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