Pileus Git - ~andy/linux/blob - 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
  92 #include <net/net_namespace.h>
  93 #include <net/arp.h>
  94 #include <net/ip.h>
  95 #include <net/protocol.h>
  96 #include <net/route.h>
  97 #include <net/sock.h>
  98 #include <net/checksum.h>
  99 #include <linux/netfilter_ipv4.h>
 100 #ifdef CONFIG_IP_MROUTE
 101 #include <linux/mroute.h>
 102 #endif
 103 #ifdef CONFIG_PROC_FS
 104 #include <linux/proc_fs.h>
 105 #include <linux/seq_file.h>
 106 #endif
 107
 108 #define IP_MAX_MEMBERSHIPS      20
 109 #define IP_MAX_MSF              10
 110
 111 #ifdef CONFIG_IP_MULTICAST
 112 /* Parameter names and values are taken from igmp-v2-06 draft */
 113
 114 #define IGMP_V1_Router_Present_Timeout          (400*HZ)
 115 #define IGMP_V2_Router_Present_Timeout          (400*HZ)
 116 #define IGMP_Unsolicited_Report_Interval        (10*HZ)
 117 #define IGMP_Query_Response_Interval            (10*HZ)
 118 #define IGMP_Unsolicited_Report_Count           2
 119
 120
 121 #define IGMP_Initial_Report_Delay               (1)
 122
 123 /* IGMP_Initial_Report_Delay is not from IGMP specs!
 124  * IGMP specs require to report membership immediately after
 125  * joining a group, but we delay the first report by a
 126  * small interval. It seems more natural and still does not
 127  * contradict to specs provided this delay is small enough.
 128  */
 129
 130 #define IGMP_V1_SEEN(in_dev) \
 131         (IPV4_DEVCONF_ALL(dev_net(in_dev->dev), FORCE_IGMP_VERSION) == 1 || \
 132          IN_DEV_CONF_GET((in_dev), FORCE_IGMP_VERSION) == 1 || \
 133          ((in_dev)->mr_v1_seen && \
 134           time_before(jiffies, (in_dev)->mr_v1_seen)))
 135 #define IGMP_V2_SEEN(in_dev) \
 136         (IPV4_DEVCONF_ALL(dev_net(in_dev->dev), FORCE_IGMP_VERSION) == 2 || \
 137          IN_DEV_CONF_GET((in_dev), FORCE_IGMP_VERSION) == 2 || \
 138          ((in_dev)->mr_v2_seen && \
 139           time_before(jiffies, (in_dev)->mr_v2_seen)))
 140
 141 static void igmpv3_add_delrec(struct in_device *in_dev, struct ip_mc_list *im);
 142 static void igmpv3_del_delrec(struct in_device *in_dev, __be32 multiaddr);
 143 static void igmpv3_clear_delrec(struct in_device *in_dev);
 144 static int sf_setstate(struct ip_mc_list *pmc);
 145 static void sf_markstate(struct ip_mc_list *pmc);
 146 #endif
 147 static void ip_mc_clear_src(struct ip_mc_list *pmc);
 148 static int ip_mc_add_src(struct in_device *in_dev, __be32 *pmca, int sfmode,
 149                          int sfcount, __be32 *psfsrc, int delta);
 150
 151 static void ip_ma_put(struct ip_mc_list *im)
 152 {
 153         if (atomic_dec_and_test(&im->refcnt)) {
 154                 in_dev_put(im->interface);
 155                 kfree_rcu(im, rcu);
 156         }
 157 }
 158
 159 #define for_each_pmc_rcu(in_dev, pmc)                           \
 160         for (pmc = rcu_dereference(in_dev->mc_list);            \
 161              pmc != NULL;                                       \
 162              pmc = rcu_dereference(pmc->next_rcu))
 163
 164 #define for_each_pmc_rtnl(in_dev, pmc)                          \
 165         for (pmc = rtnl_dereference(in_dev->mc_list);           \
 166              pmc != NULL;                                       \
 167              pmc = rtnl_dereference(pmc->next_rcu))
 168
 169 #ifdef CONFIG_IP_MULTICAST
 170
 171 /*
 172  *      Timer management
 173  */
 174
 175 static void igmp_stop_timer(struct ip_mc_list *im)
 176 {
 177         spin_lock_bh(&im->lock);
 178         if (del_timer(&im->timer))
 179                 atomic_dec(&im->refcnt);
 180         im->tm_running = 0;
 181         im->reporter = 0;
 182         im->unsolicit_count = 0;
 183         spin_unlock_bh(&im->lock);
 184 }
 185
 186 /* It must be called with locked im->lock */
 187 static void igmp_start_timer(struct ip_mc_list *im, int max_delay)
 188 {
 189         int tv = net_random() % max_delay;
 190
 191         im->tm_running = 1;
 192         if (!mod_timer(&im->timer, jiffies+tv+2))
 193                 atomic_inc(&im->refcnt);
 194 }
 195
 196 static void igmp_gq_start_timer(struct in_device *in_dev)
 197 {
 198         int tv = net_random() % in_dev->mr_maxdelay;
 199
 200         in_dev->mr_gq_running = 1;
 201         if (!mod_timer(&in_dev->mr_gq_timer, jiffies+tv+2))
 202                 in_dev_hold(in_dev);
 203 }
 204
 205 static void igmp_ifc_start_timer(struct in_device *in_dev, int delay)
 206 {
 207         int tv = net_random() % delay;
 208
 209         if (!mod_timer(&in_dev->mr_ifc_timer, jiffies+tv+2))
 210                 in_dev_hold(in_dev);
 211 }
 212
 213 static void igmp_mod_timer(struct ip_mc_list *im, int max_delay)
 214 {
 215         spin_lock_bh(&im->lock);
 216         im->unsolicit_count = 0;
 217         if (del_timer(&im->timer)) {
 218                 if ((long)(im->timer.expires-jiffies) < max_delay) {
 219                         add_timer(&im->timer);
 220                         im->tm_running = 1;
 221                         spin_unlock_bh(&im->lock);
 222                         return;
 223                 }
 224                 atomic_dec(&im->refcnt);
 225         }
 226         igmp_start_timer(im, max_delay);
 227         spin_unlock_bh(&im->lock);
 228 }
 229
 230
 231 /*
 232  *      Send an IGMP report.
 233  */
 234
 235 #define IGMP_SIZE (sizeof(struct igmphdr)+sizeof(struct iphdr)+4)
 236
 237
 238 static int is_in(struct ip_mc_list *pmc, struct ip_sf_list *psf, int type,
 239         int gdeleted, int sdeleted)
 240 {
 241         switch (type) {
 242         case IGMPV3_MODE_IS_INCLUDE:
 243         case IGMPV3_MODE_IS_EXCLUDE:
 244                 if (gdeleted || sdeleted)
 245                         return 0;
 246                 if (!(pmc->gsquery && !psf->sf_gsresp)) {
 247                         if (pmc->sfmode == MCAST_INCLUDE)
 248                                 return 1;
 249                         /* don't include if this source is excluded
 250                          * in all filters
 251                          */
 252                         if (psf->sf_count[MCAST_INCLUDE])
 253                                 return type == IGMPV3_MODE_IS_INCLUDE;
 254                         return pmc->sfcount[MCAST_EXCLUDE] ==
 255                                 psf->sf_count[MCAST_EXCLUDE];
 256                 }
 257                 return 0;
 258         case IGMPV3_CHANGE_TO_INCLUDE:
 259                 if (gdeleted || sdeleted)
 260                         return 0;
 261                 return psf->sf_count[MCAST_INCLUDE] != 0;
 262         case IGMPV3_CHANGE_TO_EXCLUDE:
 263                 if (gdeleted || sdeleted)
 264                         return 0;
 265                 if (pmc->sfcount[MCAST_EXCLUDE] == 0 ||
 266                     psf->sf_count[MCAST_INCLUDE])
 267                         return 0;
 268                 return pmc->sfcount[MCAST_EXCLUDE] ==
 269                         psf->sf_count[MCAST_EXCLUDE];
 270         case IGMPV3_ALLOW_NEW_SOURCES:
 271                 if (gdeleted || !psf->sf_crcount)
 272                         return 0;
 273                 return (pmc->sfmode == MCAST_INCLUDE) ^ sdeleted;
 274         case IGMPV3_BLOCK_OLD_SOURCES:
 275                 if (pmc->sfmode == MCAST_INCLUDE)
 276                         return gdeleted || (psf->sf_crcount && sdeleted);
 277                 return psf->sf_crcount && !gdeleted && !sdeleted;
 278         }
 279         return 0;
 280 }
 281
 282 static int
 283 igmp_scount(struct ip_mc_list *pmc, int type, int gdeleted, int sdeleted)
 284 {
 285         struct ip_sf_list *psf;
 286         int scount = 0;
 287
 288         for (psf=pmc->sources; psf; psf=psf->sf_next) {
 289                 if (!is_in(pmc, psf, type, gdeleted, sdeleted))
 290                         continue;
 291                 scount++;
 292         }
 293         return scount;
 294 }
 295
 296 #define igmp_skb_size(skb) (*(unsigned int *)((skb)->cb))
 297
 298 static struct sk_buff *igmpv3_newpack(struct net_device *dev, int size)
 299 {
 300         struct sk_buff *skb;
 301         struct rtable *rt;
 302         struct iphdr *pip;
 303         struct igmpv3_report *pig;
 304         struct net *net = dev_net(dev);
 305         struct flowi4 fl4;
 306         int hlen = LL_RESERVED_SPACE(dev);
 307         int tlen = dev->needed_tailroom;
 308
 309         while (1) {
 310                 skb = alloc_skb(size + hlen + tlen,
 311                                 GFP_ATOMIC | __GFP_NOWARN);
 312                 if (skb)
 313                         break;
 314                 size >>= 1;
 315                 if (size < 256)
 316                         return NULL;
 317         }
 318         igmp_skb_size(skb) = size;
 319
 320         rt = ip_route_output_ports(net, &fl4, NULL, IGMPV3_ALL_MCR, 0,
 321                                    0, 0,
 322                                    IPPROTO_IGMP, 0, dev->ifindex);
 323         if (IS_ERR(rt)) {
 324                 kfree_skb(skb);
 325                 return NULL;
 326         }
 327
 328         skb_dst_set(skb, &rt->dst);
 329         skb->dev = dev;
 330
 331         skb_reserve(skb, hlen);
 332
 333         skb_reset_network_header(skb);
 334         pip = ip_hdr(skb);
 335         skb_put(skb, sizeof(struct iphdr) + 4);
 336
 337         pip->version  = 4;
 338         pip->ihl      = (sizeof(struct iphdr)+4)>>2;
 339         pip->tos      = 0xc0;
 340         pip->frag_off = htons(IP_DF);
 341         pip->ttl      = 1;
 342         pip->daddr    = fl4.daddr;
 343         pip->saddr    = fl4.saddr;
 344         pip->protocol = IPPROTO_IGMP;
 345         pip->tot_len  = 0;      /* filled in later */
 346         ip_select_ident(pip, &rt->dst, NULL);
 347         ((u8 *)&pip[1])[0] = IPOPT_RA;
 348         ((u8 *)&pip[1])[1] = 4;
 349         ((u8 *)&pip[1])[2] = 0;
 350         ((u8 *)&pip[1])[3] = 0;
 351
 352         skb->transport_header = skb->network_header + sizeof(struct iphdr) + 4;
 353         skb_put(skb, sizeof(*pig));
 354         pig = igmpv3_report_hdr(skb);
 355         pig->type = IGMPV3_HOST_MEMBERSHIP_REPORT;
 356         pig->resv1 = 0;
 357         pig->csum = 0;
 358         pig->resv2 = 0;
 359         pig->ngrec = 0;
 360         return skb;
 361 }
 362
 363 static int igmpv3_sendpack(struct sk_buff *skb)
 364 {
 365         struct igmphdr *pig = igmp_hdr(skb);
 366         const int igmplen = skb_tail_pointer(skb) - skb_transport_header(skb);
 367
 368         pig->csum = ip_compute_csum(igmp_hdr(skb), igmplen);
 369
 370         return ip_local_out(skb);
 371 }
 372
 373 static int grec_size(struct ip_mc_list *pmc, int type, int gdel, int sdel)
 374 {
 375         return sizeof(struct igmpv3_grec) + 4*igmp_scount(pmc, type, gdel, sdel);
 376 }
 377
 378 static struct sk_buff *add_grhead(struct sk_buff *skb, struct ip_mc_list *pmc,
 379         int type, struct igmpv3_grec **ppgr)
 380 {
 381         struct net_device *dev = pmc->interface->dev;
 382         struct igmpv3_report *pih;
 383         struct igmpv3_grec *pgr;
 384
 385         if (!skb)
 386                 skb = igmpv3_newpack(dev, dev->mtu);
 387         if (!skb)
 388                 return NULL;
 389         pgr = (struct igmpv3_grec *)skb_put(skb, sizeof(struct igmpv3_grec));
 390         pgr->grec_type = type;
 391         pgr->grec_auxwords = 0;
 392         pgr->grec_nsrcs = 0;
 393         pgr->grec_mca = pmc->multiaddr;
 394         pih = igmpv3_report_hdr(skb);
 395         pih->ngrec = htons(ntohs(pih->ngrec)+1);
 396         *ppgr = pgr;
 397         return skb;
 398 }
 399
 400 #define AVAILABLE(skb) ((skb) ? ((skb)->dev ? igmp_skb_size(skb) - (skb)->len : \
 401         skb_tailroom(skb)) : 0)
 402
 403 static struct sk_buff *add_grec(struct sk_buff *skb, struct ip_mc_list *pmc,
 404         int type, int gdeleted, int sdeleted)
 405 {
 406         struct net_device *dev = pmc->interface->dev;
 407         struct igmpv3_report *pih;
 408         struct igmpv3_grec *pgr = NULL;
 409         struct ip_sf_list *psf, *psf_next, *psf_prev, **psf_list;
 410         int scount, stotal, first, isquery, truncate;
 411
 412         if (pmc->multiaddr == IGMP_ALL_HOSTS)
 413                 return skb;
 414
 415         isquery = type == IGMPV3_MODE_IS_INCLUDE ||
 416                   type == IGMPV3_MODE_IS_EXCLUDE;
 417         truncate = type == IGMPV3_MODE_IS_EXCLUDE ||
 418                     type == IGMPV3_CHANGE_TO_EXCLUDE;
 419
 420         stotal = scount = 0;
 421
 422         psf_list = sdeleted ? &pmc->tomb : &pmc->sources;
 423
 424         if (!*psf_list)
 425                 goto empty_source;
 426
 427         pih = skb ? igmpv3_report_hdr(skb) : NULL;
 428
 429         /* EX and TO_EX get a fresh packet, if needed */
 430         if (truncate) {
 431                 if (pih && pih->ngrec &&
 432                     AVAILABLE(skb) < grec_size(pmc, type, gdeleted, sdeleted)) {
 433                         if (skb)
 434                                 igmpv3_sendpack(skb);
 435                         skb = igmpv3_newpack(dev, dev->mtu);
 436                 }
 437         }
 438         first = 1;
 439         psf_prev = NULL;
 440         for (psf=*psf_list; psf; psf=psf_next) {
 441                 __be32 *psrc;
 442
 443                 psf_next = psf->sf_next;
 444
 445                 if (!is_in(pmc, psf, type, gdeleted, sdeleted)) {
 446                         psf_prev = psf;
 447                         continue;
 448                 }
 449
 450                 /* clear marks on query responses */
 451                 if (isquery)
 452                         psf->sf_gsresp = 0;
 453
 454                 if (AVAILABLE(skb) < sizeof(__be32) +
 455                     first*sizeof(struct igmpv3_grec)) {
 456                         if (truncate && !first)
 457                                 break;   /* truncate these */
 458                         if (pgr)
 459                                 pgr->grec_nsrcs = htons(scount);
 460                         if (skb)
 461                                 igmpv3_sendpack(skb);
 462                         skb = igmpv3_newpack(dev, dev->mtu);
 463                         first = 1;
 464                         scount = 0;
 465                 }
 466                 if (first) {
 467                         skb = add_grhead(skb, pmc, type, &pgr);
 468                         first = 0;
 469                 }
 470                 if (!skb)
 471                         return NULL;
 472                 psrc = (__be32 *)skb_put(skb, sizeof(__be32));
 473                 *psrc = psf->sf_inaddr;
 474                 scount++; stotal++;
 475                 if ((type == IGMPV3_ALLOW_NEW_SOURCES ||
 476                      type == IGMPV3_BLOCK_OLD_SOURCES) && psf->sf_crcount) {
 477                         psf->sf_crcount--;
 478                         if ((sdeleted || gdeleted) && psf->sf_crcount == 0) {
 479                                 if (psf_prev)
 480                                         psf_prev->sf_next = psf->sf_next;
 481                                 else
 482                                         *psf_list = psf->sf_next;
 483                                 kfree(psf);
 484                                 continue;
 485                         }
 486                 }
 487                 psf_prev = psf;
 488         }
 489
 490 empty_source:
 491         if (!stotal) {
 492                 if (type == IGMPV3_ALLOW_NEW_SOURCES ||
 493                     type == IGMPV3_BLOCK_OLD_SOURCES)
 494                         return skb;
 495                 if (pmc->crcount || isquery) {
 496                         /* make sure we have room for group header */
 497                         if (skb && AVAILABLE(skb)<sizeof(struct igmpv3_grec)) {
 498                                 igmpv3_sendpack(skb);
 499                                 skb = NULL; /* add_grhead will get a new one */
 500                         }
 501                         skb = add_grhead(skb, pmc, type, &pgr);
 502                 }
 503         }
 504         if (pgr)
 505                 pgr->grec_nsrcs = htons(scount);
 506
 507         if (isquery)
 508                 pmc->gsquery = 0;       /* clear query state on report */
 509         return skb;
 510 }
 511
 512 static int igmpv3_send_report(struct in_device *in_dev, struct ip_mc_list *pmc)
 513 {
 514         struct sk_buff *skb = NULL;
 515         int type;
 516
 517         if (!pmc) {
 518                 rcu_read_lock();
 519                 for_each_pmc_rcu(in_dev, pmc) {
 520                         if (pmc->multiaddr == IGMP_ALL_HOSTS)
 521                                 continue;
 522                         spin_lock_bh(&pmc->lock);
 523                         if (pmc->sfcount[MCAST_EXCLUDE])
 524                                 type = IGMPV3_MODE_IS_EXCLUDE;
 525                         else
 526                                 type = IGMPV3_MODE_IS_INCLUDE;
 527                         skb = add_grec(skb, pmc, type, 0, 0);
 528                         spin_unlock_bh(&pmc->lock);
 529                 }
 530                 rcu_read_unlock();
 531         } else {
 532                 spin_lock_bh(&pmc->lock);
 533                 if (pmc->sfcount[MCAST_EXCLUDE])
 534                         type = IGMPV3_MODE_IS_EXCLUDE;
 535                 else
 536                         type = IGMPV3_MODE_IS_INCLUDE;
 537                 skb = add_grec(skb, pmc, type, 0, 0);
 538                 spin_unlock_bh(&pmc->lock);
 539         }
 540         if (!skb)
 541                 return 0;
 542         return igmpv3_sendpack(skb);
 543 }
 544
 545 /*
 546  * remove zero-count source records from a source filter list
 547  */
 548 static void igmpv3_clear_zeros(struct ip_sf_list **ppsf)
 549 {
 550         struct ip_sf_list *psf_prev, *psf_next, *psf;
 551
 552         psf_prev = NULL;
 553         for (psf=*ppsf; psf; psf = psf_next) {
 554                 psf_next = psf->sf_next;
 555                 if (psf->sf_crcount == 0) {
 556                         if (psf_prev)
 557                                 psf_prev->sf_next = psf->sf_next;
 558                         else
 559                                 *ppsf = psf->sf_next;
 560                         kfree(psf);
 561                 } else
 562                         psf_prev = psf;
 563         }
 564 }
 565
 566 static void igmpv3_send_cr(struct in_device *in_dev)
 567 {
 568         struct ip_mc_list *pmc, *pmc_prev, *pmc_next;
 569         struct sk_buff *skb = NULL;
 570         int type, dtype;
 571
 572         rcu_read_lock();
 573         spin_lock_bh(&in_dev->mc_tomb_lock);
 574
 575         /* deleted MCA's */
 576         pmc_prev = NULL;
 577         for (pmc=in_dev->mc_tomb; pmc; pmc=pmc_next) {
 578                 pmc_next = pmc->next;
 579                 if (pmc->sfmode == MCAST_INCLUDE) {
 580                         type = IGMPV3_BLOCK_OLD_SOURCES;
 581                         dtype = IGMPV3_BLOCK_OLD_SOURCES;
 582                         skb = add_grec(skb, pmc, type, 1, 0);
 583                         skb = add_grec(skb, pmc, dtype, 1, 1);
 584                 }
 585                 if (pmc->crcount) {
 586                         if (pmc->sfmode == MCAST_EXCLUDE) {
 587                                 type = IGMPV3_CHANGE_TO_INCLUDE;
 588                                 skb = add_grec(skb, pmc, type, 1, 0);
 589                         }
 590                         pmc->crcount--;
 591                         if (pmc->crcount == 0) {
 592                                 igmpv3_clear_zeros(&pmc->tomb);
 593                                 igmpv3_clear_zeros(&pmc->sources);
 594                         }
 595                 }
 596                 if (pmc->crcount == 0 && !pmc->tomb && !pmc->sources) {
 597                         if (pmc_prev)
 598                                 pmc_prev->next = pmc_next;
 599                         else
 600                                 in_dev->mc_tomb = pmc_next;
 601                         in_dev_put(pmc->interface);
 602                         kfree(pmc);
 603                 } else
 604                         pmc_prev = pmc;
 605         }
 606         spin_unlock_bh(&in_dev->mc_tomb_lock);
 607
 608         /* change recs */
 609         for_each_pmc_rcu(in_dev, pmc) {
 610                 spin_lock_bh(&pmc->lock);
 611                 if (pmc->sfcount[MCAST_EXCLUDE]) {
 612                         type = IGMPV3_BLOCK_OLD_SOURCES;
 613                         dtype = IGMPV3_ALLOW_NEW_SOURCES;
 614                 } else {
 615                         type = IGMPV3_ALLOW_NEW_SOURCES;
 616                         dtype = IGMPV3_BLOCK_OLD_SOURCES;
 617                 }
 618                 skb = add_grec(skb, pmc, type, 0, 0);
 619                 skb = add_grec(skb, pmc, dtype, 0, 1);  /* deleted sources */
 620
 621                 /* filter mode changes */
 622                 if (pmc->crcount) {
 623                         if (pmc->sfmode == MCAST_EXCLUDE)
 624                                 type = IGMPV3_CHANGE_TO_EXCLUDE;
 625                         else
 626                                 type = IGMPV3_CHANGE_TO_INCLUDE;
 627                         skb = add_grec(skb, pmc, type, 0, 0);
 628                         pmc->crcount--;
 629                 }
 630                 spin_unlock_bh(&pmc->lock);
 631         }
 632         rcu_read_unlock();
 633
 634         if (!skb)
 635                 return;
 636         (void) igmpv3_sendpack(skb);
 637 }
 638
 639 static int igmp_send_report(struct in_device *in_dev, struct ip_mc_list *pmc,
 640         int type)
 641 {
 642         struct sk_buff *skb;
 643         struct iphdr *iph;
 644         struct igmphdr *ih;
 645         struct rtable *rt;
 646         struct net_device *dev = in_dev->dev;
 647         struct net *net = dev_net(dev);
 648         __be32  group = pmc ? pmc->multiaddr : 0;
 649         struct flowi4 fl4;
 650         __be32  dst;
 651         int hlen, tlen;
 652
 653         if (type == IGMPV3_HOST_MEMBERSHIP_REPORT)
 654                 return igmpv3_send_report(in_dev, pmc);
 655         else if (type == IGMP_HOST_LEAVE_MESSAGE)
 656                 dst = IGMP_ALL_ROUTER;
 657         else
 658                 dst = group;
 659
 660         rt = ip_route_output_ports(net, &fl4, NULL, dst, 0,
 661                                    0, 0,
 662                                    IPPROTO_IGMP, 0, dev->ifindex);
 663         if (IS_ERR(rt))
 664                 return -1;
 665
 666         hlen = LL_RESERVED_SPACE(dev);
 667         tlen = dev->needed_tailroom;
 668         skb = alloc_skb(IGMP_SIZE + hlen + tlen, GFP_ATOMIC);
 669         if (skb == NULL) {
 670                 ip_rt_put(rt);
 671                 return -1;
 672         }
 673
 674         skb_dst_set(skb, &rt->dst);
 675
 676         skb_reserve(skb, hlen);
 677
 678         skb_reset_network_header(skb);
 679         iph = ip_hdr(skb);
 680         skb_put(skb, sizeof(struct iphdr) + 4);
 681
 682         iph->version  = 4;
 683         iph->ihl      = (sizeof(struct iphdr)+4)>>2;
 684         iph->tos      = 0xc0;
 685         iph->frag_off = htons(IP_DF);
 686         iph->ttl      = 1;
 687         iph->daddr    = dst;
 688         iph->saddr    = fl4.saddr;
 689         iph->protocol = IPPROTO_IGMP;
 690         ip_select_ident(iph, &rt->dst, NULL);
 691         ((u8 *)&iph[1])[0] = IPOPT_RA;
 692         ((u8 *)&iph[1])[1] = 4;
 693         ((u8 *)&iph[1])[2] = 0;
 694         ((u8 *)&iph[1])[3] = 0;
 695
 696         ih = (struct igmphdr *)skb_put(skb, sizeof(struct igmphdr));
 697         ih->type = type;
 698         ih->code = 0;
 699         ih->csum = 0;
 700         ih->group = group;
 701         ih->csum = ip_compute_csum((void *)ih, sizeof(struct igmphdr));
 702
 703         return ip_local_out(skb);
 704 }
 705
 706 static void igmp_gq_timer_expire(unsigned long data)
 707 {
 708         struct in_device *in_dev = (struct in_device *)data;
 709
 710         in_dev->mr_gq_running = 0;
 711         igmpv3_send_report(in_dev, NULL);
 712         __in_dev_put(in_dev);
 713 }
 714
 715 static void igmp_ifc_timer_expire(unsigned long data)
 716 {
 717         struct in_device *in_dev = (struct in_device *)data;
 718
 719         igmpv3_send_cr(in_dev);
 720         if (in_dev->mr_ifc_count) {
 721                 in_dev->mr_ifc_count--;
 722                 igmp_ifc_start_timer(in_dev, IGMP_Unsolicited_Report_Interval);
 723         }
 724         __in_dev_put(in_dev);
 725 }
 726
 727 static void igmp_ifc_event(struct in_device *in_dev)
 728 {
 729         if (IGMP_V1_SEEN(in_dev) || IGMP_V2_SEEN(in_dev))
 730                 return;
 731         in_dev->mr_ifc_count = in_dev->mr_qrv ? in_dev->mr_qrv :
 732                 IGMP_Unsolicited_Report_Count;
 733         igmp_ifc_start_timer(in_dev, 1);
 734 }
 735
 736
 737 static void igmp_timer_expire(unsigned long data)
 738 {
 739         struct ip_mc_list *im=(struct ip_mc_list *)data;
 740         struct in_device *in_dev = im->interface;
 741
 742         spin_lock(&im->lock);
 743         im->tm_running = 0;
 744
 745         if (im->unsolicit_count) {
 746                 im->unsolicit_count--;
 747                 igmp_start_timer(im, IGMP_Unsolicited_Report_Interval);
 748         }
 749         im->reporter = 1;
 750         spin_unlock(&im->lock);
 751
 752         if (IGMP_V1_SEEN(in_dev))
 753                 igmp_send_report(in_dev, im, IGMP_HOST_MEMBERSHIP_REPORT);
 754         else if (IGMP_V2_SEEN(in_dev))
 755                 igmp_send_report(in_dev, im, IGMPV2_HOST_MEMBERSHIP_REPORT);
 756         else
 757                 igmp_send_report(in_dev, im, IGMPV3_HOST_MEMBERSHIP_REPORT);
 758
 759         ip_ma_put(im);
 760 }
 761
 762 /* mark EXCLUDE-mode sources */
 763 static int igmp_xmarksources(struct ip_mc_list *pmc, int nsrcs, __be32 *srcs)
 764 {
 765         struct ip_sf_list *psf;
 766         int i, scount;
 767
 768         scount = 0;
 769         for (psf=pmc->sources; psf; psf=psf->sf_next) {
 770                 if (scount == nsrcs)
 771                         break;
 772                 for (i=0; i<nsrcs; i++) {
 773                         /* skip inactive filters */
 774                         if (psf->sf_count[MCAST_INCLUDE] ||
 775                             pmc->sfcount[MCAST_EXCLUDE] !=
 776                             psf->sf_count[MCAST_EXCLUDE])
 777                                 break;
 778                         if (srcs[i] == psf->sf_inaddr) {
 779                                 scount++;
 780                                 break;
 781                         }
 782                 }
 783         }
 784         pmc->gsquery = 0;
 785         if (scount == nsrcs)    /* all sources excluded */
 786                 return 0;
 787         return 1;
 788 }
 789
 790 static int igmp_marksources(struct ip_mc_list *pmc, int nsrcs, __be32 *srcs)
 791 {
 792         struct ip_sf_list *psf;
 793         int i, scount;
 794
 795         if (pmc->sfmode == MCAST_EXCLUDE)
 796                 return igmp_xmarksources(pmc, nsrcs, srcs);
 797
 798         /* mark INCLUDE-mode sources */
 799         scount = 0;
 800         for (psf=pmc->sources; psf; psf=psf->sf_next) {
 801                 if (scount == nsrcs)
 802                         break;
 803                 for (i=0; i<nsrcs; i++)
 804                         if (srcs[i] == psf->sf_inaddr) {
 805                                 psf->sf_gsresp = 1;
 806                                 scount++;
 807                                 break;
 808                         }
 809         }
 810         if (!scount) {
 811                 pmc->gsquery = 0;
 812                 return 0;
 813         }
 814         pmc->gsquery = 1;
 815         return 1;
 816 }
 817
 818 /* return true if packet was dropped */
 819 static bool igmp_heard_report(struct in_device *in_dev, __be32 group)
 820 {
 821         struct ip_mc_list *im;
 822
 823         /* Timers are only set for non-local groups */
 824
 825         if (group == IGMP_ALL_HOSTS)
 826                 return false;
 827
 828         rcu_read_lock();
 829         for_each_pmc_rcu(in_dev, im) {
 830                 if (im->multiaddr == group) {
 831                         igmp_stop_timer(im);
 832                         break;
 833                 }
 834         }
 835         rcu_read_unlock();
 836         return false;
 837 }
 838
 839 /* return true if packet was dropped */
 840 static bool igmp_heard_query(struct in_device *in_dev, struct sk_buff *skb,
 841         int len)
 842 {
 843         struct igmphdr          *ih = igmp_hdr(skb);
 844         struct igmpv3_query *ih3 = igmpv3_query_hdr(skb);
 845         struct ip_mc_list       *im;
 846         __be32                  group = ih->group;
 847         int                     max_delay;
 848         int                     mark = 0;
 849
 850
 851         if (len == 8) {
 852                 if (ih->code == 0) {
 853                         /* Alas, old v1 router presents here. */
 854
 855                         max_delay = IGMP_Query_Response_Interval;
 856                         in_dev->mr_v1_seen = jiffies +
 857                                 IGMP_V1_Router_Present_Timeout;
 858                         group = 0;
 859                 } else {
 860                         /* v2 router present */
 861                         max_delay = ih->code*(HZ/IGMP_TIMER_SCALE);
 862                         in_dev->mr_v2_seen = jiffies +
 863                                 IGMP_V2_Router_Present_Timeout;
 864                 }
 865                 /* cancel the interface change timer */
 866                 in_dev->mr_ifc_count = 0;
 867                 if (del_timer(&in_dev->mr_ifc_timer))
 868                         __in_dev_put(in_dev);
 869                 /* clear deleted report items */
 870                 igmpv3_clear_delrec(in_dev);
 871         } else if (len < 12) {
 872                 return true;    /* ignore bogus packet; freed by caller */
 873         } else if (IGMP_V1_SEEN(in_dev)) {
 874                 /* This is a v3 query with v1 queriers present */
 875                 max_delay = IGMP_Query_Response_Interval;
 876                 group = 0;
 877         } else if (IGMP_V2_SEEN(in_dev)) {
 878                 /* this is a v3 query with v2 queriers present;
 879                  * Interpretation of the max_delay code is problematic here.
 880                  * A real v2 host would use ih_code directly, while v3 has a
 881                  * different encoding. We use the v3 encoding as more likely
 882                  * to be intended in a v3 query.
 883                  */
 884                 max_delay = IGMPV3_MRC(ih3->code)*(HZ/IGMP_TIMER_SCALE);
 885                 if (!max_delay)
 886                         max_delay = 1;  /* can't mod w/ 0 */
 887         } else { /* v3 */
 888                 if (!pskb_may_pull(skb, sizeof(struct igmpv3_query)))
 889                         return true;
 890
 891                 ih3 = igmpv3_query_hdr(skb);
 892                 if (ih3->nsrcs) {
 893                         if (!pskb_may_pull(skb, sizeof(struct igmpv3_query)
 894                                            + ntohs(ih3->nsrcs)*sizeof(__be32)))
 895                                 return true;
 896                         ih3 = igmpv3_query_hdr(skb);
 897                 }
 898
 899                 max_delay = IGMPV3_MRC(ih3->code)*(HZ/IGMP_TIMER_SCALE);
 900                 if (!max_delay)
 901                         max_delay = 1;  /* can't mod w/ 0 */
 902                 in_dev->mr_maxdelay = max_delay;
 903                 if (ih3->qrv)
 904                         in_dev->mr_qrv = ih3->qrv;
 905                 if (!group) { /* general query */
 906                         if (ih3->nsrcs)
 907                                 return false;   /* no sources allowed */
 908                         igmp_gq_start_timer(in_dev);
 909                         return false;
 910                 }
 911                 /* mark sources to include, if group & source-specific */
 912                 mark = ih3->nsrcs != 0;
 913         }
 914
 915         /*
 916          * - Start the timers in all of our membership records
 917          *   that the query applies to for the interface on
 918          *   which the query arrived excl. those that belong
 919          *   to a "local" group (224.0.0.X)
 920          * - For timers already running check if they need to
 921          *   be reset.
 922          * - Use the igmp->igmp_code field as the maximum
 923          *   delay possible
 924          */
 925         rcu_read_lock();
 926         for_each_pmc_rcu(in_dev, im) {
 927                 int changed;
 928
 929                 if (group && group != im->multiaddr)
 930                         continue;
 931                 if (im->multiaddr == IGMP_ALL_HOSTS)
 932                         continue;
 933                 spin_lock_bh(&im->lock);
 934                 if (im->tm_running)
 935                         im->gsquery = im->gsquery && mark;
 936                 else
 937                         im->gsquery = mark;
 938                 changed = !im->gsquery ||
 939                         igmp_marksources(im, ntohs(ih3->nsrcs), ih3->srcs);
 940                 spin_unlock_bh(&im->lock);
 941                 if (changed)
 942                         igmp_mod_timer(im, max_delay);
 943         }
 944         rcu_read_unlock();
 945         return false;
 946 }
 947
 948 /* called in rcu_read_lock() section */
 949 int igmp_rcv(struct sk_buff *skb)
 950 {
 951         /* This basically follows the spec line by line -- see RFC1112 */
 952         struct igmphdr *ih;
 953         struct in_device *in_dev = __in_dev_get_rcu(skb->dev);
 954         int len = skb->len;
 955         bool dropped = true;
 956
 957         if (in_dev == NULL)
 958                 goto drop;
 959
 960         if (!pskb_may_pull(skb, sizeof(struct igmphdr)))
 961                 goto drop;
 962
 963         switch (skb->ip_summed) {
 964         case CHECKSUM_COMPLETE:
 965                 if (!csum_fold(skb->csum))
 966                         break;
 967                 /* fall through */
 968         case CHECKSUM_NONE:
 969                 skb->csum = 0;
 970                 if (__skb_checksum_complete(skb))
 971                         goto drop;
 972         }
 973
 974         ih = igmp_hdr(skb);
 975         switch (ih->type) {
 976         case IGMP_HOST_MEMBERSHIP_QUERY:
 977                 dropped = igmp_heard_query(in_dev, skb, len);
 978                 break;
 979         case IGMP_HOST_MEMBERSHIP_REPORT:
 980         case IGMPV2_HOST_MEMBERSHIP_REPORT:
 981                 /* Is it our report looped back? */
 982                 if (rt_is_output_route(skb_rtable(skb)))
 983                         break;
 984                 /* don't rely on MC router hearing unicast reports */
 985                 if (skb->pkt_type == PACKET_MULTICAST ||
 986                     skb->pkt_type == PACKET_BROADCAST)
 987                         dropped = igmp_heard_report(in_dev, ih->group);
 988                 break;
 989         case IGMP_PIM:
 990 #ifdef CONFIG_IP_PIMSM_V1
 991                 return pim_rcv_v1(skb);
 992 #endif
 993         case IGMPV3_HOST_MEMBERSHIP_REPORT:
 994         case IGMP_DVMRP:
 995         case IGMP_TRACE:
 996         case IGMP_HOST_LEAVE_MESSAGE:
 997         case IGMP_MTRACE:
 998         case IGMP_MTRACE_RESP:
 999                 break;
1000         default:
1001                 break;
1002         }
1003
1004 drop:
1005         if (dropped)
1006                 kfree_skb(skb);
1007         else
1008                 consume_skb(skb);
1009         return 0;
1010 }
1011
1012 #endif
1013
1014
1015 /*
1016  *      Add a filter to a device
1017  */
1018
1019 static void ip_mc_filter_add(struct in_device *in_dev, __be32 addr)
1020 {
1021         char buf[MAX_ADDR_LEN];
1022         struct net_device *dev = in_dev->dev;
1023
1024         /* Checking for IFF_MULTICAST here is WRONG-WRONG-WRONG.
1025            We will get multicast token leakage, when IFF_MULTICAST
1026            is changed. This check should be done in ndo_set_rx_mode
1027            routine. Something sort of:
1028            if (dev->mc_list && dev->flags&IFF_MULTICAST) { do it; }
1029            --ANK
1030            */
1031         if (arp_mc_map(addr, buf, dev, 0) == 0)
1032                 dev_mc_add(dev, buf);
1033 }
1034
1035 /*
1036  *      Remove a filter from a device
1037  */
1038
1039 static void ip_mc_filter_del(struct in_device *in_dev, __be32 addr)
1040 {
1041         char buf[MAX_ADDR_LEN];
1042         struct net_device *dev = in_dev->dev;
1043
1044         if (arp_mc_map(addr, buf, dev, 0) == 0)
1045                 dev_mc_del(dev, buf);
1046 }
1047
1048 #ifdef CONFIG_IP_MULTICAST
1049 /*
1050  * deleted ip_mc_list manipulation
1051  */
1052 static void igmpv3_add_delrec(struct in_device *in_dev, struct ip_mc_list *im)
1053 {
1054         struct ip_mc_list *pmc;
1055
1056         /* this is an "ip_mc_list" for convenience; only the fields below
1057          * are actually used. In particular, the refcnt and users are not
1058          * used for management of the delete list. Using the same structure
1059          * for deleted items allows change reports to use common code with
1060          * non-deleted or query-response MCA's.
1061          */
1062         pmc = kzalloc(sizeof(*pmc), GFP_KERNEL);
1063         if (!pmc)
1064                 return;
1065         spin_lock_bh(&im->lock);
1066         pmc->interface = im->interface;
1067         in_dev_hold(in_dev);
1068         pmc->multiaddr = im->multiaddr;
1069         pmc->crcount = in_dev->mr_qrv ? in_dev->mr_qrv :
1070                 IGMP_Unsolicited_Report_Count;
1071         pmc->sfmode = im->sfmode;
1072         if (pmc->sfmode == MCAST_INCLUDE) {
1073                 struct ip_sf_list *psf;
1074
1075                 pmc->tomb = im->tomb;
1076                 pmc->sources = im->sources;
1077                 im->tomb = im->sources = NULL;
1078                 for (psf=pmc->sources; psf; psf=psf->sf_next)
1079                         psf->sf_crcount = pmc->crcount;
1080         }
1081         spin_unlock_bh(&im->lock);
1082
1083         spin_lock_bh(&in_dev->mc_tomb_lock);
1084         pmc->next = in_dev->mc_tomb;
1085         in_dev->mc_tomb = pmc;
1086         spin_unlock_bh(&in_dev->mc_tomb_lock);
1087 }
1088
1089 static void igmpv3_del_delrec(struct in_device *in_dev, __be32 multiaddr)
1090 {
1091         struct ip_mc_list *pmc, *pmc_prev;
1092         struct ip_sf_list *psf, *psf_next;
1093
1094         spin_lock_bh(&in_dev->mc_tomb_lock);
1095         pmc_prev = NULL;
1096         for (pmc=in_dev->mc_tomb; pmc; pmc=pmc->next) {
1097                 if (pmc->multiaddr == multiaddr)
1098                         break;
1099                 pmc_prev = pmc;
1100         }
1101         if (pmc) {
1102                 if (pmc_prev)
1103                         pmc_prev->next = pmc->next;
1104                 else
1105                         in_dev->mc_tomb = pmc->next;
1106         }
1107         spin_unlock_bh(&in_dev->mc_tomb_lock);
1108         if (pmc) {
1109                 for (psf=pmc->tomb; psf; psf=psf_next) {
1110                         psf_next = psf->sf_next;
1111                         kfree(psf);
1112                 }
1113                 in_dev_put(pmc->interface);
1114                 kfree(pmc);
1115         }
1116 }
1117
1118 static void igmpv3_clear_delrec(struct in_device *in_dev)
1119 {
1120         struct ip_mc_list *pmc, *nextpmc;
1121
1122         spin_lock_bh(&in_dev->mc_tomb_lock);
1123         pmc = in_dev->mc_tomb;
1124         in_dev->mc_tomb = NULL;
1125         spin_unlock_bh(&in_dev->mc_tomb_lock);
1126
1127         for (; pmc; pmc = nextpmc) {
1128                 nextpmc = pmc->next;
1129                 ip_mc_clear_src(pmc);
1130                 in_dev_put(pmc->interface);
1131                 kfree(pmc);
1132         }
1133         /* clear dead sources, too */
1134         rcu_read_lock();
1135         for_each_pmc_rcu(in_dev, pmc) {
1136                 struct ip_sf_list *psf, *psf_next;
1137
1138                 spin_lock_bh(&pmc->lock);
1139                 psf = pmc->tomb;
1140                 pmc->tomb = NULL;
1141                 spin_unlock_bh(&pmc->lock);
1142                 for (; psf; psf=psf_next) {
1143                         psf_next = psf->sf_next;
1144                         kfree(psf);
1145                 }
1146         }
1147         rcu_read_unlock();
1148 }
1149 #endif
1150
1151 static void igmp_group_dropped(struct ip_mc_list *im)
1152 {
1153         struct in_device *in_dev = im->interface;
1154 #ifdef CONFIG_IP_MULTICAST
1155         int reporter;
1156 #endif
1157
1158         if (im->loaded) {
1159                 im->loaded = 0;
1160                 ip_mc_filter_del(in_dev, im->multiaddr);
1161         }
1162
1163 #ifdef CONFIG_IP_MULTICAST
1164         if (im->multiaddr == IGMP_ALL_HOSTS)
1165                 return;
1166
1167         reporter = im->reporter;
1168         igmp_stop_timer(im);
1169
1170         if (!in_dev->dead) {
1171                 if (IGMP_V1_SEEN(in_dev))
1172                         return;
1173                 if (IGMP_V2_SEEN(in_dev)) {
1174                         if (reporter)
1175                                 igmp_send_report(in_dev, im, IGMP_HOST_LEAVE_MESSAGE);
1176                         return;
1177                 }
1178                 /* IGMPv3 */
1179                 igmpv3_add_delrec(in_dev, im);
1180
1181                 igmp_ifc_event(in_dev);
1182         }
1183 #endif
1184 }
1185
1186 static void igmp_group_added(struct ip_mc_list *im)
1187 {
1188         struct in_device *in_dev = im->interface;
1189
1190         if (im->loaded == 0) {
1191                 im->loaded = 1;
1192                 ip_mc_filter_add(in_dev, im->multiaddr);
1193         }
1194
1195 #ifdef CONFIG_IP_MULTICAST
1196         if (im->multiaddr == IGMP_ALL_HOSTS)
1197                 return;
1198
1199         if (in_dev->dead)
1200                 return;
1201         if (IGMP_V1_SEEN(in_dev) || IGMP_V2_SEEN(in_dev)) {
1202                 spin_lock_bh(&im->lock);
1203                 igmp_start_timer(im, IGMP_Initial_Report_Delay);
1204                 spin_unlock_bh(&im->lock);
1205                 return;
1206         }
1207         /* else, v3 */
1208
1209         im->crcount = in_dev->mr_qrv ? in_dev->mr_qrv :
1210                 IGMP_Unsolicited_Report_Count;
1211         igmp_ifc_event(in_dev);
1212 #endif
1213 }
1214
1215
1216 /*
1217  *      Multicast list managers
1218  */
1219
1220 static u32 ip_mc_hash(const struct ip_mc_list *im)
1221 {
1222         return hash_32((__force u32)im->multiaddr, MC_HASH_SZ_LOG);
1223 }
1224
1225 static void ip_mc_hash_add(struct in_device *in_dev,
1226                            struct ip_mc_list *im)
1227 {
1228         struct ip_mc_list __rcu **mc_hash;
1229         u32 hash;
1230
1231         mc_hash = rtnl_dereference(in_dev->mc_hash);
1232         if (mc_hash) {
1233                 hash = ip_mc_hash(im);
1234                 im->next_hash = mc_hash[hash];
1235                 rcu_assign_pointer(mc_hash[hash], im);
1236                 return;
1237         }
1238
1239         /* do not use a hash table for small number of items */
1240         if (in_dev->mc_count < 4)
1241                 return;
1242
1243         mc_hash = kzalloc(sizeof(struct ip_mc_list *) << MC_HASH_SZ_LOG,
1244                           GFP_KERNEL);
1245         if (!mc_hash)
1246                 return;
1247
1248         for_each_pmc_rtnl(in_dev, im) {
1249                 hash = ip_mc_hash(im);
1250                 im->next_hash = mc_hash[hash];
1251                 RCU_INIT_POINTER(mc_hash[hash], im);
1252         }
1253
1254         rcu_assign_pointer(in_dev->mc_hash, mc_hash);
1255 }
1256
1257 static void ip_mc_hash_remove(struct in_device *in_dev,
1258                               struct ip_mc_list *im)
1259 {
1260         struct ip_mc_list __rcu **mc_hash = rtnl_dereference(in_dev->mc_hash);
1261         struct ip_mc_list *aux;
1262
1263         if (!mc_hash)
1264                 return;
1265         mc_hash += ip_mc_hash(im);
1266         while ((aux = rtnl_dereference(*mc_hash)) != im)
1267                 mc_hash = &aux->next_hash;
1268         *mc_hash = im->next_hash;
1269 }
1270
1271
1272 /*
1273  *      A socket has joined a multicast group on device dev.
1274  */
1275
1276 void ip_mc_inc_group(struct in_device *in_dev, __be32 addr)
1277 {
1278         struct ip_mc_list *im;
1279
1280         ASSERT_RTNL();
1281
1282         for_each_pmc_rtnl(in_dev, im) {
1283                 if (im->multiaddr == addr) {
1284                         im->users++;
1285                         ip_mc_add_src(in_dev, &addr, MCAST_EXCLUDE, 0, NULL, 0);
1286                         goto out;
1287                 }
1288         }
1289
1290         im = kzalloc(sizeof(*im), GFP_KERNEL);
1291         if (!im)
1292                 goto out;
1293
1294         im->users = 1;
1295         im->interface = in_dev;
1296         in_dev_hold(in_dev);
1297         im->multiaddr = addr;
1298         /* initial mode is (EX, empty) */
1299         im->sfmode = MCAST_EXCLUDE;
1300         im->sfcount[MCAST_EXCLUDE] = 1;
1301         atomic_set(&im->refcnt, 1);
1302         spin_lock_init(&im->lock);
1303 #ifdef CONFIG_IP_MULTICAST
1304         setup_timer(&im->timer, &igmp_timer_expire, (unsigned long)im);
1305         im->unsolicit_count = IGMP_Unsolicited_Report_Count;
1306 #endif
1307
1308         im->next_rcu = in_dev->mc_list;
1309         in_dev->mc_count++;
1310         rcu_assign_pointer(in_dev->mc_list, im);
1311
1312         ip_mc_hash_add(in_dev, im);
1313
1314 #ifdef CONFIG_IP_MULTICAST
1315         igmpv3_del_delrec(in_dev, im->multiaddr);
1316 #endif
1317         igmp_group_added(im);
1318         if (!in_dev->dead)
1319                 ip_rt_multicast_event(in_dev);
1320 out:
1321         return;
1322 }
1323 EXPORT_SYMBOL(ip_mc_inc_group);
1324
1325 /*
1326  *      Resend IGMP JOIN report; used for bonding.
1327  *      Called with rcu_read_lock()
1328  */
1329 void ip_mc_rejoin_groups(struct in_device *in_dev)
1330 {
1331 #ifdef CONFIG_IP_MULTICAST
1332         struct ip_mc_list *im;
1333         int type;
1334
1335         for_each_pmc_rcu(in_dev, im) {
1336                 if (im->multiaddr == IGMP_ALL_HOSTS)
1337                         continue;
1338
1339                 /* a failover is happening and switches
1340                  * must be notified immediately
1341                  */
1342                 if (IGMP_V1_SEEN(in_dev))
1343                         type = IGMP_HOST_MEMBERSHIP_REPORT;
1344                 else if (IGMP_V2_SEEN(in_dev))
1345                         type = IGMPV2_HOST_MEMBERSHIP_REPORT;
1346                 else
1347                         type = IGMPV3_HOST_MEMBERSHIP_REPORT;
1348                 igmp_send_report(in_dev, im, type);
1349         }
1350 #endif
1351 }
1352 EXPORT_SYMBOL(ip_mc_rejoin_groups);
1353
1354 /*
1355  *      A socket has left a multicast group on device dev
1356  */
1357
1358 void ip_mc_dec_group(struct in_device *in_dev, __be32 addr)
1359 {
1360         struct ip_mc_list *i;
1361         struct ip_mc_list __rcu **ip;
1362
1363         ASSERT_RTNL();
1364
1365         for (ip = &in_dev->mc_list;
1366              (i = rtnl_dereference(*ip)) != NULL;
1367              ip = &i->next_rcu) {
1368                 if (i->multiaddr == addr) {
1369                         if (--i->users == 0) {
1370                                 ip_mc_hash_remove(in_dev, i);
1371                                 *ip = i->next_rcu;
1372                                 in_dev->mc_count--;
1373                                 igmp_group_dropped(i);
1374                                 ip_mc_clear_src(i);
1375
1376                                 if (!in_dev->dead)
1377                                         ip_rt_multicast_event(in_dev);
1378
1379                                 ip_ma_put(i);
1380                                 return;
1381                         }
1382                         break;
1383                 }
1384         }
1385 }
1386 EXPORT_SYMBOL(ip_mc_dec_group);
1387
1388 /* Device changing type */
1389
1390 void ip_mc_unmap(struct in_device *in_dev)
1391 {
1392         struct ip_mc_list *pmc;
1393
1394         ASSERT_RTNL();
1395
1396         for_each_pmc_rtnl(in_dev, pmc)
1397                 igmp_group_dropped(pmc);
1398 }
1399
1400 void ip_mc_remap(struct in_device *in_dev)
1401 {
1402         struct ip_mc_list *pmc;
1403
1404         ASSERT_RTNL();
1405
1406         for_each_pmc_rtnl(in_dev, pmc)
1407                 igmp_group_added(pmc);
1408 }
1409
1410 /* Device going down */
1411
1412 void ip_mc_down(struct in_device *in_dev)
1413 {
1414         struct ip_mc_list *pmc;
1415
1416         ASSERT_RTNL();
1417
1418         for_each_pmc_rtnl(in_dev, pmc)
1419                 igmp_group_dropped(pmc);
1420
1421 #ifdef CONFIG_IP_MULTICAST
1422         in_dev->mr_ifc_count = 0;
1423         if (del_timer(&in_dev->mr_ifc_timer))
1424                 __in_dev_put(in_dev);
1425         in_dev->mr_gq_running = 0;
1426         if (del_timer(&in_dev->mr_gq_timer))
1427                 __in_dev_put(in_dev);
1428         igmpv3_clear_delrec(in_dev);
1429 #endif
1430
1431         ip_mc_dec_group(in_dev, IGMP_ALL_HOSTS);
1432 }
1433
1434 void ip_mc_init_dev(struct in_device *in_dev)
1435 {
1436         ASSERT_RTNL();
1437
1438 #ifdef CONFIG_IP_MULTICAST
1439         setup_timer(&in_dev->mr_gq_timer, igmp_gq_timer_expire,
1440                         (unsigned long)in_dev);
1441         setup_timer(&in_dev->mr_ifc_timer, igmp_ifc_timer_expire,
1442                         (unsigned long)in_dev);
1443         in_dev->mr_qrv = IGMP_Unsolicited_Report_Count;
1444 #endif
1445
1446         spin_lock_init(&in_dev->mc_tomb_lock);
1447 }
1448
1449 /* Device going up */
1450
1451 void ip_mc_up(struct in_device *in_dev)
1452 {
1453         struct ip_mc_list *pmc;
1454
1455         ASSERT_RTNL();
1456
1457         ip_mc_inc_group(in_dev, IGMP_ALL_HOSTS);
1458
1459         for_each_pmc_rtnl(in_dev, pmc)
1460                 igmp_group_added(pmc);
1461 }
1462
1463 /*
1464  *      Device is about to be destroyed: clean up.
1465  */
1466
1467 void ip_mc_destroy_dev(struct in_device *in_dev)
1468 {
1469         struct ip_mc_list *i;
1470
1471         ASSERT_RTNL();
1472
1473         /* Deactivate timers */
1474         ip_mc_down(in_dev);
1475
1476         while ((i = rtnl_dereference(in_dev->mc_list)) != NULL) {
1477                 in_dev->mc_list = i->next_rcu;
1478                 in_dev->mc_count--;
1479
1480                 /* We've dropped the groups in ip_mc_down already */
1481                 ip_mc_clear_src(i);
1482                 ip_ma_put(i);
1483         }
1484 }
1485
1486 /* RTNL is locked */
1487 static struct in_device *ip_mc_find_dev(struct net *net, struct ip_mreqn *imr)
1488 {
1489         struct net_device *dev = NULL;
1490         struct in_device *idev = NULL;
1491
1492         if (imr->imr_ifindex) {
1493                 idev = inetdev_by_index(net, imr->imr_ifindex);
1494                 return idev;
1495         }
1496         if (imr->imr_address.s_addr) {
1497                 dev = __ip_dev_find(net, imr->imr_address.s_addr, false);
1498                 if (!dev)
1499                         return NULL;
1500         }
1501
1502         if (!dev) {
1503                 struct rtable *rt = ip_route_output(net,
1504                                                     imr->imr_multiaddr.s_addr,
1505                                                     0, 0, 0);
1506                 if (!IS_ERR(rt)) {
1507                         dev = rt->dst.dev;
1508                         ip_rt_put(rt);
1509                 }
1510         }
1511         if (dev) {
1512                 imr->imr_ifindex = dev->ifindex;
1513                 idev = __in_dev_get_rtnl(dev);
1514         }
1515         return idev;
1516 }
1517
1518 /*
1519  *      Join a socket to a group
1520  */
1521 int sysctl_igmp_max_memberships __read_mostly = IP_MAX_MEMBERSHIPS;
1522 int sysctl_igmp_max_msf __read_mostly = IP_MAX_MSF;
1523
1524
1525 static int ip_mc_del1_src(struct ip_mc_list *pmc, int sfmode,
1526         __be32 *psfsrc)
1527 {
1528         struct ip_sf_list *psf, *psf_prev;
1529         int rv = 0;
1530
1531         psf_prev = NULL;
1532         for (psf=pmc->sources; psf; psf=psf->sf_next) {
1533                 if (psf->sf_inaddr == *psfsrc)
1534                         break;
1535                 psf_prev = psf;
1536         }
1537         if (!psf || psf->sf_count[sfmode] == 0) {
1538                 /* source filter not found, or count wrong =>  bug */
1539                 return -ESRCH;
1540         }
1541         psf->sf_count[sfmode]--;
1542         if (psf->sf_count[sfmode] == 0) {
1543                 ip_rt_multicast_event(pmc->interface);
1544         }
1545         if (!psf->sf_count[MCAST_INCLUDE] && !psf->sf_count[MCAST_EXCLUDE]) {
1546 #ifdef CONFIG_IP_MULTICAST
1547                 struct in_device *in_dev = pmc->interface;
1548 #endif
1549
1550                 /* no more filters for this source */
1551                 if (psf_prev)
1552                         psf_prev->sf_next = psf->sf_next;
1553                 else
1554                         pmc->sources = psf->sf_next;
1555 #ifdef CONFIG_IP_MULTICAST
1556                 if (psf->sf_oldin &&
1557                     !IGMP_V1_SEEN(in_dev) && !IGMP_V2_SEEN(in_dev)) {
1558                         psf->sf_crcount = in_dev->mr_qrv ? in_dev->mr_qrv :
1559                                 IGMP_Unsolicited_Report_Count;
1560                         psf->sf_next = pmc->tomb;
1561                         pmc->tomb = psf;
1562                         rv = 1;
1563                 } else
1564 #endif
1565                         kfree(psf);
1566         }
1567         return rv;
1568 }
1569
1570 #ifndef CONFIG_IP_MULTICAST
1571 #define igmp_ifc_event(x)       do { } while (0)
1572 #endif
1573
1574 static int ip_mc_del_src(struct in_device *in_dev, __be32 *pmca, int sfmode,
1575                          int sfcount, __be32 *psfsrc, int delta)
1576 {
1577         struct ip_mc_list *pmc;
1578         int     changerec = 0;
1579         int     i, err;
1580
1581         if (!in_dev)
1582                 return -ENODEV;
1583         rcu_read_lock();
1584         for_each_pmc_rcu(in_dev, pmc) {
1585                 if (*pmca == pmc->multiaddr)
1586                         break;
1587         }
1588         if (!pmc) {
1589                 /* MCA not found?? bug */
1590                 rcu_read_unlock();
1591                 return -ESRCH;
1592         }
1593         spin_lock_bh(&pmc->lock);
1594         rcu_read_unlock();
1595 #ifdef CONFIG_IP_MULTICAST
1596         sf_markstate(pmc);
1597 #endif
1598         if (!delta) {
1599                 err = -EINVAL;
1600                 if (!pmc->sfcount[sfmode])
1601                         goto out_unlock;
1602                 pmc->sfcount[sfmode]--;
1603         }
1604         err = 0;
1605         for (i=0; i<sfcount; i++) {
1606                 int rv = ip_mc_del1_src(pmc, sfmode, &psfsrc[i]);
1607
1608                 changerec |= rv > 0;
1609                 if (!err && rv < 0)
1610                         err = rv;
1611         }
1612         if (pmc->sfmode == MCAST_EXCLUDE &&
1613             pmc->sfcount[MCAST_EXCLUDE] == 0 &&
1614             pmc->sfcount[MCAST_INCLUDE]) {
1615 #ifdef CONFIG_IP_MULTICAST
1616                 struct ip_sf_list *psf;
1617 #endif
1618
1619                 /* filter mode change */
1620                 pmc->sfmode = MCAST_INCLUDE;
1621 #ifdef CONFIG_IP_MULTICAST
1622                 pmc->crcount = in_dev->mr_qrv ? in_dev->mr_qrv :
1623                         IGMP_Unsolicited_Report_Count;
1624                 in_dev->mr_ifc_count = pmc->crcount;
1625                 for (psf=pmc->sources; psf; psf = psf->sf_next)
1626                         psf->sf_crcount = 0;
1627                 igmp_ifc_event(pmc->interface);
1628         } else if (sf_setstate(pmc) || changerec) {
1629                 igmp_ifc_event(pmc->interface);
1630 #endif
1631         }
1632 out_unlock:
1633         spin_unlock_bh(&pmc->lock);
1634         return err;
1635 }
1636
1637 /*
1638  * Add multicast single-source filter to the interface list
1639  */
1640 static int ip_mc_add1_src(struct ip_mc_list *pmc, int sfmode,
1641         __be32 *psfsrc)
1642 {
1643         struct ip_sf_list *psf, *psf_prev;
1644
1645         psf_prev = NULL;
1646         for (psf=pmc->sources; psf; psf=psf->sf_next) {
1647                 if (psf->sf_inaddr == *psfsrc)
1648                         break;
1649                 psf_prev = psf;
1650         }
1651         if (!psf) {
1652                 psf = kzalloc(sizeof(*psf), GFP_ATOMIC);
1653                 if (!psf)
1654                         return -ENOBUFS;
1655                 psf->sf_inaddr = *psfsrc;
1656                 if (psf_prev) {
1657                         psf_prev->sf_next = psf;
1658                 } else
1659                         pmc->sources = psf;
1660         }
1661         psf->sf_count[sfmode]++;
1662         if (psf->sf_count[sfmode] == 1) {
1663                 ip_rt_multicast_event(pmc->interface);
1664         }
1665         return 0;
1666 }
1667
1668 #ifdef CONFIG_IP_MULTICAST
1669 static void sf_markstate(struct ip_mc_list *pmc)
1670 {
1671         struct ip_sf_list *psf;
1672         int mca_xcount = pmc->sfcount[MCAST_EXCLUDE];
1673
1674         for (psf=pmc->sources; psf; psf=psf->sf_next)
1675                 if (pmc->sfcount[MCAST_EXCLUDE]) {
1676                         psf->sf_oldin = mca_xcount ==
1677                                 psf->sf_count[MCAST_EXCLUDE] &&
1678                                 !psf->sf_count[MCAST_INCLUDE];
1679                 } else
1680                         psf->sf_oldin = psf->sf_count[MCAST_INCLUDE] != 0;
1681 }
1682
1683 static int sf_setstate(struct ip_mc_list *pmc)
1684 {
1685         struct ip_sf_list *psf, *dpsf;
1686         int mca_xcount = pmc->sfcount[MCAST_EXCLUDE];
1687         int qrv = pmc->interface->mr_qrv;
1688         int new_in, rv;
1689
1690         rv = 0;
1691         for (psf=pmc->sources; psf; psf=psf->sf_next) {
1692                 if (pmc->sfcount[MCAST_EXCLUDE]) {
1693                         new_in = mca_xcount == psf->sf_count[MCAST_EXCLUDE] &&
1694                                 !psf->sf_count[MCAST_INCLUDE];
1695                 } else
1696                         new_in = psf->sf_count[MCAST_INCLUDE] != 0;
1697                 if (new_in) {
1698                         if (!psf->sf_oldin) {
1699                                 struct ip_sf_list *prev = NULL;
1700
1701                                 for (dpsf=pmc->tomb; dpsf; dpsf=dpsf->sf_next) {
1702                                         if (dpsf->sf_inaddr == psf->sf_inaddr)
1703                                                 break;
1704                                         prev = dpsf;
1705                                 }
1706                                 if (dpsf) {
1707                                         if (prev)
1708                                                 prev->sf_next = dpsf->sf_next;
1709                                         else
1710                                                 pmc->tomb = dpsf->sf_next;
1711                                         kfree(dpsf);
1712                                 }
1713                                 psf->sf_crcount = qrv;
1714                                 rv++;
1715                         }
1716                 } else if (psf->sf_oldin) {
1717
1718                         psf->sf_crcount = 0;
1719                         /*
1720                          * add or update "delete" records if an active filter
1721                          * is now inactive
1722                          */
1723                         for (dpsf=pmc->tomb; dpsf; dpsf=dpsf->sf_next)
1724                                 if (dpsf->sf_inaddr == psf->sf_inaddr)
1725                                         break;
1726                         if (!dpsf) {
1727                                 dpsf = kmalloc(sizeof(*dpsf), GFP_ATOMIC);
1728                                 if (!dpsf)
1729                                         continue;
1730                                 *dpsf = *psf;
1731                                 /* pmc->lock held by callers */
1732                                 dpsf->sf_next = pmc->tomb;
1733                                 pmc->tomb = dpsf;
1734                         }
1735                         dpsf->sf_crcount = qrv;
1736                         rv++;
1737                 }
1738         }
1739         return rv;
1740 }
1741 #endif
1742
1743 /*
1744  * Add multicast source filter list to the interface list
1745  */
1746 static int ip_mc_add_src(struct in_device *in_dev, __be32 *pmca, int sfmode,
1747                          int sfcount, __be32 *psfsrc, int delta)
1748 {
1749         struct ip_mc_list *pmc;
1750         int     isexclude;
1751         int     i, err;
1752
1753         if (!in_dev)
1754                 return -ENODEV;
1755         rcu_read_lock();
1756         for_each_pmc_rcu(in_dev, pmc) {
1757                 if (*pmca == pmc->multiaddr)
1758                         break;
1759         }
1760         if (!pmc) {
1761                 /* MCA not found?? bug */
1762                 rcu_read_unlock();
1763                 return -ESRCH;
1764         }
1765         spin_lock_bh(&pmc->lock);
1766         rcu_read_unlock();
1767
1768 #ifdef CONFIG_IP_MULTICAST
1769         sf_markstate(pmc);
1770 #endif
1771         isexclude = pmc->sfmode == MCAST_EXCLUDE;
1772         if (!delta)
1773                 pmc->sfcount[sfmode]++;
1774         err = 0;
1775         for (i=0; i<sfcount; i++) {
1776                 err = ip_mc_add1_src(pmc, sfmode, &psfsrc[i]);
1777                 if (err)
1778                         break;
1779         }
1780         if (err) {
1781                 int j;
1782
1783                 if (!delta)
1784                         pmc->sfcount[sfmode]--;
1785                 for (j=0; j<i; j++)
1786                         (void) ip_mc_del1_src(pmc, sfmode, &psfsrc[j]);
1787         } else if (isexclude != (pmc->sfcount[MCAST_EXCLUDE] != 0)) {
1788 #ifdef CONFIG_IP_MULTICAST
1789                 struct ip_sf_list *psf;
1790                 in_dev = pmc->interface;
1791 #endif
1792
1793                 /* filter mode change */
1794                 if (pmc->sfcount[MCAST_EXCLUDE])
1795                         pmc->sfmode = MCAST_EXCLUDE;
1796                 else if (pmc->sfcount[MCAST_INCLUDE])
1797                         pmc->sfmode = MCAST_INCLUDE;
1798 #ifdef CONFIG_IP_MULTICAST
1799                 /* else no filters; keep old mode for reports */
1800
1801                 pmc->crcount = in_dev->mr_qrv ? in_dev->mr_qrv :
1802                         IGMP_Unsolicited_Report_Count;
1803                 in_dev->mr_ifc_count = pmc->crcount;
1804                 for (psf=pmc->sources; psf; psf = psf->sf_next)
1805                         psf->sf_crcount = 0;
1806                 igmp_ifc_event(in_dev);
1807         } else if (sf_setstate(pmc)) {
1808                 igmp_ifc_event(in_dev);
1809 #endif
1810         }
1811         spin_unlock_bh(&pmc->lock);
1812         return err;
1813 }
1814
1815 static void ip_mc_clear_src(struct ip_mc_list *pmc)
1816 {
1817         struct ip_sf_list *psf, *nextpsf;
1818
1819         for (psf=pmc->tomb; psf; psf=nextpsf) {
1820                 nextpsf = psf->sf_next;
1821                 kfree(psf);
1822         }
1823         pmc->tomb = NULL;
1824         for (psf=pmc->sources; psf; psf=nextpsf) {
1825                 nextpsf = psf->sf_next;
1826                 kfree(psf);
1827         }
1828         pmc->sources = NULL;
1829         pmc->sfmode = MCAST_EXCLUDE;
1830         pmc->sfcount[MCAST_INCLUDE] = 0;
1831         pmc->sfcount[MCAST_EXCLUDE] = 1;
1832 }
1833
1834
1835 /*
1836  * Join a multicast group
1837  */
1838 int ip_mc_join_group(struct sock *sk , struct ip_mreqn *imr)
1839 {
1840         int err;
1841         __be32 addr = imr->imr_multiaddr.s_addr;
1842         struct ip_mc_socklist *iml = NULL, *i;
1843         struct in_device *in_dev;
1844         struct inet_sock *inet = inet_sk(sk);
1845         struct net *net = sock_net(sk);
1846         int ifindex;
1847         int count = 0;
1848
1849         if (!ipv4_is_multicast(addr))
1850                 return -EINVAL;
1851
1852         rtnl_lock();
1853
1854         in_dev = ip_mc_find_dev(net, imr);
1855
1856         if (!in_dev) {
1857                 iml = NULL;
1858                 err = -ENODEV;
1859                 goto done;
1860         }
1861
1862         err = -EADDRINUSE;
1863         ifindex = imr->imr_ifindex;
1864         for_each_pmc_rtnl(inet, i) {
1865                 if (i->multi.imr_multiaddr.s_addr == addr &&
1866                     i->multi.imr_ifindex == ifindex)
1867                         goto done;
1868                 count++;
1869         }
1870         err = -ENOBUFS;
1871         if (count >= sysctl_igmp_max_memberships)
1872                 goto done;
1873         iml = sock_kmalloc(sk, sizeof(*iml), GFP_KERNEL);
1874         if (iml == NULL)
1875                 goto done;
1876
1877         memcpy(&iml->multi, imr, sizeof(*imr));
1878         iml->next_rcu = inet->mc_list;
1879         iml->sflist = NULL;
1880         iml->sfmode = MCAST_EXCLUDE;
1881         rcu_assign_pointer(inet->mc_list, iml);
1882         ip_mc_inc_group(in_dev, addr);
1883         err = 0;
1884 done:
1885         rtnl_unlock();
1886         return err;
1887 }
1888 EXPORT_SYMBOL(ip_mc_join_group);
1889
1890 static int ip_mc_leave_src(struct sock *sk, struct ip_mc_socklist *iml,
1891                            struct in_device *in_dev)
1892 {
1893         struct ip_sf_socklist *psf = rtnl_dereference(iml->sflist);
1894         int err;
1895
1896         if (psf == NULL) {
1897                 /* any-source empty exclude case */
1898                 return ip_mc_del_src(in_dev, &iml->multi.imr_multiaddr.s_addr,
1899                         iml->sfmode, 0, NULL, 0);
1900         }
1901         err = ip_mc_del_src(in_dev, &iml->multi.imr_multiaddr.s_addr,
1902                         iml->sfmode, psf->sl_count, psf->sl_addr, 0);
1903         RCU_INIT_POINTER(iml->sflist, NULL);
1904         /* decrease mem now to avoid the memleak warning */
1905         atomic_sub(IP_SFLSIZE(psf->sl_max), &sk->sk_omem_alloc);
1906         kfree_rcu(psf, rcu);
1907         return err;
1908 }
1909
1910 /*
1911  *      Ask a socket to leave a group.
1912  */
1913
1914 int ip_mc_leave_group(struct sock *sk, struct ip_mreqn *imr)
1915 {
1916         struct inet_sock *inet = inet_sk(sk);
1917         struct ip_mc_socklist *iml;
1918         struct ip_mc_socklist __rcu **imlp;
1919         struct in_device *in_dev;
1920         struct net *net = sock_net(sk);
1921         __be32 group = imr->imr_multiaddr.s_addr;
1922         u32 ifindex;
1923         int ret = -EADDRNOTAVAIL;
1924
1925         rtnl_lock();
1926         in_dev = ip_mc_find_dev(net, imr);
1927         ifindex = imr->imr_ifindex;
1928         for (imlp = &inet->mc_list;
1929              (iml = rtnl_dereference(*imlp)) != NULL;
1930              imlp = &iml->next_rcu) {
1931                 if (iml->multi.imr_multiaddr.s_addr != group)
1932                         continue;
1933                 if (ifindex) {
1934                         if (iml->multi.imr_ifindex != ifindex)
1935                                 continue;
1936                 } else if (imr->imr_address.s_addr && imr->imr_address.s_addr !=
1937                                 iml->multi.imr_address.s_addr)
1938                         continue;
1939
1940                 (void) ip_mc_leave_src(sk, iml, in_dev);
1941
1942                 *imlp = iml->next_rcu;
1943
1944                 if (in_dev)
1945                         ip_mc_dec_group(in_dev, group);
1946                 rtnl_unlock();
1947                 /* decrease mem now to avoid the memleak warning */
1948                 atomic_sub(sizeof(*iml), &sk->sk_omem_alloc);
1949                 kfree_rcu(iml, rcu);
1950                 return 0;
1951         }
1952         if (!in_dev)
1953                 ret = -ENODEV;
1954         rtnl_unlock();
1955         return ret;
1956 }
1957 EXPORT_SYMBOL(ip_mc_leave_group);
1958
1959 int ip_mc_source(int add, int omode, struct sock *sk, struct
1960         ip_mreq_source *mreqs, int ifindex)
1961 {
1962         int err;
1963         struct ip_mreqn imr;
1964         __be32 addr = mreqs->imr_multiaddr;
1965         struct ip_mc_socklist *pmc;
1966         struct in_device *in_dev = NULL;
1967         struct inet_sock *inet = inet_sk(sk);
1968         struct ip_sf_socklist *psl;
1969         struct net *net = sock_net(sk);
1970         int leavegroup = 0;
1971         int i, j, rv;
1972
1973         if (!ipv4_is_multicast(addr))
1974                 return -EINVAL;
1975
1976         rtnl_lock();
1977
1978         imr.imr_multiaddr.s_addr = mreqs->imr_multiaddr;
1979         imr.imr_address.s_addr = mreqs->imr_interface;
1980         imr.imr_ifindex = ifindex;
1981         in_dev = ip_mc_find_dev(net, &imr);
1982
1983         if (!in_dev) {
1984                 err = -ENODEV;
1985                 goto done;
1986         }
1987         err = -EADDRNOTAVAIL;
1988
1989         for_each_pmc_rtnl(inet, pmc) {
1990                 if ((pmc->multi.imr_multiaddr.s_addr ==
1991                      imr.imr_multiaddr.s_addr) &&
1992                     (pmc->multi.imr_ifindex == imr.imr_ifindex))
1993                         break;
1994         }
1995         if (!pmc) {             /* must have a prior join */
1996                 err = -EINVAL;
1997                 goto done;
1998         }
1999         /* if a source filter was set, must be the same mode as before */
2000         if (pmc->sflist) {
2001                 if (pmc->sfmode != omode) {
2002                         err = -EINVAL;
2003                         goto done;
2004                 }
2005         } else if (pmc->sfmode != omode) {
2006                 /* allow mode switches for empty-set filters */
2007                 ip_mc_add_src(in_dev, &mreqs->imr_multiaddr, omode, 0, NULL, 0);
2008                 ip_mc_del_src(in_dev, &mreqs->imr_multiaddr, pmc->sfmode, 0,
2009                         NULL, 0);
2010                 pmc->sfmode = omode;
2011         }
2012
2013         psl = rtnl_dereference(pmc->sflist);
2014         if (!add) {
2015                 if (!psl)
2016                         goto done;      /* err = -EADDRNOTAVAIL */
2017                 rv = !0;
2018                 for (i=0; i<psl->sl_count; i++) {
2019                         rv = memcmp(&psl->sl_addr[i], &mreqs->imr_sourceaddr,
2020                                 sizeof(__be32));
2021                         if (rv == 0)
2022                                 break;
2023                 }
2024                 if (rv)         /* source not found */
2025                         goto done;      /* err = -EADDRNOTAVAIL */
2026
2027                 /* special case - (INCLUDE, empty) == LEAVE_GROUP */
2028                 if (psl->sl_count == 1 && omode == MCAST_INCLUDE) {
2029                         leavegroup = 1;
2030                         goto done;
2031                 }
2032
2033                 /* update the interface filter */
2034                 ip_mc_del_src(in_dev, &mreqs->imr_multiaddr, omode, 1,
2035                         &mreqs->imr_sourceaddr, 1);
2036
2037                 for (j=i+1; j<psl->sl_count; j++)
2038                         psl->sl_addr[j-1] = psl->sl_addr[j];
2039                 psl->sl_count--;
2040                 err = 0;
2041                 goto done;
2042         }
2043         /* else, add a new source to the filter */
2044
2045         if (psl && psl->sl_count >= sysctl_igmp_max_msf) {
2046                 err = -ENOBUFS;
2047                 goto done;
2048         }
2049         if (!psl || psl->sl_count == psl->sl_max) {
2050                 struct ip_sf_socklist *newpsl;
2051                 int count = IP_SFBLOCK;
2052
2053                 if (psl)
2054                         count += psl->sl_max;
2055                 newpsl = sock_kmalloc(sk, IP_SFLSIZE(count), GFP_KERNEL);
2056                 if (!newpsl) {
2057                         err = -ENOBUFS;
2058                         goto done;
2059                 }
2060                 newpsl->sl_max = count;
2061                 newpsl->sl_count = count - IP_SFBLOCK;
2062                 if (psl) {
2063                         for (i=0; i<psl->sl_count; i++)
2064                                 newpsl->sl_addr[i] = psl->sl_addr[i];
2065                         /* decrease mem now to avoid the memleak warning */
2066                         atomic_sub(IP_SFLSIZE(psl->sl_max), &sk->sk_omem_alloc);
2067                         kfree_rcu(psl, rcu);
2068                 }
2069                 rcu_assign_pointer(pmc->sflist, newpsl);
2070                 psl = newpsl;
2071         }
2072         rv = 1; /* > 0 for insert logic below if sl_count is 0 */
2073         for (i=0; i<psl->sl_count; i++) {
2074                 rv = memcmp(&psl->sl_addr[i], &mreqs->imr_sourceaddr,
2075                         sizeof(__be32));
2076                 if (rv == 0)
2077                         break;
2078         }
2079         if (rv == 0)            /* address already there is an error */
2080                 goto done;
2081         for (j=psl->sl_count-1; j>=i; j--)
2082                 psl->sl_addr[j+1] = psl->sl_addr[j];
2083         psl->sl_addr[i] = mreqs->imr_sourceaddr;
2084         psl->sl_count++;
2085         err = 0;
2086         /* update the interface list */
2087         ip_mc_add_src(in_dev, &mreqs->imr_multiaddr, omode, 1,
2088                 &mreqs->imr_sourceaddr, 1);
2089 done:
2090         rtnl_unlock();
2091         if (leavegroup)
2092                 return ip_mc_leave_group(sk, &imr);
2093         return err;
2094 }
2095
2096 int ip_mc_msfilter(struct sock *sk, struct ip_msfilter *msf, int ifindex)
2097 {
2098         int err = 0;
2099         struct ip_mreqn imr;
2100         __be32 addr = msf->imsf_multiaddr;
2101         struct ip_mc_socklist *pmc;
2102         struct in_device *in_dev;
2103         struct inet_sock *inet = inet_sk(sk);
2104         struct ip_sf_socklist *newpsl, *psl;
2105         struct net *net = sock_net(sk);
2106         int leavegroup = 0;
2107
2108         if (!ipv4_is_multicast(addr))
2109                 return -EINVAL;
2110         if (msf->imsf_fmode != MCAST_INCLUDE &&
2111             msf->imsf_fmode != MCAST_EXCLUDE)
2112                 return -EINVAL;
2113
2114         rtnl_lock();
2115
2116         imr.imr_multiaddr.s_addr = msf->imsf_multiaddr;
2117         imr.imr_address.s_addr = msf->imsf_interface;
2118         imr.imr_ifindex = ifindex;
2119         in_dev = ip_mc_find_dev(net, &imr);
2120
2121         if (!in_dev) {
2122                 err = -ENODEV;
2123                 goto done;
2124         }
2125
2126         /* special case - (INCLUDE, empty) == LEAVE_GROUP */
2127         if (msf->imsf_fmode == MCAST_INCLUDE && msf->imsf_numsrc == 0) {
2128                 leavegroup = 1;
2129                 goto done;
2130         }
2131
2132         for_each_pmc_rtnl(inet, pmc) {
2133                 if (pmc->multi.imr_multiaddr.s_addr == msf->imsf_multiaddr &&
2134                     pmc->multi.imr_ifindex == imr.imr_ifindex)
2135                         break;
2136         }
2137         if (!pmc) {             /* must have a prior join */
2138                 err = -EINVAL;
2139                 goto done;
2140         }
2141         if (msf->imsf_numsrc) {
2142                 newpsl = sock_kmalloc(sk, IP_SFLSIZE(msf->imsf_numsrc),
2143                                                            GFP_KERNEL);
2144                 if (!newpsl) {
2145                         err = -ENOBUFS;
2146                         goto done;
2147                 }
2148                 newpsl->sl_max = newpsl->sl_count = msf->imsf_numsrc;
2149                 memcpy(newpsl->sl_addr, msf->imsf_slist,
2150                         msf->imsf_numsrc * sizeof(msf->imsf_slist[0]));
2151                 err = ip_mc_add_src(in_dev, &msf->imsf_multiaddr,
2152                         msf->imsf_fmode, newpsl->sl_count, newpsl->sl_addr, 0);
2153                 if (err) {
2154                         sock_kfree_s(sk, newpsl, IP_SFLSIZE(newpsl->sl_max));
2155                         goto done;
2156                 }
2157         } else {
2158                 newpsl = NULL;
2159                 (void) ip_mc_add_src(in_dev, &msf->imsf_multiaddr,
2160                                      msf->imsf_fmode, 0, NULL, 0);
2161         }
2162         psl = rtnl_dereference(pmc->sflist);
2163         if (psl) {
2164                 (void) ip_mc_del_src(in_dev, &msf->imsf_multiaddr, pmc->sfmode,
2165                         psl->sl_count, psl->sl_addr, 0);
2166                 /* decrease mem now to avoid the memleak warning */
2167                 atomic_sub(IP_SFLSIZE(psl->sl_max), &sk->sk_omem_alloc);
2168                 kfree_rcu(psl, rcu);
2169         } else
2170                 (void) ip_mc_del_src(in_dev, &msf->imsf_multiaddr, pmc->sfmode,
2171                         0, NULL, 0);
2172         rcu_assign_pointer(pmc->sflist, newpsl);
2173         pmc->sfmode = msf->imsf_fmode;
2174         err = 0;
2175 done:
2176         rtnl_unlock();
2177         if (leavegroup)
2178                 err = ip_mc_leave_group(sk, &imr);
2179         return err;
2180 }
2181
2182 int ip_mc_msfget(struct sock *sk, struct ip_msfilter *msf,
2183         struct ip_msfilter __user *optval, int __user *optlen)
2184 {
2185         int err, len, count, copycount;
2186         struct ip_mreqn imr;
2187         __be32 addr = msf->imsf_multiaddr;
2188         struct ip_mc_socklist *pmc;
2189         struct in_device *in_dev;
2190         struct inet_sock *inet = inet_sk(sk);
2191         struct ip_sf_socklist *psl;
2192         struct net *net = sock_net(sk);
2193
2194         if (!ipv4_is_multicast(addr))
2195                 return -EINVAL;
2196
2197         rtnl_lock();
2198
2199         imr.imr_multiaddr.s_addr = msf->imsf_multiaddr;
2200         imr.imr_address.s_addr = msf->imsf_interface;
2201         imr.imr_ifindex = 0;
2202         in_dev = ip_mc_find_dev(net, &imr);
2203
2204         if (!in_dev) {
2205                 err = -ENODEV;
2206                 goto done;
2207         }
2208         err = -EADDRNOTAVAIL;
2209
2210         for_each_pmc_rtnl(inet, pmc) {
2211                 if (pmc->multi.imr_multiaddr.s_addr == msf->imsf_multiaddr &&
2212                     pmc->multi.imr_ifindex == imr.imr_ifindex)
2213                         break;
2214         }
2215         if (!pmc)               /* must have a prior join */
2216                 goto done;
2217         msf->imsf_fmode = pmc->sfmode;
2218         psl = rtnl_dereference(pmc->sflist);
2219         rtnl_unlock();
2220         if (!psl) {
2221                 len = 0;
2222                 count = 0;
2223         } else {
2224                 count = psl->sl_count;
2225         }
2226         copycount = count < msf->imsf_numsrc ? count : msf->imsf_numsrc;
2227         len = copycount * sizeof(psl->sl_addr[0]);
2228         msf->imsf_numsrc = count;
2229         if (put_user(IP_MSFILTER_SIZE(copycount), optlen) ||
2230             copy_to_user(optval, msf, IP_MSFILTER_SIZE(0))) {
2231                 return -EFAULT;
2232         }
2233         if (len &&
2234             copy_to_user(&optval->imsf_slist[0], psl->sl_addr, len))
2235                 return -EFAULT;
2236         return 0;
2237 done:
2238         rtnl_unlock();
2239         return err;
2240 }
2241
2242 int ip_mc_gsfget(struct sock *sk, struct group_filter *gsf,
2243         struct group_filter __user *optval, int __user *optlen)
2244 {
2245         int err, i, count, copycount;
2246         struct sockaddr_in *psin;
2247         __be32 addr;
2248         struct ip_mc_socklist *pmc;
2249         struct inet_sock *inet = inet_sk(sk);
2250         struct ip_sf_socklist *psl;
2251
2252         psin = (struct sockaddr_in *)&gsf->gf_group;
2253         if (psin->sin_family != AF_INET)
2254                 return -EINVAL;
2255         addr = psin->sin_addr.s_addr;
2256         if (!ipv4_is_multicast(addr))
2257                 return -EINVAL;
2258
2259         rtnl_lock();
2260
2261         err = -EADDRNOTAVAIL;
2262
2263         for_each_pmc_rtnl(inet, pmc) {
2264                 if (pmc->multi.imr_multiaddr.s_addr == addr &&
2265                     pmc->multi.imr_ifindex == gsf->gf_interface)
2266                         break;
2267         }
2268         if (!pmc)               /* must have a prior join */
2269                 goto done;
2270         gsf->gf_fmode = pmc->sfmode;
2271         psl = rtnl_dereference(pmc->sflist);
2272         rtnl_unlock();
2273         count = psl ? psl->sl_count : 0;
2274         copycount = count < gsf->gf_numsrc ? count : gsf->gf_numsrc;
2275         gsf->gf_numsrc = count;
2276         if (put_user(GROUP_FILTER_SIZE(copycount), optlen) ||
2277             copy_to_user(optval, gsf, GROUP_FILTER_SIZE(0))) {
2278                 return -EFAULT;
2279         }
2280         for (i=0; i<copycount; i++) {
2281                 struct sockaddr_storage ss;
2282
2283                 psin = (struct sockaddr_in *)&ss;
2284                 memset(&ss, 0, sizeof(ss));
2285                 psin->sin_family = AF_INET;
2286                 psin->sin_addr.s_addr = psl->sl_addr[i];
2287                 if (copy_to_user(&optval->gf_slist[i], &ss, sizeof(ss)))
2288                         return -EFAULT;
2289         }
2290         return 0;
2291 done:
2292         rtnl_unlock();
2293         return err;
2294 }
2295
2296 /*
2297  * check if a multicast source filter allows delivery for a given <src,dst,intf>
2298  */
2299 int ip_mc_sf_allow(struct sock *sk, __be32 loc_addr, __be32 rmt_addr, int dif)
2300 {
2301         struct inet_sock *inet = inet_sk(sk);
2302         struct ip_mc_socklist *pmc;
2303         struct ip_sf_socklist *psl;
2304         int i;
2305         int ret;
2306
2307         ret = 1;
2308         if (!ipv4_is_multicast(loc_addr))
2309                 goto out;
2310
2311         rcu_read_lock();
2312         for_each_pmc_rcu(inet, pmc) {
2313                 if (pmc->multi.imr_multiaddr.s_addr == loc_addr &&
2314                     pmc->multi.imr_ifindex == dif)
2315                         break;
2316         }
2317         ret = inet->mc_all;
2318         if (!pmc)
2319                 goto unlock;
2320         psl = rcu_dereference(pmc->sflist);
2321         ret = (pmc->sfmode == MCAST_EXCLUDE);
2322         if (!psl)
2323                 goto unlock;
2324
2325         for (i=0; i<psl->sl_count; i++) {
2326                 if (psl->sl_addr[i] == rmt_addr)
2327                         break;
2328         }
2329         ret = 0;
2330         if (pmc->sfmode == MCAST_INCLUDE && i >= psl->sl_count)
2331                 goto unlock;
2332         if (pmc->sfmode == MCAST_EXCLUDE && i < psl->sl_count)
2333                 goto unlock;
2334         ret = 1;
2335 unlock:
2336         rcu_read_unlock();
2337 out:
2338         return ret;
2339 }
2340
2341 /*
2342  *      A socket is closing.
2343  */
2344
2345 void ip_mc_drop_socket(struct sock *sk)
2346 {
2347         struct inet_sock *inet = inet_sk(sk);
2348         struct ip_mc_socklist *iml;
2349         struct net *net = sock_net(sk);
2350
2351         if (inet->mc_list == NULL)
2352                 return;
2353
2354         rtnl_lock();
2355         while ((iml = rtnl_dereference(inet->mc_list)) != NULL) {
2356                 struct in_device *in_dev;
2357
2358                 inet->mc_list = iml->next_rcu;
2359                 in_dev = inetdev_by_index(net, iml->multi.imr_ifindex);
2360                 (void) ip_mc_leave_src(sk, iml, in_dev);
2361                 if (in_dev != NULL)
2362                         ip_mc_dec_group(in_dev, iml->multi.imr_multiaddr.s_addr);
2363                 /* decrease mem now to avoid the memleak warning */
2364                 atomic_sub(sizeof(*iml), &sk->sk_omem_alloc);
2365                 kfree_rcu(iml, rcu);
2366         }
2367         rtnl_unlock();
2368 }
2369
2370 /* called with rcu_read_lock() */
2371 int ip_check_mc_rcu(struct in_device *in_dev, __be32 mc_addr, __be32 src_addr, u16 proto)
2372 {
2373         struct ip_mc_list *im;
2374         struct ip_mc_list __rcu **mc_hash;
2375         struct ip_sf_list *psf;
2376         int rv = 0;
2377
2378         mc_hash = rcu_dereference(in_dev->mc_hash);
2379         if (mc_hash) {
2380                 u32 hash = hash_32((__force u32)mc_addr, MC_HASH_SZ_LOG);
2381
2382                 for (im = rcu_dereference(mc_hash[hash]);
2383                      im != NULL;
2384                      im = rcu_dereference(im->next_hash)) {
2385                         if (im->multiaddr == mc_addr)
2386                                 break;
2387                 }
2388         } else {
2389                 for_each_pmc_rcu(in_dev, im) {
2390                         if (im->multiaddr == mc_addr)
2391                                 break;
2392                 }
2393         }
2394         if (im && proto == IPPROTO_IGMP) {
2395                 rv = 1;
2396         } else if (im) {
2397                 if (src_addr) {
2398                         for (psf=im->sources; psf; psf=psf->sf_next) {
2399                                 if (psf->sf_inaddr == src_addr)
2400                                         break;
2401                         }
2402                         if (psf)
2403                                 rv = psf->sf_count[MCAST_INCLUDE] ||
2404                                         psf->sf_count[MCAST_EXCLUDE] !=
2405                                         im->sfcount[MCAST_EXCLUDE];
2406                         else
2407                                 rv = im->sfcount[MCAST_EXCLUDE] != 0;
2408                 } else
2409                         rv = 1; /* unspecified source; tentatively allow */
2410         }
2411         return rv;
2412 }
2413
2414 #if defined(CONFIG_PROC_FS)
2415 struct igmp_mc_iter_state {
2416         struct seq_net_private p;
2417         struct net_device *dev;
2418         struct in_device *in_dev;
2419 };
2420
2421 #define igmp_mc_seq_private(seq)        ((struct igmp_mc_iter_state *)(seq)->private)
2422
2423 static inline struct ip_mc_list *igmp_mc_get_first(struct seq_file *seq)
2424 {
2425         struct net *net = seq_file_net(seq);
2426         struct ip_mc_list *im = NULL;
2427         struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2428
2429         state->in_dev = NULL;
2430         for_each_netdev_rcu(net, state->dev) {
2431                 struct in_device *in_dev;
2432
2433                 in_dev = __in_dev_get_rcu(state->dev);
2434                 if (!in_dev)
2435                         continue;
2436                 im = rcu_dereference(in_dev->mc_list);
2437                 if (im) {
2438                         state->in_dev = in_dev;
2439                         break;
2440                 }
2441         }
2442         return im;
2443 }
2444
2445 static struct ip_mc_list *igmp_mc_get_next(struct seq_file *seq, struct ip_mc_list *im)
2446 {
2447         struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2448
2449         im = rcu_dereference(im->next_rcu);
2450         while (!im) {
2451                 state->dev = next_net_device_rcu(state->dev);
2452                 if (!state->dev) {
2453                         state->in_dev = NULL;
2454                         break;
2455                 }
2456                 state->in_dev = __in_dev_get_rcu(state->dev);
2457                 if (!state->in_dev)
2458                         continue;
2459                 im = rcu_dereference(state->in_dev->mc_list);
2460         }
2461         return im;
2462 }
2463
2464 static struct ip_mc_list *igmp_mc_get_idx(struct seq_file *seq, loff_t pos)
2465 {
2466         struct ip_mc_list *im = igmp_mc_get_first(seq);
2467         if (im)
2468                 while (pos && (im = igmp_mc_get_next(seq, im)) != NULL)
2469                         --pos;
2470         return pos ? NULL : im;
2471 }
2472
2473 static void *igmp_mc_seq_start(struct seq_file *seq, loff_t *pos)
2474         __acquires(rcu)
2475 {
2476         rcu_read_lock();
2477         return *pos ? igmp_mc_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
2478 }
2479
2480 static void *igmp_mc_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2481 {
2482         struct ip_mc_list *im;
2483         if (v == SEQ_START_TOKEN)
2484                 im = igmp_mc_get_first(seq);
2485         else
2486                 im = igmp_mc_get_next(seq, v);
2487         ++*pos;
2488         return im;
2489 }
2490
2491 static void igmp_mc_seq_stop(struct seq_file *seq, void *v)
2492         __releases(rcu)
2493 {
2494         struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2495
2496         state->in_dev = NULL;
2497         state->dev = NULL;
2498         rcu_read_unlock();
2499 }
2500
2501 static int igmp_mc_seq_show(struct seq_file *seq, void *v)
2502 {
2503         if (v == SEQ_START_TOKEN)
2504                 seq_puts(seq,
2505                          "Idx\tDevice    : Count Querier\tGroup    Users Timer\tReporter\n");
2506         else {
2507                 struct ip_mc_list *im = (struct ip_mc_list *)v;
2508                 struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2509                 char   *querier;
2510                 long delta;
2511
2512 #ifdef CONFIG_IP_MULTICAST
2513                 querier = IGMP_V1_SEEN(state->in_dev) ? "V1" :
2514                           IGMP_V2_SEEN(state->in_dev) ? "V2" :
2515                           "V3";
2516 #else
2517                 querier = "NONE";
2518 #endif
2519
2520                 if (rcu_dereference(state->in_dev->mc_list) == im) {
2521                         seq_printf(seq, "%d\t%-10s: %5d %7s\n",
2522                                    state->dev->ifindex, state->dev->name, state->in_dev->mc_count, querier);
2523                 }
2524
2525                 delta = im->timer.expires - jiffies;
2526                 seq_printf(seq,
2527                            "\t\t\t\t%08X %5d %d:%08lX\t\t%d\n",
2528                            im->multiaddr, im->users,
2529                            im->tm_running,
2530                            im->tm_running ? jiffies_delta_to_clock_t(delta) : 0,
2531                            im->reporter);
2532         }
2533         return 0;
2534 }
2535
2536 static const struct seq_operations igmp_mc_seq_ops = {
2537         .start  =       igmp_mc_seq_start,
2538         .next   =       igmp_mc_seq_next,
2539         .stop   =       igmp_mc_seq_stop,
2540         .show   =       igmp_mc_seq_show,
2541 };
2542
2543 static int igmp_mc_seq_open(struct inode *inode, struct file *file)
2544 {
2545         return seq_open_net(inode, file, &igmp_mc_seq_ops,
2546                         sizeof(struct igmp_mc_iter_state));
2547 }
2548
2549 static const struct file_operations igmp_mc_seq_fops = {
2550         .owner          =       THIS_MODULE,
2551         .open           =       igmp_mc_seq_open,
2552         .read           =       seq_read,
2553         .llseek         =       seq_lseek,
2554         .release        =       seq_release_net,
2555 };
2556
2557 struct igmp_mcf_iter_state {
2558         struct seq_net_private p;
2559         struct net_device *dev;
2560         struct in_device *idev;
2561         struct ip_mc_list *im;
2562 };
2563
2564 #define igmp_mcf_seq_private(seq)       ((struct igmp_mcf_iter_state *)(seq)->private)
2565
2566 static inline struct ip_sf_list *igmp_mcf_get_first(struct seq_file *seq)
2567 {
2568         struct net *net = seq_file_net(seq);
2569         struct ip_sf_list *psf = NULL;
2570         struct ip_mc_list *im = NULL;
2571         struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2572
2573         state->idev = NULL;
2574         state->im = NULL;
2575         for_each_netdev_rcu(net, state->dev) {
2576                 struct in_device *idev;
2577                 idev = __in_dev_get_rcu(state->dev);
2578                 if (unlikely(idev == NULL))
2579                         continue;
2580                 im = rcu_dereference(idev->mc_list);
2581                 if (likely(im != NULL)) {
2582                         spin_lock_bh(&im->lock);
2583                         psf = im->sources;
2584                         if (likely(psf != NULL)) {
2585                                 state->im = im;
2586                                 state->idev = idev;
2587                                 break;
2588                         }
2589                         spin_unlock_bh(&im->lock);
2590                 }
2591         }
2592         return psf;
2593 }
2594
2595 static struct ip_sf_list *igmp_mcf_get_next(struct seq_file *seq, struct ip_sf_list *psf)
2596 {
2597         struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2598
2599         psf = psf->sf_next;
2600         while (!psf) {
2601                 spin_unlock_bh(&state->im->lock);
2602                 state->im = state->im->next;
2603                 while (!state->im) {
2604                         state->dev = next_net_device_rcu(state->dev);
2605                         if (!state->dev) {
2606                                 state->idev = NULL;
2607                                 goto out;
2608                         }
2609                         state->idev = __in_dev_get_rcu(state->dev);
2610                         if (!state->idev)
2611                                 continue;
2612                         state->im = rcu_dereference(state->idev->mc_list);
2613                 }
2614                 if (!state->im)
2615                         break;
2616                 spin_lock_bh(&state->im->lock);
2617                 psf = state->im->sources;
2618         }
2619 out:
2620         return psf;
2621 }
2622
2623 static struct ip_sf_list *igmp_mcf_get_idx(struct seq_file *seq, loff_t pos)
2624 {
2625         struct ip_sf_list *psf = igmp_mcf_get_first(seq);
2626         if (psf)
2627                 while (pos && (psf = igmp_mcf_get_next(seq, psf)) != NULL)
2628                         --pos;
2629         return pos ? NULL : psf;
2630 }
2631
2632 static void *igmp_mcf_seq_start(struct seq_file *seq, loff_t *pos)
2633         __acquires(rcu)
2634 {
2635         rcu_read_lock();
2636         return *pos ? igmp_mcf_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
2637 }
2638
2639 static void *igmp_mcf_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2640 {
2641         struct ip_sf_list *psf;
2642         if (v == SEQ_START_TOKEN)
2643                 psf = igmp_mcf_get_first(seq);
2644         else
2645                 psf = igmp_mcf_get_next(seq, v);
2646         ++*pos;
2647         return psf;
2648 }
2649
2650 static void igmp_mcf_seq_stop(struct seq_file *seq, void *v)
2651         __releases(rcu)
2652 {
2653         struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2654         if (likely(state->im != NULL)) {
2655                 spin_unlock_bh(&state->im->lock);
2656                 state->im = NULL;
2657         }
2658         state->idev = NULL;
2659         state->dev = NULL;
2660         rcu_read_unlock();
2661 }
2662
2663 static int igmp_mcf_seq_show(struct seq_file *seq, void *v)
2664 {
2665         struct ip_sf_list *psf = (struct ip_sf_list *)v;
2666         struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2667
2668         if (v == SEQ_START_TOKEN) {
2669                 seq_printf(seq,
2670                            "%3s %6s "
2671                            "%10s %10s %6s %6s\n", "Idx",
2672                            "Device", "MCA",
2673                            "SRC", "INC", "EXC");
2674         } else {
2675                 seq_printf(seq,
2676                            "%3d %6.6s 0x%08x "
2677                            "0x%08x %6lu %6lu\n",
2678                            state->dev->ifindex, state->dev->name,
2679                            ntohl(state->im->multiaddr),
2680                            ntohl(psf->sf_inaddr),
2681                            psf->sf_count[MCAST_INCLUDE],
2682                            psf->sf_count[MCAST_EXCLUDE]);
2683         }
2684         return 0;
2685 }
2686
2687 static const struct seq_operations igmp_mcf_seq_ops = {
2688         .start  =       igmp_mcf_seq_start,
2689         .next   =       igmp_mcf_seq_next,
2690         .stop   =       igmp_mcf_seq_stop,
2691         .show   =       igmp_mcf_seq_show,
2692 };
2693
2694 static int igmp_mcf_seq_open(struct inode *inode, struct file *file)
2695 {
2696         return seq_open_net(inode, file, &igmp_mcf_seq_ops,
2697                         sizeof(struct igmp_mcf_iter_state));
2698 }
2699
2700 static const struct file_operations igmp_mcf_seq_fops = {
2701         .owner          =       THIS_MODULE,
2702         .open           =       igmp_mcf_seq_open,
2703         .read           =       seq_read,
2704         .llseek         =       seq_lseek,
2705         .release        =       seq_release_net,
2706 };
2707
2708 static int __net_init igmp_net_init(struct net *net)
2709 {
2710         struct proc_dir_entry *pde;
2711
2712         pde = proc_create("igmp", S_IRUGO, net->proc_net, &igmp_mc_seq_fops);
2713         if (!pde)
2714                 goto out_igmp;
2715         pde = proc_create("mcfilter", S_IRUGO, net->proc_net,
2716                           &igmp_mcf_seq_fops);
2717         if (!pde)
2718                 goto out_mcfilter;
2719         return 0;
2720
2721 out_mcfilter:
2722         remove_proc_entry("igmp", net->proc_net);
2723 out_igmp:
2724         return -ENOMEM;
2725 }
2726
2727 static void __net_exit igmp_net_exit(struct net *net)
2728 {
2729         remove_proc_entry("mcfilter", net->proc_net);
2730         remove_proc_entry("igmp", net->proc_net);
2731 }
2732
2733 static struct pernet_operations igmp_net_ops = {
2734         .init = igmp_net_init,
2735         .exit = igmp_net_exit,
2736 };
2737
2738 int __init igmp_mc_proc_init(void)
2739 {
2740         return register_pernet_subsys(&igmp_net_ops);
2741 }
2742 #endif