net/ipv4/igmp.c

   1 /*
   2  *      Linux NET3:     Internet Group Management Protocol  [IGMP]
   3  *
   4  *      This code implements the IGMP protocol as defined in RFC1112. There has
   5  *      been a further revision of this protocol since which is now supported.
   6  *
   7  *      If you have trouble with this module be careful what gcc you have used,
   8  *      the older version didn't come out right using gcc 2.5.8, the newer one
   9  *      seems to fall out with gcc 2.6.2.
  10  *
  11  *      Authors:
  12  *              Alan Cox <alan@lxorguk.ukuu.org.uk>
  13  *
  14  *      This program is free software; you can redistribute it and/or
  15  *      modify it under the terms of the GNU General Public License
  16  *      as published by the Free Software Foundation; either version
  17  *      2 of the License, or (at your option) any later version.
  18  *
  19  *      Fixes:
  20  *
  21  *              Alan Cox        :       Added lots of __inline__ to optimise
  22  *                                      the memory usage of all the tiny little
  23  *                                      functions.
  24  *              Alan Cox        :       Dumped the header building experiment.
  25  *              Alan Cox        :       Minor tweaks ready for multicast routing
  26  *                                      and extended IGMP protocol.
  27  *              Alan Cox        :       Removed a load of inline directives. Gcc 2.5.8
  28  *                                      writes utterly bogus code otherwise (sigh)
  29  *                                      fixed IGMP loopback to behave in the manner
  30  *                                      desired by mrouted, fixed the fact it has been
  31  *                                      broken since 1.3.6 and cleaned up a few minor
  32  *                                      points.
  33  *
  34  *              Chih-Jen Chang  :       Tried to revise IGMP to Version 2
  35  *              Tsu-Sheng Tsao          E-mail: chihjenc@scf.usc.edu and tsusheng@scf.usc.edu
  36  *                                      The enhancements are mainly based on Steve Deering's
  37  *                                      ipmulti-3.5 source code.
  38  *              Chih-Jen Chang  :       Added the igmp_get_mrouter_info and
  39  *              Tsu-Sheng Tsao          igmp_set_mrouter_info to keep track of
  40  *                                      the mrouted version on that device.
  41  *              Chih-Jen Chang  :       Added the max_resp_time parameter to
  42  *              Tsu-Sheng Tsao          igmp_heard_query(). Using this parameter
  43  *                                      to identify the multicast router version
  44  *                                      and do what the IGMP version 2 specified.
  45  *              Chih-Jen Chang  :       Added a timer to revert to IGMP V2 router
  46  *              Tsu-Sheng Tsao          if the specified time expired.
  47  *              Alan Cox        :       Stop IGMP from 0.0.0.0 being accepted.
  48  *              Alan Cox        :       Use GFP_ATOMIC in the right places.
  49  *              Christian Daudt :       igmp timer wasn't set for local group
  50  *                                      memberships but was being deleted,
  51  *                                      which caused a "del_timer() called
  52  *                                      from %p with timer not initialized\n"
  53  *                                      message (960131).
  54  *              Christian Daudt :       removed del_timer from
  55  *                                      igmp_timer_expire function (960205).
  56  *             Christian Daudt :       igmp_heard_report now only calls
  57  *                                     igmp_timer_expire if tm->running is
  58  *                                     true (960216).
  59  *              Malcolm Beattie :       ttl comparison wrong in igmp_rcv made
  60  *                                      igmp_heard_query never trigger. Expiry
  61  *                                      miscalculation fixed in igmp_heard_query
  62  *                                      and random() made to return unsigned to
  63  *                                      prevent negative expiry times.
  64  *              Alexey Kuznetsov:       Wrong group leaving behaviour, backport
  65  *                                      fix from pending 2.1.x patches.
  66  *              Alan Cox:               Forget to enable FDDI support earlier.
  67  *              Alexey Kuznetsov:       Fixed leaving groups on device down.
  68  *              Alexey Kuznetsov:       Accordance to igmp-v2-06 draft.
  69  *              David L Stevens:        IGMPv3 support, with help from
  70  *                                      Vinay Kulkarni
  71  */
  72
  73 #include <linux/module.h>
  74 #include <linux/slab.h>
  75 #include <asm/uaccess.h>
  76 #include <linux/types.h>
  77 #include <linux/kernel.h>
  78 #include <linux/jiffies.h>
  79 #include <linux/string.h>
  80 #include <linux/socket.h>
  81 #include <linux/sockios.h>
  82 #include <linux/in.h>
  83 #include <linux/inet.h>
  84 #include <linux/netdevice.h>
  85 #include <linux/skbuff.h>
  86 #include <linux/inetdevice.h>
  87 #include <linux/igmp.h>
  88 #include <linux/if_arp.h>
  89 #include <linux/rtnetlink.h>
  90 #include <linux/times.h>
  91 #include <linux/pkt_sched.h>
  92
  93 #include <net/net_namespace.h>
  94 #include <net/arp.h>
  95 #include <net/ip.h>
  96 #include <net/protocol.h>
  97 #include <net/route.h>
  98 #include <net/sock.h>
  99 #include <net/checksum.h>
 100 #include <net/inet_common.h>
 101 #include <linux/netfilter_ipv4.h>
 102 #ifdef CONFIG_IP_MROUTE
 103 #include <linux/mroute.h>
 104 #endif
 105 #ifdef CONFIG_PROC_FS
 106 #include <linux/proc_fs.h>
 107 #include <linux/seq_file.h>
 108 #endif
 109
 110 #define IP_MAX_MEMBERSHIPS      20
 111 #define IP_MAX_MSF              10
 112
 113 /* IGMP reports for link-local multicast groups are enabled by default */
 114 int sysctl_igmp_llm_reports __read_mostly = 1;
 115
 116 #ifdef CONFIG_IP_MULTICAST
 117 /* Parameter names and values are taken from igmp-v2-06 draft */
 118
 119 #define IGMP_V1_ROUTER_PRESENT_TIMEOUT          (400*HZ)
 120 #define IGMP_V2_ROUTER_PRESENT_TIMEOUT          (400*HZ)
 121 #define IGMP_V2_UNSOLICITED_REPORT_INTERVAL     (10*HZ)
 122 #define IGMP_V3_UNSOLICITED_REPORT_INTERVAL     (1*HZ)
 123 #define IGMP_QUERY_RESPONSE_INTERVAL            (10*HZ)
 124 #define IGMP_QUERY_ROBUSTNESS_VARIABLE          2
 125
 126
 127 #define IGMP_INITIAL_REPORT_DELAY               (1)
 128
 129 /* IGMP_INITIAL_REPORT_DELAY is not from IGMP specs!
 130  * IGMP specs require to report membership immediately after
 131  * joining a group, but we delay the first report by a
 132  * small interval. It seems more natural and still does not
 133  * contradict to specs provided this delay is small enough.
 134  */
 135
 136 #define IGMP_V1_SEEN(in_dev) \
 137         (IPV4_DEVCONF_ALL(dev_net(in_dev->dev), FORCE_IGMP_VERSION) == 1 || \
 138          IN_DEV_CONF_GET((in_dev), FORCE_IGMP_VERSION) == 1 || \
 139          ((in_dev)->mr_v1_seen && \
 140           time_before(jiffies, (in_dev)->mr_v1_seen)))
 141 #define IGMP_V2_SEEN(in_dev) \
 142         (IPV4_DEVCONF_ALL(dev_net(in_dev->dev), FORCE_IGMP_VERSION) == 2 || \
 143          IN_DEV_CONF_GET((in_dev), FORCE_IGMP_VERSION) == 2 || \
 144          ((in_dev)->mr_v2_seen && \
 145           time_before(jiffies, (in_dev)->mr_v2_seen)))
 146
 147 static int unsolicited_report_interval(struct in_device *in_dev)
 148 {
 149         int interval_ms, interval_jiffies;
 150
 151         if (IGMP_V1_SEEN(in_dev) || IGMP_V2_SEEN(in_dev))
 152                 interval_ms = IN_DEV_CONF_GET(
 153                         in_dev,
 154                         IGMPV2_UNSOLICITED_REPORT_INTERVAL);
 155         else /* v3 */
 156                 interval_ms = IN_DEV_CONF_GET(
 157                         in_dev,
 158                         IGMPV3_UNSOLICITED_REPORT_INTERVAL);
 159
 160         interval_jiffies = msecs_to_jiffies(interval_ms);
 161
 162         /* _timer functions can't handle a delay of 0 jiffies so ensure
 163          *  we always return a positive value.
 164          */
 165         if (interval_jiffies <= 0)
 166                 interval_jiffies = 1;
 167         return interval_jiffies;
 168 }
 169
 170 static void igmpv3_add_delrec(struct in_device *in_dev, struct ip_mc_list *im);
 171 static void igmpv3_del_delrec(struct in_device *in_dev, __be32 multiaddr);
 172 static void igmpv3_clear_delrec(struct in_device *in_dev);
 173 static int sf_setstate(struct ip_mc_list *pmc);
 174 static void sf_markstate(struct ip_mc_list *pmc);
 175 #endif
 176 static void ip_mc_clear_src(struct ip_mc_list *pmc);
 177 static int ip_mc_add_src(struct in_device *in_dev, __be32 *pmca, int sfmode,
 178                          int sfcount, __be32 *psfsrc, int delta);
 179
 180 static void ip_ma_put(struct ip_mc_list *im)
 181 {
 182         if (atomic_dec_and_test(&im->refcnt)) {
 183                 in_dev_put(im->interface);
 184                 kfree_rcu(im, rcu);
 185         }
 186 }
 187
 188 #define for_each_pmc_rcu(in_dev, pmc)                           \
 189         for (pmc = rcu_dereference(in_dev->mc_list);            \
 190              pmc != NULL;                                       \
 191              pmc = rcu_dereference(pmc->next_rcu))
 192
 193 #define for_each_pmc_rtnl(in_dev, pmc)                          \
 194         for (pmc = rtnl_dereference(in_dev->mc_list);           \
 195              pmc != NULL;                                       \
 196              pmc = rtnl_dereference(pmc->next_rcu))
 197
 198 #ifdef CONFIG_IP_MULTICAST
 199
 200 /*
 201  *      Timer management
 202  */
 203
 204 static void igmp_stop_timer(struct ip_mc_list *im)
 205 {
 206         spin_lock_bh(&im->lock);
 207         if (del_timer(&im->timer))
 208                 atomic_dec(&im->refcnt);
 209         im->tm_running = 0;
 210         im->reporter = 0;
 211         im->unsolicit_count = 0;
 212         spin_unlock_bh(&im->lock);
 213 }
 214
 215 /* It must be called with locked im->lock */
 216 static void igmp_start_timer(struct ip_mc_list *im, int max_delay)
 217 {
 218         int tv = prandom_u32() % max_delay;
 219
 220         im->tm_running = 1;
 221         if (!mod_timer(&im->timer, jiffies+tv+2))
 222                 atomic_inc(&im->refcnt);
 223 }
 224
 225 static void igmp_gq_start_timer(struct in_device *in_dev)
 226 {
 227         int tv = prandom_u32() % in_dev->mr_maxdelay;
 228         unsigned long exp = jiffies + tv + 2;
 229
 230         if (in_dev->mr_gq_running &&
 231             time_after_eq(exp, (in_dev->mr_gq_timer).expires))
 232                 return;
 233
 234         in_dev->mr_gq_running = 1;
 235         if (!mod_timer(&in_dev->mr_gq_timer, exp))
 236                 in_dev_hold(in_dev);
 237 }
 238
 239 static void igmp_ifc_start_timer(struct in_device *in_dev, int delay)
 240 {
 241         int tv = prandom_u32() % delay;
 242
 243         if (!mod_timer(&in_dev->mr_ifc_timer, jiffies+tv+2))
 244                 in_dev_hold(in_dev);
 245 }
 246
 247 static void igmp_mod_timer(struct ip_mc_list *im, int max_delay)
 248 {
 249         spin_lock_bh(&im->lock);
 250         im->unsolicit_count = 0;
 251         if (del_timer(&im->timer)) {
 252                 if ((long)(im->timer.expires-jiffies) < max_delay) {
 253                         add_timer(&im->timer);
 254                         im->tm_running = 1;
 255                         spin_unlock_bh(&im->lock);
 256                         return;
 257                 }
 258                 atomic_dec(&im->refcnt);
 259         }
 260         igmp_start_timer(im, max_delay);
 261         spin_unlock_bh(&im->lock);
 262 }
 263
 264
 265 /*
 266  *      Send an IGMP report.
 267  */
 268
 269 #define IGMP_SIZE (sizeof(struct igmphdr)+sizeof(struct iphdr)+4)
 270
 271
 272 static int is_in(struct ip_mc_list *pmc, struct ip_sf_list *psf, int type,
 273         int gdeleted, int sdeleted)
 274 {
 275         switch (type) {
 276         case IGMPV3_MODE_IS_INCLUDE:
 277         case IGMPV3_MODE_IS_EXCLUDE:
 278                 if (gdeleted || sdeleted)
 279                         return 0;
 280                 if (!(pmc->gsquery && !psf->sf_gsresp)) {
 281                         if (pmc->sfmode == MCAST_INCLUDE)
 282                                 return 1;
 283                         /* don't include if this source is excluded
 284                          * in all filters
 285                          */
 286                         if (psf->sf_count[MCAST_INCLUDE])
 287                                 return type == IGMPV3_MODE_IS_INCLUDE;
 288                         return pmc->sfcount[MCAST_EXCLUDE] ==
 289                                 psf->sf_count[MCAST_EXCLUDE];
 290                 }
 291                 return 0;
 292         case IGMPV3_CHANGE_TO_INCLUDE:
 293                 if (gdeleted || sdeleted)
 294                         return 0;
 295                 return psf->sf_count[MCAST_INCLUDE] != 0;
 296         case IGMPV3_CHANGE_TO_EXCLUDE:
 297                 if (gdeleted || sdeleted)
 298                         return 0;
 299                 if (pmc->sfcount[MCAST_EXCLUDE] == 0 ||
 300                     psf->sf_count[MCAST_INCLUDE])
 301                         return 0;
 302                 return pmc->sfcount[MCAST_EXCLUDE] ==
 303                         psf->sf_count[MCAST_EXCLUDE];
 304         case IGMPV3_ALLOW_NEW_SOURCES:
 305                 if (gdeleted || !psf->sf_crcount)
 306                         return 0;
 307                 return (pmc->sfmode == MCAST_INCLUDE) ^ sdeleted;
 308         case IGMPV3_BLOCK_OLD_SOURCES:
 309                 if (pmc->sfmode == MCAST_INCLUDE)
 310                         return gdeleted || (psf->sf_crcount && sdeleted);
 311                 return psf->sf_crcount && !gdeleted && !sdeleted;
 312         }
 313         return 0;
 314 }
 315
 316 static int
 317 igmp_scount(struct ip_mc_list *pmc, int type, int gdeleted, int sdeleted)
 318 {
 319         struct ip_sf_list *psf;
 320         int scount = 0;
 321
 322         for (psf = pmc->sources; psf; psf = psf->sf_next) {
 323                 if (!is_in(pmc, psf, type, gdeleted, sdeleted))
 324                         continue;
 325                 scount++;
 326         }
 327         return scount;
 328 }
 329
 330 static struct sk_buff *igmpv3_newpack(struct net_device *dev, unsigned int mtu)
 331 {
 332         struct sk_buff *skb;
 333         struct rtable *rt;
 334         struct iphdr *pip;
 335         struct igmpv3_report *pig;
 336         struct net *net = dev_net(dev);
 337         struct flowi4 fl4;
 338         int hlen = LL_RESERVED_SPACE(dev);
 339         int tlen = dev->needed_tailroom;
 340         unsigned int size = mtu;
 341
 342         while (1) {
 343                 skb = alloc_skb(size + hlen + tlen,
 344                                 GFP_ATOMIC | __GFP_NOWARN);
 345                 if (skb)
 346                         break;
 347                 size >>= 1;
 348                 if (size < 256)
 349                         return NULL;
 350         }
 351         skb->priority = TC_PRIO_CONTROL;
 352
 353         rt = ip_route_output_ports(net, &fl4, NULL, IGMPV3_ALL_MCR, 0,
 354                                    0, 0,
 355                                    IPPROTO_IGMP, 0, dev->ifindex);
 356         if (IS_ERR(rt)) {
 357                 kfree_skb(skb);
 358                 return NULL;
 359         }
 360
 361         skb_dst_set(skb, &rt->dst);
 362         skb->dev = dev;
 363
 364         skb_reserve(skb, hlen);
 365         skb_tailroom_reserve(skb, mtu, tlen);
 366
 367         skb_reset_network_header(skb);
 368         pip = ip_hdr(skb);
 369         skb_put(skb, sizeof(struct iphdr) + 4);
 370
 371         pip->version  = 4;
 372         pip->ihl      = (sizeof(struct iphdr)+4)>>2;
 373         pip->tos      = 0xc0;
 374         pip->frag_off = htons(IP_DF);
 375         pip->ttl      = 1;
 376         pip->daddr    = fl4.daddr;
 377         pip->saddr    = fl4.saddr;
 378         pip->protocol = IPPROTO_IGMP;
 379         pip->tot_len  = 0;      /* filled in later */
 380         ip_select_ident(net, skb, NULL);
 381         ((u8 *)&pip[1])[0] = IPOPT_RA;
 382         ((u8 *)&pip[1])[1] = 4;
 383         ((u8 *)&pip[1])[2] = 0;
 384         ((u8 *)&pip[1])[3] = 0;
 385
 386         skb->transport_header = skb->network_header + sizeof(struct iphdr) + 4;
 387         skb_put(skb, sizeof(*pig));
 388         pig = igmpv3_report_hdr(skb);
 389         pig->type = IGMPV3_HOST_MEMBERSHIP_REPORT;
 390         pig->resv1 = 0;
 391         pig->csum = 0;
 392         pig->resv2 = 0;
 393         pig->ngrec = 0;
 394         return skb;
 395 }
 396
 397 static int igmpv3_sendpack(struct sk_buff *skb)
 398 {
 399         struct igmphdr *pig = igmp_hdr(skb);
 400         const int igmplen = skb_tail_pointer(skb) - skb_transport_header(skb);
 401
 402         pig->csum = ip_compute_csum(igmp_hdr(skb), igmplen);
 403
 404         return ip_local_out(dev_net(skb_dst(skb)->dev), skb->sk, skb);
 405 }
 406
 407 static int grec_size(struct ip_mc_list *pmc, int type, int gdel, int sdel)
 408 {
 409         return sizeof(struct igmpv3_grec) + 4*igmp_scount(pmc, type, gdel, sdel);
 410 }
 411
 412 static struct sk_buff *add_grhead(struct sk_buff *skb, struct ip_mc_list *pmc,
 413         int type, struct igmpv3_grec **ppgr)
 414 {
 415         struct net_device *dev = pmc->interface->dev;
 416         struct igmpv3_report *pih;
 417         struct igmpv3_grec *pgr;
 418
 419         if (!skb)
 420                 skb = igmpv3_newpack(dev, dev->mtu);
 421         if (!skb)
 422                 return NULL;
 423         pgr = (struct igmpv3_grec *)skb_put(skb, sizeof(struct igmpv3_grec));
 424         pgr->grec_type = type;
 425         pgr->grec_auxwords = 0;
 426         pgr->grec_nsrcs = 0;
 427         pgr->grec_mca = pmc->multiaddr;
 428         pih = igmpv3_report_hdr(skb);
 429         pih->ngrec = htons(ntohs(pih->ngrec)+1);
 430         *ppgr = pgr;
 431         return skb;
 432 }
 433
 434 #define AVAILABLE(skb)  ((skb) ? skb_availroom(skb) : 0)
 435
 436 static struct sk_buff *add_grec(struct sk_buff *skb, struct ip_mc_list *pmc,
 437         int type, int gdeleted, int sdeleted)
 438 {
 439         struct net_device *dev = pmc->interface->dev;
 440         struct igmpv3_report *pih;
 441         struct igmpv3_grec *pgr = NULL;
 442         struct ip_sf_list *psf, *psf_next, *psf_prev, **psf_list;
 443         int scount, stotal, first, isquery, truncate;
 444
 445         if (pmc->multiaddr == IGMP_ALL_HOSTS)
 446                 return skb;
 447         if (ipv4_is_local_multicast(pmc->multiaddr) && !sysctl_igmp_llm_reports)
 448                 return skb;
 449
 450         isquery = type == IGMPV3_MODE_IS_INCLUDE ||
 451                   type == IGMPV3_MODE_IS_EXCLUDE;
 452         truncate = type == IGMPV3_MODE_IS_EXCLUDE ||
 453                     type == IGMPV3_CHANGE_TO_EXCLUDE;
 454
 455         stotal = scount = 0;
 456
 457         psf_list = sdeleted ? &pmc->tomb : &pmc->sources;
 458
 459         if (!*psf_list)
 460                 goto empty_source;
 461
 462         pih = skb ? igmpv3_report_hdr(skb) : NULL;
 463
 464         /* EX and TO_EX get a fresh packet, if needed */
 465         if (truncate) {
 466                 if (pih && pih->ngrec &&
 467                     AVAILABLE(skb) < grec_size(pmc, type, gdeleted, sdeleted)) {
 468                         if (skb)
 469                                 igmpv3_sendpack(skb);
 470                         skb = igmpv3_newpack(dev, dev->mtu);
 471                 }
 472         }
 473         first = 1;
 474         psf_prev = NULL;
 475         for (psf = *psf_list; psf; psf = psf_next) {
 476                 __be32 *psrc;
 477
 478                 psf_next = psf->sf_next;
 479
 480                 if (!is_in(pmc, psf, type, gdeleted, sdeleted)) {
 481                         psf_prev = psf;
 482                         continue;
 483                 }
 484
 485                 /* clear marks on query responses */
 486                 if (isquery)
 487                         psf->sf_gsresp = 0;
 488
 489                 if (AVAILABLE(skb) < sizeof(__be32) +
 490                     first*sizeof(struct igmpv3_grec)) {
 491                         if (truncate && !first)
 492                                 break;   /* truncate these */
 493                         if (pgr)
 494                                 pgr->grec_nsrcs = htons(scount);
 495                         if (skb)
 496                                 igmpv3_sendpack(skb);
 497                         skb = igmpv3_newpack(dev, dev->mtu);
 498                         first = 1;
 499                         scount = 0;
 500                 }
 501                 if (first) {
 502                         skb = add_grhead(skb, pmc, type, &pgr);
 503                         first = 0;
 504                 }
 505                 if (!skb)
 506                         return NULL;
 507                 psrc = (__be32 *)skb_put(skb, sizeof(__be32));
 508                 *psrc = psf->sf_inaddr;
 509                 scount++; stotal++;
 510                 if ((type == IGMPV3_ALLOW_NEW_SOURCES ||
 511                      type == IGMPV3_BLOCK_OLD_SOURCES) && psf->sf_crcount) {
 512                         psf->sf_crcount--;
 513                         if ((sdeleted || gdeleted) && psf->sf_crcount == 0) {
 514                                 if (psf_prev)
 515                                         psf_prev->sf_next = psf->sf_next;
 516                                 else
 517                                         *psf_list = psf->sf_next;
 518                                 kfree(psf);
 519                                 continue;
 520                         }
 521                 }
 522                 psf_prev = psf;
 523         }
 524
 525 empty_source:
 526         if (!stotal) {
 527                 if (type == IGMPV3_ALLOW_NEW_SOURCES ||
 528                     type == IGMPV3_BLOCK_OLD_SOURCES)
 529                         return skb;
 530                 if (pmc->crcount || isquery) {
 531                         /* make sure we have room for group header */
 532                         if (skb && AVAILABLE(skb) < sizeof(struct igmpv3_grec)) {
 533                                 igmpv3_sendpack(skb);
 534                                 skb = NULL; /* add_grhead will get a new one */
 535                         }
 536                         skb = add_grhead(skb, pmc, type, &pgr);
 537                 }
 538         }
 539         if (pgr)
 540                 pgr->grec_nsrcs = htons(scount);
 541
 542         if (isquery)
 543                 pmc->gsquery = 0;       /* clear query state on report */
 544         return skb;
 545 }
 546
 547 static int igmpv3_send_report(struct in_device *in_dev, struct ip_mc_list *pmc)
 548 {
 549         struct sk_buff *skb = NULL;
 550         int type;
 551
 552         if (!pmc) {
 553                 rcu_read_lock();
 554                 for_each_pmc_rcu(in_dev, pmc) {
 555                         if (pmc->multiaddr == IGMP_ALL_HOSTS)
 556                                 continue;
 557                         if (ipv4_is_local_multicast(pmc->multiaddr) &&
 558                              !sysctl_igmp_llm_reports)
 559                                 continue;
 560                         spin_lock_bh(&pmc->lock);
 561                         if (pmc->sfcount[MCAST_EXCLUDE])
 562                                 type = IGMPV3_MODE_IS_EXCLUDE;
 563                         else
 564                                 type = IGMPV3_MODE_IS_INCLUDE;
 565                         skb = add_grec(skb, pmc, type, 0, 0);
 566                         spin_unlock_bh(&pmc->lock);
 567                 }
 568                 rcu_read_unlock();
 569         } else {
 570                 spin_lock_bh(&pmc->lock);
 571                 if (pmc->sfcount[MCAST_EXCLUDE])
 572                         type = IGMPV3_MODE_IS_EXCLUDE;
 573                 else
 574                         type = IGMPV3_MODE_IS_INCLUDE;
 575                 skb = add_grec(skb, pmc, type, 0, 0);
 576                 spin_unlock_bh(&pmc->lock);
 577         }
 578         if (!skb)
 579                 return 0;
 580         return igmpv3_sendpack(skb);
 581 }
 582
 583 /*
 584  * remove zero-count source records from a source filter list
 585  */
 586 static void igmpv3_clear_zeros(struct ip_sf_list **ppsf)
 587 {
 588         struct ip_sf_list *psf_prev, *psf_next, *psf;
 589
 590         psf_prev = NULL;
 591         for (psf = *ppsf; psf; psf = psf_next) {
 592                 psf_next = psf->sf_next;
 593                 if (psf->sf_crcount == 0) {
 594                         if (psf_prev)
 595                                 psf_prev->sf_next = psf->sf_next;
 596                         else
 597                                 *ppsf = psf->sf_next;
 598                         kfree(psf);
 599                 } else
 600                         psf_prev = psf;
 601         }
 602 }
 603
 604 static void igmpv3_send_cr(struct in_device *in_dev)
 605 {
 606         struct ip_mc_list *pmc, *pmc_prev, *pmc_next;
 607         struct sk_buff *skb = NULL;
 608         int type, dtype;
 609
 610         rcu_read_lock();
 611         spin_lock_bh(&in_dev->mc_tomb_lock);
 612
 613         /* deleted MCA's */
 614         pmc_prev = NULL;
 615         for (pmc = in_dev->mc_tomb; pmc; pmc = pmc_next) {
 616                 pmc_next = pmc->next;
 617                 if (pmc->sfmode == MCAST_INCLUDE) {
 618                         type = IGMPV3_BLOCK_OLD_SOURCES;
 619                         dtype = IGMPV3_BLOCK_OLD_SOURCES;
 620                         skb = add_grec(skb, pmc, type, 1, 0);
 621                         skb = add_grec(skb, pmc, dtype, 1, 1);
 622                 }
 623                 if (pmc->crcount) {
 624                         if (pmc->sfmode == MCAST_EXCLUDE) {
 625                                 type = IGMPV3_CHANGE_TO_INCLUDE;
 626                                 skb = add_grec(skb, pmc, type, 1, 0);
 627                         }
 628                         pmc->crcount--;
 629                         if (pmc->crcount == 0) {
 630                                 igmpv3_clear_zeros(&pmc->tomb);
 631                                 igmpv3_clear_zeros(&pmc->sources);
 632                         }
 633                 }
 634                 if (pmc->crcount == 0 && !pmc->tomb && !pmc->sources) {
 635                         if (pmc_prev)
 636                                 pmc_prev->next = pmc_next;
 637                         else
 638                                 in_dev->mc_tomb = pmc_next;
 639                         in_dev_put(pmc->interface);
 640                         kfree(pmc);
 641                 } else
 642                         pmc_prev = pmc;
 643         }
 644         spin_unlock_bh(&in_dev->mc_tomb_lock);
 645
 646         /* change recs */
 647         for_each_pmc_rcu(in_dev, pmc) {
 648                 spin_lock_bh(&pmc->lock);
 649                 if (pmc->sfcount[MCAST_EXCLUDE]) {
 650                         type = IGMPV3_BLOCK_OLD_SOURCES;
 651                         dtype = IGMPV3_ALLOW_NEW_SOURCES;
 652                 } else {
 653                         type = IGMPV3_ALLOW_NEW_SOURCES;
 654                         dtype = IGMPV3_BLOCK_OLD_SOURCES;
 655                 }
 656                 skb = add_grec(skb, pmc, type, 0, 0);
 657                 skb = add_grec(skb, pmc, dtype, 0, 1);  /* deleted sources */
 658
 659                 /* filter mode changes */
 660                 if (pmc->crcount) {
 661                         if (pmc->sfmode == MCAST_EXCLUDE)
 662                                 type = IGMPV3_CHANGE_TO_EXCLUDE;
 663                         else
 664                                 type = IGMPV3_CHANGE_TO_INCLUDE;
 665                         skb = add_grec(skb, pmc, type, 0, 0);
 666                         pmc->crcount--;
 667                 }
 668                 spin_unlock_bh(&pmc->lock);
 669         }
 670         rcu_read_unlock();
 671
 672         if (!skb)
 673                 return;
 674         (void) igmpv3_sendpack(skb);
 675 }
 676
 677 static int igmp_send_report(struct in_device *in_dev, struct ip_mc_list *pmc,
 678         int type)
 679 {
 680         struct sk_buff *skb;
 681         struct iphdr *iph;
 682         struct igmphdr *ih;
 683         struct rtable *rt;
 684         struct net_device *dev = in_dev->dev;
 685         struct net *net = dev_net(dev);
 686         __be32  group = pmc ? pmc->multiaddr : 0;
 687         struct flowi4 fl4;
 688         __be32  dst;
 689         int hlen, tlen;
 690
 691         if (type == IGMPV3_HOST_MEMBERSHIP_REPORT)
 692                 return igmpv3_send_report(in_dev, pmc);
 693
 694         if (ipv4_is_local_multicast(group) && !sysctl_igmp_llm_reports)
 695                 return 0;
 696
 697         if (type == IGMP_HOST_LEAVE_MESSAGE)
 698                 dst = IGMP_ALL_ROUTER;
 699         else
 700                 dst = group;
 701
 702         rt = ip_route_output_ports(net, &fl4, NULL, dst, 0,
 703                                    0, 0,
 704                                    IPPROTO_IGMP, 0, dev->ifindex);
 705         if (IS_ERR(rt))
 706                 return -1;
 707
 708         hlen = LL_RESERVED_SPACE(dev);
 709         tlen = dev->needed_tailroom;
 710         skb = alloc_skb(IGMP_SIZE + hlen + tlen, GFP_ATOMIC);
 711         if (!skb) {
 712                 ip_rt_put(rt);
 713                 return -1;
 714         }
 715         skb->priority = TC_PRIO_CONTROL;
 716
 717         skb_dst_set(skb, &rt->dst);
 718
 719         skb_reserve(skb, hlen);
 720
 721         skb_reset_network_header(skb);
 722         iph = ip_hdr(skb);
 723         skb_put(skb, sizeof(struct iphdr) + 4);
 724
 725         iph->version  = 4;
 726         iph->ihl      = (sizeof(struct iphdr)+4)>>2;
 727         iph->tos      = 0xc0;
 728         iph->frag_off = htons(IP_DF);
 729         iph->ttl      = 1;
 730         iph->daddr    = dst;
 731         iph->saddr    = fl4.saddr;
 732         iph->protocol = IPPROTO_IGMP;
 733         ip_select_ident(net, skb, NULL);
 734         ((u8 *)&iph[1])[0] = IPOPT_RA;
 735         ((u8 *)&iph[1])[1] = 4;
 736         ((u8 *)&iph[1])[2] = 0;
 737         ((u8 *)&iph[1])[3] = 0;
 738
 739         ih = (struct igmphdr *)skb_put(skb, sizeof(struct igmphdr));
 740         ih->type = type;
 741         ih->code = 0;
 742         ih->csum = 0;
 743         ih->group = group;
 744         ih->csum = ip_compute_csum((void *)ih, sizeof(struct igmphdr));
 745
 746         return ip_local_out(net, skb->sk, skb);
 747 }
 748
 749 static void igmp_gq_timer_expire(unsigned long data)
 750 {
 751         struct in_device *in_dev = (struct in_device *)data;
 752
 753         in_dev->mr_gq_running = 0;
 754         igmpv3_send_report(in_dev, NULL);
 755         in_dev_put(in_dev);
 756 }
 757
 758 static void igmp_ifc_timer_expire(unsigned long data)
 759 {
 760         struct in_device *in_dev = (struct in_device *)data;
 761
 762         igmpv3_send_cr(in_dev);
 763         if (in_dev->mr_ifc_count) {
 764                 in_dev->mr_ifc_count--;
 765                 igmp_ifc_start_timer(in_dev,
 766                                      unsolicited_report_interval(in_dev));
 767         }
 768         in_dev_put(in_dev);
 769 }
 770
 771 static void igmp_ifc_event(struct in_device *in_dev)
 772 {
 773         if (IGMP_V1_SEEN(in_dev) || IGMP_V2_SEEN(in_dev))
 774                 return;
 775         in_dev->mr_ifc_count = in_dev->mr_qrv ?: sysctl_igmp_qrv;
 776         igmp_ifc_start_timer(in_dev, 1);
 777 }
 778
 779
 780 static void igmp_timer_expire(unsigned long data)
 781 {
 782         struct ip_mc_list *im = (struct ip_mc_list *)data;
 783         struct in_device *in_dev = im->interface;
 784
 785         spin_lock(&im->lock);
 786         im->tm_running = 0;
 787
 788         if (im->unsolicit_count) {
 789                 im->unsolicit_count--;
 790                 igmp_start_timer(im, unsolicited_report_interval(in_dev));
 791         }
 792         im->reporter = 1;
 793         spin_unlock(&im->lock);
 794
 795         if (IGMP_V1_SEEN(in_dev))
 796                 igmp_send_report(in_dev, im, IGMP_HOST_MEMBERSHIP_REPORT);
 797         else if (IGMP_V2_SEEN(in_dev))
 798                 igmp_send_report(in_dev, im, IGMPV2_HOST_MEMBERSHIP_REPORT);
 799         else
 800                 igmp_send_report(in_dev, im, IGMPV3_HOST_MEMBERSHIP_REPORT);
 801
 802         ip_ma_put(im);
 803 }
 804
 805 /* mark EXCLUDE-mode sources */
 806 static int igmp_xmarksources(struct ip_mc_list *pmc, int nsrcs, __be32 *srcs)
 807 {
 808         struct ip_sf_list *psf;
 809         int i, scount;
 810
 811         scount = 0;
 812         for (psf = pmc->sources; psf; psf = psf->sf_next) {
 813                 if (scount == nsrcs)
 814                         break;
 815                 for (i = 0; i < nsrcs; i++) {
 816                         /* skip inactive filters */
 817                         if (psf->sf_count[MCAST_INCLUDE] ||
 818                             pmc->sfcount[MCAST_EXCLUDE] !=
 819                             psf->sf_count[MCAST_EXCLUDE])
 820                                 break;
 821                         if (srcs[i] == psf->sf_inaddr) {
 822                                 scount++;
 823                                 break;
 824                         }
 825                 }
 826         }
 827         pmc->gsquery = 0;
 828         if (scount == nsrcs)    /* all sources excluded */
 829                 return 0;
 830         return 1;
 831 }
 832
 833 static int igmp_marksources(struct ip_mc_list *pmc, int nsrcs, __be32 *srcs)
 834 {
 835         struct ip_sf_list *psf;
 836         int i, scount;
 837
 838         if (pmc->sfmode == MCAST_EXCLUDE)
 839                 return igmp_xmarksources(pmc, nsrcs, srcs);
 840
 841         /* mark INCLUDE-mode sources */
 842         scount = 0;
 843         for (psf = pmc->sources; psf; psf = psf->sf_next) {
 844                 if (scount == nsrcs)
 845                         break;
 846                 for (i = 0; i < nsrcs; i++)
 847                         if (srcs[i] == psf->sf_inaddr) {
 848                                 psf->sf_gsresp = 1;
 849                                 scount++;
 850                                 break;
 851                         }
 852         }
 853         if (!scount) {
 854                 pmc->gsquery = 0;
 855                 return 0;
 856         }
 857         pmc->gsquery = 1;
 858         return 1;
 859 }
 860
 861 /* return true if packet was dropped */
 862 static bool igmp_heard_report(struct in_device *in_dev, __be32 group)
 863 {
 864         struct ip_mc_list *im;
 865
 866         /* Timers are only set for non-local groups */
 867
 868         if (group == IGMP_ALL_HOSTS)
 869                 return false;
 870         if (ipv4_is_local_multicast(group) && !sysctl_igmp_llm_reports)
 871                 return false;
 872
 873         rcu_read_lock();
 874         for_each_pmc_rcu(in_dev, im) {
 875                 if (im->multiaddr == group) {
 876                         igmp_stop_timer(im);
 877                         break;
 878                 }
 879         }
 880         rcu_read_unlock();
 881         return false;
 882 }
 883
 884 /* return true if packet was dropped */
 885 static bool igmp_heard_query(struct in_device *in_dev, struct sk_buff *skb,
 886         int len)
 887 {
 888         struct igmphdr          *ih = igmp_hdr(skb);
 889         struct igmpv3_query *ih3 = igmpv3_query_hdr(skb);
 890         struct ip_mc_list       *im;
 891         __be32                  group = ih->group;
 892         int                     max_delay;
 893         int                     mark = 0;
 894
 895
 896         if (len == 8) {
 897                 if (ih->code == 0) {
 898                         /* Alas, old v1 router presents here. */
 899
 900                         max_delay = IGMP_QUERY_RESPONSE_INTERVAL;
 901                         in_dev->mr_v1_seen = jiffies +
 902                                 IGMP_V1_ROUTER_PRESENT_TIMEOUT;
 903                         group = 0;
 904                 } else {
 905                         /* v2 router present */
 906                         max_delay = ih->code*(HZ/IGMP_TIMER_SCALE);
 907                         in_dev->mr_v2_seen = jiffies +
 908                                 IGMP_V2_ROUTER_PRESENT_TIMEOUT;
 909                 }
 910                 /* cancel the interface change timer */
 911                 in_dev->mr_ifc_count = 0;
 912                 if (del_timer(&in_dev->mr_ifc_timer))
 913                         __in_dev_put(in_dev);
 914                 /* clear deleted report items */
 915                 igmpv3_clear_delrec(in_dev);
 916         } else if (len < 12) {
 917                 return true;    /* ignore bogus packet; freed by caller */
 918         } else if (IGMP_V1_SEEN(in_dev)) {
 919                 /* This is a v3 query with v1 queriers present */
 920                 max_delay = IGMP_QUERY_RESPONSE_INTERVAL;
 921                 group = 0;
 922         } else if (IGMP_V2_SEEN(in_dev)) {
 923                 /* this is a v3 query with v2 queriers present;
 924                  * Interpretation of the max_delay code is problematic here.
 925                  * A real v2 host would use ih_code directly, while v3 has a
 926                  * different encoding. We use the v3 encoding as more likely
 927                  * to be intended in a v3 query.
 928                  */
 929                 max_delay = IGMPV3_MRC(ih3->code)*(HZ/IGMP_TIMER_SCALE);
 930                 if (!max_delay)
 931                         max_delay = 1;  /* can't mod w/ 0 */
 932         } else { /* v3 */
 933                 if (!pskb_may_pull(skb, sizeof(struct igmpv3_query)))
 934                         return true;
 935
 936                 ih3 = igmpv3_query_hdr(skb);
 937                 if (ih3->nsrcs) {
 938                         if (!pskb_may_pull(skb, sizeof(struct igmpv3_query)
 939                                            + ntohs(ih3->nsrcs)*sizeof(__be32)))
 940                                 return true;
 941                         ih3 = igmpv3_query_hdr(skb);
 942                 }
 943
 944                 max_delay = IGMPV3_MRC(ih3->code)*(HZ/IGMP_TIMER_SCALE);
 945                 if (!max_delay)
 946                         max_delay = 1;  /* can't mod w/ 0 */
 947                 in_dev->mr_maxdelay = max_delay;
 948                 if (ih3->qrv)
 949                         in_dev->mr_qrv = ih3->qrv;
 950                 if (!group) { /* general query */
 951                         if (ih3->nsrcs)
 952                                 return true;    /* no sources allowed */
 953                         igmp_gq_start_timer(in_dev);
 954                         return false;
 955                 }
 956                 /* mark sources to include, if group & source-specific */
 957                 mark = ih3->nsrcs != 0;
 958         }
 959
 960         /*
 961          * - Start the timers in all of our membership records
 962          *   that the query applies to for the interface on
 963          *   which the query arrived excl. those that belong
 964          *   to a "local" group (224.0.0.X)
 965          * - For timers already running check if they need to
 966          *   be reset.
 967          * - Use the igmp->igmp_code field as the maximum
 968          *   delay possible
 969          */
 970         rcu_read_lock();
 971         for_each_pmc_rcu(in_dev, im) {
 972                 int changed;
 973
 974                 if (group && group != im->multiaddr)
 975                         continue;
 976                 if (im->multiaddr == IGMP_ALL_HOSTS)
 977                         continue;
 978                 if (ipv4_is_local_multicast(im->multiaddr) &&
 979                     !sysctl_igmp_llm_reports)
 980                         continue;
 981                 spin_lock_bh(&im->lock);
 982                 if (im->tm_running)
 983                         im->gsquery = im->gsquery && mark;
 984                 else
 985                         im->gsquery = mark;
 986                 changed = !im->gsquery ||
 987                         igmp_marksources(im, ntohs(ih3->nsrcs), ih3->srcs);
 988                 spin_unlock_bh(&im->lock);
 989                 if (changed)
 990                         igmp_mod_timer(im, max_delay);
 991         }
 992         rcu_read_unlock();
 993         return false;
 994 }
 995
 996 /* called in rcu_read_lock() section */
 997 int igmp_rcv(struct sk_buff *skb)
 998 {
 999         /* This basically follows the spec line by line -- see RFC1112 */
1000         struct igmphdr *ih;
1001         struct in_device *in_dev = __in_dev_get_rcu(skb->dev);
1002         int len = skb->len;
1003         bool dropped = true;
1004
1005         if (!in_dev)
1006                 goto drop;
1007
1008         if (!pskb_may_pull(skb, sizeof(struct igmphdr)))
1009                 goto drop;
1010
1011         if (skb_checksum_simple_validate(skb))
1012                 goto drop;
1013
1014         ih = igmp_hdr(skb);
1015         switch (ih->type) {
1016         case IGMP_HOST_MEMBERSHIP_QUERY:
1017                 dropped = igmp_heard_query(in_dev, skb, len);
1018                 break;
1019         case IGMP_HOST_MEMBERSHIP_REPORT:
1020         case IGMPV2_HOST_MEMBERSHIP_REPORT:
1021                 /* Is it our report looped back? */
1022                 if (rt_is_output_route(skb_rtable(skb)))
1023                         break;
1024                 /* don't rely on MC router hearing unicast reports */
1025                 if (skb->pkt_type == PACKET_MULTICAST ||
1026                     skb->pkt_type == PACKET_BROADCAST)
1027                         dropped = igmp_heard_report(in_dev, ih->group);
1028                 break;
1029         case IGMP_PIM:
1030 #ifdef CONFIG_IP_PIMSM_V1
1031                 return pim_rcv_v1(skb);
1032 #endif
1033         case IGMPV3_HOST_MEMBERSHIP_REPORT:
1034         case IGMP_DVMRP:
1035         case IGMP_TRACE:
1036         case IGMP_HOST_LEAVE_MESSAGE:
1037         case IGMP_MTRACE:
1038         case IGMP_MTRACE_RESP:
1039                 break;
1040         default:
1041                 break;
1042         }
1043
1044 drop:
1045         if (dropped)
1046                 kfree_skb(skb);
1047         else
1048                 consume_skb(skb);
1049         return 0;
1050 }
1051
1052 #endif
1053
1054
1055 /*
1056  *      Add a filter to a device
1057  */
1058
1059 static void ip_mc_filter_add(struct in_device *in_dev, __be32 addr)
1060 {
1061         char buf[MAX_ADDR_LEN];
1062         struct net_device *dev = in_dev->dev;
1063
1064         /* Checking for IFF_MULTICAST here is WRONG-WRONG-WRONG.
1065            We will get multicast token leakage, when IFF_MULTICAST
1066            is changed. This check should be done in ndo_set_rx_mode
1067            routine. Something sort of:
1068            if (dev->mc_list && dev->flags&IFF_MULTICAST) { do it; }
1069            --ANK
1070            */
1071         if (arp_mc_map(addr, buf, dev, 0) == 0)
1072                 dev_mc_add(dev, buf);
1073 }
1074
1075 /*
1076  *      Remove a filter from a device
1077  */
1078
1079 static void ip_mc_filter_del(struct in_device *in_dev, __be32 addr)
1080 {
1081         char buf[MAX_ADDR_LEN];
1082         struct net_device *dev = in_dev->dev;
1083
1084         if (arp_mc_map(addr, buf, dev, 0) == 0)
1085                 dev_mc_del(dev, buf);
1086 }
1087
1088 #ifdef CONFIG_IP_MULTICAST
1089 /*
1090  * deleted ip_mc_list manipulation
1091  */
1092 static void igmpv3_add_delrec(struct in_device *in_dev, struct ip_mc_list *im)
1093 {
1094         struct ip_mc_list *pmc;
1095
1096         /* this is an "ip_mc_list" for convenience; only the fields below
1097          * are actually used. In particular, the refcnt and users are not
1098          * used for management of the delete list. Using the same structure
1099          * for deleted items allows change reports to use common code with
1100          * non-deleted or query-response MCA's.
1101          */
1102         pmc = kzalloc(sizeof(*pmc), GFP_KERNEL);
1103         if (!pmc)
1104                 return;
1105         spin_lock_init(&pmc->lock);
1106         spin_lock_bh(&im->lock);
1107         pmc->interface = im->interface;
1108         in_dev_hold(in_dev);
1109         pmc->multiaddr = im->multiaddr;
1110         pmc->crcount = in_dev->mr_qrv ?: sysctl_igmp_qrv;
1111         pmc->sfmode = im->sfmode;
1112         if (pmc->sfmode == MCAST_INCLUDE) {
1113                 struct ip_sf_list *psf;
1114
1115                 pmc->tomb = im->tomb;
1116                 pmc->sources = im->sources;
1117                 im->tomb = im->sources = NULL;
1118                 for (psf = pmc->sources; psf; psf = psf->sf_next)
1119                         psf->sf_crcount = pmc->crcount;
1120         }
1121         spin_unlock_bh(&im->lock);
1122
1123         spin_lock_bh(&in_dev->mc_tomb_lock);
1124         pmc->next = in_dev->mc_tomb;
1125         in_dev->mc_tomb = pmc;
1126         spin_unlock_bh(&in_dev->mc_tomb_lock);
1127 }
1128
1129 static void igmpv3_del_delrec(struct in_device *in_dev, __be32 multiaddr)
1130 {
1131         struct ip_mc_list *pmc, *pmc_prev;
1132         struct ip_sf_list *psf, *psf_next;
1133
1134         spin_lock_bh(&in_dev->mc_tomb_lock);
1135         pmc_prev = NULL;
1136         for (pmc = in_dev->mc_tomb; pmc; pmc = pmc->next) {
1137                 if (pmc->multiaddr == multiaddr)
1138                         break;
1139                 pmc_prev = pmc;
1140         }
1141         if (pmc) {
1142                 if (pmc_prev)
1143                         pmc_prev->next = pmc->next;
1144                 else
1145                         in_dev->mc_tomb = pmc->next;
1146         }
1147         spin_unlock_bh(&in_dev->mc_tomb_lock);
1148         if (pmc) {
1149                 for (psf = pmc->tomb; psf; psf = psf_next) {
1150                         psf_next = psf->sf_next;
1151                         kfree(psf);
1152                 }
1153                 in_dev_put(pmc->interface);
1154                 kfree(pmc);
1155         }
1156 }
1157
1158 static void igmpv3_clear_delrec(struct in_device *in_dev)
1159 {
1160         struct ip_mc_list *pmc, *nextpmc;
1161
1162         spin_lock_bh(&in_dev->mc_tomb_lock);
1163         pmc = in_dev->mc_tomb;
1164         in_dev->mc_tomb = NULL;
1165         spin_unlock_bh(&in_dev->mc_tomb_lock);
1166
1167         for (; pmc; pmc = nextpmc) {
1168                 nextpmc = pmc->next;
1169                 ip_mc_clear_src(pmc);
1170                 in_dev_put(pmc->interface);
1171                 kfree(pmc);
1172         }
1173         /* clear dead sources, too */
1174         rcu_read_lock();
1175         for_each_pmc_rcu(in_dev, pmc) {
1176                 struct ip_sf_list *psf, *psf_next;
1177
1178                 spin_lock_bh(&pmc->lock);
1179                 psf = pmc->tomb;
1180                 pmc->tomb = NULL;
1181                 spin_unlock_bh(&pmc->lock);
1182                 for (; psf; psf = psf_next) {
1183                         psf_next = psf->sf_next;
1184                         kfree(psf);
1185                 }
1186         }
1187         rcu_read_unlock();
1188 }
1189 #endif
1190
1191 static void igmp_group_dropped(struct ip_mc_list *im)
1192 {
1193         struct in_device *in_dev = im->interface;
1194 #ifdef CONFIG_IP_MULTICAST
1195         int reporter;
1196 #endif
1197
1198         if (im->loaded) {
1199                 im->loaded = 0;
1200                 ip_mc_filter_del(in_dev, im->multiaddr);
1201         }
1202
1203 #ifdef CONFIG_IP_MULTICAST
1204         if (im->multiaddr == IGMP_ALL_HOSTS)
1205                 return;
1206         if (ipv4_is_local_multicast(im->multiaddr) && !sysctl_igmp_llm_reports)
1207                 return;
1208
1209         reporter = im->reporter;
1210         igmp_stop_timer(im);
1211
1212         if (!in_dev->dead) {
1213                 if (IGMP_V1_SEEN(in_dev))
1214                         return;
1215                 if (IGMP_V2_SEEN(in_dev)) {
1216                         if (reporter)
1217                                 igmp_send_report(in_dev, im, IGMP_HOST_LEAVE_MESSAGE);
1218                         return;
1219                 }
1220                 /* IGMPv3 */
1221                 igmpv3_add_delrec(in_dev, im);
1222
1223                 igmp_ifc_event(in_dev);
1224         }
1225 #endif
1226 }
1227
1228 static void igmp_group_added(struct ip_mc_list *im)
1229 {
1230         struct in_device *in_dev = im->interface;
1231
1232         if (im->loaded == 0) {
1233                 im->loaded = 1;
1234                 ip_mc_filter_add(in_dev, im->multiaddr);
1235         }
1236
1237 #ifdef CONFIG_IP_MULTICAST
1238         if (im->multiaddr == IGMP_ALL_HOSTS)
1239                 return;
1240         if (ipv4_is_local_multicast(im->multiaddr) && !sysctl_igmp_llm_reports)
1241                 return;
1242
1243         if (in_dev->dead)
1244                 return;
1245         if (IGMP_V1_SEEN(in_dev) || IGMP_V2_SEEN(in_dev)) {
1246                 spin_lock_bh(&im->lock);
1247                 igmp_start_timer(im, IGMP_INITIAL_REPORT_DELAY);
1248                 spin_unlock_bh(&im->lock);
1249                 return;
1250         }
1251         /* else, v3 */
1252
1253         im->crcount = in_dev->mr_qrv ?: sysctl_igmp_qrv;
1254         igmp_ifc_event(in_dev);
1255 #endif
1256 }
1257
1258
1259 /*
1260  *      Multicast list managers
1261  */
1262
1263 static u32 ip_mc_hash(const struct ip_mc_list *im)
1264 {
1265         return hash_32((__force u32)im->multiaddr, MC_HASH_SZ_LOG);
1266 }
1267
1268 static void ip_mc_hash_add(struct in_device *in_dev,
1269                            struct ip_mc_list *im)
1270 {
1271         struct ip_mc_list __rcu **mc_hash;
1272         u32 hash;
1273
1274         mc_hash = rtnl_dereference(in_dev->mc_hash);
1275         if (mc_hash) {
1276                 hash = ip_mc_hash(im);
1277                 im->next_hash = mc_hash[hash];
1278                 rcu_assign_pointer(mc_hash[hash], im);
1279                 return;
1280         }
1281
1282         /* do not use a hash table for small number of items */
1283         if (in_dev->mc_count < 4)
1284                 return;
1285
1286         mc_hash = kzalloc(sizeof(struct ip_mc_list *) << MC_HASH_SZ_LOG,
1287                           GFP_KERNEL);
1288         if (!mc_hash)
1289                 return;
1290
1291         for_each_pmc_rtnl(in_dev, im) {
1292                 hash = ip_mc_hash(im);
1293                 im->next_hash = mc_hash[hash];
1294                 RCU_INIT_POINTER(mc_hash[hash], im);
1295         }
1296
1297         rcu_assign_pointer(in_dev->mc_hash, mc_hash);
1298 }
1299
1300 static void ip_mc_hash_remove(struct in_device *in_dev,
1301                               struct ip_mc_list *im)
1302 {
1303         struct ip_mc_list __rcu **mc_hash = rtnl_dereference(in_dev->mc_hash);
1304         struct ip_mc_list *aux;
1305
1306         if (!mc_hash)
1307                 return;
1308         mc_hash += ip_mc_hash(im);
1309         while ((aux = rtnl_dereference(*mc_hash)) != im)
1310                 mc_hash = &aux->next_hash;
1311         *mc_hash = im->next_hash;
1312 }
1313
1314
1315 /*
1316  *      A socket has joined a multicast group on device dev.
1317  */
1318
1319 void ip_mc_inc_group(struct in_device *in_dev, __be32 addr)
1320 {
1321         struct ip_mc_list *im;
1322
1323         ASSERT_RTNL();
1324
1325         for_each_pmc_rtnl(in_dev, im) {
1326                 if (im->multiaddr == addr) {
1327                         im->users++;
1328                         ip_mc_add_src(in_dev, &addr, MCAST_EXCLUDE, 0, NULL, 0);
1329                         goto out;
1330                 }
1331         }
1332
1333         im = kzalloc(sizeof(*im), GFP_KERNEL);
1334         if (!im)
1335                 goto out;
1336
1337         im->users = 1;
1338         im->interface = in_dev;
1339         in_dev_hold(in_dev);
1340         im->multiaddr = addr;
1341         /* initial mode is (EX, empty) */
1342         im->sfmode = MCAST_EXCLUDE;
1343         im->sfcount[MCAST_EXCLUDE] = 1;
1344         atomic_set(&im->refcnt, 1);
1345         spin_lock_init(&im->lock);
1346 #ifdef CONFIG_IP_MULTICAST
1347         setup_timer(&im->timer, igmp_timer_expire, (unsigned long)im);
1348         im->unsolicit_count = sysctl_igmp_qrv;
1349 #endif
1350
1351         im->next_rcu = in_dev->mc_list;
1352         in_dev->mc_count++;
1353         rcu_assign_pointer(in_dev->mc_list, im);
1354
1355         ip_mc_hash_add(in_dev, im);
1356
1357 #ifdef CONFIG_IP_MULTICAST
1358         igmpv3_del_delrec(in_dev, im->multiaddr);
1359 #endif
1360         igmp_group_added(im);
1361         if (!in_dev->dead)
1362                 ip_rt_multicast_event(in_dev);
1363 out:
1364         return;
1365 }
1366 EXPORT_SYMBOL(ip_mc_inc_group);
1367
1368 static int ip_mc_check_iphdr(struct sk_buff *skb)
1369 {
1370         const struct iphdr *iph;
1371         unsigned int len;
1372         unsigned int offset = skb_network_offset(skb) + sizeof(*iph);
1373
1374         if (!pskb_may_pull(skb, offset))
1375                 return -EINVAL;
1376
1377         iph = ip_hdr(skb);
1378
1379         if (iph->version != 4 || ip_hdrlen(skb) < sizeof(*iph))
1380                 return -EINVAL;
1381
1382         offset += ip_hdrlen(skb) - sizeof(*iph);
1383
1384         if (!pskb_may_pull(skb, offset))
1385                 return -EINVAL;
1386
1387         iph = ip_hdr(skb);
1388
1389         if (unlikely(ip_fast_csum((u8 *)iph, iph->ihl)))
1390                 return -EINVAL;
1391
1392         len = skb_network_offset(skb) + ntohs(iph->tot_len);
1393         if (skb->len < len || len < offset)
1394                 return -EINVAL;
1395
1396         skb_set_transport_header(skb, offset);
1397
1398         return 0;
1399 }
1400
1401 static int ip_mc_check_igmp_reportv3(struct sk_buff *skb)
1402 {
1403         unsigned int len = skb_transport_offset(skb);
1404
1405         len += sizeof(struct igmpv3_report);
1406
1407         return pskb_may_pull(skb, len) ? 0 : -EINVAL;
1408 }
1409
1410 static int ip_mc_check_igmp_query(struct sk_buff *skb)
1411 {
1412         unsigned int len = skb_transport_offset(skb);
1413
1414         len += sizeof(struct igmphdr);
1415         if (skb->len < len)
1416                 return -EINVAL;
1417
1418         /* IGMPv{1,2}? */
1419         if (skb->len != len) {
1420                 /* or IGMPv3? */
1421                 len += sizeof(struct igmpv3_query) - sizeof(struct igmphdr);
1422                 if (skb->len < len || !pskb_may_pull(skb, len))
1423                         return -EINVAL;
1424         }
1425
1426         /* RFC2236+RFC3376 (IGMPv2+IGMPv3) require the multicast link layer
1427          * all-systems destination addresses (224.0.0.1) for general queries
1428          */
1429         if (!igmp_hdr(skb)->group &&
1430             ip_hdr(skb)->daddr != htonl(INADDR_ALLHOSTS_GROUP))
1431                 return -EINVAL;
1432
1433         return 0;
1434 }
1435
1436 static int ip_mc_check_igmp_msg(struct sk_buff *skb)
1437 {
1438         switch (igmp_hdr(skb)->type) {
1439         case IGMP_HOST_LEAVE_MESSAGE:
1440         case IGMP_HOST_MEMBERSHIP_REPORT:
1441         case IGMPV2_HOST_MEMBERSHIP_REPORT:
1442                 /* fall through */
1443                 return 0;
1444         case IGMPV3_HOST_MEMBERSHIP_REPORT:
1445                 return ip_mc_check_igmp_reportv3(skb);
1446         case IGMP_HOST_MEMBERSHIP_QUERY:
1447                 return ip_mc_check_igmp_query(skb);
1448         default:
1449                 return -ENOMSG;
1450         }
1451 }
1452
1453 static inline __sum16 ip_mc_validate_checksum(struct sk_buff *skb)
1454 {
1455         return skb_checksum_simple_validate(skb);
1456 }
1457
1458 static int __ip_mc_check_igmp(struct sk_buff *skb, struct sk_buff **skb_trimmed)
1459
1460 {
1461         struct sk_buff *skb_chk;
1462         unsigned int transport_len;
1463         unsigned int len = skb_transport_offset(skb) + sizeof(struct igmphdr);
1464         int ret = -EINVAL;
1465
1466         transport_len = ntohs(ip_hdr(skb)->tot_len) - ip_hdrlen(skb);
1467
1468         skb_chk = skb_checksum_trimmed(skb, transport_len,
1469                                        ip_mc_validate_checksum);
1470         if (!skb_chk)
1471                 goto err;
1472
1473         if (!pskb_may_pull(skb_chk, len))
1474                 goto err;
1475
1476         ret = ip_mc_check_igmp_msg(skb_chk);
1477         if (ret)
1478                 goto err;
1479
1480         if (skb_trimmed)
1481                 *skb_trimmed = skb_chk;
1482         /* free now unneeded clone */
1483         else if (skb_chk != skb)
1484                 kfree_skb(skb_chk);
1485
1486         ret = 0;
1487
1488 err:
1489         if (ret && skb_chk && skb_chk != skb)
1490                 kfree_skb(skb_chk);
1491
1492         return ret;
1493 }
1494
1495 /**
1496  * ip_mc_check_igmp - checks whether this is a sane IGMP packet
1497  * @skb: the skb to validate
1498  * @skb_trimmed: to store an skb pointer trimmed to IPv4 packet tail (optional)
1499  *
1500  * Checks whether an IPv4 packet is a valid IGMP packet. If so sets
1501  * skb transport header accordingly and returns zero.
1502  *
1503  * -EINVAL: A broken packet was detected, i.e. it violates some internet
1504  *  standard
1505  * -ENOMSG: IP header validation succeeded but it is not an IGMP packet.
1506  * -ENOMEM: A memory allocation failure happened.
1507  *
1508  * Optionally, an skb pointer might be provided via skb_trimmed (or set it
1509  * to NULL): After parsing an IGMP packet successfully it will point to
1510  * an skb which has its tail aligned to the IP packet end. This might
1511  * either be the originally provided skb or a trimmed, cloned version if
1512  * the skb frame had data beyond the IP packet. A cloned skb allows us
1513  * to leave the original skb and its full frame unchanged (which might be
1514  * desirable for layer 2 frame jugglers).
1515  *
1516  * Caller needs to set the skb network header and free any returned skb if it
1517  * differs from the provided skb.
1518  */
1519 int ip_mc_check_igmp(struct sk_buff *skb, struct sk_buff **skb_trimmed)
1520 {
1521         int ret = ip_mc_check_iphdr(skb);
1522
1523         if (ret < 0)
1524                 return ret;
1525
1526         if (ip_hdr(skb)->protocol != IPPROTO_IGMP)
1527                 return -ENOMSG;
1528
1529         return __ip_mc_check_igmp(skb, skb_trimmed);
1530 }
1531 EXPORT_SYMBOL(ip_mc_check_igmp);
1532
1533 /*
1534  *      Resend IGMP JOIN report; used by netdev notifier.
1535  */
1536 static void ip_mc_rejoin_groups(struct in_device *in_dev)
1537 {
1538 #ifdef CONFIG_IP_MULTICAST
1539         struct ip_mc_list *im;
1540         int type;
1541
1542         ASSERT_RTNL();
1543
1544         for_each_pmc_rtnl(in_dev, im) {
1545                 if (im->multiaddr == IGMP_ALL_HOSTS)
1546                         continue;
1547                 if (ipv4_is_local_multicast(im->multiaddr) &&
1548                     !sysctl_igmp_llm_reports)
1549                         continue;
1550
1551                 /* a failover is happening and switches
1552                  * must be notified immediately
1553                  */
1554                 if (IGMP_V1_SEEN(in_dev))
1555                         type = IGMP_HOST_MEMBERSHIP_REPORT;
1556                 else if (IGMP_V2_SEEN(in_dev))
1557                         type = IGMPV2_HOST_MEMBERSHIP_REPORT;
1558                 else
1559                         type = IGMPV3_HOST_MEMBERSHIP_REPORT;
1560                 igmp_send_report(in_dev, im, type);
1561         }
1562 #endif
1563 }
1564
1565 /*
1566  *      A socket has left a multicast group on device dev
1567  */
1568
1569 void ip_mc_dec_group(struct in_device *in_dev, __be32 addr)
1570 {
1571         struct ip_mc_list *i;
1572         struct ip_mc_list __rcu **ip;
1573
1574         ASSERT_RTNL();
1575
1576         for (ip = &in_dev->mc_list;
1577              (i = rtnl_dereference(*ip)) != NULL;
1578              ip = &i->next_rcu) {
1579                 if (i->multiaddr == addr) {
1580                         if (--i->users == 0) {
1581                                 ip_mc_hash_remove(in_dev, i);
1582                                 *ip = i->next_rcu;
1583                                 in_dev->mc_count--;
1584                                 igmp_group_dropped(i);
1585                                 ip_mc_clear_src(i);
1586
1587                                 if (!in_dev->dead)
1588                                         ip_rt_multicast_event(in_dev);
1589
1590                                 ip_ma_put(i);
1591                                 return;
1592                         }
1593                         break;
1594                 }
1595         }
1596 }
1597 EXPORT_SYMBOL(ip_mc_dec_group);
1598
1599 /* Device changing type */
1600
1601 void ip_mc_unmap(struct in_device *in_dev)
1602 {
1603         struct ip_mc_list *pmc;
1604
1605         ASSERT_RTNL();
1606
1607         for_each_pmc_rtnl(in_dev, pmc)
1608                 igmp_group_dropped(pmc);
1609 }
1610
1611 void ip_mc_remap(struct in_device *in_dev)
1612 {
1613         struct ip_mc_list *pmc;
1614
1615         ASSERT_RTNL();
1616
1617         for_each_pmc_rtnl(in_dev, pmc)
1618                 igmp_group_added(pmc);
1619 }
1620
1621 /* Device going down */
1622
1623 void ip_mc_down(struct in_device *in_dev)
1624 {
1625         struct ip_mc_list *pmc;
1626
1627         ASSERT_RTNL();
1628
1629         for_each_pmc_rtnl(in_dev, pmc)
1630                 igmp_group_dropped(pmc);
1631
1632 #ifdef CONFIG_IP_MULTICAST
1633         in_dev->mr_ifc_count = 0;
1634         if (del_timer(&in_dev->mr_ifc_timer))
1635                 __in_dev_put(in_dev);
1636         in_dev->mr_gq_running = 0;
1637         if (del_timer(&in_dev->mr_gq_timer))
1638                 __in_dev_put(in_dev);
1639         igmpv3_clear_delrec(in_dev);
1640 #endif
1641
1642         ip_mc_dec_group(in_dev, IGMP_ALL_HOSTS);
1643 }
1644
1645 void ip_mc_init_dev(struct in_device *in_dev)
1646 {
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 = 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
1666         ASSERT_RTNL();
1667
1668 #ifdef CONFIG_IP_MULTICAST
1669         in_dev->mr_qrv = sysctl_igmp_qrv;
1670 #endif
1671         ip_mc_inc_group(in_dev, IGMP_ALL_HOSTS);
1672
1673         for_each_pmc_rtnl(in_dev, pmc)
1674                 igmp_group_added(pmc);
1675 }
1676
1677 /*
1678  *      Device is about to be destroyed: clean up.
1679  */
1680
1681 void ip_mc_destroy_dev(struct in_device *in_dev)
1682 {
1683         struct ip_mc_list *i;
1684
1685         ASSERT_RTNL();
1686
1687         /* Deactivate timers */
1688         ip_mc_down(in_dev);
1689
1690         while ((i = rtnl_dereference(in_dev->mc_list)) != NULL) {
1691                 in_dev->mc_list = i->next_rcu;
1692                 in_dev->mc_count--;
1693
1694                 /* We've dropped the groups in ip_mc_down already */
1695                 ip_mc_clear_src(i);
1696                 ip_ma_put(i);
1697         }
1698 }
1699
1700 /* RTNL is locked */
1701 static struct in_device *ip_mc_find_dev(struct net *net, struct ip_mreqn *imr)
1702 {
1703         struct net_device *dev = NULL;
1704         struct in_device *idev = NULL;
1705
1706         if (imr->imr_ifindex) {
1707                 idev = inetdev_by_index(net, imr->imr_ifindex);
1708                 return idev;
1709         }
1710         if (imr->imr_address.s_addr) {
1711                 dev = __ip_dev_find(net, imr->imr_address.s_addr, false);
1712                 if (!dev)
1713                         return NULL;
1714         }
1715
1716         if (!dev) {
1717                 struct rtable *rt = ip_route_output(net,
1718                                                     imr->imr_multiaddr.s_addr,
1719                                                     0, 0, 0);
1720                 if (!IS_ERR(rt)) {
1721                         dev = rt->dst.dev;
1722                         ip_rt_put(rt);
1723                 }
1724         }
1725         if (dev) {
1726                 imr->imr_ifindex = dev->ifindex;
1727                 idev = __in_dev_get_rtnl(dev);
1728         }
1729         return idev;
1730 }
1731
1732 /*
1733  *      Join a socket to a group
1734  */
1735 int sysctl_igmp_max_memberships __read_mostly = IP_MAX_MEMBERSHIPS;
1736 int sysctl_igmp_max_msf __read_mostly = IP_MAX_MSF;
1737 #ifdef CONFIG_IP_MULTICAST
1738 int sysctl_igmp_qrv __read_mostly = IGMP_QUERY_ROBUSTNESS_VARIABLE;
1739 #endif
1740
1741 static int ip_mc_del1_src(struct ip_mc_list *pmc, int sfmode,
1742         __be32 *psfsrc)
1743 {
1744         struct ip_sf_list *psf, *psf_prev;
1745         int rv = 0;
1746
1747         psf_prev = NULL;
1748         for (psf = pmc->sources; psf; psf = psf->sf_next) {
1749                 if (psf->sf_inaddr == *psfsrc)
1750                         break;
1751                 psf_prev = psf;
1752         }
1753         if (!psf || psf->sf_count[sfmode] == 0) {
1754                 /* source filter not found, or count wrong =>  bug */
1755                 return -ESRCH;
1756         }
1757         psf->sf_count[sfmode]--;
1758         if (psf->sf_count[sfmode] == 0) {
1759                 ip_rt_multicast_event(pmc->interface);
1760         }
1761         if (!psf->sf_count[MCAST_INCLUDE] && !psf->sf_count[MCAST_EXCLUDE]) {
1762 #ifdef CONFIG_IP_MULTICAST
1763                 struct in_device *in_dev = pmc->interface;
1764 #endif
1765
1766                 /* no more filters for this source */
1767                 if (psf_prev)
1768                         psf_prev->sf_next = psf->sf_next;
1769                 else
1770                         pmc->sources = psf->sf_next;
1771 #ifdef CONFIG_IP_MULTICAST
1772                 if (psf->sf_oldin &&
1773                     !IGMP_V1_SEEN(in_dev) && !IGMP_V2_SEEN(in_dev)) {
1774                         psf->sf_crcount = in_dev->mr_qrv ?: sysctl_igmp_qrv;
1775                         psf->sf_next = pmc->tomb;
1776                         pmc->tomb = psf;
1777                         rv = 1;
1778                 } else
1779 #endif
1780                         kfree(psf);
1781         }
1782         return rv;
1783 }
1784
1785 #ifndef CONFIG_IP_MULTICAST
1786 #define igmp_ifc_event(x)       do { } while (0)
1787 #endif
1788
1789 static int ip_mc_del_src(struct in_device *in_dev, __be32 *pmca, int sfmode,
1790                          int sfcount, __be32 *psfsrc, int delta)
1791 {
1792         struct ip_mc_list *pmc;
1793         int     changerec = 0;
1794         int     i, err;
1795
1796         if (!in_dev)
1797                 return -ENODEV;
1798         rcu_read_lock();
1799         for_each_pmc_rcu(in_dev, pmc) {
1800                 if (*pmca == pmc->multiaddr)
1801                         break;
1802         }
1803         if (!pmc) {
1804                 /* MCA not found?? bug */
1805                 rcu_read_unlock();
1806                 return -ESRCH;
1807         }
1808         spin_lock_bh(&pmc->lock);
1809         rcu_read_unlock();
1810 #ifdef CONFIG_IP_MULTICAST
1811         sf_markstate(pmc);
1812 #endif
1813         if (!delta) {
1814                 err = -EINVAL;
1815                 if (!pmc->sfcount[sfmode])
1816                         goto out_unlock;
1817                 pmc->sfcount[sfmode]--;
1818         }
1819         err = 0;
1820         for (i = 0; i < sfcount; i++) {
1821                 int rv = ip_mc_del1_src(pmc, sfmode, &psfsrc[i]);
1822
1823                 changerec |= rv > 0;
1824                 if (!err && rv < 0)
1825                         err = rv;
1826         }
1827         if (pmc->sfmode == MCAST_EXCLUDE &&
1828             pmc->sfcount[MCAST_EXCLUDE] == 0 &&
1829             pmc->sfcount[MCAST_INCLUDE]) {
1830 #ifdef CONFIG_IP_MULTICAST
1831                 struct ip_sf_list *psf;
1832 #endif
1833
1834                 /* filter mode change */
1835                 pmc->sfmode = MCAST_INCLUDE;
1836 #ifdef CONFIG_IP_MULTICAST
1837                 pmc->crcount = in_dev->mr_qrv ?: sysctl_igmp_qrv;
1838                 in_dev->mr_ifc_count = pmc->crcount;
1839                 for (psf = pmc->sources; psf; psf = psf->sf_next)
1840                         psf->sf_crcount = 0;
1841                 igmp_ifc_event(pmc->interface);
1842         } else if (sf_setstate(pmc) || changerec) {
1843                 igmp_ifc_event(pmc->interface);
1844 #endif
1845         }
1846 out_unlock:
1847         spin_unlock_bh(&pmc->lock);
1848         return err;
1849 }
1850
1851 /*
1852  * Add multicast single-source filter to the interface list
1853  */
1854 static int ip_mc_add1_src(struct ip_mc_list *pmc, int sfmode,
1855         __be32 *psfsrc)
1856 {
1857         struct ip_sf_list *psf, *psf_prev;
1858
1859         psf_prev = NULL;
1860         for (psf = pmc->sources; psf; psf = psf->sf_next) {
1861                 if (psf->sf_inaddr == *psfsrc)
1862                         break;
1863                 psf_prev = psf;
1864         }
1865         if (!psf) {
1866                 psf = kzalloc(sizeof(*psf), GFP_ATOMIC);
1867                 if (!psf)
1868                         return -ENOBUFS;
1869                 psf->sf_inaddr = *psfsrc;
1870                 if (psf_prev) {
1871                         psf_prev->sf_next = psf;
1872                 } else
1873                         pmc->sources = psf;
1874         }
1875         psf->sf_count[sfmode]++;
1876         if (psf->sf_count[sfmode] == 1) {
1877                 ip_rt_multicast_event(pmc->interface);
1878         }
1879         return 0;
1880 }
1881
1882 #ifdef CONFIG_IP_MULTICAST
1883 static void sf_markstate(struct ip_mc_list *pmc)
1884 {
1885         struct ip_sf_list *psf;
1886         int mca_xcount = pmc->sfcount[MCAST_EXCLUDE];
1887
1888         for (psf = pmc->sources; psf; psf = psf->sf_next)
1889                 if (pmc->sfcount[MCAST_EXCLUDE]) {
1890                         psf->sf_oldin = mca_xcount ==
1891                                 psf->sf_count[MCAST_EXCLUDE] &&
1892                                 !psf->sf_count[MCAST_INCLUDE];
1893                 } else
1894                         psf->sf_oldin = psf->sf_count[MCAST_INCLUDE] != 0;
1895 }
1896
1897 static int sf_setstate(struct ip_mc_list *pmc)
1898 {
1899         struct ip_sf_list *psf, *dpsf;
1900         int mca_xcount = pmc->sfcount[MCAST_EXCLUDE];
1901         int qrv = pmc->interface->mr_qrv;
1902         int new_in, rv;
1903
1904         rv = 0;
1905         for (psf = pmc->sources; psf; psf = psf->sf_next) {
1906                 if (pmc->sfcount[MCAST_EXCLUDE]) {
1907                         new_in = mca_xcount == psf->sf_count[MCAST_EXCLUDE] &&
1908                                 !psf->sf_count[MCAST_INCLUDE];
1909                 } else
1910                         new_in = psf->sf_count[MCAST_INCLUDE] != 0;
1911                 if (new_in) {
1912                         if (!psf->sf_oldin) {
1913                                 struct ip_sf_list *prev = NULL;
1914
1915                                 for (dpsf = pmc->tomb; dpsf; dpsf = dpsf->sf_next) {
1916                                         if (dpsf->sf_inaddr == psf->sf_inaddr)
1917                                                 break;
1918                                         prev = dpsf;
1919                                 }
1920                                 if (dpsf) {
1921                                         if (prev)
1922                                                 prev->sf_next = dpsf->sf_next;
1923                                         else
1924                                                 pmc->tomb = dpsf->sf_next;
1925                                         kfree(dpsf);
1926                                 }
1927                                 psf->sf_crcount = qrv;
1928                                 rv++;
1929                         }
1930                 } else if (psf->sf_oldin) {
1931
1932                         psf->sf_crcount = 0;
1933                         /*
1934                          * add or update "delete" records if an active filter
1935                          * is now inactive
1936                          */
1937                         for (dpsf = pmc->tomb; dpsf; dpsf = dpsf->sf_next)
1938                                 if (dpsf->sf_inaddr == psf->sf_inaddr)
1939                                         break;
1940                         if (!dpsf) {
1941                                 dpsf = kmalloc(sizeof(*dpsf), GFP_ATOMIC);
1942                                 if (!dpsf)
1943                                         continue;
1944                                 *dpsf = *psf;
1945                                 /* pmc->lock held by callers */
1946                                 dpsf->sf_next = pmc->tomb;
1947                                 pmc->tomb = dpsf;
1948                         }
1949                         dpsf->sf_crcount = qrv;
1950                         rv++;
1951                 }
1952         }
1953         return rv;
1954 }
1955 #endif
1956
1957 /*
1958  * Add multicast source filter list to the interface list
1959  */
1960 static int ip_mc_add_src(struct in_device *in_dev, __be32 *pmca, int sfmode,
1961                          int sfcount, __be32 *psfsrc, int delta)
1962 {
1963         struct ip_mc_list *pmc;
1964         int     isexclude;
1965         int     i, err;
1966
1967         if (!in_dev)
1968                 return -ENODEV;
1969         rcu_read_lock();
1970         for_each_pmc_rcu(in_dev, pmc) {
1971                 if (*pmca == pmc->multiaddr)
1972                         break;
1973         }
1974         if (!pmc) {
1975                 /* MCA not found?? bug */
1976                 rcu_read_unlock();
1977                 return -ESRCH;
1978         }
1979         spin_lock_bh(&pmc->lock);
1980         rcu_read_unlock();
1981
1982 #ifdef CONFIG_IP_MULTICAST
1983         sf_markstate(pmc);
1984 #endif
1985         isexclude = pmc->sfmode == MCAST_EXCLUDE;
1986         if (!delta)
1987                 pmc->sfcount[sfmode]++;
1988         err = 0;
1989         for (i = 0; i < sfcount; i++) {
1990                 err = ip_mc_add1_src(pmc, sfmode, &psfsrc[i]);
1991                 if (err)
1992                         break;
1993         }
1994         if (err) {
1995                 int j;
1996
1997                 if (!delta)
1998                         pmc->sfcount[sfmode]--;
1999                 for (j = 0; j < i; j++)
2000                         (void) ip_mc_del1_src(pmc, sfmode, &psfsrc[j]);
2001         } else if (isexclude != (pmc->sfcount[MCAST_EXCLUDE] != 0)) {
2002 #ifdef CONFIG_IP_MULTICAST
2003                 struct ip_sf_list *psf;
2004                 in_dev = pmc->interface;
2005 #endif
2006
2007                 /* filter mode change */
2008                 if (pmc->sfcount[MCAST_EXCLUDE])
2009                         pmc->sfmode = MCAST_EXCLUDE;
2010                 else if (pmc->sfcount[MCAST_INCLUDE])
2011                         pmc->sfmode = MCAST_INCLUDE;
2012 #ifdef CONFIG_IP_MULTICAST
2013                 /* else no filters; keep old mode for reports */
2014
2015                 pmc->crcount = in_dev->mr_qrv ?: sysctl_igmp_qrv;
2016                 in_dev->mr_ifc_count = pmc->crcount;
2017                 for (psf = pmc->sources; psf; psf = psf->sf_next)
2018                         psf->sf_crcount = 0;
2019                 igmp_ifc_event(in_dev);
2020         } else if (sf_setstate(pmc)) {
2021                 igmp_ifc_event(in_dev);
2022 #endif
2023         }
2024         spin_unlock_bh(&pmc->lock);
2025         return err;
2026 }
2027
2028 static void ip_mc_clear_src(struct ip_mc_list *pmc)
2029 {
2030         struct ip_sf_list *psf, *nextpsf, *tomb, *sources;
2031
2032         spin_lock_bh(&pmc->lock);
2033         tomb = pmc->tomb;
2034         pmc->tomb = NULL;
2035         sources = pmc->sources;
2036         pmc->sources = NULL;
2037         pmc->sfmode = MCAST_EXCLUDE;
2038         pmc->sfcount[MCAST_INCLUDE] = 0;
2039         pmc->sfcount[MCAST_EXCLUDE] = 1;
2040         spin_unlock_bh(&pmc->lock);
2041
2042         for (psf = tomb; psf; psf = nextpsf) {
2043                 nextpsf = psf->sf_next;
2044                 kfree(psf);
2045         }
2046         for (psf = sources; psf; psf = nextpsf) {
2047                 nextpsf = psf->sf_next;
2048                 kfree(psf);
2049         }
2050 }
2051
2052 /* Join a multicast group
2053  */
2054
2055 int ip_mc_join_group(struct sock *sk, struct ip_mreqn *imr)
2056 {
2057         __be32 addr = imr->imr_multiaddr.s_addr;
2058         struct ip_mc_socklist *iml, *i;
2059         struct in_device *in_dev;
2060         struct inet_sock *inet = inet_sk(sk);
2061         struct net *net = sock_net(sk);
2062         int ifindex;
2063         int count = 0;
2064         int err;
2065
2066         ASSERT_RTNL();
2067
2068         if (!ipv4_is_multicast(addr))
2069                 return -EINVAL;
2070
2071         in_dev = ip_mc_find_dev(net, imr);
2072
2073         if (!in_dev) {
2074                 err = -ENODEV;
2075                 goto done;
2076         }
2077
2078         err = -EADDRINUSE;
2079         ifindex = imr->imr_ifindex;
2080         for_each_pmc_rtnl(inet, i) {
2081                 if (i->multi.imr_multiaddr.s_addr == addr &&
2082                     i->multi.imr_ifindex == ifindex)
2083                         goto done;
2084                 count++;
2085         }
2086         err = -ENOBUFS;
2087         if (count >= sysctl_igmp_max_memberships)
2088                 goto done;
2089         iml = sock_kmalloc(sk, sizeof(*iml), GFP_KERNEL);
2090         if (!iml)
2091                 goto done;
2092
2093         memcpy(&iml->multi, imr, sizeof(*imr));
2094         iml->next_rcu = inet->mc_list;
2095         iml->sflist = NULL;
2096         iml->sfmode = MCAST_EXCLUDE;
2097         rcu_assign_pointer(inet->mc_list, iml);
2098         ip_mc_inc_group(in_dev, addr);
2099         err = 0;
2100 done:
2101         return err;
2102 }
2103 EXPORT_SYMBOL(ip_mc_join_group);
2104
2105 static int ip_mc_leave_src(struct sock *sk, struct ip_mc_socklist *iml,
2106                            struct in_device *in_dev)
2107 {
2108         struct ip_sf_socklist *psf = rtnl_dereference(iml->sflist);
2109         int err;
2110
2111         if (!psf) {
2112                 /* any-source empty exclude case */
2113                 return ip_mc_del_src(in_dev, &iml->multi.imr_multiaddr.s_addr,
2114                         iml->sfmode, 0, NULL, 0);
2115         }
2116         err = ip_mc_del_src(in_dev, &iml->multi.imr_multiaddr.s_addr,
2117                         iml->sfmode, psf->sl_count, psf->sl_addr, 0);
2118         RCU_INIT_POINTER(iml->sflist, NULL);
2119         /* decrease mem now to avoid the memleak warning */
2120         atomic_sub(IP_SFLSIZE(psf->sl_max), &sk->sk_omem_alloc);
2121         kfree_rcu(psf, rcu);
2122         return err;
2123 }
2124
2125 int ip_mc_leave_group(struct sock *sk, struct ip_mreqn *imr)
2126 {
2127         struct inet_sock *inet = inet_sk(sk);
2128         struct ip_mc_socklist *iml;
2129         struct ip_mc_socklist __rcu **imlp;
2130         struct in_device *in_dev;
2131         struct net *net = sock_net(sk);
2132         __be32 group = imr->imr_multiaddr.s_addr;
2133         u32 ifindex;
2134         int ret = -EADDRNOTAVAIL;
2135
2136         ASSERT_RTNL();
2137
2138         in_dev = ip_mc_find_dev(net, imr);
2139         if (!imr->imr_ifindex && !imr->imr_address.s_addr && !in_dev) {
2140                 ret = -ENODEV;
2141                 goto out;
2142         }
2143         ifindex = imr->imr_ifindex;
2144         for (imlp = &inet->mc_list;
2145              (iml = rtnl_dereference(*imlp)) != NULL;
2146              imlp = &iml->next_rcu) {
2147                 if (iml->multi.imr_multiaddr.s_addr != group)
2148                         continue;
2149                 if (ifindex) {
2150                         if (iml->multi.imr_ifindex != ifindex)
2151                                 continue;
2152                 } else if (imr->imr_address.s_addr && imr->imr_address.s_addr !=
2153                                 iml->multi.imr_address.s_addr)
2154                         continue;
2155
2156                 (void) ip_mc_leave_src(sk, iml, in_dev);
2157
2158                 *imlp = iml->next_rcu;
2159
2160                 if (in_dev)
2161                         ip_mc_dec_group(in_dev, group);
2162
2163                 /* decrease mem now to avoid the memleak warning */
2164                 atomic_sub(sizeof(*iml), &sk->sk_omem_alloc);
2165                 kfree_rcu(iml, rcu);
2166                 return 0;
2167         }
2168 out:
2169         return ret;
2170 }
2171 EXPORT_SYMBOL(ip_mc_leave_group);
2172
2173 int ip_mc_source(int add, int omode, struct sock *sk, struct
2174         ip_mreq_source *mreqs, int ifindex)
2175 {
2176         int err;
2177         struct ip_mreqn imr;
2178         __be32 addr = mreqs->imr_multiaddr;
2179         struct ip_mc_socklist *pmc;
2180         struct in_device *in_dev = NULL;
2181         struct inet_sock *inet = inet_sk(sk);
2182         struct ip_sf_socklist *psl;
2183         struct net *net = sock_net(sk);
2184         int leavegroup = 0;
2185         int i, j, rv;
2186
2187         if (!ipv4_is_multicast(addr))
2188                 return -EINVAL;
2189
2190         ASSERT_RTNL();
2191
2192         imr.imr_multiaddr.s_addr = mreqs->imr_multiaddr;
2193         imr.imr_address.s_addr = mreqs->imr_interface;
2194         imr.imr_ifindex = ifindex;
2195         in_dev = ip_mc_find_dev(net, &imr);
2196
2197         if (!in_dev) {
2198                 err = -ENODEV;
2199                 goto done;
2200         }
2201         err = -EADDRNOTAVAIL;
2202
2203         for_each_pmc_rtnl(inet, pmc) {
2204                 if ((pmc->multi.imr_multiaddr.s_addr ==
2205                      imr.imr_multiaddr.s_addr) &&
2206                     (pmc->multi.imr_ifindex == imr.imr_ifindex))
2207                         break;
2208         }
2209         if (!pmc) {             /* must have a prior join */
2210                 err = -EINVAL;
2211                 goto done;
2212         }
2213         /* if a source filter was set, must be the same mode as before */
2214         if (pmc->sflist) {
2215                 if (pmc->sfmode != omode) {
2216                         err = -EINVAL;
2217                         goto done;
2218                 }
2219         } else if (pmc->sfmode != omode) {
2220                 /* allow mode switches for empty-set filters */
2221                 ip_mc_add_src(in_dev, &mreqs->imr_multiaddr, omode, 0, NULL, 0);
2222                 ip_mc_del_src(in_dev, &mreqs->imr_multiaddr, pmc->sfmode, 0,
2223                         NULL, 0);
2224                 pmc->sfmode = omode;
2225         }
2226
2227         psl = rtnl_dereference(pmc->sflist);
2228         if (!add) {
2229                 if (!psl)
2230                         goto done;      /* err = -EADDRNOTAVAIL */
2231                 rv = !0;
2232                 for (i = 0; i < psl->sl_count; i++) {
2233                         rv = memcmp(&psl->sl_addr[i], &mreqs->imr_sourceaddr,
2234                                 sizeof(__be32));
2235                         if (rv == 0)
2236                                 break;
2237                 }
2238                 if (rv)         /* source not found */
2239                         goto done;      /* err = -EADDRNOTAVAIL */
2240
2241                 /* special case - (INCLUDE, empty) == LEAVE_GROUP */
2242                 if (psl->sl_count == 1 && omode == MCAST_INCLUDE) {
2243                         leavegroup = 1;
2244                         goto done;
2245                 }
2246
2247                 /* update the interface filter */
2248                 ip_mc_del_src(in_dev, &mreqs->imr_multiaddr, omode, 1,
2249                         &mreqs->imr_sourceaddr, 1);
2250
2251                 for (j = i+1; j < psl->sl_count; j++)
2252                         psl->sl_addr[j-1] = psl->sl_addr[j];
2253                 psl->sl_count--;
2254                 err = 0;
2255                 goto done;
2256         }
2257         /* else, add a new source to the filter */
2258
2259         if (psl && psl->sl_count >= sysctl_igmp_max_msf) {
2260                 err = -ENOBUFS;
2261                 goto done;
2262         }
2263         if (!psl || psl->sl_count == psl->sl_max) {
2264                 struct ip_sf_socklist *newpsl;
2265                 int count = IP_SFBLOCK;
2266
2267                 if (psl)
2268                         count += psl->sl_max;
2269                 newpsl = sock_kmalloc(sk, IP_SFLSIZE(count), GFP_KERNEL);
2270                 if (!newpsl) {
2271                         err = -ENOBUFS;
2272                         goto done;
2273                 }
2274                 newpsl->sl_max = count;
2275                 newpsl->sl_count = count - IP_SFBLOCK;
2276                 if (psl) {
2277                         for (i = 0; i < psl->sl_count; i++)
2278                                 newpsl->sl_addr[i] = psl->sl_addr[i];
2279                         /* decrease mem now to avoid the memleak warning */
2280                         atomic_sub(IP_SFLSIZE(psl->sl_max), &sk->sk_omem_alloc);
2281                         kfree_rcu(psl, rcu);
2282                 }
2283                 rcu_assign_pointer(pmc->sflist, newpsl);
2284                 psl = newpsl;
2285         }
2286         rv = 1; /* > 0 for insert logic below if sl_count is 0 */
2287         for (i = 0; i < psl->sl_count; i++) {
2288                 rv = memcmp(&psl->sl_addr[i], &mreqs->imr_sourceaddr,
2289                         sizeof(__be32));
2290                 if (rv == 0)
2291                         break;
2292         }
2293         if (rv == 0)            /* address already there is an error */
2294                 goto done;
2295         for (j = psl->sl_count-1; j >= i; j--)
2296                 psl->sl_addr[j+1] = psl->sl_addr[j];
2297         psl->sl_addr[i] = mreqs->imr_sourceaddr;
2298         psl->sl_count++;
2299         err = 0;
2300         /* update the interface list */
2301         ip_mc_add_src(in_dev, &mreqs->imr_multiaddr, omode, 1,
2302                 &mreqs->imr_sourceaddr, 1);
2303 done:
2304         if (leavegroup)
2305                 err = ip_mc_leave_group(sk, &imr);
2306         return err;
2307 }
2308
2309 int ip_mc_msfilter(struct sock *sk, struct ip_msfilter *msf, int ifindex)
2310 {
2311         int err = 0;
2312         struct ip_mreqn imr;
2313         __be32 addr = msf->imsf_multiaddr;
2314         struct ip_mc_socklist *pmc;
2315         struct in_device *in_dev;
2316         struct inet_sock *inet = inet_sk(sk);
2317         struct ip_sf_socklist *newpsl, *psl;
2318         struct net *net = sock_net(sk);
2319         int leavegroup = 0;
2320
2321         if (!ipv4_is_multicast(addr))
2322                 return -EINVAL;
2323         if (msf->imsf_fmode != MCAST_INCLUDE &&
2324             msf->imsf_fmode != MCAST_EXCLUDE)
2325                 return -EINVAL;
2326
2327         ASSERT_RTNL();
2328
2329         imr.imr_multiaddr.s_addr = msf->imsf_multiaddr;
2330         imr.imr_address.s_addr = msf->imsf_interface;
2331         imr.imr_ifindex = ifindex;
2332         in_dev = ip_mc_find_dev(net, &imr);
2333
2334         if (!in_dev) {
2335                 err = -ENODEV;
2336                 goto done;
2337         }
2338
2339         /* special case - (INCLUDE, empty) == LEAVE_GROUP */
2340         if (msf->imsf_fmode == MCAST_INCLUDE && msf->imsf_numsrc == 0) {
2341                 leavegroup = 1;
2342                 goto done;
2343         }
2344
2345         for_each_pmc_rtnl(inet, pmc) {
2346                 if (pmc->multi.imr_multiaddr.s_addr == msf->imsf_multiaddr &&
2347                     pmc->multi.imr_ifindex == imr.imr_ifindex)
2348                         break;
2349         }
2350         if (!pmc) {             /* must have a prior join */
2351                 err = -EINVAL;
2352                 goto done;
2353         }
2354         if (msf->imsf_numsrc) {
2355                 newpsl = sock_kmalloc(sk, IP_SFLSIZE(msf->imsf_numsrc),
2356                                                            GFP_KERNEL);
2357                 if (!newpsl) {
2358                         err = -ENOBUFS;
2359                         goto done;
2360                 }
2361                 newpsl->sl_max = newpsl->sl_count = msf->imsf_numsrc;
2362                 memcpy(newpsl->sl_addr, msf->imsf_slist,
2363                         msf->imsf_numsrc * sizeof(msf->imsf_slist[0]));
2364                 err = ip_mc_add_src(in_dev, &msf->imsf_multiaddr,
2365                         msf->imsf_fmode, newpsl->sl_count, newpsl->sl_addr, 0);
2366                 if (err) {
2367                         sock_kfree_s(sk, newpsl, IP_SFLSIZE(newpsl->sl_max));
2368                         goto done;
2369                 }
2370         } else {
2371                 newpsl = NULL;
2372                 (void) ip_mc_add_src(in_dev, &msf->imsf_multiaddr,
2373                                      msf->imsf_fmode, 0, NULL, 0);
2374         }
2375         psl = rtnl_dereference(pmc->sflist);
2376         if (psl) {
2377                 (void) ip_mc_del_src(in_dev, &msf->imsf_multiaddr, pmc->sfmode,
2378                         psl->sl_count, psl->sl_addr, 0);
2379                 /* decrease mem now to avoid the memleak warning */
2380                 atomic_sub(IP_SFLSIZE(psl->sl_max), &sk->sk_omem_alloc);
2381                 kfree_rcu(psl, rcu);
2382         } else
2383                 (void) ip_mc_del_src(in_dev, &msf->imsf_multiaddr, pmc->sfmode,
2384                         0, NULL, 0);
2385         rcu_assign_pointer(pmc->sflist, newpsl);
2386         pmc->sfmode = msf->imsf_fmode;
2387         err = 0;
2388 done:
2389         if (leavegroup)
2390                 err = ip_mc_leave_group(sk, &imr);
2391         return err;
2392 }
2393
2394 int ip_mc_msfget(struct sock *sk, struct ip_msfilter *msf,
2395         struct ip_msfilter __user *optval, int __user *optlen)
2396 {
2397         int err, len, count, copycount;
2398         struct ip_mreqn imr;
2399         __be32 addr = msf->imsf_multiaddr;
2400         struct ip_mc_socklist *pmc;
2401         struct in_device *in_dev;
2402         struct inet_sock *inet = inet_sk(sk);
2403         struct ip_sf_socklist *psl;
2404         struct net *net = sock_net(sk);
2405
2406         ASSERT_RTNL();
2407
2408         if (!ipv4_is_multicast(addr))
2409                 return -EINVAL;
2410
2411         imr.imr_multiaddr.s_addr = msf->imsf_multiaddr;
2412         imr.imr_address.s_addr = msf->imsf_interface;
2413         imr.imr_ifindex = 0;
2414         in_dev = ip_mc_find_dev(net, &imr);
2415
2416         if (!in_dev) {
2417                 err = -ENODEV;
2418                 goto done;
2419         }
2420         err = -EADDRNOTAVAIL;
2421
2422         for_each_pmc_rtnl(inet, pmc) {
2423                 if (pmc->multi.imr_multiaddr.s_addr == msf->imsf_multiaddr &&
2424                     pmc->multi.imr_ifindex == imr.imr_ifindex)
2425                         break;
2426         }
2427         if (!pmc)               /* must have a prior join */
2428                 goto done;
2429         msf->imsf_fmode = pmc->sfmode;
2430         psl = rtnl_dereference(pmc->sflist);
2431         if (!psl) {
2432                 len = 0;
2433                 count = 0;
2434         } else {
2435                 count = psl->sl_count;
2436         }
2437         copycount = count < msf->imsf_numsrc ? count : msf->imsf_numsrc;
2438         len = copycount * sizeof(psl->sl_addr[0]);
2439         msf->imsf_numsrc = count;
2440         if (put_user(IP_MSFILTER_SIZE(copycount), optlen) ||
2441             copy_to_user(optval, msf, IP_MSFILTER_SIZE(0))) {
2442                 return -EFAULT;
2443         }
2444         if (len &&
2445             copy_to_user(&optval->imsf_slist[0], psl->sl_addr, len))
2446                 return -EFAULT;
2447         return 0;
2448 done:
2449         return err;
2450 }
2451
2452 int ip_mc_gsfget(struct sock *sk, struct group_filter *gsf,
2453         struct group_filter __user *optval, int __user *optlen)
2454 {
2455         int err, i, count, copycount;
2456         struct sockaddr_in *psin;
2457         __be32 addr;
2458         struct ip_mc_socklist *pmc;
2459         struct inet_sock *inet = inet_sk(sk);
2460         struct ip_sf_socklist *psl;
2461
2462         ASSERT_RTNL();
2463
2464         psin = (struct sockaddr_in *)&gsf->gf_group;
2465         if (psin->sin_family != AF_INET)
2466                 return -EINVAL;
2467         addr = psin->sin_addr.s_addr;
2468         if (!ipv4_is_multicast(addr))
2469                 return -EINVAL;
2470
2471         err = -EADDRNOTAVAIL;
2472
2473         for_each_pmc_rtnl(inet, pmc) {
2474                 if (pmc->multi.imr_multiaddr.s_addr == addr &&
2475                     pmc->multi.imr_ifindex == gsf->gf_interface)
2476                         break;
2477         }
2478         if (!pmc)               /* must have a prior join */
2479                 goto done;
2480         gsf->gf_fmode = pmc->sfmode;
2481         psl = rtnl_dereference(pmc->sflist);
2482         count = psl ? psl->sl_count : 0;
2483         copycount = count < gsf->gf_numsrc ? count : gsf->gf_numsrc;
2484         gsf->gf_numsrc = count;
2485         if (put_user(GROUP_FILTER_SIZE(copycount), optlen) ||
2486             copy_to_user(optval, gsf, GROUP_FILTER_SIZE(0))) {
2487                 return -EFAULT;
2488         }
2489         for (i = 0; i < copycount; i++) {
2490                 struct sockaddr_storage ss;
2491
2492                 psin = (struct sockaddr_in *)&ss;
2493                 memset(&ss, 0, sizeof(ss));
2494                 psin->sin_family = AF_INET;
2495                 psin->sin_addr.s_addr = psl->sl_addr[i];
2496                 if (copy_to_user(&optval->gf_slist[i], &ss, sizeof(ss)))
2497                         return -EFAULT;
2498         }
2499         return 0;
2500 done:
2501         return err;
2502 }
2503
2504 /*
2505  * check if a multicast source filter allows delivery for a given <src,dst,intf>
2506  */
2507 int ip_mc_sf_allow(struct sock *sk, __be32 loc_addr, __be32 rmt_addr, int dif)
2508 {
2509         struct inet_sock *inet = inet_sk(sk);
2510         struct ip_mc_socklist *pmc;
2511         struct ip_sf_socklist *psl;
2512         int i;
2513         int ret;
2514
2515         ret = 1;
2516         if (!ipv4_is_multicast(loc_addr))
2517                 goto out;
2518
2519         rcu_read_lock();
2520         for_each_pmc_rcu(inet, pmc) {
2521                 if (pmc->multi.imr_multiaddr.s_addr == loc_addr &&
2522                     pmc->multi.imr_ifindex == dif)
2523                         break;
2524         }
2525         ret = inet->mc_all;
2526         if (!pmc)
2527                 goto unlock;
2528         psl = rcu_dereference(pmc->sflist);
2529         ret = (pmc->sfmode == MCAST_EXCLUDE);
2530         if (!psl)
2531                 goto unlock;
2532
2533         for (i = 0; i < psl->sl_count; i++) {
2534                 if (psl->sl_addr[i] == rmt_addr)
2535                         break;
2536         }
2537         ret = 0;
2538         if (pmc->sfmode == MCAST_INCLUDE && i >= psl->sl_count)
2539                 goto unlock;
2540         if (pmc->sfmode == MCAST_EXCLUDE && i < psl->sl_count)
2541                 goto unlock;
2542         ret = 1;
2543 unlock:
2544         rcu_read_unlock();
2545 out:
2546         return ret;
2547 }
2548
2549 /*
2550  *      A socket is closing.
2551  */
2552
2553 void ip_mc_drop_socket(struct sock *sk)
2554 {
2555         struct inet_sock *inet = inet_sk(sk);
2556         struct ip_mc_socklist *iml;
2557         struct net *net = sock_net(sk);
2558
2559         if (!inet->mc_list)
2560                 return;
2561
2562         rtnl_lock();
2563         while ((iml = rtnl_dereference(inet->mc_list)) != NULL) {
2564                 struct in_device *in_dev;
2565
2566                 inet->mc_list = iml->next_rcu;
2567                 in_dev = inetdev_by_index(net, iml->multi.imr_ifindex);
2568                 (void) ip_mc_leave_src(sk, iml, in_dev);
2569                 if (in_dev)
2570                         ip_mc_dec_group(in_dev, iml->multi.imr_multiaddr.s_addr);
2571                 /* decrease mem now to avoid the memleak warning */
2572                 atomic_sub(sizeof(*iml), &sk->sk_omem_alloc);
2573                 kfree_rcu(iml, rcu);
2574         }
2575         rtnl_unlock();
2576 }
2577
2578 /* called with rcu_read_lock() */
2579 int ip_check_mc_rcu(struct in_device *in_dev, __be32 mc_addr, __be32 src_addr, u8 proto)
2580 {
2581         struct ip_mc_list *im;
2582         struct ip_mc_list __rcu **mc_hash;
2583         struct ip_sf_list *psf;
2584         int rv = 0;
2585
2586         mc_hash = rcu_dereference(in_dev->mc_hash);
2587         if (mc_hash) {
2588                 u32 hash = hash_32((__force u32)mc_addr, MC_HASH_SZ_LOG);
2589
2590                 for (im = rcu_dereference(mc_hash[hash]);
2591                      im != NULL;
2592                      im = rcu_dereference(im->next_hash)) {
2593                         if (im->multiaddr == mc_addr)
2594                                 break;
2595                 }
2596         } else {
2597                 for_each_pmc_rcu(in_dev, im) {
2598                         if (im->multiaddr == mc_addr)
2599                                 break;
2600                 }
2601         }
2602         if (im && proto == IPPROTO_IGMP) {
2603                 rv = 1;
2604         } else if (im) {
2605                 if (src_addr) {
2606                         for (psf = im->sources; psf; psf = psf->sf_next) {
2607                                 if (psf->sf_inaddr == src_addr)
2608                                         break;
2609                         }
2610                         if (psf)
2611                                 rv = psf->sf_count[MCAST_INCLUDE] ||
2612                                         psf->sf_count[MCAST_EXCLUDE] !=
2613                                         im->sfcount[MCAST_EXCLUDE];
2614                         else
2615                                 rv = im->sfcount[MCAST_EXCLUDE] != 0;
2616                 } else
2617                         rv = 1; /* unspecified source; tentatively allow */
2618         }
2619         return rv;
2620 }
2621
2622 #if defined(CONFIG_PROC_FS)
2623 struct igmp_mc_iter_state {
2624         struct seq_net_private p;
2625         struct net_device *dev;
2626         struct in_device *in_dev;
2627 };
2628
2629 #define igmp_mc_seq_private(seq)        ((struct igmp_mc_iter_state *)(seq)->private)
2630
2631 static inline struct ip_mc_list *igmp_mc_get_first(struct seq_file *seq)
2632 {
2633         struct net *net = seq_file_net(seq);
2634         struct ip_mc_list *im = NULL;
2635         struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2636
2637         state->in_dev = NULL;
2638         for_each_netdev_rcu(net, state->dev) {
2639                 struct in_device *in_dev;
2640
2641                 in_dev = __in_dev_get_rcu(state->dev);
2642                 if (!in_dev)
2643                         continue;
2644                 im = rcu_dereference(in_dev->mc_list);
2645                 if (im) {
2646                         state->in_dev = in_dev;
2647                         break;
2648                 }
2649         }
2650         return im;
2651 }
2652
2653 static struct ip_mc_list *igmp_mc_get_next(struct seq_file *seq, struct ip_mc_list *im)
2654 {
2655         struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2656
2657         im = rcu_dereference(im->next_rcu);
2658         while (!im) {
2659                 state->dev = next_net_device_rcu(state->dev);
2660                 if (!state->dev) {
2661                         state->in_dev = NULL;
2662                         break;
2663                 }
2664                 state->in_dev = __in_dev_get_rcu(state->dev);
2665                 if (!state->in_dev)
2666                         continue;
2667                 im = rcu_dereference(state->in_dev->mc_list);
2668         }
2669         return im;
2670 }
2671
2672 static struct ip_mc_list *igmp_mc_get_idx(struct seq_file *seq, loff_t pos)
2673 {
2674         struct ip_mc_list *im = igmp_mc_get_first(seq);
2675         if (im)
2676                 while (pos && (im = igmp_mc_get_next(seq, im)) != NULL)
2677                         --pos;
2678         return pos ? NULL : im;
2679 }
2680
2681 static void *igmp_mc_seq_start(struct seq_file *seq, loff_t *pos)
2682         __acquires(rcu)
2683 {
2684         rcu_read_lock();
2685         return *pos ? igmp_mc_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
2686 }
2687
2688 static void *igmp_mc_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2689 {
2690         struct ip_mc_list *im;
2691         if (v == SEQ_START_TOKEN)
2692                 im = igmp_mc_get_first(seq);
2693         else
2694                 im = igmp_mc_get_next(seq, v);
2695         ++*pos;
2696         return im;
2697 }
2698
2699 static void igmp_mc_seq_stop(struct seq_file *seq, void *v)
2700         __releases(rcu)
2701 {
2702         struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2703
2704         state->in_dev = NULL;
2705         state->dev = NULL;
2706         rcu_read_unlock();
2707 }
2708
2709 static int igmp_mc_seq_show(struct seq_file *seq, void *v)
2710 {
2711         if (v == SEQ_START_TOKEN)
2712                 seq_puts(seq,
2713                          "Idx\tDevice    : Count Querier\tGroup    Users Timer\tReporter\n");
2714         else {
2715                 struct ip_mc_list *im = (struct ip_mc_list *)v;
2716                 struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2717                 char   *querier;
2718                 long delta;
2719
2720 #ifdef CONFIG_IP_MULTICAST
2721                 querier = IGMP_V1_SEEN(state->in_dev) ? "V1" :
2722                           IGMP_V2_SEEN(state->in_dev) ? "V2" :
2723                           "V3";
2724 #else
2725                 querier = "NONE";
2726 #endif
2727
2728                 if (rcu_access_pointer(state->in_dev->mc_list) == im) {
2729                         seq_printf(seq, "%d\t%-10s: %5d %7s\n",
2730                                    state->dev->ifindex, state->dev->name, state->in_dev->mc_count, querier);
2731                 }
2732
2733                 delta = im->timer.expires - jiffies;
2734                 seq_printf(seq,
2735                            "\t\t\t\t%08X %5d %d:%08lX\t\t%d\n",
2736                            im->multiaddr, im->users,
2737                            im->tm_running,
2738                            im->tm_running ? jiffies_delta_to_clock_t(delta) : 0,
2739                            im->reporter);
2740         }
2741         return 0;
2742 }
2743
2744 static const struct seq_operations igmp_mc_seq_ops = {
2745         .start  =       igmp_mc_seq_start,
2746         .next   =       igmp_mc_seq_next,
2747         .stop   =       igmp_mc_seq_stop,
2748         .show   =       igmp_mc_seq_show,
2749 };
2750
2751 static int igmp_mc_seq_open(struct inode *inode, struct file *file)
2752 {
2753         return seq_open_net(inode, file, &igmp_mc_seq_ops,
2754                         sizeof(struct igmp_mc_iter_state));
2755 }
2756
2757 static const struct file_operations igmp_mc_seq_fops = {
2758         .owner          =       THIS_MODULE,
2759         .open           =       igmp_mc_seq_open,
2760         .read           =       seq_read,
2761         .llseek         =       seq_lseek,
2762         .release        =       seq_release_net,
2763 };
2764
2765 struct igmp_mcf_iter_state {
2766         struct seq_net_private p;
2767         struct net_device *dev;
2768         struct in_device *idev;
2769         struct ip_mc_list *im;
2770 };
2771
2772 #define igmp_mcf_seq_private(seq)       ((struct igmp_mcf_iter_state *)(seq)->private)
2773
2774 static inline struct ip_sf_list *igmp_mcf_get_first(struct seq_file *seq)
2775 {
2776         struct net *net = seq_file_net(seq);
2777         struct ip_sf_list *psf = NULL;
2778         struct ip_mc_list *im = NULL;
2779         struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2780
2781         state->idev = NULL;
2782         state->im = NULL;
2783         for_each_netdev_rcu(net, state->dev) {
2784                 struct in_device *idev;
2785                 idev = __in_dev_get_rcu(state->dev);
2786                 if (unlikely(!idev))
2787                         continue;
2788                 im = rcu_dereference(idev->mc_list);
2789                 if (likely(im)) {
2790                         spin_lock_bh(&im->lock);
2791                         psf = im->sources;
2792                         if (likely(psf)) {
2793                                 state->im = im;
2794                                 state->idev = idev;
2795                                 break;
2796                         }
2797                         spin_unlock_bh(&im->lock);
2798                 }
2799         }
2800         return psf;
2801 }
2802
2803 static struct ip_sf_list *igmp_mcf_get_next(struct seq_file *seq, struct ip_sf_list *psf)
2804 {
2805         struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2806
2807         psf = psf->sf_next;
2808         while (!psf) {
2809                 spin_unlock_bh(&state->im->lock);
2810                 state->im = state->im->next;
2811                 while (!state->im) {
2812                         state->dev = next_net_device_rcu(state->dev);
2813                         if (!state->dev) {
2814                                 state->idev = NULL;
2815                                 goto out;
2816                         }
2817                         state->idev = __in_dev_get_rcu(state->dev);
2818                         if (!state->idev)
2819                                 continue;
2820                         state->im = rcu_dereference(state->idev->mc_list);
2821                 }
2822                 if (!state->im)
2823                         break;
2824                 spin_lock_bh(&state->im->lock);
2825                 psf = state->im->sources;
2826         }
2827 out:
2828         return psf;
2829 }
2830
2831 static struct ip_sf_list *igmp_mcf_get_idx(struct seq_file *seq, loff_t pos)
2832 {
2833         struct ip_sf_list *psf = igmp_mcf_get_first(seq);
2834         if (psf)
2835                 while (pos && (psf = igmp_mcf_get_next(seq, psf)) != NULL)
2836                         --pos;
2837         return pos ? NULL : psf;
2838 }
2839
2840 static void *igmp_mcf_seq_start(struct seq_file *seq, loff_t *pos)
2841         __acquires(rcu)
2842 {
2843         rcu_read_lock();
2844         return *pos ? igmp_mcf_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
2845 }
2846
2847 static void *igmp_mcf_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2848 {
2849         struct ip_sf_list *psf;
2850         if (v == SEQ_START_TOKEN)
2851                 psf = igmp_mcf_get_first(seq);
2852         else
2853                 psf = igmp_mcf_get_next(seq, v);
2854         ++*pos;
2855         return psf;
2856 }
2857
2858 static void igmp_mcf_seq_stop(struct seq_file *seq, void *v)
2859         __releases(rcu)
2860 {
2861         struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2862         if (likely(state->im)) {
2863                 spin_unlock_bh(&state->im->lock);
2864                 state->im = NULL;
2865         }
2866         state->idev = NULL;
2867         state->dev = NULL;
2868         rcu_read_unlock();
2869 }
2870
2871 static int igmp_mcf_seq_show(struct seq_file *seq, void *v)
2872 {
2873         struct ip_sf_list *psf = (struct ip_sf_list *)v;
2874         struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2875
2876         if (v == SEQ_START_TOKEN) {
2877                 seq_puts(seq, "Idx Device        MCA        SRC    INC    EXC\n");
2878         } else {
2879                 seq_printf(seq,
2880                            "%3d %6.6s 0x%08x "
2881                            "0x%08x %6lu %6lu\n",
2882                            state->dev->ifindex, state->dev->name,
2883                            ntohl(state->im->multiaddr),
2884                            ntohl(psf->sf_inaddr),
2885                            psf->sf_count[MCAST_INCLUDE],
2886                            psf->sf_count[MCAST_EXCLUDE]);
2887         }
2888         return 0;
2889 }
2890
2891 static const struct seq_operations igmp_mcf_seq_ops = {
2892         .start  =       igmp_mcf_seq_start,
2893         .next   =       igmp_mcf_seq_next,
2894         .stop   =       igmp_mcf_seq_stop,
2895         .show   =       igmp_mcf_seq_show,
2896 };
2897
2898 static int igmp_mcf_seq_open(struct inode *inode, struct file *file)
2899 {
2900         return seq_open_net(inode, file, &igmp_mcf_seq_ops,
2901                         sizeof(struct igmp_mcf_iter_state));
2902 }
2903
2904 static const struct file_operations igmp_mcf_seq_fops = {
2905         .owner          =       THIS_MODULE,
2906         .open           =       igmp_mcf_seq_open,
2907         .read           =       seq_read,
2908         .llseek         =       seq_lseek,
2909         .release        =       seq_release_net,
2910 };
2911
2912 static int __net_init igmp_net_init(struct net *net)
2913 {
2914         struct proc_dir_entry *pde;
2915         int err;
2916
2917         pde = proc_create("igmp", S_IRUGO, net->proc_net, &igmp_mc_seq_fops);
2918         if (!pde)
2919                 goto out_igmp;
2920         pde = proc_create("mcfilter", S_IRUGO, net->proc_net,
2921                           &igmp_mcf_seq_fops);
2922         if (!pde)
2923                 goto out_mcfilter;
2924         err = inet_ctl_sock_create(&net->ipv4.mc_autojoin_sk, AF_INET,
2925                                    SOCK_DGRAM, 0, net);
2926         if (err < 0) {
2927                 pr_err("Failed to initialize the IGMP autojoin socket (err %d)\n",
2928                        err);
2929                 goto out_sock;
2930         }
2931
2932         return 0;
2933
2934 out_sock:
2935         remove_proc_entry("mcfilter", net->proc_net);
2936 out_mcfilter:
2937         remove_proc_entry("igmp", net->proc_net);
2938 out_igmp:
2939         return -ENOMEM;
2940 }
2941
2942 static void __net_exit igmp_net_exit(struct net *net)
2943 {
2944         remove_proc_entry("mcfilter", net->proc_net);
2945         remove_proc_entry("igmp", net->proc_net);
2946         inet_ctl_sock_destroy(net->ipv4.mc_autojoin_sk);
2947 }
2948
2949 static struct pernet_operations igmp_net_ops = {
2950         .init = igmp_net_init,
2951         .exit = igmp_net_exit,
2952 };
2953 #endif
2954
2955 static int igmp_netdev_event(struct notifier_block *this,
2956                              unsigned long event, void *ptr)
2957 {
2958         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2959         struct in_device *in_dev;
2960
2961         switch (event) {
2962         case NETDEV_RESEND_IGMP:
2963                 in_dev = __in_dev_get_rtnl(dev);
2964                 if (in_dev)
2965                         ip_mc_rejoin_groups(in_dev);
2966                 break;
2967         default:
2968                 break;
2969         }
2970         return NOTIFY_DONE;
2971 }
2972
2973 static struct notifier_block igmp_notifier = {
2974         .notifier_call = igmp_netdev_event,
2975 };
2976
2977 int __init igmp_mc_init(void)
2978 {
2979 #if defined(CONFIG_PROC_FS)
2980         int err;
2981
2982         err = register_pernet_subsys(&igmp_net_ops);
2983         if (err)
2984                 return err;
2985         err = register_netdevice_notifier(&igmp_notifier);
2986         if (err)
2987                 goto reg_notif_fail;
2988         return 0;
2989
2990 reg_notif_fail:
2991         unregister_pernet_subsys(&igmp_net_ops);
2992         return err;
2993 #else
2994         return register_netdevice_notifier(&igmp_notifier);
2995 #endif
2996 }