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_SPINLOCK(mrt_lock);
  81
  82 static struct net_device *vif_dev_read(const struct vif_device *vif)
  83 {
  84         return rcu_dereference(vif->dev);
  85 }
  86
  87 /* Multicast router control variables */
  88
  89 /* Special spinlock for queue of unresolved entries */
  90 static DEFINE_SPINLOCK(mfc_unres_lock);
  91
  92 /* We return to original Alan's scheme. Hash table of resolved
  93  * entries is changed only in process context and protected
  94  * with weak lock mrt_lock. Queue of unresolved entries is protected
  95  * with strong spinlock mfc_unres_lock.
  96  *
  97  * In this case data path is free of exclusive locks at all.
  98  */
  99
 100 static struct kmem_cache *mrt_cachep __ro_after_init;
 101
 102 static struct mr_table *ipmr_new_table(struct net *net, u32 id);
 103 static void ipmr_free_table(struct mr_table *mrt);
 104
 105 static void ip_mr_forward(struct net *net, struct mr_table *mrt,
 106                           struct net_device *dev, struct sk_buff *skb,
 107                           struct mfc_cache *cache, int local);
 108 static int ipmr_cache_report(const struct mr_table *mrt,
 109                              struct sk_buff *pkt, vifi_t vifi, int assert);
 110 static void mroute_netlink_event(struct mr_table *mrt, struct mfc_cache *mfc,
 111                                  int cmd);
 112 static void igmpmsg_netlink_event(const struct mr_table *mrt, struct sk_buff *pkt);
 113 static void mroute_clean_tables(struct mr_table *mrt, int flags);
 114 static void ipmr_expire_process(struct timer_list *t);
 115
 116 #ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES
 117 #define ipmr_for_each_table(mrt, net)                                   \
 118         list_for_each_entry_rcu(mrt, &net->ipv4.mr_tables, list,        \
 119                                 lockdep_rtnl_is_held() ||               \
 120                                 list_empty(&net->ipv4.mr_tables))
 121
 122 static struct mr_table *ipmr_mr_table_iter(struct net *net,
 123                                            struct mr_table *mrt)
 124 {
 125         struct mr_table *ret;
 126
 127         if (!mrt)
 128                 ret = list_entry_rcu(net->ipv4.mr_tables.next,
 129                                      struct mr_table, list);
 130         else
 131                 ret = list_entry_rcu(mrt->list.next,
 132                                      struct mr_table, list);
 133
 134         if (&ret->list == &net->ipv4.mr_tables)
 135                 return NULL;
 136         return ret;
 137 }
 138
 139 static struct mr_table *ipmr_get_table(struct net *net, u32 id)
 140 {
 141         struct mr_table *mrt;
 142
 143         ipmr_for_each_table(mrt, net) {
 144                 if (mrt->id == id)
 145                         return mrt;
 146         }
 147         return NULL;
 148 }
 149
 150 static int ipmr_fib_lookup(struct net *net, struct flowi4 *flp4,
 151                            struct mr_table **mrt)
 152 {
 153         int err;
 154         struct ipmr_result res;
 155         struct fib_lookup_arg arg = {
 156                 .result = &res,
 157                 .flags = FIB_LOOKUP_NOREF,
 158         };
 159
 160         /* update flow if oif or iif point to device enslaved to l3mdev */
 161         l3mdev_update_flow(net, flowi4_to_flowi(flp4));
 162
 163         err = fib_rules_lookup(net->ipv4.mr_rules_ops,
 164                                flowi4_to_flowi(flp4), 0, &arg);
 165         if (err < 0)
 166                 return err;
 167         *mrt = res.mrt;
 168         return 0;
 169 }
 170
 171 static int ipmr_rule_action(struct fib_rule *rule, struct flowi *flp,
 172                             int flags, struct fib_lookup_arg *arg)
 173 {
 174         struct ipmr_result *res = arg->result;
 175         struct mr_table *mrt;
 176
 177         switch (rule->action) {
 178         case FR_ACT_TO_TBL:
 179                 break;
 180         case FR_ACT_UNREACHABLE:
 181                 return -ENETUNREACH;
 182         case FR_ACT_PROHIBIT:
 183                 return -EACCES;
 184         case FR_ACT_BLACKHOLE:
 185         default:
 186                 return -EINVAL;
 187         }
 188
 189         arg->table = fib_rule_get_table(rule, arg);
 190
 191         mrt = ipmr_get_table(rule->fr_net, arg->table);
 192         if (!mrt)
 193                 return -EAGAIN;
 194         res->mrt = mrt;
 195         return 0;
 196 }
 197
 198 static int ipmr_rule_match(struct fib_rule *rule, struct flowi *fl, int flags)
 199 {
 200         return 1;
 201 }
 202
 203 static int ipmr_rule_configure(struct fib_rule *rule, struct sk_buff *skb,
 204                                struct fib_rule_hdr *frh, struct nlattr **tb,
 205                                struct netlink_ext_ack *extack)
 206 {
 207         return 0;
 208 }
 209
 210 static int ipmr_rule_compare(struct fib_rule *rule, struct fib_rule_hdr *frh,
 211                              struct nlattr **tb)
 212 {
 213         return 1;
 214 }
 215
 216 static int ipmr_rule_fill(struct fib_rule *rule, struct sk_buff *skb,
 217                           struct fib_rule_hdr *frh)
 218 {
 219         frh->dst_len = 0;
 220         frh->src_len = 0;
 221         frh->tos     = 0;
 222         return 0;
 223 }
 224
 225 static const struct fib_rules_ops __net_initconst ipmr_rules_ops_template = {
 226         .family         = RTNL_FAMILY_IPMR,
 227         .rule_size      = sizeof(struct ipmr_rule),
 228         .addr_size      = sizeof(u32),
 229         .action         = ipmr_rule_action,
 230         .match          = ipmr_rule_match,
 231         .configure      = ipmr_rule_configure,
 232         .compare        = ipmr_rule_compare,
 233         .fill           = ipmr_rule_fill,
 234         .nlgroup        = RTNLGRP_IPV4_RULE,
 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);
 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         ASSERT_RTNL();
 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 }
 283
 284 static int ipmr_rules_dump(struct net *net, struct notifier_block *nb,
 285                            struct netlink_ext_ack *extack)
 286 {
 287         return fib_rules_dump(net, nb, RTNL_FAMILY_IPMR, extack);
 288 }
 289
 290 static unsigned int ipmr_rules_seq_read(struct net *net)
 291 {
 292         return fib_rules_seq_read(net, RTNL_FAMILY_IPMR);
 293 }
 294
 295 bool ipmr_rule_default(const struct fib_rule *rule)
 296 {
 297         return fib_rule_matchall(rule) && rule->table == RT_TABLE_DEFAULT;
 298 }
 299 EXPORT_SYMBOL(ipmr_rule_default);
 300 #else
 301 #define ipmr_for_each_table(mrt, net) \
 302         for (mrt = net->ipv4.mrt; mrt; mrt = NULL)
 303
 304 static struct mr_table *ipmr_mr_table_iter(struct net *net,
 305                                            struct mr_table *mrt)
 306 {
 307         if (!mrt)
 308                 return net->ipv4.mrt;
 309         return NULL;
 310 }
 311
 312 static struct mr_table *ipmr_get_table(struct net *net, u32 id)
 313 {
 314         return net->ipv4.mrt;
 315 }
 316
 317 static int ipmr_fib_lookup(struct net *net, struct flowi4 *flp4,
 318                            struct mr_table **mrt)
 319 {
 320         *mrt = net->ipv4.mrt;
 321         return 0;
 322 }
 323
 324 static int __net_init ipmr_rules_init(struct net *net)
 325 {
 326         struct mr_table *mrt;
 327
 328         mrt = ipmr_new_table(net, RT_TABLE_DEFAULT);
 329         if (IS_ERR(mrt))
 330                 return PTR_ERR(mrt);
 331         net->ipv4.mrt = mrt;
 332         return 0;
 333 }
 334
 335 static void __net_exit ipmr_rules_exit(struct net *net)
 336 {
 337         ASSERT_RTNL();
 338         ipmr_free_table(net->ipv4.mrt);
 339         net->ipv4.mrt = NULL;
 340 }
 341
 342 static int ipmr_rules_dump(struct net *net, struct notifier_block *nb,
 343                            struct netlink_ext_ack *extack)
 344 {
 345         return 0;
 346 }
 347
 348 static unsigned int ipmr_rules_seq_read(struct net *net)
 349 {
 350         return 0;
 351 }
 352
 353 bool ipmr_rule_default(const struct fib_rule *rule)
 354 {
 355         return true;
 356 }
 357 EXPORT_SYMBOL(ipmr_rule_default);
 358 #endif
 359
 360 static inline int ipmr_hash_cmp(struct rhashtable_compare_arg *arg,
 361                                 const void *ptr)
 362 {
 363         const struct mfc_cache_cmp_arg *cmparg = arg->key;
 364         const struct mfc_cache *c = ptr;
 365
 366         return cmparg->mfc_mcastgrp != c->mfc_mcastgrp ||
 367                cmparg->mfc_origin != c->mfc_origin;
 368 }
 369
 370 static const struct rhashtable_params ipmr_rht_params = {
 371         .head_offset = offsetof(struct mr_mfc, mnode),
 372         .key_offset = offsetof(struct mfc_cache, cmparg),
 373         .key_len = sizeof(struct mfc_cache_cmp_arg),
 374         .nelem_hint = 3,
 375         .obj_cmpfn = ipmr_hash_cmp,
 376         .automatic_shrinking = true,
 377 };
 378
 379 static void ipmr_new_table_set(struct mr_table *mrt,
 380                                struct net *net)
 381 {
 382 #ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES
 383         list_add_tail_rcu(&mrt->list, &net->ipv4.mr_tables);
 384 #endif
 385 }
 386
 387 static struct mfc_cache_cmp_arg ipmr_mr_table_ops_cmparg_any = {
 388         .mfc_mcastgrp = htonl(INADDR_ANY),
 389         .mfc_origin = htonl(INADDR_ANY),
 390 };
 391
 392 static struct mr_table_ops ipmr_mr_table_ops = {
 393         .rht_params = &ipmr_rht_params,
 394         .cmparg_any = &ipmr_mr_table_ops_cmparg_any,
 395 };
 396
 397 static struct mr_table *ipmr_new_table(struct net *net, u32 id)
 398 {
 399         struct mr_table *mrt;
 400
 401         /* "pimreg%u" should not exceed 16 bytes (IFNAMSIZ) */
 402         if (id != RT_TABLE_DEFAULT && id >= 1000000000)
 403                 return ERR_PTR(-EINVAL);
 404
 405         mrt = ipmr_get_table(net, id);
 406         if (mrt)
 407                 return mrt;
 408
 409         return mr_table_alloc(net, id, &ipmr_mr_table_ops,
 410                               ipmr_expire_process, ipmr_new_table_set);
 411 }
 412
 413 static void ipmr_free_table(struct mr_table *mrt)
 414 {
 415         timer_shutdown_sync(&mrt->ipmr_expire_timer);
 416         mroute_clean_tables(mrt, MRT_FLUSH_VIFS | MRT_FLUSH_VIFS_STATIC |
 417                                  MRT_FLUSH_MFC | MRT_FLUSH_MFC_STATIC);
 418         rhltable_destroy(&mrt->mfc_hash);
 419         kfree(mrt);
 420 }
 421
 422 /* Service routines creating virtual interfaces: DVMRP tunnels and PIMREG */
 423
 424 /* Initialize ipmr pimreg/tunnel in_device */
 425 static bool ipmr_init_vif_indev(const struct net_device *dev)
 426 {
 427         struct in_device *in_dev;
 428
 429         ASSERT_RTNL();
 430
 431         in_dev = __in_dev_get_rtnl(dev);
 432         if (!in_dev)
 433                 return false;
 434         ipv4_devconf_setall(in_dev);
 435         neigh_parms_data_state_setall(in_dev->arp_parms);
 436         IPV4_DEVCONF(in_dev->cnf, RP_FILTER) = 0;
 437
 438         return true;
 439 }
 440
 441 static struct net_device *ipmr_new_tunnel(struct net *net, struct vifctl *v)
 442 {
 443         struct net_device *tunnel_dev, *new_dev;
 444         struct ip_tunnel_parm p = { };
 445         int err;
 446
 447         tunnel_dev = __dev_get_by_name(net, "tunl0");
 448         if (!tunnel_dev)
 449                 goto out;
 450
 451         p.iph.daddr = v->vifc_rmt_addr.s_addr;
 452         p.iph.saddr = v->vifc_lcl_addr.s_addr;
 453         p.iph.version = 4;
 454         p.iph.ihl = 5;
 455         p.iph.protocol = IPPROTO_IPIP;
 456         sprintf(p.name, "dvmrp%d", v->vifc_vifi);
 457
 458         if (!tunnel_dev->netdev_ops->ndo_tunnel_ctl)
 459                 goto out;
 460         err = tunnel_dev->netdev_ops->ndo_tunnel_ctl(tunnel_dev, &p,
 461                         SIOCADDTUNNEL);
 462         if (err)
 463                 goto out;
 464
 465         new_dev = __dev_get_by_name(net, p.name);
 466         if (!new_dev)
 467                 goto out;
 468
 469         new_dev->flags |= IFF_MULTICAST;
 470         if (!ipmr_init_vif_indev(new_dev))
 471                 goto out_unregister;
 472         if (dev_open(new_dev, NULL))
 473                 goto out_unregister;
 474         dev_hold(new_dev);
 475         err = dev_set_allmulti(new_dev, 1);
 476         if (err) {
 477                 dev_close(new_dev);
 478                 tunnel_dev->netdev_ops->ndo_tunnel_ctl(tunnel_dev, &p,
 479                                 SIOCDELTUNNEL);
 480                 dev_put(new_dev);
 481                 new_dev = ERR_PTR(err);
 482         }
 483         return new_dev;
 484
 485 out_unregister:
 486         unregister_netdevice(new_dev);
 487 out:
 488         return ERR_PTR(-ENOBUFS);
 489 }
 490
 491 #if defined(CONFIG_IP_PIMSM_V1) || defined(CONFIG_IP_PIMSM_V2)
 492 static netdev_tx_t reg_vif_xmit(struct sk_buff *skb, struct net_device *dev)
 493 {
 494         struct net *net = dev_net(dev);
 495         struct mr_table *mrt;
 496         struct flowi4 fl4 = {
 497                 .flowi4_oif     = dev->ifindex,
 498                 .flowi4_iif     = skb->skb_iif ? : LOOPBACK_IFINDEX,
 499                 .flowi4_mark    = skb->mark,
 500         };
 501         int err;
 502
 503         err = ipmr_fib_lookup(net, &fl4, &mrt);
 504         if (err < 0) {
 505                 kfree_skb(skb);
 506                 return err;
 507         }
 508
 509         DEV_STATS_ADD(dev, tx_bytes, skb->len);
 510         DEV_STATS_INC(dev, tx_packets);
 511         rcu_read_lock();
 512
 513         /* Pairs with WRITE_ONCE() in vif_add() and vif_delete() */
 514         ipmr_cache_report(mrt, skb, READ_ONCE(mrt->mroute_reg_vif_num),
 515                           IGMPMSG_WHOLEPKT);
 516
 517         rcu_read_unlock();
 518         kfree_skb(skb);
 519         return NETDEV_TX_OK;
 520 }
 521
 522 static int reg_vif_get_iflink(const struct net_device *dev)
 523 {
 524         return 0;
 525 }
 526
 527 static const struct net_device_ops reg_vif_netdev_ops = {
 528         .ndo_start_xmit = reg_vif_xmit,
 529         .ndo_get_iflink = reg_vif_get_iflink,
 530 };
 531
 532 static void reg_vif_setup(struct net_device *dev)
 533 {
 534         dev->type               = ARPHRD_PIMREG;
 535         dev->mtu                = ETH_DATA_LEN - sizeof(struct iphdr) - 8;
 536         dev->flags              = IFF_NOARP;
 537         dev->netdev_ops         = &reg_vif_netdev_ops;
 538         dev->needs_free_netdev  = true;
 539         dev->features           |= NETIF_F_NETNS_LOCAL;
 540 }
 541
 542 static struct net_device *ipmr_reg_vif(struct net *net, struct mr_table *mrt)
 543 {
 544         struct net_device *dev;
 545         char name[IFNAMSIZ];
 546
 547         if (mrt->id == RT_TABLE_DEFAULT)
 548                 sprintf(name, "pimreg");
 549         else
 550                 sprintf(name, "pimreg%u", mrt->id);
 551
 552         dev = alloc_netdev(0, name, NET_NAME_UNKNOWN, reg_vif_setup);
 553
 554         if (!dev)
 555                 return NULL;
 556
 557         dev_net_set(dev, net);
 558
 559         if (register_netdevice(dev)) {
 560                 free_netdev(dev);
 561                 return NULL;
 562         }
 563
 564         if (!ipmr_init_vif_indev(dev))
 565                 goto failure;
 566         if (dev_open(dev, NULL))
 567                 goto failure;
 568
 569         dev_hold(dev);
 570
 571         return dev;
 572
 573 failure:
 574         unregister_netdevice(dev);
 575         return NULL;
 576 }
 577
 578 /* called with rcu_read_lock() */
 579 static int __pim_rcv(struct mr_table *mrt, struct sk_buff *skb,
 580                      unsigned int pimlen)
 581 {
 582         struct net_device *reg_dev = NULL;
 583         struct iphdr *encap;
 584         int vif_num;
 585
 586         encap = (struct iphdr *)(skb_transport_header(skb) + pimlen);
 587         /* Check that:
 588          * a. packet is really sent to a multicast group
 589          * b. packet is not a NULL-REGISTER
 590          * c. packet is not truncated
 591          */
 592         if (!ipv4_is_multicast(encap->daddr) ||
 593             encap->tot_len == 0 ||
 594             ntohs(encap->tot_len) + pimlen > skb->len)
 595                 return 1;
 596
 597         /* Pairs with WRITE_ONCE() in vif_add()/vid_delete() */
 598         vif_num = READ_ONCE(mrt->mroute_reg_vif_num);
 599         if (vif_num >= 0)
 600                 reg_dev = vif_dev_read(&mrt->vif_table[vif_num]);
 601         if (!reg_dev)
 602                 return 1;
 603
 604         skb->mac_header = skb->network_header;
 605         skb_pull(skb, (u8 *)encap - skb->data);
 606         skb_reset_network_header(skb);
 607         skb->protocol = htons(ETH_P_IP);
 608         skb->ip_summed = CHECKSUM_NONE;
 609
 610         skb_tunnel_rx(skb, reg_dev, dev_net(reg_dev));
 611
 612         netif_rx(skb);
 613
 614         return NET_RX_SUCCESS;
 615 }
 616 #else
 617 static struct net_device *ipmr_reg_vif(struct net *net, struct mr_table *mrt)
 618 {
 619         return NULL;
 620 }
 621 #endif
 622
 623 static int call_ipmr_vif_entry_notifiers(struct net *net,
 624                                          enum fib_event_type event_type,
 625                                          struct vif_device *vif,
 626                                          struct net_device *vif_dev,
 627                                          vifi_t vif_index, u32 tb_id)
 628 {
 629         return mr_call_vif_notifiers(net, RTNL_FAMILY_IPMR, event_type,
 630                                      vif, vif_dev, vif_index, tb_id,
 631                                      &net->ipv4.ipmr_seq);
 632 }
 633
 634 static int call_ipmr_mfc_entry_notifiers(struct net *net,
 635                                          enum fib_event_type event_type,
 636                                          struct mfc_cache *mfc, u32 tb_id)
 637 {
 638         return mr_call_mfc_notifiers(net, RTNL_FAMILY_IPMR, event_type,
 639                                      &mfc->_c, tb_id, &net->ipv4.ipmr_seq);
 640 }
 641
 642 /**
 643  *      vif_delete - Delete a VIF entry
 644  *      @mrt: Table to delete from
 645  *      @vifi: VIF identifier to delete
 646  *      @notify: Set to 1, if the caller is a notifier_call
 647  *      @head: if unregistering the VIF, place it on this queue
 648  */
 649 static int vif_delete(struct mr_table *mrt, int vifi, int notify,
 650                       struct list_head *head)
 651 {
 652         struct net *net = read_pnet(&mrt->net);
 653         struct vif_device *v;
 654         struct net_device *dev;
 655         struct in_device *in_dev;
 656
 657         if (vifi < 0 || vifi >= mrt->maxvif)
 658                 return -EADDRNOTAVAIL;
 659
 660         v = &mrt->vif_table[vifi];
 661
 662         dev = rtnl_dereference(v->dev);
 663         if (!dev)
 664                 return -EADDRNOTAVAIL;
 665
 666         spin_lock(&mrt_lock);
 667         call_ipmr_vif_entry_notifiers(net, FIB_EVENT_VIF_DEL, v, dev,
 668                                       vifi, mrt->id);
 669         RCU_INIT_POINTER(v->dev, NULL);
 670
 671         if (vifi == mrt->mroute_reg_vif_num) {
 672                 /* Pairs with READ_ONCE() in ipmr_cache_report() and reg_vif_xmit() */
 673                 WRITE_ONCE(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                 WRITE_ONCE(mrt->maxvif, tmp + 1);
 683         }
 684
 685         spin_unlock(&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         netdev_put(dev, &v->dev_tracker);
 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 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         spin_lock(&mrt_lock);
 900         rcu_assign_pointer(v->dev, dev);
 901         netdev_tracker_alloc(dev, &v->dev_tracker, GFP_ATOMIC);
 902         if (v->flags & VIFF_REGISTER) {
 903                 /* Pairs with READ_ONCE() in ipmr_cache_report() and reg_vif_xmit() */
 904                 WRITE_ONCE(mrt->mroute_reg_vif_num, vifi);
 905         }
 906         if (vifi+1 > mrt->maxvif)
 907                 WRITE_ONCE(mrt->maxvif, vifi + 1);
 908         spin_unlock(&mrt_lock);
 909         call_ipmr_vif_entry_notifiers(net, FIB_EVENT_VIF_ADD, v, dev,
 910                                       vifi, mrt->id);
 911         return 0;
 912 }
 913
 914 /* called with rcu_read_lock() */
 915 static struct mfc_cache *ipmr_cache_find(struct mr_table *mrt,
 916                                          __be32 origin,
 917                                          __be32 mcastgrp)
 918 {
 919         struct mfc_cache_cmp_arg arg = {
 920                         .mfc_mcastgrp = mcastgrp,
 921                         .mfc_origin = origin
 922         };
 923
 924         return mr_mfc_find(mrt, &arg);
 925 }
 926
 927 /* Look for a (*,G) entry */
 928 static struct mfc_cache *ipmr_cache_find_any(struct mr_table *mrt,
 929                                              __be32 mcastgrp, int vifi)
 930 {
 931         struct mfc_cache_cmp_arg arg = {
 932                         .mfc_mcastgrp = mcastgrp,
 933                         .mfc_origin = htonl(INADDR_ANY)
 934         };
 935
 936         if (mcastgrp == htonl(INADDR_ANY))
 937                 return mr_mfc_find_any_parent(mrt, vifi);
 938         return mr_mfc_find_any(mrt, vifi, &arg);
 939 }
 940
 941 /* Look for a (S,G,iif) entry if parent != -1 */
 942 static struct mfc_cache *ipmr_cache_find_parent(struct mr_table *mrt,
 943                                                 __be32 origin, __be32 mcastgrp,
 944                                                 int parent)
 945 {
 946         struct mfc_cache_cmp_arg arg = {
 947                         .mfc_mcastgrp = mcastgrp,
 948                         .mfc_origin = origin,
 949         };
 950
 951         return mr_mfc_find_parent(mrt, &arg, parent);
 952 }
 953
 954 /* Allocate a multicast cache entry */
 955 static struct mfc_cache *ipmr_cache_alloc(void)
 956 {
 957         struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_KERNEL);
 958
 959         if (c) {
 960                 c->_c.mfc_un.res.last_assert = jiffies - MFC_ASSERT_THRESH - 1;
 961                 c->_c.mfc_un.res.minvif = MAXVIFS;
 962                 c->_c.free = ipmr_cache_free_rcu;
 963                 refcount_set(&c->_c.mfc_un.res.refcount, 1);
 964         }
 965         return c;
 966 }
 967
 968 static struct mfc_cache *ipmr_cache_alloc_unres(void)
 969 {
 970         struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_ATOMIC);
 971
 972         if (c) {
 973                 skb_queue_head_init(&c->_c.mfc_un.unres.unresolved);
 974                 c->_c.mfc_un.unres.expires = jiffies + 10 * HZ;
 975         }
 976         return c;
 977 }
 978
 979 /* A cache entry has gone into a resolved state from queued */
 980 static void ipmr_cache_resolve(struct net *net, struct mr_table *mrt,
 981                                struct mfc_cache *uc, struct mfc_cache *c)
 982 {
 983         struct sk_buff *skb;
 984         struct nlmsgerr *e;
 985
 986         /* Play the pending entries through our router */
 987         while ((skb = __skb_dequeue(&uc->_c.mfc_un.unres.unresolved))) {
 988                 if (ip_hdr(skb)->version == 0) {
 989                         struct nlmsghdr *nlh = skb_pull(skb,
 990                                                         sizeof(struct iphdr));
 991
 992                         if (mr_fill_mroute(mrt, skb, &c->_c,
 993                                            nlmsg_data(nlh)) > 0) {
 994                                 nlh->nlmsg_len = skb_tail_pointer(skb) -
 995                                                  (u8 *)nlh;
 996                         } else {
 997                                 nlh->nlmsg_type = NLMSG_ERROR;
 998                                 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
 999                                 skb_trim(skb, nlh->nlmsg_len);
1000                                 e = nlmsg_data(nlh);
1001                                 e->error = -EMSGSIZE;
1002                                 memset(&e->msg, 0, sizeof(e->msg));
1003                         }
1004
1005                         rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
1006                 } else {
1007                         rcu_read_lock();
1008                         ip_mr_forward(net, mrt, skb->dev, skb, c, 0);
1009                         rcu_read_unlock();
1010                 }
1011         }
1012 }
1013
1014 /* Bounce a cache query up to mrouted and netlink.
1015  *
1016  * Called under rcu_read_lock().
1017  */
1018 static int ipmr_cache_report(const struct mr_table *mrt,
1019                              struct sk_buff *pkt, vifi_t vifi, int assert)
1020 {
1021         const int ihl = ip_hdrlen(pkt);
1022         struct sock *mroute_sk;
1023         struct igmphdr *igmp;
1024         struct igmpmsg *msg;
1025         struct sk_buff *skb;
1026         int ret;
1027
1028         mroute_sk = rcu_dereference(mrt->mroute_sk);
1029         if (!mroute_sk)
1030                 return -EINVAL;
1031
1032         if (assert == IGMPMSG_WHOLEPKT || assert == IGMPMSG_WRVIFWHOLE)
1033                 skb = skb_realloc_headroom(pkt, sizeof(struct iphdr));
1034         else
1035                 skb = alloc_skb(128, GFP_ATOMIC);
1036
1037         if (!skb)
1038                 return -ENOBUFS;
1039
1040         if (assert == IGMPMSG_WHOLEPKT || assert == IGMPMSG_WRVIFWHOLE) {
1041                 /* Ugly, but we have no choice with this interface.
1042                  * Duplicate old header, fix ihl, length etc.
1043                  * And all this only to mangle msg->im_msgtype and
1044                  * to set msg->im_mbz to "mbz" :-)
1045                  */
1046                 skb_push(skb, sizeof(struct iphdr));
1047                 skb_reset_network_header(skb);
1048                 skb_reset_transport_header(skb);
1049                 msg = (struct igmpmsg *)skb_network_header(skb);
1050                 memcpy(msg, skb_network_header(pkt), sizeof(struct iphdr));
1051                 msg->im_msgtype = assert;
1052                 msg->im_mbz = 0;
1053                 if (assert == IGMPMSG_WRVIFWHOLE) {
1054                         msg->im_vif = vifi;
1055                         msg->im_vif_hi = vifi >> 8;
1056                 } else {
1057                         /* Pairs with WRITE_ONCE() in vif_add() and vif_delete() */
1058                         int vif_num = READ_ONCE(mrt->mroute_reg_vif_num);
1059
1060                         msg->im_vif = vif_num;
1061                         msg->im_vif_hi = vif_num >> 8;
1062                 }
1063                 ip_hdr(skb)->ihl = sizeof(struct iphdr) >> 2;
1064                 ip_hdr(skb)->tot_len = htons(ntohs(ip_hdr(pkt)->tot_len) +
1065                                              sizeof(struct iphdr));
1066         } else {
1067                 /* Copy the IP header */
1068                 skb_set_network_header(skb, skb->len);
1069                 skb_put(skb, ihl);
1070                 skb_copy_to_linear_data(skb, pkt->data, ihl);
1071                 /* Flag to the kernel this is a route add */
1072                 ip_hdr(skb)->protocol = 0;
1073                 msg = (struct igmpmsg *)skb_network_header(skb);
1074                 msg->im_vif = vifi;
1075                 msg->im_vif_hi = vifi >> 8;
1076                 ipv4_pktinfo_prepare(mroute_sk, pkt);
1077                 memcpy(skb->cb, pkt->cb, sizeof(skb->cb));
1078                 /* Add our header */
1079                 igmp = skb_put(skb, sizeof(struct igmphdr));
1080                 igmp->type = assert;
1081                 msg->im_msgtype = assert;
1082                 igmp->code = 0;
1083                 ip_hdr(skb)->tot_len = htons(skb->len); /* Fix the length */
1084                 skb->transport_header = skb->network_header;
1085         }
1086
1087         igmpmsg_netlink_event(mrt, skb);
1088
1089         /* Deliver to mrouted */
1090         ret = sock_queue_rcv_skb(mroute_sk, skb);
1091
1092         if (ret < 0) {
1093                 net_warn_ratelimited("mroute: pending queue full, dropping entries\n");
1094                 kfree_skb(skb);
1095         }
1096
1097         return ret;
1098 }
1099
1100 /* Queue a packet for resolution. It gets locked cache entry! */
1101 /* Called under rcu_read_lock() */
1102 static int ipmr_cache_unresolved(struct mr_table *mrt, vifi_t vifi,
1103                                  struct sk_buff *skb, struct net_device *dev)
1104 {
1105         const struct iphdr *iph = ip_hdr(skb);
1106         struct mfc_cache *c;
1107         bool found = false;
1108         int err;
1109
1110         spin_lock_bh(&mfc_unres_lock);
1111         list_for_each_entry(c, &mrt->mfc_unres_queue, _c.list) {
1112                 if (c->mfc_mcastgrp == iph->daddr &&
1113                     c->mfc_origin == iph->saddr) {
1114                         found = true;
1115                         break;
1116                 }
1117         }
1118
1119         if (!found) {
1120                 /* Create a new entry if allowable */
1121                 c = ipmr_cache_alloc_unres();
1122                 if (!c) {
1123                         spin_unlock_bh(&mfc_unres_lock);
1124
1125                         kfree_skb(skb);
1126                         return -ENOBUFS;
1127                 }
1128
1129                 /* Fill in the new cache entry */
1130                 c->_c.mfc_parent = -1;
1131                 c->mfc_origin   = iph->saddr;
1132                 c->mfc_mcastgrp = iph->daddr;
1133
1134                 /* Reflect first query at mrouted. */
1135                 err = ipmr_cache_report(mrt, skb, vifi, IGMPMSG_NOCACHE);
1136
1137                 if (err < 0) {
1138                         /* If the report failed throw the cache entry
1139                            out - Brad Parker
1140                          */
1141                         spin_unlock_bh(&mfc_unres_lock);
1142
1143                         ipmr_cache_free(c);
1144                         kfree_skb(skb);
1145                         return err;
1146                 }
1147
1148                 atomic_inc(&mrt->cache_resolve_queue_len);
1149                 list_add(&c->_c.list, &mrt->mfc_unres_queue);
1150                 mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1151
1152                 if (atomic_read(&mrt->cache_resolve_queue_len) == 1)
1153                         mod_timer(&mrt->ipmr_expire_timer,
1154                                   c->_c.mfc_un.unres.expires);
1155         }
1156
1157         /* See if we can append the packet */
1158         if (c->_c.mfc_un.unres.unresolved.qlen > 3) {
1159                 kfree_skb(skb);
1160                 err = -ENOBUFS;
1161         } else {
1162                 if (dev) {
1163                         skb->dev = dev;
1164                         skb->skb_iif = dev->ifindex;
1165                 }
1166                 skb_queue_tail(&c->_c.mfc_un.unres.unresolved, skb);
1167                 err = 0;
1168         }
1169
1170         spin_unlock_bh(&mfc_unres_lock);
1171         return err;
1172 }
1173
1174 /* MFC cache manipulation by user space mroute daemon */
1175
1176 static int ipmr_mfc_delete(struct mr_table *mrt, struct mfcctl *mfc, int parent)
1177 {
1178         struct net *net = read_pnet(&mrt->net);
1179         struct mfc_cache *c;
1180
1181         /* The entries are added/deleted only under RTNL */
1182         rcu_read_lock();
1183         c = ipmr_cache_find_parent(mrt, mfc->mfcc_origin.s_addr,
1184                                    mfc->mfcc_mcastgrp.s_addr, parent);
1185         rcu_read_unlock();
1186         if (!c)
1187                 return -ENOENT;
1188         rhltable_remove(&mrt->mfc_hash, &c->_c.mnode, ipmr_rht_params);
1189         list_del_rcu(&c->_c.list);
1190         call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_DEL, c, mrt->id);
1191         mroute_netlink_event(mrt, c, RTM_DELROUTE);
1192         mr_cache_put(&c->_c);
1193
1194         return 0;
1195 }
1196
1197 static int ipmr_mfc_add(struct net *net, struct mr_table *mrt,
1198                         struct mfcctl *mfc, int mrtsock, int parent)
1199 {
1200         struct mfc_cache *uc, *c;
1201         struct mr_mfc *_uc;
1202         bool found;
1203         int ret;
1204
1205         if (mfc->mfcc_parent >= MAXVIFS)
1206                 return -ENFILE;
1207
1208         /* The entries are added/deleted only under RTNL */
1209         rcu_read_lock();
1210         c = ipmr_cache_find_parent(mrt, mfc->mfcc_origin.s_addr,
1211                                    mfc->mfcc_mcastgrp.s_addr, parent);
1212         rcu_read_unlock();
1213         if (c) {
1214                 spin_lock(&mrt_lock);
1215                 c->_c.mfc_parent = mfc->mfcc_parent;
1216                 ipmr_update_thresholds(mrt, &c->_c, mfc->mfcc_ttls);
1217                 if (!mrtsock)
1218                         c->_c.mfc_flags |= MFC_STATIC;
1219                 spin_unlock(&mrt_lock);
1220                 call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_REPLACE, c,
1221                                               mrt->id);
1222                 mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1223                 return 0;
1224         }
1225
1226         if (mfc->mfcc_mcastgrp.s_addr != htonl(INADDR_ANY) &&
1227             !ipv4_is_multicast(mfc->mfcc_mcastgrp.s_addr))
1228                 return -EINVAL;
1229
1230         c = ipmr_cache_alloc();
1231         if (!c)
1232                 return -ENOMEM;
1233
1234         c->mfc_origin = mfc->mfcc_origin.s_addr;
1235         c->mfc_mcastgrp = mfc->mfcc_mcastgrp.s_addr;
1236         c->_c.mfc_parent = mfc->mfcc_parent;
1237         ipmr_update_thresholds(mrt, &c->_c, mfc->mfcc_ttls);
1238         if (!mrtsock)
1239                 c->_c.mfc_flags |= MFC_STATIC;
1240
1241         ret = rhltable_insert_key(&mrt->mfc_hash, &c->cmparg, &c->_c.mnode,
1242                                   ipmr_rht_params);
1243         if (ret) {
1244                 pr_err("ipmr: rhtable insert error %d\n", ret);
1245                 ipmr_cache_free(c);
1246                 return ret;
1247         }
1248         list_add_tail_rcu(&c->_c.list, &mrt->mfc_cache_list);
1249         /* Check to see if we resolved a queued list. If so we
1250          * need to send on the frames and tidy up.
1251          */
1252         found = false;
1253         spin_lock_bh(&mfc_unres_lock);
1254         list_for_each_entry(_uc, &mrt->mfc_unres_queue, list) {
1255                 uc = (struct mfc_cache *)_uc;
1256                 if (uc->mfc_origin == c->mfc_origin &&
1257                     uc->mfc_mcastgrp == c->mfc_mcastgrp) {
1258                         list_del(&_uc->list);
1259                         atomic_dec(&mrt->cache_resolve_queue_len);
1260                         found = true;
1261                         break;
1262                 }
1263         }
1264         if (list_empty(&mrt->mfc_unres_queue))
1265                 del_timer(&mrt->ipmr_expire_timer);
1266         spin_unlock_bh(&mfc_unres_lock);
1267
1268         if (found) {
1269                 ipmr_cache_resolve(net, mrt, uc, c);
1270                 ipmr_cache_free(uc);
1271         }
1272         call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_ADD, c, mrt->id);
1273         mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1274         return 0;
1275 }
1276
1277 /* Close the multicast socket, and clear the vif tables etc */
1278 static void mroute_clean_tables(struct mr_table *mrt, int flags)
1279 {
1280         struct net *net = read_pnet(&mrt->net);
1281         struct mr_mfc *c, *tmp;
1282         struct mfc_cache *cache;
1283         LIST_HEAD(list);
1284         int i;
1285
1286         /* Shut down all active vif entries */
1287         if (flags & (MRT_FLUSH_VIFS | MRT_FLUSH_VIFS_STATIC)) {
1288                 for (i = 0; i < mrt->maxvif; i++) {
1289                         if (((mrt->vif_table[i].flags & VIFF_STATIC) &&
1290                              !(flags & MRT_FLUSH_VIFS_STATIC)) ||
1291                             (!(mrt->vif_table[i].flags & VIFF_STATIC) && !(flags & MRT_FLUSH_VIFS)))
1292                                 continue;
1293                         vif_delete(mrt, i, 0, &list);
1294                 }
1295                 unregister_netdevice_many(&list);
1296         }
1297
1298         /* Wipe the cache */
1299         if (flags & (MRT_FLUSH_MFC | MRT_FLUSH_MFC_STATIC)) {
1300                 list_for_each_entry_safe(c, tmp, &mrt->mfc_cache_list, list) {
1301                         if (((c->mfc_flags & MFC_STATIC) && !(flags & MRT_FLUSH_MFC_STATIC)) ||
1302                             (!(c->mfc_flags & MFC_STATIC) && !(flags & MRT_FLUSH_MFC)))
1303                                 continue;
1304                         rhltable_remove(&mrt->mfc_hash, &c->mnode, ipmr_rht_params);
1305                         list_del_rcu(&c->list);
1306                         cache = (struct mfc_cache *)c;
1307                         call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_DEL, cache,
1308                                                       mrt->id);
1309                         mroute_netlink_event(mrt, cache, RTM_DELROUTE);
1310                         mr_cache_put(c);
1311                 }
1312         }
1313
1314         if (flags & MRT_FLUSH_MFC) {
1315                 if (atomic_read(&mrt->cache_resolve_queue_len) != 0) {
1316                         spin_lock_bh(&mfc_unres_lock);
1317                         list_for_each_entry_safe(c, tmp, &mrt->mfc_unres_queue, list) {
1318                                 list_del(&c->list);
1319                                 cache = (struct mfc_cache *)c;
1320                                 mroute_netlink_event(mrt, cache, RTM_DELROUTE);
1321                                 ipmr_destroy_unres(mrt, cache);
1322                         }
1323                         spin_unlock_bh(&mfc_unres_lock);
1324                 }
1325         }
1326 }
1327
1328 /* called from ip_ra_control(), before an RCU grace period,
1329  * we don't need to call synchronize_rcu() here
1330  */
1331 static void mrtsock_destruct(struct sock *sk)
1332 {
1333         struct net *net = sock_net(sk);
1334         struct mr_table *mrt;
1335
1336         rtnl_lock();
1337         ipmr_for_each_table(mrt, net) {
1338                 if (sk == rtnl_dereference(mrt->mroute_sk)) {
1339                         IPV4_DEVCONF_ALL(net, MC_FORWARDING)--;
1340                         inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
1341                                                     NETCONFA_MC_FORWARDING,
1342                                                     NETCONFA_IFINDEX_ALL,
1343                                                     net->ipv4.devconf_all);
1344                         RCU_INIT_POINTER(mrt->mroute_sk, NULL);
1345                         mroute_clean_tables(mrt, MRT_FLUSH_VIFS | MRT_FLUSH_MFC);
1346                 }
1347         }
1348         rtnl_unlock();
1349 }
1350
1351 /* Socket options and virtual interface manipulation. The whole
1352  * virtual interface system is a complete heap, but unfortunately
1353  * that's how BSD mrouted happens to think. Maybe one day with a proper
1354  * MOSPF/PIM router set up we can clean this up.
1355  */
1356
1357 int ip_mroute_setsockopt(struct sock *sk, int optname, sockptr_t optval,
1358                          unsigned int optlen)
1359 {
1360         struct net *net = sock_net(sk);
1361         int val, ret = 0, parent = 0;
1362         struct mr_table *mrt;
1363         struct vifctl vif;
1364         struct mfcctl mfc;
1365         bool do_wrvifwhole;
1366         u32 uval;
1367
1368         /* There's one exception to the lock - MRT_DONE which needs to unlock */
1369         rtnl_lock();
1370         if (sk->sk_type != SOCK_RAW ||
1371             inet_sk(sk)->inet_num != IPPROTO_IGMP) {
1372                 ret = -EOPNOTSUPP;
1373                 goto out_unlock;
1374         }
1375
1376         mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1377         if (!mrt) {
1378                 ret = -ENOENT;
1379                 goto out_unlock;
1380         }
1381         if (optname != MRT_INIT) {
1382                 if (sk != rcu_access_pointer(mrt->mroute_sk) &&
1383                     !ns_capable(net->user_ns, CAP_NET_ADMIN)) {
1384                         ret = -EACCES;
1385                         goto out_unlock;
1386                 }
1387         }
1388
1389         switch (optname) {
1390         case MRT_INIT:
1391                 if (optlen != sizeof(int)) {
1392                         ret = -EINVAL;
1393                         break;
1394                 }
1395                 if (rtnl_dereference(mrt->mroute_sk)) {
1396                         ret = -EADDRINUSE;
1397                         break;
1398                 }
1399
1400                 ret = ip_ra_control(sk, 1, mrtsock_destruct);
1401                 if (ret == 0) {
1402                         rcu_assign_pointer(mrt->mroute_sk, sk);
1403                         IPV4_DEVCONF_ALL(net, MC_FORWARDING)++;
1404                         inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
1405                                                     NETCONFA_MC_FORWARDING,
1406                                                     NETCONFA_IFINDEX_ALL,
1407                                                     net->ipv4.devconf_all);
1408                 }
1409                 break;
1410         case MRT_DONE:
1411                 if (sk != rcu_access_pointer(mrt->mroute_sk)) {
1412                         ret = -EACCES;
1413                 } else {
1414                         /* We need to unlock here because mrtsock_destruct takes
1415                          * care of rtnl itself and we can't change that due to
1416                          * the IP_ROUTER_ALERT setsockopt which runs without it.
1417                          */
1418                         rtnl_unlock();
1419                         ret = ip_ra_control(sk, 0, NULL);
1420                         goto out;
1421                 }
1422                 break;
1423         case MRT_ADD_VIF:
1424         case MRT_DEL_VIF:
1425                 if (optlen != sizeof(vif)) {
1426                         ret = -EINVAL;
1427                         break;
1428                 }
1429                 if (copy_from_sockptr(&vif, optval, sizeof(vif))) {
1430                         ret = -EFAULT;
1431                         break;
1432                 }
1433                 if (vif.vifc_vifi >= MAXVIFS) {
1434                         ret = -ENFILE;
1435                         break;
1436                 }
1437                 if (optname == MRT_ADD_VIF) {
1438                         ret = vif_add(net, mrt, &vif,
1439                                       sk == rtnl_dereference(mrt->mroute_sk));
1440                 } else {
1441                         ret = vif_delete(mrt, vif.vifc_vifi, 0, NULL);
1442                 }
1443                 break;
1444         /* Manipulate the forwarding caches. These live
1445          * in a sort of kernel/user symbiosis.
1446          */
1447         case MRT_ADD_MFC:
1448         case MRT_DEL_MFC:
1449                 parent = -1;
1450                 fallthrough;
1451         case MRT_ADD_MFC_PROXY:
1452         case MRT_DEL_MFC_PROXY:
1453                 if (optlen != sizeof(mfc)) {
1454                         ret = -EINVAL;
1455                         break;
1456                 }
1457                 if (copy_from_sockptr(&mfc, optval, sizeof(mfc))) {
1458                         ret = -EFAULT;
1459                         break;
1460                 }
1461                 if (parent == 0)
1462                         parent = mfc.mfcc_parent;
1463                 if (optname == MRT_DEL_MFC || optname == MRT_DEL_MFC_PROXY)
1464                         ret = ipmr_mfc_delete(mrt, &mfc, parent);
1465                 else
1466                         ret = ipmr_mfc_add(net, mrt, &mfc,
1467                                            sk == rtnl_dereference(mrt->mroute_sk),
1468                                            parent);
1469                 break;
1470         case MRT_FLUSH:
1471                 if (optlen != sizeof(val)) {
1472                         ret = -EINVAL;
1473                         break;
1474                 }
1475                 if (copy_from_sockptr(&val, optval, sizeof(val))) {
1476                         ret = -EFAULT;
1477                         break;
1478                 }
1479                 mroute_clean_tables(mrt, val);
1480                 break;
1481         /* Control PIM assert. */
1482         case MRT_ASSERT:
1483                 if (optlen != sizeof(val)) {
1484                         ret = -EINVAL;
1485                         break;
1486                 }
1487                 if (copy_from_sockptr(&val, optval, sizeof(val))) {
1488                         ret = -EFAULT;
1489                         break;
1490                 }
1491                 mrt->mroute_do_assert = val;
1492                 break;
1493         case MRT_PIM:
1494                 if (!ipmr_pimsm_enabled()) {
1495                         ret = -ENOPROTOOPT;
1496                         break;
1497                 }
1498                 if (optlen != sizeof(val)) {
1499                         ret = -EINVAL;
1500                         break;
1501                 }
1502                 if (copy_from_sockptr(&val, optval, sizeof(val))) {
1503                         ret = -EFAULT;
1504                         break;
1505                 }
1506
1507                 do_wrvifwhole = (val == IGMPMSG_WRVIFWHOLE);
1508                 val = !!val;
1509                 if (val != mrt->mroute_do_pim) {
1510                         mrt->mroute_do_pim = val;
1511                         mrt->mroute_do_assert = val;
1512                         mrt->mroute_do_wrvifwhole = do_wrvifwhole;
1513                 }
1514                 break;
1515         case MRT_TABLE:
1516                 if (!IS_BUILTIN(CONFIG_IP_MROUTE_MULTIPLE_TABLES)) {
1517                         ret = -ENOPROTOOPT;
1518                         break;
1519                 }
1520                 if (optlen != sizeof(uval)) {
1521                         ret = -EINVAL;
1522                         break;
1523                 }
1524                 if (copy_from_sockptr(&uval, optval, sizeof(uval))) {
1525                         ret = -EFAULT;
1526                         break;
1527                 }
1528
1529                 if (sk == rtnl_dereference(mrt->mroute_sk)) {
1530                         ret = -EBUSY;
1531                 } else {
1532                         mrt = ipmr_new_table(net, uval);
1533                         if (IS_ERR(mrt))
1534                                 ret = PTR_ERR(mrt);
1535                         else
1536                                 raw_sk(sk)->ipmr_table = uval;
1537                 }
1538                 break;
1539         /* Spurious command, or MRT_VERSION which you cannot set. */
1540         default:
1541                 ret = -ENOPROTOOPT;
1542         }
1543 out_unlock:
1544         rtnl_unlock();
1545 out:
1546         return ret;
1547 }
1548
1549 /* Execute if this ioctl is a special mroute ioctl */
1550 int ipmr_sk_ioctl(struct sock *sk, unsigned int cmd, void __user *arg)
1551 {
1552         switch (cmd) {
1553         /* These userspace buffers will be consumed by ipmr_ioctl() */
1554         case SIOCGETVIFCNT: {
1555                 struct sioc_vif_req buffer;
1556
1557                 return sock_ioctl_inout(sk, cmd, arg, &buffer,
1558                                       sizeof(buffer));
1559                 }
1560         case SIOCGETSGCNT: {
1561                 struct sioc_sg_req buffer;
1562
1563                 return sock_ioctl_inout(sk, cmd, arg, &buffer,
1564                                       sizeof(buffer));
1565                 }
1566         }
1567         /* return code > 0 means that the ioctl was not executed */
1568         return 1;
1569 }
1570
1571 /* Getsock opt support for the multicast routing system. */
1572 int ip_mroute_getsockopt(struct sock *sk, int optname, sockptr_t optval,
1573                          sockptr_t optlen)
1574 {
1575         int olr;
1576         int val;
1577         struct net *net = sock_net(sk);
1578         struct mr_table *mrt;
1579
1580         if (sk->sk_type != SOCK_RAW ||
1581             inet_sk(sk)->inet_num != IPPROTO_IGMP)
1582                 return -EOPNOTSUPP;
1583
1584         mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1585         if (!mrt)
1586                 return -ENOENT;
1587
1588         switch (optname) {
1589         case MRT_VERSION:
1590                 val = 0x0305;
1591                 break;
1592         case MRT_PIM:
1593                 if (!ipmr_pimsm_enabled())
1594                         return -ENOPROTOOPT;
1595                 val = mrt->mroute_do_pim;
1596                 break;
1597         case MRT_ASSERT:
1598                 val = mrt->mroute_do_assert;
1599                 break;
1600         default:
1601                 return -ENOPROTOOPT;
1602         }
1603
1604         if (copy_from_sockptr(&olr, optlen, sizeof(int)))
1605                 return -EFAULT;
1606         olr = min_t(unsigned int, olr, sizeof(int));
1607         if (olr < 0)
1608                 return -EINVAL;
1609         if (copy_to_sockptr(optlen, &olr, sizeof(int)))
1610                 return -EFAULT;
1611         if (copy_to_sockptr(optval, &val, olr))
1612                 return -EFAULT;
1613         return 0;
1614 }
1615
1616 /* The IP multicast ioctl support routines. */
1617 int ipmr_ioctl(struct sock *sk, int cmd, void *arg)
1618 {
1619         struct vif_device *vif;
1620         struct mfc_cache *c;
1621         struct net *net = sock_net(sk);
1622         struct sioc_vif_req *vr;
1623         struct sioc_sg_req *sr;
1624         struct mr_table *mrt;
1625
1626         mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1627         if (!mrt)
1628                 return -ENOENT;
1629
1630         switch (cmd) {
1631         case SIOCGETVIFCNT:
1632                 vr = (struct sioc_vif_req *)arg;
1633                 if (vr->vifi >= mrt->maxvif)
1634                         return -EINVAL;
1635                 vr->vifi = array_index_nospec(vr->vifi, mrt->maxvif);
1636                 rcu_read_lock();
1637                 vif = &mrt->vif_table[vr->vifi];
1638                 if (VIF_EXISTS(mrt, vr->vifi)) {
1639                         vr->icount = READ_ONCE(vif->pkt_in);
1640                         vr->ocount = READ_ONCE(vif->pkt_out);
1641                         vr->ibytes = READ_ONCE(vif->bytes_in);
1642                         vr->obytes = READ_ONCE(vif->bytes_out);
1643                         rcu_read_unlock();
1644
1645                         return 0;
1646                 }
1647                 rcu_read_unlock();
1648                 return -EADDRNOTAVAIL;
1649         case SIOCGETSGCNT:
1650                 sr = (struct sioc_sg_req *)arg;
1651
1652                 rcu_read_lock();
1653                 c = ipmr_cache_find(mrt, sr->src.s_addr, sr->grp.s_addr);
1654                 if (c) {
1655                         sr->pktcnt = c->_c.mfc_un.res.pkt;
1656                         sr->bytecnt = c->_c.mfc_un.res.bytes;
1657                         sr->wrong_if = c->_c.mfc_un.res.wrong_if;
1658                         rcu_read_unlock();
1659                         return 0;
1660                 }
1661                 rcu_read_unlock();
1662                 return -EADDRNOTAVAIL;
1663         default:
1664                 return -ENOIOCTLCMD;
1665         }
1666 }
1667
1668 #ifdef CONFIG_COMPAT
1669 struct compat_sioc_sg_req {
1670         struct in_addr src;
1671         struct in_addr grp;
1672         compat_ulong_t pktcnt;
1673         compat_ulong_t bytecnt;
1674         compat_ulong_t wrong_if;
1675 };
1676
1677 struct compat_sioc_vif_req {
1678         vifi_t  vifi;           /* Which iface */
1679         compat_ulong_t icount;
1680         compat_ulong_t ocount;
1681         compat_ulong_t ibytes;
1682         compat_ulong_t obytes;
1683 };
1684
1685 int ipmr_compat_ioctl(struct sock *sk, unsigned int cmd, void __user *arg)
1686 {
1687         struct compat_sioc_sg_req sr;
1688         struct compat_sioc_vif_req vr;
1689         struct vif_device *vif;
1690         struct mfc_cache *c;
1691         struct net *net = sock_net(sk);
1692         struct mr_table *mrt;
1693
1694         mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1695         if (!mrt)
1696                 return -ENOENT;
1697
1698         switch (cmd) {
1699         case SIOCGETVIFCNT:
1700                 if (copy_from_user(&vr, arg, sizeof(vr)))
1701                         return -EFAULT;
1702                 if (vr.vifi >= mrt->maxvif)
1703                         return -EINVAL;
1704                 vr.vifi = array_index_nospec(vr.vifi, mrt->maxvif);
1705                 rcu_read_lock();
1706                 vif = &mrt->vif_table[vr.vifi];
1707                 if (VIF_EXISTS(mrt, vr.vifi)) {
1708                         vr.icount = READ_ONCE(vif->pkt_in);
1709                         vr.ocount = READ_ONCE(vif->pkt_out);
1710                         vr.ibytes = READ_ONCE(vif->bytes_in);
1711                         vr.obytes = READ_ONCE(vif->bytes_out);
1712                         rcu_read_unlock();
1713
1714                         if (copy_to_user(arg, &vr, sizeof(vr)))
1715                                 return -EFAULT;
1716                         return 0;
1717                 }
1718                 rcu_read_unlock();
1719                 return -EADDRNOTAVAIL;
1720         case SIOCGETSGCNT:
1721                 if (copy_from_user(&sr, arg, sizeof(sr)))
1722                         return -EFAULT;
1723
1724                 rcu_read_lock();
1725                 c = ipmr_cache_find(mrt, sr.src.s_addr, sr.grp.s_addr);
1726                 if (c) {
1727                         sr.pktcnt = c->_c.mfc_un.res.pkt;
1728                         sr.bytecnt = c->_c.mfc_un.res.bytes;
1729                         sr.wrong_if = c->_c.mfc_un.res.wrong_if;
1730                         rcu_read_unlock();
1731
1732                         if (copy_to_user(arg, &sr, sizeof(sr)))
1733                                 return -EFAULT;
1734                         return 0;
1735                 }
1736                 rcu_read_unlock();
1737                 return -EADDRNOTAVAIL;
1738         default:
1739                 return -ENOIOCTLCMD;
1740         }
1741 }
1742 #endif
1743
1744 static int ipmr_device_event(struct notifier_block *this, unsigned long event, void *ptr)
1745 {
1746         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1747         struct net *net = dev_net(dev);
1748         struct mr_table *mrt;
1749         struct vif_device *v;
1750         int ct;
1751
1752         if (event != NETDEV_UNREGISTER)
1753                 return NOTIFY_DONE;
1754
1755         ipmr_for_each_table(mrt, net) {
1756                 v = &mrt->vif_table[0];
1757                 for (ct = 0; ct < mrt->maxvif; ct++, v++) {
1758                         if (rcu_access_pointer(v->dev) == dev)
1759                                 vif_delete(mrt, ct, 1, NULL);
1760                 }
1761         }
1762         return NOTIFY_DONE;
1763 }
1764
1765 static struct notifier_block ip_mr_notifier = {
1766         .notifier_call = ipmr_device_event,
1767 };
1768
1769 /* Encapsulate a packet by attaching a valid IPIP header to it.
1770  * This avoids tunnel drivers and other mess and gives us the speed so
1771  * important for multicast video.
1772  */
1773 static void ip_encap(struct net *net, struct sk_buff *skb,
1774                      __be32 saddr, __be32 daddr)
1775 {
1776         struct iphdr *iph;
1777         const struct iphdr *old_iph = ip_hdr(skb);
1778
1779         skb_push(skb, sizeof(struct iphdr));
1780         skb->transport_header = skb->network_header;
1781         skb_reset_network_header(skb);
1782         iph = ip_hdr(skb);
1783
1784         iph->version    =       4;
1785         iph->tos        =       old_iph->tos;
1786         iph->ttl        =       old_iph->ttl;
1787         iph->frag_off   =       0;
1788         iph->daddr      =       daddr;
1789         iph->saddr      =       saddr;
1790         iph->protocol   =       IPPROTO_IPIP;
1791         iph->ihl        =       5;
1792         iph->tot_len    =       htons(skb->len);
1793         ip_select_ident(net, skb, NULL);
1794         ip_send_check(iph);
1795
1796         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1797         nf_reset_ct(skb);
1798 }
1799
1800 static inline int ipmr_forward_finish(struct net *net, struct sock *sk,
1801                                       struct sk_buff *skb)
1802 {
1803         struct ip_options *opt = &(IPCB(skb)->opt);
1804
1805         IP_INC_STATS(net, IPSTATS_MIB_OUTFORWDATAGRAMS);
1806
1807         if (unlikely(opt->optlen))
1808                 ip_forward_options(skb);
1809
1810         return dst_output(net, sk, skb);
1811 }
1812
1813 #ifdef CONFIG_NET_SWITCHDEV
1814 static bool ipmr_forward_offloaded(struct sk_buff *skb, struct mr_table *mrt,
1815                                    int in_vifi, int out_vifi)
1816 {
1817         struct vif_device *out_vif = &mrt->vif_table[out_vifi];
1818         struct vif_device *in_vif = &mrt->vif_table[in_vifi];
1819
1820         if (!skb->offload_l3_fwd_mark)
1821                 return false;
1822         if (!out_vif->dev_parent_id.id_len || !in_vif->dev_parent_id.id_len)
1823                 return false;
1824         return netdev_phys_item_id_same(&out_vif->dev_parent_id,
1825                                         &in_vif->dev_parent_id);
1826 }
1827 #else
1828 static bool ipmr_forward_offloaded(struct sk_buff *skb, struct mr_table *mrt,
1829                                    int in_vifi, int out_vifi)
1830 {
1831         return false;
1832 }
1833 #endif
1834
1835 /* Processing handlers for ipmr_forward, under rcu_read_lock() */
1836
1837 static void ipmr_queue_xmit(struct net *net, struct mr_table *mrt,
1838                             int in_vifi, struct sk_buff *skb, int vifi)
1839 {
1840         const struct iphdr *iph = ip_hdr(skb);
1841         struct vif_device *vif = &mrt->vif_table[vifi];
1842         struct net_device *vif_dev;
1843         struct net_device *dev;
1844         struct rtable *rt;
1845         struct flowi4 fl4;
1846         int    encap = 0;
1847
1848         vif_dev = vif_dev_read(vif);
1849         if (!vif_dev)
1850                 goto out_free;
1851
1852         if (vif->flags & VIFF_REGISTER) {
1853                 WRITE_ONCE(vif->pkt_out, vif->pkt_out + 1);
1854                 WRITE_ONCE(vif->bytes_out, vif->bytes_out + skb->len);
1855                 DEV_STATS_ADD(vif_dev, tx_bytes, skb->len);
1856                 DEV_STATS_INC(vif_dev, tx_packets);
1857                 ipmr_cache_report(mrt, skb, vifi, IGMPMSG_WHOLEPKT);
1858                 goto out_free;
1859         }
1860
1861         if (ipmr_forward_offloaded(skb, mrt, in_vifi, vifi))
1862                 goto out_free;
1863
1864         if (vif->flags & VIFF_TUNNEL) {
1865                 rt = ip_route_output_ports(net, &fl4, NULL,
1866                                            vif->remote, vif->local,
1867                                            0, 0,
1868                                            IPPROTO_IPIP,
1869                                            RT_TOS(iph->tos), vif->link);
1870                 if (IS_ERR(rt))
1871                         goto out_free;
1872                 encap = sizeof(struct iphdr);
1873         } else {
1874                 rt = ip_route_output_ports(net, &fl4, NULL, iph->daddr, 0,
1875                                            0, 0,
1876                                            IPPROTO_IPIP,
1877                                            RT_TOS(iph->tos), vif->link);
1878                 if (IS_ERR(rt))
1879                         goto out_free;
1880         }
1881
1882         dev = rt->dst.dev;
1883
1884         if (skb->len+encap > dst_mtu(&rt->dst) && (ntohs(iph->frag_off) & IP_DF)) {
1885                 /* Do not fragment multicasts. Alas, IPv4 does not
1886                  * allow to send ICMP, so that packets will disappear
1887                  * to blackhole.
1888                  */
1889                 IP_INC_STATS(net, IPSTATS_MIB_FRAGFAILS);
1890                 ip_rt_put(rt);
1891                 goto out_free;
1892         }
1893
1894         encap += LL_RESERVED_SPACE(dev) + rt->dst.header_len;
1895
1896         if (skb_cow(skb, encap)) {
1897                 ip_rt_put(rt);
1898                 goto out_free;
1899         }
1900
1901         WRITE_ONCE(vif->pkt_out, vif->pkt_out + 1);
1902         WRITE_ONCE(vif->bytes_out, vif->bytes_out + skb->len);
1903
1904         skb_dst_drop(skb);
1905         skb_dst_set(skb, &rt->dst);
1906         ip_decrease_ttl(ip_hdr(skb));
1907
1908         /* FIXME: forward and output firewalls used to be called here.
1909          * What do we do with netfilter? -- RR
1910          */
1911         if (vif->flags & VIFF_TUNNEL) {
1912                 ip_encap(net, skb, vif->local, vif->remote);
1913                 /* FIXME: extra output firewall step used to be here. --RR */
1914                 DEV_STATS_INC(vif_dev, tx_packets);
1915                 DEV_STATS_ADD(vif_dev, tx_bytes, skb->len);
1916         }
1917
1918         IPCB(skb)->flags |= IPSKB_FORWARDED;
1919
1920         /* RFC1584 teaches, that DVMRP/PIM router must deliver packets locally
1921          * not only before forwarding, but after forwarding on all output
1922          * interfaces. It is clear, if mrouter runs a multicasting
1923          * program, it should receive packets not depending to what interface
1924          * program is joined.
1925          * If we will not make it, the program will have to join on all
1926          * interfaces. On the other hand, multihoming host (or router, but
1927          * not mrouter) cannot join to more than one interface - it will
1928          * result in receiving multiple packets.
1929          */
1930         NF_HOOK(NFPROTO_IPV4, NF_INET_FORWARD,
1931                 net, NULL, skb, skb->dev, dev,
1932                 ipmr_forward_finish);
1933         return;
1934
1935 out_free:
1936         kfree_skb(skb);
1937 }
1938
1939 /* Called with mrt_lock or rcu_read_lock() */
1940 static int ipmr_find_vif(const struct mr_table *mrt, struct net_device *dev)
1941 {
1942         int ct;
1943         /* Pairs with WRITE_ONCE() in vif_delete()/vif_add() */
1944         for (ct = READ_ONCE(mrt->maxvif) - 1; ct >= 0; ct--) {
1945                 if (rcu_access_pointer(mrt->vif_table[ct].dev) == dev)
1946                         break;
1947         }
1948         return ct;
1949 }
1950
1951 /* "local" means that we should preserve one skb (for local delivery) */
1952 /* Called uner rcu_read_lock() */
1953 static void ip_mr_forward(struct net *net, struct mr_table *mrt,
1954                           struct net_device *dev, struct sk_buff *skb,
1955                           struct mfc_cache *c, int local)
1956 {
1957         int true_vifi = ipmr_find_vif(mrt, dev);
1958         int psend = -1;
1959         int vif, ct;
1960
1961         vif = c->_c.mfc_parent;
1962         c->_c.mfc_un.res.pkt++;
1963         c->_c.mfc_un.res.bytes += skb->len;
1964         c->_c.mfc_un.res.lastuse = jiffies;
1965
1966         if (c->mfc_origin == htonl(INADDR_ANY) && true_vifi >= 0) {
1967                 struct mfc_cache *cache_proxy;
1968
1969                 /* For an (*,G) entry, we only check that the incoming
1970                  * interface is part of the static tree.
1971                  */
1972                 cache_proxy = mr_mfc_find_any_parent(mrt, vif);
1973                 if (cache_proxy &&
1974                     cache_proxy->_c.mfc_un.res.ttls[true_vifi] < 255)
1975                         goto forward;
1976         }
1977
1978         /* Wrong interface: drop packet and (maybe) send PIM assert. */
1979         if (rcu_access_pointer(mrt->vif_table[vif].dev) != dev) {
1980                 if (rt_is_output_route(skb_rtable(skb))) {
1981                         /* It is our own packet, looped back.
1982                          * Very complicated situation...
1983                          *
1984                          * The best workaround until routing daemons will be
1985                          * fixed is not to redistribute packet, if it was
1986                          * send through wrong interface. It means, that
1987                          * multicast applications WILL NOT work for
1988                          * (S,G), which have default multicast route pointing
1989                          * to wrong oif. In any case, it is not a good
1990                          * idea to use multicasting applications on router.
1991                          */
1992                         goto dont_forward;
1993                 }
1994
1995                 c->_c.mfc_un.res.wrong_if++;
1996
1997                 if (true_vifi >= 0 && mrt->mroute_do_assert &&
1998                     /* pimsm uses asserts, when switching from RPT to SPT,
1999                      * so that we cannot check that packet arrived on an oif.
2000                      * It is bad, but otherwise we would need to move pretty
2001                      * large chunk of pimd to kernel. Ough... --ANK
2002                      */
2003                     (mrt->mroute_do_pim ||
2004                      c->_c.mfc_un.res.ttls[true_vifi] < 255) &&
2005                     time_after(jiffies,
2006                                c->_c.mfc_un.res.last_assert +
2007                                MFC_ASSERT_THRESH)) {
2008                         c->_c.mfc_un.res.last_assert = jiffies;
2009                         ipmr_cache_report(mrt, skb, true_vifi, IGMPMSG_WRONGVIF);
2010                         if (mrt->mroute_do_wrvifwhole)
2011                                 ipmr_cache_report(mrt, skb, true_vifi,
2012                                                   IGMPMSG_WRVIFWHOLE);
2013                 }
2014                 goto dont_forward;
2015         }
2016
2017 forward:
2018         WRITE_ONCE(mrt->vif_table[vif].pkt_in,
2019                    mrt->vif_table[vif].pkt_in + 1);
2020         WRITE_ONCE(mrt->vif_table[vif].bytes_in,
2021                    mrt->vif_table[vif].bytes_in + skb->len);
2022
2023         /* Forward the frame */
2024         if (c->mfc_origin == htonl(INADDR_ANY) &&
2025             c->mfc_mcastgrp == htonl(INADDR_ANY)) {
2026                 if (true_vifi >= 0 &&
2027                     true_vifi != c->_c.mfc_parent &&
2028                     ip_hdr(skb)->ttl >
2029                                 c->_c.mfc_un.res.ttls[c->_c.mfc_parent]) {
2030                         /* It's an (*,*) entry and the packet is not coming from
2031                          * the upstream: forward the packet to the upstream
2032                          * only.
2033                          */
2034                         psend = c->_c.mfc_parent;
2035                         goto last_forward;
2036                 }
2037                 goto dont_forward;
2038         }
2039         for (ct = c->_c.mfc_un.res.maxvif - 1;
2040              ct >= c->_c.mfc_un.res.minvif; ct--) {
2041                 /* For (*,G) entry, don't forward to the incoming interface */
2042                 if ((c->mfc_origin != htonl(INADDR_ANY) ||
2043                      ct != true_vifi) &&
2044                     ip_hdr(skb)->ttl > c->_c.mfc_un.res.ttls[ct]) {
2045                         if (psend != -1) {
2046                                 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2047
2048                                 if (skb2)
2049                                         ipmr_queue_xmit(net, mrt, true_vifi,
2050                                                         skb2, psend);
2051                         }
2052                         psend = ct;
2053                 }
2054         }
2055 last_forward:
2056         if (psend != -1) {
2057                 if (local) {
2058                         struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2059
2060                         if (skb2)
2061                                 ipmr_queue_xmit(net, mrt, true_vifi, skb2,
2062                                                 psend);
2063                 } else {
2064                         ipmr_queue_xmit(net, mrt, true_vifi, skb, psend);
2065                         return;
2066                 }
2067         }
2068
2069 dont_forward:
2070         if (!local)
2071                 kfree_skb(skb);
2072 }
2073
2074 static struct mr_table *ipmr_rt_fib_lookup(struct net *net, struct sk_buff *skb)
2075 {
2076         struct rtable *rt = skb_rtable(skb);
2077         struct iphdr *iph = ip_hdr(skb);
2078         struct flowi4 fl4 = {
2079                 .daddr = iph->daddr,
2080                 .saddr = iph->saddr,
2081                 .flowi4_tos = RT_TOS(iph->tos),
2082                 .flowi4_oif = (rt_is_output_route(rt) ?
2083                                skb->dev->ifindex : 0),
2084                 .flowi4_iif = (rt_is_output_route(rt) ?
2085                                LOOPBACK_IFINDEX :
2086                                skb->dev->ifindex),
2087                 .flowi4_mark = skb->mark,
2088         };
2089         struct mr_table *mrt;
2090         int err;
2091
2092         err = ipmr_fib_lookup(net, &fl4, &mrt);
2093         if (err)
2094                 return ERR_PTR(err);
2095         return mrt;
2096 }
2097
2098 /* Multicast packets for forwarding arrive here
2099  * Called with rcu_read_lock();
2100  */
2101 int ip_mr_input(struct sk_buff *skb)
2102 {
2103         struct mfc_cache *cache;
2104         struct net *net = dev_net(skb->dev);
2105         int local = skb_rtable(skb)->rt_flags & RTCF_LOCAL;
2106         struct mr_table *mrt;
2107         struct net_device *dev;
2108
2109         /* skb->dev passed in is the loX master dev for vrfs.
2110          * As there are no vifs associated with loopback devices,
2111          * get the proper interface that does have a vif associated with it.
2112          */
2113         dev = skb->dev;
2114         if (netif_is_l3_master(skb->dev)) {
2115                 dev = dev_get_by_index_rcu(net, IPCB(skb)->iif);
2116                 if (!dev) {
2117                         kfree_skb(skb);
2118                         return -ENODEV;
2119                 }
2120         }
2121
2122         /* Packet is looped back after forward, it should not be
2123          * forwarded second time, but still can be delivered locally.
2124          */
2125         if (IPCB(skb)->flags & IPSKB_FORWARDED)
2126                 goto dont_forward;
2127
2128         mrt = ipmr_rt_fib_lookup(net, skb);
2129         if (IS_ERR(mrt)) {
2130                 kfree_skb(skb);
2131                 return PTR_ERR(mrt);
2132         }
2133         if (!local) {
2134                 if (IPCB(skb)->opt.router_alert) {
2135                         if (ip_call_ra_chain(skb))
2136                                 return 0;
2137                 } else if (ip_hdr(skb)->protocol == IPPROTO_IGMP) {
2138                         /* IGMPv1 (and broken IGMPv2 implementations sort of
2139                          * Cisco IOS <= 11.2(8)) do not put router alert
2140                          * option to IGMP packets destined to routable
2141                          * groups. It is very bad, because it means
2142                          * that we can forward NO IGMP messages.
2143                          */
2144                         struct sock *mroute_sk;
2145
2146                         mroute_sk = rcu_dereference(mrt->mroute_sk);
2147                         if (mroute_sk) {
2148                                 nf_reset_ct(skb);
2149                                 raw_rcv(mroute_sk, skb);
2150                                 return 0;
2151                         }
2152                 }
2153         }
2154
2155         /* already under rcu_read_lock() */
2156         cache = ipmr_cache_find(mrt, ip_hdr(skb)->saddr, ip_hdr(skb)->daddr);
2157         if (!cache) {
2158                 int vif = ipmr_find_vif(mrt, dev);
2159
2160                 if (vif >= 0)
2161                         cache = ipmr_cache_find_any(mrt, ip_hdr(skb)->daddr,
2162                                                     vif);
2163         }
2164
2165         /* No usable cache entry */
2166         if (!cache) {
2167                 int vif;
2168
2169                 if (local) {
2170                         struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2171                         ip_local_deliver(skb);
2172                         if (!skb2)
2173                                 return -ENOBUFS;
2174                         skb = skb2;
2175                 }
2176
2177                 vif = ipmr_find_vif(mrt, dev);
2178                 if (vif >= 0)
2179                         return ipmr_cache_unresolved(mrt, vif, skb, dev);
2180                 kfree_skb(skb);
2181                 return -ENODEV;
2182         }
2183
2184         ip_mr_forward(net, mrt, dev, skb, cache, local);
2185
2186         if (local)
2187                 return ip_local_deliver(skb);
2188
2189         return 0;
2190
2191 dont_forward:
2192         if (local)
2193                 return ip_local_deliver(skb);
2194         kfree_skb(skb);
2195         return 0;
2196 }
2197
2198 #ifdef CONFIG_IP_PIMSM_V1
2199 /* Handle IGMP messages of PIMv1 */
2200 int pim_rcv_v1(struct sk_buff *skb)
2201 {
2202         struct igmphdr *pim;
2203         struct net *net = dev_net(skb->dev);
2204         struct mr_table *mrt;
2205
2206         if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr)))
2207                 goto drop;
2208
2209         pim = igmp_hdr(skb);
2210
2211         mrt = ipmr_rt_fib_lookup(net, skb);
2212         if (IS_ERR(mrt))
2213                 goto drop;
2214         if (!mrt->mroute_do_pim ||
2215             pim->group != PIM_V1_VERSION || pim->code != PIM_V1_REGISTER)
2216                 goto drop;
2217
2218         if (__pim_rcv(mrt, skb, sizeof(*pim))) {
2219 drop:
2220                 kfree_skb(skb);
2221         }
2222         return 0;
2223 }
2224 #endif
2225
2226 #ifdef CONFIG_IP_PIMSM_V2
2227 static int pim_rcv(struct sk_buff *skb)
2228 {
2229         struct pimreghdr *pim;
2230         struct net *net = dev_net(skb->dev);
2231         struct mr_table *mrt;
2232
2233         if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr)))
2234                 goto drop;
2235
2236         pim = (struct pimreghdr *)skb_transport_header(skb);
2237         if (pim->type != ((PIM_VERSION << 4) | (PIM_TYPE_REGISTER)) ||
2238             (pim->flags & PIM_NULL_REGISTER) ||
2239             (ip_compute_csum((void *)pim, sizeof(*pim)) != 0 &&
2240              csum_fold(skb_checksum(skb, 0, skb->len, 0))))
2241                 goto drop;
2242
2243         mrt = ipmr_rt_fib_lookup(net, skb);
2244         if (IS_ERR(mrt))
2245                 goto drop;
2246         if (__pim_rcv(mrt, skb, sizeof(*pim))) {
2247 drop:
2248                 kfree_skb(skb);
2249         }
2250         return 0;
2251 }
2252 #endif
2253
2254 int ipmr_get_route(struct net *net, struct sk_buff *skb,
2255                    __be32 saddr, __be32 daddr,
2256                    struct rtmsg *rtm, u32 portid)
2257 {
2258         struct mfc_cache *cache;
2259         struct mr_table *mrt;
2260         int err;
2261
2262         mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2263         if (!mrt)
2264                 return -ENOENT;
2265
2266         rcu_read_lock();
2267         cache = ipmr_cache_find(mrt, saddr, daddr);
2268         if (!cache && skb->dev) {
2269                 int vif = ipmr_find_vif(mrt, skb->dev);
2270
2271                 if (vif >= 0)
2272                         cache = ipmr_cache_find_any(mrt, daddr, vif);
2273         }
2274         if (!cache) {
2275                 struct sk_buff *skb2;
2276                 struct iphdr *iph;
2277                 struct net_device *dev;
2278                 int vif = -1;
2279
2280                 dev = skb->dev;
2281                 if (dev)
2282                         vif = ipmr_find_vif(mrt, dev);
2283                 if (vif < 0) {
2284                         rcu_read_unlock();
2285                         return -ENODEV;
2286                 }
2287
2288                 skb2 = skb_realloc_headroom(skb, sizeof(struct iphdr));
2289                 if (!skb2) {
2290                         rcu_read_unlock();
2291                         return -ENOMEM;
2292                 }
2293
2294                 NETLINK_CB(skb2).portid = portid;
2295                 skb_push(skb2, sizeof(struct iphdr));
2296                 skb_reset_network_header(skb2);
2297                 iph = ip_hdr(skb2);
2298                 iph->ihl = sizeof(struct iphdr) >> 2;
2299                 iph->saddr = saddr;
2300                 iph->daddr = daddr;
2301                 iph->version = 0;
2302                 err = ipmr_cache_unresolved(mrt, vif, skb2, dev);
2303                 rcu_read_unlock();
2304                 return err;
2305         }
2306
2307         err = mr_fill_mroute(mrt, skb, &cache->_c, rtm);
2308         rcu_read_unlock();
2309         return err;
2310 }
2311
2312 static int ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
2313                             u32 portid, u32 seq, struct mfc_cache *c, int cmd,
2314                             int flags)
2315 {
2316         struct nlmsghdr *nlh;
2317         struct rtmsg *rtm;
2318         int err;
2319
2320         nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rtm), flags);
2321         if (!nlh)
2322                 return -EMSGSIZE;
2323
2324         rtm = nlmsg_data(nlh);
2325         rtm->rtm_family   = RTNL_FAMILY_IPMR;
2326         rtm->rtm_dst_len  = 32;
2327         rtm->rtm_src_len  = 32;
2328         rtm->rtm_tos      = 0;
2329         rtm->rtm_table    = mrt->id;
2330         if (nla_put_u32(skb, RTA_TABLE, mrt->id))
2331                 goto nla_put_failure;
2332         rtm->rtm_type     = RTN_MULTICAST;
2333         rtm->rtm_scope    = RT_SCOPE_UNIVERSE;
2334         if (c->_c.mfc_flags & MFC_STATIC)
2335                 rtm->rtm_protocol = RTPROT_STATIC;
2336         else
2337                 rtm->rtm_protocol = RTPROT_MROUTED;
2338         rtm->rtm_flags    = 0;
2339
2340         if (nla_put_in_addr(skb, RTA_SRC, c->mfc_origin) ||
2341             nla_put_in_addr(skb, RTA_DST, c->mfc_mcastgrp))
2342                 goto nla_put_failure;
2343         err = mr_fill_mroute(mrt, skb, &c->_c, rtm);
2344         /* do not break the dump if cache is unresolved */
2345         if (err < 0 && err != -ENOENT)
2346                 goto nla_put_failure;
2347
2348         nlmsg_end(skb, nlh);
2349         return 0;
2350
2351 nla_put_failure:
2352         nlmsg_cancel(skb, nlh);
2353         return -EMSGSIZE;
2354 }
2355
2356 static int _ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
2357                              u32 portid, u32 seq, struct mr_mfc *c, int cmd,
2358                              int flags)
2359 {
2360         return ipmr_fill_mroute(mrt, skb, portid, seq, (struct mfc_cache *)c,
2361                                 cmd, flags);
2362 }
2363
2364 static size_t mroute_msgsize(bool unresolved, int maxvif)
2365 {
2366         size_t len =
2367                 NLMSG_ALIGN(sizeof(struct rtmsg))
2368                 + nla_total_size(4)     /* RTA_TABLE */
2369                 + nla_total_size(4)     /* RTA_SRC */
2370                 + nla_total_size(4)     /* RTA_DST */
2371                 ;
2372
2373         if (!unresolved)
2374                 len = len
2375                       + nla_total_size(4)       /* RTA_IIF */
2376                       + nla_total_size(0)       /* RTA_MULTIPATH */
2377                       + maxvif * NLA_ALIGN(sizeof(struct rtnexthop))
2378                                                 /* RTA_MFC_STATS */
2379                       + nla_total_size_64bit(sizeof(struct rta_mfc_stats))
2380                 ;
2381
2382         return len;
2383 }
2384
2385 static void mroute_netlink_event(struct mr_table *mrt, struct mfc_cache *mfc,
2386                                  int cmd)
2387 {
2388         struct net *net = read_pnet(&mrt->net);
2389         struct sk_buff *skb;
2390         int err = -ENOBUFS;
2391
2392         skb = nlmsg_new(mroute_msgsize(mfc->_c.mfc_parent >= MAXVIFS,
2393                                        mrt->maxvif),
2394                         GFP_ATOMIC);
2395         if (!skb)
2396                 goto errout;
2397
2398         err = ipmr_fill_mroute(mrt, skb, 0, 0, mfc, cmd, 0);
2399         if (err < 0)
2400                 goto errout;
2401
2402         rtnl_notify(skb, net, 0, RTNLGRP_IPV4_MROUTE, NULL, GFP_ATOMIC);
2403         return;
2404
2405 errout:
2406         kfree_skb(skb);
2407         if (err < 0)
2408                 rtnl_set_sk_err(net, RTNLGRP_IPV4_MROUTE, err);
2409 }
2410
2411 static size_t igmpmsg_netlink_msgsize(size_t payloadlen)
2412 {
2413         size_t len =
2414                 NLMSG_ALIGN(sizeof(struct rtgenmsg))
2415                 + nla_total_size(1)     /* IPMRA_CREPORT_MSGTYPE */
2416                 + nla_total_size(4)     /* IPMRA_CREPORT_VIF_ID */
2417                 + nla_total_size(4)     /* IPMRA_CREPORT_SRC_ADDR */
2418                 + nla_total_size(4)     /* IPMRA_CREPORT_DST_ADDR */
2419                 + nla_total_size(4)     /* IPMRA_CREPORT_TABLE */
2420                                         /* IPMRA_CREPORT_PKT */
2421                 + nla_total_size(payloadlen)
2422                 ;
2423
2424         return len;
2425 }
2426
2427 static void igmpmsg_netlink_event(const struct mr_table *mrt, struct sk_buff *pkt)
2428 {
2429         struct net *net = read_pnet(&mrt->net);
2430         struct nlmsghdr *nlh;
2431         struct rtgenmsg *rtgenm;
2432         struct igmpmsg *msg;
2433         struct sk_buff *skb;
2434         struct nlattr *nla;
2435         int payloadlen;
2436
2437         payloadlen = pkt->len - sizeof(struct igmpmsg);
2438         msg = (struct igmpmsg *)skb_network_header(pkt);
2439
2440         skb = nlmsg_new(igmpmsg_netlink_msgsize(payloadlen), GFP_ATOMIC);
2441         if (!skb)
2442                 goto errout;
2443
2444         nlh = nlmsg_put(skb, 0, 0, RTM_NEWCACHEREPORT,
2445                         sizeof(struct rtgenmsg), 0);
2446         if (!nlh)
2447                 goto errout;
2448         rtgenm = nlmsg_data(nlh);
2449         rtgenm->rtgen_family = RTNL_FAMILY_IPMR;
2450         if (nla_put_u8(skb, IPMRA_CREPORT_MSGTYPE, msg->im_msgtype) ||
2451             nla_put_u32(skb, IPMRA_CREPORT_VIF_ID, msg->im_vif | (msg->im_vif_hi << 8)) ||
2452             nla_put_in_addr(skb, IPMRA_CREPORT_SRC_ADDR,
2453                             msg->im_src.s_addr) ||
2454             nla_put_in_addr(skb, IPMRA_CREPORT_DST_ADDR,
2455                             msg->im_dst.s_addr) ||
2456             nla_put_u32(skb, IPMRA_CREPORT_TABLE, mrt->id))
2457                 goto nla_put_failure;
2458
2459         nla = nla_reserve(skb, IPMRA_CREPORT_PKT, payloadlen);
2460         if (!nla || skb_copy_bits(pkt, sizeof(struct igmpmsg),
2461                                   nla_data(nla), payloadlen))
2462                 goto nla_put_failure;
2463
2464         nlmsg_end(skb, nlh);
2465
2466         rtnl_notify(skb, net, 0, RTNLGRP_IPV4_MROUTE_R, NULL, GFP_ATOMIC);
2467         return;
2468
2469 nla_put_failure:
2470         nlmsg_cancel(skb, nlh);
2471 errout:
2472         kfree_skb(skb);
2473         rtnl_set_sk_err(net, RTNLGRP_IPV4_MROUTE_R, -ENOBUFS);
2474 }
2475
2476 static int ipmr_rtm_valid_getroute_req(struct sk_buff *skb,
2477                                        const struct nlmsghdr *nlh,
2478                                        struct nlattr **tb,
2479                                        struct netlink_ext_ack *extack)
2480 {
2481         struct rtmsg *rtm;
2482         int i, err;
2483
2484         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*rtm))) {
2485                 NL_SET_ERR_MSG(extack, "ipv4: Invalid header for multicast route get request");
2486                 return -EINVAL;
2487         }
2488
2489         if (!netlink_strict_get_check(skb))
2490                 return nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX,
2491                                               rtm_ipv4_policy, extack);
2492
2493         rtm = nlmsg_data(nlh);
2494         if ((rtm->rtm_src_len && rtm->rtm_src_len != 32) ||
2495             (rtm->rtm_dst_len && rtm->rtm_dst_len != 32) ||
2496             rtm->rtm_tos || rtm->rtm_table || rtm->rtm_protocol ||
2497             rtm->rtm_scope || rtm->rtm_type || rtm->rtm_flags) {
2498                 NL_SET_ERR_MSG(extack, "ipv4: Invalid values in header for multicast route get request");
2499                 return -EINVAL;
2500         }
2501
2502         err = nlmsg_parse_deprecated_strict(nlh, sizeof(*rtm), tb, RTA_MAX,
2503                                             rtm_ipv4_policy, extack);
2504         if (err)
2505                 return err;
2506
2507         if ((tb[RTA_SRC] && !rtm->rtm_src_len) ||
2508             (tb[RTA_DST] && !rtm->rtm_dst_len)) {
2509                 NL_SET_ERR_MSG(extack, "ipv4: rtm_src_len and rtm_dst_len must be 32 for IPv4");
2510                 return -EINVAL;
2511         }
2512
2513         for (i = 0; i <= RTA_MAX; i++) {
2514                 if (!tb[i])
2515                         continue;
2516
2517                 switch (i) {
2518                 case RTA_SRC:
2519                 case RTA_DST:
2520                 case RTA_TABLE:
2521                         break;
2522                 default:
2523                         NL_SET_ERR_MSG(extack, "ipv4: Unsupported attribute in multicast route get request");
2524                         return -EINVAL;
2525                 }
2526         }
2527
2528         return 0;
2529 }
2530
2531 static int ipmr_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh,
2532                              struct netlink_ext_ack *extack)
2533 {
2534         struct net *net = sock_net(in_skb->sk);
2535         struct nlattr *tb[RTA_MAX + 1];
2536         struct sk_buff *skb = NULL;
2537         struct mfc_cache *cache;
2538         struct mr_table *mrt;
2539         __be32 src, grp;
2540         u32 tableid;
2541         int err;
2542
2543         err = ipmr_rtm_valid_getroute_req(in_skb, nlh, tb, extack);
2544         if (err < 0)
2545                 goto errout;
2546
2547         src = tb[RTA_SRC] ? nla_get_in_addr(tb[RTA_SRC]) : 0;
2548         grp = tb[RTA_DST] ? nla_get_in_addr(tb[RTA_DST]) : 0;
2549         tableid = tb[RTA_TABLE] ? nla_get_u32(tb[RTA_TABLE]) : 0;
2550
2551         mrt = ipmr_get_table(net, tableid ? tableid : RT_TABLE_DEFAULT);
2552         if (!mrt) {
2553                 err = -ENOENT;
2554                 goto errout_free;
2555         }
2556
2557         /* entries are added/deleted only under RTNL */
2558         rcu_read_lock();
2559         cache = ipmr_cache_find(mrt, src, grp);
2560         rcu_read_unlock();
2561         if (!cache) {
2562                 err = -ENOENT;
2563                 goto errout_free;
2564         }
2565
2566         skb = nlmsg_new(mroute_msgsize(false, mrt->maxvif), GFP_KERNEL);
2567         if (!skb) {
2568                 err = -ENOBUFS;
2569                 goto errout_free;
2570         }
2571
2572         err = ipmr_fill_mroute(mrt, skb, NETLINK_CB(in_skb).portid,
2573                                nlh->nlmsg_seq, cache,
2574                                RTM_NEWROUTE, 0);
2575         if (err < 0)
2576                 goto errout_free;
2577
2578         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
2579
2580 errout:
2581         return err;
2582
2583 errout_free:
2584         kfree_skb(skb);
2585         goto errout;
2586 }
2587
2588 static int ipmr_rtm_dumproute(struct sk_buff *skb, struct netlink_callback *cb)
2589 {
2590         struct fib_dump_filter filter = {};
2591         int err;
2592
2593         if (cb->strict_check) {
2594                 err = ip_valid_fib_dump_req(sock_net(skb->sk), cb->nlh,
2595                                             &filter, cb);
2596                 if (err < 0)
2597                         return err;
2598         }
2599
2600         if (filter.table_id) {
2601                 struct mr_table *mrt;
2602
2603                 mrt = ipmr_get_table(sock_net(skb->sk), filter.table_id);
2604                 if (!mrt) {
2605                         if (rtnl_msg_family(cb->nlh) != RTNL_FAMILY_IPMR)
2606                                 return skb->len;
2607
2608                         NL_SET_ERR_MSG(cb->extack, "ipv4: MR table does not exist");
2609                         return -ENOENT;
2610                 }
2611                 err = mr_table_dump(mrt, skb, cb, _ipmr_fill_mroute,
2612                                     &mfc_unres_lock, &filter);
2613                 return skb->len ? : err;
2614         }
2615
2616         return mr_rtm_dumproute(skb, cb, ipmr_mr_table_iter,
2617                                 _ipmr_fill_mroute, &mfc_unres_lock, &filter);
2618 }
2619
2620 static const struct nla_policy rtm_ipmr_policy[RTA_MAX + 1] = {
2621         [RTA_SRC]       = { .type = NLA_U32 },
2622         [RTA_DST]       = { .type = NLA_U32 },
2623         [RTA_IIF]       = { .type = NLA_U32 },
2624         [RTA_TABLE]     = { .type = NLA_U32 },
2625         [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
2626 };
2627
2628 static bool ipmr_rtm_validate_proto(unsigned char rtm_protocol)
2629 {
2630         switch (rtm_protocol) {
2631         case RTPROT_STATIC:
2632         case RTPROT_MROUTED:
2633                 return true;
2634         }
2635         return false;
2636 }
2637
2638 static int ipmr_nla_get_ttls(const struct nlattr *nla, struct mfcctl *mfcc)
2639 {
2640         struct rtnexthop *rtnh = nla_data(nla);
2641         int remaining = nla_len(nla), vifi = 0;
2642
2643         while (rtnh_ok(rtnh, remaining)) {
2644                 mfcc->mfcc_ttls[vifi] = rtnh->rtnh_hops;
2645                 if (++vifi == MAXVIFS)
2646                         break;
2647                 rtnh = rtnh_next(rtnh, &remaining);
2648         }
2649
2650         return remaining > 0 ? -EINVAL : vifi;
2651 }
2652
2653 /* returns < 0 on error, 0 for ADD_MFC and 1 for ADD_MFC_PROXY */
2654 static int rtm_to_ipmr_mfcc(struct net *net, struct nlmsghdr *nlh,
2655                             struct mfcctl *mfcc, int *mrtsock,
2656                             struct mr_table **mrtret,
2657                             struct netlink_ext_ack *extack)
2658 {
2659         struct net_device *dev = NULL;
2660         u32 tblid = RT_TABLE_DEFAULT;
2661         struct mr_table *mrt;
2662         struct nlattr *attr;
2663         struct rtmsg *rtm;
2664         int ret, rem;
2665
2666         ret = nlmsg_validate_deprecated(nlh, sizeof(*rtm), RTA_MAX,
2667                                         rtm_ipmr_policy, extack);
2668         if (ret < 0)
2669                 goto out;
2670         rtm = nlmsg_data(nlh);
2671
2672         ret = -EINVAL;
2673         if (rtm->rtm_family != RTNL_FAMILY_IPMR || rtm->rtm_dst_len != 32 ||
2674             rtm->rtm_type != RTN_MULTICAST ||
2675             rtm->rtm_scope != RT_SCOPE_UNIVERSE ||
2676             !ipmr_rtm_validate_proto(rtm->rtm_protocol))
2677                 goto out;
2678
2679         memset(mfcc, 0, sizeof(*mfcc));
2680         mfcc->mfcc_parent = -1;
2681         ret = 0;
2682         nlmsg_for_each_attr(attr, nlh, sizeof(struct rtmsg), rem) {
2683                 switch (nla_type(attr)) {
2684                 case RTA_SRC:
2685                         mfcc->mfcc_origin.s_addr = nla_get_be32(attr);
2686                         break;
2687                 case RTA_DST:
2688                         mfcc->mfcc_mcastgrp.s_addr = nla_get_be32(attr);
2689                         break;
2690                 case RTA_IIF:
2691                         dev = __dev_get_by_index(net, nla_get_u32(attr));
2692                         if (!dev) {
2693                                 ret = -ENODEV;
2694                                 goto out;
2695                         }
2696                         break;
2697                 case RTA_MULTIPATH:
2698                         if (ipmr_nla_get_ttls(attr, mfcc) < 0) {
2699                                 ret = -EINVAL;
2700                                 goto out;
2701                         }
2702                         break;
2703                 case RTA_PREFSRC:
2704                         ret = 1;
2705                         break;
2706                 case RTA_TABLE:
2707                         tblid = nla_get_u32(attr);
2708                         break;
2709                 }
2710         }
2711         mrt = ipmr_get_table(net, tblid);
2712         if (!mrt) {
2713                 ret = -ENOENT;
2714                 goto out;
2715         }
2716         *mrtret = mrt;
2717         *mrtsock = rtm->rtm_protocol == RTPROT_MROUTED ? 1 : 0;
2718         if (dev)
2719                 mfcc->mfcc_parent = ipmr_find_vif(mrt, dev);
2720
2721 out:
2722         return ret;
2723 }
2724
2725 /* takes care of both newroute and delroute */
2726 static int ipmr_rtm_route(struct sk_buff *skb, struct nlmsghdr *nlh,
2727                           struct netlink_ext_ack *extack)
2728 {
2729         struct net *net = sock_net(skb->sk);
2730         int ret, mrtsock, parent;
2731         struct mr_table *tbl;
2732         struct mfcctl mfcc;
2733
2734         mrtsock = 0;
2735         tbl = NULL;
2736         ret = rtm_to_ipmr_mfcc(net, nlh, &mfcc, &mrtsock, &tbl, extack);
2737         if (ret < 0)
2738                 return ret;
2739
2740         parent = ret ? mfcc.mfcc_parent : -1;
2741         if (nlh->nlmsg_type == RTM_NEWROUTE)
2742                 return ipmr_mfc_add(net, tbl, &mfcc, mrtsock, parent);
2743         else
2744                 return ipmr_mfc_delete(tbl, &mfcc, parent);
2745 }
2746
2747 static bool ipmr_fill_table(struct mr_table *mrt, struct sk_buff *skb)
2748 {
2749         u32 queue_len = atomic_read(&mrt->cache_resolve_queue_len);
2750
2751         if (nla_put_u32(skb, IPMRA_TABLE_ID, mrt->id) ||
2752             nla_put_u32(skb, IPMRA_TABLE_CACHE_RES_QUEUE_LEN, queue_len) ||
2753             nla_put_s32(skb, IPMRA_TABLE_MROUTE_REG_VIF_NUM,
2754                         mrt->mroute_reg_vif_num) ||
2755             nla_put_u8(skb, IPMRA_TABLE_MROUTE_DO_ASSERT,
2756                        mrt->mroute_do_assert) ||
2757             nla_put_u8(skb, IPMRA_TABLE_MROUTE_DO_PIM, mrt->mroute_do_pim) ||
2758             nla_put_u8(skb, IPMRA_TABLE_MROUTE_DO_WRVIFWHOLE,
2759                        mrt->mroute_do_wrvifwhole))
2760                 return false;
2761
2762         return true;
2763 }
2764
2765 static bool ipmr_fill_vif(struct mr_table *mrt, u32 vifid, struct sk_buff *skb)
2766 {
2767         struct net_device *vif_dev;
2768         struct nlattr *vif_nest;
2769         struct vif_device *vif;
2770
2771         vif = &mrt->vif_table[vifid];
2772         vif_dev = rtnl_dereference(vif->dev);
2773         /* if the VIF doesn't exist just continue */
2774         if (!vif_dev)
2775                 return true;
2776
2777         vif_nest = nla_nest_start_noflag(skb, IPMRA_VIF);
2778         if (!vif_nest)
2779                 return false;
2780
2781         if (nla_put_u32(skb, IPMRA_VIFA_IFINDEX, vif_dev->ifindex) ||
2782             nla_put_u32(skb, IPMRA_VIFA_VIF_ID, vifid) ||
2783             nla_put_u16(skb, IPMRA_VIFA_FLAGS, vif->flags) ||
2784             nla_put_u64_64bit(skb, IPMRA_VIFA_BYTES_IN, vif->bytes_in,
2785                               IPMRA_VIFA_PAD) ||
2786             nla_put_u64_64bit(skb, IPMRA_VIFA_BYTES_OUT, vif->bytes_out,
2787                               IPMRA_VIFA_PAD) ||
2788             nla_put_u64_64bit(skb, IPMRA_VIFA_PACKETS_IN, vif->pkt_in,
2789                               IPMRA_VIFA_PAD) ||
2790             nla_put_u64_64bit(skb, IPMRA_VIFA_PACKETS_OUT, vif->pkt_out,
2791                               IPMRA_VIFA_PAD) ||
2792             nla_put_be32(skb, IPMRA_VIFA_LOCAL_ADDR, vif->local) ||
2793             nla_put_be32(skb, IPMRA_VIFA_REMOTE_ADDR, vif->remote)) {
2794                 nla_nest_cancel(skb, vif_nest);
2795                 return false;
2796         }
2797         nla_nest_end(skb, vif_nest);
2798
2799         return true;
2800 }
2801
2802 static int ipmr_valid_dumplink(const struct nlmsghdr *nlh,
2803                                struct netlink_ext_ack *extack)
2804 {
2805         struct ifinfomsg *ifm;
2806
2807         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
2808                 NL_SET_ERR_MSG(extack, "ipv4: Invalid header for ipmr link dump");
2809                 return -EINVAL;
2810         }
2811
2812         if (nlmsg_attrlen(nlh, sizeof(*ifm))) {
2813                 NL_SET_ERR_MSG(extack, "Invalid data after header in ipmr link dump");
2814                 return -EINVAL;
2815         }
2816
2817         ifm = nlmsg_data(nlh);
2818         if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
2819             ifm->ifi_change || ifm->ifi_index) {
2820                 NL_SET_ERR_MSG(extack, "Invalid values in header for ipmr link dump request");
2821                 return -EINVAL;
2822         }
2823
2824         return 0;
2825 }
2826
2827 static int ipmr_rtm_dumplink(struct sk_buff *skb, struct netlink_callback *cb)
2828 {
2829         struct net *net = sock_net(skb->sk);
2830         struct nlmsghdr *nlh = NULL;
2831         unsigned int t = 0, s_t;
2832         unsigned int e = 0, s_e;
2833         struct mr_table *mrt;
2834
2835         if (cb->strict_check) {
2836                 int err = ipmr_valid_dumplink(cb->nlh, cb->extack);
2837
2838                 if (err < 0)
2839                         return err;
2840         }
2841
2842         s_t = cb->args[0];
2843         s_e = cb->args[1];
2844
2845         ipmr_for_each_table(mrt, net) {
2846                 struct nlattr *vifs, *af;
2847                 struct ifinfomsg *hdr;
2848                 u32 i;
2849
2850                 if (t < s_t)
2851                         goto skip_table;
2852                 nlh = nlmsg_put(skb, NETLINK_CB(cb->skb).portid,
2853                                 cb->nlh->nlmsg_seq, RTM_NEWLINK,
2854                                 sizeof(*hdr), NLM_F_MULTI);
2855                 if (!nlh)
2856                         break;
2857
2858                 hdr = nlmsg_data(nlh);
2859                 memset(hdr, 0, sizeof(*hdr));
2860                 hdr->ifi_family = RTNL_FAMILY_IPMR;
2861
2862                 af = nla_nest_start_noflag(skb, IFLA_AF_SPEC);
2863                 if (!af) {
2864                         nlmsg_cancel(skb, nlh);
2865                         goto out;
2866                 }
2867
2868                 if (!ipmr_fill_table(mrt, skb)) {
2869                         nlmsg_cancel(skb, nlh);
2870                         goto out;
2871                 }
2872
2873                 vifs = nla_nest_start_noflag(skb, IPMRA_TABLE_VIFS);
2874                 if (!vifs) {
2875                         nla_nest_end(skb, af);
2876                         nlmsg_end(skb, nlh);
2877                         goto out;
2878                 }
2879                 for (i = 0; i < mrt->maxvif; i++) {
2880                         if (e < s_e)
2881                                 goto skip_entry;
2882                         if (!ipmr_fill_vif(mrt, i, skb)) {
2883                                 nla_nest_end(skb, vifs);
2884                                 nla_nest_end(skb, af);
2885                                 nlmsg_end(skb, nlh);
2886                                 goto out;
2887                         }
2888 skip_entry:
2889                         e++;
2890                 }
2891                 s_e = 0;
2892                 e = 0;
2893                 nla_nest_end(skb, vifs);
2894                 nla_nest_end(skb, af);
2895                 nlmsg_end(skb, nlh);
2896 skip_table:
2897                 t++;
2898         }
2899
2900 out:
2901         cb->args[1] = e;
2902         cb->args[0] = t;
2903
2904         return skb->len;
2905 }
2906
2907 #ifdef CONFIG_PROC_FS
2908 /* The /proc interfaces to multicast routing :
2909  * /proc/net/ip_mr_cache & /proc/net/ip_mr_vif
2910  */
2911
2912 static void *ipmr_vif_seq_start(struct seq_file *seq, loff_t *pos)
2913         __acquires(RCU)
2914 {
2915         struct mr_vif_iter *iter = seq->private;
2916         struct net *net = seq_file_net(seq);
2917         struct mr_table *mrt;
2918
2919         mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2920         if (!mrt)
2921                 return ERR_PTR(-ENOENT);
2922
2923         iter->mrt = mrt;
2924
2925         rcu_read_lock();
2926         return mr_vif_seq_start(seq, pos);
2927 }
2928
2929 static void ipmr_vif_seq_stop(struct seq_file *seq, void *v)
2930         __releases(RCU)
2931 {
2932         rcu_read_unlock();
2933 }
2934
2935 static int ipmr_vif_seq_show(struct seq_file *seq, void *v)
2936 {
2937         struct mr_vif_iter *iter = seq->private;
2938         struct mr_table *mrt = iter->mrt;
2939
2940         if (v == SEQ_START_TOKEN) {
2941                 seq_puts(seq,
2942                          "Interface      BytesIn  PktsIn  BytesOut PktsOut Flags Local    Remote\n");
2943         } else {
2944                 const struct vif_device *vif = v;
2945                 const struct net_device *vif_dev;
2946                 const char *name;
2947
2948                 vif_dev = vif_dev_read(vif);
2949                 name = vif_dev ? vif_dev->name : "none";
2950                 seq_printf(seq,
2951                            "%2td %-10s %8ld %7ld  %8ld %7ld %05X %08X %08X\n",
2952                            vif - mrt->vif_table,
2953                            name, vif->bytes_in, vif->pkt_in,
2954                            vif->bytes_out, vif->pkt_out,
2955                            vif->flags, vif->local, vif->remote);
2956         }
2957         return 0;
2958 }
2959
2960 static const struct seq_operations ipmr_vif_seq_ops = {
2961         .start = ipmr_vif_seq_start,
2962         .next  = mr_vif_seq_next,
2963         .stop  = ipmr_vif_seq_stop,
2964         .show  = ipmr_vif_seq_show,
2965 };
2966
2967 static void *ipmr_mfc_seq_start(struct seq_file *seq, loff_t *pos)
2968 {
2969         struct net *net = seq_file_net(seq);
2970         struct mr_table *mrt;
2971
2972         mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2973         if (!mrt)
2974                 return ERR_PTR(-ENOENT);
2975
2976         return mr_mfc_seq_start(seq, pos, mrt, &mfc_unres_lock);
2977 }
2978
2979 static int ipmr_mfc_seq_show(struct seq_file *seq, void *v)
2980 {
2981         int n;
2982
2983         if (v == SEQ_START_TOKEN) {
2984                 seq_puts(seq,
2985                  "Group    Origin   Iif     Pkts    Bytes    Wrong Oifs\n");
2986         } else {
2987                 const struct mfc_cache *mfc = v;
2988                 const struct mr_mfc_iter *it = seq->private;
2989                 const struct mr_table *mrt = it->mrt;
2990
2991                 seq_printf(seq, "%08X %08X %-3hd",
2992                            (__force u32) mfc->mfc_mcastgrp,
2993                            (__force u32) mfc->mfc_origin,
2994                            mfc->_c.mfc_parent);
2995
2996                 if (it->cache != &mrt->mfc_unres_queue) {
2997                         seq_printf(seq, " %8lu %8lu %8lu",
2998                                    mfc->_c.mfc_un.res.pkt,
2999                                    mfc->_c.mfc_un.res.bytes,
3000                                    mfc->_c.mfc_un.res.wrong_if);
3001                         for (n = mfc->_c.mfc_un.res.minvif;
3002                              n < mfc->_c.mfc_un.res.maxvif; n++) {
3003                                 if (VIF_EXISTS(mrt, n) &&
3004                                     mfc->_c.mfc_un.res.ttls[n] < 255)
3005                                         seq_printf(seq,
3006                                            " %2d:%-3d",
3007                                            n, mfc->_c.mfc_un.res.ttls[n]);
3008                         }
3009                 } else {
3010                         /* unresolved mfc_caches don't contain
3011                          * pkt, bytes and wrong_if values
3012                          */
3013                         seq_printf(seq, " %8lu %8lu %8lu", 0ul, 0ul, 0ul);
3014                 }
3015                 seq_putc(seq, '\n');
3016         }
3017         return 0;
3018 }
3019
3020 static const struct seq_operations ipmr_mfc_seq_ops = {
3021         .start = ipmr_mfc_seq_start,
3022         .next  = mr_mfc_seq_next,
3023         .stop  = mr_mfc_seq_stop,
3024         .show  = ipmr_mfc_seq_show,
3025 };
3026 #endif
3027
3028 #ifdef CONFIG_IP_PIMSM_V2
3029 static const struct net_protocol pim_protocol = {
3030         .handler        =       pim_rcv,
3031 };
3032 #endif
3033
3034 static unsigned int ipmr_seq_read(struct net *net)
3035 {
3036         ASSERT_RTNL();
3037
3038         return net->ipv4.ipmr_seq + ipmr_rules_seq_read(net);
3039 }
3040
3041 static int ipmr_dump(struct net *net, struct notifier_block *nb,
3042                      struct netlink_ext_ack *extack)
3043 {
3044         return mr_dump(net, nb, RTNL_FAMILY_IPMR, ipmr_rules_dump,
3045                        ipmr_mr_table_iter, extack);
3046 }
3047
3048 static const struct fib_notifier_ops ipmr_notifier_ops_template = {
3049         .family         = RTNL_FAMILY_IPMR,
3050         .fib_seq_read   = ipmr_seq_read,
3051         .fib_dump       = ipmr_dump,
3052         .owner          = THIS_MODULE,
3053 };
3054
3055 static int __net_init ipmr_notifier_init(struct net *net)
3056 {
3057         struct fib_notifier_ops *ops;
3058
3059         net->ipv4.ipmr_seq = 0;
3060
3061         ops = fib_notifier_ops_register(&ipmr_notifier_ops_template, net);
3062         if (IS_ERR(ops))
3063                 return PTR_ERR(ops);
3064         net->ipv4.ipmr_notifier_ops = ops;
3065
3066         return 0;
3067 }
3068
3069 static void __net_exit ipmr_notifier_exit(struct net *net)
3070 {
3071         fib_notifier_ops_unregister(net->ipv4.ipmr_notifier_ops);
3072         net->ipv4.ipmr_notifier_ops = NULL;
3073 }
3074
3075 /* Setup for IP multicast routing */
3076 static int __net_init ipmr_net_init(struct net *net)
3077 {
3078         int err;
3079
3080         err = ipmr_notifier_init(net);
3081         if (err)
3082                 goto ipmr_notifier_fail;
3083
3084         err = ipmr_rules_init(net);
3085         if (err < 0)
3086                 goto ipmr_rules_fail;
3087
3088 #ifdef CONFIG_PROC_FS
3089         err = -ENOMEM;
3090         if (!proc_create_net("ip_mr_vif", 0, net->proc_net, &ipmr_vif_seq_ops,
3091                         sizeof(struct mr_vif_iter)))
3092                 goto proc_vif_fail;
3093         if (!proc_create_net("ip_mr_cache", 0, net->proc_net, &ipmr_mfc_seq_ops,
3094                         sizeof(struct mr_mfc_iter)))
3095                 goto proc_cache_fail;
3096 #endif
3097         return 0;
3098
3099 #ifdef CONFIG_PROC_FS
3100 proc_cache_fail:
3101         remove_proc_entry("ip_mr_vif", net->proc_net);
3102 proc_vif_fail:
3103         rtnl_lock();
3104         ipmr_rules_exit(net);
3105         rtnl_unlock();
3106 #endif
3107 ipmr_rules_fail:
3108         ipmr_notifier_exit(net);
3109 ipmr_notifier_fail:
3110         return err;
3111 }
3112
3113 static void __net_exit ipmr_net_exit(struct net *net)
3114 {
3115 #ifdef CONFIG_PROC_FS
3116         remove_proc_entry("ip_mr_cache", net->proc_net);
3117         remove_proc_entry("ip_mr_vif", net->proc_net);
3118 #endif
3119         ipmr_notifier_exit(net);
3120 }
3121
3122 static void __net_exit ipmr_net_exit_batch(struct list_head *net_list)
3123 {
3124         struct net *net;
3125
3126         rtnl_lock();
3127         list_for_each_entry(net, net_list, exit_list)
3128                 ipmr_rules_exit(net);
3129         rtnl_unlock();
3130 }
3131
3132 static struct pernet_operations ipmr_net_ops = {
3133         .init = ipmr_net_init,
3134         .exit = ipmr_net_exit,
3135         .exit_batch = ipmr_net_exit_batch,
3136 };
3137
3138 int __init ip_mr_init(void)
3139 {
3140         int err;
3141
3142         mrt_cachep = kmem_cache_create("ip_mrt_cache",
3143                                        sizeof(struct mfc_cache),
3144                                        0, SLAB_HWCACHE_ALIGN | SLAB_PANIC,
3145                                        NULL);
3146
3147         err = register_pernet_subsys(&ipmr_net_ops);
3148         if (err)
3149                 goto reg_pernet_fail;
3150
3151         err = register_netdevice_notifier(&ip_mr_notifier);
3152         if (err)
3153                 goto reg_notif_fail;
3154 #ifdef CONFIG_IP_PIMSM_V2
3155         if (inet_add_protocol(&pim_protocol, IPPROTO_PIM) < 0) {
3156                 pr_err("%s: can't add PIM protocol\n", __func__);
3157                 err = -EAGAIN;
3158                 goto add_proto_fail;
3159         }
3160 #endif
3161         rtnl_register(RTNL_FAMILY_IPMR, RTM_GETROUTE,
3162                       ipmr_rtm_getroute, ipmr_rtm_dumproute, 0);
3163         rtnl_register(RTNL_FAMILY_IPMR, RTM_NEWROUTE,
3164                       ipmr_rtm_route, NULL, 0);
3165         rtnl_register(RTNL_FAMILY_IPMR, RTM_DELROUTE,
3166                       ipmr_rtm_route, NULL, 0);
3167
3168         rtnl_register(RTNL_FAMILY_IPMR, RTM_GETLINK,
3169                       NULL, ipmr_rtm_dumplink, 0);
3170         return 0;
3171
3172 #ifdef CONFIG_IP_PIMSM_V2
3173 add_proto_fail:
3174         unregister_netdevice_notifier(&ip_mr_notifier);
3175 #endif
3176 reg_notif_fail:
3177         unregister_pernet_subsys(&ipmr_net_ops);
3178 reg_pernet_fail:
3179         kmem_cache_destroy(mrt_cachep);
3180         return err;
3181 }