net/ipv6/route.c

   1 /*
   2  *      Linux INET6 implementation
   3  *      FIB front-end.
   4  *
   5  *      Authors:
   6  *      Pedro Roque             <roque@di.fc.ul.pt>
   7  *
   8  *      This program is free software; you can redistribute it and/or
   9  *      modify it under the terms of the GNU General Public License
  10  *      as published by the Free Software Foundation; either version
  11  *      2 of the License, or (at your option) any later version.
  12  */
  13
  14 /*      Changes:
  15  *
  16  *      YOSHIFUJI Hideaki @USAGI
  17  *              reworked default router selection.
  18  *              - respect outgoing interface
  19  *              - select from (probably) reachable routers (i.e.
  20  *              routers in REACHABLE, STALE, DELAY or PROBE states).
  21  *              - always select the same router if it is (probably)
  22  *              reachable.  otherwise, round-robin the list.
  23  *      Ville Nuorvala
  24  *              Fixed routing subtrees.
  25  */
  26
  27 #define pr_fmt(fmt) "IPv6: " fmt
  28
  29 #include <linux/capability.h>
  30 #include <linux/errno.h>
  31 #include <linux/export.h>
  32 #include <linux/types.h>
  33 #include <linux/times.h>
  34 #include <linux/socket.h>
  35 #include <linux/sockios.h>
  36 #include <linux/net.h>
  37 #include <linux/route.h>
  38 #include <linux/netdevice.h>
  39 #include <linux/in6.h>
  40 #include <linux/mroute6.h>
  41 #include <linux/init.h>
  42 #include <linux/if_arp.h>
  43 #include <linux/proc_fs.h>
  44 #include <linux/seq_file.h>
  45 #include <linux/nsproxy.h>
  46 #include <linux/slab.h>
  47 #include <net/net_namespace.h>
  48 #include <net/snmp.h>
  49 #include <net/ipv6.h>
  50 #include <net/ip6_fib.h>
  51 #include <net/ip6_route.h>
  52 #include <net/ndisc.h>
  53 #include <net/addrconf.h>
  54 #include <net/tcp.h>
  55 #include <linux/rtnetlink.h>
  56 #include <net/dst.h>
  57 #include <net/xfrm.h>
  58 #include <net/netevent.h>
  59 #include <net/netlink.h>
  60 #include <net/nexthop.h>
  61
  62 #include <asm/uaccess.h>
  63
  64 #ifdef CONFIG_SYSCTL
  65 #include <linux/sysctl.h>
  66 #endif
  67
  68 enum rt6_nud_state {
  69         RT6_NUD_FAIL_HARD = -2,
  70         RT6_NUD_FAIL_SOFT = -1,
  71         RT6_NUD_SUCCEED = 1
  72 };
  73
  74 static struct rt6_info *ip6_rt_copy(struct rt6_info *ort,
  75                                     const struct in6_addr *dest);
  76 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie);
  77 static unsigned int      ip6_default_advmss(const struct dst_entry *dst);
  78 static unsigned int      ip6_mtu(const struct dst_entry *dst);
  79 static struct dst_entry *ip6_negative_advice(struct dst_entry *);
  80 static void             ip6_dst_destroy(struct dst_entry *);
  81 static void             ip6_dst_ifdown(struct dst_entry *,
  82                                        struct net_device *dev, int how);
  83 static int               ip6_dst_gc(struct dst_ops *ops);
  84
  85 static int              ip6_pkt_discard(struct sk_buff *skb);
  86 static int              ip6_pkt_discard_out(struct sk_buff *skb);
  87 static int              ip6_pkt_prohibit(struct sk_buff *skb);
  88 static int              ip6_pkt_prohibit_out(struct sk_buff *skb);
  89 static void             ip6_link_failure(struct sk_buff *skb);
  90 static void             ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
  91                                            struct sk_buff *skb, u32 mtu);
  92 static void             rt6_do_redirect(struct dst_entry *dst, struct sock *sk,
  93                                         struct sk_buff *skb);
  94
  95 #ifdef CONFIG_IPV6_ROUTE_INFO
  96 static struct rt6_info *rt6_add_route_info(struct net_device *dev,
  97                                            const struct in6_addr *prefix, int prefixlen,
  98                                            const struct in6_addr *gwaddr, unsigned int pref);
  99 static struct rt6_info *rt6_get_route_info(struct net_device *dev,
 100                                            const struct in6_addr *prefix, int prefixlen,
 101                                            const struct in6_addr *gwaddr);
 102 #endif
 103
 104 static u32 *ipv6_cow_metrics(struct dst_entry *dst, unsigned long old)
 105 {
 106         struct rt6_info *rt = (struct rt6_info *) dst;
 107         struct inet_peer *peer;
 108         u32 *p = NULL;
 109
 110         if (!(rt->dst.flags & DST_HOST))
 111                 return dst_cow_metrics_generic(dst, old);
 112
 113         peer = rt6_get_peer_create(rt);
 114         if (peer) {
 115                 u32 *old_p = __DST_METRICS_PTR(old);
 116                 unsigned long prev, new;
 117
 118                 p = peer->metrics;
 119                 if (inet_metrics_new(peer))
 120                         memcpy(p, old_p, sizeof(u32) * RTAX_MAX);
 121
 122                 new = (unsigned long) p;
 123                 prev = cmpxchg(&dst->_metrics, old, new);
 124
 125                 if (prev != old) {
 126                         p = __DST_METRICS_PTR(prev);
 127                         if (prev & DST_METRICS_READ_ONLY)
 128                                 p = NULL;
 129                 }
 130         }
 131         return p;
 132 }
 133
 134 static inline const void *choose_neigh_daddr(struct rt6_info *rt,
 135                                              struct sk_buff *skb,
 136                                              const void *daddr)
 137 {
 138         struct in6_addr *p = &rt->rt6i_gateway;
 139
 140         if (!ipv6_addr_any(p))
 141                 return (const void *) p;
 142         else if (skb)
 143                 return &ipv6_hdr(skb)->daddr;
 144         return daddr;
 145 }
 146
 147 static struct neighbour *ip6_neigh_lookup(const struct dst_entry *dst,
 148                                           struct sk_buff *skb,
 149                                           const void *daddr)
 150 {
 151         struct rt6_info *rt = (struct rt6_info *) dst;
 152         struct neighbour *n;
 153
 154         daddr = choose_neigh_daddr(rt, skb, daddr);
 155         n = __ipv6_neigh_lookup(dst->dev, daddr);
 156         if (n)
 157                 return n;
 158         return neigh_create(&nd_tbl, daddr, dst->dev);
 159 }
 160
 161 static struct dst_ops ip6_dst_ops_template = {
 162         .family                 =       AF_INET6,
 163         .protocol               =       cpu_to_be16(ETH_P_IPV6),
 164         .gc                     =       ip6_dst_gc,
 165         .gc_thresh              =       1024,
 166         .check                  =       ip6_dst_check,
 167         .default_advmss         =       ip6_default_advmss,
 168         .mtu                    =       ip6_mtu,
 169         .cow_metrics            =       ipv6_cow_metrics,
 170         .destroy                =       ip6_dst_destroy,
 171         .ifdown                 =       ip6_dst_ifdown,
 172         .negative_advice        =       ip6_negative_advice,
 173         .link_failure           =       ip6_link_failure,
 174         .update_pmtu            =       ip6_rt_update_pmtu,
 175         .redirect               =       rt6_do_redirect,
 176         .local_out              =       __ip6_local_out,
 177         .neigh_lookup           =       ip6_neigh_lookup,
 178 };
 179
 180 static unsigned int ip6_blackhole_mtu(const struct dst_entry *dst)
 181 {
 182         unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
 183
 184         return mtu ? : dst->dev->mtu;
 185 }
 186
 187 static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, struct sock *sk,
 188                                          struct sk_buff *skb, u32 mtu)
 189 {
 190 }
 191
 192 static void ip6_rt_blackhole_redirect(struct dst_entry *dst, struct sock *sk,
 193                                       struct sk_buff *skb)
 194 {
 195 }
 196
 197 static u32 *ip6_rt_blackhole_cow_metrics(struct dst_entry *dst,
 198                                          unsigned long old)
 199 {
 200         return NULL;
 201 }
 202
 203 static struct dst_ops ip6_dst_blackhole_ops = {
 204         .family                 =       AF_INET6,
 205         .protocol               =       cpu_to_be16(ETH_P_IPV6),
 206         .destroy                =       ip6_dst_destroy,
 207         .check                  =       ip6_dst_check,
 208         .mtu                    =       ip6_blackhole_mtu,
 209         .default_advmss         =       ip6_default_advmss,
 210         .update_pmtu            =       ip6_rt_blackhole_update_pmtu,
 211         .redirect               =       ip6_rt_blackhole_redirect,
 212         .cow_metrics            =       ip6_rt_blackhole_cow_metrics,
 213         .neigh_lookup           =       ip6_neigh_lookup,
 214 };
 215
 216 static const u32 ip6_template_metrics[RTAX_MAX] = {
 217         [RTAX_HOPLIMIT - 1] = 0,
 218 };
 219
 220 static const struct rt6_info ip6_null_entry_template = {
 221         .dst = {
 222                 .__refcnt       = ATOMIC_INIT(1),
 223                 .__use          = 1,
 224                 .obsolete       = DST_OBSOLETE_FORCE_CHK,
 225                 .error          = -ENETUNREACH,
 226                 .input          = ip6_pkt_discard,
 227                 .output         = ip6_pkt_discard_out,
 228         },
 229         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
 230         .rt6i_protocol  = RTPROT_KERNEL,
 231         .rt6i_metric    = ~(u32) 0,
 232         .rt6i_ref       = ATOMIC_INIT(1),
 233 };
 234
 235 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
 236
 237 static const struct rt6_info ip6_prohibit_entry_template = {
 238         .dst = {
 239                 .__refcnt       = ATOMIC_INIT(1),
 240                 .__use          = 1,
 241                 .obsolete       = DST_OBSOLETE_FORCE_CHK,
 242                 .error          = -EACCES,
 243                 .input          = ip6_pkt_prohibit,
 244                 .output         = ip6_pkt_prohibit_out,
 245         },
 246         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
 247         .rt6i_protocol  = RTPROT_KERNEL,
 248         .rt6i_metric    = ~(u32) 0,
 249         .rt6i_ref       = ATOMIC_INIT(1),
 250 };
 251
 252 static const struct rt6_info ip6_blk_hole_entry_template = {
 253         .dst = {
 254                 .__refcnt       = ATOMIC_INIT(1),
 255                 .__use          = 1,
 256                 .obsolete       = DST_OBSOLETE_FORCE_CHK,
 257                 .error          = -EINVAL,
 258                 .input          = dst_discard,
 259                 .output         = dst_discard,
 260         },
 261         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
 262         .rt6i_protocol  = RTPROT_KERNEL,
 263         .rt6i_metric    = ~(u32) 0,
 264         .rt6i_ref       = ATOMIC_INIT(1),
 265 };
 266
 267 #endif
 268
 269 /* allocate dst with ip6_dst_ops */
 270 static inline struct rt6_info *ip6_dst_alloc(struct net *net,
 271                                              struct net_device *dev,
 272                                              int flags,
 273                                              struct fib6_table *table)
 274 {
 275         struct rt6_info *rt = dst_alloc(&net->ipv6.ip6_dst_ops, dev,
 276                                         0, DST_OBSOLETE_FORCE_CHK, flags);
 277
 278         if (rt) {
 279                 struct dst_entry *dst = &rt->dst;
 280
 281                 memset(dst + 1, 0, sizeof(*rt) - sizeof(*dst));
 282                 rt6_init_peer(rt, table ? &table->tb6_peers : net->ipv6.peers);
 283                 rt->rt6i_genid = rt_genid(net);
 284                 INIT_LIST_HEAD(&rt->rt6i_siblings);
 285                 rt->rt6i_nsiblings = 0;
 286         }
 287         return rt;
 288 }
 289
 290 static void ip6_dst_destroy(struct dst_entry *dst)
 291 {
 292         struct rt6_info *rt = (struct rt6_info *)dst;
 293         struct inet6_dev *idev = rt->rt6i_idev;
 294         struct dst_entry *from = dst->from;
 295
 296         if (!(rt->dst.flags & DST_HOST))
 297                 dst_destroy_metrics_generic(dst);
 298
 299         if (idev) {
 300                 rt->rt6i_idev = NULL;
 301                 in6_dev_put(idev);
 302         }
 303
 304         dst->from = NULL;
 305         dst_release(from);
 306
 307         if (rt6_has_peer(rt)) {
 308                 struct inet_peer *peer = rt6_peer_ptr(rt);
 309                 inet_putpeer(peer);
 310         }
 311 }
 312
 313 void rt6_bind_peer(struct rt6_info *rt, int create)
 314 {
 315         struct inet_peer_base *base;
 316         struct inet_peer *peer;
 317
 318         base = inetpeer_base_ptr(rt->_rt6i_peer);
 319         if (!base)
 320                 return;
 321
 322         peer = inet_getpeer_v6(base, &rt->rt6i_dst.addr, create);
 323         if (peer) {
 324                 if (!rt6_set_peer(rt, peer))
 325                         inet_putpeer(peer);
 326         }
 327 }
 328
 329 static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
 330                            int how)
 331 {
 332         struct rt6_info *rt = (struct rt6_info *)dst;
 333         struct inet6_dev *idev = rt->rt6i_idev;
 334         struct net_device *loopback_dev =
 335                 dev_net(dev)->loopback_dev;
 336
 337         if (dev != loopback_dev) {
 338                 if (idev && idev->dev == dev) {
 339                         struct inet6_dev *loopback_idev =
 340                                 in6_dev_get(loopback_dev);
 341                         if (loopback_idev) {
 342                                 rt->rt6i_idev = loopback_idev;
 343                                 in6_dev_put(idev);
 344                         }
 345                 }
 346         }
 347 }
 348
 349 static bool rt6_check_expired(const struct rt6_info *rt)
 350 {
 351         if (rt->rt6i_flags & RTF_EXPIRES) {
 352                 if (time_after(jiffies, rt->dst.expires))
 353                         return true;
 354         } else if (rt->dst.from) {
 355                 return rt6_check_expired((struct rt6_info *) rt->dst.from);
 356         }
 357         return false;
 358 }
 359
 360 static bool rt6_need_strict(const struct in6_addr *daddr)
 361 {
 362         return ipv6_addr_type(daddr) &
 363                 (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
 364 }
 365
 366 /* Multipath route selection:
 367  *   Hash based function using packet header and flowlabel.
 368  * Adapted from fib_info_hashfn()
 369  */
 370 static int rt6_info_hash_nhsfn(unsigned int candidate_count,
 371                                const struct flowi6 *fl6)
 372 {
 373         unsigned int val = fl6->flowi6_proto;
 374
 375         val ^= ipv6_addr_hash(&fl6->daddr);
 376         val ^= ipv6_addr_hash(&fl6->saddr);
 377
 378         /* Work only if this not encapsulated */
 379         switch (fl6->flowi6_proto) {
 380         case IPPROTO_UDP:
 381         case IPPROTO_TCP:
 382         case IPPROTO_SCTP:
 383                 val ^= (__force u16)fl6->fl6_sport;
 384                 val ^= (__force u16)fl6->fl6_dport;
 385                 break;
 386
 387         case IPPROTO_ICMPV6:
 388                 val ^= (__force u16)fl6->fl6_icmp_type;
 389                 val ^= (__force u16)fl6->fl6_icmp_code;
 390                 break;
 391         }
 392         /* RFC6438 recommands to use flowlabel */
 393         val ^= (__force u32)fl6->flowlabel;
 394
 395         /* Perhaps, we need to tune, this function? */
 396         val = val ^ (val >> 7) ^ (val >> 12);
 397         return val % candidate_count;
 398 }
 399
 400 static struct rt6_info *rt6_multipath_select(struct rt6_info *match,
 401                                              struct flowi6 *fl6)
 402 {
 403         struct rt6_info *sibling, *next_sibling;
 404         int route_choosen;
 405
 406         route_choosen = rt6_info_hash_nhsfn(match->rt6i_nsiblings + 1, fl6);
 407         /* Don't change the route, if route_choosen == 0
 408          * (siblings does not include ourself)
 409          */
 410         if (route_choosen)
 411                 list_for_each_entry_safe(sibling, next_sibling,
 412                                 &match->rt6i_siblings, rt6i_siblings) {
 413                         route_choosen--;
 414                         if (route_choosen == 0) {
 415                                 match = sibling;
 416                                 break;
 417                         }
 418                 }
 419         return match;
 420 }
 421
 422 /*
 423  *      Route lookup. Any table->tb6_lock is implied.
 424  */
 425
 426 static inline struct rt6_info *rt6_device_match(struct net *net,
 427                                                     struct rt6_info *rt,
 428                                                     const struct in6_addr *saddr,
 429                                                     int oif,
 430                                                     int flags)
 431 {
 432         struct rt6_info *local = NULL;
 433         struct rt6_info *sprt;
 434
 435         if (!oif && ipv6_addr_any(saddr))
 436                 goto out;
 437
 438         for (sprt = rt; sprt; sprt = sprt->dst.rt6_next) {
 439                 struct net_device *dev = sprt->dst.dev;
 440
 441                 if (oif) {
 442                         if (dev->ifindex == oif)
 443                                 return sprt;
 444                         if (dev->flags & IFF_LOOPBACK) {
 445                                 if (!sprt->rt6i_idev ||
 446                                     sprt->rt6i_idev->dev->ifindex != oif) {
 447                                         if (flags & RT6_LOOKUP_F_IFACE && oif)
 448                                                 continue;
 449                                         if (local && (!oif ||
 450                                                       local->rt6i_idev->dev->ifindex == oif))
 451                                                 continue;
 452                                 }
 453                                 local = sprt;
 454                         }
 455                 } else {
 456                         if (ipv6_chk_addr(net, saddr, dev,
 457                                           flags & RT6_LOOKUP_F_IFACE))
 458                                 return sprt;
 459                 }
 460         }
 461
 462         if (oif) {
 463                 if (local)
 464                         return local;
 465
 466                 if (flags & RT6_LOOKUP_F_IFACE)
 467                         return net->ipv6.ip6_null_entry;
 468         }
 469 out:
 470         return rt;
 471 }
 472
 473 #ifdef CONFIG_IPV6_ROUTER_PREF
 474 struct __rt6_probe_work {
 475         struct work_struct work;
 476         struct in6_addr target;
 477         struct net_device *dev;
 478 };
 479
 480 static void rt6_probe_deferred(struct work_struct *w)
 481 {
 482         struct in6_addr mcaddr;
 483         struct __rt6_probe_work *work =
 484                 container_of(w, struct __rt6_probe_work, work);
 485
 486         addrconf_addr_solict_mult(&work->target, &mcaddr);
 487         ndisc_send_ns(work->dev, NULL, &work->target, &mcaddr, NULL);
 488         dev_put(work->dev);
 489         kfree(w);
 490 }
 491
 492 static void rt6_probe(struct rt6_info *rt)
 493 {
 494         struct neighbour *neigh;
 495         /*
 496          * Okay, this does not seem to be appropriate
 497          * for now, however, we need to check if it
 498          * is really so; aka Router Reachability Probing.
 499          *
 500          * Router Reachability Probe MUST be rate-limited
 501          * to no more than one per minute.
 502          */
 503         if (!rt || !(rt->rt6i_flags & RTF_GATEWAY))
 504                 return;
 505         rcu_read_lock_bh();
 506         neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
 507         if (neigh) {
 508                 write_lock(&neigh->lock);
 509                 if (neigh->nud_state & NUD_VALID)
 510                         goto out;
 511         }
 512
 513         if (!neigh ||
 514             time_after(jiffies, neigh->updated + rt->rt6i_idev->cnf.rtr_probe_interval)) {
 515                 struct __rt6_probe_work *work;
 516
 517                 work = kmalloc(sizeof(*work), GFP_ATOMIC);
 518
 519                 if (neigh && work)
 520                         neigh->updated = jiffies;
 521
 522                 if (neigh)
 523                         write_unlock(&neigh->lock);
 524
 525                 if (work) {
 526                         INIT_WORK(&work->work, rt6_probe_deferred);
 527                         work->target = rt->rt6i_gateway;
 528                         dev_hold(rt->dst.dev);
 529                         work->dev = rt->dst.dev;
 530                         schedule_work(&work->work);
 531                 }
 532         } else {
 533 out:
 534                 write_unlock(&neigh->lock);
 535         }
 536         rcu_read_unlock_bh();
 537 }
 538 #else
 539 static inline void rt6_probe(struct rt6_info *rt)
 540 {
 541 }
 542 #endif
 543
 544 /*
 545  * Default Router Selection (RFC 2461 6.3.6)
 546  */
 547 static inline int rt6_check_dev(struct rt6_info *rt, int oif)
 548 {
 549         struct net_device *dev = rt->dst.dev;
 550         if (!oif || dev->ifindex == oif)
 551                 return 2;
 552         if ((dev->flags & IFF_LOOPBACK) &&
 553             rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
 554                 return 1;
 555         return 0;
 556 }
 557
 558 static inline enum rt6_nud_state rt6_check_neigh(struct rt6_info *rt)
 559 {
 560         struct neighbour *neigh;
 561         enum rt6_nud_state ret = RT6_NUD_FAIL_HARD;
 562
 563         if (rt->rt6i_flags & RTF_NONEXTHOP ||
 564             !(rt->rt6i_flags & RTF_GATEWAY))
 565                 return RT6_NUD_SUCCEED;
 566
 567         rcu_read_lock_bh();
 568         neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
 569         if (neigh) {
 570                 read_lock(&neigh->lock);
 571                 if (neigh->nud_state & NUD_VALID)
 572                         ret = RT6_NUD_SUCCEED;
 573 #ifdef CONFIG_IPV6_ROUTER_PREF
 574                 else if (!(neigh->nud_state & NUD_FAILED))
 575                         ret = RT6_NUD_SUCCEED;
 576 #endif
 577                 read_unlock(&neigh->lock);
 578         } else {
 579                 ret = IS_ENABLED(CONFIG_IPV6_ROUTER_PREF) ?
 580                       RT6_NUD_SUCCEED : RT6_NUD_FAIL_SOFT;
 581         }
 582         rcu_read_unlock_bh();
 583
 584         return ret;
 585 }
 586
 587 static int rt6_score_route(struct rt6_info *rt, int oif,
 588                            int strict)
 589 {
 590         int m;
 591
 592         m = rt6_check_dev(rt, oif);
 593         if (!m && (strict & RT6_LOOKUP_F_IFACE))
 594                 return RT6_NUD_FAIL_HARD;
 595 #ifdef CONFIG_IPV6_ROUTER_PREF
 596         m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
 597 #endif
 598         if (strict & RT6_LOOKUP_F_REACHABLE) {
 599                 int n = rt6_check_neigh(rt);
 600                 if (n < 0)
 601                         return n;
 602         }
 603         return m;
 604 }
 605
 606 static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
 607                                    int *mpri, struct rt6_info *match,
 608                                    bool *do_rr)
 609 {
 610         int m;
 611         bool match_do_rr = false;
 612
 613         if (rt6_check_expired(rt))
 614                 goto out;
 615
 616         m = rt6_score_route(rt, oif, strict);
 617         if (m == RT6_NUD_FAIL_SOFT && !IS_ENABLED(CONFIG_IPV6_ROUTER_PREF)) {
 618                 match_do_rr = true;
 619                 m = 0; /* lowest valid score */
 620         } else if (m < 0) {
 621                 goto out;
 622         }
 623
 624         if (strict & RT6_LOOKUP_F_REACHABLE)
 625                 rt6_probe(rt);
 626
 627         if (m > *mpri) {
 628                 *do_rr = match_do_rr;
 629                 *mpri = m;
 630                 match = rt;
 631         }
 632 out:
 633         return match;
 634 }
 635
 636 static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
 637                                      struct rt6_info *rr_head,
 638                                      u32 metric, int oif, int strict,
 639                                      bool *do_rr)
 640 {
 641         struct rt6_info *rt, *match;
 642         int mpri = -1;
 643
 644         match = NULL;
 645         for (rt = rr_head; rt && rt->rt6i_metric == metric;
 646              rt = rt->dst.rt6_next)
 647                 match = find_match(rt, oif, strict, &mpri, match, do_rr);
 648         for (rt = fn->leaf; rt && rt != rr_head && rt->rt6i_metric == metric;
 649              rt = rt->dst.rt6_next)
 650                 match = find_match(rt, oif, strict, &mpri, match, do_rr);
 651
 652         return match;
 653 }
 654
 655 static struct rt6_info *rt6_select(struct fib6_node *fn, int oif, int strict)
 656 {
 657         struct rt6_info *match, *rt0;
 658         struct net *net;
 659         bool do_rr = false;
 660
 661         rt0 = fn->rr_ptr;
 662         if (!rt0)
 663                 fn->rr_ptr = rt0 = fn->leaf;
 664
 665         match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict,
 666                              &do_rr);
 667
 668         if (do_rr) {
 669                 struct rt6_info *next = rt0->dst.rt6_next;
 670
 671                 /* no entries matched; do round-robin */
 672                 if (!next || next->rt6i_metric != rt0->rt6i_metric)
 673                         next = fn->leaf;
 674
 675                 if (next != rt0)
 676                         fn->rr_ptr = next;
 677         }
 678
 679         net = dev_net(rt0->dst.dev);
 680         return match ? match : net->ipv6.ip6_null_entry;
 681 }
 682
 683 #ifdef CONFIG_IPV6_ROUTE_INFO
 684 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
 685                   const struct in6_addr *gwaddr)
 686 {
 687         struct route_info *rinfo = (struct route_info *) opt;
 688         struct in6_addr prefix_buf, *prefix;
 689         unsigned int pref;
 690         unsigned long lifetime;
 691         struct rt6_info *rt;
 692
 693         if (len < sizeof(struct route_info)) {
 694                 return -EINVAL;
 695         }
 696
 697         /* Sanity check for prefix_len and length */
 698         if (rinfo->length > 3) {
 699                 return -EINVAL;
 700         } else if (rinfo->prefix_len > 128) {
 701                 return -EINVAL;
 702         } else if (rinfo->prefix_len > 64) {
 703                 if (rinfo->length < 2) {
 704                         return -EINVAL;
 705                 }
 706         } else if (rinfo->prefix_len > 0) {
 707                 if (rinfo->length < 1) {
 708                         return -EINVAL;
 709                 }
 710         }
 711
 712         pref = rinfo->route_pref;
 713         if (pref == ICMPV6_ROUTER_PREF_INVALID)
 714                 return -EINVAL;
 715
 716         lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
 717
 718         if (rinfo->length == 3)
 719                 prefix = (struct in6_addr *)rinfo->prefix;
 720         else {
 721                 /* this function is safe */
 722                 ipv6_addr_prefix(&prefix_buf,
 723                                  (struct in6_addr *)rinfo->prefix,
 724                                  rinfo->prefix_len);
 725                 prefix = &prefix_buf;
 726         }
 727
 728         if (rinfo->prefix_len == 0)
 729                 rt = rt6_get_dflt_router(gwaddr, dev);
 730         else
 731                 rt = rt6_get_route_info(dev, prefix, rinfo->prefix_len, gwaddr);
 732
 733         if (rt && !lifetime) {
 734                 ip6_del_rt(rt);
 735                 rt = NULL;
 736         }
 737
 738         if (!rt && lifetime)
 739                 rt = rt6_add_route_info(dev, prefix, rinfo->prefix_len, gwaddr, pref);
 740         else if (rt)
 741                 rt->rt6i_flags = RTF_ROUTEINFO |
 742                                  (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
 743
 744         if (rt) {
 745                 if (!addrconf_finite_timeout(lifetime))
 746                         rt6_clean_expires(rt);
 747                 else
 748                         rt6_set_expires(rt, jiffies + HZ * lifetime);
 749
 750                 ip6_rt_put(rt);
 751         }
 752         return 0;
 753 }
 754 #endif
 755
 756 #define BACKTRACK(__net, saddr)                 \
 757 do { \
 758         if (rt == __net->ipv6.ip6_null_entry) { \
 759                 struct fib6_node *pn; \
 760                 while (1) { \
 761                         if (fn->fn_flags & RTN_TL_ROOT) \
 762                                 goto out; \
 763                         pn = fn->parent; \
 764                         if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn) \
 765                                 fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr); \
 766                         else \
 767                                 fn = pn; \
 768                         if (fn->fn_flags & RTN_RTINFO) \
 769                                 goto restart; \
 770                 } \
 771         } \
 772 } while (0)
 773
 774 static struct rt6_info *ip6_pol_route_lookup(struct net *net,
 775                                              struct fib6_table *table,
 776                                              struct flowi6 *fl6, int flags)
 777 {
 778         struct fib6_node *fn;
 779         struct rt6_info *rt;
 780
 781         read_lock_bh(&table->tb6_lock);
 782         fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
 783 restart:
 784         rt = fn->leaf;
 785         rt = rt6_device_match(net, rt, &fl6->saddr, fl6->flowi6_oif, flags);
 786         if (rt->rt6i_nsiblings && fl6->flowi6_oif == 0)
 787                 rt = rt6_multipath_select(rt, fl6);
 788         BACKTRACK(net, &fl6->saddr);
 789 out:
 790         dst_use(&rt->dst, jiffies);
 791         read_unlock_bh(&table->tb6_lock);
 792         return rt;
 793
 794 }
 795
 796 struct dst_entry * ip6_route_lookup(struct net *net, struct flowi6 *fl6,
 797                                     int flags)
 798 {
 799         return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_lookup);
 800 }
 801 EXPORT_SYMBOL_GPL(ip6_route_lookup);
 802
 803 struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
 804                             const struct in6_addr *saddr, int oif, int strict)
 805 {
 806         struct flowi6 fl6 = {
 807                 .flowi6_oif = oif,
 808                 .daddr = *daddr,
 809         };
 810         struct dst_entry *dst;
 811         int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
 812
 813         if (saddr) {
 814                 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
 815                 flags |= RT6_LOOKUP_F_HAS_SADDR;
 816         }
 817
 818         dst = fib6_rule_lookup(net, &fl6, flags, ip6_pol_route_lookup);
 819         if (dst->error == 0)
 820                 return (struct rt6_info *) dst;
 821
 822         dst_release(dst);
 823
 824         return NULL;
 825 }
 826
 827 EXPORT_SYMBOL(rt6_lookup);
 828
 829 /* ip6_ins_rt is called with FREE table->tb6_lock.
 830    It takes new route entry, the addition fails by any reason the
 831    route is freed. In any case, if caller does not hold it, it may
 832    be destroyed.
 833  */
 834
 835 static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info)
 836 {
 837         int err;
 838         struct fib6_table *table;
 839
 840         table = rt->rt6i_table;
 841         write_lock_bh(&table->tb6_lock);
 842         err = fib6_add(&table->tb6_root, rt, info);
 843         write_unlock_bh(&table->tb6_lock);
 844
 845         return err;
 846 }
 847
 848 int ip6_ins_rt(struct rt6_info *rt)
 849 {
 850         struct nl_info info = {
 851                 .nl_net = dev_net(rt->dst.dev),
 852         };
 853         return __ip6_ins_rt(rt, &info);
 854 }
 855
 856 static struct rt6_info *rt6_alloc_cow(struct rt6_info *ort,
 857                                       const struct in6_addr *daddr,
 858                                       const struct in6_addr *saddr)
 859 {
 860         struct rt6_info *rt;
 861
 862         /*
 863          *      Clone the route.
 864          */
 865
 866         rt = ip6_rt_copy(ort, daddr);
 867
 868         if (rt) {
 869                 if (!(rt->rt6i_flags & RTF_GATEWAY)) {
 870                         if (ort->rt6i_dst.plen != 128 &&
 871                             ipv6_addr_equal(&ort->rt6i_dst.addr, daddr))
 872                                 rt->rt6i_flags |= RTF_ANYCAST;
 873                 }
 874
 875                 rt->rt6i_flags |= RTF_CACHE;
 876
 877 #ifdef CONFIG_IPV6_SUBTREES
 878                 if (rt->rt6i_src.plen && saddr) {
 879                         rt->rt6i_src.addr = *saddr;
 880                         rt->rt6i_src.plen = 128;
 881                 }
 882 #endif
 883         }
 884
 885         return rt;
 886 }
 887
 888 static struct rt6_info *rt6_alloc_clone(struct rt6_info *ort,
 889                                         const struct in6_addr *daddr)
 890 {
 891         struct rt6_info *rt = ip6_rt_copy(ort, daddr);
 892
 893         if (rt)
 894                 rt->rt6i_flags |= RTF_CACHE;
 895         return rt;
 896 }
 897
 898 static struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table, int oif,
 899                                       struct flowi6 *fl6, int flags)
 900 {
 901         struct fib6_node *fn;
 902         struct rt6_info *rt, *nrt;
 903         int strict = 0;
 904         int attempts = 3;
 905         int err;
 906         int reachable = net->ipv6.devconf_all->forwarding ? 0 : RT6_LOOKUP_F_REACHABLE;
 907
 908         strict |= flags & RT6_LOOKUP_F_IFACE;
 909
 910 relookup:
 911         read_lock_bh(&table->tb6_lock);
 912
 913 restart_2:
 914         fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
 915
 916 restart:
 917         rt = rt6_select(fn, oif, strict | reachable);
 918         if (rt->rt6i_nsiblings && oif == 0)
 919                 rt = rt6_multipath_select(rt, fl6);
 920         BACKTRACK(net, &fl6->saddr);
 921         if (rt == net->ipv6.ip6_null_entry ||
 922             rt->rt6i_flags & RTF_CACHE)
 923                 goto out;
 924
 925         dst_hold(&rt->dst);
 926         read_unlock_bh(&table->tb6_lock);
 927
 928         if (!(rt->rt6i_flags & (RTF_NONEXTHOP | RTF_GATEWAY)))
 929                 nrt = rt6_alloc_cow(rt, &fl6->daddr, &fl6->saddr);
 930         else if (!(rt->dst.flags & DST_HOST))
 931                 nrt = rt6_alloc_clone(rt, &fl6->daddr);
 932         else
 933                 goto out2;
 934
 935         ip6_rt_put(rt);
 936         rt = nrt ? : net->ipv6.ip6_null_entry;
 937
 938         dst_hold(&rt->dst);
 939         if (nrt) {
 940                 err = ip6_ins_rt(nrt);
 941                 if (!err)
 942                         goto out2;
 943         }
 944
 945         if (--attempts <= 0)
 946                 goto out2;
 947
 948         /*
 949          * Race condition! In the gap, when table->tb6_lock was
 950          * released someone could insert this route.  Relookup.
 951          */
 952         ip6_rt_put(rt);
 953         goto relookup;
 954
 955 out:
 956         if (reachable) {
 957                 reachable = 0;
 958                 goto restart_2;
 959         }
 960         dst_hold(&rt->dst);
 961         read_unlock_bh(&table->tb6_lock);
 962 out2:
 963         rt->dst.lastuse = jiffies;
 964         rt->dst.__use++;
 965
 966         return rt;
 967 }
 968
 969 static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
 970                                             struct flowi6 *fl6, int flags)
 971 {
 972         return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, flags);
 973 }
 974
 975 static struct dst_entry *ip6_route_input_lookup(struct net *net,
 976                                                 struct net_device *dev,
 977                                                 struct flowi6 *fl6, int flags)
 978 {
 979         if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
 980                 flags |= RT6_LOOKUP_F_IFACE;
 981
 982         return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_input);
 983 }
 984
 985 void ip6_route_input(struct sk_buff *skb)
 986 {
 987         const struct ipv6hdr *iph = ipv6_hdr(skb);
 988         struct net *net = dev_net(skb->dev);
 989         int flags = RT6_LOOKUP_F_HAS_SADDR;
 990         struct flowi6 fl6 = {
 991                 .flowi6_iif = skb->dev->ifindex,
 992                 .daddr = iph->daddr,
 993                 .saddr = iph->saddr,
 994                 .flowlabel = ip6_flowinfo(iph),
 995                 .flowi6_mark = skb->mark,
 996                 .flowi6_proto = iph->nexthdr,
 997         };
 998
 999         skb_dst_set(skb, ip6_route_input_lookup(net, skb->dev, &fl6, flags));
1000 }
1001
1002 static struct rt6_info *ip6_pol_route_output(struct net *net, struct fib6_table *table,
1003                                              struct flowi6 *fl6, int flags)
1004 {
1005         return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, flags);
1006 }
1007
1008 struct dst_entry * ip6_route_output(struct net *net, const struct sock *sk,
1009                                     struct flowi6 *fl6)
1010 {
1011         int flags = 0;
1012
1013         fl6->flowi6_iif = LOOPBACK_IFINDEX;
1014
1015         if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr))
1016                 flags |= RT6_LOOKUP_F_IFACE;
1017
1018         if (!ipv6_addr_any(&fl6->saddr))
1019                 flags |= RT6_LOOKUP_F_HAS_SADDR;
1020         else if (sk)
1021                 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
1022
1023         return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_output);
1024 }
1025
1026 EXPORT_SYMBOL(ip6_route_output);
1027
1028 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
1029 {
1030         struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
1031         struct dst_entry *new = NULL;
1032
1033         rt = dst_alloc(&ip6_dst_blackhole_ops, ort->dst.dev, 1, DST_OBSOLETE_NONE, 0);
1034         if (rt) {
1035                 new = &rt->dst;
1036
1037                 memset(new + 1, 0, sizeof(*rt) - sizeof(*new));
1038                 rt6_init_peer(rt, net->ipv6.peers);
1039
1040                 new->__use = 1;
1041                 new->input = dst_discard;
1042                 new->output = dst_discard;
1043
1044                 if (dst_metrics_read_only(&ort->dst))
1045                         new->_metrics = ort->dst._metrics;
1046                 else
1047                         dst_copy_metrics(new, &ort->dst);
1048                 rt->rt6i_idev = ort->rt6i_idev;
1049                 if (rt->rt6i_idev)
1050                         in6_dev_hold(rt->rt6i_idev);
1051
1052                 rt->rt6i_gateway = ort->rt6i_gateway;
1053                 rt->rt6i_flags = ort->rt6i_flags;
1054                 rt->rt6i_metric = 0;
1055
1056                 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1057 #ifdef CONFIG_IPV6_SUBTREES
1058                 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1059 #endif
1060
1061                 dst_free(new);
1062         }
1063
1064         dst_release(dst_orig);
1065         return new ? new : ERR_PTR(-ENOMEM);
1066 }
1067
1068 /*
1069  *      Destination cache support functions
1070  */
1071
1072 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
1073 {
1074         struct rt6_info *rt;
1075
1076         rt = (struct rt6_info *) dst;
1077
1078         /* All IPV6 dsts are created with ->obsolete set to the value
1079          * DST_OBSOLETE_FORCE_CHK which forces validation calls down
1080          * into this function always.
1081          */
1082         if (rt->rt6i_genid != rt_genid(dev_net(rt->dst.dev)))
1083                 return NULL;
1084
1085         if (!rt->rt6i_node || (rt->rt6i_node->fn_sernum != cookie))
1086                 return NULL;
1087
1088         if (rt6_check_expired(rt))
1089                 return NULL;
1090
1091         return dst;
1092 }
1093
1094 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
1095 {
1096         struct rt6_info *rt = (struct rt6_info *) dst;
1097
1098         if (rt) {
1099                 if (rt->rt6i_flags & RTF_CACHE) {
1100                         if (rt6_check_expired(rt)) {
1101                                 ip6_del_rt(rt);
1102                                 dst = NULL;
1103                         }
1104                 } else {
1105                         dst_release(dst);
1106                         dst = NULL;
1107                 }
1108         }
1109         return dst;
1110 }
1111
1112 static void ip6_link_failure(struct sk_buff *skb)
1113 {
1114         struct rt6_info *rt;
1115
1116         icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
1117
1118         rt = (struct rt6_info *) skb_dst(skb);
1119         if (rt) {
1120                 if (rt->rt6i_flags & RTF_CACHE) {
1121                         dst_hold(&rt->dst);
1122                         if (ip6_del_rt(rt))
1123                                 dst_free(&rt->dst);
1124                 } else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT)) {
1125                         rt->rt6i_node->fn_sernum = -1;
1126                 }
1127         }
1128 }
1129
1130 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
1131                                struct sk_buff *skb, u32 mtu)
1132 {
1133         struct rt6_info *rt6 = (struct rt6_info*)dst;
1134
1135         dst_confirm(dst);
1136         if (mtu < dst_mtu(dst) && rt6->rt6i_dst.plen == 128) {
1137                 struct net *net = dev_net(dst->dev);
1138
1139                 rt6->rt6i_flags |= RTF_MODIFIED;
1140                 if (mtu < IPV6_MIN_MTU)
1141                         mtu = IPV6_MIN_MTU;
1142
1143                 dst_metric_set(dst, RTAX_MTU, mtu);
1144                 rt6_update_expires(rt6, net->ipv6.sysctl.ip6_rt_mtu_expires);
1145         }
1146 }
1147
1148 void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
1149                      int oif, u32 mark, kuid_t uid)
1150 {
1151         const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1152         struct dst_entry *dst;
1153         struct flowi6 fl6;
1154
1155         memset(&fl6, 0, sizeof(fl6));
1156         fl6.flowi6_oif = oif;
1157         fl6.flowi6_mark = mark ? mark : IP6_REPLY_MARK(net, skb->mark);
1158         fl6.flowi6_flags = 0;
1159         fl6.daddr = iph->daddr;
1160         fl6.saddr = iph->saddr;
1161         fl6.flowlabel = ip6_flowinfo(iph);
1162         fl6.flowi6_uid = uid;
1163
1164         dst = ip6_route_output(net, NULL, &fl6);
1165         if (!dst->error)
1166                 ip6_rt_update_pmtu(dst, NULL, skb, ntohl(mtu));
1167         dst_release(dst);
1168 }
1169 EXPORT_SYMBOL_GPL(ip6_update_pmtu);
1170
1171 void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
1172 {
1173         ip6_update_pmtu(skb, sock_net(sk), mtu,
1174                         sk->sk_bound_dev_if, sk->sk_mark, sock_i_uid(sk));
1175 }
1176 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
1177
1178 void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark)
1179 {
1180         const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1181         struct dst_entry *dst;
1182         struct flowi6 fl6;
1183
1184         memset(&fl6, 0, sizeof(fl6));
1185         fl6.flowi6_oif = oif;
1186         fl6.flowi6_mark = mark;
1187         fl6.flowi6_flags = 0;
1188         fl6.daddr = iph->daddr;
1189         fl6.saddr = iph->saddr;
1190         fl6.flowlabel = ip6_flowinfo(iph);
1191
1192         dst = ip6_route_output(net, NULL, &fl6);
1193         if (!dst->error)
1194                 rt6_do_redirect(dst, NULL, skb);
1195         dst_release(dst);
1196 }
1197 EXPORT_SYMBOL_GPL(ip6_redirect);
1198
1199 void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
1200 {
1201         ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark);
1202 }
1203 EXPORT_SYMBOL_GPL(ip6_sk_redirect);
1204
1205 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1206 {
1207         struct net_device *dev = dst->dev;
1208         unsigned int mtu = dst_mtu(dst);
1209         struct net *net = dev_net(dev);
1210
1211         mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
1212
1213         if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
1214                 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1215
1216         /*
1217          * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1218          * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1219          * IPV6_MAXPLEN is also valid and means: "any MSS,
1220          * rely only on pmtu discovery"
1221          */
1222         if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
1223                 mtu = IPV6_MAXPLEN;
1224         return mtu;
1225 }
1226
1227 static unsigned int ip6_mtu(const struct dst_entry *dst)
1228 {
1229         struct inet6_dev *idev;
1230         unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
1231
1232         if (mtu)
1233                 goto out;
1234
1235         mtu = IPV6_MIN_MTU;
1236
1237         rcu_read_lock();
1238         idev = __in6_dev_get(dst->dev);
1239         if (idev)
1240                 mtu = idev->cnf.mtu6;
1241         rcu_read_unlock();
1242
1243 out:
1244         return min_t(unsigned int, mtu, IP6_MAX_MTU);
1245 }
1246
1247 static struct dst_entry *icmp6_dst_gc_list;
1248 static DEFINE_SPINLOCK(icmp6_dst_lock);
1249
1250 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
1251                                   struct flowi6 *fl6)
1252 {
1253         struct dst_entry *dst;
1254         struct rt6_info *rt;
1255         struct inet6_dev *idev = in6_dev_get(dev);
1256         struct net *net = dev_net(dev);
1257
1258         if (unlikely(!idev))
1259                 return ERR_PTR(-ENODEV);
1260
1261         rt = ip6_dst_alloc(net, dev, 0, NULL);
1262         if (unlikely(!rt)) {
1263                 in6_dev_put(idev);
1264                 dst = ERR_PTR(-ENOMEM);
1265                 goto out;
1266         }
1267
1268         rt->dst.flags |= DST_HOST;
1269         rt->dst.output  = ip6_output;
1270         atomic_set(&rt->dst.__refcnt, 1);
1271         rt->rt6i_gateway  = fl6->daddr;
1272         rt->rt6i_dst.addr = fl6->daddr;
1273         rt->rt6i_dst.plen = 128;
1274         rt->rt6i_idev     = idev;
1275         dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
1276
1277         spin_lock_bh(&icmp6_dst_lock);
1278         rt->dst.next = icmp6_dst_gc_list;
1279         icmp6_dst_gc_list = &rt->dst;
1280         spin_unlock_bh(&icmp6_dst_lock);
1281
1282         fib6_force_start_gc(net);
1283
1284         dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
1285
1286 out:
1287         return dst;
1288 }
1289
1290 int icmp6_dst_gc(void)
1291 {
1292         struct dst_entry *dst, **pprev;
1293         int more = 0;
1294
1295         spin_lock_bh(&icmp6_dst_lock);
1296         pprev = &icmp6_dst_gc_list;
1297
1298         while ((dst = *pprev) != NULL) {
1299                 if (!atomic_read(&dst->__refcnt)) {
1300                         *pprev = dst->next;
1301                         dst_free(dst);
1302                 } else {
1303                         pprev = &dst->next;
1304                         ++more;
1305                 }
1306         }
1307
1308         spin_unlock_bh(&icmp6_dst_lock);
1309
1310         return more;
1311 }
1312
1313 static void icmp6_clean_all(int (*func)(struct rt6_info *rt, void *arg),
1314                             void *arg)
1315 {
1316         struct dst_entry *dst, **pprev;
1317
1318         spin_lock_bh(&icmp6_dst_lock);
1319         pprev = &icmp6_dst_gc_list;
1320         while ((dst = *pprev) != NULL) {
1321                 struct rt6_info *rt = (struct rt6_info *) dst;
1322                 if (func(rt, arg)) {
1323                         *pprev = dst->next;
1324                         dst_free(dst);
1325                 } else {
1326                         pprev = &dst->next;
1327                 }
1328         }
1329         spin_unlock_bh(&icmp6_dst_lock);
1330 }
1331
1332 static int ip6_dst_gc(struct dst_ops *ops)
1333 {
1334         unsigned long now = jiffies;
1335         struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
1336         int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1337         int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1338         int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1339         int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1340         unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1341         int entries;
1342
1343         entries = dst_entries_get_fast(ops);
1344         if (time_after(rt_last_gc + rt_min_interval, now) &&
1345             entries <= rt_max_size)
1346                 goto out;
1347
1348         net->ipv6.ip6_rt_gc_expire++;
1349         fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net);
1350         net->ipv6.ip6_rt_last_gc = now;
1351         entries = dst_entries_get_slow(ops);
1352         if (entries < ops->gc_thresh)
1353                 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1354 out:
1355         net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1356         return entries > rt_max_size;
1357 }
1358
1359 int ip6_dst_hoplimit(struct dst_entry *dst)
1360 {
1361         int hoplimit = dst_metric_raw(dst, RTAX_HOPLIMIT);
1362         if (hoplimit == 0) {
1363                 struct net_device *dev = dst->dev;
1364                 struct inet6_dev *idev;
1365
1366                 rcu_read_lock();
1367                 idev = __in6_dev_get(dev);
1368                 if (idev)
1369                         hoplimit = idev->cnf.hop_limit;
1370                 else
1371                         hoplimit = dev_net(dev)->ipv6.devconf_all->hop_limit;
1372                 rcu_read_unlock();
1373         }
1374         return hoplimit;
1375 }
1376 EXPORT_SYMBOL(ip6_dst_hoplimit);
1377
1378 /*
1379  *
1380  */
1381
1382 int ip6_route_add(struct fib6_config *cfg)
1383 {
1384         int err;
1385         struct net *net = cfg->fc_nlinfo.nl_net;
1386         struct rt6_info *rt = NULL;
1387         struct net_device *dev = NULL;
1388         struct inet6_dev *idev = NULL;
1389         struct fib6_table *table;
1390         int addr_type;
1391
1392         if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1393                 return -EINVAL;
1394 #ifndef CONFIG_IPV6_SUBTREES
1395         if (cfg->fc_src_len)
1396                 return -EINVAL;
1397 #endif
1398         if (cfg->fc_ifindex) {
1399                 err = -ENODEV;
1400                 dev = dev_get_by_index(net, cfg->fc_ifindex);
1401                 if (!dev)
1402                         goto out;
1403                 idev = in6_dev_get(dev);
1404                 if (!idev)
1405                         goto out;
1406         }
1407
1408         if (cfg->fc_metric == 0)
1409                 cfg->fc_metric = IP6_RT_PRIO_USER;
1410
1411         err = -ENOBUFS;
1412         if (cfg->fc_nlinfo.nlh &&
1413             !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
1414                 table = fib6_get_table(net, cfg->fc_table);
1415                 if (!table) {
1416                         pr_warn("NLM_F_CREATE should be specified when creating new route\n");
1417                         table = fib6_new_table(net, cfg->fc_table);
1418                 }
1419         } else {
1420                 table = fib6_new_table(net, cfg->fc_table);
1421         }
1422
1423         if (!table)
1424                 goto out;
1425
1426         rt = ip6_dst_alloc(net, NULL, (cfg->fc_flags & RTF_ADDRCONF) ? 0 : DST_NOCOUNT, table);
1427
1428         if (!rt) {
1429                 err = -ENOMEM;
1430                 goto out;
1431         }
1432
1433         if (cfg->fc_flags & RTF_EXPIRES)
1434                 rt6_set_expires(rt, jiffies +
1435                                 clock_t_to_jiffies(cfg->fc_expires));
1436         else
1437                 rt6_clean_expires(rt);
1438
1439         if (cfg->fc_protocol == RTPROT_UNSPEC)
1440                 cfg->fc_protocol = RTPROT_BOOT;
1441         rt->rt6i_protocol = cfg->fc_protocol;
1442
1443         addr_type = ipv6_addr_type(&cfg->fc_dst);
1444
1445         if (addr_type & IPV6_ADDR_MULTICAST)
1446                 rt->dst.input = ip6_mc_input;
1447         else if (cfg->fc_flags & RTF_LOCAL)
1448                 rt->dst.input = ip6_input;
1449         else
1450                 rt->dst.input = ip6_forward;
1451
1452         rt->dst.output = ip6_output;
1453
1454         ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1455         rt->rt6i_dst.plen = cfg->fc_dst_len;
1456         if (rt->rt6i_dst.plen == 128)
1457                rt->dst.flags |= DST_HOST;
1458
1459         if (!(rt->dst.flags & DST_HOST) && cfg->fc_mx) {
1460                 u32 *metrics = kzalloc(sizeof(u32) * RTAX_MAX, GFP_KERNEL);
1461                 if (!metrics) {
1462                         err = -ENOMEM;
1463                         goto out;
1464                 }
1465                 dst_init_metrics(&rt->dst, metrics, 0);
1466         }
1467 #ifdef CONFIG_IPV6_SUBTREES
1468         ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1469         rt->rt6i_src.plen = cfg->fc_src_len;
1470 #endif
1471
1472         rt->rt6i_metric = cfg->fc_metric;
1473
1474         /* We cannot add true routes via loopback here,
1475            they would result in kernel looping; promote them to reject routes
1476          */
1477         if ((cfg->fc_flags & RTF_REJECT) ||
1478             (dev && (dev->flags & IFF_LOOPBACK) &&
1479              !(addr_type & IPV6_ADDR_LOOPBACK) &&
1480              !(cfg->fc_flags & RTF_LOCAL))) {
1481                 /* hold loopback dev/idev if we haven't done so. */
1482                 if (dev != net->loopback_dev) {
1483                         if (dev) {
1484                                 dev_put(dev);
1485                                 in6_dev_put(idev);
1486                         }
1487                         dev = net->loopback_dev;
1488                         dev_hold(dev);
1489                         idev = in6_dev_get(dev);
1490                         if (!idev) {
1491                                 err = -ENODEV;
1492                                 goto out;
1493                         }
1494                 }
1495                 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1496                 switch (cfg->fc_type) {
1497                 case RTN_BLACKHOLE:
1498                         rt->dst.error = -EINVAL;
1499                         rt->dst.output = dst_discard;
1500                         rt->dst.input = dst_discard;
1501                         break;
1502                 case RTN_PROHIBIT:
1503                         rt->dst.error = -EACCES;
1504                         rt->dst.output = ip6_pkt_prohibit_out;
1505                         rt->dst.input = ip6_pkt_prohibit;
1506                         break;
1507                 case RTN_THROW:
1508                 default:
1509                         rt->dst.error = (cfg->fc_type == RTN_THROW) ? -EAGAIN
1510                                         : -ENETUNREACH;
1511                         rt->dst.output = ip6_pkt_discard_out;
1512                         rt->dst.input = ip6_pkt_discard;
1513                         break;
1514                 }
1515                 goto install_route;
1516         }
1517
1518         if (cfg->fc_flags & RTF_GATEWAY) {
1519                 const struct in6_addr *gw_addr;
1520                 int gwa_type;
1521
1522                 gw_addr = &cfg->fc_gateway;
1523                 rt->rt6i_gateway = *gw_addr;
1524                 gwa_type = ipv6_addr_type(gw_addr);
1525
1526                 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1527                         struct rt6_info *grt;
1528
1529                         /* IPv6 strictly inhibits using not link-local
1530                            addresses as nexthop address.
1531                            Otherwise, router will not able to send redirects.
1532                            It is very good, but in some (rare!) circumstances
1533                            (SIT, PtP, NBMA NOARP links) it is handy to allow
1534                            some exceptions. --ANK
1535                          */
1536                         err = -EINVAL;
1537                         if (!(gwa_type & IPV6_ADDR_UNICAST))
1538                                 goto out;
1539
1540                         grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, 1);
1541
1542                         err = -EHOSTUNREACH;
1543                         if (!grt)
1544                                 goto out;
1545                         if (dev) {
1546                                 if (dev != grt->dst.dev) {
1547                                         ip6_rt_put(grt);
1548                                         goto out;
1549                                 }
1550                         } else {
1551                                 dev = grt->dst.dev;
1552                                 idev = grt->rt6i_idev;
1553                                 dev_hold(dev);
1554                                 in6_dev_hold(grt->rt6i_idev);
1555                         }
1556                         if (!(grt->rt6i_flags & RTF_GATEWAY))
1557                                 err = 0;
1558                         ip6_rt_put(grt);
1559
1560                         if (err)
1561                                 goto out;
1562                 }
1563                 err = -EINVAL;
1564                 if (!dev || (dev->flags & IFF_LOOPBACK))
1565                         goto out;
1566         }
1567
1568         err = -ENODEV;
1569         if (!dev)
1570                 goto out;
1571
1572         if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
1573                 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
1574                         err = -EINVAL;
1575                         goto out;
1576                 }
1577                 rt->rt6i_prefsrc.addr = cfg->fc_prefsrc;
1578                 rt->rt6i_prefsrc.plen = 128;
1579         } else
1580                 rt->rt6i_prefsrc.plen = 0;
1581
1582         rt->rt6i_flags = cfg->fc_flags;
1583
1584 install_route:
1585         if (cfg->fc_mx) {
1586                 struct nlattr *nla;
1587                 int remaining;
1588
1589                 nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1590                         int type = nla_type(nla);
1591
1592                         if (type) {
1593                                 if (type > RTAX_MAX) {
1594                                         err = -EINVAL;
1595                                         goto out;
1596                                 }
1597
1598                                 dst_metric_set(&rt->dst, type, nla_get_u32(nla));
1599                         }
1600                 }
1601         }
1602
1603         rt->dst.dev = dev;
1604         rt->rt6i_idev = idev;
1605         rt->rt6i_table = table;
1606
1607         cfg->fc_nlinfo.nl_net = dev_net(dev);
1608
1609         return __ip6_ins_rt(rt, &cfg->fc_nlinfo);
1610
1611 out:
1612         if (dev)
1613                 dev_put(dev);
1614         if (idev)
1615                 in6_dev_put(idev);
1616         if (rt)
1617                 dst_free(&rt->dst);
1618         return err;
1619 }
1620
1621 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
1622 {
1623         int err;
1624         struct fib6_table *table;
1625         struct net *net = dev_net(rt->dst.dev);
1626
1627         if (rt == net->ipv6.ip6_null_entry) {
1628                 err = -ENOENT;
1629                 goto out;
1630         }
1631
1632         table = rt->rt6i_table;
1633         write_lock_bh(&table->tb6_lock);
1634         err = fib6_del(rt, info);
1635         write_unlock_bh(&table->tb6_lock);
1636
1637 out:
1638         ip6_rt_put(rt);
1639         return err;
1640 }
1641
1642 int ip6_del_rt(struct rt6_info *rt)
1643 {
1644         struct nl_info info = {
1645                 .nl_net = dev_net(rt->dst.dev),
1646         };
1647         return __ip6_del_rt(rt, &info);
1648 }
1649
1650 static int ip6_route_del(struct fib6_config *cfg)
1651 {
1652         struct fib6_table *table;
1653         struct fib6_node *fn;
1654         struct rt6_info *rt;
1655         int err = -ESRCH;
1656
1657         table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
1658         if (!table)
1659                 return err;
1660
1661         read_lock_bh(&table->tb6_lock);
1662
1663         fn = fib6_locate(&table->tb6_root,
1664                          &cfg->fc_dst, cfg->fc_dst_len,
1665                          &cfg->fc_src, cfg->fc_src_len);
1666
1667         if (fn) {
1668                 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1669                         if (cfg->fc_ifindex &&
1670                             (!rt->dst.dev ||
1671                              rt->dst.dev->ifindex != cfg->fc_ifindex))
1672                                 continue;
1673                         if (cfg->fc_flags & RTF_GATEWAY &&
1674                             !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
1675                                 continue;
1676                         if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
1677                                 continue;
1678                         dst_hold(&rt->dst);
1679                         read_unlock_bh(&table->tb6_lock);
1680
1681                         return __ip6_del_rt(rt, &cfg->fc_nlinfo);
1682                 }
1683         }
1684         read_unlock_bh(&table->tb6_lock);
1685
1686         return err;
1687 }
1688
1689 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
1690 {
1691         struct net *net = dev_net(skb->dev);
1692         struct netevent_redirect netevent;
1693         struct rt6_info *rt, *nrt = NULL;
1694         struct ndisc_options ndopts;
1695         struct inet6_dev *in6_dev;
1696         struct neighbour *neigh;
1697         struct rd_msg *msg;
1698         int optlen, on_link;
1699         u8 *lladdr;
1700
1701         optlen = skb->tail - skb->transport_header;
1702         optlen -= sizeof(*msg);
1703
1704         if (optlen < 0) {
1705                 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
1706                 return;
1707         }
1708
1709         msg = (struct rd_msg *)icmp6_hdr(skb);
1710
1711         if (ipv6_addr_is_multicast(&msg->dest)) {
1712                 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
1713                 return;
1714         }
1715
1716         on_link = 0;
1717         if (ipv6_addr_equal(&msg->dest, &msg->target)) {
1718                 on_link = 1;
1719         } else if (ipv6_addr_type(&msg->target) !=
1720                    (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
1721                 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
1722                 return;
1723         }
1724
1725         in6_dev = __in6_dev_get(skb->dev);
1726         if (!in6_dev)
1727                 return;
1728         if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
1729                 return;
1730
1731         /* RFC2461 8.1:
1732          *      The IP source address of the Redirect MUST be the same as the current
1733          *      first-hop router for the specified ICMP Destination Address.
1734          */
1735
1736         if (!ndisc_parse_options(msg->opt, optlen, &ndopts)) {
1737                 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
1738                 return;
1739         }
1740
1741         lladdr = NULL;
1742         if (ndopts.nd_opts_tgt_lladdr) {
1743                 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
1744                                              skb->dev);
1745                 if (!lladdr) {
1746                         net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
1747                         return;
1748                 }
1749         }
1750
1751         rt = (struct rt6_info *) dst;
1752         if (rt == net->ipv6.ip6_null_entry) {
1753                 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
1754                 return;
1755         }
1756
1757         /* Redirect received -> path was valid.
1758          * Look, redirects are sent only in response to data packets,
1759          * so that this nexthop apparently is reachable. --ANK
1760          */
1761         dst_confirm(&rt->dst);
1762
1763         neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
1764         if (!neigh)
1765                 return;
1766
1767         /*
1768          *      We have finally decided to accept it.
1769          */
1770
1771         neigh_update(neigh, lladdr, NUD_STALE,
1772                      NEIGH_UPDATE_F_WEAK_OVERRIDE|
1773                      NEIGH_UPDATE_F_OVERRIDE|
1774                      (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1775                                      NEIGH_UPDATE_F_ISROUTER))
1776                      );
1777
1778         nrt = ip6_rt_copy(rt, &msg->dest);
1779         if (!nrt)
1780                 goto out;
1781
1782         nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
1783         if (on_link)
1784                 nrt->rt6i_flags &= ~RTF_GATEWAY;
1785
1786         nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
1787
1788         if (ip6_ins_rt(nrt))
1789                 goto out;
1790
1791         netevent.old = &rt->dst;
1792         netevent.new = &nrt->dst;
1793         netevent.daddr = &msg->dest;
1794         netevent.neigh = neigh;
1795         call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
1796
1797         if (rt->rt6i_flags & RTF_CACHE) {
1798                 rt = (struct rt6_info *) dst_clone(&rt->dst);
1799                 ip6_del_rt(rt);
1800         }
1801
1802 out:
1803         neigh_release(neigh);
1804 }
1805
1806 /*
1807  *      Misc support functions
1808  */
1809
1810 static struct rt6_info *ip6_rt_copy(struct rt6_info *ort,
1811                                     const struct in6_addr *dest)
1812 {
1813         struct net *net = dev_net(ort->dst.dev);
1814         struct rt6_info *rt = ip6_dst_alloc(net, ort->dst.dev, 0,
1815                                             ort->rt6i_table);
1816
1817         if (rt) {
1818                 rt->dst.input = ort->dst.input;
1819                 rt->dst.output = ort->dst.output;
1820                 rt->dst.flags |= DST_HOST;
1821
1822                 rt->rt6i_dst.addr = *dest;
1823                 rt->rt6i_dst.plen = 128;
1824                 dst_copy_metrics(&rt->dst, &ort->dst);
1825                 rt->dst.error = ort->dst.error;
1826                 rt->rt6i_idev = ort->rt6i_idev;
1827                 if (rt->rt6i_idev)
1828                         in6_dev_hold(rt->rt6i_idev);
1829                 rt->dst.lastuse = jiffies;
1830
1831                 if (ort->rt6i_flags & RTF_GATEWAY)
1832                         rt->rt6i_gateway = ort->rt6i_gateway;
1833                 else
1834                         rt->rt6i_gateway = *dest;
1835                 rt->rt6i_flags = ort->rt6i_flags;
1836                 rt6_set_from(rt, ort);
1837                 rt->rt6i_metric = 0;
1838
1839 #ifdef CONFIG_IPV6_SUBTREES
1840                 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1841 #endif
1842                 memcpy(&rt->rt6i_prefsrc, &ort->rt6i_prefsrc, sizeof(struct rt6key));
1843                 rt->rt6i_table = ort->rt6i_table;
1844         }
1845         return rt;
1846 }
1847
1848 #ifdef CONFIG_IPV6_ROUTE_INFO
1849 static struct rt6_info *rt6_get_route_info(struct net_device *dev,
1850                                            const struct in6_addr *prefix, int prefixlen,
1851                                            const struct in6_addr *gwaddr)
1852 {
1853         struct fib6_node *fn;
1854         struct rt6_info *rt = NULL;
1855         struct fib6_table *table;
1856
1857         table = fib6_get_table(dev_net(dev),
1858                                addrconf_rt_table(dev, RT6_TABLE_INFO));
1859         if (!table)
1860                 return NULL;
1861
1862         read_lock_bh(&table->tb6_lock);
1863         fn = fib6_locate(&table->tb6_root, prefix ,prefixlen, NULL, 0);
1864         if (!fn)
1865                 goto out;
1866
1867         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1868                 if (rt->dst.dev->ifindex != dev->ifindex)
1869                         continue;
1870                 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
1871                         continue;
1872                 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
1873                         continue;
1874                 dst_hold(&rt->dst);
1875                 break;
1876         }
1877 out:
1878         read_unlock_bh(&table->tb6_lock);
1879         return rt;
1880 }
1881
1882 static struct rt6_info *rt6_add_route_info(struct net_device *dev,
1883                                            const struct in6_addr *prefix, int prefixlen,
1884                                            const struct in6_addr *gwaddr, unsigned int pref)
1885 {
1886         struct fib6_config cfg = {
1887                 .fc_table       = addrconf_rt_table(dev, RT6_TABLE_INFO),
1888                 .fc_metric      = IP6_RT_PRIO_USER,
1889                 .fc_ifindex     = dev->ifindex,
1890                 .fc_dst_len     = prefixlen,
1891                 .fc_flags       = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
1892                                   RTF_UP | RTF_PREF(pref),
1893                 .fc_nlinfo.portid = 0,
1894                 .fc_nlinfo.nlh = NULL,
1895                 .fc_nlinfo.nl_net = dev_net(dev),
1896         };
1897
1898         cfg.fc_dst = *prefix;
1899         cfg.fc_gateway = *gwaddr;
1900
1901         /* We should treat it as a default route if prefix length is 0. */
1902         if (!prefixlen)
1903                 cfg.fc_flags |= RTF_DEFAULT;
1904
1905         ip6_route_add(&cfg);
1906
1907         return rt6_get_route_info(dev, prefix, prefixlen, gwaddr);
1908 }
1909 #endif
1910
1911 struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
1912 {
1913         struct rt6_info *rt;
1914         struct fib6_table *table;
1915
1916         table = fib6_get_table(dev_net(dev),
1917                                addrconf_rt_table(dev, RT6_TABLE_MAIN));
1918         if (!table)
1919                 return NULL;
1920
1921         read_lock_bh(&table->tb6_lock);
1922         for (rt = table->tb6_root.leaf; rt; rt=rt->dst.rt6_next) {
1923                 if (dev == rt->dst.dev &&
1924                     ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
1925                     ipv6_addr_equal(&rt->rt6i_gateway, addr))
1926                         break;
1927         }
1928         if (rt)
1929                 dst_hold(&rt->dst);
1930         read_unlock_bh(&table->tb6_lock);
1931         return rt;
1932 }
1933
1934 struct rt6_info *rt6_add_dflt_router(const struct in6_addr *gwaddr,
1935                                      struct net_device *dev,
1936                                      unsigned int pref)
1937 {
1938         struct fib6_config cfg = {
1939                 .fc_table       = addrconf_rt_table(dev, RT6_TABLE_DFLT),
1940                 .fc_metric      = IP6_RT_PRIO_USER,
1941                 .fc_ifindex     = dev->ifindex,
1942                 .fc_flags       = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
1943                                   RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
1944                 .fc_nlinfo.portid = 0,
1945                 .fc_nlinfo.nlh = NULL,
1946                 .fc_nlinfo.nl_net = dev_net(dev),
1947         };
1948
1949         cfg.fc_gateway = *gwaddr;
1950
1951         ip6_route_add(&cfg);
1952
1953         return rt6_get_dflt_router(gwaddr, dev);
1954 }
1955
1956
1957 int rt6_addrconf_purge(struct rt6_info *rt, void *arg) {
1958         if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
1959             (!rt->rt6i_idev || rt->rt6i_idev->cnf.accept_ra != 2))
1960                 return -1;
1961         return 0;
1962 }
1963
1964 void rt6_purge_dflt_routers(struct net *net)
1965 {
1966         fib6_clean_all(net, rt6_addrconf_purge, 0, NULL);
1967 }
1968
1969 static void rtmsg_to_fib6_config(struct net *net,
1970                                  struct in6_rtmsg *rtmsg,
1971                                  struct fib6_config *cfg)
1972 {
1973         memset(cfg, 0, sizeof(*cfg));
1974
1975         cfg->fc_table = RT6_TABLE_MAIN;
1976         cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
1977         cfg->fc_metric = rtmsg->rtmsg_metric;
1978         cfg->fc_expires = rtmsg->rtmsg_info;
1979         cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
1980         cfg->fc_src_len = rtmsg->rtmsg_src_len;
1981         cfg->fc_flags = rtmsg->rtmsg_flags;
1982
1983         cfg->fc_nlinfo.nl_net = net;
1984
1985         cfg->fc_dst = rtmsg->rtmsg_dst;
1986         cfg->fc_src = rtmsg->rtmsg_src;
1987         cfg->fc_gateway = rtmsg->rtmsg_gateway;
1988 }
1989
1990 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
1991 {
1992         struct fib6_config cfg;
1993         struct in6_rtmsg rtmsg;
1994         int err;
1995
1996         switch(cmd) {
1997         case SIOCADDRT:         /* Add a route */
1998         case SIOCDELRT:         /* Delete a route */
1999                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2000                         return -EPERM;
2001                 err = copy_from_user(&rtmsg, arg,
2002                                      sizeof(struct in6_rtmsg));
2003                 if (err)
2004                         return -EFAULT;
2005
2006                 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
2007
2008                 rtnl_lock();
2009                 switch (cmd) {
2010                 case SIOCADDRT:
2011                         err = ip6_route_add(&cfg);
2012                         break;
2013                 case SIOCDELRT:
2014                         err = ip6_route_del(&cfg);
2015                         break;
2016                 default:
2017                         err = -EINVAL;
2018                 }
2019                 rtnl_unlock();
2020
2021                 return err;
2022         }
2023
2024         return -EINVAL;
2025 }
2026
2027 /*
2028  *      Drop the packet on the floor
2029  */
2030
2031 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
2032 {
2033         int type;
2034         struct dst_entry *dst = skb_dst(skb);
2035         switch (ipstats_mib_noroutes) {
2036         case IPSTATS_MIB_INNOROUTES:
2037                 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
2038                 if (type == IPV6_ADDR_ANY) {
2039                         IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2040                                       IPSTATS_MIB_INADDRERRORS);
2041                         break;
2042                 }
2043                 /* FALLTHROUGH */
2044         case IPSTATS_MIB_OUTNOROUTES:
2045                 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2046                               ipstats_mib_noroutes);
2047                 break;
2048         }
2049         icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
2050         kfree_skb(skb);
2051         return 0;
2052 }
2053
2054 static int ip6_pkt_discard(struct sk_buff *skb)
2055 {
2056         return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
2057 }
2058
2059 static int ip6_pkt_discard_out(struct sk_buff *skb)
2060 {
2061         skb->dev = skb_dst(skb)->dev;
2062         return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
2063 }
2064
2065 static int ip6_pkt_prohibit(struct sk_buff *skb)
2066 {
2067         return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
2068 }
2069
2070 static int ip6_pkt_prohibit_out(struct sk_buff *skb)
2071 {
2072         skb->dev = skb_dst(skb)->dev;
2073         return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
2074 }
2075
2076 /*
2077  *      Allocate a dst for local (unicast / anycast) address.
2078  */
2079
2080 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
2081                                     const struct in6_addr *addr,
2082                                     bool anycast)
2083 {
2084         struct net *net = dev_net(idev->dev);
2085         struct rt6_info *rt = ip6_dst_alloc(net, net->loopback_dev,
2086                                             DST_NOCOUNT, NULL);
2087         if (!rt)
2088                 return ERR_PTR(-ENOMEM);
2089
2090         in6_dev_hold(idev);
2091
2092         rt->dst.flags |= DST_HOST;
2093         rt->dst.input = ip6_input;
2094         rt->dst.output = ip6_output;
2095         rt->rt6i_idev = idev;
2096
2097         rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
2098         if (anycast)
2099                 rt->rt6i_flags |= RTF_ANYCAST;
2100         else
2101                 rt->rt6i_flags |= RTF_LOCAL;
2102
2103         rt->rt6i_gateway  = *addr;
2104         rt->rt6i_dst.addr = *addr;
2105         rt->rt6i_dst.plen = 128;
2106         rt->rt6i_table = fib6_get_table(net, RT6_TABLE_LOCAL);
2107
2108         atomic_set(&rt->dst.__refcnt, 1);
2109
2110         return rt;
2111 }
2112
2113 int ip6_route_get_saddr(struct net *net,
2114                         struct rt6_info *rt,
2115                         const struct in6_addr *daddr,
2116                         unsigned int prefs,
2117                         struct in6_addr *saddr)
2118 {
2119         struct inet6_dev *idev = ip6_dst_idev((struct dst_entry*)rt);
2120         int err = 0;
2121         if (rt->rt6i_prefsrc.plen)
2122                 *saddr = rt->rt6i_prefsrc.addr;
2123         else
2124                 err = ipv6_dev_get_saddr(net, idev ? idev->dev : NULL,
2125                                          daddr, prefs, saddr);
2126         return err;
2127 }
2128
2129 /* remove deleted ip from prefsrc entries */
2130 struct arg_dev_net_ip {
2131         struct net_device *dev;
2132         struct net *net;
2133         struct in6_addr *addr;
2134 };
2135
2136 static int fib6_remove_prefsrc(struct rt6_info *rt, void *arg)
2137 {
2138         struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
2139         struct net *net = ((struct arg_dev_net_ip *)arg)->net;
2140         struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
2141
2142         if (((void *)rt->dst.dev == dev || !dev) &&
2143             rt != net->ipv6.ip6_null_entry &&
2144             ipv6_addr_equal(addr, &rt->rt6i_prefsrc.addr)) {
2145                 /* remove prefsrc entry */
2146                 rt->rt6i_prefsrc.plen = 0;
2147         }
2148         return 0;
2149 }
2150
2151 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
2152 {
2153         struct net *net = dev_net(ifp->idev->dev);
2154         struct arg_dev_net_ip adni = {
2155                 .dev = ifp->idev->dev,
2156                 .net = net,
2157                 .addr = &ifp->addr,
2158         };
2159         fib6_clean_all(net, fib6_remove_prefsrc, 0, &adni);
2160 }
2161
2162 struct arg_dev_net {
2163         struct net_device *dev;
2164         struct net *net;
2165 };
2166
2167 static int fib6_ifdown(struct rt6_info *rt, void *arg)
2168 {
2169         const struct arg_dev_net *adn = arg;
2170         const struct net_device *dev = adn->dev;
2171
2172         if ((rt->dst.dev == dev || !dev) &&
2173             rt != adn->net->ipv6.ip6_null_entry)
2174                 return -1;
2175
2176         return 0;
2177 }
2178
2179 void rt6_ifdown(struct net *net, struct net_device *dev)
2180 {
2181         struct arg_dev_net adn = {
2182                 .dev = dev,
2183                 .net = net,
2184         };
2185
2186         fib6_clean_all(net, fib6_ifdown, 0, &adn);
2187         icmp6_clean_all(fib6_ifdown, &adn);
2188 }
2189
2190 struct rt6_mtu_change_arg {
2191         struct net_device *dev;
2192         unsigned int mtu;
2193 };
2194
2195 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2196 {
2197         struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2198         struct inet6_dev *idev;
2199
2200         /* In IPv6 pmtu discovery is not optional,
2201            so that RTAX_MTU lock cannot disable it.
2202            We still use this lock to block changes
2203            caused by addrconf/ndisc.
2204         */
2205
2206         idev = __in6_dev_get(arg->dev);
2207         if (!idev)
2208                 return 0;
2209
2210         /* For administrative MTU increase, there is no way to discover
2211            IPv6 PMTU increase, so PMTU increase should be updated here.
2212            Since RFC 1981 doesn't include administrative MTU increase
2213            update PMTU increase is a MUST. (i.e. jumbo frame)
2214          */
2215         /*
2216            If new MTU is less than route PMTU, this new MTU will be the
2217            lowest MTU in the path, update the route PMTU to reflect PMTU
2218            decreases; if new MTU is greater than route PMTU, and the
2219            old MTU is the lowest MTU in the path, update the route PMTU
2220            to reflect the increase. In this case if the other nodes' MTU
2221            also have the lowest MTU, TOO BIG MESSAGE will be lead to
2222            PMTU discouvery.
2223          */
2224         if (rt->dst.dev == arg->dev &&
2225             !dst_metric_locked(&rt->dst, RTAX_MTU) &&
2226             (dst_mtu(&rt->dst) >= arg->mtu ||
2227              (dst_mtu(&rt->dst) < arg->mtu &&
2228               dst_mtu(&rt->dst) == idev->cnf.mtu6))) {
2229                 dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
2230         }
2231         return 0;
2232 }
2233
2234 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
2235 {
2236         struct rt6_mtu_change_arg arg = {
2237                 .dev = dev,
2238                 .mtu = mtu,
2239         };
2240
2241         fib6_clean_all(dev_net(dev), rt6_mtu_change_route, 0, &arg);
2242 }
2243
2244 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2245         [RTA_GATEWAY]           = { .len = sizeof(struct in6_addr) },
2246         [RTA_OIF]               = { .type = NLA_U32 },
2247         [RTA_IIF]               = { .type = NLA_U32 },
2248         [RTA_PRIORITY]          = { .type = NLA_U32 },
2249         [RTA_METRICS]           = { .type = NLA_NESTED },
2250         [RTA_MULTIPATH]         = { .len = sizeof(struct rtnexthop) },
2251         [RTA_UID]               = { .type = NLA_U32 },
2252 };
2253
2254 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2255                               struct fib6_config *cfg)
2256 {
2257         struct rtmsg *rtm;
2258         struct nlattr *tb[RTA_MAX+1];
2259         int err;
2260
2261         err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2262         if (err < 0)
2263                 goto errout;
2264
2265         err = -EINVAL;
2266         rtm = nlmsg_data(nlh);
2267         memset(cfg, 0, sizeof(*cfg));
2268
2269         cfg->fc_table = rtm->rtm_table;
2270         cfg->fc_dst_len = rtm->rtm_dst_len;
2271         cfg->fc_src_len = rtm->rtm_src_len;
2272         cfg->fc_flags = RTF_UP;
2273         cfg->fc_protocol = rtm->rtm_protocol;
2274         cfg->fc_type = rtm->rtm_type;
2275
2276         if (rtm->rtm_type == RTN_UNREACHABLE ||
2277             rtm->rtm_type == RTN_BLACKHOLE ||
2278             rtm->rtm_type == RTN_PROHIBIT ||
2279             rtm->rtm_type == RTN_THROW)
2280                 cfg->fc_flags |= RTF_REJECT;
2281
2282         if (rtm->rtm_type == RTN_LOCAL)
2283                 cfg->fc_flags |= RTF_LOCAL;
2284
2285         cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid;
2286         cfg->fc_nlinfo.nlh = nlh;
2287         cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2288
2289         if (tb[RTA_GATEWAY]) {
2290                 nla_memcpy(&cfg->fc_gateway, tb[RTA_GATEWAY], 16);
2291                 cfg->fc_flags |= RTF_GATEWAY;
2292         }
2293
2294         if (tb[RTA_DST]) {
2295                 int plen = (rtm->rtm_dst_len + 7) >> 3;
2296
2297                 if (nla_len(tb[RTA_DST]) < plen)
2298                         goto errout;
2299
2300                 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2301         }
2302
2303         if (tb[RTA_SRC]) {
2304                 int plen = (rtm->rtm_src_len + 7) >> 3;
2305
2306                 if (nla_len(tb[RTA_SRC]) < plen)
2307                         goto errout;
2308
2309                 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2310         }
2311
2312         if (tb[RTA_PREFSRC])
2313                 nla_memcpy(&cfg->fc_prefsrc, tb[RTA_PREFSRC], 16);
2314
2315         if (tb[RTA_OIF])
2316                 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2317
2318         if (tb[RTA_PRIORITY])
2319                 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2320
2321         if (tb[RTA_METRICS]) {
2322                 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2323                 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2324         }
2325
2326         if (tb[RTA_TABLE])
2327                 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2328
2329         if (tb[RTA_MULTIPATH]) {
2330                 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
2331                 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
2332         }
2333
2334         err = 0;
2335 errout:
2336         return err;
2337 }
2338
2339 static int ip6_route_multipath(struct fib6_config *cfg, int add)
2340 {
2341         struct fib6_config r_cfg;
2342         struct rtnexthop *rtnh;
2343         int remaining;
2344         int attrlen;
2345         int err = 0, last_err = 0;
2346
2347 beginning:
2348         rtnh = (struct rtnexthop *)cfg->fc_mp;
2349         remaining = cfg->fc_mp_len;
2350
2351         /* Parse a Multipath Entry */
2352         while (rtnh_ok(rtnh, remaining)) {
2353                 memcpy(&r_cfg, cfg, sizeof(*cfg));
2354                 if (rtnh->rtnh_ifindex)
2355                         r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
2356
2357                 attrlen = rtnh_attrlen(rtnh);
2358                 if (attrlen > 0) {
2359                         struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
2360
2361                         nla = nla_find(attrs, attrlen, RTA_GATEWAY);
2362                         if (nla) {
2363                                 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
2364                                 r_cfg.fc_flags |= RTF_GATEWAY;
2365                         }
2366                 }
2367                 err = add ? ip6_route_add(&r_cfg) : ip6_route_del(&r_cfg);
2368                 if (err) {
2369                         last_err = err;
2370                         /* If we are trying to remove a route, do not stop the
2371                          * loop when ip6_route_del() fails (because next hop is
2372                          * already gone), we should try to remove all next hops.
2373                          */
2374                         if (add) {
2375                                 /* If add fails, we should try to delete all
2376                                  * next hops that have been already added.
2377                                  */
2378                                 add = 0;
2379                                 goto beginning;
2380                         }
2381                 }
2382                 /* Because each route is added like a single route we remove
2383                  * this flag after the first nexthop (if there is a collision,
2384                  * we have already fail to add the first nexthop:
2385                  * fib6_add_rt2node() has reject it).
2386                  */
2387                 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~NLM_F_EXCL;
2388                 rtnh = rtnh_next(rtnh, &remaining);
2389         }
2390
2391         return last_err;
2392 }
2393
2394 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr* nlh)
2395 {
2396         struct fib6_config cfg;
2397         int err;
2398
2399         err = rtm_to_fib6_config(skb, nlh, &cfg);
2400         if (err < 0)
2401                 return err;
2402
2403         if (cfg.fc_mp)
2404                 return ip6_route_multipath(&cfg, 0);
2405         else
2406                 return ip6_route_del(&cfg);
2407 }
2408
2409 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr* nlh)
2410 {
2411         struct fib6_config cfg;
2412         int err;
2413
2414         err = rtm_to_fib6_config(skb, nlh, &cfg);
2415         if (err < 0)
2416                 return err;
2417
2418         if (cfg.fc_mp)
2419                 return ip6_route_multipath(&cfg, 1);
2420         else
2421                 return ip6_route_add(&cfg);
2422 }
2423
2424 static inline size_t rt6_nlmsg_size(void)
2425 {
2426         return NLMSG_ALIGN(sizeof(struct rtmsg))
2427                + nla_total_size(16) /* RTA_SRC */
2428                + nla_total_size(16) /* RTA_DST */
2429                + nla_total_size(16) /* RTA_GATEWAY */
2430                + nla_total_size(16) /* RTA_PREFSRC */
2431                + nla_total_size(4) /* RTA_TABLE */
2432                + nla_total_size(4) /* RTA_IIF */
2433                + nla_total_size(4) /* RTA_OIF */
2434                + nla_total_size(4) /* RTA_PRIORITY */
2435                + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
2436                + nla_total_size(sizeof(struct rta_cacheinfo));
2437 }
2438
2439 static int rt6_fill_node(struct net *net,
2440                          struct sk_buff *skb, struct rt6_info *rt,
2441                          struct in6_addr *dst, struct in6_addr *src,
2442                          int iif, int type, u32 portid, u32 seq,
2443                          int prefix, int nowait, unsigned int flags)
2444 {
2445         struct rtmsg *rtm;
2446         struct nlmsghdr *nlh;
2447         long expires;
2448         u32 table;
2449
2450         if (prefix) {   /* user wants prefix routes only */
2451                 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
2452                         /* success since this is not a prefix route */
2453                         return 1;
2454                 }
2455         }
2456
2457         nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
2458         if (!nlh)
2459                 return -EMSGSIZE;
2460
2461         rtm = nlmsg_data(nlh);
2462         rtm->rtm_family = AF_INET6;
2463         rtm->rtm_dst_len = rt->rt6i_dst.plen;
2464         rtm->rtm_src_len = rt->rt6i_src.plen;
2465         rtm->rtm_tos = 0;
2466         if (rt->rt6i_table)
2467                 table = rt->rt6i_table->tb6_id;
2468         else
2469                 table = RT6_TABLE_UNSPEC;
2470         rtm->rtm_table = table;
2471         if (nla_put_u32(skb, RTA_TABLE, table))
2472                 goto nla_put_failure;
2473         if (rt->rt6i_flags & RTF_REJECT) {
2474                 switch (rt->dst.error) {
2475                 case -EINVAL:
2476                         rtm->rtm_type = RTN_BLACKHOLE;
2477                         break;
2478                 case -EACCES:
2479                         rtm->rtm_type = RTN_PROHIBIT;
2480                         break;
2481                 case -EAGAIN:
2482                         rtm->rtm_type = RTN_THROW;
2483                         break;
2484                 default:
2485                         rtm->rtm_type = RTN_UNREACHABLE;
2486                         break;
2487                 }
2488         }
2489         else if (rt->rt6i_flags & RTF_LOCAL)
2490                 rtm->rtm_type = RTN_LOCAL;
2491         else if (rt->dst.dev && (rt->dst.dev->flags & IFF_LOOPBACK))
2492                 rtm->rtm_type = RTN_LOCAL;
2493         else
2494                 rtm->rtm_type = RTN_UNICAST;
2495         rtm->rtm_flags = 0;
2496         rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2497         rtm->rtm_protocol = rt->rt6i_protocol;
2498         if (rt->rt6i_flags & RTF_DYNAMIC)
2499                 rtm->rtm_protocol = RTPROT_REDIRECT;
2500         else if (rt->rt6i_flags & RTF_ADDRCONF) {
2501                 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ROUTEINFO))
2502                         rtm->rtm_protocol = RTPROT_RA;
2503                 else
2504                         rtm->rtm_protocol = RTPROT_KERNEL;
2505         }
2506
2507         if (rt->rt6i_flags & RTF_CACHE)
2508                 rtm->rtm_flags |= RTM_F_CLONED;
2509
2510         if (dst) {
2511                 if (nla_put(skb, RTA_DST, 16, dst))
2512                         goto nla_put_failure;
2513                 rtm->rtm_dst_len = 128;
2514         } else if (rtm->rtm_dst_len)
2515                 if (nla_put(skb, RTA_DST, 16, &rt->rt6i_dst.addr))
2516                         goto nla_put_failure;
2517 #ifdef CONFIG_IPV6_SUBTREES
2518         if (src) {
2519                 if (nla_put(skb, RTA_SRC, 16, src))
2520                         goto nla_put_failure;
2521                 rtm->rtm_src_len = 128;
2522         } else if (rtm->rtm_src_len &&
2523                    nla_put(skb, RTA_SRC, 16, &rt->rt6i_src.addr))
2524                 goto nla_put_failure;
2525 #endif
2526         if (iif) {
2527 #ifdef CONFIG_IPV6_MROUTE
2528                 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
2529                         int err = ip6mr_get_route(net, skb, rtm, nowait);
2530                         if (err <= 0) {
2531                                 if (!nowait) {
2532                                         if (err == 0)
2533                                                 return 0;
2534                                         goto nla_put_failure;
2535                                 } else {
2536                                         if (err == -EMSGSIZE)
2537                                                 goto nla_put_failure;
2538                                 }
2539                         }
2540                 } else
2541 #endif
2542                         if (nla_put_u32(skb, RTA_IIF, iif))
2543                                 goto nla_put_failure;
2544         } else if (dst) {
2545                 struct in6_addr saddr_buf;
2546                 if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0 &&
2547                     nla_put(skb, RTA_PREFSRC, 16, &saddr_buf))
2548                         goto nla_put_failure;
2549         }
2550
2551         if (rt->rt6i_prefsrc.plen) {
2552                 struct in6_addr saddr_buf;
2553                 saddr_buf = rt->rt6i_prefsrc.addr;
2554                 if (nla_put(skb, RTA_PREFSRC, 16, &saddr_buf))
2555                         goto nla_put_failure;
2556         }
2557
2558         if (rtnetlink_put_metrics(skb, dst_metrics_ptr(&rt->dst)) < 0)
2559                 goto nla_put_failure;
2560
2561         if (rt->rt6i_flags & RTF_GATEWAY) {
2562                 if (nla_put(skb, RTA_GATEWAY, 16, &rt->rt6i_gateway) < 0)
2563                         goto nla_put_failure;
2564         }
2565
2566         if (rt->dst.dev &&
2567             nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex))
2568                 goto nla_put_failure;
2569         if (nla_put_u32(skb, RTA_PRIORITY, rt->rt6i_metric))
2570                 goto nla_put_failure;
2571
2572         expires = (rt->rt6i_flags & RTF_EXPIRES) ? rt->dst.expires - jiffies : 0;
2573
2574         if (rtnl_put_cacheinfo(skb, &rt->dst, 0, expires, rt->dst.error) < 0)
2575                 goto nla_put_failure;
2576
2577         return nlmsg_end(skb, nlh);
2578
2579 nla_put_failure:
2580         nlmsg_cancel(skb, nlh);
2581         return -EMSGSIZE;
2582 }
2583
2584 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
2585 {
2586         struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
2587         int prefix;
2588
2589         if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
2590                 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
2591                 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
2592         } else
2593                 prefix = 0;
2594
2595         return rt6_fill_node(arg->net,
2596                      arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
2597                      NETLINK_CB(arg->cb->skb).portid, arg->cb->nlh->nlmsg_seq,
2598                      prefix, 0, NLM_F_MULTI);
2599 }
2600
2601 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr* nlh)
2602 {
2603         struct net *net = sock_net(in_skb->sk);
2604         struct nlattr *tb[RTA_MAX+1];
2605         struct rt6_info *rt;
2606         struct sk_buff *skb;
2607         struct rtmsg *rtm;
2608         struct flowi6 fl6;
2609         int err, iif = 0, oif = 0;
2610
2611         err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2612         if (err < 0)
2613                 goto errout;
2614
2615         err = -EINVAL;
2616         memset(&fl6, 0, sizeof(fl6));
2617
2618         if (tb[RTA_SRC]) {
2619                 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
2620                         goto errout;
2621
2622                 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
2623         }
2624
2625         if (tb[RTA_DST]) {
2626                 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
2627                         goto errout;
2628
2629                 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
2630         }
2631
2632         if (tb[RTA_IIF])
2633                 iif = nla_get_u32(tb[RTA_IIF]);
2634
2635         if (tb[RTA_OIF])
2636                 oif = nla_get_u32(tb[RTA_OIF]);
2637
2638         if (tb[RTA_MARK])
2639                 fl6.flowi6_mark = nla_get_u32(tb[RTA_MARK]);
2640
2641         if (tb[RTA_UID])
2642                 fl6.flowi6_uid = make_kuid(current_user_ns(),
2643                                            nla_get_u32(tb[RTA_UID]));
2644         else
2645                 fl6.flowi6_uid = iif ? INVALID_UID : current_uid();
2646
2647         if (iif) {
2648                 struct net_device *dev;
2649                 int flags = 0;
2650
2651                 dev = __dev_get_by_index(net, iif);
2652                 if (!dev) {
2653                         err = -ENODEV;
2654                         goto errout;
2655                 }
2656
2657                 fl6.flowi6_iif = iif;
2658
2659                 if (!ipv6_addr_any(&fl6.saddr))
2660                         flags |= RT6_LOOKUP_F_HAS_SADDR;
2661
2662                 rt = (struct rt6_info *)ip6_route_input_lookup(net, dev, &fl6,
2663                                                                flags);
2664         } else {
2665                 fl6.flowi6_oif = oif;
2666
2667                 rt = (struct rt6_info *)ip6_route_output(net, NULL, &fl6);
2668         }
2669
2670         skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
2671         if (!skb) {
2672                 ip6_rt_put(rt);
2673                 err = -ENOBUFS;
2674                 goto errout;
2675         }
2676
2677         /* Reserve room for dummy headers, this skb can pass
2678            through good chunk of routing engine.
2679          */
2680         skb_reset_mac_header(skb);
2681         skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
2682
2683         skb_dst_set(skb, &rt->dst);
2684
2685         err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
2686                             RTM_NEWROUTE, NETLINK_CB(in_skb).portid,
2687                             nlh->nlmsg_seq, 0, 0, 0);
2688         if (err < 0) {
2689                 kfree_skb(skb);
2690                 goto errout;
2691         }
2692
2693         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
2694 errout:
2695         return err;
2696 }
2697
2698 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info)
2699 {
2700         struct sk_buff *skb;
2701         struct net *net = info->nl_net;
2702         u32 seq;
2703         int err;
2704
2705         err = -ENOBUFS;
2706         seq = info->nlh ? info->nlh->nlmsg_seq : 0;
2707
2708         skb = nlmsg_new(rt6_nlmsg_size(), gfp_any());
2709         if (!skb)
2710                 goto errout;
2711
2712         err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
2713                                 event, info->portid, seq, 0, 0, 0);
2714         if (err < 0) {
2715                 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
2716                 WARN_ON(err == -EMSGSIZE);
2717                 kfree_skb(skb);
2718                 goto errout;
2719         }
2720         rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
2721                     info->nlh, gfp_any());
2722         return;
2723 errout:
2724         if (err < 0)
2725                 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
2726 }
2727
2728 static int ip6_route_dev_notify(struct notifier_block *this,
2729                                 unsigned long event, void *data)
2730 {
2731         struct net_device *dev = (struct net_device *)data;
2732         struct net *net = dev_net(dev);
2733
2734         if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
2735                 net->ipv6.ip6_null_entry->dst.dev = dev;
2736                 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
2737 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2738                 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
2739                 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
2740                 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
2741                 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
2742 #endif
2743         }
2744
2745         return NOTIFY_OK;
2746 }
2747
2748 /*
2749  *      /proc
2750  */
2751
2752 #ifdef CONFIG_PROC_FS
2753
2754 struct rt6_proc_arg
2755 {
2756         char *buffer;
2757         int offset;
2758         int length;
2759         int skip;
2760         int len;
2761 };
2762
2763 static int rt6_info_route(struct rt6_info *rt, void *p_arg)
2764 {
2765         struct seq_file *m = p_arg;
2766
2767         seq_printf(m, "%pi6 %02x ", &rt->rt6i_dst.addr, rt->rt6i_dst.plen);
2768
2769 #ifdef CONFIG_IPV6_SUBTREES
2770         seq_printf(m, "%pi6 %02x ", &rt->rt6i_src.addr, rt->rt6i_src.plen);
2771 #else
2772         seq_puts(m, "00000000000000000000000000000000 00 ");
2773 #endif
2774         if (rt->rt6i_flags & RTF_GATEWAY) {
2775                 seq_printf(m, "%pi6", &rt->rt6i_gateway);
2776         } else {
2777                 seq_puts(m, "00000000000000000000000000000000");
2778         }
2779         seq_printf(m, " %08x %08x %08x %08x %8s\n",
2780                    rt->rt6i_metric, atomic_read(&rt->dst.__refcnt),
2781                    rt->dst.__use, rt->rt6i_flags,
2782                    rt->dst.dev ? rt->dst.dev->name : "");
2783         return 0;
2784 }
2785
2786 static int ipv6_route_show(struct seq_file *m, void *v)
2787 {
2788         struct net *net = (struct net *)m->private;
2789         fib6_clean_all_ro(net, rt6_info_route, 0, m);
2790         return 0;
2791 }
2792
2793 static int ipv6_route_open(struct inode *inode, struct file *file)
2794 {
2795         return single_open_net(inode, file, ipv6_route_show);
2796 }
2797
2798 static const struct file_operations ipv6_route_proc_fops = {
2799         .owner          = THIS_MODULE,
2800         .open           = ipv6_route_open,
2801         .read           = seq_read,
2802         .llseek         = seq_lseek,
2803         .release        = single_release_net,
2804 };
2805
2806 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
2807 {
2808         struct net *net = (struct net *)seq->private;
2809         seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
2810                    net->ipv6.rt6_stats->fib_nodes,
2811                    net->ipv6.rt6_stats->fib_route_nodes,
2812                    net->ipv6.rt6_stats->fib_rt_alloc,
2813                    net->ipv6.rt6_stats->fib_rt_entries,
2814                    net->ipv6.rt6_stats->fib_rt_cache,
2815                    dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
2816                    net->ipv6.rt6_stats->fib_discarded_routes);
2817
2818         return 0;
2819 }
2820
2821 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
2822 {
2823         return single_open_net(inode, file, rt6_stats_seq_show);
2824 }
2825
2826 static const struct file_operations rt6_stats_seq_fops = {
2827         .owner   = THIS_MODULE,
2828         .open    = rt6_stats_seq_open,
2829         .read    = seq_read,
2830         .llseek  = seq_lseek,
2831         .release = single_release_net,
2832 };
2833 #endif  /* CONFIG_PROC_FS */
2834
2835 #ifdef CONFIG_SYSCTL
2836
2837 static
2838 int ipv6_sysctl_rtcache_flush(ctl_table *ctl, int write,
2839                               void __user *buffer, size_t *lenp, loff_t *ppos)
2840 {
2841         struct net *net;
2842         int delay;
2843         if (!write)
2844                 return -EINVAL;
2845
2846         net = (struct net *)ctl->extra1;
2847         delay = net->ipv6.sysctl.flush_delay;
2848         proc_dointvec(ctl, write, buffer, lenp, ppos);
2849         fib6_run_gc(delay <= 0 ? ~0UL : (unsigned long)delay, net);
2850         return 0;
2851 }
2852
2853 ctl_table ipv6_route_table_template[] = {
2854         {
2855                 .procname       =       "flush",
2856                 .data           =       &init_net.ipv6.sysctl.flush_delay,
2857                 .maxlen         =       sizeof(int),
2858                 .mode           =       0200,
2859                 .proc_handler   =       ipv6_sysctl_rtcache_flush
2860         },
2861         {
2862                 .procname       =       "gc_thresh",
2863                 .data           =       &ip6_dst_ops_template.gc_thresh,
2864                 .maxlen         =       sizeof(int),
2865                 .mode           =       0644,
2866                 .proc_handler   =       proc_dointvec,
2867         },
2868         {
2869                 .procname       =       "max_size",
2870                 .data           =       &init_net.ipv6.sysctl.ip6_rt_max_size,
2871                 .maxlen         =       sizeof(int),
2872                 .mode           =       0644,
2873                 .proc_handler   =       proc_dointvec,
2874         },
2875         {
2876                 .procname       =       "gc_min_interval",
2877                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2878                 .maxlen         =       sizeof(int),
2879                 .mode           =       0644,
2880                 .proc_handler   =       proc_dointvec_jiffies,
2881         },
2882         {
2883                 .procname       =       "gc_timeout",
2884                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
2885                 .maxlen         =       sizeof(int),
2886                 .mode           =       0644,
2887                 .proc_handler   =       proc_dointvec_jiffies,
2888         },
2889         {
2890                 .procname       =       "gc_interval",
2891                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_interval,
2892                 .maxlen         =       sizeof(int),
2893                 .mode           =       0644,
2894                 .proc_handler   =       proc_dointvec_jiffies,
2895         },
2896         {
2897                 .procname       =       "gc_elasticity",
2898                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
2899                 .maxlen         =       sizeof(int),
2900                 .mode           =       0644,
2901                 .proc_handler   =       proc_dointvec,
2902         },
2903         {
2904                 .procname       =       "mtu_expires",
2905                 .data           =       &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
2906                 .maxlen         =       sizeof(int),
2907                 .mode           =       0644,
2908                 .proc_handler   =       proc_dointvec_jiffies,
2909         },
2910         {
2911                 .procname       =       "min_adv_mss",
2912                 .data           =       &init_net.ipv6.sysctl.ip6_rt_min_advmss,
2913                 .maxlen         =       sizeof(int),
2914                 .mode           =       0644,
2915                 .proc_handler   =       proc_dointvec,
2916         },
2917         {
2918                 .procname       =       "gc_min_interval_ms",
2919                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2920                 .maxlen         =       sizeof(int),
2921                 .mode           =       0644,
2922                 .proc_handler   =       proc_dointvec_ms_jiffies,
2923         },
2924         { }
2925 };
2926
2927 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
2928 {
2929         struct ctl_table *table;
2930
2931         table = kmemdup(ipv6_route_table_template,
2932                         sizeof(ipv6_route_table_template),
2933                         GFP_KERNEL);
2934
2935         if (table) {
2936                 table[0].data = &net->ipv6.sysctl.flush_delay;
2937                 table[0].extra1 = net;
2938                 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
2939                 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
2940                 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2941                 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
2942                 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
2943                 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
2944                 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
2945                 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
2946                 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2947
2948                 /* Don't export sysctls to unprivileged users */
2949                 if (net->user_ns != &init_user_ns)
2950                         table[0].procname = NULL;
2951         }
2952
2953         return table;
2954 }
2955 #endif
2956
2957 static int __net_init ip6_route_net_init(struct net *net)
2958 {
2959         int ret = -ENOMEM;
2960
2961         memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
2962                sizeof(net->ipv6.ip6_dst_ops));
2963
2964         if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
2965                 goto out_ip6_dst_ops;
2966
2967         net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
2968                                            sizeof(*net->ipv6.ip6_null_entry),
2969                                            GFP_KERNEL);
2970         if (!net->ipv6.ip6_null_entry)
2971                 goto out_ip6_dst_entries;
2972         net->ipv6.ip6_null_entry->dst.path =
2973                 (struct dst_entry *)net->ipv6.ip6_null_entry;
2974         net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
2975         dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
2976                          ip6_template_metrics, true);
2977
2978 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2979         net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
2980                                                sizeof(*net->ipv6.ip6_prohibit_entry),
2981                                                GFP_KERNEL);
2982         if (!net->ipv6.ip6_prohibit_entry)
2983                 goto out_ip6_null_entry;
2984         net->ipv6.ip6_prohibit_entry->dst.path =
2985                 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
2986         net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
2987         dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
2988                          ip6_template_metrics, true);
2989
2990         net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
2991                                                sizeof(*net->ipv6.ip6_blk_hole_entry),
2992                                                GFP_KERNEL);
2993         if (!net->ipv6.ip6_blk_hole_entry)
2994                 goto out_ip6_prohibit_entry;
2995         net->ipv6.ip6_blk_hole_entry->dst.path =
2996                 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
2997         net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
2998         dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
2999                          ip6_template_metrics, true);
3000 #endif
3001
3002         net->ipv6.sysctl.flush_delay = 0;
3003         net->ipv6.sysctl.ip6_rt_max_size = 4096;
3004         net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
3005         net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
3006         net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
3007         net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
3008         net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
3009         net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
3010
3011         net->ipv6.ip6_rt_gc_expire = 30*HZ;
3012
3013         ret = 0;
3014 out:
3015         return ret;
3016
3017 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3018 out_ip6_prohibit_entry:
3019         kfree(net->ipv6.ip6_prohibit_entry);
3020 out_ip6_null_entry:
3021         kfree(net->ipv6.ip6_null_entry);
3022 #endif
3023 out_ip6_dst_entries:
3024         dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3025 out_ip6_dst_ops:
3026         goto out;
3027 }
3028
3029 static void __net_exit ip6_route_net_exit(struct net *net)
3030 {
3031         kfree(net->ipv6.ip6_null_entry);
3032 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3033         kfree(net->ipv6.ip6_prohibit_entry);
3034         kfree(net->ipv6.ip6_blk_hole_entry);
3035 #endif
3036         dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3037 }
3038
3039 static int __net_init ip6_route_net_init_late(struct net *net)
3040 {
3041 #ifdef CONFIG_PROC_FS
3042         proc_create("ipv6_route", 0, net->proc_net, &ipv6_route_proc_fops);
3043         proc_create("rt6_stats", S_IRUGO, net->proc_net, &rt6_stats_seq_fops);
3044 #endif
3045         return 0;
3046 }
3047
3048 static void __net_exit ip6_route_net_exit_late(struct net *net)
3049 {
3050 #ifdef CONFIG_PROC_FS
3051         remove_proc_entry("ipv6_route", net->proc_net);
3052         remove_proc_entry("rt6_stats", net->proc_net);
3053 #endif
3054 }
3055
3056 static struct pernet_operations ip6_route_net_ops = {
3057         .init = ip6_route_net_init,
3058         .exit = ip6_route_net_exit,
3059 };
3060
3061 static int __net_init ipv6_inetpeer_init(struct net *net)
3062 {
3063         struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
3064
3065         if (!bp)
3066                 return -ENOMEM;
3067         inet_peer_base_init(bp);
3068         net->ipv6.peers = bp;
3069         return 0;
3070 }
3071
3072 static void __net_exit ipv6_inetpeer_exit(struct net *net)
3073 {
3074         struct inet_peer_base *bp = net->ipv6.peers;
3075
3076         net->ipv6.peers = NULL;
3077         inetpeer_invalidate_tree(bp);
3078         kfree(bp);
3079 }
3080
3081 static struct pernet_operations ipv6_inetpeer_ops = {
3082         .init   =       ipv6_inetpeer_init,
3083         .exit   =       ipv6_inetpeer_exit,
3084 };
3085
3086 static struct pernet_operations ip6_route_net_late_ops = {
3087         .init = ip6_route_net_init_late,
3088         .exit = ip6_route_net_exit_late,
3089 };
3090
3091 static struct notifier_block ip6_route_dev_notifier = {
3092         .notifier_call = ip6_route_dev_notify,
3093         .priority = 0,
3094 };
3095
3096 int __init ip6_route_init(void)
3097 {
3098         int ret;
3099
3100         ret = -ENOMEM;
3101         ip6_dst_ops_template.kmem_cachep =
3102                 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
3103                                   SLAB_HWCACHE_ALIGN, NULL);
3104         if (!ip6_dst_ops_template.kmem_cachep)
3105                 goto out;
3106
3107         ret = dst_entries_init(&ip6_dst_blackhole_ops);
3108         if (ret)
3109                 goto out_kmem_cache;
3110
3111         ret = register_pernet_subsys(&ipv6_inetpeer_ops);
3112         if (ret)
3113                 goto out_dst_entries;
3114
3115         ret = register_pernet_subsys(&ip6_route_net_ops);
3116         if (ret)
3117                 goto out_register_inetpeer;
3118
3119         ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
3120
3121         /* Registering of the loopback is done before this portion of code,
3122          * the loopback reference in rt6_info will not be taken, do it
3123          * manually for init_net */
3124         init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
3125         init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3126   #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3127         init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
3128         init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3129         init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
3130         init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3131   #endif
3132         ret = fib6_init();
3133         if (ret)
3134                 goto out_register_subsys;
3135
3136         ret = xfrm6_init();
3137         if (ret)
3138                 goto out_fib6_init;
3139
3140         ret = fib6_rules_init();
3141         if (ret)
3142                 goto xfrm6_init;
3143
3144         ret = register_pernet_subsys(&ip6_route_net_late_ops);
3145         if (ret)
3146                 goto fib6_rules_init;
3147
3148         ret = -ENOBUFS;
3149         if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL, NULL) ||
3150             __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL, NULL) ||
3151             __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL, NULL))
3152                 goto out_register_late_subsys;
3153
3154         ret = register_netdevice_notifier(&ip6_route_dev_notifier);
3155         if (ret)
3156                 goto out_register_late_subsys;
3157
3158 out:
3159         return ret;
3160
3161 out_register_late_subsys:
3162         unregister_pernet_subsys(&ip6_route_net_late_ops);
3163 fib6_rules_init:
3164         fib6_rules_cleanup();
3165 xfrm6_init:
3166         xfrm6_fini();
3167 out_fib6_init:
3168         fib6_gc_cleanup();
3169 out_register_subsys:
3170         unregister_pernet_subsys(&ip6_route_net_ops);
3171 out_register_inetpeer:
3172         unregister_pernet_subsys(&ipv6_inetpeer_ops);
3173 out_dst_entries:
3174         dst_entries_destroy(&ip6_dst_blackhole_ops);
3175 out_kmem_cache:
3176         kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3177         goto out;
3178 }
3179
3180 void ip6_route_cleanup(void)
3181 {
3182         unregister_netdevice_notifier(&ip6_route_dev_notifier);
3183         unregister_pernet_subsys(&ip6_route_net_late_ops);
3184         fib6_rules_cleanup();
3185         xfrm6_fini();
3186         fib6_gc_cleanup();
3187         unregister_pernet_subsys(&ipv6_inetpeer_ops);
3188         unregister_pernet_subsys(&ip6_route_net_ops);
3189         dst_entries_destroy(&ip6_dst_blackhole_ops);
3190         kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3191 }