SAFE public projects git trees. - safe/jmp/linux-2.6/blob - net/ipv6/route.c

   1 /*
   2  *      Linux INET6 implementation
   3  *      FIB front-end.
   4  *
   5  *      Authors:
   6  *      Pedro Roque             <roque@di.fc.ul.pt>
   7  *
   8  *      This program is free software; you can redistribute it and/or
   9  *      modify it under the terms of the GNU General Public License
  10  *      as published by the Free Software Foundation; either version
  11  *      2 of the License, or (at your option) any later version.
  12  */
  13
  14 /*      Changes:
  15  *
  16  *      YOSHIFUJI Hideaki @USAGI
  17  *              reworked default router selection.
  18  *              - respect outgoing interface
  19  *              - select from (probably) reachable routers (i.e.
  20  *              routers in REACHABLE, STALE, DELAY or PROBE states).
  21  *              - always select the same router if it is (probably)
  22  *              reachable.  otherwise, round-robin the list.
  23  *      Ville Nuorvala
  24  *              Fixed routing subtrees.
  25  */
  26
  27 #include <linux/capability.h>
  28 #include <linux/errno.h>
  29 #include <linux/types.h>
  30 #include <linux/times.h>
  31 #include <linux/socket.h>
  32 #include <linux/sockios.h>
  33 #include <linux/net.h>
  34 #include <linux/route.h>
  35 #include <linux/netdevice.h>
  36 #include <linux/in6.h>
  37 #include <linux/mroute6.h>
  38 #include <linux/init.h>
  39 #include <linux/if_arp.h>
  40 #include <linux/proc_fs.h>
  41 #include <linux/seq_file.h>
  42 #include <linux/nsproxy.h>
  43 #include <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 net *net,
  91                                            struct in6_addr *prefix, int prefixlen,
  92                                            struct in6_addr *gwaddr, int ifindex,
  93                                            unsigned pref);
  94 static struct rt6_info *rt6_get_route_info(struct net *net,
  95                                            struct in6_addr *prefix, int prefixlen,
  96                                            struct in6_addr *gwaddr, int ifindex);
  97 #endif
  98
  99 static struct dst_ops ip6_dst_ops_template = {
 100         .family                 =       AF_INET6,
 101         .protocol               =       cpu_to_be16(ETH_P_IPV6),
 102         .gc                     =       ip6_dst_gc,
 103         .gc_thresh              =       1024,
 104         .check                  =       ip6_dst_check,
 105         .destroy                =       ip6_dst_destroy,
 106         .ifdown                 =       ip6_dst_ifdown,
 107         .negative_advice        =       ip6_negative_advice,
 108         .link_failure           =       ip6_link_failure,
 109         .update_pmtu            =       ip6_rt_update_pmtu,
 110         .local_out              =       __ip6_local_out,
 111         .entries                =       ATOMIC_INIT(0),
 112 };
 113
 114 static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, u32 mtu)
 115 {
 116 }
 117
 118 static struct dst_ops ip6_dst_blackhole_ops = {
 119         .family                 =       AF_INET6,
 120         .protocol               =       cpu_to_be16(ETH_P_IPV6),
 121         .destroy                =       ip6_dst_destroy,
 122         .check                  =       ip6_dst_check,
 123         .update_pmtu            =       ip6_rt_blackhole_update_pmtu,
 124         .entries                =       ATOMIC_INIT(0),
 125 };
 126
 127 static struct rt6_info ip6_null_entry_template = {
 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                 }
 138         },
 139         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
 140         .rt6i_protocol  = RTPROT_KERNEL,
 141         .rt6i_metric    = ~(u32) 0,
 142         .rt6i_ref       = ATOMIC_INIT(1),
 143 };
 144
 145 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
 146
 147 static int ip6_pkt_prohibit(struct sk_buff *skb);
 148 static int ip6_pkt_prohibit_out(struct sk_buff *skb);
 149
 150 static struct rt6_info ip6_prohibit_entry_template = {
 151         .u = {
 152                 .dst = {
 153                         .__refcnt       = ATOMIC_INIT(1),
 154                         .__use          = 1,
 155                         .obsolete       = -1,
 156                         .error          = -EACCES,
 157                         .metrics        = { [RTAX_HOPLIMIT - 1] = 255, },
 158                         .input          = ip6_pkt_prohibit,
 159                         .output         = ip6_pkt_prohibit_out,
 160                 }
 161         },
 162         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
 163         .rt6i_protocol  = RTPROT_KERNEL,
 164         .rt6i_metric    = ~(u32) 0,
 165         .rt6i_ref       = ATOMIC_INIT(1),
 166 };
 167
 168 static struct rt6_info ip6_blk_hole_entry_template = {
 169         .u = {
 170                 .dst = {
 171                         .__refcnt       = ATOMIC_INIT(1),
 172                         .__use          = 1,
 173                         .obsolete       = -1,
 174                         .error          = -EINVAL,
 175                         .metrics        = { [RTAX_HOPLIMIT - 1] = 255, },
 176                         .input          = dst_discard,
 177                         .output         = dst_discard,
 178                 }
 179         },
 180         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
 181         .rt6i_protocol  = RTPROT_KERNEL,
 182         .rt6i_metric    = ~(u32) 0,
 183         .rt6i_ref       = ATOMIC_INIT(1),
 184 };
 185
 186 #endif
 187
 188 /* allocate dst with ip6_dst_ops */
 189 static inline struct rt6_info *ip6_dst_alloc(struct dst_ops *ops)
 190 {
 191         return (struct rt6_info *)dst_alloc(ops);
 192 }
 193
 194 static void ip6_dst_destroy(struct dst_entry *dst)
 195 {
 196         struct rt6_info *rt = (struct rt6_info *)dst;
 197         struct inet6_dev *idev = rt->rt6i_idev;
 198
 199         if (idev != NULL) {
 200                 rt->rt6i_idev = NULL;
 201                 in6_dev_put(idev);
 202         }
 203 }
 204
 205 static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
 206                            int how)
 207 {
 208         struct rt6_info *rt = (struct rt6_info *)dst;
 209         struct inet6_dev *idev = rt->rt6i_idev;
 210         struct net_device *loopback_dev =
 211                 dev_net(dev)->loopback_dev;
 212
 213         if (dev != loopback_dev && idev != NULL && idev->dev == dev) {
 214                 struct inet6_dev *loopback_idev =
 215                         in6_dev_get(loopback_dev);
 216                 if (loopback_idev != NULL) {
 217                         rt->rt6i_idev = loopback_idev;
 218                         in6_dev_put(idev);
 219                 }
 220         }
 221 }
 222
 223 static __inline__ int rt6_check_expired(const struct rt6_info *rt)
 224 {
 225         return (rt->rt6i_flags & RTF_EXPIRES &&
 226                 time_after(jiffies, rt->rt6i_expires));
 227 }
 228
 229 static inline int rt6_need_strict(struct in6_addr *daddr)
 230 {
 231         return (ipv6_addr_type(daddr) &
 232                 (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK));
 233 }
 234
 235 /*
 236  *      Route lookup. Any table->tb6_lock is implied.
 237  */
 238
 239 static inline struct rt6_info *rt6_device_match(struct net *net,
 240                                                     struct rt6_info *rt,
 241                                                     struct in6_addr *saddr,
 242                                                     int oif,
 243                                                     int flags)
 244 {
 245         struct rt6_info *local = NULL;
 246         struct rt6_info *sprt;
 247
 248         if (!oif && ipv6_addr_any(saddr))
 249                 goto out;
 250
 251         for (sprt = rt; sprt; sprt = sprt->u.dst.rt6_next) {
 252                 struct net_device *dev = sprt->rt6i_dev;
 253
 254                 if (oif) {
 255                         if (dev->ifindex == oif)
 256                                 return sprt;
 257                         if (dev->flags & IFF_LOOPBACK) {
 258                                 if (sprt->rt6i_idev == NULL ||
 259                                     sprt->rt6i_idev->dev->ifindex != oif) {
 260                                         if (flags & RT6_LOOKUP_F_IFACE && oif)
 261                                                 continue;
 262                                         if (local && (!oif ||
 263                                                       local->rt6i_idev->dev->ifindex == oif))
 264                                                 continue;
 265                                 }
 266                                 local = sprt;
 267                         }
 268                 } else {
 269                         if (ipv6_chk_addr(net, saddr, dev,
 270                                           flags & RT6_LOOKUP_F_IFACE))
 271                                 return sprt;
 272                 }
 273         }
 274
 275         if (oif) {
 276                 if (local)
 277                         return local;
 278
 279                 if (flags & RT6_LOOKUP_F_IFACE)
 280                         return net->ipv6.ip6_null_entry;
 281         }
 282 out:
 283         return rt;
 284 }
 285
 286 #ifdef CONFIG_IPV6_ROUTER_PREF
 287 static void rt6_probe(struct rt6_info *rt)
 288 {
 289         struct neighbour *neigh = rt ? rt->rt6i_nexthop : NULL;
 290         /*
 291          * Okay, this does not seem to be appropriate
 292          * for now, however, we need to check if it
 293          * is really so; aka Router Reachability Probing.
 294          *
 295          * Router Reachability Probe MUST be rate-limited
 296          * to no more than one per minute.
 297          */
 298         if (!neigh || (neigh->nud_state & NUD_VALID))
 299                 return;
 300         read_lock_bh(&neigh->lock);
 301         if (!(neigh->nud_state & NUD_VALID) &&
 302             time_after(jiffies, neigh->updated + rt->rt6i_idev->cnf.rtr_probe_interval)) {
 303                 struct in6_addr mcaddr;
 304                 struct in6_addr *target;
 305
 306                 neigh->updated = jiffies;
 307                 read_unlock_bh(&neigh->lock);
 308
 309                 target = (struct in6_addr *)&neigh->primary_key;
 310                 addrconf_addr_solict_mult(target, &mcaddr);
 311                 ndisc_send_ns(rt->rt6i_dev, NULL, target, &mcaddr, NULL);
 312         } else
 313                 read_unlock_bh(&neigh->lock);
 314 }
 315 #else
 316 static inline void rt6_probe(struct rt6_info *rt)
 317 {
 318         return;
 319 }
 320 #endif
 321
 322 /*
 323  * Default Router Selection (RFC 2461 6.3.6)
 324  */
 325 static inline int rt6_check_dev(struct rt6_info *rt, int oif)
 326 {
 327         struct net_device *dev = rt->rt6i_dev;
 328         if (!oif || dev->ifindex == oif)
 329                 return 2;
 330         if ((dev->flags & IFF_LOOPBACK) &&
 331             rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
 332                 return 1;
 333         return 0;
 334 }
 335
 336 static inline int rt6_check_neigh(struct rt6_info *rt)
 337 {
 338         struct neighbour *neigh = rt->rt6i_nexthop;
 339         int m;
 340         if (rt->rt6i_flags & RTF_NONEXTHOP ||
 341             !(rt->rt6i_flags & RTF_GATEWAY))
 342                 m = 1;
 343         else if (neigh) {
 344                 read_lock_bh(&neigh->lock);
 345                 if (neigh->nud_state & NUD_VALID)
 346                         m = 2;
 347 #ifdef CONFIG_IPV6_ROUTER_PREF
 348                 else if (neigh->nud_state & NUD_FAILED)
 349                         m = 0;
 350 #endif
 351                 else
 352                         m = 1;
 353                 read_unlock_bh(&neigh->lock);
 354         } else
 355                 m = 0;
 356         return m;
 357 }
 358
 359 static int rt6_score_route(struct rt6_info *rt, int oif,
 360                            int strict)
 361 {
 362         int m, n;
 363
 364         m = rt6_check_dev(rt, oif);
 365         if (!m && (strict & RT6_LOOKUP_F_IFACE))
 366                 return -1;
 367 #ifdef CONFIG_IPV6_ROUTER_PREF
 368         m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
 369 #endif
 370         n = rt6_check_neigh(rt);
 371         if (!n && (strict & RT6_LOOKUP_F_REACHABLE))
 372                 return -1;
 373         return m;
 374 }
 375
 376 static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
 377                                    int *mpri, struct rt6_info *match)
 378 {
 379         int m;
 380
 381         if (rt6_check_expired(rt))
 382                 goto out;
 383
 384         m = rt6_score_route(rt, oif, strict);
 385         if (m < 0)
 386                 goto out;
 387
 388         if (m > *mpri) {
 389                 if (strict & RT6_LOOKUP_F_REACHABLE)
 390                         rt6_probe(match);
 391                 *mpri = m;
 392                 match = rt;
 393         } else if (strict & RT6_LOOKUP_F_REACHABLE) {
 394                 rt6_probe(rt);
 395         }
 396
 397 out:
 398         return match;
 399 }
 400
 401 static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
 402                                      struct rt6_info *rr_head,
 403                                      u32 metric, int oif, int strict)
 404 {
 405         struct rt6_info *rt, *match;
 406         int mpri = -1;
 407
 408         match = NULL;
 409         for (rt = rr_head; rt && rt->rt6i_metric == metric;
 410              rt = rt->u.dst.rt6_next)
 411                 match = find_match(rt, oif, strict, &mpri, match);
 412         for (rt = fn->leaf; rt && rt != rr_head && rt->rt6i_metric == metric;
 413              rt = rt->u.dst.rt6_next)
 414                 match = find_match(rt, oif, strict, &mpri, match);
 415
 416         return match;
 417 }
 418
 419 static struct rt6_info *rt6_select(struct fib6_node *fn, int oif, int strict)
 420 {
 421         struct rt6_info *match, *rt0;
 422         struct net *net;
 423
 424         RT6_TRACE("%s(fn->leaf=%p, oif=%d)\n",
 425                   __func__, fn->leaf, oif);
 426
 427         rt0 = fn->rr_ptr;
 428         if (!rt0)
 429                 fn->rr_ptr = rt0 = fn->leaf;
 430
 431         match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict);
 432
 433         if (!match &&
 434             (strict & RT6_LOOKUP_F_REACHABLE)) {
 435                 struct rt6_info *next = rt0->u.dst.rt6_next;
 436
 437                 /* no entries matched; do round-robin */
 438                 if (!next || next->rt6i_metric != rt0->rt6i_metric)
 439                         next = fn->leaf;
 440
 441                 if (next != rt0)
 442                         fn->rr_ptr = next;
 443         }
 444
 445         RT6_TRACE("%s() => %p\n",
 446                   __func__, match);
 447
 448         net = dev_net(rt0->rt6i_dev);
 449         return (match ? match : net->ipv6.ip6_null_entry);
 450 }
 451
 452 #ifdef CONFIG_IPV6_ROUTE_INFO
 453 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
 454                   struct in6_addr *gwaddr)
 455 {
 456         struct net *net = dev_net(dev);
 457         struct route_info *rinfo = (struct route_info *) opt;
 458         struct in6_addr prefix_buf, *prefix;
 459         unsigned int pref;
 460         unsigned long lifetime;
 461         struct rt6_info *rt;
 462
 463         if (len < sizeof(struct route_info)) {
 464                 return -EINVAL;
 465         }
 466
 467         /* Sanity check for prefix_len and length */
 468         if (rinfo->length > 3) {
 469                 return -EINVAL;
 470         } else if (rinfo->prefix_len > 128) {
 471                 return -EINVAL;
 472         } else if (rinfo->prefix_len > 64) {
 473                 if (rinfo->length < 2) {
 474                         return -EINVAL;
 475                 }
 476         } else if (rinfo->prefix_len > 0) {
 477                 if (rinfo->length < 1) {
 478                         return -EINVAL;
 479                 }
 480         }
 481
 482         pref = rinfo->route_pref;
 483         if (pref == ICMPV6_ROUTER_PREF_INVALID)
 484                 return -EINVAL;
 485
 486         lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
 487
 488         if (rinfo->length == 3)
 489                 prefix = (struct in6_addr *)rinfo->prefix;
 490         else {
 491                 /* this function is safe */
 492                 ipv6_addr_prefix(&prefix_buf,
 493                                  (struct in6_addr *)rinfo->prefix,
 494                                  rinfo->prefix_len);
 495                 prefix = &prefix_buf;
 496         }
 497
 498         rt = rt6_get_route_info(net, prefix, rinfo->prefix_len, gwaddr,
 499                                 dev->ifindex);
 500
 501         if (rt && !lifetime) {
 502                 ip6_del_rt(rt);
 503                 rt = NULL;
 504         }
 505
 506         if (!rt && lifetime)
 507                 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr, dev->ifindex,
 508                                         pref);
 509         else if (rt)
 510                 rt->rt6i_flags = RTF_ROUTEINFO |
 511                                  (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
 512
 513         if (rt) {
 514                 if (!addrconf_finite_timeout(lifetime)) {
 515                         rt->rt6i_flags &= ~RTF_EXPIRES;
 516                 } else {
 517                         rt->rt6i_expires = jiffies + HZ * lifetime;
 518                         rt->rt6i_flags |= RTF_EXPIRES;
 519                 }
 520                 dst_release(&rt->u.dst);
 521         }
 522         return 0;
 523 }
 524 #endif
 525
 526 #define BACKTRACK(__net, saddr)                 \
 527 do { \
 528         if (rt == __net->ipv6.ip6_null_entry) { \
 529                 struct fib6_node *pn; \
 530                 while (1) { \
 531                         if (fn->fn_flags & RTN_TL_ROOT) \
 532                                 goto out; \
 533                         pn = fn->parent; \
 534                         if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn) \
 535                                 fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr); \
 536                         else \
 537                                 fn = pn; \
 538                         if (fn->fn_flags & RTN_RTINFO) \
 539                                 goto restart; \
 540                 } \
 541         } \
 542 } while(0)
 543
 544 static struct rt6_info *ip6_pol_route_lookup(struct net *net,
 545                                              struct fib6_table *table,
 546                                              struct flowi *fl, int flags)
 547 {
 548         struct fib6_node *fn;
 549         struct rt6_info *rt;
 550
 551         read_lock_bh(&table->tb6_lock);
 552         fn = fib6_lookup(&table->tb6_root, &fl->fl6_dst, &fl->fl6_src);
 553 restart:
 554         rt = fn->leaf;
 555         rt = rt6_device_match(net, rt, &fl->fl6_src, fl->oif, flags);
 556         BACKTRACK(net, &fl->fl6_src);
 557 out:
 558         dst_use(&rt->u.dst, jiffies);
 559         read_unlock_bh(&table->tb6_lock);
 560         return rt;
 561
 562 }
 563
 564 struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
 565                             const struct in6_addr *saddr, int oif, int strict)
 566 {
 567         struct flowi fl = {
 568                 .oif = oif,
 569                 .nl_u = {
 570                         .ip6_u = {
 571                                 .daddr = *daddr,
 572                         },
 573                 },
 574         };
 575         struct dst_entry *dst;
 576         int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
 577
 578         if (saddr) {
 579                 memcpy(&fl.fl6_src, saddr, sizeof(*saddr));
 580                 flags |= RT6_LOOKUP_F_HAS_SADDR;
 581         }
 582
 583         dst = fib6_rule_lookup(net, &fl, flags, ip6_pol_route_lookup);
 584         if (dst->error == 0)
 585                 return (struct rt6_info *) dst;
 586
 587         dst_release(dst);
 588
 589         return NULL;
 590 }
 591
 592 EXPORT_SYMBOL(rt6_lookup);
 593
 594 /* ip6_ins_rt is called with FREE table->tb6_lock.
 595    It takes new route entry, the addition fails by any reason the
 596    route is freed. In any case, if caller does not hold it, it may
 597    be destroyed.
 598  */
 599
 600 static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info)
 601 {
 602         int err;
 603         struct fib6_table *table;
 604
 605         table = rt->rt6i_table;
 606         write_lock_bh(&table->tb6_lock);
 607         err = fib6_add(&table->tb6_root, rt, info);
 608         write_unlock_bh(&table->tb6_lock);
 609
 610         return err;
 611 }
 612
 613 int ip6_ins_rt(struct rt6_info *rt)
 614 {
 615         struct nl_info info = {
 616                 .nl_net = dev_net(rt->rt6i_dev),
 617         };
 618         return __ip6_ins_rt(rt, &info);
 619 }
 620
 621 static struct rt6_info *rt6_alloc_cow(struct rt6_info *ort, struct in6_addr *daddr,
 622                                       struct in6_addr *saddr)
 623 {
 624         struct rt6_info *rt;
 625
 626         /*
 627          *      Clone the route.
 628          */
 629
 630         rt = ip6_rt_copy(ort);
 631
 632         if (rt) {
 633                 struct neighbour *neigh;
 634                 int attempts = !in_softirq();
 635
 636                 if (!(rt->rt6i_flags&RTF_GATEWAY)) {
 637                         if (rt->rt6i_dst.plen != 128 &&
 638                             ipv6_addr_equal(&rt->rt6i_dst.addr, daddr))
 639                                 rt->rt6i_flags |= RTF_ANYCAST;
 640                         ipv6_addr_copy(&rt->rt6i_gateway, daddr);
 641                 }
 642
 643                 ipv6_addr_copy(&rt->rt6i_dst.addr, daddr);
 644                 rt->rt6i_dst.plen = 128;
 645                 rt->rt6i_flags |= RTF_CACHE;
 646                 rt->u.dst.flags |= DST_HOST;
 647
 648 #ifdef CONFIG_IPV6_SUBTREES
 649                 if (rt->rt6i_src.plen && saddr) {
 650                         ipv6_addr_copy(&rt->rt6i_src.addr, saddr);
 651                         rt->rt6i_src.plen = 128;
 652                 }
 653 #endif
 654
 655         retry:
 656                 neigh = ndisc_get_neigh(rt->rt6i_dev, &rt->rt6i_gateway);
 657                 if (IS_ERR(neigh)) {
 658                         struct net *net = dev_net(rt->rt6i_dev);
 659                         int saved_rt_min_interval =
 660                                 net->ipv6.sysctl.ip6_rt_gc_min_interval;
 661                         int saved_rt_elasticity =
 662                                 net->ipv6.sysctl.ip6_rt_gc_elasticity;
 663
 664                         if (attempts-- > 0) {
 665                                 net->ipv6.sysctl.ip6_rt_gc_elasticity = 1;
 666                                 net->ipv6.sysctl.ip6_rt_gc_min_interval = 0;
 667
 668                                 ip6_dst_gc(&net->ipv6.ip6_dst_ops);
 669
 670                                 net->ipv6.sysctl.ip6_rt_gc_elasticity =
 671                                         saved_rt_elasticity;
 672                                 net->ipv6.sysctl.ip6_rt_gc_min_interval =
 673                                         saved_rt_min_interval;
 674                                 goto retry;
 675                         }
 676
 677                         if (net_ratelimit())
 678                                 printk(KERN_WARNING
 679                                        "Neighbour table overflow.\n");
 680                         dst_free(&rt->u.dst);
 681                         return NULL;
 682                 }
 683                 rt->rt6i_nexthop = neigh;
 684
 685         }
 686
 687         return rt;
 688 }
 689
 690 static struct rt6_info *rt6_alloc_clone(struct rt6_info *ort, struct in6_addr *daddr)
 691 {
 692         struct rt6_info *rt = ip6_rt_copy(ort);
 693         if (rt) {
 694                 ipv6_addr_copy(&rt->rt6i_dst.addr, daddr);
 695                 rt->rt6i_dst.plen = 128;
 696                 rt->rt6i_flags |= RTF_CACHE;
 697                 rt->u.dst.flags |= DST_HOST;
 698                 rt->rt6i_nexthop = neigh_clone(ort->rt6i_nexthop);
 699         }
 700         return rt;
 701 }
 702
 703 static struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table, int oif,
 704                                       struct flowi *fl, int flags)
 705 {
 706         struct fib6_node *fn;
 707         struct rt6_info *rt, *nrt;
 708         int strict = 0;
 709         int attempts = 3;
 710         int err;
 711         int reachable = net->ipv6.devconf_all->forwarding ? 0 : RT6_LOOKUP_F_REACHABLE;
 712
 713         strict |= flags & RT6_LOOKUP_F_IFACE;
 714
 715 relookup:
 716         read_lock_bh(&table->tb6_lock);
 717
 718 restart_2:
 719         fn = fib6_lookup(&table->tb6_root, &fl->fl6_dst, &fl->fl6_src);
 720
 721 restart:
 722         rt = rt6_select(fn, oif, strict | reachable);
 723
 724         BACKTRACK(net, &fl->fl6_src);
 725         if (rt == net->ipv6.ip6_null_entry ||
 726             rt->rt6i_flags & RTF_CACHE)
 727                 goto out;
 728
 729         dst_hold(&rt->u.dst);
 730         read_unlock_bh(&table->tb6_lock);
 731
 732         if (!rt->rt6i_nexthop && !(rt->rt6i_flags & RTF_NONEXTHOP))
 733                 nrt = rt6_alloc_cow(rt, &fl->fl6_dst, &fl->fl6_src);
 734         else {
 735 #if CLONE_OFFLINK_ROUTE
 736                 nrt = rt6_alloc_clone(rt, &fl->fl6_dst);
 737 #else
 738                 goto out2;
 739 #endif
 740         }
 741
 742         dst_release(&rt->u.dst);
 743         rt = nrt ? : net->ipv6.ip6_null_entry;
 744
 745         dst_hold(&rt->u.dst);
 746         if (nrt) {
 747                 err = ip6_ins_rt(nrt);
 748                 if (!err)
 749                         goto out2;
 750         }
 751
 752         if (--attempts <= 0)
 753                 goto out2;
 754
 755         /*
 756          * Race condition! In the gap, when table->tb6_lock was
 757          * released someone could insert this route.  Relookup.
 758          */
 759         dst_release(&rt->u.dst);
 760         goto relookup;
 761
 762 out:
 763         if (reachable) {
 764                 reachable = 0;
 765                 goto restart_2;
 766         }
 767         dst_hold(&rt->u.dst);
 768         read_unlock_bh(&table->tb6_lock);
 769 out2:
 770         rt->u.dst.lastuse = jiffies;
 771         rt->u.dst.__use++;
 772
 773         return rt;
 774 }
 775
 776 static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
 777                                             struct flowi *fl, int flags)
 778 {
 779         return ip6_pol_route(net, table, fl->iif, fl, flags);
 780 }
 781
 782 void ip6_route_input(struct sk_buff *skb)
 783 {
 784         struct ipv6hdr *iph = ipv6_hdr(skb);
 785         struct net *net = dev_net(skb->dev);
 786         int flags = RT6_LOOKUP_F_HAS_SADDR;
 787         struct flowi fl = {
 788                 .iif = skb->dev->ifindex,
 789                 .nl_u = {
 790                         .ip6_u = {
 791                                 .daddr = iph->daddr,
 792                                 .saddr = iph->saddr,
 793                                 .flowlabel = (* (__be32 *) iph)&IPV6_FLOWINFO_MASK,
 794                         },
 795                 },
 796                 .mark = skb->mark,
 797                 .proto = iph->nexthdr,
 798         };
 799
 800         if (rt6_need_strict(&iph->daddr) && skb->dev->type != ARPHRD_PIMREG)
 801                 flags |= RT6_LOOKUP_F_IFACE;
 802
 803         skb_dst_set(skb, fib6_rule_lookup(net, &fl, flags, ip6_pol_route_input));
 804 }
 805
 806 static struct rt6_info *ip6_pol_route_output(struct net *net, struct fib6_table *table,
 807                                              struct flowi *fl, int flags)
 808 {
 809         return ip6_pol_route(net, table, fl->oif, fl, flags);
 810 }
 811
 812 struct dst_entry * ip6_route_output(struct net *net, struct sock *sk,
 813                                     struct flowi *fl)
 814 {
 815         int flags = 0;
 816
 817         if (rt6_need_strict(&fl->fl6_dst))
 818                 flags |= RT6_LOOKUP_F_IFACE;
 819
 820         if (!ipv6_addr_any(&fl->fl6_src))
 821                 flags |= RT6_LOOKUP_F_HAS_SADDR;
 822         else if (sk) {
 823                 unsigned int prefs = inet6_sk(sk)->srcprefs;
 824                 if (prefs & IPV6_PREFER_SRC_TMP)
 825                         flags |= RT6_LOOKUP_F_SRCPREF_TMP;
 826                 if (prefs & IPV6_PREFER_SRC_PUBLIC)
 827                         flags |= RT6_LOOKUP_F_SRCPREF_PUBLIC;
 828                 if (prefs & IPV6_PREFER_SRC_COA)
 829                         flags |= RT6_LOOKUP_F_SRCPREF_COA;
 830         }
 831
 832         return fib6_rule_lookup(net, fl, flags, ip6_pol_route_output);
 833 }
 834
 835 EXPORT_SYMBOL(ip6_route_output);
 836
 837 int ip6_dst_blackhole(struct sock *sk, struct dst_entry **dstp, struct flowi *fl)
 838 {
 839         struct rt6_info *ort = (struct rt6_info *) *dstp;
 840         struct rt6_info *rt = (struct rt6_info *)
 841                 dst_alloc(&ip6_dst_blackhole_ops);
 842         struct dst_entry *new = NULL;
 843
 844         if (rt) {
 845                 new = &rt->u.dst;
 846
 847                 atomic_set(&new->__refcnt, 1);
 848                 new->__use = 1;
 849                 new->input = dst_discard;
 850                 new->output = dst_discard;
 851
 852                 memcpy(new->metrics, ort->u.dst.metrics, RTAX_MAX*sizeof(u32));
 853                 new->dev = ort->u.dst.dev;
 854                 if (new->dev)
 855                         dev_hold(new->dev);
 856                 rt->rt6i_idev = ort->rt6i_idev;
 857                 if (rt->rt6i_idev)
 858                         in6_dev_hold(rt->rt6i_idev);
 859                 rt->rt6i_expires = 0;
 860
 861                 ipv6_addr_copy(&rt->rt6i_gateway, &ort->rt6i_gateway);
 862                 rt->rt6i_flags = ort->rt6i_flags & ~RTF_EXPIRES;
 863                 rt->rt6i_metric = 0;
 864
 865                 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
 866 #ifdef CONFIG_IPV6_SUBTREES
 867                 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
 868 #endif
 869
 870                 dst_free(new);
 871         }
 872
 873         dst_release(*dstp);
 874         *dstp = new;
 875         return (new ? 0 : -ENOMEM);
 876 }
 877 EXPORT_SYMBOL_GPL(ip6_dst_blackhole);
 878
 879 /*
 880  *      Destination cache support functions
 881  */
 882
 883 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
 884 {
 885         struct rt6_info *rt;
 886
 887         rt = (struct rt6_info *) dst;
 888
 889         if (rt && rt->rt6i_node && (rt->rt6i_node->fn_sernum == cookie))
 890                 return dst;
 891
 892         return NULL;
 893 }
 894
 895 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
 896 {
 897         struct rt6_info *rt = (struct rt6_info *) dst;
 898
 899         if (rt) {
 900                 if (rt->rt6i_flags & RTF_CACHE)
 901                         ip6_del_rt(rt);
 902                 else
 903                         dst_release(dst);
 904         }
 905         return NULL;
 906 }
 907
 908 static void ip6_link_failure(struct sk_buff *skb)
 909 {
 910         struct rt6_info *rt;
 911
 912         icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0, skb->dev);
 913
 914         rt = (struct rt6_info *) skb_dst(skb);
 915         if (rt) {
 916                 if (rt->rt6i_flags&RTF_CACHE) {
 917                         dst_set_expires(&rt->u.dst, 0);
 918                         rt->rt6i_flags |= RTF_EXPIRES;
 919                 } else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT))
 920                         rt->rt6i_node->fn_sernum = -1;
 921         }
 922 }
 923
 924 static void ip6_rt_update_pmtu(struct dst_entry *dst, u32 mtu)
 925 {
 926         struct rt6_info *rt6 = (struct rt6_info*)dst;
 927
 928         if (mtu < dst_mtu(dst) && rt6->rt6i_dst.plen == 128) {
 929                 rt6->rt6i_flags |= RTF_MODIFIED;
 930                 if (mtu < IPV6_MIN_MTU) {
 931                         mtu = IPV6_MIN_MTU;
 932                         dst->metrics[RTAX_FEATURES-1] |= RTAX_FEATURE_ALLFRAG;
 933                 }
 934                 dst->metrics[RTAX_MTU-1] = mtu;
 935                 call_netevent_notifiers(NETEVENT_PMTU_UPDATE, dst);
 936         }
 937 }
 938
 939 static int ipv6_get_mtu(struct net_device *dev);
 940
 941 static inline unsigned int ipv6_advmss(struct net *net, unsigned int mtu)
 942 {
 943         mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
 944
 945         if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
 946                 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
 947
 948         /*
 949          * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
 950          * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
 951          * IPV6_MAXPLEN is also valid and means: "any MSS,
 952          * rely only on pmtu discovery"
 953          */
 954         if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
 955                 mtu = IPV6_MAXPLEN;
 956         return mtu;
 957 }
 958
 959 static struct dst_entry *icmp6_dst_gc_list;
 960 static DEFINE_SPINLOCK(icmp6_dst_lock);
 961
 962 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
 963                                   struct neighbour *neigh,
 964                                   const struct in6_addr *addr)
 965 {
 966         struct rt6_info *rt;
 967         struct inet6_dev *idev = in6_dev_get(dev);
 968         struct net *net = dev_net(dev);
 969
 970         if (unlikely(idev == NULL))
 971                 return NULL;
 972
 973         rt = ip6_dst_alloc(&net->ipv6.ip6_dst_ops);
 974         if (unlikely(rt == NULL)) {
 975                 in6_dev_put(idev);
 976                 goto out;
 977         }
 978
 979         dev_hold(dev);
 980         if (neigh)
 981                 neigh_hold(neigh);
 982         else {
 983                 neigh = ndisc_get_neigh(dev, addr);
 984                 if (IS_ERR(neigh))
 985                         neigh = NULL;
 986         }
 987
 988         rt->rt6i_dev      = dev;
 989         rt->rt6i_idev     = idev;
 990         rt->rt6i_nexthop  = neigh;
 991         atomic_set(&rt->u.dst.__refcnt, 1);
 992         rt->u.dst.metrics[RTAX_HOPLIMIT-1] = 255;
 993         rt->u.dst.metrics[RTAX_MTU-1] = ipv6_get_mtu(rt->rt6i_dev);
 994         rt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(net, dst_mtu(&rt->u.dst));
 995         rt->u.dst.output  = ip6_output;
 996
 997 #if 0   /* there's no chance to use these for ndisc */
 998         rt->u.dst.flags   = ipv6_addr_type(addr) & IPV6_ADDR_UNICAST
 999                                 ? DST_HOST
1000                                 : 0;
1001         ipv6_addr_copy(&rt->rt6i_dst.addr, addr);
1002         rt->rt6i_dst.plen = 128;
1003 #endif
1004
1005         spin_lock_bh(&icmp6_dst_lock);
1006         rt->u.dst.next = icmp6_dst_gc_list;
1007         icmp6_dst_gc_list = &rt->u.dst;
1008         spin_unlock_bh(&icmp6_dst_lock);
1009
1010         fib6_force_start_gc(net);
1011
1012 out:
1013         return &rt->u.dst;
1014 }
1015
1016 int icmp6_dst_gc(void)
1017 {
1018         struct dst_entry *dst, *next, **pprev;
1019         int more = 0;
1020
1021         next = NULL;
1022
1023         spin_lock_bh(&icmp6_dst_lock);
1024         pprev = &icmp6_dst_gc_list;
1025
1026         while ((dst = *pprev) != NULL) {
1027                 if (!atomic_read(&dst->__refcnt)) {
1028                         *pprev = dst->next;
1029                         dst_free(dst);
1030                 } else {
1031                         pprev = &dst->next;
1032                         ++more;
1033                 }
1034         }
1035
1036         spin_unlock_bh(&icmp6_dst_lock);
1037
1038         return more;
1039 }
1040
1041 static void icmp6_clean_all(int (*func)(struct rt6_info *rt, void *arg),
1042                             void *arg)
1043 {
1044         struct dst_entry *dst, **pprev;
1045
1046         spin_lock_bh(&icmp6_dst_lock);
1047         pprev = &icmp6_dst_gc_list;
1048         while ((dst = *pprev) != NULL) {
1049                 struct rt6_info *rt = (struct rt6_info *) dst;
1050                 if (func(rt, arg)) {
1051                         *pprev = dst->next;
1052                         dst_free(dst);
1053                 } else {
1054                         pprev = &dst->next;
1055                 }
1056         }
1057         spin_unlock_bh(&icmp6_dst_lock);
1058 }
1059
1060 static int ip6_dst_gc(struct dst_ops *ops)
1061 {
1062         unsigned long now = jiffies;
1063         struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
1064         int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1065         int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1066         int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1067         int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1068         unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1069
1070         if (time_after(rt_last_gc + rt_min_interval, now) &&
1071             atomic_read(&ops->entries) <= rt_max_size)
1072                 goto out;
1073
1074         net->ipv6.ip6_rt_gc_expire++;
1075         fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net);
1076         net->ipv6.ip6_rt_last_gc = now;
1077         if (atomic_read(&ops->entries) < ops->gc_thresh)
1078                 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1079 out:
1080         net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1081         return (atomic_read(&ops->entries) > rt_max_size);
1082 }
1083
1084 /* Clean host part of a prefix. Not necessary in radix tree,
1085    but results in cleaner routing tables.
1086
1087    Remove it only when all the things will work!
1088  */
1089
1090 static int ipv6_get_mtu(struct net_device *dev)
1091 {
1092         int mtu = IPV6_MIN_MTU;
1093         struct inet6_dev *idev;
1094
1095         idev = in6_dev_get(dev);
1096         if (idev) {
1097                 mtu = idev->cnf.mtu6;
1098                 in6_dev_put(idev);
1099         }
1100         return mtu;
1101 }
1102
1103 int ip6_dst_hoplimit(struct dst_entry *dst)
1104 {
1105         int hoplimit = dst_metric(dst, RTAX_HOPLIMIT);
1106         if (hoplimit < 0) {
1107                 struct net_device *dev = dst->dev;
1108                 struct inet6_dev *idev = in6_dev_get(dev);
1109                 if (idev) {
1110                         hoplimit = idev->cnf.hop_limit;
1111                         in6_dev_put(idev);
1112                 } else
1113                         hoplimit = dev_net(dev)->ipv6.devconf_all->hop_limit;
1114         }
1115         return hoplimit;
1116 }
1117
1118 /*
1119  *
1120  */
1121
1122 int ip6_route_add(struct fib6_config *cfg)
1123 {
1124         int err;
1125         struct net *net = cfg->fc_nlinfo.nl_net;
1126         struct rt6_info *rt = NULL;
1127         struct net_device *dev = NULL;
1128         struct inet6_dev *idev = NULL;
1129         struct fib6_table *table;
1130         int addr_type;
1131
1132         if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1133                 return -EINVAL;
1134 #ifndef CONFIG_IPV6_SUBTREES
1135         if (cfg->fc_src_len)
1136                 return -EINVAL;
1137 #endif
1138         if (cfg->fc_ifindex) {
1139                 err = -ENODEV;
1140                 dev = dev_get_by_index(net, cfg->fc_ifindex);
1141                 if (!dev)
1142                         goto out;
1143                 idev = in6_dev_get(dev);
1144                 if (!idev)
1145                         goto out;
1146         }
1147
1148         if (cfg->fc_metric == 0)
1149                 cfg->fc_metric = IP6_RT_PRIO_USER;
1150
1151         table = fib6_new_table(net, cfg->fc_table);
1152         if (table == NULL) {
1153                 err = -ENOBUFS;
1154                 goto out;
1155         }
1156
1157         rt = ip6_dst_alloc(&net->ipv6.ip6_dst_ops);
1158
1159         if (rt == NULL) {
1160                 err = -ENOMEM;
1161                 goto out;
1162         }
1163
1164         rt->u.dst.obsolete = -1;
1165         rt->rt6i_expires = (cfg->fc_flags & RTF_EXPIRES) ?
1166                                 jiffies + clock_t_to_jiffies(cfg->fc_expires) :
1167                                 0;
1168
1169         if (cfg->fc_protocol == RTPROT_UNSPEC)
1170                 cfg->fc_protocol = RTPROT_BOOT;
1171         rt->rt6i_protocol = cfg->fc_protocol;
1172
1173         addr_type = ipv6_addr_type(&cfg->fc_dst);
1174
1175         if (addr_type & IPV6_ADDR_MULTICAST)
1176                 rt->u.dst.input = ip6_mc_input;
1177         else
1178                 rt->u.dst.input = ip6_forward;
1179
1180         rt->u.dst.output = ip6_output;
1181
1182         ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1183         rt->rt6i_dst.plen = cfg->fc_dst_len;
1184         if (rt->rt6i_dst.plen == 128)
1185                rt->u.dst.flags = DST_HOST;
1186
1187 #ifdef CONFIG_IPV6_SUBTREES
1188         ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1189         rt->rt6i_src.plen = cfg->fc_src_len;
1190 #endif
1191
1192         rt->rt6i_metric = cfg->fc_metric;
1193
1194         /* We cannot add true routes via loopback here,
1195            they would result in kernel looping; promote them to reject routes
1196          */
1197         if ((cfg->fc_flags & RTF_REJECT) ||
1198             (dev && (dev->flags&IFF_LOOPBACK) && !(addr_type&IPV6_ADDR_LOOPBACK))) {
1199                 /* hold loopback dev/idev if we haven't done so. */
1200                 if (dev != net->loopback_dev) {
1201                         if (dev) {
1202                                 dev_put(dev);
1203                                 in6_dev_put(idev);
1204                         }
1205                         dev = net->loopback_dev;
1206                         dev_hold(dev);
1207                         idev = in6_dev_get(dev);
1208                         if (!idev) {
1209                                 err = -ENODEV;
1210                                 goto out;
1211                         }
1212                 }
1213                 rt->u.dst.output = ip6_pkt_discard_out;
1214                 rt->u.dst.input = ip6_pkt_discard;
1215                 rt->u.dst.error = -ENETUNREACH;
1216                 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1217                 goto install_route;
1218         }
1219
1220         if (cfg->fc_flags & RTF_GATEWAY) {
1221                 struct in6_addr *gw_addr;
1222                 int gwa_type;
1223
1224                 gw_addr = &cfg->fc_gateway;
1225                 ipv6_addr_copy(&rt->rt6i_gateway, gw_addr);
1226                 gwa_type = ipv6_addr_type(gw_addr);
1227
1228                 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1229                         struct rt6_info *grt;
1230
1231                         /* IPv6 strictly inhibits using not link-local
1232                            addresses as nexthop address.
1233                            Otherwise, router will not able to send redirects.
1234                            It is very good, but in some (rare!) circumstances
1235                            (SIT, PtP, NBMA NOARP links) it is handy to allow
1236                            some exceptions. --ANK
1237                          */
1238                         err = -EINVAL;
1239                         if (!(gwa_type&IPV6_ADDR_UNICAST))
1240                                 goto out;
1241
1242                         grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, 1);
1243
1244                         err = -EHOSTUNREACH;
1245                         if (grt == NULL)
1246                                 goto out;
1247                         if (dev) {
1248                                 if (dev != grt->rt6i_dev) {
1249                                         dst_release(&grt->u.dst);
1250                                         goto out;
1251                                 }
1252                         } else {
1253                                 dev = grt->rt6i_dev;
1254                                 idev = grt->rt6i_idev;
1255                                 dev_hold(dev);
1256                                 in6_dev_hold(grt->rt6i_idev);
1257                         }
1258                         if (!(grt->rt6i_flags&RTF_GATEWAY))
1259                                 err = 0;
1260                         dst_release(&grt->u.dst);
1261
1262                         if (err)
1263                                 goto out;
1264                 }
1265                 err = -EINVAL;
1266                 if (dev == NULL || (dev->flags&IFF_LOOPBACK))
1267                         goto out;
1268         }
1269
1270         err = -ENODEV;
1271         if (dev == NULL)
1272                 goto out;
1273
1274         if (cfg->fc_flags & (RTF_GATEWAY | RTF_NONEXTHOP)) {
1275                 rt->rt6i_nexthop = __neigh_lookup_errno(&nd_tbl, &rt->rt6i_gateway, dev);
1276                 if (IS_ERR(rt->rt6i_nexthop)) {
1277                         err = PTR_ERR(rt->rt6i_nexthop);
1278                         rt->rt6i_nexthop = NULL;
1279                         goto out;
1280                 }
1281         }
1282
1283         rt->rt6i_flags = cfg->fc_flags;
1284
1285 install_route:
1286         if (cfg->fc_mx) {
1287                 struct nlattr *nla;
1288                 int remaining;
1289
1290                 nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1291                         int type = nla_type(nla);
1292
1293                         if (type) {
1294                                 if (type > RTAX_MAX) {
1295                                         err = -EINVAL;
1296                                         goto out;
1297                                 }
1298
1299                                 rt->u.dst.metrics[type - 1] = nla_get_u32(nla);
1300                         }
1301                 }
1302         }
1303
1304         if (dst_metric(&rt->u.dst, RTAX_HOPLIMIT) == 0)
1305                 rt->u.dst.metrics[RTAX_HOPLIMIT-1] = -1;
1306         if (!dst_mtu(&rt->u.dst))
1307                 rt->u.dst.metrics[RTAX_MTU-1] = ipv6_get_mtu(dev);
1308         if (!dst_metric(&rt->u.dst, RTAX_ADVMSS))
1309                 rt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(net, dst_mtu(&rt->u.dst));
1310         rt->u.dst.dev = dev;
1311         rt->rt6i_idev = idev;
1312         rt->rt6i_table = table;
1313
1314         cfg->fc_nlinfo.nl_net = dev_net(dev);
1315
1316         return __ip6_ins_rt(rt, &cfg->fc_nlinfo);
1317
1318 out:
1319         if (dev)
1320                 dev_put(dev);
1321         if (idev)
1322                 in6_dev_put(idev);
1323         if (rt)
1324                 dst_free(&rt->u.dst);
1325         return err;
1326 }
1327
1328 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
1329 {
1330         int err;
1331         struct fib6_table *table;
1332         struct net *net = dev_net(rt->rt6i_dev);
1333
1334         if (rt == net->ipv6.ip6_null_entry)
1335                 return -ENOENT;
1336
1337         table = rt->rt6i_table;
1338         write_lock_bh(&table->tb6_lock);
1339
1340         err = fib6_del(rt, info);
1341         dst_release(&rt->u.dst);
1342
1343         write_unlock_bh(&table->tb6_lock);
1344
1345         return err;
1346 }
1347
1348 int ip6_del_rt(struct rt6_info *rt)
1349 {
1350         struct nl_info info = {
1351                 .nl_net = dev_net(rt->rt6i_dev),
1352         };
1353         return __ip6_del_rt(rt, &info);
1354 }
1355
1356 static int ip6_route_del(struct fib6_config *cfg)
1357 {
1358         struct fib6_table *table;
1359         struct fib6_node *fn;
1360         struct rt6_info *rt;
1361         int err = -ESRCH;
1362
1363         table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
1364         if (table == NULL)
1365                 return err;
1366
1367         read_lock_bh(&table->tb6_lock);
1368
1369         fn = fib6_locate(&table->tb6_root,
1370                          &cfg->fc_dst, cfg->fc_dst_len,
1371                          &cfg->fc_src, cfg->fc_src_len);
1372
1373         if (fn) {
1374                 for (rt = fn->leaf; rt; rt = rt->u.dst.rt6_next) {
1375                         if (cfg->fc_ifindex &&
1376                             (rt->rt6i_dev == NULL ||
1377                              rt->rt6i_dev->ifindex != cfg->fc_ifindex))
1378                                 continue;
1379                         if (cfg->fc_flags & RTF_GATEWAY &&
1380                             !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
1381                                 continue;
1382                         if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
1383                                 continue;
1384                         dst_hold(&rt->u.dst);
1385                         read_unlock_bh(&table->tb6_lock);
1386
1387                         return __ip6_del_rt(rt, &cfg->fc_nlinfo);
1388                 }
1389         }
1390         read_unlock_bh(&table->tb6_lock);
1391
1392         return err;
1393 }
1394
1395 /*
1396  *      Handle redirects
1397  */
1398 struct ip6rd_flowi {
1399         struct flowi fl;
1400         struct in6_addr gateway;
1401 };
1402
1403 static struct rt6_info *__ip6_route_redirect(struct net *net,
1404                                              struct fib6_table *table,
1405                                              struct flowi *fl,
1406                                              int flags)
1407 {
1408         struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl;
1409         struct rt6_info *rt;
1410         struct fib6_node *fn;
1411
1412         /*
1413          * Get the "current" route for this destination and
1414          * check if the redirect has come from approriate router.
1415          *
1416          * RFC 2461 specifies that redirects should only be
1417          * accepted if they come from the nexthop to the target.
1418          * Due to the way the routes are chosen, this notion
1419          * is a bit fuzzy and one might need to check all possible
1420          * routes.
1421          */
1422
1423         read_lock_bh(&table->tb6_lock);
1424         fn = fib6_lookup(&table->tb6_root, &fl->fl6_dst, &fl->fl6_src);
1425 restart:
1426         for (rt = fn->leaf; rt; rt = rt->u.dst.rt6_next) {
1427                 /*
1428                  * Current route is on-link; redirect is always invalid.
1429                  *
1430                  * Seems, previous statement is not true. It could
1431                  * be node, which looks for us as on-link (f.e. proxy ndisc)
1432                  * But then router serving it might decide, that we should
1433                  * know truth 8)8) --ANK (980726).
1434                  */
1435                 if (rt6_check_expired(rt))
1436                         continue;
1437                 if (!(rt->rt6i_flags & RTF_GATEWAY))
1438                         continue;
1439                 if (fl->oif != rt->rt6i_dev->ifindex)
1440                         continue;
1441                 if (!ipv6_addr_equal(&rdfl->gateway, &rt->rt6i_gateway))
1442                         continue;
1443                 break;
1444         }
1445
1446         if (!rt)
1447                 rt = net->ipv6.ip6_null_entry;
1448         BACKTRACK(net, &fl->fl6_src);
1449 out:
1450         dst_hold(&rt->u.dst);
1451
1452         read_unlock_bh(&table->tb6_lock);
1453
1454         return rt;
1455 };
1456
1457 static struct rt6_info *ip6_route_redirect(struct in6_addr *dest,
1458                                            struct in6_addr *src,
1459                                            struct in6_addr *gateway,
1460                                            struct net_device *dev)
1461 {
1462         int flags = RT6_LOOKUP_F_HAS_SADDR;
1463         struct net *net = dev_net(dev);
1464         struct ip6rd_flowi rdfl = {
1465                 .fl = {
1466                         .oif = dev->ifindex,
1467                         .nl_u = {
1468                                 .ip6_u = {
1469                                         .daddr = *dest,
1470                                         .saddr = *src,
1471                                 },
1472                         },
1473                 },
1474         };
1475
1476         ipv6_addr_copy(&rdfl.gateway, gateway);
1477
1478         if (rt6_need_strict(dest))
1479                 flags |= RT6_LOOKUP_F_IFACE;
1480
1481         return (struct rt6_info *)fib6_rule_lookup(net, (struct flowi *)&rdfl,
1482                                                    flags, __ip6_route_redirect);
1483 }
1484
1485 void rt6_redirect(struct in6_addr *dest, struct in6_addr *src,
1486                   struct in6_addr *saddr,
1487                   struct neighbour *neigh, u8 *lladdr, int on_link)
1488 {
1489         struct rt6_info *rt, *nrt = NULL;
1490         struct netevent_redirect netevent;
1491         struct net *net = dev_net(neigh->dev);
1492
1493         rt = ip6_route_redirect(dest, src, saddr, neigh->dev);
1494
1495         if (rt == net->ipv6.ip6_null_entry) {
1496                 if (net_ratelimit())
1497                         printk(KERN_DEBUG "rt6_redirect: source isn't a valid nexthop "
1498                                "for redirect target\n");
1499                 goto out;
1500         }
1501
1502         /*
1503          *      We have finally decided to accept it.
1504          */
1505
1506         neigh_update(neigh, lladdr, NUD_STALE,
1507                      NEIGH_UPDATE_F_WEAK_OVERRIDE|
1508                      NEIGH_UPDATE_F_OVERRIDE|
1509                      (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1510                                      NEIGH_UPDATE_F_ISROUTER))
1511                      );
1512
1513         /*
1514          * Redirect received -> path was valid.
1515          * Look, redirects are sent only in response to data packets,
1516          * so that this nexthop apparently is reachable. --ANK
1517          */
1518         dst_confirm(&rt->u.dst);
1519
1520         /* Duplicate redirect: silently ignore. */
1521         if (neigh == rt->u.dst.neighbour)
1522                 goto out;
1523
1524         nrt = ip6_rt_copy(rt);
1525         if (nrt == NULL)
1526                 goto out;
1527
1528         nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
1529         if (on_link)
1530                 nrt->rt6i_flags &= ~RTF_GATEWAY;
1531
1532         ipv6_addr_copy(&nrt->rt6i_dst.addr, dest);
1533         nrt->rt6i_dst.plen = 128;
1534         nrt->u.dst.flags |= DST_HOST;
1535
1536         ipv6_addr_copy(&nrt->rt6i_gateway, (struct in6_addr*)neigh->primary_key);
1537         nrt->rt6i_nexthop = neigh_clone(neigh);
1538         /* Reset pmtu, it may be better */
1539         nrt->u.dst.metrics[RTAX_MTU-1] = ipv6_get_mtu(neigh->dev);
1540         nrt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(dev_net(neigh->dev),
1541                                                         dst_mtu(&nrt->u.dst));
1542
1543         if (ip6_ins_rt(nrt))
1544                 goto out;
1545
1546         netevent.old = &rt->u.dst;
1547         netevent.new = &nrt->u.dst;
1548         call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
1549
1550         if (rt->rt6i_flags&RTF_CACHE) {
1551                 ip6_del_rt(rt);
1552                 return;
1553         }
1554
1555 out:
1556         dst_release(&rt->u.dst);
1557         return;
1558 }
1559
1560 /*
1561  *      Handle ICMP "packet too big" messages
1562  *      i.e. Path MTU discovery
1563  */
1564
1565 void rt6_pmtu_discovery(struct in6_addr *daddr, struct in6_addr *saddr,
1566                         struct net_device *dev, u32 pmtu)
1567 {
1568         struct rt6_info *rt, *nrt;
1569         struct net *net = dev_net(dev);
1570         int allfrag = 0;
1571
1572         rt = rt6_lookup(net, daddr, saddr, dev->ifindex, 0);
1573         if (rt == NULL)
1574                 return;
1575
1576         if (pmtu >= dst_mtu(&rt->u.dst))
1577                 goto out;
1578
1579         if (pmtu < IPV6_MIN_MTU) {
1580                 /*
1581                  * According to RFC2460, PMTU is set to the IPv6 Minimum Link
1582                  * MTU (1280) and a fragment header should always be included
1583                  * after a node receiving Too Big message reporting PMTU is
1584                  * less than the IPv6 Minimum Link MTU.
1585                  */
1586                 pmtu = IPV6_MIN_MTU;
1587                 allfrag = 1;
1588         }
1589
1590         /* New mtu received -> path was valid.
1591            They are sent only in response to data packets,
1592            so that this nexthop apparently is reachable. --ANK
1593          */
1594         dst_confirm(&rt->u.dst);
1595
1596         /* Host route. If it is static, it would be better
1597            not to override it, but add new one, so that
1598            when cache entry will expire old pmtu
1599            would return automatically.
1600          */
1601         if (rt->rt6i_flags & RTF_CACHE) {
1602                 rt->u.dst.metrics[RTAX_MTU-1] = pmtu;
1603                 if (allfrag)
1604                         rt->u.dst.metrics[RTAX_FEATURES-1] |= RTAX_FEATURE_ALLFRAG;
1605                 dst_set_expires(&rt->u.dst, net->ipv6.sysctl.ip6_rt_mtu_expires);
1606                 rt->rt6i_flags |= RTF_MODIFIED|RTF_EXPIRES;
1607                 goto out;
1608         }
1609
1610         /* Network route.
1611            Two cases are possible:
1612            1. It is connected route. Action: COW
1613            2. It is gatewayed route or NONEXTHOP route. Action: clone it.
1614          */
1615         if (!rt->rt6i_nexthop && !(rt->rt6i_flags & RTF_NONEXTHOP))
1616                 nrt = rt6_alloc_cow(rt, daddr, saddr);
1617         else
1618                 nrt = rt6_alloc_clone(rt, daddr);
1619
1620         if (nrt) {
1621                 nrt->u.dst.metrics[RTAX_MTU-1] = pmtu;
1622                 if (allfrag)
1623                         nrt->u.dst.metrics[RTAX_FEATURES-1] |= RTAX_FEATURE_ALLFRAG;
1624
1625                 /* According to RFC 1981, detecting PMTU increase shouldn't be
1626                  * happened within 5 mins, the recommended timer is 10 mins.
1627                  * Here this route expiration time is set to ip6_rt_mtu_expires
1628                  * which is 10 mins. After 10 mins the decreased pmtu is expired
1629                  * and detecting PMTU increase will be automatically happened.
1630                  */
1631                 dst_set_expires(&nrt->u.dst, net->ipv6.sysctl.ip6_rt_mtu_expires);
1632                 nrt->rt6i_flags |= RTF_DYNAMIC|RTF_EXPIRES;
1633
1634                 ip6_ins_rt(nrt);
1635         }
1636 out:
1637         dst_release(&rt->u.dst);
1638 }
1639
1640 /*
1641  *      Misc support functions
1642  */
1643
1644 static struct rt6_info * ip6_rt_copy(struct rt6_info *ort)
1645 {
1646         struct net *net = dev_net(ort->rt6i_dev);
1647         struct rt6_info *rt = ip6_dst_alloc(&net->ipv6.ip6_dst_ops);
1648
1649         if (rt) {
1650                 rt->u.dst.input = ort->u.dst.input;
1651                 rt->u.dst.output = ort->u.dst.output;
1652
1653                 memcpy(rt->u.dst.metrics, ort->u.dst.metrics, RTAX_MAX*sizeof(u32));
1654                 rt->u.dst.error = ort->u.dst.error;
1655                 rt->u.dst.dev = ort->u.dst.dev;
1656                 if (rt->u.dst.dev)
1657                         dev_hold(rt->u.dst.dev);
1658                 rt->rt6i_idev = ort->rt6i_idev;
1659                 if (rt->rt6i_idev)
1660                         in6_dev_hold(rt->rt6i_idev);
1661                 rt->u.dst.lastuse = jiffies;
1662                 rt->rt6i_expires = 0;
1663
1664                 ipv6_addr_copy(&rt->rt6i_gateway, &ort->rt6i_gateway);
1665                 rt->rt6i_flags = ort->rt6i_flags & ~RTF_EXPIRES;
1666                 rt->rt6i_metric = 0;
1667
1668                 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1669 #ifdef CONFIG_IPV6_SUBTREES
1670                 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1671 #endif
1672                 rt->rt6i_table = ort->rt6i_table;
1673         }
1674         return rt;
1675 }
1676
1677 #ifdef CONFIG_IPV6_ROUTE_INFO
1678 static struct rt6_info *rt6_get_route_info(struct net *net,
1679                                            struct in6_addr *prefix, int prefixlen,
1680                                            struct in6_addr *gwaddr, int ifindex)
1681 {
1682         struct fib6_node *fn;
1683         struct rt6_info *rt = NULL;
1684         struct fib6_table *table;
1685
1686         table = fib6_get_table(net, RT6_TABLE_INFO);
1687         if (table == NULL)
1688                 return NULL;
1689
1690         write_lock_bh(&table->tb6_lock);
1691         fn = fib6_locate(&table->tb6_root, prefix ,prefixlen, NULL, 0);
1692         if (!fn)
1693                 goto out;
1694
1695         for (rt = fn->leaf; rt; rt = rt->u.dst.rt6_next) {
1696                 if (rt->rt6i_dev->ifindex != ifindex)
1697                         continue;
1698                 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
1699                         continue;
1700                 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
1701                         continue;
1702                 dst_hold(&rt->u.dst);
1703                 break;
1704         }
1705 out:
1706         write_unlock_bh(&table->tb6_lock);
1707         return rt;
1708 }
1709
1710 static struct rt6_info *rt6_add_route_info(struct net *net,
1711                                            struct in6_addr *prefix, int prefixlen,
1712                                            struct in6_addr *gwaddr, int ifindex,
1713                                            unsigned pref)
1714 {
1715         struct fib6_config cfg = {
1716                 .fc_table       = RT6_TABLE_INFO,
1717                 .fc_metric      = IP6_RT_PRIO_USER,
1718                 .fc_ifindex     = ifindex,
1719                 .fc_dst_len     = prefixlen,
1720                 .fc_flags       = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
1721                                   RTF_UP | RTF_PREF(pref),
1722                 .fc_nlinfo.pid = 0,
1723                 .fc_nlinfo.nlh = NULL,
1724                 .fc_nlinfo.nl_net = net,
1725         };
1726
1727         ipv6_addr_copy(&cfg.fc_dst, prefix);
1728         ipv6_addr_copy(&cfg.fc_gateway, gwaddr);
1729
1730         /* We should treat it as a default route if prefix length is 0. */
1731         if (!prefixlen)
1732                 cfg.fc_flags |= RTF_DEFAULT;
1733
1734         ip6_route_add(&cfg);
1735
1736         return rt6_get_route_info(net, prefix, prefixlen, gwaddr, ifindex);
1737 }
1738 #endif
1739
1740 struct rt6_info *rt6_get_dflt_router(struct in6_addr *addr, struct net_device *dev)
1741 {
1742         struct rt6_info *rt;
1743         struct fib6_table *table;
1744
1745         table = fib6_get_table(dev_net(dev), RT6_TABLE_DFLT);
1746         if (table == NULL)
1747                 return NULL;
1748
1749         write_lock_bh(&table->tb6_lock);
1750         for (rt = table->tb6_root.leaf; rt; rt=rt->u.dst.rt6_next) {
1751                 if (dev == rt->rt6i_dev &&
1752                     ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
1753                     ipv6_addr_equal(&rt->rt6i_gateway, addr))
1754                         break;
1755         }
1756         if (rt)
1757                 dst_hold(&rt->u.dst);
1758         write_unlock_bh(&table->tb6_lock);
1759         return rt;
1760 }
1761
1762 struct rt6_info *rt6_add_dflt_router(struct in6_addr *gwaddr,
1763                                      struct net_device *dev,
1764                                      unsigned int pref)
1765 {
1766         struct fib6_config cfg = {
1767                 .fc_table       = RT6_TABLE_DFLT,
1768                 .fc_metric      = IP6_RT_PRIO_USER,
1769                 .fc_ifindex     = dev->ifindex,
1770                 .fc_flags       = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
1771                                   RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
1772                 .fc_nlinfo.pid = 0,
1773                 .fc_nlinfo.nlh = NULL,
1774                 .fc_nlinfo.nl_net = dev_net(dev),
1775         };
1776
1777         ipv6_addr_copy(&cfg.fc_gateway, gwaddr);
1778
1779         ip6_route_add(&cfg);
1780
1781         return rt6_get_dflt_router(gwaddr, dev);
1782 }
1783
1784 void rt6_purge_dflt_routers(struct net *net)
1785 {
1786         struct rt6_info *rt;
1787         struct fib6_table *table;
1788
1789         /* NOTE: Keep consistent with rt6_get_dflt_router */
1790         table = fib6_get_table(net, RT6_TABLE_DFLT);
1791         if (table == NULL)
1792                 return;
1793
1794 restart:
1795         read_lock_bh(&table->tb6_lock);
1796         for (rt = table->tb6_root.leaf; rt; rt = rt->u.dst.rt6_next) {
1797                 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF)) {
1798                         dst_hold(&rt->u.dst);
1799                         read_unlock_bh(&table->tb6_lock);
1800                         ip6_del_rt(rt);
1801                         goto restart;
1802                 }
1803         }
1804         read_unlock_bh(&table->tb6_lock);
1805 }
1806
1807 static void rtmsg_to_fib6_config(struct net *net,
1808                                  struct in6_rtmsg *rtmsg,
1809                                  struct fib6_config *cfg)
1810 {
1811         memset(cfg, 0, sizeof(*cfg));
1812
1813         cfg->fc_table = RT6_TABLE_MAIN;
1814         cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
1815         cfg->fc_metric = rtmsg->rtmsg_metric;
1816         cfg->fc_expires = rtmsg->rtmsg_info;
1817         cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
1818         cfg->fc_src_len = rtmsg->rtmsg_src_len;
1819         cfg->fc_flags = rtmsg->rtmsg_flags;
1820
1821         cfg->fc_nlinfo.nl_net = net;
1822
1823         ipv6_addr_copy(&cfg->fc_dst, &rtmsg->rtmsg_dst);
1824         ipv6_addr_copy(&cfg->fc_src, &rtmsg->rtmsg_src);
1825         ipv6_addr_copy(&cfg->fc_gateway, &rtmsg->rtmsg_gateway);
1826 }
1827
1828 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
1829 {
1830         struct fib6_config cfg;
1831         struct in6_rtmsg rtmsg;
1832         int err;
1833
1834         switch(cmd) {
1835         case SIOCADDRT:         /* Add a route */
1836         case SIOCDELRT:         /* Delete a route */
1837                 if (!capable(CAP_NET_ADMIN))
1838                         return -EPERM;
1839                 err = copy_from_user(&rtmsg, arg,
1840                                      sizeof(struct in6_rtmsg));
1841                 if (err)
1842                         return -EFAULT;
1843
1844                 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
1845
1846                 rtnl_lock();
1847                 switch (cmd) {
1848                 case SIOCADDRT:
1849                         err = ip6_route_add(&cfg);
1850                         break;
1851                 case SIOCDELRT:
1852                         err = ip6_route_del(&cfg);
1853                         break;
1854                 default:
1855                         err = -EINVAL;
1856                 }
1857                 rtnl_unlock();
1858
1859                 return err;
1860         }
1861
1862         return -EINVAL;
1863 }
1864
1865 /*
1866  *      Drop the packet on the floor
1867  */
1868
1869 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
1870 {
1871         int type;
1872         struct dst_entry *dst = skb_dst(skb);
1873         switch (ipstats_mib_noroutes) {
1874         case IPSTATS_MIB_INNOROUTES:
1875                 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
1876                 if (type == IPV6_ADDR_ANY || type == IPV6_ADDR_RESERVED) {
1877                         IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
1878                                       IPSTATS_MIB_INADDRERRORS);
1879                         break;
1880                 }
1881                 /* FALLTHROUGH */
1882         case IPSTATS_MIB_OUTNOROUTES:
1883                 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
1884                               ipstats_mib_noroutes);
1885                 break;
1886         }
1887         icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0, skb->dev);
1888         kfree_skb(skb);
1889         return 0;
1890 }
1891
1892 static int ip6_pkt_discard(struct sk_buff *skb)
1893 {
1894         return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
1895 }
1896
1897 static int ip6_pkt_discard_out(struct sk_buff *skb)
1898 {
1899         skb->dev = skb_dst(skb)->dev;
1900         return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
1901 }
1902
1903 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
1904
1905 static int ip6_pkt_prohibit(struct sk_buff *skb)
1906 {
1907         return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
1908 }
1909
1910 static int ip6_pkt_prohibit_out(struct sk_buff *skb)
1911 {
1912         skb->dev = skb_dst(skb)->dev;
1913         return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
1914 }
1915
1916 #endif
1917
1918 /*
1919  *      Allocate a dst for local (unicast / anycast) address.
1920  */
1921
1922 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
1923                                     const struct in6_addr *addr,
1924                                     int anycast)
1925 {
1926         struct net *net = dev_net(idev->dev);
1927         struct rt6_info *rt = ip6_dst_alloc(&net->ipv6.ip6_dst_ops);
1928         struct neighbour *neigh;
1929
1930         if (rt == NULL)
1931                 return ERR_PTR(-ENOMEM);
1932
1933         dev_hold(net->loopback_dev);
1934         in6_dev_hold(idev);
1935
1936         rt->u.dst.flags = DST_HOST;
1937         rt->u.dst.input = ip6_input;
1938         rt->u.dst.output = ip6_output;
1939         rt->rt6i_dev = net->loopback_dev;
1940         rt->rt6i_idev = idev;
1941         rt->u.dst.metrics[RTAX_MTU-1] = ipv6_get_mtu(rt->rt6i_dev);
1942         rt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(net, dst_mtu(&rt->u.dst));
1943         rt->u.dst.metrics[RTAX_HOPLIMIT-1] = -1;
1944         rt->u.dst.obsolete = -1;
1945
1946         rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
1947         if (anycast)
1948                 rt->rt6i_flags |= RTF_ANYCAST;
1949         else
1950                 rt->rt6i_flags |= RTF_LOCAL;
1951         neigh = ndisc_get_neigh(rt->rt6i_dev, &rt->rt6i_gateway);
1952         if (IS_ERR(neigh)) {
1953                 dst_free(&rt->u.dst);
1954
1955                 /* We are casting this because that is the return
1956                  * value type.  But an errno encoded pointer is the
1957                  * same regardless of the underlying pointer type,
1958                  * and that's what we are returning.  So this is OK.
1959                  */
1960                 return (struct rt6_info *) neigh;
1961         }
1962         rt->rt6i_nexthop = neigh;
1963
1964         ipv6_addr_copy(&rt->rt6i_dst.addr, addr);
1965         rt->rt6i_dst.plen = 128;
1966         rt->rt6i_table = fib6_get_table(net, RT6_TABLE_LOCAL);
1967
1968         atomic_set(&rt->u.dst.__refcnt, 1);
1969
1970         return rt;
1971 }
1972
1973 struct arg_dev_net {
1974         struct net_device *dev;
1975         struct net *net;
1976 };
1977
1978 static int fib6_ifdown(struct rt6_info *rt, void *arg)
1979 {
1980         struct net_device *dev = ((struct arg_dev_net *)arg)->dev;
1981         struct net *net = ((struct arg_dev_net *)arg)->net;
1982
1983         if (((void *)rt->rt6i_dev == dev || dev == NULL) &&
1984             rt != net->ipv6.ip6_null_entry) {
1985                 RT6_TRACE("deleted by ifdown %p\n", rt);
1986                 return -1;
1987         }
1988         return 0;
1989 }
1990
1991 void rt6_ifdown(struct net *net, struct net_device *dev)
1992 {
1993         struct arg_dev_net adn = {
1994                 .dev = dev,
1995                 .net = net,
1996         };
1997
1998         fib6_clean_all(net, fib6_ifdown, 0, &adn);
1999         icmp6_clean_all(fib6_ifdown, &adn);
2000 }
2001
2002 struct rt6_mtu_change_arg
2003 {
2004         struct net_device *dev;
2005         unsigned mtu;
2006 };
2007
2008 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2009 {
2010         struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2011         struct inet6_dev *idev;
2012         struct net *net = dev_net(arg->dev);
2013
2014         /* In IPv6 pmtu discovery is not optional,
2015            so that RTAX_MTU lock cannot disable it.
2016            We still use this lock to block changes
2017            caused by addrconf/ndisc.
2018         */
2019
2020         idev = __in6_dev_get(arg->dev);
2021         if (idev == NULL)
2022                 return 0;
2023
2024         /* For administrative MTU increase, there is no way to discover
2025            IPv6 PMTU increase, so PMTU increase should be updated here.
2026            Since RFC 1981 doesn't include administrative MTU increase
2027            update PMTU increase is a MUST. (i.e. jumbo frame)
2028          */
2029         /*
2030            If new MTU is less than route PMTU, this new MTU will be the
2031            lowest MTU in the path, update the route PMTU to reflect PMTU
2032            decreases; if new MTU is greater than route PMTU, and the
2033            old MTU is the lowest MTU in the path, update the route PMTU
2034            to reflect the increase. In this case if the other nodes' MTU
2035            also have the lowest MTU, TOO BIG MESSAGE will be lead to
2036            PMTU discouvery.
2037          */
2038         if (rt->rt6i_dev == arg->dev &&
2039             !dst_metric_locked(&rt->u.dst, RTAX_MTU) &&
2040             (dst_mtu(&rt->u.dst) >= arg->mtu ||
2041              (dst_mtu(&rt->u.dst) < arg->mtu &&
2042               dst_mtu(&rt->u.dst) == idev->cnf.mtu6))) {
2043                 rt->u.dst.metrics[RTAX_MTU-1] = arg->mtu;
2044                 rt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(net, arg->mtu);
2045         }
2046         return 0;
2047 }
2048
2049 void rt6_mtu_change(struct net_device *dev, unsigned mtu)
2050 {
2051         struct rt6_mtu_change_arg arg = {
2052                 .dev = dev,
2053                 .mtu = mtu,
2054         };
2055
2056         fib6_clean_all(dev_net(dev), rt6_mtu_change_route, 0, &arg);
2057 }
2058
2059 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2060         [RTA_GATEWAY]           = { .len = sizeof(struct in6_addr) },
2061         [RTA_OIF]               = { .type = NLA_U32 },
2062         [RTA_IIF]               = { .type = NLA_U32 },
2063         [RTA_PRIORITY]          = { .type = NLA_U32 },
2064         [RTA_METRICS]           = { .type = NLA_NESTED },
2065 };
2066
2067 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2068                               struct fib6_config *cfg)
2069 {
2070         struct rtmsg *rtm;
2071         struct nlattr *tb[RTA_MAX+1];
2072         int err;
2073
2074         err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2075         if (err < 0)
2076                 goto errout;
2077
2078         err = -EINVAL;
2079         rtm = nlmsg_data(nlh);
2080         memset(cfg, 0, sizeof(*cfg));
2081
2082         cfg->fc_table = rtm->rtm_table;
2083         cfg->fc_dst_len = rtm->rtm_dst_len;
2084         cfg->fc_src_len = rtm->rtm_src_len;
2085         cfg->fc_flags = RTF_UP;
2086         cfg->fc_protocol = rtm->rtm_protocol;
2087
2088         if (rtm->rtm_type == RTN_UNREACHABLE)
2089                 cfg->fc_flags |= RTF_REJECT;
2090
2091         cfg->fc_nlinfo.pid = NETLINK_CB(skb).pid;
2092         cfg->fc_nlinfo.nlh = nlh;
2093         cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2094
2095         if (tb[RTA_GATEWAY]) {
2096                 nla_memcpy(&cfg->fc_gateway, tb[RTA_GATEWAY], 16);
2097                 cfg->fc_flags |= RTF_GATEWAY;
2098         }
2099
2100         if (tb[RTA_DST]) {
2101                 int plen = (rtm->rtm_dst_len + 7) >> 3;
2102
2103                 if (nla_len(tb[RTA_DST]) < plen)
2104                         goto errout;
2105
2106                 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2107         }
2108
2109         if (tb[RTA_SRC]) {
2110                 int plen = (rtm->rtm_src_len + 7) >> 3;
2111
2112                 if (nla_len(tb[RTA_SRC]) < plen)
2113                         goto errout;
2114
2115                 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2116         }
2117
2118         if (tb[RTA_OIF])
2119                 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2120
2121         if (tb[RTA_PRIORITY])
2122                 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2123
2124         if (tb[RTA_METRICS]) {
2125                 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2126                 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2127         }
2128
2129         if (tb[RTA_TABLE])
2130                 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2131
2132         err = 0;
2133 errout:
2134         return err;
2135 }
2136
2137 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
2138 {
2139         struct fib6_config cfg;
2140         int err;
2141
2142         err = rtm_to_fib6_config(skb, nlh, &cfg);
2143         if (err < 0)
2144                 return err;
2145
2146         return ip6_route_del(&cfg);
2147 }
2148
2149 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
2150 {
2151         struct fib6_config cfg;
2152         int err;
2153
2154         err = rtm_to_fib6_config(skb, nlh, &cfg);
2155         if (err < 0)
2156                 return err;
2157
2158         return ip6_route_add(&cfg);
2159 }
2160
2161 static inline size_t rt6_nlmsg_size(void)
2162 {
2163         return NLMSG_ALIGN(sizeof(struct rtmsg))
2164                + nla_total_size(16) /* RTA_SRC */
2165                + nla_total_size(16) /* RTA_DST */
2166                + nla_total_size(16) /* RTA_GATEWAY */
2167                + nla_total_size(16) /* RTA_PREFSRC */
2168                + nla_total_size(4) /* RTA_TABLE */
2169                + nla_total_size(4) /* RTA_IIF */
2170                + nla_total_size(4) /* RTA_OIF */
2171                + nla_total_size(4) /* RTA_PRIORITY */
2172                + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
2173                + nla_total_size(sizeof(struct rta_cacheinfo));
2174 }
2175
2176 static int rt6_fill_node(struct net *net,
2177                          struct sk_buff *skb, struct rt6_info *rt,
2178                          struct in6_addr *dst, struct in6_addr *src,
2179                          int iif, int type, u32 pid, u32 seq,
2180                          int prefix, int nowait, unsigned int flags)
2181 {
2182         struct rtmsg *rtm;
2183         struct nlmsghdr *nlh;
2184         long expires;
2185         u32 table;
2186
2187         if (prefix) {   /* user wants prefix routes only */
2188                 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
2189                         /* success since this is not a prefix route */
2190                         return 1;
2191                 }
2192         }
2193
2194         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*rtm), flags);
2195         if (nlh == NULL)
2196                 return -EMSGSIZE;
2197
2198         rtm = nlmsg_data(nlh);
2199         rtm->rtm_family = AF_INET6;
2200         rtm->rtm_dst_len = rt->rt6i_dst.plen;
2201         rtm->rtm_src_len = rt->rt6i_src.plen;
2202         rtm->rtm_tos = 0;
2203         if (rt->rt6i_table)
2204                 table = rt->rt6i_table->tb6_id;
2205         else
2206                 table = RT6_TABLE_UNSPEC;
2207         rtm->rtm_table = table;
2208         NLA_PUT_U32(skb, RTA_TABLE, table);
2209         if (rt->rt6i_flags&RTF_REJECT)
2210                 rtm->rtm_type = RTN_UNREACHABLE;
2211         else if (rt->rt6i_dev && (rt->rt6i_dev->flags&IFF_LOOPBACK))
2212                 rtm->rtm_type = RTN_LOCAL;
2213         else
2214                 rtm->rtm_type = RTN_UNICAST;
2215         rtm->rtm_flags = 0;
2216         rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2217         rtm->rtm_protocol = rt->rt6i_protocol;
2218         if (rt->rt6i_flags&RTF_DYNAMIC)
2219                 rtm->rtm_protocol = RTPROT_REDIRECT;
2220         else if (rt->rt6i_flags & RTF_ADDRCONF)
2221                 rtm->rtm_protocol = RTPROT_KERNEL;
2222         else if (rt->rt6i_flags&RTF_DEFAULT)
2223                 rtm->rtm_protocol = RTPROT_RA;
2224
2225         if (rt->rt6i_flags&RTF_CACHE)
2226                 rtm->rtm_flags |= RTM_F_CLONED;
2227
2228         if (dst) {
2229                 NLA_PUT(skb, RTA_DST, 16, dst);
2230                 rtm->rtm_dst_len = 128;
2231         } else if (rtm->rtm_dst_len)
2232                 NLA_PUT(skb, RTA_DST, 16, &rt->rt6i_dst.addr);
2233 #ifdef CONFIG_IPV6_SUBTREES
2234         if (src) {
2235                 NLA_PUT(skb, RTA_SRC, 16, src);
2236                 rtm->rtm_src_len = 128;
2237         } else if (rtm->rtm_src_len)
2238                 NLA_PUT(skb, RTA_SRC, 16, &rt->rt6i_src.addr);
2239 #endif
2240         if (iif) {
2241 #ifdef CONFIG_IPV6_MROUTE
2242                 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
2243                         int err = ip6mr_get_route(net, skb, rtm, nowait);
2244                         if (err <= 0) {
2245                                 if (!nowait) {
2246                                         if (err == 0)
2247                                                 return 0;
2248                                         goto nla_put_failure;
2249                                 } else {
2250                                         if (err == -EMSGSIZE)
2251                                                 goto nla_put_failure;
2252                                 }
2253                         }
2254                 } else
2255 #endif
2256                         NLA_PUT_U32(skb, RTA_IIF, iif);
2257         } else if (dst) {
2258                 struct inet6_dev *idev = ip6_dst_idev(&rt->u.dst);
2259                 struct in6_addr saddr_buf;
2260                 if (ipv6_dev_get_saddr(net, idev ? idev->dev : NULL,
2261                                        dst, 0, &saddr_buf) == 0)
2262                         NLA_PUT(skb, RTA_PREFSRC, 16, &saddr_buf);
2263         }
2264
2265         if (rtnetlink_put_metrics(skb, rt->u.dst.metrics) < 0)
2266                 goto nla_put_failure;
2267
2268         if (rt->u.dst.neighbour)
2269                 NLA_PUT(skb, RTA_GATEWAY, 16, &rt->u.dst.neighbour->primary_key);
2270
2271         if (rt->u.dst.dev)
2272                 NLA_PUT_U32(skb, RTA_OIF, rt->rt6i_dev->ifindex);
2273
2274         NLA_PUT_U32(skb, RTA_PRIORITY, rt->rt6i_metric);
2275
2276         if (!(rt->rt6i_flags & RTF_EXPIRES))
2277                 expires = 0;
2278         else if (rt->rt6i_expires - jiffies < INT_MAX)
2279                 expires = rt->rt6i_expires - jiffies;
2280         else
2281                 expires = INT_MAX;
2282
2283         if (rtnl_put_cacheinfo(skb, &rt->u.dst, 0, 0, 0,
2284                                expires, rt->u.dst.error) < 0)
2285                 goto nla_put_failure;
2286
2287         return nlmsg_end(skb, nlh);
2288
2289 nla_put_failure:
2290         nlmsg_cancel(skb, nlh);
2291         return -EMSGSIZE;
2292 }
2293
2294 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
2295 {
2296         struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
2297         int prefix;
2298
2299         if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
2300                 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
2301                 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
2302         } else
2303                 prefix = 0;
2304
2305         return rt6_fill_node(arg->net,
2306                      arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
2307                      NETLINK_CB(arg->cb->skb).pid, arg->cb->nlh->nlmsg_seq,
2308                      prefix, 0, NLM_F_MULTI);
2309 }
2310
2311 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr* nlh, void *arg)
2312 {
2313         struct net *net = sock_net(in_skb->sk);
2314         struct nlattr *tb[RTA_MAX+1];
2315         struct rt6_info *rt;
2316         struct sk_buff *skb;
2317         struct rtmsg *rtm;
2318         struct flowi fl;
2319         int err, iif = 0;
2320
2321         err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2322         if (err < 0)
2323                 goto errout;
2324
2325         err = -EINVAL;
2326         memset(&fl, 0, sizeof(fl));
2327
2328         if (tb[RTA_SRC]) {
2329                 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
2330                         goto errout;
2331
2332                 ipv6_addr_copy(&fl.fl6_src, nla_data(tb[RTA_SRC]));
2333         }
2334
2335         if (tb[RTA_DST]) {
2336                 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
2337                         goto errout;
2338
2339                 ipv6_addr_copy(&fl.fl6_dst, nla_data(tb[RTA_DST]));
2340         }
2341
2342         if (tb[RTA_IIF])
2343                 iif = nla_get_u32(tb[RTA_IIF]);
2344
2345         if (tb[RTA_OIF])
2346                 fl.oif = nla_get_u32(tb[RTA_OIF]);
2347
2348         if (iif) {
2349                 struct net_device *dev;
2350                 dev = __dev_get_by_index(net, iif);
2351                 if (!dev) {
2352                         err = -ENODEV;
2353                         goto errout;
2354                 }
2355         }
2356
2357         skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
2358         if (skb == NULL) {
2359                 err = -ENOBUFS;
2360                 goto errout;
2361         }
2362
2363         /* Reserve room for dummy headers, this skb can pass
2364            through good chunk of routing engine.
2365          */
2366         skb_reset_mac_header(skb);
2367         skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
2368
2369         rt = (struct rt6_info*) ip6_route_output(net, NULL, &fl);
2370         skb_dst_set(skb, &rt->u.dst);
2371
2372         err = rt6_fill_node(net, skb, rt, &fl.fl6_dst, &fl.fl6_src, iif,
2373                             RTM_NEWROUTE, NETLINK_CB(in_skb).pid,
2374                             nlh->nlmsg_seq, 0, 0, 0);
2375         if (err < 0) {
2376                 kfree_skb(skb);
2377                 goto errout;
2378         }
2379
2380         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
2381 errout:
2382         return err;
2383 }
2384
2385 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info)
2386 {
2387         struct sk_buff *skb;
2388         struct net *net = info->nl_net;
2389         u32 seq;
2390         int err;
2391
2392         err = -ENOBUFS;
2393         seq = info->nlh != NULL ? info->nlh->nlmsg_seq : 0;
2394
2395         skb = nlmsg_new(rt6_nlmsg_size(), gfp_any());
2396         if (skb == NULL)
2397                 goto errout;
2398
2399         err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
2400                                 event, info->pid, seq, 0, 0, 0);
2401         if (err < 0) {
2402                 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
2403                 WARN_ON(err == -EMSGSIZE);
2404                 kfree_skb(skb);
2405                 goto errout;
2406         }
2407         rtnl_notify(skb, net, info->pid, RTNLGRP_IPV6_ROUTE,
2408                     info->nlh, gfp_any());
2409         return;
2410 errout:
2411         if (err < 0)
2412                 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
2413 }
2414
2415 static int ip6_route_dev_notify(struct notifier_block *this,
2416                                 unsigned long event, void *data)
2417 {
2418         struct net_device *dev = (struct net_device *)data;
2419         struct net *net = dev_net(dev);
2420
2421         if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
2422                 net->ipv6.ip6_null_entry->u.dst.dev = dev;
2423                 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
2424 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2425                 net->ipv6.ip6_prohibit_entry->u.dst.dev = dev;
2426                 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
2427                 net->ipv6.ip6_blk_hole_entry->u.dst.dev = dev;
2428                 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
2429 #endif
2430         }
2431
2432         return NOTIFY_OK;
2433 }
2434
2435 /*
2436  *      /proc
2437  */
2438
2439 #ifdef CONFIG_PROC_FS
2440
2441 #define RT6_INFO_LEN (32 + 4 + 32 + 4 + 32 + 40 + 5 + 1)
2442
2443 struct rt6_proc_arg
2444 {
2445         char *buffer;
2446         int offset;
2447         int length;
2448         int skip;
2449         int len;
2450 };
2451
2452 static int rt6_info_route(struct rt6_info *rt, void *p_arg)
2453 {
2454         struct seq_file *m = p_arg;
2455
2456         seq_printf(m, "%pi6 %02x ", &rt->rt6i_dst.addr, rt->rt6i_dst.plen);
2457
2458 #ifdef CONFIG_IPV6_SUBTREES
2459         seq_printf(m, "%pi6 %02x ", &rt->rt6i_src.addr, rt->rt6i_src.plen);
2460 #else
2461         seq_puts(m, "00000000000000000000000000000000 00 ");
2462 #endif
2463
2464         if (rt->rt6i_nexthop) {
2465                 seq_printf(m, "%pi6", rt->rt6i_nexthop->primary_key);
2466         } else {
2467                 seq_puts(m, "00000000000000000000000000000000");
2468         }
2469         seq_printf(m, " %08x %08x %08x %08x %8s\n",
2470                    rt->rt6i_metric, atomic_read(&rt->u.dst.__refcnt),
2471                    rt->u.dst.__use, rt->rt6i_flags,
2472                    rt->rt6i_dev ? rt->rt6i_dev->name : "");
2473         return 0;
2474 }
2475
2476 static int ipv6_route_show(struct seq_file *m, void *v)
2477 {
2478         struct net *net = (struct net *)m->private;
2479         fib6_clean_all(net, rt6_info_route, 0, m);
2480         return 0;
2481 }
2482
2483 static int ipv6_route_open(struct inode *inode, struct file *file)
2484 {
2485         return single_open_net(inode, file, ipv6_route_show);
2486 }
2487
2488 static const struct file_operations ipv6_route_proc_fops = {
2489         .owner          = THIS_MODULE,
2490         .open           = ipv6_route_open,
2491         .read           = seq_read,
2492         .llseek         = seq_lseek,
2493         .release        = single_release_net,
2494 };
2495
2496 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
2497 {
2498         struct net *net = (struct net *)seq->private;
2499         seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
2500                    net->ipv6.rt6_stats->fib_nodes,
2501                    net->ipv6.rt6_stats->fib_route_nodes,
2502                    net->ipv6.rt6_stats->fib_rt_alloc,
2503                    net->ipv6.rt6_stats->fib_rt_entries,
2504                    net->ipv6.rt6_stats->fib_rt_cache,
2505                    atomic_read(&net->ipv6.ip6_dst_ops.entries),
2506                    net->ipv6.rt6_stats->fib_discarded_routes);
2507
2508         return 0;
2509 }
2510
2511 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
2512 {
2513         return single_open_net(inode, file, rt6_stats_seq_show);
2514 }
2515
2516 static const struct file_operations rt6_stats_seq_fops = {
2517         .owner   = THIS_MODULE,
2518         .open    = rt6_stats_seq_open,
2519         .read    = seq_read,
2520         .llseek  = seq_lseek,
2521         .release = single_release_net,
2522 };
2523 #endif  /* CONFIG_PROC_FS */
2524
2525 #ifdef CONFIG_SYSCTL
2526
2527 static
2528 int ipv6_sysctl_rtcache_flush(ctl_table *ctl, int write,
2529                               void __user *buffer, size_t *lenp, loff_t *ppos)
2530 {
2531         struct net *net = current->nsproxy->net_ns;
2532         int delay = net->ipv6.sysctl.flush_delay;
2533         if (write) {
2534                 proc_dointvec(ctl, write, buffer, lenp, ppos);
2535                 fib6_run_gc(delay <= 0 ? ~0UL : (unsigned long)delay, net);
2536                 return 0;
2537         } else
2538                 return -EINVAL;
2539 }
2540
2541 ctl_table ipv6_route_table_template[] = {
2542         {
2543                 .procname       =       "flush",
2544                 .data           =       &init_net.ipv6.sysctl.flush_delay,
2545                 .maxlen         =       sizeof(int),
2546                 .mode           =       0200,
2547                 .proc_handler   =       ipv6_sysctl_rtcache_flush
2548         },
2549         {
2550                 .procname       =       "gc_thresh",
2551                 .data           =       &ip6_dst_ops_template.gc_thresh,
2552                 .maxlen         =       sizeof(int),
2553                 .mode           =       0644,
2554                 .proc_handler   =       proc_dointvec,
2555         },
2556         {
2557                 .procname       =       "max_size",
2558                 .data           =       &init_net.ipv6.sysctl.ip6_rt_max_size,
2559                 .maxlen         =       sizeof(int),
2560                 .mode           =       0644,
2561                 .proc_handler   =       proc_dointvec,
2562         },
2563         {
2564                 .procname       =       "gc_min_interval",
2565                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2566                 .maxlen         =       sizeof(int),
2567                 .mode           =       0644,
2568                 .proc_handler   =       proc_dointvec_jiffies,
2569         },
2570         {
2571                 .procname       =       "gc_timeout",
2572                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
2573                 .maxlen         =       sizeof(int),
2574                 .mode           =       0644,
2575                 .proc_handler   =       proc_dointvec_jiffies,
2576         },
2577         {
2578                 .procname       =       "gc_interval",
2579                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_interval,
2580                 .maxlen         =       sizeof(int),
2581                 .mode           =       0644,
2582                 .proc_handler   =       proc_dointvec_jiffies,
2583         },
2584         {
2585                 .procname       =       "gc_elasticity",
2586                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
2587                 .maxlen         =       sizeof(int),
2588                 .mode           =       0644,
2589                 .proc_handler   =       proc_dointvec_jiffies,
2590         },
2591         {
2592                 .procname       =       "mtu_expires",
2593                 .data           =       &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
2594                 .maxlen         =       sizeof(int),
2595                 .mode           =       0644,
2596                 .proc_handler   =       proc_dointvec_jiffies,
2597         },
2598         {
2599                 .procname       =       "min_adv_mss",
2600                 .data           =       &init_net.ipv6.sysctl.ip6_rt_min_advmss,
2601                 .maxlen         =       sizeof(int),
2602                 .mode           =       0644,
2603                 .proc_handler   =       proc_dointvec_jiffies,
2604         },
2605         {
2606                 .procname       =       "gc_min_interval_ms",
2607                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2608                 .maxlen         =       sizeof(int),
2609                 .mode           =       0644,
2610                 .proc_handler   =       proc_dointvec_ms_jiffies,
2611         },
2612         { }
2613 };
2614
2615 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
2616 {
2617         struct ctl_table *table;
2618
2619         table = kmemdup(ipv6_route_table_template,
2620                         sizeof(ipv6_route_table_template),
2621                         GFP_KERNEL);
2622
2623         if (table) {
2624                 table[0].data = &net->ipv6.sysctl.flush_delay;
2625                 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
2626                 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
2627                 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2628                 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
2629                 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
2630                 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
2631                 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
2632                 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
2633                 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2634         }
2635
2636         return table;
2637 }
2638 #endif
2639
2640 static int __net_init ip6_route_net_init(struct net *net)
2641 {
2642         int ret = -ENOMEM;
2643
2644         memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
2645                sizeof(net->ipv6.ip6_dst_ops));
2646
2647         net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
2648                                            sizeof(*net->ipv6.ip6_null_entry),
2649                                            GFP_KERNEL);
2650         if (!net->ipv6.ip6_null_entry)
2651                 goto out_ip6_dst_ops;
2652         net->ipv6.ip6_null_entry->u.dst.path =
2653                 (struct dst_entry *)net->ipv6.ip6_null_entry;
2654         net->ipv6.ip6_null_entry->u.dst.ops = &net->ipv6.ip6_dst_ops;
2655
2656 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2657         net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
2658                                                sizeof(*net->ipv6.ip6_prohibit_entry),
2659                                                GFP_KERNEL);
2660         if (!net->ipv6.ip6_prohibit_entry)
2661                 goto out_ip6_null_entry;
2662         net->ipv6.ip6_prohibit_entry->u.dst.path =
2663                 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
2664         net->ipv6.ip6_prohibit_entry->u.dst.ops = &net->ipv6.ip6_dst_ops;
2665
2666         net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
2667                                                sizeof(*net->ipv6.ip6_blk_hole_entry),
2668                                                GFP_KERNEL);
2669         if (!net->ipv6.ip6_blk_hole_entry)
2670                 goto out_ip6_prohibit_entry;
2671         net->ipv6.ip6_blk_hole_entry->u.dst.path =
2672                 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
2673         net->ipv6.ip6_blk_hole_entry->u.dst.ops = &net->ipv6.ip6_dst_ops;
2674 #endif
2675
2676         net->ipv6.sysctl.flush_delay = 0;
2677         net->ipv6.sysctl.ip6_rt_max_size = 4096;
2678         net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
2679         net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
2680         net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
2681         net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
2682         net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
2683         net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
2684
2685 #ifdef CONFIG_PROC_FS
2686         proc_net_fops_create(net, "ipv6_route", 0, &ipv6_route_proc_fops);
2687         proc_net_fops_create(net, "rt6_stats", S_IRUGO, &rt6_stats_seq_fops);
2688 #endif
2689         net->ipv6.ip6_rt_gc_expire = 30*HZ;
2690
2691         ret = 0;
2692 out:
2693         return ret;
2694
2695 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2696 out_ip6_prohibit_entry:
2697         kfree(net->ipv6.ip6_prohibit_entry);
2698 out_ip6_null_entry:
2699         kfree(net->ipv6.ip6_null_entry);
2700 #endif
2701 out_ip6_dst_ops:
2702         goto out;
2703 }
2704
2705 static void __net_exit ip6_route_net_exit(struct net *net)
2706 {
2707 #ifdef CONFIG_PROC_FS
2708         proc_net_remove(net, "ipv6_route");
2709         proc_net_remove(net, "rt6_stats");
2710 #endif
2711         kfree(net->ipv6.ip6_null_entry);
2712 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2713         kfree(net->ipv6.ip6_prohibit_entry);
2714         kfree(net->ipv6.ip6_blk_hole_entry);
2715 #endif
2716 }
2717
2718 static struct pernet_operations ip6_route_net_ops = {
2719         .init = ip6_route_net_init,
2720         .exit = ip6_route_net_exit,
2721 };
2722
2723 static struct notifier_block ip6_route_dev_notifier = {
2724         .notifier_call = ip6_route_dev_notify,
2725         .priority = 0,
2726 };
2727
2728 int __init ip6_route_init(void)
2729 {
2730         int ret;
2731
2732         ret = -ENOMEM;
2733         ip6_dst_ops_template.kmem_cachep =
2734                 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
2735                                   SLAB_HWCACHE_ALIGN, NULL);
2736         if (!ip6_dst_ops_template.kmem_cachep)
2737                 goto out;
2738
2739         ret = register_pernet_subsys(&ip6_route_net_ops);
2740         if (ret)
2741                 goto out_kmem_cache;
2742
2743         ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
2744
2745         /* Registering of the loopback is done before this portion of code,
2746          * the loopback reference in rt6_info will not be taken, do it
2747          * manually for init_net */
2748         init_net.ipv6.ip6_null_entry->u.dst.dev = init_net.loopback_dev;
2749         init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
2750   #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2751         init_net.ipv6.ip6_prohibit_entry->u.dst.dev = init_net.loopback_dev;
2752         init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
2753         init_net.ipv6.ip6_blk_hole_entry->u.dst.dev = init_net.loopback_dev;
2754         init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
2755   #endif
2756         ret = fib6_init();
2757         if (ret)
2758                 goto out_register_subsys;
2759
2760         ret = xfrm6_init();
2761         if (ret)
2762                 goto out_fib6_init;
2763
2764         ret = fib6_rules_init();
2765         if (ret)
2766                 goto xfrm6_init;
2767
2768         ret = -ENOBUFS;
2769         if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL) ||
2770             __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL) ||
2771             __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL))
2772                 goto fib6_rules_init;
2773
2774         ret = register_netdevice_notifier(&ip6_route_dev_notifier);
2775         if (ret)
2776                 goto fib6_rules_init;
2777
2778 out:
2779         return ret;
2780
2781 fib6_rules_init:
2782         fib6_rules_cleanup();
2783 xfrm6_init:
2784         xfrm6_fini();
2785 out_fib6_init:
2786         fib6_gc_cleanup();
2787 out_register_subsys:
2788         unregister_pernet_subsys(&ip6_route_net_ops);
2789 out_kmem_cache:
2790         kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
2791         goto out;
2792 }
2793
2794 void ip6_route_cleanup(void)
2795 {
2796         unregister_netdevice_notifier(&ip6_route_dev_notifier);
2797         fib6_rules_cleanup();
2798         xfrm6_fini();
2799         fib6_gc_cleanup();
2800         unregister_pernet_subsys(&ip6_route_net_ops);
2801         kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
2802 }