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 static struct rt6_info *ip6_rt_copy(struct rt6_info *ort,
  69                                     const struct in6_addr *dest);
  70 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie);
  71 static unsigned int      ip6_default_advmss(const struct dst_entry *dst);
  72 static unsigned int      ip6_mtu(const struct dst_entry *dst);
  73 static struct dst_entry *ip6_negative_advice(struct dst_entry *);
  74 static void             ip6_dst_destroy(struct dst_entry *);
  75 static void             ip6_dst_ifdown(struct dst_entry *,
  76                                        struct net_device *dev, int how);
  77 static int               ip6_dst_gc(struct dst_ops *ops);
  78
  79 static int              ip6_pkt_discard(struct sk_buff *skb);
  80 static int              ip6_pkt_discard_out(struct sk_buff *skb);
  81 static void             ip6_link_failure(struct sk_buff *skb);
  82 static void             ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
  83                                            struct sk_buff *skb, u32 mtu);
  84 static void             rt6_do_redirect(struct dst_entry *dst, struct sock *sk,
  85                                         struct sk_buff *skb);
  86
  87 #ifdef CONFIG_IPV6_ROUTE_INFO
  88 static struct rt6_info *rt6_add_route_info(struct net *net,
  89                                            const struct in6_addr *prefix, int prefixlen,
  90                                            const struct in6_addr *gwaddr, int ifindex,
  91                                            unsigned int pref);
  92 static struct rt6_info *rt6_get_route_info(struct net *net,
  93                                            const struct in6_addr *prefix, int prefixlen,
  94                                            const struct in6_addr *gwaddr, int ifindex);
  95 #endif
  96
  97 static u32 *ipv6_cow_metrics(struct dst_entry *dst, unsigned long old)
  98 {
  99         struct rt6_info *rt = (struct rt6_info *) dst;
 100         struct inet_peer *peer;
 101         u32 *p = NULL;
 102
 103         if (!(rt->dst.flags & DST_HOST))
 104                 return NULL;
 105
 106         peer = rt6_get_peer_create(rt);
 107         if (peer) {
 108                 u32 *old_p = __DST_METRICS_PTR(old);
 109                 unsigned long prev, new;
 110
 111                 p = peer->metrics;
 112                 if (inet_metrics_new(peer))
 113                         memcpy(p, old_p, sizeof(u32) * RTAX_MAX);
 114
 115                 new = (unsigned long) p;
 116                 prev = cmpxchg(&dst->_metrics, old, new);
 117
 118                 if (prev != old) {
 119                         p = __DST_METRICS_PTR(prev);
 120                         if (prev & DST_METRICS_READ_ONLY)
 121                                 p = NULL;
 122                 }
 123         }
 124         return p;
 125 }
 126
 127 static inline const void *choose_neigh_daddr(struct rt6_info *rt,
 128                                              struct sk_buff *skb,
 129                                              const void *daddr)
 130 {
 131         struct in6_addr *p = &rt->rt6i_gateway;
 132
 133         if (!ipv6_addr_any(p))
 134                 return (const void *) p;
 135         else if (skb)
 136                 return &ipv6_hdr(skb)->daddr;
 137         return daddr;
 138 }
 139
 140 static struct neighbour *ip6_neigh_lookup(const struct dst_entry *dst,
 141                                           struct sk_buff *skb,
 142                                           const void *daddr)
 143 {
 144         struct rt6_info *rt = (struct rt6_info *) dst;
 145         struct neighbour *n;
 146
 147         daddr = choose_neigh_daddr(rt, skb, daddr);
 148         n = __ipv6_neigh_lookup(&nd_tbl, dst->dev, daddr);
 149         if (n)
 150                 return n;
 151         return neigh_create(&nd_tbl, daddr, dst->dev);
 152 }
 153
 154 static int rt6_bind_neighbour(struct rt6_info *rt, struct net_device *dev)
 155 {
 156         struct neighbour *n = __ipv6_neigh_lookup(&nd_tbl, dev, &rt->rt6i_gateway);
 157         if (!n) {
 158                 n = neigh_create(&nd_tbl, &rt->rt6i_gateway, dev);
 159                 if (IS_ERR(n))
 160                         return PTR_ERR(n);
 161         }
 162         rt->n = n;
 163
 164         return 0;
 165 }
 166
 167 static struct dst_ops ip6_dst_ops_template = {
 168         .family                 =       AF_INET6,
 169         .protocol               =       cpu_to_be16(ETH_P_IPV6),
 170         .gc                     =       ip6_dst_gc,
 171         .gc_thresh              =       1024,
 172         .check                  =       ip6_dst_check,
 173         .default_advmss         =       ip6_default_advmss,
 174         .mtu                    =       ip6_mtu,
 175         .cow_metrics            =       ipv6_cow_metrics,
 176         .destroy                =       ip6_dst_destroy,
 177         .ifdown                 =       ip6_dst_ifdown,
 178         .negative_advice        =       ip6_negative_advice,
 179         .link_failure           =       ip6_link_failure,
 180         .update_pmtu            =       ip6_rt_update_pmtu,
 181         .redirect               =       rt6_do_redirect,
 182         .local_out              =       __ip6_local_out,
 183         .neigh_lookup           =       ip6_neigh_lookup,
 184 };
 185
 186 static unsigned int ip6_blackhole_mtu(const struct dst_entry *dst)
 187 {
 188         unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
 189
 190         return mtu ? : dst->dev->mtu;
 191 }
 192
 193 static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, struct sock *sk,
 194                                          struct sk_buff *skb, u32 mtu)
 195 {
 196 }
 197
 198 static void ip6_rt_blackhole_redirect(struct dst_entry *dst, struct sock *sk,
 199                                       struct sk_buff *skb)
 200 {
 201 }
 202
 203 static u32 *ip6_rt_blackhole_cow_metrics(struct dst_entry *dst,
 204                                          unsigned long old)
 205 {
 206         return NULL;
 207 }
 208
 209 static struct dst_ops ip6_dst_blackhole_ops = {
 210         .family                 =       AF_INET6,
 211         .protocol               =       cpu_to_be16(ETH_P_IPV6),
 212         .destroy                =       ip6_dst_destroy,
 213         .check                  =       ip6_dst_check,
 214         .mtu                    =       ip6_blackhole_mtu,
 215         .default_advmss         =       ip6_default_advmss,
 216         .update_pmtu            =       ip6_rt_blackhole_update_pmtu,
 217         .redirect               =       ip6_rt_blackhole_redirect,
 218         .cow_metrics            =       ip6_rt_blackhole_cow_metrics,
 219         .neigh_lookup           =       ip6_neigh_lookup,
 220 };
 221
 222 static const u32 ip6_template_metrics[RTAX_MAX] = {
 223         [RTAX_HOPLIMIT - 1] = 0,
 224 };
 225
 226 static const struct rt6_info ip6_null_entry_template = {
 227         .dst = {
 228                 .__refcnt       = ATOMIC_INIT(1),
 229                 .__use          = 1,
 230                 .obsolete       = DST_OBSOLETE_FORCE_CHK,
 231                 .error          = -ENETUNREACH,
 232                 .input          = ip6_pkt_discard,
 233                 .output         = ip6_pkt_discard_out,
 234         },
 235         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
 236         .rt6i_protocol  = RTPROT_KERNEL,
 237         .rt6i_metric    = ~(u32) 0,
 238         .rt6i_ref       = ATOMIC_INIT(1),
 239 };
 240
 241 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
 242
 243 static int ip6_pkt_prohibit(struct sk_buff *skb);
 244 static int ip6_pkt_prohibit_out(struct sk_buff *skb);
 245
 246 static const struct rt6_info ip6_prohibit_entry_template = {
 247         .dst = {
 248                 .__refcnt       = ATOMIC_INIT(1),
 249                 .__use          = 1,
 250                 .obsolete       = DST_OBSOLETE_FORCE_CHK,
 251                 .error          = -EACCES,
 252                 .input          = ip6_pkt_prohibit,
 253                 .output         = ip6_pkt_prohibit_out,
 254         },
 255         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
 256         .rt6i_protocol  = RTPROT_KERNEL,
 257         .rt6i_metric    = ~(u32) 0,
 258         .rt6i_ref       = ATOMIC_INIT(1),
 259 };
 260
 261 static const struct rt6_info ip6_blk_hole_entry_template = {
 262         .dst = {
 263                 .__refcnt       = ATOMIC_INIT(1),
 264                 .__use          = 1,
 265                 .obsolete       = DST_OBSOLETE_FORCE_CHK,
 266                 .error          = -EINVAL,
 267                 .input          = dst_discard,
 268                 .output         = dst_discard,
 269         },
 270         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
 271         .rt6i_protocol  = RTPROT_KERNEL,
 272         .rt6i_metric    = ~(u32) 0,
 273         .rt6i_ref       = ATOMIC_INIT(1),
 274 };
 275
 276 #endif
 277
 278 /* allocate dst with ip6_dst_ops */
 279 static inline struct rt6_info *ip6_dst_alloc(struct net *net,
 280                                              struct net_device *dev,
 281                                              int flags,
 282                                              struct fib6_table *table)
 283 {
 284         struct rt6_info *rt = dst_alloc(&net->ipv6.ip6_dst_ops, dev,
 285                                         0, DST_OBSOLETE_FORCE_CHK, flags);
 286
 287         if (rt) {
 288                 struct dst_entry *dst = &rt->dst;
 289
 290                 memset(dst + 1, 0, sizeof(*rt) - sizeof(*dst));
 291                 rt6_init_peer(rt, table ? &table->tb6_peers : net->ipv6.peers);
 292                 rt->rt6i_genid = rt_genid(net);
 293                 INIT_LIST_HEAD(&rt->rt6i_siblings);
 294                 rt->rt6i_nsiblings = 0;
 295         }
 296         return rt;
 297 }
 298
 299 static void ip6_dst_destroy(struct dst_entry *dst)
 300 {
 301         struct rt6_info *rt = (struct rt6_info *)dst;
 302         struct inet6_dev *idev = rt->rt6i_idev;
 303
 304         if (rt->n)
 305                 neigh_release(rt->n);
 306
 307         if (!(rt->dst.flags & DST_HOST))
 308                 dst_destroy_metrics_generic(dst);
 309
 310         if (idev) {
 311                 rt->rt6i_idev = NULL;
 312                 in6_dev_put(idev);
 313         }
 314
 315         if (!(rt->rt6i_flags & RTF_EXPIRES) && dst->from)
 316                 dst_release(dst->from);
 317
 318         if (rt6_has_peer(rt)) {
 319                 struct inet_peer *peer = rt6_peer_ptr(rt);
 320                 inet_putpeer(peer);
 321         }
 322 }
 323
 324 void rt6_bind_peer(struct rt6_info *rt, int create)
 325 {
 326         struct inet_peer_base *base;
 327         struct inet_peer *peer;
 328
 329         base = inetpeer_base_ptr(rt->_rt6i_peer);
 330         if (!base)
 331                 return;
 332
 333         peer = inet_getpeer_v6(base, &rt->rt6i_dst.addr, create);
 334         if (peer) {
 335                 if (!rt6_set_peer(rt, peer))
 336                         inet_putpeer(peer);
 337         }
 338 }
 339
 340 static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
 341                            int how)
 342 {
 343         struct rt6_info *rt = (struct rt6_info *)dst;
 344         struct inet6_dev *idev = rt->rt6i_idev;
 345         struct net_device *loopback_dev =
 346                 dev_net(dev)->loopback_dev;
 347
 348         if (dev != loopback_dev) {
 349                 if (idev && idev->dev == dev) {
 350                         struct inet6_dev *loopback_idev =
 351                                 in6_dev_get(loopback_dev);
 352                         if (loopback_idev) {
 353                                 rt->rt6i_idev = loopback_idev;
 354                                 in6_dev_put(idev);
 355                         }
 356                 }
 357                 if (rt->n && rt->n->dev == dev) {
 358                         rt->n->dev = loopback_dev;
 359                         dev_hold(loopback_dev);
 360                         dev_put(dev);
 361                 }
 362         }
 363 }
 364
 365 static bool rt6_check_expired(const struct rt6_info *rt)
 366 {
 367         if (rt->rt6i_flags & RTF_EXPIRES) {
 368                 if (time_after(jiffies, rt->dst.expires))
 369                         return true;
 370         } else if (rt->dst.from) {
 371                 return rt6_check_expired((struct rt6_info *) rt->dst.from);
 372         }
 373         return false;
 374 }
 375
 376 static bool rt6_need_strict(const struct in6_addr *daddr)
 377 {
 378         return ipv6_addr_type(daddr) &
 379                 (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
 380 }
 381
 382 /* Multipath route selection:
 383  *   Hash based function using packet header and flowlabel.
 384  * Adapted from fib_info_hashfn()
 385  */
 386 static int rt6_info_hash_nhsfn(unsigned int candidate_count,
 387                                const struct flowi6 *fl6)
 388 {
 389         unsigned int val = fl6->flowi6_proto;
 390
 391         val ^= ipv6_addr_hash(&fl6->daddr);
 392         val ^= ipv6_addr_hash(&fl6->saddr);
 393
 394         /* Work only if this not encapsulated */
 395         switch (fl6->flowi6_proto) {
 396         case IPPROTO_UDP:
 397         case IPPROTO_TCP:
 398         case IPPROTO_SCTP:
 399                 val ^= (__force u16)fl6->fl6_sport;
 400                 val ^= (__force u16)fl6->fl6_dport;
 401                 break;
 402
 403         case IPPROTO_ICMPV6:
 404                 val ^= (__force u16)fl6->fl6_icmp_type;
 405                 val ^= (__force u16)fl6->fl6_icmp_code;
 406                 break;
 407         }
 408         /* RFC6438 recommands to use flowlabel */
 409         val ^= (__force u32)fl6->flowlabel;
 410
 411         /* Perhaps, we need to tune, this function? */
 412         val = val ^ (val >> 7) ^ (val >> 12);
 413         return val % candidate_count;
 414 }
 415
 416 static struct rt6_info *rt6_multipath_select(struct rt6_info *match,
 417                                              struct flowi6 *fl6)
 418 {
 419         struct rt6_info *sibling, *next_sibling;
 420         int route_choosen;
 421
 422         route_choosen = rt6_info_hash_nhsfn(match->rt6i_nsiblings + 1, fl6);
 423         /* Don't change the route, if route_choosen == 0
 424          * (siblings does not include ourself)
 425          */
 426         if (route_choosen)
 427                 list_for_each_entry_safe(sibling, next_sibling,
 428                                 &match->rt6i_siblings, rt6i_siblings) {
 429                         route_choosen--;
 430                         if (route_choosen == 0) {
 431                                 match = sibling;
 432                                 break;
 433                         }
 434                 }
 435         return match;
 436 }
 437
 438 /*
 439  *      Route lookup. Any table->tb6_lock is implied.
 440  */
 441
 442 static inline struct rt6_info *rt6_device_match(struct net *net,
 443                                                     struct rt6_info *rt,
 444                                                     const struct in6_addr *saddr,
 445                                                     int oif,
 446                                                     int flags)
 447 {
 448         struct rt6_info *local = NULL;
 449         struct rt6_info *sprt;
 450
 451         if (!oif && ipv6_addr_any(saddr))
 452                 goto out;
 453
 454         for (sprt = rt; sprt; sprt = sprt->dst.rt6_next) {
 455                 struct net_device *dev = sprt->dst.dev;
 456
 457                 if (oif) {
 458                         if (dev->ifindex == oif)
 459                                 return sprt;
 460                         if (dev->flags & IFF_LOOPBACK) {
 461                                 if (!sprt->rt6i_idev ||
 462                                     sprt->rt6i_idev->dev->ifindex != oif) {
 463                                         if (flags & RT6_LOOKUP_F_IFACE && oif)
 464                                                 continue;
 465                                         if (local && (!oif ||
 466                                                       local->rt6i_idev->dev->ifindex == oif))
 467                                                 continue;
 468                                 }
 469                                 local = sprt;
 470                         }
 471                 } else {
 472                         if (ipv6_chk_addr(net, saddr, dev,
 473                                           flags & RT6_LOOKUP_F_IFACE))
 474                                 return sprt;
 475                 }
 476         }
 477
 478         if (oif) {
 479                 if (local)
 480                         return local;
 481
 482                 if (flags & RT6_LOOKUP_F_IFACE)
 483                         return net->ipv6.ip6_null_entry;
 484         }
 485 out:
 486         return rt;
 487 }
 488
 489 #ifdef CONFIG_IPV6_ROUTER_PREF
 490 static void rt6_probe(struct rt6_info *rt)
 491 {
 492         struct neighbour *neigh;
 493         /*
 494          * Okay, this does not seem to be appropriate
 495          * for now, however, we need to check if it
 496          * is really so; aka Router Reachability Probing.
 497          *
 498          * Router Reachability Probe MUST be rate-limited
 499          * to no more than one per minute.
 500          */
 501         neigh = rt ? rt->n : NULL;
 502         if (!neigh || (neigh->nud_state & NUD_VALID))
 503                 return;
 504         read_lock_bh(&neigh->lock);
 505         if (!(neigh->nud_state & NUD_VALID) &&
 506             time_after(jiffies, neigh->updated + rt->rt6i_idev->cnf.rtr_probe_interval)) {
 507                 struct in6_addr mcaddr;
 508                 struct in6_addr *target;
 509
 510                 neigh->updated = jiffies;
 511                 read_unlock_bh(&neigh->lock);
 512
 513                 target = (struct in6_addr *)&neigh->primary_key;
 514                 addrconf_addr_solict_mult(target, &mcaddr);
 515                 ndisc_send_ns(rt->dst.dev, NULL, target, &mcaddr, NULL);
 516         } else {
 517                 read_unlock_bh(&neigh->lock);
 518         }
 519 }
 520 #else
 521 static inline void rt6_probe(struct rt6_info *rt)
 522 {
 523 }
 524 #endif
 525
 526 /*
 527  * Default Router Selection (RFC 2461 6.3.6)
 528  */
 529 static inline int rt6_check_dev(struct rt6_info *rt, int oif)
 530 {
 531         struct net_device *dev = rt->dst.dev;
 532         if (!oif || dev->ifindex == oif)
 533                 return 2;
 534         if ((dev->flags & IFF_LOOPBACK) &&
 535             rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
 536                 return 1;
 537         return 0;
 538 }
 539
 540 static inline bool rt6_check_neigh(struct rt6_info *rt)
 541 {
 542         struct neighbour *neigh;
 543         bool ret = false;
 544
 545         neigh = rt->n;
 546         if (rt->rt6i_flags & RTF_NONEXTHOP ||
 547             !(rt->rt6i_flags & RTF_GATEWAY))
 548                 ret = true;
 549         else if (neigh) {
 550                 read_lock_bh(&neigh->lock);
 551                 if (neigh->nud_state & NUD_VALID)
 552                         ret = true;
 553 #ifdef CONFIG_IPV6_ROUTER_PREF
 554                 else if (!(neigh->nud_state & NUD_FAILED))
 555                         ret = true;
 556 #endif
 557                 read_unlock_bh(&neigh->lock);
 558         }
 559         return ret;
 560 }
 561
 562 static int rt6_score_route(struct rt6_info *rt, int oif,
 563                            int strict)
 564 {
 565         int m;
 566
 567         m = rt6_check_dev(rt, oif);
 568         if (!m && (strict & RT6_LOOKUP_F_IFACE))
 569                 return -1;
 570 #ifdef CONFIG_IPV6_ROUTER_PREF
 571         m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
 572 #endif
 573         if (!rt6_check_neigh(rt) && (strict & RT6_LOOKUP_F_REACHABLE))
 574                 return -1;
 575         return m;
 576 }
 577
 578 static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
 579                                    int *mpri, struct rt6_info *match)
 580 {
 581         int m;
 582
 583         if (rt6_check_expired(rt))
 584                 goto out;
 585
 586         m = rt6_score_route(rt, oif, strict);
 587         if (m < 0)
 588                 goto out;
 589
 590         if (m > *mpri) {
 591                 if (strict & RT6_LOOKUP_F_REACHABLE)
 592                         rt6_probe(match);
 593                 *mpri = m;
 594                 match = rt;
 595         } else if (strict & RT6_LOOKUP_F_REACHABLE) {
 596                 rt6_probe(rt);
 597         }
 598
 599 out:
 600         return match;
 601 }
 602
 603 static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
 604                                      struct rt6_info *rr_head,
 605                                      u32 metric, int oif, int strict)
 606 {
 607         struct rt6_info *rt, *match;
 608         int mpri = -1;
 609
 610         match = NULL;
 611         for (rt = rr_head; rt && rt->rt6i_metric == metric;
 612              rt = rt->dst.rt6_next)
 613                 match = find_match(rt, oif, strict, &mpri, match);
 614         for (rt = fn->leaf; rt && rt != rr_head && rt->rt6i_metric == metric;
 615              rt = rt->dst.rt6_next)
 616                 match = find_match(rt, oif, strict, &mpri, match);
 617
 618         return match;
 619 }
 620
 621 static struct rt6_info *rt6_select(struct fib6_node *fn, int oif, int strict)
 622 {
 623         struct rt6_info *match, *rt0;
 624         struct net *net;
 625
 626         rt0 = fn->rr_ptr;
 627         if (!rt0)
 628                 fn->rr_ptr = rt0 = fn->leaf;
 629
 630         match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict);
 631
 632         if (!match &&
 633             (strict & RT6_LOOKUP_F_REACHABLE)) {
 634                 struct rt6_info *next = rt0->dst.rt6_next;
 635
 636                 /* no entries matched; do round-robin */
 637                 if (!next || next->rt6i_metric != rt0->rt6i_metric)
 638                         next = fn->leaf;
 639
 640                 if (next != rt0)
 641                         fn->rr_ptr = next;
 642         }
 643
 644         net = dev_net(rt0->dst.dev);
 645         return match ? match : net->ipv6.ip6_null_entry;
 646 }
 647
 648 #ifdef CONFIG_IPV6_ROUTE_INFO
 649 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
 650                   const struct in6_addr *gwaddr)
 651 {
 652         struct net *net = dev_net(dev);
 653         struct route_info *rinfo = (struct route_info *) opt;
 654         struct in6_addr prefix_buf, *prefix;
 655         unsigned int pref;
 656         unsigned long lifetime;
 657         struct rt6_info *rt;
 658
 659         if (len < sizeof(struct route_info)) {
 660                 return -EINVAL;
 661         }
 662
 663         /* Sanity check for prefix_len and length */
 664         if (rinfo->length > 3) {
 665                 return -EINVAL;
 666         } else if (rinfo->prefix_len > 128) {
 667                 return -EINVAL;
 668         } else if (rinfo->prefix_len > 64) {
 669                 if (rinfo->length < 2) {
 670                         return -EINVAL;
 671                 }
 672         } else if (rinfo->prefix_len > 0) {
 673                 if (rinfo->length < 1) {
 674                         return -EINVAL;
 675                 }
 676         }
 677
 678         pref = rinfo->route_pref;
 679         if (pref == ICMPV6_ROUTER_PREF_INVALID)
 680                 return -EINVAL;
 681
 682         lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
 683
 684         if (rinfo->length == 3)
 685                 prefix = (struct in6_addr *)rinfo->prefix;
 686         else {
 687                 /* this function is safe */
 688                 ipv6_addr_prefix(&prefix_buf,
 689                                  (struct in6_addr *)rinfo->prefix,
 690                                  rinfo->prefix_len);
 691                 prefix = &prefix_buf;
 692         }
 693
 694         rt = rt6_get_route_info(net, prefix, rinfo->prefix_len, gwaddr,
 695                                 dev->ifindex);
 696
 697         if (rt && !lifetime) {
 698                 ip6_del_rt(rt);
 699                 rt = NULL;
 700         }
 701
 702         if (!rt && lifetime)
 703                 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr, dev->ifindex,
 704                                         pref);
 705         else if (rt)
 706                 rt->rt6i_flags = RTF_ROUTEINFO |
 707                                  (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
 708
 709         if (rt) {
 710                 if (!addrconf_finite_timeout(lifetime))
 711                         rt6_clean_expires(rt);
 712                 else
 713                         rt6_set_expires(rt, jiffies + HZ * lifetime);
 714
 715                 ip6_rt_put(rt);
 716         }
 717         return 0;
 718 }
 719 #endif
 720
 721 #define BACKTRACK(__net, saddr)                 \
 722 do { \
 723         if (rt == __net->ipv6.ip6_null_entry) { \
 724                 struct fib6_node *pn; \
 725                 while (1) { \
 726                         if (fn->fn_flags & RTN_TL_ROOT) \
 727                                 goto out; \
 728                         pn = fn->parent; \
 729                         if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn) \
 730                                 fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr); \
 731                         else \
 732                                 fn = pn; \
 733                         if (fn->fn_flags & RTN_RTINFO) \
 734                                 goto restart; \
 735                 } \
 736         } \
 737 } while (0)
 738
 739 static struct rt6_info *ip6_pol_route_lookup(struct net *net,
 740                                              struct fib6_table *table,
 741                                              struct flowi6 *fl6, int flags)
 742 {
 743         struct fib6_node *fn;
 744         struct rt6_info *rt;
 745
 746         read_lock_bh(&table->tb6_lock);
 747         fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
 748 restart:
 749         rt = fn->leaf;
 750         rt = rt6_device_match(net, rt, &fl6->saddr, fl6->flowi6_oif, flags);
 751         if (rt->rt6i_nsiblings && fl6->flowi6_oif == 0)
 752                 rt = rt6_multipath_select(rt, fl6);
 753         BACKTRACK(net, &fl6->saddr);
 754 out:
 755         dst_use(&rt->dst, jiffies);
 756         read_unlock_bh(&table->tb6_lock);
 757         return rt;
 758
 759 }
 760
 761 struct dst_entry * ip6_route_lookup(struct net *net, struct flowi6 *fl6,
 762                                     int flags)
 763 {
 764         return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_lookup);
 765 }
 766 EXPORT_SYMBOL_GPL(ip6_route_lookup);
 767
 768 struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
 769                             const struct in6_addr *saddr, int oif, int strict)
 770 {
 771         struct flowi6 fl6 = {
 772                 .flowi6_oif = oif,
 773                 .daddr = *daddr,
 774         };
 775         struct dst_entry *dst;
 776         int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
 777
 778         if (saddr) {
 779                 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
 780                 flags |= RT6_LOOKUP_F_HAS_SADDR;
 781         }
 782
 783         dst = fib6_rule_lookup(net, &fl6, flags, ip6_pol_route_lookup);
 784         if (dst->error == 0)
 785                 return (struct rt6_info *) dst;
 786
 787         dst_release(dst);
 788
 789         return NULL;
 790 }
 791
 792 EXPORT_SYMBOL(rt6_lookup);
 793
 794 /* ip6_ins_rt is called with FREE table->tb6_lock.
 795    It takes new route entry, the addition fails by any reason the
 796    route is freed. In any case, if caller does not hold it, it may
 797    be destroyed.
 798  */
 799
 800 static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info)
 801 {
 802         int err;
 803         struct fib6_table *table;
 804
 805         table = rt->rt6i_table;
 806         write_lock_bh(&table->tb6_lock);
 807         err = fib6_add(&table->tb6_root, rt, info);
 808         write_unlock_bh(&table->tb6_lock);
 809
 810         return err;
 811 }
 812
 813 int ip6_ins_rt(struct rt6_info *rt)
 814 {
 815         struct nl_info info = {
 816                 .nl_net = dev_net(rt->dst.dev),
 817         };
 818         return __ip6_ins_rt(rt, &info);
 819 }
 820
 821 static struct rt6_info *rt6_alloc_cow(struct rt6_info *ort,
 822                                       const struct in6_addr *daddr,
 823                                       const struct in6_addr *saddr)
 824 {
 825         struct rt6_info *rt;
 826
 827         /*
 828          *      Clone the route.
 829          */
 830
 831         rt = ip6_rt_copy(ort, daddr);
 832
 833         if (rt) {
 834                 int attempts = !in_softirq();
 835
 836                 if (!(rt->rt6i_flags & RTF_GATEWAY)) {
 837                         if (ort->rt6i_dst.plen != 128 &&
 838                             ipv6_addr_equal(&ort->rt6i_dst.addr, daddr))
 839                                 rt->rt6i_flags |= RTF_ANYCAST;
 840                         rt->rt6i_gateway = *daddr;
 841                 }
 842
 843                 rt->rt6i_flags |= RTF_CACHE;
 844
 845 #ifdef CONFIG_IPV6_SUBTREES
 846                 if (rt->rt6i_src.plen && saddr) {
 847                         rt->rt6i_src.addr = *saddr;
 848                         rt->rt6i_src.plen = 128;
 849                 }
 850 #endif
 851
 852         retry:
 853                 if (rt6_bind_neighbour(rt, rt->dst.dev)) {
 854                         struct net *net = dev_net(rt->dst.dev);
 855                         int saved_rt_min_interval =
 856                                 net->ipv6.sysctl.ip6_rt_gc_min_interval;
 857                         int saved_rt_elasticity =
 858                                 net->ipv6.sysctl.ip6_rt_gc_elasticity;
 859
 860                         if (attempts-- > 0) {
 861                                 net->ipv6.sysctl.ip6_rt_gc_elasticity = 1;
 862                                 net->ipv6.sysctl.ip6_rt_gc_min_interval = 0;
 863
 864                                 ip6_dst_gc(&net->ipv6.ip6_dst_ops);
 865
 866                                 net->ipv6.sysctl.ip6_rt_gc_elasticity =
 867                                         saved_rt_elasticity;
 868                                 net->ipv6.sysctl.ip6_rt_gc_min_interval =
 869                                         saved_rt_min_interval;
 870                                 goto retry;
 871                         }
 872
 873                         net_warn_ratelimited("Neighbour table overflow\n");
 874                         dst_free(&rt->dst);
 875                         return NULL;
 876                 }
 877         }
 878
 879         return rt;
 880 }
 881
 882 static struct rt6_info *rt6_alloc_clone(struct rt6_info *ort,
 883                                         const struct in6_addr *daddr)
 884 {
 885         struct rt6_info *rt = ip6_rt_copy(ort, daddr);
 886
 887         if (rt) {
 888                 rt->rt6i_flags |= RTF_CACHE;
 889                 rt->n = neigh_clone(ort->n);
 890         }
 891         return rt;
 892 }
 893
 894 static struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table, int oif,
 895                                       struct flowi6 *fl6, int flags)
 896 {
 897         struct fib6_node *fn;
 898         struct rt6_info *rt, *nrt;
 899         int strict = 0;
 900         int attempts = 3;
 901         int err;
 902         int reachable = net->ipv6.devconf_all->forwarding ? 0 : RT6_LOOKUP_F_REACHABLE;
 903
 904         strict |= flags & RT6_LOOKUP_F_IFACE;
 905
 906 relookup:
 907         read_lock_bh(&table->tb6_lock);
 908
 909 restart_2:
 910         fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
 911
 912 restart:
 913         rt = rt6_select(fn, oif, strict | reachable);
 914         if (rt->rt6i_nsiblings && oif == 0)
 915                 rt = rt6_multipath_select(rt, fl6);
 916         BACKTRACK(net, &fl6->saddr);
 917         if (rt == net->ipv6.ip6_null_entry ||
 918             rt->rt6i_flags & RTF_CACHE)
 919                 goto out;
 920
 921         dst_hold(&rt->dst);
 922         read_unlock_bh(&table->tb6_lock);
 923
 924         if (!rt->n && !(rt->rt6i_flags & RTF_NONEXTHOP))
 925                 nrt = rt6_alloc_cow(rt, &fl6->daddr, &fl6->saddr);
 926         else if (!(rt->dst.flags & DST_HOST))
 927                 nrt = rt6_alloc_clone(rt, &fl6->daddr);
 928         else
 929                 goto out2;
 930
 931         ip6_rt_put(rt);
 932         rt = nrt ? : net->ipv6.ip6_null_entry;
 933
 934         dst_hold(&rt->dst);
 935         if (nrt) {
 936                 err = ip6_ins_rt(nrt);
 937                 if (!err)
 938                         goto out2;
 939         }
 940
 941         if (--attempts <= 0)
 942                 goto out2;
 943
 944         /*
 945          * Race condition! In the gap, when table->tb6_lock was
 946          * released someone could insert this route.  Relookup.
 947          */
 948         ip6_rt_put(rt);
 949         goto relookup;
 950
 951 out:
 952         if (reachable) {
 953                 reachable = 0;
 954                 goto restart_2;
 955         }
 956         dst_hold(&rt->dst);
 957         read_unlock_bh(&table->tb6_lock);
 958 out2:
 959         rt->dst.lastuse = jiffies;
 960         rt->dst.__use++;
 961
 962         return rt;
 963 }
 964
 965 static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
 966                                             struct flowi6 *fl6, int flags)
 967 {
 968         return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, flags);
 969 }
 970
 971 static struct dst_entry *ip6_route_input_lookup(struct net *net,
 972                                                 struct net_device *dev,
 973                                                 struct flowi6 *fl6, int flags)
 974 {
 975         if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
 976                 flags |= RT6_LOOKUP_F_IFACE;
 977
 978         return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_input);
 979 }
 980
 981 void ip6_route_input(struct sk_buff *skb)
 982 {
 983         const struct ipv6hdr *iph = ipv6_hdr(skb);
 984         struct net *net = dev_net(skb->dev);
 985         int flags = RT6_LOOKUP_F_HAS_SADDR;
 986         struct flowi6 fl6 = {
 987                 .flowi6_iif = skb->dev->ifindex,
 988                 .daddr = iph->daddr,
 989                 .saddr = iph->saddr,
 990                 .flowlabel = ip6_flowinfo(iph),
 991                 .flowi6_mark = skb->mark,
 992                 .flowi6_proto = iph->nexthdr,
 993         };
 994
 995         skb_dst_set(skb, ip6_route_input_lookup(net, skb->dev, &fl6, flags));
 996 }
 997
 998 static struct rt6_info *ip6_pol_route_output(struct net *net, struct fib6_table *table,
 999                                              struct flowi6 *fl6, int flags)
1000 {
1001         return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, flags);
1002 }
1003
1004 struct dst_entry * ip6_route_output(struct net *net, const struct sock *sk,
1005                                     struct flowi6 *fl6)
1006 {
1007         int flags = 0;
1008
1009         fl6->flowi6_iif = LOOPBACK_IFINDEX;
1010
1011         if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr))
1012                 flags |= RT6_LOOKUP_F_IFACE;
1013
1014         if (!ipv6_addr_any(&fl6->saddr))
1015                 flags |= RT6_LOOKUP_F_HAS_SADDR;
1016         else if (sk)
1017                 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
1018
1019         return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_output);
1020 }
1021
1022 EXPORT_SYMBOL(ip6_route_output);
1023
1024 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
1025 {
1026         struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
1027         struct dst_entry *new = NULL;
1028
1029         rt = dst_alloc(&ip6_dst_blackhole_ops, ort->dst.dev, 1, DST_OBSOLETE_NONE, 0);
1030         if (rt) {
1031                 new = &rt->dst;
1032
1033                 memset(new + 1, 0, sizeof(*rt) - sizeof(*new));
1034                 rt6_init_peer(rt, net->ipv6.peers);
1035
1036                 new->__use = 1;
1037                 new->input = dst_discard;
1038                 new->output = dst_discard;
1039
1040                 if (dst_metrics_read_only(&ort->dst))
1041                         new->_metrics = ort->dst._metrics;
1042                 else
1043                         dst_copy_metrics(new, &ort->dst);
1044                 rt->rt6i_idev = ort->rt6i_idev;
1045                 if (rt->rt6i_idev)
1046                         in6_dev_hold(rt->rt6i_idev);
1047
1048                 rt->rt6i_gateway = ort->rt6i_gateway;
1049                 rt->rt6i_flags = ort->rt6i_flags;
1050                 rt6_clean_expires(rt);
1051                 rt->rt6i_metric = 0;
1052
1053                 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1054 #ifdef CONFIG_IPV6_SUBTREES
1055                 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1056 #endif
1057
1058                 dst_free(new);
1059         }
1060
1061         dst_release(dst_orig);
1062         return new ? new : ERR_PTR(-ENOMEM);
1063 }
1064
1065 /*
1066  *      Destination cache support functions
1067  */
1068
1069 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
1070 {
1071         struct rt6_info *rt;
1072
1073         rt = (struct rt6_info *) dst;
1074
1075         /* All IPV6 dsts are created with ->obsolete set to the value
1076          * DST_OBSOLETE_FORCE_CHK which forces validation calls down
1077          * into this function always.
1078          */
1079         if (rt->rt6i_genid != rt_genid(dev_net(rt->dst.dev)))
1080                 return NULL;
1081
1082         if (rt->rt6i_node && (rt->rt6i_node->fn_sernum == cookie))
1083                 return dst;
1084
1085         return NULL;
1086 }
1087
1088 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
1089 {
1090         struct rt6_info *rt = (struct rt6_info *) dst;
1091
1092         if (rt) {
1093                 if (rt->rt6i_flags & RTF_CACHE) {
1094                         if (rt6_check_expired(rt)) {
1095                                 ip6_del_rt(rt);
1096                                 dst = NULL;
1097                         }
1098                 } else {
1099                         dst_release(dst);
1100                         dst = NULL;
1101                 }
1102         }
1103         return dst;
1104 }
1105
1106 static void ip6_link_failure(struct sk_buff *skb)
1107 {
1108         struct rt6_info *rt;
1109
1110         icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
1111
1112         rt = (struct rt6_info *) skb_dst(skb);
1113         if (rt) {
1114                 if (rt->rt6i_flags & RTF_CACHE)
1115                         rt6_update_expires(rt, 0);
1116                 else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT))
1117                         rt->rt6i_node->fn_sernum = -1;
1118         }
1119 }
1120
1121 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
1122                                struct sk_buff *skb, u32 mtu)
1123 {
1124         struct rt6_info *rt6 = (struct rt6_info*)dst;
1125
1126         dst_confirm(dst);
1127         if (mtu < dst_mtu(dst) && rt6->rt6i_dst.plen == 128) {
1128                 struct net *net = dev_net(dst->dev);
1129
1130                 rt6->rt6i_flags |= RTF_MODIFIED;
1131                 if (mtu < IPV6_MIN_MTU) {
1132                         u32 features = dst_metric(dst, RTAX_FEATURES);
1133                         mtu = IPV6_MIN_MTU;
1134                         features |= RTAX_FEATURE_ALLFRAG;
1135                         dst_metric_set(dst, RTAX_FEATURES, features);
1136                 }
1137                 dst_metric_set(dst, RTAX_MTU, mtu);
1138                 rt6_update_expires(rt6, net->ipv6.sysctl.ip6_rt_mtu_expires);
1139         }
1140 }
1141
1142 void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
1143                      int oif, u32 mark)
1144 {
1145         const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1146         struct dst_entry *dst;
1147         struct flowi6 fl6;
1148
1149         memset(&fl6, 0, sizeof(fl6));
1150         fl6.flowi6_oif = oif;
1151         fl6.flowi6_mark = mark;
1152         fl6.flowi6_flags = 0;
1153         fl6.daddr = iph->daddr;
1154         fl6.saddr = iph->saddr;
1155         fl6.flowlabel = ip6_flowinfo(iph);
1156
1157         dst = ip6_route_output(net, NULL, &fl6);
1158         if (!dst->error)
1159                 ip6_rt_update_pmtu(dst, NULL, skb, ntohl(mtu));
1160         dst_release(dst);
1161 }
1162 EXPORT_SYMBOL_GPL(ip6_update_pmtu);
1163
1164 void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
1165 {
1166         ip6_update_pmtu(skb, sock_net(sk), mtu,
1167                         sk->sk_bound_dev_if, sk->sk_mark);
1168 }
1169 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
1170
1171 void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark)
1172 {
1173         const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1174         struct dst_entry *dst;
1175         struct flowi6 fl6;
1176
1177         memset(&fl6, 0, sizeof(fl6));
1178         fl6.flowi6_oif = oif;
1179         fl6.flowi6_mark = mark;
1180         fl6.flowi6_flags = 0;
1181         fl6.daddr = iph->daddr;
1182         fl6.saddr = iph->saddr;
1183         fl6.flowlabel = ip6_flowinfo(iph);
1184
1185         dst = ip6_route_output(net, NULL, &fl6);
1186         if (!dst->error)
1187                 rt6_do_redirect(dst, NULL, skb);
1188         dst_release(dst);
1189 }
1190 EXPORT_SYMBOL_GPL(ip6_redirect);
1191
1192 void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
1193 {
1194         ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark);
1195 }
1196 EXPORT_SYMBOL_GPL(ip6_sk_redirect);
1197
1198 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1199 {
1200         struct net_device *dev = dst->dev;
1201         unsigned int mtu = dst_mtu(dst);
1202         struct net *net = dev_net(dev);
1203
1204         mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
1205
1206         if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
1207                 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1208
1209         /*
1210          * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1211          * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1212          * IPV6_MAXPLEN is also valid and means: "any MSS,
1213          * rely only on pmtu discovery"
1214          */
1215         if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
1216                 mtu = IPV6_MAXPLEN;
1217         return mtu;
1218 }
1219
1220 static unsigned int ip6_mtu(const struct dst_entry *dst)
1221 {
1222         struct inet6_dev *idev;
1223         unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
1224
1225         if (mtu)
1226                 return mtu;
1227
1228         mtu = IPV6_MIN_MTU;
1229
1230         rcu_read_lock();
1231         idev = __in6_dev_get(dst->dev);
1232         if (idev)
1233                 mtu = idev->cnf.mtu6;
1234         rcu_read_unlock();
1235
1236         return mtu;
1237 }
1238
1239 static struct dst_entry *icmp6_dst_gc_list;
1240 static DEFINE_SPINLOCK(icmp6_dst_lock);
1241
1242 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
1243                                   struct neighbour *neigh,
1244                                   struct flowi6 *fl6)
1245 {
1246         struct dst_entry *dst;
1247         struct rt6_info *rt;
1248         struct inet6_dev *idev = in6_dev_get(dev);
1249         struct net *net = dev_net(dev);
1250
1251         if (unlikely(!idev))
1252                 return ERR_PTR(-ENODEV);
1253
1254         rt = ip6_dst_alloc(net, dev, 0, NULL);
1255         if (unlikely(!rt)) {
1256                 in6_dev_put(idev);
1257                 dst = ERR_PTR(-ENOMEM);
1258                 goto out;
1259         }
1260
1261         if (neigh)
1262                 neigh_hold(neigh);
1263         else {
1264                 neigh = ip6_neigh_lookup(&rt->dst, NULL, &fl6->daddr);
1265                 if (IS_ERR(neigh)) {
1266                         in6_dev_put(idev);
1267                         dst_free(&rt->dst);
1268                         return ERR_CAST(neigh);
1269                 }
1270         }
1271
1272         rt->dst.flags |= DST_HOST;
1273         rt->dst.output  = ip6_output;
1274         rt->n = neigh;
1275         atomic_set(&rt->dst.__refcnt, 1);
1276         rt->rt6i_dst.addr = fl6->daddr;
1277         rt->rt6i_dst.plen = 128;
1278         rt->rt6i_idev     = idev;
1279         dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
1280
1281         spin_lock_bh(&icmp6_dst_lock);
1282         rt->dst.next = icmp6_dst_gc_list;
1283         icmp6_dst_gc_list = &rt->dst;
1284         spin_unlock_bh(&icmp6_dst_lock);
1285
1286         fib6_force_start_gc(net);
1287
1288         dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
1289
1290 out:
1291         return dst;
1292 }
1293
1294 int icmp6_dst_gc(void)
1295 {
1296         struct dst_entry *dst, **pprev;
1297         int more = 0;
1298
1299         spin_lock_bh(&icmp6_dst_lock);
1300         pprev = &icmp6_dst_gc_list;
1301
1302         while ((dst = *pprev) != NULL) {
1303                 if (!atomic_read(&dst->__refcnt)) {
1304                         *pprev = dst->next;
1305                         dst_free(dst);
1306                 } else {
1307                         pprev = &dst->next;
1308                         ++more;
1309                 }
1310         }
1311
1312         spin_unlock_bh(&icmp6_dst_lock);
1313
1314         return more;
1315 }
1316
1317 static void icmp6_clean_all(int (*func)(struct rt6_info *rt, void *arg),
1318                             void *arg)
1319 {
1320         struct dst_entry *dst, **pprev;
1321
1322         spin_lock_bh(&icmp6_dst_lock);
1323         pprev = &icmp6_dst_gc_list;
1324         while ((dst = *pprev) != NULL) {
1325                 struct rt6_info *rt = (struct rt6_info *) dst;
1326                 if (func(rt, arg)) {
1327                         *pprev = dst->next;
1328                         dst_free(dst);
1329                 } else {
1330                         pprev = &dst->next;
1331                 }
1332         }
1333         spin_unlock_bh(&icmp6_dst_lock);
1334 }
1335
1336 static int ip6_dst_gc(struct dst_ops *ops)
1337 {
1338         unsigned long now = jiffies;
1339         struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
1340         int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1341         int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1342         int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1343         int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1344         unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1345         int entries;
1346
1347         entries = dst_entries_get_fast(ops);
1348         if (time_after(rt_last_gc + rt_min_interval, now) &&
1349             entries <= rt_max_size)
1350                 goto out;
1351
1352         net->ipv6.ip6_rt_gc_expire++;
1353         fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net);
1354         net->ipv6.ip6_rt_last_gc = now;
1355         entries = dst_entries_get_slow(ops);
1356         if (entries < ops->gc_thresh)
1357                 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1358 out:
1359         net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1360         return entries > rt_max_size;
1361 }
1362
1363 int ip6_dst_hoplimit(struct dst_entry *dst)
1364 {
1365         int hoplimit = dst_metric_raw(dst, RTAX_HOPLIMIT);
1366         if (hoplimit == 0) {
1367                 struct net_device *dev = dst->dev;
1368                 struct inet6_dev *idev;
1369
1370                 rcu_read_lock();
1371                 idev = __in6_dev_get(dev);
1372                 if (idev)
1373                         hoplimit = idev->cnf.hop_limit;
1374                 else
1375                         hoplimit = dev_net(dev)->ipv6.devconf_all->hop_limit;
1376                 rcu_read_unlock();
1377         }
1378         return hoplimit;
1379 }
1380 EXPORT_SYMBOL(ip6_dst_hoplimit);
1381
1382 /*
1383  *
1384  */
1385
1386 int ip6_route_add(struct fib6_config *cfg)
1387 {
1388         int err;
1389         struct net *net = cfg->fc_nlinfo.nl_net;
1390         struct rt6_info *rt = NULL;
1391         struct net_device *dev = NULL;
1392         struct inet6_dev *idev = NULL;
1393         struct fib6_table *table;
1394         int addr_type;
1395
1396         if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1397                 return -EINVAL;
1398 #ifndef CONFIG_IPV6_SUBTREES
1399         if (cfg->fc_src_len)
1400                 return -EINVAL;
1401 #endif
1402         if (cfg->fc_ifindex) {
1403                 err = -ENODEV;
1404                 dev = dev_get_by_index(net, cfg->fc_ifindex);
1405                 if (!dev)
1406                         goto out;
1407                 idev = in6_dev_get(dev);
1408                 if (!idev)
1409                         goto out;
1410         }
1411
1412         if (cfg->fc_metric == 0)
1413                 cfg->fc_metric = IP6_RT_PRIO_USER;
1414
1415         err = -ENOBUFS;
1416         if (cfg->fc_nlinfo.nlh &&
1417             !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
1418                 table = fib6_get_table(net, cfg->fc_table);
1419                 if (!table) {
1420                         pr_warn("NLM_F_CREATE should be specified when creating new route\n");
1421                         table = fib6_new_table(net, cfg->fc_table);
1422                 }
1423         } else {
1424                 table = fib6_new_table(net, cfg->fc_table);
1425         }
1426
1427         if (!table)
1428                 goto out;
1429
1430         rt = ip6_dst_alloc(net, NULL, DST_NOCOUNT, table);
1431
1432         if (!rt) {
1433                 err = -ENOMEM;
1434                 goto out;
1435         }
1436
1437         if (cfg->fc_flags & RTF_EXPIRES)
1438                 rt6_set_expires(rt, jiffies +
1439                                 clock_t_to_jiffies(cfg->fc_expires));
1440         else
1441                 rt6_clean_expires(rt);
1442
1443         if (cfg->fc_protocol == RTPROT_UNSPEC)
1444                 cfg->fc_protocol = RTPROT_BOOT;
1445         rt->rt6i_protocol = cfg->fc_protocol;
1446
1447         addr_type = ipv6_addr_type(&cfg->fc_dst);
1448
1449         if (addr_type & IPV6_ADDR_MULTICAST)
1450                 rt->dst.input = ip6_mc_input;
1451         else if (cfg->fc_flags & RTF_LOCAL)
1452                 rt->dst.input = ip6_input;
1453         else
1454                 rt->dst.input = ip6_forward;
1455
1456         rt->dst.output = ip6_output;
1457
1458         ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1459         rt->rt6i_dst.plen = cfg->fc_dst_len;
1460         if (rt->rt6i_dst.plen == 128)
1461                rt->dst.flags |= DST_HOST;
1462
1463         if (!(rt->dst.flags & DST_HOST) && cfg->fc_mx) {
1464                 u32 *metrics = kzalloc(sizeof(u32) * RTAX_MAX, GFP_KERNEL);
1465                 if (!metrics) {
1466                         err = -ENOMEM;
1467                         goto out;
1468                 }
1469                 dst_init_metrics(&rt->dst, metrics, 0);
1470         }
1471 #ifdef CONFIG_IPV6_SUBTREES
1472         ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1473         rt->rt6i_src.plen = cfg->fc_src_len;
1474 #endif
1475
1476         rt->rt6i_metric = cfg->fc_metric;
1477
1478         /* We cannot add true routes via loopback here,
1479            they would result in kernel looping; promote them to reject routes
1480          */
1481         if ((cfg->fc_flags & RTF_REJECT) ||
1482             (dev && (dev->flags & IFF_LOOPBACK) &&
1483              !(addr_type & IPV6_ADDR_LOOPBACK) &&
1484              !(cfg->fc_flags & RTF_LOCAL))) {
1485                 /* hold loopback dev/idev if we haven't done so. */
1486                 if (dev != net->loopback_dev) {
1487                         if (dev) {
1488                                 dev_put(dev);
1489                                 in6_dev_put(idev);
1490                         }
1491                         dev = net->loopback_dev;
1492                         dev_hold(dev);
1493                         idev = in6_dev_get(dev);
1494                         if (!idev) {
1495                                 err = -ENODEV;
1496                                 goto out;
1497                         }
1498                 }
1499                 rt->dst.output = ip6_pkt_discard_out;
1500                 rt->dst.input = ip6_pkt_discard;
1501                 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1502                 switch (cfg->fc_type) {
1503                 case RTN_BLACKHOLE:
1504                         rt->dst.error = -EINVAL;
1505                         break;
1506                 case RTN_PROHIBIT:
1507                         rt->dst.error = -EACCES;
1508                         break;
1509                 case RTN_THROW:
1510                         rt->dst.error = -EAGAIN;
1511                         break;
1512                 default:
1513                         rt->dst.error = -ENETUNREACH;
1514                         break;
1515                 }
1516                 goto install_route;
1517         }
1518
1519         if (cfg->fc_flags & RTF_GATEWAY) {
1520                 const struct in6_addr *gw_addr;
1521                 int gwa_type;
1522
1523                 gw_addr = &cfg->fc_gateway;
1524                 rt->rt6i_gateway = *gw_addr;
1525                 gwa_type = ipv6_addr_type(gw_addr);
1526
1527                 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1528                         struct rt6_info *grt;
1529
1530                         /* IPv6 strictly inhibits using not link-local
1531                            addresses as nexthop address.
1532                            Otherwise, router will not able to send redirects.
1533                            It is very good, but in some (rare!) circumstances
1534                            (SIT, PtP, NBMA NOARP links) it is handy to allow
1535                            some exceptions. --ANK
1536                          */
1537                         err = -EINVAL;
1538                         if (!(gwa_type & IPV6_ADDR_UNICAST))
1539                                 goto out;
1540
1541                         grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, 1);
1542
1543                         err = -EHOSTUNREACH;
1544                         if (!grt)
1545                                 goto out;
1546                         if (dev) {
1547                                 if (dev != grt->dst.dev) {
1548                                         ip6_rt_put(grt);
1549                                         goto out;
1550                                 }
1551                         } else {
1552                                 dev = grt->dst.dev;
1553                                 idev = grt->rt6i_idev;
1554                                 dev_hold(dev);
1555                                 in6_dev_hold(grt->rt6i_idev);
1556                         }
1557                         if (!(grt->rt6i_flags & RTF_GATEWAY))
1558                                 err = 0;
1559                         ip6_rt_put(grt);
1560
1561                         if (err)
1562                                 goto out;
1563                 }
1564                 err = -EINVAL;
1565                 if (!dev || (dev->flags & IFF_LOOPBACK))
1566                         goto out;
1567         }
1568
1569         err = -ENODEV;
1570         if (!dev)
1571                 goto out;
1572
1573         if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
1574                 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
1575                         err = -EINVAL;
1576                         goto out;
1577                 }
1578                 rt->rt6i_prefsrc.addr = cfg->fc_prefsrc;
1579                 rt->rt6i_prefsrc.plen = 128;
1580         } else
1581                 rt->rt6i_prefsrc.plen = 0;
1582
1583         if (cfg->fc_flags & (RTF_GATEWAY | RTF_NONEXTHOP)) {
1584                 err = rt6_bind_neighbour(rt, dev);
1585                 if (err)
1586                         goto out;
1587         }
1588
1589         rt->rt6i_flags = cfg->fc_flags;
1590
1591 install_route:
1592         if (cfg->fc_mx) {
1593                 struct nlattr *nla;
1594                 int remaining;
1595
1596                 nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1597                         int type = nla_type(nla);
1598
1599                         if (type) {
1600                                 if (type > RTAX_MAX) {
1601                                         err = -EINVAL;
1602                                         goto out;
1603                                 }
1604
1605                                 dst_metric_set(&rt->dst, type, nla_get_u32(nla));
1606                         }
1607                 }
1608         }
1609
1610         rt->dst.dev = dev;
1611         rt->rt6i_idev = idev;
1612         rt->rt6i_table = table;
1613
1614         cfg->fc_nlinfo.nl_net = dev_net(dev);
1615
1616         return __ip6_ins_rt(rt, &cfg->fc_nlinfo);
1617
1618 out:
1619         if (dev)
1620                 dev_put(dev);
1621         if (idev)
1622                 in6_dev_put(idev);
1623         if (rt)
1624                 dst_free(&rt->dst);
1625         return err;
1626 }
1627
1628 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
1629 {
1630         int err;
1631         struct fib6_table *table;
1632         struct net *net = dev_net(rt->dst.dev);
1633
1634         if (rt == net->ipv6.ip6_null_entry) {
1635                 err = -ENOENT;
1636                 goto out;
1637         }
1638
1639         table = rt->rt6i_table;
1640         write_lock_bh(&table->tb6_lock);
1641         err = fib6_del(rt, info);
1642         write_unlock_bh(&table->tb6_lock);
1643
1644 out:
1645         ip6_rt_put(rt);
1646         return err;
1647 }
1648
1649 int ip6_del_rt(struct rt6_info *rt)
1650 {
1651         struct nl_info info = {
1652                 .nl_net = dev_net(rt->dst.dev),
1653         };
1654         return __ip6_del_rt(rt, &info);
1655 }
1656
1657 static int ip6_route_del(struct fib6_config *cfg)
1658 {
1659         struct fib6_table *table;
1660         struct fib6_node *fn;
1661         struct rt6_info *rt;
1662         int err = -ESRCH;
1663
1664         table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
1665         if (!table)
1666                 return err;
1667
1668         read_lock_bh(&table->tb6_lock);
1669
1670         fn = fib6_locate(&table->tb6_root,
1671                          &cfg->fc_dst, cfg->fc_dst_len,
1672                          &cfg->fc_src, cfg->fc_src_len);
1673
1674         if (fn) {
1675                 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1676                         if (cfg->fc_ifindex &&
1677                             (!rt->dst.dev ||
1678                              rt->dst.dev->ifindex != cfg->fc_ifindex))
1679                                 continue;
1680                         if (cfg->fc_flags & RTF_GATEWAY &&
1681                             !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
1682                                 continue;
1683                         if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
1684                                 continue;
1685                         dst_hold(&rt->dst);
1686                         read_unlock_bh(&table->tb6_lock);
1687
1688                         return __ip6_del_rt(rt, &cfg->fc_nlinfo);
1689                 }
1690         }
1691         read_unlock_bh(&table->tb6_lock);
1692
1693         return err;
1694 }
1695
1696 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
1697 {
1698         struct net *net = dev_net(skb->dev);
1699         struct netevent_redirect netevent;
1700         struct rt6_info *rt, *nrt = NULL;
1701         struct ndisc_options ndopts;
1702         struct neighbour *old_neigh;
1703         struct inet6_dev *in6_dev;
1704         struct neighbour *neigh;
1705         struct rd_msg *msg;
1706         int optlen, on_link;
1707         u8 *lladdr;
1708
1709         optlen = skb->tail - skb->transport_header;
1710         optlen -= sizeof(*msg);
1711
1712         if (optlen < 0) {
1713                 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
1714                 return;
1715         }
1716
1717         msg = (struct rd_msg *)icmp6_hdr(skb);
1718
1719         if (ipv6_addr_is_multicast(&msg->dest)) {
1720                 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
1721                 return;
1722         }
1723
1724         on_link = 0;
1725         if (ipv6_addr_equal(&msg->dest, &msg->target)) {
1726                 on_link = 1;
1727         } else if (ipv6_addr_type(&msg->target) !=
1728                    (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
1729                 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
1730                 return;
1731         }
1732
1733         in6_dev = __in6_dev_get(skb->dev);
1734         if (!in6_dev)
1735                 return;
1736         if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
1737                 return;
1738
1739         /* RFC2461 8.1:
1740          *      The IP source address of the Redirect MUST be the same as the current
1741          *      first-hop router for the specified ICMP Destination Address.
1742          */
1743
1744         if (!ndisc_parse_options(msg->opt, optlen, &ndopts)) {
1745                 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
1746                 return;
1747         }
1748
1749         lladdr = NULL;
1750         if (ndopts.nd_opts_tgt_lladdr) {
1751                 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
1752                                              skb->dev);
1753                 if (!lladdr) {
1754                         net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
1755                         return;
1756                 }
1757         }
1758
1759         rt = (struct rt6_info *) dst;
1760         if (rt == net->ipv6.ip6_null_entry) {
1761                 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
1762                 return;
1763         }
1764
1765         /* Redirect received -> path was valid.
1766          * Look, redirects are sent only in response to data packets,
1767          * so that this nexthop apparently is reachable. --ANK
1768          */
1769         dst_confirm(&rt->dst);
1770
1771         neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
1772         if (!neigh)
1773                 return;
1774
1775         /* Duplicate redirect: silently ignore. */
1776         old_neigh = rt->n;
1777         if (neigh == old_neigh)
1778                 goto out;
1779
1780         /*
1781          *      We have finally decided to accept it.
1782          */
1783
1784         neigh_update(neigh, lladdr, NUD_STALE,
1785                      NEIGH_UPDATE_F_WEAK_OVERRIDE|
1786                      NEIGH_UPDATE_F_OVERRIDE|
1787                      (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1788                                      NEIGH_UPDATE_F_ISROUTER))
1789                      );
1790
1791         nrt = ip6_rt_copy(rt, &msg->dest);
1792         if (!nrt)
1793                 goto out;
1794
1795         nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
1796         if (on_link)
1797                 nrt->rt6i_flags &= ~RTF_GATEWAY;
1798
1799         nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
1800         nrt->n = neigh_clone(neigh);
1801
1802         if (ip6_ins_rt(nrt))
1803                 goto out;
1804
1805         netevent.old = &rt->dst;
1806         netevent.old_neigh = old_neigh;
1807         netevent.new = &nrt->dst;
1808         netevent.new_neigh = neigh;
1809         netevent.daddr = &msg->dest;
1810         call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
1811
1812         if (rt->rt6i_flags & RTF_CACHE) {
1813                 rt = (struct rt6_info *) dst_clone(&rt->dst);
1814                 ip6_del_rt(rt);
1815         }
1816
1817 out:
1818         neigh_release(neigh);
1819 }
1820
1821 /*
1822  *      Misc support functions
1823  */
1824
1825 static struct rt6_info *ip6_rt_copy(struct rt6_info *ort,
1826                                     const struct in6_addr *dest)
1827 {
1828         struct net *net = dev_net(ort->dst.dev);
1829         struct rt6_info *rt = ip6_dst_alloc(net, ort->dst.dev, 0,
1830                                             ort->rt6i_table);
1831
1832         if (rt) {
1833                 rt->dst.input = ort->dst.input;
1834                 rt->dst.output = ort->dst.output;
1835                 rt->dst.flags |= DST_HOST;
1836
1837                 rt->rt6i_dst.addr = *dest;
1838                 rt->rt6i_dst.plen = 128;
1839                 dst_copy_metrics(&rt->dst, &ort->dst);
1840                 rt->dst.error = ort->dst.error;
1841                 rt->rt6i_idev = ort->rt6i_idev;
1842                 if (rt->rt6i_idev)
1843                         in6_dev_hold(rt->rt6i_idev);
1844                 rt->dst.lastuse = jiffies;
1845
1846                 rt->rt6i_gateway = ort->rt6i_gateway;
1847                 rt->rt6i_flags = ort->rt6i_flags;
1848                 if ((ort->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF)) ==
1849                     (RTF_DEFAULT | RTF_ADDRCONF))
1850                         rt6_set_from(rt, ort);
1851                 else
1852                         rt6_clean_expires(rt);
1853                 rt->rt6i_metric = 0;
1854
1855 #ifdef CONFIG_IPV6_SUBTREES
1856                 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1857 #endif
1858                 memcpy(&rt->rt6i_prefsrc, &ort->rt6i_prefsrc, sizeof(struct rt6key));
1859                 rt->rt6i_table = ort->rt6i_table;
1860         }
1861         return rt;
1862 }
1863
1864 #ifdef CONFIG_IPV6_ROUTE_INFO
1865 static struct rt6_info *rt6_get_route_info(struct net *net,
1866                                            const struct in6_addr *prefix, int prefixlen,
1867                                            const struct in6_addr *gwaddr, int ifindex)
1868 {
1869         struct fib6_node *fn;
1870         struct rt6_info *rt = NULL;
1871         struct fib6_table *table;
1872
1873         table = fib6_get_table(net, RT6_TABLE_INFO);
1874         if (!table)
1875                 return NULL;
1876
1877         read_lock_bh(&table->tb6_lock);
1878         fn = fib6_locate(&table->tb6_root, prefix ,prefixlen, NULL, 0);
1879         if (!fn)
1880                 goto out;
1881
1882         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1883                 if (rt->dst.dev->ifindex != ifindex)
1884                         continue;
1885                 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
1886                         continue;
1887                 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
1888                         continue;
1889                 dst_hold(&rt->dst);
1890                 break;
1891         }
1892 out:
1893         read_unlock_bh(&table->tb6_lock);
1894         return rt;
1895 }
1896
1897 static struct rt6_info *rt6_add_route_info(struct net *net,
1898                                            const struct in6_addr *prefix, int prefixlen,
1899                                            const struct in6_addr *gwaddr, int ifindex,
1900                                            unsigned int pref)
1901 {
1902         struct fib6_config cfg = {
1903                 .fc_table       = RT6_TABLE_INFO,
1904                 .fc_metric      = IP6_RT_PRIO_USER,
1905                 .fc_ifindex     = ifindex,
1906                 .fc_dst_len     = prefixlen,
1907                 .fc_flags       = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
1908                                   RTF_UP | RTF_PREF(pref),
1909                 .fc_nlinfo.portid = 0,
1910                 .fc_nlinfo.nlh = NULL,
1911                 .fc_nlinfo.nl_net = net,
1912         };
1913
1914         cfg.fc_dst = *prefix;
1915         cfg.fc_gateway = *gwaddr;
1916
1917         /* We should treat it as a default route if prefix length is 0. */
1918         if (!prefixlen)
1919                 cfg.fc_flags |= RTF_DEFAULT;
1920
1921         ip6_route_add(&cfg);
1922
1923         return rt6_get_route_info(net, prefix, prefixlen, gwaddr, ifindex);
1924 }
1925 #endif
1926
1927 struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
1928 {
1929         struct rt6_info *rt;
1930         struct fib6_table *table;
1931
1932         table = fib6_get_table(dev_net(dev), RT6_TABLE_DFLT);
1933         if (!table)
1934                 return NULL;
1935
1936         read_lock_bh(&table->tb6_lock);
1937         for (rt = table->tb6_root.leaf; rt; rt=rt->dst.rt6_next) {
1938                 if (dev == rt->dst.dev &&
1939                     ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
1940                     ipv6_addr_equal(&rt->rt6i_gateway, addr))
1941                         break;
1942         }
1943         if (rt)
1944                 dst_hold(&rt->dst);
1945         read_unlock_bh(&table->tb6_lock);
1946         return rt;
1947 }
1948
1949 struct rt6_info *rt6_add_dflt_router(const struct in6_addr *gwaddr,
1950                                      struct net_device *dev,
1951                                      unsigned int pref)
1952 {
1953         struct fib6_config cfg = {
1954                 .fc_table       = RT6_TABLE_DFLT,
1955                 .fc_metric      = IP6_RT_PRIO_USER,
1956                 .fc_ifindex     = dev->ifindex,
1957                 .fc_flags       = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
1958                                   RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
1959                 .fc_nlinfo.portid = 0,
1960                 .fc_nlinfo.nlh = NULL,
1961                 .fc_nlinfo.nl_net = dev_net(dev),
1962         };
1963
1964         cfg.fc_gateway = *gwaddr;
1965
1966         ip6_route_add(&cfg);
1967
1968         return rt6_get_dflt_router(gwaddr, dev);
1969 }
1970
1971 void rt6_purge_dflt_routers(struct net *net)
1972 {
1973         struct rt6_info *rt;
1974         struct fib6_table *table;
1975
1976         /* NOTE: Keep consistent with rt6_get_dflt_router */
1977         table = fib6_get_table(net, RT6_TABLE_DFLT);
1978         if (!table)
1979                 return;
1980
1981 restart:
1982         read_lock_bh(&table->tb6_lock);
1983         for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
1984                 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF)) {
1985                         dst_hold(&rt->dst);
1986                         read_unlock_bh(&table->tb6_lock);
1987                         ip6_del_rt(rt);
1988                         goto restart;
1989                 }
1990         }
1991         read_unlock_bh(&table->tb6_lock);
1992 }
1993
1994 static void rtmsg_to_fib6_config(struct net *net,
1995                                  struct in6_rtmsg *rtmsg,
1996                                  struct fib6_config *cfg)
1997 {
1998         memset(cfg, 0, sizeof(*cfg));
1999
2000         cfg->fc_table = RT6_TABLE_MAIN;
2001         cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
2002         cfg->fc_metric = rtmsg->rtmsg_metric;
2003         cfg->fc_expires = rtmsg->rtmsg_info;
2004         cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
2005         cfg->fc_src_len = rtmsg->rtmsg_src_len;
2006         cfg->fc_flags = rtmsg->rtmsg_flags;
2007
2008         cfg->fc_nlinfo.nl_net = net;
2009
2010         cfg->fc_dst = rtmsg->rtmsg_dst;
2011         cfg->fc_src = rtmsg->rtmsg_src;
2012         cfg->fc_gateway = rtmsg->rtmsg_gateway;
2013 }
2014
2015 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
2016 {
2017         struct fib6_config cfg;
2018         struct in6_rtmsg rtmsg;
2019         int err;
2020
2021         switch(cmd) {
2022         case SIOCADDRT:         /* Add a route */
2023         case SIOCDELRT:         /* Delete a route */
2024                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2025                         return -EPERM;
2026                 err = copy_from_user(&rtmsg, arg,
2027                                      sizeof(struct in6_rtmsg));
2028                 if (err)
2029                         return -EFAULT;
2030
2031                 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
2032
2033                 rtnl_lock();
2034                 switch (cmd) {
2035                 case SIOCADDRT:
2036                         err = ip6_route_add(&cfg);
2037                         break;
2038                 case SIOCDELRT:
2039                         err = ip6_route_del(&cfg);
2040                         break;
2041                 default:
2042                         err = -EINVAL;
2043                 }
2044                 rtnl_unlock();
2045
2046                 return err;
2047         }
2048
2049         return -EINVAL;
2050 }
2051
2052 /*
2053  *      Drop the packet on the floor
2054  */
2055
2056 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
2057 {
2058         int type;
2059         struct dst_entry *dst = skb_dst(skb);
2060         switch (ipstats_mib_noroutes) {
2061         case IPSTATS_MIB_INNOROUTES:
2062                 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
2063                 if (type == IPV6_ADDR_ANY) {
2064                         IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2065                                       IPSTATS_MIB_INADDRERRORS);
2066                         break;
2067                 }
2068                 /* FALLTHROUGH */
2069         case IPSTATS_MIB_OUTNOROUTES:
2070                 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2071                               ipstats_mib_noroutes);
2072                 break;
2073         }
2074         icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
2075         kfree_skb(skb);
2076         return 0;
2077 }
2078
2079 static int ip6_pkt_discard(struct sk_buff *skb)
2080 {
2081         return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
2082 }
2083
2084 static int ip6_pkt_discard_out(struct sk_buff *skb)
2085 {
2086         skb->dev = skb_dst(skb)->dev;
2087         return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
2088 }
2089
2090 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2091
2092 static int ip6_pkt_prohibit(struct sk_buff *skb)
2093 {
2094         return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
2095 }
2096
2097 static int ip6_pkt_prohibit_out(struct sk_buff *skb)
2098 {
2099         skb->dev = skb_dst(skb)->dev;
2100         return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
2101 }
2102
2103 #endif
2104
2105 /*
2106  *      Allocate a dst for local (unicast / anycast) address.
2107  */
2108
2109 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
2110                                     const struct in6_addr *addr,
2111                                     bool anycast)
2112 {
2113         struct net *net = dev_net(idev->dev);
2114         struct rt6_info *rt = ip6_dst_alloc(net, net->loopback_dev, 0, NULL);
2115         int err;
2116
2117         if (!rt) {
2118                 net_warn_ratelimited("Maximum number of routes reached, consider increasing route/max_size\n");
2119                 return ERR_PTR(-ENOMEM);
2120         }
2121
2122         in6_dev_hold(idev);
2123
2124         rt->dst.flags |= DST_HOST;
2125         rt->dst.input = ip6_input;
2126         rt->dst.output = ip6_output;
2127         rt->rt6i_idev = idev;
2128
2129         rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
2130         if (anycast)
2131                 rt->rt6i_flags |= RTF_ANYCAST;
2132         else
2133                 rt->rt6i_flags |= RTF_LOCAL;
2134         err = rt6_bind_neighbour(rt, rt->dst.dev);
2135         if (err) {
2136                 dst_free(&rt->dst);
2137                 return ERR_PTR(err);
2138         }
2139
2140         rt->rt6i_dst.addr = *addr;
2141         rt->rt6i_dst.plen = 128;
2142         rt->rt6i_table = fib6_get_table(net, RT6_TABLE_LOCAL);
2143
2144         atomic_set(&rt->dst.__refcnt, 1);
2145
2146         return rt;
2147 }
2148
2149 int ip6_route_get_saddr(struct net *net,
2150                         struct rt6_info *rt,
2151                         const struct in6_addr *daddr,
2152                         unsigned int prefs,
2153                         struct in6_addr *saddr)
2154 {
2155         struct inet6_dev *idev = ip6_dst_idev((struct dst_entry*)rt);
2156         int err = 0;
2157         if (rt->rt6i_prefsrc.plen)
2158                 *saddr = rt->rt6i_prefsrc.addr;
2159         else
2160                 err = ipv6_dev_get_saddr(net, idev ? idev->dev : NULL,
2161                                          daddr, prefs, saddr);
2162         return err;
2163 }
2164
2165 /* remove deleted ip from prefsrc entries */
2166 struct arg_dev_net_ip {
2167         struct net_device *dev;
2168         struct net *net;
2169         struct in6_addr *addr;
2170 };
2171
2172 static int fib6_remove_prefsrc(struct rt6_info *rt, void *arg)
2173 {
2174         struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
2175         struct net *net = ((struct arg_dev_net_ip *)arg)->net;
2176         struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
2177
2178         if (((void *)rt->dst.dev == dev || !dev) &&
2179             rt != net->ipv6.ip6_null_entry &&
2180             ipv6_addr_equal(addr, &rt->rt6i_prefsrc.addr)) {
2181                 /* remove prefsrc entry */
2182                 rt->rt6i_prefsrc.plen = 0;
2183         }
2184         return 0;
2185 }
2186
2187 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
2188 {
2189         struct net *net = dev_net(ifp->idev->dev);
2190         struct arg_dev_net_ip adni = {
2191                 .dev = ifp->idev->dev,
2192                 .net = net,
2193                 .addr = &ifp->addr,
2194         };
2195         fib6_clean_all(net, fib6_remove_prefsrc, 0, &adni);
2196 }
2197
2198 struct arg_dev_net {
2199         struct net_device *dev;
2200         struct net *net;
2201 };
2202
2203 static int fib6_ifdown(struct rt6_info *rt, void *arg)
2204 {
2205         const struct arg_dev_net *adn = arg;
2206         const struct net_device *dev = adn->dev;
2207
2208         if ((rt->dst.dev == dev || !dev) &&
2209             rt != adn->net->ipv6.ip6_null_entry)
2210                 return -1;
2211
2212         return 0;
2213 }
2214
2215 void rt6_ifdown(struct net *net, struct net_device *dev)
2216 {
2217         struct arg_dev_net adn = {
2218                 .dev = dev,
2219                 .net = net,
2220         };
2221
2222         fib6_clean_all(net, fib6_ifdown, 0, &adn);
2223         icmp6_clean_all(fib6_ifdown, &adn);
2224 }
2225
2226 struct rt6_mtu_change_arg {
2227         struct net_device *dev;
2228         unsigned int mtu;
2229 };
2230
2231 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2232 {
2233         struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2234         struct inet6_dev *idev;
2235
2236         /* In IPv6 pmtu discovery is not optional,
2237            so that RTAX_MTU lock cannot disable it.
2238            We still use this lock to block changes
2239            caused by addrconf/ndisc.
2240         */
2241
2242         idev = __in6_dev_get(arg->dev);
2243         if (!idev)
2244                 return 0;
2245
2246         /* For administrative MTU increase, there is no way to discover
2247            IPv6 PMTU increase, so PMTU increase should be updated here.
2248            Since RFC 1981 doesn't include administrative MTU increase
2249            update PMTU increase is a MUST. (i.e. jumbo frame)
2250          */
2251         /*
2252            If new MTU is less than route PMTU, this new MTU will be the
2253            lowest MTU in the path, update the route PMTU to reflect PMTU
2254            decreases; if new MTU is greater than route PMTU, and the
2255            old MTU is the lowest MTU in the path, update the route PMTU
2256            to reflect the increase. In this case if the other nodes' MTU
2257            also have the lowest MTU, TOO BIG MESSAGE will be lead to
2258            PMTU discouvery.
2259          */
2260         if (rt->dst.dev == arg->dev &&
2261             !dst_metric_locked(&rt->dst, RTAX_MTU) &&
2262             (dst_mtu(&rt->dst) >= arg->mtu ||
2263              (dst_mtu(&rt->dst) < arg->mtu &&
2264               dst_mtu(&rt->dst) == idev->cnf.mtu6))) {
2265                 dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
2266         }
2267         return 0;
2268 }
2269
2270 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
2271 {
2272         struct rt6_mtu_change_arg arg = {
2273                 .dev = dev,
2274                 .mtu = mtu,
2275         };
2276
2277         fib6_clean_all(dev_net(dev), rt6_mtu_change_route, 0, &arg);
2278 }
2279
2280 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2281         [RTA_GATEWAY]           = { .len = sizeof(struct in6_addr) },
2282         [RTA_OIF]               = { .type = NLA_U32 },
2283         [RTA_IIF]               = { .type = NLA_U32 },
2284         [RTA_PRIORITY]          = { .type = NLA_U32 },
2285         [RTA_METRICS]           = { .type = NLA_NESTED },
2286         [RTA_MULTIPATH]         = { .len = sizeof(struct rtnexthop) },
2287 };
2288
2289 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2290                               struct fib6_config *cfg)
2291 {
2292         struct rtmsg *rtm;
2293         struct nlattr *tb[RTA_MAX+1];
2294         int err;
2295
2296         err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2297         if (err < 0)
2298                 goto errout;
2299
2300         err = -EINVAL;
2301         rtm = nlmsg_data(nlh);
2302         memset(cfg, 0, sizeof(*cfg));
2303
2304         cfg->fc_table = rtm->rtm_table;
2305         cfg->fc_dst_len = rtm->rtm_dst_len;
2306         cfg->fc_src_len = rtm->rtm_src_len;
2307         cfg->fc_flags = RTF_UP;
2308         cfg->fc_protocol = rtm->rtm_protocol;
2309         cfg->fc_type = rtm->rtm_type;
2310
2311         if (rtm->rtm_type == RTN_UNREACHABLE ||
2312             rtm->rtm_type == RTN_BLACKHOLE ||
2313             rtm->rtm_type == RTN_PROHIBIT ||
2314             rtm->rtm_type == RTN_THROW)
2315                 cfg->fc_flags |= RTF_REJECT;
2316
2317         if (rtm->rtm_type == RTN_LOCAL)
2318                 cfg->fc_flags |= RTF_LOCAL;
2319
2320         cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid;
2321         cfg->fc_nlinfo.nlh = nlh;
2322         cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2323
2324         if (tb[RTA_GATEWAY]) {
2325                 nla_memcpy(&cfg->fc_gateway, tb[RTA_GATEWAY], 16);
2326                 cfg->fc_flags |= RTF_GATEWAY;
2327         }
2328
2329         if (tb[RTA_DST]) {
2330                 int plen = (rtm->rtm_dst_len + 7) >> 3;
2331
2332                 if (nla_len(tb[RTA_DST]) < plen)
2333                         goto errout;
2334
2335                 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2336         }
2337
2338         if (tb[RTA_SRC]) {
2339                 int plen = (rtm->rtm_src_len + 7) >> 3;
2340
2341                 if (nla_len(tb[RTA_SRC]) < plen)
2342                         goto errout;
2343
2344                 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2345         }
2346
2347         if (tb[RTA_PREFSRC])
2348                 nla_memcpy(&cfg->fc_prefsrc, tb[RTA_PREFSRC], 16);
2349
2350         if (tb[RTA_OIF])
2351                 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2352
2353         if (tb[RTA_PRIORITY])
2354                 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2355
2356         if (tb[RTA_METRICS]) {
2357                 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2358                 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2359         }
2360
2361         if (tb[RTA_TABLE])
2362                 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2363
2364         if (tb[RTA_MULTIPATH]) {
2365                 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
2366                 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
2367         }
2368
2369         err = 0;
2370 errout:
2371         return err;
2372 }
2373
2374 static int ip6_route_multipath(struct fib6_config *cfg, int add)
2375 {
2376         struct fib6_config r_cfg;
2377         struct rtnexthop *rtnh;
2378         int remaining;
2379         int attrlen;
2380         int err = 0, last_err = 0;
2381
2382 beginning:
2383         rtnh = (struct rtnexthop *)cfg->fc_mp;
2384         remaining = cfg->fc_mp_len;
2385
2386         /* Parse a Multipath Entry */
2387         while (rtnh_ok(rtnh, remaining)) {
2388                 memcpy(&r_cfg, cfg, sizeof(*cfg));
2389                 if (rtnh->rtnh_ifindex)
2390                         r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
2391
2392                 attrlen = rtnh_attrlen(rtnh);
2393                 if (attrlen > 0) {
2394                         struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
2395
2396                         nla = nla_find(attrs, attrlen, RTA_GATEWAY);
2397                         if (nla) {
2398                                 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
2399                                 r_cfg.fc_flags |= RTF_GATEWAY;
2400                         }
2401                 }
2402                 err = add ? ip6_route_add(&r_cfg) : ip6_route_del(&r_cfg);
2403                 if (err) {
2404                         last_err = err;
2405                         /* If we are trying to remove a route, do not stop the
2406                          * loop when ip6_route_del() fails (because next hop is
2407                          * already gone), we should try to remove all next hops.
2408                          */
2409                         if (add) {
2410                                 /* If add fails, we should try to delete all
2411                                  * next hops that have been already added.
2412                                  */
2413                                 add = 0;
2414                                 goto beginning;
2415                         }
2416                 }
2417                 /* Because each route is added like a single route we remove
2418                  * this flag after the first nexthop (if there is a collision,
2419                  * we have already fail to add the first nexthop:
2420                  * fib6_add_rt2node() has reject it).
2421                  */
2422                 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~NLM_F_EXCL;
2423                 rtnh = rtnh_next(rtnh, &remaining);
2424         }
2425
2426         return last_err;
2427 }
2428
2429 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
2430 {
2431         struct fib6_config cfg;
2432         int err;
2433
2434         err = rtm_to_fib6_config(skb, nlh, &cfg);
2435         if (err < 0)
2436                 return err;
2437
2438         if (cfg.fc_mp)
2439                 return ip6_route_multipath(&cfg, 0);
2440         else
2441                 return ip6_route_del(&cfg);
2442 }
2443
2444 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
2445 {
2446         struct fib6_config cfg;
2447         int err;
2448
2449         err = rtm_to_fib6_config(skb, nlh, &cfg);
2450         if (err < 0)
2451                 return err;
2452
2453         if (cfg.fc_mp)
2454                 return ip6_route_multipath(&cfg, 1);
2455         else
2456                 return ip6_route_add(&cfg);
2457 }
2458
2459 static inline size_t rt6_nlmsg_size(void)
2460 {
2461         return NLMSG_ALIGN(sizeof(struct rtmsg))
2462                + nla_total_size(16) /* RTA_SRC */
2463                + nla_total_size(16) /* RTA_DST */
2464                + nla_total_size(16) /* RTA_GATEWAY */
2465                + nla_total_size(16) /* RTA_PREFSRC */
2466                + nla_total_size(4) /* RTA_TABLE */
2467                + nla_total_size(4) /* RTA_IIF */
2468                + nla_total_size(4) /* RTA_OIF */
2469                + nla_total_size(4) /* RTA_PRIORITY */
2470                + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
2471                + nla_total_size(sizeof(struct rta_cacheinfo));
2472 }
2473
2474 static int rt6_fill_node(struct net *net,
2475                          struct sk_buff *skb, struct rt6_info *rt,
2476                          struct in6_addr *dst, struct in6_addr *src,
2477                          int iif, int type, u32 portid, u32 seq,
2478                          int prefix, int nowait, unsigned int flags)
2479 {
2480         struct rtmsg *rtm;
2481         struct nlmsghdr *nlh;
2482         long expires;
2483         u32 table;
2484         struct neighbour *n;
2485
2486         if (prefix) {   /* user wants prefix routes only */
2487                 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
2488                         /* success since this is not a prefix route */
2489                         return 1;
2490                 }
2491         }
2492
2493         nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
2494         if (!nlh)
2495                 return -EMSGSIZE;
2496
2497         rtm = nlmsg_data(nlh);
2498         rtm->rtm_family = AF_INET6;
2499         rtm->rtm_dst_len = rt->rt6i_dst.plen;
2500         rtm->rtm_src_len = rt->rt6i_src.plen;
2501         rtm->rtm_tos = 0;
2502         if (rt->rt6i_table)
2503                 table = rt->rt6i_table->tb6_id;
2504         else
2505                 table = RT6_TABLE_UNSPEC;
2506         rtm->rtm_table = table;
2507         if (nla_put_u32(skb, RTA_TABLE, table))
2508                 goto nla_put_failure;
2509         if (rt->rt6i_flags & RTF_REJECT) {
2510                 switch (rt->dst.error) {
2511                 case -EINVAL:
2512                         rtm->rtm_type = RTN_BLACKHOLE;
2513                         break;
2514                 case -EACCES:
2515                         rtm->rtm_type = RTN_PROHIBIT;
2516                         break;
2517                 case -EAGAIN:
2518                         rtm->rtm_type = RTN_THROW;
2519                         break;
2520                 default:
2521                         rtm->rtm_type = RTN_UNREACHABLE;
2522                         break;
2523                 }
2524         }
2525         else if (rt->rt6i_flags & RTF_LOCAL)
2526                 rtm->rtm_type = RTN_LOCAL;
2527         else if (rt->dst.dev && (rt->dst.dev->flags & IFF_LOOPBACK))
2528                 rtm->rtm_type = RTN_LOCAL;
2529         else
2530                 rtm->rtm_type = RTN_UNICAST;
2531         rtm->rtm_flags = 0;
2532         rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2533         rtm->rtm_protocol = rt->rt6i_protocol;
2534         if (rt->rt6i_flags & RTF_DYNAMIC)
2535                 rtm->rtm_protocol = RTPROT_REDIRECT;
2536         else if (rt->rt6i_flags & RTF_ADDRCONF) {
2537                 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ROUTEINFO))
2538                         rtm->rtm_protocol = RTPROT_RA;
2539                 else
2540                         rtm->rtm_protocol = RTPROT_KERNEL;
2541         }
2542
2543         if (rt->rt6i_flags & RTF_CACHE)
2544                 rtm->rtm_flags |= RTM_F_CLONED;
2545
2546         if (dst) {
2547                 if (nla_put(skb, RTA_DST, 16, dst))
2548                         goto nla_put_failure;
2549                 rtm->rtm_dst_len = 128;
2550         } else if (rtm->rtm_dst_len)
2551                 if (nla_put(skb, RTA_DST, 16, &rt->rt6i_dst.addr))
2552                         goto nla_put_failure;
2553 #ifdef CONFIG_IPV6_SUBTREES
2554         if (src) {
2555                 if (nla_put(skb, RTA_SRC, 16, src))
2556                         goto nla_put_failure;
2557                 rtm->rtm_src_len = 128;
2558         } else if (rtm->rtm_src_len &&
2559                    nla_put(skb, RTA_SRC, 16, &rt->rt6i_src.addr))
2560                 goto nla_put_failure;
2561 #endif
2562         if (iif) {
2563 #ifdef CONFIG_IPV6_MROUTE
2564                 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
2565                         int err = ip6mr_get_route(net, skb, rtm, nowait);
2566                         if (err <= 0) {
2567                                 if (!nowait) {
2568                                         if (err == 0)
2569                                                 return 0;
2570                                         goto nla_put_failure;
2571                                 } else {
2572                                         if (err == -EMSGSIZE)
2573                                                 goto nla_put_failure;
2574                                 }
2575                         }
2576                 } else
2577 #endif
2578                         if (nla_put_u32(skb, RTA_IIF, iif))
2579                                 goto nla_put_failure;
2580         } else if (dst) {
2581                 struct in6_addr saddr_buf;
2582                 if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0 &&
2583                     nla_put(skb, RTA_PREFSRC, 16, &saddr_buf))
2584                         goto nla_put_failure;
2585         }
2586
2587         if (rt->rt6i_prefsrc.plen) {
2588                 struct in6_addr saddr_buf;
2589                 saddr_buf = rt->rt6i_prefsrc.addr;
2590                 if (nla_put(skb, RTA_PREFSRC, 16, &saddr_buf))
2591                         goto nla_put_failure;
2592         }
2593
2594         if (rtnetlink_put_metrics(skb, dst_metrics_ptr(&rt->dst)) < 0)
2595                 goto nla_put_failure;
2596
2597         n = rt->n;
2598         if (n) {
2599                 if (nla_put(skb, RTA_GATEWAY, 16, &n->primary_key) < 0)
2600                         goto nla_put_failure;
2601         }
2602
2603         if (rt->dst.dev &&
2604             nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex))
2605                 goto nla_put_failure;
2606         if (nla_put_u32(skb, RTA_PRIORITY, rt->rt6i_metric))
2607                 goto nla_put_failure;
2608
2609         expires = (rt->rt6i_flags & RTF_EXPIRES) ? rt->dst.expires - jiffies : 0;
2610
2611         if (rtnl_put_cacheinfo(skb, &rt->dst, 0, expires, rt->dst.error) < 0)
2612                 goto nla_put_failure;
2613
2614         return nlmsg_end(skb, nlh);
2615
2616 nla_put_failure:
2617         nlmsg_cancel(skb, nlh);
2618         return -EMSGSIZE;
2619 }
2620
2621 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
2622 {
2623         struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
2624         int prefix;
2625
2626         if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
2627                 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
2628                 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
2629         } else
2630                 prefix = 0;
2631
2632         return rt6_fill_node(arg->net,
2633                      arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
2634                      NETLINK_CB(arg->cb->skb).portid, arg->cb->nlh->nlmsg_seq,
2635                      prefix, 0, NLM_F_MULTI);
2636 }
2637
2638 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr* nlh, void *arg)
2639 {
2640         struct net *net = sock_net(in_skb->sk);
2641         struct nlattr *tb[RTA_MAX+1];
2642         struct rt6_info *rt;
2643         struct sk_buff *skb;
2644         struct rtmsg *rtm;
2645         struct flowi6 fl6;
2646         int err, iif = 0, oif = 0;
2647
2648         err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2649         if (err < 0)
2650                 goto errout;
2651
2652         err = -EINVAL;
2653         memset(&fl6, 0, sizeof(fl6));
2654
2655         if (tb[RTA_SRC]) {
2656                 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
2657                         goto errout;
2658
2659                 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
2660         }
2661
2662         if (tb[RTA_DST]) {
2663                 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
2664                         goto errout;
2665
2666                 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
2667         }
2668
2669         if (tb[RTA_IIF])
2670                 iif = nla_get_u32(tb[RTA_IIF]);
2671
2672         if (tb[RTA_OIF])
2673                 oif = nla_get_u32(tb[RTA_OIF]);
2674
2675         if (iif) {
2676                 struct net_device *dev;
2677                 int flags = 0;
2678
2679                 dev = __dev_get_by_index(net, iif);
2680                 if (!dev) {
2681                         err = -ENODEV;
2682                         goto errout;
2683                 }
2684
2685                 fl6.flowi6_iif = iif;
2686
2687                 if (!ipv6_addr_any(&fl6.saddr))
2688                         flags |= RT6_LOOKUP_F_HAS_SADDR;
2689
2690                 rt = (struct rt6_info *)ip6_route_input_lookup(net, dev, &fl6,
2691                                                                flags);
2692         } else {
2693                 fl6.flowi6_oif = oif;
2694
2695                 rt = (struct rt6_info *)ip6_route_output(net, NULL, &fl6);
2696         }
2697
2698         skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
2699         if (!skb) {
2700                 ip6_rt_put(rt);
2701                 err = -ENOBUFS;
2702                 goto errout;
2703         }
2704
2705         /* Reserve room for dummy headers, this skb can pass
2706            through good chunk of routing engine.
2707          */
2708         skb_reset_mac_header(skb);
2709         skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
2710
2711         skb_dst_set(skb, &rt->dst);
2712
2713         err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
2714                             RTM_NEWROUTE, NETLINK_CB(in_skb).portid,
2715                             nlh->nlmsg_seq, 0, 0, 0);
2716         if (err < 0) {
2717                 kfree_skb(skb);
2718                 goto errout;
2719         }
2720
2721         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
2722 errout:
2723         return err;
2724 }
2725
2726 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info)
2727 {
2728         struct sk_buff *skb;
2729         struct net *net = info->nl_net;
2730         u32 seq;
2731         int err;
2732
2733         err = -ENOBUFS;
2734         seq = info->nlh ? info->nlh->nlmsg_seq : 0;
2735
2736         skb = nlmsg_new(rt6_nlmsg_size(), gfp_any());
2737         if (!skb)
2738                 goto errout;
2739
2740         err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
2741                                 event, info->portid, seq, 0, 0, 0);
2742         if (err < 0) {
2743                 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
2744                 WARN_ON(err == -EMSGSIZE);
2745                 kfree_skb(skb);
2746                 goto errout;
2747         }
2748         rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
2749                     info->nlh, gfp_any());
2750         return;
2751 errout:
2752         if (err < 0)
2753                 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
2754 }
2755
2756 static int ip6_route_dev_notify(struct notifier_block *this,
2757                                 unsigned long event, void *data)
2758 {
2759         struct net_device *dev = (struct net_device *)data;
2760         struct net *net = dev_net(dev);
2761
2762         if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
2763                 net->ipv6.ip6_null_entry->dst.dev = dev;
2764                 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
2765 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2766                 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
2767                 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
2768                 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
2769                 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
2770 #endif
2771         }
2772
2773         return NOTIFY_OK;
2774 }
2775
2776 /*
2777  *      /proc
2778  */
2779
2780 #ifdef CONFIG_PROC_FS
2781
2782 struct rt6_proc_arg
2783 {
2784         char *buffer;
2785         int offset;
2786         int length;
2787         int skip;
2788         int len;
2789 };
2790
2791 static int rt6_info_route(struct rt6_info *rt, void *p_arg)
2792 {
2793         struct seq_file *m = p_arg;
2794         struct neighbour *n;
2795
2796         seq_printf(m, "%pi6 %02x ", &rt->rt6i_dst.addr, rt->rt6i_dst.plen);
2797
2798 #ifdef CONFIG_IPV6_SUBTREES
2799         seq_printf(m, "%pi6 %02x ", &rt->rt6i_src.addr, rt->rt6i_src.plen);
2800 #else
2801         seq_puts(m, "00000000000000000000000000000000 00 ");
2802 #endif
2803         n = rt->n;
2804         if (n) {
2805                 seq_printf(m, "%pi6", n->primary_key);
2806         } else {
2807                 seq_puts(m, "00000000000000000000000000000000");
2808         }
2809         seq_printf(m, " %08x %08x %08x %08x %8s\n",
2810                    rt->rt6i_metric, atomic_read(&rt->dst.__refcnt),
2811                    rt->dst.__use, rt->rt6i_flags,
2812                    rt->dst.dev ? rt->dst.dev->name : "");
2813         return 0;
2814 }
2815
2816 static int ipv6_route_show(struct seq_file *m, void *v)
2817 {
2818         struct net *net = (struct net *)m->private;
2819         fib6_clean_all_ro(net, rt6_info_route, 0, m);
2820         return 0;
2821 }
2822
2823 static int ipv6_route_open(struct inode *inode, struct file *file)
2824 {
2825         return single_open_net(inode, file, ipv6_route_show);
2826 }
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        = single_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(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 ? ~0UL : (unsigned long)delay, net);
2880         return 0;
2881 }
2882
2883 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_net_fops_create(net, "ipv6_route", 0, &ipv6_route_proc_fops);
3073         proc_net_fops_create(net, "rt6_stats", S_IRUGO, &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         proc_net_remove(net, "ipv6_route");
3082         proc_net_remove(net, "rt6_stats");
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 }