net/ipv6/route.c

   1 /*
   2  *      Linux INET6 implementation
   3  *      FIB front-end.
   4  *
   5  *      Authors:
   6  *      Pedro Roque             <roque@di.fc.ul.pt>
   7  *
   8  *      $Id: route.c,v 1.56 2001/10/31 21:55:55 davem Exp $
   9  *
  10  *      This program is free software; you can redistribute it and/or
  11  *      modify it under the terms of the GNU General Public License
  12  *      as published by the Free Software Foundation; either version
  13  *      2 of the License, or (at your option) any later version.
  14  */
  15
  16 /*      Changes:
  17  *
  18  *      YOSHIFUJI Hideaki @USAGI
  19  *              reworked default router selection.
  20  *              - respect outgoing interface
  21  *              - select from (probably) reachable routers (i.e.
  22  *              routers in REACHABLE, STALE, DELAY or PROBE states).
  23  *              - always select the same router if it is (probably)
  24  *              reachable.  otherwise, round-robin the list.
  25  *      Ville Nuorvala
  26  *              Fixed routing subtrees.
  27  */
  28
  29 #include <linux/capability.h>
  30 #include <linux/errno.h>
  31 #include <linux/types.h>
  32 #include <linux/times.h>
  33 #include <linux/socket.h>
  34 #include <linux/sockios.h>
  35 #include <linux/net.h>
  36 #include <linux/route.h>
  37 #include <linux/netdevice.h>
  38 #include <linux/in6.h>
  39 #include <linux/mroute6.h>
  40 #include <linux/init.h>
  41 #include <linux/if_arp.h>
  42 #include <linux/proc_fs.h>
  43 #include <linux/seq_file.h>
  44 #include <linux/nsproxy.h>
  45 #include <net/net_namespace.h>
  46 #include <net/snmp.h>
  47 #include <net/ipv6.h>
  48 #include <net/ip6_fib.h>
  49 #include <net/ip6_route.h>
  50 #include <net/ndisc.h>
  51 #include <net/addrconf.h>
  52 #include <net/tcp.h>
  53 #include <linux/rtnetlink.h>
  54 #include <net/dst.h>
  55 #include <net/xfrm.h>
  56 #include <net/netevent.h>
  57 #include <net/netlink.h>
  58
  59 #include <asm/uaccess.h>
  60
  61 #ifdef CONFIG_SYSCTL
  62 #include <linux/sysctl.h>
  63 #endif
  64
  65 /* Set to 3 to get tracing. */
  66 #define RT6_DEBUG 2
  67
  68 #if RT6_DEBUG >= 3
  69 #define RDBG(x) printk x
  70 #define RT6_TRACE(x...) printk(KERN_DEBUG x)
  71 #else
  72 #define RDBG(x)
  73 #define RT6_TRACE(x...) do { ; } while (0)
  74 #endif
  75
  76 #define CLONE_OFFLINK_ROUTE 0
  77
  78 static struct rt6_info * ip6_rt_copy(struct rt6_info *ort);
  79 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie);
  80 static struct dst_entry *ip6_negative_advice(struct dst_entry *);
  81 static void             ip6_dst_destroy(struct dst_entry *);
  82 static void             ip6_dst_ifdown(struct dst_entry *,
  83                                        struct net_device *dev, int how);
  84 static int               ip6_dst_gc(struct dst_ops *ops);
  85
  86 static int              ip6_pkt_discard(struct sk_buff *skb);
  87 static int              ip6_pkt_discard_out(struct sk_buff *skb);
  88 static void             ip6_link_failure(struct sk_buff *skb);
  89 static void             ip6_rt_update_pmtu(struct dst_entry *dst, u32 mtu);
  90
  91 #ifdef CONFIG_IPV6_ROUTE_INFO
  92 static struct rt6_info *rt6_add_route_info(struct net *net,
  93                                            struct in6_addr *prefix, int prefixlen,
  94                                            struct in6_addr *gwaddr, int ifindex,
  95                                            unsigned pref);
  96 static struct rt6_info *rt6_get_route_info(struct net *net,
  97                                            struct in6_addr *prefix, int prefixlen,
  98                                            struct in6_addr *gwaddr, int ifindex);
  99 #endif
 100
 101 static struct dst_ops ip6_dst_ops_template = {
 102         .family                 =       AF_INET6,
 103         .protocol               =       __constant_htons(ETH_P_IPV6),
 104         .gc                     =       ip6_dst_gc,
 105         .gc_thresh              =       1024,
 106         .check                  =       ip6_dst_check,
 107         .destroy                =       ip6_dst_destroy,
 108         .ifdown                 =       ip6_dst_ifdown,
 109         .negative_advice        =       ip6_negative_advice,
 110         .link_failure           =       ip6_link_failure,
 111         .update_pmtu            =       ip6_rt_update_pmtu,
 112         .local_out              =       ip6_local_out,
 113         .entry_size             =       sizeof(struct rt6_info),
 114         .entries                =       ATOMIC_INIT(0),
 115 };
 116
 117 static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, u32 mtu)
 118 {
 119 }
 120
 121 static struct dst_ops ip6_dst_blackhole_ops = {
 122         .family                 =       AF_INET6,
 123         .protocol               =       __constant_htons(ETH_P_IPV6),
 124         .destroy                =       ip6_dst_destroy,
 125         .check                  =       ip6_dst_check,
 126         .update_pmtu            =       ip6_rt_blackhole_update_pmtu,
 127         .entry_size             =       sizeof(struct rt6_info),
 128         .entries                =       ATOMIC_INIT(0),
 129 };
 130
 131 static struct rt6_info ip6_null_entry_template = {
 132         .u = {
 133                 .dst = {
 134                         .__refcnt       = ATOMIC_INIT(1),
 135                         .__use          = 1,
 136                         .obsolete       = -1,
 137                         .error          = -ENETUNREACH,
 138                         .metrics        = { [RTAX_HOPLIMIT - 1] = 255, },
 139                         .input          = ip6_pkt_discard,
 140                         .output         = ip6_pkt_discard_out,
 141                 }
 142         },
 143         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
 144         .rt6i_metric    = ~(u32) 0,
 145         .rt6i_ref       = ATOMIC_INIT(1),
 146 };
 147
 148 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
 149
 150 static int ip6_pkt_prohibit(struct sk_buff *skb);
 151 static int ip6_pkt_prohibit_out(struct sk_buff *skb);
 152
 153 struct rt6_info ip6_prohibit_entry_template = {
 154         .u = {
 155                 .dst = {
 156                         .__refcnt       = ATOMIC_INIT(1),
 157                         .__use          = 1,
 158                         .obsolete       = -1,
 159                         .error          = -EACCES,
 160                         .metrics        = { [RTAX_HOPLIMIT - 1] = 255, },
 161                         .input          = ip6_pkt_prohibit,
 162                         .output         = ip6_pkt_prohibit_out,
 163                 }
 164         },
 165         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
 166         .rt6i_metric    = ~(u32) 0,
 167         .rt6i_ref       = ATOMIC_INIT(1),
 168 };
 169
 170 static struct rt6_info ip6_blk_hole_entry_template = {
 171         .u = {
 172                 .dst = {
 173                         .__refcnt       = ATOMIC_INIT(1),
 174                         .__use          = 1,
 175                         .obsolete       = -1,
 176                         .error          = -EINVAL,
 177                         .metrics        = { [RTAX_HOPLIMIT - 1] = 255, },
 178                         .input          = dst_discard,
 179                         .output         = dst_discard,
 180                 }
 181         },
 182         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
 183         .rt6i_metric    = ~(u32) 0,
 184         .rt6i_ref       = ATOMIC_INIT(1),
 185 };
 186
 187 #endif
 188
 189 /* allocate dst with ip6_dst_ops */
 190 static inline struct rt6_info *ip6_dst_alloc(struct dst_ops *ops)
 191 {
 192         return (struct rt6_info *)dst_alloc(ops);
 193 }
 194
 195 static void ip6_dst_destroy(struct dst_entry *dst)
 196 {
 197         struct rt6_info *rt = (struct rt6_info *)dst;
 198         struct inet6_dev *idev = rt->rt6i_idev;
 199
 200         if (idev != NULL) {
 201                 rt->rt6i_idev = NULL;
 202                 in6_dev_put(idev);
 203         }
 204 }
 205
 206 static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
 207                            int how)
 208 {
 209         struct rt6_info *rt = (struct rt6_info *)dst;
 210         struct inet6_dev *idev = rt->rt6i_idev;
 211         struct net_device *loopback_dev =
 212                 dev_net(dev)->loopback_dev;
 213
 214         if (dev != loopback_dev && idev != NULL && idev->dev == dev) {
 215                 struct inet6_dev *loopback_idev =
 216                         in6_dev_get(loopback_dev);
 217                 if (loopback_idev != NULL) {
 218                         rt->rt6i_idev = loopback_idev;
 219                         in6_dev_put(idev);
 220                 }
 221         }
 222 }
 223
 224 static __inline__ int rt6_check_expired(const struct rt6_info *rt)
 225 {
 226         return (rt->rt6i_flags & RTF_EXPIRES &&
 227                 time_after(jiffies, rt->rt6i_expires));
 228 }
 229
 230 static inline int rt6_need_strict(struct in6_addr *daddr)
 231 {
 232         return (ipv6_addr_type(daddr) &
 233                 (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL));
 234 }
 235
 236 /*
 237  *      Route lookup. Any table->tb6_lock is implied.
 238  */
 239
 240 static inline struct rt6_info *rt6_device_match(struct net *net,
 241                                                     struct rt6_info *rt,
 242                                                     int oif,
 243                                                     int strict)
 244 {
 245         struct rt6_info *local = NULL;
 246         struct rt6_info *sprt;
 247
 248         if (oif) {
 249                 for (sprt = rt; sprt; sprt = sprt->u.dst.rt6_next) {
 250                         struct net_device *dev = sprt->rt6i_dev;
 251                         if (dev->ifindex == oif)
 252                                 return sprt;
 253                         if (dev->flags & IFF_LOOPBACK) {
 254                                 if (sprt->rt6i_idev == NULL ||
 255                                     sprt->rt6i_idev->dev->ifindex != oif) {
 256                                         if (strict && oif)
 257                                                 continue;
 258                                         if (local && (!oif ||
 259                                                       local->rt6i_idev->dev->ifindex == oif))
 260                                                 continue;
 261                                 }
 262                                 local = sprt;
 263                         }
 264                 }
 265
 266                 if (local)
 267                         return local;
 268
 269                 if (strict)
 270                         return net->ipv6.ip6_null_entry;
 271         }
 272         return rt;
 273 }
 274
 275 #ifdef CONFIG_IPV6_ROUTER_PREF
 276 static void rt6_probe(struct rt6_info *rt)
 277 {
 278         struct neighbour *neigh = rt ? rt->rt6i_nexthop : NULL;
 279         /*
 280          * Okay, this does not seem to be appropriate
 281          * for now, however, we need to check if it
 282          * is really so; aka Router Reachability Probing.
 283          *
 284          * Router Reachability Probe MUST be rate-limited
 285          * to no more than one per minute.
 286          */
 287         if (!neigh || (neigh->nud_state & NUD_VALID))
 288                 return;
 289         read_lock_bh(&neigh->lock);
 290         if (!(neigh->nud_state & NUD_VALID) &&
 291             time_after(jiffies, neigh->updated + rt->rt6i_idev->cnf.rtr_probe_interval)) {
 292                 struct in6_addr mcaddr;
 293                 struct in6_addr *target;
 294
 295                 neigh->updated = jiffies;
 296                 read_unlock_bh(&neigh->lock);
 297
 298                 target = (struct in6_addr *)&neigh->primary_key;
 299                 addrconf_addr_solict_mult(target, &mcaddr);
 300                 ndisc_send_ns(rt->rt6i_dev, NULL, target, &mcaddr, NULL);
 301         } else
 302                 read_unlock_bh(&neigh->lock);
 303 }
 304 #else
 305 static inline void rt6_probe(struct rt6_info *rt)
 306 {
 307         return;
 308 }
 309 #endif
 310
 311 /*
 312  * Default Router Selection (RFC 2461 6.3.6)
 313  */
 314 static inline int rt6_check_dev(struct rt6_info *rt, int oif)
 315 {
 316         struct net_device *dev = rt->rt6i_dev;
 317         if (!oif || dev->ifindex == oif)
 318                 return 2;
 319         if ((dev->flags & IFF_LOOPBACK) &&
 320             rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
 321                 return 1;
 322         return 0;
 323 }
 324
 325 static inline int rt6_check_neigh(struct rt6_info *rt)
 326 {
 327         struct neighbour *neigh = rt->rt6i_nexthop;
 328         int m;
 329         if (rt->rt6i_flags & RTF_NONEXTHOP ||
 330             !(rt->rt6i_flags & RTF_GATEWAY))
 331                 m = 1;
 332         else if (neigh) {
 333                 read_lock_bh(&neigh->lock);
 334                 if (neigh->nud_state & NUD_VALID)
 335                         m = 2;
 336 #ifdef CONFIG_IPV6_ROUTER_PREF
 337                 else if (neigh->nud_state & NUD_FAILED)
 338                         m = 0;
 339 #endif
 340                 else
 341                         m = 1;
 342                 read_unlock_bh(&neigh->lock);
 343         } else
 344                 m = 0;
 345         return m;
 346 }
 347
 348 static int rt6_score_route(struct rt6_info *rt, int oif,
 349                            int strict)
 350 {
 351         int m, n;
 352
 353         m = rt6_check_dev(rt, oif);
 354         if (!m && (strict & RT6_LOOKUP_F_IFACE))
 355                 return -1;
 356 #ifdef CONFIG_IPV6_ROUTER_PREF
 357         m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
 358 #endif
 359         n = rt6_check_neigh(rt);
 360         if (!n && (strict & RT6_LOOKUP_F_REACHABLE))
 361                 return -1;
 362         return m;
 363 }
 364
 365 static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
 366                                    int *mpri, struct rt6_info *match)
 367 {
 368         int m;
 369
 370         if (rt6_check_expired(rt))
 371                 goto out;
 372
 373         m = rt6_score_route(rt, oif, strict);
 374         if (m < 0)
 375                 goto out;
 376
 377         if (m > *mpri) {
 378                 if (strict & RT6_LOOKUP_F_REACHABLE)
 379                         rt6_probe(match);
 380                 *mpri = m;
 381                 match = rt;
 382         } else if (strict & RT6_LOOKUP_F_REACHABLE) {
 383                 rt6_probe(rt);
 384         }
 385
 386 out:
 387         return match;
 388 }
 389
 390 static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
 391                                      struct rt6_info *rr_head,
 392                                      u32 metric, int oif, int strict)
 393 {
 394         struct rt6_info *rt, *match;
 395         int mpri = -1;
 396
 397         match = NULL;
 398         for (rt = rr_head; rt && rt->rt6i_metric == metric;
 399              rt = rt->u.dst.rt6_next)
 400                 match = find_match(rt, oif, strict, &mpri, match);
 401         for (rt = fn->leaf; rt && rt != rr_head && rt->rt6i_metric == metric;
 402              rt = rt->u.dst.rt6_next)
 403                 match = find_match(rt, oif, strict, &mpri, match);
 404
 405         return match;
 406 }
 407
 408 static struct rt6_info *rt6_select(struct fib6_node *fn, int oif, int strict)
 409 {
 410         struct rt6_info *match, *rt0;
 411         struct net *net;
 412
 413         RT6_TRACE("%s(fn->leaf=%p, oif=%d)\n",
 414                   __func__, fn->leaf, oif);
 415
 416         rt0 = fn->rr_ptr;
 417         if (!rt0)
 418                 fn->rr_ptr = rt0 = fn->leaf;
 419
 420         match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict);
 421
 422         if (!match &&
 423             (strict & RT6_LOOKUP_F_REACHABLE)) {
 424                 struct rt6_info *next = rt0->u.dst.rt6_next;
 425
 426                 /* no entries matched; do round-robin */
 427                 if (!next || next->rt6i_metric != rt0->rt6i_metric)
 428                         next = fn->leaf;
 429
 430                 if (next != rt0)
 431                         fn->rr_ptr = next;
 432         }
 433
 434         RT6_TRACE("%s() => %p\n",
 435                   __func__, match);
 436
 437         net = dev_net(rt0->rt6i_dev);
 438         return (match ? match : net->ipv6.ip6_null_entry);
 439 }
 440
 441 #ifdef CONFIG_IPV6_ROUTE_INFO
 442 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
 443                   struct in6_addr *gwaddr)
 444 {
 445         struct net *net = dev_net(dev);
 446         struct route_info *rinfo = (struct route_info *) opt;
 447         struct in6_addr prefix_buf, *prefix;
 448         unsigned int pref;
 449         u32 lifetime;
 450         struct rt6_info *rt;
 451
 452         if (len < sizeof(struct route_info)) {
 453                 return -EINVAL;
 454         }
 455
 456         /* Sanity check for prefix_len and length */
 457         if (rinfo->length > 3) {
 458                 return -EINVAL;
 459         } else if (rinfo->prefix_len > 128) {
 460                 return -EINVAL;
 461         } else if (rinfo->prefix_len > 64) {
 462                 if (rinfo->length < 2) {
 463                         return -EINVAL;
 464                 }
 465         } else if (rinfo->prefix_len > 0) {
 466                 if (rinfo->length < 1) {
 467                         return -EINVAL;
 468                 }
 469         }
 470
 471         pref = rinfo->route_pref;
 472         if (pref == ICMPV6_ROUTER_PREF_INVALID)
 473                 pref = ICMPV6_ROUTER_PREF_MEDIUM;
 474
 475         lifetime = ntohl(rinfo->lifetime);
 476         if (lifetime == 0xffffffff) {
 477                 /* infinity */
 478         } else if (lifetime > 0x7fffffff/HZ) {
 479                 /* Avoid arithmetic overflow */
 480                 lifetime = 0x7fffffff/HZ - 1;
 481         }
 482
 483         if (rinfo->length == 3)
 484                 prefix = (struct in6_addr *)rinfo->prefix;
 485         else {
 486                 /* this function is safe */
 487                 ipv6_addr_prefix(&prefix_buf,
 488                                  (struct in6_addr *)rinfo->prefix,
 489                                  rinfo->prefix_len);
 490                 prefix = &prefix_buf;
 491         }
 492
 493         rt = rt6_get_route_info(net, prefix, rinfo->prefix_len, gwaddr,
 494                                 dev->ifindex);
 495
 496         if (rt && !lifetime) {
 497                 ip6_del_rt(rt);
 498                 rt = NULL;
 499         }
 500
 501         if (!rt && lifetime)
 502                 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr, dev->ifindex,
 503                                         pref);
 504         else if (rt)
 505                 rt->rt6i_flags = RTF_ROUTEINFO |
 506                                  (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
 507
 508         if (rt) {
 509                 if (lifetime == 0xffffffff) {
 510                         rt->rt6i_flags &= ~RTF_EXPIRES;
 511                 } else {
 512                         rt->rt6i_expires = jiffies + HZ * lifetime;
 513                         rt->rt6i_flags |= RTF_EXPIRES;
 514                 }
 515                 dst_release(&rt->u.dst);
 516         }
 517         return 0;
 518 }
 519 #endif
 520
 521 #define BACKTRACK(__net, saddr)                 \
 522 do { \
 523         if (rt == __net->ipv6.ip6_null_entry) { \
 524                 struct fib6_node *pn; \
 525                 while (1) { \
 526                         if (fn->fn_flags & RTN_TL_ROOT) \
 527                                 goto out; \
 528                         pn = fn->parent; \
 529                         if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn) \
 530                                 fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr); \
 531                         else \
 532                                 fn = pn; \
 533                         if (fn->fn_flags & RTN_RTINFO) \
 534                                 goto restart; \
 535                 } \
 536         } \
 537 } while(0)
 538
 539 static struct rt6_info *ip6_pol_route_lookup(struct net *net,
 540                                              struct fib6_table *table,
 541                                              struct flowi *fl, int flags)
 542 {
 543         struct fib6_node *fn;
 544         struct rt6_info *rt;
 545
 546         read_lock_bh(&table->tb6_lock);
 547         fn = fib6_lookup(&table->tb6_root, &fl->fl6_dst, &fl->fl6_src);
 548 restart:
 549         rt = fn->leaf;
 550         rt = rt6_device_match(net, rt, fl->oif, flags);
 551         BACKTRACK(net, &fl->fl6_src);
 552 out:
 553         dst_use(&rt->u.dst, jiffies);
 554         read_unlock_bh(&table->tb6_lock);
 555         return rt;
 556
 557 }
 558
 559 struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
 560                             const struct in6_addr *saddr, int oif, int strict)
 561 {
 562         struct flowi fl = {
 563                 .oif = oif,
 564                 .nl_u = {
 565                         .ip6_u = {
 566                                 .daddr = *daddr,
 567                         },
 568                 },
 569         };
 570         struct dst_entry *dst;
 571         int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
 572
 573         if (saddr) {
 574                 memcpy(&fl.fl6_src, saddr, sizeof(*saddr));
 575                 flags |= RT6_LOOKUP_F_HAS_SADDR;
 576         }
 577
 578         dst = fib6_rule_lookup(net, &fl, flags, ip6_pol_route_lookup);
 579         if (dst->error == 0)
 580                 return (struct rt6_info *) dst;
 581
 582         dst_release(dst);
 583
 584         return NULL;
 585 }
 586
 587 EXPORT_SYMBOL(rt6_lookup);
 588
 589 /* ip6_ins_rt is called with FREE table->tb6_lock.
 590    It takes new route entry, the addition fails by any reason the
 591    route is freed. In any case, if caller does not hold it, it may
 592    be destroyed.
 593  */
 594
 595 static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info)
 596 {
 597         int err;
 598         struct fib6_table *table;
 599
 600         table = rt->rt6i_table;
 601         write_lock_bh(&table->tb6_lock);
 602         err = fib6_add(&table->tb6_root, rt, info);
 603         write_unlock_bh(&table->tb6_lock);
 604
 605         return err;
 606 }
 607
 608 int ip6_ins_rt(struct rt6_info *rt)
 609 {
 610         struct nl_info info = {
 611                 .nl_net = dev_net(rt->rt6i_dev),
 612         };
 613         return __ip6_ins_rt(rt, &info);
 614 }
 615
 616 static struct rt6_info *rt6_alloc_cow(struct rt6_info *ort, struct in6_addr *daddr,
 617                                       struct in6_addr *saddr)
 618 {
 619         struct rt6_info *rt;
 620
 621         /*
 622          *      Clone the route.
 623          */
 624
 625         rt = ip6_rt_copy(ort);
 626
 627         if (rt) {
 628                 if (!(rt->rt6i_flags&RTF_GATEWAY)) {
 629                         if (rt->rt6i_dst.plen != 128 &&
 630                             ipv6_addr_equal(&rt->rt6i_dst.addr, daddr))
 631                                 rt->rt6i_flags |= RTF_ANYCAST;
 632                         ipv6_addr_copy(&rt->rt6i_gateway, daddr);
 633                 }
 634
 635                 ipv6_addr_copy(&rt->rt6i_dst.addr, daddr);
 636                 rt->rt6i_dst.plen = 128;
 637                 rt->rt6i_flags |= RTF_CACHE;
 638                 rt->u.dst.flags |= DST_HOST;
 639
 640 #ifdef CONFIG_IPV6_SUBTREES
 641                 if (rt->rt6i_src.plen && saddr) {
 642                         ipv6_addr_copy(&rt->rt6i_src.addr, saddr);
 643                         rt->rt6i_src.plen = 128;
 644                 }
 645 #endif
 646
 647                 rt->rt6i_nexthop = ndisc_get_neigh(rt->rt6i_dev, &rt->rt6i_gateway);
 648
 649         }
 650
 651         return rt;
 652 }
 653
 654 static struct rt6_info *rt6_alloc_clone(struct rt6_info *ort, struct in6_addr *daddr)
 655 {
 656         struct rt6_info *rt = ip6_rt_copy(ort);
 657         if (rt) {
 658                 ipv6_addr_copy(&rt->rt6i_dst.addr, daddr);
 659                 rt->rt6i_dst.plen = 128;
 660                 rt->rt6i_flags |= RTF_CACHE;
 661                 rt->u.dst.flags |= DST_HOST;
 662                 rt->rt6i_nexthop = neigh_clone(ort->rt6i_nexthop);
 663         }
 664         return rt;
 665 }
 666
 667 static struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table, int oif,
 668                                       struct flowi *fl, int flags)
 669 {
 670         struct fib6_node *fn;
 671         struct rt6_info *rt, *nrt;
 672         int strict = 0;
 673         int attempts = 3;
 674         int err;
 675         int reachable = ipv6_devconf.forwarding ? 0 : RT6_LOOKUP_F_REACHABLE;
 676
 677         strict |= flags & RT6_LOOKUP_F_IFACE;
 678
 679 relookup:
 680         read_lock_bh(&table->tb6_lock);
 681
 682 restart_2:
 683         fn = fib6_lookup(&table->tb6_root, &fl->fl6_dst, &fl->fl6_src);
 684
 685 restart:
 686         rt = rt6_select(fn, oif, strict | reachable);
 687
 688         BACKTRACK(net, &fl->fl6_src);
 689         if (rt == net->ipv6.ip6_null_entry ||
 690             rt->rt6i_flags & RTF_CACHE)
 691                 goto out;
 692
 693         dst_hold(&rt->u.dst);
 694         read_unlock_bh(&table->tb6_lock);
 695
 696         if (!rt->rt6i_nexthop && !(rt->rt6i_flags & RTF_NONEXTHOP))
 697                 nrt = rt6_alloc_cow(rt, &fl->fl6_dst, &fl->fl6_src);
 698         else {
 699 #if CLONE_OFFLINK_ROUTE
 700                 nrt = rt6_alloc_clone(rt, &fl->fl6_dst);
 701 #else
 702                 goto out2;
 703 #endif
 704         }
 705
 706         dst_release(&rt->u.dst);
 707         rt = nrt ? : net->ipv6.ip6_null_entry;
 708
 709         dst_hold(&rt->u.dst);
 710         if (nrt) {
 711                 err = ip6_ins_rt(nrt);
 712                 if (!err)
 713                         goto out2;
 714         }
 715
 716         if (--attempts <= 0)
 717                 goto out2;
 718
 719         /*
 720          * Race condition! In the gap, when table->tb6_lock was
 721          * released someone could insert this route.  Relookup.
 722          */
 723         dst_release(&rt->u.dst);
 724         goto relookup;
 725
 726 out:
 727         if (reachable) {
 728                 reachable = 0;
 729                 goto restart_2;
 730         }
 731         dst_hold(&rt->u.dst);
 732         read_unlock_bh(&table->tb6_lock);
 733 out2:
 734         rt->u.dst.lastuse = jiffies;
 735         rt->u.dst.__use++;
 736
 737         return rt;
 738 }
 739
 740 static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
 741                                             struct flowi *fl, int flags)
 742 {
 743         return ip6_pol_route(net, table, fl->iif, fl, flags);
 744 }
 745
 746 void ip6_route_input(struct sk_buff *skb)
 747 {
 748         struct ipv6hdr *iph = ipv6_hdr(skb);
 749         struct net *net = dev_net(skb->dev);
 750         int flags = RT6_LOOKUP_F_HAS_SADDR;
 751         struct flowi fl = {
 752                 .iif = skb->dev->ifindex,
 753                 .nl_u = {
 754                         .ip6_u = {
 755                                 .daddr = iph->daddr,
 756                                 .saddr = iph->saddr,
 757                                 .flowlabel = (* (__be32 *) iph)&IPV6_FLOWINFO_MASK,
 758                         },
 759                 },
 760                 .mark = skb->mark,
 761                 .proto = iph->nexthdr,
 762         };
 763
 764         if (rt6_need_strict(&iph->daddr))
 765                 flags |= RT6_LOOKUP_F_IFACE;
 766
 767         skb->dst = fib6_rule_lookup(net, &fl, flags, ip6_pol_route_input);
 768 }
 769
 770 static struct rt6_info *ip6_pol_route_output(struct net *net, struct fib6_table *table,
 771                                              struct flowi *fl, int flags)
 772 {
 773         return ip6_pol_route(net, table, fl->oif, fl, flags);
 774 }
 775
 776 struct dst_entry * ip6_route_output(struct net *net, struct sock *sk,
 777                                     struct flowi *fl)
 778 {
 779         int flags = 0;
 780
 781         if (rt6_need_strict(&fl->fl6_dst))
 782                 flags |= RT6_LOOKUP_F_IFACE;
 783
 784         if (!ipv6_addr_any(&fl->fl6_src))
 785                 flags |= RT6_LOOKUP_F_HAS_SADDR;
 786         else if (sk) {
 787                 unsigned int prefs = inet6_sk(sk)->srcprefs;
 788                 if (prefs & IPV6_PREFER_SRC_TMP)
 789                         flags |= RT6_LOOKUP_F_SRCPREF_TMP;
 790                 if (prefs & IPV6_PREFER_SRC_PUBLIC)
 791                         flags |= RT6_LOOKUP_F_SRCPREF_PUBLIC;
 792                 if (prefs & IPV6_PREFER_SRC_COA)
 793                         flags |= RT6_LOOKUP_F_SRCPREF_COA;
 794         }
 795
 796         return fib6_rule_lookup(net, fl, flags, ip6_pol_route_output);
 797 }
 798
 799 EXPORT_SYMBOL(ip6_route_output);
 800
 801 int ip6_dst_blackhole(struct sock *sk, struct dst_entry **dstp, struct flowi *fl)
 802 {
 803         struct rt6_info *ort = (struct rt6_info *) *dstp;
 804         struct rt6_info *rt = (struct rt6_info *)
 805                 dst_alloc(&ip6_dst_blackhole_ops);
 806         struct dst_entry *new = NULL;
 807
 808         if (rt) {
 809                 new = &rt->u.dst;
 810
 811                 atomic_set(&new->__refcnt, 1);
 812                 new->__use = 1;
 813                 new->input = dst_discard;
 814                 new->output = dst_discard;
 815
 816                 memcpy(new->metrics, ort->u.dst.metrics, RTAX_MAX*sizeof(u32));
 817                 new->dev = ort->u.dst.dev;
 818                 if (new->dev)
 819                         dev_hold(new->dev);
 820                 rt->rt6i_idev = ort->rt6i_idev;
 821                 if (rt->rt6i_idev)
 822                         in6_dev_hold(rt->rt6i_idev);
 823                 rt->rt6i_expires = 0;
 824
 825                 ipv6_addr_copy(&rt->rt6i_gateway, &ort->rt6i_gateway);
 826                 rt->rt6i_flags = ort->rt6i_flags & ~RTF_EXPIRES;
 827                 rt->rt6i_metric = 0;
 828
 829                 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
 830 #ifdef CONFIG_IPV6_SUBTREES
 831                 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
 832 #endif
 833
 834                 dst_free(new);
 835         }
 836
 837         dst_release(*dstp);
 838         *dstp = new;
 839         return (new ? 0 : -ENOMEM);
 840 }
 841 EXPORT_SYMBOL_GPL(ip6_dst_blackhole);
 842
 843 /*
 844  *      Destination cache support functions
 845  */
 846
 847 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
 848 {
 849         struct rt6_info *rt;
 850
 851         rt = (struct rt6_info *) dst;
 852
 853         if (rt && rt->rt6i_node && (rt->rt6i_node->fn_sernum == cookie))
 854                 return dst;
 855
 856         return NULL;
 857 }
 858
 859 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
 860 {
 861         struct rt6_info *rt = (struct rt6_info *) dst;
 862
 863         if (rt) {
 864                 if (rt->rt6i_flags & RTF_CACHE)
 865                         ip6_del_rt(rt);
 866                 else
 867                         dst_release(dst);
 868         }
 869         return NULL;
 870 }
 871
 872 static void ip6_link_failure(struct sk_buff *skb)
 873 {
 874         struct rt6_info *rt;
 875
 876         icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0, skb->dev);
 877
 878         rt = (struct rt6_info *) skb->dst;
 879         if (rt) {
 880                 if (rt->rt6i_flags&RTF_CACHE) {
 881                         dst_set_expires(&rt->u.dst, 0);
 882                         rt->rt6i_flags |= RTF_EXPIRES;
 883                 } else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT))
 884                         rt->rt6i_node->fn_sernum = -1;
 885         }
 886 }
 887
 888 static void ip6_rt_update_pmtu(struct dst_entry *dst, u32 mtu)
 889 {
 890         struct rt6_info *rt6 = (struct rt6_info*)dst;
 891
 892         if (mtu < dst_mtu(dst) && rt6->rt6i_dst.plen == 128) {
 893                 rt6->rt6i_flags |= RTF_MODIFIED;
 894                 if (mtu < IPV6_MIN_MTU) {
 895                         mtu = IPV6_MIN_MTU;
 896                         dst->metrics[RTAX_FEATURES-1] |= RTAX_FEATURE_ALLFRAG;
 897                 }
 898                 dst->metrics[RTAX_MTU-1] = mtu;
 899                 call_netevent_notifiers(NETEVENT_PMTU_UPDATE, dst);
 900         }
 901 }
 902
 903 static int ipv6_get_mtu(struct net_device *dev);
 904
 905 static inline unsigned int ipv6_advmss(struct net *net, unsigned int mtu)
 906 {
 907         mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
 908
 909         if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
 910                 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
 911
 912         /*
 913          * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
 914          * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
 915          * IPV6_MAXPLEN is also valid and means: "any MSS,
 916          * rely only on pmtu discovery"
 917          */
 918         if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
 919                 mtu = IPV6_MAXPLEN;
 920         return mtu;
 921 }
 922
 923 static struct dst_entry *icmp6_dst_gc_list;
 924 static DEFINE_SPINLOCK(icmp6_dst_lock);
 925
 926 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
 927                                   struct neighbour *neigh,
 928                                   const struct in6_addr *addr)
 929 {
 930         struct rt6_info *rt;
 931         struct inet6_dev *idev = in6_dev_get(dev);
 932         struct net *net = dev_net(dev);
 933
 934         if (unlikely(idev == NULL))
 935                 return NULL;
 936
 937         rt = ip6_dst_alloc(net->ipv6.ip6_dst_ops);
 938         if (unlikely(rt == NULL)) {
 939                 in6_dev_put(idev);
 940                 goto out;
 941         }
 942
 943         dev_hold(dev);
 944         if (neigh)
 945                 neigh_hold(neigh);
 946         else
 947                 neigh = ndisc_get_neigh(dev, addr);
 948
 949         rt->rt6i_dev      = dev;
 950         rt->rt6i_idev     = idev;
 951         rt->rt6i_nexthop  = neigh;
 952         atomic_set(&rt->u.dst.__refcnt, 1);
 953         rt->u.dst.metrics[RTAX_HOPLIMIT-1] = 255;
 954         rt->u.dst.metrics[RTAX_MTU-1] = ipv6_get_mtu(rt->rt6i_dev);
 955         rt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(net, dst_mtu(&rt->u.dst));
 956         rt->u.dst.output  = ip6_output;
 957
 958 #if 0   /* there's no chance to use these for ndisc */
 959         rt->u.dst.flags   = ipv6_addr_type(addr) & IPV6_ADDR_UNICAST
 960                                 ? DST_HOST
 961                                 : 0;
 962         ipv6_addr_copy(&rt->rt6i_dst.addr, addr);
 963         rt->rt6i_dst.plen = 128;
 964 #endif
 965
 966         spin_lock_bh(&icmp6_dst_lock);
 967         rt->u.dst.next = icmp6_dst_gc_list;
 968         icmp6_dst_gc_list = &rt->u.dst;
 969         spin_unlock_bh(&icmp6_dst_lock);
 970
 971         fib6_force_start_gc(net);
 972
 973 out:
 974         return &rt->u.dst;
 975 }
 976
 977 int icmp6_dst_gc(int *more)
 978 {
 979         struct dst_entry *dst, *next, **pprev;
 980         int freed;
 981
 982         next = NULL;
 983         freed = 0;
 984
 985         spin_lock_bh(&icmp6_dst_lock);
 986         pprev = &icmp6_dst_gc_list;
 987
 988         while ((dst = *pprev) != NULL) {
 989                 if (!atomic_read(&dst->__refcnt)) {
 990                         *pprev = dst->next;
 991                         dst_free(dst);
 992                         freed++;
 993                 } else {
 994                         pprev = &dst->next;
 995                         (*more)++;
 996                 }
 997         }
 998
 999         spin_unlock_bh(&icmp6_dst_lock);
1000
1001         return freed;
1002 }
1003
1004 static int ip6_dst_gc(struct dst_ops *ops)
1005 {
1006         unsigned long now = jiffies;
1007         struct net *net = ops->dst_net;
1008         int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1009         int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1010         int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1011         int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1012         unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1013
1014         if (time_after(rt_last_gc + rt_min_interval, now) &&
1015             atomic_read(&ops->entries) <= rt_max_size)
1016                 goto out;
1017
1018         net->ipv6.ip6_rt_gc_expire++;
1019         fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net);
1020         net->ipv6.ip6_rt_last_gc = now;
1021         if (atomic_read(&ops->entries) < ops->gc_thresh)
1022                 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1023 out:
1024         net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1025         return (atomic_read(&ops->entries) > rt_max_size);
1026 }
1027
1028 /* Clean host part of a prefix. Not necessary in radix tree,
1029    but results in cleaner routing tables.
1030
1031    Remove it only when all the things will work!
1032  */
1033
1034 static int ipv6_get_mtu(struct net_device *dev)
1035 {
1036         int mtu = IPV6_MIN_MTU;
1037         struct inet6_dev *idev;
1038
1039         idev = in6_dev_get(dev);
1040         if (idev) {
1041                 mtu = idev->cnf.mtu6;
1042                 in6_dev_put(idev);
1043         }
1044         return mtu;
1045 }
1046
1047 int ip6_dst_hoplimit(struct dst_entry *dst)
1048 {
1049         int hoplimit = dst_metric(dst, RTAX_HOPLIMIT);
1050         if (hoplimit < 0) {
1051                 struct net_device *dev = dst->dev;
1052                 struct inet6_dev *idev = in6_dev_get(dev);
1053                 if (idev) {
1054                         hoplimit = idev->cnf.hop_limit;
1055                         in6_dev_put(idev);
1056                 } else
1057                         hoplimit = ipv6_devconf.hop_limit;
1058         }
1059         return hoplimit;
1060 }
1061
1062 /*
1063  *
1064  */
1065
1066 int ip6_route_add(struct fib6_config *cfg)
1067 {
1068         int err;
1069         struct net *net = cfg->fc_nlinfo.nl_net;
1070         struct rt6_info *rt = NULL;
1071         struct net_device *dev = NULL;
1072         struct inet6_dev *idev = NULL;
1073         struct fib6_table *table;
1074         int addr_type;
1075
1076         if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1077                 return -EINVAL;
1078 #ifndef CONFIG_IPV6_SUBTREES
1079         if (cfg->fc_src_len)
1080                 return -EINVAL;
1081 #endif
1082         if (cfg->fc_ifindex) {
1083                 err = -ENODEV;
1084                 dev = dev_get_by_index(net, cfg->fc_ifindex);
1085                 if (!dev)
1086                         goto out;
1087                 idev = in6_dev_get(dev);
1088                 if (!idev)
1089                         goto out;
1090         }
1091
1092         if (cfg->fc_metric == 0)
1093                 cfg->fc_metric = IP6_RT_PRIO_USER;
1094
1095         table = fib6_new_table(net, cfg->fc_table);
1096         if (table == NULL) {
1097                 err = -ENOBUFS;
1098                 goto out;
1099         }
1100
1101         rt = ip6_dst_alloc(net->ipv6.ip6_dst_ops);
1102
1103         if (rt == NULL) {
1104                 err = -ENOMEM;
1105                 goto out;
1106         }
1107
1108         rt->u.dst.obsolete = -1;
1109         rt->rt6i_expires = jiffies + clock_t_to_jiffies(cfg->fc_expires);
1110
1111         if (cfg->fc_protocol == RTPROT_UNSPEC)
1112                 cfg->fc_protocol = RTPROT_BOOT;
1113         rt->rt6i_protocol = cfg->fc_protocol;
1114
1115         addr_type = ipv6_addr_type(&cfg->fc_dst);
1116
1117         if (addr_type & IPV6_ADDR_MULTICAST)
1118                 rt->u.dst.input = ip6_mc_input;
1119         else
1120                 rt->u.dst.input = ip6_forward;
1121
1122         rt->u.dst.output = ip6_output;
1123
1124         ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1125         rt->rt6i_dst.plen = cfg->fc_dst_len;
1126         if (rt->rt6i_dst.plen == 128)
1127                rt->u.dst.flags = DST_HOST;
1128
1129 #ifdef CONFIG_IPV6_SUBTREES
1130         ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1131         rt->rt6i_src.plen = cfg->fc_src_len;
1132 #endif
1133
1134         rt->rt6i_metric = cfg->fc_metric;
1135
1136         /* We cannot add true routes via loopback here,
1137            they would result in kernel looping; promote them to reject routes
1138          */
1139         if ((cfg->fc_flags & RTF_REJECT) ||
1140             (dev && (dev->flags&IFF_LOOPBACK) && !(addr_type&IPV6_ADDR_LOOPBACK))) {
1141                 /* hold loopback dev/idev if we haven't done so. */
1142                 if (dev != net->loopback_dev) {
1143                         if (dev) {
1144                                 dev_put(dev);
1145                                 in6_dev_put(idev);
1146                         }
1147                         dev = net->loopback_dev;
1148                         dev_hold(dev);
1149                         idev = in6_dev_get(dev);
1150                         if (!idev) {
1151                                 err = -ENODEV;
1152                                 goto out;
1153                         }
1154                 }
1155                 rt->u.dst.output = ip6_pkt_discard_out;
1156                 rt->u.dst.input = ip6_pkt_discard;
1157                 rt->u.dst.error = -ENETUNREACH;
1158                 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1159                 goto install_route;
1160         }
1161
1162         if (cfg->fc_flags & RTF_GATEWAY) {
1163                 struct in6_addr *gw_addr;
1164                 int gwa_type;
1165
1166                 gw_addr = &cfg->fc_gateway;
1167                 ipv6_addr_copy(&rt->rt6i_gateway, gw_addr);
1168                 gwa_type = ipv6_addr_type(gw_addr);
1169
1170                 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1171                         struct rt6_info *grt;
1172
1173                         /* IPv6 strictly inhibits using not link-local
1174                            addresses as nexthop address.
1175                            Otherwise, router will not able to send redirects.
1176                            It is very good, but in some (rare!) circumstances
1177                            (SIT, PtP, NBMA NOARP links) it is handy to allow
1178                            some exceptions. --ANK
1179                          */
1180                         err = -EINVAL;
1181                         if (!(gwa_type&IPV6_ADDR_UNICAST))
1182                                 goto out;
1183
1184                         grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, 1);
1185
1186                         err = -EHOSTUNREACH;
1187                         if (grt == NULL)
1188                                 goto out;
1189                         if (dev) {
1190                                 if (dev != grt->rt6i_dev) {
1191                                         dst_release(&grt->u.dst);
1192                                         goto out;
1193                                 }
1194                         } else {
1195                                 dev = grt->rt6i_dev;
1196                                 idev = grt->rt6i_idev;
1197                                 dev_hold(dev);
1198                                 in6_dev_hold(grt->rt6i_idev);
1199                         }
1200                         if (!(grt->rt6i_flags&RTF_GATEWAY))
1201                                 err = 0;
1202                         dst_release(&grt->u.dst);
1203
1204                         if (err)
1205                                 goto out;
1206                 }
1207                 err = -EINVAL;
1208                 if (dev == NULL || (dev->flags&IFF_LOOPBACK))
1209                         goto out;
1210         }
1211
1212         err = -ENODEV;
1213         if (dev == NULL)
1214                 goto out;
1215
1216         if (cfg->fc_flags & (RTF_GATEWAY | RTF_NONEXTHOP)) {
1217                 rt->rt6i_nexthop = __neigh_lookup_errno(&nd_tbl, &rt->rt6i_gateway, dev);
1218                 if (IS_ERR(rt->rt6i_nexthop)) {
1219                         err = PTR_ERR(rt->rt6i_nexthop);
1220                         rt->rt6i_nexthop = NULL;
1221                         goto out;
1222                 }
1223         }
1224
1225         rt->rt6i_flags = cfg->fc_flags;
1226
1227 install_route:
1228         if (cfg->fc_mx) {
1229                 struct nlattr *nla;
1230                 int remaining;
1231
1232                 nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1233                         int type = nla_type(nla);
1234
1235                         if (type) {
1236                                 if (type > RTAX_MAX) {
1237                                         err = -EINVAL;
1238                                         goto out;
1239                                 }
1240
1241                                 rt->u.dst.metrics[type - 1] = nla_get_u32(nla);
1242                         }
1243                 }
1244         }
1245
1246         if (rt->u.dst.metrics[RTAX_HOPLIMIT-1] == 0)
1247                 rt->u.dst.metrics[RTAX_HOPLIMIT-1] = -1;
1248         if (!rt->u.dst.metrics[RTAX_MTU-1])
1249                 rt->u.dst.metrics[RTAX_MTU-1] = ipv6_get_mtu(dev);
1250         if (!rt->u.dst.metrics[RTAX_ADVMSS-1])
1251                 rt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(net, dst_mtu(&rt->u.dst));
1252         rt->u.dst.dev = dev;
1253         rt->rt6i_idev = idev;
1254         rt->rt6i_table = table;
1255
1256         cfg->fc_nlinfo.nl_net = dev_net(dev);
1257
1258         return __ip6_ins_rt(rt, &cfg->fc_nlinfo);
1259
1260 out:
1261         if (dev)
1262                 dev_put(dev);
1263         if (idev)
1264                 in6_dev_put(idev);
1265         if (rt)
1266                 dst_free(&rt->u.dst);
1267         return err;
1268 }
1269
1270 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
1271 {
1272         int err;
1273         struct fib6_table *table;
1274         struct net *net = dev_net(rt->rt6i_dev);
1275
1276         if (rt == net->ipv6.ip6_null_entry)
1277                 return -ENOENT;
1278
1279         table = rt->rt6i_table;
1280         write_lock_bh(&table->tb6_lock);
1281
1282         err = fib6_del(rt, info);
1283         dst_release(&rt->u.dst);
1284
1285         write_unlock_bh(&table->tb6_lock);
1286
1287         return err;
1288 }
1289
1290 int ip6_del_rt(struct rt6_info *rt)
1291 {
1292         struct nl_info info = {
1293                 .nl_net = dev_net(rt->rt6i_dev),
1294         };
1295         return __ip6_del_rt(rt, &info);
1296 }
1297
1298 static int ip6_route_del(struct fib6_config *cfg)
1299 {
1300         struct fib6_table *table;
1301         struct fib6_node *fn;
1302         struct rt6_info *rt;
1303         int err = -ESRCH;
1304
1305         table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
1306         if (table == NULL)
1307                 return err;
1308
1309         read_lock_bh(&table->tb6_lock);
1310
1311         fn = fib6_locate(&table->tb6_root,
1312                          &cfg->fc_dst, cfg->fc_dst_len,
1313                          &cfg->fc_src, cfg->fc_src_len);
1314
1315         if (fn) {
1316                 for (rt = fn->leaf; rt; rt = rt->u.dst.rt6_next) {
1317                         if (cfg->fc_ifindex &&
1318                             (rt->rt6i_dev == NULL ||
1319                              rt->rt6i_dev->ifindex != cfg->fc_ifindex))
1320                                 continue;
1321                         if (cfg->fc_flags & RTF_GATEWAY &&
1322                             !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
1323                                 continue;
1324                         if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
1325                                 continue;
1326                         dst_hold(&rt->u.dst);
1327                         read_unlock_bh(&table->tb6_lock);
1328
1329                         return __ip6_del_rt(rt, &cfg->fc_nlinfo);
1330                 }
1331         }
1332         read_unlock_bh(&table->tb6_lock);
1333
1334         return err;
1335 }
1336
1337 /*
1338  *      Handle redirects
1339  */
1340 struct ip6rd_flowi {
1341         struct flowi fl;
1342         struct in6_addr gateway;
1343 };
1344
1345 static struct rt6_info *__ip6_route_redirect(struct net *net,
1346                                              struct fib6_table *table,
1347                                              struct flowi *fl,
1348                                              int flags)
1349 {
1350         struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl;
1351         struct rt6_info *rt;
1352         struct fib6_node *fn;
1353
1354         /*
1355          * Get the "current" route for this destination and
1356          * check if the redirect has come from approriate router.
1357          *
1358          * RFC 2461 specifies that redirects should only be
1359          * accepted if they come from the nexthop to the target.
1360          * Due to the way the routes are chosen, this notion
1361          * is a bit fuzzy and one might need to check all possible
1362          * routes.
1363          */
1364
1365         read_lock_bh(&table->tb6_lock);
1366         fn = fib6_lookup(&table->tb6_root, &fl->fl6_dst, &fl->fl6_src);
1367 restart:
1368         for (rt = fn->leaf; rt; rt = rt->u.dst.rt6_next) {
1369                 /*
1370                  * Current route is on-link; redirect is always invalid.
1371                  *
1372                  * Seems, previous statement is not true. It could
1373                  * be node, which looks for us as on-link (f.e. proxy ndisc)
1374                  * But then router serving it might decide, that we should
1375                  * know truth 8)8) --ANK (980726).
1376                  */
1377                 if (rt6_check_expired(rt))
1378                         continue;
1379                 if (!(rt->rt6i_flags & RTF_GATEWAY))
1380                         continue;
1381                 if (fl->oif != rt->rt6i_dev->ifindex)
1382                         continue;
1383                 if (!ipv6_addr_equal(&rdfl->gateway, &rt->rt6i_gateway))
1384                         continue;
1385                 break;
1386         }
1387
1388         if (!rt)
1389                 rt = net->ipv6.ip6_null_entry;
1390         BACKTRACK(net, &fl->fl6_src);
1391 out:
1392         dst_hold(&rt->u.dst);
1393
1394         read_unlock_bh(&table->tb6_lock);
1395
1396         return rt;
1397 };
1398
1399 static struct rt6_info *ip6_route_redirect(struct in6_addr *dest,
1400                                            struct in6_addr *src,
1401                                            struct in6_addr *gateway,
1402                                            struct net_device *dev)
1403 {
1404         int flags = RT6_LOOKUP_F_HAS_SADDR;
1405         struct net *net = dev_net(dev);
1406         struct ip6rd_flowi rdfl = {
1407                 .fl = {
1408                         .oif = dev->ifindex,
1409                         .nl_u = {
1410                                 .ip6_u = {
1411                                         .daddr = *dest,
1412                                         .saddr = *src,
1413                                 },
1414                         },
1415                 },
1416                 .gateway = *gateway,
1417         };
1418
1419         if (rt6_need_strict(dest))
1420                 flags |= RT6_LOOKUP_F_IFACE;
1421
1422         return (struct rt6_info *)fib6_rule_lookup(net, (struct flowi *)&rdfl,
1423                                                    flags, __ip6_route_redirect);
1424 }
1425
1426 void rt6_redirect(struct in6_addr *dest, struct in6_addr *src,
1427                   struct in6_addr *saddr,
1428                   struct neighbour *neigh, u8 *lladdr, int on_link)
1429 {
1430         struct rt6_info *rt, *nrt = NULL;
1431         struct netevent_redirect netevent;
1432         struct net *net = dev_net(neigh->dev);
1433
1434         rt = ip6_route_redirect(dest, src, saddr, neigh->dev);
1435
1436         if (rt == net->ipv6.ip6_null_entry) {
1437                 if (net_ratelimit())
1438                         printk(KERN_DEBUG "rt6_redirect: source isn't a valid nexthop "
1439                                "for redirect target\n");
1440                 goto out;
1441         }
1442
1443         /*
1444          *      We have finally decided to accept it.
1445          */
1446
1447         neigh_update(neigh, lladdr, NUD_STALE,
1448                      NEIGH_UPDATE_F_WEAK_OVERRIDE|
1449                      NEIGH_UPDATE_F_OVERRIDE|
1450                      (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1451                                      NEIGH_UPDATE_F_ISROUTER))
1452                      );
1453
1454         /*
1455          * Redirect received -> path was valid.
1456          * Look, redirects are sent only in response to data packets,
1457          * so that this nexthop apparently is reachable. --ANK
1458          */
1459         dst_confirm(&rt->u.dst);
1460
1461         /* Duplicate redirect: silently ignore. */
1462         if (neigh == rt->u.dst.neighbour)
1463                 goto out;
1464
1465         nrt = ip6_rt_copy(rt);
1466         if (nrt == NULL)
1467                 goto out;
1468
1469         nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
1470         if (on_link)
1471                 nrt->rt6i_flags &= ~RTF_GATEWAY;
1472
1473         ipv6_addr_copy(&nrt->rt6i_dst.addr, dest);
1474         nrt->rt6i_dst.plen = 128;
1475         nrt->u.dst.flags |= DST_HOST;
1476
1477         ipv6_addr_copy(&nrt->rt6i_gateway, (struct in6_addr*)neigh->primary_key);
1478         nrt->rt6i_nexthop = neigh_clone(neigh);
1479         /* Reset pmtu, it may be better */
1480         nrt->u.dst.metrics[RTAX_MTU-1] = ipv6_get_mtu(neigh->dev);
1481         nrt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(dev_net(neigh->dev),
1482                                                         dst_mtu(&nrt->u.dst));
1483
1484         if (ip6_ins_rt(nrt))
1485                 goto out;
1486
1487         netevent.old = &rt->u.dst;
1488         netevent.new = &nrt->u.dst;
1489         call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
1490
1491         if (rt->rt6i_flags&RTF_CACHE) {
1492                 ip6_del_rt(rt);
1493                 return;
1494         }
1495
1496 out:
1497         dst_release(&rt->u.dst);
1498         return;
1499 }
1500
1501 /*
1502  *      Handle ICMP "packet too big" messages
1503  *      i.e. Path MTU discovery
1504  */
1505
1506 void rt6_pmtu_discovery(struct in6_addr *daddr, struct in6_addr *saddr,
1507                         struct net_device *dev, u32 pmtu)
1508 {
1509         struct rt6_info *rt, *nrt;
1510         struct net *net = dev_net(dev);
1511         int allfrag = 0;
1512
1513         rt = rt6_lookup(net, daddr, saddr, dev->ifindex, 0);
1514         if (rt == NULL)
1515                 return;
1516
1517         if (pmtu >= dst_mtu(&rt->u.dst))
1518                 goto out;
1519
1520         if (pmtu < IPV6_MIN_MTU) {
1521                 /*
1522                  * According to RFC2460, PMTU is set to the IPv6 Minimum Link
1523                  * MTU (1280) and a fragment header should always be included
1524                  * after a node receiving Too Big message reporting PMTU is
1525                  * less than the IPv6 Minimum Link MTU.
1526                  */
1527                 pmtu = IPV6_MIN_MTU;
1528                 allfrag = 1;
1529         }
1530
1531         /* New mtu received -> path was valid.
1532            They are sent only in response to data packets,
1533            so that this nexthop apparently is reachable. --ANK
1534          */
1535         dst_confirm(&rt->u.dst);
1536
1537         /* Host route. If it is static, it would be better
1538            not to override it, but add new one, so that
1539            when cache entry will expire old pmtu
1540            would return automatically.
1541          */
1542         if (rt->rt6i_flags & RTF_CACHE) {
1543                 rt->u.dst.metrics[RTAX_MTU-1] = pmtu;
1544                 if (allfrag)
1545                         rt->u.dst.metrics[RTAX_FEATURES-1] |= RTAX_FEATURE_ALLFRAG;
1546                 dst_set_expires(&rt->u.dst, net->ipv6.sysctl.ip6_rt_mtu_expires);
1547                 rt->rt6i_flags |= RTF_MODIFIED|RTF_EXPIRES;
1548                 goto out;
1549         }
1550
1551         /* Network route.
1552            Two cases are possible:
1553            1. It is connected route. Action: COW
1554            2. It is gatewayed route or NONEXTHOP route. Action: clone it.
1555          */
1556         if (!rt->rt6i_nexthop && !(rt->rt6i_flags & RTF_NONEXTHOP))
1557                 nrt = rt6_alloc_cow(rt, daddr, saddr);
1558         else
1559                 nrt = rt6_alloc_clone(rt, daddr);
1560
1561         if (nrt) {
1562                 nrt->u.dst.metrics[RTAX_MTU-1] = pmtu;
1563                 if (allfrag)
1564                         nrt->u.dst.metrics[RTAX_FEATURES-1] |= RTAX_FEATURE_ALLFRAG;
1565
1566                 /* According to RFC 1981, detecting PMTU increase shouldn't be
1567                  * happened within 5 mins, the recommended timer is 10 mins.
1568                  * Here this route expiration time is set to ip6_rt_mtu_expires
1569                  * which is 10 mins. After 10 mins the decreased pmtu is expired
1570                  * and detecting PMTU increase will be automatically happened.
1571                  */
1572                 dst_set_expires(&nrt->u.dst, net->ipv6.sysctl.ip6_rt_mtu_expires);
1573                 nrt->rt6i_flags |= RTF_DYNAMIC|RTF_EXPIRES;
1574
1575                 ip6_ins_rt(nrt);
1576         }
1577 out:
1578         dst_release(&rt->u.dst);
1579 }
1580
1581 /*
1582  *      Misc support functions
1583  */
1584
1585 static struct rt6_info * ip6_rt_copy(struct rt6_info *ort)
1586 {
1587         struct net *net = dev_net(ort->rt6i_dev);
1588         struct rt6_info *rt = ip6_dst_alloc(net->ipv6.ip6_dst_ops);
1589
1590         if (rt) {
1591                 rt->u.dst.input = ort->u.dst.input;
1592                 rt->u.dst.output = ort->u.dst.output;
1593
1594                 memcpy(rt->u.dst.metrics, ort->u.dst.metrics, RTAX_MAX*sizeof(u32));
1595                 rt->u.dst.error = ort->u.dst.error;
1596                 rt->u.dst.dev = ort->u.dst.dev;
1597                 if (rt->u.dst.dev)
1598                         dev_hold(rt->u.dst.dev);
1599                 rt->rt6i_idev = ort->rt6i_idev;
1600                 if (rt->rt6i_idev)
1601                         in6_dev_hold(rt->rt6i_idev);
1602                 rt->u.dst.lastuse = jiffies;
1603                 rt->rt6i_expires = 0;
1604
1605                 ipv6_addr_copy(&rt->rt6i_gateway, &ort->rt6i_gateway);
1606                 rt->rt6i_flags = ort->rt6i_flags & ~RTF_EXPIRES;
1607                 rt->rt6i_metric = 0;
1608
1609                 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1610 #ifdef CONFIG_IPV6_SUBTREES
1611                 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1612 #endif
1613                 rt->rt6i_table = ort->rt6i_table;
1614         }
1615         return rt;
1616 }
1617
1618 #ifdef CONFIG_IPV6_ROUTE_INFO
1619 static struct rt6_info *rt6_get_route_info(struct net *net,
1620                                            struct in6_addr *prefix, int prefixlen,
1621                                            struct in6_addr *gwaddr, int ifindex)
1622 {
1623         struct fib6_node *fn;
1624         struct rt6_info *rt = NULL;
1625         struct fib6_table *table;
1626
1627         table = fib6_get_table(net, RT6_TABLE_INFO);
1628         if (table == NULL)
1629                 return NULL;
1630
1631         write_lock_bh(&table->tb6_lock);
1632         fn = fib6_locate(&table->tb6_root, prefix ,prefixlen, NULL, 0);
1633         if (!fn)
1634                 goto out;
1635
1636         for (rt = fn->leaf; rt; rt = rt->u.dst.rt6_next) {
1637                 if (rt->rt6i_dev->ifindex != ifindex)
1638                         continue;
1639                 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
1640                         continue;
1641                 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
1642                         continue;
1643                 dst_hold(&rt->u.dst);
1644                 break;
1645         }
1646 out:
1647         write_unlock_bh(&table->tb6_lock);
1648         return rt;
1649 }
1650
1651 static struct rt6_info *rt6_add_route_info(struct net *net,
1652                                            struct in6_addr *prefix, int prefixlen,
1653                                            struct in6_addr *gwaddr, int ifindex,
1654                                            unsigned pref)
1655 {
1656         struct fib6_config cfg = {
1657                 .fc_table       = RT6_TABLE_INFO,
1658                 .fc_metric      = IP6_RT_PRIO_USER,
1659                 .fc_ifindex     = ifindex,
1660                 .fc_dst_len     = prefixlen,
1661                 .fc_flags       = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
1662                                   RTF_UP | RTF_PREF(pref),
1663                 .fc_nlinfo.pid = 0,
1664                 .fc_nlinfo.nlh = NULL,
1665                 .fc_nlinfo.nl_net = net,
1666         };
1667
1668         ipv6_addr_copy(&cfg.fc_dst, prefix);
1669         ipv6_addr_copy(&cfg.fc_gateway, gwaddr);
1670
1671         /* We should treat it as a default route if prefix length is 0. */
1672         if (!prefixlen)
1673                 cfg.fc_flags |= RTF_DEFAULT;
1674
1675         ip6_route_add(&cfg);
1676
1677         return rt6_get_route_info(net, prefix, prefixlen, gwaddr, ifindex);
1678 }
1679 #endif
1680
1681 struct rt6_info *rt6_get_dflt_router(struct in6_addr *addr, struct net_device *dev)
1682 {
1683         struct rt6_info *rt;
1684         struct fib6_table *table;
1685
1686         table = fib6_get_table(dev_net(dev), RT6_TABLE_DFLT);
1687         if (table == NULL)
1688                 return NULL;
1689
1690         write_lock_bh(&table->tb6_lock);
1691         for (rt = table->tb6_root.leaf; rt; rt=rt->u.dst.rt6_next) {
1692                 if (dev == rt->rt6i_dev &&
1693                     ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
1694                     ipv6_addr_equal(&rt->rt6i_gateway, addr))
1695                         break;
1696         }
1697         if (rt)
1698                 dst_hold(&rt->u.dst);
1699         write_unlock_bh(&table->tb6_lock);
1700         return rt;
1701 }
1702
1703 struct rt6_info *rt6_add_dflt_router(struct in6_addr *gwaddr,
1704                                      struct net_device *dev,
1705                                      unsigned int pref)
1706 {
1707         struct fib6_config cfg = {
1708                 .fc_table       = RT6_TABLE_DFLT,
1709                 .fc_metric      = IP6_RT_PRIO_USER,
1710                 .fc_ifindex     = dev->ifindex,
1711                 .fc_flags       = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
1712                                   RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
1713                 .fc_nlinfo.pid = 0,
1714                 .fc_nlinfo.nlh = NULL,
1715                 .fc_nlinfo.nl_net = dev_net(dev),
1716         };
1717
1718         ipv6_addr_copy(&cfg.fc_gateway, gwaddr);
1719
1720         ip6_route_add(&cfg);
1721
1722         return rt6_get_dflt_router(gwaddr, dev);
1723 }
1724
1725 void rt6_purge_dflt_routers(struct net *net)
1726 {
1727         struct rt6_info *rt;
1728         struct fib6_table *table;
1729
1730         /* NOTE: Keep consistent with rt6_get_dflt_router */
1731         table = fib6_get_table(net, RT6_TABLE_DFLT);
1732         if (table == NULL)
1733                 return;
1734
1735 restart:
1736         read_lock_bh(&table->tb6_lock);
1737         for (rt = table->tb6_root.leaf; rt; rt = rt->u.dst.rt6_next) {
1738                 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF)) {
1739                         dst_hold(&rt->u.dst);
1740                         read_unlock_bh(&table->tb6_lock);
1741                         ip6_del_rt(rt);
1742                         goto restart;
1743                 }
1744         }
1745         read_unlock_bh(&table->tb6_lock);
1746 }
1747
1748 static void rtmsg_to_fib6_config(struct net *net,
1749                                  struct in6_rtmsg *rtmsg,
1750                                  struct fib6_config *cfg)
1751 {
1752         memset(cfg, 0, sizeof(*cfg));
1753
1754         cfg->fc_table = RT6_TABLE_MAIN;
1755         cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
1756         cfg->fc_metric = rtmsg->rtmsg_metric;
1757         cfg->fc_expires = rtmsg->rtmsg_info;
1758         cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
1759         cfg->fc_src_len = rtmsg->rtmsg_src_len;
1760         cfg->fc_flags = rtmsg->rtmsg_flags;
1761
1762         cfg->fc_nlinfo.nl_net = net;
1763
1764         ipv6_addr_copy(&cfg->fc_dst, &rtmsg->rtmsg_dst);
1765         ipv6_addr_copy(&cfg->fc_src, &rtmsg->rtmsg_src);
1766         ipv6_addr_copy(&cfg->fc_gateway, &rtmsg->rtmsg_gateway);
1767 }
1768
1769 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
1770 {
1771         struct fib6_config cfg;
1772         struct in6_rtmsg rtmsg;
1773         int err;
1774
1775         switch(cmd) {
1776         case SIOCADDRT:         /* Add a route */
1777         case SIOCDELRT:         /* Delete a route */
1778                 if (!capable(CAP_NET_ADMIN))
1779                         return -EPERM;
1780                 err = copy_from_user(&rtmsg, arg,
1781                                      sizeof(struct in6_rtmsg));
1782                 if (err)
1783                         return -EFAULT;
1784
1785                 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
1786
1787                 rtnl_lock();
1788                 switch (cmd) {
1789                 case SIOCADDRT:
1790                         err = ip6_route_add(&cfg);
1791                         break;
1792                 case SIOCDELRT:
1793                         err = ip6_route_del(&cfg);
1794                         break;
1795                 default:
1796                         err = -EINVAL;
1797                 }
1798                 rtnl_unlock();
1799
1800                 return err;
1801         }
1802
1803         return -EINVAL;
1804 }
1805
1806 /*
1807  *      Drop the packet on the floor
1808  */
1809
1810 static int ip6_pkt_drop(struct sk_buff *skb, int code, int ipstats_mib_noroutes)
1811 {
1812         int type;
1813         switch (ipstats_mib_noroutes) {
1814         case IPSTATS_MIB_INNOROUTES:
1815                 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
1816                 if (type == IPV6_ADDR_ANY || type == IPV6_ADDR_RESERVED) {
1817                         IP6_INC_STATS(ip6_dst_idev(skb->dst), IPSTATS_MIB_INADDRERRORS);
1818                         break;
1819                 }
1820                 /* FALLTHROUGH */
1821         case IPSTATS_MIB_OUTNOROUTES:
1822                 IP6_INC_STATS(ip6_dst_idev(skb->dst), ipstats_mib_noroutes);
1823                 break;
1824         }
1825         icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0, skb->dev);
1826         kfree_skb(skb);
1827         return 0;
1828 }
1829
1830 static int ip6_pkt_discard(struct sk_buff *skb)
1831 {
1832         return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
1833 }
1834
1835 static int ip6_pkt_discard_out(struct sk_buff *skb)
1836 {
1837         skb->dev = skb->dst->dev;
1838         return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
1839 }
1840
1841 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
1842
1843 static int ip6_pkt_prohibit(struct sk_buff *skb)
1844 {
1845         return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
1846 }
1847
1848 static int ip6_pkt_prohibit_out(struct sk_buff *skb)
1849 {
1850         skb->dev = skb->dst->dev;
1851         return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
1852 }
1853
1854 #endif
1855
1856 /*
1857  *      Allocate a dst for local (unicast / anycast) address.
1858  */
1859
1860 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
1861                                     const struct in6_addr *addr,
1862                                     int anycast)
1863 {
1864         struct net *net = dev_net(idev->dev);
1865         struct rt6_info *rt = ip6_dst_alloc(net->ipv6.ip6_dst_ops);
1866
1867         if (rt == NULL)
1868                 return ERR_PTR(-ENOMEM);
1869
1870         dev_hold(net->loopback_dev);
1871         in6_dev_hold(idev);
1872
1873         rt->u.dst.flags = DST_HOST;
1874         rt->u.dst.input = ip6_input;
1875         rt->u.dst.output = ip6_output;
1876         rt->rt6i_dev = net->loopback_dev;
1877         rt->rt6i_idev = idev;
1878         rt->u.dst.metrics[RTAX_MTU-1] = ipv6_get_mtu(rt->rt6i_dev);
1879         rt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(net, dst_mtu(&rt->u.dst));
1880         rt->u.dst.metrics[RTAX_HOPLIMIT-1] = -1;
1881         rt->u.dst.obsolete = -1;
1882
1883         rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
1884         if (anycast)
1885                 rt->rt6i_flags |= RTF_ANYCAST;
1886         else
1887                 rt->rt6i_flags |= RTF_LOCAL;
1888         rt->rt6i_nexthop = ndisc_get_neigh(rt->rt6i_dev, &rt->rt6i_gateway);
1889         if (rt->rt6i_nexthop == NULL) {
1890                 dst_free(&rt->u.dst);
1891                 return ERR_PTR(-ENOMEM);
1892         }
1893
1894         ipv6_addr_copy(&rt->rt6i_dst.addr, addr);
1895         rt->rt6i_dst.plen = 128;
1896         rt->rt6i_table = fib6_get_table(net, RT6_TABLE_LOCAL);
1897
1898         atomic_set(&rt->u.dst.__refcnt, 1);
1899
1900         return rt;
1901 }
1902
1903 struct arg_dev_net {
1904         struct net_device *dev;
1905         struct net *net;
1906 };
1907
1908 static int fib6_ifdown(struct rt6_info *rt, void *arg)
1909 {
1910         struct net_device *dev = ((struct arg_dev_net *)arg)->dev;
1911         struct net *net = ((struct arg_dev_net *)arg)->net;
1912
1913         if (((void *)rt->rt6i_dev == dev || dev == NULL) &&
1914             rt != net->ipv6.ip6_null_entry) {
1915                 RT6_TRACE("deleted by ifdown %p\n", rt);
1916                 return -1;
1917         }
1918         return 0;
1919 }
1920
1921 void rt6_ifdown(struct net *net, struct net_device *dev)
1922 {
1923         struct arg_dev_net adn = {
1924                 .dev = dev,
1925                 .net = net,
1926         };
1927
1928         fib6_clean_all(net, fib6_ifdown, 0, &adn);
1929 }
1930
1931 struct rt6_mtu_change_arg
1932 {
1933         struct net_device *dev;
1934         unsigned mtu;
1935 };
1936
1937 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
1938 {
1939         struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
1940         struct inet6_dev *idev;
1941         struct net *net = dev_net(arg->dev);
1942
1943         /* In IPv6 pmtu discovery is not optional,
1944            so that RTAX_MTU lock cannot disable it.
1945            We still use this lock to block changes
1946            caused by addrconf/ndisc.
1947         */
1948
1949         idev = __in6_dev_get(arg->dev);
1950         if (idev == NULL)
1951                 return 0;
1952
1953         /* For administrative MTU increase, there is no way to discover
1954            IPv6 PMTU increase, so PMTU increase should be updated here.
1955            Since RFC 1981 doesn't include administrative MTU increase
1956            update PMTU increase is a MUST. (i.e. jumbo frame)
1957          */
1958         /*
1959            If new MTU is less than route PMTU, this new MTU will be the
1960            lowest MTU in the path, update the route PMTU to reflect PMTU
1961            decreases; if new MTU is greater than route PMTU, and the
1962            old MTU is the lowest MTU in the path, update the route PMTU
1963            to reflect the increase. In this case if the other nodes' MTU
1964            also have the lowest MTU, TOO BIG MESSAGE will be lead to
1965            PMTU discouvery.
1966          */
1967         if (rt->rt6i_dev == arg->dev &&
1968             !dst_metric_locked(&rt->u.dst, RTAX_MTU) &&
1969             (dst_mtu(&rt->u.dst) >= arg->mtu ||
1970              (dst_mtu(&rt->u.dst) < arg->mtu &&
1971               dst_mtu(&rt->u.dst) == idev->cnf.mtu6))) {
1972                 rt->u.dst.metrics[RTAX_MTU-1] = arg->mtu;
1973                 rt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(net, arg->mtu);
1974         }
1975         return 0;
1976 }
1977
1978 void rt6_mtu_change(struct net_device *dev, unsigned mtu)
1979 {
1980         struct rt6_mtu_change_arg arg = {
1981                 .dev = dev,
1982                 .mtu = mtu,
1983         };
1984
1985         fib6_clean_all(dev_net(dev), rt6_mtu_change_route, 0, &arg);
1986 }
1987
1988 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
1989         [RTA_GATEWAY]           = { .len = sizeof(struct in6_addr) },
1990         [RTA_OIF]               = { .type = NLA_U32 },
1991         [RTA_IIF]               = { .type = NLA_U32 },
1992         [RTA_PRIORITY]          = { .type = NLA_U32 },
1993         [RTA_METRICS]           = { .type = NLA_NESTED },
1994 };
1995
1996 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
1997                               struct fib6_config *cfg)
1998 {
1999         struct rtmsg *rtm;
2000         struct nlattr *tb[RTA_MAX+1];
2001         int err;
2002
2003         err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2004         if (err < 0)
2005                 goto errout;
2006
2007         err = -EINVAL;
2008         rtm = nlmsg_data(nlh);
2009         memset(cfg, 0, sizeof(*cfg));
2010
2011         cfg->fc_table = rtm->rtm_table;
2012         cfg->fc_dst_len = rtm->rtm_dst_len;
2013         cfg->fc_src_len = rtm->rtm_src_len;
2014         cfg->fc_flags = RTF_UP;
2015         cfg->fc_protocol = rtm->rtm_protocol;
2016
2017         if (rtm->rtm_type == RTN_UNREACHABLE)
2018                 cfg->fc_flags |= RTF_REJECT;
2019
2020         cfg->fc_nlinfo.pid = NETLINK_CB(skb).pid;
2021         cfg->fc_nlinfo.nlh = nlh;
2022         cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2023
2024         if (tb[RTA_GATEWAY]) {
2025                 nla_memcpy(&cfg->fc_gateway, tb[RTA_GATEWAY], 16);
2026                 cfg->fc_flags |= RTF_GATEWAY;
2027         }
2028
2029         if (tb[RTA_DST]) {
2030                 int plen = (rtm->rtm_dst_len + 7) >> 3;
2031
2032                 if (nla_len(tb[RTA_DST]) < plen)
2033                         goto errout;
2034
2035                 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2036         }
2037
2038         if (tb[RTA_SRC]) {
2039                 int plen = (rtm->rtm_src_len + 7) >> 3;
2040
2041                 if (nla_len(tb[RTA_SRC]) < plen)
2042                         goto errout;
2043
2044                 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2045         }
2046
2047         if (tb[RTA_OIF])
2048                 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2049
2050         if (tb[RTA_PRIORITY])
2051                 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2052
2053         if (tb[RTA_METRICS]) {
2054                 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2055                 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2056         }
2057
2058         if (tb[RTA_TABLE])
2059                 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2060
2061         err = 0;
2062 errout:
2063         return err;
2064 }
2065
2066 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
2067 {
2068         struct fib6_config cfg;
2069         int err;
2070
2071         err = rtm_to_fib6_config(skb, nlh, &cfg);
2072         if (err < 0)
2073                 return err;
2074
2075         return ip6_route_del(&cfg);
2076 }
2077
2078 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
2079 {
2080         struct fib6_config cfg;
2081         int err;
2082
2083         err = rtm_to_fib6_config(skb, nlh, &cfg);
2084         if (err < 0)
2085                 return err;
2086
2087         return ip6_route_add(&cfg);
2088 }
2089
2090 static inline size_t rt6_nlmsg_size(void)
2091 {
2092         return NLMSG_ALIGN(sizeof(struct rtmsg))
2093                + nla_total_size(16) /* RTA_SRC */
2094                + nla_total_size(16) /* RTA_DST */
2095                + nla_total_size(16) /* RTA_GATEWAY */
2096                + nla_total_size(16) /* RTA_PREFSRC */
2097                + nla_total_size(4) /* RTA_TABLE */
2098                + nla_total_size(4) /* RTA_IIF */
2099                + nla_total_size(4) /* RTA_OIF */
2100                + nla_total_size(4) /* RTA_PRIORITY */
2101                + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
2102                + nla_total_size(sizeof(struct rta_cacheinfo));
2103 }
2104
2105 static int rt6_fill_node(struct sk_buff *skb, struct rt6_info *rt,
2106                          struct in6_addr *dst, struct in6_addr *src,
2107                          int iif, int type, u32 pid, u32 seq,
2108                          int prefix, int nowait, unsigned int flags)
2109 {
2110         struct rtmsg *rtm;
2111         struct nlmsghdr *nlh;
2112         long expires;
2113         u32 table;
2114
2115         if (prefix) {   /* user wants prefix routes only */
2116                 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
2117                         /* success since this is not a prefix route */
2118                         return 1;
2119                 }
2120         }
2121
2122         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*rtm), flags);
2123         if (nlh == NULL)
2124                 return -EMSGSIZE;
2125
2126         rtm = nlmsg_data(nlh);
2127         rtm->rtm_family = AF_INET6;
2128         rtm->rtm_dst_len = rt->rt6i_dst.plen;
2129         rtm->rtm_src_len = rt->rt6i_src.plen;
2130         rtm->rtm_tos = 0;
2131         if (rt->rt6i_table)
2132                 table = rt->rt6i_table->tb6_id;
2133         else
2134                 table = RT6_TABLE_UNSPEC;
2135         rtm->rtm_table = table;
2136         NLA_PUT_U32(skb, RTA_TABLE, table);
2137         if (rt->rt6i_flags&RTF_REJECT)
2138                 rtm->rtm_type = RTN_UNREACHABLE;
2139         else if (rt->rt6i_dev && (rt->rt6i_dev->flags&IFF_LOOPBACK))
2140                 rtm->rtm_type = RTN_LOCAL;
2141         else
2142                 rtm->rtm_type = RTN_UNICAST;
2143         rtm->rtm_flags = 0;
2144         rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2145         rtm->rtm_protocol = rt->rt6i_protocol;
2146         if (rt->rt6i_flags&RTF_DYNAMIC)
2147                 rtm->rtm_protocol = RTPROT_REDIRECT;
2148         else if (rt->rt6i_flags & RTF_ADDRCONF)
2149                 rtm->rtm_protocol = RTPROT_KERNEL;
2150         else if (rt->rt6i_flags&RTF_DEFAULT)
2151                 rtm->rtm_protocol = RTPROT_RA;
2152
2153         if (rt->rt6i_flags&RTF_CACHE)
2154                 rtm->rtm_flags |= RTM_F_CLONED;
2155
2156         if (dst) {
2157                 NLA_PUT(skb, RTA_DST, 16, dst);
2158                 rtm->rtm_dst_len = 128;
2159         } else if (rtm->rtm_dst_len)
2160                 NLA_PUT(skb, RTA_DST, 16, &rt->rt6i_dst.addr);
2161 #ifdef CONFIG_IPV6_SUBTREES
2162         if (src) {
2163                 NLA_PUT(skb, RTA_SRC, 16, src);
2164                 rtm->rtm_src_len = 128;
2165         } else if (rtm->rtm_src_len)
2166                 NLA_PUT(skb, RTA_SRC, 16, &rt->rt6i_src.addr);
2167 #endif
2168         if (iif) {
2169 #ifdef CONFIG_IPV6_MROUTE
2170                 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
2171                         int err = ip6mr_get_route(skb, rtm, nowait);
2172                         if (err <= 0) {
2173                                 if (!nowait) {
2174                                         if (err == 0)
2175                                                 return 0;
2176                                         goto nla_put_failure;
2177                                 } else {
2178                                         if (err == -EMSGSIZE)
2179                                                 goto nla_put_failure;
2180                                 }
2181                         }
2182                 } else
2183 #endif
2184                         NLA_PUT_U32(skb, RTA_IIF, iif);
2185         } else if (dst) {
2186                 struct in6_addr saddr_buf;
2187                 if (ipv6_dev_get_saddr(ip6_dst_idev(&rt->u.dst)->dev,
2188                                        dst, 0, &saddr_buf) == 0)
2189                         NLA_PUT(skb, RTA_PREFSRC, 16, &saddr_buf);
2190         }
2191
2192         if (rtnetlink_put_metrics(skb, rt->u.dst.metrics) < 0)
2193                 goto nla_put_failure;
2194
2195         if (rt->u.dst.neighbour)
2196                 NLA_PUT(skb, RTA_GATEWAY, 16, &rt->u.dst.neighbour->primary_key);
2197
2198         if (rt->u.dst.dev)
2199                 NLA_PUT_U32(skb, RTA_OIF, rt->rt6i_dev->ifindex);
2200
2201         NLA_PUT_U32(skb, RTA_PRIORITY, rt->rt6i_metric);
2202
2203         expires = rt->rt6i_expires ? rt->rt6i_expires - jiffies : 0;
2204         if (rtnl_put_cacheinfo(skb, &rt->u.dst, 0, 0, 0,
2205                                expires, rt->u.dst.error) < 0)
2206                 goto nla_put_failure;
2207
2208         return nlmsg_end(skb, nlh);
2209
2210 nla_put_failure:
2211         nlmsg_cancel(skb, nlh);
2212         return -EMSGSIZE;
2213 }
2214
2215 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
2216 {
2217         struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
2218         int prefix;
2219
2220         if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
2221                 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
2222                 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
2223         } else
2224                 prefix = 0;
2225
2226         return rt6_fill_node(arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
2227                      NETLINK_CB(arg->cb->skb).pid, arg->cb->nlh->nlmsg_seq,
2228                      prefix, 0, NLM_F_MULTI);
2229 }
2230
2231 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr* nlh, void *arg)
2232 {
2233         struct net *net = sock_net(in_skb->sk);
2234         struct nlattr *tb[RTA_MAX+1];
2235         struct rt6_info *rt;
2236         struct sk_buff *skb;
2237         struct rtmsg *rtm;
2238         struct flowi fl;
2239         int err, iif = 0;
2240
2241         err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2242         if (err < 0)
2243                 goto errout;
2244
2245         err = -EINVAL;
2246         memset(&fl, 0, sizeof(fl));
2247
2248         if (tb[RTA_SRC]) {
2249                 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
2250                         goto errout;
2251
2252                 ipv6_addr_copy(&fl.fl6_src, nla_data(tb[RTA_SRC]));
2253         }
2254
2255         if (tb[RTA_DST]) {
2256                 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
2257                         goto errout;
2258
2259                 ipv6_addr_copy(&fl.fl6_dst, nla_data(tb[RTA_DST]));
2260         }
2261
2262         if (tb[RTA_IIF])
2263                 iif = nla_get_u32(tb[RTA_IIF]);
2264
2265         if (tb[RTA_OIF])
2266                 fl.oif = nla_get_u32(tb[RTA_OIF]);
2267
2268         if (iif) {
2269                 struct net_device *dev;
2270                 dev = __dev_get_by_index(net, iif);
2271                 if (!dev) {
2272                         err = -ENODEV;
2273                         goto errout;
2274                 }
2275         }
2276
2277         skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
2278         if (skb == NULL) {
2279                 err = -ENOBUFS;
2280                 goto errout;
2281         }
2282
2283         /* Reserve room for dummy headers, this skb can pass
2284            through good chunk of routing engine.
2285          */
2286         skb_reset_mac_header(skb);
2287         skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
2288
2289         rt = (struct rt6_info*) ip6_route_output(net, NULL, &fl);
2290         skb->dst = &rt->u.dst;
2291
2292         err = rt6_fill_node(skb, rt, &fl.fl6_dst, &fl.fl6_src, iif,
2293                             RTM_NEWROUTE, NETLINK_CB(in_skb).pid,
2294                             nlh->nlmsg_seq, 0, 0, 0);
2295         if (err < 0) {
2296                 kfree_skb(skb);
2297                 goto errout;
2298         }
2299
2300         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
2301 errout:
2302         return err;
2303 }
2304
2305 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info)
2306 {
2307         struct sk_buff *skb;
2308         struct net *net = info->nl_net;
2309         u32 seq;
2310         int err;
2311
2312         err = -ENOBUFS;
2313         seq = info->nlh != NULL ? info->nlh->nlmsg_seq : 0;
2314
2315         skb = nlmsg_new(rt6_nlmsg_size(), gfp_any());
2316         if (skb == NULL)
2317                 goto errout;
2318
2319         err = rt6_fill_node(skb, rt, NULL, NULL, 0,
2320                                 event, info->pid, seq, 0, 0, 0);
2321         if (err < 0) {
2322                 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
2323                 WARN_ON(err == -EMSGSIZE);
2324                 kfree_skb(skb);
2325                 goto errout;
2326         }
2327         err = rtnl_notify(skb, net, info->pid, RTNLGRP_IPV6_ROUTE,
2328                           info->nlh, gfp_any());
2329 errout:
2330         if (err < 0)
2331                 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
2332 }
2333
2334 static int ip6_route_dev_notify(struct notifier_block *this,
2335                                 unsigned long event, void *data)
2336 {
2337         struct net_device *dev = (struct net_device *)data;
2338         struct net *net = dev_net(dev);
2339
2340         if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
2341                 net->ipv6.ip6_null_entry->u.dst.dev = dev;
2342                 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
2343 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2344                 net->ipv6.ip6_prohibit_entry->u.dst.dev = dev;
2345                 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
2346                 net->ipv6.ip6_blk_hole_entry->u.dst.dev = dev;
2347                 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
2348 #endif
2349         }
2350
2351         return NOTIFY_OK;
2352 }
2353
2354 /*
2355  *      /proc
2356  */
2357
2358 #ifdef CONFIG_PROC_FS
2359
2360 #define RT6_INFO_LEN (32 + 4 + 32 + 4 + 32 + 40 + 5 + 1)
2361
2362 struct rt6_proc_arg
2363 {
2364         char *buffer;
2365         int offset;
2366         int length;
2367         int skip;
2368         int len;
2369 };
2370
2371 static int rt6_info_route(struct rt6_info *rt, void *p_arg)
2372 {
2373         struct seq_file *m = p_arg;
2374
2375         seq_printf(m, NIP6_SEQFMT " %02x ", NIP6(rt->rt6i_dst.addr),
2376                    rt->rt6i_dst.plen);
2377
2378 #ifdef CONFIG_IPV6_SUBTREES
2379         seq_printf(m, NIP6_SEQFMT " %02x ", NIP6(rt->rt6i_src.addr),
2380                    rt->rt6i_src.plen);
2381 #else
2382         seq_puts(m, "00000000000000000000000000000000 00 ");
2383 #endif
2384
2385         if (rt->rt6i_nexthop) {
2386                 seq_printf(m, NIP6_SEQFMT,
2387                            NIP6(*((struct in6_addr *)rt->rt6i_nexthop->primary_key)));
2388         } else {
2389                 seq_puts(m, "00000000000000000000000000000000");
2390         }
2391         seq_printf(m, " %08x %08x %08x %08x %8s\n",
2392                    rt->rt6i_metric, atomic_read(&rt->u.dst.__refcnt),
2393                    rt->u.dst.__use, rt->rt6i_flags,
2394                    rt->rt6i_dev ? rt->rt6i_dev->name : "");
2395         return 0;
2396 }
2397
2398 static int ipv6_route_show(struct seq_file *m, void *v)
2399 {
2400         struct net *net = (struct net *)m->private;
2401         fib6_clean_all(net, rt6_info_route, 0, m);
2402         return 0;
2403 }
2404
2405 static int ipv6_route_open(struct inode *inode, struct file *file)
2406 {
2407         int err;
2408         struct net *net = get_proc_net(inode);
2409         if (!net)
2410                 return -ENXIO;
2411
2412         err = single_open(file, ipv6_route_show, net);
2413         if (err < 0) {
2414                 put_net(net);
2415                 return err;
2416         }
2417
2418         return 0;
2419 }
2420
2421 static int ipv6_route_release(struct inode *inode, struct file *file)
2422 {
2423         struct seq_file *seq = file->private_data;
2424         struct net *net = seq->private;
2425         put_net(net);
2426         return single_release(inode, file);
2427 }
2428
2429 static const struct file_operations ipv6_route_proc_fops = {
2430         .owner          = THIS_MODULE,
2431         .open           = ipv6_route_open,
2432         .read           = seq_read,
2433         .llseek         = seq_lseek,
2434         .release        = ipv6_route_release,
2435 };
2436
2437 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
2438 {
2439         struct net *net = (struct net *)seq->private;
2440         seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
2441                    net->ipv6.rt6_stats->fib_nodes,
2442                    net->ipv6.rt6_stats->fib_route_nodes,
2443                    net->ipv6.rt6_stats->fib_rt_alloc,
2444                    net->ipv6.rt6_stats->fib_rt_entries,
2445                    net->ipv6.rt6_stats->fib_rt_cache,
2446                    atomic_read(&net->ipv6.ip6_dst_ops->entries),
2447                    net->ipv6.rt6_stats->fib_discarded_routes);
2448
2449         return 0;
2450 }
2451
2452 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
2453 {
2454         int err;
2455         struct net *net = get_proc_net(inode);
2456         if (!net)
2457                 return -ENXIO;
2458
2459         err = single_open(file, rt6_stats_seq_show, net);
2460         if (err < 0) {
2461                 put_net(net);
2462                 return err;
2463         }
2464
2465         return 0;
2466 }
2467
2468 static int rt6_stats_seq_release(struct inode *inode, struct file *file)
2469 {
2470         struct seq_file *seq = file->private_data;
2471         struct net *net = (struct net *)seq->private;
2472         put_net(net);
2473         return single_release(inode, file);
2474 }
2475
2476 static const struct file_operations rt6_stats_seq_fops = {
2477         .owner   = THIS_MODULE,
2478         .open    = rt6_stats_seq_open,
2479         .read    = seq_read,
2480         .llseek  = seq_lseek,
2481         .release = rt6_stats_seq_release,
2482 };
2483 #endif  /* CONFIG_PROC_FS */
2484
2485 #ifdef CONFIG_SYSCTL
2486
2487 static
2488 int ipv6_sysctl_rtcache_flush(ctl_table *ctl, int write, struct file * filp,
2489                               void __user *buffer, size_t *lenp, loff_t *ppos)
2490 {
2491         struct net *net = current->nsproxy->net_ns;
2492         int delay = net->ipv6.sysctl.flush_delay;
2493         if (write) {
2494                 proc_dointvec(ctl, write, filp, buffer, lenp, ppos);
2495                 fib6_run_gc(delay <= 0 ? ~0UL : (unsigned long)delay, net);
2496                 return 0;
2497         } else
2498                 return -EINVAL;
2499 }
2500
2501 ctl_table ipv6_route_table_template[] = {
2502         {
2503                 .procname       =       "flush",
2504                 .data           =       &init_net.ipv6.sysctl.flush_delay,
2505                 .maxlen         =       sizeof(int),
2506                 .mode           =       0200,
2507                 .proc_handler   =       &ipv6_sysctl_rtcache_flush
2508         },
2509         {
2510                 .ctl_name       =       NET_IPV6_ROUTE_GC_THRESH,
2511                 .procname       =       "gc_thresh",
2512                 .data           =       &ip6_dst_ops_template.gc_thresh,
2513                 .maxlen         =       sizeof(int),
2514                 .mode           =       0644,
2515                 .proc_handler   =       &proc_dointvec,
2516         },
2517         {
2518                 .ctl_name       =       NET_IPV6_ROUTE_MAX_SIZE,
2519                 .procname       =       "max_size",
2520                 .data           =       &init_net.ipv6.sysctl.ip6_rt_max_size,
2521                 .maxlen         =       sizeof(int),
2522                 .mode           =       0644,
2523                 .proc_handler   =       &proc_dointvec,
2524         },
2525         {
2526                 .ctl_name       =       NET_IPV6_ROUTE_GC_MIN_INTERVAL,
2527                 .procname       =       "gc_min_interval",
2528                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2529                 .maxlen         =       sizeof(int),
2530                 .mode           =       0644,
2531                 .proc_handler   =       &proc_dointvec_jiffies,
2532                 .strategy       =       &sysctl_jiffies,
2533         },
2534         {
2535                 .ctl_name       =       NET_IPV6_ROUTE_GC_TIMEOUT,
2536                 .procname       =       "gc_timeout",
2537                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
2538                 .maxlen         =       sizeof(int),
2539                 .mode           =       0644,
2540                 .proc_handler   =       &proc_dointvec_jiffies,
2541                 .strategy       =       &sysctl_jiffies,
2542         },
2543         {
2544                 .ctl_name       =       NET_IPV6_ROUTE_GC_INTERVAL,
2545                 .procname       =       "gc_interval",
2546                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_interval,
2547                 .maxlen         =       sizeof(int),
2548                 .mode           =       0644,
2549                 .proc_handler   =       &proc_dointvec_jiffies,
2550                 .strategy       =       &sysctl_jiffies,
2551         },
2552         {
2553                 .ctl_name       =       NET_IPV6_ROUTE_GC_ELASTICITY,
2554                 .procname       =       "gc_elasticity",
2555                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
2556                 .maxlen         =       sizeof(int),
2557                 .mode           =       0644,
2558                 .proc_handler   =       &proc_dointvec_jiffies,
2559                 .strategy       =       &sysctl_jiffies,
2560         },
2561         {
2562                 .ctl_name       =       NET_IPV6_ROUTE_MTU_EXPIRES,
2563                 .procname       =       "mtu_expires",
2564                 .data           =       &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
2565                 .maxlen         =       sizeof(int),
2566                 .mode           =       0644,
2567                 .proc_handler   =       &proc_dointvec_jiffies,
2568                 .strategy       =       &sysctl_jiffies,
2569         },
2570         {
2571                 .ctl_name       =       NET_IPV6_ROUTE_MIN_ADVMSS,
2572                 .procname       =       "min_adv_mss",
2573                 .data           =       &init_net.ipv6.sysctl.ip6_rt_min_advmss,
2574                 .maxlen         =       sizeof(int),
2575                 .mode           =       0644,
2576                 .proc_handler   =       &proc_dointvec_jiffies,
2577                 .strategy       =       &sysctl_jiffies,
2578         },
2579         {
2580                 .ctl_name       =       NET_IPV6_ROUTE_GC_MIN_INTERVAL_MS,
2581                 .procname       =       "gc_min_interval_ms",
2582                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2583                 .maxlen         =       sizeof(int),
2584                 .mode           =       0644,
2585                 .proc_handler   =       &proc_dointvec_ms_jiffies,
2586                 .strategy       =       &sysctl_ms_jiffies,
2587         },
2588         { .ctl_name = 0 }
2589 };
2590
2591 struct ctl_table *ipv6_route_sysctl_init(struct net *net)
2592 {
2593         struct ctl_table *table;
2594
2595         table = kmemdup(ipv6_route_table_template,
2596                         sizeof(ipv6_route_table_template),
2597                         GFP_KERNEL);
2598
2599         if (table) {
2600                 table[0].data = &net->ipv6.sysctl.flush_delay;
2601                 table[1].data = &net->ipv6.ip6_dst_ops->gc_thresh;
2602                 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
2603                 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2604                 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
2605                 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
2606                 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
2607                 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
2608                 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
2609         }
2610
2611         return table;
2612 }
2613 #endif
2614
2615 static int ip6_route_net_init(struct net *net)
2616 {
2617         int ret = 0;
2618
2619         ret = -ENOMEM;
2620         net->ipv6.ip6_dst_ops = kmemdup(&ip6_dst_ops_template,
2621                                         sizeof(*net->ipv6.ip6_dst_ops),
2622                                         GFP_KERNEL);
2623         if (!net->ipv6.ip6_dst_ops)
2624                 goto out;
2625         net->ipv6.ip6_dst_ops->dst_net = net;
2626
2627         net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
2628                                            sizeof(*net->ipv6.ip6_null_entry),
2629                                            GFP_KERNEL);
2630         if (!net->ipv6.ip6_null_entry)
2631                 goto out_ip6_dst_ops;
2632         net->ipv6.ip6_null_entry->u.dst.path =
2633                 (struct dst_entry *)net->ipv6.ip6_null_entry;
2634         net->ipv6.ip6_null_entry->u.dst.ops = net->ipv6.ip6_dst_ops;
2635
2636 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2637         net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
2638                                                sizeof(*net->ipv6.ip6_prohibit_entry),
2639                                                GFP_KERNEL);
2640         if (!net->ipv6.ip6_prohibit_entry) {
2641                 kfree(net->ipv6.ip6_null_entry);
2642                 goto out;
2643         }
2644         net->ipv6.ip6_prohibit_entry->u.dst.path =
2645                 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
2646         net->ipv6.ip6_prohibit_entry->u.dst.ops = net->ipv6.ip6_dst_ops;
2647
2648         net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
2649                                                sizeof(*net->ipv6.ip6_blk_hole_entry),
2650                                                GFP_KERNEL);
2651         if (!net->ipv6.ip6_blk_hole_entry) {
2652                 kfree(net->ipv6.ip6_null_entry);
2653                 kfree(net->ipv6.ip6_prohibit_entry);
2654                 goto out;
2655         }
2656         net->ipv6.ip6_blk_hole_entry->u.dst.path =
2657                 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
2658         net->ipv6.ip6_blk_hole_entry->u.dst.ops = net->ipv6.ip6_dst_ops;
2659 #endif
2660
2661 #ifdef CONFIG_PROC_FS
2662         proc_net_fops_create(net, "ipv6_route", 0, &ipv6_route_proc_fops);
2663         proc_net_fops_create(net, "rt6_stats", S_IRUGO, &rt6_stats_seq_fops);
2664 #endif
2665         net->ipv6.ip6_rt_gc_expire = 30*HZ;
2666
2667         ret = 0;
2668 out:
2669         return ret;
2670
2671 out_ip6_dst_ops:
2672         kfree(net->ipv6.ip6_dst_ops);
2673         goto out;
2674 }
2675
2676 static void ip6_route_net_exit(struct net *net)
2677 {
2678 #ifdef CONFIG_PROC_FS
2679         proc_net_remove(net, "ipv6_route");
2680         proc_net_remove(net, "rt6_stats");
2681 #endif
2682         kfree(net->ipv6.ip6_null_entry);
2683 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2684         kfree(net->ipv6.ip6_prohibit_entry);
2685         kfree(net->ipv6.ip6_blk_hole_entry);
2686 #endif
2687         kfree(net->ipv6.ip6_dst_ops);
2688 }
2689
2690 static struct pernet_operations ip6_route_net_ops = {
2691         .init = ip6_route_net_init,
2692         .exit = ip6_route_net_exit,
2693 };
2694
2695 static struct notifier_block ip6_route_dev_notifier = {
2696         .notifier_call = ip6_route_dev_notify,
2697         .priority = 0,
2698 };
2699
2700 int __init ip6_route_init(void)
2701 {
2702         int ret;
2703
2704         ret = -ENOMEM;
2705         ip6_dst_ops_template.kmem_cachep =
2706                 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
2707                                   SLAB_HWCACHE_ALIGN, NULL);
2708         if (!ip6_dst_ops_template.kmem_cachep)
2709                 goto out;;
2710
2711         ret = register_pernet_subsys(&ip6_route_net_ops);
2712         if (ret)
2713                 goto out_kmem_cache;
2714
2715         /* Registering of the loopback is done before this portion of code,
2716          * the loopback reference in rt6_info will not be taken, do it
2717          * manually for init_net */
2718         init_net.ipv6.ip6_null_entry->u.dst.dev = init_net.loopback_dev;
2719         init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
2720   #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2721         init_net.ipv6.ip6_prohibit_entry->u.dst.dev = init_net.loopback_dev;
2722         init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
2723         init_net.ipv6.ip6_blk_hole_entry->u.dst.dev = init_net.loopback_dev;
2724         init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
2725   #endif
2726         ret = fib6_init();
2727         if (ret)
2728                 goto out_register_subsys;
2729
2730         ret = xfrm6_init();
2731         if (ret)
2732                 goto out_fib6_init;
2733
2734         ret = fib6_rules_init();
2735         if (ret)
2736                 goto xfrm6_init;
2737
2738         ret = -ENOBUFS;
2739         if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL) ||
2740             __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL) ||
2741             __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL))
2742                 goto fib6_rules_init;
2743
2744         ret = register_netdevice_notifier(&ip6_route_dev_notifier);
2745         if (ret)
2746                 goto fib6_rules_init;
2747
2748 out:
2749         return ret;
2750
2751 fib6_rules_init:
2752         fib6_rules_cleanup();
2753 xfrm6_init:
2754         xfrm6_fini();
2755 out_fib6_init:
2756         fib6_gc_cleanup();
2757 out_register_subsys:
2758         unregister_pernet_subsys(&ip6_route_net_ops);
2759 out_kmem_cache:
2760         kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
2761         goto out;
2762 }
2763
2764 void ip6_route_cleanup(void)
2765 {
2766         unregister_netdevice_notifier(&ip6_route_dev_notifier);
2767         fib6_rules_cleanup();
2768         xfrm6_fini();
2769         fib6_gc_cleanup();
2770         unregister_pernet_subsys(&ip6_route_net_ops);
2771         kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
2772 }