net/ipv4/igmp.c

   1 /*
   2  *      Linux NET3:     Internet Group Management Protocol  [IGMP]
   3  *
   4  *      This code implements the IGMP protocol as defined in RFC1112. There has
   5  *      been a further revision of this protocol since which is now supported.
   6  *
   7  *      If you have trouble with this module be careful what gcc you have used,
   8  *      the older version didn't come out right using gcc 2.5.8, the newer one
   9  *      seems to fall out with gcc 2.6.2.
  10  *
  11  *      Authors:
  12  *              Alan Cox <alan@lxorguk.ukuu.org.uk>
  13  *
  14  *      This program is free software; you can redistribute it and/or
  15  *      modify it under the terms of the GNU General Public License
  16  *      as published by the Free Software Foundation; either version
  17  *      2 of the License, or (at your option) any later version.
  18  *
  19  *      Fixes:
  20  *
  21  *              Alan Cox        :       Added lots of __inline__ to optimise
  22  *                                      the memory usage of all the tiny little
  23  *                                      functions.
  24  *              Alan Cox        :       Dumped the header building experiment.
  25  *              Alan Cox        :       Minor tweaks ready for multicast routing
  26  *                                      and extended IGMP protocol.
  27  *              Alan Cox        :       Removed a load of inline directives. Gcc 2.5.8
  28  *                                      writes utterly bogus code otherwise (sigh)
  29  *                                      fixed IGMP loopback to behave in the manner
  30  *                                      desired by mrouted, fixed the fact it has been
  31  *                                      broken since 1.3.6 and cleaned up a few minor
  32  *                                      points.
  33  *
  34  *              Chih-Jen Chang  :       Tried to revise IGMP to Version 2
  35  *              Tsu-Sheng Tsao          E-mail: chihjenc@scf.usc.edu and tsusheng@scf.usc.edu
  36  *                                      The enhancements are mainly based on Steve Deering's
  37  *                                      ipmulti-3.5 source code.
  38  *              Chih-Jen Chang  :       Added the igmp_get_mrouter_info and
  39  *              Tsu-Sheng Tsao          igmp_set_mrouter_info to keep track of
  40  *                                      the mrouted version on that device.
  41  *              Chih-Jen Chang  :       Added the max_resp_time parameter to
  42  *              Tsu-Sheng Tsao          igmp_heard_query(). Using this parameter
  43  *                                      to identify the multicast router version
  44  *                                      and do what the IGMP version 2 specified.
  45  *              Chih-Jen Chang  :       Added a timer to revert to IGMP V2 router
  46  *              Tsu-Sheng Tsao          if the specified time expired.
  47  *              Alan Cox        :       Stop IGMP from 0.0.0.0 being accepted.
  48  *              Alan Cox        :       Use GFP_ATOMIC in the right places.
  49  *              Christian Daudt :       igmp timer wasn't set for local group
  50  *                                      memberships but was being deleted,
  51  *                                      which caused a "del_timer() called
  52  *                                      from %p with timer not initialized\n"
  53  *                                      message (960131).
  54  *              Christian Daudt :       removed del_timer from
  55  *                                      igmp_timer_expire function (960205).
  56  *             Christian Daudt :       igmp_heard_report now only calls
  57  *                                     igmp_timer_expire if tm->running is
  58  *                                     true (960216).
  59  *              Malcolm Beattie :       ttl comparison wrong in igmp_rcv made
  60  *                                      igmp_heard_query never trigger. Expiry
  61  *                                      miscalculation fixed in igmp_heard_query
  62  *                                      and random() made to return unsigned to
  63  *                                      prevent negative expiry times.
  64  *              Alexey Kuznetsov:       Wrong group leaving behaviour, backport
  65  *                                      fix from pending 2.1.x patches.
  66  *              Alan Cox:               Forget to enable FDDI support earlier.
  67  *              Alexey Kuznetsov:       Fixed leaving groups on device down.
  68  *              Alexey Kuznetsov:       Accordance to igmp-v2-06 draft.
  69  *              David L Stevens:        IGMPv3 support, with help from
  70  *                                      Vinay Kulkarni
  71  */
  72
  73 #include <linux/module.h>
  74 #include <linux/slab.h>
  75 #include <asm/uaccess.h>
  76 #include <asm/system.h>
  77 #include <linux/types.h>
  78 #include <linux/kernel.h>
  79 #include <linux/jiffies.h>
  80 #include <linux/string.h>
  81 #include <linux/socket.h>
  82 #include <linux/sockios.h>
  83 #include <linux/in.h>
  84 #include <linux/inet.h>
  85 #include <linux/netdevice.h>
  86 #include <linux/skbuff.h>
  87 #include <linux/inetdevice.h>
  88 #include <linux/igmp.h>
  89 #include <linux/if_arp.h>
  90 #include <linux/rtnetlink.h>
  91 #include <linux/times.h>
  92
  93 #include <net/net_namespace.h>
  94 #include <net/arp.h>
  95 #include <net/ip.h>
  96 #include <net/protocol.h>
  97 #include <net/route.h>
  98 #include <net/sock.h>
  99 #include <net/checksum.h>
 100 #include <linux/netfilter_ipv4.h>
 101 #ifdef CONFIG_IP_MROUTE
 102 #include <linux/mroute.h>
 103 #endif
 104 #ifdef CONFIG_PROC_FS
 105 #include <linux/proc_fs.h>
 106 #include <linux/seq_file.h>
 107 #endif
 108
 109 #define IP_MAX_MEMBERSHIPS      20
 110 #define IP_MAX_MSF              10
 111
 112 #ifdef CONFIG_IP_MULTICAST
 113 /* Parameter names and values are taken from igmp-v2-06 draft */
 114
 115 #define IGMP_V1_Router_Present_Timeout          (400*HZ)
 116 #define IGMP_V2_Router_Present_Timeout          (400*HZ)
 117 #define IGMP_Unsolicited_Report_Interval        (10*HZ)
 118 #define IGMP_Query_Response_Interval            (10*HZ)
 119 #define IGMP_Unsolicited_Report_Count           2
 120
 121
 122 #define IGMP_Initial_Report_Delay               (1)
 123
 124 /* IGMP_Initial_Report_Delay is not from IGMP specs!
 125  * IGMP specs require to report membership immediately after
 126  * joining a group, but we delay the first report by a
 127  * small interval. It seems more natural and still does not
 128  * contradict to specs provided this delay is small enough.
 129  */
 130
 131 #define IGMP_V1_SEEN(in_dev) \
 132         (IPV4_DEVCONF_ALL(dev_net(in_dev->dev), FORCE_IGMP_VERSION) == 1 || \
 133          IN_DEV_CONF_GET((in_dev), FORCE_IGMP_VERSION) == 1 || \
 134          ((in_dev)->mr_v1_seen && \
 135           time_before(jiffies, (in_dev)->mr_v1_seen)))
 136 #define IGMP_V2_SEEN(in_dev) \
 137         (IPV4_DEVCONF_ALL(dev_net(in_dev->dev), FORCE_IGMP_VERSION) == 2 || \
 138          IN_DEV_CONF_GET((in_dev), FORCE_IGMP_VERSION) == 2 || \
 139          ((in_dev)->mr_v2_seen && \
 140           time_before(jiffies, (in_dev)->mr_v2_seen)))
 141
 142 static void igmpv3_add_delrec(struct in_device *in_dev, struct ip_mc_list *im);
 143 static void igmpv3_del_delrec(struct in_device *in_dev, __be32 multiaddr);
 144 static void igmpv3_clear_delrec(struct in_device *in_dev);
 145 static int sf_setstate(struct ip_mc_list *pmc);
 146 static void sf_markstate(struct ip_mc_list *pmc);
 147 #endif
 148 static void ip_mc_clear_src(struct ip_mc_list *pmc);
 149 static int ip_mc_add_src(struct in_device *in_dev, __be32 *pmca, int sfmode,
 150                          int sfcount, __be32 *psfsrc, int delta);
 151
 152 static void ip_ma_put(struct ip_mc_list *im)
 153 {
 154         if (atomic_dec_and_test(&im->refcnt)) {
 155                 in_dev_put(im->interface);
 156                 kfree_rcu(im, rcu);
 157         }
 158 }
 159
 160 #define for_each_pmc_rcu(in_dev, pmc)                           \
 161         for (pmc = rcu_dereference(in_dev->mc_list);            \
 162              pmc != NULL;                                       \
 163              pmc = rcu_dereference(pmc->next_rcu))
 164
 165 #define for_each_pmc_rtnl(in_dev, pmc)                          \
 166         for (pmc = rtnl_dereference(in_dev->mc_list);           \
 167              pmc != NULL;                                       \
 168              pmc = rtnl_dereference(pmc->next_rcu))
 169
 170 #ifdef CONFIG_IP_MULTICAST
 171
 172 /*
 173  *      Timer management
 174  */
 175
 176 static void igmp_stop_timer(struct ip_mc_list *im)
 177 {
 178         spin_lock_bh(&im->lock);
 179         if (del_timer(&im->timer))
 180                 atomic_dec(&im->refcnt);
 181         im->tm_running = 0;
 182         im->reporter = 0;
 183         im->unsolicit_count = 0;
 184         spin_unlock_bh(&im->lock);
 185 }
 186
 187 /* It must be called with locked im->lock */
 188 static void igmp_start_timer(struct ip_mc_list *im, int max_delay)
 189 {
 190         int tv = net_random() % max_delay;
 191
 192         im->tm_running = 1;
 193         if (!mod_timer(&im->timer, jiffies+tv+2))
 194                 atomic_inc(&im->refcnt);
 195 }
 196
 197 static void igmp_gq_start_timer(struct in_device *in_dev)
 198 {
 199         int tv = net_random() % in_dev->mr_maxdelay;
 200
 201         in_dev->mr_gq_running = 1;
 202         if (!mod_timer(&in_dev->mr_gq_timer, jiffies+tv+2))
 203                 in_dev_hold(in_dev);
 204 }
 205
 206 static void igmp_ifc_start_timer(struct in_device *in_dev, int delay)
 207 {
 208         int tv = net_random() % delay;
 209
 210         if (!mod_timer(&in_dev->mr_ifc_timer, jiffies+tv+2))
 211                 in_dev_hold(in_dev);
 212 }
 213
 214 static void igmp_mod_timer(struct ip_mc_list *im, int max_delay)
 215 {
 216         spin_lock_bh(&im->lock);
 217         im->unsolicit_count = 0;
 218         if (del_timer(&im->timer)) {
 219                 if ((long)(im->timer.expires-jiffies) < max_delay) {
 220                         add_timer(&im->timer);
 221                         im->tm_running = 1;
 222                         spin_unlock_bh(&im->lock);
 223                         return;
 224                 }
 225                 atomic_dec(&im->refcnt);
 226         }
 227         igmp_start_timer(im, max_delay);
 228         spin_unlock_bh(&im->lock);
 229 }
 230
 231
 232 /*
 233  *      Send an IGMP report.
 234  */
 235
 236 #define IGMP_SIZE (sizeof(struct igmphdr)+sizeof(struct iphdr)+4)
 237
 238
 239 static int is_in(struct ip_mc_list *pmc, struct ip_sf_list *psf, int type,
 240         int gdeleted, int sdeleted)
 241 {
 242         switch (type) {
 243         case IGMPV3_MODE_IS_INCLUDE:
 244         case IGMPV3_MODE_IS_EXCLUDE:
 245                 if (gdeleted || sdeleted)
 246                         return 0;
 247                 if (!(pmc->gsquery && !psf->sf_gsresp)) {
 248                         if (pmc->sfmode == MCAST_INCLUDE)
 249                                 return 1;
 250                         /* don't include if this source is excluded
 251                          * in all filters
 252                          */
 253                         if (psf->sf_count[MCAST_INCLUDE])
 254                                 return type == IGMPV3_MODE_IS_INCLUDE;
 255                         return pmc->sfcount[MCAST_EXCLUDE] ==
 256                                 psf->sf_count[MCAST_EXCLUDE];
 257                 }
 258                 return 0;
 259         case IGMPV3_CHANGE_TO_INCLUDE:
 260                 if (gdeleted || sdeleted)
 261                         return 0;
 262                 return psf->sf_count[MCAST_INCLUDE] != 0;
 263         case IGMPV3_CHANGE_TO_EXCLUDE:
 264                 if (gdeleted || sdeleted)
 265                         return 0;
 266                 if (pmc->sfcount[MCAST_EXCLUDE] == 0 ||
 267                     psf->sf_count[MCAST_INCLUDE])
 268                         return 0;
 269                 return pmc->sfcount[MCAST_EXCLUDE] ==
 270                         psf->sf_count[MCAST_EXCLUDE];
 271         case IGMPV3_ALLOW_NEW_SOURCES:
 272                 if (gdeleted || !psf->sf_crcount)
 273                         return 0;
 274                 return (pmc->sfmode == MCAST_INCLUDE) ^ sdeleted;
 275         case IGMPV3_BLOCK_OLD_SOURCES:
 276                 if (pmc->sfmode == MCAST_INCLUDE)
 277                         return gdeleted || (psf->sf_crcount && sdeleted);
 278                 return psf->sf_crcount && !gdeleted && !sdeleted;
 279         }
 280         return 0;
 281 }
 282
 283 static int
 284 igmp_scount(struct ip_mc_list *pmc, int type, int gdeleted, int sdeleted)
 285 {
 286         struct ip_sf_list *psf;
 287         int scount = 0;
 288
 289         for (psf=pmc->sources; psf; psf=psf->sf_next) {
 290                 if (!is_in(pmc, psf, type, gdeleted, sdeleted))
 291                         continue;
 292                 scount++;
 293         }
 294         return scount;
 295 }
 296
 297 #define igmp_skb_size(skb) (*(unsigned int *)((skb)->cb))
 298
 299 static struct sk_buff *igmpv3_newpack(struct net_device *dev, int size)
 300 {
 301         struct sk_buff *skb;
 302         struct rtable *rt;
 303         struct iphdr *pip;
 304         struct igmpv3_report *pig;
 305         struct net *net = dev_net(dev);
 306
 307         while (1) {
 308                 skb = alloc_skb(size + LL_ALLOCATED_SPACE(dev),
 309                                 GFP_ATOMIC | __GFP_NOWARN);
 310                 if (skb)
 311                         break;
 312                 size >>= 1;
 313                 if (size < 256)
 314                         return NULL;
 315         }
 316         igmp_skb_size(skb) = size;
 317
 318         rt = ip_route_output_ports(net, NULL, IGMPV3_ALL_MCR, 0,
 319                                    0, 0,
 320                                    IPPROTO_IGMP, 0, dev->ifindex);
 321         if (IS_ERR(rt)) {
 322                 kfree_skb(skb);
 323                 return NULL;
 324         }
 325         if (rt->rt_src == 0) {
 326                 kfree_skb(skb);
 327                 ip_rt_put(rt);
 328                 return NULL;
 329         }
 330
 331         skb_dst_set(skb, &rt->dst);
 332         skb->dev = dev;
 333
 334         skb_reserve(skb, LL_RESERVED_SPACE(dev));
 335
 336         skb_reset_network_header(skb);
 337         pip = ip_hdr(skb);
 338         skb_put(skb, sizeof(struct iphdr) + 4);
 339
 340         pip->version  = 4;
 341         pip->ihl      = (sizeof(struct iphdr)+4)>>2;
 342         pip->tos      = 0xc0;
 343         pip->frag_off = htons(IP_DF);
 344         pip->ttl      = 1;
 345         pip->daddr    = rt->rt_dst;
 346         pip->saddr    = rt->rt_src;
 347         pip->protocol = IPPROTO_IGMP;
 348         pip->tot_len  = 0;      /* filled in later */
 349         ip_select_ident(pip, &rt->dst, NULL);
 350         ((u8*)&pip[1])[0] = IPOPT_RA;
 351         ((u8*)&pip[1])[1] = 4;
 352         ((u8*)&pip[1])[2] = 0;
 353         ((u8*)&pip[1])[3] = 0;
 354
 355         skb->transport_header = skb->network_header + sizeof(struct iphdr) + 4;
 356         skb_put(skb, sizeof(*pig));
 357         pig = igmpv3_report_hdr(skb);
 358         pig->type = IGMPV3_HOST_MEMBERSHIP_REPORT;
 359         pig->resv1 = 0;
 360         pig->csum = 0;
 361         pig->resv2 = 0;
 362         pig->ngrec = 0;
 363         return skb;
 364 }
 365
 366 static int igmpv3_sendpack(struct sk_buff *skb)
 367 {
 368         struct igmphdr *pig = igmp_hdr(skb);
 369         const int igmplen = skb->tail - skb->transport_header;
 370
 371         pig->csum = ip_compute_csum(igmp_hdr(skb), igmplen);
 372
 373         return ip_local_out(skb);
 374 }
 375
 376 static int grec_size(struct ip_mc_list *pmc, int type, int gdel, int sdel)
 377 {
 378         return sizeof(struct igmpv3_grec) + 4*igmp_scount(pmc, type, gdel, sdel);
 379 }
 380
 381 static struct sk_buff *add_grhead(struct sk_buff *skb, struct ip_mc_list *pmc,
 382         int type, struct igmpv3_grec **ppgr)
 383 {
 384         struct net_device *dev = pmc->interface->dev;
 385         struct igmpv3_report *pih;
 386         struct igmpv3_grec *pgr;
 387
 388         if (!skb)
 389                 skb = igmpv3_newpack(dev, dev->mtu);
 390         if (!skb)
 391                 return NULL;
 392         pgr = (struct igmpv3_grec *)skb_put(skb, sizeof(struct igmpv3_grec));
 393         pgr->grec_type = type;
 394         pgr->grec_auxwords = 0;
 395         pgr->grec_nsrcs = 0;
 396         pgr->grec_mca = pmc->multiaddr;
 397         pih = igmpv3_report_hdr(skb);
 398         pih->ngrec = htons(ntohs(pih->ngrec)+1);
 399         *ppgr = pgr;
 400         return skb;
 401 }
 402
 403 #define AVAILABLE(skb) ((skb) ? ((skb)->dev ? igmp_skb_size(skb) - (skb)->len : \
 404         skb_tailroom(skb)) : 0)
 405
 406 static struct sk_buff *add_grec(struct sk_buff *skb, struct ip_mc_list *pmc,
 407         int type, int gdeleted, int sdeleted)
 408 {
 409         struct net_device *dev = pmc->interface->dev;
 410         struct igmpv3_report *pih;
 411         struct igmpv3_grec *pgr = NULL;
 412         struct ip_sf_list *psf, *psf_next, *psf_prev, **psf_list;
 413         int scount, stotal, first, isquery, truncate;
 414
 415         if (pmc->multiaddr == IGMP_ALL_HOSTS)
 416                 return skb;
 417
 418         isquery = type == IGMPV3_MODE_IS_INCLUDE ||
 419                   type == IGMPV3_MODE_IS_EXCLUDE;
 420         truncate = type == IGMPV3_MODE_IS_EXCLUDE ||
 421                     type == IGMPV3_CHANGE_TO_EXCLUDE;
 422
 423         stotal = scount = 0;
 424
 425         psf_list = sdeleted ? &pmc->tomb : &pmc->sources;
 426
 427         if (!*psf_list)
 428                 goto empty_source;
 429
 430         pih = skb ? igmpv3_report_hdr(skb) : NULL;
 431
 432         /* EX and TO_EX get a fresh packet, if needed */
 433         if (truncate) {
 434                 if (pih && pih->ngrec &&
 435                     AVAILABLE(skb) < grec_size(pmc, type, gdeleted, sdeleted)) {
 436                         if (skb)
 437                                 igmpv3_sendpack(skb);
 438                         skb = igmpv3_newpack(dev, dev->mtu);
 439                 }
 440         }
 441         first = 1;
 442         psf_prev = NULL;
 443         for (psf=*psf_list; psf; psf=psf_next) {
 444                 __be32 *psrc;
 445
 446                 psf_next = psf->sf_next;
 447
 448                 if (!is_in(pmc, psf, type, gdeleted, sdeleted)) {
 449                         psf_prev = psf;
 450                         continue;
 451                 }
 452
 453                 /* clear marks on query responses */
 454                 if (isquery)
 455                         psf->sf_gsresp = 0;
 456
 457                 if (AVAILABLE(skb) < sizeof(__be32) +
 458                     first*sizeof(struct igmpv3_grec)) {
 459                         if (truncate && !first)
 460                                 break;   /* truncate these */
 461                         if (pgr)
 462                                 pgr->grec_nsrcs = htons(scount);
 463                         if (skb)
 464                                 igmpv3_sendpack(skb);
 465                         skb = igmpv3_newpack(dev, dev->mtu);
 466                         first = 1;
 467                         scount = 0;
 468                 }
 469                 if (first) {
 470                         skb = add_grhead(skb, pmc, type, &pgr);
 471                         first = 0;
 472                 }
 473                 if (!skb)
 474                         return NULL;
 475                 psrc = (__be32 *)skb_put(skb, sizeof(__be32));
 476                 *psrc = psf->sf_inaddr;
 477                 scount++; stotal++;
 478                 if ((type == IGMPV3_ALLOW_NEW_SOURCES ||
 479                      type == IGMPV3_BLOCK_OLD_SOURCES) && psf->sf_crcount) {
 480                         psf->sf_crcount--;
 481                         if ((sdeleted || gdeleted) && psf->sf_crcount == 0) {
 482                                 if (psf_prev)
 483                                         psf_prev->sf_next = psf->sf_next;
 484                                 else
 485                                         *psf_list = psf->sf_next;
 486                                 kfree(psf);
 487                                 continue;
 488                         }
 489                 }
 490                 psf_prev = psf;
 491         }
 492
 493 empty_source:
 494         if (!stotal) {
 495                 if (type == IGMPV3_ALLOW_NEW_SOURCES ||
 496                     type == IGMPV3_BLOCK_OLD_SOURCES)
 497                         return skb;
 498                 if (pmc->crcount || isquery) {
 499                         /* make sure we have room for group header */
 500                         if (skb && AVAILABLE(skb)<sizeof(struct igmpv3_grec)) {
 501                                 igmpv3_sendpack(skb);
 502                                 skb = NULL; /* add_grhead will get a new one */
 503                         }
 504                         skb = add_grhead(skb, pmc, type, &pgr);
 505                 }
 506         }
 507         if (pgr)
 508                 pgr->grec_nsrcs = htons(scount);
 509
 510         if (isquery)
 511                 pmc->gsquery = 0;       /* clear query state on report */
 512         return skb;
 513 }
 514
 515 static int igmpv3_send_report(struct in_device *in_dev, struct ip_mc_list *pmc)
 516 {
 517         struct sk_buff *skb = NULL;
 518         int type;
 519
 520         if (!pmc) {
 521                 rcu_read_lock();
 522                 for_each_pmc_rcu(in_dev, pmc) {
 523                         if (pmc->multiaddr == IGMP_ALL_HOSTS)
 524                                 continue;
 525                         spin_lock_bh(&pmc->lock);
 526                         if (pmc->sfcount[MCAST_EXCLUDE])
 527                                 type = IGMPV3_MODE_IS_EXCLUDE;
 528                         else
 529                                 type = IGMPV3_MODE_IS_INCLUDE;
 530                         skb = add_grec(skb, pmc, type, 0, 0);
 531                         spin_unlock_bh(&pmc->lock);
 532                 }
 533                 rcu_read_unlock();
 534         } else {
 535                 spin_lock_bh(&pmc->lock);
 536                 if (pmc->sfcount[MCAST_EXCLUDE])
 537                         type = IGMPV3_MODE_IS_EXCLUDE;
 538                 else
 539                         type = IGMPV3_MODE_IS_INCLUDE;
 540                 skb = add_grec(skb, pmc, type, 0, 0);
 541                 spin_unlock_bh(&pmc->lock);
 542         }
 543         if (!skb)
 544                 return 0;
 545         return igmpv3_sendpack(skb);
 546 }
 547
 548 /*
 549  * remove zero-count source records from a source filter list
 550  */
 551 static void igmpv3_clear_zeros(struct ip_sf_list **ppsf)
 552 {
 553         struct ip_sf_list *psf_prev, *psf_next, *psf;
 554
 555         psf_prev = NULL;
 556         for (psf=*ppsf; psf; psf = psf_next) {
 557                 psf_next = psf->sf_next;
 558                 if (psf->sf_crcount == 0) {
 559                         if (psf_prev)
 560                                 psf_prev->sf_next = psf->sf_next;
 561                         else
 562                                 *ppsf = psf->sf_next;
 563                         kfree(psf);
 564                 } else
 565                         psf_prev = psf;
 566         }
 567 }
 568
 569 static void igmpv3_send_cr(struct in_device *in_dev)
 570 {
 571         struct ip_mc_list *pmc, *pmc_prev, *pmc_next;
 572         struct sk_buff *skb = NULL;
 573         int type, dtype;
 574
 575         rcu_read_lock();
 576         spin_lock_bh(&in_dev->mc_tomb_lock);
 577
 578         /* deleted MCA's */
 579         pmc_prev = NULL;
 580         for (pmc=in_dev->mc_tomb; pmc; pmc=pmc_next) {
 581                 pmc_next = pmc->next;
 582                 if (pmc->sfmode == MCAST_INCLUDE) {
 583                         type = IGMPV3_BLOCK_OLD_SOURCES;
 584                         dtype = IGMPV3_BLOCK_OLD_SOURCES;
 585                         skb = add_grec(skb, pmc, type, 1, 0);
 586                         skb = add_grec(skb, pmc, dtype, 1, 1);
 587                 }
 588                 if (pmc->crcount) {
 589                         if (pmc->sfmode == MCAST_EXCLUDE) {
 590                                 type = IGMPV3_CHANGE_TO_INCLUDE;
 591                                 skb = add_grec(skb, pmc, type, 1, 0);
 592                         }
 593                         pmc->crcount--;
 594                         if (pmc->crcount == 0) {
 595                                 igmpv3_clear_zeros(&pmc->tomb);
 596                                 igmpv3_clear_zeros(&pmc->sources);
 597                         }
 598                 }
 599                 if (pmc->crcount == 0 && !pmc->tomb && !pmc->sources) {
 600                         if (pmc_prev)
 601                                 pmc_prev->next = pmc_next;
 602                         else
 603                                 in_dev->mc_tomb = pmc_next;
 604                         in_dev_put(pmc->interface);
 605                         kfree(pmc);
 606                 } else
 607                         pmc_prev = pmc;
 608         }
 609         spin_unlock_bh(&in_dev->mc_tomb_lock);
 610
 611         /* change recs */
 612         for_each_pmc_rcu(in_dev, pmc) {
 613                 spin_lock_bh(&pmc->lock);
 614                 if (pmc->sfcount[MCAST_EXCLUDE]) {
 615                         type = IGMPV3_BLOCK_OLD_SOURCES;
 616                         dtype = IGMPV3_ALLOW_NEW_SOURCES;
 617                 } else {
 618                         type = IGMPV3_ALLOW_NEW_SOURCES;
 619                         dtype = IGMPV3_BLOCK_OLD_SOURCES;
 620                 }
 621                 skb = add_grec(skb, pmc, type, 0, 0);
 622                 skb = add_grec(skb, pmc, dtype, 0, 1);  /* deleted sources */
 623
 624                 /* filter mode changes */
 625                 if (pmc->crcount) {
 626                         if (pmc->sfmode == MCAST_EXCLUDE)
 627                                 type = IGMPV3_CHANGE_TO_EXCLUDE;
 628                         else
 629                                 type = IGMPV3_CHANGE_TO_INCLUDE;
 630                         skb = add_grec(skb, pmc, type, 0, 0);
 631                         pmc->crcount--;
 632                 }
 633                 spin_unlock_bh(&pmc->lock);
 634         }
 635         rcu_read_unlock();
 636
 637         if (!skb)
 638                 return;
 639         (void) igmpv3_sendpack(skb);
 640 }
 641
 642 static int igmp_send_report(struct in_device *in_dev, struct ip_mc_list *pmc,
 643         int type)
 644 {
 645         struct sk_buff *skb;
 646         struct iphdr *iph;
 647         struct igmphdr *ih;
 648         struct rtable *rt;
 649         struct net_device *dev = in_dev->dev;
 650         struct net *net = dev_net(dev);
 651         __be32  group = pmc ? pmc->multiaddr : 0;
 652         __be32  dst;
 653
 654         if (type == IGMPV3_HOST_MEMBERSHIP_REPORT)
 655                 return igmpv3_send_report(in_dev, pmc);
 656         else if (type == IGMP_HOST_LEAVE_MESSAGE)
 657                 dst = IGMP_ALL_ROUTER;
 658         else
 659                 dst = group;
 660
 661         rt = ip_route_output_ports(net, NULL, dst, 0,
 662                                    0, 0,
 663                                    IPPROTO_IGMP, 0, dev->ifindex);
 664         if (IS_ERR(rt))
 665                 return -1;
 666
 667         if (rt->rt_src == 0) {
 668                 ip_rt_put(rt);
 669                 return -1;
 670         }
 671
 672         skb = alloc_skb(IGMP_SIZE+LL_ALLOCATED_SPACE(dev), GFP_ATOMIC);
 673         if (skb == NULL) {
 674                 ip_rt_put(rt);
 675                 return -1;
 676         }
 677
 678         skb_dst_set(skb, &rt->dst);
 679
 680         skb_reserve(skb, LL_RESERVED_SPACE(dev));
 681
 682         skb_reset_network_header(skb);
 683         iph = ip_hdr(skb);
 684         skb_put(skb, sizeof(struct iphdr) + 4);
 685
 686         iph->version  = 4;
 687         iph->ihl      = (sizeof(struct iphdr)+4)>>2;
 688         iph->tos      = 0xc0;
 689         iph->frag_off = htons(IP_DF);
 690         iph->ttl      = 1;
 691         iph->daddr    = dst;
 692         iph->saddr    = rt->rt_src;
 693         iph->protocol = IPPROTO_IGMP;
 694         ip_select_ident(iph, &rt->dst, NULL);
 695         ((u8*)&iph[1])[0] = IPOPT_RA;
 696         ((u8*)&iph[1])[1] = 4;
 697         ((u8*)&iph[1])[2] = 0;
 698         ((u8*)&iph[1])[3] = 0;
 699
 700         ih = (struct igmphdr *)skb_put(skb, sizeof(struct igmphdr));
 701         ih->type = type;
 702         ih->code = 0;
 703         ih->csum = 0;
 704         ih->group = group;
 705         ih->csum = ip_compute_csum((void *)ih, sizeof(struct igmphdr));
 706
 707         return ip_local_out(skb);
 708 }
 709
 710 static void igmp_gq_timer_expire(unsigned long data)
 711 {
 712         struct in_device *in_dev = (struct in_device *)data;
 713
 714         in_dev->mr_gq_running = 0;
 715         igmpv3_send_report(in_dev, NULL);
 716         __in_dev_put(in_dev);
 717 }
 718
 719 static void igmp_ifc_timer_expire(unsigned long data)
 720 {
 721         struct in_device *in_dev = (struct in_device *)data;
 722
 723         igmpv3_send_cr(in_dev);
 724         if (in_dev->mr_ifc_count) {
 725                 in_dev->mr_ifc_count--;
 726                 igmp_ifc_start_timer(in_dev, IGMP_Unsolicited_Report_Interval);
 727         }
 728         __in_dev_put(in_dev);
 729 }
 730
 731 static void igmp_ifc_event(struct in_device *in_dev)
 732 {
 733         if (IGMP_V1_SEEN(in_dev) || IGMP_V2_SEEN(in_dev))
 734                 return;
 735         in_dev->mr_ifc_count = in_dev->mr_qrv ? in_dev->mr_qrv :
 736                 IGMP_Unsolicited_Report_Count;
 737         igmp_ifc_start_timer(in_dev, 1);
 738 }
 739
 740
 741 static void igmp_timer_expire(unsigned long data)
 742 {
 743         struct ip_mc_list *im=(struct ip_mc_list *)data;
 744         struct in_device *in_dev = im->interface;
 745
 746         spin_lock(&im->lock);
 747         im->tm_running = 0;
 748
 749         if (im->unsolicit_count) {
 750                 im->unsolicit_count--;
 751                 igmp_start_timer(im, IGMP_Unsolicited_Report_Interval);
 752         }
 753         im->reporter = 1;
 754         spin_unlock(&im->lock);
 755
 756         if (IGMP_V1_SEEN(in_dev))
 757                 igmp_send_report(in_dev, im, IGMP_HOST_MEMBERSHIP_REPORT);
 758         else if (IGMP_V2_SEEN(in_dev))
 759                 igmp_send_report(in_dev, im, IGMPV2_HOST_MEMBERSHIP_REPORT);
 760         else
 761                 igmp_send_report(in_dev, im, IGMPV3_HOST_MEMBERSHIP_REPORT);
 762
 763         ip_ma_put(im);
 764 }
 765
 766 /* mark EXCLUDE-mode sources */
 767 static int igmp_xmarksources(struct ip_mc_list *pmc, int nsrcs, __be32 *srcs)
 768 {
 769         struct ip_sf_list *psf;
 770         int i, scount;
 771
 772         scount = 0;
 773         for (psf=pmc->sources; psf; psf=psf->sf_next) {
 774                 if (scount == nsrcs)
 775                         break;
 776                 for (i=0; i<nsrcs; i++) {
 777                         /* skip inactive filters */
 778                         if (pmc->sfcount[MCAST_INCLUDE] ||
 779                             pmc->sfcount[MCAST_EXCLUDE] !=
 780                             psf->sf_count[MCAST_EXCLUDE])
 781                                 continue;
 782                         if (srcs[i] == psf->sf_inaddr) {
 783                                 scount++;
 784                                 break;
 785                         }
 786                 }
 787         }
 788         pmc->gsquery = 0;
 789         if (scount == nsrcs)    /* all sources excluded */
 790                 return 0;
 791         return 1;
 792 }
 793
 794 static int igmp_marksources(struct ip_mc_list *pmc, int nsrcs, __be32 *srcs)
 795 {
 796         struct ip_sf_list *psf;
 797         int i, scount;
 798
 799         if (pmc->sfmode == MCAST_EXCLUDE)
 800                 return igmp_xmarksources(pmc, nsrcs, srcs);
 801
 802         /* mark INCLUDE-mode sources */
 803         scount = 0;
 804         for (psf=pmc->sources; psf; psf=psf->sf_next) {
 805                 if (scount == nsrcs)
 806                         break;
 807                 for (i=0; i<nsrcs; i++)
 808                         if (srcs[i] == psf->sf_inaddr) {
 809                                 psf->sf_gsresp = 1;
 810                                 scount++;
 811                                 break;
 812                         }
 813         }
 814         if (!scount) {
 815                 pmc->gsquery = 0;
 816                 return 0;
 817         }
 818         pmc->gsquery = 1;
 819         return 1;
 820 }
 821
 822 static void igmp_heard_report(struct in_device *in_dev, __be32 group)
 823 {
 824         struct ip_mc_list *im;
 825
 826         /* Timers are only set for non-local groups */
 827
 828         if (group == IGMP_ALL_HOSTS)
 829                 return;
 830
 831         rcu_read_lock();
 832         for_each_pmc_rcu(in_dev, im) {
 833                 if (im->multiaddr == group) {
 834                         igmp_stop_timer(im);
 835                         break;
 836                 }
 837         }
 838         rcu_read_unlock();
 839 }
 840
 841 static void igmp_heard_query(struct in_device *in_dev, struct sk_buff *skb,
 842         int len)
 843 {
 844         struct igmphdr          *ih = igmp_hdr(skb);
 845         struct igmpv3_query *ih3 = igmpv3_query_hdr(skb);
 846         struct ip_mc_list       *im;
 847         __be32                  group = ih->group;
 848         int                     max_delay;
 849         int                     mark = 0;
 850
 851
 852         if (len == 8) {
 853                 if (ih->code == 0) {
 854                         /* Alas, old v1 router presents here. */
 855
 856                         max_delay = IGMP_Query_Response_Interval;
 857                         in_dev->mr_v1_seen = jiffies +
 858                                 IGMP_V1_Router_Present_Timeout;
 859                         group = 0;
 860                 } else {
 861                         /* v2 router present */
 862                         max_delay = ih->code*(HZ/IGMP_TIMER_SCALE);
 863                         in_dev->mr_v2_seen = jiffies +
 864                                 IGMP_V2_Router_Present_Timeout;
 865                 }
 866                 /* cancel the interface change timer */
 867                 in_dev->mr_ifc_count = 0;
 868                 if (del_timer(&in_dev->mr_ifc_timer))
 869                         __in_dev_put(in_dev);
 870                 /* clear deleted report items */
 871                 igmpv3_clear_delrec(in_dev);
 872         } else if (len < 12) {
 873                 return; /* ignore bogus packet; freed by caller */
 874         } else if (IGMP_V1_SEEN(in_dev)) {
 875                 /* This is a v3 query with v1 queriers present */
 876                 max_delay = IGMP_Query_Response_Interval;
 877                 group = 0;
 878         } else if (IGMP_V2_SEEN(in_dev)) {
 879                 /* this is a v3 query with v2 queriers present;
 880                  * Interpretation of the max_delay code is problematic here.
 881                  * A real v2 host would use ih_code directly, while v3 has a
 882                  * different encoding. We use the v3 encoding as more likely
 883                  * to be intended in a v3 query.
 884                  */
 885                 max_delay = IGMPV3_MRC(ih3->code)*(HZ/IGMP_TIMER_SCALE);
 886         } else { /* v3 */
 887                 if (!pskb_may_pull(skb, sizeof(struct igmpv3_query)))
 888                         return;
 889
 890                 ih3 = igmpv3_query_hdr(skb);
 891                 if (ih3->nsrcs) {
 892                         if (!pskb_may_pull(skb, sizeof(struct igmpv3_query)
 893                                            + ntohs(ih3->nsrcs)*sizeof(__be32)))
 894                                 return;
 895                         ih3 = igmpv3_query_hdr(skb);
 896                 }
 897
 898                 max_delay = IGMPV3_MRC(ih3->code)*(HZ/IGMP_TIMER_SCALE);
 899                 if (!max_delay)
 900                         max_delay = 1;  /* can't mod w/ 0 */
 901                 in_dev->mr_maxdelay = max_delay;
 902                 if (ih3->qrv)
 903                         in_dev->mr_qrv = ih3->qrv;
 904                 if (!group) { /* general query */
 905                         if (ih3->nsrcs)
 906                                 return; /* no sources allowed */
 907                         igmp_gq_start_timer(in_dev);
 908                         return;
 909                 }
 910                 /* mark sources to include, if group & source-specific */
 911                 mark = ih3->nsrcs != 0;
 912         }
 913
 914         /*
 915          * - Start the timers in all of our membership records
 916          *   that the query applies to for the interface on
 917          *   which the query arrived excl. those that belong
 918          *   to a "local" group (224.0.0.X)
 919          * - For timers already running check if they need to
 920          *   be reset.
 921          * - Use the igmp->igmp_code field as the maximum
 922          *   delay possible
 923          */
 924         rcu_read_lock();
 925         for_each_pmc_rcu(in_dev, im) {
 926                 int changed;
 927
 928                 if (group && group != im->multiaddr)
 929                         continue;
 930                 if (im->multiaddr == IGMP_ALL_HOSTS)
 931                         continue;
 932                 spin_lock_bh(&im->lock);
 933                 if (im->tm_running)
 934                         im->gsquery = im->gsquery && mark;
 935                 else
 936                         im->gsquery = mark;
 937                 changed = !im->gsquery ||
 938                         igmp_marksources(im, ntohs(ih3->nsrcs), ih3->srcs);
 939                 spin_unlock_bh(&im->lock);
 940                 if (changed)
 941                         igmp_mod_timer(im, max_delay);
 942         }
 943         rcu_read_unlock();
 944 }
 945
 946 /* called in rcu_read_lock() section */
 947 int igmp_rcv(struct sk_buff *skb)
 948 {
 949         /* This basically follows the spec line by line -- see RFC1112 */
 950         struct igmphdr *ih;
 951         struct in_device *in_dev = __in_dev_get_rcu(skb->dev);
 952         int len = skb->len;
 953
 954         if (in_dev == NULL)
 955                 goto drop;
 956
 957         if (!pskb_may_pull(skb, sizeof(struct igmphdr)))
 958                 goto drop;
 959
 960         switch (skb->ip_summed) {
 961         case CHECKSUM_COMPLETE:
 962                 if (!csum_fold(skb->csum))
 963                         break;
 964                 /* fall through */
 965         case CHECKSUM_NONE:
 966                 skb->csum = 0;
 967                 if (__skb_checksum_complete(skb))
 968                         goto drop;
 969         }
 970
 971         ih = igmp_hdr(skb);
 972         switch (ih->type) {
 973         case IGMP_HOST_MEMBERSHIP_QUERY:
 974                 igmp_heard_query(in_dev, skb, len);
 975                 break;
 976         case IGMP_HOST_MEMBERSHIP_REPORT:
 977         case IGMPV2_HOST_MEMBERSHIP_REPORT:
 978                 /* Is it our report looped back? */
 979                 if (rt_is_output_route(skb_rtable(skb)))
 980                         break;
 981                 /* don't rely on MC router hearing unicast reports */
 982                 if (skb->pkt_type == PACKET_MULTICAST ||
 983                     skb->pkt_type == PACKET_BROADCAST)
 984                         igmp_heard_report(in_dev, ih->group);
 985                 break;
 986         case IGMP_PIM:
 987 #ifdef CONFIG_IP_PIMSM_V1
 988                 return pim_rcv_v1(skb);
 989 #endif
 990         case IGMPV3_HOST_MEMBERSHIP_REPORT:
 991         case IGMP_DVMRP:
 992         case IGMP_TRACE:
 993         case IGMP_HOST_LEAVE_MESSAGE:
 994         case IGMP_MTRACE:
 995         case IGMP_MTRACE_RESP:
 996                 break;
 997         default:
 998                 break;
 999         }
1000
1001 drop:
1002         kfree_skb(skb);
1003         return 0;
1004 }
1005
1006 #endif
1007
1008
1009 /*
1010  *      Add a filter to a device
1011  */
1012
1013 static void ip_mc_filter_add(struct in_device *in_dev, __be32 addr)
1014 {
1015         char buf[MAX_ADDR_LEN];
1016         struct net_device *dev = in_dev->dev;
1017
1018         /* Checking for IFF_MULTICAST here is WRONG-WRONG-WRONG.
1019            We will get multicast token leakage, when IFF_MULTICAST
1020            is changed. This check should be done in dev->set_multicast_list
1021            routine. Something sort of:
1022            if (dev->mc_list && dev->flags&IFF_MULTICAST) { do it; }
1023            --ANK
1024            */
1025         if (arp_mc_map(addr, buf, dev, 0) == 0)
1026                 dev_mc_add(dev, buf);
1027 }
1028
1029 /*
1030  *      Remove a filter from a device
1031  */
1032
1033 static void ip_mc_filter_del(struct in_device *in_dev, __be32 addr)
1034 {
1035         char buf[MAX_ADDR_LEN];
1036         struct net_device *dev = in_dev->dev;
1037
1038         if (arp_mc_map(addr, buf, dev, 0) == 0)
1039                 dev_mc_del(dev, buf);
1040 }
1041
1042 #ifdef CONFIG_IP_MULTICAST
1043 /*
1044  * deleted ip_mc_list manipulation
1045  */
1046 static void igmpv3_add_delrec(struct in_device *in_dev, struct ip_mc_list *im)
1047 {
1048         struct ip_mc_list *pmc;
1049
1050         /* this is an "ip_mc_list" for convenience; only the fields below
1051          * are actually used. In particular, the refcnt and users are not
1052          * used for management of the delete list. Using the same structure
1053          * for deleted items allows change reports to use common code with
1054          * non-deleted or query-response MCA's.
1055          */
1056         pmc = kzalloc(sizeof(*pmc), GFP_KERNEL);
1057         if (!pmc)
1058                 return;
1059         spin_lock_bh(&im->lock);
1060         pmc->interface = im->interface;
1061         in_dev_hold(in_dev);
1062         pmc->multiaddr = im->multiaddr;
1063         pmc->crcount = in_dev->mr_qrv ? in_dev->mr_qrv :
1064                 IGMP_Unsolicited_Report_Count;
1065         pmc->sfmode = im->sfmode;
1066         if (pmc->sfmode == MCAST_INCLUDE) {
1067                 struct ip_sf_list *psf;
1068
1069                 pmc->tomb = im->tomb;
1070                 pmc->sources = im->sources;
1071                 im->tomb = im->sources = NULL;
1072                 for (psf=pmc->sources; psf; psf=psf->sf_next)
1073                         psf->sf_crcount = pmc->crcount;
1074         }
1075         spin_unlock_bh(&im->lock);
1076
1077         spin_lock_bh(&in_dev->mc_tomb_lock);
1078         pmc->next = in_dev->mc_tomb;
1079         in_dev->mc_tomb = pmc;
1080         spin_unlock_bh(&in_dev->mc_tomb_lock);
1081 }
1082
1083 static void igmpv3_del_delrec(struct in_device *in_dev, __be32 multiaddr)
1084 {
1085         struct ip_mc_list *pmc, *pmc_prev;
1086         struct ip_sf_list *psf, *psf_next;
1087
1088         spin_lock_bh(&in_dev->mc_tomb_lock);
1089         pmc_prev = NULL;
1090         for (pmc=in_dev->mc_tomb; pmc; pmc=pmc->next) {
1091                 if (pmc->multiaddr == multiaddr)
1092                         break;
1093                 pmc_prev = pmc;
1094         }
1095         if (pmc) {
1096                 if (pmc_prev)
1097                         pmc_prev->next = pmc->next;
1098                 else
1099                         in_dev->mc_tomb = pmc->next;
1100         }
1101         spin_unlock_bh(&in_dev->mc_tomb_lock);
1102         if (pmc) {
1103                 for (psf=pmc->tomb; psf; psf=psf_next) {
1104                         psf_next = psf->sf_next;
1105                         kfree(psf);
1106                 }
1107                 in_dev_put(pmc->interface);
1108                 kfree(pmc);
1109         }
1110 }
1111
1112 static void igmpv3_clear_delrec(struct in_device *in_dev)
1113 {
1114         struct ip_mc_list *pmc, *nextpmc;
1115
1116         spin_lock_bh(&in_dev->mc_tomb_lock);
1117         pmc = in_dev->mc_tomb;
1118         in_dev->mc_tomb = NULL;
1119         spin_unlock_bh(&in_dev->mc_tomb_lock);
1120
1121         for (; pmc; pmc = nextpmc) {
1122                 nextpmc = pmc->next;
1123                 ip_mc_clear_src(pmc);
1124                 in_dev_put(pmc->interface);
1125                 kfree(pmc);
1126         }
1127         /* clear dead sources, too */
1128         rcu_read_lock();
1129         for_each_pmc_rcu(in_dev, pmc) {
1130                 struct ip_sf_list *psf, *psf_next;
1131
1132                 spin_lock_bh(&pmc->lock);
1133                 psf = pmc->tomb;
1134                 pmc->tomb = NULL;
1135                 spin_unlock_bh(&pmc->lock);
1136                 for (; psf; psf=psf_next) {
1137                         psf_next = psf->sf_next;
1138                         kfree(psf);
1139                 }
1140         }
1141         rcu_read_unlock();
1142 }
1143 #endif
1144
1145 static void igmp_group_dropped(struct ip_mc_list *im)
1146 {
1147         struct in_device *in_dev = im->interface;
1148 #ifdef CONFIG_IP_MULTICAST
1149         int reporter;
1150 #endif
1151
1152         if (im->loaded) {
1153                 im->loaded = 0;
1154                 ip_mc_filter_del(in_dev, im->multiaddr);
1155         }
1156
1157 #ifdef CONFIG_IP_MULTICAST
1158         if (im->multiaddr == IGMP_ALL_HOSTS)
1159                 return;
1160
1161         reporter = im->reporter;
1162         igmp_stop_timer(im);
1163
1164         if (!in_dev->dead) {
1165                 if (IGMP_V1_SEEN(in_dev))
1166                         goto done;
1167                 if (IGMP_V2_SEEN(in_dev)) {
1168                         if (reporter)
1169                                 igmp_send_report(in_dev, im, IGMP_HOST_LEAVE_MESSAGE);
1170                         goto done;
1171                 }
1172                 /* IGMPv3 */
1173                 igmpv3_add_delrec(in_dev, im);
1174
1175                 igmp_ifc_event(in_dev);
1176         }
1177 done:
1178 #endif
1179         ip_mc_clear_src(im);
1180 }
1181
1182 static void igmp_group_added(struct ip_mc_list *im)
1183 {
1184         struct in_device *in_dev = im->interface;
1185
1186         if (im->loaded == 0) {
1187                 im->loaded = 1;
1188                 ip_mc_filter_add(in_dev, im->multiaddr);
1189         }
1190
1191 #ifdef CONFIG_IP_MULTICAST
1192         if (im->multiaddr == IGMP_ALL_HOSTS)
1193                 return;
1194
1195         if (in_dev->dead)
1196                 return;
1197         if (IGMP_V1_SEEN(in_dev) || IGMP_V2_SEEN(in_dev)) {
1198                 spin_lock_bh(&im->lock);
1199                 igmp_start_timer(im, IGMP_Initial_Report_Delay);
1200                 spin_unlock_bh(&im->lock);
1201                 return;
1202         }
1203         /* else, v3 */
1204
1205         im->crcount = in_dev->mr_qrv ? in_dev->mr_qrv :
1206                 IGMP_Unsolicited_Report_Count;
1207         igmp_ifc_event(in_dev);
1208 #endif
1209 }
1210
1211
1212 /*
1213  *      Multicast list managers
1214  */
1215
1216
1217 /*
1218  *      A socket has joined a multicast group on device dev.
1219  */
1220
1221 void ip_mc_inc_group(struct in_device *in_dev, __be32 addr)
1222 {
1223         struct ip_mc_list *im;
1224
1225         ASSERT_RTNL();
1226
1227         for_each_pmc_rtnl(in_dev, im) {
1228                 if (im->multiaddr == addr) {
1229                         im->users++;
1230                         ip_mc_add_src(in_dev, &addr, MCAST_EXCLUDE, 0, NULL, 0);
1231                         goto out;
1232                 }
1233         }
1234
1235         im = kzalloc(sizeof(*im), GFP_KERNEL);
1236         if (!im)
1237                 goto out;
1238
1239         im->users = 1;
1240         im->interface = in_dev;
1241         in_dev_hold(in_dev);
1242         im->multiaddr = addr;
1243         /* initial mode is (EX, empty) */
1244         im->sfmode = MCAST_EXCLUDE;
1245         im->sfcount[MCAST_EXCLUDE] = 1;
1246         atomic_set(&im->refcnt, 1);
1247         spin_lock_init(&im->lock);
1248 #ifdef CONFIG_IP_MULTICAST
1249         setup_timer(&im->timer, &igmp_timer_expire, (unsigned long)im);
1250         im->unsolicit_count = IGMP_Unsolicited_Report_Count;
1251 #endif
1252
1253         im->next_rcu = in_dev->mc_list;
1254         in_dev->mc_count++;
1255         rcu_assign_pointer(in_dev->mc_list, im);
1256
1257 #ifdef CONFIG_IP_MULTICAST
1258         igmpv3_del_delrec(in_dev, im->multiaddr);
1259 #endif
1260         igmp_group_added(im);
1261         if (!in_dev->dead)
1262                 ip_rt_multicast_event(in_dev);
1263 out:
1264         return;
1265 }
1266 EXPORT_SYMBOL(ip_mc_inc_group);
1267
1268 /*
1269  *      Resend IGMP JOIN report; used for bonding.
1270  *      Called with rcu_read_lock()
1271  */
1272 void ip_mc_rejoin_groups(struct in_device *in_dev)
1273 {
1274 #ifdef CONFIG_IP_MULTICAST
1275         struct ip_mc_list *im;
1276         int type;
1277
1278         for_each_pmc_rcu(in_dev, im) {
1279                 if (im->multiaddr == IGMP_ALL_HOSTS)
1280                         continue;
1281
1282                 /* a failover is happening and switches
1283                  * must be notified immediately
1284                  */
1285                 if (IGMP_V1_SEEN(in_dev))
1286                         type = IGMP_HOST_MEMBERSHIP_REPORT;
1287                 else if (IGMP_V2_SEEN(in_dev))
1288                         type = IGMPV2_HOST_MEMBERSHIP_REPORT;
1289                 else
1290                         type = IGMPV3_HOST_MEMBERSHIP_REPORT;
1291                 igmp_send_report(in_dev, im, type);
1292         }
1293 #endif
1294 }
1295 EXPORT_SYMBOL(ip_mc_rejoin_groups);
1296
1297 /*
1298  *      A socket has left a multicast group on device dev
1299  */
1300
1301 void ip_mc_dec_group(struct in_device *in_dev, __be32 addr)
1302 {
1303         struct ip_mc_list *i;
1304         struct ip_mc_list __rcu **ip;
1305
1306         ASSERT_RTNL();
1307
1308         for (ip = &in_dev->mc_list;
1309              (i = rtnl_dereference(*ip)) != NULL;
1310              ip = &i->next_rcu) {
1311                 if (i->multiaddr == addr) {
1312                         if (--i->users == 0) {
1313                                 *ip = i->next_rcu;
1314                                 in_dev->mc_count--;
1315                                 igmp_group_dropped(i);
1316
1317                                 if (!in_dev->dead)
1318                                         ip_rt_multicast_event(in_dev);
1319
1320                                 ip_ma_put(i);
1321                                 return;
1322                         }
1323                         break;
1324                 }
1325         }
1326 }
1327 EXPORT_SYMBOL(ip_mc_dec_group);
1328
1329 /* Device changing type */
1330
1331 void ip_mc_unmap(struct in_device *in_dev)
1332 {
1333         struct ip_mc_list *pmc;
1334
1335         ASSERT_RTNL();
1336
1337         for_each_pmc_rtnl(in_dev, pmc)
1338                 igmp_group_dropped(pmc);
1339 }
1340
1341 void ip_mc_remap(struct in_device *in_dev)
1342 {
1343         struct ip_mc_list *pmc;
1344
1345         ASSERT_RTNL();
1346
1347         for_each_pmc_rtnl(in_dev, pmc)
1348                 igmp_group_added(pmc);
1349 }
1350
1351 /* Device going down */
1352
1353 void ip_mc_down(struct in_device *in_dev)
1354 {
1355         struct ip_mc_list *pmc;
1356
1357         ASSERT_RTNL();
1358
1359         for_each_pmc_rtnl(in_dev, pmc)
1360                 igmp_group_dropped(pmc);
1361
1362 #ifdef CONFIG_IP_MULTICAST
1363         in_dev->mr_ifc_count = 0;
1364         if (del_timer(&in_dev->mr_ifc_timer))
1365                 __in_dev_put(in_dev);
1366         in_dev->mr_gq_running = 0;
1367         if (del_timer(&in_dev->mr_gq_timer))
1368                 __in_dev_put(in_dev);
1369         igmpv3_clear_delrec(in_dev);
1370 #endif
1371
1372         ip_mc_dec_group(in_dev, IGMP_ALL_HOSTS);
1373 }
1374
1375 void ip_mc_init_dev(struct in_device *in_dev)
1376 {
1377         ASSERT_RTNL();
1378
1379         in_dev->mc_tomb = NULL;
1380 #ifdef CONFIG_IP_MULTICAST
1381         in_dev->mr_gq_running = 0;
1382         setup_timer(&in_dev->mr_gq_timer, igmp_gq_timer_expire,
1383                         (unsigned long)in_dev);
1384         in_dev->mr_ifc_count = 0;
1385         in_dev->mc_count     = 0;
1386         setup_timer(&in_dev->mr_ifc_timer, igmp_ifc_timer_expire,
1387                         (unsigned long)in_dev);
1388         in_dev->mr_qrv = IGMP_Unsolicited_Report_Count;
1389 #endif
1390
1391         spin_lock_init(&in_dev->mc_tomb_lock);
1392 }
1393
1394 /* Device going up */
1395
1396 void ip_mc_up(struct in_device *in_dev)
1397 {
1398         struct ip_mc_list *pmc;
1399
1400         ASSERT_RTNL();
1401
1402         ip_mc_inc_group(in_dev, IGMP_ALL_HOSTS);
1403
1404         for_each_pmc_rtnl(in_dev, pmc)
1405                 igmp_group_added(pmc);
1406 }
1407
1408 /*
1409  *      Device is about to be destroyed: clean up.
1410  */
1411
1412 void ip_mc_destroy_dev(struct in_device *in_dev)
1413 {
1414         struct ip_mc_list *i;
1415
1416         ASSERT_RTNL();
1417
1418         /* Deactivate timers */
1419         ip_mc_down(in_dev);
1420
1421         while ((i = rtnl_dereference(in_dev->mc_list)) != NULL) {
1422                 in_dev->mc_list = i->next_rcu;
1423                 in_dev->mc_count--;
1424
1425                 igmp_group_dropped(i);
1426                 ip_ma_put(i);
1427         }
1428 }
1429
1430 /* RTNL is locked */
1431 static struct in_device *ip_mc_find_dev(struct net *net, struct ip_mreqn *imr)
1432 {
1433         struct net_device *dev = NULL;
1434         struct in_device *idev = NULL;
1435
1436         if (imr->imr_ifindex) {
1437                 idev = inetdev_by_index(net, imr->imr_ifindex);
1438                 return idev;
1439         }
1440         if (imr->imr_address.s_addr) {
1441                 dev = __ip_dev_find(net, imr->imr_address.s_addr, false);
1442                 if (!dev)
1443                         return NULL;
1444         }
1445
1446         if (!dev) {
1447                 struct rtable *rt = ip_route_output(net,
1448                                                     imr->imr_multiaddr.s_addr,
1449                                                     0, 0, 0);
1450                 if (!IS_ERR(rt)) {
1451                         dev = rt->dst.dev;
1452                         ip_rt_put(rt);
1453                 }
1454         }
1455         if (dev) {
1456                 imr->imr_ifindex = dev->ifindex;
1457                 idev = __in_dev_get_rtnl(dev);
1458         }
1459         return idev;
1460 }
1461
1462 /*
1463  *      Join a socket to a group
1464  */
1465 int sysctl_igmp_max_memberships __read_mostly = IP_MAX_MEMBERSHIPS;
1466 int sysctl_igmp_max_msf __read_mostly = IP_MAX_MSF;
1467
1468
1469 static int ip_mc_del1_src(struct ip_mc_list *pmc, int sfmode,
1470         __be32 *psfsrc)
1471 {
1472         struct ip_sf_list *psf, *psf_prev;
1473         int rv = 0;
1474
1475         psf_prev = NULL;
1476         for (psf=pmc->sources; psf; psf=psf->sf_next) {
1477                 if (psf->sf_inaddr == *psfsrc)
1478                         break;
1479                 psf_prev = psf;
1480         }
1481         if (!psf || psf->sf_count[sfmode] == 0) {
1482                 /* source filter not found, or count wrong =>  bug */
1483                 return -ESRCH;
1484         }
1485         psf->sf_count[sfmode]--;
1486         if (psf->sf_count[sfmode] == 0) {
1487                 ip_rt_multicast_event(pmc->interface);
1488         }
1489         if (!psf->sf_count[MCAST_INCLUDE] && !psf->sf_count[MCAST_EXCLUDE]) {
1490 #ifdef CONFIG_IP_MULTICAST
1491                 struct in_device *in_dev = pmc->interface;
1492 #endif
1493
1494                 /* no more filters for this source */
1495                 if (psf_prev)
1496                         psf_prev->sf_next = psf->sf_next;
1497                 else
1498                         pmc->sources = psf->sf_next;
1499 #ifdef CONFIG_IP_MULTICAST
1500                 if (psf->sf_oldin &&
1501                     !IGMP_V1_SEEN(in_dev) && !IGMP_V2_SEEN(in_dev)) {
1502                         psf->sf_crcount = in_dev->mr_qrv ? in_dev->mr_qrv :
1503                                 IGMP_Unsolicited_Report_Count;
1504                         psf->sf_next = pmc->tomb;
1505                         pmc->tomb = psf;
1506                         rv = 1;
1507                 } else
1508 #endif
1509                         kfree(psf);
1510         }
1511         return rv;
1512 }
1513
1514 #ifndef CONFIG_IP_MULTICAST
1515 #define igmp_ifc_event(x)       do { } while (0)
1516 #endif
1517
1518 static int ip_mc_del_src(struct in_device *in_dev, __be32 *pmca, int sfmode,
1519                          int sfcount, __be32 *psfsrc, int delta)
1520 {
1521         struct ip_mc_list *pmc;
1522         int     changerec = 0;
1523         int     i, err;
1524
1525         if (!in_dev)
1526                 return -ENODEV;
1527         rcu_read_lock();
1528         for_each_pmc_rcu(in_dev, pmc) {
1529                 if (*pmca == pmc->multiaddr)
1530                         break;
1531         }
1532         if (!pmc) {
1533                 /* MCA not found?? bug */
1534                 rcu_read_unlock();
1535                 return -ESRCH;
1536         }
1537         spin_lock_bh(&pmc->lock);
1538         rcu_read_unlock();
1539 #ifdef CONFIG_IP_MULTICAST
1540         sf_markstate(pmc);
1541 #endif
1542         if (!delta) {
1543                 err = -EINVAL;
1544                 if (!pmc->sfcount[sfmode])
1545                         goto out_unlock;
1546                 pmc->sfcount[sfmode]--;
1547         }
1548         err = 0;
1549         for (i=0; i<sfcount; i++) {
1550                 int rv = ip_mc_del1_src(pmc, sfmode, &psfsrc[i]);
1551
1552                 changerec |= rv > 0;
1553                 if (!err && rv < 0)
1554                         err = rv;
1555         }
1556         if (pmc->sfmode == MCAST_EXCLUDE &&
1557             pmc->sfcount[MCAST_EXCLUDE] == 0 &&
1558             pmc->sfcount[MCAST_INCLUDE]) {
1559 #ifdef CONFIG_IP_MULTICAST
1560                 struct ip_sf_list *psf;
1561 #endif
1562
1563                 /* filter mode change */
1564                 pmc->sfmode = MCAST_INCLUDE;
1565 #ifdef CONFIG_IP_MULTICAST
1566                 pmc->crcount = in_dev->mr_qrv ? in_dev->mr_qrv :
1567                         IGMP_Unsolicited_Report_Count;
1568                 in_dev->mr_ifc_count = pmc->crcount;
1569                 for (psf=pmc->sources; psf; psf = psf->sf_next)
1570                         psf->sf_crcount = 0;
1571                 igmp_ifc_event(pmc->interface);
1572         } else if (sf_setstate(pmc) || changerec) {
1573                 igmp_ifc_event(pmc->interface);
1574 #endif
1575         }
1576 out_unlock:
1577         spin_unlock_bh(&pmc->lock);
1578         return err;
1579 }
1580
1581 /*
1582  * Add multicast single-source filter to the interface list
1583  */
1584 static int ip_mc_add1_src(struct ip_mc_list *pmc, int sfmode,
1585         __be32 *psfsrc, int delta)
1586 {
1587         struct ip_sf_list *psf, *psf_prev;
1588
1589         psf_prev = NULL;
1590         for (psf=pmc->sources; psf; psf=psf->sf_next) {
1591                 if (psf->sf_inaddr == *psfsrc)
1592                         break;
1593                 psf_prev = psf;
1594         }
1595         if (!psf) {
1596                 psf = kzalloc(sizeof(*psf), GFP_ATOMIC);
1597                 if (!psf)
1598                         return -ENOBUFS;
1599                 psf->sf_inaddr = *psfsrc;
1600                 if (psf_prev) {
1601                         psf_prev->sf_next = psf;
1602                 } else
1603                         pmc->sources = psf;
1604         }
1605         psf->sf_count[sfmode]++;
1606         if (psf->sf_count[sfmode] == 1) {
1607                 ip_rt_multicast_event(pmc->interface);
1608         }
1609         return 0;
1610 }
1611
1612 #ifdef CONFIG_IP_MULTICAST
1613 static void sf_markstate(struct ip_mc_list *pmc)
1614 {
1615         struct ip_sf_list *psf;
1616         int mca_xcount = pmc->sfcount[MCAST_EXCLUDE];
1617
1618         for (psf=pmc->sources; psf; psf=psf->sf_next)
1619                 if (pmc->sfcount[MCAST_EXCLUDE]) {
1620                         psf->sf_oldin = mca_xcount ==
1621                                 psf->sf_count[MCAST_EXCLUDE] &&
1622                                 !psf->sf_count[MCAST_INCLUDE];
1623                 } else
1624                         psf->sf_oldin = psf->sf_count[MCAST_INCLUDE] != 0;
1625 }
1626
1627 static int sf_setstate(struct ip_mc_list *pmc)
1628 {
1629         struct ip_sf_list *psf, *dpsf;
1630         int mca_xcount = pmc->sfcount[MCAST_EXCLUDE];
1631         int qrv = pmc->interface->mr_qrv;
1632         int new_in, rv;
1633
1634         rv = 0;
1635         for (psf=pmc->sources; psf; psf=psf->sf_next) {
1636                 if (pmc->sfcount[MCAST_EXCLUDE]) {
1637                         new_in = mca_xcount == psf->sf_count[MCAST_EXCLUDE] &&
1638                                 !psf->sf_count[MCAST_INCLUDE];
1639                 } else
1640                         new_in = psf->sf_count[MCAST_INCLUDE] != 0;
1641                 if (new_in) {
1642                         if (!psf->sf_oldin) {
1643                                 struct ip_sf_list *prev = NULL;
1644
1645                                 for (dpsf=pmc->tomb; dpsf; dpsf=dpsf->sf_next) {
1646                                         if (dpsf->sf_inaddr == psf->sf_inaddr)
1647                                                 break;
1648                                         prev = dpsf;
1649                                 }
1650                                 if (dpsf) {
1651                                         if (prev)
1652                                                 prev->sf_next = dpsf->sf_next;
1653                                         else
1654                                                 pmc->tomb = dpsf->sf_next;
1655                                         kfree(dpsf);
1656                                 }
1657                                 psf->sf_crcount = qrv;
1658                                 rv++;
1659                         }
1660                 } else if (psf->sf_oldin) {
1661
1662                         psf->sf_crcount = 0;
1663                         /*
1664                          * add or update "delete" records if an active filter
1665                          * is now inactive
1666                          */
1667                         for (dpsf=pmc->tomb; dpsf; dpsf=dpsf->sf_next)
1668                                 if (dpsf->sf_inaddr == psf->sf_inaddr)
1669                                         break;
1670                         if (!dpsf) {
1671                                 dpsf = kmalloc(sizeof(*dpsf), GFP_ATOMIC);
1672                                 if (!dpsf)
1673                                         continue;
1674                                 *dpsf = *psf;
1675                                 /* pmc->lock held by callers */
1676                                 dpsf->sf_next = pmc->tomb;
1677                                 pmc->tomb = dpsf;
1678                         }
1679                         dpsf->sf_crcount = qrv;
1680                         rv++;
1681                 }
1682         }
1683         return rv;
1684 }
1685 #endif
1686
1687 /*
1688  * Add multicast source filter list to the interface list
1689  */
1690 static int ip_mc_add_src(struct in_device *in_dev, __be32 *pmca, int sfmode,
1691                          int sfcount, __be32 *psfsrc, int delta)
1692 {
1693         struct ip_mc_list *pmc;
1694         int     isexclude;
1695         int     i, err;
1696
1697         if (!in_dev)
1698                 return -ENODEV;
1699         rcu_read_lock();
1700         for_each_pmc_rcu(in_dev, pmc) {
1701                 if (*pmca == pmc->multiaddr)
1702                         break;
1703         }
1704         if (!pmc) {
1705                 /* MCA not found?? bug */
1706                 rcu_read_unlock();
1707                 return -ESRCH;
1708         }
1709         spin_lock_bh(&pmc->lock);
1710         rcu_read_unlock();
1711
1712 #ifdef CONFIG_IP_MULTICAST
1713         sf_markstate(pmc);
1714 #endif
1715         isexclude = pmc->sfmode == MCAST_EXCLUDE;
1716         if (!delta)
1717                 pmc->sfcount[sfmode]++;
1718         err = 0;
1719         for (i=0; i<sfcount; i++) {
1720                 err = ip_mc_add1_src(pmc, sfmode, &psfsrc[i], delta);
1721                 if (err)
1722                         break;
1723         }
1724         if (err) {
1725                 int j;
1726
1727                 pmc->sfcount[sfmode]--;
1728                 for (j=0; j<i; j++)
1729                         (void) ip_mc_del1_src(pmc, sfmode, &psfsrc[i]);
1730         } else if (isexclude != (pmc->sfcount[MCAST_EXCLUDE] != 0)) {
1731 #ifdef CONFIG_IP_MULTICAST
1732                 struct ip_sf_list *psf;
1733                 in_dev = pmc->interface;
1734 #endif
1735
1736                 /* filter mode change */
1737                 if (pmc->sfcount[MCAST_EXCLUDE])
1738                         pmc->sfmode = MCAST_EXCLUDE;
1739                 else if (pmc->sfcount[MCAST_INCLUDE])
1740                         pmc->sfmode = MCAST_INCLUDE;
1741 #ifdef CONFIG_IP_MULTICAST
1742                 /* else no filters; keep old mode for reports */
1743
1744                 pmc->crcount = in_dev->mr_qrv ? in_dev->mr_qrv :
1745                         IGMP_Unsolicited_Report_Count;
1746                 in_dev->mr_ifc_count = pmc->crcount;
1747                 for (psf=pmc->sources; psf; psf = psf->sf_next)
1748                         psf->sf_crcount = 0;
1749                 igmp_ifc_event(in_dev);
1750         } else if (sf_setstate(pmc)) {
1751                 igmp_ifc_event(in_dev);
1752 #endif
1753         }
1754         spin_unlock_bh(&pmc->lock);
1755         return err;
1756 }
1757
1758 static void ip_mc_clear_src(struct ip_mc_list *pmc)
1759 {
1760         struct ip_sf_list *psf, *nextpsf;
1761
1762         for (psf=pmc->tomb; psf; psf=nextpsf) {
1763                 nextpsf = psf->sf_next;
1764                 kfree(psf);
1765         }
1766         pmc->tomb = NULL;
1767         for (psf=pmc->sources; psf; psf=nextpsf) {
1768                 nextpsf = psf->sf_next;
1769                 kfree(psf);
1770         }
1771         pmc->sources = NULL;
1772         pmc->sfmode = MCAST_EXCLUDE;
1773         pmc->sfcount[MCAST_INCLUDE] = 0;
1774         pmc->sfcount[MCAST_EXCLUDE] = 1;
1775 }
1776
1777
1778 /*
1779  * Join a multicast group
1780  */
1781 int ip_mc_join_group(struct sock *sk , struct ip_mreqn *imr)
1782 {
1783         int err;
1784         __be32 addr = imr->imr_multiaddr.s_addr;
1785         struct ip_mc_socklist *iml = NULL, *i;
1786         struct in_device *in_dev;
1787         struct inet_sock *inet = inet_sk(sk);
1788         struct net *net = sock_net(sk);
1789         int ifindex;
1790         int count = 0;
1791
1792         if (!ipv4_is_multicast(addr))
1793                 return -EINVAL;
1794
1795         rtnl_lock();
1796
1797         in_dev = ip_mc_find_dev(net, imr);
1798
1799         if (!in_dev) {
1800                 iml = NULL;
1801                 err = -ENODEV;
1802                 goto done;
1803         }
1804
1805         err = -EADDRINUSE;
1806         ifindex = imr->imr_ifindex;
1807         for_each_pmc_rtnl(inet, i) {
1808                 if (i->multi.imr_multiaddr.s_addr == addr &&
1809                     i->multi.imr_ifindex == ifindex)
1810                         goto done;
1811                 count++;
1812         }
1813         err = -ENOBUFS;
1814         if (count >= sysctl_igmp_max_memberships)
1815                 goto done;
1816         iml = sock_kmalloc(sk, sizeof(*iml), GFP_KERNEL);
1817         if (iml == NULL)
1818                 goto done;
1819
1820         memcpy(&iml->multi, imr, sizeof(*imr));
1821         iml->next_rcu = inet->mc_list;
1822         iml->sflist = NULL;
1823         iml->sfmode = MCAST_EXCLUDE;
1824         rcu_assign_pointer(inet->mc_list, iml);
1825         ip_mc_inc_group(in_dev, addr);
1826         err = 0;
1827 done:
1828         rtnl_unlock();
1829         return err;
1830 }
1831 EXPORT_SYMBOL(ip_mc_join_group);
1832
1833 static void ip_sf_socklist_reclaim(struct rcu_head *rp)
1834 {
1835         kfree(container_of(rp, struct ip_sf_socklist, rcu));
1836         /* sk_omem_alloc should have been decreased by the caller*/
1837 }
1838
1839 static int ip_mc_leave_src(struct sock *sk, struct ip_mc_socklist *iml,
1840                            struct in_device *in_dev)
1841 {
1842         struct ip_sf_socklist *psf = rtnl_dereference(iml->sflist);
1843         int err;
1844
1845         if (psf == NULL) {
1846                 /* any-source empty exclude case */
1847                 return ip_mc_del_src(in_dev, &iml->multi.imr_multiaddr.s_addr,
1848                         iml->sfmode, 0, NULL, 0);
1849         }
1850         err = ip_mc_del_src(in_dev, &iml->multi.imr_multiaddr.s_addr,
1851                         iml->sfmode, psf->sl_count, psf->sl_addr, 0);
1852         rcu_assign_pointer(iml->sflist, NULL);
1853         /* decrease mem now to avoid the memleak warning */
1854         atomic_sub(IP_SFLSIZE(psf->sl_max), &sk->sk_omem_alloc);
1855         call_rcu(&psf->rcu, ip_sf_socklist_reclaim);
1856         return err;
1857 }
1858
1859
1860 static void ip_mc_socklist_reclaim(struct rcu_head *rp)
1861 {
1862         kfree(container_of(rp, struct ip_mc_socklist, rcu));
1863         /* sk_omem_alloc should have been decreased by the caller*/
1864 }
1865
1866
1867 /*
1868  *      Ask a socket to leave a group.
1869  */
1870
1871 int ip_mc_leave_group(struct sock *sk, struct ip_mreqn *imr)
1872 {
1873         struct inet_sock *inet = inet_sk(sk);
1874         struct ip_mc_socklist *iml;
1875         struct ip_mc_socklist __rcu **imlp;
1876         struct in_device *in_dev;
1877         struct net *net = sock_net(sk);
1878         __be32 group = imr->imr_multiaddr.s_addr;
1879         u32 ifindex;
1880         int ret = -EADDRNOTAVAIL;
1881
1882         rtnl_lock();
1883         in_dev = ip_mc_find_dev(net, imr);
1884         ifindex = imr->imr_ifindex;
1885         for (imlp = &inet->mc_list;
1886              (iml = rtnl_dereference(*imlp)) != NULL;
1887              imlp = &iml->next_rcu) {
1888                 if (iml->multi.imr_multiaddr.s_addr != group)
1889                         continue;
1890                 if (ifindex) {
1891                         if (iml->multi.imr_ifindex != ifindex)
1892                                 continue;
1893                 } else if (imr->imr_address.s_addr && imr->imr_address.s_addr !=
1894                                 iml->multi.imr_address.s_addr)
1895                         continue;
1896
1897                 (void) ip_mc_leave_src(sk, iml, in_dev);
1898
1899                 *imlp = iml->next_rcu;
1900
1901                 if (in_dev)
1902                         ip_mc_dec_group(in_dev, group);
1903                 rtnl_unlock();
1904                 /* decrease mem now to avoid the memleak warning */
1905                 atomic_sub(sizeof(*iml), &sk->sk_omem_alloc);
1906                 call_rcu(&iml->rcu, ip_mc_socklist_reclaim);
1907                 return 0;
1908         }
1909         if (!in_dev)
1910                 ret = -ENODEV;
1911         rtnl_unlock();
1912         return ret;
1913 }
1914
1915 int ip_mc_source(int add, int omode, struct sock *sk, struct
1916         ip_mreq_source *mreqs, int ifindex)
1917 {
1918         int err;
1919         struct ip_mreqn imr;
1920         __be32 addr = mreqs->imr_multiaddr;
1921         struct ip_mc_socklist *pmc;
1922         struct in_device *in_dev = NULL;
1923         struct inet_sock *inet = inet_sk(sk);
1924         struct ip_sf_socklist *psl;
1925         struct net *net = sock_net(sk);
1926         int leavegroup = 0;
1927         int i, j, rv;
1928
1929         if (!ipv4_is_multicast(addr))
1930                 return -EINVAL;
1931
1932         rtnl_lock();
1933
1934         imr.imr_multiaddr.s_addr = mreqs->imr_multiaddr;
1935         imr.imr_address.s_addr = mreqs->imr_interface;
1936         imr.imr_ifindex = ifindex;
1937         in_dev = ip_mc_find_dev(net, &imr);
1938
1939         if (!in_dev) {
1940                 err = -ENODEV;
1941                 goto done;
1942         }
1943         err = -EADDRNOTAVAIL;
1944
1945         for_each_pmc_rtnl(inet, pmc) {
1946                 if ((pmc->multi.imr_multiaddr.s_addr ==
1947                      imr.imr_multiaddr.s_addr) &&
1948                     (pmc->multi.imr_ifindex == imr.imr_ifindex))
1949                         break;
1950         }
1951         if (!pmc) {             /* must have a prior join */
1952                 err = -EINVAL;
1953                 goto done;
1954         }
1955         /* if a source filter was set, must be the same mode as before */
1956         if (pmc->sflist) {
1957                 if (pmc->sfmode != omode) {
1958                         err = -EINVAL;
1959                         goto done;
1960                 }
1961         } else if (pmc->sfmode != omode) {
1962                 /* allow mode switches for empty-set filters */
1963                 ip_mc_add_src(in_dev, &mreqs->imr_multiaddr, omode, 0, NULL, 0);
1964                 ip_mc_del_src(in_dev, &mreqs->imr_multiaddr, pmc->sfmode, 0,
1965                         NULL, 0);
1966                 pmc->sfmode = omode;
1967         }
1968
1969         psl = rtnl_dereference(pmc->sflist);
1970         if (!add) {
1971                 if (!psl)
1972                         goto done;      /* err = -EADDRNOTAVAIL */
1973                 rv = !0;
1974                 for (i=0; i<psl->sl_count; i++) {
1975                         rv = memcmp(&psl->sl_addr[i], &mreqs->imr_sourceaddr,
1976                                 sizeof(__be32));
1977                         if (rv == 0)
1978                                 break;
1979                 }
1980                 if (rv)         /* source not found */
1981                         goto done;      /* err = -EADDRNOTAVAIL */
1982
1983                 /* special case - (INCLUDE, empty) == LEAVE_GROUP */
1984                 if (psl->sl_count == 1 && omode == MCAST_INCLUDE) {
1985                         leavegroup = 1;
1986                         goto done;
1987                 }
1988
1989                 /* update the interface filter */
1990                 ip_mc_del_src(in_dev, &mreqs->imr_multiaddr, omode, 1,
1991                         &mreqs->imr_sourceaddr, 1);
1992
1993                 for (j=i+1; j<psl->sl_count; j++)
1994                         psl->sl_addr[j-1] = psl->sl_addr[j];
1995                 psl->sl_count--;
1996                 err = 0;
1997                 goto done;
1998         }
1999         /* else, add a new source to the filter */
2000
2001         if (psl && psl->sl_count >= sysctl_igmp_max_msf) {
2002                 err = -ENOBUFS;
2003                 goto done;
2004         }
2005         if (!psl || psl->sl_count == psl->sl_max) {
2006                 struct ip_sf_socklist *newpsl;
2007                 int count = IP_SFBLOCK;
2008
2009                 if (psl)
2010                         count += psl->sl_max;
2011                 newpsl = sock_kmalloc(sk, IP_SFLSIZE(count), GFP_KERNEL);
2012                 if (!newpsl) {
2013                         err = -ENOBUFS;
2014                         goto done;
2015                 }
2016                 newpsl->sl_max = count;
2017                 newpsl->sl_count = count - IP_SFBLOCK;
2018                 if (psl) {
2019                         for (i=0; i<psl->sl_count; i++)
2020                                 newpsl->sl_addr[i] = psl->sl_addr[i];
2021                         /* decrease mem now to avoid the memleak warning */
2022                         atomic_sub(IP_SFLSIZE(psl->sl_max), &sk->sk_omem_alloc);
2023                         call_rcu(&psl->rcu, ip_sf_socklist_reclaim);
2024                 }
2025                 rcu_assign_pointer(pmc->sflist, newpsl);
2026                 psl = newpsl;
2027         }
2028         rv = 1; /* > 0 for insert logic below if sl_count is 0 */
2029         for (i=0; i<psl->sl_count; i++) {
2030                 rv = memcmp(&psl->sl_addr[i], &mreqs->imr_sourceaddr,
2031                         sizeof(__be32));
2032                 if (rv == 0)
2033                         break;
2034         }
2035         if (rv == 0)            /* address already there is an error */
2036                 goto done;
2037         for (j=psl->sl_count-1; j>=i; j--)
2038                 psl->sl_addr[j+1] = psl->sl_addr[j];
2039         psl->sl_addr[i] = mreqs->imr_sourceaddr;
2040         psl->sl_count++;
2041         err = 0;
2042         /* update the interface list */
2043         ip_mc_add_src(in_dev, &mreqs->imr_multiaddr, omode, 1,
2044                 &mreqs->imr_sourceaddr, 1);
2045 done:
2046         rtnl_unlock();
2047         if (leavegroup)
2048                 return ip_mc_leave_group(sk, &imr);
2049         return err;
2050 }
2051
2052 int ip_mc_msfilter(struct sock *sk, struct ip_msfilter *msf, int ifindex)
2053 {
2054         int err = 0;
2055         struct ip_mreqn imr;
2056         __be32 addr = msf->imsf_multiaddr;
2057         struct ip_mc_socklist *pmc;
2058         struct in_device *in_dev;
2059         struct inet_sock *inet = inet_sk(sk);
2060         struct ip_sf_socklist *newpsl, *psl;
2061         struct net *net = sock_net(sk);
2062         int leavegroup = 0;
2063
2064         if (!ipv4_is_multicast(addr))
2065                 return -EINVAL;
2066         if (msf->imsf_fmode != MCAST_INCLUDE &&
2067             msf->imsf_fmode != MCAST_EXCLUDE)
2068                 return -EINVAL;
2069
2070         rtnl_lock();
2071
2072         imr.imr_multiaddr.s_addr = msf->imsf_multiaddr;
2073         imr.imr_address.s_addr = msf->imsf_interface;
2074         imr.imr_ifindex = ifindex;
2075         in_dev = ip_mc_find_dev(net, &imr);
2076
2077         if (!in_dev) {
2078                 err = -ENODEV;
2079                 goto done;
2080         }
2081
2082         /* special case - (INCLUDE, empty) == LEAVE_GROUP */
2083         if (msf->imsf_fmode == MCAST_INCLUDE && msf->imsf_numsrc == 0) {
2084                 leavegroup = 1;
2085                 goto done;
2086         }
2087
2088         for_each_pmc_rtnl(inet, pmc) {
2089                 if (pmc->multi.imr_multiaddr.s_addr == msf->imsf_multiaddr &&
2090                     pmc->multi.imr_ifindex == imr.imr_ifindex)
2091                         break;
2092         }
2093         if (!pmc) {             /* must have a prior join */
2094                 err = -EINVAL;
2095                 goto done;
2096         }
2097         if (msf->imsf_numsrc) {
2098                 newpsl = sock_kmalloc(sk, IP_SFLSIZE(msf->imsf_numsrc),
2099                                                            GFP_KERNEL);
2100                 if (!newpsl) {
2101                         err = -ENOBUFS;
2102                         goto done;
2103                 }
2104                 newpsl->sl_max = newpsl->sl_count = msf->imsf_numsrc;
2105                 memcpy(newpsl->sl_addr, msf->imsf_slist,
2106                         msf->imsf_numsrc * sizeof(msf->imsf_slist[0]));
2107                 err = ip_mc_add_src(in_dev, &msf->imsf_multiaddr,
2108                         msf->imsf_fmode, newpsl->sl_count, newpsl->sl_addr, 0);
2109                 if (err) {
2110                         sock_kfree_s(sk, newpsl, IP_SFLSIZE(newpsl->sl_max));
2111                         goto done;
2112                 }
2113         } else {
2114                 newpsl = NULL;
2115                 (void) ip_mc_add_src(in_dev, &msf->imsf_multiaddr,
2116                                      msf->imsf_fmode, 0, NULL, 0);
2117         }
2118         psl = rtnl_dereference(pmc->sflist);
2119         if (psl) {
2120                 (void) ip_mc_del_src(in_dev, &msf->imsf_multiaddr, pmc->sfmode,
2121                         psl->sl_count, psl->sl_addr, 0);
2122                 /* decrease mem now to avoid the memleak warning */
2123                 atomic_sub(IP_SFLSIZE(psl->sl_max), &sk->sk_omem_alloc);
2124                 call_rcu(&psl->rcu, ip_sf_socklist_reclaim);
2125         } else
2126                 (void) ip_mc_del_src(in_dev, &msf->imsf_multiaddr, pmc->sfmode,
2127                         0, NULL, 0);
2128         rcu_assign_pointer(pmc->sflist, newpsl);
2129         pmc->sfmode = msf->imsf_fmode;
2130         err = 0;
2131 done:
2132         rtnl_unlock();
2133         if (leavegroup)
2134                 err = ip_mc_leave_group(sk, &imr);
2135         return err;
2136 }
2137
2138 int ip_mc_msfget(struct sock *sk, struct ip_msfilter *msf,
2139         struct ip_msfilter __user *optval, int __user *optlen)
2140 {
2141         int err, len, count, copycount;
2142         struct ip_mreqn imr;
2143         __be32 addr = msf->imsf_multiaddr;
2144         struct ip_mc_socklist *pmc;
2145         struct in_device *in_dev;
2146         struct inet_sock *inet = inet_sk(sk);
2147         struct ip_sf_socklist *psl;
2148         struct net *net = sock_net(sk);
2149
2150         if (!ipv4_is_multicast(addr))
2151                 return -EINVAL;
2152
2153         rtnl_lock();
2154
2155         imr.imr_multiaddr.s_addr = msf->imsf_multiaddr;
2156         imr.imr_address.s_addr = msf->imsf_interface;
2157         imr.imr_ifindex = 0;
2158         in_dev = ip_mc_find_dev(net, &imr);
2159
2160         if (!in_dev) {
2161                 err = -ENODEV;
2162                 goto done;
2163         }
2164         err = -EADDRNOTAVAIL;
2165
2166         for_each_pmc_rtnl(inet, pmc) {
2167                 if (pmc->multi.imr_multiaddr.s_addr == msf->imsf_multiaddr &&
2168                     pmc->multi.imr_ifindex == imr.imr_ifindex)
2169                         break;
2170         }
2171         if (!pmc)               /* must have a prior join */
2172                 goto done;
2173         msf->imsf_fmode = pmc->sfmode;
2174         psl = rtnl_dereference(pmc->sflist);
2175         rtnl_unlock();
2176         if (!psl) {
2177                 len = 0;
2178                 count = 0;
2179         } else {
2180                 count = psl->sl_count;
2181         }
2182         copycount = count < msf->imsf_numsrc ? count : msf->imsf_numsrc;
2183         len = copycount * sizeof(psl->sl_addr[0]);
2184         msf->imsf_numsrc = count;
2185         if (put_user(IP_MSFILTER_SIZE(copycount), optlen) ||
2186             copy_to_user(optval, msf, IP_MSFILTER_SIZE(0))) {
2187                 return -EFAULT;
2188         }
2189         if (len &&
2190             copy_to_user(&optval->imsf_slist[0], psl->sl_addr, len))
2191                 return -EFAULT;
2192         return 0;
2193 done:
2194         rtnl_unlock();
2195         return err;
2196 }
2197
2198 int ip_mc_gsfget(struct sock *sk, struct group_filter *gsf,
2199         struct group_filter __user *optval, int __user *optlen)
2200 {
2201         int err, i, count, copycount;
2202         struct sockaddr_in *psin;
2203         __be32 addr;
2204         struct ip_mc_socklist *pmc;
2205         struct inet_sock *inet = inet_sk(sk);
2206         struct ip_sf_socklist *psl;
2207
2208         psin = (struct sockaddr_in *)&gsf->gf_group;
2209         if (psin->sin_family != AF_INET)
2210                 return -EINVAL;
2211         addr = psin->sin_addr.s_addr;
2212         if (!ipv4_is_multicast(addr))
2213                 return -EINVAL;
2214
2215         rtnl_lock();
2216
2217         err = -EADDRNOTAVAIL;
2218
2219         for_each_pmc_rtnl(inet, pmc) {
2220                 if (pmc->multi.imr_multiaddr.s_addr == addr &&
2221                     pmc->multi.imr_ifindex == gsf->gf_interface)
2222                         break;
2223         }
2224         if (!pmc)               /* must have a prior join */
2225                 goto done;
2226         gsf->gf_fmode = pmc->sfmode;
2227         psl = rtnl_dereference(pmc->sflist);
2228         rtnl_unlock();
2229         count = psl ? psl->sl_count : 0;
2230         copycount = count < gsf->gf_numsrc ? count : gsf->gf_numsrc;
2231         gsf->gf_numsrc = count;
2232         if (put_user(GROUP_FILTER_SIZE(copycount), optlen) ||
2233             copy_to_user(optval, gsf, GROUP_FILTER_SIZE(0))) {
2234                 return -EFAULT;
2235         }
2236         for (i=0; i<copycount; i++) {
2237                 struct sockaddr_storage ss;
2238
2239                 psin = (struct sockaddr_in *)&ss;
2240                 memset(&ss, 0, sizeof(ss));
2241                 psin->sin_family = AF_INET;
2242                 psin->sin_addr.s_addr = psl->sl_addr[i];
2243                 if (copy_to_user(&optval->gf_slist[i], &ss, sizeof(ss)))
2244                         return -EFAULT;
2245         }
2246         return 0;
2247 done:
2248         rtnl_unlock();
2249         return err;
2250 }
2251
2252 /*
2253  * check if a multicast source filter allows delivery for a given <src,dst,intf>
2254  */
2255 int ip_mc_sf_allow(struct sock *sk, __be32 loc_addr, __be32 rmt_addr, int dif)
2256 {
2257         struct inet_sock *inet = inet_sk(sk);
2258         struct ip_mc_socklist *pmc;
2259         struct ip_sf_socklist *psl;
2260         int i;
2261         int ret;
2262
2263         ret = 1;
2264         if (!ipv4_is_multicast(loc_addr))
2265                 goto out;
2266
2267         rcu_read_lock();
2268         for_each_pmc_rcu(inet, pmc) {
2269                 if (pmc->multi.imr_multiaddr.s_addr == loc_addr &&
2270                     pmc->multi.imr_ifindex == dif)
2271                         break;
2272         }
2273         ret = inet->mc_all;
2274         if (!pmc)
2275                 goto unlock;
2276         psl = rcu_dereference(pmc->sflist);
2277         ret = (pmc->sfmode == MCAST_EXCLUDE);
2278         if (!psl)
2279                 goto unlock;
2280
2281         for (i=0; i<psl->sl_count; i++) {
2282                 if (psl->sl_addr[i] == rmt_addr)
2283                         break;
2284         }
2285         ret = 0;
2286         if (pmc->sfmode == MCAST_INCLUDE && i >= psl->sl_count)
2287                 goto unlock;
2288         if (pmc->sfmode == MCAST_EXCLUDE && i < psl->sl_count)
2289                 goto unlock;
2290         ret = 1;
2291 unlock:
2292         rcu_read_unlock();
2293 out:
2294         return ret;
2295 }
2296
2297 /*
2298  *      A socket is closing.
2299  */
2300
2301 void ip_mc_drop_socket(struct sock *sk)
2302 {
2303         struct inet_sock *inet = inet_sk(sk);
2304         struct ip_mc_socklist *iml;
2305         struct net *net = sock_net(sk);
2306
2307         if (inet->mc_list == NULL)
2308                 return;
2309
2310         rtnl_lock();
2311         while ((iml = rtnl_dereference(inet->mc_list)) != NULL) {
2312                 struct in_device *in_dev;
2313
2314                 inet->mc_list = iml->next_rcu;
2315                 in_dev = inetdev_by_index(net, iml->multi.imr_ifindex);
2316                 (void) ip_mc_leave_src(sk, iml, in_dev);
2317                 if (in_dev != NULL)
2318                         ip_mc_dec_group(in_dev, iml->multi.imr_multiaddr.s_addr);
2319                 /* decrease mem now to avoid the memleak warning */
2320                 atomic_sub(sizeof(*iml), &sk->sk_omem_alloc);
2321                 call_rcu(&iml->rcu, ip_mc_socklist_reclaim);
2322         }
2323         rtnl_unlock();
2324 }
2325
2326 /* called with rcu_read_lock() */
2327 int ip_check_mc_rcu(struct in_device *in_dev, __be32 mc_addr, __be32 src_addr, u16 proto)
2328 {
2329         struct ip_mc_list *im;
2330         struct ip_sf_list *psf;
2331         int rv = 0;
2332
2333         for_each_pmc_rcu(in_dev, im) {
2334                 if (im->multiaddr == mc_addr)
2335                         break;
2336         }
2337         if (im && proto == IPPROTO_IGMP) {
2338                 rv = 1;
2339         } else if (im) {
2340                 if (src_addr) {
2341                         for (psf=im->sources; psf; psf=psf->sf_next) {
2342                                 if (psf->sf_inaddr == src_addr)
2343                                         break;
2344                         }
2345                         if (psf)
2346                                 rv = psf->sf_count[MCAST_INCLUDE] ||
2347                                         psf->sf_count[MCAST_EXCLUDE] !=
2348                                         im->sfcount[MCAST_EXCLUDE];
2349                         else
2350                                 rv = im->sfcount[MCAST_EXCLUDE] != 0;
2351                 } else
2352                         rv = 1; /* unspecified source; tentatively allow */
2353         }
2354         return rv;
2355 }
2356
2357 #if defined(CONFIG_PROC_FS)
2358 struct igmp_mc_iter_state {
2359         struct seq_net_private p;
2360         struct net_device *dev;
2361         struct in_device *in_dev;
2362 };
2363
2364 #define igmp_mc_seq_private(seq)        ((struct igmp_mc_iter_state *)(seq)->private)
2365
2366 static inline struct ip_mc_list *igmp_mc_get_first(struct seq_file *seq)
2367 {
2368         struct net *net = seq_file_net(seq);
2369         struct ip_mc_list *im = NULL;
2370         struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2371
2372         state->in_dev = NULL;
2373         for_each_netdev_rcu(net, state->dev) {
2374                 struct in_device *in_dev;
2375
2376                 in_dev = __in_dev_get_rcu(state->dev);
2377                 if (!in_dev)
2378                         continue;
2379                 im = rcu_dereference(in_dev->mc_list);
2380                 if (im) {
2381                         state->in_dev = in_dev;
2382                         break;
2383                 }
2384         }
2385         return im;
2386 }
2387
2388 static struct ip_mc_list *igmp_mc_get_next(struct seq_file *seq, struct ip_mc_list *im)
2389 {
2390         struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2391
2392         im = rcu_dereference(im->next_rcu);
2393         while (!im) {
2394                 state->dev = next_net_device_rcu(state->dev);
2395                 if (!state->dev) {
2396                         state->in_dev = NULL;
2397                         break;
2398                 }
2399                 state->in_dev = __in_dev_get_rcu(state->dev);
2400                 if (!state->in_dev)
2401                         continue;
2402                 im = rcu_dereference(state->in_dev->mc_list);
2403         }
2404         return im;
2405 }
2406
2407 static struct ip_mc_list *igmp_mc_get_idx(struct seq_file *seq, loff_t pos)
2408 {
2409         struct ip_mc_list *im = igmp_mc_get_first(seq);
2410         if (im)
2411                 while (pos && (im = igmp_mc_get_next(seq, im)) != NULL)
2412                         --pos;
2413         return pos ? NULL : im;
2414 }
2415
2416 static void *igmp_mc_seq_start(struct seq_file *seq, loff_t *pos)
2417         __acquires(rcu)
2418 {
2419         rcu_read_lock();
2420         return *pos ? igmp_mc_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
2421 }
2422
2423 static void *igmp_mc_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2424 {
2425         struct ip_mc_list *im;
2426         if (v == SEQ_START_TOKEN)
2427                 im = igmp_mc_get_first(seq);
2428         else
2429                 im = igmp_mc_get_next(seq, v);
2430         ++*pos;
2431         return im;
2432 }
2433
2434 static void igmp_mc_seq_stop(struct seq_file *seq, void *v)
2435         __releases(rcu)
2436 {
2437         struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2438
2439         state->in_dev = NULL;
2440         state->dev = NULL;
2441         rcu_read_unlock();
2442 }
2443
2444 static int igmp_mc_seq_show(struct seq_file *seq, void *v)
2445 {
2446         if (v == SEQ_START_TOKEN)
2447                 seq_puts(seq,
2448                          "Idx\tDevice    : Count Querier\tGroup    Users Timer\tReporter\n");
2449         else {
2450                 struct ip_mc_list *im = (struct ip_mc_list *)v;
2451                 struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2452                 char   *querier;
2453 #ifdef CONFIG_IP_MULTICAST
2454                 querier = IGMP_V1_SEEN(state->in_dev) ? "V1" :
2455                           IGMP_V2_SEEN(state->in_dev) ? "V2" :
2456                           "V3";
2457 #else
2458                 querier = "NONE";
2459 #endif
2460
2461                 if (rcu_dereference(state->in_dev->mc_list) == im) {
2462                         seq_printf(seq, "%d\t%-10s: %5d %7s\n",
2463                                    state->dev->ifindex, state->dev->name, state->in_dev->mc_count, querier);
2464                 }
2465
2466                 seq_printf(seq,
2467                            "\t\t\t\t%08X %5d %d:%08lX\t\t%d\n",
2468                            im->multiaddr, im->users,
2469                            im->tm_running, im->tm_running ?
2470                            jiffies_to_clock_t(im->timer.expires-jiffies) : 0,
2471                            im->reporter);
2472         }
2473         return 0;
2474 }
2475
2476 static const struct seq_operations igmp_mc_seq_ops = {
2477         .start  =       igmp_mc_seq_start,
2478         .next   =       igmp_mc_seq_next,
2479         .stop   =       igmp_mc_seq_stop,
2480         .show   =       igmp_mc_seq_show,
2481 };
2482
2483 static int igmp_mc_seq_open(struct inode *inode, struct file *file)
2484 {
2485         return seq_open_net(inode, file, &igmp_mc_seq_ops,
2486                         sizeof(struct igmp_mc_iter_state));
2487 }
2488
2489 static const struct file_operations igmp_mc_seq_fops = {
2490         .owner          =       THIS_MODULE,
2491         .open           =       igmp_mc_seq_open,
2492         .read           =       seq_read,
2493         .llseek         =       seq_lseek,
2494         .release        =       seq_release_net,
2495 };
2496
2497 struct igmp_mcf_iter_state {
2498         struct seq_net_private p;
2499         struct net_device *dev;
2500         struct in_device *idev;
2501         struct ip_mc_list *im;
2502 };
2503
2504 #define igmp_mcf_seq_private(seq)       ((struct igmp_mcf_iter_state *)(seq)->private)
2505
2506 static inline struct ip_sf_list *igmp_mcf_get_first(struct seq_file *seq)
2507 {
2508         struct net *net = seq_file_net(seq);
2509         struct ip_sf_list *psf = NULL;
2510         struct ip_mc_list *im = NULL;
2511         struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2512
2513         state->idev = NULL;
2514         state->im = NULL;
2515         for_each_netdev_rcu(net, state->dev) {
2516                 struct in_device *idev;
2517                 idev = __in_dev_get_rcu(state->dev);
2518                 if (unlikely(idev == NULL))
2519                         continue;
2520                 im = rcu_dereference(idev->mc_list);
2521                 if (likely(im != NULL)) {
2522                         spin_lock_bh(&im->lock);
2523                         psf = im->sources;
2524                         if (likely(psf != NULL)) {
2525                                 state->im = im;
2526                                 state->idev = idev;
2527                                 break;
2528                         }
2529                         spin_unlock_bh(&im->lock);
2530                 }
2531         }
2532         return psf;
2533 }
2534
2535 static struct ip_sf_list *igmp_mcf_get_next(struct seq_file *seq, struct ip_sf_list *psf)
2536 {
2537         struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2538
2539         psf = psf->sf_next;
2540         while (!psf) {
2541                 spin_unlock_bh(&state->im->lock);
2542                 state->im = state->im->next;
2543                 while (!state->im) {
2544                         state->dev = next_net_device_rcu(state->dev);
2545                         if (!state->dev) {
2546                                 state->idev = NULL;
2547                                 goto out;
2548                         }
2549                         state->idev = __in_dev_get_rcu(state->dev);
2550                         if (!state->idev)
2551                                 continue;
2552                         state->im = rcu_dereference(state->idev->mc_list);
2553                 }
2554                 if (!state->im)
2555                         break;
2556                 spin_lock_bh(&state->im->lock);
2557                 psf = state->im->sources;
2558         }
2559 out:
2560         return psf;
2561 }
2562
2563 static struct ip_sf_list *igmp_mcf_get_idx(struct seq_file *seq, loff_t pos)
2564 {
2565         struct ip_sf_list *psf = igmp_mcf_get_first(seq);
2566         if (psf)
2567                 while (pos && (psf = igmp_mcf_get_next(seq, psf)) != NULL)
2568                         --pos;
2569         return pos ? NULL : psf;
2570 }
2571
2572 static void *igmp_mcf_seq_start(struct seq_file *seq, loff_t *pos)
2573         __acquires(rcu)
2574 {
2575         rcu_read_lock();
2576         return *pos ? igmp_mcf_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
2577 }
2578
2579 static void *igmp_mcf_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2580 {
2581         struct ip_sf_list *psf;
2582         if (v == SEQ_START_TOKEN)
2583                 psf = igmp_mcf_get_first(seq);
2584         else
2585                 psf = igmp_mcf_get_next(seq, v);
2586         ++*pos;
2587         return psf;
2588 }
2589
2590 static void igmp_mcf_seq_stop(struct seq_file *seq, void *v)
2591         __releases(rcu)
2592 {
2593         struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2594         if (likely(state->im != NULL)) {
2595                 spin_unlock_bh(&state->im->lock);
2596                 state->im = NULL;
2597         }
2598         state->idev = NULL;
2599         state->dev = NULL;
2600         rcu_read_unlock();
2601 }
2602
2603 static int igmp_mcf_seq_show(struct seq_file *seq, void *v)
2604 {
2605         struct ip_sf_list *psf = (struct ip_sf_list *)v;
2606         struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2607
2608         if (v == SEQ_START_TOKEN) {
2609                 seq_printf(seq,
2610                            "%3s %6s "
2611                            "%10s %10s %6s %6s\n", "Idx",
2612                            "Device", "MCA",
2613                            "SRC", "INC", "EXC");
2614         } else {
2615                 seq_printf(seq,
2616                            "%3d %6.6s 0x%08x "
2617                            "0x%08x %6lu %6lu\n",
2618                            state->dev->ifindex, state->dev->name,
2619                            ntohl(state->im->multiaddr),
2620                            ntohl(psf->sf_inaddr),
2621                            psf->sf_count[MCAST_INCLUDE],
2622                            psf->sf_count[MCAST_EXCLUDE]);
2623         }
2624         return 0;
2625 }
2626
2627 static const struct seq_operations igmp_mcf_seq_ops = {
2628         .start  =       igmp_mcf_seq_start,
2629         .next   =       igmp_mcf_seq_next,
2630         .stop   =       igmp_mcf_seq_stop,
2631         .show   =       igmp_mcf_seq_show,
2632 };
2633
2634 static int igmp_mcf_seq_open(struct inode *inode, struct file *file)
2635 {
2636         return seq_open_net(inode, file, &igmp_mcf_seq_ops,
2637                         sizeof(struct igmp_mcf_iter_state));
2638 }
2639
2640 static const struct file_operations igmp_mcf_seq_fops = {
2641         .owner          =       THIS_MODULE,
2642         .open           =       igmp_mcf_seq_open,
2643         .read           =       seq_read,
2644         .llseek         =       seq_lseek,
2645         .release        =       seq_release_net,
2646 };
2647
2648 static int __net_init igmp_net_init(struct net *net)
2649 {
2650         struct proc_dir_entry *pde;
2651
2652         pde = proc_net_fops_create(net, "igmp", S_IRUGO, &igmp_mc_seq_fops);
2653         if (!pde)
2654                 goto out_igmp;
2655         pde = proc_net_fops_create(net, "mcfilter", S_IRUGO, &igmp_mcf_seq_fops);
2656         if (!pde)
2657                 goto out_mcfilter;
2658         return 0;
2659
2660 out_mcfilter:
2661         proc_net_remove(net, "igmp");
2662 out_igmp:
2663         return -ENOMEM;
2664 }
2665
2666 static void __net_exit igmp_net_exit(struct net *net)
2667 {
2668         proc_net_remove(net, "mcfilter");
2669         proc_net_remove(net, "igmp");
2670 }
2671
2672 static struct pernet_operations igmp_net_ops = {
2673         .init = igmp_net_init,
2674         .exit = igmp_net_exit,
2675 };
2676
2677 int __init igmp_mc_proc_init(void)
2678 {
2679         return register_pernet_subsys(&igmp_net_ops);
2680 }
2681 #endif