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
  93 #include <net/net_namespace.h>
  94 #include <net/arp.h>
  95 #include <net/ip.h>
  96 #include <net/protocol.h>
  97 #include <net/route.h>
  98 #include <net/sock.h>
  99 #include <net/checksum.h>
 100 #include <linux/netfilter_ipv4.h>
 101 #ifdef CONFIG_IP_MROUTE
 102 #include <linux/mroute.h>
 103 #endif
 104 #ifdef CONFIG_PROC_FS
 105 #include <linux/proc_fs.h>
 106 #include <linux/seq_file.h>
 107 #endif
 108
 109 #define IP_MAX_MEMBERSHIPS      20
 110 #define IP_MAX_MSF              10
 111
 112 #ifdef CONFIG_IP_MULTICAST
 113 /* Parameter names and values are taken from igmp-v2-06 draft */
 114
 115 #define IGMP_V1_Router_Present_Timeout          (400*HZ)
 116 #define IGMP_V2_Router_Present_Timeout          (400*HZ)
 117 #define IGMP_Unsolicited_Report_Interval        (10*HZ)
 118 #define IGMP_Query_Response_Interval            (10*HZ)
 119 #define IGMP_Unsolicited_Report_Count           2
 120
 121
 122 #define IGMP_Initial_Report_Delay               (1)
 123
 124 /* IGMP_Initial_Report_Delay is not from IGMP specs!
 125  * IGMP specs require to report membership immediately after
 126  * joining a group, but we delay the first report by a
 127  * small interval. It seems more natural and still does not
 128  * contradict to specs provided this delay is small enough.
 129  */
 130
 131 #define IGMP_V1_SEEN(in_dev) \
 132         (IPV4_DEVCONF_ALL(dev_net(in_dev->dev), FORCE_IGMP_VERSION) == 1 || \
 133          IN_DEV_CONF_GET((in_dev), FORCE_IGMP_VERSION) == 1 || \
 134          ((in_dev)->mr_v1_seen && \
 135           time_before(jiffies, (in_dev)->mr_v1_seen)))
 136 #define IGMP_V2_SEEN(in_dev) \
 137         (IPV4_DEVCONF_ALL(dev_net(in_dev->dev), FORCE_IGMP_VERSION) == 2 || \
 138          IN_DEV_CONF_GET((in_dev), FORCE_IGMP_VERSION) == 2 || \
 139          ((in_dev)->mr_v2_seen && \
 140           time_before(jiffies, (in_dev)->mr_v2_seen)))
 141
 142 static void igmpv3_add_delrec(struct in_device *in_dev, struct ip_mc_list *im);
 143 static void igmpv3_del_delrec(struct in_device *in_dev, __be32 multiaddr);
 144 static void igmpv3_clear_delrec(struct in_device *in_dev);
 145 static int sf_setstate(struct ip_mc_list *pmc);
 146 static void sf_markstate(struct ip_mc_list *pmc);
 147 #endif
 148 static void ip_mc_clear_src(struct ip_mc_list *pmc);
 149 static int ip_mc_add_src(struct in_device *in_dev, __be32 *pmca, int sfmode,
 150                          int sfcount, __be32 *psfsrc, int delta);
 151
 152
 153 static void ip_mc_list_reclaim(struct rcu_head *head)
 154 {
 155         kfree(container_of(head, struct ip_mc_list, rcu));
 156 }
 157
 158 static void ip_ma_put(struct ip_mc_list *im)
 159 {
 160         if (atomic_dec_and_test(&im->refcnt)) {
 161                 in_dev_put(im->interface);
 162                 call_rcu(&im->rcu, ip_mc_list_reclaim);
 163         }
 164 }
 165
 166 #define for_each_pmc_rcu(in_dev, pmc)                           \
 167         for (pmc = rcu_dereference(in_dev->mc_list);            \
 168              pmc != NULL;                                       \
 169              pmc = rcu_dereference(pmc->next_rcu))
 170
 171 #define for_each_pmc_rtnl(in_dev, pmc)                          \
 172         for (pmc = rtnl_dereference(in_dev->mc_list);           \
 173              pmc != NULL;                                       \
 174              pmc = rtnl_dereference(pmc->next_rcu))
 175
 176 #ifdef CONFIG_IP_MULTICAST
 177
 178 /*
 179  *      Timer management
 180  */
 181
 182 static void igmp_stop_timer(struct ip_mc_list *im)
 183 {
 184         spin_lock_bh(&im->lock);
 185         if (del_timer(&im->timer))
 186                 atomic_dec(&im->refcnt);
 187         im->tm_running = 0;
 188         im->reporter = 0;
 189         im->unsolicit_count = 0;
 190         spin_unlock_bh(&im->lock);
 191 }
 192
 193 /* It must be called with locked im->lock */
 194 static void igmp_start_timer(struct ip_mc_list *im, int max_delay)
 195 {
 196         int tv = net_random() % max_delay;
 197
 198         im->tm_running = 1;
 199         if (!mod_timer(&im->timer, jiffies+tv+2))
 200                 atomic_inc(&im->refcnt);
 201 }
 202
 203 static void igmp_gq_start_timer(struct in_device *in_dev)
 204 {
 205         int tv = net_random() % in_dev->mr_maxdelay;
 206
 207         in_dev->mr_gq_running = 1;
 208         if (!mod_timer(&in_dev->mr_gq_timer, jiffies+tv+2))
 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 static struct sk_buff *igmpv3_newpack(struct net_device *dev, int size)
 304 {
 305         struct sk_buff *skb;
 306         struct rtable *rt;
 307         struct iphdr *pip;
 308         struct igmpv3_report *pig;
 309         struct net *net = dev_net(dev);
 310
 311         skb = alloc_skb(size + LL_ALLOCATED_SPACE(dev), GFP_ATOMIC);
 312         if (skb == NULL)
 313                 return NULL;
 314
 315         {
 316                 struct flowi fl = { .oif = dev->ifindex,
 317                                     .nl_u = { .ip4_u = {
 318                                     .daddr = IGMPV3_ALL_MCR } },
 319                                     .proto = IPPROTO_IGMP };
 320                 if (ip_route_output_key(net, &rt, &fl)) {
 321                         kfree_skb(skb);
 322                         return NULL;
 323                 }
 324         }
 325         if (rt->rt_src == 0) {
 326                 kfree_skb(skb);
 327                 ip_rt_put(rt);
 328                 return NULL;
 329         }
 330
 331         skb_dst_set(skb, &rt->dst);
 332         skb->dev = dev;
 333
 334         skb_reserve(skb, LL_RESERVED_SPACE(dev));
 335
 336         skb_reset_network_header(skb);
 337         pip = ip_hdr(skb);
 338         skb_put(skb, sizeof(struct iphdr) + 4);
 339
 340         pip->version  = 4;
 341         pip->ihl      = (sizeof(struct iphdr)+4)>>2;
 342         pip->tos      = 0xc0;
 343         pip->frag_off = htons(IP_DF);
 344         pip->ttl      = 1;
 345         pip->daddr    = rt->rt_dst;
 346         pip->saddr    = rt->rt_src;
 347         pip->protocol = IPPROTO_IGMP;
 348         pip->tot_len  = 0;      /* filled in later */
 349         ip_select_ident(pip, &rt->dst, NULL);
 350         ((u8*)&pip[1])[0] = IPOPT_RA;
 351         ((u8*)&pip[1])[1] = 4;
 352         ((u8*)&pip[1])[2] = 0;
 353         ((u8*)&pip[1])[3] = 0;
 354
 355         skb->transport_header = skb->network_header + sizeof(struct iphdr) + 4;
 356         skb_put(skb, sizeof(*pig));
 357         pig = igmpv3_report_hdr(skb);
 358         pig->type = IGMPV3_HOST_MEMBERSHIP_REPORT;
 359         pig->resv1 = 0;
 360         pig->csum = 0;
 361         pig->resv2 = 0;
 362         pig->ngrec = 0;
 363         return skb;
 364 }
 365
 366 static int igmpv3_sendpack(struct sk_buff *skb)
 367 {
 368         struct igmphdr *pig = igmp_hdr(skb);
 369         const int igmplen = skb->tail - skb->transport_header;
 370
 371         pig->csum = ip_compute_csum(igmp_hdr(skb), igmplen);
 372
 373         return ip_local_out(skb);
 374 }
 375
 376 static int grec_size(struct ip_mc_list *pmc, int type, int gdel, int sdel)
 377 {
 378         return sizeof(struct igmpv3_grec) + 4*igmp_scount(pmc, type, gdel, sdel);
 379 }
 380
 381 static struct sk_buff *add_grhead(struct sk_buff *skb, struct ip_mc_list *pmc,
 382         int type, struct igmpv3_grec **ppgr)
 383 {
 384         struct net_device *dev = pmc->interface->dev;
 385         struct igmpv3_report *pih;
 386         struct igmpv3_grec *pgr;
 387
 388         if (!skb)
 389                 skb = igmpv3_newpack(dev, dev->mtu);
 390         if (!skb)
 391                 return NULL;
 392         pgr = (struct igmpv3_grec *)skb_put(skb, sizeof(struct igmpv3_grec));
 393         pgr->grec_type = type;
 394         pgr->grec_auxwords = 0;
 395         pgr->grec_nsrcs = 0;
 396         pgr->grec_mca = pmc->multiaddr;
 397         pih = igmpv3_report_hdr(skb);
 398         pih->ngrec = htons(ntohs(pih->ngrec)+1);
 399         *ppgr = pgr;
 400         return skb;
 401 }
 402
 403 #define AVAILABLE(skb) ((skb) ? ((skb)->dev ? (skb)->dev->mtu - (skb)->len : \
 404         skb_tailroom(skb)) : 0)
 405
 406 static struct sk_buff *add_grec(struct sk_buff *skb, struct ip_mc_list *pmc,
 407         int type, int gdeleted, int sdeleted)
 408 {
 409         struct net_device *dev = pmc->interface->dev;
 410         struct igmpv3_report *pih;
 411         struct igmpv3_grec *pgr = NULL;
 412         struct ip_sf_list *psf, *psf_next, *psf_prev, **psf_list;
 413         int scount, stotal, first, isquery, truncate;
 414
 415         if (pmc->multiaddr == IGMP_ALL_HOSTS)
 416                 return skb;
 417
 418         isquery = type == IGMPV3_MODE_IS_INCLUDE ||
 419                   type == IGMPV3_MODE_IS_EXCLUDE;
 420         truncate = type == IGMPV3_MODE_IS_EXCLUDE ||
 421                     type == IGMPV3_CHANGE_TO_EXCLUDE;
 422
 423         stotal = scount = 0;
 424
 425         psf_list = sdeleted ? &pmc->tomb : &pmc->sources;
 426
 427         if (!*psf_list)
 428                 goto empty_source;
 429
 430         pih = skb ? igmpv3_report_hdr(skb) : NULL;
 431
 432         /* EX and TO_EX get a fresh packet, if needed */
 433         if (truncate) {
 434                 if (pih && pih->ngrec &&
 435                     AVAILABLE(skb) < grec_size(pmc, type, gdeleted, sdeleted)) {
 436                         if (skb)
 437                                 igmpv3_sendpack(skb);
 438                         skb = igmpv3_newpack(dev, dev->mtu);
 439                 }
 440         }
 441         first = 1;
 442         psf_prev = NULL;
 443         for (psf=*psf_list; psf; psf=psf_next) {
 444                 __be32 *psrc;
 445
 446                 psf_next = psf->sf_next;
 447
 448                 if (!is_in(pmc, psf, type, gdeleted, sdeleted)) {
 449                         psf_prev = psf;
 450                         continue;
 451                 }
 452
 453                 /* clear marks on query responses */
 454                 if (isquery)
 455                         psf->sf_gsresp = 0;
 456
 457                 if (AVAILABLE(skb) < sizeof(__be32) +
 458                     first*sizeof(struct igmpv3_grec)) {
 459                         if (truncate && !first)
 460                                 break;   /* truncate these */
 461                         if (pgr)
 462                                 pgr->grec_nsrcs = htons(scount);
 463                         if (skb)
 464                                 igmpv3_sendpack(skb);
 465                         skb = igmpv3_newpack(dev, dev->mtu);
 466                         first = 1;
 467                         scount = 0;
 468                 }
 469                 if (first) {
 470                         skb = add_grhead(skb, pmc, type, &pgr);
 471                         first = 0;
 472                 }
 473                 if (!skb)
 474                         return NULL;
 475                 psrc = (__be32 *)skb_put(skb, sizeof(__be32));
 476                 *psrc = psf->sf_inaddr;
 477                 scount++; stotal++;
 478                 if ((type == IGMPV3_ALLOW_NEW_SOURCES ||
 479                      type == IGMPV3_BLOCK_OLD_SOURCES) && psf->sf_crcount) {
 480                         psf->sf_crcount--;
 481                         if ((sdeleted || gdeleted) && psf->sf_crcount == 0) {
 482                                 if (psf_prev)
 483                                         psf_prev->sf_next = psf->sf_next;
 484                                 else
 485                                         *psf_list = psf->sf_next;
 486                                 kfree(psf);
 487                                 continue;
 488                         }
 489                 }
 490                 psf_prev = psf;
 491         }
 492
 493 empty_source:
 494         if (!stotal) {
 495                 if (type == IGMPV3_ALLOW_NEW_SOURCES ||
 496                     type == IGMPV3_BLOCK_OLD_SOURCES)
 497                         return skb;
 498                 if (pmc->crcount || isquery) {
 499                         /* make sure we have room for group header */
 500                         if (skb && AVAILABLE(skb)<sizeof(struct igmpv3_grec)) {
 501                                 igmpv3_sendpack(skb);
 502                                 skb = NULL; /* add_grhead will get a new one */
 503                         }
 504                         skb = add_grhead(skb, pmc, type, &pgr);
 505                 }
 506         }
 507         if (pgr)
 508                 pgr->grec_nsrcs = htons(scount);
 509
 510         if (isquery)
 511                 pmc->gsquery = 0;       /* clear query state on report */
 512         return skb;
 513 }
 514
 515 static int igmpv3_send_report(struct in_device *in_dev, struct ip_mc_list *pmc)
 516 {
 517         struct sk_buff *skb = NULL;
 518         int type;
 519
 520         if (!pmc) {
 521                 rcu_read_lock();
 522                 for_each_pmc_rcu(in_dev, pmc) {
 523                         if (pmc->multiaddr == IGMP_ALL_HOSTS)
 524                                 continue;
 525                         spin_lock_bh(&pmc->lock);
 526                         if (pmc->sfcount[MCAST_EXCLUDE])
 527                                 type = IGMPV3_MODE_IS_EXCLUDE;
 528                         else
 529                                 type = IGMPV3_MODE_IS_INCLUDE;
 530                         skb = add_grec(skb, pmc, type, 0, 0);
 531                         spin_unlock_bh(&pmc->lock);
 532                 }
 533                 rcu_read_unlock();
 534         } else {
 535                 spin_lock_bh(&pmc->lock);
 536                 if (pmc->sfcount[MCAST_EXCLUDE])
 537                         type = IGMPV3_MODE_IS_EXCLUDE;
 538                 else
 539                         type = IGMPV3_MODE_IS_INCLUDE;
 540                 skb = add_grec(skb, pmc, type, 0, 0);
 541                 spin_unlock_bh(&pmc->lock);
 542         }
 543         if (!skb)
 544                 return 0;
 545         return igmpv3_sendpack(skb);
 546 }
 547
 548 /*
 549  * remove zero-count source records from a source filter list
 550  */
 551 static void igmpv3_clear_zeros(struct ip_sf_list **ppsf)
 552 {
 553         struct ip_sf_list *psf_prev, *psf_next, *psf;
 554
 555         psf_prev = NULL;
 556         for (psf=*ppsf; psf; psf = psf_next) {
 557                 psf_next = psf->sf_next;
 558                 if (psf->sf_crcount == 0) {
 559                         if (psf_prev)
 560                                 psf_prev->sf_next = psf->sf_next;
 561                         else
 562                                 *ppsf = psf->sf_next;
 563                         kfree(psf);
 564                 } else
 565                         psf_prev = psf;
 566         }
 567 }
 568
 569 static void igmpv3_send_cr(struct in_device *in_dev)
 570 {
 571         struct ip_mc_list *pmc, *pmc_prev, *pmc_next;
 572         struct sk_buff *skb = NULL;
 573         int type, dtype;
 574
 575         rcu_read_lock();
 576         spin_lock_bh(&in_dev->mc_tomb_lock);
 577
 578         /* deleted MCA's */
 579         pmc_prev = NULL;
 580         for (pmc=in_dev->mc_tomb; pmc; pmc=pmc_next) {
 581                 pmc_next = pmc->next;
 582                 if (pmc->sfmode == MCAST_INCLUDE) {
 583                         type = IGMPV3_BLOCK_OLD_SOURCES;
 584                         dtype = IGMPV3_BLOCK_OLD_SOURCES;
 585                         skb = add_grec(skb, pmc, type, 1, 0);
 586                         skb = add_grec(skb, pmc, dtype, 1, 1);
 587                 }
 588                 if (pmc->crcount) {
 589                         if (pmc->sfmode == MCAST_EXCLUDE) {
 590                                 type = IGMPV3_CHANGE_TO_INCLUDE;
 591                                 skb = add_grec(skb, pmc, type, 1, 0);
 592                         }
 593                         pmc->crcount--;
 594                         if (pmc->crcount == 0) {
 595                                 igmpv3_clear_zeros(&pmc->tomb);
 596                                 igmpv3_clear_zeros(&pmc->sources);
 597                         }
 598                 }
 599                 if (pmc->crcount == 0 && !pmc->tomb && !pmc->sources) {
 600                         if (pmc_prev)
 601                                 pmc_prev->next = pmc_next;
 602                         else
 603                                 in_dev->mc_tomb = pmc_next;
 604                         in_dev_put(pmc->interface);
 605                         kfree(pmc);
 606                 } else
 607                         pmc_prev = pmc;
 608         }
 609         spin_unlock_bh(&in_dev->mc_tomb_lock);
 610
 611         /* change recs */
 612         for_each_pmc_rcu(in_dev, pmc) {
 613                 spin_lock_bh(&pmc->lock);
 614                 if (pmc->sfcount[MCAST_EXCLUDE]) {
 615                         type = IGMPV3_BLOCK_OLD_SOURCES;
 616                         dtype = IGMPV3_ALLOW_NEW_SOURCES;
 617                 } else {
 618                         type = IGMPV3_ALLOW_NEW_SOURCES;
 619                         dtype = IGMPV3_BLOCK_OLD_SOURCES;
 620                 }
 621                 skb = add_grec(skb, pmc, type, 0, 0);
 622                 skb = add_grec(skb, pmc, dtype, 0, 1);  /* deleted sources */
 623
 624                 /* filter mode changes */
 625                 if (pmc->crcount) {
 626                         if (pmc->sfmode == MCAST_EXCLUDE)
 627                                 type = IGMPV3_CHANGE_TO_EXCLUDE;
 628                         else
 629                                 type = IGMPV3_CHANGE_TO_INCLUDE;
 630                         skb = add_grec(skb, pmc, type, 0, 0);
 631                         pmc->crcount--;
 632                 }
 633                 spin_unlock_bh(&pmc->lock);
 634         }
 635         rcu_read_unlock();
 636
 637         if (!skb)
 638                 return;
 639         (void) igmpv3_sendpack(skb);
 640 }
 641
 642 static int igmp_send_report(struct in_device *in_dev, struct ip_mc_list *pmc,
 643         int type)
 644 {
 645         struct sk_buff *skb;
 646         struct iphdr *iph;
 647         struct igmphdr *ih;
 648         struct rtable *rt;
 649         struct net_device *dev = in_dev->dev;
 650         struct net *net = dev_net(dev);
 651         __be32  group = pmc ? pmc->multiaddr : 0;
 652         __be32  dst;
 653
 654         if (type == IGMPV3_HOST_MEMBERSHIP_REPORT)
 655                 return igmpv3_send_report(in_dev, pmc);
 656         else if (type == IGMP_HOST_LEAVE_MESSAGE)
 657                 dst = IGMP_ALL_ROUTER;
 658         else
 659                 dst = group;
 660
 661         {
 662                 struct flowi fl = { .oif = dev->ifindex,
 663                                     .nl_u = { .ip4_u = { .daddr = dst } },
 664                                     .proto = IPPROTO_IGMP };
 665                 if (ip_route_output_key(net, &rt, &fl))
 666                         return -1;
 667         }
 668         if (rt->rt_src == 0) {
 669                 ip_rt_put(rt);
 670                 return -1;
 671         }
 672
 673         skb = alloc_skb(IGMP_SIZE+LL_ALLOCATED_SPACE(dev), GFP_ATOMIC);
 674         if (skb == NULL) {
 675                 ip_rt_put(rt);
 676                 return -1;
 677         }
 678
 679         skb_dst_set(skb, &rt->dst);
 680
 681         skb_reserve(skb, LL_RESERVED_SPACE(dev));
 682
 683         skb_reset_network_header(skb);
 684         iph = ip_hdr(skb);
 685         skb_put(skb, sizeof(struct iphdr) + 4);
 686
 687         iph->version  = 4;
 688         iph->ihl      = (sizeof(struct iphdr)+4)>>2;
 689         iph->tos      = 0xc0;
 690         iph->frag_off = htons(IP_DF);
 691         iph->ttl      = 1;
 692         iph->daddr    = dst;
 693         iph->saddr    = rt->rt_src;
 694         iph->protocol = IPPROTO_IGMP;
 695         ip_select_ident(iph, &rt->dst, NULL);
 696         ((u8*)&iph[1])[0] = IPOPT_RA;
 697         ((u8*)&iph[1])[1] = 4;
 698         ((u8*)&iph[1])[2] = 0;
 699         ((u8*)&iph[1])[3] = 0;
 700
 701         ih = (struct igmphdr *)skb_put(skb, sizeof(struct igmphdr));
 702         ih->type = type;
 703         ih->code = 0;
 704         ih->csum = 0;
 705         ih->group = group;
 706         ih->csum = ip_compute_csum((void *)ih, sizeof(struct igmphdr));
 707
 708         return ip_local_out(skb);
 709 }
 710
 711 static void igmp_gq_timer_expire(unsigned long data)
 712 {
 713         struct in_device *in_dev = (struct in_device *)data;
 714
 715         in_dev->mr_gq_running = 0;
 716         igmpv3_send_report(in_dev, NULL);
 717         __in_dev_put(in_dev);
 718 }
 719
 720 static void igmp_ifc_timer_expire(unsigned long data)
 721 {
 722         struct in_device *in_dev = (struct in_device *)data;
 723
 724         igmpv3_send_cr(in_dev);
 725         if (in_dev->mr_ifc_count) {
 726                 in_dev->mr_ifc_count--;
 727                 igmp_ifc_start_timer(in_dev, IGMP_Unsolicited_Report_Interval);
 728         }
 729         __in_dev_put(in_dev);
 730 }
 731
 732 static void igmp_ifc_event(struct in_device *in_dev)
 733 {
 734         if (IGMP_V1_SEEN(in_dev) || IGMP_V2_SEEN(in_dev))
 735                 return;
 736         in_dev->mr_ifc_count = in_dev->mr_qrv ? in_dev->mr_qrv :
 737                 IGMP_Unsolicited_Report_Count;
 738         igmp_ifc_start_timer(in_dev, 1);
 739 }
 740
 741
 742 static void igmp_timer_expire(unsigned long data)
 743 {
 744         struct ip_mc_list *im=(struct ip_mc_list *)data;
 745         struct in_device *in_dev = im->interface;
 746
 747         spin_lock(&im->lock);
 748         im->tm_running = 0;
 749
 750         if (im->unsolicit_count) {
 751                 im->unsolicit_count--;
 752                 igmp_start_timer(im, IGMP_Unsolicited_Report_Interval);
 753         }
 754         im->reporter = 1;
 755         spin_unlock(&im->lock);
 756
 757         if (IGMP_V1_SEEN(in_dev))
 758                 igmp_send_report(in_dev, im, IGMP_HOST_MEMBERSHIP_REPORT);
 759         else if (IGMP_V2_SEEN(in_dev))
 760                 igmp_send_report(in_dev, im, IGMPV2_HOST_MEMBERSHIP_REPORT);
 761         else
 762                 igmp_send_report(in_dev, im, IGMPV3_HOST_MEMBERSHIP_REPORT);
 763
 764         ip_ma_put(im);
 765 }
 766
 767 /* mark EXCLUDE-mode sources */
 768 static int igmp_xmarksources(struct ip_mc_list *pmc, int nsrcs, __be32 *srcs)
 769 {
 770         struct ip_sf_list *psf;
 771         int i, scount;
 772
 773         scount = 0;
 774         for (psf=pmc->sources; psf; psf=psf->sf_next) {
 775                 if (scount == nsrcs)
 776                         break;
 777                 for (i=0; i<nsrcs; i++) {
 778                         /* skip inactive filters */
 779                         if (pmc->sfcount[MCAST_INCLUDE] ||
 780                             pmc->sfcount[MCAST_EXCLUDE] !=
 781                             psf->sf_count[MCAST_EXCLUDE])
 782                                 continue;
 783                         if (srcs[i] == psf->sf_inaddr) {
 784                                 scount++;
 785                                 break;
 786                         }
 787                 }
 788         }
 789         pmc->gsquery = 0;
 790         if (scount == nsrcs)    /* all sources excluded */
 791                 return 0;
 792         return 1;
 793 }
 794
 795 static int igmp_marksources(struct ip_mc_list *pmc, int nsrcs, __be32 *srcs)
 796 {
 797         struct ip_sf_list *psf;
 798         int i, scount;
 799
 800         if (pmc->sfmode == MCAST_EXCLUDE)
 801                 return igmp_xmarksources(pmc, nsrcs, srcs);
 802
 803         /* mark INCLUDE-mode sources */
 804         scount = 0;
 805         for (psf=pmc->sources; psf; psf=psf->sf_next) {
 806                 if (scount == nsrcs)
 807                         break;
 808                 for (i=0; i<nsrcs; i++)
 809                         if (srcs[i] == psf->sf_inaddr) {
 810                                 psf->sf_gsresp = 1;
 811                                 scount++;
 812                                 break;
 813                         }
 814         }
 815         if (!scount) {
 816                 pmc->gsquery = 0;
 817                 return 0;
 818         }
 819         pmc->gsquery = 1;
 820         return 1;
 821 }
 822
 823 static void igmp_heard_report(struct in_device *in_dev, __be32 group)
 824 {
 825         struct ip_mc_list *im;
 826
 827         /* Timers are only set for non-local groups */
 828
 829         if (group == IGMP_ALL_HOSTS)
 830                 return;
 831
 832         rcu_read_lock();
 833         for_each_pmc_rcu(in_dev, im) {
 834                 if (im->multiaddr == group) {
 835                         igmp_stop_timer(im);
 836                         break;
 837                 }
 838         }
 839         rcu_read_unlock();
 840 }
 841
 842 static void igmp_heard_query(struct in_device *in_dev, struct sk_buff *skb,
 843         int len)
 844 {
 845         struct igmphdr          *ih = igmp_hdr(skb);
 846         struct igmpv3_query *ih3 = igmpv3_query_hdr(skb);
 847         struct ip_mc_list       *im;
 848         __be32                  group = ih->group;
 849         int                     max_delay;
 850         int                     mark = 0;
 851
 852
 853         if (len == 8) {
 854                 if (ih->code == 0) {
 855                         /* Alas, old v1 router presents here. */
 856
 857                         max_delay = IGMP_Query_Response_Interval;
 858                         in_dev->mr_v1_seen = jiffies +
 859                                 IGMP_V1_Router_Present_Timeout;
 860                         group = 0;
 861                 } else {
 862                         /* v2 router present */
 863                         max_delay = ih->code*(HZ/IGMP_TIMER_SCALE);
 864                         in_dev->mr_v2_seen = jiffies +
 865                                 IGMP_V2_Router_Present_Timeout;
 866                 }
 867                 /* cancel the interface change timer */
 868                 in_dev->mr_ifc_count = 0;
 869                 if (del_timer(&in_dev->mr_ifc_timer))
 870                         __in_dev_put(in_dev);
 871                 /* clear deleted report items */
 872                 igmpv3_clear_delrec(in_dev);
 873         } else if (len < 12) {
 874                 return; /* ignore bogus packet; freed by caller */
 875         } else if (IGMP_V1_SEEN(in_dev)) {
 876                 /* This is a v3 query with v1 queriers present */
 877                 max_delay = IGMP_Query_Response_Interval;
 878                 group = 0;
 879         } else if (IGMP_V2_SEEN(in_dev)) {
 880                 /* this is a v3 query with v2 queriers present;
 881                  * Interpretation of the max_delay code is problematic here.
 882                  * A real v2 host would use ih_code directly, while v3 has a
 883                  * different encoding. We use the v3 encoding as more likely
 884                  * to be intended in a v3 query.
 885                  */
 886                 max_delay = IGMPV3_MRC(ih3->code)*(HZ/IGMP_TIMER_SCALE);
 887         } else { /* v3 */
 888                 if (!pskb_may_pull(skb, sizeof(struct igmpv3_query)))
 889                         return;
 890
 891                 ih3 = igmpv3_query_hdr(skb);
 892                 if (ih3->nsrcs) {
 893                         if (!pskb_may_pull(skb, sizeof(struct igmpv3_query)
 894                                            + ntohs(ih3->nsrcs)*sizeof(__be32)))
 895                                 return;
 896                         ih3 = igmpv3_query_hdr(skb);
 897                 }
 898
 899                 max_delay = IGMPV3_MRC(ih3->code)*(HZ/IGMP_TIMER_SCALE);
 900                 if (!max_delay)
 901                         max_delay = 1;  /* can't mod w/ 0 */
 902                 in_dev->mr_maxdelay = max_delay;
 903                 if (ih3->qrv)
 904                         in_dev->mr_qrv = ih3->qrv;
 905                 if (!group) { /* general query */
 906                         if (ih3->nsrcs)
 907                                 return; /* no sources allowed */
 908                         igmp_gq_start_timer(in_dev);
 909                         return;
 910                 }
 911                 /* mark sources to include, if group & source-specific */
 912                 mark = ih3->nsrcs != 0;
 913         }
 914
 915         /*
 916          * - Start the timers in all of our membership records
 917          *   that the query applies to for the interface on
 918          *   which the query arrived excl. those that belong
 919          *   to a "local" group (224.0.0.X)
 920          * - For timers already running check if they need to
 921          *   be reset.
 922          * - Use the igmp->igmp_code field as the maximum
 923          *   delay possible
 924          */
 925         rcu_read_lock();
 926         for_each_pmc_rcu(in_dev, im) {
 927                 int changed;
 928
 929                 if (group && group != im->multiaddr)
 930                         continue;
 931                 if (im->multiaddr == IGMP_ALL_HOSTS)
 932                         continue;
 933                 spin_lock_bh(&im->lock);
 934                 if (im->tm_running)
 935                         im->gsquery = im->gsquery && mark;
 936                 else
 937                         im->gsquery = mark;
 938                 changed = !im->gsquery ||
 939                         igmp_marksources(im, ntohs(ih3->nsrcs), ih3->srcs);
 940                 spin_unlock_bh(&im->lock);
 941                 if (changed)
 942                         igmp_mod_timer(im, max_delay);
 943         }
 944         rcu_read_unlock();
 945 }
 946
 947 /* called in rcu_read_lock() section */
 948 int igmp_rcv(struct sk_buff *skb)
 949 {
 950         /* This basically follows the spec line by line -- see RFC1112 */
 951         struct igmphdr *ih;
 952         struct in_device *in_dev = __in_dev_get_rcu(skb->dev);
 953         int len = skb->len;
 954
 955         if (in_dev == NULL)
 956                 goto drop;
 957
 958         if (!pskb_may_pull(skb, sizeof(struct igmphdr)))
 959                 goto drop;
 960
 961         switch (skb->ip_summed) {
 962         case CHECKSUM_COMPLETE:
 963                 if (!csum_fold(skb->csum))
 964                         break;
 965                 /* fall through */
 966         case CHECKSUM_NONE:
 967                 skb->csum = 0;
 968                 if (__skb_checksum_complete(skb))
 969                         goto drop;
 970         }
 971
 972         ih = igmp_hdr(skb);
 973         switch (ih->type) {
 974         case IGMP_HOST_MEMBERSHIP_QUERY:
 975                 igmp_heard_query(in_dev, skb, len);
 976                 break;
 977         case IGMP_HOST_MEMBERSHIP_REPORT:
 978         case IGMPV2_HOST_MEMBERSHIP_REPORT:
 979                 /* Is it our report looped back? */
 980                 if (rt_is_output_route(skb_rtable(skb)))
 981                         break;
 982                 /* don't rely on MC router hearing unicast reports */
 983                 if (skb->pkt_type == PACKET_MULTICAST ||
 984                     skb->pkt_type == PACKET_BROADCAST)
 985                         igmp_heard_report(in_dev, ih->group);
 986                 break;
 987         case IGMP_PIM:
 988 #ifdef CONFIG_IP_PIMSM_V1
 989                 return pim_rcv_v1(skb);
 990 #endif
 991         case IGMPV3_HOST_MEMBERSHIP_REPORT:
 992         case IGMP_DVMRP:
 993         case IGMP_TRACE:
 994         case IGMP_HOST_LEAVE_MESSAGE:
 995         case IGMP_MTRACE:
 996         case IGMP_MTRACE_RESP:
 997                 break;
 998         default:
 999                 break;
1000         }
1001
1002 drop:
1003         kfree_skb(skb);
1004         return 0;
1005 }
1006
1007 #endif
1008
1009
1010 /*
1011  *      Add a filter to a device
1012  */
1013
1014 static void ip_mc_filter_add(struct in_device *in_dev, __be32 addr)
1015 {
1016         char buf[MAX_ADDR_LEN];
1017         struct net_device *dev = in_dev->dev;
1018
1019         /* Checking for IFF_MULTICAST here is WRONG-WRONG-WRONG.
1020            We will get multicast token leakage, when IFF_MULTICAST
1021            is changed. This check should be done in dev->set_multicast_list
1022            routine. Something sort of:
1023            if (dev->mc_list && dev->flags&IFF_MULTICAST) { do it; }
1024            --ANK
1025            */
1026         if (arp_mc_map(addr, buf, dev, 0) == 0)
1027                 dev_mc_add(dev, buf);
1028 }
1029
1030 /*
1031  *      Remove a filter from a device
1032  */
1033
1034 static void ip_mc_filter_del(struct in_device *in_dev, __be32 addr)
1035 {
1036         char buf[MAX_ADDR_LEN];
1037         struct net_device *dev = in_dev->dev;
1038
1039         if (arp_mc_map(addr, buf, dev, 0) == 0)
1040                 dev_mc_del(dev, buf);
1041 }
1042
1043 #ifdef CONFIG_IP_MULTICAST
1044 /*
1045  * deleted ip_mc_list manipulation
1046  */
1047 static void igmpv3_add_delrec(struct in_device *in_dev, struct ip_mc_list *im)
1048 {
1049         struct ip_mc_list *pmc;
1050
1051         /* this is an "ip_mc_list" for convenience; only the fields below
1052          * are actually used. In particular, the refcnt and users are not
1053          * used for management of the delete list. Using the same structure
1054          * for deleted items allows change reports to use common code with
1055          * non-deleted or query-response MCA's.
1056          */
1057         pmc = kzalloc(sizeof(*pmc), GFP_KERNEL);
1058         if (!pmc)
1059                 return;
1060         spin_lock_bh(&im->lock);
1061         pmc->interface = im->interface;
1062         in_dev_hold(in_dev);
1063         pmc->multiaddr = im->multiaddr;
1064         pmc->crcount = in_dev->mr_qrv ? in_dev->mr_qrv :
1065                 IGMP_Unsolicited_Report_Count;
1066         pmc->sfmode = im->sfmode;
1067         if (pmc->sfmode == MCAST_INCLUDE) {
1068                 struct ip_sf_list *psf;
1069
1070                 pmc->tomb = im->tomb;
1071                 pmc->sources = im->sources;
1072                 im->tomb = im->sources = NULL;
1073                 for (psf=pmc->sources; psf; psf=psf->sf_next)
1074                         psf->sf_crcount = pmc->crcount;
1075         }
1076         spin_unlock_bh(&im->lock);
1077
1078         spin_lock_bh(&in_dev->mc_tomb_lock);
1079         pmc->next = in_dev->mc_tomb;
1080         in_dev->mc_tomb = pmc;
1081         spin_unlock_bh(&in_dev->mc_tomb_lock);
1082 }
1083
1084 static void igmpv3_del_delrec(struct in_device *in_dev, __be32 multiaddr)
1085 {
1086         struct ip_mc_list *pmc, *pmc_prev;
1087         struct ip_sf_list *psf, *psf_next;
1088
1089         spin_lock_bh(&in_dev->mc_tomb_lock);
1090         pmc_prev = NULL;
1091         for (pmc=in_dev->mc_tomb; pmc; pmc=pmc->next) {
1092                 if (pmc->multiaddr == multiaddr)
1093                         break;
1094                 pmc_prev = pmc;
1095         }
1096         if (pmc) {
1097                 if (pmc_prev)
1098                         pmc_prev->next = pmc->next;
1099                 else
1100                         in_dev->mc_tomb = pmc->next;
1101         }
1102         spin_unlock_bh(&in_dev->mc_tomb_lock);
1103         if (pmc) {
1104                 for (psf=pmc->tomb; psf; psf=psf_next) {
1105                         psf_next = psf->sf_next;
1106                         kfree(psf);
1107                 }
1108                 in_dev_put(pmc->interface);
1109                 kfree(pmc);
1110         }
1111 }
1112
1113 static void igmpv3_clear_delrec(struct in_device *in_dev)
1114 {
1115         struct ip_mc_list *pmc, *nextpmc;
1116
1117         spin_lock_bh(&in_dev->mc_tomb_lock);
1118         pmc = in_dev->mc_tomb;
1119         in_dev->mc_tomb = NULL;
1120         spin_unlock_bh(&in_dev->mc_tomb_lock);
1121
1122         for (; pmc; pmc = nextpmc) {
1123                 nextpmc = pmc->next;
1124                 ip_mc_clear_src(pmc);
1125                 in_dev_put(pmc->interface);
1126                 kfree(pmc);
1127         }
1128         /* clear dead sources, too */
1129         rcu_read_lock();
1130         for_each_pmc_rcu(in_dev, pmc) {
1131                 struct ip_sf_list *psf, *psf_next;
1132
1133                 spin_lock_bh(&pmc->lock);
1134                 psf = pmc->tomb;
1135                 pmc->tomb = NULL;
1136                 spin_unlock_bh(&pmc->lock);
1137                 for (; psf; psf=psf_next) {
1138                         psf_next = psf->sf_next;
1139                         kfree(psf);
1140                 }
1141         }
1142         rcu_read_unlock();
1143 }
1144 #endif
1145
1146 static void igmp_group_dropped(struct ip_mc_list *im)
1147 {
1148         struct in_device *in_dev = im->interface;
1149 #ifdef CONFIG_IP_MULTICAST
1150         int reporter;
1151 #endif
1152
1153         if (im->loaded) {
1154                 im->loaded = 0;
1155                 ip_mc_filter_del(in_dev, im->multiaddr);
1156         }
1157
1158 #ifdef CONFIG_IP_MULTICAST
1159         if (im->multiaddr == IGMP_ALL_HOSTS)
1160                 return;
1161
1162         reporter = im->reporter;
1163         igmp_stop_timer(im);
1164
1165         if (!in_dev->dead) {
1166                 if (IGMP_V1_SEEN(in_dev))
1167                         goto done;
1168                 if (IGMP_V2_SEEN(in_dev)) {
1169                         if (reporter)
1170                                 igmp_send_report(in_dev, im, IGMP_HOST_LEAVE_MESSAGE);
1171                         goto done;
1172                 }
1173                 /* IGMPv3 */
1174                 igmpv3_add_delrec(in_dev, im);
1175
1176                 igmp_ifc_event(in_dev);
1177         }
1178 done:
1179 #endif
1180         ip_mc_clear_src(im);
1181 }
1182
1183 static void igmp_group_added(struct ip_mc_list *im)
1184 {
1185         struct in_device *in_dev = im->interface;
1186
1187         if (im->loaded == 0) {
1188                 im->loaded = 1;
1189                 ip_mc_filter_add(in_dev, im->multiaddr);
1190         }
1191
1192 #ifdef CONFIG_IP_MULTICAST
1193         if (im->multiaddr == IGMP_ALL_HOSTS)
1194                 return;
1195
1196         if (in_dev->dead)
1197                 return;
1198         if (IGMP_V1_SEEN(in_dev) || IGMP_V2_SEEN(in_dev)) {
1199                 spin_lock_bh(&im->lock);
1200                 igmp_start_timer(im, IGMP_Initial_Report_Delay);
1201                 spin_unlock_bh(&im->lock);
1202                 return;
1203         }
1204         /* else, v3 */
1205
1206         im->crcount = in_dev->mr_qrv ? in_dev->mr_qrv :
1207                 IGMP_Unsolicited_Report_Count;
1208         igmp_ifc_event(in_dev);
1209 #endif
1210 }
1211
1212
1213 /*
1214  *      Multicast list managers
1215  */
1216
1217
1218 /*
1219  *      A socket has joined a multicast group on device dev.
1220  */
1221
1222 void ip_mc_inc_group(struct in_device *in_dev, __be32 addr)
1223 {
1224         struct ip_mc_list *im;
1225
1226         ASSERT_RTNL();
1227
1228         for_each_pmc_rtnl(in_dev, im) {
1229                 if (im->multiaddr == addr) {
1230                         im->users++;
1231                         ip_mc_add_src(in_dev, &addr, MCAST_EXCLUDE, 0, NULL, 0);
1232                         goto out;
1233                 }
1234         }
1235
1236         im = kzalloc(sizeof(*im), GFP_KERNEL);
1237         if (!im)
1238                 goto out;
1239
1240         im->users = 1;
1241         im->interface = in_dev;
1242         in_dev_hold(in_dev);
1243         im->multiaddr = addr;
1244         /* initial mode is (EX, empty) */
1245         im->sfmode = MCAST_EXCLUDE;
1246         im->sfcount[MCAST_EXCLUDE] = 1;
1247         atomic_set(&im->refcnt, 1);
1248         spin_lock_init(&im->lock);
1249 #ifdef CONFIG_IP_MULTICAST
1250         setup_timer(&im->timer, &igmp_timer_expire, (unsigned long)im);
1251         im->unsolicit_count = IGMP_Unsolicited_Report_Count;
1252 #endif
1253
1254         im->next_rcu = in_dev->mc_list;
1255         in_dev->mc_count++;
1256         rcu_assign_pointer(in_dev->mc_list, im);
1257
1258 #ifdef CONFIG_IP_MULTICAST
1259         igmpv3_del_delrec(in_dev, im->multiaddr);
1260 #endif
1261         igmp_group_added(im);
1262         if (!in_dev->dead)
1263                 ip_rt_multicast_event(in_dev);
1264 out:
1265         return;
1266 }
1267 EXPORT_SYMBOL(ip_mc_inc_group);
1268
1269 /*
1270  *      Resend IGMP JOIN report; used for bonding.
1271  */
1272 void ip_mc_rejoin_group(struct ip_mc_list *im)
1273 {
1274 #ifdef CONFIG_IP_MULTICAST
1275         struct in_device *in_dev = im->interface;
1276
1277         if (im->multiaddr == IGMP_ALL_HOSTS)
1278                 return;
1279
1280         /* a failover is happening and switches
1281          * must be notified immediately */
1282         if (IGMP_V1_SEEN(in_dev))
1283                 igmp_send_report(in_dev, im, IGMP_HOST_MEMBERSHIP_REPORT);
1284         else if (IGMP_V2_SEEN(in_dev))
1285                 igmp_send_report(in_dev, im, IGMPV2_HOST_MEMBERSHIP_REPORT);
1286         else
1287                 igmp_send_report(in_dev, im, IGMPV3_HOST_MEMBERSHIP_REPORT);
1288 #endif
1289 }
1290 EXPORT_SYMBOL(ip_mc_rejoin_group);
1291
1292 /*
1293  *      A socket has left a multicast group on device dev
1294  */
1295
1296 void ip_mc_dec_group(struct in_device *in_dev, __be32 addr)
1297 {
1298         struct ip_mc_list *i;
1299         struct ip_mc_list __rcu **ip;
1300
1301         ASSERT_RTNL();
1302
1303         for (ip = &in_dev->mc_list;
1304              (i = rtnl_dereference(*ip)) != NULL;
1305              ip = &i->next_rcu) {
1306                 if (i->multiaddr == addr) {
1307                         if (--i->users == 0) {
1308                                 *ip = i->next_rcu;
1309                                 in_dev->mc_count--;
1310                                 igmp_group_dropped(i);
1311
1312                                 if (!in_dev->dead)
1313                                         ip_rt_multicast_event(in_dev);
1314
1315                                 ip_ma_put(i);
1316                                 return;
1317                         }
1318                         break;
1319                 }
1320         }
1321 }
1322 EXPORT_SYMBOL(ip_mc_dec_group);
1323
1324 /* Device changing type */
1325
1326 void ip_mc_unmap(struct in_device *in_dev)
1327 {
1328         struct ip_mc_list *pmc;
1329
1330         ASSERT_RTNL();
1331
1332         for_each_pmc_rtnl(in_dev, pmc)
1333                 igmp_group_dropped(pmc);
1334 }
1335
1336 void ip_mc_remap(struct in_device *in_dev)
1337 {
1338         struct ip_mc_list *pmc;
1339
1340         ASSERT_RTNL();
1341
1342         for_each_pmc_rtnl(in_dev, pmc)
1343                 igmp_group_added(pmc);
1344 }
1345
1346 /* Device going down */
1347
1348 void ip_mc_down(struct in_device *in_dev)
1349 {
1350         struct ip_mc_list *pmc;
1351
1352         ASSERT_RTNL();
1353
1354         for_each_pmc_rtnl(in_dev, pmc)
1355                 igmp_group_dropped(pmc);
1356
1357 #ifdef CONFIG_IP_MULTICAST
1358         in_dev->mr_ifc_count = 0;
1359         if (del_timer(&in_dev->mr_ifc_timer))
1360                 __in_dev_put(in_dev);
1361         in_dev->mr_gq_running = 0;
1362         if (del_timer(&in_dev->mr_gq_timer))
1363                 __in_dev_put(in_dev);
1364         igmpv3_clear_delrec(in_dev);
1365 #endif
1366
1367         ip_mc_dec_group(in_dev, IGMP_ALL_HOSTS);
1368 }
1369
1370 void ip_mc_init_dev(struct in_device *in_dev)
1371 {
1372         ASSERT_RTNL();
1373
1374         in_dev->mc_tomb = NULL;
1375 #ifdef CONFIG_IP_MULTICAST
1376         in_dev->mr_gq_running = 0;
1377         setup_timer(&in_dev->mr_gq_timer, igmp_gq_timer_expire,
1378                         (unsigned long)in_dev);
1379         in_dev->mr_ifc_count = 0;
1380         in_dev->mc_count     = 0;
1381         setup_timer(&in_dev->mr_ifc_timer, igmp_ifc_timer_expire,
1382                         (unsigned long)in_dev);
1383         in_dev->mr_qrv = IGMP_Unsolicited_Report_Count;
1384 #endif
1385
1386         spin_lock_init(&in_dev->mc_tomb_lock);
1387 }
1388
1389 /* Device going up */
1390
1391 void ip_mc_up(struct in_device *in_dev)
1392 {
1393         struct ip_mc_list *pmc;
1394
1395         ASSERT_RTNL();
1396
1397         ip_mc_inc_group(in_dev, IGMP_ALL_HOSTS);
1398
1399         for_each_pmc_rtnl(in_dev, pmc)
1400                 igmp_group_added(pmc);
1401 }
1402
1403 /*
1404  *      Device is about to be destroyed: clean up.
1405  */
1406
1407 void ip_mc_destroy_dev(struct in_device *in_dev)
1408 {
1409         struct ip_mc_list *i;
1410
1411         ASSERT_RTNL();
1412
1413         /* Deactivate timers */
1414         ip_mc_down(in_dev);
1415
1416         while ((i = rtnl_dereference(in_dev->mc_list)) != NULL) {
1417                 in_dev->mc_list = i->next_rcu;
1418                 in_dev->mc_count--;
1419
1420                 igmp_group_dropped(i);
1421                 ip_ma_put(i);
1422         }
1423 }
1424
1425 /* RTNL is locked */
1426 static struct in_device *ip_mc_find_dev(struct net *net, struct ip_mreqn *imr)
1427 {
1428         struct flowi fl = { .nl_u = { .ip4_u =
1429                                       { .daddr = imr->imr_multiaddr.s_addr } } };
1430         struct rtable *rt;
1431         struct net_device *dev = NULL;
1432         struct in_device *idev = NULL;
1433
1434         if (imr->imr_ifindex) {
1435                 idev = inetdev_by_index(net, imr->imr_ifindex);
1436                 return idev;
1437         }
1438         if (imr->imr_address.s_addr) {
1439                 dev = __ip_dev_find(net, imr->imr_address.s_addr, false);
1440                 if (!dev)
1441                         return NULL;
1442         }
1443
1444         if (!dev && !ip_route_output_key(net, &rt, &fl)) {
1445                 dev = rt->dst.dev;
1446                 ip_rt_put(rt);
1447         }
1448         if (dev) {
1449                 imr->imr_ifindex = dev->ifindex;
1450                 idev = __in_dev_get_rtnl(dev);
1451         }
1452         return idev;
1453 }
1454
1455 /*
1456  *      Join a socket to a group
1457  */
1458 int sysctl_igmp_max_memberships __read_mostly = IP_MAX_MEMBERSHIPS;
1459 int sysctl_igmp_max_msf __read_mostly = IP_MAX_MSF;
1460
1461
1462 static int ip_mc_del1_src(struct ip_mc_list *pmc, int sfmode,
1463         __be32 *psfsrc)
1464 {
1465         struct ip_sf_list *psf, *psf_prev;
1466         int rv = 0;
1467
1468         psf_prev = NULL;
1469         for (psf=pmc->sources; psf; psf=psf->sf_next) {
1470                 if (psf->sf_inaddr == *psfsrc)
1471                         break;
1472                 psf_prev = psf;
1473         }
1474         if (!psf || psf->sf_count[sfmode] == 0) {
1475                 /* source filter not found, or count wrong =>  bug */
1476                 return -ESRCH;
1477         }
1478         psf->sf_count[sfmode]--;
1479         if (psf->sf_count[sfmode] == 0) {
1480                 ip_rt_multicast_event(pmc->interface);
1481         }
1482         if (!psf->sf_count[MCAST_INCLUDE] && !psf->sf_count[MCAST_EXCLUDE]) {
1483 #ifdef CONFIG_IP_MULTICAST
1484                 struct in_device *in_dev = pmc->interface;
1485 #endif
1486
1487                 /* no more filters for this source */
1488                 if (psf_prev)
1489                         psf_prev->sf_next = psf->sf_next;
1490                 else
1491                         pmc->sources = psf->sf_next;
1492 #ifdef CONFIG_IP_MULTICAST
1493                 if (psf->sf_oldin &&
1494                     !IGMP_V1_SEEN(in_dev) && !IGMP_V2_SEEN(in_dev)) {
1495                         psf->sf_crcount = in_dev->mr_qrv ? in_dev->mr_qrv :
1496                                 IGMP_Unsolicited_Report_Count;
1497                         psf->sf_next = pmc->tomb;
1498                         pmc->tomb = psf;
1499                         rv = 1;
1500                 } else
1501 #endif
1502                         kfree(psf);
1503         }
1504         return rv;
1505 }
1506
1507 #ifndef CONFIG_IP_MULTICAST
1508 #define igmp_ifc_event(x)       do { } while (0)
1509 #endif
1510
1511 static int ip_mc_del_src(struct in_device *in_dev, __be32 *pmca, int sfmode,
1512                          int sfcount, __be32 *psfsrc, int delta)
1513 {
1514         struct ip_mc_list *pmc;
1515         int     changerec = 0;
1516         int     i, err;
1517
1518         if (!in_dev)
1519                 return -ENODEV;
1520         rcu_read_lock();
1521         for_each_pmc_rcu(in_dev, pmc) {
1522                 if (*pmca == pmc->multiaddr)
1523                         break;
1524         }
1525         if (!pmc) {
1526                 /* MCA not found?? bug */
1527                 rcu_read_unlock();
1528                 return -ESRCH;
1529         }
1530         spin_lock_bh(&pmc->lock);
1531         rcu_read_unlock();
1532 #ifdef CONFIG_IP_MULTICAST
1533         sf_markstate(pmc);
1534 #endif
1535         if (!delta) {
1536                 err = -EINVAL;
1537                 if (!pmc->sfcount[sfmode])
1538                         goto out_unlock;
1539                 pmc->sfcount[sfmode]--;
1540         }
1541         err = 0;
1542         for (i=0; i<sfcount; i++) {
1543                 int rv = ip_mc_del1_src(pmc, sfmode, &psfsrc[i]);
1544
1545                 changerec |= rv > 0;
1546                 if (!err && rv < 0)
1547                         err = rv;
1548         }
1549         if (pmc->sfmode == MCAST_EXCLUDE &&
1550             pmc->sfcount[MCAST_EXCLUDE] == 0 &&
1551             pmc->sfcount[MCAST_INCLUDE]) {
1552 #ifdef CONFIG_IP_MULTICAST
1553                 struct ip_sf_list *psf;
1554 #endif
1555
1556                 /* filter mode change */
1557                 pmc->sfmode = MCAST_INCLUDE;
1558 #ifdef CONFIG_IP_MULTICAST
1559                 pmc->crcount = in_dev->mr_qrv ? in_dev->mr_qrv :
1560                         IGMP_Unsolicited_Report_Count;
1561                 in_dev->mr_ifc_count = pmc->crcount;
1562                 for (psf=pmc->sources; psf; psf = psf->sf_next)
1563                         psf->sf_crcount = 0;
1564                 igmp_ifc_event(pmc->interface);
1565         } else if (sf_setstate(pmc) || changerec) {
1566                 igmp_ifc_event(pmc->interface);
1567 #endif
1568         }
1569 out_unlock:
1570         spin_unlock_bh(&pmc->lock);
1571         return err;
1572 }
1573
1574 /*
1575  * Add multicast single-source filter to the interface list
1576  */
1577 static int ip_mc_add1_src(struct ip_mc_list *pmc, int sfmode,
1578         __be32 *psfsrc, int delta)
1579 {
1580         struct ip_sf_list *psf, *psf_prev;
1581
1582         psf_prev = NULL;
1583         for (psf=pmc->sources; psf; psf=psf->sf_next) {
1584                 if (psf->sf_inaddr == *psfsrc)
1585                         break;
1586                 psf_prev = psf;
1587         }
1588         if (!psf) {
1589                 psf = kzalloc(sizeof(*psf), GFP_ATOMIC);
1590                 if (!psf)
1591                         return -ENOBUFS;
1592                 psf->sf_inaddr = *psfsrc;
1593                 if (psf_prev) {
1594                         psf_prev->sf_next = psf;
1595                 } else
1596                         pmc->sources = psf;
1597         }
1598         psf->sf_count[sfmode]++;
1599         if (psf->sf_count[sfmode] == 1) {
1600                 ip_rt_multicast_event(pmc->interface);
1601         }
1602         return 0;
1603 }
1604
1605 #ifdef CONFIG_IP_MULTICAST
1606 static void sf_markstate(struct ip_mc_list *pmc)
1607 {
1608         struct ip_sf_list *psf;
1609         int mca_xcount = pmc->sfcount[MCAST_EXCLUDE];
1610
1611         for (psf=pmc->sources; psf; psf=psf->sf_next)
1612                 if (pmc->sfcount[MCAST_EXCLUDE]) {
1613                         psf->sf_oldin = mca_xcount ==
1614                                 psf->sf_count[MCAST_EXCLUDE] &&
1615                                 !psf->sf_count[MCAST_INCLUDE];
1616                 } else
1617                         psf->sf_oldin = psf->sf_count[MCAST_INCLUDE] != 0;
1618 }
1619
1620 static int sf_setstate(struct ip_mc_list *pmc)
1621 {
1622         struct ip_sf_list *psf, *dpsf;
1623         int mca_xcount = pmc->sfcount[MCAST_EXCLUDE];
1624         int qrv = pmc->interface->mr_qrv;
1625         int new_in, rv;
1626
1627         rv = 0;
1628         for (psf=pmc->sources; psf; psf=psf->sf_next) {
1629                 if (pmc->sfcount[MCAST_EXCLUDE]) {
1630                         new_in = mca_xcount == psf->sf_count[MCAST_EXCLUDE] &&
1631                                 !psf->sf_count[MCAST_INCLUDE];
1632                 } else
1633                         new_in = psf->sf_count[MCAST_INCLUDE] != 0;
1634                 if (new_in) {
1635                         if (!psf->sf_oldin) {
1636                                 struct ip_sf_list *prev = NULL;
1637
1638                                 for (dpsf=pmc->tomb; dpsf; dpsf=dpsf->sf_next) {
1639                                         if (dpsf->sf_inaddr == psf->sf_inaddr)
1640                                                 break;
1641                                         prev = dpsf;
1642                                 }
1643                                 if (dpsf) {
1644                                         if (prev)
1645                                                 prev->sf_next = dpsf->sf_next;
1646                                         else
1647                                                 pmc->tomb = dpsf->sf_next;
1648                                         kfree(dpsf);
1649                                 }
1650                                 psf->sf_crcount = qrv;
1651                                 rv++;
1652                         }
1653                 } else if (psf->sf_oldin) {
1654
1655                         psf->sf_crcount = 0;
1656                         /*
1657                          * add or update "delete" records if an active filter
1658                          * is now inactive
1659                          */
1660                         for (dpsf=pmc->tomb; dpsf; dpsf=dpsf->sf_next)
1661                                 if (dpsf->sf_inaddr == psf->sf_inaddr)
1662                                         break;
1663                         if (!dpsf) {
1664                                 dpsf = kmalloc(sizeof(*dpsf), GFP_ATOMIC);
1665                                 if (!dpsf)
1666                                         continue;
1667                                 *dpsf = *psf;
1668                                 /* pmc->lock held by callers */
1669                                 dpsf->sf_next = pmc->tomb;
1670                                 pmc->tomb = dpsf;
1671                         }
1672                         dpsf->sf_crcount = qrv;
1673                         rv++;
1674                 }
1675         }
1676         return rv;
1677 }
1678 #endif
1679
1680 /*
1681  * Add multicast source filter list to the interface list
1682  */
1683 static int ip_mc_add_src(struct in_device *in_dev, __be32 *pmca, int sfmode,
1684                          int sfcount, __be32 *psfsrc, int delta)
1685 {
1686         struct ip_mc_list *pmc;
1687         int     isexclude;
1688         int     i, err;
1689
1690         if (!in_dev)
1691                 return -ENODEV;
1692         rcu_read_lock();
1693         for_each_pmc_rcu(in_dev, pmc) {
1694                 if (*pmca == pmc->multiaddr)
1695                         break;
1696         }
1697         if (!pmc) {
1698                 /* MCA not found?? bug */
1699                 rcu_read_unlock();
1700                 return -ESRCH;
1701         }
1702         spin_lock_bh(&pmc->lock);
1703         rcu_read_unlock();
1704
1705 #ifdef CONFIG_IP_MULTICAST
1706         sf_markstate(pmc);
1707 #endif
1708         isexclude = pmc->sfmode == MCAST_EXCLUDE;
1709         if (!delta)
1710                 pmc->sfcount[sfmode]++;
1711         err = 0;
1712         for (i=0; i<sfcount; i++) {
1713                 err = ip_mc_add1_src(pmc, sfmode, &psfsrc[i], delta);
1714                 if (err)
1715                         break;
1716         }
1717         if (err) {
1718                 int j;
1719
1720                 pmc->sfcount[sfmode]--;
1721                 for (j=0; j<i; j++)
1722                         (void) ip_mc_del1_src(pmc, sfmode, &psfsrc[i]);
1723         } else if (isexclude != (pmc->sfcount[MCAST_EXCLUDE] != 0)) {
1724 #ifdef CONFIG_IP_MULTICAST
1725                 struct ip_sf_list *psf;
1726                 in_dev = pmc->interface;
1727 #endif
1728
1729                 /* filter mode change */
1730                 if (pmc->sfcount[MCAST_EXCLUDE])
1731                         pmc->sfmode = MCAST_EXCLUDE;
1732                 else if (pmc->sfcount[MCAST_INCLUDE])
1733                         pmc->sfmode = MCAST_INCLUDE;
1734 #ifdef CONFIG_IP_MULTICAST
1735                 /* else no filters; keep old mode for reports */
1736
1737                 pmc->crcount = in_dev->mr_qrv ? in_dev->mr_qrv :
1738                         IGMP_Unsolicited_Report_Count;
1739                 in_dev->mr_ifc_count = pmc->crcount;
1740                 for (psf=pmc->sources; psf; psf = psf->sf_next)
1741                         psf->sf_crcount = 0;
1742                 igmp_ifc_event(in_dev);
1743         } else if (sf_setstate(pmc)) {
1744                 igmp_ifc_event(in_dev);
1745 #endif
1746         }
1747         spin_unlock_bh(&pmc->lock);
1748         return err;
1749 }
1750
1751 static void ip_mc_clear_src(struct ip_mc_list *pmc)
1752 {
1753         struct ip_sf_list *psf, *nextpsf;
1754
1755         for (psf=pmc->tomb; psf; psf=nextpsf) {
1756                 nextpsf = psf->sf_next;
1757                 kfree(psf);
1758         }
1759         pmc->tomb = NULL;
1760         for (psf=pmc->sources; psf; psf=nextpsf) {
1761                 nextpsf = psf->sf_next;
1762                 kfree(psf);
1763         }
1764         pmc->sources = NULL;
1765         pmc->sfmode = MCAST_EXCLUDE;
1766         pmc->sfcount[MCAST_INCLUDE] = 0;
1767         pmc->sfcount[MCAST_EXCLUDE] = 1;
1768 }
1769
1770
1771 /*
1772  * Join a multicast group
1773  */
1774 int ip_mc_join_group(struct sock *sk , struct ip_mreqn *imr)
1775 {
1776         int err;
1777         __be32 addr = imr->imr_multiaddr.s_addr;
1778         struct ip_mc_socklist *iml = NULL, *i;
1779         struct in_device *in_dev;
1780         struct inet_sock *inet = inet_sk(sk);
1781         struct net *net = sock_net(sk);
1782         int ifindex;
1783         int count = 0;
1784
1785         if (!ipv4_is_multicast(addr))
1786                 return -EINVAL;
1787
1788         rtnl_lock();
1789
1790         in_dev = ip_mc_find_dev(net, imr);
1791
1792         if (!in_dev) {
1793                 iml = NULL;
1794                 err = -ENODEV;
1795                 goto done;
1796         }
1797
1798         err = -EADDRINUSE;
1799         ifindex = imr->imr_ifindex;
1800         for_each_pmc_rtnl(inet, i) {
1801                 if (i->multi.imr_multiaddr.s_addr == addr &&
1802                     i->multi.imr_ifindex == ifindex)
1803                         goto done;
1804                 count++;
1805         }
1806         err = -ENOBUFS;
1807         if (count >= sysctl_igmp_max_memberships)
1808                 goto done;
1809         iml = sock_kmalloc(sk, sizeof(*iml), GFP_KERNEL);
1810         if (iml == NULL)
1811                 goto done;
1812
1813         memcpy(&iml->multi, imr, sizeof(*imr));
1814         iml->next_rcu = inet->mc_list;
1815         iml->sflist = NULL;
1816         iml->sfmode = MCAST_EXCLUDE;
1817         rcu_assign_pointer(inet->mc_list, iml);
1818         ip_mc_inc_group(in_dev, addr);
1819         err = 0;
1820 done:
1821         rtnl_unlock();
1822         return err;
1823 }
1824 EXPORT_SYMBOL(ip_mc_join_group);
1825
1826 static void ip_sf_socklist_reclaim(struct rcu_head *rp)
1827 {
1828         kfree(container_of(rp, struct ip_sf_socklist, rcu));
1829         /* sk_omem_alloc should have been decreased by the caller*/
1830 }
1831
1832 static int ip_mc_leave_src(struct sock *sk, struct ip_mc_socklist *iml,
1833                            struct in_device *in_dev)
1834 {
1835         struct ip_sf_socklist *psf = rtnl_dereference(iml->sflist);
1836         int err;
1837
1838         if (psf == NULL) {
1839                 /* any-source empty exclude case */
1840                 return ip_mc_del_src(in_dev, &iml->multi.imr_multiaddr.s_addr,
1841                         iml->sfmode, 0, NULL, 0);
1842         }
1843         err = ip_mc_del_src(in_dev, &iml->multi.imr_multiaddr.s_addr,
1844                         iml->sfmode, psf->sl_count, psf->sl_addr, 0);
1845         rcu_assign_pointer(iml->sflist, NULL);
1846         /* decrease mem now to avoid the memleak warning */
1847         atomic_sub(IP_SFLSIZE(psf->sl_max), &sk->sk_omem_alloc);
1848         call_rcu(&psf->rcu, ip_sf_socklist_reclaim);
1849         return err;
1850 }
1851
1852
1853 static void ip_mc_socklist_reclaim(struct rcu_head *rp)
1854 {
1855         kfree(container_of(rp, struct ip_mc_socklist, rcu));
1856         /* sk_omem_alloc should have been decreased by the caller*/
1857 }
1858
1859
1860 /*
1861  *      Ask a socket to leave a group.
1862  */
1863
1864 int ip_mc_leave_group(struct sock *sk, struct ip_mreqn *imr)
1865 {
1866         struct inet_sock *inet = inet_sk(sk);
1867         struct ip_mc_socklist *iml;
1868         struct ip_mc_socklist __rcu **imlp;
1869         struct in_device *in_dev;
1870         struct net *net = sock_net(sk);
1871         __be32 group = imr->imr_multiaddr.s_addr;
1872         u32 ifindex;
1873         int ret = -EADDRNOTAVAIL;
1874
1875         rtnl_lock();
1876         in_dev = ip_mc_find_dev(net, imr);
1877         ifindex = imr->imr_ifindex;
1878         for (imlp = &inet->mc_list;
1879              (iml = rtnl_dereference(*imlp)) != NULL;
1880              imlp = &iml->next_rcu) {
1881                 if (iml->multi.imr_multiaddr.s_addr != group)
1882                         continue;
1883                 if (ifindex) {
1884                         if (iml->multi.imr_ifindex != ifindex)
1885                                 continue;
1886                 } else if (imr->imr_address.s_addr && imr->imr_address.s_addr !=
1887                                 iml->multi.imr_address.s_addr)
1888                         continue;
1889
1890                 (void) ip_mc_leave_src(sk, iml, in_dev);
1891
1892                 *imlp = iml->next_rcu;
1893
1894                 if (in_dev)
1895                         ip_mc_dec_group(in_dev, group);
1896                 rtnl_unlock();
1897                 /* decrease mem now to avoid the memleak warning */
1898                 atomic_sub(sizeof(*iml), &sk->sk_omem_alloc);
1899                 call_rcu(&iml->rcu, ip_mc_socklist_reclaim);
1900                 return 0;
1901         }
1902         if (!in_dev)
1903                 ret = -ENODEV;
1904         rtnl_unlock();
1905         return ret;
1906 }
1907
1908 int ip_mc_source(int add, int omode, struct sock *sk, struct
1909         ip_mreq_source *mreqs, int ifindex)
1910 {
1911         int err;
1912         struct ip_mreqn imr;
1913         __be32 addr = mreqs->imr_multiaddr;
1914         struct ip_mc_socklist *pmc;
1915         struct in_device *in_dev = NULL;
1916         struct inet_sock *inet = inet_sk(sk);
1917         struct ip_sf_socklist *psl;
1918         struct net *net = sock_net(sk);
1919         int leavegroup = 0;
1920         int i, j, rv;
1921
1922         if (!ipv4_is_multicast(addr))
1923                 return -EINVAL;
1924
1925         rtnl_lock();
1926
1927         imr.imr_multiaddr.s_addr = mreqs->imr_multiaddr;
1928         imr.imr_address.s_addr = mreqs->imr_interface;
1929         imr.imr_ifindex = ifindex;
1930         in_dev = ip_mc_find_dev(net, &imr);
1931
1932         if (!in_dev) {
1933                 err = -ENODEV;
1934                 goto done;
1935         }
1936         err = -EADDRNOTAVAIL;
1937
1938         for_each_pmc_rtnl(inet, pmc) {
1939                 if ((pmc->multi.imr_multiaddr.s_addr ==
1940                      imr.imr_multiaddr.s_addr) &&
1941                     (pmc->multi.imr_ifindex == imr.imr_ifindex))
1942                         break;
1943         }
1944         if (!pmc) {             /* must have a prior join */
1945                 err = -EINVAL;
1946                 goto done;
1947         }
1948         /* if a source filter was set, must be the same mode as before */
1949         if (pmc->sflist) {
1950                 if (pmc->sfmode != omode) {
1951                         err = -EINVAL;
1952                         goto done;
1953                 }
1954         } else if (pmc->sfmode != omode) {
1955                 /* allow mode switches for empty-set filters */
1956                 ip_mc_add_src(in_dev, &mreqs->imr_multiaddr, omode, 0, NULL, 0);
1957                 ip_mc_del_src(in_dev, &mreqs->imr_multiaddr, pmc->sfmode, 0,
1958                         NULL, 0);
1959                 pmc->sfmode = omode;
1960         }
1961
1962         psl = rtnl_dereference(pmc->sflist);
1963         if (!add) {
1964                 if (!psl)
1965                         goto done;      /* err = -EADDRNOTAVAIL */
1966                 rv = !0;
1967                 for (i=0; i<psl->sl_count; i++) {
1968                         rv = memcmp(&psl->sl_addr[i], &mreqs->imr_sourceaddr,
1969                                 sizeof(__be32));
1970                         if (rv == 0)
1971                                 break;
1972                 }
1973                 if (rv)         /* source not found */
1974                         goto done;      /* err = -EADDRNOTAVAIL */
1975
1976                 /* special case - (INCLUDE, empty) == LEAVE_GROUP */
1977                 if (psl->sl_count == 1 && omode == MCAST_INCLUDE) {
1978                         leavegroup = 1;
1979                         goto done;
1980                 }
1981
1982                 /* update the interface filter */
1983                 ip_mc_del_src(in_dev, &mreqs->imr_multiaddr, omode, 1,
1984                         &mreqs->imr_sourceaddr, 1);
1985
1986                 for (j=i+1; j<psl->sl_count; j++)
1987                         psl->sl_addr[j-1] = psl->sl_addr[j];
1988                 psl->sl_count--;
1989                 err = 0;
1990                 goto done;
1991         }
1992         /* else, add a new source to the filter */
1993
1994         if (psl && psl->sl_count >= sysctl_igmp_max_msf) {
1995                 err = -ENOBUFS;
1996                 goto done;
1997         }
1998         if (!psl || psl->sl_count == psl->sl_max) {
1999                 struct ip_sf_socklist *newpsl;
2000                 int count = IP_SFBLOCK;
2001
2002                 if (psl)
2003                         count += psl->sl_max;
2004                 newpsl = sock_kmalloc(sk, IP_SFLSIZE(count), GFP_KERNEL);
2005                 if (!newpsl) {
2006                         err = -ENOBUFS;
2007                         goto done;
2008                 }
2009                 newpsl->sl_max = count;
2010                 newpsl->sl_count = count - IP_SFBLOCK;
2011                 if (psl) {
2012                         for (i=0; i<psl->sl_count; i++)
2013                                 newpsl->sl_addr[i] = psl->sl_addr[i];
2014                         /* decrease mem now to avoid the memleak warning */
2015                         atomic_sub(IP_SFLSIZE(psl->sl_max), &sk->sk_omem_alloc);
2016                         call_rcu(&psl->rcu, ip_sf_socklist_reclaim);
2017                 }
2018                 rcu_assign_pointer(pmc->sflist, newpsl);
2019                 psl = newpsl;
2020         }
2021         rv = 1; /* > 0 for insert logic below if sl_count is 0 */
2022         for (i=0; i<psl->sl_count; i++) {
2023                 rv = memcmp(&psl->sl_addr[i], &mreqs->imr_sourceaddr,
2024                         sizeof(__be32));
2025                 if (rv == 0)
2026                         break;
2027         }
2028         if (rv == 0)            /* address already there is an error */
2029                 goto done;
2030         for (j=psl->sl_count-1; j>=i; j--)
2031                 psl->sl_addr[j+1] = psl->sl_addr[j];
2032         psl->sl_addr[i] = mreqs->imr_sourceaddr;
2033         psl->sl_count++;
2034         err = 0;
2035         /* update the interface list */
2036         ip_mc_add_src(in_dev, &mreqs->imr_multiaddr, omode, 1,
2037                 &mreqs->imr_sourceaddr, 1);
2038 done:
2039         rtnl_unlock();
2040         if (leavegroup)
2041                 return ip_mc_leave_group(sk, &imr);
2042         return err;
2043 }
2044
2045 int ip_mc_msfilter(struct sock *sk, struct ip_msfilter *msf, int ifindex)
2046 {
2047         int err = 0;
2048         struct ip_mreqn imr;
2049         __be32 addr = msf->imsf_multiaddr;
2050         struct ip_mc_socklist *pmc;
2051         struct in_device *in_dev;
2052         struct inet_sock *inet = inet_sk(sk);
2053         struct ip_sf_socklist *newpsl, *psl;
2054         struct net *net = sock_net(sk);
2055         int leavegroup = 0;
2056
2057         if (!ipv4_is_multicast(addr))
2058                 return -EINVAL;
2059         if (msf->imsf_fmode != MCAST_INCLUDE &&
2060             msf->imsf_fmode != MCAST_EXCLUDE)
2061                 return -EINVAL;
2062
2063         rtnl_lock();
2064
2065         imr.imr_multiaddr.s_addr = msf->imsf_multiaddr;
2066         imr.imr_address.s_addr = msf->imsf_interface;
2067         imr.imr_ifindex = ifindex;
2068         in_dev = ip_mc_find_dev(net, &imr);
2069
2070         if (!in_dev) {
2071                 err = -ENODEV;
2072                 goto done;
2073         }
2074
2075         /* special case - (INCLUDE, empty) == LEAVE_GROUP */
2076         if (msf->imsf_fmode == MCAST_INCLUDE && msf->imsf_numsrc == 0) {
2077                 leavegroup = 1;
2078                 goto done;
2079         }
2080
2081         for_each_pmc_rtnl(inet, pmc) {
2082                 if (pmc->multi.imr_multiaddr.s_addr == msf->imsf_multiaddr &&
2083                     pmc->multi.imr_ifindex == imr.imr_ifindex)
2084                         break;
2085         }
2086         if (!pmc) {             /* must have a prior join */
2087                 err = -EINVAL;
2088                 goto done;
2089         }
2090         if (msf->imsf_numsrc) {
2091                 newpsl = sock_kmalloc(sk, IP_SFLSIZE(msf->imsf_numsrc),
2092                                                            GFP_KERNEL);
2093                 if (!newpsl) {
2094                         err = -ENOBUFS;
2095                         goto done;
2096                 }
2097                 newpsl->sl_max = newpsl->sl_count = msf->imsf_numsrc;
2098                 memcpy(newpsl->sl_addr, msf->imsf_slist,
2099                         msf->imsf_numsrc * sizeof(msf->imsf_slist[0]));
2100                 err = ip_mc_add_src(in_dev, &msf->imsf_multiaddr,
2101                         msf->imsf_fmode, newpsl->sl_count, newpsl->sl_addr, 0);
2102                 if (err) {
2103                         sock_kfree_s(sk, newpsl, IP_SFLSIZE(newpsl->sl_max));
2104                         goto done;
2105                 }
2106         } else {
2107                 newpsl = NULL;
2108                 (void) ip_mc_add_src(in_dev, &msf->imsf_multiaddr,
2109                                      msf->imsf_fmode, 0, NULL, 0);
2110         }
2111         psl = rtnl_dereference(pmc->sflist);
2112         if (psl) {
2113                 (void) ip_mc_del_src(in_dev, &msf->imsf_multiaddr, pmc->sfmode,
2114                         psl->sl_count, psl->sl_addr, 0);
2115                 /* decrease mem now to avoid the memleak warning */
2116                 atomic_sub(IP_SFLSIZE(psl->sl_max), &sk->sk_omem_alloc);
2117                 call_rcu(&psl->rcu, ip_sf_socklist_reclaim);
2118         } else
2119                 (void) ip_mc_del_src(in_dev, &msf->imsf_multiaddr, pmc->sfmode,
2120                         0, NULL, 0);
2121         rcu_assign_pointer(pmc->sflist, newpsl);
2122         pmc->sfmode = msf->imsf_fmode;
2123         err = 0;
2124 done:
2125         rtnl_unlock();
2126         if (leavegroup)
2127                 err = ip_mc_leave_group(sk, &imr);
2128         return err;
2129 }
2130
2131 int ip_mc_msfget(struct sock *sk, struct ip_msfilter *msf,
2132         struct ip_msfilter __user *optval, int __user *optlen)
2133 {
2134         int err, len, count, copycount;
2135         struct ip_mreqn imr;
2136         __be32 addr = msf->imsf_multiaddr;
2137         struct ip_mc_socklist *pmc;
2138         struct in_device *in_dev;
2139         struct inet_sock *inet = inet_sk(sk);
2140         struct ip_sf_socklist *psl;
2141         struct net *net = sock_net(sk);
2142
2143         if (!ipv4_is_multicast(addr))
2144                 return -EINVAL;
2145
2146         rtnl_lock();
2147
2148         imr.imr_multiaddr.s_addr = msf->imsf_multiaddr;
2149         imr.imr_address.s_addr = msf->imsf_interface;
2150         imr.imr_ifindex = 0;
2151         in_dev = ip_mc_find_dev(net, &imr);
2152
2153         if (!in_dev) {
2154                 err = -ENODEV;
2155                 goto done;
2156         }
2157         err = -EADDRNOTAVAIL;
2158
2159         for_each_pmc_rtnl(inet, pmc) {
2160                 if (pmc->multi.imr_multiaddr.s_addr == msf->imsf_multiaddr &&
2161                     pmc->multi.imr_ifindex == imr.imr_ifindex)
2162                         break;
2163         }
2164         if (!pmc)               /* must have a prior join */
2165                 goto done;
2166         msf->imsf_fmode = pmc->sfmode;
2167         psl = rtnl_dereference(pmc->sflist);
2168         rtnl_unlock();
2169         if (!psl) {
2170                 len = 0;
2171                 count = 0;
2172         } else {
2173                 count = psl->sl_count;
2174         }
2175         copycount = count < msf->imsf_numsrc ? count : msf->imsf_numsrc;
2176         len = copycount * sizeof(psl->sl_addr[0]);
2177         msf->imsf_numsrc = count;
2178         if (put_user(IP_MSFILTER_SIZE(copycount), optlen) ||
2179             copy_to_user(optval, msf, IP_MSFILTER_SIZE(0))) {
2180                 return -EFAULT;
2181         }
2182         if (len &&
2183             copy_to_user(&optval->imsf_slist[0], psl->sl_addr, len))
2184                 return -EFAULT;
2185         return 0;
2186 done:
2187         rtnl_unlock();
2188         return err;
2189 }
2190
2191 int ip_mc_gsfget(struct sock *sk, struct group_filter *gsf,
2192         struct group_filter __user *optval, int __user *optlen)
2193 {
2194         int err, i, count, copycount;
2195         struct sockaddr_in *psin;
2196         __be32 addr;
2197         struct ip_mc_socklist *pmc;
2198         struct inet_sock *inet = inet_sk(sk);
2199         struct ip_sf_socklist *psl;
2200
2201         psin = (struct sockaddr_in *)&gsf->gf_group;
2202         if (psin->sin_family != AF_INET)
2203                 return -EINVAL;
2204         addr = psin->sin_addr.s_addr;
2205         if (!ipv4_is_multicast(addr))
2206                 return -EINVAL;
2207
2208         rtnl_lock();
2209
2210         err = -EADDRNOTAVAIL;
2211
2212         for_each_pmc_rtnl(inet, pmc) {
2213                 if (pmc->multi.imr_multiaddr.s_addr == addr &&
2214                     pmc->multi.imr_ifindex == gsf->gf_interface)
2215                         break;
2216         }
2217         if (!pmc)               /* must have a prior join */
2218                 goto done;
2219         gsf->gf_fmode = pmc->sfmode;
2220         psl = rtnl_dereference(pmc->sflist);
2221         rtnl_unlock();
2222         count = psl ? psl->sl_count : 0;
2223         copycount = count < gsf->gf_numsrc ? count : gsf->gf_numsrc;
2224         gsf->gf_numsrc = count;
2225         if (put_user(GROUP_FILTER_SIZE(copycount), optlen) ||
2226             copy_to_user(optval, gsf, GROUP_FILTER_SIZE(0))) {
2227                 return -EFAULT;
2228         }
2229         for (i=0; i<copycount; i++) {
2230                 struct sockaddr_storage ss;
2231
2232                 psin = (struct sockaddr_in *)&ss;
2233                 memset(&ss, 0, sizeof(ss));
2234                 psin->sin_family = AF_INET;
2235                 psin->sin_addr.s_addr = psl->sl_addr[i];
2236                 if (copy_to_user(&optval->gf_slist[i], &ss, sizeof(ss)))
2237                         return -EFAULT;
2238         }
2239         return 0;
2240 done:
2241         rtnl_unlock();
2242         return err;
2243 }
2244
2245 /*
2246  * check if a multicast source filter allows delivery for a given <src,dst,intf>
2247  */
2248 int ip_mc_sf_allow(struct sock *sk, __be32 loc_addr, __be32 rmt_addr, int dif)
2249 {
2250         struct inet_sock *inet = inet_sk(sk);
2251         struct ip_mc_socklist *pmc;
2252         struct ip_sf_socklist *psl;
2253         int i;
2254         int ret;
2255
2256         ret = 1;
2257         if (!ipv4_is_multicast(loc_addr))
2258                 goto out;
2259
2260         rcu_read_lock();
2261         for_each_pmc_rcu(inet, pmc) {
2262                 if (pmc->multi.imr_multiaddr.s_addr == loc_addr &&
2263                     pmc->multi.imr_ifindex == dif)
2264                         break;
2265         }
2266         ret = inet->mc_all;
2267         if (!pmc)
2268                 goto unlock;
2269         psl = rcu_dereference(pmc->sflist);
2270         ret = (pmc->sfmode == MCAST_EXCLUDE);
2271         if (!psl)
2272                 goto unlock;
2273
2274         for (i=0; i<psl->sl_count; i++) {
2275                 if (psl->sl_addr[i] == rmt_addr)
2276                         break;
2277         }
2278         ret = 0;
2279         if (pmc->sfmode == MCAST_INCLUDE && i >= psl->sl_count)
2280                 goto unlock;
2281         if (pmc->sfmode == MCAST_EXCLUDE && i < psl->sl_count)
2282                 goto unlock;
2283         ret = 1;
2284 unlock:
2285         rcu_read_unlock();
2286 out:
2287         return ret;
2288 }
2289
2290 /*
2291  *      A socket is closing.
2292  */
2293
2294 void ip_mc_drop_socket(struct sock *sk)
2295 {
2296         struct inet_sock *inet = inet_sk(sk);
2297         struct ip_mc_socklist *iml;
2298         struct net *net = sock_net(sk);
2299
2300         if (inet->mc_list == NULL)
2301                 return;
2302
2303         rtnl_lock();
2304         while ((iml = rtnl_dereference(inet->mc_list)) != NULL) {
2305                 struct in_device *in_dev;
2306
2307                 inet->mc_list = iml->next_rcu;
2308                 in_dev = inetdev_by_index(net, iml->multi.imr_ifindex);
2309                 (void) ip_mc_leave_src(sk, iml, in_dev);
2310                 if (in_dev != NULL)
2311                         ip_mc_dec_group(in_dev, iml->multi.imr_multiaddr.s_addr);
2312                 /* decrease mem now to avoid the memleak warning */
2313                 atomic_sub(sizeof(*iml), &sk->sk_omem_alloc);
2314                 call_rcu(&iml->rcu, ip_mc_socklist_reclaim);
2315         }
2316         rtnl_unlock();
2317 }
2318
2319 int ip_check_mc(struct in_device *in_dev, __be32 mc_addr, __be32 src_addr, u16 proto)
2320 {
2321         struct ip_mc_list *im;
2322         struct ip_sf_list *psf;
2323         int rv = 0;
2324
2325         rcu_read_lock();
2326         for_each_pmc_rcu(in_dev, im) {
2327                 if (im->multiaddr == mc_addr)
2328                         break;
2329         }
2330         if (im && proto == IPPROTO_IGMP) {
2331                 rv = 1;
2332         } else if (im) {
2333                 if (src_addr) {
2334                         for (psf=im->sources; psf; psf=psf->sf_next) {
2335                                 if (psf->sf_inaddr == src_addr)
2336                                         break;
2337                         }
2338                         if (psf)
2339                                 rv = psf->sf_count[MCAST_INCLUDE] ||
2340                                         psf->sf_count[MCAST_EXCLUDE] !=
2341                                         im->sfcount[MCAST_EXCLUDE];
2342                         else
2343                                 rv = im->sfcount[MCAST_EXCLUDE] != 0;
2344                 } else
2345                         rv = 1; /* unspecified source; tentatively allow */
2346         }
2347         rcu_read_unlock();
2348         return rv;
2349 }
2350
2351 #if defined(CONFIG_PROC_FS)
2352 struct igmp_mc_iter_state {
2353         struct seq_net_private p;
2354         struct net_device *dev;
2355         struct in_device *in_dev;
2356 };
2357
2358 #define igmp_mc_seq_private(seq)        ((struct igmp_mc_iter_state *)(seq)->private)
2359
2360 static inline struct ip_mc_list *igmp_mc_get_first(struct seq_file *seq)
2361 {
2362         struct net *net = seq_file_net(seq);
2363         struct ip_mc_list *im = NULL;
2364         struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2365
2366         state->in_dev = NULL;
2367         for_each_netdev_rcu(net, state->dev) {
2368                 struct in_device *in_dev;
2369
2370                 in_dev = __in_dev_get_rcu(state->dev);
2371                 if (!in_dev)
2372                         continue;
2373                 im = rcu_dereference(in_dev->mc_list);
2374                 if (im) {
2375                         state->in_dev = in_dev;
2376                         break;
2377                 }
2378         }
2379         return im;
2380 }
2381
2382 static struct ip_mc_list *igmp_mc_get_next(struct seq_file *seq, struct ip_mc_list *im)
2383 {
2384         struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2385
2386         im = rcu_dereference(im->next_rcu);
2387         while (!im) {
2388                 state->dev = next_net_device_rcu(state->dev);
2389                 if (!state->dev) {
2390                         state->in_dev = NULL;
2391                         break;
2392                 }
2393                 state->in_dev = __in_dev_get_rcu(state->dev);
2394                 if (!state->in_dev)
2395                         continue;
2396                 im = rcu_dereference(state->in_dev->mc_list);
2397         }
2398         return im;
2399 }
2400
2401 static struct ip_mc_list *igmp_mc_get_idx(struct seq_file *seq, loff_t pos)
2402 {
2403         struct ip_mc_list *im = igmp_mc_get_first(seq);
2404         if (im)
2405                 while (pos && (im = igmp_mc_get_next(seq, im)) != NULL)
2406                         --pos;
2407         return pos ? NULL : im;
2408 }
2409
2410 static void *igmp_mc_seq_start(struct seq_file *seq, loff_t *pos)
2411         __acquires(rcu)
2412 {
2413         rcu_read_lock();
2414         return *pos ? igmp_mc_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
2415 }
2416
2417 static void *igmp_mc_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2418 {
2419         struct ip_mc_list *im;
2420         if (v == SEQ_START_TOKEN)
2421                 im = igmp_mc_get_first(seq);
2422         else
2423                 im = igmp_mc_get_next(seq, v);
2424         ++*pos;
2425         return im;
2426 }
2427
2428 static void igmp_mc_seq_stop(struct seq_file *seq, void *v)
2429         __releases(rcu)
2430 {
2431         struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2432
2433         state->in_dev = NULL;
2434         state->dev = NULL;
2435         rcu_read_unlock();
2436 }
2437
2438 static int igmp_mc_seq_show(struct seq_file *seq, void *v)
2439 {
2440         if (v == SEQ_START_TOKEN)
2441                 seq_puts(seq,
2442                          "Idx\tDevice    : Count Querier\tGroup    Users Timer\tReporter\n");
2443         else {
2444                 struct ip_mc_list *im = (struct ip_mc_list *)v;
2445                 struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2446                 char   *querier;
2447 #ifdef CONFIG_IP_MULTICAST
2448                 querier = IGMP_V1_SEEN(state->in_dev) ? "V1" :
2449                           IGMP_V2_SEEN(state->in_dev) ? "V2" :
2450                           "V3";
2451 #else
2452                 querier = "NONE";
2453 #endif
2454
2455                 if (rcu_dereference(state->in_dev->mc_list) == im) {
2456                         seq_printf(seq, "%d\t%-10s: %5d %7s\n",
2457                                    state->dev->ifindex, state->dev->name, state->in_dev->mc_count, querier);
2458                 }
2459
2460                 seq_printf(seq,
2461                            "\t\t\t\t%08X %5d %d:%08lX\t\t%d\n",
2462                            im->multiaddr, im->users,
2463                            im->tm_running, im->tm_running ?
2464                            jiffies_to_clock_t(im->timer.expires-jiffies) : 0,
2465                            im->reporter);
2466         }
2467         return 0;
2468 }
2469
2470 static const struct seq_operations igmp_mc_seq_ops = {
2471         .start  =       igmp_mc_seq_start,
2472         .next   =       igmp_mc_seq_next,
2473         .stop   =       igmp_mc_seq_stop,
2474         .show   =       igmp_mc_seq_show,
2475 };
2476
2477 static int igmp_mc_seq_open(struct inode *inode, struct file *file)
2478 {
2479         return seq_open_net(inode, file, &igmp_mc_seq_ops,
2480                         sizeof(struct igmp_mc_iter_state));
2481 }
2482
2483 static const struct file_operations igmp_mc_seq_fops = {
2484         .owner          =       THIS_MODULE,
2485         .open           =       igmp_mc_seq_open,
2486         .read           =       seq_read,
2487         .llseek         =       seq_lseek,
2488         .release        =       seq_release_net,
2489 };
2490
2491 struct igmp_mcf_iter_state {
2492         struct seq_net_private p;
2493         struct net_device *dev;
2494         struct in_device *idev;
2495         struct ip_mc_list *im;
2496 };
2497
2498 #define igmp_mcf_seq_private(seq)       ((struct igmp_mcf_iter_state *)(seq)->private)
2499
2500 static inline struct ip_sf_list *igmp_mcf_get_first(struct seq_file *seq)
2501 {
2502         struct net *net = seq_file_net(seq);
2503         struct ip_sf_list *psf = NULL;
2504         struct ip_mc_list *im = NULL;
2505         struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2506
2507         state->idev = NULL;
2508         state->im = NULL;
2509         for_each_netdev_rcu(net, state->dev) {
2510                 struct in_device *idev;
2511                 idev = __in_dev_get_rcu(state->dev);
2512                 if (unlikely(idev == NULL))
2513                         continue;
2514                 im = rcu_dereference(idev->mc_list);
2515                 if (likely(im != NULL)) {
2516                         spin_lock_bh(&im->lock);
2517                         psf = im->sources;
2518                         if (likely(psf != NULL)) {
2519                                 state->im = im;
2520                                 state->idev = idev;
2521                                 break;
2522                         }
2523                         spin_unlock_bh(&im->lock);
2524                 }
2525         }
2526         return psf;
2527 }
2528
2529 static struct ip_sf_list *igmp_mcf_get_next(struct seq_file *seq, struct ip_sf_list *psf)
2530 {
2531         struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2532
2533         psf = psf->sf_next;
2534         while (!psf) {
2535                 spin_unlock_bh(&state->im->lock);
2536                 state->im = state->im->next;
2537                 while (!state->im) {
2538                         state->dev = next_net_device_rcu(state->dev);
2539                         if (!state->dev) {
2540                                 state->idev = NULL;
2541                                 goto out;
2542                         }
2543                         state->idev = __in_dev_get_rcu(state->dev);
2544                         if (!state->idev)
2545                                 continue;
2546                         state->im = rcu_dereference(state->idev->mc_list);
2547                 }
2548                 if (!state->im)
2549                         break;
2550                 spin_lock_bh(&state->im->lock);
2551                 psf = state->im->sources;
2552         }
2553 out:
2554         return psf;
2555 }
2556
2557 static struct ip_sf_list *igmp_mcf_get_idx(struct seq_file *seq, loff_t pos)
2558 {
2559         struct ip_sf_list *psf = igmp_mcf_get_first(seq);
2560         if (psf)
2561                 while (pos && (psf = igmp_mcf_get_next(seq, psf)) != NULL)
2562                         --pos;
2563         return pos ? NULL : psf;
2564 }
2565
2566 static void *igmp_mcf_seq_start(struct seq_file *seq, loff_t *pos)
2567         __acquires(rcu)
2568 {
2569         rcu_read_lock();
2570         return *pos ? igmp_mcf_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
2571 }
2572
2573 static void *igmp_mcf_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2574 {
2575         struct ip_sf_list *psf;
2576         if (v == SEQ_START_TOKEN)
2577                 psf = igmp_mcf_get_first(seq);
2578         else
2579                 psf = igmp_mcf_get_next(seq, v);
2580         ++*pos;
2581         return psf;
2582 }
2583
2584 static void igmp_mcf_seq_stop(struct seq_file *seq, void *v)
2585         __releases(rcu)
2586 {
2587         struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2588         if (likely(state->im != NULL)) {
2589                 spin_unlock_bh(&state->im->lock);
2590                 state->im = NULL;
2591         }
2592         state->idev = NULL;
2593         state->dev = NULL;
2594         rcu_read_unlock();
2595 }
2596
2597 static int igmp_mcf_seq_show(struct seq_file *seq, void *v)
2598 {
2599         struct ip_sf_list *psf = (struct ip_sf_list *)v;
2600         struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2601
2602         if (v == SEQ_START_TOKEN) {
2603                 seq_printf(seq,
2604                            "%3s %6s "
2605                            "%10s %10s %6s %6s\n", "Idx",
2606                            "Device", "MCA",
2607                            "SRC", "INC", "EXC");
2608         } else {
2609                 seq_printf(seq,
2610                            "%3d %6.6s 0x%08x "
2611                            "0x%08x %6lu %6lu\n",
2612                            state->dev->ifindex, state->dev->name,
2613                            ntohl(state->im->multiaddr),
2614                            ntohl(psf->sf_inaddr),
2615                            psf->sf_count[MCAST_INCLUDE],
2616                            psf->sf_count[MCAST_EXCLUDE]);
2617         }
2618         return 0;
2619 }
2620
2621 static const struct seq_operations igmp_mcf_seq_ops = {
2622         .start  =       igmp_mcf_seq_start,
2623         .next   =       igmp_mcf_seq_next,
2624         .stop   =       igmp_mcf_seq_stop,
2625         .show   =       igmp_mcf_seq_show,
2626 };
2627
2628 static int igmp_mcf_seq_open(struct inode *inode, struct file *file)
2629 {
2630         return seq_open_net(inode, file, &igmp_mcf_seq_ops,
2631                         sizeof(struct igmp_mcf_iter_state));
2632 }
2633
2634 static const struct file_operations igmp_mcf_seq_fops = {
2635         .owner          =       THIS_MODULE,
2636         .open           =       igmp_mcf_seq_open,
2637         .read           =       seq_read,
2638         .llseek         =       seq_lseek,
2639         .release        =       seq_release_net,
2640 };
2641
2642 static int __net_init igmp_net_init(struct net *net)
2643 {
2644         struct proc_dir_entry *pde;
2645
2646         pde = proc_net_fops_create(net, "igmp", S_IRUGO, &igmp_mc_seq_fops);
2647         if (!pde)
2648                 goto out_igmp;
2649         pde = proc_net_fops_create(net, "mcfilter", S_IRUGO, &igmp_mcf_seq_fops);
2650         if (!pde)
2651                 goto out_mcfilter;
2652         return 0;
2653
2654 out_mcfilter:
2655         proc_net_remove(net, "igmp");
2656 out_igmp:
2657         return -ENOMEM;
2658 }
2659
2660 static void __net_exit igmp_net_exit(struct net *net)
2661 {
2662         proc_net_remove(net, "mcfilter");
2663         proc_net_remove(net, "igmp");
2664 }
2665
2666 static struct pernet_operations igmp_net_ops = {
2667         .init = igmp_net_init,
2668         .exit = igmp_net_exit,
2669 };
2670
2671 int __init igmp_mc_proc_init(void)
2672 {
2673         return register_pernet_subsys(&igmp_net_ops);
2674 }
2675 #endif