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 #ifdef CONFIG_IP_MULTICAST
 111 /* Parameter names and values are taken from igmp-v2-06 draft */
 112
 113 #define IGMP_V1_ROUTER_PRESENT_TIMEOUT          (400*HZ)
 114 #define IGMP_V2_ROUTER_PRESENT_TIMEOUT          (400*HZ)
 115 #define IGMP_V2_UNSOLICITED_REPORT_INTERVAL     (10*HZ)
 116 #define IGMP_V3_UNSOLICITED_REPORT_INTERVAL     (1*HZ)
 117 #define IGMP_QUERY_RESPONSE_INTERVAL            (10*HZ)
 118 #define IGMP_QUERY_ROBUSTNESS_VARIABLE          2
 119
 120
 121 #define IGMP_INITIAL_REPORT_DELAY               (1)
 122
 123 /* IGMP_INITIAL_REPORT_DELAY is not from IGMP specs!
 124  * IGMP specs require to report membership immediately after
 125  * joining a group, but we delay the first report by a
 126  * small interval. It seems more natural and still does not
 127  * contradict to specs provided this delay is small enough.
 128  */
 129
 130 #define IGMP_V1_SEEN(in_dev) \
 131         (IPV4_DEVCONF_ALL(dev_net(in_dev->dev), FORCE_IGMP_VERSION) == 1 || \
 132          IN_DEV_CONF_GET((in_dev), FORCE_IGMP_VERSION) == 1 || \
 133          ((in_dev)->mr_v1_seen && \
 134           time_before(jiffies, (in_dev)->mr_v1_seen)))
 135 #define IGMP_V2_SEEN(in_dev) \
 136         (IPV4_DEVCONF_ALL(dev_net(in_dev->dev), FORCE_IGMP_VERSION) == 2 || \
 137          IN_DEV_CONF_GET((in_dev), FORCE_IGMP_VERSION) == 2 || \
 138          ((in_dev)->mr_v2_seen && \
 139           time_before(jiffies, (in_dev)->mr_v2_seen)))
 140
 141 static int unsolicited_report_interval(struct in_device *in_dev)
 142 {
 143         int interval_ms, interval_jiffies;
 144
 145         if (IGMP_V1_SEEN(in_dev) || IGMP_V2_SEEN(in_dev))
 146                 interval_ms = IN_DEV_CONF_GET(
 147                         in_dev,
 148                         IGMPV2_UNSOLICITED_REPORT_INTERVAL);
 149         else /* v3 */
 150                 interval_ms = IN_DEV_CONF_GET(
 151                         in_dev,
 152                         IGMPV3_UNSOLICITED_REPORT_INTERVAL);
 153
 154         interval_jiffies = msecs_to_jiffies(interval_ms);
 155
 156         /* _timer functions can't handle a delay of 0 jiffies so ensure
 157          *  we always return a positive value.
 158          */
 159         if (interval_jiffies <= 0)
 160                 interval_jiffies = 1;
 161         return interval_jiffies;
 162 }
 163
 164 static void igmpv3_add_delrec(struct in_device *in_dev, struct ip_mc_list *im);
 165 static void igmpv3_del_delrec(struct in_device *in_dev, __be32 multiaddr);
 166 static void igmpv3_clear_delrec(struct in_device *in_dev);
 167 static int sf_setstate(struct ip_mc_list *pmc);
 168 static void sf_markstate(struct ip_mc_list *pmc);
 169 #endif
 170 static void ip_mc_clear_src(struct ip_mc_list *pmc);
 171 static int ip_mc_add_src(struct in_device *in_dev, __be32 *pmca, int sfmode,
 172                          int sfcount, __be32 *psfsrc, int delta);
 173
 174 static void ip_ma_put(struct ip_mc_list *im)
 175 {
 176         if (atomic_dec_and_test(&im->refcnt)) {
 177                 in_dev_put(im->interface);
 178                 kfree_rcu(im, rcu);
 179         }
 180 }
 181
 182 #define for_each_pmc_rcu(in_dev, pmc)                           \
 183         for (pmc = rcu_dereference(in_dev->mc_list);            \
 184              pmc != NULL;                                       \
 185              pmc = rcu_dereference(pmc->next_rcu))
 186
 187 #define for_each_pmc_rtnl(in_dev, pmc)                          \
 188         for (pmc = rtnl_dereference(in_dev->mc_list);           \
 189              pmc != NULL;                                       \
 190              pmc = rtnl_dereference(pmc->next_rcu))
 191
 192 #ifdef CONFIG_IP_MULTICAST
 193
 194 /*
 195  *      Timer management
 196  */
 197
 198 static void igmp_stop_timer(struct ip_mc_list *im)
 199 {
 200         spin_lock_bh(&im->lock);
 201         if (del_timer(&im->timer))
 202                 atomic_dec(&im->refcnt);
 203         im->tm_running = 0;
 204         im->reporter = 0;
 205         im->unsolicit_count = 0;
 206         spin_unlock_bh(&im->lock);
 207 }
 208
 209 /* It must be called with locked im->lock */
 210 static void igmp_start_timer(struct ip_mc_list *im, int max_delay)
 211 {
 212         int tv = prandom_u32() % max_delay;
 213
 214         im->tm_running = 1;
 215         if (!mod_timer(&im->timer, jiffies+tv+2))
 216                 atomic_inc(&im->refcnt);
 217 }
 218
 219 static void igmp_gq_start_timer(struct in_device *in_dev)
 220 {
 221         int tv = prandom_u32() % in_dev->mr_maxdelay;
 222
 223         in_dev->mr_gq_running = 1;
 224         if (!mod_timer(&in_dev->mr_gq_timer, jiffies+tv+2))
 225                 in_dev_hold(in_dev);
 226 }
 227
 228 static void igmp_ifc_start_timer(struct in_device *in_dev, int delay)
 229 {
 230         int tv = prandom_u32() % delay;
 231
 232         if (!mod_timer(&in_dev->mr_ifc_timer, jiffies+tv+2))
 233                 in_dev_hold(in_dev);
 234 }
 235
 236 static void igmp_mod_timer(struct ip_mc_list *im, int max_delay)
 237 {
 238         spin_lock_bh(&im->lock);
 239         im->unsolicit_count = 0;
 240         if (del_timer(&im->timer)) {
 241                 if ((long)(im->timer.expires-jiffies) < max_delay) {
 242                         add_timer(&im->timer);
 243                         im->tm_running = 1;
 244                         spin_unlock_bh(&im->lock);
 245                         return;
 246                 }
 247                 atomic_dec(&im->refcnt);
 248         }
 249         igmp_start_timer(im, max_delay);
 250         spin_unlock_bh(&im->lock);
 251 }
 252
 253
 254 /*
 255  *      Send an IGMP report.
 256  */
 257
 258 #define IGMP_SIZE (sizeof(struct igmphdr)+sizeof(struct iphdr)+4)
 259
 260
 261 static int is_in(struct ip_mc_list *pmc, struct ip_sf_list *psf, int type,
 262         int gdeleted, int sdeleted)
 263 {
 264         switch (type) {
 265         case IGMPV3_MODE_IS_INCLUDE:
 266         case IGMPV3_MODE_IS_EXCLUDE:
 267                 if (gdeleted || sdeleted)
 268                         return 0;
 269                 if (!(pmc->gsquery && !psf->sf_gsresp)) {
 270                         if (pmc->sfmode == MCAST_INCLUDE)
 271                                 return 1;
 272                         /* don't include if this source is excluded
 273                          * in all filters
 274                          */
 275                         if (psf->sf_count[MCAST_INCLUDE])
 276                                 return type == IGMPV3_MODE_IS_INCLUDE;
 277                         return pmc->sfcount[MCAST_EXCLUDE] ==
 278                                 psf->sf_count[MCAST_EXCLUDE];
 279                 }
 280                 return 0;
 281         case IGMPV3_CHANGE_TO_INCLUDE:
 282                 if (gdeleted || sdeleted)
 283                         return 0;
 284                 return psf->sf_count[MCAST_INCLUDE] != 0;
 285         case IGMPV3_CHANGE_TO_EXCLUDE:
 286                 if (gdeleted || sdeleted)
 287                         return 0;
 288                 if (pmc->sfcount[MCAST_EXCLUDE] == 0 ||
 289                     psf->sf_count[MCAST_INCLUDE])
 290                         return 0;
 291                 return pmc->sfcount[MCAST_EXCLUDE] ==
 292                         psf->sf_count[MCAST_EXCLUDE];
 293         case IGMPV3_ALLOW_NEW_SOURCES:
 294                 if (gdeleted || !psf->sf_crcount)
 295                         return 0;
 296                 return (pmc->sfmode == MCAST_INCLUDE) ^ sdeleted;
 297         case IGMPV3_BLOCK_OLD_SOURCES:
 298                 if (pmc->sfmode == MCAST_INCLUDE)
 299                         return gdeleted || (psf->sf_crcount && sdeleted);
 300                 return psf->sf_crcount && !gdeleted && !sdeleted;
 301         }
 302         return 0;
 303 }
 304
 305 static int
 306 igmp_scount(struct ip_mc_list *pmc, int type, int gdeleted, int sdeleted)
 307 {
 308         struct ip_sf_list *psf;
 309         int scount = 0;
 310
 311         for (psf = pmc->sources; psf; psf = psf->sf_next) {
 312                 if (!is_in(pmc, psf, type, gdeleted, sdeleted))
 313                         continue;
 314                 scount++;
 315         }
 316         return scount;
 317 }
 318
 319 static struct sk_buff *igmpv3_newpack(struct net_device *dev, unsigned int mtu)
 320 {
 321         struct sk_buff *skb;
 322         struct rtable *rt;
 323         struct iphdr *pip;
 324         struct igmpv3_report *pig;
 325         struct net *net = dev_net(dev);
 326         struct flowi4 fl4;
 327         int hlen = LL_RESERVED_SPACE(dev);
 328         int tlen = dev->needed_tailroom;
 329         unsigned int size = mtu;
 330
 331         while (1) {
 332                 skb = alloc_skb(size + hlen + tlen,
 333                                 GFP_ATOMIC | __GFP_NOWARN);
 334                 if (skb)
 335                         break;
 336                 size >>= 1;
 337                 if (size < 256)
 338                         return NULL;
 339         }
 340         skb->priority = TC_PRIO_CONTROL;
 341
 342         rt = ip_route_output_ports(net, &fl4, NULL, IGMPV3_ALL_MCR, 0,
 343                                    0, 0,
 344                                    IPPROTO_IGMP, 0, dev->ifindex);
 345         if (IS_ERR(rt)) {
 346                 kfree_skb(skb);
 347                 return NULL;
 348         }
 349
 350         skb_dst_set(skb, &rt->dst);
 351         skb->dev = dev;
 352
 353         skb->reserved_tailroom = skb_end_offset(skb) -
 354                                  min(mtu, skb_end_offset(skb));
 355         skb_reserve(skb, hlen);
 356
 357         skb_reset_network_header(skb);
 358         pip = ip_hdr(skb);
 359         skb_put(skb, sizeof(struct iphdr) + 4);
 360
 361         pip->version  = 4;
 362         pip->ihl      = (sizeof(struct iphdr)+4)>>2;
 363         pip->tos      = 0xc0;
 364         pip->frag_off = htons(IP_DF);
 365         pip->ttl      = 1;
 366         pip->daddr    = fl4.daddr;
 367         pip->saddr    = fl4.saddr;
 368         pip->protocol = IPPROTO_IGMP;
 369         pip->tot_len  = 0;      /* filled in later */
 370         ip_select_ident(net, skb, NULL);
 371         ((u8 *)&pip[1])[0] = IPOPT_RA;
 372         ((u8 *)&pip[1])[1] = 4;
 373         ((u8 *)&pip[1])[2] = 0;
 374         ((u8 *)&pip[1])[3] = 0;
 375
 376         skb->transport_header = skb->network_header + sizeof(struct iphdr) + 4;
 377         skb_put(skb, sizeof(*pig));
 378         pig = igmpv3_report_hdr(skb);
 379         pig->type = IGMPV3_HOST_MEMBERSHIP_REPORT;
 380         pig->resv1 = 0;
 381         pig->csum = 0;
 382         pig->resv2 = 0;
 383         pig->ngrec = 0;
 384         return skb;
 385 }
 386
 387 static int igmpv3_sendpack(struct sk_buff *skb)
 388 {
 389         struct igmphdr *pig = igmp_hdr(skb);
 390         const int igmplen = skb_tail_pointer(skb) - skb_transport_header(skb);
 391
 392         pig->csum = ip_compute_csum(igmp_hdr(skb), igmplen);
 393
 394         return ip_local_out(dev_net(skb_dst(skb)->dev), skb->sk, skb);
 395 }
 396
 397 static int grec_size(struct ip_mc_list *pmc, int type, int gdel, int sdel)
 398 {
 399         return sizeof(struct igmpv3_grec) + 4*igmp_scount(pmc, type, gdel, sdel);
 400 }
 401
 402 static struct sk_buff *add_grhead(struct sk_buff *skb, struct ip_mc_list *pmc,
 403         int type, struct igmpv3_grec **ppgr)
 404 {
 405         struct net_device *dev = pmc->interface->dev;
 406         struct igmpv3_report *pih;
 407         struct igmpv3_grec *pgr;
 408
 409         if (!skb)
 410                 skb = igmpv3_newpack(dev, dev->mtu);
 411         if (!skb)
 412                 return NULL;
 413         pgr = (struct igmpv3_grec *)skb_put(skb, sizeof(struct igmpv3_grec));
 414         pgr->grec_type = type;
 415         pgr->grec_auxwords = 0;
 416         pgr->grec_nsrcs = 0;
 417         pgr->grec_mca = pmc->multiaddr;
 418         pih = igmpv3_report_hdr(skb);
 419         pih->ngrec = htons(ntohs(pih->ngrec)+1);
 420         *ppgr = pgr;
 421         return skb;
 422 }
 423
 424 #define AVAILABLE(skb)  ((skb) ? skb_availroom(skb) : 0)
 425
 426 static struct sk_buff *add_grec(struct sk_buff *skb, struct ip_mc_list *pmc,
 427         int type, int gdeleted, int sdeleted)
 428 {
 429         struct net_device *dev = pmc->interface->dev;
 430         struct net *net = dev_net(dev);
 431         struct igmpv3_report *pih;
 432         struct igmpv3_grec *pgr = NULL;
 433         struct ip_sf_list *psf, *psf_next, *psf_prev, **psf_list;
 434         int scount, stotal, first, isquery, truncate;
 435
 436         if (pmc->multiaddr == IGMP_ALL_HOSTS)
 437                 return skb;
 438         if (ipv4_is_local_multicast(pmc->multiaddr) && !net->ipv4.sysctl_igmp_llm_reports)
 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         struct net *net = dev_net(in_dev->dev);
 542         int type;
 543
 544         if (!pmc) {
 545                 rcu_read_lock();
 546                 for_each_pmc_rcu(in_dev, pmc) {
 547                         if (pmc->multiaddr == IGMP_ALL_HOSTS)
 548                                 continue;
 549                         if (ipv4_is_local_multicast(pmc->multiaddr) &&
 550                              !net->ipv4.sysctl_igmp_llm_reports)
 551                                 continue;
 552                         spin_lock_bh(&pmc->lock);
 553                         if (pmc->sfcount[MCAST_EXCLUDE])
 554                                 type = IGMPV3_MODE_IS_EXCLUDE;
 555                         else
 556                                 type = IGMPV3_MODE_IS_INCLUDE;
 557                         skb = add_grec(skb, pmc, type, 0, 0);
 558                         spin_unlock_bh(&pmc->lock);
 559                 }
 560                 rcu_read_unlock();
 561         } else {
 562                 spin_lock_bh(&pmc->lock);
 563                 if (pmc->sfcount[MCAST_EXCLUDE])
 564                         type = IGMPV3_MODE_IS_EXCLUDE;
 565                 else
 566                         type = IGMPV3_MODE_IS_INCLUDE;
 567                 skb = add_grec(skb, pmc, type, 0, 0);
 568                 spin_unlock_bh(&pmc->lock);
 569         }
 570         if (!skb)
 571                 return 0;
 572         return igmpv3_sendpack(skb);
 573 }
 574
 575 /*
 576  * remove zero-count source records from a source filter list
 577  */
 578 static void igmpv3_clear_zeros(struct ip_sf_list **ppsf)
 579 {
 580         struct ip_sf_list *psf_prev, *psf_next, *psf;
 581
 582         psf_prev = NULL;
 583         for (psf = *ppsf; psf; psf = psf_next) {
 584                 psf_next = psf->sf_next;
 585                 if (psf->sf_crcount == 0) {
 586                         if (psf_prev)
 587                                 psf_prev->sf_next = psf->sf_next;
 588                         else
 589                                 *ppsf = psf->sf_next;
 590                         kfree(psf);
 591                 } else
 592                         psf_prev = psf;
 593         }
 594 }
 595
 596 static void igmpv3_send_cr(struct in_device *in_dev)
 597 {
 598         struct ip_mc_list *pmc, *pmc_prev, *pmc_next;
 599         struct sk_buff *skb = NULL;
 600         int type, dtype;
 601
 602         rcu_read_lock();
 603         spin_lock_bh(&in_dev->mc_tomb_lock);
 604
 605         /* deleted MCA's */
 606         pmc_prev = NULL;
 607         for (pmc = in_dev->mc_tomb; pmc; pmc = pmc_next) {
 608                 pmc_next = pmc->next;
 609                 if (pmc->sfmode == MCAST_INCLUDE) {
 610                         type = IGMPV3_BLOCK_OLD_SOURCES;
 611                         dtype = IGMPV3_BLOCK_OLD_SOURCES;
 612                         skb = add_grec(skb, pmc, type, 1, 0);
 613                         skb = add_grec(skb, pmc, dtype, 1, 1);
 614                 }
 615                 if (pmc->crcount) {
 616                         if (pmc->sfmode == MCAST_EXCLUDE) {
 617                                 type = IGMPV3_CHANGE_TO_INCLUDE;
 618                                 skb = add_grec(skb, pmc, type, 1, 0);
 619                         }
 620                         pmc->crcount--;
 621                         if (pmc->crcount == 0) {
 622                                 igmpv3_clear_zeros(&pmc->tomb);
 623                                 igmpv3_clear_zeros(&pmc->sources);
 624                         }
 625                 }
 626                 if (pmc->crcount == 0 && !pmc->tomb && !pmc->sources) {
 627                         if (pmc_prev)
 628                                 pmc_prev->next = pmc_next;
 629                         else
 630                                 in_dev->mc_tomb = pmc_next;
 631                         in_dev_put(pmc->interface);
 632                         kfree(pmc);
 633                 } else
 634                         pmc_prev = pmc;
 635         }
 636         spin_unlock_bh(&in_dev->mc_tomb_lock);
 637
 638         /* change recs */
 639         for_each_pmc_rcu(in_dev, pmc) {
 640                 spin_lock_bh(&pmc->lock);
 641                 if (pmc->sfcount[MCAST_EXCLUDE]) {
 642                         type = IGMPV3_BLOCK_OLD_SOURCES;
 643                         dtype = IGMPV3_ALLOW_NEW_SOURCES;
 644                 } else {
 645                         type = IGMPV3_ALLOW_NEW_SOURCES;
 646                         dtype = IGMPV3_BLOCK_OLD_SOURCES;
 647                 }
 648                 skb = add_grec(skb, pmc, type, 0, 0);
 649                 skb = add_grec(skb, pmc, dtype, 0, 1);  /* deleted sources */
 650
 651                 /* filter mode changes */
 652                 if (pmc->crcount) {
 653                         if (pmc->sfmode == MCAST_EXCLUDE)
 654                                 type = IGMPV3_CHANGE_TO_EXCLUDE;
 655                         else
 656                                 type = IGMPV3_CHANGE_TO_INCLUDE;
 657                         skb = add_grec(skb, pmc, type, 0, 0);
 658                         pmc->crcount--;
 659                 }
 660                 spin_unlock_bh(&pmc->lock);
 661         }
 662         rcu_read_unlock();
 663
 664         if (!skb)
 665                 return;
 666         (void) igmpv3_sendpack(skb);
 667 }
 668
 669 static int igmp_send_report(struct in_device *in_dev, struct ip_mc_list *pmc,
 670         int type)
 671 {
 672         struct sk_buff *skb;
 673         struct iphdr *iph;
 674         struct igmphdr *ih;
 675         struct rtable *rt;
 676         struct net_device *dev = in_dev->dev;
 677         struct net *net = dev_net(dev);
 678         __be32  group = pmc ? pmc->multiaddr : 0;
 679         struct flowi4 fl4;
 680         __be32  dst;
 681         int hlen, tlen;
 682
 683         if (type == IGMPV3_HOST_MEMBERSHIP_REPORT)
 684                 return igmpv3_send_report(in_dev, pmc);
 685
 686         if (ipv4_is_local_multicast(group) && !net->ipv4.sysctl_igmp_llm_reports)
 687                 return 0;
 688
 689         if (type == IGMP_HOST_LEAVE_MESSAGE)
 690                 dst = IGMP_ALL_ROUTER;
 691         else
 692                 dst = group;
 693
 694         rt = ip_route_output_ports(net, &fl4, NULL, dst, 0,
 695                                    0, 0,
 696                                    IPPROTO_IGMP, 0, dev->ifindex);
 697         if (IS_ERR(rt))
 698                 return -1;
 699
 700         hlen = LL_RESERVED_SPACE(dev);
 701         tlen = dev->needed_tailroom;
 702         skb = alloc_skb(IGMP_SIZE + hlen + tlen, GFP_ATOMIC);
 703         if (!skb) {
 704                 ip_rt_put(rt);
 705                 return -1;
 706         }
 707         skb->priority = TC_PRIO_CONTROL;
 708
 709         skb_dst_set(skb, &rt->dst);
 710
 711         skb_reserve(skb, hlen);
 712
 713         skb_reset_network_header(skb);
 714         iph = ip_hdr(skb);
 715         skb_put(skb, sizeof(struct iphdr) + 4);
 716
 717         iph->version  = 4;
 718         iph->ihl      = (sizeof(struct iphdr)+4)>>2;
 719         iph->tos      = 0xc0;
 720         iph->frag_off = htons(IP_DF);
 721         iph->ttl      = 1;
 722         iph->daddr    = dst;
 723         iph->saddr    = fl4.saddr;
 724         iph->protocol = IPPROTO_IGMP;
 725         ip_select_ident(net, skb, NULL);
 726         ((u8 *)&iph[1])[0] = IPOPT_RA;
 727         ((u8 *)&iph[1])[1] = 4;
 728         ((u8 *)&iph[1])[2] = 0;
 729         ((u8 *)&iph[1])[3] = 0;
 730
 731         ih = (struct igmphdr *)skb_put(skb, sizeof(struct igmphdr));
 732         ih->type = type;
 733         ih->code = 0;
 734         ih->csum = 0;
 735         ih->group = group;
 736         ih->csum = ip_compute_csum((void *)ih, sizeof(struct igmphdr));
 737
 738         return ip_local_out(net, skb->sk, skb);
 739 }
 740
 741 static void igmp_gq_timer_expire(unsigned long data)
 742 {
 743         struct in_device *in_dev = (struct in_device *)data;
 744
 745         in_dev->mr_gq_running = 0;
 746         igmpv3_send_report(in_dev, NULL);
 747         in_dev_put(in_dev);
 748 }
 749
 750 static void igmp_ifc_timer_expire(unsigned long data)
 751 {
 752         struct in_device *in_dev = (struct in_device *)data;
 753
 754         igmpv3_send_cr(in_dev);
 755         if (in_dev->mr_ifc_count) {
 756                 in_dev->mr_ifc_count--;
 757                 igmp_ifc_start_timer(in_dev,
 758                                      unsolicited_report_interval(in_dev));
 759         }
 760         in_dev_put(in_dev);
 761 }
 762
 763 static void igmp_ifc_event(struct in_device *in_dev)
 764 {
 765         struct net *net = dev_net(in_dev->dev);
 766         if (IGMP_V1_SEEN(in_dev) || IGMP_V2_SEEN(in_dev))
 767                 return;
 768         in_dev->mr_ifc_count = in_dev->mr_qrv ?: net->ipv4.sysctl_igmp_qrv;
 769         igmp_ifc_start_timer(in_dev, 1);
 770 }
 771
 772
 773 static void igmp_timer_expire(unsigned long data)
 774 {
 775         struct ip_mc_list *im = (struct ip_mc_list *)data;
 776         struct in_device *in_dev = im->interface;
 777
 778         spin_lock(&im->lock);
 779         im->tm_running = 0;
 780
 781         if (im->unsolicit_count) {
 782                 im->unsolicit_count--;
 783                 igmp_start_timer(im, unsolicited_report_interval(in_dev));
 784         }
 785         im->reporter = 1;
 786         spin_unlock(&im->lock);
 787
 788         if (IGMP_V1_SEEN(in_dev))
 789                 igmp_send_report(in_dev, im, IGMP_HOST_MEMBERSHIP_REPORT);
 790         else if (IGMP_V2_SEEN(in_dev))
 791                 igmp_send_report(in_dev, im, IGMPV2_HOST_MEMBERSHIP_REPORT);
 792         else
 793                 igmp_send_report(in_dev, im, IGMPV3_HOST_MEMBERSHIP_REPORT);
 794
 795         ip_ma_put(im);
 796 }
 797
 798 /* mark EXCLUDE-mode sources */
 799 static int igmp_xmarksources(struct ip_mc_list *pmc, int nsrcs, __be32 *srcs)
 800 {
 801         struct ip_sf_list *psf;
 802         int i, scount;
 803
 804         scount = 0;
 805         for (psf = pmc->sources; psf; psf = psf->sf_next) {
 806                 if (scount == nsrcs)
 807                         break;
 808                 for (i = 0; i < nsrcs; i++) {
 809                         /* skip inactive filters */
 810                         if (psf->sf_count[MCAST_INCLUDE] ||
 811                             pmc->sfcount[MCAST_EXCLUDE] !=
 812                             psf->sf_count[MCAST_EXCLUDE])
 813                                 break;
 814                         if (srcs[i] == psf->sf_inaddr) {
 815                                 scount++;
 816                                 break;
 817                         }
 818                 }
 819         }
 820         pmc->gsquery = 0;
 821         if (scount == nsrcs)    /* all sources excluded */
 822                 return 0;
 823         return 1;
 824 }
 825
 826 static int igmp_marksources(struct ip_mc_list *pmc, int nsrcs, __be32 *srcs)
 827 {
 828         struct ip_sf_list *psf;
 829         int i, scount;
 830
 831         if (pmc->sfmode == MCAST_EXCLUDE)
 832                 return igmp_xmarksources(pmc, nsrcs, srcs);
 833
 834         /* mark INCLUDE-mode sources */
 835         scount = 0;
 836         for (psf = pmc->sources; psf; psf = psf->sf_next) {
 837                 if (scount == nsrcs)
 838                         break;
 839                 for (i = 0; i < nsrcs; i++)
 840                         if (srcs[i] == psf->sf_inaddr) {
 841                                 psf->sf_gsresp = 1;
 842                                 scount++;
 843                                 break;
 844                         }
 845         }
 846         if (!scount) {
 847                 pmc->gsquery = 0;
 848                 return 0;
 849         }
 850         pmc->gsquery = 1;
 851         return 1;
 852 }
 853
 854 /* return true if packet was dropped */
 855 static bool igmp_heard_report(struct in_device *in_dev, __be32 group)
 856 {
 857         struct ip_mc_list *im;
 858         struct net *net = dev_net(in_dev->dev);
 859
 860         /* Timers are only set for non-local groups */
 861
 862         if (group == IGMP_ALL_HOSTS)
 863                 return false;
 864         if (ipv4_is_local_multicast(group) && !net->ipv4.sysctl_igmp_llm_reports)
 865                 return false;
 866
 867         rcu_read_lock();
 868         for_each_pmc_rcu(in_dev, im) {
 869                 if (im->multiaddr == group) {
 870                         igmp_stop_timer(im);
 871                         break;
 872                 }
 873         }
 874         rcu_read_unlock();
 875         return false;
 876 }
 877
 878 /* return true if packet was dropped */
 879 static bool igmp_heard_query(struct in_device *in_dev, struct sk_buff *skb,
 880         int len)
 881 {
 882         struct igmphdr          *ih = igmp_hdr(skb);
 883         struct igmpv3_query *ih3 = igmpv3_query_hdr(skb);
 884         struct ip_mc_list       *im;
 885         __be32                  group = ih->group;
 886         int                     max_delay;
 887         int                     mark = 0;
 888         struct net              *net = dev_net(in_dev->dev);
 889
 890
 891         if (len == 8) {
 892                 if (ih->code == 0) {
 893                         /* Alas, old v1 router presents here. */
 894
 895                         max_delay = IGMP_QUERY_RESPONSE_INTERVAL;
 896                         in_dev->mr_v1_seen = jiffies +
 897                                 IGMP_V1_ROUTER_PRESENT_TIMEOUT;
 898                         group = 0;
 899                 } else {
 900                         /* v2 router present */
 901                         max_delay = ih->code*(HZ/IGMP_TIMER_SCALE);
 902                         in_dev->mr_v2_seen = jiffies +
 903                                 IGMP_V2_ROUTER_PRESENT_TIMEOUT;
 904                 }
 905                 /* cancel the interface change timer */
 906                 in_dev->mr_ifc_count = 0;
 907                 if (del_timer(&in_dev->mr_ifc_timer))
 908                         __in_dev_put(in_dev);
 909                 /* clear deleted report items */
 910                 igmpv3_clear_delrec(in_dev);
 911         } else if (len < 12) {
 912                 return true;    /* ignore bogus packet; freed by caller */
 913         } else if (IGMP_V1_SEEN(in_dev)) {
 914                 /* This is a v3 query with v1 queriers present */
 915                 max_delay = IGMP_QUERY_RESPONSE_INTERVAL;
 916                 group = 0;
 917         } else if (IGMP_V2_SEEN(in_dev)) {
 918                 /* this is a v3 query with v2 queriers present;
 919                  * Interpretation of the max_delay code is problematic here.
 920                  * A real v2 host would use ih_code directly, while v3 has a
 921                  * different encoding. We use the v3 encoding as more likely
 922                  * to be intended in a v3 query.
 923                  */
 924                 max_delay = IGMPV3_MRC(ih3->code)*(HZ/IGMP_TIMER_SCALE);
 925                 if (!max_delay)
 926                         max_delay = 1;  /* can't mod w/ 0 */
 927         } else { /* v3 */
 928                 if (!pskb_may_pull(skb, sizeof(struct igmpv3_query)))
 929                         return true;
 930
 931                 ih3 = igmpv3_query_hdr(skb);
 932                 if (ih3->nsrcs) {
 933                         if (!pskb_may_pull(skb, sizeof(struct igmpv3_query)
 934                                            + ntohs(ih3->nsrcs)*sizeof(__be32)))
 935                                 return true;
 936                         ih3 = igmpv3_query_hdr(skb);
 937                 }
 938
 939                 max_delay = IGMPV3_MRC(ih3->code)*(HZ/IGMP_TIMER_SCALE);
 940                 if (!max_delay)
 941                         max_delay = 1;  /* can't mod w/ 0 */
 942                 in_dev->mr_maxdelay = max_delay;
 943                 if (ih3->qrv)
 944                         in_dev->mr_qrv = ih3->qrv;
 945                 if (!group) { /* general query */
 946                         if (ih3->nsrcs)
 947                                 return true;    /* no sources allowed */
 948                         igmp_gq_start_timer(in_dev);
 949                         return false;
 950                 }
 951                 /* mark sources to include, if group & source-specific */
 952                 mark = ih3->nsrcs != 0;
 953         }
 954
 955         /*
 956          * - Start the timers in all of our membership records
 957          *   that the query applies to for the interface on
 958          *   which the query arrived excl. those that belong
 959          *   to a "local" group (224.0.0.X)
 960          * - For timers already running check if they need to
 961          *   be reset.
 962          * - Use the igmp->igmp_code field as the maximum
 963          *   delay possible
 964          */
 965         rcu_read_lock();
 966         for_each_pmc_rcu(in_dev, im) {
 967                 int changed;
 968
 969                 if (group && group != im->multiaddr)
 970                         continue;
 971                 if (im->multiaddr == IGMP_ALL_HOSTS)
 972                         continue;
 973                 if (ipv4_is_local_multicast(im->multiaddr) &&
 974                     !net->ipv4.sysctl_igmp_llm_reports)
 975                         continue;
 976                 spin_lock_bh(&im->lock);
 977                 if (im->tm_running)
 978                         im->gsquery = im->gsquery && mark;
 979                 else
 980                         im->gsquery = mark;
 981                 changed = !im->gsquery ||
 982                         igmp_marksources(im, ntohs(ih3->nsrcs), ih3->srcs);
 983                 spin_unlock_bh(&im->lock);
 984                 if (changed)
 985                         igmp_mod_timer(im, max_delay);
 986         }
 987         rcu_read_unlock();
 988         return false;
 989 }
 990
 991 /* called in rcu_read_lock() section */
 992 int igmp_rcv(struct sk_buff *skb)
 993 {
 994         /* This basically follows the spec line by line -- see RFC1112 */
 995         struct igmphdr *ih;
 996         struct in_device *in_dev = __in_dev_get_rcu(skb->dev);
 997         int len = skb->len;
 998         bool dropped = true;
 999
1000         if (!in_dev)
1001                 goto drop;
1002
1003         if (!pskb_may_pull(skb, sizeof(struct igmphdr)))
1004                 goto drop;
1005
1006         if (skb_checksum_simple_validate(skb))
1007                 goto drop;
1008
1009         ih = igmp_hdr(skb);
1010         switch (ih->type) {
1011         case IGMP_HOST_MEMBERSHIP_QUERY:
1012                 dropped = igmp_heard_query(in_dev, skb, len);
1013                 break;
1014         case IGMP_HOST_MEMBERSHIP_REPORT:
1015         case IGMPV2_HOST_MEMBERSHIP_REPORT:
1016                 /* Is it our report looped back? */
1017                 if (rt_is_output_route(skb_rtable(skb)))
1018                         break;
1019                 /* don't rely on MC router hearing unicast reports */
1020                 if (skb->pkt_type == PACKET_MULTICAST ||
1021                     skb->pkt_type == PACKET_BROADCAST)
1022                         dropped = igmp_heard_report(in_dev, ih->group);
1023                 break;
1024         case IGMP_PIM:
1025 #ifdef CONFIG_IP_PIMSM_V1
1026                 return pim_rcv_v1(skb);
1027 #endif
1028         case IGMPV3_HOST_MEMBERSHIP_REPORT:
1029         case IGMP_DVMRP:
1030         case IGMP_TRACE:
1031         case IGMP_HOST_LEAVE_MESSAGE:
1032         case IGMP_MTRACE:
1033         case IGMP_MTRACE_RESP:
1034                 break;
1035         default:
1036                 break;
1037         }
1038
1039 drop:
1040         if (dropped)
1041                 kfree_skb(skb);
1042         else
1043                 consume_skb(skb);
1044         return 0;
1045 }
1046
1047 #endif
1048
1049
1050 /*
1051  *      Add a filter to a device
1052  */
1053
1054 static void ip_mc_filter_add(struct in_device *in_dev, __be32 addr)
1055 {
1056         char buf[MAX_ADDR_LEN];
1057         struct net_device *dev = in_dev->dev;
1058
1059         /* Checking for IFF_MULTICAST here is WRONG-WRONG-WRONG.
1060            We will get multicast token leakage, when IFF_MULTICAST
1061            is changed. This check should be done in ndo_set_rx_mode
1062            routine. Something sort of:
1063            if (dev->mc_list && dev->flags&IFF_MULTICAST) { do it; }
1064            --ANK
1065            */
1066         if (arp_mc_map(addr, buf, dev, 0) == 0)
1067                 dev_mc_add(dev, buf);
1068 }
1069
1070 /*
1071  *      Remove a filter from a device
1072  */
1073
1074 static void ip_mc_filter_del(struct in_device *in_dev, __be32 addr)
1075 {
1076         char buf[MAX_ADDR_LEN];
1077         struct net_device *dev = in_dev->dev;
1078
1079         if (arp_mc_map(addr, buf, dev, 0) == 0)
1080                 dev_mc_del(dev, buf);
1081 }
1082
1083 #ifdef CONFIG_IP_MULTICAST
1084 /*
1085  * deleted ip_mc_list manipulation
1086  */
1087 static void igmpv3_add_delrec(struct in_device *in_dev, struct ip_mc_list *im)
1088 {
1089         struct ip_mc_list *pmc;
1090         struct net *net = dev_net(in_dev->dev);
1091
1092         /* this is an "ip_mc_list" for convenience; only the fields below
1093          * are actually used. In particular, the refcnt and users are not
1094          * used for management of the delete list. Using the same structure
1095          * for deleted items allows change reports to use common code with
1096          * non-deleted or query-response MCA's.
1097          */
1098         pmc = kzalloc(sizeof(*pmc), GFP_KERNEL);
1099         if (!pmc)
1100                 return;
1101         spin_lock_bh(&im->lock);
1102         pmc->interface = im->interface;
1103         in_dev_hold(in_dev);
1104         pmc->multiaddr = im->multiaddr;
1105         pmc->crcount = in_dev->mr_qrv ?: net->ipv4.sysctl_igmp_qrv;
1106         pmc->sfmode = im->sfmode;
1107         if (pmc->sfmode == MCAST_INCLUDE) {
1108                 struct ip_sf_list *psf;
1109
1110                 pmc->tomb = im->tomb;
1111                 pmc->sources = im->sources;
1112                 im->tomb = im->sources = NULL;
1113                 for (psf = pmc->sources; psf; psf = psf->sf_next)
1114                         psf->sf_crcount = pmc->crcount;
1115         }
1116         spin_unlock_bh(&im->lock);
1117
1118         spin_lock_bh(&in_dev->mc_tomb_lock);
1119         pmc->next = in_dev->mc_tomb;
1120         in_dev->mc_tomb = pmc;
1121         spin_unlock_bh(&in_dev->mc_tomb_lock);
1122 }
1123
1124 static void igmpv3_del_delrec(struct in_device *in_dev, __be32 multiaddr)
1125 {
1126         struct ip_mc_list *pmc, *pmc_prev;
1127         struct ip_sf_list *psf, *psf_next;
1128
1129         spin_lock_bh(&in_dev->mc_tomb_lock);
1130         pmc_prev = NULL;
1131         for (pmc = in_dev->mc_tomb; pmc; pmc = pmc->next) {
1132                 if (pmc->multiaddr == multiaddr)
1133                         break;
1134                 pmc_prev = pmc;
1135         }
1136         if (pmc) {
1137                 if (pmc_prev)
1138                         pmc_prev->next = pmc->next;
1139                 else
1140                         in_dev->mc_tomb = pmc->next;
1141         }
1142         spin_unlock_bh(&in_dev->mc_tomb_lock);
1143         if (pmc) {
1144                 for (psf = pmc->tomb; psf; psf = psf_next) {
1145                         psf_next = psf->sf_next;
1146                         kfree(psf);
1147                 }
1148                 in_dev_put(pmc->interface);
1149                 kfree(pmc);
1150         }
1151 }
1152
1153 static void igmpv3_clear_delrec(struct in_device *in_dev)
1154 {
1155         struct ip_mc_list *pmc, *nextpmc;
1156
1157         spin_lock_bh(&in_dev->mc_tomb_lock);
1158         pmc = in_dev->mc_tomb;
1159         in_dev->mc_tomb = NULL;
1160         spin_unlock_bh(&in_dev->mc_tomb_lock);
1161
1162         for (; pmc; pmc = nextpmc) {
1163                 nextpmc = pmc->next;
1164                 ip_mc_clear_src(pmc);
1165                 in_dev_put(pmc->interface);
1166                 kfree(pmc);
1167         }
1168         /* clear dead sources, too */
1169         rcu_read_lock();
1170         for_each_pmc_rcu(in_dev, pmc) {
1171                 struct ip_sf_list *psf, *psf_next;
1172
1173                 spin_lock_bh(&pmc->lock);
1174                 psf = pmc->tomb;
1175                 pmc->tomb = NULL;
1176                 spin_unlock_bh(&pmc->lock);
1177                 for (; psf; psf = psf_next) {
1178                         psf_next = psf->sf_next;
1179                         kfree(psf);
1180                 }
1181         }
1182         rcu_read_unlock();
1183 }
1184 #endif
1185
1186 static void igmp_group_dropped(struct ip_mc_list *im)
1187 {
1188         struct in_device *in_dev = im->interface;
1189 #ifdef CONFIG_IP_MULTICAST
1190         struct net *net = dev_net(in_dev->dev);
1191         int reporter;
1192 #endif
1193
1194         if (im->loaded) {
1195                 im->loaded = 0;
1196                 ip_mc_filter_del(in_dev, im->multiaddr);
1197         }
1198
1199 #ifdef CONFIG_IP_MULTICAST
1200         if (im->multiaddr == IGMP_ALL_HOSTS)
1201                 return;
1202         if (ipv4_is_local_multicast(im->multiaddr) && !net->ipv4.sysctl_igmp_llm_reports)
1203                 return;
1204
1205         reporter = im->reporter;
1206         igmp_stop_timer(im);
1207
1208         if (!in_dev->dead) {
1209                 if (IGMP_V1_SEEN(in_dev))
1210                         return;
1211                 if (IGMP_V2_SEEN(in_dev)) {
1212                         if (reporter)
1213                                 igmp_send_report(in_dev, im, IGMP_HOST_LEAVE_MESSAGE);
1214                         return;
1215                 }
1216                 /* IGMPv3 */
1217                 igmpv3_add_delrec(in_dev, im);
1218
1219                 igmp_ifc_event(in_dev);
1220         }
1221 #endif
1222 }
1223
1224 static void igmp_group_added(struct ip_mc_list *im)
1225 {
1226         struct in_device *in_dev = im->interface;
1227         struct net *net = dev_net(in_dev->dev);
1228
1229         if (im->loaded == 0) {
1230                 im->loaded = 1;
1231                 ip_mc_filter_add(in_dev, im->multiaddr);
1232         }
1233
1234 #ifdef CONFIG_IP_MULTICAST
1235         if (im->multiaddr == IGMP_ALL_HOSTS)
1236                 return;
1237         if (ipv4_is_local_multicast(im->multiaddr) && !net->ipv4.sysctl_igmp_llm_reports)
1238                 return;
1239
1240         if (in_dev->dead)
1241                 return;
1242         if (IGMP_V1_SEEN(in_dev) || IGMP_V2_SEEN(in_dev)) {
1243                 spin_lock_bh(&im->lock);
1244                 igmp_start_timer(im, IGMP_INITIAL_REPORT_DELAY);
1245                 spin_unlock_bh(&im->lock);
1246                 return;
1247         }
1248         /* else, v3 */
1249
1250         im->crcount = in_dev->mr_qrv ?: net->ipv4.sysctl_igmp_qrv;
1251         igmp_ifc_event(in_dev);
1252 #endif
1253 }
1254
1255
1256 /*
1257  *      Multicast list managers
1258  */
1259
1260 static u32 ip_mc_hash(const struct ip_mc_list *im)
1261 {
1262         return hash_32((__force u32)im->multiaddr, MC_HASH_SZ_LOG);
1263 }
1264
1265 static void ip_mc_hash_add(struct in_device *in_dev,
1266                            struct ip_mc_list *im)
1267 {
1268         struct ip_mc_list __rcu **mc_hash;
1269         u32 hash;
1270
1271         mc_hash = rtnl_dereference(in_dev->mc_hash);
1272         if (mc_hash) {
1273                 hash = ip_mc_hash(im);
1274                 im->next_hash = mc_hash[hash];
1275                 rcu_assign_pointer(mc_hash[hash], im);
1276                 return;
1277         }
1278
1279         /* do not use a hash table for small number of items */
1280         if (in_dev->mc_count < 4)
1281                 return;
1282
1283         mc_hash = kzalloc(sizeof(struct ip_mc_list *) << MC_HASH_SZ_LOG,
1284                           GFP_KERNEL);
1285         if (!mc_hash)
1286                 return;
1287
1288         for_each_pmc_rtnl(in_dev, im) {
1289                 hash = ip_mc_hash(im);
1290                 im->next_hash = mc_hash[hash];
1291                 RCU_INIT_POINTER(mc_hash[hash], im);
1292         }
1293
1294         rcu_assign_pointer(in_dev->mc_hash, mc_hash);
1295 }
1296
1297 static void ip_mc_hash_remove(struct in_device *in_dev,
1298                               struct ip_mc_list *im)
1299 {
1300         struct ip_mc_list __rcu **mc_hash = rtnl_dereference(in_dev->mc_hash);
1301         struct ip_mc_list *aux;
1302
1303         if (!mc_hash)
1304                 return;
1305         mc_hash += ip_mc_hash(im);
1306         while ((aux = rtnl_dereference(*mc_hash)) != im)
1307                 mc_hash = &aux->next_hash;
1308         *mc_hash = im->next_hash;
1309 }
1310
1311
1312 /*
1313  *      A socket has joined a multicast group on device dev.
1314  */
1315
1316 void ip_mc_inc_group(struct in_device *in_dev, __be32 addr)
1317 {
1318         struct ip_mc_list *im;
1319         struct net *net = dev_net(in_dev->dev);
1320
1321         ASSERT_RTNL();
1322
1323         for_each_pmc_rtnl(in_dev, im) {
1324                 if (im->multiaddr == addr) {
1325                         im->users++;
1326                         ip_mc_add_src(in_dev, &addr, MCAST_EXCLUDE, 0, NULL, 0);
1327                         goto out;
1328                 }
1329         }
1330
1331         im = kzalloc(sizeof(*im), GFP_KERNEL);
1332         if (!im)
1333                 goto out;
1334
1335         im->users = 1;
1336         im->interface = in_dev;
1337         in_dev_hold(in_dev);
1338         im->multiaddr = addr;
1339         /* initial mode is (EX, empty) */
1340         im->sfmode = MCAST_EXCLUDE;
1341         im->sfcount[MCAST_EXCLUDE] = 1;
1342         atomic_set(&im->refcnt, 1);
1343         spin_lock_init(&im->lock);
1344 #ifdef CONFIG_IP_MULTICAST
1345         setup_timer(&im->timer, igmp_timer_expire, (unsigned long)im);
1346         im->unsolicit_count = net->ipv4.sysctl_igmp_qrv;
1347 #endif
1348
1349         im->next_rcu = in_dev->mc_list;
1350         in_dev->mc_count++;
1351         rcu_assign_pointer(in_dev->mc_list, im);
1352
1353         ip_mc_hash_add(in_dev, im);
1354
1355 #ifdef CONFIG_IP_MULTICAST
1356         igmpv3_del_delrec(in_dev, im->multiaddr);
1357 #endif
1358         igmp_group_added(im);
1359         if (!in_dev->dead)
1360                 ip_rt_multicast_event(in_dev);
1361 out:
1362         return;
1363 }
1364 EXPORT_SYMBOL(ip_mc_inc_group);
1365
1366 static int ip_mc_check_iphdr(struct sk_buff *skb)
1367 {
1368         const struct iphdr *iph;
1369         unsigned int len;
1370         unsigned int offset = skb_network_offset(skb) + sizeof(*iph);
1371
1372         if (!pskb_may_pull(skb, offset))
1373                 return -EINVAL;
1374
1375         iph = ip_hdr(skb);
1376
1377         if (iph->version != 4 || ip_hdrlen(skb) < sizeof(*iph))
1378                 return -EINVAL;
1379
1380         offset += ip_hdrlen(skb) - sizeof(*iph);
1381
1382         if (!pskb_may_pull(skb, offset))
1383                 return -EINVAL;
1384
1385         iph = ip_hdr(skb);
1386
1387         if (unlikely(ip_fast_csum((u8 *)iph, iph->ihl)))
1388                 return -EINVAL;
1389
1390         len = skb_network_offset(skb) + ntohs(iph->tot_len);
1391         if (skb->len < len || len < offset)
1392                 return -EINVAL;
1393
1394         skb_set_transport_header(skb, offset);
1395
1396         return 0;
1397 }
1398
1399 static int ip_mc_check_igmp_reportv3(struct sk_buff *skb)
1400 {
1401         unsigned int len = skb_transport_offset(skb);
1402
1403         len += sizeof(struct igmpv3_report);
1404
1405         return pskb_may_pull(skb, len) ? 0 : -EINVAL;
1406 }
1407
1408 static int ip_mc_check_igmp_query(struct sk_buff *skb)
1409 {
1410         unsigned int len = skb_transport_offset(skb);
1411
1412         len += sizeof(struct igmphdr);
1413         if (skb->len < len)
1414                 return -EINVAL;
1415
1416         /* IGMPv{1,2}? */
1417         if (skb->len != len) {
1418                 /* or IGMPv3? */
1419                 len += sizeof(struct igmpv3_query) - sizeof(struct igmphdr);
1420                 if (skb->len < len || !pskb_may_pull(skb, len))
1421                         return -EINVAL;
1422         }
1423
1424         /* RFC2236+RFC3376 (IGMPv2+IGMPv3) require the multicast link layer
1425          * all-systems destination addresses (224.0.0.1) for general queries
1426          */
1427         if (!igmp_hdr(skb)->group &&
1428             ip_hdr(skb)->daddr != htonl(INADDR_ALLHOSTS_GROUP))
1429                 return -EINVAL;
1430
1431         return 0;
1432 }
1433
1434 static int ip_mc_check_igmp_msg(struct sk_buff *skb)
1435 {
1436         switch (igmp_hdr(skb)->type) {
1437         case IGMP_HOST_LEAVE_MESSAGE:
1438         case IGMP_HOST_MEMBERSHIP_REPORT:
1439         case IGMPV2_HOST_MEMBERSHIP_REPORT:
1440                 /* fall through */
1441                 return 0;
1442         case IGMPV3_HOST_MEMBERSHIP_REPORT:
1443                 return ip_mc_check_igmp_reportv3(skb);
1444         case IGMP_HOST_MEMBERSHIP_QUERY:
1445                 return ip_mc_check_igmp_query(skb);
1446         default:
1447                 return -ENOMSG;
1448         }
1449 }
1450
1451 static inline __sum16 ip_mc_validate_checksum(struct sk_buff *skb)
1452 {
1453         return skb_checksum_simple_validate(skb);
1454 }
1455
1456 static int __ip_mc_check_igmp(struct sk_buff *skb, struct sk_buff **skb_trimmed)
1457
1458 {
1459         struct sk_buff *skb_chk;
1460         unsigned int transport_len;
1461         unsigned int len = skb_transport_offset(skb) + sizeof(struct igmphdr);
1462         int ret = -EINVAL;
1463
1464         transport_len = ntohs(ip_hdr(skb)->tot_len) - ip_hdrlen(skb);
1465
1466         skb_chk = skb_checksum_trimmed(skb, transport_len,
1467                                        ip_mc_validate_checksum);
1468         if (!skb_chk)
1469                 goto err;
1470
1471         if (!pskb_may_pull(skb_chk, len))
1472                 goto err;
1473
1474         ret = ip_mc_check_igmp_msg(skb_chk);
1475         if (ret)
1476                 goto err;
1477
1478         if (skb_trimmed)
1479                 *skb_trimmed = skb_chk;
1480         /* free now unneeded clone */
1481         else if (skb_chk != skb)
1482                 kfree_skb(skb_chk);
1483
1484         ret = 0;
1485
1486 err:
1487         if (ret && skb_chk && skb_chk != skb)
1488                 kfree_skb(skb_chk);
1489
1490         return ret;
1491 }
1492
1493 /**
1494  * ip_mc_check_igmp - checks whether this is a sane IGMP packet
1495  * @skb: the skb to validate
1496  * @skb_trimmed: to store an skb pointer trimmed to IPv4 packet tail (optional)
1497  *
1498  * Checks whether an IPv4 packet is a valid IGMP packet. If so sets
1499  * skb transport header accordingly and returns zero.
1500  *
1501  * -EINVAL: A broken packet was detected, i.e. it violates some internet
1502  *  standard
1503  * -ENOMSG: IP header validation succeeded but it is not an IGMP packet.
1504  * -ENOMEM: A memory allocation failure happened.
1505  *
1506  * Optionally, an skb pointer might be provided via skb_trimmed (or set it
1507  * to NULL): After parsing an IGMP packet successfully it will point to
1508  * an skb which has its tail aligned to the IP packet end. This might
1509  * either be the originally provided skb or a trimmed, cloned version if
1510  * the skb frame had data beyond the IP packet. A cloned skb allows us
1511  * to leave the original skb and its full frame unchanged (which might be
1512  * desirable for layer 2 frame jugglers).
1513  *
1514  * Caller needs to set the skb network header and free any returned skb if it
1515  * differs from the provided skb.
1516  */
1517 int ip_mc_check_igmp(struct sk_buff *skb, struct sk_buff **skb_trimmed)
1518 {
1519         int ret = ip_mc_check_iphdr(skb);
1520
1521         if (ret < 0)
1522                 return ret;
1523
1524         if (ip_hdr(skb)->protocol != IPPROTO_IGMP)
1525                 return -ENOMSG;
1526
1527         return __ip_mc_check_igmp(skb, skb_trimmed);
1528 }
1529 EXPORT_SYMBOL(ip_mc_check_igmp);
1530
1531 /*
1532  *      Resend IGMP JOIN report; used by netdev notifier.
1533  */
1534 static void ip_mc_rejoin_groups(struct in_device *in_dev)
1535 {
1536 #ifdef CONFIG_IP_MULTICAST
1537         struct ip_mc_list *im;
1538         int type;
1539         struct net *net = dev_net(in_dev->dev);
1540
1541         ASSERT_RTNL();
1542
1543         for_each_pmc_rtnl(in_dev, im) {
1544                 if (im->multiaddr == IGMP_ALL_HOSTS)
1545                         continue;
1546                 if (ipv4_is_local_multicast(im->multiaddr) &&
1547                     !net->ipv4.sysctl_igmp_llm_reports)
1548                         continue;
1549
1550                 /* a failover is happening and switches
1551                  * must be notified immediately
1552                  */
1553                 if (IGMP_V1_SEEN(in_dev))
1554                         type = IGMP_HOST_MEMBERSHIP_REPORT;
1555                 else if (IGMP_V2_SEEN(in_dev))
1556                         type = IGMPV2_HOST_MEMBERSHIP_REPORT;
1557                 else
1558                         type = IGMPV3_HOST_MEMBERSHIP_REPORT;
1559                 igmp_send_report(in_dev, im, type);
1560         }
1561 #endif
1562 }
1563
1564 /*
1565  *      A socket has left a multicast group on device dev
1566  */
1567
1568 void ip_mc_dec_group(struct in_device *in_dev, __be32 addr)
1569 {
1570         struct ip_mc_list *i;
1571         struct ip_mc_list __rcu **ip;
1572
1573         ASSERT_RTNL();
1574
1575         for (ip = &in_dev->mc_list;
1576              (i = rtnl_dereference(*ip)) != NULL;
1577              ip = &i->next_rcu) {
1578                 if (i->multiaddr == addr) {
1579                         if (--i->users == 0) {
1580                                 ip_mc_hash_remove(in_dev, i);
1581                                 *ip = i->next_rcu;
1582                                 in_dev->mc_count--;
1583                                 igmp_group_dropped(i);
1584                                 ip_mc_clear_src(i);
1585
1586                                 if (!in_dev->dead)
1587                                         ip_rt_multicast_event(in_dev);
1588
1589                                 ip_ma_put(i);
1590                                 return;
1591                         }
1592                         break;
1593                 }
1594         }
1595 }
1596 EXPORT_SYMBOL(ip_mc_dec_group);
1597
1598 /* Device changing type */
1599
1600 void ip_mc_unmap(struct in_device *in_dev)
1601 {
1602         struct ip_mc_list *pmc;
1603
1604         ASSERT_RTNL();
1605
1606         for_each_pmc_rtnl(in_dev, pmc)
1607                 igmp_group_dropped(pmc);
1608 }
1609
1610 void ip_mc_remap(struct in_device *in_dev)
1611 {
1612         struct ip_mc_list *pmc;
1613
1614         ASSERT_RTNL();
1615
1616         for_each_pmc_rtnl(in_dev, pmc)
1617                 igmp_group_added(pmc);
1618 }
1619
1620 /* Device going down */
1621
1622 void ip_mc_down(struct in_device *in_dev)
1623 {
1624         struct ip_mc_list *pmc;
1625
1626         ASSERT_RTNL();
1627
1628         for_each_pmc_rtnl(in_dev, pmc)
1629                 igmp_group_dropped(pmc);
1630
1631 #ifdef CONFIG_IP_MULTICAST
1632         in_dev->mr_ifc_count = 0;
1633         if (del_timer(&in_dev->mr_ifc_timer))
1634                 __in_dev_put(in_dev);
1635         in_dev->mr_gq_running = 0;
1636         if (del_timer(&in_dev->mr_gq_timer))
1637                 __in_dev_put(in_dev);
1638         igmpv3_clear_delrec(in_dev);
1639 #endif
1640
1641         ip_mc_dec_group(in_dev, IGMP_ALL_HOSTS);
1642 }
1643
1644 void ip_mc_init_dev(struct in_device *in_dev)
1645 {
1646         struct net *net = dev_net(in_dev->dev);
1647         ASSERT_RTNL();
1648
1649 #ifdef CONFIG_IP_MULTICAST
1650         setup_timer(&in_dev->mr_gq_timer, igmp_gq_timer_expire,
1651                         (unsigned long)in_dev);
1652         setup_timer(&in_dev->mr_ifc_timer, igmp_ifc_timer_expire,
1653                         (unsigned long)in_dev);
1654         in_dev->mr_qrv = net->ipv4.sysctl_igmp_qrv;
1655 #endif
1656
1657         spin_lock_init(&in_dev->mc_tomb_lock);
1658 }
1659
1660 /* Device going up */
1661
1662 void ip_mc_up(struct in_device *in_dev)
1663 {
1664         struct ip_mc_list *pmc;
1665         struct net *net = dev_net(in_dev->dev);
1666
1667         ASSERT_RTNL();
1668
1669 #ifdef CONFIG_IP_MULTICAST
1670         in_dev->mr_qrv = net->ipv4.sysctl_igmp_qrv;
1671 #endif
1672         ip_mc_inc_group(in_dev, IGMP_ALL_HOSTS);
1673
1674         for_each_pmc_rtnl(in_dev, pmc)
1675                 igmp_group_added(pmc);
1676 }
1677
1678 /*
1679  *      Device is about to be destroyed: clean up.
1680  */
1681
1682 void ip_mc_destroy_dev(struct in_device *in_dev)
1683 {
1684         struct ip_mc_list *i;
1685
1686         ASSERT_RTNL();
1687
1688         /* Deactivate timers */
1689         ip_mc_down(in_dev);
1690
1691         while ((i = rtnl_dereference(in_dev->mc_list)) != NULL) {
1692                 in_dev->mc_list = i->next_rcu;
1693                 in_dev->mc_count--;
1694
1695                 /* We've dropped the groups in ip_mc_down already */
1696                 ip_mc_clear_src(i);
1697                 ip_ma_put(i);
1698         }
1699 }
1700
1701 /* RTNL is locked */
1702 static struct in_device *ip_mc_find_dev(struct net *net, struct ip_mreqn *imr)
1703 {
1704         struct net_device *dev = NULL;
1705         struct in_device *idev = NULL;
1706
1707         if (imr->imr_ifindex) {
1708                 idev = inetdev_by_index(net, imr->imr_ifindex);
1709                 return idev;
1710         }
1711         if (imr->imr_address.s_addr) {
1712                 dev = __ip_dev_find(net, imr->imr_address.s_addr, false);
1713                 if (!dev)
1714                         return NULL;
1715         }
1716
1717         if (!dev) {
1718                 struct rtable *rt = ip_route_output(net,
1719                                                     imr->imr_multiaddr.s_addr,
1720                                                     0, 0, 0);
1721                 if (!IS_ERR(rt)) {
1722                         dev = rt->dst.dev;
1723                         ip_rt_put(rt);
1724                 }
1725         }
1726         if (dev) {
1727                 imr->imr_ifindex = dev->ifindex;
1728                 idev = __in_dev_get_rtnl(dev);
1729         }
1730         return idev;
1731 }
1732
1733 /*
1734  *      Join a socket to a group
1735  */
1736
1737 static int ip_mc_del1_src(struct ip_mc_list *pmc, int sfmode,
1738         __be32 *psfsrc)
1739 {
1740         struct ip_sf_list *psf, *psf_prev;
1741         int rv = 0;
1742
1743         psf_prev = NULL;
1744         for (psf = pmc->sources; psf; psf = psf->sf_next) {
1745                 if (psf->sf_inaddr == *psfsrc)
1746                         break;
1747                 psf_prev = psf;
1748         }
1749         if (!psf || psf->sf_count[sfmode] == 0) {
1750                 /* source filter not found, or count wrong =>  bug */
1751                 return -ESRCH;
1752         }
1753         psf->sf_count[sfmode]--;
1754         if (psf->sf_count[sfmode] == 0) {
1755                 ip_rt_multicast_event(pmc->interface);
1756         }
1757         if (!psf->sf_count[MCAST_INCLUDE] && !psf->sf_count[MCAST_EXCLUDE]) {
1758 #ifdef CONFIG_IP_MULTICAST
1759                 struct in_device *in_dev = pmc->interface;
1760                 struct net *net = dev_net(in_dev->dev);
1761 #endif
1762
1763                 /* no more filters for this source */
1764                 if (psf_prev)
1765                         psf_prev->sf_next = psf->sf_next;
1766                 else
1767                         pmc->sources = psf->sf_next;
1768 #ifdef CONFIG_IP_MULTICAST
1769                 if (psf->sf_oldin &&
1770                     !IGMP_V1_SEEN(in_dev) && !IGMP_V2_SEEN(in_dev)) {
1771                         psf->sf_crcount = in_dev->mr_qrv ?: net->ipv4.sysctl_igmp_qrv;
1772                         psf->sf_next = pmc->tomb;
1773                         pmc->tomb = psf;
1774                         rv = 1;
1775                 } else
1776 #endif
1777                         kfree(psf);
1778         }
1779         return rv;
1780 }
1781
1782 #ifndef CONFIG_IP_MULTICAST
1783 #define igmp_ifc_event(x)       do { } while (0)
1784 #endif
1785
1786 static int ip_mc_del_src(struct in_device *in_dev, __be32 *pmca, int sfmode,
1787                          int sfcount, __be32 *psfsrc, int delta)
1788 {
1789         struct ip_mc_list *pmc;
1790         int     changerec = 0;
1791         int     i, err;
1792
1793         if (!in_dev)
1794                 return -ENODEV;
1795         rcu_read_lock();
1796         for_each_pmc_rcu(in_dev, pmc) {
1797                 if (*pmca == pmc->multiaddr)
1798                         break;
1799         }
1800         if (!pmc) {
1801                 /* MCA not found?? bug */
1802                 rcu_read_unlock();
1803                 return -ESRCH;
1804         }
1805         spin_lock_bh(&pmc->lock);
1806         rcu_read_unlock();
1807 #ifdef CONFIG_IP_MULTICAST
1808         sf_markstate(pmc);
1809 #endif
1810         if (!delta) {
1811                 err = -EINVAL;
1812                 if (!pmc->sfcount[sfmode])
1813                         goto out_unlock;
1814                 pmc->sfcount[sfmode]--;
1815         }
1816         err = 0;
1817         for (i = 0; i < sfcount; i++) {
1818                 int rv = ip_mc_del1_src(pmc, sfmode, &psfsrc[i]);
1819
1820                 changerec |= rv > 0;
1821                 if (!err && rv < 0)
1822                         err = rv;
1823         }
1824         if (pmc->sfmode == MCAST_EXCLUDE &&
1825             pmc->sfcount[MCAST_EXCLUDE] == 0 &&
1826             pmc->sfcount[MCAST_INCLUDE]) {
1827 #ifdef CONFIG_IP_MULTICAST
1828                 struct ip_sf_list *psf;
1829                 struct net *net = dev_net(in_dev->dev);
1830 #endif
1831
1832                 /* filter mode change */
1833                 pmc->sfmode = MCAST_INCLUDE;
1834 #ifdef CONFIG_IP_MULTICAST
1835                 pmc->crcount = in_dev->mr_qrv ?: net->ipv4.sysctl_igmp_qrv;
1836                 in_dev->mr_ifc_count = pmc->crcount;
1837                 for (psf = pmc->sources; psf; psf = psf->sf_next)
1838                         psf->sf_crcount = 0;
1839                 igmp_ifc_event(pmc->interface);
1840         } else if (sf_setstate(pmc) || changerec) {
1841                 igmp_ifc_event(pmc->interface);
1842 #endif
1843         }
1844 out_unlock:
1845         spin_unlock_bh(&pmc->lock);
1846         return err;
1847 }
1848
1849 /*
1850  * Add multicast single-source filter to the interface list
1851  */
1852 static int ip_mc_add1_src(struct ip_mc_list *pmc, int sfmode,
1853         __be32 *psfsrc)
1854 {
1855         struct ip_sf_list *psf, *psf_prev;
1856
1857         psf_prev = NULL;
1858         for (psf = pmc->sources; psf; psf = psf->sf_next) {
1859                 if (psf->sf_inaddr == *psfsrc)
1860                         break;
1861                 psf_prev = psf;
1862         }
1863         if (!psf) {
1864                 psf = kzalloc(sizeof(*psf), GFP_ATOMIC);
1865                 if (!psf)
1866                         return -ENOBUFS;
1867                 psf->sf_inaddr = *psfsrc;
1868                 if (psf_prev) {
1869                         psf_prev->sf_next = psf;
1870                 } else
1871                         pmc->sources = psf;
1872         }
1873         psf->sf_count[sfmode]++;
1874         if (psf->sf_count[sfmode] == 1) {
1875                 ip_rt_multicast_event(pmc->interface);
1876         }
1877         return 0;
1878 }
1879
1880 #ifdef CONFIG_IP_MULTICAST
1881 static void sf_markstate(struct ip_mc_list *pmc)
1882 {
1883         struct ip_sf_list *psf;
1884         int mca_xcount = pmc->sfcount[MCAST_EXCLUDE];
1885
1886         for (psf = pmc->sources; psf; psf = psf->sf_next)
1887                 if (pmc->sfcount[MCAST_EXCLUDE]) {
1888                         psf->sf_oldin = mca_xcount ==
1889                                 psf->sf_count[MCAST_EXCLUDE] &&
1890                                 !psf->sf_count[MCAST_INCLUDE];
1891                 } else
1892                         psf->sf_oldin = psf->sf_count[MCAST_INCLUDE] != 0;
1893 }
1894
1895 static int sf_setstate(struct ip_mc_list *pmc)
1896 {
1897         struct ip_sf_list *psf, *dpsf;
1898         int mca_xcount = pmc->sfcount[MCAST_EXCLUDE];
1899         int qrv = pmc->interface->mr_qrv;
1900         int new_in, rv;
1901
1902         rv = 0;
1903         for (psf = pmc->sources; psf; psf = psf->sf_next) {
1904                 if (pmc->sfcount[MCAST_EXCLUDE]) {
1905                         new_in = mca_xcount == psf->sf_count[MCAST_EXCLUDE] &&
1906                                 !psf->sf_count[MCAST_INCLUDE];
1907                 } else
1908                         new_in = psf->sf_count[MCAST_INCLUDE] != 0;
1909                 if (new_in) {
1910                         if (!psf->sf_oldin) {
1911                                 struct ip_sf_list *prev = NULL;
1912
1913                                 for (dpsf = pmc->tomb; dpsf; dpsf = dpsf->sf_next) {
1914                                         if (dpsf->sf_inaddr == psf->sf_inaddr)
1915                                                 break;
1916                                         prev = dpsf;
1917                                 }
1918                                 if (dpsf) {
1919                                         if (prev)
1920                                                 prev->sf_next = dpsf->sf_next;
1921                                         else
1922                                                 pmc->tomb = dpsf->sf_next;
1923                                         kfree(dpsf);
1924                                 }
1925                                 psf->sf_crcount = qrv;
1926                                 rv++;
1927                         }
1928                 } else if (psf->sf_oldin) {
1929
1930                         psf->sf_crcount = 0;
1931                         /*
1932                          * add or update "delete" records if an active filter
1933                          * is now inactive
1934                          */
1935                         for (dpsf = pmc->tomb; dpsf; dpsf = dpsf->sf_next)
1936                                 if (dpsf->sf_inaddr == psf->sf_inaddr)
1937                                         break;
1938                         if (!dpsf) {
1939                                 dpsf = kmalloc(sizeof(*dpsf), GFP_ATOMIC);
1940                                 if (!dpsf)
1941                                         continue;
1942                                 *dpsf = *psf;
1943                                 /* pmc->lock held by callers */
1944                                 dpsf->sf_next = pmc->tomb;
1945                                 pmc->tomb = dpsf;
1946                         }
1947                         dpsf->sf_crcount = qrv;
1948                         rv++;
1949                 }
1950         }
1951         return rv;
1952 }
1953 #endif
1954
1955 /*
1956  * Add multicast source filter list to the interface list
1957  */
1958 static int ip_mc_add_src(struct in_device *in_dev, __be32 *pmca, int sfmode,
1959                          int sfcount, __be32 *psfsrc, int delta)
1960 {
1961         struct ip_mc_list *pmc;
1962         int     isexclude;
1963         int     i, err;
1964
1965         if (!in_dev)
1966                 return -ENODEV;
1967         rcu_read_lock();
1968         for_each_pmc_rcu(in_dev, pmc) {
1969                 if (*pmca == pmc->multiaddr)
1970                         break;
1971         }
1972         if (!pmc) {
1973                 /* MCA not found?? bug */
1974                 rcu_read_unlock();
1975                 return -ESRCH;
1976         }
1977         spin_lock_bh(&pmc->lock);
1978         rcu_read_unlock();
1979
1980 #ifdef CONFIG_IP_MULTICAST
1981         sf_markstate(pmc);
1982 #endif
1983         isexclude = pmc->sfmode == MCAST_EXCLUDE;
1984         if (!delta)
1985                 pmc->sfcount[sfmode]++;
1986         err = 0;
1987         for (i = 0; i < sfcount; i++) {
1988                 err = ip_mc_add1_src(pmc, sfmode, &psfsrc[i]);
1989                 if (err)
1990                         break;
1991         }
1992         if (err) {
1993                 int j;
1994
1995                 if (!delta)
1996                         pmc->sfcount[sfmode]--;
1997                 for (j = 0; j < i; j++)
1998                         (void) ip_mc_del1_src(pmc, sfmode, &psfsrc[j]);
1999         } else if (isexclude != (pmc->sfcount[MCAST_EXCLUDE] != 0)) {
2000 #ifdef CONFIG_IP_MULTICAST
2001                 struct ip_sf_list *psf;
2002                 struct net *net = dev_net(pmc->interface->dev);
2003                 in_dev = pmc->interface;
2004 #endif
2005
2006                 /* filter mode change */
2007                 if (pmc->sfcount[MCAST_EXCLUDE])
2008                         pmc->sfmode = MCAST_EXCLUDE;
2009                 else if (pmc->sfcount[MCAST_INCLUDE])
2010                         pmc->sfmode = MCAST_INCLUDE;
2011 #ifdef CONFIG_IP_MULTICAST
2012                 /* else no filters; keep old mode for reports */
2013
2014                 pmc->crcount = in_dev->mr_qrv ?: net->ipv4.sysctl_igmp_qrv;
2015                 in_dev->mr_ifc_count = pmc->crcount;
2016                 for (psf = pmc->sources; psf; psf = psf->sf_next)
2017                         psf->sf_crcount = 0;
2018                 igmp_ifc_event(in_dev);
2019         } else if (sf_setstate(pmc)) {
2020                 igmp_ifc_event(in_dev);
2021 #endif
2022         }
2023         spin_unlock_bh(&pmc->lock);
2024         return err;
2025 }
2026
2027 static void ip_mc_clear_src(struct ip_mc_list *pmc)
2028 {
2029         struct ip_sf_list *psf, *nextpsf;
2030
2031         for (psf = pmc->tomb; psf; psf = nextpsf) {
2032                 nextpsf = psf->sf_next;
2033                 kfree(psf);
2034         }
2035         pmc->tomb = NULL;
2036         for (psf = pmc->sources; psf; psf = nextpsf) {
2037                 nextpsf = psf->sf_next;
2038                 kfree(psf);
2039         }
2040         pmc->sources = NULL;
2041         pmc->sfmode = MCAST_EXCLUDE;
2042         pmc->sfcount[MCAST_INCLUDE] = 0;
2043         pmc->sfcount[MCAST_EXCLUDE] = 1;
2044 }
2045
2046 /* Join a multicast group
2047  */
2048
2049 int ip_mc_join_group(struct sock *sk, struct ip_mreqn *imr)
2050 {
2051         __be32 addr = imr->imr_multiaddr.s_addr;
2052         struct ip_mc_socklist *iml, *i;
2053         struct in_device *in_dev;
2054         struct inet_sock *inet = inet_sk(sk);
2055         struct net *net = sock_net(sk);
2056         int ifindex;
2057         int count = 0;
2058         int err;
2059
2060         ASSERT_RTNL();
2061
2062         if (!ipv4_is_multicast(addr))
2063                 return -EINVAL;
2064
2065         in_dev = ip_mc_find_dev(net, imr);
2066
2067         if (!in_dev) {
2068                 err = -ENODEV;
2069                 goto done;
2070         }
2071
2072         err = -EADDRINUSE;
2073         ifindex = imr->imr_ifindex;
2074         for_each_pmc_rtnl(inet, i) {
2075                 if (i->multi.imr_multiaddr.s_addr == addr &&
2076                     i->multi.imr_ifindex == ifindex)
2077                         goto done;
2078                 count++;
2079         }
2080         err = -ENOBUFS;
2081         if (count >= net->ipv4.sysctl_igmp_max_memberships)
2082                 goto done;
2083         iml = sock_kmalloc(sk, sizeof(*iml), GFP_KERNEL);
2084         if (!iml)
2085                 goto done;
2086
2087         memcpy(&iml->multi, imr, sizeof(*imr));
2088         iml->next_rcu = inet->mc_list;
2089         iml->sflist = NULL;
2090         iml->sfmode = MCAST_EXCLUDE;
2091         rcu_assign_pointer(inet->mc_list, iml);
2092         ip_mc_inc_group(in_dev, addr);
2093         err = 0;
2094 done:
2095         return err;
2096 }
2097 EXPORT_SYMBOL(ip_mc_join_group);
2098
2099 static int ip_mc_leave_src(struct sock *sk, struct ip_mc_socklist *iml,
2100                            struct in_device *in_dev)
2101 {
2102         struct ip_sf_socklist *psf = rtnl_dereference(iml->sflist);
2103         int err;
2104
2105         if (!psf) {
2106                 /* any-source empty exclude case */
2107                 return ip_mc_del_src(in_dev, &iml->multi.imr_multiaddr.s_addr,
2108                         iml->sfmode, 0, NULL, 0);
2109         }
2110         err = ip_mc_del_src(in_dev, &iml->multi.imr_multiaddr.s_addr,
2111                         iml->sfmode, psf->sl_count, psf->sl_addr, 0);
2112         RCU_INIT_POINTER(iml->sflist, NULL);
2113         /* decrease mem now to avoid the memleak warning */
2114         atomic_sub(IP_SFLSIZE(psf->sl_max), &sk->sk_omem_alloc);
2115         kfree_rcu(psf, rcu);
2116         return err;
2117 }
2118
2119 int ip_mc_leave_group(struct sock *sk, struct ip_mreqn *imr)
2120 {
2121         struct inet_sock *inet = inet_sk(sk);
2122         struct ip_mc_socklist *iml;
2123         struct ip_mc_socklist __rcu **imlp;
2124         struct in_device *in_dev;
2125         struct net *net = sock_net(sk);
2126         __be32 group = imr->imr_multiaddr.s_addr;
2127         u32 ifindex;
2128         int ret = -EADDRNOTAVAIL;
2129
2130         ASSERT_RTNL();
2131
2132         in_dev = ip_mc_find_dev(net, imr);
2133         if (!imr->imr_ifindex && !imr->imr_address.s_addr && !in_dev) {
2134                 ret = -ENODEV;
2135                 goto out;
2136         }
2137         ifindex = imr->imr_ifindex;
2138         for (imlp = &inet->mc_list;
2139              (iml = rtnl_dereference(*imlp)) != NULL;
2140              imlp = &iml->next_rcu) {
2141                 if (iml->multi.imr_multiaddr.s_addr != group)
2142                         continue;
2143                 if (ifindex) {
2144                         if (iml->multi.imr_ifindex != ifindex)
2145                                 continue;
2146                 } else if (imr->imr_address.s_addr && imr->imr_address.s_addr !=
2147                                 iml->multi.imr_address.s_addr)
2148                         continue;
2149
2150                 (void) ip_mc_leave_src(sk, iml, in_dev);
2151
2152                 *imlp = iml->next_rcu;
2153
2154                 if (in_dev)
2155                         ip_mc_dec_group(in_dev, group);
2156
2157                 /* decrease mem now to avoid the memleak warning */
2158                 atomic_sub(sizeof(*iml), &sk->sk_omem_alloc);
2159                 kfree_rcu(iml, rcu);
2160                 return 0;
2161         }
2162 out:
2163         return ret;
2164 }
2165 EXPORT_SYMBOL(ip_mc_leave_group);
2166
2167 int ip_mc_source(int add, int omode, struct sock *sk, struct
2168         ip_mreq_source *mreqs, int ifindex)
2169 {
2170         int err;
2171         struct ip_mreqn imr;
2172         __be32 addr = mreqs->imr_multiaddr;
2173         struct ip_mc_socklist *pmc;
2174         struct in_device *in_dev = NULL;
2175         struct inet_sock *inet = inet_sk(sk);
2176         struct ip_sf_socklist *psl;
2177         struct net *net = sock_net(sk);
2178         int leavegroup = 0;
2179         int i, j, rv;
2180
2181         if (!ipv4_is_multicast(addr))
2182                 return -EINVAL;
2183
2184         ASSERT_RTNL();
2185
2186         imr.imr_multiaddr.s_addr = mreqs->imr_multiaddr;
2187         imr.imr_address.s_addr = mreqs->imr_interface;
2188         imr.imr_ifindex = ifindex;
2189         in_dev = ip_mc_find_dev(net, &imr);
2190
2191         if (!in_dev) {
2192                 err = -ENODEV;
2193                 goto done;
2194         }
2195         err = -EADDRNOTAVAIL;
2196
2197         for_each_pmc_rtnl(inet, pmc) {
2198                 if ((pmc->multi.imr_multiaddr.s_addr ==
2199                      imr.imr_multiaddr.s_addr) &&
2200                     (pmc->multi.imr_ifindex == imr.imr_ifindex))
2201                         break;
2202         }
2203         if (!pmc) {             /* must have a prior join */
2204                 err = -EINVAL;
2205                 goto done;
2206         }
2207         /* if a source filter was set, must be the same mode as before */
2208         if (pmc->sflist) {
2209                 if (pmc->sfmode != omode) {
2210                         err = -EINVAL;
2211                         goto done;
2212                 }
2213         } else if (pmc->sfmode != omode) {
2214                 /* allow mode switches for empty-set filters */
2215                 ip_mc_add_src(in_dev, &mreqs->imr_multiaddr, omode, 0, NULL, 0);
2216                 ip_mc_del_src(in_dev, &mreqs->imr_multiaddr, pmc->sfmode, 0,
2217                         NULL, 0);
2218                 pmc->sfmode = omode;
2219         }
2220
2221         psl = rtnl_dereference(pmc->sflist);
2222         if (!add) {
2223                 if (!psl)
2224                         goto done;      /* err = -EADDRNOTAVAIL */
2225                 rv = !0;
2226                 for (i = 0; i < psl->sl_count; i++) {
2227                         rv = memcmp(&psl->sl_addr[i], &mreqs->imr_sourceaddr,
2228                                 sizeof(__be32));
2229                         if (rv == 0)
2230                                 break;
2231                 }
2232                 if (rv)         /* source not found */
2233                         goto done;      /* err = -EADDRNOTAVAIL */
2234
2235                 /* special case - (INCLUDE, empty) == LEAVE_GROUP */
2236                 if (psl->sl_count == 1 && omode == MCAST_INCLUDE) {
2237                         leavegroup = 1;
2238                         goto done;
2239                 }
2240
2241                 /* update the interface filter */
2242                 ip_mc_del_src(in_dev, &mreqs->imr_multiaddr, omode, 1,
2243                         &mreqs->imr_sourceaddr, 1);
2244
2245                 for (j = i+1; j < psl->sl_count; j++)
2246                         psl->sl_addr[j-1] = psl->sl_addr[j];
2247                 psl->sl_count--;
2248                 err = 0;
2249                 goto done;
2250         }
2251         /* else, add a new source to the filter */
2252
2253         if (psl && psl->sl_count >= net->ipv4.sysctl_igmp_max_msf) {
2254                 err = -ENOBUFS;
2255                 goto done;
2256         }
2257         if (!psl || psl->sl_count == psl->sl_max) {
2258                 struct ip_sf_socklist *newpsl;
2259                 int count = IP_SFBLOCK;
2260
2261                 if (psl)
2262                         count += psl->sl_max;
2263                 newpsl = sock_kmalloc(sk, IP_SFLSIZE(count), GFP_KERNEL);
2264                 if (!newpsl) {
2265                         err = -ENOBUFS;
2266                         goto done;
2267                 }
2268                 newpsl->sl_max = count;
2269                 newpsl->sl_count = count - IP_SFBLOCK;
2270                 if (psl) {
2271                         for (i = 0; i < psl->sl_count; i++)
2272                                 newpsl->sl_addr[i] = psl->sl_addr[i];
2273                         /* decrease mem now to avoid the memleak warning */
2274                         atomic_sub(IP_SFLSIZE(psl->sl_max), &sk->sk_omem_alloc);
2275                         kfree_rcu(psl, rcu);
2276                 }
2277                 rcu_assign_pointer(pmc->sflist, newpsl);
2278                 psl = newpsl;
2279         }
2280         rv = 1; /* > 0 for insert logic below if sl_count is 0 */
2281         for (i = 0; i < psl->sl_count; i++) {
2282                 rv = memcmp(&psl->sl_addr[i], &mreqs->imr_sourceaddr,
2283                         sizeof(__be32));
2284                 if (rv == 0)
2285                         break;
2286         }
2287         if (rv == 0)            /* address already there is an error */
2288                 goto done;
2289         for (j = psl->sl_count-1; j >= i; j--)
2290                 psl->sl_addr[j+1] = psl->sl_addr[j];
2291         psl->sl_addr[i] = mreqs->imr_sourceaddr;
2292         psl->sl_count++;
2293         err = 0;
2294         /* update the interface list */
2295         ip_mc_add_src(in_dev, &mreqs->imr_multiaddr, omode, 1,
2296                 &mreqs->imr_sourceaddr, 1);
2297 done:
2298         if (leavegroup)
2299                 err = ip_mc_leave_group(sk, &imr);
2300         return err;
2301 }
2302
2303 int ip_mc_msfilter(struct sock *sk, struct ip_msfilter *msf, int ifindex)
2304 {
2305         int err = 0;
2306         struct ip_mreqn imr;
2307         __be32 addr = msf->imsf_multiaddr;
2308         struct ip_mc_socklist *pmc;
2309         struct in_device *in_dev;
2310         struct inet_sock *inet = inet_sk(sk);
2311         struct ip_sf_socklist *newpsl, *psl;
2312         struct net *net = sock_net(sk);
2313         int leavegroup = 0;
2314
2315         if (!ipv4_is_multicast(addr))
2316                 return -EINVAL;
2317         if (msf->imsf_fmode != MCAST_INCLUDE &&
2318             msf->imsf_fmode != MCAST_EXCLUDE)
2319                 return -EINVAL;
2320
2321         ASSERT_RTNL();
2322
2323         imr.imr_multiaddr.s_addr = msf->imsf_multiaddr;
2324         imr.imr_address.s_addr = msf->imsf_interface;
2325         imr.imr_ifindex = ifindex;
2326         in_dev = ip_mc_find_dev(net, &imr);
2327
2328         if (!in_dev) {
2329                 err = -ENODEV;
2330                 goto done;
2331         }
2332
2333         /* special case - (INCLUDE, empty) == LEAVE_GROUP */
2334         if (msf->imsf_fmode == MCAST_INCLUDE && msf->imsf_numsrc == 0) {
2335                 leavegroup = 1;
2336                 goto done;
2337         }
2338
2339         for_each_pmc_rtnl(inet, pmc) {
2340                 if (pmc->multi.imr_multiaddr.s_addr == msf->imsf_multiaddr &&
2341                     pmc->multi.imr_ifindex == imr.imr_ifindex)
2342                         break;
2343         }
2344         if (!pmc) {             /* must have a prior join */
2345                 err = -EINVAL;
2346                 goto done;
2347         }
2348         if (msf->imsf_numsrc) {
2349                 newpsl = sock_kmalloc(sk, IP_SFLSIZE(msf->imsf_numsrc),
2350                                                            GFP_KERNEL);
2351                 if (!newpsl) {
2352                         err = -ENOBUFS;
2353                         goto done;
2354                 }
2355                 newpsl->sl_max = newpsl->sl_count = msf->imsf_numsrc;
2356                 memcpy(newpsl->sl_addr, msf->imsf_slist,
2357                         msf->imsf_numsrc * sizeof(msf->imsf_slist[0]));
2358                 err = ip_mc_add_src(in_dev, &msf->imsf_multiaddr,
2359                         msf->imsf_fmode, newpsl->sl_count, newpsl->sl_addr, 0);
2360                 if (err) {
2361                         sock_kfree_s(sk, newpsl, IP_SFLSIZE(newpsl->sl_max));
2362                         goto done;
2363                 }
2364         } else {
2365                 newpsl = NULL;
2366                 (void) ip_mc_add_src(in_dev, &msf->imsf_multiaddr,
2367                                      msf->imsf_fmode, 0, NULL, 0);
2368         }
2369         psl = rtnl_dereference(pmc->sflist);
2370         if (psl) {
2371                 (void) ip_mc_del_src(in_dev, &msf->imsf_multiaddr, pmc->sfmode,
2372                         psl->sl_count, psl->sl_addr, 0);
2373                 /* decrease mem now to avoid the memleak warning */
2374                 atomic_sub(IP_SFLSIZE(psl->sl_max), &sk->sk_omem_alloc);
2375                 kfree_rcu(psl, rcu);
2376         } else
2377                 (void) ip_mc_del_src(in_dev, &msf->imsf_multiaddr, pmc->sfmode,
2378                         0, NULL, 0);
2379         rcu_assign_pointer(pmc->sflist, newpsl);
2380         pmc->sfmode = msf->imsf_fmode;
2381         err = 0;
2382 done:
2383         if (leavegroup)
2384                 err = ip_mc_leave_group(sk, &imr);
2385         return err;
2386 }
2387
2388 int ip_mc_msfget(struct sock *sk, struct ip_msfilter *msf,
2389         struct ip_msfilter __user *optval, int __user *optlen)
2390 {
2391         int err, len, count, copycount;
2392         struct ip_mreqn imr;
2393         __be32 addr = msf->imsf_multiaddr;
2394         struct ip_mc_socklist *pmc;
2395         struct in_device *in_dev;
2396         struct inet_sock *inet = inet_sk(sk);
2397         struct ip_sf_socklist *psl;
2398         struct net *net = sock_net(sk);
2399
2400         ASSERT_RTNL();
2401
2402         if (!ipv4_is_multicast(addr))
2403                 return -EINVAL;
2404
2405         imr.imr_multiaddr.s_addr = msf->imsf_multiaddr;
2406         imr.imr_address.s_addr = msf->imsf_interface;
2407         imr.imr_ifindex = 0;
2408         in_dev = ip_mc_find_dev(net, &imr);
2409
2410         if (!in_dev) {
2411                 err = -ENODEV;
2412                 goto done;
2413         }
2414         err = -EADDRNOTAVAIL;
2415
2416         for_each_pmc_rtnl(inet, pmc) {
2417                 if (pmc->multi.imr_multiaddr.s_addr == msf->imsf_multiaddr &&
2418                     pmc->multi.imr_ifindex == imr.imr_ifindex)
2419                         break;
2420         }
2421         if (!pmc)               /* must have a prior join */
2422                 goto done;
2423         msf->imsf_fmode = pmc->sfmode;
2424         psl = rtnl_dereference(pmc->sflist);
2425         if (!psl) {
2426                 len = 0;
2427                 count = 0;
2428         } else {
2429                 count = psl->sl_count;
2430         }
2431         copycount = count < msf->imsf_numsrc ? count : msf->imsf_numsrc;
2432         len = copycount * sizeof(psl->sl_addr[0]);
2433         msf->imsf_numsrc = count;
2434         if (put_user(IP_MSFILTER_SIZE(copycount), optlen) ||
2435             copy_to_user(optval, msf, IP_MSFILTER_SIZE(0))) {
2436                 return -EFAULT;
2437         }
2438         if (len &&
2439             copy_to_user(&optval->imsf_slist[0], psl->sl_addr, len))
2440                 return -EFAULT;
2441         return 0;
2442 done:
2443         return err;
2444 }
2445
2446 int ip_mc_gsfget(struct sock *sk, struct group_filter *gsf,
2447         struct group_filter __user *optval, int __user *optlen)
2448 {
2449         int err, i, count, copycount;
2450         struct sockaddr_in *psin;
2451         __be32 addr;
2452         struct ip_mc_socklist *pmc;
2453         struct inet_sock *inet = inet_sk(sk);
2454         struct ip_sf_socklist *psl;
2455
2456         ASSERT_RTNL();
2457
2458         psin = (struct sockaddr_in *)&gsf->gf_group;
2459         if (psin->sin_family != AF_INET)
2460                 return -EINVAL;
2461         addr = psin->sin_addr.s_addr;
2462         if (!ipv4_is_multicast(addr))
2463                 return -EINVAL;
2464
2465         err = -EADDRNOTAVAIL;
2466
2467         for_each_pmc_rtnl(inet, pmc) {
2468                 if (pmc->multi.imr_multiaddr.s_addr == addr &&
2469                     pmc->multi.imr_ifindex == gsf->gf_interface)
2470                         break;
2471         }
2472         if (!pmc)               /* must have a prior join */
2473                 goto done;
2474         gsf->gf_fmode = pmc->sfmode;
2475         psl = rtnl_dereference(pmc->sflist);
2476         count = psl ? psl->sl_count : 0;
2477         copycount = count < gsf->gf_numsrc ? count : gsf->gf_numsrc;
2478         gsf->gf_numsrc = count;
2479         if (put_user(GROUP_FILTER_SIZE(copycount), optlen) ||
2480             copy_to_user(optval, gsf, GROUP_FILTER_SIZE(0))) {
2481                 return -EFAULT;
2482         }
2483         for (i = 0; i < copycount; i++) {
2484                 struct sockaddr_storage ss;
2485
2486                 psin = (struct sockaddr_in *)&ss;
2487                 memset(&ss, 0, sizeof(ss));
2488                 psin->sin_family = AF_INET;
2489                 psin->sin_addr.s_addr = psl->sl_addr[i];
2490                 if (copy_to_user(&optval->gf_slist[i], &ss, sizeof(ss)))
2491                         return -EFAULT;
2492         }
2493         return 0;
2494 done:
2495         return err;
2496 }
2497
2498 /*
2499  * check if a multicast source filter allows delivery for a given <src,dst,intf>
2500  */
2501 int ip_mc_sf_allow(struct sock *sk, __be32 loc_addr, __be32 rmt_addr, int dif)
2502 {
2503         struct inet_sock *inet = inet_sk(sk);
2504         struct ip_mc_socklist *pmc;
2505         struct ip_sf_socklist *psl;
2506         int i;
2507         int ret;
2508
2509         ret = 1;
2510         if (!ipv4_is_multicast(loc_addr))
2511                 goto out;
2512
2513         rcu_read_lock();
2514         for_each_pmc_rcu(inet, pmc) {
2515                 if (pmc->multi.imr_multiaddr.s_addr == loc_addr &&
2516                     pmc->multi.imr_ifindex == dif)
2517                         break;
2518         }
2519         ret = inet->mc_all;
2520         if (!pmc)
2521                 goto unlock;
2522         psl = rcu_dereference(pmc->sflist);
2523         ret = (pmc->sfmode == MCAST_EXCLUDE);
2524         if (!psl)
2525                 goto unlock;
2526
2527         for (i = 0; i < psl->sl_count; i++) {
2528                 if (psl->sl_addr[i] == rmt_addr)
2529                         break;
2530         }
2531         ret = 0;
2532         if (pmc->sfmode == MCAST_INCLUDE && i >= psl->sl_count)
2533                 goto unlock;
2534         if (pmc->sfmode == MCAST_EXCLUDE && i < psl->sl_count)
2535                 goto unlock;
2536         ret = 1;
2537 unlock:
2538         rcu_read_unlock();
2539 out:
2540         return ret;
2541 }
2542
2543 /*
2544  *      A socket is closing.
2545  */
2546
2547 void ip_mc_drop_socket(struct sock *sk)
2548 {
2549         struct inet_sock *inet = inet_sk(sk);
2550         struct ip_mc_socklist *iml;
2551         struct net *net = sock_net(sk);
2552
2553         if (!inet->mc_list)
2554                 return;
2555
2556         rtnl_lock();
2557         while ((iml = rtnl_dereference(inet->mc_list)) != NULL) {
2558                 struct in_device *in_dev;
2559
2560                 inet->mc_list = iml->next_rcu;
2561                 in_dev = inetdev_by_index(net, iml->multi.imr_ifindex);
2562                 (void) ip_mc_leave_src(sk, iml, in_dev);
2563                 if (in_dev)
2564                         ip_mc_dec_group(in_dev, iml->multi.imr_multiaddr.s_addr);
2565                 /* decrease mem now to avoid the memleak warning */
2566                 atomic_sub(sizeof(*iml), &sk->sk_omem_alloc);
2567                 kfree_rcu(iml, rcu);
2568         }
2569         rtnl_unlock();
2570 }
2571
2572 /* called with rcu_read_lock() */
2573 int ip_check_mc_rcu(struct in_device *in_dev, __be32 mc_addr, __be32 src_addr, u8 proto)
2574 {
2575         struct ip_mc_list *im;
2576         struct ip_mc_list __rcu **mc_hash;
2577         struct ip_sf_list *psf;
2578         int rv = 0;
2579
2580         mc_hash = rcu_dereference(in_dev->mc_hash);
2581         if (mc_hash) {
2582                 u32 hash = hash_32((__force u32)mc_addr, MC_HASH_SZ_LOG);
2583
2584                 for (im = rcu_dereference(mc_hash[hash]);
2585                      im != NULL;
2586                      im = rcu_dereference(im->next_hash)) {
2587                         if (im->multiaddr == mc_addr)
2588                                 break;
2589                 }
2590         } else {
2591                 for_each_pmc_rcu(in_dev, im) {
2592                         if (im->multiaddr == mc_addr)
2593                                 break;
2594                 }
2595         }
2596         if (im && proto == IPPROTO_IGMP) {
2597                 rv = 1;
2598         } else if (im) {
2599                 if (src_addr) {
2600                         for (psf = im->sources; psf; psf = psf->sf_next) {
2601                                 if (psf->sf_inaddr == src_addr)
2602                                         break;
2603                         }
2604                         if (psf)
2605                                 rv = psf->sf_count[MCAST_INCLUDE] ||
2606                                         psf->sf_count[MCAST_EXCLUDE] !=
2607                                         im->sfcount[MCAST_EXCLUDE];
2608                         else
2609                                 rv = im->sfcount[MCAST_EXCLUDE] != 0;
2610                 } else
2611                         rv = 1; /* unspecified source; tentatively allow */
2612         }
2613         return rv;
2614 }
2615
2616 #if defined(CONFIG_PROC_FS)
2617 struct igmp_mc_iter_state {
2618         struct seq_net_private p;
2619         struct net_device *dev;
2620         struct in_device *in_dev;
2621 };
2622
2623 #define igmp_mc_seq_private(seq)        ((struct igmp_mc_iter_state *)(seq)->private)
2624
2625 static inline struct ip_mc_list *igmp_mc_get_first(struct seq_file *seq)
2626 {
2627         struct net *net = seq_file_net(seq);
2628         struct ip_mc_list *im = NULL;
2629         struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2630
2631         state->in_dev = NULL;
2632         for_each_netdev_rcu(net, state->dev) {
2633                 struct in_device *in_dev;
2634
2635                 in_dev = __in_dev_get_rcu(state->dev);
2636                 if (!in_dev)
2637                         continue;
2638                 im = rcu_dereference(in_dev->mc_list);
2639                 if (im) {
2640                         state->in_dev = in_dev;
2641                         break;
2642                 }
2643         }
2644         return im;
2645 }
2646
2647 static struct ip_mc_list *igmp_mc_get_next(struct seq_file *seq, struct ip_mc_list *im)
2648 {
2649         struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2650
2651         im = rcu_dereference(im->next_rcu);
2652         while (!im) {
2653                 state->dev = next_net_device_rcu(state->dev);
2654                 if (!state->dev) {
2655                         state->in_dev = NULL;
2656                         break;
2657                 }
2658                 state->in_dev = __in_dev_get_rcu(state->dev);
2659                 if (!state->in_dev)
2660                         continue;
2661                 im = rcu_dereference(state->in_dev->mc_list);
2662         }
2663         return im;
2664 }
2665
2666 static struct ip_mc_list *igmp_mc_get_idx(struct seq_file *seq, loff_t pos)
2667 {
2668         struct ip_mc_list *im = igmp_mc_get_first(seq);
2669         if (im)
2670                 while (pos && (im = igmp_mc_get_next(seq, im)) != NULL)
2671                         --pos;
2672         return pos ? NULL : im;
2673 }
2674
2675 static void *igmp_mc_seq_start(struct seq_file *seq, loff_t *pos)
2676         __acquires(rcu)
2677 {
2678         rcu_read_lock();
2679         return *pos ? igmp_mc_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
2680 }
2681
2682 static void *igmp_mc_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2683 {
2684         struct ip_mc_list *im;
2685         if (v == SEQ_START_TOKEN)
2686                 im = igmp_mc_get_first(seq);
2687         else
2688                 im = igmp_mc_get_next(seq, v);
2689         ++*pos;
2690         return im;
2691 }
2692
2693 static void igmp_mc_seq_stop(struct seq_file *seq, void *v)
2694         __releases(rcu)
2695 {
2696         struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2697
2698         state->in_dev = NULL;
2699         state->dev = NULL;
2700         rcu_read_unlock();
2701 }
2702
2703 static int igmp_mc_seq_show(struct seq_file *seq, void *v)
2704 {
2705         if (v == SEQ_START_TOKEN)
2706                 seq_puts(seq,
2707                          "Idx\tDevice    : Count Querier\tGroup    Users Timer\tReporter\n");
2708         else {
2709                 struct ip_mc_list *im = (struct ip_mc_list *)v;
2710                 struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2711                 char   *querier;
2712                 long delta;
2713
2714 #ifdef CONFIG_IP_MULTICAST
2715                 querier = IGMP_V1_SEEN(state->in_dev) ? "V1" :
2716                           IGMP_V2_SEEN(state->in_dev) ? "V2" :
2717                           "V3";
2718 #else
2719                 querier = "NONE";
2720 #endif
2721
2722                 if (rcu_access_pointer(state->in_dev->mc_list) == im) {
2723                         seq_printf(seq, "%d\t%-10s: %5d %7s\n",
2724                                    state->dev->ifindex, state->dev->name, state->in_dev->mc_count, querier);
2725                 }
2726
2727                 delta = im->timer.expires - jiffies;
2728                 seq_printf(seq,
2729                            "\t\t\t\t%08X %5d %d:%08lX\t\t%d\n",
2730                            im->multiaddr, im->users,
2731                            im->tm_running,
2732                            im->tm_running ? jiffies_delta_to_clock_t(delta) : 0,
2733                            im->reporter);
2734         }
2735         return 0;
2736 }
2737
2738 static const struct seq_operations igmp_mc_seq_ops = {
2739         .start  =       igmp_mc_seq_start,
2740         .next   =       igmp_mc_seq_next,
2741         .stop   =       igmp_mc_seq_stop,
2742         .show   =       igmp_mc_seq_show,
2743 };
2744
2745 static int igmp_mc_seq_open(struct inode *inode, struct file *file)
2746 {
2747         return seq_open_net(inode, file, &igmp_mc_seq_ops,
2748                         sizeof(struct igmp_mc_iter_state));
2749 }
2750
2751 static const struct file_operations igmp_mc_seq_fops = {
2752         .owner          =       THIS_MODULE,
2753         .open           =       igmp_mc_seq_open,
2754         .read           =       seq_read,
2755         .llseek         =       seq_lseek,
2756         .release        =       seq_release_net,
2757 };
2758
2759 struct igmp_mcf_iter_state {
2760         struct seq_net_private p;
2761         struct net_device *dev;
2762         struct in_device *idev;
2763         struct ip_mc_list *im;
2764 };
2765
2766 #define igmp_mcf_seq_private(seq)       ((struct igmp_mcf_iter_state *)(seq)->private)
2767
2768 static inline struct ip_sf_list *igmp_mcf_get_first(struct seq_file *seq)
2769 {
2770         struct net *net = seq_file_net(seq);
2771         struct ip_sf_list *psf = NULL;
2772         struct ip_mc_list *im = NULL;
2773         struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2774
2775         state->idev = NULL;
2776         state->im = NULL;
2777         for_each_netdev_rcu(net, state->dev) {
2778                 struct in_device *idev;
2779                 idev = __in_dev_get_rcu(state->dev);
2780                 if (unlikely(!idev))
2781                         continue;
2782                 im = rcu_dereference(idev->mc_list);
2783                 if (likely(im)) {
2784                         spin_lock_bh(&im->lock);
2785                         psf = im->sources;
2786                         if (likely(psf)) {
2787                                 state->im = im;
2788                                 state->idev = idev;
2789                                 break;
2790                         }
2791                         spin_unlock_bh(&im->lock);
2792                 }
2793         }
2794         return psf;
2795 }
2796
2797 static struct ip_sf_list *igmp_mcf_get_next(struct seq_file *seq, struct ip_sf_list *psf)
2798 {
2799         struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2800
2801         psf = psf->sf_next;
2802         while (!psf) {
2803                 spin_unlock_bh(&state->im->lock);
2804                 state->im = state->im->next;
2805                 while (!state->im) {
2806                         state->dev = next_net_device_rcu(state->dev);
2807                         if (!state->dev) {
2808                                 state->idev = NULL;
2809                                 goto out;
2810                         }
2811                         state->idev = __in_dev_get_rcu(state->dev);
2812                         if (!state->idev)
2813                                 continue;
2814                         state->im = rcu_dereference(state->idev->mc_list);
2815                 }
2816                 if (!state->im)
2817                         break;
2818                 spin_lock_bh(&state->im->lock);
2819                 psf = state->im->sources;
2820         }
2821 out:
2822         return psf;
2823 }
2824
2825 static struct ip_sf_list *igmp_mcf_get_idx(struct seq_file *seq, loff_t pos)
2826 {
2827         struct ip_sf_list *psf = igmp_mcf_get_first(seq);
2828         if (psf)
2829                 while (pos && (psf = igmp_mcf_get_next(seq, psf)) != NULL)
2830                         --pos;
2831         return pos ? NULL : psf;
2832 }
2833
2834 static void *igmp_mcf_seq_start(struct seq_file *seq, loff_t *pos)
2835         __acquires(rcu)
2836 {
2837         rcu_read_lock();
2838         return *pos ? igmp_mcf_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
2839 }
2840
2841 static void *igmp_mcf_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2842 {
2843         struct ip_sf_list *psf;
2844         if (v == SEQ_START_TOKEN)
2845                 psf = igmp_mcf_get_first(seq);
2846         else
2847                 psf = igmp_mcf_get_next(seq, v);
2848         ++*pos;
2849         return psf;
2850 }
2851
2852 static void igmp_mcf_seq_stop(struct seq_file *seq, void *v)
2853         __releases(rcu)
2854 {
2855         struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2856         if (likely(state->im)) {
2857                 spin_unlock_bh(&state->im->lock);
2858                 state->im = NULL;
2859         }
2860         state->idev = NULL;
2861         state->dev = NULL;
2862         rcu_read_unlock();
2863 }
2864
2865 static int igmp_mcf_seq_show(struct seq_file *seq, void *v)
2866 {
2867         struct ip_sf_list *psf = (struct ip_sf_list *)v;
2868         struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2869
2870         if (v == SEQ_START_TOKEN) {
2871                 seq_puts(seq, "Idx Device        MCA        SRC    INC    EXC\n");
2872         } else {
2873                 seq_printf(seq,
2874                            "%3d %6.6s 0x%08x "
2875                            "0x%08x %6lu %6lu\n",
2876                            state->dev->ifindex, state->dev->name,
2877                            ntohl(state->im->multiaddr),
2878                            ntohl(psf->sf_inaddr),
2879                            psf->sf_count[MCAST_INCLUDE],
2880                            psf->sf_count[MCAST_EXCLUDE]);
2881         }
2882         return 0;
2883 }
2884
2885 static const struct seq_operations igmp_mcf_seq_ops = {
2886         .start  =       igmp_mcf_seq_start,
2887         .next   =       igmp_mcf_seq_next,
2888         .stop   =       igmp_mcf_seq_stop,
2889         .show   =       igmp_mcf_seq_show,
2890 };
2891
2892 static int igmp_mcf_seq_open(struct inode *inode, struct file *file)
2893 {
2894         return seq_open_net(inode, file, &igmp_mcf_seq_ops,
2895                         sizeof(struct igmp_mcf_iter_state));
2896 }
2897
2898 static const struct file_operations igmp_mcf_seq_fops = {
2899         .owner          =       THIS_MODULE,
2900         .open           =       igmp_mcf_seq_open,
2901         .read           =       seq_read,
2902         .llseek         =       seq_lseek,
2903         .release        =       seq_release_net,
2904 };
2905
2906 static int __net_init igmp_net_init(struct net *net)
2907 {
2908         struct proc_dir_entry *pde;
2909         int err;
2910
2911         pde = proc_create("igmp", S_IRUGO, net->proc_net, &igmp_mc_seq_fops);
2912         if (!pde)
2913                 goto out_igmp;
2914         pde = proc_create("mcfilter", S_IRUGO, net->proc_net,
2915                           &igmp_mcf_seq_fops);
2916         if (!pde)
2917                 goto out_mcfilter;
2918         err = inet_ctl_sock_create(&net->ipv4.mc_autojoin_sk, AF_INET,
2919                                    SOCK_DGRAM, 0, net);
2920         if (err < 0) {
2921                 pr_err("Failed to initialize the IGMP autojoin socket (err %d)\n",
2922                        err);
2923                 goto out_sock;
2924         }
2925
2926         return 0;
2927
2928 out_sock:
2929         remove_proc_entry("mcfilter", net->proc_net);
2930 out_mcfilter:
2931         remove_proc_entry("igmp", net->proc_net);
2932 out_igmp:
2933         return -ENOMEM;
2934 }
2935
2936 static void __net_exit igmp_net_exit(struct net *net)
2937 {
2938         remove_proc_entry("mcfilter", net->proc_net);
2939         remove_proc_entry("igmp", net->proc_net);
2940         inet_ctl_sock_destroy(net->ipv4.mc_autojoin_sk);
2941 }
2942
2943 static struct pernet_operations igmp_net_ops = {
2944         .init = igmp_net_init,
2945         .exit = igmp_net_exit,
2946 };
2947 #endif
2948
2949 static int igmp_netdev_event(struct notifier_block *this,
2950                              unsigned long event, void *ptr)
2951 {
2952         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2953         struct in_device *in_dev;
2954
2955         switch (event) {
2956         case NETDEV_RESEND_IGMP:
2957                 in_dev = __in_dev_get_rtnl(dev);
2958                 if (in_dev)
2959                         ip_mc_rejoin_groups(in_dev);
2960                 break;
2961         default:
2962                 break;
2963         }
2964         return NOTIFY_DONE;
2965 }
2966
2967 static struct notifier_block igmp_notifier = {
2968         .notifier_call = igmp_netdev_event,
2969 };
2970
2971 int __init igmp_mc_init(void)
2972 {
2973 #if defined(CONFIG_PROC_FS)
2974         int err;
2975
2976         err = register_pernet_subsys(&igmp_net_ops);
2977         if (err)
2978                 return err;
2979         err = register_netdevice_notifier(&igmp_notifier);
2980         if (err)
2981                 goto reg_notif_fail;
2982         return 0;
2983
2984 reg_notif_fail:
2985         unregister_pernet_subsys(&igmp_net_ops);
2986         return err;
2987 #else
2988         return register_netdevice_notifier(&igmp_notifier);
2989 #endif
2990 }