net/ipv4/ipmr.c

   1 // SPDX-License-Identifier: GPL-2.0-or-later
   2 /*
   3  *      IP multicast routing support for mrouted 3.6/3.8
   4  *
   5  *              (c) 1995 Alan Cox, <alan@lxorguk.ukuu.org.uk>
   6  *        Linux Consultancy and Custom Driver Development
   7  *
   8  *      Fixes:
   9  *      Michael Chastain        :       Incorrect size of copying.
  10  *      Alan Cox                :       Added the cache manager code
  11  *      Alan Cox                :       Fixed the clone/copy bug and device race.
  12  *      Mike McLagan            :       Routing by source
  13  *      Malcolm Beattie         :       Buffer handling fixes.
  14  *      Alexey Kuznetsov        :       Double buffer free and other fixes.
  15  *      SVR Anand               :       Fixed several multicast bugs and problems.
  16  *      Alexey Kuznetsov        :       Status, optimisations and more.
  17  *      Brad Parker             :       Better behaviour on mrouted upcall
  18  *                                      overflow.
  19  *      Carlos Picoto           :       PIMv1 Support
  20  *      Pavlin Ivanov Radoslavov:       PIMv2 Registers must checksum only PIM header
  21  *                                      Relax this requirement to work with older peers.
  22  */
  23
  24 #include <linux/uaccess.h>
  25 #include <linux/types.h>
  26 #include <linux/cache.h>
  27 #include <linux/capability.h>
  28 #include <linux/errno.h>
  29 #include <linux/mm.h>
  30 #include <linux/kernel.h>
  31 #include <linux/fcntl.h>
  32 #include <linux/stat.h>
  33 #include <linux/socket.h>
  34 #include <linux/in.h>
  35 #include <linux/inet.h>
  36 #include <linux/netdevice.h>
  37 #include <linux/inetdevice.h>
  38 #include <linux/igmp.h>
  39 #include <linux/proc_fs.h>
  40 #include <linux/seq_file.h>
  41 #include <linux/mroute.h>
  42 #include <linux/init.h>
  43 #include <linux/if_ether.h>
  44 #include <linux/slab.h>
  45 #include <net/net_namespace.h>
  46 #include <net/ip.h>
  47 #include <net/protocol.h>
  48 #include <linux/skbuff.h>
  49 #include <net/route.h>
  50 #include <net/icmp.h>
  51 #include <net/udp.h>
  52 #include <net/raw.h>
  53 #include <linux/notifier.h>
  54 #include <linux/if_arp.h>
  55 #include <linux/netfilter_ipv4.h>
  56 #include <linux/compat.h>
  57 #include <linux/export.h>
  58 #include <linux/rhashtable.h>
  59 #include <net/ip_tunnels.h>
  60 #include <net/checksum.h>
  61 #include <net/netlink.h>
  62 #include <net/fib_rules.h>
  63 #include <linux/netconf.h>
  64 #include <net/rtnh.h>
  65
  66 #include <linux/nospec.h>
  67
  68 struct ipmr_rule {
  69         struct fib_rule         common;
  70 };
  71
  72 struct ipmr_result {
  73         struct mr_table         *mrt;
  74 };
  75
  76 /* Big lock, protecting vif table, mrt cache and mroute socket state.
  77  * Note that the changes are semaphored via rtnl_lock.
  78  */
  79
  80 static DEFINE_RWLOCK(mrt_lock);
  81
  82 /* Multicast router control variables */
  83
  84 /* Special spinlock for queue of unresolved entries */
  85 static DEFINE_SPINLOCK(mfc_unres_lock);
  86
  87 /* We return to original Alan's scheme. Hash table of resolved
  88  * entries is changed only in process context and protected
  89  * with weak lock mrt_lock. Queue of unresolved entries is protected
  90  * with strong spinlock mfc_unres_lock.
  91  *
  92  * In this case data path is free of exclusive locks at all.
  93  */
  94
  95 static struct kmem_cache *mrt_cachep __ro_after_init;
  96
  97 static struct mr_table *ipmr_new_table(struct net *net, u32 id);
  98 static void ipmr_free_table(struct mr_table *mrt);
  99
 100 static void ip_mr_forward(struct net *net, struct mr_table *mrt,
 101                           struct net_device *dev, struct sk_buff *skb,
 102                           struct mfc_cache *cache, int local);
 103 static int ipmr_cache_report(struct mr_table *mrt,
 104                              struct sk_buff *pkt, vifi_t vifi, int assert);
 105 static void mroute_netlink_event(struct mr_table *mrt, struct mfc_cache *mfc,
 106                                  int cmd);
 107 static void igmpmsg_netlink_event(struct mr_table *mrt, struct sk_buff *pkt);
 108 static void mroute_clean_tables(struct mr_table *mrt, int flags);
 109 static void ipmr_expire_process(struct timer_list *t);
 110
 111 #ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES
 112 #define ipmr_for_each_table(mrt, net)                                   \
 113         list_for_each_entry_rcu(mrt, &net->ipv4.mr_tables, list,        \
 114                                 lockdep_rtnl_is_held() ||               \
 115                                 list_empty(&net->ipv4.mr_tables))
 116
 117 static struct mr_table *ipmr_mr_table_iter(struct net *net,
 118                                            struct mr_table *mrt)
 119 {
 120         struct mr_table *ret;
 121
 122         if (!mrt)
 123                 ret = list_entry_rcu(net->ipv4.mr_tables.next,
 124                                      struct mr_table, list);
 125         else
 126                 ret = list_entry_rcu(mrt->list.next,
 127                                      struct mr_table, list);
 128
 129         if (&ret->list == &net->ipv4.mr_tables)
 130                 return NULL;
 131         return ret;
 132 }
 133
 134 static struct mr_table *ipmr_get_table(struct net *net, u32 id)
 135 {
 136         struct mr_table *mrt;
 137
 138         ipmr_for_each_table(mrt, net) {
 139                 if (mrt->id == id)
 140                         return mrt;
 141         }
 142         return NULL;
 143 }
 144
 145 static int ipmr_fib_lookup(struct net *net, struct flowi4 *flp4,
 146                            struct mr_table **mrt)
 147 {
 148         int err;
 149         struct ipmr_result res;
 150         struct fib_lookup_arg arg = {
 151                 .result = &res,
 152                 .flags = FIB_LOOKUP_NOREF,
 153         };
 154
 155         /* update flow if oif or iif point to device enslaved to l3mdev */
 156         l3mdev_update_flow(net, flowi4_to_flowi(flp4));
 157
 158         err = fib_rules_lookup(net->ipv4.mr_rules_ops,
 159                                flowi4_to_flowi(flp4), 0, &arg);
 160         if (err < 0)
 161                 return err;
 162         *mrt = res.mrt;
 163         return 0;
 164 }
 165
 166 static int ipmr_rule_action(struct fib_rule *rule, struct flowi *flp,
 167                             int flags, struct fib_lookup_arg *arg)
 168 {
 169         struct ipmr_result *res = arg->result;
 170         struct mr_table *mrt;
 171
 172         switch (rule->action) {
 173         case FR_ACT_TO_TBL:
 174                 break;
 175         case FR_ACT_UNREACHABLE:
 176                 return -ENETUNREACH;
 177         case FR_ACT_PROHIBIT:
 178                 return -EACCES;
 179         case FR_ACT_BLACKHOLE:
 180         default:
 181                 return -EINVAL;
 182         }
 183
 184         arg->table = fib_rule_get_table(rule, arg);
 185
 186         mrt = ipmr_get_table(rule->fr_net, arg->table);
 187         if (!mrt)
 188                 return -EAGAIN;
 189         res->mrt = mrt;
 190         return 0;
 191 }
 192
 193 static int ipmr_rule_match(struct fib_rule *rule, struct flowi *fl, int flags)
 194 {
 195         return 1;
 196 }
 197
 198 static const struct nla_policy ipmr_rule_policy[FRA_MAX + 1] = {
 199         FRA_GENERIC_POLICY,
 200 };
 201
 202 static int ipmr_rule_configure(struct fib_rule *rule, struct sk_buff *skb,
 203                                struct fib_rule_hdr *frh, struct nlattr **tb,
 204                                struct netlink_ext_ack *extack)
 205 {
 206         return 0;
 207 }
 208
 209 static int ipmr_rule_compare(struct fib_rule *rule, struct fib_rule_hdr *frh,
 210                              struct nlattr **tb)
 211 {
 212         return 1;
 213 }
 214
 215 static int ipmr_rule_fill(struct fib_rule *rule, struct sk_buff *skb,
 216                           struct fib_rule_hdr *frh)
 217 {
 218         frh->dst_len = 0;
 219         frh->src_len = 0;
 220         frh->tos     = 0;
 221         return 0;
 222 }
 223
 224 static const struct fib_rules_ops __net_initconst ipmr_rules_ops_template = {
 225         .family         = RTNL_FAMILY_IPMR,
 226         .rule_size      = sizeof(struct ipmr_rule),
 227         .addr_size      = sizeof(u32),
 228         .action         = ipmr_rule_action,
 229         .match          = ipmr_rule_match,
 230         .configure      = ipmr_rule_configure,
 231         .compare        = ipmr_rule_compare,
 232         .fill           = ipmr_rule_fill,
 233         .nlgroup        = RTNLGRP_IPV4_RULE,
 234         .policy         = ipmr_rule_policy,
 235         .owner          = THIS_MODULE,
 236 };
 237
 238 static int __net_init ipmr_rules_init(struct net *net)
 239 {
 240         struct fib_rules_ops *ops;
 241         struct mr_table *mrt;
 242         int err;
 243
 244         ops = fib_rules_register(&ipmr_rules_ops_template, net);
 245         if (IS_ERR(ops))
 246                 return PTR_ERR(ops);
 247
 248         INIT_LIST_HEAD(&net->ipv4.mr_tables);
 249
 250         mrt = ipmr_new_table(net, RT_TABLE_DEFAULT);
 251         if (IS_ERR(mrt)) {
 252                 err = PTR_ERR(mrt);
 253                 goto err1;
 254         }
 255
 256         err = fib_default_rule_add(ops, 0x7fff, RT_TABLE_DEFAULT, 0);
 257         if (err < 0)
 258                 goto err2;
 259
 260         net->ipv4.mr_rules_ops = ops;
 261         return 0;
 262
 263 err2:
 264         rtnl_lock();
 265         ipmr_free_table(mrt);
 266         rtnl_unlock();
 267 err1:
 268         fib_rules_unregister(ops);
 269         return err;
 270 }
 271
 272 static void __net_exit ipmr_rules_exit(struct net *net)
 273 {
 274         struct mr_table *mrt, *next;
 275
 276         rtnl_lock();
 277         list_for_each_entry_safe(mrt, next, &net->ipv4.mr_tables, list) {
 278                 list_del(&mrt->list);
 279                 ipmr_free_table(mrt);
 280         }
 281         fib_rules_unregister(net->ipv4.mr_rules_ops);
 282         rtnl_unlock();
 283 }
 284
 285 static int ipmr_rules_dump(struct net *net, struct notifier_block *nb,
 286                            struct netlink_ext_ack *extack)
 287 {
 288         return fib_rules_dump(net, nb, RTNL_FAMILY_IPMR, extack);
 289 }
 290
 291 static unsigned int ipmr_rules_seq_read(struct net *net)
 292 {
 293         return fib_rules_seq_read(net, RTNL_FAMILY_IPMR);
 294 }
 295
 296 bool ipmr_rule_default(const struct fib_rule *rule)
 297 {
 298         return fib_rule_matchall(rule) && rule->table == RT_TABLE_DEFAULT;
 299 }
 300 EXPORT_SYMBOL(ipmr_rule_default);
 301 #else
 302 #define ipmr_for_each_table(mrt, net) \
 303         for (mrt = net->ipv4.mrt; mrt; mrt = NULL)
 304
 305 static struct mr_table *ipmr_mr_table_iter(struct net *net,
 306                                            struct mr_table *mrt)
 307 {
 308         if (!mrt)
 309                 return net->ipv4.mrt;
 310         return NULL;
 311 }
 312
 313 static struct mr_table *ipmr_get_table(struct net *net, u32 id)
 314 {
 315         return net->ipv4.mrt;
 316 }
 317
 318 static int ipmr_fib_lookup(struct net *net, struct flowi4 *flp4,
 319                            struct mr_table **mrt)
 320 {
 321         *mrt = net->ipv4.mrt;
 322         return 0;
 323 }
 324
 325 static int __net_init ipmr_rules_init(struct net *net)
 326 {
 327         struct mr_table *mrt;
 328
 329         mrt = ipmr_new_table(net, RT_TABLE_DEFAULT);
 330         if (IS_ERR(mrt))
 331                 return PTR_ERR(mrt);
 332         net->ipv4.mrt = mrt;
 333         return 0;
 334 }
 335
 336 static void __net_exit ipmr_rules_exit(struct net *net)
 337 {
 338         rtnl_lock();
 339         ipmr_free_table(net->ipv4.mrt);
 340         net->ipv4.mrt = NULL;
 341         rtnl_unlock();
 342 }
 343
 344 static int ipmr_rules_dump(struct net *net, struct notifier_block *nb,
 345                            struct netlink_ext_ack *extack)
 346 {
 347         return 0;
 348 }
 349
 350 static unsigned int ipmr_rules_seq_read(struct net *net)
 351 {
 352         return 0;
 353 }
 354
 355 bool ipmr_rule_default(const struct fib_rule *rule)
 356 {
 357         return true;
 358 }
 359 EXPORT_SYMBOL(ipmr_rule_default);
 360 #endif
 361
 362 static inline int ipmr_hash_cmp(struct rhashtable_compare_arg *arg,
 363                                 const void *ptr)
 364 {
 365         const struct mfc_cache_cmp_arg *cmparg = arg->key;
 366         struct mfc_cache *c = (struct mfc_cache *)ptr;
 367
 368         return cmparg->mfc_mcastgrp != c->mfc_mcastgrp ||
 369                cmparg->mfc_origin != c->mfc_origin;
 370 }
 371
 372 static const struct rhashtable_params ipmr_rht_params = {
 373         .head_offset = offsetof(struct mr_mfc, mnode),
 374         .key_offset = offsetof(struct mfc_cache, cmparg),
 375         .key_len = sizeof(struct mfc_cache_cmp_arg),
 376         .nelem_hint = 3,
 377         .obj_cmpfn = ipmr_hash_cmp,
 378         .automatic_shrinking = true,
 379 };
 380
 381 static void ipmr_new_table_set(struct mr_table *mrt,
 382                                struct net *net)
 383 {
 384 #ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES
 385         list_add_tail_rcu(&mrt->list, &net->ipv4.mr_tables);
 386 #endif
 387 }
 388
 389 static struct mfc_cache_cmp_arg ipmr_mr_table_ops_cmparg_any = {
 390         .mfc_mcastgrp = htonl(INADDR_ANY),
 391         .mfc_origin = htonl(INADDR_ANY),
 392 };
 393
 394 static struct mr_table_ops ipmr_mr_table_ops = {
 395         .rht_params = &ipmr_rht_params,
 396         .cmparg_any = &ipmr_mr_table_ops_cmparg_any,
 397 };
 398
 399 static struct mr_table *ipmr_new_table(struct net *net, u32 id)
 400 {
 401         struct mr_table *mrt;
 402
 403         /* "pimreg%u" should not exceed 16 bytes (IFNAMSIZ) */
 404         if (id != RT_TABLE_DEFAULT && id >= 1000000000)
 405                 return ERR_PTR(-EINVAL);
 406
 407         mrt = ipmr_get_table(net, id);
 408         if (mrt)
 409                 return mrt;
 410
 411         return mr_table_alloc(net, id, &ipmr_mr_table_ops,
 412                               ipmr_expire_process, ipmr_new_table_set);
 413 }
 414
 415 static void ipmr_free_table(struct mr_table *mrt)
 416 {
 417         del_timer_sync(&mrt->ipmr_expire_timer);
 418         mroute_clean_tables(mrt, MRT_FLUSH_VIFS | MRT_FLUSH_VIFS_STATIC |
 419                                  MRT_FLUSH_MFC | MRT_FLUSH_MFC_STATIC);
 420         rhltable_destroy(&mrt->mfc_hash);
 421         kfree(mrt);
 422 }
 423
 424 /* Service routines creating virtual interfaces: DVMRP tunnels and PIMREG */
 425
 426 /* Initialize ipmr pimreg/tunnel in_device */
 427 static bool ipmr_init_vif_indev(const struct net_device *dev)
 428 {
 429         struct in_device *in_dev;
 430
 431         ASSERT_RTNL();
 432
 433         in_dev = __in_dev_get_rtnl(dev);
 434         if (!in_dev)
 435                 return false;
 436         ipv4_devconf_setall(in_dev);
 437         neigh_parms_data_state_setall(in_dev->arp_parms);
 438         IPV4_DEVCONF(in_dev->cnf, RP_FILTER) = 0;
 439
 440         return true;
 441 }
 442
 443 static struct net_device *ipmr_new_tunnel(struct net *net, struct vifctl *v)
 444 {
 445         struct net_device *tunnel_dev, *new_dev;
 446         struct ip_tunnel_parm p = { };
 447         int err;
 448
 449         tunnel_dev = __dev_get_by_name(net, "tunl0");
 450         if (!tunnel_dev)
 451                 goto out;
 452
 453         p.iph.daddr = v->vifc_rmt_addr.s_addr;
 454         p.iph.saddr = v->vifc_lcl_addr.s_addr;
 455         p.iph.version = 4;
 456         p.iph.ihl = 5;
 457         p.iph.protocol = IPPROTO_IPIP;
 458         sprintf(p.name, "dvmrp%d", v->vifc_vifi);
 459
 460         if (!tunnel_dev->netdev_ops->ndo_tunnel_ctl)
 461                 goto out;
 462         err = tunnel_dev->netdev_ops->ndo_tunnel_ctl(tunnel_dev, &p,
 463                         SIOCADDTUNNEL);
 464         if (err)
 465                 goto out;
 466
 467         new_dev = __dev_get_by_name(net, p.name);
 468         if (!new_dev)
 469                 goto out;
 470
 471         new_dev->flags |= IFF_MULTICAST;
 472         if (!ipmr_init_vif_indev(new_dev))
 473                 goto out_unregister;
 474         if (dev_open(new_dev, NULL))
 475                 goto out_unregister;
 476         dev_hold(new_dev);
 477         err = dev_set_allmulti(new_dev, 1);
 478         if (err) {
 479                 dev_close(new_dev);
 480                 tunnel_dev->netdev_ops->ndo_tunnel_ctl(tunnel_dev, &p,
 481                                 SIOCDELTUNNEL);
 482                 dev_put(new_dev);
 483                 new_dev = ERR_PTR(err);
 484         }
 485         return new_dev;
 486
 487 out_unregister:
 488         unregister_netdevice(new_dev);
 489 out:
 490         return ERR_PTR(-ENOBUFS);
 491 }
 492
 493 #if defined(CONFIG_IP_PIMSM_V1) || defined(CONFIG_IP_PIMSM_V2)
 494 static netdev_tx_t reg_vif_xmit(struct sk_buff *skb, struct net_device *dev)
 495 {
 496         struct net *net = dev_net(dev);
 497         struct mr_table *mrt;
 498         struct flowi4 fl4 = {
 499                 .flowi4_oif     = dev->ifindex,
 500                 .flowi4_iif     = skb->skb_iif ? : LOOPBACK_IFINDEX,
 501                 .flowi4_mark    = skb->mark,
 502         };
 503         int err;
 504
 505         err = ipmr_fib_lookup(net, &fl4, &mrt);
 506         if (err < 0) {
 507                 kfree_skb(skb);
 508                 return err;
 509         }
 510
 511         read_lock(&mrt_lock);
 512         dev->stats.tx_bytes += skb->len;
 513         dev->stats.tx_packets++;
 514         ipmr_cache_report(mrt, skb, mrt->mroute_reg_vif_num, IGMPMSG_WHOLEPKT);
 515         read_unlock(&mrt_lock);
 516         kfree_skb(skb);
 517         return NETDEV_TX_OK;
 518 }
 519
 520 static int reg_vif_get_iflink(const struct net_device *dev)
 521 {
 522         return 0;
 523 }
 524
 525 static const struct net_device_ops reg_vif_netdev_ops = {
 526         .ndo_start_xmit = reg_vif_xmit,
 527         .ndo_get_iflink = reg_vif_get_iflink,
 528 };
 529
 530 static void reg_vif_setup(struct net_device *dev)
 531 {
 532         dev->type               = ARPHRD_PIMREG;
 533         dev->mtu                = ETH_DATA_LEN - sizeof(struct iphdr) - 8;
 534         dev->flags              = IFF_NOARP;
 535         dev->netdev_ops         = &reg_vif_netdev_ops;
 536         dev->needs_free_netdev  = true;
 537         dev->features           |= NETIF_F_NETNS_LOCAL;
 538 }
 539
 540 static struct net_device *ipmr_reg_vif(struct net *net, struct mr_table *mrt)
 541 {
 542         struct net_device *dev;
 543         char name[IFNAMSIZ];
 544
 545         if (mrt->id == RT_TABLE_DEFAULT)
 546                 sprintf(name, "pimreg");
 547         else
 548                 sprintf(name, "pimreg%u", mrt->id);
 549
 550         dev = alloc_netdev(0, name, NET_NAME_UNKNOWN, reg_vif_setup);
 551
 552         if (!dev)
 553                 return NULL;
 554
 555         dev_net_set(dev, net);
 556
 557         if (register_netdevice(dev)) {
 558                 free_netdev(dev);
 559                 return NULL;
 560         }
 561
 562         if (!ipmr_init_vif_indev(dev))
 563                 goto failure;
 564         if (dev_open(dev, NULL))
 565                 goto failure;
 566
 567         dev_hold(dev);
 568
 569         return dev;
 570
 571 failure:
 572         unregister_netdevice(dev);
 573         return NULL;
 574 }
 575
 576 /* called with rcu_read_lock() */
 577 static int __pim_rcv(struct mr_table *mrt, struct sk_buff *skb,
 578                      unsigned int pimlen)
 579 {
 580         struct net_device *reg_dev = NULL;
 581         struct iphdr *encap;
 582
 583         encap = (struct iphdr *)(skb_transport_header(skb) + pimlen);
 584         /* Check that:
 585          * a. packet is really sent to a multicast group
 586          * b. packet is not a NULL-REGISTER
 587          * c. packet is not truncated
 588          */
 589         if (!ipv4_is_multicast(encap->daddr) ||
 590             encap->tot_len == 0 ||
 591             ntohs(encap->tot_len) + pimlen > skb->len)
 592                 return 1;
 593
 594         read_lock(&mrt_lock);
 595         if (mrt->mroute_reg_vif_num >= 0)
 596                 reg_dev = mrt->vif_table[mrt->mroute_reg_vif_num].dev;
 597         read_unlock(&mrt_lock);
 598
 599         if (!reg_dev)
 600                 return 1;
 601
 602         skb->mac_header = skb->network_header;
 603         skb_pull(skb, (u8 *)encap - skb->data);
 604         skb_reset_network_header(skb);
 605         skb->protocol = htons(ETH_P_IP);
 606         skb->ip_summed = CHECKSUM_NONE;
 607
 608         skb_tunnel_rx(skb, reg_dev, dev_net(reg_dev));
 609
 610         netif_rx(skb);
 611
 612         return NET_RX_SUCCESS;
 613 }
 614 #else
 615 static struct net_device *ipmr_reg_vif(struct net *net, struct mr_table *mrt)
 616 {
 617         return NULL;
 618 }
 619 #endif
 620
 621 static int call_ipmr_vif_entry_notifiers(struct net *net,
 622                                          enum fib_event_type event_type,
 623                                          struct vif_device *vif,
 624                                          vifi_t vif_index, u32 tb_id)
 625 {
 626         return mr_call_vif_notifiers(net, RTNL_FAMILY_IPMR, event_type,
 627                                      vif, vif_index, tb_id,
 628                                      &net->ipv4.ipmr_seq);
 629 }
 630
 631 static int call_ipmr_mfc_entry_notifiers(struct net *net,
 632                                          enum fib_event_type event_type,
 633                                          struct mfc_cache *mfc, u32 tb_id)
 634 {
 635         return mr_call_mfc_notifiers(net, RTNL_FAMILY_IPMR, event_type,
 636                                      &mfc->_c, tb_id, &net->ipv4.ipmr_seq);
 637 }
 638
 639 /**
 640  *      vif_delete - Delete a VIF entry
 641  *      @mrt: Table to delete from
 642  *      @vifi: VIF identifier to delete
 643  *      @notify: Set to 1, if the caller is a notifier_call
 644  *      @head: if unregistering the VIF, place it on this queue
 645  */
 646 static int vif_delete(struct mr_table *mrt, int vifi, int notify,
 647                       struct list_head *head)
 648 {
 649         struct net *net = read_pnet(&mrt->net);
 650         struct vif_device *v;
 651         struct net_device *dev;
 652         struct in_device *in_dev;
 653
 654         if (vifi < 0 || vifi >= mrt->maxvif)
 655                 return -EADDRNOTAVAIL;
 656
 657         v = &mrt->vif_table[vifi];
 658
 659         if (VIF_EXISTS(mrt, vifi))
 660                 call_ipmr_vif_entry_notifiers(net, FIB_EVENT_VIF_DEL, v, vifi,
 661                                               mrt->id);
 662
 663         write_lock_bh(&mrt_lock);
 664         dev = v->dev;
 665         v->dev = NULL;
 666
 667         if (!dev) {
 668                 write_unlock_bh(&mrt_lock);
 669                 return -EADDRNOTAVAIL;
 670         }
 671
 672         if (vifi == mrt->mroute_reg_vif_num)
 673                 mrt->mroute_reg_vif_num = -1;
 674
 675         if (vifi + 1 == mrt->maxvif) {
 676                 int tmp;
 677
 678                 for (tmp = vifi - 1; tmp >= 0; tmp--) {
 679                         if (VIF_EXISTS(mrt, tmp))
 680                                 break;
 681                 }
 682                 mrt->maxvif = tmp+1;
 683         }
 684
 685         write_unlock_bh(&mrt_lock);
 686
 687         dev_set_allmulti(dev, -1);
 688
 689         in_dev = __in_dev_get_rtnl(dev);
 690         if (in_dev) {
 691                 IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)--;
 692                 inet_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
 693                                             NETCONFA_MC_FORWARDING,
 694                                             dev->ifindex, &in_dev->cnf);
 695                 ip_rt_multicast_event(in_dev);
 696         }
 697
 698         if (v->flags & (VIFF_TUNNEL | VIFF_REGISTER) && !notify)
 699                 unregister_netdevice_queue(dev, head);
 700
 701         dev_put(dev);
 702         return 0;
 703 }
 704
 705 static void ipmr_cache_free_rcu(struct rcu_head *head)
 706 {
 707         struct mr_mfc *c = container_of(head, struct mr_mfc, rcu);
 708
 709         kmem_cache_free(mrt_cachep, (struct mfc_cache *)c);
 710 }
 711
 712 static void ipmr_cache_free(struct mfc_cache *c)
 713 {
 714         call_rcu(&c->_c.rcu, ipmr_cache_free_rcu);
 715 }
 716
 717 /* Destroy an unresolved cache entry, killing queued skbs
 718  * and reporting error to netlink readers.
 719  */
 720 static void ipmr_destroy_unres(struct mr_table *mrt, struct mfc_cache *c)
 721 {
 722         struct net *net = read_pnet(&mrt->net);
 723         struct sk_buff *skb;
 724         struct nlmsgerr *e;
 725
 726         atomic_dec(&mrt->cache_resolve_queue_len);
 727
 728         while ((skb = skb_dequeue(&c->_c.mfc_un.unres.unresolved))) {
 729                 if (ip_hdr(skb)->version == 0) {
 730                         struct nlmsghdr *nlh = skb_pull(skb,
 731                                                         sizeof(struct iphdr));
 732                         nlh->nlmsg_type = NLMSG_ERROR;
 733                         nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
 734                         skb_trim(skb, nlh->nlmsg_len);
 735                         e = nlmsg_data(nlh);
 736                         e->error = -ETIMEDOUT;
 737                         memset(&e->msg, 0, sizeof(e->msg));
 738
 739                         rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
 740                 } else {
 741                         kfree_skb(skb);
 742                 }
 743         }
 744
 745         ipmr_cache_free(c);
 746 }
 747
 748 /* Timer process for the unresolved queue. */
 749 static void ipmr_expire_process(struct timer_list *t)
 750 {
 751         struct mr_table *mrt = from_timer(mrt, t, ipmr_expire_timer);
 752         struct mr_mfc *c, *next;
 753         unsigned long expires;
 754         unsigned long now;
 755
 756         if (!spin_trylock(&mfc_unres_lock)) {
 757                 mod_timer(&mrt->ipmr_expire_timer, jiffies+HZ/10);
 758                 return;
 759         }
 760
 761         if (list_empty(&mrt->mfc_unres_queue))
 762                 goto out;
 763
 764         now = jiffies;
 765         expires = 10*HZ;
 766
 767         list_for_each_entry_safe(c, next, &mrt->mfc_unres_queue, list) {
 768                 if (time_after(c->mfc_un.unres.expires, now)) {
 769                         unsigned long interval = c->mfc_un.unres.expires - now;
 770                         if (interval < expires)
 771                                 expires = interval;
 772                         continue;
 773                 }
 774
 775                 list_del(&c->list);
 776                 mroute_netlink_event(mrt, (struct mfc_cache *)c, RTM_DELROUTE);
 777                 ipmr_destroy_unres(mrt, (struct mfc_cache *)c);
 778         }
 779
 780         if (!list_empty(&mrt->mfc_unres_queue))
 781                 mod_timer(&mrt->ipmr_expire_timer, jiffies + expires);
 782
 783 out:
 784         spin_unlock(&mfc_unres_lock);
 785 }
 786
 787 /* Fill oifs list. It is called under write locked mrt_lock. */
 788 static void ipmr_update_thresholds(struct mr_table *mrt, struct mr_mfc *cache,
 789                                    unsigned char *ttls)
 790 {
 791         int vifi;
 792
 793         cache->mfc_un.res.minvif = MAXVIFS;
 794         cache->mfc_un.res.maxvif = 0;
 795         memset(cache->mfc_un.res.ttls, 255, MAXVIFS);
 796
 797         for (vifi = 0; vifi < mrt->maxvif; vifi++) {
 798                 if (VIF_EXISTS(mrt, vifi) &&
 799                     ttls[vifi] && ttls[vifi] < 255) {
 800                         cache->mfc_un.res.ttls[vifi] = ttls[vifi];
 801                         if (cache->mfc_un.res.minvif > vifi)
 802                                 cache->mfc_un.res.minvif = vifi;
 803                         if (cache->mfc_un.res.maxvif <= vifi)
 804                                 cache->mfc_un.res.maxvif = vifi + 1;
 805                 }
 806         }
 807         cache->mfc_un.res.lastuse = jiffies;
 808 }
 809
 810 static int vif_add(struct net *net, struct mr_table *mrt,
 811                    struct vifctl *vifc, int mrtsock)
 812 {
 813         struct netdev_phys_item_id ppid = { };
 814         int vifi = vifc->vifc_vifi;
 815         struct vif_device *v = &mrt->vif_table[vifi];
 816         struct net_device *dev;
 817         struct in_device *in_dev;
 818         int err;
 819
 820         /* Is vif busy ? */
 821         if (VIF_EXISTS(mrt, vifi))
 822                 return -EADDRINUSE;
 823
 824         switch (vifc->vifc_flags) {
 825         case VIFF_REGISTER:
 826                 if (!ipmr_pimsm_enabled())
 827                         return -EINVAL;
 828                 /* Special Purpose VIF in PIM
 829                  * All the packets will be sent to the daemon
 830                  */
 831                 if (mrt->mroute_reg_vif_num >= 0)
 832                         return -EADDRINUSE;
 833                 dev = ipmr_reg_vif(net, mrt);
 834                 if (!dev)
 835                         return -ENOBUFS;
 836                 err = dev_set_allmulti(dev, 1);
 837                 if (err) {
 838                         unregister_netdevice(dev);
 839                         dev_put(dev);
 840                         return err;
 841                 }
 842                 break;
 843         case VIFF_TUNNEL:
 844                 dev = ipmr_new_tunnel(net, vifc);
 845                 if (IS_ERR(dev))
 846                         return PTR_ERR(dev);
 847                 break;
 848         case VIFF_USE_IFINDEX:
 849         case 0:
 850                 if (vifc->vifc_flags == VIFF_USE_IFINDEX) {
 851                         dev = dev_get_by_index(net, vifc->vifc_lcl_ifindex);
 852                         if (dev && !__in_dev_get_rtnl(dev)) {
 853                                 dev_put(dev);
 854                                 return -EADDRNOTAVAIL;
 855                         }
 856                 } else {
 857                         dev = ip_dev_find(net, vifc->vifc_lcl_addr.s_addr);
 858                 }
 859                 if (!dev)
 860                         return -EADDRNOTAVAIL;
 861                 err = dev_set_allmulti(dev, 1);
 862                 if (err) {
 863                         dev_put(dev);
 864                         return err;
 865                 }
 866                 break;
 867         default:
 868                 return -EINVAL;
 869         }
 870
 871         in_dev = __in_dev_get_rtnl(dev);
 872         if (!in_dev) {
 873                 dev_put(dev);
 874                 return -EADDRNOTAVAIL;
 875         }
 876         IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)++;
 877         inet_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_MC_FORWARDING,
 878                                     dev->ifindex, &in_dev->cnf);
 879         ip_rt_multicast_event(in_dev);
 880
 881         /* Fill in the VIF structures */
 882         vif_device_init(v, dev, vifc->vifc_rate_limit,
 883                         vifc->vifc_threshold,
 884                         vifc->vifc_flags | (!mrtsock ? VIFF_STATIC : 0),
 885                         (VIFF_TUNNEL | VIFF_REGISTER));
 886
 887         err = dev_get_port_parent_id(dev, &ppid, true);
 888         if (err == 0) {
 889                 memcpy(v->dev_parent_id.id, ppid.id, ppid.id_len);
 890                 v->dev_parent_id.id_len = ppid.id_len;
 891         } else {
 892                 v->dev_parent_id.id_len = 0;
 893         }
 894
 895         v->local = vifc->vifc_lcl_addr.s_addr;
 896         v->remote = vifc->vifc_rmt_addr.s_addr;
 897
 898         /* And finish update writing critical data */
 899         write_lock_bh(&mrt_lock);
 900         v->dev = dev;
 901         if (v->flags & VIFF_REGISTER)
 902                 mrt->mroute_reg_vif_num = vifi;
 903         if (vifi+1 > mrt->maxvif)
 904                 mrt->maxvif = vifi+1;
 905         write_unlock_bh(&mrt_lock);
 906         call_ipmr_vif_entry_notifiers(net, FIB_EVENT_VIF_ADD, v, vifi, mrt->id);
 907         return 0;
 908 }
 909
 910 /* called with rcu_read_lock() */
 911 static struct mfc_cache *ipmr_cache_find(struct mr_table *mrt,
 912                                          __be32 origin,
 913                                          __be32 mcastgrp)
 914 {
 915         struct mfc_cache_cmp_arg arg = {
 916                         .mfc_mcastgrp = mcastgrp,
 917                         .mfc_origin = origin
 918         };
 919
 920         return mr_mfc_find(mrt, &arg);
 921 }
 922
 923 /* Look for a (*,G) entry */
 924 static struct mfc_cache *ipmr_cache_find_any(struct mr_table *mrt,
 925                                              __be32 mcastgrp, int vifi)
 926 {
 927         struct mfc_cache_cmp_arg arg = {
 928                         .mfc_mcastgrp = mcastgrp,
 929                         .mfc_origin = htonl(INADDR_ANY)
 930         };
 931
 932         if (mcastgrp == htonl(INADDR_ANY))
 933                 return mr_mfc_find_any_parent(mrt, vifi);
 934         return mr_mfc_find_any(mrt, vifi, &arg);
 935 }
 936
 937 /* Look for a (S,G,iif) entry if parent != -1 */
 938 static struct mfc_cache *ipmr_cache_find_parent(struct mr_table *mrt,
 939                                                 __be32 origin, __be32 mcastgrp,
 940                                                 int parent)
 941 {
 942         struct mfc_cache_cmp_arg arg = {
 943                         .mfc_mcastgrp = mcastgrp,
 944                         .mfc_origin = origin,
 945         };
 946
 947         return mr_mfc_find_parent(mrt, &arg, parent);
 948 }
 949
 950 /* Allocate a multicast cache entry */
 951 static struct mfc_cache *ipmr_cache_alloc(void)
 952 {
 953         struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_KERNEL);
 954
 955         if (c) {
 956                 c->_c.mfc_un.res.last_assert = jiffies - MFC_ASSERT_THRESH - 1;
 957                 c->_c.mfc_un.res.minvif = MAXVIFS;
 958                 c->_c.free = ipmr_cache_free_rcu;
 959                 refcount_set(&c->_c.mfc_un.res.refcount, 1);
 960         }
 961         return c;
 962 }
 963
 964 static struct mfc_cache *ipmr_cache_alloc_unres(void)
 965 {
 966         struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_ATOMIC);
 967
 968         if (c) {
 969                 skb_queue_head_init(&c->_c.mfc_un.unres.unresolved);
 970                 c->_c.mfc_un.unres.expires = jiffies + 10 * HZ;
 971         }
 972         return c;
 973 }
 974
 975 /* A cache entry has gone into a resolved state from queued */
 976 static void ipmr_cache_resolve(struct net *net, struct mr_table *mrt,
 977                                struct mfc_cache *uc, struct mfc_cache *c)
 978 {
 979         struct sk_buff *skb;
 980         struct nlmsgerr *e;
 981
 982         /* Play the pending entries through our router */
 983         while ((skb = __skb_dequeue(&uc->_c.mfc_un.unres.unresolved))) {
 984                 if (ip_hdr(skb)->version == 0) {
 985                         struct nlmsghdr *nlh = skb_pull(skb,
 986                                                         sizeof(struct iphdr));
 987
 988                         if (mr_fill_mroute(mrt, skb, &c->_c,
 989                                            nlmsg_data(nlh)) > 0) {
 990                                 nlh->nlmsg_len = skb_tail_pointer(skb) -
 991                                                  (u8 *)nlh;
 992                         } else {
 993                                 nlh->nlmsg_type = NLMSG_ERROR;
 994                                 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
 995                                 skb_trim(skb, nlh->nlmsg_len);
 996                                 e = nlmsg_data(nlh);
 997                                 e->error = -EMSGSIZE;
 998                                 memset(&e->msg, 0, sizeof(e->msg));
 999                         }
1000
1001                         rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
1002                 } else {
1003                         ip_mr_forward(net, mrt, skb->dev, skb, c, 0);
1004                 }
1005         }
1006 }
1007
1008 /* Bounce a cache query up to mrouted and netlink.
1009  *
1010  * Called under mrt_lock.
1011  */
1012 static int ipmr_cache_report(struct mr_table *mrt,
1013                              struct sk_buff *pkt, vifi_t vifi, int assert)
1014 {
1015         const int ihl = ip_hdrlen(pkt);
1016         struct sock *mroute_sk;
1017         struct igmphdr *igmp;
1018         struct igmpmsg *msg;
1019         struct sk_buff *skb;
1020         int ret;
1021
1022         if (assert == IGMPMSG_WHOLEPKT || assert == IGMPMSG_WRVIFWHOLE)
1023                 skb = skb_realloc_headroom(pkt, sizeof(struct iphdr));
1024         else
1025                 skb = alloc_skb(128, GFP_ATOMIC);
1026
1027         if (!skb)
1028                 return -ENOBUFS;
1029
1030         if (assert == IGMPMSG_WHOLEPKT || assert == IGMPMSG_WRVIFWHOLE) {
1031                 /* Ugly, but we have no choice with this interface.
1032                  * Duplicate old header, fix ihl, length etc.
1033                  * And all this only to mangle msg->im_msgtype and
1034                  * to set msg->im_mbz to "mbz" :-)
1035                  */
1036                 skb_push(skb, sizeof(struct iphdr));
1037                 skb_reset_network_header(skb);
1038                 skb_reset_transport_header(skb);
1039                 msg = (struct igmpmsg *)skb_network_header(skb);
1040                 memcpy(msg, skb_network_header(pkt), sizeof(struct iphdr));
1041                 msg->im_msgtype = assert;
1042                 msg->im_mbz = 0;
1043                 if (assert == IGMPMSG_WRVIFWHOLE) {
1044                         msg->im_vif = vifi;
1045                         msg->im_vif_hi = vifi >> 8;
1046                 } else {
1047                         msg->im_vif = mrt->mroute_reg_vif_num;
1048                         msg->im_vif_hi = mrt->mroute_reg_vif_num >> 8;
1049                 }
1050                 ip_hdr(skb)->ihl = sizeof(struct iphdr) >> 2;
1051                 ip_hdr(skb)->tot_len = htons(ntohs(ip_hdr(pkt)->tot_len) +
1052                                              sizeof(struct iphdr));
1053         } else {
1054                 /* Copy the IP header */
1055                 skb_set_network_header(skb, skb->len);
1056                 skb_put(skb, ihl);
1057                 skb_copy_to_linear_data(skb, pkt->data, ihl);
1058                 /* Flag to the kernel this is a route add */
1059                 ip_hdr(skb)->protocol = 0;
1060                 msg = (struct igmpmsg *)skb_network_header(skb);
1061                 msg->im_vif = vifi;
1062                 msg->im_vif_hi = vifi >> 8;
1063                 skb_dst_set(skb, dst_clone(skb_dst(pkt)));
1064                 /* Add our header */
1065                 igmp = skb_put(skb, sizeof(struct igmphdr));
1066                 igmp->type = assert;
1067                 msg->im_msgtype = assert;
1068                 igmp->code = 0;
1069                 ip_hdr(skb)->tot_len = htons(skb->len); /* Fix the length */
1070                 skb->transport_header = skb->network_header;
1071         }
1072
1073         rcu_read_lock();
1074         mroute_sk = rcu_dereference(mrt->mroute_sk);
1075         if (!mroute_sk) {
1076                 rcu_read_unlock();
1077                 kfree_skb(skb);
1078                 return -EINVAL;
1079         }
1080
1081         igmpmsg_netlink_event(mrt, skb);
1082
1083         /* Deliver to mrouted */
1084         ret = sock_queue_rcv_skb(mroute_sk, skb);
1085         rcu_read_unlock();
1086         if (ret < 0) {
1087                 net_warn_ratelimited("mroute: pending queue full, dropping entries\n");
1088                 kfree_skb(skb);
1089         }
1090
1091         return ret;
1092 }
1093
1094 /* Queue a packet for resolution. It gets locked cache entry! */
1095 static int ipmr_cache_unresolved(struct mr_table *mrt, vifi_t vifi,
1096                                  struct sk_buff *skb, struct net_device *dev)
1097 {
1098         const struct iphdr *iph = ip_hdr(skb);
1099         struct mfc_cache *c;
1100         bool found = false;
1101         int err;
1102
1103         spin_lock_bh(&mfc_unres_lock);
1104         list_for_each_entry(c, &mrt->mfc_unres_queue, _c.list) {
1105                 if (c->mfc_mcastgrp == iph->daddr &&
1106                     c->mfc_origin == iph->saddr) {
1107                         found = true;
1108                         break;
1109                 }
1110         }
1111
1112         if (!found) {
1113                 /* Create a new entry if allowable */
1114                 c = ipmr_cache_alloc_unres();
1115                 if (!c) {
1116                         spin_unlock_bh(&mfc_unres_lock);
1117
1118                         kfree_skb(skb);
1119                         return -ENOBUFS;
1120                 }
1121
1122                 /* Fill in the new cache entry */
1123                 c->_c.mfc_parent = -1;
1124                 c->mfc_origin   = iph->saddr;
1125                 c->mfc_mcastgrp = iph->daddr;
1126
1127                 /* Reflect first query at mrouted. */
1128                 err = ipmr_cache_report(mrt, skb, vifi, IGMPMSG_NOCACHE);
1129
1130                 if (err < 0) {
1131                         /* If the report failed throw the cache entry
1132                            out - Brad Parker
1133                          */
1134                         spin_unlock_bh(&mfc_unres_lock);
1135
1136                         ipmr_cache_free(c);
1137                         kfree_skb(skb);
1138                         return err;
1139                 }
1140
1141                 atomic_inc(&mrt->cache_resolve_queue_len);
1142                 list_add(&c->_c.list, &mrt->mfc_unres_queue);
1143                 mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1144
1145                 if (atomic_read(&mrt->cache_resolve_queue_len) == 1)
1146                         mod_timer(&mrt->ipmr_expire_timer,
1147                                   c->_c.mfc_un.unres.expires);
1148         }
1149
1150         /* See if we can append the packet */
1151         if (c->_c.mfc_un.unres.unresolved.qlen > 3) {
1152                 kfree_skb(skb);
1153                 err = -ENOBUFS;
1154         } else {
1155                 if (dev) {
1156                         skb->dev = dev;
1157                         skb->skb_iif = dev->ifindex;
1158                 }
1159                 skb_queue_tail(&c->_c.mfc_un.unres.unresolved, skb);
1160                 err = 0;
1161         }
1162
1163         spin_unlock_bh(&mfc_unres_lock);
1164         return err;
1165 }
1166
1167 /* MFC cache manipulation by user space mroute daemon */
1168
1169 static int ipmr_mfc_delete(struct mr_table *mrt, struct mfcctl *mfc, int parent)
1170 {
1171         struct net *net = read_pnet(&mrt->net);
1172         struct mfc_cache *c;
1173
1174         /* The entries are added/deleted only under RTNL */
1175         rcu_read_lock();
1176         c = ipmr_cache_find_parent(mrt, mfc->mfcc_origin.s_addr,
1177                                    mfc->mfcc_mcastgrp.s_addr, parent);
1178         rcu_read_unlock();
1179         if (!c)
1180                 return -ENOENT;
1181         rhltable_remove(&mrt->mfc_hash, &c->_c.mnode, ipmr_rht_params);
1182         list_del_rcu(&c->_c.list);
1183         call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_DEL, c, mrt->id);
1184         mroute_netlink_event(mrt, c, RTM_DELROUTE);
1185         mr_cache_put(&c->_c);
1186
1187         return 0;
1188 }
1189
1190 static int ipmr_mfc_add(struct net *net, struct mr_table *mrt,
1191                         struct mfcctl *mfc, int mrtsock, int parent)
1192 {
1193         struct mfc_cache *uc, *c;
1194         struct mr_mfc *_uc;
1195         bool found;
1196         int ret;
1197
1198         if (mfc->mfcc_parent >= MAXVIFS)
1199                 return -ENFILE;
1200
1201         /* The entries are added/deleted only under RTNL */
1202         rcu_read_lock();
1203         c = ipmr_cache_find_parent(mrt, mfc->mfcc_origin.s_addr,
1204                                    mfc->mfcc_mcastgrp.s_addr, parent);
1205         rcu_read_unlock();
1206         if (c) {
1207                 write_lock_bh(&mrt_lock);
1208                 c->_c.mfc_parent = mfc->mfcc_parent;
1209                 ipmr_update_thresholds(mrt, &c->_c, mfc->mfcc_ttls);
1210                 if (!mrtsock)
1211                         c->_c.mfc_flags |= MFC_STATIC;
1212                 write_unlock_bh(&mrt_lock);
1213                 call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_REPLACE, c,
1214                                               mrt->id);
1215                 mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1216                 return 0;
1217         }
1218
1219         if (mfc->mfcc_mcastgrp.s_addr != htonl(INADDR_ANY) &&
1220             !ipv4_is_multicast(mfc->mfcc_mcastgrp.s_addr))
1221                 return -EINVAL;
1222
1223         c = ipmr_cache_alloc();
1224         if (!c)
1225                 return -ENOMEM;
1226
1227         c->mfc_origin = mfc->mfcc_origin.s_addr;
1228         c->mfc_mcastgrp = mfc->mfcc_mcastgrp.s_addr;
1229         c->_c.mfc_parent = mfc->mfcc_parent;
1230         ipmr_update_thresholds(mrt, &c->_c, mfc->mfcc_ttls);
1231         if (!mrtsock)
1232                 c->_c.mfc_flags |= MFC_STATIC;
1233
1234         ret = rhltable_insert_key(&mrt->mfc_hash, &c->cmparg, &c->_c.mnode,
1235                                   ipmr_rht_params);
1236         if (ret) {
1237                 pr_err("ipmr: rhtable insert error %d\n", ret);
1238                 ipmr_cache_free(c);
1239                 return ret;
1240         }
1241         list_add_tail_rcu(&c->_c.list, &mrt->mfc_cache_list);
1242         /* Check to see if we resolved a queued list. If so we
1243          * need to send on the frames and tidy up.
1244          */
1245         found = false;
1246         spin_lock_bh(&mfc_unres_lock);
1247         list_for_each_entry(_uc, &mrt->mfc_unres_queue, list) {
1248                 uc = (struct mfc_cache *)_uc;
1249                 if (uc->mfc_origin == c->mfc_origin &&
1250                     uc->mfc_mcastgrp == c->mfc_mcastgrp) {
1251                         list_del(&_uc->list);
1252                         atomic_dec(&mrt->cache_resolve_queue_len);
1253                         found = true;
1254                         break;
1255                 }
1256         }
1257         if (list_empty(&mrt->mfc_unres_queue))
1258                 del_timer(&mrt->ipmr_expire_timer);
1259         spin_unlock_bh(&mfc_unres_lock);
1260
1261         if (found) {
1262                 ipmr_cache_resolve(net, mrt, uc, c);
1263                 ipmr_cache_free(uc);
1264         }
1265         call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_ADD, c, mrt->id);
1266         mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1267         return 0;
1268 }
1269
1270 /* Close the multicast socket, and clear the vif tables etc */
1271 static void mroute_clean_tables(struct mr_table *mrt, int flags)
1272 {
1273         struct net *net = read_pnet(&mrt->net);
1274         struct mr_mfc *c, *tmp;
1275         struct mfc_cache *cache;
1276         LIST_HEAD(list);
1277         int i;
1278
1279         /* Shut down all active vif entries */
1280         if (flags & (MRT_FLUSH_VIFS | MRT_FLUSH_VIFS_STATIC)) {
1281                 for (i = 0; i < mrt->maxvif; i++) {
1282                         if (((mrt->vif_table[i].flags & VIFF_STATIC) &&
1283                              !(flags & MRT_FLUSH_VIFS_STATIC)) ||
1284                             (!(mrt->vif_table[i].flags & VIFF_STATIC) && !(flags & MRT_FLUSH_VIFS)))
1285                                 continue;
1286                         vif_delete(mrt, i, 0, &list);
1287                 }
1288                 unregister_netdevice_many(&list);
1289         }
1290
1291         /* Wipe the cache */
1292         if (flags & (MRT_FLUSH_MFC | MRT_FLUSH_MFC_STATIC)) {
1293                 list_for_each_entry_safe(c, tmp, &mrt->mfc_cache_list, list) {
1294                         if (((c->mfc_flags & MFC_STATIC) && !(flags & MRT_FLUSH_MFC_STATIC)) ||
1295                             (!(c->mfc_flags & MFC_STATIC) && !(flags & MRT_FLUSH_MFC)))
1296                                 continue;
1297                         rhltable_remove(&mrt->mfc_hash, &c->mnode, ipmr_rht_params);
1298                         list_del_rcu(&c->list);
1299                         cache = (struct mfc_cache *)c;
1300                         call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_DEL, cache,
1301                                                       mrt->id);
1302                         mroute_netlink_event(mrt, cache, RTM_DELROUTE);
1303                         mr_cache_put(c);
1304                 }
1305         }
1306
1307         if (flags & MRT_FLUSH_MFC) {
1308                 if (atomic_read(&mrt->cache_resolve_queue_len) != 0) {
1309                         spin_lock_bh(&mfc_unres_lock);
1310                         list_for_each_entry_safe(c, tmp, &mrt->mfc_unres_queue, list) {
1311                                 list_del(&c->list);
1312                                 cache = (struct mfc_cache *)c;
1313                                 mroute_netlink_event(mrt, cache, RTM_DELROUTE);
1314                                 ipmr_destroy_unres(mrt, cache);
1315                         }
1316                         spin_unlock_bh(&mfc_unres_lock);
1317                 }
1318         }
1319 }
1320
1321 /* called from ip_ra_control(), before an RCU grace period,
1322  * we dont need to call synchronize_rcu() here
1323  */
1324 static void mrtsock_destruct(struct sock *sk)
1325 {
1326         struct net *net = sock_net(sk);
1327         struct mr_table *mrt;
1328
1329         rtnl_lock();
1330         ipmr_for_each_table(mrt, net) {
1331                 if (sk == rtnl_dereference(mrt->mroute_sk)) {
1332                         IPV4_DEVCONF_ALL(net, MC_FORWARDING)--;
1333                         inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
1334                                                     NETCONFA_MC_FORWARDING,
1335                                                     NETCONFA_IFINDEX_ALL,
1336                                                     net->ipv4.devconf_all);
1337                         RCU_INIT_POINTER(mrt->mroute_sk, NULL);
1338                         mroute_clean_tables(mrt, MRT_FLUSH_VIFS | MRT_FLUSH_MFC);
1339                 }
1340         }
1341         rtnl_unlock();
1342 }
1343
1344 /* Socket options and virtual interface manipulation. The whole
1345  * virtual interface system is a complete heap, but unfortunately
1346  * that's how BSD mrouted happens to think. Maybe one day with a proper
1347  * MOSPF/PIM router set up we can clean this up.
1348  */
1349
1350 int ip_mroute_setsockopt(struct sock *sk, int optname, sockptr_t optval,
1351                          unsigned int optlen)
1352 {
1353         struct net *net = sock_net(sk);
1354         int val, ret = 0, parent = 0;
1355         struct mr_table *mrt;
1356         struct vifctl vif;
1357         struct mfcctl mfc;
1358         bool do_wrvifwhole;
1359         u32 uval;
1360
1361         /* There's one exception to the lock - MRT_DONE which needs to unlock */
1362         rtnl_lock();
1363         if (sk->sk_type != SOCK_RAW ||
1364             inet_sk(sk)->inet_num != IPPROTO_IGMP) {
1365                 ret = -EOPNOTSUPP;
1366                 goto out_unlock;
1367         }
1368
1369         mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1370         if (!mrt) {
1371                 ret = -ENOENT;
1372                 goto out_unlock;
1373         }
1374         if (optname != MRT_INIT) {
1375                 if (sk != rcu_access_pointer(mrt->mroute_sk) &&
1376                     !ns_capable(net->user_ns, CAP_NET_ADMIN)) {
1377                         ret = -EACCES;
1378                         goto out_unlock;
1379                 }
1380         }
1381
1382         switch (optname) {
1383         case MRT_INIT:
1384                 if (optlen != sizeof(int)) {
1385                         ret = -EINVAL;
1386                         break;
1387                 }
1388                 if (rtnl_dereference(mrt->mroute_sk)) {
1389                         ret = -EADDRINUSE;
1390                         break;
1391                 }
1392
1393                 ret = ip_ra_control(sk, 1, mrtsock_destruct);
1394                 if (ret == 0) {
1395                         rcu_assign_pointer(mrt->mroute_sk, sk);
1396                         IPV4_DEVCONF_ALL(net, MC_FORWARDING)++;
1397                         inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
1398                                                     NETCONFA_MC_FORWARDING,
1399                                                     NETCONFA_IFINDEX_ALL,
1400                                                     net->ipv4.devconf_all);
1401                 }
1402                 break;
1403         case MRT_DONE:
1404                 if (sk != rcu_access_pointer(mrt->mroute_sk)) {
1405                         ret = -EACCES;
1406                 } else {
1407                         /* We need to unlock here because mrtsock_destruct takes
1408                          * care of rtnl itself and we can't change that due to
1409                          * the IP_ROUTER_ALERT setsockopt which runs without it.
1410                          */
1411                         rtnl_unlock();
1412                         ret = ip_ra_control(sk, 0, NULL);
1413                         goto out;
1414                 }
1415                 break;
1416         case MRT_ADD_VIF:
1417         case MRT_DEL_VIF:
1418                 if (optlen != sizeof(vif)) {
1419                         ret = -EINVAL;
1420                         break;
1421                 }
1422                 if (copy_from_sockptr(&vif, optval, sizeof(vif))) {
1423                         ret = -EFAULT;
1424                         break;
1425                 }
1426                 if (vif.vifc_vifi >= MAXVIFS) {
1427                         ret = -ENFILE;
1428                         break;
1429                 }
1430                 if (optname == MRT_ADD_VIF) {
1431                         ret = vif_add(net, mrt, &vif,
1432                                       sk == rtnl_dereference(mrt->mroute_sk));
1433                 } else {
1434                         ret = vif_delete(mrt, vif.vifc_vifi, 0, NULL);
1435                 }
1436                 break;
1437         /* Manipulate the forwarding caches. These live
1438          * in a sort of kernel/user symbiosis.
1439          */
1440         case MRT_ADD_MFC:
1441         case MRT_DEL_MFC:
1442                 parent = -1;
1443                 fallthrough;
1444         case MRT_ADD_MFC_PROXY:
1445         case MRT_DEL_MFC_PROXY:
1446                 if (optlen != sizeof(mfc)) {
1447                         ret = -EINVAL;
1448                         break;
1449                 }
1450                 if (copy_from_sockptr(&mfc, optval, sizeof(mfc))) {
1451                         ret = -EFAULT;
1452                         break;
1453                 }
1454                 if (parent == 0)
1455                         parent = mfc.mfcc_parent;
1456                 if (optname == MRT_DEL_MFC || optname == MRT_DEL_MFC_PROXY)
1457                         ret = ipmr_mfc_delete(mrt, &mfc, parent);
1458                 else
1459                         ret = ipmr_mfc_add(net, mrt, &mfc,
1460                                            sk == rtnl_dereference(mrt->mroute_sk),
1461                                            parent);
1462                 break;
1463         case MRT_FLUSH:
1464                 if (optlen != sizeof(val)) {
1465                         ret = -EINVAL;
1466                         break;
1467                 }
1468                 if (copy_from_sockptr(&val, optval, sizeof(val))) {
1469                         ret = -EFAULT;
1470                         break;
1471                 }
1472                 mroute_clean_tables(mrt, val);
1473                 break;
1474         /* Control PIM assert. */
1475         case MRT_ASSERT:
1476                 if (optlen != sizeof(val)) {
1477                         ret = -EINVAL;
1478                         break;
1479                 }
1480                 if (copy_from_sockptr(&val, optval, sizeof(val))) {
1481                         ret = -EFAULT;
1482                         break;
1483                 }
1484                 mrt->mroute_do_assert = val;
1485                 break;
1486         case MRT_PIM:
1487                 if (!ipmr_pimsm_enabled()) {
1488                         ret = -ENOPROTOOPT;
1489                         break;
1490                 }
1491                 if (optlen != sizeof(val)) {
1492                         ret = -EINVAL;
1493                         break;
1494                 }
1495                 if (copy_from_sockptr(&val, optval, sizeof(val))) {
1496                         ret = -EFAULT;
1497                         break;
1498                 }
1499
1500                 do_wrvifwhole = (val == IGMPMSG_WRVIFWHOLE);
1501                 val = !!val;
1502                 if (val != mrt->mroute_do_pim) {
1503                         mrt->mroute_do_pim = val;
1504                         mrt->mroute_do_assert = val;
1505                         mrt->mroute_do_wrvifwhole = do_wrvifwhole;
1506                 }
1507                 break;
1508         case MRT_TABLE:
1509                 if (!IS_BUILTIN(CONFIG_IP_MROUTE_MULTIPLE_TABLES)) {
1510                         ret = -ENOPROTOOPT;
1511                         break;
1512                 }
1513                 if (optlen != sizeof(uval)) {
1514                         ret = -EINVAL;
1515                         break;
1516                 }
1517                 if (copy_from_sockptr(&uval, optval, sizeof(uval))) {
1518                         ret = -EFAULT;
1519                         break;
1520                 }
1521
1522                 if (sk == rtnl_dereference(mrt->mroute_sk)) {
1523                         ret = -EBUSY;
1524                 } else {
1525                         mrt = ipmr_new_table(net, uval);
1526                         if (IS_ERR(mrt))
1527                                 ret = PTR_ERR(mrt);
1528                         else
1529                                 raw_sk(sk)->ipmr_table = uval;
1530                 }
1531                 break;
1532         /* Spurious command, or MRT_VERSION which you cannot set. */
1533         default:
1534                 ret = -ENOPROTOOPT;
1535         }
1536 out_unlock:
1537         rtnl_unlock();
1538 out:
1539         return ret;
1540 }
1541
1542 /* Getsock opt support for the multicast routing system. */
1543 int ip_mroute_getsockopt(struct sock *sk, int optname, char __user *optval, int __user *optlen)
1544 {
1545         int olr;
1546         int val;
1547         struct net *net = sock_net(sk);
1548         struct mr_table *mrt;
1549
1550         if (sk->sk_type != SOCK_RAW ||
1551             inet_sk(sk)->inet_num != IPPROTO_IGMP)
1552                 return -EOPNOTSUPP;
1553
1554         mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1555         if (!mrt)
1556                 return -ENOENT;
1557
1558         switch (optname) {
1559         case MRT_VERSION:
1560                 val = 0x0305;
1561                 break;
1562         case MRT_PIM:
1563                 if (!ipmr_pimsm_enabled())
1564                         return -ENOPROTOOPT;
1565                 val = mrt->mroute_do_pim;
1566                 break;
1567         case MRT_ASSERT:
1568                 val = mrt->mroute_do_assert;
1569                 break;
1570         default:
1571                 return -ENOPROTOOPT;
1572         }
1573
1574         if (get_user(olr, optlen))
1575                 return -EFAULT;
1576         olr = min_t(unsigned int, olr, sizeof(int));
1577         if (olr < 0)
1578                 return -EINVAL;
1579         if (put_user(olr, optlen))
1580                 return -EFAULT;
1581         if (copy_to_user(optval, &val, olr))
1582                 return -EFAULT;
1583         return 0;
1584 }
1585
1586 /* The IP multicast ioctl support routines. */
1587 int ipmr_ioctl(struct sock *sk, int cmd, void __user *arg)
1588 {
1589         struct sioc_sg_req sr;
1590         struct sioc_vif_req vr;
1591         struct vif_device *vif;
1592         struct mfc_cache *c;
1593         struct net *net = sock_net(sk);
1594         struct mr_table *mrt;
1595
1596         mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1597         if (!mrt)
1598                 return -ENOENT;
1599
1600         switch (cmd) {
1601         case SIOCGETVIFCNT:
1602                 if (copy_from_user(&vr, arg, sizeof(vr)))
1603                         return -EFAULT;
1604                 if (vr.vifi >= mrt->maxvif)
1605                         return -EINVAL;
1606                 vr.vifi = array_index_nospec(vr.vifi, mrt->maxvif);
1607                 read_lock(&mrt_lock);
1608                 vif = &mrt->vif_table[vr.vifi];
1609                 if (VIF_EXISTS(mrt, vr.vifi)) {
1610                         vr.icount = vif->pkt_in;
1611                         vr.ocount = vif->pkt_out;
1612                         vr.ibytes = vif->bytes_in;
1613                         vr.obytes = vif->bytes_out;
1614                         read_unlock(&mrt_lock);
1615
1616                         if (copy_to_user(arg, &vr, sizeof(vr)))
1617                                 return -EFAULT;
1618                         return 0;
1619                 }
1620                 read_unlock(&mrt_lock);
1621                 return -EADDRNOTAVAIL;
1622         case SIOCGETSGCNT:
1623                 if (copy_from_user(&sr, arg, sizeof(sr)))
1624                         return -EFAULT;
1625
1626                 rcu_read_lock();
1627                 c = ipmr_cache_find(mrt, sr.src.s_addr, sr.grp.s_addr);
1628                 if (c) {
1629                         sr.pktcnt = c->_c.mfc_un.res.pkt;
1630                         sr.bytecnt = c->_c.mfc_un.res.bytes;
1631                         sr.wrong_if = c->_c.mfc_un.res.wrong_if;
1632                         rcu_read_unlock();
1633
1634                         if (copy_to_user(arg, &sr, sizeof(sr)))
1635                                 return -EFAULT;
1636                         return 0;
1637                 }
1638                 rcu_read_unlock();
1639                 return -EADDRNOTAVAIL;
1640         default:
1641                 return -ENOIOCTLCMD;
1642         }
1643 }
1644
1645 #ifdef CONFIG_COMPAT
1646 struct compat_sioc_sg_req {
1647         struct in_addr src;
1648         struct in_addr grp;
1649         compat_ulong_t pktcnt;
1650         compat_ulong_t bytecnt;
1651         compat_ulong_t wrong_if;
1652 };
1653
1654 struct compat_sioc_vif_req {
1655         vifi_t  vifi;           /* Which iface */
1656         compat_ulong_t icount;
1657         compat_ulong_t ocount;
1658         compat_ulong_t ibytes;
1659         compat_ulong_t obytes;
1660 };
1661
1662 int ipmr_compat_ioctl(struct sock *sk, unsigned int cmd, void __user *arg)
1663 {
1664         struct compat_sioc_sg_req sr;
1665         struct compat_sioc_vif_req vr;
1666         struct vif_device *vif;
1667         struct mfc_cache *c;
1668         struct net *net = sock_net(sk);
1669         struct mr_table *mrt;
1670
1671         mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1672         if (!mrt)
1673                 return -ENOENT;
1674
1675         switch (cmd) {
1676         case SIOCGETVIFCNT:
1677                 if (copy_from_user(&vr, arg, sizeof(vr)))
1678                         return -EFAULT;
1679                 if (vr.vifi >= mrt->maxvif)
1680                         return -EINVAL;
1681                 vr.vifi = array_index_nospec(vr.vifi, mrt->maxvif);
1682                 read_lock(&mrt_lock);
1683                 vif = &mrt->vif_table[vr.vifi];
1684                 if (VIF_EXISTS(mrt, vr.vifi)) {
1685                         vr.icount = vif->pkt_in;
1686                         vr.ocount = vif->pkt_out;
1687                         vr.ibytes = vif->bytes_in;
1688                         vr.obytes = vif->bytes_out;
1689                         read_unlock(&mrt_lock);
1690
1691                         if (copy_to_user(arg, &vr, sizeof(vr)))
1692                                 return -EFAULT;
1693                         return 0;
1694                 }
1695                 read_unlock(&mrt_lock);
1696                 return -EADDRNOTAVAIL;
1697         case SIOCGETSGCNT:
1698                 if (copy_from_user(&sr, arg, sizeof(sr)))
1699                         return -EFAULT;
1700
1701                 rcu_read_lock();
1702                 c = ipmr_cache_find(mrt, sr.src.s_addr, sr.grp.s_addr);
1703                 if (c) {
1704                         sr.pktcnt = c->_c.mfc_un.res.pkt;
1705                         sr.bytecnt = c->_c.mfc_un.res.bytes;
1706                         sr.wrong_if = c->_c.mfc_un.res.wrong_if;
1707                         rcu_read_unlock();
1708
1709                         if (copy_to_user(arg, &sr, sizeof(sr)))
1710                                 return -EFAULT;
1711                         return 0;
1712                 }
1713                 rcu_read_unlock();
1714                 return -EADDRNOTAVAIL;
1715         default:
1716                 return -ENOIOCTLCMD;
1717         }
1718 }
1719 #endif
1720
1721 static int ipmr_device_event(struct notifier_block *this, unsigned long event, void *ptr)
1722 {
1723         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1724         struct net *net = dev_net(dev);
1725         struct mr_table *mrt;
1726         struct vif_device *v;
1727         int ct;
1728
1729         if (event != NETDEV_UNREGISTER)
1730                 return NOTIFY_DONE;
1731
1732         ipmr_for_each_table(mrt, net) {
1733                 v = &mrt->vif_table[0];
1734                 for (ct = 0; ct < mrt->maxvif; ct++, v++) {
1735                         if (v->dev == dev)
1736                                 vif_delete(mrt, ct, 1, NULL);
1737                 }
1738         }
1739         return NOTIFY_DONE;
1740 }
1741
1742 static struct notifier_block ip_mr_notifier = {
1743         .notifier_call = ipmr_device_event,
1744 };
1745
1746 /* Encapsulate a packet by attaching a valid IPIP header to it.
1747  * This avoids tunnel drivers and other mess and gives us the speed so
1748  * important for multicast video.
1749  */
1750 static void ip_encap(struct net *net, struct sk_buff *skb,
1751                      __be32 saddr, __be32 daddr)
1752 {
1753         struct iphdr *iph;
1754         const struct iphdr *old_iph = ip_hdr(skb);
1755
1756         skb_push(skb, sizeof(struct iphdr));
1757         skb->transport_header = skb->network_header;
1758         skb_reset_network_header(skb);
1759         iph = ip_hdr(skb);
1760
1761         iph->version    =       4;
1762         iph->tos        =       old_iph->tos;
1763         iph->ttl        =       old_iph->ttl;
1764         iph->frag_off   =       0;
1765         iph->daddr      =       daddr;
1766         iph->saddr      =       saddr;
1767         iph->protocol   =       IPPROTO_IPIP;
1768         iph->ihl        =       5;
1769         iph->tot_len    =       htons(skb->len);
1770         ip_select_ident(net, skb, NULL);
1771         ip_send_check(iph);
1772
1773         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1774         nf_reset_ct(skb);
1775 }
1776
1777 static inline int ipmr_forward_finish(struct net *net, struct sock *sk,
1778                                       struct sk_buff *skb)
1779 {
1780         struct ip_options *opt = &(IPCB(skb)->opt);
1781
1782         IP_INC_STATS(net, IPSTATS_MIB_OUTFORWDATAGRAMS);
1783         IP_ADD_STATS(net, IPSTATS_MIB_OUTOCTETS, skb->len);
1784
1785         if (unlikely(opt->optlen))
1786                 ip_forward_options(skb);
1787
1788         return dst_output(net, sk, skb);
1789 }
1790
1791 #ifdef CONFIG_NET_SWITCHDEV
1792 static bool ipmr_forward_offloaded(struct sk_buff *skb, struct mr_table *mrt,
1793                                    int in_vifi, int out_vifi)
1794 {
1795         struct vif_device *out_vif = &mrt->vif_table[out_vifi];
1796         struct vif_device *in_vif = &mrt->vif_table[in_vifi];
1797
1798         if (!skb->offload_l3_fwd_mark)
1799                 return false;
1800         if (!out_vif->dev_parent_id.id_len || !in_vif->dev_parent_id.id_len)
1801                 return false;
1802         return netdev_phys_item_id_same(&out_vif->dev_parent_id,
1803                                         &in_vif->dev_parent_id);
1804 }
1805 #else
1806 static bool ipmr_forward_offloaded(struct sk_buff *skb, struct mr_table *mrt,
1807                                    int in_vifi, int out_vifi)
1808 {
1809         return false;
1810 }
1811 #endif
1812
1813 /* Processing handlers for ipmr_forward */
1814
1815 static void ipmr_queue_xmit(struct net *net, struct mr_table *mrt,
1816                             int in_vifi, struct sk_buff *skb, int vifi)
1817 {
1818         const struct iphdr *iph = ip_hdr(skb);
1819         struct vif_device *vif = &mrt->vif_table[vifi];
1820         struct net_device *dev;
1821         struct rtable *rt;
1822         struct flowi4 fl4;
1823         int    encap = 0;
1824
1825         if (!vif->dev)
1826                 goto out_free;
1827
1828         if (vif->flags & VIFF_REGISTER) {
1829                 vif->pkt_out++;
1830                 vif->bytes_out += skb->len;
1831                 vif->dev->stats.tx_bytes += skb->len;
1832                 vif->dev->stats.tx_packets++;
1833                 ipmr_cache_report(mrt, skb, vifi, IGMPMSG_WHOLEPKT);
1834                 goto out_free;
1835         }
1836
1837         if (ipmr_forward_offloaded(skb, mrt, in_vifi, vifi))
1838                 goto out_free;
1839
1840         if (vif->flags & VIFF_TUNNEL) {
1841                 rt = ip_route_output_ports(net, &fl4, NULL,
1842                                            vif->remote, vif->local,
1843                                            0, 0,
1844                                            IPPROTO_IPIP,
1845                                            RT_TOS(iph->tos), vif->link);
1846                 if (IS_ERR(rt))
1847                         goto out_free;
1848                 encap = sizeof(struct iphdr);
1849         } else {
1850                 rt = ip_route_output_ports(net, &fl4, NULL, iph->daddr, 0,
1851                                            0, 0,
1852                                            IPPROTO_IPIP,
1853                                            RT_TOS(iph->tos), vif->link);
1854                 if (IS_ERR(rt))
1855                         goto out_free;
1856         }
1857
1858         dev = rt->dst.dev;
1859
1860         if (skb->len+encap > dst_mtu(&rt->dst) && (ntohs(iph->frag_off) & IP_DF)) {
1861                 /* Do not fragment multicasts. Alas, IPv4 does not
1862                  * allow to send ICMP, so that packets will disappear
1863                  * to blackhole.
1864                  */
1865                 IP_INC_STATS(net, IPSTATS_MIB_FRAGFAILS);
1866                 ip_rt_put(rt);
1867                 goto out_free;
1868         }
1869
1870         encap += LL_RESERVED_SPACE(dev) + rt->dst.header_len;
1871
1872         if (skb_cow(skb, encap)) {
1873                 ip_rt_put(rt);
1874                 goto out_free;
1875         }
1876
1877         vif->pkt_out++;
1878         vif->bytes_out += skb->len;
1879
1880         skb_dst_drop(skb);
1881         skb_dst_set(skb, &rt->dst);
1882         ip_decrease_ttl(ip_hdr(skb));
1883
1884         /* FIXME: forward and output firewalls used to be called here.
1885          * What do we do with netfilter? -- RR
1886          */
1887         if (vif->flags & VIFF_TUNNEL) {
1888                 ip_encap(net, skb, vif->local, vif->remote);
1889                 /* FIXME: extra output firewall step used to be here. --RR */
1890                 vif->dev->stats.tx_packets++;
1891                 vif->dev->stats.tx_bytes += skb->len;
1892         }
1893
1894         IPCB(skb)->flags |= IPSKB_FORWARDED;
1895
1896         /* RFC1584 teaches, that DVMRP/PIM router must deliver packets locally
1897          * not only before forwarding, but after forwarding on all output
1898          * interfaces. It is clear, if mrouter runs a multicasting
1899          * program, it should receive packets not depending to what interface
1900          * program is joined.
1901          * If we will not make it, the program will have to join on all
1902          * interfaces. On the other hand, multihoming host (or router, but
1903          * not mrouter) cannot join to more than one interface - it will
1904          * result in receiving multiple packets.
1905          */
1906         NF_HOOK(NFPROTO_IPV4, NF_INET_FORWARD,
1907                 net, NULL, skb, skb->dev, dev,
1908                 ipmr_forward_finish);
1909         return;
1910
1911 out_free:
1912         kfree_skb(skb);
1913 }
1914
1915 static int ipmr_find_vif(struct mr_table *mrt, struct net_device *dev)
1916 {
1917         int ct;
1918
1919         for (ct = mrt->maxvif-1; ct >= 0; ct--) {
1920                 if (mrt->vif_table[ct].dev == dev)
1921                         break;
1922         }
1923         return ct;
1924 }
1925
1926 /* "local" means that we should preserve one skb (for local delivery) */
1927 static void ip_mr_forward(struct net *net, struct mr_table *mrt,
1928                           struct net_device *dev, struct sk_buff *skb,
1929                           struct mfc_cache *c, int local)
1930 {
1931         int true_vifi = ipmr_find_vif(mrt, dev);
1932         int psend = -1;
1933         int vif, ct;
1934
1935         vif = c->_c.mfc_parent;
1936         c->_c.mfc_un.res.pkt++;
1937         c->_c.mfc_un.res.bytes += skb->len;
1938         c->_c.mfc_un.res.lastuse = jiffies;
1939
1940         if (c->mfc_origin == htonl(INADDR_ANY) && true_vifi >= 0) {
1941                 struct mfc_cache *cache_proxy;
1942
1943                 /* For an (*,G) entry, we only check that the incomming
1944                  * interface is part of the static tree.
1945                  */
1946                 cache_proxy = mr_mfc_find_any_parent(mrt, vif);
1947                 if (cache_proxy &&
1948                     cache_proxy->_c.mfc_un.res.ttls[true_vifi] < 255)
1949                         goto forward;
1950         }
1951
1952         /* Wrong interface: drop packet and (maybe) send PIM assert. */
1953         if (mrt->vif_table[vif].dev != dev) {
1954                 if (rt_is_output_route(skb_rtable(skb))) {
1955                         /* It is our own packet, looped back.
1956                          * Very complicated situation...
1957                          *
1958                          * The best workaround until routing daemons will be
1959                          * fixed is not to redistribute packet, if it was
1960                          * send through wrong interface. It means, that
1961                          * multicast applications WILL NOT work for
1962                          * (S,G), which have default multicast route pointing
1963                          * to wrong oif. In any case, it is not a good
1964                          * idea to use multicasting applications on router.
1965                          */
1966                         goto dont_forward;
1967                 }
1968
1969                 c->_c.mfc_un.res.wrong_if++;
1970
1971                 if (true_vifi >= 0 && mrt->mroute_do_assert &&
1972                     /* pimsm uses asserts, when switching from RPT to SPT,
1973                      * so that we cannot check that packet arrived on an oif.
1974                      * It is bad, but otherwise we would need to move pretty
1975                      * large chunk of pimd to kernel. Ough... --ANK
1976                      */
1977                     (mrt->mroute_do_pim ||
1978                      c->_c.mfc_un.res.ttls[true_vifi] < 255) &&
1979                     time_after(jiffies,
1980                                c->_c.mfc_un.res.last_assert +
1981                                MFC_ASSERT_THRESH)) {
1982                         c->_c.mfc_un.res.last_assert = jiffies;
1983                         ipmr_cache_report(mrt, skb, true_vifi, IGMPMSG_WRONGVIF);
1984                         if (mrt->mroute_do_wrvifwhole)
1985                                 ipmr_cache_report(mrt, skb, true_vifi,
1986                                                   IGMPMSG_WRVIFWHOLE);
1987                 }
1988                 goto dont_forward;
1989         }
1990
1991 forward:
1992         mrt->vif_table[vif].pkt_in++;
1993         mrt->vif_table[vif].bytes_in += skb->len;
1994
1995         /* Forward the frame */
1996         if (c->mfc_origin == htonl(INADDR_ANY) &&
1997             c->mfc_mcastgrp == htonl(INADDR_ANY)) {
1998                 if (true_vifi >= 0 &&
1999                     true_vifi != c->_c.mfc_parent &&
2000                     ip_hdr(skb)->ttl >
2001                                 c->_c.mfc_un.res.ttls[c->_c.mfc_parent]) {
2002                         /* It's an (*,*) entry and the packet is not coming from
2003                          * the upstream: forward the packet to the upstream
2004                          * only.
2005                          */
2006                         psend = c->_c.mfc_parent;
2007                         goto last_forward;
2008                 }
2009                 goto dont_forward;
2010         }
2011         for (ct = c->_c.mfc_un.res.maxvif - 1;
2012              ct >= c->_c.mfc_un.res.minvif; ct--) {
2013                 /* For (*,G) entry, don't forward to the incoming interface */
2014                 if ((c->mfc_origin != htonl(INADDR_ANY) ||
2015                      ct != true_vifi) &&
2016                     ip_hdr(skb)->ttl > c->_c.mfc_un.res.ttls[ct]) {
2017                         if (psend != -1) {
2018                                 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2019
2020                                 if (skb2)
2021                                         ipmr_queue_xmit(net, mrt, true_vifi,
2022                                                         skb2, psend);
2023                         }
2024                         psend = ct;
2025                 }
2026         }
2027 last_forward:
2028         if (psend != -1) {
2029                 if (local) {
2030                         struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2031
2032                         if (skb2)
2033                                 ipmr_queue_xmit(net, mrt, true_vifi, skb2,
2034                                                 psend);
2035                 } else {
2036                         ipmr_queue_xmit(net, mrt, true_vifi, skb, psend);
2037                         return;
2038                 }
2039         }
2040
2041 dont_forward:
2042         if (!local)
2043                 kfree_skb(skb);
2044 }
2045
2046 static struct mr_table *ipmr_rt_fib_lookup(struct net *net, struct sk_buff *skb)
2047 {
2048         struct rtable *rt = skb_rtable(skb);
2049         struct iphdr *iph = ip_hdr(skb);
2050         struct flowi4 fl4 = {
2051                 .daddr = iph->daddr,
2052                 .saddr = iph->saddr,
2053                 .flowi4_tos = RT_TOS(iph->tos),
2054                 .flowi4_oif = (rt_is_output_route(rt) ?
2055                                skb->dev->ifindex : 0),
2056                 .flowi4_iif = (rt_is_output_route(rt) ?
2057                                LOOPBACK_IFINDEX :
2058                                skb->dev->ifindex),
2059                 .flowi4_mark = skb->mark,
2060         };
2061         struct mr_table *mrt;
2062         int err;
2063
2064         err = ipmr_fib_lookup(net, &fl4, &mrt);
2065         if (err)
2066                 return ERR_PTR(err);
2067         return mrt;
2068 }
2069
2070 /* Multicast packets for forwarding arrive here
2071  * Called with rcu_read_lock();
2072  */
2073 int ip_mr_input(struct sk_buff *skb)
2074 {
2075         struct mfc_cache *cache;
2076         struct net *net = dev_net(skb->dev);
2077         int local = skb_rtable(skb)->rt_flags & RTCF_LOCAL;
2078         struct mr_table *mrt;
2079         struct net_device *dev;
2080
2081         /* skb->dev passed in is the loX master dev for vrfs.
2082          * As there are no vifs associated with loopback devices,
2083          * get the proper interface that does have a vif associated with it.
2084          */
2085         dev = skb->dev;
2086         if (netif_is_l3_master(skb->dev)) {
2087                 dev = dev_get_by_index_rcu(net, IPCB(skb)->iif);
2088                 if (!dev) {
2089                         kfree_skb(skb);
2090                         return -ENODEV;
2091                 }
2092         }
2093
2094         /* Packet is looped back after forward, it should not be
2095          * forwarded second time, but still can be delivered locally.
2096          */
2097         if (IPCB(skb)->flags & IPSKB_FORWARDED)
2098                 goto dont_forward;
2099
2100         mrt = ipmr_rt_fib_lookup(net, skb);
2101         if (IS_ERR(mrt)) {
2102                 kfree_skb(skb);
2103                 return PTR_ERR(mrt);
2104         }
2105         if (!local) {
2106                 if (IPCB(skb)->opt.router_alert) {
2107                         if (ip_call_ra_chain(skb))
2108                                 return 0;
2109                 } else if (ip_hdr(skb)->protocol == IPPROTO_IGMP) {
2110                         /* IGMPv1 (and broken IGMPv2 implementations sort of
2111                          * Cisco IOS <= 11.2(8)) do not put router alert
2112                          * option to IGMP packets destined to routable
2113                          * groups. It is very bad, because it means
2114                          * that we can forward NO IGMP messages.
2115                          */
2116                         struct sock *mroute_sk;
2117
2118                         mroute_sk = rcu_dereference(mrt->mroute_sk);
2119                         if (mroute_sk) {
2120                                 nf_reset_ct(skb);
2121                                 raw_rcv(mroute_sk, skb);
2122                                 return 0;
2123                         }
2124                     }
2125         }
2126
2127         /* already under rcu_read_lock() */
2128         cache = ipmr_cache_find(mrt, ip_hdr(skb)->saddr, ip_hdr(skb)->daddr);
2129         if (!cache) {
2130                 int vif = ipmr_find_vif(mrt, dev);
2131
2132                 if (vif >= 0)
2133                         cache = ipmr_cache_find_any(mrt, ip_hdr(skb)->daddr,
2134                                                     vif);
2135         }
2136
2137         /* No usable cache entry */
2138         if (!cache) {
2139                 int vif;
2140
2141                 if (local) {
2142                         struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2143                         ip_local_deliver(skb);
2144                         if (!skb2)
2145                                 return -ENOBUFS;
2146                         skb = skb2;
2147                 }
2148
2149                 read_lock(&mrt_lock);
2150                 vif = ipmr_find_vif(mrt, dev);
2151                 if (vif >= 0) {
2152                         int err2 = ipmr_cache_unresolved(mrt, vif, skb, dev);
2153                         read_unlock(&mrt_lock);
2154
2155                         return err2;
2156                 }
2157                 read_unlock(&mrt_lock);
2158                 kfree_skb(skb);
2159                 return -ENODEV;
2160         }
2161
2162         read_lock(&mrt_lock);
2163         ip_mr_forward(net, mrt, dev, skb, cache, local);
2164         read_unlock(&mrt_lock);
2165
2166         if (local)
2167                 return ip_local_deliver(skb);
2168
2169         return 0;
2170
2171 dont_forward:
2172         if (local)
2173                 return ip_local_deliver(skb);
2174         kfree_skb(skb);
2175         return 0;
2176 }
2177
2178 #ifdef CONFIG_IP_PIMSM_V1
2179 /* Handle IGMP messages of PIMv1 */
2180 int pim_rcv_v1(struct sk_buff *skb)
2181 {
2182         struct igmphdr *pim;
2183         struct net *net = dev_net(skb->dev);
2184         struct mr_table *mrt;
2185
2186         if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr)))
2187                 goto drop;
2188
2189         pim = igmp_hdr(skb);
2190
2191         mrt = ipmr_rt_fib_lookup(net, skb);
2192         if (IS_ERR(mrt))
2193                 goto drop;
2194         if (!mrt->mroute_do_pim ||
2195             pim->group != PIM_V1_VERSION || pim->code != PIM_V1_REGISTER)
2196                 goto drop;
2197
2198         if (__pim_rcv(mrt, skb, sizeof(*pim))) {
2199 drop:
2200                 kfree_skb(skb);
2201         }
2202         return 0;
2203 }
2204 #endif
2205
2206 #ifdef CONFIG_IP_PIMSM_V2
2207 static int pim_rcv(struct sk_buff *skb)
2208 {
2209         struct pimreghdr *pim;
2210         struct net *net = dev_net(skb->dev);
2211         struct mr_table *mrt;
2212
2213         if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr)))
2214                 goto drop;
2215
2216         pim = (struct pimreghdr *)skb_transport_header(skb);
2217         if (pim->type != ((PIM_VERSION << 4) | (PIM_TYPE_REGISTER)) ||
2218             (pim->flags & PIM_NULL_REGISTER) ||
2219             (ip_compute_csum((void *)pim, sizeof(*pim)) != 0 &&
2220              csum_fold(skb_checksum(skb, 0, skb->len, 0))))
2221                 goto drop;
2222
2223         mrt = ipmr_rt_fib_lookup(net, skb);
2224         if (IS_ERR(mrt))
2225                 goto drop;
2226         if (__pim_rcv(mrt, skb, sizeof(*pim))) {
2227 drop:
2228                 kfree_skb(skb);
2229         }
2230         return 0;
2231 }
2232 #endif
2233
2234 int ipmr_get_route(struct net *net, struct sk_buff *skb,
2235                    __be32 saddr, __be32 daddr,
2236                    struct rtmsg *rtm, u32 portid)
2237 {
2238         struct mfc_cache *cache;
2239         struct mr_table *mrt;
2240         int err;
2241
2242         mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2243         if (!mrt)
2244                 return -ENOENT;
2245
2246         rcu_read_lock();
2247         cache = ipmr_cache_find(mrt, saddr, daddr);
2248         if (!cache && skb->dev) {
2249                 int vif = ipmr_find_vif(mrt, skb->dev);
2250
2251                 if (vif >= 0)
2252                         cache = ipmr_cache_find_any(mrt, daddr, vif);
2253         }
2254         if (!cache) {
2255                 struct sk_buff *skb2;
2256                 struct iphdr *iph;
2257                 struct net_device *dev;
2258                 int vif = -1;
2259
2260                 dev = skb->dev;
2261                 read_lock(&mrt_lock);
2262                 if (dev)
2263                         vif = ipmr_find_vif(mrt, dev);
2264                 if (vif < 0) {
2265                         read_unlock(&mrt_lock);
2266                         rcu_read_unlock();
2267                         return -ENODEV;
2268                 }
2269
2270                 skb2 = skb_realloc_headroom(skb, sizeof(struct iphdr));
2271                 if (!skb2) {
2272                         read_unlock(&mrt_lock);
2273                         rcu_read_unlock();
2274                         return -ENOMEM;
2275                 }
2276
2277                 NETLINK_CB(skb2).portid = portid;
2278                 skb_push(skb2, sizeof(struct iphdr));
2279                 skb_reset_network_header(skb2);
2280                 iph = ip_hdr(skb2);
2281                 iph->ihl = sizeof(struct iphdr) >> 2;
2282                 iph->saddr = saddr;
2283                 iph->daddr = daddr;
2284                 iph->version = 0;
2285                 err = ipmr_cache_unresolved(mrt, vif, skb2, dev);
2286                 read_unlock(&mrt_lock);
2287                 rcu_read_unlock();
2288                 return err;
2289         }
2290
2291         read_lock(&mrt_lock);
2292         err = mr_fill_mroute(mrt, skb, &cache->_c, rtm);
2293         read_unlock(&mrt_lock);
2294         rcu_read_unlock();
2295         return err;
2296 }
2297
2298 static int ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
2299                             u32 portid, u32 seq, struct mfc_cache *c, int cmd,
2300                             int flags)
2301 {
2302         struct nlmsghdr *nlh;
2303         struct rtmsg *rtm;
2304         int err;
2305
2306         nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rtm), flags);
2307         if (!nlh)
2308                 return -EMSGSIZE;
2309
2310         rtm = nlmsg_data(nlh);
2311         rtm->rtm_family   = RTNL_FAMILY_IPMR;
2312         rtm->rtm_dst_len  = 32;
2313         rtm->rtm_src_len  = 32;
2314         rtm->rtm_tos      = 0;
2315         rtm->rtm_table    = mrt->id;
2316         if (nla_put_u32(skb, RTA_TABLE, mrt->id))
2317                 goto nla_put_failure;
2318         rtm->rtm_type     = RTN_MULTICAST;
2319         rtm->rtm_scope    = RT_SCOPE_UNIVERSE;
2320         if (c->_c.mfc_flags & MFC_STATIC)
2321                 rtm->rtm_protocol = RTPROT_STATIC;
2322         else
2323                 rtm->rtm_protocol = RTPROT_MROUTED;
2324         rtm->rtm_flags    = 0;
2325
2326         if (nla_put_in_addr(skb, RTA_SRC, c->mfc_origin) ||
2327             nla_put_in_addr(skb, RTA_DST, c->mfc_mcastgrp))
2328                 goto nla_put_failure;
2329         err = mr_fill_mroute(mrt, skb, &c->_c, rtm);
2330         /* do not break the dump if cache is unresolved */
2331         if (err < 0 && err != -ENOENT)
2332                 goto nla_put_failure;
2333
2334         nlmsg_end(skb, nlh);
2335         return 0;
2336
2337 nla_put_failure:
2338         nlmsg_cancel(skb, nlh);
2339         return -EMSGSIZE;
2340 }
2341
2342 static int _ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
2343                              u32 portid, u32 seq, struct mr_mfc *c, int cmd,
2344                              int flags)
2345 {
2346         return ipmr_fill_mroute(mrt, skb, portid, seq, (struct mfc_cache *)c,
2347                                 cmd, flags);
2348 }
2349
2350 static size_t mroute_msgsize(bool unresolved, int maxvif)
2351 {
2352         size_t len =
2353                 NLMSG_ALIGN(sizeof(struct rtmsg))
2354                 + nla_total_size(4)     /* RTA_TABLE */
2355                 + nla_total_size(4)     /* RTA_SRC */
2356                 + nla_total_size(4)     /* RTA_DST */
2357                 ;
2358
2359         if (!unresolved)
2360                 len = len
2361                       + nla_total_size(4)       /* RTA_IIF */
2362                       + nla_total_size(0)       /* RTA_MULTIPATH */
2363                       + maxvif * NLA_ALIGN(sizeof(struct rtnexthop))
2364                                                 /* RTA_MFC_STATS */
2365                       + nla_total_size_64bit(sizeof(struct rta_mfc_stats))
2366                 ;
2367
2368         return len;
2369 }
2370
2371 static void mroute_netlink_event(struct mr_table *mrt, struct mfc_cache *mfc,
2372                                  int cmd)
2373 {
2374         struct net *net = read_pnet(&mrt->net);
2375         struct sk_buff *skb;
2376         int err = -ENOBUFS;
2377
2378         skb = nlmsg_new(mroute_msgsize(mfc->_c.mfc_parent >= MAXVIFS,
2379                                        mrt->maxvif),
2380                         GFP_ATOMIC);
2381         if (!skb)
2382                 goto errout;
2383
2384         err = ipmr_fill_mroute(mrt, skb, 0, 0, mfc, cmd, 0);
2385         if (err < 0)
2386                 goto errout;
2387
2388         rtnl_notify(skb, net, 0, RTNLGRP_IPV4_MROUTE, NULL, GFP_ATOMIC);
2389         return;
2390
2391 errout:
2392         kfree_skb(skb);
2393         if (err < 0)
2394                 rtnl_set_sk_err(net, RTNLGRP_IPV4_MROUTE, err);
2395 }
2396
2397 static size_t igmpmsg_netlink_msgsize(size_t payloadlen)
2398 {
2399         size_t len =
2400                 NLMSG_ALIGN(sizeof(struct rtgenmsg))
2401                 + nla_total_size(1)     /* IPMRA_CREPORT_MSGTYPE */
2402                 + nla_total_size(4)     /* IPMRA_CREPORT_VIF_ID */
2403                 + nla_total_size(4)     /* IPMRA_CREPORT_SRC_ADDR */
2404                 + nla_total_size(4)     /* IPMRA_CREPORT_DST_ADDR */
2405                 + nla_total_size(4)     /* IPMRA_CREPORT_TABLE */
2406                                         /* IPMRA_CREPORT_PKT */
2407                 + nla_total_size(payloadlen)
2408                 ;
2409
2410         return len;
2411 }
2412
2413 static void igmpmsg_netlink_event(struct mr_table *mrt, struct sk_buff *pkt)
2414 {
2415         struct net *net = read_pnet(&mrt->net);
2416         struct nlmsghdr *nlh;
2417         struct rtgenmsg *rtgenm;
2418         struct igmpmsg *msg;
2419         struct sk_buff *skb;
2420         struct nlattr *nla;
2421         int payloadlen;
2422
2423         payloadlen = pkt->len - sizeof(struct igmpmsg);
2424         msg = (struct igmpmsg *)skb_network_header(pkt);
2425
2426         skb = nlmsg_new(igmpmsg_netlink_msgsize(payloadlen), GFP_ATOMIC);
2427         if (!skb)
2428                 goto errout;
2429
2430         nlh = nlmsg_put(skb, 0, 0, RTM_NEWCACHEREPORT,
2431                         sizeof(struct rtgenmsg), 0);
2432         if (!nlh)
2433                 goto errout;
2434         rtgenm = nlmsg_data(nlh);
2435         rtgenm->rtgen_family = RTNL_FAMILY_IPMR;
2436         if (nla_put_u8(skb, IPMRA_CREPORT_MSGTYPE, msg->im_msgtype) ||
2437             nla_put_u32(skb, IPMRA_CREPORT_VIF_ID, msg->im_vif | (msg->im_vif_hi << 8)) ||
2438             nla_put_in_addr(skb, IPMRA_CREPORT_SRC_ADDR,
2439                             msg->im_src.s_addr) ||
2440             nla_put_in_addr(skb, IPMRA_CREPORT_DST_ADDR,
2441                             msg->im_dst.s_addr) ||
2442             nla_put_u32(skb, IPMRA_CREPORT_TABLE, mrt->id))
2443                 goto nla_put_failure;
2444
2445         nla = nla_reserve(skb, IPMRA_CREPORT_PKT, payloadlen);
2446         if (!nla || skb_copy_bits(pkt, sizeof(struct igmpmsg),
2447                                   nla_data(nla), payloadlen))
2448                 goto nla_put_failure;
2449
2450         nlmsg_end(skb, nlh);
2451
2452         rtnl_notify(skb, net, 0, RTNLGRP_IPV4_MROUTE_R, NULL, GFP_ATOMIC);
2453         return;
2454
2455 nla_put_failure:
2456         nlmsg_cancel(skb, nlh);
2457 errout:
2458         kfree_skb(skb);
2459         rtnl_set_sk_err(net, RTNLGRP_IPV4_MROUTE_R, -ENOBUFS);
2460 }
2461
2462 static int ipmr_rtm_valid_getroute_req(struct sk_buff *skb,
2463                                        const struct nlmsghdr *nlh,
2464                                        struct nlattr **tb,
2465                                        struct netlink_ext_ack *extack)
2466 {
2467         struct rtmsg *rtm;
2468         int i, err;
2469
2470         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*rtm))) {
2471                 NL_SET_ERR_MSG(extack, "ipv4: Invalid header for multicast route get request");
2472                 return -EINVAL;
2473         }
2474
2475         if (!netlink_strict_get_check(skb))
2476                 return nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX,
2477                                               rtm_ipv4_policy, extack);
2478
2479         rtm = nlmsg_data(nlh);
2480         if ((rtm->rtm_src_len && rtm->rtm_src_len != 32) ||
2481             (rtm->rtm_dst_len && rtm->rtm_dst_len != 32) ||
2482             rtm->rtm_tos || rtm->rtm_table || rtm->rtm_protocol ||
2483             rtm->rtm_scope || rtm->rtm_type || rtm->rtm_flags) {
2484                 NL_SET_ERR_MSG(extack, "ipv4: Invalid values in header for multicast route get request");
2485                 return -EINVAL;
2486         }
2487
2488         err = nlmsg_parse_deprecated_strict(nlh, sizeof(*rtm), tb, RTA_MAX,
2489                                             rtm_ipv4_policy, extack);
2490         if (err)
2491                 return err;
2492
2493         if ((tb[RTA_SRC] && !rtm->rtm_src_len) ||
2494             (tb[RTA_DST] && !rtm->rtm_dst_len)) {
2495                 NL_SET_ERR_MSG(extack, "ipv4: rtm_src_len and rtm_dst_len must be 32 for IPv4");
2496                 return -EINVAL;
2497         }
2498
2499         for (i = 0; i <= RTA_MAX; i++) {
2500                 if (!tb[i])
2501                         continue;
2502
2503                 switch (i) {
2504                 case RTA_SRC:
2505                 case RTA_DST:
2506                 case RTA_TABLE:
2507                         break;
2508                 default:
2509                         NL_SET_ERR_MSG(extack, "ipv4: Unsupported attribute in multicast route get request");
2510                         return -EINVAL;
2511                 }
2512         }
2513
2514         return 0;
2515 }
2516
2517 static int ipmr_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh,
2518                              struct netlink_ext_ack *extack)
2519 {
2520         struct net *net = sock_net(in_skb->sk);
2521         struct nlattr *tb[RTA_MAX + 1];
2522         struct sk_buff *skb = NULL;
2523         struct mfc_cache *cache;
2524         struct mr_table *mrt;
2525         __be32 src, grp;
2526         u32 tableid;
2527         int err;
2528
2529         err = ipmr_rtm_valid_getroute_req(in_skb, nlh, tb, extack);
2530         if (err < 0)
2531                 goto errout;
2532
2533         src = tb[RTA_SRC] ? nla_get_in_addr(tb[RTA_SRC]) : 0;
2534         grp = tb[RTA_DST] ? nla_get_in_addr(tb[RTA_DST]) : 0;
2535         tableid = tb[RTA_TABLE] ? nla_get_u32(tb[RTA_TABLE]) : 0;
2536
2537         mrt = ipmr_get_table(net, tableid ? tableid : RT_TABLE_DEFAULT);
2538         if (!mrt) {
2539                 err = -ENOENT;
2540                 goto errout_free;
2541         }
2542
2543         /* entries are added/deleted only under RTNL */
2544         rcu_read_lock();
2545         cache = ipmr_cache_find(mrt, src, grp);
2546         rcu_read_unlock();
2547         if (!cache) {
2548                 err = -ENOENT;
2549                 goto errout_free;
2550         }
2551
2552         skb = nlmsg_new(mroute_msgsize(false, mrt->maxvif), GFP_KERNEL);
2553         if (!skb) {
2554                 err = -ENOBUFS;
2555                 goto errout_free;
2556         }
2557
2558         err = ipmr_fill_mroute(mrt, skb, NETLINK_CB(in_skb).portid,
2559                                nlh->nlmsg_seq, cache,
2560                                RTM_NEWROUTE, 0);
2561         if (err < 0)
2562                 goto errout_free;
2563
2564         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
2565
2566 errout:
2567         return err;
2568
2569 errout_free:
2570         kfree_skb(skb);
2571         goto errout;
2572 }
2573
2574 static int ipmr_rtm_dumproute(struct sk_buff *skb, struct netlink_callback *cb)
2575 {
2576         struct fib_dump_filter filter = {};
2577         int err;
2578
2579         if (cb->strict_check) {
2580                 err = ip_valid_fib_dump_req(sock_net(skb->sk), cb->nlh,
2581                                             &filter, cb);
2582                 if (err < 0)
2583                         return err;
2584         }
2585
2586         if (filter.table_id) {
2587                 struct mr_table *mrt;
2588
2589                 mrt = ipmr_get_table(sock_net(skb->sk), filter.table_id);
2590                 if (!mrt) {
2591                         if (rtnl_msg_family(cb->nlh) != RTNL_FAMILY_IPMR)
2592                                 return skb->len;
2593
2594                         NL_SET_ERR_MSG(cb->extack, "ipv4: MR table does not exist");
2595                         return -ENOENT;
2596                 }
2597                 err = mr_table_dump(mrt, skb, cb, _ipmr_fill_mroute,
2598                                     &mfc_unres_lock, &filter);
2599                 return skb->len ? : err;
2600         }
2601
2602         return mr_rtm_dumproute(skb, cb, ipmr_mr_table_iter,
2603                                 _ipmr_fill_mroute, &mfc_unres_lock, &filter);
2604 }
2605
2606 static const struct nla_policy rtm_ipmr_policy[RTA_MAX + 1] = {
2607         [RTA_SRC]       = { .type = NLA_U32 },
2608         [RTA_DST]       = { .type = NLA_U32 },
2609         [RTA_IIF]       = { .type = NLA_U32 },
2610         [RTA_TABLE]     = { .type = NLA_U32 },
2611         [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
2612 };
2613
2614 static bool ipmr_rtm_validate_proto(unsigned char rtm_protocol)
2615 {
2616         switch (rtm_protocol) {
2617         case RTPROT_STATIC:
2618         case RTPROT_MROUTED:
2619                 return true;
2620         }
2621         return false;
2622 }
2623
2624 static int ipmr_nla_get_ttls(const struct nlattr *nla, struct mfcctl *mfcc)
2625 {
2626         struct rtnexthop *rtnh = nla_data(nla);
2627         int remaining = nla_len(nla), vifi = 0;
2628
2629         while (rtnh_ok(rtnh, remaining)) {
2630                 mfcc->mfcc_ttls[vifi] = rtnh->rtnh_hops;
2631                 if (++vifi == MAXVIFS)
2632                         break;
2633                 rtnh = rtnh_next(rtnh, &remaining);
2634         }
2635
2636         return remaining > 0 ? -EINVAL : vifi;
2637 }
2638
2639 /* returns < 0 on error, 0 for ADD_MFC and 1 for ADD_MFC_PROXY */
2640 static int rtm_to_ipmr_mfcc(struct net *net, struct nlmsghdr *nlh,
2641                             struct mfcctl *mfcc, int *mrtsock,
2642                             struct mr_table **mrtret,
2643                             struct netlink_ext_ack *extack)
2644 {
2645         struct net_device *dev = NULL;
2646         u32 tblid = RT_TABLE_DEFAULT;
2647         struct mr_table *mrt;
2648         struct nlattr *attr;
2649         struct rtmsg *rtm;
2650         int ret, rem;
2651
2652         ret = nlmsg_validate_deprecated(nlh, sizeof(*rtm), RTA_MAX,
2653                                         rtm_ipmr_policy, extack);
2654         if (ret < 0)
2655                 goto out;
2656         rtm = nlmsg_data(nlh);
2657
2658         ret = -EINVAL;
2659         if (rtm->rtm_family != RTNL_FAMILY_IPMR || rtm->rtm_dst_len != 32 ||
2660             rtm->rtm_type != RTN_MULTICAST ||
2661             rtm->rtm_scope != RT_SCOPE_UNIVERSE ||
2662             !ipmr_rtm_validate_proto(rtm->rtm_protocol))
2663                 goto out;
2664
2665         memset(mfcc, 0, sizeof(*mfcc));
2666         mfcc->mfcc_parent = -1;
2667         ret = 0;
2668         nlmsg_for_each_attr(attr, nlh, sizeof(struct rtmsg), rem) {
2669                 switch (nla_type(attr)) {
2670                 case RTA_SRC:
2671                         mfcc->mfcc_origin.s_addr = nla_get_be32(attr);
2672                         break;
2673                 case RTA_DST:
2674                         mfcc->mfcc_mcastgrp.s_addr = nla_get_be32(attr);
2675                         break;
2676                 case RTA_IIF:
2677                         dev = __dev_get_by_index(net, nla_get_u32(attr));
2678                         if (!dev) {
2679                                 ret = -ENODEV;
2680                                 goto out;
2681                         }
2682                         break;
2683                 case RTA_MULTIPATH:
2684                         if (ipmr_nla_get_ttls(attr, mfcc) < 0) {
2685                                 ret = -EINVAL;
2686                                 goto out;
2687                         }
2688                         break;
2689                 case RTA_PREFSRC:
2690                         ret = 1;
2691                         break;
2692                 case RTA_TABLE:
2693                         tblid = nla_get_u32(attr);
2694                         break;
2695                 }
2696         }
2697         mrt = ipmr_get_table(net, tblid);
2698         if (!mrt) {
2699                 ret = -ENOENT;
2700                 goto out;
2701         }
2702         *mrtret = mrt;
2703         *mrtsock = rtm->rtm_protocol == RTPROT_MROUTED ? 1 : 0;
2704         if (dev)
2705                 mfcc->mfcc_parent = ipmr_find_vif(mrt, dev);
2706
2707 out:
2708         return ret;
2709 }
2710
2711 /* takes care of both newroute and delroute */
2712 static int ipmr_rtm_route(struct sk_buff *skb, struct nlmsghdr *nlh,
2713                           struct netlink_ext_ack *extack)
2714 {
2715         struct net *net = sock_net(skb->sk);
2716         int ret, mrtsock, parent;
2717         struct mr_table *tbl;
2718         struct mfcctl mfcc;
2719
2720         mrtsock = 0;
2721         tbl = NULL;
2722         ret = rtm_to_ipmr_mfcc(net, nlh, &mfcc, &mrtsock, &tbl, extack);
2723         if (ret < 0)
2724                 return ret;
2725
2726         parent = ret ? mfcc.mfcc_parent : -1;
2727         if (nlh->nlmsg_type == RTM_NEWROUTE)
2728                 return ipmr_mfc_add(net, tbl, &mfcc, mrtsock, parent);
2729         else
2730                 return ipmr_mfc_delete(tbl, &mfcc, parent);
2731 }
2732
2733 static bool ipmr_fill_table(struct mr_table *mrt, struct sk_buff *skb)
2734 {
2735         u32 queue_len = atomic_read(&mrt->cache_resolve_queue_len);
2736
2737         if (nla_put_u32(skb, IPMRA_TABLE_ID, mrt->id) ||
2738             nla_put_u32(skb, IPMRA_TABLE_CACHE_RES_QUEUE_LEN, queue_len) ||
2739             nla_put_s32(skb, IPMRA_TABLE_MROUTE_REG_VIF_NUM,
2740                         mrt->mroute_reg_vif_num) ||
2741             nla_put_u8(skb, IPMRA_TABLE_MROUTE_DO_ASSERT,
2742                        mrt->mroute_do_assert) ||
2743             nla_put_u8(skb, IPMRA_TABLE_MROUTE_DO_PIM, mrt->mroute_do_pim) ||
2744             nla_put_u8(skb, IPMRA_TABLE_MROUTE_DO_WRVIFWHOLE,
2745                        mrt->mroute_do_wrvifwhole))
2746                 return false;
2747
2748         return true;
2749 }
2750
2751 static bool ipmr_fill_vif(struct mr_table *mrt, u32 vifid, struct sk_buff *skb)
2752 {
2753         struct nlattr *vif_nest;
2754         struct vif_device *vif;
2755
2756         /* if the VIF doesn't exist just continue */
2757         if (!VIF_EXISTS(mrt, vifid))
2758                 return true;
2759
2760         vif = &mrt->vif_table[vifid];
2761         vif_nest = nla_nest_start_noflag(skb, IPMRA_VIF);
2762         if (!vif_nest)
2763                 return false;
2764         if (nla_put_u32(skb, IPMRA_VIFA_IFINDEX, vif->dev->ifindex) ||
2765             nla_put_u32(skb, IPMRA_VIFA_VIF_ID, vifid) ||
2766             nla_put_u16(skb, IPMRA_VIFA_FLAGS, vif->flags) ||
2767             nla_put_u64_64bit(skb, IPMRA_VIFA_BYTES_IN, vif->bytes_in,
2768                               IPMRA_VIFA_PAD) ||
2769             nla_put_u64_64bit(skb, IPMRA_VIFA_BYTES_OUT, vif->bytes_out,
2770                               IPMRA_VIFA_PAD) ||
2771             nla_put_u64_64bit(skb, IPMRA_VIFA_PACKETS_IN, vif->pkt_in,
2772                               IPMRA_VIFA_PAD) ||
2773             nla_put_u64_64bit(skb, IPMRA_VIFA_PACKETS_OUT, vif->pkt_out,
2774                               IPMRA_VIFA_PAD) ||
2775             nla_put_be32(skb, IPMRA_VIFA_LOCAL_ADDR, vif->local) ||
2776             nla_put_be32(skb, IPMRA_VIFA_REMOTE_ADDR, vif->remote)) {
2777                 nla_nest_cancel(skb, vif_nest);
2778                 return false;
2779         }
2780         nla_nest_end(skb, vif_nest);
2781
2782         return true;
2783 }
2784
2785 static int ipmr_valid_dumplink(const struct nlmsghdr *nlh,
2786                                struct netlink_ext_ack *extack)
2787 {
2788         struct ifinfomsg *ifm;
2789
2790         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
2791                 NL_SET_ERR_MSG(extack, "ipv4: Invalid header for ipmr link dump");
2792                 return -EINVAL;
2793         }
2794
2795         if (nlmsg_attrlen(nlh, sizeof(*ifm))) {
2796                 NL_SET_ERR_MSG(extack, "Invalid data after header in ipmr link dump");
2797                 return -EINVAL;
2798         }
2799
2800         ifm = nlmsg_data(nlh);
2801         if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
2802             ifm->ifi_change || ifm->ifi_index) {
2803                 NL_SET_ERR_MSG(extack, "Invalid values in header for ipmr link dump request");
2804                 return -EINVAL;
2805         }
2806
2807         return 0;
2808 }
2809
2810 static int ipmr_rtm_dumplink(struct sk_buff *skb, struct netlink_callback *cb)
2811 {
2812         struct net *net = sock_net(skb->sk);
2813         struct nlmsghdr *nlh = NULL;
2814         unsigned int t = 0, s_t;
2815         unsigned int e = 0, s_e;
2816         struct mr_table *mrt;
2817
2818         if (cb->strict_check) {
2819                 int err = ipmr_valid_dumplink(cb->nlh, cb->extack);
2820
2821                 if (err < 0)
2822                         return err;
2823         }
2824
2825         s_t = cb->args[0];
2826         s_e = cb->args[1];
2827
2828         ipmr_for_each_table(mrt, net) {
2829                 struct nlattr *vifs, *af;
2830                 struct ifinfomsg *hdr;
2831                 u32 i;
2832
2833                 if (t < s_t)
2834                         goto skip_table;
2835                 nlh = nlmsg_put(skb, NETLINK_CB(cb->skb).portid,
2836                                 cb->nlh->nlmsg_seq, RTM_NEWLINK,
2837                                 sizeof(*hdr), NLM_F_MULTI);
2838                 if (!nlh)
2839                         break;
2840
2841                 hdr = nlmsg_data(nlh);
2842                 memset(hdr, 0, sizeof(*hdr));
2843                 hdr->ifi_family = RTNL_FAMILY_IPMR;
2844
2845                 af = nla_nest_start_noflag(skb, IFLA_AF_SPEC);
2846                 if (!af) {
2847                         nlmsg_cancel(skb, nlh);
2848                         goto out;
2849                 }
2850
2851                 if (!ipmr_fill_table(mrt, skb)) {
2852                         nlmsg_cancel(skb, nlh);
2853                         goto out;
2854                 }
2855
2856                 vifs = nla_nest_start_noflag(skb, IPMRA_TABLE_VIFS);
2857                 if (!vifs) {
2858                         nla_nest_end(skb, af);
2859                         nlmsg_end(skb, nlh);
2860                         goto out;
2861                 }
2862                 for (i = 0; i < mrt->maxvif; i++) {
2863                         if (e < s_e)
2864                                 goto skip_entry;
2865                         if (!ipmr_fill_vif(mrt, i, skb)) {
2866                                 nla_nest_end(skb, vifs);
2867                                 nla_nest_end(skb, af);
2868                                 nlmsg_end(skb, nlh);
2869                                 goto out;
2870                         }
2871 skip_entry:
2872                         e++;
2873                 }
2874                 s_e = 0;
2875                 e = 0;
2876                 nla_nest_end(skb, vifs);
2877                 nla_nest_end(skb, af);
2878                 nlmsg_end(skb, nlh);
2879 skip_table:
2880                 t++;
2881         }
2882
2883 out:
2884         cb->args[1] = e;
2885         cb->args[0] = t;
2886
2887         return skb->len;
2888 }
2889
2890 #ifdef CONFIG_PROC_FS
2891 /* The /proc interfaces to multicast routing :
2892  * /proc/net/ip_mr_cache & /proc/net/ip_mr_vif
2893  */
2894
2895 static void *ipmr_vif_seq_start(struct seq_file *seq, loff_t *pos)
2896         __acquires(mrt_lock)
2897 {
2898         struct mr_vif_iter *iter = seq->private;
2899         struct net *net = seq_file_net(seq);
2900         struct mr_table *mrt;
2901
2902         mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2903         if (!mrt)
2904                 return ERR_PTR(-ENOENT);
2905
2906         iter->mrt = mrt;
2907
2908         read_lock(&mrt_lock);
2909         return mr_vif_seq_start(seq, pos);
2910 }
2911
2912 static void ipmr_vif_seq_stop(struct seq_file *seq, void *v)
2913         __releases(mrt_lock)
2914 {
2915         read_unlock(&mrt_lock);
2916 }
2917
2918 static int ipmr_vif_seq_show(struct seq_file *seq, void *v)
2919 {
2920         struct mr_vif_iter *iter = seq->private;
2921         struct mr_table *mrt = iter->mrt;
2922
2923         if (v == SEQ_START_TOKEN) {
2924                 seq_puts(seq,
2925                          "Interface      BytesIn  PktsIn  BytesOut PktsOut Flags Local    Remote\n");
2926         } else {
2927                 const struct vif_device *vif = v;
2928                 const char *name =  vif->dev ?
2929                                     vif->dev->name : "none";
2930
2931                 seq_printf(seq,
2932                            "%2td %-10s %8ld %7ld  %8ld %7ld %05X %08X %08X\n",
2933                            vif - mrt->vif_table,
2934                            name, vif->bytes_in, vif->pkt_in,
2935                            vif->bytes_out, vif->pkt_out,
2936                            vif->flags, vif->local, vif->remote);
2937         }
2938         return 0;
2939 }
2940
2941 static const struct seq_operations ipmr_vif_seq_ops = {
2942         .start = ipmr_vif_seq_start,
2943         .next  = mr_vif_seq_next,
2944         .stop  = ipmr_vif_seq_stop,
2945         .show  = ipmr_vif_seq_show,
2946 };
2947
2948 static void *ipmr_mfc_seq_start(struct seq_file *seq, loff_t *pos)
2949 {
2950         struct net *net = seq_file_net(seq);
2951         struct mr_table *mrt;
2952
2953         mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2954         if (!mrt)
2955                 return ERR_PTR(-ENOENT);
2956
2957         return mr_mfc_seq_start(seq, pos, mrt, &mfc_unres_lock);
2958 }
2959
2960 static int ipmr_mfc_seq_show(struct seq_file *seq, void *v)
2961 {
2962         int n;
2963
2964         if (v == SEQ_START_TOKEN) {
2965                 seq_puts(seq,
2966                  "Group    Origin   Iif     Pkts    Bytes    Wrong Oifs\n");
2967         } else {
2968                 const struct mfc_cache *mfc = v;
2969                 const struct mr_mfc_iter *it = seq->private;
2970                 const struct mr_table *mrt = it->mrt;
2971
2972                 seq_printf(seq, "%08X %08X %-3hd",
2973                            (__force u32) mfc->mfc_mcastgrp,
2974                            (__force u32) mfc->mfc_origin,
2975                            mfc->_c.mfc_parent);
2976
2977                 if (it->cache != &mrt->mfc_unres_queue) {
2978                         seq_printf(seq, " %8lu %8lu %8lu",
2979                                    mfc->_c.mfc_un.res.pkt,
2980                                    mfc->_c.mfc_un.res.bytes,
2981                                    mfc->_c.mfc_un.res.wrong_if);
2982                         for (n = mfc->_c.mfc_un.res.minvif;
2983                              n < mfc->_c.mfc_un.res.maxvif; n++) {
2984                                 if (VIF_EXISTS(mrt, n) &&
2985                                     mfc->_c.mfc_un.res.ttls[n] < 255)
2986                                         seq_printf(seq,
2987                                            " %2d:%-3d",
2988                                            n, mfc->_c.mfc_un.res.ttls[n]);
2989                         }
2990                 } else {
2991                         /* unresolved mfc_caches don't contain
2992                          * pkt, bytes and wrong_if values
2993                          */
2994                         seq_printf(seq, " %8lu %8lu %8lu", 0ul, 0ul, 0ul);
2995                 }
2996                 seq_putc(seq, '\n');
2997         }
2998         return 0;
2999 }
3000
3001 static const struct seq_operations ipmr_mfc_seq_ops = {
3002         .start = ipmr_mfc_seq_start,
3003         .next  = mr_mfc_seq_next,
3004         .stop  = mr_mfc_seq_stop,
3005         .show  = ipmr_mfc_seq_show,
3006 };
3007 #endif
3008
3009 #ifdef CONFIG_IP_PIMSM_V2
3010 static const struct net_protocol pim_protocol = {
3011         .handler        =       pim_rcv,
3012         .netns_ok       =       1,
3013 };
3014 #endif
3015
3016 static unsigned int ipmr_seq_read(struct net *net)
3017 {
3018         ASSERT_RTNL();
3019
3020         return net->ipv4.ipmr_seq + ipmr_rules_seq_read(net);
3021 }
3022
3023 static int ipmr_dump(struct net *net, struct notifier_block *nb,
3024                      struct netlink_ext_ack *extack)
3025 {
3026         return mr_dump(net, nb, RTNL_FAMILY_IPMR, ipmr_rules_dump,
3027                        ipmr_mr_table_iter, &mrt_lock, extack);
3028 }
3029
3030 static const struct fib_notifier_ops ipmr_notifier_ops_template = {
3031         .family         = RTNL_FAMILY_IPMR,
3032         .fib_seq_read   = ipmr_seq_read,
3033         .fib_dump       = ipmr_dump,
3034         .owner          = THIS_MODULE,
3035 };
3036
3037 static int __net_init ipmr_notifier_init(struct net *net)
3038 {
3039         struct fib_notifier_ops *ops;
3040
3041         net->ipv4.ipmr_seq = 0;
3042
3043         ops = fib_notifier_ops_register(&ipmr_notifier_ops_template, net);
3044         if (IS_ERR(ops))
3045                 return PTR_ERR(ops);
3046         net->ipv4.ipmr_notifier_ops = ops;
3047
3048         return 0;
3049 }
3050
3051 static void __net_exit ipmr_notifier_exit(struct net *net)
3052 {
3053         fib_notifier_ops_unregister(net->ipv4.ipmr_notifier_ops);
3054         net->ipv4.ipmr_notifier_ops = NULL;
3055 }
3056
3057 /* Setup for IP multicast routing */
3058 static int __net_init ipmr_net_init(struct net *net)
3059 {
3060         int err;
3061
3062         err = ipmr_notifier_init(net);
3063         if (err)
3064                 goto ipmr_notifier_fail;
3065
3066         err = ipmr_rules_init(net);
3067         if (err < 0)
3068                 goto ipmr_rules_fail;
3069
3070 #ifdef CONFIG_PROC_FS
3071         err = -ENOMEM;
3072         if (!proc_create_net("ip_mr_vif", 0, net->proc_net, &ipmr_vif_seq_ops,
3073                         sizeof(struct mr_vif_iter)))
3074                 goto proc_vif_fail;
3075         if (!proc_create_net("ip_mr_cache", 0, net->proc_net, &ipmr_mfc_seq_ops,
3076                         sizeof(struct mr_mfc_iter)))
3077                 goto proc_cache_fail;
3078 #endif
3079         return 0;
3080
3081 #ifdef CONFIG_PROC_FS
3082 proc_cache_fail:
3083         remove_proc_entry("ip_mr_vif", net->proc_net);
3084 proc_vif_fail:
3085         ipmr_rules_exit(net);
3086 #endif
3087 ipmr_rules_fail:
3088         ipmr_notifier_exit(net);
3089 ipmr_notifier_fail:
3090         return err;
3091 }
3092
3093 static void __net_exit ipmr_net_exit(struct net *net)
3094 {
3095 #ifdef CONFIG_PROC_FS
3096         remove_proc_entry("ip_mr_cache", net->proc_net);
3097         remove_proc_entry("ip_mr_vif", net->proc_net);
3098 #endif
3099         ipmr_notifier_exit(net);
3100         ipmr_rules_exit(net);
3101 }
3102
3103 static struct pernet_operations ipmr_net_ops = {
3104         .init = ipmr_net_init,
3105         .exit = ipmr_net_exit,
3106 };
3107
3108 int __init ip_mr_init(void)
3109 {
3110         int err;
3111
3112         mrt_cachep = kmem_cache_create("ip_mrt_cache",
3113                                        sizeof(struct mfc_cache),
3114                                        0, SLAB_HWCACHE_ALIGN | SLAB_PANIC,
3115                                        NULL);
3116
3117         err = register_pernet_subsys(&ipmr_net_ops);
3118         if (err)
3119                 goto reg_pernet_fail;
3120
3121         err = register_netdevice_notifier(&ip_mr_notifier);
3122         if (err)
3123                 goto reg_notif_fail;
3124 #ifdef CONFIG_IP_PIMSM_V2
3125         if (inet_add_protocol(&pim_protocol, IPPROTO_PIM) < 0) {
3126                 pr_err("%s: can't add PIM protocol\n", __func__);
3127                 err = -EAGAIN;
3128                 goto add_proto_fail;
3129         }
3130 #endif
3131         rtnl_register(RTNL_FAMILY_IPMR, RTM_GETROUTE,
3132                       ipmr_rtm_getroute, ipmr_rtm_dumproute, 0);
3133         rtnl_register(RTNL_FAMILY_IPMR, RTM_NEWROUTE,
3134                       ipmr_rtm_route, NULL, 0);
3135         rtnl_register(RTNL_FAMILY_IPMR, RTM_DELROUTE,
3136                       ipmr_rtm_route, NULL, 0);
3137
3138         rtnl_register(RTNL_FAMILY_IPMR, RTM_GETLINK,
3139                       NULL, ipmr_rtm_dumplink, 0);
3140         return 0;
3141
3142 #ifdef CONFIG_IP_PIMSM_V2
3143 add_proto_fail:
3144         unregister_netdevice_notifier(&ip_mr_notifier);
3145 #endif
3146 reg_notif_fail:
3147         unregister_pernet_subsys(&ipmr_net_ops);
3148 reg_pernet_fail:
3149         kmem_cache_destroy(mrt_cachep);
3150         return err;
3151 }