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         possible_net_t          net;
  77         u32                     id;
  78         struct sock __rcu       *mroute_sk;
  79         struct timer_list       ipmr_expire_timer;
  80         struct list_head        mfc_unres_queue;
  81         struct list_head        mfc_cache_array[MFC_LINES];
  82         struct vif_device       vif_table[MAXVIFS];
  83         int                     maxvif;
  84         atomic_t                cache_resolve_queue_len;
  85         bool                    mroute_do_assert;
  86         bool                    mroute_do_pim;
  87 #if defined(CONFIG_IP_PIMSM_V1) || defined(CONFIG_IP_PIMSM_V2)
  88         int                     mroute_reg_vif_num;
  89 #endif
  90 };
  91
  92 struct ipmr_rule {
  93         struct fib_rule         common;
  94 };
  95
  96 struct ipmr_result {
  97         struct mr_table         *mrt;
  98 };
  99
 100 /* Big lock, protecting vif table, mrt cache and mroute socket state.
 101  * Note that the changes are semaphored via rtnl_lock.
 102  */
 103
 104 static DEFINE_RWLOCK(mrt_lock);
 105
 106 /*
 107  *      Multicast router control variables
 108  */
 109
 110 #define VIF_EXISTS(_mrt, _idx) ((_mrt)->vif_table[_idx].dev != NULL)
 111
 112 /* Special spinlock for queue of unresolved entries */
 113 static DEFINE_SPINLOCK(mfc_unres_lock);
 114
 115 /* We return to original Alan's scheme. Hash table of resolved
 116  * entries is changed only in process context and protected
 117  * with weak lock mrt_lock. Queue of unresolved entries is protected
 118  * with strong spinlock mfc_unres_lock.
 119  *
 120  * In this case data path is free of exclusive locks at all.
 121  */
 122
 123 static struct kmem_cache *mrt_cachep __read_mostly;
 124
 125 static struct mr_table *ipmr_new_table(struct net *net, u32 id);
 126 static void ipmr_free_table(struct mr_table *mrt);
 127
 128 static void ip_mr_forward(struct net *net, struct mr_table *mrt,
 129                           struct sk_buff *skb, struct mfc_cache *cache,
 130                           int local);
 131 static int ipmr_cache_report(struct mr_table *mrt,
 132                              struct sk_buff *pkt, vifi_t vifi, int assert);
 133 static int __ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
 134                               struct mfc_cache *c, struct rtmsg *rtm);
 135 static void mroute_netlink_event(struct mr_table *mrt, struct mfc_cache *mfc,
 136                                  int cmd);
 137 static void mroute_clean_tables(struct mr_table *mrt, bool all);
 138 static void ipmr_expire_process(unsigned long arg);
 139
 140 #ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES
 141 #define ipmr_for_each_table(mrt, net) \
 142         list_for_each_entry_rcu(mrt, &net->ipv4.mr_tables, list)
 143
 144 static struct mr_table *ipmr_get_table(struct net *net, u32 id)
 145 {
 146         struct mr_table *mrt;
 147
 148         ipmr_for_each_table(mrt, net) {
 149                 if (mrt->id == id)
 150                         return mrt;
 151         }
 152         return NULL;
 153 }
 154
 155 static int ipmr_fib_lookup(struct net *net, struct flowi4 *flp4,
 156                            struct mr_table **mrt)
 157 {
 158         int err;
 159         struct ipmr_result res;
 160         struct fib_lookup_arg arg = {
 161                 .result = &res,
 162                 .flags = FIB_LOOKUP_NOREF,
 163         };
 164
 165         err = fib_rules_lookup(net->ipv4.mr_rules_ops,
 166                                flowi4_to_flowi(flp4), 0, &arg);
 167         if (err < 0)
 168                 return err;
 169         *mrt = res.mrt;
 170         return 0;
 171 }
 172
 173 static int ipmr_rule_action(struct fib_rule *rule, struct flowi *flp,
 174                             int flags, struct fib_lookup_arg *arg)
 175 {
 176         struct ipmr_result *res = arg->result;
 177         struct mr_table *mrt;
 178
 179         switch (rule->action) {
 180         case FR_ACT_TO_TBL:
 181                 break;
 182         case FR_ACT_UNREACHABLE:
 183                 return -ENETUNREACH;
 184         case FR_ACT_PROHIBIT:
 185                 return -EACCES;
 186         case FR_ACT_BLACKHOLE:
 187         default:
 188                 return -EINVAL;
 189         }
 190
 191         mrt = ipmr_get_table(rule->fr_net, rule->table);
 192         if (!mrt)
 193                 return -EAGAIN;
 194         res->mrt = mrt;
 195         return 0;
 196 }
 197
 198 static int ipmr_rule_match(struct fib_rule *rule, struct flowi *fl, int flags)
 199 {
 200         return 1;
 201 }
 202
 203 static const struct nla_policy ipmr_rule_policy[FRA_MAX + 1] = {
 204         FRA_GENERIC_POLICY,
 205 };
 206
 207 static int ipmr_rule_configure(struct fib_rule *rule, struct sk_buff *skb,
 208                                struct fib_rule_hdr *frh, struct nlattr **tb)
 209 {
 210         return 0;
 211 }
 212
 213 static int ipmr_rule_compare(struct fib_rule *rule, struct fib_rule_hdr *frh,
 214                              struct nlattr **tb)
 215 {
 216         return 1;
 217 }
 218
 219 static int ipmr_rule_fill(struct fib_rule *rule, struct sk_buff *skb,
 220                           struct fib_rule_hdr *frh)
 221 {
 222         frh->dst_len = 0;
 223         frh->src_len = 0;
 224         frh->tos     = 0;
 225         return 0;
 226 }
 227
 228 static const struct fib_rules_ops __net_initconst ipmr_rules_ops_template = {
 229         .family         = RTNL_FAMILY_IPMR,
 230         .rule_size      = sizeof(struct ipmr_rule),
 231         .addr_size      = sizeof(u32),
 232         .action         = ipmr_rule_action,
 233         .match          = ipmr_rule_match,
 234         .configure      = ipmr_rule_configure,
 235         .compare        = ipmr_rule_compare,
 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) {
 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         ipmr_free_table(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         rtnl_lock();
 279         list_for_each_entry_safe(mrt, next, &net->ipv4.mr_tables, list) {
 280                 list_del(&mrt->list);
 281                 ipmr_free_table(mrt);
 282         }
 283         fib_rules_unregister(net->ipv4.mr_rules_ops);
 284         rtnl_unlock();
 285 }
 286 #else
 287 #define ipmr_for_each_table(mrt, net) \
 288         for (mrt = net->ipv4.mrt; mrt; mrt = NULL)
 289
 290 static struct mr_table *ipmr_get_table(struct net *net, u32 id)
 291 {
 292         return net->ipv4.mrt;
 293 }
 294
 295 static int ipmr_fib_lookup(struct net *net, struct flowi4 *flp4,
 296                            struct mr_table **mrt)
 297 {
 298         *mrt = net->ipv4.mrt;
 299         return 0;
 300 }
 301
 302 static int __net_init ipmr_rules_init(struct net *net)
 303 {
 304         net->ipv4.mrt = ipmr_new_table(net, RT_TABLE_DEFAULT);
 305         return net->ipv4.mrt ? 0 : -ENOMEM;
 306 }
 307
 308 static void __net_exit ipmr_rules_exit(struct net *net)
 309 {
 310         rtnl_lock();
 311         ipmr_free_table(net->ipv4.mrt);
 312         net->ipv4.mrt = NULL;
 313         rtnl_unlock();
 314 }
 315 #endif
 316
 317 static struct mr_table *ipmr_new_table(struct net *net, u32 id)
 318 {
 319         struct mr_table *mrt;
 320         unsigned int i;
 321
 322         mrt = ipmr_get_table(net, id);
 323         if (mrt)
 324                 return mrt;
 325
 326         mrt = kzalloc(sizeof(*mrt), GFP_KERNEL);
 327         if (!mrt)
 328                 return NULL;
 329         write_pnet(&mrt->net, net);
 330         mrt->id = id;
 331
 332         /* Forwarding cache */
 333         for (i = 0; i < MFC_LINES; i++)
 334                 INIT_LIST_HEAD(&mrt->mfc_cache_array[i]);
 335
 336         INIT_LIST_HEAD(&mrt->mfc_unres_queue);
 337
 338         setup_timer(&mrt->ipmr_expire_timer, ipmr_expire_process,
 339                     (unsigned long)mrt);
 340
 341 #ifdef CONFIG_IP_PIMSM
 342         mrt->mroute_reg_vif_num = -1;
 343 #endif
 344 #ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES
 345         list_add_tail_rcu(&mrt->list, &net->ipv4.mr_tables);
 346 #endif
 347         return mrt;
 348 }
 349
 350 static void ipmr_free_table(struct mr_table *mrt)
 351 {
 352         del_timer_sync(&mrt->ipmr_expire_timer);
 353         mroute_clean_tables(mrt, true);
 354         kfree(mrt);
 355 }
 356
 357 /* Service routines creating virtual interfaces: DVMRP tunnels and PIMREG */
 358
 359 static void ipmr_del_tunnel(struct net_device *dev, struct vifctl *v)
 360 {
 361         struct net *net = dev_net(dev);
 362
 363         dev_close(dev);
 364
 365         dev = __dev_get_by_name(net, "tunl0");
 366         if (dev) {
 367                 const struct net_device_ops *ops = dev->netdev_ops;
 368                 struct ifreq ifr;
 369                 struct ip_tunnel_parm p;
 370
 371                 memset(&p, 0, sizeof(p));
 372                 p.iph.daddr = v->vifc_rmt_addr.s_addr;
 373                 p.iph.saddr = v->vifc_lcl_addr.s_addr;
 374                 p.iph.version = 4;
 375                 p.iph.ihl = 5;
 376                 p.iph.protocol = IPPROTO_IPIP;
 377                 sprintf(p.name, "dvmrp%d", v->vifc_vifi);
 378                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
 379
 380                 if (ops->ndo_do_ioctl) {
 381                         mm_segment_t oldfs = get_fs();
 382
 383                         set_fs(KERNEL_DS);
 384                         ops->ndo_do_ioctl(dev, &ifr, SIOCDELTUNNEL);
 385                         set_fs(oldfs);
 386                 }
 387         }
 388 }
 389
 390 static
 391 struct net_device *ipmr_new_tunnel(struct net *net, struct vifctl *v)
 392 {
 393         struct net_device  *dev;
 394
 395         dev = __dev_get_by_name(net, "tunl0");
 396
 397         if (dev) {
 398                 const struct net_device_ops *ops = dev->netdev_ops;
 399                 int err;
 400                 struct ifreq ifr;
 401                 struct ip_tunnel_parm p;
 402                 struct in_device  *in_dev;
 403
 404                 memset(&p, 0, sizeof(p));
 405                 p.iph.daddr = v->vifc_rmt_addr.s_addr;
 406                 p.iph.saddr = v->vifc_lcl_addr.s_addr;
 407                 p.iph.version = 4;
 408                 p.iph.ihl = 5;
 409                 p.iph.protocol = IPPROTO_IPIP;
 410                 sprintf(p.name, "dvmrp%d", v->vifc_vifi);
 411                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
 412
 413                 if (ops->ndo_do_ioctl) {
 414                         mm_segment_t oldfs = get_fs();
 415
 416                         set_fs(KERNEL_DS);
 417                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
 418                         set_fs(oldfs);
 419                 } else {
 420                         err = -EOPNOTSUPP;
 421                 }
 422                 dev = NULL;
 423
 424                 if (err == 0 &&
 425                     (dev = __dev_get_by_name(net, p.name)) != NULL) {
 426                         dev->flags |= IFF_MULTICAST;
 427
 428                         in_dev = __in_dev_get_rtnl(dev);
 429                         if (!in_dev)
 430                                 goto failure;
 431
 432                         ipv4_devconf_setall(in_dev);
 433                         neigh_parms_data_state_setall(in_dev->arp_parms);
 434                         IPV4_DEVCONF(in_dev->cnf, RP_FILTER) = 0;
 435
 436                         if (dev_open(dev))
 437                                 goto failure;
 438                         dev_hold(dev);
 439                 }
 440         }
 441         return dev;
 442
 443 failure:
 444         unregister_netdevice(dev);
 445         return NULL;
 446 }
 447
 448 #ifdef CONFIG_IP_PIMSM
 449
 450 static netdev_tx_t reg_vif_xmit(struct sk_buff *skb, struct net_device *dev)
 451 {
 452         struct net *net = dev_net(dev);
 453         struct mr_table *mrt;
 454         struct flowi4 fl4 = {
 455                 .flowi4_oif     = dev->ifindex,
 456                 .flowi4_iif     = skb->skb_iif ? : LOOPBACK_IFINDEX,
 457                 .flowi4_mark    = skb->mark,
 458         };
 459         int err;
 460
 461         err = ipmr_fib_lookup(net, &fl4, &mrt);
 462         if (err < 0) {
 463                 kfree_skb(skb);
 464                 return err;
 465         }
 466
 467         read_lock(&mrt_lock);
 468         dev->stats.tx_bytes += skb->len;
 469         dev->stats.tx_packets++;
 470         ipmr_cache_report(mrt, skb, mrt->mroute_reg_vif_num, IGMPMSG_WHOLEPKT);
 471         read_unlock(&mrt_lock);
 472         kfree_skb(skb);
 473         return NETDEV_TX_OK;
 474 }
 475
 476 static int reg_vif_get_iflink(const struct net_device *dev)
 477 {
 478         return 0;
 479 }
 480
 481 static const struct net_device_ops reg_vif_netdev_ops = {
 482         .ndo_start_xmit = reg_vif_xmit,
 483         .ndo_get_iflink = reg_vif_get_iflink,
 484 };
 485
 486 static void reg_vif_setup(struct net_device *dev)
 487 {
 488         dev->type               = ARPHRD_PIMREG;
 489         dev->mtu                = ETH_DATA_LEN - sizeof(struct iphdr) - 8;
 490         dev->flags              = IFF_NOARP;
 491         dev->netdev_ops         = &reg_vif_netdev_ops;
 492         dev->destructor         = free_netdev;
 493         dev->features           |= NETIF_F_NETNS_LOCAL;
 494 }
 495
 496 static struct net_device *ipmr_reg_vif(struct net *net, struct mr_table *mrt)
 497 {
 498         struct net_device *dev;
 499         struct in_device *in_dev;
 500         char name[IFNAMSIZ];
 501
 502         if (mrt->id == RT_TABLE_DEFAULT)
 503                 sprintf(name, "pimreg");
 504         else
 505                 sprintf(name, "pimreg%u", mrt->id);
 506
 507         dev = alloc_netdev(0, name, NET_NAME_UNKNOWN, reg_vif_setup);
 508
 509         if (!dev)
 510                 return NULL;
 511
 512         dev_net_set(dev, net);
 513
 514         if (register_netdevice(dev)) {
 515                 free_netdev(dev);
 516                 return NULL;
 517         }
 518
 519         rcu_read_lock();
 520         in_dev = __in_dev_get_rcu(dev);
 521         if (!in_dev) {
 522                 rcu_read_unlock();
 523                 goto failure;
 524         }
 525
 526         ipv4_devconf_setall(in_dev);
 527         neigh_parms_data_state_setall(in_dev->arp_parms);
 528         IPV4_DEVCONF(in_dev->cnf, RP_FILTER) = 0;
 529         rcu_read_unlock();
 530
 531         if (dev_open(dev))
 532                 goto failure;
 533
 534         dev_hold(dev);
 535
 536         return dev;
 537
 538 failure:
 539         unregister_netdevice(dev);
 540         return NULL;
 541 }
 542 #endif
 543
 544 /**
 545  *      vif_delete - Delete a VIF entry
 546  *      @notify: Set to 1, if the caller is a notifier_call
 547  */
 548
 549 static int vif_delete(struct mr_table *mrt, int vifi, int notify,
 550                       struct list_head *head)
 551 {
 552         struct vif_device *v;
 553         struct net_device *dev;
 554         struct in_device *in_dev;
 555
 556         if (vifi < 0 || vifi >= mrt->maxvif)
 557                 return -EADDRNOTAVAIL;
 558
 559         v = &mrt->vif_table[vifi];
 560
 561         write_lock_bh(&mrt_lock);
 562         dev = v->dev;
 563         v->dev = NULL;
 564
 565         if (!dev) {
 566                 write_unlock_bh(&mrt_lock);
 567                 return -EADDRNOTAVAIL;
 568         }
 569
 570 #ifdef CONFIG_IP_PIMSM
 571         if (vifi == mrt->mroute_reg_vif_num)
 572                 mrt->mroute_reg_vif_num = -1;
 573 #endif
 574
 575         if (vifi + 1 == mrt->maxvif) {
 576                 int tmp;
 577
 578                 for (tmp = vifi - 1; tmp >= 0; tmp--) {
 579                         if (VIF_EXISTS(mrt, tmp))
 580                                 break;
 581                 }
 582                 mrt->maxvif = tmp+1;
 583         }
 584
 585         write_unlock_bh(&mrt_lock);
 586
 587         dev_set_allmulti(dev, -1);
 588
 589         in_dev = __in_dev_get_rtnl(dev);
 590         if (in_dev) {
 591                 IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)--;
 592                 inet_netconf_notify_devconf(dev_net(dev),
 593                                             NETCONFA_MC_FORWARDING,
 594                                             dev->ifindex, &in_dev->cnf);
 595                 ip_rt_multicast_event(in_dev);
 596         }
 597
 598         if (v->flags & (VIFF_TUNNEL | VIFF_REGISTER) && !notify)
 599                 unregister_netdevice_queue(dev, head);
 600
 601         dev_put(dev);
 602         return 0;
 603 }
 604
 605 static void ipmr_cache_free_rcu(struct rcu_head *head)
 606 {
 607         struct mfc_cache *c = container_of(head, struct mfc_cache, rcu);
 608
 609         kmem_cache_free(mrt_cachep, c);
 610 }
 611
 612 static inline void ipmr_cache_free(struct mfc_cache *c)
 613 {
 614         call_rcu(&c->rcu, ipmr_cache_free_rcu);
 615 }
 616
 617 /* Destroy an unresolved cache entry, killing queued skbs
 618  * and reporting error to netlink readers.
 619  */
 620
 621 static void ipmr_destroy_unres(struct mr_table *mrt, struct mfc_cache *c)
 622 {
 623         struct net *net = read_pnet(&mrt->net);
 624         struct sk_buff *skb;
 625         struct nlmsgerr *e;
 626
 627         atomic_dec(&mrt->cache_resolve_queue_len);
 628
 629         while ((skb = skb_dequeue(&c->mfc_un.unres.unresolved))) {
 630                 if (ip_hdr(skb)->version == 0) {
 631                         struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct iphdr));
 632                         nlh->nlmsg_type = NLMSG_ERROR;
 633                         nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
 634                         skb_trim(skb, nlh->nlmsg_len);
 635                         e = nlmsg_data(nlh);
 636                         e->error = -ETIMEDOUT;
 637                         memset(&e->msg, 0, sizeof(e->msg));
 638
 639                         rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
 640                 } else {
 641                         kfree_skb(skb);
 642                 }
 643         }
 644
 645         ipmr_cache_free(c);
 646 }
 647
 648
 649 /* Timer process for the unresolved queue. */
 650
 651 static void ipmr_expire_process(unsigned long arg)
 652 {
 653         struct mr_table *mrt = (struct mr_table *)arg;
 654         unsigned long now;
 655         unsigned long expires;
 656         struct mfc_cache *c, *next;
 657
 658         if (!spin_trylock(&mfc_unres_lock)) {
 659                 mod_timer(&mrt->ipmr_expire_timer, jiffies+HZ/10);
 660                 return;
 661         }
 662
 663         if (list_empty(&mrt->mfc_unres_queue))
 664                 goto out;
 665
 666         now = jiffies;
 667         expires = 10*HZ;
 668
 669         list_for_each_entry_safe(c, next, &mrt->mfc_unres_queue, list) {
 670                 if (time_after(c->mfc_un.unres.expires, now)) {
 671                         unsigned long interval = c->mfc_un.unres.expires - now;
 672                         if (interval < expires)
 673                                 expires = interval;
 674                         continue;
 675                 }
 676
 677                 list_del(&c->list);
 678                 mroute_netlink_event(mrt, c, RTM_DELROUTE);
 679                 ipmr_destroy_unres(mrt, c);
 680         }
 681
 682         if (!list_empty(&mrt->mfc_unres_queue))
 683                 mod_timer(&mrt->ipmr_expire_timer, jiffies + expires);
 684
 685 out:
 686         spin_unlock(&mfc_unres_lock);
 687 }
 688
 689 /* Fill oifs list. It is called under write locked mrt_lock. */
 690
 691 static void ipmr_update_thresholds(struct mr_table *mrt, struct mfc_cache *cache,
 692                                    unsigned char *ttls)
 693 {
 694         int vifi;
 695
 696         cache->mfc_un.res.minvif = MAXVIFS;
 697         cache->mfc_un.res.maxvif = 0;
 698         memset(cache->mfc_un.res.ttls, 255, MAXVIFS);
 699
 700         for (vifi = 0; vifi < mrt->maxvif; vifi++) {
 701                 if (VIF_EXISTS(mrt, vifi) &&
 702                     ttls[vifi] && ttls[vifi] < 255) {
 703                         cache->mfc_un.res.ttls[vifi] = ttls[vifi];
 704                         if (cache->mfc_un.res.minvif > vifi)
 705                                 cache->mfc_un.res.minvif = vifi;
 706                         if (cache->mfc_un.res.maxvif <= vifi)
 707                                 cache->mfc_un.res.maxvif = vifi + 1;
 708                 }
 709         }
 710 }
 711
 712 static int vif_add(struct net *net, struct mr_table *mrt,
 713                    struct vifctl *vifc, int mrtsock)
 714 {
 715         int vifi = vifc->vifc_vifi;
 716         struct vif_device *v = &mrt->vif_table[vifi];
 717         struct net_device *dev;
 718         struct in_device *in_dev;
 719         int err;
 720
 721         /* Is vif busy ? */
 722         if (VIF_EXISTS(mrt, vifi))
 723                 return -EADDRINUSE;
 724
 725         switch (vifc->vifc_flags) {
 726 #ifdef CONFIG_IP_PIMSM
 727         case VIFF_REGISTER:
 728                 /*
 729                  * Special Purpose VIF in PIM
 730                  * All the packets will be sent to the daemon
 731                  */
 732                 if (mrt->mroute_reg_vif_num >= 0)
 733                         return -EADDRINUSE;
 734                 dev = ipmr_reg_vif(net, mrt);
 735                 if (!dev)
 736                         return -ENOBUFS;
 737                 err = dev_set_allmulti(dev, 1);
 738                 if (err) {
 739                         unregister_netdevice(dev);
 740                         dev_put(dev);
 741                         return err;
 742                 }
 743                 break;
 744 #endif
 745         case VIFF_TUNNEL:
 746                 dev = ipmr_new_tunnel(net, vifc);
 747                 if (!dev)
 748                         return -ENOBUFS;
 749                 err = dev_set_allmulti(dev, 1);
 750                 if (err) {
 751                         ipmr_del_tunnel(dev, vifc);
 752                         dev_put(dev);
 753                         return err;
 754                 }
 755                 break;
 756
 757         case VIFF_USE_IFINDEX:
 758         case 0:
 759                 if (vifc->vifc_flags == VIFF_USE_IFINDEX) {
 760                         dev = dev_get_by_index(net, vifc->vifc_lcl_ifindex);
 761                         if (dev && !__in_dev_get_rtnl(dev)) {
 762                                 dev_put(dev);
 763                                 return -EADDRNOTAVAIL;
 764                         }
 765                 } else {
 766                         dev = ip_dev_find(net, vifc->vifc_lcl_addr.s_addr);
 767                 }
 768                 if (!dev)
 769                         return -EADDRNOTAVAIL;
 770                 err = dev_set_allmulti(dev, 1);
 771                 if (err) {
 772                         dev_put(dev);
 773                         return err;
 774                 }
 775                 break;
 776         default:
 777                 return -EINVAL;
 778         }
 779
 780         in_dev = __in_dev_get_rtnl(dev);
 781         if (!in_dev) {
 782                 dev_put(dev);
 783                 return -EADDRNOTAVAIL;
 784         }
 785         IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)++;
 786         inet_netconf_notify_devconf(net, NETCONFA_MC_FORWARDING, dev->ifindex,
 787                                     &in_dev->cnf);
 788         ip_rt_multicast_event(in_dev);
 789
 790         /* Fill in the VIF structures */
 791
 792         v->rate_limit = vifc->vifc_rate_limit;
 793         v->local = vifc->vifc_lcl_addr.s_addr;
 794         v->remote = vifc->vifc_rmt_addr.s_addr;
 795         v->flags = vifc->vifc_flags;
 796         if (!mrtsock)
 797                 v->flags |= VIFF_STATIC;
 798         v->threshold = vifc->vifc_threshold;
 799         v->bytes_in = 0;
 800         v->bytes_out = 0;
 801         v->pkt_in = 0;
 802         v->pkt_out = 0;
 803         v->link = dev->ifindex;
 804         if (v->flags & (VIFF_TUNNEL | VIFF_REGISTER))
 805                 v->link = dev_get_iflink(dev);
 806
 807         /* And finish update writing critical data */
 808         write_lock_bh(&mrt_lock);
 809         v->dev = dev;
 810 #ifdef CONFIG_IP_PIMSM
 811         if (v->flags & VIFF_REGISTER)
 812                 mrt->mroute_reg_vif_num = vifi;
 813 #endif
 814         if (vifi+1 > mrt->maxvif)
 815                 mrt->maxvif = vifi+1;
 816         write_unlock_bh(&mrt_lock);
 817         return 0;
 818 }
 819
 820 /* called with rcu_read_lock() */
 821 static struct mfc_cache *ipmr_cache_find(struct mr_table *mrt,
 822                                          __be32 origin,
 823                                          __be32 mcastgrp)
 824 {
 825         int line = MFC_HASH(mcastgrp, origin);
 826         struct mfc_cache *c;
 827
 828         list_for_each_entry_rcu(c, &mrt->mfc_cache_array[line], list) {
 829                 if (c->mfc_origin == origin && c->mfc_mcastgrp == mcastgrp)
 830                         return c;
 831         }
 832         return NULL;
 833 }
 834
 835 /* Look for a (*,*,oif) entry */
 836 static struct mfc_cache *ipmr_cache_find_any_parent(struct mr_table *mrt,
 837                                                     int vifi)
 838 {
 839         int line = MFC_HASH(htonl(INADDR_ANY), htonl(INADDR_ANY));
 840         struct mfc_cache *c;
 841
 842         list_for_each_entry_rcu(c, &mrt->mfc_cache_array[line], list)
 843                 if (c->mfc_origin == htonl(INADDR_ANY) &&
 844                     c->mfc_mcastgrp == htonl(INADDR_ANY) &&
 845                     c->mfc_un.res.ttls[vifi] < 255)
 846                         return c;
 847
 848         return NULL;
 849 }
 850
 851 /* Look for a (*,G) entry */
 852 static struct mfc_cache *ipmr_cache_find_any(struct mr_table *mrt,
 853                                              __be32 mcastgrp, int vifi)
 854 {
 855         int line = MFC_HASH(mcastgrp, htonl(INADDR_ANY));
 856         struct mfc_cache *c, *proxy;
 857
 858         if (mcastgrp == htonl(INADDR_ANY))
 859                 goto skip;
 860
 861         list_for_each_entry_rcu(c, &mrt->mfc_cache_array[line], list)
 862                 if (c->mfc_origin == htonl(INADDR_ANY) &&
 863                     c->mfc_mcastgrp == mcastgrp) {
 864                         if (c->mfc_un.res.ttls[vifi] < 255)
 865                                 return c;
 866
 867                         /* It's ok if the vifi is part of the static tree */
 868                         proxy = ipmr_cache_find_any_parent(mrt,
 869                                                            c->mfc_parent);
 870                         if (proxy && proxy->mfc_un.res.ttls[vifi] < 255)
 871                                 return c;
 872                 }
 873
 874 skip:
 875         return ipmr_cache_find_any_parent(mrt, vifi);
 876 }
 877
 878 /*
 879  *      Allocate a multicast cache entry
 880  */
 881 static struct mfc_cache *ipmr_cache_alloc(void)
 882 {
 883         struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_KERNEL);
 884
 885         if (c)
 886                 c->mfc_un.res.minvif = MAXVIFS;
 887         return c;
 888 }
 889
 890 static struct mfc_cache *ipmr_cache_alloc_unres(void)
 891 {
 892         struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_ATOMIC);
 893
 894         if (c) {
 895                 skb_queue_head_init(&c->mfc_un.unres.unresolved);
 896                 c->mfc_un.unres.expires = jiffies + 10*HZ;
 897         }
 898         return c;
 899 }
 900
 901 /*
 902  *      A cache entry has gone into a resolved state from queued
 903  */
 904
 905 static void ipmr_cache_resolve(struct net *net, struct mr_table *mrt,
 906                                struct mfc_cache *uc, struct mfc_cache *c)
 907 {
 908         struct sk_buff *skb;
 909         struct nlmsgerr *e;
 910
 911         /* Play the pending entries through our router */
 912
 913         while ((skb = __skb_dequeue(&uc->mfc_un.unres.unresolved))) {
 914                 if (ip_hdr(skb)->version == 0) {
 915                         struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct iphdr));
 916
 917                         if (__ipmr_fill_mroute(mrt, skb, c, nlmsg_data(nlh)) > 0) {
 918                                 nlh->nlmsg_len = skb_tail_pointer(skb) -
 919                                                  (u8 *)nlh;
 920                         } else {
 921                                 nlh->nlmsg_type = NLMSG_ERROR;
 922                                 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
 923                                 skb_trim(skb, nlh->nlmsg_len);
 924                                 e = nlmsg_data(nlh);
 925                                 e->error = -EMSGSIZE;
 926                                 memset(&e->msg, 0, sizeof(e->msg));
 927                         }
 928
 929                         rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
 930                 } else {
 931                         ip_mr_forward(net, mrt, skb, c, 0);
 932                 }
 933         }
 934 }
 935
 936 /*
 937  *      Bounce a cache query up to mrouted. We could use netlink for this but mrouted
 938  *      expects the following bizarre scheme.
 939  *
 940  *      Called under mrt_lock.
 941  */
 942
 943 static int ipmr_cache_report(struct mr_table *mrt,
 944                              struct sk_buff *pkt, vifi_t vifi, int assert)
 945 {
 946         struct sk_buff *skb;
 947         const int ihl = ip_hdrlen(pkt);
 948         struct igmphdr *igmp;
 949         struct igmpmsg *msg;
 950         struct sock *mroute_sk;
 951         int ret;
 952
 953 #ifdef CONFIG_IP_PIMSM
 954         if (assert == IGMPMSG_WHOLEPKT)
 955                 skb = skb_realloc_headroom(pkt, sizeof(struct iphdr));
 956         else
 957 #endif
 958                 skb = alloc_skb(128, GFP_ATOMIC);
 959
 960         if (!skb)
 961                 return -ENOBUFS;
 962
 963 #ifdef CONFIG_IP_PIMSM
 964         if (assert == IGMPMSG_WHOLEPKT) {
 965                 /* Ugly, but we have no choice with this interface.
 966                  * Duplicate old header, fix ihl, length etc.
 967                  * And all this only to mangle msg->im_msgtype and
 968                  * to set msg->im_mbz to "mbz" :-)
 969                  */
 970                 skb_push(skb, sizeof(struct iphdr));
 971                 skb_reset_network_header(skb);
 972                 skb_reset_transport_header(skb);
 973                 msg = (struct igmpmsg *)skb_network_header(skb);
 974                 memcpy(msg, skb_network_header(pkt), sizeof(struct iphdr));
 975                 msg->im_msgtype = IGMPMSG_WHOLEPKT;
 976                 msg->im_mbz = 0;
 977                 msg->im_vif = mrt->mroute_reg_vif_num;
 978                 ip_hdr(skb)->ihl = sizeof(struct iphdr) >> 2;
 979                 ip_hdr(skb)->tot_len = htons(ntohs(ip_hdr(pkt)->tot_len) +
 980                                              sizeof(struct iphdr));
 981         } else
 982 #endif
 983         {
 984
 985         /* Copy the IP header */
 986
 987         skb_set_network_header(skb, skb->len);
 988         skb_put(skb, ihl);
 989         skb_copy_to_linear_data(skb, pkt->data, ihl);
 990         ip_hdr(skb)->protocol = 0;      /* Flag to the kernel this is a route add */
 991         msg = (struct igmpmsg *)skb_network_header(skb);
 992         msg->im_vif = vifi;
 993         skb_dst_set(skb, dst_clone(skb_dst(pkt)));
 994
 995         /* Add our header */
 996
 997         igmp = (struct igmphdr *)skb_put(skb, sizeof(struct igmphdr));
 998         igmp->type      =
 999         msg->im_msgtype = assert;
1000         igmp->code      = 0;
1001         ip_hdr(skb)->tot_len = htons(skb->len);         /* Fix the length */
1002         skb->transport_header = skb->network_header;
1003         }
1004
1005         rcu_read_lock();
1006         mroute_sk = rcu_dereference(mrt->mroute_sk);
1007         if (!mroute_sk) {
1008                 rcu_read_unlock();
1009                 kfree_skb(skb);
1010                 return -EINVAL;
1011         }
1012
1013         /* Deliver to mrouted */
1014
1015         ret = sock_queue_rcv_skb(mroute_sk, skb);
1016         rcu_read_unlock();
1017         if (ret < 0) {
1018                 net_warn_ratelimited("mroute: pending queue full, dropping entries\n");
1019                 kfree_skb(skb);
1020         }
1021
1022         return ret;
1023 }
1024
1025 /*
1026  *      Queue a packet for resolution. It gets locked cache entry!
1027  */
1028
1029 static int
1030 ipmr_cache_unresolved(struct mr_table *mrt, vifi_t vifi, struct sk_buff *skb)
1031 {
1032         bool found = false;
1033         int err;
1034         struct mfc_cache *c;
1035         const struct iphdr *iph = ip_hdr(skb);
1036
1037         spin_lock_bh(&mfc_unres_lock);
1038         list_for_each_entry(c, &mrt->mfc_unres_queue, list) {
1039                 if (c->mfc_mcastgrp == iph->daddr &&
1040                     c->mfc_origin == iph->saddr) {
1041                         found = true;
1042                         break;
1043                 }
1044         }
1045
1046         if (!found) {
1047                 /* Create a new entry if allowable */
1048
1049                 if (atomic_read(&mrt->cache_resolve_queue_len) >= 10 ||
1050                     (c = ipmr_cache_alloc_unres()) == NULL) {
1051                         spin_unlock_bh(&mfc_unres_lock);
1052
1053                         kfree_skb(skb);
1054                         return -ENOBUFS;
1055                 }
1056
1057                 /* Fill in the new cache entry */
1058
1059                 c->mfc_parent   = -1;
1060                 c->mfc_origin   = iph->saddr;
1061                 c->mfc_mcastgrp = iph->daddr;
1062
1063                 /* Reflect first query at mrouted. */
1064
1065                 err = ipmr_cache_report(mrt, skb, vifi, IGMPMSG_NOCACHE);
1066                 if (err < 0) {
1067                         /* If the report failed throw the cache entry
1068                            out - Brad Parker
1069                          */
1070                         spin_unlock_bh(&mfc_unres_lock);
1071
1072                         ipmr_cache_free(c);
1073                         kfree_skb(skb);
1074                         return err;
1075                 }
1076
1077                 atomic_inc(&mrt->cache_resolve_queue_len);
1078                 list_add(&c->list, &mrt->mfc_unres_queue);
1079                 mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1080
1081                 if (atomic_read(&mrt->cache_resolve_queue_len) == 1)
1082                         mod_timer(&mrt->ipmr_expire_timer, c->mfc_un.unres.expires);
1083         }
1084
1085         /* See if we can append the packet */
1086
1087         if (c->mfc_un.unres.unresolved.qlen > 3) {
1088                 kfree_skb(skb);
1089                 err = -ENOBUFS;
1090         } else {
1091                 skb_queue_tail(&c->mfc_un.unres.unresolved, skb);
1092                 err = 0;
1093         }
1094
1095         spin_unlock_bh(&mfc_unres_lock);
1096         return err;
1097 }
1098
1099 /*
1100  *      MFC cache manipulation by user space mroute daemon
1101  */
1102
1103 static int ipmr_mfc_delete(struct mr_table *mrt, struct mfcctl *mfc, int parent)
1104 {
1105         int line;
1106         struct mfc_cache *c, *next;
1107
1108         line = MFC_HASH(mfc->mfcc_mcastgrp.s_addr, mfc->mfcc_origin.s_addr);
1109
1110         list_for_each_entry_safe(c, next, &mrt->mfc_cache_array[line], list) {
1111                 if (c->mfc_origin == mfc->mfcc_origin.s_addr &&
1112                     c->mfc_mcastgrp == mfc->mfcc_mcastgrp.s_addr &&
1113                     (parent == -1 || parent == c->mfc_parent)) {
1114                         list_del_rcu(&c->list);
1115                         mroute_netlink_event(mrt, c, RTM_DELROUTE);
1116                         ipmr_cache_free(c);
1117                         return 0;
1118                 }
1119         }
1120         return -ENOENT;
1121 }
1122
1123 static int ipmr_mfc_add(struct net *net, struct mr_table *mrt,
1124                         struct mfcctl *mfc, int mrtsock, int parent)
1125 {
1126         bool found = false;
1127         int line;
1128         struct mfc_cache *uc, *c;
1129
1130         if (mfc->mfcc_parent >= MAXVIFS)
1131                 return -ENFILE;
1132
1133         line = MFC_HASH(mfc->mfcc_mcastgrp.s_addr, mfc->mfcc_origin.s_addr);
1134
1135         list_for_each_entry(c, &mrt->mfc_cache_array[line], list) {
1136                 if (c->mfc_origin == mfc->mfcc_origin.s_addr &&
1137                     c->mfc_mcastgrp == mfc->mfcc_mcastgrp.s_addr &&
1138                     (parent == -1 || parent == c->mfc_parent)) {
1139                         found = true;
1140                         break;
1141                 }
1142         }
1143
1144         if (found) {
1145                 write_lock_bh(&mrt_lock);
1146                 c->mfc_parent = mfc->mfcc_parent;
1147                 ipmr_update_thresholds(mrt, c, mfc->mfcc_ttls);
1148                 if (!mrtsock)
1149                         c->mfc_flags |= MFC_STATIC;
1150                 write_unlock_bh(&mrt_lock);
1151                 mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1152                 return 0;
1153         }
1154
1155         if (mfc->mfcc_mcastgrp.s_addr != htonl(INADDR_ANY) &&
1156             !ipv4_is_multicast(mfc->mfcc_mcastgrp.s_addr))
1157                 return -EINVAL;
1158
1159         c = ipmr_cache_alloc();
1160         if (!c)
1161                 return -ENOMEM;
1162
1163         c->mfc_origin = mfc->mfcc_origin.s_addr;
1164         c->mfc_mcastgrp = mfc->mfcc_mcastgrp.s_addr;
1165         c->mfc_parent = mfc->mfcc_parent;
1166         ipmr_update_thresholds(mrt, c, mfc->mfcc_ttls);
1167         if (!mrtsock)
1168                 c->mfc_flags |= MFC_STATIC;
1169
1170         list_add_rcu(&c->list, &mrt->mfc_cache_array[line]);
1171
1172         /*
1173          *      Check to see if we resolved a queued list. If so we
1174          *      need to send on the frames and tidy up.
1175          */
1176         found = false;
1177         spin_lock_bh(&mfc_unres_lock);
1178         list_for_each_entry(uc, &mrt->mfc_unres_queue, list) {
1179                 if (uc->mfc_origin == c->mfc_origin &&
1180                     uc->mfc_mcastgrp == c->mfc_mcastgrp) {
1181                         list_del(&uc->list);
1182                         atomic_dec(&mrt->cache_resolve_queue_len);
1183                         found = true;
1184                         break;
1185                 }
1186         }
1187         if (list_empty(&mrt->mfc_unres_queue))
1188                 del_timer(&mrt->ipmr_expire_timer);
1189         spin_unlock_bh(&mfc_unres_lock);
1190
1191         if (found) {
1192                 ipmr_cache_resolve(net, mrt, uc, c);
1193                 ipmr_cache_free(uc);
1194         }
1195         mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1196         return 0;
1197 }
1198
1199 /*
1200  *      Close the multicast socket, and clear the vif tables etc
1201  */
1202
1203 static void mroute_clean_tables(struct mr_table *mrt, bool all)
1204 {
1205         int i;
1206         LIST_HEAD(list);
1207         struct mfc_cache *c, *next;
1208
1209         /* Shut down all active vif entries */
1210
1211         for (i = 0; i < mrt->maxvif; i++) {
1212                 if (!all && (mrt->vif_table[i].flags & VIFF_STATIC))
1213                         continue;
1214                 vif_delete(mrt, i, 0, &list);
1215         }
1216         unregister_netdevice_many(&list);
1217
1218         /* Wipe the cache */
1219
1220         for (i = 0; i < MFC_LINES; i++) {
1221                 list_for_each_entry_safe(c, next, &mrt->mfc_cache_array[i], list) {
1222                         if (!all && (c->mfc_flags & MFC_STATIC))
1223                                 continue;
1224                         list_del_rcu(&c->list);
1225                         mroute_netlink_event(mrt, c, RTM_DELROUTE);
1226                         ipmr_cache_free(c);
1227                 }
1228         }
1229
1230         if (atomic_read(&mrt->cache_resolve_queue_len) != 0) {
1231                 spin_lock_bh(&mfc_unres_lock);
1232                 list_for_each_entry_safe(c, next, &mrt->mfc_unres_queue, list) {
1233                         list_del(&c->list);
1234                         mroute_netlink_event(mrt, c, RTM_DELROUTE);
1235                         ipmr_destroy_unres(mrt, c);
1236                 }
1237                 spin_unlock_bh(&mfc_unres_lock);
1238         }
1239 }
1240
1241 /* called from ip_ra_control(), before an RCU grace period,
1242  * we dont need to call synchronize_rcu() here
1243  */
1244 static void mrtsock_destruct(struct sock *sk)
1245 {
1246         struct net *net = sock_net(sk);
1247         struct mr_table *mrt;
1248
1249         rtnl_lock();
1250         ipmr_for_each_table(mrt, net) {
1251                 if (sk == rtnl_dereference(mrt->mroute_sk)) {
1252                         IPV4_DEVCONF_ALL(net, MC_FORWARDING)--;
1253                         inet_netconf_notify_devconf(net, NETCONFA_MC_FORWARDING,
1254                                                     NETCONFA_IFINDEX_ALL,
1255                                                     net->ipv4.devconf_all);
1256                         RCU_INIT_POINTER(mrt->mroute_sk, NULL);
1257                         mroute_clean_tables(mrt, false);
1258                 }
1259         }
1260         rtnl_unlock();
1261 }
1262
1263 /*
1264  *      Socket options and virtual interface manipulation. The whole
1265  *      virtual interface system is a complete heap, but unfortunately
1266  *      that's how BSD mrouted happens to think. Maybe one day with a proper
1267  *      MOSPF/PIM router set up we can clean this up.
1268  */
1269
1270 int ip_mroute_setsockopt(struct sock *sk, int optname, char __user *optval, unsigned int optlen)
1271 {
1272         int ret, parent = 0;
1273         struct vifctl vif;
1274         struct mfcctl mfc;
1275         struct net *net = sock_net(sk);
1276         struct mr_table *mrt;
1277
1278         if (sk->sk_type != SOCK_RAW ||
1279             inet_sk(sk)->inet_num != IPPROTO_IGMP)
1280                 return -EOPNOTSUPP;
1281
1282         mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1283         if (!mrt)
1284                 return -ENOENT;
1285
1286         if (optname != MRT_INIT) {
1287                 if (sk != rcu_access_pointer(mrt->mroute_sk) &&
1288                     !ns_capable(net->user_ns, CAP_NET_ADMIN))
1289                         return -EACCES;
1290         }
1291
1292         switch (optname) {
1293         case MRT_INIT:
1294                 if (optlen != sizeof(int))
1295                         return -EINVAL;
1296
1297                 rtnl_lock();
1298                 if (rtnl_dereference(mrt->mroute_sk)) {
1299                         rtnl_unlock();
1300                         return -EADDRINUSE;
1301                 }
1302
1303                 ret = ip_ra_control(sk, 1, mrtsock_destruct);
1304                 if (ret == 0) {
1305                         rcu_assign_pointer(mrt->mroute_sk, sk);
1306                         IPV4_DEVCONF_ALL(net, MC_FORWARDING)++;
1307                         inet_netconf_notify_devconf(net, NETCONFA_MC_FORWARDING,
1308                                                     NETCONFA_IFINDEX_ALL,
1309                                                     net->ipv4.devconf_all);
1310                 }
1311                 rtnl_unlock();
1312                 return ret;
1313         case MRT_DONE:
1314                 if (sk != rcu_access_pointer(mrt->mroute_sk))
1315                         return -EACCES;
1316                 return ip_ra_control(sk, 0, NULL);
1317         case MRT_ADD_VIF:
1318         case MRT_DEL_VIF:
1319                 if (optlen != sizeof(vif))
1320                         return -EINVAL;
1321                 if (copy_from_user(&vif, optval, sizeof(vif)))
1322                         return -EFAULT;
1323                 if (vif.vifc_vifi >= MAXVIFS)
1324                         return -ENFILE;
1325                 rtnl_lock();
1326                 if (optname == MRT_ADD_VIF) {
1327                         ret = vif_add(net, mrt, &vif,
1328                                       sk == rtnl_dereference(mrt->mroute_sk));
1329                 } else {
1330                         ret = vif_delete(mrt, vif.vifc_vifi, 0, NULL);
1331                 }
1332                 rtnl_unlock();
1333                 return ret;
1334
1335                 /*
1336                  *      Manipulate the forwarding caches. These live
1337                  *      in a sort of kernel/user symbiosis.
1338                  */
1339         case MRT_ADD_MFC:
1340         case MRT_DEL_MFC:
1341                 parent = -1;
1342         case MRT_ADD_MFC_PROXY:
1343         case MRT_DEL_MFC_PROXY:
1344                 if (optlen != sizeof(mfc))
1345                         return -EINVAL;
1346                 if (copy_from_user(&mfc, optval, sizeof(mfc)))
1347                         return -EFAULT;
1348                 if (parent == 0)
1349                         parent = mfc.mfcc_parent;
1350                 rtnl_lock();
1351                 if (optname == MRT_DEL_MFC || optname == MRT_DEL_MFC_PROXY)
1352                         ret = ipmr_mfc_delete(mrt, &mfc, parent);
1353                 else
1354                         ret = ipmr_mfc_add(net, mrt, &mfc,
1355                                            sk == rtnl_dereference(mrt->mroute_sk),
1356                                            parent);
1357                 rtnl_unlock();
1358                 return ret;
1359                 /*
1360                  *      Control PIM assert.
1361                  */
1362         case MRT_ASSERT:
1363         {
1364                 int v;
1365                 if (optlen != sizeof(v))
1366                         return -EINVAL;
1367                 if (get_user(v, (int __user *)optval))
1368                         return -EFAULT;
1369                 mrt->mroute_do_assert = v;
1370                 return 0;
1371         }
1372 #ifdef CONFIG_IP_PIMSM
1373         case MRT_PIM:
1374         {
1375                 int v;
1376
1377                 if (optlen != sizeof(v))
1378                         return -EINVAL;
1379                 if (get_user(v, (int __user *)optval))
1380                         return -EFAULT;
1381                 v = !!v;
1382
1383                 rtnl_lock();
1384                 ret = 0;
1385                 if (v != mrt->mroute_do_pim) {
1386                         mrt->mroute_do_pim = v;
1387                         mrt->mroute_do_assert = v;
1388                 }
1389                 rtnl_unlock();
1390                 return ret;
1391         }
1392 #endif
1393 #ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES
1394         case MRT_TABLE:
1395         {
1396                 u32 v;
1397
1398                 if (optlen != sizeof(u32))
1399                         return -EINVAL;
1400                 if (get_user(v, (u32 __user *)optval))
1401                         return -EFAULT;
1402
1403                 /* "pimreg%u" should not exceed 16 bytes (IFNAMSIZ) */
1404                 if (v != RT_TABLE_DEFAULT && v >= 1000000000)
1405                         return -EINVAL;
1406
1407                 rtnl_lock();
1408                 ret = 0;
1409                 if (sk == rtnl_dereference(mrt->mroute_sk)) {
1410                         ret = -EBUSY;
1411                 } else {
1412                         if (!ipmr_new_table(net, v))
1413                                 ret = -ENOMEM;
1414                         else
1415                                 raw_sk(sk)->ipmr_table = v;
1416                 }
1417                 rtnl_unlock();
1418                 return ret;
1419         }
1420 #endif
1421         /*
1422          *      Spurious command, or MRT_VERSION which you cannot
1423          *      set.
1424          */
1425         default:
1426                 return -ENOPROTOOPT;
1427         }
1428 }
1429
1430 /*
1431  *      Getsock opt support for the multicast routing system.
1432  */
1433
1434 int ip_mroute_getsockopt(struct sock *sk, int optname, char __user *optval, int __user *optlen)
1435 {
1436         int olr;
1437         int val;
1438         struct net *net = sock_net(sk);
1439         struct mr_table *mrt;
1440
1441         if (sk->sk_type != SOCK_RAW ||
1442             inet_sk(sk)->inet_num != IPPROTO_IGMP)
1443                 return -EOPNOTSUPP;
1444
1445         mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1446         if (!mrt)
1447                 return -ENOENT;
1448
1449         if (optname != MRT_VERSION &&
1450 #ifdef CONFIG_IP_PIMSM
1451            optname != MRT_PIM &&
1452 #endif
1453            optname != MRT_ASSERT)
1454                 return -ENOPROTOOPT;
1455
1456         if (get_user(olr, optlen))
1457                 return -EFAULT;
1458
1459         olr = min_t(unsigned int, olr, sizeof(int));
1460         if (olr < 0)
1461                 return -EINVAL;
1462
1463         if (put_user(olr, optlen))
1464                 return -EFAULT;
1465         if (optname == MRT_VERSION)
1466                 val = 0x0305;
1467 #ifdef CONFIG_IP_PIMSM
1468         else if (optname == MRT_PIM)
1469                 val = mrt->mroute_do_pim;
1470 #endif
1471         else
1472                 val = mrt->mroute_do_assert;
1473         if (copy_to_user(optval, &val, olr))
1474                 return -EFAULT;
1475         return 0;
1476 }
1477
1478 /*
1479  *      The IP multicast ioctl support routines.
1480  */
1481
1482 int ipmr_ioctl(struct sock *sk, int cmd, void __user *arg)
1483 {
1484         struct sioc_sg_req sr;
1485         struct sioc_vif_req vr;
1486         struct vif_device *vif;
1487         struct mfc_cache *c;
1488         struct net *net = sock_net(sk);
1489         struct mr_table *mrt;
1490
1491         mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1492         if (!mrt)
1493                 return -ENOENT;
1494
1495         switch (cmd) {
1496         case SIOCGETVIFCNT:
1497                 if (copy_from_user(&vr, arg, sizeof(vr)))
1498                         return -EFAULT;
1499                 if (vr.vifi >= mrt->maxvif)
1500                         return -EINVAL;
1501                 read_lock(&mrt_lock);
1502                 vif = &mrt->vif_table[vr.vifi];
1503                 if (VIF_EXISTS(mrt, vr.vifi)) {
1504                         vr.icount = vif->pkt_in;
1505                         vr.ocount = vif->pkt_out;
1506                         vr.ibytes = vif->bytes_in;
1507                         vr.obytes = vif->bytes_out;
1508                         read_unlock(&mrt_lock);
1509
1510                         if (copy_to_user(arg, &vr, sizeof(vr)))
1511                                 return -EFAULT;
1512                         return 0;
1513                 }
1514                 read_unlock(&mrt_lock);
1515                 return -EADDRNOTAVAIL;
1516         case SIOCGETSGCNT:
1517                 if (copy_from_user(&sr, arg, sizeof(sr)))
1518                         return -EFAULT;
1519
1520                 rcu_read_lock();
1521                 c = ipmr_cache_find(mrt, sr.src.s_addr, sr.grp.s_addr);
1522                 if (c) {
1523                         sr.pktcnt = c->mfc_un.res.pkt;
1524                         sr.bytecnt = c->mfc_un.res.bytes;
1525                         sr.wrong_if = c->mfc_un.res.wrong_if;
1526                         rcu_read_unlock();
1527
1528                         if (copy_to_user(arg, &sr, sizeof(sr)))
1529                                 return -EFAULT;
1530                         return 0;
1531                 }
1532                 rcu_read_unlock();
1533                 return -EADDRNOTAVAIL;
1534         default:
1535                 return -ENOIOCTLCMD;
1536         }
1537 }
1538
1539 #ifdef CONFIG_COMPAT
1540 struct compat_sioc_sg_req {
1541         struct in_addr src;
1542         struct in_addr grp;
1543         compat_ulong_t pktcnt;
1544         compat_ulong_t bytecnt;
1545         compat_ulong_t wrong_if;
1546 };
1547
1548 struct compat_sioc_vif_req {
1549         vifi_t  vifi;           /* Which iface */
1550         compat_ulong_t icount;
1551         compat_ulong_t ocount;
1552         compat_ulong_t ibytes;
1553         compat_ulong_t obytes;
1554 };
1555
1556 int ipmr_compat_ioctl(struct sock *sk, unsigned int cmd, void __user *arg)
1557 {
1558         struct compat_sioc_sg_req sr;
1559         struct compat_sioc_vif_req vr;
1560         struct vif_device *vif;
1561         struct mfc_cache *c;
1562         struct net *net = sock_net(sk);
1563         struct mr_table *mrt;
1564
1565         mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1566         if (!mrt)
1567                 return -ENOENT;
1568
1569         switch (cmd) {
1570         case SIOCGETVIFCNT:
1571                 if (copy_from_user(&vr, arg, sizeof(vr)))
1572                         return -EFAULT;
1573                 if (vr.vifi >= mrt->maxvif)
1574                         return -EINVAL;
1575                 read_lock(&mrt_lock);
1576                 vif = &mrt->vif_table[vr.vifi];
1577                 if (VIF_EXISTS(mrt, vr.vifi)) {
1578                         vr.icount = vif->pkt_in;
1579                         vr.ocount = vif->pkt_out;
1580                         vr.ibytes = vif->bytes_in;
1581                         vr.obytes = vif->bytes_out;
1582                         read_unlock(&mrt_lock);
1583
1584                         if (copy_to_user(arg, &vr, sizeof(vr)))
1585                                 return -EFAULT;
1586                         return 0;
1587                 }
1588                 read_unlock(&mrt_lock);
1589                 return -EADDRNOTAVAIL;
1590         case SIOCGETSGCNT:
1591                 if (copy_from_user(&sr, arg, sizeof(sr)))
1592                         return -EFAULT;
1593
1594                 rcu_read_lock();
1595                 c = ipmr_cache_find(mrt, sr.src.s_addr, sr.grp.s_addr);
1596                 if (c) {
1597                         sr.pktcnt = c->mfc_un.res.pkt;
1598                         sr.bytecnt = c->mfc_un.res.bytes;
1599                         sr.wrong_if = c->mfc_un.res.wrong_if;
1600                         rcu_read_unlock();
1601
1602                         if (copy_to_user(arg, &sr, sizeof(sr)))
1603                                 return -EFAULT;
1604                         return 0;
1605                 }
1606                 rcu_read_unlock();
1607                 return -EADDRNOTAVAIL;
1608         default:
1609                 return -ENOIOCTLCMD;
1610         }
1611 }
1612 #endif
1613
1614
1615 static int ipmr_device_event(struct notifier_block *this, unsigned long event, void *ptr)
1616 {
1617         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1618         struct net *net = dev_net(dev);
1619         struct mr_table *mrt;
1620         struct vif_device *v;
1621         int ct;
1622
1623         if (event != NETDEV_UNREGISTER)
1624                 return NOTIFY_DONE;
1625
1626         ipmr_for_each_table(mrt, net) {
1627                 v = &mrt->vif_table[0];
1628                 for (ct = 0; ct < mrt->maxvif; ct++, v++) {
1629                         if (v->dev == dev)
1630                                 vif_delete(mrt, ct, 1, NULL);
1631                 }
1632         }
1633         return NOTIFY_DONE;
1634 }
1635
1636
1637 static struct notifier_block ip_mr_notifier = {
1638         .notifier_call = ipmr_device_event,
1639 };
1640
1641 /*
1642  *      Encapsulate a packet by attaching a valid IPIP header to it.
1643  *      This avoids tunnel drivers and other mess and gives us the speed so
1644  *      important for multicast video.
1645  */
1646
1647 static void ip_encap(struct net *net, struct sk_buff *skb,
1648                      __be32 saddr, __be32 daddr)
1649 {
1650         struct iphdr *iph;
1651         const struct iphdr *old_iph = ip_hdr(skb);
1652
1653         skb_push(skb, sizeof(struct iphdr));
1654         skb->transport_header = skb->network_header;
1655         skb_reset_network_header(skb);
1656         iph = ip_hdr(skb);
1657
1658         iph->version    =       4;
1659         iph->tos        =       old_iph->tos;
1660         iph->ttl        =       old_iph->ttl;
1661         iph->frag_off   =       0;
1662         iph->daddr      =       daddr;
1663         iph->saddr      =       saddr;
1664         iph->protocol   =       IPPROTO_IPIP;
1665         iph->ihl        =       5;
1666         iph->tot_len    =       htons(skb->len);
1667         ip_select_ident(net, skb, NULL);
1668         ip_send_check(iph);
1669
1670         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1671         nf_reset(skb);
1672 }
1673
1674 static inline int ipmr_forward_finish(struct net *net, struct sock *sk,
1675                                       struct sk_buff *skb)
1676 {
1677         struct ip_options *opt = &(IPCB(skb)->opt);
1678
1679         IP_INC_STATS(net, IPSTATS_MIB_OUTFORWDATAGRAMS);
1680         IP_ADD_STATS(net, IPSTATS_MIB_OUTOCTETS, skb->len);
1681
1682         if (unlikely(opt->optlen))
1683                 ip_forward_options(skb);
1684
1685         return dst_output(net, sk, skb);
1686 }
1687
1688 /*
1689  *      Processing handlers for ipmr_forward
1690  */
1691
1692 static void ipmr_queue_xmit(struct net *net, struct mr_table *mrt,
1693                             struct sk_buff *skb, struct mfc_cache *c, int vifi)
1694 {
1695         const struct iphdr *iph = ip_hdr(skb);
1696         struct vif_device *vif = &mrt->vif_table[vifi];
1697         struct net_device *dev;
1698         struct rtable *rt;
1699         struct flowi4 fl4;
1700         int    encap = 0;
1701
1702         if (!vif->dev)
1703                 goto out_free;
1704
1705 #ifdef CONFIG_IP_PIMSM
1706         if (vif->flags & VIFF_REGISTER) {
1707                 vif->pkt_out++;
1708                 vif->bytes_out += skb->len;
1709                 vif->dev->stats.tx_bytes += skb->len;
1710                 vif->dev->stats.tx_packets++;
1711                 ipmr_cache_report(mrt, skb, vifi, IGMPMSG_WHOLEPKT);
1712                 goto out_free;
1713         }
1714 #endif
1715
1716         if (vif->flags & VIFF_TUNNEL) {
1717                 rt = ip_route_output_ports(net, &fl4, NULL,
1718                                            vif->remote, vif->local,
1719                                            0, 0,
1720                                            IPPROTO_IPIP,
1721                                            RT_TOS(iph->tos), vif->link);
1722                 if (IS_ERR(rt))
1723                         goto out_free;
1724                 encap = sizeof(struct iphdr);
1725         } else {
1726                 rt = ip_route_output_ports(net, &fl4, NULL, iph->daddr, 0,
1727                                            0, 0,
1728                                            IPPROTO_IPIP,
1729                                            RT_TOS(iph->tos), vif->link);
1730                 if (IS_ERR(rt))
1731                         goto out_free;
1732         }
1733
1734         dev = rt->dst.dev;
1735
1736         if (skb->len+encap > dst_mtu(&rt->dst) && (ntohs(iph->frag_off) & IP_DF)) {
1737                 /* Do not fragment multicasts. Alas, IPv4 does not
1738                  * allow to send ICMP, so that packets will disappear
1739                  * to blackhole.
1740                  */
1741
1742                 IP_INC_STATS(net, IPSTATS_MIB_FRAGFAILS);
1743                 ip_rt_put(rt);
1744                 goto out_free;
1745         }
1746
1747         encap += LL_RESERVED_SPACE(dev) + rt->dst.header_len;
1748
1749         if (skb_cow(skb, encap)) {
1750                 ip_rt_put(rt);
1751                 goto out_free;
1752         }
1753
1754         vif->pkt_out++;
1755         vif->bytes_out += skb->len;
1756
1757         skb_dst_drop(skb);
1758         skb_dst_set(skb, &rt->dst);
1759         ip_decrease_ttl(ip_hdr(skb));
1760
1761         /* FIXME: forward and output firewalls used to be called here.
1762          * What do we do with netfilter? -- RR
1763          */
1764         if (vif->flags & VIFF_TUNNEL) {
1765                 ip_encap(net, skb, vif->local, vif->remote);
1766                 /* FIXME: extra output firewall step used to be here. --RR */
1767                 vif->dev->stats.tx_packets++;
1768                 vif->dev->stats.tx_bytes += skb->len;
1769         }
1770
1771         IPCB(skb)->flags |= IPSKB_FORWARDED;
1772
1773         /*
1774          * RFC1584 teaches, that DVMRP/PIM router must deliver packets locally
1775          * not only before forwarding, but after forwarding on all output
1776          * interfaces. It is clear, if mrouter runs a multicasting
1777          * program, it should receive packets not depending to what interface
1778          * program is joined.
1779          * If we will not make it, the program will have to join on all
1780          * interfaces. On the other hand, multihoming host (or router, but
1781          * not mrouter) cannot join to more than one interface - it will
1782          * result in receiving multiple packets.
1783          */
1784         NF_HOOK(NFPROTO_IPV4, NF_INET_FORWARD,
1785                 net, NULL, skb, skb->dev, dev,
1786                 ipmr_forward_finish);
1787         return;
1788
1789 out_free:
1790         kfree_skb(skb);
1791 }
1792
1793 static int ipmr_find_vif(struct mr_table *mrt, struct net_device *dev)
1794 {
1795         int ct;
1796
1797         for (ct = mrt->maxvif-1; ct >= 0; ct--) {
1798                 if (mrt->vif_table[ct].dev == dev)
1799                         break;
1800         }
1801         return ct;
1802 }
1803
1804 /* "local" means that we should preserve one skb (for local delivery) */
1805
1806 static void ip_mr_forward(struct net *net, struct mr_table *mrt,
1807                           struct sk_buff *skb, struct mfc_cache *cache,
1808                           int local)
1809 {
1810         int psend = -1;
1811         int vif, ct;
1812         int true_vifi = ipmr_find_vif(mrt, skb->dev);
1813
1814         vif = cache->mfc_parent;
1815         cache->mfc_un.res.pkt++;
1816         cache->mfc_un.res.bytes += skb->len;
1817
1818         if (cache->mfc_origin == htonl(INADDR_ANY) && true_vifi >= 0) {
1819                 struct mfc_cache *cache_proxy;
1820
1821                 /* For an (*,G) entry, we only check that the incomming
1822                  * interface is part of the static tree.
1823                  */
1824                 cache_proxy = ipmr_cache_find_any_parent(mrt, vif);
1825                 if (cache_proxy &&
1826                     cache_proxy->mfc_un.res.ttls[true_vifi] < 255)
1827                         goto forward;
1828         }
1829
1830         /*
1831          * Wrong interface: drop packet and (maybe) send PIM assert.
1832          */
1833         if (mrt->vif_table[vif].dev != skb->dev) {
1834                 if (rt_is_output_route(skb_rtable(skb))) {
1835                         /* It is our own packet, looped back.
1836                          * Very complicated situation...
1837                          *
1838                          * The best workaround until routing daemons will be
1839                          * fixed is not to redistribute packet, if it was
1840                          * send through wrong interface. It means, that
1841                          * multicast applications WILL NOT work for
1842                          * (S,G), which have default multicast route pointing
1843                          * to wrong oif. In any case, it is not a good
1844                          * idea to use multicasting applications on router.
1845                          */
1846                         goto dont_forward;
1847                 }
1848
1849                 cache->mfc_un.res.wrong_if++;
1850
1851                 if (true_vifi >= 0 && mrt->mroute_do_assert &&
1852                     /* pimsm uses asserts, when switching from RPT to SPT,
1853                      * so that we cannot check that packet arrived on an oif.
1854                      * It is bad, but otherwise we would need to move pretty
1855                      * large chunk of pimd to kernel. Ough... --ANK
1856                      */
1857                     (mrt->mroute_do_pim ||
1858                      cache->mfc_un.res.ttls[true_vifi] < 255) &&
1859                     time_after(jiffies,
1860                                cache->mfc_un.res.last_assert + MFC_ASSERT_THRESH)) {
1861                         cache->mfc_un.res.last_assert = jiffies;
1862                         ipmr_cache_report(mrt, skb, true_vifi, IGMPMSG_WRONGVIF);
1863                 }
1864                 goto dont_forward;
1865         }
1866
1867 forward:
1868         mrt->vif_table[vif].pkt_in++;
1869         mrt->vif_table[vif].bytes_in += skb->len;
1870
1871         /*
1872          *      Forward the frame
1873          */
1874         if (cache->mfc_origin == htonl(INADDR_ANY) &&
1875             cache->mfc_mcastgrp == htonl(INADDR_ANY)) {
1876                 if (true_vifi >= 0 &&
1877                     true_vifi != cache->mfc_parent &&
1878                     ip_hdr(skb)->ttl >
1879                                 cache->mfc_un.res.ttls[cache->mfc_parent]) {
1880                         /* It's an (*,*) entry and the packet is not coming from
1881                          * the upstream: forward the packet to the upstream
1882                          * only.
1883                          */
1884                         psend = cache->mfc_parent;
1885                         goto last_forward;
1886                 }
1887                 goto dont_forward;
1888         }
1889         for (ct = cache->mfc_un.res.maxvif - 1;
1890              ct >= cache->mfc_un.res.minvif; ct--) {
1891                 /* For (*,G) entry, don't forward to the incoming interface */
1892                 if ((cache->mfc_origin != htonl(INADDR_ANY) ||
1893                      ct != true_vifi) &&
1894                     ip_hdr(skb)->ttl > cache->mfc_un.res.ttls[ct]) {
1895                         if (psend != -1) {
1896                                 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
1897
1898                                 if (skb2)
1899                                         ipmr_queue_xmit(net, mrt, skb2, cache,
1900                                                         psend);
1901                         }
1902                         psend = ct;
1903                 }
1904         }
1905 last_forward:
1906         if (psend != -1) {
1907                 if (local) {
1908                         struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
1909
1910                         if (skb2)
1911                                 ipmr_queue_xmit(net, mrt, skb2, cache, psend);
1912                 } else {
1913                         ipmr_queue_xmit(net, mrt, skb, cache, psend);
1914                         return;
1915                 }
1916         }
1917
1918 dont_forward:
1919         if (!local)
1920                 kfree_skb(skb);
1921 }
1922
1923 static struct mr_table *ipmr_rt_fib_lookup(struct net *net, struct sk_buff *skb)
1924 {
1925         struct rtable *rt = skb_rtable(skb);
1926         struct iphdr *iph = ip_hdr(skb);
1927         struct flowi4 fl4 = {
1928                 .daddr = iph->daddr,
1929                 .saddr = iph->saddr,
1930                 .flowi4_tos = RT_TOS(iph->tos),
1931                 .flowi4_oif = (rt_is_output_route(rt) ?
1932                                skb->dev->ifindex : 0),
1933                 .flowi4_iif = (rt_is_output_route(rt) ?
1934                                LOOPBACK_IFINDEX :
1935                                skb->dev->ifindex),
1936                 .flowi4_mark = skb->mark,
1937         };
1938         struct mr_table *mrt;
1939         int err;
1940
1941         err = ipmr_fib_lookup(net, &fl4, &mrt);
1942         if (err)
1943                 return ERR_PTR(err);
1944         return mrt;
1945 }
1946
1947 /*
1948  *      Multicast packets for forwarding arrive here
1949  *      Called with rcu_read_lock();
1950  */
1951
1952 int ip_mr_input(struct sk_buff *skb)
1953 {
1954         struct mfc_cache *cache;
1955         struct net *net = dev_net(skb->dev);
1956         int local = skb_rtable(skb)->rt_flags & RTCF_LOCAL;
1957         struct mr_table *mrt;
1958
1959         /* Packet is looped back after forward, it should not be
1960          * forwarded second time, but still can be delivered locally.
1961          */
1962         if (IPCB(skb)->flags & IPSKB_FORWARDED)
1963                 goto dont_forward;
1964
1965         mrt = ipmr_rt_fib_lookup(net, skb);
1966         if (IS_ERR(mrt)) {
1967                 kfree_skb(skb);
1968                 return PTR_ERR(mrt);
1969         }
1970         if (!local) {
1971                 if (IPCB(skb)->opt.router_alert) {
1972                         if (ip_call_ra_chain(skb))
1973                                 return 0;
1974                 } else if (ip_hdr(skb)->protocol == IPPROTO_IGMP) {
1975                         /* IGMPv1 (and broken IGMPv2 implementations sort of
1976                          * Cisco IOS <= 11.2(8)) do not put router alert
1977                          * option to IGMP packets destined to routable
1978                          * groups. It is very bad, because it means
1979                          * that we can forward NO IGMP messages.
1980                          */
1981                         struct sock *mroute_sk;
1982
1983                         mroute_sk = rcu_dereference(mrt->mroute_sk);
1984                         if (mroute_sk) {
1985                                 nf_reset(skb);
1986                                 raw_rcv(mroute_sk, skb);
1987                                 return 0;
1988                         }
1989                     }
1990         }
1991
1992         /* already under rcu_read_lock() */
1993         cache = ipmr_cache_find(mrt, ip_hdr(skb)->saddr, ip_hdr(skb)->daddr);
1994         if (!cache) {
1995                 int vif = ipmr_find_vif(mrt, skb->dev);
1996
1997                 if (vif >= 0)
1998                         cache = ipmr_cache_find_any(mrt, ip_hdr(skb)->daddr,
1999                                                     vif);
2000         }
2001
2002         /*
2003          *      No usable cache entry
2004          */
2005         if (!cache) {
2006                 int vif;
2007
2008                 if (local) {
2009                         struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2010                         ip_local_deliver(skb);
2011                         if (!skb2)
2012                                 return -ENOBUFS;
2013                         skb = skb2;
2014                 }
2015
2016                 read_lock(&mrt_lock);
2017                 vif = ipmr_find_vif(mrt, skb->dev);
2018                 if (vif >= 0) {
2019                         int err2 = ipmr_cache_unresolved(mrt, vif, skb);
2020                         read_unlock(&mrt_lock);
2021
2022                         return err2;
2023                 }
2024                 read_unlock(&mrt_lock);
2025                 kfree_skb(skb);
2026                 return -ENODEV;
2027         }
2028
2029         read_lock(&mrt_lock);
2030         ip_mr_forward(net, mrt, skb, cache, local);
2031         read_unlock(&mrt_lock);
2032
2033         if (local)
2034                 return ip_local_deliver(skb);
2035
2036         return 0;
2037
2038 dont_forward:
2039         if (local)
2040                 return ip_local_deliver(skb);
2041         kfree_skb(skb);
2042         return 0;
2043 }
2044
2045 #ifdef CONFIG_IP_PIMSM
2046 /* called with rcu_read_lock() */
2047 static int __pim_rcv(struct mr_table *mrt, struct sk_buff *skb,
2048                      unsigned int pimlen)
2049 {
2050         struct net_device *reg_dev = NULL;
2051         struct iphdr *encap;
2052
2053         encap = (struct iphdr *)(skb_transport_header(skb) + pimlen);
2054         /*
2055          * Check that:
2056          * a. packet is really sent to a multicast group
2057          * b. packet is not a NULL-REGISTER
2058          * c. packet is not truncated
2059          */
2060         if (!ipv4_is_multicast(encap->daddr) ||
2061             encap->tot_len == 0 ||
2062             ntohs(encap->tot_len) + pimlen > skb->len)
2063                 return 1;
2064
2065         read_lock(&mrt_lock);
2066         if (mrt->mroute_reg_vif_num >= 0)
2067                 reg_dev = mrt->vif_table[mrt->mroute_reg_vif_num].dev;
2068         read_unlock(&mrt_lock);
2069
2070         if (!reg_dev)
2071                 return 1;
2072
2073         skb->mac_header = skb->network_header;
2074         skb_pull(skb, (u8 *)encap - skb->data);
2075         skb_reset_network_header(skb);
2076         skb->protocol = htons(ETH_P_IP);
2077         skb->ip_summed = CHECKSUM_NONE;
2078
2079         skb_tunnel_rx(skb, reg_dev, dev_net(reg_dev));
2080
2081         netif_rx(skb);
2082
2083         return NET_RX_SUCCESS;
2084 }
2085 #endif
2086
2087 #ifdef CONFIG_IP_PIMSM_V1
2088 /*
2089  * Handle IGMP messages of PIMv1
2090  */
2091
2092 int pim_rcv_v1(struct sk_buff *skb)
2093 {
2094         struct igmphdr *pim;
2095         struct net *net = dev_net(skb->dev);
2096         struct mr_table *mrt;
2097
2098         if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr)))
2099                 goto drop;
2100
2101         pim = igmp_hdr(skb);
2102
2103         mrt = ipmr_rt_fib_lookup(net, skb);
2104         if (IS_ERR(mrt))
2105                 goto drop;
2106         if (!mrt->mroute_do_pim ||
2107             pim->group != PIM_V1_VERSION || pim->code != PIM_V1_REGISTER)
2108                 goto drop;
2109
2110         if (__pim_rcv(mrt, skb, sizeof(*pim))) {
2111 drop:
2112                 kfree_skb(skb);
2113         }
2114         return 0;
2115 }
2116 #endif
2117
2118 #ifdef CONFIG_IP_PIMSM_V2
2119 static int pim_rcv(struct sk_buff *skb)
2120 {
2121         struct pimreghdr *pim;
2122         struct net *net = dev_net(skb->dev);
2123         struct mr_table *mrt;
2124
2125         if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr)))
2126                 goto drop;
2127
2128         pim = (struct pimreghdr *)skb_transport_header(skb);
2129         if (pim->type != ((PIM_VERSION << 4) | (PIM_REGISTER)) ||
2130             (pim->flags & PIM_NULL_REGISTER) ||
2131             (ip_compute_csum((void *)pim, sizeof(*pim)) != 0 &&
2132              csum_fold(skb_checksum(skb, 0, skb->len, 0))))
2133                 goto drop;
2134
2135         mrt = ipmr_rt_fib_lookup(net, skb);
2136         if (IS_ERR(mrt))
2137                 goto drop;
2138         if (__pim_rcv(mrt, skb, sizeof(*pim))) {
2139 drop:
2140                 kfree_skb(skb);
2141         }
2142         return 0;
2143 }
2144 #endif
2145
2146 static int __ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
2147                               struct mfc_cache *c, struct rtmsg *rtm)
2148 {
2149         int ct;
2150         struct rtnexthop *nhp;
2151         struct nlattr *mp_attr;
2152         struct rta_mfc_stats mfcs;
2153
2154         /* If cache is unresolved, don't try to parse IIF and OIF */
2155         if (c->mfc_parent >= MAXVIFS)
2156                 return -ENOENT;
2157
2158         if (VIF_EXISTS(mrt, c->mfc_parent) &&
2159             nla_put_u32(skb, RTA_IIF, mrt->vif_table[c->mfc_parent].dev->ifindex) < 0)
2160                 return -EMSGSIZE;
2161
2162         if (!(mp_attr = nla_nest_start(skb, RTA_MULTIPATH)))
2163                 return -EMSGSIZE;
2164
2165         for (ct = c->mfc_un.res.minvif; ct < c->mfc_un.res.maxvif; ct++) {
2166                 if (VIF_EXISTS(mrt, ct) && c->mfc_un.res.ttls[ct] < 255) {
2167                         if (!(nhp = nla_reserve_nohdr(skb, sizeof(*nhp)))) {
2168                                 nla_nest_cancel(skb, mp_attr);
2169                                 return -EMSGSIZE;
2170                         }
2171
2172                         nhp->rtnh_flags = 0;
2173                         nhp->rtnh_hops = c->mfc_un.res.ttls[ct];
2174                         nhp->rtnh_ifindex = mrt->vif_table[ct].dev->ifindex;
2175                         nhp->rtnh_len = sizeof(*nhp);
2176                 }
2177         }
2178
2179         nla_nest_end(skb, mp_attr);
2180
2181         mfcs.mfcs_packets = c->mfc_un.res.pkt;
2182         mfcs.mfcs_bytes = c->mfc_un.res.bytes;
2183         mfcs.mfcs_wrong_if = c->mfc_un.res.wrong_if;
2184         if (nla_put(skb, RTA_MFC_STATS, sizeof(mfcs), &mfcs) < 0)
2185                 return -EMSGSIZE;
2186
2187         rtm->rtm_type = RTN_MULTICAST;
2188         return 1;
2189 }
2190
2191 int ipmr_get_route(struct net *net, struct sk_buff *skb,
2192                    __be32 saddr, __be32 daddr,
2193                    struct rtmsg *rtm, int nowait)
2194 {
2195         struct mfc_cache *cache;
2196         struct mr_table *mrt;
2197         int err;
2198
2199         mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2200         if (!mrt)
2201                 return -ENOENT;
2202
2203         rcu_read_lock();
2204         cache = ipmr_cache_find(mrt, saddr, daddr);
2205         if (!cache && skb->dev) {
2206                 int vif = ipmr_find_vif(mrt, skb->dev);
2207
2208                 if (vif >= 0)
2209                         cache = ipmr_cache_find_any(mrt, daddr, vif);
2210         }
2211         if (!cache) {
2212                 struct sk_buff *skb2;
2213                 struct iphdr *iph;
2214                 struct net_device *dev;
2215                 int vif = -1;
2216
2217                 if (nowait) {
2218                         rcu_read_unlock();
2219                         return -EAGAIN;
2220                 }
2221
2222                 dev = skb->dev;
2223                 read_lock(&mrt_lock);
2224                 if (dev)
2225                         vif = ipmr_find_vif(mrt, dev);
2226                 if (vif < 0) {
2227                         read_unlock(&mrt_lock);
2228                         rcu_read_unlock();
2229                         return -ENODEV;
2230                 }
2231                 skb2 = skb_clone(skb, GFP_ATOMIC);
2232                 if (!skb2) {
2233                         read_unlock(&mrt_lock);
2234                         rcu_read_unlock();
2235                         return -ENOMEM;
2236                 }
2237
2238                 skb_push(skb2, sizeof(struct iphdr));
2239                 skb_reset_network_header(skb2);
2240                 iph = ip_hdr(skb2);
2241                 iph->ihl = sizeof(struct iphdr) >> 2;
2242                 iph->saddr = saddr;
2243                 iph->daddr = daddr;
2244                 iph->version = 0;
2245                 err = ipmr_cache_unresolved(mrt, vif, skb2);
2246                 read_unlock(&mrt_lock);
2247                 rcu_read_unlock();
2248                 return err;
2249         }
2250
2251         read_lock(&mrt_lock);
2252         if (!nowait && (rtm->rtm_flags & RTM_F_NOTIFY))
2253                 cache->mfc_flags |= MFC_NOTIFY;
2254         err = __ipmr_fill_mroute(mrt, skb, cache, rtm);
2255         read_unlock(&mrt_lock);
2256         rcu_read_unlock();
2257         return err;
2258 }
2259
2260 static int ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
2261                             u32 portid, u32 seq, struct mfc_cache *c, int cmd,
2262                             int flags)
2263 {
2264         struct nlmsghdr *nlh;
2265         struct rtmsg *rtm;
2266         int err;
2267
2268         nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rtm), flags);
2269         if (!nlh)
2270                 return -EMSGSIZE;
2271
2272         rtm = nlmsg_data(nlh);
2273         rtm->rtm_family   = RTNL_FAMILY_IPMR;
2274         rtm->rtm_dst_len  = 32;
2275         rtm->rtm_src_len  = 32;
2276         rtm->rtm_tos      = 0;
2277         rtm->rtm_table    = mrt->id;
2278         if (nla_put_u32(skb, RTA_TABLE, mrt->id))
2279                 goto nla_put_failure;
2280         rtm->rtm_type     = RTN_MULTICAST;
2281         rtm->rtm_scope    = RT_SCOPE_UNIVERSE;
2282         if (c->mfc_flags & MFC_STATIC)
2283                 rtm->rtm_protocol = RTPROT_STATIC;
2284         else
2285                 rtm->rtm_protocol = RTPROT_MROUTED;
2286         rtm->rtm_flags    = 0;
2287
2288         if (nla_put_in_addr(skb, RTA_SRC, c->mfc_origin) ||
2289             nla_put_in_addr(skb, RTA_DST, c->mfc_mcastgrp))
2290                 goto nla_put_failure;
2291         err = __ipmr_fill_mroute(mrt, skb, c, rtm);
2292         /* do not break the dump if cache is unresolved */
2293         if (err < 0 && err != -ENOENT)
2294                 goto nla_put_failure;
2295
2296         nlmsg_end(skb, nlh);
2297         return 0;
2298
2299 nla_put_failure:
2300         nlmsg_cancel(skb, nlh);
2301         return -EMSGSIZE;
2302 }
2303
2304 static size_t mroute_msgsize(bool unresolved, int maxvif)
2305 {
2306         size_t len =
2307                 NLMSG_ALIGN(sizeof(struct rtmsg))
2308                 + nla_total_size(4)     /* RTA_TABLE */
2309                 + nla_total_size(4)     /* RTA_SRC */
2310                 + nla_total_size(4)     /* RTA_DST */
2311                 ;
2312
2313         if (!unresolved)
2314                 len = len
2315                       + nla_total_size(4)       /* RTA_IIF */
2316                       + nla_total_size(0)       /* RTA_MULTIPATH */
2317                       + maxvif * NLA_ALIGN(sizeof(struct rtnexthop))
2318                                                 /* RTA_MFC_STATS */
2319                       + nla_total_size(sizeof(struct rta_mfc_stats))
2320                 ;
2321
2322         return len;
2323 }
2324
2325 static void mroute_netlink_event(struct mr_table *mrt, struct mfc_cache *mfc,
2326                                  int cmd)
2327 {
2328         struct net *net = read_pnet(&mrt->net);
2329         struct sk_buff *skb;
2330         int err = -ENOBUFS;
2331
2332         skb = nlmsg_new(mroute_msgsize(mfc->mfc_parent >= MAXVIFS, mrt->maxvif),
2333                         GFP_ATOMIC);
2334         if (!skb)
2335                 goto errout;
2336
2337         err = ipmr_fill_mroute(mrt, skb, 0, 0, mfc, cmd, 0);
2338         if (err < 0)
2339                 goto errout;
2340
2341         rtnl_notify(skb, net, 0, RTNLGRP_IPV4_MROUTE, NULL, GFP_ATOMIC);
2342         return;
2343
2344 errout:
2345         kfree_skb(skb);
2346         if (err < 0)
2347                 rtnl_set_sk_err(net, RTNLGRP_IPV4_MROUTE, err);
2348 }
2349
2350 static int ipmr_rtm_dumproute(struct sk_buff *skb, struct netlink_callback *cb)
2351 {
2352         struct net *net = sock_net(skb->sk);
2353         struct mr_table *mrt;
2354         struct mfc_cache *mfc;
2355         unsigned int t = 0, s_t;
2356         unsigned int h = 0, s_h;
2357         unsigned int e = 0, s_e;
2358
2359         s_t = cb->args[0];
2360         s_h = cb->args[1];
2361         s_e = cb->args[2];
2362
2363         rcu_read_lock();
2364         ipmr_for_each_table(mrt, net) {
2365                 if (t < s_t)
2366                         goto next_table;
2367                 if (t > s_t)
2368                         s_h = 0;
2369                 for (h = s_h; h < MFC_LINES; h++) {
2370                         list_for_each_entry_rcu(mfc, &mrt->mfc_cache_array[h], list) {
2371                                 if (e < s_e)
2372                                         goto next_entry;
2373                                 if (ipmr_fill_mroute(mrt, skb,
2374                                                      NETLINK_CB(cb->skb).portid,
2375                                                      cb->nlh->nlmsg_seq,
2376                                                      mfc, RTM_NEWROUTE,
2377                                                      NLM_F_MULTI) < 0)
2378                                         goto done;
2379 next_entry:
2380                                 e++;
2381                         }
2382                         e = s_e = 0;
2383                 }
2384                 spin_lock_bh(&mfc_unres_lock);
2385                 list_for_each_entry(mfc, &mrt->mfc_unres_queue, list) {
2386                         if (e < s_e)
2387                                 goto next_entry2;
2388                         if (ipmr_fill_mroute(mrt, skb,
2389                                              NETLINK_CB(cb->skb).portid,
2390                                              cb->nlh->nlmsg_seq,
2391                                              mfc, RTM_NEWROUTE,
2392                                              NLM_F_MULTI) < 0) {
2393                                 spin_unlock_bh(&mfc_unres_lock);
2394                                 goto done;
2395                         }
2396 next_entry2:
2397                         e++;
2398                 }
2399                 spin_unlock_bh(&mfc_unres_lock);
2400                 e = s_e = 0;
2401                 s_h = 0;
2402 next_table:
2403                 t++;
2404         }
2405 done:
2406         rcu_read_unlock();
2407
2408         cb->args[2] = e;
2409         cb->args[1] = h;
2410         cb->args[0] = t;
2411
2412         return skb->len;
2413 }
2414
2415 #ifdef CONFIG_PROC_FS
2416 /*
2417  *      The /proc interfaces to multicast routing :
2418  *      /proc/net/ip_mr_cache & /proc/net/ip_mr_vif
2419  */
2420 struct ipmr_vif_iter {
2421         struct seq_net_private p;
2422         struct mr_table *mrt;
2423         int ct;
2424 };
2425
2426 static struct vif_device *ipmr_vif_seq_idx(struct net *net,
2427                                            struct ipmr_vif_iter *iter,
2428                                            loff_t pos)
2429 {
2430         struct mr_table *mrt = iter->mrt;
2431
2432         for (iter->ct = 0; iter->ct < mrt->maxvif; ++iter->ct) {
2433                 if (!VIF_EXISTS(mrt, iter->ct))
2434                         continue;
2435                 if (pos-- == 0)
2436                         return &mrt->vif_table[iter->ct];
2437         }
2438         return NULL;
2439 }
2440
2441 static void *ipmr_vif_seq_start(struct seq_file *seq, loff_t *pos)
2442         __acquires(mrt_lock)
2443 {
2444         struct ipmr_vif_iter *iter = seq->private;
2445         struct net *net = seq_file_net(seq);
2446         struct mr_table *mrt;
2447
2448         mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2449         if (!mrt)
2450                 return ERR_PTR(-ENOENT);
2451
2452         iter->mrt = mrt;
2453
2454         read_lock(&mrt_lock);
2455         return *pos ? ipmr_vif_seq_idx(net, seq->private, *pos - 1)
2456                 : SEQ_START_TOKEN;
2457 }
2458
2459 static void *ipmr_vif_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2460 {
2461         struct ipmr_vif_iter *iter = seq->private;
2462         struct net *net = seq_file_net(seq);
2463         struct mr_table *mrt = iter->mrt;
2464
2465         ++*pos;
2466         if (v == SEQ_START_TOKEN)
2467                 return ipmr_vif_seq_idx(net, iter, 0);
2468
2469         while (++iter->ct < mrt->maxvif) {
2470                 if (!VIF_EXISTS(mrt, iter->ct))
2471                         continue;
2472                 return &mrt->vif_table[iter->ct];
2473         }
2474         return NULL;
2475 }
2476
2477 static void ipmr_vif_seq_stop(struct seq_file *seq, void *v)
2478         __releases(mrt_lock)
2479 {
2480         read_unlock(&mrt_lock);
2481 }
2482
2483 static int ipmr_vif_seq_show(struct seq_file *seq, void *v)
2484 {
2485         struct ipmr_vif_iter *iter = seq->private;
2486         struct mr_table *mrt = iter->mrt;
2487
2488         if (v == SEQ_START_TOKEN) {
2489                 seq_puts(seq,
2490                          "Interface      BytesIn  PktsIn  BytesOut PktsOut Flags Local    Remote\n");
2491         } else {
2492                 const struct vif_device *vif = v;
2493                 const char *name =  vif->dev ? vif->dev->name : "none";
2494
2495                 seq_printf(seq,
2496                            "%2Zd %-10s %8ld %7ld  %8ld %7ld %05X %08X %08X\n",
2497                            vif - mrt->vif_table,
2498                            name, vif->bytes_in, vif->pkt_in,
2499                            vif->bytes_out, vif->pkt_out,
2500                            vif->flags, vif->local, vif->remote);
2501         }
2502         return 0;
2503 }
2504
2505 static const struct seq_operations ipmr_vif_seq_ops = {
2506         .start = ipmr_vif_seq_start,
2507         .next  = ipmr_vif_seq_next,
2508         .stop  = ipmr_vif_seq_stop,
2509         .show  = ipmr_vif_seq_show,
2510 };
2511
2512 static int ipmr_vif_open(struct inode *inode, struct file *file)
2513 {
2514         return seq_open_net(inode, file, &ipmr_vif_seq_ops,
2515                             sizeof(struct ipmr_vif_iter));
2516 }
2517
2518 static const struct file_operations ipmr_vif_fops = {
2519         .owner   = THIS_MODULE,
2520         .open    = ipmr_vif_open,
2521         .read    = seq_read,
2522         .llseek  = seq_lseek,
2523         .release = seq_release_net,
2524 };
2525
2526 struct ipmr_mfc_iter {
2527         struct seq_net_private p;
2528         struct mr_table *mrt;
2529         struct list_head *cache;
2530         int ct;
2531 };
2532
2533
2534 static struct mfc_cache *ipmr_mfc_seq_idx(struct net *net,
2535                                           struct ipmr_mfc_iter *it, loff_t pos)
2536 {
2537         struct mr_table *mrt = it->mrt;
2538         struct mfc_cache *mfc;
2539
2540         rcu_read_lock();
2541         for (it->ct = 0; it->ct < MFC_LINES; it->ct++) {
2542                 it->cache = &mrt->mfc_cache_array[it->ct];
2543                 list_for_each_entry_rcu(mfc, it->cache, list)
2544                         if (pos-- == 0)
2545                                 return mfc;
2546         }
2547         rcu_read_unlock();
2548
2549         spin_lock_bh(&mfc_unres_lock);
2550         it->cache = &mrt->mfc_unres_queue;
2551         list_for_each_entry(mfc, it->cache, list)
2552                 if (pos-- == 0)
2553                         return mfc;
2554         spin_unlock_bh(&mfc_unres_lock);
2555
2556         it->cache = NULL;
2557         return NULL;
2558 }
2559
2560
2561 static void *ipmr_mfc_seq_start(struct seq_file *seq, loff_t *pos)
2562 {
2563         struct ipmr_mfc_iter *it = seq->private;
2564         struct net *net = seq_file_net(seq);
2565         struct mr_table *mrt;
2566
2567         mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2568         if (!mrt)
2569                 return ERR_PTR(-ENOENT);
2570
2571         it->mrt = mrt;
2572         it->cache = NULL;
2573         it->ct = 0;
2574         return *pos ? ipmr_mfc_seq_idx(net, seq->private, *pos - 1)
2575                 : SEQ_START_TOKEN;
2576 }
2577
2578 static void *ipmr_mfc_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2579 {
2580         struct mfc_cache *mfc = v;
2581         struct ipmr_mfc_iter *it = seq->private;
2582         struct net *net = seq_file_net(seq);
2583         struct mr_table *mrt = it->mrt;
2584
2585         ++*pos;
2586
2587         if (v == SEQ_START_TOKEN)
2588                 return ipmr_mfc_seq_idx(net, seq->private, 0);
2589
2590         if (mfc->list.next != it->cache)
2591                 return list_entry(mfc->list.next, struct mfc_cache, list);
2592
2593         if (it->cache == &mrt->mfc_unres_queue)
2594                 goto end_of_list;
2595
2596         BUG_ON(it->cache != &mrt->mfc_cache_array[it->ct]);
2597
2598         while (++it->ct < MFC_LINES) {
2599                 it->cache = &mrt->mfc_cache_array[it->ct];
2600                 if (list_empty(it->cache))
2601                         continue;
2602                 return list_first_entry(it->cache, struct mfc_cache, list);
2603         }
2604
2605         /* exhausted cache_array, show unresolved */
2606         rcu_read_unlock();
2607         it->cache = &mrt->mfc_unres_queue;
2608         it->ct = 0;
2609
2610         spin_lock_bh(&mfc_unres_lock);
2611         if (!list_empty(it->cache))
2612                 return list_first_entry(it->cache, struct mfc_cache, list);
2613
2614 end_of_list:
2615         spin_unlock_bh(&mfc_unres_lock);
2616         it->cache = NULL;
2617
2618         return NULL;
2619 }
2620
2621 static void ipmr_mfc_seq_stop(struct seq_file *seq, void *v)
2622 {
2623         struct ipmr_mfc_iter *it = seq->private;
2624         struct mr_table *mrt = it->mrt;
2625
2626         if (it->cache == &mrt->mfc_unres_queue)
2627                 spin_unlock_bh(&mfc_unres_lock);
2628         else if (it->cache == &mrt->mfc_cache_array[it->ct])
2629                 rcu_read_unlock();
2630 }
2631
2632 static int ipmr_mfc_seq_show(struct seq_file *seq, void *v)
2633 {
2634         int n;
2635
2636         if (v == SEQ_START_TOKEN) {
2637                 seq_puts(seq,
2638                  "Group    Origin   Iif     Pkts    Bytes    Wrong Oifs\n");
2639         } else {
2640                 const struct mfc_cache *mfc = v;
2641                 const struct ipmr_mfc_iter *it = seq->private;
2642                 const struct mr_table *mrt = it->mrt;
2643
2644                 seq_printf(seq, "%08X %08X %-3hd",
2645                            (__force u32) mfc->mfc_mcastgrp,
2646                            (__force u32) mfc->mfc_origin,
2647                            mfc->mfc_parent);
2648
2649                 if (it->cache != &mrt->mfc_unres_queue) {
2650                         seq_printf(seq, " %8lu %8lu %8lu",
2651                                    mfc->mfc_un.res.pkt,
2652                                    mfc->mfc_un.res.bytes,
2653                                    mfc->mfc_un.res.wrong_if);
2654                         for (n = mfc->mfc_un.res.minvif;
2655                              n < mfc->mfc_un.res.maxvif; n++) {
2656                                 if (VIF_EXISTS(mrt, n) &&
2657                                     mfc->mfc_un.res.ttls[n] < 255)
2658                                         seq_printf(seq,
2659                                            " %2d:%-3d",
2660                                            n, mfc->mfc_un.res.ttls[n]);
2661                         }
2662                 } else {
2663                         /* unresolved mfc_caches don't contain
2664                          * pkt, bytes and wrong_if values
2665                          */
2666                         seq_printf(seq, " %8lu %8lu %8lu", 0ul, 0ul, 0ul);
2667                 }
2668                 seq_putc(seq, '\n');
2669         }
2670         return 0;
2671 }
2672
2673 static const struct seq_operations ipmr_mfc_seq_ops = {
2674         .start = ipmr_mfc_seq_start,
2675         .next  = ipmr_mfc_seq_next,
2676         .stop  = ipmr_mfc_seq_stop,
2677         .show  = ipmr_mfc_seq_show,
2678 };
2679
2680 static int ipmr_mfc_open(struct inode *inode, struct file *file)
2681 {
2682         return seq_open_net(inode, file, &ipmr_mfc_seq_ops,
2683                             sizeof(struct ipmr_mfc_iter));
2684 }
2685
2686 static const struct file_operations ipmr_mfc_fops = {
2687         .owner   = THIS_MODULE,
2688         .open    = ipmr_mfc_open,
2689         .read    = seq_read,
2690         .llseek  = seq_lseek,
2691         .release = seq_release_net,
2692 };
2693 #endif
2694
2695 #ifdef CONFIG_IP_PIMSM_V2
2696 static const struct net_protocol pim_protocol = {
2697         .handler        =       pim_rcv,
2698         .netns_ok       =       1,
2699 };
2700 #endif
2701
2702
2703 /*
2704  *      Setup for IP multicast routing
2705  */
2706 static int __net_init ipmr_net_init(struct net *net)
2707 {
2708         int err;
2709
2710         err = ipmr_rules_init(net);
2711         if (err < 0)
2712                 goto fail;
2713
2714 #ifdef CONFIG_PROC_FS
2715         err = -ENOMEM;
2716         if (!proc_create("ip_mr_vif", 0, net->proc_net, &ipmr_vif_fops))
2717                 goto proc_vif_fail;
2718         if (!proc_create("ip_mr_cache", 0, net->proc_net, &ipmr_mfc_fops))
2719                 goto proc_cache_fail;
2720 #endif
2721         return 0;
2722
2723 #ifdef CONFIG_PROC_FS
2724 proc_cache_fail:
2725         remove_proc_entry("ip_mr_vif", net->proc_net);
2726 proc_vif_fail:
2727         ipmr_rules_exit(net);
2728 #endif
2729 fail:
2730         return err;
2731 }
2732
2733 static void __net_exit ipmr_net_exit(struct net *net)
2734 {
2735 #ifdef CONFIG_PROC_FS
2736         remove_proc_entry("ip_mr_cache", net->proc_net);
2737         remove_proc_entry("ip_mr_vif", net->proc_net);
2738 #endif
2739         ipmr_rules_exit(net);
2740 }
2741
2742 static struct pernet_operations ipmr_net_ops = {
2743         .init = ipmr_net_init,
2744         .exit = ipmr_net_exit,
2745 };
2746
2747 int __init ip_mr_init(void)
2748 {
2749         int err;
2750
2751         mrt_cachep = kmem_cache_create("ip_mrt_cache",
2752                                        sizeof(struct mfc_cache),
2753                                        0, SLAB_HWCACHE_ALIGN | SLAB_PANIC,
2754                                        NULL);
2755         if (!mrt_cachep)
2756                 return -ENOMEM;
2757
2758         err = register_pernet_subsys(&ipmr_net_ops);
2759         if (err)
2760                 goto reg_pernet_fail;
2761
2762         err = register_netdevice_notifier(&ip_mr_notifier);
2763         if (err)
2764                 goto reg_notif_fail;
2765 #ifdef CONFIG_IP_PIMSM_V2
2766         if (inet_add_protocol(&pim_protocol, IPPROTO_PIM) < 0) {
2767                 pr_err("%s: can't add PIM protocol\n", __func__);
2768                 err = -EAGAIN;
2769                 goto add_proto_fail;
2770         }
2771 #endif
2772         rtnl_register(RTNL_FAMILY_IPMR, RTM_GETROUTE,
2773                       NULL, ipmr_rtm_dumproute, NULL);
2774         return 0;
2775
2776 #ifdef CONFIG_IP_PIMSM_V2
2777 add_proto_fail:
2778         unregister_netdevice_notifier(&ip_mr_notifier);
2779 #endif
2780 reg_notif_fail:
2781         unregister_pernet_subsys(&ipmr_net_ops);
2782 reg_pernet_fail:
2783         kmem_cache_destroy(mrt_cachep);
2784         return err;
2785 }