net/ipv6/route.c

   1 /*
   2  *      Linux INET6 implementation
   3  *      FIB front-end.
   4  *
   5  *      Authors:
   6  *      Pedro Roque             <roque@di.fc.ul.pt>
   7  *
   8  *      This program is free software; you can redistribute it and/or
   9  *      modify it under the terms of the GNU General Public License
  10  *      as published by the Free Software Foundation; either version
  11  *      2 of the License, or (at your option) any later version.
  12  */
  13
  14 /*      Changes:
  15  *
  16  *      YOSHIFUJI Hideaki @USAGI
  17  *              reworked default router selection.
  18  *              - respect outgoing interface
  19  *              - select from (probably) reachable routers (i.e.
  20  *              routers in REACHABLE, STALE, DELAY or PROBE states).
  21  *              - always select the same router if it is (probably)
  22  *              reachable.  otherwise, round-robin the list.
  23  *      Ville Nuorvala
  24  *              Fixed routing subtrees.
  25  */
  26
  27 #define pr_fmt(fmt) "IPv6: " fmt
  28
  29 #include <linux/capability.h>
  30 #include <linux/errno.h>
  31 #include <linux/export.h>
  32 #include <linux/types.h>
  33 #include <linux/times.h>
  34 #include <linux/socket.h>
  35 #include <linux/sockios.h>
  36 #include <linux/net.h>
  37 #include <linux/route.h>
  38 #include <linux/netdevice.h>
  39 #include <linux/in6.h>
  40 #include <linux/mroute6.h>
  41 #include <linux/init.h>
  42 #include <linux/if_arp.h>
  43 #include <linux/proc_fs.h>
  44 #include <linux/seq_file.h>
  45 #include <linux/nsproxy.h>
  46 #include <linux/slab.h>
  47 #include <net/net_namespace.h>
  48 #include <net/snmp.h>
  49 #include <net/ipv6.h>
  50 #include <net/ip6_fib.h>
  51 #include <net/ip6_route.h>
  52 #include <net/ndisc.h>
  53 #include <net/addrconf.h>
  54 #include <net/tcp.h>
  55 #include <linux/rtnetlink.h>
  56 #include <net/dst.h>
  57 #include <net/xfrm.h>
  58 #include <net/netevent.h>
  59 #include <net/netlink.h>
  60 #include <net/nexthop.h>
  61
  62 #include <asm/uaccess.h>
  63
  64 #ifdef CONFIG_SYSCTL
  65 #include <linux/sysctl.h>
  66 #endif
  67
  68 enum rt6_nud_state {
  69         RT6_NUD_FAIL_HARD = -3,
  70         RT6_NUD_FAIL_PROBE = -2,
  71         RT6_NUD_FAIL_DO_RR = -1,
  72         RT6_NUD_SUCCEED = 1
  73 };
  74
  75 static struct rt6_info *ip6_rt_copy(struct rt6_info *ort,
  76                                     const struct in6_addr *dest);
  77 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie);
  78 static unsigned int      ip6_default_advmss(const struct dst_entry *dst);
  79 static unsigned int      ip6_mtu(const struct dst_entry *dst);
  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 sock *sk, struct sk_buff *skb);
  88 static int              ip6_pkt_prohibit(struct sk_buff *skb);
  89 static int              ip6_pkt_prohibit_out(struct sock *sk, struct sk_buff *skb);
  90 static void             ip6_link_failure(struct sk_buff *skb);
  91 static void             ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
  92                                            struct sk_buff *skb, u32 mtu);
  93 static void             rt6_do_redirect(struct dst_entry *dst, struct sock *sk,
  94                                         struct sk_buff *skb);
  95 static int rt6_score_route(struct rt6_info *rt, int oif, int strict);
  96
  97 #ifdef CONFIG_IPV6_ROUTE_INFO
  98 static struct rt6_info *rt6_add_route_info(struct net *net,
  99                                            const struct in6_addr *prefix, int prefixlen,
 100                                            const struct in6_addr *gwaddr, int ifindex,
 101                                            unsigned int pref);
 102 static struct rt6_info *rt6_get_route_info(struct net *net,
 103                                            const struct in6_addr *prefix, int prefixlen,
 104                                            const struct in6_addr *gwaddr, int ifindex);
 105 #endif
 106
 107 static void rt6_bind_peer(struct rt6_info *rt, int create)
 108 {
 109         struct inet_peer_base *base;
 110         struct inet_peer *peer;
 111
 112         base = inetpeer_base_ptr(rt->_rt6i_peer);
 113         if (!base)
 114                 return;
 115
 116         peer = inet_getpeer_v6(base, &rt->rt6i_dst.addr, create);
 117         if (peer) {
 118                 if (!rt6_set_peer(rt, peer))
 119                         inet_putpeer(peer);
 120         }
 121 }
 122
 123 static struct inet_peer *__rt6_get_peer(struct rt6_info *rt, int create)
 124 {
 125         if (rt6_has_peer(rt))
 126                 return rt6_peer_ptr(rt);
 127
 128         rt6_bind_peer(rt, create);
 129         return (rt6_has_peer(rt) ? rt6_peer_ptr(rt) : NULL);
 130 }
 131
 132 static struct inet_peer *rt6_get_peer_create(struct rt6_info *rt)
 133 {
 134         return __rt6_get_peer(rt, 1);
 135 }
 136
 137 static u32 *ipv6_cow_metrics(struct dst_entry *dst, unsigned long old)
 138 {
 139         struct rt6_info *rt = (struct rt6_info *) dst;
 140         struct inet_peer *peer;
 141         u32 *p = NULL;
 142
 143         if (!(rt->dst.flags & DST_HOST))
 144                 return NULL;
 145
 146         peer = rt6_get_peer_create(rt);
 147         if (peer) {
 148                 u32 *old_p = __DST_METRICS_PTR(old);
 149                 unsigned long prev, new;
 150
 151                 p = peer->metrics;
 152                 if (inet_metrics_new(peer) ||
 153                     (old & DST_METRICS_FORCE_OVERWRITE))
 154                         memcpy(p, old_p, sizeof(u32) * RTAX_MAX);
 155
 156                 new = (unsigned long) p;
 157                 prev = cmpxchg(&dst->_metrics, old, new);
 158
 159                 if (prev != old) {
 160                         p = __DST_METRICS_PTR(prev);
 161                         if (prev & DST_METRICS_READ_ONLY)
 162                                 p = NULL;
 163                 }
 164         }
 165         return p;
 166 }
 167
 168 static inline const void *choose_neigh_daddr(struct rt6_info *rt,
 169                                              struct sk_buff *skb,
 170                                              const void *daddr)
 171 {
 172         struct in6_addr *p = &rt->rt6i_gateway;
 173
 174         if (!ipv6_addr_any(p))
 175                 return (const void *) p;
 176         else if (skb)
 177                 return &ipv6_hdr(skb)->daddr;
 178         return daddr;
 179 }
 180
 181 static struct neighbour *ip6_neigh_lookup(const struct dst_entry *dst,
 182                                           struct sk_buff *skb,
 183                                           const void *daddr)
 184 {
 185         struct rt6_info *rt = (struct rt6_info *) dst;
 186         struct neighbour *n;
 187
 188         daddr = choose_neigh_daddr(rt, skb, daddr);
 189         n = __ipv6_neigh_lookup(dst->dev, daddr);
 190         if (n)
 191                 return n;
 192         return neigh_create(&nd_tbl, daddr, dst->dev);
 193 }
 194
 195 static struct dst_ops ip6_dst_ops_template = {
 196         .family                 =       AF_INET6,
 197         .protocol               =       cpu_to_be16(ETH_P_IPV6),
 198         .gc                     =       ip6_dst_gc,
 199         .gc_thresh              =       1024,
 200         .check                  =       ip6_dst_check,
 201         .default_advmss         =       ip6_default_advmss,
 202         .mtu                    =       ip6_mtu,
 203         .cow_metrics            =       ipv6_cow_metrics,
 204         .destroy                =       ip6_dst_destroy,
 205         .ifdown                 =       ip6_dst_ifdown,
 206         .negative_advice        =       ip6_negative_advice,
 207         .link_failure           =       ip6_link_failure,
 208         .update_pmtu            =       ip6_rt_update_pmtu,
 209         .redirect               =       rt6_do_redirect,
 210         .local_out              =       __ip6_local_out,
 211         .neigh_lookup           =       ip6_neigh_lookup,
 212 };
 213
 214 static unsigned int ip6_blackhole_mtu(const struct dst_entry *dst)
 215 {
 216         unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
 217
 218         return mtu ? : dst->dev->mtu;
 219 }
 220
 221 static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, struct sock *sk,
 222                                          struct sk_buff *skb, u32 mtu)
 223 {
 224 }
 225
 226 static void ip6_rt_blackhole_redirect(struct dst_entry *dst, struct sock *sk,
 227                                       struct sk_buff *skb)
 228 {
 229 }
 230
 231 static u32 *ip6_rt_blackhole_cow_metrics(struct dst_entry *dst,
 232                                          unsigned long old)
 233 {
 234         return NULL;
 235 }
 236
 237 static struct dst_ops ip6_dst_blackhole_ops = {
 238         .family                 =       AF_INET6,
 239         .protocol               =       cpu_to_be16(ETH_P_IPV6),
 240         .destroy                =       ip6_dst_destroy,
 241         .check                  =       ip6_dst_check,
 242         .mtu                    =       ip6_blackhole_mtu,
 243         .default_advmss         =       ip6_default_advmss,
 244         .update_pmtu            =       ip6_rt_blackhole_update_pmtu,
 245         .redirect               =       ip6_rt_blackhole_redirect,
 246         .cow_metrics            =       ip6_rt_blackhole_cow_metrics,
 247         .neigh_lookup           =       ip6_neigh_lookup,
 248 };
 249
 250 static const u32 ip6_template_metrics[RTAX_MAX] = {
 251         [RTAX_HOPLIMIT - 1] = 0,
 252 };
 253
 254 static const struct rt6_info ip6_null_entry_template = {
 255         .dst = {
 256                 .__refcnt       = ATOMIC_INIT(1),
 257                 .__use          = 1,
 258                 .obsolete       = DST_OBSOLETE_FORCE_CHK,
 259                 .error          = -ENETUNREACH,
 260                 .input          = ip6_pkt_discard,
 261                 .output         = ip6_pkt_discard_out,
 262         },
 263         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
 264         .rt6i_protocol  = RTPROT_KERNEL,
 265         .rt6i_metric    = ~(u32) 0,
 266         .rt6i_ref       = ATOMIC_INIT(1),
 267 };
 268
 269 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
 270
 271 static const struct rt6_info ip6_prohibit_entry_template = {
 272         .dst = {
 273                 .__refcnt       = ATOMIC_INIT(1),
 274                 .__use          = 1,
 275                 .obsolete       = DST_OBSOLETE_FORCE_CHK,
 276                 .error          = -EACCES,
 277                 .input          = ip6_pkt_prohibit,
 278                 .output         = ip6_pkt_prohibit_out,
 279         },
 280         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
 281         .rt6i_protocol  = RTPROT_KERNEL,
 282         .rt6i_metric    = ~(u32) 0,
 283         .rt6i_ref       = ATOMIC_INIT(1),
 284 };
 285
 286 static const struct rt6_info ip6_blk_hole_entry_template = {
 287         .dst = {
 288                 .__refcnt       = ATOMIC_INIT(1),
 289                 .__use          = 1,
 290                 .obsolete       = DST_OBSOLETE_FORCE_CHK,
 291                 .error          = -EINVAL,
 292                 .input          = dst_discard,
 293                 .output         = dst_discard_sk,
 294         },
 295         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
 296         .rt6i_protocol  = RTPROT_KERNEL,
 297         .rt6i_metric    = ~(u32) 0,
 298         .rt6i_ref       = ATOMIC_INIT(1),
 299 };
 300
 301 #endif
 302
 303 /* allocate dst with ip6_dst_ops */
 304 static inline struct rt6_info *ip6_dst_alloc(struct net *net,
 305                                              struct net_device *dev,
 306                                              int flags,
 307                                              struct fib6_table *table)
 308 {
 309         struct rt6_info *rt = dst_alloc(&net->ipv6.ip6_dst_ops, dev,
 310                                         0, DST_OBSOLETE_FORCE_CHK, flags);
 311
 312         if (rt) {
 313                 struct dst_entry *dst = &rt->dst;
 314
 315                 memset(dst + 1, 0, sizeof(*rt) - sizeof(*dst));
 316                 rt6_init_peer(rt, table ? &table->tb6_peers : net->ipv6.peers);
 317                 rt->rt6i_genid = rt_genid_ipv6(net);
 318                 INIT_LIST_HEAD(&rt->rt6i_siblings);
 319         }
 320         return rt;
 321 }
 322
 323 static void ip6_dst_destroy(struct dst_entry *dst)
 324 {
 325         struct rt6_info *rt = (struct rt6_info *)dst;
 326         struct inet6_dev *idev = rt->rt6i_idev;
 327         struct dst_entry *from = dst->from;
 328
 329         if (!(rt->dst.flags & DST_HOST))
 330                 dst_destroy_metrics_generic(dst);
 331
 332         if (idev) {
 333                 rt->rt6i_idev = NULL;
 334                 in6_dev_put(idev);
 335         }
 336
 337         dst->from = NULL;
 338         dst_release(from);
 339
 340         if (rt6_has_peer(rt)) {
 341                 struct inet_peer *peer = rt6_peer_ptr(rt);
 342                 inet_putpeer(peer);
 343         }
 344 }
 345
 346 static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
 347                            int how)
 348 {
 349         struct rt6_info *rt = (struct rt6_info *)dst;
 350         struct inet6_dev *idev = rt->rt6i_idev;
 351         struct net_device *loopback_dev =
 352                 dev_net(dev)->loopback_dev;
 353
 354         if (dev != loopback_dev) {
 355                 if (idev && idev->dev == dev) {
 356                         struct inet6_dev *loopback_idev =
 357                                 in6_dev_get(loopback_dev);
 358                         if (loopback_idev) {
 359                                 rt->rt6i_idev = loopback_idev;
 360                                 in6_dev_put(idev);
 361                         }
 362                 }
 363         }
 364 }
 365
 366 static bool rt6_check_expired(const struct rt6_info *rt)
 367 {
 368         if (rt->rt6i_flags & RTF_EXPIRES) {
 369                 if (time_after(jiffies, rt->dst.expires))
 370                         return true;
 371         } else if (rt->dst.from) {
 372                 return rt6_check_expired((struct rt6_info *) rt->dst.from);
 373         }
 374         return false;
 375 }
 376
 377 /* Multipath route selection:
 378  *   Hash based function using packet header and flowlabel.
 379  * Adapted from fib_info_hashfn()
 380  */
 381 static int rt6_info_hash_nhsfn(unsigned int candidate_count,
 382                                const struct flowi6 *fl6)
 383 {
 384         unsigned int val = fl6->flowi6_proto;
 385
 386         val ^= ipv6_addr_hash(&fl6->daddr);
 387         val ^= ipv6_addr_hash(&fl6->saddr);
 388
 389         /* Work only if this not encapsulated */
 390         switch (fl6->flowi6_proto) {
 391         case IPPROTO_UDP:
 392         case IPPROTO_TCP:
 393         case IPPROTO_SCTP:
 394                 val ^= (__force u16)fl6->fl6_sport;
 395                 val ^= (__force u16)fl6->fl6_dport;
 396                 break;
 397
 398         case IPPROTO_ICMPV6:
 399                 val ^= (__force u16)fl6->fl6_icmp_type;
 400                 val ^= (__force u16)fl6->fl6_icmp_code;
 401                 break;
 402         }
 403         /* RFC6438 recommands to use flowlabel */
 404         val ^= (__force u32)fl6->flowlabel;
 405
 406         /* Perhaps, we need to tune, this function? */
 407         val = val ^ (val >> 7) ^ (val >> 12);
 408         return val % candidate_count;
 409 }
 410
 411 static struct rt6_info *rt6_multipath_select(struct rt6_info *match,
 412                                              struct flowi6 *fl6, int oif,
 413                                              int strict)
 414 {
 415         struct rt6_info *sibling, *next_sibling;
 416         int route_choosen;
 417
 418         route_choosen = rt6_info_hash_nhsfn(match->rt6i_nsiblings + 1, fl6);
 419         /* Don't change the route, if route_choosen == 0
 420          * (siblings does not include ourself)
 421          */
 422         if (route_choosen)
 423                 list_for_each_entry_safe(sibling, next_sibling,
 424                                 &match->rt6i_siblings, rt6i_siblings) {
 425                         route_choosen--;
 426                         if (route_choosen == 0) {
 427                                 if (rt6_score_route(sibling, oif, strict) < 0)
 428                                         break;
 429                                 match = sibling;
 430                                 break;
 431                         }
 432                 }
 433         return match;
 434 }
 435
 436 /*
 437  *      Route lookup. Any table->tb6_lock is implied.
 438  */
 439
 440 static inline struct rt6_info *rt6_device_match(struct net *net,
 441                                                     struct rt6_info *rt,
 442                                                     const struct in6_addr *saddr,
 443                                                     int oif,
 444                                                     int flags)
 445 {
 446         struct rt6_info *local = NULL;
 447         struct rt6_info *sprt;
 448
 449         if (!oif && ipv6_addr_any(saddr))
 450                 goto out;
 451
 452         for (sprt = rt; sprt; sprt = sprt->dst.rt6_next) {
 453                 struct net_device *dev = sprt->dst.dev;
 454
 455                 if (oif) {
 456                         if (dev->ifindex == oif)
 457                                 return sprt;
 458                         if (dev->flags & IFF_LOOPBACK) {
 459                                 if (!sprt->rt6i_idev ||
 460                                     sprt->rt6i_idev->dev->ifindex != oif) {
 461                                         if (flags & RT6_LOOKUP_F_IFACE && oif)
 462                                                 continue;
 463                                         if (local && (!oif ||
 464                                                       local->rt6i_idev->dev->ifindex == oif))
 465                                                 continue;
 466                                 }
 467                                 local = sprt;
 468                         }
 469                 } else {
 470                         if (ipv6_chk_addr(net, saddr, dev,
 471                                           flags & RT6_LOOKUP_F_IFACE))
 472                                 return sprt;
 473                 }
 474         }
 475
 476         if (oif) {
 477                 if (local)
 478                         return local;
 479
 480                 if (flags & RT6_LOOKUP_F_IFACE)
 481                         return net->ipv6.ip6_null_entry;
 482         }
 483 out:
 484         return rt;
 485 }
 486
 487 #ifdef CONFIG_IPV6_ROUTER_PREF
 488 struct __rt6_probe_work {
 489         struct work_struct work;
 490         struct in6_addr target;
 491         struct net_device *dev;
 492 };
 493
 494 static void rt6_probe_deferred(struct work_struct *w)
 495 {
 496         struct in6_addr mcaddr;
 497         struct __rt6_probe_work *work =
 498                 container_of(w, struct __rt6_probe_work, work);
 499
 500         addrconf_addr_solict_mult(&work->target, &mcaddr);
 501         ndisc_send_ns(work->dev, NULL, &work->target, &mcaddr, NULL);
 502         dev_put(work->dev);
 503         kfree(w);
 504 }
 505
 506 static void rt6_probe(struct rt6_info *rt)
 507 {
 508         struct neighbour *neigh;
 509         /*
 510          * Okay, this does not seem to be appropriate
 511          * for now, however, we need to check if it
 512          * is really so; aka Router Reachability Probing.
 513          *
 514          * Router Reachability Probe MUST be rate-limited
 515          * to no more than one per minute.
 516          */
 517         if (!rt || !(rt->rt6i_flags & RTF_GATEWAY))
 518                 return;
 519         rcu_read_lock_bh();
 520         neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
 521         if (neigh) {
 522                 write_lock(&neigh->lock);
 523                 if (neigh->nud_state & NUD_VALID)
 524                         goto out;
 525         }
 526
 527         if (!neigh ||
 528             time_after(jiffies, neigh->updated + rt->rt6i_idev->cnf.rtr_probe_interval)) {
 529                 struct __rt6_probe_work *work;
 530
 531                 work = kmalloc(sizeof(*work), GFP_ATOMIC);
 532
 533                 if (neigh && work)
 534                         __neigh_set_probe_once(neigh);
 535
 536                 if (neigh)
 537                         write_unlock(&neigh->lock);
 538
 539                 if (work) {
 540                         INIT_WORK(&work->work, rt6_probe_deferred);
 541                         work->target = rt->rt6i_gateway;
 542                         dev_hold(rt->dst.dev);
 543                         work->dev = rt->dst.dev;
 544                         schedule_work(&work->work);
 545                 }
 546         } else {
 547 out:
 548                 write_unlock(&neigh->lock);
 549         }
 550         rcu_read_unlock_bh();
 551 }
 552 #else
 553 static inline void rt6_probe(struct rt6_info *rt)
 554 {
 555 }
 556 #endif
 557
 558 /*
 559  * Default Router Selection (RFC 2461 6.3.6)
 560  */
 561 static inline int rt6_check_dev(struct rt6_info *rt, int oif)
 562 {
 563         struct net_device *dev = rt->dst.dev;
 564         if (!oif || dev->ifindex == oif)
 565                 return 2;
 566         if ((dev->flags & IFF_LOOPBACK) &&
 567             rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
 568                 return 1;
 569         return 0;
 570 }
 571
 572 static inline enum rt6_nud_state rt6_check_neigh(struct rt6_info *rt)
 573 {
 574         struct neighbour *neigh;
 575         enum rt6_nud_state ret = RT6_NUD_FAIL_HARD;
 576
 577         if (rt->rt6i_flags & RTF_NONEXTHOP ||
 578             !(rt->rt6i_flags & RTF_GATEWAY))
 579                 return RT6_NUD_SUCCEED;
 580
 581         rcu_read_lock_bh();
 582         neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
 583         if (neigh) {
 584                 read_lock(&neigh->lock);
 585                 if (neigh->nud_state & NUD_VALID)
 586                         ret = RT6_NUD_SUCCEED;
 587 #ifdef CONFIG_IPV6_ROUTER_PREF
 588                 else if (!(neigh->nud_state & NUD_FAILED))
 589                         ret = RT6_NUD_SUCCEED;
 590                 else
 591                         ret = RT6_NUD_FAIL_PROBE;
 592 #endif
 593                 read_unlock(&neigh->lock);
 594         } else {
 595                 ret = IS_ENABLED(CONFIG_IPV6_ROUTER_PREF) ?
 596                       RT6_NUD_SUCCEED : RT6_NUD_FAIL_DO_RR;
 597         }
 598         rcu_read_unlock_bh();
 599
 600         return ret;
 601 }
 602
 603 static int rt6_score_route(struct rt6_info *rt, int oif,
 604                            int strict)
 605 {
 606         int m;
 607
 608         m = rt6_check_dev(rt, oif);
 609         if (!m && (strict & RT6_LOOKUP_F_IFACE))
 610                 return RT6_NUD_FAIL_HARD;
 611 #ifdef CONFIG_IPV6_ROUTER_PREF
 612         m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
 613 #endif
 614         if (strict & RT6_LOOKUP_F_REACHABLE) {
 615                 int n = rt6_check_neigh(rt);
 616                 if (n < 0)
 617                         return n;
 618         }
 619         return m;
 620 }
 621
 622 static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
 623                                    int *mpri, struct rt6_info *match,
 624                                    bool *do_rr)
 625 {
 626         int m;
 627         bool match_do_rr = false;
 628
 629         if (rt6_check_expired(rt))
 630                 goto out;
 631
 632         m = rt6_score_route(rt, oif, strict);
 633         if (m == RT6_NUD_FAIL_DO_RR) {
 634                 match_do_rr = true;
 635                 m = 0; /* lowest valid score */
 636         } else if (m == RT6_NUD_FAIL_HARD) {
 637                 goto out;
 638         }
 639
 640         if (strict & RT6_LOOKUP_F_REACHABLE)
 641                 rt6_probe(rt);
 642
 643         /* note that m can be RT6_NUD_FAIL_PROBE at this point */
 644         if (m > *mpri) {
 645                 *do_rr = match_do_rr;
 646                 *mpri = m;
 647                 match = rt;
 648         }
 649 out:
 650         return match;
 651 }
 652
 653 static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
 654                                      struct rt6_info *rr_head,
 655                                      u32 metric, int oif, int strict,
 656                                      bool *do_rr)
 657 {
 658         struct rt6_info *rt, *match;
 659         int mpri = -1;
 660
 661         match = NULL;
 662         for (rt = rr_head; rt && rt->rt6i_metric == metric;
 663              rt = rt->dst.rt6_next)
 664                 match = find_match(rt, oif, strict, &mpri, match, do_rr);
 665         for (rt = fn->leaf; rt && rt != rr_head && rt->rt6i_metric == metric;
 666              rt = rt->dst.rt6_next)
 667                 match = find_match(rt, oif, strict, &mpri, match, do_rr);
 668
 669         return match;
 670 }
 671
 672 static struct rt6_info *rt6_select(struct fib6_node *fn, int oif, int strict)
 673 {
 674         struct rt6_info *match, *rt0;
 675         struct net *net;
 676         bool do_rr = false;
 677
 678         rt0 = fn->rr_ptr;
 679         if (!rt0)
 680                 fn->rr_ptr = rt0 = fn->leaf;
 681
 682         match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict,
 683                              &do_rr);
 684
 685         if (do_rr) {
 686                 struct rt6_info *next = rt0->dst.rt6_next;
 687
 688                 /* no entries matched; do round-robin */
 689                 if (!next || next->rt6i_metric != rt0->rt6i_metric)
 690                         next = fn->leaf;
 691
 692                 if (next != rt0)
 693                         fn->rr_ptr = next;
 694         }
 695
 696         net = dev_net(rt0->dst.dev);
 697         return match ? match : net->ipv6.ip6_null_entry;
 698 }
 699
 700 #ifdef CONFIG_IPV6_ROUTE_INFO
 701 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
 702                   const struct in6_addr *gwaddr)
 703 {
 704         struct net *net = dev_net(dev);
 705         struct route_info *rinfo = (struct route_info *) opt;
 706         struct in6_addr prefix_buf, *prefix;
 707         unsigned int pref;
 708         unsigned long lifetime;
 709         struct rt6_info *rt;
 710
 711         if (len < sizeof(struct route_info)) {
 712                 return -EINVAL;
 713         }
 714
 715         /* Sanity check for prefix_len and length */
 716         if (rinfo->length > 3) {
 717                 return -EINVAL;
 718         } else if (rinfo->prefix_len > 128) {
 719                 return -EINVAL;
 720         } else if (rinfo->prefix_len > 64) {
 721                 if (rinfo->length < 2) {
 722                         return -EINVAL;
 723                 }
 724         } else if (rinfo->prefix_len > 0) {
 725                 if (rinfo->length < 1) {
 726                         return -EINVAL;
 727                 }
 728         }
 729
 730         pref = rinfo->route_pref;
 731         if (pref == ICMPV6_ROUTER_PREF_INVALID)
 732                 return -EINVAL;
 733
 734         lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
 735
 736         if (rinfo->length == 3)
 737                 prefix = (struct in6_addr *)rinfo->prefix;
 738         else {
 739                 /* this function is safe */
 740                 ipv6_addr_prefix(&prefix_buf,
 741                                  (struct in6_addr *)rinfo->prefix,
 742                                  rinfo->prefix_len);
 743                 prefix = &prefix_buf;
 744         }
 745
 746         if (rinfo->prefix_len == 0)
 747                 rt = rt6_get_dflt_router(gwaddr, dev);
 748         else
 749                 rt = rt6_get_route_info(net, prefix, rinfo->prefix_len,
 750                                         gwaddr, dev->ifindex);
 751
 752         if (rt && !lifetime) {
 753                 ip6_del_rt(rt);
 754                 rt = NULL;
 755         }
 756
 757         if (!rt && lifetime)
 758                 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr, dev->ifindex,
 759                                         pref);
 760         else if (rt)
 761                 rt->rt6i_flags = RTF_ROUTEINFO |
 762                                  (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
 763
 764         if (rt) {
 765                 if (!addrconf_finite_timeout(lifetime))
 766                         rt6_clean_expires(rt);
 767                 else
 768                         rt6_set_expires(rt, jiffies + HZ * lifetime);
 769
 770                 ip6_rt_put(rt);
 771         }
 772         return 0;
 773 }
 774 #endif
 775
 776 #define BACKTRACK(__net, saddr)                 \
 777 do { \
 778         if (rt == __net->ipv6.ip6_null_entry) { \
 779                 struct fib6_node *pn; \
 780                 while (1) { \
 781                         if (fn->fn_flags & RTN_TL_ROOT) \
 782                                 goto out; \
 783                         pn = fn->parent; \
 784                         if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn) \
 785                                 fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr); \
 786                         else \
 787                                 fn = pn; \
 788                         if (fn->fn_flags & RTN_RTINFO) \
 789                                 goto restart; \
 790                 } \
 791         } \
 792 } while (0)
 793
 794 static struct rt6_info *ip6_pol_route_lookup(struct net *net,
 795                                              struct fib6_table *table,
 796                                              struct flowi6 *fl6, int flags)
 797 {
 798         struct fib6_node *fn;
 799         struct rt6_info *rt;
 800
 801         read_lock_bh(&table->tb6_lock);
 802         fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
 803 restart:
 804         rt = fn->leaf;
 805         rt = rt6_device_match(net, rt, &fl6->saddr, fl6->flowi6_oif, flags);
 806         if (rt->rt6i_nsiblings && fl6->flowi6_oif == 0)
 807                 rt = rt6_multipath_select(rt, fl6, fl6->flowi6_oif, flags);
 808         BACKTRACK(net, &fl6->saddr);
 809 out:
 810         dst_use(&rt->dst, jiffies);
 811         read_unlock_bh(&table->tb6_lock);
 812         return rt;
 813
 814 }
 815
 816 struct dst_entry * ip6_route_lookup(struct net *net, struct flowi6 *fl6,
 817                                     int flags)
 818 {
 819         return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_lookup);
 820 }
 821 EXPORT_SYMBOL_GPL(ip6_route_lookup);
 822
 823 struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
 824                             const struct in6_addr *saddr, int oif, int strict)
 825 {
 826         struct flowi6 fl6 = {
 827                 .flowi6_oif = oif,
 828                 .daddr = *daddr,
 829         };
 830         struct dst_entry *dst;
 831         int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
 832
 833         if (saddr) {
 834                 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
 835                 flags |= RT6_LOOKUP_F_HAS_SADDR;
 836         }
 837
 838         dst = fib6_rule_lookup(net, &fl6, flags, ip6_pol_route_lookup);
 839         if (dst->error == 0)
 840                 return (struct rt6_info *) dst;
 841
 842         dst_release(dst);
 843
 844         return NULL;
 845 }
 846
 847 EXPORT_SYMBOL(rt6_lookup);
 848
 849 /* ip6_ins_rt is called with FREE table->tb6_lock.
 850    It takes new route entry, the addition fails by any reason the
 851    route is freed. In any case, if caller does not hold it, it may
 852    be destroyed.
 853  */
 854
 855 static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info,
 856                         struct nlattr *mx, int mx_len)
 857 {
 858         int err;
 859         struct fib6_table *table;
 860
 861         table = rt->rt6i_table;
 862         write_lock_bh(&table->tb6_lock);
 863         err = fib6_add(&table->tb6_root, rt, info, mx, mx_len);
 864         write_unlock_bh(&table->tb6_lock);
 865
 866         return err;
 867 }
 868
 869 int ip6_ins_rt(struct rt6_info *rt)
 870 {
 871         struct nl_info info = {
 872                 .nl_net = dev_net(rt->dst.dev),
 873         };
 874         return __ip6_ins_rt(rt, &info, NULL, 0);
 875 }
 876
 877 static struct rt6_info *rt6_alloc_cow(struct rt6_info *ort,
 878                                       const struct in6_addr *daddr,
 879                                       const struct in6_addr *saddr)
 880 {
 881         struct rt6_info *rt;
 882
 883         /*
 884          *      Clone the route.
 885          */
 886
 887         rt = ip6_rt_copy(ort, daddr);
 888
 889         if (rt) {
 890                 if (ort->rt6i_dst.plen != 128 &&
 891                     ipv6_addr_equal(&ort->rt6i_dst.addr, daddr))
 892                         rt->rt6i_flags |= RTF_ANYCAST;
 893
 894                 rt->rt6i_flags |= RTF_CACHE;
 895
 896 #ifdef CONFIG_IPV6_SUBTREES
 897                 if (rt->rt6i_src.plen && saddr) {
 898                         rt->rt6i_src.addr = *saddr;
 899                         rt->rt6i_src.plen = 128;
 900                 }
 901 #endif
 902         }
 903
 904         return rt;
 905 }
 906
 907 static struct rt6_info *rt6_alloc_clone(struct rt6_info *ort,
 908                                         const struct in6_addr *daddr)
 909 {
 910         struct rt6_info *rt = ip6_rt_copy(ort, daddr);
 911
 912         if (rt)
 913                 rt->rt6i_flags |= RTF_CACHE;
 914         return rt;
 915 }
 916
 917 static struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table, int oif,
 918                                       struct flowi6 *fl6, int flags)
 919 {
 920         struct fib6_node *fn;
 921         struct rt6_info *rt, *nrt;
 922         int strict = 0;
 923         int attempts = 3;
 924         int err;
 925         int reachable = net->ipv6.devconf_all->forwarding ? 0 : RT6_LOOKUP_F_REACHABLE;
 926
 927         strict |= flags & RT6_LOOKUP_F_IFACE;
 928
 929 relookup:
 930         read_lock_bh(&table->tb6_lock);
 931
 932 restart_2:
 933         fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
 934
 935 restart:
 936         rt = rt6_select(fn, oif, strict | reachable);
 937         if (rt->rt6i_nsiblings)
 938                 rt = rt6_multipath_select(rt, fl6, oif, strict | reachable);
 939         BACKTRACK(net, &fl6->saddr);
 940         if (rt == net->ipv6.ip6_null_entry ||
 941             rt->rt6i_flags & RTF_CACHE)
 942                 goto out;
 943
 944         dst_hold(&rt->dst);
 945         read_unlock_bh(&table->tb6_lock);
 946
 947         if (!(rt->rt6i_flags & (RTF_NONEXTHOP | RTF_GATEWAY)))
 948                 nrt = rt6_alloc_cow(rt, &fl6->daddr, &fl6->saddr);
 949         else if (!(rt->dst.flags & DST_HOST))
 950                 nrt = rt6_alloc_clone(rt, &fl6->daddr);
 951         else
 952                 goto out2;
 953
 954         ip6_rt_put(rt);
 955         rt = nrt ? : net->ipv6.ip6_null_entry;
 956
 957         dst_hold(&rt->dst);
 958         if (nrt) {
 959                 err = ip6_ins_rt(nrt);
 960                 if (!err)
 961                         goto out2;
 962         }
 963
 964         if (--attempts <= 0)
 965                 goto out2;
 966
 967         /*
 968          * Race condition! In the gap, when table->tb6_lock was
 969          * released someone could insert this route.  Relookup.
 970          */
 971         ip6_rt_put(rt);
 972         goto relookup;
 973
 974 out:
 975         if (reachable) {
 976                 reachable = 0;
 977                 goto restart_2;
 978         }
 979         dst_hold(&rt->dst);
 980         read_unlock_bh(&table->tb6_lock);
 981 out2:
 982         rt->dst.lastuse = jiffies;
 983         rt->dst.__use++;
 984
 985         return rt;
 986 }
 987
 988 static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
 989                                             struct flowi6 *fl6, int flags)
 990 {
 991         return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, flags);
 992 }
 993
 994 static struct dst_entry *ip6_route_input_lookup(struct net *net,
 995                                                 struct net_device *dev,
 996                                                 struct flowi6 *fl6, int flags)
 997 {
 998         if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
 999                 flags |= RT6_LOOKUP_F_IFACE;
1000
1001         return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_input);
1002 }
1003
1004 void ip6_route_input(struct sk_buff *skb)
1005 {
1006         const struct ipv6hdr *iph = ipv6_hdr(skb);
1007         struct net *net = dev_net(skb->dev);
1008         int flags = RT6_LOOKUP_F_HAS_SADDR;
1009         struct flowi6 fl6 = {
1010                 .flowi6_iif = skb->dev->ifindex,
1011                 .daddr = iph->daddr,
1012                 .saddr = iph->saddr,
1013                 .flowlabel = ip6_flowinfo(iph),
1014                 .flowi6_mark = skb->mark,
1015                 .flowi6_proto = iph->nexthdr,
1016         };
1017
1018         skb_dst_set(skb, ip6_route_input_lookup(net, skb->dev, &fl6, flags));
1019 }
1020
1021 static struct rt6_info *ip6_pol_route_output(struct net *net, struct fib6_table *table,
1022                                              struct flowi6 *fl6, int flags)
1023 {
1024         return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, flags);
1025 }
1026
1027 struct dst_entry * ip6_route_output(struct net *net, const struct sock *sk,
1028                                     struct flowi6 *fl6)
1029 {
1030         int flags = 0;
1031
1032         fl6->flowi6_iif = LOOPBACK_IFINDEX;
1033
1034         if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr))
1035                 flags |= RT6_LOOKUP_F_IFACE;
1036
1037         if (!ipv6_addr_any(&fl6->saddr))
1038                 flags |= RT6_LOOKUP_F_HAS_SADDR;
1039         else if (sk)
1040                 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
1041
1042         return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_output);
1043 }
1044
1045 EXPORT_SYMBOL(ip6_route_output);
1046
1047 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
1048 {
1049         struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
1050         struct dst_entry *new = NULL;
1051
1052         rt = dst_alloc(&ip6_dst_blackhole_ops, ort->dst.dev, 1, DST_OBSOLETE_NONE, 0);
1053         if (rt) {
1054                 new = &rt->dst;
1055
1056                 memset(new + 1, 0, sizeof(*rt) - sizeof(*new));
1057                 rt6_init_peer(rt, net->ipv6.peers);
1058
1059                 new->__use = 1;
1060                 new->input = dst_discard;
1061                 new->output = dst_discard_sk;
1062
1063                 if (dst_metrics_read_only(&ort->dst))
1064                         new->_metrics = ort->dst._metrics;
1065                 else
1066                         dst_copy_metrics(new, &ort->dst);
1067                 rt->rt6i_idev = ort->rt6i_idev;
1068                 if (rt->rt6i_idev)
1069                         in6_dev_hold(rt->rt6i_idev);
1070
1071                 rt->rt6i_gateway = ort->rt6i_gateway;
1072                 rt->rt6i_flags = ort->rt6i_flags;
1073                 rt->rt6i_metric = 0;
1074
1075                 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1076 #ifdef CONFIG_IPV6_SUBTREES
1077                 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1078 #endif
1079
1080                 dst_free(new);
1081         }
1082
1083         dst_release(dst_orig);
1084         return new ? new : ERR_PTR(-ENOMEM);
1085 }
1086
1087 /*
1088  *      Destination cache support functions
1089  */
1090
1091 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
1092 {
1093         struct rt6_info *rt;
1094
1095         rt = (struct rt6_info *) dst;
1096
1097         /* All IPV6 dsts are created with ->obsolete set to the value
1098          * DST_OBSOLETE_FORCE_CHK which forces validation calls down
1099          * into this function always.
1100          */
1101         if (rt->rt6i_genid != rt_genid_ipv6(dev_net(rt->dst.dev)))
1102                 return NULL;
1103
1104         if (!rt->rt6i_node || (rt->rt6i_node->fn_sernum != cookie))
1105                 return NULL;
1106
1107         if (rt6_check_expired(rt))
1108                 return NULL;
1109
1110         return dst;
1111 }
1112
1113 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
1114 {
1115         struct rt6_info *rt = (struct rt6_info *) dst;
1116
1117         if (rt) {
1118                 if (rt->rt6i_flags & RTF_CACHE) {
1119                         if (rt6_check_expired(rt)) {
1120                                 ip6_del_rt(rt);
1121                                 dst = NULL;
1122                         }
1123                 } else {
1124                         dst_release(dst);
1125                         dst = NULL;
1126                 }
1127         }
1128         return dst;
1129 }
1130
1131 static void ip6_link_failure(struct sk_buff *skb)
1132 {
1133         struct rt6_info *rt;
1134
1135         icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
1136
1137         rt = (struct rt6_info *) skb_dst(skb);
1138         if (rt) {
1139                 if (rt->rt6i_flags & RTF_CACHE) {
1140                         dst_hold(&rt->dst);
1141                         if (ip6_del_rt(rt))
1142                                 dst_free(&rt->dst);
1143                 } else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT)) {
1144                         rt->rt6i_node->fn_sernum = -1;
1145                 }
1146         }
1147 }
1148
1149 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
1150                                struct sk_buff *skb, u32 mtu)
1151 {
1152         struct rt6_info *rt6 = (struct rt6_info*)dst;
1153
1154         dst_confirm(dst);
1155         if (mtu < dst_mtu(dst) && rt6->rt6i_dst.plen == 128) {
1156                 struct net *net = dev_net(dst->dev);
1157
1158                 rt6->rt6i_flags |= RTF_MODIFIED;
1159                 if (mtu < IPV6_MIN_MTU) {
1160                         u32 features = dst_metric(dst, RTAX_FEATURES);
1161                         mtu = IPV6_MIN_MTU;
1162                         features |= RTAX_FEATURE_ALLFRAG;
1163                         dst_metric_set(dst, RTAX_FEATURES, features);
1164                 }
1165                 dst_metric_set(dst, RTAX_MTU, mtu);
1166                 rt6_update_expires(rt6, net->ipv6.sysctl.ip6_rt_mtu_expires);
1167         }
1168 }
1169
1170 void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
1171                      int oif, u32 mark)
1172 {
1173         const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1174         struct dst_entry *dst;
1175         struct flowi6 fl6;
1176
1177         memset(&fl6, 0, sizeof(fl6));
1178         fl6.flowi6_oif = oif;
1179         fl6.flowi6_mark = mark;
1180         fl6.daddr = iph->daddr;
1181         fl6.saddr = iph->saddr;
1182         fl6.flowlabel = ip6_flowinfo(iph);
1183
1184         dst = ip6_route_output(net, NULL, &fl6);
1185         if (!dst->error)
1186                 ip6_rt_update_pmtu(dst, NULL, skb, ntohl(mtu));
1187         dst_release(dst);
1188 }
1189 EXPORT_SYMBOL_GPL(ip6_update_pmtu);
1190
1191 void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
1192 {
1193         ip6_update_pmtu(skb, sock_net(sk), mtu,
1194                         sk->sk_bound_dev_if, sk->sk_mark);
1195 }
1196 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
1197
1198 /* Handle redirects */
1199 struct ip6rd_flowi {
1200         struct flowi6 fl6;
1201         struct in6_addr gateway;
1202 };
1203
1204 static struct rt6_info *__ip6_route_redirect(struct net *net,
1205                                              struct fib6_table *table,
1206                                              struct flowi6 *fl6,
1207                                              int flags)
1208 {
1209         struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
1210         struct rt6_info *rt;
1211         struct fib6_node *fn;
1212
1213         /* Get the "current" route for this destination and
1214          * check if the redirect has come from approriate router.
1215          *
1216          * RFC 4861 specifies that redirects should only be
1217          * accepted if they come from the nexthop to the target.
1218          * Due to the way the routes are chosen, this notion
1219          * is a bit fuzzy and one might need to check all possible
1220          * routes.
1221          */
1222
1223         read_lock_bh(&table->tb6_lock);
1224         fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1225 restart:
1226         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1227                 if (rt6_check_expired(rt))
1228                         continue;
1229                 if (rt->dst.error)
1230                         break;
1231                 if (!(rt->rt6i_flags & RTF_GATEWAY))
1232                         continue;
1233                 if (fl6->flowi6_oif != rt->dst.dev->ifindex)
1234                         continue;
1235                 if (!ipv6_addr_equal(&rdfl->gateway, &rt->rt6i_gateway))
1236                         continue;
1237                 break;
1238         }
1239
1240         if (!rt)
1241                 rt = net->ipv6.ip6_null_entry;
1242         else if (rt->dst.error) {
1243                 rt = net->ipv6.ip6_null_entry;
1244                 goto out;
1245         }
1246         BACKTRACK(net, &fl6->saddr);
1247 out:
1248         dst_hold(&rt->dst);
1249
1250         read_unlock_bh(&table->tb6_lock);
1251
1252         return rt;
1253 };
1254
1255 static struct dst_entry *ip6_route_redirect(struct net *net,
1256                                         const struct flowi6 *fl6,
1257                                         const struct in6_addr *gateway)
1258 {
1259         int flags = RT6_LOOKUP_F_HAS_SADDR;
1260         struct ip6rd_flowi rdfl;
1261
1262         rdfl.fl6 = *fl6;
1263         rdfl.gateway = *gateway;
1264
1265         return fib6_rule_lookup(net, &rdfl.fl6,
1266                                 flags, __ip6_route_redirect);
1267 }
1268
1269 void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark)
1270 {
1271         const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1272         struct dst_entry *dst;
1273         struct flowi6 fl6;
1274
1275         memset(&fl6, 0, sizeof(fl6));
1276         fl6.flowi6_oif = oif;
1277         fl6.flowi6_mark = mark;
1278         fl6.daddr = iph->daddr;
1279         fl6.saddr = iph->saddr;
1280         fl6.flowlabel = ip6_flowinfo(iph);
1281
1282         dst = ip6_route_redirect(net, &fl6, &ipv6_hdr(skb)->saddr);
1283         rt6_do_redirect(dst, NULL, skb);
1284         dst_release(dst);
1285 }
1286 EXPORT_SYMBOL_GPL(ip6_redirect);
1287
1288 void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif,
1289                             u32 mark)
1290 {
1291         const struct ipv6hdr *iph = ipv6_hdr(skb);
1292         const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
1293         struct dst_entry *dst;
1294         struct flowi6 fl6;
1295
1296         memset(&fl6, 0, sizeof(fl6));
1297         fl6.flowi6_oif = oif;
1298         fl6.flowi6_mark = mark;
1299         fl6.daddr = msg->dest;
1300         fl6.saddr = iph->daddr;
1301
1302         dst = ip6_route_redirect(net, &fl6, &iph->saddr);
1303         rt6_do_redirect(dst, NULL, skb);
1304         dst_release(dst);
1305 }
1306
1307 void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
1308 {
1309         ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark);
1310 }
1311 EXPORT_SYMBOL_GPL(ip6_sk_redirect);
1312
1313 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1314 {
1315         struct net_device *dev = dst->dev;
1316         unsigned int mtu = dst_mtu(dst);
1317         struct net *net = dev_net(dev);
1318
1319         mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
1320
1321         if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
1322                 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1323
1324         /*
1325          * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1326          * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1327          * IPV6_MAXPLEN is also valid and means: "any MSS,
1328          * rely only on pmtu discovery"
1329          */
1330         if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
1331                 mtu = IPV6_MAXPLEN;
1332         return mtu;
1333 }
1334
1335 static unsigned int ip6_mtu(const struct dst_entry *dst)
1336 {
1337         struct inet6_dev *idev;
1338         unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
1339
1340         if (mtu)
1341                 goto out;
1342
1343         mtu = IPV6_MIN_MTU;
1344
1345         rcu_read_lock();
1346         idev = __in6_dev_get(dst->dev);
1347         if (idev)
1348                 mtu = idev->cnf.mtu6;
1349         rcu_read_unlock();
1350
1351 out:
1352         return min_t(unsigned int, mtu, IP6_MAX_MTU);
1353 }
1354
1355 static struct dst_entry *icmp6_dst_gc_list;
1356 static DEFINE_SPINLOCK(icmp6_dst_lock);
1357
1358 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
1359                                   struct flowi6 *fl6)
1360 {
1361         struct dst_entry *dst;
1362         struct rt6_info *rt;
1363         struct inet6_dev *idev = in6_dev_get(dev);
1364         struct net *net = dev_net(dev);
1365
1366         if (unlikely(!idev))
1367                 return ERR_PTR(-ENODEV);
1368
1369         rt = ip6_dst_alloc(net, dev, 0, NULL);
1370         if (unlikely(!rt)) {
1371                 in6_dev_put(idev);
1372                 dst = ERR_PTR(-ENOMEM);
1373                 goto out;
1374         }
1375
1376         rt->dst.flags |= DST_HOST;
1377         rt->dst.output  = ip6_output;
1378         atomic_set(&rt->dst.__refcnt, 1);
1379         rt->rt6i_gateway  = fl6->daddr;
1380         rt->rt6i_dst.addr = fl6->daddr;
1381         rt->rt6i_dst.plen = 128;
1382         rt->rt6i_idev     = idev;
1383         dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
1384
1385         spin_lock_bh(&icmp6_dst_lock);
1386         rt->dst.next = icmp6_dst_gc_list;
1387         icmp6_dst_gc_list = &rt->dst;
1388         spin_unlock_bh(&icmp6_dst_lock);
1389
1390         fib6_force_start_gc(net);
1391
1392         dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
1393
1394 out:
1395         return dst;
1396 }
1397
1398 int icmp6_dst_gc(void)
1399 {
1400         struct dst_entry *dst, **pprev;
1401         int more = 0;
1402
1403         spin_lock_bh(&icmp6_dst_lock);
1404         pprev = &icmp6_dst_gc_list;
1405
1406         while ((dst = *pprev) != NULL) {
1407                 if (!atomic_read(&dst->__refcnt)) {
1408                         *pprev = dst->next;
1409                         dst_free(dst);
1410                 } else {
1411                         pprev = &dst->next;
1412                         ++more;
1413                 }
1414         }
1415
1416         spin_unlock_bh(&icmp6_dst_lock);
1417
1418         return more;
1419 }
1420
1421 static void icmp6_clean_all(int (*func)(struct rt6_info *rt, void *arg),
1422                             void *arg)
1423 {
1424         struct dst_entry *dst, **pprev;
1425
1426         spin_lock_bh(&icmp6_dst_lock);
1427         pprev = &icmp6_dst_gc_list;
1428         while ((dst = *pprev) != NULL) {
1429                 struct rt6_info *rt = (struct rt6_info *) dst;
1430                 if (func(rt, arg)) {
1431                         *pprev = dst->next;
1432                         dst_free(dst);
1433                 } else {
1434                         pprev = &dst->next;
1435                 }
1436         }
1437         spin_unlock_bh(&icmp6_dst_lock);
1438 }
1439
1440 static int ip6_dst_gc(struct dst_ops *ops)
1441 {
1442         struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
1443         int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1444         int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1445         int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1446         int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1447         unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1448         int entries;
1449
1450         entries = dst_entries_get_fast(ops);
1451         if (time_after(rt_last_gc + rt_min_interval, jiffies) &&
1452             entries <= rt_max_size)
1453                 goto out;
1454
1455         net->ipv6.ip6_rt_gc_expire++;
1456         fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net, entries > rt_max_size);
1457         entries = dst_entries_get_slow(ops);
1458         if (entries < ops->gc_thresh)
1459                 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1460 out:
1461         net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1462         return entries > rt_max_size;
1463 }
1464
1465 /*
1466  *
1467  */
1468
1469 int ip6_route_add(struct fib6_config *cfg)
1470 {
1471         int err;
1472         struct net *net = cfg->fc_nlinfo.nl_net;
1473         struct rt6_info *rt = NULL;
1474         struct net_device *dev = NULL;
1475         struct inet6_dev *idev = NULL;
1476         struct fib6_table *table;
1477         int addr_type;
1478
1479         if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1480                 return -EINVAL;
1481 #ifndef CONFIG_IPV6_SUBTREES
1482         if (cfg->fc_src_len)
1483                 return -EINVAL;
1484 #endif
1485         if (cfg->fc_ifindex) {
1486                 err = -ENODEV;
1487                 dev = dev_get_by_index(net, cfg->fc_ifindex);
1488                 if (!dev)
1489                         goto out;
1490                 idev = in6_dev_get(dev);
1491                 if (!idev)
1492                         goto out;
1493         }
1494
1495         if (cfg->fc_metric == 0)
1496                 cfg->fc_metric = IP6_RT_PRIO_USER;
1497
1498         err = -ENOBUFS;
1499         if (cfg->fc_nlinfo.nlh &&
1500             !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
1501                 table = fib6_get_table(net, cfg->fc_table);
1502                 if (!table) {
1503                         pr_warn("NLM_F_CREATE should be specified when creating new route\n");
1504                         table = fib6_new_table(net, cfg->fc_table);
1505                 }
1506         } else {
1507                 table = fib6_new_table(net, cfg->fc_table);
1508         }
1509
1510         if (!table)
1511                 goto out;
1512
1513         rt = ip6_dst_alloc(net, NULL, (cfg->fc_flags & RTF_ADDRCONF) ? 0 : DST_NOCOUNT, table);
1514
1515         if (!rt) {
1516                 err = -ENOMEM;
1517                 goto out;
1518         }
1519
1520         if (cfg->fc_flags & RTF_EXPIRES)
1521                 rt6_set_expires(rt, jiffies +
1522                                 clock_t_to_jiffies(cfg->fc_expires));
1523         else
1524                 rt6_clean_expires(rt);
1525
1526         if (cfg->fc_protocol == RTPROT_UNSPEC)
1527                 cfg->fc_protocol = RTPROT_BOOT;
1528         rt->rt6i_protocol = cfg->fc_protocol;
1529
1530         addr_type = ipv6_addr_type(&cfg->fc_dst);
1531
1532         if (addr_type & IPV6_ADDR_MULTICAST)
1533                 rt->dst.input = ip6_mc_input;
1534         else if (cfg->fc_flags & RTF_LOCAL)
1535                 rt->dst.input = ip6_input;
1536         else
1537                 rt->dst.input = ip6_forward;
1538
1539         rt->dst.output = ip6_output;
1540
1541         ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1542         rt->rt6i_dst.plen = cfg->fc_dst_len;
1543         if (rt->rt6i_dst.plen == 128) {
1544                 rt->dst.flags |= DST_HOST;
1545                 dst_metrics_set_force_overwrite(&rt->dst);
1546         }
1547
1548 #ifdef CONFIG_IPV6_SUBTREES
1549         ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1550         rt->rt6i_src.plen = cfg->fc_src_len;
1551 #endif
1552
1553         rt->rt6i_metric = cfg->fc_metric;
1554
1555         /* We cannot add true routes via loopback here,
1556            they would result in kernel looping; promote them to reject routes
1557          */
1558         if ((cfg->fc_flags & RTF_REJECT) ||
1559             (dev && (dev->flags & IFF_LOOPBACK) &&
1560              !(addr_type & IPV6_ADDR_LOOPBACK) &&
1561              !(cfg->fc_flags & RTF_LOCAL))) {
1562                 /* hold loopback dev/idev if we haven't done so. */
1563                 if (dev != net->loopback_dev) {
1564                         if (dev) {
1565                                 dev_put(dev);
1566                                 in6_dev_put(idev);
1567                         }
1568                         dev = net->loopback_dev;
1569                         dev_hold(dev);
1570                         idev = in6_dev_get(dev);
1571                         if (!idev) {
1572                                 err = -ENODEV;
1573                                 goto out;
1574                         }
1575                 }
1576                 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1577                 switch (cfg->fc_type) {
1578                 case RTN_BLACKHOLE:
1579                         rt->dst.error = -EINVAL;
1580                         rt->dst.output = dst_discard_sk;
1581                         rt->dst.input = dst_discard;
1582                         break;
1583                 case RTN_PROHIBIT:
1584                         rt->dst.error = -EACCES;
1585                         rt->dst.output = ip6_pkt_prohibit_out;
1586                         rt->dst.input = ip6_pkt_prohibit;
1587                         break;
1588                 case RTN_THROW:
1589                 default:
1590                         rt->dst.error = (cfg->fc_type == RTN_THROW) ? -EAGAIN
1591                                         : -ENETUNREACH;
1592                         rt->dst.output = ip6_pkt_discard_out;
1593                         rt->dst.input = ip6_pkt_discard;
1594                         break;
1595                 }
1596                 goto install_route;
1597         }
1598
1599         if (cfg->fc_flags & RTF_GATEWAY) {
1600                 const struct in6_addr *gw_addr;
1601                 int gwa_type;
1602
1603                 gw_addr = &cfg->fc_gateway;
1604                 rt->rt6i_gateway = *gw_addr;
1605                 gwa_type = ipv6_addr_type(gw_addr);
1606
1607                 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1608                         struct rt6_info *grt;
1609
1610                         /* IPv6 strictly inhibits using not link-local
1611                            addresses as nexthop address.
1612                            Otherwise, router will not able to send redirects.
1613                            It is very good, but in some (rare!) circumstances
1614                            (SIT, PtP, NBMA NOARP links) it is handy to allow
1615                            some exceptions. --ANK
1616                          */
1617                         err = -EINVAL;
1618                         if (!(gwa_type & IPV6_ADDR_UNICAST))
1619                                 goto out;
1620
1621                         grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, 1);
1622
1623                         err = -EHOSTUNREACH;
1624                         if (!grt)
1625                                 goto out;
1626                         if (dev) {
1627                                 if (dev != grt->dst.dev) {
1628                                         ip6_rt_put(grt);
1629                                         goto out;
1630                                 }
1631                         } else {
1632                                 dev = grt->dst.dev;
1633                                 idev = grt->rt6i_idev;
1634                                 dev_hold(dev);
1635                                 in6_dev_hold(grt->rt6i_idev);
1636                         }
1637                         if (!(grt->rt6i_flags & RTF_GATEWAY))
1638                                 err = 0;
1639                         ip6_rt_put(grt);
1640
1641                         if (err)
1642                                 goto out;
1643                 }
1644                 err = -EINVAL;
1645                 if (!dev || (dev->flags & IFF_LOOPBACK))
1646                         goto out;
1647         }
1648
1649         err = -ENODEV;
1650         if (!dev)
1651                 goto out;
1652
1653         if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
1654                 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
1655                         err = -EINVAL;
1656                         goto out;
1657                 }
1658                 rt->rt6i_prefsrc.addr = cfg->fc_prefsrc;
1659                 rt->rt6i_prefsrc.plen = 128;
1660         } else
1661                 rt->rt6i_prefsrc.plen = 0;
1662
1663         rt->rt6i_flags = cfg->fc_flags;
1664
1665 install_route:
1666         rt->dst.dev = dev;
1667         rt->rt6i_idev = idev;
1668         rt->rt6i_table = table;
1669
1670         cfg->fc_nlinfo.nl_net = dev_net(dev);
1671
1672         return __ip6_ins_rt(rt, &cfg->fc_nlinfo, cfg->fc_mx, cfg->fc_mx_len);
1673
1674 out:
1675         if (dev)
1676                 dev_put(dev);
1677         if (idev)
1678                 in6_dev_put(idev);
1679         if (rt)
1680                 dst_free(&rt->dst);
1681         return err;
1682 }
1683
1684 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
1685 {
1686         int err;
1687         struct fib6_table *table;
1688         struct net *net = dev_net(rt->dst.dev);
1689
1690         if (rt == net->ipv6.ip6_null_entry) {
1691                 err = -ENOENT;
1692                 goto out;
1693         }
1694
1695         table = rt->rt6i_table;
1696         write_lock_bh(&table->tb6_lock);
1697         err = fib6_del(rt, info);
1698         write_unlock_bh(&table->tb6_lock);
1699
1700 out:
1701         ip6_rt_put(rt);
1702         return err;
1703 }
1704
1705 int ip6_del_rt(struct rt6_info *rt)
1706 {
1707         struct nl_info info = {
1708                 .nl_net = dev_net(rt->dst.dev),
1709         };
1710         return __ip6_del_rt(rt, &info);
1711 }
1712
1713 static int ip6_route_del(struct fib6_config *cfg)
1714 {
1715         struct fib6_table *table;
1716         struct fib6_node *fn;
1717         struct rt6_info *rt;
1718         int err = -ESRCH;
1719
1720         table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
1721         if (!table)
1722                 return err;
1723
1724         read_lock_bh(&table->tb6_lock);
1725
1726         fn = fib6_locate(&table->tb6_root,
1727                          &cfg->fc_dst, cfg->fc_dst_len,
1728                          &cfg->fc_src, cfg->fc_src_len);
1729
1730         if (fn) {
1731                 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1732                         if (cfg->fc_ifindex &&
1733                             (!rt->dst.dev ||
1734                              rt->dst.dev->ifindex != cfg->fc_ifindex))
1735                                 continue;
1736                         if (cfg->fc_flags & RTF_GATEWAY &&
1737                             !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
1738                                 continue;
1739                         if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
1740                                 continue;
1741                         dst_hold(&rt->dst);
1742                         read_unlock_bh(&table->tb6_lock);
1743
1744                         return __ip6_del_rt(rt, &cfg->fc_nlinfo);
1745                 }
1746         }
1747         read_unlock_bh(&table->tb6_lock);
1748
1749         return err;
1750 }
1751
1752 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
1753 {
1754         struct net *net = dev_net(skb->dev);
1755         struct netevent_redirect netevent;
1756         struct rt6_info *rt, *nrt = NULL;
1757         struct ndisc_options ndopts;
1758         struct inet6_dev *in6_dev;
1759         struct neighbour *neigh;
1760         struct rd_msg *msg;
1761         int optlen, on_link;
1762         u8 *lladdr;
1763
1764         optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
1765         optlen -= sizeof(*msg);
1766
1767         if (optlen < 0) {
1768                 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
1769                 return;
1770         }
1771
1772         msg = (struct rd_msg *)icmp6_hdr(skb);
1773
1774         if (ipv6_addr_is_multicast(&msg->dest)) {
1775                 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
1776                 return;
1777         }
1778
1779         on_link = 0;
1780         if (ipv6_addr_equal(&msg->dest, &msg->target)) {
1781                 on_link = 1;
1782         } else if (ipv6_addr_type(&msg->target) !=
1783                    (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
1784                 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
1785                 return;
1786         }
1787
1788         in6_dev = __in6_dev_get(skb->dev);
1789         if (!in6_dev)
1790                 return;
1791         if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
1792                 return;
1793
1794         /* RFC2461 8.1:
1795          *      The IP source address of the Redirect MUST be the same as the current
1796          *      first-hop router for the specified ICMP Destination Address.
1797          */
1798
1799         if (!ndisc_parse_options(msg->opt, optlen, &ndopts)) {
1800                 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
1801                 return;
1802         }
1803
1804         lladdr = NULL;
1805         if (ndopts.nd_opts_tgt_lladdr) {
1806                 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
1807                                              skb->dev);
1808                 if (!lladdr) {
1809                         net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
1810                         return;
1811                 }
1812         }
1813
1814         rt = (struct rt6_info *) dst;
1815         if (rt == net->ipv6.ip6_null_entry) {
1816                 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
1817                 return;
1818         }
1819
1820         /* Redirect received -> path was valid.
1821          * Look, redirects are sent only in response to data packets,
1822          * so that this nexthop apparently is reachable. --ANK
1823          */
1824         dst_confirm(&rt->dst);
1825
1826         neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
1827         if (!neigh)
1828                 return;
1829
1830         /*
1831          *      We have finally decided to accept it.
1832          */
1833
1834         neigh_update(neigh, lladdr, NUD_STALE,
1835                      NEIGH_UPDATE_F_WEAK_OVERRIDE|
1836                      NEIGH_UPDATE_F_OVERRIDE|
1837                      (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1838                                      NEIGH_UPDATE_F_ISROUTER))
1839                      );
1840
1841         nrt = ip6_rt_copy(rt, &msg->dest);
1842         if (!nrt)
1843                 goto out;
1844
1845         nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
1846         if (on_link)
1847                 nrt->rt6i_flags &= ~RTF_GATEWAY;
1848
1849         nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
1850
1851         if (ip6_ins_rt(nrt))
1852                 goto out;
1853
1854         netevent.old = &rt->dst;
1855         netevent.new = &nrt->dst;
1856         netevent.daddr = &msg->dest;
1857         netevent.neigh = neigh;
1858         call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
1859
1860         if (rt->rt6i_flags & RTF_CACHE) {
1861                 rt = (struct rt6_info *) dst_clone(&rt->dst);
1862                 ip6_del_rt(rt);
1863         }
1864
1865 out:
1866         neigh_release(neigh);
1867 }
1868
1869 /*
1870  *      Misc support functions
1871  */
1872
1873 static struct rt6_info *ip6_rt_copy(struct rt6_info *ort,
1874                                     const struct in6_addr *dest)
1875 {
1876         struct net *net = dev_net(ort->dst.dev);
1877         struct rt6_info *rt = ip6_dst_alloc(net, ort->dst.dev, 0,
1878                                             ort->rt6i_table);
1879
1880         if (rt) {
1881                 rt->dst.input = ort->dst.input;
1882                 rt->dst.output = ort->dst.output;
1883                 rt->dst.flags |= DST_HOST;
1884
1885                 rt->rt6i_dst.addr = *dest;
1886                 rt->rt6i_dst.plen = 128;
1887                 dst_copy_metrics(&rt->dst, &ort->dst);
1888                 rt->dst.error = ort->dst.error;
1889                 rt->rt6i_idev = ort->rt6i_idev;
1890                 if (rt->rt6i_idev)
1891                         in6_dev_hold(rt->rt6i_idev);
1892                 rt->dst.lastuse = jiffies;
1893
1894                 if (ort->rt6i_flags & RTF_GATEWAY)
1895                         rt->rt6i_gateway = ort->rt6i_gateway;
1896                 else
1897                         rt->rt6i_gateway = *dest;
1898                 rt->rt6i_flags = ort->rt6i_flags;
1899                 rt6_set_from(rt, ort);
1900                 rt->rt6i_metric = 0;
1901
1902 #ifdef CONFIG_IPV6_SUBTREES
1903                 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1904 #endif
1905                 memcpy(&rt->rt6i_prefsrc, &ort->rt6i_prefsrc, sizeof(struct rt6key));
1906                 rt->rt6i_table = ort->rt6i_table;
1907         }
1908         return rt;
1909 }
1910
1911 #ifdef CONFIG_IPV6_ROUTE_INFO
1912 static struct rt6_info *rt6_get_route_info(struct net *net,
1913                                            const struct in6_addr *prefix, int prefixlen,
1914                                            const struct in6_addr *gwaddr, int ifindex)
1915 {
1916         struct fib6_node *fn;
1917         struct rt6_info *rt = NULL;
1918         struct fib6_table *table;
1919
1920         table = fib6_get_table(net, RT6_TABLE_INFO);
1921         if (!table)
1922                 return NULL;
1923
1924         read_lock_bh(&table->tb6_lock);
1925         fn = fib6_locate(&table->tb6_root, prefix ,prefixlen, NULL, 0);
1926         if (!fn)
1927                 goto out;
1928
1929         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1930                 if (rt->dst.dev->ifindex != ifindex)
1931                         continue;
1932                 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
1933                         continue;
1934                 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
1935                         continue;
1936                 dst_hold(&rt->dst);
1937                 break;
1938         }
1939 out:
1940         read_unlock_bh(&table->tb6_lock);
1941         return rt;
1942 }
1943
1944 static struct rt6_info *rt6_add_route_info(struct net *net,
1945                                            const struct in6_addr *prefix, int prefixlen,
1946                                            const struct in6_addr *gwaddr, int ifindex,
1947                                            unsigned int pref)
1948 {
1949         struct fib6_config cfg = {
1950                 .fc_table       = RT6_TABLE_INFO,
1951                 .fc_metric      = IP6_RT_PRIO_USER,
1952                 .fc_ifindex     = ifindex,
1953                 .fc_dst_len     = prefixlen,
1954                 .fc_flags       = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
1955                                   RTF_UP | RTF_PREF(pref),
1956                 .fc_nlinfo.portid = 0,
1957                 .fc_nlinfo.nlh = NULL,
1958                 .fc_nlinfo.nl_net = net,
1959         };
1960
1961         cfg.fc_dst = *prefix;
1962         cfg.fc_gateway = *gwaddr;
1963
1964         /* We should treat it as a default route if prefix length is 0. */
1965         if (!prefixlen)
1966                 cfg.fc_flags |= RTF_DEFAULT;
1967
1968         ip6_route_add(&cfg);
1969
1970         return rt6_get_route_info(net, prefix, prefixlen, gwaddr, ifindex);
1971 }
1972 #endif
1973
1974 struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
1975 {
1976         struct rt6_info *rt;
1977         struct fib6_table *table;
1978
1979         table = fib6_get_table(dev_net(dev), RT6_TABLE_DFLT);
1980         if (!table)
1981                 return NULL;
1982
1983         read_lock_bh(&table->tb6_lock);
1984         for (rt = table->tb6_root.leaf; rt; rt=rt->dst.rt6_next) {
1985                 if (dev == rt->dst.dev &&
1986                     ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
1987                     ipv6_addr_equal(&rt->rt6i_gateway, addr))
1988                         break;
1989         }
1990         if (rt)
1991                 dst_hold(&rt->dst);
1992         read_unlock_bh(&table->tb6_lock);
1993         return rt;
1994 }
1995
1996 struct rt6_info *rt6_add_dflt_router(const struct in6_addr *gwaddr,
1997                                      struct net_device *dev,
1998                                      unsigned int pref)
1999 {
2000         struct fib6_config cfg = {
2001                 .fc_table       = RT6_TABLE_DFLT,
2002                 .fc_metric      = IP6_RT_PRIO_USER,
2003                 .fc_ifindex     = dev->ifindex,
2004                 .fc_flags       = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
2005                                   RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
2006                 .fc_nlinfo.portid = 0,
2007                 .fc_nlinfo.nlh = NULL,
2008                 .fc_nlinfo.nl_net = dev_net(dev),
2009         };
2010
2011         cfg.fc_gateway = *gwaddr;
2012
2013         ip6_route_add(&cfg);
2014
2015         return rt6_get_dflt_router(gwaddr, dev);
2016 }
2017
2018 void rt6_purge_dflt_routers(struct net *net)
2019 {
2020         struct rt6_info *rt;
2021         struct fib6_table *table;
2022
2023         /* NOTE: Keep consistent with rt6_get_dflt_router */
2024         table = fib6_get_table(net, RT6_TABLE_DFLT);
2025         if (!table)
2026                 return;
2027
2028 restart:
2029         read_lock_bh(&table->tb6_lock);
2030         for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2031                 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
2032                     (!rt->rt6i_idev || rt->rt6i_idev->cnf.accept_ra != 2)) {
2033                         dst_hold(&rt->dst);
2034                         read_unlock_bh(&table->tb6_lock);
2035                         ip6_del_rt(rt);
2036                         goto restart;
2037                 }
2038         }
2039         read_unlock_bh(&table->tb6_lock);
2040 }
2041
2042 static void rtmsg_to_fib6_config(struct net *net,
2043                                  struct in6_rtmsg *rtmsg,
2044                                  struct fib6_config *cfg)
2045 {
2046         memset(cfg, 0, sizeof(*cfg));
2047
2048         cfg->fc_table = RT6_TABLE_MAIN;
2049         cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
2050         cfg->fc_metric = rtmsg->rtmsg_metric;
2051         cfg->fc_expires = rtmsg->rtmsg_info;
2052         cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
2053         cfg->fc_src_len = rtmsg->rtmsg_src_len;
2054         cfg->fc_flags = rtmsg->rtmsg_flags;
2055
2056         cfg->fc_nlinfo.nl_net = net;
2057
2058         cfg->fc_dst = rtmsg->rtmsg_dst;
2059         cfg->fc_src = rtmsg->rtmsg_src;
2060         cfg->fc_gateway = rtmsg->rtmsg_gateway;
2061 }
2062
2063 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
2064 {
2065         struct fib6_config cfg;
2066         struct in6_rtmsg rtmsg;
2067         int err;
2068
2069         switch(cmd) {
2070         case SIOCADDRT:         /* Add a route */
2071         case SIOCDELRT:         /* Delete a route */
2072                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2073                         return -EPERM;
2074                 err = copy_from_user(&rtmsg, arg,
2075                                      sizeof(struct in6_rtmsg));
2076                 if (err)
2077                         return -EFAULT;
2078
2079                 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
2080
2081                 rtnl_lock();
2082                 switch (cmd) {
2083                 case SIOCADDRT:
2084                         err = ip6_route_add(&cfg);
2085                         break;
2086                 case SIOCDELRT:
2087                         err = ip6_route_del(&cfg);
2088                         break;
2089                 default:
2090                         err = -EINVAL;
2091                 }
2092                 rtnl_unlock();
2093
2094                 return err;
2095         }
2096
2097         return -EINVAL;
2098 }
2099
2100 /*
2101  *      Drop the packet on the floor
2102  */
2103
2104 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
2105 {
2106         int type;
2107         struct dst_entry *dst = skb_dst(skb);
2108         switch (ipstats_mib_noroutes) {
2109         case IPSTATS_MIB_INNOROUTES:
2110                 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
2111                 if (type == IPV6_ADDR_ANY) {
2112                         IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2113                                       IPSTATS_MIB_INADDRERRORS);
2114                         break;
2115                 }
2116                 /* FALLTHROUGH */
2117         case IPSTATS_MIB_OUTNOROUTES:
2118                 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2119                               ipstats_mib_noroutes);
2120                 break;
2121         }
2122         icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
2123         kfree_skb(skb);
2124         return 0;
2125 }
2126
2127 static int ip6_pkt_discard(struct sk_buff *skb)
2128 {
2129         return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
2130 }
2131
2132 static int ip6_pkt_discard_out(struct sock *sk, struct sk_buff *skb)
2133 {
2134         skb->dev = skb_dst(skb)->dev;
2135         return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
2136 }
2137
2138 static int ip6_pkt_prohibit(struct sk_buff *skb)
2139 {
2140         return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
2141 }
2142
2143 static int ip6_pkt_prohibit_out(struct sock *sk, struct sk_buff *skb)
2144 {
2145         skb->dev = skb_dst(skb)->dev;
2146         return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
2147 }
2148
2149 /*
2150  *      Allocate a dst for local (unicast / anycast) address.
2151  */
2152
2153 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
2154                                     const struct in6_addr *addr,
2155                                     bool anycast)
2156 {
2157         struct net *net = dev_net(idev->dev);
2158         struct rt6_info *rt = ip6_dst_alloc(net, net->loopback_dev,
2159                                             DST_NOCOUNT, NULL);
2160         if (!rt)
2161                 return ERR_PTR(-ENOMEM);
2162
2163         in6_dev_hold(idev);
2164
2165         rt->dst.flags |= DST_HOST;
2166         rt->dst.input = ip6_input;
2167         rt->dst.output = ip6_output;
2168         rt->rt6i_idev = idev;
2169
2170         rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
2171         if (anycast)
2172                 rt->rt6i_flags |= RTF_ANYCAST;
2173         else
2174                 rt->rt6i_flags |= RTF_LOCAL;
2175
2176         rt->rt6i_gateway  = *addr;
2177         rt->rt6i_dst.addr = *addr;
2178         rt->rt6i_dst.plen = 128;
2179         rt->rt6i_table = fib6_get_table(net, RT6_TABLE_LOCAL);
2180
2181         atomic_set(&rt->dst.__refcnt, 1);
2182
2183         return rt;
2184 }
2185
2186 int ip6_route_get_saddr(struct net *net,
2187                         struct rt6_info *rt,
2188                         const struct in6_addr *daddr,
2189                         unsigned int prefs,
2190                         struct in6_addr *saddr)
2191 {
2192         struct inet6_dev *idev = ip6_dst_idev((struct dst_entry*)rt);
2193         int err = 0;
2194         if (rt->rt6i_prefsrc.plen)
2195                 *saddr = rt->rt6i_prefsrc.addr;
2196         else
2197                 err = ipv6_dev_get_saddr(net, idev ? idev->dev : NULL,
2198                                          daddr, prefs, saddr);
2199         return err;
2200 }
2201
2202 /* remove deleted ip from prefsrc entries */
2203 struct arg_dev_net_ip {
2204         struct net_device *dev;
2205         struct net *net;
2206         struct in6_addr *addr;
2207 };
2208
2209 static int fib6_remove_prefsrc(struct rt6_info *rt, void *arg)
2210 {
2211         struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
2212         struct net *net = ((struct arg_dev_net_ip *)arg)->net;
2213         struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
2214
2215         if (((void *)rt->dst.dev == dev || !dev) &&
2216             rt != net->ipv6.ip6_null_entry &&
2217             ipv6_addr_equal(addr, &rt->rt6i_prefsrc.addr)) {
2218                 /* remove prefsrc entry */
2219                 rt->rt6i_prefsrc.plen = 0;
2220         }
2221         return 0;
2222 }
2223
2224 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
2225 {
2226         struct net *net = dev_net(ifp->idev->dev);
2227         struct arg_dev_net_ip adni = {
2228                 .dev = ifp->idev->dev,
2229                 .net = net,
2230                 .addr = &ifp->addr,
2231         };
2232         fib6_clean_all(net, fib6_remove_prefsrc, &adni);
2233 }
2234
2235 struct arg_dev_net {
2236         struct net_device *dev;
2237         struct net *net;
2238 };
2239
2240 static int fib6_ifdown(struct rt6_info *rt, void *arg)
2241 {
2242         const struct arg_dev_net *adn = arg;
2243         const struct net_device *dev = adn->dev;
2244
2245         if ((rt->dst.dev == dev || !dev) &&
2246             rt != adn->net->ipv6.ip6_null_entry)
2247                 return -1;
2248
2249         return 0;
2250 }
2251
2252 void rt6_ifdown(struct net *net, struct net_device *dev)
2253 {
2254         struct arg_dev_net adn = {
2255                 .dev = dev,
2256                 .net = net,
2257         };
2258
2259         fib6_clean_all(net, fib6_ifdown, &adn);
2260         icmp6_clean_all(fib6_ifdown, &adn);
2261 }
2262
2263 struct rt6_mtu_change_arg {
2264         struct net_device *dev;
2265         unsigned int mtu;
2266 };
2267
2268 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2269 {
2270         struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2271         struct inet6_dev *idev;
2272
2273         /* In IPv6 pmtu discovery is not optional,
2274            so that RTAX_MTU lock cannot disable it.
2275            We still use this lock to block changes
2276            caused by addrconf/ndisc.
2277         */
2278
2279         idev = __in6_dev_get(arg->dev);
2280         if (!idev)
2281                 return 0;
2282
2283         /* For administrative MTU increase, there is no way to discover
2284            IPv6 PMTU increase, so PMTU increase should be updated here.
2285            Since RFC 1981 doesn't include administrative MTU increase
2286            update PMTU increase is a MUST. (i.e. jumbo frame)
2287          */
2288         /*
2289            If new MTU is less than route PMTU, this new MTU will be the
2290            lowest MTU in the path, update the route PMTU to reflect PMTU
2291            decreases; if new MTU is greater than route PMTU, and the
2292            old MTU is the lowest MTU in the path, update the route PMTU
2293            to reflect the increase. In this case if the other nodes' MTU
2294            also have the lowest MTU, TOO BIG MESSAGE will be lead to
2295            PMTU discouvery.
2296          */
2297         if (rt->dst.dev == arg->dev &&
2298             !dst_metric_locked(&rt->dst, RTAX_MTU) &&
2299             (dst_mtu(&rt->dst) >= arg->mtu ||
2300              (dst_mtu(&rt->dst) < arg->mtu &&
2301               dst_mtu(&rt->dst) == idev->cnf.mtu6))) {
2302                 dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
2303         }
2304         return 0;
2305 }
2306
2307 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
2308 {
2309         struct rt6_mtu_change_arg arg = {
2310                 .dev = dev,
2311                 .mtu = mtu,
2312         };
2313
2314         fib6_clean_all(dev_net(dev), rt6_mtu_change_route, &arg);
2315 }
2316
2317 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2318         [RTA_GATEWAY]           = { .len = sizeof(struct in6_addr) },
2319         [RTA_OIF]               = { .type = NLA_U32 },
2320         [RTA_IIF]               = { .type = NLA_U32 },
2321         [RTA_PRIORITY]          = { .type = NLA_U32 },
2322         [RTA_METRICS]           = { .type = NLA_NESTED },
2323         [RTA_MULTIPATH]         = { .len = sizeof(struct rtnexthop) },
2324 };
2325
2326 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2327                               struct fib6_config *cfg)
2328 {
2329         struct rtmsg *rtm;
2330         struct nlattr *tb[RTA_MAX+1];
2331         int err;
2332
2333         err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2334         if (err < 0)
2335                 goto errout;
2336
2337         err = -EINVAL;
2338         rtm = nlmsg_data(nlh);
2339         memset(cfg, 0, sizeof(*cfg));
2340
2341         cfg->fc_table = rtm->rtm_table;
2342         cfg->fc_dst_len = rtm->rtm_dst_len;
2343         cfg->fc_src_len = rtm->rtm_src_len;
2344         cfg->fc_flags = RTF_UP;
2345         cfg->fc_protocol = rtm->rtm_protocol;
2346         cfg->fc_type = rtm->rtm_type;
2347
2348         if (rtm->rtm_type == RTN_UNREACHABLE ||
2349             rtm->rtm_type == RTN_BLACKHOLE ||
2350             rtm->rtm_type == RTN_PROHIBIT ||
2351             rtm->rtm_type == RTN_THROW)
2352                 cfg->fc_flags |= RTF_REJECT;
2353
2354         if (rtm->rtm_type == RTN_LOCAL)
2355                 cfg->fc_flags |= RTF_LOCAL;
2356
2357         cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid;
2358         cfg->fc_nlinfo.nlh = nlh;
2359         cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2360
2361         if (tb[RTA_GATEWAY]) {
2362                 nla_memcpy(&cfg->fc_gateway, tb[RTA_GATEWAY], 16);
2363                 cfg->fc_flags |= RTF_GATEWAY;
2364         }
2365
2366         if (tb[RTA_DST]) {
2367                 int plen = (rtm->rtm_dst_len + 7) >> 3;
2368
2369                 if (nla_len(tb[RTA_DST]) < plen)
2370                         goto errout;
2371
2372                 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2373         }
2374
2375         if (tb[RTA_SRC]) {
2376                 int plen = (rtm->rtm_src_len + 7) >> 3;
2377
2378                 if (nla_len(tb[RTA_SRC]) < plen)
2379                         goto errout;
2380
2381                 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2382         }
2383
2384         if (tb[RTA_PREFSRC])
2385                 nla_memcpy(&cfg->fc_prefsrc, tb[RTA_PREFSRC], 16);
2386
2387         if (tb[RTA_OIF])
2388                 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2389
2390         if (tb[RTA_PRIORITY])
2391                 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2392
2393         if (tb[RTA_METRICS]) {
2394                 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2395                 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2396         }
2397
2398         if (tb[RTA_TABLE])
2399                 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2400
2401         if (tb[RTA_MULTIPATH]) {
2402                 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
2403                 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
2404         }
2405
2406         err = 0;
2407 errout:
2408         return err;
2409 }
2410
2411 static int ip6_route_multipath(struct fib6_config *cfg, int add)
2412 {
2413         struct fib6_config r_cfg;
2414         struct rtnexthop *rtnh;
2415         int remaining;
2416         int attrlen;
2417         int err = 0, last_err = 0;
2418
2419 beginning:
2420         rtnh = (struct rtnexthop *)cfg->fc_mp;
2421         remaining = cfg->fc_mp_len;
2422
2423         /* Parse a Multipath Entry */
2424         while (rtnh_ok(rtnh, remaining)) {
2425                 memcpy(&r_cfg, cfg, sizeof(*cfg));
2426                 if (rtnh->rtnh_ifindex)
2427                         r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
2428
2429                 attrlen = rtnh_attrlen(rtnh);
2430                 if (attrlen > 0) {
2431                         struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
2432
2433                         nla = nla_find(attrs, attrlen, RTA_GATEWAY);
2434                         if (nla) {
2435                                 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
2436                                 r_cfg.fc_flags |= RTF_GATEWAY;
2437                         }
2438                 }
2439                 err = add ? ip6_route_add(&r_cfg) : ip6_route_del(&r_cfg);
2440                 if (err) {
2441                         last_err = err;
2442                         /* If we are trying to remove a route, do not stop the
2443                          * loop when ip6_route_del() fails (because next hop is
2444                          * already gone), we should try to remove all next hops.
2445                          */
2446                         if (add) {
2447                                 /* If add fails, we should try to delete all
2448                                  * next hops that have been already added.
2449                                  */
2450                                 add = 0;
2451                                 goto beginning;
2452                         }
2453                 }
2454                 /* Because each route is added like a single route we remove
2455                  * this flag after the first nexthop (if there is a collision,
2456                  * we have already fail to add the first nexthop:
2457                  * fib6_add_rt2node() has reject it).
2458                  */
2459                 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~NLM_F_EXCL;
2460                 rtnh = rtnh_next(rtnh, &remaining);
2461         }
2462
2463         return last_err;
2464 }
2465
2466 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr* nlh)
2467 {
2468         struct fib6_config cfg;
2469         int err;
2470
2471         err = rtm_to_fib6_config(skb, nlh, &cfg);
2472         if (err < 0)
2473                 return err;
2474
2475         if (cfg.fc_mp)
2476                 return ip6_route_multipath(&cfg, 0);
2477         else
2478                 return ip6_route_del(&cfg);
2479 }
2480
2481 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr* nlh)
2482 {
2483         struct fib6_config cfg;
2484         int err;
2485
2486         err = rtm_to_fib6_config(skb, nlh, &cfg);
2487         if (err < 0)
2488                 return err;
2489
2490         if (cfg.fc_mp)
2491                 return ip6_route_multipath(&cfg, 1);
2492         else
2493                 return ip6_route_add(&cfg);
2494 }
2495
2496 static inline size_t rt6_nlmsg_size(void)
2497 {
2498         return NLMSG_ALIGN(sizeof(struct rtmsg))
2499                + nla_total_size(16) /* RTA_SRC */
2500                + nla_total_size(16) /* RTA_DST */
2501                + nla_total_size(16) /* RTA_GATEWAY */
2502                + nla_total_size(16) /* RTA_PREFSRC */
2503                + nla_total_size(4) /* RTA_TABLE */
2504                + nla_total_size(4) /* RTA_IIF */
2505                + nla_total_size(4) /* RTA_OIF */
2506                + nla_total_size(4) /* RTA_PRIORITY */
2507                + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
2508                + nla_total_size(sizeof(struct rta_cacheinfo));
2509 }
2510
2511 static int rt6_fill_node(struct net *net,
2512                          struct sk_buff *skb, struct rt6_info *rt,
2513                          struct in6_addr *dst, struct in6_addr *src,
2514                          int iif, int type, u32 portid, u32 seq,
2515                          int prefix, int nowait, unsigned int flags)
2516 {
2517         struct rtmsg *rtm;
2518         struct nlmsghdr *nlh;
2519         long expires;
2520         u32 table;
2521
2522         if (prefix) {   /* user wants prefix routes only */
2523                 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
2524                         /* success since this is not a prefix route */
2525                         return 1;
2526                 }
2527         }
2528
2529         nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
2530         if (!nlh)
2531                 return -EMSGSIZE;
2532
2533         rtm = nlmsg_data(nlh);
2534         rtm->rtm_family = AF_INET6;
2535         rtm->rtm_dst_len = rt->rt6i_dst.plen;
2536         rtm->rtm_src_len = rt->rt6i_src.plen;
2537         rtm->rtm_tos = 0;
2538         if (rt->rt6i_table)
2539                 table = rt->rt6i_table->tb6_id;
2540         else
2541                 table = RT6_TABLE_UNSPEC;
2542         rtm->rtm_table = table;
2543         if (nla_put_u32(skb, RTA_TABLE, table))
2544                 goto nla_put_failure;
2545         if (rt->rt6i_flags & RTF_REJECT) {
2546                 switch (rt->dst.error) {
2547                 case -EINVAL:
2548                         rtm->rtm_type = RTN_BLACKHOLE;
2549                         break;
2550                 case -EACCES:
2551                         rtm->rtm_type = RTN_PROHIBIT;
2552                         break;
2553                 case -EAGAIN:
2554                         rtm->rtm_type = RTN_THROW;
2555                         break;
2556                 default:
2557                         rtm->rtm_type = RTN_UNREACHABLE;
2558                         break;
2559                 }
2560         }
2561         else if (rt->rt6i_flags & RTF_LOCAL)
2562                 rtm->rtm_type = RTN_LOCAL;
2563         else if (rt->dst.dev && (rt->dst.dev->flags & IFF_LOOPBACK))
2564                 rtm->rtm_type = RTN_LOCAL;
2565         else
2566                 rtm->rtm_type = RTN_UNICAST;
2567         rtm->rtm_flags = 0;
2568         rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2569         rtm->rtm_protocol = rt->rt6i_protocol;
2570         if (rt->rt6i_flags & RTF_DYNAMIC)
2571                 rtm->rtm_protocol = RTPROT_REDIRECT;
2572         else if (rt->rt6i_flags & RTF_ADDRCONF) {
2573                 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ROUTEINFO))
2574                         rtm->rtm_protocol = RTPROT_RA;
2575                 else
2576                         rtm->rtm_protocol = RTPROT_KERNEL;
2577         }
2578
2579         if (rt->rt6i_flags & RTF_CACHE)
2580                 rtm->rtm_flags |= RTM_F_CLONED;
2581
2582         if (dst) {
2583                 if (nla_put(skb, RTA_DST, 16, dst))
2584                         goto nla_put_failure;
2585                 rtm->rtm_dst_len = 128;
2586         } else if (rtm->rtm_dst_len)
2587                 if (nla_put(skb, RTA_DST, 16, &rt->rt6i_dst.addr))
2588                         goto nla_put_failure;
2589 #ifdef CONFIG_IPV6_SUBTREES
2590         if (src) {
2591                 if (nla_put(skb, RTA_SRC, 16, src))
2592                         goto nla_put_failure;
2593                 rtm->rtm_src_len = 128;
2594         } else if (rtm->rtm_src_len &&
2595                    nla_put(skb, RTA_SRC, 16, &rt->rt6i_src.addr))
2596                 goto nla_put_failure;
2597 #endif
2598         if (iif) {
2599 #ifdef CONFIG_IPV6_MROUTE
2600                 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
2601                         int err = ip6mr_get_route(net, skb, rtm, nowait);
2602                         if (err <= 0) {
2603                                 if (!nowait) {
2604                                         if (err == 0)
2605                                                 return 0;
2606                                         goto nla_put_failure;
2607                                 } else {
2608                                         if (err == -EMSGSIZE)
2609                                                 goto nla_put_failure;
2610                                 }
2611                         }
2612                 } else
2613 #endif
2614                         if (nla_put_u32(skb, RTA_IIF, iif))
2615                                 goto nla_put_failure;
2616         } else if (dst) {
2617                 struct in6_addr saddr_buf;
2618                 if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0 &&
2619                     nla_put(skb, RTA_PREFSRC, 16, &saddr_buf))
2620                         goto nla_put_failure;
2621         }
2622
2623         if (rt->rt6i_prefsrc.plen) {
2624                 struct in6_addr saddr_buf;
2625                 saddr_buf = rt->rt6i_prefsrc.addr;
2626                 if (nla_put(skb, RTA_PREFSRC, 16, &saddr_buf))
2627                         goto nla_put_failure;
2628         }
2629
2630         if (rtnetlink_put_metrics(skb, dst_metrics_ptr(&rt->dst)) < 0)
2631                 goto nla_put_failure;
2632
2633         if (rt->rt6i_flags & RTF_GATEWAY) {
2634                 if (nla_put(skb, RTA_GATEWAY, 16, &rt->rt6i_gateway) < 0)
2635                         goto nla_put_failure;
2636         }
2637
2638         if (rt->dst.dev &&
2639             nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex))
2640                 goto nla_put_failure;
2641         if (nla_put_u32(skb, RTA_PRIORITY, rt->rt6i_metric))
2642                 goto nla_put_failure;
2643
2644         expires = (rt->rt6i_flags & RTF_EXPIRES) ? rt->dst.expires - jiffies : 0;
2645
2646         if (rtnl_put_cacheinfo(skb, &rt->dst, 0, expires, rt->dst.error) < 0)
2647                 goto nla_put_failure;
2648
2649         return nlmsg_end(skb, nlh);
2650
2651 nla_put_failure:
2652         nlmsg_cancel(skb, nlh);
2653         return -EMSGSIZE;
2654 }
2655
2656 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
2657 {
2658         struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
2659         int prefix;
2660
2661         if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
2662                 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
2663                 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
2664         } else
2665                 prefix = 0;
2666
2667         return rt6_fill_node(arg->net,
2668                      arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
2669                      NETLINK_CB(arg->cb->skb).portid, arg->cb->nlh->nlmsg_seq,
2670                      prefix, 0, NLM_F_MULTI);
2671 }
2672
2673 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr* nlh)
2674 {
2675         struct net *net = sock_net(in_skb->sk);
2676         struct nlattr *tb[RTA_MAX+1];
2677         struct rt6_info *rt;
2678         struct sk_buff *skb;
2679         struct rtmsg *rtm;
2680         struct flowi6 fl6;
2681         int err, iif = 0, oif = 0;
2682
2683         err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2684         if (err < 0)
2685                 goto errout;
2686
2687         err = -EINVAL;
2688         memset(&fl6, 0, sizeof(fl6));
2689
2690         if (tb[RTA_SRC]) {
2691                 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
2692                         goto errout;
2693
2694                 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
2695         }
2696
2697         if (tb[RTA_DST]) {
2698                 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
2699                         goto errout;
2700
2701                 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
2702         }
2703
2704         if (tb[RTA_IIF])
2705                 iif = nla_get_u32(tb[RTA_IIF]);
2706
2707         if (tb[RTA_OIF])
2708                 oif = nla_get_u32(tb[RTA_OIF]);
2709
2710         if (iif) {
2711                 struct net_device *dev;
2712                 int flags = 0;
2713
2714                 dev = __dev_get_by_index(net, iif);
2715                 if (!dev) {
2716                         err = -ENODEV;
2717                         goto errout;
2718                 }
2719
2720                 fl6.flowi6_iif = iif;
2721
2722                 if (!ipv6_addr_any(&fl6.saddr))
2723                         flags |= RT6_LOOKUP_F_HAS_SADDR;
2724
2725                 rt = (struct rt6_info *)ip6_route_input_lookup(net, dev, &fl6,
2726                                                                flags);
2727         } else {
2728                 fl6.flowi6_oif = oif;
2729
2730                 rt = (struct rt6_info *)ip6_route_output(net, NULL, &fl6);
2731         }
2732
2733         skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
2734         if (!skb) {
2735                 ip6_rt_put(rt);
2736                 err = -ENOBUFS;
2737                 goto errout;
2738         }
2739
2740         /* Reserve room for dummy headers, this skb can pass
2741            through good chunk of routing engine.
2742          */
2743         skb_reset_mac_header(skb);
2744         skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
2745
2746         skb_dst_set(skb, &rt->dst);
2747
2748         err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
2749                             RTM_NEWROUTE, NETLINK_CB(in_skb).portid,
2750                             nlh->nlmsg_seq, 0, 0, 0);
2751         if (err < 0) {
2752                 kfree_skb(skb);
2753                 goto errout;
2754         }
2755
2756         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
2757 errout:
2758         return err;
2759 }
2760
2761 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info)
2762 {
2763         struct sk_buff *skb;
2764         struct net *net = info->nl_net;
2765         u32 seq;
2766         int err;
2767
2768         err = -ENOBUFS;
2769         seq = info->nlh ? info->nlh->nlmsg_seq : 0;
2770
2771         skb = nlmsg_new(rt6_nlmsg_size(), gfp_any());
2772         if (!skb)
2773                 goto errout;
2774
2775         err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
2776                                 event, info->portid, seq, 0, 0, 0);
2777         if (err < 0) {
2778                 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
2779                 WARN_ON(err == -EMSGSIZE);
2780                 kfree_skb(skb);
2781                 goto errout;
2782         }
2783         rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
2784                     info->nlh, gfp_any());
2785         return;
2786 errout:
2787         if (err < 0)
2788                 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
2789 }
2790
2791 static int ip6_route_dev_notify(struct notifier_block *this,
2792                                 unsigned long event, void *ptr)
2793 {
2794         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2795         struct net *net = dev_net(dev);
2796
2797         if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
2798                 net->ipv6.ip6_null_entry->dst.dev = dev;
2799                 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
2800 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2801                 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
2802                 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
2803                 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
2804                 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
2805 #endif
2806         }
2807
2808         return NOTIFY_OK;
2809 }
2810
2811 /*
2812  *      /proc
2813  */
2814
2815 #ifdef CONFIG_PROC_FS
2816
2817 static const struct file_operations ipv6_route_proc_fops = {
2818         .owner          = THIS_MODULE,
2819         .open           = ipv6_route_open,
2820         .read           = seq_read,
2821         .llseek         = seq_lseek,
2822         .release        = seq_release_net,
2823 };
2824
2825 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
2826 {
2827         struct net *net = (struct net *)seq->private;
2828         seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
2829                    net->ipv6.rt6_stats->fib_nodes,
2830                    net->ipv6.rt6_stats->fib_route_nodes,
2831                    net->ipv6.rt6_stats->fib_rt_alloc,
2832                    net->ipv6.rt6_stats->fib_rt_entries,
2833                    net->ipv6.rt6_stats->fib_rt_cache,
2834                    dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
2835                    net->ipv6.rt6_stats->fib_discarded_routes);
2836
2837         return 0;
2838 }
2839
2840 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
2841 {
2842         return single_open_net(inode, file, rt6_stats_seq_show);
2843 }
2844
2845 static const struct file_operations rt6_stats_seq_fops = {
2846         .owner   = THIS_MODULE,
2847         .open    = rt6_stats_seq_open,
2848         .read    = seq_read,
2849         .llseek  = seq_lseek,
2850         .release = single_release_net,
2851 };
2852 #endif  /* CONFIG_PROC_FS */
2853
2854 #ifdef CONFIG_SYSCTL
2855
2856 static
2857 int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
2858                               void __user *buffer, size_t *lenp, loff_t *ppos)
2859 {
2860         struct net *net;
2861         int delay;
2862         if (!write)
2863                 return -EINVAL;
2864
2865         net = (struct net *)ctl->extra1;
2866         delay = net->ipv6.sysctl.flush_delay;
2867         proc_dointvec(ctl, write, buffer, lenp, ppos);
2868         fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
2869         return 0;
2870 }
2871
2872 struct ctl_table ipv6_route_table_template[] = {
2873         {
2874                 .procname       =       "flush",
2875                 .data           =       &init_net.ipv6.sysctl.flush_delay,
2876                 .maxlen         =       sizeof(int),
2877                 .mode           =       0200,
2878                 .proc_handler   =       ipv6_sysctl_rtcache_flush
2879         },
2880         {
2881                 .procname       =       "gc_thresh",
2882                 .data           =       &ip6_dst_ops_template.gc_thresh,
2883                 .maxlen         =       sizeof(int),
2884                 .mode           =       0644,
2885                 .proc_handler   =       proc_dointvec,
2886         },
2887         {
2888                 .procname       =       "max_size",
2889                 .data           =       &init_net.ipv6.sysctl.ip6_rt_max_size,
2890                 .maxlen         =       sizeof(int),
2891                 .mode           =       0644,
2892                 .proc_handler   =       proc_dointvec,
2893         },
2894         {
2895                 .procname       =       "gc_min_interval",
2896                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2897                 .maxlen         =       sizeof(int),
2898                 .mode           =       0644,
2899                 .proc_handler   =       proc_dointvec_jiffies,
2900         },
2901         {
2902                 .procname       =       "gc_timeout",
2903                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
2904                 .maxlen         =       sizeof(int),
2905                 .mode           =       0644,
2906                 .proc_handler   =       proc_dointvec_jiffies,
2907         },
2908         {
2909                 .procname       =       "gc_interval",
2910                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_interval,
2911                 .maxlen         =       sizeof(int),
2912                 .mode           =       0644,
2913                 .proc_handler   =       proc_dointvec_jiffies,
2914         },
2915         {
2916                 .procname       =       "gc_elasticity",
2917                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
2918                 .maxlen         =       sizeof(int),
2919                 .mode           =       0644,
2920                 .proc_handler   =       proc_dointvec,
2921         },
2922         {
2923                 .procname       =       "mtu_expires",
2924                 .data           =       &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
2925                 .maxlen         =       sizeof(int),
2926                 .mode           =       0644,
2927                 .proc_handler   =       proc_dointvec_jiffies,
2928         },
2929         {
2930                 .procname       =       "min_adv_mss",
2931                 .data           =       &init_net.ipv6.sysctl.ip6_rt_min_advmss,
2932                 .maxlen         =       sizeof(int),
2933                 .mode           =       0644,
2934                 .proc_handler   =       proc_dointvec,
2935         },
2936         {
2937                 .procname       =       "gc_min_interval_ms",
2938                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2939                 .maxlen         =       sizeof(int),
2940                 .mode           =       0644,
2941                 .proc_handler   =       proc_dointvec_ms_jiffies,
2942         },
2943         { }
2944 };
2945
2946 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
2947 {
2948         struct ctl_table *table;
2949
2950         table = kmemdup(ipv6_route_table_template,
2951                         sizeof(ipv6_route_table_template),
2952                         GFP_KERNEL);
2953
2954         if (table) {
2955                 table[0].data = &net->ipv6.sysctl.flush_delay;
2956                 table[0].extra1 = net;
2957                 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
2958                 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
2959                 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2960                 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
2961                 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
2962                 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
2963                 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
2964                 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
2965                 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2966
2967                 /* Don't export sysctls to unprivileged users */
2968                 if (net->user_ns != &init_user_ns)
2969                         table[0].procname = NULL;
2970         }
2971
2972         return table;
2973 }
2974 #endif
2975
2976 static int __net_init ip6_route_net_init(struct net *net)
2977 {
2978         int ret = -ENOMEM;
2979
2980         memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
2981                sizeof(net->ipv6.ip6_dst_ops));
2982
2983         if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
2984                 goto out_ip6_dst_ops;
2985
2986         net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
2987                                            sizeof(*net->ipv6.ip6_null_entry),
2988                                            GFP_KERNEL);
2989         if (!net->ipv6.ip6_null_entry)
2990                 goto out_ip6_dst_entries;
2991         net->ipv6.ip6_null_entry->dst.path =
2992                 (struct dst_entry *)net->ipv6.ip6_null_entry;
2993         net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
2994         dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
2995                          ip6_template_metrics, true);
2996
2997 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2998         net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
2999                                                sizeof(*net->ipv6.ip6_prohibit_entry),
3000                                                GFP_KERNEL);
3001         if (!net->ipv6.ip6_prohibit_entry)
3002                 goto out_ip6_null_entry;
3003         net->ipv6.ip6_prohibit_entry->dst.path =
3004                 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
3005         net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3006         dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
3007                          ip6_template_metrics, true);
3008
3009         net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
3010                                                sizeof(*net->ipv6.ip6_blk_hole_entry),
3011                                                GFP_KERNEL);
3012         if (!net->ipv6.ip6_blk_hole_entry)
3013                 goto out_ip6_prohibit_entry;
3014         net->ipv6.ip6_blk_hole_entry->dst.path =
3015                 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
3016         net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3017         dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
3018                          ip6_template_metrics, true);
3019 #endif
3020
3021         net->ipv6.sysctl.flush_delay = 0;
3022         net->ipv6.sysctl.ip6_rt_max_size = 4096;
3023         net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
3024         net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
3025         net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
3026         net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
3027         net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
3028         net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
3029
3030         net->ipv6.ip6_rt_gc_expire = 30*HZ;
3031
3032         ret = 0;
3033 out:
3034         return ret;
3035
3036 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3037 out_ip6_prohibit_entry:
3038         kfree(net->ipv6.ip6_prohibit_entry);
3039 out_ip6_null_entry:
3040         kfree(net->ipv6.ip6_null_entry);
3041 #endif
3042 out_ip6_dst_entries:
3043         dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3044 out_ip6_dst_ops:
3045         goto out;
3046 }
3047
3048 static void __net_exit ip6_route_net_exit(struct net *net)
3049 {
3050         kfree(net->ipv6.ip6_null_entry);
3051 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3052         kfree(net->ipv6.ip6_prohibit_entry);
3053         kfree(net->ipv6.ip6_blk_hole_entry);
3054 #endif
3055         dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3056 }
3057
3058 static int __net_init ip6_route_net_init_late(struct net *net)
3059 {
3060 #ifdef CONFIG_PROC_FS
3061         proc_create("ipv6_route", 0, net->proc_net, &ipv6_route_proc_fops);
3062         proc_create("rt6_stats", S_IRUGO, net->proc_net, &rt6_stats_seq_fops);
3063 #endif
3064         return 0;
3065 }
3066
3067 static void __net_exit ip6_route_net_exit_late(struct net *net)
3068 {
3069 #ifdef CONFIG_PROC_FS
3070         remove_proc_entry("ipv6_route", net->proc_net);
3071         remove_proc_entry("rt6_stats", net->proc_net);
3072 #endif
3073 }
3074
3075 static struct pernet_operations ip6_route_net_ops = {
3076         .init = ip6_route_net_init,
3077         .exit = ip6_route_net_exit,
3078 };
3079
3080 static int __net_init ipv6_inetpeer_init(struct net *net)
3081 {
3082         struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
3083
3084         if (!bp)
3085                 return -ENOMEM;
3086         inet_peer_base_init(bp);
3087         net->ipv6.peers = bp;
3088         return 0;
3089 }
3090
3091 static void __net_exit ipv6_inetpeer_exit(struct net *net)
3092 {
3093         struct inet_peer_base *bp = net->ipv6.peers;
3094
3095         net->ipv6.peers = NULL;
3096         inetpeer_invalidate_tree(bp);
3097         kfree(bp);
3098 }
3099
3100 static struct pernet_operations ipv6_inetpeer_ops = {
3101         .init   =       ipv6_inetpeer_init,
3102         .exit   =       ipv6_inetpeer_exit,
3103 };
3104
3105 static struct pernet_operations ip6_route_net_late_ops = {
3106         .init = ip6_route_net_init_late,
3107         .exit = ip6_route_net_exit_late,
3108 };
3109
3110 static struct notifier_block ip6_route_dev_notifier = {
3111         .notifier_call = ip6_route_dev_notify,
3112         .priority = 0,
3113 };
3114
3115 int __init ip6_route_init(void)
3116 {
3117         int ret;
3118
3119         ret = -ENOMEM;
3120         ip6_dst_ops_template.kmem_cachep =
3121                 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
3122                                   SLAB_HWCACHE_ALIGN, NULL);
3123         if (!ip6_dst_ops_template.kmem_cachep)
3124                 goto out;
3125
3126         ret = dst_entries_init(&ip6_dst_blackhole_ops);
3127         if (ret)
3128                 goto out_kmem_cache;
3129
3130         ret = register_pernet_subsys(&ipv6_inetpeer_ops);
3131         if (ret)
3132                 goto out_dst_entries;
3133
3134         ret = register_pernet_subsys(&ip6_route_net_ops);
3135         if (ret)
3136                 goto out_register_inetpeer;
3137
3138         ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
3139
3140         /* Registering of the loopback is done before this portion of code,
3141          * the loopback reference in rt6_info will not be taken, do it
3142          * manually for init_net */
3143         init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
3144         init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3145   #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3146         init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
3147         init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3148         init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
3149         init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3150   #endif
3151         ret = fib6_init();
3152         if (ret)
3153                 goto out_register_subsys;
3154
3155         ret = xfrm6_init();
3156         if (ret)
3157                 goto out_fib6_init;
3158
3159         ret = fib6_rules_init();
3160         if (ret)
3161                 goto xfrm6_init;
3162
3163         ret = register_pernet_subsys(&ip6_route_net_late_ops);
3164         if (ret)
3165                 goto fib6_rules_init;
3166
3167         ret = -ENOBUFS;
3168         if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL, NULL) ||
3169             __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL, NULL) ||
3170             __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL, NULL))
3171                 goto out_register_late_subsys;
3172
3173         ret = register_netdevice_notifier(&ip6_route_dev_notifier);
3174         if (ret)
3175                 goto out_register_late_subsys;
3176
3177 out:
3178         return ret;
3179
3180 out_register_late_subsys:
3181         unregister_pernet_subsys(&ip6_route_net_late_ops);
3182 fib6_rules_init:
3183         fib6_rules_cleanup();
3184 xfrm6_init:
3185         xfrm6_fini();
3186 out_fib6_init:
3187         fib6_gc_cleanup();
3188 out_register_subsys:
3189         unregister_pernet_subsys(&ip6_route_net_ops);
3190 out_register_inetpeer:
3191         unregister_pernet_subsys(&ipv6_inetpeer_ops);
3192 out_dst_entries:
3193         dst_entries_destroy(&ip6_dst_blackhole_ops);
3194 out_kmem_cache:
3195         kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3196         goto out;
3197 }
3198
3199 void ip6_route_cleanup(void)
3200 {
3201         unregister_netdevice_notifier(&ip6_route_dev_notifier);
3202         unregister_pernet_subsys(&ip6_route_net_late_ops);
3203         fib6_rules_cleanup();
3204         xfrm6_fini();
3205         fib6_gc_cleanup();
3206         unregister_pernet_subsys(&ipv6_inetpeer_ops);
3207         unregister_pernet_subsys(&ip6_route_net_ops);
3208         dst_entries_destroy(&ip6_dst_blackhole_ops);
3209         kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3210 }