net/ipv6/route.c

   1 /*
   2  *      Linux INET6 implementation
   3  *      FIB front-end.
   4  *
   5  *      Authors:
   6  *      Pedro Roque             <roque@di.fc.ul.pt>
   7  *
   8  *      This program is free software; you can redistribute it and/or
   9  *      modify it under the terms of the GNU General Public License
  10  *      as published by the Free Software Foundation; either version
  11  *      2 of the License, or (at your option) any later version.
  12  */
  13
  14 /*      Changes:
  15  *
  16  *      YOSHIFUJI Hideaki @USAGI
  17  *              reworked default router selection.
  18  *              - respect outgoing interface
  19  *              - select from (probably) reachable routers (i.e.
  20  *              routers in REACHABLE, STALE, DELAY or PROBE states).
  21  *              - always select the same router if it is (probably)
  22  *              reachable.  otherwise, round-robin the list.
  23  *      Ville Nuorvala
  24  *              Fixed routing subtrees.
  25  */
  26
  27 #include <linux/capability.h>
  28 #include <linux/errno.h>
  29 #include <linux/types.h>
  30 #include <linux/times.h>
  31 #include <linux/socket.h>
  32 #include <linux/sockios.h>
  33 #include <linux/net.h>
  34 #include <linux/route.h>
  35 #include <linux/netdevice.h>
  36 #include <linux/in6.h>
  37 #include <linux/mroute6.h>
  38 #include <linux/init.h>
  39 #include <linux/if_arp.h>
  40 #include <linux/proc_fs.h>
  41 #include <linux/seq_file.h>
  42 #include <linux/nsproxy.h>
  43 #include <linux/slab.h>
  44 #include <net/net_namespace.h>
  45 #include <net/snmp.h>
  46 #include <net/ipv6.h>
  47 #include <net/ip6_fib.h>
  48 #include <net/ip6_route.h>
  49 #include <net/ndisc.h>
  50 #include <net/addrconf.h>
  51 #include <net/tcp.h>
  52 #include <linux/rtnetlink.h>
  53 #include <net/dst.h>
  54 #include <net/xfrm.h>
  55 #include <net/netevent.h>
  56 #include <net/netlink.h>
  57
  58 #include <asm/uaccess.h>
  59
  60 #ifdef CONFIG_SYSCTL
  61 #include <linux/sysctl.h>
  62 #endif
  63
  64 /* Set to 3 to get tracing. */
  65 #define RT6_DEBUG 2
  66
  67 #if RT6_DEBUG >= 3
  68 #define RDBG(x) printk x
  69 #define RT6_TRACE(x...) printk(KERN_DEBUG x)
  70 #else
  71 #define RDBG(x)
  72 #define RT6_TRACE(x...) do { ; } while (0)
  73 #endif
  74
  75 #define CLONE_OFFLINK_ROUTE 0
  76
  77 static struct rt6_info * ip6_rt_copy(struct rt6_info *ort);
  78 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie);
  79 static struct dst_entry *ip6_negative_advice(struct dst_entry *);
  80 static void             ip6_dst_destroy(struct dst_entry *);
  81 static void             ip6_dst_ifdown(struct dst_entry *,
  82                                        struct net_device *dev, int how);
  83 static int               ip6_dst_gc(struct dst_ops *ops);
  84
  85 static int              ip6_pkt_discard(struct sk_buff *skb);
  86 static int              ip6_pkt_discard_out(struct sk_buff *skb);
  87 static void             ip6_link_failure(struct sk_buff *skb);
  88 static void             ip6_rt_update_pmtu(struct dst_entry *dst, u32 mtu);
  89
  90 #ifdef CONFIG_IPV6_ROUTE_INFO
  91 static struct rt6_info *rt6_add_route_info(struct net *net,
  92                                            struct in6_addr *prefix, int prefixlen,
  93                                            struct in6_addr *gwaddr, int ifindex,
  94                                            unsigned pref);
  95 static struct rt6_info *rt6_get_route_info(struct net *net,
  96                                            struct in6_addr *prefix, int prefixlen,
  97                                            struct in6_addr *gwaddr, int ifindex);
  98 #endif
  99
 100 static struct dst_ops ip6_dst_ops_template = {
 101         .family                 =       AF_INET6,
 102         .protocol               =       cpu_to_be16(ETH_P_IPV6),
 103         .gc                     =       ip6_dst_gc,
 104         .gc_thresh              =       1024,
 105         .check                  =       ip6_dst_check,
 106         .destroy                =       ip6_dst_destroy,
 107         .ifdown                 =       ip6_dst_ifdown,
 108         .negative_advice        =       ip6_negative_advice,
 109         .link_failure           =       ip6_link_failure,
 110         .update_pmtu            =       ip6_rt_update_pmtu,
 111         .local_out              =       __ip6_local_out,
 112         .entries                =       ATOMIC_INIT(0),
 113 };
 114
 115 static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, u32 mtu)
 116 {
 117 }
 118
 119 static struct dst_ops ip6_dst_blackhole_ops = {
 120         .family                 =       AF_INET6,
 121         .protocol               =       cpu_to_be16(ETH_P_IPV6),
 122         .destroy                =       ip6_dst_destroy,
 123         .check                  =       ip6_dst_check,
 124         .update_pmtu            =       ip6_rt_blackhole_update_pmtu,
 125         .entries                =       ATOMIC_INIT(0),
 126 };
 127
 128 static struct rt6_info ip6_null_entry_template = {
 129         .u = {
 130                 .dst = {
 131                         .__refcnt       = ATOMIC_INIT(1),
 132                         .__use          = 1,
 133                         .obsolete       = -1,
 134                         .error          = -ENETUNREACH,
 135                         .metrics        = { [RTAX_HOPLIMIT - 1] = 255, },
 136                         .input          = ip6_pkt_discard,
 137                         .output         = ip6_pkt_discard_out,
 138                 }
 139         },
 140         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
 141         .rt6i_protocol  = RTPROT_KERNEL,
 142         .rt6i_metric    = ~(u32) 0,
 143         .rt6i_ref       = ATOMIC_INIT(1),
 144 };
 145
 146 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
 147
 148 static int ip6_pkt_prohibit(struct sk_buff *skb);
 149 static int ip6_pkt_prohibit_out(struct sk_buff *skb);
 150
 151 static struct rt6_info ip6_prohibit_entry_template = {
 152         .u = {
 153                 .dst = {
 154                         .__refcnt       = ATOMIC_INIT(1),
 155                         .__use          = 1,
 156                         .obsolete       = -1,
 157                         .error          = -EACCES,
 158                         .metrics        = { [RTAX_HOPLIMIT - 1] = 255, },
 159                         .input          = ip6_pkt_prohibit,
 160                         .output         = ip6_pkt_prohibit_out,
 161                 }
 162         },
 163         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
 164         .rt6i_protocol  = RTPROT_KERNEL,
 165         .rt6i_metric    = ~(u32) 0,
 166         .rt6i_ref       = ATOMIC_INIT(1),
 167 };
 168
 169 static struct rt6_info ip6_blk_hole_entry_template = {
 170         .u = {
 171                 .dst = {
 172                         .__refcnt       = ATOMIC_INIT(1),
 173                         .__use          = 1,
 174                         .obsolete       = -1,
 175                         .error          = -EINVAL,
 176                         .metrics        = { [RTAX_HOPLIMIT - 1] = 255, },
 177                         .input          = dst_discard,
 178                         .output         = dst_discard,
 179                 }
 180         },
 181         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
 182         .rt6i_protocol  = RTPROT_KERNEL,
 183         .rt6i_metric    = ~(u32) 0,
 184         .rt6i_ref       = ATOMIC_INIT(1),
 185 };
 186
 187 #endif
 188
 189 /* allocate dst with ip6_dst_ops */
 190 static inline struct rt6_info *ip6_dst_alloc(struct dst_ops *ops)
 191 {
 192         return (struct rt6_info *)dst_alloc(ops);
 193 }
 194
 195 static void ip6_dst_destroy(struct dst_entry *dst)
 196 {
 197         struct rt6_info *rt = (struct rt6_info *)dst;
 198         struct inet6_dev *idev = rt->rt6i_idev;
 199
 200         if (idev != NULL) {
 201                 rt->rt6i_idev = NULL;
 202                 in6_dev_put(idev);
 203         }
 204 }
 205
 206 static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
 207                            int how)
 208 {
 209         struct rt6_info *rt = (struct rt6_info *)dst;
 210         struct inet6_dev *idev = rt->rt6i_idev;
 211         struct net_device *loopback_dev =
 212                 dev_net(dev)->loopback_dev;
 213
 214         if (dev != loopback_dev && idev != NULL && idev->dev == dev) {
 215                 struct inet6_dev *loopback_idev =
 216                         in6_dev_get(loopback_dev);
 217                 if (loopback_idev != NULL) {
 218                         rt->rt6i_idev = loopback_idev;
 219                         in6_dev_put(idev);
 220                 }
 221         }
 222 }
 223
 224 static __inline__ int rt6_check_expired(const struct rt6_info *rt)
 225 {
 226         return (rt->rt6i_flags & RTF_EXPIRES &&
 227                 time_after(jiffies, rt->rt6i_expires));
 228 }
 229
 230 static inline int rt6_need_strict(struct in6_addr *daddr)
 231 {
 232         return (ipv6_addr_type(daddr) &
 233                 (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK));
 234 }
 235
 236 /*
 237  *      Route lookup. Any table->tb6_lock is implied.
 238  */
 239
 240 static inline struct rt6_info *rt6_device_match(struct net *net,
 241                                                     struct rt6_info *rt,
 242                                                     struct in6_addr *saddr,
 243                                                     int oif,
 244                                                     int flags)
 245 {
 246         struct rt6_info *local = NULL;
 247         struct rt6_info *sprt;
 248
 249         if (!oif && ipv6_addr_any(saddr))
 250                 goto out;
 251
 252         for (sprt = rt; sprt; sprt = sprt->u.dst.rt6_next) {
 253                 struct net_device *dev = sprt->rt6i_dev;
 254
 255                 if (oif) {
 256                         if (dev->ifindex == oif)
 257                                 return sprt;
 258                         if (dev->flags & IFF_LOOPBACK) {
 259                                 if (sprt->rt6i_idev == NULL ||
 260                                     sprt->rt6i_idev->dev->ifindex != oif) {
 261                                         if (flags & RT6_LOOKUP_F_IFACE && oif)
 262                                                 continue;
 263                                         if (local && (!oif ||
 264                                                       local->rt6i_idev->dev->ifindex == oif))
 265                                                 continue;
 266                                 }
 267                                 local = sprt;
 268                         }
 269                 } else {
 270                         if (ipv6_chk_addr(net, saddr, dev,
 271                                           flags & RT6_LOOKUP_F_IFACE))
 272                                 return sprt;
 273                 }
 274         }
 275
 276         if (oif) {
 277                 if (local)
 278                         return local;
 279
 280                 if (flags & RT6_LOOKUP_F_IFACE)
 281                         return net->ipv6.ip6_null_entry;
 282         }
 283 out:
 284         return rt;
 285 }
 286
 287 #ifdef CONFIG_IPV6_ROUTER_PREF
 288 static void rt6_probe(struct rt6_info *rt)
 289 {
 290         struct neighbour *neigh = rt ? rt->rt6i_nexthop : NULL;
 291         /*
 292          * Okay, this does not seem to be appropriate
 293          * for now, however, we need to check if it
 294          * is really so; aka Router Reachability Probing.
 295          *
 296          * Router Reachability Probe MUST be rate-limited
 297          * to no more than one per minute.
 298          */
 299         if (!neigh || (neigh->nud_state & NUD_VALID))
 300                 return;
 301         read_lock_bh(&neigh->lock);
 302         if (!(neigh->nud_state & NUD_VALID) &&
 303             time_after(jiffies, neigh->updated + rt->rt6i_idev->cnf.rtr_probe_interval)) {
 304                 struct in6_addr mcaddr;
 305                 struct in6_addr *target;
 306
 307                 neigh->updated = jiffies;
 308                 read_unlock_bh(&neigh->lock);
 309
 310                 target = (struct in6_addr *)&neigh->primary_key;
 311                 addrconf_addr_solict_mult(target, &mcaddr);
 312                 ndisc_send_ns(rt->rt6i_dev, NULL, target, &mcaddr, NULL);
 313         } else
 314                 read_unlock_bh(&neigh->lock);
 315 }
 316 #else
 317 static inline void rt6_probe(struct rt6_info *rt)
 318 {
 319         return;
 320 }
 321 #endif
 322
 323 /*
 324  * Default Router Selection (RFC 2461 6.3.6)
 325  */
 326 static inline int rt6_check_dev(struct rt6_info *rt, int oif)
 327 {
 328         struct net_device *dev = rt->rt6i_dev;
 329         if (!oif || dev->ifindex == oif)
 330                 return 2;
 331         if ((dev->flags & IFF_LOOPBACK) &&
 332             rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
 333                 return 1;
 334         return 0;
 335 }
 336
 337 static inline int rt6_check_neigh(struct rt6_info *rt)
 338 {
 339         struct neighbour *neigh = rt->rt6i_nexthop;
 340         int m;
 341         if (rt->rt6i_flags & RTF_NONEXTHOP ||
 342             !(rt->rt6i_flags & RTF_GATEWAY))
 343                 m = 1;
 344         else if (neigh) {
 345                 read_lock_bh(&neigh->lock);
 346                 if (neigh->nud_state & NUD_VALID)
 347                         m = 2;
 348 #ifdef CONFIG_IPV6_ROUTER_PREF
 349                 else if (neigh->nud_state & NUD_FAILED)
 350                         m = 0;
 351 #endif
 352                 else
 353                         m = 1;
 354                 read_unlock_bh(&neigh->lock);
 355         } else
 356                 m = 0;
 357         return m;
 358 }
 359
 360 static int rt6_score_route(struct rt6_info *rt, int oif,
 361                            int strict)
 362 {
 363         int m, n;
 364
 365         m = rt6_check_dev(rt, oif);
 366         if (!m && (strict & RT6_LOOKUP_F_IFACE))
 367                 return -1;
 368 #ifdef CONFIG_IPV6_ROUTER_PREF
 369         m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
 370 #endif
 371         n = rt6_check_neigh(rt);
 372         if (!n && (strict & RT6_LOOKUP_F_REACHABLE))
 373                 return -1;
 374         return m;
 375 }
 376
 377 static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
 378                                    int *mpri, struct rt6_info *match)
 379 {
 380         int m;
 381
 382         if (rt6_check_expired(rt))
 383                 goto out;
 384
 385         m = rt6_score_route(rt, oif, strict);
 386         if (m < 0)
 387                 goto out;
 388
 389         if (m > *mpri) {
 390                 if (strict & RT6_LOOKUP_F_REACHABLE)
 391                         rt6_probe(match);
 392                 *mpri = m;
 393                 match = rt;
 394         } else if (strict & RT6_LOOKUP_F_REACHABLE) {
 395                 rt6_probe(rt);
 396         }
 397
 398 out:
 399         return match;
 400 }
 401
 402 static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
 403                                      struct rt6_info *rr_head,
 404                                      u32 metric, int oif, int strict)
 405 {
 406         struct rt6_info *rt, *match;
 407         int mpri = -1;
 408
 409         match = NULL;
 410         for (rt = rr_head; rt && rt->rt6i_metric == metric;
 411              rt = rt->u.dst.rt6_next)
 412                 match = find_match(rt, oif, strict, &mpri, match);
 413         for (rt = fn->leaf; rt && rt != rr_head && rt->rt6i_metric == metric;
 414              rt = rt->u.dst.rt6_next)
 415                 match = find_match(rt, oif, strict, &mpri, match);
 416
 417         return match;
 418 }
 419
 420 static struct rt6_info *rt6_select(struct fib6_node *fn, int oif, int strict)
 421 {
 422         struct rt6_info *match, *rt0;
 423         struct net *net;
 424
 425         RT6_TRACE("%s(fn->leaf=%p, oif=%d)\n",
 426                   __func__, fn->leaf, oif);
 427
 428         rt0 = fn->rr_ptr;
 429         if (!rt0)
 430                 fn->rr_ptr = rt0 = fn->leaf;
 431
 432         match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict);
 433
 434         if (!match &&
 435             (strict & RT6_LOOKUP_F_REACHABLE)) {
 436                 struct rt6_info *next = rt0->u.dst.rt6_next;
 437
 438                 /* no entries matched; do round-robin */
 439                 if (!next || next->rt6i_metric != rt0->rt6i_metric)
 440                         next = fn->leaf;
 441
 442                 if (next != rt0)
 443                         fn->rr_ptr = next;
 444         }
 445
 446         RT6_TRACE("%s() => %p\n",
 447                   __func__, match);
 448
 449         net = dev_net(rt0->rt6i_dev);
 450         return (match ? match : net->ipv6.ip6_null_entry);
 451 }
 452
 453 #ifdef CONFIG_IPV6_ROUTE_INFO
 454 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
 455                   struct in6_addr *gwaddr)
 456 {
 457         struct net *net = dev_net(dev);
 458         struct route_info *rinfo = (struct route_info *) opt;
 459         struct in6_addr prefix_buf, *prefix;
 460         unsigned int pref;
 461         unsigned long lifetime;
 462         struct rt6_info *rt;
 463
 464         if (len < sizeof(struct route_info)) {
 465                 return -EINVAL;
 466         }
 467
 468         /* Sanity check for prefix_len and length */
 469         if (rinfo->length > 3) {
 470                 return -EINVAL;
 471         } else if (rinfo->prefix_len > 128) {
 472                 return -EINVAL;
 473         } else if (rinfo->prefix_len > 64) {
 474                 if (rinfo->length < 2) {
 475                         return -EINVAL;
 476                 }
 477         } else if (rinfo->prefix_len > 0) {
 478                 if (rinfo->length < 1) {
 479                         return -EINVAL;
 480                 }
 481         }
 482
 483         pref = rinfo->route_pref;
 484         if (pref == ICMPV6_ROUTER_PREF_INVALID)
 485                 return -EINVAL;
 486
 487         lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
 488
 489         if (rinfo->length == 3)
 490                 prefix = (struct in6_addr *)rinfo->prefix;
 491         else {
 492                 /* this function is safe */
 493                 ipv6_addr_prefix(&prefix_buf,
 494                                  (struct in6_addr *)rinfo->prefix,
 495                                  rinfo->prefix_len);
 496                 prefix = &prefix_buf;
 497         }
 498
 499         rt = rt6_get_route_info(net, prefix, rinfo->prefix_len, gwaddr,
 500                                 dev->ifindex);
 501
 502         if (rt && !lifetime) {
 503                 ip6_del_rt(rt);
 504                 rt = NULL;
 505         }
 506
 507         if (!rt && lifetime)
 508                 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr, dev->ifindex,
 509                                         pref);
 510         else if (rt)
 511                 rt->rt6i_flags = RTF_ROUTEINFO |
 512                                  (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
 513
 514         if (rt) {
 515                 if (!addrconf_finite_timeout(lifetime)) {
 516                         rt->rt6i_flags &= ~RTF_EXPIRES;
 517                 } else {
 518                         rt->rt6i_expires = jiffies + HZ * lifetime;
 519                         rt->rt6i_flags |= RTF_EXPIRES;
 520                 }
 521                 dst_release(&rt->u.dst);
 522         }
 523         return 0;
 524 }
 525 #endif
 526
 527 #define BACKTRACK(__net, saddr)                 \
 528 do { \
 529         if (rt == __net->ipv6.ip6_null_entry) { \
 530                 struct fib6_node *pn; \
 531                 while (1) { \
 532                         if (fn->fn_flags & RTN_TL_ROOT) \
 533                                 goto out; \
 534                         pn = fn->parent; \
 535                         if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn) \
 536                                 fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr); \
 537                         else \
 538                                 fn = pn; \
 539                         if (fn->fn_flags & RTN_RTINFO) \
 540                                 goto restart; \
 541                 } \
 542         } \
 543 } while(0)
 544
 545 static struct rt6_info *ip6_pol_route_lookup(struct net *net,
 546                                              struct fib6_table *table,
 547                                              struct flowi *fl, int flags)
 548 {
 549         struct fib6_node *fn;
 550         struct rt6_info *rt;
 551
 552         read_lock_bh(&table->tb6_lock);
 553         fn = fib6_lookup(&table->tb6_root, &fl->fl6_dst, &fl->fl6_src);
 554 restart:
 555         rt = fn->leaf;
 556         rt = rt6_device_match(net, rt, &fl->fl6_src, fl->oif, flags);
 557         BACKTRACK(net, &fl->fl6_src);
 558 out:
 559         dst_use(&rt->u.dst, jiffies);
 560         read_unlock_bh(&table->tb6_lock);
 561         return rt;
 562
 563 }
 564
 565 struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
 566                             const struct in6_addr *saddr, int oif, int strict)
 567 {
 568         struct flowi fl = {
 569                 .oif = oif,
 570                 .nl_u = {
 571                         .ip6_u = {
 572                                 .daddr = *daddr,
 573                         },
 574                 },
 575         };
 576         struct dst_entry *dst;
 577         int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
 578
 579         if (saddr) {
 580                 memcpy(&fl.fl6_src, saddr, sizeof(*saddr));
 581                 flags |= RT6_LOOKUP_F_HAS_SADDR;
 582         }
 583
 584         dst = fib6_rule_lookup(net, &fl, flags, ip6_pol_route_lookup);
 585         if (dst->error == 0)
 586                 return (struct rt6_info *) dst;
 587
 588         dst_release(dst);
 589
 590         return NULL;
 591 }
 592
 593 EXPORT_SYMBOL(rt6_lookup);
 594
 595 /* ip6_ins_rt is called with FREE table->tb6_lock.
 596    It takes new route entry, the addition fails by any reason the
 597    route is freed. In any case, if caller does not hold it, it may
 598    be destroyed.
 599  */
 600
 601 static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info)
 602 {
 603         int err;
 604         struct fib6_table *table;
 605
 606         table = rt->rt6i_table;
 607         write_lock_bh(&table->tb6_lock);
 608         err = fib6_add(&table->tb6_root, rt, info);
 609         write_unlock_bh(&table->tb6_lock);
 610
 611         return err;
 612 }
 613
 614 int ip6_ins_rt(struct rt6_info *rt)
 615 {
 616         struct nl_info info = {
 617                 .nl_net = dev_net(rt->rt6i_dev),
 618         };
 619         return __ip6_ins_rt(rt, &info);
 620 }
 621
 622 static struct rt6_info *rt6_alloc_cow(struct rt6_info *ort, struct in6_addr *daddr,
 623                                       struct in6_addr *saddr)
 624 {
 625         struct rt6_info *rt;
 626
 627         /*
 628          *      Clone the route.
 629          */
 630
 631         rt = ip6_rt_copy(ort);
 632
 633         if (rt) {
 634                 struct neighbour *neigh;
 635                 int attempts = !in_softirq();
 636
 637                 if (!(rt->rt6i_flags&RTF_GATEWAY)) {
 638                         if (rt->rt6i_dst.plen != 128 &&
 639                             ipv6_addr_equal(&rt->rt6i_dst.addr, daddr))
 640                                 rt->rt6i_flags |= RTF_ANYCAST;
 641                         ipv6_addr_copy(&rt->rt6i_gateway, daddr);
 642                 }
 643
 644                 ipv6_addr_copy(&rt->rt6i_dst.addr, daddr);
 645                 rt->rt6i_dst.plen = 128;
 646                 rt->rt6i_flags |= RTF_CACHE;
 647                 rt->u.dst.flags |= DST_HOST;
 648
 649 #ifdef CONFIG_IPV6_SUBTREES
 650                 if (rt->rt6i_src.plen && saddr) {
 651                         ipv6_addr_copy(&rt->rt6i_src.addr, saddr);
 652                         rt->rt6i_src.plen = 128;
 653                 }
 654 #endif
 655
 656         retry:
 657                 neigh = ndisc_get_neigh(rt->rt6i_dev, &rt->rt6i_gateway);
 658                 if (IS_ERR(neigh)) {
 659                         struct net *net = dev_net(rt->rt6i_dev);
 660                         int saved_rt_min_interval =
 661                                 net->ipv6.sysctl.ip6_rt_gc_min_interval;
 662                         int saved_rt_elasticity =
 663                                 net->ipv6.sysctl.ip6_rt_gc_elasticity;
 664
 665                         if (attempts-- > 0) {
 666                                 net->ipv6.sysctl.ip6_rt_gc_elasticity = 1;
 667                                 net->ipv6.sysctl.ip6_rt_gc_min_interval = 0;
 668
 669                                 ip6_dst_gc(&net->ipv6.ip6_dst_ops);
 670
 671                                 net->ipv6.sysctl.ip6_rt_gc_elasticity =
 672                                         saved_rt_elasticity;
 673                                 net->ipv6.sysctl.ip6_rt_gc_min_interval =
 674                                         saved_rt_min_interval;
 675                                 goto retry;
 676                         }
 677
 678                         if (net_ratelimit())
 679                                 printk(KERN_WARNING
 680                                        "Neighbour table overflow.\n");
 681                         dst_free(&rt->u.dst);
 682                         return NULL;
 683                 }
 684                 rt->rt6i_nexthop = neigh;
 685
 686         }
 687
 688         return rt;
 689 }
 690
 691 static struct rt6_info *rt6_alloc_clone(struct rt6_info *ort, struct in6_addr *daddr)
 692 {
 693         struct rt6_info *rt = ip6_rt_copy(ort);
 694         if (rt) {
 695                 ipv6_addr_copy(&rt->rt6i_dst.addr, daddr);
 696                 rt->rt6i_dst.plen = 128;
 697                 rt->rt6i_flags |= RTF_CACHE;
 698                 rt->u.dst.flags |= DST_HOST;
 699                 rt->rt6i_nexthop = neigh_clone(ort->rt6i_nexthop);
 700         }
 701         return rt;
 702 }
 703
 704 static struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table, int oif,
 705                                       struct flowi *fl, int flags)
 706 {
 707         struct fib6_node *fn;
 708         struct rt6_info *rt, *nrt;
 709         int strict = 0;
 710         int attempts = 3;
 711         int err;
 712         int reachable = net->ipv6.devconf_all->forwarding ? 0 : RT6_LOOKUP_F_REACHABLE;
 713
 714         strict |= flags & RT6_LOOKUP_F_IFACE;
 715
 716 relookup:
 717         read_lock_bh(&table->tb6_lock);
 718
 719 restart_2:
 720         fn = fib6_lookup(&table->tb6_root, &fl->fl6_dst, &fl->fl6_src);
 721
 722 restart:
 723         rt = rt6_select(fn, oif, strict | reachable);
 724
 725         BACKTRACK(net, &fl->fl6_src);
 726         if (rt == net->ipv6.ip6_null_entry ||
 727             rt->rt6i_flags & RTF_CACHE)
 728                 goto out;
 729
 730         dst_hold(&rt->u.dst);
 731         read_unlock_bh(&table->tb6_lock);
 732
 733         if (!rt->rt6i_nexthop && !(rt->rt6i_flags & RTF_NONEXTHOP))
 734                 nrt = rt6_alloc_cow(rt, &fl->fl6_dst, &fl->fl6_src);
 735         else {
 736 #if CLONE_OFFLINK_ROUTE
 737                 nrt = rt6_alloc_clone(rt, &fl->fl6_dst);
 738 #else
 739                 goto out2;
 740 #endif
 741         }
 742
 743         dst_release(&rt->u.dst);
 744         rt = nrt ? : net->ipv6.ip6_null_entry;
 745
 746         dst_hold(&rt->u.dst);
 747         if (nrt) {
 748                 err = ip6_ins_rt(nrt);
 749                 if (!err)
 750                         goto out2;
 751         }
 752
 753         if (--attempts <= 0)
 754                 goto out2;
 755
 756         /*
 757          * Race condition! In the gap, when table->tb6_lock was
 758          * released someone could insert this route.  Relookup.
 759          */
 760         dst_release(&rt->u.dst);
 761         goto relookup;
 762
 763 out:
 764         if (reachable) {
 765                 reachable = 0;
 766                 goto restart_2;
 767         }
 768         dst_hold(&rt->u.dst);
 769         read_unlock_bh(&table->tb6_lock);
 770 out2:
 771         rt->u.dst.lastuse = jiffies;
 772         rt->u.dst.__use++;
 773
 774         return rt;
 775 }
 776
 777 static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
 778                                             struct flowi *fl, int flags)
 779 {
 780         return ip6_pol_route(net, table, fl->iif, fl, flags);
 781 }
 782
 783 void ip6_route_input(struct sk_buff *skb)
 784 {
 785         struct ipv6hdr *iph = ipv6_hdr(skb);
 786         struct net *net = dev_net(skb->dev);
 787         int flags = RT6_LOOKUP_F_HAS_SADDR;
 788         struct flowi fl = {
 789                 .iif = skb->dev->ifindex,
 790                 .nl_u = {
 791                         .ip6_u = {
 792                                 .daddr = iph->daddr,
 793                                 .saddr = iph->saddr,
 794                                 .flowlabel = (* (__be32 *) iph)&IPV6_FLOWINFO_MASK,
 795                         },
 796                 },
 797                 .mark = skb->mark,
 798                 .proto = iph->nexthdr,
 799         };
 800
 801         if (rt6_need_strict(&iph->daddr) && skb->dev->type != ARPHRD_PIMREG)
 802                 flags |= RT6_LOOKUP_F_IFACE;
 803
 804         skb_dst_set(skb, fib6_rule_lookup(net, &fl, flags, ip6_pol_route_input));
 805 }
 806
 807 static struct rt6_info *ip6_pol_route_output(struct net *net, struct fib6_table *table,
 808                                              struct flowi *fl, int flags)
 809 {
 810         return ip6_pol_route(net, table, fl->oif, fl, flags);
 811 }
 812
 813 struct dst_entry * ip6_route_output(struct net *net, struct sock *sk,
 814                                     struct flowi *fl)
 815 {
 816         int flags = 0;
 817
 818         if (fl->oif || rt6_need_strict(&fl->fl6_dst))
 819                 flags |= RT6_LOOKUP_F_IFACE;
 820
 821         if (!ipv6_addr_any(&fl->fl6_src))
 822                 flags |= RT6_LOOKUP_F_HAS_SADDR;
 823         else if (sk)
 824                 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
 825
 826         return fib6_rule_lookup(net, fl, flags, ip6_pol_route_output);
 827 }
 828
 829 EXPORT_SYMBOL(ip6_route_output);
 830
 831 int ip6_dst_blackhole(struct sock *sk, struct dst_entry **dstp, struct flowi *fl)
 832 {
 833         struct rt6_info *ort = (struct rt6_info *) *dstp;
 834         struct rt6_info *rt = (struct rt6_info *)
 835                 dst_alloc(&ip6_dst_blackhole_ops);
 836         struct dst_entry *new = NULL;
 837
 838         if (rt) {
 839                 new = &rt->u.dst;
 840
 841                 atomic_set(&new->__refcnt, 1);
 842                 new->__use = 1;
 843                 new->input = dst_discard;
 844                 new->output = dst_discard;
 845
 846                 memcpy(new->metrics, ort->u.dst.metrics, RTAX_MAX*sizeof(u32));
 847                 new->dev = ort->u.dst.dev;
 848                 if (new->dev)
 849                         dev_hold(new->dev);
 850                 rt->rt6i_idev = ort->rt6i_idev;
 851                 if (rt->rt6i_idev)
 852                         in6_dev_hold(rt->rt6i_idev);
 853                 rt->rt6i_expires = 0;
 854
 855                 ipv6_addr_copy(&rt->rt6i_gateway, &ort->rt6i_gateway);
 856                 rt->rt6i_flags = ort->rt6i_flags & ~RTF_EXPIRES;
 857                 rt->rt6i_metric = 0;
 858
 859                 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
 860 #ifdef CONFIG_IPV6_SUBTREES
 861                 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
 862 #endif
 863
 864                 dst_free(new);
 865         }
 866
 867         dst_release(*dstp);
 868         *dstp = new;
 869         return (new ? 0 : -ENOMEM);
 870 }
 871 EXPORT_SYMBOL_GPL(ip6_dst_blackhole);
 872
 873 /*
 874  *      Destination cache support functions
 875  */
 876
 877 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
 878 {
 879         struct rt6_info *rt;
 880
 881         rt = (struct rt6_info *) dst;
 882
 883         if (rt->rt6i_node && (rt->rt6i_node->fn_sernum == cookie))
 884                 return dst;
 885
 886         return NULL;
 887 }
 888
 889 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
 890 {
 891         struct rt6_info *rt = (struct rt6_info *) dst;
 892
 893         if (rt) {
 894                 if (rt->rt6i_flags & RTF_CACHE) {
 895                         if (rt6_check_expired(rt)) {
 896                                 ip6_del_rt(rt);
 897                                 dst = NULL;
 898                         }
 899                 } else {
 900                         dst_release(dst);
 901                         dst = NULL;
 902                 }
 903         }
 904         return dst;
 905 }
 906
 907 static void ip6_link_failure(struct sk_buff *skb)
 908 {
 909         struct rt6_info *rt;
 910
 911         icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
 912
 913         rt = (struct rt6_info *) skb_dst(skb);
 914         if (rt) {
 915                 if (rt->rt6i_flags&RTF_CACHE) {
 916                         dst_set_expires(&rt->u.dst, 0);
 917                         rt->rt6i_flags |= RTF_EXPIRES;
 918                 } else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT))
 919                         rt->rt6i_node->fn_sernum = -1;
 920         }
 921 }
 922
 923 static void ip6_rt_update_pmtu(struct dst_entry *dst, u32 mtu)
 924 {
 925         struct rt6_info *rt6 = (struct rt6_info*)dst;
 926
 927         if (mtu < dst_mtu(dst) && rt6->rt6i_dst.plen == 128) {
 928                 rt6->rt6i_flags |= RTF_MODIFIED;
 929                 if (mtu < IPV6_MIN_MTU) {
 930                         mtu = IPV6_MIN_MTU;
 931                         dst->metrics[RTAX_FEATURES-1] |= RTAX_FEATURE_ALLFRAG;
 932                 }
 933                 dst->metrics[RTAX_MTU-1] = mtu;
 934                 call_netevent_notifiers(NETEVENT_PMTU_UPDATE, dst);
 935         }
 936 }
 937
 938 static int ipv6_get_mtu(struct net_device *dev);
 939
 940 static inline unsigned int ipv6_advmss(struct net *net, unsigned int mtu)
 941 {
 942         mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
 943
 944         if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
 945                 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
 946
 947         /*
 948          * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
 949          * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
 950          * IPV6_MAXPLEN is also valid and means: "any MSS,
 951          * rely only on pmtu discovery"
 952          */
 953         if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
 954                 mtu = IPV6_MAXPLEN;
 955         return mtu;
 956 }
 957
 958 static struct dst_entry *icmp6_dst_gc_list;
 959 static DEFINE_SPINLOCK(icmp6_dst_lock);
 960
 961 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
 962                                   struct neighbour *neigh,
 963                                   const struct in6_addr *addr)
 964 {
 965         struct rt6_info *rt;
 966         struct inet6_dev *idev = in6_dev_get(dev);
 967         struct net *net = dev_net(dev);
 968
 969         if (unlikely(idev == NULL))
 970                 return NULL;
 971
 972         rt = ip6_dst_alloc(&net->ipv6.ip6_dst_ops);
 973         if (unlikely(rt == NULL)) {
 974                 in6_dev_put(idev);
 975                 goto out;
 976         }
 977
 978         dev_hold(dev);
 979         if (neigh)
 980                 neigh_hold(neigh);
 981         else {
 982                 neigh = ndisc_get_neigh(dev, addr);
 983                 if (IS_ERR(neigh))
 984                         neigh = NULL;
 985         }
 986
 987         rt->rt6i_dev      = dev;
 988         rt->rt6i_idev     = idev;
 989         rt->rt6i_nexthop  = neigh;
 990         atomic_set(&rt->u.dst.__refcnt, 1);
 991         rt->u.dst.metrics[RTAX_HOPLIMIT-1] = 255;
 992         rt->u.dst.metrics[RTAX_MTU-1] = ipv6_get_mtu(rt->rt6i_dev);
 993         rt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(net, dst_mtu(&rt->u.dst));
 994         rt->u.dst.output  = ip6_output;
 995
 996 #if 0   /* there's no chance to use these for ndisc */
 997         rt->u.dst.flags   = ipv6_addr_type(addr) & IPV6_ADDR_UNICAST
 998                                 ? DST_HOST
 999                                 : 0;
1000         ipv6_addr_copy(&rt->rt6i_dst.addr, addr);
1001         rt->rt6i_dst.plen = 128;
1002 #endif
1003
1004         spin_lock_bh(&icmp6_dst_lock);
1005         rt->u.dst.next = icmp6_dst_gc_list;
1006         icmp6_dst_gc_list = &rt->u.dst;
1007         spin_unlock_bh(&icmp6_dst_lock);
1008
1009         fib6_force_start_gc(net);
1010
1011 out:
1012         return &rt->u.dst;
1013 }
1014
1015 int icmp6_dst_gc(void)
1016 {
1017         struct dst_entry *dst, *next, **pprev;
1018         int more = 0;
1019
1020         next = NULL;
1021
1022         spin_lock_bh(&icmp6_dst_lock);
1023         pprev = &icmp6_dst_gc_list;
1024
1025         while ((dst = *pprev) != NULL) {
1026                 if (!atomic_read(&dst->__refcnt)) {
1027                         *pprev = dst->next;
1028                         dst_free(dst);
1029                 } else {
1030                         pprev = &dst->next;
1031                         ++more;
1032                 }
1033         }
1034
1035         spin_unlock_bh(&icmp6_dst_lock);
1036
1037         return more;
1038 }
1039
1040 static void icmp6_clean_all(int (*func)(struct rt6_info *rt, void *arg),
1041                             void *arg)
1042 {
1043         struct dst_entry *dst, **pprev;
1044
1045         spin_lock_bh(&icmp6_dst_lock);
1046         pprev = &icmp6_dst_gc_list;
1047         while ((dst = *pprev) != NULL) {
1048                 struct rt6_info *rt = (struct rt6_info *) dst;
1049                 if (func(rt, arg)) {
1050                         *pprev = dst->next;
1051                         dst_free(dst);
1052                 } else {
1053                         pprev = &dst->next;
1054                 }
1055         }
1056         spin_unlock_bh(&icmp6_dst_lock);
1057 }
1058
1059 static int ip6_dst_gc(struct dst_ops *ops)
1060 {
1061         unsigned long now = jiffies;
1062         struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
1063         int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1064         int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1065         int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1066         int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1067         unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1068
1069         if (time_after(rt_last_gc + rt_min_interval, now) &&
1070             atomic_read(&ops->entries) <= rt_max_size)
1071                 goto out;
1072
1073         net->ipv6.ip6_rt_gc_expire++;
1074         fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net);
1075         net->ipv6.ip6_rt_last_gc = now;
1076         if (atomic_read(&ops->entries) < ops->gc_thresh)
1077                 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1078 out:
1079         net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1080         return (atomic_read(&ops->entries) > rt_max_size);
1081 }
1082
1083 /* Clean host part of a prefix. Not necessary in radix tree,
1084    but results in cleaner routing tables.
1085
1086    Remove it only when all the things will work!
1087  */
1088
1089 static int ipv6_get_mtu(struct net_device *dev)
1090 {
1091         int mtu = IPV6_MIN_MTU;
1092         struct inet6_dev *idev;
1093
1094         idev = in6_dev_get(dev);
1095         if (idev) {
1096                 mtu = idev->cnf.mtu6;
1097                 in6_dev_put(idev);
1098         }
1099         return mtu;
1100 }
1101
1102 int ip6_dst_hoplimit(struct dst_entry *dst)
1103 {
1104         int hoplimit = dst_metric(dst, RTAX_HOPLIMIT);
1105         if (hoplimit < 0) {
1106                 struct net_device *dev = dst->dev;
1107                 struct inet6_dev *idev = in6_dev_get(dev);
1108                 if (idev) {
1109                         hoplimit = idev->cnf.hop_limit;
1110                         in6_dev_put(idev);
1111                 } else
1112                         hoplimit = dev_net(dev)->ipv6.devconf_all->hop_limit;
1113         }
1114         return hoplimit;
1115 }
1116
1117 /*
1118  *
1119  */
1120
1121 int ip6_route_add(struct fib6_config *cfg)
1122 {
1123         int err;
1124         struct net *net = cfg->fc_nlinfo.nl_net;
1125         struct rt6_info *rt = NULL;
1126         struct net_device *dev = NULL;
1127         struct inet6_dev *idev = NULL;
1128         struct fib6_table *table;
1129         int addr_type;
1130
1131         if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1132                 return -EINVAL;
1133 #ifndef CONFIG_IPV6_SUBTREES
1134         if (cfg->fc_src_len)
1135                 return -EINVAL;
1136 #endif
1137         if (cfg->fc_ifindex) {
1138                 err = -ENODEV;
1139                 dev = dev_get_by_index(net, cfg->fc_ifindex);
1140                 if (!dev)
1141                         goto out;
1142                 idev = in6_dev_get(dev);
1143                 if (!idev)
1144                         goto out;
1145         }
1146
1147         if (cfg->fc_metric == 0)
1148                 cfg->fc_metric = IP6_RT_PRIO_USER;
1149
1150         table = fib6_new_table(net, cfg->fc_table);
1151         if (table == NULL) {
1152                 err = -ENOBUFS;
1153                 goto out;
1154         }
1155
1156         rt = ip6_dst_alloc(&net->ipv6.ip6_dst_ops);
1157
1158         if (rt == NULL) {
1159                 err = -ENOMEM;
1160                 goto out;
1161         }
1162
1163         rt->u.dst.obsolete = -1;
1164         rt->rt6i_expires = (cfg->fc_flags & RTF_EXPIRES) ?
1165                                 jiffies + clock_t_to_jiffies(cfg->fc_expires) :
1166                                 0;
1167
1168         if (cfg->fc_protocol == RTPROT_UNSPEC)
1169                 cfg->fc_protocol = RTPROT_BOOT;
1170         rt->rt6i_protocol = cfg->fc_protocol;
1171
1172         addr_type = ipv6_addr_type(&cfg->fc_dst);
1173
1174         if (addr_type & IPV6_ADDR_MULTICAST)
1175                 rt->u.dst.input = ip6_mc_input;
1176         else
1177                 rt->u.dst.input = ip6_forward;
1178
1179         rt->u.dst.output = ip6_output;
1180
1181         ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1182         rt->rt6i_dst.plen = cfg->fc_dst_len;
1183         if (rt->rt6i_dst.plen == 128)
1184                rt->u.dst.flags = DST_HOST;
1185
1186 #ifdef CONFIG_IPV6_SUBTREES
1187         ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1188         rt->rt6i_src.plen = cfg->fc_src_len;
1189 #endif
1190
1191         rt->rt6i_metric = cfg->fc_metric;
1192
1193         /* We cannot add true routes via loopback here,
1194            they would result in kernel looping; promote them to reject routes
1195          */
1196         if ((cfg->fc_flags & RTF_REJECT) ||
1197             (dev && (dev->flags&IFF_LOOPBACK) && !(addr_type&IPV6_ADDR_LOOPBACK))) {
1198                 /* hold loopback dev/idev if we haven't done so. */
1199                 if (dev != net->loopback_dev) {
1200                         if (dev) {
1201                                 dev_put(dev);
1202                                 in6_dev_put(idev);
1203                         }
1204                         dev = net->loopback_dev;
1205                         dev_hold(dev);
1206                         idev = in6_dev_get(dev);
1207                         if (!idev) {
1208                                 err = -ENODEV;
1209                                 goto out;
1210                         }
1211                 }
1212                 rt->u.dst.output = ip6_pkt_discard_out;
1213                 rt->u.dst.input = ip6_pkt_discard;
1214                 rt->u.dst.error = -ENETUNREACH;
1215                 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1216                 goto install_route;
1217         }
1218
1219         if (cfg->fc_flags & RTF_GATEWAY) {
1220                 struct in6_addr *gw_addr;
1221                 int gwa_type;
1222
1223                 gw_addr = &cfg->fc_gateway;
1224                 ipv6_addr_copy(&rt->rt6i_gateway, gw_addr);
1225                 gwa_type = ipv6_addr_type(gw_addr);
1226
1227                 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1228                         struct rt6_info *grt;
1229
1230                         /* IPv6 strictly inhibits using not link-local
1231                            addresses as nexthop address.
1232                            Otherwise, router will not able to send redirects.
1233                            It is very good, but in some (rare!) circumstances
1234                            (SIT, PtP, NBMA NOARP links) it is handy to allow
1235                            some exceptions. --ANK
1236                          */
1237                         err = -EINVAL;
1238                         if (!(gwa_type&IPV6_ADDR_UNICAST))
1239                                 goto out;
1240
1241                         grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, 1);
1242
1243                         err = -EHOSTUNREACH;
1244                         if (grt == NULL)
1245                                 goto out;
1246                         if (dev) {
1247                                 if (dev != grt->rt6i_dev) {
1248                                         dst_release(&grt->u.dst);
1249                                         goto out;
1250                                 }
1251                         } else {
1252                                 dev = grt->rt6i_dev;
1253                                 idev = grt->rt6i_idev;
1254                                 dev_hold(dev);
1255                                 in6_dev_hold(grt->rt6i_idev);
1256                         }
1257                         if (!(grt->rt6i_flags&RTF_GATEWAY))
1258                                 err = 0;
1259                         dst_release(&grt->u.dst);
1260
1261                         if (err)
1262                                 goto out;
1263                 }
1264                 err = -EINVAL;
1265                 if (dev == NULL || (dev->flags&IFF_LOOPBACK))
1266                         goto out;
1267         }
1268
1269         err = -ENODEV;
1270         if (dev == NULL)
1271                 goto out;
1272
1273         if (cfg->fc_flags & (RTF_GATEWAY | RTF_NONEXTHOP)) {
1274                 rt->rt6i_nexthop = __neigh_lookup_errno(&nd_tbl, &rt->rt6i_gateway, dev);
1275                 if (IS_ERR(rt->rt6i_nexthop)) {
1276                         err = PTR_ERR(rt->rt6i_nexthop);
1277                         rt->rt6i_nexthop = NULL;
1278                         goto out;
1279                 }
1280         }
1281
1282         rt->rt6i_flags = cfg->fc_flags;
1283
1284 install_route:
1285         if (cfg->fc_mx) {
1286                 struct nlattr *nla;
1287                 int remaining;
1288
1289                 nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1290                         int type = nla_type(nla);
1291
1292                         if (type) {
1293                                 if (type > RTAX_MAX) {
1294                                         err = -EINVAL;
1295                                         goto out;
1296                                 }
1297
1298                                 rt->u.dst.metrics[type - 1] = nla_get_u32(nla);
1299                         }
1300                 }
1301         }
1302
1303         if (dst_metric(&rt->u.dst, RTAX_HOPLIMIT) == 0)
1304                 rt->u.dst.metrics[RTAX_HOPLIMIT-1] = -1;
1305         if (!dst_mtu(&rt->u.dst))
1306                 rt->u.dst.metrics[RTAX_MTU-1] = ipv6_get_mtu(dev);
1307         if (!dst_metric(&rt->u.dst, RTAX_ADVMSS))
1308                 rt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(net, dst_mtu(&rt->u.dst));
1309         rt->u.dst.dev = dev;
1310         rt->rt6i_idev = idev;
1311         rt->rt6i_table = table;
1312
1313         cfg->fc_nlinfo.nl_net = dev_net(dev);
1314
1315         return __ip6_ins_rt(rt, &cfg->fc_nlinfo);
1316
1317 out:
1318         if (dev)
1319                 dev_put(dev);
1320         if (idev)
1321                 in6_dev_put(idev);
1322         if (rt)
1323                 dst_free(&rt->u.dst);
1324         return err;
1325 }
1326
1327 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
1328 {
1329         int err;
1330         struct fib6_table *table;
1331         struct net *net = dev_net(rt->rt6i_dev);
1332
1333         if (rt == net->ipv6.ip6_null_entry)
1334                 return -ENOENT;
1335
1336         table = rt->rt6i_table;
1337         write_lock_bh(&table->tb6_lock);
1338
1339         err = fib6_del(rt, info);
1340         dst_release(&rt->u.dst);
1341
1342         write_unlock_bh(&table->tb6_lock);
1343
1344         return err;
1345 }
1346
1347 int ip6_del_rt(struct rt6_info *rt)
1348 {
1349         struct nl_info info = {
1350                 .nl_net = dev_net(rt->rt6i_dev),
1351         };
1352         return __ip6_del_rt(rt, &info);
1353 }
1354
1355 static int ip6_route_del(struct fib6_config *cfg)
1356 {
1357         struct fib6_table *table;
1358         struct fib6_node *fn;
1359         struct rt6_info *rt;
1360         int err = -ESRCH;
1361
1362         table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
1363         if (table == NULL)
1364                 return err;
1365
1366         read_lock_bh(&table->tb6_lock);
1367
1368         fn = fib6_locate(&table->tb6_root,
1369                          &cfg->fc_dst, cfg->fc_dst_len,
1370                          &cfg->fc_src, cfg->fc_src_len);
1371
1372         if (fn) {
1373                 for (rt = fn->leaf; rt; rt = rt->u.dst.rt6_next) {
1374                         if (cfg->fc_ifindex &&
1375                             (rt->rt6i_dev == NULL ||
1376                              rt->rt6i_dev->ifindex != cfg->fc_ifindex))
1377                                 continue;
1378                         if (cfg->fc_flags & RTF_GATEWAY &&
1379                             !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
1380                                 continue;
1381                         if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
1382                                 continue;
1383                         dst_hold(&rt->u.dst);
1384                         read_unlock_bh(&table->tb6_lock);
1385
1386                         return __ip6_del_rt(rt, &cfg->fc_nlinfo);
1387                 }
1388         }
1389         read_unlock_bh(&table->tb6_lock);
1390
1391         return err;
1392 }
1393
1394 /*
1395  *      Handle redirects
1396  */
1397 struct ip6rd_flowi {
1398         struct flowi fl;
1399         struct in6_addr gateway;
1400 };
1401
1402 static struct rt6_info *__ip6_route_redirect(struct net *net,
1403                                              struct fib6_table *table,
1404                                              struct flowi *fl,
1405                                              int flags)
1406 {
1407         struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl;
1408         struct rt6_info *rt;
1409         struct fib6_node *fn;
1410
1411         /*
1412          * Get the "current" route for this destination and
1413          * check if the redirect has come from approriate router.
1414          *
1415          * RFC 2461 specifies that redirects should only be
1416          * accepted if they come from the nexthop to the target.
1417          * Due to the way the routes are chosen, this notion
1418          * is a bit fuzzy and one might need to check all possible
1419          * routes.
1420          */
1421
1422         read_lock_bh(&table->tb6_lock);
1423         fn = fib6_lookup(&table->tb6_root, &fl->fl6_dst, &fl->fl6_src);
1424 restart:
1425         for (rt = fn->leaf; rt; rt = rt->u.dst.rt6_next) {
1426                 /*
1427                  * Current route is on-link; redirect is always invalid.
1428                  *
1429                  * Seems, previous statement is not true. It could
1430                  * be node, which looks for us as on-link (f.e. proxy ndisc)
1431                  * But then router serving it might decide, that we should
1432                  * know truth 8)8) --ANK (980726).
1433                  */
1434                 if (rt6_check_expired(rt))
1435                         continue;
1436                 if (!(rt->rt6i_flags & RTF_GATEWAY))
1437                         continue;
1438                 if (fl->oif != rt->rt6i_dev->ifindex)
1439                         continue;
1440                 if (!ipv6_addr_equal(&rdfl->gateway, &rt->rt6i_gateway))
1441                         continue;
1442                 break;
1443         }
1444
1445         if (!rt)
1446                 rt = net->ipv6.ip6_null_entry;
1447         BACKTRACK(net, &fl->fl6_src);
1448 out:
1449         dst_hold(&rt->u.dst);
1450
1451         read_unlock_bh(&table->tb6_lock);
1452
1453         return rt;
1454 };
1455
1456 static struct rt6_info *ip6_route_redirect(struct in6_addr *dest,
1457                                            struct in6_addr *src,
1458                                            struct in6_addr *gateway,
1459                                            struct net_device *dev)
1460 {
1461         int flags = RT6_LOOKUP_F_HAS_SADDR;
1462         struct net *net = dev_net(dev);
1463         struct ip6rd_flowi rdfl = {
1464                 .fl = {
1465                         .oif = dev->ifindex,
1466                         .nl_u = {
1467                                 .ip6_u = {
1468                                         .daddr = *dest,
1469                                         .saddr = *src,
1470                                 },
1471                         },
1472                 },
1473         };
1474
1475         ipv6_addr_copy(&rdfl.gateway, gateway);
1476
1477         if (rt6_need_strict(dest))
1478                 flags |= RT6_LOOKUP_F_IFACE;
1479
1480         return (struct rt6_info *)fib6_rule_lookup(net, (struct flowi *)&rdfl,
1481                                                    flags, __ip6_route_redirect);
1482 }
1483
1484 void rt6_redirect(struct in6_addr *dest, struct in6_addr *src,
1485                   struct in6_addr *saddr,
1486                   struct neighbour *neigh, u8 *lladdr, int on_link)
1487 {
1488         struct rt6_info *rt, *nrt = NULL;
1489         struct netevent_redirect netevent;
1490         struct net *net = dev_net(neigh->dev);
1491
1492         rt = ip6_route_redirect(dest, src, saddr, neigh->dev);
1493
1494         if (rt == net->ipv6.ip6_null_entry) {
1495                 if (net_ratelimit())
1496                         printk(KERN_DEBUG "rt6_redirect: source isn't a valid nexthop "
1497                                "for redirect target\n");
1498                 goto out;
1499         }
1500
1501         /*
1502          *      We have finally decided to accept it.
1503          */
1504
1505         neigh_update(neigh, lladdr, NUD_STALE,
1506                      NEIGH_UPDATE_F_WEAK_OVERRIDE|
1507                      NEIGH_UPDATE_F_OVERRIDE|
1508                      (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1509                                      NEIGH_UPDATE_F_ISROUTER))
1510                      );
1511
1512         /*
1513          * Redirect received -> path was valid.
1514          * Look, redirects are sent only in response to data packets,
1515          * so that this nexthop apparently is reachable. --ANK
1516          */
1517         dst_confirm(&rt->u.dst);
1518
1519         /* Duplicate redirect: silently ignore. */
1520         if (neigh == rt->u.dst.neighbour)
1521                 goto out;
1522
1523         nrt = ip6_rt_copy(rt);
1524         if (nrt == NULL)
1525                 goto out;
1526
1527         nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
1528         if (on_link)
1529                 nrt->rt6i_flags &= ~RTF_GATEWAY;
1530
1531         ipv6_addr_copy(&nrt->rt6i_dst.addr, dest);
1532         nrt->rt6i_dst.plen = 128;
1533         nrt->u.dst.flags |= DST_HOST;
1534
1535         ipv6_addr_copy(&nrt->rt6i_gateway, (struct in6_addr*)neigh->primary_key);
1536         nrt->rt6i_nexthop = neigh_clone(neigh);
1537         /* Reset pmtu, it may be better */
1538         nrt->u.dst.metrics[RTAX_MTU-1] = ipv6_get_mtu(neigh->dev);
1539         nrt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(dev_net(neigh->dev),
1540                                                         dst_mtu(&nrt->u.dst));
1541
1542         if (ip6_ins_rt(nrt))
1543                 goto out;
1544
1545         netevent.old = &rt->u.dst;
1546         netevent.new = &nrt->u.dst;
1547         call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
1548
1549         if (rt->rt6i_flags&RTF_CACHE) {
1550                 ip6_del_rt(rt);
1551                 return;
1552         }
1553
1554 out:
1555         dst_release(&rt->u.dst);
1556         return;
1557 }
1558
1559 /*
1560  *      Handle ICMP "packet too big" messages
1561  *      i.e. Path MTU discovery
1562  */
1563
1564 void rt6_pmtu_discovery(struct in6_addr *daddr, struct in6_addr *saddr,
1565                         struct net_device *dev, u32 pmtu)
1566 {
1567         struct rt6_info *rt, *nrt;
1568         struct net *net = dev_net(dev);
1569         int allfrag = 0;
1570
1571         rt = rt6_lookup(net, daddr, saddr, dev->ifindex, 0);
1572         if (rt == NULL)
1573                 return;
1574
1575         if (pmtu >= dst_mtu(&rt->u.dst))
1576                 goto out;
1577
1578         if (pmtu < IPV6_MIN_MTU) {
1579                 /*
1580                  * According to RFC2460, PMTU is set to the IPv6 Minimum Link
1581                  * MTU (1280) and a fragment header should always be included
1582                  * after a node receiving Too Big message reporting PMTU is
1583                  * less than the IPv6 Minimum Link MTU.
1584                  */
1585                 pmtu = IPV6_MIN_MTU;
1586                 allfrag = 1;
1587         }
1588
1589         /* New mtu received -> path was valid.
1590            They are sent only in response to data packets,
1591            so that this nexthop apparently is reachable. --ANK
1592          */
1593         dst_confirm(&rt->u.dst);
1594
1595         /* Host route. If it is static, it would be better
1596            not to override it, but add new one, so that
1597            when cache entry will expire old pmtu
1598            would return automatically.
1599          */
1600         if (rt->rt6i_flags & RTF_CACHE) {
1601                 rt->u.dst.metrics[RTAX_MTU-1] = pmtu;
1602                 if (allfrag)
1603                         rt->u.dst.metrics[RTAX_FEATURES-1] |= RTAX_FEATURE_ALLFRAG;
1604                 dst_set_expires(&rt->u.dst, net->ipv6.sysctl.ip6_rt_mtu_expires);
1605                 rt->rt6i_flags |= RTF_MODIFIED|RTF_EXPIRES;
1606                 goto out;
1607         }
1608
1609         /* Network route.
1610            Two cases are possible:
1611            1. It is connected route. Action: COW
1612            2. It is gatewayed route or NONEXTHOP route. Action: clone it.
1613          */
1614         if (!rt->rt6i_nexthop && !(rt->rt6i_flags & RTF_NONEXTHOP))
1615                 nrt = rt6_alloc_cow(rt, daddr, saddr);
1616         else
1617                 nrt = rt6_alloc_clone(rt, daddr);
1618
1619         if (nrt) {
1620                 nrt->u.dst.metrics[RTAX_MTU-1] = pmtu;
1621                 if (allfrag)
1622                         nrt->u.dst.metrics[RTAX_FEATURES-1] |= RTAX_FEATURE_ALLFRAG;
1623
1624                 /* According to RFC 1981, detecting PMTU increase shouldn't be
1625                  * happened within 5 mins, the recommended timer is 10 mins.
1626                  * Here this route expiration time is set to ip6_rt_mtu_expires
1627                  * which is 10 mins. After 10 mins the decreased pmtu is expired
1628                  * and detecting PMTU increase will be automatically happened.
1629                  */
1630                 dst_set_expires(&nrt->u.dst, net->ipv6.sysctl.ip6_rt_mtu_expires);
1631                 nrt->rt6i_flags |= RTF_DYNAMIC|RTF_EXPIRES;
1632
1633                 ip6_ins_rt(nrt);
1634         }
1635 out:
1636         dst_release(&rt->u.dst);
1637 }
1638
1639 /*
1640  *      Misc support functions
1641  */
1642
1643 static struct rt6_info * ip6_rt_copy(struct rt6_info *ort)
1644 {
1645         struct net *net = dev_net(ort->rt6i_dev);
1646         struct rt6_info *rt = ip6_dst_alloc(&net->ipv6.ip6_dst_ops);
1647
1648         if (rt) {
1649                 rt->u.dst.input = ort->u.dst.input;
1650                 rt->u.dst.output = ort->u.dst.output;
1651
1652                 memcpy(rt->u.dst.metrics, ort->u.dst.metrics, RTAX_MAX*sizeof(u32));
1653                 rt->u.dst.error = ort->u.dst.error;
1654                 rt->u.dst.dev = ort->u.dst.dev;
1655                 if (rt->u.dst.dev)
1656                         dev_hold(rt->u.dst.dev);
1657                 rt->rt6i_idev = ort->rt6i_idev;
1658                 if (rt->rt6i_idev)
1659                         in6_dev_hold(rt->rt6i_idev);
1660                 rt->u.dst.lastuse = jiffies;
1661                 rt->rt6i_expires = 0;
1662
1663                 ipv6_addr_copy(&rt->rt6i_gateway, &ort->rt6i_gateway);
1664                 rt->rt6i_flags = ort->rt6i_flags & ~RTF_EXPIRES;
1665                 rt->rt6i_metric = 0;
1666
1667                 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1668 #ifdef CONFIG_IPV6_SUBTREES
1669                 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1670 #endif
1671                 rt->rt6i_table = ort->rt6i_table;
1672         }
1673         return rt;
1674 }
1675
1676 #ifdef CONFIG_IPV6_ROUTE_INFO
1677 static struct rt6_info *rt6_get_route_info(struct net *net,
1678                                            struct in6_addr *prefix, int prefixlen,
1679                                            struct in6_addr *gwaddr, int ifindex)
1680 {
1681         struct fib6_node *fn;
1682         struct rt6_info *rt = NULL;
1683         struct fib6_table *table;
1684
1685         table = fib6_get_table(net, RT6_TABLE_INFO);
1686         if (table == NULL)
1687                 return NULL;
1688
1689         write_lock_bh(&table->tb6_lock);
1690         fn = fib6_locate(&table->tb6_root, prefix ,prefixlen, NULL, 0);
1691         if (!fn)
1692                 goto out;
1693
1694         for (rt = fn->leaf; rt; rt = rt->u.dst.rt6_next) {
1695                 if (rt->rt6i_dev->ifindex != ifindex)
1696                         continue;
1697                 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
1698                         continue;
1699                 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
1700                         continue;
1701                 dst_hold(&rt->u.dst);
1702                 break;
1703         }
1704 out:
1705         write_unlock_bh(&table->tb6_lock);
1706         return rt;
1707 }
1708
1709 static struct rt6_info *rt6_add_route_info(struct net *net,
1710                                            struct in6_addr *prefix, int prefixlen,
1711                                            struct in6_addr *gwaddr, int ifindex,
1712                                            unsigned pref)
1713 {
1714         struct fib6_config cfg = {
1715                 .fc_table       = RT6_TABLE_INFO,
1716                 .fc_metric      = IP6_RT_PRIO_USER,
1717                 .fc_ifindex     = ifindex,
1718                 .fc_dst_len     = prefixlen,
1719                 .fc_flags       = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
1720                                   RTF_UP | RTF_PREF(pref),
1721                 .fc_nlinfo.pid = 0,
1722                 .fc_nlinfo.nlh = NULL,
1723                 .fc_nlinfo.nl_net = net,
1724         };
1725
1726         ipv6_addr_copy(&cfg.fc_dst, prefix);
1727         ipv6_addr_copy(&cfg.fc_gateway, gwaddr);
1728
1729         /* We should treat it as a default route if prefix length is 0. */
1730         if (!prefixlen)
1731                 cfg.fc_flags |= RTF_DEFAULT;
1732
1733         ip6_route_add(&cfg);
1734
1735         return rt6_get_route_info(net, prefix, prefixlen, gwaddr, ifindex);
1736 }
1737 #endif
1738
1739 struct rt6_info *rt6_get_dflt_router(struct in6_addr *addr, struct net_device *dev)
1740 {
1741         struct rt6_info *rt;
1742         struct fib6_table *table;
1743
1744         table = fib6_get_table(dev_net(dev), RT6_TABLE_DFLT);
1745         if (table == NULL)
1746                 return NULL;
1747
1748         write_lock_bh(&table->tb6_lock);
1749         for (rt = table->tb6_root.leaf; rt; rt=rt->u.dst.rt6_next) {
1750                 if (dev == rt->rt6i_dev &&
1751                     ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
1752                     ipv6_addr_equal(&rt->rt6i_gateway, addr))
1753                         break;
1754         }
1755         if (rt)
1756                 dst_hold(&rt->u.dst);
1757         write_unlock_bh(&table->tb6_lock);
1758         return rt;
1759 }
1760
1761 struct rt6_info *rt6_add_dflt_router(struct in6_addr *gwaddr,
1762                                      struct net_device *dev,
1763                                      unsigned int pref)
1764 {
1765         struct fib6_config cfg = {
1766                 .fc_table       = RT6_TABLE_DFLT,
1767                 .fc_metric      = IP6_RT_PRIO_USER,
1768                 .fc_ifindex     = dev->ifindex,
1769                 .fc_flags       = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
1770                                   RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
1771                 .fc_nlinfo.pid = 0,
1772                 .fc_nlinfo.nlh = NULL,
1773                 .fc_nlinfo.nl_net = dev_net(dev),
1774         };
1775
1776         ipv6_addr_copy(&cfg.fc_gateway, gwaddr);
1777
1778         ip6_route_add(&cfg);
1779
1780         return rt6_get_dflt_router(gwaddr, dev);
1781 }
1782
1783 void rt6_purge_dflt_routers(struct net *net)
1784 {
1785         struct rt6_info *rt;
1786         struct fib6_table *table;
1787
1788         /* NOTE: Keep consistent with rt6_get_dflt_router */
1789         table = fib6_get_table(net, RT6_TABLE_DFLT);
1790         if (table == NULL)
1791                 return;
1792
1793 restart:
1794         read_lock_bh(&table->tb6_lock);
1795         for (rt = table->tb6_root.leaf; rt; rt = rt->u.dst.rt6_next) {
1796                 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF)) {
1797                         dst_hold(&rt->u.dst);
1798                         read_unlock_bh(&table->tb6_lock);
1799                         ip6_del_rt(rt);
1800                         goto restart;
1801                 }
1802         }
1803         read_unlock_bh(&table->tb6_lock);
1804 }
1805
1806 static void rtmsg_to_fib6_config(struct net *net,
1807                                  struct in6_rtmsg *rtmsg,
1808                                  struct fib6_config *cfg)
1809 {
1810         memset(cfg, 0, sizeof(*cfg));
1811
1812         cfg->fc_table = RT6_TABLE_MAIN;
1813         cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
1814         cfg->fc_metric = rtmsg->rtmsg_metric;
1815         cfg->fc_expires = rtmsg->rtmsg_info;
1816         cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
1817         cfg->fc_src_len = rtmsg->rtmsg_src_len;
1818         cfg->fc_flags = rtmsg->rtmsg_flags;
1819
1820         cfg->fc_nlinfo.nl_net = net;
1821
1822         ipv6_addr_copy(&cfg->fc_dst, &rtmsg->rtmsg_dst);
1823         ipv6_addr_copy(&cfg->fc_src, &rtmsg->rtmsg_src);
1824         ipv6_addr_copy(&cfg->fc_gateway, &rtmsg->rtmsg_gateway);
1825 }
1826
1827 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
1828 {
1829         struct fib6_config cfg;
1830         struct in6_rtmsg rtmsg;
1831         int err;
1832
1833         switch(cmd) {
1834         case SIOCADDRT:         /* Add a route */
1835         case SIOCDELRT:         /* Delete a route */
1836                 if (!capable(CAP_NET_ADMIN))
1837                         return -EPERM;
1838                 err = copy_from_user(&rtmsg, arg,
1839                                      sizeof(struct in6_rtmsg));
1840                 if (err)
1841                         return -EFAULT;
1842
1843                 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
1844
1845                 rtnl_lock();
1846                 switch (cmd) {
1847                 case SIOCADDRT:
1848                         err = ip6_route_add(&cfg);
1849                         break;
1850                 case SIOCDELRT:
1851                         err = ip6_route_del(&cfg);
1852                         break;
1853                 default:
1854                         err = -EINVAL;
1855                 }
1856                 rtnl_unlock();
1857
1858                 return err;
1859         }
1860
1861         return -EINVAL;
1862 }
1863
1864 /*
1865  *      Drop the packet on the floor
1866  */
1867
1868 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
1869 {
1870         int type;
1871         struct dst_entry *dst = skb_dst(skb);
1872         switch (ipstats_mib_noroutes) {
1873         case IPSTATS_MIB_INNOROUTES:
1874                 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
1875                 if (type == IPV6_ADDR_ANY) {
1876                         IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
1877                                       IPSTATS_MIB_INADDRERRORS);
1878                         break;
1879                 }
1880                 /* FALLTHROUGH */
1881         case IPSTATS_MIB_OUTNOROUTES:
1882                 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
1883                               ipstats_mib_noroutes);
1884                 break;
1885         }
1886         icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
1887         kfree_skb(skb);
1888         return 0;
1889 }
1890
1891 static int ip6_pkt_discard(struct sk_buff *skb)
1892 {
1893         return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
1894 }
1895
1896 static int ip6_pkt_discard_out(struct sk_buff *skb)
1897 {
1898         skb->dev = skb_dst(skb)->dev;
1899         return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
1900 }
1901
1902 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
1903
1904 static int ip6_pkt_prohibit(struct sk_buff *skb)
1905 {
1906         return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
1907 }
1908
1909 static int ip6_pkt_prohibit_out(struct sk_buff *skb)
1910 {
1911         skb->dev = skb_dst(skb)->dev;
1912         return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
1913 }
1914
1915 #endif
1916
1917 /*
1918  *      Allocate a dst for local (unicast / anycast) address.
1919  */
1920
1921 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
1922                                     const struct in6_addr *addr,
1923                                     int anycast)
1924 {
1925         struct net *net = dev_net(idev->dev);
1926         struct rt6_info *rt = ip6_dst_alloc(&net->ipv6.ip6_dst_ops);
1927         struct neighbour *neigh;
1928
1929         if (rt == NULL)
1930                 return ERR_PTR(-ENOMEM);
1931
1932         dev_hold(net->loopback_dev);
1933         in6_dev_hold(idev);
1934
1935         rt->u.dst.flags = DST_HOST;
1936         rt->u.dst.input = ip6_input;
1937         rt->u.dst.output = ip6_output;
1938         rt->rt6i_dev = net->loopback_dev;
1939         rt->rt6i_idev = idev;
1940         rt->u.dst.metrics[RTAX_MTU-1] = ipv6_get_mtu(rt->rt6i_dev);
1941         rt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(net, dst_mtu(&rt->u.dst));
1942         rt->u.dst.metrics[RTAX_HOPLIMIT-1] = -1;
1943         rt->u.dst.obsolete = -1;
1944
1945         rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
1946         if (anycast)
1947                 rt->rt6i_flags |= RTF_ANYCAST;
1948         else
1949                 rt->rt6i_flags |= RTF_LOCAL;
1950         neigh = ndisc_get_neigh(rt->rt6i_dev, &rt->rt6i_gateway);
1951         if (IS_ERR(neigh)) {
1952                 dst_free(&rt->u.dst);
1953
1954                 /* We are casting this because that is the return
1955                  * value type.  But an errno encoded pointer is the
1956                  * same regardless of the underlying pointer type,
1957                  * and that's what we are returning.  So this is OK.
1958                  */
1959                 return (struct rt6_info *) neigh;
1960         }
1961         rt->rt6i_nexthop = neigh;
1962
1963         ipv6_addr_copy(&rt->rt6i_dst.addr, addr);
1964         rt->rt6i_dst.plen = 128;
1965         rt->rt6i_table = fib6_get_table(net, RT6_TABLE_LOCAL);
1966
1967         atomic_set(&rt->u.dst.__refcnt, 1);
1968
1969         return rt;
1970 }
1971
1972 struct arg_dev_net {
1973         struct net_device *dev;
1974         struct net *net;
1975 };
1976
1977 static int fib6_ifdown(struct rt6_info *rt, void *arg)
1978 {
1979         struct net_device *dev = ((struct arg_dev_net *)arg)->dev;
1980         struct net *net = ((struct arg_dev_net *)arg)->net;
1981
1982         if (((void *)rt->rt6i_dev == dev || dev == NULL) &&
1983             rt != net->ipv6.ip6_null_entry) {
1984                 RT6_TRACE("deleted by ifdown %p\n", rt);
1985                 return -1;
1986         }
1987         return 0;
1988 }
1989
1990 void rt6_ifdown(struct net *net, struct net_device *dev)
1991 {
1992         struct arg_dev_net adn = {
1993                 .dev = dev,
1994                 .net = net,
1995         };
1996
1997         fib6_clean_all(net, fib6_ifdown, 0, &adn);
1998         icmp6_clean_all(fib6_ifdown, &adn);
1999 }
2000
2001 struct rt6_mtu_change_arg
2002 {
2003         struct net_device *dev;
2004         unsigned mtu;
2005 };
2006
2007 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2008 {
2009         struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2010         struct inet6_dev *idev;
2011         struct net *net = dev_net(arg->dev);
2012
2013         /* In IPv6 pmtu discovery is not optional,
2014            so that RTAX_MTU lock cannot disable it.
2015            We still use this lock to block changes
2016            caused by addrconf/ndisc.
2017         */
2018
2019         idev = __in6_dev_get(arg->dev);
2020         if (idev == NULL)
2021                 return 0;
2022
2023         /* For administrative MTU increase, there is no way to discover
2024            IPv6 PMTU increase, so PMTU increase should be updated here.
2025            Since RFC 1981 doesn't include administrative MTU increase
2026            update PMTU increase is a MUST. (i.e. jumbo frame)
2027          */
2028         /*
2029            If new MTU is less than route PMTU, this new MTU will be the
2030            lowest MTU in the path, update the route PMTU to reflect PMTU
2031            decreases; if new MTU is greater than route PMTU, and the
2032            old MTU is the lowest MTU in the path, update the route PMTU
2033            to reflect the increase. In this case if the other nodes' MTU
2034            also have the lowest MTU, TOO BIG MESSAGE will be lead to
2035            PMTU discouvery.
2036          */
2037         if (rt->rt6i_dev == arg->dev &&
2038             !dst_metric_locked(&rt->u.dst, RTAX_MTU) &&
2039             (dst_mtu(&rt->u.dst) >= arg->mtu ||
2040              (dst_mtu(&rt->u.dst) < arg->mtu &&
2041               dst_mtu(&rt->u.dst) == idev->cnf.mtu6))) {
2042                 rt->u.dst.metrics[RTAX_MTU-1] = arg->mtu;
2043                 rt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(net, arg->mtu);
2044         }
2045         return 0;
2046 }
2047
2048 void rt6_mtu_change(struct net_device *dev, unsigned mtu)
2049 {
2050         struct rt6_mtu_change_arg arg = {
2051                 .dev = dev,
2052                 .mtu = mtu,
2053         };
2054
2055         fib6_clean_all(dev_net(dev), rt6_mtu_change_route, 0, &arg);
2056 }
2057
2058 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2059         [RTA_GATEWAY]           = { .len = sizeof(struct in6_addr) },
2060         [RTA_OIF]               = { .type = NLA_U32 },
2061         [RTA_IIF]               = { .type = NLA_U32 },
2062         [RTA_PRIORITY]          = { .type = NLA_U32 },
2063         [RTA_METRICS]           = { .type = NLA_NESTED },
2064 };
2065
2066 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2067                               struct fib6_config *cfg)
2068 {
2069         struct rtmsg *rtm;
2070         struct nlattr *tb[RTA_MAX+1];
2071         int err;
2072
2073         err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2074         if (err < 0)
2075                 goto errout;
2076
2077         err = -EINVAL;
2078         rtm = nlmsg_data(nlh);
2079         memset(cfg, 0, sizeof(*cfg));
2080
2081         cfg->fc_table = rtm->rtm_table;
2082         cfg->fc_dst_len = rtm->rtm_dst_len;
2083         cfg->fc_src_len = rtm->rtm_src_len;
2084         cfg->fc_flags = RTF_UP;
2085         cfg->fc_protocol = rtm->rtm_protocol;
2086
2087         if (rtm->rtm_type == RTN_UNREACHABLE)
2088                 cfg->fc_flags |= RTF_REJECT;
2089
2090         cfg->fc_nlinfo.pid = NETLINK_CB(skb).pid;
2091         cfg->fc_nlinfo.nlh = nlh;
2092         cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2093
2094         if (tb[RTA_GATEWAY]) {
2095                 nla_memcpy(&cfg->fc_gateway, tb[RTA_GATEWAY], 16);
2096                 cfg->fc_flags |= RTF_GATEWAY;
2097         }
2098
2099         if (tb[RTA_DST]) {
2100                 int plen = (rtm->rtm_dst_len + 7) >> 3;
2101
2102                 if (nla_len(tb[RTA_DST]) < plen)
2103                         goto errout;
2104
2105                 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2106         }
2107
2108         if (tb[RTA_SRC]) {
2109                 int plen = (rtm->rtm_src_len + 7) >> 3;
2110
2111                 if (nla_len(tb[RTA_SRC]) < plen)
2112                         goto errout;
2113
2114                 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2115         }
2116
2117         if (tb[RTA_OIF])
2118                 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2119
2120         if (tb[RTA_PRIORITY])
2121                 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2122
2123         if (tb[RTA_METRICS]) {
2124                 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2125                 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2126         }
2127
2128         if (tb[RTA_TABLE])
2129                 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2130
2131         err = 0;
2132 errout:
2133         return err;
2134 }
2135
2136 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
2137 {
2138         struct fib6_config cfg;
2139         int err;
2140
2141         err = rtm_to_fib6_config(skb, nlh, &cfg);
2142         if (err < 0)
2143                 return err;
2144
2145         return ip6_route_del(&cfg);
2146 }
2147
2148 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
2149 {
2150         struct fib6_config cfg;
2151         int err;
2152
2153         err = rtm_to_fib6_config(skb, nlh, &cfg);
2154         if (err < 0)
2155                 return err;
2156
2157         return ip6_route_add(&cfg);
2158 }
2159
2160 static inline size_t rt6_nlmsg_size(void)
2161 {
2162         return NLMSG_ALIGN(sizeof(struct rtmsg))
2163                + nla_total_size(16) /* RTA_SRC */
2164                + nla_total_size(16) /* RTA_DST */
2165                + nla_total_size(16) /* RTA_GATEWAY */
2166                + nla_total_size(16) /* RTA_PREFSRC */
2167                + nla_total_size(4) /* RTA_TABLE */
2168                + nla_total_size(4) /* RTA_IIF */
2169                + nla_total_size(4) /* RTA_OIF */
2170                + nla_total_size(4) /* RTA_PRIORITY */
2171                + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
2172                + nla_total_size(sizeof(struct rta_cacheinfo));
2173 }
2174
2175 static int rt6_fill_node(struct net *net,
2176                          struct sk_buff *skb, struct rt6_info *rt,
2177                          struct in6_addr *dst, struct in6_addr *src,
2178                          int iif, int type, u32 pid, u32 seq,
2179                          int prefix, int nowait, unsigned int flags)
2180 {
2181         struct rtmsg *rtm;
2182         struct nlmsghdr *nlh;
2183         long expires;
2184         u32 table;
2185
2186         if (prefix) {   /* user wants prefix routes only */
2187                 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
2188                         /* success since this is not a prefix route */
2189                         return 1;
2190                 }
2191         }
2192
2193         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*rtm), flags);
2194         if (nlh == NULL)
2195                 return -EMSGSIZE;
2196
2197         rtm = nlmsg_data(nlh);
2198         rtm->rtm_family = AF_INET6;
2199         rtm->rtm_dst_len = rt->rt6i_dst.plen;
2200         rtm->rtm_src_len = rt->rt6i_src.plen;
2201         rtm->rtm_tos = 0;
2202         if (rt->rt6i_table)
2203                 table = rt->rt6i_table->tb6_id;
2204         else
2205                 table = RT6_TABLE_UNSPEC;
2206         rtm->rtm_table = table;
2207         NLA_PUT_U32(skb, RTA_TABLE, table);
2208         if (rt->rt6i_flags&RTF_REJECT)
2209                 rtm->rtm_type = RTN_UNREACHABLE;
2210         else if (rt->rt6i_dev && (rt->rt6i_dev->flags&IFF_LOOPBACK))
2211                 rtm->rtm_type = RTN_LOCAL;
2212         else
2213                 rtm->rtm_type = RTN_UNICAST;
2214         rtm->rtm_flags = 0;
2215         rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2216         rtm->rtm_protocol = rt->rt6i_protocol;
2217         if (rt->rt6i_flags&RTF_DYNAMIC)
2218                 rtm->rtm_protocol = RTPROT_REDIRECT;
2219         else if (rt->rt6i_flags & RTF_ADDRCONF)
2220                 rtm->rtm_protocol = RTPROT_KERNEL;
2221         else if (rt->rt6i_flags&RTF_DEFAULT)
2222                 rtm->rtm_protocol = RTPROT_RA;
2223
2224         if (rt->rt6i_flags&RTF_CACHE)
2225                 rtm->rtm_flags |= RTM_F_CLONED;
2226
2227         if (dst) {
2228                 NLA_PUT(skb, RTA_DST, 16, dst);
2229                 rtm->rtm_dst_len = 128;
2230         } else if (rtm->rtm_dst_len)
2231                 NLA_PUT(skb, RTA_DST, 16, &rt->rt6i_dst.addr);
2232 #ifdef CONFIG_IPV6_SUBTREES
2233         if (src) {
2234                 NLA_PUT(skb, RTA_SRC, 16, src);
2235                 rtm->rtm_src_len = 128;
2236         } else if (rtm->rtm_src_len)
2237                 NLA_PUT(skb, RTA_SRC, 16, &rt->rt6i_src.addr);
2238 #endif
2239         if (iif) {
2240 #ifdef CONFIG_IPV6_MROUTE
2241                 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
2242                         int err = ip6mr_get_route(net, skb, rtm, nowait);
2243                         if (err <= 0) {
2244                                 if (!nowait) {
2245                                         if (err == 0)
2246                                                 return 0;
2247                                         goto nla_put_failure;
2248                                 } else {
2249                                         if (err == -EMSGSIZE)
2250                                                 goto nla_put_failure;
2251                                 }
2252                         }
2253                 } else
2254 #endif
2255                         NLA_PUT_U32(skb, RTA_IIF, iif);
2256         } else if (dst) {
2257                 struct inet6_dev *idev = ip6_dst_idev(&rt->u.dst);
2258                 struct in6_addr saddr_buf;
2259                 if (ipv6_dev_get_saddr(net, idev ? idev->dev : NULL,
2260                                        dst, 0, &saddr_buf) == 0)
2261                         NLA_PUT(skb, RTA_PREFSRC, 16, &saddr_buf);
2262         }
2263
2264         if (rtnetlink_put_metrics(skb, rt->u.dst.metrics) < 0)
2265                 goto nla_put_failure;
2266
2267         if (rt->u.dst.neighbour)
2268                 NLA_PUT(skb, RTA_GATEWAY, 16, &rt->u.dst.neighbour->primary_key);
2269
2270         if (rt->u.dst.dev)
2271                 NLA_PUT_U32(skb, RTA_OIF, rt->rt6i_dev->ifindex);
2272
2273         NLA_PUT_U32(skb, RTA_PRIORITY, rt->rt6i_metric);
2274
2275         if (!(rt->rt6i_flags & RTF_EXPIRES))
2276                 expires = 0;
2277         else if (rt->rt6i_expires - jiffies < INT_MAX)
2278                 expires = rt->rt6i_expires - jiffies;
2279         else
2280                 expires = INT_MAX;
2281
2282         if (rtnl_put_cacheinfo(skb, &rt->u.dst, 0, 0, 0,
2283                                expires, rt->u.dst.error) < 0)
2284                 goto nla_put_failure;
2285
2286         return nlmsg_end(skb, nlh);
2287
2288 nla_put_failure:
2289         nlmsg_cancel(skb, nlh);
2290         return -EMSGSIZE;
2291 }
2292
2293 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
2294 {
2295         struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
2296         int prefix;
2297
2298         if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
2299                 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
2300                 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
2301         } else
2302                 prefix = 0;
2303
2304         return rt6_fill_node(arg->net,
2305                      arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
2306                      NETLINK_CB(arg->cb->skb).pid, arg->cb->nlh->nlmsg_seq,
2307                      prefix, 0, NLM_F_MULTI);
2308 }
2309
2310 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr* nlh, void *arg)
2311 {
2312         struct net *net = sock_net(in_skb->sk);
2313         struct nlattr *tb[RTA_MAX+1];
2314         struct rt6_info *rt;
2315         struct sk_buff *skb;
2316         struct rtmsg *rtm;
2317         struct flowi fl;
2318         int err, iif = 0;
2319
2320         err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2321         if (err < 0)
2322                 goto errout;
2323
2324         err = -EINVAL;
2325         memset(&fl, 0, sizeof(fl));
2326
2327         if (tb[RTA_SRC]) {
2328                 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
2329                         goto errout;
2330
2331                 ipv6_addr_copy(&fl.fl6_src, nla_data(tb[RTA_SRC]));
2332         }
2333
2334         if (tb[RTA_DST]) {
2335                 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
2336                         goto errout;
2337
2338                 ipv6_addr_copy(&fl.fl6_dst, nla_data(tb[RTA_DST]));
2339         }
2340
2341         if (tb[RTA_IIF])
2342                 iif = nla_get_u32(tb[RTA_IIF]);
2343
2344         if (tb[RTA_OIF])
2345                 fl.oif = nla_get_u32(tb[RTA_OIF]);
2346
2347         if (iif) {
2348                 struct net_device *dev;
2349                 dev = __dev_get_by_index(net, iif);
2350                 if (!dev) {
2351                         err = -ENODEV;
2352                         goto errout;
2353                 }
2354         }
2355
2356         skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
2357         if (skb == NULL) {
2358                 err = -ENOBUFS;
2359                 goto errout;
2360         }
2361
2362         /* Reserve room for dummy headers, this skb can pass
2363            through good chunk of routing engine.
2364          */
2365         skb_reset_mac_header(skb);
2366         skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
2367
2368         rt = (struct rt6_info*) ip6_route_output(net, NULL, &fl);
2369         skb_dst_set(skb, &rt->u.dst);
2370
2371         err = rt6_fill_node(net, skb, rt, &fl.fl6_dst, &fl.fl6_src, iif,
2372                             RTM_NEWROUTE, NETLINK_CB(in_skb).pid,
2373                             nlh->nlmsg_seq, 0, 0, 0);
2374         if (err < 0) {
2375                 kfree_skb(skb);
2376                 goto errout;
2377         }
2378
2379         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
2380 errout:
2381         return err;
2382 }
2383
2384 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info)
2385 {
2386         struct sk_buff *skb;
2387         struct net *net = info->nl_net;
2388         u32 seq;
2389         int err;
2390
2391         err = -ENOBUFS;
2392         seq = info->nlh != NULL ? info->nlh->nlmsg_seq : 0;
2393
2394         skb = nlmsg_new(rt6_nlmsg_size(), gfp_any());
2395         if (skb == NULL)
2396                 goto errout;
2397
2398         err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
2399                                 event, info->pid, seq, 0, 0, 0);
2400         if (err < 0) {
2401                 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
2402                 WARN_ON(err == -EMSGSIZE);
2403                 kfree_skb(skb);
2404                 goto errout;
2405         }
2406         rtnl_notify(skb, net, info->pid, RTNLGRP_IPV6_ROUTE,
2407                     info->nlh, gfp_any());
2408         return;
2409 errout:
2410         if (err < 0)
2411                 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
2412 }
2413
2414 static int ip6_route_dev_notify(struct notifier_block *this,
2415                                 unsigned long event, void *data)
2416 {
2417         struct net_device *dev = (struct net_device *)data;
2418         struct net *net = dev_net(dev);
2419
2420         if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
2421                 net->ipv6.ip6_null_entry->u.dst.dev = dev;
2422                 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
2423 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2424                 net->ipv6.ip6_prohibit_entry->u.dst.dev = dev;
2425                 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
2426                 net->ipv6.ip6_blk_hole_entry->u.dst.dev = dev;
2427                 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
2428 #endif
2429         }
2430
2431         return NOTIFY_OK;
2432 }
2433
2434 /*
2435  *      /proc
2436  */
2437
2438 #ifdef CONFIG_PROC_FS
2439
2440 #define RT6_INFO_LEN (32 + 4 + 32 + 4 + 32 + 40 + 5 + 1)
2441
2442 struct rt6_proc_arg
2443 {
2444         char *buffer;
2445         int offset;
2446         int length;
2447         int skip;
2448         int len;
2449 };
2450
2451 static int rt6_info_route(struct rt6_info *rt, void *p_arg)
2452 {
2453         struct seq_file *m = p_arg;
2454
2455         seq_printf(m, "%pi6 %02x ", &rt->rt6i_dst.addr, rt->rt6i_dst.plen);
2456
2457 #ifdef CONFIG_IPV6_SUBTREES
2458         seq_printf(m, "%pi6 %02x ", &rt->rt6i_src.addr, rt->rt6i_src.plen);
2459 #else
2460         seq_puts(m, "00000000000000000000000000000000 00 ");
2461 #endif
2462
2463         if (rt->rt6i_nexthop) {
2464                 seq_printf(m, "%pi6", rt->rt6i_nexthop->primary_key);
2465         } else {
2466                 seq_puts(m, "00000000000000000000000000000000");
2467         }
2468         seq_printf(m, " %08x %08x %08x %08x %8s\n",
2469                    rt->rt6i_metric, atomic_read(&rt->u.dst.__refcnt),
2470                    rt->u.dst.__use, rt->rt6i_flags,
2471                    rt->rt6i_dev ? rt->rt6i_dev->name : "");
2472         return 0;
2473 }
2474
2475 static int ipv6_route_show(struct seq_file *m, void *v)
2476 {
2477         struct net *net = (struct net *)m->private;
2478         fib6_clean_all(net, rt6_info_route, 0, m);
2479         return 0;
2480 }
2481
2482 static int ipv6_route_open(struct inode *inode, struct file *file)
2483 {
2484         return single_open_net(inode, file, ipv6_route_show);
2485 }
2486
2487 static const struct file_operations ipv6_route_proc_fops = {
2488         .owner          = THIS_MODULE,
2489         .open           = ipv6_route_open,
2490         .read           = seq_read,
2491         .llseek         = seq_lseek,
2492         .release        = single_release_net,
2493 };
2494
2495 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
2496 {
2497         struct net *net = (struct net *)seq->private;
2498         seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
2499                    net->ipv6.rt6_stats->fib_nodes,
2500                    net->ipv6.rt6_stats->fib_route_nodes,
2501                    net->ipv6.rt6_stats->fib_rt_alloc,
2502                    net->ipv6.rt6_stats->fib_rt_entries,
2503                    net->ipv6.rt6_stats->fib_rt_cache,
2504                    atomic_read(&net->ipv6.ip6_dst_ops.entries),
2505                    net->ipv6.rt6_stats->fib_discarded_routes);
2506
2507         return 0;
2508 }
2509
2510 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
2511 {
2512         return single_open_net(inode, file, rt6_stats_seq_show);
2513 }
2514
2515 static const struct file_operations rt6_stats_seq_fops = {
2516         .owner   = THIS_MODULE,
2517         .open    = rt6_stats_seq_open,
2518         .read    = seq_read,
2519         .llseek  = seq_lseek,
2520         .release = single_release_net,
2521 };
2522 #endif  /* CONFIG_PROC_FS */
2523
2524 #ifdef CONFIG_SYSCTL
2525
2526 static
2527 int ipv6_sysctl_rtcache_flush(ctl_table *ctl, int write,
2528                               void __user *buffer, size_t *lenp, loff_t *ppos)
2529 {
2530         struct net *net = current->nsproxy->net_ns;
2531         int delay = net->ipv6.sysctl.flush_delay;
2532         if (write) {
2533                 proc_dointvec(ctl, write, buffer, lenp, ppos);
2534                 fib6_run_gc(delay <= 0 ? ~0UL : (unsigned long)delay, net);
2535                 return 0;
2536         } else
2537                 return -EINVAL;
2538 }
2539
2540 ctl_table ipv6_route_table_template[] = {
2541         {
2542                 .procname       =       "flush",
2543                 .data           =       &init_net.ipv6.sysctl.flush_delay,
2544                 .maxlen         =       sizeof(int),
2545                 .mode           =       0200,
2546                 .proc_handler   =       ipv6_sysctl_rtcache_flush
2547         },
2548         {
2549                 .procname       =       "gc_thresh",
2550                 .data           =       &ip6_dst_ops_template.gc_thresh,
2551                 .maxlen         =       sizeof(int),
2552                 .mode           =       0644,
2553                 .proc_handler   =       proc_dointvec,
2554         },
2555         {
2556                 .procname       =       "max_size",
2557                 .data           =       &init_net.ipv6.sysctl.ip6_rt_max_size,
2558                 .maxlen         =       sizeof(int),
2559                 .mode           =       0644,
2560                 .proc_handler   =       proc_dointvec,
2561         },
2562         {
2563                 .procname       =       "gc_min_interval",
2564                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2565                 .maxlen         =       sizeof(int),
2566                 .mode           =       0644,
2567                 .proc_handler   =       proc_dointvec_jiffies,
2568         },
2569         {
2570                 .procname       =       "gc_timeout",
2571                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
2572                 .maxlen         =       sizeof(int),
2573                 .mode           =       0644,
2574                 .proc_handler   =       proc_dointvec_jiffies,
2575         },
2576         {
2577                 .procname       =       "gc_interval",
2578                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_interval,
2579                 .maxlen         =       sizeof(int),
2580                 .mode           =       0644,
2581                 .proc_handler   =       proc_dointvec_jiffies,
2582         },
2583         {
2584                 .procname       =       "gc_elasticity",
2585                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
2586                 .maxlen         =       sizeof(int),
2587                 .mode           =       0644,
2588                 .proc_handler   =       proc_dointvec_jiffies,
2589         },
2590         {
2591                 .procname       =       "mtu_expires",
2592                 .data           =       &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
2593                 .maxlen         =       sizeof(int),
2594                 .mode           =       0644,
2595                 .proc_handler   =       proc_dointvec_jiffies,
2596         },
2597         {
2598                 .procname       =       "min_adv_mss",
2599                 .data           =       &init_net.ipv6.sysctl.ip6_rt_min_advmss,
2600                 .maxlen         =       sizeof(int),
2601                 .mode           =       0644,
2602                 .proc_handler   =       proc_dointvec_jiffies,
2603         },
2604         {
2605                 .procname       =       "gc_min_interval_ms",
2606                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2607                 .maxlen         =       sizeof(int),
2608                 .mode           =       0644,
2609                 .proc_handler   =       proc_dointvec_ms_jiffies,
2610         },
2611         { }
2612 };
2613
2614 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
2615 {
2616         struct ctl_table *table;
2617
2618         table = kmemdup(ipv6_route_table_template,
2619                         sizeof(ipv6_route_table_template),
2620                         GFP_KERNEL);
2621
2622         if (table) {
2623                 table[0].data = &net->ipv6.sysctl.flush_delay;
2624                 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
2625                 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
2626                 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2627                 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
2628                 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
2629                 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
2630                 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
2631                 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
2632                 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2633         }
2634
2635         return table;
2636 }
2637 #endif
2638
2639 static int __net_init ip6_route_net_init(struct net *net)
2640 {
2641         int ret = -ENOMEM;
2642
2643         memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
2644                sizeof(net->ipv6.ip6_dst_ops));
2645
2646         net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
2647                                            sizeof(*net->ipv6.ip6_null_entry),
2648                                            GFP_KERNEL);
2649         if (!net->ipv6.ip6_null_entry)
2650                 goto out_ip6_dst_ops;
2651         net->ipv6.ip6_null_entry->u.dst.path =
2652                 (struct dst_entry *)net->ipv6.ip6_null_entry;
2653         net->ipv6.ip6_null_entry->u.dst.ops = &net->ipv6.ip6_dst_ops;
2654
2655 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2656         net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
2657                                                sizeof(*net->ipv6.ip6_prohibit_entry),
2658                                                GFP_KERNEL);
2659         if (!net->ipv6.ip6_prohibit_entry)
2660                 goto out_ip6_null_entry;
2661         net->ipv6.ip6_prohibit_entry->u.dst.path =
2662                 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
2663         net->ipv6.ip6_prohibit_entry->u.dst.ops = &net->ipv6.ip6_dst_ops;
2664
2665         net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
2666                                                sizeof(*net->ipv6.ip6_blk_hole_entry),
2667                                                GFP_KERNEL);
2668         if (!net->ipv6.ip6_blk_hole_entry)
2669                 goto out_ip6_prohibit_entry;
2670         net->ipv6.ip6_blk_hole_entry->u.dst.path =
2671                 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
2672         net->ipv6.ip6_blk_hole_entry->u.dst.ops = &net->ipv6.ip6_dst_ops;
2673 #endif
2674
2675         net->ipv6.sysctl.flush_delay = 0;
2676         net->ipv6.sysctl.ip6_rt_max_size = 4096;
2677         net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
2678         net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
2679         net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
2680         net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
2681         net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
2682         net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
2683
2684 #ifdef CONFIG_PROC_FS
2685         proc_net_fops_create(net, "ipv6_route", 0, &ipv6_route_proc_fops);
2686         proc_net_fops_create(net, "rt6_stats", S_IRUGO, &rt6_stats_seq_fops);
2687 #endif
2688         net->ipv6.ip6_rt_gc_expire = 30*HZ;
2689
2690         ret = 0;
2691 out:
2692         return ret;
2693
2694 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2695 out_ip6_prohibit_entry:
2696         kfree(net->ipv6.ip6_prohibit_entry);
2697 out_ip6_null_entry:
2698         kfree(net->ipv6.ip6_null_entry);
2699 #endif
2700 out_ip6_dst_ops:
2701         goto out;
2702 }
2703
2704 static void __net_exit ip6_route_net_exit(struct net *net)
2705 {
2706 #ifdef CONFIG_PROC_FS
2707         proc_net_remove(net, "ipv6_route");
2708         proc_net_remove(net, "rt6_stats");
2709 #endif
2710         kfree(net->ipv6.ip6_null_entry);
2711 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2712         kfree(net->ipv6.ip6_prohibit_entry);
2713         kfree(net->ipv6.ip6_blk_hole_entry);
2714 #endif
2715 }
2716
2717 static struct pernet_operations ip6_route_net_ops = {
2718         .init = ip6_route_net_init,
2719         .exit = ip6_route_net_exit,
2720 };
2721
2722 static struct notifier_block ip6_route_dev_notifier = {
2723         .notifier_call = ip6_route_dev_notify,
2724         .priority = 0,
2725 };
2726
2727 int __init ip6_route_init(void)
2728 {
2729         int ret;
2730
2731         ret = -ENOMEM;
2732         ip6_dst_ops_template.kmem_cachep =
2733                 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
2734                                   SLAB_HWCACHE_ALIGN, NULL);
2735         if (!ip6_dst_ops_template.kmem_cachep)
2736                 goto out;
2737
2738         ret = register_pernet_subsys(&ip6_route_net_ops);
2739         if (ret)
2740                 goto out_kmem_cache;
2741
2742         ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
2743
2744         /* Registering of the loopback is done before this portion of code,
2745          * the loopback reference in rt6_info will not be taken, do it
2746          * manually for init_net */
2747         init_net.ipv6.ip6_null_entry->u.dst.dev = init_net.loopback_dev;
2748         init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
2749   #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2750         init_net.ipv6.ip6_prohibit_entry->u.dst.dev = init_net.loopback_dev;
2751         init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
2752         init_net.ipv6.ip6_blk_hole_entry->u.dst.dev = init_net.loopback_dev;
2753         init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
2754   #endif
2755         ret = fib6_init();
2756         if (ret)
2757                 goto out_register_subsys;
2758
2759         ret = xfrm6_init();
2760         if (ret)
2761                 goto out_fib6_init;
2762
2763         ret = fib6_rules_init();
2764         if (ret)
2765                 goto xfrm6_init;
2766
2767         ret = -ENOBUFS;
2768         if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL) ||
2769             __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL) ||
2770             __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL))
2771                 goto fib6_rules_init;
2772
2773         ret = register_netdevice_notifier(&ip6_route_dev_notifier);
2774         if (ret)
2775                 goto fib6_rules_init;
2776
2777 out:
2778         return ret;
2779
2780 fib6_rules_init:
2781         fib6_rules_cleanup();
2782 xfrm6_init:
2783         xfrm6_fini();
2784 out_fib6_init:
2785         fib6_gc_cleanup();
2786 out_register_subsys:
2787         unregister_pernet_subsys(&ip6_route_net_ops);
2788 out_kmem_cache:
2789         kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
2790         goto out;
2791 }
2792
2793 void ip6_route_cleanup(void)
2794 {
2795         unregister_netdevice_notifier(&ip6_route_dev_notifier);
2796         fib6_rules_cleanup();
2797         xfrm6_fini();
2798         fib6_gc_cleanup();
2799         unregister_pernet_subsys(&ip6_route_net_ops);
2800         kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
2801 }