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 = -2,
  70         RT6_NUD_FAIL_SOFT = -1,
  71         RT6_NUD_SUCCEED = 1
  72 };
  73
  74 static struct rt6_info *ip6_rt_copy(struct rt6_info *ort,
  75                                     const struct in6_addr *dest);
  76 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie);
  77 static unsigned int      ip6_default_advmss(const struct dst_entry *dst);
  78 static unsigned int      ip6_mtu(const struct dst_entry *dst);
  79 static struct dst_entry *ip6_negative_advice(struct dst_entry *);
  80 static void             ip6_dst_destroy(struct dst_entry *);
  81 static void             ip6_dst_ifdown(struct dst_entry *,
  82                                        struct net_device *dev, int how);
  83 static int               ip6_dst_gc(struct dst_ops *ops);
  84
  85 static int              ip6_pkt_discard(struct sk_buff *skb);
  86 static int              ip6_pkt_discard_out(struct sk_buff *skb);
  87 static int              ip6_pkt_prohibit(struct sk_buff *skb);
  88 static int              ip6_pkt_prohibit_out(struct sk_buff *skb);
  89 static void             ip6_link_failure(struct sk_buff *skb);
  90 static void             ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
  91                                            struct sk_buff *skb, u32 mtu);
  92 static void             rt6_do_redirect(struct dst_entry *dst, struct sock *sk,
  93                                         struct sk_buff *skb);
  94 static int rt6_score_route(struct rt6_info *rt, int oif, int strict);
  95
  96 #ifdef CONFIG_IPV6_ROUTE_INFO
  97 static struct rt6_info *rt6_add_route_info(struct net *net,
  98                                            const struct in6_addr *prefix, int prefixlen,
  99                                            const struct in6_addr *gwaddr, int ifindex,
 100                                            unsigned int pref);
 101 static struct rt6_info *rt6_get_route_info(struct net *net,
 102                                            const struct in6_addr *prefix, int prefixlen,
 103                                            const struct in6_addr *gwaddr, int ifindex);
 104 #endif
 105
 106 static u32 *ipv6_cow_metrics(struct dst_entry *dst, unsigned long old)
 107 {
 108         struct rt6_info *rt = (struct rt6_info *) dst;
 109         struct inet_peer *peer;
 110         u32 *p = NULL;
 111
 112         if (!(rt->dst.flags & DST_HOST))
 113                 return NULL;
 114
 115         peer = rt6_get_peer_create(rt);
 116         if (peer) {
 117                 u32 *old_p = __DST_METRICS_PTR(old);
 118                 unsigned long prev, new;
 119
 120                 p = peer->metrics;
 121                 if (inet_metrics_new(peer))
 122                         memcpy(p, old_p, sizeof(u32) * RTAX_MAX);
 123
 124                 new = (unsigned long) p;
 125                 prev = cmpxchg(&dst->_metrics, old, new);
 126
 127                 if (prev != old) {
 128                         p = __DST_METRICS_PTR(prev);
 129                         if (prev & DST_METRICS_READ_ONLY)
 130                                 p = NULL;
 131                 }
 132         }
 133         return p;
 134 }
 135
 136 static inline const void *choose_neigh_daddr(struct rt6_info *rt,
 137                                              struct sk_buff *skb,
 138                                              const void *daddr)
 139 {
 140         struct in6_addr *p = &rt->rt6i_gateway;
 141
 142         if (!ipv6_addr_any(p))
 143                 return (const void *) p;
 144         else if (skb)
 145                 return &ipv6_hdr(skb)->daddr;
 146         return daddr;
 147 }
 148
 149 static struct neighbour *ip6_neigh_lookup(const struct dst_entry *dst,
 150                                           struct sk_buff *skb,
 151                                           const void *daddr)
 152 {
 153         struct rt6_info *rt = (struct rt6_info *) dst;
 154         struct neighbour *n;
 155
 156         daddr = choose_neigh_daddr(rt, skb, daddr);
 157         n = __ipv6_neigh_lookup(dst->dev, daddr);
 158         if (n)
 159                 return n;
 160         return neigh_create(&nd_tbl, daddr, dst->dev);
 161 }
 162
 163 static struct dst_ops ip6_dst_ops_template = {
 164         .family                 =       AF_INET6,
 165         .protocol               =       cpu_to_be16(ETH_P_IPV6),
 166         .gc                     =       ip6_dst_gc,
 167         .gc_thresh              =       1024,
 168         .check                  =       ip6_dst_check,
 169         .default_advmss         =       ip6_default_advmss,
 170         .mtu                    =       ip6_mtu,
 171         .cow_metrics            =       ipv6_cow_metrics,
 172         .destroy                =       ip6_dst_destroy,
 173         .ifdown                 =       ip6_dst_ifdown,
 174         .negative_advice        =       ip6_negative_advice,
 175         .link_failure           =       ip6_link_failure,
 176         .update_pmtu            =       ip6_rt_update_pmtu,
 177         .redirect               =       rt6_do_redirect,
 178         .local_out              =       __ip6_local_out,
 179         .neigh_lookup           =       ip6_neigh_lookup,
 180 };
 181
 182 static unsigned int ip6_blackhole_mtu(const struct dst_entry *dst)
 183 {
 184         unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
 185
 186         return mtu ? : dst->dev->mtu;
 187 }
 188
 189 static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, struct sock *sk,
 190                                          struct sk_buff *skb, u32 mtu)
 191 {
 192 }
 193
 194 static void ip6_rt_blackhole_redirect(struct dst_entry *dst, struct sock *sk,
 195                                       struct sk_buff *skb)
 196 {
 197 }
 198
 199 static u32 *ip6_rt_blackhole_cow_metrics(struct dst_entry *dst,
 200                                          unsigned long old)
 201 {
 202         return NULL;
 203 }
 204
 205 static struct dst_ops ip6_dst_blackhole_ops = {
 206         .family                 =       AF_INET6,
 207         .protocol               =       cpu_to_be16(ETH_P_IPV6),
 208         .destroy                =       ip6_dst_destroy,
 209         .check                  =       ip6_dst_check,
 210         .mtu                    =       ip6_blackhole_mtu,
 211         .default_advmss         =       ip6_default_advmss,
 212         .update_pmtu            =       ip6_rt_blackhole_update_pmtu,
 213         .redirect               =       ip6_rt_blackhole_redirect,
 214         .cow_metrics            =       ip6_rt_blackhole_cow_metrics,
 215         .neigh_lookup           =       ip6_neigh_lookup,
 216 };
 217
 218 static const u32 ip6_template_metrics[RTAX_MAX] = {
 219         [RTAX_HOPLIMIT - 1] = 0,
 220 };
 221
 222 static const struct rt6_info ip6_null_entry_template = {
 223         .dst = {
 224                 .__refcnt       = ATOMIC_INIT(1),
 225                 .__use          = 1,
 226                 .obsolete       = DST_OBSOLETE_FORCE_CHK,
 227                 .error          = -ENETUNREACH,
 228                 .input          = ip6_pkt_discard,
 229                 .output         = ip6_pkt_discard_out,
 230         },
 231         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
 232         .rt6i_protocol  = RTPROT_KERNEL,
 233         .rt6i_metric    = ~(u32) 0,
 234         .rt6i_ref       = ATOMIC_INIT(1),
 235 };
 236
 237 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
 238
 239 static const struct rt6_info ip6_prohibit_entry_template = {
 240         .dst = {
 241                 .__refcnt       = ATOMIC_INIT(1),
 242                 .__use          = 1,
 243                 .obsolete       = DST_OBSOLETE_FORCE_CHK,
 244                 .error          = -EACCES,
 245                 .input          = ip6_pkt_prohibit,
 246                 .output         = ip6_pkt_prohibit_out,
 247         },
 248         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
 249         .rt6i_protocol  = RTPROT_KERNEL,
 250         .rt6i_metric    = ~(u32) 0,
 251         .rt6i_ref       = ATOMIC_INIT(1),
 252 };
 253
 254 static const struct rt6_info ip6_blk_hole_entry_template = {
 255         .dst = {
 256                 .__refcnt       = ATOMIC_INIT(1),
 257                 .__use          = 1,
 258                 .obsolete       = DST_OBSOLETE_FORCE_CHK,
 259                 .error          = -EINVAL,
 260                 .input          = dst_discard,
 261                 .output         = dst_discard,
 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 #endif
 270
 271 /* allocate dst with ip6_dst_ops */
 272 static inline struct rt6_info *ip6_dst_alloc(struct net *net,
 273                                              struct net_device *dev,
 274                                              int flags,
 275                                              struct fib6_table *table)
 276 {
 277         struct rt6_info *rt = dst_alloc(&net->ipv6.ip6_dst_ops, dev,
 278                                         0, DST_OBSOLETE_FORCE_CHK, flags);
 279
 280         if (rt) {
 281                 struct dst_entry *dst = &rt->dst;
 282
 283                 memset(dst + 1, 0, sizeof(*rt) - sizeof(*dst));
 284                 rt6_init_peer(rt, table ? &table->tb6_peers : net->ipv6.peers);
 285                 rt->rt6i_genid = rt_genid_ipv6(net);
 286                 INIT_LIST_HEAD(&rt->rt6i_siblings);
 287         }
 288         return rt;
 289 }
 290
 291 static void ip6_dst_destroy(struct dst_entry *dst)
 292 {
 293         struct rt6_info *rt = (struct rt6_info *)dst;
 294         struct inet6_dev *idev = rt->rt6i_idev;
 295         struct dst_entry *from = dst->from;
 296
 297         if (!(rt->dst.flags & DST_HOST))
 298                 dst_destroy_metrics_generic(dst);
 299
 300         if (idev) {
 301                 rt->rt6i_idev = NULL;
 302                 in6_dev_put(idev);
 303         }
 304
 305         dst->from = NULL;
 306         dst_release(from);
 307
 308         if (rt6_has_peer(rt)) {
 309                 struct inet_peer *peer = rt6_peer_ptr(rt);
 310                 inet_putpeer(peer);
 311         }
 312 }
 313
 314 void rt6_bind_peer(struct rt6_info *rt, int create)
 315 {
 316         struct inet_peer_base *base;
 317         struct inet_peer *peer;
 318
 319         base = inetpeer_base_ptr(rt->_rt6i_peer);
 320         if (!base)
 321                 return;
 322
 323         peer = inet_getpeer_v6(base, &rt->rt6i_dst.addr, create);
 324         if (peer) {
 325                 if (!rt6_set_peer(rt, peer))
 326                         inet_putpeer(peer);
 327         }
 328 }
 329
 330 static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
 331                            int how)
 332 {
 333         struct rt6_info *rt = (struct rt6_info *)dst;
 334         struct inet6_dev *idev = rt->rt6i_idev;
 335         struct net_device *loopback_dev =
 336                 dev_net(dev)->loopback_dev;
 337
 338         if (dev != loopback_dev) {
 339                 if (idev && idev->dev == dev) {
 340                         struct inet6_dev *loopback_idev =
 341                                 in6_dev_get(loopback_dev);
 342                         if (loopback_idev) {
 343                                 rt->rt6i_idev = loopback_idev;
 344                                 in6_dev_put(idev);
 345                         }
 346                 }
 347         }
 348 }
 349
 350 static bool rt6_check_expired(const struct rt6_info *rt)
 351 {
 352         if (rt->rt6i_flags & RTF_EXPIRES) {
 353                 if (time_after(jiffies, rt->dst.expires))
 354                         return true;
 355         } else if (rt->dst.from) {
 356                 return rt6_check_expired((struct rt6_info *) rt->dst.from);
 357         }
 358         return false;
 359 }
 360
 361 static bool rt6_need_strict(const struct in6_addr *daddr)
 362 {
 363         return ipv6_addr_type(daddr) &
 364                 (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
 365 }
 366
 367 /* Multipath route selection:
 368  *   Hash based function using packet header and flowlabel.
 369  * Adapted from fib_info_hashfn()
 370  */
 371 static int rt6_info_hash_nhsfn(unsigned int candidate_count,
 372                                const struct flowi6 *fl6)
 373 {
 374         unsigned int val = fl6->flowi6_proto;
 375
 376         val ^= ipv6_addr_hash(&fl6->daddr);
 377         val ^= ipv6_addr_hash(&fl6->saddr);
 378
 379         /* Work only if this not encapsulated */
 380         switch (fl6->flowi6_proto) {
 381         case IPPROTO_UDP:
 382         case IPPROTO_TCP:
 383         case IPPROTO_SCTP:
 384                 val ^= (__force u16)fl6->fl6_sport;
 385                 val ^= (__force u16)fl6->fl6_dport;
 386                 break;
 387
 388         case IPPROTO_ICMPV6:
 389                 val ^= (__force u16)fl6->fl6_icmp_type;
 390                 val ^= (__force u16)fl6->fl6_icmp_code;
 391                 break;
 392         }
 393         /* RFC6438 recommands to use flowlabel */
 394         val ^= (__force u32)fl6->flowlabel;
 395
 396         /* Perhaps, we need to tune, this function? */
 397         val = val ^ (val >> 7) ^ (val >> 12);
 398         return val % candidate_count;
 399 }
 400
 401 static struct rt6_info *rt6_multipath_select(struct rt6_info *match,
 402                                              struct flowi6 *fl6, int oif,
 403                                              int strict)
 404 {
 405         struct rt6_info *sibling, *next_sibling;
 406         int route_choosen;
 407
 408         route_choosen = rt6_info_hash_nhsfn(match->rt6i_nsiblings + 1, fl6);
 409         /* Don't change the route, if route_choosen == 0
 410          * (siblings does not include ourself)
 411          */
 412         if (route_choosen)
 413                 list_for_each_entry_safe(sibling, next_sibling,
 414                                 &match->rt6i_siblings, rt6i_siblings) {
 415                         route_choosen--;
 416                         if (route_choosen == 0) {
 417                                 if (rt6_score_route(sibling, oif, strict) < 0)
 418                                         break;
 419                                 match = sibling;
 420                                 break;
 421                         }
 422                 }
 423         return match;
 424 }
 425
 426 /*
 427  *      Route lookup. Any table->tb6_lock is implied.
 428  */
 429
 430 static inline struct rt6_info *rt6_device_match(struct net *net,
 431                                                     struct rt6_info *rt,
 432                                                     const struct in6_addr *saddr,
 433                                                     int oif,
 434                                                     int flags)
 435 {
 436         struct rt6_info *local = NULL;
 437         struct rt6_info *sprt;
 438
 439         if (!oif && ipv6_addr_any(saddr))
 440                 goto out;
 441
 442         for (sprt = rt; sprt; sprt = sprt->dst.rt6_next) {
 443                 struct net_device *dev = sprt->dst.dev;
 444
 445                 if (oif) {
 446                         if (dev->ifindex == oif)
 447                                 return sprt;
 448                         if (dev->flags & IFF_LOOPBACK) {
 449                                 if (!sprt->rt6i_idev ||
 450                                     sprt->rt6i_idev->dev->ifindex != oif) {
 451                                         if (flags & RT6_LOOKUP_F_IFACE && oif)
 452                                                 continue;
 453                                         if (local && (!oif ||
 454                                                       local->rt6i_idev->dev->ifindex == oif))
 455                                                 continue;
 456                                 }
 457                                 local = sprt;
 458                         }
 459                 } else {
 460                         if (ipv6_chk_addr(net, saddr, dev,
 461                                           flags & RT6_LOOKUP_F_IFACE))
 462                                 return sprt;
 463                 }
 464         }
 465
 466         if (oif) {
 467                 if (local)
 468                         return local;
 469
 470                 if (flags & RT6_LOOKUP_F_IFACE)
 471                         return net->ipv6.ip6_null_entry;
 472         }
 473 out:
 474         return rt;
 475 }
 476
 477 #ifdef CONFIG_IPV6_ROUTER_PREF
 478 struct __rt6_probe_work {
 479         struct work_struct work;
 480         struct in6_addr target;
 481         struct net_device *dev;
 482 };
 483
 484 static void rt6_probe_deferred(struct work_struct *w)
 485 {
 486         struct in6_addr mcaddr;
 487         struct __rt6_probe_work *work =
 488                 container_of(w, struct __rt6_probe_work, work);
 489
 490         addrconf_addr_solict_mult(&work->target, &mcaddr);
 491         ndisc_send_ns(work->dev, NULL, &work->target, &mcaddr, NULL);
 492         dev_put(work->dev);
 493         kfree(w);
 494 }
 495
 496 static void rt6_probe(struct rt6_info *rt)
 497 {
 498         struct neighbour *neigh;
 499         /*
 500          * Okay, this does not seem to be appropriate
 501          * for now, however, we need to check if it
 502          * is really so; aka Router Reachability Probing.
 503          *
 504          * Router Reachability Probe MUST be rate-limited
 505          * to no more than one per minute.
 506          */
 507         if (!rt || !(rt->rt6i_flags & RTF_GATEWAY))
 508                 return;
 509         rcu_read_lock_bh();
 510         neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
 511         if (neigh) {
 512                 write_lock(&neigh->lock);
 513                 if (neigh->nud_state & NUD_VALID)
 514                         goto out;
 515         }
 516
 517         if (!neigh ||
 518             time_after(jiffies, neigh->updated + rt->rt6i_idev->cnf.rtr_probe_interval)) {
 519                 struct __rt6_probe_work *work;
 520
 521                 work = kmalloc(sizeof(*work), GFP_ATOMIC);
 522
 523                 if (neigh && work)
 524                         neigh->updated = jiffies;
 525
 526                 if (neigh)
 527                         write_unlock(&neigh->lock);
 528
 529                 if (work) {
 530                         INIT_WORK(&work->work, rt6_probe_deferred);
 531                         work->target = rt->rt6i_gateway;
 532                         dev_hold(rt->dst.dev);
 533                         work->dev = rt->dst.dev;
 534                         schedule_work(&work->work);
 535                 }
 536         } else {
 537 out:
 538                 write_unlock(&neigh->lock);
 539         }
 540         rcu_read_unlock_bh();
 541 }
 542 #else
 543 static inline void rt6_probe(struct rt6_info *rt)
 544 {
 545 }
 546 #endif
 547
 548 /*
 549  * Default Router Selection (RFC 2461 6.3.6)
 550  */
 551 static inline int rt6_check_dev(struct rt6_info *rt, int oif)
 552 {
 553         struct net_device *dev = rt->dst.dev;
 554         if (!oif || dev->ifindex == oif)
 555                 return 2;
 556         if ((dev->flags & IFF_LOOPBACK) &&
 557             rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
 558                 return 1;
 559         return 0;
 560 }
 561
 562 static inline enum rt6_nud_state rt6_check_neigh(struct rt6_info *rt)
 563 {
 564         struct neighbour *neigh;
 565         enum rt6_nud_state ret = RT6_NUD_FAIL_HARD;
 566
 567         if (rt->rt6i_flags & RTF_NONEXTHOP ||
 568             !(rt->rt6i_flags & RTF_GATEWAY))
 569                 return RT6_NUD_SUCCEED;
 570
 571         rcu_read_lock_bh();
 572         neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
 573         if (neigh) {
 574                 read_lock(&neigh->lock);
 575                 if (neigh->nud_state & NUD_VALID)
 576                         ret = RT6_NUD_SUCCEED;
 577 #ifdef CONFIG_IPV6_ROUTER_PREF
 578                 else if (!(neigh->nud_state & NUD_FAILED))
 579                         ret = RT6_NUD_SUCCEED;
 580 #endif
 581                 read_unlock(&neigh->lock);
 582         } else {
 583                 ret = IS_ENABLED(CONFIG_IPV6_ROUTER_PREF) ?
 584                       RT6_NUD_SUCCEED : RT6_NUD_FAIL_SOFT;
 585         }
 586         rcu_read_unlock_bh();
 587
 588         return ret;
 589 }
 590
 591 static int rt6_score_route(struct rt6_info *rt, int oif,
 592                            int strict)
 593 {
 594         int m;
 595
 596         m = rt6_check_dev(rt, oif);
 597         if (!m && (strict & RT6_LOOKUP_F_IFACE))
 598                 return RT6_NUD_FAIL_HARD;
 599 #ifdef CONFIG_IPV6_ROUTER_PREF
 600         m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
 601 #endif
 602         if (strict & RT6_LOOKUP_F_REACHABLE) {
 603                 int n = rt6_check_neigh(rt);
 604                 if (n < 0)
 605                         return n;
 606         }
 607         return m;
 608 }
 609
 610 static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
 611                                    int *mpri, struct rt6_info *match,
 612                                    bool *do_rr)
 613 {
 614         int m;
 615         bool match_do_rr = false;
 616
 617         if (rt6_check_expired(rt))
 618                 goto out;
 619
 620         m = rt6_score_route(rt, oif, strict);
 621         if (m == RT6_NUD_FAIL_SOFT) {
 622                 match_do_rr = true;
 623                 m = 0; /* lowest valid score */
 624         } else if (m < 0) {
 625                 goto out;
 626         }
 627
 628         if (strict & RT6_LOOKUP_F_REACHABLE)
 629                 rt6_probe(rt);
 630
 631         if (m > *mpri) {
 632                 *do_rr = match_do_rr;
 633                 *mpri = m;
 634                 match = rt;
 635         }
 636 out:
 637         return match;
 638 }
 639
 640 static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
 641                                      struct rt6_info *rr_head,
 642                                      u32 metric, int oif, int strict,
 643                                      bool *do_rr)
 644 {
 645         struct rt6_info *rt, *match;
 646         int mpri = -1;
 647
 648         match = NULL;
 649         for (rt = rr_head; rt && rt->rt6i_metric == metric;
 650              rt = rt->dst.rt6_next)
 651                 match = find_match(rt, oif, strict, &mpri, match, do_rr);
 652         for (rt = fn->leaf; rt && rt != rr_head && rt->rt6i_metric == metric;
 653              rt = rt->dst.rt6_next)
 654                 match = find_match(rt, oif, strict, &mpri, match, do_rr);
 655
 656         return match;
 657 }
 658
 659 static struct rt6_info *rt6_select(struct fib6_node *fn, int oif, int strict)
 660 {
 661         struct rt6_info *match, *rt0;
 662         struct net *net;
 663         bool do_rr = false;
 664
 665         rt0 = fn->rr_ptr;
 666         if (!rt0)
 667                 fn->rr_ptr = rt0 = fn->leaf;
 668
 669         match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict,
 670                              &do_rr);
 671
 672         if (do_rr) {
 673                 struct rt6_info *next = rt0->dst.rt6_next;
 674
 675                 /* no entries matched; do round-robin */
 676                 if (!next || next->rt6i_metric != rt0->rt6i_metric)
 677                         next = fn->leaf;
 678
 679                 if (next != rt0)
 680                         fn->rr_ptr = next;
 681         }
 682
 683         net = dev_net(rt0->dst.dev);
 684         return match ? match : net->ipv6.ip6_null_entry;
 685 }
 686
 687 #ifdef CONFIG_IPV6_ROUTE_INFO
 688 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
 689                   const struct in6_addr *gwaddr)
 690 {
 691         struct net *net = dev_net(dev);
 692         struct route_info *rinfo = (struct route_info *) opt;
 693         struct in6_addr prefix_buf, *prefix;
 694         unsigned int pref;
 695         unsigned long lifetime;
 696         struct rt6_info *rt;
 697
 698         if (len < sizeof(struct route_info)) {
 699                 return -EINVAL;
 700         }
 701
 702         /* Sanity check for prefix_len and length */
 703         if (rinfo->length > 3) {
 704                 return -EINVAL;
 705         } else if (rinfo->prefix_len > 128) {
 706                 return -EINVAL;
 707         } else if (rinfo->prefix_len > 64) {
 708                 if (rinfo->length < 2) {
 709                         return -EINVAL;
 710                 }
 711         } else if (rinfo->prefix_len > 0) {
 712                 if (rinfo->length < 1) {
 713                         return -EINVAL;
 714                 }
 715         }
 716
 717         pref = rinfo->route_pref;
 718         if (pref == ICMPV6_ROUTER_PREF_INVALID)
 719                 return -EINVAL;
 720
 721         lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
 722
 723         if (rinfo->length == 3)
 724                 prefix = (struct in6_addr *)rinfo->prefix;
 725         else {
 726                 /* this function is safe */
 727                 ipv6_addr_prefix(&prefix_buf,
 728                                  (struct in6_addr *)rinfo->prefix,
 729                                  rinfo->prefix_len);
 730                 prefix = &prefix_buf;
 731         }
 732
 733         if (rinfo->prefix_len == 0)
 734                 rt = rt6_get_dflt_router(gwaddr, dev);
 735         else
 736                 rt = rt6_get_route_info(net, prefix, rinfo->prefix_len,
 737                                         gwaddr, dev->ifindex);
 738
 739         if (rt && !lifetime) {
 740                 ip6_del_rt(rt);
 741                 rt = NULL;
 742         }
 743
 744         if (!rt && lifetime)
 745                 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr, dev->ifindex,
 746                                         pref);
 747         else if (rt)
 748                 rt->rt6i_flags = RTF_ROUTEINFO |
 749                                  (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
 750
 751         if (rt) {
 752                 if (!addrconf_finite_timeout(lifetime))
 753                         rt6_clean_expires(rt);
 754                 else
 755                         rt6_set_expires(rt, jiffies + HZ * lifetime);
 756
 757                 ip6_rt_put(rt);
 758         }
 759         return 0;
 760 }
 761 #endif
 762
 763 #define BACKTRACK(__net, saddr)                 \
 764 do { \
 765         if (rt == __net->ipv6.ip6_null_entry) { \
 766                 struct fib6_node *pn; \
 767                 while (1) { \
 768                         if (fn->fn_flags & RTN_TL_ROOT) \
 769                                 goto out; \
 770                         pn = fn->parent; \
 771                         if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn) \
 772                                 fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr); \
 773                         else \
 774                                 fn = pn; \
 775                         if (fn->fn_flags & RTN_RTINFO) \
 776                                 goto restart; \
 777                 } \
 778         } \
 779 } while (0)
 780
 781 static struct rt6_info *ip6_pol_route_lookup(struct net *net,
 782                                              struct fib6_table *table,
 783                                              struct flowi6 *fl6, int flags)
 784 {
 785         struct fib6_node *fn;
 786         struct rt6_info *rt;
 787
 788         read_lock_bh(&table->tb6_lock);
 789         fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
 790 restart:
 791         rt = fn->leaf;
 792         rt = rt6_device_match(net, rt, &fl6->saddr, fl6->flowi6_oif, flags);
 793         if (rt->rt6i_nsiblings && fl6->flowi6_oif == 0)
 794                 rt = rt6_multipath_select(rt, fl6, fl6->flowi6_oif, flags);
 795         BACKTRACK(net, &fl6->saddr);
 796 out:
 797         dst_use(&rt->dst, jiffies);
 798         read_unlock_bh(&table->tb6_lock);
 799         return rt;
 800
 801 }
 802
 803 struct dst_entry * ip6_route_lookup(struct net *net, struct flowi6 *fl6,
 804                                     int flags)
 805 {
 806         return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_lookup);
 807 }
 808 EXPORT_SYMBOL_GPL(ip6_route_lookup);
 809
 810 struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
 811                             const struct in6_addr *saddr, int oif, int strict)
 812 {
 813         struct flowi6 fl6 = {
 814                 .flowi6_oif = oif,
 815                 .daddr = *daddr,
 816         };
 817         struct dst_entry *dst;
 818         int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
 819
 820         if (saddr) {
 821                 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
 822                 flags |= RT6_LOOKUP_F_HAS_SADDR;
 823         }
 824
 825         dst = fib6_rule_lookup(net, &fl6, flags, ip6_pol_route_lookup);
 826         if (dst->error == 0)
 827                 return (struct rt6_info *) dst;
 828
 829         dst_release(dst);
 830
 831         return NULL;
 832 }
 833
 834 EXPORT_SYMBOL(rt6_lookup);
 835
 836 /* ip6_ins_rt is called with FREE table->tb6_lock.
 837    It takes new route entry, the addition fails by any reason the
 838    route is freed. In any case, if caller does not hold it, it may
 839    be destroyed.
 840  */
 841
 842 static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info)
 843 {
 844         int err;
 845         struct fib6_table *table;
 846
 847         table = rt->rt6i_table;
 848         write_lock_bh(&table->tb6_lock);
 849         err = fib6_add(&table->tb6_root, rt, info);
 850         write_unlock_bh(&table->tb6_lock);
 851
 852         return err;
 853 }
 854
 855 int ip6_ins_rt(struct rt6_info *rt)
 856 {
 857         struct nl_info info = {
 858                 .nl_net = dev_net(rt->dst.dev),
 859         };
 860         return __ip6_ins_rt(rt, &info);
 861 }
 862
 863 static struct rt6_info *rt6_alloc_cow(struct rt6_info *ort,
 864                                       const struct in6_addr *daddr,
 865                                       const struct in6_addr *saddr)
 866 {
 867         struct rt6_info *rt;
 868
 869         /*
 870          *      Clone the route.
 871          */
 872
 873         rt = ip6_rt_copy(ort, daddr);
 874
 875         if (rt) {
 876                 if (ort->rt6i_dst.plen != 128 &&
 877                     ipv6_addr_equal(&ort->rt6i_dst.addr, daddr))
 878                         rt->rt6i_flags |= RTF_ANYCAST;
 879
 880                 rt->rt6i_flags |= RTF_CACHE;
 881
 882 #ifdef CONFIG_IPV6_SUBTREES
 883                 if (rt->rt6i_src.plen && saddr) {
 884                         rt->rt6i_src.addr = *saddr;
 885                         rt->rt6i_src.plen = 128;
 886                 }
 887 #endif
 888         }
 889
 890         return rt;
 891 }
 892
 893 static struct rt6_info *rt6_alloc_clone(struct rt6_info *ort,
 894                                         const struct in6_addr *daddr)
 895 {
 896         struct rt6_info *rt = ip6_rt_copy(ort, daddr);
 897
 898         if (rt)
 899                 rt->rt6i_flags |= RTF_CACHE;
 900         return rt;
 901 }
 902
 903 static struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table, int oif,
 904                                       struct flowi6 *fl6, int flags)
 905 {
 906         struct fib6_node *fn;
 907         struct rt6_info *rt, *nrt;
 908         int strict = 0;
 909         int attempts = 3;
 910         int err;
 911         int reachable = net->ipv6.devconf_all->forwarding ? 0 : RT6_LOOKUP_F_REACHABLE;
 912
 913         strict |= flags & RT6_LOOKUP_F_IFACE;
 914
 915 relookup:
 916         read_lock_bh(&table->tb6_lock);
 917
 918 restart_2:
 919         fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
 920
 921 restart:
 922         rt = rt6_select(fn, oif, strict | reachable);
 923         if (rt->rt6i_nsiblings)
 924                 rt = rt6_multipath_select(rt, fl6, oif, strict | reachable);
 925         BACKTRACK(net, &fl6->saddr);
 926         if (rt == net->ipv6.ip6_null_entry ||
 927             rt->rt6i_flags & RTF_CACHE)
 928                 goto out;
 929
 930         dst_hold(&rt->dst);
 931         read_unlock_bh(&table->tb6_lock);
 932
 933         if (!(rt->rt6i_flags & (RTF_NONEXTHOP | RTF_GATEWAY)))
 934                 nrt = rt6_alloc_cow(rt, &fl6->daddr, &fl6->saddr);
 935         else if (!(rt->dst.flags & DST_HOST))
 936                 nrt = rt6_alloc_clone(rt, &fl6->daddr);
 937         else
 938                 goto out2;
 939
 940         ip6_rt_put(rt);
 941         rt = nrt ? : net->ipv6.ip6_null_entry;
 942
 943         dst_hold(&rt->dst);
 944         if (nrt) {
 945                 err = ip6_ins_rt(nrt);
 946                 if (!err)
 947                         goto out2;
 948         }
 949
 950         if (--attempts <= 0)
 951                 goto out2;
 952
 953         /*
 954          * Race condition! In the gap, when table->tb6_lock was
 955          * released someone could insert this route.  Relookup.
 956          */
 957         ip6_rt_put(rt);
 958         goto relookup;
 959
 960 out:
 961         if (reachable) {
 962                 reachable = 0;
 963                 goto restart_2;
 964         }
 965         dst_hold(&rt->dst);
 966         read_unlock_bh(&table->tb6_lock);
 967 out2:
 968         rt->dst.lastuse = jiffies;
 969         rt->dst.__use++;
 970
 971         return rt;
 972 }
 973
 974 static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
 975                                             struct flowi6 *fl6, int flags)
 976 {
 977         return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, flags);
 978 }
 979
 980 static struct dst_entry *ip6_route_input_lookup(struct net *net,
 981                                                 struct net_device *dev,
 982                                                 struct flowi6 *fl6, int flags)
 983 {
 984         if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
 985                 flags |= RT6_LOOKUP_F_IFACE;
 986
 987         return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_input);
 988 }
 989
 990 void ip6_route_input(struct sk_buff *skb)
 991 {
 992         const struct ipv6hdr *iph = ipv6_hdr(skb);
 993         struct net *net = dev_net(skb->dev);
 994         int flags = RT6_LOOKUP_F_HAS_SADDR;
 995         struct flowi6 fl6 = {
 996                 .flowi6_iif = skb->dev->ifindex,
 997                 .daddr = iph->daddr,
 998                 .saddr = iph->saddr,
 999                 .flowlabel = ip6_flowinfo(iph),
1000                 .flowi6_mark = skb->mark,
1001                 .flowi6_proto = iph->nexthdr,
1002         };
1003
1004         skb_dst_set(skb, ip6_route_input_lookup(net, skb->dev, &fl6, flags));
1005 }
1006
1007 static struct rt6_info *ip6_pol_route_output(struct net *net, struct fib6_table *table,
1008                                              struct flowi6 *fl6, int flags)
1009 {
1010         return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, flags);
1011 }
1012
1013 struct dst_entry * ip6_route_output(struct net *net, const struct sock *sk,
1014                                     struct flowi6 *fl6)
1015 {
1016         int flags = 0;
1017
1018         fl6->flowi6_iif = LOOPBACK_IFINDEX;
1019
1020         if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr))
1021                 flags |= RT6_LOOKUP_F_IFACE;
1022
1023         if (!ipv6_addr_any(&fl6->saddr))
1024                 flags |= RT6_LOOKUP_F_HAS_SADDR;
1025         else if (sk)
1026                 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
1027
1028         return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_output);
1029 }
1030
1031 EXPORT_SYMBOL(ip6_route_output);
1032
1033 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
1034 {
1035         struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
1036         struct dst_entry *new = NULL;
1037
1038         rt = dst_alloc(&ip6_dst_blackhole_ops, ort->dst.dev, 1, DST_OBSOLETE_NONE, 0);
1039         if (rt) {
1040                 new = &rt->dst;
1041
1042                 memset(new + 1, 0, sizeof(*rt) - sizeof(*new));
1043                 rt6_init_peer(rt, net->ipv6.peers);
1044
1045                 new->__use = 1;
1046                 new->input = dst_discard;
1047                 new->output = dst_discard;
1048
1049                 if (dst_metrics_read_only(&ort->dst))
1050                         new->_metrics = ort->dst._metrics;
1051                 else
1052                         dst_copy_metrics(new, &ort->dst);
1053                 rt->rt6i_idev = ort->rt6i_idev;
1054                 if (rt->rt6i_idev)
1055                         in6_dev_hold(rt->rt6i_idev);
1056
1057                 rt->rt6i_gateway = ort->rt6i_gateway;
1058                 rt->rt6i_flags = ort->rt6i_flags;
1059                 rt->rt6i_metric = 0;
1060
1061                 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1062 #ifdef CONFIG_IPV6_SUBTREES
1063                 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1064 #endif
1065
1066                 dst_free(new);
1067         }
1068
1069         dst_release(dst_orig);
1070         return new ? new : ERR_PTR(-ENOMEM);
1071 }
1072
1073 /*
1074  *      Destination cache support functions
1075  */
1076
1077 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
1078 {
1079         struct rt6_info *rt;
1080
1081         rt = (struct rt6_info *) dst;
1082
1083         /* All IPV6 dsts are created with ->obsolete set to the value
1084          * DST_OBSOLETE_FORCE_CHK which forces validation calls down
1085          * into this function always.
1086          */
1087         if (rt->rt6i_genid != rt_genid_ipv6(dev_net(rt->dst.dev)))
1088                 return NULL;
1089
1090         if (!rt->rt6i_node || (rt->rt6i_node->fn_sernum != cookie))
1091                 return NULL;
1092
1093         if (rt6_check_expired(rt))
1094                 return NULL;
1095
1096         return dst;
1097 }
1098
1099 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
1100 {
1101         struct rt6_info *rt = (struct rt6_info *) dst;
1102
1103         if (rt) {
1104                 if (rt->rt6i_flags & RTF_CACHE) {
1105                         if (rt6_check_expired(rt)) {
1106                                 ip6_del_rt(rt);
1107                                 dst = NULL;
1108                         }
1109                 } else {
1110                         dst_release(dst);
1111                         dst = NULL;
1112                 }
1113         }
1114         return dst;
1115 }
1116
1117 static void ip6_link_failure(struct sk_buff *skb)
1118 {
1119         struct rt6_info *rt;
1120
1121         icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
1122
1123         rt = (struct rt6_info *) skb_dst(skb);
1124         if (rt) {
1125                 if (rt->rt6i_flags & RTF_CACHE) {
1126                         dst_hold(&rt->dst);
1127                         if (ip6_del_rt(rt))
1128                                 dst_free(&rt->dst);
1129                 } else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT)) {
1130                         rt->rt6i_node->fn_sernum = -1;
1131                 }
1132         }
1133 }
1134
1135 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
1136                                struct sk_buff *skb, u32 mtu)
1137 {
1138         struct rt6_info *rt6 = (struct rt6_info*)dst;
1139
1140         dst_confirm(dst);
1141         if (mtu < dst_mtu(dst) && rt6->rt6i_dst.plen == 128) {
1142                 struct net *net = dev_net(dst->dev);
1143
1144                 rt6->rt6i_flags |= RTF_MODIFIED;
1145                 if (mtu < IPV6_MIN_MTU) {
1146                         u32 features = dst_metric(dst, RTAX_FEATURES);
1147                         mtu = IPV6_MIN_MTU;
1148                         features |= RTAX_FEATURE_ALLFRAG;
1149                         dst_metric_set(dst, RTAX_FEATURES, features);
1150                 }
1151                 dst_metric_set(dst, RTAX_MTU, mtu);
1152                 rt6_update_expires(rt6, net->ipv6.sysctl.ip6_rt_mtu_expires);
1153         }
1154 }
1155
1156 void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
1157                      int oif, u32 mark)
1158 {
1159         const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1160         struct dst_entry *dst;
1161         struct flowi6 fl6;
1162
1163         memset(&fl6, 0, sizeof(fl6));
1164         fl6.flowi6_oif = oif;
1165         fl6.flowi6_mark = mark;
1166         fl6.daddr = iph->daddr;
1167         fl6.saddr = iph->saddr;
1168         fl6.flowlabel = ip6_flowinfo(iph);
1169
1170         dst = ip6_route_output(net, NULL, &fl6);
1171         if (!dst->error)
1172                 ip6_rt_update_pmtu(dst, NULL, skb, ntohl(mtu));
1173         dst_release(dst);
1174 }
1175 EXPORT_SYMBOL_GPL(ip6_update_pmtu);
1176
1177 void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
1178 {
1179         ip6_update_pmtu(skb, sock_net(sk), mtu,
1180                         sk->sk_bound_dev_if, sk->sk_mark);
1181 }
1182 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
1183
1184 /* Handle redirects */
1185 struct ip6rd_flowi {
1186         struct flowi6 fl6;
1187         struct in6_addr gateway;
1188 };
1189
1190 static struct rt6_info *__ip6_route_redirect(struct net *net,
1191                                              struct fib6_table *table,
1192                                              struct flowi6 *fl6,
1193                                              int flags)
1194 {
1195         struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
1196         struct rt6_info *rt;
1197         struct fib6_node *fn;
1198
1199         /* Get the "current" route for this destination and
1200          * check if the redirect has come from approriate router.
1201          *
1202          * RFC 4861 specifies that redirects should only be
1203          * accepted if they come from the nexthop to the target.
1204          * Due to the way the routes are chosen, this notion
1205          * is a bit fuzzy and one might need to check all possible
1206          * routes.
1207          */
1208
1209         read_lock_bh(&table->tb6_lock);
1210         fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1211 restart:
1212         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1213                 if (rt6_check_expired(rt))
1214                         continue;
1215                 if (rt->dst.error)
1216                         break;
1217                 if (!(rt->rt6i_flags & RTF_GATEWAY))
1218                         continue;
1219                 if (fl6->flowi6_oif != rt->dst.dev->ifindex)
1220                         continue;
1221                 if (!ipv6_addr_equal(&rdfl->gateway, &rt->rt6i_gateway))
1222                         continue;
1223                 break;
1224         }
1225
1226         if (!rt)
1227                 rt = net->ipv6.ip6_null_entry;
1228         else if (rt->dst.error) {
1229                 rt = net->ipv6.ip6_null_entry;
1230                 goto out;
1231         }
1232         BACKTRACK(net, &fl6->saddr);
1233 out:
1234         dst_hold(&rt->dst);
1235
1236         read_unlock_bh(&table->tb6_lock);
1237
1238         return rt;
1239 };
1240
1241 static struct dst_entry *ip6_route_redirect(struct net *net,
1242                                         const struct flowi6 *fl6,
1243                                         const struct in6_addr *gateway)
1244 {
1245         int flags = RT6_LOOKUP_F_HAS_SADDR;
1246         struct ip6rd_flowi rdfl;
1247
1248         rdfl.fl6 = *fl6;
1249         rdfl.gateway = *gateway;
1250
1251         return fib6_rule_lookup(net, &rdfl.fl6,
1252                                 flags, __ip6_route_redirect);
1253 }
1254
1255 void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark)
1256 {
1257         const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1258         struct dst_entry *dst;
1259         struct flowi6 fl6;
1260
1261         memset(&fl6, 0, sizeof(fl6));
1262         fl6.flowi6_oif = oif;
1263         fl6.flowi6_mark = mark;
1264         fl6.daddr = iph->daddr;
1265         fl6.saddr = iph->saddr;
1266         fl6.flowlabel = ip6_flowinfo(iph);
1267
1268         dst = ip6_route_redirect(net, &fl6, &ipv6_hdr(skb)->saddr);
1269         rt6_do_redirect(dst, NULL, skb);
1270         dst_release(dst);
1271 }
1272 EXPORT_SYMBOL_GPL(ip6_redirect);
1273
1274 void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif,
1275                             u32 mark)
1276 {
1277         const struct ipv6hdr *iph = ipv6_hdr(skb);
1278         const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
1279         struct dst_entry *dst;
1280         struct flowi6 fl6;
1281
1282         memset(&fl6, 0, sizeof(fl6));
1283         fl6.flowi6_oif = oif;
1284         fl6.flowi6_mark = mark;
1285         fl6.daddr = msg->dest;
1286         fl6.saddr = iph->daddr;
1287
1288         dst = ip6_route_redirect(net, &fl6, &iph->saddr);
1289         rt6_do_redirect(dst, NULL, skb);
1290         dst_release(dst);
1291 }
1292
1293 void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
1294 {
1295         ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark);
1296 }
1297 EXPORT_SYMBOL_GPL(ip6_sk_redirect);
1298
1299 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1300 {
1301         struct net_device *dev = dst->dev;
1302         unsigned int mtu = dst_mtu(dst);
1303         struct net *net = dev_net(dev);
1304
1305         mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
1306
1307         if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
1308                 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1309
1310         /*
1311          * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1312          * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1313          * IPV6_MAXPLEN is also valid and means: "any MSS,
1314          * rely only on pmtu discovery"
1315          */
1316         if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
1317                 mtu = IPV6_MAXPLEN;
1318         return mtu;
1319 }
1320
1321 static unsigned int ip6_mtu(const struct dst_entry *dst)
1322 {
1323         struct inet6_dev *idev;
1324         unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
1325
1326         if (mtu)
1327                 return mtu;
1328
1329         mtu = IPV6_MIN_MTU;
1330
1331         rcu_read_lock();
1332         idev = __in6_dev_get(dst->dev);
1333         if (idev)
1334                 mtu = idev->cnf.mtu6;
1335         rcu_read_unlock();
1336
1337         return mtu;
1338 }
1339
1340 static struct dst_entry *icmp6_dst_gc_list;
1341 static DEFINE_SPINLOCK(icmp6_dst_lock);
1342
1343 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
1344                                   struct flowi6 *fl6)
1345 {
1346         struct dst_entry *dst;
1347         struct rt6_info *rt;
1348         struct inet6_dev *idev = in6_dev_get(dev);
1349         struct net *net = dev_net(dev);
1350
1351         if (unlikely(!idev))
1352                 return ERR_PTR(-ENODEV);
1353
1354         rt = ip6_dst_alloc(net, dev, 0, NULL);
1355         if (unlikely(!rt)) {
1356                 in6_dev_put(idev);
1357                 dst = ERR_PTR(-ENOMEM);
1358                 goto out;
1359         }
1360
1361         rt->dst.flags |= DST_HOST;
1362         rt->dst.output  = ip6_output;
1363         atomic_set(&rt->dst.__refcnt, 1);
1364         rt->rt6i_gateway  = fl6->daddr;
1365         rt->rt6i_dst.addr = fl6->daddr;
1366         rt->rt6i_dst.plen = 128;
1367         rt->rt6i_idev     = idev;
1368         dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
1369
1370         spin_lock_bh(&icmp6_dst_lock);
1371         rt->dst.next = icmp6_dst_gc_list;
1372         icmp6_dst_gc_list = &rt->dst;
1373         spin_unlock_bh(&icmp6_dst_lock);
1374
1375         fib6_force_start_gc(net);
1376
1377         dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
1378
1379 out:
1380         return dst;
1381 }
1382
1383 int icmp6_dst_gc(void)
1384 {
1385         struct dst_entry *dst, **pprev;
1386         int more = 0;
1387
1388         spin_lock_bh(&icmp6_dst_lock);
1389         pprev = &icmp6_dst_gc_list;
1390
1391         while ((dst = *pprev) != NULL) {
1392                 if (!atomic_read(&dst->__refcnt)) {
1393                         *pprev = dst->next;
1394                         dst_free(dst);
1395                 } else {
1396                         pprev = &dst->next;
1397                         ++more;
1398                 }
1399         }
1400
1401         spin_unlock_bh(&icmp6_dst_lock);
1402
1403         return more;
1404 }
1405
1406 static void icmp6_clean_all(int (*func)(struct rt6_info *rt, void *arg),
1407                             void *arg)
1408 {
1409         struct dst_entry *dst, **pprev;
1410
1411         spin_lock_bh(&icmp6_dst_lock);
1412         pprev = &icmp6_dst_gc_list;
1413         while ((dst = *pprev) != NULL) {
1414                 struct rt6_info *rt = (struct rt6_info *) dst;
1415                 if (func(rt, arg)) {
1416                         *pprev = dst->next;
1417                         dst_free(dst);
1418                 } else {
1419                         pprev = &dst->next;
1420                 }
1421         }
1422         spin_unlock_bh(&icmp6_dst_lock);
1423 }
1424
1425 static int ip6_dst_gc(struct dst_ops *ops)
1426 {
1427         struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
1428         int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1429         int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1430         int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1431         int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1432         unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1433         int entries;
1434
1435         entries = dst_entries_get_fast(ops);
1436         if (time_after(rt_last_gc + rt_min_interval, jiffies) &&
1437             entries <= rt_max_size)
1438                 goto out;
1439
1440         net->ipv6.ip6_rt_gc_expire++;
1441         fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net, entries > rt_max_size);
1442         entries = dst_entries_get_slow(ops);
1443         if (entries < ops->gc_thresh)
1444                 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1445 out:
1446         net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1447         return entries > rt_max_size;
1448 }
1449
1450 /*
1451  *
1452  */
1453
1454 int ip6_route_add(struct fib6_config *cfg)
1455 {
1456         int err;
1457         struct net *net = cfg->fc_nlinfo.nl_net;
1458         struct rt6_info *rt = NULL;
1459         struct net_device *dev = NULL;
1460         struct inet6_dev *idev = NULL;
1461         struct fib6_table *table;
1462         int addr_type;
1463
1464         if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1465                 return -EINVAL;
1466 #ifndef CONFIG_IPV6_SUBTREES
1467         if (cfg->fc_src_len)
1468                 return -EINVAL;
1469 #endif
1470         if (cfg->fc_ifindex) {
1471                 err = -ENODEV;
1472                 dev = dev_get_by_index(net, cfg->fc_ifindex);
1473                 if (!dev)
1474                         goto out;
1475                 idev = in6_dev_get(dev);
1476                 if (!idev)
1477                         goto out;
1478         }
1479
1480         if (cfg->fc_metric == 0)
1481                 cfg->fc_metric = IP6_RT_PRIO_USER;
1482
1483         err = -ENOBUFS;
1484         if (cfg->fc_nlinfo.nlh &&
1485             !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
1486                 table = fib6_get_table(net, cfg->fc_table);
1487                 if (!table) {
1488                         pr_warn("NLM_F_CREATE should be specified when creating new route\n");
1489                         table = fib6_new_table(net, cfg->fc_table);
1490                 }
1491         } else {
1492                 table = fib6_new_table(net, cfg->fc_table);
1493         }
1494
1495         if (!table)
1496                 goto out;
1497
1498         rt = ip6_dst_alloc(net, NULL, DST_NOCOUNT, table);
1499
1500         if (!rt) {
1501                 err = -ENOMEM;
1502                 goto out;
1503         }
1504
1505         if (cfg->fc_flags & RTF_EXPIRES)
1506                 rt6_set_expires(rt, jiffies +
1507                                 clock_t_to_jiffies(cfg->fc_expires));
1508         else
1509                 rt6_clean_expires(rt);
1510
1511         if (cfg->fc_protocol == RTPROT_UNSPEC)
1512                 cfg->fc_protocol = RTPROT_BOOT;
1513         rt->rt6i_protocol = cfg->fc_protocol;
1514
1515         addr_type = ipv6_addr_type(&cfg->fc_dst);
1516
1517         if (addr_type & IPV6_ADDR_MULTICAST)
1518                 rt->dst.input = ip6_mc_input;
1519         else if (cfg->fc_flags & RTF_LOCAL)
1520                 rt->dst.input = ip6_input;
1521         else
1522                 rt->dst.input = ip6_forward;
1523
1524         rt->dst.output = ip6_output;
1525
1526         ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1527         rt->rt6i_dst.plen = cfg->fc_dst_len;
1528         if (rt->rt6i_dst.plen == 128)
1529                rt->dst.flags |= DST_HOST;
1530
1531         if (!(rt->dst.flags & DST_HOST) && cfg->fc_mx) {
1532                 u32 *metrics = kzalloc(sizeof(u32) * RTAX_MAX, GFP_KERNEL);
1533                 if (!metrics) {
1534                         err = -ENOMEM;
1535                         goto out;
1536                 }
1537                 dst_init_metrics(&rt->dst, metrics, 0);
1538         }
1539 #ifdef CONFIG_IPV6_SUBTREES
1540         ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1541         rt->rt6i_src.plen = cfg->fc_src_len;
1542 #endif
1543
1544         rt->rt6i_metric = cfg->fc_metric;
1545
1546         /* We cannot add true routes via loopback here,
1547            they would result in kernel looping; promote them to reject routes
1548          */
1549         if ((cfg->fc_flags & RTF_REJECT) ||
1550             (dev && (dev->flags & IFF_LOOPBACK) &&
1551              !(addr_type & IPV6_ADDR_LOOPBACK) &&
1552              !(cfg->fc_flags & RTF_LOCAL))) {
1553                 /* hold loopback dev/idev if we haven't done so. */
1554                 if (dev != net->loopback_dev) {
1555                         if (dev) {
1556                                 dev_put(dev);
1557                                 in6_dev_put(idev);
1558                         }
1559                         dev = net->loopback_dev;
1560                         dev_hold(dev);
1561                         idev = in6_dev_get(dev);
1562                         if (!idev) {
1563                                 err = -ENODEV;
1564                                 goto out;
1565                         }
1566                 }
1567                 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1568                 switch (cfg->fc_type) {
1569                 case RTN_BLACKHOLE:
1570                         rt->dst.error = -EINVAL;
1571                         rt->dst.output = dst_discard;
1572                         rt->dst.input = dst_discard;
1573                         break;
1574                 case RTN_PROHIBIT:
1575                         rt->dst.error = -EACCES;
1576                         rt->dst.output = ip6_pkt_prohibit_out;
1577                         rt->dst.input = ip6_pkt_prohibit;
1578                         break;
1579                 case RTN_THROW:
1580                 default:
1581                         rt->dst.error = (cfg->fc_type == RTN_THROW) ? -EAGAIN
1582                                         : -ENETUNREACH;
1583                         rt->dst.output = ip6_pkt_discard_out;
1584                         rt->dst.input = ip6_pkt_discard;
1585                         break;
1586                 }
1587                 goto install_route;
1588         }
1589
1590         if (cfg->fc_flags & RTF_GATEWAY) {
1591                 const struct in6_addr *gw_addr;
1592                 int gwa_type;
1593
1594                 gw_addr = &cfg->fc_gateway;
1595                 rt->rt6i_gateway = *gw_addr;
1596                 gwa_type = ipv6_addr_type(gw_addr);
1597
1598                 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1599                         struct rt6_info *grt;
1600
1601                         /* IPv6 strictly inhibits using not link-local
1602                            addresses as nexthop address.
1603                            Otherwise, router will not able to send redirects.
1604                            It is very good, but in some (rare!) circumstances
1605                            (SIT, PtP, NBMA NOARP links) it is handy to allow
1606                            some exceptions. --ANK
1607                          */
1608                         err = -EINVAL;
1609                         if (!(gwa_type & IPV6_ADDR_UNICAST))
1610                                 goto out;
1611
1612                         grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, 1);
1613
1614                         err = -EHOSTUNREACH;
1615                         if (!grt)
1616                                 goto out;
1617                         if (dev) {
1618                                 if (dev != grt->dst.dev) {
1619                                         ip6_rt_put(grt);
1620                                         goto out;
1621                                 }
1622                         } else {
1623                                 dev = grt->dst.dev;
1624                                 idev = grt->rt6i_idev;
1625                                 dev_hold(dev);
1626                                 in6_dev_hold(grt->rt6i_idev);
1627                         }
1628                         if (!(grt->rt6i_flags & RTF_GATEWAY))
1629                                 err = 0;
1630                         ip6_rt_put(grt);
1631
1632                         if (err)
1633                                 goto out;
1634                 }
1635                 err = -EINVAL;
1636                 if (!dev || (dev->flags & IFF_LOOPBACK))
1637                         goto out;
1638         }
1639
1640         err = -ENODEV;
1641         if (!dev)
1642                 goto out;
1643
1644         if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
1645                 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
1646                         err = -EINVAL;
1647                         goto out;
1648                 }
1649                 rt->rt6i_prefsrc.addr = cfg->fc_prefsrc;
1650                 rt->rt6i_prefsrc.plen = 128;
1651         } else
1652                 rt->rt6i_prefsrc.plen = 0;
1653
1654         rt->rt6i_flags = cfg->fc_flags;
1655
1656 install_route:
1657         if (cfg->fc_mx) {
1658                 struct nlattr *nla;
1659                 int remaining;
1660
1661                 nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1662                         int type = nla_type(nla);
1663
1664                         if (type) {
1665                                 if (type > RTAX_MAX) {
1666                                         err = -EINVAL;
1667                                         goto out;
1668                                 }
1669
1670                                 dst_metric_set(&rt->dst, type, nla_get_u32(nla));
1671                         }
1672                 }
1673         }
1674
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);
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                 if ((ort->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF)) ==
1909                     (RTF_DEFAULT | RTF_ADDRCONF))
1910                         rt6_set_from(rt, ort);
1911                 rt->rt6i_metric = 0;
1912
1913 #ifdef CONFIG_IPV6_SUBTREES
1914                 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1915 #endif
1916                 memcpy(&rt->rt6i_prefsrc, &ort->rt6i_prefsrc, sizeof(struct rt6key));
1917                 rt->rt6i_table = ort->rt6i_table;
1918         }
1919         return rt;
1920 }
1921
1922 #ifdef CONFIG_IPV6_ROUTE_INFO
1923 static struct rt6_info *rt6_get_route_info(struct net *net,
1924                                            const struct in6_addr *prefix, int prefixlen,
1925                                            const struct in6_addr *gwaddr, int ifindex)
1926 {
1927         struct fib6_node *fn;
1928         struct rt6_info *rt = NULL;
1929         struct fib6_table *table;
1930
1931         table = fib6_get_table(net, RT6_TABLE_INFO);
1932         if (!table)
1933                 return NULL;
1934
1935         read_lock_bh(&table->tb6_lock);
1936         fn = fib6_locate(&table->tb6_root, prefix ,prefixlen, NULL, 0);
1937         if (!fn)
1938                 goto out;
1939
1940         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1941                 if (rt->dst.dev->ifindex != ifindex)
1942                         continue;
1943                 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
1944                         continue;
1945                 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
1946                         continue;
1947                 dst_hold(&rt->dst);
1948                 break;
1949         }
1950 out:
1951         read_unlock_bh(&table->tb6_lock);
1952         return rt;
1953 }
1954
1955 static struct rt6_info *rt6_add_route_info(struct net *net,
1956                                            const struct in6_addr *prefix, int prefixlen,
1957                                            const struct in6_addr *gwaddr, int ifindex,
1958                                            unsigned int pref)
1959 {
1960         struct fib6_config cfg = {
1961                 .fc_table       = RT6_TABLE_INFO,
1962                 .fc_metric      = IP6_RT_PRIO_USER,
1963                 .fc_ifindex     = ifindex,
1964                 .fc_dst_len     = prefixlen,
1965                 .fc_flags       = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
1966                                   RTF_UP | RTF_PREF(pref),
1967                 .fc_nlinfo.portid = 0,
1968                 .fc_nlinfo.nlh = NULL,
1969                 .fc_nlinfo.nl_net = net,
1970         };
1971
1972         cfg.fc_dst = *prefix;
1973         cfg.fc_gateway = *gwaddr;
1974
1975         /* We should treat it as a default route if prefix length is 0. */
1976         if (!prefixlen)
1977                 cfg.fc_flags |= RTF_DEFAULT;
1978
1979         ip6_route_add(&cfg);
1980
1981         return rt6_get_route_info(net, prefix, prefixlen, gwaddr, ifindex);
1982 }
1983 #endif
1984
1985 struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
1986 {
1987         struct rt6_info *rt;
1988         struct fib6_table *table;
1989
1990         table = fib6_get_table(dev_net(dev), RT6_TABLE_DFLT);
1991         if (!table)
1992                 return NULL;
1993
1994         read_lock_bh(&table->tb6_lock);
1995         for (rt = table->tb6_root.leaf; rt; rt=rt->dst.rt6_next) {
1996                 if (dev == rt->dst.dev &&
1997                     ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
1998                     ipv6_addr_equal(&rt->rt6i_gateway, addr))
1999                         break;
2000         }
2001         if (rt)
2002                 dst_hold(&rt->dst);
2003         read_unlock_bh(&table->tb6_lock);
2004         return rt;
2005 }
2006
2007 struct rt6_info *rt6_add_dflt_router(const struct in6_addr *gwaddr,
2008                                      struct net_device *dev,
2009                                      unsigned int pref)
2010 {
2011         struct fib6_config cfg = {
2012                 .fc_table       = RT6_TABLE_DFLT,
2013                 .fc_metric      = IP6_RT_PRIO_USER,
2014                 .fc_ifindex     = dev->ifindex,
2015                 .fc_flags       = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
2016                                   RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
2017                 .fc_nlinfo.portid = 0,
2018                 .fc_nlinfo.nlh = NULL,
2019                 .fc_nlinfo.nl_net = dev_net(dev),
2020         };
2021
2022         cfg.fc_gateway = *gwaddr;
2023
2024         ip6_route_add(&cfg);
2025
2026         return rt6_get_dflt_router(gwaddr, dev);
2027 }
2028
2029 void rt6_purge_dflt_routers(struct net *net)
2030 {
2031         struct rt6_info *rt;
2032         struct fib6_table *table;
2033
2034         /* NOTE: Keep consistent with rt6_get_dflt_router */
2035         table = fib6_get_table(net, RT6_TABLE_DFLT);
2036         if (!table)
2037                 return;
2038
2039 restart:
2040         read_lock_bh(&table->tb6_lock);
2041         for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2042                 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
2043                     (!rt->rt6i_idev || rt->rt6i_idev->cnf.accept_ra != 2)) {
2044                         dst_hold(&rt->dst);
2045                         read_unlock_bh(&table->tb6_lock);
2046                         ip6_del_rt(rt);
2047                         goto restart;
2048                 }
2049         }
2050         read_unlock_bh(&table->tb6_lock);
2051 }
2052
2053 static void rtmsg_to_fib6_config(struct net *net,
2054                                  struct in6_rtmsg *rtmsg,
2055                                  struct fib6_config *cfg)
2056 {
2057         memset(cfg, 0, sizeof(*cfg));
2058
2059         cfg->fc_table = RT6_TABLE_MAIN;
2060         cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
2061         cfg->fc_metric = rtmsg->rtmsg_metric;
2062         cfg->fc_expires = rtmsg->rtmsg_info;
2063         cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
2064         cfg->fc_src_len = rtmsg->rtmsg_src_len;
2065         cfg->fc_flags = rtmsg->rtmsg_flags;
2066
2067         cfg->fc_nlinfo.nl_net = net;
2068
2069         cfg->fc_dst = rtmsg->rtmsg_dst;
2070         cfg->fc_src = rtmsg->rtmsg_src;
2071         cfg->fc_gateway = rtmsg->rtmsg_gateway;
2072 }
2073
2074 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
2075 {
2076         struct fib6_config cfg;
2077         struct in6_rtmsg rtmsg;
2078         int err;
2079
2080         switch(cmd) {
2081         case SIOCADDRT:         /* Add a route */
2082         case SIOCDELRT:         /* Delete a route */
2083                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2084                         return -EPERM;
2085                 err = copy_from_user(&rtmsg, arg,
2086                                      sizeof(struct in6_rtmsg));
2087                 if (err)
2088                         return -EFAULT;
2089
2090                 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
2091
2092                 rtnl_lock();
2093                 switch (cmd) {
2094                 case SIOCADDRT:
2095                         err = ip6_route_add(&cfg);
2096                         break;
2097                 case SIOCDELRT:
2098                         err = ip6_route_del(&cfg);
2099                         break;
2100                 default:
2101                         err = -EINVAL;
2102                 }
2103                 rtnl_unlock();
2104
2105                 return err;
2106         }
2107
2108         return -EINVAL;
2109 }
2110
2111 /*
2112  *      Drop the packet on the floor
2113  */
2114
2115 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
2116 {
2117         int type;
2118         struct dst_entry *dst = skb_dst(skb);
2119         switch (ipstats_mib_noroutes) {
2120         case IPSTATS_MIB_INNOROUTES:
2121                 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
2122                 if (type == IPV6_ADDR_ANY) {
2123                         IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2124                                       IPSTATS_MIB_INADDRERRORS);
2125                         break;
2126                 }
2127                 /* FALLTHROUGH */
2128         case IPSTATS_MIB_OUTNOROUTES:
2129                 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2130                               ipstats_mib_noroutes);
2131                 break;
2132         }
2133         icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
2134         kfree_skb(skb);
2135         return 0;
2136 }
2137
2138 static int ip6_pkt_discard(struct sk_buff *skb)
2139 {
2140         return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
2141 }
2142
2143 static int ip6_pkt_discard_out(struct sk_buff *skb)
2144 {
2145         skb->dev = skb_dst(skb)->dev;
2146         return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
2147 }
2148
2149 static int ip6_pkt_prohibit(struct sk_buff *skb)
2150 {
2151         return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
2152 }
2153
2154 static int ip6_pkt_prohibit_out(struct sk_buff *skb)
2155 {
2156         skb->dev = skb_dst(skb)->dev;
2157         return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
2158 }
2159
2160 /*
2161  *      Allocate a dst for local (unicast / anycast) address.
2162  */
2163
2164 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
2165                                     const struct in6_addr *addr,
2166                                     bool anycast)
2167 {
2168         struct net *net = dev_net(idev->dev);
2169         struct rt6_info *rt = ip6_dst_alloc(net, net->loopback_dev,
2170                                             DST_NOCOUNT, NULL);
2171         if (!rt)
2172                 return ERR_PTR(-ENOMEM);
2173
2174         in6_dev_hold(idev);
2175
2176         rt->dst.flags |= DST_HOST;
2177         rt->dst.input = ip6_input;
2178         rt->dst.output = ip6_output;
2179         rt->rt6i_idev = idev;
2180
2181         rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
2182         if (anycast)
2183                 rt->rt6i_flags |= RTF_ANYCAST;
2184         else
2185                 rt->rt6i_flags |= RTF_LOCAL;
2186
2187         rt->rt6i_gateway  = *addr;
2188         rt->rt6i_dst.addr = *addr;
2189         rt->rt6i_dst.plen = 128;
2190         rt->rt6i_table = fib6_get_table(net, RT6_TABLE_LOCAL);
2191
2192         atomic_set(&rt->dst.__refcnt, 1);
2193
2194         return rt;
2195 }
2196
2197 int ip6_route_get_saddr(struct net *net,
2198                         struct rt6_info *rt,
2199                         const struct in6_addr *daddr,
2200                         unsigned int prefs,
2201                         struct in6_addr *saddr)
2202 {
2203         struct inet6_dev *idev = ip6_dst_idev((struct dst_entry*)rt);
2204         int err = 0;
2205         if (rt->rt6i_prefsrc.plen)
2206                 *saddr = rt->rt6i_prefsrc.addr;
2207         else
2208                 err = ipv6_dev_get_saddr(net, idev ? idev->dev : NULL,
2209                                          daddr, prefs, saddr);
2210         return err;
2211 }
2212
2213 /* remove deleted ip from prefsrc entries */
2214 struct arg_dev_net_ip {
2215         struct net_device *dev;
2216         struct net *net;
2217         struct in6_addr *addr;
2218 };
2219
2220 static int fib6_remove_prefsrc(struct rt6_info *rt, void *arg)
2221 {
2222         struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
2223         struct net *net = ((struct arg_dev_net_ip *)arg)->net;
2224         struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
2225
2226         if (((void *)rt->dst.dev == dev || !dev) &&
2227             rt != net->ipv6.ip6_null_entry &&
2228             ipv6_addr_equal(addr, &rt->rt6i_prefsrc.addr)) {
2229                 /* remove prefsrc entry */
2230                 rt->rt6i_prefsrc.plen = 0;
2231         }
2232         return 0;
2233 }
2234
2235 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
2236 {
2237         struct net *net = dev_net(ifp->idev->dev);
2238         struct arg_dev_net_ip adni = {
2239                 .dev = ifp->idev->dev,
2240                 .net = net,
2241                 .addr = &ifp->addr,
2242         };
2243         fib6_clean_all(net, fib6_remove_prefsrc, 0, &adni);
2244 }
2245
2246 struct arg_dev_net {
2247         struct net_device *dev;
2248         struct net *net;
2249 };
2250
2251 static int fib6_ifdown(struct rt6_info *rt, void *arg)
2252 {
2253         const struct arg_dev_net *adn = arg;
2254         const struct net_device *dev = adn->dev;
2255
2256         if ((rt->dst.dev == dev || !dev) &&
2257             rt != adn->net->ipv6.ip6_null_entry)
2258                 return -1;
2259
2260         return 0;
2261 }
2262
2263 void rt6_ifdown(struct net *net, struct net_device *dev)
2264 {
2265         struct arg_dev_net adn = {
2266                 .dev = dev,
2267                 .net = net,
2268         };
2269
2270         fib6_clean_all(net, fib6_ifdown, 0, &adn);
2271         icmp6_clean_all(fib6_ifdown, &adn);
2272 }
2273
2274 struct rt6_mtu_change_arg {
2275         struct net_device *dev;
2276         unsigned int mtu;
2277 };
2278
2279 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2280 {
2281         struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2282         struct inet6_dev *idev;
2283
2284         /* In IPv6 pmtu discovery is not optional,
2285            so that RTAX_MTU lock cannot disable it.
2286            We still use this lock to block changes
2287            caused by addrconf/ndisc.
2288         */
2289
2290         idev = __in6_dev_get(arg->dev);
2291         if (!idev)
2292                 return 0;
2293
2294         /* For administrative MTU increase, there is no way to discover
2295            IPv6 PMTU increase, so PMTU increase should be updated here.
2296            Since RFC 1981 doesn't include administrative MTU increase
2297            update PMTU increase is a MUST. (i.e. jumbo frame)
2298          */
2299         /*
2300            If new MTU is less than route PMTU, this new MTU will be the
2301            lowest MTU in the path, update the route PMTU to reflect PMTU
2302            decreases; if new MTU is greater than route PMTU, and the
2303            old MTU is the lowest MTU in the path, update the route PMTU
2304            to reflect the increase. In this case if the other nodes' MTU
2305            also have the lowest MTU, TOO BIG MESSAGE will be lead to
2306            PMTU discouvery.
2307          */
2308         if (rt->dst.dev == arg->dev &&
2309             !dst_metric_locked(&rt->dst, RTAX_MTU) &&
2310             (dst_mtu(&rt->dst) >= arg->mtu ||
2311              (dst_mtu(&rt->dst) < arg->mtu &&
2312               dst_mtu(&rt->dst) == idev->cnf.mtu6))) {
2313                 dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
2314         }
2315         return 0;
2316 }
2317
2318 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
2319 {
2320         struct rt6_mtu_change_arg arg = {
2321                 .dev = dev,
2322                 .mtu = mtu,
2323         };
2324
2325         fib6_clean_all(dev_net(dev), rt6_mtu_change_route, 0, &arg);
2326 }
2327
2328 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2329         [RTA_GATEWAY]           = { .len = sizeof(struct in6_addr) },
2330         [RTA_OIF]               = { .type = NLA_U32 },
2331         [RTA_IIF]               = { .type = NLA_U32 },
2332         [RTA_PRIORITY]          = { .type = NLA_U32 },
2333         [RTA_METRICS]           = { .type = NLA_NESTED },
2334         [RTA_MULTIPATH]         = { .len = sizeof(struct rtnexthop) },
2335 };
2336
2337 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2338                               struct fib6_config *cfg)
2339 {
2340         struct rtmsg *rtm;
2341         struct nlattr *tb[RTA_MAX+1];
2342         int err;
2343
2344         err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2345         if (err < 0)
2346                 goto errout;
2347
2348         err = -EINVAL;
2349         rtm = nlmsg_data(nlh);
2350         memset(cfg, 0, sizeof(*cfg));
2351
2352         cfg->fc_table = rtm->rtm_table;
2353         cfg->fc_dst_len = rtm->rtm_dst_len;
2354         cfg->fc_src_len = rtm->rtm_src_len;
2355         cfg->fc_flags = RTF_UP;
2356         cfg->fc_protocol = rtm->rtm_protocol;
2357         cfg->fc_type = rtm->rtm_type;
2358
2359         if (rtm->rtm_type == RTN_UNREACHABLE ||
2360             rtm->rtm_type == RTN_BLACKHOLE ||
2361             rtm->rtm_type == RTN_PROHIBIT ||
2362             rtm->rtm_type == RTN_THROW)
2363                 cfg->fc_flags |= RTF_REJECT;
2364
2365         if (rtm->rtm_type == RTN_LOCAL)
2366                 cfg->fc_flags |= RTF_LOCAL;
2367
2368         cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid;
2369         cfg->fc_nlinfo.nlh = nlh;
2370         cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2371
2372         if (tb[RTA_GATEWAY]) {
2373                 nla_memcpy(&cfg->fc_gateway, tb[RTA_GATEWAY], 16);
2374                 cfg->fc_flags |= RTF_GATEWAY;
2375         }
2376
2377         if (tb[RTA_DST]) {
2378                 int plen = (rtm->rtm_dst_len + 7) >> 3;
2379
2380                 if (nla_len(tb[RTA_DST]) < plen)
2381                         goto errout;
2382
2383                 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2384         }
2385
2386         if (tb[RTA_SRC]) {
2387                 int plen = (rtm->rtm_src_len + 7) >> 3;
2388
2389                 if (nla_len(tb[RTA_SRC]) < plen)
2390                         goto errout;
2391
2392                 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2393         }
2394
2395         if (tb[RTA_PREFSRC])
2396                 nla_memcpy(&cfg->fc_prefsrc, tb[RTA_PREFSRC], 16);
2397
2398         if (tb[RTA_OIF])
2399                 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2400
2401         if (tb[RTA_PRIORITY])
2402                 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2403
2404         if (tb[RTA_METRICS]) {
2405                 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2406                 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2407         }
2408
2409         if (tb[RTA_TABLE])
2410                 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2411
2412         if (tb[RTA_MULTIPATH]) {
2413                 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
2414                 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
2415         }
2416
2417         err = 0;
2418 errout:
2419         return err;
2420 }
2421
2422 static int ip6_route_multipath(struct fib6_config *cfg, int add)
2423 {
2424         struct fib6_config r_cfg;
2425         struct rtnexthop *rtnh;
2426         int remaining;
2427         int attrlen;
2428         int err = 0, last_err = 0;
2429
2430 beginning:
2431         rtnh = (struct rtnexthop *)cfg->fc_mp;
2432         remaining = cfg->fc_mp_len;
2433
2434         /* Parse a Multipath Entry */
2435         while (rtnh_ok(rtnh, remaining)) {
2436                 memcpy(&r_cfg, cfg, sizeof(*cfg));
2437                 if (rtnh->rtnh_ifindex)
2438                         r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
2439
2440                 attrlen = rtnh_attrlen(rtnh);
2441                 if (attrlen > 0) {
2442                         struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
2443
2444                         nla = nla_find(attrs, attrlen, RTA_GATEWAY);
2445                         if (nla) {
2446                                 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
2447                                 r_cfg.fc_flags |= RTF_GATEWAY;
2448                         }
2449                 }
2450                 err = add ? ip6_route_add(&r_cfg) : ip6_route_del(&r_cfg);
2451                 if (err) {
2452                         last_err = err;
2453                         /* If we are trying to remove a route, do not stop the
2454                          * loop when ip6_route_del() fails (because next hop is
2455                          * already gone), we should try to remove all next hops.
2456                          */
2457                         if (add) {
2458                                 /* If add fails, we should try to delete all
2459                                  * next hops that have been already added.
2460                                  */
2461                                 add = 0;
2462                                 goto beginning;
2463                         }
2464                 }
2465                 /* Because each route is added like a single route we remove
2466                  * this flag after the first nexthop (if there is a collision,
2467                  * we have already fail to add the first nexthop:
2468                  * fib6_add_rt2node() has reject it).
2469                  */
2470                 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~NLM_F_EXCL;
2471                 rtnh = rtnh_next(rtnh, &remaining);
2472         }
2473
2474         return last_err;
2475 }
2476
2477 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr* nlh)
2478 {
2479         struct fib6_config cfg;
2480         int err;
2481
2482         err = rtm_to_fib6_config(skb, nlh, &cfg);
2483         if (err < 0)
2484                 return err;
2485
2486         if (cfg.fc_mp)
2487                 return ip6_route_multipath(&cfg, 0);
2488         else
2489                 return ip6_route_del(&cfg);
2490 }
2491
2492 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr* nlh)
2493 {
2494         struct fib6_config cfg;
2495         int err;
2496
2497         err = rtm_to_fib6_config(skb, nlh, &cfg);
2498         if (err < 0)
2499                 return err;
2500
2501         if (cfg.fc_mp)
2502                 return ip6_route_multipath(&cfg, 1);
2503         else
2504                 return ip6_route_add(&cfg);
2505 }
2506
2507 static inline size_t rt6_nlmsg_size(void)
2508 {
2509         return NLMSG_ALIGN(sizeof(struct rtmsg))
2510                + nla_total_size(16) /* RTA_SRC */
2511                + nla_total_size(16) /* RTA_DST */
2512                + nla_total_size(16) /* RTA_GATEWAY */
2513                + nla_total_size(16) /* RTA_PREFSRC */
2514                + nla_total_size(4) /* RTA_TABLE */
2515                + nla_total_size(4) /* RTA_IIF */
2516                + nla_total_size(4) /* RTA_OIF */
2517                + nla_total_size(4) /* RTA_PRIORITY */
2518                + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
2519                + nla_total_size(sizeof(struct rta_cacheinfo));
2520 }
2521
2522 static int rt6_fill_node(struct net *net,
2523                          struct sk_buff *skb, struct rt6_info *rt,
2524                          struct in6_addr *dst, struct in6_addr *src,
2525                          int iif, int type, u32 portid, u32 seq,
2526                          int prefix, int nowait, unsigned int flags)
2527 {
2528         struct rtmsg *rtm;
2529         struct nlmsghdr *nlh;
2530         long expires;
2531         u32 table;
2532
2533         if (prefix) {   /* user wants prefix routes only */
2534                 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
2535                         /* success since this is not a prefix route */
2536                         return 1;
2537                 }
2538         }
2539
2540         nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
2541         if (!nlh)
2542                 return -EMSGSIZE;
2543
2544         rtm = nlmsg_data(nlh);
2545         rtm->rtm_family = AF_INET6;
2546         rtm->rtm_dst_len = rt->rt6i_dst.plen;
2547         rtm->rtm_src_len = rt->rt6i_src.plen;
2548         rtm->rtm_tos = 0;
2549         if (rt->rt6i_table)
2550                 table = rt->rt6i_table->tb6_id;
2551         else
2552                 table = RT6_TABLE_UNSPEC;
2553         rtm->rtm_table = table;
2554         if (nla_put_u32(skb, RTA_TABLE, table))
2555                 goto nla_put_failure;
2556         if (rt->rt6i_flags & RTF_REJECT) {
2557                 switch (rt->dst.error) {
2558                 case -EINVAL:
2559                         rtm->rtm_type = RTN_BLACKHOLE;
2560                         break;
2561                 case -EACCES:
2562                         rtm->rtm_type = RTN_PROHIBIT;
2563                         break;
2564                 case -EAGAIN:
2565                         rtm->rtm_type = RTN_THROW;
2566                         break;
2567                 default:
2568                         rtm->rtm_type = RTN_UNREACHABLE;
2569                         break;
2570                 }
2571         }
2572         else if (rt->rt6i_flags & RTF_LOCAL)
2573                 rtm->rtm_type = RTN_LOCAL;
2574         else if (rt->dst.dev && (rt->dst.dev->flags & IFF_LOOPBACK))
2575                 rtm->rtm_type = RTN_LOCAL;
2576         else
2577                 rtm->rtm_type = RTN_UNICAST;
2578         rtm->rtm_flags = 0;
2579         rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2580         rtm->rtm_protocol = rt->rt6i_protocol;
2581         if (rt->rt6i_flags & RTF_DYNAMIC)
2582                 rtm->rtm_protocol = RTPROT_REDIRECT;
2583         else if (rt->rt6i_flags & RTF_ADDRCONF) {
2584                 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ROUTEINFO))
2585                         rtm->rtm_protocol = RTPROT_RA;
2586                 else
2587                         rtm->rtm_protocol = RTPROT_KERNEL;
2588         }
2589
2590         if (rt->rt6i_flags & RTF_CACHE)
2591                 rtm->rtm_flags |= RTM_F_CLONED;
2592
2593         if (dst) {
2594                 if (nla_put(skb, RTA_DST, 16, dst))
2595                         goto nla_put_failure;
2596                 rtm->rtm_dst_len = 128;
2597         } else if (rtm->rtm_dst_len)
2598                 if (nla_put(skb, RTA_DST, 16, &rt->rt6i_dst.addr))
2599                         goto nla_put_failure;
2600 #ifdef CONFIG_IPV6_SUBTREES
2601         if (src) {
2602                 if (nla_put(skb, RTA_SRC, 16, src))
2603                         goto nla_put_failure;
2604                 rtm->rtm_src_len = 128;
2605         } else if (rtm->rtm_src_len &&
2606                    nla_put(skb, RTA_SRC, 16, &rt->rt6i_src.addr))
2607                 goto nla_put_failure;
2608 #endif
2609         if (iif) {
2610 #ifdef CONFIG_IPV6_MROUTE
2611                 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
2612                         int err = ip6mr_get_route(net, skb, rtm, nowait);
2613                         if (err <= 0) {
2614                                 if (!nowait) {
2615                                         if (err == 0)
2616                                                 return 0;
2617                                         goto nla_put_failure;
2618                                 } else {
2619                                         if (err == -EMSGSIZE)
2620                                                 goto nla_put_failure;
2621                                 }
2622                         }
2623                 } else
2624 #endif
2625                         if (nla_put_u32(skb, RTA_IIF, iif))
2626                                 goto nla_put_failure;
2627         } else if (dst) {
2628                 struct in6_addr saddr_buf;
2629                 if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0 &&
2630                     nla_put(skb, RTA_PREFSRC, 16, &saddr_buf))
2631                         goto nla_put_failure;
2632         }
2633
2634         if (rt->rt6i_prefsrc.plen) {
2635                 struct in6_addr saddr_buf;
2636                 saddr_buf = rt->rt6i_prefsrc.addr;
2637                 if (nla_put(skb, RTA_PREFSRC, 16, &saddr_buf))
2638                         goto nla_put_failure;
2639         }
2640
2641         if (rtnetlink_put_metrics(skb, dst_metrics_ptr(&rt->dst)) < 0)
2642                 goto nla_put_failure;
2643
2644         if (rt->rt6i_flags & RTF_GATEWAY) {
2645                 if (nla_put(skb, RTA_GATEWAY, 16, &rt->rt6i_gateway) < 0)
2646                         goto nla_put_failure;
2647         }
2648
2649         if (rt->dst.dev &&
2650             nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex))
2651                 goto nla_put_failure;
2652         if (nla_put_u32(skb, RTA_PRIORITY, rt->rt6i_metric))
2653                 goto nla_put_failure;
2654
2655         expires = (rt->rt6i_flags & RTF_EXPIRES) ? rt->dst.expires - jiffies : 0;
2656
2657         if (rtnl_put_cacheinfo(skb, &rt->dst, 0, expires, rt->dst.error) < 0)
2658                 goto nla_put_failure;
2659
2660         return nlmsg_end(skb, nlh);
2661
2662 nla_put_failure:
2663         nlmsg_cancel(skb, nlh);
2664         return -EMSGSIZE;
2665 }
2666
2667 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
2668 {
2669         struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
2670         int prefix;
2671
2672         if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
2673                 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
2674                 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
2675         } else
2676                 prefix = 0;
2677
2678         return rt6_fill_node(arg->net,
2679                      arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
2680                      NETLINK_CB(arg->cb->skb).portid, arg->cb->nlh->nlmsg_seq,
2681                      prefix, 0, NLM_F_MULTI);
2682 }
2683
2684 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr* nlh)
2685 {
2686         struct net *net = sock_net(in_skb->sk);
2687         struct nlattr *tb[RTA_MAX+1];
2688         struct rt6_info *rt;
2689         struct sk_buff *skb;
2690         struct rtmsg *rtm;
2691         struct flowi6 fl6;
2692         int err, iif = 0, oif = 0;
2693
2694         err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2695         if (err < 0)
2696                 goto errout;
2697
2698         err = -EINVAL;
2699         memset(&fl6, 0, sizeof(fl6));
2700
2701         if (tb[RTA_SRC]) {
2702                 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
2703                         goto errout;
2704
2705                 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
2706         }
2707
2708         if (tb[RTA_DST]) {
2709                 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
2710                         goto errout;
2711
2712                 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
2713         }
2714
2715         if (tb[RTA_IIF])
2716                 iif = nla_get_u32(tb[RTA_IIF]);
2717
2718         if (tb[RTA_OIF])
2719                 oif = nla_get_u32(tb[RTA_OIF]);
2720
2721         if (iif) {
2722                 struct net_device *dev;
2723                 int flags = 0;
2724
2725                 dev = __dev_get_by_index(net, iif);
2726                 if (!dev) {
2727                         err = -ENODEV;
2728                         goto errout;
2729                 }
2730
2731                 fl6.flowi6_iif = iif;
2732
2733                 if (!ipv6_addr_any(&fl6.saddr))
2734                         flags |= RT6_LOOKUP_F_HAS_SADDR;
2735
2736                 rt = (struct rt6_info *)ip6_route_input_lookup(net, dev, &fl6,
2737                                                                flags);
2738         } else {
2739                 fl6.flowi6_oif = oif;
2740
2741                 rt = (struct rt6_info *)ip6_route_output(net, NULL, &fl6);
2742         }
2743
2744         skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
2745         if (!skb) {
2746                 ip6_rt_put(rt);
2747                 err = -ENOBUFS;
2748                 goto errout;
2749         }
2750
2751         /* Reserve room for dummy headers, this skb can pass
2752            through good chunk of routing engine.
2753          */
2754         skb_reset_mac_header(skb);
2755         skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
2756
2757         skb_dst_set(skb, &rt->dst);
2758
2759         err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
2760                             RTM_NEWROUTE, NETLINK_CB(in_skb).portid,
2761                             nlh->nlmsg_seq, 0, 0, 0);
2762         if (err < 0) {
2763                 kfree_skb(skb);
2764                 goto errout;
2765         }
2766
2767         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
2768 errout:
2769         return err;
2770 }
2771
2772 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info)
2773 {
2774         struct sk_buff *skb;
2775         struct net *net = info->nl_net;
2776         u32 seq;
2777         int err;
2778
2779         err = -ENOBUFS;
2780         seq = info->nlh ? info->nlh->nlmsg_seq : 0;
2781
2782         skb = nlmsg_new(rt6_nlmsg_size(), gfp_any());
2783         if (!skb)
2784                 goto errout;
2785
2786         err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
2787                                 event, info->portid, seq, 0, 0, 0);
2788         if (err < 0) {
2789                 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
2790                 WARN_ON(err == -EMSGSIZE);
2791                 kfree_skb(skb);
2792                 goto errout;
2793         }
2794         rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
2795                     info->nlh, gfp_any());
2796         return;
2797 errout:
2798         if (err < 0)
2799                 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
2800 }
2801
2802 static int ip6_route_dev_notify(struct notifier_block *this,
2803                                 unsigned long event, void *ptr)
2804 {
2805         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2806         struct net *net = dev_net(dev);
2807
2808         if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
2809                 net->ipv6.ip6_null_entry->dst.dev = dev;
2810                 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
2811 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2812                 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
2813                 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
2814                 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
2815                 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
2816 #endif
2817         }
2818
2819         return NOTIFY_OK;
2820 }
2821
2822 /*
2823  *      /proc
2824  */
2825
2826 #ifdef CONFIG_PROC_FS
2827
2828 static const struct file_operations ipv6_route_proc_fops = {
2829         .owner          = THIS_MODULE,
2830         .open           = ipv6_route_open,
2831         .read           = seq_read,
2832         .llseek         = seq_lseek,
2833         .release        = seq_release_net,
2834 };
2835
2836 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
2837 {
2838         struct net *net = (struct net *)seq->private;
2839         seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
2840                    net->ipv6.rt6_stats->fib_nodes,
2841                    net->ipv6.rt6_stats->fib_route_nodes,
2842                    net->ipv6.rt6_stats->fib_rt_alloc,
2843                    net->ipv6.rt6_stats->fib_rt_entries,
2844                    net->ipv6.rt6_stats->fib_rt_cache,
2845                    dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
2846                    net->ipv6.rt6_stats->fib_discarded_routes);
2847
2848         return 0;
2849 }
2850
2851 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
2852 {
2853         return single_open_net(inode, file, rt6_stats_seq_show);
2854 }
2855
2856 static const struct file_operations rt6_stats_seq_fops = {
2857         .owner   = THIS_MODULE,
2858         .open    = rt6_stats_seq_open,
2859         .read    = seq_read,
2860         .llseek  = seq_lseek,
2861         .release = single_release_net,
2862 };
2863 #endif  /* CONFIG_PROC_FS */
2864
2865 #ifdef CONFIG_SYSCTL
2866
2867 static
2868 int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
2869                               void __user *buffer, size_t *lenp, loff_t *ppos)
2870 {
2871         struct net *net;
2872         int delay;
2873         if (!write)
2874                 return -EINVAL;
2875
2876         net = (struct net *)ctl->extra1;
2877         delay = net->ipv6.sysctl.flush_delay;
2878         proc_dointvec(ctl, write, buffer, lenp, ppos);
2879         fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
2880         return 0;
2881 }
2882
2883 struct ctl_table ipv6_route_table_template[] = {
2884         {
2885                 .procname       =       "flush",
2886                 .data           =       &init_net.ipv6.sysctl.flush_delay,
2887                 .maxlen         =       sizeof(int),
2888                 .mode           =       0200,
2889                 .proc_handler   =       ipv6_sysctl_rtcache_flush
2890         },
2891         {
2892                 .procname       =       "gc_thresh",
2893                 .data           =       &ip6_dst_ops_template.gc_thresh,
2894                 .maxlen         =       sizeof(int),
2895                 .mode           =       0644,
2896                 .proc_handler   =       proc_dointvec,
2897         },
2898         {
2899                 .procname       =       "max_size",
2900                 .data           =       &init_net.ipv6.sysctl.ip6_rt_max_size,
2901                 .maxlen         =       sizeof(int),
2902                 .mode           =       0644,
2903                 .proc_handler   =       proc_dointvec,
2904         },
2905         {
2906                 .procname       =       "gc_min_interval",
2907                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2908                 .maxlen         =       sizeof(int),
2909                 .mode           =       0644,
2910                 .proc_handler   =       proc_dointvec_jiffies,
2911         },
2912         {
2913                 .procname       =       "gc_timeout",
2914                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
2915                 .maxlen         =       sizeof(int),
2916                 .mode           =       0644,
2917                 .proc_handler   =       proc_dointvec_jiffies,
2918         },
2919         {
2920                 .procname       =       "gc_interval",
2921                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_interval,
2922                 .maxlen         =       sizeof(int),
2923                 .mode           =       0644,
2924                 .proc_handler   =       proc_dointvec_jiffies,
2925         },
2926         {
2927                 .procname       =       "gc_elasticity",
2928                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
2929                 .maxlen         =       sizeof(int),
2930                 .mode           =       0644,
2931                 .proc_handler   =       proc_dointvec,
2932         },
2933         {
2934                 .procname       =       "mtu_expires",
2935                 .data           =       &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
2936                 .maxlen         =       sizeof(int),
2937                 .mode           =       0644,
2938                 .proc_handler   =       proc_dointvec_jiffies,
2939         },
2940         {
2941                 .procname       =       "min_adv_mss",
2942                 .data           =       &init_net.ipv6.sysctl.ip6_rt_min_advmss,
2943                 .maxlen         =       sizeof(int),
2944                 .mode           =       0644,
2945                 .proc_handler   =       proc_dointvec,
2946         },
2947         {
2948                 .procname       =       "gc_min_interval_ms",
2949                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2950                 .maxlen         =       sizeof(int),
2951                 .mode           =       0644,
2952                 .proc_handler   =       proc_dointvec_ms_jiffies,
2953         },
2954         { }
2955 };
2956
2957 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
2958 {
2959         struct ctl_table *table;
2960
2961         table = kmemdup(ipv6_route_table_template,
2962                         sizeof(ipv6_route_table_template),
2963                         GFP_KERNEL);
2964
2965         if (table) {
2966                 table[0].data = &net->ipv6.sysctl.flush_delay;
2967                 table[0].extra1 = net;
2968                 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
2969                 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
2970                 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2971                 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
2972                 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
2973                 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
2974                 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
2975                 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
2976                 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2977
2978                 /* Don't export sysctls to unprivileged users */
2979                 if (net->user_ns != &init_user_ns)
2980                         table[0].procname = NULL;
2981         }
2982
2983         return table;
2984 }
2985 #endif
2986
2987 static int __net_init ip6_route_net_init(struct net *net)
2988 {
2989         int ret = -ENOMEM;
2990
2991         memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
2992                sizeof(net->ipv6.ip6_dst_ops));
2993
2994         if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
2995                 goto out_ip6_dst_ops;
2996
2997         net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
2998                                            sizeof(*net->ipv6.ip6_null_entry),
2999                                            GFP_KERNEL);
3000         if (!net->ipv6.ip6_null_entry)
3001                 goto out_ip6_dst_entries;
3002         net->ipv6.ip6_null_entry->dst.path =
3003                 (struct dst_entry *)net->ipv6.ip6_null_entry;
3004         net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3005         dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
3006                          ip6_template_metrics, true);
3007
3008 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3009         net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
3010                                                sizeof(*net->ipv6.ip6_prohibit_entry),
3011                                                GFP_KERNEL);
3012         if (!net->ipv6.ip6_prohibit_entry)
3013                 goto out_ip6_null_entry;
3014         net->ipv6.ip6_prohibit_entry->dst.path =
3015                 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
3016         net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3017         dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
3018                          ip6_template_metrics, true);
3019
3020         net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
3021                                                sizeof(*net->ipv6.ip6_blk_hole_entry),
3022                                                GFP_KERNEL);
3023         if (!net->ipv6.ip6_blk_hole_entry)
3024                 goto out_ip6_prohibit_entry;
3025         net->ipv6.ip6_blk_hole_entry->dst.path =
3026                 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
3027         net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3028         dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
3029                          ip6_template_metrics, true);
3030 #endif
3031
3032         net->ipv6.sysctl.flush_delay = 0;
3033         net->ipv6.sysctl.ip6_rt_max_size = 4096;
3034         net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
3035         net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
3036         net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
3037         net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
3038         net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
3039         net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
3040
3041         net->ipv6.ip6_rt_gc_expire = 30*HZ;
3042
3043         ret = 0;
3044 out:
3045         return ret;
3046
3047 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3048 out_ip6_prohibit_entry:
3049         kfree(net->ipv6.ip6_prohibit_entry);
3050 out_ip6_null_entry:
3051         kfree(net->ipv6.ip6_null_entry);
3052 #endif
3053 out_ip6_dst_entries:
3054         dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3055 out_ip6_dst_ops:
3056         goto out;
3057 }
3058
3059 static void __net_exit ip6_route_net_exit(struct net *net)
3060 {
3061         kfree(net->ipv6.ip6_null_entry);
3062 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3063         kfree(net->ipv6.ip6_prohibit_entry);
3064         kfree(net->ipv6.ip6_blk_hole_entry);
3065 #endif
3066         dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3067 }
3068
3069 static int __net_init ip6_route_net_init_late(struct net *net)
3070 {
3071 #ifdef CONFIG_PROC_FS
3072         proc_create("ipv6_route", 0, net->proc_net, &ipv6_route_proc_fops);
3073         proc_create("rt6_stats", S_IRUGO, net->proc_net, &rt6_stats_seq_fops);
3074 #endif
3075         return 0;
3076 }
3077
3078 static void __net_exit ip6_route_net_exit_late(struct net *net)
3079 {
3080 #ifdef CONFIG_PROC_FS
3081         remove_proc_entry("ipv6_route", net->proc_net);
3082         remove_proc_entry("rt6_stats", net->proc_net);
3083 #endif
3084 }
3085
3086 static struct pernet_operations ip6_route_net_ops = {
3087         .init = ip6_route_net_init,
3088         .exit = ip6_route_net_exit,
3089 };
3090
3091 static int __net_init ipv6_inetpeer_init(struct net *net)
3092 {
3093         struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
3094
3095         if (!bp)
3096                 return -ENOMEM;
3097         inet_peer_base_init(bp);
3098         net->ipv6.peers = bp;
3099         return 0;
3100 }
3101
3102 static void __net_exit ipv6_inetpeer_exit(struct net *net)
3103 {
3104         struct inet_peer_base *bp = net->ipv6.peers;
3105
3106         net->ipv6.peers = NULL;
3107         inetpeer_invalidate_tree(bp);
3108         kfree(bp);
3109 }
3110
3111 static struct pernet_operations ipv6_inetpeer_ops = {
3112         .init   =       ipv6_inetpeer_init,
3113         .exit   =       ipv6_inetpeer_exit,
3114 };
3115
3116 static struct pernet_operations ip6_route_net_late_ops = {
3117         .init = ip6_route_net_init_late,
3118         .exit = ip6_route_net_exit_late,
3119 };
3120
3121 static struct notifier_block ip6_route_dev_notifier = {
3122         .notifier_call = ip6_route_dev_notify,
3123         .priority = 0,
3124 };
3125
3126 int __init ip6_route_init(void)
3127 {
3128         int ret;
3129
3130         ret = -ENOMEM;
3131         ip6_dst_ops_template.kmem_cachep =
3132                 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
3133                                   SLAB_HWCACHE_ALIGN, NULL);
3134         if (!ip6_dst_ops_template.kmem_cachep)
3135                 goto out;
3136
3137         ret = dst_entries_init(&ip6_dst_blackhole_ops);
3138         if (ret)
3139                 goto out_kmem_cache;
3140
3141         ret = register_pernet_subsys(&ipv6_inetpeer_ops);
3142         if (ret)
3143                 goto out_dst_entries;
3144
3145         ret = register_pernet_subsys(&ip6_route_net_ops);
3146         if (ret)
3147                 goto out_register_inetpeer;
3148
3149         ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
3150
3151         /* Registering of the loopback is done before this portion of code,
3152          * the loopback reference in rt6_info will not be taken, do it
3153          * manually for init_net */
3154         init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
3155         init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3156   #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3157         init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
3158         init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3159         init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
3160         init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3161   #endif
3162         ret = fib6_init();
3163         if (ret)
3164                 goto out_register_subsys;
3165
3166         ret = xfrm6_init();
3167         if (ret)
3168                 goto out_fib6_init;
3169
3170         ret = fib6_rules_init();
3171         if (ret)
3172                 goto xfrm6_init;
3173
3174         ret = register_pernet_subsys(&ip6_route_net_late_ops);
3175         if (ret)
3176                 goto fib6_rules_init;
3177
3178         ret = -ENOBUFS;
3179         if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL, NULL) ||
3180             __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL, NULL) ||
3181             __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL, NULL))
3182                 goto out_register_late_subsys;
3183
3184         ret = register_netdevice_notifier(&ip6_route_dev_notifier);
3185         if (ret)
3186                 goto out_register_late_subsys;
3187
3188 out:
3189         return ret;
3190
3191 out_register_late_subsys:
3192         unregister_pernet_subsys(&ip6_route_net_late_ops);
3193 fib6_rules_init:
3194         fib6_rules_cleanup();
3195 xfrm6_init:
3196         xfrm6_fini();
3197 out_fib6_init:
3198         fib6_gc_cleanup();
3199 out_register_subsys:
3200         unregister_pernet_subsys(&ip6_route_net_ops);
3201 out_register_inetpeer:
3202         unregister_pernet_subsys(&ipv6_inetpeer_ops);
3203 out_dst_entries:
3204         dst_entries_destroy(&ip6_dst_blackhole_ops);
3205 out_kmem_cache:
3206         kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3207         goto out;
3208 }
3209
3210 void ip6_route_cleanup(void)
3211 {
3212         unregister_netdevice_notifier(&ip6_route_dev_notifier);
3213         unregister_pernet_subsys(&ip6_route_net_late_ops);
3214         fib6_rules_cleanup();
3215         xfrm6_fini();
3216         fib6_gc_cleanup();
3217         unregister_pernet_subsys(&ipv6_inetpeer_ops);
3218         unregister_pernet_subsys(&ip6_route_net_ops);
3219         dst_entries_destroy(&ip6_dst_blackhole_ops);
3220         kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3221 }