net/ipv4/igmp.c

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