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