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  *      $Id: route.c,v 1.56 2001/10/31 21:55:55 davem Exp $
   9  *
  10  *      This program is free software; you can redistribute it and/or
  11  *      modify it under the terms of the GNU General Public License
  12  *      as published by the Free Software Foundation; either version
  13  *      2 of the License, or (at your option) any later version.
  14  */
  15
  16 /*      Changes:
  17  *
  18  *      YOSHIFUJI Hideaki @USAGI
  19  *              reworked default router selection.
  20  *              - respect outgoing interface
  21  *              - select from (probably) reachable routers (i.e.
  22  *              routers in REACHABLE, STALE, DELAY or PROBE states).
  23  *              - always select the same router if it is (probably)
  24  *              reachable.  otherwise, round-robin the list.
  25  *      Ville Nuorvala
  26  *              Fixed routing subtrees.
  27  */
  28
  29 #include <linux/capability.h>
  30 #include <linux/errno.h>
  31 #include <linux/types.h>
  32 #include <linux/times.h>
  33 #include <linux/socket.h>
  34 #include <linux/sockios.h>
  35 #include <linux/net.h>
  36 #include <linux/route.h>
  37 #include <linux/netdevice.h>
  38 #include <linux/in6.h>
  39 #include <linux/init.h>
  40 #include <linux/if_arp.h>
  41 #include <linux/proc_fs.h>
  42 #include <linux/seq_file.h>
  43 #include <net/net_namespace.h>
  44 #include <net/snmp.h>
  45 #include <net/ipv6.h>
  46 #include <net/ip6_fib.h>
  47 #include <net/ip6_route.h>
  48 #include <net/ndisc.h>
  49 #include <net/addrconf.h>
  50 #include <net/tcp.h>
  51 #include <linux/rtnetlink.h>
  52 #include <net/dst.h>
  53 #include <net/xfrm.h>
  54 #include <net/netevent.h>
  55 #include <net/netlink.h>
  56
  57 #include <asm/uaccess.h>
  58
  59 #ifdef CONFIG_SYSCTL
  60 #include <linux/sysctl.h>
  61 #endif
  62
  63 /* Set to 3 to get tracing. */
  64 #define RT6_DEBUG 2
  65
  66 #if RT6_DEBUG >= 3
  67 #define RDBG(x) printk x
  68 #define RT6_TRACE(x...) printk(KERN_DEBUG x)
  69 #else
  70 #define RDBG(x)
  71 #define RT6_TRACE(x...) do { ; } while (0)
  72 #endif
  73
  74 #define CLONE_OFFLINK_ROUTE 0
  75
  76 static struct rt6_info * ip6_rt_copy(struct rt6_info *ort);
  77 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie);
  78 static struct dst_entry *ip6_negative_advice(struct dst_entry *);
  79 static void             ip6_dst_destroy(struct dst_entry *);
  80 static void             ip6_dst_ifdown(struct dst_entry *,
  81                                        struct net_device *dev, int how);
  82 static int               ip6_dst_gc(struct dst_ops *ops);
  83
  84 static int              ip6_pkt_discard(struct sk_buff *skb);
  85 static int              ip6_pkt_discard_out(struct sk_buff *skb);
  86 static void             ip6_link_failure(struct sk_buff *skb);
  87 static void             ip6_rt_update_pmtu(struct dst_entry *dst, u32 mtu);
  88
  89 #ifdef CONFIG_IPV6_ROUTE_INFO
  90 static struct rt6_info *rt6_add_route_info(struct in6_addr *prefix, int prefixlen,
  91                                            struct in6_addr *gwaddr, int ifindex,
  92                                            unsigned pref);
  93 static struct rt6_info *rt6_get_route_info(struct in6_addr *prefix, int prefixlen,
  94                                            struct in6_addr *gwaddr, int ifindex);
  95 #endif
  96
  97 static struct dst_ops ip6_dst_ops = {
  98         .family                 =       AF_INET6,
  99         .protocol               =       __constant_htons(ETH_P_IPV6),
 100         .gc                     =       ip6_dst_gc,
 101         .gc_thresh              =       1024,
 102         .check                  =       ip6_dst_check,
 103         .destroy                =       ip6_dst_destroy,
 104         .ifdown                 =       ip6_dst_ifdown,
 105         .negative_advice        =       ip6_negative_advice,
 106         .link_failure           =       ip6_link_failure,
 107         .update_pmtu            =       ip6_rt_update_pmtu,
 108         .local_out              =       ip6_local_out,
 109         .entry_size             =       sizeof(struct rt6_info),
 110         .entries                =       ATOMIC_INIT(0),
 111 };
 112
 113 static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, u32 mtu)
 114 {
 115 }
 116
 117 static struct dst_ops ip6_dst_blackhole_ops = {
 118         .family                 =       AF_INET6,
 119         .protocol               =       __constant_htons(ETH_P_IPV6),
 120         .destroy                =       ip6_dst_destroy,
 121         .check                  =       ip6_dst_check,
 122         .update_pmtu            =       ip6_rt_blackhole_update_pmtu,
 123         .entry_size             =       sizeof(struct rt6_info),
 124         .entries                =       ATOMIC_INIT(0),
 125 };
 126
 127 struct rt6_info ip6_null_entry = {
 128         .u = {
 129                 .dst = {
 130                         .__refcnt       = ATOMIC_INIT(1),
 131                         .__use          = 1,
 132                         .obsolete       = -1,
 133                         .error          = -ENETUNREACH,
 134                         .metrics        = { [RTAX_HOPLIMIT - 1] = 255, },
 135                         .input          = ip6_pkt_discard,
 136                         .output         = ip6_pkt_discard_out,
 137                         .ops            = &ip6_dst_ops,
 138                         .path           = (struct dst_entry*)&ip6_null_entry,
 139                 }
 140         },
 141         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
 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 struct rt6_info ip6_prohibit_entry = {
 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                         .ops            = &ip6_dst_ops,
 162                         .path           = (struct dst_entry*)&ip6_prohibit_entry,
 163                 }
 164         },
 165         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
 166         .rt6i_metric    = ~(u32) 0,
 167         .rt6i_ref       = ATOMIC_INIT(1),
 168 };
 169
 170 struct rt6_info ip6_blk_hole_entry = {
 171         .u = {
 172                 .dst = {
 173                         .__refcnt       = ATOMIC_INIT(1),
 174                         .__use          = 1,
 175                         .obsolete       = -1,
 176                         .error          = -EINVAL,
 177                         .metrics        = { [RTAX_HOPLIMIT - 1] = 255, },
 178                         .input          = dst_discard,
 179                         .output         = dst_discard,
 180                         .ops            = &ip6_dst_ops,
 181                         .path           = (struct dst_entry*)&ip6_blk_hole_entry,
 182                 }
 183         },
 184         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
 185         .rt6i_metric    = ~(u32) 0,
 186         .rt6i_ref       = ATOMIC_INIT(1),
 187 };
 188
 189 #endif
 190
 191 /* allocate dst with ip6_dst_ops */
 192 static __inline__ struct rt6_info *ip6_dst_alloc(void)
 193 {
 194         return (struct rt6_info *)dst_alloc(&ip6_dst_ops);
 195 }
 196
 197 static void ip6_dst_destroy(struct dst_entry *dst)
 198 {
 199         struct rt6_info *rt = (struct rt6_info *)dst;
 200         struct inet6_dev *idev = rt->rt6i_idev;
 201
 202         if (idev != NULL) {
 203                 rt->rt6i_idev = NULL;
 204                 in6_dev_put(idev);
 205         }
 206 }
 207
 208 static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
 209                            int how)
 210 {
 211         struct rt6_info *rt = (struct rt6_info *)dst;
 212         struct inet6_dev *idev = rt->rt6i_idev;
 213         struct net_device *loopback_dev =
 214                 dev->nd_net->loopback_dev;
 215
 216         if (dev != loopback_dev && idev != NULL && idev->dev == dev) {
 217                 struct inet6_dev *loopback_idev =
 218                         in6_dev_get(loopback_dev);
 219                 if (loopback_idev != NULL) {
 220                         rt->rt6i_idev = loopback_idev;
 221                         in6_dev_put(idev);
 222                 }
 223         }
 224 }
 225
 226 static __inline__ int rt6_check_expired(const struct rt6_info *rt)
 227 {
 228         return (rt->rt6i_flags & RTF_EXPIRES &&
 229                 time_after(jiffies, rt->rt6i_expires));
 230 }
 231
 232 static inline int rt6_need_strict(struct in6_addr *daddr)
 233 {
 234         return (ipv6_addr_type(daddr) &
 235                 (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL));
 236 }
 237
 238 /*
 239  *      Route lookup. Any table->tb6_lock is implied.
 240  */
 241
 242 static __inline__ struct rt6_info *rt6_device_match(struct rt6_info *rt,
 243                                                     int oif,
 244                                                     int strict)
 245 {
 246         struct rt6_info *local = NULL;
 247         struct rt6_info *sprt;
 248
 249         if (oif) {
 250                 for (sprt = rt; sprt; sprt = sprt->u.dst.rt6_next) {
 251                         struct net_device *dev = sprt->rt6i_dev;
 252                         if (dev->ifindex == oif)
 253                                 return sprt;
 254                         if (dev->flags & IFF_LOOPBACK) {
 255                                 if (sprt->rt6i_idev == NULL ||
 256                                     sprt->rt6i_idev->dev->ifindex != oif) {
 257                                         if (strict && oif)
 258                                                 continue;
 259                                         if (local && (!oif ||
 260                                                       local->rt6i_idev->dev->ifindex == oif))
 261                                                 continue;
 262                                 }
 263                                 local = sprt;
 264                         }
 265                 }
 266
 267                 if (local)
 268                         return local;
 269
 270                 if (strict)
 271                         return &ip6_null_entry;
 272         }
 273         return rt;
 274 }
 275
 276 #ifdef CONFIG_IPV6_ROUTER_PREF
 277 static void rt6_probe(struct rt6_info *rt)
 278 {
 279         struct neighbour *neigh = rt ? rt->rt6i_nexthop : NULL;
 280         /*
 281          * Okay, this does not seem to be appropriate
 282          * for now, however, we need to check if it
 283          * is really so; aka Router Reachability Probing.
 284          *
 285          * Router Reachability Probe MUST be rate-limited
 286          * to no more than one per minute.
 287          */
 288         if (!neigh || (neigh->nud_state & NUD_VALID))
 289                 return;
 290         read_lock_bh(&neigh->lock);
 291         if (!(neigh->nud_state & NUD_VALID) &&
 292             time_after(jiffies, neigh->updated + rt->rt6i_idev->cnf.rtr_probe_interval)) {
 293                 struct in6_addr mcaddr;
 294                 struct in6_addr *target;
 295
 296                 neigh->updated = jiffies;
 297                 read_unlock_bh(&neigh->lock);
 298
 299                 target = (struct in6_addr *)&neigh->primary_key;
 300                 addrconf_addr_solict_mult(target, &mcaddr);
 301                 ndisc_send_ns(rt->rt6i_dev, NULL, target, &mcaddr, NULL);
 302         } else
 303                 read_unlock_bh(&neigh->lock);
 304 }
 305 #else
 306 static inline void rt6_probe(struct rt6_info *rt)
 307 {
 308         return;
 309 }
 310 #endif
 311
 312 /*
 313  * Default Router Selection (RFC 2461 6.3.6)
 314  */
 315 static inline int rt6_check_dev(struct rt6_info *rt, int oif)
 316 {
 317         struct net_device *dev = rt->rt6i_dev;
 318         if (!oif || dev->ifindex == oif)
 319                 return 2;
 320         if ((dev->flags & IFF_LOOPBACK) &&
 321             rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
 322                 return 1;
 323         return 0;
 324 }
 325
 326 static inline int rt6_check_neigh(struct rt6_info *rt)
 327 {
 328         struct neighbour *neigh = rt->rt6i_nexthop;
 329         int m;
 330         if (rt->rt6i_flags & RTF_NONEXTHOP ||
 331             !(rt->rt6i_flags & RTF_GATEWAY))
 332                 m = 1;
 333         else if (neigh) {
 334                 read_lock_bh(&neigh->lock);
 335                 if (neigh->nud_state & NUD_VALID)
 336                         m = 2;
 337 #ifdef CONFIG_IPV6_ROUTER_PREF
 338                 else if (neigh->nud_state & NUD_FAILED)
 339                         m = 0;
 340 #endif
 341                 else
 342                         m = 1;
 343                 read_unlock_bh(&neigh->lock);
 344         } else
 345                 m = 0;
 346         return m;
 347 }
 348
 349 static int rt6_score_route(struct rt6_info *rt, int oif,
 350                            int strict)
 351 {
 352         int m, n;
 353
 354         m = rt6_check_dev(rt, oif);
 355         if (!m && (strict & RT6_LOOKUP_F_IFACE))
 356                 return -1;
 357 #ifdef CONFIG_IPV6_ROUTER_PREF
 358         m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
 359 #endif
 360         n = rt6_check_neigh(rt);
 361         if (!n && (strict & RT6_LOOKUP_F_REACHABLE))
 362                 return -1;
 363         return m;
 364 }
 365
 366 static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
 367                                    int *mpri, struct rt6_info *match)
 368 {
 369         int m;
 370
 371         if (rt6_check_expired(rt))
 372                 goto out;
 373
 374         m = rt6_score_route(rt, oif, strict);
 375         if (m < 0)
 376                 goto out;
 377
 378         if (m > *mpri) {
 379                 if (strict & RT6_LOOKUP_F_REACHABLE)
 380                         rt6_probe(match);
 381                 *mpri = m;
 382                 match = rt;
 383         } else if (strict & RT6_LOOKUP_F_REACHABLE) {
 384                 rt6_probe(rt);
 385         }
 386
 387 out:
 388         return match;
 389 }
 390
 391 static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
 392                                      struct rt6_info *rr_head,
 393                                      u32 metric, int oif, int strict)
 394 {
 395         struct rt6_info *rt, *match;
 396         int mpri = -1;
 397
 398         match = NULL;
 399         for (rt = rr_head; rt && rt->rt6i_metric == metric;
 400              rt = rt->u.dst.rt6_next)
 401                 match = find_match(rt, oif, strict, &mpri, match);
 402         for (rt = fn->leaf; rt && rt != rr_head && rt->rt6i_metric == metric;
 403              rt = rt->u.dst.rt6_next)
 404                 match = find_match(rt, oif, strict, &mpri, match);
 405
 406         return match;
 407 }
 408
 409 static struct rt6_info *rt6_select(struct fib6_node *fn, int oif, int strict)
 410 {
 411         struct rt6_info *match, *rt0;
 412
 413         RT6_TRACE("%s(fn->leaf=%p, oif=%d)\n",
 414                   __FUNCTION__, fn->leaf, oif);
 415
 416         rt0 = fn->rr_ptr;
 417         if (!rt0)
 418                 fn->rr_ptr = rt0 = fn->leaf;
 419
 420         match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict);
 421
 422         if (!match &&
 423             (strict & RT6_LOOKUP_F_REACHABLE)) {
 424                 struct rt6_info *next = rt0->u.dst.rt6_next;
 425
 426                 /* no entries matched; do round-robin */
 427                 if (!next || next->rt6i_metric != rt0->rt6i_metric)
 428                         next = fn->leaf;
 429
 430                 if (next != rt0)
 431                         fn->rr_ptr = next;
 432         }
 433
 434         RT6_TRACE("%s() => %p\n",
 435                   __FUNCTION__, match);
 436
 437         return (match ? match : &ip6_null_entry);
 438 }
 439
 440 #ifdef CONFIG_IPV6_ROUTE_INFO
 441 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
 442                   struct in6_addr *gwaddr)
 443 {
 444         struct route_info *rinfo = (struct route_info *) opt;
 445         struct in6_addr prefix_buf, *prefix;
 446         unsigned int pref;
 447         u32 lifetime;
 448         struct rt6_info *rt;
 449
 450         if (len < sizeof(struct route_info)) {
 451                 return -EINVAL;
 452         }
 453
 454         /* Sanity check for prefix_len and length */
 455         if (rinfo->length > 3) {
 456                 return -EINVAL;
 457         } else if (rinfo->prefix_len > 128) {
 458                 return -EINVAL;
 459         } else if (rinfo->prefix_len > 64) {
 460                 if (rinfo->length < 2) {
 461                         return -EINVAL;
 462                 }
 463         } else if (rinfo->prefix_len > 0) {
 464                 if (rinfo->length < 1) {
 465                         return -EINVAL;
 466                 }
 467         }
 468
 469         pref = rinfo->route_pref;
 470         if (pref == ICMPV6_ROUTER_PREF_INVALID)
 471                 pref = ICMPV6_ROUTER_PREF_MEDIUM;
 472
 473         lifetime = ntohl(rinfo->lifetime);
 474         if (lifetime == 0xffffffff) {
 475                 /* infinity */
 476         } else if (lifetime > 0x7fffffff/HZ) {
 477                 /* Avoid arithmetic overflow */
 478                 lifetime = 0x7fffffff/HZ - 1;
 479         }
 480
 481         if (rinfo->length == 3)
 482                 prefix = (struct in6_addr *)rinfo->prefix;
 483         else {
 484                 /* this function is safe */
 485                 ipv6_addr_prefix(&prefix_buf,
 486                                  (struct in6_addr *)rinfo->prefix,
 487                                  rinfo->prefix_len);
 488                 prefix = &prefix_buf;
 489         }
 490
 491         rt = rt6_get_route_info(prefix, rinfo->prefix_len, gwaddr, dev->ifindex);
 492
 493         if (rt && !lifetime) {
 494                 ip6_del_rt(rt);
 495                 rt = NULL;
 496         }
 497
 498         if (!rt && lifetime)
 499                 rt = rt6_add_route_info(prefix, rinfo->prefix_len, gwaddr, dev->ifindex,
 500                                         pref);
 501         else if (rt)
 502                 rt->rt6i_flags = RTF_ROUTEINFO |
 503                                  (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
 504
 505         if (rt) {
 506                 if (lifetime == 0xffffffff) {
 507                         rt->rt6i_flags &= ~RTF_EXPIRES;
 508                 } else {
 509                         rt->rt6i_expires = jiffies + HZ * lifetime;
 510                         rt->rt6i_flags |= RTF_EXPIRES;
 511                 }
 512                 dst_release(&rt->u.dst);
 513         }
 514         return 0;
 515 }
 516 #endif
 517
 518 #define BACKTRACK(saddr) \
 519 do { \
 520         if (rt == &ip6_null_entry) { \
 521                 struct fib6_node *pn; \
 522                 while (1) { \
 523                         if (fn->fn_flags & RTN_TL_ROOT) \
 524                                 goto out; \
 525                         pn = fn->parent; \
 526                         if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn) \
 527                                 fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr); \
 528                         else \
 529                                 fn = pn; \
 530                         if (fn->fn_flags & RTN_RTINFO) \
 531                                 goto restart; \
 532                 } \
 533         } \
 534 } while(0)
 535
 536 static struct rt6_info *ip6_pol_route_lookup(struct fib6_table *table,
 537                                              struct flowi *fl, int flags)
 538 {
 539         struct fib6_node *fn;
 540         struct rt6_info *rt;
 541
 542         read_lock_bh(&table->tb6_lock);
 543         fn = fib6_lookup(&table->tb6_root, &fl->fl6_dst, &fl->fl6_src);
 544 restart:
 545         rt = fn->leaf;
 546         rt = rt6_device_match(rt, fl->oif, flags);
 547         BACKTRACK(&fl->fl6_src);
 548 out:
 549         dst_use(&rt->u.dst, jiffies);
 550         read_unlock_bh(&table->tb6_lock);
 551         return rt;
 552
 553 }
 554
 555 struct rt6_info *rt6_lookup(struct in6_addr *daddr, struct in6_addr *saddr,
 556                             int oif, int strict)
 557 {
 558         struct flowi fl = {
 559                 .oif = oif,
 560                 .nl_u = {
 561                         .ip6_u = {
 562                                 .daddr = *daddr,
 563                         },
 564                 },
 565         };
 566         struct dst_entry *dst;
 567         int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
 568
 569         if (saddr) {
 570                 memcpy(&fl.fl6_src, saddr, sizeof(*saddr));
 571                 flags |= RT6_LOOKUP_F_HAS_SADDR;
 572         }
 573
 574         dst = fib6_rule_lookup(&fl, flags, ip6_pol_route_lookup);
 575         if (dst->error == 0)
 576                 return (struct rt6_info *) dst;
 577
 578         dst_release(dst);
 579
 580         return NULL;
 581 }
 582
 583 EXPORT_SYMBOL(rt6_lookup);
 584
 585 /* ip6_ins_rt is called with FREE table->tb6_lock.
 586    It takes new route entry, the addition fails by any reason the
 587    route is freed. In any case, if caller does not hold it, it may
 588    be destroyed.
 589  */
 590
 591 static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info)
 592 {
 593         int err;
 594         struct fib6_table *table;
 595
 596         table = rt->rt6i_table;
 597         write_lock_bh(&table->tb6_lock);
 598         err = fib6_add(&table->tb6_root, rt, info);
 599         write_unlock_bh(&table->tb6_lock);
 600
 601         return err;
 602 }
 603
 604 int ip6_ins_rt(struct rt6_info *rt)
 605 {
 606         struct nl_info info = {
 607                 .nl_net = &init_net,
 608         };
 609         return __ip6_ins_rt(rt, &info);
 610 }
 611
 612 static struct rt6_info *rt6_alloc_cow(struct rt6_info *ort, struct in6_addr *daddr,
 613                                       struct in6_addr *saddr)
 614 {
 615         struct rt6_info *rt;
 616
 617         /*
 618          *      Clone the route.
 619          */
 620
 621         rt = ip6_rt_copy(ort);
 622
 623         if (rt) {
 624                 if (!(rt->rt6i_flags&RTF_GATEWAY)) {
 625                         if (rt->rt6i_dst.plen != 128 &&
 626                             ipv6_addr_equal(&rt->rt6i_dst.addr, daddr))
 627                                 rt->rt6i_flags |= RTF_ANYCAST;
 628                         ipv6_addr_copy(&rt->rt6i_gateway, daddr);
 629                 }
 630
 631                 ipv6_addr_copy(&rt->rt6i_dst.addr, daddr);
 632                 rt->rt6i_dst.plen = 128;
 633                 rt->rt6i_flags |= RTF_CACHE;
 634                 rt->u.dst.flags |= DST_HOST;
 635
 636 #ifdef CONFIG_IPV6_SUBTREES
 637                 if (rt->rt6i_src.plen && saddr) {
 638                         ipv6_addr_copy(&rt->rt6i_src.addr, saddr);
 639                         rt->rt6i_src.plen = 128;
 640                 }
 641 #endif
 642
 643                 rt->rt6i_nexthop = ndisc_get_neigh(rt->rt6i_dev, &rt->rt6i_gateway);
 644
 645         }
 646
 647         return rt;
 648 }
 649
 650 static struct rt6_info *rt6_alloc_clone(struct rt6_info *ort, struct in6_addr *daddr)
 651 {
 652         struct rt6_info *rt = ip6_rt_copy(ort);
 653         if (rt) {
 654                 ipv6_addr_copy(&rt->rt6i_dst.addr, daddr);
 655                 rt->rt6i_dst.plen = 128;
 656                 rt->rt6i_flags |= RTF_CACHE;
 657                 rt->u.dst.flags |= DST_HOST;
 658                 rt->rt6i_nexthop = neigh_clone(ort->rt6i_nexthop);
 659         }
 660         return rt;
 661 }
 662
 663 static struct rt6_info *ip6_pol_route(struct fib6_table *table, int oif,
 664                                             struct flowi *fl, int flags)
 665 {
 666         struct fib6_node *fn;
 667         struct rt6_info *rt, *nrt;
 668         int strict = 0;
 669         int attempts = 3;
 670         int err;
 671         int reachable = ipv6_devconf.forwarding ? 0 : RT6_LOOKUP_F_REACHABLE;
 672
 673         strict |= flags & RT6_LOOKUP_F_IFACE;
 674
 675 relookup:
 676         read_lock_bh(&table->tb6_lock);
 677
 678 restart_2:
 679         fn = fib6_lookup(&table->tb6_root, &fl->fl6_dst, &fl->fl6_src);
 680
 681 restart:
 682         rt = rt6_select(fn, oif, strict | reachable);
 683         BACKTRACK(&fl->fl6_src);
 684         if (rt == &ip6_null_entry ||
 685             rt->rt6i_flags & RTF_CACHE)
 686                 goto out;
 687
 688         dst_hold(&rt->u.dst);
 689         read_unlock_bh(&table->tb6_lock);
 690
 691         if (!rt->rt6i_nexthop && !(rt->rt6i_flags & RTF_NONEXTHOP))
 692                 nrt = rt6_alloc_cow(rt, &fl->fl6_dst, &fl->fl6_src);
 693         else {
 694 #if CLONE_OFFLINK_ROUTE
 695                 nrt = rt6_alloc_clone(rt, &fl->fl6_dst);
 696 #else
 697                 goto out2;
 698 #endif
 699         }
 700
 701         dst_release(&rt->u.dst);
 702         rt = nrt ? : &ip6_null_entry;
 703
 704         dst_hold(&rt->u.dst);
 705         if (nrt) {
 706                 err = ip6_ins_rt(nrt);
 707                 if (!err)
 708                         goto out2;
 709         }
 710
 711         if (--attempts <= 0)
 712                 goto out2;
 713
 714         /*
 715          * Race condition! In the gap, when table->tb6_lock was
 716          * released someone could insert this route.  Relookup.
 717          */
 718         dst_release(&rt->u.dst);
 719         goto relookup;
 720
 721 out:
 722         if (reachable) {
 723                 reachable = 0;
 724                 goto restart_2;
 725         }
 726         dst_hold(&rt->u.dst);
 727         read_unlock_bh(&table->tb6_lock);
 728 out2:
 729         rt->u.dst.lastuse = jiffies;
 730         rt->u.dst.__use++;
 731
 732         return rt;
 733 }
 734
 735 static struct rt6_info *ip6_pol_route_input(struct fib6_table *table,
 736                                             struct flowi *fl, int flags)
 737 {
 738         return ip6_pol_route(table, fl->iif, fl, flags);
 739 }
 740
 741 void ip6_route_input(struct sk_buff *skb)
 742 {
 743         struct ipv6hdr *iph = ipv6_hdr(skb);
 744         int flags = RT6_LOOKUP_F_HAS_SADDR;
 745         struct flowi fl = {
 746                 .iif = skb->dev->ifindex,
 747                 .nl_u = {
 748                         .ip6_u = {
 749                                 .daddr = iph->daddr,
 750                                 .saddr = iph->saddr,
 751                                 .flowlabel = (* (__be32 *) iph)&IPV6_FLOWINFO_MASK,
 752                         },
 753                 },
 754                 .mark = skb->mark,
 755                 .proto = iph->nexthdr,
 756         };
 757
 758         if (rt6_need_strict(&iph->daddr))
 759                 flags |= RT6_LOOKUP_F_IFACE;
 760
 761         skb->dst = fib6_rule_lookup(&fl, flags, ip6_pol_route_input);
 762 }
 763
 764 static struct rt6_info *ip6_pol_route_output(struct fib6_table *table,
 765                                              struct flowi *fl, int flags)
 766 {
 767         return ip6_pol_route(table, fl->oif, fl, flags);
 768 }
 769
 770 struct dst_entry * ip6_route_output(struct sock *sk, struct flowi *fl)
 771 {
 772         int flags = 0;
 773
 774         if (rt6_need_strict(&fl->fl6_dst))
 775                 flags |= RT6_LOOKUP_F_IFACE;
 776
 777         if (!ipv6_addr_any(&fl->fl6_src))
 778                 flags |= RT6_LOOKUP_F_HAS_SADDR;
 779
 780         return fib6_rule_lookup(fl, flags, ip6_pol_route_output);
 781 }
 782
 783 EXPORT_SYMBOL(ip6_route_output);
 784
 785 int ip6_dst_blackhole(struct sock *sk, struct dst_entry **dstp, struct flowi *fl)
 786 {
 787         struct rt6_info *ort = (struct rt6_info *) *dstp;
 788         struct rt6_info *rt = (struct rt6_info *)
 789                 dst_alloc(&ip6_dst_blackhole_ops);
 790         struct dst_entry *new = NULL;
 791
 792         if (rt) {
 793                 new = &rt->u.dst;
 794
 795                 atomic_set(&new->__refcnt, 1);
 796                 new->__use = 1;
 797                 new->input = dst_discard;
 798                 new->output = dst_discard;
 799
 800                 memcpy(new->metrics, ort->u.dst.metrics, RTAX_MAX*sizeof(u32));
 801                 new->dev = ort->u.dst.dev;
 802                 if (new->dev)
 803                         dev_hold(new->dev);
 804                 rt->rt6i_idev = ort->rt6i_idev;
 805                 if (rt->rt6i_idev)
 806                         in6_dev_hold(rt->rt6i_idev);
 807                 rt->rt6i_expires = 0;
 808
 809                 ipv6_addr_copy(&rt->rt6i_gateway, &ort->rt6i_gateway);
 810                 rt->rt6i_flags = ort->rt6i_flags & ~RTF_EXPIRES;
 811                 rt->rt6i_metric = 0;
 812
 813                 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
 814 #ifdef CONFIG_IPV6_SUBTREES
 815                 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
 816 #endif
 817
 818                 dst_free(new);
 819         }
 820
 821         dst_release(*dstp);
 822         *dstp = new;
 823         return (new ? 0 : -ENOMEM);
 824 }
 825 EXPORT_SYMBOL_GPL(ip6_dst_blackhole);
 826
 827 /*
 828  *      Destination cache support functions
 829  */
 830
 831 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
 832 {
 833         struct rt6_info *rt;
 834
 835         rt = (struct rt6_info *) dst;
 836
 837         if (rt && rt->rt6i_node && (rt->rt6i_node->fn_sernum == cookie))
 838                 return dst;
 839
 840         return NULL;
 841 }
 842
 843 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
 844 {
 845         struct rt6_info *rt = (struct rt6_info *) dst;
 846
 847         if (rt) {
 848                 if (rt->rt6i_flags & RTF_CACHE)
 849                         ip6_del_rt(rt);
 850                 else
 851                         dst_release(dst);
 852         }
 853         return NULL;
 854 }
 855
 856 static void ip6_link_failure(struct sk_buff *skb)
 857 {
 858         struct rt6_info *rt;
 859
 860         icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0, skb->dev);
 861
 862         rt = (struct rt6_info *) skb->dst;
 863         if (rt) {
 864                 if (rt->rt6i_flags&RTF_CACHE) {
 865                         dst_set_expires(&rt->u.dst, 0);
 866                         rt->rt6i_flags |= RTF_EXPIRES;
 867                 } else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT))
 868                         rt->rt6i_node->fn_sernum = -1;
 869         }
 870 }
 871
 872 static void ip6_rt_update_pmtu(struct dst_entry *dst, u32 mtu)
 873 {
 874         struct rt6_info *rt6 = (struct rt6_info*)dst;
 875
 876         if (mtu < dst_mtu(dst) && rt6->rt6i_dst.plen == 128) {
 877                 rt6->rt6i_flags |= RTF_MODIFIED;
 878                 if (mtu < IPV6_MIN_MTU) {
 879                         mtu = IPV6_MIN_MTU;
 880                         dst->metrics[RTAX_FEATURES-1] |= RTAX_FEATURE_ALLFRAG;
 881                 }
 882                 dst->metrics[RTAX_MTU-1] = mtu;
 883                 call_netevent_notifiers(NETEVENT_PMTU_UPDATE, dst);
 884         }
 885 }
 886
 887 static int ipv6_get_mtu(struct net_device *dev);
 888
 889 static inline unsigned int ipv6_advmss(unsigned int mtu)
 890 {
 891         mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
 892
 893         if (mtu < init_net.ipv6.sysctl.ip6_rt_min_advmss)
 894                 mtu = init_net.ipv6.sysctl.ip6_rt_min_advmss;
 895
 896         /*
 897          * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
 898          * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
 899          * IPV6_MAXPLEN is also valid and means: "any MSS,
 900          * rely only on pmtu discovery"
 901          */
 902         if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
 903                 mtu = IPV6_MAXPLEN;
 904         return mtu;
 905 }
 906
 907 static struct dst_entry *icmp6_dst_gc_list;
 908 static DEFINE_SPINLOCK(icmp6_dst_lock);
 909
 910 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
 911                                   struct neighbour *neigh,
 912                                   struct in6_addr *addr)
 913 {
 914         struct rt6_info *rt;
 915         struct inet6_dev *idev = in6_dev_get(dev);
 916
 917         if (unlikely(idev == NULL))
 918                 return NULL;
 919
 920         rt = ip6_dst_alloc();
 921         if (unlikely(rt == NULL)) {
 922                 in6_dev_put(idev);
 923                 goto out;
 924         }
 925
 926         dev_hold(dev);
 927         if (neigh)
 928                 neigh_hold(neigh);
 929         else
 930                 neigh = ndisc_get_neigh(dev, addr);
 931
 932         rt->rt6i_dev      = dev;
 933         rt->rt6i_idev     = idev;
 934         rt->rt6i_nexthop  = neigh;
 935         atomic_set(&rt->u.dst.__refcnt, 1);
 936         rt->u.dst.metrics[RTAX_HOPLIMIT-1] = 255;
 937         rt->u.dst.metrics[RTAX_MTU-1] = ipv6_get_mtu(rt->rt6i_dev);
 938         rt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(dst_mtu(&rt->u.dst));
 939         rt->u.dst.output  = ip6_output;
 940
 941 #if 0   /* there's no chance to use these for ndisc */
 942         rt->u.dst.flags   = ipv6_addr_type(addr) & IPV6_ADDR_UNICAST
 943                                 ? DST_HOST
 944                                 : 0;
 945         ipv6_addr_copy(&rt->rt6i_dst.addr, addr);
 946         rt->rt6i_dst.plen = 128;
 947 #endif
 948
 949         spin_lock_bh(&icmp6_dst_lock);
 950         rt->u.dst.next = icmp6_dst_gc_list;
 951         icmp6_dst_gc_list = &rt->u.dst;
 952         spin_unlock_bh(&icmp6_dst_lock);
 953
 954         fib6_force_start_gc();
 955
 956 out:
 957         return &rt->u.dst;
 958 }
 959
 960 int icmp6_dst_gc(int *more)
 961 {
 962         struct dst_entry *dst, *next, **pprev;
 963         int freed;
 964
 965         next = NULL;
 966         freed = 0;
 967
 968         spin_lock_bh(&icmp6_dst_lock);
 969         pprev = &icmp6_dst_gc_list;
 970
 971         while ((dst = *pprev) != NULL) {
 972                 if (!atomic_read(&dst->__refcnt)) {
 973                         *pprev = dst->next;
 974                         dst_free(dst);
 975                         freed++;
 976                 } else {
 977                         pprev = &dst->next;
 978                         (*more)++;
 979                 }
 980         }
 981
 982         spin_unlock_bh(&icmp6_dst_lock);
 983
 984         return freed;
 985 }
 986
 987 static int ip6_dst_gc(struct dst_ops *ops)
 988 {
 989         static unsigned expire = 30*HZ;
 990         static unsigned long last_gc;
 991         unsigned long now = jiffies;
 992
 993         if (time_after(last_gc + init_net.ipv6.sysctl.ip6_rt_gc_min_interval, now) &&
 994             atomic_read(&ip6_dst_ops.entries) <= init_net.ipv6.sysctl.ip6_rt_max_size)
 995                 goto out;
 996
 997         expire++;
 998         fib6_run_gc(expire);
 999         last_gc = now;
1000         if (atomic_read(&ip6_dst_ops.entries) < ip6_dst_ops.gc_thresh)
1001                 expire = init_net.ipv6.sysctl.ip6_rt_gc_timeout>>1;
1002
1003 out:
1004         expire -= expire>>init_net.ipv6.sysctl.ip6_rt_gc_elasticity;
1005         return (atomic_read(&ip6_dst_ops.entries) > init_net.ipv6.sysctl.ip6_rt_max_size);
1006 }
1007
1008 /* Clean host part of a prefix. Not necessary in radix tree,
1009    but results in cleaner routing tables.
1010
1011    Remove it only when all the things will work!
1012  */
1013
1014 static int ipv6_get_mtu(struct net_device *dev)
1015 {
1016         int mtu = IPV6_MIN_MTU;
1017         struct inet6_dev *idev;
1018
1019         idev = in6_dev_get(dev);
1020         if (idev) {
1021                 mtu = idev->cnf.mtu6;
1022                 in6_dev_put(idev);
1023         }
1024         return mtu;
1025 }
1026
1027 int ipv6_get_hoplimit(struct net_device *dev)
1028 {
1029         int hoplimit = ipv6_devconf.hop_limit;
1030         struct inet6_dev *idev;
1031
1032         idev = in6_dev_get(dev);
1033         if (idev) {
1034                 hoplimit = idev->cnf.hop_limit;
1035                 in6_dev_put(idev);
1036         }
1037         return hoplimit;
1038 }
1039
1040 /*
1041  *
1042  */
1043
1044 int ip6_route_add(struct fib6_config *cfg)
1045 {
1046         int err;
1047         struct rt6_info *rt = NULL;
1048         struct net_device *dev = NULL;
1049         struct inet6_dev *idev = NULL;
1050         struct fib6_table *table;
1051         int addr_type;
1052
1053         if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1054                 return -EINVAL;
1055 #ifndef CONFIG_IPV6_SUBTREES
1056         if (cfg->fc_src_len)
1057                 return -EINVAL;
1058 #endif
1059         if (cfg->fc_ifindex) {
1060                 err = -ENODEV;
1061                 dev = dev_get_by_index(&init_net, cfg->fc_ifindex);
1062                 if (!dev)
1063                         goto out;
1064                 idev = in6_dev_get(dev);
1065                 if (!idev)
1066                         goto out;
1067         }
1068
1069         if (cfg->fc_metric == 0)
1070                 cfg->fc_metric = IP6_RT_PRIO_USER;
1071
1072         table = fib6_new_table(cfg->fc_table);
1073         if (table == NULL) {
1074                 err = -ENOBUFS;
1075                 goto out;
1076         }
1077
1078         rt = ip6_dst_alloc();
1079
1080         if (rt == NULL) {
1081                 err = -ENOMEM;
1082                 goto out;
1083         }
1084
1085         rt->u.dst.obsolete = -1;
1086         rt->rt6i_expires = jiffies + clock_t_to_jiffies(cfg->fc_expires);
1087
1088         if (cfg->fc_protocol == RTPROT_UNSPEC)
1089                 cfg->fc_protocol = RTPROT_BOOT;
1090         rt->rt6i_protocol = cfg->fc_protocol;
1091
1092         addr_type = ipv6_addr_type(&cfg->fc_dst);
1093
1094         if (addr_type & IPV6_ADDR_MULTICAST)
1095                 rt->u.dst.input = ip6_mc_input;
1096         else
1097                 rt->u.dst.input = ip6_forward;
1098
1099         rt->u.dst.output = ip6_output;
1100
1101         ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1102         rt->rt6i_dst.plen = cfg->fc_dst_len;
1103         if (rt->rt6i_dst.plen == 128)
1104                rt->u.dst.flags = DST_HOST;
1105
1106 #ifdef CONFIG_IPV6_SUBTREES
1107         ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1108         rt->rt6i_src.plen = cfg->fc_src_len;
1109 #endif
1110
1111         rt->rt6i_metric = cfg->fc_metric;
1112
1113         /* We cannot add true routes via loopback here,
1114            they would result in kernel looping; promote them to reject routes
1115          */
1116         if ((cfg->fc_flags & RTF_REJECT) ||
1117             (dev && (dev->flags&IFF_LOOPBACK) && !(addr_type&IPV6_ADDR_LOOPBACK))) {
1118                 /* hold loopback dev/idev if we haven't done so. */
1119                 if (dev != init_net.loopback_dev) {
1120                         if (dev) {
1121                                 dev_put(dev);
1122                                 in6_dev_put(idev);
1123                         }
1124                         dev = init_net.loopback_dev;
1125                         dev_hold(dev);
1126                         idev = in6_dev_get(dev);
1127                         if (!idev) {
1128                                 err = -ENODEV;
1129                                 goto out;
1130                         }
1131                 }
1132                 rt->u.dst.output = ip6_pkt_discard_out;
1133                 rt->u.dst.input = ip6_pkt_discard;
1134                 rt->u.dst.error = -ENETUNREACH;
1135                 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1136                 goto install_route;
1137         }
1138
1139         if (cfg->fc_flags & RTF_GATEWAY) {
1140                 struct in6_addr *gw_addr;
1141                 int gwa_type;
1142
1143                 gw_addr = &cfg->fc_gateway;
1144                 ipv6_addr_copy(&rt->rt6i_gateway, gw_addr);
1145                 gwa_type = ipv6_addr_type(gw_addr);
1146
1147                 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1148                         struct rt6_info *grt;
1149
1150                         /* IPv6 strictly inhibits using not link-local
1151                            addresses as nexthop address.
1152                            Otherwise, router will not able to send redirects.
1153                            It is very good, but in some (rare!) circumstances
1154                            (SIT, PtP, NBMA NOARP links) it is handy to allow
1155                            some exceptions. --ANK
1156                          */
1157                         err = -EINVAL;
1158                         if (!(gwa_type&IPV6_ADDR_UNICAST))
1159                                 goto out;
1160
1161                         grt = rt6_lookup(gw_addr, NULL, cfg->fc_ifindex, 1);
1162
1163                         err = -EHOSTUNREACH;
1164                         if (grt == NULL)
1165                                 goto out;
1166                         if (dev) {
1167                                 if (dev != grt->rt6i_dev) {
1168                                         dst_release(&grt->u.dst);
1169                                         goto out;
1170                                 }
1171                         } else {
1172                                 dev = grt->rt6i_dev;
1173                                 idev = grt->rt6i_idev;
1174                                 dev_hold(dev);
1175                                 in6_dev_hold(grt->rt6i_idev);
1176                         }
1177                         if (!(grt->rt6i_flags&RTF_GATEWAY))
1178                                 err = 0;
1179                         dst_release(&grt->u.dst);
1180
1181                         if (err)
1182                                 goto out;
1183                 }
1184                 err = -EINVAL;
1185                 if (dev == NULL || (dev->flags&IFF_LOOPBACK))
1186                         goto out;
1187         }
1188
1189         err = -ENODEV;
1190         if (dev == NULL)
1191                 goto out;
1192
1193         if (cfg->fc_flags & (RTF_GATEWAY | RTF_NONEXTHOP)) {
1194                 rt->rt6i_nexthop = __neigh_lookup_errno(&nd_tbl, &rt->rt6i_gateway, dev);
1195                 if (IS_ERR(rt->rt6i_nexthop)) {
1196                         err = PTR_ERR(rt->rt6i_nexthop);
1197                         rt->rt6i_nexthop = NULL;
1198                         goto out;
1199                 }
1200         }
1201
1202         rt->rt6i_flags = cfg->fc_flags;
1203
1204 install_route:
1205         if (cfg->fc_mx) {
1206                 struct nlattr *nla;
1207                 int remaining;
1208
1209                 nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1210                         int type = nla_type(nla);
1211
1212                         if (type) {
1213                                 if (type > RTAX_MAX) {
1214                                         err = -EINVAL;
1215                                         goto out;
1216                                 }
1217
1218                                 rt->u.dst.metrics[type - 1] = nla_get_u32(nla);
1219                         }
1220                 }
1221         }
1222
1223         if (rt->u.dst.metrics[RTAX_HOPLIMIT-1] == 0)
1224                 rt->u.dst.metrics[RTAX_HOPLIMIT-1] = -1;
1225         if (!rt->u.dst.metrics[RTAX_MTU-1])
1226                 rt->u.dst.metrics[RTAX_MTU-1] = ipv6_get_mtu(dev);
1227         if (!rt->u.dst.metrics[RTAX_ADVMSS-1])
1228                 rt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(dst_mtu(&rt->u.dst));
1229         rt->u.dst.dev = dev;
1230         rt->rt6i_idev = idev;
1231         rt->rt6i_table = table;
1232         return __ip6_ins_rt(rt, &cfg->fc_nlinfo);
1233
1234 out:
1235         if (dev)
1236                 dev_put(dev);
1237         if (idev)
1238                 in6_dev_put(idev);
1239         if (rt)
1240                 dst_free(&rt->u.dst);
1241         return err;
1242 }
1243
1244 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
1245 {
1246         int err;
1247         struct fib6_table *table;
1248
1249         if (rt == &ip6_null_entry)
1250                 return -ENOENT;
1251
1252         table = rt->rt6i_table;
1253         write_lock_bh(&table->tb6_lock);
1254
1255         err = fib6_del(rt, info);
1256         dst_release(&rt->u.dst);
1257
1258         write_unlock_bh(&table->tb6_lock);
1259
1260         return err;
1261 }
1262
1263 int ip6_del_rt(struct rt6_info *rt)
1264 {
1265         struct nl_info info = {
1266                 .nl_net = &init_net,
1267         };
1268         return __ip6_del_rt(rt, &info);
1269 }
1270
1271 static int ip6_route_del(struct fib6_config *cfg)
1272 {
1273         struct fib6_table *table;
1274         struct fib6_node *fn;
1275         struct rt6_info *rt;
1276         int err = -ESRCH;
1277
1278         table = fib6_get_table(cfg->fc_table);
1279         if (table == NULL)
1280                 return err;
1281
1282         read_lock_bh(&table->tb6_lock);
1283
1284         fn = fib6_locate(&table->tb6_root,
1285                          &cfg->fc_dst, cfg->fc_dst_len,
1286                          &cfg->fc_src, cfg->fc_src_len);
1287
1288         if (fn) {
1289                 for (rt = fn->leaf; rt; rt = rt->u.dst.rt6_next) {
1290                         if (cfg->fc_ifindex &&
1291                             (rt->rt6i_dev == NULL ||
1292                              rt->rt6i_dev->ifindex != cfg->fc_ifindex))
1293                                 continue;
1294                         if (cfg->fc_flags & RTF_GATEWAY &&
1295                             !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
1296                                 continue;
1297                         if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
1298                                 continue;
1299                         dst_hold(&rt->u.dst);
1300                         read_unlock_bh(&table->tb6_lock);
1301
1302                         return __ip6_del_rt(rt, &cfg->fc_nlinfo);
1303                 }
1304         }
1305         read_unlock_bh(&table->tb6_lock);
1306
1307         return err;
1308 }
1309
1310 /*
1311  *      Handle redirects
1312  */
1313 struct ip6rd_flowi {
1314         struct flowi fl;
1315         struct in6_addr gateway;
1316 };
1317
1318 static struct rt6_info *__ip6_route_redirect(struct fib6_table *table,
1319                                              struct flowi *fl,
1320                                              int flags)
1321 {
1322         struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl;
1323         struct rt6_info *rt;
1324         struct fib6_node *fn;
1325
1326         /*
1327          * Get the "current" route for this destination and
1328          * check if the redirect has come from approriate router.
1329          *
1330          * RFC 2461 specifies that redirects should only be
1331          * accepted if they come from the nexthop to the target.
1332          * Due to the way the routes are chosen, this notion
1333          * is a bit fuzzy and one might need to check all possible
1334          * routes.
1335          */
1336
1337         read_lock_bh(&table->tb6_lock);
1338         fn = fib6_lookup(&table->tb6_root, &fl->fl6_dst, &fl->fl6_src);
1339 restart:
1340         for (rt = fn->leaf; rt; rt = rt->u.dst.rt6_next) {
1341                 /*
1342                  * Current route is on-link; redirect is always invalid.
1343                  *
1344                  * Seems, previous statement is not true. It could
1345                  * be node, which looks for us as on-link (f.e. proxy ndisc)
1346                  * But then router serving it might decide, that we should
1347                  * know truth 8)8) --ANK (980726).
1348                  */
1349                 if (rt6_check_expired(rt))
1350                         continue;
1351                 if (!(rt->rt6i_flags & RTF_GATEWAY))
1352                         continue;
1353                 if (fl->oif != rt->rt6i_dev->ifindex)
1354                         continue;
1355                 if (!ipv6_addr_equal(&rdfl->gateway, &rt->rt6i_gateway))
1356                         continue;
1357                 break;
1358         }
1359
1360         if (!rt)
1361                 rt = &ip6_null_entry;
1362         BACKTRACK(&fl->fl6_src);
1363 out:
1364         dst_hold(&rt->u.dst);
1365
1366         read_unlock_bh(&table->tb6_lock);
1367
1368         return rt;
1369 };
1370
1371 static struct rt6_info *ip6_route_redirect(struct in6_addr *dest,
1372                                            struct in6_addr *src,
1373                                            struct in6_addr *gateway,
1374                                            struct net_device *dev)
1375 {
1376         int flags = RT6_LOOKUP_F_HAS_SADDR;
1377         struct ip6rd_flowi rdfl = {
1378                 .fl = {
1379                         .oif = dev->ifindex,
1380                         .nl_u = {
1381                                 .ip6_u = {
1382                                         .daddr = *dest,
1383                                         .saddr = *src,
1384                                 },
1385                         },
1386                 },
1387                 .gateway = *gateway,
1388         };
1389
1390         if (rt6_need_strict(dest))
1391                 flags |= RT6_LOOKUP_F_IFACE;
1392
1393         return (struct rt6_info *)fib6_rule_lookup((struct flowi *)&rdfl, flags, __ip6_route_redirect);
1394 }
1395
1396 void rt6_redirect(struct in6_addr *dest, struct in6_addr *src,
1397                   struct in6_addr *saddr,
1398                   struct neighbour *neigh, u8 *lladdr, int on_link)
1399 {
1400         struct rt6_info *rt, *nrt = NULL;
1401         struct netevent_redirect netevent;
1402
1403         rt = ip6_route_redirect(dest, src, saddr, neigh->dev);
1404
1405         if (rt == &ip6_null_entry) {
1406                 if (net_ratelimit())
1407                         printk(KERN_DEBUG "rt6_redirect: source isn't a valid nexthop "
1408                                "for redirect target\n");
1409                 goto out;
1410         }
1411
1412         /*
1413          *      We have finally decided to accept it.
1414          */
1415
1416         neigh_update(neigh, lladdr, NUD_STALE,
1417                      NEIGH_UPDATE_F_WEAK_OVERRIDE|
1418                      NEIGH_UPDATE_F_OVERRIDE|
1419                      (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1420                                      NEIGH_UPDATE_F_ISROUTER))
1421                      );
1422
1423         /*
1424          * Redirect received -> path was valid.
1425          * Look, redirects are sent only in response to data packets,
1426          * so that this nexthop apparently is reachable. --ANK
1427          */
1428         dst_confirm(&rt->u.dst);
1429
1430         /* Duplicate redirect: silently ignore. */
1431         if (neigh == rt->u.dst.neighbour)
1432                 goto out;
1433
1434         nrt = ip6_rt_copy(rt);
1435         if (nrt == NULL)
1436                 goto out;
1437
1438         nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
1439         if (on_link)
1440                 nrt->rt6i_flags &= ~RTF_GATEWAY;
1441
1442         ipv6_addr_copy(&nrt->rt6i_dst.addr, dest);
1443         nrt->rt6i_dst.plen = 128;
1444         nrt->u.dst.flags |= DST_HOST;
1445
1446         ipv6_addr_copy(&nrt->rt6i_gateway, (struct in6_addr*)neigh->primary_key);
1447         nrt->rt6i_nexthop = neigh_clone(neigh);
1448         /* Reset pmtu, it may be better */
1449         nrt->u.dst.metrics[RTAX_MTU-1] = ipv6_get_mtu(neigh->dev);
1450         nrt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(dst_mtu(&nrt->u.dst));
1451
1452         if (ip6_ins_rt(nrt))
1453                 goto out;
1454
1455         netevent.old = &rt->u.dst;
1456         netevent.new = &nrt->u.dst;
1457         call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
1458
1459         if (rt->rt6i_flags&RTF_CACHE) {
1460                 ip6_del_rt(rt);
1461                 return;
1462         }
1463
1464 out:
1465         dst_release(&rt->u.dst);
1466         return;
1467 }
1468
1469 /*
1470  *      Handle ICMP "packet too big" messages
1471  *      i.e. Path MTU discovery
1472  */
1473
1474 void rt6_pmtu_discovery(struct in6_addr *daddr, struct in6_addr *saddr,
1475                         struct net_device *dev, u32 pmtu)
1476 {
1477         struct rt6_info *rt, *nrt;
1478         int allfrag = 0;
1479
1480         rt = rt6_lookup(daddr, saddr, dev->ifindex, 0);
1481         if (rt == NULL)
1482                 return;
1483
1484         if (pmtu >= dst_mtu(&rt->u.dst))
1485                 goto out;
1486
1487         if (pmtu < IPV6_MIN_MTU) {
1488                 /*
1489                  * According to RFC2460, PMTU is set to the IPv6 Minimum Link
1490                  * MTU (1280) and a fragment header should always be included
1491                  * after a node receiving Too Big message reporting PMTU is
1492                  * less than the IPv6 Minimum Link MTU.
1493                  */
1494                 pmtu = IPV6_MIN_MTU;
1495                 allfrag = 1;
1496         }
1497
1498         /* New mtu received -> path was valid.
1499            They are sent only in response to data packets,
1500            so that this nexthop apparently is reachable. --ANK
1501          */
1502         dst_confirm(&rt->u.dst);
1503
1504         /* Host route. If it is static, it would be better
1505            not to override it, but add new one, so that
1506            when cache entry will expire old pmtu
1507            would return automatically.
1508          */
1509         if (rt->rt6i_flags & RTF_CACHE) {
1510                 rt->u.dst.metrics[RTAX_MTU-1] = pmtu;
1511                 if (allfrag)
1512                         rt->u.dst.metrics[RTAX_FEATURES-1] |= RTAX_FEATURE_ALLFRAG;
1513                 dst_set_expires(&rt->u.dst, init_net.ipv6.sysctl.ip6_rt_mtu_expires);
1514                 rt->rt6i_flags |= RTF_MODIFIED|RTF_EXPIRES;
1515                 goto out;
1516         }
1517
1518         /* Network route.
1519            Two cases are possible:
1520            1. It is connected route. Action: COW
1521            2. It is gatewayed route or NONEXTHOP route. Action: clone it.
1522          */
1523         if (!rt->rt6i_nexthop && !(rt->rt6i_flags & RTF_NONEXTHOP))
1524                 nrt = rt6_alloc_cow(rt, daddr, saddr);
1525         else
1526                 nrt = rt6_alloc_clone(rt, daddr);
1527
1528         if (nrt) {
1529                 nrt->u.dst.metrics[RTAX_MTU-1] = pmtu;
1530                 if (allfrag)
1531                         nrt->u.dst.metrics[RTAX_FEATURES-1] |= RTAX_FEATURE_ALLFRAG;
1532
1533                 /* According to RFC 1981, detecting PMTU increase shouldn't be
1534                  * happened within 5 mins, the recommended timer is 10 mins.
1535                  * Here this route expiration time is set to ip6_rt_mtu_expires
1536                  * which is 10 mins. After 10 mins the decreased pmtu is expired
1537                  * and detecting PMTU increase will be automatically happened.
1538                  */
1539                 dst_set_expires(&nrt->u.dst, init_net.ipv6.sysctl.ip6_rt_mtu_expires);
1540                 nrt->rt6i_flags |= RTF_DYNAMIC|RTF_EXPIRES;
1541
1542                 ip6_ins_rt(nrt);
1543         }
1544 out:
1545         dst_release(&rt->u.dst);
1546 }
1547
1548 /*
1549  *      Misc support functions
1550  */
1551
1552 static struct rt6_info * ip6_rt_copy(struct rt6_info *ort)
1553 {
1554         struct rt6_info *rt = ip6_dst_alloc();
1555
1556         if (rt) {
1557                 rt->u.dst.input = ort->u.dst.input;
1558                 rt->u.dst.output = ort->u.dst.output;
1559
1560                 memcpy(rt->u.dst.metrics, ort->u.dst.metrics, RTAX_MAX*sizeof(u32));
1561                 rt->u.dst.error = ort->u.dst.error;
1562                 rt->u.dst.dev = ort->u.dst.dev;
1563                 if (rt->u.dst.dev)
1564                         dev_hold(rt->u.dst.dev);
1565                 rt->rt6i_idev = ort->rt6i_idev;
1566                 if (rt->rt6i_idev)
1567                         in6_dev_hold(rt->rt6i_idev);
1568                 rt->u.dst.lastuse = jiffies;
1569                 rt->rt6i_expires = 0;
1570
1571                 ipv6_addr_copy(&rt->rt6i_gateway, &ort->rt6i_gateway);
1572                 rt->rt6i_flags = ort->rt6i_flags & ~RTF_EXPIRES;
1573                 rt->rt6i_metric = 0;
1574
1575                 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1576 #ifdef CONFIG_IPV6_SUBTREES
1577                 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1578 #endif
1579                 rt->rt6i_table = ort->rt6i_table;
1580         }
1581         return rt;
1582 }
1583
1584 #ifdef CONFIG_IPV6_ROUTE_INFO
1585 static struct rt6_info *rt6_get_route_info(struct in6_addr *prefix, int prefixlen,
1586                                            struct in6_addr *gwaddr, int ifindex)
1587 {
1588         struct fib6_node *fn;
1589         struct rt6_info *rt = NULL;
1590         struct fib6_table *table;
1591
1592         table = fib6_get_table(RT6_TABLE_INFO);
1593         if (table == NULL)
1594                 return NULL;
1595
1596         write_lock_bh(&table->tb6_lock);
1597         fn = fib6_locate(&table->tb6_root, prefix ,prefixlen, NULL, 0);
1598         if (!fn)
1599                 goto out;
1600
1601         for (rt = fn->leaf; rt; rt = rt->u.dst.rt6_next) {
1602                 if (rt->rt6i_dev->ifindex != ifindex)
1603                         continue;
1604                 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
1605                         continue;
1606                 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
1607                         continue;
1608                 dst_hold(&rt->u.dst);
1609                 break;
1610         }
1611 out:
1612         write_unlock_bh(&table->tb6_lock);
1613         return rt;
1614 }
1615
1616 static struct rt6_info *rt6_add_route_info(struct in6_addr *prefix, int prefixlen,
1617                                            struct in6_addr *gwaddr, int ifindex,
1618                                            unsigned pref)
1619 {
1620         struct fib6_config cfg = {
1621                 .fc_table       = RT6_TABLE_INFO,
1622                 .fc_metric      = IP6_RT_PRIO_USER,
1623                 .fc_ifindex     = ifindex,
1624                 .fc_dst_len     = prefixlen,
1625                 .fc_flags       = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
1626                                   RTF_UP | RTF_PREF(pref),
1627         };
1628
1629         ipv6_addr_copy(&cfg.fc_dst, prefix);
1630         ipv6_addr_copy(&cfg.fc_gateway, gwaddr);
1631
1632         /* We should treat it as a default route if prefix length is 0. */
1633         if (!prefixlen)
1634                 cfg.fc_flags |= RTF_DEFAULT;
1635
1636         ip6_route_add(&cfg);
1637
1638         return rt6_get_route_info(prefix, prefixlen, gwaddr, ifindex);
1639 }
1640 #endif
1641
1642 struct rt6_info *rt6_get_dflt_router(struct in6_addr *addr, struct net_device *dev)
1643 {
1644         struct rt6_info *rt;
1645         struct fib6_table *table;
1646
1647         table = fib6_get_table(RT6_TABLE_DFLT);
1648         if (table == NULL)
1649                 return NULL;
1650
1651         write_lock_bh(&table->tb6_lock);
1652         for (rt = table->tb6_root.leaf; rt; rt=rt->u.dst.rt6_next) {
1653                 if (dev == rt->rt6i_dev &&
1654                     ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
1655                     ipv6_addr_equal(&rt->rt6i_gateway, addr))
1656                         break;
1657         }
1658         if (rt)
1659                 dst_hold(&rt->u.dst);
1660         write_unlock_bh(&table->tb6_lock);
1661         return rt;
1662 }
1663
1664 EXPORT_SYMBOL(rt6_get_dflt_router);
1665
1666 struct rt6_info *rt6_add_dflt_router(struct in6_addr *gwaddr,
1667                                      struct net_device *dev,
1668                                      unsigned int pref)
1669 {
1670         struct fib6_config cfg = {
1671                 .fc_table       = RT6_TABLE_DFLT,
1672                 .fc_metric      = IP6_RT_PRIO_USER,
1673                 .fc_ifindex     = dev->ifindex,
1674                 .fc_flags       = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
1675                                   RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
1676         };
1677
1678         ipv6_addr_copy(&cfg.fc_gateway, gwaddr);
1679
1680         ip6_route_add(&cfg);
1681
1682         return rt6_get_dflt_router(gwaddr, dev);
1683 }
1684
1685 void rt6_purge_dflt_routers(void)
1686 {
1687         struct rt6_info *rt;
1688         struct fib6_table *table;
1689
1690         /* NOTE: Keep consistent with rt6_get_dflt_router */
1691         table = fib6_get_table(RT6_TABLE_DFLT);
1692         if (table == NULL)
1693                 return;
1694
1695 restart:
1696         read_lock_bh(&table->tb6_lock);
1697         for (rt = table->tb6_root.leaf; rt; rt = rt->u.dst.rt6_next) {
1698                 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF)) {
1699                         dst_hold(&rt->u.dst);
1700                         read_unlock_bh(&table->tb6_lock);
1701                         ip6_del_rt(rt);
1702                         goto restart;
1703                 }
1704         }
1705         read_unlock_bh(&table->tb6_lock);
1706 }
1707
1708 static void rtmsg_to_fib6_config(struct in6_rtmsg *rtmsg,
1709                                  struct fib6_config *cfg)
1710 {
1711         memset(cfg, 0, sizeof(*cfg));
1712
1713         cfg->fc_table = RT6_TABLE_MAIN;
1714         cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
1715         cfg->fc_metric = rtmsg->rtmsg_metric;
1716         cfg->fc_expires = rtmsg->rtmsg_info;
1717         cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
1718         cfg->fc_src_len = rtmsg->rtmsg_src_len;
1719         cfg->fc_flags = rtmsg->rtmsg_flags;
1720
1721         cfg->fc_nlinfo.nl_net = &init_net;
1722
1723         ipv6_addr_copy(&cfg->fc_dst, &rtmsg->rtmsg_dst);
1724         ipv6_addr_copy(&cfg->fc_src, &rtmsg->rtmsg_src);
1725         ipv6_addr_copy(&cfg->fc_gateway, &rtmsg->rtmsg_gateway);
1726 }
1727
1728 int ipv6_route_ioctl(unsigned int cmd, void __user *arg)
1729 {
1730         struct fib6_config cfg;
1731         struct in6_rtmsg rtmsg;
1732         int err;
1733
1734         switch(cmd) {
1735         case SIOCADDRT:         /* Add a route */
1736         case SIOCDELRT:         /* Delete a route */
1737                 if (!capable(CAP_NET_ADMIN))
1738                         return -EPERM;
1739                 err = copy_from_user(&rtmsg, arg,
1740                                      sizeof(struct in6_rtmsg));
1741                 if (err)
1742                         return -EFAULT;
1743
1744                 rtmsg_to_fib6_config(&rtmsg, &cfg);
1745
1746                 rtnl_lock();
1747                 switch (cmd) {
1748                 case SIOCADDRT:
1749                         err = ip6_route_add(&cfg);
1750                         break;
1751                 case SIOCDELRT:
1752                         err = ip6_route_del(&cfg);
1753                         break;
1754                 default:
1755                         err = -EINVAL;
1756                 }
1757                 rtnl_unlock();
1758
1759                 return err;
1760         }
1761
1762         return -EINVAL;
1763 }
1764
1765 /*
1766  *      Drop the packet on the floor
1767  */
1768
1769 static int ip6_pkt_drop(struct sk_buff *skb, int code, int ipstats_mib_noroutes)
1770 {
1771         int type;
1772         switch (ipstats_mib_noroutes) {
1773         case IPSTATS_MIB_INNOROUTES:
1774                 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
1775                 if (type == IPV6_ADDR_ANY || type == IPV6_ADDR_RESERVED) {
1776                         IP6_INC_STATS(ip6_dst_idev(skb->dst), IPSTATS_MIB_INADDRERRORS);
1777                         break;
1778                 }
1779                 /* FALLTHROUGH */
1780         case IPSTATS_MIB_OUTNOROUTES:
1781                 IP6_INC_STATS(ip6_dst_idev(skb->dst), ipstats_mib_noroutes);
1782                 break;
1783         }
1784         icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0, skb->dev);
1785         kfree_skb(skb);
1786         return 0;
1787 }
1788
1789 static int ip6_pkt_discard(struct sk_buff *skb)
1790 {
1791         return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
1792 }
1793
1794 static int ip6_pkt_discard_out(struct sk_buff *skb)
1795 {
1796         skb->dev = skb->dst->dev;
1797         return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
1798 }
1799
1800 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
1801
1802 static int ip6_pkt_prohibit(struct sk_buff *skb)
1803 {
1804         return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
1805 }
1806
1807 static int ip6_pkt_prohibit_out(struct sk_buff *skb)
1808 {
1809         skb->dev = skb->dst->dev;
1810         return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
1811 }
1812
1813 #endif
1814
1815 /*
1816  *      Allocate a dst for local (unicast / anycast) address.
1817  */
1818
1819 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
1820                                     const struct in6_addr *addr,
1821                                     int anycast)
1822 {
1823         struct rt6_info *rt = ip6_dst_alloc();
1824
1825         if (rt == NULL)
1826                 return ERR_PTR(-ENOMEM);
1827
1828         dev_hold(init_net.loopback_dev);
1829         in6_dev_hold(idev);
1830
1831         rt->u.dst.flags = DST_HOST;
1832         rt->u.dst.input = ip6_input;
1833         rt->u.dst.output = ip6_output;
1834         rt->rt6i_dev = init_net.loopback_dev;
1835         rt->rt6i_idev = idev;
1836         rt->u.dst.metrics[RTAX_MTU-1] = ipv6_get_mtu(rt->rt6i_dev);
1837         rt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(dst_mtu(&rt->u.dst));
1838         rt->u.dst.metrics[RTAX_HOPLIMIT-1] = -1;
1839         rt->u.dst.obsolete = -1;
1840
1841         rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
1842         if (anycast)
1843                 rt->rt6i_flags |= RTF_ANYCAST;
1844         else
1845                 rt->rt6i_flags |= RTF_LOCAL;
1846         rt->rt6i_nexthop = ndisc_get_neigh(rt->rt6i_dev, &rt->rt6i_gateway);
1847         if (rt->rt6i_nexthop == NULL) {
1848                 dst_free(&rt->u.dst);
1849                 return ERR_PTR(-ENOMEM);
1850         }
1851
1852         ipv6_addr_copy(&rt->rt6i_dst.addr, addr);
1853         rt->rt6i_dst.plen = 128;
1854         rt->rt6i_table = fib6_get_table(RT6_TABLE_LOCAL);
1855
1856         atomic_set(&rt->u.dst.__refcnt, 1);
1857
1858         return rt;
1859 }
1860
1861 static int fib6_ifdown(struct rt6_info *rt, void *arg)
1862 {
1863         if (((void*)rt->rt6i_dev == arg || arg == NULL) &&
1864             rt != &ip6_null_entry) {
1865                 RT6_TRACE("deleted by ifdown %p\n", rt);
1866                 return -1;
1867         }
1868         return 0;
1869 }
1870
1871 void rt6_ifdown(struct net_device *dev)
1872 {
1873         fib6_clean_all(fib6_ifdown, 0, dev);
1874 }
1875
1876 struct rt6_mtu_change_arg
1877 {
1878         struct net_device *dev;
1879         unsigned mtu;
1880 };
1881
1882 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
1883 {
1884         struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
1885         struct inet6_dev *idev;
1886
1887         /* In IPv6 pmtu discovery is not optional,
1888            so that RTAX_MTU lock cannot disable it.
1889            We still use this lock to block changes
1890            caused by addrconf/ndisc.
1891         */
1892
1893         idev = __in6_dev_get(arg->dev);
1894         if (idev == NULL)
1895                 return 0;
1896
1897         /* For administrative MTU increase, there is no way to discover
1898            IPv6 PMTU increase, so PMTU increase should be updated here.
1899            Since RFC 1981 doesn't include administrative MTU increase
1900            update PMTU increase is a MUST. (i.e. jumbo frame)
1901          */
1902         /*
1903            If new MTU is less than route PMTU, this new MTU will be the
1904            lowest MTU in the path, update the route PMTU to reflect PMTU
1905            decreases; if new MTU is greater than route PMTU, and the
1906            old MTU is the lowest MTU in the path, update the route PMTU
1907            to reflect the increase. In this case if the other nodes' MTU
1908            also have the lowest MTU, TOO BIG MESSAGE will be lead to
1909            PMTU discouvery.
1910          */
1911         if (rt->rt6i_dev == arg->dev &&
1912             !dst_metric_locked(&rt->u.dst, RTAX_MTU) &&
1913             (dst_mtu(&rt->u.dst) >= arg->mtu ||
1914              (dst_mtu(&rt->u.dst) < arg->mtu &&
1915               dst_mtu(&rt->u.dst) == idev->cnf.mtu6))) {
1916                 rt->u.dst.metrics[RTAX_MTU-1] = arg->mtu;
1917                 rt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(arg->mtu);
1918         }
1919         return 0;
1920 }
1921
1922 void rt6_mtu_change(struct net_device *dev, unsigned mtu)
1923 {
1924         struct rt6_mtu_change_arg arg = {
1925                 .dev = dev,
1926                 .mtu = mtu,
1927         };
1928
1929         fib6_clean_all(rt6_mtu_change_route, 0, &arg);
1930 }
1931
1932 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
1933         [RTA_GATEWAY]           = { .len = sizeof(struct in6_addr) },
1934         [RTA_OIF]               = { .type = NLA_U32 },
1935         [RTA_IIF]               = { .type = NLA_U32 },
1936         [RTA_PRIORITY]          = { .type = NLA_U32 },
1937         [RTA_METRICS]           = { .type = NLA_NESTED },
1938 };
1939
1940 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
1941                               struct fib6_config *cfg)
1942 {
1943         struct rtmsg *rtm;
1944         struct nlattr *tb[RTA_MAX+1];
1945         int err;
1946
1947         err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
1948         if (err < 0)
1949                 goto errout;
1950
1951         err = -EINVAL;
1952         rtm = nlmsg_data(nlh);
1953         memset(cfg, 0, sizeof(*cfg));
1954
1955         cfg->fc_table = rtm->rtm_table;
1956         cfg->fc_dst_len = rtm->rtm_dst_len;
1957         cfg->fc_src_len = rtm->rtm_src_len;
1958         cfg->fc_flags = RTF_UP;
1959         cfg->fc_protocol = rtm->rtm_protocol;
1960
1961         if (rtm->rtm_type == RTN_UNREACHABLE)
1962                 cfg->fc_flags |= RTF_REJECT;
1963
1964         cfg->fc_nlinfo.pid = NETLINK_CB(skb).pid;
1965         cfg->fc_nlinfo.nlh = nlh;
1966         cfg->fc_nlinfo.nl_net = skb->sk->sk_net;
1967
1968         if (tb[RTA_GATEWAY]) {
1969                 nla_memcpy(&cfg->fc_gateway, tb[RTA_GATEWAY], 16);
1970                 cfg->fc_flags |= RTF_GATEWAY;
1971         }
1972
1973         if (tb[RTA_DST]) {
1974                 int plen = (rtm->rtm_dst_len + 7) >> 3;
1975
1976                 if (nla_len(tb[RTA_DST]) < plen)
1977                         goto errout;
1978
1979                 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
1980         }
1981
1982         if (tb[RTA_SRC]) {
1983                 int plen = (rtm->rtm_src_len + 7) >> 3;
1984
1985                 if (nla_len(tb[RTA_SRC]) < plen)
1986                         goto errout;
1987
1988                 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
1989         }
1990
1991         if (tb[RTA_OIF])
1992                 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
1993
1994         if (tb[RTA_PRIORITY])
1995                 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
1996
1997         if (tb[RTA_METRICS]) {
1998                 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
1999                 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2000         }
2001
2002         if (tb[RTA_TABLE])
2003                 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2004
2005         err = 0;
2006 errout:
2007         return err;
2008 }
2009
2010 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
2011 {
2012         struct net *net = skb->sk->sk_net;
2013         struct fib6_config cfg;
2014         int err;
2015
2016         if (net != &init_net)
2017                 return -EINVAL;
2018
2019         err = rtm_to_fib6_config(skb, nlh, &cfg);
2020         if (err < 0)
2021                 return err;
2022
2023         return ip6_route_del(&cfg);
2024 }
2025
2026 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
2027 {
2028         struct net *net = skb->sk->sk_net;
2029         struct fib6_config cfg;
2030         int err;
2031
2032         if (net != &init_net)
2033                 return -EINVAL;
2034
2035         err = rtm_to_fib6_config(skb, nlh, &cfg);
2036         if (err < 0)
2037                 return err;
2038
2039         return ip6_route_add(&cfg);
2040 }
2041
2042 static inline size_t rt6_nlmsg_size(void)
2043 {
2044         return NLMSG_ALIGN(sizeof(struct rtmsg))
2045                + nla_total_size(16) /* RTA_SRC */
2046                + nla_total_size(16) /* RTA_DST */
2047                + nla_total_size(16) /* RTA_GATEWAY */
2048                + nla_total_size(16) /* RTA_PREFSRC */
2049                + nla_total_size(4) /* RTA_TABLE */
2050                + nla_total_size(4) /* RTA_IIF */
2051                + nla_total_size(4) /* RTA_OIF */
2052                + nla_total_size(4) /* RTA_PRIORITY */
2053                + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
2054                + nla_total_size(sizeof(struct rta_cacheinfo));
2055 }
2056
2057 static int rt6_fill_node(struct sk_buff *skb, struct rt6_info *rt,
2058                          struct in6_addr *dst, struct in6_addr *src,
2059                          int iif, int type, u32 pid, u32 seq,
2060                          int prefix, unsigned int flags)
2061 {
2062         struct rtmsg *rtm;
2063         struct nlmsghdr *nlh;
2064         long expires;
2065         u32 table;
2066
2067         if (prefix) {   /* user wants prefix routes only */
2068                 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
2069                         /* success since this is not a prefix route */
2070                         return 1;
2071                 }
2072         }
2073
2074         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*rtm), flags);
2075         if (nlh == NULL)
2076                 return -EMSGSIZE;
2077
2078         rtm = nlmsg_data(nlh);
2079         rtm->rtm_family = AF_INET6;
2080         rtm->rtm_dst_len = rt->rt6i_dst.plen;
2081         rtm->rtm_src_len = rt->rt6i_src.plen;
2082         rtm->rtm_tos = 0;
2083         if (rt->rt6i_table)
2084                 table = rt->rt6i_table->tb6_id;
2085         else
2086                 table = RT6_TABLE_UNSPEC;
2087         rtm->rtm_table = table;
2088         NLA_PUT_U32(skb, RTA_TABLE, table);
2089         if (rt->rt6i_flags&RTF_REJECT)
2090                 rtm->rtm_type = RTN_UNREACHABLE;
2091         else if (rt->rt6i_dev && (rt->rt6i_dev->flags&IFF_LOOPBACK))
2092                 rtm->rtm_type = RTN_LOCAL;
2093         else
2094                 rtm->rtm_type = RTN_UNICAST;
2095         rtm->rtm_flags = 0;
2096         rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2097         rtm->rtm_protocol = rt->rt6i_protocol;
2098         if (rt->rt6i_flags&RTF_DYNAMIC)
2099                 rtm->rtm_protocol = RTPROT_REDIRECT;
2100         else if (rt->rt6i_flags & RTF_ADDRCONF)
2101                 rtm->rtm_protocol = RTPROT_KERNEL;
2102         else if (rt->rt6i_flags&RTF_DEFAULT)
2103                 rtm->rtm_protocol = RTPROT_RA;
2104
2105         if (rt->rt6i_flags&RTF_CACHE)
2106                 rtm->rtm_flags |= RTM_F_CLONED;
2107
2108         if (dst) {
2109                 NLA_PUT(skb, RTA_DST, 16, dst);
2110                 rtm->rtm_dst_len = 128;
2111         } else if (rtm->rtm_dst_len)
2112                 NLA_PUT(skb, RTA_DST, 16, &rt->rt6i_dst.addr);
2113 #ifdef CONFIG_IPV6_SUBTREES
2114         if (src) {
2115                 NLA_PUT(skb, RTA_SRC, 16, src);
2116                 rtm->rtm_src_len = 128;
2117         } else if (rtm->rtm_src_len)
2118                 NLA_PUT(skb, RTA_SRC, 16, &rt->rt6i_src.addr);
2119 #endif
2120         if (iif)
2121                 NLA_PUT_U32(skb, RTA_IIF, iif);
2122         else if (dst) {
2123                 struct in6_addr saddr_buf;
2124                 if (ipv6_dev_get_saddr(ip6_dst_idev(&rt->u.dst)->dev,
2125                                        dst, &saddr_buf) == 0)
2126                         NLA_PUT(skb, RTA_PREFSRC, 16, &saddr_buf);
2127         }
2128
2129         if (rtnetlink_put_metrics(skb, rt->u.dst.metrics) < 0)
2130                 goto nla_put_failure;
2131
2132         if (rt->u.dst.neighbour)
2133                 NLA_PUT(skb, RTA_GATEWAY, 16, &rt->u.dst.neighbour->primary_key);
2134
2135         if (rt->u.dst.dev)
2136                 NLA_PUT_U32(skb, RTA_OIF, rt->rt6i_dev->ifindex);
2137
2138         NLA_PUT_U32(skb, RTA_PRIORITY, rt->rt6i_metric);
2139
2140         expires = rt->rt6i_expires ? rt->rt6i_expires - jiffies : 0;
2141         if (rtnl_put_cacheinfo(skb, &rt->u.dst, 0, 0, 0,
2142                                expires, rt->u.dst.error) < 0)
2143                 goto nla_put_failure;
2144
2145         return nlmsg_end(skb, nlh);
2146
2147 nla_put_failure:
2148         nlmsg_cancel(skb, nlh);
2149         return -EMSGSIZE;
2150 }
2151
2152 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
2153 {
2154         struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
2155         int prefix;
2156
2157         if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
2158                 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
2159                 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
2160         } else
2161                 prefix = 0;
2162
2163         return rt6_fill_node(arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
2164                      NETLINK_CB(arg->cb->skb).pid, arg->cb->nlh->nlmsg_seq,
2165                      prefix, NLM_F_MULTI);
2166 }
2167
2168 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr* nlh, void *arg)
2169 {
2170         struct net *net = in_skb->sk->sk_net;
2171         struct nlattr *tb[RTA_MAX+1];
2172         struct rt6_info *rt;
2173         struct sk_buff *skb;
2174         struct rtmsg *rtm;
2175         struct flowi fl;
2176         int err, iif = 0;
2177
2178         if (net != &init_net)
2179                 return -EINVAL;
2180
2181         err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2182         if (err < 0)
2183                 goto errout;
2184
2185         err = -EINVAL;
2186         memset(&fl, 0, sizeof(fl));
2187
2188         if (tb[RTA_SRC]) {
2189                 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
2190                         goto errout;
2191
2192                 ipv6_addr_copy(&fl.fl6_src, nla_data(tb[RTA_SRC]));
2193         }
2194
2195         if (tb[RTA_DST]) {
2196                 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
2197                         goto errout;
2198
2199                 ipv6_addr_copy(&fl.fl6_dst, nla_data(tb[RTA_DST]));
2200         }
2201
2202         if (tb[RTA_IIF])
2203                 iif = nla_get_u32(tb[RTA_IIF]);
2204
2205         if (tb[RTA_OIF])
2206                 fl.oif = nla_get_u32(tb[RTA_OIF]);
2207
2208         if (iif) {
2209                 struct net_device *dev;
2210                 dev = __dev_get_by_index(&init_net, iif);
2211                 if (!dev) {
2212                         err = -ENODEV;
2213                         goto errout;
2214                 }
2215         }
2216
2217         skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
2218         if (skb == NULL) {
2219                 err = -ENOBUFS;
2220                 goto errout;
2221         }
2222
2223         /* Reserve room for dummy headers, this skb can pass
2224            through good chunk of routing engine.
2225          */
2226         skb_reset_mac_header(skb);
2227         skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
2228
2229         rt = (struct rt6_info*) ip6_route_output(NULL, &fl);
2230         skb->dst = &rt->u.dst;
2231
2232         err = rt6_fill_node(skb, rt, &fl.fl6_dst, &fl.fl6_src, iif,
2233                             RTM_NEWROUTE, NETLINK_CB(in_skb).pid,
2234                             nlh->nlmsg_seq, 0, 0);
2235         if (err < 0) {
2236                 kfree_skb(skb);
2237                 goto errout;
2238         }
2239
2240         err = rtnl_unicast(skb, &init_net, NETLINK_CB(in_skb).pid);
2241 errout:
2242         return err;
2243 }
2244
2245 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info)
2246 {
2247         struct sk_buff *skb;
2248         u32 seq;
2249         int err;
2250
2251         err = -ENOBUFS;
2252         seq = info->nlh != NULL ? info->nlh->nlmsg_seq : 0;
2253
2254         skb = nlmsg_new(rt6_nlmsg_size(), gfp_any());
2255         if (skb == NULL)
2256                 goto errout;
2257
2258         err = rt6_fill_node(skb, rt, NULL, NULL, 0,
2259                                 event, info->pid, seq, 0, 0);
2260         if (err < 0) {
2261                 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
2262                 WARN_ON(err == -EMSGSIZE);
2263                 kfree_skb(skb);
2264                 goto errout;
2265         }
2266         err = rtnl_notify(skb, &init_net, info->pid,
2267                                 RTNLGRP_IPV6_ROUTE, info->nlh, gfp_any());
2268 errout:
2269         if (err < 0)
2270                 rtnl_set_sk_err(&init_net, RTNLGRP_IPV6_ROUTE, err);
2271 }
2272
2273 /*
2274  *      /proc
2275  */
2276
2277 #ifdef CONFIG_PROC_FS
2278
2279 #define RT6_INFO_LEN (32 + 4 + 32 + 4 + 32 + 40 + 5 + 1)
2280
2281 struct rt6_proc_arg
2282 {
2283         char *buffer;
2284         int offset;
2285         int length;
2286         int skip;
2287         int len;
2288 };
2289
2290 static int rt6_info_route(struct rt6_info *rt, void *p_arg)
2291 {
2292         struct seq_file *m = p_arg;
2293
2294         seq_printf(m, NIP6_SEQFMT " %02x ", NIP6(rt->rt6i_dst.addr),
2295                    rt->rt6i_dst.plen);
2296
2297 #ifdef CONFIG_IPV6_SUBTREES
2298         seq_printf(m, NIP6_SEQFMT " %02x ", NIP6(rt->rt6i_src.addr),
2299                    rt->rt6i_src.plen);
2300 #else
2301         seq_puts(m, "00000000000000000000000000000000 00 ");
2302 #endif
2303
2304         if (rt->rt6i_nexthop) {
2305                 seq_printf(m, NIP6_SEQFMT,
2306                            NIP6(*((struct in6_addr *)rt->rt6i_nexthop->primary_key)));
2307         } else {
2308                 seq_puts(m, "00000000000000000000000000000000");
2309         }
2310         seq_printf(m, " %08x %08x %08x %08x %8s\n",
2311                    rt->rt6i_metric, atomic_read(&rt->u.dst.__refcnt),
2312                    rt->u.dst.__use, rt->rt6i_flags,
2313                    rt->rt6i_dev ? rt->rt6i_dev->name : "");
2314         return 0;
2315 }
2316
2317 static int ipv6_route_show(struct seq_file *m, void *v)
2318 {
2319         fib6_clean_all(rt6_info_route, 0, m);
2320         return 0;
2321 }
2322
2323 static int ipv6_route_open(struct inode *inode, struct file *file)
2324 {
2325         return single_open(file, ipv6_route_show, NULL);
2326 }
2327
2328 static const struct file_operations ipv6_route_proc_fops = {
2329         .owner          = THIS_MODULE,
2330         .open           = ipv6_route_open,
2331         .read           = seq_read,
2332         .llseek         = seq_lseek,
2333         .release        = single_release,
2334 };
2335
2336 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
2337 {
2338         seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
2339                       rt6_stats.fib_nodes, rt6_stats.fib_route_nodes,
2340                       rt6_stats.fib_rt_alloc, rt6_stats.fib_rt_entries,
2341                       rt6_stats.fib_rt_cache,
2342                       atomic_read(&ip6_dst_ops.entries),
2343                       rt6_stats.fib_discarded_routes);
2344
2345         return 0;
2346 }
2347
2348 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
2349 {
2350         return single_open(file, rt6_stats_seq_show, NULL);
2351 }
2352
2353 static const struct file_operations rt6_stats_seq_fops = {
2354         .owner   = THIS_MODULE,
2355         .open    = rt6_stats_seq_open,
2356         .read    = seq_read,
2357         .llseek  = seq_lseek,
2358         .release = single_release,
2359 };
2360
2361 static int ipv6_route_proc_init(struct net *net)
2362 {
2363         int ret = -ENOMEM;
2364         if (!proc_net_fops_create(net, "ipv6_route",
2365                                   0, &ipv6_route_proc_fops))
2366                 goto out;
2367
2368         if (!proc_net_fops_create(net, "rt6_stats",
2369                                   S_IRUGO, &rt6_stats_seq_fops))
2370                 goto out_ipv6_route;
2371
2372         ret = 0;
2373 out:
2374         return ret;
2375 out_ipv6_route:
2376         proc_net_remove(net, "ipv6_route");
2377         goto out;
2378 }
2379
2380 static void ipv6_route_proc_fini(struct net *net)
2381 {
2382         proc_net_remove(net, "ipv6_route");
2383         proc_net_remove(net, "rt6_stats");
2384 }
2385 #else
2386 static inline int ipv6_route_proc_init(struct net *net)
2387 {
2388         return 0;
2389 }
2390 static inline void ipv6_route_proc_fini(struct net *net)
2391 {
2392         return ;
2393 }
2394 #endif  /* CONFIG_PROC_FS */
2395
2396 #ifdef CONFIG_SYSCTL
2397
2398 static
2399 int ipv6_sysctl_rtcache_flush(ctl_table *ctl, int write, struct file * filp,
2400                               void __user *buffer, size_t *lenp, loff_t *ppos)
2401 {
2402         int delay = init_net.ipv6.sysctl.flush_delay;
2403         if (write) {
2404                 proc_dointvec(ctl, write, filp, buffer, lenp, ppos);
2405                 fib6_run_gc(delay <= 0 ? ~0UL : (unsigned long)delay);
2406                 return 0;
2407         } else
2408                 return -EINVAL;
2409 }
2410
2411 ctl_table ipv6_route_table_template[] = {
2412         {
2413                 .procname       =       "flush",
2414                 .data           =       &init_net.ipv6.sysctl.flush_delay,
2415                 .maxlen         =       sizeof(int),
2416                 .mode           =       0200,
2417                 .proc_handler   =       &ipv6_sysctl_rtcache_flush
2418         },
2419         {
2420                 .ctl_name       =       NET_IPV6_ROUTE_GC_THRESH,
2421                 .procname       =       "gc_thresh",
2422                 .data           =       &ip6_dst_ops.gc_thresh,
2423                 .maxlen         =       sizeof(int),
2424                 .mode           =       0644,
2425                 .proc_handler   =       &proc_dointvec,
2426         },
2427         {
2428                 .ctl_name       =       NET_IPV6_ROUTE_MAX_SIZE,
2429                 .procname       =       "max_size",
2430                 .data           =       &init_net.ipv6.sysctl.ip6_rt_max_size,
2431                 .maxlen         =       sizeof(int),
2432                 .mode           =       0644,
2433                 .proc_handler   =       &proc_dointvec,
2434         },
2435         {
2436                 .ctl_name       =       NET_IPV6_ROUTE_GC_MIN_INTERVAL,
2437                 .procname       =       "gc_min_interval",
2438                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2439                 .maxlen         =       sizeof(int),
2440                 .mode           =       0644,
2441                 .proc_handler   =       &proc_dointvec_jiffies,
2442                 .strategy       =       &sysctl_jiffies,
2443         },
2444         {
2445                 .ctl_name       =       NET_IPV6_ROUTE_GC_TIMEOUT,
2446                 .procname       =       "gc_timeout",
2447                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
2448                 .maxlen         =       sizeof(int),
2449                 .mode           =       0644,
2450                 .proc_handler   =       &proc_dointvec_jiffies,
2451                 .strategy       =       &sysctl_jiffies,
2452         },
2453         {
2454                 .ctl_name       =       NET_IPV6_ROUTE_GC_INTERVAL,
2455                 .procname       =       "gc_interval",
2456                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_interval,
2457                 .maxlen         =       sizeof(int),
2458                 .mode           =       0644,
2459                 .proc_handler   =       &proc_dointvec_jiffies,
2460                 .strategy       =       &sysctl_jiffies,
2461         },
2462         {
2463                 .ctl_name       =       NET_IPV6_ROUTE_GC_ELASTICITY,
2464                 .procname       =       "gc_elasticity",
2465                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
2466                 .maxlen         =       sizeof(int),
2467                 .mode           =       0644,
2468                 .proc_handler   =       &proc_dointvec_jiffies,
2469                 .strategy       =       &sysctl_jiffies,
2470         },
2471         {
2472                 .ctl_name       =       NET_IPV6_ROUTE_MTU_EXPIRES,
2473                 .procname       =       "mtu_expires",
2474                 .data           =       &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
2475                 .maxlen         =       sizeof(int),
2476                 .mode           =       0644,
2477                 .proc_handler   =       &proc_dointvec_jiffies,
2478                 .strategy       =       &sysctl_jiffies,
2479         },
2480         {
2481                 .ctl_name       =       NET_IPV6_ROUTE_MIN_ADVMSS,
2482                 .procname       =       "min_adv_mss",
2483                 .data           =       &init_net.ipv6.sysctl.ip6_rt_min_advmss,
2484                 .maxlen         =       sizeof(int),
2485                 .mode           =       0644,
2486                 .proc_handler   =       &proc_dointvec_jiffies,
2487                 .strategy       =       &sysctl_jiffies,
2488         },
2489         {
2490                 .ctl_name       =       NET_IPV6_ROUTE_GC_MIN_INTERVAL_MS,
2491                 .procname       =       "gc_min_interval_ms",
2492                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2493                 .maxlen         =       sizeof(int),
2494                 .mode           =       0644,
2495                 .proc_handler   =       &proc_dointvec_ms_jiffies,
2496                 .strategy       =       &sysctl_ms_jiffies,
2497         },
2498         { .ctl_name = 0 }
2499 };
2500
2501 struct ctl_table *ipv6_route_sysctl_init(struct net *net)
2502 {
2503         struct ctl_table *table;
2504
2505         table = kmemdup(ipv6_route_table_template,
2506                         sizeof(ipv6_route_table_template),
2507                         GFP_KERNEL);
2508
2509         if (table) {
2510                 table[0].data = &net->ipv6.sysctl.flush_delay;
2511                 /* table[1].data will be handled when we have
2512                    routes per namespace */
2513                 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
2514                 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2515                 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
2516                 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
2517                 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
2518                 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
2519                 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
2520         }
2521
2522         return table;
2523 }
2524 #endif
2525
2526 int __init ip6_route_init(void)
2527 {
2528         int ret;
2529
2530         ip6_dst_ops.kmem_cachep =
2531                 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
2532                                   SLAB_HWCACHE_ALIGN, NULL);
2533         if (!ip6_dst_ops.kmem_cachep)
2534                 return -ENOMEM;
2535
2536         ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops.kmem_cachep;
2537
2538         ret = fib6_init();
2539         if (ret)
2540                 goto out_kmem_cache;
2541
2542         ret = ipv6_route_proc_init(&init_net);
2543         if (ret)
2544                 goto out_fib6_init;
2545
2546         ret = xfrm6_init();
2547         if (ret)
2548                 goto out_proc_init;
2549
2550         ret = fib6_rules_init();
2551         if (ret)
2552                 goto xfrm6_init;
2553
2554         ret = -ENOBUFS;
2555         if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL) ||
2556             __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL) ||
2557             __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL))
2558                 goto fib6_rules_init;
2559
2560         ret = 0;
2561 out:
2562         return ret;
2563
2564 fib6_rules_init:
2565         fib6_rules_cleanup();
2566 xfrm6_init:
2567         xfrm6_fini();
2568 out_proc_init:
2569         ipv6_route_proc_fini(&init_net);
2570 out_fib6_init:
2571         rt6_ifdown(NULL);
2572         fib6_gc_cleanup();
2573 out_kmem_cache:
2574         kmem_cache_destroy(ip6_dst_ops.kmem_cachep);
2575         goto out;
2576 }
2577
2578 void ip6_route_cleanup(void)
2579 {
2580         fib6_rules_cleanup();
2581         ipv6_route_proc_fini(&init_net);
2582         xfrm6_fini();
2583         rt6_ifdown(NULL);
2584         fib6_gc_cleanup();
2585         kmem_cache_destroy(ip6_dst_ops.kmem_cachep);
2586 }