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_iif = LOOPBACK_IFINDEX;
1277         fl6.flowi6_oif = oif;
1278         fl6.flowi6_mark = mark;
1279         fl6.daddr = iph->daddr;
1280         fl6.saddr = iph->saddr;
1281         fl6.flowlabel = ip6_flowinfo(iph);
1282
1283         dst = ip6_route_redirect(net, &fl6, &ipv6_hdr(skb)->saddr);
1284         rt6_do_redirect(dst, NULL, skb);
1285         dst_release(dst);
1286 }
1287 EXPORT_SYMBOL_GPL(ip6_redirect);
1288
1289 void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif,
1290                             u32 mark)
1291 {
1292         const struct ipv6hdr *iph = ipv6_hdr(skb);
1293         const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
1294         struct dst_entry *dst;
1295         struct flowi6 fl6;
1296
1297         memset(&fl6, 0, sizeof(fl6));
1298         fl6.flowi6_iif = LOOPBACK_IFINDEX;
1299         fl6.flowi6_oif = oif;
1300         fl6.flowi6_mark = mark;
1301         fl6.daddr = msg->dest;
1302         fl6.saddr = iph->daddr;
1303
1304         dst = ip6_route_redirect(net, &fl6, &iph->saddr);
1305         rt6_do_redirect(dst, NULL, skb);
1306         dst_release(dst);
1307 }
1308
1309 void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
1310 {
1311         ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark);
1312 }
1313 EXPORT_SYMBOL_GPL(ip6_sk_redirect);
1314
1315 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1316 {
1317         struct net_device *dev = dst->dev;
1318         unsigned int mtu = dst_mtu(dst);
1319         struct net *net = dev_net(dev);
1320
1321         mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
1322
1323         if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
1324                 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1325
1326         /*
1327          * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1328          * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1329          * IPV6_MAXPLEN is also valid and means: "any MSS,
1330          * rely only on pmtu discovery"
1331          */
1332         if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
1333                 mtu = IPV6_MAXPLEN;
1334         return mtu;
1335 }
1336
1337 static unsigned int ip6_mtu(const struct dst_entry *dst)
1338 {
1339         struct inet6_dev *idev;
1340         unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
1341
1342         if (mtu)
1343                 goto out;
1344
1345         mtu = IPV6_MIN_MTU;
1346
1347         rcu_read_lock();
1348         idev = __in6_dev_get(dst->dev);
1349         if (idev)
1350                 mtu = idev->cnf.mtu6;
1351         rcu_read_unlock();
1352
1353 out:
1354         return min_t(unsigned int, mtu, IP6_MAX_MTU);
1355 }
1356
1357 static struct dst_entry *icmp6_dst_gc_list;
1358 static DEFINE_SPINLOCK(icmp6_dst_lock);
1359
1360 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
1361                                   struct flowi6 *fl6)
1362 {
1363         struct dst_entry *dst;
1364         struct rt6_info *rt;
1365         struct inet6_dev *idev = in6_dev_get(dev);
1366         struct net *net = dev_net(dev);
1367
1368         if (unlikely(!idev))
1369                 return ERR_PTR(-ENODEV);
1370
1371         rt = ip6_dst_alloc(net, dev, 0, NULL);
1372         if (unlikely(!rt)) {
1373                 in6_dev_put(idev);
1374                 dst = ERR_PTR(-ENOMEM);
1375                 goto out;
1376         }
1377
1378         rt->dst.flags |= DST_HOST;
1379         rt->dst.output  = ip6_output;
1380         atomic_set(&rt->dst.__refcnt, 1);
1381         rt->rt6i_gateway  = fl6->daddr;
1382         rt->rt6i_dst.addr = fl6->daddr;
1383         rt->rt6i_dst.plen = 128;
1384         rt->rt6i_idev     = idev;
1385         dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
1386
1387         spin_lock_bh(&icmp6_dst_lock);
1388         rt->dst.next = icmp6_dst_gc_list;
1389         icmp6_dst_gc_list = &rt->dst;
1390         spin_unlock_bh(&icmp6_dst_lock);
1391
1392         fib6_force_start_gc(net);
1393
1394         dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
1395
1396 out:
1397         return dst;
1398 }
1399
1400 int icmp6_dst_gc(void)
1401 {
1402         struct dst_entry *dst, **pprev;
1403         int more = 0;
1404
1405         spin_lock_bh(&icmp6_dst_lock);
1406         pprev = &icmp6_dst_gc_list;
1407
1408         while ((dst = *pprev) != NULL) {
1409                 if (!atomic_read(&dst->__refcnt)) {
1410                         *pprev = dst->next;
1411                         dst_free(dst);
1412                 } else {
1413                         pprev = &dst->next;
1414                         ++more;
1415                 }
1416         }
1417
1418         spin_unlock_bh(&icmp6_dst_lock);
1419
1420         return more;
1421 }
1422
1423 static void icmp6_clean_all(int (*func)(struct rt6_info *rt, void *arg),
1424                             void *arg)
1425 {
1426         struct dst_entry *dst, **pprev;
1427
1428         spin_lock_bh(&icmp6_dst_lock);
1429         pprev = &icmp6_dst_gc_list;
1430         while ((dst = *pprev) != NULL) {
1431                 struct rt6_info *rt = (struct rt6_info *) dst;
1432                 if (func(rt, arg)) {
1433                         *pprev = dst->next;
1434                         dst_free(dst);
1435                 } else {
1436                         pprev = &dst->next;
1437                 }
1438         }
1439         spin_unlock_bh(&icmp6_dst_lock);
1440 }
1441
1442 static int ip6_dst_gc(struct dst_ops *ops)
1443 {
1444         struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
1445         int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1446         int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1447         int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1448         int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1449         unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1450         int entries;
1451
1452         entries = dst_entries_get_fast(ops);
1453         if (time_after(rt_last_gc + rt_min_interval, jiffies) &&
1454             entries <= rt_max_size)
1455                 goto out;
1456
1457         net->ipv6.ip6_rt_gc_expire++;
1458         fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net, entries > rt_max_size);
1459         entries = dst_entries_get_slow(ops);
1460         if (entries < ops->gc_thresh)
1461                 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1462 out:
1463         net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1464         return entries > rt_max_size;
1465 }
1466
1467 /*
1468  *
1469  */
1470
1471 int ip6_route_add(struct fib6_config *cfg)
1472 {
1473         int err;
1474         struct net *net = cfg->fc_nlinfo.nl_net;
1475         struct rt6_info *rt = NULL;
1476         struct net_device *dev = NULL;
1477         struct inet6_dev *idev = NULL;
1478         struct fib6_table *table;
1479         int addr_type;
1480
1481         if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1482                 return -EINVAL;
1483 #ifndef CONFIG_IPV6_SUBTREES
1484         if (cfg->fc_src_len)
1485                 return -EINVAL;
1486 #endif
1487         if (cfg->fc_ifindex) {
1488                 err = -ENODEV;
1489                 dev = dev_get_by_index(net, cfg->fc_ifindex);
1490                 if (!dev)
1491                         goto out;
1492                 idev = in6_dev_get(dev);
1493                 if (!idev)
1494                         goto out;
1495         }
1496
1497         if (cfg->fc_metric == 0)
1498                 cfg->fc_metric = IP6_RT_PRIO_USER;
1499
1500         err = -ENOBUFS;
1501         if (cfg->fc_nlinfo.nlh &&
1502             !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
1503                 table = fib6_get_table(net, cfg->fc_table);
1504                 if (!table) {
1505                         pr_warn("NLM_F_CREATE should be specified when creating new route\n");
1506                         table = fib6_new_table(net, cfg->fc_table);
1507                 }
1508         } else {
1509                 table = fib6_new_table(net, cfg->fc_table);
1510         }
1511
1512         if (!table)
1513                 goto out;
1514
1515         rt = ip6_dst_alloc(net, NULL, (cfg->fc_flags & RTF_ADDRCONF) ? 0 : DST_NOCOUNT, table);
1516
1517         if (!rt) {
1518                 err = -ENOMEM;
1519                 goto out;
1520         }
1521
1522         if (cfg->fc_flags & RTF_EXPIRES)
1523                 rt6_set_expires(rt, jiffies +
1524                                 clock_t_to_jiffies(cfg->fc_expires));
1525         else
1526                 rt6_clean_expires(rt);
1527
1528         if (cfg->fc_protocol == RTPROT_UNSPEC)
1529                 cfg->fc_protocol = RTPROT_BOOT;
1530         rt->rt6i_protocol = cfg->fc_protocol;
1531
1532         addr_type = ipv6_addr_type(&cfg->fc_dst);
1533
1534         if (addr_type & IPV6_ADDR_MULTICAST)
1535                 rt->dst.input = ip6_mc_input;
1536         else if (cfg->fc_flags & RTF_LOCAL)
1537                 rt->dst.input = ip6_input;
1538         else
1539                 rt->dst.input = ip6_forward;
1540
1541         rt->dst.output = ip6_output;
1542
1543         ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1544         rt->rt6i_dst.plen = cfg->fc_dst_len;
1545         if (rt->rt6i_dst.plen == 128) {
1546                 rt->dst.flags |= DST_HOST;
1547                 dst_metrics_set_force_overwrite(&rt->dst);
1548         }
1549
1550 #ifdef CONFIG_IPV6_SUBTREES
1551         ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1552         rt->rt6i_src.plen = cfg->fc_src_len;
1553 #endif
1554
1555         rt->rt6i_metric = cfg->fc_metric;
1556
1557         /* We cannot add true routes via loopback here,
1558            they would result in kernel looping; promote them to reject routes
1559          */
1560         if ((cfg->fc_flags & RTF_REJECT) ||
1561             (dev && (dev->flags & IFF_LOOPBACK) &&
1562              !(addr_type & IPV6_ADDR_LOOPBACK) &&
1563              !(cfg->fc_flags & RTF_LOCAL))) {
1564                 /* hold loopback dev/idev if we haven't done so. */
1565                 if (dev != net->loopback_dev) {
1566                         if (dev) {
1567                                 dev_put(dev);
1568                                 in6_dev_put(idev);
1569                         }
1570                         dev = net->loopback_dev;
1571                         dev_hold(dev);
1572                         idev = in6_dev_get(dev);
1573                         if (!idev) {
1574                                 err = -ENODEV;
1575                                 goto out;
1576                         }
1577                 }
1578                 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1579                 switch (cfg->fc_type) {
1580                 case RTN_BLACKHOLE:
1581                         rt->dst.error = -EINVAL;
1582                         rt->dst.output = dst_discard_sk;
1583                         rt->dst.input = dst_discard;
1584                         break;
1585                 case RTN_PROHIBIT:
1586                         rt->dst.error = -EACCES;
1587                         rt->dst.output = ip6_pkt_prohibit_out;
1588                         rt->dst.input = ip6_pkt_prohibit;
1589                         break;
1590                 case RTN_THROW:
1591                 default:
1592                         rt->dst.error = (cfg->fc_type == RTN_THROW) ? -EAGAIN
1593                                         : -ENETUNREACH;
1594                         rt->dst.output = ip6_pkt_discard_out;
1595                         rt->dst.input = ip6_pkt_discard;
1596                         break;
1597                 }
1598                 goto install_route;
1599         }
1600
1601         if (cfg->fc_flags & RTF_GATEWAY) {
1602                 const struct in6_addr *gw_addr;
1603                 int gwa_type;
1604
1605                 gw_addr = &cfg->fc_gateway;
1606                 rt->rt6i_gateway = *gw_addr;
1607                 gwa_type = ipv6_addr_type(gw_addr);
1608
1609                 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1610                         struct rt6_info *grt;
1611
1612                         /* IPv6 strictly inhibits using not link-local
1613                            addresses as nexthop address.
1614                            Otherwise, router will not able to send redirects.
1615                            It is very good, but in some (rare!) circumstances
1616                            (SIT, PtP, NBMA NOARP links) it is handy to allow
1617                            some exceptions. --ANK
1618                          */
1619                         err = -EINVAL;
1620                         if (!(gwa_type & IPV6_ADDR_UNICAST))
1621                                 goto out;
1622
1623                         grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, 1);
1624
1625                         err = -EHOSTUNREACH;
1626                         if (!grt)
1627                                 goto out;
1628                         if (dev) {
1629                                 if (dev != grt->dst.dev) {
1630                                         ip6_rt_put(grt);
1631                                         goto out;
1632                                 }
1633                         } else {
1634                                 dev = grt->dst.dev;
1635                                 idev = grt->rt6i_idev;
1636                                 dev_hold(dev);
1637                                 in6_dev_hold(grt->rt6i_idev);
1638                         }
1639                         if (!(grt->rt6i_flags & RTF_GATEWAY))
1640                                 err = 0;
1641                         ip6_rt_put(grt);
1642
1643                         if (err)
1644                                 goto out;
1645                 }
1646                 err = -EINVAL;
1647                 if (!dev || (dev->flags & IFF_LOOPBACK))
1648                         goto out;
1649         }
1650
1651         err = -ENODEV;
1652         if (!dev)
1653                 goto out;
1654
1655         if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
1656                 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
1657                         err = -EINVAL;
1658                         goto out;
1659                 }
1660                 rt->rt6i_prefsrc.addr = cfg->fc_prefsrc;
1661                 rt->rt6i_prefsrc.plen = 128;
1662         } else
1663                 rt->rt6i_prefsrc.plen = 0;
1664
1665         rt->rt6i_flags = cfg->fc_flags;
1666
1667 install_route:
1668         rt->dst.dev = dev;
1669         rt->rt6i_idev = idev;
1670         rt->rt6i_table = table;
1671
1672         cfg->fc_nlinfo.nl_net = dev_net(dev);
1673
1674         return __ip6_ins_rt(rt, &cfg->fc_nlinfo, cfg->fc_mx, cfg->fc_mx_len);
1675
1676 out:
1677         if (dev)
1678                 dev_put(dev);
1679         if (idev)
1680                 in6_dev_put(idev);
1681         if (rt)
1682                 dst_free(&rt->dst);
1683         return err;
1684 }
1685
1686 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
1687 {
1688         int err;
1689         struct fib6_table *table;
1690         struct net *net = dev_net(rt->dst.dev);
1691
1692         if (rt == net->ipv6.ip6_null_entry) {
1693                 err = -ENOENT;
1694                 goto out;
1695         }
1696
1697         table = rt->rt6i_table;
1698         write_lock_bh(&table->tb6_lock);
1699         err = fib6_del(rt, info);
1700         write_unlock_bh(&table->tb6_lock);
1701
1702 out:
1703         ip6_rt_put(rt);
1704         return err;
1705 }
1706
1707 int ip6_del_rt(struct rt6_info *rt)
1708 {
1709         struct nl_info info = {
1710                 .nl_net = dev_net(rt->dst.dev),
1711         };
1712         return __ip6_del_rt(rt, &info);
1713 }
1714
1715 static int ip6_route_del(struct fib6_config *cfg)
1716 {
1717         struct fib6_table *table;
1718         struct fib6_node *fn;
1719         struct rt6_info *rt;
1720         int err = -ESRCH;
1721
1722         table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
1723         if (!table)
1724                 return err;
1725
1726         read_lock_bh(&table->tb6_lock);
1727
1728         fn = fib6_locate(&table->tb6_root,
1729                          &cfg->fc_dst, cfg->fc_dst_len,
1730                          &cfg->fc_src, cfg->fc_src_len);
1731
1732         if (fn) {
1733                 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1734                         if (cfg->fc_ifindex &&
1735                             (!rt->dst.dev ||
1736                              rt->dst.dev->ifindex != cfg->fc_ifindex))
1737                                 continue;
1738                         if (cfg->fc_flags & RTF_GATEWAY &&
1739                             !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
1740                                 continue;
1741                         if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
1742                                 continue;
1743                         dst_hold(&rt->dst);
1744                         read_unlock_bh(&table->tb6_lock);
1745
1746                         return __ip6_del_rt(rt, &cfg->fc_nlinfo);
1747                 }
1748         }
1749         read_unlock_bh(&table->tb6_lock);
1750
1751         return err;
1752 }
1753
1754 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
1755 {
1756         struct net *net = dev_net(skb->dev);
1757         struct netevent_redirect netevent;
1758         struct rt6_info *rt, *nrt = NULL;
1759         struct ndisc_options ndopts;
1760         struct inet6_dev *in6_dev;
1761         struct neighbour *neigh;
1762         struct rd_msg *msg;
1763         int optlen, on_link;
1764         u8 *lladdr;
1765
1766         optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
1767         optlen -= sizeof(*msg);
1768
1769         if (optlen < 0) {
1770                 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
1771                 return;
1772         }
1773
1774         msg = (struct rd_msg *)icmp6_hdr(skb);
1775
1776         if (ipv6_addr_is_multicast(&msg->dest)) {
1777                 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
1778                 return;
1779         }
1780
1781         on_link = 0;
1782         if (ipv6_addr_equal(&msg->dest, &msg->target)) {
1783                 on_link = 1;
1784         } else if (ipv6_addr_type(&msg->target) !=
1785                    (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
1786                 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
1787                 return;
1788         }
1789
1790         in6_dev = __in6_dev_get(skb->dev);
1791         if (!in6_dev)
1792                 return;
1793         if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
1794                 return;
1795
1796         /* RFC2461 8.1:
1797          *      The IP source address of the Redirect MUST be the same as the current
1798          *      first-hop router for the specified ICMP Destination Address.
1799          */
1800
1801         if (!ndisc_parse_options(msg->opt, optlen, &ndopts)) {
1802                 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
1803                 return;
1804         }
1805
1806         lladdr = NULL;
1807         if (ndopts.nd_opts_tgt_lladdr) {
1808                 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
1809                                              skb->dev);
1810                 if (!lladdr) {
1811                         net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
1812                         return;
1813                 }
1814         }
1815
1816         rt = (struct rt6_info *) dst;
1817         if (rt == net->ipv6.ip6_null_entry) {
1818                 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
1819                 return;
1820         }
1821
1822         /* Redirect received -> path was valid.
1823          * Look, redirects are sent only in response to data packets,
1824          * so that this nexthop apparently is reachable. --ANK
1825          */
1826         dst_confirm(&rt->dst);
1827
1828         neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
1829         if (!neigh)
1830                 return;
1831
1832         /*
1833          *      We have finally decided to accept it.
1834          */
1835
1836         neigh_update(neigh, lladdr, NUD_STALE,
1837                      NEIGH_UPDATE_F_WEAK_OVERRIDE|
1838                      NEIGH_UPDATE_F_OVERRIDE|
1839                      (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1840                                      NEIGH_UPDATE_F_ISROUTER))
1841                      );
1842
1843         nrt = ip6_rt_copy(rt, &msg->dest);
1844         if (!nrt)
1845                 goto out;
1846
1847         nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
1848         if (on_link)
1849                 nrt->rt6i_flags &= ~RTF_GATEWAY;
1850
1851         nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
1852
1853         if (ip6_ins_rt(nrt))
1854                 goto out;
1855
1856         netevent.old = &rt->dst;
1857         netevent.new = &nrt->dst;
1858         netevent.daddr = &msg->dest;
1859         netevent.neigh = neigh;
1860         call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
1861
1862         if (rt->rt6i_flags & RTF_CACHE) {
1863                 rt = (struct rt6_info *) dst_clone(&rt->dst);
1864                 ip6_del_rt(rt);
1865         }
1866
1867 out:
1868         neigh_release(neigh);
1869 }
1870
1871 /*
1872  *      Misc support functions
1873  */
1874
1875 static struct rt6_info *ip6_rt_copy(struct rt6_info *ort,
1876                                     const struct in6_addr *dest)
1877 {
1878         struct net *net = dev_net(ort->dst.dev);
1879         struct rt6_info *rt = ip6_dst_alloc(net, ort->dst.dev, 0,
1880                                             ort->rt6i_table);
1881
1882         if (rt) {
1883                 rt->dst.input = ort->dst.input;
1884                 rt->dst.output = ort->dst.output;
1885                 rt->dst.flags |= DST_HOST;
1886
1887                 rt->rt6i_dst.addr = *dest;
1888                 rt->rt6i_dst.plen = 128;
1889                 dst_copy_metrics(&rt->dst, &ort->dst);
1890                 rt->dst.error = ort->dst.error;
1891                 rt->rt6i_idev = ort->rt6i_idev;
1892                 if (rt->rt6i_idev)
1893                         in6_dev_hold(rt->rt6i_idev);
1894                 rt->dst.lastuse = jiffies;
1895
1896                 if (ort->rt6i_flags & RTF_GATEWAY)
1897                         rt->rt6i_gateway = ort->rt6i_gateway;
1898                 else
1899                         rt->rt6i_gateway = *dest;
1900                 rt->rt6i_flags = ort->rt6i_flags;
1901                 rt6_set_from(rt, ort);
1902                 rt->rt6i_metric = 0;
1903
1904 #ifdef CONFIG_IPV6_SUBTREES
1905                 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1906 #endif
1907                 memcpy(&rt->rt6i_prefsrc, &ort->rt6i_prefsrc, sizeof(struct rt6key));
1908                 rt->rt6i_table = ort->rt6i_table;
1909         }
1910         return rt;
1911 }
1912
1913 #ifdef CONFIG_IPV6_ROUTE_INFO
1914 static struct rt6_info *rt6_get_route_info(struct net *net,
1915                                            const struct in6_addr *prefix, int prefixlen,
1916                                            const struct in6_addr *gwaddr, int ifindex)
1917 {
1918         struct fib6_node *fn;
1919         struct rt6_info *rt = NULL;
1920         struct fib6_table *table;
1921
1922         table = fib6_get_table(net, RT6_TABLE_INFO);
1923         if (!table)
1924                 return NULL;
1925
1926         read_lock_bh(&table->tb6_lock);
1927         fn = fib6_locate(&table->tb6_root, prefix ,prefixlen, NULL, 0);
1928         if (!fn)
1929                 goto out;
1930
1931         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1932                 if (rt->dst.dev->ifindex != ifindex)
1933                         continue;
1934                 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
1935                         continue;
1936                 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
1937                         continue;
1938                 dst_hold(&rt->dst);
1939                 break;
1940         }
1941 out:
1942         read_unlock_bh(&table->tb6_lock);
1943         return rt;
1944 }
1945
1946 static struct rt6_info *rt6_add_route_info(struct net *net,
1947                                            const struct in6_addr *prefix, int prefixlen,
1948                                            const struct in6_addr *gwaddr, int ifindex,
1949                                            unsigned int pref)
1950 {
1951         struct fib6_config cfg = {
1952                 .fc_table       = RT6_TABLE_INFO,
1953                 .fc_metric      = IP6_RT_PRIO_USER,
1954                 .fc_ifindex     = ifindex,
1955                 .fc_dst_len     = prefixlen,
1956                 .fc_flags       = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
1957                                   RTF_UP | RTF_PREF(pref),
1958                 .fc_nlinfo.portid = 0,
1959                 .fc_nlinfo.nlh = NULL,
1960                 .fc_nlinfo.nl_net = net,
1961         };
1962
1963         cfg.fc_dst = *prefix;
1964         cfg.fc_gateway = *gwaddr;
1965
1966         /* We should treat it as a default route if prefix length is 0. */
1967         if (!prefixlen)
1968                 cfg.fc_flags |= RTF_DEFAULT;
1969
1970         ip6_route_add(&cfg);
1971
1972         return rt6_get_route_info(net, prefix, prefixlen, gwaddr, ifindex);
1973 }
1974 #endif
1975
1976 struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
1977 {
1978         struct rt6_info *rt;
1979         struct fib6_table *table;
1980
1981         table = fib6_get_table(dev_net(dev), RT6_TABLE_DFLT);
1982         if (!table)
1983                 return NULL;
1984
1985         read_lock_bh(&table->tb6_lock);
1986         for (rt = table->tb6_root.leaf; rt; rt=rt->dst.rt6_next) {
1987                 if (dev == rt->dst.dev &&
1988                     ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
1989                     ipv6_addr_equal(&rt->rt6i_gateway, addr))
1990                         break;
1991         }
1992         if (rt)
1993                 dst_hold(&rt->dst);
1994         read_unlock_bh(&table->tb6_lock);
1995         return rt;
1996 }
1997
1998 struct rt6_info *rt6_add_dflt_router(const struct in6_addr *gwaddr,
1999                                      struct net_device *dev,
2000                                      unsigned int pref)
2001 {
2002         struct fib6_config cfg = {
2003                 .fc_table       = RT6_TABLE_DFLT,
2004                 .fc_metric      = IP6_RT_PRIO_USER,
2005                 .fc_ifindex     = dev->ifindex,
2006                 .fc_flags       = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
2007                                   RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
2008                 .fc_nlinfo.portid = 0,
2009                 .fc_nlinfo.nlh = NULL,
2010                 .fc_nlinfo.nl_net = dev_net(dev),
2011         };
2012
2013         cfg.fc_gateway = *gwaddr;
2014
2015         ip6_route_add(&cfg);
2016
2017         return rt6_get_dflt_router(gwaddr, dev);
2018 }
2019
2020 void rt6_purge_dflt_routers(struct net *net)
2021 {
2022         struct rt6_info *rt;
2023         struct fib6_table *table;
2024
2025         /* NOTE: Keep consistent with rt6_get_dflt_router */
2026         table = fib6_get_table(net, RT6_TABLE_DFLT);
2027         if (!table)
2028                 return;
2029
2030 restart:
2031         read_lock_bh(&table->tb6_lock);
2032         for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2033                 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
2034                     (!rt->rt6i_idev || rt->rt6i_idev->cnf.accept_ra != 2)) {
2035                         dst_hold(&rt->dst);
2036                         read_unlock_bh(&table->tb6_lock);
2037                         ip6_del_rt(rt);
2038                         goto restart;
2039                 }
2040         }
2041         read_unlock_bh(&table->tb6_lock);
2042 }
2043
2044 static void rtmsg_to_fib6_config(struct net *net,
2045                                  struct in6_rtmsg *rtmsg,
2046                                  struct fib6_config *cfg)
2047 {
2048         memset(cfg, 0, sizeof(*cfg));
2049
2050         cfg->fc_table = RT6_TABLE_MAIN;
2051         cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
2052         cfg->fc_metric = rtmsg->rtmsg_metric;
2053         cfg->fc_expires = rtmsg->rtmsg_info;
2054         cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
2055         cfg->fc_src_len = rtmsg->rtmsg_src_len;
2056         cfg->fc_flags = rtmsg->rtmsg_flags;
2057
2058         cfg->fc_nlinfo.nl_net = net;
2059
2060         cfg->fc_dst = rtmsg->rtmsg_dst;
2061         cfg->fc_src = rtmsg->rtmsg_src;
2062         cfg->fc_gateway = rtmsg->rtmsg_gateway;
2063 }
2064
2065 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
2066 {
2067         struct fib6_config cfg;
2068         struct in6_rtmsg rtmsg;
2069         int err;
2070
2071         switch(cmd) {
2072         case SIOCADDRT:         /* Add a route */
2073         case SIOCDELRT:         /* Delete a route */
2074                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2075                         return -EPERM;
2076                 err = copy_from_user(&rtmsg, arg,
2077                                      sizeof(struct in6_rtmsg));
2078                 if (err)
2079                         return -EFAULT;
2080
2081                 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
2082
2083                 rtnl_lock();
2084                 switch (cmd) {
2085                 case SIOCADDRT:
2086                         err = ip6_route_add(&cfg);
2087                         break;
2088                 case SIOCDELRT:
2089                         err = ip6_route_del(&cfg);
2090                         break;
2091                 default:
2092                         err = -EINVAL;
2093                 }
2094                 rtnl_unlock();
2095
2096                 return err;
2097         }
2098
2099         return -EINVAL;
2100 }
2101
2102 /*
2103  *      Drop the packet on the floor
2104  */
2105
2106 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
2107 {
2108         int type;
2109         struct dst_entry *dst = skb_dst(skb);
2110         switch (ipstats_mib_noroutes) {
2111         case IPSTATS_MIB_INNOROUTES:
2112                 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
2113                 if (type == IPV6_ADDR_ANY) {
2114                         IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2115                                       IPSTATS_MIB_INADDRERRORS);
2116                         break;
2117                 }
2118                 /* FALLTHROUGH */
2119         case IPSTATS_MIB_OUTNOROUTES:
2120                 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2121                               ipstats_mib_noroutes);
2122                 break;
2123         }
2124         icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
2125         kfree_skb(skb);
2126         return 0;
2127 }
2128
2129 static int ip6_pkt_discard(struct sk_buff *skb)
2130 {
2131         return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
2132 }
2133
2134 static int ip6_pkt_discard_out(struct sock *sk, struct sk_buff *skb)
2135 {
2136         skb->dev = skb_dst(skb)->dev;
2137         return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
2138 }
2139
2140 static int ip6_pkt_prohibit(struct sk_buff *skb)
2141 {
2142         return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
2143 }
2144
2145 static int ip6_pkt_prohibit_out(struct sock *sk, struct sk_buff *skb)
2146 {
2147         skb->dev = skb_dst(skb)->dev;
2148         return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
2149 }
2150
2151 /*
2152  *      Allocate a dst for local (unicast / anycast) address.
2153  */
2154
2155 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
2156                                     const struct in6_addr *addr,
2157                                     bool anycast)
2158 {
2159         struct net *net = dev_net(idev->dev);
2160         struct rt6_info *rt = ip6_dst_alloc(net, net->loopback_dev,
2161                                             DST_NOCOUNT, NULL);
2162         if (!rt)
2163                 return ERR_PTR(-ENOMEM);
2164
2165         in6_dev_hold(idev);
2166
2167         rt->dst.flags |= DST_HOST;
2168         rt->dst.input = ip6_input;
2169         rt->dst.output = ip6_output;
2170         rt->rt6i_idev = idev;
2171
2172         rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
2173         if (anycast)
2174                 rt->rt6i_flags |= RTF_ANYCAST;
2175         else
2176                 rt->rt6i_flags |= RTF_LOCAL;
2177
2178         rt->rt6i_gateway  = *addr;
2179         rt->rt6i_dst.addr = *addr;
2180         rt->rt6i_dst.plen = 128;
2181         rt->rt6i_table = fib6_get_table(net, RT6_TABLE_LOCAL);
2182
2183         atomic_set(&rt->dst.__refcnt, 1);
2184
2185         return rt;
2186 }
2187
2188 int ip6_route_get_saddr(struct net *net,
2189                         struct rt6_info *rt,
2190                         const struct in6_addr *daddr,
2191                         unsigned int prefs,
2192                         struct in6_addr *saddr)
2193 {
2194         struct inet6_dev *idev = ip6_dst_idev((struct dst_entry*)rt);
2195         int err = 0;
2196         if (rt->rt6i_prefsrc.plen)
2197                 *saddr = rt->rt6i_prefsrc.addr;
2198         else
2199                 err = ipv6_dev_get_saddr(net, idev ? idev->dev : NULL,
2200                                          daddr, prefs, saddr);
2201         return err;
2202 }
2203
2204 /* remove deleted ip from prefsrc entries */
2205 struct arg_dev_net_ip {
2206         struct net_device *dev;
2207         struct net *net;
2208         struct in6_addr *addr;
2209 };
2210
2211 static int fib6_remove_prefsrc(struct rt6_info *rt, void *arg)
2212 {
2213         struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
2214         struct net *net = ((struct arg_dev_net_ip *)arg)->net;
2215         struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
2216
2217         if (((void *)rt->dst.dev == dev || !dev) &&
2218             rt != net->ipv6.ip6_null_entry &&
2219             ipv6_addr_equal(addr, &rt->rt6i_prefsrc.addr)) {
2220                 /* remove prefsrc entry */
2221                 rt->rt6i_prefsrc.plen = 0;
2222         }
2223         return 0;
2224 }
2225
2226 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
2227 {
2228         struct net *net = dev_net(ifp->idev->dev);
2229         struct arg_dev_net_ip adni = {
2230                 .dev = ifp->idev->dev,
2231                 .net = net,
2232                 .addr = &ifp->addr,
2233         };
2234         fib6_clean_all(net, fib6_remove_prefsrc, &adni);
2235 }
2236
2237 struct arg_dev_net {
2238         struct net_device *dev;
2239         struct net *net;
2240 };
2241
2242 static int fib6_ifdown(struct rt6_info *rt, void *arg)
2243 {
2244         const struct arg_dev_net *adn = arg;
2245         const struct net_device *dev = adn->dev;
2246
2247         if ((rt->dst.dev == dev || !dev) &&
2248             rt != adn->net->ipv6.ip6_null_entry)
2249                 return -1;
2250
2251         return 0;
2252 }
2253
2254 void rt6_ifdown(struct net *net, struct net_device *dev)
2255 {
2256         struct arg_dev_net adn = {
2257                 .dev = dev,
2258                 .net = net,
2259         };
2260
2261         fib6_clean_all(net, fib6_ifdown, &adn);
2262         icmp6_clean_all(fib6_ifdown, &adn);
2263 }
2264
2265 struct rt6_mtu_change_arg {
2266         struct net_device *dev;
2267         unsigned int mtu;
2268 };
2269
2270 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2271 {
2272         struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2273         struct inet6_dev *idev;
2274
2275         /* In IPv6 pmtu discovery is not optional,
2276            so that RTAX_MTU lock cannot disable it.
2277            We still use this lock to block changes
2278            caused by addrconf/ndisc.
2279         */
2280
2281         idev = __in6_dev_get(arg->dev);
2282         if (!idev)
2283                 return 0;
2284
2285         /* For administrative MTU increase, there is no way to discover
2286            IPv6 PMTU increase, so PMTU increase should be updated here.
2287            Since RFC 1981 doesn't include administrative MTU increase
2288            update PMTU increase is a MUST. (i.e. jumbo frame)
2289          */
2290         /*
2291            If new MTU is less than route PMTU, this new MTU will be the
2292            lowest MTU in the path, update the route PMTU to reflect PMTU
2293            decreases; if new MTU is greater than route PMTU, and the
2294            old MTU is the lowest MTU in the path, update the route PMTU
2295            to reflect the increase. In this case if the other nodes' MTU
2296            also have the lowest MTU, TOO BIG MESSAGE will be lead to
2297            PMTU discouvery.
2298          */
2299         if (rt->dst.dev == arg->dev &&
2300             !dst_metric_locked(&rt->dst, RTAX_MTU) &&
2301             (dst_mtu(&rt->dst) >= arg->mtu ||
2302              (dst_mtu(&rt->dst) < arg->mtu &&
2303               dst_mtu(&rt->dst) == idev->cnf.mtu6))) {
2304                 dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
2305         }
2306         return 0;
2307 }
2308
2309 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
2310 {
2311         struct rt6_mtu_change_arg arg = {
2312                 .dev = dev,
2313                 .mtu = mtu,
2314         };
2315
2316         fib6_clean_all(dev_net(dev), rt6_mtu_change_route, &arg);
2317 }
2318
2319 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2320         [RTA_GATEWAY]           = { .len = sizeof(struct in6_addr) },
2321         [RTA_OIF]               = { .type = NLA_U32 },
2322         [RTA_IIF]               = { .type = NLA_U32 },
2323         [RTA_PRIORITY]          = { .type = NLA_U32 },
2324         [RTA_METRICS]           = { .type = NLA_NESTED },
2325         [RTA_MULTIPATH]         = { .len = sizeof(struct rtnexthop) },
2326 };
2327
2328 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2329                               struct fib6_config *cfg)
2330 {
2331         struct rtmsg *rtm;
2332         struct nlattr *tb[RTA_MAX+1];
2333         int err;
2334
2335         err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2336         if (err < 0)
2337                 goto errout;
2338
2339         err = -EINVAL;
2340         rtm = nlmsg_data(nlh);
2341         memset(cfg, 0, sizeof(*cfg));
2342
2343         cfg->fc_table = rtm->rtm_table;
2344         cfg->fc_dst_len = rtm->rtm_dst_len;
2345         cfg->fc_src_len = rtm->rtm_src_len;
2346         cfg->fc_flags = RTF_UP;
2347         cfg->fc_protocol = rtm->rtm_protocol;
2348         cfg->fc_type = rtm->rtm_type;
2349
2350         if (rtm->rtm_type == RTN_UNREACHABLE ||
2351             rtm->rtm_type == RTN_BLACKHOLE ||
2352             rtm->rtm_type == RTN_PROHIBIT ||
2353             rtm->rtm_type == RTN_THROW)
2354                 cfg->fc_flags |= RTF_REJECT;
2355
2356         if (rtm->rtm_type == RTN_LOCAL)
2357                 cfg->fc_flags |= RTF_LOCAL;
2358
2359         cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid;
2360         cfg->fc_nlinfo.nlh = nlh;
2361         cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2362
2363         if (tb[RTA_GATEWAY]) {
2364                 nla_memcpy(&cfg->fc_gateway, tb[RTA_GATEWAY], 16);
2365                 cfg->fc_flags |= RTF_GATEWAY;
2366         }
2367
2368         if (tb[RTA_DST]) {
2369                 int plen = (rtm->rtm_dst_len + 7) >> 3;
2370
2371                 if (nla_len(tb[RTA_DST]) < plen)
2372                         goto errout;
2373
2374                 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2375         }
2376
2377         if (tb[RTA_SRC]) {
2378                 int plen = (rtm->rtm_src_len + 7) >> 3;
2379
2380                 if (nla_len(tb[RTA_SRC]) < plen)
2381                         goto errout;
2382
2383                 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2384         }
2385
2386         if (tb[RTA_PREFSRC])
2387                 nla_memcpy(&cfg->fc_prefsrc, tb[RTA_PREFSRC], 16);
2388
2389         if (tb[RTA_OIF])
2390                 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2391
2392         if (tb[RTA_PRIORITY])
2393                 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2394
2395         if (tb[RTA_METRICS]) {
2396                 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2397                 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2398         }
2399
2400         if (tb[RTA_TABLE])
2401                 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2402
2403         if (tb[RTA_MULTIPATH]) {
2404                 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
2405                 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
2406         }
2407
2408         err = 0;
2409 errout:
2410         return err;
2411 }
2412
2413 static int ip6_route_multipath(struct fib6_config *cfg, int add)
2414 {
2415         struct fib6_config r_cfg;
2416         struct rtnexthop *rtnh;
2417         int remaining;
2418         int attrlen;
2419         int err = 0, last_err = 0;
2420
2421 beginning:
2422         rtnh = (struct rtnexthop *)cfg->fc_mp;
2423         remaining = cfg->fc_mp_len;
2424
2425         /* Parse a Multipath Entry */
2426         while (rtnh_ok(rtnh, remaining)) {
2427                 memcpy(&r_cfg, cfg, sizeof(*cfg));
2428                 if (rtnh->rtnh_ifindex)
2429                         r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
2430
2431                 attrlen = rtnh_attrlen(rtnh);
2432                 if (attrlen > 0) {
2433                         struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
2434
2435                         nla = nla_find(attrs, attrlen, RTA_GATEWAY);
2436                         if (nla) {
2437                                 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
2438                                 r_cfg.fc_flags |= RTF_GATEWAY;
2439                         }
2440                 }
2441                 err = add ? ip6_route_add(&r_cfg) : ip6_route_del(&r_cfg);
2442                 if (err) {
2443                         last_err = err;
2444                         /* If we are trying to remove a route, do not stop the
2445                          * loop when ip6_route_del() fails (because next hop is
2446                          * already gone), we should try to remove all next hops.
2447                          */
2448                         if (add) {
2449                                 /* If add fails, we should try to delete all
2450                                  * next hops that have been already added.
2451                                  */
2452                                 add = 0;
2453                                 goto beginning;
2454                         }
2455                 }
2456                 /* Because each route is added like a single route we remove
2457                  * this flag after the first nexthop (if there is a collision,
2458                  * we have already fail to add the first nexthop:
2459                  * fib6_add_rt2node() has reject it).
2460                  */
2461                 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~NLM_F_EXCL;
2462                 rtnh = rtnh_next(rtnh, &remaining);
2463         }
2464
2465         return last_err;
2466 }
2467
2468 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr* nlh)
2469 {
2470         struct fib6_config cfg;
2471         int err;
2472
2473         err = rtm_to_fib6_config(skb, nlh, &cfg);
2474         if (err < 0)
2475                 return err;
2476
2477         if (cfg.fc_mp)
2478                 return ip6_route_multipath(&cfg, 0);
2479         else
2480                 return ip6_route_del(&cfg);
2481 }
2482
2483 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr* nlh)
2484 {
2485         struct fib6_config cfg;
2486         int err;
2487
2488         err = rtm_to_fib6_config(skb, nlh, &cfg);
2489         if (err < 0)
2490                 return err;
2491
2492         if (cfg.fc_mp)
2493                 return ip6_route_multipath(&cfg, 1);
2494         else
2495                 return ip6_route_add(&cfg);
2496 }
2497
2498 static inline size_t rt6_nlmsg_size(void)
2499 {
2500         return NLMSG_ALIGN(sizeof(struct rtmsg))
2501                + nla_total_size(16) /* RTA_SRC */
2502                + nla_total_size(16) /* RTA_DST */
2503                + nla_total_size(16) /* RTA_GATEWAY */
2504                + nla_total_size(16) /* RTA_PREFSRC */
2505                + nla_total_size(4) /* RTA_TABLE */
2506                + nla_total_size(4) /* RTA_IIF */
2507                + nla_total_size(4) /* RTA_OIF */
2508                + nla_total_size(4) /* RTA_PRIORITY */
2509                + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
2510                + nla_total_size(sizeof(struct rta_cacheinfo));
2511 }
2512
2513 static int rt6_fill_node(struct net *net,
2514                          struct sk_buff *skb, struct rt6_info *rt,
2515                          struct in6_addr *dst, struct in6_addr *src,
2516                          int iif, int type, u32 portid, u32 seq,
2517                          int prefix, int nowait, unsigned int flags)
2518 {
2519         struct rtmsg *rtm;
2520         struct nlmsghdr *nlh;
2521         long expires;
2522         u32 table;
2523
2524         if (prefix) {   /* user wants prefix routes only */
2525                 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
2526                         /* success since this is not a prefix route */
2527                         return 1;
2528                 }
2529         }
2530
2531         nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
2532         if (!nlh)
2533                 return -EMSGSIZE;
2534
2535         rtm = nlmsg_data(nlh);
2536         rtm->rtm_family = AF_INET6;
2537         rtm->rtm_dst_len = rt->rt6i_dst.plen;
2538         rtm->rtm_src_len = rt->rt6i_src.plen;
2539         rtm->rtm_tos = 0;
2540         if (rt->rt6i_table)
2541                 table = rt->rt6i_table->tb6_id;
2542         else
2543                 table = RT6_TABLE_UNSPEC;
2544         rtm->rtm_table = table;
2545         if (nla_put_u32(skb, RTA_TABLE, table))
2546                 goto nla_put_failure;
2547         if (rt->rt6i_flags & RTF_REJECT) {
2548                 switch (rt->dst.error) {
2549                 case -EINVAL:
2550                         rtm->rtm_type = RTN_BLACKHOLE;
2551                         break;
2552                 case -EACCES:
2553                         rtm->rtm_type = RTN_PROHIBIT;
2554                         break;
2555                 case -EAGAIN:
2556                         rtm->rtm_type = RTN_THROW;
2557                         break;
2558                 default:
2559                         rtm->rtm_type = RTN_UNREACHABLE;
2560                         break;
2561                 }
2562         }
2563         else if (rt->rt6i_flags & RTF_LOCAL)
2564                 rtm->rtm_type = RTN_LOCAL;
2565         else if (rt->dst.dev && (rt->dst.dev->flags & IFF_LOOPBACK))
2566                 rtm->rtm_type = RTN_LOCAL;
2567         else
2568                 rtm->rtm_type = RTN_UNICAST;
2569         rtm->rtm_flags = 0;
2570         rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2571         rtm->rtm_protocol = rt->rt6i_protocol;
2572         if (rt->rt6i_flags & RTF_DYNAMIC)
2573                 rtm->rtm_protocol = RTPROT_REDIRECT;
2574         else if (rt->rt6i_flags & RTF_ADDRCONF) {
2575                 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ROUTEINFO))
2576                         rtm->rtm_protocol = RTPROT_RA;
2577                 else
2578                         rtm->rtm_protocol = RTPROT_KERNEL;
2579         }
2580
2581         if (rt->rt6i_flags & RTF_CACHE)
2582                 rtm->rtm_flags |= RTM_F_CLONED;
2583
2584         if (dst) {
2585                 if (nla_put(skb, RTA_DST, 16, dst))
2586                         goto nla_put_failure;
2587                 rtm->rtm_dst_len = 128;
2588         } else if (rtm->rtm_dst_len)
2589                 if (nla_put(skb, RTA_DST, 16, &rt->rt6i_dst.addr))
2590                         goto nla_put_failure;
2591 #ifdef CONFIG_IPV6_SUBTREES
2592         if (src) {
2593                 if (nla_put(skb, RTA_SRC, 16, src))
2594                         goto nla_put_failure;
2595                 rtm->rtm_src_len = 128;
2596         } else if (rtm->rtm_src_len &&
2597                    nla_put(skb, RTA_SRC, 16, &rt->rt6i_src.addr))
2598                 goto nla_put_failure;
2599 #endif
2600         if (iif) {
2601 #ifdef CONFIG_IPV6_MROUTE
2602                 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
2603                         int err = ip6mr_get_route(net, skb, rtm, nowait);
2604                         if (err <= 0) {
2605                                 if (!nowait) {
2606                                         if (err == 0)
2607                                                 return 0;
2608                                         goto nla_put_failure;
2609                                 } else {
2610                                         if (err == -EMSGSIZE)
2611                                                 goto nla_put_failure;
2612                                 }
2613                         }
2614                 } else
2615 #endif
2616                         if (nla_put_u32(skb, RTA_IIF, iif))
2617                                 goto nla_put_failure;
2618         } else if (dst) {
2619                 struct in6_addr saddr_buf;
2620                 if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0 &&
2621                     nla_put(skb, RTA_PREFSRC, 16, &saddr_buf))
2622                         goto nla_put_failure;
2623         }
2624
2625         if (rt->rt6i_prefsrc.plen) {
2626                 struct in6_addr saddr_buf;
2627                 saddr_buf = rt->rt6i_prefsrc.addr;
2628                 if (nla_put(skb, RTA_PREFSRC, 16, &saddr_buf))
2629                         goto nla_put_failure;
2630         }
2631
2632         if (rtnetlink_put_metrics(skb, dst_metrics_ptr(&rt->dst)) < 0)
2633                 goto nla_put_failure;
2634
2635         if (rt->rt6i_flags & RTF_GATEWAY) {
2636                 if (nla_put(skb, RTA_GATEWAY, 16, &rt->rt6i_gateway) < 0)
2637                         goto nla_put_failure;
2638         }
2639
2640         if (rt->dst.dev &&
2641             nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex))
2642                 goto nla_put_failure;
2643         if (nla_put_u32(skb, RTA_PRIORITY, rt->rt6i_metric))
2644                 goto nla_put_failure;
2645
2646         expires = (rt->rt6i_flags & RTF_EXPIRES) ? rt->dst.expires - jiffies : 0;
2647
2648         if (rtnl_put_cacheinfo(skb, &rt->dst, 0, expires, rt->dst.error) < 0)
2649                 goto nla_put_failure;
2650
2651         return nlmsg_end(skb, nlh);
2652
2653 nla_put_failure:
2654         nlmsg_cancel(skb, nlh);
2655         return -EMSGSIZE;
2656 }
2657
2658 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
2659 {
2660         struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
2661         int prefix;
2662
2663         if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
2664                 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
2665                 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
2666         } else
2667                 prefix = 0;
2668
2669         return rt6_fill_node(arg->net,
2670                      arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
2671                      NETLINK_CB(arg->cb->skb).portid, arg->cb->nlh->nlmsg_seq,
2672                      prefix, 0, NLM_F_MULTI);
2673 }
2674
2675 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr* nlh)
2676 {
2677         struct net *net = sock_net(in_skb->sk);
2678         struct nlattr *tb[RTA_MAX+1];
2679         struct rt6_info *rt;
2680         struct sk_buff *skb;
2681         struct rtmsg *rtm;
2682         struct flowi6 fl6;
2683         int err, iif = 0, oif = 0;
2684
2685         err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2686         if (err < 0)
2687                 goto errout;
2688
2689         err = -EINVAL;
2690         memset(&fl6, 0, sizeof(fl6));
2691
2692         if (tb[RTA_SRC]) {
2693                 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
2694                         goto errout;
2695
2696                 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
2697         }
2698
2699         if (tb[RTA_DST]) {
2700                 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
2701                         goto errout;
2702
2703                 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
2704         }
2705
2706         if (tb[RTA_IIF])
2707                 iif = nla_get_u32(tb[RTA_IIF]);
2708
2709         if (tb[RTA_OIF])
2710                 oif = nla_get_u32(tb[RTA_OIF]);
2711
2712         if (iif) {
2713                 struct net_device *dev;
2714                 int flags = 0;
2715
2716                 dev = __dev_get_by_index(net, iif);
2717                 if (!dev) {
2718                         err = -ENODEV;
2719                         goto errout;
2720                 }
2721
2722                 fl6.flowi6_iif = iif;
2723
2724                 if (!ipv6_addr_any(&fl6.saddr))
2725                         flags |= RT6_LOOKUP_F_HAS_SADDR;
2726
2727                 rt = (struct rt6_info *)ip6_route_input_lookup(net, dev, &fl6,
2728                                                                flags);
2729         } else {
2730                 fl6.flowi6_oif = oif;
2731
2732                 rt = (struct rt6_info *)ip6_route_output(net, NULL, &fl6);
2733         }
2734
2735         skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
2736         if (!skb) {
2737                 ip6_rt_put(rt);
2738                 err = -ENOBUFS;
2739                 goto errout;
2740         }
2741
2742         /* Reserve room for dummy headers, this skb can pass
2743            through good chunk of routing engine.
2744          */
2745         skb_reset_mac_header(skb);
2746         skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
2747
2748         skb_dst_set(skb, &rt->dst);
2749
2750         err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
2751                             RTM_NEWROUTE, NETLINK_CB(in_skb).portid,
2752                             nlh->nlmsg_seq, 0, 0, 0);
2753         if (err < 0) {
2754                 kfree_skb(skb);
2755                 goto errout;
2756         }
2757
2758         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
2759 errout:
2760         return err;
2761 }
2762
2763 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info)
2764 {
2765         struct sk_buff *skb;
2766         struct net *net = info->nl_net;
2767         u32 seq;
2768         int err;
2769
2770         err = -ENOBUFS;
2771         seq = info->nlh ? info->nlh->nlmsg_seq : 0;
2772
2773         skb = nlmsg_new(rt6_nlmsg_size(), gfp_any());
2774         if (!skb)
2775                 goto errout;
2776
2777         err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
2778                                 event, info->portid, seq, 0, 0, 0);
2779         if (err < 0) {
2780                 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
2781                 WARN_ON(err == -EMSGSIZE);
2782                 kfree_skb(skb);
2783                 goto errout;
2784         }
2785         rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
2786                     info->nlh, gfp_any());
2787         return;
2788 errout:
2789         if (err < 0)
2790                 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
2791 }
2792
2793 static int ip6_route_dev_notify(struct notifier_block *this,
2794                                 unsigned long event, void *ptr)
2795 {
2796         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2797         struct net *net = dev_net(dev);
2798
2799         if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
2800                 net->ipv6.ip6_null_entry->dst.dev = dev;
2801                 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
2802 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2803                 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
2804                 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
2805                 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
2806                 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
2807 #endif
2808         }
2809
2810         return NOTIFY_OK;
2811 }
2812
2813 /*
2814  *      /proc
2815  */
2816
2817 #ifdef CONFIG_PROC_FS
2818
2819 static const struct file_operations ipv6_route_proc_fops = {
2820         .owner          = THIS_MODULE,
2821         .open           = ipv6_route_open,
2822         .read           = seq_read,
2823         .llseek         = seq_lseek,
2824         .release        = seq_release_net,
2825 };
2826
2827 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
2828 {
2829         struct net *net = (struct net *)seq->private;
2830         seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
2831                    net->ipv6.rt6_stats->fib_nodes,
2832                    net->ipv6.rt6_stats->fib_route_nodes,
2833                    net->ipv6.rt6_stats->fib_rt_alloc,
2834                    net->ipv6.rt6_stats->fib_rt_entries,
2835                    net->ipv6.rt6_stats->fib_rt_cache,
2836                    dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
2837                    net->ipv6.rt6_stats->fib_discarded_routes);
2838
2839         return 0;
2840 }
2841
2842 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
2843 {
2844         return single_open_net(inode, file, rt6_stats_seq_show);
2845 }
2846
2847 static const struct file_operations rt6_stats_seq_fops = {
2848         .owner   = THIS_MODULE,
2849         .open    = rt6_stats_seq_open,
2850         .read    = seq_read,
2851         .llseek  = seq_lseek,
2852         .release = single_release_net,
2853 };
2854 #endif  /* CONFIG_PROC_FS */
2855
2856 #ifdef CONFIG_SYSCTL
2857
2858 static
2859 int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
2860                               void __user *buffer, size_t *lenp, loff_t *ppos)
2861 {
2862         struct net *net;
2863         int delay;
2864         if (!write)
2865                 return -EINVAL;
2866
2867         net = (struct net *)ctl->extra1;
2868         delay = net->ipv6.sysctl.flush_delay;
2869         proc_dointvec(ctl, write, buffer, lenp, ppos);
2870         fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
2871         return 0;
2872 }
2873
2874 struct ctl_table ipv6_route_table_template[] = {
2875         {
2876                 .procname       =       "flush",
2877                 .data           =       &init_net.ipv6.sysctl.flush_delay,
2878                 .maxlen         =       sizeof(int),
2879                 .mode           =       0200,
2880                 .proc_handler   =       ipv6_sysctl_rtcache_flush
2881         },
2882         {
2883                 .procname       =       "gc_thresh",
2884                 .data           =       &ip6_dst_ops_template.gc_thresh,
2885                 .maxlen         =       sizeof(int),
2886                 .mode           =       0644,
2887                 .proc_handler   =       proc_dointvec,
2888         },
2889         {
2890                 .procname       =       "max_size",
2891                 .data           =       &init_net.ipv6.sysctl.ip6_rt_max_size,
2892                 .maxlen         =       sizeof(int),
2893                 .mode           =       0644,
2894                 .proc_handler   =       proc_dointvec,
2895         },
2896         {
2897                 .procname       =       "gc_min_interval",
2898                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2899                 .maxlen         =       sizeof(int),
2900                 .mode           =       0644,
2901                 .proc_handler   =       proc_dointvec_jiffies,
2902         },
2903         {
2904                 .procname       =       "gc_timeout",
2905                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
2906                 .maxlen         =       sizeof(int),
2907                 .mode           =       0644,
2908                 .proc_handler   =       proc_dointvec_jiffies,
2909         },
2910         {
2911                 .procname       =       "gc_interval",
2912                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_interval,
2913                 .maxlen         =       sizeof(int),
2914                 .mode           =       0644,
2915                 .proc_handler   =       proc_dointvec_jiffies,
2916         },
2917         {
2918                 .procname       =       "gc_elasticity",
2919                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
2920                 .maxlen         =       sizeof(int),
2921                 .mode           =       0644,
2922                 .proc_handler   =       proc_dointvec,
2923         },
2924         {
2925                 .procname       =       "mtu_expires",
2926                 .data           =       &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
2927                 .maxlen         =       sizeof(int),
2928                 .mode           =       0644,
2929                 .proc_handler   =       proc_dointvec_jiffies,
2930         },
2931         {
2932                 .procname       =       "min_adv_mss",
2933                 .data           =       &init_net.ipv6.sysctl.ip6_rt_min_advmss,
2934                 .maxlen         =       sizeof(int),
2935                 .mode           =       0644,
2936                 .proc_handler   =       proc_dointvec,
2937         },
2938         {
2939                 .procname       =       "gc_min_interval_ms",
2940                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2941                 .maxlen         =       sizeof(int),
2942                 .mode           =       0644,
2943                 .proc_handler   =       proc_dointvec_ms_jiffies,
2944         },
2945         { }
2946 };
2947
2948 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
2949 {
2950         struct ctl_table *table;
2951
2952         table = kmemdup(ipv6_route_table_template,
2953                         sizeof(ipv6_route_table_template),
2954                         GFP_KERNEL);
2955
2956         if (table) {
2957                 table[0].data = &net->ipv6.sysctl.flush_delay;
2958                 table[0].extra1 = net;
2959                 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
2960                 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
2961                 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2962                 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
2963                 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
2964                 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
2965                 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
2966                 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
2967                 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2968
2969                 /* Don't export sysctls to unprivileged users */
2970                 if (net->user_ns != &init_user_ns)
2971                         table[0].procname = NULL;
2972         }
2973
2974         return table;
2975 }
2976 #endif
2977
2978 static int __net_init ip6_route_net_init(struct net *net)
2979 {
2980         int ret = -ENOMEM;
2981
2982         memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
2983                sizeof(net->ipv6.ip6_dst_ops));
2984
2985         if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
2986                 goto out_ip6_dst_ops;
2987
2988         net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
2989                                            sizeof(*net->ipv6.ip6_null_entry),
2990                                            GFP_KERNEL);
2991         if (!net->ipv6.ip6_null_entry)
2992                 goto out_ip6_dst_entries;
2993         net->ipv6.ip6_null_entry->dst.path =
2994                 (struct dst_entry *)net->ipv6.ip6_null_entry;
2995         net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
2996         dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
2997                          ip6_template_metrics, true);
2998
2999 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3000         net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
3001                                                sizeof(*net->ipv6.ip6_prohibit_entry),
3002                                                GFP_KERNEL);
3003         if (!net->ipv6.ip6_prohibit_entry)
3004                 goto out_ip6_null_entry;
3005         net->ipv6.ip6_prohibit_entry->dst.path =
3006                 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
3007         net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3008         dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
3009                          ip6_template_metrics, true);
3010
3011         net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
3012                                                sizeof(*net->ipv6.ip6_blk_hole_entry),
3013                                                GFP_KERNEL);
3014         if (!net->ipv6.ip6_blk_hole_entry)
3015                 goto out_ip6_prohibit_entry;
3016         net->ipv6.ip6_blk_hole_entry->dst.path =
3017                 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
3018         net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3019         dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
3020                          ip6_template_metrics, true);
3021 #endif
3022
3023         net->ipv6.sysctl.flush_delay = 0;
3024         net->ipv6.sysctl.ip6_rt_max_size = 4096;
3025         net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
3026         net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
3027         net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
3028         net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
3029         net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
3030         net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
3031
3032         net->ipv6.ip6_rt_gc_expire = 30*HZ;
3033
3034         ret = 0;
3035 out:
3036         return ret;
3037
3038 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3039 out_ip6_prohibit_entry:
3040         kfree(net->ipv6.ip6_prohibit_entry);
3041 out_ip6_null_entry:
3042         kfree(net->ipv6.ip6_null_entry);
3043 #endif
3044 out_ip6_dst_entries:
3045         dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3046 out_ip6_dst_ops:
3047         goto out;
3048 }
3049
3050 static void __net_exit ip6_route_net_exit(struct net *net)
3051 {
3052         kfree(net->ipv6.ip6_null_entry);
3053 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3054         kfree(net->ipv6.ip6_prohibit_entry);
3055         kfree(net->ipv6.ip6_blk_hole_entry);
3056 #endif
3057         dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3058 }
3059
3060 static int __net_init ip6_route_net_init_late(struct net *net)
3061 {
3062 #ifdef CONFIG_PROC_FS
3063         proc_create("ipv6_route", 0, net->proc_net, &ipv6_route_proc_fops);
3064         proc_create("rt6_stats", S_IRUGO, net->proc_net, &rt6_stats_seq_fops);
3065 #endif
3066         return 0;
3067 }
3068
3069 static void __net_exit ip6_route_net_exit_late(struct net *net)
3070 {
3071 #ifdef CONFIG_PROC_FS
3072         remove_proc_entry("ipv6_route", net->proc_net);
3073         remove_proc_entry("rt6_stats", net->proc_net);
3074 #endif
3075 }
3076
3077 static struct pernet_operations ip6_route_net_ops = {
3078         .init = ip6_route_net_init,
3079         .exit = ip6_route_net_exit,
3080 };
3081
3082 static int __net_init ipv6_inetpeer_init(struct net *net)
3083 {
3084         struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
3085
3086         if (!bp)
3087                 return -ENOMEM;
3088         inet_peer_base_init(bp);
3089         net->ipv6.peers = bp;
3090         return 0;
3091 }
3092
3093 static void __net_exit ipv6_inetpeer_exit(struct net *net)
3094 {
3095         struct inet_peer_base *bp = net->ipv6.peers;
3096
3097         net->ipv6.peers = NULL;
3098         inetpeer_invalidate_tree(bp);
3099         kfree(bp);
3100 }
3101
3102 static struct pernet_operations ipv6_inetpeer_ops = {
3103         .init   =       ipv6_inetpeer_init,
3104         .exit   =       ipv6_inetpeer_exit,
3105 };
3106
3107 static struct pernet_operations ip6_route_net_late_ops = {
3108         .init = ip6_route_net_init_late,
3109         .exit = ip6_route_net_exit_late,
3110 };
3111
3112 static struct notifier_block ip6_route_dev_notifier = {
3113         .notifier_call = ip6_route_dev_notify,
3114         .priority = 0,
3115 };
3116
3117 int __init ip6_route_init(void)
3118 {
3119         int ret;
3120
3121         ret = -ENOMEM;
3122         ip6_dst_ops_template.kmem_cachep =
3123                 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
3124                                   SLAB_HWCACHE_ALIGN, NULL);
3125         if (!ip6_dst_ops_template.kmem_cachep)
3126                 goto out;
3127
3128         ret = dst_entries_init(&ip6_dst_blackhole_ops);
3129         if (ret)
3130                 goto out_kmem_cache;
3131
3132         ret = register_pernet_subsys(&ipv6_inetpeer_ops);
3133         if (ret)
3134                 goto out_dst_entries;
3135
3136         ret = register_pernet_subsys(&ip6_route_net_ops);
3137         if (ret)
3138                 goto out_register_inetpeer;
3139
3140         ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
3141
3142         /* Registering of the loopback is done before this portion of code,
3143          * the loopback reference in rt6_info will not be taken, do it
3144          * manually for init_net */
3145         init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
3146         init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3147   #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3148         init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
3149         init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3150         init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
3151         init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3152   #endif
3153         ret = fib6_init();
3154         if (ret)
3155                 goto out_register_subsys;
3156
3157         ret = xfrm6_init();
3158         if (ret)
3159                 goto out_fib6_init;
3160
3161         ret = fib6_rules_init();
3162         if (ret)
3163                 goto xfrm6_init;
3164
3165         ret = register_pernet_subsys(&ip6_route_net_late_ops);
3166         if (ret)
3167                 goto fib6_rules_init;
3168
3169         ret = -ENOBUFS;
3170         if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL, NULL) ||
3171             __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL, NULL) ||
3172             __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL, NULL))
3173                 goto out_register_late_subsys;
3174
3175         ret = register_netdevice_notifier(&ip6_route_dev_notifier);
3176         if (ret)
3177                 goto out_register_late_subsys;
3178
3179 out:
3180         return ret;
3181
3182 out_register_late_subsys:
3183         unregister_pernet_subsys(&ip6_route_net_late_ops);
3184 fib6_rules_init:
3185         fib6_rules_cleanup();
3186 xfrm6_init:
3187         xfrm6_fini();
3188 out_fib6_init:
3189         fib6_gc_cleanup();
3190 out_register_subsys:
3191         unregister_pernet_subsys(&ip6_route_net_ops);
3192 out_register_inetpeer:
3193         unregister_pernet_subsys(&ipv6_inetpeer_ops);
3194 out_dst_entries:
3195         dst_entries_destroy(&ip6_dst_blackhole_ops);
3196 out_kmem_cache:
3197         kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3198         goto out;
3199 }
3200
3201 void ip6_route_cleanup(void)
3202 {
3203         unregister_netdevice_notifier(&ip6_route_dev_notifier);
3204         unregister_pernet_subsys(&ip6_route_net_late_ops);
3205         fib6_rules_cleanup();
3206         xfrm6_fini();
3207         fib6_gc_cleanup();
3208         unregister_pernet_subsys(&ipv6_inetpeer_ops);
3209         unregister_pernet_subsys(&ip6_route_net_ops);
3210         dst_entries_destroy(&ip6_dst_blackhole_ops);
3211         kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3212 }