net/ipv4/ipmr.c

   1 /*
   2  *      IP multicast routing support for mrouted 3.6/3.8
   3  *
   4  *              (c) 1995 Alan Cox, <alan@lxorguk.ukuu.org.uk>
   5  *        Linux Consultancy and Custom Driver Development
   6  *
   7  *      This program is free software; you can redistribute it and/or
   8  *      modify it under the terms of the GNU General Public License
   9  *      as published by the Free Software Foundation; either version
  10  *      2 of the License, or (at your option) any later version.
  11  *
  12  *      Fixes:
  13  *      Michael Chastain        :       Incorrect size of copying.
  14  *      Alan Cox                :       Added the cache manager code
  15  *      Alan Cox                :       Fixed the clone/copy bug and device race.
  16  *      Mike McLagan            :       Routing by source
  17  *      Malcolm Beattie         :       Buffer handling fixes.
  18  *      Alexey Kuznetsov        :       Double buffer free and other fixes.
  19  *      SVR Anand               :       Fixed several multicast bugs and problems.
  20  *      Alexey Kuznetsov        :       Status, optimisations and more.
  21  *      Brad Parker             :       Better behaviour on mrouted upcall
  22  *                                      overflow.
  23  *      Carlos Picoto           :       PIMv1 Support
  24  *      Pavlin Ivanov Radoslavov:       PIMv2 Registers must checksum only PIM header
  25  *                                      Relax this requirement to work with older peers.
  26  *
  27  */
  28
  29 #include <linux/uaccess.h>
  30 #include <linux/types.h>
  31 #include <linux/capability.h>
  32 #include <linux/errno.h>
  33 #include <linux/timer.h>
  34 #include <linux/mm.h>
  35 #include <linux/kernel.h>
  36 #include <linux/fcntl.h>
  37 #include <linux/stat.h>
  38 #include <linux/socket.h>
  39 #include <linux/in.h>
  40 #include <linux/inet.h>
  41 #include <linux/netdevice.h>
  42 #include <linux/inetdevice.h>
  43 #include <linux/igmp.h>
  44 #include <linux/proc_fs.h>
  45 #include <linux/seq_file.h>
  46 #include <linux/mroute.h>
  47 #include <linux/init.h>
  48 #include <linux/if_ether.h>
  49 #include <linux/slab.h>
  50 #include <net/net_namespace.h>
  51 #include <net/ip.h>
  52 #include <net/protocol.h>
  53 #include <linux/skbuff.h>
  54 #include <net/route.h>
  55 #include <net/sock.h>
  56 #include <net/icmp.h>
  57 #include <net/udp.h>
  58 #include <net/raw.h>
  59 #include <linux/notifier.h>
  60 #include <linux/if_arp.h>
  61 #include <linux/netfilter_ipv4.h>
  62 #include <linux/compat.h>
  63 #include <linux/export.h>
  64 #include <net/ip_tunnels.h>
  65 #include <net/checksum.h>
  66 #include <net/netlink.h>
  67 #include <net/fib_rules.h>
  68 #include <linux/netconf.h>
  69 #include <net/nexthop.h>
  70
  71 struct ipmr_rule {
  72         struct fib_rule         common;
  73 };
  74
  75 struct ipmr_result {
  76         struct mr_table         *mrt;
  77 };
  78
  79 /* Big lock, protecting vif table, mrt cache and mroute socket state.
  80  * Note that the changes are semaphored via rtnl_lock.
  81  */
  82
  83 static DEFINE_RWLOCK(mrt_lock);
  84
  85 /* Multicast router control variables */
  86
  87 /* Special spinlock for queue of unresolved entries */
  88 static DEFINE_SPINLOCK(mfc_unres_lock);
  89
  90 /* We return to original Alan's scheme. Hash table of resolved
  91  * entries is changed only in process context and protected
  92  * with weak lock mrt_lock. Queue of unresolved entries is protected
  93  * with strong spinlock mfc_unres_lock.
  94  *
  95  * In this case data path is free of exclusive locks at all.
  96  */
  97
  98 static struct kmem_cache *mrt_cachep __read_mostly;
  99
 100 static struct mr_table *ipmr_new_table(struct net *net, u32 id);
 101 static void ipmr_free_table(struct mr_table *mrt);
 102
 103 static void ip_mr_forward(struct net *net, struct mr_table *mrt,
 104                           struct net_device *dev, struct sk_buff *skb,
 105                           struct mfc_cache *cache, int local);
 106 static int ipmr_cache_report(struct mr_table *mrt,
 107                              struct sk_buff *pkt, vifi_t vifi, int assert);
 108 static int __ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
 109                               struct mfc_cache *c, struct rtmsg *rtm);
 110 static void mroute_netlink_event(struct mr_table *mrt, struct mfc_cache *mfc,
 111                                  int cmd);
 112 static void igmpmsg_netlink_event(struct mr_table *mrt, struct sk_buff *pkt);
 113 static void mroute_clean_tables(struct mr_table *mrt, bool all);
 114 static void ipmr_expire_process(unsigned long arg);
 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
 120 static struct mr_table *ipmr_get_table(struct net *net, u32 id)
 121 {
 122         struct mr_table *mrt;
 123
 124         ipmr_for_each_table(mrt, net) {
 125                 if (mrt->id == id)
 126                         return mrt;
 127         }
 128         return NULL;
 129 }
 130
 131 static int ipmr_fib_lookup(struct net *net, struct flowi4 *flp4,
 132                            struct mr_table **mrt)
 133 {
 134         int err;
 135         struct ipmr_result res;
 136         struct fib_lookup_arg arg = {
 137                 .result = &res,
 138                 .flags = FIB_LOOKUP_NOREF,
 139         };
 140
 141         /* update flow if oif or iif point to device enslaved to l3mdev */
 142         l3mdev_update_flow(net, flowi4_to_flowi(flp4));
 143
 144         err = fib_rules_lookup(net->ipv4.mr_rules_ops,
 145                                flowi4_to_flowi(flp4), 0, &arg);
 146         if (err < 0)
 147                 return err;
 148         *mrt = res.mrt;
 149         return 0;
 150 }
 151
 152 static int ipmr_rule_action(struct fib_rule *rule, struct flowi *flp,
 153                             int flags, struct fib_lookup_arg *arg)
 154 {
 155         struct ipmr_result *res = arg->result;
 156         struct mr_table *mrt;
 157
 158         switch (rule->action) {
 159         case FR_ACT_TO_TBL:
 160                 break;
 161         case FR_ACT_UNREACHABLE:
 162                 return -ENETUNREACH;
 163         case FR_ACT_PROHIBIT:
 164                 return -EACCES;
 165         case FR_ACT_BLACKHOLE:
 166         default:
 167                 return -EINVAL;
 168         }
 169
 170         arg->table = fib_rule_get_table(rule, arg);
 171
 172         mrt = ipmr_get_table(rule->fr_net, arg->table);
 173         if (!mrt)
 174                 return -EAGAIN;
 175         res->mrt = mrt;
 176         return 0;
 177 }
 178
 179 static int ipmr_rule_match(struct fib_rule *rule, struct flowi *fl, int flags)
 180 {
 181         return 1;
 182 }
 183
 184 static const struct nla_policy ipmr_rule_policy[FRA_MAX + 1] = {
 185         FRA_GENERIC_POLICY,
 186 };
 187
 188 static int ipmr_rule_configure(struct fib_rule *rule, struct sk_buff *skb,
 189                                struct fib_rule_hdr *frh, struct nlattr **tb)
 190 {
 191         return 0;
 192 }
 193
 194 static int ipmr_rule_compare(struct fib_rule *rule, struct fib_rule_hdr *frh,
 195                              struct nlattr **tb)
 196 {
 197         return 1;
 198 }
 199
 200 static int ipmr_rule_fill(struct fib_rule *rule, struct sk_buff *skb,
 201                           struct fib_rule_hdr *frh)
 202 {
 203         frh->dst_len = 0;
 204         frh->src_len = 0;
 205         frh->tos     = 0;
 206         return 0;
 207 }
 208
 209 static const struct fib_rules_ops __net_initconst ipmr_rules_ops_template = {
 210         .family         = RTNL_FAMILY_IPMR,
 211         .rule_size      = sizeof(struct ipmr_rule),
 212         .addr_size      = sizeof(u32),
 213         .action         = ipmr_rule_action,
 214         .match          = ipmr_rule_match,
 215         .configure      = ipmr_rule_configure,
 216         .compare        = ipmr_rule_compare,
 217         .fill           = ipmr_rule_fill,
 218         .nlgroup        = RTNLGRP_IPV4_RULE,
 219         .policy         = ipmr_rule_policy,
 220         .owner          = THIS_MODULE,
 221 };
 222
 223 static int __net_init ipmr_rules_init(struct net *net)
 224 {
 225         struct fib_rules_ops *ops;
 226         struct mr_table *mrt;
 227         int err;
 228
 229         ops = fib_rules_register(&ipmr_rules_ops_template, net);
 230         if (IS_ERR(ops))
 231                 return PTR_ERR(ops);
 232
 233         INIT_LIST_HEAD(&net->ipv4.mr_tables);
 234
 235         mrt = ipmr_new_table(net, RT_TABLE_DEFAULT);
 236         if (IS_ERR(mrt)) {
 237                 err = PTR_ERR(mrt);
 238                 goto err1;
 239         }
 240
 241         err = fib_default_rule_add(ops, 0x7fff, RT_TABLE_DEFAULT, 0);
 242         if (err < 0)
 243                 goto err2;
 244
 245         net->ipv4.mr_rules_ops = ops;
 246         return 0;
 247
 248 err2:
 249         ipmr_free_table(mrt);
 250 err1:
 251         fib_rules_unregister(ops);
 252         return err;
 253 }
 254
 255 static void __net_exit ipmr_rules_exit(struct net *net)
 256 {
 257         struct mr_table *mrt, *next;
 258
 259         rtnl_lock();
 260         list_for_each_entry_safe(mrt, next, &net->ipv4.mr_tables, list) {
 261                 list_del(&mrt->list);
 262                 ipmr_free_table(mrt);
 263         }
 264         fib_rules_unregister(net->ipv4.mr_rules_ops);
 265         rtnl_unlock();
 266 }
 267
 268 static int ipmr_rules_dump(struct net *net, struct notifier_block *nb)
 269 {
 270         return fib_rules_dump(net, nb, RTNL_FAMILY_IPMR);
 271 }
 272
 273 static unsigned int ipmr_rules_seq_read(struct net *net)
 274 {
 275         return fib_rules_seq_read(net, RTNL_FAMILY_IPMR);
 276 }
 277
 278 bool ipmr_rule_default(const struct fib_rule *rule)
 279 {
 280         return fib_rule_matchall(rule) && rule->table == RT_TABLE_DEFAULT;
 281 }
 282 EXPORT_SYMBOL(ipmr_rule_default);
 283 #else
 284 #define ipmr_for_each_table(mrt, net) \
 285         for (mrt = net->ipv4.mrt; mrt; mrt = NULL)
 286
 287 static struct mr_table *ipmr_get_table(struct net *net, u32 id)
 288 {
 289         return net->ipv4.mrt;
 290 }
 291
 292 static int ipmr_fib_lookup(struct net *net, struct flowi4 *flp4,
 293                            struct mr_table **mrt)
 294 {
 295         *mrt = net->ipv4.mrt;
 296         return 0;
 297 }
 298
 299 static int __net_init ipmr_rules_init(struct net *net)
 300 {
 301         struct mr_table *mrt;
 302
 303         mrt = ipmr_new_table(net, RT_TABLE_DEFAULT);
 304         if (IS_ERR(mrt))
 305                 return PTR_ERR(mrt);
 306         net->ipv4.mrt = mrt;
 307         return 0;
 308 }
 309
 310 static void __net_exit ipmr_rules_exit(struct net *net)
 311 {
 312         rtnl_lock();
 313         ipmr_free_table(net->ipv4.mrt);
 314         net->ipv4.mrt = NULL;
 315         rtnl_unlock();
 316 }
 317
 318 static int ipmr_rules_dump(struct net *net, struct notifier_block *nb)
 319 {
 320         return 0;
 321 }
 322
 323 static unsigned int ipmr_rules_seq_read(struct net *net)
 324 {
 325         return 0;
 326 }
 327
 328 bool ipmr_rule_default(const struct fib_rule *rule)
 329 {
 330         return true;
 331 }
 332 EXPORT_SYMBOL(ipmr_rule_default);
 333 #endif
 334
 335 static inline int ipmr_hash_cmp(struct rhashtable_compare_arg *arg,
 336                                 const void *ptr)
 337 {
 338         const struct mfc_cache_cmp_arg *cmparg = arg->key;
 339         struct mfc_cache *c = (struct mfc_cache *)ptr;
 340
 341         return cmparg->mfc_mcastgrp != c->mfc_mcastgrp ||
 342                cmparg->mfc_origin != c->mfc_origin;
 343 }
 344
 345 static const struct rhashtable_params ipmr_rht_params = {
 346         .head_offset = offsetof(struct mfc_cache, mnode),
 347         .key_offset = offsetof(struct mfc_cache, cmparg),
 348         .key_len = sizeof(struct mfc_cache_cmp_arg),
 349         .nelem_hint = 3,
 350         .locks_mul = 1,
 351         .obj_cmpfn = ipmr_hash_cmp,
 352         .automatic_shrinking = true,
 353 };
 354
 355 static struct mr_table *ipmr_new_table(struct net *net, u32 id)
 356 {
 357         struct mr_table *mrt;
 358
 359         /* "pimreg%u" should not exceed 16 bytes (IFNAMSIZ) */
 360         if (id != RT_TABLE_DEFAULT && id >= 1000000000)
 361                 return ERR_PTR(-EINVAL);
 362
 363         mrt = ipmr_get_table(net, id);
 364         if (mrt)
 365                 return mrt;
 366
 367         mrt = kzalloc(sizeof(*mrt), GFP_KERNEL);
 368         if (!mrt)
 369                 return ERR_PTR(-ENOMEM);
 370         write_pnet(&mrt->net, net);
 371         mrt->id = id;
 372
 373         rhltable_init(&mrt->mfc_hash, &ipmr_rht_params);
 374         INIT_LIST_HEAD(&mrt->mfc_cache_list);
 375         INIT_LIST_HEAD(&mrt->mfc_unres_queue);
 376
 377         setup_timer(&mrt->ipmr_expire_timer, ipmr_expire_process,
 378                     (unsigned long)mrt);
 379
 380         mrt->mroute_reg_vif_num = -1;
 381 #ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES
 382         list_add_tail_rcu(&mrt->list, &net->ipv4.mr_tables);
 383 #endif
 384         return mrt;
 385 }
 386
 387 static void ipmr_free_table(struct mr_table *mrt)
 388 {
 389         del_timer_sync(&mrt->ipmr_expire_timer);
 390         mroute_clean_tables(mrt, true);
 391         rhltable_destroy(&mrt->mfc_hash);
 392         kfree(mrt);
 393 }
 394
 395 /* Service routines creating virtual interfaces: DVMRP tunnels and PIMREG */
 396
 397 static void ipmr_del_tunnel(struct net_device *dev, struct vifctl *v)
 398 {
 399         struct net *net = dev_net(dev);
 400
 401         dev_close(dev);
 402
 403         dev = __dev_get_by_name(net, "tunl0");
 404         if (dev) {
 405                 const struct net_device_ops *ops = dev->netdev_ops;
 406                 struct ifreq ifr;
 407                 struct ip_tunnel_parm p;
 408
 409                 memset(&p, 0, sizeof(p));
 410                 p.iph.daddr = v->vifc_rmt_addr.s_addr;
 411                 p.iph.saddr = v->vifc_lcl_addr.s_addr;
 412                 p.iph.version = 4;
 413                 p.iph.ihl = 5;
 414                 p.iph.protocol = IPPROTO_IPIP;
 415                 sprintf(p.name, "dvmrp%d", v->vifc_vifi);
 416                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
 417
 418                 if (ops->ndo_do_ioctl) {
 419                         mm_segment_t oldfs = get_fs();
 420
 421                         set_fs(KERNEL_DS);
 422                         ops->ndo_do_ioctl(dev, &ifr, SIOCDELTUNNEL);
 423                         set_fs(oldfs);
 424                 }
 425         }
 426 }
 427
 428 /* Initialize ipmr pimreg/tunnel in_device */
 429 static bool ipmr_init_vif_indev(const struct net_device *dev)
 430 {
 431         struct in_device *in_dev;
 432
 433         ASSERT_RTNL();
 434
 435         in_dev = __in_dev_get_rtnl(dev);
 436         if (!in_dev)
 437                 return false;
 438         ipv4_devconf_setall(in_dev);
 439         neigh_parms_data_state_setall(in_dev->arp_parms);
 440         IPV4_DEVCONF(in_dev->cnf, RP_FILTER) = 0;
 441
 442         return true;
 443 }
 444
 445 static struct net_device *ipmr_new_tunnel(struct net *net, struct vifctl *v)
 446 {
 447         struct net_device  *dev;
 448
 449         dev = __dev_get_by_name(net, "tunl0");
 450
 451         if (dev) {
 452                 const struct net_device_ops *ops = dev->netdev_ops;
 453                 int err;
 454                 struct ifreq ifr;
 455                 struct ip_tunnel_parm p;
 456
 457                 memset(&p, 0, sizeof(p));
 458                 p.iph.daddr = v->vifc_rmt_addr.s_addr;
 459                 p.iph.saddr = v->vifc_lcl_addr.s_addr;
 460                 p.iph.version = 4;
 461                 p.iph.ihl = 5;
 462                 p.iph.protocol = IPPROTO_IPIP;
 463                 sprintf(p.name, "dvmrp%d", v->vifc_vifi);
 464                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
 465
 466                 if (ops->ndo_do_ioctl) {
 467                         mm_segment_t oldfs = get_fs();
 468
 469                         set_fs(KERNEL_DS);
 470                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
 471                         set_fs(oldfs);
 472                 } else {
 473                         err = -EOPNOTSUPP;
 474                 }
 475                 dev = NULL;
 476
 477                 if (err == 0 &&
 478                     (dev = __dev_get_by_name(net, p.name)) != NULL) {
 479                         dev->flags |= IFF_MULTICAST;
 480                         if (!ipmr_init_vif_indev(dev))
 481                                 goto failure;
 482                         if (dev_open(dev))
 483                                 goto failure;
 484                         dev_hold(dev);
 485                 }
 486         }
 487         return dev;
 488
 489 failure:
 490         unregister_netdevice(dev);
 491         return NULL;
 492 }
 493
 494 #if defined(CONFIG_IP_PIMSM_V1) || defined(CONFIG_IP_PIMSM_V2)
 495 static netdev_tx_t reg_vif_xmit(struct sk_buff *skb, struct net_device *dev)
 496 {
 497         struct net *net = dev_net(dev);
 498         struct mr_table *mrt;
 499         struct flowi4 fl4 = {
 500                 .flowi4_oif     = dev->ifindex,
 501                 .flowi4_iif     = skb->skb_iif ? : LOOPBACK_IFINDEX,
 502                 .flowi4_mark    = skb->mark,
 503         };
 504         int err;
 505
 506         err = ipmr_fib_lookup(net, &fl4, &mrt);
 507         if (err < 0) {
 508                 kfree_skb(skb);
 509                 return err;
 510         }
 511
 512         read_lock(&mrt_lock);
 513         dev->stats.tx_bytes += skb->len;
 514         dev->stats.tx_packets++;
 515         ipmr_cache_report(mrt, skb, mrt->mroute_reg_vif_num, IGMPMSG_WHOLEPKT);
 516         read_unlock(&mrt_lock);
 517         kfree_skb(skb);
 518         return NETDEV_TX_OK;
 519 }
 520
 521 static int reg_vif_get_iflink(const struct net_device *dev)
 522 {
 523         return 0;
 524 }
 525
 526 static const struct net_device_ops reg_vif_netdev_ops = {
 527         .ndo_start_xmit = reg_vif_xmit,
 528         .ndo_get_iflink = reg_vif_get_iflink,
 529 };
 530
 531 static void reg_vif_setup(struct net_device *dev)
 532 {
 533         dev->type               = ARPHRD_PIMREG;
 534         dev->mtu                = ETH_DATA_LEN - sizeof(struct iphdr) - 8;
 535         dev->flags              = IFF_NOARP;
 536         dev->netdev_ops         = &reg_vif_netdev_ops;
 537         dev->needs_free_netdev  = true;
 538         dev->features           |= NETIF_F_NETNS_LOCAL;
 539 }
 540
 541 static struct net_device *ipmr_reg_vif(struct net *net, struct mr_table *mrt)
 542 {
 543         struct net_device *dev;
 544         char name[IFNAMSIZ];
 545
 546         if (mrt->id == RT_TABLE_DEFAULT)
 547                 sprintf(name, "pimreg");
 548         else
 549                 sprintf(name, "pimreg%u", mrt->id);
 550
 551         dev = alloc_netdev(0, name, NET_NAME_UNKNOWN, reg_vif_setup);
 552
 553         if (!dev)
 554                 return NULL;
 555
 556         dev_net_set(dev, net);
 557
 558         if (register_netdevice(dev)) {
 559                 free_netdev(dev);
 560                 return NULL;
 561         }
 562
 563         if (!ipmr_init_vif_indev(dev))
 564                 goto failure;
 565         if (dev_open(dev))
 566                 goto failure;
 567
 568         dev_hold(dev);
 569
 570         return dev;
 571
 572 failure:
 573         unregister_netdevice(dev);
 574         return NULL;
 575 }
 576
 577 /* called with rcu_read_lock() */
 578 static int __pim_rcv(struct mr_table *mrt, struct sk_buff *skb,
 579                      unsigned int pimlen)
 580 {
 581         struct net_device *reg_dev = NULL;
 582         struct iphdr *encap;
 583
 584         encap = (struct iphdr *)(skb_transport_header(skb) + pimlen);
 585         /* Check that:
 586          * a. packet is really sent to a multicast group
 587          * b. packet is not a NULL-REGISTER
 588          * c. packet is not truncated
 589          */
 590         if (!ipv4_is_multicast(encap->daddr) ||
 591             encap->tot_len == 0 ||
 592             ntohs(encap->tot_len) + pimlen > skb->len)
 593                 return 1;
 594
 595         read_lock(&mrt_lock);
 596         if (mrt->mroute_reg_vif_num >= 0)
 597                 reg_dev = mrt->vif_table[mrt->mroute_reg_vif_num].dev;
 598         read_unlock(&mrt_lock);
 599
 600         if (!reg_dev)
 601                 return 1;
 602
 603         skb->mac_header = skb->network_header;
 604         skb_pull(skb, (u8 *)encap - skb->data);
 605         skb_reset_network_header(skb);
 606         skb->protocol = htons(ETH_P_IP);
 607         skb->ip_summed = CHECKSUM_NONE;
 608
 609         skb_tunnel_rx(skb, reg_dev, dev_net(reg_dev));
 610
 611         netif_rx(skb);
 612
 613         return NET_RX_SUCCESS;
 614 }
 615 #else
 616 static struct net_device *ipmr_reg_vif(struct net *net, struct mr_table *mrt)
 617 {
 618         return NULL;
 619 }
 620 #endif
 621
 622 static int call_ipmr_vif_entry_notifier(struct notifier_block *nb,
 623                                         struct net *net,
 624                                         enum fib_event_type event_type,
 625                                         struct vif_device *vif,
 626                                         vifi_t vif_index, u32 tb_id)
 627 {
 628         struct vif_entry_notifier_info info = {
 629                 .info = {
 630                         .family = RTNL_FAMILY_IPMR,
 631                         .net = net,
 632                 },
 633                 .dev = vif->dev,
 634                 .vif_index = vif_index,
 635                 .vif_flags = vif->flags,
 636                 .tb_id = tb_id,
 637         };
 638
 639         return call_fib_notifier(nb, net, event_type, &info.info);
 640 }
 641
 642 static int call_ipmr_vif_entry_notifiers(struct net *net,
 643                                          enum fib_event_type event_type,
 644                                          struct vif_device *vif,
 645                                          vifi_t vif_index, u32 tb_id)
 646 {
 647         struct vif_entry_notifier_info info = {
 648                 .info = {
 649                         .family = RTNL_FAMILY_IPMR,
 650                         .net = net,
 651                 },
 652                 .dev = vif->dev,
 653                 .vif_index = vif_index,
 654                 .vif_flags = vif->flags,
 655                 .tb_id = tb_id,
 656         };
 657
 658         ASSERT_RTNL();
 659         net->ipv4.ipmr_seq++;
 660         return call_fib_notifiers(net, event_type, &info.info);
 661 }
 662
 663 static int call_ipmr_mfc_entry_notifier(struct notifier_block *nb,
 664                                         struct net *net,
 665                                         enum fib_event_type event_type,
 666                                         struct mfc_cache *mfc, u32 tb_id)
 667 {
 668         struct mfc_entry_notifier_info info = {
 669                 .info = {
 670                         .family = RTNL_FAMILY_IPMR,
 671                         .net = net,
 672                 },
 673                 .mfc = mfc,
 674                 .tb_id = tb_id
 675         };
 676
 677         return call_fib_notifier(nb, net, event_type, &info.info);
 678 }
 679
 680 static int call_ipmr_mfc_entry_notifiers(struct net *net,
 681                                          enum fib_event_type event_type,
 682                                          struct mfc_cache *mfc, u32 tb_id)
 683 {
 684         struct mfc_entry_notifier_info info = {
 685                 .info = {
 686                         .family = RTNL_FAMILY_IPMR,
 687                         .net = net,
 688                 },
 689                 .mfc = mfc,
 690                 .tb_id = tb_id
 691         };
 692
 693         ASSERT_RTNL();
 694         net->ipv4.ipmr_seq++;
 695         return call_fib_notifiers(net, event_type, &info.info);
 696 }
 697
 698 /**
 699  *      vif_delete - Delete a VIF entry
 700  *      @notify: Set to 1, if the caller is a notifier_call
 701  */
 702 static int vif_delete(struct mr_table *mrt, int vifi, int notify,
 703                       struct list_head *head)
 704 {
 705         struct net *net = read_pnet(&mrt->net);
 706         struct vif_device *v;
 707         struct net_device *dev;
 708         struct in_device *in_dev;
 709
 710         if (vifi < 0 || vifi >= mrt->maxvif)
 711                 return -EADDRNOTAVAIL;
 712
 713         v = &mrt->vif_table[vifi];
 714
 715         if (VIF_EXISTS(mrt, vifi))
 716                 call_ipmr_vif_entry_notifiers(net, FIB_EVENT_VIF_DEL, v, vifi,
 717                                               mrt->id);
 718
 719         write_lock_bh(&mrt_lock);
 720         dev = v->dev;
 721         v->dev = NULL;
 722
 723         if (!dev) {
 724                 write_unlock_bh(&mrt_lock);
 725                 return -EADDRNOTAVAIL;
 726         }
 727
 728         if (vifi == mrt->mroute_reg_vif_num)
 729                 mrt->mroute_reg_vif_num = -1;
 730
 731         if (vifi + 1 == mrt->maxvif) {
 732                 int tmp;
 733
 734                 for (tmp = vifi - 1; tmp >= 0; tmp--) {
 735                         if (VIF_EXISTS(mrt, tmp))
 736                                 break;
 737                 }
 738                 mrt->maxvif = tmp+1;
 739         }
 740
 741         write_unlock_bh(&mrt_lock);
 742
 743         dev_set_allmulti(dev, -1);
 744
 745         in_dev = __in_dev_get_rtnl(dev);
 746         if (in_dev) {
 747                 IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)--;
 748                 inet_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
 749                                             NETCONFA_MC_FORWARDING,
 750                                             dev->ifindex, &in_dev->cnf);
 751                 ip_rt_multicast_event(in_dev);
 752         }
 753
 754         if (v->flags & (VIFF_TUNNEL | VIFF_REGISTER) && !notify)
 755                 unregister_netdevice_queue(dev, head);
 756
 757         dev_put(dev);
 758         return 0;
 759 }
 760
 761 static void ipmr_cache_free_rcu(struct rcu_head *head)
 762 {
 763         struct mfc_cache *c = container_of(head, struct mfc_cache, rcu);
 764
 765         kmem_cache_free(mrt_cachep, c);
 766 }
 767
 768 void ipmr_cache_free(struct mfc_cache *c)
 769 {
 770         call_rcu(&c->rcu, ipmr_cache_free_rcu);
 771 }
 772 EXPORT_SYMBOL(ipmr_cache_free);
 773
 774 /* Destroy an unresolved cache entry, killing queued skbs
 775  * and reporting error to netlink readers.
 776  */
 777 static void ipmr_destroy_unres(struct mr_table *mrt, struct mfc_cache *c)
 778 {
 779         struct net *net = read_pnet(&mrt->net);
 780         struct sk_buff *skb;
 781         struct nlmsgerr *e;
 782
 783         atomic_dec(&mrt->cache_resolve_queue_len);
 784
 785         while ((skb = skb_dequeue(&c->mfc_un.unres.unresolved))) {
 786                 if (ip_hdr(skb)->version == 0) {
 787                         struct nlmsghdr *nlh = skb_pull(skb,
 788                                                         sizeof(struct iphdr));
 789                         nlh->nlmsg_type = NLMSG_ERROR;
 790                         nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
 791                         skb_trim(skb, nlh->nlmsg_len);
 792                         e = nlmsg_data(nlh);
 793                         e->error = -ETIMEDOUT;
 794                         memset(&e->msg, 0, sizeof(e->msg));
 795
 796                         rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
 797                 } else {
 798                         kfree_skb(skb);
 799                 }
 800         }
 801
 802         ipmr_cache_free(c);
 803 }
 804
 805 /* Timer process for the unresolved queue. */
 806 static void ipmr_expire_process(unsigned long arg)
 807 {
 808         struct mr_table *mrt = (struct mr_table *)arg;
 809         unsigned long now;
 810         unsigned long expires;
 811         struct mfc_cache *c, *next;
 812
 813         if (!spin_trylock(&mfc_unres_lock)) {
 814                 mod_timer(&mrt->ipmr_expire_timer, jiffies+HZ/10);
 815                 return;
 816         }
 817
 818         if (list_empty(&mrt->mfc_unres_queue))
 819                 goto out;
 820
 821         now = jiffies;
 822         expires = 10*HZ;
 823
 824         list_for_each_entry_safe(c, next, &mrt->mfc_unres_queue, list) {
 825                 if (time_after(c->mfc_un.unres.expires, now)) {
 826                         unsigned long interval = c->mfc_un.unres.expires - now;
 827                         if (interval < expires)
 828                                 expires = interval;
 829                         continue;
 830                 }
 831
 832                 list_del(&c->list);
 833                 mroute_netlink_event(mrt, c, RTM_DELROUTE);
 834                 ipmr_destroy_unres(mrt, c);
 835         }
 836
 837         if (!list_empty(&mrt->mfc_unres_queue))
 838                 mod_timer(&mrt->ipmr_expire_timer, jiffies + expires);
 839
 840 out:
 841         spin_unlock(&mfc_unres_lock);
 842 }
 843
 844 /* Fill oifs list. It is called under write locked mrt_lock. */
 845 static void ipmr_update_thresholds(struct mr_table *mrt, struct mfc_cache *cache,
 846                                    unsigned char *ttls)
 847 {
 848         int vifi;
 849
 850         cache->mfc_un.res.minvif = MAXVIFS;
 851         cache->mfc_un.res.maxvif = 0;
 852         memset(cache->mfc_un.res.ttls, 255, MAXVIFS);
 853
 854         for (vifi = 0; vifi < mrt->maxvif; vifi++) {
 855                 if (VIF_EXISTS(mrt, vifi) &&
 856                     ttls[vifi] && ttls[vifi] < 255) {
 857                         cache->mfc_un.res.ttls[vifi] = ttls[vifi];
 858                         if (cache->mfc_un.res.minvif > vifi)
 859                                 cache->mfc_un.res.minvif = vifi;
 860                         if (cache->mfc_un.res.maxvif <= vifi)
 861                                 cache->mfc_un.res.maxvif = vifi + 1;
 862                 }
 863         }
 864         cache->mfc_un.res.lastuse = jiffies;
 865 }
 866
 867 static int vif_add(struct net *net, struct mr_table *mrt,
 868                    struct vifctl *vifc, int mrtsock)
 869 {
 870         int vifi = vifc->vifc_vifi;
 871         struct vif_device *v = &mrt->vif_table[vifi];
 872         struct net_device *dev;
 873         struct in_device *in_dev;
 874         int err;
 875
 876         /* Is vif busy ? */
 877         if (VIF_EXISTS(mrt, vifi))
 878                 return -EADDRINUSE;
 879
 880         switch (vifc->vifc_flags) {
 881         case VIFF_REGISTER:
 882                 if (!ipmr_pimsm_enabled())
 883                         return -EINVAL;
 884                 /* Special Purpose VIF in PIM
 885                  * All the packets will be sent to the daemon
 886                  */
 887                 if (mrt->mroute_reg_vif_num >= 0)
 888                         return -EADDRINUSE;
 889                 dev = ipmr_reg_vif(net, mrt);
 890                 if (!dev)
 891                         return -ENOBUFS;
 892                 err = dev_set_allmulti(dev, 1);
 893                 if (err) {
 894                         unregister_netdevice(dev);
 895                         dev_put(dev);
 896                         return err;
 897                 }
 898                 break;
 899         case VIFF_TUNNEL:
 900                 dev = ipmr_new_tunnel(net, vifc);
 901                 if (!dev)
 902                         return -ENOBUFS;
 903                 err = dev_set_allmulti(dev, 1);
 904                 if (err) {
 905                         ipmr_del_tunnel(dev, vifc);
 906                         dev_put(dev);
 907                         return err;
 908                 }
 909                 break;
 910         case VIFF_USE_IFINDEX:
 911         case 0:
 912                 if (vifc->vifc_flags == VIFF_USE_IFINDEX) {
 913                         dev = dev_get_by_index(net, vifc->vifc_lcl_ifindex);
 914                         if (dev && !__in_dev_get_rtnl(dev)) {
 915                                 dev_put(dev);
 916                                 return -EADDRNOTAVAIL;
 917                         }
 918                 } else {
 919                         dev = ip_dev_find(net, vifc->vifc_lcl_addr.s_addr);
 920                 }
 921                 if (!dev)
 922                         return -EADDRNOTAVAIL;
 923                 err = dev_set_allmulti(dev, 1);
 924                 if (err) {
 925                         dev_put(dev);
 926                         return err;
 927                 }
 928                 break;
 929         default:
 930                 return -EINVAL;
 931         }
 932
 933         in_dev = __in_dev_get_rtnl(dev);
 934         if (!in_dev) {
 935                 dev_put(dev);
 936                 return -EADDRNOTAVAIL;
 937         }
 938         IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)++;
 939         inet_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_MC_FORWARDING,
 940                                     dev->ifindex, &in_dev->cnf);
 941         ip_rt_multicast_event(in_dev);
 942
 943         /* Fill in the VIF structures */
 944
 945         v->rate_limit = vifc->vifc_rate_limit;
 946         v->local = vifc->vifc_lcl_addr.s_addr;
 947         v->remote = vifc->vifc_rmt_addr.s_addr;
 948         v->flags = vifc->vifc_flags;
 949         if (!mrtsock)
 950                 v->flags |= VIFF_STATIC;
 951         v->threshold = vifc->vifc_threshold;
 952         v->bytes_in = 0;
 953         v->bytes_out = 0;
 954         v->pkt_in = 0;
 955         v->pkt_out = 0;
 956         v->link = dev->ifindex;
 957         if (v->flags & (VIFF_TUNNEL | VIFF_REGISTER))
 958                 v->link = dev_get_iflink(dev);
 959
 960         /* And finish update writing critical data */
 961         write_lock_bh(&mrt_lock);
 962         v->dev = dev;
 963         if (v->flags & VIFF_REGISTER)
 964                 mrt->mroute_reg_vif_num = vifi;
 965         if (vifi+1 > mrt->maxvif)
 966                 mrt->maxvif = vifi+1;
 967         write_unlock_bh(&mrt_lock);
 968         call_ipmr_vif_entry_notifiers(net, FIB_EVENT_VIF_ADD, v, vifi, mrt->id);
 969         return 0;
 970 }
 971
 972 /* called with rcu_read_lock() */
 973 static struct mfc_cache *ipmr_cache_find(struct mr_table *mrt,
 974                                          __be32 origin,
 975                                          __be32 mcastgrp)
 976 {
 977         struct mfc_cache_cmp_arg arg = {
 978                         .mfc_mcastgrp = mcastgrp,
 979                         .mfc_origin = origin
 980         };
 981         struct rhlist_head *tmp, *list;
 982         struct mfc_cache *c;
 983
 984         list = rhltable_lookup(&mrt->mfc_hash, &arg, ipmr_rht_params);
 985         rhl_for_each_entry_rcu(c, tmp, list, mnode)
 986                 return c;
 987
 988         return NULL;
 989 }
 990
 991 /* Look for a (*,*,oif) entry */
 992 static struct mfc_cache *ipmr_cache_find_any_parent(struct mr_table *mrt,
 993                                                     int vifi)
 994 {
 995         struct mfc_cache_cmp_arg arg = {
 996                         .mfc_mcastgrp = htonl(INADDR_ANY),
 997                         .mfc_origin = htonl(INADDR_ANY)
 998         };
 999         struct rhlist_head *tmp, *list;
1000         struct mfc_cache *c;
1001
1002         list = rhltable_lookup(&mrt->mfc_hash, &arg, ipmr_rht_params);
1003         rhl_for_each_entry_rcu(c, tmp, list, mnode)
1004                 if (c->mfc_un.res.ttls[vifi] < 255)
1005                         return c;
1006
1007         return NULL;
1008 }
1009
1010 /* Look for a (*,G) entry */
1011 static struct mfc_cache *ipmr_cache_find_any(struct mr_table *mrt,
1012                                              __be32 mcastgrp, int vifi)
1013 {
1014         struct mfc_cache_cmp_arg arg = {
1015                         .mfc_mcastgrp = mcastgrp,
1016                         .mfc_origin = htonl(INADDR_ANY)
1017         };
1018         struct rhlist_head *tmp, *list;
1019         struct mfc_cache *c, *proxy;
1020
1021         if (mcastgrp == htonl(INADDR_ANY))
1022                 goto skip;
1023
1024         list = rhltable_lookup(&mrt->mfc_hash, &arg, ipmr_rht_params);
1025         rhl_for_each_entry_rcu(c, tmp, list, mnode) {
1026                 if (c->mfc_un.res.ttls[vifi] < 255)
1027                         return c;
1028
1029                 /* It's ok if the vifi is part of the static tree */
1030                 proxy = ipmr_cache_find_any_parent(mrt, c->mfc_parent);
1031                 if (proxy && proxy->mfc_un.res.ttls[vifi] < 255)
1032                         return c;
1033         }
1034
1035 skip:
1036         return ipmr_cache_find_any_parent(mrt, vifi);
1037 }
1038
1039 /* Look for a (S,G,iif) entry if parent != -1 */
1040 static struct mfc_cache *ipmr_cache_find_parent(struct mr_table *mrt,
1041                                                 __be32 origin, __be32 mcastgrp,
1042                                                 int parent)
1043 {
1044         struct mfc_cache_cmp_arg arg = {
1045                         .mfc_mcastgrp = mcastgrp,
1046                         .mfc_origin = origin,
1047         };
1048         struct rhlist_head *tmp, *list;
1049         struct mfc_cache *c;
1050
1051         list = rhltable_lookup(&mrt->mfc_hash, &arg, ipmr_rht_params);
1052         rhl_for_each_entry_rcu(c, tmp, list, mnode)
1053                 if (parent == -1 || parent == c->mfc_parent)
1054                         return c;
1055
1056         return NULL;
1057 }
1058
1059 /* Allocate a multicast cache entry */
1060 static struct mfc_cache *ipmr_cache_alloc(void)
1061 {
1062         struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_KERNEL);
1063
1064         if (c) {
1065                 c->mfc_un.res.last_assert = jiffies - MFC_ASSERT_THRESH - 1;
1066                 c->mfc_un.res.minvif = MAXVIFS;
1067                 refcount_set(&c->mfc_un.res.refcount, 1);
1068         }
1069         return c;
1070 }
1071
1072 static struct mfc_cache *ipmr_cache_alloc_unres(void)
1073 {
1074         struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_ATOMIC);
1075
1076         if (c) {
1077                 skb_queue_head_init(&c->mfc_un.unres.unresolved);
1078                 c->mfc_un.unres.expires = jiffies + 10*HZ;
1079         }
1080         return c;
1081 }
1082
1083 /* A cache entry has gone into a resolved state from queued */
1084 static void ipmr_cache_resolve(struct net *net, struct mr_table *mrt,
1085                                struct mfc_cache *uc, struct mfc_cache *c)
1086 {
1087         struct sk_buff *skb;
1088         struct nlmsgerr *e;
1089
1090         /* Play the pending entries through our router */
1091         while ((skb = __skb_dequeue(&uc->mfc_un.unres.unresolved))) {
1092                 if (ip_hdr(skb)->version == 0) {
1093                         struct nlmsghdr *nlh = skb_pull(skb,
1094                                                         sizeof(struct iphdr));
1095
1096                         if (__ipmr_fill_mroute(mrt, skb, c, nlmsg_data(nlh)) > 0) {
1097                                 nlh->nlmsg_len = skb_tail_pointer(skb) -
1098                                                  (u8 *)nlh;
1099                         } else {
1100                                 nlh->nlmsg_type = NLMSG_ERROR;
1101                                 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
1102                                 skb_trim(skb, nlh->nlmsg_len);
1103                                 e = nlmsg_data(nlh);
1104                                 e->error = -EMSGSIZE;
1105                                 memset(&e->msg, 0, sizeof(e->msg));
1106                         }
1107
1108                         rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
1109                 } else {
1110                         ip_mr_forward(net, mrt, skb->dev, skb, c, 0);
1111                 }
1112         }
1113 }
1114
1115 /* Bounce a cache query up to mrouted and netlink.
1116  *
1117  * Called under mrt_lock.
1118  */
1119 static int ipmr_cache_report(struct mr_table *mrt,
1120                              struct sk_buff *pkt, vifi_t vifi, int assert)
1121 {
1122         const int ihl = ip_hdrlen(pkt);
1123         struct sock *mroute_sk;
1124         struct igmphdr *igmp;
1125         struct igmpmsg *msg;
1126         struct sk_buff *skb;
1127         int ret;
1128
1129         if (assert == IGMPMSG_WHOLEPKT)
1130                 skb = skb_realloc_headroom(pkt, sizeof(struct iphdr));
1131         else
1132                 skb = alloc_skb(128, GFP_ATOMIC);
1133
1134         if (!skb)
1135                 return -ENOBUFS;
1136
1137         if (assert == IGMPMSG_WHOLEPKT) {
1138                 /* Ugly, but we have no choice with this interface.
1139                  * Duplicate old header, fix ihl, length etc.
1140                  * And all this only to mangle msg->im_msgtype and
1141                  * to set msg->im_mbz to "mbz" :-)
1142                  */
1143                 skb_push(skb, sizeof(struct iphdr));
1144                 skb_reset_network_header(skb);
1145                 skb_reset_transport_header(skb);
1146                 msg = (struct igmpmsg *)skb_network_header(skb);
1147                 memcpy(msg, skb_network_header(pkt), sizeof(struct iphdr));
1148                 msg->im_msgtype = IGMPMSG_WHOLEPKT;
1149                 msg->im_mbz = 0;
1150                 msg->im_vif = mrt->mroute_reg_vif_num;
1151                 ip_hdr(skb)->ihl = sizeof(struct iphdr) >> 2;
1152                 ip_hdr(skb)->tot_len = htons(ntohs(ip_hdr(pkt)->tot_len) +
1153                                              sizeof(struct iphdr));
1154         } else {
1155                 /* Copy the IP header */
1156                 skb_set_network_header(skb, skb->len);
1157                 skb_put(skb, ihl);
1158                 skb_copy_to_linear_data(skb, pkt->data, ihl);
1159                 /* Flag to the kernel this is a route add */
1160                 ip_hdr(skb)->protocol = 0;
1161                 msg = (struct igmpmsg *)skb_network_header(skb);
1162                 msg->im_vif = vifi;
1163                 skb_dst_set(skb, dst_clone(skb_dst(pkt)));
1164                 /* Add our header */
1165                 igmp = skb_put(skb, sizeof(struct igmphdr));
1166                 igmp->type = assert;
1167                 msg->im_msgtype = assert;
1168                 igmp->code = 0;
1169                 ip_hdr(skb)->tot_len = htons(skb->len); /* Fix the length */
1170                 skb->transport_header = skb->network_header;
1171         }
1172
1173         rcu_read_lock();
1174         mroute_sk = rcu_dereference(mrt->mroute_sk);
1175         if (!mroute_sk) {
1176                 rcu_read_unlock();
1177                 kfree_skb(skb);
1178                 return -EINVAL;
1179         }
1180
1181         igmpmsg_netlink_event(mrt, skb);
1182
1183         /* Deliver to mrouted */
1184         ret = sock_queue_rcv_skb(mroute_sk, skb);
1185         rcu_read_unlock();
1186         if (ret < 0) {
1187                 net_warn_ratelimited("mroute: pending queue full, dropping entries\n");
1188                 kfree_skb(skb);
1189         }
1190
1191         return ret;
1192 }
1193
1194 /* Queue a packet for resolution. It gets locked cache entry! */
1195 static int ipmr_cache_unresolved(struct mr_table *mrt, vifi_t vifi,
1196                                  struct sk_buff *skb, struct net_device *dev)
1197 {
1198         const struct iphdr *iph = ip_hdr(skb);
1199         struct mfc_cache *c;
1200         bool found = false;
1201         int err;
1202
1203         spin_lock_bh(&mfc_unres_lock);
1204         list_for_each_entry(c, &mrt->mfc_unres_queue, list) {
1205                 if (c->mfc_mcastgrp == iph->daddr &&
1206                     c->mfc_origin == iph->saddr) {
1207                         found = true;
1208                         break;
1209                 }
1210         }
1211
1212         if (!found) {
1213                 /* Create a new entry if allowable */
1214                 if (atomic_read(&mrt->cache_resolve_queue_len) >= 10 ||
1215                     (c = ipmr_cache_alloc_unres()) == NULL) {
1216                         spin_unlock_bh(&mfc_unres_lock);
1217
1218                         kfree_skb(skb);
1219                         return -ENOBUFS;
1220                 }
1221
1222                 /* Fill in the new cache entry */
1223                 c->mfc_parent   = -1;
1224                 c->mfc_origin   = iph->saddr;
1225                 c->mfc_mcastgrp = iph->daddr;
1226
1227                 /* Reflect first query at mrouted. */
1228                 err = ipmr_cache_report(mrt, skb, vifi, IGMPMSG_NOCACHE);
1229                 if (err < 0) {
1230                         /* If the report failed throw the cache entry
1231                            out - Brad Parker
1232                          */
1233                         spin_unlock_bh(&mfc_unres_lock);
1234
1235                         ipmr_cache_free(c);
1236                         kfree_skb(skb);
1237                         return err;
1238                 }
1239
1240                 atomic_inc(&mrt->cache_resolve_queue_len);
1241                 list_add(&c->list, &mrt->mfc_unres_queue);
1242                 mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1243
1244                 if (atomic_read(&mrt->cache_resolve_queue_len) == 1)
1245                         mod_timer(&mrt->ipmr_expire_timer, c->mfc_un.unres.expires);
1246         }
1247
1248         /* See if we can append the packet */
1249         if (c->mfc_un.unres.unresolved.qlen > 3) {
1250                 kfree_skb(skb);
1251                 err = -ENOBUFS;
1252         } else {
1253                 if (dev) {
1254                         skb->dev = dev;
1255                         skb->skb_iif = dev->ifindex;
1256                 }
1257                 skb_queue_tail(&c->mfc_un.unres.unresolved, skb);
1258                 err = 0;
1259         }
1260
1261         spin_unlock_bh(&mfc_unres_lock);
1262         return err;
1263 }
1264
1265 /* MFC cache manipulation by user space mroute daemon */
1266
1267 static int ipmr_mfc_delete(struct mr_table *mrt, struct mfcctl *mfc, int parent)
1268 {
1269         struct net *net = read_pnet(&mrt->net);
1270         struct mfc_cache *c;
1271
1272         /* The entries are added/deleted only under RTNL */
1273         rcu_read_lock();
1274         c = ipmr_cache_find_parent(mrt, mfc->mfcc_origin.s_addr,
1275                                    mfc->mfcc_mcastgrp.s_addr, parent);
1276         rcu_read_unlock();
1277         if (!c)
1278                 return -ENOENT;
1279         rhltable_remove(&mrt->mfc_hash, &c->mnode, ipmr_rht_params);
1280         list_del_rcu(&c->list);
1281         call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_DEL, c, mrt->id);
1282         mroute_netlink_event(mrt, c, RTM_DELROUTE);
1283         ipmr_cache_put(c);
1284
1285         return 0;
1286 }
1287
1288 static int ipmr_mfc_add(struct net *net, struct mr_table *mrt,
1289                         struct mfcctl *mfc, int mrtsock, int parent)
1290 {
1291         struct mfc_cache *uc, *c;
1292         bool found;
1293         int ret;
1294
1295         if (mfc->mfcc_parent >= MAXVIFS)
1296                 return -ENFILE;
1297
1298         /* The entries are added/deleted only under RTNL */
1299         rcu_read_lock();
1300         c = ipmr_cache_find_parent(mrt, mfc->mfcc_origin.s_addr,
1301                                    mfc->mfcc_mcastgrp.s_addr, parent);
1302         rcu_read_unlock();
1303         if (c) {
1304                 write_lock_bh(&mrt_lock);
1305                 c->mfc_parent = mfc->mfcc_parent;
1306                 ipmr_update_thresholds(mrt, c, mfc->mfcc_ttls);
1307                 if (!mrtsock)
1308                         c->mfc_flags |= MFC_STATIC;
1309                 write_unlock_bh(&mrt_lock);
1310                 call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_REPLACE, c,
1311                                               mrt->id);
1312                 mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1313                 return 0;
1314         }
1315
1316         if (mfc->mfcc_mcastgrp.s_addr != htonl(INADDR_ANY) &&
1317             !ipv4_is_multicast(mfc->mfcc_mcastgrp.s_addr))
1318                 return -EINVAL;
1319
1320         c = ipmr_cache_alloc();
1321         if (!c)
1322                 return -ENOMEM;
1323
1324         c->mfc_origin = mfc->mfcc_origin.s_addr;
1325         c->mfc_mcastgrp = mfc->mfcc_mcastgrp.s_addr;
1326         c->mfc_parent = mfc->mfcc_parent;
1327         ipmr_update_thresholds(mrt, c, mfc->mfcc_ttls);
1328         if (!mrtsock)
1329                 c->mfc_flags |= MFC_STATIC;
1330
1331         ret = rhltable_insert_key(&mrt->mfc_hash, &c->cmparg, &c->mnode,
1332                                   ipmr_rht_params);
1333         if (ret) {
1334                 pr_err("ipmr: rhtable insert error %d\n", ret);
1335                 ipmr_cache_free(c);
1336                 return ret;
1337         }
1338         list_add_tail_rcu(&c->list, &mrt->mfc_cache_list);
1339         /* Check to see if we resolved a queued list. If so we
1340          * need to send on the frames and tidy up.
1341          */
1342         found = false;
1343         spin_lock_bh(&mfc_unres_lock);
1344         list_for_each_entry(uc, &mrt->mfc_unres_queue, list) {
1345                 if (uc->mfc_origin == c->mfc_origin &&
1346                     uc->mfc_mcastgrp == c->mfc_mcastgrp) {
1347                         list_del(&uc->list);
1348                         atomic_dec(&mrt->cache_resolve_queue_len);
1349                         found = true;
1350                         break;
1351                 }
1352         }
1353         if (list_empty(&mrt->mfc_unres_queue))
1354                 del_timer(&mrt->ipmr_expire_timer);
1355         spin_unlock_bh(&mfc_unres_lock);
1356
1357         if (found) {
1358                 ipmr_cache_resolve(net, mrt, uc, c);
1359                 ipmr_cache_free(uc);
1360         }
1361         call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_ADD, c, mrt->id);
1362         mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1363         return 0;
1364 }
1365
1366 /* Close the multicast socket, and clear the vif tables etc */
1367 static void mroute_clean_tables(struct mr_table *mrt, bool all)
1368 {
1369         struct net *net = read_pnet(&mrt->net);
1370         struct mfc_cache *c, *tmp;
1371         LIST_HEAD(list);
1372         int i;
1373
1374         /* Shut down all active vif entries */
1375         for (i = 0; i < mrt->maxvif; i++) {
1376                 if (!all && (mrt->vif_table[i].flags & VIFF_STATIC))
1377                         continue;
1378                 vif_delete(mrt, i, 0, &list);
1379         }
1380         unregister_netdevice_many(&list);
1381
1382         /* Wipe the cache */
1383         list_for_each_entry_safe(c, tmp, &mrt->mfc_cache_list, list) {
1384                 if (!all && (c->mfc_flags & MFC_STATIC))
1385                         continue;
1386                 rhltable_remove(&mrt->mfc_hash, &c->mnode, ipmr_rht_params);
1387                 list_del_rcu(&c->list);
1388                 call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_DEL, c,
1389                                               mrt->id);
1390                 mroute_netlink_event(mrt, c, RTM_DELROUTE);
1391                 ipmr_cache_put(c);
1392         }
1393
1394         if (atomic_read(&mrt->cache_resolve_queue_len) != 0) {
1395                 spin_lock_bh(&mfc_unres_lock);
1396                 list_for_each_entry_safe(c, tmp, &mrt->mfc_unres_queue, list) {
1397                         list_del(&c->list);
1398                         mroute_netlink_event(mrt, c, RTM_DELROUTE);
1399                         ipmr_destroy_unres(mrt, c);
1400                 }
1401                 spin_unlock_bh(&mfc_unres_lock);
1402         }
1403 }
1404
1405 /* called from ip_ra_control(), before an RCU grace period,
1406  * we dont need to call synchronize_rcu() here
1407  */
1408 static void mrtsock_destruct(struct sock *sk)
1409 {
1410         struct net *net = sock_net(sk);
1411         struct mr_table *mrt;
1412
1413         ASSERT_RTNL();
1414         ipmr_for_each_table(mrt, net) {
1415                 if (sk == rtnl_dereference(mrt->mroute_sk)) {
1416                         IPV4_DEVCONF_ALL(net, MC_FORWARDING)--;
1417                         inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
1418                                                     NETCONFA_MC_FORWARDING,
1419                                                     NETCONFA_IFINDEX_ALL,
1420                                                     net->ipv4.devconf_all);
1421                         RCU_INIT_POINTER(mrt->mroute_sk, NULL);
1422                         mroute_clean_tables(mrt, false);
1423                 }
1424         }
1425 }
1426
1427 /* Socket options and virtual interface manipulation. The whole
1428  * virtual interface system is a complete heap, but unfortunately
1429  * that's how BSD mrouted happens to think. Maybe one day with a proper
1430  * MOSPF/PIM router set up we can clean this up.
1431  */
1432
1433 int ip_mroute_setsockopt(struct sock *sk, int optname, char __user *optval,
1434                          unsigned int optlen)
1435 {
1436         struct net *net = sock_net(sk);
1437         int val, ret = 0, parent = 0;
1438         struct mr_table *mrt;
1439         struct vifctl vif;
1440         struct mfcctl mfc;
1441         u32 uval;
1442
1443         /* There's one exception to the lock - MRT_DONE which needs to unlock */
1444         rtnl_lock();
1445         if (sk->sk_type != SOCK_RAW ||
1446             inet_sk(sk)->inet_num != IPPROTO_IGMP) {
1447                 ret = -EOPNOTSUPP;
1448                 goto out_unlock;
1449         }
1450
1451         mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1452         if (!mrt) {
1453                 ret = -ENOENT;
1454                 goto out_unlock;
1455         }
1456         if (optname != MRT_INIT) {
1457                 if (sk != rcu_access_pointer(mrt->mroute_sk) &&
1458                     !ns_capable(net->user_ns, CAP_NET_ADMIN)) {
1459                         ret = -EACCES;
1460                         goto out_unlock;
1461                 }
1462         }
1463
1464         switch (optname) {
1465         case MRT_INIT:
1466                 if (optlen != sizeof(int)) {
1467                         ret = -EINVAL;
1468                         break;
1469                 }
1470                 if (rtnl_dereference(mrt->mroute_sk)) {
1471                         ret = -EADDRINUSE;
1472                         break;
1473                 }
1474
1475                 ret = ip_ra_control(sk, 1, mrtsock_destruct);
1476                 if (ret == 0) {
1477                         rcu_assign_pointer(mrt->mroute_sk, sk);
1478                         IPV4_DEVCONF_ALL(net, MC_FORWARDING)++;
1479                         inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
1480                                                     NETCONFA_MC_FORWARDING,
1481                                                     NETCONFA_IFINDEX_ALL,
1482                                                     net->ipv4.devconf_all);
1483                 }
1484                 break;
1485         case MRT_DONE:
1486                 if (sk != rcu_access_pointer(mrt->mroute_sk)) {
1487                         ret = -EACCES;
1488                 } else {
1489                         ret = ip_ra_control(sk, 0, NULL);
1490                         goto out_unlock;
1491                 }
1492                 break;
1493         case MRT_ADD_VIF:
1494         case MRT_DEL_VIF:
1495                 if (optlen != sizeof(vif)) {
1496                         ret = -EINVAL;
1497                         break;
1498                 }
1499                 if (copy_from_user(&vif, optval, sizeof(vif))) {
1500                         ret = -EFAULT;
1501                         break;
1502                 }
1503                 if (vif.vifc_vifi >= MAXVIFS) {
1504                         ret = -ENFILE;
1505                         break;
1506                 }
1507                 if (optname == MRT_ADD_VIF) {
1508                         ret = vif_add(net, mrt, &vif,
1509                                       sk == rtnl_dereference(mrt->mroute_sk));
1510                 } else {
1511                         ret = vif_delete(mrt, vif.vifc_vifi, 0, NULL);
1512                 }
1513                 break;
1514         /* Manipulate the forwarding caches. These live
1515          * in a sort of kernel/user symbiosis.
1516          */
1517         case MRT_ADD_MFC:
1518         case MRT_DEL_MFC:
1519                 parent = -1;
1520         case MRT_ADD_MFC_PROXY:
1521         case MRT_DEL_MFC_PROXY:
1522                 if (optlen != sizeof(mfc)) {
1523                         ret = -EINVAL;
1524                         break;
1525                 }
1526                 if (copy_from_user(&mfc, optval, sizeof(mfc))) {
1527                         ret = -EFAULT;
1528                         break;
1529                 }
1530                 if (parent == 0)
1531                         parent = mfc.mfcc_parent;
1532                 if (optname == MRT_DEL_MFC || optname == MRT_DEL_MFC_PROXY)
1533                         ret = ipmr_mfc_delete(mrt, &mfc, parent);
1534                 else
1535                         ret = ipmr_mfc_add(net, mrt, &mfc,
1536                                            sk == rtnl_dereference(mrt->mroute_sk),
1537                                            parent);
1538                 break;
1539         /* Control PIM assert. */
1540         case MRT_ASSERT:
1541                 if (optlen != sizeof(val)) {
1542                         ret = -EINVAL;
1543                         break;
1544                 }
1545                 if (get_user(val, (int __user *)optval)) {
1546                         ret = -EFAULT;
1547                         break;
1548                 }
1549                 mrt->mroute_do_assert = val;
1550                 break;
1551         case MRT_PIM:
1552                 if (!ipmr_pimsm_enabled()) {
1553                         ret = -ENOPROTOOPT;
1554                         break;
1555                 }
1556                 if (optlen != sizeof(val)) {
1557                         ret = -EINVAL;
1558                         break;
1559                 }
1560                 if (get_user(val, (int __user *)optval)) {
1561                         ret = -EFAULT;
1562                         break;
1563                 }
1564
1565                 val = !!val;
1566                 if (val != mrt->mroute_do_pim) {
1567                         mrt->mroute_do_pim = val;
1568                         mrt->mroute_do_assert = val;
1569                 }
1570                 break;
1571         case MRT_TABLE:
1572                 if (!IS_BUILTIN(CONFIG_IP_MROUTE_MULTIPLE_TABLES)) {
1573                         ret = -ENOPROTOOPT;
1574                         break;
1575                 }
1576                 if (optlen != sizeof(uval)) {
1577                         ret = -EINVAL;
1578                         break;
1579                 }
1580                 if (get_user(uval, (u32 __user *)optval)) {
1581                         ret = -EFAULT;
1582                         break;
1583                 }
1584
1585                 if (sk == rtnl_dereference(mrt->mroute_sk)) {
1586                         ret = -EBUSY;
1587                 } else {
1588                         mrt = ipmr_new_table(net, uval);
1589                         if (IS_ERR(mrt))
1590                                 ret = PTR_ERR(mrt);
1591                         else
1592                                 raw_sk(sk)->ipmr_table = uval;
1593                 }
1594                 break;
1595         /* Spurious command, or MRT_VERSION which you cannot set. */
1596         default:
1597                 ret = -ENOPROTOOPT;
1598         }
1599 out_unlock:
1600         rtnl_unlock();
1601         return ret;
1602 }
1603
1604 /* Getsock opt support for the multicast routing system. */
1605 int ip_mroute_getsockopt(struct sock *sk, int optname, char __user *optval, int __user *optlen)
1606 {
1607         int olr;
1608         int val;
1609         struct net *net = sock_net(sk);
1610         struct mr_table *mrt;
1611
1612         if (sk->sk_type != SOCK_RAW ||
1613             inet_sk(sk)->inet_num != IPPROTO_IGMP)
1614                 return -EOPNOTSUPP;
1615
1616         mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1617         if (!mrt)
1618                 return -ENOENT;
1619
1620         switch (optname) {
1621         case MRT_VERSION:
1622                 val = 0x0305;
1623                 break;
1624         case MRT_PIM:
1625                 if (!ipmr_pimsm_enabled())
1626                         return -ENOPROTOOPT;
1627                 val = mrt->mroute_do_pim;
1628                 break;
1629         case MRT_ASSERT:
1630                 val = mrt->mroute_do_assert;
1631                 break;
1632         default:
1633                 return -ENOPROTOOPT;
1634         }
1635
1636         if (get_user(olr, optlen))
1637                 return -EFAULT;
1638         olr = min_t(unsigned int, olr, sizeof(int));
1639         if (olr < 0)
1640                 return -EINVAL;
1641         if (put_user(olr, optlen))
1642                 return -EFAULT;
1643         if (copy_to_user(optval, &val, olr))
1644                 return -EFAULT;
1645         return 0;
1646 }
1647
1648 /* The IP multicast ioctl support routines. */
1649 int ipmr_ioctl(struct sock *sk, int cmd, void __user *arg)
1650 {
1651         struct sioc_sg_req sr;
1652         struct sioc_vif_req vr;
1653         struct vif_device *vif;
1654         struct mfc_cache *c;
1655         struct net *net = sock_net(sk);
1656         struct mr_table *mrt;
1657
1658         mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1659         if (!mrt)
1660                 return -ENOENT;
1661
1662         switch (cmd) {
1663         case SIOCGETVIFCNT:
1664                 if (copy_from_user(&vr, arg, sizeof(vr)))
1665                         return -EFAULT;
1666                 if (vr.vifi >= mrt->maxvif)
1667                         return -EINVAL;
1668                 read_lock(&mrt_lock);
1669                 vif = &mrt->vif_table[vr.vifi];
1670                 if (VIF_EXISTS(mrt, vr.vifi)) {
1671                         vr.icount = vif->pkt_in;
1672                         vr.ocount = vif->pkt_out;
1673                         vr.ibytes = vif->bytes_in;
1674                         vr.obytes = vif->bytes_out;
1675                         read_unlock(&mrt_lock);
1676
1677                         if (copy_to_user(arg, &vr, sizeof(vr)))
1678                                 return -EFAULT;
1679                         return 0;
1680                 }
1681                 read_unlock(&mrt_lock);
1682                 return -EADDRNOTAVAIL;
1683         case SIOCGETSGCNT:
1684                 if (copy_from_user(&sr, arg, sizeof(sr)))
1685                         return -EFAULT;
1686
1687                 rcu_read_lock();
1688                 c = ipmr_cache_find(mrt, sr.src.s_addr, sr.grp.s_addr);
1689                 if (c) {
1690                         sr.pktcnt = c->mfc_un.res.pkt;
1691                         sr.bytecnt = c->mfc_un.res.bytes;
1692                         sr.wrong_if = c->mfc_un.res.wrong_if;
1693                         rcu_read_unlock();
1694
1695                         if (copy_to_user(arg, &sr, sizeof(sr)))
1696                                 return -EFAULT;
1697                         return 0;
1698                 }
1699                 rcu_read_unlock();
1700                 return -EADDRNOTAVAIL;
1701         default:
1702                 return -ENOIOCTLCMD;
1703         }
1704 }
1705
1706 #ifdef CONFIG_COMPAT
1707 struct compat_sioc_sg_req {
1708         struct in_addr src;
1709         struct in_addr grp;
1710         compat_ulong_t pktcnt;
1711         compat_ulong_t bytecnt;
1712         compat_ulong_t wrong_if;
1713 };
1714
1715 struct compat_sioc_vif_req {
1716         vifi_t  vifi;           /* Which iface */
1717         compat_ulong_t icount;
1718         compat_ulong_t ocount;
1719         compat_ulong_t ibytes;
1720         compat_ulong_t obytes;
1721 };
1722
1723 int ipmr_compat_ioctl(struct sock *sk, unsigned int cmd, void __user *arg)
1724 {
1725         struct compat_sioc_sg_req sr;
1726         struct compat_sioc_vif_req vr;
1727         struct vif_device *vif;
1728         struct mfc_cache *c;
1729         struct net *net = sock_net(sk);
1730         struct mr_table *mrt;
1731
1732         mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1733         if (!mrt)
1734                 return -ENOENT;
1735
1736         switch (cmd) {
1737         case SIOCGETVIFCNT:
1738                 if (copy_from_user(&vr, arg, sizeof(vr)))
1739                         return -EFAULT;
1740                 if (vr.vifi >= mrt->maxvif)
1741                         return -EINVAL;
1742                 read_lock(&mrt_lock);
1743                 vif = &mrt->vif_table[vr.vifi];
1744                 if (VIF_EXISTS(mrt, vr.vifi)) {
1745                         vr.icount = vif->pkt_in;
1746                         vr.ocount = vif->pkt_out;
1747                         vr.ibytes = vif->bytes_in;
1748                         vr.obytes = vif->bytes_out;
1749                         read_unlock(&mrt_lock);
1750
1751                         if (copy_to_user(arg, &vr, sizeof(vr)))
1752                                 return -EFAULT;
1753                         return 0;
1754                 }
1755                 read_unlock(&mrt_lock);
1756                 return -EADDRNOTAVAIL;
1757         case SIOCGETSGCNT:
1758                 if (copy_from_user(&sr, arg, sizeof(sr)))
1759                         return -EFAULT;
1760
1761                 rcu_read_lock();
1762                 c = ipmr_cache_find(mrt, sr.src.s_addr, sr.grp.s_addr);
1763                 if (c) {
1764                         sr.pktcnt = c->mfc_un.res.pkt;
1765                         sr.bytecnt = c->mfc_un.res.bytes;
1766                         sr.wrong_if = c->mfc_un.res.wrong_if;
1767                         rcu_read_unlock();
1768
1769                         if (copy_to_user(arg, &sr, sizeof(sr)))
1770                                 return -EFAULT;
1771                         return 0;
1772                 }
1773                 rcu_read_unlock();
1774                 return -EADDRNOTAVAIL;
1775         default:
1776                 return -ENOIOCTLCMD;
1777         }
1778 }
1779 #endif
1780
1781 static int ipmr_device_event(struct notifier_block *this, unsigned long event, void *ptr)
1782 {
1783         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1784         struct net *net = dev_net(dev);
1785         struct mr_table *mrt;
1786         struct vif_device *v;
1787         int ct;
1788
1789         if (event != NETDEV_UNREGISTER)
1790                 return NOTIFY_DONE;
1791
1792         ipmr_for_each_table(mrt, net) {
1793                 v = &mrt->vif_table[0];
1794                 for (ct = 0; ct < mrt->maxvif; ct++, v++) {
1795                         if (v->dev == dev)
1796                                 vif_delete(mrt, ct, 1, NULL);
1797                 }
1798         }
1799         return NOTIFY_DONE;
1800 }
1801
1802 static struct notifier_block ip_mr_notifier = {
1803         .notifier_call = ipmr_device_event,
1804 };
1805
1806 /* Encapsulate a packet by attaching a valid IPIP header to it.
1807  * This avoids tunnel drivers and other mess and gives us the speed so
1808  * important for multicast video.
1809  */
1810 static void ip_encap(struct net *net, struct sk_buff *skb,
1811                      __be32 saddr, __be32 daddr)
1812 {
1813         struct iphdr *iph;
1814         const struct iphdr *old_iph = ip_hdr(skb);
1815
1816         skb_push(skb, sizeof(struct iphdr));
1817         skb->transport_header = skb->network_header;
1818         skb_reset_network_header(skb);
1819         iph = ip_hdr(skb);
1820
1821         iph->version    =       4;
1822         iph->tos        =       old_iph->tos;
1823         iph->ttl        =       old_iph->ttl;
1824         iph->frag_off   =       0;
1825         iph->daddr      =       daddr;
1826         iph->saddr      =       saddr;
1827         iph->protocol   =       IPPROTO_IPIP;
1828         iph->ihl        =       5;
1829         iph->tot_len    =       htons(skb->len);
1830         ip_select_ident(net, skb, NULL);
1831         ip_send_check(iph);
1832
1833         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1834         nf_reset(skb);
1835 }
1836
1837 static inline int ipmr_forward_finish(struct net *net, struct sock *sk,
1838                                       struct sk_buff *skb)
1839 {
1840         struct ip_options *opt = &(IPCB(skb)->opt);
1841
1842         IP_INC_STATS(net, IPSTATS_MIB_OUTFORWDATAGRAMS);
1843         IP_ADD_STATS(net, IPSTATS_MIB_OUTOCTETS, skb->len);
1844
1845         if (unlikely(opt->optlen))
1846                 ip_forward_options(skb);
1847
1848         return dst_output(net, sk, skb);
1849 }
1850
1851 /* Processing handlers for ipmr_forward */
1852
1853 static void ipmr_queue_xmit(struct net *net, struct mr_table *mrt,
1854                             struct sk_buff *skb, struct mfc_cache *c, int vifi)
1855 {
1856         const struct iphdr *iph = ip_hdr(skb);
1857         struct vif_device *vif = &mrt->vif_table[vifi];
1858         struct net_device *dev;
1859         struct rtable *rt;
1860         struct flowi4 fl4;
1861         int    encap = 0;
1862
1863         if (!vif->dev)
1864                 goto out_free;
1865
1866         if (vif->flags & VIFF_REGISTER) {
1867                 vif->pkt_out++;
1868                 vif->bytes_out += skb->len;
1869                 vif->dev->stats.tx_bytes += skb->len;
1870                 vif->dev->stats.tx_packets++;
1871                 ipmr_cache_report(mrt, skb, vifi, IGMPMSG_WHOLEPKT);
1872                 goto out_free;
1873         }
1874
1875         if (vif->flags & VIFF_TUNNEL) {
1876                 rt = ip_route_output_ports(net, &fl4, NULL,
1877                                            vif->remote, vif->local,
1878                                            0, 0,
1879                                            IPPROTO_IPIP,
1880                                            RT_TOS(iph->tos), vif->link);
1881                 if (IS_ERR(rt))
1882                         goto out_free;
1883                 encap = sizeof(struct iphdr);
1884         } else {
1885                 rt = ip_route_output_ports(net, &fl4, NULL, iph->daddr, 0,
1886                                            0, 0,
1887                                            IPPROTO_IPIP,
1888                                            RT_TOS(iph->tos), vif->link);
1889                 if (IS_ERR(rt))
1890                         goto out_free;
1891         }
1892
1893         dev = rt->dst.dev;
1894
1895         if (skb->len+encap > dst_mtu(&rt->dst) && (ntohs(iph->frag_off) & IP_DF)) {
1896                 /* Do not fragment multicasts. Alas, IPv4 does not
1897                  * allow to send ICMP, so that packets will disappear
1898                  * to blackhole.
1899                  */
1900                 IP_INC_STATS(net, IPSTATS_MIB_FRAGFAILS);
1901                 ip_rt_put(rt);
1902                 goto out_free;
1903         }
1904
1905         encap += LL_RESERVED_SPACE(dev) + rt->dst.header_len;
1906
1907         if (skb_cow(skb, encap)) {
1908                 ip_rt_put(rt);
1909                 goto out_free;
1910         }
1911
1912         vif->pkt_out++;
1913         vif->bytes_out += skb->len;
1914
1915         skb_dst_drop(skb);
1916         skb_dst_set(skb, &rt->dst);
1917         ip_decrease_ttl(ip_hdr(skb));
1918
1919         /* FIXME: forward and output firewalls used to be called here.
1920          * What do we do with netfilter? -- RR
1921          */
1922         if (vif->flags & VIFF_TUNNEL) {
1923                 ip_encap(net, skb, vif->local, vif->remote);
1924                 /* FIXME: extra output firewall step used to be here. --RR */
1925                 vif->dev->stats.tx_packets++;
1926                 vif->dev->stats.tx_bytes += skb->len;
1927         }
1928
1929         IPCB(skb)->flags |= IPSKB_FORWARDED;
1930
1931         /* RFC1584 teaches, that DVMRP/PIM router must deliver packets locally
1932          * not only before forwarding, but after forwarding on all output
1933          * interfaces. It is clear, if mrouter runs a multicasting
1934          * program, it should receive packets not depending to what interface
1935          * program is joined.
1936          * If we will not make it, the program will have to join on all
1937          * interfaces. On the other hand, multihoming host (or router, but
1938          * not mrouter) cannot join to more than one interface - it will
1939          * result in receiving multiple packets.
1940          */
1941         NF_HOOK(NFPROTO_IPV4, NF_INET_FORWARD,
1942                 net, NULL, skb, skb->dev, dev,
1943                 ipmr_forward_finish);
1944         return;
1945
1946 out_free:
1947         kfree_skb(skb);
1948 }
1949
1950 static int ipmr_find_vif(struct mr_table *mrt, struct net_device *dev)
1951 {
1952         int ct;
1953
1954         for (ct = mrt->maxvif-1; ct >= 0; ct--) {
1955                 if (mrt->vif_table[ct].dev == dev)
1956                         break;
1957         }
1958         return ct;
1959 }
1960
1961 /* "local" means that we should preserve one skb (for local delivery) */
1962 static void ip_mr_forward(struct net *net, struct mr_table *mrt,
1963                           struct net_device *dev, struct sk_buff *skb,
1964                           struct mfc_cache *cache, int local)
1965 {
1966         int true_vifi = ipmr_find_vif(mrt, dev);
1967         int psend = -1;
1968         int vif, ct;
1969
1970         vif = cache->mfc_parent;
1971         cache->mfc_un.res.pkt++;
1972         cache->mfc_un.res.bytes += skb->len;
1973         cache->mfc_un.res.lastuse = jiffies;
1974
1975         if (cache->mfc_origin == htonl(INADDR_ANY) && true_vifi >= 0) {
1976                 struct mfc_cache *cache_proxy;
1977
1978                 /* For an (*,G) entry, we only check that the incomming
1979                  * interface is part of the static tree.
1980                  */
1981                 cache_proxy = ipmr_cache_find_any_parent(mrt, vif);
1982                 if (cache_proxy &&
1983                     cache_proxy->mfc_un.res.ttls[true_vifi] < 255)
1984                         goto forward;
1985         }
1986
1987         /* Wrong interface: drop packet and (maybe) send PIM assert. */
1988         if (mrt->vif_table[vif].dev != dev) {
1989                 if (rt_is_output_route(skb_rtable(skb))) {
1990                         /* It is our own packet, looped back.
1991                          * Very complicated situation...
1992                          *
1993                          * The best workaround until routing daemons will be
1994                          * fixed is not to redistribute packet, if it was
1995                          * send through wrong interface. It means, that
1996                          * multicast applications WILL NOT work for
1997                          * (S,G), which have default multicast route pointing
1998                          * to wrong oif. In any case, it is not a good
1999                          * idea to use multicasting applications on router.
2000                          */
2001                         goto dont_forward;
2002                 }
2003
2004                 cache->mfc_un.res.wrong_if++;
2005
2006                 if (true_vifi >= 0 && mrt->mroute_do_assert &&
2007                     /* pimsm uses asserts, when switching from RPT to SPT,
2008                      * so that we cannot check that packet arrived on an oif.
2009                      * It is bad, but otherwise we would need to move pretty
2010                      * large chunk of pimd to kernel. Ough... --ANK
2011                      */
2012                     (mrt->mroute_do_pim ||
2013                      cache->mfc_un.res.ttls[true_vifi] < 255) &&
2014                     time_after(jiffies,
2015                                cache->mfc_un.res.last_assert + MFC_ASSERT_THRESH)) {
2016                         cache->mfc_un.res.last_assert = jiffies;
2017                         ipmr_cache_report(mrt, skb, true_vifi, IGMPMSG_WRONGVIF);
2018                 }
2019                 goto dont_forward;
2020         }
2021
2022 forward:
2023         mrt->vif_table[vif].pkt_in++;
2024         mrt->vif_table[vif].bytes_in += skb->len;
2025
2026         /* Forward the frame */
2027         if (cache->mfc_origin == htonl(INADDR_ANY) &&
2028             cache->mfc_mcastgrp == htonl(INADDR_ANY)) {
2029                 if (true_vifi >= 0 &&
2030                     true_vifi != cache->mfc_parent &&
2031                     ip_hdr(skb)->ttl >
2032                                 cache->mfc_un.res.ttls[cache->mfc_parent]) {
2033                         /* It's an (*,*) entry and the packet is not coming from
2034                          * the upstream: forward the packet to the upstream
2035                          * only.
2036                          */
2037                         psend = cache->mfc_parent;
2038                         goto last_forward;
2039                 }
2040                 goto dont_forward;
2041         }
2042         for (ct = cache->mfc_un.res.maxvif - 1;
2043              ct >= cache->mfc_un.res.minvif; ct--) {
2044                 /* For (*,G) entry, don't forward to the incoming interface */
2045                 if ((cache->mfc_origin != htonl(INADDR_ANY) ||
2046                      ct != true_vifi) &&
2047                     ip_hdr(skb)->ttl > cache->mfc_un.res.ttls[ct]) {
2048                         if (psend != -1) {
2049                                 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2050
2051                                 if (skb2)
2052                                         ipmr_queue_xmit(net, mrt, skb2, cache,
2053                                                         psend);
2054                         }
2055                         psend = ct;
2056                 }
2057         }
2058 last_forward:
2059         if (psend != -1) {
2060                 if (local) {
2061                         struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2062
2063                         if (skb2)
2064                                 ipmr_queue_xmit(net, mrt, skb2, cache, psend);
2065                 } else {
2066                         ipmr_queue_xmit(net, mrt, skb, cache, psend);
2067                         return;
2068                 }
2069         }
2070
2071 dont_forward:
2072         if (!local)
2073                 kfree_skb(skb);
2074 }
2075
2076 static struct mr_table *ipmr_rt_fib_lookup(struct net *net, struct sk_buff *skb)
2077 {
2078         struct rtable *rt = skb_rtable(skb);
2079         struct iphdr *iph = ip_hdr(skb);
2080         struct flowi4 fl4 = {
2081                 .daddr = iph->daddr,
2082                 .saddr = iph->saddr,
2083                 .flowi4_tos = RT_TOS(iph->tos),
2084                 .flowi4_oif = (rt_is_output_route(rt) ?
2085                                skb->dev->ifindex : 0),
2086                 .flowi4_iif = (rt_is_output_route(rt) ?
2087                                LOOPBACK_IFINDEX :
2088                                skb->dev->ifindex),
2089                 .flowi4_mark = skb->mark,
2090         };
2091         struct mr_table *mrt;
2092         int err;
2093
2094         err = ipmr_fib_lookup(net, &fl4, &mrt);
2095         if (err)
2096                 return ERR_PTR(err);
2097         return mrt;
2098 }
2099
2100 /* Multicast packets for forwarding arrive here
2101  * Called with rcu_read_lock();
2102  */
2103 int ip_mr_input(struct sk_buff *skb)
2104 {
2105         struct mfc_cache *cache;
2106         struct net *net = dev_net(skb->dev);
2107         int local = skb_rtable(skb)->rt_flags & RTCF_LOCAL;
2108         struct mr_table *mrt;
2109         struct net_device *dev;
2110
2111         /* skb->dev passed in is the loX master dev for vrfs.
2112          * As there are no vifs associated with loopback devices,
2113          * get the proper interface that does have a vif associated with it.
2114          */
2115         dev = skb->dev;
2116         if (netif_is_l3_master(skb->dev)) {
2117                 dev = dev_get_by_index_rcu(net, IPCB(skb)->iif);
2118                 if (!dev) {
2119                         kfree_skb(skb);
2120                         return -ENODEV;
2121                 }
2122         }
2123
2124         /* Packet is looped back after forward, it should not be
2125          * forwarded second time, but still can be delivered locally.
2126          */
2127         if (IPCB(skb)->flags & IPSKB_FORWARDED)
2128                 goto dont_forward;
2129
2130         mrt = ipmr_rt_fib_lookup(net, skb);
2131         if (IS_ERR(mrt)) {
2132                 kfree_skb(skb);
2133                 return PTR_ERR(mrt);
2134         }
2135         if (!local) {
2136                 if (IPCB(skb)->opt.router_alert) {
2137                         if (ip_call_ra_chain(skb))
2138                                 return 0;
2139                 } else if (ip_hdr(skb)->protocol == IPPROTO_IGMP) {
2140                         /* IGMPv1 (and broken IGMPv2 implementations sort of
2141                          * Cisco IOS <= 11.2(8)) do not put router alert
2142                          * option to IGMP packets destined to routable
2143                          * groups. It is very bad, because it means
2144                          * that we can forward NO IGMP messages.
2145                          */
2146                         struct sock *mroute_sk;
2147
2148                         mroute_sk = rcu_dereference(mrt->mroute_sk);
2149                         if (mroute_sk) {
2150                                 nf_reset(skb);
2151                                 raw_rcv(mroute_sk, skb);
2152                                 return 0;
2153                         }
2154                     }
2155         }
2156
2157         /* already under rcu_read_lock() */
2158         cache = ipmr_cache_find(mrt, ip_hdr(skb)->saddr, ip_hdr(skb)->daddr);
2159         if (!cache) {
2160                 int vif = ipmr_find_vif(mrt, dev);
2161
2162                 if (vif >= 0)
2163                         cache = ipmr_cache_find_any(mrt, ip_hdr(skb)->daddr,
2164                                                     vif);
2165         }
2166
2167         /* No usable cache entry */
2168         if (!cache) {
2169                 int vif;
2170
2171                 if (local) {
2172                         struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2173                         ip_local_deliver(skb);
2174                         if (!skb2)
2175                                 return -ENOBUFS;
2176                         skb = skb2;
2177                 }
2178
2179                 read_lock(&mrt_lock);
2180                 vif = ipmr_find_vif(mrt, dev);
2181                 if (vif >= 0) {
2182                         int err2 = ipmr_cache_unresolved(mrt, vif, skb, dev);
2183                         read_unlock(&mrt_lock);
2184
2185                         return err2;
2186                 }
2187                 read_unlock(&mrt_lock);
2188                 kfree_skb(skb);
2189                 return -ENODEV;
2190         }
2191
2192         read_lock(&mrt_lock);
2193         ip_mr_forward(net, mrt, dev, skb, cache, local);
2194         read_unlock(&mrt_lock);
2195
2196         if (local)
2197                 return ip_local_deliver(skb);
2198
2199         return 0;
2200
2201 dont_forward:
2202         if (local)
2203                 return ip_local_deliver(skb);
2204         kfree_skb(skb);
2205         return 0;
2206 }
2207
2208 #ifdef CONFIG_IP_PIMSM_V1
2209 /* Handle IGMP messages of PIMv1 */
2210 int pim_rcv_v1(struct sk_buff *skb)
2211 {
2212         struct igmphdr *pim;
2213         struct net *net = dev_net(skb->dev);
2214         struct mr_table *mrt;
2215
2216         if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr)))
2217                 goto drop;
2218
2219         pim = igmp_hdr(skb);
2220
2221         mrt = ipmr_rt_fib_lookup(net, skb);
2222         if (IS_ERR(mrt))
2223                 goto drop;
2224         if (!mrt->mroute_do_pim ||
2225             pim->group != PIM_V1_VERSION || pim->code != PIM_V1_REGISTER)
2226                 goto drop;
2227
2228         if (__pim_rcv(mrt, skb, sizeof(*pim))) {
2229 drop:
2230                 kfree_skb(skb);
2231         }
2232         return 0;
2233 }
2234 #endif
2235
2236 #ifdef CONFIG_IP_PIMSM_V2
2237 static int pim_rcv(struct sk_buff *skb)
2238 {
2239         struct pimreghdr *pim;
2240         struct net *net = dev_net(skb->dev);
2241         struct mr_table *mrt;
2242
2243         if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr)))
2244                 goto drop;
2245
2246         pim = (struct pimreghdr *)skb_transport_header(skb);
2247         if (pim->type != ((PIM_VERSION << 4) | (PIM_TYPE_REGISTER)) ||
2248             (pim->flags & PIM_NULL_REGISTER) ||
2249             (ip_compute_csum((void *)pim, sizeof(*pim)) != 0 &&
2250              csum_fold(skb_checksum(skb, 0, skb->len, 0))))
2251                 goto drop;
2252
2253         mrt = ipmr_rt_fib_lookup(net, skb);
2254         if (IS_ERR(mrt))
2255                 goto drop;
2256         if (__pim_rcv(mrt, skb, sizeof(*pim))) {
2257 drop:
2258                 kfree_skb(skb);
2259         }
2260         return 0;
2261 }
2262 #endif
2263
2264 static int __ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
2265                               struct mfc_cache *c, struct rtmsg *rtm)
2266 {
2267         struct rta_mfc_stats mfcs;
2268         struct nlattr *mp_attr;
2269         struct rtnexthop *nhp;
2270         unsigned long lastuse;
2271         int ct;
2272
2273         /* If cache is unresolved, don't try to parse IIF and OIF */
2274         if (c->mfc_parent >= MAXVIFS) {
2275                 rtm->rtm_flags |= RTNH_F_UNRESOLVED;
2276                 return -ENOENT;
2277         }
2278
2279         if (VIF_EXISTS(mrt, c->mfc_parent) &&
2280             nla_put_u32(skb, RTA_IIF, mrt->vif_table[c->mfc_parent].dev->ifindex) < 0)
2281                 return -EMSGSIZE;
2282
2283         if (c->mfc_flags & MFC_OFFLOAD)
2284                 rtm->rtm_flags |= RTNH_F_OFFLOAD;
2285
2286         if (!(mp_attr = nla_nest_start(skb, RTA_MULTIPATH)))
2287                 return -EMSGSIZE;
2288
2289         for (ct = c->mfc_un.res.minvif; ct < c->mfc_un.res.maxvif; ct++) {
2290                 if (VIF_EXISTS(mrt, ct) && c->mfc_un.res.ttls[ct] < 255) {
2291                         if (!(nhp = nla_reserve_nohdr(skb, sizeof(*nhp)))) {
2292                                 nla_nest_cancel(skb, mp_attr);
2293                                 return -EMSGSIZE;
2294                         }
2295
2296                         nhp->rtnh_flags = 0;
2297                         nhp->rtnh_hops = c->mfc_un.res.ttls[ct];
2298                         nhp->rtnh_ifindex = mrt->vif_table[ct].dev->ifindex;
2299                         nhp->rtnh_len = sizeof(*nhp);
2300                 }
2301         }
2302
2303         nla_nest_end(skb, mp_attr);
2304
2305         lastuse = READ_ONCE(c->mfc_un.res.lastuse);
2306         lastuse = time_after_eq(jiffies, lastuse) ? jiffies - lastuse : 0;
2307
2308         mfcs.mfcs_packets = c->mfc_un.res.pkt;
2309         mfcs.mfcs_bytes = c->mfc_un.res.bytes;
2310         mfcs.mfcs_wrong_if = c->mfc_un.res.wrong_if;
2311         if (nla_put_64bit(skb, RTA_MFC_STATS, sizeof(mfcs), &mfcs, RTA_PAD) ||
2312             nla_put_u64_64bit(skb, RTA_EXPIRES, jiffies_to_clock_t(lastuse),
2313                               RTA_PAD))
2314                 return -EMSGSIZE;
2315
2316         rtm->rtm_type = RTN_MULTICAST;
2317         return 1;
2318 }
2319
2320 int ipmr_get_route(struct net *net, struct sk_buff *skb,
2321                    __be32 saddr, __be32 daddr,
2322                    struct rtmsg *rtm, u32 portid)
2323 {
2324         struct mfc_cache *cache;
2325         struct mr_table *mrt;
2326         int err;
2327
2328         mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2329         if (!mrt)
2330                 return -ENOENT;
2331
2332         rcu_read_lock();
2333         cache = ipmr_cache_find(mrt, saddr, daddr);
2334         if (!cache && skb->dev) {
2335                 int vif = ipmr_find_vif(mrt, skb->dev);
2336
2337                 if (vif >= 0)
2338                         cache = ipmr_cache_find_any(mrt, daddr, vif);
2339         }
2340         if (!cache) {
2341                 struct sk_buff *skb2;
2342                 struct iphdr *iph;
2343                 struct net_device *dev;
2344                 int vif = -1;
2345
2346                 dev = skb->dev;
2347                 read_lock(&mrt_lock);
2348                 if (dev)
2349                         vif = ipmr_find_vif(mrt, dev);
2350                 if (vif < 0) {
2351                         read_unlock(&mrt_lock);
2352                         rcu_read_unlock();
2353                         return -ENODEV;
2354                 }
2355                 skb2 = skb_clone(skb, GFP_ATOMIC);
2356                 if (!skb2) {
2357                         read_unlock(&mrt_lock);
2358                         rcu_read_unlock();
2359                         return -ENOMEM;
2360                 }
2361
2362                 NETLINK_CB(skb2).portid = portid;
2363                 skb_push(skb2, sizeof(struct iphdr));
2364                 skb_reset_network_header(skb2);
2365                 iph = ip_hdr(skb2);
2366                 iph->ihl = sizeof(struct iphdr) >> 2;
2367                 iph->saddr = saddr;
2368                 iph->daddr = daddr;
2369                 iph->version = 0;
2370                 err = ipmr_cache_unresolved(mrt, vif, skb2, dev);
2371                 read_unlock(&mrt_lock);
2372                 rcu_read_unlock();
2373                 return err;
2374         }
2375
2376         read_lock(&mrt_lock);
2377         err = __ipmr_fill_mroute(mrt, skb, cache, rtm);
2378         read_unlock(&mrt_lock);
2379         rcu_read_unlock();
2380         return err;
2381 }
2382
2383 static int ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
2384                             u32 portid, u32 seq, struct mfc_cache *c, int cmd,
2385                             int flags)
2386 {
2387         struct nlmsghdr *nlh;
2388         struct rtmsg *rtm;
2389         int err;
2390
2391         nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rtm), flags);
2392         if (!nlh)
2393                 return -EMSGSIZE;
2394
2395         rtm = nlmsg_data(nlh);
2396         rtm->rtm_family   = RTNL_FAMILY_IPMR;
2397         rtm->rtm_dst_len  = 32;
2398         rtm->rtm_src_len  = 32;
2399         rtm->rtm_tos      = 0;
2400         rtm->rtm_table    = mrt->id;
2401         if (nla_put_u32(skb, RTA_TABLE, mrt->id))
2402                 goto nla_put_failure;
2403         rtm->rtm_type     = RTN_MULTICAST;
2404         rtm->rtm_scope    = RT_SCOPE_UNIVERSE;
2405         if (c->mfc_flags & MFC_STATIC)
2406                 rtm->rtm_protocol = RTPROT_STATIC;
2407         else
2408                 rtm->rtm_protocol = RTPROT_MROUTED;
2409         rtm->rtm_flags    = 0;
2410
2411         if (nla_put_in_addr(skb, RTA_SRC, c->mfc_origin) ||
2412             nla_put_in_addr(skb, RTA_DST, c->mfc_mcastgrp))
2413                 goto nla_put_failure;
2414         err = __ipmr_fill_mroute(mrt, skb, c, rtm);
2415         /* do not break the dump if cache is unresolved */
2416         if (err < 0 && err != -ENOENT)
2417                 goto nla_put_failure;
2418
2419         nlmsg_end(skb, nlh);
2420         return 0;
2421
2422 nla_put_failure:
2423         nlmsg_cancel(skb, nlh);
2424         return -EMSGSIZE;
2425 }
2426
2427 static size_t mroute_msgsize(bool unresolved, int maxvif)
2428 {
2429         size_t len =
2430                 NLMSG_ALIGN(sizeof(struct rtmsg))
2431                 + nla_total_size(4)     /* RTA_TABLE */
2432                 + nla_total_size(4)     /* RTA_SRC */
2433                 + nla_total_size(4)     /* RTA_DST */
2434                 ;
2435
2436         if (!unresolved)
2437                 len = len
2438                       + nla_total_size(4)       /* RTA_IIF */
2439                       + nla_total_size(0)       /* RTA_MULTIPATH */
2440                       + maxvif * NLA_ALIGN(sizeof(struct rtnexthop))
2441                                                 /* RTA_MFC_STATS */
2442                       + nla_total_size_64bit(sizeof(struct rta_mfc_stats))
2443                 ;
2444
2445         return len;
2446 }
2447
2448 static void mroute_netlink_event(struct mr_table *mrt, struct mfc_cache *mfc,
2449                                  int cmd)
2450 {
2451         struct net *net = read_pnet(&mrt->net);
2452         struct sk_buff *skb;
2453         int err = -ENOBUFS;
2454
2455         skb = nlmsg_new(mroute_msgsize(mfc->mfc_parent >= MAXVIFS, mrt->maxvif),
2456                         GFP_ATOMIC);
2457         if (!skb)
2458                 goto errout;
2459
2460         err = ipmr_fill_mroute(mrt, skb, 0, 0, mfc, cmd, 0);
2461         if (err < 0)
2462                 goto errout;
2463
2464         rtnl_notify(skb, net, 0, RTNLGRP_IPV4_MROUTE, NULL, GFP_ATOMIC);
2465         return;
2466
2467 errout:
2468         kfree_skb(skb);
2469         if (err < 0)
2470                 rtnl_set_sk_err(net, RTNLGRP_IPV4_MROUTE, err);
2471 }
2472
2473 static size_t igmpmsg_netlink_msgsize(size_t payloadlen)
2474 {
2475         size_t len =
2476                 NLMSG_ALIGN(sizeof(struct rtgenmsg))
2477                 + nla_total_size(1)     /* IPMRA_CREPORT_MSGTYPE */
2478                 + nla_total_size(4)     /* IPMRA_CREPORT_VIF_ID */
2479                 + nla_total_size(4)     /* IPMRA_CREPORT_SRC_ADDR */
2480                 + nla_total_size(4)     /* IPMRA_CREPORT_DST_ADDR */
2481                                         /* IPMRA_CREPORT_PKT */
2482                 + nla_total_size(payloadlen)
2483                 ;
2484
2485         return len;
2486 }
2487
2488 static void igmpmsg_netlink_event(struct mr_table *mrt, struct sk_buff *pkt)
2489 {
2490         struct net *net = read_pnet(&mrt->net);
2491         struct nlmsghdr *nlh;
2492         struct rtgenmsg *rtgenm;
2493         struct igmpmsg *msg;
2494         struct sk_buff *skb;
2495         struct nlattr *nla;
2496         int payloadlen;
2497
2498         payloadlen = pkt->len - sizeof(struct igmpmsg);
2499         msg = (struct igmpmsg *)skb_network_header(pkt);
2500
2501         skb = nlmsg_new(igmpmsg_netlink_msgsize(payloadlen), GFP_ATOMIC);
2502         if (!skb)
2503                 goto errout;
2504
2505         nlh = nlmsg_put(skb, 0, 0, RTM_NEWCACHEREPORT,
2506                         sizeof(struct rtgenmsg), 0);
2507         if (!nlh)
2508                 goto errout;
2509         rtgenm = nlmsg_data(nlh);
2510         rtgenm->rtgen_family = RTNL_FAMILY_IPMR;
2511         if (nla_put_u8(skb, IPMRA_CREPORT_MSGTYPE, msg->im_msgtype) ||
2512             nla_put_u32(skb, IPMRA_CREPORT_VIF_ID, msg->im_vif) ||
2513             nla_put_in_addr(skb, IPMRA_CREPORT_SRC_ADDR,
2514                             msg->im_src.s_addr) ||
2515             nla_put_in_addr(skb, IPMRA_CREPORT_DST_ADDR,
2516                             msg->im_dst.s_addr))
2517                 goto nla_put_failure;
2518
2519         nla = nla_reserve(skb, IPMRA_CREPORT_PKT, payloadlen);
2520         if (!nla || skb_copy_bits(pkt, sizeof(struct igmpmsg),
2521                                   nla_data(nla), payloadlen))
2522                 goto nla_put_failure;
2523
2524         nlmsg_end(skb, nlh);
2525
2526         rtnl_notify(skb, net, 0, RTNLGRP_IPV4_MROUTE_R, NULL, GFP_ATOMIC);
2527         return;
2528
2529 nla_put_failure:
2530         nlmsg_cancel(skb, nlh);
2531 errout:
2532         kfree_skb(skb);
2533         rtnl_set_sk_err(net, RTNLGRP_IPV4_MROUTE_R, -ENOBUFS);
2534 }
2535
2536 static int ipmr_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh,
2537                              struct netlink_ext_ack *extack)
2538 {
2539         struct net *net = sock_net(in_skb->sk);
2540         struct nlattr *tb[RTA_MAX + 1];
2541         struct sk_buff *skb = NULL;
2542         struct mfc_cache *cache;
2543         struct mr_table *mrt;
2544         struct rtmsg *rtm;
2545         __be32 src, grp;
2546         u32 tableid;
2547         int err;
2548
2549         err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX,
2550                           rtm_ipv4_policy, extack);
2551         if (err < 0)
2552                 goto errout;
2553
2554         rtm = nlmsg_data(nlh);
2555
2556         src = tb[RTA_SRC] ? nla_get_in_addr(tb[RTA_SRC]) : 0;
2557         grp = tb[RTA_DST] ? nla_get_in_addr(tb[RTA_DST]) : 0;
2558         tableid = tb[RTA_TABLE] ? nla_get_u32(tb[RTA_TABLE]) : 0;
2559
2560         mrt = ipmr_get_table(net, tableid ? tableid : RT_TABLE_DEFAULT);
2561         if (!mrt) {
2562                 err = -ENOENT;
2563                 goto errout_free;
2564         }
2565
2566         /* entries are added/deleted only under RTNL */
2567         rcu_read_lock();
2568         cache = ipmr_cache_find(mrt, src, grp);
2569         rcu_read_unlock();
2570         if (!cache) {
2571                 err = -ENOENT;
2572                 goto errout_free;
2573         }
2574
2575         skb = nlmsg_new(mroute_msgsize(false, mrt->maxvif), GFP_KERNEL);
2576         if (!skb) {
2577                 err = -ENOBUFS;
2578                 goto errout_free;
2579         }
2580
2581         err = ipmr_fill_mroute(mrt, skb, NETLINK_CB(in_skb).portid,
2582                                nlh->nlmsg_seq, cache,
2583                                RTM_NEWROUTE, 0);
2584         if (err < 0)
2585                 goto errout_free;
2586
2587         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
2588
2589 errout:
2590         return err;
2591
2592 errout_free:
2593         kfree_skb(skb);
2594         goto errout;
2595 }
2596
2597 static int ipmr_rtm_dumproute(struct sk_buff *skb, struct netlink_callback *cb)
2598 {
2599         struct net *net = sock_net(skb->sk);
2600         struct mr_table *mrt;
2601         struct mfc_cache *mfc;
2602         unsigned int t = 0, s_t;
2603         unsigned int e = 0, s_e;
2604
2605         s_t = cb->args[0];
2606         s_e = cb->args[1];
2607
2608         rcu_read_lock();
2609         ipmr_for_each_table(mrt, net) {
2610                 if (t < s_t)
2611                         goto next_table;
2612                 list_for_each_entry_rcu(mfc, &mrt->mfc_cache_list, list) {
2613                         if (e < s_e)
2614                                 goto next_entry;
2615                         if (ipmr_fill_mroute(mrt, skb,
2616                                              NETLINK_CB(cb->skb).portid,
2617                                              cb->nlh->nlmsg_seq,
2618                                              mfc, RTM_NEWROUTE,
2619                                              NLM_F_MULTI) < 0)
2620                                 goto done;
2621 next_entry:
2622                         e++;
2623                 }
2624                 e = 0;
2625                 s_e = 0;
2626
2627                 spin_lock_bh(&mfc_unres_lock);
2628                 list_for_each_entry(mfc, &mrt->mfc_unres_queue, list) {
2629                         if (e < s_e)
2630                                 goto next_entry2;
2631                         if (ipmr_fill_mroute(mrt, skb,
2632                                              NETLINK_CB(cb->skb).portid,
2633                                              cb->nlh->nlmsg_seq,
2634                                              mfc, RTM_NEWROUTE,
2635                                              NLM_F_MULTI) < 0) {
2636                                 spin_unlock_bh(&mfc_unres_lock);
2637                                 goto done;
2638                         }
2639 next_entry2:
2640                         e++;
2641                 }
2642                 spin_unlock_bh(&mfc_unres_lock);
2643                 e = 0;
2644                 s_e = 0;
2645 next_table:
2646                 t++;
2647         }
2648 done:
2649         rcu_read_unlock();
2650
2651         cb->args[1] = e;
2652         cb->args[0] = t;
2653
2654         return skb->len;
2655 }
2656
2657 static const struct nla_policy rtm_ipmr_policy[RTA_MAX + 1] = {
2658         [RTA_SRC]       = { .type = NLA_U32 },
2659         [RTA_DST]       = { .type = NLA_U32 },
2660         [RTA_IIF]       = { .type = NLA_U32 },
2661         [RTA_TABLE]     = { .type = NLA_U32 },
2662         [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
2663 };
2664
2665 static bool ipmr_rtm_validate_proto(unsigned char rtm_protocol)
2666 {
2667         switch (rtm_protocol) {
2668         case RTPROT_STATIC:
2669         case RTPROT_MROUTED:
2670                 return true;
2671         }
2672         return false;
2673 }
2674
2675 static int ipmr_nla_get_ttls(const struct nlattr *nla, struct mfcctl *mfcc)
2676 {
2677         struct rtnexthop *rtnh = nla_data(nla);
2678         int remaining = nla_len(nla), vifi = 0;
2679
2680         while (rtnh_ok(rtnh, remaining)) {
2681                 mfcc->mfcc_ttls[vifi] = rtnh->rtnh_hops;
2682                 if (++vifi == MAXVIFS)
2683                         break;
2684                 rtnh = rtnh_next(rtnh, &remaining);
2685         }
2686
2687         return remaining > 0 ? -EINVAL : vifi;
2688 }
2689
2690 /* returns < 0 on error, 0 for ADD_MFC and 1 for ADD_MFC_PROXY */
2691 static int rtm_to_ipmr_mfcc(struct net *net, struct nlmsghdr *nlh,
2692                             struct mfcctl *mfcc, int *mrtsock,
2693                             struct mr_table **mrtret,
2694                             struct netlink_ext_ack *extack)
2695 {
2696         struct net_device *dev = NULL;
2697         u32 tblid = RT_TABLE_DEFAULT;
2698         struct mr_table *mrt;
2699         struct nlattr *attr;
2700         struct rtmsg *rtm;
2701         int ret, rem;
2702
2703         ret = nlmsg_validate(nlh, sizeof(*rtm), RTA_MAX, rtm_ipmr_policy,
2704                              extack);
2705         if (ret < 0)
2706                 goto out;
2707         rtm = nlmsg_data(nlh);
2708
2709         ret = -EINVAL;
2710         if (rtm->rtm_family != RTNL_FAMILY_IPMR || rtm->rtm_dst_len != 32 ||
2711             rtm->rtm_type != RTN_MULTICAST ||
2712             rtm->rtm_scope != RT_SCOPE_UNIVERSE ||
2713             !ipmr_rtm_validate_proto(rtm->rtm_protocol))
2714                 goto out;
2715
2716         memset(mfcc, 0, sizeof(*mfcc));
2717         mfcc->mfcc_parent = -1;
2718         ret = 0;
2719         nlmsg_for_each_attr(attr, nlh, sizeof(struct rtmsg), rem) {
2720                 switch (nla_type(attr)) {
2721                 case RTA_SRC:
2722                         mfcc->mfcc_origin.s_addr = nla_get_be32(attr);
2723                         break;
2724                 case RTA_DST:
2725                         mfcc->mfcc_mcastgrp.s_addr = nla_get_be32(attr);
2726                         break;
2727                 case RTA_IIF:
2728                         dev = __dev_get_by_index(net, nla_get_u32(attr));
2729                         if (!dev) {
2730                                 ret = -ENODEV;
2731                                 goto out;
2732                         }
2733                         break;
2734                 case RTA_MULTIPATH:
2735                         if (ipmr_nla_get_ttls(attr, mfcc) < 0) {
2736                                 ret = -EINVAL;
2737                                 goto out;
2738                         }
2739                         break;
2740                 case RTA_PREFSRC:
2741                         ret = 1;
2742                         break;
2743                 case RTA_TABLE:
2744                         tblid = nla_get_u32(attr);
2745                         break;
2746                 }
2747         }
2748         mrt = ipmr_get_table(net, tblid);
2749         if (!mrt) {
2750                 ret = -ENOENT;
2751                 goto out;
2752         }
2753         *mrtret = mrt;
2754         *mrtsock = rtm->rtm_protocol == RTPROT_MROUTED ? 1 : 0;
2755         if (dev)
2756                 mfcc->mfcc_parent = ipmr_find_vif(mrt, dev);
2757
2758 out:
2759         return ret;
2760 }
2761
2762 /* takes care of both newroute and delroute */
2763 static int ipmr_rtm_route(struct sk_buff *skb, struct nlmsghdr *nlh,
2764                           struct netlink_ext_ack *extack)
2765 {
2766         struct net *net = sock_net(skb->sk);
2767         int ret, mrtsock, parent;
2768         struct mr_table *tbl;
2769         struct mfcctl mfcc;
2770
2771         mrtsock = 0;
2772         tbl = NULL;
2773         ret = rtm_to_ipmr_mfcc(net, nlh, &mfcc, &mrtsock, &tbl, extack);
2774         if (ret < 0)
2775                 return ret;
2776
2777         parent = ret ? mfcc.mfcc_parent : -1;
2778         if (nlh->nlmsg_type == RTM_NEWROUTE)
2779                 return ipmr_mfc_add(net, tbl, &mfcc, mrtsock, parent);
2780         else
2781                 return ipmr_mfc_delete(tbl, &mfcc, parent);
2782 }
2783
2784 static bool ipmr_fill_table(struct mr_table *mrt, struct sk_buff *skb)
2785 {
2786         u32 queue_len = atomic_read(&mrt->cache_resolve_queue_len);
2787
2788         if (nla_put_u32(skb, IPMRA_TABLE_ID, mrt->id) ||
2789             nla_put_u32(skb, IPMRA_TABLE_CACHE_RES_QUEUE_LEN, queue_len) ||
2790             nla_put_s32(skb, IPMRA_TABLE_MROUTE_REG_VIF_NUM,
2791                         mrt->mroute_reg_vif_num) ||
2792             nla_put_u8(skb, IPMRA_TABLE_MROUTE_DO_ASSERT,
2793                        mrt->mroute_do_assert) ||
2794             nla_put_u8(skb, IPMRA_TABLE_MROUTE_DO_PIM, mrt->mroute_do_pim))
2795                 return false;
2796
2797         return true;
2798 }
2799
2800 static bool ipmr_fill_vif(struct mr_table *mrt, u32 vifid, struct sk_buff *skb)
2801 {
2802         struct nlattr *vif_nest;
2803         struct vif_device *vif;
2804
2805         /* if the VIF doesn't exist just continue */
2806         if (!VIF_EXISTS(mrt, vifid))
2807                 return true;
2808
2809         vif = &mrt->vif_table[vifid];
2810         vif_nest = nla_nest_start(skb, IPMRA_VIF);
2811         if (!vif_nest)
2812                 return false;
2813         if (nla_put_u32(skb, IPMRA_VIFA_IFINDEX, vif->dev->ifindex) ||
2814             nla_put_u32(skb, IPMRA_VIFA_VIF_ID, vifid) ||
2815             nla_put_u16(skb, IPMRA_VIFA_FLAGS, vif->flags) ||
2816             nla_put_u64_64bit(skb, IPMRA_VIFA_BYTES_IN, vif->bytes_in,
2817                               IPMRA_VIFA_PAD) ||
2818             nla_put_u64_64bit(skb, IPMRA_VIFA_BYTES_OUT, vif->bytes_out,
2819                               IPMRA_VIFA_PAD) ||
2820             nla_put_u64_64bit(skb, IPMRA_VIFA_PACKETS_IN, vif->pkt_in,
2821                               IPMRA_VIFA_PAD) ||
2822             nla_put_u64_64bit(skb, IPMRA_VIFA_PACKETS_OUT, vif->pkt_out,
2823                               IPMRA_VIFA_PAD) ||
2824             nla_put_be32(skb, IPMRA_VIFA_LOCAL_ADDR, vif->local) ||
2825             nla_put_be32(skb, IPMRA_VIFA_REMOTE_ADDR, vif->remote)) {
2826                 nla_nest_cancel(skb, vif_nest);
2827                 return false;
2828         }
2829         nla_nest_end(skb, vif_nest);
2830
2831         return true;
2832 }
2833
2834 static int ipmr_rtm_dumplink(struct sk_buff *skb, struct netlink_callback *cb)
2835 {
2836         struct net *net = sock_net(skb->sk);
2837         struct nlmsghdr *nlh = NULL;
2838         unsigned int t = 0, s_t;
2839         unsigned int e = 0, s_e;
2840         struct mr_table *mrt;
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(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(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 struct ipmr_vif_iter {
2912         struct seq_net_private p;
2913         struct mr_table *mrt;
2914         int ct;
2915 };
2916
2917 static struct vif_device *ipmr_vif_seq_idx(struct net *net,
2918                                            struct ipmr_vif_iter *iter,
2919                                            loff_t pos)
2920 {
2921         struct mr_table *mrt = iter->mrt;
2922
2923         for (iter->ct = 0; iter->ct < mrt->maxvif; ++iter->ct) {
2924                 if (!VIF_EXISTS(mrt, iter->ct))
2925                         continue;
2926                 if (pos-- == 0)
2927                         return &mrt->vif_table[iter->ct];
2928         }
2929         return NULL;
2930 }
2931
2932 static void *ipmr_vif_seq_start(struct seq_file *seq, loff_t *pos)
2933         __acquires(mrt_lock)
2934 {
2935         struct ipmr_vif_iter *iter = seq->private;
2936         struct net *net = seq_file_net(seq);
2937         struct mr_table *mrt;
2938
2939         mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2940         if (!mrt)
2941                 return ERR_PTR(-ENOENT);
2942
2943         iter->mrt = mrt;
2944
2945         read_lock(&mrt_lock);
2946         return *pos ? ipmr_vif_seq_idx(net, seq->private, *pos - 1)
2947                 : SEQ_START_TOKEN;
2948 }
2949
2950 static void *ipmr_vif_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2951 {
2952         struct ipmr_vif_iter *iter = seq->private;
2953         struct net *net = seq_file_net(seq);
2954         struct mr_table *mrt = iter->mrt;
2955
2956         ++*pos;
2957         if (v == SEQ_START_TOKEN)
2958                 return ipmr_vif_seq_idx(net, iter, 0);
2959
2960         while (++iter->ct < mrt->maxvif) {
2961                 if (!VIF_EXISTS(mrt, iter->ct))
2962                         continue;
2963                 return &mrt->vif_table[iter->ct];
2964         }
2965         return NULL;
2966 }
2967
2968 static void ipmr_vif_seq_stop(struct seq_file *seq, void *v)
2969         __releases(mrt_lock)
2970 {
2971         read_unlock(&mrt_lock);
2972 }
2973
2974 static int ipmr_vif_seq_show(struct seq_file *seq, void *v)
2975 {
2976         struct ipmr_vif_iter *iter = seq->private;
2977         struct mr_table *mrt = iter->mrt;
2978
2979         if (v == SEQ_START_TOKEN) {
2980                 seq_puts(seq,
2981                          "Interface      BytesIn  PktsIn  BytesOut PktsOut Flags Local    Remote\n");
2982         } else {
2983                 const struct vif_device *vif = v;
2984                 const char *name =  vif->dev ? vif->dev->name : "none";
2985
2986                 seq_printf(seq,
2987                            "%2zd %-10s %8ld %7ld  %8ld %7ld %05X %08X %08X\n",
2988                            vif - mrt->vif_table,
2989                            name, vif->bytes_in, vif->pkt_in,
2990                            vif->bytes_out, vif->pkt_out,
2991                            vif->flags, vif->local, vif->remote);
2992         }
2993         return 0;
2994 }
2995
2996 static const struct seq_operations ipmr_vif_seq_ops = {
2997         .start = ipmr_vif_seq_start,
2998         .next  = ipmr_vif_seq_next,
2999         .stop  = ipmr_vif_seq_stop,
3000         .show  = ipmr_vif_seq_show,
3001 };
3002
3003 static int ipmr_vif_open(struct inode *inode, struct file *file)
3004 {
3005         return seq_open_net(inode, file, &ipmr_vif_seq_ops,
3006                             sizeof(struct ipmr_vif_iter));
3007 }
3008
3009 static const struct file_operations ipmr_vif_fops = {
3010         .owner   = THIS_MODULE,
3011         .open    = ipmr_vif_open,
3012         .read    = seq_read,
3013         .llseek  = seq_lseek,
3014         .release = seq_release_net,
3015 };
3016
3017 struct ipmr_mfc_iter {
3018         struct seq_net_private p;
3019         struct mr_table *mrt;
3020         struct list_head *cache;
3021 };
3022
3023 static struct mfc_cache *ipmr_mfc_seq_idx(struct net *net,
3024                                           struct ipmr_mfc_iter *it, loff_t pos)
3025 {
3026         struct mr_table *mrt = it->mrt;
3027         struct mfc_cache *mfc;
3028
3029         rcu_read_lock();
3030         it->cache = &mrt->mfc_cache_list;
3031         list_for_each_entry_rcu(mfc, &mrt->mfc_cache_list, list)
3032                 if (pos-- == 0)
3033                         return mfc;
3034         rcu_read_unlock();
3035
3036         spin_lock_bh(&mfc_unres_lock);
3037         it->cache = &mrt->mfc_unres_queue;
3038         list_for_each_entry(mfc, it->cache, list)
3039                 if (pos-- == 0)
3040                         return mfc;
3041         spin_unlock_bh(&mfc_unres_lock);
3042
3043         it->cache = NULL;
3044         return NULL;
3045 }
3046
3047
3048 static void *ipmr_mfc_seq_start(struct seq_file *seq, loff_t *pos)
3049 {
3050         struct ipmr_mfc_iter *it = seq->private;
3051         struct net *net = seq_file_net(seq);
3052         struct mr_table *mrt;
3053
3054         mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
3055         if (!mrt)
3056                 return ERR_PTR(-ENOENT);
3057
3058         it->mrt = mrt;
3059         it->cache = NULL;
3060         return *pos ? ipmr_mfc_seq_idx(net, seq->private, *pos - 1)
3061                 : SEQ_START_TOKEN;
3062 }
3063
3064 static void *ipmr_mfc_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3065 {
3066         struct ipmr_mfc_iter *it = seq->private;
3067         struct net *net = seq_file_net(seq);
3068         struct mr_table *mrt = it->mrt;
3069         struct mfc_cache *mfc = v;
3070
3071         ++*pos;
3072
3073         if (v == SEQ_START_TOKEN)
3074                 return ipmr_mfc_seq_idx(net, seq->private, 0);
3075
3076         if (mfc->list.next != it->cache)
3077                 return list_entry(mfc->list.next, struct mfc_cache, list);
3078
3079         if (it->cache == &mrt->mfc_unres_queue)
3080                 goto end_of_list;
3081
3082         /* exhausted cache_array, show unresolved */
3083         rcu_read_unlock();
3084         it->cache = &mrt->mfc_unres_queue;
3085
3086         spin_lock_bh(&mfc_unres_lock);
3087         if (!list_empty(it->cache))
3088                 return list_first_entry(it->cache, struct mfc_cache, list);
3089
3090 end_of_list:
3091         spin_unlock_bh(&mfc_unres_lock);
3092         it->cache = NULL;
3093
3094         return NULL;
3095 }
3096
3097 static void ipmr_mfc_seq_stop(struct seq_file *seq, void *v)
3098 {
3099         struct ipmr_mfc_iter *it = seq->private;
3100         struct mr_table *mrt = it->mrt;
3101
3102         if (it->cache == &mrt->mfc_unres_queue)
3103                 spin_unlock_bh(&mfc_unres_lock);
3104         else if (it->cache == &mrt->mfc_cache_list)
3105                 rcu_read_unlock();
3106 }
3107
3108 static int ipmr_mfc_seq_show(struct seq_file *seq, void *v)
3109 {
3110         int n;
3111
3112         if (v == SEQ_START_TOKEN) {
3113                 seq_puts(seq,
3114                  "Group    Origin   Iif     Pkts    Bytes    Wrong Oifs\n");
3115         } else {
3116                 const struct mfc_cache *mfc = v;
3117                 const struct ipmr_mfc_iter *it = seq->private;
3118                 const struct mr_table *mrt = it->mrt;
3119
3120                 seq_printf(seq, "%08X %08X %-3hd",
3121                            (__force u32) mfc->mfc_mcastgrp,
3122                            (__force u32) mfc->mfc_origin,
3123                            mfc->mfc_parent);
3124
3125                 if (it->cache != &mrt->mfc_unres_queue) {
3126                         seq_printf(seq, " %8lu %8lu %8lu",
3127                                    mfc->mfc_un.res.pkt,
3128                                    mfc->mfc_un.res.bytes,
3129                                    mfc->mfc_un.res.wrong_if);
3130                         for (n = mfc->mfc_un.res.minvif;
3131                              n < mfc->mfc_un.res.maxvif; n++) {
3132                                 if (VIF_EXISTS(mrt, n) &&
3133                                     mfc->mfc_un.res.ttls[n] < 255)
3134                                         seq_printf(seq,
3135                                            " %2d:%-3d",
3136                                            n, mfc->mfc_un.res.ttls[n]);
3137                         }
3138                 } else {
3139                         /* unresolved mfc_caches don't contain
3140                          * pkt, bytes and wrong_if values
3141                          */
3142                         seq_printf(seq, " %8lu %8lu %8lu", 0ul, 0ul, 0ul);
3143                 }
3144                 seq_putc(seq, '\n');
3145         }
3146         return 0;
3147 }
3148
3149 static const struct seq_operations ipmr_mfc_seq_ops = {
3150         .start = ipmr_mfc_seq_start,
3151         .next  = ipmr_mfc_seq_next,
3152         .stop  = ipmr_mfc_seq_stop,
3153         .show  = ipmr_mfc_seq_show,
3154 };
3155
3156 static int ipmr_mfc_open(struct inode *inode, struct file *file)
3157 {
3158         return seq_open_net(inode, file, &ipmr_mfc_seq_ops,
3159                             sizeof(struct ipmr_mfc_iter));
3160 }
3161
3162 static const struct file_operations ipmr_mfc_fops = {
3163         .owner   = THIS_MODULE,
3164         .open    = ipmr_mfc_open,
3165         .read    = seq_read,
3166         .llseek  = seq_lseek,
3167         .release = seq_release_net,
3168 };
3169 #endif
3170
3171 #ifdef CONFIG_IP_PIMSM_V2
3172 static const struct net_protocol pim_protocol = {
3173         .handler        =       pim_rcv,
3174         .netns_ok       =       1,
3175 };
3176 #endif
3177
3178 static unsigned int ipmr_seq_read(struct net *net)
3179 {
3180         ASSERT_RTNL();
3181
3182         return net->ipv4.ipmr_seq + ipmr_rules_seq_read(net);
3183 }
3184
3185 static int ipmr_dump(struct net *net, struct notifier_block *nb)
3186 {
3187         struct mr_table *mrt;
3188         int err;
3189
3190         err = ipmr_rules_dump(net, nb);
3191         if (err)
3192                 return err;
3193
3194         ipmr_for_each_table(mrt, net) {
3195                 struct vif_device *v = &mrt->vif_table[0];
3196                 struct mfc_cache *mfc;
3197                 int vifi;
3198
3199                 /* Notifiy on table VIF entries */
3200                 read_lock(&mrt_lock);
3201                 for (vifi = 0; vifi < mrt->maxvif; vifi++, v++) {
3202                         if (!v->dev)
3203                                 continue;
3204
3205                         call_ipmr_vif_entry_notifier(nb, net, FIB_EVENT_VIF_ADD,
3206                                                      v, vifi, mrt->id);
3207                 }
3208                 read_unlock(&mrt_lock);
3209
3210                 /* Notify on table MFC entries */
3211                 list_for_each_entry_rcu(mfc, &mrt->mfc_cache_list, list)
3212                         call_ipmr_mfc_entry_notifier(nb, net,
3213                                                      FIB_EVENT_ENTRY_ADD, mfc,
3214                                                      mrt->id);
3215         }
3216
3217         return 0;
3218 }
3219
3220 static const struct fib_notifier_ops ipmr_notifier_ops_template = {
3221         .family         = RTNL_FAMILY_IPMR,
3222         .fib_seq_read   = ipmr_seq_read,
3223         .fib_dump       = ipmr_dump,
3224         .owner          = THIS_MODULE,
3225 };
3226
3227 static int __net_init ipmr_notifier_init(struct net *net)
3228 {
3229         struct fib_notifier_ops *ops;
3230
3231         net->ipv4.ipmr_seq = 0;
3232
3233         ops = fib_notifier_ops_register(&ipmr_notifier_ops_template, net);
3234         if (IS_ERR(ops))
3235                 return PTR_ERR(ops);
3236         net->ipv4.ipmr_notifier_ops = ops;
3237
3238         return 0;
3239 }
3240
3241 static void __net_exit ipmr_notifier_exit(struct net *net)
3242 {
3243         fib_notifier_ops_unregister(net->ipv4.ipmr_notifier_ops);
3244         net->ipv4.ipmr_notifier_ops = NULL;
3245 }
3246
3247 /* Setup for IP multicast routing */
3248 static int __net_init ipmr_net_init(struct net *net)
3249 {
3250         int err;
3251
3252         err = ipmr_notifier_init(net);
3253         if (err)
3254                 goto ipmr_notifier_fail;
3255
3256         err = ipmr_rules_init(net);
3257         if (err < 0)
3258                 goto ipmr_rules_fail;
3259
3260 #ifdef CONFIG_PROC_FS
3261         err = -ENOMEM;
3262         if (!proc_create("ip_mr_vif", 0, net->proc_net, &ipmr_vif_fops))
3263                 goto proc_vif_fail;
3264         if (!proc_create("ip_mr_cache", 0, net->proc_net, &ipmr_mfc_fops))
3265                 goto proc_cache_fail;
3266 #endif
3267         return 0;
3268
3269 #ifdef CONFIG_PROC_FS
3270 proc_cache_fail:
3271         remove_proc_entry("ip_mr_vif", net->proc_net);
3272 proc_vif_fail:
3273         ipmr_rules_exit(net);
3274 #endif
3275 ipmr_rules_fail:
3276         ipmr_notifier_exit(net);
3277 ipmr_notifier_fail:
3278         return err;
3279 }
3280
3281 static void __net_exit ipmr_net_exit(struct net *net)
3282 {
3283 #ifdef CONFIG_PROC_FS
3284         remove_proc_entry("ip_mr_cache", net->proc_net);
3285         remove_proc_entry("ip_mr_vif", net->proc_net);
3286 #endif
3287         ipmr_notifier_exit(net);
3288         ipmr_rules_exit(net);
3289 }
3290
3291 static struct pernet_operations ipmr_net_ops = {
3292         .init = ipmr_net_init,
3293         .exit = ipmr_net_exit,
3294 };
3295
3296 int __init ip_mr_init(void)
3297 {
3298         int err;
3299
3300         mrt_cachep = kmem_cache_create("ip_mrt_cache",
3301                                        sizeof(struct mfc_cache),
3302                                        0, SLAB_HWCACHE_ALIGN | SLAB_PANIC,
3303                                        NULL);
3304
3305         err = register_pernet_subsys(&ipmr_net_ops);
3306         if (err)
3307                 goto reg_pernet_fail;
3308
3309         err = register_netdevice_notifier(&ip_mr_notifier);
3310         if (err)
3311                 goto reg_notif_fail;
3312 #ifdef CONFIG_IP_PIMSM_V2
3313         if (inet_add_protocol(&pim_protocol, IPPROTO_PIM) < 0) {
3314                 pr_err("%s: can't add PIM protocol\n", __func__);
3315                 err = -EAGAIN;
3316                 goto add_proto_fail;
3317         }
3318 #endif
3319         rtnl_register(RTNL_FAMILY_IPMR, RTM_GETROUTE,
3320                       ipmr_rtm_getroute, ipmr_rtm_dumproute, 0);
3321         rtnl_register(RTNL_FAMILY_IPMR, RTM_NEWROUTE,
3322                       ipmr_rtm_route, NULL, 0);
3323         rtnl_register(RTNL_FAMILY_IPMR, RTM_DELROUTE,
3324                       ipmr_rtm_route, NULL, 0);
3325
3326         rtnl_register(RTNL_FAMILY_IPMR, RTM_GETLINK,
3327                       NULL, ipmr_rtm_dumplink, 0);
3328         return 0;
3329
3330 #ifdef CONFIG_IP_PIMSM_V2
3331 add_proto_fail:
3332         unregister_netdevice_notifier(&ip_mr_notifier);
3333 #endif
3334 reg_notif_fail:
3335         unregister_pernet_subsys(&ipmr_net_ops);
3336 reg_pernet_fail:
3337         kmem_cache_destroy(mrt_cachep);
3338         return err;
3339 }