pmc, type, 1, 0, 0);
1928                         }
1929                         pmc->mca_crcount--;
1930                         if (pmc->mca_crcount == 0) {
1931                                 mld_clear_zeros(&pmc->mca_tomb);
1932                                 mld_clear_zeros(&pmc->mca_sources);
1933                         }
1934                 }
1935                 if (pmc->mca_crcount == 0 && !pmc->mca_tomb &&
1936                     !pmc->mca_sources) {
1937                         if (pmc_prev)
1938                                 pmc_prev->next = pmc_next;
1939                         else
1940                                 idev->mc_tomb = pmc_next;
1941                         in6_dev_put(pmc->idev);
1942                         kfree(pmc);
1943                 } else
1944                         pmc_prev = pmc;
1945         }
1946         spin_unlock(&idev->mc_lock);
1947
1948         /* change recs */
1949         for (pmc = idev->mc_list; pmc; pmc = pmc->next) {
1950                 spin_lock_bh(&pmc->mca_lock);
1951                 if (pmc->mca_sfcount[MCAST_EXCLUDE]) {
1952                         type = MLD2_BLOCK_OLD_SOURCES;
1953                         dtype = MLD2_ALLOW_NEW_SOURCES;
1954                 } else {
1955                         type = MLD2_ALLOW_NEW_SOURCES;
1956                         dtype = MLD2_BLOCK_OLD_SOURCES;
1957                 }
1958                 skb = add_grec(skb, pmc, type, 0, 0, 0);
1959                 skb = add_grec(skb, pmc, dtype, 0, 1, 0);       /* deleted sources */
1960
1961                 /* filter mode changes */
1962                 if (pmc->mca_crcount) {
1963                         if (pmc->mca_sfmode == MCAST_EXCLUDE)
1964                                 type = MLD2_CHANGE_TO_EXCLUDE;
1965                         else
1966                                 type = MLD2_CHANGE_TO_INCLUDE;
1967                         skb = add_grec(skb, pmc, type, 0, 0, 0);
1968                         pmc->mca_crcount--;
1969                 }
1970                 spin_unlock_bh(&pmc->mca_lock);
1971         }
1972         read_unlock_bh(&idev->lock);
1973         if (!skb)
1974                 return;
1975         (void) mld_sendpack(skb);
1976 }
1977
1978 static void igmp6_send(struct in6_addr *addr, struct net_device *dev, int type)
1979 {
1980         struct net *net = dev_net(dev);
1981         struct sock *sk = net->ipv6.igmp_sk;
1982         struct inet6_dev *idev;
1983         struct sk_buff *skb;
1984         struct mld_msg *hdr;
1985         const struct in6_addr *snd_addr, *saddr;
1986         struct in6_addr addr_buf;
1987         int hlen = LL_RESERVED_SPACE(dev);
1988         int tlen = dev->needed_tailroom;
1989         int err, len, payload_len, full_len;
1990         u8 ra[8] = { IPPROTO_ICMPV6, 0,
1991                      IPV6_TLV_ROUTERALERT, 2, 0, 0,
1992                      IPV6_TLV_PADN, 0 };
1993         struct flowi6 fl6;
1994         struct dst_entry *dst;
1995
1996         if (type == ICMPV6_MGM_REDUCTION)
1997                 snd_addr = &in6addr_linklocal_allrouters;
1998         else
1999                 snd_addr = addr;
2000
2001         len = sizeof(struct icmp6hdr) + sizeof(struct in6_addr);
2002         payload_len = len + sizeof(ra);
2003         full_len = sizeof(struct ipv6hdr) + payload_len;
2004
2005         rcu_read_lock();
2006         IP6_UPD_PO_STATS(net, __in6_dev_get(dev),
2007                       IPSTATS_MIB_OUT, full_len);
2008         rcu_read_unlock();
2009
2010         skb = sock_alloc_send_skb(sk, hlen + tlen + full_len, 1, &err);
2011
2012         if (!skb) {
2013                 rcu_read_lock();
2014                 IP6_INC_STATS(net, __in6_dev_get(dev),
2015                               IPSTATS_MIB_OUTDISCARDS);
2016                 rcu_read_unlock();
2017                 return;
2018         }
2019         skb->priority = TC_PRIO_CONTROL;
2020         skb_reserve(skb, hlen);
2021
2022         if (ipv6_get_lladdr(dev, &addr_buf, IFA_F_TENTATIVE)) {
2023                 /* <draft-ietf-magma-mld-source-05.txt>:
2024                  * use unspecified address as the source address
2025                  * when a valid link-local address is not available.
2026                  */
2027                 saddr = &in6addr_any;
2028         } else
2029                 saddr = &addr_buf;
2030
2031         ip6_mc_hdr(sk, skb, dev, saddr, snd_addr, NEXTHDR_HOP, payload_len);
2032
2033         skb_put_data(skb, ra, sizeof(ra));
2034
2035         hdr = skb_put_zero(skb, sizeof(struct mld_msg));
2036         hdr->mld_type = type;
2037         hdr->mld_mca = *addr;
2038
2039         hdr->mld_cksum = csum_ipv6_magic(saddr, snd_addr, len,
2040                                          IPPROTO_ICMPV6,
2041                                          csum_partial(hdr, len, 0));
2042
2043         rcu_read_lock();
2044         idev = __in6_dev_get(skb->dev);
2045
2046         icmpv6_flow_init(sk, &fl6, type,
2047                          &ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr,
2048                          skb->dev->ifindex);
2049         dst = icmp6_dst_alloc(skb->dev, &fl6);
2050         if (IS_ERR(dst)) {
2051                 err = PTR_ERR(dst);
2052                 goto err_out;
2053         }
2054
2055         skb_dst_set(skb, dst);
2056         err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT,
2057                       net, sk, skb, NULL, skb->dev,
2058                       dst_output);
2059 out:
2060         if (!err) {
2061                 ICMP6MSGOUT_INC_STATS(net, idev, type);
2062                 ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTMSGS);
2063         } else
2064                 IP6_INC_STATS(net, idev, IPSTATS_MIB_OUTDISCARDS);
2065
2066         rcu_read_unlock();
2067         return;
2068
2069 err_out:
2070         kfree_skb(skb);
2071         goto out;
2072 }
2073
2074 static void mld_send_initial_cr(struct inet6_dev *idev)
2075 {
2076         struct sk_buff *skb;
2077         struct ifmcaddr6 *pmc;
2078         int type;
2079
2080         if (mld_in_v1_mode(idev))
2081                 return;
2082
2083         skb = NULL;
2084         read_lock_bh(&idev->lock);
2085         for (pmc = idev->mc_list; pmc; pmc = pmc->next) {
2086                 spin_lock_bh(&pmc->mca_lock);
2087                 if (pmc->mca_sfcount[MCAST_EXCLUDE])
2088                         type = MLD2_CHANGE_TO_EXCLUDE;
2089                 else
2090                         type = MLD2_ALLOW_NEW_SOURCES;
2091                 skb = add_grec(skb, pmc, type, 0, 0, 1);
2092                 spin_unlock_bh(&pmc->mca_lock);
2093         }
2094         read_unlock_bh(&idev->lock);
2095         if (skb)
2096                 mld_sendpack(skb);
2097 }
2098
2099 void ipv6_mc_dad_complete(struct inet6_dev *idev)
2100 {
2101         idev->mc_dad_count = idev->mc_qrv;
2102         if (idev->mc_dad_count) {
2103                 mld_send_initial_cr(idev);
2104                 idev->mc_dad_count--;
2105                 if (idev->mc_dad_count)
2106                         mld_dad_start_timer(idev,
2107                                             unsolicited_report_interval(idev));
2108         }
2109 }
2110
2111 static void mld_dad_timer_expire(struct timer_list *t)
2112 {
2113         struct inet6_dev *idev = from_timer(idev, t, mc_dad_timer);
2114
2115         mld_send_initial_cr(idev);
2116         if (idev->mc_dad_count) {
2117                 idev->mc_dad_count--;
2118                 if (idev->mc_dad_count)
2119                         mld_dad_start_timer(idev,
2120                                             unsolicited_report_interval(idev));
2121         }
2122         in6_dev_put(idev);
2123 }
2124
2125 static int ip6_mc_del1_src(struct ifmcaddr6 *pmc, int sfmode,
2126         const struct in6_addr *psfsrc)
2127 {
2128         struct ip6_sf_list *psf, *psf_prev;
2129         int rv = 0;
2130
2131         psf_prev = NULL;
2132         for (psf = pmc->mca_sources; psf; psf = psf->sf_next) {
2133                 if (ipv6_addr_equal(&psf->sf_addr, psfsrc))
2134                         break;
2135                 psf_prev = psf;
2136         }
2137         if (!psf || psf->sf_count[sfmode] == 0) {
2138                 /* source filter not found, or count wrong =>  bug */
2139                 return -ESRCH;
2140         }
2141         psf->sf_count[sfmode]--;
2142         if (!psf->sf_count[MCAST_INCLUDE] && !psf->sf_count[MCAST_EXCLUDE]) {
2143                 struct inet6_dev *idev = pmc->idev;
2144
2145                 /* no more filters for this source */
2146                 if (psf_prev)
2147                         psf_prev->sf_next = psf->sf_next;
2148                 else
2149                         pmc->mca_sources = psf->sf_next;
2150                 if (psf->sf_oldin && !(pmc->mca_flags & MAF_NOREPORT) &&
2151                     !mld_in_v1_mode(idev)) {
2152                         psf->sf_crcount = idev->mc_qrv;
2153                         psf->sf_next = pmc->mca_tomb;
2154                         pmc->mca_tomb = psf;
2155                         rv = 1;
2156                 } else
2157                         kfree(psf);
2158         }
2159         return rv;
2160 }
2161
2162 static int ip6_mc_del_src(struct inet6_dev *idev, const struct in6_addr *pmca,
2163                           int sfmode, int sfcount, const struct in6_addr *psfsrc,
2164                           int delta)
2165 {
2166         struct ifmcaddr6 *pmc;
2167         int     changerec = 0;
2168         int     i, err;
2169
2170         if (!idev)
2171                 return -ENODEV;
2172         read_lock_bh(&idev->lock);
2173         for (pmc = idev->mc_list; pmc; pmc = pmc->next) {
2174                 if (ipv6_addr_equal(pmca, &pmc->mca_addr))
2175                         break;
2176         }
2177         if (!pmc) {
2178                 /* MCA not found?? bug */
2179                 read_unlock_bh(&idev->lock);
2180                 return -ESRCH;
2181         }
2182         spin_lock_bh(&pmc->mca_lock);
2183         sf_markstate(pmc);
2184         if (!delta) {
2185                 if (!pmc->mca_sfcount[sfmode]) {
2186                         spin_unlock_bh(&pmc->mca_lock);
2187                         read_unlock_bh(&idev->lock);
2188                         return -EINVAL;
2189                 }
2190                 pmc->mca_sfcount[sfmode]--;
2191         }
2192         err = 0;
2193         for (i = 0; i < sfcount; i++) {
2194                 int rv = ip6_mc_del1_src(pmc, sfmode, &psfsrc[i]);
2195
2196                 changerec |= rv > 0;
2197                 if (!err && rv < 0)
2198                         err = rv;
2199         }
2200         if (pmc->mca_sfmode == MCAST_EXCLUDE &&
2201             pmc->mca_sfcount[MCAST_EXCLUDE] == 0 &&
2202             pmc->mca_sfcount[MCAST_INCLUDE]) {
2203                 struct ip6_sf_list *psf;
2204
2205                 /* filter mode change */
2206                 pmc->mca_sfmode = MCAST_INCLUDE;
2207                 pmc->mca_crcount = idev->mc_qrv;
2208                 idev->mc_ifc_count = pmc->mca_crcount;
2209                 for (psf = pmc->mca_sources; psf; psf = psf->sf_next)
2210                         psf->sf_crcount = 0;
2211                 mld_ifc_event(pmc->idev);
2212         } else if (sf_setstate(pmc) || changerec)
2213                 mld_ifc_event(pmc->idev);
2214         spin_unlock_bh(&pmc->mca_lock);
2215         read_unlock_bh(&idev->lock);
2216         return err;
2217 }
2218
2219 /*
2220  * Add multicast single-source filter to the interface list
2221  */
2222 static int ip6_mc_add1_src(struct ifmcaddr6 *pmc, int sfmode,
2223         const struct in6_addr *psfsrc)
2224 {
2225         struct ip6_sf_list *psf, *psf_prev;
2226
2227         psf_prev = NULL;
2228         for (psf = pmc->mca_sources; psf; psf = psf->sf_next) {
2229                 if (ipv6_addr_equal(&psf->sf_addr, psfsrc))
2230                         break;
2231                 psf_prev = psf;
2232         }
2233         if (!psf) {
2234                 psf = kzalloc(sizeof(*psf), GFP_ATOMIC);
2235                 if (!psf)
2236                         return -ENOBUFS;
2237
2238                 psf->sf_addr = *psfsrc;
2239                 if (psf_prev) {
2240                         psf_prev->sf_next = psf;
2241                 } else
2242                         pmc->mca_sources = psf;
2243         }
2244         psf->sf_count[sfmode]++;
2245         return 0;
2246 }
2247
2248 static void sf_markstate(struct ifmcaddr6 *pmc)
2249 {
2250         struct ip6_sf_list *psf;
2251         int mca_xcount = pmc->mca_sfcount[MCAST_EXCLUDE];
2252
2253         for (psf = pmc->mca_sources; psf; psf = psf->sf_next)
2254                 if (pmc->mca_sfcount[MCAST_EXCLUDE]) {
2255                         psf->sf_oldin = mca_xcount ==
2256                                 psf->sf_count[MCAST_EXCLUDE] &&
2257                                 !psf->sf_count[MCAST_INCLUDE];
2258                 } else
2259                         psf->sf_oldin = psf->sf_count[MCAST_INCLUDE] != 0;
2260 }
2261
2262 static int sf_setstate(struct ifmcaddr6 *pmc)
2263 {
2264         struct ip6_sf_list *psf, *dpsf;
2265         int mca_xcount = pmc->mca_sfcount[MCAST_EXCLUDE];
2266         int qrv = pmc->idev->mc_qrv;
2267         int new_in, rv;
2268
2269         rv = 0;
2270         for (psf = pmc->mca_sources; psf; psf = psf->sf_next) {
2271                 if (pmc->mca_sfcount[MCAST_EXCLUDE]) {
2272                         new_in = mca_xcount == psf->sf_count[MCAST_EXCLUDE] &&
2273                                 !psf->sf_count[MCAST_INCLUDE];
2274                 } else
2275                         new_in = psf->sf_count[MCAST_INCLUDE] != 0;
2276                 if (new_in) {
2277                         if (!psf->sf_oldin) {
2278                                 struct ip6_sf_list *prev = NULL;
2279
2280                                 for (dpsf = pmc->mca_tomb; dpsf;
2281                                      dpsf = dpsf->sf_next) {
2282                                         if (ipv6_addr_equal(&dpsf->sf_addr,
2283                                             &psf->sf_addr))
2284                                                 break;
2285                                         prev = dpsf;
2286                                 }
2287                                 if (dpsf) {
2288                                         if (prev)
2289                                                 prev->sf_next = dpsf->sf_next;
2290                                         else
2291                                                 pmc->mca_tomb = dpsf->sf_next;
2292                                         kfree(dpsf);
2293                                 }
2294                                 psf->sf_crcount = qrv;
2295                                 rv++;
2296                         }
2297                 } else if (psf->sf_oldin) {
2298                         psf->sf_crcount = 0;
2299                         /*
2300                          * add or update "delete" records if an active filter
2301                          * is now inactive
2302                          */
2303                         for (dpsf = pmc->mca_tomb; dpsf; dpsf = dpsf->sf_next)
2304                                 if (ipv6_addr_equal(&dpsf->sf_addr,
2305                                     &psf->sf_addr))
2306                                         break;
2307                         if (!dpsf) {
2308                                 dpsf = kmalloc(sizeof(*dpsf), GFP_ATOMIC);
2309                                 if (!dpsf)
2310                                         continue;
2311                                 *dpsf = *psf;
2312                                 /* pmc->mca_lock held by callers */
2313                                 dpsf->sf_next = pmc->mca_tomb;
2314                                 pmc->mca_tomb = dpsf;
2315                         }
2316                         dpsf->sf_crcount = qrv;
2317                         rv++;
2318                 }
2319         }
2320         return rv;
2321 }
2322
2323 /*
2324  * Add multicast source filter list to the interface list
2325  */
2326 static int ip6_mc_add_src(struct inet6_dev *idev, const struct in6_addr *pmca,
2327                           int sfmode, int sfcount, const struct in6_addr *psfsrc,
2328                           int delta)
2329 {
2330         struct ifmcaddr6 *pmc;
2331         int     isexclude;
2332         int     i, err;
2333
2334         if (!idev)
2335                 return -ENODEV;
2336         read_lock_bh(&idev->lock);
2337         for (pmc = idev->mc_list; pmc; pmc = pmc->next) {
2338                 if (ipv6_addr_equal(pmca, &pmc->mca_addr))
2339                         break;
2340         }
2341         if (!pmc) {
2342                 /* MCA not found?? bug */
2343                 read_unlock_bh(&idev->lock);
2344                 return -ESRCH;
2345         }
2346         spin_lock_bh(&pmc->mca_lock);
2347
2348         sf_markstate(pmc);
2349         isexclude = pmc->mca_sfmode == MCAST_EXCLUDE;
2350         if (!delta)
2351                 pmc->mca_sfcount[sfmode]++;
2352         err = 0;
2353         for (i = 0; i < sfcount; i++) {
2354                 err = ip6_mc_add1_src(pmc, sfmode, &psfsrc[i]);
2355                 if (err)
2356                         break;
2357         }
2358         if (err) {
2359                 int j;
2360
2361                 if (!delta)
2362                         pmc->mca_sfcount[sfmode]--;
2363                 for (j = 0; j < i; j++)
2364                         ip6_mc_del1_src(pmc, sfmode, &psfsrc[j]);
2365         } else if (isexclude != (pmc->mca_sfcount[MCAST_EXCLUDE] != 0)) {
2366                 struct ip6_sf_list *psf;
2367
2368                 /* filter mode change */
2369                 if (pmc->mca_sfcount[MCAST_EXCLUDE])
2370                         pmc->mca_sfmode = MCAST_EXCLUDE;
2371                 else if (pmc->mca_sfcount[MCAST_INCLUDE])
2372                         pmc->mca_sfmode = MCAST_INCLUDE;
2373                 /* else no filters; keep old mode for reports */
2374
2375                 pmc->mca_crcount = idev->mc_qrv;
2376                 idev->mc_ifc_count = pmc->mca_crcount;
2377                 for (psf = pmc->mca_sources; psf; psf = psf->sf_next)
2378                         psf->sf_crcount = 0;
2379                 mld_ifc_event(idev);
2380         } else if (sf_setstate(pmc))
2381                 mld_ifc_event(idev);
2382         spin_unlock_bh(&pmc->mca_lock);
2383         read_unlock_bh(&idev->lock);
2384         return err;
2385 }
2386
2387 static void ip6_mc_clear_src(struct ifmcaddr6 *pmc)
2388 {
2389         struct ip6_sf_list *psf, *nextpsf;
2390
2391         for (psf = pmc->mca_tomb; psf; psf = nextpsf) {
2392                 nextpsf = psf->sf_next;
2393                 kfree(psf);
2394         }
2395         pmc->mca_tomb = NULL;
2396         for (psf = pmc->mca_sources; psf; psf = nextpsf) {
2397                 nextpsf = psf->sf_next;
2398                 kfree(psf);
2399         }
2400         pmc->mca_sources = NULL;
2401         pmc->mca_sfmode = MCAST_EXCLUDE;
2402         pmc->mca_sfcount[MCAST_INCLUDE] = 0;
2403         pmc->mca_sfcount[MCAST_EXCLUDE] = 1;
2404 }
2405
2406
2407 static void igmp6_join_group(struct ifmcaddr6 *ma)
2408 {
2409         unsigned long delay;
2410
2411         if (ma->mca_flags & MAF_NOREPORT)
2412                 return;
2413
2414         igmp6_send(&ma->mca_addr, ma->idev->dev, ICMPV6_MGM_REPORT);
2415
2416         delay = prandom_u32() % unsolicited_report_interval(ma->idev);
2417
2418         spin_lock_bh(&ma->mca_lock);
2419         if (del_timer(&ma->mca_timer)) {
2420                 refcount_dec(&ma->mca_refcnt);
2421                 delay = ma->mca_timer.expires - jiffies;
2422         }
2423
2424         if (!mod_timer(&ma->mca_timer, jiffies + delay))
2425                 refcount_inc(&ma->mca_refcnt);
2426         ma->mca_flags |= MAF_TIMER_RUNNING | MAF_LAST_REPORTER;
2427         spin_unlock_bh(&ma->mca_lock);
2428 }
2429
2430 static int ip6_mc_leave_src(struct sock *sk, struct ipv6_mc_socklist *iml,
2431                             struct inet6_dev *idev)
2432 {
2433         int err;
2434
2435         write_lock_bh(&iml->sflock);
2436         if (!iml->sflist) {
2437                 /* any-source empty exclude case */
2438                 err = ip6_mc_del_src(idev, &iml->addr, iml->sfmode, 0, NULL, 0);
2439         } else {
2440                 err = ip6_mc_del_src(idev, &iml->addr, iml->sfmode,
2441                                 iml->sflist->sl_count, iml->sflist->sl_addr, 0);
2442                 sock_kfree_s(sk, iml->sflist, IP6_SFLSIZE(iml->sflist->sl_max));
2443                 iml->sflist = NULL;
2444         }
2445         write_unlock_bh(&iml->sflock);
2446         return err;
2447 }
2448
2449 static void igmp6_leave_group(struct ifmcaddr6 *ma)
2450 {
2451         if (mld_in_v1_mode(ma->idev)) {
2452                 if (ma->mca_flags & MAF_LAST_REPORTER)
2453                         igmp6_send(&ma->mca_addr, ma->idev->dev,
2454                                 ICMPV6_MGM_REDUCTION);
2455         } else {
2456                 mld_add_delrec(ma->idev, ma);
2457                 mld_ifc_event(ma->idev);
2458         }
2459 }
2460
2461 static void mld_gq_timer_expire(struct timer_list *t)
2462 {
2463         struct inet6_dev *idev = from_timer(idev, t, mc_gq_timer);
2464
2465         idev->mc_gq_running = 0;
2466         mld_send_report(idev, NULL);
2467         in6_dev_put(idev);
2468 }
2469
2470 static void mld_ifc_timer_expire(struct timer_list *t)
2471 {
2472         struct inet6_dev *idev = from_timer(idev, t, mc_ifc_timer);
2473
2474         mld_send_cr(idev);
2475         if (idev->mc_ifc_count) {
2476                 idev->mc_ifc_count--;
2477                 if (idev->mc_ifc_count)
2478                         mld_ifc_start_timer(idev,
2479                                             unsolicited_report_interval(idev));
2480         }
2481         in6_dev_put(idev);
2482 }
2483
2484 static void mld_ifc_event(struct inet6_dev *idev)
2485 {
2486         if (mld_in_v1_mode(idev))
2487                 return;
2488         idev->mc_ifc_count = idev->mc_qrv;
2489         mld_ifc_start_timer(idev, 1);
2490 }
2491
2492 static void igmp6_timer_handler(struct timer_list *t)
2493 {
2494         struct ifmcaddr6 *ma = from_timer(ma, t, mca_timer);
2495
2496         if (mld_in_v1_mode(ma->idev))
2497                 igmp6_send(&ma->mca_addr, ma->idev->dev, ICMPV6_MGM_REPORT);
2498         else
2499                 mld_send_report(ma->idev, ma);
2500
2501         spin_lock(&ma->mca_lock);
2502         ma->mca_flags |=  MAF_LAST_REPORTER;
2503         ma->mca_flags &= ~MAF_TIMER_RUNNING;
2504         spin_unlock(&ma->mca_lock);
2505         ma_put(ma);
2506 }
2507
2508 /* Device changing type */
2509
2510 void ipv6_mc_unmap(struct inet6_dev *idev)
2511 {
2512         struct ifmcaddr6 *i;
2513
2514         /* Install multicast list, except for all-nodes (already installed) */
2515
2516         read_lock_bh(&idev->lock);
2517         for (i = idev->mc_list; i; i = i->next)
2518                 igmp6_group_dropped(i);
2519         read_unlock_bh(&idev->lock);
2520 }
2521
2522 void ipv6_mc_remap(struct inet6_dev *idev)
2523 {
2524         ipv6_mc_up(idev);
2525 }
2526
2527 /* Device going down */
2528
2529 void ipv6_mc_down(struct inet6_dev *idev)
2530 {
2531         struct ifmcaddr6 *i;
2532
2533         /* Withdraw multicast list */
2534
2535         read_lock_bh(&idev->lock);
2536
2537         for (i = idev->mc_list; i; i = i->next)
2538                 igmp6_group_dropped(i);
2539
2540         /* Should stop timer after group drop. or we will
2541          * start timer again in mld_ifc_event()
2542          */
2543         mld_ifc_stop_timer(idev);
2544         mld_gq_stop_timer(idev);
2545         mld_dad_stop_timer(idev);
2546         read_unlock_bh(&idev->lock);
2547 }
2548
2549 static void ipv6_mc_reset(struct inet6_dev *idev)
2550 {
2551         idev->mc_qrv = sysctl_mld_qrv;
2552         idev->mc_qi = MLD_QI_DEFAULT;
2553         idev->mc_qri = MLD_QRI_DEFAULT;
2554         idev->mc_v1_seen = 0;
2555         idev->mc_maxdelay = unsolicited_report_interval(idev);
2556 }
2557
2558 /* Device going up */
2559
2560 void ipv6_mc_up(struct inet6_dev *idev)
2561 {
2562         struct ifmcaddr6 *i;
2563
2564         /* Install multicast list, except for all-nodes (already installed) */
2565
2566         read_lock_bh(&idev->lock);
2567         ipv6_mc_reset(idev);
2568         for (i = idev->mc_list; i; i = i->next) {
2569                 mld_del_delrec(idev, i);
2570                 igmp6_group_added(i);
2571         }
2572         read_unlock_bh(&idev->lock);
2573 }
2574
2575 /* IPv6 device initialization. */
2576
2577 void ipv6_mc_init_dev(struct inet6_dev *idev)
2578 {
2579         write_lock_bh(&idev->lock);
2580         spin_lock_init(&idev->mc_lock);
2581         idev->mc_gq_running = 0;
2582         timer_setup(&idev->mc_gq_timer, mld_gq_timer_expire, 0);
2583         idev->mc_tomb = NULL;
2584         idev->mc_ifc_count = 0;
2585         timer_setup(&idev->mc_ifc_timer, mld_ifc_timer_expire, 0);
2586         timer_setup(&idev->mc_dad_timer, mld_dad_timer_expire, 0);
2587         ipv6_mc_reset(idev);
2588         write_unlock_bh(&idev->lock);
2589 }
2590
2591 /*
2592  *      Device is about to be destroyed: clean up.
2593  */
2594
2595 void ipv6_mc_destroy_dev(struct inet6_dev *idev)
2596 {
2597         struct ifmcaddr6 *i;
2598
2599         /* Deactivate timers */
2600         ipv6_mc_down(idev);
2601         mld_clear_delrec(idev);
2602
2603         /* Delete all-nodes address. */
2604         /* We cannot call ipv6_dev_mc_dec() directly, our caller in
2605          * addrconf.c has NULL'd out dev->ip6_ptr so in6_dev_get() will
2606          * fail.
2607          */
2608         __ipv6_dev_mc_dec(idev, &in6addr_linklocal_allnodes);
2609
2610         if (idev->cnf.forwarding)
2611                 __ipv6_dev_mc_dec(idev, &in6addr_linklocal_allrouters);
2612
2613         write_lock_bh(&idev->lock);
2614         while ((i = idev->mc_list) != NULL) {
2615                 idev->mc_list = i->next;
2616
2617                 write_unlock_bh(&idev->lock);
2618                 ip6_mc_clear_src(i);
2619                 ma_put(i);
2620                 write_lock_bh(&idev->lock);
2621         }
2622         write_unlock_bh(&idev->lock);
2623 }
2624
2625 static void ipv6_mc_rejoin_groups(struct inet6_dev *idev)
2626 {
2627         struct ifmcaddr6 *pmc;
2628
2629         ASSERT_RTNL();
2630
2631         if (mld_in_v1_mode(idev)) {
2632                 read_lock_bh(&idev->lock);
2633                 for (pmc = idev->mc_list; pmc; pmc = pmc->next)
2634                         igmp6_join_group(pmc);
2635                 read_unlock_bh(&idev->lock);
2636         } else
2637                 mld_send_report(idev, NULL);
2638 }
2639
2640 static int ipv6_mc_netdev_event(struct notifier_block *this,
2641                                 unsigned long event,
2642                                 void *ptr)
2643 {
2644         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2645         struct inet6_dev *idev = __in6_dev_get(dev);
2646
2647         switch (event) {
2648         case NETDEV_RESEND_IGMP:
2649                 if (idev)
2650                         ipv6_mc_rejoin_groups(idev);
2651                 break;
2652         default:
2653                 break;
2654         }
2655
2656         return NOTIFY_DONE;
2657 }
2658
2659 static struct notifier_block igmp6_netdev_notifier = {
2660         .notifier_call = ipv6_mc_netdev_event,
2661 };
2662
2663 #ifdef CONFIG_PROC_FS
2664 struct igmp6_mc_iter_state {
2665         struct seq_net_private p;
2666         struct net_device *dev;
2667         struct inet6_dev *idev;
2668 };
2669
2670 #define igmp6_mc_seq_private(seq)       ((struct igmp6_mc_iter_state *)(seq)->private)
2671
2672 static inline struct ifmcaddr6 *igmp6_mc_get_first(struct seq_file *seq)
2673 {
2674         struct ifmcaddr6 *im = NULL;
2675         struct igmp6_mc_iter_state *state = igmp6_mc_seq_private(seq);
2676         struct net *net = seq_file_net(seq);
2677
2678         state->idev = NULL;
2679         for_each_netdev_rcu(net, state->dev) {
2680                 struct inet6_dev *idev;
2681                 idev = __in6_dev_get(state->dev);
2682                 if (!idev)
2683                         continue;
2684                 read_lock_bh(&idev->lock);
2685                 im = idev->mc_list;
2686                 if (im) {
2687                         state->idev = idev;
2688                         break;
2689                 }
2690                 read_unlock_bh(&idev->lock);
2691         }
2692         return im;
2693 }
2694
2695 static struct ifmcaddr6 *igmp6_mc_get_next(struct seq_file *seq, struct ifmcaddr6 *im)
2696 {
2697         struct igmp6_mc_iter_state *state = igmp6_mc_seq_private(seq);
2698
2699         im = im->next;
2700         while (!im) {
2701                 if (likely(state->idev))
2702                         read_unlock_bh(&state->idev->lock);
2703
2704                 state->dev = next_net_device_rcu(state->dev);
2705                 if (!state->dev) {
2706                         state->idev = NULL;
2707                         break;
2708                 }
2709                 state->idev = __in6_dev_get(state->dev);
2710                 if (!state->idev)
2711                         continue;
2712                 read_lock_bh(&state->idev->lock);
2713                 im = state->idev->mc_list;
2714         }
2715         return im;
2716 }
2717
2718 static struct ifmcaddr6 *igmp6_mc_get_idx(struct seq_file *seq, loff_t pos)
2719 {
2720         struct ifmcaddr6 *im = igmp6_mc_get_first(seq);
2721         if (im)
2722                 while (pos && (im = igmp6_mc_get_next(seq, im)) != NULL)
2723                         --pos;
2724         return pos ? NULL : im;
2725 }
2726
2727 static void *igmp6_mc_seq_start(struct seq_file *seq, loff_t *pos)
2728         __acquires(RCU)
2729 {
2730         rcu_read_lock();
2731         return igmp6_mc_get_idx(seq, *pos);
2732 }
2733
2734 static void *igmp6_mc_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2735 {
2736         struct ifmcaddr6 *im = igmp6_mc_get_next(seq, v);
2737
2738         ++*pos;
2739         return im;
2740 }
2741
2742 static void igmp6_mc_seq_stop(struct seq_file *seq, void *v)
2743         __releases(RCU)
2744 {
2745         struct igmp6_mc_iter_state *state = igmp6_mc_seq_private(seq);
2746
2747         if (likely(state->idev)) {
2748                 read_unlock_bh(&state->idev->lock);
2749                 state->idev = NULL;
2750         }
2751         state->dev = NULL;
2752         rcu_read_unlock();
2753 }
2754
2755 static int igmp6_mc_seq_show(struct seq_file *seq, void *v)
2756 {
2757         struct ifmcaddr6 *im = (struct ifmcaddr6 *)v;
2758         struct igmp6_mc_iter_state *state = igmp6_mc_seq_private(seq);
2759
2760         seq_printf(seq,
2761                    "%-4d %-15s %pi6 %5d %08X %ld\n",
2762                    state->dev->ifindex, state->dev->name,
2763                    &im->mca_addr,
2764                    im->mca_users, im->mca_flags,
2765                    (im->mca_flags&MAF_TIMER_RUNNING) ?
2766                    jiffies_to_clock_t(im->mca_timer.expires-jiffies) : 0);
2767         return 0;
2768 }
2769
2770 static const struct seq_operations igmp6_mc_seq_ops = {
2771         .start  =       igmp6_mc_seq_start,
2772         .next   =       igmp6_mc_seq_next,
2773         .stop   =       igmp6_mc_seq_stop,
2774         .show   =       igmp6_mc_seq_show,
2775 };
2776
2777 struct igmp6_mcf_iter_state {
2778         struct seq_net_private p;
2779         struct net_device *dev;
2780         struct inet6_dev *idev;
2781         struct ifmcaddr6 *im;
2782 };
2783
2784 #define igmp6_mcf_seq_private(seq)      ((struct igmp6_mcf_iter_state *)(seq)->private)
2785
2786 static inline struct ip6_sf_list *igmp6_mcf_get_first(struct seq_file *seq)
2787 {
2788         struct ip6_sf_list *psf = NULL;
2789         struct ifmcaddr6 *im = NULL;
2790         struct igmp6_mcf_iter_state *state = igmp6_mcf_seq_private(seq);
2791         struct net *net = seq_file_net(seq);
2792
2793         state->idev = NULL;
2794         state->im = NULL;
2795         for_each_netdev_rcu(net, state->dev) {
2796                 struct inet6_dev *idev;
2797                 idev = __in6_dev_get(state->dev);
2798                 if (unlikely(idev == NULL))
2799                         continue;
2800                 read_lock_bh(&idev->lock);
2801                 im = idev->mc_list;
2802                 if (likely(im)) {
2803                         spin_lock_bh(&im->mca_lock);
2804                         psf = im->mca_sources;
2805                         if (likely(psf)) {
2806                                 state->im = im;
2807                                 state->idev = idev;
2808                                 break;
2809                         }
2810                         spin_unlock_bh(&im->mca_lock);
2811                 }
2812                 read_unlock_bh(&idev->lock);
2813         }
2814         return psf;
2815 }
2816
2817 static struct ip6_sf_list *igmp6_mcf_get_next(struct seq_file *seq, struct ip6_sf_list *psf)
2818 {
2819         struct igmp6_mcf_iter_state *state = igmp6_mcf_seq_private(seq);
2820
2821         psf = psf->sf_next;
2822         while (!psf) {
2823                 spin_unlock_bh(&state->im->mca_lock);
2824                 state->im = state->im->next;
2825                 while (!state->im) {
2826                         if (likely(state->idev))
2827                                 read_unlock_bh(&state->idev->lock);
2828
2829                         state->dev = next_net_device_rcu(state->dev);
2830                         if (!state->dev) {
2831                                 state->idev = NULL;
2832                                 goto out;
2833                         }
2834                         state->idev = __in6_dev_get(state->dev);
2835                         if (!state->idev)
2836                                 continue;
2837                         read_lock_bh(&state->idev->lock);
2838                         state->im = state->idev->mc_list;
2839                 }
2840                 if (!state->im)
2841                         break;
2842                 spin_lock_bh(&state->im->mca_lock);
2843                 psf = state->im->mca_sources;
2844         }
2845 out:
2846         return psf;
2847 }
2848
2849 static struct ip6_sf_list *igmp6_mcf_get_idx(struct seq_file *seq, loff_t pos)
2850 {
2851         struct ip6_sf_list *psf = igmp6_mcf_get_first(seq);
2852         if (psf)
2853                 while (pos && (psf = igmp6_mcf_get_next(seq, psf)) != NULL)
2854                         --pos;
2855         return pos ? NULL : psf;
2856 }
2857
2858 static void *igmp6_mcf_seq_start(struct seq_file *seq, loff_t *pos)
2859         __acquires(RCU)
2860 {
2861         rcu_read_lock();
2862         return *pos ? igmp6_mcf_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
2863 }
2864
2865 static void *igmp6_mcf_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2866 {
2867         struct ip6_sf_list *psf;
2868         if (v == SEQ_START_TOKEN)
2869                 psf = igmp6_mcf_get_first(seq);
2870         else
2871                 psf = igmp6_mcf_get_next(seq, v);
2872         ++*pos;
2873         return psf;
2874 }
2875
2876 static void igmp6_mcf_seq_stop(struct seq_file *seq, void *v)
2877         __releases(RCU)
2878 {
2879         struct igmp6_mcf_iter_state *state = igmp6_mcf_seq_private(seq);
2880         if (likely(state->im)) {
2881                 spin_unlock_bh(&state->im->mca_lock);
2882                 state->im = NULL;
2883         }
2884         if (likely(state->idev)) {
2885                 read_unlock_bh(&state->idev->lock);
2886                 state->idev = NULL;
2887         }
2888         state->dev = NULL;
2889         rcu_read_unlock();
2890 }
2891
2892 static int igmp6_mcf_seq_show(struct seq_file *seq, void *v)
2893 {
2894         struct ip6_sf_list *psf = (struct ip6_sf_list *)v;
2895         struct igmp6_mcf_iter_state *state = igmp6_mcf_seq_private(seq);
2896
2897         if (v == SEQ_START_TOKEN) {
2898                 seq_puts(seq, "Idx Device                Multicast Address                   Source Address    INC    EXC\n");
2899         } else {
2900                 seq_printf(seq,
2901                            "%3d %6.6s %pi6 %pi6 %6lu %6lu\n",
2902                            state->dev->ifindex, state->dev->name,
2903                            &state->im->mca_addr,
2904                            &psf->sf_addr,
2905                            psf->sf_count[MCAST_INCLUDE],
2906                            psf->sf_count[MCAST_EXCLUDE]);
2907         }
2908         return 0;
2909 }
2910
2911 static const struct seq_operations igmp6_mcf_seq_ops = {
2912         .start  =       igmp6_mcf_seq_start,
2913         .next   =       igmp6_mcf_seq_next,
2914         .stop   =       igmp6_mcf_seq_stop,
2915         .show   =       igmp6_mcf_seq_show,
2916 };
2917
2918 static int __net_init igmp6_proc_init(struct net *net)
2919 {
2920         int err;
2921
2922         err = -ENOMEM;
2923         if (!proc_create_net("igmp6", 0444, net->proc_net, &igmp6_mc_seq_ops,
2924                         sizeof(struct igmp6_mc_iter_state)))
2925                 goto out;
2926         if (!proc_create_net("mcfilter6", 0444, net->proc_net,
2927                         &igmp6_mcf_seq_ops,
2928                         sizeof(struct igmp6_mcf_iter_state)))
2929                 goto out_proc_net_igmp6;
2930
2931         err = 0;
2932 out:
2933         return err;
2934
2935 out_proc_net_igmp6:
2936         remove_proc_entry("igmp6", net->proc_net);
2937         goto out;
2938 }
2939
2940 static void __net_exit igmp6_proc_exit(struct net *net)
2941 {
2942         remove_proc_entry("mcfilter6", net->proc_net);
2943         remove_proc_entry("igmp6", net->proc_net);
2944 }
2945 #else
2946 static inline int igmp6_proc_init(struct net *net)
2947 {
2948         return 0;
2949 }
2950 static inline void igmp6_proc_exit(struct net *net)
2951 {
2952 }
2953 #endif
2954
2955 static int __net_init igmp6_net_init(struct net *net)
2956 {
2957         int err;
2958
2959         err = inet_ctl_sock_create(&net->ipv6.igmp_sk, PF_INET6,
2960                                    SOCK_RAW, IPPROTO_ICMPV6, net);
2961         if (err < 0) {
2962                 pr_err("Failed to initialize the IGMP6 control socket (err %d)\n",
2963                        err);
2964                 goto out;
2965         }
2966
2967         inet6_sk(net->ipv6.igmp_sk)->hop_limit = 1;
2968
2969         err = inet_ctl_sock_create(&net->ipv6.mc_autojoin_sk, PF_INET6,
2970                                    SOCK_RAW, IPPROTO_ICMPV6, net);
2971         if (err < 0) {
2972                 pr_err("Failed to initialize the IGMP6 autojoin socket (err %d)\n",
2973                        err);
2974                 goto out_sock_create;
2975         }
2976
2977         err = igmp6_proc_init(net);
2978         if (err)
2979                 goto out_sock_create_autojoin;
2980
2981         return 0;
2982
2983 out_sock_create_autojoin:
2984         inet_ctl_sock_destroy(net->ipv6.mc_autojoin_sk);
2985 out_sock_create:
2986         inet_ctl_sock_destroy(net->ipv6.igmp_sk);
2987 out:
2988         return err;
2989 }
2990
2991 static void __net_exit igmp6_net_exit(struct net *net)
2992 {
2993         inet_ctl_sock_destroy(net->ipv6.igmp_sk);
2994         inet_ctl_sock_destroy(net->ipv6.mc_autojoin_sk);
2995         igmp6_proc_exit(net);
2996 }
2997
2998 static struct pernet_operations igmp6_net_ops = {
2999         .init = igmp6_net_init,
3000         .exit = igmp6_net_exit,
3001 };
3002
3003 int __init igmp6_init(void)
3004 {
3005         return register_pernet_subsys(&igmp6_net_ops);
3006 }
3007
3008 int __init igmp6_late_init(void)
3009 {
3010         return register_netdevice_notifier(&igmp6_netdev_notifier);
3011 }
3012
3013 void igmp6_cleanup(void)
3014 {
3015         unregister_pernet_subsys(&igmp6_net_ops);
3016 }
3017
3018 void igmp6_late_cleanup(void)
3019 {
3020         unregister_netdevice_notifier(&igmp6_netdev_notifier);
3021 }