net/ipv6/mcast.c

   1 /*
   2  *      Multicast support for IPv6
   3  *      Linux INET6 implementation
   4  *
   5  *      Authors:
   6  *      Pedro Roque             <roque@di.fc.ul.pt>
   7  *
   8  *      Based on linux/ipv4/igmp.c and linux/ipv4/ip_sockglue.c
   9  *
  10  *      This program is free software; you can redistribute it and/or
  11  *      modify it under the terms of the GNU General Public License
  12  *      as published by the Free Software Foundation; either version
  13  *      2 of the License, or (at your option) any later version.
  14  */
  15
  16 /* Changes:
  17  *
  18  *      yoshfuji        : fix format of router-alert option
  19  *      YOSHIFUJI Hideaki @USAGI:
  20  *              Fixed source address for MLD message based on
  21  *              <draft-ietf-magma-mld-source-05.txt>.
  22  *      YOSHIFUJI Hideaki @USAGI:
  23  *              - Ignore Queries for invalid addresses.
  24  *              - MLD for link-local addresses.
  25  *      David L Stevens <dlstevens@us.ibm.com>:
  26  *              - MLDv2 support
  27  */
  28
  29 #include <linux/module.h>
  30 #include <linux/errno.h>
  31 #include <linux/types.h>
  32 #include <linux/string.h>
  33 #include <linux/socket.h>
  34 #include <linux/sockios.h>
  35 #include <linux/jiffies.h>
  36 #include <linux/times.h>
  37 #include <linux/net.h>
  38 #include <linux/in.h>
  39 #include <linux/in6.h>
  40 #include <linux/netdevice.h>
  41 #include <linux/if_arp.h>
  42 #include <linux/route.h>
  43 #include <linux/init.h>
  44 #include <linux/proc_fs.h>
  45 #include <linux/seq_file.h>
  46 #include <linux/slab.h>
  47 #include <linux/pkt_sched.h>
  48 #include <net/mld.h>
  49
  50 #include <linux/netfilter.h>
  51 #include <linux/netfilter_ipv6.h>
  52
  53 #include <net/net_namespace.h>
  54 #include <net/sock.h>
  55 #include <net/snmp.h>
  56
  57 #include <net/ipv6.h>
  58 #include <net/protocol.h>
  59 #include <net/if_inet6.h>
  60 #include <net/ndisc.h>
  61 #include <net/addrconf.h>
  62 #include <net/ip6_route.h>
  63 #include <net/inet_common.h>
  64
  65 #include <net/ip6_checksum.h>
  66
  67 /* Ensure that we have struct in6_addr aligned on 32bit word. */
  68 static int __mld2_query_bugs[] __attribute__((__unused__)) = {
  69         BUILD_BUG_ON_ZERO(offsetof(struct mld2_query, mld2q_srcs) % 4),
  70         BUILD_BUG_ON_ZERO(offsetof(struct mld2_report, mld2r_grec) % 4),
  71         BUILD_BUG_ON_ZERO(offsetof(struct mld2_grec, grec_mca) % 4)
  72 };
  73
  74 static struct in6_addr mld2_all_mcr = MLD2_ALL_MCR_INIT;
  75
  76 static void igmp6_join_group(struct ifmcaddr6 *ma);
  77 static void igmp6_leave_group(struct ifmcaddr6 *ma);
  78 static void igmp6_timer_handler(struct timer_list *t);
  79
  80 static void mld_gq_timer_expire(struct timer_list *t);
  81 static void mld_ifc_timer_expire(struct timer_list *t);
  82 static void mld_ifc_event(struct inet6_dev *idev);
  83 static void mld_add_delrec(struct inet6_dev *idev, struct ifmcaddr6 *pmc);
  84 static void mld_del_delrec(struct inet6_dev *idev, struct ifmcaddr6 *pmc);
  85 static void mld_clear_delrec(struct inet6_dev *idev);
  86 static bool mld_in_v1_mode(const struct inet6_dev *idev);
  87 static int sf_setstate(struct ifmcaddr6 *pmc);
  88 static void sf_markstate(struct ifmcaddr6 *pmc);
  89 static void ip6_mc_clear_src(struct ifmcaddr6 *pmc);
  90 static int ip6_mc_del_src(struct inet6_dev *idev, const struct in6_addr *pmca,
  91                           int sfmode, int sfcount, const struct in6_addr *psfsrc,
  92                           int delta);
  93 static int ip6_mc_add_src(struct inet6_dev *idev, const struct in6_addr *pmca,
  94                           int sfmode, int sfcount, const struct in6_addr *psfsrc,
  95                           int delta);
  96 static int ip6_mc_leave_src(struct sock *sk, struct ipv6_mc_socklist *iml,
  97                             struct inet6_dev *idev);
  98 static int __ipv6_dev_mc_inc(struct net_device *dev,
  99                              const struct in6_addr *addr, unsigned int mode);
 100
 101 #define MLD_QRV_DEFAULT         2
 102 /* RFC3810, 9.2. Query Interval */
 103 #define MLD_QI_DEFAULT          (125 * HZ)
 104 /* RFC3810, 9.3. Query Response Interval */
 105 #define MLD_QRI_DEFAULT         (10 * HZ)
 106
 107 /* RFC3810, 8.1 Query Version Distinctions */
 108 #define MLD_V1_QUERY_LEN        24
 109 #define MLD_V2_QUERY_LEN_MIN    28
 110
 111 #define IPV6_MLD_MAX_MSF        64
 112
 113 int sysctl_mld_max_msf __read_mostly = IPV6_MLD_MAX_MSF;
 114 int sysctl_mld_qrv __read_mostly = MLD_QRV_DEFAULT;
 115
 116 /*
 117  *      socket join on multicast group
 118  */
 119
 120 #define for_each_pmc_rcu(np, pmc)                               \
 121         for (pmc = rcu_dereference(np->ipv6_mc_list);           \
 122              pmc != NULL;                                       \
 123              pmc = rcu_dereference(pmc->next))
 124
 125 static int unsolicited_report_interval(struct inet6_dev *idev)
 126 {
 127         int iv;
 128
 129         if (mld_in_v1_mode(idev))
 130                 iv = idev->cnf.mldv1_unsolicited_report_interval;
 131         else
 132                 iv = idev->cnf.mldv2_unsolicited_report_interval;
 133
 134         return iv > 0 ? iv : 1;
 135 }
 136
 137 static int __ipv6_sock_mc_join(struct sock *sk, int ifindex,
 138                                const struct in6_addr *addr, unsigned int mode)
 139 {
 140         struct net_device *dev = NULL;
 141         struct ipv6_mc_socklist *mc_lst;
 142         struct ipv6_pinfo *np = inet6_sk(sk);
 143         struct net *net = sock_net(sk);
 144         int err;
 145
 146         ASSERT_RTNL();
 147
 148         if (!ipv6_addr_is_multicast(addr))
 149                 return -EINVAL;
 150
 151         rcu_read_lock();
 152         for_each_pmc_rcu(np, mc_lst) {
 153                 if ((ifindex == 0 || mc_lst->ifindex == ifindex) &&
 154                     ipv6_addr_equal(&mc_lst->addr, addr)) {
 155                         rcu_read_unlock();
 156                         return -EADDRINUSE;
 157                 }
 158         }
 159         rcu_read_unlock();
 160
 161         mc_lst = sock_kmalloc(sk, sizeof(struct ipv6_mc_socklist), GFP_KERNEL);
 162
 163         if (!mc_lst)
 164                 return -ENOMEM;
 165
 166         mc_lst->next = NULL;
 167         mc_lst->addr = *addr;
 168
 169         if (ifindex == 0) {
 170                 struct rt6_info *rt;
 171                 rt = rt6_lookup(net, addr, NULL, 0, NULL, 0);
 172                 if (rt) {
 173                         dev = rt->dst.dev;
 174                         ip6_rt_put(rt);
 175                 }
 176         } else
 177                 dev = __dev_get_by_index(net, ifindex);
 178
 179         if (!dev) {
 180                 sock_kfree_s(sk, mc_lst, sizeof(*mc_lst));
 181                 return -ENODEV;
 182         }
 183
 184         mc_lst->ifindex = dev->ifindex;
 185         mc_lst->sfmode = mode;
 186         rwlock_init(&mc_lst->sflock);
 187         mc_lst->sflist = NULL;
 188
 189         /*
 190          *      now add/increase the group membership on the device
 191          */
 192
 193         err = __ipv6_dev_mc_inc(dev, addr, mode);
 194
 195         if (err) {
 196                 sock_kfree_s(sk, mc_lst, sizeof(*mc_lst));
 197                 return err;
 198         }
 199
 200         mc_lst->next = np->ipv6_mc_list;
 201         rcu_assign_pointer(np->ipv6_mc_list, mc_lst);
 202
 203         return 0;
 204 }
 205
 206 int ipv6_sock_mc_join(struct sock *sk, int ifindex, const struct in6_addr *addr)
 207 {
 208         return __ipv6_sock_mc_join(sk, ifindex, addr, MCAST_EXCLUDE);
 209 }
 210 EXPORT_SYMBOL(ipv6_sock_mc_join);
 211
 212 int ipv6_sock_mc_join_ssm(struct sock *sk, int ifindex,
 213                           const struct in6_addr *addr, unsigned int mode)
 214 {
 215         return __ipv6_sock_mc_join(sk, ifindex, addr, mode);
 216 }
 217
 218 /*
 219  *      socket leave on multicast group
 220  */
 221 int ipv6_sock_mc_drop(struct sock *sk, int ifindex, const struct in6_addr *addr)
 222 {
 223         struct ipv6_pinfo *np = inet6_sk(sk);
 224         struct ipv6_mc_socklist *mc_lst;
 225         struct ipv6_mc_socklist __rcu **lnk;
 226         struct net *net = sock_net(sk);
 227
 228         ASSERT_RTNL();
 229
 230         if (!ipv6_addr_is_multicast(addr))
 231                 return -EINVAL;
 232
 233         for (lnk = &np->ipv6_mc_list;
 234              (mc_lst = rtnl_dereference(*lnk)) != NULL;
 235               lnk = &mc_lst->next) {
 236                 if ((ifindex == 0 || mc_lst->ifindex == ifindex) &&
 237                     ipv6_addr_equal(&mc_lst->addr, addr)) {
 238                         struct net_device *dev;
 239
 240                         *lnk = mc_lst->next;
 241
 242                         dev = __dev_get_by_index(net, mc_lst->ifindex);
 243                         if (dev) {
 244                                 struct inet6_dev *idev = __in6_dev_get(dev);
 245
 246                                 (void) ip6_mc_leave_src(sk, mc_lst, idev);
 247                                 if (idev)
 248                                         __ipv6_dev_mc_dec(idev, &mc_lst->addr);
 249                         } else
 250                                 (void) ip6_mc_leave_src(sk, mc_lst, NULL);
 251
 252                         atomic_sub(sizeof(*mc_lst), &sk->sk_omem_alloc);
 253                         kfree_rcu(mc_lst, rcu);
 254                         return 0;
 255                 }
 256         }
 257
 258         return -EADDRNOTAVAIL;
 259 }
 260 EXPORT_SYMBOL(ipv6_sock_mc_drop);
 261
 262 /* called with rcu_read_lock() */
 263 static struct inet6_dev *ip6_mc_find_dev_rcu(struct net *net,
 264                                              const struct in6_addr *group,
 265                                              int ifindex)
 266 {
 267         struct net_device *dev = NULL;
 268         struct inet6_dev *idev = NULL;
 269
 270         if (ifindex == 0) {
 271                 struct rt6_info *rt = rt6_lookup(net, group, NULL, 0, NULL, 0);
 272
 273                 if (rt) {
 274                         dev = rt->dst.dev;
 275                         ip6_rt_put(rt);
 276                 }
 277         } else
 278                 dev = dev_get_by_index_rcu(net, ifindex);
 279
 280         if (!dev)
 281                 return NULL;
 282         idev = __in6_dev_get(dev);
 283         if (!idev)
 284                 return NULL;
 285         read_lock_bh(&idev->lock);
 286         if (idev->dead) {
 287                 read_unlock_bh(&idev->lock);
 288                 return NULL;
 289         }
 290         return idev;
 291 }
 292
 293 void __ipv6_sock_mc_close(struct sock *sk)
 294 {
 295         struct ipv6_pinfo *np = inet6_sk(sk);
 296         struct ipv6_mc_socklist *mc_lst;
 297         struct net *net = sock_net(sk);
 298
 299         ASSERT_RTNL();
 300
 301         while ((mc_lst = rtnl_dereference(np->ipv6_mc_list)) != NULL) {
 302                 struct net_device *dev;
 303
 304                 np->ipv6_mc_list = mc_lst->next;
 305
 306                 dev = __dev_get_by_index(net, mc_lst->ifindex);
 307                 if (dev) {
 308                         struct inet6_dev *idev = __in6_dev_get(dev);
 309
 310                         (void) ip6_mc_leave_src(sk, mc_lst, idev);
 311                         if (idev)
 312                                 __ipv6_dev_mc_dec(idev, &mc_lst->addr);
 313                 } else
 314                         (void) ip6_mc_leave_src(sk, mc_lst, NULL);
 315
 316                 atomic_sub(sizeof(*mc_lst), &sk->sk_omem_alloc);
 317                 kfree_rcu(mc_lst, rcu);
 318         }
 319 }
 320
 321 void ipv6_sock_mc_close(struct sock *sk)
 322 {
 323         struct ipv6_pinfo *np = inet6_sk(sk);
 324
 325         if (!rcu_access_pointer(np->ipv6_mc_list))
 326                 return;
 327         rtnl_lock();
 328         __ipv6_sock_mc_close(sk);
 329         rtnl_unlock();
 330 }
 331
 332 int ip6_mc_source(int add, int omode, struct sock *sk,
 333         struct group_source_req *pgsr)
 334 {
 335         struct in6_addr *source, *group;
 336         struct ipv6_mc_socklist *pmc;
 337         struct inet6_dev *idev;
 338         struct ipv6_pinfo *inet6 = inet6_sk(sk);
 339         struct ip6_sf_socklist *psl;
 340         struct net *net = sock_net(sk);
 341         int i, j, rv;
 342         int leavegroup = 0;
 343         int pmclocked = 0;
 344         int err;
 345
 346         source = &((struct sockaddr_in6 *)&pgsr->gsr_source)->sin6_addr;
 347         group = &((struct sockaddr_in6 *)&pgsr->gsr_group)->sin6_addr;
 348
 349         if (!ipv6_addr_is_multicast(group))
 350                 return -EINVAL;
 351
 352         rcu_read_lock();
 353         idev = ip6_mc_find_dev_rcu(net, group, pgsr->gsr_interface);
 354         if (!idev) {
 355                 rcu_read_unlock();
 356                 return -ENODEV;
 357         }
 358
 359         err = -EADDRNOTAVAIL;
 360
 361         for_each_pmc_rcu(inet6, pmc) {
 362                 if (pgsr->gsr_interface && pmc->ifindex != pgsr->gsr_interface)
 363                         continue;
 364                 if (ipv6_addr_equal(&pmc->addr, group))
 365                         break;
 366         }
 367         if (!pmc) {             /* must have a prior join */
 368                 err = -EINVAL;
 369                 goto done;
 370         }
 371         /* if a source filter was set, must be the same mode as before */
 372         if (pmc->sflist) {
 373                 if (pmc->sfmode != omode) {
 374                         err = -EINVAL;
 375                         goto done;
 376                 }
 377         } else if (pmc->sfmode != omode) {
 378                 /* allow mode switches for empty-set filters */
 379                 ip6_mc_add_src(idev, group, omode, 0, NULL, 0);
 380                 ip6_mc_del_src(idev, group, pmc->sfmode, 0, NULL, 0);
 381                 pmc->sfmode = omode;
 382         }
 383
 384         write_lock(&pmc->sflock);
 385         pmclocked = 1;
 386
 387         psl = pmc->sflist;
 388         if (!add) {
 389                 if (!psl)
 390                         goto done;      /* err = -EADDRNOTAVAIL */
 391                 rv = !0;
 392                 for (i = 0; i < psl->sl_count; i++) {
 393                         rv = !ipv6_addr_equal(&psl->sl_addr[i], source);
 394                         if (rv == 0)
 395                                 break;
 396                 }
 397                 if (rv)         /* source not found */
 398                         goto done;      /* err = -EADDRNOTAVAIL */
 399
 400                 /* special case - (INCLUDE, empty) == LEAVE_GROUP */
 401                 if (psl->sl_count == 1 && omode == MCAST_INCLUDE) {
 402                         leavegroup = 1;
 403                         goto done;
 404                 }
 405
 406                 /* update the interface filter */
 407                 ip6_mc_del_src(idev, group, omode, 1, source, 1);
 408
 409                 for (j = i+1; j < psl->sl_count; j++)
 410                         psl->sl_addr[j-1] = psl->sl_addr[j];
 411                 psl->sl_count--;
 412                 err = 0;
 413                 goto done;
 414         }
 415         /* else, add a new source to the filter */
 416
 417         if (psl && psl->sl_count >= sysctl_mld_max_msf) {
 418                 err = -ENOBUFS;
 419                 goto done;
 420         }
 421         if (!psl || psl->sl_count == psl->sl_max) {
 422                 struct ip6_sf_socklist *newpsl;
 423                 int count = IP6_SFBLOCK;
 424
 425                 if (psl)
 426                         count += psl->sl_max;
 427                 newpsl = sock_kmalloc(sk, IP6_SFLSIZE(count), GFP_ATOMIC);
 428                 if (!newpsl) {
 429                         err = -ENOBUFS;
 430                         goto done;
 431                 }
 432                 newpsl->sl_max = count;
 433                 newpsl->sl_count = count - IP6_SFBLOCK;
 434                 if (psl) {
 435                         for (i = 0; i < psl->sl_count; i++)
 436                                 newpsl->sl_addr[i] = psl->sl_addr[i];
 437                         sock_kfree_s(sk, psl, IP6_SFLSIZE(psl->sl_max));
 438                 }
 439                 pmc->sflist = psl = newpsl;
 440         }
 441         rv = 1; /* > 0 for insert logic below if sl_count is 0 */
 442         for (i = 0; i < psl->sl_count; i++) {
 443                 rv = !ipv6_addr_equal(&psl->sl_addr[i], source);
 444                 if (rv == 0) /* There is an error in the address. */
 445                         goto done;
 446         }
 447         for (j = psl->sl_count-1; j >= i; j--)
 448                 psl->sl_addr[j+1] = psl->sl_addr[j];
 449         psl->sl_addr[i] = *source;
 450         psl->sl_count++;
 451         err = 0;
 452         /* update the interface list */
 453         ip6_mc_add_src(idev, group, omode, 1, source, 1);
 454 done:
 455         if (pmclocked)
 456                 write_unlock(&pmc->sflock);
 457         read_unlock_bh(&idev->lock);
 458         rcu_read_unlock();
 459         if (leavegroup)
 460                 err = ipv6_sock_mc_drop(sk, pgsr->gsr_interface, group);
 461         return err;
 462 }
 463
 464 int ip6_mc_msfilter(struct sock *sk, struct group_filter *gsf)
 465 {
 466         const struct in6_addr *group;
 467         struct ipv6_mc_socklist *pmc;
 468         struct inet6_dev *idev;
 469         struct ipv6_pinfo *inet6 = inet6_sk(sk);
 470         struct ip6_sf_socklist *newpsl, *psl;
 471         struct net *net = sock_net(sk);
 472         int leavegroup = 0;
 473         int i, err;
 474
 475         group = &((struct sockaddr_in6 *)&gsf->gf_group)->sin6_addr;
 476
 477         if (!ipv6_addr_is_multicast(group))
 478                 return -EINVAL;
 479         if (gsf->gf_fmode != MCAST_INCLUDE &&
 480             gsf->gf_fmode != MCAST_EXCLUDE)
 481                 return -EINVAL;
 482
 483         rcu_read_lock();
 484         idev = ip6_mc_find_dev_rcu(net, group, gsf->gf_interface);
 485
 486         if (!idev) {
 487                 rcu_read_unlock();
 488                 return -ENODEV;
 489         }
 490
 491         err = 0;
 492
 493         if (gsf->gf_fmode == MCAST_INCLUDE && gsf->gf_numsrc == 0) {
 494                 leavegroup = 1;
 495                 goto done;
 496         }
 497
 498         for_each_pmc_rcu(inet6, pmc) {
 499                 if (pmc->ifindex != gsf->gf_interface)
 500                         continue;
 501                 if (ipv6_addr_equal(&pmc->addr, group))
 502                         break;
 503         }
 504         if (!pmc) {             /* must have a prior join */
 505                 err = -EINVAL;
 506                 goto done;
 507         }
 508         if (gsf->gf_numsrc) {
 509                 newpsl = sock_kmalloc(sk, IP6_SFLSIZE(gsf->gf_numsrc),
 510                                                           GFP_ATOMIC);
 511                 if (!newpsl) {
 512                         err = -ENOBUFS;
 513                         goto done;
 514                 }
 515                 newpsl->sl_max = newpsl->sl_count = gsf->gf_numsrc;
 516                 for (i = 0; i < newpsl->sl_count; ++i) {
 517                         struct sockaddr_in6 *psin6;
 518
 519                         psin6 = (struct sockaddr_in6 *)&gsf->gf_slist[i];
 520                         newpsl->sl_addr[i] = psin6->sin6_addr;
 521                 }
 522                 err = ip6_mc_add_src(idev, group, gsf->gf_fmode,
 523                         newpsl->sl_count, newpsl->sl_addr, 0);
 524                 if (err) {
 525                         sock_kfree_s(sk, newpsl, IP6_SFLSIZE(newpsl->sl_max));
 526                         goto done;
 527                 }
 528         } else {
 529                 newpsl = NULL;
 530                 (void) ip6_mc_add_src(idev, group, gsf->gf_fmode, 0, NULL, 0);
 531         }
 532
 533         write_lock(&pmc->sflock);
 534         psl = pmc->sflist;
 535         if (psl) {
 536                 (void) ip6_mc_del_src(idev, group, pmc->sfmode,
 537                         psl->sl_count, psl->sl_addr, 0);
 538                 sock_kfree_s(sk, psl, IP6_SFLSIZE(psl->sl_max));
 539         } else
 540                 (void) ip6_mc_del_src(idev, group, pmc->sfmode, 0, NULL, 0);
 541         pmc->sflist = newpsl;
 542         pmc->sfmode = gsf->gf_fmode;
 543         write_unlock(&pmc->sflock);
 544         err = 0;
 545 done:
 546         read_unlock_bh(&idev->lock);
 547         rcu_read_unlock();
 548         if (leavegroup)
 549                 err = ipv6_sock_mc_drop(sk, gsf->gf_interface, group);
 550         return err;
 551 }
 552
 553 int ip6_mc_msfget(struct sock *sk, struct group_filter *gsf,
 554         struct group_filter __user *optval, int __user *optlen)
 555 {
 556         int err, i, count, copycount;
 557         const struct in6_addr *group;
 558         struct ipv6_mc_socklist *pmc;
 559         struct inet6_dev *idev;
 560         struct ipv6_pinfo *inet6 = inet6_sk(sk);
 561         struct ip6_sf_socklist *psl;
 562         struct net *net = sock_net(sk);
 563
 564         group = &((struct sockaddr_in6 *)&gsf->gf_group)->sin6_addr;
 565
 566         if (!ipv6_addr_is_multicast(group))
 567                 return -EINVAL;
 568
 569         rcu_read_lock();
 570         idev = ip6_mc_find_dev_rcu(net, group, gsf->gf_interface);
 571
 572         if (!idev) {
 573                 rcu_read_unlock();
 574                 return -ENODEV;
 575         }
 576
 577         err = -EADDRNOTAVAIL;
 578         /* changes to the ipv6_mc_list require the socket lock and
 579          * rtnl lock. We have the socket lock and rcu read lock,
 580          * so reading the list is safe.
 581          */
 582
 583         for_each_pmc_rcu(inet6, pmc) {
 584                 if (pmc->ifindex != gsf->gf_interface)
 585                         continue;
 586                 if (ipv6_addr_equal(group, &pmc->addr))
 587                         break;
 588         }
 589         if (!pmc)               /* must have a prior join */
 590                 goto done;
 591         gsf->gf_fmode = pmc->sfmode;
 592         psl = pmc->sflist;
 593         count = psl ? psl->sl_count : 0;
 594         read_unlock_bh(&idev->lock);
 595         rcu_read_unlock();
 596
 597         copycount = count < gsf->gf_numsrc ? count : gsf->gf_numsrc;
 598         gsf->gf_numsrc = count;
 599         if (put_user(GROUP_FILTER_SIZE(copycount), optlen) ||
 600             copy_to_user(optval, gsf, GROUP_FILTER_SIZE(0))) {
 601                 return -EFAULT;
 602         }
 603         /* changes to psl require the socket lock, and a write lock
 604          * on pmc->sflock. We have the socket lock so reading here is safe.
 605          */
 606         for (i = 0; i < copycount; i++) {
 607                 struct sockaddr_in6 *psin6;
 608                 struct sockaddr_storage ss;
 609
 610                 psin6 = (struct sockaddr_in6 *)&ss;
 611                 memset(&ss, 0, sizeof(ss));
 612                 psin6->sin6_family = AF_INET6;
 613                 psin6->sin6_addr = psl->sl_addr[i];
 614                 if (copy_to_user(&optval->gf_slist[i], &ss, sizeof(ss)))
 615                         return -EFAULT;
 616         }
 617         return 0;
 618 done:
 619         read_unlock_bh(&idev->lock);
 620         rcu_read_unlock();
 621         return err;
 622 }
 623
 624 bool inet6_mc_check(struct sock *sk, const struct in6_addr *mc_addr,
 625                     const struct in6_addr *src_addr)
 626 {
 627         struct ipv6_pinfo *np = inet6_sk(sk);
 628         struct ipv6_mc_socklist *mc;
 629         struct ip6_sf_socklist *psl;
 630         bool rv = true;
 631
 632         rcu_read_lock();
 633         for_each_pmc_rcu(np, mc) {
 634                 if (ipv6_addr_equal(&mc->addr, mc_addr))
 635                         break;
 636         }
 637         if (!mc) {
 638                 rcu_read_unlock();
 639                 return true;
 640         }
 641         read_lock(&mc->sflock);
 642         psl = mc->sflist;
 643         if (!psl) {
 644                 rv = mc->sfmode == MCAST_EXCLUDE;
 645         } else {
 646                 int i;
 647
 648                 for (i = 0; i < psl->sl_count; i++) {
 649                         if (ipv6_addr_equal(&psl->sl_addr[i], src_addr))
 650                                 break;
 651                 }
 652                 if (mc->sfmode == MCAST_INCLUDE && i >= psl->sl_count)
 653                         rv = false;
 654                 if (mc->sfmode == MCAST_EXCLUDE && i < psl->sl_count)
 655                         rv = false;
 656         }
 657         read_unlock(&mc->sflock);
 658         rcu_read_unlock();
 659
 660         return rv;
 661 }
 662
 663 static void igmp6_group_added(struct ifmcaddr6 *mc, unsigned int mode)
 664 {
 665         struct net_device *dev = mc->idev->dev;
 666         char buf[MAX_ADDR_LEN];
 667
 668         if (IPV6_ADDR_MC_SCOPE(&mc->mca_addr) <
 669             IPV6_ADDR_SCOPE_LINKLOCAL)
 670                 return;
 671
 672         spin_lock_bh(&mc->mca_lock);
 673         if (!(mc->mca_flags&MAF_LOADED)) {
 674                 mc->mca_flags |= MAF_LOADED;
 675                 if (ndisc_mc_map(&mc->mca_addr, buf, dev, 0) == 0)
 676                         dev_mc_add(dev, buf);
 677         }
 678         spin_unlock_bh(&mc->mca_lock);
 679
 680         if (!(dev->flags & IFF_UP) || (mc->mca_flags & MAF_NOREPORT))
 681                 return;
 682
 683         if (mld_in_v1_mode(mc->idev)) {
 684                 igmp6_join_group(mc);
 685                 return;
 686         }
 687         /* else v2 */
 688
 689         /* Based on RFC3810 6.1, for newly added INCLUDE SSM, we
 690          * should not send filter-mode change record as the mode
 691          * should be from IN() to IN(A).
 692          */
 693         if (mode == MCAST_EXCLUDE)
 694                 mc->mca_crcount = mc->idev->mc_qrv;
 695
 696         mld_ifc_event(mc->idev);
 697 }
 698
 699 static void igmp6_group_dropped(struct ifmcaddr6 *mc)
 700 {
 701         struct net_device *dev = mc->idev->dev;
 702         char buf[MAX_ADDR_LEN];
 703
 704         if (IPV6_ADDR_MC_SCOPE(&mc->mca_addr) <
 705             IPV6_ADDR_SCOPE_LINKLOCAL)
 706                 return;
 707
 708         spin_lock_bh(&mc->mca_lock);
 709         if (mc->mca_flags&MAF_LOADED) {
 710                 mc->mca_flags &= ~MAF_LOADED;
 711                 if (ndisc_mc_map(&mc->mca_addr, buf, dev, 0) == 0)
 712                         dev_mc_del(dev, buf);
 713         }
 714
 715         spin_unlock_bh(&mc->mca_lock);
 716         if (mc->mca_flags & MAF_NOREPORT)
 717                 return;
 718
 719         if (!mc->idev->dead)
 720                 igmp6_leave_group(mc);
 721
 722         spin_lock_bh(&mc->mca_lock);
 723         if (del_timer(&mc->mca_timer))
 724                 refcount_dec(&mc->mca_refcnt);
 725         spin_unlock_bh(&mc->mca_lock);
 726 }
 727
 728 /*
 729  * deleted ifmcaddr6 manipulation
 730  */
 731 static void mld_add_delrec(struct inet6_dev *idev, struct ifmcaddr6 *im)
 732 {
 733         struct ifmcaddr6 *pmc;
 734
 735         /* this is an "ifmcaddr6" for convenience; only the fields below
 736          * are actually used. In particular, the refcnt and users are not
 737          * used for management of the delete list. Using the same structure
 738          * for deleted items allows change reports to use common code with
 739          * non-deleted or query-response MCA's.
 740          */
 741         pmc = kzalloc(sizeof(*pmc), GFP_ATOMIC);
 742         if (!pmc)
 743                 return;
 744
 745         spin_lock_bh(&im->mca_lock);
 746         spin_lock_init(&pmc->mca_lock);
 747         pmc->idev = im->idev;
 748         in6_dev_hold(idev);
 749         pmc->mca_addr = im->mca_addr;
 750         pmc->mca_crcount = idev->mc_qrv;
 751         pmc->mca_sfmode = im->mca_sfmode;
 752         if (pmc->mca_sfmode == MCAST_INCLUDE) {
 753                 struct ip6_sf_list *psf;
 754
 755                 pmc->mca_tomb = im->mca_tomb;
 756                 pmc->mca_sources = im->mca_sources;
 757                 im->mca_tomb = im->mca_sources = NULL;
 758                 for (psf = pmc->mca_sources; psf; psf = psf->sf_next)
 759                         psf->sf_crcount = pmc->mca_crcount;
 760         }
 761         spin_unlock_bh(&im->mca_lock);
 762
 763         spin_lock_bh(&idev->mc_lock);
 764         pmc->next = idev->mc_tomb;
 765         idev->mc_tomb = pmc;
 766         spin_unlock_bh(&idev->mc_lock);
 767 }
 768
 769 static void mld_del_delrec(struct inet6_dev *idev, struct ifmcaddr6 *im)
 770 {
 771         struct ifmcaddr6 *pmc, *pmc_prev;
 772         struct ip6_sf_list *psf;
 773         struct in6_addr *pmca = &im->mca_addr;
 774
 775         spin_lock_bh(&idev->mc_lock);
 776         pmc_prev = NULL;
 777         for (pmc = idev->mc_tomb; pmc; pmc = pmc->next) {
 778                 if (ipv6_addr_equal(&pmc->mca_addr, pmca))
 779                         break;
 780                 pmc_prev = pmc;
 781         }
 782         if (pmc) {
 783                 if (pmc_prev)
 784                         pmc_prev->next = pmc->next;
 785                 else
 786                         idev->mc_tomb = pmc->next;
 787         }
 788         spin_unlock_bh(&idev->mc_lock);
 789
 790         spin_lock_bh(&im->mca_lock);
 791         if (pmc) {
 792                 im->idev = pmc->idev;
 793                 if (im->mca_sfmode == MCAST_INCLUDE) {
 794                         im->mca_tomb = pmc->mca_tomb;
 795                         im->mca_sources = pmc->mca_sources;
 796                         for (psf = im->mca_sources; psf; psf = psf->sf_next)
 797                                 psf->sf_crcount = idev->mc_qrv;
 798                 } else {
 799                         im->mca_crcount = idev->mc_qrv;
 800                 }
 801                 in6_dev_put(pmc->idev);
 802                 kfree(pmc);
 803         }
 804         spin_unlock_bh(&im->mca_lock);
 805 }
 806
 807 static void mld_clear_delrec(struct inet6_dev *idev)
 808 {
 809         struct ifmcaddr6 *pmc, *nextpmc;
 810
 811         spin_lock_bh(&idev->mc_lock);
 812         pmc = idev->mc_tomb;
 813         idev->mc_tomb = NULL;
 814         spin_unlock_bh(&idev->mc_lock);
 815
 816         for (; pmc; pmc = nextpmc) {
 817                 nextpmc = pmc->next;
 818                 ip6_mc_clear_src(pmc);
 819                 in6_dev_put(pmc->idev);
 820                 kfree(pmc);
 821         }
 822
 823         /* clear dead sources, too */
 824         read_lock_bh(&idev->lock);
 825         for (pmc = idev->mc_list; pmc; pmc = pmc->next) {
 826                 struct ip6_sf_list *psf, *psf_next;
 827
 828                 spin_lock_bh(&pmc->mca_lock);
 829                 psf = pmc->mca_tomb;
 830                 pmc->mca_tomb = NULL;
 831                 spin_unlock_bh(&pmc->mca_lock);
 832                 for (; psf; psf = psf_next) {
 833                         psf_next = psf->sf_next;
 834                         kfree(psf);
 835                 }
 836         }
 837         read_unlock_bh(&idev->lock);
 838 }
 839
 840 static void mca_get(struct ifmcaddr6 *mc)
 841 {
 842         refcount_inc(&mc->mca_refcnt);
 843 }
 844
 845 static void ma_put(struct ifmcaddr6 *mc)
 846 {
 847         if (refcount_dec_and_test(&mc->mca_refcnt)) {
 848                 in6_dev_put(mc->idev);
 849                 kfree(mc);
 850         }
 851 }
 852
 853 static struct ifmcaddr6 *mca_alloc(struct inet6_dev *idev,
 854                                    const struct in6_addr *addr,
 855                                    unsigned int mode)
 856 {
 857         struct ifmcaddr6 *mc;
 858
 859         mc = kzalloc(sizeof(*mc), GFP_ATOMIC);
 860         if (!mc)
 861                 return NULL;
 862
 863         timer_setup(&mc->mca_timer, igmp6_timer_handler, 0);
 864
 865         mc->mca_addr = *addr;
 866         mc->idev = idev; /* reference taken by caller */
 867         mc->mca_users = 1;
 868         /* mca_stamp should be updated upon changes */
 869         mc->mca_cstamp = mc->mca_tstamp = jiffies;
 870         refcount_set(&mc->mca_refcnt, 1);
 871         spin_lock_init(&mc->mca_lock);
 872
 873         mc->mca_sfmode = mode;
 874         mc->mca_sfcount[mode] = 1;
 875
 876         if (ipv6_addr_is_ll_all_nodes(&mc->mca_addr) ||
 877             IPV6_ADDR_MC_SCOPE(&mc->mca_addr) < IPV6_ADDR_SCOPE_LINKLOCAL)
 878                 mc->mca_flags |= MAF_NOREPORT;
 879
 880         return mc;
 881 }
 882
 883 /*
 884  *      device multicast group inc (add if not found)
 885  */
 886 static int __ipv6_dev_mc_inc(struct net_device *dev,
 887                              const struct in6_addr *addr, unsigned int mode)
 888 {
 889         struct ifmcaddr6 *mc;
 890         struct inet6_dev *idev;
 891
 892         ASSERT_RTNL();
 893
 894         /* we need to take a reference on idev */
 895         idev = in6_dev_get(dev);
 896
 897         if (!idev)
 898                 return -EINVAL;
 899
 900         write_lock_bh(&idev->lock);
 901         if (idev->dead) {
 902                 write_unlock_bh(&idev->lock);
 903                 in6_dev_put(idev);
 904                 return -ENODEV;
 905         }
 906
 907         for (mc = idev->mc_list; mc; mc = mc->next) {
 908                 if (ipv6_addr_equal(&mc->mca_addr, addr)) {
 909                         mc->mca_users++;
 910                         write_unlock_bh(&idev->lock);
 911                         ip6_mc_add_src(idev, &mc->mca_addr, mode, 0, NULL, 0);
 912                         in6_dev_put(idev);
 913                         return 0;
 914                 }
 915         }
 916
 917         mc = mca_alloc(idev, addr, mode);
 918         if (!mc) {
 919                 write_unlock_bh(&idev->lock);
 920                 in6_dev_put(idev);
 921                 return -ENOMEM;
 922         }
 923
 924         mc->next = idev->mc_list;
 925         idev->mc_list = mc;
 926
 927         /* Hold this for the code below before we unlock,
 928          * it is already exposed via idev->mc_list.
 929          */
 930         mca_get(mc);
 931         write_unlock_bh(&idev->lock);
 932
 933         mld_del_delrec(idev, mc);
 934         igmp6_group_added(mc, mode);
 935         ma_put(mc);
 936         return 0;
 937 }
 938
 939 int ipv6_dev_mc_inc(struct net_device *dev, const struct in6_addr *addr)
 940 {
 941         return __ipv6_dev_mc_inc(dev, addr, MCAST_EXCLUDE);
 942 }
 943
 944 /*
 945  *      device multicast group del
 946  */
 947 int __ipv6_dev_mc_dec(struct inet6_dev *idev, const struct in6_addr *addr)
 948 {
 949         struct ifmcaddr6 *ma, **map;
 950
 951         ASSERT_RTNL();
 952
 953         write_lock_bh(&idev->lock);
 954         for (map = &idev->mc_list; (ma = *map) != NULL; map = &ma->next) {
 955                 if (ipv6_addr_equal(&ma->mca_addr, addr)) {
 956                         if (--ma->mca_users == 0) {
 957                                 *map = ma->next;
 958                                 write_unlock_bh(&idev->lock);
 959
 960                                 igmp6_group_dropped(ma);
 961                                 ip6_mc_clear_src(ma);
 962
 963                                 ma_put(ma);
 964                                 return 0;
 965                         }
 966                         write_unlock_bh(&idev->lock);
 967                         return 0;
 968                 }
 969         }
 970         write_unlock_bh(&idev->lock);
 971
 972         return -ENOENT;
 973 }
 974
 975 int ipv6_dev_mc_dec(struct net_device *dev, const struct in6_addr *addr)
 976 {
 977         struct inet6_dev *idev;
 978         int err;
 979
 980         ASSERT_RTNL();
 981
 982         idev = __in6_dev_get(dev);
 983         if (!idev)
 984                 err = -ENODEV;
 985         else
 986                 err = __ipv6_dev_mc_dec(idev, addr);
 987
 988         return err;
 989 }
 990
 991 /*
 992  *      check if the interface/address pair is valid
 993  */
 994 bool ipv6_chk_mcast_addr(struct net_device *dev, const struct in6_addr *group,
 995                          const struct in6_addr *src_addr)
 996 {
 997         struct inet6_dev *idev;
 998         struct ifmcaddr6 *mc;
 999         bool rv = false;
1000
1001         rcu_read_lock();
1002         idev = __in6_dev_get(dev);
1003         if (idev) {
1004                 read_lock_bh(&idev->lock);
1005                 for (mc = idev->mc_list; mc; mc = mc->next) {
1006                         if (ipv6_addr_equal(&mc->mca_addr, group))
1007                                 break;
1008                 }
1009                 if (mc) {
1010                         if (src_addr && !ipv6_addr_any(src_addr)) {
1011                                 struct ip6_sf_list *psf;
1012
1013                                 spin_lock_bh(&mc->mca_lock);
1014                                 for (psf = mc->mca_sources; psf; psf = psf->sf_next) {
1015                                         if (ipv6_addr_equal(&psf->sf_addr, src_addr))
1016                                                 break;
1017                                 }
1018                                 if (psf)
1019                                         rv = psf->sf_count[MCAST_INCLUDE] ||
1020                                                 psf->sf_count[MCAST_EXCLUDE] !=
1021                                                 mc->mca_sfcount[MCAST_EXCLUDE];
1022                                 else
1023                                         rv = mc->mca_sfcount[MCAST_EXCLUDE] != 0;
1024                                 spin_unlock_bh(&mc->mca_lock);
1025                         } else
1026                                 rv = true; /* don't filter unspecified source */
1027                 }
1028                 read_unlock_bh(&idev->lock);
1029         }
1030         rcu_read_unlock();
1031         return rv;
1032 }
1033
1034 static void mld_gq_start_timer(struct inet6_dev *idev)
1035 {
1036         unsigned long tv = prandom_u32() % idev->mc_maxdelay;
1037
1038         idev->mc_gq_running = 1;
1039         if (!mod_timer(&idev->mc_gq_timer, jiffies+tv+2))
1040                 in6_dev_hold(idev);
1041 }
1042
1043 static void mld_gq_stop_timer(struct inet6_dev *idev)
1044 {
1045         idev->mc_gq_running = 0;
1046         if (del_timer(&idev->mc_gq_timer))
1047                 __in6_dev_put(idev);
1048 }
1049
1050 static void mld_ifc_start_timer(struct inet6_dev *idev, unsigned long delay)
1051 {
1052         unsigned long tv = prandom_u32() % delay;
1053
1054         if (!mod_timer(&idev->mc_ifc_timer, jiffies+tv+2))
1055                 in6_dev_hold(idev);
1056 }
1057
1058 static void mld_ifc_stop_timer(struct inet6_dev *idev)
1059 {
1060         idev->mc_ifc_count = 0;
1061         if (del_timer(&idev->mc_ifc_timer))
1062                 __in6_dev_put(idev);
1063 }
1064
1065 static void mld_dad_start_timer(struct inet6_dev *idev, unsigned long delay)
1066 {
1067         unsigned long tv = prandom_u32() % delay;
1068
1069         if (!mod_timer(&idev->mc_dad_timer, jiffies+tv+2))
1070                 in6_dev_hold(idev);
1071 }
1072
1073 static void mld_dad_stop_timer(struct inet6_dev *idev)
1074 {
1075         if (del_timer(&idev->mc_dad_timer))
1076                 __in6_dev_put(idev);
1077 }
1078
1079 /*
1080  *      IGMP handling (alias multicast ICMPv6 messages)
1081  */
1082
1083 static void igmp6_group_queried(struct ifmcaddr6 *ma, unsigned long resptime)
1084 {
1085         unsigned long delay = resptime;
1086
1087         /* Do not start timer for these addresses */
1088         if (ipv6_addr_is_ll_all_nodes(&ma->mca_addr) ||
1089             IPV6_ADDR_MC_SCOPE(&ma->mca_addr) < IPV6_ADDR_SCOPE_LINKLOCAL)
1090                 return;
1091
1092         if (del_timer(&ma->mca_timer)) {
1093                 refcount_dec(&ma->mca_refcnt);
1094                 delay = ma->mca_timer.expires - jiffies;
1095         }
1096
1097         if (delay >= resptime)
1098                 delay = prandom_u32() % resptime;
1099
1100         ma->mca_timer.expires = jiffies + delay;
1101         if (!mod_timer(&ma->mca_timer, jiffies + delay))
1102                 refcount_inc(&ma->mca_refcnt);
1103         ma->mca_flags |= MAF_TIMER_RUNNING;
1104 }
1105
1106 /* mark EXCLUDE-mode sources */
1107 static bool mld_xmarksources(struct ifmcaddr6 *pmc, int nsrcs,
1108                              const struct in6_addr *srcs)
1109 {
1110         struct ip6_sf_list *psf;
1111         int i, scount;
1112
1113         scount = 0;
1114         for (psf = pmc->mca_sources; psf; psf = psf->sf_next) {
1115                 if (scount == nsrcs)
1116                         break;
1117                 for (i = 0; i < nsrcs; i++) {
1118                         /* skip inactive filters */
1119                         if (psf->sf_count[MCAST_INCLUDE] ||
1120                             pmc->mca_sfcount[MCAST_EXCLUDE] !=
1121                             psf->sf_count[MCAST_EXCLUDE])
1122                                 break;
1123                         if (ipv6_addr_equal(&srcs[i], &psf->sf_addr)) {
1124                                 scount++;
1125                                 break;
1126                         }
1127                 }
1128         }
1129         pmc->mca_flags &= ~MAF_GSQUERY;
1130         if (scount == nsrcs)    /* all sources excluded */
1131                 return false;
1132         return true;
1133 }
1134
1135 static bool mld_marksources(struct ifmcaddr6 *pmc, int nsrcs,
1136                             const struct in6_addr *srcs)
1137 {
1138         struct ip6_sf_list *psf;
1139         int i, scount;
1140
1141         if (pmc->mca_sfmode == MCAST_EXCLUDE)
1142                 return mld_xmarksources(pmc, nsrcs, srcs);
1143
1144         /* mark INCLUDE-mode sources */
1145
1146         scount = 0;
1147         for (psf = pmc->mca_sources; psf; psf = psf->sf_next) {
1148                 if (scount == nsrcs)
1149                         break;
1150                 for (i = 0; i < nsrcs; i++) {
1151                         if (ipv6_addr_equal(&srcs[i], &psf->sf_addr)) {
1152                                 psf->sf_gsresp = 1;
1153                                 scount++;
1154                                 break;
1155                         }
1156                 }
1157         }
1158         if (!scount) {
1159                 pmc->mca_flags &= ~MAF_GSQUERY;
1160                 return false;
1161         }
1162         pmc->mca_flags |= MAF_GSQUERY;
1163         return true;
1164 }
1165
1166 static int mld_force_mld_version(const struct inet6_dev *idev)
1167 {
1168         /* Normally, both are 0 here. If enforcement to a particular is
1169          * being used, individual device enforcement will have a lower
1170          * precedence over 'all' device (.../conf/all/force_mld_version).
1171          */
1172
1173         if (dev_net(idev->dev)->ipv6.devconf_all->force_mld_version != 0)
1174                 return dev_net(idev->dev)->ipv6.devconf_all->force_mld_version;
1175         else
1176                 return idev->cnf.force_mld_version;
1177 }
1178
1179 static bool mld_in_v2_mode_only(const struct inet6_dev *idev)
1180 {
1181         return mld_force_mld_version(idev) == 2;
1182 }
1183
1184 static bool mld_in_v1_mode_only(const struct inet6_dev *idev)
1185 {
1186         return mld_force_mld_version(idev) == 1;
1187 }
1188
1189 static bool mld_in_v1_mode(const struct inet6_dev *idev)
1190 {
1191         if (mld_in_v2_mode_only(idev))
1192                 return false;
1193         if (mld_in_v1_mode_only(idev))
1194                 return true;
1195         if (idev->mc_v1_seen && time_before(jiffies, idev->mc_v1_seen))
1196                 return true;
1197
1198         return false;
1199 }
1200
1201 static void mld_set_v1_mode(struct inet6_dev *idev)
1202 {
1203         /* RFC3810, relevant sections:
1204          *  - 9.1. Robustness Variable
1205          *  - 9.2. Query Interval
1206          *  - 9.3. Query Response Interval
1207          *  - 9.12. Older Version Querier Present Timeout
1208          */
1209         unsigned long switchback;
1210
1211         switchback = (idev->mc_qrv * idev->mc_qi) + idev->mc_qri;
1212
1213         idev->mc_v1_seen = jiffies + switchback;
1214 }
1215
1216 static void mld_update_qrv(struct inet6_dev *idev,
1217                            const struct mld2_query *mlh2)
1218 {
1219         /* RFC3810, relevant sections:
1220          *  - 5.1.8. QRV (Querier's Robustness Variable)
1221          *  - 9.1. Robustness Variable
1222          */
1223
1224         /* The value of the Robustness Variable MUST NOT be zero,
1225          * and SHOULD NOT be one. Catch this here if we ever run
1226          * into such a case in future.
1227          */
1228         const int min_qrv = min(MLD_QRV_DEFAULT, sysctl_mld_qrv);
1229         WARN_ON(idev->mc_qrv == 0);
1230
1231         if (mlh2->mld2q_qrv > 0)
1232                 idev->mc_qrv = mlh2->mld2q_qrv;
1233
1234         if (unlikely(idev->mc_qrv < min_qrv)) {
1235                 net_warn_ratelimited("IPv6: MLD: clamping QRV from %u to %u!\n",
1236                                      idev->mc_qrv, min_qrv);
1237                 idev->mc_qrv = min_qrv;
1238         }
1239 }
1240
1241 static void mld_update_qi(struct inet6_dev *idev,
1242                           const struct mld2_query *mlh2)
1243 {
1244         /* RFC3810, relevant sections:
1245          *  - 5.1.9. QQIC (Querier's Query Interval Code)
1246          *  - 9.2. Query Interval
1247          *  - 9.12. Older Version Querier Present Timeout
1248          *    (the [Query Interval] in the last Query received)
1249          */
1250         unsigned long mc_qqi;
1251
1252         if (mlh2->mld2q_qqic < 128) {
1253                 mc_qqi = mlh2->mld2q_qqic;
1254         } else {
1255                 unsigned long mc_man, mc_exp;
1256
1257                 mc_exp = MLDV2_QQIC_EXP(mlh2->mld2q_qqic);
1258                 mc_man = MLDV2_QQIC_MAN(mlh2->mld2q_qqic);
1259
1260                 mc_qqi = (mc_man | 0x10) << (mc_exp + 3);
1261         }
1262
1263         idev->mc_qi = mc_qqi * HZ;
1264 }
1265
1266 static void mld_update_qri(struct inet6_dev *idev,
1267                            const struct mld2_query *mlh2)
1268 {
1269         /* RFC3810, relevant sections:
1270          *  - 5.1.3. Maximum Response Code
1271          *  - 9.3. Query Response Interval
1272          */
1273         idev->mc_qri = msecs_to_jiffies(mldv2_mrc(mlh2));
1274 }
1275
1276 static int mld_process_v1(struct inet6_dev *idev, struct mld_msg *mld,
1277                           unsigned long *max_delay, bool v1_query)
1278 {
1279         unsigned long mldv1_md;
1280
1281         /* Ignore v1 queries */
1282         if (mld_in_v2_mode_only(idev))
1283                 return -EINVAL;
1284
1285         mldv1_md = ntohs(mld->mld_maxdelay);
1286
1287         /* When in MLDv1 fallback and a MLDv2 router start-up being
1288          * unaware of current MLDv1 operation, the MRC == MRD mapping
1289          * only works when the exponential algorithm is not being
1290          * used (as MLDv1 is unaware of such things).
1291          *
1292          * According to the RFC author, the MLDv2 implementations
1293          * he's aware of all use a MRC < 32768 on start up queries.
1294          *
1295          * Thus, should we *ever* encounter something else larger
1296          * than that, just assume the maximum possible within our
1297          * reach.
1298          */
1299         if (!v1_query)
1300                 mldv1_md = min(mldv1_md, MLDV1_MRD_MAX_COMPAT);
1301
1302         *max_delay = max(msecs_to_jiffies(mldv1_md), 1UL);
1303
1304         /* MLDv1 router present: we need to go into v1 mode *only*
1305          * when an MLDv1 query is received as per section 9.12. of
1306          * RFC3810! And we know from RFC2710 section 3.7 that MLDv1
1307          * queries MUST be of exactly 24 octets.
1308          */
1309         if (v1_query)
1310                 mld_set_v1_mode(idev);
1311
1312         /* cancel MLDv2 report timer */
1313         mld_gq_stop_timer(idev);
1314         /* cancel the interface change timer */
1315         mld_ifc_stop_timer(idev);
1316         /* clear deleted report items */
1317         mld_clear_delrec(idev);
1318
1319         return 0;
1320 }
1321
1322 static int mld_process_v2(struct inet6_dev *idev, struct mld2_query *mld,
1323                           unsigned long *max_delay)
1324 {
1325         *max_delay = max(msecs_to_jiffies(mldv2_mrc(mld)), 1UL);
1326
1327         mld_update_qrv(idev, mld);
1328         mld_update_qi(idev, mld);
1329         mld_update_qri(idev, mld);
1330
1331         idev->mc_maxdelay = *max_delay;
1332
1333         return 0;
1334 }
1335
1336 /* called with rcu_read_lock() */
1337 int igmp6_event_query(struct sk_buff *skb)
1338 {
1339         struct mld2_query *mlh2 = NULL;
1340         struct ifmcaddr6 *ma;
1341         const struct in6_addr *group;
1342         unsigned long max_delay;
1343         struct inet6_dev *idev;
1344         struct mld_msg *mld;
1345         int group_type;
1346         int mark = 0;
1347         int len, err;
1348
1349         if (!pskb_may_pull(skb, sizeof(struct in6_addr)))
1350                 return -EINVAL;
1351
1352         /* compute payload length excluding extension headers */
1353         len = ntohs(ipv6_hdr(skb)->payload_len) + sizeof(struct ipv6hdr);
1354         len -= skb_network_header_len(skb);
1355
1356         /* RFC3810 6.2
1357          * Upon reception of an MLD message that contains a Query, the node
1358          * checks if the source address of the message is a valid link-local
1359          * address, if the Hop Limit is set to 1, and if the Router Alert
1360          * option is present in the Hop-By-Hop Options header of the IPv6
1361          * packet.  If any of these checks fails, the packet is dropped.
1362          */
1363         if (!(ipv6_addr_type(&ipv6_hdr(skb)->saddr) & IPV6_ADDR_LINKLOCAL) ||
1364             ipv6_hdr(skb)->hop_limit != 1 ||
1365             !(IP6CB(skb)->flags & IP6SKB_ROUTERALERT) ||
1366             IP6CB(skb)->ra != htons(IPV6_OPT_ROUTERALERT_MLD))
1367                 return -EINVAL;
1368
1369         idev = __in6_dev_get(skb->dev);
1370         if (!idev)
1371                 return 0;
1372
1373         mld = (struct mld_msg *)icmp6_hdr(skb);
1374         group = &mld->mld_mca;
1375         group_type = ipv6_addr_type(group);
1376
1377         if (group_type != IPV6_ADDR_ANY &&
1378             !(group_type&IPV6_ADDR_MULTICAST))
1379                 return -EINVAL;
1380
1381         if (len < MLD_V1_QUERY_LEN) {
1382                 return -EINVAL;
1383         } else if (len == MLD_V1_QUERY_LEN || mld_in_v1_mode(idev)) {
1384                 err = mld_process_v1(idev, mld, &max_delay,
1385                                      len == MLD_V1_QUERY_LEN);
1386                 if (err < 0)
1387                         return err;
1388         } else if (len >= MLD_V2_QUERY_LEN_MIN) {
1389                 int srcs_offset = sizeof(struct mld2_query) -
1390                                   sizeof(struct icmp6hdr);
1391
1392                 if (!pskb_may_pull(skb, srcs_offset))
1393                         return -EINVAL;
1394
1395                 mlh2 = (struct mld2_query *)skb_transport_header(skb);
1396
1397                 err = mld_process_v2(idev, mlh2, &max_delay);
1398                 if (err < 0)
1399                         return err;
1400
1401                 if (group_type == IPV6_ADDR_ANY) { /* general query */
1402                         if (mlh2->mld2q_nsrcs)
1403                                 return -EINVAL; /* no sources allowed */
1404
1405                         mld_gq_start_timer(idev);
1406                         return 0;
1407                 }
1408                 /* mark sources to include, if group & source-specific */
1409                 if (mlh2->mld2q_nsrcs != 0) {
1410                         if (!pskb_may_pull(skb, srcs_offset +
1411                             ntohs(mlh2->mld2q_nsrcs) * sizeof(struct in6_addr)))
1412                                 return -EINVAL;
1413
1414                         mlh2 = (struct mld2_query *)skb_transport_header(skb);
1415                         mark = 1;
1416                 }
1417         } else {
1418                 return -EINVAL;
1419         }
1420
1421         read_lock_bh(&idev->lock);
1422         if (group_type == IPV6_ADDR_ANY) {
1423                 for (ma = idev->mc_list; ma; ma = ma->next) {
1424                         spin_lock_bh(&ma->mca_lock);
1425                         igmp6_group_queried(ma, max_delay);
1426                         spin_unlock_bh(&ma->mca_lock);
1427                 }
1428         } else {
1429                 for (ma = idev->mc_list; ma; ma = ma->next) {
1430                         if (!ipv6_addr_equal(group, &ma->mca_addr))
1431                                 continue;
1432                         spin_lock_bh(&ma->mca_lock);
1433                         if (ma->mca_flags & MAF_TIMER_RUNNING) {
1434                                 /* gsquery <- gsquery && mark */
1435                                 if (!mark)
1436                                         ma->mca_flags &= ~MAF_GSQUERY;
1437                         } else {
1438                                 /* gsquery <- mark */
1439                                 if (mark)
1440                                         ma->mca_flags |= MAF_GSQUERY;
1441                                 else
1442                                         ma->mca_flags &= ~MAF_GSQUERY;
1443                         }
1444                         if (!(ma->mca_flags & MAF_GSQUERY) ||
1445                             mld_marksources(ma, ntohs(mlh2->mld2q_nsrcs), mlh2->mld2q_srcs))
1446                                 igmp6_group_queried(ma, max_delay);
1447                         spin_unlock_bh(&ma->mca_lock);
1448                         break;
1449                 }
1450         }
1451         read_unlock_bh(&idev->lock);
1452
1453         return 0;
1454 }
1455
1456 /* called with rcu_read_lock() */
1457 int igmp6_event_report(struct sk_buff *skb)
1458 {
1459         struct ifmcaddr6 *ma;
1460         struct inet6_dev *idev;
1461         struct mld_msg *mld;
1462         int addr_type;
1463
1464         /* Our own report looped back. Ignore it. */
1465         if (skb->pkt_type == PACKET_LOOPBACK)
1466                 return 0;
1467
1468         /* send our report if the MC router may not have heard this report */
1469         if (skb->pkt_type != PACKET_MULTICAST &&
1470             skb->pkt_type != PACKET_BROADCAST)
1471                 return 0;
1472
1473         if (!pskb_may_pull(skb, sizeof(*mld) - sizeof(struct icmp6hdr)))
1474                 return -EINVAL;
1475
1476         mld = (struct mld_msg *)icmp6_hdr(skb);
1477
1478         /* Drop reports with not link local source */
1479         addr_type = ipv6_addr_type(&ipv6_hdr(skb)->saddr);
1480         if (addr_type != IPV6_ADDR_ANY &&
1481             !(addr_type&IPV6_ADDR_LINKLOCAL))
1482                 return -EINVAL;
1483
1484         idev = __in6_dev_get(skb->dev);
1485         if (!idev)
1486                 return -ENODEV;
1487
1488         /*
1489          *      Cancel the timer for this group
1490          */
1491
1492         read_lock_bh(&idev->lock);
1493         for (ma = idev->mc_list; ma; ma = ma->next) {
1494                 if (ipv6_addr_equal(&ma->mca_addr, &mld->mld_mca)) {
1495                         spin_lock(&ma->mca_lock);
1496                         if (del_timer(&ma->mca_timer))
1497                                 refcount_dec(&ma->mca_refcnt);
1498                         ma->mca_flags &= ~(MAF_LAST_REPORTER|MAF_TIMER_RUNNING);
1499                         spin_unlock(&ma->mca_lock);
1500                         break;
1501                 }
1502         }
1503         read_unlock_bh(&idev->lock);
1504         return 0;
1505 }
1506
1507 static bool is_in(struct ifmcaddr6 *pmc, struct ip6_sf_list *psf, int type,
1508                   int gdeleted, int sdeleted)
1509 {
1510         switch (type) {
1511         case MLD2_MODE_IS_INCLUDE:
1512         case MLD2_MODE_IS_EXCLUDE:
1513                 if (gdeleted || sdeleted)
1514                         return false;
1515                 if (!((pmc->mca_flags & MAF_GSQUERY) && !psf->sf_gsresp)) {
1516                         if (pmc->mca_sfmode == MCAST_INCLUDE)
1517                                 return true;
1518                         /* don't include if this source is excluded
1519                          * in all filters
1520                          */
1521                         if (psf->sf_count[MCAST_INCLUDE])
1522                                 return type == MLD2_MODE_IS_INCLUDE;
1523                         return pmc->mca_sfcount[MCAST_EXCLUDE] ==
1524                                 psf->sf_count[MCAST_EXCLUDE];
1525                 }
1526                 return false;
1527         case MLD2_CHANGE_TO_INCLUDE:
1528                 if (gdeleted || sdeleted)
1529                         return false;
1530                 return psf->sf_count[MCAST_INCLUDE] != 0;
1531         case MLD2_CHANGE_TO_EXCLUDE:
1532                 if (gdeleted || sdeleted)
1533                         return false;
1534                 if (pmc->mca_sfcount[MCAST_EXCLUDE] == 0 ||
1535                     psf->sf_count[MCAST_INCLUDE])
1536                         return false;
1537                 return pmc->mca_sfcount[MCAST_EXCLUDE] ==
1538                         psf->sf_count[MCAST_EXCLUDE];
1539         case MLD2_ALLOW_NEW_SOURCES:
1540                 if (gdeleted || !psf->sf_crcount)
1541                         return false;
1542                 return (pmc->mca_sfmode == MCAST_INCLUDE) ^ sdeleted;
1543         case MLD2_BLOCK_OLD_SOURCES:
1544                 if (pmc->mca_sfmode == MCAST_INCLUDE)
1545                         return gdeleted || (psf->sf_crcount && sdeleted);
1546                 return psf->sf_crcount && !gdeleted && !sdeleted;
1547         }
1548         return false;
1549 }
1550
1551 static int
1552 mld_scount(struct ifmcaddr6 *pmc, int type, int gdeleted, int sdeleted)
1553 {
1554         struct ip6_sf_list *psf;
1555         int scount = 0;
1556
1557         for (psf = pmc->mca_sources; psf; psf = psf->sf_next) {
1558                 if (!is_in(pmc, psf, type, gdeleted, sdeleted))
1559                         continue;
1560                 scount++;
1561         }
1562         return scount;
1563 }
1564
1565 static void ip6_mc_hdr(struct sock *sk, struct sk_buff *skb,
1566                        struct net_device *dev,
1567                        const struct in6_addr *saddr,
1568                        const struct in6_addr *daddr,
1569                        int proto, int len)
1570 {
1571         struct ipv6hdr *hdr;
1572
1573         skb->protocol = htons(ETH_P_IPV6);
1574         skb->dev = dev;
1575
1576         skb_reset_network_header(skb);
1577         skb_put(skb, sizeof(struct ipv6hdr));
1578         hdr = ipv6_hdr(skb);
1579
1580         ip6_flow_hdr(hdr, 0, 0);
1581
1582         hdr->payload_len = htons(len);
1583         hdr->nexthdr = proto;
1584         hdr->hop_limit = inet6_sk(sk)->hop_limit;
1585
1586         hdr->saddr = *saddr;
1587         hdr->daddr = *daddr;
1588 }
1589
1590 static struct sk_buff *mld_newpack(struct inet6_dev *idev, unsigned int mtu)
1591 {
1592         struct net_device *dev = idev->dev;
1593         struct net *net = dev_net(dev);
1594         struct sock *sk = net->ipv6.igmp_sk;
1595         struct sk_buff *skb;
1596         struct mld2_report *pmr;
1597         struct in6_addr addr_buf;
1598         const struct in6_addr *saddr;
1599         int hlen = LL_RESERVED_SPACE(dev);
1600         int tlen = dev->needed_tailroom;
1601         unsigned int size = mtu + hlen + tlen;
1602         int err;
1603         u8 ra[8] = { IPPROTO_ICMPV6, 0,
1604                      IPV6_TLV_ROUTERALERT, 2, 0, 0,
1605                      IPV6_TLV_PADN, 0 };
1606
1607         /* we assume size > sizeof(ra) here */
1608         /* limit our allocations to order-0 page */
1609         size = min_t(int, size, SKB_MAX_ORDER(0, 0));
1610         skb = sock_alloc_send_skb(sk, size, 1, &err);
1611
1612         if (!skb)
1613                 return NULL;
1614
1615         skb->priority = TC_PRIO_CONTROL;
1616         skb_reserve(skb, hlen);
1617         skb_tailroom_reserve(skb, mtu, tlen);
1618
1619         if (__ipv6_get_lladdr(idev, &addr_buf, IFA_F_TENTATIVE)) {
1620                 /* <draft-ietf-magma-mld-source-05.txt>:
1621                  * use unspecified address as the source address
1622                  * when a valid link-local address is not available.
1623                  */
1624                 saddr = &in6addr_any;
1625         } else
1626                 saddr = &addr_buf;
1627
1628         ip6_mc_hdr(sk, skb, dev, saddr, &mld2_all_mcr, NEXTHDR_HOP, 0);
1629
1630         skb_put_data(skb, ra, sizeof(ra));
1631
1632         skb_set_transport_header(skb, skb_tail_pointer(skb) - skb->data);
1633         skb_put(skb, sizeof(*pmr));
1634         pmr = (struct mld2_report *)skb_transport_header(skb);
1635         pmr->mld2r_type = ICMPV6_MLD2_REPORT;
1636         pmr->mld2r_resv1 = 0;
1637         pmr->mld2r_cksum = 0;
1638         pmr->mld2r_resv2 = 0;
1639         pmr->mld2r_ngrec = 0;
1640         return skb;
1641 }
1642
1643 static void mld_sendpack(struct sk_buff *skb)
1644 {
1645         struct ipv6hdr *pip6 = ipv6_hdr(skb);
1646         struct mld2_report *pmr =
1647                               (struct mld2_report *)skb_transport_header(skb);
1648         int payload_len, mldlen;
1649         struct inet6_dev *idev;
1650         struct net *net = dev_net(skb->dev);
1651         int err;
1652         struct flowi6 fl6;
1653         struct dst_entry *dst;
1654
1655         rcu_read_lock();
1656         idev = __in6_dev_get(skb->dev);
1657         IP6_UPD_PO_STATS(net, idev, IPSTATS_MIB_OUT, skb->len);
1658
1659         payload_len = (skb_tail_pointer(skb) - skb_network_header(skb)) -
1660                 sizeof(*pip6);
1661         mldlen = skb_tail_pointer(skb) - skb_transport_header(skb);
1662         pip6->payload_len = htons(payload_len);
1663
1664         pmr->mld2r_cksum = csum_ipv6_magic(&pip6->saddr, &pip6->daddr, mldlen,
1665                                            IPPROTO_ICMPV6,
1666                                            csum_partial(skb_transport_header(skb),
1667                                                         mldlen, 0));
1668
1669         icmpv6_flow_init(net->ipv6.igmp_sk, &fl6, ICMPV6_MLD2_REPORT,
1670                          &ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr,
1671                          skb->dev->ifindex);
1672         dst = icmp6_dst_alloc(skb->dev, &fl6);
1673
1674         err = 0;
1675         if (IS_ERR(dst)) {
1676                 err = PTR_ERR(dst);
1677                 dst = NULL;
1678         }
1679         skb_dst_set(skb, dst);
1680         if (err)
1681                 goto err_out;
1682
1683         err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT,
1684                       net, net->ipv6.igmp_sk, skb, NULL, skb->dev,
1685                       dst_output);
1686 out:
1687         if (!err) {
1688                 ICMP6MSGOUT_INC_STATS(net, idev, ICMPV6_MLD2_REPORT);
1689                 ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTMSGS);
1690         } else {
1691                 IP6_INC_STATS(net, idev, IPSTATS_MIB_OUTDISCARDS);
1692         }
1693
1694         rcu_read_unlock();
1695         return;
1696
1697 err_out:
1698         kfree_skb(skb);
1699         goto out;
1700 }
1701
1702 static int grec_size(struct ifmcaddr6 *pmc, int type, int gdel, int sdel)
1703 {
1704         return sizeof(struct mld2_grec) + 16 * mld_scount(pmc,type,gdel,sdel);
1705 }
1706
1707 static struct sk_buff *add_grhead(struct sk_buff *skb, struct ifmcaddr6 *pmc,
1708         int type, struct mld2_grec **ppgr, unsigned int mtu)
1709 {
1710         struct mld2_report *pmr;
1711         struct mld2_grec *pgr;
1712
1713         if (!skb) {
1714                 skb = mld_newpack(pmc->idev, mtu);
1715                 if (!skb)
1716                         return NULL;
1717         }
1718         pgr = skb_put(skb, sizeof(struct mld2_grec));
1719         pgr->grec_type = type;
1720         pgr->grec_auxwords = 0;
1721         pgr->grec_nsrcs = 0;
1722         pgr->grec_mca = pmc->mca_addr;  /* structure copy */
1723         pmr = (struct mld2_report *)skb_transport_header(skb);
1724         pmr->mld2r_ngrec = htons(ntohs(pmr->mld2r_ngrec)+1);
1725         *ppgr = pgr;
1726         return skb;
1727 }
1728
1729 #define AVAILABLE(skb)  ((skb) ? skb_availroom(skb) : 0)
1730
1731 static struct sk_buff *add_grec(struct sk_buff *skb, struct ifmcaddr6 *pmc,
1732         int type, int gdeleted, int sdeleted, int crsend)
1733 {
1734         struct inet6_dev *idev = pmc->idev;
1735         struct net_device *dev = idev->dev;
1736         struct mld2_report *pmr;
1737         struct mld2_grec *pgr = NULL;
1738         struct ip6_sf_list *psf, *psf_next, *psf_prev, **psf_list;
1739         int scount, stotal, first, isquery, truncate;
1740         unsigned int mtu;
1741
1742         if (pmc->mca_flags & MAF_NOREPORT)
1743                 return skb;
1744
1745         mtu = READ_ONCE(dev->mtu);
1746         if (mtu < IPV6_MIN_MTU)
1747                 return skb;
1748
1749         isquery = type == MLD2_MODE_IS_INCLUDE ||
1750                   type == MLD2_MODE_IS_EXCLUDE;
1751         truncate = type == MLD2_MODE_IS_EXCLUDE ||
1752                     type == MLD2_CHANGE_TO_EXCLUDE;
1753
1754         stotal = scount = 0;
1755
1756         psf_list = sdeleted ? &pmc->mca_tomb : &pmc->mca_sources;
1757
1758         if (!*psf_list)
1759                 goto empty_source;
1760
1761         pmr = skb ? (struct mld2_report *)skb_transport_header(skb) : NULL;
1762
1763         /* EX and TO_EX get a fresh packet, if needed */
1764         if (truncate) {
1765                 if (pmr && pmr->mld2r_ngrec &&
1766                     AVAILABLE(skb) < grec_size(pmc, type, gdeleted, sdeleted)) {
1767                         if (skb)
1768                                 mld_sendpack(skb);
1769                         skb = mld_newpack(idev, mtu);
1770                 }
1771         }
1772         first = 1;
1773         psf_prev = NULL;
1774         for (psf = *psf_list; psf; psf = psf_next) {
1775                 struct in6_addr *psrc;
1776
1777                 psf_next = psf->sf_next;
1778
1779                 if (!is_in(pmc, psf, type, gdeleted, sdeleted) && !crsend) {
1780                         psf_prev = psf;
1781                         continue;
1782                 }
1783
1784                 /* Based on RFC3810 6.1. Should not send source-list change
1785                  * records when there is a filter mode change.
1786                  */
1787                 if (((gdeleted && pmc->mca_sfmode == MCAST_EXCLUDE) ||
1788                      (!gdeleted && pmc->mca_crcount)) &&
1789                     (type == MLD2_ALLOW_NEW_SOURCES ||
1790                      type == MLD2_BLOCK_OLD_SOURCES) && psf->sf_crcount)
1791                         goto decrease_sf_crcount;
1792
1793                 /* clear marks on query responses */
1794                 if (isquery)
1795                         psf->sf_gsresp = 0;
1796
1797                 if (AVAILABLE(skb) < sizeof(*psrc) +
1798                     first*sizeof(struct mld2_grec)) {
1799                         if (truncate && !first)
1800                                 break;   /* truncate these */
1801                         if (pgr)
1802                                 pgr->grec_nsrcs = htons(scount);
1803                         if (skb)
1804                                 mld_sendpack(skb);
1805                         skb = mld_newpack(idev, mtu);
1806                         first = 1;
1807                         scount = 0;
1808                 }
1809                 if (first) {
1810                         skb = add_grhead(skb, pmc, type, &pgr, mtu);
1811                         first = 0;
1812                 }
1813                 if (!skb)
1814                         return NULL;
1815                 psrc = skb_put(skb, sizeof(*psrc));
1816                 *psrc = psf->sf_addr;
1817                 scount++; stotal++;
1818                 if ((type == MLD2_ALLOW_NEW_SOURCES ||
1819                      type == MLD2_BLOCK_OLD_SOURCES) && psf->sf_crcount) {
1820 decrease_sf_crcount:
1821                         psf->sf_crcount--;
1822                         if ((sdeleted || gdeleted) && psf->sf_crcount == 0) {
1823                                 if (psf_prev)
1824                                         psf_prev->sf_next = psf->sf_next;
1825                                 else
1826                                         *psf_list = psf->sf_next;
1827                                 kfree(psf);
1828                                 continue;
1829                         }
1830                 }
1831                 psf_prev = psf;
1832         }
1833
1834 empty_source:
1835         if (!stotal) {
1836                 if (type == MLD2_ALLOW_NEW_SOURCES ||
1837                     type == MLD2_BLOCK_OLD_SOURCES)
1838                         return skb;
1839                 if (pmc->mca_crcount || isquery || crsend) {
1840                         /* make sure we have room for group header */
1841                         if (skb && AVAILABLE(skb) < sizeof(struct mld2_grec)) {
1842                                 mld_sendpack(skb);
1843                                 skb = NULL; /* add_grhead will get a new one */
1844                         }
1845                         skb = add_grhead(skb, pmc, type, &pgr, mtu);
1846                 }
1847         }
1848         if (pgr)
1849                 pgr->grec_nsrcs = htons(scount);
1850
1851         if (isquery)
1852                 pmc->mca_flags &= ~MAF_GSQUERY; /* clear query state */
1853         return skb;
1854 }
1855
1856 static void mld_send_report(struct inet6_dev *idev, struct ifmcaddr6 *pmc)
1857 {
1858         struct sk_buff *skb = NULL;
1859         int type;
1860
1861         read_lock_bh(&idev->lock);
1862         if (!pmc) {
1863                 for (pmc = idev->mc_list; pmc; pmc = pmc->next) {
1864                         if (pmc->mca_flags & MAF_NOREPORT)
1865                                 continue;
1866                         spin_lock_bh(&pmc->mca_lock);
1867                         if (pmc->mca_sfcount[MCAST_EXCLUDE])
1868                                 type = MLD2_MODE_IS_EXCLUDE;
1869                         else
1870                                 type = MLD2_MODE_IS_INCLUDE;
1871                         skb = add_grec(skb, pmc, type, 0, 0, 0);
1872                         spin_unlock_bh(&pmc->mca_lock);
1873                 }
1874         } else {
1875                 spin_lock_bh(&pmc->mca_lock);
1876                 if (pmc->mca_sfcount[MCAST_EXCLUDE])
1877                         type = MLD2_MODE_IS_EXCLUDE;
1878                 else
1879                         type = MLD2_MODE_IS_INCLUDE;
1880                 skb = add_grec(skb, pmc, type, 0, 0, 0);
1881                 spin_unlock_bh(&pmc->mca_lock);
1882         }
1883         read_unlock_bh(&idev->lock);
1884         if (skb)
1885                 mld_sendpack(skb);
1886 }
1887
1888 /*
1889  * remove zero-count source records from a source filter list
1890  */
1891 static void mld_clear_zeros(struct ip6_sf_list **ppsf)
1892 {
1893         struct ip6_sf_list *psf_prev, *psf_next, *psf;
1894
1895         psf_prev = NULL;
1896         for (psf = *ppsf; psf; psf = psf_next) {
1897                 psf_next = psf->sf_next;
1898                 if (psf->sf_crcount == 0) {
1899                         if (psf_prev)
1900                                 psf_prev->sf_next = psf->sf_next;
1901                         else
1902                                 *ppsf = psf->sf_next;
1903                         kfree(psf);
1904                 } else
1905                         psf_prev = psf;
1906         }
1907 }
1908
1909 static void mld_send_cr(struct inet6_dev *idev)
1910 {
1911         struct ifmcaddr6 *pmc, *pmc_prev, *pmc_next;
1912         struct sk_buff *skb = NULL;
1913         int type, dtype;
1914
1915         read_lock_bh(&idev->lock);
1916         spin_lock(&idev->mc_lock);
1917
1918         /* deleted MCA's */
1919         pmc_prev = NULL;
1920         for (pmc = idev->mc_tomb; pmc; pmc = pmc_next) {
1921                 pmc_next = pmc->next;
1922                 if (pmc->mca_sfmode == MCAST_INCLUDE) {
1923                         type = MLD2_BLOCK_OLD_SOURCES;
1924                         dtype = MLD2_BLOCK_OLD_SOURCES;
1925                         skb = add_grec(skb, pmc, type, 1, 0, 0);
1926                         skb = add_grec(skb, pmc, dtype, 1, 1, 0);
1927                 }
1928                 if (pmc->mca_crcount) {
1929                         if (pmc->mca_sfmode == MCAST_EXCLUDE) {
1930                                 type = MLD2_CHANGE_TO_INCLUDE;
1931                                 skb = add_grec(skb, pmc, type, 1, 0, 0);
1932                         }
1933                         pmc->mca_crcount--;
1934                         if (pmc->mca_crcount == 0) {
1935                                 mld_clear_zeros(&pmc->mca_tomb);
1936                                 mld_clear_zeros(&pmc->mca_sources);
1937                         }
1938                 }
1939                 if (pmc->mca_crcount == 0 && !pmc->mca_tomb &&
1940                     !pmc->mca_sources) {
1941                         if (pmc_prev)
1942                                 pmc_prev->next = pmc_next;
1943                         else
1944                                 idev->mc_tomb = pmc_next;
1945                         in6_dev_put(pmc->idev);
1946                         kfree(pmc);
1947                 } else
1948                         pmc_prev = pmc;
1949         }
1950         spin_unlock(&idev->mc_lock);
1951
1952         /* change recs */
1953         for (pmc = idev->mc_list; pmc; pmc = pmc->next) {
1954                 spin_lock_bh(&pmc->mca_lock);
1955                 if (pmc->mca_sfcount[MCAST_EXCLUDE]) {
1956                         type = MLD2_BLOCK_OLD_SOURCES;
1957                         dtype = MLD2_ALLOW_NEW_SOURCES;
1958                 } else {
1959                         type = MLD2_ALLOW_NEW_SOURCES;
1960                         dtype = MLD2_BLOCK_OLD_SOURCES;
1961                 }
1962                 skb = add_grec(skb, pmc, type, 0, 0, 0);
1963                 skb = add_grec(skb, pmc, dtype, 0, 1, 0);       /* deleted sources */
1964
1965                 /* filter mode changes */
1966                 if (pmc->mca_crcount) {
1967                         if (pmc->mca_sfmode == MCAST_EXCLUDE)
1968                                 type = MLD2_CHANGE_TO_EXCLUDE;
1969                         else
1970                                 type = MLD2_CHANGE_TO_INCLUDE;
1971                         skb = add_grec(skb, pmc, type, 0, 0, 0);
1972                         pmc->mca_crcount--;
1973                 }
1974                 spin_unlock_bh(&pmc->mca_lock);
1975         }
1976         read_unlock_bh(&idev->lock);
1977         if (!skb)
1978                 return;
1979         (void) mld_sendpack(skb);
1980 }
1981
1982 static void igmp6_send(struct in6_addr *addr, struct net_device *dev, int type)
1983 {
1984         struct net *net = dev_net(dev);
1985         struct sock *sk = net->ipv6.igmp_sk;
1986         struct inet6_dev *idev;
1987         struct sk_buff *skb;
1988         struct mld_msg *hdr;
1989         const struct in6_addr *snd_addr, *saddr;
1990         struct in6_addr addr_buf;
1991         int hlen = LL_RESERVED_SPACE(dev);
1992         int tlen = dev->needed_tailroom;
1993         int err, len, payload_len, full_len;
1994         u8 ra[8] = { IPPROTO_ICMPV6, 0,
1995                      IPV6_TLV_ROUTERALERT, 2, 0, 0,
1996                      IPV6_TLV_PADN, 0 };
1997         struct flowi6 fl6;
1998         struct dst_entry *dst;
1999
2000         if (type == ICMPV6_MGM_REDUCTION)
2001                 snd_addr = &in6addr_linklocal_allrouters;
2002         else
2003                 snd_addr = addr;
2004
2005         len = sizeof(struct icmp6hdr) + sizeof(struct in6_addr);
2006         payload_len = len + sizeof(ra);
2007         full_len = sizeof(struct ipv6hdr) + payload_len;
2008
2009         rcu_read_lock();
2010         IP6_UPD_PO_STATS(net, __in6_dev_get(dev),
2011                       IPSTATS_MIB_OUT, full_len);
2012         rcu_read_unlock();
2013
2014         skb = sock_alloc_send_skb(sk, hlen + tlen + full_len, 1, &err);
2015
2016         if (!skb) {
2017                 rcu_read_lock();
2018                 IP6_INC_STATS(net, __in6_dev_get(dev),
2019                               IPSTATS_MIB_OUTDISCARDS);
2020                 rcu_read_unlock();
2021                 return;
2022         }
2023         skb->priority = TC_PRIO_CONTROL;
2024         skb_reserve(skb, hlen);
2025
2026         if (ipv6_get_lladdr(dev, &addr_buf, IFA_F_TENTATIVE)) {
2027                 /* <draft-ietf-magma-mld-source-05.txt>:
2028                  * use unspecified address as the source address
2029                  * when a valid link-local address is not available.
2030                  */
2031                 saddr = &in6addr_any;
2032         } else
2033                 saddr = &addr_buf;
2034
2035         ip6_mc_hdr(sk, skb, dev, saddr, snd_addr, NEXTHDR_HOP, payload_len);
2036
2037         skb_put_data(skb, ra, sizeof(ra));
2038
2039         hdr = skb_put_zero(skb, sizeof(struct mld_msg));
2040         hdr->mld_type = type;
2041         hdr->mld_mca = *addr;
2042
2043         hdr->mld_cksum = csum_ipv6_magic(saddr, snd_addr, len,
2044                                          IPPROTO_ICMPV6,
2045                                          csum_partial(hdr, len, 0));
2046
2047         rcu_read_lock();
2048         idev = __in6_dev_get(skb->dev);
2049
2050         icmpv6_flow_init(sk, &fl6, type,
2051                          &ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr,
2052                          skb->dev->ifindex);
2053         dst = icmp6_dst_alloc(skb->dev, &fl6);
2054         if (IS_ERR(dst)) {
2055                 err = PTR_ERR(dst);
2056                 goto err_out;
2057         }
2058
2059         skb_dst_set(skb, dst);
2060         err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT,
2061                       net, sk, skb, NULL, skb->dev,
2062                       dst_output);
2063 out:
2064         if (!err) {
2065                 ICMP6MSGOUT_INC_STATS(net, idev, type);
2066                 ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTMSGS);
2067         } else
2068                 IP6_INC_STATS(net, idev, IPSTATS_MIB_OUTDISCARDS);
2069
2070         rcu_read_unlock();
2071         return;
2072
2073 err_out:
2074         kfree_skb(skb);
2075         goto out;
2076 }
2077
2078 static void mld_send_initial_cr(struct inet6_dev *idev)
2079 {
2080         struct sk_buff *skb;
2081         struct ifmcaddr6 *pmc;
2082         int type;
2083
2084         if (mld_in_v1_mode(idev))
2085                 return;
2086
2087         skb = NULL;
2088         read_lock_bh(&idev->lock);
2089         for (pmc = idev->mc_list; pmc; pmc = pmc->next) {
2090                 spin_lock_bh(&pmc->mca_lock);
2091                 if (pmc->mca_sfcount[MCAST_EXCLUDE])
2092                         type = MLD2_CHANGE_TO_EXCLUDE;
2093                 else
2094                         type = MLD2_ALLOW_NEW_SOURCES;
2095                 skb = add_grec(skb, pmc, type, 0, 0, 1);
2096                 spin_unlock_bh(&pmc->mca_lock);
2097         }
2098         read_unlock_bh(&idev->lock);
2099         if (skb)
2100                 mld_sendpack(skb);
2101 }
2102
2103 void ipv6_mc_dad_complete(struct inet6_dev *idev)
2104 {
2105         idev->mc_dad_count = idev->mc_qrv;
2106         if (idev->mc_dad_count) {
2107                 mld_send_initial_cr(idev);
2108                 idev->mc_dad_count--;
2109                 if (idev->mc_dad_count)
2110                         mld_dad_start_timer(idev,
2111                                             unsolicited_report_interval(idev));
2112         }
2113 }
2114
2115 static void mld_dad_timer_expire(struct timer_list *t)
2116 {
2117         struct inet6_dev *idev = from_timer(idev, t, mc_dad_timer);
2118
2119         mld_send_initial_cr(idev);
2120         if (idev->mc_dad_count) {
2121                 idev->mc_dad_count--;
2122                 if (idev->mc_dad_count)
2123                         mld_dad_start_timer(idev,
2124                                             unsolicited_report_interval(idev));
2125         }
2126         in6_dev_put(idev);
2127 }
2128
2129 static int ip6_mc_del1_src(struct ifmcaddr6 *pmc, int sfmode,
2130         const struct in6_addr *psfsrc)
2131 {
2132         struct ip6_sf_list *psf, *psf_prev;
2133         int rv = 0;
2134
2135         psf_prev = NULL;
2136         for (psf = pmc->mca_sources; psf; psf = psf->sf_next) {
2137                 if (ipv6_addr_equal(&psf->sf_addr, psfsrc))
2138                         break;
2139                 psf_prev = psf;
2140         }
2141         if (!psf || psf->sf_count[sfmode] == 0) {
2142                 /* source filter not found, or count wrong =>  bug */
2143                 return -ESRCH;
2144         }
2145         psf->sf_count[sfmode]--;
2146         if (!psf->sf_count[MCAST_INCLUDE] && !psf->sf_count[MCAST_EXCLUDE]) {
2147                 struct inet6_dev *idev = pmc->idev;
2148
2149                 /* no more filters for this source */
2150                 if (psf_prev)
2151                         psf_prev->sf_next = psf->sf_next;
2152                 else
2153                         pmc->mca_sources = psf->sf_next;
2154                 if (psf->sf_oldin && !(pmc->mca_flags & MAF_NOREPORT) &&
2155                     !mld_in_v1_mode(idev)) {
2156                         psf->sf_crcount = idev->mc_qrv;
2157                         psf->sf_next = pmc->mca_tomb;
2158                         pmc->mca_tomb = psf;
2159                         rv = 1;
2160                 } else
2161                         kfree(psf);
2162         }
2163         return rv;
2164 }
2165
2166 static int ip6_mc_del_src(struct inet6_dev *idev, const struct in6_addr *pmca,
2167                           int sfmode, int sfcount, const struct in6_addr *psfsrc,
2168                           int delta)
2169 {
2170         struct ifmcaddr6 *pmc;
2171         int     changerec = 0;
2172         int     i, err;
2173
2174         if (!idev)
2175                 return -ENODEV;
2176         read_lock_bh(&idev->lock);
2177         for (pmc = idev->mc_list; pmc; pmc = pmc->next) {
2178                 if (ipv6_addr_equal(pmca, &pmc->mca_addr))
2179                         break;
2180         }
2181         if (!pmc) {
2182                 /* MCA not found?? bug */
2183                 read_unlock_bh(&idev->lock);
2184                 return -ESRCH;
2185         }
2186         spin_lock_bh(&pmc->mca_lock);
2187         sf_markstate(pmc);
2188         if (!delta) {
2189                 if (!pmc->mca_sfcount[sfmode]) {
2190                         spin_unlock_bh(&pmc->mca_lock);
2191                         read_unlock_bh(&idev->lock);
2192                         return -EINVAL;
2193                 }
2194                 pmc->mca_sfcount[sfmode]--;
2195         }
2196         err = 0;
2197         for (i = 0; i < sfcount; i++) {
2198                 int rv = ip6_mc_del1_src(pmc, sfmode, &psfsrc[i]);
2199
2200                 changerec |= rv > 0;
2201                 if (!err && rv < 0)
2202                         err = rv;
2203         }
2204         if (pmc->mca_sfmode == MCAST_EXCLUDE &&
2205             pmc->mca_sfcount[MCAST_EXCLUDE] == 0 &&
2206             pmc->mca_sfcount[MCAST_INCLUDE]) {
2207                 struct ip6_sf_list *psf;
2208
2209                 /* filter mode change */
2210                 pmc->mca_sfmode = MCAST_INCLUDE;
2211                 pmc->mca_crcount = idev->mc_qrv;
2212                 idev->mc_ifc_count = pmc->mca_crcount;
2213                 for (psf = pmc->mca_sources; psf; psf = psf->sf_next)
2214                         psf->sf_crcount = 0;
2215                 mld_ifc_event(pmc->idev);
2216         } else if (sf_setstate(pmc) || changerec)
2217                 mld_ifc_event(pmc->idev);
2218         spin_unlock_bh(&pmc->mca_lock);
2219         read_unlock_bh(&idev->lock);
2220         return err;
2221 }
2222
2223 /*
2224  * Add multicast single-source filter to the interface list
2225  */
2226 static int ip6_mc_add1_src(struct ifmcaddr6 *pmc, int sfmode,
2227         const struct in6_addr *psfsrc)
2228 {
2229         struct ip6_sf_list *psf, *psf_prev;
2230
2231         psf_prev = NULL;
2232         for (psf = pmc->mca_sources; psf; psf = psf->sf_next) {
2233                 if (ipv6_addr_equal(&psf->sf_addr, psfsrc))
2234                         break;
2235                 psf_prev = psf;
2236         }
2237         if (!psf) {
2238                 psf = kzalloc(sizeof(*psf), GFP_ATOMIC);
2239                 if (!psf)
2240                         return -ENOBUFS;
2241
2242                 psf->sf_addr = *psfsrc;
2243                 if (psf_prev) {
2244                         psf_prev->sf_next = psf;
2245                 } else
2246                         pmc->mca_sources = psf;
2247         }
2248         psf->sf_count[sfmode]++;
2249         return 0;
2250 }
2251
2252 static void sf_markstate(struct ifmcaddr6 *pmc)
2253 {
2254         struct ip6_sf_list *psf;
2255         int mca_xcount = pmc->mca_sfcount[MCAST_EXCLUDE];
2256
2257         for (psf = pmc->mca_sources; psf; psf = psf->sf_next)
2258                 if (pmc->mca_sfcount[MCAST_EXCLUDE]) {
2259                         psf->sf_oldin = mca_xcount ==
2260                                 psf->sf_count[MCAST_EXCLUDE] &&
2261                                 !psf->sf_count[MCAST_INCLUDE];
2262                 } else
2263                         psf->sf_oldin = psf->sf_count[MCAST_INCLUDE] != 0;
2264 }
2265
2266 static int sf_setstate(struct ifmcaddr6 *pmc)
2267 {
2268         struct ip6_sf_list *psf, *dpsf;
2269         int mca_xcount = pmc->mca_sfcount[MCAST_EXCLUDE];
2270         int qrv = pmc->idev->mc_qrv;
2271         int new_in, rv;
2272
2273         rv = 0;
2274         for (psf = pmc->mca_sources; psf; psf = psf->sf_next) {
2275                 if (pmc->mca_sfcount[MCAST_EXCLUDE]) {
2276                         new_in = mca_xcount == psf->sf_count[MCAST_EXCLUDE] &&
2277                                 !psf->sf_count[MCAST_INCLUDE];
2278                 } else
2279                         new_in = psf->sf_count[MCAST_INCLUDE] != 0;
2280                 if (new_in) {
2281                         if (!psf->sf_oldin) {
2282                                 struct ip6_sf_list *prev = NULL;
2283
2284                                 for (dpsf = pmc->mca_tomb; dpsf;
2285                                      dpsf = dpsf->sf_next) {
2286                                         if (ipv6_addr_equal(&dpsf->sf_addr,
2287                                             &psf->sf_addr))
2288                                                 break;
2289                                         prev = dpsf;
2290                                 }
2291                                 if (dpsf) {
2292                                         if (prev)
2293                                                 prev->sf_next = dpsf->sf_next;
2294                                         else
2295                                                 pmc->mca_tomb = dpsf->sf_next;
2296                                         kfree(dpsf);
2297                                 }
2298                                 psf->sf_crcount = qrv;
2299                                 rv++;
2300                         }
2301                 } else if (psf->sf_oldin) {
2302                         psf->sf_crcount = 0;
2303                         /*
2304                          * add or update "delete" records if an active filter
2305                          * is now inactive
2306                          */
2307                         for (dpsf = pmc->mca_tomb; dpsf; dpsf = dpsf->sf_next)
2308                                 if (ipv6_addr_equal(&dpsf->sf_addr,
2309                                     &psf->sf_addr))
2310                                         break;
2311                         if (!dpsf) {
2312                                 dpsf = kmalloc(sizeof(*dpsf), GFP_ATOMIC);
2313                                 if (!dpsf)
2314                                         continue;
2315                                 *dpsf = *psf;
2316                                 /* pmc->mca_lock held by callers */
2317                                 dpsf->sf_next = pmc->mca_tomb;
2318                                 pmc->mca_tomb = dpsf;
2319                         }
2320                         dpsf->sf_crcount = qrv;
2321                         rv++;
2322                 }
2323         }
2324         return rv;
2325 }
2326
2327 /*
2328  * Add multicast source filter list to the interface list
2329  */
2330 static int ip6_mc_add_src(struct inet6_dev *idev, const struct in6_addr *pmca,
2331                           int sfmode, int sfcount, const struct in6_addr *psfsrc,
2332                           int delta)
2333 {
2334         struct ifmcaddr6 *pmc;
2335         int     isexclude;
2336         int     i, err;
2337
2338         if (!idev)
2339                 return -ENODEV;
2340         read_lock_bh(&idev->lock);
2341         for (pmc = idev->mc_list; pmc; pmc = pmc->next) {
2342                 if (ipv6_addr_equal(pmca, &pmc->mca_addr))
2343                         break;
2344         }
2345         if (!pmc) {
2346                 /* MCA not found?? bug */
2347                 read_unlock_bh(&idev->lock);
2348                 return -ESRCH;
2349         }
2350         spin_lock_bh(&pmc->mca_lock);
2351
2352         sf_markstate(pmc);
2353         isexclude = pmc->mca_sfmode == MCAST_EXCLUDE;
2354         if (!delta)
2355                 pmc->mca_sfcount[sfmode]++;
2356         err = 0;
2357         for (i = 0; i < sfcount; i++) {
2358                 err = ip6_mc_add1_src(pmc, sfmode, &psfsrc[i]);
2359                 if (err)
2360                         break;
2361         }
2362         if (err) {
2363                 int j;
2364
2365                 if (!delta)
2366                         pmc->mca_sfcount[sfmode]--;
2367                 for (j = 0; j < i; j++)
2368                         ip6_mc_del1_src(pmc, sfmode, &psfsrc[j]);
2369         } else if (isexclude != (pmc->mca_sfcount[MCAST_EXCLUDE] != 0)) {
2370                 struct ip6_sf_list *psf;
2371
2372                 /* filter mode change */
2373                 if (pmc->mca_sfcount[MCAST_EXCLUDE])
2374                         pmc->mca_sfmode = MCAST_EXCLUDE;
2375                 else if (pmc->mca_sfcount[MCAST_INCLUDE])
2376                         pmc->mca_sfmode = MCAST_INCLUDE;
2377                 /* else no filters; keep old mode for reports */
2378
2379                 pmc->mca_crcount = idev->mc_qrv;
2380                 idev->mc_ifc_count = pmc->mca_crcount;
2381                 for (psf = pmc->mca_sources; psf; psf = psf->sf_next)
2382                         psf->sf_crcount = 0;
2383                 mld_ifc_event(idev);
2384         } else if (sf_setstate(pmc))
2385                 mld_ifc_event(idev);
2386         spin_unlock_bh(&pmc->mca_lock);
2387         read_unlock_bh(&idev->lock);
2388         return err;
2389 }
2390
2391 static void ip6_mc_clear_src(struct ifmcaddr6 *pmc)
2392 {
2393         struct ip6_sf_list *psf, *nextpsf;
2394
2395         for (psf = pmc->mca_tomb; psf; psf = nextpsf) {
2396                 nextpsf = psf->sf_next;
2397                 kfree(psf);
2398         }
2399         pmc->mca_tomb = NULL;
2400         for (psf = pmc->mca_sources; psf; psf = nextpsf) {
2401                 nextpsf = psf->sf_next;
2402                 kfree(psf);
2403         }
2404         pmc->mca_sources = NULL;
2405         pmc->mca_sfmode = MCAST_EXCLUDE;
2406         pmc->mca_sfcount[MCAST_INCLUDE] = 0;
2407         pmc->mca_sfcount[MCAST_EXCLUDE] = 1;
2408 }
2409
2410
2411 static void igmp6_join_group(struct ifmcaddr6 *ma)
2412 {
2413         unsigned long delay;
2414
2415         if (ma->mca_flags & MAF_NOREPORT)
2416                 return;
2417
2418         igmp6_send(&ma->mca_addr, ma->idev->dev, ICMPV6_MGM_REPORT);
2419
2420         delay = prandom_u32() % unsolicited_report_interval(ma->idev);
2421
2422         spin_lock_bh(&ma->mca_lock);
2423         if (del_timer(&ma->mca_timer)) {
2424                 refcount_dec(&ma->mca_refcnt);
2425                 delay = ma->mca_timer.expires - jiffies;
2426         }
2427
2428         if (!mod_timer(&ma->mca_timer, jiffies + delay))
2429                 refcount_inc(&ma->mca_refcnt);
2430         ma->mca_flags |= MAF_TIMER_RUNNING | MAF_LAST_REPORTER;
2431         spin_unlock_bh(&ma->mca_lock);
2432 }
2433
2434 static int ip6_mc_leave_src(struct sock *sk, struct ipv6_mc_socklist *iml,
2435                             struct inet6_dev *idev)
2436 {
2437         int err;
2438
2439         /* callers have the socket lock and rtnl lock
2440          * so no other readers or writers of iml or its sflist
2441          */
2442         if (!iml->sflist) {
2443                 /* any-source empty exclude case */
2444                 return ip6_mc_del_src(idev, &iml->addr, iml->sfmode, 0, NULL, 0);
2445         }
2446         err = ip6_mc_del_src(idev, &iml->addr, iml->sfmode,
2447                 iml->sflist->sl_count, iml->sflist->sl_addr, 0);
2448         sock_kfree_s(sk, iml->sflist, IP6_SFLSIZE(iml->sflist->sl_max));
2449         iml->sflist = NULL;
2450         return err;
2451 }
2452
2453 static void igmp6_leave_group(struct ifmcaddr6 *ma)
2454 {
2455         if (mld_in_v1_mode(ma->idev)) {
2456                 if (ma->mca_flags & MAF_LAST_REPORTER)
2457                         igmp6_send(&ma->mca_addr, ma->idev->dev,
2458                                 ICMPV6_MGM_REDUCTION);
2459         } else {
2460                 mld_add_delrec(ma->idev, ma);
2461                 mld_ifc_event(ma->idev);
2462         }
2463 }
2464
2465 static void mld_gq_timer_expire(struct timer_list *t)
2466 {
2467         struct inet6_dev *idev = from_timer(idev, t, mc_gq_timer);
2468
2469         idev->mc_gq_running = 0;
2470         mld_send_report(idev, NULL);
2471         in6_dev_put(idev);
2472 }
2473
2474 static void mld_ifc_timer_expire(struct timer_list *t)
2475 {
2476         struct inet6_dev *idev = from_timer(idev, t, mc_ifc_timer);
2477
2478         mld_send_cr(idev);
2479         if (idev->mc_ifc_count) {
2480                 idev->mc_ifc_count--;
2481                 if (idev->mc_ifc_count)
2482                         mld_ifc_start_timer(idev,
2483                                             unsolicited_report_interval(idev));
2484         }
2485         in6_dev_put(idev);
2486 }
2487
2488 static void mld_ifc_event(struct inet6_dev *idev)
2489 {
2490         if (mld_in_v1_mode(idev))
2491                 return;
2492         idev->mc_ifc_count = idev->mc_qrv;
2493         mld_ifc_start_timer(idev, 1);
2494 }
2495
2496 static void igmp6_timer_handler(struct timer_list *t)
2497 {
2498         struct ifmcaddr6 *ma = from_timer(ma, t, mca_timer);
2499
2500         if (mld_in_v1_mode(ma->idev))
2501                 igmp6_send(&ma->mca_addr, ma->idev->dev, ICMPV6_MGM_REPORT);
2502         else
2503                 mld_send_report(ma->idev, ma);
2504
2505         spin_lock(&ma->mca_lock);
2506         ma->mca_flags |=  MAF_LAST_REPORTER;
2507         ma->mca_flags &= ~MAF_TIMER_RUNNING;
2508         spin_unlock(&ma->mca_lock);
2509         ma_put(ma);
2510 }
2511
2512 /* Device changing type */
2513
2514 void ipv6_mc_unmap(struct inet6_dev *idev)
2515 {
2516         struct ifmcaddr6 *i;
2517
2518         /* Install multicast list, except for all-nodes (already installed) */
2519
2520         read_lock_bh(&idev->lock);
2521         for (i = idev->mc_list; i; i = i->next)
2522                 igmp6_group_dropped(i);
2523         read_unlock_bh(&idev->lock);
2524 }
2525
2526 void ipv6_mc_remap(struct inet6_dev *idev)
2527 {
2528         ipv6_mc_up(idev);
2529 }
2530
2531 /* Device going down */
2532
2533 void ipv6_mc_down(struct inet6_dev *idev)
2534 {
2535         struct ifmcaddr6 *i;
2536
2537         /* Withdraw multicast list */
2538
2539         read_lock_bh(&idev->lock);
2540
2541         for (i = idev->mc_list; i; i = i->next)
2542                 igmp6_group_dropped(i);
2543
2544         /* Should stop timer after group drop. or we will
2545          * start timer again in mld_ifc_event()
2546          */
2547         mld_ifc_stop_timer(idev);
2548         mld_gq_stop_timer(idev);
2549         mld_dad_stop_timer(idev);
2550         read_unlock_bh(&idev->lock);
2551 }
2552
2553 static void ipv6_mc_reset(struct inet6_dev *idev)
2554 {
2555         idev->mc_qrv = sysctl_mld_qrv;
2556         idev->mc_qi = MLD_QI_DEFAULT;
2557         idev->mc_qri = MLD_QRI_DEFAULT;
2558         idev->mc_v1_seen = 0;
2559         idev->mc_maxdelay = unsolicited_report_interval(idev);
2560 }
2561
2562 /* Device going up */
2563
2564 void ipv6_mc_up(struct inet6_dev *idev)
2565 {
2566         struct ifmcaddr6 *i;
2567
2568         /* Install multicast list, except for all-nodes (already installed) */
2569
2570         read_lock_bh(&idev->lock);
2571         ipv6_mc_reset(idev);
2572         for (i = idev->mc_list; i; i = i->next) {
2573                 mld_del_delrec(idev, i);
2574                 igmp6_group_added(i, i->mca_sfmode);
2575         }
2576         read_unlock_bh(&idev->lock);
2577 }
2578
2579 /* IPv6 device initialization. */
2580
2581 void ipv6_mc_init_dev(struct inet6_dev *idev)
2582 {
2583         write_lock_bh(&idev->lock);
2584         spin_lock_init(&idev->mc_lock);
2585         idev->mc_gq_running = 0;
2586         timer_setup(&idev->mc_gq_timer, mld_gq_timer_expire, 0);
2587         idev->mc_tomb = NULL;
2588         idev->mc_ifc_count = 0;
2589         timer_setup(&idev->mc_ifc_timer, mld_ifc_timer_expire, 0);
2590         timer_setup(&idev->mc_dad_timer, mld_dad_timer_expire, 0);
2591         ipv6_mc_reset(idev);
2592         write_unlock_bh(&idev->lock);
2593 }
2594
2595 /*
2596  *      Device is about to be destroyed: clean up.
2597  */
2598
2599 void ipv6_mc_destroy_dev(struct inet6_dev *idev)
2600 {
2601         struct ifmcaddr6 *i;
2602
2603         /* Deactivate timers */
2604         ipv6_mc_down(idev);
2605         mld_clear_delrec(idev);
2606
2607         /* Delete all-nodes address. */
2608         /* We cannot call ipv6_dev_mc_dec() directly, our caller in
2609          * addrconf.c has NULL'd out dev->ip6_ptr so in6_dev_get() will
2610          * fail.
2611          */
2612         __ipv6_dev_mc_dec(idev, &in6addr_linklocal_allnodes);
2613
2614         if (idev->cnf.forwarding)
2615                 __ipv6_dev_mc_dec(idev, &in6addr_linklocal_allrouters);
2616
2617         write_lock_bh(&idev->lock);
2618         while ((i = idev->mc_list) != NULL) {
2619                 idev->mc_list = i->next;
2620
2621                 write_unlock_bh(&idev->lock);
2622                 ma_put(i);
2623                 write_lock_bh(&idev->lock);
2624         }
2625         write_unlock_bh(&idev->lock);
2626 }
2627
2628 static void ipv6_mc_rejoin_groups(struct inet6_dev *idev)
2629 {
2630         struct ifmcaddr6 *pmc;
2631
2632         ASSERT_RTNL();
2633
2634         if (mld_in_v1_mode(idev)) {
2635                 read_lock_bh(&idev->lock);
2636                 for (pmc = idev->mc_list; pmc; pmc = pmc->next)
2637                         igmp6_join_group(pmc);
2638                 read_unlock_bh(&idev->lock);
2639         } else
2640                 mld_send_report(idev, NULL);
2641 }
2642
2643 static int ipv6_mc_netdev_event(struct notifier_block *this,
2644                                 unsigned long event,
2645                                 void *ptr)
2646 {
2647         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2648         struct inet6_dev *idev = __in6_dev_get(dev);
2649
2650         switch (event) {
2651         case NETDEV_RESEND_IGMP:
2652                 if (idev)
2653                         ipv6_mc_rejoin_groups(idev);
2654                 break;
2655         default:
2656                 break;
2657         }
2658
2659         return NOTIFY_DONE;
2660 }
2661
2662 static struct notifier_block igmp6_netdev_notifier = {
2663         .notifier_call = ipv6_mc_netdev_event,
2664 };
2665
2666 #ifdef CONFIG_PROC_FS
2667 struct igmp6_mc_iter_state {
2668         struct seq_net_private p;
2669         struct net_device *dev;
2670         struct inet6_dev *idev;
2671 };
2672
2673 #define igmp6_mc_seq_private(seq)       ((struct igmp6_mc_iter_state *)(seq)->private)
2674
2675 static inline struct ifmcaddr6 *igmp6_mc_get_first(struct seq_file *seq)
2676 {
2677         struct ifmcaddr6 *im = NULL;
2678         struct igmp6_mc_iter_state *state = igmp6_mc_seq_private(seq);
2679         struct net *net = seq_file_net(seq);
2680
2681         state->idev = NULL;
2682         for_each_netdev_rcu(net, state->dev) {
2683                 struct inet6_dev *idev;
2684                 idev = __in6_dev_get(state->dev);
2685                 if (!idev)
2686                         continue;
2687                 read_lock_bh(&idev->lock);
2688                 im = idev->mc_list;
2689                 if (im) {
2690                         state->idev = idev;
2691                         break;
2692                 }
2693                 read_unlock_bh(&idev->lock);
2694         }
2695         return im;
2696 }
2697
2698 static struct ifmcaddr6 *igmp6_mc_get_next(struct seq_file *seq, struct ifmcaddr6 *im)
2699 {
2700         struct igmp6_mc_iter_state *state = igmp6_mc_seq_private(seq);
2701
2702         im = im->next;
2703         while (!im) {
2704                 if (likely(state->idev))
2705                         read_unlock_bh(&state->idev->lock);
2706
2707                 state->dev = next_net_device_rcu(state->dev);
2708                 if (!state->dev) {
2709                         state->idev = NULL;
2710                         break;
2711                 }
2712                 state->idev = __in6_dev_get(state->dev);
2713                 if (!state->idev)
2714                         continue;
2715                 read_lock_bh(&state->idev->lock);
2716                 im = state->idev->mc_list;
2717         }
2718         return im;
2719 }
2720
2721 static struct ifmcaddr6 *igmp6_mc_get_idx(struct seq_file *seq, loff_t pos)
2722 {
2723         struct ifmcaddr6 *im = igmp6_mc_get_first(seq);
2724         if (im)
2725                 while (pos && (im = igmp6_mc_get_next(seq, im)) != NULL)
2726                         --pos;
2727         return pos ? NULL : im;
2728 }
2729
2730 static void *igmp6_mc_seq_start(struct seq_file *seq, loff_t *pos)
2731         __acquires(RCU)
2732 {
2733         rcu_read_lock();
2734         return igmp6_mc_get_idx(seq, *pos);
2735 }
2736
2737 static void *igmp6_mc_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2738 {
2739         struct ifmcaddr6 *im = igmp6_mc_get_next(seq, v);
2740
2741         ++*pos;
2742         return im;
2743 }
2744
2745 static void igmp6_mc_seq_stop(struct seq_file *seq, void *v)
2746         __releases(RCU)
2747 {
2748         struct igmp6_mc_iter_state *state = igmp6_mc_seq_private(seq);
2749
2750         if (likely(state->idev)) {
2751                 read_unlock_bh(&state->idev->lock);
2752                 state->idev = NULL;
2753         }
2754         state->dev = NULL;
2755         rcu_read_unlock();
2756 }
2757
2758 static int igmp6_mc_seq_show(struct seq_file *seq, void *v)
2759 {
2760         struct ifmcaddr6 *im = (struct ifmcaddr6 *)v;
2761         struct igmp6_mc_iter_state *state = igmp6_mc_seq_private(seq);
2762
2763         seq_printf(seq,
2764                    "%-4d %-15s %pi6 %5d %08X %ld\n",
2765                    state->dev->ifindex, state->dev->name,
2766                    &im->mca_addr,
2767                    im->mca_users, im->mca_flags,
2768                    (im->mca_flags&MAF_TIMER_RUNNING) ?
2769                    jiffies_to_clock_t(im->mca_timer.expires-jiffies) : 0);
2770         return 0;
2771 }
2772
2773 static const struct seq_operations igmp6_mc_seq_ops = {
2774         .start  =       igmp6_mc_seq_start,
2775         .next   =       igmp6_mc_seq_next,
2776         .stop   =       igmp6_mc_seq_stop,
2777         .show   =       igmp6_mc_seq_show,
2778 };
2779
2780 struct igmp6_mcf_iter_state {
2781         struct seq_net_private p;
2782         struct net_device *dev;
2783         struct inet6_dev *idev;
2784         struct ifmcaddr6 *im;
2785 };
2786
2787 #define igmp6_mcf_seq_private(seq)      ((struct igmp6_mcf_iter_state *)(seq)->private)
2788
2789 static inline struct ip6_sf_list *igmp6_mcf_get_first(struct seq_file *seq)
2790 {
2791         struct ip6_sf_list *psf = NULL;
2792         struct ifmcaddr6 *im = NULL;
2793         struct igmp6_mcf_iter_state *state = igmp6_mcf_seq_private(seq);
2794         struct net *net = seq_file_net(seq);
2795
2796         state->idev = NULL;
2797         state->im = NULL;
2798         for_each_netdev_rcu(net, state->dev) {
2799                 struct inet6_dev *idev;
2800                 idev = __in6_dev_get(state->dev);
2801                 if (unlikely(idev == NULL))
2802                         continue;
2803                 read_lock_bh(&idev->lock);
2804                 im = idev->mc_list;
2805                 if (likely(im)) {
2806                         spin_lock_bh(&im->mca_lock);
2807                         psf = im->mca_sources;
2808                         if (likely(psf)) {
2809                                 state->im = im;
2810                                 state->idev = idev;
2811                                 break;
2812                         }
2813                         spin_unlock_bh(&im->mca_lock);
2814                 }
2815                 read_unlock_bh(&idev->lock);
2816         }
2817         return psf;
2818 }
2819
2820 static struct ip6_sf_list *igmp6_mcf_get_next(struct seq_file *seq, struct ip6_sf_list *psf)
2821 {
2822         struct igmp6_mcf_iter_state *state = igmp6_mcf_seq_private(seq);
2823
2824         psf = psf->sf_next;
2825         while (!psf) {
2826                 spin_unlock_bh(&state->im->mca_lock);
2827                 state->im = state->im->next;
2828                 while (!state->im) {
2829                         if (likely(state->idev))
2830                                 read_unlock_bh(&state->idev->lock);
2831
2832                         state->dev = next_net_device_rcu(state->dev);
2833                         if (!state->dev) {
2834                                 state->idev = NULL;
2835                                 goto out;
2836                         }
2837                         state->idev = __in6_dev_get(state->dev);
2838                         if (!state->idev)
2839                                 continue;
2840                         read_lock_bh(&state->idev->lock);
2841                         state->im = state->idev->mc_list;
2842                 }
2843                 if (!state->im)
2844                         break;
2845                 spin_lock_bh(&state->im->mca_lock);
2846                 psf = state->im->mca_sources;
2847         }
2848 out:
2849         return psf;
2850 }
2851
2852 static struct ip6_sf_list *igmp6_mcf_get_idx(struct seq_file *seq, loff_t pos)
2853 {
2854         struct ip6_sf_list *psf = igmp6_mcf_get_first(seq);
2855         if (psf)
2856                 while (pos && (psf = igmp6_mcf_get_next(seq, psf)) != NULL)
2857                         --pos;
2858         return pos ? NULL : psf;
2859 }
2860
2861 static void *igmp6_mcf_seq_start(struct seq_file *seq, loff_t *pos)
2862         __acquires(RCU)
2863 {
2864         rcu_read_lock();
2865         return *pos ? igmp6_mcf_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
2866 }
2867
2868 static void *igmp6_mcf_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2869 {
2870         struct ip6_sf_list *psf;
2871         if (v == SEQ_START_TOKEN)
2872                 psf = igmp6_mcf_get_first(seq);
2873         else
2874                 psf = igmp6_mcf_get_next(seq, v);
2875         ++*pos;
2876         return psf;
2877 }
2878
2879 static void igmp6_mcf_seq_stop(struct seq_file *seq, void *v)
2880         __releases(RCU)
2881 {
2882         struct igmp6_mcf_iter_state *state = igmp6_mcf_seq_private(seq);
2883         if (likely(state->im)) {
2884                 spin_unlock_bh(&state->im->mca_lock);
2885                 state->im = NULL;
2886         }
2887         if (likely(state->idev)) {
2888                 read_unlock_bh(&state->idev->lock);
2889                 state->idev = NULL;
2890         }
2891         state->dev = NULL;
2892         rcu_read_unlock();
2893 }
2894
2895 static int igmp6_mcf_seq_show(struct seq_file *seq, void *v)
2896 {
2897         struct ip6_sf_list *psf = (struct ip6_sf_list *)v;
2898         struct igmp6_mcf_iter_state *state = igmp6_mcf_seq_private(seq);
2899
2900         if (v == SEQ_START_TOKEN) {
2901                 seq_puts(seq, "Idx Device                Multicast Address                   Source Address    INC    EXC\n");
2902         } else {
2903                 seq_printf(seq,
2904                            "%3d %6.6s %pi6 %pi6 %6lu %6lu\n",
2905                            state->dev->ifindex, state->dev->name,
2906                            &state->im->mca_addr,
2907                            &psf->sf_addr,
2908                            psf->sf_count[MCAST_INCLUDE],
2909                            psf->sf_count[MCAST_EXCLUDE]);
2910         }
2911         return 0;
2912 }
2913
2914 static const struct seq_operations igmp6_mcf_seq_ops = {
2915         .start  =       igmp6_mcf_seq_start,
2916         .next   =       igmp6_mcf_seq_next,
2917         .stop   =       igmp6_mcf_seq_stop,
2918         .show   =       igmp6_mcf_seq_show,
2919 };
2920
2921 static int __net_init igmp6_proc_init(struct net *net)
2922 {
2923         int err;
2924
2925         err = -ENOMEM;
2926         if (!proc_create_net("igmp6", 0444, net->proc_net, &igmp6_mc_seq_ops,
2927                         sizeof(struct igmp6_mc_iter_state)))
2928                 goto out;
2929         if (!proc_create_net("mcfilter6", 0444, net->proc_net,
2930                         &igmp6_mcf_seq_ops,
2931                         sizeof(struct igmp6_mcf_iter_state)))
2932                 goto out_proc_net_igmp6;
2933
2934         err = 0;
2935 out:
2936         return err;
2937
2938 out_proc_net_igmp6:
2939         remove_proc_entry("igmp6", net->proc_net);
2940         goto out;
2941 }
2942
2943 static void __net_exit igmp6_proc_exit(struct net *net)
2944 {
2945         remove_proc_entry("mcfilter6", net->proc_net);
2946         remove_proc_entry("igmp6", net->proc_net);
2947 }
2948 #else
2949 static inline int igmp6_proc_init(struct net *net)
2950 {
2951         return 0;
2952 }
2953 static inline void igmp6_proc_exit(struct net *net)
2954 {
2955 }
2956 #endif
2957
2958 static int __net_init igmp6_net_init(struct net *net)
2959 {
2960         int err;
2961
2962         err = inet_ctl_sock_create(&net->ipv6.igmp_sk, PF_INET6,
2963                                    SOCK_RAW, IPPROTO_ICMPV6, net);
2964         if (err < 0) {
2965                 pr_err("Failed to initialize the IGMP6 control socket (err %d)\n",
2966                        err);
2967                 goto out;
2968         }
2969
2970         inet6_sk(net->ipv6.igmp_sk)->hop_limit = 1;
2971
2972         err = inet_ctl_sock_create(&net->ipv6.mc_autojoin_sk, PF_INET6,
2973                                    SOCK_RAW, IPPROTO_ICMPV6, net);
2974         if (err < 0) {
2975                 pr_err("Failed to initialize the IGMP6 autojoin socket (err %d)\n",
2976                        err);
2977                 goto out_sock_create;
2978         }
2979
2980         err = igmp6_proc_init(net);
2981         if (err)
2982                 goto out_sock_create_autojoin;
2983
2984         return 0;
2985
2986 out_sock_create_autojoin:
2987         inet_ctl_sock_destroy(net->ipv6.mc_autojoin_sk);
2988 out_sock_create:
2989         inet_ctl_sock_destroy(net->ipv6.igmp_sk);
2990 out:
2991         return err;
2992 }
2993
2994 static void __net_exit igmp6_net_exit(struct net *net)
2995 {
2996         inet_ctl_sock_destroy(net->ipv6.igmp_sk);
2997         inet_ctl_sock_destroy(net->ipv6.mc_autojoin_sk);
2998         igmp6_proc_exit(net);
2999 }
3000
3001 static struct pernet_operations igmp6_net_ops = {
3002         .init = igmp6_net_init,
3003         .exit = igmp6_net_exit,
3004 };
3005
3006 int __init igmp6_init(void)
3007 {
3008         return register_pernet_subsys(&igmp6_net_ops);
3009 }
3010
3011 int __init igmp6_late_init(void)
3012 {
3013         return register_netdevice_notifier(&igmp6_netdev_notifier);
3014 }
3015
3016 void igmp6_cleanup(void)
3017 {
3018         unregister_pernet_subsys(&igmp6_net_ops);
3019 }
3020
3021 void igmp6_late_cleanup(void)
3022 {
3023         unregister_netdevice_notifier(&igmp6_netdev_notifier);
3024 }