net/ipv4/igmp.c

   1 /*
   2  *      Linux NET3:     Internet Group Management Protocol  [IGMP]
   3  *
   4  *      This code implements the IGMP protocol as defined in RFC1112. There has
   5  *      been a further revision of this protocol since which is now supported.
   6  *
   7  *      If you have trouble with this module be careful what gcc you have used,
   8  *      the older version didn't come out right using gcc 2.5.8, the newer one
   9  *      seems to fall out with gcc 2.6.2.
  10  *
  11  *      Authors:
  12  *              Alan Cox <alan@lxorguk.ukuu.org.uk>
  13  *
  14  *      This program is free software; you can redistribute it and/or
  15  *      modify it under the terms of the GNU General Public License
  16  *      as published by the Free Software Foundation; either version
  17  *      2 of the License, or (at your option) any later version.
  18  *
  19  *      Fixes:
  20  *
  21  *              Alan Cox        :       Added lots of __inline__ to optimise
  22  *                                      the memory usage of all the tiny little
  23  *                                      functions.
  24  *              Alan Cox        :       Dumped the header building experiment.
  25  *              Alan Cox        :       Minor tweaks ready for multicast routing
  26  *                                      and extended IGMP protocol.
  27  *              Alan Cox        :       Removed a load of inline directives. Gcc 2.5.8
  28  *                                      writes utterly bogus code otherwise (sigh)
  29  *                                      fixed IGMP loopback to behave in the manner
  30  *                                      desired by mrouted, fixed the fact it has been
  31  *                                      broken since 1.3.6 and cleaned up a few minor
  32  *                                      points.
  33  *
  34  *              Chih-Jen Chang  :       Tried to revise IGMP to Version 2
  35  *              Tsu-Sheng Tsao          E-mail: chihjenc@scf.usc.edu and tsusheng@scf.usc.edu
  36  *                                      The enhancements are mainly based on Steve Deering's
  37  *                                      ipmulti-3.5 source code.
  38  *              Chih-Jen Chang  :       Added the igmp_get_mrouter_info and
  39  *              Tsu-Sheng Tsao          igmp_set_mrouter_info to keep track of
  40  *                                      the mrouted version on that device.
  41  *              Chih-Jen Chang  :       Added the max_resp_time parameter to
  42  *              Tsu-Sheng Tsao          igmp_heard_query(). Using this parameter
  43  *                                      to identify the multicast router version
  44  *                                      and do what the IGMP version 2 specified.
  45  *              Chih-Jen Chang  :       Added a timer to revert to IGMP V2 router
  46  *              Tsu-Sheng Tsao          if the specified time expired.
  47  *              Alan Cox        :       Stop IGMP from 0.0.0.0 being accepted.
  48  *              Alan Cox        :       Use GFP_ATOMIC in the right places.
  49  *              Christian Daudt :       igmp timer wasn't set for local group
  50  *                                      memberships but was being deleted,
  51  *                                      which caused a "del_timer() called
  52  *                                      from %p with timer not initialized\n"
  53  *                                      message (960131).
  54  *              Christian Daudt :       removed del_timer from
  55  *                                      igmp_timer_expire function (960205).
  56  *             Christian Daudt :       igmp_heard_report now only calls
  57  *                                     igmp_timer_expire if tm->running is
  58  *                                     true (960216).
  59  *              Malcolm Beattie :       ttl comparison wrong in igmp_rcv made
  60  *                                      igmp_heard_query never trigger. Expiry
  61  *                                      miscalculation fixed in igmp_heard_query
  62  *                                      and random() made to return unsigned to
  63  *                                      prevent negative expiry times.
  64  *              Alexey Kuznetsov:       Wrong group leaving behaviour, backport
  65  *                                      fix from pending 2.1.x patches.
  66  *              Alan Cox:               Forget to enable FDDI support earlier.
  67  *              Alexey Kuznetsov:       Fixed leaving groups on device down.
  68  *              Alexey Kuznetsov:       Accordance to igmp-v2-06 draft.
  69  *              David L Stevens:        IGMPv3 support, with help from
  70  *                                      Vinay Kulkarni
  71  */
  72
  73 #include <linux/module.h>
  74 #include <linux/slab.h>
  75 #include <asm/uaccess.h>
  76 #include <asm/system.h>
  77 #include <linux/types.h>
  78 #include <linux/kernel.h>
  79 #include <linux/jiffies.h>
  80 #include <linux/string.h>
  81 #include <linux/socket.h>
  82 #include <linux/sockios.h>
  83 #include <linux/in.h>
  84 #include <linux/inet.h>
  85 #include <linux/netdevice.h>
  86 #include <linux/skbuff.h>
  87 #include <linux/inetdevice.h>
  88 #include <linux/igmp.h>
  89 #include <linux/if_arp.h>
  90 #include <linux/rtnetlink.h>
  91 #include <linux/times.h>
  92 #include <linux/byteorder/generic.h>
  93
  94 #include <net/net_namespace.h>
  95 #include <net/arp.h>
  96 #include <net/ip.h>
  97 #include <net/protocol.h>
  98 #include <net/route.h>
  99 #include <net/sock.h>
 100 #include <net/checksum.h>
 101 #include <linux/netfilter_ipv4.h>
 102 #ifdef CONFIG_IP_MROUTE
 103 #include <linux/mroute.h>
 104 #endif
 105 #ifdef CONFIG_PROC_FS
 106 #include <linux/proc_fs.h>
 107 #include <linux/seq_file.h>
 108 #endif
 109
 110 #define IP_MAX_MEMBERSHIPS      20
 111 #define IP_MAX_MSF              10
 112
 113 #ifdef CONFIG_IP_MULTICAST
 114 /* Parameter names and values are taken from igmp-v2-06 draft */
 115
 116 #define IGMP_V1_Router_Present_Timeout          (400*HZ)
 117 #define IGMP_V2_Router_Present_Timeout          (400*HZ)
 118 #define IGMP_Unsolicited_Report_Interval        (10*HZ)
 119 #define IGMP_Query_Response_Interval            (10*HZ)
 120 #define IGMP_Unsolicited_Report_Count           2
 121
 122
 123 #define IGMP_Initial_Report_Delay               (1)
 124
 125 /* IGMP_Initial_Report_Delay is not from IGMP specs!
 126  * IGMP specs require to report membership immediately after
 127  * joining a group, but we delay the first report by a
 128  * small interval. It seems more natural and still does not
 129  * contradict to specs provided this delay is small enough.
 130  */
 131
 132 #define IGMP_V1_SEEN(in_dev) \
 133         (IPV4_DEVCONF_ALL(dev_net(in_dev->dev), FORCE_IGMP_VERSION) == 1 || \
 134          IN_DEV_CONF_GET((in_dev), FORCE_IGMP_VERSION) == 1 || \
 135          ((in_dev)->mr_v1_seen && \
 136           time_before(jiffies, (in_dev)->mr_v1_seen)))
 137 #define IGMP_V2_SEEN(in_dev) \
 138         (IPV4_DEVCONF_ALL(dev_net(in_dev->dev), FORCE_IGMP_VERSION) == 2 || \
 139          IN_DEV_CONF_GET((in_dev), FORCE_IGMP_VERSION) == 2 || \
 140          ((in_dev)->mr_v2_seen && \
 141           time_before(jiffies, (in_dev)->mr_v2_seen)))
 142
 143 static void igmpv3_add_delrec(struct in_device *in_dev, struct ip_mc_list *im);
 144 static void igmpv3_del_delrec(struct in_device *in_dev, struct ip_mc_list *im);
 145 static void igmpv3_clear_delrec(struct in_device *in_dev);
 146 static int sf_setstate(struct ip_mc_list *pmc);
 147 static void sf_markstate(struct ip_mc_list *pmc);
 148 #endif
 149 static void ip_mc_clear_src(struct ip_mc_list *pmc);
 150 static int ip_mc_add_src(struct in_device *in_dev, __be32 *pmca, int sfmode,
 151                          int sfcount, __be32 *psfsrc, int delta);
 152
 153 static void ip_ma_put(struct ip_mc_list *im)
 154 {
 155         if (atomic_dec_and_test(&im->refcnt)) {
 156                 in_dev_put(im->interface);
 157                 kfree_rcu(im, rcu);
 158         }
 159 }
 160
 161 #define for_each_pmc_rcu(in_dev, pmc)                           \
 162         for (pmc = rcu_dereference(in_dev->mc_list);            \
 163              pmc != NULL;                                       \
 164              pmc = rcu_dereference(pmc->next_rcu))
 165
 166 #define for_each_pmc_rtnl(in_dev, pmc)                          \
 167         for (pmc = rtnl_dereference(in_dev->mc_list);           \
 168              pmc != NULL;                                       \
 169              pmc = rtnl_dereference(pmc->next_rcu))
 170
 171 #ifdef CONFIG_IP_MULTICAST
 172
 173 /*
 174  *      Timer management
 175  */
 176
 177 static void igmp_stop_timer(struct ip_mc_list *im)
 178 {
 179         spin_lock_bh(&im->lock);
 180         if (del_timer(&im->timer))
 181                 atomic_dec(&im->refcnt);
 182         im->tm_running = 0;
 183         im->reporter = 0;
 184         im->unsolicit_count = 0;
 185         spin_unlock_bh(&im->lock);
 186 }
 187
 188 /* It must be called with locked im->lock */
 189 static void igmp_start_timer(struct ip_mc_list *im, int max_delay)
 190 {
 191         int tv = net_random() % max_delay;
 192
 193         im->tm_running = 1;
 194         if (!mod_timer(&im->timer, jiffies+tv+2))
 195                 atomic_inc(&im->refcnt);
 196 }
 197
 198 static void igmp_gq_start_timer(struct in_device *in_dev)
 199 {
 200         int tv = net_random() % in_dev->mr_maxdelay;
 201         unsigned long exp = jiffies + tv + 2;
 202
 203         if (in_dev->mr_gq_running &&
 204             time_after_eq(exp, (in_dev->mr_gq_timer).expires))
 205                 return;
 206
 207         in_dev->mr_gq_running = 1;
 208         if (!mod_timer(&in_dev->mr_gq_timer, exp))
 209                 in_dev_hold(in_dev);
 210 }
 211
 212 static void igmp_ifc_start_timer(struct in_device *in_dev, int delay)
 213 {
 214         int tv = net_random() % delay;
 215
 216         if (!mod_timer(&in_dev->mr_ifc_timer, jiffies+tv+2))
 217                 in_dev_hold(in_dev);
 218 }
 219
 220 static void igmp_mod_timer(struct ip_mc_list *im, int max_delay)
 221 {
 222         spin_lock_bh(&im->lock);
 223         im->unsolicit_count = 0;
 224         if (del_timer(&im->timer)) {
 225                 if ((long)(im->timer.expires-jiffies) < max_delay) {
 226                         add_timer(&im->timer);
 227                         im->tm_running = 1;
 228                         spin_unlock_bh(&im->lock);
 229                         return;
 230                 }
 231                 atomic_dec(&im->refcnt);
 232         }
 233         igmp_start_timer(im, max_delay);
 234         spin_unlock_bh(&im->lock);
 235 }
 236
 237
 238 /*
 239  *      Send an IGMP report.
 240  */
 241
 242 #define IGMP_SIZE (sizeof(struct igmphdr)+sizeof(struct iphdr)+4)
 243
 244
 245 static int is_in(struct ip_mc_list *pmc, struct ip_sf_list *psf, int type,
 246         int gdeleted, int sdeleted)
 247 {
 248         switch (type) {
 249         case IGMPV3_MODE_IS_INCLUDE:
 250         case IGMPV3_MODE_IS_EXCLUDE:
 251                 if (gdeleted || sdeleted)
 252                         return 0;
 253                 if (!(pmc->gsquery && !psf->sf_gsresp)) {
 254                         if (pmc->sfmode == MCAST_INCLUDE)
 255                                 return 1;
 256                         /* don't include if this source is excluded
 257                          * in all filters
 258                          */
 259                         if (psf->sf_count[MCAST_INCLUDE])
 260                                 return type == IGMPV3_MODE_IS_INCLUDE;
 261                         return pmc->sfcount[MCAST_EXCLUDE] ==
 262                                 psf->sf_count[MCAST_EXCLUDE];
 263                 }
 264                 return 0;
 265         case IGMPV3_CHANGE_TO_INCLUDE:
 266                 if (gdeleted || sdeleted)
 267                         return 0;
 268                 return psf->sf_count[MCAST_INCLUDE] != 0;
 269         case IGMPV3_CHANGE_TO_EXCLUDE:
 270                 if (gdeleted || sdeleted)
 271                         return 0;
 272                 if (pmc->sfcount[MCAST_EXCLUDE] == 0 ||
 273                     psf->sf_count[MCAST_INCLUDE])
 274                         return 0;
 275                 return pmc->sfcount[MCAST_EXCLUDE] ==
 276                         psf->sf_count[MCAST_EXCLUDE];
 277         case IGMPV3_ALLOW_NEW_SOURCES:
 278                 if (gdeleted || !psf->sf_crcount)
 279                         return 0;
 280                 return (pmc->sfmode == MCAST_INCLUDE) ^ sdeleted;
 281         case IGMPV3_BLOCK_OLD_SOURCES:
 282                 if (pmc->sfmode == MCAST_INCLUDE)
 283                         return gdeleted || (psf->sf_crcount && sdeleted);
 284                 return psf->sf_crcount && !gdeleted && !sdeleted;
 285         }
 286         return 0;
 287 }
 288
 289 static int
 290 igmp_scount(struct ip_mc_list *pmc, int type, int gdeleted, int sdeleted)
 291 {
 292         struct ip_sf_list *psf;
 293         int scount = 0;
 294
 295         for (psf=pmc->sources; psf; psf=psf->sf_next) {
 296                 if (!is_in(pmc, psf, type, gdeleted, sdeleted))
 297                         continue;
 298                 scount++;
 299         }
 300         return scount;
 301 }
 302
 303 /* source address selection per RFC 3376 section 4.2.13 */
 304 static __be32 igmpv3_get_srcaddr(struct net_device *dev,
 305                                  const struct flowi4 *fl4)
 306 {
 307         struct in_device *in_dev = __in_dev_get_rcu(dev);
 308
 309         if (!in_dev)
 310                 return htonl(INADDR_ANY);
 311
 312         for_ifa(in_dev) {
 313                 if (fl4->saddr == ifa->ifa_local)
 314                         return fl4->saddr;
 315         } endfor_ifa(in_dev);
 316
 317         return htonl(INADDR_ANY);
 318 }
 319
 320 static struct sk_buff *igmpv3_newpack(struct net_device *dev, unsigned int mtu)
 321 {
 322         struct sk_buff *skb;
 323         struct rtable *rt;
 324         struct iphdr *pip;
 325         struct igmpv3_report *pig;
 326         struct net *net = dev_net(dev);
 327         struct flowi4 fl4;
 328         int hlen = LL_RESERVED_SPACE(dev);
 329         int tlen = dev->needed_tailroom;
 330         unsigned int size = mtu;
 331
 332         while (1) {
 333                 skb = alloc_skb(size + hlen + tlen,
 334                                 GFP_ATOMIC | __GFP_NOWARN);
 335                 if (skb)
 336                         break;
 337                 size >>= 1;
 338                 if (size < 256)
 339                         return NULL;
 340         }
 341
 342         rt = ip_route_output_ports(net, &fl4, NULL, IGMPV3_ALL_MCR, 0,
 343                                    0, 0,
 344                                    IPPROTO_IGMP, 0, dev->ifindex);
 345         if (IS_ERR(rt)) {
 346                 kfree_skb(skb);
 347                 return NULL;
 348         }
 349
 350         skb_dst_set(skb, &rt->dst);
 351         skb->dev = dev;
 352
 353         skb_reserve(skb, hlen);
 354         skb_tailroom_reserve(skb, mtu, tlen);
 355
 356         skb_reset_network_header(skb);
 357         pip = ip_hdr(skb);
 358         skb_put(skb, sizeof(struct iphdr) + 4);
 359
 360         pip->version  = 4;
 361         pip->ihl      = (sizeof(struct iphdr)+4)>>2;
 362         pip->tos      = 0xc0;
 363         pip->frag_off = htons(IP_DF);
 364         pip->ttl      = 1;
 365         pip->daddr    = fl4.daddr;
 366
 367         rcu_read_lock();
 368         pip->saddr    = igmpv3_get_srcaddr(dev, &fl4);
 369         rcu_read_unlock();
 370
 371         pip->protocol = IPPROTO_IGMP;
 372         pip->tot_len  = 0;      /* filled in later */
 373         ip_select_ident(skb, NULL);
 374         ((u8*)&pip[1])[0] = IPOPT_RA;
 375         ((u8*)&pip[1])[1] = 4;
 376         ((u8*)&pip[1])[2] = 0;
 377         ((u8*)&pip[1])[3] = 0;
 378
 379         skb->transport_header = skb->network_header + sizeof(struct iphdr) + 4;
 380         skb_put(skb, sizeof(*pig));
 381         pig = igmpv3_report_hdr(skb);
 382         pig->type = IGMPV3_HOST_MEMBERSHIP_REPORT;
 383         pig->resv1 = 0;
 384         pig->csum = 0;
 385         pig->resv2 = 0;
 386         pig->ngrec = 0;
 387         return skb;
 388 }
 389
 390 static int igmpv3_sendpack(struct sk_buff *skb)
 391 {
 392         struct igmphdr *pig = igmp_hdr(skb);
 393         const int igmplen = skb->tail - skb->transport_header;
 394
 395         pig->csum = ip_compute_csum(igmp_hdr(skb), igmplen);
 396
 397         return ip_local_out(skb);
 398 }
 399
 400 static int grec_size(struct ip_mc_list *pmc, int type, int gdel, int sdel)
 401 {
 402         return sizeof(struct igmpv3_grec) + 4*igmp_scount(pmc, type, gdel, sdel);
 403 }
 404
 405 static struct sk_buff *add_grhead(struct sk_buff *skb, struct ip_mc_list *pmc,
 406         int type, struct igmpv3_grec **ppgr)
 407 {
 408         struct net_device *dev = pmc->interface->dev;
 409         struct igmpv3_report *pih;
 410         struct igmpv3_grec *pgr;
 411
 412         if (!skb)
 413                 skb = igmpv3_newpack(dev, dev->mtu);
 414         if (!skb)
 415                 return NULL;
 416         pgr = (struct igmpv3_grec *)skb_put(skb, sizeof(struct igmpv3_grec));
 417         pgr->grec_type = type;
 418         pgr->grec_auxwords = 0;
 419         pgr->grec_nsrcs = 0;
 420         pgr->grec_mca = pmc->multiaddr;
 421         pih = igmpv3_report_hdr(skb);
 422         pih->ngrec = htons(ntohs(pih->ngrec)+1);
 423         *ppgr = pgr;
 424         return skb;
 425 }
 426
 427 #define AVAILABLE(skb)  ((skb) ? skb_availroom(skb) : 0)
 428
 429 static struct sk_buff *add_grec(struct sk_buff *skb, struct ip_mc_list *pmc,
 430         int type, int gdeleted, int sdeleted)
 431 {
 432         struct net_device *dev = pmc->interface->dev;
 433         struct igmpv3_report *pih;
 434         struct igmpv3_grec *pgr = NULL;
 435         struct ip_sf_list *psf, *psf_next, *psf_prev, **psf_list;
 436         int scount, stotal, first, isquery, truncate;
 437
 438         if (pmc->multiaddr == IGMP_ALL_HOSTS)
 439                 return skb;
 440
 441         isquery = type == IGMPV3_MODE_IS_INCLUDE ||
 442                   type == IGMPV3_MODE_IS_EXCLUDE;
 443         truncate = type == IGMPV3_MODE_IS_EXCLUDE ||
 444                     type == IGMPV3_CHANGE_TO_EXCLUDE;
 445
 446         stotal = scount = 0;
 447
 448         psf_list = sdeleted ? &pmc->tomb : &pmc->sources;
 449
 450         if (!*psf_list)
 451                 goto empty_source;
 452
 453         pih = skb ? igmpv3_report_hdr(skb) : NULL;
 454
 455         /* EX and TO_EX get a fresh packet, if needed */
 456         if (truncate) {
 457                 if (pih && pih->ngrec &&
 458                     AVAILABLE(skb) < grec_size(pmc, type, gdeleted, sdeleted)) {
 459                         if (skb)
 460                                 igmpv3_sendpack(skb);
 461                         skb = igmpv3_newpack(dev, dev->mtu);
 462                 }
 463         }
 464         first = 1;
 465         psf_prev = NULL;
 466         for (psf=*psf_list; psf; psf=psf_next) {
 467                 __be32 *psrc;
 468
 469                 psf_next = psf->sf_next;
 470
 471                 if (!is_in(pmc, psf, type, gdeleted, sdeleted)) {
 472                         psf_prev = psf;
 473                         continue;
 474                 }
 475
 476                 /* clear marks on query responses */
 477                 if (isquery)
 478                         psf->sf_gsresp = 0;
 479
 480                 if (AVAILABLE(skb) < sizeof(__be32) +
 481                     first*sizeof(struct igmpv3_grec)) {
 482                         if (truncate && !first)
 483                                 break;   /* truncate these */
 484                         if (pgr)
 485                                 pgr->grec_nsrcs = htons(scount);
 486                         if (skb)
 487                                 igmpv3_sendpack(skb);
 488                         skb = igmpv3_newpack(dev, dev->mtu);
 489                         first = 1;
 490                         scount = 0;
 491                 }
 492                 if (first) {
 493                         skb = add_grhead(skb, pmc, type, &pgr);
 494                         first = 0;
 495                 }
 496                 if (!skb)
 497                         return NULL;
 498                 psrc = (__be32 *)skb_put(skb, sizeof(__be32));
 499                 *psrc = psf->sf_inaddr;
 500                 scount++; stotal++;
 501                 if ((type == IGMPV3_ALLOW_NEW_SOURCES ||
 502                      type == IGMPV3_BLOCK_OLD_SOURCES) && psf->sf_crcount) {
 503                         psf->sf_crcount--;
 504                         if ((sdeleted || gdeleted) && psf->sf_crcount == 0) {
 505                                 if (psf_prev)
 506                                         psf_prev->sf_next = psf->sf_next;
 507                                 else
 508                                         *psf_list = psf->sf_next;
 509                                 kfree(psf);
 510                                 continue;
 511                         }
 512                 }
 513                 psf_prev = psf;
 514         }
 515
 516 empty_source:
 517         if (!stotal) {
 518                 if (type == IGMPV3_ALLOW_NEW_SOURCES ||
 519                     type == IGMPV3_BLOCK_OLD_SOURCES)
 520                         return skb;
 521                 if (pmc->crcount || isquery) {
 522                         /* make sure we have room for group header */
 523                         if (skb && AVAILABLE(skb)<sizeof(struct igmpv3_grec)) {
 524                                 igmpv3_sendpack(skb);
 525                                 skb = NULL; /* add_grhead will get a new one */
 526                         }
 527                         skb = add_grhead(skb, pmc, type, &pgr);
 528                 }
 529         }
 530         if (pgr)
 531                 pgr->grec_nsrcs = htons(scount);
 532
 533         if (isquery)
 534                 pmc->gsquery = 0;       /* clear query state on report */
 535         return skb;
 536 }
 537
 538 static int igmpv3_send_report(struct in_device *in_dev, struct ip_mc_list *pmc)
 539 {
 540         struct sk_buff *skb = NULL;
 541         int type;
 542
 543         if (!pmc) {
 544                 rcu_read_lock();
 545                 for_each_pmc_rcu(in_dev, pmc) {
 546                         if (pmc->multiaddr == IGMP_ALL_HOSTS)
 547                                 continue;
 548                         spin_lock_bh(&pmc->lock);
 549                         if (pmc->sfcount[MCAST_EXCLUDE])
 550                                 type = IGMPV3_MODE_IS_EXCLUDE;
 551                         else
 552                                 type = IGMPV3_MODE_IS_INCLUDE;
 553                         skb = add_grec(skb, pmc, type, 0, 0);
 554                         spin_unlock_bh(&pmc->lock);
 555                 }
 556                 rcu_read_unlock();
 557         } else {
 558                 spin_lock_bh(&pmc->lock);
 559                 if (pmc->sfcount[MCAST_EXCLUDE])
 560                         type = IGMPV3_MODE_IS_EXCLUDE;
 561                 else
 562                         type = IGMPV3_MODE_IS_INCLUDE;
 563                 skb = add_grec(skb, pmc, type, 0, 0);
 564                 spin_unlock_bh(&pmc->lock);
 565         }
 566         if (!skb)
 567                 return 0;
 568         return igmpv3_sendpack(skb);
 569 }
 570
 571 /*
 572  * remove zero-count source records from a source filter list
 573  */
 574 static void igmpv3_clear_zeros(struct ip_sf_list **ppsf)
 575 {
 576         struct ip_sf_list *psf_prev, *psf_next, *psf;
 577
 578         psf_prev = NULL;
 579         for (psf=*ppsf; psf; psf = psf_next) {
 580                 psf_next = psf->sf_next;
 581                 if (psf->sf_crcount == 0) {
 582                         if (psf_prev)
 583                                 psf_prev->sf_next = psf->sf_next;
 584                         else
 585                                 *ppsf = psf->sf_next;
 586                         kfree(psf);
 587                 } else
 588                         psf_prev = psf;
 589         }
 590 }
 591
 592 static void igmpv3_send_cr(struct in_device *in_dev)
 593 {
 594         struct ip_mc_list *pmc, *pmc_prev, *pmc_next;
 595         struct sk_buff *skb = NULL;
 596         int type, dtype;
 597
 598         rcu_read_lock();
 599         spin_lock_bh(&in_dev->mc_tomb_lock);
 600
 601         /* deleted MCA's */
 602         pmc_prev = NULL;
 603         for (pmc=in_dev->mc_tomb; pmc; pmc=pmc_next) {
 604                 pmc_next = pmc->next;
 605                 if (pmc->sfmode == MCAST_INCLUDE) {
 606                         type = IGMPV3_BLOCK_OLD_SOURCES;
 607                         dtype = IGMPV3_BLOCK_OLD_SOURCES;
 608                         skb = add_grec(skb, pmc, type, 1, 0);
 609                         skb = add_grec(skb, pmc, dtype, 1, 1);
 610                 }
 611                 if (pmc->crcount) {
 612                         if (pmc->sfmode == MCAST_EXCLUDE) {
 613                                 type = IGMPV3_CHANGE_TO_INCLUDE;
 614                                 skb = add_grec(skb, pmc, type, 1, 0);
 615                         }
 616                         pmc->crcount--;
 617                         if (pmc->crcount == 0) {
 618                                 igmpv3_clear_zeros(&pmc->tomb);
 619                                 igmpv3_clear_zeros(&pmc->sources);
 620                         }
 621                 }
 622                 if (pmc->crcount == 0 && !pmc->tomb && !pmc->sources) {
 623                         if (pmc_prev)
 624                                 pmc_prev->next = pmc_next;
 625                         else
 626                                 in_dev->mc_tomb = pmc_next;
 627                         in_dev_put(pmc->interface);
 628                         kfree(pmc);
 629                 } else
 630                         pmc_prev = pmc;
 631         }
 632         spin_unlock_bh(&in_dev->mc_tomb_lock);
 633
 634         /* change recs */
 635         for_each_pmc_rcu(in_dev, pmc) {
 636                 spin_lock_bh(&pmc->lock);
 637                 if (pmc->sfcount[MCAST_EXCLUDE]) {
 638                         type = IGMPV3_BLOCK_OLD_SOURCES;
 639                         dtype = IGMPV3_ALLOW_NEW_SOURCES;
 640                 } else {
 641                         type = IGMPV3_ALLOW_NEW_SOURCES;
 642                         dtype = IGMPV3_BLOCK_OLD_SOURCES;
 643                 }
 644                 skb = add_grec(skb, pmc, type, 0, 0);
 645                 skb = add_grec(skb, pmc, dtype, 0, 1);  /* deleted sources */
 646
 647                 /* filter mode changes */
 648                 if (pmc->crcount) {
 649                         if (pmc->sfmode == MCAST_EXCLUDE)
 650                                 type = IGMPV3_CHANGE_TO_EXCLUDE;
 651                         else
 652                                 type = IGMPV3_CHANGE_TO_INCLUDE;
 653                         skb = add_grec(skb, pmc, type, 0, 0);
 654                         pmc->crcount--;
 655                 }
 656                 spin_unlock_bh(&pmc->lock);
 657         }
 658         rcu_read_unlock();
 659
 660         if (!skb)
 661                 return;
 662         (void) igmpv3_sendpack(skb);
 663 }
 664
 665 static int igmp_send_report(struct in_device *in_dev, struct ip_mc_list *pmc,
 666         int type)
 667 {
 668         struct sk_buff *skb;
 669         struct iphdr *iph;
 670         struct igmphdr *ih;
 671         struct rtable *rt;
 672         struct net_device *dev = in_dev->dev;
 673         struct net *net = dev_net(dev);
 674         __be32  group = pmc ? pmc->multiaddr : 0;
 675         struct flowi4 fl4;
 676         __be32  dst;
 677         int hlen, tlen;
 678
 679         if (type == IGMPV3_HOST_MEMBERSHIP_REPORT)
 680                 return igmpv3_send_report(in_dev, pmc);
 681         else if (type == IGMP_HOST_LEAVE_MESSAGE)
 682                 dst = IGMP_ALL_ROUTER;
 683         else
 684                 dst = group;
 685
 686         rt = ip_route_output_ports(net, &fl4, NULL, dst, 0,
 687                                    0, 0,
 688                                    IPPROTO_IGMP, 0, dev->ifindex);
 689         if (IS_ERR(rt))
 690                 return -1;
 691
 692         hlen = LL_RESERVED_SPACE(dev);
 693         tlen = dev->needed_tailroom;
 694         skb = alloc_skb(IGMP_SIZE + hlen + tlen, GFP_ATOMIC);
 695         if (skb == NULL) {
 696                 ip_rt_put(rt);
 697                 return -1;
 698         }
 699
 700         skb_dst_set(skb, &rt->dst);
 701
 702         skb_reserve(skb, hlen);
 703
 704         skb_reset_network_header(skb);
 705         iph = ip_hdr(skb);
 706         skb_put(skb, sizeof(struct iphdr) + 4);
 707
 708         iph->version  = 4;
 709         iph->ihl      = (sizeof(struct iphdr)+4)>>2;
 710         iph->tos      = 0xc0;
 711         iph->frag_off = htons(IP_DF);
 712         iph->ttl      = 1;
 713         iph->daddr    = dst;
 714         iph->saddr    = fl4.saddr;
 715         iph->protocol = IPPROTO_IGMP;
 716         ip_select_ident(skb, NULL);
 717         ((u8*)&iph[1])[0] = IPOPT_RA;
 718         ((u8*)&iph[1])[1] = 4;
 719         ((u8*)&iph[1])[2] = 0;
 720         ((u8*)&iph[1])[3] = 0;
 721
 722         ih = (struct igmphdr *)skb_put(skb, sizeof(struct igmphdr));
 723         ih->type = type;
 724         ih->code = 0;
 725         ih->csum = 0;
 726         ih->group = group;
 727         ih->csum = ip_compute_csum((void *)ih, sizeof(struct igmphdr));
 728
 729         return ip_local_out(skb);
 730 }
 731
 732 static void igmp_gq_timer_expire(unsigned long data)
 733 {
 734         struct in_device *in_dev = (struct in_device *)data;
 735
 736         in_dev->mr_gq_running = 0;
 737         igmpv3_send_report(in_dev, NULL);
 738         in_dev_put(in_dev);
 739 }
 740
 741 static void igmp_ifc_timer_expire(unsigned long data)
 742 {
 743         struct in_device *in_dev = (struct in_device *)data;
 744
 745         igmpv3_send_cr(in_dev);
 746         if (in_dev->mr_ifc_count) {
 747                 in_dev->mr_ifc_count--;
 748                 igmp_ifc_start_timer(in_dev, IGMP_Unsolicited_Report_Interval);
 749         }
 750         in_dev_put(in_dev);
 751 }
 752
 753 static void igmp_ifc_event(struct in_device *in_dev)
 754 {
 755         if (IGMP_V1_SEEN(in_dev) || IGMP_V2_SEEN(in_dev))
 756                 return;
 757         in_dev->mr_ifc_count = in_dev->mr_qrv ? in_dev->mr_qrv :
 758                 IGMP_Unsolicited_Report_Count;
 759         igmp_ifc_start_timer(in_dev, 1);
 760 }
 761
 762
 763 static void igmp_timer_expire(unsigned long data)
 764 {
 765         struct ip_mc_list *im=(struct ip_mc_list *)data;
 766         struct in_device *in_dev = im->interface;
 767
 768         spin_lock(&im->lock);
 769         im->tm_running = 0;
 770
 771         if (im->unsolicit_count) {
 772                 im->unsolicit_count--;
 773                 igmp_start_timer(im, IGMP_Unsolicited_Report_Interval);
 774         }
 775         im->reporter = 1;
 776         spin_unlock(&im->lock);
 777
 778         if (IGMP_V1_SEEN(in_dev))
 779                 igmp_send_report(in_dev, im, IGMP_HOST_MEMBERSHIP_REPORT);
 780         else if (IGMP_V2_SEEN(in_dev))
 781                 igmp_send_report(in_dev, im, IGMPV2_HOST_MEMBERSHIP_REPORT);
 782         else
 783                 igmp_send_report(in_dev, im, IGMPV3_HOST_MEMBERSHIP_REPORT);
 784
 785         ip_ma_put(im);
 786 }
 787
 788 /* mark EXCLUDE-mode sources */
 789 static int igmp_xmarksources(struct ip_mc_list *pmc, int nsrcs, __be32 *srcs)
 790 {
 791         struct ip_sf_list *psf;
 792         int i, scount;
 793
 794         scount = 0;
 795         for (psf=pmc->sources; psf; psf=psf->sf_next) {
 796                 if (scount == nsrcs)
 797                         break;
 798                 for (i=0; i<nsrcs; i++) {
 799                         /* skip inactive filters */
 800                         if (psf->sf_count[MCAST_INCLUDE] ||
 801                             pmc->sfcount[MCAST_EXCLUDE] !=
 802                             psf->sf_count[MCAST_EXCLUDE])
 803                                 continue;
 804                         if (srcs[i] == psf->sf_inaddr) {
 805                                 scount++;
 806                                 break;
 807                         }
 808                 }
 809         }
 810         pmc->gsquery = 0;
 811         if (scount == nsrcs)    /* all sources excluded */
 812                 return 0;
 813         return 1;
 814 }
 815
 816 static int igmp_marksources(struct ip_mc_list *pmc, int nsrcs, __be32 *srcs)
 817 {
 818         struct ip_sf_list *psf;
 819         int i, scount;
 820
 821         if (pmc->sfmode == MCAST_EXCLUDE)
 822                 return igmp_xmarksources(pmc, nsrcs, srcs);
 823
 824         /* mark INCLUDE-mode sources */
 825         scount = 0;
 826         for (psf=pmc->sources; psf; psf=psf->sf_next) {
 827                 if (scount == nsrcs)
 828                         break;
 829                 for (i=0; i<nsrcs; i++)
 830                         if (srcs[i] == psf->sf_inaddr) {
 831                                 psf->sf_gsresp = 1;
 832                                 scount++;
 833                                 break;
 834                         }
 835         }
 836         if (!scount) {
 837                 pmc->gsquery = 0;
 838                 return 0;
 839         }
 840         pmc->gsquery = 1;
 841         return 1;
 842 }
 843
 844 static void igmp_heard_report(struct in_device *in_dev, __be32 group)
 845 {
 846         struct ip_mc_list *im;
 847
 848         /* Timers are only set for non-local groups */
 849
 850         if (group == IGMP_ALL_HOSTS)
 851                 return;
 852
 853         rcu_read_lock();
 854         for_each_pmc_rcu(in_dev, im) {
 855                 if (im->multiaddr == group) {
 856                         igmp_stop_timer(im);
 857                         break;
 858                 }
 859         }
 860         rcu_read_unlock();
 861 }
 862
 863 static void igmp_heard_query(struct in_device *in_dev, struct sk_buff *skb,
 864         int len)
 865 {
 866         struct igmphdr          *ih = igmp_hdr(skb);
 867         struct igmpv3_query *ih3 = igmpv3_query_hdr(skb);
 868         struct ip_mc_list       *im;
 869         __be32                  group = ih->group;
 870         int                     max_delay;
 871         int                     mark = 0;
 872
 873
 874         if (len == 8) {
 875                 if (ih->code == 0) {
 876                         /* Alas, old v1 router presents here. */
 877
 878                         max_delay = IGMP_Query_Response_Interval;
 879                         in_dev->mr_v1_seen = jiffies +
 880                                 IGMP_V1_Router_Present_Timeout;
 881                         group = 0;
 882                 } else {
 883                         /* v2 router present */
 884                         max_delay = ih->code*(HZ/IGMP_TIMER_SCALE);
 885                         in_dev->mr_v2_seen = jiffies +
 886                                 IGMP_V2_Router_Present_Timeout;
 887                 }
 888                 /* cancel the interface change timer */
 889                 in_dev->mr_ifc_count = 0;
 890                 if (del_timer(&in_dev->mr_ifc_timer))
 891                         __in_dev_put(in_dev);
 892                 /* clear deleted report items */
 893                 igmpv3_clear_delrec(in_dev);
 894         } else if (len < 12) {
 895                 return; /* ignore bogus packet; freed by caller */
 896         } else if (IGMP_V1_SEEN(in_dev)) {
 897                 /* This is a v3 query with v1 queriers present */
 898                 max_delay = IGMP_Query_Response_Interval;
 899                 group = 0;
 900         } else if (IGMP_V2_SEEN(in_dev)) {
 901                 /* this is a v3 query with v2 queriers present;
 902                  * Interpretation of the max_delay code is problematic here.
 903                  * A real v2 host would use ih_code directly, while v3 has a
 904                  * different encoding. We use the v3 encoding as more likely
 905                  * to be intended in a v3 query.
 906                  */
 907                 max_delay = IGMPV3_MRC(ih3->code)*(HZ/IGMP_TIMER_SCALE);
 908                 if (!max_delay)
 909                         max_delay = 1;  /* can't mod w/ 0 */
 910         } else { /* v3 */
 911                 if (!pskb_may_pull(skb, sizeof(struct igmpv3_query)))
 912                         return;
 913
 914                 ih3 = igmpv3_query_hdr(skb);
 915                 if (ih3->nsrcs) {
 916                         if (!pskb_may_pull(skb, sizeof(struct igmpv3_query)
 917                                            + ntohs(ih3->nsrcs)*sizeof(__be32)))
 918                                 return;
 919                         ih3 = igmpv3_query_hdr(skb);
 920                 }
 921
 922                 max_delay = IGMPV3_MRC(ih3->code)*(HZ/IGMP_TIMER_SCALE);
 923                 if (!max_delay)
 924                         max_delay = 1;  /* can't mod w/ 0 */
 925                 in_dev->mr_maxdelay = max_delay;
 926                 if (ih3->qrv)
 927                         in_dev->mr_qrv = ih3->qrv;
 928                 if (!group) { /* general query */
 929                         if (ih3->nsrcs)
 930                                 return; /* no sources allowed */
 931                         igmp_gq_start_timer(in_dev);
 932                         return;
 933                 }
 934                 /* mark sources to include, if group & source-specific */
 935                 mark = ih3->nsrcs != 0;
 936         }
 937
 938         /*
 939          * - Start the timers in all of our membership records
 940          *   that the query applies to for the interface on
 941          *   which the query arrived excl. those that belong
 942          *   to a "local" group (224.0.0.X)
 943          * - For timers already running check if they need to
 944          *   be reset.
 945          * - Use the igmp->igmp_code field as the maximum
 946          *   delay possible
 947          */
 948         rcu_read_lock();
 949         for_each_pmc_rcu(in_dev, im) {
 950                 int changed;
 951
 952                 if (group && group != im->multiaddr)
 953                         continue;
 954                 if (im->multiaddr == IGMP_ALL_HOSTS)
 955                         continue;
 956                 spin_lock_bh(&im->lock);
 957                 if (im->tm_running)
 958                         im->gsquery = im->gsquery && mark;
 959                 else
 960                         im->gsquery = mark;
 961                 changed = !im->gsquery ||
 962                         igmp_marksources(im, ntohs(ih3->nsrcs), ih3->srcs);
 963                 spin_unlock_bh(&im->lock);
 964                 if (changed)
 965                         igmp_mod_timer(im, max_delay);
 966         }
 967         rcu_read_unlock();
 968 }
 969
 970 /* called in rcu_read_lock() section */
 971 int igmp_rcv(struct sk_buff *skb)
 972 {
 973         /* This basically follows the spec line by line -- see RFC1112 */
 974         struct igmphdr *ih;
 975         struct in_device *in_dev = __in_dev_get_rcu(skb->dev);
 976         int len = skb->len;
 977
 978         if (in_dev == NULL)
 979                 goto drop;
 980
 981         if (!pskb_may_pull(skb, sizeof(struct igmphdr)))
 982                 goto drop;
 983
 984         switch (skb->ip_summed) {
 985         case CHECKSUM_COMPLETE:
 986                 if (!csum_fold(skb->csum))
 987                         break;
 988                 /* fall through */
 989         case CHECKSUM_NONE:
 990                 skb->csum = 0;
 991                 if (__skb_checksum_complete(skb))
 992                         goto drop;
 993         }
 994
 995         ih = igmp_hdr(skb);
 996         switch (ih->type) {
 997         case IGMP_HOST_MEMBERSHIP_QUERY:
 998                 igmp_heard_query(in_dev, skb, len);
 999                 break;
1000         case IGMP_HOST_MEMBERSHIP_REPORT:
1001         case IGMPV2_HOST_MEMBERSHIP_REPORT:
1002                 /* Is it our report looped back? */
1003                 if (rt_is_output_route(skb_rtable(skb)))
1004                         break;
1005                 /* don't rely on MC router hearing unicast reports */
1006                 if (skb->pkt_type == PACKET_MULTICAST ||
1007                     skb->pkt_type == PACKET_BROADCAST)
1008                         igmp_heard_report(in_dev, ih->group);
1009                 break;
1010         case IGMP_PIM:
1011 #ifdef CONFIG_IP_PIMSM_V1
1012                 return pim_rcv_v1(skb);
1013 #endif
1014         case IGMPV3_HOST_MEMBERSHIP_REPORT:
1015         case IGMP_DVMRP:
1016         case IGMP_TRACE:
1017         case IGMP_HOST_LEAVE_MESSAGE:
1018         case IGMP_MTRACE:
1019         case IGMP_MTRACE_RESP:
1020                 break;
1021         default:
1022                 break;
1023         }
1024
1025 drop:
1026         kfree_skb(skb);
1027         return 0;
1028 }
1029
1030 #endif
1031
1032
1033 /*
1034  *      Add a filter to a device
1035  */
1036
1037 static void ip_mc_filter_add(struct in_device *in_dev, __be32 addr)
1038 {
1039         char buf[MAX_ADDR_LEN];
1040         struct net_device *dev = in_dev->dev;
1041
1042         /* Checking for IFF_MULTICAST here is WRONG-WRONG-WRONG.
1043            We will get multicast token leakage, when IFF_MULTICAST
1044            is changed. This check should be done in ndo_set_rx_mode
1045            routine. Something sort of:
1046            if (dev->mc_list && dev->flags&IFF_MULTICAST) { do it; }
1047            --ANK
1048            */
1049         if (arp_mc_map(addr, buf, dev, 0) == 0)
1050                 dev_mc_add(dev, buf);
1051 }
1052
1053 /*
1054  *      Remove a filter from a device
1055  */
1056
1057 static void ip_mc_filter_del(struct in_device *in_dev, __be32 addr)
1058 {
1059         char buf[MAX_ADDR_LEN];
1060         struct net_device *dev = in_dev->dev;
1061
1062         if (arp_mc_map(addr, buf, dev, 0) == 0)
1063                 dev_mc_del(dev, buf);
1064 }
1065
1066 #ifdef CONFIG_IP_MULTICAST
1067 /*
1068  * deleted ip_mc_list manipulation
1069  */
1070 static void igmpv3_add_delrec(struct in_device *in_dev, struct ip_mc_list *im)
1071 {
1072         struct ip_mc_list *pmc;
1073
1074         /* this is an "ip_mc_list" for convenience; only the fields below
1075          * are actually used. In particular, the refcnt and users are not
1076          * used for management of the delete list. Using the same structure
1077          * for deleted items allows change reports to use common code with
1078          * non-deleted or query-response MCA's.
1079          */
1080         pmc = kzalloc(sizeof(*pmc), GFP_KERNEL);
1081         if (!pmc)
1082                 return;
1083         spin_lock_bh(&im->lock);
1084         pmc->interface = im->interface;
1085         in_dev_hold(in_dev);
1086         pmc->multiaddr = im->multiaddr;
1087         pmc->crcount = in_dev->mr_qrv ? in_dev->mr_qrv :
1088                 IGMP_Unsolicited_Report_Count;
1089         pmc->sfmode = im->sfmode;
1090         if (pmc->sfmode == MCAST_INCLUDE) {
1091                 struct ip_sf_list *psf;
1092
1093                 pmc->tomb = im->tomb;
1094                 pmc->sources = im->sources;
1095                 im->tomb = im->sources = NULL;
1096                 for (psf=pmc->sources; psf; psf=psf->sf_next)
1097                         psf->sf_crcount = pmc->crcount;
1098         }
1099         spin_unlock_bh(&im->lock);
1100
1101         spin_lock_bh(&in_dev->mc_tomb_lock);
1102         pmc->next = in_dev->mc_tomb;
1103         in_dev->mc_tomb = pmc;
1104         spin_unlock_bh(&in_dev->mc_tomb_lock);
1105 }
1106
1107 /*
1108  * restore ip_mc_list deleted records
1109  */
1110 static void igmpv3_del_delrec(struct in_device *in_dev, struct ip_mc_list *im)
1111 {
1112         struct ip_mc_list *pmc, *pmc_prev;
1113         struct ip_sf_list *psf;
1114         __be32 multiaddr = im->multiaddr;
1115
1116         spin_lock_bh(&in_dev->mc_tomb_lock);
1117         pmc_prev = NULL;
1118         for (pmc=in_dev->mc_tomb; pmc; pmc=pmc->next) {
1119                 if (pmc->multiaddr == multiaddr)
1120                         break;
1121                 pmc_prev = pmc;
1122         }
1123         if (pmc) {
1124                 if (pmc_prev)
1125                         pmc_prev->next = pmc->next;
1126                 else
1127                         in_dev->mc_tomb = pmc->next;
1128         }
1129         spin_unlock_bh(&in_dev->mc_tomb_lock);
1130
1131         spin_lock_bh(&im->lock);
1132         if (pmc) {
1133                 im->interface = pmc->interface;
1134                 im->crcount = in_dev->mr_qrv ?: IGMP_Unsolicited_Report_Count;
1135                 im->sfmode = pmc->sfmode;
1136                 if (pmc->sfmode == MCAST_INCLUDE) {
1137                         im->tomb = pmc->tomb;
1138                         im->sources = pmc->sources;
1139                         for (psf = im->sources; psf; psf = psf->sf_next)
1140                                 psf->sf_crcount = im->crcount;
1141                 }
1142                 in_dev_put(pmc->interface);
1143                 kfree(pmc);
1144         }
1145         spin_unlock_bh(&im->lock);
1146 }
1147
1148 /*
1149  * flush ip_mc_list deleted records
1150  */
1151 static void igmpv3_clear_delrec(struct in_device *in_dev)
1152 {
1153         struct ip_mc_list *pmc, *nextpmc;
1154
1155         spin_lock_bh(&in_dev->mc_tomb_lock);
1156         pmc = in_dev->mc_tomb;
1157         in_dev->mc_tomb = NULL;
1158         spin_unlock_bh(&in_dev->mc_tomb_lock);
1159
1160         for (; pmc; pmc = nextpmc) {
1161                 nextpmc = pmc->next;
1162                 ip_mc_clear_src(pmc);
1163                 in_dev_put(pmc->interface);
1164                 kfree(pmc);
1165         }
1166         /* clear dead sources, too */
1167         rcu_read_lock();
1168         for_each_pmc_rcu(in_dev, pmc) {
1169                 struct ip_sf_list *psf, *psf_next;
1170
1171                 spin_lock_bh(&pmc->lock);
1172                 psf = pmc->tomb;
1173                 pmc->tomb = NULL;
1174                 spin_unlock_bh(&pmc->lock);
1175                 for (; psf; psf=psf_next) {
1176                         psf_next = psf->sf_next;
1177                         kfree(psf);
1178                 }
1179         }
1180         rcu_read_unlock();
1181 }
1182 #endif
1183
1184 static void igmp_group_dropped(struct ip_mc_list *im)
1185 {
1186         struct in_device *in_dev = im->interface;
1187 #ifdef CONFIG_IP_MULTICAST
1188         int reporter;
1189 #endif
1190
1191         if (im->loaded) {
1192                 im->loaded = 0;
1193                 ip_mc_filter_del(in_dev, im->multiaddr);
1194         }
1195
1196 #ifdef CONFIG_IP_MULTICAST
1197         if (im->multiaddr == IGMP_ALL_HOSTS)
1198                 return;
1199
1200         reporter = im->reporter;
1201         igmp_stop_timer(im);
1202
1203         if (!in_dev->dead) {
1204                 if (IGMP_V1_SEEN(in_dev))
1205                         return;
1206                 if (IGMP_V2_SEEN(in_dev)) {
1207                         if (reporter)
1208                                 igmp_send_report(in_dev, im, IGMP_HOST_LEAVE_MESSAGE);
1209                         return;
1210                 }
1211                 /* IGMPv3 */
1212                 igmpv3_add_delrec(in_dev, im);
1213
1214                 igmp_ifc_event(in_dev);
1215         }
1216 #endif
1217 }
1218
1219 static void igmp_group_added(struct ip_mc_list *im)
1220 {
1221         struct in_device *in_dev = im->interface;
1222
1223         if (im->loaded == 0) {
1224                 im->loaded = 1;
1225                 ip_mc_filter_add(in_dev, im->multiaddr);
1226         }
1227
1228 #ifdef CONFIG_IP_MULTICAST
1229         if (im->multiaddr == IGMP_ALL_HOSTS)
1230                 return;
1231
1232         if (in_dev->dead)
1233                 return;
1234         if (IGMP_V1_SEEN(in_dev) || IGMP_V2_SEEN(in_dev)) {
1235                 spin_lock_bh(&im->lock);
1236                 igmp_start_timer(im, IGMP_Initial_Report_Delay);
1237                 spin_unlock_bh(&im->lock);
1238                 return;
1239         }
1240         /* else, v3 */
1241
1242         im->crcount = in_dev->mr_qrv ? in_dev->mr_qrv :
1243                 IGMP_Unsolicited_Report_Count;
1244         igmp_ifc_event(in_dev);
1245 #endif
1246 }
1247
1248
1249 /*
1250  *      Multicast list managers
1251  */
1252
1253
1254 /*
1255  *      A socket has joined a multicast group on device dev.
1256  */
1257
1258 void ip_mc_inc_group(struct in_device *in_dev, __be32 addr)
1259 {
1260         struct ip_mc_list *im;
1261
1262         ASSERT_RTNL();
1263
1264         for_each_pmc_rtnl(in_dev, im) {
1265                 if (im->multiaddr == addr) {
1266                         im->users++;
1267                         ip_mc_add_src(in_dev, &addr, MCAST_EXCLUDE, 0, NULL, 0);
1268                         goto out;
1269                 }
1270         }
1271
1272         im = kzalloc(sizeof(*im), GFP_KERNEL);
1273         if (!im)
1274                 goto out;
1275
1276         im->users = 1;
1277         im->interface = in_dev;
1278         in_dev_hold(in_dev);
1279         im->multiaddr = addr;
1280         /* initial mode is (EX, empty) */
1281         im->sfmode = MCAST_EXCLUDE;
1282         im->sfcount[MCAST_EXCLUDE] = 1;
1283         atomic_set(&im->refcnt, 1);
1284         spin_lock_init(&im->lock);
1285 #ifdef CONFIG_IP_MULTICAST
1286         setup_timer(&im->timer, &igmp_timer_expire, (unsigned long)im);
1287         im->unsolicit_count = IGMP_Unsolicited_Report_Count;
1288 #endif
1289
1290         im->next_rcu = in_dev->mc_list;
1291         in_dev->mc_count++;
1292         rcu_assign_pointer(in_dev->mc_list, im);
1293
1294 #ifdef CONFIG_IP_MULTICAST
1295         igmpv3_del_delrec(in_dev, im);
1296 #endif
1297         igmp_group_added(im);
1298         if (!in_dev->dead)
1299                 ip_rt_multicast_event(in_dev);
1300 out:
1301         return;
1302 }
1303 EXPORT_SYMBOL(ip_mc_inc_group);
1304
1305 /*
1306  *      Resend IGMP JOIN report; used for bonding.
1307  *      Called with rcu_read_lock()
1308  */
1309 void ip_mc_rejoin_groups(struct in_device *in_dev)
1310 {
1311 #ifdef CONFIG_IP_MULTICAST
1312         struct ip_mc_list *im;
1313         int type;
1314
1315         for_each_pmc_rcu(in_dev, im) {
1316                 if (im->multiaddr == IGMP_ALL_HOSTS)
1317                         continue;
1318
1319                 /* a failover is happening and switches
1320                  * must be notified immediately
1321                  */
1322                 if (IGMP_V1_SEEN(in_dev))
1323                         type = IGMP_HOST_MEMBERSHIP_REPORT;
1324                 else if (IGMP_V2_SEEN(in_dev))
1325                         type = IGMPV2_HOST_MEMBERSHIP_REPORT;
1326                 else
1327                         type = IGMPV3_HOST_MEMBERSHIP_REPORT;
1328                 igmp_send_report(in_dev, im, type);
1329         }
1330 #endif
1331 }
1332 EXPORT_SYMBOL(ip_mc_rejoin_groups);
1333
1334 /*
1335  *      A socket has left a multicast group on device dev
1336  */
1337
1338 void ip_mc_dec_group(struct in_device *in_dev, __be32 addr)
1339 {
1340         struct ip_mc_list *i;
1341         struct ip_mc_list __rcu **ip;
1342
1343         ASSERT_RTNL();
1344
1345         for (ip = &in_dev->mc_list;
1346              (i = rtnl_dereference(*ip)) != NULL;
1347              ip = &i->next_rcu) {
1348                 if (i->multiaddr == addr) {
1349                         if (--i->users == 0) {
1350                                 *ip = i->next_rcu;
1351                                 in_dev->mc_count--;
1352                                 igmp_group_dropped(i);
1353                                 ip_mc_clear_src(i);
1354
1355                                 if (!in_dev->dead)
1356                                         ip_rt_multicast_event(in_dev);
1357
1358                                 ip_ma_put(i);
1359                                 return;
1360                         }
1361                         break;
1362                 }
1363         }
1364 }
1365 EXPORT_SYMBOL(ip_mc_dec_group);
1366
1367 /* Device changing type */
1368
1369 void ip_mc_unmap(struct in_device *in_dev)
1370 {
1371         struct ip_mc_list *pmc;
1372
1373         ASSERT_RTNL();
1374
1375         for_each_pmc_rtnl(in_dev, pmc)
1376                 igmp_group_dropped(pmc);
1377 }
1378
1379 void ip_mc_remap(struct in_device *in_dev)
1380 {
1381         struct ip_mc_list *pmc;
1382
1383         ASSERT_RTNL();
1384
1385         for_each_pmc_rtnl(in_dev, pmc) {
1386 #ifdef CONFIG_IP_MULTICAST
1387                 igmpv3_del_delrec(in_dev, pmc);
1388 #endif
1389                 igmp_group_added(pmc);
1390         }
1391 }
1392
1393 /* Device going down */
1394
1395 void ip_mc_down(struct in_device *in_dev)
1396 {
1397         struct ip_mc_list *pmc;
1398
1399         ASSERT_RTNL();
1400
1401         for_each_pmc_rtnl(in_dev, pmc)
1402                 igmp_group_dropped(pmc);
1403
1404 #ifdef CONFIG_IP_MULTICAST
1405         in_dev->mr_ifc_count = 0;
1406         if (del_timer(&in_dev->mr_ifc_timer))
1407                 __in_dev_put(in_dev);
1408         in_dev->mr_gq_running = 0;
1409         if (del_timer(&in_dev->mr_gq_timer))
1410                 __in_dev_put(in_dev);
1411 #endif
1412
1413         ip_mc_dec_group(in_dev, IGMP_ALL_HOSTS);
1414 }
1415
1416 void ip_mc_init_dev(struct in_device *in_dev)
1417 {
1418         ASSERT_RTNL();
1419
1420         in_dev->mc_tomb = NULL;
1421 #ifdef CONFIG_IP_MULTICAST
1422         in_dev->mr_gq_running = 0;
1423         setup_timer(&in_dev->mr_gq_timer, igmp_gq_timer_expire,
1424                         (unsigned long)in_dev);
1425         in_dev->mr_ifc_count = 0;
1426         in_dev->mc_count     = 0;
1427         setup_timer(&in_dev->mr_ifc_timer, igmp_ifc_timer_expire,
1428                         (unsigned long)in_dev);
1429         in_dev->mr_qrv = IGMP_Unsolicited_Report_Count;
1430 #endif
1431
1432         spin_lock_init(&in_dev->mc_tomb_lock);
1433 }
1434
1435 /* Device going up */
1436
1437 void ip_mc_up(struct in_device *in_dev)
1438 {
1439         struct ip_mc_list *pmc;
1440
1441         ASSERT_RTNL();
1442
1443         ip_mc_inc_group(in_dev, IGMP_ALL_HOSTS);
1444
1445         for_each_pmc_rtnl(in_dev, pmc) {
1446 #ifdef CONFIG_IP_MULTICAST
1447                 igmpv3_del_delrec(in_dev, pmc);
1448 #endif
1449                 igmp_group_added(pmc);
1450         }
1451 }
1452
1453 /*
1454  *      Device is about to be destroyed: clean up.
1455  */
1456
1457 void ip_mc_destroy_dev(struct in_device *in_dev)
1458 {
1459         struct ip_mc_list *i;
1460
1461         ASSERT_RTNL();
1462
1463         /* Deactivate timers */
1464         ip_mc_down(in_dev);
1465 #ifdef CONFIG_IP_MULTICAST
1466         igmpv3_clear_delrec(in_dev);
1467 #endif
1468
1469         while ((i = rtnl_dereference(in_dev->mc_list)) != NULL) {
1470                 in_dev->mc_list = i->next_rcu;
1471                 in_dev->mc_count--;
1472                 ip_ma_put(i);
1473         }
1474 }
1475
1476 /* RTNL is locked */
1477 static struct in_device *ip_mc_find_dev(struct net *net, struct ip_mreqn *imr)
1478 {
1479         struct net_device *dev = NULL;
1480         struct in_device *idev = NULL;
1481
1482         if (imr->imr_ifindex) {
1483                 idev = inetdev_by_index(net, imr->imr_ifindex);
1484                 return idev;
1485         }
1486         if (imr->imr_address.s_addr) {
1487                 dev = __ip_dev_find(net, imr->imr_address.s_addr, false);
1488                 if (!dev)
1489                         return NULL;
1490         }
1491
1492         if (!dev) {
1493                 struct rtable *rt = ip_route_output(net,
1494                                                     imr->imr_multiaddr.s_addr,
1495                                                     0, 0, 0);
1496                 if (!IS_ERR(rt)) {
1497                         dev = rt->dst.dev;
1498                         ip_rt_put(rt);
1499                 }
1500         }
1501         if (dev) {
1502                 imr->imr_ifindex = dev->ifindex;
1503                 idev = __in_dev_get_rtnl(dev);
1504         }
1505         return idev;
1506 }
1507
1508 /*
1509  *      Join a socket to a group
1510  */
1511 int sysctl_igmp_max_memberships __read_mostly = IP_MAX_MEMBERSHIPS;
1512 int sysctl_igmp_max_msf __read_mostly = IP_MAX_MSF;
1513
1514
1515 static int ip_mc_del1_src(struct ip_mc_list *pmc, int sfmode,
1516         __be32 *psfsrc)
1517 {
1518         struct ip_sf_list *psf, *psf_prev;
1519         int rv = 0;
1520
1521         psf_prev = NULL;
1522         for (psf=pmc->sources; psf; psf=psf->sf_next) {
1523                 if (psf->sf_inaddr == *psfsrc)
1524                         break;
1525                 psf_prev = psf;
1526         }
1527         if (!psf || psf->sf_count[sfmode] == 0) {
1528                 /* source filter not found, or count wrong =>  bug */
1529                 return -ESRCH;
1530         }
1531         psf->sf_count[sfmode]--;
1532         if (psf->sf_count[sfmode] == 0) {
1533                 ip_rt_multicast_event(pmc->interface);
1534         }
1535         if (!psf->sf_count[MCAST_INCLUDE] && !psf->sf_count[MCAST_EXCLUDE]) {
1536 #ifdef CONFIG_IP_MULTICAST
1537                 struct in_device *in_dev = pmc->interface;
1538 #endif
1539
1540                 /* no more filters for this source */
1541                 if (psf_prev)
1542                         psf_prev->sf_next = psf->sf_next;
1543                 else
1544                         pmc->sources = psf->sf_next;
1545 #ifdef CONFIG_IP_MULTICAST
1546                 if (psf->sf_oldin &&
1547                     !IGMP_V1_SEEN(in_dev) && !IGMP_V2_SEEN(in_dev)) {
1548                         psf->sf_crcount = in_dev->mr_qrv ? in_dev->mr_qrv :
1549                                 IGMP_Unsolicited_Report_Count;
1550                         psf->sf_next = pmc->tomb;
1551                         pmc->tomb = psf;
1552                         rv = 1;
1553                 } else
1554 #endif
1555                         kfree(psf);
1556         }
1557         return rv;
1558 }
1559
1560 #ifndef CONFIG_IP_MULTICAST
1561 #define igmp_ifc_event(x)       do { } while (0)
1562 #endif
1563
1564 static int ip_mc_del_src(struct in_device *in_dev, __be32 *pmca, int sfmode,
1565                          int sfcount, __be32 *psfsrc, int delta)
1566 {
1567         struct ip_mc_list *pmc;
1568         int     changerec = 0;
1569         int     i, err;
1570
1571         if (!in_dev)
1572                 return -ENODEV;
1573         rcu_read_lock();
1574         for_each_pmc_rcu(in_dev, pmc) {
1575                 if (*pmca == pmc->multiaddr)
1576                         break;
1577         }
1578         if (!pmc) {
1579                 /* MCA not found?? bug */
1580                 rcu_read_unlock();
1581                 return -ESRCH;
1582         }
1583         spin_lock_bh(&pmc->lock);
1584         rcu_read_unlock();
1585 #ifdef CONFIG_IP_MULTICAST
1586         sf_markstate(pmc);
1587 #endif
1588         if (!delta) {
1589                 err = -EINVAL;
1590                 if (!pmc->sfcount[sfmode])
1591                         goto out_unlock;
1592                 pmc->sfcount[sfmode]--;
1593         }
1594         err = 0;
1595         for (i=0; i<sfcount; i++) {
1596                 int rv = ip_mc_del1_src(pmc, sfmode, &psfsrc[i]);
1597
1598                 changerec |= rv > 0;
1599                 if (!err && rv < 0)
1600                         err = rv;
1601         }
1602         if (pmc->sfmode == MCAST_EXCLUDE &&
1603             pmc->sfcount[MCAST_EXCLUDE] == 0 &&
1604             pmc->sfcount[MCAST_INCLUDE]) {
1605 #ifdef CONFIG_IP_MULTICAST
1606                 struct ip_sf_list *psf;
1607 #endif
1608
1609                 /* filter mode change */
1610                 pmc->sfmode = MCAST_INCLUDE;
1611 #ifdef CONFIG_IP_MULTICAST
1612                 pmc->crcount = in_dev->mr_qrv ? in_dev->mr_qrv :
1613                         IGMP_Unsolicited_Report_Count;
1614                 in_dev->mr_ifc_count = pmc->crcount;
1615                 for (psf=pmc->sources; psf; psf = psf->sf_next)
1616                         psf->sf_crcount = 0;
1617                 igmp_ifc_event(pmc->interface);
1618         } else if (sf_setstate(pmc) || changerec) {
1619                 igmp_ifc_event(pmc->interface);
1620 #endif
1621         }
1622 out_unlock:
1623         spin_unlock_bh(&pmc->lock);
1624         return err;
1625 }
1626
1627 /*
1628  * Add multicast single-source filter to the interface list
1629  */
1630 static int ip_mc_add1_src(struct ip_mc_list *pmc, int sfmode,
1631         __be32 *psfsrc, int delta)
1632 {
1633         struct ip_sf_list *psf, *psf_prev;
1634
1635         psf_prev = NULL;
1636         for (psf=pmc->sources; psf; psf=psf->sf_next) {
1637                 if (psf->sf_inaddr == *psfsrc)
1638                         break;
1639                 psf_prev = psf;
1640         }
1641         if (!psf) {
1642                 psf = kzalloc(sizeof(*psf), GFP_ATOMIC);
1643                 if (!psf)
1644                         return -ENOBUFS;
1645                 psf->sf_inaddr = *psfsrc;
1646                 if (psf_prev) {
1647                         psf_prev->sf_next = psf;
1648                 } else
1649                         pmc->sources = psf;
1650         }
1651         psf->sf_count[sfmode]++;
1652         if (psf->sf_count[sfmode] == 1) {
1653                 ip_rt_multicast_event(pmc->interface);
1654         }
1655         return 0;
1656 }
1657
1658 #ifdef CONFIG_IP_MULTICAST
1659 static void sf_markstate(struct ip_mc_list *pmc)
1660 {
1661         struct ip_sf_list *psf;
1662         int mca_xcount = pmc->sfcount[MCAST_EXCLUDE];
1663
1664         for (psf=pmc->sources; psf; psf=psf->sf_next)
1665                 if (pmc->sfcount[MCAST_EXCLUDE]) {
1666                         psf->sf_oldin = mca_xcount ==
1667                                 psf->sf_count[MCAST_EXCLUDE] &&
1668                                 !psf->sf_count[MCAST_INCLUDE];
1669                 } else
1670                         psf->sf_oldin = psf->sf_count[MCAST_INCLUDE] != 0;
1671 }
1672
1673 static int sf_setstate(struct ip_mc_list *pmc)
1674 {
1675         struct ip_sf_list *psf, *dpsf;
1676         int mca_xcount = pmc->sfcount[MCAST_EXCLUDE];
1677         int qrv = pmc->interface->mr_qrv;
1678         int new_in, rv;
1679
1680         rv = 0;
1681         for (psf=pmc->sources; psf; psf=psf->sf_next) {
1682                 if (pmc->sfcount[MCAST_EXCLUDE]) {
1683                         new_in = mca_xcount == psf->sf_count[MCAST_EXCLUDE] &&
1684                                 !psf->sf_count[MCAST_INCLUDE];
1685                 } else
1686                         new_in = psf->sf_count[MCAST_INCLUDE] != 0;
1687                 if (new_in) {
1688                         if (!psf->sf_oldin) {
1689                                 struct ip_sf_list *prev = NULL;
1690
1691                                 for (dpsf=pmc->tomb; dpsf; dpsf=dpsf->sf_next) {
1692                                         if (dpsf->sf_inaddr == psf->sf_inaddr)
1693                                                 break;
1694                                         prev = dpsf;
1695                                 }
1696                                 if (dpsf) {
1697                                         if (prev)
1698                                                 prev->sf_next = dpsf->sf_next;
1699                                         else
1700                                                 pmc->tomb = dpsf->sf_next;
1701                                         kfree(dpsf);
1702                                 }
1703                                 psf->sf_crcount = qrv;
1704                                 rv++;
1705                         }
1706                 } else if (psf->sf_oldin) {
1707
1708                         psf->sf_crcount = 0;
1709                         /*
1710                          * add or update "delete" records if an active filter
1711                          * is now inactive
1712                          */
1713                         for (dpsf=pmc->tomb; dpsf; dpsf=dpsf->sf_next)
1714                                 if (dpsf->sf_inaddr == psf->sf_inaddr)
1715                                         break;
1716                         if (!dpsf) {
1717                                 dpsf = kmalloc(sizeof(*dpsf), GFP_ATOMIC);
1718                                 if (!dpsf)
1719                                         continue;
1720                                 *dpsf = *psf;
1721                                 /* pmc->lock held by callers */
1722                                 dpsf->sf_next = pmc->tomb;
1723                                 pmc->tomb = dpsf;
1724                         }
1725                         dpsf->sf_crcount = qrv;
1726                         rv++;
1727                 }
1728         }
1729         return rv;
1730 }
1731 #endif
1732
1733 /*
1734  * Add multicast source filter list to the interface list
1735  */
1736 static int ip_mc_add_src(struct in_device *in_dev, __be32 *pmca, int sfmode,
1737                          int sfcount, __be32 *psfsrc, int delta)
1738 {
1739         struct ip_mc_list *pmc;
1740         int     isexclude;
1741         int     i, err;
1742
1743         if (!in_dev)
1744                 return -ENODEV;
1745         rcu_read_lock();
1746         for_each_pmc_rcu(in_dev, pmc) {
1747                 if (*pmca == pmc->multiaddr)
1748                         break;
1749         }
1750         if (!pmc) {
1751                 /* MCA not found?? bug */
1752                 rcu_read_unlock();
1753                 return -ESRCH;
1754         }
1755         spin_lock_bh(&pmc->lock);
1756         rcu_read_unlock();
1757
1758 #ifdef CONFIG_IP_MULTICAST
1759         sf_markstate(pmc);
1760 #endif
1761         isexclude = pmc->sfmode == MCAST_EXCLUDE;
1762         if (!delta)
1763                 pmc->sfcount[sfmode]++;
1764         err = 0;
1765         for (i=0; i<sfcount; i++) {
1766                 err = ip_mc_add1_src(pmc, sfmode, &psfsrc[i], delta);
1767                 if (err)
1768                         break;
1769         }
1770         if (err) {
1771                 int j;
1772
1773                 if (!delta)
1774                         pmc->sfcount[sfmode]--;
1775                 for (j=0; j<i; j++)
1776                         (void) ip_mc_del1_src(pmc, sfmode, &psfsrc[j]);
1777         } else if (isexclude != (pmc->sfcount[MCAST_EXCLUDE] != 0)) {
1778 #ifdef CONFIG_IP_MULTICAST
1779                 struct ip_sf_list *psf;
1780                 in_dev = pmc->interface;
1781 #endif
1782
1783                 /* filter mode change */
1784                 if (pmc->sfcount[MCAST_EXCLUDE])
1785                         pmc->sfmode = MCAST_EXCLUDE;
1786                 else if (pmc->sfcount[MCAST_INCLUDE])
1787                         pmc->sfmode = MCAST_INCLUDE;
1788 #ifdef CONFIG_IP_MULTICAST
1789                 /* else no filters; keep old mode for reports */
1790
1791                 pmc->crcount = in_dev->mr_qrv ? in_dev->mr_qrv :
1792                         IGMP_Unsolicited_Report_Count;
1793                 in_dev->mr_ifc_count = pmc->crcount;
1794                 for (psf=pmc->sources; psf; psf = psf->sf_next)
1795                         psf->sf_crcount = 0;
1796                 igmp_ifc_event(in_dev);
1797         } else if (sf_setstate(pmc)) {
1798                 igmp_ifc_event(in_dev);
1799 #endif
1800         }
1801         spin_unlock_bh(&pmc->lock);
1802         return err;
1803 }
1804
1805 static void ip_mc_clear_src(struct ip_mc_list *pmc)
1806 {
1807         struct ip_sf_list *psf, *nextpsf;
1808
1809         for (psf=pmc->tomb; psf; psf=nextpsf) {
1810                 nextpsf = psf->sf_next;
1811                 kfree(psf);
1812         }
1813         pmc->tomb = NULL;
1814         for (psf=pmc->sources; psf; psf=nextpsf) {
1815                 nextpsf = psf->sf_next;
1816                 kfree(psf);
1817         }
1818         pmc->sources = NULL;
1819         pmc->sfmode = MCAST_EXCLUDE;
1820         pmc->sfcount[MCAST_INCLUDE] = 0;
1821         pmc->sfcount[MCAST_EXCLUDE] = 1;
1822 }
1823
1824
1825 /*
1826  * Join a multicast group
1827  */
1828 int ip_mc_join_group(struct sock *sk , struct ip_mreqn *imr)
1829 {
1830         int err;
1831         __be32 addr = imr->imr_multiaddr.s_addr;
1832         struct ip_mc_socklist *iml = NULL, *i;
1833         struct in_device *in_dev;
1834         struct inet_sock *inet = inet_sk(sk);
1835         struct net *net = sock_net(sk);
1836         int ifindex;
1837         int count = 0;
1838
1839         if (!ipv4_is_multicast(addr))
1840                 return -EINVAL;
1841
1842         rtnl_lock();
1843
1844         in_dev = ip_mc_find_dev(net, imr);
1845
1846         if (!in_dev) {
1847                 iml = NULL;
1848                 err = -ENODEV;
1849                 goto done;
1850         }
1851
1852         err = -EADDRINUSE;
1853         ifindex = imr->imr_ifindex;
1854         for_each_pmc_rtnl(inet, i) {
1855                 if (i->multi.imr_multiaddr.s_addr == addr &&
1856                     i->multi.imr_ifindex == ifindex)
1857                         goto done;
1858                 count++;
1859         }
1860         err = -ENOBUFS;
1861         if (count >= sysctl_igmp_max_memberships)
1862                 goto done;
1863         iml = sock_kmalloc(sk, sizeof(*iml), GFP_KERNEL);
1864         if (iml == NULL)
1865                 goto done;
1866
1867         memcpy(&iml->multi, imr, sizeof(*imr));
1868         iml->next_rcu = inet->mc_list;
1869         iml->sflist = NULL;
1870         iml->sfmode = MCAST_EXCLUDE;
1871         rcu_assign_pointer(inet->mc_list, iml);
1872         ip_mc_inc_group(in_dev, addr);
1873         err = 0;
1874 done:
1875         rtnl_unlock();
1876         return err;
1877 }
1878 EXPORT_SYMBOL(ip_mc_join_group);
1879
1880 static int ip_mc_leave_src(struct sock *sk, struct ip_mc_socklist *iml,
1881                            struct in_device *in_dev)
1882 {
1883         struct ip_sf_socklist *psf = rtnl_dereference(iml->sflist);
1884         int err;
1885
1886         if (psf == NULL) {
1887                 /* any-source empty exclude case */
1888                 return ip_mc_del_src(in_dev, &iml->multi.imr_multiaddr.s_addr,
1889                         iml->sfmode, 0, NULL, 0);
1890         }
1891         err = ip_mc_del_src(in_dev, &iml->multi.imr_multiaddr.s_addr,
1892                         iml->sfmode, psf->sl_count, psf->sl_addr, 0);
1893         RCU_INIT_POINTER(iml->sflist, NULL);
1894         /* decrease mem now to avoid the memleak warning */
1895         atomic_sub(IP_SFLSIZE(psf->sl_max), &sk->sk_omem_alloc);
1896         kfree_rcu(psf, rcu);
1897         return err;
1898 }
1899
1900 /*
1901  *      Ask a socket to leave a group.
1902  */
1903
1904 int ip_mc_leave_group(struct sock *sk, struct ip_mreqn *imr)
1905 {
1906         struct inet_sock *inet = inet_sk(sk);
1907         struct ip_mc_socklist *iml;
1908         struct ip_mc_socklist __rcu **imlp;
1909         struct in_device *in_dev;
1910         struct net *net = sock_net(sk);
1911         __be32 group = imr->imr_multiaddr.s_addr;
1912         u32 ifindex;
1913         int ret = -EADDRNOTAVAIL;
1914
1915         rtnl_lock();
1916         in_dev = ip_mc_find_dev(net, imr);
1917         if (!imr->imr_ifindex && !imr->imr_address.s_addr && !in_dev) {
1918                 ret = -ENODEV;
1919                 goto out;
1920         }
1921         ifindex = imr->imr_ifindex;
1922         for (imlp = &inet->mc_list;
1923              (iml = rtnl_dereference(*imlp)) != NULL;
1924              imlp = &iml->next_rcu) {
1925                 if (iml->multi.imr_multiaddr.s_addr != group)
1926                         continue;
1927                 if (ifindex) {
1928                         if (iml->multi.imr_ifindex != ifindex)
1929                                 continue;
1930                 } else if (imr->imr_address.s_addr && imr->imr_address.s_addr !=
1931                                 iml->multi.imr_address.s_addr)
1932                         continue;
1933
1934                 (void) ip_mc_leave_src(sk, iml, in_dev);
1935
1936                 *imlp = iml->next_rcu;
1937
1938                 if (in_dev)
1939                         ip_mc_dec_group(in_dev, group);
1940                 rtnl_unlock();
1941                 /* decrease mem now to avoid the memleak warning */
1942                 atomic_sub(sizeof(*iml), &sk->sk_omem_alloc);
1943                 kfree_rcu(iml, rcu);
1944                 return 0;
1945         }
1946 out:
1947         rtnl_unlock();
1948         return ret;
1949 }
1950
1951 int ip_mc_source(int add, int omode, struct sock *sk, struct
1952         ip_mreq_source *mreqs, int ifindex)
1953 {
1954         int err;
1955         struct ip_mreqn imr;
1956         __be32 addr = mreqs->imr_multiaddr;
1957         struct ip_mc_socklist *pmc;
1958         struct in_device *in_dev = NULL;
1959         struct inet_sock *inet = inet_sk(sk);
1960         struct ip_sf_socklist *psl;
1961         struct net *net = sock_net(sk);
1962         int leavegroup = 0;
1963         int i, j, rv;
1964
1965         if (!ipv4_is_multicast(addr))
1966                 return -EINVAL;
1967
1968         rtnl_lock();
1969
1970         imr.imr_multiaddr.s_addr = mreqs->imr_multiaddr;
1971         imr.imr_address.s_addr = mreqs->imr_interface;
1972         imr.imr_ifindex = ifindex;
1973         in_dev = ip_mc_find_dev(net, &imr);
1974
1975         if (!in_dev) {
1976                 err = -ENODEV;
1977                 goto done;
1978         }
1979         err = -EADDRNOTAVAIL;
1980
1981         for_each_pmc_rtnl(inet, pmc) {
1982                 if ((pmc->multi.imr_multiaddr.s_addr ==
1983                      imr.imr_multiaddr.s_addr) &&
1984                     (pmc->multi.imr_ifindex == imr.imr_ifindex))
1985                         break;
1986         }
1987         if (!pmc) {             /* must have a prior join */
1988                 err = -EINVAL;
1989                 goto done;
1990         }
1991         /* if a source filter was set, must be the same mode as before */
1992         if (pmc->sflist) {
1993                 if (pmc->sfmode != omode) {
1994                         err = -EINVAL;
1995                         goto done;
1996                 }
1997         } else if (pmc->sfmode != omode) {
1998                 /* allow mode switches for empty-set filters */
1999                 ip_mc_add_src(in_dev, &mreqs->imr_multiaddr, omode, 0, NULL, 0);
2000                 ip_mc_del_src(in_dev, &mreqs->imr_multiaddr, pmc->sfmode, 0,
2001                         NULL, 0);
2002                 pmc->sfmode = omode;
2003         }
2004
2005         psl = rtnl_dereference(pmc->sflist);
2006         if (!add) {
2007                 if (!psl)
2008                         goto done;      /* err = -EADDRNOTAVAIL */
2009                 rv = !0;
2010                 for (i=0; i<psl->sl_count; i++) {
2011                         rv = memcmp(&psl->sl_addr[i], &mreqs->imr_sourceaddr,
2012                                 sizeof(__be32));
2013                         if (rv == 0)
2014                                 break;
2015                 }
2016                 if (rv)         /* source not found */
2017                         goto done;      /* err = -EADDRNOTAVAIL */
2018
2019                 /* special case - (INCLUDE, empty) == LEAVE_GROUP */
2020                 if (psl->sl_count == 1 && omode == MCAST_INCLUDE) {
2021                         leavegroup = 1;
2022                         goto done;
2023                 }
2024
2025                 /* update the interface filter */
2026                 ip_mc_del_src(in_dev, &mreqs->imr_multiaddr, omode, 1,
2027                         &mreqs->imr_sourceaddr, 1);
2028
2029                 for (j=i+1; j<psl->sl_count; j++)
2030                         psl->sl_addr[j-1] = psl->sl_addr[j];
2031                 psl->sl_count--;
2032                 err = 0;
2033                 goto done;
2034         }
2035         /* else, add a new source to the filter */
2036
2037         if (psl && psl->sl_count >= sysctl_igmp_max_msf) {
2038                 err = -ENOBUFS;
2039                 goto done;
2040         }
2041         if (!psl || psl->sl_count == psl->sl_max) {
2042                 struct ip_sf_socklist *newpsl;
2043                 int count = IP_SFBLOCK;
2044
2045                 if (psl)
2046                         count += psl->sl_max;
2047                 newpsl = sock_kmalloc(sk, IP_SFLSIZE(count), GFP_KERNEL);
2048                 if (!newpsl) {
2049                         err = -ENOBUFS;
2050                         goto done;
2051                 }
2052                 newpsl->sl_max = count;
2053                 newpsl->sl_count = count - IP_SFBLOCK;
2054                 if (psl) {
2055                         for (i=0; i<psl->sl_count; i++)
2056                                 newpsl->sl_addr[i] = psl->sl_addr[i];
2057                         /* decrease mem now to avoid the memleak warning */
2058                         atomic_sub(IP_SFLSIZE(psl->sl_max), &sk->sk_omem_alloc);
2059                         kfree_rcu(psl, rcu);
2060                 }
2061                 rcu_assign_pointer(pmc->sflist, newpsl);
2062                 psl = newpsl;
2063         }
2064         rv = 1; /* > 0 for insert logic below if sl_count is 0 */
2065         for (i=0; i<psl->sl_count; i++) {
2066                 rv = memcmp(&psl->sl_addr[i], &mreqs->imr_sourceaddr,
2067                         sizeof(__be32));
2068                 if (rv == 0)
2069                         break;
2070         }
2071         if (rv == 0)            /* address already there is an error */
2072                 goto done;
2073         for (j=psl->sl_count-1; j>=i; j--)
2074                 psl->sl_addr[j+1] = psl->sl_addr[j];
2075         psl->sl_addr[i] = mreqs->imr_sourceaddr;
2076         psl->sl_count++;
2077         err = 0;
2078         /* update the interface list */
2079         ip_mc_add_src(in_dev, &mreqs->imr_multiaddr, omode, 1,
2080                 &mreqs->imr_sourceaddr, 1);
2081 done:
2082         rtnl_unlock();
2083         if (leavegroup)
2084                 return ip_mc_leave_group(sk, &imr);
2085         return err;
2086 }
2087
2088 int ip_mc_msfilter(struct sock *sk, struct ip_msfilter *msf, int ifindex)
2089 {
2090         int err = 0;
2091         struct ip_mreqn imr;
2092         __be32 addr = msf->imsf_multiaddr;
2093         struct ip_mc_socklist *pmc;
2094         struct in_device *in_dev;
2095         struct inet_sock *inet = inet_sk(sk);
2096         struct ip_sf_socklist *newpsl, *psl;
2097         struct net *net = sock_net(sk);
2098         int leavegroup = 0;
2099
2100         if (!ipv4_is_multicast(addr))
2101                 return -EINVAL;
2102         if (msf->imsf_fmode != MCAST_INCLUDE &&
2103             msf->imsf_fmode != MCAST_EXCLUDE)
2104                 return -EINVAL;
2105
2106         rtnl_lock();
2107
2108         imr.imr_multiaddr.s_addr = msf->imsf_multiaddr;
2109         imr.imr_address.s_addr = msf->imsf_interface;
2110         imr.imr_ifindex = ifindex;
2111         in_dev = ip_mc_find_dev(net, &imr);
2112
2113         if (!in_dev) {
2114                 err = -ENODEV;
2115                 goto done;
2116         }
2117
2118         /* special case - (INCLUDE, empty) == LEAVE_GROUP */
2119         if (msf->imsf_fmode == MCAST_INCLUDE && msf->imsf_numsrc == 0) {
2120                 leavegroup = 1;
2121                 goto done;
2122         }
2123
2124         for_each_pmc_rtnl(inet, pmc) {
2125                 if (pmc->multi.imr_multiaddr.s_addr == msf->imsf_multiaddr &&
2126                     pmc->multi.imr_ifindex == imr.imr_ifindex)
2127                         break;
2128         }
2129         if (!pmc) {             /* must have a prior join */
2130                 err = -EINVAL;
2131                 goto done;
2132         }
2133         if (msf->imsf_numsrc) {
2134                 newpsl = sock_kmalloc(sk, IP_SFLSIZE(msf->imsf_numsrc),
2135                                                            GFP_KERNEL);
2136                 if (!newpsl) {
2137                         err = -ENOBUFS;
2138                         goto done;
2139                 }
2140                 newpsl->sl_max = newpsl->sl_count = msf->imsf_numsrc;
2141                 memcpy(newpsl->sl_addr, msf->imsf_slist,
2142                         msf->imsf_numsrc * sizeof(msf->imsf_slist[0]));
2143                 err = ip_mc_add_src(in_dev, &msf->imsf_multiaddr,
2144                         msf->imsf_fmode, newpsl->sl_count, newpsl->sl_addr, 0);
2145                 if (err) {
2146                         sock_kfree_s(sk, newpsl, IP_SFLSIZE(newpsl->sl_max));
2147                         goto done;
2148                 }
2149         } else {
2150                 newpsl = NULL;
2151                 (void) ip_mc_add_src(in_dev, &msf->imsf_multiaddr,
2152                                      msf->imsf_fmode, 0, NULL, 0);
2153         }
2154         psl = rtnl_dereference(pmc->sflist);
2155         if (psl) {
2156                 (void) ip_mc_del_src(in_dev, &msf->imsf_multiaddr, pmc->sfmode,
2157                         psl->sl_count, psl->sl_addr, 0);
2158                 /* decrease mem now to avoid the memleak warning */
2159                 atomic_sub(IP_SFLSIZE(psl->sl_max), &sk->sk_omem_alloc);
2160                 kfree_rcu(psl, rcu);
2161         } else
2162                 (void) ip_mc_del_src(in_dev, &msf->imsf_multiaddr, pmc->sfmode,
2163                         0, NULL, 0);
2164         rcu_assign_pointer(pmc->sflist, newpsl);
2165         pmc->sfmode = msf->imsf_fmode;
2166         err = 0;
2167 done:
2168         rtnl_unlock();
2169         if (leavegroup)
2170                 err = ip_mc_leave_group(sk, &imr);
2171         return err;
2172 }
2173
2174 int ip_mc_msfget(struct sock *sk, struct ip_msfilter *msf,
2175         struct ip_msfilter __user *optval, int __user *optlen)
2176 {
2177         int err, len, count, copycount;
2178         struct ip_mreqn imr;
2179         __be32 addr = msf->imsf_multiaddr;
2180         struct ip_mc_socklist *pmc;
2181         struct in_device *in_dev;
2182         struct inet_sock *inet = inet_sk(sk);
2183         struct ip_sf_socklist *psl;
2184         struct net *net = sock_net(sk);
2185
2186         if (!ipv4_is_multicast(addr))
2187                 return -EINVAL;
2188
2189         rtnl_lock();
2190
2191         imr.imr_multiaddr.s_addr = msf->imsf_multiaddr;
2192         imr.imr_address.s_addr = msf->imsf_interface;
2193         imr.imr_ifindex = 0;
2194         in_dev = ip_mc_find_dev(net, &imr);
2195
2196         if (!in_dev) {
2197                 err = -ENODEV;
2198                 goto done;
2199         }
2200         err = -EADDRNOTAVAIL;
2201
2202         for_each_pmc_rtnl(inet, pmc) {
2203                 if (pmc->multi.imr_multiaddr.s_addr == msf->imsf_multiaddr &&
2204                     pmc->multi.imr_ifindex == imr.imr_ifindex)
2205                         break;
2206         }
2207         if (!pmc)               /* must have a prior join */
2208                 goto done;
2209         msf->imsf_fmode = pmc->sfmode;
2210         psl = rtnl_dereference(pmc->sflist);
2211         rtnl_unlock();
2212         if (!psl) {
2213                 len = 0;
2214                 count = 0;
2215         } else {
2216                 count = psl->sl_count;
2217         }
2218         copycount = count < msf->imsf_numsrc ? count : msf->imsf_numsrc;
2219         len = copycount * sizeof(psl->sl_addr[0]);
2220         msf->imsf_numsrc = count;
2221         if (put_user(IP_MSFILTER_SIZE(copycount), optlen) ||
2222             copy_to_user(optval, msf, IP_MSFILTER_SIZE(0))) {
2223                 return -EFAULT;
2224         }
2225         if (len &&
2226             copy_to_user(&optval->imsf_slist[0], psl->sl_addr, len))
2227                 return -EFAULT;
2228         return 0;
2229 done:
2230         rtnl_unlock();
2231         return err;
2232 }
2233
2234 int ip_mc_gsfget(struct sock *sk, struct group_filter *gsf,
2235         struct group_filter __user *optval, int __user *optlen)
2236 {
2237         int err, i, count, copycount;
2238         struct sockaddr_in *psin;
2239         __be32 addr;
2240         struct ip_mc_socklist *pmc;
2241         struct inet_sock *inet = inet_sk(sk);
2242         struct ip_sf_socklist *psl;
2243
2244         psin = (struct sockaddr_in *)&gsf->gf_group;
2245         if (psin->sin_family != AF_INET)
2246                 return -EINVAL;
2247         addr = psin->sin_addr.s_addr;
2248         if (!ipv4_is_multicast(addr))
2249                 return -EINVAL;
2250
2251         rtnl_lock();
2252
2253         err = -EADDRNOTAVAIL;
2254
2255         for_each_pmc_rtnl(inet, pmc) {
2256                 if (pmc->multi.imr_multiaddr.s_addr == addr &&
2257                     pmc->multi.imr_ifindex == gsf->gf_interface)
2258                         break;
2259         }
2260         if (!pmc)               /* must have a prior join */
2261                 goto done;
2262         gsf->gf_fmode = pmc->sfmode;
2263         psl = rtnl_dereference(pmc->sflist);
2264         rtnl_unlock();
2265         count = psl ? psl->sl_count : 0;
2266         copycount = count < gsf->gf_numsrc ? count : gsf->gf_numsrc;
2267         gsf->gf_numsrc = count;
2268         if (put_user(GROUP_FILTER_SIZE(copycount), optlen) ||
2269             copy_to_user(optval, gsf, GROUP_FILTER_SIZE(0))) {
2270                 return -EFAULT;
2271         }
2272         for (i=0; i<copycount; i++) {
2273                 struct sockaddr_storage ss;
2274
2275                 psin = (struct sockaddr_in *)&ss;
2276                 memset(&ss, 0, sizeof(ss));
2277                 psin->sin_family = AF_INET;
2278                 psin->sin_addr.s_addr = psl->sl_addr[i];
2279                 if (copy_to_user(&optval->gf_slist[i], &ss, sizeof(ss)))
2280                         return -EFAULT;
2281         }
2282         return 0;
2283 done:
2284         rtnl_unlock();
2285         return err;
2286 }
2287
2288 /*
2289  * check if a multicast source filter allows delivery for a given <src,dst,intf>
2290  */
2291 int ip_mc_sf_allow(struct sock *sk, __be32 loc_addr, __be32 rmt_addr, int dif)
2292 {
2293         struct inet_sock *inet = inet_sk(sk);
2294         struct ip_mc_socklist *pmc;
2295         struct ip_sf_socklist *psl;
2296         int i;
2297         int ret;
2298
2299         ret = 1;
2300         if (!ipv4_is_multicast(loc_addr))
2301                 goto out;
2302
2303         rcu_read_lock();
2304         for_each_pmc_rcu(inet, pmc) {
2305                 if (pmc->multi.imr_multiaddr.s_addr == loc_addr &&
2306                     pmc->multi.imr_ifindex == dif)
2307                         break;
2308         }
2309         ret = inet->mc_all;
2310         if (!pmc)
2311                 goto unlock;
2312         psl = rcu_dereference(pmc->sflist);
2313         ret = (pmc->sfmode == MCAST_EXCLUDE);
2314         if (!psl)
2315                 goto unlock;
2316
2317         for (i=0; i<psl->sl_count; i++) {
2318                 if (psl->sl_addr[i] == rmt_addr)
2319                         break;
2320         }
2321         ret = 0;
2322         if (pmc->sfmode == MCAST_INCLUDE && i >= psl->sl_count)
2323                 goto unlock;
2324         if (pmc->sfmode == MCAST_EXCLUDE && i < psl->sl_count)
2325                 goto unlock;
2326         ret = 1;
2327 unlock:
2328         rcu_read_unlock();
2329 out:
2330         return ret;
2331 }
2332
2333 /*
2334  *      A socket is closing.
2335  */
2336
2337 void ip_mc_drop_socket(struct sock *sk)
2338 {
2339         struct inet_sock *inet = inet_sk(sk);
2340         struct ip_mc_socklist *iml;
2341         struct net *net = sock_net(sk);
2342
2343         if (inet->mc_list == NULL)
2344                 return;
2345
2346         rtnl_lock();
2347         while ((iml = rtnl_dereference(inet->mc_list)) != NULL) {
2348                 struct in_device *in_dev;
2349
2350                 inet->mc_list = iml->next_rcu;
2351                 in_dev = inetdev_by_index(net, iml->multi.imr_ifindex);
2352                 (void) ip_mc_leave_src(sk, iml, in_dev);
2353                 if (in_dev != NULL)
2354                         ip_mc_dec_group(in_dev, iml->multi.imr_multiaddr.s_addr);
2355                 /* decrease mem now to avoid the memleak warning */
2356                 atomic_sub(sizeof(*iml), &sk->sk_omem_alloc);
2357                 kfree_rcu(iml, rcu);
2358         }
2359         rtnl_unlock();
2360 }
2361
2362 /* called with rcu_read_lock() */
2363 int ip_check_mc_rcu(struct in_device *in_dev, __be32 mc_addr, __be32 src_addr, u16 proto)
2364 {
2365         struct ip_mc_list *im;
2366         struct ip_sf_list *psf;
2367         int rv = 0;
2368
2369         for_each_pmc_rcu(in_dev, im) {
2370                 if (im->multiaddr == mc_addr)
2371                         break;
2372         }
2373         if (im && proto == IPPROTO_IGMP) {
2374                 rv = 1;
2375         } else if (im) {
2376                 if (src_addr) {
2377                         for (psf=im->sources; psf; psf=psf->sf_next) {
2378                                 if (psf->sf_inaddr == src_addr)
2379                                         break;
2380                         }
2381                         if (psf)
2382                                 rv = psf->sf_count[MCAST_INCLUDE] ||
2383                                         psf->sf_count[MCAST_EXCLUDE] !=
2384                                         im->sfcount[MCAST_EXCLUDE];
2385                         else
2386                                 rv = im->sfcount[MCAST_EXCLUDE] != 0;
2387                 } else
2388                         rv = 1; /* unspecified source; tentatively allow */
2389         }
2390         return rv;
2391 }
2392
2393 #if defined(CONFIG_PROC_FS)
2394 struct igmp_mc_iter_state {
2395         struct seq_net_private p;
2396         struct net_device *dev;
2397         struct in_device *in_dev;
2398 };
2399
2400 #define igmp_mc_seq_private(seq)        ((struct igmp_mc_iter_state *)(seq)->private)
2401
2402 static inline struct ip_mc_list *igmp_mc_get_first(struct seq_file *seq)
2403 {
2404         struct net *net = seq_file_net(seq);
2405         struct ip_mc_list *im = NULL;
2406         struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2407
2408         state->in_dev = NULL;
2409         for_each_netdev_rcu(net, state->dev) {
2410                 struct in_device *in_dev;
2411
2412                 in_dev = __in_dev_get_rcu(state->dev);
2413                 if (!in_dev)
2414                         continue;
2415                 im = rcu_dereference(in_dev->mc_list);
2416                 if (im) {
2417                         state->in_dev = in_dev;
2418                         break;
2419                 }
2420         }
2421         return im;
2422 }
2423
2424 static struct ip_mc_list *igmp_mc_get_next(struct seq_file *seq, struct ip_mc_list *im)
2425 {
2426         struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2427
2428         im = rcu_dereference(im->next_rcu);
2429         while (!im) {
2430                 state->dev = next_net_device_rcu(state->dev);
2431                 if (!state->dev) {
2432                         state->in_dev = NULL;
2433                         break;
2434                 }
2435                 state->in_dev = __in_dev_get_rcu(state->dev);
2436                 if (!state->in_dev)
2437                         continue;
2438                 im = rcu_dereference(state->in_dev->mc_list);
2439         }
2440         return im;
2441 }
2442
2443 static struct ip_mc_list *igmp_mc_get_idx(struct seq_file *seq, loff_t pos)
2444 {
2445         struct ip_mc_list *im = igmp_mc_get_first(seq);
2446         if (im)
2447                 while (pos && (im = igmp_mc_get_next(seq, im)) != NULL)
2448                         --pos;
2449         return pos ? NULL : im;
2450 }
2451
2452 static void *igmp_mc_seq_start(struct seq_file *seq, loff_t *pos)
2453         __acquires(rcu)
2454 {
2455         rcu_read_lock();
2456         return *pos ? igmp_mc_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
2457 }
2458
2459 static void *igmp_mc_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2460 {
2461         struct ip_mc_list *im;
2462         if (v == SEQ_START_TOKEN)
2463                 im = igmp_mc_get_first(seq);
2464         else
2465                 im = igmp_mc_get_next(seq, v);
2466         ++*pos;
2467         return im;
2468 }
2469
2470 static void igmp_mc_seq_stop(struct seq_file *seq, void *v)
2471         __releases(rcu)
2472 {
2473         struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2474
2475         state->in_dev = NULL;
2476         state->dev = NULL;
2477         rcu_read_unlock();
2478 }
2479
2480 static int igmp_mc_seq_show(struct seq_file *seq, void *v)
2481 {
2482         if (v == SEQ_START_TOKEN)
2483                 seq_puts(seq,
2484                          "Idx\tDevice    : Count Querier\tGroup    Users Timer\tReporter\n");
2485         else {
2486                 struct ip_mc_list *im = (struct ip_mc_list *)v;
2487                 struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2488                 char   *querier;
2489 #ifdef CONFIG_IP_MULTICAST
2490                 querier = IGMP_V1_SEEN(state->in_dev) ? "V1" :
2491                           IGMP_V2_SEEN(state->in_dev) ? "V2" :
2492                           "V3";
2493 #else
2494                 querier = "NONE";
2495 #endif
2496
2497                 if (rcu_dereference(state->in_dev->mc_list) == im) {
2498                         seq_printf(seq, "%d\t%-10s: %5d %7s\n",
2499                                    state->dev->ifindex, state->dev->name, state->in_dev->mc_count, querier);
2500                 }
2501
2502                 seq_printf(seq,
2503                            "\t\t\t\t%08X %5d %d:%08lX\t\t%d\n",
2504                            im->multiaddr, im->users,
2505                            im->tm_running, im->tm_running ?
2506                            jiffies_to_clock_t(im->timer.expires-jiffies) : 0,
2507                            im->reporter);
2508         }
2509         return 0;
2510 }
2511
2512 static const struct seq_operations igmp_mc_seq_ops = {
2513         .start  =       igmp_mc_seq_start,
2514         .next   =       igmp_mc_seq_next,
2515         .stop   =       igmp_mc_seq_stop,
2516         .show   =       igmp_mc_seq_show,
2517 };
2518
2519 static int igmp_mc_seq_open(struct inode *inode, struct file *file)
2520 {
2521         return seq_open_net(inode, file, &igmp_mc_seq_ops,
2522                         sizeof(struct igmp_mc_iter_state));
2523 }
2524
2525 static const struct file_operations igmp_mc_seq_fops = {
2526         .owner          =       THIS_MODULE,
2527         .open           =       igmp_mc_seq_open,
2528         .read           =       seq_read,
2529         .llseek         =       seq_lseek,
2530         .release        =       seq_release_net,
2531 };
2532
2533 struct igmp_mcf_iter_state {
2534         struct seq_net_private p;
2535         struct net_device *dev;
2536         struct in_device *idev;
2537         struct ip_mc_list *im;
2538 };
2539
2540 #define igmp_mcf_seq_private(seq)       ((struct igmp_mcf_iter_state *)(seq)->private)
2541
2542 static inline struct ip_sf_list *igmp_mcf_get_first(struct seq_file *seq)
2543 {
2544         struct net *net = seq_file_net(seq);
2545         struct ip_sf_list *psf = NULL;
2546         struct ip_mc_list *im = NULL;
2547         struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2548
2549         state->idev = NULL;
2550         state->im = NULL;
2551         for_each_netdev_rcu(net, state->dev) {
2552                 struct in_device *idev;
2553                 idev = __in_dev_get_rcu(state->dev);
2554                 if (unlikely(idev == NULL))
2555                         continue;
2556                 im = rcu_dereference(idev->mc_list);
2557                 if (likely(im != NULL)) {
2558                         spin_lock_bh(&im->lock);
2559                         psf = im->sources;
2560                         if (likely(psf != NULL)) {
2561                                 state->im = im;
2562                                 state->idev = idev;
2563                                 break;
2564                         }
2565                         spin_unlock_bh(&im->lock);
2566                 }
2567         }
2568         return psf;
2569 }
2570
2571 static struct ip_sf_list *igmp_mcf_get_next(struct seq_file *seq, struct ip_sf_list *psf)
2572 {
2573         struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2574
2575         psf = psf->sf_next;
2576         while (!psf) {
2577                 spin_unlock_bh(&state->im->lock);
2578                 state->im = state->im->next;
2579                 while (!state->im) {
2580                         state->dev = next_net_device_rcu(state->dev);
2581                         if (!state->dev) {
2582                                 state->idev = NULL;
2583                                 goto out;
2584                         }
2585                         state->idev = __in_dev_get_rcu(state->dev);
2586                         if (!state->idev)
2587                                 continue;
2588                         state->im = rcu_dereference(state->idev->mc_list);
2589                 }
2590                 if (!state->im)
2591                         break;
2592                 spin_lock_bh(&state->im->lock);
2593                 psf = state->im->sources;
2594         }
2595 out:
2596         return psf;
2597 }
2598
2599 static struct ip_sf_list *igmp_mcf_get_idx(struct seq_file *seq, loff_t pos)
2600 {
2601         struct ip_sf_list *psf = igmp_mcf_get_first(seq);
2602         if (psf)
2603                 while (pos && (psf = igmp_mcf_get_next(seq, psf)) != NULL)
2604                         --pos;
2605         return pos ? NULL : psf;
2606 }
2607
2608 static void *igmp_mcf_seq_start(struct seq_file *seq, loff_t *pos)
2609         __acquires(rcu)
2610 {
2611         rcu_read_lock();
2612         return *pos ? igmp_mcf_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
2613 }
2614
2615 static void *igmp_mcf_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2616 {
2617         struct ip_sf_list *psf;
2618         if (v == SEQ_START_TOKEN)
2619                 psf = igmp_mcf_get_first(seq);
2620         else
2621                 psf = igmp_mcf_get_next(seq, v);
2622         ++*pos;
2623         return psf;
2624 }
2625
2626 static void igmp_mcf_seq_stop(struct seq_file *seq, void *v)
2627         __releases(rcu)
2628 {
2629         struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2630         if (likely(state->im != NULL)) {
2631                 spin_unlock_bh(&state->im->lock);
2632                 state->im = NULL;
2633         }
2634         state->idev = NULL;
2635         state->dev = NULL;
2636         rcu_read_unlock();
2637 }
2638
2639 static int igmp_mcf_seq_show(struct seq_file *seq, void *v)
2640 {
2641         struct ip_sf_list *psf = (struct ip_sf_list *)v;
2642         struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2643
2644         if (v == SEQ_START_TOKEN) {
2645                 seq_printf(seq,
2646                            "%3s %6s "
2647                            "%10s %10s %6s %6s\n", "Idx",
2648                            "Device", "MCA",
2649                            "SRC", "INC", "EXC");
2650         } else {
2651                 seq_printf(seq,
2652                            "%3d %6.6s 0x%08x "
2653                            "0x%08x %6lu %6lu\n",
2654                            state->dev->ifindex, state->dev->name,
2655                            ntohl(state->im->multiaddr),
2656                            ntohl(psf->sf_inaddr),
2657                            psf->sf_count[MCAST_INCLUDE],
2658                            psf->sf_count[MCAST_EXCLUDE]);
2659         }
2660         return 0;
2661 }
2662
2663 static const struct seq_operations igmp_mcf_seq_ops = {
2664         .start  =       igmp_mcf_seq_start,
2665         .next   =       igmp_mcf_seq_next,
2666         .stop   =       igmp_mcf_seq_stop,
2667         .show   =       igmp_mcf_seq_show,
2668 };
2669
2670 static int igmp_mcf_seq_open(struct inode *inode, struct file *file)
2671 {
2672         return seq_open_net(inode, file, &igmp_mcf_seq_ops,
2673                         sizeof(struct igmp_mcf_iter_state));
2674 }
2675
2676 static const struct file_operations igmp_mcf_seq_fops = {
2677         .owner          =       THIS_MODULE,
2678         .open           =       igmp_mcf_seq_open,
2679         .read           =       seq_read,
2680         .llseek         =       seq_lseek,
2681         .release        =       seq_release_net,
2682 };
2683
2684 static int __net_init igmp_net_init(struct net *net)
2685 {
2686         struct proc_dir_entry *pde;
2687
2688         pde = proc_net_fops_create(net, "igmp", S_IRUGO, &igmp_mc_seq_fops);
2689         if (!pde)
2690                 goto out_igmp;
2691         pde = proc_net_fops_create(net, "mcfilter", S_IRUGO, &igmp_mcf_seq_fops);
2692         if (!pde)
2693                 goto out_mcfilter;
2694         return 0;
2695
2696 out_mcfilter:
2697         proc_net_remove(net, "igmp");
2698 out_igmp:
2699         return -ENOMEM;
2700 }
2701
2702 static void __net_exit igmp_net_exit(struct net *net)
2703 {
2704         proc_net_remove(net, "mcfilter");
2705         proc_net_remove(net, "igmp");
2706 }
2707
2708 static struct pernet_operations igmp_net_ops = {
2709         .init = igmp_net_init,
2710         .exit = igmp_net_exit,
2711 };
2712
2713 int __init igmp_mc_proc_init(void)
2714 {
2715         return register_pernet_subsys(&igmp_net_ops);
2716 }
2717 #endif