net/ipv4/ipmr.c

   1 /*
   2  *      IP multicast routing support for mrouted 3.6/3.8
   3  *
   4  *              (c) 1995 Alan Cox, <alan@lxorguk.ukuu.org.uk>
   5  *        Linux Consultancy and Custom Driver Development
   6  *
   7  *      This program is free software; you can redistribute it and/or
   8  *      modify it under the terms of the GNU General Public License
   9  *      as published by the Free Software Foundation; either version
  10  *      2 of the License, or (at your option) any later version.
  11  *
  12  *      Fixes:
  13  *      Michael Chastain        :       Incorrect size of copying.
  14  *      Alan Cox                :       Added the cache manager code
  15  *      Alan Cox                :       Fixed the clone/copy bug and device race.
  16  *      Mike McLagan            :       Routing by source
  17  *      Malcolm Beattie         :       Buffer handling fixes.
  18  *      Alexey Kuznetsov        :       Double buffer free and other fixes.
  19  *      SVR Anand               :       Fixed several multicast bugs and problems.
  20  *      Alexey Kuznetsov        :       Status, optimisations and more.
  21  *      Brad Parker             :       Better behaviour on mrouted upcall
  22  *                                      overflow.
  23  *      Carlos Picoto           :       PIMv1 Support
  24  *      Pavlin Ivanov Radoslavov:       PIMv2 Registers must checksum only PIM header
  25  *                                      Relax this requirement to work with older peers.
  26  *
  27  */
  28
  29 #include <asm/uaccess.h>
  30 #include <linux/types.h>
  31 #include <linux/capability.h>
  32 #include <linux/errno.h>
  33 #include <linux/timer.h>
  34 #include <linux/mm.h>
  35 #include <linux/kernel.h>
  36 #include <linux/fcntl.h>
  37 #include <linux/stat.h>
  38 #include <linux/socket.h>
  39 #include <linux/in.h>
  40 #include <linux/inet.h>
  41 #include <linux/netdevice.h>
  42 #include <linux/inetdevice.h>
  43 #include <linux/igmp.h>
  44 #include <linux/proc_fs.h>
  45 #include <linux/seq_file.h>
  46 #include <linux/mroute.h>
  47 #include <linux/init.h>
  48 #include <linux/if_ether.h>
  49 #include <linux/slab.h>
  50 #include <net/net_namespace.h>
  51 #include <net/ip.h>
  52 #include <net/protocol.h>
  53 #include <linux/skbuff.h>
  54 #include <net/route.h>
  55 #include <net/sock.h>
  56 #include <net/icmp.h>
  57 #include <net/udp.h>
  58 #include <net/raw.h>
  59 #include <linux/notifier.h>
  60 #include <linux/if_arp.h>
  61 #include <linux/netfilter_ipv4.h>
  62 #include <linux/compat.h>
  63 #include <linux/export.h>
  64 #include <net/ip_tunnels.h>
  65 #include <net/checksum.h>
  66 #include <net/netlink.h>
  67 #include <net/fib_rules.h>
  68 #include <linux/netconf.h>
  69
  70 #if defined(CONFIG_IP_PIMSM_V1) || defined(CONFIG_IP_PIMSM_V2)
  71 #define CONFIG_IP_PIMSM 1
  72 #endif
  73
  74 struct mr_table {
  75         struct list_head        list;
  76 #ifdef CONFIG_NET_NS
  77         struct net              *net;
  78 #endif
  79         u32                     id;
  80         struct sock __rcu       *mroute_sk;
  81         struct timer_list       ipmr_expire_timer;
  82         struct list_head        mfc_unres_queue;
  83         struct list_head        mfc_cache_array[MFC_LINES];
  84         struct vif_device       vif_table[MAXVIFS];
  85         int                     maxvif;
  86         atomic_t                cache_resolve_queue_len;
  87         bool                    mroute_do_assert;
  88         bool                    mroute_do_pim;
  89 #if defined(CONFIG_IP_PIMSM_V1) || defined(CONFIG_IP_PIMSM_V2)
  90         int                     mroute_reg_vif_num;
  91 #endif
  92 };
  93
  94 struct ipmr_rule {
  95         struct fib_rule         common;
  96 };
  97
  98 struct ipmr_result {
  99         struct mr_table         *mrt;
 100 };
 101
 102 /* Big lock, protecting vif table, mrt cache and mroute socket state.
 103  * Note that the changes are semaphored via rtnl_lock.
 104  */
 105
 106 static DEFINE_RWLOCK(mrt_lock);
 107
 108 /*
 109  *      Multicast router control variables
 110  */
 111
 112 #define VIF_EXISTS(_mrt, _idx) ((_mrt)->vif_table[_idx].dev != NULL)
 113
 114 /* Special spinlock for queue of unresolved entries */
 115 static DEFINE_SPINLOCK(mfc_unres_lock);
 116
 117 /* We return to original Alan's scheme. Hash table of resolved
 118  * entries is changed only in process context and protected
 119  * with weak lock mrt_lock. Queue of unresolved entries is protected
 120  * with strong spinlock mfc_unres_lock.
 121  *
 122  * In this case data path is free of exclusive locks at all.
 123  */
 124
 125 static struct kmem_cache *mrt_cachep __read_mostly;
 126
 127 static struct mr_table *ipmr_new_table(struct net *net, u32 id);
 128 static void ipmr_free_table(struct mr_table *mrt);
 129
 130 static void ip_mr_forward(struct net *net, struct mr_table *mrt,
 131                           struct sk_buff *skb, struct mfc_cache *cache,
 132                           int local);
 133 static int ipmr_cache_report(struct mr_table *mrt,
 134                              struct sk_buff *pkt, vifi_t vifi, int assert);
 135 static int __ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
 136                               struct mfc_cache *c, struct rtmsg *rtm);
 137 static void mroute_netlink_event(struct mr_table *mrt, struct mfc_cache *mfc,
 138                                  int cmd);
 139 static void mroute_clean_tables(struct mr_table *mrt);
 140 static void ipmr_expire_process(unsigned long arg);
 141
 142 #ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES
 143 #define ipmr_for_each_table(mrt, net) \
 144         list_for_each_entry_rcu(mrt, &net->ipv4.mr_tables, list)
 145
 146 static struct mr_table *ipmr_get_table(struct net *net, u32 id)
 147 {
 148         struct mr_table *mrt;
 149
 150         ipmr_for_each_table(mrt, net) {
 151                 if (mrt->id == id)
 152                         return mrt;
 153         }
 154         return NULL;
 155 }
 156
 157 static int ipmr_fib_lookup(struct net *net, struct flowi4 *flp4,
 158                            struct mr_table **mrt)
 159 {
 160         struct ipmr_result res;
 161         struct fib_lookup_arg arg = { .result = &res, };
 162         int err;
 163
 164         err = fib_rules_lookup(net->ipv4.mr_rules_ops,
 165                                flowi4_to_flowi(flp4), 0, &arg);
 166         if (err < 0)
 167                 return err;
 168         *mrt = res.mrt;
 169         return 0;
 170 }
 171
 172 static int ipmr_rule_action(struct fib_rule *rule, struct flowi *flp,
 173                             int flags, struct fib_lookup_arg *arg)
 174 {
 175         struct ipmr_result *res = arg->result;
 176         struct mr_table *mrt;
 177
 178         switch (rule->action) {
 179         case FR_ACT_TO_TBL:
 180                 break;
 181         case FR_ACT_UNREACHABLE:
 182                 return -ENETUNREACH;
 183         case FR_ACT_PROHIBIT:
 184                 return -EACCES;
 185         case FR_ACT_BLACKHOLE:
 186         default:
 187                 return -EINVAL;
 188         }
 189
 190         mrt = ipmr_get_table(rule->fr_net, rule->table);
 191         if (mrt == NULL)
 192                 return -EAGAIN;
 193         res->mrt = mrt;
 194         return 0;
 195 }
 196
 197 static int ipmr_rule_match(struct fib_rule *rule, struct flowi *fl, int flags)
 198 {
 199         return 1;
 200 }
 201
 202 static const struct nla_policy ipmr_rule_policy[FRA_MAX + 1] = {
 203         FRA_GENERIC_POLICY,
 204 };
 205
 206 static int ipmr_rule_configure(struct fib_rule *rule, struct sk_buff *skb,
 207                                struct fib_rule_hdr *frh, struct nlattr **tb)
 208 {
 209         return 0;
 210 }
 211
 212 static int ipmr_rule_compare(struct fib_rule *rule, struct fib_rule_hdr *frh,
 213                              struct nlattr **tb)
 214 {
 215         return 1;
 216 }
 217
 218 static int ipmr_rule_fill(struct fib_rule *rule, struct sk_buff *skb,
 219                           struct fib_rule_hdr *frh)
 220 {
 221         frh->dst_len = 0;
 222         frh->src_len = 0;
 223         frh->tos     = 0;
 224         return 0;
 225 }
 226
 227 static const struct fib_rules_ops __net_initconst ipmr_rules_ops_template = {
 228         .family         = RTNL_FAMILY_IPMR,
 229         .rule_size      = sizeof(struct ipmr_rule),
 230         .addr_size      = sizeof(u32),
 231         .action         = ipmr_rule_action,
 232         .match          = ipmr_rule_match,
 233         .configure      = ipmr_rule_configure,
 234         .compare        = ipmr_rule_compare,
 235         .default_pref   = fib_default_rule_pref,
 236         .fill           = ipmr_rule_fill,
 237         .nlgroup        = RTNLGRP_IPV4_RULE,
 238         .policy         = ipmr_rule_policy,
 239         .owner          = THIS_MODULE,
 240 };
 241
 242 static int __net_init ipmr_rules_init(struct net *net)
 243 {
 244         struct fib_rules_ops *ops;
 245         struct mr_table *mrt;
 246         int err;
 247
 248         ops = fib_rules_register(&ipmr_rules_ops_template, net);
 249         if (IS_ERR(ops))
 250                 return PTR_ERR(ops);
 251
 252         INIT_LIST_HEAD(&net->ipv4.mr_tables);
 253
 254         mrt = ipmr_new_table(net, RT_TABLE_DEFAULT);
 255         if (mrt == NULL) {
 256                 err = -ENOMEM;
 257                 goto err1;
 258         }
 259
 260         err = fib_default_rule_add(ops, 0x7fff, RT_TABLE_DEFAULT, 0);
 261         if (err < 0)
 262                 goto err2;
 263
 264         net->ipv4.mr_rules_ops = ops;
 265         return 0;
 266
 267 err2:
 268         kfree(mrt);
 269 err1:
 270         fib_rules_unregister(ops);
 271         return err;
 272 }
 273
 274 static void __net_exit ipmr_rules_exit(struct net *net)
 275 {
 276         struct mr_table *mrt, *next;
 277
 278         list_for_each_entry_safe(mrt, next, &net->ipv4.mr_tables, list) {
 279                 list_del(&mrt->list);
 280                 ipmr_free_table(mrt);
 281         }
 282         fib_rules_unregister(net->ipv4.mr_rules_ops);
 283 }
 284 #else
 285 #define ipmr_for_each_table(mrt, net) \
 286         for (mrt = net->ipv4.mrt; mrt; mrt = NULL)
 287
 288 static struct mr_table *ipmr_get_table(struct net *net, u32 id)
 289 {
 290         return net->ipv4.mrt;
 291 }
 292
 293 static int ipmr_fib_lookup(struct net *net, struct flowi4 *flp4,
 294                            struct mr_table **mrt)
 295 {
 296         *mrt = net->ipv4.mrt;
 297         return 0;
 298 }
 299
 300 static int __net_init ipmr_rules_init(struct net *net)
 301 {
 302         net->ipv4.mrt = ipmr_new_table(net, RT_TABLE_DEFAULT);
 303         return net->ipv4.mrt ? 0 : -ENOMEM;
 304 }
 305
 306 static void __net_exit ipmr_rules_exit(struct net *net)
 307 {
 308         ipmr_free_table(net->ipv4.mrt);
 309 }
 310 #endif
 311
 312 static struct mr_table *ipmr_new_table(struct net *net, u32 id)
 313 {
 314         struct mr_table *mrt;
 315         unsigned int i;
 316
 317         mrt = ipmr_get_table(net, id);
 318         if (mrt != NULL)
 319                 return mrt;
 320
 321         mrt = kzalloc(sizeof(*mrt), GFP_KERNEL);
 322         if (mrt == NULL)
 323                 return NULL;
 324         write_pnet(&mrt->net, net);
 325         mrt->id = id;
 326
 327         /* Forwarding cache */
 328         for (i = 0; i < MFC_LINES; i++)
 329                 INIT_LIST_HEAD(&mrt->mfc_cache_array[i]);
 330
 331         INIT_LIST_HEAD(&mrt->mfc_unres_queue);
 332
 333         setup_timer(&mrt->ipmr_expire_timer, ipmr_expire_process,
 334                     (unsigned long)mrt);
 335
 336 #ifdef CONFIG_IP_PIMSM
 337         mrt->mroute_reg_vif_num = -1;
 338 #endif
 339 #ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES
 340         list_add_tail_rcu(&mrt->list, &net->ipv4.mr_tables);
 341 #endif
 342         return mrt;
 343 }
 344
 345 static void ipmr_free_table(struct mr_table *mrt)
 346 {
 347         del_timer_sync(&mrt->ipmr_expire_timer);
 348         mroute_clean_tables(mrt);
 349         kfree(mrt);
 350 }
 351
 352 /* Service routines creating virtual interfaces: DVMRP tunnels and PIMREG */
 353
 354 static void ipmr_del_tunnel(struct net_device *dev, struct vifctl *v)
 355 {
 356         struct net *net = dev_net(dev);
 357
 358         dev_close(dev);
 359
 360         dev = __dev_get_by_name(net, "tunl0");
 361         if (dev) {
 362                 const struct net_device_ops *ops = dev->netdev_ops;
 363                 struct ifreq ifr;
 364                 struct ip_tunnel_parm p;
 365
 366                 memset(&p, 0, sizeof(p));
 367                 p.iph.daddr = v->vifc_rmt_addr.s_addr;
 368                 p.iph.saddr = v->vifc_lcl_addr.s_addr;
 369                 p.iph.version = 4;
 370                 p.iph.ihl = 5;
 371                 p.iph.protocol = IPPROTO_IPIP;
 372                 sprintf(p.name, "dvmrp%d", v->vifc_vifi);
 373                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
 374
 375                 if (ops->ndo_do_ioctl) {
 376                         mm_segment_t oldfs = get_fs();
 377
 378                         set_fs(KERNEL_DS);
 379                         ops->ndo_do_ioctl(dev, &ifr, SIOCDELTUNNEL);
 380                         set_fs(oldfs);
 381                 }
 382         }
 383 }
 384
 385 static
 386 struct net_device *ipmr_new_tunnel(struct net *net, struct vifctl *v)
 387 {
 388         struct net_device  *dev;
 389
 390         dev = __dev_get_by_name(net, "tunl0");
 391
 392         if (dev) {
 393                 const struct net_device_ops *ops = dev->netdev_ops;
 394                 int err;
 395                 struct ifreq ifr;
 396                 struct ip_tunnel_parm p;
 397                 struct in_device  *in_dev;
 398
 399                 memset(&p, 0, sizeof(p));
 400                 p.iph.daddr = v->vifc_rmt_addr.s_addr;
 401                 p.iph.saddr = v->vifc_lcl_addr.s_addr;
 402                 p.iph.version = 4;
 403                 p.iph.ihl = 5;
 404                 p.iph.protocol = IPPROTO_IPIP;
 405                 sprintf(p.name, "dvmrp%d", v->vifc_vifi);
 406                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
 407
 408                 if (ops->ndo_do_ioctl) {
 409                         mm_segment_t oldfs = get_fs();
 410
 411                         set_fs(KERNEL_DS);
 412                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
 413                         set_fs(oldfs);
 414                 } else {
 415                         err = -EOPNOTSUPP;
 416                 }
 417                 dev = NULL;
 418
 419                 if (err == 0 &&
 420                     (dev = __dev_get_by_name(net, p.name)) != NULL) {
 421                         dev->flags |= IFF_MULTICAST;
 422
 423                         in_dev = __in_dev_get_rtnl(dev);
 424                         if (in_dev == NULL)
 425                                 goto failure;
 426
 427                         ipv4_devconf_setall(in_dev);
 428                         IPV4_DEVCONF(in_dev->cnf, RP_FILTER) = 0;
 429
 430                         if (dev_open(dev))
 431                                 goto failure;
 432                         dev_hold(dev);
 433                 }
 434         }
 435         return dev;
 436
 437 failure:
 438         /* allow the register to be completed before unregistering. */
 439         rtnl_unlock();
 440         rtnl_lock();
 441
 442         unregister_netdevice(dev);
 443         return NULL;
 444 }
 445
 446 #ifdef CONFIG_IP_PIMSM
 447
 448 static netdev_tx_t reg_vif_xmit(struct sk_buff *skb, struct net_device *dev)
 449 {
 450         struct net *net = dev_net(dev);
 451         struct mr_table *mrt;
 452         struct flowi4 fl4 = {
 453                 .flowi4_oif     = dev->ifindex,
 454                 .flowi4_iif     = skb->skb_iif,
 455                 .flowi4_mark    = skb->mark,
 456         };
 457         int err;
 458
 459         err = ipmr_fib_lookup(net, &fl4, &mrt);
 460         if (err < 0) {
 461                 kfree_skb(skb);
 462                 return err;
 463         }
 464
 465         read_lock(&mrt_lock);
 466         dev->stats.tx_bytes += skb->len;
 467         dev->stats.tx_packets++;
 468         ipmr_cache_report(mrt, skb, mrt->mroute_reg_vif_num, IGMPMSG_WHOLEPKT);
 469         read_unlock(&mrt_lock);
 470         kfree_skb(skb);
 471         return NETDEV_TX_OK;
 472 }
 473
 474 static const struct net_device_ops reg_vif_netdev_ops = {
 475         .ndo_start_xmit = reg_vif_xmit,
 476 };
 477
 478 static void reg_vif_setup(struct net_device *dev)
 479 {
 480         dev->type               = ARPHRD_PIMREG;
 481         dev->mtu                = ETH_DATA_LEN - sizeof(struct iphdr) - 8;
 482         dev->flags              = IFF_NOARP;
 483         dev->netdev_ops         = &reg_vif_netdev_ops,
 484         dev->destructor         = free_netdev;
 485         dev->features           |= NETIF_F_NETNS_LOCAL;
 486 }
 487
 488 static struct net_device *ipmr_reg_vif(struct net *net, struct mr_table *mrt)
 489 {
 490         struct net_device *dev;
 491         struct in_device *in_dev;
 492         char name[IFNAMSIZ];
 493
 494         if (mrt->id == RT_TABLE_DEFAULT)
 495                 sprintf(name, "pimreg");
 496         else
 497                 sprintf(name, "pimreg%u", mrt->id);
 498
 499         dev = alloc_netdev(0, name, reg_vif_setup);
 500
 501         if (dev == NULL)
 502                 return NULL;
 503
 504         dev_net_set(dev, net);
 505
 506         if (register_netdevice(dev)) {
 507                 free_netdev(dev);
 508                 return NULL;
 509         }
 510         dev->iflink = 0;
 511
 512         rcu_read_lock();
 513         in_dev = __in_dev_get_rcu(dev);
 514         if (!in_dev) {
 515                 rcu_read_unlock();
 516                 goto failure;
 517         }
 518
 519         ipv4_devconf_setall(in_dev);
 520         IPV4_DEVCONF(in_dev->cnf, RP_FILTER) = 0;
 521         rcu_read_unlock();
 522
 523         if (dev_open(dev))
 524                 goto failure;
 525
 526         dev_hold(dev);
 527
 528         return dev;
 529
 530 failure:
 531         /* allow the register to be completed before unregistering. */
 532         rtnl_unlock();
 533         rtnl_lock();
 534
 535         unregister_netdevice(dev);
 536         return NULL;
 537 }
 538 #endif
 539
 540 /**
 541  *      vif_delete - Delete a VIF entry
 542  *      @notify: Set to 1, if the caller is a notifier_call
 543  */
 544
 545 static int vif_delete(struct mr_table *mrt, int vifi, int notify,
 546                       struct list_head *head)
 547 {
 548         struct vif_device *v;
 549         struct net_device *dev;
 550         struct in_device *in_dev;
 551
 552         if (vifi < 0 || vifi >= mrt->maxvif)
 553                 return -EADDRNOTAVAIL;
 554
 555         v = &mrt->vif_table[vifi];
 556
 557         write_lock_bh(&mrt_lock);
 558         dev = v->dev;
 559         v->dev = NULL;
 560
 561         if (!dev) {
 562                 write_unlock_bh(&mrt_lock);
 563                 return -EADDRNOTAVAIL;
 564         }
 565
 566 #ifdef CONFIG_IP_PIMSM
 567         if (vifi == mrt->mroute_reg_vif_num)
 568                 mrt->mroute_reg_vif_num = -1;
 569 #endif
 570
 571         if (vifi + 1 == mrt->maxvif) {
 572                 int tmp;
 573
 574                 for (tmp = vifi - 1; tmp >= 0; tmp--) {
 575                         if (VIF_EXISTS(mrt, tmp))
 576                                 break;
 577                 }
 578                 mrt->maxvif = tmp+1;
 579         }
 580
 581         write_unlock_bh(&mrt_lock);
 582
 583         dev_set_allmulti(dev, -1);
 584
 585         in_dev = __in_dev_get_rtnl(dev);
 586         if (in_dev) {
 587                 IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)--;
 588                 inet_netconf_notify_devconf(dev_net(dev),
 589                                             NETCONFA_MC_FORWARDING,
 590                                             dev->ifindex, &in_dev->cnf);
 591                 ip_rt_multicast_event(in_dev);
 592         }
 593
 594         if (v->flags & (VIFF_TUNNEL | VIFF_REGISTER) && !notify)
 595                 unregister_netdevice_queue(dev, head);
 596
 597         dev_put(dev);
 598         return 0;
 599 }
 600
 601 static void ipmr_cache_free_rcu(struct rcu_head *head)
 602 {
 603         struct mfc_cache *c = container_of(head, struct mfc_cache, rcu);
 604
 605         kmem_cache_free(mrt_cachep, c);
 606 }
 607
 608 static inline void ipmr_cache_free(struct mfc_cache *c)
 609 {
 610         call_rcu(&c->rcu, ipmr_cache_free_rcu);
 611 }
 612
 613 /* Destroy an unresolved cache entry, killing queued skbs
 614  * and reporting error to netlink readers.
 615  */
 616
 617 static void ipmr_destroy_unres(struct mr_table *mrt, struct mfc_cache *c)
 618 {
 619         struct net *net = read_pnet(&mrt->net);
 620         struct sk_buff *skb;
 621         struct nlmsgerr *e;
 622
 623         atomic_dec(&mrt->cache_resolve_queue_len);
 624
 625         while ((skb = skb_dequeue(&c->mfc_un.unres.unresolved))) {
 626                 if (ip_hdr(skb)->version == 0) {
 627                         struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct iphdr));
 628                         nlh->nlmsg_type = NLMSG_ERROR;
 629                         nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
 630                         skb_trim(skb, nlh->nlmsg_len);
 631                         e = nlmsg_data(nlh);
 632                         e->error = -ETIMEDOUT;
 633                         memset(&e->msg, 0, sizeof(e->msg));
 634
 635                         rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
 636                 } else {
 637                         kfree_skb(skb);
 638                 }
 639         }
 640
 641         ipmr_cache_free(c);
 642 }
 643
 644
 645 /* Timer process for the unresolved queue. */
 646
 647 static void ipmr_expire_process(unsigned long arg)
 648 {
 649         struct mr_table *mrt = (struct mr_table *)arg;
 650         unsigned long now;
 651         unsigned long expires;
 652         struct mfc_cache *c, *next;
 653
 654         if (!spin_trylock(&mfc_unres_lock)) {
 655                 mod_timer(&mrt->ipmr_expire_timer, jiffies+HZ/10);
 656                 return;
 657         }
 658
 659         if (list_empty(&mrt->mfc_unres_queue))
 660                 goto out;
 661
 662         now = jiffies;
 663         expires = 10*HZ;
 664
 665         list_for_each_entry_safe(c, next, &mrt->mfc_unres_queue, list) {
 666                 if (time_after(c->mfc_un.unres.expires, now)) {
 667                         unsigned long interval = c->mfc_un.unres.expires - now;
 668                         if (interval < expires)
 669                                 expires = interval;
 670                         continue;
 671                 }
 672
 673                 list_del(&c->list);
 674                 mroute_netlink_event(mrt, c, RTM_DELROUTE);
 675                 ipmr_destroy_unres(mrt, c);
 676         }
 677
 678         if (!list_empty(&mrt->mfc_unres_queue))
 679                 mod_timer(&mrt->ipmr_expire_timer, jiffies + expires);
 680
 681 out:
 682         spin_unlock(&mfc_unres_lock);
 683 }
 684
 685 /* Fill oifs list. It is called under write locked mrt_lock. */
 686
 687 static void ipmr_update_thresholds(struct mr_table *mrt, struct mfc_cache *cache,
 688                                    unsigned char *ttls)
 689 {
 690         int vifi;
 691
 692         cache->mfc_un.res.minvif = MAXVIFS;
 693         cache->mfc_un.res.maxvif = 0;
 694         memset(cache->mfc_un.res.ttls, 255, MAXVIFS);
 695
 696         for (vifi = 0; vifi < mrt->maxvif; vifi++) {
 697                 if (VIF_EXISTS(mrt, vifi) &&
 698                     ttls[vifi] && ttls[vifi] < 255) {
 699                         cache->mfc_un.res.ttls[vifi] = ttls[vifi];
 700                         if (cache->mfc_un.res.minvif > vifi)
 701                                 cache->mfc_un.res.minvif = vifi;
 702                         if (cache->mfc_un.res.maxvif <= vifi)
 703                                 cache->mfc_un.res.maxvif = vifi + 1;
 704                 }
 705         }
 706 }
 707
 708 static int vif_add(struct net *net, struct mr_table *mrt,
 709                    struct vifctl *vifc, int mrtsock)
 710 {
 711         int vifi = vifc->vifc_vifi;
 712         struct vif_device *v = &mrt->vif_table[vifi];
 713         struct net_device *dev;
 714         struct in_device *in_dev;
 715         int err;
 716
 717         /* Is vif busy ? */
 718         if (VIF_EXISTS(mrt, vifi))
 719                 return -EADDRINUSE;
 720
 721         switch (vifc->vifc_flags) {
 722 #ifdef CONFIG_IP_PIMSM
 723         case VIFF_REGISTER:
 724                 /*
 725                  * Special Purpose VIF in PIM
 726                  * All the packets will be sent to the daemon
 727                  */
 728                 if (mrt->mroute_reg_vif_num >= 0)
 729                         return -EADDRINUSE;
 730                 dev = ipmr_reg_vif(net, mrt);
 731                 if (!dev)
 732                         return -ENOBUFS;
 733                 err = dev_set_allmulti(dev, 1);
 734                 if (err) {
 735                         unregister_netdevice(dev);
 736                         dev_put(dev);
 737                         return err;
 738                 }
 739                 break;
 740 #endif
 741         case VIFF_TUNNEL:
 742                 dev = ipmr_new_tunnel(net, vifc);
 743                 if (!dev)
 744                         return -ENOBUFS;
 745                 err = dev_set_allmulti(dev, 1);
 746                 if (err) {
 747                         ipmr_del_tunnel(dev, vifc);
 748                         dev_put(dev);
 749                         return err;
 750                 }
 751                 break;
 752
 753         case VIFF_USE_IFINDEX:
 754         case 0:
 755                 if (vifc->vifc_flags == VIFF_USE_IFINDEX) {
 756                         dev = dev_get_by_index(net, vifc->vifc_lcl_ifindex);
 757                         if (dev && __in_dev_get_rtnl(dev) == NULL) {
 758                                 dev_put(dev);
 759                                 return -EADDRNOTAVAIL;
 760                         }
 761                 } else {
 762                         dev = ip_dev_find(net, vifc->vifc_lcl_addr.s_addr);
 763                 }
 764                 if (!dev)
 765                         return -EADDRNOTAVAIL;
 766                 err = dev_set_allmulti(dev, 1);
 767                 if (err) {
 768                         dev_put(dev);
 769                         return err;
 770                 }
 771                 break;
 772         default:
 773                 return -EINVAL;
 774         }
 775
 776         in_dev = __in_dev_get_rtnl(dev);
 777         if (!in_dev) {
 778                 dev_put(dev);
 779                 return -EADDRNOTAVAIL;
 780         }
 781         IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)++;
 782         inet_netconf_notify_devconf(net, NETCONFA_MC_FORWARDING, dev->ifindex,
 783                                     &in_dev->cnf);
 784         ip_rt_multicast_event(in_dev);
 785
 786         /* Fill in the VIF structures */
 787
 788         v->rate_limit = vifc->vifc_rate_limit;
 789         v->local = vifc->vifc_lcl_addr.s_addr;
 790         v->remote = vifc->vifc_rmt_addr.s_addr;
 791         v->flags = vifc->vifc_flags;
 792         if (!mrtsock)
 793                 v->flags |= VIFF_STATIC;
 794         v->threshold = vifc->vifc_threshold;
 795         v->bytes_in = 0;
 796         v->bytes_out = 0;
 797         v->pkt_in = 0;
 798         v->pkt_out = 0;
 799         v->link = dev->ifindex;
 800         if (v->flags & (VIFF_TUNNEL | VIFF_REGISTER))
 801                 v->link = dev->iflink;
 802
 803         /* And finish update writing critical data */
 804         write_lock_bh(&mrt_lock);
 805         v->dev = dev;
 806 #ifdef CONFIG_IP_PIMSM
 807         if (v->flags & VIFF_REGISTER)
 808                 mrt->mroute_reg_vif_num = vifi;
 809 #endif
 810         if (vifi+1 > mrt->maxvif)
 811                 mrt->maxvif = vifi+1;
 812         write_unlock_bh(&mrt_lock);
 813         return 0;
 814 }
 815
 816 /* called with rcu_read_lock() */
 817 static struct mfc_cache *ipmr_cache_find(struct mr_table *mrt,
 818                                          __be32 origin,
 819                                          __be32 mcastgrp)
 820 {
 821         int line = MFC_HASH(mcastgrp, origin);
 822         struct mfc_cache *c;
 823
 824         list_for_each_entry_rcu(c, &mrt->mfc_cache_array[line], list) {
 825                 if (c->mfc_origin == origin && c->mfc_mcastgrp == mcastgrp)
 826                         return c;
 827         }
 828         return NULL;
 829 }
 830
 831 /* Look for a (*,*,oif) entry */
 832 static struct mfc_cache *ipmr_cache_find_any_parent(struct mr_table *mrt,
 833                                                     int vifi)
 834 {
 835         int line = MFC_HASH(htonl(INADDR_ANY), htonl(INADDR_ANY));
 836         struct mfc_cache *c;
 837
 838         list_for_each_entry_rcu(c, &mrt->mfc_cache_array[line], list)
 839                 if (c->mfc_origin == htonl(INADDR_ANY) &&
 840                     c->mfc_mcastgrp == htonl(INADDR_ANY) &&
 841                     c->mfc_un.res.ttls[vifi] < 255)
 842                         return c;
 843
 844         return NULL;
 845 }
 846
 847 /* Look for a (*,G) entry */
 848 static struct mfc_cache *ipmr_cache_find_any(struct mr_table *mrt,
 849                                              __be32 mcastgrp, int vifi)
 850 {
 851         int line = MFC_HASH(mcastgrp, htonl(INADDR_ANY));
 852         struct mfc_cache *c, *proxy;
 853
 854         if (mcastgrp == htonl(INADDR_ANY))
 855                 goto skip;
 856
 857         list_for_each_entry_rcu(c, &mrt->mfc_cache_array[line], list)
 858                 if (c->mfc_origin == htonl(INADDR_ANY) &&
 859                     c->mfc_mcastgrp == mcastgrp) {
 860                         if (c->mfc_un.res.ttls[vifi] < 255)
 861                                 return c;
 862
 863                         /* It's ok if the vifi is part of the static tree */
 864                         proxy = ipmr_cache_find_any_parent(mrt,
 865                                                            c->mfc_parent);
 866                         if (proxy && proxy->mfc_un.res.ttls[vifi] < 255)
 867                                 return c;
 868                 }
 869
 870 skip:
 871         return ipmr_cache_find_any_parent(mrt, vifi);
 872 }
 873
 874 /*
 875  *      Allocate a multicast cache entry
 876  */
 877 static struct mfc_cache *ipmr_cache_alloc(void)
 878 {
 879         struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_KERNEL);
 880
 881         if (c)
 882                 c->mfc_un.res.minvif = MAXVIFS;
 883         return c;
 884 }
 885
 886 static struct mfc_cache *ipmr_cache_alloc_unres(void)
 887 {
 888         struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_ATOMIC);
 889
 890         if (c) {
 891                 skb_queue_head_init(&c->mfc_un.unres.unresolved);
 892                 c->mfc_un.unres.expires = jiffies + 10*HZ;
 893         }
 894         return c;
 895 }
 896
 897 /*
 898  *      A cache entry has gone into a resolved state from queued
 899  */
 900
 901 static void ipmr_cache_resolve(struct net *net, struct mr_table *mrt,
 902                                struct mfc_cache *uc, struct mfc_cache *c)
 903 {
 904         struct sk_buff *skb;
 905         struct nlmsgerr *e;
 906
 907         /* Play the pending entries through our router */
 908
 909         while ((skb = __skb_dequeue(&uc->mfc_un.unres.unresolved))) {
 910                 if (ip_hdr(skb)->version == 0) {
 911                         struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct iphdr));
 912
 913                         if (__ipmr_fill_mroute(mrt, skb, c, nlmsg_data(nlh)) > 0) {
 914                                 nlh->nlmsg_len = skb_tail_pointer(skb) -
 915                                                  (u8 *)nlh;
 916                         } else {
 917                                 nlh->nlmsg_type = NLMSG_ERROR;
 918                                 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
 919                                 skb_trim(skb, nlh->nlmsg_len);
 920                                 e = nlmsg_data(nlh);
 921                                 e->error = -EMSGSIZE;
 922                                 memset(&e->msg, 0, sizeof(e->msg));
 923                         }
 924
 925                         rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
 926                 } else {
 927                         ip_mr_forward(net, mrt, skb, c, 0);
 928                 }
 929         }
 930 }
 931
 932 /*
 933  *      Bounce a cache query up to mrouted. We could use netlink for this but mrouted
 934  *      expects the following bizarre scheme.
 935  *
 936  *      Called under mrt_lock.
 937  */
 938
 939 static int ipmr_cache_report(struct mr_table *mrt,
 940                              struct sk_buff *pkt, vifi_t vifi, int assert)
 941 {
 942         struct sk_buff *skb;
 943         const int ihl = ip_hdrlen(pkt);
 944         struct igmphdr *igmp;
 945         struct igmpmsg *msg;
 946         struct sock *mroute_sk;
 947         int ret;
 948
 949 #ifdef CONFIG_IP_PIMSM
 950         if (assert == IGMPMSG_WHOLEPKT)
 951                 skb = skb_realloc_headroom(pkt, sizeof(struct iphdr));
 952         else
 953 #endif
 954                 skb = alloc_skb(128, GFP_ATOMIC);
 955
 956         if (!skb)
 957                 return -ENOBUFS;
 958
 959 #ifdef CONFIG_IP_PIMSM
 960         if (assert == IGMPMSG_WHOLEPKT) {
 961                 /* Ugly, but we have no choice with this interface.
 962                  * Duplicate old header, fix ihl, length etc.
 963                  * And all this only to mangle msg->im_msgtype and
 964                  * to set msg->im_mbz to "mbz" :-)
 965                  */
 966                 skb_push(skb, sizeof(struct iphdr));
 967                 skb_reset_network_header(skb);
 968                 skb_reset_transport_header(skb);
 969                 msg = (struct igmpmsg *)skb_network_header(skb);
 970                 memcpy(msg, skb_network_header(pkt), sizeof(struct iphdr));
 971                 msg->im_msgtype = IGMPMSG_WHOLEPKT;
 972                 msg->im_mbz = 0;
 973                 msg->im_vif = mrt->mroute_reg_vif_num;
 974                 ip_hdr(skb)->ihl = sizeof(struct iphdr) >> 2;
 975                 ip_hdr(skb)->tot_len = htons(ntohs(ip_hdr(pkt)->tot_len) +
 976                                              sizeof(struct iphdr));
 977         } else
 978 #endif
 979         {
 980
 981         /* Copy the IP header */
 982
 983         skb_set_network_header(skb, skb->len);
 984         skb_put(skb, ihl);
 985         skb_copy_to_linear_data(skb, pkt->data, ihl);
 986         ip_hdr(skb)->protocol = 0;      /* Flag to the kernel this is a route add */
 987         msg = (struct igmpmsg *)skb_network_header(skb);
 988         msg->im_vif = vifi;
 989         skb_dst_set(skb, dst_clone(skb_dst(pkt)));
 990
 991         /* Add our header */
 992
 993         igmp = (struct igmphdr *)skb_put(skb, sizeof(struct igmphdr));
 994         igmp->type      =
 995         msg->im_msgtype = assert;
 996         igmp->code      = 0;
 997         ip_hdr(skb)->tot_len = htons(skb->len);         /* Fix the length */
 998         skb->transport_header = skb->network_header;
 999         }
1000
1001         rcu_read_lock();
1002         mroute_sk = rcu_dereference(mrt->mroute_sk);
1003         if (mroute_sk == NULL) {
1004                 rcu_read_unlock();
1005                 kfree_skb(skb);
1006                 return -EINVAL;
1007         }
1008
1009         /* Deliver to mrouted */
1010
1011         ret = sock_queue_rcv_skb(mroute_sk, skb);
1012         rcu_read_unlock();
1013         if (ret < 0) {
1014                 net_warn_ratelimited("mroute: pending queue full, dropping entries\n");
1015                 kfree_skb(skb);
1016         }
1017
1018         return ret;
1019 }
1020
1021 /*
1022  *      Queue a packet for resolution. It gets locked cache entry!
1023  */
1024
1025 static int
1026 ipmr_cache_unresolved(struct mr_table *mrt, vifi_t vifi, struct sk_buff *skb)
1027 {
1028         bool found = false;
1029         int err;
1030         struct mfc_cache *c;
1031         const struct iphdr *iph = ip_hdr(skb);
1032
1033         spin_lock_bh(&mfc_unres_lock);
1034         list_for_each_entry(c, &mrt->mfc_unres_queue, list) {
1035                 if (c->mfc_mcastgrp == iph->daddr &&
1036                     c->mfc_origin == iph->saddr) {
1037                         found = true;
1038                         break;
1039                 }
1040         }
1041
1042         if (!found) {
1043                 /* Create a new entry if allowable */
1044
1045                 if (atomic_read(&mrt->cache_resolve_queue_len) >= 10 ||
1046                     (c = ipmr_cache_alloc_unres()) == NULL) {
1047                         spin_unlock_bh(&mfc_unres_lock);
1048
1049                         kfree_skb(skb);
1050                         return -ENOBUFS;
1051                 }
1052
1053                 /* Fill in the new cache entry */
1054
1055                 c->mfc_parent   = -1;
1056                 c->mfc_origin   = iph->saddr;
1057                 c->mfc_mcastgrp = iph->daddr;
1058
1059                 /* Reflect first query at mrouted. */
1060
1061                 err = ipmr_cache_report(mrt, skb, vifi, IGMPMSG_NOCACHE);
1062                 if (err < 0) {
1063                         /* If the report failed throw the cache entry
1064                            out - Brad Parker
1065                          */
1066                         spin_unlock_bh(&mfc_unres_lock);
1067
1068                         ipmr_cache_free(c);
1069                         kfree_skb(skb);
1070                         return err;
1071                 }
1072
1073                 atomic_inc(&mrt->cache_resolve_queue_len);
1074                 list_add(&c->list, &mrt->mfc_unres_queue);
1075                 mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1076
1077                 if (atomic_read(&mrt->cache_resolve_queue_len) == 1)
1078                         mod_timer(&mrt->ipmr_expire_timer, c->mfc_un.unres.expires);
1079         }
1080
1081         /* See if we can append the packet */
1082
1083         if (c->mfc_un.unres.unresolved.qlen > 3) {
1084                 kfree_skb(skb);
1085                 err = -ENOBUFS;
1086         } else {
1087                 skb_queue_tail(&c->mfc_un.unres.unresolved, skb);
1088                 err = 0;
1089         }
1090
1091         spin_unlock_bh(&mfc_unres_lock);
1092         return err;
1093 }
1094
1095 /*
1096  *      MFC cache manipulation by user space mroute daemon
1097  */
1098
1099 static int ipmr_mfc_delete(struct mr_table *mrt, struct mfcctl *mfc, int parent)
1100 {
1101         int line;
1102         struct mfc_cache *c, *next;
1103
1104         line = MFC_HASH(mfc->mfcc_mcastgrp.s_addr, mfc->mfcc_origin.s_addr);
1105
1106         list_for_each_entry_safe(c, next, &mrt->mfc_cache_array[line], list) {
1107                 if (c->mfc_origin == mfc->mfcc_origin.s_addr &&
1108                     c->mfc_mcastgrp == mfc->mfcc_mcastgrp.s_addr &&
1109                     (parent == -1 || parent == c->mfc_parent)) {
1110                         list_del_rcu(&c->list);
1111                         mroute_netlink_event(mrt, c, RTM_DELROUTE);
1112                         ipmr_cache_free(c);
1113                         return 0;
1114                 }
1115         }
1116         return -ENOENT;
1117 }
1118
1119 static int ipmr_mfc_add(struct net *net, struct mr_table *mrt,
1120                         struct mfcctl *mfc, int mrtsock, int parent)
1121 {
1122         bool found = false;
1123         int line;
1124         struct mfc_cache *uc, *c;
1125
1126         if (mfc->mfcc_parent >= MAXVIFS)
1127                 return -ENFILE;
1128
1129         line = MFC_HASH(mfc->mfcc_mcastgrp.s_addr, mfc->mfcc_origin.s_addr);
1130
1131         list_for_each_entry(c, &mrt->mfc_cache_array[line], list) {
1132                 if (c->mfc_origin == mfc->mfcc_origin.s_addr &&
1133                     c->mfc_mcastgrp == mfc->mfcc_mcastgrp.s_addr &&
1134                     (parent == -1 || parent == c->mfc_parent)) {
1135                         found = true;
1136                         break;
1137                 }
1138         }
1139
1140         if (found) {
1141                 write_lock_bh(&mrt_lock);
1142                 c->mfc_parent = mfc->mfcc_parent;
1143                 ipmr_update_thresholds(mrt, c, mfc->mfcc_ttls);
1144                 if (!mrtsock)
1145                         c->mfc_flags |= MFC_STATIC;
1146                 write_unlock_bh(&mrt_lock);
1147                 mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1148                 return 0;
1149         }
1150
1151         if (mfc->mfcc_mcastgrp.s_addr != htonl(INADDR_ANY) &&
1152             !ipv4_is_multicast(mfc->mfcc_mcastgrp.s_addr))
1153                 return -EINVAL;
1154
1155         c = ipmr_cache_alloc();
1156         if (c == NULL)
1157                 return -ENOMEM;
1158
1159         c->mfc_origin = mfc->mfcc_origin.s_addr;
1160         c->mfc_mcastgrp = mfc->mfcc_mcastgrp.s_addr;
1161         c->mfc_parent = mfc->mfcc_parent;
1162         ipmr_update_thresholds(mrt, c, mfc->mfcc_ttls);
1163         if (!mrtsock)
1164                 c->mfc_flags |= MFC_STATIC;
1165
1166         list_add_rcu(&c->list, &mrt->mfc_cache_array[line]);
1167
1168         /*
1169          *      Check to see if we resolved a queued list. If so we
1170          *      need to send on the frames and tidy up.
1171          */
1172         found = false;
1173         spin_lock_bh(&mfc_unres_lock);
1174         list_for_each_entry(uc, &mrt->mfc_unres_queue, list) {
1175                 if (uc->mfc_origin == c->mfc_origin &&
1176                     uc->mfc_mcastgrp == c->mfc_mcastgrp) {
1177                         list_del(&uc->list);
1178                         atomic_dec(&mrt->cache_resolve_queue_len);
1179                         found = true;
1180                         break;
1181                 }
1182         }
1183         if (list_empty(&mrt->mfc_unres_queue))
1184                 del_timer(&mrt->ipmr_expire_timer);
1185         spin_unlock_bh(&mfc_unres_lock);
1186
1187         if (found) {
1188                 ipmr_cache_resolve(net, mrt, uc, c);
1189                 ipmr_cache_free(uc);
1190         }
1191         mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1192         return 0;
1193 }
1194
1195 /*
1196  *      Close the multicast socket, and clear the vif tables etc
1197  */
1198
1199 static void mroute_clean_tables(struct mr_table *mrt)
1200 {
1201         int i;
1202         LIST_HEAD(list);
1203         struct mfc_cache *c, *next;
1204
1205         /* Shut down all active vif entries */
1206
1207         for (i = 0; i < mrt->maxvif; i++) {
1208                 if (!(mrt->vif_table[i].flags & VIFF_STATIC))
1209                         vif_delete(mrt, i, 0, &list);
1210         }
1211         unregister_netdevice_many(&list);
1212
1213         /* Wipe the cache */
1214
1215         for (i = 0; i < MFC_LINES; i++) {
1216                 list_for_each_entry_safe(c, next, &mrt->mfc_cache_array[i], list) {
1217                         if (c->mfc_flags & MFC_STATIC)
1218                                 continue;
1219                         list_del_rcu(&c->list);
1220                         mroute_netlink_event(mrt, c, RTM_DELROUTE);
1221                         ipmr_cache_free(c);
1222                 }
1223         }
1224
1225         if (atomic_read(&mrt->cache_resolve_queue_len) != 0) {
1226                 spin_lock_bh(&mfc_unres_lock);
1227                 list_for_each_entry_safe(c, next, &mrt->mfc_unres_queue, list) {
1228                         list_del(&c->list);
1229                         mroute_netlink_event(mrt, c, RTM_DELROUTE);
1230                         ipmr_destroy_unres(mrt, c);
1231                 }
1232                 spin_unlock_bh(&mfc_unres_lock);
1233         }
1234 }
1235
1236 /* called from ip_ra_control(), before an RCU grace period,
1237  * we dont need to call synchronize_rcu() here
1238  */
1239 static void mrtsock_destruct(struct sock *sk)
1240 {
1241         struct net *net = sock_net(sk);
1242         struct mr_table *mrt;
1243
1244         rtnl_lock();
1245         ipmr_for_each_table(mrt, net) {
1246                 if (sk == rtnl_dereference(mrt->mroute_sk)) {
1247                         IPV4_DEVCONF_ALL(net, MC_FORWARDING)--;
1248                         inet_netconf_notify_devconf(net, NETCONFA_MC_FORWARDING,
1249                                                     NETCONFA_IFINDEX_ALL,
1250                                                     net->ipv4.devconf_all);
1251                         RCU_INIT_POINTER(mrt->mroute_sk, NULL);
1252                         mroute_clean_tables(mrt);
1253                 }
1254         }
1255         rtnl_unlock();
1256 }
1257
1258 /*
1259  *      Socket options and virtual interface manipulation. The whole
1260  *      virtual interface system is a complete heap, but unfortunately
1261  *      that's how BSD mrouted happens to think. Maybe one day with a proper
1262  *      MOSPF/PIM router set up we can clean this up.
1263  */
1264
1265 int ip_mroute_setsockopt(struct sock *sk, int optname, char __user *optval, unsigned int optlen)
1266 {
1267         int ret, parent = 0;
1268         struct vifctl vif;
1269         struct mfcctl mfc;
1270         struct net *net = sock_net(sk);
1271         struct mr_table *mrt;
1272
1273         if (sk->sk_type != SOCK_RAW ||
1274             inet_sk(sk)->inet_num != IPPROTO_IGMP)
1275                 return -EOPNOTSUPP;
1276
1277         mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1278         if (mrt == NULL)
1279                 return -ENOENT;
1280
1281         if (optname != MRT_INIT) {
1282                 if (sk != rcu_access_pointer(mrt->mroute_sk) &&
1283                     !ns_capable(net->user_ns, CAP_NET_ADMIN))
1284                         return -EACCES;
1285         }
1286
1287         switch (optname) {
1288         case MRT_INIT:
1289                 if (optlen != sizeof(int))
1290                         return -EINVAL;
1291
1292                 rtnl_lock();
1293                 if (rtnl_dereference(mrt->mroute_sk)) {
1294                         rtnl_unlock();
1295                         return -EADDRINUSE;
1296                 }
1297
1298                 ret = ip_ra_control(sk, 1, mrtsock_destruct);
1299                 if (ret == 0) {
1300                         rcu_assign_pointer(mrt->mroute_sk, sk);
1301                         IPV4_DEVCONF_ALL(net, MC_FORWARDING)++;
1302                         inet_netconf_notify_devconf(net, NETCONFA_MC_FORWARDING,
1303                                                     NETCONFA_IFINDEX_ALL,
1304                                                     net->ipv4.devconf_all);
1305                 }
1306                 rtnl_unlock();
1307                 return ret;
1308         case MRT_DONE:
1309                 if (sk != rcu_access_pointer(mrt->mroute_sk))
1310                         return -EACCES;
1311                 return ip_ra_control(sk, 0, NULL);
1312         case MRT_ADD_VIF:
1313         case MRT_DEL_VIF:
1314                 if (optlen != sizeof(vif))
1315                         return -EINVAL;
1316                 if (copy_from_user(&vif, optval, sizeof(vif)))
1317                         return -EFAULT;
1318                 if (vif.vifc_vifi >= MAXVIFS)
1319                         return -ENFILE;
1320                 rtnl_lock();
1321                 if (optname == MRT_ADD_VIF) {
1322                         ret = vif_add(net, mrt, &vif,
1323                                       sk == rtnl_dereference(mrt->mroute_sk));
1324                 } else {
1325                         ret = vif_delete(mrt, vif.vifc_vifi, 0, NULL);
1326                 }
1327                 rtnl_unlock();
1328                 return ret;
1329
1330                 /*
1331                  *      Manipulate the forwarding caches. These live
1332                  *      in a sort of kernel/user symbiosis.
1333                  */
1334         case MRT_ADD_MFC:
1335         case MRT_DEL_MFC:
1336                 parent = -1;
1337         case MRT_ADD_MFC_PROXY:
1338         case MRT_DEL_MFC_PROXY:
1339                 if (optlen != sizeof(mfc))
1340                         return -EINVAL;
1341                 if (copy_from_user(&mfc, optval, sizeof(mfc)))
1342                         return -EFAULT;
1343                 if (parent == 0)
1344                         parent = mfc.mfcc_parent;
1345                 rtnl_lock();
1346                 if (optname == MRT_DEL_MFC || optname == MRT_DEL_MFC_PROXY)
1347                         ret = ipmr_mfc_delete(mrt, &mfc, parent);
1348                 else
1349                         ret = ipmr_mfc_add(net, mrt, &mfc,
1350                                            sk == rtnl_dereference(mrt->mroute_sk),
1351                                            parent);
1352                 rtnl_unlock();
1353                 return ret;
1354                 /*
1355                  *      Control PIM assert.
1356                  */
1357         case MRT_ASSERT:
1358         {
1359                 int v;
1360                 if (optlen != sizeof(v))
1361                         return -EINVAL;
1362                 if (get_user(v, (int __user *)optval))
1363                         return -EFAULT;
1364                 mrt->mroute_do_assert = v;
1365                 return 0;
1366         }
1367 #ifdef CONFIG_IP_PIMSM
1368         case MRT_PIM:
1369         {
1370                 int v;
1371
1372                 if (optlen != sizeof(v))
1373                         return -EINVAL;
1374                 if (get_user(v, (int __user *)optval))
1375                         return -EFAULT;
1376                 v = !!v;
1377
1378                 rtnl_lock();
1379                 ret = 0;
1380                 if (v != mrt->mroute_do_pim) {
1381                         mrt->mroute_do_pim = v;
1382                         mrt->mroute_do_assert = v;
1383                 }
1384                 rtnl_unlock();
1385                 return ret;
1386         }
1387 #endif
1388 #ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES
1389         case MRT_TABLE:
1390         {
1391                 u32 v;
1392
1393                 if (optlen != sizeof(u32))
1394                         return -EINVAL;
1395                 if (get_user(v, (u32 __user *)optval))
1396                         return -EFAULT;
1397
1398                 /* "pimreg%u" should not exceed 16 bytes (IFNAMSIZ) */
1399                 if (v != RT_TABLE_DEFAULT && v >= 1000000000)
1400                         return -EINVAL;
1401
1402                 rtnl_lock();
1403                 ret = 0;
1404                 if (sk == rtnl_dereference(mrt->mroute_sk)) {
1405                         ret = -EBUSY;
1406                 } else {
1407                         if (!ipmr_new_table(net, v))
1408                                 ret = -ENOMEM;
1409                         else
1410                                 raw_sk(sk)->ipmr_table = v;
1411                 }
1412                 rtnl_unlock();
1413                 return ret;
1414         }
1415 #endif
1416         /*
1417          *      Spurious command, or MRT_VERSION which you cannot
1418          *      set.
1419          */
1420         default:
1421                 return -ENOPROTOOPT;
1422         }
1423 }
1424
1425 /*
1426  *      Getsock opt support for the multicast routing system.
1427  */
1428
1429 int ip_mroute_getsockopt(struct sock *sk, int optname, char __user *optval, int __user *optlen)
1430 {
1431         int olr;
1432         int val;
1433         struct net *net = sock_net(sk);
1434         struct mr_table *mrt;
1435
1436         if (sk->sk_type != SOCK_RAW ||
1437             inet_sk(sk)->inet_num != IPPROTO_IGMP)
1438                 return -EOPNOTSUPP;
1439
1440         mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1441         if (mrt == NULL)
1442                 return -ENOENT;
1443
1444         if (optname != MRT_VERSION &&
1445 #ifdef CONFIG_IP_PIMSM
1446            optname != MRT_PIM &&
1447 #endif
1448            optname != MRT_ASSERT)
1449                 return -ENOPROTOOPT;
1450
1451         if (get_user(olr, optlen))
1452                 return -EFAULT;
1453
1454         olr = min_t(unsigned int, olr, sizeof(int));
1455         if (olr < 0)
1456                 return -EINVAL;
1457
1458         if (put_user(olr, optlen))
1459                 return -EFAULT;
1460         if (optname == MRT_VERSION)
1461                 val = 0x0305;
1462 #ifdef CONFIG_IP_PIMSM
1463         else if (optname == MRT_PIM)
1464                 val = mrt->mroute_do_pim;
1465 #endif
1466         else
1467                 val = mrt->mroute_do_assert;
1468         if (copy_to_user(optval, &val, olr))
1469                 return -EFAULT;
1470         return 0;
1471 }
1472
1473 /*
1474  *      The IP multicast ioctl support routines.
1475  */
1476
1477 int ipmr_ioctl(struct sock *sk, int cmd, void __user *arg)
1478 {
1479         struct sioc_sg_req sr;
1480         struct sioc_vif_req vr;
1481         struct vif_device *vif;
1482         struct mfc_cache *c;
1483         struct net *net = sock_net(sk);
1484         struct mr_table *mrt;
1485
1486         mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1487         if (mrt == NULL)
1488                 return -ENOENT;
1489
1490         switch (cmd) {
1491         case SIOCGETVIFCNT:
1492                 if (copy_from_user(&vr, arg, sizeof(vr)))
1493                         return -EFAULT;
1494                 if (vr.vifi >= mrt->maxvif)
1495                         return -EINVAL;
1496                 read_lock(&mrt_lock);
1497                 vif = &mrt->vif_table[vr.vifi];
1498                 if (VIF_EXISTS(mrt, vr.vifi)) {
1499                         vr.icount = vif->pkt_in;
1500                         vr.ocount = vif->pkt_out;
1501                         vr.ibytes = vif->bytes_in;
1502                         vr.obytes = vif->bytes_out;
1503                         read_unlock(&mrt_lock);
1504
1505                         if (copy_to_user(arg, &vr, sizeof(vr)))
1506                                 return -EFAULT;
1507                         return 0;
1508                 }
1509                 read_unlock(&mrt_lock);
1510                 return -EADDRNOTAVAIL;
1511         case SIOCGETSGCNT:
1512                 if (copy_from_user(&sr, arg, sizeof(sr)))
1513                         return -EFAULT;
1514
1515                 rcu_read_lock();
1516                 c = ipmr_cache_find(mrt, sr.src.s_addr, sr.grp.s_addr);
1517                 if (c) {
1518                         sr.pktcnt = c->mfc_un.res.pkt;
1519                         sr.bytecnt = c->mfc_un.res.bytes;
1520                         sr.wrong_if = c->mfc_un.res.wrong_if;
1521                         rcu_read_unlock();
1522
1523                         if (copy_to_user(arg, &sr, sizeof(sr)))
1524                                 return -EFAULT;
1525                         return 0;
1526                 }
1527                 rcu_read_unlock();
1528                 return -EADDRNOTAVAIL;
1529         default:
1530                 return -ENOIOCTLCMD;
1531         }
1532 }
1533
1534 #ifdef CONFIG_COMPAT
1535 struct compat_sioc_sg_req {
1536         struct in_addr src;
1537         struct in_addr grp;
1538         compat_ulong_t pktcnt;
1539         compat_ulong_t bytecnt;
1540         compat_ulong_t wrong_if;
1541 };
1542
1543 struct compat_sioc_vif_req {
1544         vifi_t  vifi;           /* Which iface */
1545         compat_ulong_t icount;
1546         compat_ulong_t ocount;
1547         compat_ulong_t ibytes;
1548         compat_ulong_t obytes;
1549 };
1550
1551 int ipmr_compat_ioctl(struct sock *sk, unsigned int cmd, void __user *arg)
1552 {
1553         struct compat_sioc_sg_req sr;
1554         struct compat_sioc_vif_req vr;
1555         struct vif_device *vif;
1556         struct mfc_cache *c;
1557         struct net *net = sock_net(sk);
1558         struct mr_table *mrt;
1559
1560         mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1561         if (mrt == NULL)
1562                 return -ENOENT;
1563
1564         switch (cmd) {
1565         case SIOCGETVIFCNT:
1566                 if (copy_from_user(&vr, arg, sizeof(vr)))
1567                         return -EFAULT;
1568                 if (vr.vifi >= mrt->maxvif)
1569                         return -EINVAL;
1570                 read_lock(&mrt_lock);
1571                 vif = &mrt->vif_table[vr.vifi];
1572                 if (VIF_EXISTS(mrt, vr.vifi)) {
1573                         vr.icount = vif->pkt_in;
1574                         vr.ocount = vif->pkt_out;
1575                         vr.ibytes = vif->bytes_in;
1576                         vr.obytes = vif->bytes_out;
1577                         read_unlock(&mrt_lock);
1578
1579                         if (copy_to_user(arg, &vr, sizeof(vr)))
1580                                 return -EFAULT;
1581                         return 0;
1582                 }
1583                 read_unlock(&mrt_lock);
1584                 return -EADDRNOTAVAIL;
1585         case SIOCGETSGCNT:
1586                 if (copy_from_user(&sr, arg, sizeof(sr)))
1587                         return -EFAULT;
1588
1589                 rcu_read_lock();
1590                 c = ipmr_cache_find(mrt, sr.src.s_addr, sr.grp.s_addr);
1591                 if (c) {
1592                         sr.pktcnt = c->mfc_un.res.pkt;
1593                         sr.bytecnt = c->mfc_un.res.bytes;
1594                         sr.wrong_if = c->mfc_un.res.wrong_if;
1595                         rcu_read_unlock();
1596
1597                         if (copy_to_user(arg, &sr, sizeof(sr)))
1598                                 return -EFAULT;
1599                         return 0;
1600                 }
1601                 rcu_read_unlock();
1602                 return -EADDRNOTAVAIL;
1603         default:
1604                 return -ENOIOCTLCMD;
1605         }
1606 }
1607 #endif
1608
1609
1610 static int ipmr_device_event(struct notifier_block *this, unsigned long event, void *ptr)
1611 {
1612         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1613         struct net *net = dev_net(dev);
1614         struct mr_table *mrt;
1615         struct vif_device *v;
1616         int ct;
1617
1618         if (event != NETDEV_UNREGISTER)
1619                 return NOTIFY_DONE;
1620
1621         ipmr_for_each_table(mrt, net) {
1622                 v = &mrt->vif_table[0];
1623                 for (ct = 0; ct < mrt->maxvif; ct++, v++) {
1624                         if (v->dev == dev)
1625                                 vif_delete(mrt, ct, 1, NULL);
1626                 }
1627         }
1628         return NOTIFY_DONE;
1629 }
1630
1631
1632 static struct notifier_block ip_mr_notifier = {
1633         .notifier_call = ipmr_device_event,
1634 };
1635
1636 /*
1637  *      Encapsulate a packet by attaching a valid IPIP header to it.
1638  *      This avoids tunnel drivers and other mess and gives us the speed so
1639  *      important for multicast video.
1640  */
1641
1642 static void ip_encap(struct sk_buff *skb, __be32 saddr, __be32 daddr)
1643 {
1644         struct iphdr *iph;
1645         const struct iphdr *old_iph = ip_hdr(skb);
1646
1647         skb_push(skb, sizeof(struct iphdr));
1648         skb->transport_header = skb->network_header;
1649         skb_reset_network_header(skb);
1650         iph = ip_hdr(skb);
1651
1652         iph->version    =       4;
1653         iph->tos        =       old_iph->tos;
1654         iph->ttl        =       old_iph->ttl;
1655         iph->frag_off   =       0;
1656         iph->daddr      =       daddr;
1657         iph->saddr      =       saddr;
1658         iph->protocol   =       IPPROTO_IPIP;
1659         iph->ihl        =       5;
1660         iph->tot_len    =       htons(skb->len);
1661         ip_select_ident(skb, skb_dst(skb), NULL);
1662         ip_send_check(iph);
1663
1664         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1665         nf_reset(skb);
1666 }
1667
1668 static inline int ipmr_forward_finish(struct sk_buff *skb)
1669 {
1670         struct ip_options *opt = &(IPCB(skb)->opt);
1671
1672         IP_INC_STATS_BH(dev_net(skb_dst(skb)->dev), IPSTATS_MIB_OUTFORWDATAGRAMS);
1673         IP_ADD_STATS_BH(dev_net(skb_dst(skb)->dev), IPSTATS_MIB_OUTOCTETS, skb->len);
1674
1675         if (unlikely(opt->optlen))
1676                 ip_forward_options(skb);
1677
1678         return dst_output(skb);
1679 }
1680
1681 /*
1682  *      Processing handlers for ipmr_forward
1683  */
1684
1685 static void ipmr_queue_xmit(struct net *net, struct mr_table *mrt,
1686                             struct sk_buff *skb, struct mfc_cache *c, int vifi)
1687 {
1688         const struct iphdr *iph = ip_hdr(skb);
1689         struct vif_device *vif = &mrt->vif_table[vifi];
1690         struct net_device *dev;
1691         struct rtable *rt;
1692         struct flowi4 fl4;
1693         int    encap = 0;
1694
1695         if (vif->dev == NULL)
1696                 goto out_free;
1697
1698 #ifdef CONFIG_IP_PIMSM
1699         if (vif->flags & VIFF_REGISTER) {
1700                 vif->pkt_out++;
1701                 vif->bytes_out += skb->len;
1702                 vif->dev->stats.tx_bytes += skb->len;
1703                 vif->dev->stats.tx_packets++;
1704                 ipmr_cache_report(mrt, skb, vifi, IGMPMSG_WHOLEPKT);
1705                 goto out_free;
1706         }
1707 #endif
1708
1709         if (vif->flags & VIFF_TUNNEL) {
1710                 rt = ip_route_output_ports(net, &fl4, NULL,
1711                                            vif->remote, vif->local,
1712                                            0, 0,
1713                                            IPPROTO_IPIP,
1714                                            RT_TOS(iph->tos), vif->link);
1715                 if (IS_ERR(rt))
1716                         goto out_free;
1717                 encap = sizeof(struct iphdr);
1718         } else {
1719                 rt = ip_route_output_ports(net, &fl4, NULL, iph->daddr, 0,
1720                                            0, 0,
1721                                            IPPROTO_IPIP,
1722                                            RT_TOS(iph->tos), vif->link);
1723                 if (IS_ERR(rt))
1724                         goto out_free;
1725         }
1726
1727         dev = rt->dst.dev;
1728
1729         if (skb->len+encap > dst_mtu(&rt->dst) && (ntohs(iph->frag_off) & IP_DF)) {
1730                 /* Do not fragment multicasts. Alas, IPv4 does not
1731                  * allow to send ICMP, so that packets will disappear
1732                  * to blackhole.
1733                  */
1734
1735                 IP_INC_STATS_BH(dev_net(dev), IPSTATS_MIB_FRAGFAILS);
1736                 ip_rt_put(rt);
1737                 goto out_free;
1738         }
1739
1740         encap += LL_RESERVED_SPACE(dev) + rt->dst.header_len;
1741
1742         if (skb_cow(skb, encap)) {
1743                 ip_rt_put(rt);
1744                 goto out_free;
1745         }
1746
1747         vif->pkt_out++;
1748         vif->bytes_out += skb->len;
1749
1750         skb_dst_drop(skb);
1751         skb_dst_set(skb, &rt->dst);
1752         ip_decrease_ttl(ip_hdr(skb));
1753
1754         /* FIXME: forward and output firewalls used to be called here.
1755          * What do we do with netfilter? -- RR
1756          */
1757         if (vif->flags & VIFF_TUNNEL) {
1758                 ip_encap(skb, vif->local, vif->remote);
1759                 /* FIXME: extra output firewall step used to be here. --RR */
1760                 vif->dev->stats.tx_packets++;
1761                 vif->dev->stats.tx_bytes += skb->len;
1762         }
1763
1764         IPCB(skb)->flags |= IPSKB_FORWARDED;
1765
1766         /*
1767          * RFC1584 teaches, that DVMRP/PIM router must deliver packets locally
1768          * not only before forwarding, but after forwarding on all output
1769          * interfaces. It is clear, if mrouter runs a multicasting
1770          * program, it should receive packets not depending to what interface
1771          * program is joined.
1772          * If we will not make it, the program will have to join on all
1773          * interfaces. On the other hand, multihoming host (or router, but
1774          * not mrouter) cannot join to more than one interface - it will
1775          * result in receiving multiple packets.
1776          */
1777         NF_HOOK(NFPROTO_IPV4, NF_INET_FORWARD, skb, skb->dev, dev,
1778                 ipmr_forward_finish);
1779         return;
1780
1781 out_free:
1782         kfree_skb(skb);
1783 }
1784
1785 static int ipmr_find_vif(struct mr_table *mrt, struct net_device *dev)
1786 {
1787         int ct;
1788
1789         for (ct = mrt->maxvif-1; ct >= 0; ct--) {
1790                 if (mrt->vif_table[ct].dev == dev)
1791                         break;
1792         }
1793         return ct;
1794 }
1795
1796 /* "local" means that we should preserve one skb (for local delivery) */
1797
1798 static void ip_mr_forward(struct net *net, struct mr_table *mrt,
1799                           struct sk_buff *skb, struct mfc_cache *cache,
1800                           int local)
1801 {
1802         int psend = -1;
1803         int vif, ct;
1804         int true_vifi = ipmr_find_vif(mrt, skb->dev);
1805
1806         vif = cache->mfc_parent;
1807         cache->mfc_un.res.pkt++;
1808         cache->mfc_un.res.bytes += skb->len;
1809
1810         if (cache->mfc_origin == htonl(INADDR_ANY) && true_vifi >= 0) {
1811                 struct mfc_cache *cache_proxy;
1812
1813                 /* For an (*,G) entry, we only check that the incomming
1814                  * interface is part of the static tree.
1815                  */
1816                 cache_proxy = ipmr_cache_find_any_parent(mrt, vif);
1817                 if (cache_proxy &&
1818                     cache_proxy->mfc_un.res.ttls[true_vifi] < 255)
1819                         goto forward;
1820         }
1821
1822         /*
1823          * Wrong interface: drop packet and (maybe) send PIM assert.
1824          */
1825         if (mrt->vif_table[vif].dev != skb->dev) {
1826                 if (rt_is_output_route(skb_rtable(skb))) {
1827                         /* It is our own packet, looped back.
1828                          * Very complicated situation...
1829                          *
1830                          * The best workaround until routing daemons will be
1831                          * fixed is not to redistribute packet, if it was
1832                          * send through wrong interface. It means, that
1833                          * multicast applications WILL NOT work for
1834                          * (S,G), which have default multicast route pointing
1835                          * to wrong oif. In any case, it is not a good
1836                          * idea to use multicasting applications on router.
1837                          */
1838                         goto dont_forward;
1839                 }
1840
1841                 cache->mfc_un.res.wrong_if++;
1842
1843                 if (true_vifi >= 0 && mrt->mroute_do_assert &&
1844                     /* pimsm uses asserts, when switching from RPT to SPT,
1845                      * so that we cannot check that packet arrived on an oif.
1846                      * It is bad, but otherwise we would need to move pretty
1847                      * large chunk of pimd to kernel. Ough... --ANK
1848                      */
1849                     (mrt->mroute_do_pim ||
1850                      cache->mfc_un.res.ttls[true_vifi] < 255) &&
1851                     time_after(jiffies,
1852                                cache->mfc_un.res.last_assert + MFC_ASSERT_THRESH)) {
1853                         cache->mfc_un.res.last_assert = jiffies;
1854                         ipmr_cache_report(mrt, skb, true_vifi, IGMPMSG_WRONGVIF);
1855                 }
1856                 goto dont_forward;
1857         }
1858
1859 forward:
1860         mrt->vif_table[vif].pkt_in++;
1861         mrt->vif_table[vif].bytes_in += skb->len;
1862
1863         /*
1864          *      Forward the frame
1865          */
1866         if (cache->mfc_origin == htonl(INADDR_ANY) &&
1867             cache->mfc_mcastgrp == htonl(INADDR_ANY)) {
1868                 if (true_vifi >= 0 &&
1869                     true_vifi != cache->mfc_parent &&
1870                     ip_hdr(skb)->ttl >
1871                                 cache->mfc_un.res.ttls[cache->mfc_parent]) {
1872                         /* It's an (*,*) entry and the packet is not coming from
1873                          * the upstream: forward the packet to the upstream
1874                          * only.
1875                          */
1876                         psend = cache->mfc_parent;
1877                         goto last_forward;
1878                 }
1879                 goto dont_forward;
1880         }
1881         for (ct = cache->mfc_un.res.maxvif - 1;
1882              ct >= cache->mfc_un.res.minvif; ct--) {
1883                 /* For (*,G) entry, don't forward to the incoming interface */
1884                 if ((cache->mfc_origin != htonl(INADDR_ANY) ||
1885                      ct != true_vifi) &&
1886                     ip_hdr(skb)->ttl > cache->mfc_un.res.ttls[ct]) {
1887                         if (psend != -1) {
1888                                 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
1889
1890                                 if (skb2)
1891                                         ipmr_queue_xmit(net, mrt, skb2, cache,
1892                                                         psend);
1893                         }
1894                         psend = ct;
1895                 }
1896         }
1897 last_forward:
1898         if (psend != -1) {
1899                 if (local) {
1900                         struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
1901
1902                         if (skb2)
1903                                 ipmr_queue_xmit(net, mrt, skb2, cache, psend);
1904                 } else {
1905                         ipmr_queue_xmit(net, mrt, skb, cache, psend);
1906                         return;
1907                 }
1908         }
1909
1910 dont_forward:
1911         if (!local)
1912                 kfree_skb(skb);
1913 }
1914
1915 static struct mr_table *ipmr_rt_fib_lookup(struct net *net, struct sk_buff *skb)
1916 {
1917         struct rtable *rt = skb_rtable(skb);
1918         struct iphdr *iph = ip_hdr(skb);
1919         struct flowi4 fl4 = {
1920                 .daddr = iph->daddr,
1921                 .saddr = iph->saddr,
1922                 .flowi4_tos = RT_TOS(iph->tos),
1923                 .flowi4_oif = (rt_is_output_route(rt) ?
1924                                skb->dev->ifindex : 0),
1925                 .flowi4_iif = (rt_is_output_route(rt) ?
1926                                LOOPBACK_IFINDEX :
1927                                skb->dev->ifindex),
1928                 .flowi4_mark = skb->mark,
1929         };
1930         struct mr_table *mrt;
1931         int err;
1932
1933         err = ipmr_fib_lookup(net, &fl4, &mrt);
1934         if (err)
1935                 return ERR_PTR(err);
1936         return mrt;
1937 }
1938
1939 /*
1940  *      Multicast packets for forwarding arrive here
1941  *      Called with rcu_read_lock();
1942  */
1943
1944 int ip_mr_input(struct sk_buff *skb)
1945 {
1946         struct mfc_cache *cache;
1947         struct net *net = dev_net(skb->dev);
1948         int local = skb_rtable(skb)->rt_flags & RTCF_LOCAL;
1949         struct mr_table *mrt;
1950
1951         /* Packet is looped back after forward, it should not be
1952          * forwarded second time, but still can be delivered locally.
1953          */
1954         if (IPCB(skb)->flags & IPSKB_FORWARDED)
1955                 goto dont_forward;
1956
1957         mrt = ipmr_rt_fib_lookup(net, skb);
1958         if (IS_ERR(mrt)) {
1959                 kfree_skb(skb);
1960                 return PTR_ERR(mrt);
1961         }
1962         if (!local) {
1963                 if (IPCB(skb)->opt.router_alert) {
1964                         if (ip_call_ra_chain(skb))
1965                                 return 0;
1966                 } else if (ip_hdr(skb)->protocol == IPPROTO_IGMP) {
1967                         /* IGMPv1 (and broken IGMPv2 implementations sort of
1968                          * Cisco IOS <= 11.2(8)) do not put router alert
1969                          * option to IGMP packets destined to routable
1970                          * groups. It is very bad, because it means
1971                          * that we can forward NO IGMP messages.
1972                          */
1973                         struct sock *mroute_sk;
1974
1975                         mroute_sk = rcu_dereference(mrt->mroute_sk);
1976                         if (mroute_sk) {
1977                                 nf_reset(skb);
1978                                 raw_rcv(mroute_sk, skb);
1979                                 return 0;
1980                         }
1981                     }
1982         }
1983
1984         /* already under rcu_read_lock() */
1985         cache = ipmr_cache_find(mrt, ip_hdr(skb)->saddr, ip_hdr(skb)->daddr);
1986         if (cache == NULL) {
1987                 int vif = ipmr_find_vif(mrt, skb->dev);
1988
1989                 if (vif >= 0)
1990                         cache = ipmr_cache_find_any(mrt, ip_hdr(skb)->daddr,
1991                                                     vif);
1992         }
1993
1994         /*
1995          *      No usable cache entry
1996          */
1997         if (cache == NULL) {
1998                 int vif;
1999
2000                 if (local) {
2001                         struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2002                         ip_local_deliver(skb);
2003                         if (skb2 == NULL)
2004                                 return -ENOBUFS;
2005                         skb = skb2;
2006                 }
2007
2008                 read_lock(&mrt_lock);
2009                 vif = ipmr_find_vif(mrt, skb->dev);
2010                 if (vif >= 0) {
2011                         int err2 = ipmr_cache_unresolved(mrt, vif, skb);
2012                         read_unlock(&mrt_lock);
2013
2014                         return err2;
2015                 }
2016                 read_unlock(&mrt_lock);
2017                 kfree_skb(skb);
2018                 return -ENODEV;
2019         }
2020
2021         read_lock(&mrt_lock);
2022         ip_mr_forward(net, mrt, skb, cache, local);
2023         read_unlock(&mrt_lock);
2024
2025         if (local)
2026                 return ip_local_deliver(skb);
2027
2028         return 0;
2029
2030 dont_forward:
2031         if (local)
2032                 return ip_local_deliver(skb);
2033         kfree_skb(skb);
2034         return 0;
2035 }
2036
2037 #ifdef CONFIG_IP_PIMSM
2038 /* called with rcu_read_lock() */
2039 static int __pim_rcv(struct mr_table *mrt, struct sk_buff *skb,
2040                      unsigned int pimlen)
2041 {
2042         struct net_device *reg_dev = NULL;
2043         struct iphdr *encap;
2044
2045         encap = (struct iphdr *)(skb_transport_header(skb) + pimlen);
2046         /*
2047          * Check that:
2048          * a. packet is really sent to a multicast group
2049          * b. packet is not a NULL-REGISTER
2050          * c. packet is not truncated
2051          */
2052         if (!ipv4_is_multicast(encap->daddr) ||
2053             encap->tot_len == 0 ||
2054             ntohs(encap->tot_len) + pimlen > skb->len)
2055                 return 1;
2056
2057         read_lock(&mrt_lock);
2058         if (mrt->mroute_reg_vif_num >= 0)
2059                 reg_dev = mrt->vif_table[mrt->mroute_reg_vif_num].dev;
2060         read_unlock(&mrt_lock);
2061
2062         if (reg_dev == NULL)
2063                 return 1;
2064
2065         skb->mac_header = skb->network_header;
2066         skb_pull(skb, (u8 *)encap - skb->data);
2067         skb_reset_network_header(skb);
2068         skb->protocol = htons(ETH_P_IP);
2069         skb->ip_summed = CHECKSUM_NONE;
2070
2071         skb_tunnel_rx(skb, reg_dev, dev_net(reg_dev));
2072
2073         netif_rx(skb);
2074
2075         return NET_RX_SUCCESS;
2076 }
2077 #endif
2078
2079 #ifdef CONFIG_IP_PIMSM_V1
2080 /*
2081  * Handle IGMP messages of PIMv1
2082  */
2083
2084 int pim_rcv_v1(struct sk_buff *skb)
2085 {
2086         struct igmphdr *pim;
2087         struct net *net = dev_net(skb->dev);
2088         struct mr_table *mrt;
2089
2090         if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr)))
2091                 goto drop;
2092
2093         pim = igmp_hdr(skb);
2094
2095         mrt = ipmr_rt_fib_lookup(net, skb);
2096         if (IS_ERR(mrt))
2097                 goto drop;
2098         if (!mrt->mroute_do_pim ||
2099             pim->group != PIM_V1_VERSION || pim->code != PIM_V1_REGISTER)
2100                 goto drop;
2101
2102         if (__pim_rcv(mrt, skb, sizeof(*pim))) {
2103 drop:
2104                 kfree_skb(skb);
2105         }
2106         return 0;
2107 }
2108 #endif
2109
2110 #ifdef CONFIG_IP_PIMSM_V2
2111 static int pim_rcv(struct sk_buff *skb)
2112 {
2113         struct pimreghdr *pim;
2114         struct net *net = dev_net(skb->dev);
2115         struct mr_table *mrt;
2116
2117         if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr)))
2118                 goto drop;
2119
2120         pim = (struct pimreghdr *)skb_transport_header(skb);
2121         if (pim->type != ((PIM_VERSION << 4) | (PIM_REGISTER)) ||
2122             (pim->flags & PIM_NULL_REGISTER) ||
2123             (ip_compute_csum((void *)pim, sizeof(*pim)) != 0 &&
2124              csum_fold(skb_checksum(skb, 0, skb->len, 0))))
2125                 goto drop;
2126
2127         mrt = ipmr_rt_fib_lookup(net, skb);
2128         if (IS_ERR(mrt))
2129                 goto drop;
2130         if (__pim_rcv(mrt, skb, sizeof(*pim))) {
2131 drop:
2132                 kfree_skb(skb);
2133         }
2134         return 0;
2135 }
2136 #endif
2137
2138 static int __ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
2139                               struct mfc_cache *c, struct rtmsg *rtm)
2140 {
2141         int ct;
2142         struct rtnexthop *nhp;
2143         struct nlattr *mp_attr;
2144         struct rta_mfc_stats mfcs;
2145
2146         /* If cache is unresolved, don't try to parse IIF and OIF */
2147         if (c->mfc_parent >= MAXVIFS)
2148                 return -ENOENT;
2149
2150         if (VIF_EXISTS(mrt, c->mfc_parent) &&
2151             nla_put_u32(skb, RTA_IIF, mrt->vif_table[c->mfc_parent].dev->ifindex) < 0)
2152                 return -EMSGSIZE;
2153
2154         if (!(mp_attr = nla_nest_start(skb, RTA_MULTIPATH)))
2155                 return -EMSGSIZE;
2156
2157         for (ct = c->mfc_un.res.minvif; ct < c->mfc_un.res.maxvif; ct++) {
2158                 if (VIF_EXISTS(mrt, ct) && c->mfc_un.res.ttls[ct] < 255) {
2159                         if (!(nhp = nla_reserve_nohdr(skb, sizeof(*nhp)))) {
2160                                 nla_nest_cancel(skb, mp_attr);
2161                                 return -EMSGSIZE;
2162                         }
2163
2164                         nhp->rtnh_flags = 0;
2165                         nhp->rtnh_hops = c->mfc_un.res.ttls[ct];
2166                         nhp->rtnh_ifindex = mrt->vif_table[ct].dev->ifindex;
2167                         nhp->rtnh_len = sizeof(*nhp);
2168                 }
2169         }
2170
2171         nla_nest_end(skb, mp_attr);
2172
2173         mfcs.mfcs_packets = c->mfc_un.res.pkt;
2174         mfcs.mfcs_bytes = c->mfc_un.res.bytes;
2175         mfcs.mfcs_wrong_if = c->mfc_un.res.wrong_if;
2176         if (nla_put(skb, RTA_MFC_STATS, sizeof(mfcs), &mfcs) < 0)
2177                 return -EMSGSIZE;
2178
2179         rtm->rtm_type = RTN_MULTICAST;
2180         return 1;
2181 }
2182
2183 int ipmr_get_route(struct net *net, struct sk_buff *skb,
2184                    __be32 saddr, __be32 daddr,
2185                    struct rtmsg *rtm, int nowait)
2186 {
2187         struct mfc_cache *cache;
2188         struct mr_table *mrt;
2189         int err;
2190
2191         mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2192         if (mrt == NULL)
2193                 return -ENOENT;
2194
2195         rcu_read_lock();
2196         cache = ipmr_cache_find(mrt, saddr, daddr);
2197         if (cache == NULL && skb->dev) {
2198                 int vif = ipmr_find_vif(mrt, skb->dev);
2199
2200                 if (vif >= 0)
2201                         cache = ipmr_cache_find_any(mrt, daddr, vif);
2202         }
2203         if (cache == NULL) {
2204                 struct sk_buff *skb2;
2205                 struct iphdr *iph;
2206                 struct net_device *dev;
2207                 int vif = -1;
2208
2209                 if (nowait) {
2210                         rcu_read_unlock();
2211                         return -EAGAIN;
2212                 }
2213
2214                 dev = skb->dev;
2215                 read_lock(&mrt_lock);
2216                 if (dev)
2217                         vif = ipmr_find_vif(mrt, dev);
2218                 if (vif < 0) {
2219                         read_unlock(&mrt_lock);
2220                         rcu_read_unlock();
2221                         return -ENODEV;
2222                 }
2223                 skb2 = skb_clone(skb, GFP_ATOMIC);
2224                 if (!skb2) {
2225                         read_unlock(&mrt_lock);
2226                         rcu_read_unlock();
2227                         return -ENOMEM;
2228                 }
2229
2230                 skb_push(skb2, sizeof(struct iphdr));
2231                 skb_reset_network_header(skb2);
2232                 iph = ip_hdr(skb2);
2233                 iph->ihl = sizeof(struct iphdr) >> 2;
2234                 iph->saddr = saddr;
2235                 iph->daddr = daddr;
2236                 iph->version = 0;
2237                 err = ipmr_cache_unresolved(mrt, vif, skb2);
2238                 read_unlock(&mrt_lock);
2239                 rcu_read_unlock();
2240                 return err;
2241         }
2242
2243         read_lock(&mrt_lock);
2244         if (!nowait && (rtm->rtm_flags & RTM_F_NOTIFY))
2245                 cache->mfc_flags |= MFC_NOTIFY;
2246         err = __ipmr_fill_mroute(mrt, skb, cache, rtm);
2247         read_unlock(&mrt_lock);
2248         rcu_read_unlock();
2249         return err;
2250 }
2251
2252 static int ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
2253                             u32 portid, u32 seq, struct mfc_cache *c, int cmd)
2254 {
2255         struct nlmsghdr *nlh;
2256         struct rtmsg *rtm;
2257         int err;
2258
2259         nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rtm), NLM_F_MULTI);
2260         if (nlh == NULL)
2261                 return -EMSGSIZE;
2262
2263         rtm = nlmsg_data(nlh);
2264         rtm->rtm_family   = RTNL_FAMILY_IPMR;
2265         rtm->rtm_dst_len  = 32;
2266         rtm->rtm_src_len  = 32;
2267         rtm->rtm_tos      = 0;
2268         rtm->rtm_table    = mrt->id;
2269         if (nla_put_u32(skb, RTA_TABLE, mrt->id))
2270                 goto nla_put_failure;
2271         rtm->rtm_type     = RTN_MULTICAST;
2272         rtm->rtm_scope    = RT_SCOPE_UNIVERSE;
2273         if (c->mfc_flags & MFC_STATIC)
2274                 rtm->rtm_protocol = RTPROT_STATIC;
2275         else
2276                 rtm->rtm_protocol = RTPROT_MROUTED;
2277         rtm->rtm_flags    = 0;
2278
2279         if (nla_put_be32(skb, RTA_SRC, c->mfc_origin) ||
2280             nla_put_be32(skb, RTA_DST, c->mfc_mcastgrp))
2281                 goto nla_put_failure;
2282         err = __ipmr_fill_mroute(mrt, skb, c, rtm);
2283         /* do not break the dump if cache is unresolved */
2284         if (err < 0 && err != -ENOENT)
2285                 goto nla_put_failure;
2286
2287         return nlmsg_end(skb, nlh);
2288
2289 nla_put_failure:
2290         nlmsg_cancel(skb, nlh);
2291         return -EMSGSIZE;
2292 }
2293
2294 static size_t mroute_msgsize(bool unresolved, int maxvif)
2295 {
2296         size_t len =
2297                 NLMSG_ALIGN(sizeof(struct rtmsg))
2298                 + nla_total_size(4)     /* RTA_TABLE */
2299                 + nla_total_size(4)     /* RTA_SRC */
2300                 + nla_total_size(4)     /* RTA_DST */
2301                 ;
2302
2303         if (!unresolved)
2304                 len = len
2305                       + nla_total_size(4)       /* RTA_IIF */
2306                       + nla_total_size(0)       /* RTA_MULTIPATH */
2307                       + maxvif * NLA_ALIGN(sizeof(struct rtnexthop))
2308                                                 /* RTA_MFC_STATS */
2309                       + nla_total_size(sizeof(struct rta_mfc_stats))
2310                 ;
2311
2312         return len;
2313 }
2314
2315 static void mroute_netlink_event(struct mr_table *mrt, struct mfc_cache *mfc,
2316                                  int cmd)
2317 {
2318         struct net *net = read_pnet(&mrt->net);
2319         struct sk_buff *skb;
2320         int err = -ENOBUFS;
2321
2322         skb = nlmsg_new(mroute_msgsize(mfc->mfc_parent >= MAXVIFS, mrt->maxvif),
2323                         GFP_ATOMIC);
2324         if (skb == NULL)
2325                 goto errout;
2326
2327         err = ipmr_fill_mroute(mrt, skb, 0, 0, mfc, cmd);
2328         if (err < 0)
2329                 goto errout;
2330
2331         rtnl_notify(skb, net, 0, RTNLGRP_IPV4_MROUTE, NULL, GFP_ATOMIC);
2332         return;
2333
2334 errout:
2335         kfree_skb(skb);
2336         if (err < 0)
2337                 rtnl_set_sk_err(net, RTNLGRP_IPV4_MROUTE, err);
2338 }
2339
2340 static int ipmr_rtm_dumproute(struct sk_buff *skb, struct netlink_callback *cb)
2341 {
2342         struct net *net = sock_net(skb->sk);
2343         struct mr_table *mrt;
2344         struct mfc_cache *mfc;
2345         unsigned int t = 0, s_t;
2346         unsigned int h = 0, s_h;
2347         unsigned int e = 0, s_e;
2348
2349         s_t = cb->args[0];
2350         s_h = cb->args[1];
2351         s_e = cb->args[2];
2352
2353         rcu_read_lock();
2354         ipmr_for_each_table(mrt, net) {
2355                 if (t < s_t)
2356                         goto next_table;
2357                 if (t > s_t)
2358                         s_h = 0;
2359                 for (h = s_h; h < MFC_LINES; h++) {
2360                         list_for_each_entry_rcu(mfc, &mrt->mfc_cache_array[h], list) {
2361                                 if (e < s_e)
2362                                         goto next_entry;
2363                                 if (ipmr_fill_mroute(mrt, skb,
2364                                                      NETLINK_CB(cb->skb).portid,
2365                                                      cb->nlh->nlmsg_seq,
2366                                                      mfc, RTM_NEWROUTE) < 0)
2367                                         goto done;
2368 next_entry:
2369                                 e++;
2370                         }
2371                         e = s_e = 0;
2372                 }
2373                 spin_lock_bh(&mfc_unres_lock);
2374                 list_for_each_entry(mfc, &mrt->mfc_unres_queue, list) {
2375                         if (e < s_e)
2376                                 goto next_entry2;
2377                         if (ipmr_fill_mroute(mrt, skb,
2378                                              NETLINK_CB(cb->skb).portid,
2379                                              cb->nlh->nlmsg_seq,
2380                                              mfc, RTM_NEWROUTE) < 0) {
2381                                 spin_unlock_bh(&mfc_unres_lock);
2382                                 goto done;
2383                         }
2384 next_entry2:
2385                         e++;
2386                 }
2387                 spin_unlock_bh(&mfc_unres_lock);
2388                 e = s_e = 0;
2389                 s_h = 0;
2390 next_table:
2391                 t++;
2392         }
2393 done:
2394         rcu_read_unlock();
2395
2396         cb->args[2] = e;
2397         cb->args[1] = h;
2398         cb->args[0] = t;
2399
2400         return skb->len;
2401 }
2402
2403 #ifdef CONFIG_PROC_FS
2404 /*
2405  *      The /proc interfaces to multicast routing :
2406  *      /proc/net/ip_mr_cache & /proc/net/ip_mr_vif
2407  */
2408 struct ipmr_vif_iter {
2409         struct seq_net_private p;
2410         struct mr_table *mrt;
2411         int ct;
2412 };
2413
2414 static struct vif_device *ipmr_vif_seq_idx(struct net *net,
2415                                            struct ipmr_vif_iter *iter,
2416                                            loff_t pos)
2417 {
2418         struct mr_table *mrt = iter->mrt;
2419
2420         for (iter->ct = 0; iter->ct < mrt->maxvif; ++iter->ct) {
2421                 if (!VIF_EXISTS(mrt, iter->ct))
2422                         continue;
2423                 if (pos-- == 0)
2424                         return &mrt->vif_table[iter->ct];
2425         }
2426         return NULL;
2427 }
2428
2429 static void *ipmr_vif_seq_start(struct seq_file *seq, loff_t *pos)
2430         __acquires(mrt_lock)
2431 {
2432         struct ipmr_vif_iter *iter = seq->private;
2433         struct net *net = seq_file_net(seq);
2434         struct mr_table *mrt;
2435
2436         mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2437         if (mrt == NULL)
2438                 return ERR_PTR(-ENOENT);
2439
2440         iter->mrt = mrt;
2441
2442         read_lock(&mrt_lock);
2443         return *pos ? ipmr_vif_seq_idx(net, seq->private, *pos - 1)
2444                 : SEQ_START_TOKEN;
2445 }
2446
2447 static void *ipmr_vif_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2448 {
2449         struct ipmr_vif_iter *iter = seq->private;
2450         struct net *net = seq_file_net(seq);
2451         struct mr_table *mrt = iter->mrt;
2452
2453         ++*pos;
2454         if (v == SEQ_START_TOKEN)
2455                 return ipmr_vif_seq_idx(net, iter, 0);
2456
2457         while (++iter->ct < mrt->maxvif) {
2458                 if (!VIF_EXISTS(mrt, iter->ct))
2459                         continue;
2460                 return &mrt->vif_table[iter->ct];
2461         }
2462         return NULL;
2463 }
2464
2465 static void ipmr_vif_seq_stop(struct seq_file *seq, void *v)
2466         __releases(mrt_lock)
2467 {
2468         read_unlock(&mrt_lock);
2469 }
2470
2471 static int ipmr_vif_seq_show(struct seq_file *seq, void *v)
2472 {
2473         struct ipmr_vif_iter *iter = seq->private;
2474         struct mr_table *mrt = iter->mrt;
2475
2476         if (v == SEQ_START_TOKEN) {
2477                 seq_puts(seq,
2478                          "Interface      BytesIn  PktsIn  BytesOut PktsOut Flags Local    Remote\n");
2479         } else {
2480                 const struct vif_device *vif = v;
2481                 const char *name =  vif->dev ? vif->dev->name : "none";
2482
2483                 seq_printf(seq,
2484                            "%2Zd %-10s %8ld %7ld  %8ld %7ld %05X %08X %08X\n",
2485                            vif - mrt->vif_table,
2486                            name, vif->bytes_in, vif->pkt_in,
2487                            vif->bytes_out, vif->pkt_out,
2488                            vif->flags, vif->local, vif->remote);
2489         }
2490         return 0;
2491 }
2492
2493 static const struct seq_operations ipmr_vif_seq_ops = {
2494         .start = ipmr_vif_seq_start,
2495         .next  = ipmr_vif_seq_next,
2496         .stop  = ipmr_vif_seq_stop,
2497         .show  = ipmr_vif_seq_show,
2498 };
2499
2500 static int ipmr_vif_open(struct inode *inode, struct file *file)
2501 {
2502         return seq_open_net(inode, file, &ipmr_vif_seq_ops,
2503                             sizeof(struct ipmr_vif_iter));
2504 }
2505
2506 static const struct file_operations ipmr_vif_fops = {
2507         .owner   = THIS_MODULE,
2508         .open    = ipmr_vif_open,
2509         .read    = seq_read,
2510         .llseek  = seq_lseek,
2511         .release = seq_release_net,
2512 };
2513
2514 struct ipmr_mfc_iter {
2515         struct seq_net_private p;
2516         struct mr_table *mrt;
2517         struct list_head *cache;
2518         int ct;
2519 };
2520
2521
2522 static struct mfc_cache *ipmr_mfc_seq_idx(struct net *net,
2523                                           struct ipmr_mfc_iter *it, loff_t pos)
2524 {
2525         struct mr_table *mrt = it->mrt;
2526         struct mfc_cache *mfc;
2527
2528         rcu_read_lock();
2529         for (it->ct = 0; it->ct < MFC_LINES; it->ct++) {
2530                 it->cache = &mrt->mfc_cache_array[it->ct];
2531                 list_for_each_entry_rcu(mfc, it->cache, list)
2532                         if (pos-- == 0)
2533                                 return mfc;
2534         }
2535         rcu_read_unlock();
2536
2537         spin_lock_bh(&mfc_unres_lock);
2538         it->cache = &mrt->mfc_unres_queue;
2539         list_for_each_entry(mfc, it->cache, list)
2540                 if (pos-- == 0)
2541                         return mfc;
2542         spin_unlock_bh(&mfc_unres_lock);
2543
2544         it->cache = NULL;
2545         return NULL;
2546 }
2547
2548
2549 static void *ipmr_mfc_seq_start(struct seq_file *seq, loff_t *pos)
2550 {
2551         struct ipmr_mfc_iter *it = seq->private;
2552         struct net *net = seq_file_net(seq);
2553         struct mr_table *mrt;
2554
2555         mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2556         if (mrt == NULL)
2557                 return ERR_PTR(-ENOENT);
2558
2559         it->mrt = mrt;
2560         it->cache = NULL;
2561         it->ct = 0;
2562         return *pos ? ipmr_mfc_seq_idx(net, seq->private, *pos - 1)
2563                 : SEQ_START_TOKEN;
2564 }
2565
2566 static void *ipmr_mfc_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2567 {
2568         struct mfc_cache *mfc = v;
2569         struct ipmr_mfc_iter *it = seq->private;
2570         struct net *net = seq_file_net(seq);
2571         struct mr_table *mrt = it->mrt;
2572
2573         ++*pos;
2574
2575         if (v == SEQ_START_TOKEN)
2576                 return ipmr_mfc_seq_idx(net, seq->private, 0);
2577
2578         if (mfc->list.next != it->cache)
2579                 return list_entry(mfc->list.next, struct mfc_cache, list);
2580
2581         if (it->cache == &mrt->mfc_unres_queue)
2582                 goto end_of_list;
2583
2584         BUG_ON(it->cache != &mrt->mfc_cache_array[it->ct]);
2585
2586         while (++it->ct < MFC_LINES) {
2587                 it->cache = &mrt->mfc_cache_array[it->ct];
2588                 if (list_empty(it->cache))
2589                         continue;
2590                 return list_first_entry(it->cache, struct mfc_cache, list);
2591         }
2592
2593         /* exhausted cache_array, show unresolved */
2594         rcu_read_unlock();
2595         it->cache = &mrt->mfc_unres_queue;
2596         it->ct = 0;
2597
2598         spin_lock_bh(&mfc_unres_lock);
2599         if (!list_empty(it->cache))
2600                 return list_first_entry(it->cache, struct mfc_cache, list);
2601
2602 end_of_list:
2603         spin_unlock_bh(&mfc_unres_lock);
2604         it->cache = NULL;
2605
2606         return NULL;
2607 }
2608
2609 static void ipmr_mfc_seq_stop(struct seq_file *seq, void *v)
2610 {
2611         struct ipmr_mfc_iter *it = seq->private;
2612         struct mr_table *mrt = it->mrt;
2613
2614         if (it->cache == &mrt->mfc_unres_queue)
2615                 spin_unlock_bh(&mfc_unres_lock);
2616         else if (it->cache == &mrt->mfc_cache_array[it->ct])
2617                 rcu_read_unlock();
2618 }
2619
2620 static int ipmr_mfc_seq_show(struct seq_file *seq, void *v)
2621 {
2622         int n;
2623
2624         if (v == SEQ_START_TOKEN) {
2625                 seq_puts(seq,
2626                  "Group    Origin   Iif     Pkts    Bytes    Wrong Oifs\n");
2627         } else {
2628                 const struct mfc_cache *mfc = v;
2629                 const struct ipmr_mfc_iter *it = seq->private;
2630                 const struct mr_table *mrt = it->mrt;
2631
2632                 seq_printf(seq, "%08X %08X %-3hd",
2633                            (__force u32) mfc->mfc_mcastgrp,
2634                            (__force u32) mfc->mfc_origin,
2635                            mfc->mfc_parent);
2636
2637                 if (it->cache != &mrt->mfc_unres_queue) {
2638                         seq_printf(seq, " %8lu %8lu %8lu",
2639                                    mfc->mfc_un.res.pkt,
2640                                    mfc->mfc_un.res.bytes,
2641                                    mfc->mfc_un.res.wrong_if);
2642                         for (n = mfc->mfc_un.res.minvif;
2643                              n < mfc->mfc_un.res.maxvif; n++) {
2644                                 if (VIF_EXISTS(mrt, n) &&
2645                                     mfc->mfc_un.res.ttls[n] < 255)
2646                                         seq_printf(seq,
2647                                            " %2d:%-3d",
2648                                            n, mfc->mfc_un.res.ttls[n]);
2649                         }
2650                 } else {
2651                         /* unresolved mfc_caches don't contain
2652                          * pkt, bytes and wrong_if values
2653                          */
2654                         seq_printf(seq, " %8lu %8lu %8lu", 0ul, 0ul, 0ul);
2655                 }
2656                 seq_putc(seq, '\n');
2657         }
2658         return 0;
2659 }
2660
2661 static const struct seq_operations ipmr_mfc_seq_ops = {
2662         .start = ipmr_mfc_seq_start,
2663         .next  = ipmr_mfc_seq_next,
2664         .stop  = ipmr_mfc_seq_stop,
2665         .show  = ipmr_mfc_seq_show,
2666 };
2667
2668 static int ipmr_mfc_open(struct inode *inode, struct file *file)
2669 {
2670         return seq_open_net(inode, file, &ipmr_mfc_seq_ops,
2671                             sizeof(struct ipmr_mfc_iter));
2672 }
2673
2674 static const struct file_operations ipmr_mfc_fops = {
2675         .owner   = THIS_MODULE,
2676         .open    = ipmr_mfc_open,
2677         .read    = seq_read,
2678         .llseek  = seq_lseek,
2679         .release = seq_release_net,
2680 };
2681 #endif
2682
2683 #ifdef CONFIG_IP_PIMSM_V2
2684 static const struct net_protocol pim_protocol = {
2685         .handler        =       pim_rcv,
2686         .netns_ok       =       1,
2687 };
2688 #endif
2689
2690
2691 /*
2692  *      Setup for IP multicast routing
2693  */
2694 static int __net_init ipmr_net_init(struct net *net)
2695 {
2696         int err;
2697
2698         err = ipmr_rules_init(net);
2699         if (err < 0)
2700                 goto fail;
2701
2702 #ifdef CONFIG_PROC_FS
2703         err = -ENOMEM;
2704         if (!proc_create("ip_mr_vif", 0, net->proc_net, &ipmr_vif_fops))
2705                 goto proc_vif_fail;
2706         if (!proc_create("ip_mr_cache", 0, net->proc_net, &ipmr_mfc_fops))
2707                 goto proc_cache_fail;
2708 #endif
2709         return 0;
2710
2711 #ifdef CONFIG_PROC_FS
2712 proc_cache_fail:
2713         remove_proc_entry("ip_mr_vif", net->proc_net);
2714 proc_vif_fail:
2715         ipmr_rules_exit(net);
2716 #endif
2717 fail:
2718         return err;
2719 }
2720
2721 static void __net_exit ipmr_net_exit(struct net *net)
2722 {
2723 #ifdef CONFIG_PROC_FS
2724         remove_proc_entry("ip_mr_cache", net->proc_net);
2725         remove_proc_entry("ip_mr_vif", net->proc_net);
2726 #endif
2727         ipmr_rules_exit(net);
2728 }
2729
2730 static struct pernet_operations ipmr_net_ops = {
2731         .init = ipmr_net_init,
2732         .exit = ipmr_net_exit,
2733 };
2734
2735 int __init ip_mr_init(void)
2736 {
2737         int err;
2738
2739         mrt_cachep = kmem_cache_create("ip_mrt_cache",
2740                                        sizeof(struct mfc_cache),
2741                                        0, SLAB_HWCACHE_ALIGN | SLAB_PANIC,
2742                                        NULL);
2743         if (!mrt_cachep)
2744                 return -ENOMEM;
2745
2746         err = register_pernet_subsys(&ipmr_net_ops);
2747         if (err)
2748                 goto reg_pernet_fail;
2749
2750         err = register_netdevice_notifier(&ip_mr_notifier);
2751         if (err)
2752                 goto reg_notif_fail;
2753 #ifdef CONFIG_IP_PIMSM_V2
2754         if (inet_add_protocol(&pim_protocol, IPPROTO_PIM) < 0) {
2755                 pr_err("%s: can't add PIM protocol\n", __func__);
2756                 err = -EAGAIN;
2757                 goto add_proto_fail;
2758         }
2759 #endif
2760         rtnl_register(RTNL_FAMILY_IPMR, RTM_GETROUTE,
2761                       NULL, ipmr_rtm_dumproute, NULL);
2762         return 0;
2763
2764 #ifdef CONFIG_IP_PIMSM_V2
2765 add_proto_fail:
2766         unregister_netdevice_notifier(&ip_mr_notifier);
2767 #endif
2768 reg_notif_fail:
2769         unregister_pernet_subsys(&ipmr_net_ops);
2770 reg_pernet_fail:
2771         kmem_cache_destroy(mrt_cachep);
2772         return err;
2773 }