net/ipv4/igmp.c

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