net/ipv4/ipmr.c

   1 /*
   2  *      IP multicast routing support for mrouted 3.6/3.8
   3  *
   4  *              (c) 1995 Alan Cox, <alan@lxorguk.ukuu.org.uk>
   5  *        Linux Consultancy and Custom Driver Development
   6  *
   7  *      This program is free software; you can redistribute it and/or
   8  *      modify it under the terms of the GNU General Public License
   9  *      as published by the Free Software Foundation; either version
  10  *      2 of the License, or (at your option) any later version.
  11  *
  12  *      Fixes:
  13  *      Michael Chastain        :       Incorrect size of copying.
  14  *      Alan Cox                :       Added the cache manager code
  15  *      Alan Cox                :       Fixed the clone/copy bug and device race.
  16  *      Mike McLagan            :       Routing by source
  17  *      Malcolm Beattie         :       Buffer handling fixes.
  18  *      Alexey Kuznetsov        :       Double buffer free and other fixes.
  19  *      SVR Anand               :       Fixed several multicast bugs and problems.
  20  *      Alexey Kuznetsov        :       Status, optimisations and more.
  21  *      Brad Parker             :       Better behaviour on mrouted upcall
  22  *                                      overflow.
  23  *      Carlos Picoto           :       PIMv1 Support
  24  *      Pavlin Ivanov Radoslavov:       PIMv2 Registers must checksum only PIM header
  25  *                                      Relax this requirement to work with older peers.
  26  *
  27  */
  28
  29 #include <asm/uaccess.h>
  30 #include <linux/types.h>
  31 #include <linux/capability.h>
  32 #include <linux/errno.h>
  33 #include <linux/timer.h>
  34 #include <linux/mm.h>
  35 #include <linux/kernel.h>
  36 #include <linux/fcntl.h>
  37 #include <linux/stat.h>
  38 #include <linux/socket.h>
  39 #include <linux/in.h>
  40 #include <linux/inet.h>
  41 #include <linux/netdevice.h>
  42 #include <linux/inetdevice.h>
  43 #include <linux/igmp.h>
  44 #include <linux/proc_fs.h>
  45 #include <linux/seq_file.h>
  46 #include <linux/mroute.h>
  47 #include <linux/init.h>
  48 #include <linux/if_ether.h>
  49 #include <linux/slab.h>
  50 #include <net/net_namespace.h>
  51 #include <net/ip.h>
  52 #include <net/protocol.h>
  53 #include <linux/skbuff.h>
  54 #include <net/route.h>
  55 #include <net/sock.h>
  56 #include <net/icmp.h>
  57 #include <net/udp.h>
  58 #include <net/raw.h>
  59 #include <linux/notifier.h>
  60 #include <linux/if_arp.h>
  61 #include <linux/netfilter_ipv4.h>
  62 #include <linux/compat.h>
  63 #include <linux/export.h>
  64 #include <net/ip_tunnels.h>
  65 #include <net/checksum.h>
  66 #include <net/netlink.h>
  67 #include <net/fib_rules.h>
  68 #include <linux/netconf.h>
  69
  70 #if defined(CONFIG_IP_PIMSM_V1) || defined(CONFIG_IP_PIMSM_V2)
  71 #define CONFIG_IP_PIMSM 1
  72 #endif
  73
  74 struct mr_table {
  75         struct list_head        list;
  76 #ifdef CONFIG_NET_NS
  77         struct net              *net;
  78 #endif
  79         u32                     id;
  80         struct sock __rcu       *mroute_sk;
  81         struct timer_list       ipmr_expire_timer;
  82         struct list_head        mfc_unres_queue;
  83         struct list_head        mfc_cache_array[MFC_LINES];
  84         struct vif_device       vif_table[MAXVIFS];
  85         int                     maxvif;
  86         atomic_t                cache_resolve_queue_len;
  87         bool                    mroute_do_assert;
  88         bool                    mroute_do_pim;
  89 #if defined(CONFIG_IP_PIMSM_V1) || defined(CONFIG_IP_PIMSM_V2)
  90         int                     mroute_reg_vif_num;
  91 #endif
  92 };
  93
  94 struct ipmr_rule {
  95         struct fib_rule         common;
  96 };
  97
  98 struct ipmr_result {
  99         struct mr_table         *mrt;
 100 };
 101
 102 /* Big lock, protecting vif table, mrt cache and mroute socket state.
 103  * Note that the changes are semaphored via rtnl_lock.
 104  */
 105
 106 static DEFINE_RWLOCK(mrt_lock);
 107
 108 /*
 109  *      Multicast router control variables
 110  */
 111
 112 #define VIF_EXISTS(_mrt, _idx) ((_mrt)->vif_table[_idx].dev != NULL)
 113
 114 /* Special spinlock for queue of unresolved entries */
 115 static DEFINE_SPINLOCK(mfc_unres_lock);
 116
 117 /* We return to original Alan's scheme. Hash table of resolved
 118  * entries is changed only in process context and protected
 119  * with weak lock mrt_lock. Queue of unresolved entries is protected
 120  * with strong spinlock mfc_unres_lock.
 121  *
 122  * In this case data path is free of exclusive locks at all.
 123  */
 124
 125 static struct kmem_cache *mrt_cachep __read_mostly;
 126
 127 static struct mr_table *ipmr_new_table(struct net *net, u32 id);
 128 static void ipmr_free_table(struct mr_table *mrt);
 129
 130 static void ip_mr_forward(struct net *net, struct mr_table *mrt,
 131                           struct sk_buff *skb, struct mfc_cache *cache,
 132                           int local);
 133 static int ipmr_cache_report(struct mr_table *mrt,
 134                              struct sk_buff *pkt, vifi_t vifi, int assert);
 135 static int __ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
 136                               struct mfc_cache *c, struct rtmsg *rtm);
 137 static void mroute_netlink_event(struct mr_table *mrt, struct mfc_cache *mfc,
 138                                  int cmd);
 139 static void mroute_clean_tables(struct mr_table *mrt);
 140 static void ipmr_expire_process(unsigned long arg);
 141
 142 #ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES
 143 #define ipmr_for_each_table(mrt, net) \
 144         list_for_each_entry_rcu(mrt, &net->ipv4.mr_tables, list)
 145
 146 static struct mr_table *ipmr_get_table(struct net *net, u32 id)
 147 {
 148         struct mr_table *mrt;
 149
 150         ipmr_for_each_table(mrt, net) {
 151                 if (mrt->id == id)
 152                         return mrt;
 153         }
 154         return NULL;
 155 }
 156
 157 static int ipmr_fib_lookup(struct net *net, struct flowi4 *flp4,
 158                            struct mr_table **mrt)
 159 {
 160         int err;
 161         struct ipmr_result res;
 162         struct fib_lookup_arg arg = {
 163                 .result = &res,
 164                 .flags = FIB_LOOKUP_NOREF,
 165         };
 166
 167         err = fib_rules_lookup(net->ipv4.mr_rules_ops,
 168                                flowi4_to_flowi(flp4), 0, &arg);
 169         if (err < 0)
 170                 return err;
 171         *mrt = res.mrt;
 172         return 0;
 173 }
 174
 175 static int ipmr_rule_action(struct fib_rule *rule, struct flowi *flp,
 176                             int flags, struct fib_lookup_arg *arg)
 177 {
 178         struct ipmr_result *res = arg->result;
 179         struct mr_table *mrt;
 180
 181         switch (rule->action) {
 182         case FR_ACT_TO_TBL:
 183                 break;
 184         case FR_ACT_UNREACHABLE:
 185                 return -ENETUNREACH;
 186         case FR_ACT_PROHIBIT:
 187                 return -EACCES;
 188         case FR_ACT_BLACKHOLE:
 189         default:
 190                 return -EINVAL;
 191         }
 192
 193         mrt = ipmr_get_table(rule->fr_net, rule->table);
 194         if (mrt == NULL)
 195                 return -EAGAIN;
 196         res->mrt = mrt;
 197         return 0;
 198 }
 199
 200 static int ipmr_rule_match(struct fib_rule *rule, struct flowi *fl, int flags)
 201 {
 202         return 1;
 203 }
 204
 205 static const struct nla_policy ipmr_rule_policy[FRA_MAX + 1] = {
 206         FRA_GENERIC_POLICY,
 207 };
 208
 209 static int ipmr_rule_configure(struct fib_rule *rule, struct sk_buff *skb,
 210                                struct fib_rule_hdr *frh, struct nlattr **tb)
 211 {
 212         return 0;
 213 }
 214
 215 static int ipmr_rule_compare(struct fib_rule *rule, struct fib_rule_hdr *frh,
 216                              struct nlattr **tb)
 217 {
 218         return 1;
 219 }
 220
 221 static int ipmr_rule_fill(struct fib_rule *rule, struct sk_buff *skb,
 222                           struct fib_rule_hdr *frh)
 223 {
 224         frh->dst_len = 0;
 225         frh->src_len = 0;
 226         frh->tos     = 0;
 227         return 0;
 228 }
 229
 230 static const struct fib_rules_ops __net_initconst ipmr_rules_ops_template = {
 231         .family         = RTNL_FAMILY_IPMR,
 232         .rule_size      = sizeof(struct ipmr_rule),
 233         .addr_size      = sizeof(u32),
 234         .action         = ipmr_rule_action,
 235         .match          = ipmr_rule_match,
 236         .configure      = ipmr_rule_configure,
 237         .compare        = ipmr_rule_compare,
 238         .default_pref   = fib_default_rule_pref,
 239         .fill           = ipmr_rule_fill,
 240         .nlgroup        = RTNLGRP_IPV4_RULE,
 241         .policy         = ipmr_rule_policy,
 242         .owner          = THIS_MODULE,
 243 };
 244
 245 static int __net_init ipmr_rules_init(struct net *net)
 246 {
 247         struct fib_rules_ops *ops;
 248         struct mr_table *mrt;
 249         int err;
 250
 251         ops = fib_rules_register(&ipmr_rules_ops_template, net);
 252         if (IS_ERR(ops))
 253                 return PTR_ERR(ops);
 254
 255         INIT_LIST_HEAD(&net->ipv4.mr_tables);
 256
 257         mrt = ipmr_new_table(net, RT_TABLE_DEFAULT);
 258         if (mrt == NULL) {
 259                 err = -ENOMEM;
 260                 goto err1;
 261         }
 262
 263         err = fib_default_rule_add(ops, 0x7fff, RT_TABLE_DEFAULT, 0);
 264         if (err < 0)
 265                 goto err2;
 266
 267         net->ipv4.mr_rules_ops = ops;
 268         return 0;
 269
 270 err2:
 271         kfree(mrt);
 272 err1:
 273         fib_rules_unregister(ops);
 274         return err;
 275 }
 276
 277 static void __net_exit ipmr_rules_exit(struct net *net)
 278 {
 279         struct mr_table *mrt, *next;
 280
 281         list_for_each_entry_safe(mrt, next, &net->ipv4.mr_tables, list) {
 282                 list_del(&mrt->list);
 283                 ipmr_free_table(mrt);
 284         }
 285         fib_rules_unregister(net->ipv4.mr_rules_ops);
 286 }
 287 #else
 288 #define ipmr_for_each_table(mrt, net) \
 289         for (mrt = net->ipv4.mrt; mrt; mrt = NULL)
 290
 291 static struct mr_table *ipmr_get_table(struct net *net, u32 id)
 292 {
 293         return net->ipv4.mrt;
 294 }
 295
 296 static int ipmr_fib_lookup(struct net *net, struct flowi4 *flp4,
 297                            struct mr_table **mrt)
 298 {
 299         *mrt = net->ipv4.mrt;
 300         return 0;
 301 }
 302
 303 static int __net_init ipmr_rules_init(struct net *net)
 304 {
 305         net->ipv4.mrt = ipmr_new_table(net, RT_TABLE_DEFAULT);
 306         return net->ipv4.mrt ? 0 : -ENOMEM;
 307 }
 308
 309 static void __net_exit ipmr_rules_exit(struct net *net)
 310 {
 311         ipmr_free_table(net->ipv4.mrt);
 312 }
 313 #endif
 314
 315 static struct mr_table *ipmr_new_table(struct net *net, u32 id)
 316 {
 317         struct mr_table *mrt;
 318         unsigned int i;
 319
 320         mrt = ipmr_get_table(net, id);
 321         if (mrt != NULL)
 322                 return mrt;
 323
 324         mrt = kzalloc(sizeof(*mrt), GFP_KERNEL);
 325         if (mrt == NULL)
 326                 return NULL;
 327         write_pnet(&mrt->net, net);
 328         mrt->id = id;
 329
 330         /* Forwarding cache */
 331         for (i = 0; i < MFC_LINES; i++)
 332                 INIT_LIST_HEAD(&mrt->mfc_cache_array[i]);
 333
 334         INIT_LIST_HEAD(&mrt->mfc_unres_queue);
 335
 336         setup_timer(&mrt->ipmr_expire_timer, ipmr_expire_process,
 337                     (unsigned long)mrt);
 338
 339 #ifdef CONFIG_IP_PIMSM
 340         mrt->mroute_reg_vif_num = -1;
 341 #endif
 342 #ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES
 343         list_add_tail_rcu(&mrt->list, &net->ipv4.mr_tables);
 344 #endif
 345         return mrt;
 346 }
 347
 348 static void ipmr_free_table(struct mr_table *mrt)
 349 {
 350         del_timer_sync(&mrt->ipmr_expire_timer);
 351         mroute_clean_tables(mrt);
 352         kfree(mrt);
 353 }
 354
 355 /* Service routines creating virtual interfaces: DVMRP tunnels and PIMREG */
 356
 357 static void ipmr_del_tunnel(struct net_device *dev, struct vifctl *v)
 358 {
 359         struct net *net = dev_net(dev);
 360
 361         dev_close(dev);
 362
 363         dev = __dev_get_by_name(net, "tunl0");
 364         if (dev) {
 365                 const struct net_device_ops *ops = dev->netdev_ops;
 366                 struct ifreq ifr;
 367                 struct ip_tunnel_parm p;
 368
 369                 memset(&p, 0, sizeof(p));
 370                 p.iph.daddr = v->vifc_rmt_addr.s_addr;
 371                 p.iph.saddr = v->vifc_lcl_addr.s_addr;
 372                 p.iph.version = 4;
 373                 p.iph.ihl = 5;
 374                 p.iph.protocol = IPPROTO_IPIP;
 375                 sprintf(p.name, "dvmrp%d", v->vifc_vifi);
 376                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
 377
 378                 if (ops->ndo_do_ioctl) {
 379                         mm_segment_t oldfs = get_fs();
 380
 381                         set_fs(KERNEL_DS);
 382                         ops->ndo_do_ioctl(dev, &ifr, SIOCDELTUNNEL);
 383                         set_fs(oldfs);
 384                 }
 385         }
 386 }
 387
 388 static
 389 struct net_device *ipmr_new_tunnel(struct net *net, struct vifctl *v)
 390 {
 391         struct net_device  *dev;
 392
 393         dev = __dev_get_by_name(net, "tunl0");
 394
 395         if (dev) {
 396                 const struct net_device_ops *ops = dev->netdev_ops;
 397                 int err;
 398                 struct ifreq ifr;
 399                 struct ip_tunnel_parm p;
 400                 struct in_device  *in_dev;
 401
 402                 memset(&p, 0, sizeof(p));
 403                 p.iph.daddr = v->vifc_rmt_addr.s_addr;
 404                 p.iph.saddr = v->vifc_lcl_addr.s_addr;
 405                 p.iph.version = 4;
 406                 p.iph.ihl = 5;
 407                 p.iph.protocol = IPPROTO_IPIP;
 408                 sprintf(p.name, "dvmrp%d", v->vifc_vifi);
 409                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
 410
 411                 if (ops->ndo_do_ioctl) {
 412                         mm_segment_t oldfs = get_fs();
 413
 414                         set_fs(KERNEL_DS);
 415                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
 416                         set_fs(oldfs);
 417                 } else {
 418                         err = -EOPNOTSUPP;
 419                 }
 420                 dev = NULL;
 421
 422                 if (err == 0 &&
 423                     (dev = __dev_get_by_name(net, p.name)) != NULL) {
 424                         dev->flags |= IFF_MULTICAST;
 425
 426                         in_dev = __in_dev_get_rtnl(dev);
 427                         if (in_dev == NULL)
 428                                 goto failure;
 429
 430                         ipv4_devconf_setall(in_dev);
 431                         IPV4_DEVCONF(in_dev->cnf, RP_FILTER) = 0;
 432
 433                         if (dev_open(dev))
 434                                 goto failure;
 435                         dev_hold(dev);
 436                 }
 437         }
 438         return dev;
 439
 440 failure:
 441         /* allow the register to be completed before unregistering. */
 442         rtnl_unlock();
 443         rtnl_lock();
 444
 445         unregister_netdevice(dev);
 446         return NULL;
 447 }
 448
 449 #ifdef CONFIG_IP_PIMSM
 450
 451 static netdev_tx_t reg_vif_xmit(struct sk_buff *skb, struct net_device *dev)
 452 {
 453         struct net *net = dev_net(dev);
 454         struct mr_table *mrt;
 455         struct flowi4 fl4 = {
 456                 .flowi4_oif     = dev->ifindex,
 457                 .flowi4_iif     = skb->skb_iif,
 458                 .flowi4_mark    = skb->mark,
 459         };
 460         int err;
 461
 462         err = ipmr_fib_lookup(net, &fl4, &mrt);
 463         if (err < 0) {
 464                 kfree_skb(skb);
 465                 return err;
 466         }
 467
 468         read_lock(&mrt_lock);
 469         dev->stats.tx_bytes += skb->len;
 470         dev->stats.tx_packets++;
 471         ipmr_cache_report(mrt, skb, mrt->mroute_reg_vif_num, IGMPMSG_WHOLEPKT);
 472         read_unlock(&mrt_lock);
 473         kfree_skb(skb);
 474         return NETDEV_TX_OK;
 475 }
 476
 477 static const struct net_device_ops reg_vif_netdev_ops = {
 478         .ndo_start_xmit = reg_vif_xmit,
 479 };
 480
 481 static void reg_vif_setup(struct net_device *dev)
 482 {
 483         dev->type               = ARPHRD_PIMREG;
 484         dev->mtu                = ETH_DATA_LEN - sizeof(struct iphdr) - 8;
 485         dev->flags              = IFF_NOARP;
 486         dev->netdev_ops         = &reg_vif_netdev_ops,
 487         dev->destructor         = free_netdev;
 488         dev->features           |= NETIF_F_NETNS_LOCAL;
 489 }
 490
 491 static struct net_device *ipmr_reg_vif(struct net *net, struct mr_table *mrt)
 492 {
 493         struct net_device *dev;
 494         struct in_device *in_dev;
 495         char name[IFNAMSIZ];
 496
 497         if (mrt->id == RT_TABLE_DEFAULT)
 498                 sprintf(name, "pimreg");
 499         else
 500                 sprintf(name, "pimreg%u", mrt->id);
 501
 502         dev = alloc_netdev(0, name, reg_vif_setup);
 503
 504         if (dev == NULL)
 505                 return NULL;
 506
 507         dev_net_set(dev, net);
 508
 509         if (register_netdevice(dev)) {
 510                 free_netdev(dev);
 511                 return NULL;
 512         }
 513         dev->iflink = 0;
 514
 515         rcu_read_lock();
 516         in_dev = __in_dev_get_rcu(dev);
 517         if (!in_dev) {
 518                 rcu_read_unlock();
 519                 goto failure;
 520         }
 521
 522         ipv4_devconf_setall(in_dev);
 523         IPV4_DEVCONF(in_dev->cnf, RP_FILTER) = 0;
 524         rcu_read_unlock();
 525
 526         if (dev_open(dev))
 527                 goto failure;
 528
 529         dev_hold(dev);
 530
 531         return dev;
 532
 533 failure:
 534         /* allow the register to be completed before unregistering. */
 535         rtnl_unlock();
 536         rtnl_lock();
 537
 538         unregister_netdevice(dev);
 539         return NULL;
 540 }
 541 #endif
 542
 543 /**
 544  *      vif_delete - Delete a VIF entry
 545  *      @notify: Set to 1, if the caller is a notifier_call
 546  */
 547
 548 static int vif_delete(struct mr_table *mrt, int vifi, int notify,
 549                       struct list_head *head)
 550 {
 551         struct vif_device *v;
 552         struct net_device *dev;
 553         struct in_device *in_dev;
 554
 555         if (vifi < 0 || vifi >= mrt->maxvif)
 556                 return -EADDRNOTAVAIL;
 557
 558         v = &mrt->vif_table[vifi];
 559
 560         write_lock_bh(&mrt_lock);
 561         dev = v->dev;
 562         v->dev = NULL;
 563
 564         if (!dev) {
 565                 write_unlock_bh(&mrt_lock);
 566                 return -EADDRNOTAVAIL;
 567         }
 568
 569 #ifdef CONFIG_IP_PIMSM
 570         if (vifi == mrt->mroute_reg_vif_num)
 571                 mrt->mroute_reg_vif_num = -1;
 572 #endif
 573
 574         if (vifi + 1 == mrt->maxvif) {
 575                 int tmp;
 576
 577                 for (tmp = vifi - 1; tmp >= 0; tmp--) {
 578                         if (VIF_EXISTS(mrt, tmp))
 579                                 break;
 580                 }
 581                 mrt->maxvif = tmp+1;
 582         }
 583
 584         write_unlock_bh(&mrt_lock);
 585
 586         dev_set_allmulti(dev, -1);
 587
 588         in_dev = __in_dev_get_rtnl(dev);
 589         if (in_dev) {
 590                 IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)--;
 591                 inet_netconf_notify_devconf(dev_net(dev),
 592                                             NETCONFA_MC_FORWARDING,
 593                                             dev->ifindex, &in_dev->cnf);
 594                 ip_rt_multicast_event(in_dev);
 595         }
 596
 597         if (v->flags & (VIFF_TUNNEL | VIFF_REGISTER) && !notify)
 598                 unregister_netdevice_queue(dev, head);
 599
 600         dev_put(dev);
 601         return 0;
 602 }
 603
 604 static void ipmr_cache_free_rcu(struct rcu_head *head)
 605 {
 606         struct mfc_cache *c = container_of(head, struct mfc_cache, rcu);
 607
 608         kmem_cache_free(mrt_cachep, c);
 609 }
 610
 611 static inline void ipmr_cache_free(struct mfc_cache *c)
 612 {
 613         call_rcu(&c->rcu, ipmr_cache_free_rcu);
 614 }
 615
 616 /* Destroy an unresolved cache entry, killing queued skbs
 617  * and reporting error to netlink readers.
 618  */
 619
 620 static void ipmr_destroy_unres(struct mr_table *mrt, struct mfc_cache *c)
 621 {
 622         struct net *net = read_pnet(&mrt->net);
 623         struct sk_buff *skb;
 624         struct nlmsgerr *e;
 625
 626         atomic_dec(&mrt->cache_resolve_queue_len);
 627
 628         while ((skb = skb_dequeue(&c->mfc_un.unres.unresolved))) {
 629                 if (ip_hdr(skb)->version == 0) {
 630                         struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct iphdr));
 631                         nlh->nlmsg_type = NLMSG_ERROR;
 632                         nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
 633                         skb_trim(skb, nlh->nlmsg_len);
 634                         e = nlmsg_data(nlh);
 635                         e->error = -ETIMEDOUT;
 636                         memset(&e->msg, 0, sizeof(e->msg));
 637
 638                         rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
 639                 } else {
 640                         kfree_skb(skb);
 641                 }
 642         }
 643
 644         ipmr_cache_free(c);
 645 }
 646
 647
 648 /* Timer process for the unresolved queue. */
 649
 650 static void ipmr_expire_process(unsigned long arg)
 651 {
 652         struct mr_table *mrt = (struct mr_table *)arg;
 653         unsigned long now;
 654         unsigned long expires;
 655         struct mfc_cache *c, *next;
 656
 657         if (!spin_trylock(&mfc_unres_lock)) {
 658                 mod_timer(&mrt->ipmr_expire_timer, jiffies+HZ/10);
 659                 return;
 660         }
 661
 662         if (list_empty(&mrt->mfc_unres_queue))
 663                 goto out;
 664
 665         now = jiffies;
 666         expires = 10*HZ;
 667
 668         list_for_each_entry_safe(c, next, &mrt->mfc_unres_queue, list) {
 669                 if (time_after(c->mfc_un.unres.expires, now)) {
 670                         unsigned long interval = c->mfc_un.unres.expires - now;
 671                         if (interval < expires)
 672                                 expires = interval;
 673                         continue;
 674                 }
 675
 676                 list_del(&c->list);
 677                 mroute_netlink_event(mrt, c, RTM_DELROUTE);
 678                 ipmr_destroy_unres(mrt, c);
 679         }
 680
 681         if (!list_empty(&mrt->mfc_unres_queue))
 682                 mod_timer(&mrt->ipmr_expire_timer, jiffies + expires);
 683
 684 out:
 685         spin_unlock(&mfc_unres_lock);
 686 }
 687
 688 /* Fill oifs list. It is called under write locked mrt_lock. */
 689
 690 static void ipmr_update_thresholds(struct mr_table *mrt, struct mfc_cache *cache,
 691                                    unsigned char *ttls)
 692 {
 693         int vifi;
 694
 695         cache->mfc_un.res.minvif = MAXVIFS;
 696         cache->mfc_un.res.maxvif = 0;
 697         memset(cache->mfc_un.res.ttls, 255, MAXVIFS);
 698
 699         for (vifi = 0; vifi < mrt->maxvif; vifi++) {
 700                 if (VIF_EXISTS(mrt, vifi) &&
 701                     ttls[vifi] && ttls[vifi] < 255) {
 702                         cache->mfc_un.res.ttls[vifi] = ttls[vifi];
 703                         if (cache->mfc_un.res.minvif > vifi)
 704                                 cache->mfc_un.res.minvif = vifi;
 705                         if (cache->mfc_un.res.maxvif <= vifi)
 706                                 cache->mfc_un.res.maxvif = vifi + 1;
 707                 }
 708         }
 709 }
 710
 711 static int vif_add(struct net *net, struct mr_table *mrt,
 712                    struct vifctl *vifc, int mrtsock)
 713 {
 714         int vifi = vifc->vifc_vifi;
 715         struct vif_device *v = &mrt->vif_table[vifi];
 716         struct net_device *dev;
 717         struct in_device *in_dev;
 718         int err;
 719
 720         /* Is vif busy ? */
 721         if (VIF_EXISTS(mrt, vifi))
 722                 return -EADDRINUSE;
 723
 724         switch (vifc->vifc_flags) {
 725 #ifdef CONFIG_IP_PIMSM
 726         case VIFF_REGISTER:
 727                 /*
 728                  * Special Purpose VIF in PIM
 729                  * All the packets will be sent to the daemon
 730                  */
 731                 if (mrt->mroute_reg_vif_num >= 0)
 732                         return -EADDRINUSE;
 733                 dev = ipmr_reg_vif(net, mrt);
 734                 if (!dev)
 735                         return -ENOBUFS;
 736                 err = dev_set_allmulti(dev, 1);
 737                 if (err) {
 738                         unregister_netdevice(dev);
 739                         dev_put(dev);
 740                         return err;
 741                 }
 742                 break;
 743 #endif
 744         case VIFF_TUNNEL:
 745                 dev = ipmr_new_tunnel(net, vifc);
 746                 if (!dev)
 747                         return -ENOBUFS;
 748                 err = dev_set_allmulti(dev, 1);
 749                 if (err) {
 750                         ipmr_del_tunnel(dev, vifc);
 751                         dev_put(dev);
 752                         return err;
 753                 }
 754                 break;
 755
 756         case VIFF_USE_IFINDEX:
 757         case 0:
 758                 if (vifc->vifc_flags == VIFF_USE_IFINDEX) {
 759                         dev = dev_get_by_index(net, vifc->vifc_lcl_ifindex);
 760                         if (dev && __in_dev_get_rtnl(dev) == NULL) {
 761                                 dev_put(dev);
 762                                 return -EADDRNOTAVAIL;
 763                         }
 764                 } else {
 765                         dev = ip_dev_find(net, vifc->vifc_lcl_addr.s_addr);
 766                 }
 767                 if (!dev)
 768                         return -EADDRNOTAVAIL;
 769                 err = dev_set_allmulti(dev, 1);
 770                 if (err) {
 771                         dev_put(dev);
 772                         return err;
 773                 }
 774                 break;
 775         default:
 776                 return -EINVAL;
 777         }
 778
 779         in_dev = __in_dev_get_rtnl(dev);
 780         if (!in_dev) {
 781                 dev_put(dev);
 782                 return -EADDRNOTAVAIL;
 783         }
 784         IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)++;
 785         inet_netconf_notify_devconf(net, NETCONFA_MC_FORWARDING, dev->ifindex,
 786                                     &in_dev->cnf);
 787         ip_rt_multicast_event(in_dev);
 788
 789         /* Fill in the VIF structures */
 790
 791         v->rate_limit = vifc->vifc_rate_limit;
 792         v->local = vifc->vifc_lcl_addr.s_addr;
 793         v->remote = vifc->vifc_rmt_addr.s_addr;
 794         v->flags = vifc->vifc_flags;
 795         if (!mrtsock)
 796                 v->flags |= VIFF_STATIC;
 797         v->threshold = vifc->vifc_threshold;
 798         v->bytes_in = 0;
 799         v->bytes_out = 0;
 800         v->pkt_in = 0;
 801         v->pkt_out = 0;
 802         v->link = dev->ifindex;
 803         if (v->flags & (VIFF_TUNNEL | VIFF_REGISTER))
 804                 v->link = dev->iflink;
 805
 806         /* And finish update writing critical data */
 807         write_lock_bh(&mrt_lock);
 808         v->dev = dev;
 809 #ifdef CONFIG_IP_PIMSM
 810         if (v->flags & VIFF_REGISTER)
 811                 mrt->mroute_reg_vif_num = vifi;
 812 #endif
 813         if (vifi+1 > mrt->maxvif)
 814                 mrt->maxvif = vifi+1;
 815         write_unlock_bh(&mrt_lock);
 816         return 0;
 817 }
 818
 819 /* called with rcu_read_lock() */
 820 static struct mfc_cache *ipmr_cache_find(struct mr_table *mrt,
 821                                          __be32 origin,
 822                                          __be32 mcastgrp)
 823 {
 824         int line = MFC_HASH(mcastgrp, origin);
 825         struct mfc_cache *c;
 826
 827         list_for_each_entry_rcu(c, &mrt->mfc_cache_array[line], list) {
 828                 if (c->mfc_origin == origin && c->mfc_mcastgrp == mcastgrp)
 829                         return c;
 830         }
 831         return NULL;
 832 }
 833
 834 /* Look for a (*,*,oif) entry */
 835 static struct mfc_cache *ipmr_cache_find_any_parent(struct mr_table *mrt,
 836                                                     int vifi)
 837 {
 838         int line = MFC_HASH(htonl(INADDR_ANY), htonl(INADDR_ANY));
 839         struct mfc_cache *c;
 840
 841         list_for_each_entry_rcu(c, &mrt->mfc_cache_array[line], list)
 842                 if (c->mfc_origin == htonl(INADDR_ANY) &&
 843                     c->mfc_mcastgrp == htonl(INADDR_ANY) &&
 844                     c->mfc_un.res.ttls[vifi] < 255)
 845                         return c;
 846
 847         return NULL;
 848 }
 849
 850 /* Look for a (*,G) entry */
 851 static struct mfc_cache *ipmr_cache_find_any(struct mr_table *mrt,
 852                                              __be32 mcastgrp, int vifi)
 853 {
 854         int line = MFC_HASH(mcastgrp, htonl(INADDR_ANY));
 855         struct mfc_cache *c, *proxy;
 856
 857         if (mcastgrp == htonl(INADDR_ANY))
 858                 goto skip;
 859
 860         list_for_each_entry_rcu(c, &mrt->mfc_cache_array[line], list)
 861                 if (c->mfc_origin == htonl(INADDR_ANY) &&
 862                     c->mfc_mcastgrp == mcastgrp) {
 863                         if (c->mfc_un.res.ttls[vifi] < 255)
 864                                 return c;
 865
 866                         /* It's ok if the vifi is part of the static tree */
 867                         proxy = ipmr_cache_find_any_parent(mrt,
 868                                                            c->mfc_parent);
 869                         if (proxy && proxy->mfc_un.res.ttls[vifi] < 255)
 870                                 return c;
 871                 }
 872
 873 skip:
 874         return ipmr_cache_find_any_parent(mrt, vifi);
 875 }
 876
 877 /*
 878  *      Allocate a multicast cache entry
 879  */
 880 static struct mfc_cache *ipmr_cache_alloc(void)
 881 {
 882         struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_KERNEL);
 883
 884         if (c)
 885                 c->mfc_un.res.minvif = MAXVIFS;
 886         return c;
 887 }
 888
 889 static struct mfc_cache *ipmr_cache_alloc_unres(void)
 890 {
 891         struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_ATOMIC);
 892
 893         if (c) {
 894                 skb_queue_head_init(&c->mfc_un.unres.unresolved);
 895                 c->mfc_un.unres.expires = jiffies + 10*HZ;
 896         }
 897         return c;
 898 }
 899
 900 /*
 901  *      A cache entry has gone into a resolved state from queued
 902  */
 903
 904 static void ipmr_cache_resolve(struct net *net, struct mr_table *mrt,
 905                                struct mfc_cache *uc, struct mfc_cache *c)
 906 {
 907         struct sk_buff *skb;
 908         struct nlmsgerr *e;
 909
 910         /* Play the pending entries through our router */
 911
 912         while ((skb = __skb_dequeue(&uc->mfc_un.unres.unresolved))) {
 913                 if (ip_hdr(skb)->version == 0) {
 914                         struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct iphdr));
 915
 916                         if (__ipmr_fill_mroute(mrt, skb, c, nlmsg_data(nlh)) > 0) {
 917                                 nlh->nlmsg_len = skb_tail_pointer(skb) -
 918                                                  (u8 *)nlh;
 919                         } else {
 920                                 nlh->nlmsg_type = NLMSG_ERROR;
 921                                 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
 922                                 skb_trim(skb, nlh->nlmsg_len);
 923                                 e = nlmsg_data(nlh);
 924                                 e->error = -EMSGSIZE;
 925                                 memset(&e->msg, 0, sizeof(e->msg));
 926                         }
 927
 928                         rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
 929                 } else {
 930                         ip_mr_forward(net, mrt, skb, c, 0);
 931                 }
 932         }
 933 }
 934
 935 /*
 936  *      Bounce a cache query up to mrouted. We could use netlink for this but mrouted
 937  *      expects the following bizarre scheme.
 938  *
 939  *      Called under mrt_lock.
 940  */
 941
 942 static int ipmr_cache_report(struct mr_table *mrt,
 943                              struct sk_buff *pkt, vifi_t vifi, int assert)
 944 {
 945         struct sk_buff *skb;
 946         const int ihl = ip_hdrlen(pkt);
 947         struct igmphdr *igmp;
 948         struct igmpmsg *msg;
 949         struct sock *mroute_sk;
 950         int ret;
 951
 952 #ifdef CONFIG_IP_PIMSM
 953         if (assert == IGMPMSG_WHOLEPKT)
 954                 skb = skb_realloc_headroom(pkt, sizeof(struct iphdr));
 955         else
 956 #endif
 957                 skb = alloc_skb(128, GFP_ATOMIC);
 958
 959         if (!skb)
 960                 return -ENOBUFS;
 961
 962 #ifdef CONFIG_IP_PIMSM
 963         if (assert == IGMPMSG_WHOLEPKT) {
 964                 /* Ugly, but we have no choice with this interface.
 965                  * Duplicate old header, fix ihl, length etc.
 966                  * And all this only to mangle msg->im_msgtype and
 967                  * to set msg->im_mbz to "mbz" :-)
 968                  */
 969                 skb_push(skb, sizeof(struct iphdr));
 970                 skb_reset_network_header(skb);
 971                 skb_reset_transport_header(skb);
 972                 msg = (struct igmpmsg *)skb_network_header(skb);
 973                 memcpy(msg, skb_network_header(pkt), sizeof(struct iphdr));
 974                 msg->im_msgtype = IGMPMSG_WHOLEPKT;
 975                 msg->im_mbz = 0;
 976                 msg->im_vif = mrt->mroute_reg_vif_num;
 977                 ip_hdr(skb)->ihl = sizeof(struct iphdr) >> 2;
 978                 ip_hdr(skb)->tot_len = htons(ntohs(ip_hdr(pkt)->tot_len) +
 979                                              sizeof(struct iphdr));
 980         } else
 981 #endif
 982         {
 983
 984         /* Copy the IP header */
 985
 986         skb_set_network_header(skb, skb->len);
 987         skb_put(skb, ihl);
 988         skb_copy_to_linear_data(skb, pkt->data, ihl);
 989         ip_hdr(skb)->protocol = 0;      /* Flag to the kernel this is a route add */
 990         msg = (struct igmpmsg *)skb_network_header(skb);
 991         msg->im_vif = vifi;
 992         skb_dst_set(skb, dst_clone(skb_dst(pkt)));
 993
 994         /* Add our header */
 995
 996         igmp = (struct igmphdr *)skb_put(skb, sizeof(struct igmphdr));
 997         igmp->type      =
 998         msg->im_msgtype = assert;
 999         igmp->code      = 0;
1000         ip_hdr(skb)->tot_len = htons(skb->len);         /* Fix the length */
1001         skb->transport_header = skb->network_header;
1002         }
1003
1004         rcu_read_lock();
1005         mroute_sk = rcu_dereference(mrt->mroute_sk);
1006         if (mroute_sk == NULL) {
1007                 rcu_read_unlock();
1008                 kfree_skb(skb);
1009                 return -EINVAL;
1010         }
1011
1012         /* Deliver to mrouted */
1013
1014         ret = sock_queue_rcv_skb(mroute_sk, skb);
1015         rcu_read_unlock();
1016         if (ret < 0) {
1017                 net_warn_ratelimited("mroute: pending queue full, dropping entries\n");
1018                 kfree_skb(skb);
1019         }
1020
1021         return ret;
1022 }
1023
1024 /*
1025  *      Queue a packet for resolution. It gets locked cache entry!
1026  */
1027
1028 static int
1029 ipmr_cache_unresolved(struct mr_table *mrt, vifi_t vifi, struct sk_buff *skb)
1030 {
1031         bool found = false;
1032         int err;
1033         struct mfc_cache *c;
1034         const struct iphdr *iph = ip_hdr(skb);
1035
1036         spin_lock_bh(&mfc_unres_lock);
1037         list_for_each_entry(c, &mrt->mfc_unres_queue, list) {
1038                 if (c->mfc_mcastgrp == iph->daddr &&
1039                     c->mfc_origin == iph->saddr) {
1040                         found = true;
1041                         break;
1042                 }
1043         }
1044
1045         if (!found) {
1046                 /* Create a new entry if allowable */
1047
1048                 if (atomic_read(&mrt->cache_resolve_queue_len) >= 10 ||
1049                     (c = ipmr_cache_alloc_unres()) == NULL) {
1050                         spin_unlock_bh(&mfc_unres_lock);
1051
1052                         kfree_skb(skb);
1053                         return -ENOBUFS;
1054                 }
1055
1056                 /* Fill in the new cache entry */
1057
1058                 c->mfc_parent   = -1;
1059                 c->mfc_origin   = iph->saddr;
1060                 c->mfc_mcastgrp = iph->daddr;
1061
1062                 /* Reflect first query at mrouted. */
1063
1064                 err = ipmr_cache_report(mrt, skb, vifi, IGMPMSG_NOCACHE);
1065                 if (err < 0) {
1066                         /* If the report failed throw the cache entry
1067                            out - Brad Parker
1068                          */
1069                         spin_unlock_bh(&mfc_unres_lock);
1070
1071                         ipmr_cache_free(c);
1072                         kfree_skb(skb);
1073                         return err;
1074                 }
1075
1076                 atomic_inc(&mrt->cache_resolve_queue_len);
1077                 list_add(&c->list, &mrt->mfc_unres_queue);
1078                 mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1079
1080                 if (atomic_read(&mrt->cache_resolve_queue_len) == 1)
1081                         mod_timer(&mrt->ipmr_expire_timer, c->mfc_un.unres.expires);
1082         }
1083
1084         /* See if we can append the packet */
1085
1086         if (c->mfc_un.unres.unresolved.qlen > 3) {
1087                 kfree_skb(skb);
1088                 err = -ENOBUFS;
1089         } else {
1090                 skb_queue_tail(&c->mfc_un.unres.unresolved, skb);
1091                 err = 0;
1092         }
1093
1094         spin_unlock_bh(&mfc_unres_lock);
1095         return err;
1096 }
1097
1098 /*
1099  *      MFC cache manipulation by user space mroute daemon
1100  */
1101
1102 static int ipmr_mfc_delete(struct mr_table *mrt, struct mfcctl *mfc, int parent)
1103 {
1104         int line;
1105         struct mfc_cache *c, *next;
1106
1107         line = MFC_HASH(mfc->mfcc_mcastgrp.s_addr, mfc->mfcc_origin.s_addr);
1108
1109         list_for_each_entry_safe(c, next, &mrt->mfc_cache_array[line], list) {
1110                 if (c->mfc_origin == mfc->mfcc_origin.s_addr &&
1111                     c->mfc_mcastgrp == mfc->mfcc_mcastgrp.s_addr &&
1112                     (parent == -1 || parent == c->mfc_parent)) {
1113                         list_del_rcu(&c->list);
1114                         mroute_netlink_event(mrt, c, RTM_DELROUTE);
1115                         ipmr_cache_free(c);
1116                         return 0;
1117                 }
1118         }
1119         return -ENOENT;
1120 }
1121
1122 static int ipmr_mfc_add(struct net *net, struct mr_table *mrt,
1123                         struct mfcctl *mfc, int mrtsock, int parent)
1124 {
1125         bool found = false;
1126         int line;
1127         struct mfc_cache *uc, *c;
1128
1129         if (mfc->mfcc_parent >= MAXVIFS)
1130                 return -ENFILE;
1131
1132         line = MFC_HASH(mfc->mfcc_mcastgrp.s_addr, mfc->mfcc_origin.s_addr);
1133
1134         list_for_each_entry(c, &mrt->mfc_cache_array[line], list) {
1135                 if (c->mfc_origin == mfc->mfcc_origin.s_addr &&
1136                     c->mfc_mcastgrp == mfc->mfcc_mcastgrp.s_addr &&
1137                     (parent == -1 || parent == c->mfc_parent)) {
1138                         found = true;
1139                         break;
1140                 }
1141         }
1142
1143         if (found) {
1144                 write_lock_bh(&mrt_lock);
1145                 c->mfc_parent = mfc->mfcc_parent;
1146                 ipmr_update_thresholds(mrt, c, mfc->mfcc_ttls);
1147                 if (!mrtsock)
1148                         c->mfc_flags |= MFC_STATIC;
1149                 write_unlock_bh(&mrt_lock);
1150                 mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1151                 return 0;
1152         }
1153
1154         if (mfc->mfcc_mcastgrp.s_addr != htonl(INADDR_ANY) &&
1155             !ipv4_is_multicast(mfc->mfcc_mcastgrp.s_addr))
1156                 return -EINVAL;
1157
1158         c = ipmr_cache_alloc();
1159         if (c == NULL)
1160                 return -ENOMEM;
1161
1162         c->mfc_origin = mfc->mfcc_origin.s_addr;
1163         c->mfc_mcastgrp = mfc->mfcc_mcastgrp.s_addr;
1164         c->mfc_parent = mfc->mfcc_parent;
1165         ipmr_update_thresholds(mrt, c, mfc->mfcc_ttls);
1166         if (!mrtsock)
1167                 c->mfc_flags |= MFC_STATIC;
1168
1169         list_add_rcu(&c->list, &mrt->mfc_cache_array[line]);
1170
1171         /*
1172          *      Check to see if we resolved a queued list. If so we
1173          *      need to send on the frames and tidy up.
1174          */
1175         found = false;
1176         spin_lock_bh(&mfc_unres_lock);
1177         list_for_each_entry(uc, &mrt->mfc_unres_queue, list) {
1178                 if (uc->mfc_origin == c->mfc_origin &&
1179                     uc->mfc_mcastgrp == c->mfc_mcastgrp) {
1180                         list_del(&uc->list);
1181                         atomic_dec(&mrt->cache_resolve_queue_len);
1182                         found = true;
1183                         break;
1184                 }
1185         }
1186         if (list_empty(&mrt->mfc_unres_queue))
1187                 del_timer(&mrt->ipmr_expire_timer);
1188         spin_unlock_bh(&mfc_unres_lock);
1189
1190         if (found) {
1191                 ipmr_cache_resolve(net, mrt, uc, c);
1192                 ipmr_cache_free(uc);
1193         }
1194         mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1195         return 0;
1196 }
1197
1198 /*
1199  *      Close the multicast socket, and clear the vif tables etc
1200  */
1201
1202 static void mroute_clean_tables(struct mr_table *mrt)
1203 {
1204         int i;
1205         LIST_HEAD(list);
1206         struct mfc_cache *c, *next;
1207
1208         /* Shut down all active vif entries */
1209
1210         for (i = 0; i < mrt->maxvif; i++) {
1211                 if (!(mrt->vif_table[i].flags & VIFF_STATIC))
1212                         vif_delete(mrt, i, 0, &list);
1213         }
1214         unregister_netdevice_many(&list);
1215
1216         /* Wipe the cache */
1217
1218         for (i = 0; i < MFC_LINES; i++) {
1219                 list_for_each_entry_safe(c, next, &mrt->mfc_cache_array[i], list) {
1220                         if (c->mfc_flags & MFC_STATIC)
1221                                 continue;
1222                         list_del_rcu(&c->list);
1223                         mroute_netlink_event(mrt, c, RTM_DELROUTE);
1224                         ipmr_cache_free(c);
1225                 }
1226         }
1227
1228         if (atomic_read(&mrt->cache_resolve_queue_len) != 0) {
1229                 spin_lock_bh(&mfc_unres_lock);
1230                 list_for_each_entry_safe(c, next, &mrt->mfc_unres_queue, list) {
1231                         list_del(&c->list);
1232                         mroute_netlink_event(mrt, c, RTM_DELROUTE);
1233                         ipmr_destroy_unres(mrt, c);
1234                 }
1235                 spin_unlock_bh(&mfc_unres_lock);
1236         }
1237 }
1238
1239 /* called from ip_ra_control(), before an RCU grace period,
1240  * we dont need to call synchronize_rcu() here
1241  */
1242 static void mrtsock_destruct(struct sock *sk)
1243 {
1244         struct net *net = sock_net(sk);
1245         struct mr_table *mrt;
1246
1247         rtnl_lock();
1248         ipmr_for_each_table(mrt, net) {
1249                 if (sk == rtnl_dereference(mrt->mroute_sk)) {
1250                         IPV4_DEVCONF_ALL(net, MC_FORWARDING)--;
1251                         inet_netconf_notify_devconf(net, NETCONFA_MC_FORWARDING,
1252                                                     NETCONFA_IFINDEX_ALL,
1253                                                     net->ipv4.devconf_all);
1254                         RCU_INIT_POINTER(mrt->mroute_sk, NULL);
1255                         mroute_clean_tables(mrt);
1256                 }
1257         }
1258         rtnl_unlock();
1259 }
1260
1261 /*
1262  *      Socket options and virtual interface manipulation. The whole
1263  *      virtual interface system is a complete heap, but unfortunately
1264  *      that's how BSD mrouted happens to think. Maybe one day with a proper
1265  *      MOSPF/PIM router set up we can clean this up.
1266  */
1267
1268 int ip_mroute_setsockopt(struct sock *sk, int optname, char __user *optval, unsigned int optlen)
1269 {
1270         int ret, parent = 0;
1271         struct vifctl vif;
1272         struct mfcctl mfc;
1273         struct net *net = sock_net(sk);
1274         struct mr_table *mrt;
1275
1276         if (sk->sk_type != SOCK_RAW ||
1277             inet_sk(sk)->inet_num != IPPROTO_IGMP)
1278                 return -EOPNOTSUPP;
1279
1280         mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1281         if (mrt == NULL)
1282                 return -ENOENT;
1283
1284         if (optname != MRT_INIT) {
1285                 if (sk != rcu_access_pointer(mrt->mroute_sk) &&
1286                     !ns_capable(net->user_ns, CAP_NET_ADMIN))
1287                         return -EACCES;
1288         }
1289
1290         switch (optname) {
1291         case MRT_INIT:
1292                 if (optlen != sizeof(int))
1293                         return -EINVAL;
1294
1295                 rtnl_lock();
1296                 if (rtnl_dereference(mrt->mroute_sk)) {
1297                         rtnl_unlock();
1298                         return -EADDRINUSE;
1299                 }
1300
1301                 ret = ip_ra_control(sk, 1, mrtsock_destruct);
1302                 if (ret == 0) {
1303                         rcu_assign_pointer(mrt->mroute_sk, sk);
1304                         IPV4_DEVCONF_ALL(net, MC_FORWARDING)++;
1305                         inet_netconf_notify_devconf(net, NETCONFA_MC_FORWARDING,
1306                                                     NETCONFA_IFINDEX_ALL,
1307                                                     net->ipv4.devconf_all);
1308                 }
1309                 rtnl_unlock();
1310                 return ret;
1311         case MRT_DONE:
1312                 if (sk != rcu_access_pointer(mrt->mroute_sk))
1313                         return -EACCES;
1314                 return ip_ra_control(sk, 0, NULL);
1315         case MRT_ADD_VIF:
1316         case MRT_DEL_VIF:
1317                 if (optlen != sizeof(vif))
1318                         return -EINVAL;
1319                 if (copy_from_user(&vif, optval, sizeof(vif)))
1320                         return -EFAULT;
1321                 if (vif.vifc_vifi >= MAXVIFS)
1322                         return -ENFILE;
1323                 rtnl_lock();
1324                 if (optname == MRT_ADD_VIF) {
1325                         ret = vif_add(net, mrt, &vif,
1326                                       sk == rtnl_dereference(mrt->mroute_sk));
1327                 } else {
1328                         ret = vif_delete(mrt, vif.vifc_vifi, 0, NULL);
1329                 }
1330                 rtnl_unlock();
1331                 return ret;
1332
1333                 /*
1334                  *      Manipulate the forwarding caches. These live
1335                  *      in a sort of kernel/user symbiosis.
1336                  */
1337         case MRT_ADD_MFC:
1338         case MRT_DEL_MFC:
1339                 parent = -1;
1340         case MRT_ADD_MFC_PROXY:
1341         case MRT_DEL_MFC_PROXY:
1342                 if (optlen != sizeof(mfc))
1343                         return -EINVAL;
1344                 if (copy_from_user(&mfc, optval, sizeof(mfc)))
1345                         return -EFAULT;
1346                 if (parent == 0)
1347                         parent = mfc.mfcc_parent;
1348                 rtnl_lock();
1349                 if (optname == MRT_DEL_MFC || optname == MRT_DEL_MFC_PROXY)
1350                         ret = ipmr_mfc_delete(mrt, &mfc, parent);
1351                 else
1352                         ret = ipmr_mfc_add(net, mrt, &mfc,
1353                                            sk == rtnl_dereference(mrt->mroute_sk),
1354                                            parent);
1355                 rtnl_unlock();
1356                 return ret;
1357                 /*
1358                  *      Control PIM assert.
1359                  */
1360         case MRT_ASSERT:
1361         {
1362                 int v;
1363                 if (optlen != sizeof(v))
1364                         return -EINVAL;
1365                 if (get_user(v, (int __user *)optval))
1366                         return -EFAULT;
1367                 mrt->mroute_do_assert = v;
1368                 return 0;
1369         }
1370 #ifdef CONFIG_IP_PIMSM
1371         case MRT_PIM:
1372         {
1373                 int v;
1374
1375                 if (optlen != sizeof(v))
1376                         return -EINVAL;
1377                 if (get_user(v, (int __user *)optval))
1378                         return -EFAULT;
1379                 v = !!v;
1380
1381                 rtnl_lock();
1382                 ret = 0;
1383                 if (v != mrt->mroute_do_pim) {
1384                         mrt->mroute_do_pim = v;
1385                         mrt->mroute_do_assert = v;
1386                 }
1387                 rtnl_unlock();
1388                 return ret;
1389         }
1390 #endif
1391 #ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES
1392         case MRT_TABLE:
1393         {
1394                 u32 v;
1395
1396                 if (optlen != sizeof(u32))
1397                         return -EINVAL;
1398                 if (get_user(v, (u32 __user *)optval))
1399                         return -EFAULT;
1400
1401                 /* "pimreg%u" should not exceed 16 bytes (IFNAMSIZ) */
1402                 if (v != RT_TABLE_DEFAULT && v >= 1000000000)
1403                         return -EINVAL;
1404
1405                 rtnl_lock();
1406                 ret = 0;
1407                 if (sk == rtnl_dereference(mrt->mroute_sk)) {
1408                         ret = -EBUSY;
1409                 } else {
1410                         if (!ipmr_new_table(net, v))
1411                                 ret = -ENOMEM;
1412                         else
1413                                 raw_sk(sk)->ipmr_table = v;
1414                 }
1415                 rtnl_unlock();
1416                 return ret;
1417         }
1418 #endif
1419         /*
1420          *      Spurious command, or MRT_VERSION which you cannot
1421          *      set.
1422          */
1423         default:
1424                 return -ENOPROTOOPT;
1425         }
1426 }
1427
1428 /*
1429  *      Getsock opt support for the multicast routing system.
1430  */
1431
1432 int ip_mroute_getsockopt(struct sock *sk, int optname, char __user *optval, int __user *optlen)
1433 {
1434         int olr;
1435         int val;
1436         struct net *net = sock_net(sk);
1437         struct mr_table *mrt;
1438
1439         if (sk->sk_type != SOCK_RAW ||
1440             inet_sk(sk)->inet_num != IPPROTO_IGMP)
1441                 return -EOPNOTSUPP;
1442
1443         mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1444         if (mrt == NULL)
1445                 return -ENOENT;
1446
1447         if (optname != MRT_VERSION &&
1448 #ifdef CONFIG_IP_PIMSM
1449            optname != MRT_PIM &&
1450 #endif
1451            optname != MRT_ASSERT)
1452                 return -ENOPROTOOPT;
1453
1454         if (get_user(olr, optlen))
1455                 return -EFAULT;
1456
1457         olr = min_t(unsigned int, olr, sizeof(int));
1458         if (olr < 0)
1459                 return -EINVAL;
1460
1461         if (put_user(olr, optlen))
1462                 return -EFAULT;
1463         if (optname == MRT_VERSION)
1464                 val = 0x0305;
1465 #ifdef CONFIG_IP_PIMSM
1466         else if (optname == MRT_PIM)
1467                 val = mrt->mroute_do_pim;
1468 #endif
1469         else
1470                 val = mrt->mroute_do_assert;
1471         if (copy_to_user(optval, &val, olr))
1472                 return -EFAULT;
1473         return 0;
1474 }
1475
1476 /*
1477  *      The IP multicast ioctl support routines.
1478  */
1479
1480 int ipmr_ioctl(struct sock *sk, int cmd, void __user *arg)
1481 {
1482         struct sioc_sg_req sr;
1483         struct sioc_vif_req vr;
1484         struct vif_device *vif;
1485         struct mfc_cache *c;
1486         struct net *net = sock_net(sk);
1487         struct mr_table *mrt;
1488
1489         mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1490         if (mrt == NULL)
1491                 return -ENOENT;
1492
1493         switch (cmd) {
1494         case SIOCGETVIFCNT:
1495                 if (copy_from_user(&vr, arg, sizeof(vr)))
1496                         return -EFAULT;
1497                 if (vr.vifi >= mrt->maxvif)
1498                         return -EINVAL;
1499                 read_lock(&mrt_lock);
1500                 vif = &mrt->vif_table[vr.vifi];
1501                 if (VIF_EXISTS(mrt, vr.vifi)) {
1502                         vr.icount = vif->pkt_in;
1503                         vr.ocount = vif->pkt_out;
1504                         vr.ibytes = vif->bytes_in;
1505                         vr.obytes = vif->bytes_out;
1506                         read_unlock(&mrt_lock);
1507
1508                         if (copy_to_user(arg, &vr, sizeof(vr)))
1509                                 return -EFAULT;
1510                         return 0;
1511                 }
1512                 read_unlock(&mrt_lock);
1513                 return -EADDRNOTAVAIL;
1514         case SIOCGETSGCNT:
1515                 if (copy_from_user(&sr, arg, sizeof(sr)))
1516                         return -EFAULT;
1517
1518                 rcu_read_lock();
1519                 c = ipmr_cache_find(mrt, sr.src.s_addr, sr.grp.s_addr);
1520                 if (c) {
1521                         sr.pktcnt = c->mfc_un.res.pkt;
1522                         sr.bytecnt = c->mfc_un.res.bytes;
1523                         sr.wrong_if = c->mfc_un.res.wrong_if;
1524                         rcu_read_unlock();
1525
1526                         if (copy_to_user(arg, &sr, sizeof(sr)))
1527                                 return -EFAULT;
1528                         return 0;
1529                 }
1530                 rcu_read_unlock();
1531                 return -EADDRNOTAVAIL;
1532         default:
1533                 return -ENOIOCTLCMD;
1534         }
1535 }
1536
1537 #ifdef CONFIG_COMPAT
1538 struct compat_sioc_sg_req {
1539         struct in_addr src;
1540         struct in_addr grp;
1541         compat_ulong_t pktcnt;
1542         compat_ulong_t bytecnt;
1543         compat_ulong_t wrong_if;
1544 };
1545
1546 struct compat_sioc_vif_req {
1547         vifi_t  vifi;           /* Which iface */
1548         compat_ulong_t icount;
1549         compat_ulong_t ocount;
1550         compat_ulong_t ibytes;
1551         compat_ulong_t obytes;
1552 };
1553
1554 int ipmr_compat_ioctl(struct sock *sk, unsigned int cmd, void __user *arg)
1555 {
1556         struct compat_sioc_sg_req sr;
1557         struct compat_sioc_vif_req vr;
1558         struct vif_device *vif;
1559         struct mfc_cache *c;
1560         struct net *net = sock_net(sk);
1561         struct mr_table *mrt;
1562
1563         mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1564         if (mrt == NULL)
1565                 return -ENOENT;
1566
1567         switch (cmd) {
1568         case SIOCGETVIFCNT:
1569                 if (copy_from_user(&vr, arg, sizeof(vr)))
1570                         return -EFAULT;
1571                 if (vr.vifi >= mrt->maxvif)
1572                         return -EINVAL;
1573                 read_lock(&mrt_lock);
1574                 vif = &mrt->vif_table[vr.vifi];
1575                 if (VIF_EXISTS(mrt, vr.vifi)) {
1576                         vr.icount = vif->pkt_in;
1577                         vr.ocount = vif->pkt_out;
1578                         vr.ibytes = vif->bytes_in;
1579                         vr.obytes = vif->bytes_out;
1580                         read_unlock(&mrt_lock);
1581
1582                         if (copy_to_user(arg, &vr, sizeof(vr)))
1583                                 return -EFAULT;
1584                         return 0;
1585                 }
1586                 read_unlock(&mrt_lock);
1587                 return -EADDRNOTAVAIL;
1588         case SIOCGETSGCNT:
1589                 if (copy_from_user(&sr, arg, sizeof(sr)))
1590                         return -EFAULT;
1591
1592                 rcu_read_lock();
1593                 c = ipmr_cache_find(mrt, sr.src.s_addr, sr.grp.s_addr);
1594                 if (c) {
1595                         sr.pktcnt = c->mfc_un.res.pkt;
1596                         sr.bytecnt = c->mfc_un.res.bytes;
1597                         sr.wrong_if = c->mfc_un.res.wrong_if;
1598                         rcu_read_unlock();
1599
1600                         if (copy_to_user(arg, &sr, sizeof(sr)))
1601                                 return -EFAULT;
1602                         return 0;
1603                 }
1604                 rcu_read_unlock();
1605                 return -EADDRNOTAVAIL;
1606         default:
1607                 return -ENOIOCTLCMD;
1608         }
1609 }
1610 #endif
1611
1612
1613 static int ipmr_device_event(struct notifier_block *this, unsigned long event, void *ptr)
1614 {
1615         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1616         struct net *net = dev_net(dev);
1617         struct mr_table *mrt;
1618         struct vif_device *v;
1619         int ct;
1620
1621         if (event != NETDEV_UNREGISTER)
1622                 return NOTIFY_DONE;
1623
1624         ipmr_for_each_table(mrt, net) {
1625                 v = &mrt->vif_table[0];
1626                 for (ct = 0; ct < mrt->maxvif; ct++, v++) {
1627                         if (v->dev == dev)
1628                                 vif_delete(mrt, ct, 1, NULL);
1629                 }
1630         }
1631         return NOTIFY_DONE;
1632 }
1633
1634
1635 static struct notifier_block ip_mr_notifier = {
1636         .notifier_call = ipmr_device_event,
1637 };
1638
1639 /*
1640  *      Encapsulate a packet by attaching a valid IPIP header to it.
1641  *      This avoids tunnel drivers and other mess and gives us the speed so
1642  *      important for multicast video.
1643  */
1644
1645 static void ip_encap(struct sk_buff *skb, __be32 saddr, __be32 daddr)
1646 {
1647         struct iphdr *iph;
1648         const struct iphdr *old_iph = ip_hdr(skb);
1649
1650         skb_push(skb, sizeof(struct iphdr));
1651         skb->transport_header = skb->network_header;
1652         skb_reset_network_header(skb);
1653         iph = ip_hdr(skb);
1654
1655         iph->version    =       4;
1656         iph->tos        =       old_iph->tos;
1657         iph->ttl        =       old_iph->ttl;
1658         iph->frag_off   =       0;
1659         iph->daddr      =       daddr;
1660         iph->saddr      =       saddr;
1661         iph->protocol   =       IPPROTO_IPIP;
1662         iph->ihl        =       5;
1663         iph->tot_len    =       htons(skb->len);
1664         ip_select_ident(skb, skb_dst(skb), NULL);
1665         ip_send_check(iph);
1666
1667         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1668         nf_reset(skb);
1669 }
1670
1671 static inline int ipmr_forward_finish(struct sk_buff *skb)
1672 {
1673         struct ip_options *opt = &(IPCB(skb)->opt);
1674
1675         IP_INC_STATS_BH(dev_net(skb_dst(skb)->dev), IPSTATS_MIB_OUTFORWDATAGRAMS);
1676         IP_ADD_STATS_BH(dev_net(skb_dst(skb)->dev), IPSTATS_MIB_OUTOCTETS, skb->len);
1677
1678         if (unlikely(opt->optlen))
1679                 ip_forward_options(skb);
1680
1681         return dst_output(skb);
1682 }
1683
1684 /*
1685  *      Processing handlers for ipmr_forward
1686  */
1687
1688 static void ipmr_queue_xmit(struct net *net, struct mr_table *mrt,
1689                             struct sk_buff *skb, struct mfc_cache *c, int vifi)
1690 {
1691         const struct iphdr *iph = ip_hdr(skb);
1692         struct vif_device *vif = &mrt->vif_table[vifi];
1693         struct net_device *dev;
1694         struct rtable *rt;
1695         struct flowi4 fl4;
1696         int    encap = 0;
1697
1698         if (vif->dev == NULL)
1699                 goto out_free;
1700
1701 #ifdef CONFIG_IP_PIMSM
1702         if (vif->flags & VIFF_REGISTER) {
1703                 vif->pkt_out++;
1704                 vif->bytes_out += skb->len;
1705                 vif->dev->stats.tx_bytes += skb->len;
1706                 vif->dev->stats.tx_packets++;
1707                 ipmr_cache_report(mrt, skb, vifi, IGMPMSG_WHOLEPKT);
1708                 goto out_free;
1709         }
1710 #endif
1711
1712         if (vif->flags & VIFF_TUNNEL) {
1713                 rt = ip_route_output_ports(net, &fl4, NULL,
1714                                            vif->remote, vif->local,
1715                                            0, 0,
1716                                            IPPROTO_IPIP,
1717                                            RT_TOS(iph->tos), vif->link);
1718                 if (IS_ERR(rt))
1719                         goto out_free;
1720                 encap = sizeof(struct iphdr);
1721         } else {
1722                 rt = ip_route_output_ports(net, &fl4, NULL, iph->daddr, 0,
1723                                            0, 0,
1724                                            IPPROTO_IPIP,
1725                                            RT_TOS(iph->tos), vif->link);
1726                 if (IS_ERR(rt))
1727                         goto out_free;
1728         }
1729
1730         dev = rt->dst.dev;
1731
1732         if (skb->len+encap > dst_mtu(&rt->dst) && (ntohs(iph->frag_off) & IP_DF)) {
1733                 /* Do not fragment multicasts. Alas, IPv4 does not
1734                  * allow to send ICMP, so that packets will disappear
1735                  * to blackhole.
1736                  */
1737
1738                 IP_INC_STATS_BH(dev_net(dev), IPSTATS_MIB_FRAGFAILS);
1739                 ip_rt_put(rt);
1740                 goto out_free;
1741         }
1742
1743         encap += LL_RESERVED_SPACE(dev) + rt->dst.header_len;
1744
1745         if (skb_cow(skb, encap)) {
1746                 ip_rt_put(rt);
1747                 goto out_free;
1748         }
1749
1750         vif->pkt_out++;
1751         vif->bytes_out += skb->len;
1752
1753         skb_dst_drop(skb);
1754         skb_dst_set(skb, &rt->dst);
1755         ip_decrease_ttl(ip_hdr(skb));
1756
1757         /* FIXME: forward and output firewalls used to be called here.
1758          * What do we do with netfilter? -- RR
1759          */
1760         if (vif->flags & VIFF_TUNNEL) {
1761                 ip_encap(skb, vif->local, vif->remote);
1762                 /* FIXME: extra output firewall step used to be here. --RR */
1763                 vif->dev->stats.tx_packets++;
1764                 vif->dev->stats.tx_bytes += skb->len;
1765         }
1766
1767         IPCB(skb)->flags |= IPSKB_FORWARDED;
1768
1769         /*
1770          * RFC1584 teaches, that DVMRP/PIM router must deliver packets locally
1771          * not only before forwarding, but after forwarding on all output
1772          * interfaces. It is clear, if mrouter runs a multicasting
1773          * program, it should receive packets not depending to what interface
1774          * program is joined.
1775          * If we will not make it, the program will have to join on all
1776          * interfaces. On the other hand, multihoming host (or router, but
1777          * not mrouter) cannot join to more than one interface - it will
1778          * result in receiving multiple packets.
1779          */
1780         NF_HOOK(NFPROTO_IPV4, NF_INET_FORWARD, skb, skb->dev, dev,
1781                 ipmr_forward_finish);
1782         return;
1783
1784 out_free:
1785         kfree_skb(skb);
1786 }
1787
1788 static int ipmr_find_vif(struct mr_table *mrt, struct net_device *dev)
1789 {
1790         int ct;
1791
1792         for (ct = mrt->maxvif-1; ct >= 0; ct--) {
1793                 if (mrt->vif_table[ct].dev == dev)
1794                         break;
1795         }
1796         return ct;
1797 }
1798
1799 /* "local" means that we should preserve one skb (for local delivery) */
1800
1801 static void ip_mr_forward(struct net *net, struct mr_table *mrt,
1802                           struct sk_buff *skb, struct mfc_cache *cache,
1803                           int local)
1804 {
1805         int psend = -1;
1806         int vif, ct;
1807         int true_vifi = ipmr_find_vif(mrt, skb->dev);
1808
1809         vif = cache->mfc_parent;
1810         cache->mfc_un.res.pkt++;
1811         cache->mfc_un.res.bytes += skb->len;
1812
1813         if (cache->mfc_origin == htonl(INADDR_ANY) && true_vifi >= 0) {
1814                 struct mfc_cache *cache_proxy;
1815
1816                 /* For an (*,G) entry, we only check that the incomming
1817                  * interface is part of the static tree.
1818                  */
1819                 cache_proxy = ipmr_cache_find_any_parent(mrt, vif);
1820                 if (cache_proxy &&
1821                     cache_proxy->mfc_un.res.ttls[true_vifi] < 255)
1822                         goto forward;
1823         }
1824
1825         /*
1826          * Wrong interface: drop packet and (maybe) send PIM assert.
1827          */
1828         if (mrt->vif_table[vif].dev != skb->dev) {
1829                 if (rt_is_output_route(skb_rtable(skb))) {
1830                         /* It is our own packet, looped back.
1831                          * Very complicated situation...
1832                          *
1833                          * The best workaround until routing daemons will be
1834                          * fixed is not to redistribute packet, if it was
1835                          * send through wrong interface. It means, that
1836                          * multicast applications WILL NOT work for
1837                          * (S,G), which have default multicast route pointing
1838                          * to wrong oif. In any case, it is not a good
1839                          * idea to use multicasting applications on router.
1840                          */
1841                         goto dont_forward;
1842                 }
1843
1844                 cache->mfc_un.res.wrong_if++;
1845
1846                 if (true_vifi >= 0 && mrt->mroute_do_assert &&
1847                     /* pimsm uses asserts, when switching from RPT to SPT,
1848                      * so that we cannot check that packet arrived on an oif.
1849                      * It is bad, but otherwise we would need to move pretty
1850                      * large chunk of pimd to kernel. Ough... --ANK
1851                      */
1852                     (mrt->mroute_do_pim ||
1853                      cache->mfc_un.res.ttls[true_vifi] < 255) &&
1854                     time_after(jiffies,
1855                                cache->mfc_un.res.last_assert + MFC_ASSERT_THRESH)) {
1856                         cache->mfc_un.res.last_assert = jiffies;
1857                         ipmr_cache_report(mrt, skb, true_vifi, IGMPMSG_WRONGVIF);
1858                 }
1859                 goto dont_forward;
1860         }
1861
1862 forward:
1863         mrt->vif_table[vif].pkt_in++;
1864         mrt->vif_table[vif].bytes_in += skb->len;
1865
1866         /*
1867          *      Forward the frame
1868          */
1869         if (cache->mfc_origin == htonl(INADDR_ANY) &&
1870             cache->mfc_mcastgrp == htonl(INADDR_ANY)) {
1871                 if (true_vifi >= 0 &&
1872                     true_vifi != cache->mfc_parent &&
1873                     ip_hdr(skb)->ttl >
1874                                 cache->mfc_un.res.ttls[cache->mfc_parent]) {
1875                         /* It's an (*,*) entry and the packet is not coming from
1876                          * the upstream: forward the packet to the upstream
1877                          * only.
1878                          */
1879                         psend = cache->mfc_parent;
1880                         goto last_forward;
1881                 }
1882                 goto dont_forward;
1883         }
1884         for (ct = cache->mfc_un.res.maxvif - 1;
1885              ct >= cache->mfc_un.res.minvif; ct--) {
1886                 /* For (*,G) entry, don't forward to the incoming interface */
1887                 if ((cache->mfc_origin != htonl(INADDR_ANY) ||
1888                      ct != true_vifi) &&
1889                     ip_hdr(skb)->ttl > cache->mfc_un.res.ttls[ct]) {
1890                         if (psend != -1) {
1891                                 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
1892
1893                                 if (skb2)
1894                                         ipmr_queue_xmit(net, mrt, skb2, cache,
1895                                                         psend);
1896                         }
1897                         psend = ct;
1898                 }
1899         }
1900 last_forward:
1901         if (psend != -1) {
1902                 if (local) {
1903                         struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
1904
1905                         if (skb2)
1906                                 ipmr_queue_xmit(net, mrt, skb2, cache, psend);
1907                 } else {
1908                         ipmr_queue_xmit(net, mrt, skb, cache, psend);
1909                         return;
1910                 }
1911         }
1912
1913 dont_forward:
1914         if (!local)
1915                 kfree_skb(skb);
1916 }
1917
1918 static struct mr_table *ipmr_rt_fib_lookup(struct net *net, struct sk_buff *skb)
1919 {
1920         struct rtable *rt = skb_rtable(skb);
1921         struct iphdr *iph = ip_hdr(skb);
1922         struct flowi4 fl4 = {
1923                 .daddr = iph->daddr,
1924                 .saddr = iph->saddr,
1925                 .flowi4_tos = RT_TOS(iph->tos),
1926                 .flowi4_oif = (rt_is_output_route(rt) ?
1927                                skb->dev->ifindex : 0),
1928                 .flowi4_iif = (rt_is_output_route(rt) ?
1929                                LOOPBACK_IFINDEX :
1930                                skb->dev->ifindex),
1931                 .flowi4_mark = skb->mark,
1932         };
1933         struct mr_table *mrt;
1934         int err;
1935
1936         err = ipmr_fib_lookup(net, &fl4, &mrt);
1937         if (err)
1938                 return ERR_PTR(err);
1939         return mrt;
1940 }
1941
1942 /*
1943  *      Multicast packets for forwarding arrive here
1944  *      Called with rcu_read_lock();
1945  */
1946
1947 int ip_mr_input(struct sk_buff *skb)
1948 {
1949         struct mfc_cache *cache;
1950         struct net *net = dev_net(skb->dev);
1951         int local = skb_rtable(skb)->rt_flags & RTCF_LOCAL;
1952         struct mr_table *mrt;
1953
1954         /* Packet is looped back after forward, it should not be
1955          * forwarded second time, but still can be delivered locally.
1956          */
1957         if (IPCB(skb)->flags & IPSKB_FORWARDED)
1958                 goto dont_forward;
1959
1960         mrt = ipmr_rt_fib_lookup(net, skb);
1961         if (IS_ERR(mrt)) {
1962                 kfree_skb(skb);
1963                 return PTR_ERR(mrt);
1964         }
1965         if (!local) {
1966                 if (IPCB(skb)->opt.router_alert) {
1967                         if (ip_call_ra_chain(skb))
1968                                 return 0;
1969                 } else if (ip_hdr(skb)->protocol == IPPROTO_IGMP) {
1970                         /* IGMPv1 (and broken IGMPv2 implementations sort of
1971                          * Cisco IOS <= 11.2(8)) do not put router alert
1972                          * option to IGMP packets destined to routable
1973                          * groups. It is very bad, because it means
1974                          * that we can forward NO IGMP messages.
1975                          */
1976                         struct sock *mroute_sk;
1977
1978                         mroute_sk = rcu_dereference(mrt->mroute_sk);
1979                         if (mroute_sk) {
1980                                 nf_reset(skb);
1981                                 raw_rcv(mroute_sk, skb);
1982                                 return 0;
1983                         }
1984                     }
1985         }
1986
1987         /* already under rcu_read_lock() */
1988         cache = ipmr_cache_find(mrt, ip_hdr(skb)->saddr, ip_hdr(skb)->daddr);
1989         if (cache == NULL) {
1990                 int vif = ipmr_find_vif(mrt, skb->dev);
1991
1992                 if (vif >= 0)
1993                         cache = ipmr_cache_find_any(mrt, ip_hdr(skb)->daddr,
1994                                                     vif);
1995         }
1996
1997         /*
1998          *      No usable cache entry
1999          */
2000         if (cache == NULL) {
2001                 int vif;
2002
2003                 if (local) {
2004                         struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2005                         ip_local_deliver(skb);
2006                         if (skb2 == NULL)
2007                                 return -ENOBUFS;
2008                         skb = skb2;
2009                 }
2010
2011                 read_lock(&mrt_lock);
2012                 vif = ipmr_find_vif(mrt, skb->dev);
2013                 if (vif >= 0) {
2014                         int err2 = ipmr_cache_unresolved(mrt, vif, skb);
2015                         read_unlock(&mrt_lock);
2016
2017                         return err2;
2018                 }
2019                 read_unlock(&mrt_lock);
2020                 kfree_skb(skb);
2021                 return -ENODEV;
2022         }
2023
2024         read_lock(&mrt_lock);
2025         ip_mr_forward(net, mrt, skb, cache, local);
2026         read_unlock(&mrt_lock);
2027
2028         if (local)
2029                 return ip_local_deliver(skb);
2030
2031         return 0;
2032
2033 dont_forward:
2034         if (local)
2035                 return ip_local_deliver(skb);
2036         kfree_skb(skb);
2037         return 0;
2038 }
2039
2040 #ifdef CONFIG_IP_PIMSM
2041 /* called with rcu_read_lock() */
2042 static int __pim_rcv(struct mr_table *mrt, struct sk_buff *skb,
2043                      unsigned int pimlen)
2044 {
2045         struct net_device *reg_dev = NULL;
2046         struct iphdr *encap;
2047
2048         encap = (struct iphdr *)(skb_transport_header(skb) + pimlen);
2049         /*
2050          * Check that:
2051          * a. packet is really sent to a multicast group
2052          * b. packet is not a NULL-REGISTER
2053          * c. packet is not truncated
2054          */
2055         if (!ipv4_is_multicast(encap->daddr) ||
2056             encap->tot_len == 0 ||
2057             ntohs(encap->tot_len) + pimlen > skb->len)
2058                 return 1;
2059
2060         read_lock(&mrt_lock);
2061         if (mrt->mroute_reg_vif_num >= 0)
2062                 reg_dev = mrt->vif_table[mrt->mroute_reg_vif_num].dev;
2063         read_unlock(&mrt_lock);
2064
2065         if (reg_dev == NULL)
2066                 return 1;
2067
2068         skb->mac_header = skb->network_header;
2069         skb_pull(skb, (u8 *)encap - skb->data);
2070         skb_reset_network_header(skb);
2071         skb->protocol = htons(ETH_P_IP);
2072         skb->ip_summed = CHECKSUM_NONE;
2073
2074         skb_tunnel_rx(skb, reg_dev, dev_net(reg_dev));
2075
2076         netif_rx(skb);
2077
2078         return NET_RX_SUCCESS;
2079 }
2080 #endif
2081
2082 #ifdef CONFIG_IP_PIMSM_V1
2083 /*
2084  * Handle IGMP messages of PIMv1
2085  */
2086
2087 int pim_rcv_v1(struct sk_buff *skb)
2088 {
2089         struct igmphdr *pim;
2090         struct net *net = dev_net(skb->dev);
2091         struct mr_table *mrt;
2092
2093         if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr)))
2094                 goto drop;
2095
2096         pim = igmp_hdr(skb);
2097
2098         mrt = ipmr_rt_fib_lookup(net, skb);
2099         if (IS_ERR(mrt))
2100                 goto drop;
2101         if (!mrt->mroute_do_pim ||
2102             pim->group != PIM_V1_VERSION || pim->code != PIM_V1_REGISTER)
2103                 goto drop;
2104
2105         if (__pim_rcv(mrt, skb, sizeof(*pim))) {
2106 drop:
2107                 kfree_skb(skb);
2108         }
2109         return 0;
2110 }
2111 #endif
2112
2113 #ifdef CONFIG_IP_PIMSM_V2
2114 static int pim_rcv(struct sk_buff *skb)
2115 {
2116         struct pimreghdr *pim;
2117         struct net *net = dev_net(skb->dev);
2118         struct mr_table *mrt;
2119
2120         if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr)))
2121                 goto drop;
2122
2123         pim = (struct pimreghdr *)skb_transport_header(skb);
2124         if (pim->type != ((PIM_VERSION << 4) | (PIM_REGISTER)) ||
2125             (pim->flags & PIM_NULL_REGISTER) ||
2126             (ip_compute_csum((void *)pim, sizeof(*pim)) != 0 &&
2127              csum_fold(skb_checksum(skb, 0, skb->len, 0))))
2128                 goto drop;
2129
2130         mrt = ipmr_rt_fib_lookup(net, skb);
2131         if (IS_ERR(mrt))
2132                 goto drop;
2133         if (__pim_rcv(mrt, skb, sizeof(*pim))) {
2134 drop:
2135                 kfree_skb(skb);
2136         }
2137         return 0;
2138 }
2139 #endif
2140
2141 static int __ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
2142                               struct mfc_cache *c, struct rtmsg *rtm)
2143 {
2144         int ct;
2145         struct rtnexthop *nhp;
2146         struct nlattr *mp_attr;
2147         struct rta_mfc_stats mfcs;
2148
2149         /* If cache is unresolved, don't try to parse IIF and OIF */
2150         if (c->mfc_parent >= MAXVIFS)
2151                 return -ENOENT;
2152
2153         if (VIF_EXISTS(mrt, c->mfc_parent) &&
2154             nla_put_u32(skb, RTA_IIF, mrt->vif_table[c->mfc_parent].dev->ifindex) < 0)
2155                 return -EMSGSIZE;
2156
2157         if (!(mp_attr = nla_nest_start(skb, RTA_MULTIPATH)))
2158                 return -EMSGSIZE;
2159
2160         for (ct = c->mfc_un.res.minvif; ct < c->mfc_un.res.maxvif; ct++) {
2161                 if (VIF_EXISTS(mrt, ct) && c->mfc_un.res.ttls[ct] < 255) {
2162                         if (!(nhp = nla_reserve_nohdr(skb, sizeof(*nhp)))) {
2163                                 nla_nest_cancel(skb, mp_attr);
2164                                 return -EMSGSIZE;
2165                         }
2166
2167                         nhp->rtnh_flags = 0;
2168                         nhp->rtnh_hops = c->mfc_un.res.ttls[ct];
2169                         nhp->rtnh_ifindex = mrt->vif_table[ct].dev->ifindex;
2170                         nhp->rtnh_len = sizeof(*nhp);
2171                 }
2172         }
2173
2174         nla_nest_end(skb, mp_attr);
2175
2176         mfcs.mfcs_packets = c->mfc_un.res.pkt;
2177         mfcs.mfcs_bytes = c->mfc_un.res.bytes;
2178         mfcs.mfcs_wrong_if = c->mfc_un.res.wrong_if;
2179         if (nla_put(skb, RTA_MFC_STATS, sizeof(mfcs), &mfcs) < 0)
2180                 return -EMSGSIZE;
2181
2182         rtm->rtm_type = RTN_MULTICAST;
2183         return 1;
2184 }
2185
2186 int ipmr_get_route(struct net *net, struct sk_buff *skb,
2187                    __be32 saddr, __be32 daddr,
2188                    struct rtmsg *rtm, int nowait)
2189 {
2190         struct mfc_cache *cache;
2191         struct mr_table *mrt;
2192         int err;
2193
2194         mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2195         if (mrt == NULL)
2196                 return -ENOENT;
2197
2198         rcu_read_lock();
2199         cache = ipmr_cache_find(mrt, saddr, daddr);
2200         if (cache == NULL && skb->dev) {
2201                 int vif = ipmr_find_vif(mrt, skb->dev);
2202
2203                 if (vif >= 0)
2204                         cache = ipmr_cache_find_any(mrt, daddr, vif);
2205         }
2206         if (cache == NULL) {
2207                 struct sk_buff *skb2;
2208                 struct iphdr *iph;
2209                 struct net_device *dev;
2210                 int vif = -1;
2211
2212                 if (nowait) {
2213                         rcu_read_unlock();
2214                         return -EAGAIN;
2215                 }
2216
2217                 dev = skb->dev;
2218                 read_lock(&mrt_lock);
2219                 if (dev)
2220                         vif = ipmr_find_vif(mrt, dev);
2221                 if (vif < 0) {
2222                         read_unlock(&mrt_lock);
2223                         rcu_read_unlock();
2224                         return -ENODEV;
2225                 }
2226                 skb2 = skb_clone(skb, GFP_ATOMIC);
2227                 if (!skb2) {
2228                         read_unlock(&mrt_lock);
2229                         rcu_read_unlock();
2230                         return -ENOMEM;
2231                 }
2232
2233                 skb_push(skb2, sizeof(struct iphdr));
2234                 skb_reset_network_header(skb2);
2235                 iph = ip_hdr(skb2);
2236                 iph->ihl = sizeof(struct iphdr) >> 2;
2237                 iph->saddr = saddr;
2238                 iph->daddr = daddr;
2239                 iph->version = 0;
2240                 err = ipmr_cache_unresolved(mrt, vif, skb2);
2241                 read_unlock(&mrt_lock);
2242                 rcu_read_unlock();
2243                 return err;
2244         }
2245
2246         read_lock(&mrt_lock);
2247         if (!nowait && (rtm->rtm_flags & RTM_F_NOTIFY))
2248                 cache->mfc_flags |= MFC_NOTIFY;
2249         err = __ipmr_fill_mroute(mrt, skb, cache, rtm);
2250         read_unlock(&mrt_lock);
2251         rcu_read_unlock();
2252         return err;
2253 }
2254
2255 static int ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
2256                             u32 portid, u32 seq, struct mfc_cache *c, int cmd)
2257 {
2258         struct nlmsghdr *nlh;
2259         struct rtmsg *rtm;
2260         int err;
2261
2262         nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rtm), NLM_F_MULTI);
2263         if (nlh == NULL)
2264                 return -EMSGSIZE;
2265
2266         rtm = nlmsg_data(nlh);
2267         rtm->rtm_family   = RTNL_FAMILY_IPMR;
2268         rtm->rtm_dst_len  = 32;
2269         rtm->rtm_src_len  = 32;
2270         rtm->rtm_tos      = 0;
2271         rtm->rtm_table    = mrt->id;
2272         if (nla_put_u32(skb, RTA_TABLE, mrt->id))
2273                 goto nla_put_failure;
2274         rtm->rtm_type     = RTN_MULTICAST;
2275         rtm->rtm_scope    = RT_SCOPE_UNIVERSE;
2276         if (c->mfc_flags & MFC_STATIC)
2277                 rtm->rtm_protocol = RTPROT_STATIC;
2278         else
2279                 rtm->rtm_protocol = RTPROT_MROUTED;
2280         rtm->rtm_flags    = 0;
2281
2282         if (nla_put_be32(skb, RTA_SRC, c->mfc_origin) ||
2283             nla_put_be32(skb, RTA_DST, c->mfc_mcastgrp))
2284                 goto nla_put_failure;
2285         err = __ipmr_fill_mroute(mrt, skb, c, rtm);
2286         /* do not break the dump if cache is unresolved */
2287         if (err < 0 && err != -ENOENT)
2288                 goto nla_put_failure;
2289
2290         return nlmsg_end(skb, nlh);
2291
2292 nla_put_failure:
2293         nlmsg_cancel(skb, nlh);
2294         return -EMSGSIZE;
2295 }
2296
2297 static size_t mroute_msgsize(bool unresolved, int maxvif)
2298 {
2299         size_t len =
2300                 NLMSG_ALIGN(sizeof(struct rtmsg))
2301                 + nla_total_size(4)     /* RTA_TABLE */
2302                 + nla_total_size(4)     /* RTA_SRC */
2303                 + nla_total_size(4)     /* RTA_DST */
2304                 ;
2305
2306         if (!unresolved)
2307                 len = len
2308                       + nla_total_size(4)       /* RTA_IIF */
2309                       + nla_total_size(0)       /* RTA_MULTIPATH */
2310                       + maxvif * NLA_ALIGN(sizeof(struct rtnexthop))
2311                                                 /* RTA_MFC_STATS */
2312                       + nla_total_size(sizeof(struct rta_mfc_stats))
2313                 ;
2314
2315         return len;
2316 }
2317
2318 static void mroute_netlink_event(struct mr_table *mrt, struct mfc_cache *mfc,
2319                                  int cmd)
2320 {
2321         struct net *net = read_pnet(&mrt->net);
2322         struct sk_buff *skb;
2323         int err = -ENOBUFS;
2324
2325         skb = nlmsg_new(mroute_msgsize(mfc->mfc_parent >= MAXVIFS, mrt->maxvif),
2326                         GFP_ATOMIC);
2327         if (skb == NULL)
2328                 goto errout;
2329
2330         err = ipmr_fill_mroute(mrt, skb, 0, 0, mfc, cmd);
2331         if (err < 0)
2332                 goto errout;
2333
2334         rtnl_notify(skb, net, 0, RTNLGRP_IPV4_MROUTE, NULL, GFP_ATOMIC);
2335         return;
2336
2337 errout:
2338         kfree_skb(skb);
2339         if (err < 0)
2340                 rtnl_set_sk_err(net, RTNLGRP_IPV4_MROUTE, err);
2341 }
2342
2343 static int ipmr_rtm_dumproute(struct sk_buff *skb, struct netlink_callback *cb)
2344 {
2345         struct net *net = sock_net(skb->sk);
2346         struct mr_table *mrt;
2347         struct mfc_cache *mfc;
2348         unsigned int t = 0, s_t;
2349         unsigned int h = 0, s_h;
2350         unsigned int e = 0, s_e;
2351
2352         s_t = cb->args[0];
2353         s_h = cb->args[1];
2354         s_e = cb->args[2];
2355
2356         rcu_read_lock();
2357         ipmr_for_each_table(mrt, net) {
2358                 if (t < s_t)
2359                         goto next_table;
2360                 if (t > s_t)
2361                         s_h = 0;
2362                 for (h = s_h; h < MFC_LINES; h++) {
2363                         list_for_each_entry_rcu(mfc, &mrt->mfc_cache_array[h], list) {
2364                                 if (e < s_e)
2365                                         goto next_entry;
2366                                 if (ipmr_fill_mroute(mrt, skb,
2367                                                      NETLINK_CB(cb->skb).portid,
2368                                                      cb->nlh->nlmsg_seq,
2369                                                      mfc, RTM_NEWROUTE) < 0)
2370                                         goto done;
2371 next_entry:
2372                                 e++;
2373                         }
2374                         e = s_e = 0;
2375                 }
2376                 spin_lock_bh(&mfc_unres_lock);
2377                 list_for_each_entry(mfc, &mrt->mfc_unres_queue, list) {
2378                         if (e < s_e)
2379                                 goto next_entry2;
2380                         if (ipmr_fill_mroute(mrt, skb,
2381                                              NETLINK_CB(cb->skb).portid,
2382                                              cb->nlh->nlmsg_seq,
2383                                              mfc, RTM_NEWROUTE) < 0) {
2384                                 spin_unlock_bh(&mfc_unres_lock);
2385                                 goto done;
2386                         }
2387 next_entry2:
2388                         e++;
2389                 }
2390                 spin_unlock_bh(&mfc_unres_lock);
2391                 e = s_e = 0;
2392                 s_h = 0;
2393 next_table:
2394                 t++;
2395         }
2396 done:
2397         rcu_read_unlock();
2398
2399         cb->args[2] = e;
2400         cb->args[1] = h;
2401         cb->args[0] = t;
2402
2403         return skb->len;
2404 }
2405
2406 #ifdef CONFIG_PROC_FS
2407 /*
2408  *      The /proc interfaces to multicast routing :
2409  *      /proc/net/ip_mr_cache & /proc/net/ip_mr_vif
2410  */
2411 struct ipmr_vif_iter {
2412         struct seq_net_private p;
2413         struct mr_table *mrt;
2414         int ct;
2415 };
2416
2417 static struct vif_device *ipmr_vif_seq_idx(struct net *net,
2418                                            struct ipmr_vif_iter *iter,
2419                                            loff_t pos)
2420 {
2421         struct mr_table *mrt = iter->mrt;
2422
2423         for (iter->ct = 0; iter->ct < mrt->maxvif; ++iter->ct) {
2424                 if (!VIF_EXISTS(mrt, iter->ct))
2425                         continue;
2426                 if (pos-- == 0)
2427                         return &mrt->vif_table[iter->ct];
2428         }
2429         return NULL;
2430 }
2431
2432 static void *ipmr_vif_seq_start(struct seq_file *seq, loff_t *pos)
2433         __acquires(mrt_lock)
2434 {
2435         struct ipmr_vif_iter *iter = seq->private;
2436         struct net *net = seq_file_net(seq);
2437         struct mr_table *mrt;
2438
2439         mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2440         if (mrt == NULL)
2441                 return ERR_PTR(-ENOENT);
2442
2443         iter->mrt = mrt;
2444
2445         read_lock(&mrt_lock);
2446         return *pos ? ipmr_vif_seq_idx(net, seq->private, *pos - 1)
2447                 : SEQ_START_TOKEN;
2448 }
2449
2450 static void *ipmr_vif_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2451 {
2452         struct ipmr_vif_iter *iter = seq->private;
2453         struct net *net = seq_file_net(seq);
2454         struct mr_table *mrt = iter->mrt;
2455
2456         ++*pos;
2457         if (v == SEQ_START_TOKEN)
2458                 return ipmr_vif_seq_idx(net, iter, 0);
2459
2460         while (++iter->ct < mrt->maxvif) {
2461                 if (!VIF_EXISTS(mrt, iter->ct))
2462                         continue;
2463                 return &mrt->vif_table[iter->ct];
2464         }
2465         return NULL;
2466 }
2467
2468 static void ipmr_vif_seq_stop(struct seq_file *seq, void *v)
2469         __releases(mrt_lock)
2470 {
2471         read_unlock(&mrt_lock);
2472 }
2473
2474 static int ipmr_vif_seq_show(struct seq_file *seq, void *v)
2475 {
2476         struct ipmr_vif_iter *iter = seq->private;
2477         struct mr_table *mrt = iter->mrt;
2478
2479         if (v == SEQ_START_TOKEN) {
2480                 seq_puts(seq,
2481                          "Interface      BytesIn  PktsIn  BytesOut PktsOut Flags Local    Remote\n");
2482         } else {
2483                 const struct vif_device *vif = v;
2484                 const char *name =  vif->dev ? vif->dev->name : "none";
2485
2486                 seq_printf(seq,
2487                            "%2Zd %-10s %8ld %7ld  %8ld %7ld %05X %08X %08X\n",
2488                            vif - mrt->vif_table,
2489                            name, vif->bytes_in, vif->pkt_in,
2490                            vif->bytes_out, vif->pkt_out,
2491                            vif->flags, vif->local, vif->remote);
2492         }
2493         return 0;
2494 }
2495
2496 static const struct seq_operations ipmr_vif_seq_ops = {
2497         .start = ipmr_vif_seq_start,
2498         .next  = ipmr_vif_seq_next,
2499         .stop  = ipmr_vif_seq_stop,
2500         .show  = ipmr_vif_seq_show,
2501 };
2502
2503 static int ipmr_vif_open(struct inode *inode, struct file *file)
2504 {
2505         return seq_open_net(inode, file, &ipmr_vif_seq_ops,
2506                             sizeof(struct ipmr_vif_iter));
2507 }
2508
2509 static const struct file_operations ipmr_vif_fops = {
2510         .owner   = THIS_MODULE,
2511         .open    = ipmr_vif_open,
2512         .read    = seq_read,
2513         .llseek  = seq_lseek,
2514         .release = seq_release_net,
2515 };
2516
2517 struct ipmr_mfc_iter {
2518         struct seq_net_private p;
2519         struct mr_table *mrt;
2520         struct list_head *cache;
2521         int ct;
2522 };
2523
2524
2525 static struct mfc_cache *ipmr_mfc_seq_idx(struct net *net,
2526                                           struct ipmr_mfc_iter *it, loff_t pos)
2527 {
2528         struct mr_table *mrt = it->mrt;
2529         struct mfc_cache *mfc;
2530
2531         rcu_read_lock();
2532         for (it->ct = 0; it->ct < MFC_LINES; it->ct++) {
2533                 it->cache = &mrt->mfc_cache_array[it->ct];
2534                 list_for_each_entry_rcu(mfc, it->cache, list)
2535                         if (pos-- == 0)
2536                                 return mfc;
2537         }
2538         rcu_read_unlock();
2539
2540         spin_lock_bh(&mfc_unres_lock);
2541         it->cache = &mrt->mfc_unres_queue;
2542         list_for_each_entry(mfc, it->cache, list)
2543                 if (pos-- == 0)
2544                         return mfc;
2545         spin_unlock_bh(&mfc_unres_lock);
2546
2547         it->cache = NULL;
2548         return NULL;
2549 }
2550
2551
2552 static void *ipmr_mfc_seq_start(struct seq_file *seq, loff_t *pos)
2553 {
2554         struct ipmr_mfc_iter *it = seq->private;
2555         struct net *net = seq_file_net(seq);
2556         struct mr_table *mrt;
2557
2558         mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2559         if (mrt == NULL)
2560                 return ERR_PTR(-ENOENT);
2561
2562         it->mrt = mrt;
2563         it->cache = NULL;
2564         it->ct = 0;
2565         return *pos ? ipmr_mfc_seq_idx(net, seq->private, *pos - 1)
2566                 : SEQ_START_TOKEN;
2567 }
2568
2569 static void *ipmr_mfc_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2570 {
2571         struct mfc_cache *mfc = v;
2572         struct ipmr_mfc_iter *it = seq->private;
2573         struct net *net = seq_file_net(seq);
2574         struct mr_table *mrt = it->mrt;
2575
2576         ++*pos;
2577
2578         if (v == SEQ_START_TOKEN)
2579                 return ipmr_mfc_seq_idx(net, seq->private, 0);
2580
2581         if (mfc->list.next != it->cache)
2582                 return list_entry(mfc->list.next, struct mfc_cache, list);
2583
2584         if (it->cache == &mrt->mfc_unres_queue)
2585                 goto end_of_list;
2586
2587         BUG_ON(it->cache != &mrt->mfc_cache_array[it->ct]);
2588
2589         while (++it->ct < MFC_LINES) {
2590                 it->cache = &mrt->mfc_cache_array[it->ct];
2591                 if (list_empty(it->cache))
2592                         continue;
2593                 return list_first_entry(it->cache, struct mfc_cache, list);
2594         }
2595
2596         /* exhausted cache_array, show unresolved */
2597         rcu_read_unlock();
2598         it->cache = &mrt->mfc_unres_queue;
2599         it->ct = 0;
2600
2601         spin_lock_bh(&mfc_unres_lock);
2602         if (!list_empty(it->cache))
2603                 return list_first_entry(it->cache, struct mfc_cache, list);
2604
2605 end_of_list:
2606         spin_unlock_bh(&mfc_unres_lock);
2607         it->cache = NULL;
2608
2609         return NULL;
2610 }
2611
2612 static void ipmr_mfc_seq_stop(struct seq_file *seq, void *v)
2613 {
2614         struct ipmr_mfc_iter *it = seq->private;
2615         struct mr_table *mrt = it->mrt;
2616
2617         if (it->cache == &mrt->mfc_unres_queue)
2618                 spin_unlock_bh(&mfc_unres_lock);
2619         else if (it->cache == &mrt->mfc_cache_array[it->ct])
2620                 rcu_read_unlock();
2621 }
2622
2623 static int ipmr_mfc_seq_show(struct seq_file *seq, void *v)
2624 {
2625         int n;
2626
2627         if (v == SEQ_START_TOKEN) {
2628                 seq_puts(seq,
2629                  "Group    Origin   Iif     Pkts    Bytes    Wrong Oifs\n");
2630         } else {
2631                 const struct mfc_cache *mfc = v;
2632                 const struct ipmr_mfc_iter *it = seq->private;
2633                 const struct mr_table *mrt = it->mrt;
2634
2635                 seq_printf(seq, "%08X %08X %-3hd",
2636                            (__force u32) mfc->mfc_mcastgrp,
2637                            (__force u32) mfc->mfc_origin,
2638                            mfc->mfc_parent);
2639
2640                 if (it->cache != &mrt->mfc_unres_queue) {
2641                         seq_printf(seq, " %8lu %8lu %8lu",
2642                                    mfc->mfc_un.res.pkt,
2643                                    mfc->mfc_un.res.bytes,
2644                                    mfc->mfc_un.res.wrong_if);
2645                         for (n = mfc->mfc_un.res.minvif;
2646                              n < mfc->mfc_un.res.maxvif; n++) {
2647                                 if (VIF_EXISTS(mrt, n) &&
2648                                     mfc->mfc_un.res.ttls[n] < 255)
2649                                         seq_printf(seq,
2650                                            " %2d:%-3d",
2651                                            n, mfc->mfc_un.res.ttls[n]);
2652                         }
2653                 } else {
2654                         /* unresolved mfc_caches don't contain
2655                          * pkt, bytes and wrong_if values
2656                          */
2657                         seq_printf(seq, " %8lu %8lu %8lu", 0ul, 0ul, 0ul);
2658                 }
2659                 seq_putc(seq, '\n');
2660         }
2661         return 0;
2662 }
2663
2664 static const struct seq_operations ipmr_mfc_seq_ops = {
2665         .start = ipmr_mfc_seq_start,
2666         .next  = ipmr_mfc_seq_next,
2667         .stop  = ipmr_mfc_seq_stop,
2668         .show  = ipmr_mfc_seq_show,
2669 };
2670
2671 static int ipmr_mfc_open(struct inode *inode, struct file *file)
2672 {
2673         return seq_open_net(inode, file, &ipmr_mfc_seq_ops,
2674                             sizeof(struct ipmr_mfc_iter));
2675 }
2676
2677 static const struct file_operations ipmr_mfc_fops = {
2678         .owner   = THIS_MODULE,
2679         .open    = ipmr_mfc_open,
2680         .read    = seq_read,
2681         .llseek  = seq_lseek,
2682         .release = seq_release_net,
2683 };
2684 #endif
2685
2686 #ifdef CONFIG_IP_PIMSM_V2
2687 static const struct net_protocol pim_protocol = {
2688         .handler        =       pim_rcv,
2689         .netns_ok       =       1,
2690 };
2691 #endif
2692
2693
2694 /*
2695  *      Setup for IP multicast routing
2696  */
2697 static int __net_init ipmr_net_init(struct net *net)
2698 {
2699         int err;
2700
2701         err = ipmr_rules_init(net);
2702         if (err < 0)
2703                 goto fail;
2704
2705 #ifdef CONFIG_PROC_FS
2706         err = -ENOMEM;
2707         if (!proc_create("ip_mr_vif", 0, net->proc_net, &ipmr_vif_fops))
2708                 goto proc_vif_fail;
2709         if (!proc_create("ip_mr_cache", 0, net->proc_net, &ipmr_mfc_fops))
2710                 goto proc_cache_fail;
2711 #endif
2712         return 0;
2713
2714 #ifdef CONFIG_PROC_FS
2715 proc_cache_fail:
2716         remove_proc_entry("ip_mr_vif", net->proc_net);
2717 proc_vif_fail:
2718         ipmr_rules_exit(net);
2719 #endif
2720 fail:
2721         return err;
2722 }
2723
2724 static void __net_exit ipmr_net_exit(struct net *net)
2725 {
2726 #ifdef CONFIG_PROC_FS
2727         remove_proc_entry("ip_mr_cache", net->proc_net);
2728         remove_proc_entry("ip_mr_vif", net->proc_net);
2729 #endif
2730         ipmr_rules_exit(net);
2731 }
2732
2733 static struct pernet_operations ipmr_net_ops = {
2734         .init = ipmr_net_init,
2735         .exit = ipmr_net_exit,
2736 };
2737
2738 int __init ip_mr_init(void)
2739 {
2740         int err;
2741
2742         mrt_cachep = kmem_cache_create("ip_mrt_cache",
2743                                        sizeof(struct mfc_cache),
2744                                        0, SLAB_HWCACHE_ALIGN | SLAB_PANIC,
2745                                        NULL);
2746         if (!mrt_cachep)
2747                 return -ENOMEM;
2748
2749         err = register_pernet_subsys(&ipmr_net_ops);
2750         if (err)
2751                 goto reg_pernet_fail;
2752
2753         err = register_netdevice_notifier(&ip_mr_notifier);
2754         if (err)
2755                 goto reg_notif_fail;
2756 #ifdef CONFIG_IP_PIMSM_V2
2757         if (inet_add_protocol(&pim_protocol, IPPROTO_PIM) < 0) {
2758                 pr_err("%s: can't add PIM protocol\n", __func__);
2759                 err = -EAGAIN;
2760                 goto add_proto_fail;
2761         }
2762 #endif
2763         rtnl_register(RTNL_FAMILY_IPMR, RTM_GETROUTE,
2764                       NULL, ipmr_rtm_dumproute, NULL);
2765         return 0;
2766
2767 #ifdef CONFIG_IP_PIMSM_V2
2768 add_proto_fail:
2769         unregister_netdevice_notifier(&ip_mr_notifier);
2770 #endif
2771 reg_notif_fail:
2772         unregister_pernet_subsys(&ipmr_net_ops);
2773 reg_pernet_fail:
2774         kmem_cache_destroy(mrt_cachep);
2775         return err;
2776 }