net/ipv4/fib_frontend.c

   1 /*
   2  * INET         An implementation of the TCP/IP protocol suite for the LINUX
   3  *              operating system.  INET is implemented using the  BSD Socket
   4  *              interface as the means of communication with the user level.
   5  *
   6  *              IPv4 Forwarding Information Base: FIB frontend.
   7  *
   8  * Authors:     Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
   9  *
  10  *              This program is free software; you can redistribute it and/or
  11  *              modify it under the terms of the GNU General Public License
  12  *              as published by the Free Software Foundation; either version
  13  *              2 of the License, or (at your option) any later version.
  14  */
  15
  16 #include <linux/module.h>
  17 #include <asm/uaccess.h>
  18 #include <linux/bitops.h>
  19 #include <linux/capability.h>
  20 #include <linux/types.h>
  21 #include <linux/kernel.h>
  22 #include <linux/mm.h>
  23 #include <linux/string.h>
  24 #include <linux/socket.h>
  25 #include <linux/sockios.h>
  26 #include <linux/errno.h>
  27 #include <linux/in.h>
  28 #include <linux/inet.h>
  29 #include <linux/inetdevice.h>
  30 #include <linux/netdevice.h>
  31 #include <linux/if_addr.h>
  32 #include <linux/if_arp.h>
  33 #include <linux/skbuff.h>
  34 #include <linux/cache.h>
  35 #include <linux/init.h>
  36 #include <linux/list.h>
  37 #include <linux/slab.h>
  38
  39 #include <net/ip.h>
  40 #include <net/protocol.h>
  41 #include <net/route.h>
  42 #include <net/tcp.h>
  43 #include <net/sock.h>
  44 #include <net/arp.h>
  45 #include <net/ip_fib.h>
  46 #include <net/rtnetlink.h>
  47 #include <net/xfrm.h>
  48 #include <net/vrf.h>
  49
  50 #ifndef CONFIG_IP_MULTIPLE_TABLES
  51
  52 static int __net_init fib4_rules_init(struct net *net)
  53 {
  54         struct fib_table *local_table, *main_table;
  55
  56         main_table  = fib_trie_table(RT_TABLE_MAIN, NULL);
  57         if (!main_table)
  58                 return -ENOMEM;
  59
  60         local_table = fib_trie_table(RT_TABLE_LOCAL, main_table);
  61         if (!local_table)
  62                 goto fail;
  63
  64         hlist_add_head_rcu(&local_table->tb_hlist,
  65                                 &net->ipv4.fib_table_hash[TABLE_LOCAL_INDEX]);
  66         hlist_add_head_rcu(&main_table->tb_hlist,
  67                                 &net->ipv4.fib_table_hash[TABLE_MAIN_INDEX]);
  68         return 0;
  69
  70 fail:
  71         fib_free_table(main_table);
  72         return -ENOMEM;
  73 }
  74 #else
  75
  76 struct fib_table *fib_new_table(struct net *net, u32 id)
  77 {
  78         struct fib_table *tb, *alias = NULL;
  79         unsigned int h;
  80
  81         if (id == 0)
  82                 id = RT_TABLE_MAIN;
  83         tb = fib_get_table(net, id);
  84         if (tb)
  85                 return tb;
  86
  87         if (id == RT_TABLE_LOCAL)
  88                 alias = fib_new_table(net, RT_TABLE_MAIN);
  89
  90         tb = fib_trie_table(id, alias);
  91         if (!tb)
  92                 return NULL;
  93
  94         switch (id) {
  95         case RT_TABLE_LOCAL:
  96                 rcu_assign_pointer(net->ipv4.fib_local, tb);
  97                 break;
  98         case RT_TABLE_MAIN:
  99                 rcu_assign_pointer(net->ipv4.fib_main, tb);
 100                 break;
 101         case RT_TABLE_DEFAULT:
 102                 rcu_assign_pointer(net->ipv4.fib_default, tb);
 103                 break;
 104         default:
 105                 break;
 106         }
 107
 108         h = id & (FIB_TABLE_HASHSZ - 1);
 109         hlist_add_head_rcu(&tb->tb_hlist, &net->ipv4.fib_table_hash[h]);
 110         return tb;
 111 }
 112
 113 /* caller must hold either rtnl or rcu read lock */
 114 struct fib_table *fib_get_table(struct net *net, u32 id)
 115 {
 116         struct fib_table *tb;
 117         struct hlist_head *head;
 118         unsigned int h;
 119
 120         if (id == 0)
 121                 id = RT_TABLE_MAIN;
 122         h = id & (FIB_TABLE_HASHSZ - 1);
 123
 124         head = &net->ipv4.fib_table_hash[h];
 125         hlist_for_each_entry_rcu(tb, head, tb_hlist) {
 126                 if (tb->tb_id == id)
 127                         return tb;
 128         }
 129         return NULL;
 130 }
 131 #endif /* CONFIG_IP_MULTIPLE_TABLES */
 132
 133 static void fib_replace_table(struct net *net, struct fib_table *old,
 134                               struct fib_table *new)
 135 {
 136 #ifdef CONFIG_IP_MULTIPLE_TABLES
 137         switch (new->tb_id) {
 138         case RT_TABLE_LOCAL:
 139                 rcu_assign_pointer(net->ipv4.fib_local, new);
 140                 break;
 141         case RT_TABLE_MAIN:
 142                 rcu_assign_pointer(net->ipv4.fib_main, new);
 143                 break;
 144         case RT_TABLE_DEFAULT:
 145                 rcu_assign_pointer(net->ipv4.fib_default, new);
 146                 break;
 147         default:
 148                 break;
 149         }
 150
 151 #endif
 152         /* replace the old table in the hlist */
 153         hlist_replace_rcu(&old->tb_hlist, &new->tb_hlist);
 154 }
 155
 156 int fib_unmerge(struct net *net)
 157 {
 158         struct fib_table *old, *new;
 159
 160         /* attempt to fetch local table if it has been allocated */
 161         old = fib_get_table(net, RT_TABLE_LOCAL);
 162         if (!old)
 163                 return 0;
 164
 165         new = fib_trie_unmerge(old);
 166         if (!new)
 167                 return -ENOMEM;
 168
 169         /* replace merged table with clean table */
 170         if (new != old) {
 171                 fib_replace_table(net, old, new);
 172                 fib_free_table(old);
 173         }
 174
 175         return 0;
 176 }
 177
 178 static void fib_flush(struct net *net)
 179 {
 180         int flushed = 0;
 181         unsigned int h;
 182
 183         for (h = 0; h < FIB_TABLE_HASHSZ; h++) {
 184                 struct hlist_head *head = &net->ipv4.fib_table_hash[h];
 185                 struct hlist_node *tmp;
 186                 struct fib_table *tb;
 187
 188                 hlist_for_each_entry_safe(tb, tmp, head, tb_hlist)
 189                         flushed += fib_table_flush(tb);
 190         }
 191
 192         if (flushed)
 193                 rt_cache_flush(net);
 194 }
 195
 196 void fib_flush_external(struct net *net)
 197 {
 198         struct fib_table *tb;
 199         struct hlist_head *head;
 200         unsigned int h;
 201
 202         for (h = 0; h < FIB_TABLE_HASHSZ; h++) {
 203                 head = &net->ipv4.fib_table_hash[h];
 204                 hlist_for_each_entry(tb, head, tb_hlist)
 205                         fib_table_flush_external(tb);
 206         }
 207 }
 208
 209 /*
 210  * Find address type as if only "dev" was present in the system. If
 211  * on_dev is NULL then all interfaces are taken into consideration.
 212  */
 213 static inline unsigned int __inet_dev_addr_type(struct net *net,
 214                                                 const struct net_device *dev,
 215                                                 __be32 addr)
 216 {
 217         struct flowi4           fl4 = { .daddr = addr };
 218         struct fib_result       res;
 219         unsigned int ret = RTN_BROADCAST;
 220         struct fib_table *local_table;
 221
 222         if (ipv4_is_zeronet(addr) || ipv4_is_lbcast(addr))
 223                 return RTN_BROADCAST;
 224         if (ipv4_is_multicast(addr))
 225                 return RTN_MULTICAST;
 226
 227         rcu_read_lock();
 228
 229         local_table = fib_get_table(net, RT_TABLE_LOCAL);
 230         if (local_table) {
 231                 ret = RTN_UNICAST;
 232                 if (!fib_table_lookup(local_table, &fl4, &res, FIB_LOOKUP_NOREF)) {
 233                         if (!dev || dev == res.fi->fib_dev)
 234                                 ret = res.type;
 235                 }
 236         }
 237
 238         rcu_read_unlock();
 239         return ret;
 240 }
 241
 242 unsigned int inet_addr_type(struct net *net, __be32 addr)
 243 {
 244         return __inet_dev_addr_type(net, NULL, addr);
 245 }
 246 EXPORT_SYMBOL(inet_addr_type);
 247
 248 unsigned int inet_dev_addr_type(struct net *net, const struct net_device *dev,
 249                                 __be32 addr)
 250 {
 251         return __inet_dev_addr_type(net, dev, addr);
 252 }
 253 EXPORT_SYMBOL(inet_dev_addr_type);
 254
 255 __be32 fib_compute_spec_dst(struct sk_buff *skb)
 256 {
 257         struct net_device *dev = skb->dev;
 258         struct in_device *in_dev;
 259         struct fib_result res;
 260         struct rtable *rt;
 261         struct flowi4 fl4;
 262         struct net *net;
 263         int scope;
 264
 265         rt = skb_rtable(skb);
 266         if ((rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST | RTCF_LOCAL)) ==
 267             RTCF_LOCAL)
 268                 return ip_hdr(skb)->daddr;
 269
 270         in_dev = __in_dev_get_rcu(dev);
 271         BUG_ON(!in_dev);
 272
 273         net = dev_net(dev);
 274
 275         scope = RT_SCOPE_UNIVERSE;
 276         if (!ipv4_is_zeronet(ip_hdr(skb)->saddr)) {
 277                 fl4.flowi4_oif = 0;
 278                 fl4.flowi4_iif = LOOPBACK_IFINDEX;
 279                 fl4.daddr = ip_hdr(skb)->saddr;
 280                 fl4.saddr = 0;
 281                 fl4.flowi4_tos = RT_TOS(ip_hdr(skb)->tos);
 282                 fl4.flowi4_scope = scope;
 283                 fl4.flowi4_mark = IN_DEV_SRC_VMARK(in_dev) ? skb->mark : 0;
 284                 fl4.flowi4_tun_key.tun_id = 0;
 285                 if (!fib_lookup(net, &fl4, &res, 0))
 286                         return FIB_RES_PREFSRC(net, res);
 287         } else {
 288                 scope = RT_SCOPE_LINK;
 289         }
 290
 291         return inet_select_addr(dev, ip_hdr(skb)->saddr, scope);
 292 }
 293
 294 /* Given (packet source, input interface) and optional (dst, oif, tos):
 295  * - (main) check, that source is valid i.e. not broadcast or our local
 296  *   address.
 297  * - figure out what "logical" interface this packet arrived
 298  *   and calculate "specific destination" address.
 299  * - check, that packet arrived from expected physical interface.
 300  * called with rcu_read_lock()
 301  */
 302 static int __fib_validate_source(struct sk_buff *skb, __be32 src, __be32 dst,
 303                                  u8 tos, int oif, struct net_device *dev,
 304                                  int rpf, struct in_device *idev, u32 *itag)
 305 {
 306         int ret, no_addr;
 307         struct fib_result res;
 308         struct flowi4 fl4;
 309         struct net *net;
 310         bool dev_match;
 311
 312         fl4.flowi4_oif = 0;
 313         fl4.flowi4_iif = vrf_master_ifindex_rcu(dev);
 314         if (!fl4.flowi4_iif)
 315                 fl4.flowi4_iif = oif ? : LOOPBACK_IFINDEX;
 316         fl4.daddr = src;
 317         fl4.saddr = dst;
 318         fl4.flowi4_tos = tos;
 319         fl4.flowi4_scope = RT_SCOPE_UNIVERSE;
 320         fl4.flowi4_tun_key.tun_id = 0;
 321
 322         no_addr = idev->ifa_list == NULL;
 323
 324         fl4.flowi4_mark = IN_DEV_SRC_VMARK(idev) ? skb->mark : 0;
 325
 326         net = dev_net(dev);
 327         if (fib_lookup(net, &fl4, &res, 0))
 328                 goto last_resort;
 329         if (res.type != RTN_UNICAST &&
 330             (res.type != RTN_LOCAL || !IN_DEV_ACCEPT_LOCAL(idev)))
 331                 goto e_inval;
 332         if (!rpf && !fib_num_tclassid_users(dev_net(dev)) &&
 333             (dev->ifindex != oif || !IN_DEV_TX_REDIRECTS(idev)))
 334                 goto last_resort;
 335         fib_combine_itag(itag, &res);
 336         dev_match = false;
 337
 338 #ifdef CONFIG_IP_ROUTE_MULTIPATH
 339         for (ret = 0; ret < res.fi->fib_nhs; ret++) {
 340                 struct fib_nh *nh = &res.fi->fib_nh[ret];
 341
 342                 if (nh->nh_dev == dev) {
 343                         dev_match = true;
 344                         break;
 345                 } else if (vrf_master_ifindex_rcu(nh->nh_dev) == dev->ifindex) {
 346                         dev_match = true;
 347                         break;
 348                 }
 349         }
 350 #else
 351         if (FIB_RES_DEV(res) == dev)
 352                 dev_match = true;
 353 #endif
 354         if (dev_match) {
 355                 ret = FIB_RES_NH(res).nh_scope >= RT_SCOPE_HOST;
 356                 return ret;
 357         }
 358         if (no_addr)
 359                 goto last_resort;
 360         if (rpf == 1)
 361                 goto e_rpf;
 362         fl4.flowi4_oif = dev->ifindex;
 363
 364         ret = 0;
 365         if (fib_lookup(net, &fl4, &res, FIB_LOOKUP_IGNORE_LINKSTATE) == 0) {
 366                 if (res.type == RTN_UNICAST)
 367                         ret = FIB_RES_NH(res).nh_scope >= RT_SCOPE_HOST;
 368         }
 369         return ret;
 370
 371 last_resort:
 372         if (rpf)
 373                 goto e_rpf;
 374         *itag = 0;
 375         return 0;
 376
 377 e_inval:
 378         return -EINVAL;
 379 e_rpf:
 380         return -EXDEV;
 381 }
 382
 383 /* Ignore rp_filter for packets protected by IPsec. */
 384 int fib_validate_source(struct sk_buff *skb, __be32 src, __be32 dst,
 385                         u8 tos, int oif, struct net_device *dev,
 386                         struct in_device *idev, u32 *itag)
 387 {
 388         int r = secpath_exists(skb) ? 0 : IN_DEV_RPFILTER(idev);
 389
 390         if (!r && !fib_num_tclassid_users(dev_net(dev)) &&
 391             IN_DEV_ACCEPT_LOCAL(idev) &&
 392             (dev->ifindex != oif || !IN_DEV_TX_REDIRECTS(idev))) {
 393                 *itag = 0;
 394                 return 0;
 395         }
 396         return __fib_validate_source(skb, src, dst, tos, oif, dev, r, idev, itag);
 397 }
 398
 399 static inline __be32 sk_extract_addr(struct sockaddr *addr)
 400 {
 401         return ((struct sockaddr_in *) addr)->sin_addr.s_addr;
 402 }
 403
 404 static int put_rtax(struct nlattr *mx, int len, int type, u32 value)
 405 {
 406         struct nlattr *nla;
 407
 408         nla = (struct nlattr *) ((char *) mx + len);
 409         nla->nla_type = type;
 410         nla->nla_len = nla_attr_size(4);
 411         *(u32 *) nla_data(nla) = value;
 412
 413         return len + nla_total_size(4);
 414 }
 415
 416 static int rtentry_to_fib_config(struct net *net, int cmd, struct rtentry *rt,
 417                                  struct fib_config *cfg)
 418 {
 419         __be32 addr;
 420         int plen;
 421
 422         memset(cfg, 0, sizeof(*cfg));
 423         cfg->fc_nlinfo.nl_net = net;
 424
 425         if (rt->rt_dst.sa_family != AF_INET)
 426                 return -EAFNOSUPPORT;
 427
 428         /*
 429          * Check mask for validity:
 430          * a) it must be contiguous.
 431          * b) destination must have all host bits clear.
 432          * c) if application forgot to set correct family (AF_INET),
 433          *    reject request unless it is absolutely clear i.e.
 434          *    both family and mask are zero.
 435          */
 436         plen = 32;
 437         addr = sk_extract_addr(&rt->rt_dst);
 438         if (!(rt->rt_flags & RTF_HOST)) {
 439                 __be32 mask = sk_extract_addr(&rt->rt_genmask);
 440
 441                 if (rt->rt_genmask.sa_family != AF_INET) {
 442                         if (mask || rt->rt_genmask.sa_family)
 443                                 return -EAFNOSUPPORT;
 444                 }
 445
 446                 if (bad_mask(mask, addr))
 447                         return -EINVAL;
 448
 449                 plen = inet_mask_len(mask);
 450         }
 451
 452         cfg->fc_dst_len = plen;
 453         cfg->fc_dst = addr;
 454
 455         if (cmd != SIOCDELRT) {
 456                 cfg->fc_nlflags = NLM_F_CREATE;
 457                 cfg->fc_protocol = RTPROT_BOOT;
 458         }
 459
 460         if (rt->rt_metric)
 461                 cfg->fc_priority = rt->rt_metric - 1;
 462
 463         if (rt->rt_flags & RTF_REJECT) {
 464                 cfg->fc_scope = RT_SCOPE_HOST;
 465                 cfg->fc_type = RTN_UNREACHABLE;
 466                 return 0;
 467         }
 468
 469         cfg->fc_scope = RT_SCOPE_NOWHERE;
 470         cfg->fc_type = RTN_UNICAST;
 471
 472         if (rt->rt_dev) {
 473                 char *colon;
 474                 struct net_device *dev;
 475                 char devname[IFNAMSIZ];
 476
 477                 if (copy_from_user(devname, rt->rt_dev, IFNAMSIZ-1))
 478                         return -EFAULT;
 479
 480                 devname[IFNAMSIZ-1] = 0;
 481                 colon = strchr(devname, ':');
 482                 if (colon)
 483                         *colon = 0;
 484                 dev = __dev_get_by_name(net, devname);
 485                 if (!dev)
 486                         return -ENODEV;
 487                 cfg->fc_oif = dev->ifindex;
 488                 if (colon) {
 489                         struct in_ifaddr *ifa;
 490                         struct in_device *in_dev = __in_dev_get_rtnl(dev);
 491                         if (!in_dev)
 492                                 return -ENODEV;
 493                         *colon = ':';
 494                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next)
 495                                 if (strcmp(ifa->ifa_label, devname) == 0)
 496                                         break;
 497                         if (!ifa)
 498                                 return -ENODEV;
 499                         cfg->fc_prefsrc = ifa->ifa_local;
 500                 }
 501         }
 502
 503         addr = sk_extract_addr(&rt->rt_gateway);
 504         if (rt->rt_gateway.sa_family == AF_INET && addr) {
 505                 cfg->fc_gw = addr;
 506                 if (rt->rt_flags & RTF_GATEWAY &&
 507                     inet_addr_type(net, addr) == RTN_UNICAST)
 508                         cfg->fc_scope = RT_SCOPE_UNIVERSE;
 509         }
 510
 511         if (cmd == SIOCDELRT)
 512                 return 0;
 513
 514         if (rt->rt_flags & RTF_GATEWAY && !cfg->fc_gw)
 515                 return -EINVAL;
 516
 517         if (cfg->fc_scope == RT_SCOPE_NOWHERE)
 518                 cfg->fc_scope = RT_SCOPE_LINK;
 519
 520         if (rt->rt_flags & (RTF_MTU | RTF_WINDOW | RTF_IRTT)) {
 521                 struct nlattr *mx;
 522                 int len = 0;
 523
 524                 mx = kzalloc(3 * nla_total_size(4), GFP_KERNEL);
 525                 if (!mx)
 526                         return -ENOMEM;
 527
 528                 if (rt->rt_flags & RTF_MTU)
 529                         len = put_rtax(mx, len, RTAX_ADVMSS, rt->rt_mtu - 40);
 530
 531                 if (rt->rt_flags & RTF_WINDOW)
 532                         len = put_rtax(mx, len, RTAX_WINDOW, rt->rt_window);
 533
 534                 if (rt->rt_flags & RTF_IRTT)
 535                         len = put_rtax(mx, len, RTAX_RTT, rt->rt_irtt << 3);
 536
 537                 cfg->fc_mx = mx;
 538                 cfg->fc_mx_len = len;
 539         }
 540
 541         return 0;
 542 }
 543
 544 /*
 545  * Handle IP routing ioctl calls.
 546  * These are used to manipulate the routing tables
 547  */
 548 int ip_rt_ioctl(struct net *net, unsigned int cmd, void __user *arg)
 549 {
 550         struct fib_config cfg;
 551         struct rtentry rt;
 552         int err;
 553
 554         switch (cmd) {
 555         case SIOCADDRT:         /* Add a route */
 556         case SIOCDELRT:         /* Delete a route */
 557                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
 558                         return -EPERM;
 559
 560                 if (copy_from_user(&rt, arg, sizeof(rt)))
 561                         return -EFAULT;
 562
 563                 rtnl_lock();
 564                 err = rtentry_to_fib_config(net, cmd, &rt, &cfg);
 565                 if (err == 0) {
 566                         struct fib_table *tb;
 567
 568                         if (cmd == SIOCDELRT) {
 569                                 tb = fib_get_table(net, cfg.fc_table);
 570                                 if (tb)
 571                                         err = fib_table_delete(tb, &cfg);
 572                                 else
 573                                         err = -ESRCH;
 574                         } else {
 575                                 tb = fib_new_table(net, cfg.fc_table);
 576                                 if (tb)
 577                                         err = fib_table_insert(tb, &cfg);
 578                                 else
 579                                         err = -ENOBUFS;
 580                         }
 581
 582                         /* allocated by rtentry_to_fib_config() */
 583                         kfree(cfg.fc_mx);
 584                 }
 585                 rtnl_unlock();
 586                 return err;
 587         }
 588         return -EINVAL;
 589 }
 590
 591 const struct nla_policy rtm_ipv4_policy[RTA_MAX + 1] = {
 592         [RTA_DST]               = { .type = NLA_U32 },
 593         [RTA_SRC]               = { .type = NLA_U32 },
 594         [RTA_IIF]               = { .type = NLA_U32 },
 595         [RTA_OIF]               = { .type = NLA_U32 },
 596         [RTA_GATEWAY]           = { .type = NLA_U32 },
 597         [RTA_PRIORITY]          = { .type = NLA_U32 },
 598         [RTA_PREFSRC]           = { .type = NLA_U32 },
 599         [RTA_METRICS]           = { .type = NLA_NESTED },
 600         [RTA_MULTIPATH]         = { .len = sizeof(struct rtnexthop) },
 601         [RTA_FLOW]              = { .type = NLA_U32 },
 602         [RTA_ENCAP_TYPE]        = { .type = NLA_U16 },
 603         [RTA_ENCAP]             = { .type = NLA_NESTED },
 604 };
 605
 606 static int rtm_to_fib_config(struct net *net, struct sk_buff *skb,
 607                              struct nlmsghdr *nlh, struct fib_config *cfg)
 608 {
 609         struct nlattr *attr;
 610         int err, remaining;
 611         struct rtmsg *rtm;
 612
 613         err = nlmsg_validate(nlh, sizeof(*rtm), RTA_MAX, rtm_ipv4_policy);
 614         if (err < 0)
 615                 goto errout;
 616
 617         memset(cfg, 0, sizeof(*cfg));
 618
 619         rtm = nlmsg_data(nlh);
 620         cfg->fc_dst_len = rtm->rtm_dst_len;
 621         cfg->fc_tos = rtm->rtm_tos;
 622         cfg->fc_table = rtm->rtm_table;
 623         cfg->fc_protocol = rtm->rtm_protocol;
 624         cfg->fc_scope = rtm->rtm_scope;
 625         cfg->fc_type = rtm->rtm_type;
 626         cfg->fc_flags = rtm->rtm_flags;
 627         cfg->fc_nlflags = nlh->nlmsg_flags;
 628
 629         cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid;
 630         cfg->fc_nlinfo.nlh = nlh;
 631         cfg->fc_nlinfo.nl_net = net;
 632
 633         if (cfg->fc_type > RTN_MAX) {
 634                 err = -EINVAL;
 635                 goto errout;
 636         }
 637
 638         nlmsg_for_each_attr(attr, nlh, sizeof(struct rtmsg), remaining) {
 639                 switch (nla_type(attr)) {
 640                 case RTA_DST:
 641                         cfg->fc_dst = nla_get_be32(attr);
 642                         break;
 643                 case RTA_OIF:
 644                         cfg->fc_oif = nla_get_u32(attr);
 645                         break;
 646                 case RTA_GATEWAY:
 647                         cfg->fc_gw = nla_get_be32(attr);
 648                         break;
 649                 case RTA_PRIORITY:
 650                         cfg->fc_priority = nla_get_u32(attr);
 651                         break;
 652                 case RTA_PREFSRC:
 653                         cfg->fc_prefsrc = nla_get_be32(attr);
 654                         break;
 655                 case RTA_METRICS:
 656                         cfg->fc_mx = nla_data(attr);
 657                         cfg->fc_mx_len = nla_len(attr);
 658                         break;
 659                 case RTA_MULTIPATH:
 660                         cfg->fc_mp = nla_data(attr);
 661                         cfg->fc_mp_len = nla_len(attr);
 662                         break;
 663                 case RTA_FLOW:
 664                         cfg->fc_flow = nla_get_u32(attr);
 665                         break;
 666                 case RTA_TABLE:
 667                         cfg->fc_table = nla_get_u32(attr);
 668                         break;
 669                 case RTA_ENCAP:
 670                         cfg->fc_encap = attr;
 671                         break;
 672                 case RTA_ENCAP_TYPE:
 673                         cfg->fc_encap_type = nla_get_u16(attr);
 674                         break;
 675                 }
 676         }
 677
 678         return 0;
 679 errout:
 680         return err;
 681 }
 682
 683 static int inet_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh)
 684 {
 685         struct net *net = sock_net(skb->sk);
 686         struct fib_config cfg;
 687         struct fib_table *tb;
 688         int err;
 689
 690         err = rtm_to_fib_config(net, skb, nlh, &cfg);
 691         if (err < 0)
 692                 goto errout;
 693
 694         tb = fib_get_table(net, cfg.fc_table);
 695         if (!tb) {
 696                 err = -ESRCH;
 697                 goto errout;
 698         }
 699
 700         err = fib_table_delete(tb, &cfg);
 701 errout:
 702         return err;
 703 }
 704
 705 static int inet_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh)
 706 {
 707         struct net *net = sock_net(skb->sk);
 708         struct fib_config cfg;
 709         struct fib_table *tb;
 710         int err;
 711
 712         err = rtm_to_fib_config(net, skb, nlh, &cfg);
 713         if (err < 0)
 714                 goto errout;
 715
 716         tb = fib_new_table(net, cfg.fc_table);
 717         if (!tb) {
 718                 err = -ENOBUFS;
 719                 goto errout;
 720         }
 721
 722         err = fib_table_insert(tb, &cfg);
 723 errout:
 724         return err;
 725 }
 726
 727 static int inet_dump_fib(struct sk_buff *skb, struct netlink_callback *cb)
 728 {
 729         struct net *net = sock_net(skb->sk);
 730         unsigned int h, s_h;
 731         unsigned int e = 0, s_e;
 732         struct fib_table *tb;
 733         struct hlist_head *head;
 734         int dumped = 0;
 735
 736         if (nlmsg_len(cb->nlh) >= sizeof(struct rtmsg) &&
 737             ((struct rtmsg *) nlmsg_data(cb->nlh))->rtm_flags & RTM_F_CLONED)
 738                 return skb->len;
 739
 740         s_h = cb->args[0];
 741         s_e = cb->args[1];
 742
 743         rcu_read_lock();
 744
 745         for (h = s_h; h < FIB_TABLE_HASHSZ; h++, s_e = 0) {
 746                 e = 0;
 747                 head = &net->ipv4.fib_table_hash[h];
 748                 hlist_for_each_entry_rcu(tb, head, tb_hlist) {
 749                         if (e < s_e)
 750                                 goto next;
 751                         if (dumped)
 752                                 memset(&cb->args[2], 0, sizeof(cb->args) -
 753                                                  2 * sizeof(cb->args[0]));
 754                         if (fib_table_dump(tb, skb, cb) < 0)
 755                                 goto out;
 756                         dumped = 1;
 757 next:
 758                         e++;
 759                 }
 760         }
 761 out:
 762         rcu_read_unlock();
 763
 764         cb->args[1] = e;
 765         cb->args[0] = h;
 766
 767         return skb->len;
 768 }
 769
 770 /* Prepare and feed intra-kernel routing request.
 771  * Really, it should be netlink message, but :-( netlink
 772  * can be not configured, so that we feed it directly
 773  * to fib engine. It is legal, because all events occur
 774  * only when netlink is already locked.
 775  */
 776 static void fib_magic(int cmd, int type, __be32 dst, int dst_len, struct in_ifaddr *ifa)
 777 {
 778         struct net *net = dev_net(ifa->ifa_dev->dev);
 779         struct fib_table *tb;
 780         struct fib_config cfg = {
 781                 .fc_protocol = RTPROT_KERNEL,
 782                 .fc_type = type,
 783                 .fc_dst = dst,
 784                 .fc_dst_len = dst_len,
 785                 .fc_prefsrc = ifa->ifa_local,
 786                 .fc_oif = ifa->ifa_dev->dev->ifindex,
 787                 .fc_nlflags = NLM_F_CREATE | NLM_F_APPEND,
 788                 .fc_nlinfo = {
 789                         .nl_net = net,
 790                 },
 791         };
 792
 793         if (type == RTN_UNICAST)
 794                 tb = fib_new_table(net, RT_TABLE_MAIN);
 795         else
 796                 tb = fib_new_table(net, RT_TABLE_LOCAL);
 797
 798         if (!tb)
 799                 return;
 800
 801         cfg.fc_table = tb->tb_id;
 802
 803         if (type != RTN_LOCAL)
 804                 cfg.fc_scope = RT_SCOPE_LINK;
 805         else
 806                 cfg.fc_scope = RT_SCOPE_HOST;
 807
 808         if (cmd == RTM_NEWROUTE)
 809                 fib_table_insert(tb, &cfg);
 810         else
 811                 fib_table_delete(tb, &cfg);
 812 }
 813
 814 void fib_add_ifaddr(struct in_ifaddr *ifa)
 815 {
 816         struct in_device *in_dev = ifa->ifa_dev;
 817         struct net_device *dev = in_dev->dev;
 818         struct in_ifaddr *prim = ifa;
 819         __be32 mask = ifa->ifa_mask;
 820         __be32 addr = ifa->ifa_local;
 821         __be32 prefix = ifa->ifa_address & mask;
 822
 823         if (ifa->ifa_flags & IFA_F_SECONDARY) {
 824                 prim = inet_ifa_byprefix(in_dev, prefix, mask);
 825                 if (!prim) {
 826                         pr_warn("%s: bug: prim == NULL\n", __func__);
 827                         return;
 828                 }
 829         }
 830
 831         fib_magic(RTM_NEWROUTE, RTN_LOCAL, addr, 32, prim);
 832
 833         if (!(dev->flags & IFF_UP))
 834                 return;
 835
 836         /* Add broadcast address, if it is explicitly assigned. */
 837         if (ifa->ifa_broadcast && ifa->ifa_broadcast != htonl(0xFFFFFFFF))
 838                 fib_magic(RTM_NEWROUTE, RTN_BROADCAST, ifa->ifa_broadcast, 32, prim);
 839
 840         if (!ipv4_is_zeronet(prefix) && !(ifa->ifa_flags & IFA_F_SECONDARY) &&
 841             (prefix != addr || ifa->ifa_prefixlen < 32)) {
 842                 fib_magic(RTM_NEWROUTE,
 843                           dev->flags & IFF_LOOPBACK ? RTN_LOCAL : RTN_UNICAST,
 844                           prefix, ifa->ifa_prefixlen, prim);
 845
 846                 /* Add network specific broadcasts, when it takes a sense */
 847                 if (ifa->ifa_prefixlen < 31) {
 848                         fib_magic(RTM_NEWROUTE, RTN_BROADCAST, prefix, 32, prim);
 849                         fib_magic(RTM_NEWROUTE, RTN_BROADCAST, prefix | ~mask,
 850                                   32, prim);
 851                 }
 852         }
 853 }
 854
 855 /* Delete primary or secondary address.
 856  * Optionally, on secondary address promotion consider the addresses
 857  * from subnet iprim as deleted, even if they are in device list.
 858  * In this case the secondary ifa can be in device list.
 859  */
 860 void fib_del_ifaddr(struct in_ifaddr *ifa, struct in_ifaddr *iprim)
 861 {
 862         struct in_device *in_dev = ifa->ifa_dev;
 863         struct net_device *dev = in_dev->dev;
 864         struct in_ifaddr *ifa1;
 865         struct in_ifaddr *prim = ifa, *prim1 = NULL;
 866         __be32 brd = ifa->ifa_address | ~ifa->ifa_mask;
 867         __be32 any = ifa->ifa_address & ifa->ifa_mask;
 868 #define LOCAL_OK        1
 869 #define BRD_OK          2
 870 #define BRD0_OK         4
 871 #define BRD1_OK         8
 872         unsigned int ok = 0;
 873         int subnet = 0;         /* Primary network */
 874         int gone = 1;           /* Address is missing */
 875         int same_prefsrc = 0;   /* Another primary with same IP */
 876
 877         if (ifa->ifa_flags & IFA_F_SECONDARY) {
 878                 prim = inet_ifa_byprefix(in_dev, any, ifa->ifa_mask);
 879                 if (!prim) {
 880                         pr_warn("%s: bug: prim == NULL\n", __func__);
 881                         return;
 882                 }
 883                 if (iprim && iprim != prim) {
 884                         pr_warn("%s: bug: iprim != prim\n", __func__);
 885                         return;
 886                 }
 887         } else if (!ipv4_is_zeronet(any) &&
 888                    (any != ifa->ifa_local || ifa->ifa_prefixlen < 32)) {
 889                 fib_magic(RTM_DELROUTE,
 890                           dev->flags & IFF_LOOPBACK ? RTN_LOCAL : RTN_UNICAST,
 891                           any, ifa->ifa_prefixlen, prim);
 892                 subnet = 1;
 893         }
 894
 895         /* Deletion is more complicated than add.
 896          * We should take care of not to delete too much :-)
 897          *
 898          * Scan address list to be sure that addresses are really gone.
 899          */
 900
 901         for (ifa1 = in_dev->ifa_list; ifa1; ifa1 = ifa1->ifa_next) {
 902                 if (ifa1 == ifa) {
 903                         /* promotion, keep the IP */
 904                         gone = 0;
 905                         continue;
 906                 }
 907                 /* Ignore IFAs from our subnet */
 908                 if (iprim && ifa1->ifa_mask == iprim->ifa_mask &&
 909                     inet_ifa_match(ifa1->ifa_address, iprim))
 910                         continue;
 911
 912                 /* Ignore ifa1 if it uses different primary IP (prefsrc) */
 913                 if (ifa1->ifa_flags & IFA_F_SECONDARY) {
 914                         /* Another address from our subnet? */
 915                         if (ifa1->ifa_mask == prim->ifa_mask &&
 916                             inet_ifa_match(ifa1->ifa_address, prim))
 917                                 prim1 = prim;
 918                         else {
 919                                 /* We reached the secondaries, so
 920                                  * same_prefsrc should be determined.
 921                                  */
 922                                 if (!same_prefsrc)
 923                                         continue;
 924                                 /* Search new prim1 if ifa1 is not
 925                                  * using the current prim1
 926                                  */
 927                                 if (!prim1 ||
 928                                     ifa1->ifa_mask != prim1->ifa_mask ||
 929                                     !inet_ifa_match(ifa1->ifa_address, prim1))
 930                                         prim1 = inet_ifa_byprefix(in_dev,
 931                                                         ifa1->ifa_address,
 932                                                         ifa1->ifa_mask);
 933                                 if (!prim1)
 934                                         continue;
 935                                 if (prim1->ifa_local != prim->ifa_local)
 936                                         continue;
 937                         }
 938                 } else {
 939                         if (prim->ifa_local != ifa1->ifa_local)
 940                                 continue;
 941                         prim1 = ifa1;
 942                         if (prim != prim1)
 943                                 same_prefsrc = 1;
 944                 }
 945                 if (ifa->ifa_local == ifa1->ifa_local)
 946                         ok |= LOCAL_OK;
 947                 if (ifa->ifa_broadcast == ifa1->ifa_broadcast)
 948                         ok |= BRD_OK;
 949                 if (brd == ifa1->ifa_broadcast)
 950                         ok |= BRD1_OK;
 951                 if (any == ifa1->ifa_broadcast)
 952                         ok |= BRD0_OK;
 953                 /* primary has network specific broadcasts */
 954                 if (prim1 == ifa1 && ifa1->ifa_prefixlen < 31) {
 955                         __be32 brd1 = ifa1->ifa_address | ~ifa1->ifa_mask;
 956                         __be32 any1 = ifa1->ifa_address & ifa1->ifa_mask;
 957
 958                         if (!ipv4_is_zeronet(any1)) {
 959                                 if (ifa->ifa_broadcast == brd1 ||
 960                                     ifa->ifa_broadcast == any1)
 961                                         ok |= BRD_OK;
 962                                 if (brd == brd1 || brd == any1)
 963                                         ok |= BRD1_OK;
 964                                 if (any == brd1 || any == any1)
 965                                         ok |= BRD0_OK;
 966                         }
 967                 }
 968         }
 969
 970         if (!(ok & BRD_OK))
 971                 fib_magic(RTM_DELROUTE, RTN_BROADCAST, ifa->ifa_broadcast, 32, prim);
 972         if (subnet && ifa->ifa_prefixlen < 31) {
 973                 if (!(ok & BRD1_OK))
 974                         fib_magic(RTM_DELROUTE, RTN_BROADCAST, brd, 32, prim);
 975                 if (!(ok & BRD0_OK))
 976                         fib_magic(RTM_DELROUTE, RTN_BROADCAST, any, 32, prim);
 977         }
 978         if (!(ok & LOCAL_OK)) {
 979                 fib_magic(RTM_DELROUTE, RTN_LOCAL, ifa->ifa_local, 32, prim);
 980
 981                 /* Check, that this local address finally disappeared. */
 982                 if (gone &&
 983                     inet_addr_type(dev_net(dev), ifa->ifa_local) != RTN_LOCAL) {
 984                         /* And the last, but not the least thing.
 985                          * We must flush stray FIB entries.
 986                          *
 987                          * First of all, we scan fib_info list searching
 988                          * for stray nexthop entries, then ignite fib_flush.
 989                          */
 990                         if (fib_sync_down_addr(dev_net(dev), ifa->ifa_local))
 991                                 fib_flush(dev_net(dev));
 992                 }
 993         }
 994 #undef LOCAL_OK
 995 #undef BRD_OK
 996 #undef BRD0_OK
 997 #undef BRD1_OK
 998 }
 999
1000 static void nl_fib_lookup(struct net *net, struct fib_result_nl *frn)
1001 {
1002
1003         struct fib_result       res;
1004         struct flowi4           fl4 = {
1005                 .flowi4_mark = frn->fl_mark,
1006                 .daddr = frn->fl_addr,
1007                 .flowi4_tos = frn->fl_tos,
1008                 .flowi4_scope = frn->fl_scope,
1009         };
1010         struct fib_table *tb;
1011
1012         rcu_read_lock();
1013
1014         tb = fib_get_table(net, frn->tb_id_in);
1015
1016         frn->err = -ENOENT;
1017         if (tb) {
1018                 local_bh_disable();
1019
1020                 frn->tb_id = tb->tb_id;
1021                 frn->err = fib_table_lookup(tb, &fl4, &res, FIB_LOOKUP_NOREF);
1022
1023                 if (!frn->err) {
1024                         frn->prefixlen = res.prefixlen;
1025                         frn->nh_sel = res.nh_sel;
1026                         frn->type = res.type;
1027                         frn->scope = res.scope;
1028                 }
1029                 local_bh_enable();
1030         }
1031
1032         rcu_read_unlock();
1033 }
1034
1035 static void nl_fib_input(struct sk_buff *skb)
1036 {
1037         struct net *net;
1038         struct fib_result_nl *frn;
1039         struct nlmsghdr *nlh;
1040         u32 portid;
1041
1042         net = sock_net(skb->sk);
1043         nlh = nlmsg_hdr(skb);
1044         if (skb->len < NLMSG_HDRLEN || skb->len < nlh->nlmsg_len ||
1045             nlmsg_len(nlh) < sizeof(*frn))
1046                 return;
1047
1048         skb = netlink_skb_clone(skb, GFP_KERNEL);
1049         if (!skb)
1050                 return;
1051         nlh = nlmsg_hdr(skb);
1052
1053         frn = (struct fib_result_nl *) nlmsg_data(nlh);
1054         nl_fib_lookup(net, frn);
1055
1056         portid = NETLINK_CB(skb).portid;      /* netlink portid */
1057         NETLINK_CB(skb).portid = 0;        /* from kernel */
1058         NETLINK_CB(skb).dst_group = 0;  /* unicast */
1059         netlink_unicast(net->ipv4.fibnl, skb, portid, MSG_DONTWAIT);
1060 }
1061
1062 static int __net_init nl_fib_lookup_init(struct net *net)
1063 {
1064         struct sock *sk;
1065         struct netlink_kernel_cfg cfg = {
1066                 .input  = nl_fib_input,
1067         };
1068
1069         sk = netlink_kernel_create(net, NETLINK_FIB_LOOKUP, &cfg);
1070         if (!sk)
1071                 return -EAFNOSUPPORT;
1072         net->ipv4.fibnl = sk;
1073         return 0;
1074 }
1075
1076 static void nl_fib_lookup_exit(struct net *net)
1077 {
1078         netlink_kernel_release(net->ipv4.fibnl);
1079         net->ipv4.fibnl = NULL;
1080 }
1081
1082 static void fib_disable_ip(struct net_device *dev, unsigned long event)
1083 {
1084         if (fib_sync_down_dev(dev, event))
1085                 fib_flush(dev_net(dev));
1086         rt_cache_flush(dev_net(dev));
1087         arp_ifdown(dev);
1088 }
1089
1090 static int fib_inetaddr_event(struct notifier_block *this, unsigned long event, void *ptr)
1091 {
1092         struct in_ifaddr *ifa = (struct in_ifaddr *)ptr;
1093         struct net_device *dev = ifa->ifa_dev->dev;
1094         struct net *net = dev_net(dev);
1095
1096         switch (event) {
1097         case NETDEV_UP:
1098                 fib_add_ifaddr(ifa);
1099 #ifdef CONFIG_IP_ROUTE_MULTIPATH
1100                 fib_sync_up(dev, RTNH_F_DEAD);
1101 #endif
1102                 atomic_inc(&net->ipv4.dev_addr_genid);
1103                 rt_cache_flush(dev_net(dev));
1104                 break;
1105         case NETDEV_DOWN:
1106                 fib_del_ifaddr(ifa, NULL);
1107                 atomic_inc(&net->ipv4.dev_addr_genid);
1108                 if (!ifa->ifa_dev->ifa_list) {
1109                         /* Last address was deleted from this interface.
1110                          * Disable IP.
1111                          */
1112                         fib_disable_ip(dev, event);
1113                 } else {
1114                         rt_cache_flush(dev_net(dev));
1115                 }
1116                 break;
1117         }
1118         return NOTIFY_DONE;
1119 }
1120
1121 static int fib_netdev_event(struct notifier_block *this, unsigned long event, void *ptr)
1122 {
1123         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1124         struct in_device *in_dev;
1125         struct net *net = dev_net(dev);
1126         unsigned int flags;
1127
1128         if (event == NETDEV_UNREGISTER) {
1129                 fib_disable_ip(dev, event);
1130                 rt_flush_dev(dev);
1131                 return NOTIFY_DONE;
1132         }
1133
1134         in_dev = __in_dev_get_rtnl(dev);
1135         if (!in_dev)
1136                 return NOTIFY_DONE;
1137
1138         switch (event) {
1139         case NETDEV_UP:
1140                 for_ifa(in_dev) {
1141                         fib_add_ifaddr(ifa);
1142                 } endfor_ifa(in_dev);
1143 #ifdef CONFIG_IP_ROUTE_MULTIPATH
1144                 fib_sync_up(dev, RTNH_F_DEAD);
1145 #endif
1146                 atomic_inc(&net->ipv4.dev_addr_genid);
1147                 rt_cache_flush(net);
1148                 break;
1149         case NETDEV_DOWN:
1150                 fib_disable_ip(dev, event);
1151                 break;
1152         case NETDEV_CHANGE:
1153                 flags = dev_get_flags(dev);
1154                 if (flags & (IFF_RUNNING | IFF_LOWER_UP))
1155                         fib_sync_up(dev, RTNH_F_LINKDOWN);
1156                 else
1157                         fib_sync_down_dev(dev, event);
1158                 /* fall through */
1159         case NETDEV_CHANGEMTU:
1160                 rt_cache_flush(net);
1161                 break;
1162         }
1163         return NOTIFY_DONE;
1164 }
1165
1166 static struct notifier_block fib_inetaddr_notifier = {
1167         .notifier_call = fib_inetaddr_event,
1168 };
1169
1170 static struct notifier_block fib_netdev_notifier = {
1171         .notifier_call = fib_netdev_event,
1172 };
1173
1174 static int __net_init ip_fib_net_init(struct net *net)
1175 {
1176         int err;
1177         size_t size = sizeof(struct hlist_head) * FIB_TABLE_HASHSZ;
1178
1179         /* Avoid false sharing : Use at least a full cache line */
1180         size = max_t(size_t, size, L1_CACHE_BYTES);
1181
1182         net->ipv4.fib_table_hash = kzalloc(size, GFP_KERNEL);
1183         if (!net->ipv4.fib_table_hash)
1184                 return -ENOMEM;
1185
1186         err = fib4_rules_init(net);
1187         if (err < 0)
1188                 goto fail;
1189         return 0;
1190
1191 fail:
1192         kfree(net->ipv4.fib_table_hash);
1193         return err;
1194 }
1195
1196 static void ip_fib_net_exit(struct net *net)
1197 {
1198         unsigned int i;
1199
1200         rtnl_lock();
1201 #ifdef CONFIG_IP_MULTIPLE_TABLES
1202         RCU_INIT_POINTER(net->ipv4.fib_local, NULL);
1203         RCU_INIT_POINTER(net->ipv4.fib_main, NULL);
1204         RCU_INIT_POINTER(net->ipv4.fib_default, NULL);
1205 #endif
1206         for (i = 0; i < FIB_TABLE_HASHSZ; i++) {
1207                 struct hlist_head *head = &net->ipv4.fib_table_hash[i];
1208                 struct hlist_node *tmp;
1209                 struct fib_table *tb;
1210
1211                 hlist_for_each_entry_safe(tb, tmp, head, tb_hlist) {
1212                         hlist_del(&tb->tb_hlist);
1213                         fib_table_flush(tb);
1214                         fib_free_table(tb);
1215                 }
1216         }
1217
1218 #ifdef CONFIG_IP_MULTIPLE_TABLES
1219         fib4_rules_exit(net);
1220 #endif
1221         rtnl_unlock();
1222         kfree(net->ipv4.fib_table_hash);
1223 }
1224
1225 static int __net_init fib_net_init(struct net *net)
1226 {
1227         int error;
1228
1229 #ifdef CONFIG_IP_ROUTE_CLASSID
1230         net->ipv4.fib_num_tclassid_users = 0;
1231 #endif
1232         error = ip_fib_net_init(net);
1233         if (error < 0)
1234                 goto out;
1235         error = nl_fib_lookup_init(net);
1236         if (error < 0)
1237                 goto out_nlfl;
1238         error = fib_proc_init(net);
1239         if (error < 0)
1240                 goto out_proc;
1241 out:
1242         return error;
1243
1244 out_proc:
1245         nl_fib_lookup_exit(net);
1246 out_nlfl:
1247         ip_fib_net_exit(net);
1248         goto out;
1249 }
1250
1251 static void __net_exit fib_net_exit(struct net *net)
1252 {
1253         fib_proc_exit(net);
1254         nl_fib_lookup_exit(net);
1255         ip_fib_net_exit(net);
1256 }
1257
1258 static struct pernet_operations fib_net_ops = {
1259         .init = fib_net_init,
1260         .exit = fib_net_exit,
1261 };
1262
1263 void __init ip_fib_init(void)
1264 {
1265         rtnl_register(PF_INET, RTM_NEWROUTE, inet_rtm_newroute, NULL, NULL);
1266         rtnl_register(PF_INET, RTM_DELROUTE, inet_rtm_delroute, NULL, NULL);
1267         rtnl_register(PF_INET, RTM_GETROUTE, NULL, inet_dump_fib, NULL);
1268
1269         register_pernet_subsys(&fib_net_ops);
1270         register_netdevice_notifier(&fib_netdev_notifier);
1271         register_inetaddr_notifier(&fib_inetaddr_notifier);
1272
1273         fib_trie_init();
1274 }