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 <linux/types.h>
  77 #include <linux/kernel.h>
  78 #include <linux/jiffies.h>
  79 #include <linux/string.h>
  80 #include <linux/socket.h>
  81 #include <linux/sockios.h>
  82 #include <linux/in.h>
  83 #include <linux/inet.h>
  84 #include <linux/netdevice.h>
  85 #include <linux/skbuff.h>
  86 #include <linux/inetdevice.h>
  87 #include <linux/igmp.h>
  88 #include <linux/if_arp.h>
  89 #include <linux/rtnetlink.h>
  90 #include <linux/times.h>
  91 #include <linux/pkt_sched.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 <net/inet_common.h>
 101 #include <linux/netfilter_ipv4.h>
 102 #ifdef CONFIG_IP_MROUTE
 103 #include <linux/mroute.h>
 104 #endif
 105 #ifdef CONFIG_PROC_FS
 106 #include <linux/proc_fs.h>
 107 #include <linux/seq_file.h>
 108 #endif
 109
 110 #define IP_MAX_MEMBERSHIPS      20
 111 #define IP_MAX_MSF              10
 112
 113 #ifdef CONFIG_IP_MULTICAST
 114 /* Parameter names and values are taken from igmp-v2-06 draft */
 115
 116 #define IGMP_V1_ROUTER_PRESENT_TIMEOUT          (400*HZ)
 117 #define IGMP_V2_ROUTER_PRESENT_TIMEOUT          (400*HZ)
 118 #define IGMP_V2_UNSOLICITED_REPORT_INTERVAL     (10*HZ)
 119 #define IGMP_V3_UNSOLICITED_REPORT_INTERVAL     (1*HZ)
 120 #define IGMP_QUERY_RESPONSE_INTERVAL            (10*HZ)
 121 #define IGMP_QUERY_ROBUSTNESS_VARIABLE          2
 122
 123
 124 #define IGMP_INITIAL_REPORT_DELAY               (1)
 125
 126 /* IGMP_INITIAL_REPORT_DELAY is not from IGMP specs!
 127  * IGMP specs require to report membership immediately after
 128  * joining a group, but we delay the first report by a
 129  * small interval. It seems more natural and still does not
 130  * contradict to specs provided this delay is small enough.
 131  */
 132
 133 #define IGMP_V1_SEEN(in_dev) \
 134         (IPV4_DEVCONF_ALL(dev_net(in_dev->dev), FORCE_IGMP_VERSION) == 1 || \
 135          IN_DEV_CONF_GET((in_dev), FORCE_IGMP_VERSION) == 1 || \
 136          ((in_dev)->mr_v1_seen && \
 137           time_before(jiffies, (in_dev)->mr_v1_seen)))
 138 #define IGMP_V2_SEEN(in_dev) \
 139         (IPV4_DEVCONF_ALL(dev_net(in_dev->dev), FORCE_IGMP_VERSION) == 2 || \
 140          IN_DEV_CONF_GET((in_dev), FORCE_IGMP_VERSION) == 2 || \
 141          ((in_dev)->mr_v2_seen && \
 142           time_before(jiffies, (in_dev)->mr_v2_seen)))
 143
 144 static int unsolicited_report_interval(struct in_device *in_dev)
 145 {
 146         int interval_ms, interval_jiffies;
 147
 148         if (IGMP_V1_SEEN(in_dev) || IGMP_V2_SEEN(in_dev))
 149                 interval_ms = IN_DEV_CONF_GET(
 150                         in_dev,
 151                         IGMPV2_UNSOLICITED_REPORT_INTERVAL);
 152         else /* v3 */
 153                 interval_ms = IN_DEV_CONF_GET(
 154                         in_dev,
 155                         IGMPV3_UNSOLICITED_REPORT_INTERVAL);
 156
 157         interval_jiffies = msecs_to_jiffies(interval_ms);
 158
 159         /* _timer functions can't handle a delay of 0 jiffies so ensure
 160          *  we always return a positive value.
 161          */
 162         if (interval_jiffies <= 0)
 163                 interval_jiffies = 1;
 164         return interval_jiffies;
 165 }
 166
 167 static void igmpv3_add_delrec(struct in_device *in_dev, struct ip_mc_list *im);
 168 static void igmpv3_del_delrec(struct in_device *in_dev, __be32 multiaddr);
 169 static void igmpv3_clear_delrec(struct in_device *in_dev);
 170 static int sf_setstate(struct ip_mc_list *pmc);
 171 static void sf_markstate(struct ip_mc_list *pmc);
 172 #endif
 173 static void ip_mc_clear_src(struct ip_mc_list *pmc);
 174 static int ip_mc_add_src(struct in_device *in_dev, __be32 *pmca, int sfmode,
 175                          int sfcount, __be32 *psfsrc, int delta);
 176
 177 static void ip_ma_put(struct ip_mc_list *im)
 178 {
 179         if (atomic_dec_and_test(&im->refcnt)) {
 180                 in_dev_put(im->interface);
 181                 kfree_rcu(im, rcu);
 182         }
 183 }
 184
 185 #define for_each_pmc_rcu(in_dev, pmc)                           \
 186         for (pmc = rcu_dereference(in_dev->mc_list);            \
 187              pmc != NULL;                                       \
 188              pmc = rcu_dereference(pmc->next_rcu))
 189
 190 #define for_each_pmc_rtnl(in_dev, pmc)                          \
 191         for (pmc = rtnl_dereference(in_dev->mc_list);           \
 192              pmc != NULL;                                       \
 193              pmc = rtnl_dereference(pmc->next_rcu))
 194
 195 #ifdef CONFIG_IP_MULTICAST
 196
 197 /*
 198  *      Timer management
 199  */
 200
 201 static void igmp_stop_timer(struct ip_mc_list *im)
 202 {
 203         spin_lock_bh(&im->lock);
 204         if (del_timer(&im->timer))
 205                 atomic_dec(&im->refcnt);
 206         im->tm_running = 0;
 207         im->reporter = 0;
 208         im->unsolicit_count = 0;
 209         spin_unlock_bh(&im->lock);
 210 }
 211
 212 /* It must be called with locked im->lock */
 213 static void igmp_start_timer(struct ip_mc_list *im, int max_delay)
 214 {
 215         int tv = prandom_u32() % max_delay;
 216
 217         im->tm_running = 1;
 218         if (!mod_timer(&im->timer, jiffies+tv+2))
 219                 atomic_inc(&im->refcnt);
 220 }
 221
 222 static void igmp_gq_start_timer(struct in_device *in_dev)
 223 {
 224         int tv = prandom_u32() % in_dev->mr_maxdelay;
 225
 226         in_dev->mr_gq_running = 1;
 227         if (!mod_timer(&in_dev->mr_gq_timer, jiffies+tv+2))
 228                 in_dev_hold(in_dev);
 229 }
 230
 231 static void igmp_ifc_start_timer(struct in_device *in_dev, int delay)
 232 {
 233         int tv = prandom_u32() % delay;
 234
 235         if (!mod_timer(&in_dev->mr_ifc_timer, jiffies+tv+2))
 236                 in_dev_hold(in_dev);
 237 }
 238
 239 static void igmp_mod_timer(struct ip_mc_list *im, int max_delay)
 240 {
 241         spin_lock_bh(&im->lock);
 242         im->unsolicit_count = 0;
 243         if (del_timer(&im->timer)) {
 244                 if ((long)(im->timer.expires-jiffies) < max_delay) {
 245                         add_timer(&im->timer);
 246                         im->tm_running = 1;
 247                         spin_unlock_bh(&im->lock);
 248                         return;
 249                 }
 250                 atomic_dec(&im->refcnt);
 251         }
 252         igmp_start_timer(im, max_delay);
 253         spin_unlock_bh(&im->lock);
 254 }
 255
 256
 257 /*
 258  *      Send an IGMP report.
 259  */
 260
 261 #define IGMP_SIZE (sizeof(struct igmphdr)+sizeof(struct iphdr)+4)
 262
 263
 264 static int is_in(struct ip_mc_list *pmc, struct ip_sf_list *psf, int type,
 265         int gdeleted, int sdeleted)
 266 {
 267         switch (type) {
 268         case IGMPV3_MODE_IS_INCLUDE:
 269         case IGMPV3_MODE_IS_EXCLUDE:
 270                 if (gdeleted || sdeleted)
 271                         return 0;
 272                 if (!(pmc->gsquery && !psf->sf_gsresp)) {
 273                         if (pmc->sfmode == MCAST_INCLUDE)
 274                                 return 1;
 275                         /* don't include if this source is excluded
 276                          * in all filters
 277                          */
 278                         if (psf->sf_count[MCAST_INCLUDE])
 279                                 return type == IGMPV3_MODE_IS_INCLUDE;
 280                         return pmc->sfcount[MCAST_EXCLUDE] ==
 281                                 psf->sf_count[MCAST_EXCLUDE];
 282                 }
 283                 return 0;
 284         case IGMPV3_CHANGE_TO_INCLUDE:
 285                 if (gdeleted || sdeleted)
 286                         return 0;
 287                 return psf->sf_count[MCAST_INCLUDE] != 0;
 288         case IGMPV3_CHANGE_TO_EXCLUDE:
 289                 if (gdeleted || sdeleted)
 290                         return 0;
 291                 if (pmc->sfcount[MCAST_EXCLUDE] == 0 ||
 292                     psf->sf_count[MCAST_INCLUDE])
 293                         return 0;
 294                 return pmc->sfcount[MCAST_EXCLUDE] ==
 295                         psf->sf_count[MCAST_EXCLUDE];
 296         case IGMPV3_ALLOW_NEW_SOURCES:
 297                 if (gdeleted || !psf->sf_crcount)
 298                         return 0;
 299                 return (pmc->sfmode == MCAST_INCLUDE) ^ sdeleted;
 300         case IGMPV3_BLOCK_OLD_SOURCES:
 301                 if (pmc->sfmode == MCAST_INCLUDE)
 302                         return gdeleted || (psf->sf_crcount && sdeleted);
 303                 return psf->sf_crcount && !gdeleted && !sdeleted;
 304         }
 305         return 0;
 306 }
 307
 308 static int
 309 igmp_scount(struct ip_mc_list *pmc, int type, int gdeleted, int sdeleted)
 310 {
 311         struct ip_sf_list *psf;
 312         int scount = 0;
 313
 314         for (psf = pmc->sources; psf; psf = psf->sf_next) {
 315                 if (!is_in(pmc, psf, type, gdeleted, sdeleted))
 316                         continue;
 317                 scount++;
 318         }
 319         return scount;
 320 }
 321
 322 static struct sk_buff *igmpv3_newpack(struct net_device *dev, unsigned int mtu)
 323 {
 324         struct sk_buff *skb;
 325         struct rtable *rt;
 326         struct iphdr *pip;
 327         struct igmpv3_report *pig;
 328         struct net *net = dev_net(dev);
 329         struct flowi4 fl4;
 330         int hlen = LL_RESERVED_SPACE(dev);
 331         int tlen = dev->needed_tailroom;
 332         unsigned int size = mtu;
 333
 334         while (1) {
 335                 skb = alloc_skb(size + hlen + tlen,
 336                                 GFP_ATOMIC | __GFP_NOWARN);
 337                 if (skb)
 338                         break;
 339                 size >>= 1;
 340                 if (size < 256)
 341                         return NULL;
 342         }
 343         skb->priority = TC_PRIO_CONTROL;
 344
 345         rt = ip_route_output_ports(net, &fl4, NULL, IGMPV3_ALL_MCR, 0,
 346                                    0, 0,
 347                                    IPPROTO_IGMP, 0, dev->ifindex);
 348         if (IS_ERR(rt)) {
 349                 kfree_skb(skb);
 350                 return NULL;
 351         }
 352
 353         skb_dst_set(skb, &rt->dst);
 354         skb->dev = dev;
 355
 356         skb->reserved_tailroom = skb_end_offset(skb) -
 357                                  min(mtu, skb_end_offset(skb));
 358         skb_reserve(skb, hlen);
 359
 360         skb_reset_network_header(skb);
 361         pip = ip_hdr(skb);
 362         skb_put(skb, sizeof(struct iphdr) + 4);
 363
 364         pip->version  = 4;
 365         pip->ihl      = (sizeof(struct iphdr)+4)>>2;
 366         pip->tos      = 0xc0;
 367         pip->frag_off = htons(IP_DF);
 368         pip->ttl      = 1;
 369         pip->daddr    = fl4.daddr;
 370         pip->saddr    = fl4.saddr;
 371         pip->protocol = IPPROTO_IGMP;
 372         pip->tot_len  = 0;      /* filled in later */
 373         ip_select_ident(net, skb, NULL);
 374         ((u8 *)&pip[1])[0] = IPOPT_RA;
 375         ((u8 *)&pip[1])[1] = 4;
 376         ((u8 *)&pip[1])[2] = 0;
 377         ((u8 *)&pip[1])[3] = 0;
 378
 379         skb->transport_header = skb->network_header + sizeof(struct iphdr) + 4;
 380         skb_put(skb, sizeof(*pig));
 381         pig = igmpv3_report_hdr(skb);
 382         pig->type = IGMPV3_HOST_MEMBERSHIP_REPORT;
 383         pig->resv1 = 0;
 384         pig->csum = 0;
 385         pig->resv2 = 0;
 386         pig->ngrec = 0;
 387         return skb;
 388 }
 389
 390 static int igmpv3_sendpack(struct sk_buff *skb)
 391 {
 392         struct igmphdr *pig = igmp_hdr(skb);
 393         const int igmplen = skb_tail_pointer(skb) - skb_transport_header(skb);
 394
 395         pig->csum = ip_compute_csum(igmp_hdr(skb), igmplen);
 396
 397         return ip_local_out(skb);
 398 }
 399
 400 static int grec_size(struct ip_mc_list *pmc, int type, int gdel, int sdel)
 401 {
 402         return sizeof(struct igmpv3_grec) + 4*igmp_scount(pmc, type, gdel, sdel);
 403 }
 404
 405 static struct sk_buff *add_grhead(struct sk_buff *skb, struct ip_mc_list *pmc,
 406         int type, struct igmpv3_grec **ppgr)
 407 {
 408         struct net_device *dev = pmc->interface->dev;
 409         struct igmpv3_report *pih;
 410         struct igmpv3_grec *pgr;
 411
 412         if (!skb)
 413                 skb = igmpv3_newpack(dev, dev->mtu);
 414         if (!skb)
 415                 return NULL;
 416         pgr = (struct igmpv3_grec *)skb_put(skb, sizeof(struct igmpv3_grec));
 417         pgr->grec_type = type;
 418         pgr->grec_auxwords = 0;
 419         pgr->grec_nsrcs = 0;
 420         pgr->grec_mca = pmc->multiaddr;
 421         pih = igmpv3_report_hdr(skb);
 422         pih->ngrec = htons(ntohs(pih->ngrec)+1);
 423         *ppgr = pgr;
 424         return skb;
 425 }
 426
 427 #define AVAILABLE(skb)  ((skb) ? skb_availroom(skb) : 0)
 428
 429 static struct sk_buff *add_grec(struct sk_buff *skb, struct ip_mc_list *pmc,
 430         int type, int gdeleted, int sdeleted)
 431 {
 432         struct net_device *dev = pmc->interface->dev;
 433         struct igmpv3_report *pih;
 434         struct igmpv3_grec *pgr = NULL;
 435         struct ip_sf_list *psf, *psf_next, *psf_prev, **psf_list;
 436         int scount, stotal, first, isquery, truncate;
 437
 438         if (pmc->multiaddr == IGMP_ALL_HOSTS)
 439                 return skb;
 440
 441         isquery = type == IGMPV3_MODE_IS_INCLUDE ||
 442                   type == IGMPV3_MODE_IS_EXCLUDE;
 443         truncate = type == IGMPV3_MODE_IS_EXCLUDE ||
 444                     type == IGMPV3_CHANGE_TO_EXCLUDE;
 445
 446         stotal = scount = 0;
 447
 448         psf_list = sdeleted ? &pmc->tomb : &pmc->sources;
 449
 450         if (!*psf_list)
 451                 goto empty_source;
 452
 453         pih = skb ? igmpv3_report_hdr(skb) : NULL;
 454
 455         /* EX and TO_EX get a fresh packet, if needed */
 456         if (truncate) {
 457                 if (pih && pih->ngrec &&
 458                     AVAILABLE(skb) < grec_size(pmc, type, gdeleted, sdeleted)) {
 459                         if (skb)
 460                                 igmpv3_sendpack(skb);
 461                         skb = igmpv3_newpack(dev, dev->mtu);
 462                 }
 463         }
 464         first = 1;
 465         psf_prev = NULL;
 466         for (psf = *psf_list; psf; psf = psf_next) {
 467                 __be32 *psrc;
 468
 469                 psf_next = psf->sf_next;
 470
 471                 if (!is_in(pmc, psf, type, gdeleted, sdeleted)) {
 472                         psf_prev = psf;
 473                         continue;
 474                 }
 475
 476                 /* clear marks on query responses */
 477                 if (isquery)
 478                         psf->sf_gsresp = 0;
 479
 480                 if (AVAILABLE(skb) < sizeof(__be32) +
 481                     first*sizeof(struct igmpv3_grec)) {
 482                         if (truncate && !first)
 483                                 break;   /* truncate these */
 484                         if (pgr)
 485                                 pgr->grec_nsrcs = htons(scount);
 486                         if (skb)
 487                                 igmpv3_sendpack(skb);
 488                         skb = igmpv3_newpack(dev, dev->mtu);
 489                         first = 1;
 490                         scount = 0;
 491                 }
 492                 if (first) {
 493                         skb = add_grhead(skb, pmc, type, &pgr);
 494                         first = 0;
 495                 }
 496                 if (!skb)
 497                         return NULL;
 498                 psrc = (__be32 *)skb_put(skb, sizeof(__be32));
 499                 *psrc = psf->sf_inaddr;
 500                 scount++; stotal++;
 501                 if ((type == IGMPV3_ALLOW_NEW_SOURCES ||
 502                      type == IGMPV3_BLOCK_OLD_SOURCES) && psf->sf_crcount) {
 503                         psf->sf_crcount--;
 504                         if ((sdeleted || gdeleted) && psf->sf_crcount == 0) {
 505                                 if (psf_prev)
 506                                         psf_prev->sf_next = psf->sf_next;
 507                                 else
 508                                         *psf_list = psf->sf_next;
 509                                 kfree(psf);
 510                                 continue;
 511                         }
 512                 }
 513                 psf_prev = psf;
 514         }
 515
 516 empty_source:
 517         if (!stotal) {
 518                 if (type == IGMPV3_ALLOW_NEW_SOURCES ||
 519                     type == IGMPV3_BLOCK_OLD_SOURCES)
 520                         return skb;
 521                 if (pmc->crcount || isquery) {
 522                         /* make sure we have room for group header */
 523                         if (skb && AVAILABLE(skb) < sizeof(struct igmpv3_grec)) {
 524                                 igmpv3_sendpack(skb);
 525                                 skb = NULL; /* add_grhead will get a new one */
 526                         }
 527                         skb = add_grhead(skb, pmc, type, &pgr);
 528                 }
 529         }
 530         if (pgr)
 531                 pgr->grec_nsrcs = htons(scount);
 532
 533         if (isquery)
 534                 pmc->gsquery = 0;       /* clear query state on report */
 535         return skb;
 536 }
 537
 538 static int igmpv3_send_report(struct in_device *in_dev, struct ip_mc_list *pmc)
 539 {
 540         struct sk_buff *skb = NULL;
 541         int type;
 542
 543         if (!pmc) {
 544                 rcu_read_lock();
 545                 for_each_pmc_rcu(in_dev, pmc) {
 546                         if (pmc->multiaddr == IGMP_ALL_HOSTS)
 547                                 continue;
 548                         spin_lock_bh(&pmc->lock);
 549                         if (pmc->sfcount[MCAST_EXCLUDE])
 550                                 type = IGMPV3_MODE_IS_EXCLUDE;
 551                         else
 552                                 type = IGMPV3_MODE_IS_INCLUDE;
 553                         skb = add_grec(skb, pmc, type, 0, 0);
 554                         spin_unlock_bh(&pmc->lock);
 555                 }
 556                 rcu_read_unlock();
 557         } else {
 558                 spin_lock_bh(&pmc->lock);
 559                 if (pmc->sfcount[MCAST_EXCLUDE])
 560                         type = IGMPV3_MODE_IS_EXCLUDE;
 561                 else
 562                         type = IGMPV3_MODE_IS_INCLUDE;
 563                 skb = add_grec(skb, pmc, type, 0, 0);
 564                 spin_unlock_bh(&pmc->lock);
 565         }
 566         if (!skb)
 567                 return 0;
 568         return igmpv3_sendpack(skb);
 569 }
 570
 571 /*
 572  * remove zero-count source records from a source filter list
 573  */
 574 static void igmpv3_clear_zeros(struct ip_sf_list **ppsf)
 575 {
 576         struct ip_sf_list *psf_prev, *psf_next, *psf;
 577
 578         psf_prev = NULL;
 579         for (psf = *ppsf; psf; psf = psf_next) {
 580                 psf_next = psf->sf_next;
 581                 if (psf->sf_crcount == 0) {
 582                         if (psf_prev)
 583                                 psf_prev->sf_next = psf->sf_next;
 584                         else
 585                                 *ppsf = psf->sf_next;
 586                         kfree(psf);
 587                 } else
 588                         psf_prev = psf;
 589         }
 590 }
 591
 592 static void igmpv3_send_cr(struct in_device *in_dev)
 593 {
 594         struct ip_mc_list *pmc, *pmc_prev, *pmc_next;
 595         struct sk_buff *skb = NULL;
 596         int type, dtype;
 597
 598         rcu_read_lock();
 599         spin_lock_bh(&in_dev->mc_tomb_lock);
 600
 601         /* deleted MCA's */
 602         pmc_prev = NULL;
 603         for (pmc = in_dev->mc_tomb; pmc; pmc = pmc_next) {
 604                 pmc_next = pmc->next;
 605                 if (pmc->sfmode == MCAST_INCLUDE) {
 606                         type = IGMPV3_BLOCK_OLD_SOURCES;
 607                         dtype = IGMPV3_BLOCK_OLD_SOURCES;
 608                         skb = add_grec(skb, pmc, type, 1, 0);
 609                         skb = add_grec(skb, pmc, dtype, 1, 1);
 610                 }
 611                 if (pmc->crcount) {
 612                         if (pmc->sfmode == MCAST_EXCLUDE) {
 613                                 type = IGMPV3_CHANGE_TO_INCLUDE;
 614                                 skb = add_grec(skb, pmc, type, 1, 0);
 615                         }
 616                         pmc->crcount--;
 617                         if (pmc->crcount == 0) {
 618                                 igmpv3_clear_zeros(&pmc->tomb);
 619                                 igmpv3_clear_zeros(&pmc->sources);
 620                         }
 621                 }
 622                 if (pmc->crcount == 0 && !pmc->tomb && !pmc->sources) {
 623                         if (pmc_prev)
 624                                 pmc_prev->next = pmc_next;
 625                         else
 626                                 in_dev->mc_tomb = pmc_next;
 627                         in_dev_put(pmc->interface);
 628                         kfree(pmc);
 629                 } else
 630                         pmc_prev = pmc;
 631         }
 632         spin_unlock_bh(&in_dev->mc_tomb_lock);
 633
 634         /* change recs */
 635         for_each_pmc_rcu(in_dev, pmc) {
 636                 spin_lock_bh(&pmc->lock);
 637                 if (pmc->sfcount[MCAST_EXCLUDE]) {
 638                         type = IGMPV3_BLOCK_OLD_SOURCES;
 639                         dtype = IGMPV3_ALLOW_NEW_SOURCES;
 640                 } else {
 641                         type = IGMPV3_ALLOW_NEW_SOURCES;
 642                         dtype = IGMPV3_BLOCK_OLD_SOURCES;
 643                 }
 644                 skb = add_grec(skb, pmc, type, 0, 0);
 645                 skb = add_grec(skb, pmc, dtype, 0, 1);  /* deleted sources */
 646
 647                 /* filter mode changes */
 648                 if (pmc->crcount) {
 649                         if (pmc->sfmode == MCAST_EXCLUDE)
 650                                 type = IGMPV3_CHANGE_TO_EXCLUDE;
 651                         else
 652                                 type = IGMPV3_CHANGE_TO_INCLUDE;
 653                         skb = add_grec(skb, pmc, type, 0, 0);
 654                         pmc->crcount--;
 655                 }
 656                 spin_unlock_bh(&pmc->lock);
 657         }
 658         rcu_read_unlock();
 659
 660         if (!skb)
 661                 return;
 662         (void) igmpv3_sendpack(skb);
 663 }
 664
 665 static int igmp_send_report(struct in_device *in_dev, struct ip_mc_list *pmc,
 666         int type)
 667 {
 668         struct sk_buff *skb;
 669         struct iphdr *iph;
 670         struct igmphdr *ih;
 671         struct rtable *rt;
 672         struct net_device *dev = in_dev->dev;
 673         struct net *net = dev_net(dev);
 674         __be32  group = pmc ? pmc->multiaddr : 0;
 675         struct flowi4 fl4;
 676         __be32  dst;
 677         int hlen, tlen;
 678
 679         if (type == IGMPV3_HOST_MEMBERSHIP_REPORT)
 680                 return igmpv3_send_report(in_dev, pmc);
 681         else if (type == IGMP_HOST_LEAVE_MESSAGE)
 682                 dst = IGMP_ALL_ROUTER;
 683         else
 684                 dst = group;
 685
 686         rt = ip_route_output_ports(net, &fl4, NULL, dst, 0,
 687                                    0, 0,
 688                                    IPPROTO_IGMP, 0, dev->ifindex);
 689         if (IS_ERR(rt))
 690                 return -1;
 691
 692         hlen = LL_RESERVED_SPACE(dev);
 693         tlen = dev->needed_tailroom;
 694         skb = alloc_skb(IGMP_SIZE + hlen + tlen, GFP_ATOMIC);
 695         if (!skb) {
 696                 ip_rt_put(rt);
 697                 return -1;
 698         }
 699         skb->priority = TC_PRIO_CONTROL;
 700
 701         skb_dst_set(skb, &rt->dst);
 702
 703         skb_reserve(skb, hlen);
 704
 705         skb_reset_network_header(skb);
 706         iph = ip_hdr(skb);
 707         skb_put(skb, sizeof(struct iphdr) + 4);
 708
 709         iph->version  = 4;
 710         iph->ihl      = (sizeof(struct iphdr)+4)>>2;
 711         iph->tos      = 0xc0;
 712         iph->frag_off = htons(IP_DF);
 713         iph->ttl      = 1;
 714         iph->daddr    = dst;
 715         iph->saddr    = fl4.saddr;
 716         iph->protocol = IPPROTO_IGMP;
 717         ip_select_ident(net, skb, NULL);
 718         ((u8 *)&iph[1])[0] = IPOPT_RA;
 719         ((u8 *)&iph[1])[1] = 4;
 720         ((u8 *)&iph[1])[2] = 0;
 721         ((u8 *)&iph[1])[3] = 0;
 722
 723         ih = (struct igmphdr *)skb_put(skb, sizeof(struct igmphdr));
 724         ih->type = type;
 725         ih->code = 0;
 726         ih->csum = 0;
 727         ih->group = group;
 728         ih->csum = ip_compute_csum((void *)ih, sizeof(struct igmphdr));
 729
 730         return ip_local_out(skb);
 731 }
 732
 733 static void igmp_gq_timer_expire(unsigned long data)
 734 {
 735         struct in_device *in_dev = (struct in_device *)data;
 736
 737         in_dev->mr_gq_running = 0;
 738         igmpv3_send_report(in_dev, NULL);
 739         in_dev_put(in_dev);
 740 }
 741
 742 static void igmp_ifc_timer_expire(unsigned long data)
 743 {
 744         struct in_device *in_dev = (struct in_device *)data;
 745
 746         igmpv3_send_cr(in_dev);
 747         if (in_dev->mr_ifc_count) {
 748                 in_dev->mr_ifc_count--;
 749                 igmp_ifc_start_timer(in_dev,
 750                                      unsolicited_report_interval(in_dev));
 751         }
 752         in_dev_put(in_dev);
 753 }
 754
 755 static void igmp_ifc_event(struct in_device *in_dev)
 756 {
 757         if (IGMP_V1_SEEN(in_dev) || IGMP_V2_SEEN(in_dev))
 758                 return;
 759         in_dev->mr_ifc_count = in_dev->mr_qrv ?: sysctl_igmp_qrv;
 760         igmp_ifc_start_timer(in_dev, 1);
 761 }
 762
 763
 764 static void igmp_timer_expire(unsigned long data)
 765 {
 766         struct ip_mc_list *im = (struct ip_mc_list *)data;
 767         struct in_device *in_dev = im->interface;
 768
 769         spin_lock(&im->lock);
 770         im->tm_running = 0;
 771
 772         if (im->unsolicit_count) {
 773                 im->unsolicit_count--;
 774                 igmp_start_timer(im, unsolicited_report_interval(in_dev));
 775         }
 776         im->reporter = 1;
 777         spin_unlock(&im->lock);
 778
 779         if (IGMP_V1_SEEN(in_dev))
 780                 igmp_send_report(in_dev, im, IGMP_HOST_MEMBERSHIP_REPORT);
 781         else if (IGMP_V2_SEEN(in_dev))
 782                 igmp_send_report(in_dev, im, IGMPV2_HOST_MEMBERSHIP_REPORT);
 783         else
 784                 igmp_send_report(in_dev, im, IGMPV3_HOST_MEMBERSHIP_REPORT);
 785
 786         ip_ma_put(im);
 787 }
 788
 789 /* mark EXCLUDE-mode sources */
 790 static int igmp_xmarksources(struct ip_mc_list *pmc, int nsrcs, __be32 *srcs)
 791 {
 792         struct ip_sf_list *psf;
 793         int i, scount;
 794
 795         scount = 0;
 796         for (psf = pmc->sources; psf; psf = psf->sf_next) {
 797                 if (scount == nsrcs)
 798                         break;
 799                 for (i = 0; i < nsrcs; i++) {
 800                         /* skip inactive filters */
 801                         if (psf->sf_count[MCAST_INCLUDE] ||
 802                             pmc->sfcount[MCAST_EXCLUDE] !=
 803                             psf->sf_count[MCAST_EXCLUDE])
 804                                 break;
 805                         if (srcs[i] == psf->sf_inaddr) {
 806                                 scount++;
 807                                 break;
 808                         }
 809                 }
 810         }
 811         pmc->gsquery = 0;
 812         if (scount == nsrcs)    /* all sources excluded */
 813                 return 0;
 814         return 1;
 815 }
 816
 817 static int igmp_marksources(struct ip_mc_list *pmc, int nsrcs, __be32 *srcs)
 818 {
 819         struct ip_sf_list *psf;
 820         int i, scount;
 821
 822         if (pmc->sfmode == MCAST_EXCLUDE)
 823                 return igmp_xmarksources(pmc, nsrcs, srcs);
 824
 825         /* mark INCLUDE-mode sources */
 826         scount = 0;
 827         for (psf = pmc->sources; psf; psf = psf->sf_next) {
 828                 if (scount == nsrcs)
 829                         break;
 830                 for (i = 0; i < nsrcs; i++)
 831                         if (srcs[i] == psf->sf_inaddr) {
 832                                 psf->sf_gsresp = 1;
 833                                 scount++;
 834                                 break;
 835                         }
 836         }
 837         if (!scount) {
 838                 pmc->gsquery = 0;
 839                 return 0;
 840         }
 841         pmc->gsquery = 1;
 842         return 1;
 843 }
 844
 845 /* return true if packet was dropped */
 846 static bool igmp_heard_report(struct in_device *in_dev, __be32 group)
 847 {
 848         struct ip_mc_list *im;
 849
 850         /* Timers are only set for non-local groups */
 851
 852         if (group == IGMP_ALL_HOSTS)
 853                 return false;
 854
 855         rcu_read_lock();
 856         for_each_pmc_rcu(in_dev, im) {
 857                 if (im->multiaddr == group) {
 858                         igmp_stop_timer(im);
 859                         break;
 860                 }
 861         }
 862         rcu_read_unlock();
 863         return false;
 864 }
 865
 866 /* return true if packet was dropped */
 867 static bool igmp_heard_query(struct in_device *in_dev, struct sk_buff *skb,
 868         int len)
 869 {
 870         struct igmphdr          *ih = igmp_hdr(skb);
 871         struct igmpv3_query *ih3 = igmpv3_query_hdr(skb);
 872         struct ip_mc_list       *im;
 873         __be32                  group = ih->group;
 874         int                     max_delay;
 875         int                     mark = 0;
 876
 877
 878         if (len == 8) {
 879                 if (ih->code == 0) {
 880                         /* Alas, old v1 router presents here. */
 881
 882                         max_delay = IGMP_QUERY_RESPONSE_INTERVAL;
 883                         in_dev->mr_v1_seen = jiffies +
 884                                 IGMP_V1_ROUTER_PRESENT_TIMEOUT;
 885                         group = 0;
 886                 } else {
 887                         /* v2 router present */
 888                         max_delay = ih->code*(HZ/IGMP_TIMER_SCALE);
 889                         in_dev->mr_v2_seen = jiffies +
 890                                 IGMP_V2_ROUTER_PRESENT_TIMEOUT;
 891                 }
 892                 /* cancel the interface change timer */
 893                 in_dev->mr_ifc_count = 0;
 894                 if (del_timer(&in_dev->mr_ifc_timer))
 895                         __in_dev_put(in_dev);
 896                 /* clear deleted report items */
 897                 igmpv3_clear_delrec(in_dev);
 898         } else if (len < 12) {
 899                 return true;    /* ignore bogus packet; freed by caller */
 900         } else if (IGMP_V1_SEEN(in_dev)) {
 901                 /* This is a v3 query with v1 queriers present */
 902                 max_delay = IGMP_QUERY_RESPONSE_INTERVAL;
 903                 group = 0;
 904         } else if (IGMP_V2_SEEN(in_dev)) {
 905                 /* this is a v3 query with v2 queriers present;
 906                  * Interpretation of the max_delay code is problematic here.
 907                  * A real v2 host would use ih_code directly, while v3 has a
 908                  * different encoding. We use the v3 encoding as more likely
 909                  * to be intended in a v3 query.
 910                  */
 911                 max_delay = IGMPV3_MRC(ih3->code)*(HZ/IGMP_TIMER_SCALE);
 912                 if (!max_delay)
 913                         max_delay = 1;  /* can't mod w/ 0 */
 914         } else { /* v3 */
 915                 if (!pskb_may_pull(skb, sizeof(struct igmpv3_query)))
 916                         return true;
 917
 918                 ih3 = igmpv3_query_hdr(skb);
 919                 if (ih3->nsrcs) {
 920                         if (!pskb_may_pull(skb, sizeof(struct igmpv3_query)
 921                                            + ntohs(ih3->nsrcs)*sizeof(__be32)))
 922                                 return true;
 923                         ih3 = igmpv3_query_hdr(skb);
 924                 }
 925
 926                 max_delay = IGMPV3_MRC(ih3->code)*(HZ/IGMP_TIMER_SCALE);
 927                 if (!max_delay)
 928                         max_delay = 1;  /* can't mod w/ 0 */
 929                 in_dev->mr_maxdelay = max_delay;
 930                 if (ih3->qrv)
 931                         in_dev->mr_qrv = ih3->qrv;
 932                 if (!group) { /* general query */
 933                         if (ih3->nsrcs)
 934                                 return true;    /* no sources allowed */
 935                         igmp_gq_start_timer(in_dev);
 936                         return false;
 937                 }
 938                 /* mark sources to include, if group & source-specific */
 939                 mark = ih3->nsrcs != 0;
 940         }
 941
 942         /*
 943          * - Start the timers in all of our membership records
 944          *   that the query applies to for the interface on
 945          *   which the query arrived excl. those that belong
 946          *   to a "local" group (224.0.0.X)
 947          * - For timers already running check if they need to
 948          *   be reset.
 949          * - Use the igmp->igmp_code field as the maximum
 950          *   delay possible
 951          */
 952         rcu_read_lock();
 953         for_each_pmc_rcu(in_dev, im) {
 954                 int changed;
 955
 956                 if (group && group != im->multiaddr)
 957                         continue;
 958                 if (im->multiaddr == IGMP_ALL_HOSTS)
 959                         continue;
 960                 spin_lock_bh(&im->lock);
 961                 if (im->tm_running)
 962                         im->gsquery = im->gsquery && mark;
 963                 else
 964                         im->gsquery = mark;
 965                 changed = !im->gsquery ||
 966                         igmp_marksources(im, ntohs(ih3->nsrcs), ih3->srcs);
 967                 spin_unlock_bh(&im->lock);
 968                 if (changed)
 969                         igmp_mod_timer(im, max_delay);
 970         }
 971         rcu_read_unlock();
 972         return false;
 973 }
 974
 975 /* called in rcu_read_lock() section */
 976 int igmp_rcv(struct sk_buff *skb)
 977 {
 978         /* This basically follows the spec line by line -- see RFC1112 */
 979         struct igmphdr *ih;
 980         struct in_device *in_dev = __in_dev_get_rcu(skb->dev);
 981         int len = skb->len;
 982         bool dropped = true;
 983
 984         if (!in_dev)
 985                 goto drop;
 986
 987         if (!pskb_may_pull(skb, sizeof(struct igmphdr)))
 988                 goto drop;
 989
 990         if (skb_checksum_simple_validate(skb))
 991                 goto drop;
 992
 993         ih = igmp_hdr(skb);
 994         switch (ih->type) {
 995         case IGMP_HOST_MEMBERSHIP_QUERY:
 996                 dropped = igmp_heard_query(in_dev, skb, len);
 997                 break;
 998         case IGMP_HOST_MEMBERSHIP_REPORT:
 999         case IGMPV2_HOST_MEMBERSHIP_REPORT:
1000                 /* Is it our report looped back? */
1001                 if (rt_is_output_route(skb_rtable(skb)))
1002                         break;
1003                 /* don't rely on MC router hearing unicast reports */
1004                 if (skb->pkt_type == PACKET_MULTICAST ||
1005                     skb->pkt_type == PACKET_BROADCAST)
1006                         dropped = igmp_heard_report(in_dev, ih->group);
1007                 break;
1008         case IGMP_PIM:
1009 #ifdef CONFIG_IP_PIMSM_V1
1010                 return pim_rcv_v1(skb);
1011 #endif
1012         case IGMPV3_HOST_MEMBERSHIP_REPORT:
1013         case IGMP_DVMRP:
1014         case IGMP_TRACE:
1015         case IGMP_HOST_LEAVE_MESSAGE:
1016         case IGMP_MTRACE:
1017         case IGMP_MTRACE_RESP:
1018                 break;
1019         default:
1020                 break;
1021         }
1022
1023 drop:
1024         if (dropped)
1025                 kfree_skb(skb);
1026         else
1027                 consume_skb(skb);
1028         return 0;
1029 }
1030
1031 #endif
1032
1033
1034 /*
1035  *      Add a filter to a device
1036  */
1037
1038 static void ip_mc_filter_add(struct in_device *in_dev, __be32 addr)
1039 {
1040         char buf[MAX_ADDR_LEN];
1041         struct net_device *dev = in_dev->dev;
1042
1043         /* Checking for IFF_MULTICAST here is WRONG-WRONG-WRONG.
1044            We will get multicast token leakage, when IFF_MULTICAST
1045            is changed. This check should be done in ndo_set_rx_mode
1046            routine. Something sort of:
1047            if (dev->mc_list && dev->flags&IFF_MULTICAST) { do it; }
1048            --ANK
1049            */
1050         if (arp_mc_map(addr, buf, dev, 0) == 0)
1051                 dev_mc_add(dev, buf);
1052 }
1053
1054 /*
1055  *      Remove a filter from a device
1056  */
1057
1058 static void ip_mc_filter_del(struct in_device *in_dev, __be32 addr)
1059 {
1060         char buf[MAX_ADDR_LEN];
1061         struct net_device *dev = in_dev->dev;
1062
1063         if (arp_mc_map(addr, buf, dev, 0) == 0)
1064                 dev_mc_del(dev, buf);
1065 }
1066
1067 #ifdef CONFIG_IP_MULTICAST
1068 /*
1069  * deleted ip_mc_list manipulation
1070  */
1071 static void igmpv3_add_delrec(struct in_device *in_dev, struct ip_mc_list *im)
1072 {
1073         struct ip_mc_list *pmc;
1074
1075         /* this is an "ip_mc_list" for convenience; only the fields below
1076          * are actually used. In particular, the refcnt and users are not
1077          * used for management of the delete list. Using the same structure
1078          * for deleted items allows change reports to use common code with
1079          * non-deleted or query-response MCA's.
1080          */
1081         pmc = kzalloc(sizeof(*pmc), GFP_KERNEL);
1082         if (!pmc)
1083                 return;
1084         spin_lock_bh(&im->lock);
1085         pmc->interface = im->interface;
1086         in_dev_hold(in_dev);
1087         pmc->multiaddr = im->multiaddr;
1088         pmc->crcount = in_dev->mr_qrv ?: sysctl_igmp_qrv;
1089         pmc->sfmode = im->sfmode;
1090         if (pmc->sfmode == MCAST_INCLUDE) {
1091                 struct ip_sf_list *psf;
1092
1093                 pmc->tomb = im->tomb;
1094                 pmc->sources = im->sources;
1095                 im->tomb = im->sources = NULL;
1096                 for (psf = pmc->sources; psf; psf = psf->sf_next)
1097                         psf->sf_crcount = pmc->crcount;
1098         }
1099         spin_unlock_bh(&im->lock);
1100
1101         spin_lock_bh(&in_dev->mc_tomb_lock);
1102         pmc->next = in_dev->mc_tomb;
1103         in_dev->mc_tomb = pmc;
1104         spin_unlock_bh(&in_dev->mc_tomb_lock);
1105 }
1106
1107 static void igmpv3_del_delrec(struct in_device *in_dev, __be32 multiaddr)
1108 {
1109         struct ip_mc_list *pmc, *pmc_prev;
1110         struct ip_sf_list *psf, *psf_next;
1111
1112         spin_lock_bh(&in_dev->mc_tomb_lock);
1113         pmc_prev = NULL;
1114         for (pmc = in_dev->mc_tomb; pmc; pmc = pmc->next) {
1115                 if (pmc->multiaddr == multiaddr)
1116                         break;
1117                 pmc_prev = pmc;
1118         }
1119         if (pmc) {
1120                 if (pmc_prev)
1121                         pmc_prev->next = pmc->next;
1122                 else
1123                         in_dev->mc_tomb = pmc->next;
1124         }
1125         spin_unlock_bh(&in_dev->mc_tomb_lock);
1126         if (pmc) {
1127                 for (psf = pmc->tomb; psf; psf = psf_next) {
1128                         psf_next = psf->sf_next;
1129                         kfree(psf);
1130                 }
1131                 in_dev_put(pmc->interface);
1132                 kfree(pmc);
1133         }
1134 }
1135
1136 static void igmpv3_clear_delrec(struct in_device *in_dev)
1137 {
1138         struct ip_mc_list *pmc, *nextpmc;
1139
1140         spin_lock_bh(&in_dev->mc_tomb_lock);
1141         pmc = in_dev->mc_tomb;
1142         in_dev->mc_tomb = NULL;
1143         spin_unlock_bh(&in_dev->mc_tomb_lock);
1144
1145         for (; pmc; pmc = nextpmc) {
1146                 nextpmc = pmc->next;
1147                 ip_mc_clear_src(pmc);
1148                 in_dev_put(pmc->interface);
1149                 kfree(pmc);
1150         }
1151         /* clear dead sources, too */
1152         rcu_read_lock();
1153         for_each_pmc_rcu(in_dev, pmc) {
1154                 struct ip_sf_list *psf, *psf_next;
1155
1156                 spin_lock_bh(&pmc->lock);
1157                 psf = pmc->tomb;
1158                 pmc->tomb = NULL;
1159                 spin_unlock_bh(&pmc->lock);
1160                 for (; psf; psf = psf_next) {
1161                         psf_next = psf->sf_next;
1162                         kfree(psf);
1163                 }
1164         }
1165         rcu_read_unlock();
1166 }
1167 #endif
1168
1169 static void igmp_group_dropped(struct ip_mc_list *im)
1170 {
1171         struct in_device *in_dev = im->interface;
1172 #ifdef CONFIG_IP_MULTICAST
1173         int reporter;
1174 #endif
1175
1176         if (im->loaded) {
1177                 im->loaded = 0;
1178                 ip_mc_filter_del(in_dev, im->multiaddr);
1179         }
1180
1181 #ifdef CONFIG_IP_MULTICAST
1182         if (im->multiaddr == IGMP_ALL_HOSTS)
1183                 return;
1184
1185         reporter = im->reporter;
1186         igmp_stop_timer(im);
1187
1188         if (!in_dev->dead) {
1189                 if (IGMP_V1_SEEN(in_dev))
1190                         return;
1191                 if (IGMP_V2_SEEN(in_dev)) {
1192                         if (reporter)
1193                                 igmp_send_report(in_dev, im, IGMP_HOST_LEAVE_MESSAGE);
1194                         return;
1195                 }
1196                 /* IGMPv3 */
1197                 igmpv3_add_delrec(in_dev, im);
1198
1199                 igmp_ifc_event(in_dev);
1200         }
1201 #endif
1202 }
1203
1204 static void igmp_group_added(struct ip_mc_list *im)
1205 {
1206         struct in_device *in_dev = im->interface;
1207
1208         if (im->loaded == 0) {
1209                 im->loaded = 1;
1210                 ip_mc_filter_add(in_dev, im->multiaddr);
1211         }
1212
1213 #ifdef CONFIG_IP_MULTICAST
1214         if (im->multiaddr == IGMP_ALL_HOSTS)
1215                 return;
1216
1217         if (in_dev->dead)
1218                 return;
1219         if (IGMP_V1_SEEN(in_dev) || IGMP_V2_SEEN(in_dev)) {
1220                 spin_lock_bh(&im->lock);
1221                 igmp_start_timer(im, IGMP_INITIAL_REPORT_DELAY);
1222                 spin_unlock_bh(&im->lock);
1223                 return;
1224         }
1225         /* else, v3 */
1226
1227         im->crcount = in_dev->mr_qrv ?: sysctl_igmp_qrv;
1228         igmp_ifc_event(in_dev);
1229 #endif
1230 }
1231
1232
1233 /*
1234  *      Multicast list managers
1235  */
1236
1237 static u32 ip_mc_hash(const struct ip_mc_list *im)
1238 {
1239         return hash_32((__force u32)im->multiaddr, MC_HASH_SZ_LOG);
1240 }
1241
1242 static void ip_mc_hash_add(struct in_device *in_dev,
1243                            struct ip_mc_list *im)
1244 {
1245         struct ip_mc_list __rcu **mc_hash;
1246         u32 hash;
1247
1248         mc_hash = rtnl_dereference(in_dev->mc_hash);
1249         if (mc_hash) {
1250                 hash = ip_mc_hash(im);
1251                 im->next_hash = mc_hash[hash];
1252                 rcu_assign_pointer(mc_hash[hash], im);
1253                 return;
1254         }
1255
1256         /* do not use a hash table for small number of items */
1257         if (in_dev->mc_count < 4)
1258                 return;
1259
1260         mc_hash = kzalloc(sizeof(struct ip_mc_list *) << MC_HASH_SZ_LOG,
1261                           GFP_KERNEL);
1262         if (!mc_hash)
1263                 return;
1264
1265         for_each_pmc_rtnl(in_dev, im) {
1266                 hash = ip_mc_hash(im);
1267                 im->next_hash = mc_hash[hash];
1268                 RCU_INIT_POINTER(mc_hash[hash], im);
1269         }
1270
1271         rcu_assign_pointer(in_dev->mc_hash, mc_hash);
1272 }
1273
1274 static void ip_mc_hash_remove(struct in_device *in_dev,
1275                               struct ip_mc_list *im)
1276 {
1277         struct ip_mc_list __rcu **mc_hash = rtnl_dereference(in_dev->mc_hash);
1278         struct ip_mc_list *aux;
1279
1280         if (!mc_hash)
1281                 return;
1282         mc_hash += ip_mc_hash(im);
1283         while ((aux = rtnl_dereference(*mc_hash)) != im)
1284                 mc_hash = &aux->next_hash;
1285         *mc_hash = im->next_hash;
1286 }
1287
1288
1289 /*
1290  *      A socket has joined a multicast group on device dev.
1291  */
1292
1293 void ip_mc_inc_group(struct in_device *in_dev, __be32 addr)
1294 {
1295         struct ip_mc_list *im;
1296
1297         ASSERT_RTNL();
1298
1299         for_each_pmc_rtnl(in_dev, im) {
1300                 if (im->multiaddr == addr) {
1301                         im->users++;
1302                         ip_mc_add_src(in_dev, &addr, MCAST_EXCLUDE, 0, NULL, 0);
1303                         goto out;
1304                 }
1305         }
1306
1307         im = kzalloc(sizeof(*im), GFP_KERNEL);
1308         if (!im)
1309                 goto out;
1310
1311         im->users = 1;
1312         im->interface = in_dev;
1313         in_dev_hold(in_dev);
1314         im->multiaddr = addr;
1315         /* initial mode is (EX, empty) */
1316         im->sfmode = MCAST_EXCLUDE;
1317         im->sfcount[MCAST_EXCLUDE] = 1;
1318         atomic_set(&im->refcnt, 1);
1319         spin_lock_init(&im->lock);
1320 #ifdef CONFIG_IP_MULTICAST
1321         setup_timer(&im->timer, igmp_timer_expire, (unsigned long)im);
1322         im->unsolicit_count = sysctl_igmp_qrv;
1323 #endif
1324
1325         im->next_rcu = in_dev->mc_list;
1326         in_dev->mc_count++;
1327         rcu_assign_pointer(in_dev->mc_list, im);
1328
1329         ip_mc_hash_add(in_dev, im);
1330
1331 #ifdef CONFIG_IP_MULTICAST
1332         igmpv3_del_delrec(in_dev, im->multiaddr);
1333 #endif
1334         igmp_group_added(im);
1335         if (!in_dev->dead)
1336                 ip_rt_multicast_event(in_dev);
1337 out:
1338         return;
1339 }
1340 EXPORT_SYMBOL(ip_mc_inc_group);
1341
1342 static int ip_mc_check_iphdr(struct sk_buff *skb)
1343 {
1344         const struct iphdr *iph;
1345         unsigned int len;
1346         unsigned int offset = skb_network_offset(skb) + sizeof(*iph);
1347
1348         if (!pskb_may_pull(skb, offset))
1349                 return -EINVAL;
1350
1351         iph = ip_hdr(skb);
1352
1353         if (iph->version != 4 || ip_hdrlen(skb) < sizeof(*iph))
1354                 return -EINVAL;
1355
1356         offset += ip_hdrlen(skb) - sizeof(*iph);
1357
1358         if (!pskb_may_pull(skb, offset))
1359                 return -EINVAL;
1360
1361         iph = ip_hdr(skb);
1362
1363         if (unlikely(ip_fast_csum((u8 *)iph, iph->ihl)))
1364                 return -EINVAL;
1365
1366         len = skb_network_offset(skb) + ntohs(iph->tot_len);
1367         if (skb->len < len || len < offset)
1368                 return -EINVAL;
1369
1370         skb_set_transport_header(skb, offset);
1371
1372         return 0;
1373 }
1374
1375 static int ip_mc_check_igmp_reportv3(struct sk_buff *skb)
1376 {
1377         unsigned int len = skb_transport_offset(skb);
1378
1379         len += sizeof(struct igmpv3_report);
1380
1381         return pskb_may_pull(skb, len) ? 0 : -EINVAL;
1382 }
1383
1384 static int ip_mc_check_igmp_query(struct sk_buff *skb)
1385 {
1386         unsigned int len = skb_transport_offset(skb);
1387
1388         len += sizeof(struct igmphdr);
1389         if (skb->len < len)
1390                 return -EINVAL;
1391
1392         /* IGMPv{1,2}? */
1393         if (skb->len != len) {
1394                 /* or IGMPv3? */
1395                 len += sizeof(struct igmpv3_query) - sizeof(struct igmphdr);
1396                 if (skb->len < len || !pskb_may_pull(skb, len))
1397                         return -EINVAL;
1398         }
1399
1400         /* RFC2236+RFC3376 (IGMPv2+IGMPv3) require the multicast link layer
1401          * all-systems destination addresses (224.0.0.1) for general queries
1402          */
1403         if (!igmp_hdr(skb)->group &&
1404             ip_hdr(skb)->daddr != htonl(INADDR_ALLHOSTS_GROUP))
1405                 return -EINVAL;
1406
1407         return 0;
1408 }
1409
1410 static int ip_mc_check_igmp_msg(struct sk_buff *skb)
1411 {
1412         switch (igmp_hdr(skb)->type) {
1413         case IGMP_HOST_LEAVE_MESSAGE:
1414         case IGMP_HOST_MEMBERSHIP_REPORT:
1415         case IGMPV2_HOST_MEMBERSHIP_REPORT:
1416                 /* fall through */
1417                 return 0;
1418         case IGMPV3_HOST_MEMBERSHIP_REPORT:
1419                 return ip_mc_check_igmp_reportv3(skb);
1420         case IGMP_HOST_MEMBERSHIP_QUERY:
1421                 return ip_mc_check_igmp_query(skb);
1422         default:
1423                 return -ENOMSG;
1424         }
1425 }
1426
1427 static inline __sum16 ip_mc_validate_checksum(struct sk_buff *skb)
1428 {
1429         return skb_checksum_simple_validate(skb);
1430 }
1431
1432 static int __ip_mc_check_igmp(struct sk_buff *skb, struct sk_buff **skb_trimmed)
1433
1434 {
1435         struct sk_buff *skb_chk;
1436         unsigned int transport_len;
1437         unsigned int len = skb_transport_offset(skb) + sizeof(struct igmphdr);
1438         int ret;
1439
1440         transport_len = ntohs(ip_hdr(skb)->tot_len) - ip_hdrlen(skb);
1441
1442         skb_get(skb);
1443         skb_chk = skb_checksum_trimmed(skb, transport_len,
1444                                        ip_mc_validate_checksum);
1445         if (!skb_chk)
1446                 return -EINVAL;
1447
1448         if (!pskb_may_pull(skb_chk, len)) {
1449                 kfree_skb(skb_chk);
1450                 return -EINVAL;
1451         }
1452
1453         ret = ip_mc_check_igmp_msg(skb_chk);
1454         if (ret) {
1455                 kfree_skb(skb_chk);
1456                 return ret;
1457         }
1458
1459         if (skb_trimmed)
1460                 *skb_trimmed = skb_chk;
1461         else
1462                 kfree_skb(skb_chk);
1463
1464         return 0;
1465 }
1466
1467 /**
1468  * ip_mc_check_igmp - checks whether this is a sane IGMP packet
1469  * @skb: the skb to validate
1470  * @skb_trimmed: to store an skb pointer trimmed to IPv4 packet tail (optional)
1471  *
1472  * Checks whether an IPv4 packet is a valid IGMP packet. If so sets
1473  * skb network and transport headers accordingly and returns zero.
1474  *
1475  * -EINVAL: A broken packet was detected, i.e. it violates some internet
1476  *  standard
1477  * -ENOMSG: IP header validation succeeded but it is not an IGMP packet.
1478  * -ENOMEM: A memory allocation failure happened.
1479  *
1480  * Optionally, an skb pointer might be provided via skb_trimmed (or set it
1481  * to NULL): After parsing an IGMP packet successfully it will point to
1482  * an skb which has its tail aligned to the IP packet end. This might
1483  * either be the originally provided skb or a trimmed, cloned version if
1484  * the skb frame had data beyond the IP packet. A cloned skb allows us
1485  * to leave the original skb and its full frame unchanged (which might be
1486  * desirable for layer 2 frame jugglers).
1487  *
1488  * The caller needs to release a reference count from any returned skb_trimmed.
1489  */
1490 int ip_mc_check_igmp(struct sk_buff *skb, struct sk_buff **skb_trimmed)
1491 {
1492         int ret = ip_mc_check_iphdr(skb);
1493
1494         if (ret < 0)
1495                 return ret;
1496
1497         if (ip_hdr(skb)->protocol != IPPROTO_IGMP)
1498                 return -ENOMSG;
1499
1500         return __ip_mc_check_igmp(skb, skb_trimmed);
1501 }
1502 EXPORT_SYMBOL(ip_mc_check_igmp);
1503
1504 /*
1505  *      Resend IGMP JOIN report; used by netdev notifier.
1506  */
1507 static void ip_mc_rejoin_groups(struct in_device *in_dev)
1508 {
1509 #ifdef CONFIG_IP_MULTICAST
1510         struct ip_mc_list *im;
1511         int type;
1512
1513         ASSERT_RTNL();
1514
1515         for_each_pmc_rtnl(in_dev, im) {
1516                 if (im->multiaddr == IGMP_ALL_HOSTS)
1517                         continue;
1518
1519                 /* a failover is happening and switches
1520                  * must be notified immediately
1521                  */
1522                 if (IGMP_V1_SEEN(in_dev))
1523                         type = IGMP_HOST_MEMBERSHIP_REPORT;
1524                 else if (IGMP_V2_SEEN(in_dev))
1525                         type = IGMPV2_HOST_MEMBERSHIP_REPORT;
1526                 else
1527                         type = IGMPV3_HOST_MEMBERSHIP_REPORT;
1528                 igmp_send_report(in_dev, im, type);
1529         }
1530 #endif
1531 }
1532
1533 /*
1534  *      A socket has left a multicast group on device dev
1535  */
1536
1537 void ip_mc_dec_group(struct in_device *in_dev, __be32 addr)
1538 {
1539         struct ip_mc_list *i;
1540         struct ip_mc_list __rcu **ip;
1541
1542         ASSERT_RTNL();
1543
1544         for (ip = &in_dev->mc_list;
1545              (i = rtnl_dereference(*ip)) != NULL;
1546              ip = &i->next_rcu) {
1547                 if (i->multiaddr == addr) {
1548                         if (--i->users == 0) {
1549                                 ip_mc_hash_remove(in_dev, i);
1550                                 *ip = i->next_rcu;
1551                                 in_dev->mc_count--;
1552                                 igmp_group_dropped(i);
1553                                 ip_mc_clear_src(i);
1554
1555                                 if (!in_dev->dead)
1556                                         ip_rt_multicast_event(in_dev);
1557
1558                                 ip_ma_put(i);
1559                                 return;
1560                         }
1561                         break;
1562                 }
1563         }
1564 }
1565 EXPORT_SYMBOL(ip_mc_dec_group);
1566
1567 /* Device changing type */
1568
1569 void ip_mc_unmap(struct in_device *in_dev)
1570 {
1571         struct ip_mc_list *pmc;
1572
1573         ASSERT_RTNL();
1574
1575         for_each_pmc_rtnl(in_dev, pmc)
1576                 igmp_group_dropped(pmc);
1577 }
1578
1579 void ip_mc_remap(struct in_device *in_dev)
1580 {
1581         struct ip_mc_list *pmc;
1582
1583         ASSERT_RTNL();
1584
1585         for_each_pmc_rtnl(in_dev, pmc)
1586                 igmp_group_added(pmc);
1587 }
1588
1589 /* Device going down */
1590
1591 void ip_mc_down(struct in_device *in_dev)
1592 {
1593         struct ip_mc_list *pmc;
1594
1595         ASSERT_RTNL();
1596
1597         for_each_pmc_rtnl(in_dev, pmc)
1598                 igmp_group_dropped(pmc);
1599
1600 #ifdef CONFIG_IP_MULTICAST
1601         in_dev->mr_ifc_count = 0;
1602         if (del_timer(&in_dev->mr_ifc_timer))
1603                 __in_dev_put(in_dev);
1604         in_dev->mr_gq_running = 0;
1605         if (del_timer(&in_dev->mr_gq_timer))
1606                 __in_dev_put(in_dev);
1607         igmpv3_clear_delrec(in_dev);
1608 #endif
1609
1610         ip_mc_dec_group(in_dev, IGMP_ALL_HOSTS);
1611 }
1612
1613 void ip_mc_init_dev(struct in_device *in_dev)
1614 {
1615         ASSERT_RTNL();
1616
1617 #ifdef CONFIG_IP_MULTICAST
1618         setup_timer(&in_dev->mr_gq_timer, igmp_gq_timer_expire,
1619                         (unsigned long)in_dev);
1620         setup_timer(&in_dev->mr_ifc_timer, igmp_ifc_timer_expire,
1621                         (unsigned long)in_dev);
1622         in_dev->mr_qrv = sysctl_igmp_qrv;
1623 #endif
1624
1625         spin_lock_init(&in_dev->mc_tomb_lock);
1626 }
1627
1628 /* Device going up */
1629
1630 void ip_mc_up(struct in_device *in_dev)
1631 {
1632         struct ip_mc_list *pmc;
1633
1634         ASSERT_RTNL();
1635
1636 #ifdef CONFIG_IP_MULTICAST
1637         in_dev->mr_qrv = sysctl_igmp_qrv;
1638 #endif
1639         ip_mc_inc_group(in_dev, IGMP_ALL_HOSTS);
1640
1641         for_each_pmc_rtnl(in_dev, pmc)
1642                 igmp_group_added(pmc);
1643 }
1644
1645 /*
1646  *      Device is about to be destroyed: clean up.
1647  */
1648
1649 void ip_mc_destroy_dev(struct in_device *in_dev)
1650 {
1651         struct ip_mc_list *i;
1652
1653         ASSERT_RTNL();
1654
1655         /* Deactivate timers */
1656         ip_mc_down(in_dev);
1657
1658         while ((i = rtnl_dereference(in_dev->mc_list)) != NULL) {
1659                 in_dev->mc_list = i->next_rcu;
1660                 in_dev->mc_count--;
1661
1662                 /* We've dropped the groups in ip_mc_down already */
1663                 ip_mc_clear_src(i);
1664                 ip_ma_put(i);
1665         }
1666 }
1667
1668 /* RTNL is locked */
1669 static struct in_device *ip_mc_find_dev(struct net *net, struct ip_mreqn *imr)
1670 {
1671         struct net_device *dev = NULL;
1672         struct in_device *idev = NULL;
1673
1674         if (imr->imr_ifindex) {
1675                 idev = inetdev_by_index(net, imr->imr_ifindex);
1676                 return idev;
1677         }
1678         if (imr->imr_address.s_addr) {
1679                 dev = __ip_dev_find(net, imr->imr_address.s_addr, false);
1680                 if (!dev)
1681                         return NULL;
1682         }
1683
1684         if (!dev) {
1685                 struct rtable *rt = ip_route_output(net,
1686                                                     imr->imr_multiaddr.s_addr,
1687                                                     0, 0, 0);
1688                 if (!IS_ERR(rt)) {
1689                         dev = rt->dst.dev;
1690                         ip_rt_put(rt);
1691                 }
1692         }
1693         if (dev) {
1694                 imr->imr_ifindex = dev->ifindex;
1695                 idev = __in_dev_get_rtnl(dev);
1696         }
1697         return idev;
1698 }
1699
1700 /*
1701  *      Join a socket to a group
1702  */
1703 int sysctl_igmp_max_memberships __read_mostly = IP_MAX_MEMBERSHIPS;
1704 int sysctl_igmp_max_msf __read_mostly = IP_MAX_MSF;
1705 #ifdef CONFIG_IP_MULTICAST
1706 int sysctl_igmp_qrv __read_mostly = IGMP_QUERY_ROBUSTNESS_VARIABLE;
1707 #endif
1708
1709 static int ip_mc_del1_src(struct ip_mc_list *pmc, int sfmode,
1710         __be32 *psfsrc)
1711 {
1712         struct ip_sf_list *psf, *psf_prev;
1713         int rv = 0;
1714
1715         psf_prev = NULL;
1716         for (psf = pmc->sources; psf; psf = psf->sf_next) {
1717                 if (psf->sf_inaddr == *psfsrc)
1718                         break;
1719                 psf_prev = psf;
1720         }
1721         if (!psf || psf->sf_count[sfmode] == 0) {
1722                 /* source filter not found, or count wrong =>  bug */
1723                 return -ESRCH;
1724         }
1725         psf->sf_count[sfmode]--;
1726         if (psf->sf_count[sfmode] == 0) {
1727                 ip_rt_multicast_event(pmc->interface);
1728         }
1729         if (!psf->sf_count[MCAST_INCLUDE] && !psf->sf_count[MCAST_EXCLUDE]) {
1730 #ifdef CONFIG_IP_MULTICAST
1731                 struct in_device *in_dev = pmc->interface;
1732 #endif
1733
1734                 /* no more filters for this source */
1735                 if (psf_prev)
1736                         psf_prev->sf_next = psf->sf_next;
1737                 else
1738                         pmc->sources = psf->sf_next;
1739 #ifdef CONFIG_IP_MULTICAST
1740                 if (psf->sf_oldin &&
1741                     !IGMP_V1_SEEN(in_dev) && !IGMP_V2_SEEN(in_dev)) {
1742                         psf->sf_crcount = in_dev->mr_qrv ?: sysctl_igmp_qrv;
1743                         psf->sf_next = pmc->tomb;
1744                         pmc->tomb = psf;
1745                         rv = 1;
1746                 } else
1747 #endif
1748                         kfree(psf);
1749         }
1750         return rv;
1751 }
1752
1753 #ifndef CONFIG_IP_MULTICAST
1754 #define igmp_ifc_event(x)       do { } while (0)
1755 #endif
1756
1757 static int ip_mc_del_src(struct in_device *in_dev, __be32 *pmca, int sfmode,
1758                          int sfcount, __be32 *psfsrc, int delta)
1759 {
1760         struct ip_mc_list *pmc;
1761         int     changerec = 0;
1762         int     i, err;
1763
1764         if (!in_dev)
1765                 return -ENODEV;
1766         rcu_read_lock();
1767         for_each_pmc_rcu(in_dev, pmc) {
1768                 if (*pmca == pmc->multiaddr)
1769                         break;
1770         }
1771         if (!pmc) {
1772                 /* MCA not found?? bug */
1773                 rcu_read_unlock();
1774                 return -ESRCH;
1775         }
1776         spin_lock_bh(&pmc->lock);
1777         rcu_read_unlock();
1778 #ifdef CONFIG_IP_MULTICAST
1779         sf_markstate(pmc);
1780 #endif
1781         if (!delta) {
1782                 err = -EINVAL;
1783                 if (!pmc->sfcount[sfmode])
1784                         goto out_unlock;
1785                 pmc->sfcount[sfmode]--;
1786         }
1787         err = 0;
1788         for (i = 0; i < sfcount; i++) {
1789                 int rv = ip_mc_del1_src(pmc, sfmode, &psfsrc[i]);
1790
1791                 changerec |= rv > 0;
1792                 if (!err && rv < 0)
1793                         err = rv;
1794         }
1795         if (pmc->sfmode == MCAST_EXCLUDE &&
1796             pmc->sfcount[MCAST_EXCLUDE] == 0 &&
1797             pmc->sfcount[MCAST_INCLUDE]) {
1798 #ifdef CONFIG_IP_MULTICAST
1799                 struct ip_sf_list *psf;
1800 #endif
1801
1802                 /* filter mode change */
1803                 pmc->sfmode = MCAST_INCLUDE;
1804 #ifdef CONFIG_IP_MULTICAST
1805                 pmc->crcount = in_dev->mr_qrv ?: sysctl_igmp_qrv;
1806                 in_dev->mr_ifc_count = pmc->crcount;
1807                 for (psf = pmc->sources; psf; psf = psf->sf_next)
1808                         psf->sf_crcount = 0;
1809                 igmp_ifc_event(pmc->interface);
1810         } else if (sf_setstate(pmc) || changerec) {
1811                 igmp_ifc_event(pmc->interface);
1812 #endif
1813         }
1814 out_unlock:
1815         spin_unlock_bh(&pmc->lock);
1816         return err;
1817 }
1818
1819 /*
1820  * Add multicast single-source filter to the interface list
1821  */
1822 static int ip_mc_add1_src(struct ip_mc_list *pmc, int sfmode,
1823         __be32 *psfsrc)
1824 {
1825         struct ip_sf_list *psf, *psf_prev;
1826
1827         psf_prev = NULL;
1828         for (psf = pmc->sources; psf; psf = psf->sf_next) {
1829                 if (psf->sf_inaddr == *psfsrc)
1830                         break;
1831                 psf_prev = psf;
1832         }
1833         if (!psf) {
1834                 psf = kzalloc(sizeof(*psf), GFP_ATOMIC);
1835                 if (!psf)
1836                         return -ENOBUFS;
1837                 psf->sf_inaddr = *psfsrc;
1838                 if (psf_prev) {
1839                         psf_prev->sf_next = psf;
1840                 } else
1841                         pmc->sources = psf;
1842         }
1843         psf->sf_count[sfmode]++;
1844         if (psf->sf_count[sfmode] == 1) {
1845                 ip_rt_multicast_event(pmc->interface);
1846         }
1847         return 0;
1848 }
1849
1850 #ifdef CONFIG_IP_MULTICAST
1851 static void sf_markstate(struct ip_mc_list *pmc)
1852 {
1853         struct ip_sf_list *psf;
1854         int mca_xcount = pmc->sfcount[MCAST_EXCLUDE];
1855
1856         for (psf = pmc->sources; psf; psf = psf->sf_next)
1857                 if (pmc->sfcount[MCAST_EXCLUDE]) {
1858                         psf->sf_oldin = mca_xcount ==
1859                                 psf->sf_count[MCAST_EXCLUDE] &&
1860                                 !psf->sf_count[MCAST_INCLUDE];
1861                 } else
1862                         psf->sf_oldin = psf->sf_count[MCAST_INCLUDE] != 0;
1863 }
1864
1865 static int sf_setstate(struct ip_mc_list *pmc)
1866 {
1867         struct ip_sf_list *psf, *dpsf;
1868         int mca_xcount = pmc->sfcount[MCAST_EXCLUDE];
1869         int qrv = pmc->interface->mr_qrv;
1870         int new_in, rv;
1871
1872         rv = 0;
1873         for (psf = pmc->sources; psf; psf = psf->sf_next) {
1874                 if (pmc->sfcount[MCAST_EXCLUDE]) {
1875                         new_in = mca_xcount == psf->sf_count[MCAST_EXCLUDE] &&
1876                                 !psf->sf_count[MCAST_INCLUDE];
1877                 } else
1878                         new_in = psf->sf_count[MCAST_INCLUDE] != 0;
1879                 if (new_in) {
1880                         if (!psf->sf_oldin) {
1881                                 struct ip_sf_list *prev = NULL;
1882
1883                                 for (dpsf = pmc->tomb; dpsf; dpsf = dpsf->sf_next) {
1884                                         if (dpsf->sf_inaddr == psf->sf_inaddr)
1885                                                 break;
1886                                         prev = dpsf;
1887                                 }
1888                                 if (dpsf) {
1889                                         if (prev)
1890                                                 prev->sf_next = dpsf->sf_next;
1891                                         else
1892                                                 pmc->tomb = dpsf->sf_next;
1893                                         kfree(dpsf);
1894                                 }
1895                                 psf->sf_crcount = qrv;
1896                                 rv++;
1897                         }
1898                 } else if (psf->sf_oldin) {
1899
1900                         psf->sf_crcount = 0;
1901                         /*
1902                          * add or update "delete" records if an active filter
1903                          * is now inactive
1904                          */
1905                         for (dpsf = pmc->tomb; dpsf; dpsf = dpsf->sf_next)
1906                                 if (dpsf->sf_inaddr == psf->sf_inaddr)
1907                                         break;
1908                         if (!dpsf) {
1909                                 dpsf = kmalloc(sizeof(*dpsf), GFP_ATOMIC);
1910                                 if (!dpsf)
1911                                         continue;
1912                                 *dpsf = *psf;
1913                                 /* pmc->lock held by callers */
1914                                 dpsf->sf_next = pmc->tomb;
1915                                 pmc->tomb = dpsf;
1916                         }
1917                         dpsf->sf_crcount = qrv;
1918                         rv++;
1919                 }
1920         }
1921         return rv;
1922 }
1923 #endif
1924
1925 /*
1926  * Add multicast source filter list to the interface list
1927  */
1928 static int ip_mc_add_src(struct in_device *in_dev, __be32 *pmca, int sfmode,
1929                          int sfcount, __be32 *psfsrc, int delta)
1930 {
1931         struct ip_mc_list *pmc;
1932         int     isexclude;
1933         int     i, err;
1934
1935         if (!in_dev)
1936                 return -ENODEV;
1937         rcu_read_lock();
1938         for_each_pmc_rcu(in_dev, pmc) {
1939                 if (*pmca == pmc->multiaddr)
1940                         break;
1941         }
1942         if (!pmc) {
1943                 /* MCA not found?? bug */
1944                 rcu_read_unlock();
1945                 return -ESRCH;
1946         }
1947         spin_lock_bh(&pmc->lock);
1948         rcu_read_unlock();
1949
1950 #ifdef CONFIG_IP_MULTICAST
1951         sf_markstate(pmc);
1952 #endif
1953         isexclude = pmc->sfmode == MCAST_EXCLUDE;
1954         if (!delta)
1955                 pmc->sfcount[sfmode]++;
1956         err = 0;
1957         for (i = 0; i < sfcount; i++) {
1958                 err = ip_mc_add1_src(pmc, sfmode, &psfsrc[i]);
1959                 if (err)
1960                         break;
1961         }
1962         if (err) {
1963                 int j;
1964
1965                 if (!delta)
1966                         pmc->sfcount[sfmode]--;
1967                 for (j = 0; j < i; j++)
1968                         (void) ip_mc_del1_src(pmc, sfmode, &psfsrc[j]);
1969         } else if (isexclude != (pmc->sfcount[MCAST_EXCLUDE] != 0)) {
1970 #ifdef CONFIG_IP_MULTICAST
1971                 struct ip_sf_list *psf;
1972                 in_dev = pmc->interface;
1973 #endif
1974
1975                 /* filter mode change */
1976                 if (pmc->sfcount[MCAST_EXCLUDE])
1977                         pmc->sfmode = MCAST_EXCLUDE;
1978                 else if (pmc->sfcount[MCAST_INCLUDE])
1979                         pmc->sfmode = MCAST_INCLUDE;
1980 #ifdef CONFIG_IP_MULTICAST
1981                 /* else no filters; keep old mode for reports */
1982
1983                 pmc->crcount = in_dev->mr_qrv ?: sysctl_igmp_qrv;
1984                 in_dev->mr_ifc_count = pmc->crcount;
1985                 for (psf = pmc->sources; psf; psf = psf->sf_next)
1986                         psf->sf_crcount = 0;
1987                 igmp_ifc_event(in_dev);
1988         } else if (sf_setstate(pmc)) {
1989                 igmp_ifc_event(in_dev);
1990 #endif
1991         }
1992         spin_unlock_bh(&pmc->lock);
1993         return err;
1994 }
1995
1996 static void ip_mc_clear_src(struct ip_mc_list *pmc)
1997 {
1998         struct ip_sf_list *psf, *nextpsf;
1999
2000         for (psf = pmc->tomb; psf; psf = nextpsf) {
2001                 nextpsf = psf->sf_next;
2002                 kfree(psf);
2003         }
2004         pmc->tomb = NULL;
2005         for (psf = pmc->sources; psf; psf = nextpsf) {
2006                 nextpsf = psf->sf_next;
2007                 kfree(psf);
2008         }
2009         pmc->sources = NULL;
2010         pmc->sfmode = MCAST_EXCLUDE;
2011         pmc->sfcount[MCAST_INCLUDE] = 0;
2012         pmc->sfcount[MCAST_EXCLUDE] = 1;
2013 }
2014
2015 /* Join a multicast group
2016  */
2017
2018 int ip_mc_join_group(struct sock *sk, struct ip_mreqn *imr)
2019 {
2020         __be32 addr = imr->imr_multiaddr.s_addr;
2021         struct ip_mc_socklist *iml, *i;
2022         struct in_device *in_dev;
2023         struct inet_sock *inet = inet_sk(sk);
2024         struct net *net = sock_net(sk);
2025         int ifindex;
2026         int count = 0;
2027         int err;
2028
2029         ASSERT_RTNL();
2030
2031         if (!ipv4_is_multicast(addr))
2032                 return -EINVAL;
2033
2034         in_dev = ip_mc_find_dev(net, imr);
2035
2036         if (!in_dev) {
2037                 err = -ENODEV;
2038                 goto done;
2039         }
2040
2041         err = -EADDRINUSE;
2042         ifindex = imr->imr_ifindex;
2043         for_each_pmc_rtnl(inet, i) {
2044                 if (i->multi.imr_multiaddr.s_addr == addr &&
2045                     i->multi.imr_ifindex == ifindex)
2046                         goto done;
2047                 count++;
2048         }
2049         err = -ENOBUFS;
2050         if (count >= sysctl_igmp_max_memberships)
2051                 goto done;
2052         iml = sock_kmalloc(sk, sizeof(*iml), GFP_KERNEL);
2053         if (!iml)
2054                 goto done;
2055
2056         memcpy(&iml->multi, imr, sizeof(*imr));
2057         iml->next_rcu = inet->mc_list;
2058         iml->sflist = NULL;
2059         iml->sfmode = MCAST_EXCLUDE;
2060         rcu_assign_pointer(inet->mc_list, iml);
2061         ip_mc_inc_group(in_dev, addr);
2062         err = 0;
2063 done:
2064         return err;
2065 }
2066 EXPORT_SYMBOL(ip_mc_join_group);
2067
2068 static int ip_mc_leave_src(struct sock *sk, struct ip_mc_socklist *iml,
2069                            struct in_device *in_dev)
2070 {
2071         struct ip_sf_socklist *psf = rtnl_dereference(iml->sflist);
2072         int err;
2073
2074         if (!psf) {
2075                 /* any-source empty exclude case */
2076                 return ip_mc_del_src(in_dev, &iml->multi.imr_multiaddr.s_addr,
2077                         iml->sfmode, 0, NULL, 0);
2078         }
2079         err = ip_mc_del_src(in_dev, &iml->multi.imr_multiaddr.s_addr,
2080                         iml->sfmode, psf->sl_count, psf->sl_addr, 0);
2081         RCU_INIT_POINTER(iml->sflist, NULL);
2082         /* decrease mem now to avoid the memleak warning */
2083         atomic_sub(IP_SFLSIZE(psf->sl_max), &sk->sk_omem_alloc);
2084         kfree_rcu(psf, rcu);
2085         return err;
2086 }
2087
2088 int ip_mc_leave_group(struct sock *sk, struct ip_mreqn *imr)
2089 {
2090         struct inet_sock *inet = inet_sk(sk);
2091         struct ip_mc_socklist *iml;
2092         struct ip_mc_socklist __rcu **imlp;
2093         struct in_device *in_dev;
2094         struct net *net = sock_net(sk);
2095         __be32 group = imr->imr_multiaddr.s_addr;
2096         u32 ifindex;
2097         int ret = -EADDRNOTAVAIL;
2098
2099         ASSERT_RTNL();
2100
2101         in_dev = ip_mc_find_dev(net, imr);
2102         if (!in_dev) {
2103                 ret = -ENODEV;
2104                 goto out;
2105         }
2106         ifindex = imr->imr_ifindex;
2107         for (imlp = &inet->mc_list;
2108              (iml = rtnl_dereference(*imlp)) != NULL;
2109              imlp = &iml->next_rcu) {
2110                 if (iml->multi.imr_multiaddr.s_addr != group)
2111                         continue;
2112                 if (ifindex) {
2113                         if (iml->multi.imr_ifindex != ifindex)
2114                                 continue;
2115                 } else if (imr->imr_address.s_addr && imr->imr_address.s_addr !=
2116                                 iml->multi.imr_address.s_addr)
2117                         continue;
2118
2119                 (void) ip_mc_leave_src(sk, iml, in_dev);
2120
2121                 *imlp = iml->next_rcu;
2122
2123                 ip_mc_dec_group(in_dev, group);
2124
2125                 /* decrease mem now to avoid the memleak warning */
2126                 atomic_sub(sizeof(*iml), &sk->sk_omem_alloc);
2127                 kfree_rcu(iml, rcu);
2128                 return 0;
2129         }
2130 out:
2131         return ret;
2132 }
2133 EXPORT_SYMBOL(ip_mc_leave_group);
2134
2135 int ip_mc_source(int add, int omode, struct sock *sk, struct
2136         ip_mreq_source *mreqs, int ifindex)
2137 {
2138         int err;
2139         struct ip_mreqn imr;
2140         __be32 addr = mreqs->imr_multiaddr;
2141         struct ip_mc_socklist *pmc;
2142         struct in_device *in_dev = NULL;
2143         struct inet_sock *inet = inet_sk(sk);
2144         struct ip_sf_socklist *psl;
2145         struct net *net = sock_net(sk);
2146         int leavegroup = 0;
2147         int i, j, rv;
2148
2149         if (!ipv4_is_multicast(addr))
2150                 return -EINVAL;
2151
2152         ASSERT_RTNL();
2153
2154         imr.imr_multiaddr.s_addr = mreqs->imr_multiaddr;
2155         imr.imr_address.s_addr = mreqs->imr_interface;
2156         imr.imr_ifindex = ifindex;
2157         in_dev = ip_mc_find_dev(net, &imr);
2158
2159         if (!in_dev) {
2160                 err = -ENODEV;
2161                 goto done;
2162         }
2163         err = -EADDRNOTAVAIL;
2164
2165         for_each_pmc_rtnl(inet, pmc) {
2166                 if ((pmc->multi.imr_multiaddr.s_addr ==
2167                      imr.imr_multiaddr.s_addr) &&
2168                     (pmc->multi.imr_ifindex == imr.imr_ifindex))
2169                         break;
2170         }
2171         if (!pmc) {             /* must have a prior join */
2172                 err = -EINVAL;
2173                 goto done;
2174         }
2175         /* if a source filter was set, must be the same mode as before */
2176         if (pmc->sflist) {
2177                 if (pmc->sfmode != omode) {
2178                         err = -EINVAL;
2179                         goto done;
2180                 }
2181         } else if (pmc->sfmode != omode) {
2182                 /* allow mode switches for empty-set filters */
2183                 ip_mc_add_src(in_dev, &mreqs->imr_multiaddr, omode, 0, NULL, 0);
2184                 ip_mc_del_src(in_dev, &mreqs->imr_multiaddr, pmc->sfmode, 0,
2185                         NULL, 0);
2186                 pmc->sfmode = omode;
2187         }
2188
2189         psl = rtnl_dereference(pmc->sflist);
2190         if (!add) {
2191                 if (!psl)
2192                         goto done;      /* err = -EADDRNOTAVAIL */
2193                 rv = !0;
2194                 for (i = 0; i < psl->sl_count; i++) {
2195                         rv = memcmp(&psl->sl_addr[i], &mreqs->imr_sourceaddr,
2196                                 sizeof(__be32));
2197                         if (rv == 0)
2198                                 break;
2199                 }
2200                 if (rv)         /* source not found */
2201                         goto done;      /* err = -EADDRNOTAVAIL */
2202
2203                 /* special case - (INCLUDE, empty) == LEAVE_GROUP */
2204                 if (psl->sl_count == 1 && omode == MCAST_INCLUDE) {
2205                         leavegroup = 1;
2206                         goto done;
2207                 }
2208
2209                 /* update the interface filter */
2210                 ip_mc_del_src(in_dev, &mreqs->imr_multiaddr, omode, 1,
2211                         &mreqs->imr_sourceaddr, 1);
2212
2213                 for (j = i+1; j < psl->sl_count; j++)
2214                         psl->sl_addr[j-1] = psl->sl_addr[j];
2215                 psl->sl_count--;
2216                 err = 0;
2217                 goto done;
2218         }
2219         /* else, add a new source to the filter */
2220
2221         if (psl && psl->sl_count >= sysctl_igmp_max_msf) {
2222                 err = -ENOBUFS;
2223                 goto done;
2224         }
2225         if (!psl || psl->sl_count == psl->sl_max) {
2226                 struct ip_sf_socklist *newpsl;
2227                 int count = IP_SFBLOCK;
2228
2229                 if (psl)
2230                         count += psl->sl_max;
2231                 newpsl = sock_kmalloc(sk, IP_SFLSIZE(count), GFP_KERNEL);
2232                 if (!newpsl) {
2233                         err = -ENOBUFS;
2234                         goto done;
2235                 }
2236                 newpsl->sl_max = count;
2237                 newpsl->sl_count = count - IP_SFBLOCK;
2238                 if (psl) {
2239                         for (i = 0; i < psl->sl_count; i++)
2240                                 newpsl->sl_addr[i] = psl->sl_addr[i];
2241                         /* decrease mem now to avoid the memleak warning */
2242                         atomic_sub(IP_SFLSIZE(psl->sl_max), &sk->sk_omem_alloc);
2243                         kfree_rcu(psl, rcu);
2244                 }
2245                 rcu_assign_pointer(pmc->sflist, newpsl);
2246                 psl = newpsl;
2247         }
2248         rv = 1; /* > 0 for insert logic below if sl_count is 0 */
2249         for (i = 0; i < psl->sl_count; i++) {
2250                 rv = memcmp(&psl->sl_addr[i], &mreqs->imr_sourceaddr,
2251                         sizeof(__be32));
2252                 if (rv == 0)
2253                         break;
2254         }
2255         if (rv == 0)            /* address already there is an error */
2256                 goto done;
2257         for (j = psl->sl_count-1; j >= i; j--)
2258                 psl->sl_addr[j+1] = psl->sl_addr[j];
2259         psl->sl_addr[i] = mreqs->imr_sourceaddr;
2260         psl->sl_count++;
2261         err = 0;
2262         /* update the interface list */
2263         ip_mc_add_src(in_dev, &mreqs->imr_multiaddr, omode, 1,
2264                 &mreqs->imr_sourceaddr, 1);
2265 done:
2266         if (leavegroup)
2267                 err = ip_mc_leave_group(sk, &imr);
2268         return err;
2269 }
2270
2271 int ip_mc_msfilter(struct sock *sk, struct ip_msfilter *msf, int ifindex)
2272 {
2273         int err = 0;
2274         struct ip_mreqn imr;
2275         __be32 addr = msf->imsf_multiaddr;
2276         struct ip_mc_socklist *pmc;
2277         struct in_device *in_dev;
2278         struct inet_sock *inet = inet_sk(sk);
2279         struct ip_sf_socklist *newpsl, *psl;
2280         struct net *net = sock_net(sk);
2281         int leavegroup = 0;
2282
2283         if (!ipv4_is_multicast(addr))
2284                 return -EINVAL;
2285         if (msf->imsf_fmode != MCAST_INCLUDE &&
2286             msf->imsf_fmode != MCAST_EXCLUDE)
2287                 return -EINVAL;
2288
2289         ASSERT_RTNL();
2290
2291         imr.imr_multiaddr.s_addr = msf->imsf_multiaddr;
2292         imr.imr_address.s_addr = msf->imsf_interface;
2293         imr.imr_ifindex = ifindex;
2294         in_dev = ip_mc_find_dev(net, &imr);
2295
2296         if (!in_dev) {
2297                 err = -ENODEV;
2298                 goto done;
2299         }
2300
2301         /* special case - (INCLUDE, empty) == LEAVE_GROUP */
2302         if (msf->imsf_fmode == MCAST_INCLUDE && msf->imsf_numsrc == 0) {
2303                 leavegroup = 1;
2304                 goto done;
2305         }
2306
2307         for_each_pmc_rtnl(inet, pmc) {
2308                 if (pmc->multi.imr_multiaddr.s_addr == msf->imsf_multiaddr &&
2309                     pmc->multi.imr_ifindex == imr.imr_ifindex)
2310                         break;
2311         }
2312         if (!pmc) {             /* must have a prior join */
2313                 err = -EINVAL;
2314                 goto done;
2315         }
2316         if (msf->imsf_numsrc) {
2317                 newpsl = sock_kmalloc(sk, IP_SFLSIZE(msf->imsf_numsrc),
2318                                                            GFP_KERNEL);
2319                 if (!newpsl) {
2320                         err = -ENOBUFS;
2321                         goto done;
2322                 }
2323                 newpsl->sl_max = newpsl->sl_count = msf->imsf_numsrc;
2324                 memcpy(newpsl->sl_addr, msf->imsf_slist,
2325                         msf->imsf_numsrc * sizeof(msf->imsf_slist[0]));
2326                 err = ip_mc_add_src(in_dev, &msf->imsf_multiaddr,
2327                         msf->imsf_fmode, newpsl->sl_count, newpsl->sl_addr, 0);
2328                 if (err) {
2329                         sock_kfree_s(sk, newpsl, IP_SFLSIZE(newpsl->sl_max));
2330                         goto done;
2331                 }
2332         } else {
2333                 newpsl = NULL;
2334                 (void) ip_mc_add_src(in_dev, &msf->imsf_multiaddr,
2335                                      msf->imsf_fmode, 0, NULL, 0);
2336         }
2337         psl = rtnl_dereference(pmc->sflist);
2338         if (psl) {
2339                 (void) ip_mc_del_src(in_dev, &msf->imsf_multiaddr, pmc->sfmode,
2340                         psl->sl_count, psl->sl_addr, 0);
2341                 /* decrease mem now to avoid the memleak warning */
2342                 atomic_sub(IP_SFLSIZE(psl->sl_max), &sk->sk_omem_alloc);
2343                 kfree_rcu(psl, rcu);
2344         } else
2345                 (void) ip_mc_del_src(in_dev, &msf->imsf_multiaddr, pmc->sfmode,
2346                         0, NULL, 0);
2347         rcu_assign_pointer(pmc->sflist, newpsl);
2348         pmc->sfmode = msf->imsf_fmode;
2349         err = 0;
2350 done:
2351         if (leavegroup)
2352                 err = ip_mc_leave_group(sk, &imr);
2353         return err;
2354 }
2355
2356 int ip_mc_msfget(struct sock *sk, struct ip_msfilter *msf,
2357         struct ip_msfilter __user *optval, int __user *optlen)
2358 {
2359         int err, len, count, copycount;
2360         struct ip_mreqn imr;
2361         __be32 addr = msf->imsf_multiaddr;
2362         struct ip_mc_socklist *pmc;
2363         struct in_device *in_dev;
2364         struct inet_sock *inet = inet_sk(sk);
2365         struct ip_sf_socklist *psl;
2366         struct net *net = sock_net(sk);
2367
2368         if (!ipv4_is_multicast(addr))
2369                 return -EINVAL;
2370
2371         rtnl_lock();
2372
2373         imr.imr_multiaddr.s_addr = msf->imsf_multiaddr;
2374         imr.imr_address.s_addr = msf->imsf_interface;
2375         imr.imr_ifindex = 0;
2376         in_dev = ip_mc_find_dev(net, &imr);
2377
2378         if (!in_dev) {
2379                 err = -ENODEV;
2380                 goto done;
2381         }
2382         err = -EADDRNOTAVAIL;
2383
2384         for_each_pmc_rtnl(inet, pmc) {
2385                 if (pmc->multi.imr_multiaddr.s_addr == msf->imsf_multiaddr &&
2386                     pmc->multi.imr_ifindex == imr.imr_ifindex)
2387                         break;
2388         }
2389         if (!pmc)               /* must have a prior join */
2390                 goto done;
2391         msf->imsf_fmode = pmc->sfmode;
2392         psl = rtnl_dereference(pmc->sflist);
2393         rtnl_unlock();
2394         if (!psl) {
2395                 len = 0;
2396                 count = 0;
2397         } else {
2398                 count = psl->sl_count;
2399         }
2400         copycount = count < msf->imsf_numsrc ? count : msf->imsf_numsrc;
2401         len = copycount * sizeof(psl->sl_addr[0]);
2402         msf->imsf_numsrc = count;
2403         if (put_user(IP_MSFILTER_SIZE(copycount), optlen) ||
2404             copy_to_user(optval, msf, IP_MSFILTER_SIZE(0))) {
2405                 return -EFAULT;
2406         }
2407         if (len &&
2408             copy_to_user(&optval->imsf_slist[0], psl->sl_addr, len))
2409                 return -EFAULT;
2410         return 0;
2411 done:
2412         rtnl_unlock();
2413         return err;
2414 }
2415
2416 int ip_mc_gsfget(struct sock *sk, struct group_filter *gsf,
2417         struct group_filter __user *optval, int __user *optlen)
2418 {
2419         int err, i, count, copycount;
2420         struct sockaddr_in *psin;
2421         __be32 addr;
2422         struct ip_mc_socklist *pmc;
2423         struct inet_sock *inet = inet_sk(sk);
2424         struct ip_sf_socklist *psl;
2425
2426         psin = (struct sockaddr_in *)&gsf->gf_group;
2427         if (psin->sin_family != AF_INET)
2428                 return -EINVAL;
2429         addr = psin->sin_addr.s_addr;
2430         if (!ipv4_is_multicast(addr))
2431                 return -EINVAL;
2432
2433         rtnl_lock();
2434
2435         err = -EADDRNOTAVAIL;
2436
2437         for_each_pmc_rtnl(inet, pmc) {
2438                 if (pmc->multi.imr_multiaddr.s_addr == addr &&
2439                     pmc->multi.imr_ifindex == gsf->gf_interface)
2440                         break;
2441         }
2442         if (!pmc)               /* must have a prior join */
2443                 goto done;
2444         gsf->gf_fmode = pmc->sfmode;
2445         psl = rtnl_dereference(pmc->sflist);
2446         rtnl_unlock();
2447         count = psl ? psl->sl_count : 0;
2448         copycount = count < gsf->gf_numsrc ? count : gsf->gf_numsrc;
2449         gsf->gf_numsrc = count;
2450         if (put_user(GROUP_FILTER_SIZE(copycount), optlen) ||
2451             copy_to_user(optval, gsf, GROUP_FILTER_SIZE(0))) {
2452                 return -EFAULT;
2453         }
2454         for (i = 0; i < copycount; i++) {
2455                 struct sockaddr_storage ss;
2456
2457                 psin = (struct sockaddr_in *)&ss;
2458                 memset(&ss, 0, sizeof(ss));
2459                 psin->sin_family = AF_INET;
2460                 psin->sin_addr.s_addr = psl->sl_addr[i];
2461                 if (copy_to_user(&optval->gf_slist[i], &ss, sizeof(ss)))
2462                         return -EFAULT;
2463         }
2464         return 0;
2465 done:
2466         rtnl_unlock();
2467         return err;
2468 }
2469
2470 /*
2471  * check if a multicast source filter allows delivery for a given <src,dst,intf>
2472  */
2473 int ip_mc_sf_allow(struct sock *sk, __be32 loc_addr, __be32 rmt_addr, int dif)
2474 {
2475         struct inet_sock *inet = inet_sk(sk);
2476         struct ip_mc_socklist *pmc;
2477         struct ip_sf_socklist *psl;
2478         int i;
2479         int ret;
2480
2481         ret = 1;
2482         if (!ipv4_is_multicast(loc_addr))
2483                 goto out;
2484
2485         rcu_read_lock();
2486         for_each_pmc_rcu(inet, pmc) {
2487                 if (pmc->multi.imr_multiaddr.s_addr == loc_addr &&
2488                     pmc->multi.imr_ifindex == dif)
2489                         break;
2490         }
2491         ret = inet->mc_all;
2492         if (!pmc)
2493                 goto unlock;
2494         psl = rcu_dereference(pmc->sflist);
2495         ret = (pmc->sfmode == MCAST_EXCLUDE);
2496         if (!psl)
2497                 goto unlock;
2498
2499         for (i = 0; i < psl->sl_count; i++) {
2500                 if (psl->sl_addr[i] == rmt_addr)
2501                         break;
2502         }
2503         ret = 0;
2504         if (pmc->sfmode == MCAST_INCLUDE && i >= psl->sl_count)
2505                 goto unlock;
2506         if (pmc->sfmode == MCAST_EXCLUDE && i < psl->sl_count)
2507                 goto unlock;
2508         ret = 1;
2509 unlock:
2510         rcu_read_unlock();
2511 out:
2512         return ret;
2513 }
2514
2515 /*
2516  *      A socket is closing.
2517  */
2518
2519 void ip_mc_drop_socket(struct sock *sk)
2520 {
2521         struct inet_sock *inet = inet_sk(sk);
2522         struct ip_mc_socklist *iml;
2523         struct net *net = sock_net(sk);
2524
2525         if (!inet->mc_list)
2526                 return;
2527
2528         rtnl_lock();
2529         while ((iml = rtnl_dereference(inet->mc_list)) != NULL) {
2530                 struct in_device *in_dev;
2531
2532                 inet->mc_list = iml->next_rcu;
2533                 in_dev = inetdev_by_index(net, iml->multi.imr_ifindex);
2534                 (void) ip_mc_leave_src(sk, iml, in_dev);
2535                 if (in_dev)
2536                         ip_mc_dec_group(in_dev, iml->multi.imr_multiaddr.s_addr);
2537                 /* decrease mem now to avoid the memleak warning */
2538                 atomic_sub(sizeof(*iml), &sk->sk_omem_alloc);
2539                 kfree_rcu(iml, rcu);
2540         }
2541         rtnl_unlock();
2542 }
2543
2544 /* called with rcu_read_lock() */
2545 int ip_check_mc_rcu(struct in_device *in_dev, __be32 mc_addr, __be32 src_addr, u16 proto)
2546 {
2547         struct ip_mc_list *im;
2548         struct ip_mc_list __rcu **mc_hash;
2549         struct ip_sf_list *psf;
2550         int rv = 0;
2551
2552         mc_hash = rcu_dereference(in_dev->mc_hash);
2553         if (mc_hash) {
2554                 u32 hash = hash_32((__force u32)mc_addr, MC_HASH_SZ_LOG);
2555
2556                 for (im = rcu_dereference(mc_hash[hash]);
2557                      im != NULL;
2558                      im = rcu_dereference(im->next_hash)) {
2559                         if (im->multiaddr == mc_addr)
2560                                 break;
2561                 }
2562         } else {
2563                 for_each_pmc_rcu(in_dev, im) {
2564                         if (im->multiaddr == mc_addr)
2565                                 break;
2566                 }
2567         }
2568         if (im && proto == IPPROTO_IGMP) {
2569                 rv = 1;
2570         } else if (im) {
2571                 if (src_addr) {
2572                         for (psf = im->sources; psf; psf = psf->sf_next) {
2573                                 if (psf->sf_inaddr == src_addr)
2574                                         break;
2575                         }
2576                         if (psf)
2577                                 rv = psf->sf_count[MCAST_INCLUDE] ||
2578                                         psf->sf_count[MCAST_EXCLUDE] !=
2579                                         im->sfcount[MCAST_EXCLUDE];
2580                         else
2581                                 rv = im->sfcount[MCAST_EXCLUDE] != 0;
2582                 } else
2583                         rv = 1; /* unspecified source; tentatively allow */
2584         }
2585         return rv;
2586 }
2587
2588 #if defined(CONFIG_PROC_FS)
2589 struct igmp_mc_iter_state {
2590         struct seq_net_private p;
2591         struct net_device *dev;
2592         struct in_device *in_dev;
2593 };
2594
2595 #define igmp_mc_seq_private(seq)        ((struct igmp_mc_iter_state *)(seq)->private)
2596
2597 static inline struct ip_mc_list *igmp_mc_get_first(struct seq_file *seq)
2598 {
2599         struct net *net = seq_file_net(seq);
2600         struct ip_mc_list *im = NULL;
2601         struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2602
2603         state->in_dev = NULL;
2604         for_each_netdev_rcu(net, state->dev) {
2605                 struct in_device *in_dev;
2606
2607                 in_dev = __in_dev_get_rcu(state->dev);
2608                 if (!in_dev)
2609                         continue;
2610                 im = rcu_dereference(in_dev->mc_list);
2611                 if (im) {
2612                         state->in_dev = in_dev;
2613                         break;
2614                 }
2615         }
2616         return im;
2617 }
2618
2619 static struct ip_mc_list *igmp_mc_get_next(struct seq_file *seq, struct ip_mc_list *im)
2620 {
2621         struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2622
2623         im = rcu_dereference(im->next_rcu);
2624         while (!im) {
2625                 state->dev = next_net_device_rcu(state->dev);
2626                 if (!state->dev) {
2627                         state->in_dev = NULL;
2628                         break;
2629                 }
2630                 state->in_dev = __in_dev_get_rcu(state->dev);
2631                 if (!state->in_dev)
2632                         continue;
2633                 im = rcu_dereference(state->in_dev->mc_list);
2634         }
2635         return im;
2636 }
2637
2638 static struct ip_mc_list *igmp_mc_get_idx(struct seq_file *seq, loff_t pos)
2639 {
2640         struct ip_mc_list *im = igmp_mc_get_first(seq);
2641         if (im)
2642                 while (pos && (im = igmp_mc_get_next(seq, im)) != NULL)
2643                         --pos;
2644         return pos ? NULL : im;
2645 }
2646
2647 static void *igmp_mc_seq_start(struct seq_file *seq, loff_t *pos)
2648         __acquires(rcu)
2649 {
2650         rcu_read_lock();
2651         return *pos ? igmp_mc_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
2652 }
2653
2654 static void *igmp_mc_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2655 {
2656         struct ip_mc_list *im;
2657         if (v == SEQ_START_TOKEN)
2658                 im = igmp_mc_get_first(seq);
2659         else
2660                 im = igmp_mc_get_next(seq, v);
2661         ++*pos;
2662         return im;
2663 }
2664
2665 static void igmp_mc_seq_stop(struct seq_file *seq, void *v)
2666         __releases(rcu)
2667 {
2668         struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2669
2670         state->in_dev = NULL;
2671         state->dev = NULL;
2672         rcu_read_unlock();
2673 }
2674
2675 static int igmp_mc_seq_show(struct seq_file *seq, void *v)
2676 {
2677         if (v == SEQ_START_TOKEN)
2678                 seq_puts(seq,
2679                          "Idx\tDevice    : Count Querier\tGroup    Users Timer\tReporter\n");
2680         else {
2681                 struct ip_mc_list *im = (struct ip_mc_list *)v;
2682                 struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2683                 char   *querier;
2684                 long delta;
2685
2686 #ifdef CONFIG_IP_MULTICAST
2687                 querier = IGMP_V1_SEEN(state->in_dev) ? "V1" :
2688                           IGMP_V2_SEEN(state->in_dev) ? "V2" :
2689                           "V3";
2690 #else
2691                 querier = "NONE";
2692 #endif
2693
2694                 if (rcu_access_pointer(state->in_dev->mc_list) == im) {
2695                         seq_printf(seq, "%d\t%-10s: %5d %7s\n",
2696                                    state->dev->ifindex, state->dev->name, state->in_dev->mc_count, querier);
2697                 }
2698
2699                 delta = im->timer.expires - jiffies;
2700                 seq_printf(seq,
2701                            "\t\t\t\t%08X %5d %d:%08lX\t\t%d\n",
2702                            im->multiaddr, im->users,
2703                            im->tm_running,
2704                            im->tm_running ? jiffies_delta_to_clock_t(delta) : 0,
2705                            im->reporter);
2706         }
2707         return 0;
2708 }
2709
2710 static const struct seq_operations igmp_mc_seq_ops = {
2711         .start  =       igmp_mc_seq_start,
2712         .next   =       igmp_mc_seq_next,
2713         .stop   =       igmp_mc_seq_stop,
2714         .show   =       igmp_mc_seq_show,
2715 };
2716
2717 static int igmp_mc_seq_open(struct inode *inode, struct file *file)
2718 {
2719         return seq_open_net(inode, file, &igmp_mc_seq_ops,
2720                         sizeof(struct igmp_mc_iter_state));
2721 }
2722
2723 static const struct file_operations igmp_mc_seq_fops = {
2724         .owner          =       THIS_MODULE,
2725         .open           =       igmp_mc_seq_open,
2726         .read           =       seq_read,
2727         .llseek         =       seq_lseek,
2728         .release        =       seq_release_net,
2729 };
2730
2731 struct igmp_mcf_iter_state {
2732         struct seq_net_private p;
2733         struct net_device *dev;
2734         struct in_device *idev;
2735         struct ip_mc_list *im;
2736 };
2737
2738 #define igmp_mcf_seq_private(seq)       ((struct igmp_mcf_iter_state *)(seq)->private)
2739
2740 static inline struct ip_sf_list *igmp_mcf_get_first(struct seq_file *seq)
2741 {
2742         struct net *net = seq_file_net(seq);
2743         struct ip_sf_list *psf = NULL;
2744         struct ip_mc_list *im = NULL;
2745         struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2746
2747         state->idev = NULL;
2748         state->im = NULL;
2749         for_each_netdev_rcu(net, state->dev) {
2750                 struct in_device *idev;
2751                 idev = __in_dev_get_rcu(state->dev);
2752                 if (unlikely(!idev))
2753                         continue;
2754                 im = rcu_dereference(idev->mc_list);
2755                 if (likely(im)) {
2756                         spin_lock_bh(&im->lock);
2757                         psf = im->sources;
2758                         if (likely(psf)) {
2759                                 state->im = im;
2760                                 state->idev = idev;
2761                                 break;
2762                         }
2763                         spin_unlock_bh(&im->lock);
2764                 }
2765         }
2766         return psf;
2767 }
2768
2769 static struct ip_sf_list *igmp_mcf_get_next(struct seq_file *seq, struct ip_sf_list *psf)
2770 {
2771         struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2772
2773         psf = psf->sf_next;
2774         while (!psf) {
2775                 spin_unlock_bh(&state->im->lock);
2776                 state->im = state->im->next;
2777                 while (!state->im) {
2778                         state->dev = next_net_device_rcu(state->dev);
2779                         if (!state->dev) {
2780                                 state->idev = NULL;
2781                                 goto out;
2782                         }
2783                         state->idev = __in_dev_get_rcu(state->dev);
2784                         if (!state->idev)
2785                                 continue;
2786                         state->im = rcu_dereference(state->idev->mc_list);
2787                 }
2788                 if (!state->im)
2789                         break;
2790                 spin_lock_bh(&state->im->lock);
2791                 psf = state->im->sources;
2792         }
2793 out:
2794         return psf;
2795 }
2796
2797 static struct ip_sf_list *igmp_mcf_get_idx(struct seq_file *seq, loff_t pos)
2798 {
2799         struct ip_sf_list *psf = igmp_mcf_get_first(seq);
2800         if (psf)
2801                 while (pos && (psf = igmp_mcf_get_next(seq, psf)) != NULL)
2802                         --pos;
2803         return pos ? NULL : psf;
2804 }
2805
2806 static void *igmp_mcf_seq_start(struct seq_file *seq, loff_t *pos)
2807         __acquires(rcu)
2808 {
2809         rcu_read_lock();
2810         return *pos ? igmp_mcf_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
2811 }
2812
2813 static void *igmp_mcf_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2814 {
2815         struct ip_sf_list *psf;
2816         if (v == SEQ_START_TOKEN)
2817                 psf = igmp_mcf_get_first(seq);
2818         else
2819                 psf = igmp_mcf_get_next(seq, v);
2820         ++*pos;
2821         return psf;
2822 }
2823
2824 static void igmp_mcf_seq_stop(struct seq_file *seq, void *v)
2825         __releases(rcu)
2826 {
2827         struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2828         if (likely(state->im)) {
2829                 spin_unlock_bh(&state->im->lock);
2830                 state->im = NULL;
2831         }
2832         state->idev = NULL;
2833         state->dev = NULL;
2834         rcu_read_unlock();
2835 }
2836
2837 static int igmp_mcf_seq_show(struct seq_file *seq, void *v)
2838 {
2839         struct ip_sf_list *psf = (struct ip_sf_list *)v;
2840         struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2841
2842         if (v == SEQ_START_TOKEN) {
2843                 seq_puts(seq, "Idx Device        MCA        SRC    INC    EXC\n");
2844         } else {
2845                 seq_printf(seq,
2846                            "%3d %6.6s 0x%08x "
2847                            "0x%08x %6lu %6lu\n",
2848                            state->dev->ifindex, state->dev->name,
2849                            ntohl(state->im->multiaddr),
2850                            ntohl(psf->sf_inaddr),
2851                            psf->sf_count[MCAST_INCLUDE],
2852                            psf->sf_count[MCAST_EXCLUDE]);
2853         }
2854         return 0;
2855 }
2856
2857 static const struct seq_operations igmp_mcf_seq_ops = {
2858         .start  =       igmp_mcf_seq_start,
2859         .next   =       igmp_mcf_seq_next,
2860         .stop   =       igmp_mcf_seq_stop,
2861         .show   =       igmp_mcf_seq_show,
2862 };
2863
2864 static int igmp_mcf_seq_open(struct inode *inode, struct file *file)
2865 {
2866         return seq_open_net(inode, file, &igmp_mcf_seq_ops,
2867                         sizeof(struct igmp_mcf_iter_state));
2868 }
2869
2870 static const struct file_operations igmp_mcf_seq_fops = {
2871         .owner          =       THIS_MODULE,
2872         .open           =       igmp_mcf_seq_open,
2873         .read           =       seq_read,
2874         .llseek         =       seq_lseek,
2875         .release        =       seq_release_net,
2876 };
2877
2878 static int __net_init igmp_net_init(struct net *net)
2879 {
2880         struct proc_dir_entry *pde;
2881         int err;
2882
2883         pde = proc_create("igmp", S_IRUGO, net->proc_net, &igmp_mc_seq_fops);
2884         if (!pde)
2885                 goto out_igmp;
2886         pde = proc_create("mcfilter", S_IRUGO, net->proc_net,
2887                           &igmp_mcf_seq_fops);
2888         if (!pde)
2889                 goto out_mcfilter;
2890         err = inet_ctl_sock_create(&net->ipv4.mc_autojoin_sk, AF_INET,
2891                                    SOCK_DGRAM, 0, net);
2892         if (err < 0) {
2893                 pr_err("Failed to initialize the IGMP autojoin socket (err %d)\n",
2894                        err);
2895                 goto out_sock;
2896         }
2897
2898         return 0;
2899
2900 out_sock:
2901         remove_proc_entry("mcfilter", net->proc_net);
2902 out_mcfilter:
2903         remove_proc_entry("igmp", net->proc_net);
2904 out_igmp:
2905         return -ENOMEM;
2906 }
2907
2908 static void __net_exit igmp_net_exit(struct net *net)
2909 {
2910         remove_proc_entry("mcfilter", net->proc_net);
2911         remove_proc_entry("igmp", net->proc_net);
2912         inet_ctl_sock_destroy(net->ipv4.mc_autojoin_sk);
2913 }
2914
2915 static struct pernet_operations igmp_net_ops = {
2916         .init = igmp_net_init,
2917         .exit = igmp_net_exit,
2918 };
2919 #endif
2920
2921 static int igmp_netdev_event(struct notifier_block *this,
2922                              unsigned long event, void *ptr)
2923 {
2924         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2925         struct in_device *in_dev;
2926
2927         switch (event) {
2928         case NETDEV_RESEND_IGMP:
2929                 in_dev = __in_dev_get_rtnl(dev);
2930                 if (in_dev)
2931                         ip_mc_rejoin_groups(in_dev);
2932                 break;
2933         default:
2934                 break;
2935         }
2936         return NOTIFY_DONE;
2937 }
2938
2939 static struct notifier_block igmp_notifier = {
2940         .notifier_call = igmp_netdev_event,
2941 };
2942
2943 int __init igmp_mc_init(void)
2944 {
2945 #if defined(CONFIG_PROC_FS)
2946         int err;
2947
2948         err = register_pernet_subsys(&igmp_net_ops);
2949         if (err)
2950                 return err;
2951         err = register_netdevice_notifier(&igmp_notifier);
2952         if (err)
2953                 goto reg_notif_fail;
2954         return 0;
2955
2956 reg_notif_fail:
2957         unregister_pernet_subsys(&igmp_net_ops);
2958         return err;
2959 #else
2960         return register_netdevice_notifier(&igmp_notifier);
2961 #endif
2962 }