net/ipv6/route.c

   1 /*
   2  *      Linux INET6 implementation
   3  *      FIB front-end.
   4  *
   5  *      Authors:
   6  *      Pedro Roque             <roque@di.fc.ul.pt>
   7  *
   8  *      This program is free software; you can redistribute it and/or
   9  *      modify it under the terms of the GNU General Public License
  10  *      as published by the Free Software Foundation; either version
  11  *      2 of the License, or (at your option) any later version.
  12  */
  13
  14 /*      Changes:
  15  *
  16  *      YOSHIFUJI Hideaki @USAGI
  17  *              reworked default router selection.
  18  *              - respect outgoing interface
  19  *              - select from (probably) reachable routers (i.e.
  20  *              routers in REACHABLE, STALE, DELAY or PROBE states).
  21  *              - always select the same router if it is (probably)
  22  *              reachable.  otherwise, round-robin the list.
  23  *      Ville Nuorvala
  24  *              Fixed routing subtrees.
  25  */
  26
  27 #define pr_fmt(fmt) "IPv6: " fmt
  28
  29 #include <linux/capability.h>
  30 #include <linux/errno.h>
  31 #include <linux/export.h>
  32 #include <linux/types.h>
  33 #include <linux/times.h>
  34 #include <linux/socket.h>
  35 #include <linux/sockios.h>
  36 #include <linux/net.h>
  37 #include <linux/route.h>
  38 #include <linux/netdevice.h>
  39 #include <linux/in6.h>
  40 #include <linux/mroute6.h>
  41 #include <linux/init.h>
  42 #include <linux/if_arp.h>
  43 #include <linux/proc_fs.h>
  44 #include <linux/seq_file.h>
  45 #include <linux/nsproxy.h>
  46 #include <linux/slab.h>
  47 #include <net/net_namespace.h>
  48 #include <net/snmp.h>
  49 #include <net/ipv6.h>
  50 #include <net/ip6_fib.h>
  51 #include <net/ip6_route.h>
  52 #include <net/ndisc.h>
  53 #include <net/addrconf.h>
  54 #include <net/tcp.h>
  55 #include <linux/rtnetlink.h>
  56 #include <net/dst.h>
  57 #include <net/xfrm.h>
  58 #include <net/netevent.h>
  59 #include <net/netlink.h>
  60 #include <net/nexthop.h>
  61
  62 #include <asm/uaccess.h>
  63
  64 #ifdef CONFIG_SYSCTL
  65 #include <linux/sysctl.h>
  66 #endif
  67
  68 enum rt6_nud_state {
  69         RT6_NUD_FAIL_HARD = -3,
  70         RT6_NUD_FAIL_PROBE = -2,
  71         RT6_NUD_FAIL_DO_RR = -1,
  72         RT6_NUD_SUCCEED = 1
  73 };
  74
  75 static struct rt6_info *ip6_rt_copy(struct rt6_info *ort,
  76                                     const struct in6_addr *dest);
  77 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie);
  78 static unsigned int      ip6_default_advmss(const struct dst_entry *dst);
  79 static unsigned int      ip6_mtu(const struct dst_entry *dst);
  80 static struct dst_entry *ip6_negative_advice(struct dst_entry *);
  81 static void             ip6_dst_destroy(struct dst_entry *);
  82 static void             ip6_dst_ifdown(struct dst_entry *,
  83                                        struct net_device *dev, int how);
  84 static int               ip6_dst_gc(struct dst_ops *ops);
  85
  86 static int              ip6_pkt_discard(struct sk_buff *skb);
  87 static int              ip6_pkt_discard_out(struct sock *sk, struct sk_buff *skb);
  88 static int              ip6_pkt_prohibit(struct sk_buff *skb);
  89 static int              ip6_pkt_prohibit_out(struct sock *sk, struct sk_buff *skb);
  90 static void             ip6_link_failure(struct sk_buff *skb);
  91 static void             ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
  92                                            struct sk_buff *skb, u32 mtu);
  93 static void             rt6_do_redirect(struct dst_entry *dst, struct sock *sk,
  94                                         struct sk_buff *skb);
  95 static int rt6_score_route(struct rt6_info *rt, int oif, int strict);
  96
  97 #ifdef CONFIG_IPV6_ROUTE_INFO
  98 static struct rt6_info *rt6_add_route_info(struct net *net,
  99                                            const struct in6_addr *prefix, int prefixlen,
 100                                            const struct in6_addr *gwaddr, int ifindex,
 101                                            unsigned int pref);
 102 static struct rt6_info *rt6_get_route_info(struct net *net,
 103                                            const struct in6_addr *prefix, int prefixlen,
 104                                            const struct in6_addr *gwaddr, int ifindex);
 105 #endif
 106
 107 static void rt6_bind_peer(struct rt6_info *rt, int create)
 108 {
 109         struct inet_peer_base *base;
 110         struct inet_peer *peer;
 111
 112         base = inetpeer_base_ptr(rt->_rt6i_peer);
 113         if (!base)
 114                 return;
 115
 116         peer = inet_getpeer_v6(base, &rt->rt6i_dst.addr, create);
 117         if (peer) {
 118                 if (!rt6_set_peer(rt, peer))
 119                         inet_putpeer(peer);
 120         }
 121 }
 122
 123 static struct inet_peer *__rt6_get_peer(struct rt6_info *rt, int create)
 124 {
 125         if (rt6_has_peer(rt))
 126                 return rt6_peer_ptr(rt);
 127
 128         rt6_bind_peer(rt, create);
 129         return (rt6_has_peer(rt) ? rt6_peer_ptr(rt) : NULL);
 130 }
 131
 132 static struct inet_peer *rt6_get_peer_create(struct rt6_info *rt)
 133 {
 134         return __rt6_get_peer(rt, 1);
 135 }
 136
 137 static u32 *ipv6_cow_metrics(struct dst_entry *dst, unsigned long old)
 138 {
 139         struct rt6_info *rt = (struct rt6_info *) dst;
 140         struct inet_peer *peer;
 141         u32 *p = NULL;
 142
 143         if (!(rt->dst.flags & DST_HOST))
 144                 return NULL;
 145
 146         peer = rt6_get_peer_create(rt);
 147         if (peer) {
 148                 u32 *old_p = __DST_METRICS_PTR(old);
 149                 unsigned long prev, new;
 150
 151                 p = peer->metrics;
 152                 if (inet_metrics_new(peer) ||
 153                     (old & DST_METRICS_FORCE_OVERWRITE))
 154                         memcpy(p, old_p, sizeof(u32) * RTAX_MAX);
 155
 156                 new = (unsigned long) p;
 157                 prev = cmpxchg(&dst->_metrics, old, new);
 158
 159                 if (prev != old) {
 160                         p = __DST_METRICS_PTR(prev);
 161                         if (prev & DST_METRICS_READ_ONLY)
 162                                 p = NULL;
 163                 }
 164         }
 165         return p;
 166 }
 167
 168 static inline const void *choose_neigh_daddr(struct rt6_info *rt,
 169                                              struct sk_buff *skb,
 170                                              const void *daddr)
 171 {
 172         struct in6_addr *p = &rt->rt6i_gateway;
 173
 174         if (!ipv6_addr_any(p))
 175                 return (const void *) p;
 176         else if (skb)
 177                 return &ipv6_hdr(skb)->daddr;
 178         return daddr;
 179 }
 180
 181 static struct neighbour *ip6_neigh_lookup(const struct dst_entry *dst,
 182                                           struct sk_buff *skb,
 183                                           const void *daddr)
 184 {
 185         struct rt6_info *rt = (struct rt6_info *) dst;
 186         struct neighbour *n;
 187
 188         daddr = choose_neigh_daddr(rt, skb, daddr);
 189         n = __ipv6_neigh_lookup(dst->dev, daddr);
 190         if (n)
 191                 return n;
 192         return neigh_create(&nd_tbl, daddr, dst->dev);
 193 }
 194
 195 static struct dst_ops ip6_dst_ops_template = {
 196         .family                 =       AF_INET6,
 197         .protocol               =       cpu_to_be16(ETH_P_IPV6),
 198         .gc                     =       ip6_dst_gc,
 199         .gc_thresh              =       1024,
 200         .check                  =       ip6_dst_check,
 201         .default_advmss         =       ip6_default_advmss,
 202         .mtu                    =       ip6_mtu,
 203         .cow_metrics            =       ipv6_cow_metrics,
 204         .destroy                =       ip6_dst_destroy,
 205         .ifdown                 =       ip6_dst_ifdown,
 206         .negative_advice        =       ip6_negative_advice,
 207         .link_failure           =       ip6_link_failure,
 208         .update_pmtu            =       ip6_rt_update_pmtu,
 209         .redirect               =       rt6_do_redirect,
 210         .local_out              =       __ip6_local_out,
 211         .neigh_lookup           =       ip6_neigh_lookup,
 212 };
 213
 214 static unsigned int ip6_blackhole_mtu(const struct dst_entry *dst)
 215 {
 216         unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
 217
 218         return mtu ? : dst->dev->mtu;
 219 }
 220
 221 static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, struct sock *sk,
 222                                          struct sk_buff *skb, u32 mtu)
 223 {
 224 }
 225
 226 static void ip6_rt_blackhole_redirect(struct dst_entry *dst, struct sock *sk,
 227                                       struct sk_buff *skb)
 228 {
 229 }
 230
 231 static u32 *ip6_rt_blackhole_cow_metrics(struct dst_entry *dst,
 232                                          unsigned long old)
 233 {
 234         return NULL;
 235 }
 236
 237 static struct dst_ops ip6_dst_blackhole_ops = {
 238         .family                 =       AF_INET6,
 239         .protocol               =       cpu_to_be16(ETH_P_IPV6),
 240         .destroy                =       ip6_dst_destroy,
 241         .check                  =       ip6_dst_check,
 242         .mtu                    =       ip6_blackhole_mtu,
 243         .default_advmss         =       ip6_default_advmss,
 244         .update_pmtu            =       ip6_rt_blackhole_update_pmtu,
 245         .redirect               =       ip6_rt_blackhole_redirect,
 246         .cow_metrics            =       ip6_rt_blackhole_cow_metrics,
 247         .neigh_lookup           =       ip6_neigh_lookup,
 248 };
 249
 250 static const u32 ip6_template_metrics[RTAX_MAX] = {
 251         [RTAX_HOPLIMIT - 1] = 0,
 252 };
 253
 254 static const struct rt6_info ip6_null_entry_template = {
 255         .dst = {
 256                 .__refcnt       = ATOMIC_INIT(1),
 257                 .__use          = 1,
 258                 .obsolete       = DST_OBSOLETE_FORCE_CHK,
 259                 .error          = -ENETUNREACH,
 260                 .input          = ip6_pkt_discard,
 261                 .output         = ip6_pkt_discard_out,
 262         },
 263         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
 264         .rt6i_protocol  = RTPROT_KERNEL,
 265         .rt6i_metric    = ~(u32) 0,
 266         .rt6i_ref       = ATOMIC_INIT(1),
 267 };
 268
 269 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
 270
 271 static const struct rt6_info ip6_prohibit_entry_template = {
 272         .dst = {
 273                 .__refcnt       = ATOMIC_INIT(1),
 274                 .__use          = 1,
 275                 .obsolete       = DST_OBSOLETE_FORCE_CHK,
 276                 .error          = -EACCES,
 277                 .input          = ip6_pkt_prohibit,
 278                 .output         = ip6_pkt_prohibit_out,
 279         },
 280         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
 281         .rt6i_protocol  = RTPROT_KERNEL,
 282         .rt6i_metric    = ~(u32) 0,
 283         .rt6i_ref       = ATOMIC_INIT(1),
 284 };
 285
 286 static const struct rt6_info ip6_blk_hole_entry_template = {
 287         .dst = {
 288                 .__refcnt       = ATOMIC_INIT(1),
 289                 .__use          = 1,
 290                 .obsolete       = DST_OBSOLETE_FORCE_CHK,
 291                 .error          = -EINVAL,
 292                 .input          = dst_discard,
 293                 .output         = dst_discard_sk,
 294         },
 295         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
 296         .rt6i_protocol  = RTPROT_KERNEL,
 297         .rt6i_metric    = ~(u32) 0,
 298         .rt6i_ref       = ATOMIC_INIT(1),
 299 };
 300
 301 #endif
 302
 303 /* allocate dst with ip6_dst_ops */
 304 static inline struct rt6_info *ip6_dst_alloc(struct net *net,
 305                                              struct net_device *dev,
 306                                              int flags,
 307                                              struct fib6_table *table)
 308 {
 309         struct rt6_info *rt = dst_alloc(&net->ipv6.ip6_dst_ops, dev,
 310                                         0, DST_OBSOLETE_FORCE_CHK, flags);
 311
 312         if (rt) {
 313                 struct dst_entry *dst = &rt->dst;
 314
 315                 memset(dst + 1, 0, sizeof(*rt) - sizeof(*dst));
 316                 rt6_init_peer(rt, table ? &table->tb6_peers : net->ipv6.peers);
 317                 INIT_LIST_HEAD(&rt->rt6i_siblings);
 318         }
 319         return rt;
 320 }
 321
 322 static void ip6_dst_destroy(struct dst_entry *dst)
 323 {
 324         struct rt6_info *rt = (struct rt6_info *)dst;
 325         struct inet6_dev *idev = rt->rt6i_idev;
 326         struct dst_entry *from = dst->from;
 327
 328         if (!(rt->dst.flags & DST_HOST))
 329                 dst_destroy_metrics_generic(dst);
 330
 331         if (idev) {
 332                 rt->rt6i_idev = NULL;
 333                 in6_dev_put(idev);
 334         }
 335
 336         dst->from = NULL;
 337         dst_release(from);
 338
 339         if (rt6_has_peer(rt)) {
 340                 struct inet_peer *peer = rt6_peer_ptr(rt);
 341                 inet_putpeer(peer);
 342         }
 343 }
 344
 345 static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
 346                            int how)
 347 {
 348         struct rt6_info *rt = (struct rt6_info *)dst;
 349         struct inet6_dev *idev = rt->rt6i_idev;
 350         struct net_device *loopback_dev =
 351                 dev_net(dev)->loopback_dev;
 352
 353         if (dev != loopback_dev) {
 354                 if (idev && idev->dev == dev) {
 355                         struct inet6_dev *loopback_idev =
 356                                 in6_dev_get(loopback_dev);
 357                         if (loopback_idev) {
 358                                 rt->rt6i_idev = loopback_idev;
 359                                 in6_dev_put(idev);
 360                         }
 361                 }
 362         }
 363 }
 364
 365 static bool rt6_check_expired(const struct rt6_info *rt)
 366 {
 367         if (rt->rt6i_flags & RTF_EXPIRES) {
 368                 if (time_after(jiffies, rt->dst.expires))
 369                         return true;
 370         } else if (rt->dst.from) {
 371                 return rt6_check_expired((struct rt6_info *) rt->dst.from);
 372         }
 373         return false;
 374 }
 375
 376 /* Multipath route selection:
 377  *   Hash based function using packet header and flowlabel.
 378  * Adapted from fib_info_hashfn()
 379  */
 380 static int rt6_info_hash_nhsfn(unsigned int candidate_count,
 381                                const struct flowi6 *fl6)
 382 {
 383         unsigned int val = fl6->flowi6_proto;
 384
 385         val ^= ipv6_addr_hash(&fl6->daddr);
 386         val ^= ipv6_addr_hash(&fl6->saddr);
 387
 388         /* Work only if this not encapsulated */
 389         switch (fl6->flowi6_proto) {
 390         case IPPROTO_UDP:
 391         case IPPROTO_TCP:
 392         case IPPROTO_SCTP:
 393                 val ^= (__force u16)fl6->fl6_sport;
 394                 val ^= (__force u16)fl6->fl6_dport;
 395                 break;
 396
 397         case IPPROTO_ICMPV6:
 398                 val ^= (__force u16)fl6->fl6_icmp_type;
 399                 val ^= (__force u16)fl6->fl6_icmp_code;
 400                 break;
 401         }
 402         /* RFC6438 recommands to use flowlabel */
 403         val ^= (__force u32)fl6->flowlabel;
 404
 405         /* Perhaps, we need to tune, this function? */
 406         val = val ^ (val >> 7) ^ (val >> 12);
 407         return val % candidate_count;
 408 }
 409
 410 static struct rt6_info *rt6_multipath_select(struct rt6_info *match,
 411                                              struct flowi6 *fl6, int oif,
 412                                              int strict)
 413 {
 414         struct rt6_info *sibling, *next_sibling;
 415         int route_choosen;
 416
 417         route_choosen = rt6_info_hash_nhsfn(match->rt6i_nsiblings + 1, fl6);
 418         /* Don't change the route, if route_choosen == 0
 419          * (siblings does not include ourself)
 420          */
 421         if (route_choosen)
 422                 list_for_each_entry_safe(sibling, next_sibling,
 423                                 &match->rt6i_siblings, rt6i_siblings) {
 424                         route_choosen--;
 425                         if (route_choosen == 0) {
 426                                 if (rt6_score_route(sibling, oif, strict) < 0)
 427                                         break;
 428                                 match = sibling;
 429                                 break;
 430                         }
 431                 }
 432         return match;
 433 }
 434
 435 /*
 436  *      Route lookup. Any table->tb6_lock is implied.
 437  */
 438
 439 static inline struct rt6_info *rt6_device_match(struct net *net,
 440                                                     struct rt6_info *rt,
 441                                                     const struct in6_addr *saddr,
 442                                                     int oif,
 443                                                     int flags)
 444 {
 445         struct rt6_info *local = NULL;
 446         struct rt6_info *sprt;
 447
 448         if (!oif && ipv6_addr_any(saddr))
 449                 goto out;
 450
 451         for (sprt = rt; sprt; sprt = sprt->dst.rt6_next) {
 452                 struct net_device *dev = sprt->dst.dev;
 453
 454                 if (oif) {
 455                         if (dev->ifindex == oif)
 456                                 return sprt;
 457                         if (dev->flags & IFF_LOOPBACK) {
 458                                 if (!sprt->rt6i_idev ||
 459                                     sprt->rt6i_idev->dev->ifindex != oif) {
 460                                         if (flags & RT6_LOOKUP_F_IFACE && oif)
 461                                                 continue;
 462                                         if (local && (!oif ||
 463                                                       local->rt6i_idev->dev->ifindex == oif))
 464                                                 continue;
 465                                 }
 466                                 local = sprt;
 467                         }
 468                 } else {
 469                         if (ipv6_chk_addr(net, saddr, dev,
 470                                           flags & RT6_LOOKUP_F_IFACE))
 471                                 return sprt;
 472                 }
 473         }
 474
 475         if (oif) {
 476                 if (local)
 477                         return local;
 478
 479                 if (flags & RT6_LOOKUP_F_IFACE)
 480                         return net->ipv6.ip6_null_entry;
 481         }
 482 out:
 483         return rt;
 484 }
 485
 486 #ifdef CONFIG_IPV6_ROUTER_PREF
 487 struct __rt6_probe_work {
 488         struct work_struct work;
 489         struct in6_addr target;
 490         struct net_device *dev;
 491 };
 492
 493 static void rt6_probe_deferred(struct work_struct *w)
 494 {
 495         struct in6_addr mcaddr;
 496         struct __rt6_probe_work *work =
 497                 container_of(w, struct __rt6_probe_work, work);
 498
 499         addrconf_addr_solict_mult(&work->target, &mcaddr);
 500         ndisc_send_ns(work->dev, NULL, &work->target, &mcaddr, NULL);
 501         dev_put(work->dev);
 502         kfree(w);
 503 }
 504
 505 static void rt6_probe(struct rt6_info *rt)
 506 {
 507         struct neighbour *neigh;
 508         /*
 509          * Okay, this does not seem to be appropriate
 510          * for now, however, we need to check if it
 511          * is really so; aka Router Reachability Probing.
 512          *
 513          * Router Reachability Probe MUST be rate-limited
 514          * to no more than one per minute.
 515          */
 516         if (!rt || !(rt->rt6i_flags & RTF_GATEWAY))
 517                 return;
 518         rcu_read_lock_bh();
 519         neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
 520         if (neigh) {
 521                 write_lock(&neigh->lock);
 522                 if (neigh->nud_state & NUD_VALID)
 523                         goto out;
 524         }
 525
 526         if (!neigh ||
 527             time_after(jiffies, neigh->updated + rt->rt6i_idev->cnf.rtr_probe_interval)) {
 528                 struct __rt6_probe_work *work;
 529
 530                 work = kmalloc(sizeof(*work), GFP_ATOMIC);
 531
 532                 if (neigh && work)
 533                         __neigh_set_probe_once(neigh);
 534
 535                 if (neigh)
 536                         write_unlock(&neigh->lock);
 537
 538                 if (work) {
 539                         INIT_WORK(&work->work, rt6_probe_deferred);
 540                         work->target = rt->rt6i_gateway;
 541                         dev_hold(rt->dst.dev);
 542                         work->dev = rt->dst.dev;
 543                         schedule_work(&work->work);
 544                 }
 545         } else {
 546 out:
 547                 write_unlock(&neigh->lock);
 548         }
 549         rcu_read_unlock_bh();
 550 }
 551 #else
 552 static inline void rt6_probe(struct rt6_info *rt)
 553 {
 554 }
 555 #endif
 556
 557 /*
 558  * Default Router Selection (RFC 2461 6.3.6)
 559  */
 560 static inline int rt6_check_dev(struct rt6_info *rt, int oif)
 561 {
 562         struct net_device *dev = rt->dst.dev;
 563         if (!oif || dev->ifindex == oif)
 564                 return 2;
 565         if ((dev->flags & IFF_LOOPBACK) &&
 566             rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
 567                 return 1;
 568         return 0;
 569 }
 570
 571 static inline enum rt6_nud_state rt6_check_neigh(struct rt6_info *rt)
 572 {
 573         struct neighbour *neigh;
 574         enum rt6_nud_state ret = RT6_NUD_FAIL_HARD;
 575
 576         if (rt->rt6i_flags & RTF_NONEXTHOP ||
 577             !(rt->rt6i_flags & RTF_GATEWAY))
 578                 return RT6_NUD_SUCCEED;
 579
 580         rcu_read_lock_bh();
 581         neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
 582         if (neigh) {
 583                 read_lock(&neigh->lock);
 584                 if (neigh->nud_state & NUD_VALID)
 585                         ret = RT6_NUD_SUCCEED;
 586 #ifdef CONFIG_IPV6_ROUTER_PREF
 587                 else if (!(neigh->nud_state & NUD_FAILED))
 588                         ret = RT6_NUD_SUCCEED;
 589                 else
 590                         ret = RT6_NUD_FAIL_PROBE;
 591 #endif
 592                 read_unlock(&neigh->lock);
 593         } else {
 594                 ret = IS_ENABLED(CONFIG_IPV6_ROUTER_PREF) ?
 595                       RT6_NUD_SUCCEED : RT6_NUD_FAIL_DO_RR;
 596         }
 597         rcu_read_unlock_bh();
 598
 599         return ret;
 600 }
 601
 602 static int rt6_score_route(struct rt6_info *rt, int oif,
 603                            int strict)
 604 {
 605         int m;
 606
 607         m = rt6_check_dev(rt, oif);
 608         if (!m && (strict & RT6_LOOKUP_F_IFACE))
 609                 return RT6_NUD_FAIL_HARD;
 610 #ifdef CONFIG_IPV6_ROUTER_PREF
 611         m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
 612 #endif
 613         if (strict & RT6_LOOKUP_F_REACHABLE) {
 614                 int n = rt6_check_neigh(rt);
 615                 if (n < 0)
 616                         return n;
 617         }
 618         return m;
 619 }
 620
 621 static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
 622                                    int *mpri, struct rt6_info *match,
 623                                    bool *do_rr)
 624 {
 625         int m;
 626         bool match_do_rr = false;
 627
 628         if (rt6_check_expired(rt))
 629                 goto out;
 630
 631         m = rt6_score_route(rt, oif, strict);
 632         if (m == RT6_NUD_FAIL_DO_RR) {
 633                 match_do_rr = true;
 634                 m = 0; /* lowest valid score */
 635         } else if (m == RT6_NUD_FAIL_HARD) {
 636                 goto out;
 637         }
 638
 639         if (strict & RT6_LOOKUP_F_REACHABLE)
 640                 rt6_probe(rt);
 641
 642         /* note that m can be RT6_NUD_FAIL_PROBE at this point */
 643         if (m > *mpri) {
 644                 *do_rr = match_do_rr;
 645                 *mpri = m;
 646                 match = rt;
 647         }
 648 out:
 649         return match;
 650 }
 651
 652 static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
 653                                      struct rt6_info *rr_head,
 654                                      u32 metric, int oif, int strict,
 655                                      bool *do_rr)
 656 {
 657         struct rt6_info *rt, *match;
 658         int mpri = -1;
 659
 660         match = NULL;
 661         for (rt = rr_head; rt && rt->rt6i_metric == metric;
 662              rt = rt->dst.rt6_next)
 663                 match = find_match(rt, oif, strict, &mpri, match, do_rr);
 664         for (rt = fn->leaf; rt && rt != rr_head && rt->rt6i_metric == metric;
 665              rt = rt->dst.rt6_next)
 666                 match = find_match(rt, oif, strict, &mpri, match, do_rr);
 667
 668         return match;
 669 }
 670
 671 static struct rt6_info *rt6_select(struct fib6_node *fn, int oif, int strict)
 672 {
 673         struct rt6_info *match, *rt0;
 674         struct net *net;
 675         bool do_rr = false;
 676
 677         rt0 = fn->rr_ptr;
 678         if (!rt0)
 679                 fn->rr_ptr = rt0 = fn->leaf;
 680
 681         match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict,
 682                              &do_rr);
 683
 684         if (do_rr) {
 685                 struct rt6_info *next = rt0->dst.rt6_next;
 686
 687                 /* no entries matched; do round-robin */
 688                 if (!next || next->rt6i_metric != rt0->rt6i_metric)
 689                         next = fn->leaf;
 690
 691                 if (next != rt0)
 692                         fn->rr_ptr = next;
 693         }
 694
 695         net = dev_net(rt0->dst.dev);
 696         return match ? match : net->ipv6.ip6_null_entry;
 697 }
 698
 699 #ifdef CONFIG_IPV6_ROUTE_INFO
 700 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
 701                   const struct in6_addr *gwaddr)
 702 {
 703         struct net *net = dev_net(dev);
 704         struct route_info *rinfo = (struct route_info *) opt;
 705         struct in6_addr prefix_buf, *prefix;
 706         unsigned int pref;
 707         unsigned long lifetime;
 708         struct rt6_info *rt;
 709
 710         if (len < sizeof(struct route_info)) {
 711                 return -EINVAL;
 712         }
 713
 714         /* Sanity check for prefix_len and length */
 715         if (rinfo->length > 3) {
 716                 return -EINVAL;
 717         } else if (rinfo->prefix_len > 128) {
 718                 return -EINVAL;
 719         } else if (rinfo->prefix_len > 64) {
 720                 if (rinfo->length < 2) {
 721                         return -EINVAL;
 722                 }
 723         } else if (rinfo->prefix_len > 0) {
 724                 if (rinfo->length < 1) {
 725                         return -EINVAL;
 726                 }
 727         }
 728
 729         pref = rinfo->route_pref;
 730         if (pref == ICMPV6_ROUTER_PREF_INVALID)
 731                 return -EINVAL;
 732
 733         lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
 734
 735         if (rinfo->length == 3)
 736                 prefix = (struct in6_addr *)rinfo->prefix;
 737         else {
 738                 /* this function is safe */
 739                 ipv6_addr_prefix(&prefix_buf,
 740                                  (struct in6_addr *)rinfo->prefix,
 741                                  rinfo->prefix_len);
 742                 prefix = &prefix_buf;
 743         }
 744
 745         if (rinfo->prefix_len == 0)
 746                 rt = rt6_get_dflt_router(gwaddr, dev);
 747         else
 748                 rt = rt6_get_route_info(net, prefix, rinfo->prefix_len,
 749                                         gwaddr, dev->ifindex);
 750
 751         if (rt && !lifetime) {
 752                 ip6_del_rt(rt);
 753                 rt = NULL;
 754         }
 755
 756         if (!rt && lifetime)
 757                 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr, dev->ifindex,
 758                                         pref);
 759         else if (rt)
 760                 rt->rt6i_flags = RTF_ROUTEINFO |
 761                                  (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
 762
 763         if (rt) {
 764                 if (!addrconf_finite_timeout(lifetime))
 765                         rt6_clean_expires(rt);
 766                 else
 767                         rt6_set_expires(rt, jiffies + HZ * lifetime);
 768
 769                 ip6_rt_put(rt);
 770         }
 771         return 0;
 772 }
 773 #endif
 774
 775 #define BACKTRACK(__net, saddr)                 \
 776 do { \
 777         if (rt == __net->ipv6.ip6_null_entry) { \
 778                 struct fib6_node *pn; \
 779                 while (1) { \
 780                         if (fn->fn_flags & RTN_TL_ROOT) \
 781                                 goto out; \
 782                         pn = fn->parent; \
 783                         if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn) \
 784                                 fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr); \
 785                         else \
 786                                 fn = pn; \
 787                         if (fn->fn_flags & RTN_RTINFO) \
 788                                 goto restart; \
 789                 } \
 790         } \
 791 } while (0)
 792
 793 static struct rt6_info *ip6_pol_route_lookup(struct net *net,
 794                                              struct fib6_table *table,
 795                                              struct flowi6 *fl6, int flags)
 796 {
 797         struct fib6_node *fn;
 798         struct rt6_info *rt;
 799
 800         read_lock_bh(&table->tb6_lock);
 801         fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
 802 restart:
 803         rt = fn->leaf;
 804         rt = rt6_device_match(net, rt, &fl6->saddr, fl6->flowi6_oif, flags);
 805         if (rt->rt6i_nsiblings && fl6->flowi6_oif == 0)
 806                 rt = rt6_multipath_select(rt, fl6, fl6->flowi6_oif, flags);
 807         BACKTRACK(net, &fl6->saddr);
 808 out:
 809         dst_use(&rt->dst, jiffies);
 810         read_unlock_bh(&table->tb6_lock);
 811         return rt;
 812
 813 }
 814
 815 struct dst_entry *ip6_route_lookup(struct net *net, struct flowi6 *fl6,
 816                                     int flags)
 817 {
 818         return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_lookup);
 819 }
 820 EXPORT_SYMBOL_GPL(ip6_route_lookup);
 821
 822 struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
 823                             const struct in6_addr *saddr, int oif, int strict)
 824 {
 825         struct flowi6 fl6 = {
 826                 .flowi6_oif = oif,
 827                 .daddr = *daddr,
 828         };
 829         struct dst_entry *dst;
 830         int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
 831
 832         if (saddr) {
 833                 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
 834                 flags |= RT6_LOOKUP_F_HAS_SADDR;
 835         }
 836
 837         dst = fib6_rule_lookup(net, &fl6, flags, ip6_pol_route_lookup);
 838         if (dst->error == 0)
 839                 return (struct rt6_info *) dst;
 840
 841         dst_release(dst);
 842
 843         return NULL;
 844 }
 845 EXPORT_SYMBOL(rt6_lookup);
 846
 847 /* ip6_ins_rt is called with FREE table->tb6_lock.
 848    It takes new route entry, the addition fails by any reason the
 849    route is freed. In any case, if caller does not hold it, it may
 850    be destroyed.
 851  */
 852
 853 static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info,
 854                         struct nlattr *mx, int mx_len)
 855 {
 856         int err;
 857         struct fib6_table *table;
 858
 859         table = rt->rt6i_table;
 860         write_lock_bh(&table->tb6_lock);
 861         err = fib6_add(&table->tb6_root, rt, info, mx, mx_len);
 862         write_unlock_bh(&table->tb6_lock);
 863
 864         return err;
 865 }
 866
 867 int ip6_ins_rt(struct rt6_info *rt)
 868 {
 869         struct nl_info info = {
 870                 .nl_net = dev_net(rt->dst.dev),
 871         };
 872         return __ip6_ins_rt(rt, &info, NULL, 0);
 873 }
 874
 875 static struct rt6_info *rt6_alloc_cow(struct rt6_info *ort,
 876                                       const struct in6_addr *daddr,
 877                                       const struct in6_addr *saddr)
 878 {
 879         struct rt6_info *rt;
 880
 881         /*
 882          *      Clone the route.
 883          */
 884
 885         rt = ip6_rt_copy(ort, daddr);
 886
 887         if (rt) {
 888                 if (ort->rt6i_dst.plen != 128 &&
 889                     ipv6_addr_equal(&ort->rt6i_dst.addr, daddr))
 890                         rt->rt6i_flags |= RTF_ANYCAST;
 891
 892                 rt->rt6i_flags |= RTF_CACHE;
 893
 894 #ifdef CONFIG_IPV6_SUBTREES
 895                 if (rt->rt6i_src.plen && saddr) {
 896                         rt->rt6i_src.addr = *saddr;
 897                         rt->rt6i_src.plen = 128;
 898                 }
 899 #endif
 900         }
 901
 902         return rt;
 903 }
 904
 905 static struct rt6_info *rt6_alloc_clone(struct rt6_info *ort,
 906                                         const struct in6_addr *daddr)
 907 {
 908         struct rt6_info *rt = ip6_rt_copy(ort, daddr);
 909
 910         if (rt)
 911                 rt->rt6i_flags |= RTF_CACHE;
 912         return rt;
 913 }
 914
 915 static struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table, int oif,
 916                                       struct flowi6 *fl6, int flags)
 917 {
 918         struct fib6_node *fn;
 919         struct rt6_info *rt, *nrt;
 920         int strict = 0;
 921         int attempts = 3;
 922         int err;
 923         int reachable = net->ipv6.devconf_all->forwarding ? 0 : RT6_LOOKUP_F_REACHABLE;
 924
 925         strict |= flags & RT6_LOOKUP_F_IFACE;
 926
 927 relookup:
 928         read_lock_bh(&table->tb6_lock);
 929
 930 restart_2:
 931         fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
 932
 933 restart:
 934         rt = rt6_select(fn, oif, strict | reachable);
 935         if (rt->rt6i_nsiblings)
 936                 rt = rt6_multipath_select(rt, fl6, oif, strict | reachable);
 937         BACKTRACK(net, &fl6->saddr);
 938         if (rt == net->ipv6.ip6_null_entry ||
 939             rt->rt6i_flags & RTF_CACHE)
 940                 goto out;
 941
 942         dst_hold(&rt->dst);
 943         read_unlock_bh(&table->tb6_lock);
 944
 945         if (!(rt->rt6i_flags & (RTF_NONEXTHOP | RTF_GATEWAY)))
 946                 nrt = rt6_alloc_cow(rt, &fl6->daddr, &fl6->saddr);
 947         else if (!(rt->dst.flags & DST_HOST))
 948                 nrt = rt6_alloc_clone(rt, &fl6->daddr);
 949         else
 950                 goto out2;
 951
 952         ip6_rt_put(rt);
 953         rt = nrt ? : net->ipv6.ip6_null_entry;
 954
 955         dst_hold(&rt->dst);
 956         if (nrt) {
 957                 err = ip6_ins_rt(nrt);
 958                 if (!err)
 959                         goto out2;
 960         }
 961
 962         if (--attempts <= 0)
 963                 goto out2;
 964
 965         /*
 966          * Race condition! In the gap, when table->tb6_lock was
 967          * released someone could insert this route.  Relookup.
 968          */
 969         ip6_rt_put(rt);
 970         goto relookup;
 971
 972 out:
 973         if (reachable) {
 974                 reachable = 0;
 975                 goto restart_2;
 976         }
 977         dst_hold(&rt->dst);
 978         read_unlock_bh(&table->tb6_lock);
 979 out2:
 980         rt->dst.lastuse = jiffies;
 981         rt->dst.__use++;
 982
 983         return rt;
 984 }
 985
 986 static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
 987                                             struct flowi6 *fl6, int flags)
 988 {
 989         return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, flags);
 990 }
 991
 992 static struct dst_entry *ip6_route_input_lookup(struct net *net,
 993                                                 struct net_device *dev,
 994                                                 struct flowi6 *fl6, int flags)
 995 {
 996         if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
 997                 flags |= RT6_LOOKUP_F_IFACE;
 998
 999         return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_input);
1000 }
1001
1002 void ip6_route_input(struct sk_buff *skb)
1003 {
1004         const struct ipv6hdr *iph = ipv6_hdr(skb);
1005         struct net *net = dev_net(skb->dev);
1006         int flags = RT6_LOOKUP_F_HAS_SADDR;
1007         struct flowi6 fl6 = {
1008                 .flowi6_iif = skb->dev->ifindex,
1009                 .daddr = iph->daddr,
1010                 .saddr = iph->saddr,
1011                 .flowlabel = ip6_flowinfo(iph),
1012                 .flowi6_mark = skb->mark,
1013                 .flowi6_proto = iph->nexthdr,
1014         };
1015
1016         skb_dst_set(skb, ip6_route_input_lookup(net, skb->dev, &fl6, flags));
1017 }
1018
1019 static struct rt6_info *ip6_pol_route_output(struct net *net, struct fib6_table *table,
1020                                              struct flowi6 *fl6, int flags)
1021 {
1022         return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, flags);
1023 }
1024
1025 struct dst_entry *ip6_route_output(struct net *net, const struct sock *sk,
1026                                     struct flowi6 *fl6)
1027 {
1028         int flags = 0;
1029
1030         fl6->flowi6_iif = LOOPBACK_IFINDEX;
1031
1032         if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr))
1033                 flags |= RT6_LOOKUP_F_IFACE;
1034
1035         if (!ipv6_addr_any(&fl6->saddr))
1036                 flags |= RT6_LOOKUP_F_HAS_SADDR;
1037         else if (sk)
1038                 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
1039
1040         return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_output);
1041 }
1042 EXPORT_SYMBOL(ip6_route_output);
1043
1044 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
1045 {
1046         struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
1047         struct dst_entry *new = NULL;
1048
1049         rt = dst_alloc(&ip6_dst_blackhole_ops, ort->dst.dev, 1, DST_OBSOLETE_NONE, 0);
1050         if (rt) {
1051                 new = &rt->dst;
1052
1053                 memset(new + 1, 0, sizeof(*rt) - sizeof(*new));
1054                 rt6_init_peer(rt, net->ipv6.peers);
1055
1056                 new->__use = 1;
1057                 new->input = dst_discard;
1058                 new->output = dst_discard_sk;
1059
1060                 if (dst_metrics_read_only(&ort->dst))
1061                         new->_metrics = ort->dst._metrics;
1062                 else
1063                         dst_copy_metrics(new, &ort->dst);
1064                 rt->rt6i_idev = ort->rt6i_idev;
1065                 if (rt->rt6i_idev)
1066                         in6_dev_hold(rt->rt6i_idev);
1067
1068                 rt->rt6i_gateway = ort->rt6i_gateway;
1069                 rt->rt6i_flags = ort->rt6i_flags;
1070                 rt->rt6i_metric = 0;
1071
1072                 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1073 #ifdef CONFIG_IPV6_SUBTREES
1074                 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1075 #endif
1076
1077                 dst_free(new);
1078         }
1079
1080         dst_release(dst_orig);
1081         return new ? new : ERR_PTR(-ENOMEM);
1082 }
1083
1084 /*
1085  *      Destination cache support functions
1086  */
1087
1088 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
1089 {
1090         struct rt6_info *rt;
1091
1092         rt = (struct rt6_info *) dst;
1093
1094         /* All IPV6 dsts are created with ->obsolete set to the value
1095          * DST_OBSOLETE_FORCE_CHK which forces validation calls down
1096          * into this function always.
1097          */
1098         if (!rt->rt6i_node || (rt->rt6i_node->fn_sernum != cookie))
1099                 return NULL;
1100
1101         if (rt6_check_expired(rt))
1102                 return NULL;
1103
1104         return dst;
1105 }
1106
1107 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
1108 {
1109         struct rt6_info *rt = (struct rt6_info *) dst;
1110
1111         if (rt) {
1112                 if (rt->rt6i_flags & RTF_CACHE) {
1113                         if (rt6_check_expired(rt)) {
1114                                 ip6_del_rt(rt);
1115                                 dst = NULL;
1116                         }
1117                 } else {
1118                         dst_release(dst);
1119                         dst = NULL;
1120                 }
1121         }
1122         return dst;
1123 }
1124
1125 static void ip6_link_failure(struct sk_buff *skb)
1126 {
1127         struct rt6_info *rt;
1128
1129         icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
1130
1131         rt = (struct rt6_info *) skb_dst(skb);
1132         if (rt) {
1133                 if (rt->rt6i_flags & RTF_CACHE) {
1134                         dst_hold(&rt->dst);
1135                         if (ip6_del_rt(rt))
1136                                 dst_free(&rt->dst);
1137                 } else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT)) {
1138                         rt->rt6i_node->fn_sernum = -1;
1139                 }
1140         }
1141 }
1142
1143 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
1144                                struct sk_buff *skb, u32 mtu)
1145 {
1146         struct rt6_info *rt6 = (struct rt6_info *)dst;
1147
1148         dst_confirm(dst);
1149         if (mtu < dst_mtu(dst) && rt6->rt6i_dst.plen == 128) {
1150                 struct net *net = dev_net(dst->dev);
1151
1152                 rt6->rt6i_flags |= RTF_MODIFIED;
1153                 if (mtu < IPV6_MIN_MTU) {
1154                         u32 features = dst_metric(dst, RTAX_FEATURES);
1155                         mtu = IPV6_MIN_MTU;
1156                         features |= RTAX_FEATURE_ALLFRAG;
1157                         dst_metric_set(dst, RTAX_FEATURES, features);
1158                 }
1159                 dst_metric_set(dst, RTAX_MTU, mtu);
1160                 rt6_update_expires(rt6, net->ipv6.sysctl.ip6_rt_mtu_expires);
1161         }
1162 }
1163
1164 void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
1165                      int oif, u32 mark)
1166 {
1167         const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1168         struct dst_entry *dst;
1169         struct flowi6 fl6;
1170
1171         memset(&fl6, 0, sizeof(fl6));
1172         fl6.flowi6_oif = oif;
1173         fl6.flowi6_mark = mark ? mark : IP6_REPLY_MARK(net, skb->mark);
1174         fl6.daddr = iph->daddr;
1175         fl6.saddr = iph->saddr;
1176         fl6.flowlabel = ip6_flowinfo(iph);
1177
1178         dst = ip6_route_output(net, NULL, &fl6);
1179         if (!dst->error)
1180                 ip6_rt_update_pmtu(dst, NULL, skb, ntohl(mtu));
1181         dst_release(dst);
1182 }
1183 EXPORT_SYMBOL_GPL(ip6_update_pmtu);
1184
1185 void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
1186 {
1187         ip6_update_pmtu(skb, sock_net(sk), mtu,
1188                         sk->sk_bound_dev_if, sk->sk_mark);
1189 }
1190 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
1191
1192 /* Handle redirects */
1193 struct ip6rd_flowi {
1194         struct flowi6 fl6;
1195         struct in6_addr gateway;
1196 };
1197
1198 static struct rt6_info *__ip6_route_redirect(struct net *net,
1199                                              struct fib6_table *table,
1200                                              struct flowi6 *fl6,
1201                                              int flags)
1202 {
1203         struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
1204         struct rt6_info *rt;
1205         struct fib6_node *fn;
1206
1207         /* Get the "current" route for this destination and
1208          * check if the redirect has come from approriate router.
1209          *
1210          * RFC 4861 specifies that redirects should only be
1211          * accepted if they come from the nexthop to the target.
1212          * Due to the way the routes are chosen, this notion
1213          * is a bit fuzzy and one might need to check all possible
1214          * routes.
1215          */
1216
1217         read_lock_bh(&table->tb6_lock);
1218         fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1219 restart:
1220         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1221                 if (rt6_check_expired(rt))
1222                         continue;
1223                 if (rt->dst.error)
1224                         break;
1225                 if (!(rt->rt6i_flags & RTF_GATEWAY))
1226                         continue;
1227                 if (fl6->flowi6_oif != rt->dst.dev->ifindex)
1228                         continue;
1229                 if (!ipv6_addr_equal(&rdfl->gateway, &rt->rt6i_gateway))
1230                         continue;
1231                 break;
1232         }
1233
1234         if (!rt)
1235                 rt = net->ipv6.ip6_null_entry;
1236         else if (rt->dst.error) {
1237                 rt = net->ipv6.ip6_null_entry;
1238                 goto out;
1239         }
1240         BACKTRACK(net, &fl6->saddr);
1241 out:
1242         dst_hold(&rt->dst);
1243
1244         read_unlock_bh(&table->tb6_lock);
1245
1246         return rt;
1247 };
1248
1249 static struct dst_entry *ip6_route_redirect(struct net *net,
1250                                         const struct flowi6 *fl6,
1251                                         const struct in6_addr *gateway)
1252 {
1253         int flags = RT6_LOOKUP_F_HAS_SADDR;
1254         struct ip6rd_flowi rdfl;
1255
1256         rdfl.fl6 = *fl6;
1257         rdfl.gateway = *gateway;
1258
1259         return fib6_rule_lookup(net, &rdfl.fl6,
1260                                 flags, __ip6_route_redirect);
1261 }
1262
1263 void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark)
1264 {
1265         const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1266         struct dst_entry *dst;
1267         struct flowi6 fl6;
1268
1269         memset(&fl6, 0, sizeof(fl6));
1270         fl6.flowi6_iif = LOOPBACK_IFINDEX;
1271         fl6.flowi6_oif = oif;
1272         fl6.flowi6_mark = mark;
1273         fl6.daddr = iph->daddr;
1274         fl6.saddr = iph->saddr;
1275         fl6.flowlabel = ip6_flowinfo(iph);
1276
1277         dst = ip6_route_redirect(net, &fl6, &ipv6_hdr(skb)->saddr);
1278         rt6_do_redirect(dst, NULL, skb);
1279         dst_release(dst);
1280 }
1281 EXPORT_SYMBOL_GPL(ip6_redirect);
1282
1283 void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif,
1284                             u32 mark)
1285 {
1286         const struct ipv6hdr *iph = ipv6_hdr(skb);
1287         const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
1288         struct dst_entry *dst;
1289         struct flowi6 fl6;
1290
1291         memset(&fl6, 0, sizeof(fl6));
1292         fl6.flowi6_iif = LOOPBACK_IFINDEX;
1293         fl6.flowi6_oif = oif;
1294         fl6.flowi6_mark = mark;
1295         fl6.daddr = msg->dest;
1296         fl6.saddr = iph->daddr;
1297
1298         dst = ip6_route_redirect(net, &fl6, &iph->saddr);
1299         rt6_do_redirect(dst, NULL, skb);
1300         dst_release(dst);
1301 }
1302
1303 void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
1304 {
1305         ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark);
1306 }
1307 EXPORT_SYMBOL_GPL(ip6_sk_redirect);
1308
1309 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1310 {
1311         struct net_device *dev = dst->dev;
1312         unsigned int mtu = dst_mtu(dst);
1313         struct net *net = dev_net(dev);
1314
1315         mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
1316
1317         if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
1318                 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1319
1320         /*
1321          * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1322          * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1323          * IPV6_MAXPLEN is also valid and means: "any MSS,
1324          * rely only on pmtu discovery"
1325          */
1326         if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
1327                 mtu = IPV6_MAXPLEN;
1328         return mtu;
1329 }
1330
1331 static unsigned int ip6_mtu(const struct dst_entry *dst)
1332 {
1333         struct inet6_dev *idev;
1334         unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
1335
1336         if (mtu)
1337                 goto out;
1338
1339         mtu = IPV6_MIN_MTU;
1340
1341         rcu_read_lock();
1342         idev = __in6_dev_get(dst->dev);
1343         if (idev)
1344                 mtu = idev->cnf.mtu6;
1345         rcu_read_unlock();
1346
1347 out:
1348         return min_t(unsigned int, mtu, IP6_MAX_MTU);
1349 }
1350
1351 static struct dst_entry *icmp6_dst_gc_list;
1352 static DEFINE_SPINLOCK(icmp6_dst_lock);
1353
1354 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
1355                                   struct flowi6 *fl6)
1356 {
1357         struct dst_entry *dst;
1358         struct rt6_info *rt;
1359         struct inet6_dev *idev = in6_dev_get(dev);
1360         struct net *net = dev_net(dev);
1361
1362         if (unlikely(!idev))
1363                 return ERR_PTR(-ENODEV);
1364
1365         rt = ip6_dst_alloc(net, dev, 0, NULL);
1366         if (unlikely(!rt)) {
1367                 in6_dev_put(idev);
1368                 dst = ERR_PTR(-ENOMEM);
1369                 goto out;
1370         }
1371
1372         rt->dst.flags |= DST_HOST;
1373         rt->dst.output  = ip6_output;
1374         atomic_set(&rt->dst.__refcnt, 1);
1375         rt->rt6i_gateway  = fl6->daddr;
1376         rt->rt6i_dst.addr = fl6->daddr;
1377         rt->rt6i_dst.plen = 128;
1378         rt->rt6i_idev     = idev;
1379         dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
1380
1381         spin_lock_bh(&icmp6_dst_lock);
1382         rt->dst.next = icmp6_dst_gc_list;
1383         icmp6_dst_gc_list = &rt->dst;
1384         spin_unlock_bh(&icmp6_dst_lock);
1385
1386         fib6_force_start_gc(net);
1387
1388         dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
1389
1390 out:
1391         return dst;
1392 }
1393
1394 int icmp6_dst_gc(void)
1395 {
1396         struct dst_entry *dst, **pprev;
1397         int more = 0;
1398
1399         spin_lock_bh(&icmp6_dst_lock);
1400         pprev = &icmp6_dst_gc_list;
1401
1402         while ((dst = *pprev) != NULL) {
1403                 if (!atomic_read(&dst->__refcnt)) {
1404                         *pprev = dst->next;
1405                         dst_free(dst);
1406                 } else {
1407                         pprev = &dst->next;
1408                         ++more;
1409                 }
1410         }
1411
1412         spin_unlock_bh(&icmp6_dst_lock);
1413
1414         return more;
1415 }
1416
1417 static void icmp6_clean_all(int (*func)(struct rt6_info *rt, void *arg),
1418                             void *arg)
1419 {
1420         struct dst_entry *dst, **pprev;
1421
1422         spin_lock_bh(&icmp6_dst_lock);
1423         pprev = &icmp6_dst_gc_list;
1424         while ((dst = *pprev) != NULL) {
1425                 struct rt6_info *rt = (struct rt6_info *) dst;
1426                 if (func(rt, arg)) {
1427                         *pprev = dst->next;
1428                         dst_free(dst);
1429                 } else {
1430                         pprev = &dst->next;
1431                 }
1432         }
1433         spin_unlock_bh(&icmp6_dst_lock);
1434 }
1435
1436 static int ip6_dst_gc(struct dst_ops *ops)
1437 {
1438         struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
1439         int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1440         int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1441         int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1442         int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1443         unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1444         int entries;
1445
1446         entries = dst_entries_get_fast(ops);
1447         if (time_after(rt_last_gc + rt_min_interval, jiffies) &&
1448             entries <= rt_max_size)
1449                 goto out;
1450
1451         net->ipv6.ip6_rt_gc_expire++;
1452         fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net, true);
1453         entries = dst_entries_get_slow(ops);
1454         if (entries < ops->gc_thresh)
1455                 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1456 out:
1457         net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1458         return entries > rt_max_size;
1459 }
1460
1461 /*
1462  *
1463  */
1464
1465 int ip6_route_add(struct fib6_config *cfg)
1466 {
1467         int err;
1468         struct net *net = cfg->fc_nlinfo.nl_net;
1469         struct rt6_info *rt = NULL;
1470         struct net_device *dev = NULL;
1471         struct inet6_dev *idev = NULL;
1472         struct fib6_table *table;
1473         int addr_type;
1474
1475         if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1476                 return -EINVAL;
1477 #ifndef CONFIG_IPV6_SUBTREES
1478         if (cfg->fc_src_len)
1479                 return -EINVAL;
1480 #endif
1481         if (cfg->fc_ifindex) {
1482                 err = -ENODEV;
1483                 dev = dev_get_by_index(net, cfg->fc_ifindex);
1484                 if (!dev)
1485                         goto out;
1486                 idev = in6_dev_get(dev);
1487                 if (!idev)
1488                         goto out;
1489         }
1490
1491         if (cfg->fc_metric == 0)
1492                 cfg->fc_metric = IP6_RT_PRIO_USER;
1493
1494         err = -ENOBUFS;
1495         if (cfg->fc_nlinfo.nlh &&
1496             !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
1497                 table = fib6_get_table(net, cfg->fc_table);
1498                 if (!table) {
1499                         pr_warn("NLM_F_CREATE should be specified when creating new route\n");
1500                         table = fib6_new_table(net, cfg->fc_table);
1501                 }
1502         } else {
1503                 table = fib6_new_table(net, cfg->fc_table);
1504         }
1505
1506         if (!table)
1507                 goto out;
1508
1509         rt = ip6_dst_alloc(net, NULL, (cfg->fc_flags & RTF_ADDRCONF) ? 0 : DST_NOCOUNT, table);
1510
1511         if (!rt) {
1512                 err = -ENOMEM;
1513                 goto out;
1514         }
1515
1516         if (cfg->fc_flags & RTF_EXPIRES)
1517                 rt6_set_expires(rt, jiffies +
1518                                 clock_t_to_jiffies(cfg->fc_expires));
1519         else
1520                 rt6_clean_expires(rt);
1521
1522         if (cfg->fc_protocol == RTPROT_UNSPEC)
1523                 cfg->fc_protocol = RTPROT_BOOT;
1524         rt->rt6i_protocol = cfg->fc_protocol;
1525
1526         addr_type = ipv6_addr_type(&cfg->fc_dst);
1527
1528         if (addr_type & IPV6_ADDR_MULTICAST)
1529                 rt->dst.input = ip6_mc_input;
1530         else if (cfg->fc_flags & RTF_LOCAL)
1531                 rt->dst.input = ip6_input;
1532         else
1533                 rt->dst.input = ip6_forward;
1534
1535         rt->dst.output = ip6_output;
1536
1537         ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1538         rt->rt6i_dst.plen = cfg->fc_dst_len;
1539         if (rt->rt6i_dst.plen == 128) {
1540                 rt->dst.flags |= DST_HOST;
1541                 dst_metrics_set_force_overwrite(&rt->dst);
1542         }
1543
1544 #ifdef CONFIG_IPV6_SUBTREES
1545         ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1546         rt->rt6i_src.plen = cfg->fc_src_len;
1547 #endif
1548
1549         rt->rt6i_metric = cfg->fc_metric;
1550
1551         /* We cannot add true routes via loopback here,
1552            they would result in kernel looping; promote them to reject routes
1553          */
1554         if ((cfg->fc_flags & RTF_REJECT) ||
1555             (dev && (dev->flags & IFF_LOOPBACK) &&
1556              !(addr_type & IPV6_ADDR_LOOPBACK) &&
1557              !(cfg->fc_flags & RTF_LOCAL))) {
1558                 /* hold loopback dev/idev if we haven't done so. */
1559                 if (dev != net->loopback_dev) {
1560                         if (dev) {
1561                                 dev_put(dev);
1562                                 in6_dev_put(idev);
1563                         }
1564                         dev = net->loopback_dev;
1565                         dev_hold(dev);
1566                         idev = in6_dev_get(dev);
1567                         if (!idev) {
1568                                 err = -ENODEV;
1569                                 goto out;
1570                         }
1571                 }
1572                 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1573                 switch (cfg->fc_type) {
1574                 case RTN_BLACKHOLE:
1575                         rt->dst.error = -EINVAL;
1576                         rt->dst.output = dst_discard_sk;
1577                         rt->dst.input = dst_discard;
1578                         break;
1579                 case RTN_PROHIBIT:
1580                         rt->dst.error = -EACCES;
1581                         rt->dst.output = ip6_pkt_prohibit_out;
1582                         rt->dst.input = ip6_pkt_prohibit;
1583                         break;
1584                 case RTN_THROW:
1585                 default:
1586                         rt->dst.error = (cfg->fc_type == RTN_THROW) ? -EAGAIN
1587                                         : -ENETUNREACH;
1588                         rt->dst.output = ip6_pkt_discard_out;
1589                         rt->dst.input = ip6_pkt_discard;
1590                         break;
1591                 }
1592                 goto install_route;
1593         }
1594
1595         if (cfg->fc_flags & RTF_GATEWAY) {
1596                 const struct in6_addr *gw_addr;
1597                 int gwa_type;
1598
1599                 gw_addr = &cfg->fc_gateway;
1600                 rt->rt6i_gateway = *gw_addr;
1601                 gwa_type = ipv6_addr_type(gw_addr);
1602
1603                 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1604                         struct rt6_info *grt;
1605
1606                         /* IPv6 strictly inhibits using not link-local
1607                            addresses as nexthop address.
1608                            Otherwise, router will not able to send redirects.
1609                            It is very good, but in some (rare!) circumstances
1610                            (SIT, PtP, NBMA NOARP links) it is handy to allow
1611                            some exceptions. --ANK
1612                          */
1613                         err = -EINVAL;
1614                         if (!(gwa_type & IPV6_ADDR_UNICAST))
1615                                 goto out;
1616
1617                         grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, 1);
1618
1619                         err = -EHOSTUNREACH;
1620                         if (!grt)
1621                                 goto out;
1622                         if (dev) {
1623                                 if (dev != grt->dst.dev) {
1624                                         ip6_rt_put(grt);
1625                                         goto out;
1626                                 }
1627                         } else {
1628                                 dev = grt->dst.dev;
1629                                 idev = grt->rt6i_idev;
1630                                 dev_hold(dev);
1631                                 in6_dev_hold(grt->rt6i_idev);
1632                         }
1633                         if (!(grt->rt6i_flags & RTF_GATEWAY))
1634                                 err = 0;
1635                         ip6_rt_put(grt);
1636
1637                         if (err)
1638                                 goto out;
1639                 }
1640                 err = -EINVAL;
1641                 if (!dev || (dev->flags & IFF_LOOPBACK))
1642                         goto out;
1643         }
1644
1645         err = -ENODEV;
1646         if (!dev)
1647                 goto out;
1648
1649         if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
1650                 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
1651                         err = -EINVAL;
1652                         goto out;
1653                 }
1654                 rt->rt6i_prefsrc.addr = cfg->fc_prefsrc;
1655                 rt->rt6i_prefsrc.plen = 128;
1656         } else
1657                 rt->rt6i_prefsrc.plen = 0;
1658
1659         rt->rt6i_flags = cfg->fc_flags;
1660
1661 install_route:
1662         rt->dst.dev = dev;
1663         rt->rt6i_idev = idev;
1664         rt->rt6i_table = table;
1665
1666         cfg->fc_nlinfo.nl_net = dev_net(dev);
1667
1668         return __ip6_ins_rt(rt, &cfg->fc_nlinfo, cfg->fc_mx, cfg->fc_mx_len);
1669
1670 out:
1671         if (dev)
1672                 dev_put(dev);
1673         if (idev)
1674                 in6_dev_put(idev);
1675         if (rt)
1676                 dst_free(&rt->dst);
1677         return err;
1678 }
1679
1680 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
1681 {
1682         int err;
1683         struct fib6_table *table;
1684         struct net *net = dev_net(rt->dst.dev);
1685
1686         if (rt == net->ipv6.ip6_null_entry) {
1687                 err = -ENOENT;
1688                 goto out;
1689         }
1690
1691         table = rt->rt6i_table;
1692         write_lock_bh(&table->tb6_lock);
1693         err = fib6_del(rt, info);
1694         write_unlock_bh(&table->tb6_lock);
1695
1696 out:
1697         ip6_rt_put(rt);
1698         return err;
1699 }
1700
1701 int ip6_del_rt(struct rt6_info *rt)
1702 {
1703         struct nl_info info = {
1704                 .nl_net = dev_net(rt->dst.dev),
1705         };
1706         return __ip6_del_rt(rt, &info);
1707 }
1708
1709 static int ip6_route_del(struct fib6_config *cfg)
1710 {
1711         struct fib6_table *table;
1712         struct fib6_node *fn;
1713         struct rt6_info *rt;
1714         int err = -ESRCH;
1715
1716         table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
1717         if (!table)
1718                 return err;
1719
1720         read_lock_bh(&table->tb6_lock);
1721
1722         fn = fib6_locate(&table->tb6_root,
1723                          &cfg->fc_dst, cfg->fc_dst_len,
1724                          &cfg->fc_src, cfg->fc_src_len);
1725
1726         if (fn) {
1727                 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1728                         if (cfg->fc_ifindex &&
1729                             (!rt->dst.dev ||
1730                              rt->dst.dev->ifindex != cfg->fc_ifindex))
1731                                 continue;
1732                         if (cfg->fc_flags & RTF_GATEWAY &&
1733                             !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
1734                                 continue;
1735                         if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
1736                                 continue;
1737                         dst_hold(&rt->dst);
1738                         read_unlock_bh(&table->tb6_lock);
1739
1740                         return __ip6_del_rt(rt, &cfg->fc_nlinfo);
1741                 }
1742         }
1743         read_unlock_bh(&table->tb6_lock);
1744
1745         return err;
1746 }
1747
1748 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
1749 {
1750         struct net *net = dev_net(skb->dev);
1751         struct netevent_redirect netevent;
1752         struct rt6_info *rt, *nrt = NULL;
1753         struct ndisc_options ndopts;
1754         struct inet6_dev *in6_dev;
1755         struct neighbour *neigh;
1756         struct rd_msg *msg;
1757         int optlen, on_link;
1758         u8 *lladdr;
1759
1760         optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
1761         optlen -= sizeof(*msg);
1762
1763         if (optlen < 0) {
1764                 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
1765                 return;
1766         }
1767
1768         msg = (struct rd_msg *)icmp6_hdr(skb);
1769
1770         if (ipv6_addr_is_multicast(&msg->dest)) {
1771                 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
1772                 return;
1773         }
1774
1775         on_link = 0;
1776         if (ipv6_addr_equal(&msg->dest, &msg->target)) {
1777                 on_link = 1;
1778         } else if (ipv6_addr_type(&msg->target) !=
1779                    (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
1780                 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
1781                 return;
1782         }
1783
1784         in6_dev = __in6_dev_get(skb->dev);
1785         if (!in6_dev)
1786                 return;
1787         if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
1788                 return;
1789
1790         /* RFC2461 8.1:
1791          *      The IP source address of the Redirect MUST be the same as the current
1792          *      first-hop router for the specified ICMP Destination Address.
1793          */
1794
1795         if (!ndisc_parse_options(msg->opt, optlen, &ndopts)) {
1796                 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
1797                 return;
1798         }
1799
1800         lladdr = NULL;
1801         if (ndopts.nd_opts_tgt_lladdr) {
1802                 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
1803                                              skb->dev);
1804                 if (!lladdr) {
1805                         net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
1806                         return;
1807                 }
1808         }
1809
1810         rt = (struct rt6_info *) dst;
1811         if (rt == net->ipv6.ip6_null_entry) {
1812                 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
1813                 return;
1814         }
1815
1816         /* Redirect received -> path was valid.
1817          * Look, redirects are sent only in response to data packets,
1818          * so that this nexthop apparently is reachable. --ANK
1819          */
1820         dst_confirm(&rt->dst);
1821
1822         neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
1823         if (!neigh)
1824                 return;
1825
1826         /*
1827          *      We have finally decided to accept it.
1828          */
1829
1830         neigh_update(neigh, lladdr, NUD_STALE,
1831                      NEIGH_UPDATE_F_WEAK_OVERRIDE|
1832                      NEIGH_UPDATE_F_OVERRIDE|
1833                      (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1834                                      NEIGH_UPDATE_F_ISROUTER))
1835                      );
1836
1837         nrt = ip6_rt_copy(rt, &msg->dest);
1838         if (!nrt)
1839                 goto out;
1840
1841         nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
1842         if (on_link)
1843                 nrt->rt6i_flags &= ~RTF_GATEWAY;
1844
1845         nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
1846
1847         if (ip6_ins_rt(nrt))
1848                 goto out;
1849
1850         netevent.old = &rt->dst;
1851         netevent.new = &nrt->dst;
1852         netevent.daddr = &msg->dest;
1853         netevent.neigh = neigh;
1854         call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
1855
1856         if (rt->rt6i_flags & RTF_CACHE) {
1857                 rt = (struct rt6_info *) dst_clone(&rt->dst);
1858                 ip6_del_rt(rt);
1859         }
1860
1861 out:
1862         neigh_release(neigh);
1863 }
1864
1865 /*
1866  *      Misc support functions
1867  */
1868
1869 static struct rt6_info *ip6_rt_copy(struct rt6_info *ort,
1870                                     const struct in6_addr *dest)
1871 {
1872         struct net *net = dev_net(ort->dst.dev);
1873         struct rt6_info *rt = ip6_dst_alloc(net, ort->dst.dev, 0,
1874                                             ort->rt6i_table);
1875
1876         if (rt) {
1877                 rt->dst.input = ort->dst.input;
1878                 rt->dst.output = ort->dst.output;
1879                 rt->dst.flags |= DST_HOST;
1880
1881                 rt->rt6i_dst.addr = *dest;
1882                 rt->rt6i_dst.plen = 128;
1883                 dst_copy_metrics(&rt->dst, &ort->dst);
1884                 rt->dst.error = ort->dst.error;
1885                 rt->rt6i_idev = ort->rt6i_idev;
1886                 if (rt->rt6i_idev)
1887                         in6_dev_hold(rt->rt6i_idev);
1888                 rt->dst.lastuse = jiffies;
1889
1890                 if (ort->rt6i_flags & RTF_GATEWAY)
1891                         rt->rt6i_gateway = ort->rt6i_gateway;
1892                 else
1893                         rt->rt6i_gateway = *dest;
1894                 rt->rt6i_flags = ort->rt6i_flags;
1895                 rt6_set_from(rt, ort);
1896                 rt->rt6i_metric = 0;
1897
1898 #ifdef CONFIG_IPV6_SUBTREES
1899                 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1900 #endif
1901                 memcpy(&rt->rt6i_prefsrc, &ort->rt6i_prefsrc, sizeof(struct rt6key));
1902                 rt->rt6i_table = ort->rt6i_table;
1903         }
1904         return rt;
1905 }
1906
1907 #ifdef CONFIG_IPV6_ROUTE_INFO
1908 static struct rt6_info *rt6_get_route_info(struct net *net,
1909                                            const struct in6_addr *prefix, int prefixlen,
1910                                            const struct in6_addr *gwaddr, int ifindex)
1911 {
1912         struct fib6_node *fn;
1913         struct rt6_info *rt = NULL;
1914         struct fib6_table *table;
1915
1916         table = fib6_get_table(net, RT6_TABLE_INFO);
1917         if (!table)
1918                 return NULL;
1919
1920         read_lock_bh(&table->tb6_lock);
1921         fn = fib6_locate(&table->tb6_root, prefix, prefixlen, NULL, 0);
1922         if (!fn)
1923                 goto out;
1924
1925         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1926                 if (rt->dst.dev->ifindex != ifindex)
1927                         continue;
1928                 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
1929                         continue;
1930                 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
1931                         continue;
1932                 dst_hold(&rt->dst);
1933                 break;
1934         }
1935 out:
1936         read_unlock_bh(&table->tb6_lock);
1937         return rt;
1938 }
1939
1940 static struct rt6_info *rt6_add_route_info(struct net *net,
1941                                            const struct in6_addr *prefix, int prefixlen,
1942                                            const struct in6_addr *gwaddr, int ifindex,
1943                                            unsigned int pref)
1944 {
1945         struct fib6_config cfg = {
1946                 .fc_table       = RT6_TABLE_INFO,
1947                 .fc_metric      = IP6_RT_PRIO_USER,
1948                 .fc_ifindex     = ifindex,
1949                 .fc_dst_len     = prefixlen,
1950                 .fc_flags       = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
1951                                   RTF_UP | RTF_PREF(pref),
1952                 .fc_nlinfo.portid = 0,
1953                 .fc_nlinfo.nlh = NULL,
1954                 .fc_nlinfo.nl_net = net,
1955         };
1956
1957         cfg.fc_dst = *prefix;
1958         cfg.fc_gateway = *gwaddr;
1959
1960         /* We should treat it as a default route if prefix length is 0. */
1961         if (!prefixlen)
1962                 cfg.fc_flags |= RTF_DEFAULT;
1963
1964         ip6_route_add(&cfg);
1965
1966         return rt6_get_route_info(net, prefix, prefixlen, gwaddr, ifindex);
1967 }
1968 #endif
1969
1970 struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
1971 {
1972         struct rt6_info *rt;
1973         struct fib6_table *table;
1974
1975         table = fib6_get_table(dev_net(dev), RT6_TABLE_DFLT);
1976         if (!table)
1977                 return NULL;
1978
1979         read_lock_bh(&table->tb6_lock);
1980         for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
1981                 if (dev == rt->dst.dev &&
1982                     ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
1983                     ipv6_addr_equal(&rt->rt6i_gateway, addr))
1984                         break;
1985         }
1986         if (rt)
1987                 dst_hold(&rt->dst);
1988         read_unlock_bh(&table->tb6_lock);
1989         return rt;
1990 }
1991
1992 struct rt6_info *rt6_add_dflt_router(const struct in6_addr *gwaddr,
1993                                      struct net_device *dev,
1994                                      unsigned int pref)
1995 {
1996         struct fib6_config cfg = {
1997                 .fc_table       = RT6_TABLE_DFLT,
1998                 .fc_metric      = IP6_RT_PRIO_USER,
1999                 .fc_ifindex     = dev->ifindex,
2000                 .fc_flags       = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
2001                                   RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
2002                 .fc_nlinfo.portid = 0,
2003                 .fc_nlinfo.nlh = NULL,
2004                 .fc_nlinfo.nl_net = dev_net(dev),
2005         };
2006
2007         cfg.fc_gateway = *gwaddr;
2008
2009         ip6_route_add(&cfg);
2010
2011         return rt6_get_dflt_router(gwaddr, dev);
2012 }
2013
2014 void rt6_purge_dflt_routers(struct net *net)
2015 {
2016         struct rt6_info *rt;
2017         struct fib6_table *table;
2018
2019         /* NOTE: Keep consistent with rt6_get_dflt_router */
2020         table = fib6_get_table(net, RT6_TABLE_DFLT);
2021         if (!table)
2022                 return;
2023
2024 restart:
2025         read_lock_bh(&table->tb6_lock);
2026         for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2027                 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
2028                     (!rt->rt6i_idev || rt->rt6i_idev->cnf.accept_ra != 2)) {
2029                         dst_hold(&rt->dst);
2030                         read_unlock_bh(&table->tb6_lock);
2031                         ip6_del_rt(rt);
2032                         goto restart;
2033                 }
2034         }
2035         read_unlock_bh(&table->tb6_lock);
2036 }
2037
2038 static void rtmsg_to_fib6_config(struct net *net,
2039                                  struct in6_rtmsg *rtmsg,
2040                                  struct fib6_config *cfg)
2041 {
2042         memset(cfg, 0, sizeof(*cfg));
2043
2044         cfg->fc_table = RT6_TABLE_MAIN;
2045         cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
2046         cfg->fc_metric = rtmsg->rtmsg_metric;
2047         cfg->fc_expires = rtmsg->rtmsg_info;
2048         cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
2049         cfg->fc_src_len = rtmsg->rtmsg_src_len;
2050         cfg->fc_flags = rtmsg->rtmsg_flags;
2051
2052         cfg->fc_nlinfo.nl_net = net;
2053
2054         cfg->fc_dst = rtmsg->rtmsg_dst;
2055         cfg->fc_src = rtmsg->rtmsg_src;
2056         cfg->fc_gateway = rtmsg->rtmsg_gateway;
2057 }
2058
2059 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
2060 {
2061         struct fib6_config cfg;
2062         struct in6_rtmsg rtmsg;
2063         int err;
2064
2065         switch (cmd) {
2066         case SIOCADDRT:         /* Add a route */
2067         case SIOCDELRT:         /* Delete a route */
2068                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2069                         return -EPERM;
2070                 err = copy_from_user(&rtmsg, arg,
2071                                      sizeof(struct in6_rtmsg));
2072                 if (err)
2073                         return -EFAULT;
2074
2075                 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
2076
2077                 rtnl_lock();
2078                 switch (cmd) {
2079                 case SIOCADDRT:
2080                         err = ip6_route_add(&cfg);
2081                         break;
2082                 case SIOCDELRT:
2083                         err = ip6_route_del(&cfg);
2084                         break;
2085                 default:
2086                         err = -EINVAL;
2087                 }
2088                 rtnl_unlock();
2089
2090                 return err;
2091         }
2092
2093         return -EINVAL;
2094 }
2095
2096 /*
2097  *      Drop the packet on the floor
2098  */
2099
2100 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
2101 {
2102         int type;
2103         struct dst_entry *dst = skb_dst(skb);
2104         switch (ipstats_mib_noroutes) {
2105         case IPSTATS_MIB_INNOROUTES:
2106                 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
2107                 if (type == IPV6_ADDR_ANY) {
2108                         IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2109                                       IPSTATS_MIB_INADDRERRORS);
2110                         break;
2111                 }
2112                 /* FALLTHROUGH */
2113         case IPSTATS_MIB_OUTNOROUTES:
2114                 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2115                               ipstats_mib_noroutes);
2116                 break;
2117         }
2118         icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
2119         kfree_skb(skb);
2120         return 0;
2121 }
2122
2123 static int ip6_pkt_discard(struct sk_buff *skb)
2124 {
2125         return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
2126 }
2127
2128 static int ip6_pkt_discard_out(struct sock *sk, struct sk_buff *skb)
2129 {
2130         skb->dev = skb_dst(skb)->dev;
2131         return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
2132 }
2133
2134 static int ip6_pkt_prohibit(struct sk_buff *skb)
2135 {
2136         return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
2137 }
2138
2139 static int ip6_pkt_prohibit_out(struct sock *sk, struct sk_buff *skb)
2140 {
2141         skb->dev = skb_dst(skb)->dev;
2142         return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
2143 }
2144
2145 /*
2146  *      Allocate a dst for local (unicast / anycast) address.
2147  */
2148
2149 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
2150                                     const struct in6_addr *addr,
2151                                     bool anycast)
2152 {
2153         struct net *net = dev_net(idev->dev);
2154         struct rt6_info *rt = ip6_dst_alloc(net, net->loopback_dev,
2155                                             DST_NOCOUNT, NULL);
2156         if (!rt)
2157                 return ERR_PTR(-ENOMEM);
2158
2159         in6_dev_hold(idev);
2160
2161         rt->dst.flags |= DST_HOST;
2162         rt->dst.input = ip6_input;
2163         rt->dst.output = ip6_output;
2164         rt->rt6i_idev = idev;
2165
2166         rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
2167         if (anycast)
2168                 rt->rt6i_flags |= RTF_ANYCAST;
2169         else
2170                 rt->rt6i_flags |= RTF_LOCAL;
2171
2172         rt->rt6i_gateway  = *addr;
2173         rt->rt6i_dst.addr = *addr;
2174         rt->rt6i_dst.plen = 128;
2175         rt->rt6i_table = fib6_get_table(net, RT6_TABLE_LOCAL);
2176
2177         atomic_set(&rt->dst.__refcnt, 1);
2178
2179         return rt;
2180 }
2181
2182 int ip6_route_get_saddr(struct net *net,
2183                         struct rt6_info *rt,
2184                         const struct in6_addr *daddr,
2185                         unsigned int prefs,
2186                         struct in6_addr *saddr)
2187 {
2188         struct inet6_dev *idev = ip6_dst_idev((struct dst_entry *)rt);
2189         int err = 0;
2190         if (rt->rt6i_prefsrc.plen)
2191                 *saddr = rt->rt6i_prefsrc.addr;
2192         else
2193                 err = ipv6_dev_get_saddr(net, idev ? idev->dev : NULL,
2194                                          daddr, prefs, saddr);
2195         return err;
2196 }
2197
2198 /* remove deleted ip from prefsrc entries */
2199 struct arg_dev_net_ip {
2200         struct net_device *dev;
2201         struct net *net;
2202         struct in6_addr *addr;
2203 };
2204
2205 static int fib6_remove_prefsrc(struct rt6_info *rt, void *arg)
2206 {
2207         struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
2208         struct net *net = ((struct arg_dev_net_ip *)arg)->net;
2209         struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
2210
2211         if (((void *)rt->dst.dev == dev || !dev) &&
2212             rt != net->ipv6.ip6_null_entry &&
2213             ipv6_addr_equal(addr, &rt->rt6i_prefsrc.addr)) {
2214                 /* remove prefsrc entry */
2215                 rt->rt6i_prefsrc.plen = 0;
2216         }
2217         return 0;
2218 }
2219
2220 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
2221 {
2222         struct net *net = dev_net(ifp->idev->dev);
2223         struct arg_dev_net_ip adni = {
2224                 .dev = ifp->idev->dev,
2225                 .net = net,
2226                 .addr = &ifp->addr,
2227         };
2228         fib6_clean_all(net, fib6_remove_prefsrc, &adni);
2229 }
2230
2231 #define RTF_RA_ROUTER           (RTF_ADDRCONF | RTF_DEFAULT | RTF_GATEWAY)
2232 #define RTF_CACHE_GATEWAY       (RTF_GATEWAY | RTF_CACHE)
2233
2234 /* Remove routers and update dst entries when gateway turn into host. */
2235 static int fib6_clean_tohost(struct rt6_info *rt, void *arg)
2236 {
2237         struct in6_addr *gateway = (struct in6_addr *)arg;
2238
2239         if ((((rt->rt6i_flags & RTF_RA_ROUTER) == RTF_RA_ROUTER) ||
2240              ((rt->rt6i_flags & RTF_CACHE_GATEWAY) == RTF_CACHE_GATEWAY)) &&
2241              ipv6_addr_equal(gateway, &rt->rt6i_gateway)) {
2242                 return -1;
2243         }
2244         return 0;
2245 }
2246
2247 void rt6_clean_tohost(struct net *net, struct in6_addr *gateway)
2248 {
2249         fib6_clean_all(net, fib6_clean_tohost, gateway);
2250 }
2251
2252 struct arg_dev_net {
2253         struct net_device *dev;
2254         struct net *net;
2255 };
2256
2257 static int fib6_ifdown(struct rt6_info *rt, void *arg)
2258 {
2259         const struct arg_dev_net *adn = arg;
2260         const struct net_device *dev = adn->dev;
2261
2262         if ((rt->dst.dev == dev || !dev) &&
2263             rt != adn->net->ipv6.ip6_null_entry)
2264                 return -1;
2265
2266         return 0;
2267 }
2268
2269 void rt6_ifdown(struct net *net, struct net_device *dev)
2270 {
2271         struct arg_dev_net adn = {
2272                 .dev = dev,
2273                 .net = net,
2274         };
2275
2276         fib6_clean_all(net, fib6_ifdown, &adn);
2277         icmp6_clean_all(fib6_ifdown, &adn);
2278 }
2279
2280 struct rt6_mtu_change_arg {
2281         struct net_device *dev;
2282         unsigned int mtu;
2283 };
2284
2285 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2286 {
2287         struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2288         struct inet6_dev *idev;
2289
2290         /* In IPv6 pmtu discovery is not optional,
2291            so that RTAX_MTU lock cannot disable it.
2292            We still use this lock to block changes
2293            caused by addrconf/ndisc.
2294         */
2295
2296         idev = __in6_dev_get(arg->dev);
2297         if (!idev)
2298                 return 0;
2299
2300         /* For administrative MTU increase, there is no way to discover
2301            IPv6 PMTU increase, so PMTU increase should be updated here.
2302            Since RFC 1981 doesn't include administrative MTU increase
2303            update PMTU increase is a MUST. (i.e. jumbo frame)
2304          */
2305         /*
2306            If new MTU is less than route PMTU, this new MTU will be the
2307            lowest MTU in the path, update the route PMTU to reflect PMTU
2308            decreases; if new MTU is greater than route PMTU, and the
2309            old MTU is the lowest MTU in the path, update the route PMTU
2310            to reflect the increase. In this case if the other nodes' MTU
2311            also have the lowest MTU, TOO BIG MESSAGE will be lead to
2312            PMTU discouvery.
2313          */
2314         if (rt->dst.dev == arg->dev &&
2315             !dst_metric_locked(&rt->dst, RTAX_MTU) &&
2316             (dst_mtu(&rt->dst) >= arg->mtu ||
2317              (dst_mtu(&rt->dst) < arg->mtu &&
2318               dst_mtu(&rt->dst) == idev->cnf.mtu6))) {
2319                 dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
2320         }
2321         return 0;
2322 }
2323
2324 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
2325 {
2326         struct rt6_mtu_change_arg arg = {
2327                 .dev = dev,
2328                 .mtu = mtu,
2329         };
2330
2331         fib6_clean_all(dev_net(dev), rt6_mtu_change_route, &arg);
2332 }
2333
2334 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2335         [RTA_GATEWAY]           = { .len = sizeof(struct in6_addr) },
2336         [RTA_OIF]               = { .type = NLA_U32 },
2337         [RTA_IIF]               = { .type = NLA_U32 },
2338         [RTA_PRIORITY]          = { .type = NLA_U32 },
2339         [RTA_METRICS]           = { .type = NLA_NESTED },
2340         [RTA_MULTIPATH]         = { .len = sizeof(struct rtnexthop) },
2341 };
2342
2343 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2344                               struct fib6_config *cfg)
2345 {
2346         struct rtmsg *rtm;
2347         struct nlattr *tb[RTA_MAX+1];
2348         int err;
2349
2350         err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2351         if (err < 0)
2352                 goto errout;
2353
2354         err = -EINVAL;
2355         rtm = nlmsg_data(nlh);
2356         memset(cfg, 0, sizeof(*cfg));
2357
2358         cfg->fc_table = rtm->rtm_table;
2359         cfg->fc_dst_len = rtm->rtm_dst_len;
2360         cfg->fc_src_len = rtm->rtm_src_len;
2361         cfg->fc_flags = RTF_UP;
2362         cfg->fc_protocol = rtm->rtm_protocol;
2363         cfg->fc_type = rtm->rtm_type;
2364
2365         if (rtm->rtm_type == RTN_UNREACHABLE ||
2366             rtm->rtm_type == RTN_BLACKHOLE ||
2367             rtm->rtm_type == RTN_PROHIBIT ||
2368             rtm->rtm_type == RTN_THROW)
2369                 cfg->fc_flags |= RTF_REJECT;
2370
2371         if (rtm->rtm_type == RTN_LOCAL)
2372                 cfg->fc_flags |= RTF_LOCAL;
2373
2374         cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid;
2375         cfg->fc_nlinfo.nlh = nlh;
2376         cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2377
2378         if (tb[RTA_GATEWAY]) {
2379                 nla_memcpy(&cfg->fc_gateway, tb[RTA_GATEWAY], 16);
2380                 cfg->fc_flags |= RTF_GATEWAY;
2381         }
2382
2383         if (tb[RTA_DST]) {
2384                 int plen = (rtm->rtm_dst_len + 7) >> 3;
2385
2386                 if (nla_len(tb[RTA_DST]) < plen)
2387                         goto errout;
2388
2389                 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2390         }
2391
2392         if (tb[RTA_SRC]) {
2393                 int plen = (rtm->rtm_src_len + 7) >> 3;
2394
2395                 if (nla_len(tb[RTA_SRC]) < plen)
2396                         goto errout;
2397
2398                 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2399         }
2400
2401         if (tb[RTA_PREFSRC])
2402                 nla_memcpy(&cfg->fc_prefsrc, tb[RTA_PREFSRC], 16);
2403
2404         if (tb[RTA_OIF])
2405                 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2406
2407         if (tb[RTA_PRIORITY])
2408                 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2409
2410         if (tb[RTA_METRICS]) {
2411                 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2412                 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2413         }
2414
2415         if (tb[RTA_TABLE])
2416                 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2417
2418         if (tb[RTA_MULTIPATH]) {
2419                 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
2420                 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
2421         }
2422
2423         err = 0;
2424 errout:
2425         return err;
2426 }
2427
2428 static int ip6_route_multipath(struct fib6_config *cfg, int add)
2429 {
2430         struct fib6_config r_cfg;
2431         struct rtnexthop *rtnh;
2432         int remaining;
2433         int attrlen;
2434         int err = 0, last_err = 0;
2435
2436 beginning:
2437         rtnh = (struct rtnexthop *)cfg->fc_mp;
2438         remaining = cfg->fc_mp_len;
2439
2440         /* Parse a Multipath Entry */
2441         while (rtnh_ok(rtnh, remaining)) {
2442                 memcpy(&r_cfg, cfg, sizeof(*cfg));
2443                 if (rtnh->rtnh_ifindex)
2444                         r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
2445
2446                 attrlen = rtnh_attrlen(rtnh);
2447                 if (attrlen > 0) {
2448                         struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
2449
2450                         nla = nla_find(attrs, attrlen, RTA_GATEWAY);
2451                         if (nla) {
2452                                 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
2453                                 r_cfg.fc_flags |= RTF_GATEWAY;
2454                         }
2455                 }
2456                 err = add ? ip6_route_add(&r_cfg) : ip6_route_del(&r_cfg);
2457                 if (err) {
2458                         last_err = err;
2459                         /* If we are trying to remove a route, do not stop the
2460                          * loop when ip6_route_del() fails (because next hop is
2461                          * already gone), we should try to remove all next hops.
2462                          */
2463                         if (add) {
2464                                 /* If add fails, we should try to delete all
2465                                  * next hops that have been already added.
2466                                  */
2467                                 add = 0;
2468                                 goto beginning;
2469                         }
2470                 }
2471                 /* Because each route is added like a single route we remove
2472                  * this flag after the first nexthop (if there is a collision,
2473                  * we have already fail to add the first nexthop:
2474                  * fib6_add_rt2node() has reject it).
2475                  */
2476                 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~NLM_F_EXCL;
2477                 rtnh = rtnh_next(rtnh, &remaining);
2478         }
2479
2480         return last_err;
2481 }
2482
2483 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh)
2484 {
2485         struct fib6_config cfg;
2486         int err;
2487
2488         err = rtm_to_fib6_config(skb, nlh, &cfg);
2489         if (err < 0)
2490                 return err;
2491
2492         if (cfg.fc_mp)
2493                 return ip6_route_multipath(&cfg, 0);
2494         else
2495                 return ip6_route_del(&cfg);
2496 }
2497
2498 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh)
2499 {
2500         struct fib6_config cfg;
2501         int err;
2502
2503         err = rtm_to_fib6_config(skb, nlh, &cfg);
2504         if (err < 0)
2505                 return err;
2506
2507         if (cfg.fc_mp)
2508                 return ip6_route_multipath(&cfg, 1);
2509         else
2510                 return ip6_route_add(&cfg);
2511 }
2512
2513 static inline size_t rt6_nlmsg_size(void)
2514 {
2515         return NLMSG_ALIGN(sizeof(struct rtmsg))
2516                + nla_total_size(16) /* RTA_SRC */
2517                + nla_total_size(16) /* RTA_DST */
2518                + nla_total_size(16) /* RTA_GATEWAY */
2519                + nla_total_size(16) /* RTA_PREFSRC */
2520                + nla_total_size(4) /* RTA_TABLE */
2521                + nla_total_size(4) /* RTA_IIF */
2522                + nla_total_size(4) /* RTA_OIF */
2523                + nla_total_size(4) /* RTA_PRIORITY */
2524                + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
2525                + nla_total_size(sizeof(struct rta_cacheinfo));
2526 }
2527
2528 static int rt6_fill_node(struct net *net,
2529                          struct sk_buff *skb, struct rt6_info *rt,
2530                          struct in6_addr *dst, struct in6_addr *src,
2531                          int iif, int type, u32 portid, u32 seq,
2532                          int prefix, int nowait, unsigned int flags)
2533 {
2534         struct rtmsg *rtm;
2535         struct nlmsghdr *nlh;
2536         long expires;
2537         u32 table;
2538
2539         if (prefix) {   /* user wants prefix routes only */
2540                 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
2541                         /* success since this is not a prefix route */
2542                         return 1;
2543                 }
2544         }
2545
2546         nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
2547         if (!nlh)
2548                 return -EMSGSIZE;
2549
2550         rtm = nlmsg_data(nlh);
2551         rtm->rtm_family = AF_INET6;
2552         rtm->rtm_dst_len = rt->rt6i_dst.plen;
2553         rtm->rtm_src_len = rt->rt6i_src.plen;
2554         rtm->rtm_tos = 0;
2555         if (rt->rt6i_table)
2556                 table = rt->rt6i_table->tb6_id;
2557         else
2558                 table = RT6_TABLE_UNSPEC;
2559         rtm->rtm_table = table;
2560         if (nla_put_u32(skb, RTA_TABLE, table))
2561                 goto nla_put_failure;
2562         if (rt->rt6i_flags & RTF_REJECT) {
2563                 switch (rt->dst.error) {
2564                 case -EINVAL:
2565                         rtm->rtm_type = RTN_BLACKHOLE;
2566                         break;
2567                 case -EACCES:
2568                         rtm->rtm_type = RTN_PROHIBIT;
2569                         break;
2570                 case -EAGAIN:
2571                         rtm->rtm_type = RTN_THROW;
2572                         break;
2573                 default:
2574                         rtm->rtm_type = RTN_UNREACHABLE;
2575                         break;
2576                 }
2577         }
2578         else if (rt->rt6i_flags & RTF_LOCAL)
2579                 rtm->rtm_type = RTN_LOCAL;
2580         else if (rt->dst.dev && (rt->dst.dev->flags & IFF_LOOPBACK))
2581                 rtm->rtm_type = RTN_LOCAL;
2582         else
2583                 rtm->rtm_type = RTN_UNICAST;
2584         rtm->rtm_flags = 0;
2585         rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2586         rtm->rtm_protocol = rt->rt6i_protocol;
2587         if (rt->rt6i_flags & RTF_DYNAMIC)
2588                 rtm->rtm_protocol = RTPROT_REDIRECT;
2589         else if (rt->rt6i_flags & RTF_ADDRCONF) {
2590                 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ROUTEINFO))
2591                         rtm->rtm_protocol = RTPROT_RA;
2592                 else
2593                         rtm->rtm_protocol = RTPROT_KERNEL;
2594         }
2595
2596         if (rt->rt6i_flags & RTF_CACHE)
2597                 rtm->rtm_flags |= RTM_F_CLONED;
2598
2599         if (dst) {
2600                 if (nla_put(skb, RTA_DST, 16, dst))
2601                         goto nla_put_failure;
2602                 rtm->rtm_dst_len = 128;
2603         } else if (rtm->rtm_dst_len)
2604                 if (nla_put(skb, RTA_DST, 16, &rt->rt6i_dst.addr))
2605                         goto nla_put_failure;
2606 #ifdef CONFIG_IPV6_SUBTREES
2607         if (src) {
2608                 if (nla_put(skb, RTA_SRC, 16, src))
2609                         goto nla_put_failure;
2610                 rtm->rtm_src_len = 128;
2611         } else if (rtm->rtm_src_len &&
2612                    nla_put(skb, RTA_SRC, 16, &rt->rt6i_src.addr))
2613                 goto nla_put_failure;
2614 #endif
2615         if (iif) {
2616 #ifdef CONFIG_IPV6_MROUTE
2617                 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
2618                         int err = ip6mr_get_route(net, skb, rtm, nowait);
2619                         if (err <= 0) {
2620                                 if (!nowait) {
2621                                         if (err == 0)
2622                                                 return 0;
2623                                         goto nla_put_failure;
2624                                 } else {
2625                                         if (err == -EMSGSIZE)
2626                                                 goto nla_put_failure;
2627                                 }
2628                         }
2629                 } else
2630 #endif
2631                         if (nla_put_u32(skb, RTA_IIF, iif))
2632                                 goto nla_put_failure;
2633         } else if (dst) {
2634                 struct in6_addr saddr_buf;
2635                 if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0 &&
2636                     nla_put(skb, RTA_PREFSRC, 16, &saddr_buf))
2637                         goto nla_put_failure;
2638         }
2639
2640         if (rt->rt6i_prefsrc.plen) {
2641                 struct in6_addr saddr_buf;
2642                 saddr_buf = rt->rt6i_prefsrc.addr;
2643                 if (nla_put(skb, RTA_PREFSRC, 16, &saddr_buf))
2644                         goto nla_put_failure;
2645         }
2646
2647         if (rtnetlink_put_metrics(skb, dst_metrics_ptr(&rt->dst)) < 0)
2648                 goto nla_put_failure;
2649
2650         if (rt->rt6i_flags & RTF_GATEWAY) {
2651                 if (nla_put(skb, RTA_GATEWAY, 16, &rt->rt6i_gateway) < 0)
2652                         goto nla_put_failure;
2653         }
2654
2655         if (rt->dst.dev &&
2656             nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex))
2657                 goto nla_put_failure;
2658         if (nla_put_u32(skb, RTA_PRIORITY, rt->rt6i_metric))
2659                 goto nla_put_failure;
2660
2661         expires = (rt->rt6i_flags & RTF_EXPIRES) ? rt->dst.expires - jiffies : 0;
2662
2663         if (rtnl_put_cacheinfo(skb, &rt->dst, 0, expires, rt->dst.error) < 0)
2664                 goto nla_put_failure;
2665
2666         return nlmsg_end(skb, nlh);
2667
2668 nla_put_failure:
2669         nlmsg_cancel(skb, nlh);
2670         return -EMSGSIZE;
2671 }
2672
2673 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
2674 {
2675         struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
2676         int prefix;
2677
2678         if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
2679                 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
2680                 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
2681         } else
2682                 prefix = 0;
2683
2684         return rt6_fill_node(arg->net,
2685                      arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
2686                      NETLINK_CB(arg->cb->skb).portid, arg->cb->nlh->nlmsg_seq,
2687                      prefix, 0, NLM_F_MULTI);
2688 }
2689
2690 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh)
2691 {
2692         struct net *net = sock_net(in_skb->sk);
2693         struct nlattr *tb[RTA_MAX+1];
2694         struct rt6_info *rt;
2695         struct sk_buff *skb;
2696         struct rtmsg *rtm;
2697         struct flowi6 fl6;
2698         int err, iif = 0, oif = 0;
2699
2700         err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2701         if (err < 0)
2702                 goto errout;
2703
2704         err = -EINVAL;
2705         memset(&fl6, 0, sizeof(fl6));
2706
2707         if (tb[RTA_SRC]) {
2708                 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
2709                         goto errout;
2710
2711                 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
2712         }
2713
2714         if (tb[RTA_DST]) {
2715                 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
2716                         goto errout;
2717
2718                 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
2719         }
2720
2721         if (tb[RTA_IIF])
2722                 iif = nla_get_u32(tb[RTA_IIF]);
2723
2724         if (tb[RTA_OIF])
2725                 oif = nla_get_u32(tb[RTA_OIF]);
2726
2727         if (tb[RTA_MARK])
2728                 fl6.flowi6_mark = nla_get_u32(tb[RTA_MARK]);
2729
2730         if (iif) {
2731                 struct net_device *dev;
2732                 int flags = 0;
2733
2734                 dev = __dev_get_by_index(net, iif);
2735                 if (!dev) {
2736                         err = -ENODEV;
2737                         goto errout;
2738                 }
2739
2740                 fl6.flowi6_iif = iif;
2741
2742                 if (!ipv6_addr_any(&fl6.saddr))
2743                         flags |= RT6_LOOKUP_F_HAS_SADDR;
2744
2745                 rt = (struct rt6_info *)ip6_route_input_lookup(net, dev, &fl6,
2746                                                                flags);
2747         } else {
2748                 fl6.flowi6_oif = oif;
2749
2750                 rt = (struct rt6_info *)ip6_route_output(net, NULL, &fl6);
2751         }
2752
2753         skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
2754         if (!skb) {
2755                 ip6_rt_put(rt);
2756                 err = -ENOBUFS;
2757                 goto errout;
2758         }
2759
2760         /* Reserve room for dummy headers, this skb can pass
2761            through good chunk of routing engine.
2762          */
2763         skb_reset_mac_header(skb);
2764         skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
2765
2766         skb_dst_set(skb, &rt->dst);
2767
2768         err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
2769                             RTM_NEWROUTE, NETLINK_CB(in_skb).portid,
2770                             nlh->nlmsg_seq, 0, 0, 0);
2771         if (err < 0) {
2772                 kfree_skb(skb);
2773                 goto errout;
2774         }
2775
2776         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
2777 errout:
2778         return err;
2779 }
2780
2781 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info)
2782 {
2783         struct sk_buff *skb;
2784         struct net *net = info->nl_net;
2785         u32 seq;
2786         int err;
2787
2788         err = -ENOBUFS;
2789         seq = info->nlh ? info->nlh->nlmsg_seq : 0;
2790
2791         skb = nlmsg_new(rt6_nlmsg_size(), gfp_any());
2792         if (!skb)
2793                 goto errout;
2794
2795         err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
2796                                 event, info->portid, seq, 0, 0, 0);
2797         if (err < 0) {
2798                 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
2799                 WARN_ON(err == -EMSGSIZE);
2800                 kfree_skb(skb);
2801                 goto errout;
2802         }
2803         rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
2804                     info->nlh, gfp_any());
2805         return;
2806 errout:
2807         if (err < 0)
2808                 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
2809 }
2810
2811 static int ip6_route_dev_notify(struct notifier_block *this,
2812                                 unsigned long event, void *ptr)
2813 {
2814         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2815         struct net *net = dev_net(dev);
2816
2817         if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
2818                 net->ipv6.ip6_null_entry->dst.dev = dev;
2819                 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
2820 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2821                 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
2822                 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
2823                 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
2824                 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
2825 #endif
2826         }
2827
2828         return NOTIFY_OK;
2829 }
2830
2831 /*
2832  *      /proc
2833  */
2834
2835 #ifdef CONFIG_PROC_FS
2836
2837 static const struct file_operations ipv6_route_proc_fops = {
2838         .owner          = THIS_MODULE,
2839         .open           = ipv6_route_open,
2840         .read           = seq_read,
2841         .llseek         = seq_lseek,
2842         .release        = seq_release_net,
2843 };
2844
2845 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
2846 {
2847         struct net *net = (struct net *)seq->private;
2848         seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
2849                    net->ipv6.rt6_stats->fib_nodes,
2850                    net->ipv6.rt6_stats->fib_route_nodes,
2851                    net->ipv6.rt6_stats->fib_rt_alloc,
2852                    net->ipv6.rt6_stats->fib_rt_entries,
2853                    net->ipv6.rt6_stats->fib_rt_cache,
2854                    dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
2855                    net->ipv6.rt6_stats->fib_discarded_routes);
2856
2857         return 0;
2858 }
2859
2860 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
2861 {
2862         return single_open_net(inode, file, rt6_stats_seq_show);
2863 }
2864
2865 static const struct file_operations rt6_stats_seq_fops = {
2866         .owner   = THIS_MODULE,
2867         .open    = rt6_stats_seq_open,
2868         .read    = seq_read,
2869         .llseek  = seq_lseek,
2870         .release = single_release_net,
2871 };
2872 #endif  /* CONFIG_PROC_FS */
2873
2874 #ifdef CONFIG_SYSCTL
2875
2876 static
2877 int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
2878                               void __user *buffer, size_t *lenp, loff_t *ppos)
2879 {
2880         struct net *net;
2881         int delay;
2882         if (!write)
2883                 return -EINVAL;
2884
2885         net = (struct net *)ctl->extra1;
2886         delay = net->ipv6.sysctl.flush_delay;
2887         proc_dointvec(ctl, write, buffer, lenp, ppos);
2888         fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
2889         return 0;
2890 }
2891
2892 struct ctl_table ipv6_route_table_template[] = {
2893         {
2894                 .procname       =       "flush",
2895                 .data           =       &init_net.ipv6.sysctl.flush_delay,
2896                 .maxlen         =       sizeof(int),
2897                 .mode           =       0200,
2898                 .proc_handler   =       ipv6_sysctl_rtcache_flush
2899         },
2900         {
2901                 .procname       =       "gc_thresh",
2902                 .data           =       &ip6_dst_ops_template.gc_thresh,
2903                 .maxlen         =       sizeof(int),
2904                 .mode           =       0644,
2905                 .proc_handler   =       proc_dointvec,
2906         },
2907         {
2908                 .procname       =       "max_size",
2909                 .data           =       &init_net.ipv6.sysctl.ip6_rt_max_size,
2910                 .maxlen         =       sizeof(int),
2911                 .mode           =       0644,
2912                 .proc_handler   =       proc_dointvec,
2913         },
2914         {
2915                 .procname       =       "gc_min_interval",
2916                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2917                 .maxlen         =       sizeof(int),
2918                 .mode           =       0644,
2919                 .proc_handler   =       proc_dointvec_jiffies,
2920         },
2921         {
2922                 .procname       =       "gc_timeout",
2923                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
2924                 .maxlen         =       sizeof(int),
2925                 .mode           =       0644,
2926                 .proc_handler   =       proc_dointvec_jiffies,
2927         },
2928         {
2929                 .procname       =       "gc_interval",
2930                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_interval,
2931                 .maxlen         =       sizeof(int),
2932                 .mode           =       0644,
2933                 .proc_handler   =       proc_dointvec_jiffies,
2934         },
2935         {
2936                 .procname       =       "gc_elasticity",
2937                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
2938                 .maxlen         =       sizeof(int),
2939                 .mode           =       0644,
2940                 .proc_handler   =       proc_dointvec,
2941         },
2942         {
2943                 .procname       =       "mtu_expires",
2944                 .data           =       &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
2945                 .maxlen         =       sizeof(int),
2946                 .mode           =       0644,
2947                 .proc_handler   =       proc_dointvec_jiffies,
2948         },
2949         {
2950                 .procname       =       "min_adv_mss",
2951                 .data           =       &init_net.ipv6.sysctl.ip6_rt_min_advmss,
2952                 .maxlen         =       sizeof(int),
2953                 .mode           =       0644,
2954                 .proc_handler   =       proc_dointvec,
2955         },
2956         {
2957                 .procname       =       "gc_min_interval_ms",
2958                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2959                 .maxlen         =       sizeof(int),
2960                 .mode           =       0644,
2961                 .proc_handler   =       proc_dointvec_ms_jiffies,
2962         },
2963         { }
2964 };
2965
2966 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
2967 {
2968         struct ctl_table *table;
2969
2970         table = kmemdup(ipv6_route_table_template,
2971                         sizeof(ipv6_route_table_template),
2972                         GFP_KERNEL);
2973
2974         if (table) {
2975                 table[0].data = &net->ipv6.sysctl.flush_delay;
2976                 table[0].extra1 = net;
2977                 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
2978                 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
2979                 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2980                 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
2981                 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
2982                 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
2983                 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
2984                 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
2985                 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2986
2987                 /* Don't export sysctls to unprivileged users */
2988                 if (net->user_ns != &init_user_ns)
2989                         table[0].procname = NULL;
2990         }
2991
2992         return table;
2993 }
2994 #endif
2995
2996 static int __net_init ip6_route_net_init(struct net *net)
2997 {
2998         int ret = -ENOMEM;
2999
3000         memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
3001                sizeof(net->ipv6.ip6_dst_ops));
3002
3003         if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
3004                 goto out_ip6_dst_ops;
3005
3006         net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
3007                                            sizeof(*net->ipv6.ip6_null_entry),
3008                                            GFP_KERNEL);
3009         if (!net->ipv6.ip6_null_entry)
3010                 goto out_ip6_dst_entries;
3011         net->ipv6.ip6_null_entry->dst.path =
3012                 (struct dst_entry *)net->ipv6.ip6_null_entry;
3013         net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3014         dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
3015                          ip6_template_metrics, true);
3016
3017 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3018         net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
3019                                                sizeof(*net->ipv6.ip6_prohibit_entry),
3020                                                GFP_KERNEL);
3021         if (!net->ipv6.ip6_prohibit_entry)
3022                 goto out_ip6_null_entry;
3023         net->ipv6.ip6_prohibit_entry->dst.path =
3024                 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
3025         net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3026         dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
3027                          ip6_template_metrics, true);
3028
3029         net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
3030                                                sizeof(*net->ipv6.ip6_blk_hole_entry),
3031                                                GFP_KERNEL);
3032         if (!net->ipv6.ip6_blk_hole_entry)
3033                 goto out_ip6_prohibit_entry;
3034         net->ipv6.ip6_blk_hole_entry->dst.path =
3035                 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
3036         net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3037         dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
3038                          ip6_template_metrics, true);
3039 #endif
3040
3041         net->ipv6.sysctl.flush_delay = 0;
3042         net->ipv6.sysctl.ip6_rt_max_size = 4096;
3043         net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
3044         net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
3045         net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
3046         net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
3047         net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
3048         net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
3049
3050         net->ipv6.ip6_rt_gc_expire = 30*HZ;
3051
3052         ret = 0;
3053 out:
3054         return ret;
3055
3056 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3057 out_ip6_prohibit_entry:
3058         kfree(net->ipv6.ip6_prohibit_entry);
3059 out_ip6_null_entry:
3060         kfree(net->ipv6.ip6_null_entry);
3061 #endif
3062 out_ip6_dst_entries:
3063         dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3064 out_ip6_dst_ops:
3065         goto out;
3066 }
3067
3068 static void __net_exit ip6_route_net_exit(struct net *net)
3069 {
3070         kfree(net->ipv6.ip6_null_entry);
3071 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3072         kfree(net->ipv6.ip6_prohibit_entry);
3073         kfree(net->ipv6.ip6_blk_hole_entry);
3074 #endif
3075         dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3076 }
3077
3078 static int __net_init ip6_route_net_init_late(struct net *net)
3079 {
3080 #ifdef CONFIG_PROC_FS
3081         proc_create("ipv6_route", 0, net->proc_net, &ipv6_route_proc_fops);
3082         proc_create("rt6_stats", S_IRUGO, net->proc_net, &rt6_stats_seq_fops);
3083 #endif
3084         return 0;
3085 }
3086
3087 static void __net_exit ip6_route_net_exit_late(struct net *net)
3088 {
3089 #ifdef CONFIG_PROC_FS
3090         remove_proc_entry("ipv6_route", net->proc_net);
3091         remove_proc_entry("rt6_stats", net->proc_net);
3092 #endif
3093 }
3094
3095 static struct pernet_operations ip6_route_net_ops = {
3096         .init = ip6_route_net_init,
3097         .exit = ip6_route_net_exit,
3098 };
3099
3100 static int __net_init ipv6_inetpeer_init(struct net *net)
3101 {
3102         struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
3103
3104         if (!bp)
3105                 return -ENOMEM;
3106         inet_peer_base_init(bp);
3107         net->ipv6.peers = bp;
3108         return 0;
3109 }
3110
3111 static void __net_exit ipv6_inetpeer_exit(struct net *net)
3112 {
3113         struct inet_peer_base *bp = net->ipv6.peers;
3114
3115         net->ipv6.peers = NULL;
3116         inetpeer_invalidate_tree(bp);
3117         kfree(bp);
3118 }
3119
3120 static struct pernet_operations ipv6_inetpeer_ops = {
3121         .init   =       ipv6_inetpeer_init,
3122         .exit   =       ipv6_inetpeer_exit,
3123 };
3124
3125 static struct pernet_operations ip6_route_net_late_ops = {
3126         .init = ip6_route_net_init_late,
3127         .exit = ip6_route_net_exit_late,
3128 };
3129
3130 static struct notifier_block ip6_route_dev_notifier = {
3131         .notifier_call = ip6_route_dev_notify,
3132         .priority = 0,
3133 };
3134
3135 int __init ip6_route_init(void)
3136 {
3137         int ret;
3138
3139         ret = -ENOMEM;
3140         ip6_dst_ops_template.kmem_cachep =
3141                 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
3142                                   SLAB_HWCACHE_ALIGN, NULL);
3143         if (!ip6_dst_ops_template.kmem_cachep)
3144                 goto out;
3145
3146         ret = dst_entries_init(&ip6_dst_blackhole_ops);
3147         if (ret)
3148                 goto out_kmem_cache;
3149
3150         ret = register_pernet_subsys(&ipv6_inetpeer_ops);
3151         if (ret)
3152                 goto out_dst_entries;
3153
3154         ret = register_pernet_subsys(&ip6_route_net_ops);
3155         if (ret)
3156                 goto out_register_inetpeer;
3157
3158         ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
3159
3160         /* Registering of the loopback is done before this portion of code,
3161          * the loopback reference in rt6_info will not be taken, do it
3162          * manually for init_net */
3163         init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
3164         init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3165   #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3166         init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
3167         init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3168         init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
3169         init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3170   #endif
3171         ret = fib6_init();
3172         if (ret)
3173                 goto out_register_subsys;
3174
3175         ret = xfrm6_init();
3176         if (ret)
3177                 goto out_fib6_init;
3178
3179         ret = fib6_rules_init();
3180         if (ret)
3181                 goto xfrm6_init;
3182
3183         ret = register_pernet_subsys(&ip6_route_net_late_ops);
3184         if (ret)
3185                 goto fib6_rules_init;
3186
3187         ret = -ENOBUFS;
3188         if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL, NULL) ||
3189             __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL, NULL) ||
3190             __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL, NULL))
3191                 goto out_register_late_subsys;
3192
3193         ret = register_netdevice_notifier(&ip6_route_dev_notifier);
3194         if (ret)
3195                 goto out_register_late_subsys;
3196
3197 out:
3198         return ret;
3199
3200 out_register_late_subsys:
3201         unregister_pernet_subsys(&ip6_route_net_late_ops);
3202 fib6_rules_init:
3203         fib6_rules_cleanup();
3204 xfrm6_init:
3205         xfrm6_fini();
3206 out_fib6_init:
3207         fib6_gc_cleanup();
3208 out_register_subsys:
3209         unregister_pernet_subsys(&ip6_route_net_ops);
3210 out_register_inetpeer:
3211         unregister_pernet_subsys(&ipv6_inetpeer_ops);
3212 out_dst_entries:
3213         dst_entries_destroy(&ip6_dst_blackhole_ops);
3214 out_kmem_cache:
3215         kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3216         goto out;
3217 }
3218
3219 void ip6_route_cleanup(void)
3220 {
3221         unregister_netdevice_notifier(&ip6_route_dev_notifier);
3222         unregister_pernet_subsys(&ip6_route_net_late_ops);
3223         fib6_rules_cleanup();
3224         xfrm6_fini();
3225         fib6_gc_cleanup();
3226         unregister_pernet_subsys(&ipv6_inetpeer_ops);
3227         unregister_pernet_subsys(&ip6_route_net_ops);
3228         dst_entries_destroy(&ip6_dst_blackhole_ops);
3229         kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3230 }