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                 INIT_LIST_HEAD(&rt->rt6i_siblings);
 318         }
 319         return rt;
 320 }
 321
 322 static void ip6_dst_destroy(struct dst_entry *dst)
 323 {
 324         struct rt6_info *rt = (struct rt6_info *)dst;
 325         struct inet6_dev *idev = rt->rt6i_idev;
 326         struct dst_entry *from = dst->from;
 327
 328         if (!(rt->dst.flags & DST_HOST))
 329                 dst_destroy_metrics_generic(dst);
 330
 331         if (idev) {
 332                 rt->rt6i_idev = NULL;
 333                 in6_dev_put(idev);
 334         }
 335
 336         dst->from = NULL;
 337         dst_release(from);
 338
 339         if (rt6_has_peer(rt)) {
 340                 struct inet_peer *peer = rt6_peer_ptr(rt);
 341                 inet_putpeer(peer);
 342         }
 343 }
 344
 345 static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
 346                            int how)
 347 {
 348         struct rt6_info *rt = (struct rt6_info *)dst;
 349         struct inet6_dev *idev = rt->rt6i_idev;
 350         struct net_device *loopback_dev =
 351                 dev_net(dev)->loopback_dev;
 352
 353         if (dev != loopback_dev) {
 354                 if (idev && idev->dev == dev) {
 355                         struct inet6_dev *loopback_idev =
 356                                 in6_dev_get(loopback_dev);
 357                         if (loopback_idev) {
 358                                 rt->rt6i_idev = loopback_idev;
 359                                 in6_dev_put(idev);
 360                         }
 361                 }
 362         }
 363 }
 364
 365 static bool rt6_check_expired(const struct rt6_info *rt)
 366 {
 367         if (rt->rt6i_flags & RTF_EXPIRES) {
 368                 if (time_after(jiffies, rt->dst.expires))
 369                         return true;
 370         } else if (rt->dst.from) {
 371                 return rt6_check_expired((struct rt6_info *) rt->dst.from);
 372         }
 373         return false;
 374 }
 375
 376 /* Multipath route selection:
 377  *   Hash based function using packet header and flowlabel.
 378  * Adapted from fib_info_hashfn()
 379  */
 380 static int rt6_info_hash_nhsfn(unsigned int candidate_count,
 381                                const struct flowi6 *fl6)
 382 {
 383         unsigned int val = fl6->flowi6_proto;
 384
 385         val ^= ipv6_addr_hash(&fl6->daddr);
 386         val ^= ipv6_addr_hash(&fl6->saddr);
 387
 388         /* Work only if this not encapsulated */
 389         switch (fl6->flowi6_proto) {
 390         case IPPROTO_UDP:
 391         case IPPROTO_TCP:
 392         case IPPROTO_SCTP:
 393                 val ^= (__force u16)fl6->fl6_sport;
 394                 val ^= (__force u16)fl6->fl6_dport;
 395                 break;
 396
 397         case IPPROTO_ICMPV6:
 398                 val ^= (__force u16)fl6->fl6_icmp_type;
 399                 val ^= (__force u16)fl6->fl6_icmp_code;
 400                 break;
 401         }
 402         /* RFC6438 recommands to use flowlabel */
 403         val ^= (__force u32)fl6->flowlabel;
 404
 405         /* Perhaps, we need to tune, this function? */
 406         val = val ^ (val >> 7) ^ (val >> 12);
 407         return val % candidate_count;
 408 }
 409
 410 static struct rt6_info *rt6_multipath_select(struct rt6_info *match,
 411                                              struct flowi6 *fl6, int oif,
 412                                              int strict)
 413 {
 414         struct rt6_info *sibling, *next_sibling;
 415         int route_choosen;
 416
 417         route_choosen = rt6_info_hash_nhsfn(match->rt6i_nsiblings + 1, fl6);
 418         /* Don't change the route, if route_choosen == 0
 419          * (siblings does not include ourself)
 420          */
 421         if (route_choosen)
 422                 list_for_each_entry_safe(sibling, next_sibling,
 423                                 &match->rt6i_siblings, rt6i_siblings) {
 424                         route_choosen--;
 425                         if (route_choosen == 0) {
 426                                 if (rt6_score_route(sibling, oif, strict) < 0)
 427                                         break;
 428                                 match = sibling;
 429                                 break;
 430                         }
 431                 }
 432         return match;
 433 }
 434
 435 /*
 436  *      Route lookup. Any table->tb6_lock is implied.
 437  */
 438
 439 static inline struct rt6_info *rt6_device_match(struct net *net,
 440                                                     struct rt6_info *rt,
 441                                                     const struct in6_addr *saddr,
 442                                                     int oif,
 443                                                     int flags)
 444 {
 445         struct rt6_info *local = NULL;
 446         struct rt6_info *sprt;
 447
 448         if (!oif && ipv6_addr_any(saddr))
 449                 goto out;
 450
 451         for (sprt = rt; sprt; sprt = sprt->dst.rt6_next) {
 452                 struct net_device *dev = sprt->dst.dev;
 453
 454                 if (oif) {
 455                         if (dev->ifindex == oif)
 456                                 return sprt;
 457                         if (dev->flags & IFF_LOOPBACK) {
 458                                 if (!sprt->rt6i_idev ||
 459                                     sprt->rt6i_idev->dev->ifindex != oif) {
 460                                         if (flags & RT6_LOOKUP_F_IFACE && oif)
 461                                                 continue;
 462                                         if (local && (!oif ||
 463                                                       local->rt6i_idev->dev->ifindex == oif))
 464                                                 continue;
 465                                 }
 466                                 local = sprt;
 467                         }
 468                 } else {
 469                         if (ipv6_chk_addr(net, saddr, dev,
 470                                           flags & RT6_LOOKUP_F_IFACE))
 471                                 return sprt;
 472                 }
 473         }
 474
 475         if (oif) {
 476                 if (local)
 477                         return local;
 478
 479                 if (flags & RT6_LOOKUP_F_IFACE)
 480                         return net->ipv6.ip6_null_entry;
 481         }
 482 out:
 483         return rt;
 484 }
 485
 486 #ifdef CONFIG_IPV6_ROUTER_PREF
 487 struct __rt6_probe_work {
 488         struct work_struct work;
 489         struct in6_addr target;
 490         struct net_device *dev;
 491 };
 492
 493 static void rt6_probe_deferred(struct work_struct *w)
 494 {
 495         struct in6_addr mcaddr;
 496         struct __rt6_probe_work *work =
 497                 container_of(w, struct __rt6_probe_work, work);
 498
 499         addrconf_addr_solict_mult(&work->target, &mcaddr);
 500         ndisc_send_ns(work->dev, NULL, &work->target, &mcaddr, NULL);
 501         dev_put(work->dev);
 502         kfree(w);
 503 }
 504
 505 static void rt6_probe(struct rt6_info *rt)
 506 {
 507         struct neighbour *neigh;
 508         /*
 509          * Okay, this does not seem to be appropriate
 510          * for now, however, we need to check if it
 511          * is really so; aka Router Reachability Probing.
 512          *
 513          * Router Reachability Probe MUST be rate-limited
 514          * to no more than one per minute.
 515          */
 516         if (!rt || !(rt->rt6i_flags & RTF_GATEWAY))
 517                 return;
 518         rcu_read_lock_bh();
 519         neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
 520         if (neigh) {
 521                 write_lock(&neigh->lock);
 522                 if (neigh->nud_state & NUD_VALID)
 523                         goto out;
 524         }
 525
 526         if (!neigh ||
 527             time_after(jiffies, neigh->updated + rt->rt6i_idev->cnf.rtr_probe_interval)) {
 528                 struct __rt6_probe_work *work;
 529
 530                 work = kmalloc(sizeof(*work), GFP_ATOMIC);
 531
 532                 if (neigh && work)
 533                         __neigh_set_probe_once(neigh);
 534
 535                 if (neigh)
 536                         write_unlock(&neigh->lock);
 537
 538                 if (work) {
 539                         INIT_WORK(&work->work, rt6_probe_deferred);
 540                         work->target = rt->rt6i_gateway;
 541                         dev_hold(rt->dst.dev);
 542                         work->dev = rt->dst.dev;
 543                         schedule_work(&work->work);
 544                 }
 545         } else {
 546 out:
 547                 write_unlock(&neigh->lock);
 548         }
 549         rcu_read_unlock_bh();
 550 }
 551 #else
 552 static inline void rt6_probe(struct rt6_info *rt)
 553 {
 554 }
 555 #endif
 556
 557 /*
 558  * Default Router Selection (RFC 2461 6.3.6)
 559  */
 560 static inline int rt6_check_dev(struct rt6_info *rt, int oif)
 561 {
 562         struct net_device *dev = rt->dst.dev;
 563         if (!oif || dev->ifindex == oif)
 564                 return 2;
 565         if ((dev->flags & IFF_LOOPBACK) &&
 566             rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
 567                 return 1;
 568         return 0;
 569 }
 570
 571 static inline enum rt6_nud_state rt6_check_neigh(struct rt6_info *rt)
 572 {
 573         struct neighbour *neigh;
 574         enum rt6_nud_state ret = RT6_NUD_FAIL_HARD;
 575
 576         if (rt->rt6i_flags & RTF_NONEXTHOP ||
 577             !(rt->rt6i_flags & RTF_GATEWAY))
 578                 return RT6_NUD_SUCCEED;
 579
 580         rcu_read_lock_bh();
 581         neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
 582         if (neigh) {
 583                 read_lock(&neigh->lock);
 584                 if (neigh->nud_state & NUD_VALID)
 585                         ret = RT6_NUD_SUCCEED;
 586 #ifdef CONFIG_IPV6_ROUTER_PREF
 587                 else if (!(neigh->nud_state & NUD_FAILED))
 588                         ret = RT6_NUD_SUCCEED;
 589                 else
 590                         ret = RT6_NUD_FAIL_PROBE;
 591 #endif
 592                 read_unlock(&neigh->lock);
 593         } else {
 594                 ret = IS_ENABLED(CONFIG_IPV6_ROUTER_PREF) ?
 595                       RT6_NUD_SUCCEED : RT6_NUD_FAIL_DO_RR;
 596         }
 597         rcu_read_unlock_bh();
 598
 599         return ret;
 600 }
 601
 602 static int rt6_score_route(struct rt6_info *rt, int oif,
 603                            int strict)
 604 {
 605         int m;
 606
 607         m = rt6_check_dev(rt, oif);
 608         if (!m && (strict & RT6_LOOKUP_F_IFACE))
 609                 return RT6_NUD_FAIL_HARD;
 610 #ifdef CONFIG_IPV6_ROUTER_PREF
 611         m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
 612 #endif
 613         if (strict & RT6_LOOKUP_F_REACHABLE) {
 614                 int n = rt6_check_neigh(rt);
 615                 if (n < 0)
 616                         return n;
 617         }
 618         return m;
 619 }
 620
 621 static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
 622                                    int *mpri, struct rt6_info *match,
 623                                    bool *do_rr)
 624 {
 625         int m;
 626         bool match_do_rr = false;
 627
 628         if (rt6_check_expired(rt))
 629                 goto out;
 630
 631         m = rt6_score_route(rt, oif, strict);
 632         if (m == RT6_NUD_FAIL_DO_RR) {
 633                 match_do_rr = true;
 634                 m = 0; /* lowest valid score */
 635         } else if (m == RT6_NUD_FAIL_HARD) {
 636                 goto out;
 637         }
 638
 639         if (strict & RT6_LOOKUP_F_REACHABLE)
 640                 rt6_probe(rt);
 641
 642         /* note that m can be RT6_NUD_FAIL_PROBE at this point */
 643         if (m > *mpri) {
 644                 *do_rr = match_do_rr;
 645                 *mpri = m;
 646                 match = rt;
 647         }
 648 out:
 649         return match;
 650 }
 651
 652 static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
 653                                      struct rt6_info *rr_head,
 654                                      u32 metric, int oif, int strict,
 655                                      bool *do_rr)
 656 {
 657         struct rt6_info *rt, *match;
 658         int mpri = -1;
 659
 660         match = NULL;
 661         for (rt = rr_head; rt && rt->rt6i_metric == metric;
 662              rt = rt->dst.rt6_next)
 663                 match = find_match(rt, oif, strict, &mpri, match, do_rr);
 664         for (rt = fn->leaf; rt && rt != rr_head && rt->rt6i_metric == metric;
 665              rt = rt->dst.rt6_next)
 666                 match = find_match(rt, oif, strict, &mpri, match, do_rr);
 667
 668         return match;
 669 }
 670
 671 static struct rt6_info *rt6_select(struct fib6_node *fn, int oif, int strict)
 672 {
 673         struct rt6_info *match, *rt0;
 674         struct net *net;
 675         bool do_rr = false;
 676
 677         rt0 = fn->rr_ptr;
 678         if (!rt0)
 679                 fn->rr_ptr = rt0 = fn->leaf;
 680
 681         match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict,
 682                              &do_rr);
 683
 684         if (do_rr) {
 685                 struct rt6_info *next = rt0->dst.rt6_next;
 686
 687                 /* no entries matched; do round-robin */
 688                 if (!next || next->rt6i_metric != rt0->rt6i_metric)
 689                         next = fn->leaf;
 690
 691                 if (next != rt0)
 692                         fn->rr_ptr = next;
 693         }
 694
 695         net = dev_net(rt0->dst.dev);
 696         return match ? match : net->ipv6.ip6_null_entry;
 697 }
 698
 699 #ifdef CONFIG_IPV6_ROUTE_INFO
 700 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
 701                   const struct in6_addr *gwaddr)
 702 {
 703         struct net *net = dev_net(dev);
 704         struct route_info *rinfo = (struct route_info *) opt;
 705         struct in6_addr prefix_buf, *prefix;
 706         unsigned int pref;
 707         unsigned long lifetime;
 708         struct rt6_info *rt;
 709
 710         if (len < sizeof(struct route_info)) {
 711                 return -EINVAL;
 712         }
 713
 714         /* Sanity check for prefix_len and length */
 715         if (rinfo->length > 3) {
 716                 return -EINVAL;
 717         } else if (rinfo->prefix_len > 128) {
 718                 return -EINVAL;
 719         } else if (rinfo->prefix_len > 64) {
 720                 if (rinfo->length < 2) {
 721                         return -EINVAL;
 722                 }
 723         } else if (rinfo->prefix_len > 0) {
 724                 if (rinfo->length < 1) {
 725                         return -EINVAL;
 726                 }
 727         }
 728
 729         pref = rinfo->route_pref;
 730         if (pref == ICMPV6_ROUTER_PREF_INVALID)
 731                 return -EINVAL;
 732
 733         lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
 734
 735         if (rinfo->length == 3)
 736                 prefix = (struct in6_addr *)rinfo->prefix;
 737         else {
 738                 /* this function is safe */
 739                 ipv6_addr_prefix(&prefix_buf,
 740                                  (struct in6_addr *)rinfo->prefix,
 741                                  rinfo->prefix_len);
 742                 prefix = &prefix_buf;
 743         }
 744
 745         if (rinfo->prefix_len == 0)
 746                 rt = rt6_get_dflt_router(gwaddr, dev);
 747         else
 748                 rt = rt6_get_route_info(net, prefix, rinfo->prefix_len,
 749                                         gwaddr, dev->ifindex);
 750
 751         if (rt && !lifetime) {
 752                 ip6_del_rt(rt);
 753                 rt = NULL;
 754         }
 755
 756         if (!rt && lifetime)
 757                 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr, dev->ifindex,
 758                                         pref);
 759         else if (rt)
 760                 rt->rt6i_flags = RTF_ROUTEINFO |
 761                                  (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
 762
 763         if (rt) {
 764                 if (!addrconf_finite_timeout(lifetime))
 765                         rt6_clean_expires(rt);
 766                 else
 767                         rt6_set_expires(rt, jiffies + HZ * lifetime);
 768
 769                 ip6_rt_put(rt);
 770         }
 771         return 0;
 772 }
 773 #endif
 774
 775 static struct fib6_node* fib6_backtrack(struct fib6_node *fn,
 776                                         struct in6_addr *saddr)
 777 {
 778         struct fib6_node *pn;
 779         while (1) {
 780                 if (fn->fn_flags & RTN_TL_ROOT)
 781                         return NULL;
 782                 pn = fn->parent;
 783                 if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn)
 784                         fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr);
 785                 else
 786                         fn = pn;
 787                 if (fn->fn_flags & RTN_RTINFO)
 788                         return fn;
 789         }
 790 }
 791
 792 static struct rt6_info *ip6_pol_route_lookup(struct net *net,
 793                                              struct fib6_table *table,
 794                                              struct flowi6 *fl6, int flags)
 795 {
 796         struct fib6_node *fn;
 797         struct rt6_info *rt;
 798
 799         read_lock_bh(&table->tb6_lock);
 800         fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
 801 restart:
 802         rt = fn->leaf;
 803         rt = rt6_device_match(net, rt, &fl6->saddr, fl6->flowi6_oif, flags);
 804         if (rt->rt6i_nsiblings && fl6->flowi6_oif == 0)
 805                 rt = rt6_multipath_select(rt, fl6, fl6->flowi6_oif, flags);
 806         if (rt == net->ipv6.ip6_null_entry) {
 807                 fn = fib6_backtrack(fn, &fl6->saddr);
 808                 if (fn)
 809                         goto restart;
 810         }
 811         dst_use(&rt->dst, jiffies);
 812         read_unlock_bh(&table->tb6_lock);
 813         return rt;
 814
 815 }
 816
 817 struct dst_entry *ip6_route_lookup(struct net *net, struct flowi6 *fl6,
 818                                     int flags)
 819 {
 820         return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_lookup);
 821 }
 822 EXPORT_SYMBOL_GPL(ip6_route_lookup);
 823
 824 struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
 825                             const struct in6_addr *saddr, int oif, int strict)
 826 {
 827         struct flowi6 fl6 = {
 828                 .flowi6_oif = oif,
 829                 .daddr = *daddr,
 830         };
 831         struct dst_entry *dst;
 832         int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
 833
 834         if (saddr) {
 835                 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
 836                 flags |= RT6_LOOKUP_F_HAS_SADDR;
 837         }
 838
 839         dst = fib6_rule_lookup(net, &fl6, flags, ip6_pol_route_lookup);
 840         if (dst->error == 0)
 841                 return (struct rt6_info *) dst;
 842
 843         dst_release(dst);
 844
 845         return NULL;
 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, *saved_fn;
 921         struct rt6_info *rt, *nrt;
 922         int strict = 0;
 923         int attempts = 3;
 924         int err;
 925
 926         strict |= flags & RT6_LOOKUP_F_IFACE;
 927         if (net->ipv6.devconf_all->forwarding == 0)
 928                 strict |= RT6_LOOKUP_F_REACHABLE;
 929
 930 redo_fib6_lookup_lock:
 931         read_lock_bh(&table->tb6_lock);
 932
 933         fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
 934         saved_fn = fn;
 935
 936 redo_rt6_select:
 937         rt = rt6_select(fn, oif, strict);
 938         if (rt->rt6i_nsiblings)
 939                 rt = rt6_multipath_select(rt, fl6, oif, strict);
 940         if (rt == net->ipv6.ip6_null_entry) {
 941                 fn = fib6_backtrack(fn, &fl6->saddr);
 942                 if (fn)
 943                         goto redo_rt6_select;
 944                 else if (strict & RT6_LOOKUP_F_REACHABLE) {
 945                         /* also consider unreachable route */
 946                         strict &= ~RT6_LOOKUP_F_REACHABLE;
 947                         fn = saved_fn;
 948                         goto redo_rt6_select;
 949                 } else {
 950                         dst_hold(&rt->dst);
 951                         read_unlock_bh(&table->tb6_lock);
 952                         goto out2;
 953                 }
 954         }
 955
 956         dst_hold(&rt->dst);
 957         read_unlock_bh(&table->tb6_lock);
 958
 959         if (rt->rt6i_flags & RTF_CACHE)
 960                 goto out2;
 961
 962         if (!(rt->rt6i_flags & (RTF_NONEXTHOP | RTF_GATEWAY)))
 963                 nrt = rt6_alloc_cow(rt, &fl6->daddr, &fl6->saddr);
 964         else if (!(rt->dst.flags & DST_HOST))
 965                 nrt = rt6_alloc_clone(rt, &fl6->daddr);
 966         else
 967                 goto out2;
 968
 969         ip6_rt_put(rt);
 970         rt = nrt ? : net->ipv6.ip6_null_entry;
 971
 972         dst_hold(&rt->dst);
 973         if (nrt) {
 974                 err = ip6_ins_rt(nrt);
 975                 if (!err)
 976                         goto out2;
 977         }
 978
 979         if (--attempts <= 0)
 980                 goto out2;
 981
 982         /*
 983          * Race condition! In the gap, when table->tb6_lock was
 984          * released someone could insert this route.  Relookup.
 985          */
 986         ip6_rt_put(rt);
 987         goto redo_fib6_lookup_lock;
 988
 989 out2:
 990         rt->dst.lastuse = jiffies;
 991         rt->dst.__use++;
 992
 993         return rt;
 994 }
 995
 996 static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
 997                                             struct flowi6 *fl6, int flags)
 998 {
 999         return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, flags);
1000 }
1001
1002 static struct dst_entry *ip6_route_input_lookup(struct net *net,
1003                                                 struct net_device *dev,
1004                                                 struct flowi6 *fl6, int flags)
1005 {
1006         if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
1007                 flags |= RT6_LOOKUP_F_IFACE;
1008
1009         return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_input);
1010 }
1011
1012 void ip6_route_input(struct sk_buff *skb)
1013 {
1014         const struct ipv6hdr *iph = ipv6_hdr(skb);
1015         struct net *net = dev_net(skb->dev);
1016         int flags = RT6_LOOKUP_F_HAS_SADDR;
1017         struct flowi6 fl6 = {
1018                 .flowi6_iif = skb->dev->ifindex,
1019                 .daddr = iph->daddr,
1020                 .saddr = iph->saddr,
1021                 .flowlabel = ip6_flowinfo(iph),
1022                 .flowi6_mark = skb->mark,
1023                 .flowi6_proto = iph->nexthdr,
1024         };
1025
1026         skb_dst_set(skb, ip6_route_input_lookup(net, skb->dev, &fl6, flags));
1027 }
1028
1029 static struct rt6_info *ip6_pol_route_output(struct net *net, struct fib6_table *table,
1030                                              struct flowi6 *fl6, int flags)
1031 {
1032         return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, flags);
1033 }
1034
1035 struct dst_entry *ip6_route_output(struct net *net, const struct sock *sk,
1036                                     struct flowi6 *fl6)
1037 {
1038         int flags = 0;
1039
1040         fl6->flowi6_iif = LOOPBACK_IFINDEX;
1041
1042         if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr))
1043                 flags |= RT6_LOOKUP_F_IFACE;
1044
1045         if (!ipv6_addr_any(&fl6->saddr))
1046                 flags |= RT6_LOOKUP_F_HAS_SADDR;
1047         else if (sk)
1048                 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
1049
1050         return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_output);
1051 }
1052 EXPORT_SYMBOL(ip6_route_output);
1053
1054 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
1055 {
1056         struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
1057         struct dst_entry *new = NULL;
1058
1059         rt = dst_alloc(&ip6_dst_blackhole_ops, ort->dst.dev, 1, DST_OBSOLETE_NONE, 0);
1060         if (rt) {
1061                 new = &rt->dst;
1062
1063                 memset(new + 1, 0, sizeof(*rt) - sizeof(*new));
1064                 rt6_init_peer(rt, net->ipv6.peers);
1065
1066                 new->__use = 1;
1067                 new->input = dst_discard;
1068                 new->output = dst_discard_sk;
1069
1070                 if (dst_metrics_read_only(&ort->dst))
1071                         new->_metrics = ort->dst._metrics;
1072                 else
1073                         dst_copy_metrics(new, &ort->dst);
1074                 rt->rt6i_idev = ort->rt6i_idev;
1075                 if (rt->rt6i_idev)
1076                         in6_dev_hold(rt->rt6i_idev);
1077
1078                 rt->rt6i_gateway = ort->rt6i_gateway;
1079                 rt->rt6i_flags = ort->rt6i_flags;
1080                 rt->rt6i_metric = 0;
1081
1082                 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1083 #ifdef CONFIG_IPV6_SUBTREES
1084                 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1085 #endif
1086
1087                 dst_free(new);
1088         }
1089
1090         dst_release(dst_orig);
1091         return new ? new : ERR_PTR(-ENOMEM);
1092 }
1093
1094 /*
1095  *      Destination cache support functions
1096  */
1097
1098 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
1099 {
1100         struct rt6_info *rt;
1101
1102         rt = (struct rt6_info *) dst;
1103
1104         /* All IPV6 dsts are created with ->obsolete set to the value
1105          * DST_OBSOLETE_FORCE_CHK which forces validation calls down
1106          * into this function always.
1107          */
1108         if (!rt->rt6i_node || (rt->rt6i_node->fn_sernum != cookie))
1109                 return NULL;
1110
1111         if (rt6_check_expired(rt))
1112                 return NULL;
1113
1114         return dst;
1115 }
1116
1117 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
1118 {
1119         struct rt6_info *rt = (struct rt6_info *) dst;
1120
1121         if (rt) {
1122                 if (rt->rt6i_flags & RTF_CACHE) {
1123                         if (rt6_check_expired(rt)) {
1124                                 ip6_del_rt(rt);
1125                                 dst = NULL;
1126                         }
1127                 } else {
1128                         dst_release(dst);
1129                         dst = NULL;
1130                 }
1131         }
1132         return dst;
1133 }
1134
1135 static void ip6_link_failure(struct sk_buff *skb)
1136 {
1137         struct rt6_info *rt;
1138
1139         icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
1140
1141         rt = (struct rt6_info *) skb_dst(skb);
1142         if (rt) {
1143                 if (rt->rt6i_flags & RTF_CACHE) {
1144                         dst_hold(&rt->dst);
1145                         if (ip6_del_rt(rt))
1146                                 dst_free(&rt->dst);
1147                 } else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT)) {
1148                         rt->rt6i_node->fn_sernum = -1;
1149                 }
1150         }
1151 }
1152
1153 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
1154                                struct sk_buff *skb, u32 mtu)
1155 {
1156         struct rt6_info *rt6 = (struct rt6_info *)dst;
1157
1158         dst_confirm(dst);
1159         if (mtu < dst_mtu(dst) && rt6->rt6i_dst.plen == 128) {
1160                 struct net *net = dev_net(dst->dev);
1161
1162                 rt6->rt6i_flags |= RTF_MODIFIED;
1163                 if (mtu < IPV6_MIN_MTU)
1164                         mtu = IPV6_MIN_MTU;
1165
1166                 dst_metric_set(dst, RTAX_MTU, mtu);
1167                 rt6_update_expires(rt6, net->ipv6.sysctl.ip6_rt_mtu_expires);
1168         }
1169 }
1170
1171 void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
1172                      int oif, u32 mark)
1173 {
1174         const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1175         struct dst_entry *dst;
1176         struct flowi6 fl6;
1177
1178         memset(&fl6, 0, sizeof(fl6));
1179         fl6.flowi6_oif = oif;
1180         fl6.flowi6_mark = mark ? mark : IP6_REPLY_MARK(net, skb->mark);
1181         fl6.daddr = iph->daddr;
1182         fl6.saddr = iph->saddr;
1183         fl6.flowlabel = ip6_flowinfo(iph);
1184
1185         dst = ip6_route_output(net, NULL, &fl6);
1186         if (!dst->error)
1187                 ip6_rt_update_pmtu(dst, NULL, skb, ntohl(mtu));
1188         dst_release(dst);
1189 }
1190 EXPORT_SYMBOL_GPL(ip6_update_pmtu);
1191
1192 void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
1193 {
1194         ip6_update_pmtu(skb, sock_net(sk), mtu,
1195                         sk->sk_bound_dev_if, sk->sk_mark);
1196 }
1197 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
1198
1199 /* Handle redirects */
1200 struct ip6rd_flowi {
1201         struct flowi6 fl6;
1202         struct in6_addr gateway;
1203 };
1204
1205 static struct rt6_info *__ip6_route_redirect(struct net *net,
1206                                              struct fib6_table *table,
1207                                              struct flowi6 *fl6,
1208                                              int flags)
1209 {
1210         struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
1211         struct rt6_info *rt;
1212         struct fib6_node *fn;
1213
1214         /* Get the "current" route for this destination and
1215          * check if the redirect has come from approriate router.
1216          *
1217          * RFC 4861 specifies that redirects should only be
1218          * accepted if they come from the nexthop to the target.
1219          * Due to the way the routes are chosen, this notion
1220          * is a bit fuzzy and one might need to check all possible
1221          * routes.
1222          */
1223
1224         read_lock_bh(&table->tb6_lock);
1225         fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1226 restart:
1227         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1228                 if (rt6_check_expired(rt))
1229                         continue;
1230                 if (rt->dst.error)
1231                         break;
1232                 if (!(rt->rt6i_flags & RTF_GATEWAY))
1233                         continue;
1234                 if (fl6->flowi6_oif != rt->dst.dev->ifindex)
1235                         continue;
1236                 if (!ipv6_addr_equal(&rdfl->gateway, &rt->rt6i_gateway))
1237                         continue;
1238                 break;
1239         }
1240
1241         if (!rt)
1242                 rt = net->ipv6.ip6_null_entry;
1243         else if (rt->dst.error) {
1244                 rt = net->ipv6.ip6_null_entry;
1245                 goto out;
1246         }
1247
1248         if (rt == net->ipv6.ip6_null_entry) {
1249                 fn = fib6_backtrack(fn, &fl6->saddr);
1250                 if (fn)
1251                         goto restart;
1252         }
1253
1254 out:
1255         dst_hold(&rt->dst);
1256
1257         read_unlock_bh(&table->tb6_lock);
1258
1259         return rt;
1260 };
1261
1262 static struct dst_entry *ip6_route_redirect(struct net *net,
1263                                         const struct flowi6 *fl6,
1264                                         const struct in6_addr *gateway)
1265 {
1266         int flags = RT6_LOOKUP_F_HAS_SADDR;
1267         struct ip6rd_flowi rdfl;
1268
1269         rdfl.fl6 = *fl6;
1270         rdfl.gateway = *gateway;
1271
1272         return fib6_rule_lookup(net, &rdfl.fl6,
1273                                 flags, __ip6_route_redirect);
1274 }
1275
1276 void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark)
1277 {
1278         const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1279         struct dst_entry *dst;
1280         struct flowi6 fl6;
1281
1282         memset(&fl6, 0, sizeof(fl6));
1283         fl6.flowi6_iif = LOOPBACK_IFINDEX;
1284         fl6.flowi6_oif = oif;
1285         fl6.flowi6_mark = mark;
1286         fl6.daddr = iph->daddr;
1287         fl6.saddr = iph->saddr;
1288         fl6.flowlabel = ip6_flowinfo(iph);
1289
1290         dst = ip6_route_redirect(net, &fl6, &ipv6_hdr(skb)->saddr);
1291         rt6_do_redirect(dst, NULL, skb);
1292         dst_release(dst);
1293 }
1294 EXPORT_SYMBOL_GPL(ip6_redirect);
1295
1296 void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif,
1297                             u32 mark)
1298 {
1299         const struct ipv6hdr *iph = ipv6_hdr(skb);
1300         const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
1301         struct dst_entry *dst;
1302         struct flowi6 fl6;
1303
1304         memset(&fl6, 0, sizeof(fl6));
1305         fl6.flowi6_iif = LOOPBACK_IFINDEX;
1306         fl6.flowi6_oif = oif;
1307         fl6.flowi6_mark = mark;
1308         fl6.daddr = msg->dest;
1309         fl6.saddr = iph->daddr;
1310
1311         dst = ip6_route_redirect(net, &fl6, &iph->saddr);
1312         rt6_do_redirect(dst, NULL, skb);
1313         dst_release(dst);
1314 }
1315
1316 void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
1317 {
1318         ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark);
1319 }
1320 EXPORT_SYMBOL_GPL(ip6_sk_redirect);
1321
1322 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1323 {
1324         struct net_device *dev = dst->dev;
1325         unsigned int mtu = dst_mtu(dst);
1326         struct net *net = dev_net(dev);
1327
1328         mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
1329
1330         if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
1331                 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1332
1333         /*
1334          * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1335          * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1336          * IPV6_MAXPLEN is also valid and means: "any MSS,
1337          * rely only on pmtu discovery"
1338          */
1339         if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
1340                 mtu = IPV6_MAXPLEN;
1341         return mtu;
1342 }
1343
1344 static unsigned int ip6_mtu(const struct dst_entry *dst)
1345 {
1346         struct inet6_dev *idev;
1347         unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
1348
1349         if (mtu)
1350                 goto out;
1351
1352         mtu = IPV6_MIN_MTU;
1353
1354         rcu_read_lock();
1355         idev = __in6_dev_get(dst->dev);
1356         if (idev)
1357                 mtu = idev->cnf.mtu6;
1358         rcu_read_unlock();
1359
1360 out:
1361         return min_t(unsigned int, mtu, IP6_MAX_MTU);
1362 }
1363
1364 static struct dst_entry *icmp6_dst_gc_list;
1365 static DEFINE_SPINLOCK(icmp6_dst_lock);
1366
1367 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
1368                                   struct flowi6 *fl6)
1369 {
1370         struct dst_entry *dst;
1371         struct rt6_info *rt;
1372         struct inet6_dev *idev = in6_dev_get(dev);
1373         struct net *net = dev_net(dev);
1374
1375         if (unlikely(!idev))
1376                 return ERR_PTR(-ENODEV);
1377
1378         rt = ip6_dst_alloc(net, dev, 0, NULL);
1379         if (unlikely(!rt)) {
1380                 in6_dev_put(idev);
1381                 dst = ERR_PTR(-ENOMEM);
1382                 goto out;
1383         }
1384
1385         rt->dst.flags |= DST_HOST;
1386         rt->dst.output  = ip6_output;
1387         atomic_set(&rt->dst.__refcnt, 1);
1388         rt->rt6i_gateway  = fl6->daddr;
1389         rt->rt6i_dst.addr = fl6->daddr;
1390         rt->rt6i_dst.plen = 128;
1391         rt->rt6i_idev     = idev;
1392         dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
1393
1394         spin_lock_bh(&icmp6_dst_lock);
1395         rt->dst.next = icmp6_dst_gc_list;
1396         icmp6_dst_gc_list = &rt->dst;
1397         spin_unlock_bh(&icmp6_dst_lock);
1398
1399         fib6_force_start_gc(net);
1400
1401         dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
1402
1403 out:
1404         return dst;
1405 }
1406
1407 int icmp6_dst_gc(void)
1408 {
1409         struct dst_entry *dst, **pprev;
1410         int more = 0;
1411
1412         spin_lock_bh(&icmp6_dst_lock);
1413         pprev = &icmp6_dst_gc_list;
1414
1415         while ((dst = *pprev) != NULL) {
1416                 if (!atomic_read(&dst->__refcnt)) {
1417                         *pprev = dst->next;
1418                         dst_free(dst);
1419                 } else {
1420                         pprev = &dst->next;
1421                         ++more;
1422                 }
1423         }
1424
1425         spin_unlock_bh(&icmp6_dst_lock);
1426
1427         return more;
1428 }
1429
1430 static void icmp6_clean_all(int (*func)(struct rt6_info *rt, void *arg),
1431                             void *arg)
1432 {
1433         struct dst_entry *dst, **pprev;
1434
1435         spin_lock_bh(&icmp6_dst_lock);
1436         pprev = &icmp6_dst_gc_list;
1437         while ((dst = *pprev) != NULL) {
1438                 struct rt6_info *rt = (struct rt6_info *) dst;
1439                 if (func(rt, arg)) {
1440                         *pprev = dst->next;
1441                         dst_free(dst);
1442                 } else {
1443                         pprev = &dst->next;
1444                 }
1445         }
1446         spin_unlock_bh(&icmp6_dst_lock);
1447 }
1448
1449 static int ip6_dst_gc(struct dst_ops *ops)
1450 {
1451         struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
1452         int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1453         int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1454         int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1455         int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1456         unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1457         int entries;
1458
1459         entries = dst_entries_get_fast(ops);
1460         if (time_after(rt_last_gc + rt_min_interval, jiffies) &&
1461             entries <= rt_max_size)
1462                 goto out;
1463
1464         net->ipv6.ip6_rt_gc_expire++;
1465         fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net, true);
1466         entries = dst_entries_get_slow(ops);
1467         if (entries < ops->gc_thresh)
1468                 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1469 out:
1470         net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1471         return entries > rt_max_size;
1472 }
1473
1474 /*
1475  *
1476  */
1477
1478 int ip6_route_add(struct fib6_config *cfg)
1479 {
1480         int err;
1481         struct net *net = cfg->fc_nlinfo.nl_net;
1482         struct rt6_info *rt = NULL;
1483         struct net_device *dev = NULL;
1484         struct inet6_dev *idev = NULL;
1485         struct fib6_table *table;
1486         int addr_type;
1487
1488         if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1489                 return -EINVAL;
1490 #ifndef CONFIG_IPV6_SUBTREES
1491         if (cfg->fc_src_len)
1492                 return -EINVAL;
1493 #endif
1494         if (cfg->fc_ifindex) {
1495                 err = -ENODEV;
1496                 dev = dev_get_by_index(net, cfg->fc_ifindex);
1497                 if (!dev)
1498                         goto out;
1499                 idev = in6_dev_get(dev);
1500                 if (!idev)
1501                         goto out;
1502         }
1503
1504         if (cfg->fc_metric == 0)
1505                 cfg->fc_metric = IP6_RT_PRIO_USER;
1506
1507         err = -ENOBUFS;
1508         if (cfg->fc_nlinfo.nlh &&
1509             !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
1510                 table = fib6_get_table(net, cfg->fc_table);
1511                 if (!table) {
1512                         pr_warn("NLM_F_CREATE should be specified when creating new route\n");
1513                         table = fib6_new_table(net, cfg->fc_table);
1514                 }
1515         } else {
1516                 table = fib6_new_table(net, cfg->fc_table);
1517         }
1518
1519         if (!table)
1520                 goto out;
1521
1522         rt = ip6_dst_alloc(net, NULL, (cfg->fc_flags & RTF_ADDRCONF) ? 0 : DST_NOCOUNT, table);
1523
1524         if (!rt) {
1525                 err = -ENOMEM;
1526                 goto out;
1527         }
1528
1529         if (cfg->fc_flags & RTF_EXPIRES)
1530                 rt6_set_expires(rt, jiffies +
1531                                 clock_t_to_jiffies(cfg->fc_expires));
1532         else
1533                 rt6_clean_expires(rt);
1534
1535         if (cfg->fc_protocol == RTPROT_UNSPEC)
1536                 cfg->fc_protocol = RTPROT_BOOT;
1537         rt->rt6i_protocol = cfg->fc_protocol;
1538
1539         addr_type = ipv6_addr_type(&cfg->fc_dst);
1540
1541         if (addr_type & IPV6_ADDR_MULTICAST)
1542                 rt->dst.input = ip6_mc_input;
1543         else if (cfg->fc_flags & RTF_LOCAL)
1544                 rt->dst.input = ip6_input;
1545         else
1546                 rt->dst.input = ip6_forward;
1547
1548         rt->dst.output = ip6_output;
1549
1550         ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1551         rt->rt6i_dst.plen = cfg->fc_dst_len;
1552         if (rt->rt6i_dst.plen == 128) {
1553                 rt->dst.flags |= DST_HOST;
1554                 dst_metrics_set_force_overwrite(&rt->dst);
1555         }
1556
1557 #ifdef CONFIG_IPV6_SUBTREES
1558         ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1559         rt->rt6i_src.plen = cfg->fc_src_len;
1560 #endif
1561
1562         rt->rt6i_metric = cfg->fc_metric;
1563
1564         /* We cannot add true routes via loopback here,
1565            they would result in kernel looping; promote them to reject routes
1566          */
1567         if ((cfg->fc_flags & RTF_REJECT) ||
1568             (dev && (dev->flags & IFF_LOOPBACK) &&
1569              !(addr_type & IPV6_ADDR_LOOPBACK) &&
1570              !(cfg->fc_flags & RTF_LOCAL))) {
1571                 /* hold loopback dev/idev if we haven't done so. */
1572                 if (dev != net->loopback_dev) {
1573                         if (dev) {
1574                                 dev_put(dev);
1575                                 in6_dev_put(idev);
1576                         }
1577                         dev = net->loopback_dev;
1578                         dev_hold(dev);
1579                         idev = in6_dev_get(dev);
1580                         if (!idev) {
1581                                 err = -ENODEV;
1582                                 goto out;
1583                         }
1584                 }
1585                 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1586                 switch (cfg->fc_type) {
1587                 case RTN_BLACKHOLE:
1588                         rt->dst.error = -EINVAL;
1589                         rt->dst.output = dst_discard_sk;
1590                         rt->dst.input = dst_discard;
1591                         break;
1592                 case RTN_PROHIBIT:
1593                         rt->dst.error = -EACCES;
1594                         rt->dst.output = ip6_pkt_prohibit_out;
1595                         rt->dst.input = ip6_pkt_prohibit;
1596                         break;
1597                 case RTN_THROW:
1598                 default:
1599                         rt->dst.error = (cfg->fc_type == RTN_THROW) ? -EAGAIN
1600                                         : -ENETUNREACH;
1601                         rt->dst.output = ip6_pkt_discard_out;
1602                         rt->dst.input = ip6_pkt_discard;
1603                         break;
1604                 }
1605                 goto install_route;
1606         }
1607
1608         if (cfg->fc_flags & RTF_GATEWAY) {
1609                 const struct in6_addr *gw_addr;
1610                 int gwa_type;
1611
1612                 gw_addr = &cfg->fc_gateway;
1613                 rt->rt6i_gateway = *gw_addr;
1614                 gwa_type = ipv6_addr_type(gw_addr);
1615
1616                 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1617                         struct rt6_info *grt;
1618
1619                         /* IPv6 strictly inhibits using not link-local
1620                            addresses as nexthop address.
1621                            Otherwise, router will not able to send redirects.
1622                            It is very good, but in some (rare!) circumstances
1623                            (SIT, PtP, NBMA NOARP links) it is handy to allow
1624                            some exceptions. --ANK
1625                          */
1626                         err = -EINVAL;
1627                         if (!(gwa_type & IPV6_ADDR_UNICAST))
1628                                 goto out;
1629
1630                         grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, 1);
1631
1632                         err = -EHOSTUNREACH;
1633                         if (!grt)
1634                                 goto out;
1635                         if (dev) {
1636                                 if (dev != grt->dst.dev) {
1637                                         ip6_rt_put(grt);
1638                                         goto out;
1639                                 }
1640                         } else {
1641                                 dev = grt->dst.dev;
1642                                 idev = grt->rt6i_idev;
1643                                 dev_hold(dev);
1644                                 in6_dev_hold(grt->rt6i_idev);
1645                         }
1646                         if (!(grt->rt6i_flags & RTF_GATEWAY))
1647                                 err = 0;
1648                         ip6_rt_put(grt);
1649
1650                         if (err)
1651                                 goto out;
1652                 }
1653                 err = -EINVAL;
1654                 if (!dev || (dev->flags & IFF_LOOPBACK))
1655                         goto out;
1656         }
1657
1658         err = -ENODEV;
1659         if (!dev)
1660                 goto out;
1661
1662         if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
1663                 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
1664                         err = -EINVAL;
1665                         goto out;
1666                 }
1667                 rt->rt6i_prefsrc.addr = cfg->fc_prefsrc;
1668                 rt->rt6i_prefsrc.plen = 128;
1669         } else
1670                 rt->rt6i_prefsrc.plen = 0;
1671
1672         rt->rt6i_flags = cfg->fc_flags;
1673
1674 install_route:
1675         rt->dst.dev = dev;
1676         rt->rt6i_idev = idev;
1677         rt->rt6i_table = table;
1678
1679         cfg->fc_nlinfo.nl_net = dev_net(dev);
1680
1681         return __ip6_ins_rt(rt, &cfg->fc_nlinfo, cfg->fc_mx, cfg->fc_mx_len);
1682
1683 out:
1684         if (dev)
1685                 dev_put(dev);
1686         if (idev)
1687                 in6_dev_put(idev);
1688         if (rt)
1689                 dst_free(&rt->dst);
1690         return err;
1691 }
1692
1693 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
1694 {
1695         int err;
1696         struct fib6_table *table;
1697         struct net *net = dev_net(rt->dst.dev);
1698
1699         if (rt == net->ipv6.ip6_null_entry) {
1700                 err = -ENOENT;
1701                 goto out;
1702         }
1703
1704         table = rt->rt6i_table;
1705         write_lock_bh(&table->tb6_lock);
1706         err = fib6_del(rt, info);
1707         write_unlock_bh(&table->tb6_lock);
1708
1709 out:
1710         ip6_rt_put(rt);
1711         return err;
1712 }
1713
1714 int ip6_del_rt(struct rt6_info *rt)
1715 {
1716         struct nl_info info = {
1717                 .nl_net = dev_net(rt->dst.dev),
1718         };
1719         return __ip6_del_rt(rt, &info);
1720 }
1721
1722 static int ip6_route_del(struct fib6_config *cfg)
1723 {
1724         struct fib6_table *table;
1725         struct fib6_node *fn;
1726         struct rt6_info *rt;
1727         int err = -ESRCH;
1728
1729         table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
1730         if (!table)
1731                 return err;
1732
1733         read_lock_bh(&table->tb6_lock);
1734
1735         fn = fib6_locate(&table->tb6_root,
1736                          &cfg->fc_dst, cfg->fc_dst_len,
1737                          &cfg->fc_src, cfg->fc_src_len);
1738
1739         if (fn) {
1740                 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1741                         if (cfg->fc_ifindex &&
1742                             (!rt->dst.dev ||
1743                              rt->dst.dev->ifindex != cfg->fc_ifindex))
1744                                 continue;
1745                         if (cfg->fc_flags & RTF_GATEWAY &&
1746                             !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
1747                                 continue;
1748                         if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
1749                                 continue;
1750                         dst_hold(&rt->dst);
1751                         read_unlock_bh(&table->tb6_lock);
1752
1753                         return __ip6_del_rt(rt, &cfg->fc_nlinfo);
1754                 }
1755         }
1756         read_unlock_bh(&table->tb6_lock);
1757
1758         return err;
1759 }
1760
1761 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
1762 {
1763         struct net *net = dev_net(skb->dev);
1764         struct netevent_redirect netevent;
1765         struct rt6_info *rt, *nrt = NULL;
1766         struct ndisc_options ndopts;
1767         struct inet6_dev *in6_dev;
1768         struct neighbour *neigh;
1769         struct rd_msg *msg;
1770         int optlen, on_link;
1771         u8 *lladdr;
1772
1773         optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
1774         optlen -= sizeof(*msg);
1775
1776         if (optlen < 0) {
1777                 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
1778                 return;
1779         }
1780
1781         msg = (struct rd_msg *)icmp6_hdr(skb);
1782
1783         if (ipv6_addr_is_multicast(&msg->dest)) {
1784                 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
1785                 return;
1786         }
1787
1788         on_link = 0;
1789         if (ipv6_addr_equal(&msg->dest, &msg->target)) {
1790                 on_link = 1;
1791         } else if (ipv6_addr_type(&msg->target) !=
1792                    (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
1793                 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
1794                 return;
1795         }
1796
1797         in6_dev = __in6_dev_get(skb->dev);
1798         if (!in6_dev)
1799                 return;
1800         if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
1801                 return;
1802
1803         /* RFC2461 8.1:
1804          *      The IP source address of the Redirect MUST be the same as the current
1805          *      first-hop router for the specified ICMP Destination Address.
1806          */
1807
1808         if (!ndisc_parse_options(msg->opt, optlen, &ndopts)) {
1809                 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
1810                 return;
1811         }
1812
1813         lladdr = NULL;
1814         if (ndopts.nd_opts_tgt_lladdr) {
1815                 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
1816                                              skb->dev);
1817                 if (!lladdr) {
1818                         net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
1819                         return;
1820                 }
1821         }
1822
1823         rt = (struct rt6_info *) dst;
1824         if (rt == net->ipv6.ip6_null_entry) {
1825                 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
1826                 return;
1827         }
1828
1829         /* Redirect received -> path was valid.
1830          * Look, redirects are sent only in response to data packets,
1831          * so that this nexthop apparently is reachable. --ANK
1832          */
1833         dst_confirm(&rt->dst);
1834
1835         neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
1836         if (!neigh)
1837                 return;
1838
1839         /*
1840          *      We have finally decided to accept it.
1841          */
1842
1843         neigh_update(neigh, lladdr, NUD_STALE,
1844                      NEIGH_UPDATE_F_WEAK_OVERRIDE|
1845                      NEIGH_UPDATE_F_OVERRIDE|
1846                      (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1847                                      NEIGH_UPDATE_F_ISROUTER))
1848                      );
1849
1850         nrt = ip6_rt_copy(rt, &msg->dest);
1851         if (!nrt)
1852                 goto out;
1853
1854         nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
1855         if (on_link)
1856                 nrt->rt6i_flags &= ~RTF_GATEWAY;
1857
1858         nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
1859
1860         if (ip6_ins_rt(nrt))
1861                 goto out;
1862
1863         netevent.old = &rt->dst;
1864         netevent.new = &nrt->dst;
1865         netevent.daddr = &msg->dest;
1866         netevent.neigh = neigh;
1867         call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
1868
1869         if (rt->rt6i_flags & RTF_CACHE) {
1870                 rt = (struct rt6_info *) dst_clone(&rt->dst);
1871                 ip6_del_rt(rt);
1872         }
1873
1874 out:
1875         neigh_release(neigh);
1876 }
1877
1878 /*
1879  *      Misc support functions
1880  */
1881
1882 static struct rt6_info *ip6_rt_copy(struct rt6_info *ort,
1883                                     const struct in6_addr *dest)
1884 {
1885         struct net *net = dev_net(ort->dst.dev);
1886         struct rt6_info *rt = ip6_dst_alloc(net, ort->dst.dev, 0,
1887                                             ort->rt6i_table);
1888
1889         if (rt) {
1890                 rt->dst.input = ort->dst.input;
1891                 rt->dst.output = ort->dst.output;
1892                 rt->dst.flags |= DST_HOST;
1893
1894                 rt->rt6i_dst.addr = *dest;
1895                 rt->rt6i_dst.plen = 128;
1896                 dst_copy_metrics(&rt->dst, &ort->dst);
1897                 rt->dst.error = ort->dst.error;
1898                 rt->rt6i_idev = ort->rt6i_idev;
1899                 if (rt->rt6i_idev)
1900                         in6_dev_hold(rt->rt6i_idev);
1901                 rt->dst.lastuse = jiffies;
1902
1903                 if (ort->rt6i_flags & RTF_GATEWAY)
1904                         rt->rt6i_gateway = ort->rt6i_gateway;
1905                 else
1906                         rt->rt6i_gateway = *dest;
1907                 rt->rt6i_flags = ort->rt6i_flags;
1908                 rt6_set_from(rt, ort);
1909                 rt->rt6i_metric = 0;
1910
1911 #ifdef CONFIG_IPV6_SUBTREES
1912                 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1913 #endif
1914                 memcpy(&rt->rt6i_prefsrc, &ort->rt6i_prefsrc, sizeof(struct rt6key));
1915                 rt->rt6i_table = ort->rt6i_table;
1916         }
1917         return rt;
1918 }
1919
1920 #ifdef CONFIG_IPV6_ROUTE_INFO
1921 static struct rt6_info *rt6_get_route_info(struct net *net,
1922                                            const struct in6_addr *prefix, int prefixlen,
1923                                            const struct in6_addr *gwaddr, int ifindex)
1924 {
1925         struct fib6_node *fn;
1926         struct rt6_info *rt = NULL;
1927         struct fib6_table *table;
1928
1929         table = fib6_get_table(net, RT6_TABLE_INFO);
1930         if (!table)
1931                 return NULL;
1932
1933         read_lock_bh(&table->tb6_lock);
1934         fn = fib6_locate(&table->tb6_root, prefix, prefixlen, NULL, 0);
1935         if (!fn)
1936                 goto out;
1937
1938         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1939                 if (rt->dst.dev->ifindex != ifindex)
1940                         continue;
1941                 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
1942                         continue;
1943                 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
1944                         continue;
1945                 dst_hold(&rt->dst);
1946                 break;
1947         }
1948 out:
1949         read_unlock_bh(&table->tb6_lock);
1950         return rt;
1951 }
1952
1953 static struct rt6_info *rt6_add_route_info(struct net *net,
1954                                            const struct in6_addr *prefix, int prefixlen,
1955                                            const struct in6_addr *gwaddr, int ifindex,
1956                                            unsigned int pref)
1957 {
1958         struct fib6_config cfg = {
1959                 .fc_table       = RT6_TABLE_INFO,
1960                 .fc_metric      = IP6_RT_PRIO_USER,
1961                 .fc_ifindex     = ifindex,
1962                 .fc_dst_len     = prefixlen,
1963                 .fc_flags       = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
1964                                   RTF_UP | RTF_PREF(pref),
1965                 .fc_nlinfo.portid = 0,
1966                 .fc_nlinfo.nlh = NULL,
1967                 .fc_nlinfo.nl_net = net,
1968         };
1969
1970         cfg.fc_dst = *prefix;
1971         cfg.fc_gateway = *gwaddr;
1972
1973         /* We should treat it as a default route if prefix length is 0. */
1974         if (!prefixlen)
1975                 cfg.fc_flags |= RTF_DEFAULT;
1976
1977         ip6_route_add(&cfg);
1978
1979         return rt6_get_route_info(net, prefix, prefixlen, gwaddr, ifindex);
1980 }
1981 #endif
1982
1983 struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
1984 {
1985         struct rt6_info *rt;
1986         struct fib6_table *table;
1987
1988         table = fib6_get_table(dev_net(dev), RT6_TABLE_DFLT);
1989         if (!table)
1990                 return NULL;
1991
1992         read_lock_bh(&table->tb6_lock);
1993         for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
1994                 if (dev == rt->dst.dev &&
1995                     ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
1996                     ipv6_addr_equal(&rt->rt6i_gateway, addr))
1997                         break;
1998         }
1999         if (rt)
2000                 dst_hold(&rt->dst);
2001         read_unlock_bh(&table->tb6_lock);
2002         return rt;
2003 }
2004
2005 struct rt6_info *rt6_add_dflt_router(const struct in6_addr *gwaddr,
2006                                      struct net_device *dev,
2007                                      unsigned int pref)
2008 {
2009         struct fib6_config cfg = {
2010                 .fc_table       = RT6_TABLE_DFLT,
2011                 .fc_metric      = IP6_RT_PRIO_USER,
2012                 .fc_ifindex     = dev->ifindex,
2013                 .fc_flags       = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
2014                                   RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
2015                 .fc_nlinfo.portid = 0,
2016                 .fc_nlinfo.nlh = NULL,
2017                 .fc_nlinfo.nl_net = dev_net(dev),
2018         };
2019
2020         cfg.fc_gateway = *gwaddr;
2021
2022         ip6_route_add(&cfg);
2023
2024         return rt6_get_dflt_router(gwaddr, dev);
2025 }
2026
2027 void rt6_purge_dflt_routers(struct net *net)
2028 {
2029         struct rt6_info *rt;
2030         struct fib6_table *table;
2031
2032         /* NOTE: Keep consistent with rt6_get_dflt_router */
2033         table = fib6_get_table(net, RT6_TABLE_DFLT);
2034         if (!table)
2035                 return;
2036
2037 restart:
2038         read_lock_bh(&table->tb6_lock);
2039         for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2040                 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
2041                     (!rt->rt6i_idev || rt->rt6i_idev->cnf.accept_ra != 2)) {
2042                         dst_hold(&rt->dst);
2043                         read_unlock_bh(&table->tb6_lock);
2044                         ip6_del_rt(rt);
2045                         goto restart;
2046                 }
2047         }
2048         read_unlock_bh(&table->tb6_lock);
2049 }
2050
2051 static void rtmsg_to_fib6_config(struct net *net,
2052                                  struct in6_rtmsg *rtmsg,
2053                                  struct fib6_config *cfg)
2054 {
2055         memset(cfg, 0, sizeof(*cfg));
2056
2057         cfg->fc_table = RT6_TABLE_MAIN;
2058         cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
2059         cfg->fc_metric = rtmsg->rtmsg_metric;
2060         cfg->fc_expires = rtmsg->rtmsg_info;
2061         cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
2062         cfg->fc_src_len = rtmsg->rtmsg_src_len;
2063         cfg->fc_flags = rtmsg->rtmsg_flags;
2064
2065         cfg->fc_nlinfo.nl_net = net;
2066
2067         cfg->fc_dst = rtmsg->rtmsg_dst;
2068         cfg->fc_src = rtmsg->rtmsg_src;
2069         cfg->fc_gateway = rtmsg->rtmsg_gateway;
2070 }
2071
2072 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
2073 {
2074         struct fib6_config cfg;
2075         struct in6_rtmsg rtmsg;
2076         int err;
2077
2078         switch (cmd) {
2079         case SIOCADDRT:         /* Add a route */
2080         case SIOCDELRT:         /* Delete a route */
2081                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2082                         return -EPERM;
2083                 err = copy_from_user(&rtmsg, arg,
2084                                      sizeof(struct in6_rtmsg));
2085                 if (err)
2086                         return -EFAULT;
2087
2088                 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
2089
2090                 rtnl_lock();
2091                 switch (cmd) {
2092                 case SIOCADDRT:
2093                         err = ip6_route_add(&cfg);
2094                         break;
2095                 case SIOCDELRT:
2096                         err = ip6_route_del(&cfg);
2097                         break;
2098                 default:
2099                         err = -EINVAL;
2100                 }
2101                 rtnl_unlock();
2102
2103                 return err;
2104         }
2105
2106         return -EINVAL;
2107 }
2108
2109 /*
2110  *      Drop the packet on the floor
2111  */
2112
2113 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
2114 {
2115         int type;
2116         struct dst_entry *dst = skb_dst(skb);
2117         switch (ipstats_mib_noroutes) {
2118         case IPSTATS_MIB_INNOROUTES:
2119                 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
2120                 if (type == IPV6_ADDR_ANY) {
2121                         IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2122                                       IPSTATS_MIB_INADDRERRORS);
2123                         break;
2124                 }
2125                 /* FALLTHROUGH */
2126         case IPSTATS_MIB_OUTNOROUTES:
2127                 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2128                               ipstats_mib_noroutes);
2129                 break;
2130         }
2131         icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
2132         kfree_skb(skb);
2133         return 0;
2134 }
2135
2136 static int ip6_pkt_discard(struct sk_buff *skb)
2137 {
2138         return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
2139 }
2140
2141 static int ip6_pkt_discard_out(struct sock *sk, struct sk_buff *skb)
2142 {
2143         skb->dev = skb_dst(skb)->dev;
2144         return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
2145 }
2146
2147 static int ip6_pkt_prohibit(struct sk_buff *skb)
2148 {
2149         return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
2150 }
2151
2152 static int ip6_pkt_prohibit_out(struct sock *sk, struct sk_buff *skb)
2153 {
2154         skb->dev = skb_dst(skb)->dev;
2155         return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
2156 }
2157
2158 /*
2159  *      Allocate a dst for local (unicast / anycast) address.
2160  */
2161
2162 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
2163                                     const struct in6_addr *addr,
2164                                     bool anycast)
2165 {
2166         struct net *net = dev_net(idev->dev);
2167         struct rt6_info *rt = ip6_dst_alloc(net, net->loopback_dev,
2168                                             DST_NOCOUNT, NULL);
2169         if (!rt)
2170                 return ERR_PTR(-ENOMEM);
2171
2172         in6_dev_hold(idev);
2173
2174         rt->dst.flags |= DST_HOST;
2175         rt->dst.input = ip6_input;
2176         rt->dst.output = ip6_output;
2177         rt->rt6i_idev = idev;
2178
2179         rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
2180         if (anycast)
2181                 rt->rt6i_flags |= RTF_ANYCAST;
2182         else
2183                 rt->rt6i_flags |= RTF_LOCAL;
2184
2185         rt->rt6i_gateway  = *addr;
2186         rt->rt6i_dst.addr = *addr;
2187         rt->rt6i_dst.plen = 128;
2188         rt->rt6i_table = fib6_get_table(net, RT6_TABLE_LOCAL);
2189
2190         atomic_set(&rt->dst.__refcnt, 1);
2191
2192         return rt;
2193 }
2194
2195 int ip6_route_get_saddr(struct net *net,
2196                         struct rt6_info *rt,
2197                         const struct in6_addr *daddr,
2198                         unsigned int prefs,
2199                         struct in6_addr *saddr)
2200 {
2201         struct inet6_dev *idev = ip6_dst_idev((struct dst_entry *)rt);
2202         int err = 0;
2203         if (rt->rt6i_prefsrc.plen)
2204                 *saddr = rt->rt6i_prefsrc.addr;
2205         else
2206                 err = ipv6_dev_get_saddr(net, idev ? idev->dev : NULL,
2207                                          daddr, prefs, saddr);
2208         return err;
2209 }
2210
2211 /* remove deleted ip from prefsrc entries */
2212 struct arg_dev_net_ip {
2213         struct net_device *dev;
2214         struct net *net;
2215         struct in6_addr *addr;
2216 };
2217
2218 static int fib6_remove_prefsrc(struct rt6_info *rt, void *arg)
2219 {
2220         struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
2221         struct net *net = ((struct arg_dev_net_ip *)arg)->net;
2222         struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
2223
2224         if (((void *)rt->dst.dev == dev || !dev) &&
2225             rt != net->ipv6.ip6_null_entry &&
2226             ipv6_addr_equal(addr, &rt->rt6i_prefsrc.addr)) {
2227                 /* remove prefsrc entry */
2228                 rt->rt6i_prefsrc.plen = 0;
2229         }
2230         return 0;
2231 }
2232
2233 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
2234 {
2235         struct net *net = dev_net(ifp->idev->dev);
2236         struct arg_dev_net_ip adni = {
2237                 .dev = ifp->idev->dev,
2238                 .net = net,
2239                 .addr = &ifp->addr,
2240         };
2241         fib6_clean_all(net, fib6_remove_prefsrc, &adni);
2242 }
2243
2244 #define RTF_RA_ROUTER           (RTF_ADDRCONF | RTF_DEFAULT | RTF_GATEWAY)
2245 #define RTF_CACHE_GATEWAY       (RTF_GATEWAY | RTF_CACHE)
2246
2247 /* Remove routers and update dst entries when gateway turn into host. */
2248 static int fib6_clean_tohost(struct rt6_info *rt, void *arg)
2249 {
2250         struct in6_addr *gateway = (struct in6_addr *)arg;
2251
2252         if ((((rt->rt6i_flags & RTF_RA_ROUTER) == RTF_RA_ROUTER) ||
2253              ((rt->rt6i_flags & RTF_CACHE_GATEWAY) == RTF_CACHE_GATEWAY)) &&
2254              ipv6_addr_equal(gateway, &rt->rt6i_gateway)) {
2255                 return -1;
2256         }
2257         return 0;
2258 }
2259
2260 void rt6_clean_tohost(struct net *net, struct in6_addr *gateway)
2261 {
2262         fib6_clean_all(net, fib6_clean_tohost, gateway);
2263 }
2264
2265 struct arg_dev_net {
2266         struct net_device *dev;
2267         struct net *net;
2268 };
2269
2270 static int fib6_ifdown(struct rt6_info *rt, void *arg)
2271 {
2272         const struct arg_dev_net *adn = arg;
2273         const struct net_device *dev = adn->dev;
2274
2275         if ((rt->dst.dev == dev || !dev) &&
2276             rt != adn->net->ipv6.ip6_null_entry)
2277                 return -1;
2278
2279         return 0;
2280 }
2281
2282 void rt6_ifdown(struct net *net, struct net_device *dev)
2283 {
2284         struct arg_dev_net adn = {
2285                 .dev = dev,
2286                 .net = net,
2287         };
2288
2289         fib6_clean_all(net, fib6_ifdown, &adn);
2290         icmp6_clean_all(fib6_ifdown, &adn);
2291 }
2292
2293 struct rt6_mtu_change_arg {
2294         struct net_device *dev;
2295         unsigned int mtu;
2296 };
2297
2298 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2299 {
2300         struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2301         struct inet6_dev *idev;
2302
2303         /* In IPv6 pmtu discovery is not optional,
2304            so that RTAX_MTU lock cannot disable it.
2305            We still use this lock to block changes
2306            caused by addrconf/ndisc.
2307         */
2308
2309         idev = __in6_dev_get(arg->dev);
2310         if (!idev)
2311                 return 0;
2312
2313         /* For administrative MTU increase, there is no way to discover
2314            IPv6 PMTU increase, so PMTU increase should be updated here.
2315            Since RFC 1981 doesn't include administrative MTU increase
2316            update PMTU increase is a MUST. (i.e. jumbo frame)
2317          */
2318         /*
2319            If new MTU is less than route PMTU, this new MTU will be the
2320            lowest MTU in the path, update the route PMTU to reflect PMTU
2321            decreases; if new MTU is greater than route PMTU, and the
2322            old MTU is the lowest MTU in the path, update the route PMTU
2323            to reflect the increase. In this case if the other nodes' MTU
2324            also have the lowest MTU, TOO BIG MESSAGE will be lead to
2325            PMTU discouvery.
2326          */
2327         if (rt->dst.dev == arg->dev &&
2328             !dst_metric_locked(&rt->dst, RTAX_MTU) &&
2329             (dst_mtu(&rt->dst) >= arg->mtu ||
2330              (dst_mtu(&rt->dst) < arg->mtu &&
2331               dst_mtu(&rt->dst) == idev->cnf.mtu6))) {
2332                 dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
2333         }
2334         return 0;
2335 }
2336
2337 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
2338 {
2339         struct rt6_mtu_change_arg arg = {
2340                 .dev = dev,
2341                 .mtu = mtu,
2342         };
2343
2344         fib6_clean_all(dev_net(dev), rt6_mtu_change_route, &arg);
2345 }
2346
2347 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2348         [RTA_GATEWAY]           = { .len = sizeof(struct in6_addr) },
2349         [RTA_OIF]               = { .type = NLA_U32 },
2350         [RTA_IIF]               = { .type = NLA_U32 },
2351         [RTA_PRIORITY]          = { .type = NLA_U32 },
2352         [RTA_METRICS]           = { .type = NLA_NESTED },
2353         [RTA_MULTIPATH]         = { .len = sizeof(struct rtnexthop) },
2354 };
2355
2356 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2357                               struct fib6_config *cfg)
2358 {
2359         struct rtmsg *rtm;
2360         struct nlattr *tb[RTA_MAX+1];
2361         int err;
2362
2363         err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2364         if (err < 0)
2365                 goto errout;
2366
2367         err = -EINVAL;
2368         rtm = nlmsg_data(nlh);
2369         memset(cfg, 0, sizeof(*cfg));
2370
2371         cfg->fc_table = rtm->rtm_table;
2372         cfg->fc_dst_len = rtm->rtm_dst_len;
2373         cfg->fc_src_len = rtm->rtm_src_len;
2374         cfg->fc_flags = RTF_UP;
2375         cfg->fc_protocol = rtm->rtm_protocol;
2376         cfg->fc_type = rtm->rtm_type;
2377
2378         if (rtm->rtm_type == RTN_UNREACHABLE ||
2379             rtm->rtm_type == RTN_BLACKHOLE ||
2380             rtm->rtm_type == RTN_PROHIBIT ||
2381             rtm->rtm_type == RTN_THROW)
2382                 cfg->fc_flags |= RTF_REJECT;
2383
2384         if (rtm->rtm_type == RTN_LOCAL)
2385                 cfg->fc_flags |= RTF_LOCAL;
2386
2387         cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid;
2388         cfg->fc_nlinfo.nlh = nlh;
2389         cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2390
2391         if (tb[RTA_GATEWAY]) {
2392                 nla_memcpy(&cfg->fc_gateway, tb[RTA_GATEWAY], 16);
2393                 cfg->fc_flags |= RTF_GATEWAY;
2394         }
2395
2396         if (tb[RTA_DST]) {
2397                 int plen = (rtm->rtm_dst_len + 7) >> 3;
2398
2399                 if (nla_len(tb[RTA_DST]) < plen)
2400                         goto errout;
2401
2402                 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2403         }
2404
2405         if (tb[RTA_SRC]) {
2406                 int plen = (rtm->rtm_src_len + 7) >> 3;
2407
2408                 if (nla_len(tb[RTA_SRC]) < plen)
2409                         goto errout;
2410
2411                 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2412         }
2413
2414         if (tb[RTA_PREFSRC])
2415                 nla_memcpy(&cfg->fc_prefsrc, tb[RTA_PREFSRC], 16);
2416
2417         if (tb[RTA_OIF])
2418                 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2419
2420         if (tb[RTA_PRIORITY])
2421                 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2422
2423         if (tb[RTA_METRICS]) {
2424                 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2425                 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2426         }
2427
2428         if (tb[RTA_TABLE])
2429                 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2430
2431         if (tb[RTA_MULTIPATH]) {
2432                 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
2433                 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
2434         }
2435
2436         err = 0;
2437 errout:
2438         return err;
2439 }
2440
2441 static int ip6_route_multipath(struct fib6_config *cfg, int add)
2442 {
2443         struct fib6_config r_cfg;
2444         struct rtnexthop *rtnh;
2445         int remaining;
2446         int attrlen;
2447         int err = 0, last_err = 0;
2448
2449 beginning:
2450         rtnh = (struct rtnexthop *)cfg->fc_mp;
2451         remaining = cfg->fc_mp_len;
2452
2453         /* Parse a Multipath Entry */
2454         while (rtnh_ok(rtnh, remaining)) {
2455                 memcpy(&r_cfg, cfg, sizeof(*cfg));
2456                 if (rtnh->rtnh_ifindex)
2457                         r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
2458
2459                 attrlen = rtnh_attrlen(rtnh);
2460                 if (attrlen > 0) {
2461                         struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
2462
2463                         nla = nla_find(attrs, attrlen, RTA_GATEWAY);
2464                         if (nla) {
2465                                 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
2466                                 r_cfg.fc_flags |= RTF_GATEWAY;
2467                         }
2468                 }
2469                 err = add ? ip6_route_add(&r_cfg) : ip6_route_del(&r_cfg);
2470                 if (err) {
2471                         last_err = err;
2472                         /* If we are trying to remove a route, do not stop the
2473                          * loop when ip6_route_del() fails (because next hop is
2474                          * already gone), we should try to remove all next hops.
2475                          */
2476                         if (add) {
2477                                 /* If add fails, we should try to delete all
2478                                  * next hops that have been already added.
2479                                  */
2480                                 add = 0;
2481                                 goto beginning;
2482                         }
2483                 }
2484                 /* Because each route is added like a single route we remove
2485                  * this flag after the first nexthop (if there is a collision,
2486                  * we have already fail to add the first nexthop:
2487                  * fib6_add_rt2node() has reject it).
2488                  */
2489                 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~NLM_F_EXCL;
2490                 rtnh = rtnh_next(rtnh, &remaining);
2491         }
2492
2493         return last_err;
2494 }
2495
2496 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh)
2497 {
2498         struct fib6_config cfg;
2499         int err;
2500
2501         err = rtm_to_fib6_config(skb, nlh, &cfg);
2502         if (err < 0)
2503                 return err;
2504
2505         if (cfg.fc_mp)
2506                 return ip6_route_multipath(&cfg, 0);
2507         else
2508                 return ip6_route_del(&cfg);
2509 }
2510
2511 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh)
2512 {
2513         struct fib6_config cfg;
2514         int err;
2515
2516         err = rtm_to_fib6_config(skb, nlh, &cfg);
2517         if (err < 0)
2518                 return err;
2519
2520         if (cfg.fc_mp)
2521                 return ip6_route_multipath(&cfg, 1);
2522         else
2523                 return ip6_route_add(&cfg);
2524 }
2525
2526 static inline size_t rt6_nlmsg_size(void)
2527 {
2528         return NLMSG_ALIGN(sizeof(struct rtmsg))
2529                + nla_total_size(16) /* RTA_SRC */
2530                + nla_total_size(16) /* RTA_DST */
2531                + nla_total_size(16) /* RTA_GATEWAY */
2532                + nla_total_size(16) /* RTA_PREFSRC */
2533                + nla_total_size(4) /* RTA_TABLE */
2534                + nla_total_size(4) /* RTA_IIF */
2535                + nla_total_size(4) /* RTA_OIF */
2536                + nla_total_size(4) /* RTA_PRIORITY */
2537                + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
2538                + nla_total_size(sizeof(struct rta_cacheinfo));
2539 }
2540
2541 static int rt6_fill_node(struct net *net,
2542                          struct sk_buff *skb, struct rt6_info *rt,
2543                          struct in6_addr *dst, struct in6_addr *src,
2544                          int iif, int type, u32 portid, u32 seq,
2545                          int prefix, int nowait, unsigned int flags)
2546 {
2547         struct rtmsg *rtm;
2548         struct nlmsghdr *nlh;
2549         long expires;
2550         u32 table;
2551
2552         if (prefix) {   /* user wants prefix routes only */
2553                 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
2554                         /* success since this is not a prefix route */
2555                         return 1;
2556                 }
2557         }
2558
2559         nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
2560         if (!nlh)
2561                 return -EMSGSIZE;
2562
2563         rtm = nlmsg_data(nlh);
2564         rtm->rtm_family = AF_INET6;
2565         rtm->rtm_dst_len = rt->rt6i_dst.plen;
2566         rtm->rtm_src_len = rt->rt6i_src.plen;
2567         rtm->rtm_tos = 0;
2568         if (rt->rt6i_table)
2569                 table = rt->rt6i_table->tb6_id;
2570         else
2571                 table = RT6_TABLE_UNSPEC;
2572         rtm->rtm_table = table;
2573         if (nla_put_u32(skb, RTA_TABLE, table))
2574                 goto nla_put_failure;
2575         if (rt->rt6i_flags & RTF_REJECT) {
2576                 switch (rt->dst.error) {
2577                 case -EINVAL:
2578                         rtm->rtm_type = RTN_BLACKHOLE;
2579                         break;
2580                 case -EACCES:
2581                         rtm->rtm_type = RTN_PROHIBIT;
2582                         break;
2583                 case -EAGAIN:
2584                         rtm->rtm_type = RTN_THROW;
2585                         break;
2586                 default:
2587                         rtm->rtm_type = RTN_UNREACHABLE;
2588                         break;
2589                 }
2590         }
2591         else if (rt->rt6i_flags & RTF_LOCAL)
2592                 rtm->rtm_type = RTN_LOCAL;
2593         else if (rt->dst.dev && (rt->dst.dev->flags & IFF_LOOPBACK))
2594                 rtm->rtm_type = RTN_LOCAL;
2595         else
2596                 rtm->rtm_type = RTN_UNICAST;
2597         rtm->rtm_flags = 0;
2598         rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2599         rtm->rtm_protocol = rt->rt6i_protocol;
2600         if (rt->rt6i_flags & RTF_DYNAMIC)
2601                 rtm->rtm_protocol = RTPROT_REDIRECT;
2602         else if (rt->rt6i_flags & RTF_ADDRCONF) {
2603                 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ROUTEINFO))
2604                         rtm->rtm_protocol = RTPROT_RA;
2605                 else
2606                         rtm->rtm_protocol = RTPROT_KERNEL;
2607         }
2608
2609         if (rt->rt6i_flags & RTF_CACHE)
2610                 rtm->rtm_flags |= RTM_F_CLONED;
2611
2612         if (dst) {
2613                 if (nla_put(skb, RTA_DST, 16, dst))
2614                         goto nla_put_failure;
2615                 rtm->rtm_dst_len = 128;
2616         } else if (rtm->rtm_dst_len)
2617                 if (nla_put(skb, RTA_DST, 16, &rt->rt6i_dst.addr))
2618                         goto nla_put_failure;
2619 #ifdef CONFIG_IPV6_SUBTREES
2620         if (src) {
2621                 if (nla_put(skb, RTA_SRC, 16, src))
2622                         goto nla_put_failure;
2623                 rtm->rtm_src_len = 128;
2624         } else if (rtm->rtm_src_len &&
2625                    nla_put(skb, RTA_SRC, 16, &rt->rt6i_src.addr))
2626                 goto nla_put_failure;
2627 #endif
2628         if (iif) {
2629 #ifdef CONFIG_IPV6_MROUTE
2630                 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
2631                         int err = ip6mr_get_route(net, skb, rtm, nowait);
2632                         if (err <= 0) {
2633                                 if (!nowait) {
2634                                         if (err == 0)
2635                                                 return 0;
2636                                         goto nla_put_failure;
2637                                 } else {
2638                                         if (err == -EMSGSIZE)
2639                                                 goto nla_put_failure;
2640                                 }
2641                         }
2642                 } else
2643 #endif
2644                         if (nla_put_u32(skb, RTA_IIF, iif))
2645                                 goto nla_put_failure;
2646         } else if (dst) {
2647                 struct in6_addr saddr_buf;
2648                 if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0 &&
2649                     nla_put(skb, RTA_PREFSRC, 16, &saddr_buf))
2650                         goto nla_put_failure;
2651         }
2652
2653         if (rt->rt6i_prefsrc.plen) {
2654                 struct in6_addr saddr_buf;
2655                 saddr_buf = rt->rt6i_prefsrc.addr;
2656                 if (nla_put(skb, RTA_PREFSRC, 16, &saddr_buf))
2657                         goto nla_put_failure;
2658         }
2659
2660         if (rtnetlink_put_metrics(skb, dst_metrics_ptr(&rt->dst)) < 0)
2661                 goto nla_put_failure;
2662
2663         if (rt->rt6i_flags & RTF_GATEWAY) {
2664                 if (nla_put(skb, RTA_GATEWAY, 16, &rt->rt6i_gateway) < 0)
2665                         goto nla_put_failure;
2666         }
2667
2668         if (rt->dst.dev &&
2669             nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex))
2670                 goto nla_put_failure;
2671         if (nla_put_u32(skb, RTA_PRIORITY, rt->rt6i_metric))
2672                 goto nla_put_failure;
2673
2674         expires = (rt->rt6i_flags & RTF_EXPIRES) ? rt->dst.expires - jiffies : 0;
2675
2676         if (rtnl_put_cacheinfo(skb, &rt->dst, 0, expires, rt->dst.error) < 0)
2677                 goto nla_put_failure;
2678
2679         return nlmsg_end(skb, nlh);
2680
2681 nla_put_failure:
2682         nlmsg_cancel(skb, nlh);
2683         return -EMSGSIZE;
2684 }
2685
2686 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
2687 {
2688         struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
2689         int prefix;
2690
2691         if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
2692                 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
2693                 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
2694         } else
2695                 prefix = 0;
2696
2697         return rt6_fill_node(arg->net,
2698                      arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
2699                      NETLINK_CB(arg->cb->skb).portid, arg->cb->nlh->nlmsg_seq,
2700                      prefix, 0, NLM_F_MULTI);
2701 }
2702
2703 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh)
2704 {
2705         struct net *net = sock_net(in_skb->sk);
2706         struct nlattr *tb[RTA_MAX+1];
2707         struct rt6_info *rt;
2708         struct sk_buff *skb;
2709         struct rtmsg *rtm;
2710         struct flowi6 fl6;
2711         int err, iif = 0, oif = 0;
2712
2713         err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2714         if (err < 0)
2715                 goto errout;
2716
2717         err = -EINVAL;
2718         memset(&fl6, 0, sizeof(fl6));
2719
2720         if (tb[RTA_SRC]) {
2721                 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
2722                         goto errout;
2723
2724                 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
2725         }
2726
2727         if (tb[RTA_DST]) {
2728                 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
2729                         goto errout;
2730
2731                 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
2732         }
2733
2734         if (tb[RTA_IIF])
2735                 iif = nla_get_u32(tb[RTA_IIF]);
2736
2737         if (tb[RTA_OIF])
2738                 oif = nla_get_u32(tb[RTA_OIF]);
2739
2740         if (tb[RTA_MARK])
2741                 fl6.flowi6_mark = nla_get_u32(tb[RTA_MARK]);
2742
2743         if (iif) {
2744                 struct net_device *dev;
2745                 int flags = 0;
2746
2747                 dev = __dev_get_by_index(net, iif);
2748                 if (!dev) {
2749                         err = -ENODEV;
2750                         goto errout;
2751                 }
2752
2753                 fl6.flowi6_iif = iif;
2754
2755                 if (!ipv6_addr_any(&fl6.saddr))
2756                         flags |= RT6_LOOKUP_F_HAS_SADDR;
2757
2758                 rt = (struct rt6_info *)ip6_route_input_lookup(net, dev, &fl6,
2759                                                                flags);
2760         } else {
2761                 fl6.flowi6_oif = oif;
2762
2763                 rt = (struct rt6_info *)ip6_route_output(net, NULL, &fl6);
2764         }
2765
2766         skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
2767         if (!skb) {
2768                 ip6_rt_put(rt);
2769                 err = -ENOBUFS;
2770                 goto errout;
2771         }
2772
2773         /* Reserve room for dummy headers, this skb can pass
2774            through good chunk of routing engine.
2775          */
2776         skb_reset_mac_header(skb);
2777         skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
2778
2779         skb_dst_set(skb, &rt->dst);
2780
2781         err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
2782                             RTM_NEWROUTE, NETLINK_CB(in_skb).portid,
2783                             nlh->nlmsg_seq, 0, 0, 0);
2784         if (err < 0) {
2785                 kfree_skb(skb);
2786                 goto errout;
2787         }
2788
2789         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
2790 errout:
2791         return err;
2792 }
2793
2794 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info)
2795 {
2796         struct sk_buff *skb;
2797         struct net *net = info->nl_net;
2798         u32 seq;
2799         int err;
2800
2801         err = -ENOBUFS;
2802         seq = info->nlh ? info->nlh->nlmsg_seq : 0;
2803
2804         skb = nlmsg_new(rt6_nlmsg_size(), gfp_any());
2805         if (!skb)
2806                 goto errout;
2807
2808         err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
2809                                 event, info->portid, seq, 0, 0, 0);
2810         if (err < 0) {
2811                 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
2812                 WARN_ON(err == -EMSGSIZE);
2813                 kfree_skb(skb);
2814                 goto errout;
2815         }
2816         rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
2817                     info->nlh, gfp_any());
2818         return;
2819 errout:
2820         if (err < 0)
2821                 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
2822 }
2823
2824 static int ip6_route_dev_notify(struct notifier_block *this,
2825                                 unsigned long event, void *ptr)
2826 {
2827         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2828         struct net *net = dev_net(dev);
2829
2830         if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
2831                 net->ipv6.ip6_null_entry->dst.dev = dev;
2832                 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
2833 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2834                 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
2835                 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
2836                 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
2837                 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
2838 #endif
2839         }
2840
2841         return NOTIFY_OK;
2842 }
2843
2844 /*
2845  *      /proc
2846  */
2847
2848 #ifdef CONFIG_PROC_FS
2849
2850 static const struct file_operations ipv6_route_proc_fops = {
2851         .owner          = THIS_MODULE,
2852         .open           = ipv6_route_open,
2853         .read           = seq_read,
2854         .llseek         = seq_lseek,
2855         .release        = seq_release_net,
2856 };
2857
2858 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
2859 {
2860         struct net *net = (struct net *)seq->private;
2861         seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
2862                    net->ipv6.rt6_stats->fib_nodes,
2863                    net->ipv6.rt6_stats->fib_route_nodes,
2864                    net->ipv6.rt6_stats->fib_rt_alloc,
2865                    net->ipv6.rt6_stats->fib_rt_entries,
2866                    net->ipv6.rt6_stats->fib_rt_cache,
2867                    dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
2868                    net->ipv6.rt6_stats->fib_discarded_routes);
2869
2870         return 0;
2871 }
2872
2873 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
2874 {
2875         return single_open_net(inode, file, rt6_stats_seq_show);
2876 }
2877
2878 static const struct file_operations rt6_stats_seq_fops = {
2879         .owner   = THIS_MODULE,
2880         .open    = rt6_stats_seq_open,
2881         .read    = seq_read,
2882         .llseek  = seq_lseek,
2883         .release = single_release_net,
2884 };
2885 #endif  /* CONFIG_PROC_FS */
2886
2887 #ifdef CONFIG_SYSCTL
2888
2889 static
2890 int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
2891                               void __user *buffer, size_t *lenp, loff_t *ppos)
2892 {
2893         struct net *net;
2894         int delay;
2895         if (!write)
2896                 return -EINVAL;
2897
2898         net = (struct net *)ctl->extra1;
2899         delay = net->ipv6.sysctl.flush_delay;
2900         proc_dointvec(ctl, write, buffer, lenp, ppos);
2901         fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
2902         return 0;
2903 }
2904
2905 struct ctl_table ipv6_route_table_template[] = {
2906         {
2907                 .procname       =       "flush",
2908                 .data           =       &init_net.ipv6.sysctl.flush_delay,
2909                 .maxlen         =       sizeof(int),
2910                 .mode           =       0200,
2911                 .proc_handler   =       ipv6_sysctl_rtcache_flush
2912         },
2913         {
2914                 .procname       =       "gc_thresh",
2915                 .data           =       &ip6_dst_ops_template.gc_thresh,
2916                 .maxlen         =       sizeof(int),
2917                 .mode           =       0644,
2918                 .proc_handler   =       proc_dointvec,
2919         },
2920         {
2921                 .procname       =       "max_size",
2922                 .data           =       &init_net.ipv6.sysctl.ip6_rt_max_size,
2923                 .maxlen         =       sizeof(int),
2924                 .mode           =       0644,
2925                 .proc_handler   =       proc_dointvec,
2926         },
2927         {
2928                 .procname       =       "gc_min_interval",
2929                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2930                 .maxlen         =       sizeof(int),
2931                 .mode           =       0644,
2932                 .proc_handler   =       proc_dointvec_jiffies,
2933         },
2934         {
2935                 .procname       =       "gc_timeout",
2936                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
2937                 .maxlen         =       sizeof(int),
2938                 .mode           =       0644,
2939                 .proc_handler   =       proc_dointvec_jiffies,
2940         },
2941         {
2942                 .procname       =       "gc_interval",
2943                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_interval,
2944                 .maxlen         =       sizeof(int),
2945                 .mode           =       0644,
2946                 .proc_handler   =       proc_dointvec_jiffies,
2947         },
2948         {
2949                 .procname       =       "gc_elasticity",
2950                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
2951                 .maxlen         =       sizeof(int),
2952                 .mode           =       0644,
2953                 .proc_handler   =       proc_dointvec,
2954         },
2955         {
2956                 .procname       =       "mtu_expires",
2957                 .data           =       &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
2958                 .maxlen         =       sizeof(int),
2959                 .mode           =       0644,
2960                 .proc_handler   =       proc_dointvec_jiffies,
2961         },
2962         {
2963                 .procname       =       "min_adv_mss",
2964                 .data           =       &init_net.ipv6.sysctl.ip6_rt_min_advmss,
2965                 .maxlen         =       sizeof(int),
2966                 .mode           =       0644,
2967                 .proc_handler   =       proc_dointvec,
2968         },
2969         {
2970                 .procname       =       "gc_min_interval_ms",
2971                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2972                 .maxlen         =       sizeof(int),
2973                 .mode           =       0644,
2974                 .proc_handler   =       proc_dointvec_ms_jiffies,
2975         },
2976         { }
2977 };
2978
2979 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
2980 {
2981         struct ctl_table *table;
2982
2983         table = kmemdup(ipv6_route_table_template,
2984                         sizeof(ipv6_route_table_template),
2985                         GFP_KERNEL);
2986
2987         if (table) {
2988                 table[0].data = &net->ipv6.sysctl.flush_delay;
2989                 table[0].extra1 = net;
2990                 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
2991                 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
2992                 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2993                 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
2994                 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
2995                 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
2996                 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
2997                 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
2998                 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2999
3000                 /* Don't export sysctls to unprivileged users */
3001                 if (net->user_ns != &init_user_ns)
3002                         table[0].procname = NULL;
3003         }
3004
3005         return table;
3006 }
3007 #endif
3008
3009 static int __net_init ip6_route_net_init(struct net *net)
3010 {
3011         int ret = -ENOMEM;
3012
3013         memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
3014                sizeof(net->ipv6.ip6_dst_ops));
3015
3016         if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
3017                 goto out_ip6_dst_ops;
3018
3019         net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
3020                                            sizeof(*net->ipv6.ip6_null_entry),
3021                                            GFP_KERNEL);
3022         if (!net->ipv6.ip6_null_entry)
3023                 goto out_ip6_dst_entries;
3024         net->ipv6.ip6_null_entry->dst.path =
3025                 (struct dst_entry *)net->ipv6.ip6_null_entry;
3026         net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3027         dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
3028                          ip6_template_metrics, true);
3029
3030 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3031         net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
3032                                                sizeof(*net->ipv6.ip6_prohibit_entry),
3033                                                GFP_KERNEL);
3034         if (!net->ipv6.ip6_prohibit_entry)
3035                 goto out_ip6_null_entry;
3036         net->ipv6.ip6_prohibit_entry->dst.path =
3037                 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
3038         net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3039         dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
3040                          ip6_template_metrics, true);
3041
3042         net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
3043                                                sizeof(*net->ipv6.ip6_blk_hole_entry),
3044                                                GFP_KERNEL);
3045         if (!net->ipv6.ip6_blk_hole_entry)
3046                 goto out_ip6_prohibit_entry;
3047         net->ipv6.ip6_blk_hole_entry->dst.path =
3048                 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
3049         net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3050         dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
3051                          ip6_template_metrics, true);
3052 #endif
3053
3054         net->ipv6.sysctl.flush_delay = 0;
3055         net->ipv6.sysctl.ip6_rt_max_size = 4096;
3056         net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
3057         net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
3058         net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
3059         net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
3060         net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
3061         net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
3062
3063         net->ipv6.ip6_rt_gc_expire = 30*HZ;
3064
3065         ret = 0;
3066 out:
3067         return ret;
3068
3069 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3070 out_ip6_prohibit_entry:
3071         kfree(net->ipv6.ip6_prohibit_entry);
3072 out_ip6_null_entry:
3073         kfree(net->ipv6.ip6_null_entry);
3074 #endif
3075 out_ip6_dst_entries:
3076         dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3077 out_ip6_dst_ops:
3078         goto out;
3079 }
3080
3081 static void __net_exit ip6_route_net_exit(struct net *net)
3082 {
3083         kfree(net->ipv6.ip6_null_entry);
3084 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3085         kfree(net->ipv6.ip6_prohibit_entry);
3086         kfree(net->ipv6.ip6_blk_hole_entry);
3087 #endif
3088         dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3089 }
3090
3091 static int __net_init ip6_route_net_init_late(struct net *net)
3092 {
3093 #ifdef CONFIG_PROC_FS
3094         proc_create("ipv6_route", 0, net->proc_net, &ipv6_route_proc_fops);
3095         proc_create("rt6_stats", S_IRUGO, net->proc_net, &rt6_stats_seq_fops);
3096 #endif
3097         return 0;
3098 }
3099
3100 static void __net_exit ip6_route_net_exit_late(struct net *net)
3101 {
3102 #ifdef CONFIG_PROC_FS
3103         remove_proc_entry("ipv6_route", net->proc_net);
3104         remove_proc_entry("rt6_stats", net->proc_net);
3105 #endif
3106 }
3107
3108 static struct pernet_operations ip6_route_net_ops = {
3109         .init = ip6_route_net_init,
3110         .exit = ip6_route_net_exit,
3111 };
3112
3113 static int __net_init ipv6_inetpeer_init(struct net *net)
3114 {
3115         struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
3116
3117         if (!bp)
3118                 return -ENOMEM;
3119         inet_peer_base_init(bp);
3120         net->ipv6.peers = bp;
3121         return 0;
3122 }
3123
3124 static void __net_exit ipv6_inetpeer_exit(struct net *net)
3125 {
3126         struct inet_peer_base *bp = net->ipv6.peers;
3127
3128         net->ipv6.peers = NULL;
3129         inetpeer_invalidate_tree(bp);
3130         kfree(bp);
3131 }
3132
3133 static struct pernet_operations ipv6_inetpeer_ops = {
3134         .init   =       ipv6_inetpeer_init,
3135         .exit   =       ipv6_inetpeer_exit,
3136 };
3137
3138 static struct pernet_operations ip6_route_net_late_ops = {
3139         .init = ip6_route_net_init_late,
3140         .exit = ip6_route_net_exit_late,
3141 };
3142
3143 static struct notifier_block ip6_route_dev_notifier = {
3144         .notifier_call = ip6_route_dev_notify,
3145         .priority = 0,
3146 };
3147
3148 int __init ip6_route_init(void)
3149 {
3150         int ret;
3151
3152         ret = -ENOMEM;
3153         ip6_dst_ops_template.kmem_cachep =
3154                 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
3155                                   SLAB_HWCACHE_ALIGN, NULL);
3156         if (!ip6_dst_ops_template.kmem_cachep)
3157                 goto out;
3158
3159         ret = dst_entries_init(&ip6_dst_blackhole_ops);
3160         if (ret)
3161                 goto out_kmem_cache;
3162
3163         ret = register_pernet_subsys(&ipv6_inetpeer_ops);
3164         if (ret)
3165                 goto out_dst_entries;
3166
3167         ret = register_pernet_subsys(&ip6_route_net_ops);
3168         if (ret)
3169                 goto out_register_inetpeer;
3170
3171         ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
3172
3173         /* Registering of the loopback is done before this portion of code,
3174          * the loopback reference in rt6_info will not be taken, do it
3175          * manually for init_net */
3176         init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
3177         init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3178   #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3179         init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
3180         init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3181         init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
3182         init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3183   #endif
3184         ret = fib6_init();
3185         if (ret)
3186                 goto out_register_subsys;
3187
3188         ret = xfrm6_init();
3189         if (ret)
3190                 goto out_fib6_init;
3191
3192         ret = fib6_rules_init();
3193         if (ret)
3194                 goto xfrm6_init;
3195
3196         ret = register_pernet_subsys(&ip6_route_net_late_ops);
3197         if (ret)
3198                 goto fib6_rules_init;
3199
3200         ret = -ENOBUFS;
3201         if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL, NULL) ||
3202             __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL, NULL) ||
3203             __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL, NULL))
3204                 goto out_register_late_subsys;
3205
3206         ret = register_netdevice_notifier(&ip6_route_dev_notifier);
3207         if (ret)
3208                 goto out_register_late_subsys;
3209
3210 out:
3211         return ret;
3212
3213 out_register_late_subsys:
3214         unregister_pernet_subsys(&ip6_route_net_late_ops);
3215 fib6_rules_init:
3216         fib6_rules_cleanup();
3217 xfrm6_init:
3218         xfrm6_fini();
3219 out_fib6_init:
3220         fib6_gc_cleanup();
3221 out_register_subsys:
3222         unregister_pernet_subsys(&ip6_route_net_ops);
3223 out_register_inetpeer:
3224         unregister_pernet_subsys(&ipv6_inetpeer_ops);
3225 out_dst_entries:
3226         dst_entries_destroy(&ip6_dst_blackhole_ops);
3227 out_kmem_cache:
3228         kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3229         goto out;
3230 }
3231
3232 void ip6_route_cleanup(void)
3233 {
3234         unregister_netdevice_notifier(&ip6_route_dev_notifier);
3235         unregister_pernet_subsys(&ip6_route_net_late_ops);
3236         fib6_rules_cleanup();
3237         xfrm6_fini();
3238         fib6_gc_cleanup();
3239         unregister_pernet_subsys(&ipv6_inetpeer_ops);
3240         unregister_pernet_subsys(&ip6_route_net_ops);
3241         dst_entries_destroy(&ip6_dst_blackhole_ops);
3242         kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3243 }