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)
 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 (mc->mca_sfmode == 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                         swap(im->mca_tomb, pmc->mca_tomb);
 795                         swap(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                 ip6_mc_clear_src(pmc);
 803                 kfree(pmc);
 804         }
 805         spin_unlock_bh(&im->mca_lock);
 806 }
 807
 808 static void mld_clear_delrec(struct inet6_dev *idev)
 809 {
 810         struct ifmcaddr6 *pmc, *nextpmc;
 811
 812         spin_lock_bh(&idev->mc_lock);
 813         pmc = idev->mc_tomb;
 814         idev->mc_tomb = NULL;
 815         spin_unlock_bh(&idev->mc_lock);
 816
 817         for (; pmc; pmc = nextpmc) {
 818                 nextpmc = pmc->next;
 819                 ip6_mc_clear_src(pmc);
 820                 in6_dev_put(pmc->idev);
 821                 kfree(pmc);
 822         }
 823
 824         /* clear dead sources, too */
 825         read_lock_bh(&idev->lock);
 826         for (pmc = idev->mc_list; pmc; pmc = pmc->next) {
 827                 struct ip6_sf_list *psf, *psf_next;
 828
 829                 spin_lock_bh(&pmc->mca_lock);
 830                 psf = pmc->mca_tomb;
 831                 pmc->mca_tomb = NULL;
 832                 spin_unlock_bh(&pmc->mca_lock);
 833                 for (; psf; psf = psf_next) {
 834                         psf_next = psf->sf_next;
 835                         kfree(psf);
 836                 }
 837         }
 838         read_unlock_bh(&idev->lock);
 839 }
 840
 841 static void mca_get(struct ifmcaddr6 *mc)
 842 {
 843         refcount_inc(&mc->mca_refcnt);
 844 }
 845
 846 static void ma_put(struct ifmcaddr6 *mc)
 847 {
 848         if (refcount_dec_and_test(&mc->mca_refcnt)) {
 849                 in6_dev_put(mc->idev);
 850                 kfree(mc);
 851         }
 852 }
 853
 854 static struct ifmcaddr6 *mca_alloc(struct inet6_dev *idev,
 855                                    const struct in6_addr *addr,
 856                                    unsigned int mode)
 857 {
 858         struct ifmcaddr6 *mc;
 859
 860         mc = kzalloc(sizeof(*mc), GFP_ATOMIC);
 861         if (!mc)
 862                 return NULL;
 863
 864         timer_setup(&mc->mca_timer, igmp6_timer_handler, 0);
 865
 866         mc->mca_addr = *addr;
 867         mc->idev = idev; /* reference taken by caller */
 868         mc->mca_users = 1;
 869         /* mca_stamp should be updated upon changes */
 870         mc->mca_cstamp = mc->mca_tstamp = jiffies;
 871         refcount_set(&mc->mca_refcnt, 1);
 872         spin_lock_init(&mc->mca_lock);
 873
 874         mc->mca_sfmode = mode;
 875         mc->mca_sfcount[mode] = 1;
 876
 877         if (ipv6_addr_is_ll_all_nodes(&mc->mca_addr) ||
 878             IPV6_ADDR_MC_SCOPE(&mc->mca_addr) < IPV6_ADDR_SCOPE_LINKLOCAL)
 879                 mc->mca_flags |= MAF_NOREPORT;
 880
 881         return mc;
 882 }
 883
 884 /*
 885  *      device multicast group inc (add if not found)
 886  */
 887 static int __ipv6_dev_mc_inc(struct net_device *dev,
 888                              const struct in6_addr *addr, unsigned int mode)
 889 {
 890         struct ifmcaddr6 *mc;
 891         struct inet6_dev *idev;
 892
 893         ASSERT_RTNL();
 894
 895         /* we need to take a reference on idev */
 896         idev = in6_dev_get(dev);
 897
 898         if (!idev)
 899                 return -EINVAL;
 900
 901         write_lock_bh(&idev->lock);
 902         if (idev->dead) {
 903                 write_unlock_bh(&idev->lock);
 904                 in6_dev_put(idev);
 905                 return -ENODEV;
 906         }
 907
 908         for (mc = idev->mc_list; mc; mc = mc->next) {
 909                 if (ipv6_addr_equal(&mc->mca_addr, addr)) {
 910                         mc->mca_users++;
 911                         write_unlock_bh(&idev->lock);
 912                         ip6_mc_add_src(idev, &mc->mca_addr, mode, 0, NULL, 0);
 913                         in6_dev_put(idev);
 914                         return 0;
 915                 }
 916         }
 917
 918         mc = mca_alloc(idev, addr, mode);
 919         if (!mc) {
 920                 write_unlock_bh(&idev->lock);
 921                 in6_dev_put(idev);
 922                 return -ENOMEM;
 923         }
 924
 925         mc->next = idev->mc_list;
 926         idev->mc_list = mc;
 927
 928         /* Hold this for the code below before we unlock,
 929          * it is already exposed via idev->mc_list.
 930          */
 931         mca_get(mc);
 932         write_unlock_bh(&idev->lock);
 933
 934         mld_del_delrec(idev, mc);
 935         igmp6_group_added(mc);
 936         ma_put(mc);
 937         return 0;
 938 }
 939
 940 int ipv6_dev_mc_inc(struct net_device *dev, const struct in6_addr *addr)
 941 {
 942         return __ipv6_dev_mc_inc(dev, addr, MCAST_EXCLUDE);
 943 }
 944
 945 /*
 946  *      device multicast group del
 947  */
 948 int __ipv6_dev_mc_dec(struct inet6_dev *idev, const struct in6_addr *addr)
 949 {
 950         struct ifmcaddr6 *ma, **map;
 951
 952         ASSERT_RTNL();
 953
 954         write_lock_bh(&idev->lock);
 955         for (map = &idev->mc_list; (ma = *map) != NULL; map = &ma->next) {
 956                 if (ipv6_addr_equal(&ma->mca_addr, addr)) {
 957                         if (--ma->mca_users == 0) {
 958                                 *map = ma->next;
 959                                 write_unlock_bh(&idev->lock);
 960
 961                                 igmp6_group_dropped(ma);
 962                                 ip6_mc_clear_src(ma);
 963
 964                                 ma_put(ma);
 965                                 return 0;
 966                         }
 967                         write_unlock_bh(&idev->lock);
 968                         return 0;
 969                 }
 970         }
 971         write_unlock_bh(&idev->lock);
 972
 973         return -ENOENT;
 974 }
 975
 976 int ipv6_dev_mc_dec(struct net_device *dev, const struct in6_addr *addr)
 977 {
 978         struct inet6_dev *idev;
 979         int err;
 980
 981         ASSERT_RTNL();
 982
 983         idev = __in6_dev_get(dev);
 984         if (!idev)
 985                 err = -ENODEV;
 986         else
 987                 err = __ipv6_dev_mc_dec(idev, addr);
 988
 989         return err;
 990 }
 991
 992 /*
 993  *      check if the interface/address pair is valid
 994  */
 995 bool ipv6_chk_mcast_addr(struct net_device *dev, const struct in6_addr *group,
 996                          const struct in6_addr *src_addr)
 997 {
 998         struct inet6_dev *idev;
 999         struct ifmcaddr6 *mc;
1000         bool rv = false;
1001
1002         rcu_read_lock();
1003         idev = __in6_dev_get(dev);
1004         if (idev) {
1005                 read_lock_bh(&idev->lock);
1006                 for (mc = idev->mc_list; mc; mc = mc->next) {
1007                         if (ipv6_addr_equal(&mc->mca_addr, group))
1008                                 break;
1009                 }
1010                 if (mc) {
1011                         if (src_addr && !ipv6_addr_any(src_addr)) {
1012                                 struct ip6_sf_list *psf;
1013
1014                                 spin_lock_bh(&mc->mca_lock);
1015                                 for (psf = mc->mca_sources; psf; psf = psf->sf_next) {
1016                                         if (ipv6_addr_equal(&psf->sf_addr, src_addr))
1017                                                 break;
1018                                 }
1019                                 if (psf)
1020                                         rv = psf->sf_count[MCAST_INCLUDE] ||
1021                                                 psf->sf_count[MCAST_EXCLUDE] !=
1022                                                 mc->mca_sfcount[MCAST_EXCLUDE];
1023                                 else
1024                                         rv = mc->mca_sfcount[MCAST_EXCLUDE] != 0;
1025                                 spin_unlock_bh(&mc->mca_lock);
1026                         } else
1027                                 rv = true; /* don't filter unspecified source */
1028                 }
1029                 read_unlock_bh(&idev->lock);
1030         }
1031         rcu_read_unlock();
1032         return rv;
1033 }
1034
1035 static void mld_gq_start_timer(struct inet6_dev *idev)
1036 {
1037         unsigned long tv = prandom_u32() % idev->mc_maxdelay;
1038
1039         idev->mc_gq_running = 1;
1040         if (!mod_timer(&idev->mc_gq_timer, jiffies+tv+2))
1041                 in6_dev_hold(idev);
1042 }
1043
1044 static void mld_gq_stop_timer(struct inet6_dev *idev)
1045 {
1046         idev->mc_gq_running = 0;
1047         if (del_timer(&idev->mc_gq_timer))
1048                 __in6_dev_put(idev);
1049 }
1050
1051 static void mld_ifc_start_timer(struct inet6_dev *idev, unsigned long delay)
1052 {
1053         unsigned long tv = prandom_u32() % delay;
1054
1055         if (!mod_timer(&idev->mc_ifc_timer, jiffies+tv+2))
1056                 in6_dev_hold(idev);
1057 }
1058
1059 static void mld_ifc_stop_timer(struct inet6_dev *idev)
1060 {
1061         idev->mc_ifc_count = 0;
1062         if (del_timer(&idev->mc_ifc_timer))
1063                 __in6_dev_put(idev);
1064 }
1065
1066 static void mld_dad_start_timer(struct inet6_dev *idev, unsigned long delay)
1067 {
1068         unsigned long tv = prandom_u32() % delay;
1069
1070         if (!mod_timer(&idev->mc_dad_timer, jiffies+tv+2))
1071                 in6_dev_hold(idev);
1072 }
1073
1074 static void mld_dad_stop_timer(struct inet6_dev *idev)
1075 {
1076         if (del_timer(&idev->mc_dad_timer))
1077                 __in6_dev_put(idev);
1078 }
1079
1080 /*
1081  *      IGMP handling (alias multicast ICMPv6 messages)
1082  */
1083
1084 static void igmp6_group_queried(struct ifmcaddr6 *ma, unsigned long resptime)
1085 {
1086         unsigned long delay = resptime;
1087
1088         /* Do not start timer for these addresses */
1089         if (ipv6_addr_is_ll_all_nodes(&ma->mca_addr) ||
1090             IPV6_ADDR_MC_SCOPE(&ma->mca_addr) < IPV6_ADDR_SCOPE_LINKLOCAL)
1091                 return;
1092
1093         if (del_timer(&ma->mca_timer)) {
1094                 refcount_dec(&ma->mca_refcnt);
1095                 delay = ma->mca_timer.expires - jiffies;
1096         }
1097
1098         if (delay >= resptime)
1099                 delay = prandom_u32() % resptime;
1100
1101         ma->mca_timer.expires = jiffies + delay;
1102         if (!mod_timer(&ma->mca_timer, jiffies + delay))
1103                 refcount_inc(&ma->mca_refcnt);
1104         ma->mca_flags |= MAF_TIMER_RUNNING;
1105 }
1106
1107 /* mark EXCLUDE-mode sources */
1108 static bool mld_xmarksources(struct ifmcaddr6 *pmc, int nsrcs,
1109                              const struct in6_addr *srcs)
1110 {
1111         struct ip6_sf_list *psf;
1112         int i, scount;
1113
1114         scount = 0;
1115         for (psf = pmc->mca_sources; psf; psf = psf->sf_next) {
1116                 if (scount == nsrcs)
1117                         break;
1118                 for (i = 0; i < nsrcs; i++) {
1119                         /* skip inactive filters */
1120                         if (psf->sf_count[MCAST_INCLUDE] ||
1121                             pmc->mca_sfcount[MCAST_EXCLUDE] !=
1122                             psf->sf_count[MCAST_EXCLUDE])
1123                                 break;
1124                         if (ipv6_addr_equal(&srcs[i], &psf->sf_addr)) {
1125                                 scount++;
1126                                 break;
1127                         }
1128                 }
1129         }
1130         pmc->mca_flags &= ~MAF_GSQUERY;
1131         if (scount == nsrcs)    /* all sources excluded */
1132                 return false;
1133         return true;
1134 }
1135
1136 static bool mld_marksources(struct ifmcaddr6 *pmc, int nsrcs,
1137                             const struct in6_addr *srcs)
1138 {
1139         struct ip6_sf_list *psf;
1140         int i, scount;
1141
1142         if (pmc->mca_sfmode == MCAST_EXCLUDE)
1143                 return mld_xmarksources(pmc, nsrcs, srcs);
1144
1145         /* mark INCLUDE-mode sources */
1146
1147         scount = 0;
1148         for (psf = pmc->mca_sources; psf; psf = psf->sf_next) {
1149                 if (scount == nsrcs)
1150                         break;
1151                 for (i = 0; i < nsrcs; i++) {
1152                         if (ipv6_addr_equal(&srcs[i], &psf->sf_addr)) {
1153                                 psf->sf_gsresp = 1;
1154                                 scount++;
1155                                 break;
1156                         }
1157                 }
1158         }
1159         if (!scount) {
1160                 pmc->mca_flags &= ~MAF_GSQUERY;
1161                 return false;
1162         }
1163         pmc->mca_flags |= MAF_GSQUERY;
1164         return true;
1165 }
1166
1167 static int mld_force_mld_version(const struct inet6_dev *idev)
1168 {
1169         /* Normally, both are 0 here. If enforcement to a particular is
1170          * being used, individual device enforcement will have a lower
1171          * precedence over 'all' device (.../conf/all/force_mld_version).
1172          */
1173
1174         if (dev_net(idev->dev)->ipv6.devconf_all->force_mld_version != 0)
1175                 return dev_net(idev->dev)->ipv6.devconf_all->force_mld_version;
1176         else
1177                 return idev->cnf.force_mld_version;
1178 }
1179
1180 static bool mld_in_v2_mode_only(const struct inet6_dev *idev)
1181 {
1182         return mld_force_mld_version(idev) == 2;
1183 }
1184
1185 static bool mld_in_v1_mode_only(const struct inet6_dev *idev)
1186 {
1187         return mld_force_mld_version(idev) == 1;
1188 }
1189
1190 static bool mld_in_v1_mode(const struct inet6_dev *idev)
1191 {
1192         if (mld_in_v2_mode_only(idev))
1193                 return false;
1194         if (mld_in_v1_mode_only(idev))
1195                 return true;
1196         if (idev->mc_v1_seen && time_before(jiffies, idev->mc_v1_seen))
1197                 return true;
1198
1199         return false;
1200 }
1201
1202 static void mld_set_v1_mode(struct inet6_dev *idev)
1203 {
1204         /* RFC3810, relevant sections:
1205          *  - 9.1. Robustness Variable
1206          *  - 9.2. Query Interval
1207          *  - 9.3. Query Response Interval
1208          *  - 9.12. Older Version Querier Present Timeout
1209          */
1210         unsigned long switchback;
1211
1212         switchback = (idev->mc_qrv * idev->mc_qi) + idev->mc_qri;
1213
1214         idev->mc_v1_seen = jiffies + switchback;
1215 }
1216
1217 static void mld_update_qrv(struct inet6_dev *idev,
1218                            const struct mld2_query *mlh2)
1219 {
1220         /* RFC3810, relevant sections:
1221          *  - 5.1.8. QRV (Querier's Robustness Variable)
1222          *  - 9.1. Robustness Variable
1223          */
1224
1225         /* The value of the Robustness Variable MUST NOT be zero,
1226          * and SHOULD NOT be one. Catch this here if we ever run
1227          * into such a case in future.
1228          */
1229         const int min_qrv = min(MLD_QRV_DEFAULT, sysctl_mld_qrv);
1230         WARN_ON(idev->mc_qrv == 0);
1231
1232         if (mlh2->mld2q_qrv > 0)
1233                 idev->mc_qrv = mlh2->mld2q_qrv;
1234
1235         if (unlikely(idev->mc_qrv < min_qrv)) {
1236                 net_warn_ratelimited("IPv6: MLD: clamping QRV from %u to %u!\n",
1237                                      idev->mc_qrv, min_qrv);
1238                 idev->mc_qrv = min_qrv;
1239         }
1240 }
1241
1242 static void mld_update_qi(struct inet6_dev *idev,
1243                           const struct mld2_query *mlh2)
1244 {
1245         /* RFC3810, relevant sections:
1246          *  - 5.1.9. QQIC (Querier's Query Interval Code)
1247          *  - 9.2. Query Interval
1248          *  - 9.12. Older Version Querier Present Timeout
1249          *    (the [Query Interval] in the last Query received)
1250          */
1251         unsigned long mc_qqi;
1252
1253         if (mlh2->mld2q_qqic < 128) {
1254                 mc_qqi = mlh2->mld2q_qqic;
1255         } else {
1256                 unsigned long mc_man, mc_exp;
1257
1258                 mc_exp = MLDV2_QQIC_EXP(mlh2->mld2q_qqic);
1259                 mc_man = MLDV2_QQIC_MAN(mlh2->mld2q_qqic);
1260
1261                 mc_qqi = (mc_man | 0x10) << (mc_exp + 3);
1262         }
1263
1264         idev->mc_qi = mc_qqi * HZ;
1265 }
1266
1267 static void mld_update_qri(struct inet6_dev *idev,
1268                            const struct mld2_query *mlh2)
1269 {
1270         /* RFC3810, relevant sections:
1271          *  - 5.1.3. Maximum Response Code
1272          *  - 9.3. Query Response Interval
1273          */
1274         idev->mc_qri = msecs_to_jiffies(mldv2_mrc(mlh2));
1275 }
1276
1277 static int mld_process_v1(struct inet6_dev *idev, struct mld_msg *mld,
1278                           unsigned long *max_delay, bool v1_query)
1279 {
1280         unsigned long mldv1_md;
1281
1282         /* Ignore v1 queries */
1283         if (mld_in_v2_mode_only(idev))
1284                 return -EINVAL;
1285
1286         mldv1_md = ntohs(mld->mld_maxdelay);
1287
1288         /* When in MLDv1 fallback and a MLDv2 router start-up being
1289          * unaware of current MLDv1 operation, the MRC == MRD mapping
1290          * only works when the exponential algorithm is not being
1291          * used (as MLDv1 is unaware of such things).
1292          *
1293          * According to the RFC author, the MLDv2 implementations
1294          * he's aware of all use a MRC < 32768 on start up queries.
1295          *
1296          * Thus, should we *ever* encounter something else larger
1297          * than that, just assume the maximum possible within our
1298          * reach.
1299          */
1300         if (!v1_query)
1301                 mldv1_md = min(mldv1_md, MLDV1_MRD_MAX_COMPAT);
1302
1303         *max_delay = max(msecs_to_jiffies(mldv1_md), 1UL);
1304
1305         /* MLDv1 router present: we need to go into v1 mode *only*
1306          * when an MLDv1 query is received as per section 9.12. of
1307          * RFC3810! And we know from RFC2710 section 3.7 that MLDv1
1308          * queries MUST be of exactly 24 octets.
1309          */
1310         if (v1_query)
1311                 mld_set_v1_mode(idev);
1312
1313         /* cancel MLDv2 report timer */
1314         mld_gq_stop_timer(idev);
1315         /* cancel the interface change timer */
1316         mld_ifc_stop_timer(idev);
1317         /* clear deleted report items */
1318         mld_clear_delrec(idev);
1319
1320         return 0;
1321 }
1322
1323 static int mld_process_v2(struct inet6_dev *idev, struct mld2_query *mld,
1324                           unsigned long *max_delay)
1325 {
1326         *max_delay = max(msecs_to_jiffies(mldv2_mrc(mld)), 1UL);
1327
1328         mld_update_qrv(idev, mld);
1329         mld_update_qi(idev, mld);
1330         mld_update_qri(idev, mld);
1331
1332         idev->mc_maxdelay = *max_delay;
1333
1334         return 0;
1335 }
1336
1337 /* called with rcu_read_lock() */
1338 int igmp6_event_query(struct sk_buff *skb)
1339 {
1340         struct mld2_query *mlh2 = NULL;
1341         struct ifmcaddr6 *ma;
1342         const struct in6_addr *group;
1343         unsigned long max_delay;
1344         struct inet6_dev *idev;
1345         struct mld_msg *mld;
1346         int group_type;
1347         int mark = 0;
1348         int len, err;
1349
1350         if (!pskb_may_pull(skb, sizeof(struct in6_addr)))
1351                 return -EINVAL;
1352
1353         /* compute payload length excluding extension headers */
1354         len = ntohs(ipv6_hdr(skb)->payload_len) + sizeof(struct ipv6hdr);
1355         len -= skb_network_header_len(skb);
1356
1357         /* RFC3810 6.2
1358          * Upon reception of an MLD message that contains a Query, the node
1359          * checks if the source address of the message is a valid link-local
1360          * address, if the Hop Limit is set to 1, and if the Router Alert
1361          * option is present in the Hop-By-Hop Options header of the IPv6
1362          * packet.  If any of these checks fails, the packet is dropped.
1363          */
1364         if (!(ipv6_addr_type(&ipv6_hdr(skb)->saddr) & IPV6_ADDR_LINKLOCAL) ||
1365             ipv6_hdr(skb)->hop_limit != 1 ||
1366             !(IP6CB(skb)->flags & IP6SKB_ROUTERALERT) ||
1367             IP6CB(skb)->ra != htons(IPV6_OPT_ROUTERALERT_MLD))
1368                 return -EINVAL;
1369
1370         idev = __in6_dev_get(skb->dev);
1371         if (!idev)
1372                 return 0;
1373
1374         mld = (struct mld_msg *)icmp6_hdr(skb);
1375         group = &mld->mld_mca;
1376         group_type = ipv6_addr_type(group);
1377
1378         if (group_type != IPV6_ADDR_ANY &&
1379             !(group_type&IPV6_ADDR_MULTICAST))
1380                 return -EINVAL;
1381
1382         if (len < MLD_V1_QUERY_LEN) {
1383                 return -EINVAL;
1384         } else if (len == MLD_V1_QUERY_LEN || mld_in_v1_mode(idev)) {
1385                 err = mld_process_v1(idev, mld, &max_delay,
1386                                      len == MLD_V1_QUERY_LEN);
1387                 if (err < 0)
1388                         return err;
1389         } else if (len >= MLD_V2_QUERY_LEN_MIN) {
1390                 int srcs_offset = sizeof(struct mld2_query) -
1391                                   sizeof(struct icmp6hdr);
1392
1393                 if (!pskb_may_pull(skb, srcs_offset))
1394                         return -EINVAL;
1395
1396                 mlh2 = (struct mld2_query *)skb_transport_header(skb);
1397
1398                 err = mld_process_v2(idev, mlh2, &max_delay);
1399                 if (err < 0)
1400                         return err;
1401
1402                 if (group_type == IPV6_ADDR_ANY) { /* general query */
1403                         if (mlh2->mld2q_nsrcs)
1404                                 return -EINVAL; /* no sources allowed */
1405
1406                         mld_gq_start_timer(idev);
1407                         return 0;
1408                 }
1409                 /* mark sources to include, if group & source-specific */
1410                 if (mlh2->mld2q_nsrcs != 0) {
1411                         if (!pskb_may_pull(skb, srcs_offset +
1412                             ntohs(mlh2->mld2q_nsrcs) * sizeof(struct in6_addr)))
1413                                 return -EINVAL;
1414
1415                         mlh2 = (struct mld2_query *)skb_transport_header(skb);
1416                         mark = 1;
1417                 }
1418         } else {
1419                 return -EINVAL;
1420         }
1421
1422         read_lock_bh(&idev->lock);
1423         if (group_type == IPV6_ADDR_ANY) {
1424                 for (ma = idev->mc_list; ma; ma = ma->next) {
1425                         spin_lock_bh(&ma->mca_lock);
1426                         igmp6_group_queried(ma, max_delay);
1427                         spin_unlock_bh(&ma->mca_lock);
1428                 }
1429         } else {
1430                 for (ma = idev->mc_list; ma; ma = ma->next) {
1431                         if (!ipv6_addr_equal(group, &ma->mca_addr))
1432                                 continue;
1433                         spin_lock_bh(&ma->mca_lock);
1434                         if (ma->mca_flags & MAF_TIMER_RUNNING) {
1435                                 /* gsquery <- gsquery && mark */
1436                                 if (!mark)
1437                                         ma->mca_flags &= ~MAF_GSQUERY;
1438                         } else {
1439                                 /* gsquery <- mark */
1440                                 if (mark)
1441                                         ma->mca_flags |= MAF_GSQUERY;
1442                                 else
1443                                         ma->mca_flags &= ~MAF_GSQUERY;
1444                         }
1445                         if (!(ma->mca_flags & MAF_GSQUERY) ||
1446                             mld_marksources(ma, ntohs(mlh2->mld2q_nsrcs), mlh2->mld2q_srcs))
1447                                 igmp6_group_queried(ma, max_delay);
1448                         spin_unlock_bh(&ma->mca_lock);
1449                         break;
1450                 }
1451         }
1452         read_unlock_bh(&idev->lock);
1453
1454         return 0;
1455 }
1456
1457 /* called with rcu_read_lock() */
1458 int igmp6_event_report(struct sk_buff *skb)
1459 {
1460         struct ifmcaddr6 *ma;
1461         struct inet6_dev *idev;
1462         struct mld_msg *mld;
1463         int addr_type;
1464
1465         /* Our own report looped back. Ignore it. */
1466         if (skb->pkt_type == PACKET_LOOPBACK)
1467                 return 0;
1468
1469         /* send our report if the MC router may not have heard this report */
1470         if (skb->pkt_type != PACKET_MULTICAST &&
1471             skb->pkt_type != PACKET_BROADCAST)
1472                 return 0;
1473
1474         if (!pskb_may_pull(skb, sizeof(*mld) - sizeof(struct icmp6hdr)))
1475                 return -EINVAL;
1476
1477         mld = (struct mld_msg *)icmp6_hdr(skb);
1478
1479         /* Drop reports with not link local source */
1480         addr_type = ipv6_addr_type(&ipv6_hdr(skb)->saddr);
1481         if (addr_type != IPV6_ADDR_ANY &&
1482             !(addr_type&IPV6_ADDR_LINKLOCAL))
1483                 return -EINVAL;
1484
1485         idev = __in6_dev_get(skb->dev);
1486         if (!idev)
1487                 return -ENODEV;
1488
1489         /*
1490          *      Cancel the timer for this group
1491          */
1492
1493         read_lock_bh(&idev->lock);
1494         for (ma = idev->mc_list; ma; ma = ma->next) {
1495                 if (ipv6_addr_equal(&ma->mca_addr, &mld->mld_mca)) {
1496                         spin_lock(&ma->mca_lock);
1497                         if (del_timer(&ma->mca_timer))
1498                                 refcount_dec(&ma->mca_refcnt);
1499                         ma->mca_flags &= ~(MAF_LAST_REPORTER|MAF_TIMER_RUNNING);
1500                         spin_unlock(&ma->mca_lock);
1501                         break;
1502                 }
1503         }
1504         read_unlock_bh(&idev->lock);
1505         return 0;
1506 }
1507
1508 static bool is_in(struct ifmcaddr6 *pmc, struct ip6_sf_list *psf, int type,
1509                   int gdeleted, int sdeleted)
1510 {
1511         switch (type) {
1512         case MLD2_MODE_IS_INCLUDE:
1513         case MLD2_MODE_IS_EXCLUDE:
1514                 if (gdeleted || sdeleted)
1515                         return false;
1516                 if (!((pmc->mca_flags & MAF_GSQUERY) && !psf->sf_gsresp)) {
1517                         if (pmc->mca_sfmode == MCAST_INCLUDE)
1518                                 return true;
1519                         /* don't include if this source is excluded
1520                          * in all filters
1521                          */
1522                         if (psf->sf_count[MCAST_INCLUDE])
1523                                 return type == MLD2_MODE_IS_INCLUDE;
1524                         return pmc->mca_sfcount[MCAST_EXCLUDE] ==
1525                                 psf->sf_count[MCAST_EXCLUDE];
1526                 }
1527                 return false;
1528         case MLD2_CHANGE_TO_INCLUDE:
1529                 if (gdeleted || sdeleted)
1530                         return false;
1531                 return psf->sf_count[MCAST_INCLUDE] != 0;
1532         case MLD2_CHANGE_TO_EXCLUDE:
1533                 if (gdeleted || sdeleted)
1534                         return false;
1535                 if (pmc->mca_sfcount[MCAST_EXCLUDE] == 0 ||
1536                     psf->sf_count[MCAST_INCLUDE])
1537                         return false;
1538                 return pmc->mca_sfcount[MCAST_EXCLUDE] ==
1539                         psf->sf_count[MCAST_EXCLUDE];
1540         case MLD2_ALLOW_NEW_SOURCES:
1541                 if (gdeleted || !psf->sf_crcount)
1542                         return false;
1543                 return (pmc->mca_sfmode == MCAST_INCLUDE) ^ sdeleted;
1544         case MLD2_BLOCK_OLD_SOURCES:
1545                 if (pmc->mca_sfmode == MCAST_INCLUDE)
1546                         return gdeleted || (psf->sf_crcount && sdeleted);
1547                 return psf->sf_crcount && !gdeleted && !sdeleted;
1548         }
1549         return false;
1550 }
1551
1552 static int
1553 mld_scount(struct ifmcaddr6 *pmc, int type, int gdeleted, int sdeleted)
1554 {
1555         struct ip6_sf_list *psf;
1556         int scount = 0;
1557
1558         for (psf = pmc->mca_sources; psf; psf = psf->sf_next) {
1559                 if (!is_in(pmc, psf, type, gdeleted, sdeleted))
1560                         continue;
1561                 scount++;
1562         }
1563         return scount;
1564 }
1565
1566 static void ip6_mc_hdr(struct sock *sk, struct sk_buff *skb,
1567                        struct net_device *dev,
1568                        const struct in6_addr *saddr,
1569                        const struct in6_addr *daddr,
1570                        int proto, int len)
1571 {
1572         struct ipv6hdr *hdr;
1573
1574         skb->protocol = htons(ETH_P_IPV6);
1575         skb->dev = dev;
1576
1577         skb_reset_network_header(skb);
1578         skb_put(skb, sizeof(struct ipv6hdr));
1579         hdr = ipv6_hdr(skb);
1580
1581         ip6_flow_hdr(hdr, 0, 0);
1582
1583         hdr->payload_len = htons(len);
1584         hdr->nexthdr = proto;
1585         hdr->hop_limit = inet6_sk(sk)->hop_limit;
1586
1587         hdr->saddr = *saddr;
1588         hdr->daddr = *daddr;
1589 }
1590
1591 static struct sk_buff *mld_newpack(struct inet6_dev *idev, unsigned int mtu)
1592 {
1593         struct net_device *dev = idev->dev;
1594         struct net *net = dev_net(dev);
1595         struct sock *sk = net->ipv6.igmp_sk;
1596         struct sk_buff *skb;
1597         struct mld2_report *pmr;
1598         struct in6_addr addr_buf;
1599         const struct in6_addr *saddr;
1600         int hlen = LL_RESERVED_SPACE(dev);
1601         int tlen = dev->needed_tailroom;
1602         unsigned int size = mtu + hlen + tlen;
1603         int err;
1604         u8 ra[8] = { IPPROTO_ICMPV6, 0,
1605                      IPV6_TLV_ROUTERALERT, 2, 0, 0,
1606                      IPV6_TLV_PADN, 0 };
1607
1608         /* we assume size > sizeof(ra) here */
1609         skb = sock_alloc_send_skb(sk, size, 1, &err);
1610         if (!skb)
1611                 return NULL;
1612
1613         skb->priority = TC_PRIO_CONTROL;
1614         skb_reserve(skb, hlen);
1615         skb_tailroom_reserve(skb, mtu, tlen);
1616
1617         if (__ipv6_get_lladdr(idev, &addr_buf, IFA_F_TENTATIVE)) {
1618                 /* <draft-ietf-magma-mld-source-05.txt>:
1619                  * use unspecified address as the source address
1620                  * when a valid link-local address is not available.
1621                  */
1622                 saddr = &in6addr_any;
1623         } else
1624                 saddr = &addr_buf;
1625
1626         ip6_mc_hdr(sk, skb, dev, saddr, &mld2_all_mcr, NEXTHDR_HOP, 0);
1627
1628         skb_put_data(skb, ra, sizeof(ra));
1629
1630         skb_set_transport_header(skb, skb_tail_pointer(skb) - skb->data);
1631         skb_put(skb, sizeof(*pmr));
1632         pmr = (struct mld2_report *)skb_transport_header(skb);
1633         pmr->mld2r_type = ICMPV6_MLD2_REPORT;
1634         pmr->mld2r_resv1 = 0;
1635         pmr->mld2r_cksum = 0;
1636         pmr->mld2r_resv2 = 0;
1637         pmr->mld2r_ngrec = 0;
1638         return skb;
1639 }
1640
1641 static void mld_sendpack(struct sk_buff *skb)
1642 {
1643         struct ipv6hdr *pip6 = ipv6_hdr(skb);
1644         struct mld2_report *pmr =
1645                               (struct mld2_report *)skb_transport_header(skb);
1646         int payload_len, mldlen;
1647         struct inet6_dev *idev;
1648         struct net *net = dev_net(skb->dev);
1649         int err;
1650         struct flowi6 fl6;
1651         struct dst_entry *dst;
1652
1653         rcu_read_lock();
1654         idev = __in6_dev_get(skb->dev);
1655         IP6_UPD_PO_STATS(net, idev, IPSTATS_MIB_OUT, skb->len);
1656
1657         payload_len = (skb_tail_pointer(skb) - skb_network_header(skb)) -
1658                 sizeof(*pip6);
1659         mldlen = skb_tail_pointer(skb) - skb_transport_header(skb);
1660         pip6->payload_len = htons(payload_len);
1661
1662         pmr->mld2r_cksum = csum_ipv6_magic(&pip6->saddr, &pip6->daddr, mldlen,
1663                                            IPPROTO_ICMPV6,
1664                                            csum_partial(skb_transport_header(skb),
1665                                                         mldlen, 0));
1666
1667         icmpv6_flow_init(net->ipv6.igmp_sk, &fl6, ICMPV6_MLD2_REPORT,
1668                          &ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr,
1669                          skb->dev->ifindex);
1670         dst = icmp6_dst_alloc(skb->dev, &fl6);
1671
1672         err = 0;
1673         if (IS_ERR(dst)) {
1674                 err = PTR_ERR(dst);
1675                 dst = NULL;
1676         }
1677         skb_dst_set(skb, dst);
1678         if (err)
1679                 goto err_out;
1680
1681         err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT,
1682                       net, net->ipv6.igmp_sk, skb, NULL, skb->dev,
1683                       dst_output);
1684 out:
1685         if (!err) {
1686                 ICMP6MSGOUT_INC_STATS(net, idev, ICMPV6_MLD2_REPORT);
1687                 ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTMSGS);
1688         } else {
1689                 IP6_INC_STATS(net, idev, IPSTATS_MIB_OUTDISCARDS);
1690         }
1691
1692         rcu_read_unlock();
1693         return;
1694
1695 err_out:
1696         kfree_skb(skb);
1697         goto out;
1698 }
1699
1700 static int grec_size(struct ifmcaddr6 *pmc, int type, int gdel, int sdel)
1701 {
1702         return sizeof(struct mld2_grec) + 16 * mld_scount(pmc,type,gdel,sdel);
1703 }
1704
1705 static struct sk_buff *add_grhead(struct sk_buff *skb, struct ifmcaddr6 *pmc,
1706         int type, struct mld2_grec **ppgr, unsigned int mtu)
1707 {
1708         struct mld2_report *pmr;
1709         struct mld2_grec *pgr;
1710
1711         if (!skb) {
1712                 skb = mld_newpack(pmc->idev, mtu);
1713                 if (!skb)
1714                         return NULL;
1715         }
1716         pgr = skb_put(skb, sizeof(struct mld2_grec));
1717         pgr->grec_type = type;
1718         pgr->grec_auxwords = 0;
1719         pgr->grec_nsrcs = 0;
1720         pgr->grec_mca = pmc->mca_addr;  /* structure copy */
1721         pmr = (struct mld2_report *)skb_transport_header(skb);
1722         pmr->mld2r_ngrec = htons(ntohs(pmr->mld2r_ngrec)+1);
1723         *ppgr = pgr;
1724         return skb;
1725 }
1726
1727 #define AVAILABLE(skb)  ((skb) ? skb_availroom(skb) : 0)
1728
1729 static struct sk_buff *add_grec(struct sk_buff *skb, struct ifmcaddr6 *pmc,
1730         int type, int gdeleted, int sdeleted, int crsend)
1731 {
1732         struct inet6_dev *idev = pmc->idev;
1733         struct net_device *dev = idev->dev;
1734         struct mld2_report *pmr;
1735         struct mld2_grec *pgr = NULL;
1736         struct ip6_sf_list *psf, *psf_next, *psf_prev, **psf_list;
1737         int scount, stotal, first, isquery, truncate;
1738         unsigned int mtu;
1739
1740         if (pmc->mca_flags & MAF_NOREPORT)
1741                 return skb;
1742
1743         mtu = READ_ONCE(dev->mtu);
1744         if (mtu < IPV6_MIN_MTU)
1745                 return skb;
1746
1747         isquery = type == MLD2_MODE_IS_INCLUDE ||
1748                   type == MLD2_MODE_IS_EXCLUDE;
1749         truncate = type == MLD2_MODE_IS_EXCLUDE ||
1750                     type == MLD2_CHANGE_TO_EXCLUDE;
1751
1752         stotal = scount = 0;
1753
1754         psf_list = sdeleted ? &pmc->mca_tomb : &pmc->mca_sources;
1755
1756         if (!*psf_list)
1757                 goto empty_source;
1758
1759         pmr = skb ? (struct mld2_report *)skb_transport_header(skb) : NULL;
1760
1761         /* EX and TO_EX get a fresh packet, if needed */
1762         if (truncate) {
1763                 if (pmr && pmr->mld2r_ngrec &&
1764                     AVAILABLE(skb) < grec_size(pmc, type, gdeleted, sdeleted)) {
1765                         if (skb)
1766                                 mld_sendpack(skb);
1767                         skb = mld_newpack(idev, mtu);
1768                 }
1769         }
1770         first = 1;
1771         psf_prev = NULL;
1772         for (psf = *psf_list; psf; psf = psf_next) {
1773                 struct in6_addr *psrc;
1774
1775                 psf_next = psf->sf_next;
1776
1777                 if (!is_in(pmc, psf, type, gdeleted, sdeleted) && !crsend) {
1778                         psf_prev = psf;
1779                         continue;
1780                 }
1781
1782                 /* Based on RFC3810 6.1. Should not send source-list change
1783                  * records when there is a filter mode change.
1784                  */
1785                 if (((gdeleted && pmc->mca_sfmode == MCAST_EXCLUDE) ||
1786                      (!gdeleted && pmc->mca_crcount)) &&
1787                     (type == MLD2_ALLOW_NEW_SOURCES ||
1788                      type == MLD2_BLOCK_OLD_SOURCES) && psf->sf_crcount)
1789                         goto decrease_sf_crcount;
1790
1791                 /* clear marks on query responses */
1792                 if (isquery)
1793                         psf->sf_gsresp = 0;
1794
1795                 if (AVAILABLE(skb) < sizeof(*psrc) +
1796                     first*sizeof(struct mld2_grec)) {
1797                         if (truncate && !first)
1798                                 break;   /* truncate these */
1799                         if (pgr)
1800                                 pgr->grec_nsrcs = htons(scount);
1801                         if (skb)
1802                                 mld_sendpack(skb);
1803                         skb = mld_newpack(idev, mtu);
1804                         first = 1;
1805                         scount = 0;
1806                 }
1807                 if (first) {
1808                         skb = add_grhead(skb, pmc, type, &pgr, mtu);
1809                         first = 0;
1810                 }
1811                 if (!skb)
1812                         return NULL;
1813                 psrc = skb_put(skb, sizeof(*psrc));
1814                 *psrc = psf->sf_addr;
1815                 scount++; stotal++;
1816                 if ((type == MLD2_ALLOW_NEW_SOURCES ||
1817                      type == MLD2_BLOCK_OLD_SOURCES) && psf->sf_crcount) {
1818 decrease_sf_crcount:
1819                         psf->sf_crcount--;
1820                         if ((sdeleted || gdeleted) && psf->sf_crcount == 0) {
1821                                 if (psf_prev)
1822                                         psf_prev->sf_next = psf->sf_next;
1823                                 else
1824                                         *psf_list = psf->sf_next;
1825                                 kfree(psf);
1826                                 continue;
1827                         }
1828                 }
1829                 psf_prev = psf;
1830         }
1831
1832 empty_source:
1833         if (!stotal) {
1834                 if (type == MLD2_ALLOW_NEW_SOURCES ||
1835                     type == MLD2_BLOCK_OLD_SOURCES)
1836                         return skb;
1837                 if (pmc->mca_crcount || isquery || crsend) {
1838                         /* make sure we have room for group header */
1839                         if (skb && AVAILABLE(skb) < sizeof(struct mld2_grec)) {
1840                                 mld_sendpack(skb);
1841                                 skb = NULL; /* add_grhead will get a new one */
1842                         }
1843                         skb = add_grhead(skb, pmc, type, &pgr, mtu);
1844                 }
1845         }
1846         if (pgr)
1847                 pgr->grec_nsrcs = htons(scount);
1848
1849         if (isquery)
1850                 pmc->mca_flags &= ~MAF_GSQUERY; /* clear query state */
1851         return skb;
1852 }
1853
1854 static void mld_send_report(struct inet6_dev *idev, struct ifmcaddr6 *pmc)
1855 {
1856         struct sk_buff *skb = NULL;
1857         int type;
1858
1859         read_lock_bh(&idev->lock);
1860         if (!pmc) {
1861                 for (pmc = idev->mc_list; pmc; pmc = pmc->next) {
1862                         if (pmc->mca_flags & MAF_NOREPORT)
1863                                 continue;
1864                         spin_lock_bh(&pmc->mca_lock);
1865                         if (pmc->mca_sfcount[MCAST_EXCLUDE])
1866                                 type = MLD2_MODE_IS_EXCLUDE;
1867                         else
1868                                 type = MLD2_MODE_IS_INCLUDE;
1869                         skb = add_grec(skb, pmc, type, 0, 0, 0);
1870                         spin_unlock_bh(&pmc->mca_lock);
1871                 }
1872         } else {
1873                 spin_lock_bh(&pmc->mca_lock);
1874                 if (pmc->mca_sfcount[MCAST_EXCLUDE])
1875                         type = MLD2_MODE_IS_EXCLUDE;
1876                 else
1877                         type = MLD2_MODE_IS_INCLUDE;
1878                 skb = add_grec(skb, pmc, type, 0, 0, 0);
1879                 spin_unlock_bh(&pmc->mca_lock);
1880         }
1881         read_unlock_bh(&idev->lock);
1882         if (skb)
1883                 mld_sendpack(skb);
1884 }
1885
1886 /*
1887  * remove zero-count source records from a source filter list
1888  */
1889 static void mld_clear_zeros(struct ip6_sf_list **ppsf)
1890 {
1891         struct ip6_sf_list *psf_prev, *psf_next, *psf;
1892
1893         psf_prev = NULL;
1894         for (psf = *ppsf; psf; psf = psf_next) {
1895                 psf_next = psf->sf_next;
1896                 if (psf->sf_crcount == 0) {
1897                         if (psf_prev)
1898                                 psf_prev->sf_next = psf->sf_next;
1899                         else
1900                                 *ppsf = psf->sf_next;
1901                         kfree(psf);
1902                 } else
1903                         psf_prev = psf;
1904         }
1905 }
1906
1907 static void mld_send_cr(struct inet6_dev *idev)
1908 {
1909         struct ifmcaddr6 *pmc, *pmc_prev, *pmc_next;
1910         struct sk_buff *skb = NULL;
1911         int type, dtype;
1912
1913         read_lock_bh(&idev->lock);
1914         spin_lock(&idev->mc_lock);
1915
1916         /* deleted MCA's */
1917         pmc_prev = NULL;
1918         for (pmc = idev->mc_tomb; pmc; pmc = pmc_next) {
1919                 pmc_next = pmc->next;
1920                 if (pmc->mca_sfmode == MCAST_INCLUDE) {
1921                         type = MLD2_BLOCK_OLD_SOURCES;
1922                         dtype = MLD2_BLOCK_OLD_SOURCES;
1923                         skb = add_grec(skb, pmc, type, 1, 0, 0);
1924                         skb = add_grec(skb, pmc, dtype, 1, 1, 0);
1925                 }
1926                 if (pmc->mca_crcount) {
1927                         if (pmc->mca_sfmode == MCAST_EXCLUDE) {
1928                                 type = MLD2_CHANGE_TO_INCLUDE;
1929                                 skb = add_grec(skb, pmc, type, 1, 0, 0);
1930                         }
1931                         pmc->mca_crcount--;
1932                         if (pmc->mca_crcount == 0) {
1933                                 mld_clear_zeros(&pmc->mca_tomb);
1934                                 mld_clear_zeros(&pmc->mca_sources);
1935                         }
1936                 }
1937                 if (pmc->mca_crcount == 0 && !pmc->mca_tomb &&
1938                     !pmc->mca_sources) {
1939                         if (pmc_prev)
1940                                 pmc_prev->next = pmc_next;
1941                         else
1942                                 idev->mc_tomb = pmc_next;
1943                         in6_dev_put(pmc->idev);
1944                         kfree(pmc);
1945                 } else
1946                         pmc_prev = pmc;
1947         }
1948         spin_unlock(&idev->mc_lock);
1949
1950         /* change recs */
1951         for (pmc = idev->mc_list; pmc; pmc = pmc->next) {
1952                 spin_lock_bh(&pmc->mca_lock);
1953                 if (pmc->mca_sfcount[MCAST_EXCLUDE]) {
1954                         type = MLD2_BLOCK_OLD_SOURCES;
1955                         dtype = MLD2_ALLOW_NEW_SOURCES;
1956                 } else {
1957                         type = MLD2_ALLOW_NEW_SOURCES;
1958                         dtype = MLD2_BLOCK_OLD_SOURCES;
1959                 }
1960                 skb = add_grec(skb, pmc, type, 0, 0, 0);
1961                 skb = add_grec(skb, pmc, dtype, 0, 1, 0);       /* deleted sources */
1962
1963                 /* filter mode changes */
1964                 if (pmc->mca_crcount) {
1965                         if (pmc->mca_sfmode == MCAST_EXCLUDE)
1966                                 type = MLD2_CHANGE_TO_EXCLUDE;
1967                         else
1968                                 type = MLD2_CHANGE_TO_INCLUDE;
1969                         skb = add_grec(skb, pmc, type, 0, 0, 0);
1970                         pmc->mca_crcount--;
1971                 }
1972                 spin_unlock_bh(&pmc->mca_lock);
1973         }
1974         read_unlock_bh(&idev->lock);
1975         if (!skb)
1976                 return;
1977         (void) mld_sendpack(skb);
1978 }
1979
1980 static void igmp6_send(struct in6_addr *addr, struct net_device *dev, int type)
1981 {
1982         struct net *net = dev_net(dev);
1983         struct sock *sk = net->ipv6.igmp_sk;
1984         struct inet6_dev *idev;
1985         struct sk_buff *skb;
1986         struct mld_msg *hdr;
1987         const struct in6_addr *snd_addr, *saddr;
1988         struct in6_addr addr_buf;
1989         int hlen = LL_RESERVED_SPACE(dev);
1990         int tlen = dev->needed_tailroom;
1991         int err, len, payload_len, full_len;
1992         u8 ra[8] = { IPPROTO_ICMPV6, 0,
1993                      IPV6_TLV_ROUTERALERT, 2, 0, 0,
1994                      IPV6_TLV_PADN, 0 };
1995         struct flowi6 fl6;
1996         struct dst_entry *dst;
1997
1998         if (type == ICMPV6_MGM_REDUCTION)
1999                 snd_addr = &in6addr_linklocal_allrouters;
2000         else
2001                 snd_addr = addr;
2002
2003         len = sizeof(struct icmp6hdr) + sizeof(struct in6_addr);
2004         payload_len = len + sizeof(ra);
2005         full_len = sizeof(struct ipv6hdr) + payload_len;
2006
2007         rcu_read_lock();
2008         IP6_UPD_PO_STATS(net, __in6_dev_get(dev),
2009                       IPSTATS_MIB_OUT, full_len);
2010         rcu_read_unlock();
2011
2012         skb = sock_alloc_send_skb(sk, hlen + tlen + full_len, 1, &err);
2013
2014         if (!skb) {
2015                 rcu_read_lock();
2016                 IP6_INC_STATS(net, __in6_dev_get(dev),
2017                               IPSTATS_MIB_OUTDISCARDS);
2018                 rcu_read_unlock();
2019                 return;
2020         }
2021         skb->priority = TC_PRIO_CONTROL;
2022         skb_reserve(skb, hlen);
2023
2024         if (ipv6_get_lladdr(dev, &addr_buf, IFA_F_TENTATIVE)) {
2025                 /* <draft-ietf-magma-mld-source-05.txt>:
2026                  * use unspecified address as the source address
2027                  * when a valid link-local address is not available.
2028                  */
2029                 saddr = &in6addr_any;
2030         } else
2031                 saddr = &addr_buf;
2032
2033         ip6_mc_hdr(sk, skb, dev, saddr, snd_addr, NEXTHDR_HOP, payload_len);
2034
2035         skb_put_data(skb, ra, sizeof(ra));
2036
2037         hdr = skb_put_zero(skb, sizeof(struct mld_msg));
2038         hdr->mld_type = type;
2039         hdr->mld_mca = *addr;
2040
2041         hdr->mld_cksum = csum_ipv6_magic(saddr, snd_addr, len,
2042                                          IPPROTO_ICMPV6,
2043                                          csum_partial(hdr, len, 0));
2044
2045         rcu_read_lock();
2046         idev = __in6_dev_get(skb->dev);
2047
2048         icmpv6_flow_init(sk, &fl6, type,
2049                          &ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr,
2050                          skb->dev->ifindex);
2051         dst = icmp6_dst_alloc(skb->dev, &fl6);
2052         if (IS_ERR(dst)) {
2053                 err = PTR_ERR(dst);
2054                 goto err_out;
2055         }
2056
2057         skb_dst_set(skb, dst);
2058         err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT,
2059                       net, sk, skb, NULL, skb->dev,
2060                       dst_output);
2061 out:
2062         if (!err) {
2063                 ICMP6MSGOUT_INC_STATS(net, idev, type);
2064                 ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTMSGS);
2065         } else
2066                 IP6_INC_STATS(net, idev, IPSTATS_MIB_OUTDISCARDS);
2067
2068         rcu_read_unlock();
2069         return;
2070
2071 err_out:
2072         kfree_skb(skb);
2073         goto out;
2074 }
2075
2076 static void mld_send_initial_cr(struct inet6_dev *idev)
2077 {
2078         struct sk_buff *skb;
2079         struct ifmcaddr6 *pmc;
2080         int type;
2081
2082         if (mld_in_v1_mode(idev))
2083                 return;
2084
2085         skb = NULL;
2086         read_lock_bh(&idev->lock);
2087         for (pmc = idev->mc_list; pmc; pmc = pmc->next) {
2088                 spin_lock_bh(&pmc->mca_lock);
2089                 if (pmc->mca_sfcount[MCAST_EXCLUDE])
2090                         type = MLD2_CHANGE_TO_EXCLUDE;
2091                 else
2092                         type = MLD2_ALLOW_NEW_SOURCES;
2093                 skb = add_grec(skb, pmc, type, 0, 0, 1);
2094                 spin_unlock_bh(&pmc->mca_lock);
2095         }
2096         read_unlock_bh(&idev->lock);
2097         if (skb)
2098                 mld_sendpack(skb);
2099 }
2100
2101 void ipv6_mc_dad_complete(struct inet6_dev *idev)
2102 {
2103         idev->mc_dad_count = idev->mc_qrv;
2104         if (idev->mc_dad_count) {
2105                 mld_send_initial_cr(idev);
2106                 idev->mc_dad_count--;
2107                 if (idev->mc_dad_count)
2108                         mld_dad_start_timer(idev,
2109                                             unsolicited_report_interval(idev));
2110         }
2111 }
2112
2113 static void mld_dad_timer_expire(struct timer_list *t)
2114 {
2115         struct inet6_dev *idev = from_timer(idev, t, mc_dad_timer);
2116
2117         mld_send_initial_cr(idev);
2118         if (idev->mc_dad_count) {
2119                 idev->mc_dad_count--;
2120                 if (idev->mc_dad_count)
2121                         mld_dad_start_timer(idev,
2122                                             unsolicited_report_interval(idev));
2123         }
2124         in6_dev_put(idev);
2125 }
2126
2127 static int ip6_mc_del1_src(struct ifmcaddr6 *pmc, int sfmode,
2128         const struct in6_addr *psfsrc)
2129 {
2130         struct ip6_sf_list *psf, *psf_prev;
2131         int rv = 0;
2132
2133         psf_prev = NULL;
2134         for (psf = pmc->mca_sources; psf; psf = psf->sf_next) {
2135                 if (ipv6_addr_equal(&psf->sf_addr, psfsrc))
2136                         break;
2137                 psf_prev = psf;
2138         }
2139         if (!psf || psf->sf_count[sfmode] == 0) {
2140                 /* source filter not found, or count wrong =>  bug */
2141                 return -ESRCH;
2142         }
2143         psf->sf_count[sfmode]--;
2144         if (!psf->sf_count[MCAST_INCLUDE] && !psf->sf_count[MCAST_EXCLUDE]) {
2145                 struct inet6_dev *idev = pmc->idev;
2146
2147                 /* no more filters for this source */
2148                 if (psf_prev)
2149                         psf_prev->sf_next = psf->sf_next;
2150                 else
2151                         pmc->mca_sources = psf->sf_next;
2152                 if (psf->sf_oldin && !(pmc->mca_flags & MAF_NOREPORT) &&
2153                     !mld_in_v1_mode(idev)) {
2154                         psf->sf_crcount = idev->mc_qrv;
2155                         psf->sf_next = pmc->mca_tomb;
2156                         pmc->mca_tomb = psf;
2157                         rv = 1;
2158                 } else
2159                         kfree(psf);
2160         }
2161         return rv;
2162 }
2163
2164 static int ip6_mc_del_src(struct inet6_dev *idev, const struct in6_addr *pmca,
2165                           int sfmode, int sfcount, const struct in6_addr *psfsrc,
2166                           int delta)
2167 {
2168         struct ifmcaddr6 *pmc;
2169         int     changerec = 0;
2170         int     i, err;
2171
2172         if (!idev)
2173                 return -ENODEV;
2174         read_lock_bh(&idev->lock);
2175         for (pmc = idev->mc_list; pmc; pmc = pmc->next) {
2176                 if (ipv6_addr_equal(pmca, &pmc->mca_addr))
2177                         break;
2178         }
2179         if (!pmc) {
2180                 /* MCA not found?? bug */
2181                 read_unlock_bh(&idev->lock);
2182                 return -ESRCH;
2183         }
2184         spin_lock_bh(&pmc->mca_lock);
2185         sf_markstate(pmc);
2186         if (!delta) {
2187                 if (!pmc->mca_sfcount[sfmode]) {
2188                         spin_unlock_bh(&pmc->mca_lock);
2189                         read_unlock_bh(&idev->lock);
2190                         return -EINVAL;
2191                 }
2192                 pmc->mca_sfcount[sfmode]--;
2193         }
2194         err = 0;
2195         for (i = 0; i < sfcount; i++) {
2196                 int rv = ip6_mc_del1_src(pmc, sfmode, &psfsrc[i]);
2197
2198                 changerec |= rv > 0;
2199                 if (!err && rv < 0)
2200                         err = rv;
2201         }
2202         if (pmc->mca_sfmode == MCAST_EXCLUDE &&
2203             pmc->mca_sfcount[MCAST_EXCLUDE] == 0 &&
2204             pmc->mca_sfcount[MCAST_INCLUDE]) {
2205                 struct ip6_sf_list *psf;
2206
2207                 /* filter mode change */
2208                 pmc->mca_sfmode = MCAST_INCLUDE;
2209                 pmc->mca_crcount = idev->mc_qrv;
2210                 idev->mc_ifc_count = pmc->mca_crcount;
2211                 for (psf = pmc->mca_sources; psf; psf = psf->sf_next)
2212                         psf->sf_crcount = 0;
2213                 mld_ifc_event(pmc->idev);
2214         } else if (sf_setstate(pmc) || changerec)
2215                 mld_ifc_event(pmc->idev);
2216         spin_unlock_bh(&pmc->mca_lock);
2217         read_unlock_bh(&idev->lock);
2218         return err;
2219 }
2220
2221 /*
2222  * Add multicast single-source filter to the interface list
2223  */
2224 static int ip6_mc_add1_src(struct ifmcaddr6 *pmc, int sfmode,
2225         const struct in6_addr *psfsrc)
2226 {
2227         struct ip6_sf_list *psf, *psf_prev;
2228
2229         psf_prev = NULL;
2230         for (psf = pmc->mca_sources; psf; psf = psf->sf_next) {
2231                 if (ipv6_addr_equal(&psf->sf_addr, psfsrc))
2232                         break;
2233                 psf_prev = psf;
2234         }
2235         if (!psf) {
2236                 psf = kzalloc(sizeof(*psf), GFP_ATOMIC);
2237                 if (!psf)
2238                         return -ENOBUFS;
2239
2240                 psf->sf_addr = *psfsrc;
2241                 if (psf_prev) {
2242                         psf_prev->sf_next = psf;
2243                 } else
2244                         pmc->mca_sources = psf;
2245         }
2246         psf->sf_count[sfmode]++;
2247         return 0;
2248 }
2249
2250 static void sf_markstate(struct ifmcaddr6 *pmc)
2251 {
2252         struct ip6_sf_list *psf;
2253         int mca_xcount = pmc->mca_sfcount[MCAST_EXCLUDE];
2254
2255         for (psf = pmc->mca_sources; psf; psf = psf->sf_next)
2256                 if (pmc->mca_sfcount[MCAST_EXCLUDE]) {
2257                         psf->sf_oldin = mca_xcount ==
2258                                 psf->sf_count[MCAST_EXCLUDE] &&
2259                                 !psf->sf_count[MCAST_INCLUDE];
2260                 } else
2261                         psf->sf_oldin = psf->sf_count[MCAST_INCLUDE] != 0;
2262 }
2263
2264 static int sf_setstate(struct ifmcaddr6 *pmc)
2265 {
2266         struct ip6_sf_list *psf, *dpsf;
2267         int mca_xcount = pmc->mca_sfcount[MCAST_EXCLUDE];
2268         int qrv = pmc->idev->mc_qrv;
2269         int new_in, rv;
2270
2271         rv = 0;
2272         for (psf = pmc->mca_sources; psf; psf = psf->sf_next) {
2273                 if (pmc->mca_sfcount[MCAST_EXCLUDE]) {
2274                         new_in = mca_xcount == psf->sf_count[MCAST_EXCLUDE] &&
2275                                 !psf->sf_count[MCAST_INCLUDE];
2276                 } else
2277                         new_in = psf->sf_count[MCAST_INCLUDE] != 0;
2278                 if (new_in) {
2279                         if (!psf->sf_oldin) {
2280                                 struct ip6_sf_list *prev = NULL;
2281
2282                                 for (dpsf = pmc->mca_tomb; dpsf;
2283                                      dpsf = dpsf->sf_next) {
2284                                         if (ipv6_addr_equal(&dpsf->sf_addr,
2285                                             &psf->sf_addr))
2286                                                 break;
2287                                         prev = dpsf;
2288                                 }
2289                                 if (dpsf) {
2290                                         if (prev)
2291                                                 prev->sf_next = dpsf->sf_next;
2292                                         else
2293                                                 pmc->mca_tomb = dpsf->sf_next;
2294                                         kfree(dpsf);
2295                                 }
2296                                 psf->sf_crcount = qrv;
2297                                 rv++;
2298                         }
2299                 } else if (psf->sf_oldin) {
2300                         psf->sf_crcount = 0;
2301                         /*
2302                          * add or update "delete" records if an active filter
2303                          * is now inactive
2304                          */
2305                         for (dpsf = pmc->mca_tomb; dpsf; dpsf = dpsf->sf_next)
2306                                 if (ipv6_addr_equal(&dpsf->sf_addr,
2307                                     &psf->sf_addr))
2308                                         break;
2309                         if (!dpsf) {
2310                                 dpsf = kmalloc(sizeof(*dpsf), GFP_ATOMIC);
2311                                 if (!dpsf)
2312                                         continue;
2313                                 *dpsf = *psf;
2314                                 /* pmc->mca_lock held by callers */
2315                                 dpsf->sf_next = pmc->mca_tomb;
2316                                 pmc->mca_tomb = dpsf;
2317                         }
2318                         dpsf->sf_crcount = qrv;
2319                         rv++;
2320                 }
2321         }
2322         return rv;
2323 }
2324
2325 /*
2326  * Add multicast source filter list to the interface list
2327  */
2328 static int ip6_mc_add_src(struct inet6_dev *idev, const struct in6_addr *pmca,
2329                           int sfmode, int sfcount, const struct in6_addr *psfsrc,
2330                           int delta)
2331 {
2332         struct ifmcaddr6 *pmc;
2333         int     isexclude;
2334         int     i, err;
2335
2336         if (!idev)
2337                 return -ENODEV;
2338         read_lock_bh(&idev->lock);
2339         for (pmc = idev->mc_list; pmc; pmc = pmc->next) {
2340                 if (ipv6_addr_equal(pmca, &pmc->mca_addr))
2341                         break;
2342         }
2343         if (!pmc) {
2344                 /* MCA not found?? bug */
2345                 read_unlock_bh(&idev->lock);
2346                 return -ESRCH;
2347         }
2348         spin_lock_bh(&pmc->mca_lock);
2349
2350         sf_markstate(pmc);
2351         isexclude = pmc->mca_sfmode == MCAST_EXCLUDE;
2352         if (!delta)
2353                 pmc->mca_sfcount[sfmode]++;
2354         err = 0;
2355         for (i = 0; i < sfcount; i++) {
2356                 err = ip6_mc_add1_src(pmc, sfmode, &psfsrc[i]);
2357                 if (err)
2358                         break;
2359         }
2360         if (err) {
2361                 int j;
2362
2363                 if (!delta)
2364                         pmc->mca_sfcount[sfmode]--;
2365                 for (j = 0; j < i; j++)
2366                         ip6_mc_del1_src(pmc, sfmode, &psfsrc[j]);
2367         } else if (isexclude != (pmc->mca_sfcount[MCAST_EXCLUDE] != 0)) {
2368                 struct ip6_sf_list *psf;
2369
2370                 /* filter mode change */
2371                 if (pmc->mca_sfcount[MCAST_EXCLUDE])
2372                         pmc->mca_sfmode = MCAST_EXCLUDE;
2373                 else if (pmc->mca_sfcount[MCAST_INCLUDE])
2374                         pmc->mca_sfmode = MCAST_INCLUDE;
2375                 /* else no filters; keep old mode for reports */
2376
2377                 pmc->mca_crcount = idev->mc_qrv;
2378                 idev->mc_ifc_count = pmc->mca_crcount;
2379                 for (psf = pmc->mca_sources; psf; psf = psf->sf_next)
2380                         psf->sf_crcount = 0;
2381                 mld_ifc_event(idev);
2382         } else if (sf_setstate(pmc))
2383                 mld_ifc_event(idev);
2384         spin_unlock_bh(&pmc->mca_lock);
2385         read_unlock_bh(&idev->lock);
2386         return err;
2387 }
2388
2389 static void ip6_mc_clear_src(struct ifmcaddr6 *pmc)
2390 {
2391         struct ip6_sf_list *psf, *nextpsf;
2392
2393         for (psf = pmc->mca_tomb; psf; psf = nextpsf) {
2394                 nextpsf = psf->sf_next;
2395                 kfree(psf);
2396         }
2397         pmc->mca_tomb = NULL;
2398         for (psf = pmc->mca_sources; psf; psf = nextpsf) {
2399                 nextpsf = psf->sf_next;
2400                 kfree(psf);
2401         }
2402         pmc->mca_sources = NULL;
2403         pmc->mca_sfmode = MCAST_EXCLUDE;
2404         pmc->mca_sfcount[MCAST_INCLUDE] = 0;
2405         pmc->mca_sfcount[MCAST_EXCLUDE] = 1;
2406 }
2407
2408
2409 static void igmp6_join_group(struct ifmcaddr6 *ma)
2410 {
2411         unsigned long delay;
2412
2413         if (ma->mca_flags & MAF_NOREPORT)
2414                 return;
2415
2416         igmp6_send(&ma->mca_addr, ma->idev->dev, ICMPV6_MGM_REPORT);
2417
2418         delay = prandom_u32() % unsolicited_report_interval(ma->idev);
2419
2420         spin_lock_bh(&ma->mca_lock);
2421         if (del_timer(&ma->mca_timer)) {
2422                 refcount_dec(&ma->mca_refcnt);
2423                 delay = ma->mca_timer.expires - jiffies;
2424         }
2425
2426         if (!mod_timer(&ma->mca_timer, jiffies + delay))
2427                 refcount_inc(&ma->mca_refcnt);
2428         ma->mca_flags |= MAF_TIMER_RUNNING | MAF_LAST_REPORTER;
2429         spin_unlock_bh(&ma->mca_lock);
2430 }
2431
2432 static int ip6_mc_leave_src(struct sock *sk, struct ipv6_mc_socklist *iml,
2433                             struct inet6_dev *idev)
2434 {
2435         int err;
2436
2437         write_lock_bh(&iml->sflock);
2438         if (!iml->sflist) {
2439                 /* any-source empty exclude case */
2440                 err = ip6_mc_del_src(idev, &iml->addr, iml->sfmode, 0, NULL, 0);
2441         } else {
2442                 err = ip6_mc_del_src(idev, &iml->addr, iml->sfmode,
2443                                 iml->sflist->sl_count, iml->sflist->sl_addr, 0);
2444                 sock_kfree_s(sk, iml->sflist, IP6_SFLSIZE(iml->sflist->sl_max));
2445                 iml->sflist = NULL;
2446         }
2447         write_unlock_bh(&iml->sflock);
2448         return err;
2449 }
2450
2451 static void igmp6_leave_group(struct ifmcaddr6 *ma)
2452 {
2453         if (mld_in_v1_mode(ma->idev)) {
2454                 if (ma->mca_flags & MAF_LAST_REPORTER)
2455                         igmp6_send(&ma->mca_addr, ma->idev->dev,
2456                                 ICMPV6_MGM_REDUCTION);
2457         } else {
2458                 mld_add_delrec(ma->idev, ma);
2459                 mld_ifc_event(ma->idev);
2460         }
2461 }
2462
2463 static void mld_gq_timer_expire(struct timer_list *t)
2464 {
2465         struct inet6_dev *idev = from_timer(idev, t, mc_gq_timer);
2466
2467         idev->mc_gq_running = 0;
2468         mld_send_report(idev, NULL);
2469         in6_dev_put(idev);
2470 }
2471
2472 static void mld_ifc_timer_expire(struct timer_list *t)
2473 {
2474         struct inet6_dev *idev = from_timer(idev, t, mc_ifc_timer);
2475
2476         mld_send_cr(idev);
2477         if (idev->mc_ifc_count) {
2478                 idev->mc_ifc_count--;
2479                 if (idev->mc_ifc_count)
2480                         mld_ifc_start_timer(idev,
2481                                             unsolicited_report_interval(idev));
2482         }
2483         in6_dev_put(idev);
2484 }
2485
2486 static void mld_ifc_event(struct inet6_dev *idev)
2487 {
2488         if (mld_in_v1_mode(idev))
2489                 return;
2490         idev->mc_ifc_count = idev->mc_qrv;
2491         mld_ifc_start_timer(idev, 1);
2492 }
2493
2494 static void igmp6_timer_handler(struct timer_list *t)
2495 {
2496         struct ifmcaddr6 *ma = from_timer(ma, t, mca_timer);
2497
2498         if (mld_in_v1_mode(ma->idev))
2499                 igmp6_send(&ma->mca_addr, ma->idev->dev, ICMPV6_MGM_REPORT);
2500         else
2501                 mld_send_report(ma->idev, ma);
2502
2503         spin_lock(&ma->mca_lock);
2504         ma->mca_flags |=  MAF_LAST_REPORTER;
2505         ma->mca_flags &= ~MAF_TIMER_RUNNING;
2506         spin_unlock(&ma->mca_lock);
2507         ma_put(ma);
2508 }
2509
2510 /* Device changing type */
2511
2512 void ipv6_mc_unmap(struct inet6_dev *idev)
2513 {
2514         struct ifmcaddr6 *i;
2515
2516         /* Install multicast list, except for all-nodes (already installed) */
2517
2518         read_lock_bh(&idev->lock);
2519         for (i = idev->mc_list; i; i = i->next)
2520                 igmp6_group_dropped(i);
2521         read_unlock_bh(&idev->lock);
2522 }
2523
2524 void ipv6_mc_remap(struct inet6_dev *idev)
2525 {
2526         ipv6_mc_up(idev);
2527 }
2528
2529 /* Device going down */
2530
2531 void ipv6_mc_down(struct inet6_dev *idev)
2532 {
2533         struct ifmcaddr6 *i;
2534
2535         /* Withdraw multicast list */
2536
2537         read_lock_bh(&idev->lock);
2538
2539         for (i = idev->mc_list; i; i = i->next)
2540                 igmp6_group_dropped(i);
2541
2542         /* Should stop timer after group drop. or we will
2543          * start timer again in mld_ifc_event()
2544          */
2545         mld_ifc_stop_timer(idev);
2546         mld_gq_stop_timer(idev);
2547         mld_dad_stop_timer(idev);
2548         read_unlock_bh(&idev->lock);
2549 }
2550
2551 static void ipv6_mc_reset(struct inet6_dev *idev)
2552 {
2553         idev->mc_qrv = sysctl_mld_qrv;
2554         idev->mc_qi = MLD_QI_DEFAULT;
2555         idev->mc_qri = MLD_QRI_DEFAULT;
2556         idev->mc_v1_seen = 0;
2557         idev->mc_maxdelay = unsolicited_report_interval(idev);
2558 }
2559
2560 /* Device going up */
2561
2562 void ipv6_mc_up(struct inet6_dev *idev)
2563 {
2564         struct ifmcaddr6 *i;
2565
2566         /* Install multicast list, except for all-nodes (already installed) */
2567
2568         read_lock_bh(&idev->lock);
2569         ipv6_mc_reset(idev);
2570         for (i = idev->mc_list; i; i = i->next) {
2571                 mld_del_delrec(idev, i);
2572                 igmp6_group_added(i);
2573         }
2574         read_unlock_bh(&idev->lock);
2575 }
2576
2577 /* IPv6 device initialization. */
2578
2579 void ipv6_mc_init_dev(struct inet6_dev *idev)
2580 {
2581         write_lock_bh(&idev->lock);
2582         spin_lock_init(&idev->mc_lock);
2583         idev->mc_gq_running = 0;
2584         timer_setup(&idev->mc_gq_timer, mld_gq_timer_expire, 0);
2585         idev->mc_tomb = NULL;
2586         idev->mc_ifc_count = 0;
2587         timer_setup(&idev->mc_ifc_timer, mld_ifc_timer_expire, 0);
2588         timer_setup(&idev->mc_dad_timer, mld_dad_timer_expire, 0);
2589         ipv6_mc_reset(idev);
2590         write_unlock_bh(&idev->lock);
2591 }
2592
2593 /*
2594  *      Device is about to be destroyed: clean up.
2595  */
2596
2597 void ipv6_mc_destroy_dev(struct inet6_dev *idev)
2598 {
2599         struct ifmcaddr6 *i;
2600
2601         /* Deactivate timers */
2602         ipv6_mc_down(idev);
2603         mld_clear_delrec(idev);
2604
2605         /* Delete all-nodes address. */
2606         /* We cannot call ipv6_dev_mc_dec() directly, our caller in
2607          * addrconf.c has NULL'd out dev->ip6_ptr so in6_dev_get() will
2608          * fail.
2609          */
2610         __ipv6_dev_mc_dec(idev, &in6addr_linklocal_allnodes);
2611
2612         if (idev->cnf.forwarding)
2613                 __ipv6_dev_mc_dec(idev, &in6addr_linklocal_allrouters);
2614
2615         write_lock_bh(&idev->lock);
2616         while ((i = idev->mc_list) != NULL) {
2617                 idev->mc_list = i->next;
2618
2619                 write_unlock_bh(&idev->lock);
2620                 ip6_mc_clear_src(i);
2621                 ma_put(i);
2622                 write_lock_bh(&idev->lock);
2623         }
2624         write_unlock_bh(&idev->lock);
2625 }
2626
2627 static void ipv6_mc_rejoin_groups(struct inet6_dev *idev)
2628 {
2629         struct ifmcaddr6 *pmc;
2630
2631         ASSERT_RTNL();
2632
2633         if (mld_in_v1_mode(idev)) {
2634                 read_lock_bh(&idev->lock);
2635                 for (pmc = idev->mc_list; pmc; pmc = pmc->next)
2636                         igmp6_join_group(pmc);
2637                 read_unlock_bh(&idev->lock);
2638         } else
2639                 mld_send_report(idev, NULL);
2640 }
2641
2642 static int ipv6_mc_netdev_event(struct notifier_block *this,
2643                                 unsigned long event,
2644                                 void *ptr)
2645 {
2646         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2647         struct inet6_dev *idev = __in6_dev_get(dev);
2648
2649         switch (event) {
2650         case NETDEV_RESEND_IGMP:
2651                 if (idev)
2652                         ipv6_mc_rejoin_groups(idev);
2653                 break;
2654         default:
2655                 break;
2656         }
2657
2658         return NOTIFY_DONE;
2659 }
2660
2661 static struct notifier_block igmp6_netdev_notifier = {
2662         .notifier_call = ipv6_mc_netdev_event,
2663 };
2664
2665 #ifdef CONFIG_PROC_FS
2666 struct igmp6_mc_iter_state {
2667         struct seq_net_private p;
2668         struct net_device *dev;
2669         struct inet6_dev *idev;
2670 };
2671
2672 #define igmp6_mc_seq_private(seq)       ((struct igmp6_mc_iter_state *)(seq)->private)
2673
2674 static inline struct ifmcaddr6 *igmp6_mc_get_first(struct seq_file *seq)
2675 {
2676         struct ifmcaddr6 *im = NULL;
2677         struct igmp6_mc_iter_state *state = igmp6_mc_seq_private(seq);
2678         struct net *net = seq_file_net(seq);
2679
2680         state->idev = NULL;
2681         for_each_netdev_rcu(net, state->dev) {
2682                 struct inet6_dev *idev;
2683                 idev = __in6_dev_get(state->dev);
2684                 if (!idev)
2685                         continue;
2686                 read_lock_bh(&idev->lock);
2687                 im = idev->mc_list;
2688                 if (im) {
2689                         state->idev = idev;
2690                         break;
2691                 }
2692                 read_unlock_bh(&idev->lock);
2693         }
2694         return im;
2695 }
2696
2697 static struct ifmcaddr6 *igmp6_mc_get_next(struct seq_file *seq, struct ifmcaddr6 *im)
2698 {
2699         struct igmp6_mc_iter_state *state = igmp6_mc_seq_private(seq);
2700
2701         im = im->next;
2702         while (!im) {
2703                 if (likely(state->idev))
2704                         read_unlock_bh(&state->idev->lock);
2705
2706                 state->dev = next_net_device_rcu(state->dev);
2707                 if (!state->dev) {
2708                         state->idev = NULL;
2709                         break;
2710                 }
2711                 state->idev = __in6_dev_get(state->dev);
2712                 if (!state->idev)
2713                         continue;
2714                 read_lock_bh(&state->idev->lock);
2715                 im = state->idev->mc_list;
2716         }
2717         return im;
2718 }
2719
2720 static struct ifmcaddr6 *igmp6_mc_get_idx(struct seq_file *seq, loff_t pos)
2721 {
2722         struct ifmcaddr6 *im = igmp6_mc_get_first(seq);
2723         if (im)
2724                 while (pos && (im = igmp6_mc_get_next(seq, im)) != NULL)
2725                         --pos;
2726         return pos ? NULL : im;
2727 }
2728
2729 static void *igmp6_mc_seq_start(struct seq_file *seq, loff_t *pos)
2730         __acquires(RCU)
2731 {
2732         rcu_read_lock();
2733         return igmp6_mc_get_idx(seq, *pos);
2734 }
2735
2736 static void *igmp6_mc_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2737 {
2738         struct ifmcaddr6 *im = igmp6_mc_get_next(seq, v);
2739
2740         ++*pos;
2741         return im;
2742 }
2743
2744 static void igmp6_mc_seq_stop(struct seq_file *seq, void *v)
2745         __releases(RCU)
2746 {
2747         struct igmp6_mc_iter_state *state = igmp6_mc_seq_private(seq);
2748
2749         if (likely(state->idev)) {
2750                 read_unlock_bh(&state->idev->lock);
2751                 state->idev = NULL;
2752         }
2753         state->dev = NULL;
2754         rcu_read_unlock();
2755 }
2756
2757 static int igmp6_mc_seq_show(struct seq_file *seq, void *v)
2758 {
2759         struct ifmcaddr6 *im = (struct ifmcaddr6 *)v;
2760         struct igmp6_mc_iter_state *state = igmp6_mc_seq_private(seq);
2761
2762         seq_printf(seq,
2763                    "%-4d %-15s %pi6 %5d %08X %ld\n",
2764                    state->dev->ifindex, state->dev->name,
2765                    &im->mca_addr,
2766                    im->mca_users, im->mca_flags,
2767                    (im->mca_flags&MAF_TIMER_RUNNING) ?
2768                    jiffies_to_clock_t(im->mca_timer.expires-jiffies) : 0);
2769         return 0;
2770 }
2771
2772 static const struct seq_operations igmp6_mc_seq_ops = {
2773         .start  =       igmp6_mc_seq_start,
2774         .next   =       igmp6_mc_seq_next,
2775         .stop   =       igmp6_mc_seq_stop,
2776         .show   =       igmp6_mc_seq_show,
2777 };
2778
2779 struct igmp6_mcf_iter_state {
2780         struct seq_net_private p;
2781         struct net_device *dev;
2782         struct inet6_dev *idev;
2783         struct ifmcaddr6 *im;
2784 };
2785
2786 #define igmp6_mcf_seq_private(seq)      ((struct igmp6_mcf_iter_state *)(seq)->private)
2787
2788 static inline struct ip6_sf_list *igmp6_mcf_get_first(struct seq_file *seq)
2789 {
2790         struct ip6_sf_list *psf = NULL;
2791         struct ifmcaddr6 *im = NULL;
2792         struct igmp6_mcf_iter_state *state = igmp6_mcf_seq_private(seq);
2793         struct net *net = seq_file_net(seq);
2794
2795         state->idev = NULL;
2796         state->im = NULL;
2797         for_each_netdev_rcu(net, state->dev) {
2798                 struct inet6_dev *idev;
2799                 idev = __in6_dev_get(state->dev);
2800                 if (unlikely(idev == NULL))
2801                         continue;
2802                 read_lock_bh(&idev->lock);
2803                 im = idev->mc_list;
2804                 if (likely(im)) {
2805                         spin_lock_bh(&im->mca_lock);
2806                         psf = im->mca_sources;
2807                         if (likely(psf)) {
2808                                 state->im = im;
2809                                 state->idev = idev;
2810                                 break;
2811                         }
2812                         spin_unlock_bh(&im->mca_lock);
2813                 }
2814                 read_unlock_bh(&idev->lock);
2815         }
2816         return psf;
2817 }
2818
2819 static struct ip6_sf_list *igmp6_mcf_get_next(struct seq_file *seq, struct ip6_sf_list *psf)
2820 {
2821         struct igmp6_mcf_iter_state *state = igmp6_mcf_seq_private(seq);
2822
2823         psf = psf->sf_next;
2824         while (!psf) {
2825                 spin_unlock_bh(&state->im->mca_lock);
2826                 state->im = state->im->next;
2827                 while (!state->im) {
2828                         if (likely(state->idev))
2829                                 read_unlock_bh(&state->idev->lock);
2830
2831                         state->dev = next_net_device_rcu(state->dev);
2832                         if (!state->dev) {
2833                                 state->idev = NULL;
2834                                 goto out;
2835                         }
2836                         state->idev = __in6_dev_get(state->dev);
2837                         if (!state->idev)
2838                                 continue;
2839                         read_lock_bh(&state->idev->lock);
2840                         state->im = state->idev->mc_list;
2841                 }
2842                 if (!state->im)
2843                         break;
2844                 spin_lock_bh(&state->im->mca_lock);
2845                 psf = state->im->mca_sources;
2846         }
2847 out:
2848         return psf;
2849 }
2850
2851 static struct ip6_sf_list *igmp6_mcf_get_idx(struct seq_file *seq, loff_t pos)
2852 {
2853         struct ip6_sf_list *psf = igmp6_mcf_get_first(seq);
2854         if (psf)
2855                 while (pos && (psf = igmp6_mcf_get_next(seq, psf)) != NULL)
2856                         --pos;
2857         return pos ? NULL : psf;
2858 }
2859
2860 static void *igmp6_mcf_seq_start(struct seq_file *seq, loff_t *pos)
2861         __acquires(RCU)
2862 {
2863         rcu_read_lock();
2864         return *pos ? igmp6_mcf_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
2865 }
2866
2867 static void *igmp6_mcf_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2868 {
2869         struct ip6_sf_list *psf;
2870         if (v == SEQ_START_TOKEN)
2871                 psf = igmp6_mcf_get_first(seq);
2872         else
2873                 psf = igmp6_mcf_get_next(seq, v);
2874         ++*pos;
2875         return psf;
2876 }
2877
2878 static void igmp6_mcf_seq_stop(struct seq_file *seq, void *v)
2879         __releases(RCU)
2880 {
2881         struct igmp6_mcf_iter_state *state = igmp6_mcf_seq_private(seq);
2882         if (likely(state->im)) {
2883                 spin_unlock_bh(&state->im->mca_lock);
2884                 state->im = NULL;
2885         }
2886         if (likely(state->idev)) {
2887                 read_unlock_bh(&state->idev->lock);
2888                 state->idev = NULL;
2889         }
2890         state->dev = NULL;
2891         rcu_read_unlock();
2892 }
2893
2894 static int igmp6_mcf_seq_show(struct seq_file *seq, void *v)
2895 {
2896         struct ip6_sf_list *psf = (struct ip6_sf_list *)v;
2897         struct igmp6_mcf_iter_state *state = igmp6_mcf_seq_private(seq);
2898
2899         if (v == SEQ_START_TOKEN) {
2900                 seq_puts(seq, "Idx Device                Multicast Address                   Source Address    INC    EXC\n");
2901         } else {
2902                 seq_printf(seq,
2903                            "%3d %6.6s %pi6 %pi6 %6lu %6lu\n",
2904                            state->dev->ifindex, state->dev->name,
2905                            &state->im->mca_addr,
2906                            &psf->sf_addr,
2907                            psf->sf_count[MCAST_INCLUDE],
2908                            psf->sf_count[MCAST_EXCLUDE]);
2909         }
2910         return 0;
2911 }
2912
2913 static const struct seq_operations igmp6_mcf_seq_ops = {
2914         .start  =       igmp6_mcf_seq_start,
2915         .next   =       igmp6_mcf_seq_next,
2916         .stop   =       igmp6_mcf_seq_stop,
2917         .show   =       igmp6_mcf_seq_show,
2918 };
2919
2920 static int __net_init igmp6_proc_init(struct net *net)
2921 {
2922         int err;
2923
2924         err = -ENOMEM;
2925         if (!proc_create_net("igmp6", 0444, net->proc_net, &igmp6_mc_seq_ops,
2926                         sizeof(struct igmp6_mc_iter_state)))
2927                 goto out;
2928         if (!proc_create_net("mcfilter6", 0444, net->proc_net,
2929                         &igmp6_mcf_seq_ops,
2930                         sizeof(struct igmp6_mcf_iter_state)))
2931                 goto out_proc_net_igmp6;
2932
2933         err = 0;
2934 out:
2935         return err;
2936
2937 out_proc_net_igmp6:
2938         remove_proc_entry("igmp6", net->proc_net);
2939         goto out;
2940 }
2941
2942 static void __net_exit igmp6_proc_exit(struct net *net)
2943 {
2944         remove_proc_entry("mcfilter6", net->proc_net);
2945         remove_proc_entry("igmp6", net->proc_net);
2946 }
2947 #else
2948 static inline int igmp6_proc_init(struct net *net)
2949 {
2950         return 0;
2951 }
2952 static inline void igmp6_proc_exit(struct net *net)
2953 {
2954 }
2955 #endif
2956
2957 static int __net_init igmp6_net_init(struct net *net)
2958 {
2959         int err;
2960
2961         err = inet_ctl_sock_create(&net->ipv6.igmp_sk, PF_INET6,
2962                                    SOCK_RAW, IPPROTO_ICMPV6, net);
2963         if (err < 0) {
2964                 pr_err("Failed to initialize the IGMP6 control socket (err %d)\n",
2965                        err);
2966                 goto out;
2967         }
2968
2969         inet6_sk(net->ipv6.igmp_sk)->hop_limit = 1;
2970
2971         err = inet_ctl_sock_create(&net->ipv6.mc_autojoin_sk, PF_INET6,
2972                                    SOCK_RAW, IPPROTO_ICMPV6, net);
2973         if (err < 0) {
2974                 pr_err("Failed to initialize the IGMP6 autojoin socket (err %d)\n",
2975                        err);
2976                 goto out_sock_create;
2977         }
2978
2979         err = igmp6_proc_init(net);
2980         if (err)
2981                 goto out_sock_create_autojoin;
2982
2983         return 0;
2984
2985 out_sock_create_autojoin:
2986         inet_ctl_sock_destroy(net->ipv6.mc_autojoin_sk);
2987 out_sock_create:
2988         inet_ctl_sock_destroy(net->ipv6.igmp_sk);
2989 out:
2990         return err;
2991 }
2992
2993 static void __net_exit igmp6_net_exit(struct net *net)
2994 {
2995         inet_ctl_sock_destroy(net->ipv6.igmp_sk);
2996         inet_ctl_sock_destroy(net->ipv6.mc_autojoin_sk);
2997         igmp6_proc_exit(net);
2998 }
2999
3000 static struct pernet_operations igmp6_net_ops = {
3001         .init = igmp6_net_init,
3002         .exit = igmp6_net_exit,
3003 };
3004
3005 int __init igmp6_init(void)
3006 {
3007         return register_pernet_subsys(&igmp6_net_ops);
3008 }
3009
3010 int __init igmp6_late_init(void)
3011 {
3012         return register_netdevice_notifier(&igmp6_netdev_notifier);
3013 }
3014
3015 void igmp6_cleanup(void)
3016 {
3017         unregister_pernet_subsys(&igmp6_net_ops);
3018 }
3019
3020 void igmp6_late_cleanup(void)
3021 {
3022         unregister_netdevice_notifier(&igmp6_netdev_notifier);
3023 }