net/ipv4/igmp.c

   1 /*
   2  *      Linux NET3:     Internet Group Management Protocol  [IGMP]
   3  *
   4  *      This code implements the IGMP protocol as defined in RFC1112. There has
   5  *      been a further revision of this protocol since which is now supported.
   6  *
   7  *      If you have trouble with this module be careful what gcc you have used,
   8  *      the older version didn't come out right using gcc 2.5.8, the newer one
   9  *      seems to fall out with gcc 2.6.2.
  10  *
  11  *      Authors:
  12  *              Alan Cox <alan@lxorguk.ukuu.org.uk>
  13  *
  14  *      This program is free software; you can redistribute it and/or
  15  *      modify it under the terms of the GNU General Public License
  16  *      as published by the Free Software Foundation; either version
  17  *      2 of the License, or (at your option) any later version.
  18  *
  19  *      Fixes:
  20  *
  21  *              Alan Cox        :       Added lots of __inline__ to optimise
  22  *                                      the memory usage of all the tiny little
  23  *                                      functions.
  24  *              Alan Cox        :       Dumped the header building experiment.
  25  *              Alan Cox        :       Minor tweaks ready for multicast routing
  26  *                                      and extended IGMP protocol.
  27  *              Alan Cox        :       Removed a load of inline directives. Gcc 2.5.8
  28  *                                      writes utterly bogus code otherwise (sigh)
  29  *                                      fixed IGMP loopback to behave in the manner
  30  *                                      desired by mrouted, fixed the fact it has been
  31  *                                      broken since 1.3.6 and cleaned up a few minor
  32  *                                      points.
  33  *
  34  *              Chih-Jen Chang  :       Tried to revise IGMP to Version 2
  35  *              Tsu-Sheng Tsao          E-mail: chihjenc@scf.usc.edu and tsusheng@scf.usc.edu
  36  *                                      The enhancements are mainly based on Steve Deering's
  37  *                                      ipmulti-3.5 source code.
  38  *              Chih-Jen Chang  :       Added the igmp_get_mrouter_info and
  39  *              Tsu-Sheng Tsao          igmp_set_mrouter_info to keep track of
  40  *                                      the mrouted version on that device.
  41  *              Chih-Jen Chang  :       Added the max_resp_time parameter to
  42  *              Tsu-Sheng Tsao          igmp_heard_query(). Using this parameter
  43  *                                      to identify the multicast router version
  44  *                                      and do what the IGMP version 2 specified.
  45  *              Chih-Jen Chang  :       Added a timer to revert to IGMP V2 router
  46  *              Tsu-Sheng Tsao          if the specified time expired.
  47  *              Alan Cox        :       Stop IGMP from 0.0.0.0 being accepted.
  48  *              Alan Cox        :       Use GFP_ATOMIC in the right places.
  49  *              Christian Daudt :       igmp timer wasn't set for local group
  50  *                                      memberships but was being deleted,
  51  *                                      which caused a "del_timer() called
  52  *                                      from %p with timer not initialized\n"
  53  *                                      message (960131).
  54  *              Christian Daudt :       removed del_timer from
  55  *                                      igmp_timer_expire function (960205).
  56  *             Christian Daudt :       igmp_heard_report now only calls
  57  *                                     igmp_timer_expire if tm->running is
  58  *                                     true (960216).
  59  *              Malcolm Beattie :       ttl comparison wrong in igmp_rcv made
  60  *                                      igmp_heard_query never trigger. Expiry
  61  *                                      miscalculation fixed in igmp_heard_query
  62  *                                      and random() made to return unsigned to
  63  *                                      prevent negative expiry times.
  64  *              Alexey Kuznetsov:       Wrong group leaving behaviour, backport
  65  *                                      fix from pending 2.1.x patches.
  66  *              Alan Cox:               Forget to enable FDDI support earlier.
  67  *              Alexey Kuznetsov:       Fixed leaving groups on device down.
  68  *              Alexey Kuznetsov:       Accordance to igmp-v2-06 draft.
  69  *              David L Stevens:        IGMPv3 support, with help from
  70  *                                      Vinay Kulkarni
  71  */
  72
  73 #include <linux/module.h>
  74 #include <linux/slab.h>
  75 #include <asm/uaccess.h>
  76 #include <asm/system.h>
  77 #include <linux/types.h>
  78 #include <linux/kernel.h>
  79 #include <linux/jiffies.h>
  80 #include <linux/string.h>
  81 #include <linux/socket.h>
  82 #include <linux/sockios.h>
  83 #include <linux/in.h>
  84 #include <linux/inet.h>
  85 #include <linux/netdevice.h>
  86 #include <linux/skbuff.h>
  87 #include <linux/inetdevice.h>
  88 #include <linux/igmp.h>
  89 #include <linux/if_arp.h>
  90 #include <linux/rtnetlink.h>
  91 #include <linux/times.h>
  92
  93 #include <net/net_namespace.h>
  94 #include <net/arp.h>
  95 #include <net/ip.h>
  96 #include <net/protocol.h>
  97 #include <net/route.h>
  98 #include <net/sock.h>
  99 #include <net/checksum.h>
 100 #include <linux/netfilter_ipv4.h>
 101 #ifdef CONFIG_IP_MROUTE
 102 #include <linux/mroute.h>
 103 #endif
 104 #ifdef CONFIG_PROC_FS
 105 #include <linux/proc_fs.h>
 106 #include <linux/seq_file.h>
 107 #endif
 108
 109 #define IP_MAX_MEMBERSHIPS      20
 110 #define IP_MAX_MSF              10
 111
 112 #ifdef CONFIG_IP_MULTICAST
 113 /* Parameter names and values are taken from igmp-v2-06 draft */
 114
 115 #define IGMP_V1_Router_Present_Timeout          (400*HZ)
 116 #define IGMP_V2_Router_Present_Timeout          (400*HZ)
 117 #define IGMP_Unsolicited_Report_Interval        (10*HZ)
 118 #define IGMP_Query_Response_Interval            (10*HZ)
 119 #define IGMP_Unsolicited_Report_Count           2
 120
 121
 122 #define IGMP_Initial_Report_Delay               (1)
 123
 124 /* IGMP_Initial_Report_Delay is not from IGMP specs!
 125  * IGMP specs require to report membership immediately after
 126  * joining a group, but we delay the first report by a
 127  * small interval. It seems more natural and still does not
 128  * contradict to specs provided this delay is small enough.
 129  */
 130
 131 #define IGMP_V1_SEEN(in_dev) \
 132         (IPV4_DEVCONF_ALL(dev_net(in_dev->dev), FORCE_IGMP_VERSION) == 1 || \
 133          IN_DEV_CONF_GET((in_dev), FORCE_IGMP_VERSION) == 1 || \
 134          ((in_dev)->mr_v1_seen && \
 135           time_before(jiffies, (in_dev)->mr_v1_seen)))
 136 #define IGMP_V2_SEEN(in_dev) \
 137         (IPV4_DEVCONF_ALL(dev_net(in_dev->dev), FORCE_IGMP_VERSION) == 2 || \
 138          IN_DEV_CONF_GET((in_dev), FORCE_IGMP_VERSION) == 2 || \
 139          ((in_dev)->mr_v2_seen && \
 140           time_before(jiffies, (in_dev)->mr_v2_seen)))
 141
 142 static void igmpv3_add_delrec(struct in_device *in_dev, struct ip_mc_list *im);
 143 static void igmpv3_del_delrec(struct in_device *in_dev, __be32 multiaddr);
 144 static void igmpv3_clear_delrec(struct in_device *in_dev);
 145 static int sf_setstate(struct ip_mc_list *pmc);
 146 static void sf_markstate(struct ip_mc_list *pmc);
 147 #endif
 148 static void ip_mc_clear_src(struct ip_mc_list *pmc);
 149 static int ip_mc_add_src(struct in_device *in_dev, __be32 *pmca, int sfmode,
 150                          int sfcount, __be32 *psfsrc, int delta);
 151
 152
 153 static void ip_mc_list_reclaim(struct rcu_head *head)
 154 {
 155         kfree(container_of(head, struct ip_mc_list, rcu));
 156 }
 157
 158 static void ip_ma_put(struct ip_mc_list *im)
 159 {
 160         if (atomic_dec_and_test(&im->refcnt)) {
 161                 in_dev_put(im->interface);
 162                 call_rcu(&im->rcu, ip_mc_list_reclaim);
 163         }
 164 }
 165
 166 #define for_each_pmc_rcu(in_dev, pmc)                           \
 167         for (pmc = rcu_dereference(in_dev->mc_list);            \
 168              pmc != NULL;                                       \
 169              pmc = rcu_dereference(pmc->next_rcu))
 170
 171 #define for_each_pmc_rtnl(in_dev, pmc)                          \
 172         for (pmc = rtnl_dereference(in_dev->mc_list);           \
 173              pmc != NULL;                                       \
 174              pmc = rtnl_dereference(pmc->next_rcu))
 175
 176 #ifdef CONFIG_IP_MULTICAST
 177
 178 /*
 179  *      Timer management
 180  */
 181
 182 static void igmp_stop_timer(struct ip_mc_list *im)
 183 {
 184         spin_lock_bh(&im->lock);
 185         if (del_timer(&im->timer))
 186                 atomic_dec(&im->refcnt);
 187         im->tm_running = 0;
 188         im->reporter = 0;
 189         im->unsolicit_count = 0;
 190         spin_unlock_bh(&im->lock);
 191 }
 192
 193 /* It must be called with locked im->lock */
 194 static void igmp_start_timer(struct ip_mc_list *im, int max_delay)
 195 {
 196         int tv = net_random() % max_delay;
 197
 198         im->tm_running = 1;
 199         if (!mod_timer(&im->timer, jiffies+tv+2))
 200                 atomic_inc(&im->refcnt);
 201 }
 202
 203 static void igmp_gq_start_timer(struct in_device *in_dev)
 204 {
 205         int tv = net_random() % in_dev->mr_maxdelay;
 206
 207         in_dev->mr_gq_running = 1;
 208         if (!mod_timer(&in_dev->mr_gq_timer, jiffies+tv+2))
 209                 in_dev_hold(in_dev);
 210 }
 211
 212 static void igmp_ifc_start_timer(struct in_device *in_dev, int delay)
 213 {
 214         int tv = net_random() % delay;
 215
 216         if (!mod_timer(&in_dev->mr_ifc_timer, jiffies+tv+2))
 217                 in_dev_hold(in_dev);
 218 }
 219
 220 static void igmp_mod_timer(struct ip_mc_list *im, int max_delay)
 221 {
 222         spin_lock_bh(&im->lock);
 223         im->unsolicit_count = 0;
 224         if (del_timer(&im->timer)) {
 225                 if ((long)(im->timer.expires-jiffies) < max_delay) {
 226                         add_timer(&im->timer);
 227                         im->tm_running = 1;
 228                         spin_unlock_bh(&im->lock);
 229                         return;
 230                 }
 231                 atomic_dec(&im->refcnt);
 232         }
 233         igmp_start_timer(im, max_delay);
 234         spin_unlock_bh(&im->lock);
 235 }
 236
 237
 238 /*
 239  *      Send an IGMP report.
 240  */
 241
 242 #define IGMP_SIZE (sizeof(struct igmphdr)+sizeof(struct iphdr)+4)
 243
 244
 245 static int is_in(struct ip_mc_list *pmc, struct ip_sf_list *psf, int type,
 246         int gdeleted, int sdeleted)
 247 {
 248         switch (type) {
 249         case IGMPV3_MODE_IS_INCLUDE:
 250         case IGMPV3_MODE_IS_EXCLUDE:
 251                 if (gdeleted || sdeleted)
 252                         return 0;
 253                 if (!(pmc->gsquery && !psf->sf_gsresp)) {
 254                         if (pmc->sfmode == MCAST_INCLUDE)
 255                                 return 1;
 256                         /* don't include if this source is excluded
 257                          * in all filters
 258                          */
 259                         if (psf->sf_count[MCAST_INCLUDE])
 260                                 return type == IGMPV3_MODE_IS_INCLUDE;
 261                         return pmc->sfcount[MCAST_EXCLUDE] ==
 262                                 psf->sf_count[MCAST_EXCLUDE];
 263                 }
 264                 return 0;
 265         case IGMPV3_CHANGE_TO_INCLUDE:
 266                 if (gdeleted || sdeleted)
 267                         return 0;
 268                 return psf->sf_count[MCAST_INCLUDE] != 0;
 269         case IGMPV3_CHANGE_TO_EXCLUDE:
 270                 if (gdeleted || sdeleted)
 271                         return 0;
 272                 if (pmc->sfcount[MCAST_EXCLUDE] == 0 ||
 273                     psf->sf_count[MCAST_INCLUDE])
 274                         return 0;
 275                 return pmc->sfcount[MCAST_EXCLUDE] ==
 276                         psf->sf_count[MCAST_EXCLUDE];
 277         case IGMPV3_ALLOW_NEW_SOURCES:
 278                 if (gdeleted || !psf->sf_crcount)
 279                         return 0;
 280                 return (pmc->sfmode == MCAST_INCLUDE) ^ sdeleted;
 281         case IGMPV3_BLOCK_OLD_SOURCES:
 282                 if (pmc->sfmode == MCAST_INCLUDE)
 283                         return gdeleted || (psf->sf_crcount && sdeleted);
 284                 return psf->sf_crcount && !gdeleted && !sdeleted;
 285         }
 286         return 0;
 287 }
 288
 289 static int
 290 igmp_scount(struct ip_mc_list *pmc, int type, int gdeleted, int sdeleted)
 291 {
 292         struct ip_sf_list *psf;
 293         int scount = 0;
 294
 295         for (psf=pmc->sources; psf; psf=psf->sf_next) {
 296                 if (!is_in(pmc, psf, type, gdeleted, sdeleted))
 297                         continue;
 298                 scount++;
 299         }
 300         return scount;
 301 }
 302
 303 #define igmp_skb_size(skb) (*(unsigned int *)((skb)->cb))
 304
 305 static struct sk_buff *igmpv3_newpack(struct net_device *dev, int size)
 306 {
 307         struct sk_buff *skb;
 308         struct rtable *rt;
 309         struct iphdr *pip;
 310         struct igmpv3_report *pig;
 311         struct net *net = dev_net(dev);
 312
 313         while (1) {
 314                 skb = alloc_skb(size + LL_ALLOCATED_SPACE(dev),
 315                                 GFP_ATOMIC | __GFP_NOWARN);
 316                 if (skb)
 317                         break;
 318                 size >>= 1;
 319                 if (size < 256)
 320                         return NULL;
 321         }
 322         igmp_skb_size(skb) = size;
 323
 324         rt = ip_route_output_ports(net, NULL, IGMPV3_ALL_MCR, 0,
 325                                    0, 0,
 326                                    IPPROTO_IGMP, 0, dev->ifindex);
 327         if (IS_ERR(rt)) {
 328                 kfree_skb(skb);
 329                 return NULL;
 330         }
 331
 332         skb_dst_set(skb, &rt->dst);
 333         skb->dev = dev;
 334
 335         skb_reserve(skb, LL_RESERVED_SPACE(dev));
 336
 337         skb_reset_network_header(skb);
 338         pip = ip_hdr(skb);
 339         skb_put(skb, sizeof(struct iphdr) + 4);
 340
 341         pip->version  = 4;
 342         pip->ihl      = (sizeof(struct iphdr)+4)>>2;
 343         pip->tos      = 0xc0;
 344         pip->frag_off = htons(IP_DF);
 345         pip->ttl      = 1;
 346         pip->daddr    = rt->rt_dst;
 347         pip->saddr    = rt->rt_src;
 348         pip->protocol = IPPROTO_IGMP;
 349         pip->tot_len  = 0;      /* filled in later */
 350         ip_select_ident(pip, &rt->dst, NULL);
 351         ((u8*)&pip[1])[0] = IPOPT_RA;
 352         ((u8*)&pip[1])[1] = 4;
 353         ((u8*)&pip[1])[2] = 0;
 354         ((u8*)&pip[1])[3] = 0;
 355
 356         skb->transport_header = skb->network_header + sizeof(struct iphdr) + 4;
 357         skb_put(skb, sizeof(*pig));
 358         pig = igmpv3_report_hdr(skb);
 359         pig->type = IGMPV3_HOST_MEMBERSHIP_REPORT;
 360         pig->resv1 = 0;
 361         pig->csum = 0;
 362         pig->resv2 = 0;
 363         pig->ngrec = 0;
 364         return skb;
 365 }
 366
 367 static int igmpv3_sendpack(struct sk_buff *skb)
 368 {
 369         struct igmphdr *pig = igmp_hdr(skb);
 370         const int igmplen = skb->tail - skb->transport_header;
 371
 372         pig->csum = ip_compute_csum(igmp_hdr(skb), igmplen);
 373
 374         return ip_local_out(skb);
 375 }
 376
 377 static int grec_size(struct ip_mc_list *pmc, int type, int gdel, int sdel)
 378 {
 379         return sizeof(struct igmpv3_grec) + 4*igmp_scount(pmc, type, gdel, sdel);
 380 }
 381
 382 static struct sk_buff *add_grhead(struct sk_buff *skb, struct ip_mc_list *pmc,
 383         int type, struct igmpv3_grec **ppgr)
 384 {
 385         struct net_device *dev = pmc->interface->dev;
 386         struct igmpv3_report *pih;
 387         struct igmpv3_grec *pgr;
 388
 389         if (!skb)
 390                 skb = igmpv3_newpack(dev, dev->mtu);
 391         if (!skb)
 392                 return NULL;
 393         pgr = (struct igmpv3_grec *)skb_put(skb, sizeof(struct igmpv3_grec));
 394         pgr->grec_type = type;
 395         pgr->grec_auxwords = 0;
 396         pgr->grec_nsrcs = 0;
 397         pgr->grec_mca = pmc->multiaddr;
 398         pih = igmpv3_report_hdr(skb);
 399         pih->ngrec = htons(ntohs(pih->ngrec)+1);
 400         *ppgr = pgr;
 401         return skb;
 402 }
 403
 404 #define AVAILABLE(skb) ((skb) ? ((skb)->dev ? igmp_skb_size(skb) - (skb)->len : \
 405         skb_tailroom(skb)) : 0)
 406
 407 static struct sk_buff *add_grec(struct sk_buff *skb, struct ip_mc_list *pmc,
 408         int type, int gdeleted, int sdeleted)
 409 {
 410         struct net_device *dev = pmc->interface->dev;
 411         struct igmpv3_report *pih;
 412         struct igmpv3_grec *pgr = NULL;
 413         struct ip_sf_list *psf, *psf_next, *psf_prev, **psf_list;
 414         int scount, stotal, first, isquery, truncate;
 415
 416         if (pmc->multiaddr == IGMP_ALL_HOSTS)
 417                 return skb;
 418
 419         isquery = type == IGMPV3_MODE_IS_INCLUDE ||
 420                   type == IGMPV3_MODE_IS_EXCLUDE;
 421         truncate = type == IGMPV3_MODE_IS_EXCLUDE ||
 422                     type == IGMPV3_CHANGE_TO_EXCLUDE;
 423
 424         stotal = scount = 0;
 425
 426         psf_list = sdeleted ? &pmc->tomb : &pmc->sources;
 427
 428         if (!*psf_list)
 429                 goto empty_source;
 430
 431         pih = skb ? igmpv3_report_hdr(skb) : NULL;
 432
 433         /* EX and TO_EX get a fresh packet, if needed */
 434         if (truncate) {
 435                 if (pih && pih->ngrec &&
 436                     AVAILABLE(skb) < grec_size(pmc, type, gdeleted, sdeleted)) {
 437                         if (skb)
 438                                 igmpv3_sendpack(skb);
 439                         skb = igmpv3_newpack(dev, dev->mtu);
 440                 }
 441         }
 442         first = 1;
 443         psf_prev = NULL;
 444         for (psf=*psf_list; psf; psf=psf_next) {
 445                 __be32 *psrc;
 446
 447                 psf_next = psf->sf_next;
 448
 449                 if (!is_in(pmc, psf, type, gdeleted, sdeleted)) {
 450                         psf_prev = psf;
 451                         continue;
 452                 }
 453
 454                 /* clear marks on query responses */
 455                 if (isquery)
 456                         psf->sf_gsresp = 0;
 457
 458                 if (AVAILABLE(skb) < sizeof(__be32) +
 459                     first*sizeof(struct igmpv3_grec)) {
 460                         if (truncate && !first)
 461                                 break;   /* truncate these */
 462                         if (pgr)
 463                                 pgr->grec_nsrcs = htons(scount);
 464                         if (skb)
 465                                 igmpv3_sendpack(skb);
 466                         skb = igmpv3_newpack(dev, dev->mtu);
 467                         first = 1;
 468                         scount = 0;
 469                 }
 470                 if (first) {
 471                         skb = add_grhead(skb, pmc, type, &pgr);
 472                         first = 0;
 473                 }
 474                 if (!skb)
 475                         return NULL;
 476                 psrc = (__be32 *)skb_put(skb, sizeof(__be32));
 477                 *psrc = psf->sf_inaddr;
 478                 scount++; stotal++;
 479                 if ((type == IGMPV3_ALLOW_NEW_SOURCES ||
 480                      type == IGMPV3_BLOCK_OLD_SOURCES) && psf->sf_crcount) {
 481                         psf->sf_crcount--;
 482                         if ((sdeleted || gdeleted) && psf->sf_crcount == 0) {
 483                                 if (psf_prev)
 484                                         psf_prev->sf_next = psf->sf_next;
 485                                 else
 486                                         *psf_list = psf->sf_next;
 487                                 kfree(psf);
 488                                 continue;
 489                         }
 490                 }
 491                 psf_prev = psf;
 492         }
 493
 494 empty_source:
 495         if (!stotal) {
 496                 if (type == IGMPV3_ALLOW_NEW_SOURCES ||
 497                     type == IGMPV3_BLOCK_OLD_SOURCES)
 498                         return skb;
 499                 if (pmc->crcount || isquery) {
 500                         /* make sure we have room for group header */
 501                         if (skb && AVAILABLE(skb)<sizeof(struct igmpv3_grec)) {
 502                                 igmpv3_sendpack(skb);
 503                                 skb = NULL; /* add_grhead will get a new one */
 504                         }
 505                         skb = add_grhead(skb, pmc, type, &pgr);
 506                 }
 507         }
 508         if (pgr)
 509                 pgr->grec_nsrcs = htons(scount);
 510
 511         if (isquery)
 512                 pmc->gsquery = 0;       /* clear query state on report */
 513         return skb;
 514 }
 515
 516 static int igmpv3_send_report(struct in_device *in_dev, struct ip_mc_list *pmc)
 517 {
 518         struct sk_buff *skb = NULL;
 519         int type;
 520
 521         if (!pmc) {
 522                 rcu_read_lock();
 523                 for_each_pmc_rcu(in_dev, pmc) {
 524                         if (pmc->multiaddr == IGMP_ALL_HOSTS)
 525                                 continue;
 526                         spin_lock_bh(&pmc->lock);
 527                         if (pmc->sfcount[MCAST_EXCLUDE])
 528                                 type = IGMPV3_MODE_IS_EXCLUDE;
 529                         else
 530                                 type = IGMPV3_MODE_IS_INCLUDE;
 531                         skb = add_grec(skb, pmc, type, 0, 0);
 532                         spin_unlock_bh(&pmc->lock);
 533                 }
 534                 rcu_read_unlock();
 535         } else {
 536                 spin_lock_bh(&pmc->lock);
 537                 if (pmc->sfcount[MCAST_EXCLUDE])
 538                         type = IGMPV3_MODE_IS_EXCLUDE;
 539                 else
 540                         type = IGMPV3_MODE_IS_INCLUDE;
 541                 skb = add_grec(skb, pmc, type, 0, 0);
 542                 spin_unlock_bh(&pmc->lock);
 543         }
 544         if (!skb)
 545                 return 0;
 546         return igmpv3_sendpack(skb);
 547 }
 548
 549 /*
 550  * remove zero-count source records from a source filter list
 551  */
 552 static void igmpv3_clear_zeros(struct ip_sf_list **ppsf)
 553 {
 554         struct ip_sf_list *psf_prev, *psf_next, *psf;
 555
 556         psf_prev = NULL;
 557         for (psf=*ppsf; psf; psf = psf_next) {
 558                 psf_next = psf->sf_next;
 559                 if (psf->sf_crcount == 0) {
 560                         if (psf_prev)
 561                                 psf_prev->sf_next = psf->sf_next;
 562                         else
 563                                 *ppsf = psf->sf_next;
 564                         kfree(psf);
 565                 } else
 566                         psf_prev = psf;
 567         }
 568 }
 569
 570 static void igmpv3_send_cr(struct in_device *in_dev)
 571 {
 572         struct ip_mc_list *pmc, *pmc_prev, *pmc_next;
 573         struct sk_buff *skb = NULL;
 574         int type, dtype;
 575
 576         rcu_read_lock();
 577         spin_lock_bh(&in_dev->mc_tomb_lock);
 578
 579         /* deleted MCA's */
 580         pmc_prev = NULL;
 581         for (pmc=in_dev->mc_tomb; pmc; pmc=pmc_next) {
 582                 pmc_next = pmc->next;
 583                 if (pmc->sfmode == MCAST_INCLUDE) {
 584                         type = IGMPV3_BLOCK_OLD_SOURCES;
 585                         dtype = IGMPV3_BLOCK_OLD_SOURCES;
 586                         skb = add_grec(skb, pmc, type, 1, 0);
 587                         skb = add_grec(skb, pmc, dtype, 1, 1);
 588                 }
 589                 if (pmc->crcount) {
 590                         if (pmc->sfmode == MCAST_EXCLUDE) {
 591                                 type = IGMPV3_CHANGE_TO_INCLUDE;
 592                                 skb = add_grec(skb, pmc, type, 1, 0);
 593                         }
 594                         pmc->crcount--;
 595                         if (pmc->crcount == 0) {
 596                                 igmpv3_clear_zeros(&pmc->tomb);
 597                                 igmpv3_clear_zeros(&pmc->sources);
 598                         }
 599                 }
 600                 if (pmc->crcount == 0 && !pmc->tomb && !pmc->sources) {
 601                         if (pmc_prev)
 602                                 pmc_prev->next = pmc_next;
 603                         else
 604                                 in_dev->mc_tomb = pmc_next;
 605                         in_dev_put(pmc->interface);
 606                         kfree(pmc);
 607                 } else
 608                         pmc_prev = pmc;
 609         }
 610         spin_unlock_bh(&in_dev->mc_tomb_lock);
 611
 612         /* change recs */
 613         for_each_pmc_rcu(in_dev, pmc) {
 614                 spin_lock_bh(&pmc->lock);
 615                 if (pmc->sfcount[MCAST_EXCLUDE]) {
 616                         type = IGMPV3_BLOCK_OLD_SOURCES;
 617                         dtype = IGMPV3_ALLOW_NEW_SOURCES;
 618                 } else {
 619                         type = IGMPV3_ALLOW_NEW_SOURCES;
 620                         dtype = IGMPV3_BLOCK_OLD_SOURCES;
 621                 }
 622                 skb = add_grec(skb, pmc, type, 0, 0);
 623                 skb = add_grec(skb, pmc, dtype, 0, 1);  /* deleted sources */
 624
 625                 /* filter mode changes */
 626                 if (pmc->crcount) {
 627                         if (pmc->sfmode == MCAST_EXCLUDE)
 628                                 type = IGMPV3_CHANGE_TO_EXCLUDE;
 629                         else
 630                                 type = IGMPV3_CHANGE_TO_INCLUDE;
 631                         skb = add_grec(skb, pmc, type, 0, 0);
 632                         pmc->crcount--;
 633                 }
 634                 spin_unlock_bh(&pmc->lock);
 635         }
 636         rcu_read_unlock();
 637
 638         if (!skb)
 639                 return;
 640         (void) igmpv3_sendpack(skb);
 641 }
 642
 643 static int igmp_send_report(struct in_device *in_dev, struct ip_mc_list *pmc,
 644         int type)
 645 {
 646         struct sk_buff *skb;
 647         struct iphdr *iph;
 648         struct igmphdr *ih;
 649         struct rtable *rt;
 650         struct net_device *dev = in_dev->dev;
 651         struct net *net = dev_net(dev);
 652         __be32  group = pmc ? pmc->multiaddr : 0;
 653         __be32  dst;
 654
 655         if (type == IGMPV3_HOST_MEMBERSHIP_REPORT)
 656                 return igmpv3_send_report(in_dev, pmc);
 657         else if (type == IGMP_HOST_LEAVE_MESSAGE)
 658                 dst = IGMP_ALL_ROUTER;
 659         else
 660                 dst = group;
 661
 662         rt = ip_route_output_ports(net, NULL, dst, 0,
 663                                    0, 0,
 664                                    IPPROTO_IGMP, 0, dev->ifindex);
 665         if (IS_ERR(rt))
 666                 return -1;
 667
 668         skb = alloc_skb(IGMP_SIZE+LL_ALLOCATED_SPACE(dev), 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, LL_RESERVED_SPACE(dev));
 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    = rt->rt_src;
 689         iph->protocol = IPPROTO_IGMP;
 690         ip_select_ident(iph, &rt->dst, 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 (pmc->sfcount[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         } else { /* v3 */
 883                 if (!pskb_may_pull(skb, sizeof(struct igmpv3_query)))
 884                         return;
 885
 886                 ih3 = igmpv3_query_hdr(skb);
 887                 if (ih3->nsrcs) {
 888                         if (!pskb_may_pull(skb, sizeof(struct igmpv3_query)
 889                                            + ntohs(ih3->nsrcs)*sizeof(__be32)))
 890                                 return;
 891                         ih3 = igmpv3_query_hdr(skb);
 892                 }
 893
 894                 max_delay = IGMPV3_MRC(ih3->code)*(HZ/IGMP_TIMER_SCALE);
 895                 if (!max_delay)
 896                         max_delay = 1;  /* can't mod w/ 0 */
 897                 in_dev->mr_maxdelay = max_delay;
 898                 if (ih3->qrv)
 899                         in_dev->mr_qrv = ih3->qrv;
 900                 if (!group) { /* general query */
 901                         if (ih3->nsrcs)
 902                                 return; /* no sources allowed */
 903                         igmp_gq_start_timer(in_dev);
 904                         return;
 905                 }
 906                 /* mark sources to include, if group & source-specific */
 907                 mark = ih3->nsrcs != 0;
 908         }
 909
 910         /*
 911          * - Start the timers in all of our membership records
 912          *   that the query applies to for the interface on
 913          *   which the query arrived excl. those that belong
 914          *   to a "local" group (224.0.0.X)
 915          * - For timers already running check if they need to
 916          *   be reset.
 917          * - Use the igmp->igmp_code field as the maximum
 918          *   delay possible
 919          */
 920         rcu_read_lock();
 921         for_each_pmc_rcu(in_dev, im) {
 922                 int changed;
 923
 924                 if (group && group != im->multiaddr)
 925                         continue;
 926                 if (im->multiaddr == IGMP_ALL_HOSTS)
 927                         continue;
 928                 spin_lock_bh(&im->lock);
 929                 if (im->tm_running)
 930                         im->gsquery = im->gsquery && mark;
 931                 else
 932                         im->gsquery = mark;
 933                 changed = !im->gsquery ||
 934                         igmp_marksources(im, ntohs(ih3->nsrcs), ih3->srcs);
 935                 spin_unlock_bh(&im->lock);
 936                 if (changed)
 937                         igmp_mod_timer(im, max_delay);
 938         }
 939         rcu_read_unlock();
 940 }
 941
 942 /* called in rcu_read_lock() section */
 943 int igmp_rcv(struct sk_buff *skb)
 944 {
 945         /* This basically follows the spec line by line -- see RFC1112 */
 946         struct igmphdr *ih;
 947         struct in_device *in_dev = __in_dev_get_rcu(skb->dev);
 948         int len = skb->len;
 949
 950         if (in_dev == NULL)
 951                 goto drop;
 952
 953         if (!pskb_may_pull(skb, sizeof(struct igmphdr)))
 954                 goto drop;
 955
 956         switch (skb->ip_summed) {
 957         case CHECKSUM_COMPLETE:
 958                 if (!csum_fold(skb->csum))
 959                         break;
 960                 /* fall through */
 961         case CHECKSUM_NONE:
 962                 skb->csum = 0;
 963                 if (__skb_checksum_complete(skb))
 964                         goto drop;
 965         }
 966
 967         ih = igmp_hdr(skb);
 968         switch (ih->type) {
 969         case IGMP_HOST_MEMBERSHIP_QUERY:
 970                 igmp_heard_query(in_dev, skb, len);
 971                 break;
 972         case IGMP_HOST_MEMBERSHIP_REPORT:
 973         case IGMPV2_HOST_MEMBERSHIP_REPORT:
 974                 /* Is it our report looped back? */
 975                 if (rt_is_output_route(skb_rtable(skb)))
 976                         break;
 977                 /* don't rely on MC router hearing unicast reports */
 978                 if (skb->pkt_type == PACKET_MULTICAST ||
 979                     skb->pkt_type == PACKET_BROADCAST)
 980                         igmp_heard_report(in_dev, ih->group);
 981                 break;
 982         case IGMP_PIM:
 983 #ifdef CONFIG_IP_PIMSM_V1
 984                 return pim_rcv_v1(skb);
 985 #endif
 986         case IGMPV3_HOST_MEMBERSHIP_REPORT:
 987         case IGMP_DVMRP:
 988         case IGMP_TRACE:
 989         case IGMP_HOST_LEAVE_MESSAGE:
 990         case IGMP_MTRACE:
 991         case IGMP_MTRACE_RESP:
 992                 break;
 993         default:
 994                 break;
 995         }
 996
 997 drop:
 998         kfree_skb(skb);
 999         return 0;
1000 }
1001
1002 #endif
1003
1004
1005 /*
1006  *      Add a filter to a device
1007  */
1008
1009 static void ip_mc_filter_add(struct in_device *in_dev, __be32 addr)
1010 {
1011         char buf[MAX_ADDR_LEN];
1012         struct net_device *dev = in_dev->dev;
1013
1014         /* Checking for IFF_MULTICAST here is WRONG-WRONG-WRONG.
1015            We will get multicast token leakage, when IFF_MULTICAST
1016            is changed. This check should be done in dev->set_multicast_list
1017            routine. Something sort of:
1018            if (dev->mc_list && dev->flags&IFF_MULTICAST) { do it; }
1019            --ANK
1020            */
1021         if (arp_mc_map(addr, buf, dev, 0) == 0)
1022                 dev_mc_add(dev, buf);
1023 }
1024
1025 /*
1026  *      Remove a filter from a device
1027  */
1028
1029 static void ip_mc_filter_del(struct in_device *in_dev, __be32 addr)
1030 {
1031         char buf[MAX_ADDR_LEN];
1032         struct net_device *dev = in_dev->dev;
1033
1034         if (arp_mc_map(addr, buf, dev, 0) == 0)
1035                 dev_mc_del(dev, buf);
1036 }
1037
1038 #ifdef CONFIG_IP_MULTICAST
1039 /*
1040  * deleted ip_mc_list manipulation
1041  */
1042 static void igmpv3_add_delrec(struct in_device *in_dev, struct ip_mc_list *im)
1043 {
1044         struct ip_mc_list *pmc;
1045
1046         /* this is an "ip_mc_list" for convenience; only the fields below
1047          * are actually used. In particular, the refcnt and users are not
1048          * used for management of the delete list. Using the same structure
1049          * for deleted items allows change reports to use common code with
1050          * non-deleted or query-response MCA's.
1051          */
1052         pmc = kzalloc(sizeof(*pmc), GFP_KERNEL);
1053         if (!pmc)
1054                 return;
1055         spin_lock_bh(&im->lock);
1056         pmc->interface = im->interface;
1057         in_dev_hold(in_dev);
1058         pmc->multiaddr = im->multiaddr;
1059         pmc->crcount = in_dev->mr_qrv ? in_dev->mr_qrv :
1060                 IGMP_Unsolicited_Report_Count;
1061         pmc->sfmode = im->sfmode;
1062         if (pmc->sfmode == MCAST_INCLUDE) {
1063                 struct ip_sf_list *psf;
1064
1065                 pmc->tomb = im->tomb;
1066                 pmc->sources = im->sources;
1067                 im->tomb = im->sources = NULL;
1068                 for (psf=pmc->sources; psf; psf=psf->sf_next)
1069                         psf->sf_crcount = pmc->crcount;
1070         }
1071         spin_unlock_bh(&im->lock);
1072
1073         spin_lock_bh(&in_dev->mc_tomb_lock);
1074         pmc->next = in_dev->mc_tomb;
1075         in_dev->mc_tomb = pmc;
1076         spin_unlock_bh(&in_dev->mc_tomb_lock);
1077 }
1078
1079 static void igmpv3_del_delrec(struct in_device *in_dev, __be32 multiaddr)
1080 {
1081         struct ip_mc_list *pmc, *pmc_prev;
1082         struct ip_sf_list *psf, *psf_next;
1083
1084         spin_lock_bh(&in_dev->mc_tomb_lock);
1085         pmc_prev = NULL;
1086         for (pmc=in_dev->mc_tomb; pmc; pmc=pmc->next) {
1087                 if (pmc->multiaddr == multiaddr)
1088                         break;
1089                 pmc_prev = pmc;
1090         }
1091         if (pmc) {
1092                 if (pmc_prev)
1093                         pmc_prev->next = pmc->next;
1094                 else
1095                         in_dev->mc_tomb = pmc->next;
1096         }
1097         spin_unlock_bh(&in_dev->mc_tomb_lock);
1098         if (pmc) {
1099                 for (psf=pmc->tomb; psf; psf=psf_next) {
1100                         psf_next = psf->sf_next;
1101                         kfree(psf);
1102                 }
1103                 in_dev_put(pmc->interface);
1104                 kfree(pmc);
1105         }
1106 }
1107
1108 static void igmpv3_clear_delrec(struct in_device *in_dev)
1109 {
1110         struct ip_mc_list *pmc, *nextpmc;
1111
1112         spin_lock_bh(&in_dev->mc_tomb_lock);
1113         pmc = in_dev->mc_tomb;
1114         in_dev->mc_tomb = NULL;
1115         spin_unlock_bh(&in_dev->mc_tomb_lock);
1116
1117         for (; pmc; pmc = nextpmc) {
1118                 nextpmc = pmc->next;
1119                 ip_mc_clear_src(pmc);
1120                 in_dev_put(pmc->interface);
1121                 kfree(pmc);
1122         }
1123         /* clear dead sources, too */
1124         rcu_read_lock();
1125         for_each_pmc_rcu(in_dev, pmc) {
1126                 struct ip_sf_list *psf, *psf_next;
1127
1128                 spin_lock_bh(&pmc->lock);
1129                 psf = pmc->tomb;
1130                 pmc->tomb = NULL;
1131                 spin_unlock_bh(&pmc->lock);
1132                 for (; psf; psf=psf_next) {
1133                         psf_next = psf->sf_next;
1134                         kfree(psf);
1135                 }
1136         }
1137         rcu_read_unlock();
1138 }
1139 #endif
1140
1141 static void igmp_group_dropped(struct ip_mc_list *im)
1142 {
1143         struct in_device *in_dev = im->interface;
1144 #ifdef CONFIG_IP_MULTICAST
1145         int reporter;
1146 #endif
1147
1148         if (im->loaded) {
1149                 im->loaded = 0;
1150                 ip_mc_filter_del(in_dev, im->multiaddr);
1151         }
1152
1153 #ifdef CONFIG_IP_MULTICAST
1154         if (im->multiaddr == IGMP_ALL_HOSTS)
1155                 return;
1156
1157         reporter = im->reporter;
1158         igmp_stop_timer(im);
1159
1160         if (!in_dev->dead) {
1161                 if (IGMP_V1_SEEN(in_dev))
1162                         goto done;
1163                 if (IGMP_V2_SEEN(in_dev)) {
1164                         if (reporter)
1165                                 igmp_send_report(in_dev, im, IGMP_HOST_LEAVE_MESSAGE);
1166                         goto done;
1167                 }
1168                 /* IGMPv3 */
1169                 igmpv3_add_delrec(in_dev, im);
1170
1171                 igmp_ifc_event(in_dev);
1172         }
1173 done:
1174 #endif
1175         ip_mc_clear_src(im);
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
1313                                 if (!in_dev->dead)
1314                                         ip_rt_multicast_event(in_dev);
1315
1316                                 ip_ma_put(i);
1317                                 return;
1318                         }
1319                         break;
1320                 }
1321         }
1322 }
1323 EXPORT_SYMBOL(ip_mc_dec_group);
1324
1325 /* Device changing type */
1326
1327 void ip_mc_unmap(struct in_device *in_dev)
1328 {
1329         struct ip_mc_list *pmc;
1330
1331         ASSERT_RTNL();
1332
1333         for_each_pmc_rtnl(in_dev, pmc)
1334                 igmp_group_dropped(pmc);
1335 }
1336
1337 void ip_mc_remap(struct in_device *in_dev)
1338 {
1339         struct ip_mc_list *pmc;
1340
1341         ASSERT_RTNL();
1342
1343         for_each_pmc_rtnl(in_dev, pmc)
1344                 igmp_group_added(pmc);
1345 }
1346
1347 /* Device going down */
1348
1349 void ip_mc_down(struct in_device *in_dev)
1350 {
1351         struct ip_mc_list *pmc;
1352
1353         ASSERT_RTNL();
1354
1355         for_each_pmc_rtnl(in_dev, pmc)
1356                 igmp_group_dropped(pmc);
1357
1358 #ifdef CONFIG_IP_MULTICAST
1359         in_dev->mr_ifc_count = 0;
1360         if (del_timer(&in_dev->mr_ifc_timer))
1361                 __in_dev_put(in_dev);
1362         in_dev->mr_gq_running = 0;
1363         if (del_timer(&in_dev->mr_gq_timer))
1364                 __in_dev_put(in_dev);
1365         igmpv3_clear_delrec(in_dev);
1366 #endif
1367
1368         ip_mc_dec_group(in_dev, IGMP_ALL_HOSTS);
1369 }
1370
1371 void ip_mc_init_dev(struct in_device *in_dev)
1372 {
1373         ASSERT_RTNL();
1374
1375         in_dev->mc_tomb = NULL;
1376 #ifdef CONFIG_IP_MULTICAST
1377         in_dev->mr_gq_running = 0;
1378         setup_timer(&in_dev->mr_gq_timer, igmp_gq_timer_expire,
1379                         (unsigned long)in_dev);
1380         in_dev->mr_ifc_count = 0;
1381         in_dev->mc_count     = 0;
1382         setup_timer(&in_dev->mr_ifc_timer, igmp_ifc_timer_expire,
1383                         (unsigned long)in_dev);
1384         in_dev->mr_qrv = IGMP_Unsolicited_Report_Count;
1385 #endif
1386
1387         spin_lock_init(&in_dev->mc_tomb_lock);
1388 }
1389
1390 /* Device going up */
1391
1392 void ip_mc_up(struct in_device *in_dev)
1393 {
1394         struct ip_mc_list *pmc;
1395
1396         ASSERT_RTNL();
1397
1398         ip_mc_inc_group(in_dev, IGMP_ALL_HOSTS);
1399
1400         for_each_pmc_rtnl(in_dev, pmc)
1401                 igmp_group_added(pmc);
1402 }
1403
1404 /*
1405  *      Device is about to be destroyed: clean up.
1406  */
1407
1408 void ip_mc_destroy_dev(struct in_device *in_dev)
1409 {
1410         struct ip_mc_list *i;
1411
1412         ASSERT_RTNL();
1413
1414         /* Deactivate timers */
1415         ip_mc_down(in_dev);
1416
1417         while ((i = rtnl_dereference(in_dev->mc_list)) != NULL) {
1418                 in_dev->mc_list = i->next_rcu;
1419                 in_dev->mc_count--;
1420
1421                 igmp_group_dropped(i);
1422                 ip_ma_put(i);
1423         }
1424 }
1425
1426 /* RTNL is locked */
1427 static struct in_device *ip_mc_find_dev(struct net *net, struct ip_mreqn *imr)
1428 {
1429         struct net_device *dev = NULL;
1430         struct in_device *idev = NULL;
1431
1432         if (imr->imr_ifindex) {
1433                 idev = inetdev_by_index(net, imr->imr_ifindex);
1434                 return idev;
1435         }
1436         if (imr->imr_address.s_addr) {
1437                 dev = __ip_dev_find(net, imr->imr_address.s_addr, false);
1438                 if (!dev)
1439                         return NULL;
1440         }
1441
1442         if (!dev) {
1443                 struct rtable *rt = ip_route_output(net,
1444                                                     imr->imr_multiaddr.s_addr,
1445                                                     0, 0, 0);
1446                 if (!IS_ERR(rt)) {
1447                         dev = rt->dst.dev;
1448                         ip_rt_put(rt);
1449                 }
1450         }
1451         if (dev) {
1452                 imr->imr_ifindex = dev->ifindex;
1453                 idev = __in_dev_get_rtnl(dev);
1454         }
1455         return idev;
1456 }
1457
1458 /*
1459  *      Join a socket to a group
1460  */
1461 int sysctl_igmp_max_memberships __read_mostly = IP_MAX_MEMBERSHIPS;
1462 int sysctl_igmp_max_msf __read_mostly = IP_MAX_MSF;
1463
1464
1465 static int ip_mc_del1_src(struct ip_mc_list *pmc, int sfmode,
1466         __be32 *psfsrc)
1467 {
1468         struct ip_sf_list *psf, *psf_prev;
1469         int rv = 0;
1470
1471         psf_prev = NULL;
1472         for (psf=pmc->sources; psf; psf=psf->sf_next) {
1473                 if (psf->sf_inaddr == *psfsrc)
1474                         break;
1475                 psf_prev = psf;
1476         }
1477         if (!psf || psf->sf_count[sfmode] == 0) {
1478                 /* source filter not found, or count wrong =>  bug */
1479                 return -ESRCH;
1480         }
1481         psf->sf_count[sfmode]--;
1482         if (psf->sf_count[sfmode] == 0) {
1483                 ip_rt_multicast_event(pmc->interface);
1484         }
1485         if (!psf->sf_count[MCAST_INCLUDE] && !psf->sf_count[MCAST_EXCLUDE]) {
1486 #ifdef CONFIG_IP_MULTICAST
1487                 struct in_device *in_dev = pmc->interface;
1488 #endif
1489
1490                 /* no more filters for this source */
1491                 if (psf_prev)
1492                         psf_prev->sf_next = psf->sf_next;
1493                 else
1494                         pmc->sources = psf->sf_next;
1495 #ifdef CONFIG_IP_MULTICAST
1496                 if (psf->sf_oldin &&
1497                     !IGMP_V1_SEEN(in_dev) && !IGMP_V2_SEEN(in_dev)) {
1498                         psf->sf_crcount = in_dev->mr_qrv ? in_dev->mr_qrv :
1499                                 IGMP_Unsolicited_Report_Count;
1500                         psf->sf_next = pmc->tomb;
1501                         pmc->tomb = psf;
1502                         rv = 1;
1503                 } else
1504 #endif
1505                         kfree(psf);
1506         }
1507         return rv;
1508 }
1509
1510 #ifndef CONFIG_IP_MULTICAST
1511 #define igmp_ifc_event(x)       do { } while (0)
1512 #endif
1513
1514 static int ip_mc_del_src(struct in_device *in_dev, __be32 *pmca, int sfmode,
1515                          int sfcount, __be32 *psfsrc, int delta)
1516 {
1517         struct ip_mc_list *pmc;
1518         int     changerec = 0;
1519         int     i, err;
1520
1521         if (!in_dev)
1522                 return -ENODEV;
1523         rcu_read_lock();
1524         for_each_pmc_rcu(in_dev, pmc) {
1525                 if (*pmca == pmc->multiaddr)
1526                         break;
1527         }
1528         if (!pmc) {
1529                 /* MCA not found?? bug */
1530                 rcu_read_unlock();
1531                 return -ESRCH;
1532         }
1533         spin_lock_bh(&pmc->lock);
1534         rcu_read_unlock();
1535 #ifdef CONFIG_IP_MULTICAST
1536         sf_markstate(pmc);
1537 #endif
1538         if (!delta) {
1539                 err = -EINVAL;
1540                 if (!pmc->sfcount[sfmode])
1541                         goto out_unlock;
1542                 pmc->sfcount[sfmode]--;
1543         }
1544         err = 0;
1545         for (i=0; i<sfcount; i++) {
1546                 int rv = ip_mc_del1_src(pmc, sfmode, &psfsrc[i]);
1547
1548                 changerec |= rv > 0;
1549                 if (!err && rv < 0)
1550                         err = rv;
1551         }
1552         if (pmc->sfmode == MCAST_EXCLUDE &&
1553             pmc->sfcount[MCAST_EXCLUDE] == 0 &&
1554             pmc->sfcount[MCAST_INCLUDE]) {
1555 #ifdef CONFIG_IP_MULTICAST
1556                 struct ip_sf_list *psf;
1557 #endif
1558
1559                 /* filter mode change */
1560                 pmc->sfmode = MCAST_INCLUDE;
1561 #ifdef CONFIG_IP_MULTICAST
1562                 pmc->crcount = in_dev->mr_qrv ? in_dev->mr_qrv :
1563                         IGMP_Unsolicited_Report_Count;
1564                 in_dev->mr_ifc_count = pmc->crcount;
1565                 for (psf=pmc->sources; psf; psf = psf->sf_next)
1566                         psf->sf_crcount = 0;
1567                 igmp_ifc_event(pmc->interface);
1568         } else if (sf_setstate(pmc) || changerec) {
1569                 igmp_ifc_event(pmc->interface);
1570 #endif
1571         }
1572 out_unlock:
1573         spin_unlock_bh(&pmc->lock);
1574         return err;
1575 }
1576
1577 /*
1578  * Add multicast single-source filter to the interface list
1579  */
1580 static int ip_mc_add1_src(struct ip_mc_list *pmc, int sfmode,
1581         __be32 *psfsrc, int delta)
1582 {
1583         struct ip_sf_list *psf, *psf_prev;
1584
1585         psf_prev = NULL;
1586         for (psf=pmc->sources; psf; psf=psf->sf_next) {
1587                 if (psf->sf_inaddr == *psfsrc)
1588                         break;
1589                 psf_prev = psf;
1590         }
1591         if (!psf) {
1592                 psf = kzalloc(sizeof(*psf), GFP_ATOMIC);
1593                 if (!psf)
1594                         return -ENOBUFS;
1595                 psf->sf_inaddr = *psfsrc;
1596                 if (psf_prev) {
1597                         psf_prev->sf_next = psf;
1598                 } else
1599                         pmc->sources = psf;
1600         }
1601         psf->sf_count[sfmode]++;
1602         if (psf->sf_count[sfmode] == 1) {
1603                 ip_rt_multicast_event(pmc->interface);
1604         }
1605         return 0;
1606 }
1607
1608 #ifdef CONFIG_IP_MULTICAST
1609 static void sf_markstate(struct ip_mc_list *pmc)
1610 {
1611         struct ip_sf_list *psf;
1612         int mca_xcount = pmc->sfcount[MCAST_EXCLUDE];
1613
1614         for (psf=pmc->sources; psf; psf=psf->sf_next)
1615                 if (pmc->sfcount[MCAST_EXCLUDE]) {
1616                         psf->sf_oldin = mca_xcount ==
1617                                 psf->sf_count[MCAST_EXCLUDE] &&
1618                                 !psf->sf_count[MCAST_INCLUDE];
1619                 } else
1620                         psf->sf_oldin = psf->sf_count[MCAST_INCLUDE] != 0;
1621 }
1622
1623 static int sf_setstate(struct ip_mc_list *pmc)
1624 {
1625         struct ip_sf_list *psf, *dpsf;
1626         int mca_xcount = pmc->sfcount[MCAST_EXCLUDE];
1627         int qrv = pmc->interface->mr_qrv;
1628         int new_in, rv;
1629
1630         rv = 0;
1631         for (psf=pmc->sources; psf; psf=psf->sf_next) {
1632                 if (pmc->sfcount[MCAST_EXCLUDE]) {
1633                         new_in = mca_xcount == psf->sf_count[MCAST_EXCLUDE] &&
1634                                 !psf->sf_count[MCAST_INCLUDE];
1635                 } else
1636                         new_in = psf->sf_count[MCAST_INCLUDE] != 0;
1637                 if (new_in) {
1638                         if (!psf->sf_oldin) {
1639                                 struct ip_sf_list *prev = NULL;
1640
1641                                 for (dpsf=pmc->tomb; dpsf; dpsf=dpsf->sf_next) {
1642                                         if (dpsf->sf_inaddr == psf->sf_inaddr)
1643                                                 break;
1644                                         prev = dpsf;
1645                                 }
1646                                 if (dpsf) {
1647                                         if (prev)
1648                                                 prev->sf_next = dpsf->sf_next;
1649                                         else
1650                                                 pmc->tomb = dpsf->sf_next;
1651                                         kfree(dpsf);
1652                                 }
1653                                 psf->sf_crcount = qrv;
1654                                 rv++;
1655                         }
1656                 } else if (psf->sf_oldin) {
1657
1658                         psf->sf_crcount = 0;
1659                         /*
1660                          * add or update "delete" records if an active filter
1661                          * is now inactive
1662                          */
1663                         for (dpsf=pmc->tomb; dpsf; dpsf=dpsf->sf_next)
1664                                 if (dpsf->sf_inaddr == psf->sf_inaddr)
1665                                         break;
1666                         if (!dpsf) {
1667                                 dpsf = kmalloc(sizeof(*dpsf), GFP_ATOMIC);
1668                                 if (!dpsf)
1669                                         continue;
1670                                 *dpsf = *psf;
1671                                 /* pmc->lock held by callers */
1672                                 dpsf->sf_next = pmc->tomb;
1673                                 pmc->tomb = dpsf;
1674                         }
1675                         dpsf->sf_crcount = qrv;
1676                         rv++;
1677                 }
1678         }
1679         return rv;
1680 }
1681 #endif
1682
1683 /*
1684  * Add multicast source filter list to the interface list
1685  */
1686 static int ip_mc_add_src(struct in_device *in_dev, __be32 *pmca, int sfmode,
1687                          int sfcount, __be32 *psfsrc, int delta)
1688 {
1689         struct ip_mc_list *pmc;
1690         int     isexclude;
1691         int     i, err;
1692
1693         if (!in_dev)
1694                 return -ENODEV;
1695         rcu_read_lock();
1696         for_each_pmc_rcu(in_dev, pmc) {
1697                 if (*pmca == pmc->multiaddr)
1698                         break;
1699         }
1700         if (!pmc) {
1701                 /* MCA not found?? bug */
1702                 rcu_read_unlock();
1703                 return -ESRCH;
1704         }
1705         spin_lock_bh(&pmc->lock);
1706         rcu_read_unlock();
1707
1708 #ifdef CONFIG_IP_MULTICAST
1709         sf_markstate(pmc);
1710 #endif
1711         isexclude = pmc->sfmode == MCAST_EXCLUDE;
1712         if (!delta)
1713                 pmc->sfcount[sfmode]++;
1714         err = 0;
1715         for (i=0; i<sfcount; i++) {
1716                 err = ip_mc_add1_src(pmc, sfmode, &psfsrc[i], delta);
1717                 if (err)
1718                         break;
1719         }
1720         if (err) {
1721                 int j;
1722
1723                 pmc->sfcount[sfmode]--;
1724                 for (j=0; j<i; j++)
1725                         (void) ip_mc_del1_src(pmc, sfmode, &psfsrc[i]);
1726         } else if (isexclude != (pmc->sfcount[MCAST_EXCLUDE] != 0)) {
1727 #ifdef CONFIG_IP_MULTICAST
1728                 struct ip_sf_list *psf;
1729                 in_dev = pmc->interface;
1730 #endif
1731
1732                 /* filter mode change */
1733                 if (pmc->sfcount[MCAST_EXCLUDE])
1734                         pmc->sfmode = MCAST_EXCLUDE;
1735                 else if (pmc->sfcount[MCAST_INCLUDE])
1736                         pmc->sfmode = MCAST_INCLUDE;
1737 #ifdef CONFIG_IP_MULTICAST
1738                 /* else no filters; keep old mode for reports */
1739
1740                 pmc->crcount = in_dev->mr_qrv ? in_dev->mr_qrv :
1741                         IGMP_Unsolicited_Report_Count;
1742                 in_dev->mr_ifc_count = pmc->crcount;
1743                 for (psf=pmc->sources; psf; psf = psf->sf_next)
1744                         psf->sf_crcount = 0;
1745                 igmp_ifc_event(in_dev);
1746         } else if (sf_setstate(pmc)) {
1747                 igmp_ifc_event(in_dev);
1748 #endif
1749         }
1750         spin_unlock_bh(&pmc->lock);
1751         return err;
1752 }
1753
1754 static void ip_mc_clear_src(struct ip_mc_list *pmc)
1755 {
1756         struct ip_sf_list *psf, *nextpsf;
1757
1758         for (psf=pmc->tomb; psf; psf=nextpsf) {
1759                 nextpsf = psf->sf_next;
1760                 kfree(psf);
1761         }
1762         pmc->tomb = NULL;
1763         for (psf=pmc->sources; psf; psf=nextpsf) {
1764                 nextpsf = psf->sf_next;
1765                 kfree(psf);
1766         }
1767         pmc->sources = NULL;
1768         pmc->sfmode = MCAST_EXCLUDE;
1769         pmc->sfcount[MCAST_INCLUDE] = 0;
1770         pmc->sfcount[MCAST_EXCLUDE] = 1;
1771 }
1772
1773
1774 /*
1775  * Join a multicast group
1776  */
1777 int ip_mc_join_group(struct sock *sk , struct ip_mreqn *imr)
1778 {
1779         int err;
1780         __be32 addr = imr->imr_multiaddr.s_addr;
1781         struct ip_mc_socklist *iml = NULL, *i;
1782         struct in_device *in_dev;
1783         struct inet_sock *inet = inet_sk(sk);
1784         struct net *net = sock_net(sk);
1785         int ifindex;
1786         int count = 0;
1787
1788         if (!ipv4_is_multicast(addr))
1789                 return -EINVAL;
1790
1791         rtnl_lock();
1792
1793         in_dev = ip_mc_find_dev(net, imr);
1794
1795         if (!in_dev) {
1796                 iml = NULL;
1797                 err = -ENODEV;
1798                 goto done;
1799         }
1800
1801         err = -EADDRINUSE;
1802         ifindex = imr->imr_ifindex;
1803         for_each_pmc_rtnl(inet, i) {
1804                 if (i->multi.imr_multiaddr.s_addr == addr &&
1805                     i->multi.imr_ifindex == ifindex)
1806                         goto done;
1807                 count++;
1808         }
1809         err = -ENOBUFS;
1810         if (count >= sysctl_igmp_max_memberships)
1811                 goto done;
1812         iml = sock_kmalloc(sk, sizeof(*iml), GFP_KERNEL);
1813         if (iml == NULL)
1814                 goto done;
1815
1816         memcpy(&iml->multi, imr, sizeof(*imr));
1817         iml->next_rcu = inet->mc_list;
1818         iml->sflist = NULL;
1819         iml->sfmode = MCAST_EXCLUDE;
1820         rcu_assign_pointer(inet->mc_list, iml);
1821         ip_mc_inc_group(in_dev, addr);
1822         err = 0;
1823 done:
1824         rtnl_unlock();
1825         return err;
1826 }
1827 EXPORT_SYMBOL(ip_mc_join_group);
1828
1829 static void ip_sf_socklist_reclaim(struct rcu_head *rp)
1830 {
1831         kfree(container_of(rp, struct ip_sf_socklist, rcu));
1832         /* sk_omem_alloc should have been decreased by the caller*/
1833 }
1834
1835 static int ip_mc_leave_src(struct sock *sk, struct ip_mc_socklist *iml,
1836                            struct in_device *in_dev)
1837 {
1838         struct ip_sf_socklist *psf = rtnl_dereference(iml->sflist);
1839         int err;
1840
1841         if (psf == NULL) {
1842                 /* any-source empty exclude case */
1843                 return ip_mc_del_src(in_dev, &iml->multi.imr_multiaddr.s_addr,
1844                         iml->sfmode, 0, NULL, 0);
1845         }
1846         err = ip_mc_del_src(in_dev, &iml->multi.imr_multiaddr.s_addr,
1847                         iml->sfmode, psf->sl_count, psf->sl_addr, 0);
1848         rcu_assign_pointer(iml->sflist, NULL);
1849         /* decrease mem now to avoid the memleak warning */
1850         atomic_sub(IP_SFLSIZE(psf->sl_max), &sk->sk_omem_alloc);
1851         call_rcu(&psf->rcu, ip_sf_socklist_reclaim);
1852         return err;
1853 }
1854
1855
1856 static void ip_mc_socklist_reclaim(struct rcu_head *rp)
1857 {
1858         kfree(container_of(rp, struct ip_mc_socklist, rcu));
1859         /* sk_omem_alloc should have been decreased by the caller*/
1860 }
1861
1862
1863 /*
1864  *      Ask a socket to leave a group.
1865  */
1866
1867 int ip_mc_leave_group(struct sock *sk, struct ip_mreqn *imr)
1868 {
1869         struct inet_sock *inet = inet_sk(sk);
1870         struct ip_mc_socklist *iml;
1871         struct ip_mc_socklist __rcu **imlp;
1872         struct in_device *in_dev;
1873         struct net *net = sock_net(sk);
1874         __be32 group = imr->imr_multiaddr.s_addr;
1875         u32 ifindex;
1876         int ret = -EADDRNOTAVAIL;
1877
1878         rtnl_lock();
1879         in_dev = ip_mc_find_dev(net, imr);
1880         ifindex = imr->imr_ifindex;
1881         for (imlp = &inet->mc_list;
1882              (iml = rtnl_dereference(*imlp)) != NULL;
1883              imlp = &iml->next_rcu) {
1884                 if (iml->multi.imr_multiaddr.s_addr != group)
1885                         continue;
1886                 if (ifindex) {
1887                         if (iml->multi.imr_ifindex != ifindex)
1888                                 continue;
1889                 } else if (imr->imr_address.s_addr && imr->imr_address.s_addr !=
1890                                 iml->multi.imr_address.s_addr)
1891                         continue;
1892
1893                 (void) ip_mc_leave_src(sk, iml, in_dev);
1894
1895                 *imlp = iml->next_rcu;
1896
1897                 if (in_dev)
1898                         ip_mc_dec_group(in_dev, group);
1899                 rtnl_unlock();
1900                 /* decrease mem now to avoid the memleak warning */
1901                 atomic_sub(sizeof(*iml), &sk->sk_omem_alloc);
1902                 call_rcu(&iml->rcu, ip_mc_socklist_reclaim);
1903                 return 0;
1904         }
1905         if (!in_dev)
1906                 ret = -ENODEV;
1907         rtnl_unlock();
1908         return ret;
1909 }
1910
1911 int ip_mc_source(int add, int omode, struct sock *sk, struct
1912         ip_mreq_source *mreqs, int ifindex)
1913 {
1914         int err;
1915         struct ip_mreqn imr;
1916         __be32 addr = mreqs->imr_multiaddr;
1917         struct ip_mc_socklist *pmc;
1918         struct in_device *in_dev = NULL;
1919         struct inet_sock *inet = inet_sk(sk);
1920         struct ip_sf_socklist *psl;
1921         struct net *net = sock_net(sk);
1922         int leavegroup = 0;
1923         int i, j, rv;
1924
1925         if (!ipv4_is_multicast(addr))
1926                 return -EINVAL;
1927
1928         rtnl_lock();
1929
1930         imr.imr_multiaddr.s_addr = mreqs->imr_multiaddr;
1931         imr.imr_address.s_addr = mreqs->imr_interface;
1932         imr.imr_ifindex = ifindex;
1933         in_dev = ip_mc_find_dev(net, &imr);
1934
1935         if (!in_dev) {
1936                 err = -ENODEV;
1937                 goto done;
1938         }
1939         err = -EADDRNOTAVAIL;
1940
1941         for_each_pmc_rtnl(inet, pmc) {
1942                 if ((pmc->multi.imr_multiaddr.s_addr ==
1943                      imr.imr_multiaddr.s_addr) &&
1944                     (pmc->multi.imr_ifindex == imr.imr_ifindex))
1945                         break;
1946         }
1947         if (!pmc) {             /* must have a prior join */
1948                 err = -EINVAL;
1949                 goto done;
1950         }
1951         /* if a source filter was set, must be the same mode as before */
1952         if (pmc->sflist) {
1953                 if (pmc->sfmode != omode) {
1954                         err = -EINVAL;
1955                         goto done;
1956                 }
1957         } else if (pmc->sfmode != omode) {
1958                 /* allow mode switches for empty-set filters */
1959                 ip_mc_add_src(in_dev, &mreqs->imr_multiaddr, omode, 0, NULL, 0);
1960                 ip_mc_del_src(in_dev, &mreqs->imr_multiaddr, pmc->sfmode, 0,
1961                         NULL, 0);
1962                 pmc->sfmode = omode;
1963         }
1964
1965         psl = rtnl_dereference(pmc->sflist);
1966         if (!add) {
1967                 if (!psl)
1968                         goto done;      /* err = -EADDRNOTAVAIL */
1969                 rv = !0;
1970                 for (i=0; i<psl->sl_count; i++) {
1971                         rv = memcmp(&psl->sl_addr[i], &mreqs->imr_sourceaddr,
1972                                 sizeof(__be32));
1973                         if (rv == 0)
1974                                 break;
1975                 }
1976                 if (rv)         /* source not found */
1977                         goto done;      /* err = -EADDRNOTAVAIL */
1978
1979                 /* special case - (INCLUDE, empty) == LEAVE_GROUP */
1980                 if (psl->sl_count == 1 && omode == MCAST_INCLUDE) {
1981                         leavegroup = 1;
1982                         goto done;
1983                 }
1984
1985                 /* update the interface filter */
1986                 ip_mc_del_src(in_dev, &mreqs->imr_multiaddr, omode, 1,
1987                         &mreqs->imr_sourceaddr, 1);
1988
1989                 for (j=i+1; j<psl->sl_count; j++)
1990                         psl->sl_addr[j-1] = psl->sl_addr[j];
1991                 psl->sl_count--;
1992                 err = 0;
1993                 goto done;
1994         }
1995         /* else, add a new source to the filter */
1996
1997         if (psl && psl->sl_count >= sysctl_igmp_max_msf) {
1998                 err = -ENOBUFS;
1999                 goto done;
2000         }
2001         if (!psl || psl->sl_count == psl->sl_max) {
2002                 struct ip_sf_socklist *newpsl;
2003                 int count = IP_SFBLOCK;
2004
2005                 if (psl)
2006                         count += psl->sl_max;
2007                 newpsl = sock_kmalloc(sk, IP_SFLSIZE(count), GFP_KERNEL);
2008                 if (!newpsl) {
2009                         err = -ENOBUFS;
2010                         goto done;
2011                 }
2012                 newpsl->sl_max = count;
2013                 newpsl->sl_count = count - IP_SFBLOCK;
2014                 if (psl) {
2015                         for (i=0; i<psl->sl_count; i++)
2016                                 newpsl->sl_addr[i] = psl->sl_addr[i];
2017                         /* decrease mem now to avoid the memleak warning */
2018                         atomic_sub(IP_SFLSIZE(psl->sl_max), &sk->sk_omem_alloc);
2019                         call_rcu(&psl->rcu, ip_sf_socklist_reclaim);
2020                 }
2021                 rcu_assign_pointer(pmc->sflist, newpsl);
2022                 psl = newpsl;
2023         }
2024         rv = 1; /* > 0 for insert logic below if sl_count is 0 */
2025         for (i=0; i<psl->sl_count; i++) {
2026                 rv = memcmp(&psl->sl_addr[i], &mreqs->imr_sourceaddr,
2027                         sizeof(__be32));
2028                 if (rv == 0)
2029                         break;
2030         }
2031         if (rv == 0)            /* address already there is an error */
2032                 goto done;
2033         for (j=psl->sl_count-1; j>=i; j--)
2034                 psl->sl_addr[j+1] = psl->sl_addr[j];
2035         psl->sl_addr[i] = mreqs->imr_sourceaddr;
2036         psl->sl_count++;
2037         err = 0;
2038         /* update the interface list */
2039         ip_mc_add_src(in_dev, &mreqs->imr_multiaddr, omode, 1,
2040                 &mreqs->imr_sourceaddr, 1);
2041 done:
2042         rtnl_unlock();
2043         if (leavegroup)
2044                 return ip_mc_leave_group(sk, &imr);
2045         return err;
2046 }
2047
2048 int ip_mc_msfilter(struct sock *sk, struct ip_msfilter *msf, int ifindex)
2049 {
2050         int err = 0;
2051         struct ip_mreqn imr;
2052         __be32 addr = msf->imsf_multiaddr;
2053         struct ip_mc_socklist *pmc;
2054         struct in_device *in_dev;
2055         struct inet_sock *inet = inet_sk(sk);
2056         struct ip_sf_socklist *newpsl, *psl;
2057         struct net *net = sock_net(sk);
2058         int leavegroup = 0;
2059
2060         if (!ipv4_is_multicast(addr))
2061                 return -EINVAL;
2062         if (msf->imsf_fmode != MCAST_INCLUDE &&
2063             msf->imsf_fmode != MCAST_EXCLUDE)
2064                 return -EINVAL;
2065
2066         rtnl_lock();
2067
2068         imr.imr_multiaddr.s_addr = msf->imsf_multiaddr;
2069         imr.imr_address.s_addr = msf->imsf_interface;
2070         imr.imr_ifindex = ifindex;
2071         in_dev = ip_mc_find_dev(net, &imr);
2072
2073         if (!in_dev) {
2074                 err = -ENODEV;
2075                 goto done;
2076         }
2077
2078         /* special case - (INCLUDE, empty) == LEAVE_GROUP */
2079         if (msf->imsf_fmode == MCAST_INCLUDE && msf->imsf_numsrc == 0) {
2080                 leavegroup = 1;
2081                 goto done;
2082         }
2083
2084         for_each_pmc_rtnl(inet, pmc) {
2085                 if (pmc->multi.imr_multiaddr.s_addr == msf->imsf_multiaddr &&
2086                     pmc->multi.imr_ifindex == imr.imr_ifindex)
2087                         break;
2088         }
2089         if (!pmc) {             /* must have a prior join */
2090                 err = -EINVAL;
2091                 goto done;
2092         }
2093         if (msf->imsf_numsrc) {
2094                 newpsl = sock_kmalloc(sk, IP_SFLSIZE(msf->imsf_numsrc),
2095                                                            GFP_KERNEL);
2096                 if (!newpsl) {
2097                         err = -ENOBUFS;
2098                         goto done;
2099                 }
2100                 newpsl->sl_max = newpsl->sl_count = msf->imsf_numsrc;
2101                 memcpy(newpsl->sl_addr, msf->imsf_slist,
2102                         msf->imsf_numsrc * sizeof(msf->imsf_slist[0]));
2103                 err = ip_mc_add_src(in_dev, &msf->imsf_multiaddr,
2104                         msf->imsf_fmode, newpsl->sl_count, newpsl->sl_addr, 0);
2105                 if (err) {
2106                         sock_kfree_s(sk, newpsl, IP_SFLSIZE(newpsl->sl_max));
2107                         goto done;
2108                 }
2109         } else {
2110                 newpsl = NULL;
2111                 (void) ip_mc_add_src(in_dev, &msf->imsf_multiaddr,
2112                                      msf->imsf_fmode, 0, NULL, 0);
2113         }
2114         psl = rtnl_dereference(pmc->sflist);
2115         if (psl) {
2116                 (void) ip_mc_del_src(in_dev, &msf->imsf_multiaddr, pmc->sfmode,
2117                         psl->sl_count, psl->sl_addr, 0);
2118                 /* decrease mem now to avoid the memleak warning */
2119                 atomic_sub(IP_SFLSIZE(psl->sl_max), &sk->sk_omem_alloc);
2120                 call_rcu(&psl->rcu, ip_sf_socklist_reclaim);
2121         } else
2122                 (void) ip_mc_del_src(in_dev, &msf->imsf_multiaddr, pmc->sfmode,
2123                         0, NULL, 0);
2124         rcu_assign_pointer(pmc->sflist, newpsl);
2125         pmc->sfmode = msf->imsf_fmode;
2126         err = 0;
2127 done:
2128         rtnl_unlock();
2129         if (leavegroup)
2130                 err = ip_mc_leave_group(sk, &imr);
2131         return err;
2132 }
2133
2134 int ip_mc_msfget(struct sock *sk, struct ip_msfilter *msf,
2135         struct ip_msfilter __user *optval, int __user *optlen)
2136 {
2137         int err, len, count, copycount;
2138         struct ip_mreqn imr;
2139         __be32 addr = msf->imsf_multiaddr;
2140         struct ip_mc_socklist *pmc;
2141         struct in_device *in_dev;
2142         struct inet_sock *inet = inet_sk(sk);
2143         struct ip_sf_socklist *psl;
2144         struct net *net = sock_net(sk);
2145
2146         if (!ipv4_is_multicast(addr))
2147                 return -EINVAL;
2148
2149         rtnl_lock();
2150
2151         imr.imr_multiaddr.s_addr = msf->imsf_multiaddr;
2152         imr.imr_address.s_addr = msf->imsf_interface;
2153         imr.imr_ifindex = 0;
2154         in_dev = ip_mc_find_dev(net, &imr);
2155
2156         if (!in_dev) {
2157                 err = -ENODEV;
2158                 goto done;
2159         }
2160         err = -EADDRNOTAVAIL;
2161
2162         for_each_pmc_rtnl(inet, pmc) {
2163                 if (pmc->multi.imr_multiaddr.s_addr == msf->imsf_multiaddr &&
2164                     pmc->multi.imr_ifindex == imr.imr_ifindex)
2165                         break;
2166         }
2167         if (!pmc)               /* must have a prior join */
2168                 goto done;
2169         msf->imsf_fmode = pmc->sfmode;
2170         psl = rtnl_dereference(pmc->sflist);
2171         rtnl_unlock();
2172         if (!psl) {
2173                 len = 0;
2174                 count = 0;
2175         } else {
2176                 count = psl->sl_count;
2177         }
2178         copycount = count < msf->imsf_numsrc ? count : msf->imsf_numsrc;
2179         len = copycount * sizeof(psl->sl_addr[0]);
2180         msf->imsf_numsrc = count;
2181         if (put_user(IP_MSFILTER_SIZE(copycount), optlen) ||
2182             copy_to_user(optval, msf, IP_MSFILTER_SIZE(0))) {
2183                 return -EFAULT;
2184         }
2185         if (len &&
2186             copy_to_user(&optval->imsf_slist[0], psl->sl_addr, len))
2187                 return -EFAULT;
2188         return 0;
2189 done:
2190         rtnl_unlock();
2191         return err;
2192 }
2193
2194 int ip_mc_gsfget(struct sock *sk, struct group_filter *gsf,
2195         struct group_filter __user *optval, int __user *optlen)
2196 {
2197         int err, i, count, copycount;
2198         struct sockaddr_in *psin;
2199         __be32 addr;
2200         struct ip_mc_socklist *pmc;
2201         struct inet_sock *inet = inet_sk(sk);
2202         struct ip_sf_socklist *psl;
2203
2204         psin = (struct sockaddr_in *)&gsf->gf_group;
2205         if (psin->sin_family != AF_INET)
2206                 return -EINVAL;
2207         addr = psin->sin_addr.s_addr;
2208         if (!ipv4_is_multicast(addr))
2209                 return -EINVAL;
2210
2211         rtnl_lock();
2212
2213         err = -EADDRNOTAVAIL;
2214
2215         for_each_pmc_rtnl(inet, pmc) {
2216                 if (pmc->multi.imr_multiaddr.s_addr == addr &&
2217                     pmc->multi.imr_ifindex == gsf->gf_interface)
2218                         break;
2219         }
2220         if (!pmc)               /* must have a prior join */
2221                 goto done;
2222         gsf->gf_fmode = pmc->sfmode;
2223         psl = rtnl_dereference(pmc->sflist);
2224         rtnl_unlock();
2225         count = psl ? psl->sl_count : 0;
2226         copycount = count < gsf->gf_numsrc ? count : gsf->gf_numsrc;
2227         gsf->gf_numsrc = count;
2228         if (put_user(GROUP_FILTER_SIZE(copycount), optlen) ||
2229             copy_to_user(optval, gsf, GROUP_FILTER_SIZE(0))) {
2230                 return -EFAULT;
2231         }
2232         for (i=0; i<copycount; i++) {
2233                 struct sockaddr_storage ss;
2234
2235                 psin = (struct sockaddr_in *)&ss;
2236                 memset(&ss, 0, sizeof(ss));
2237                 psin->sin_family = AF_INET;
2238                 psin->sin_addr.s_addr = psl->sl_addr[i];
2239                 if (copy_to_user(&optval->gf_slist[i], &ss, sizeof(ss)))
2240                         return -EFAULT;
2241         }
2242         return 0;
2243 done:
2244         rtnl_unlock();
2245         return err;
2246 }
2247
2248 /*
2249  * check if a multicast source filter allows delivery for a given <src,dst,intf>
2250  */
2251 int ip_mc_sf_allow(struct sock *sk, __be32 loc_addr, __be32 rmt_addr, int dif)
2252 {
2253         struct inet_sock *inet = inet_sk(sk);
2254         struct ip_mc_socklist *pmc;
2255         struct ip_sf_socklist *psl;
2256         int i;
2257         int ret;
2258
2259         ret = 1;
2260         if (!ipv4_is_multicast(loc_addr))
2261                 goto out;
2262
2263         rcu_read_lock();
2264         for_each_pmc_rcu(inet, pmc) {
2265                 if (pmc->multi.imr_multiaddr.s_addr == loc_addr &&
2266                     pmc->multi.imr_ifindex == dif)
2267                         break;
2268         }
2269         ret = inet->mc_all;
2270         if (!pmc)
2271                 goto unlock;
2272         psl = rcu_dereference(pmc->sflist);
2273         ret = (pmc->sfmode == MCAST_EXCLUDE);
2274         if (!psl)
2275                 goto unlock;
2276
2277         for (i=0; i<psl->sl_count; i++) {
2278                 if (psl->sl_addr[i] == rmt_addr)
2279                         break;
2280         }
2281         ret = 0;
2282         if (pmc->sfmode == MCAST_INCLUDE && i >= psl->sl_count)
2283                 goto unlock;
2284         if (pmc->sfmode == MCAST_EXCLUDE && i < psl->sl_count)
2285                 goto unlock;
2286         ret = 1;
2287 unlock:
2288         rcu_read_unlock();
2289 out:
2290         return ret;
2291 }
2292
2293 /*
2294  *      A socket is closing.
2295  */
2296
2297 void ip_mc_drop_socket(struct sock *sk)
2298 {
2299         struct inet_sock *inet = inet_sk(sk);
2300         struct ip_mc_socklist *iml;
2301         struct net *net = sock_net(sk);
2302
2303         if (inet->mc_list == NULL)
2304                 return;
2305
2306         rtnl_lock();
2307         while ((iml = rtnl_dereference(inet->mc_list)) != NULL) {
2308                 struct in_device *in_dev;
2309
2310                 inet->mc_list = iml->next_rcu;
2311                 in_dev = inetdev_by_index(net, iml->multi.imr_ifindex);
2312                 (void) ip_mc_leave_src(sk, iml, in_dev);
2313                 if (in_dev != NULL)
2314                         ip_mc_dec_group(in_dev, iml->multi.imr_multiaddr.s_addr);
2315                 /* decrease mem now to avoid the memleak warning */
2316                 atomic_sub(sizeof(*iml), &sk->sk_omem_alloc);
2317                 call_rcu(&iml->rcu, ip_mc_socklist_reclaim);
2318         }
2319         rtnl_unlock();
2320 }
2321
2322 /* called with rcu_read_lock() */
2323 int ip_check_mc_rcu(struct in_device *in_dev, __be32 mc_addr, __be32 src_addr, u16 proto)
2324 {
2325         struct ip_mc_list *im;
2326         struct ip_sf_list *psf;
2327         int rv = 0;
2328
2329         for_each_pmc_rcu(in_dev, im) {
2330                 if (im->multiaddr == mc_addr)
2331                         break;
2332         }
2333         if (im && proto == IPPROTO_IGMP) {
2334                 rv = 1;
2335         } else if (im) {
2336                 if (src_addr) {
2337                         for (psf=im->sources; psf; psf=psf->sf_next) {
2338                                 if (psf->sf_inaddr == src_addr)
2339                                         break;
2340                         }
2341                         if (psf)
2342                                 rv = psf->sf_count[MCAST_INCLUDE] ||
2343                                         psf->sf_count[MCAST_EXCLUDE] !=
2344                                         im->sfcount[MCAST_EXCLUDE];
2345                         else
2346                                 rv = im->sfcount[MCAST_EXCLUDE] != 0;
2347                 } else
2348                         rv = 1; /* unspecified source; tentatively allow */
2349         }
2350         return rv;
2351 }
2352
2353 #if defined(CONFIG_PROC_FS)
2354 struct igmp_mc_iter_state {
2355         struct seq_net_private p;
2356         struct net_device *dev;
2357         struct in_device *in_dev;
2358 };
2359
2360 #define igmp_mc_seq_private(seq)        ((struct igmp_mc_iter_state *)(seq)->private)
2361
2362 static inline struct ip_mc_list *igmp_mc_get_first(struct seq_file *seq)
2363 {
2364         struct net *net = seq_file_net(seq);
2365         struct ip_mc_list *im = NULL;
2366         struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2367
2368         state->in_dev = NULL;
2369         for_each_netdev_rcu(net, state->dev) {
2370                 struct in_device *in_dev;
2371
2372                 in_dev = __in_dev_get_rcu(state->dev);
2373                 if (!in_dev)
2374                         continue;
2375                 im = rcu_dereference(in_dev->mc_list);
2376                 if (im) {
2377                         state->in_dev = in_dev;
2378                         break;
2379                 }
2380         }
2381         return im;
2382 }
2383
2384 static struct ip_mc_list *igmp_mc_get_next(struct seq_file *seq, struct ip_mc_list *im)
2385 {
2386         struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2387
2388         im = rcu_dereference(im->next_rcu);
2389         while (!im) {
2390                 state->dev = next_net_device_rcu(state->dev);
2391                 if (!state->dev) {
2392                         state->in_dev = NULL;
2393                         break;
2394                 }
2395                 state->in_dev = __in_dev_get_rcu(state->dev);
2396                 if (!state->in_dev)
2397                         continue;
2398                 im = rcu_dereference(state->in_dev->mc_list);
2399         }
2400         return im;
2401 }
2402
2403 static struct ip_mc_list *igmp_mc_get_idx(struct seq_file *seq, loff_t pos)
2404 {
2405         struct ip_mc_list *im = igmp_mc_get_first(seq);
2406         if (im)
2407                 while (pos && (im = igmp_mc_get_next(seq, im)) != NULL)
2408                         --pos;
2409         return pos ? NULL : im;
2410 }
2411
2412 static void *igmp_mc_seq_start(struct seq_file *seq, loff_t *pos)
2413         __acquires(rcu)
2414 {
2415         rcu_read_lock();
2416         return *pos ? igmp_mc_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
2417 }
2418
2419 static void *igmp_mc_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2420 {
2421         struct ip_mc_list *im;
2422         if (v == SEQ_START_TOKEN)
2423                 im = igmp_mc_get_first(seq);
2424         else
2425                 im = igmp_mc_get_next(seq, v);
2426         ++*pos;
2427         return im;
2428 }
2429
2430 static void igmp_mc_seq_stop(struct seq_file *seq, void *v)
2431         __releases(rcu)
2432 {
2433         struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2434
2435         state->in_dev = NULL;
2436         state->dev = NULL;
2437         rcu_read_unlock();
2438 }
2439
2440 static int igmp_mc_seq_show(struct seq_file *seq, void *v)
2441 {
2442         if (v == SEQ_START_TOKEN)
2443                 seq_puts(seq,
2444                          "Idx\tDevice    : Count Querier\tGroup    Users Timer\tReporter\n");
2445         else {
2446                 struct ip_mc_list *im = (struct ip_mc_list *)v;
2447                 struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2448                 char   *querier;
2449 #ifdef CONFIG_IP_MULTICAST
2450                 querier = IGMP_V1_SEEN(state->in_dev) ? "V1" :
2451                           IGMP_V2_SEEN(state->in_dev) ? "V2" :
2452                           "V3";
2453 #else
2454                 querier = "NONE";
2455 #endif
2456
2457                 if (rcu_dereference(state->in_dev->mc_list) == im) {
2458                         seq_printf(seq, "%d\t%-10s: %5d %7s\n",
2459                                    state->dev->ifindex, state->dev->name, state->in_dev->mc_count, querier);
2460                 }
2461
2462                 seq_printf(seq,
2463                            "\t\t\t\t%08X %5d %d:%08lX\t\t%d\n",
2464                            im->multiaddr, im->users,
2465                            im->tm_running, im->tm_running ?
2466                            jiffies_to_clock_t(im->timer.expires-jiffies) : 0,
2467                            im->reporter);
2468         }
2469         return 0;
2470 }
2471
2472 static const struct seq_operations igmp_mc_seq_ops = {
2473         .start  =       igmp_mc_seq_start,
2474         .next   =       igmp_mc_seq_next,
2475         .stop   =       igmp_mc_seq_stop,
2476         .show   =       igmp_mc_seq_show,
2477 };
2478
2479 static int igmp_mc_seq_open(struct inode *inode, struct file *file)
2480 {
2481         return seq_open_net(inode, file, &igmp_mc_seq_ops,
2482                         sizeof(struct igmp_mc_iter_state));
2483 }
2484
2485 static const struct file_operations igmp_mc_seq_fops = {
2486         .owner          =       THIS_MODULE,
2487         .open           =       igmp_mc_seq_open,
2488         .read           =       seq_read,
2489         .llseek         =       seq_lseek,
2490         .release        =       seq_release_net,
2491 };
2492
2493 struct igmp_mcf_iter_state {
2494         struct seq_net_private p;
2495         struct net_device *dev;
2496         struct in_device *idev;
2497         struct ip_mc_list *im;
2498 };
2499
2500 #define igmp_mcf_seq_private(seq)       ((struct igmp_mcf_iter_state *)(seq)->private)
2501
2502 static inline struct ip_sf_list *igmp_mcf_get_first(struct seq_file *seq)
2503 {
2504         struct net *net = seq_file_net(seq);
2505         struct ip_sf_list *psf = NULL;
2506         struct ip_mc_list *im = NULL;
2507         struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2508
2509         state->idev = NULL;
2510         state->im = NULL;
2511         for_each_netdev_rcu(net, state->dev) {
2512                 struct in_device *idev;
2513                 idev = __in_dev_get_rcu(state->dev);
2514                 if (unlikely(idev == NULL))
2515                         continue;
2516                 im = rcu_dereference(idev->mc_list);
2517                 if (likely(im != NULL)) {
2518                         spin_lock_bh(&im->lock);
2519                         psf = im->sources;
2520                         if (likely(psf != NULL)) {
2521                                 state->im = im;
2522                                 state->idev = idev;
2523                                 break;
2524                         }
2525                         spin_unlock_bh(&im->lock);
2526                 }
2527         }
2528         return psf;
2529 }
2530
2531 static struct ip_sf_list *igmp_mcf_get_next(struct seq_file *seq, struct ip_sf_list *psf)
2532 {
2533         struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2534
2535         psf = psf->sf_next;
2536         while (!psf) {
2537                 spin_unlock_bh(&state->im->lock);
2538                 state->im = state->im->next;
2539                 while (!state->im) {
2540                         state->dev = next_net_device_rcu(state->dev);
2541                         if (!state->dev) {
2542                                 state->idev = NULL;
2543                                 goto out;
2544                         }
2545                         state->idev = __in_dev_get_rcu(state->dev);
2546                         if (!state->idev)
2547                                 continue;
2548                         state->im = rcu_dereference(state->idev->mc_list);
2549                 }
2550                 if (!state->im)
2551                         break;
2552                 spin_lock_bh(&state->im->lock);
2553                 psf = state->im->sources;
2554         }
2555 out:
2556         return psf;
2557 }
2558
2559 static struct ip_sf_list *igmp_mcf_get_idx(struct seq_file *seq, loff_t pos)
2560 {
2561         struct ip_sf_list *psf = igmp_mcf_get_first(seq);
2562         if (psf)
2563                 while (pos && (psf = igmp_mcf_get_next(seq, psf)) != NULL)
2564                         --pos;
2565         return pos ? NULL : psf;
2566 }
2567
2568 static void *igmp_mcf_seq_start(struct seq_file *seq, loff_t *pos)
2569         __acquires(rcu)
2570 {
2571         rcu_read_lock();
2572         return *pos ? igmp_mcf_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
2573 }
2574
2575 static void *igmp_mcf_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2576 {
2577         struct ip_sf_list *psf;
2578         if (v == SEQ_START_TOKEN)
2579                 psf = igmp_mcf_get_first(seq);
2580         else
2581                 psf = igmp_mcf_get_next(seq, v);
2582         ++*pos;
2583         return psf;
2584 }
2585
2586 static void igmp_mcf_seq_stop(struct seq_file *seq, void *v)
2587         __releases(rcu)
2588 {
2589         struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2590         if (likely(state->im != NULL)) {
2591                 spin_unlock_bh(&state->im->lock);
2592                 state->im = NULL;
2593         }
2594         state->idev = NULL;
2595         state->dev = NULL;
2596         rcu_read_unlock();
2597 }
2598
2599 static int igmp_mcf_seq_show(struct seq_file *seq, void *v)
2600 {
2601         struct ip_sf_list *psf = (struct ip_sf_list *)v;
2602         struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2603
2604         if (v == SEQ_START_TOKEN) {
2605                 seq_printf(seq,
2606                            "%3s %6s "
2607                            "%10s %10s %6s %6s\n", "Idx",
2608                            "Device", "MCA",
2609                            "SRC", "INC", "EXC");
2610         } else {
2611                 seq_printf(seq,
2612                            "%3d %6.6s 0x%08x "
2613                            "0x%08x %6lu %6lu\n",
2614                            state->dev->ifindex, state->dev->name,
2615                            ntohl(state->im->multiaddr),
2616                            ntohl(psf->sf_inaddr),
2617                            psf->sf_count[MCAST_INCLUDE],
2618                            psf->sf_count[MCAST_EXCLUDE]);
2619         }
2620         return 0;
2621 }
2622
2623 static const struct seq_operations igmp_mcf_seq_ops = {
2624         .start  =       igmp_mcf_seq_start,
2625         .next   =       igmp_mcf_seq_next,
2626         .stop   =       igmp_mcf_seq_stop,
2627         .show   =       igmp_mcf_seq_show,
2628 };
2629
2630 static int igmp_mcf_seq_open(struct inode *inode, struct file *file)
2631 {
2632         return seq_open_net(inode, file, &igmp_mcf_seq_ops,
2633                         sizeof(struct igmp_mcf_iter_state));
2634 }
2635
2636 static const struct file_operations igmp_mcf_seq_fops = {
2637         .owner          =       THIS_MODULE,
2638         .open           =       igmp_mcf_seq_open,
2639         .read           =       seq_read,
2640         .llseek         =       seq_lseek,
2641         .release        =       seq_release_net,
2642 };
2643
2644 static int __net_init igmp_net_init(struct net *net)
2645 {
2646         struct proc_dir_entry *pde;
2647
2648         pde = proc_net_fops_create(net, "igmp", S_IRUGO, &igmp_mc_seq_fops);
2649         if (!pde)
2650                 goto out_igmp;
2651         pde = proc_net_fops_create(net, "mcfilter", S_IRUGO, &igmp_mcf_seq_fops);
2652         if (!pde)
2653                 goto out_mcfilter;
2654         return 0;
2655
2656 out_mcfilter:
2657         proc_net_remove(net, "igmp");
2658 out_igmp:
2659         return -ENOMEM;
2660 }
2661
2662 static void __net_exit igmp_net_exit(struct net *net)
2663 {
2664         proc_net_remove(net, "mcfilter");
2665         proc_net_remove(net, "igmp");
2666 }
2667
2668 static struct pernet_operations igmp_net_ops = {
2669         .init = igmp_net_init,
2670         .exit = igmp_net_exit,
2671 };
2672
2673 int __init igmp_mc_proc_init(void)
2674 {
2675         return register_pernet_subsys(&igmp_net_ops);
2676 }
2677 #endif