net/ipv6/route.c

   1 /*
   2  *      Linux INET6 implementation
   3  *      FIB front-end.
   4  *
   5  *      Authors:
   6  *      Pedro Roque             <roque@di.fc.ul.pt>
   7  *
   8  *      This program is free software; you can redistribute it and/or
   9  *      modify it under the terms of the GNU General Public License
  10  *      as published by the Free Software Foundation; either version
  11  *      2 of the License, or (at your option) any later version.
  12  */
  13
  14 /*      Changes:
  15  *
  16  *      YOSHIFUJI Hideaki @USAGI
  17  *              reworked default router selection.
  18  *              - respect outgoing interface
  19  *              - select from (probably) reachable routers (i.e.
  20  *              routers in REACHABLE, STALE, DELAY or PROBE states).
  21  *              - always select the same router if it is (probably)
  22  *              reachable.  otherwise, round-robin the list.
  23  *      Ville Nuorvala
  24  *              Fixed routing subtrees.
  25  */
  26
  27 #define pr_fmt(fmt) "IPv6: " fmt
  28
  29 #include <linux/capability.h>
  30 #include <linux/errno.h>
  31 #include <linux/export.h>
  32 #include <linux/types.h>
  33 #include <linux/times.h>
  34 #include <linux/socket.h>
  35 #include <linux/sockios.h>
  36 #include <linux/net.h>
  37 #include <linux/route.h>
  38 #include <linux/netdevice.h>
  39 #include <linux/in6.h>
  40 #include <linux/mroute6.h>
  41 #include <linux/init.h>
  42 #include <linux/if_arp.h>
  43 #include <linux/proc_fs.h>
  44 #include <linux/seq_file.h>
  45 #include <linux/nsproxy.h>
  46 #include <linux/slab.h>
  47 #include <net/net_namespace.h>
  48 #include <net/snmp.h>
  49 #include <net/ipv6.h>
  50 #include <net/ip6_fib.h>
  51 #include <net/ip6_route.h>
  52 #include <net/ndisc.h>
  53 #include <net/addrconf.h>
  54 #include <net/tcp.h>
  55 #include <linux/rtnetlink.h>
  56 #include <net/dst.h>
  57 #include <net/xfrm.h>
  58 #include <net/netevent.h>
  59 #include <net/netlink.h>
  60 #include <net/nexthop.h>
  61
  62 #include <asm/uaccess.h>
  63
  64 #ifdef CONFIG_SYSCTL
  65 #include <linux/sysctl.h>
  66 #endif
  67
  68 enum rt6_nud_state {
  69         RT6_NUD_FAIL_HARD = -3,
  70         RT6_NUD_FAIL_PROBE = -2,
  71         RT6_NUD_FAIL_DO_RR = -1,
  72         RT6_NUD_SUCCEED = 1
  73 };
  74
  75 static struct rt6_info *ip6_rt_copy(struct rt6_info *ort,
  76                                     const struct in6_addr *dest);
  77 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie);
  78 static unsigned int      ip6_default_advmss(const struct dst_entry *dst);
  79 static unsigned int      ip6_mtu(const struct dst_entry *dst);
  80 static struct dst_entry *ip6_negative_advice(struct dst_entry *);
  81 static void             ip6_dst_destroy(struct dst_entry *);
  82 static void             ip6_dst_ifdown(struct dst_entry *,
  83                                        struct net_device *dev, int how);
  84 static int               ip6_dst_gc(struct dst_ops *ops);
  85
  86 static int              ip6_pkt_discard(struct sk_buff *skb);
  87 static int              ip6_pkt_discard_out(struct sock *sk, struct sk_buff *skb);
  88 static int              ip6_pkt_prohibit(struct sk_buff *skb);
  89 static int              ip6_pkt_prohibit_out(struct sock *sk, struct sk_buff *skb);
  90 static void             ip6_link_failure(struct sk_buff *skb);
  91 static void             ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
  92                                            struct sk_buff *skb, u32 mtu);
  93 static void             rt6_do_redirect(struct dst_entry *dst, struct sock *sk,
  94                                         struct sk_buff *skb);
  95 static int rt6_score_route(struct rt6_info *rt, int oif, int strict);
  96
  97 #ifdef CONFIG_IPV6_ROUTE_INFO
  98 static struct rt6_info *rt6_add_route_info(struct net *net,
  99                                            const struct in6_addr *prefix, int prefixlen,
 100                                            const struct in6_addr *gwaddr, int ifindex,
 101                                            unsigned int pref);
 102 static struct rt6_info *rt6_get_route_info(struct net *net,
 103                                            const struct in6_addr *prefix, int prefixlen,
 104                                            const struct in6_addr *gwaddr, int ifindex);
 105 #endif
 106
 107 static void rt6_bind_peer(struct rt6_info *rt, int create)
 108 {
 109         struct inet_peer_base *base;
 110         struct inet_peer *peer;
 111
 112         base = inetpeer_base_ptr(rt->_rt6i_peer);
 113         if (!base)
 114                 return;
 115
 116         peer = inet_getpeer_v6(base, &rt->rt6i_dst.addr, create);
 117         if (peer) {
 118                 if (!rt6_set_peer(rt, peer))
 119                         inet_putpeer(peer);
 120         }
 121 }
 122
 123 static struct inet_peer *__rt6_get_peer(struct rt6_info *rt, int create)
 124 {
 125         if (rt6_has_peer(rt))
 126                 return rt6_peer_ptr(rt);
 127
 128         rt6_bind_peer(rt, create);
 129         return (rt6_has_peer(rt) ? rt6_peer_ptr(rt) : NULL);
 130 }
 131
 132 static struct inet_peer *rt6_get_peer_create(struct rt6_info *rt)
 133 {
 134         return __rt6_get_peer(rt, 1);
 135 }
 136
 137 static u32 *ipv6_cow_metrics(struct dst_entry *dst, unsigned long old)
 138 {
 139         struct rt6_info *rt = (struct rt6_info *) dst;
 140         struct inet_peer *peer;
 141         u32 *p = NULL;
 142
 143         if (!(rt->dst.flags & DST_HOST))
 144                 return NULL;
 145
 146         peer = rt6_get_peer_create(rt);
 147         if (peer) {
 148                 u32 *old_p = __DST_METRICS_PTR(old);
 149                 unsigned long prev, new;
 150
 151                 p = peer->metrics;
 152                 if (inet_metrics_new(peer) ||
 153                     (old & DST_METRICS_FORCE_OVERWRITE))
 154                         memcpy(p, old_p, sizeof(u32) * RTAX_MAX);
 155
 156                 new = (unsigned long) p;
 157                 prev = cmpxchg(&dst->_metrics, old, new);
 158
 159                 if (prev != old) {
 160                         p = __DST_METRICS_PTR(prev);
 161                         if (prev & DST_METRICS_READ_ONLY)
 162                                 p = NULL;
 163                 }
 164         }
 165         return p;
 166 }
 167
 168 static inline const void *choose_neigh_daddr(struct rt6_info *rt,
 169                                              struct sk_buff *skb,
 170                                              const void *daddr)
 171 {
 172         struct in6_addr *p = &rt->rt6i_gateway;
 173
 174         if (!ipv6_addr_any(p))
 175                 return (const void *) p;
 176         else if (skb)
 177                 return &ipv6_hdr(skb)->daddr;
 178         return daddr;
 179 }
 180
 181 static struct neighbour *ip6_neigh_lookup(const struct dst_entry *dst,
 182                                           struct sk_buff *skb,
 183                                           const void *daddr)
 184 {
 185         struct rt6_info *rt = (struct rt6_info *) dst;
 186         struct neighbour *n;
 187
 188         daddr = choose_neigh_daddr(rt, skb, daddr);
 189         n = __ipv6_neigh_lookup(dst->dev, daddr);
 190         if (n)
 191                 return n;
 192         return neigh_create(&nd_tbl, daddr, dst->dev);
 193 }
 194
 195 static struct dst_ops ip6_dst_ops_template = {
 196         .family                 =       AF_INET6,
 197         .protocol               =       cpu_to_be16(ETH_P_IPV6),
 198         .gc                     =       ip6_dst_gc,
 199         .gc_thresh              =       1024,
 200         .check                  =       ip6_dst_check,
 201         .default_advmss         =       ip6_default_advmss,
 202         .mtu                    =       ip6_mtu,
 203         .cow_metrics            =       ipv6_cow_metrics,
 204         .destroy                =       ip6_dst_destroy,
 205         .ifdown                 =       ip6_dst_ifdown,
 206         .negative_advice        =       ip6_negative_advice,
 207         .link_failure           =       ip6_link_failure,
 208         .update_pmtu            =       ip6_rt_update_pmtu,
 209         .redirect               =       rt6_do_redirect,
 210         .local_out              =       __ip6_local_out,
 211         .neigh_lookup           =       ip6_neigh_lookup,
 212 };
 213
 214 static unsigned int ip6_blackhole_mtu(const struct dst_entry *dst)
 215 {
 216         unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
 217
 218         return mtu ? : dst->dev->mtu;
 219 }
 220
 221 static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, struct sock *sk,
 222                                          struct sk_buff *skb, u32 mtu)
 223 {
 224 }
 225
 226 static void ip6_rt_blackhole_redirect(struct dst_entry *dst, struct sock *sk,
 227                                       struct sk_buff *skb)
 228 {
 229 }
 230
 231 static u32 *ip6_rt_blackhole_cow_metrics(struct dst_entry *dst,
 232                                          unsigned long old)
 233 {
 234         return NULL;
 235 }
 236
 237 static struct dst_ops ip6_dst_blackhole_ops = {
 238         .family                 =       AF_INET6,
 239         .protocol               =       cpu_to_be16(ETH_P_IPV6),
 240         .destroy                =       ip6_dst_destroy,
 241         .check                  =       ip6_dst_check,
 242         .mtu                    =       ip6_blackhole_mtu,
 243         .default_advmss         =       ip6_default_advmss,
 244         .update_pmtu            =       ip6_rt_blackhole_update_pmtu,
 245         .redirect               =       ip6_rt_blackhole_redirect,
 246         .cow_metrics            =       ip6_rt_blackhole_cow_metrics,
 247         .neigh_lookup           =       ip6_neigh_lookup,
 248 };
 249
 250 static const u32 ip6_template_metrics[RTAX_MAX] = {
 251         [RTAX_HOPLIMIT - 1] = 0,
 252 };
 253
 254 static const struct rt6_info ip6_null_entry_template = {
 255         .dst = {
 256                 .__refcnt       = ATOMIC_INIT(1),
 257                 .__use          = 1,
 258                 .obsolete       = DST_OBSOLETE_FORCE_CHK,
 259                 .error          = -ENETUNREACH,
 260                 .input          = ip6_pkt_discard,
 261                 .output         = ip6_pkt_discard_out,
 262         },
 263         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
 264         .rt6i_protocol  = RTPROT_KERNEL,
 265         .rt6i_metric    = ~(u32) 0,
 266         .rt6i_ref       = ATOMIC_INIT(1),
 267 };
 268
 269 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
 270
 271 static const struct rt6_info ip6_prohibit_entry_template = {
 272         .dst = {
 273                 .__refcnt       = ATOMIC_INIT(1),
 274                 .__use          = 1,
 275                 .obsolete       = DST_OBSOLETE_FORCE_CHK,
 276                 .error          = -EACCES,
 277                 .input          = ip6_pkt_prohibit,
 278                 .output         = ip6_pkt_prohibit_out,
 279         },
 280         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
 281         .rt6i_protocol  = RTPROT_KERNEL,
 282         .rt6i_metric    = ~(u32) 0,
 283         .rt6i_ref       = ATOMIC_INIT(1),
 284 };
 285
 286 static const struct rt6_info ip6_blk_hole_entry_template = {
 287         .dst = {
 288                 .__refcnt       = ATOMIC_INIT(1),
 289                 .__use          = 1,
 290                 .obsolete       = DST_OBSOLETE_FORCE_CHK,
 291                 .error          = -EINVAL,
 292                 .input          = dst_discard,
 293                 .output         = dst_discard_sk,
 294         },
 295         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
 296         .rt6i_protocol  = RTPROT_KERNEL,
 297         .rt6i_metric    = ~(u32) 0,
 298         .rt6i_ref       = ATOMIC_INIT(1),
 299 };
 300
 301 #endif
 302
 303 /* allocate dst with ip6_dst_ops */
 304 static inline struct rt6_info *ip6_dst_alloc(struct net *net,
 305                                              struct net_device *dev,
 306                                              int flags,
 307                                              struct fib6_table *table)
 308 {
 309         struct rt6_info *rt = dst_alloc(&net->ipv6.ip6_dst_ops, dev,
 310                                         0, DST_OBSOLETE_FORCE_CHK, flags);
 311
 312         if (rt) {
 313                 struct dst_entry *dst = &rt->dst;
 314
 315                 memset(dst + 1, 0, sizeof(*rt) - sizeof(*dst));
 316                 rt6_init_peer(rt, table ? &table->tb6_peers : net->ipv6.peers);
 317                 INIT_LIST_HEAD(&rt->rt6i_siblings);
 318         }
 319         return rt;
 320 }
 321
 322 static void ip6_dst_destroy(struct dst_entry *dst)
 323 {
 324         struct rt6_info *rt = (struct rt6_info *)dst;
 325         struct inet6_dev *idev = rt->rt6i_idev;
 326         struct dst_entry *from = dst->from;
 327
 328         if (!(rt->dst.flags & DST_HOST))
 329                 dst_destroy_metrics_generic(dst);
 330
 331         if (idev) {
 332                 rt->rt6i_idev = NULL;
 333                 in6_dev_put(idev);
 334         }
 335
 336         dst->from = NULL;
 337         dst_release(from);
 338
 339         if (rt6_has_peer(rt)) {
 340                 struct inet_peer *peer = rt6_peer_ptr(rt);
 341                 inet_putpeer(peer);
 342         }
 343 }
 344
 345 static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
 346                            int how)
 347 {
 348         struct rt6_info *rt = (struct rt6_info *)dst;
 349         struct inet6_dev *idev = rt->rt6i_idev;
 350         struct net_device *loopback_dev =
 351                 dev_net(dev)->loopback_dev;
 352
 353         if (dev != loopback_dev) {
 354                 if (idev && idev->dev == dev) {
 355                         struct inet6_dev *loopback_idev =
 356                                 in6_dev_get(loopback_dev);
 357                         if (loopback_idev) {
 358                                 rt->rt6i_idev = loopback_idev;
 359                                 in6_dev_put(idev);
 360                         }
 361                 }
 362         }
 363 }
 364
 365 static bool rt6_check_expired(const struct rt6_info *rt)
 366 {
 367         if (rt->rt6i_flags & RTF_EXPIRES) {
 368                 if (time_after(jiffies, rt->dst.expires))
 369                         return true;
 370         } else if (rt->dst.from) {
 371                 return rt6_check_expired((struct rt6_info *) rt->dst.from);
 372         }
 373         return false;
 374 }
 375
 376 /* Multipath route selection:
 377  *   Hash based function using packet header and flowlabel.
 378  * Adapted from fib_info_hashfn()
 379  */
 380 static int rt6_info_hash_nhsfn(unsigned int candidate_count,
 381                                const struct flowi6 *fl6)
 382 {
 383         unsigned int val = fl6->flowi6_proto;
 384
 385         val ^= ipv6_addr_hash(&fl6->daddr);
 386         val ^= ipv6_addr_hash(&fl6->saddr);
 387
 388         /* Work only if this not encapsulated */
 389         switch (fl6->flowi6_proto) {
 390         case IPPROTO_UDP:
 391         case IPPROTO_TCP:
 392         case IPPROTO_SCTP:
 393                 val ^= (__force u16)fl6->fl6_sport;
 394                 val ^= (__force u16)fl6->fl6_dport;
 395                 break;
 396
 397         case IPPROTO_ICMPV6:
 398                 val ^= (__force u16)fl6->fl6_icmp_type;
 399                 val ^= (__force u16)fl6->fl6_icmp_code;
 400                 break;
 401         }
 402         /* RFC6438 recommands to use flowlabel */
 403         val ^= (__force u32)fl6->flowlabel;
 404
 405         /* Perhaps, we need to tune, this function? */
 406         val = val ^ (val >> 7) ^ (val >> 12);
 407         return val % candidate_count;
 408 }
 409
 410 static struct rt6_info *rt6_multipath_select(struct rt6_info *match,
 411                                              struct flowi6 *fl6, int oif,
 412                                              int strict)
 413 {
 414         struct rt6_info *sibling, *next_sibling;
 415         int route_choosen;
 416
 417         route_choosen = rt6_info_hash_nhsfn(match->rt6i_nsiblings + 1, fl6);
 418         /* Don't change the route, if route_choosen == 0
 419          * (siblings does not include ourself)
 420          */
 421         if (route_choosen)
 422                 list_for_each_entry_safe(sibling, next_sibling,
 423                                 &match->rt6i_siblings, rt6i_siblings) {
 424                         route_choosen--;
 425                         if (route_choosen == 0) {
 426                                 if (rt6_score_route(sibling, oif, strict) < 0)
 427                                         break;
 428                                 match = sibling;
 429                                 break;
 430                         }
 431                 }
 432         return match;
 433 }
 434
 435 /*
 436  *      Route lookup. Any table->tb6_lock is implied.
 437  */
 438
 439 static inline struct rt6_info *rt6_device_match(struct net *net,
 440                                                     struct rt6_info *rt,
 441                                                     const struct in6_addr *saddr,
 442                                                     int oif,
 443                                                     int flags)
 444 {
 445         struct rt6_info *local = NULL;
 446         struct rt6_info *sprt;
 447
 448         if (!oif && ipv6_addr_any(saddr))
 449                 goto out;
 450
 451         for (sprt = rt; sprt; sprt = sprt->dst.rt6_next) {
 452                 struct net_device *dev = sprt->dst.dev;
 453
 454                 if (oif) {
 455                         if (dev->ifindex == oif)
 456                                 return sprt;
 457                         if (dev->flags & IFF_LOOPBACK) {
 458                                 if (!sprt->rt6i_idev ||
 459                                     sprt->rt6i_idev->dev->ifindex != oif) {
 460                                         if (flags & RT6_LOOKUP_F_IFACE && oif)
 461                                                 continue;
 462                                         if (local && (!oif ||
 463                                                       local->rt6i_idev->dev->ifindex == oif))
 464                                                 continue;
 465                                 }
 466                                 local = sprt;
 467                         }
 468                 } else {
 469                         if (ipv6_chk_addr(net, saddr, dev,
 470                                           flags & RT6_LOOKUP_F_IFACE))
 471                                 return sprt;
 472                 }
 473         }
 474
 475         if (oif) {
 476                 if (local)
 477                         return local;
 478
 479                 if (flags & RT6_LOOKUP_F_IFACE)
 480                         return net->ipv6.ip6_null_entry;
 481         }
 482 out:
 483         return rt;
 484 }
 485
 486 #ifdef CONFIG_IPV6_ROUTER_PREF
 487 struct __rt6_probe_work {
 488         struct work_struct work;
 489         struct in6_addr target;
 490         struct net_device *dev;
 491 };
 492
 493 static void rt6_probe_deferred(struct work_struct *w)
 494 {
 495         struct in6_addr mcaddr;
 496         struct __rt6_probe_work *work =
 497                 container_of(w, struct __rt6_probe_work, work);
 498
 499         addrconf_addr_solict_mult(&work->target, &mcaddr);
 500         ndisc_send_ns(work->dev, NULL, &work->target, &mcaddr, NULL);
 501         dev_put(work->dev);
 502         kfree(w);
 503 }
 504
 505 static void rt6_probe(struct rt6_info *rt)
 506 {
 507         struct neighbour *neigh;
 508         /*
 509          * Okay, this does not seem to be appropriate
 510          * for now, however, we need to check if it
 511          * is really so; aka Router Reachability Probing.
 512          *
 513          * Router Reachability Probe MUST be rate-limited
 514          * to no more than one per minute.
 515          */
 516         if (!rt || !(rt->rt6i_flags & RTF_GATEWAY))
 517                 return;
 518         rcu_read_lock_bh();
 519         neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
 520         if (neigh) {
 521                 write_lock(&neigh->lock);
 522                 if (neigh->nud_state & NUD_VALID)
 523                         goto out;
 524         }
 525
 526         if (!neigh ||
 527             time_after(jiffies, neigh->updated + rt->rt6i_idev->cnf.rtr_probe_interval)) {
 528                 struct __rt6_probe_work *work;
 529
 530                 work = kmalloc(sizeof(*work), GFP_ATOMIC);
 531
 532                 if (neigh && work)
 533                         __neigh_set_probe_once(neigh);
 534
 535                 if (neigh)
 536                         write_unlock(&neigh->lock);
 537
 538                 if (work) {
 539                         INIT_WORK(&work->work, rt6_probe_deferred);
 540                         work->target = rt->rt6i_gateway;
 541                         dev_hold(rt->dst.dev);
 542                         work->dev = rt->dst.dev;
 543                         schedule_work(&work->work);
 544                 }
 545         } else {
 546 out:
 547                 write_unlock(&neigh->lock);
 548         }
 549         rcu_read_unlock_bh();
 550 }
 551 #else
 552 static inline void rt6_probe(struct rt6_info *rt)
 553 {
 554 }
 555 #endif
 556
 557 /*
 558  * Default Router Selection (RFC 2461 6.3.6)
 559  */
 560 static inline int rt6_check_dev(struct rt6_info *rt, int oif)
 561 {
 562         struct net_device *dev = rt->dst.dev;
 563         if (!oif || dev->ifindex == oif)
 564                 return 2;
 565         if ((dev->flags & IFF_LOOPBACK) &&
 566             rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
 567                 return 1;
 568         return 0;
 569 }
 570
 571 static inline enum rt6_nud_state rt6_check_neigh(struct rt6_info *rt)
 572 {
 573         struct neighbour *neigh;
 574         enum rt6_nud_state ret = RT6_NUD_FAIL_HARD;
 575
 576         if (rt->rt6i_flags & RTF_NONEXTHOP ||
 577             !(rt->rt6i_flags & RTF_GATEWAY))
 578                 return RT6_NUD_SUCCEED;
 579
 580         rcu_read_lock_bh();
 581         neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
 582         if (neigh) {
 583                 read_lock(&neigh->lock);
 584                 if (neigh->nud_state & NUD_VALID)
 585                         ret = RT6_NUD_SUCCEED;
 586 #ifdef CONFIG_IPV6_ROUTER_PREF
 587                 else if (!(neigh->nud_state & NUD_FAILED))
 588                         ret = RT6_NUD_SUCCEED;
 589                 else
 590                         ret = RT6_NUD_FAIL_PROBE;
 591 #endif
 592                 read_unlock(&neigh->lock);
 593         } else {
 594                 ret = IS_ENABLED(CONFIG_IPV6_ROUTER_PREF) ?
 595                       RT6_NUD_SUCCEED : RT6_NUD_FAIL_DO_RR;
 596         }
 597         rcu_read_unlock_bh();
 598
 599         return ret;
 600 }
 601
 602 static int rt6_score_route(struct rt6_info *rt, int oif,
 603                            int strict)
 604 {
 605         int m;
 606
 607         m = rt6_check_dev(rt, oif);
 608         if (!m && (strict & RT6_LOOKUP_F_IFACE))
 609                 return RT6_NUD_FAIL_HARD;
 610 #ifdef CONFIG_IPV6_ROUTER_PREF
 611         m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
 612 #endif
 613         if (strict & RT6_LOOKUP_F_REACHABLE) {
 614                 int n = rt6_check_neigh(rt);
 615                 if (n < 0)
 616                         return n;
 617         }
 618         return m;
 619 }
 620
 621 static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
 622                                    int *mpri, struct rt6_info *match,
 623                                    bool *do_rr)
 624 {
 625         int m;
 626         bool match_do_rr = false;
 627
 628         if (rt6_check_expired(rt))
 629                 goto out;
 630
 631         m = rt6_score_route(rt, oif, strict);
 632         if (m == RT6_NUD_FAIL_DO_RR) {
 633                 match_do_rr = true;
 634                 m = 0; /* lowest valid score */
 635         } else if (m == RT6_NUD_FAIL_HARD) {
 636                 goto out;
 637         }
 638
 639         if (strict & RT6_LOOKUP_F_REACHABLE)
 640                 rt6_probe(rt);
 641
 642         /* note that m can be RT6_NUD_FAIL_PROBE at this point */
 643         if (m > *mpri) {
 644                 *do_rr = match_do_rr;
 645                 *mpri = m;
 646                 match = rt;
 647         }
 648 out:
 649         return match;
 650 }
 651
 652 static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
 653                                      struct rt6_info *rr_head,
 654                                      u32 metric, int oif, int strict,
 655                                      bool *do_rr)
 656 {
 657         struct rt6_info *rt, *match;
 658         int mpri = -1;
 659
 660         match = NULL;
 661         for (rt = rr_head; rt && rt->rt6i_metric == metric;
 662              rt = rt->dst.rt6_next)
 663                 match = find_match(rt, oif, strict, &mpri, match, do_rr);
 664         for (rt = fn->leaf; rt && rt != rr_head && rt->rt6i_metric == metric;
 665              rt = rt->dst.rt6_next)
 666                 match = find_match(rt, oif, strict, &mpri, match, do_rr);
 667
 668         return match;
 669 }
 670
 671 static struct rt6_info *rt6_select(struct fib6_node *fn, int oif, int strict)
 672 {
 673         struct rt6_info *match, *rt0;
 674         struct net *net;
 675         bool do_rr = false;
 676
 677         rt0 = fn->rr_ptr;
 678         if (!rt0)
 679                 fn->rr_ptr = rt0 = fn->leaf;
 680
 681         match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict,
 682                              &do_rr);
 683
 684         if (do_rr) {
 685                 struct rt6_info *next = rt0->dst.rt6_next;
 686
 687                 /* no entries matched; do round-robin */
 688                 if (!next || next->rt6i_metric != rt0->rt6i_metric)
 689                         next = fn->leaf;
 690
 691                 if (next != rt0)
 692                         fn->rr_ptr = next;
 693         }
 694
 695         net = dev_net(rt0->dst.dev);
 696         return match ? match : net->ipv6.ip6_null_entry;
 697 }
 698
 699 #ifdef CONFIG_IPV6_ROUTE_INFO
 700 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
 701                   const struct in6_addr *gwaddr)
 702 {
 703         struct net *net = dev_net(dev);
 704         struct route_info *rinfo = (struct route_info *) opt;
 705         struct in6_addr prefix_buf, *prefix;
 706         unsigned int pref;
 707         unsigned long lifetime;
 708         struct rt6_info *rt;
 709
 710         if (len < sizeof(struct route_info)) {
 711                 return -EINVAL;
 712         }
 713
 714         /* Sanity check for prefix_len and length */
 715         if (rinfo->length > 3) {
 716                 return -EINVAL;
 717         } else if (rinfo->prefix_len > 128) {
 718                 return -EINVAL;
 719         } else if (rinfo->prefix_len > 64) {
 720                 if (rinfo->length < 2) {
 721                         return -EINVAL;
 722                 }
 723         } else if (rinfo->prefix_len > 0) {
 724                 if (rinfo->length < 1) {
 725                         return -EINVAL;
 726                 }
 727         }
 728
 729         pref = rinfo->route_pref;
 730         if (pref == ICMPV6_ROUTER_PREF_INVALID)
 731                 return -EINVAL;
 732
 733         lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
 734
 735         if (rinfo->length == 3)
 736                 prefix = (struct in6_addr *)rinfo->prefix;
 737         else {
 738                 /* this function is safe */
 739                 ipv6_addr_prefix(&prefix_buf,
 740                                  (struct in6_addr *)rinfo->prefix,
 741                                  rinfo->prefix_len);
 742                 prefix = &prefix_buf;
 743         }
 744
 745         if (rinfo->prefix_len == 0)
 746                 rt = rt6_get_dflt_router(gwaddr, dev);
 747         else
 748                 rt = rt6_get_route_info(net, prefix, rinfo->prefix_len,
 749                                         gwaddr, dev->ifindex);
 750
 751         if (rt && !lifetime) {
 752                 ip6_del_rt(rt);
 753                 rt = NULL;
 754         }
 755
 756         if (!rt && lifetime)
 757                 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr, dev->ifindex,
 758                                         pref);
 759         else if (rt)
 760                 rt->rt6i_flags = RTF_ROUTEINFO |
 761                                  (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
 762
 763         if (rt) {
 764                 if (!addrconf_finite_timeout(lifetime))
 765                         rt6_clean_expires(rt);
 766                 else
 767                         rt6_set_expires(rt, jiffies + HZ * lifetime);
 768
 769                 ip6_rt_put(rt);
 770         }
 771         return 0;
 772 }
 773 #endif
 774
 775 static struct fib6_node* fib6_backtrack(struct fib6_node *fn,
 776                                         struct in6_addr *saddr)
 777 {
 778         struct fib6_node *pn;
 779         while (1) {
 780                 if (fn->fn_flags & RTN_TL_ROOT)
 781                         return NULL;
 782                 pn = fn->parent;
 783                 if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn)
 784                         fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr);
 785                 else
 786                         fn = pn;
 787                 if (fn->fn_flags & RTN_RTINFO)
 788                         return fn;
 789         }
 790 }
 791
 792 static struct rt6_info *ip6_pol_route_lookup(struct net *net,
 793                                              struct fib6_table *table,
 794                                              struct flowi6 *fl6, int flags)
 795 {
 796         struct fib6_node *fn;
 797         struct rt6_info *rt;
 798
 799         read_lock_bh(&table->tb6_lock);
 800         fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
 801 restart:
 802         rt = fn->leaf;
 803         rt = rt6_device_match(net, rt, &fl6->saddr, fl6->flowi6_oif, flags);
 804         if (rt->rt6i_nsiblings && fl6->flowi6_oif == 0)
 805                 rt = rt6_multipath_select(rt, fl6, fl6->flowi6_oif, flags);
 806         if (rt == net->ipv6.ip6_null_entry) {
 807                 fn = fib6_backtrack(fn, &fl6->saddr);
 808                 if (fn)
 809                         goto restart;
 810         }
 811         dst_use(&rt->dst, jiffies);
 812         read_unlock_bh(&table->tb6_lock);
 813         return rt;
 814
 815 }
 816
 817 struct dst_entry *ip6_route_lookup(struct net *net, struct flowi6 *fl6,
 818                                     int flags)
 819 {
 820         return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_lookup);
 821 }
 822 EXPORT_SYMBOL_GPL(ip6_route_lookup);
 823
 824 struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
 825                             const struct in6_addr *saddr, int oif, int strict)
 826 {
 827         struct flowi6 fl6 = {
 828                 .flowi6_oif = oif,
 829                 .daddr = *daddr,
 830         };
 831         struct dst_entry *dst;
 832         int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
 833
 834         if (saddr) {
 835                 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
 836                 flags |= RT6_LOOKUP_F_HAS_SADDR;
 837         }
 838
 839         dst = fib6_rule_lookup(net, &fl6, flags, ip6_pol_route_lookup);
 840         if (dst->error == 0)
 841                 return (struct rt6_info *) dst;
 842
 843         dst_release(dst);
 844
 845         return NULL;
 846 }
 847 EXPORT_SYMBOL(rt6_lookup);
 848
 849 /* ip6_ins_rt is called with FREE table->tb6_lock.
 850    It takes new route entry, the addition fails by any reason the
 851    route is freed. In any case, if caller does not hold it, it may
 852    be destroyed.
 853  */
 854
 855 static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info,
 856                         struct nlattr *mx, int mx_len)
 857 {
 858         int err;
 859         struct fib6_table *table;
 860
 861         table = rt->rt6i_table;
 862         write_lock_bh(&table->tb6_lock);
 863         err = fib6_add(&table->tb6_root, rt, info, mx, mx_len);
 864         write_unlock_bh(&table->tb6_lock);
 865
 866         return err;
 867 }
 868
 869 int ip6_ins_rt(struct rt6_info *rt)
 870 {
 871         struct nl_info info = {
 872                 .nl_net = dev_net(rt->dst.dev),
 873         };
 874         return __ip6_ins_rt(rt, &info, NULL, 0);
 875 }
 876
 877 static struct rt6_info *rt6_alloc_cow(struct rt6_info *ort,
 878                                       const struct in6_addr *daddr,
 879                                       const struct in6_addr *saddr)
 880 {
 881         struct rt6_info *rt;
 882
 883         /*
 884          *      Clone the route.
 885          */
 886
 887         rt = ip6_rt_copy(ort, daddr);
 888
 889         if (rt) {
 890                 if (ort->rt6i_dst.plen != 128 &&
 891                     ipv6_addr_equal(&ort->rt6i_dst.addr, daddr))
 892                         rt->rt6i_flags |= RTF_ANYCAST;
 893
 894                 rt->rt6i_flags |= RTF_CACHE;
 895
 896 #ifdef CONFIG_IPV6_SUBTREES
 897                 if (rt->rt6i_src.plen && saddr) {
 898                         rt->rt6i_src.addr = *saddr;
 899                         rt->rt6i_src.plen = 128;
 900                 }
 901 #endif
 902         }
 903
 904         return rt;
 905 }
 906
 907 static struct rt6_info *rt6_alloc_clone(struct rt6_info *ort,
 908                                         const struct in6_addr *daddr)
 909 {
 910         struct rt6_info *rt = ip6_rt_copy(ort, daddr);
 911
 912         if (rt)
 913                 rt->rt6i_flags |= RTF_CACHE;
 914         return rt;
 915 }
 916
 917 static struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table, int oif,
 918                                       struct flowi6 *fl6, int flags)
 919 {
 920         struct fib6_node *fn, *saved_fn;
 921         struct rt6_info *rt, *nrt;
 922         int strict = 0;
 923         int attempts = 3;
 924         int err;
 925
 926         strict |= flags & RT6_LOOKUP_F_IFACE;
 927         if (net->ipv6.devconf_all->forwarding == 0)
 928                 strict |= RT6_LOOKUP_F_REACHABLE;
 929
 930 redo_fib6_lookup_lock:
 931         read_lock_bh(&table->tb6_lock);
 932
 933         fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
 934         saved_fn = fn;
 935
 936 redo_rt6_select:
 937         rt = rt6_select(fn, oif, strict);
 938         if (rt->rt6i_nsiblings)
 939                 rt = rt6_multipath_select(rt, fl6, oif, strict);
 940         if (rt == net->ipv6.ip6_null_entry) {
 941                 fn = fib6_backtrack(fn, &fl6->saddr);
 942                 if (fn)
 943                         goto redo_rt6_select;
 944                 else if (strict & RT6_LOOKUP_F_REACHABLE) {
 945                         /* also consider unreachable route */
 946                         strict &= ~RT6_LOOKUP_F_REACHABLE;
 947                         fn = saved_fn;
 948                         goto redo_rt6_select;
 949                 } else {
 950                         dst_hold(&rt->dst);
 951                         read_unlock_bh(&table->tb6_lock);
 952                         goto out2;
 953                 }
 954         }
 955
 956         dst_hold(&rt->dst);
 957         read_unlock_bh(&table->tb6_lock);
 958
 959         if (rt->rt6i_flags & RTF_CACHE)
 960                 goto out2;
 961
 962         if (!(rt->rt6i_flags & (RTF_NONEXTHOP | RTF_GATEWAY)))
 963                 nrt = rt6_alloc_cow(rt, &fl6->daddr, &fl6->saddr);
 964         else if (!(rt->dst.flags & DST_HOST))
 965                 nrt = rt6_alloc_clone(rt, &fl6->daddr);
 966         else
 967                 goto out2;
 968
 969         ip6_rt_put(rt);
 970         rt = nrt ? : net->ipv6.ip6_null_entry;
 971
 972         dst_hold(&rt->dst);
 973         if (nrt) {
 974                 err = ip6_ins_rt(nrt);
 975                 if (!err)
 976                         goto out2;
 977         }
 978
 979         if (--attempts <= 0)
 980                 goto out2;
 981
 982         /*
 983          * Race condition! In the gap, when table->tb6_lock was
 984          * released someone could insert this route.  Relookup.
 985          */
 986         ip6_rt_put(rt);
 987         goto redo_fib6_lookup_lock;
 988
 989 out2:
 990         rt->dst.lastuse = jiffies;
 991         rt->dst.__use++;
 992
 993         return rt;
 994 }
 995
 996 static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
 997                                             struct flowi6 *fl6, int flags)
 998 {
 999         return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, flags);
1000 }
1001
1002 static struct dst_entry *ip6_route_input_lookup(struct net *net,
1003                                                 struct net_device *dev,
1004                                                 struct flowi6 *fl6, int flags)
1005 {
1006         if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
1007                 flags |= RT6_LOOKUP_F_IFACE;
1008
1009         return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_input);
1010 }
1011
1012 void ip6_route_input(struct sk_buff *skb)
1013 {
1014         const struct ipv6hdr *iph = ipv6_hdr(skb);
1015         struct net *net = dev_net(skb->dev);
1016         int flags = RT6_LOOKUP_F_HAS_SADDR;
1017         struct flowi6 fl6 = {
1018                 .flowi6_iif = skb->dev->ifindex,
1019                 .daddr = iph->daddr,
1020                 .saddr = iph->saddr,
1021                 .flowlabel = ip6_flowinfo(iph),
1022                 .flowi6_mark = skb->mark,
1023                 .flowi6_proto = iph->nexthdr,
1024         };
1025
1026         skb_dst_set(skb, ip6_route_input_lookup(net, skb->dev, &fl6, flags));
1027 }
1028
1029 static struct rt6_info *ip6_pol_route_output(struct net *net, struct fib6_table *table,
1030                                              struct flowi6 *fl6, int flags)
1031 {
1032         return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, flags);
1033 }
1034
1035 struct dst_entry *ip6_route_output(struct net *net, const struct sock *sk,
1036                                     struct flowi6 *fl6)
1037 {
1038         int flags = 0;
1039
1040         fl6->flowi6_iif = LOOPBACK_IFINDEX;
1041
1042         if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr))
1043                 flags |= RT6_LOOKUP_F_IFACE;
1044
1045         if (!ipv6_addr_any(&fl6->saddr))
1046                 flags |= RT6_LOOKUP_F_HAS_SADDR;
1047         else if (sk)
1048                 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
1049
1050         return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_output);
1051 }
1052 EXPORT_SYMBOL(ip6_route_output);
1053
1054 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
1055 {
1056         struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
1057         struct dst_entry *new = NULL;
1058
1059         rt = dst_alloc(&ip6_dst_blackhole_ops, ort->dst.dev, 1, DST_OBSOLETE_NONE, 0);
1060         if (rt) {
1061                 new = &rt->dst;
1062
1063                 memset(new + 1, 0, sizeof(*rt) - sizeof(*new));
1064                 rt6_init_peer(rt, net->ipv6.peers);
1065
1066                 new->__use = 1;
1067                 new->input = dst_discard;
1068                 new->output = dst_discard_sk;
1069
1070                 if (dst_metrics_read_only(&ort->dst))
1071                         new->_metrics = ort->dst._metrics;
1072                 else
1073                         dst_copy_metrics(new, &ort->dst);
1074                 rt->rt6i_idev = ort->rt6i_idev;
1075                 if (rt->rt6i_idev)
1076                         in6_dev_hold(rt->rt6i_idev);
1077
1078                 rt->rt6i_gateway = ort->rt6i_gateway;
1079                 rt->rt6i_flags = ort->rt6i_flags;
1080                 rt->rt6i_metric = 0;
1081
1082                 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1083 #ifdef CONFIG_IPV6_SUBTREES
1084                 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1085 #endif
1086
1087                 dst_free(new);
1088         }
1089
1090         dst_release(dst_orig);
1091         return new ? new : ERR_PTR(-ENOMEM);
1092 }
1093
1094 /*
1095  *      Destination cache support functions
1096  */
1097
1098 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
1099 {
1100         struct rt6_info *rt;
1101
1102         rt = (struct rt6_info *) dst;
1103
1104         /* All IPV6 dsts are created with ->obsolete set to the value
1105          * DST_OBSOLETE_FORCE_CHK which forces validation calls down
1106          * into this function always.
1107          */
1108         if (!rt->rt6i_node || (rt->rt6i_node->fn_sernum != cookie))
1109                 return NULL;
1110
1111         if (rt6_check_expired(rt))
1112                 return NULL;
1113
1114         return dst;
1115 }
1116
1117 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
1118 {
1119         struct rt6_info *rt = (struct rt6_info *) dst;
1120
1121         if (rt) {
1122                 if (rt->rt6i_flags & RTF_CACHE) {
1123                         if (rt6_check_expired(rt)) {
1124                                 ip6_del_rt(rt);
1125                                 dst = NULL;
1126                         }
1127                 } else {
1128                         dst_release(dst);
1129                         dst = NULL;
1130                 }
1131         }
1132         return dst;
1133 }
1134
1135 static void ip6_link_failure(struct sk_buff *skb)
1136 {
1137         struct rt6_info *rt;
1138
1139         icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
1140
1141         rt = (struct rt6_info *) skb_dst(skb);
1142         if (rt) {
1143                 if (rt->rt6i_flags & RTF_CACHE) {
1144                         dst_hold(&rt->dst);
1145                         if (ip6_del_rt(rt))
1146                                 dst_free(&rt->dst);
1147                 } else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT)) {
1148                         rt->rt6i_node->fn_sernum = -1;
1149                 }
1150         }
1151 }
1152
1153 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
1154                                struct sk_buff *skb, u32 mtu)
1155 {
1156         struct rt6_info *rt6 = (struct rt6_info *)dst;
1157
1158         dst_confirm(dst);
1159         if (mtu < dst_mtu(dst) && rt6->rt6i_dst.plen == 128) {
1160                 struct net *net = dev_net(dst->dev);
1161
1162                 rt6->rt6i_flags |= RTF_MODIFIED;
1163                 if (mtu < IPV6_MIN_MTU)
1164                         mtu = IPV6_MIN_MTU;
1165
1166                 dst_metric_set(dst, RTAX_MTU, mtu);
1167                 rt6_update_expires(rt6, net->ipv6.sysctl.ip6_rt_mtu_expires);
1168         }
1169 }
1170
1171 void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
1172                      int oif, u32 mark)
1173 {
1174         const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1175         struct dst_entry *dst;
1176         struct flowi6 fl6;
1177
1178         memset(&fl6, 0, sizeof(fl6));
1179         fl6.flowi6_oif = oif;
1180         fl6.flowi6_mark = mark ? mark : IP6_REPLY_MARK(net, skb->mark);
1181         fl6.daddr = iph->daddr;
1182         fl6.saddr = iph->saddr;
1183         fl6.flowlabel = ip6_flowinfo(iph);
1184
1185         dst = ip6_route_output(net, NULL, &fl6);
1186         if (!dst->error)
1187                 ip6_rt_update_pmtu(dst, NULL, skb, ntohl(mtu));
1188         dst_release(dst);
1189 }
1190 EXPORT_SYMBOL_GPL(ip6_update_pmtu);
1191
1192 void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
1193 {
1194         ip6_update_pmtu(skb, sock_net(sk), mtu,
1195                         sk->sk_bound_dev_if, sk->sk_mark);
1196 }
1197 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
1198
1199 /* Handle redirects */
1200 struct ip6rd_flowi {
1201         struct flowi6 fl6;
1202         struct in6_addr gateway;
1203 };
1204
1205 static struct rt6_info *__ip6_route_redirect(struct net *net,
1206                                              struct fib6_table *table,
1207                                              struct flowi6 *fl6,
1208                                              int flags)
1209 {
1210         struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
1211         struct rt6_info *rt;
1212         struct fib6_node *fn;
1213
1214         /* Get the "current" route for this destination and
1215          * check if the redirect has come from approriate router.
1216          *
1217          * RFC 4861 specifies that redirects should only be
1218          * accepted if they come from the nexthop to the target.
1219          * Due to the way the routes are chosen, this notion
1220          * is a bit fuzzy and one might need to check all possible
1221          * routes.
1222          */
1223
1224         read_lock_bh(&table->tb6_lock);
1225         fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1226 restart:
1227         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1228                 if (rt6_check_expired(rt))
1229                         continue;
1230                 if (rt->dst.error)
1231                         break;
1232                 if (!(rt->rt6i_flags & RTF_GATEWAY))
1233                         continue;
1234                 if (fl6->flowi6_oif != rt->dst.dev->ifindex)
1235                         continue;
1236                 if (!ipv6_addr_equal(&rdfl->gateway, &rt->rt6i_gateway))
1237                         continue;
1238                 break;
1239         }
1240
1241         if (!rt)
1242                 rt = net->ipv6.ip6_null_entry;
1243         else if (rt->dst.error) {
1244                 rt = net->ipv6.ip6_null_entry;
1245         } else if (rt == net->ipv6.ip6_null_entry) {
1246                 fn = fib6_backtrack(fn, &fl6->saddr);
1247                 if (fn)
1248                         goto restart;
1249         }
1250
1251         dst_hold(&rt->dst);
1252
1253         read_unlock_bh(&table->tb6_lock);
1254
1255         return rt;
1256 };
1257
1258 static struct dst_entry *ip6_route_redirect(struct net *net,
1259                                         const struct flowi6 *fl6,
1260                                         const struct in6_addr *gateway)
1261 {
1262         int flags = RT6_LOOKUP_F_HAS_SADDR;
1263         struct ip6rd_flowi rdfl;
1264
1265         rdfl.fl6 = *fl6;
1266         rdfl.gateway = *gateway;
1267
1268         return fib6_rule_lookup(net, &rdfl.fl6,
1269                                 flags, __ip6_route_redirect);
1270 }
1271
1272 void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark)
1273 {
1274         const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1275         struct dst_entry *dst;
1276         struct flowi6 fl6;
1277
1278         memset(&fl6, 0, sizeof(fl6));
1279         fl6.flowi6_iif = LOOPBACK_IFINDEX;
1280         fl6.flowi6_oif = oif;
1281         fl6.flowi6_mark = mark;
1282         fl6.daddr = iph->daddr;
1283         fl6.saddr = iph->saddr;
1284         fl6.flowlabel = ip6_flowinfo(iph);
1285
1286         dst = ip6_route_redirect(net, &fl6, &ipv6_hdr(skb)->saddr);
1287         rt6_do_redirect(dst, NULL, skb);
1288         dst_release(dst);
1289 }
1290 EXPORT_SYMBOL_GPL(ip6_redirect);
1291
1292 void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif,
1293                             u32 mark)
1294 {
1295         const struct ipv6hdr *iph = ipv6_hdr(skb);
1296         const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
1297         struct dst_entry *dst;
1298         struct flowi6 fl6;
1299
1300         memset(&fl6, 0, sizeof(fl6));
1301         fl6.flowi6_iif = LOOPBACK_IFINDEX;
1302         fl6.flowi6_oif = oif;
1303         fl6.flowi6_mark = mark;
1304         fl6.daddr = msg->dest;
1305         fl6.saddr = iph->daddr;
1306
1307         dst = ip6_route_redirect(net, &fl6, &iph->saddr);
1308         rt6_do_redirect(dst, NULL, skb);
1309         dst_release(dst);
1310 }
1311
1312 void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
1313 {
1314         ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark);
1315 }
1316 EXPORT_SYMBOL_GPL(ip6_sk_redirect);
1317
1318 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1319 {
1320         struct net_device *dev = dst->dev;
1321         unsigned int mtu = dst_mtu(dst);
1322         struct net *net = dev_net(dev);
1323
1324         mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
1325
1326         if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
1327                 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1328
1329         /*
1330          * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1331          * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1332          * IPV6_MAXPLEN is also valid and means: "any MSS,
1333          * rely only on pmtu discovery"
1334          */
1335         if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
1336                 mtu = IPV6_MAXPLEN;
1337         return mtu;
1338 }
1339
1340 static unsigned int ip6_mtu(const struct dst_entry *dst)
1341 {
1342         struct inet6_dev *idev;
1343         unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
1344
1345         if (mtu)
1346                 goto out;
1347
1348         mtu = IPV6_MIN_MTU;
1349
1350         rcu_read_lock();
1351         idev = __in6_dev_get(dst->dev);
1352         if (idev)
1353                 mtu = idev->cnf.mtu6;
1354         rcu_read_unlock();
1355
1356 out:
1357         return min_t(unsigned int, mtu, IP6_MAX_MTU);
1358 }
1359
1360 static struct dst_entry *icmp6_dst_gc_list;
1361 static DEFINE_SPINLOCK(icmp6_dst_lock);
1362
1363 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
1364                                   struct flowi6 *fl6)
1365 {
1366         struct dst_entry *dst;
1367         struct rt6_info *rt;
1368         struct inet6_dev *idev = in6_dev_get(dev);
1369         struct net *net = dev_net(dev);
1370
1371         if (unlikely(!idev))
1372                 return ERR_PTR(-ENODEV);
1373
1374         rt = ip6_dst_alloc(net, dev, 0, NULL);
1375         if (unlikely(!rt)) {
1376                 in6_dev_put(idev);
1377                 dst = ERR_PTR(-ENOMEM);
1378                 goto out;
1379         }
1380
1381         rt->dst.flags |= DST_HOST;
1382         rt->dst.output  = ip6_output;
1383         atomic_set(&rt->dst.__refcnt, 1);
1384         rt->rt6i_gateway  = fl6->daddr;
1385         rt->rt6i_dst.addr = fl6->daddr;
1386         rt->rt6i_dst.plen = 128;
1387         rt->rt6i_idev     = idev;
1388         dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
1389
1390         spin_lock_bh(&icmp6_dst_lock);
1391         rt->dst.next = icmp6_dst_gc_list;
1392         icmp6_dst_gc_list = &rt->dst;
1393         spin_unlock_bh(&icmp6_dst_lock);
1394
1395         fib6_force_start_gc(net);
1396
1397         dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
1398
1399 out:
1400         return dst;
1401 }
1402
1403 int icmp6_dst_gc(void)
1404 {
1405         struct dst_entry *dst, **pprev;
1406         int more = 0;
1407
1408         spin_lock_bh(&icmp6_dst_lock);
1409         pprev = &icmp6_dst_gc_list;
1410
1411         while ((dst = *pprev) != NULL) {
1412                 if (!atomic_read(&dst->__refcnt)) {
1413                         *pprev = dst->next;
1414                         dst_free(dst);
1415                 } else {
1416                         pprev = &dst->next;
1417                         ++more;
1418                 }
1419         }
1420
1421         spin_unlock_bh(&icmp6_dst_lock);
1422
1423         return more;
1424 }
1425
1426 static void icmp6_clean_all(int (*func)(struct rt6_info *rt, void *arg),
1427                             void *arg)
1428 {
1429         struct dst_entry *dst, **pprev;
1430
1431         spin_lock_bh(&icmp6_dst_lock);
1432         pprev = &icmp6_dst_gc_list;
1433         while ((dst = *pprev) != NULL) {
1434                 struct rt6_info *rt = (struct rt6_info *) dst;
1435                 if (func(rt, arg)) {
1436                         *pprev = dst->next;
1437                         dst_free(dst);
1438                 } else {
1439                         pprev = &dst->next;
1440                 }
1441         }
1442         spin_unlock_bh(&icmp6_dst_lock);
1443 }
1444
1445 static int ip6_dst_gc(struct dst_ops *ops)
1446 {
1447         struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
1448         int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1449         int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1450         int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1451         int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1452         unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1453         int entries;
1454
1455         entries = dst_entries_get_fast(ops);
1456         if (time_after(rt_last_gc + rt_min_interval, jiffies) &&
1457             entries <= rt_max_size)
1458                 goto out;
1459
1460         net->ipv6.ip6_rt_gc_expire++;
1461         fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net, true);
1462         entries = dst_entries_get_slow(ops);
1463         if (entries < ops->gc_thresh)
1464                 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1465 out:
1466         net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1467         return entries > rt_max_size;
1468 }
1469
1470 /*
1471  *
1472  */
1473
1474 int ip6_route_add(struct fib6_config *cfg)
1475 {
1476         int err;
1477         struct net *net = cfg->fc_nlinfo.nl_net;
1478         struct rt6_info *rt = NULL;
1479         struct net_device *dev = NULL;
1480         struct inet6_dev *idev = NULL;
1481         struct fib6_table *table;
1482         int addr_type;
1483
1484         if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1485                 return -EINVAL;
1486 #ifndef CONFIG_IPV6_SUBTREES
1487         if (cfg->fc_src_len)
1488                 return -EINVAL;
1489 #endif
1490         if (cfg->fc_ifindex) {
1491                 err = -ENODEV;
1492                 dev = dev_get_by_index(net, cfg->fc_ifindex);
1493                 if (!dev)
1494                         goto out;
1495                 idev = in6_dev_get(dev);
1496                 if (!idev)
1497                         goto out;
1498         }
1499
1500         if (cfg->fc_metric == 0)
1501                 cfg->fc_metric = IP6_RT_PRIO_USER;
1502
1503         err = -ENOBUFS;
1504         if (cfg->fc_nlinfo.nlh &&
1505             !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
1506                 table = fib6_get_table(net, cfg->fc_table);
1507                 if (!table) {
1508                         pr_warn("NLM_F_CREATE should be specified when creating new route\n");
1509                         table = fib6_new_table(net, cfg->fc_table);
1510                 }
1511         } else {
1512                 table = fib6_new_table(net, cfg->fc_table);
1513         }
1514
1515         if (!table)
1516                 goto out;
1517
1518         rt = ip6_dst_alloc(net, NULL, (cfg->fc_flags & RTF_ADDRCONF) ? 0 : DST_NOCOUNT, table);
1519
1520         if (!rt) {
1521                 err = -ENOMEM;
1522                 goto out;
1523         }
1524
1525         if (cfg->fc_flags & RTF_EXPIRES)
1526                 rt6_set_expires(rt, jiffies +
1527                                 clock_t_to_jiffies(cfg->fc_expires));
1528         else
1529                 rt6_clean_expires(rt);
1530
1531         if (cfg->fc_protocol == RTPROT_UNSPEC)
1532                 cfg->fc_protocol = RTPROT_BOOT;
1533         rt->rt6i_protocol = cfg->fc_protocol;
1534
1535         addr_type = ipv6_addr_type(&cfg->fc_dst);
1536
1537         if (addr_type & IPV6_ADDR_MULTICAST)
1538                 rt->dst.input = ip6_mc_input;
1539         else if (cfg->fc_flags & RTF_LOCAL)
1540                 rt->dst.input = ip6_input;
1541         else
1542                 rt->dst.input = ip6_forward;
1543
1544         rt->dst.output = ip6_output;
1545
1546         ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1547         rt->rt6i_dst.plen = cfg->fc_dst_len;
1548         if (rt->rt6i_dst.plen == 128) {
1549                 rt->dst.flags |= DST_HOST;
1550                 dst_metrics_set_force_overwrite(&rt->dst);
1551         }
1552
1553 #ifdef CONFIG_IPV6_SUBTREES
1554         ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1555         rt->rt6i_src.plen = cfg->fc_src_len;
1556 #endif
1557
1558         rt->rt6i_metric = cfg->fc_metric;
1559
1560         /* We cannot add true routes via loopback here,
1561            they would result in kernel looping; promote them to reject routes
1562          */
1563         if ((cfg->fc_flags & RTF_REJECT) ||
1564             (dev && (dev->flags & IFF_LOOPBACK) &&
1565              !(addr_type & IPV6_ADDR_LOOPBACK) &&
1566              !(cfg->fc_flags & RTF_LOCAL))) {
1567                 /* hold loopback dev/idev if we haven't done so. */
1568                 if (dev != net->loopback_dev) {
1569                         if (dev) {
1570                                 dev_put(dev);
1571                                 in6_dev_put(idev);
1572                         }
1573                         dev = net->loopback_dev;
1574                         dev_hold(dev);
1575                         idev = in6_dev_get(dev);
1576                         if (!idev) {
1577                                 err = -ENODEV;
1578                                 goto out;
1579                         }
1580                 }
1581                 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1582                 switch (cfg->fc_type) {
1583                 case RTN_BLACKHOLE:
1584                         rt->dst.error = -EINVAL;
1585                         rt->dst.output = dst_discard_sk;
1586                         rt->dst.input = dst_discard;
1587                         break;
1588                 case RTN_PROHIBIT:
1589                         rt->dst.error = -EACCES;
1590                         rt->dst.output = ip6_pkt_prohibit_out;
1591                         rt->dst.input = ip6_pkt_prohibit;
1592                         break;
1593                 case RTN_THROW:
1594                 default:
1595                         rt->dst.error = (cfg->fc_type == RTN_THROW) ? -EAGAIN
1596                                         : -ENETUNREACH;
1597                         rt->dst.output = ip6_pkt_discard_out;
1598                         rt->dst.input = ip6_pkt_discard;
1599                         break;
1600                 }
1601                 goto install_route;
1602         }
1603
1604         if (cfg->fc_flags & RTF_GATEWAY) {
1605                 const struct in6_addr *gw_addr;
1606                 int gwa_type;
1607
1608                 gw_addr = &cfg->fc_gateway;
1609                 rt->rt6i_gateway = *gw_addr;
1610                 gwa_type = ipv6_addr_type(gw_addr);
1611
1612                 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1613                         struct rt6_info *grt;
1614
1615                         /* IPv6 strictly inhibits using not link-local
1616                            addresses as nexthop address.
1617                            Otherwise, router will not able to send redirects.
1618                            It is very good, but in some (rare!) circumstances
1619                            (SIT, PtP, NBMA NOARP links) it is handy to allow
1620                            some exceptions. --ANK
1621                          */
1622                         err = -EINVAL;
1623                         if (!(gwa_type & IPV6_ADDR_UNICAST))
1624                                 goto out;
1625
1626                         grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, 1);
1627
1628                         err = -EHOSTUNREACH;
1629                         if (!grt)
1630                                 goto out;
1631                         if (dev) {
1632                                 if (dev != grt->dst.dev) {
1633                                         ip6_rt_put(grt);
1634                                         goto out;
1635                                 }
1636                         } else {
1637                                 dev = grt->dst.dev;
1638                                 idev = grt->rt6i_idev;
1639                                 dev_hold(dev);
1640                                 in6_dev_hold(grt->rt6i_idev);
1641                         }
1642                         if (!(grt->rt6i_flags & RTF_GATEWAY))
1643                                 err = 0;
1644                         ip6_rt_put(grt);
1645
1646                         if (err)
1647                                 goto out;
1648                 }
1649                 err = -EINVAL;
1650                 if (!dev || (dev->flags & IFF_LOOPBACK))
1651                         goto out;
1652         }
1653
1654         err = -ENODEV;
1655         if (!dev)
1656                 goto out;
1657
1658         if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
1659                 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
1660                         err = -EINVAL;
1661                         goto out;
1662                 }
1663                 rt->rt6i_prefsrc.addr = cfg->fc_prefsrc;
1664                 rt->rt6i_prefsrc.plen = 128;
1665         } else
1666                 rt->rt6i_prefsrc.plen = 0;
1667
1668         rt->rt6i_flags = cfg->fc_flags;
1669
1670 install_route:
1671         rt->dst.dev = dev;
1672         rt->rt6i_idev = idev;
1673         rt->rt6i_table = table;
1674
1675         cfg->fc_nlinfo.nl_net = dev_net(dev);
1676
1677         return __ip6_ins_rt(rt, &cfg->fc_nlinfo, cfg->fc_mx, cfg->fc_mx_len);
1678
1679 out:
1680         if (dev)
1681                 dev_put(dev);
1682         if (idev)
1683                 in6_dev_put(idev);
1684         if (rt)
1685                 dst_free(&rt->dst);
1686         return err;
1687 }
1688
1689 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
1690 {
1691         int err;
1692         struct fib6_table *table;
1693         struct net *net = dev_net(rt->dst.dev);
1694
1695         if (rt == net->ipv6.ip6_null_entry) {
1696                 err = -ENOENT;
1697                 goto out;
1698         }
1699
1700         table = rt->rt6i_table;
1701         write_lock_bh(&table->tb6_lock);
1702         err = fib6_del(rt, info);
1703         write_unlock_bh(&table->tb6_lock);
1704
1705 out:
1706         ip6_rt_put(rt);
1707         return err;
1708 }
1709
1710 int ip6_del_rt(struct rt6_info *rt)
1711 {
1712         struct nl_info info = {
1713                 .nl_net = dev_net(rt->dst.dev),
1714         };
1715         return __ip6_del_rt(rt, &info);
1716 }
1717
1718 static int ip6_route_del(struct fib6_config *cfg)
1719 {
1720         struct fib6_table *table;
1721         struct fib6_node *fn;
1722         struct rt6_info *rt;
1723         int err = -ESRCH;
1724
1725         table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
1726         if (!table)
1727                 return err;
1728
1729         read_lock_bh(&table->tb6_lock);
1730
1731         fn = fib6_locate(&table->tb6_root,
1732                          &cfg->fc_dst, cfg->fc_dst_len,
1733                          &cfg->fc_src, cfg->fc_src_len);
1734
1735         if (fn) {
1736                 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1737                         if (cfg->fc_ifindex &&
1738                             (!rt->dst.dev ||
1739                              rt->dst.dev->ifindex != cfg->fc_ifindex))
1740                                 continue;
1741                         if (cfg->fc_flags & RTF_GATEWAY &&
1742                             !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
1743                                 continue;
1744                         if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
1745                                 continue;
1746                         dst_hold(&rt->dst);
1747                         read_unlock_bh(&table->tb6_lock);
1748
1749                         return __ip6_del_rt(rt, &cfg->fc_nlinfo);
1750                 }
1751         }
1752         read_unlock_bh(&table->tb6_lock);
1753
1754         return err;
1755 }
1756
1757 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
1758 {
1759         struct net *net = dev_net(skb->dev);
1760         struct netevent_redirect netevent;
1761         struct rt6_info *rt, *nrt = NULL;
1762         struct ndisc_options ndopts;
1763         struct inet6_dev *in6_dev;
1764         struct neighbour *neigh;
1765         struct rd_msg *msg;
1766         int optlen, on_link;
1767         u8 *lladdr;
1768
1769         optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
1770         optlen -= sizeof(*msg);
1771
1772         if (optlen < 0) {
1773                 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
1774                 return;
1775         }
1776
1777         msg = (struct rd_msg *)icmp6_hdr(skb);
1778
1779         if (ipv6_addr_is_multicast(&msg->dest)) {
1780                 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
1781                 return;
1782         }
1783
1784         on_link = 0;
1785         if (ipv6_addr_equal(&msg->dest, &msg->target)) {
1786                 on_link = 1;
1787         } else if (ipv6_addr_type(&msg->target) !=
1788                    (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
1789                 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
1790                 return;
1791         }
1792
1793         in6_dev = __in6_dev_get(skb->dev);
1794         if (!in6_dev)
1795                 return;
1796         if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
1797                 return;
1798
1799         /* RFC2461 8.1:
1800          *      The IP source address of the Redirect MUST be the same as the current
1801          *      first-hop router for the specified ICMP Destination Address.
1802          */
1803
1804         if (!ndisc_parse_options(msg->opt, optlen, &ndopts)) {
1805                 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
1806                 return;
1807         }
1808
1809         lladdr = NULL;
1810         if (ndopts.nd_opts_tgt_lladdr) {
1811                 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
1812                                              skb->dev);
1813                 if (!lladdr) {
1814                         net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
1815                         return;
1816                 }
1817         }
1818
1819         rt = (struct rt6_info *) dst;
1820         if (rt == net->ipv6.ip6_null_entry) {
1821                 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
1822                 return;
1823         }
1824
1825         /* Redirect received -> path was valid.
1826          * Look, redirects are sent only in response to data packets,
1827          * so that this nexthop apparently is reachable. --ANK
1828          */
1829         dst_confirm(&rt->dst);
1830
1831         neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
1832         if (!neigh)
1833                 return;
1834
1835         /*
1836          *      We have finally decided to accept it.
1837          */
1838
1839         neigh_update(neigh, lladdr, NUD_STALE,
1840                      NEIGH_UPDATE_F_WEAK_OVERRIDE|
1841                      NEIGH_UPDATE_F_OVERRIDE|
1842                      (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1843                                      NEIGH_UPDATE_F_ISROUTER))
1844                      );
1845
1846         nrt = ip6_rt_copy(rt, &msg->dest);
1847         if (!nrt)
1848                 goto out;
1849
1850         nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
1851         if (on_link)
1852                 nrt->rt6i_flags &= ~RTF_GATEWAY;
1853
1854         nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
1855
1856         if (ip6_ins_rt(nrt))
1857                 goto out;
1858
1859         netevent.old = &rt->dst;
1860         netevent.new = &nrt->dst;
1861         netevent.daddr = &msg->dest;
1862         netevent.neigh = neigh;
1863         call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
1864
1865         if (rt->rt6i_flags & RTF_CACHE) {
1866                 rt = (struct rt6_info *) dst_clone(&rt->dst);
1867                 ip6_del_rt(rt);
1868         }
1869
1870 out:
1871         neigh_release(neigh);
1872 }
1873
1874 /*
1875  *      Misc support functions
1876  */
1877
1878 static struct rt6_info *ip6_rt_copy(struct rt6_info *ort,
1879                                     const struct in6_addr *dest)
1880 {
1881         struct net *net = dev_net(ort->dst.dev);
1882         struct rt6_info *rt = ip6_dst_alloc(net, ort->dst.dev, 0,
1883                                             ort->rt6i_table);
1884
1885         if (rt) {
1886                 rt->dst.input = ort->dst.input;
1887                 rt->dst.output = ort->dst.output;
1888                 rt->dst.flags |= DST_HOST;
1889
1890                 rt->rt6i_dst.addr = *dest;
1891                 rt->rt6i_dst.plen = 128;
1892                 dst_copy_metrics(&rt->dst, &ort->dst);
1893                 rt->dst.error = ort->dst.error;
1894                 rt->rt6i_idev = ort->rt6i_idev;
1895                 if (rt->rt6i_idev)
1896                         in6_dev_hold(rt->rt6i_idev);
1897                 rt->dst.lastuse = jiffies;
1898
1899                 if (ort->rt6i_flags & RTF_GATEWAY)
1900                         rt->rt6i_gateway = ort->rt6i_gateway;
1901                 else
1902                         rt->rt6i_gateway = *dest;
1903                 rt->rt6i_flags = ort->rt6i_flags;
1904                 rt6_set_from(rt, ort);
1905                 rt->rt6i_metric = 0;
1906
1907 #ifdef CONFIG_IPV6_SUBTREES
1908                 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1909 #endif
1910                 memcpy(&rt->rt6i_prefsrc, &ort->rt6i_prefsrc, sizeof(struct rt6key));
1911                 rt->rt6i_table = ort->rt6i_table;
1912         }
1913         return rt;
1914 }
1915
1916 #ifdef CONFIG_IPV6_ROUTE_INFO
1917 static struct rt6_info *rt6_get_route_info(struct net *net,
1918                                            const struct in6_addr *prefix, int prefixlen,
1919                                            const struct in6_addr *gwaddr, int ifindex)
1920 {
1921         struct fib6_node *fn;
1922         struct rt6_info *rt = NULL;
1923         struct fib6_table *table;
1924
1925         table = fib6_get_table(net, RT6_TABLE_INFO);
1926         if (!table)
1927                 return NULL;
1928
1929         read_lock_bh(&table->tb6_lock);
1930         fn = fib6_locate(&table->tb6_root, prefix, prefixlen, NULL, 0);
1931         if (!fn)
1932                 goto out;
1933
1934         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1935                 if (rt->dst.dev->ifindex != ifindex)
1936                         continue;
1937                 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
1938                         continue;
1939                 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
1940                         continue;
1941                 dst_hold(&rt->dst);
1942                 break;
1943         }
1944 out:
1945         read_unlock_bh(&table->tb6_lock);
1946         return rt;
1947 }
1948
1949 static struct rt6_info *rt6_add_route_info(struct net *net,
1950                                            const struct in6_addr *prefix, int prefixlen,
1951                                            const struct in6_addr *gwaddr, int ifindex,
1952                                            unsigned int pref)
1953 {
1954         struct fib6_config cfg = {
1955                 .fc_table       = RT6_TABLE_INFO,
1956                 .fc_metric      = IP6_RT_PRIO_USER,
1957                 .fc_ifindex     = ifindex,
1958                 .fc_dst_len     = prefixlen,
1959                 .fc_flags       = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
1960                                   RTF_UP | RTF_PREF(pref),
1961                 .fc_nlinfo.portid = 0,
1962                 .fc_nlinfo.nlh = NULL,
1963                 .fc_nlinfo.nl_net = net,
1964         };
1965
1966         cfg.fc_dst = *prefix;
1967         cfg.fc_gateway = *gwaddr;
1968
1969         /* We should treat it as a default route if prefix length is 0. */
1970         if (!prefixlen)
1971                 cfg.fc_flags |= RTF_DEFAULT;
1972
1973         ip6_route_add(&cfg);
1974
1975         return rt6_get_route_info(net, prefix, prefixlen, gwaddr, ifindex);
1976 }
1977 #endif
1978
1979 struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
1980 {
1981         struct rt6_info *rt;
1982         struct fib6_table *table;
1983
1984         table = fib6_get_table(dev_net(dev), RT6_TABLE_DFLT);
1985         if (!table)
1986                 return NULL;
1987
1988         read_lock_bh(&table->tb6_lock);
1989         for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
1990                 if (dev == rt->dst.dev &&
1991                     ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
1992                     ipv6_addr_equal(&rt->rt6i_gateway, addr))
1993                         break;
1994         }
1995         if (rt)
1996                 dst_hold(&rt->dst);
1997         read_unlock_bh(&table->tb6_lock);
1998         return rt;
1999 }
2000
2001 struct rt6_info *rt6_add_dflt_router(const struct in6_addr *gwaddr,
2002                                      struct net_device *dev,
2003                                      unsigned int pref)
2004 {
2005         struct fib6_config cfg = {
2006                 .fc_table       = RT6_TABLE_DFLT,
2007                 .fc_metric      = IP6_RT_PRIO_USER,
2008                 .fc_ifindex     = dev->ifindex,
2009                 .fc_flags       = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
2010                                   RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
2011                 .fc_nlinfo.portid = 0,
2012                 .fc_nlinfo.nlh = NULL,
2013                 .fc_nlinfo.nl_net = dev_net(dev),
2014         };
2015
2016         cfg.fc_gateway = *gwaddr;
2017
2018         ip6_route_add(&cfg);
2019
2020         return rt6_get_dflt_router(gwaddr, dev);
2021 }
2022
2023 void rt6_purge_dflt_routers(struct net *net)
2024 {
2025         struct rt6_info *rt;
2026         struct fib6_table *table;
2027
2028         /* NOTE: Keep consistent with rt6_get_dflt_router */
2029         table = fib6_get_table(net, RT6_TABLE_DFLT);
2030         if (!table)
2031                 return;
2032
2033 restart:
2034         read_lock_bh(&table->tb6_lock);
2035         for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2036                 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
2037                     (!rt->rt6i_idev || rt->rt6i_idev->cnf.accept_ra != 2)) {
2038                         dst_hold(&rt->dst);
2039                         read_unlock_bh(&table->tb6_lock);
2040                         ip6_del_rt(rt);
2041                         goto restart;
2042                 }
2043         }
2044         read_unlock_bh(&table->tb6_lock);
2045 }
2046
2047 static void rtmsg_to_fib6_config(struct net *net,
2048                                  struct in6_rtmsg *rtmsg,
2049                                  struct fib6_config *cfg)
2050 {
2051         memset(cfg, 0, sizeof(*cfg));
2052
2053         cfg->fc_table = RT6_TABLE_MAIN;
2054         cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
2055         cfg->fc_metric = rtmsg->rtmsg_metric;
2056         cfg->fc_expires = rtmsg->rtmsg_info;
2057         cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
2058         cfg->fc_src_len = rtmsg->rtmsg_src_len;
2059         cfg->fc_flags = rtmsg->rtmsg_flags;
2060
2061         cfg->fc_nlinfo.nl_net = net;
2062
2063         cfg->fc_dst = rtmsg->rtmsg_dst;
2064         cfg->fc_src = rtmsg->rtmsg_src;
2065         cfg->fc_gateway = rtmsg->rtmsg_gateway;
2066 }
2067
2068 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
2069 {
2070         struct fib6_config cfg;
2071         struct in6_rtmsg rtmsg;
2072         int err;
2073
2074         switch (cmd) {
2075         case SIOCADDRT:         /* Add a route */
2076         case SIOCDELRT:         /* Delete a route */
2077                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2078                         return -EPERM;
2079                 err = copy_from_user(&rtmsg, arg,
2080                                      sizeof(struct in6_rtmsg));
2081                 if (err)
2082                         return -EFAULT;
2083
2084                 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
2085
2086                 rtnl_lock();
2087                 switch (cmd) {
2088                 case SIOCADDRT:
2089                         err = ip6_route_add(&cfg);
2090                         break;
2091                 case SIOCDELRT:
2092                         err = ip6_route_del(&cfg);
2093                         break;
2094                 default:
2095                         err = -EINVAL;
2096                 }
2097                 rtnl_unlock();
2098
2099                 return err;
2100         }
2101
2102         return -EINVAL;
2103 }
2104
2105 /*
2106  *      Drop the packet on the floor
2107  */
2108
2109 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
2110 {
2111         int type;
2112         struct dst_entry *dst = skb_dst(skb);
2113         switch (ipstats_mib_noroutes) {
2114         case IPSTATS_MIB_INNOROUTES:
2115                 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
2116                 if (type == IPV6_ADDR_ANY) {
2117                         IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2118                                       IPSTATS_MIB_INADDRERRORS);
2119                         break;
2120                 }
2121                 /* FALLTHROUGH */
2122         case IPSTATS_MIB_OUTNOROUTES:
2123                 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2124                               ipstats_mib_noroutes);
2125                 break;
2126         }
2127         icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
2128         kfree_skb(skb);
2129         return 0;
2130 }
2131
2132 static int ip6_pkt_discard(struct sk_buff *skb)
2133 {
2134         return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
2135 }
2136
2137 static int ip6_pkt_discard_out(struct sock *sk, struct sk_buff *skb)
2138 {
2139         skb->dev = skb_dst(skb)->dev;
2140         return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
2141 }
2142
2143 static int ip6_pkt_prohibit(struct sk_buff *skb)
2144 {
2145         return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
2146 }
2147
2148 static int ip6_pkt_prohibit_out(struct sock *sk, struct sk_buff *skb)
2149 {
2150         skb->dev = skb_dst(skb)->dev;
2151         return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
2152 }
2153
2154 /*
2155  *      Allocate a dst for local (unicast / anycast) address.
2156  */
2157
2158 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
2159                                     const struct in6_addr *addr,
2160                                     bool anycast)
2161 {
2162         struct net *net = dev_net(idev->dev);
2163         struct rt6_info *rt = ip6_dst_alloc(net, net->loopback_dev,
2164                                             DST_NOCOUNT, NULL);
2165         if (!rt)
2166                 return ERR_PTR(-ENOMEM);
2167
2168         in6_dev_hold(idev);
2169
2170         rt->dst.flags |= DST_HOST;
2171         rt->dst.input = ip6_input;
2172         rt->dst.output = ip6_output;
2173         rt->rt6i_idev = idev;
2174
2175         rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
2176         if (anycast)
2177                 rt->rt6i_flags |= RTF_ANYCAST;
2178         else
2179                 rt->rt6i_flags |= RTF_LOCAL;
2180
2181         rt->rt6i_gateway  = *addr;
2182         rt->rt6i_dst.addr = *addr;
2183         rt->rt6i_dst.plen = 128;
2184         rt->rt6i_table = fib6_get_table(net, RT6_TABLE_LOCAL);
2185
2186         atomic_set(&rt->dst.__refcnt, 1);
2187
2188         return rt;
2189 }
2190
2191 int ip6_route_get_saddr(struct net *net,
2192                         struct rt6_info *rt,
2193                         const struct in6_addr *daddr,
2194                         unsigned int prefs,
2195                         struct in6_addr *saddr)
2196 {
2197         struct inet6_dev *idev = ip6_dst_idev((struct dst_entry *)rt);
2198         int err = 0;
2199         if (rt->rt6i_prefsrc.plen)
2200                 *saddr = rt->rt6i_prefsrc.addr;
2201         else
2202                 err = ipv6_dev_get_saddr(net, idev ? idev->dev : NULL,
2203                                          daddr, prefs, saddr);
2204         return err;
2205 }
2206
2207 /* remove deleted ip from prefsrc entries */
2208 struct arg_dev_net_ip {
2209         struct net_device *dev;
2210         struct net *net;
2211         struct in6_addr *addr;
2212 };
2213
2214 static int fib6_remove_prefsrc(struct rt6_info *rt, void *arg)
2215 {
2216         struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
2217         struct net *net = ((struct arg_dev_net_ip *)arg)->net;
2218         struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
2219
2220         if (((void *)rt->dst.dev == dev || !dev) &&
2221             rt != net->ipv6.ip6_null_entry &&
2222             ipv6_addr_equal(addr, &rt->rt6i_prefsrc.addr)) {
2223                 /* remove prefsrc entry */
2224                 rt->rt6i_prefsrc.plen = 0;
2225         }
2226         return 0;
2227 }
2228
2229 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
2230 {
2231         struct net *net = dev_net(ifp->idev->dev);
2232         struct arg_dev_net_ip adni = {
2233                 .dev = ifp->idev->dev,
2234                 .net = net,
2235                 .addr = &ifp->addr,
2236         };
2237         fib6_clean_all(net, fib6_remove_prefsrc, &adni);
2238 }
2239
2240 #define RTF_RA_ROUTER           (RTF_ADDRCONF | RTF_DEFAULT | RTF_GATEWAY)
2241 #define RTF_CACHE_GATEWAY       (RTF_GATEWAY | RTF_CACHE)
2242
2243 /* Remove routers and update dst entries when gateway turn into host. */
2244 static int fib6_clean_tohost(struct rt6_info *rt, void *arg)
2245 {
2246         struct in6_addr *gateway = (struct in6_addr *)arg;
2247
2248         if ((((rt->rt6i_flags & RTF_RA_ROUTER) == RTF_RA_ROUTER) ||
2249              ((rt->rt6i_flags & RTF_CACHE_GATEWAY) == RTF_CACHE_GATEWAY)) &&
2250              ipv6_addr_equal(gateway, &rt->rt6i_gateway)) {
2251                 return -1;
2252         }
2253         return 0;
2254 }
2255
2256 void rt6_clean_tohost(struct net *net, struct in6_addr *gateway)
2257 {
2258         fib6_clean_all(net, fib6_clean_tohost, gateway);
2259 }
2260
2261 struct arg_dev_net {
2262         struct net_device *dev;
2263         struct net *net;
2264 };
2265
2266 static int fib6_ifdown(struct rt6_info *rt, void *arg)
2267 {
2268         const struct arg_dev_net *adn = arg;
2269         const struct net_device *dev = adn->dev;
2270
2271         if ((rt->dst.dev == dev || !dev) &&
2272             rt != adn->net->ipv6.ip6_null_entry)
2273                 return -1;
2274
2275         return 0;
2276 }
2277
2278 void rt6_ifdown(struct net *net, struct net_device *dev)
2279 {
2280         struct arg_dev_net adn = {
2281                 .dev = dev,
2282                 .net = net,
2283         };
2284
2285         fib6_clean_all(net, fib6_ifdown, &adn);
2286         icmp6_clean_all(fib6_ifdown, &adn);
2287 }
2288
2289 struct rt6_mtu_change_arg {
2290         struct net_device *dev;
2291         unsigned int mtu;
2292 };
2293
2294 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2295 {
2296         struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2297         struct inet6_dev *idev;
2298
2299         /* In IPv6 pmtu discovery is not optional,
2300            so that RTAX_MTU lock cannot disable it.
2301            We still use this lock to block changes
2302            caused by addrconf/ndisc.
2303         */
2304
2305         idev = __in6_dev_get(arg->dev);
2306         if (!idev)
2307                 return 0;
2308
2309         /* For administrative MTU increase, there is no way to discover
2310            IPv6 PMTU increase, so PMTU increase should be updated here.
2311            Since RFC 1981 doesn't include administrative MTU increase
2312            update PMTU increase is a MUST. (i.e. jumbo frame)
2313          */
2314         /*
2315            If new MTU is less than route PMTU, this new MTU will be the
2316            lowest MTU in the path, update the route PMTU to reflect PMTU
2317            decreases; if new MTU is greater than route PMTU, and the
2318            old MTU is the lowest MTU in the path, update the route PMTU
2319            to reflect the increase. In this case if the other nodes' MTU
2320            also have the lowest MTU, TOO BIG MESSAGE will be lead to
2321            PMTU discouvery.
2322          */
2323         if (rt->dst.dev == arg->dev &&
2324             !dst_metric_locked(&rt->dst, RTAX_MTU) &&
2325             (dst_mtu(&rt->dst) >= arg->mtu ||
2326              (dst_mtu(&rt->dst) < arg->mtu &&
2327               dst_mtu(&rt->dst) == idev->cnf.mtu6))) {
2328                 dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
2329         }
2330         return 0;
2331 }
2332
2333 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
2334 {
2335         struct rt6_mtu_change_arg arg = {
2336                 .dev = dev,
2337                 .mtu = mtu,
2338         };
2339
2340         fib6_clean_all(dev_net(dev), rt6_mtu_change_route, &arg);
2341 }
2342
2343 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2344         [RTA_GATEWAY]           = { .len = sizeof(struct in6_addr) },
2345         [RTA_OIF]               = { .type = NLA_U32 },
2346         [RTA_IIF]               = { .type = NLA_U32 },
2347         [RTA_PRIORITY]          = { .type = NLA_U32 },
2348         [RTA_METRICS]           = { .type = NLA_NESTED },
2349         [RTA_MULTIPATH]         = { .len = sizeof(struct rtnexthop) },
2350 };
2351
2352 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2353                               struct fib6_config *cfg)
2354 {
2355         struct rtmsg *rtm;
2356         struct nlattr *tb[RTA_MAX+1];
2357         int err;
2358
2359         err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2360         if (err < 0)
2361                 goto errout;
2362
2363         err = -EINVAL;
2364         rtm = nlmsg_data(nlh);
2365         memset(cfg, 0, sizeof(*cfg));
2366
2367         cfg->fc_table = rtm->rtm_table;
2368         cfg->fc_dst_len = rtm->rtm_dst_len;
2369         cfg->fc_src_len = rtm->rtm_src_len;
2370         cfg->fc_flags = RTF_UP;
2371         cfg->fc_protocol = rtm->rtm_protocol;
2372         cfg->fc_type = rtm->rtm_type;
2373
2374         if (rtm->rtm_type == RTN_UNREACHABLE ||
2375             rtm->rtm_type == RTN_BLACKHOLE ||
2376             rtm->rtm_type == RTN_PROHIBIT ||
2377             rtm->rtm_type == RTN_THROW)
2378                 cfg->fc_flags |= RTF_REJECT;
2379
2380         if (rtm->rtm_type == RTN_LOCAL)
2381                 cfg->fc_flags |= RTF_LOCAL;
2382
2383         cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid;
2384         cfg->fc_nlinfo.nlh = nlh;
2385         cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2386
2387         if (tb[RTA_GATEWAY]) {
2388                 nla_memcpy(&cfg->fc_gateway, tb[RTA_GATEWAY], 16);
2389                 cfg->fc_flags |= RTF_GATEWAY;
2390         }
2391
2392         if (tb[RTA_DST]) {
2393                 int plen = (rtm->rtm_dst_len + 7) >> 3;
2394
2395                 if (nla_len(tb[RTA_DST]) < plen)
2396                         goto errout;
2397
2398                 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2399         }
2400
2401         if (tb[RTA_SRC]) {
2402                 int plen = (rtm->rtm_src_len + 7) >> 3;
2403
2404                 if (nla_len(tb[RTA_SRC]) < plen)
2405                         goto errout;
2406
2407                 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2408         }
2409
2410         if (tb[RTA_PREFSRC])
2411                 nla_memcpy(&cfg->fc_prefsrc, tb[RTA_PREFSRC], 16);
2412
2413         if (tb[RTA_OIF])
2414                 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2415
2416         if (tb[RTA_PRIORITY])
2417                 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2418
2419         if (tb[RTA_METRICS]) {
2420                 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2421                 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2422         }
2423
2424         if (tb[RTA_TABLE])
2425                 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2426
2427         if (tb[RTA_MULTIPATH]) {
2428                 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
2429                 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
2430         }
2431
2432         err = 0;
2433 errout:
2434         return err;
2435 }
2436
2437 static int ip6_route_multipath(struct fib6_config *cfg, int add)
2438 {
2439         struct fib6_config r_cfg;
2440         struct rtnexthop *rtnh;
2441         int remaining;
2442         int attrlen;
2443         int err = 0, last_err = 0;
2444
2445 beginning:
2446         rtnh = (struct rtnexthop *)cfg->fc_mp;
2447         remaining = cfg->fc_mp_len;
2448
2449         /* Parse a Multipath Entry */
2450         while (rtnh_ok(rtnh, remaining)) {
2451                 memcpy(&r_cfg, cfg, sizeof(*cfg));
2452                 if (rtnh->rtnh_ifindex)
2453                         r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
2454
2455                 attrlen = rtnh_attrlen(rtnh);
2456                 if (attrlen > 0) {
2457                         struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
2458
2459                         nla = nla_find(attrs, attrlen, RTA_GATEWAY);
2460                         if (nla) {
2461                                 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
2462                                 r_cfg.fc_flags |= RTF_GATEWAY;
2463                         }
2464                 }
2465                 err = add ? ip6_route_add(&r_cfg) : ip6_route_del(&r_cfg);
2466                 if (err) {
2467                         last_err = err;
2468                         /* If we are trying to remove a route, do not stop the
2469                          * loop when ip6_route_del() fails (because next hop is
2470                          * already gone), we should try to remove all next hops.
2471                          */
2472                         if (add) {
2473                                 /* If add fails, we should try to delete all
2474                                  * next hops that have been already added.
2475                                  */
2476                                 add = 0;
2477                                 goto beginning;
2478                         }
2479                 }
2480                 /* Because each route is added like a single route we remove
2481                  * this flag after the first nexthop (if there is a collision,
2482                  * we have already fail to add the first nexthop:
2483                  * fib6_add_rt2node() has reject it).
2484                  */
2485                 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~NLM_F_EXCL;
2486                 rtnh = rtnh_next(rtnh, &remaining);
2487         }
2488
2489         return last_err;
2490 }
2491
2492 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh)
2493 {
2494         struct fib6_config cfg;
2495         int err;
2496
2497         err = rtm_to_fib6_config(skb, nlh, &cfg);
2498         if (err < 0)
2499                 return err;
2500
2501         if (cfg.fc_mp)
2502                 return ip6_route_multipath(&cfg, 0);
2503         else
2504                 return ip6_route_del(&cfg);
2505 }
2506
2507 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh)
2508 {
2509         struct fib6_config cfg;
2510         int err;
2511
2512         err = rtm_to_fib6_config(skb, nlh, &cfg);
2513         if (err < 0)
2514                 return err;
2515
2516         if (cfg.fc_mp)
2517                 return ip6_route_multipath(&cfg, 1);
2518         else
2519                 return ip6_route_add(&cfg);
2520 }
2521
2522 static inline size_t rt6_nlmsg_size(void)
2523 {
2524         return NLMSG_ALIGN(sizeof(struct rtmsg))
2525                + nla_total_size(16) /* RTA_SRC */
2526                + nla_total_size(16) /* RTA_DST */
2527                + nla_total_size(16) /* RTA_GATEWAY */
2528                + nla_total_size(16) /* RTA_PREFSRC */
2529                + nla_total_size(4) /* RTA_TABLE */
2530                + nla_total_size(4) /* RTA_IIF */
2531                + nla_total_size(4) /* RTA_OIF */
2532                + nla_total_size(4) /* RTA_PRIORITY */
2533                + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
2534                + nla_total_size(sizeof(struct rta_cacheinfo));
2535 }
2536
2537 static int rt6_fill_node(struct net *net,
2538                          struct sk_buff *skb, struct rt6_info *rt,
2539                          struct in6_addr *dst, struct in6_addr *src,
2540                          int iif, int type, u32 portid, u32 seq,
2541                          int prefix, int nowait, unsigned int flags)
2542 {
2543         struct rtmsg *rtm;
2544         struct nlmsghdr *nlh;
2545         long expires;
2546         u32 table;
2547
2548         if (prefix) {   /* user wants prefix routes only */
2549                 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
2550                         /* success since this is not a prefix route */
2551                         return 1;
2552                 }
2553         }
2554
2555         nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
2556         if (!nlh)
2557                 return -EMSGSIZE;
2558
2559         rtm = nlmsg_data(nlh);
2560         rtm->rtm_family = AF_INET6;
2561         rtm->rtm_dst_len = rt->rt6i_dst.plen;
2562         rtm->rtm_src_len = rt->rt6i_src.plen;
2563         rtm->rtm_tos = 0;
2564         if (rt->rt6i_table)
2565                 table = rt->rt6i_table->tb6_id;
2566         else
2567                 table = RT6_TABLE_UNSPEC;
2568         rtm->rtm_table = table;
2569         if (nla_put_u32(skb, RTA_TABLE, table))
2570                 goto nla_put_failure;
2571         if (rt->rt6i_flags & RTF_REJECT) {
2572                 switch (rt->dst.error) {
2573                 case -EINVAL:
2574                         rtm->rtm_type = RTN_BLACKHOLE;
2575                         break;
2576                 case -EACCES:
2577                         rtm->rtm_type = RTN_PROHIBIT;
2578                         break;
2579                 case -EAGAIN:
2580                         rtm->rtm_type = RTN_THROW;
2581                         break;
2582                 default:
2583                         rtm->rtm_type = RTN_UNREACHABLE;
2584                         break;
2585                 }
2586         }
2587         else if (rt->rt6i_flags & RTF_LOCAL)
2588                 rtm->rtm_type = RTN_LOCAL;
2589         else if (rt->dst.dev && (rt->dst.dev->flags & IFF_LOOPBACK))
2590                 rtm->rtm_type = RTN_LOCAL;
2591         else
2592                 rtm->rtm_type = RTN_UNICAST;
2593         rtm->rtm_flags = 0;
2594         rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2595         rtm->rtm_protocol = rt->rt6i_protocol;
2596         if (rt->rt6i_flags & RTF_DYNAMIC)
2597                 rtm->rtm_protocol = RTPROT_REDIRECT;
2598         else if (rt->rt6i_flags & RTF_ADDRCONF) {
2599                 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ROUTEINFO))
2600                         rtm->rtm_protocol = RTPROT_RA;
2601                 else
2602                         rtm->rtm_protocol = RTPROT_KERNEL;
2603         }
2604
2605         if (rt->rt6i_flags & RTF_CACHE)
2606                 rtm->rtm_flags |= RTM_F_CLONED;
2607
2608         if (dst) {
2609                 if (nla_put(skb, RTA_DST, 16, dst))
2610                         goto nla_put_failure;
2611                 rtm->rtm_dst_len = 128;
2612         } else if (rtm->rtm_dst_len)
2613                 if (nla_put(skb, RTA_DST, 16, &rt->rt6i_dst.addr))
2614                         goto nla_put_failure;
2615 #ifdef CONFIG_IPV6_SUBTREES
2616         if (src) {
2617                 if (nla_put(skb, RTA_SRC, 16, src))
2618                         goto nla_put_failure;
2619                 rtm->rtm_src_len = 128;
2620         } else if (rtm->rtm_src_len &&
2621                    nla_put(skb, RTA_SRC, 16, &rt->rt6i_src.addr))
2622                 goto nla_put_failure;
2623 #endif
2624         if (iif) {
2625 #ifdef CONFIG_IPV6_MROUTE
2626                 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
2627                         int err = ip6mr_get_route(net, skb, rtm, nowait);
2628                         if (err <= 0) {
2629                                 if (!nowait) {
2630                                         if (err == 0)
2631                                                 return 0;
2632                                         goto nla_put_failure;
2633                                 } else {
2634                                         if (err == -EMSGSIZE)
2635                                                 goto nla_put_failure;
2636                                 }
2637                         }
2638                 } else
2639 #endif
2640                         if (nla_put_u32(skb, RTA_IIF, iif))
2641                                 goto nla_put_failure;
2642         } else if (dst) {
2643                 struct in6_addr saddr_buf;
2644                 if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0 &&
2645                     nla_put(skb, RTA_PREFSRC, 16, &saddr_buf))
2646                         goto nla_put_failure;
2647         }
2648
2649         if (rt->rt6i_prefsrc.plen) {
2650                 struct in6_addr saddr_buf;
2651                 saddr_buf = rt->rt6i_prefsrc.addr;
2652                 if (nla_put(skb, RTA_PREFSRC, 16, &saddr_buf))
2653                         goto nla_put_failure;
2654         }
2655
2656         if (rtnetlink_put_metrics(skb, dst_metrics_ptr(&rt->dst)) < 0)
2657                 goto nla_put_failure;
2658
2659         if (rt->rt6i_flags & RTF_GATEWAY) {
2660                 if (nla_put(skb, RTA_GATEWAY, 16, &rt->rt6i_gateway) < 0)
2661                         goto nla_put_failure;
2662         }
2663
2664         if (rt->dst.dev &&
2665             nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex))
2666                 goto nla_put_failure;
2667         if (nla_put_u32(skb, RTA_PRIORITY, rt->rt6i_metric))
2668                 goto nla_put_failure;
2669
2670         expires = (rt->rt6i_flags & RTF_EXPIRES) ? rt->dst.expires - jiffies : 0;
2671
2672         if (rtnl_put_cacheinfo(skb, &rt->dst, 0, expires, rt->dst.error) < 0)
2673                 goto nla_put_failure;
2674
2675         return nlmsg_end(skb, nlh);
2676
2677 nla_put_failure:
2678         nlmsg_cancel(skb, nlh);
2679         return -EMSGSIZE;
2680 }
2681
2682 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
2683 {
2684         struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
2685         int prefix;
2686
2687         if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
2688                 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
2689                 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
2690         } else
2691                 prefix = 0;
2692
2693         return rt6_fill_node(arg->net,
2694                      arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
2695                      NETLINK_CB(arg->cb->skb).portid, arg->cb->nlh->nlmsg_seq,
2696                      prefix, 0, NLM_F_MULTI);
2697 }
2698
2699 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh)
2700 {
2701         struct net *net = sock_net(in_skb->sk);
2702         struct nlattr *tb[RTA_MAX+1];
2703         struct rt6_info *rt;
2704         struct sk_buff *skb;
2705         struct rtmsg *rtm;
2706         struct flowi6 fl6;
2707         int err, iif = 0, oif = 0;
2708
2709         err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2710         if (err < 0)
2711                 goto errout;
2712
2713         err = -EINVAL;
2714         memset(&fl6, 0, sizeof(fl6));
2715
2716         if (tb[RTA_SRC]) {
2717                 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
2718                         goto errout;
2719
2720                 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
2721         }
2722
2723         if (tb[RTA_DST]) {
2724                 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
2725                         goto errout;
2726
2727                 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
2728         }
2729
2730         if (tb[RTA_IIF])
2731                 iif = nla_get_u32(tb[RTA_IIF]);
2732
2733         if (tb[RTA_OIF])
2734                 oif = nla_get_u32(tb[RTA_OIF]);
2735
2736         if (tb[RTA_MARK])
2737                 fl6.flowi6_mark = nla_get_u32(tb[RTA_MARK]);
2738
2739         if (iif) {
2740                 struct net_device *dev;
2741                 int flags = 0;
2742
2743                 dev = __dev_get_by_index(net, iif);
2744                 if (!dev) {
2745                         err = -ENODEV;
2746                         goto errout;
2747                 }
2748
2749                 fl6.flowi6_iif = iif;
2750
2751                 if (!ipv6_addr_any(&fl6.saddr))
2752                         flags |= RT6_LOOKUP_F_HAS_SADDR;
2753
2754                 rt = (struct rt6_info *)ip6_route_input_lookup(net, dev, &fl6,
2755                                                                flags);
2756         } else {
2757                 fl6.flowi6_oif = oif;
2758
2759                 rt = (struct rt6_info *)ip6_route_output(net, NULL, &fl6);
2760         }
2761
2762         skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
2763         if (!skb) {
2764                 ip6_rt_put(rt);
2765                 err = -ENOBUFS;
2766                 goto errout;
2767         }
2768
2769         /* Reserve room for dummy headers, this skb can pass
2770            through good chunk of routing engine.
2771          */
2772         skb_reset_mac_header(skb);
2773         skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
2774
2775         skb_dst_set(skb, &rt->dst);
2776
2777         err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
2778                             RTM_NEWROUTE, NETLINK_CB(in_skb).portid,
2779                             nlh->nlmsg_seq, 0, 0, 0);
2780         if (err < 0) {
2781                 kfree_skb(skb);
2782                 goto errout;
2783         }
2784
2785         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
2786 errout:
2787         return err;
2788 }
2789
2790 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info)
2791 {
2792         struct sk_buff *skb;
2793         struct net *net = info->nl_net;
2794         u32 seq;
2795         int err;
2796
2797         err = -ENOBUFS;
2798         seq = info->nlh ? info->nlh->nlmsg_seq : 0;
2799
2800         skb = nlmsg_new(rt6_nlmsg_size(), gfp_any());
2801         if (!skb)
2802                 goto errout;
2803
2804         err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
2805                                 event, info->portid, seq, 0, 0, 0);
2806         if (err < 0) {
2807                 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
2808                 WARN_ON(err == -EMSGSIZE);
2809                 kfree_skb(skb);
2810                 goto errout;
2811         }
2812         rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
2813                     info->nlh, gfp_any());
2814         return;
2815 errout:
2816         if (err < 0)
2817                 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
2818 }
2819
2820 static int ip6_route_dev_notify(struct notifier_block *this,
2821                                 unsigned long event, void *ptr)
2822 {
2823         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2824         struct net *net = dev_net(dev);
2825
2826         if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
2827                 net->ipv6.ip6_null_entry->dst.dev = dev;
2828                 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
2829 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2830                 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
2831                 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
2832                 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
2833                 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
2834 #endif
2835         }
2836
2837         return NOTIFY_OK;
2838 }
2839
2840 /*
2841  *      /proc
2842  */
2843
2844 #ifdef CONFIG_PROC_FS
2845
2846 static const struct file_operations ipv6_route_proc_fops = {
2847         .owner          = THIS_MODULE,
2848         .open           = ipv6_route_open,
2849         .read           = seq_read,
2850         .llseek         = seq_lseek,
2851         .release        = seq_release_net,
2852 };
2853
2854 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
2855 {
2856         struct net *net = (struct net *)seq->private;
2857         seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
2858                    net->ipv6.rt6_stats->fib_nodes,
2859                    net->ipv6.rt6_stats->fib_route_nodes,
2860                    net->ipv6.rt6_stats->fib_rt_alloc,
2861                    net->ipv6.rt6_stats->fib_rt_entries,
2862                    net->ipv6.rt6_stats->fib_rt_cache,
2863                    dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
2864                    net->ipv6.rt6_stats->fib_discarded_routes);
2865
2866         return 0;
2867 }
2868
2869 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
2870 {
2871         return single_open_net(inode, file, rt6_stats_seq_show);
2872 }
2873
2874 static const struct file_operations rt6_stats_seq_fops = {
2875         .owner   = THIS_MODULE,
2876         .open    = rt6_stats_seq_open,
2877         .read    = seq_read,
2878         .llseek  = seq_lseek,
2879         .release = single_release_net,
2880 };
2881 #endif  /* CONFIG_PROC_FS */
2882
2883 #ifdef CONFIG_SYSCTL
2884
2885 static
2886 int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
2887                               void __user *buffer, size_t *lenp, loff_t *ppos)
2888 {
2889         struct net *net;
2890         int delay;
2891         if (!write)
2892                 return -EINVAL;
2893
2894         net = (struct net *)ctl->extra1;
2895         delay = net->ipv6.sysctl.flush_delay;
2896         proc_dointvec(ctl, write, buffer, lenp, ppos);
2897         fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
2898         return 0;
2899 }
2900
2901 struct ctl_table ipv6_route_table_template[] = {
2902         {
2903                 .procname       =       "flush",
2904                 .data           =       &init_net.ipv6.sysctl.flush_delay,
2905                 .maxlen         =       sizeof(int),
2906                 .mode           =       0200,
2907                 .proc_handler   =       ipv6_sysctl_rtcache_flush
2908         },
2909         {
2910                 .procname       =       "gc_thresh",
2911                 .data           =       &ip6_dst_ops_template.gc_thresh,
2912                 .maxlen         =       sizeof(int),
2913                 .mode           =       0644,
2914                 .proc_handler   =       proc_dointvec,
2915         },
2916         {
2917                 .procname       =       "max_size",
2918                 .data           =       &init_net.ipv6.sysctl.ip6_rt_max_size,
2919                 .maxlen         =       sizeof(int),
2920                 .mode           =       0644,
2921                 .proc_handler   =       proc_dointvec,
2922         },
2923         {
2924                 .procname       =       "gc_min_interval",
2925                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2926                 .maxlen         =       sizeof(int),
2927                 .mode           =       0644,
2928                 .proc_handler   =       proc_dointvec_jiffies,
2929         },
2930         {
2931                 .procname       =       "gc_timeout",
2932                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
2933                 .maxlen         =       sizeof(int),
2934                 .mode           =       0644,
2935                 .proc_handler   =       proc_dointvec_jiffies,
2936         },
2937         {
2938                 .procname       =       "gc_interval",
2939                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_interval,
2940                 .maxlen         =       sizeof(int),
2941                 .mode           =       0644,
2942                 .proc_handler   =       proc_dointvec_jiffies,
2943         },
2944         {
2945                 .procname       =       "gc_elasticity",
2946                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
2947                 .maxlen         =       sizeof(int),
2948                 .mode           =       0644,
2949                 .proc_handler   =       proc_dointvec,
2950         },
2951         {
2952                 .procname       =       "mtu_expires",
2953                 .data           =       &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
2954                 .maxlen         =       sizeof(int),
2955                 .mode           =       0644,
2956                 .proc_handler   =       proc_dointvec_jiffies,
2957         },
2958         {
2959                 .procname       =       "min_adv_mss",
2960                 .data           =       &init_net.ipv6.sysctl.ip6_rt_min_advmss,
2961                 .maxlen         =       sizeof(int),
2962                 .mode           =       0644,
2963                 .proc_handler   =       proc_dointvec,
2964         },
2965         {
2966                 .procname       =       "gc_min_interval_ms",
2967                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2968                 .maxlen         =       sizeof(int),
2969                 .mode           =       0644,
2970                 .proc_handler   =       proc_dointvec_ms_jiffies,
2971         },
2972         { }
2973 };
2974
2975 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
2976 {
2977         struct ctl_table *table;
2978
2979         table = kmemdup(ipv6_route_table_template,
2980                         sizeof(ipv6_route_table_template),
2981                         GFP_KERNEL);
2982
2983         if (table) {
2984                 table[0].data = &net->ipv6.sysctl.flush_delay;
2985                 table[0].extra1 = net;
2986                 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
2987                 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
2988                 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2989                 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
2990                 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
2991                 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
2992                 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
2993                 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
2994                 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2995
2996                 /* Don't export sysctls to unprivileged users */
2997                 if (net->user_ns != &init_user_ns)
2998                         table[0].procname = NULL;
2999         }
3000
3001         return table;
3002 }
3003 #endif
3004
3005 static int __net_init ip6_route_net_init(struct net *net)
3006 {
3007         int ret = -ENOMEM;
3008
3009         memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
3010                sizeof(net->ipv6.ip6_dst_ops));
3011
3012         if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
3013                 goto out_ip6_dst_ops;
3014
3015         net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
3016                                            sizeof(*net->ipv6.ip6_null_entry),
3017                                            GFP_KERNEL);
3018         if (!net->ipv6.ip6_null_entry)
3019                 goto out_ip6_dst_entries;
3020         net->ipv6.ip6_null_entry->dst.path =
3021                 (struct dst_entry *)net->ipv6.ip6_null_entry;
3022         net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3023         dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
3024                          ip6_template_metrics, true);
3025
3026 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3027         net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
3028                                                sizeof(*net->ipv6.ip6_prohibit_entry),
3029                                                GFP_KERNEL);
3030         if (!net->ipv6.ip6_prohibit_entry)
3031                 goto out_ip6_null_entry;
3032         net->ipv6.ip6_prohibit_entry->dst.path =
3033                 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
3034         net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3035         dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
3036                          ip6_template_metrics, true);
3037
3038         net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
3039                                                sizeof(*net->ipv6.ip6_blk_hole_entry),
3040                                                GFP_KERNEL);
3041         if (!net->ipv6.ip6_blk_hole_entry)
3042                 goto out_ip6_prohibit_entry;
3043         net->ipv6.ip6_blk_hole_entry->dst.path =
3044                 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
3045         net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3046         dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
3047                          ip6_template_metrics, true);
3048 #endif
3049
3050         net->ipv6.sysctl.flush_delay = 0;
3051         net->ipv6.sysctl.ip6_rt_max_size = 4096;
3052         net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
3053         net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
3054         net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
3055         net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
3056         net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
3057         net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
3058
3059         net->ipv6.ip6_rt_gc_expire = 30*HZ;
3060
3061         ret = 0;
3062 out:
3063         return ret;
3064
3065 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3066 out_ip6_prohibit_entry:
3067         kfree(net->ipv6.ip6_prohibit_entry);
3068 out_ip6_null_entry:
3069         kfree(net->ipv6.ip6_null_entry);
3070 #endif
3071 out_ip6_dst_entries:
3072         dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3073 out_ip6_dst_ops:
3074         goto out;
3075 }
3076
3077 static void __net_exit ip6_route_net_exit(struct net *net)
3078 {
3079         kfree(net->ipv6.ip6_null_entry);
3080 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3081         kfree(net->ipv6.ip6_prohibit_entry);
3082         kfree(net->ipv6.ip6_blk_hole_entry);
3083 #endif
3084         dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3085 }
3086
3087 static int __net_init ip6_route_net_init_late(struct net *net)
3088 {
3089 #ifdef CONFIG_PROC_FS
3090         proc_create("ipv6_route", 0, net->proc_net, &ipv6_route_proc_fops);
3091         proc_create("rt6_stats", S_IRUGO, net->proc_net, &rt6_stats_seq_fops);
3092 #endif
3093         return 0;
3094 }
3095
3096 static void __net_exit ip6_route_net_exit_late(struct net *net)
3097 {
3098 #ifdef CONFIG_PROC_FS
3099         remove_proc_entry("ipv6_route", net->proc_net);
3100         remove_proc_entry("rt6_stats", net->proc_net);
3101 #endif
3102 }
3103
3104 static struct pernet_operations ip6_route_net_ops = {
3105         .init = ip6_route_net_init,
3106         .exit = ip6_route_net_exit,
3107 };
3108
3109 static int __net_init ipv6_inetpeer_init(struct net *net)
3110 {
3111         struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
3112
3113         if (!bp)
3114                 return -ENOMEM;
3115         inet_peer_base_init(bp);
3116         net->ipv6.peers = bp;
3117         return 0;
3118 }
3119
3120 static void __net_exit ipv6_inetpeer_exit(struct net *net)
3121 {
3122         struct inet_peer_base *bp = net->ipv6.peers;
3123
3124         net->ipv6.peers = NULL;
3125         inetpeer_invalidate_tree(bp);
3126         kfree(bp);
3127 }
3128
3129 static struct pernet_operations ipv6_inetpeer_ops = {
3130         .init   =       ipv6_inetpeer_init,
3131         .exit   =       ipv6_inetpeer_exit,
3132 };
3133
3134 static struct pernet_operations ip6_route_net_late_ops = {
3135         .init = ip6_route_net_init_late,
3136         .exit = ip6_route_net_exit_late,
3137 };
3138
3139 static struct notifier_block ip6_route_dev_notifier = {
3140         .notifier_call = ip6_route_dev_notify,
3141         .priority = 0,
3142 };
3143
3144 int __init ip6_route_init(void)
3145 {
3146         int ret;
3147
3148         ret = -ENOMEM;
3149         ip6_dst_ops_template.kmem_cachep =
3150                 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
3151                                   SLAB_HWCACHE_ALIGN, NULL);
3152         if (!ip6_dst_ops_template.kmem_cachep)
3153                 goto out;
3154
3155         ret = dst_entries_init(&ip6_dst_blackhole_ops);
3156         if (ret)
3157                 goto out_kmem_cache;
3158
3159         ret = register_pernet_subsys(&ipv6_inetpeer_ops);
3160         if (ret)
3161                 goto out_dst_entries;
3162
3163         ret = register_pernet_subsys(&ip6_route_net_ops);
3164         if (ret)
3165                 goto out_register_inetpeer;
3166
3167         ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
3168
3169         /* Registering of the loopback is done before this portion of code,
3170          * the loopback reference in rt6_info will not be taken, do it
3171          * manually for init_net */
3172         init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
3173         init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3174   #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3175         init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
3176         init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3177         init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
3178         init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3179   #endif
3180         ret = fib6_init();
3181         if (ret)
3182                 goto out_register_subsys;
3183
3184         ret = xfrm6_init();
3185         if (ret)
3186                 goto out_fib6_init;
3187
3188         ret = fib6_rules_init();
3189         if (ret)
3190                 goto xfrm6_init;
3191
3192         ret = register_pernet_subsys(&ip6_route_net_late_ops);
3193         if (ret)
3194                 goto fib6_rules_init;
3195
3196         ret = -ENOBUFS;
3197         if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL, NULL) ||
3198             __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL, NULL) ||
3199             __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL, NULL))
3200                 goto out_register_late_subsys;
3201
3202         ret = register_netdevice_notifier(&ip6_route_dev_notifier);
3203         if (ret)
3204                 goto out_register_late_subsys;
3205
3206 out:
3207         return ret;
3208
3209 out_register_late_subsys:
3210         unregister_pernet_subsys(&ip6_route_net_late_ops);
3211 fib6_rules_init:
3212         fib6_rules_cleanup();
3213 xfrm6_init:
3214         xfrm6_fini();
3215 out_fib6_init:
3216         fib6_gc_cleanup();
3217 out_register_subsys:
3218         unregister_pernet_subsys(&ip6_route_net_ops);
3219 out_register_inetpeer:
3220         unregister_pernet_subsys(&ipv6_inetpeer_ops);
3221 out_dst_entries:
3222         dst_entries_destroy(&ip6_dst_blackhole_ops);
3223 out_kmem_cache:
3224         kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3225         goto out;
3226 }
3227
3228 void ip6_route_cleanup(void)
3229 {
3230         unregister_netdevice_notifier(&ip6_route_dev_notifier);
3231         unregister_pernet_subsys(&ip6_route_net_late_ops);
3232         fib6_rules_cleanup();
3233         xfrm6_fini();
3234         fib6_gc_cleanup();
3235         unregister_pernet_subsys(&ipv6_inetpeer_ops);
3236         unregister_pernet_subsys(&ip6_route_net_ops);
3237         dst_entries_destroy(&ip6_dst_blackhole_ops);
3238         kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3239 }