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 dst_cow_metrics_generic(dst, old);
 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(work);
 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 mx6_config *mxc)
 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, mxc);
 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 = { .nl_net = dev_net(rt->dst.dev), };
 872         struct mx6_config mxc = { .mx = NULL, };
 873
 874         return __ip6_ins_rt(rt, &info, &mxc);
 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                 goto out;
1246         }
1247
1248         if (rt == net->ipv6.ip6_null_entry) {
1249                 fn = fib6_backtrack(fn, &fl6->saddr);
1250                 if (fn)
1251                         goto restart;
1252         }
1253
1254 out:
1255         dst_hold(&rt->dst);
1256
1257         read_unlock_bh(&table->tb6_lock);
1258
1259         return rt;
1260 };
1261
1262 static struct dst_entry *ip6_route_redirect(struct net *net,
1263                                         const struct flowi6 *fl6,
1264                                         const struct in6_addr *gateway)
1265 {
1266         int flags = RT6_LOOKUP_F_HAS_SADDR;
1267         struct ip6rd_flowi rdfl;
1268
1269         rdfl.fl6 = *fl6;
1270         rdfl.gateway = *gateway;
1271
1272         return fib6_rule_lookup(net, &rdfl.fl6,
1273                                 flags, __ip6_route_redirect);
1274 }
1275
1276 void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark)
1277 {
1278         const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1279         struct dst_entry *dst;
1280         struct flowi6 fl6;
1281
1282         memset(&fl6, 0, sizeof(fl6));
1283         fl6.flowi6_iif = LOOPBACK_IFINDEX;
1284         fl6.flowi6_oif = oif;
1285         fl6.flowi6_mark = mark;
1286         fl6.daddr = iph->daddr;
1287         fl6.saddr = iph->saddr;
1288         fl6.flowlabel = ip6_flowinfo(iph);
1289
1290         dst = ip6_route_redirect(net, &fl6, &ipv6_hdr(skb)->saddr);
1291         rt6_do_redirect(dst, NULL, skb);
1292         dst_release(dst);
1293 }
1294 EXPORT_SYMBOL_GPL(ip6_redirect);
1295
1296 void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif,
1297                             u32 mark)
1298 {
1299         const struct ipv6hdr *iph = ipv6_hdr(skb);
1300         const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
1301         struct dst_entry *dst;
1302         struct flowi6 fl6;
1303
1304         memset(&fl6, 0, sizeof(fl6));
1305         fl6.flowi6_iif = LOOPBACK_IFINDEX;
1306         fl6.flowi6_oif = oif;
1307         fl6.flowi6_mark = mark;
1308         fl6.daddr = msg->dest;
1309         fl6.saddr = iph->daddr;
1310
1311         dst = ip6_route_redirect(net, &fl6, &iph->saddr);
1312         rt6_do_redirect(dst, NULL, skb);
1313         dst_release(dst);
1314 }
1315
1316 void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
1317 {
1318         ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark);
1319 }
1320 EXPORT_SYMBOL_GPL(ip6_sk_redirect);
1321
1322 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1323 {
1324         struct net_device *dev = dst->dev;
1325         unsigned int mtu = dst_mtu(dst);
1326         struct net *net = dev_net(dev);
1327
1328         mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
1329
1330         if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
1331                 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1332
1333         /*
1334          * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1335          * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1336          * IPV6_MAXPLEN is also valid and means: "any MSS,
1337          * rely only on pmtu discovery"
1338          */
1339         if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
1340                 mtu = IPV6_MAXPLEN;
1341         return mtu;
1342 }
1343
1344 static unsigned int ip6_mtu(const struct dst_entry *dst)
1345 {
1346         struct inet6_dev *idev;
1347         unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
1348
1349         if (mtu)
1350                 goto out;
1351
1352         mtu = IPV6_MIN_MTU;
1353
1354         rcu_read_lock();
1355         idev = __in6_dev_get(dst->dev);
1356         if (idev)
1357                 mtu = idev->cnf.mtu6;
1358         rcu_read_unlock();
1359
1360 out:
1361         return min_t(unsigned int, mtu, IP6_MAX_MTU);
1362 }
1363
1364 static struct dst_entry *icmp6_dst_gc_list;
1365 static DEFINE_SPINLOCK(icmp6_dst_lock);
1366
1367 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
1368                                   struct flowi6 *fl6)
1369 {
1370         struct dst_entry *dst;
1371         struct rt6_info *rt;
1372         struct inet6_dev *idev = in6_dev_get(dev);
1373         struct net *net = dev_net(dev);
1374
1375         if (unlikely(!idev))
1376                 return ERR_PTR(-ENODEV);
1377
1378         rt = ip6_dst_alloc(net, dev, 0, NULL);
1379         if (unlikely(!rt)) {
1380                 in6_dev_put(idev);
1381                 dst = ERR_PTR(-ENOMEM);
1382                 goto out;
1383         }
1384
1385         rt->dst.flags |= DST_HOST;
1386         rt->dst.output  = ip6_output;
1387         atomic_set(&rt->dst.__refcnt, 1);
1388         rt->rt6i_gateway  = fl6->daddr;
1389         rt->rt6i_dst.addr = fl6->daddr;
1390         rt->rt6i_dst.plen = 128;
1391         rt->rt6i_idev     = idev;
1392         dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
1393
1394         spin_lock_bh(&icmp6_dst_lock);
1395         rt->dst.next = icmp6_dst_gc_list;
1396         icmp6_dst_gc_list = &rt->dst;
1397         spin_unlock_bh(&icmp6_dst_lock);
1398
1399         fib6_force_start_gc(net);
1400
1401         dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
1402
1403 out:
1404         return dst;
1405 }
1406
1407 int icmp6_dst_gc(void)
1408 {
1409         struct dst_entry *dst, **pprev;
1410         int more = 0;
1411
1412         spin_lock_bh(&icmp6_dst_lock);
1413         pprev = &icmp6_dst_gc_list;
1414
1415         while ((dst = *pprev) != NULL) {
1416                 if (!atomic_read(&dst->__refcnt)) {
1417                         *pprev = dst->next;
1418                         dst_free(dst);
1419                 } else {
1420                         pprev = &dst->next;
1421                         ++more;
1422                 }
1423         }
1424
1425         spin_unlock_bh(&icmp6_dst_lock);
1426
1427         return more;
1428 }
1429
1430 static void icmp6_clean_all(int (*func)(struct rt6_info *rt, void *arg),
1431                             void *arg)
1432 {
1433         struct dst_entry *dst, **pprev;
1434
1435         spin_lock_bh(&icmp6_dst_lock);
1436         pprev = &icmp6_dst_gc_list;
1437         while ((dst = *pprev) != NULL) {
1438                 struct rt6_info *rt = (struct rt6_info *) dst;
1439                 if (func(rt, arg)) {
1440                         *pprev = dst->next;
1441                         dst_free(dst);
1442                 } else {
1443                         pprev = &dst->next;
1444                 }
1445         }
1446         spin_unlock_bh(&icmp6_dst_lock);
1447 }
1448
1449 static int ip6_dst_gc(struct dst_ops *ops)
1450 {
1451         struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
1452         int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1453         int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1454         int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1455         int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1456         unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1457         int entries;
1458
1459         entries = dst_entries_get_fast(ops);
1460         if (time_after(rt_last_gc + rt_min_interval, jiffies) &&
1461             entries <= rt_max_size)
1462                 goto out;
1463
1464         net->ipv6.ip6_rt_gc_expire++;
1465         fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net, true);
1466         entries = dst_entries_get_slow(ops);
1467         if (entries < ops->gc_thresh)
1468                 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1469 out:
1470         net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1471         return entries > rt_max_size;
1472 }
1473
1474 static int ip6_convert_metrics(struct mx6_config *mxc,
1475                                const struct fib6_config *cfg)
1476 {
1477         struct nlattr *nla;
1478         int remaining;
1479         u32 *mp;
1480
1481         if (cfg->fc_mx == NULL)
1482                 return 0;
1483
1484         mp = kzalloc(sizeof(u32) * RTAX_MAX, GFP_KERNEL);
1485         if (unlikely(!mp))
1486                 return -ENOMEM;
1487
1488         nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1489                 int type = nla_type(nla);
1490
1491                 if (type) {
1492                         u32 val;
1493
1494                         if (unlikely(type > RTAX_MAX))
1495                                 goto err;
1496                         if (type == RTAX_CC_ALGO) {
1497                                 char tmp[TCP_CA_NAME_MAX];
1498
1499                                 nla_strlcpy(tmp, nla, sizeof(tmp));
1500                                 val = tcp_ca_get_key_by_name(tmp);
1501                                 if (val == TCP_CA_UNSPEC)
1502                                         goto err;
1503                         } else {
1504                                 val = nla_get_u32(nla);
1505                         }
1506
1507                         mp[type - 1] = val;
1508                         __set_bit(type - 1, mxc->mx_valid);
1509                 }
1510         }
1511
1512         mxc->mx = mp;
1513
1514         return 0;
1515  err:
1516         kfree(mp);
1517         return -EINVAL;
1518 }
1519
1520 int ip6_route_add(struct fib6_config *cfg)
1521 {
1522         int err;
1523         struct net *net = cfg->fc_nlinfo.nl_net;
1524         struct rt6_info *rt = NULL;
1525         struct net_device *dev = NULL;
1526         struct inet6_dev *idev = NULL;
1527         struct fib6_table *table;
1528         struct mx6_config mxc = { .mx = NULL, };
1529         int addr_type;
1530
1531         if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1532                 return -EINVAL;
1533 #ifndef CONFIG_IPV6_SUBTREES
1534         if (cfg->fc_src_len)
1535                 return -EINVAL;
1536 #endif
1537         if (cfg->fc_ifindex) {
1538                 err = -ENODEV;
1539                 dev = dev_get_by_index(net, cfg->fc_ifindex);
1540                 if (!dev)
1541                         goto out;
1542                 idev = in6_dev_get(dev);
1543                 if (!idev)
1544                         goto out;
1545         }
1546
1547         if (cfg->fc_metric == 0)
1548                 cfg->fc_metric = IP6_RT_PRIO_USER;
1549
1550         err = -ENOBUFS;
1551         if (cfg->fc_nlinfo.nlh &&
1552             !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
1553                 table = fib6_get_table(net, cfg->fc_table);
1554                 if (!table) {
1555                         pr_warn("NLM_F_CREATE should be specified when creating new route\n");
1556                         table = fib6_new_table(net, cfg->fc_table);
1557                 }
1558         } else {
1559                 table = fib6_new_table(net, cfg->fc_table);
1560         }
1561
1562         if (!table)
1563                 goto out;
1564
1565         rt = ip6_dst_alloc(net, NULL, (cfg->fc_flags & RTF_ADDRCONF) ? 0 : DST_NOCOUNT, table);
1566
1567         if (!rt) {
1568                 err = -ENOMEM;
1569                 goto out;
1570         }
1571
1572         if (cfg->fc_flags & RTF_EXPIRES)
1573                 rt6_set_expires(rt, jiffies +
1574                                 clock_t_to_jiffies(cfg->fc_expires));
1575         else
1576                 rt6_clean_expires(rt);
1577
1578         if (cfg->fc_protocol == RTPROT_UNSPEC)
1579                 cfg->fc_protocol = RTPROT_BOOT;
1580         rt->rt6i_protocol = cfg->fc_protocol;
1581
1582         addr_type = ipv6_addr_type(&cfg->fc_dst);
1583
1584         if (addr_type & IPV6_ADDR_MULTICAST)
1585                 rt->dst.input = ip6_mc_input;
1586         else if (cfg->fc_flags & RTF_LOCAL)
1587                 rt->dst.input = ip6_input;
1588         else
1589                 rt->dst.input = ip6_forward;
1590
1591         rt->dst.output = ip6_output;
1592
1593         ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1594         rt->rt6i_dst.plen = cfg->fc_dst_len;
1595         if (rt->rt6i_dst.plen == 128) {
1596                 rt->dst.flags |= DST_HOST;
1597                 dst_metrics_set_force_overwrite(&rt->dst);
1598         }
1599
1600 #ifdef CONFIG_IPV6_SUBTREES
1601         ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1602         rt->rt6i_src.plen = cfg->fc_src_len;
1603 #endif
1604
1605         rt->rt6i_metric = cfg->fc_metric;
1606
1607         /* We cannot add true routes via loopback here,
1608            they would result in kernel looping; promote them to reject routes
1609          */
1610         if ((cfg->fc_flags & RTF_REJECT) ||
1611             (dev && (dev->flags & IFF_LOOPBACK) &&
1612              !(addr_type & IPV6_ADDR_LOOPBACK) &&
1613              !(cfg->fc_flags & RTF_LOCAL))) {
1614                 /* hold loopback dev/idev if we haven't done so. */
1615                 if (dev != net->loopback_dev) {
1616                         if (dev) {
1617                                 dev_put(dev);
1618                                 in6_dev_put(idev);
1619                         }
1620                         dev = net->loopback_dev;
1621                         dev_hold(dev);
1622                         idev = in6_dev_get(dev);
1623                         if (!idev) {
1624                                 err = -ENODEV;
1625                                 goto out;
1626                         }
1627                 }
1628                 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1629                 switch (cfg->fc_type) {
1630                 case RTN_BLACKHOLE:
1631                         rt->dst.error = -EINVAL;
1632                         rt->dst.output = dst_discard_sk;
1633                         rt->dst.input = dst_discard;
1634                         break;
1635                 case RTN_PROHIBIT:
1636                         rt->dst.error = -EACCES;
1637                         rt->dst.output = ip6_pkt_prohibit_out;
1638                         rt->dst.input = ip6_pkt_prohibit;
1639                         break;
1640                 case RTN_THROW:
1641                 default:
1642                         rt->dst.error = (cfg->fc_type == RTN_THROW) ? -EAGAIN
1643                                         : -ENETUNREACH;
1644                         rt->dst.output = ip6_pkt_discard_out;
1645                         rt->dst.input = ip6_pkt_discard;
1646                         break;
1647                 }
1648                 goto install_route;
1649         }
1650
1651         if (cfg->fc_flags & RTF_GATEWAY) {
1652                 const struct in6_addr *gw_addr;
1653                 int gwa_type;
1654
1655                 gw_addr = &cfg->fc_gateway;
1656                 rt->rt6i_gateway = *gw_addr;
1657                 gwa_type = ipv6_addr_type(gw_addr);
1658
1659                 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1660                         struct rt6_info *grt;
1661
1662                         /* IPv6 strictly inhibits using not link-local
1663                            addresses as nexthop address.
1664                            Otherwise, router will not able to send redirects.
1665                            It is very good, but in some (rare!) circumstances
1666                            (SIT, PtP, NBMA NOARP links) it is handy to allow
1667                            some exceptions. --ANK
1668                          */
1669                         err = -EINVAL;
1670                         if (!(gwa_type & IPV6_ADDR_UNICAST))
1671                                 goto out;
1672
1673                         grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, 1);
1674
1675                         err = -EHOSTUNREACH;
1676                         if (!grt)
1677                                 goto out;
1678                         if (dev) {
1679                                 if (dev != grt->dst.dev) {
1680                                         ip6_rt_put(grt);
1681                                         goto out;
1682                                 }
1683                         } else {
1684                                 dev = grt->dst.dev;
1685                                 idev = grt->rt6i_idev;
1686                                 dev_hold(dev);
1687                                 in6_dev_hold(grt->rt6i_idev);
1688                         }
1689                         if (!(grt->rt6i_flags & RTF_GATEWAY))
1690                                 err = 0;
1691                         ip6_rt_put(grt);
1692
1693                         if (err)
1694                                 goto out;
1695                 }
1696                 err = -EINVAL;
1697                 if (!dev || (dev->flags & IFF_LOOPBACK))
1698                         goto out;
1699         }
1700
1701         err = -ENODEV;
1702         if (!dev)
1703                 goto out;
1704
1705         if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
1706                 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
1707                         err = -EINVAL;
1708                         goto out;
1709                 }
1710                 rt->rt6i_prefsrc.addr = cfg->fc_prefsrc;
1711                 rt->rt6i_prefsrc.plen = 128;
1712         } else
1713                 rt->rt6i_prefsrc.plen = 0;
1714
1715         rt->rt6i_flags = cfg->fc_flags;
1716
1717 install_route:
1718         rt->dst.dev = dev;
1719         rt->rt6i_idev = idev;
1720         rt->rt6i_table = table;
1721
1722         cfg->fc_nlinfo.nl_net = dev_net(dev);
1723
1724         err = ip6_convert_metrics(&mxc, cfg);
1725         if (err)
1726                 goto out;
1727
1728         err = __ip6_ins_rt(rt, &cfg->fc_nlinfo, &mxc);
1729
1730         kfree(mxc.mx);
1731         return err;
1732 out:
1733         if (dev)
1734                 dev_put(dev);
1735         if (idev)
1736                 in6_dev_put(idev);
1737         if (rt)
1738                 dst_free(&rt->dst);
1739         return err;
1740 }
1741
1742 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
1743 {
1744         int err;
1745         struct fib6_table *table;
1746         struct net *net = dev_net(rt->dst.dev);
1747
1748         if (rt == net->ipv6.ip6_null_entry) {
1749                 err = -ENOENT;
1750                 goto out;
1751         }
1752
1753         table = rt->rt6i_table;
1754         write_lock_bh(&table->tb6_lock);
1755         err = fib6_del(rt, info);
1756         write_unlock_bh(&table->tb6_lock);
1757
1758 out:
1759         ip6_rt_put(rt);
1760         return err;
1761 }
1762
1763 int ip6_del_rt(struct rt6_info *rt)
1764 {
1765         struct nl_info info = {
1766                 .nl_net = dev_net(rt->dst.dev),
1767         };
1768         return __ip6_del_rt(rt, &info);
1769 }
1770
1771 static int ip6_route_del(struct fib6_config *cfg)
1772 {
1773         struct fib6_table *table;
1774         struct fib6_node *fn;
1775         struct rt6_info *rt;
1776         int err = -ESRCH;
1777
1778         table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
1779         if (!table)
1780                 return err;
1781
1782         read_lock_bh(&table->tb6_lock);
1783
1784         fn = fib6_locate(&table->tb6_root,
1785                          &cfg->fc_dst, cfg->fc_dst_len,
1786                          &cfg->fc_src, cfg->fc_src_len);
1787
1788         if (fn) {
1789                 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1790                         if (cfg->fc_ifindex &&
1791                             (!rt->dst.dev ||
1792                              rt->dst.dev->ifindex != cfg->fc_ifindex))
1793                                 continue;
1794                         if (cfg->fc_flags & RTF_GATEWAY &&
1795                             !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
1796                                 continue;
1797                         if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
1798                                 continue;
1799                         dst_hold(&rt->dst);
1800                         read_unlock_bh(&table->tb6_lock);
1801
1802                         return __ip6_del_rt(rt, &cfg->fc_nlinfo);
1803                 }
1804         }
1805         read_unlock_bh(&table->tb6_lock);
1806
1807         return err;
1808 }
1809
1810 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
1811 {
1812         struct net *net = dev_net(skb->dev);
1813         struct netevent_redirect netevent;
1814         struct rt6_info *rt, *nrt = NULL;
1815         struct ndisc_options ndopts;
1816         struct inet6_dev *in6_dev;
1817         struct neighbour *neigh;
1818         struct rd_msg *msg;
1819         int optlen, on_link;
1820         u8 *lladdr;
1821
1822         optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
1823         optlen -= sizeof(*msg);
1824
1825         if (optlen < 0) {
1826                 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
1827                 return;
1828         }
1829
1830         msg = (struct rd_msg *)icmp6_hdr(skb);
1831
1832         if (ipv6_addr_is_multicast(&msg->dest)) {
1833                 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
1834                 return;
1835         }
1836
1837         on_link = 0;
1838         if (ipv6_addr_equal(&msg->dest, &msg->target)) {
1839                 on_link = 1;
1840         } else if (ipv6_addr_type(&msg->target) !=
1841                    (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
1842                 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
1843                 return;
1844         }
1845
1846         in6_dev = __in6_dev_get(skb->dev);
1847         if (!in6_dev)
1848                 return;
1849         if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
1850                 return;
1851
1852         /* RFC2461 8.1:
1853          *      The IP source address of the Redirect MUST be the same as the current
1854          *      first-hop router for the specified ICMP Destination Address.
1855          */
1856
1857         if (!ndisc_parse_options(msg->opt, optlen, &ndopts)) {
1858                 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
1859                 return;
1860         }
1861
1862         lladdr = NULL;
1863         if (ndopts.nd_opts_tgt_lladdr) {
1864                 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
1865                                              skb->dev);
1866                 if (!lladdr) {
1867                         net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
1868                         return;
1869                 }
1870         }
1871
1872         rt = (struct rt6_info *) dst;
1873         if (rt == net->ipv6.ip6_null_entry) {
1874                 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
1875                 return;
1876         }
1877
1878         /* Redirect received -> path was valid.
1879          * Look, redirects are sent only in response to data packets,
1880          * so that this nexthop apparently is reachable. --ANK
1881          */
1882         dst_confirm(&rt->dst);
1883
1884         neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
1885         if (!neigh)
1886                 return;
1887
1888         /*
1889          *      We have finally decided to accept it.
1890          */
1891
1892         neigh_update(neigh, lladdr, NUD_STALE,
1893                      NEIGH_UPDATE_F_WEAK_OVERRIDE|
1894                      NEIGH_UPDATE_F_OVERRIDE|
1895                      (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1896                                      NEIGH_UPDATE_F_ISROUTER))
1897                      );
1898
1899         nrt = ip6_rt_copy(rt, &msg->dest);
1900         if (!nrt)
1901                 goto out;
1902
1903         nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
1904         if (on_link)
1905                 nrt->rt6i_flags &= ~RTF_GATEWAY;
1906
1907         nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
1908
1909         if (ip6_ins_rt(nrt))
1910                 goto out;
1911
1912         netevent.old = &rt->dst;
1913         netevent.new = &nrt->dst;
1914         netevent.daddr = &msg->dest;
1915         netevent.neigh = neigh;
1916         call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
1917
1918         if (rt->rt6i_flags & RTF_CACHE) {
1919                 rt = (struct rt6_info *) dst_clone(&rt->dst);
1920                 ip6_del_rt(rt);
1921         }
1922
1923 out:
1924         neigh_release(neigh);
1925 }
1926
1927 /*
1928  *      Misc support functions
1929  */
1930
1931 static struct rt6_info *ip6_rt_copy(struct rt6_info *ort,
1932                                     const struct in6_addr *dest)
1933 {
1934         struct net *net = dev_net(ort->dst.dev);
1935         struct rt6_info *rt = ip6_dst_alloc(net, ort->dst.dev, 0,
1936                                             ort->rt6i_table);
1937
1938         if (rt) {
1939                 rt->dst.input = ort->dst.input;
1940                 rt->dst.output = ort->dst.output;
1941                 rt->dst.flags |= DST_HOST;
1942
1943                 rt->rt6i_dst.addr = *dest;
1944                 rt->rt6i_dst.plen = 128;
1945                 dst_copy_metrics(&rt->dst, &ort->dst);
1946                 rt->dst.error = ort->dst.error;
1947                 rt->rt6i_idev = ort->rt6i_idev;
1948                 if (rt->rt6i_idev)
1949                         in6_dev_hold(rt->rt6i_idev);
1950                 rt->dst.lastuse = jiffies;
1951
1952                 if (ort->rt6i_flags & RTF_GATEWAY)
1953                         rt->rt6i_gateway = ort->rt6i_gateway;
1954                 else
1955                         rt->rt6i_gateway = *dest;
1956                 rt->rt6i_flags = ort->rt6i_flags;
1957                 rt6_set_from(rt, ort);
1958                 rt->rt6i_metric = 0;
1959
1960 #ifdef CONFIG_IPV6_SUBTREES
1961                 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1962 #endif
1963                 memcpy(&rt->rt6i_prefsrc, &ort->rt6i_prefsrc, sizeof(struct rt6key));
1964                 rt->rt6i_table = ort->rt6i_table;
1965         }
1966         return rt;
1967 }
1968
1969 #ifdef CONFIG_IPV6_ROUTE_INFO
1970 static struct rt6_info *rt6_get_route_info(struct net *net,
1971                                            const struct in6_addr *prefix, int prefixlen,
1972                                            const struct in6_addr *gwaddr, int ifindex)
1973 {
1974         struct fib6_node *fn;
1975         struct rt6_info *rt = NULL;
1976         struct fib6_table *table;
1977
1978         table = fib6_get_table(net, RT6_TABLE_INFO);
1979         if (!table)
1980                 return NULL;
1981
1982         read_lock_bh(&table->tb6_lock);
1983         fn = fib6_locate(&table->tb6_root, prefix, prefixlen, NULL, 0);
1984         if (!fn)
1985                 goto out;
1986
1987         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1988                 if (rt->dst.dev->ifindex != ifindex)
1989                         continue;
1990                 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
1991                         continue;
1992                 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
1993                         continue;
1994                 dst_hold(&rt->dst);
1995                 break;
1996         }
1997 out:
1998         read_unlock_bh(&table->tb6_lock);
1999         return rt;
2000 }
2001
2002 static struct rt6_info *rt6_add_route_info(struct net *net,
2003                                            const struct in6_addr *prefix, int prefixlen,
2004                                            const struct in6_addr *gwaddr, int ifindex,
2005                                            unsigned int pref)
2006 {
2007         struct fib6_config cfg = {
2008                 .fc_table       = RT6_TABLE_INFO,
2009                 .fc_metric      = IP6_RT_PRIO_USER,
2010                 .fc_ifindex     = ifindex,
2011                 .fc_dst_len     = prefixlen,
2012                 .fc_flags       = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
2013                                   RTF_UP | RTF_PREF(pref),
2014                 .fc_nlinfo.portid = 0,
2015                 .fc_nlinfo.nlh = NULL,
2016                 .fc_nlinfo.nl_net = net,
2017         };
2018
2019         cfg.fc_dst = *prefix;
2020         cfg.fc_gateway = *gwaddr;
2021
2022         /* We should treat it as a default route if prefix length is 0. */
2023         if (!prefixlen)
2024                 cfg.fc_flags |= RTF_DEFAULT;
2025
2026         ip6_route_add(&cfg);
2027
2028         return rt6_get_route_info(net, prefix, prefixlen, gwaddr, ifindex);
2029 }
2030 #endif
2031
2032 struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
2033 {
2034         struct rt6_info *rt;
2035         struct fib6_table *table;
2036
2037         table = fib6_get_table(dev_net(dev), RT6_TABLE_DFLT);
2038         if (!table)
2039                 return NULL;
2040
2041         read_lock_bh(&table->tb6_lock);
2042         for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2043                 if (dev == rt->dst.dev &&
2044                     ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
2045                     ipv6_addr_equal(&rt->rt6i_gateway, addr))
2046                         break;
2047         }
2048         if (rt)
2049                 dst_hold(&rt->dst);
2050         read_unlock_bh(&table->tb6_lock);
2051         return rt;
2052 }
2053
2054 struct rt6_info *rt6_add_dflt_router(const struct in6_addr *gwaddr,
2055                                      struct net_device *dev,
2056                                      unsigned int pref)
2057 {
2058         struct fib6_config cfg = {
2059                 .fc_table       = RT6_TABLE_DFLT,
2060                 .fc_metric      = IP6_RT_PRIO_USER,
2061                 .fc_ifindex     = dev->ifindex,
2062                 .fc_flags       = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
2063                                   RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
2064                 .fc_nlinfo.portid = 0,
2065                 .fc_nlinfo.nlh = NULL,
2066                 .fc_nlinfo.nl_net = dev_net(dev),
2067         };
2068
2069         cfg.fc_gateway = *gwaddr;
2070
2071         ip6_route_add(&cfg);
2072
2073         return rt6_get_dflt_router(gwaddr, dev);
2074 }
2075
2076 void rt6_purge_dflt_routers(struct net *net)
2077 {
2078         struct rt6_info *rt;
2079         struct fib6_table *table;
2080
2081         /* NOTE: Keep consistent with rt6_get_dflt_router */
2082         table = fib6_get_table(net, RT6_TABLE_DFLT);
2083         if (!table)
2084                 return;
2085
2086 restart:
2087         read_lock_bh(&table->tb6_lock);
2088         for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2089                 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
2090                     (!rt->rt6i_idev || rt->rt6i_idev->cnf.accept_ra != 2)) {
2091                         dst_hold(&rt->dst);
2092                         read_unlock_bh(&table->tb6_lock);
2093                         ip6_del_rt(rt);
2094                         goto restart;
2095                 }
2096         }
2097         read_unlock_bh(&table->tb6_lock);
2098 }
2099
2100 static void rtmsg_to_fib6_config(struct net *net,
2101                                  struct in6_rtmsg *rtmsg,
2102                                  struct fib6_config *cfg)
2103 {
2104         memset(cfg, 0, sizeof(*cfg));
2105
2106         cfg->fc_table = RT6_TABLE_MAIN;
2107         cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
2108         cfg->fc_metric = rtmsg->rtmsg_metric;
2109         cfg->fc_expires = rtmsg->rtmsg_info;
2110         cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
2111         cfg->fc_src_len = rtmsg->rtmsg_src_len;
2112         cfg->fc_flags = rtmsg->rtmsg_flags;
2113
2114         cfg->fc_nlinfo.nl_net = net;
2115
2116         cfg->fc_dst = rtmsg->rtmsg_dst;
2117         cfg->fc_src = rtmsg->rtmsg_src;
2118         cfg->fc_gateway = rtmsg->rtmsg_gateway;
2119 }
2120
2121 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
2122 {
2123         struct fib6_config cfg;
2124         struct in6_rtmsg rtmsg;
2125         int err;
2126
2127         switch (cmd) {
2128         case SIOCADDRT:         /* Add a route */
2129         case SIOCDELRT:         /* Delete a route */
2130                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2131                         return -EPERM;
2132                 err = copy_from_user(&rtmsg, arg,
2133                                      sizeof(struct in6_rtmsg));
2134                 if (err)
2135                         return -EFAULT;
2136
2137                 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
2138
2139                 rtnl_lock();
2140                 switch (cmd) {
2141                 case SIOCADDRT:
2142                         err = ip6_route_add(&cfg);
2143                         break;
2144                 case SIOCDELRT:
2145                         err = ip6_route_del(&cfg);
2146                         break;
2147                 default:
2148                         err = -EINVAL;
2149                 }
2150                 rtnl_unlock();
2151
2152                 return err;
2153         }
2154
2155         return -EINVAL;
2156 }
2157
2158 /*
2159  *      Drop the packet on the floor
2160  */
2161
2162 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
2163 {
2164         int type;
2165         struct dst_entry *dst = skb_dst(skb);
2166         switch (ipstats_mib_noroutes) {
2167         case IPSTATS_MIB_INNOROUTES:
2168                 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
2169                 if (type == IPV6_ADDR_ANY) {
2170                         IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2171                                       IPSTATS_MIB_INADDRERRORS);
2172                         break;
2173                 }
2174                 /* FALLTHROUGH */
2175         case IPSTATS_MIB_OUTNOROUTES:
2176                 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2177                               ipstats_mib_noroutes);
2178                 break;
2179         }
2180         icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
2181         kfree_skb(skb);
2182         return 0;
2183 }
2184
2185 static int ip6_pkt_discard(struct sk_buff *skb)
2186 {
2187         return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
2188 }
2189
2190 static int ip6_pkt_discard_out(struct sock *sk, struct sk_buff *skb)
2191 {
2192         skb->dev = skb_dst(skb)->dev;
2193         return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
2194 }
2195
2196 static int ip6_pkt_prohibit(struct sk_buff *skb)
2197 {
2198         return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
2199 }
2200
2201 static int ip6_pkt_prohibit_out(struct sock *sk, struct sk_buff *skb)
2202 {
2203         skb->dev = skb_dst(skb)->dev;
2204         return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
2205 }
2206
2207 /*
2208  *      Allocate a dst for local (unicast / anycast) address.
2209  */
2210
2211 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
2212                                     const struct in6_addr *addr,
2213                                     bool anycast)
2214 {
2215         struct net *net = dev_net(idev->dev);
2216         struct rt6_info *rt = ip6_dst_alloc(net, net->loopback_dev,
2217                                             DST_NOCOUNT, NULL);
2218         if (!rt)
2219                 return ERR_PTR(-ENOMEM);
2220
2221         in6_dev_hold(idev);
2222
2223         rt->dst.flags |= DST_HOST;
2224         rt->dst.input = ip6_input;
2225         rt->dst.output = ip6_output;
2226         rt->rt6i_idev = idev;
2227
2228         rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
2229         if (anycast)
2230                 rt->rt6i_flags |= RTF_ANYCAST;
2231         else
2232                 rt->rt6i_flags |= RTF_LOCAL;
2233
2234         rt->rt6i_gateway  = *addr;
2235         rt->rt6i_dst.addr = *addr;
2236         rt->rt6i_dst.plen = 128;
2237         rt->rt6i_table = fib6_get_table(net, RT6_TABLE_LOCAL);
2238
2239         atomic_set(&rt->dst.__refcnt, 1);
2240
2241         return rt;
2242 }
2243
2244 int ip6_route_get_saddr(struct net *net,
2245                         struct rt6_info *rt,
2246                         const struct in6_addr *daddr,
2247                         unsigned int prefs,
2248                         struct in6_addr *saddr)
2249 {
2250         struct inet6_dev *idev = ip6_dst_idev((struct dst_entry *)rt);
2251         int err = 0;
2252         if (rt->rt6i_prefsrc.plen)
2253                 *saddr = rt->rt6i_prefsrc.addr;
2254         else
2255                 err = ipv6_dev_get_saddr(net, idev ? idev->dev : NULL,
2256                                          daddr, prefs, saddr);
2257         return err;
2258 }
2259
2260 /* remove deleted ip from prefsrc entries */
2261 struct arg_dev_net_ip {
2262         struct net_device *dev;
2263         struct net *net;
2264         struct in6_addr *addr;
2265 };
2266
2267 static int fib6_remove_prefsrc(struct rt6_info *rt, void *arg)
2268 {
2269         struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
2270         struct net *net = ((struct arg_dev_net_ip *)arg)->net;
2271         struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
2272
2273         if (((void *)rt->dst.dev == dev || !dev) &&
2274             rt != net->ipv6.ip6_null_entry &&
2275             ipv6_addr_equal(addr, &rt->rt6i_prefsrc.addr)) {
2276                 /* remove prefsrc entry */
2277                 rt->rt6i_prefsrc.plen = 0;
2278         }
2279         return 0;
2280 }
2281
2282 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
2283 {
2284         struct net *net = dev_net(ifp->idev->dev);
2285         struct arg_dev_net_ip adni = {
2286                 .dev = ifp->idev->dev,
2287                 .net = net,
2288                 .addr = &ifp->addr,
2289         };
2290         fib6_clean_all(net, fib6_remove_prefsrc, &adni);
2291 }
2292
2293 #define RTF_RA_ROUTER           (RTF_ADDRCONF | RTF_DEFAULT | RTF_GATEWAY)
2294 #define RTF_CACHE_GATEWAY       (RTF_GATEWAY | RTF_CACHE)
2295
2296 /* Remove routers and update dst entries when gateway turn into host. */
2297 static int fib6_clean_tohost(struct rt6_info *rt, void *arg)
2298 {
2299         struct in6_addr *gateway = (struct in6_addr *)arg;
2300
2301         if ((((rt->rt6i_flags & RTF_RA_ROUTER) == RTF_RA_ROUTER) ||
2302              ((rt->rt6i_flags & RTF_CACHE_GATEWAY) == RTF_CACHE_GATEWAY)) &&
2303              ipv6_addr_equal(gateway, &rt->rt6i_gateway)) {
2304                 return -1;
2305         }
2306         return 0;
2307 }
2308
2309 void rt6_clean_tohost(struct net *net, struct in6_addr *gateway)
2310 {
2311         fib6_clean_all(net, fib6_clean_tohost, gateway);
2312 }
2313
2314 struct arg_dev_net {
2315         struct net_device *dev;
2316         struct net *net;
2317 };
2318
2319 static int fib6_ifdown(struct rt6_info *rt, void *arg)
2320 {
2321         const struct arg_dev_net *adn = arg;
2322         const struct net_device *dev = adn->dev;
2323
2324         if ((rt->dst.dev == dev || !dev) &&
2325             rt != adn->net->ipv6.ip6_null_entry)
2326                 return -1;
2327
2328         return 0;
2329 }
2330
2331 void rt6_ifdown(struct net *net, struct net_device *dev)
2332 {
2333         struct arg_dev_net adn = {
2334                 .dev = dev,
2335                 .net = net,
2336         };
2337
2338         fib6_clean_all(net, fib6_ifdown, &adn);
2339         icmp6_clean_all(fib6_ifdown, &adn);
2340 }
2341
2342 struct rt6_mtu_change_arg {
2343         struct net_device *dev;
2344         unsigned int mtu;
2345 };
2346
2347 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2348 {
2349         struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2350         struct inet6_dev *idev;
2351
2352         /* In IPv6 pmtu discovery is not optional,
2353            so that RTAX_MTU lock cannot disable it.
2354            We still use this lock to block changes
2355            caused by addrconf/ndisc.
2356         */
2357
2358         idev = __in6_dev_get(arg->dev);
2359         if (!idev)
2360                 return 0;
2361
2362         /* For administrative MTU increase, there is no way to discover
2363            IPv6 PMTU increase, so PMTU increase should be updated here.
2364            Since RFC 1981 doesn't include administrative MTU increase
2365            update PMTU increase is a MUST. (i.e. jumbo frame)
2366          */
2367         /*
2368            If new MTU is less than route PMTU, this new MTU will be the
2369            lowest MTU in the path, update the route PMTU to reflect PMTU
2370            decreases; if new MTU is greater than route PMTU, and the
2371            old MTU is the lowest MTU in the path, update the route PMTU
2372            to reflect the increase. In this case if the other nodes' MTU
2373            also have the lowest MTU, TOO BIG MESSAGE will be lead to
2374            PMTU discouvery.
2375          */
2376         if (rt->dst.dev == arg->dev &&
2377             !dst_metric_locked(&rt->dst, RTAX_MTU) &&
2378             (dst_mtu(&rt->dst) >= arg->mtu ||
2379              (dst_mtu(&rt->dst) < arg->mtu &&
2380               dst_mtu(&rt->dst) == idev->cnf.mtu6))) {
2381                 dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
2382         }
2383         return 0;
2384 }
2385
2386 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
2387 {
2388         struct rt6_mtu_change_arg arg = {
2389                 .dev = dev,
2390                 .mtu = mtu,
2391         };
2392
2393         fib6_clean_all(dev_net(dev), rt6_mtu_change_route, &arg);
2394 }
2395
2396 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2397         [RTA_GATEWAY]           = { .len = sizeof(struct in6_addr) },
2398         [RTA_OIF]               = { .type = NLA_U32 },
2399         [RTA_IIF]               = { .type = NLA_U32 },
2400         [RTA_PRIORITY]          = { .type = NLA_U32 },
2401         [RTA_METRICS]           = { .type = NLA_NESTED },
2402         [RTA_MULTIPATH]         = { .len = sizeof(struct rtnexthop) },
2403 };
2404
2405 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2406                               struct fib6_config *cfg)
2407 {
2408         struct rtmsg *rtm;
2409         struct nlattr *tb[RTA_MAX+1];
2410         int err;
2411
2412         err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2413         if (err < 0)
2414                 goto errout;
2415
2416         err = -EINVAL;
2417         rtm = nlmsg_data(nlh);
2418         memset(cfg, 0, sizeof(*cfg));
2419
2420         cfg->fc_table = rtm->rtm_table;
2421         cfg->fc_dst_len = rtm->rtm_dst_len;
2422         cfg->fc_src_len = rtm->rtm_src_len;
2423         cfg->fc_flags = RTF_UP;
2424         cfg->fc_protocol = rtm->rtm_protocol;
2425         cfg->fc_type = rtm->rtm_type;
2426
2427         if (rtm->rtm_type == RTN_UNREACHABLE ||
2428             rtm->rtm_type == RTN_BLACKHOLE ||
2429             rtm->rtm_type == RTN_PROHIBIT ||
2430             rtm->rtm_type == RTN_THROW)
2431                 cfg->fc_flags |= RTF_REJECT;
2432
2433         if (rtm->rtm_type == RTN_LOCAL)
2434                 cfg->fc_flags |= RTF_LOCAL;
2435
2436         cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid;
2437         cfg->fc_nlinfo.nlh = nlh;
2438         cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2439
2440         if (tb[RTA_GATEWAY]) {
2441                 nla_memcpy(&cfg->fc_gateway, tb[RTA_GATEWAY], 16);
2442                 cfg->fc_flags |= RTF_GATEWAY;
2443         }
2444
2445         if (tb[RTA_DST]) {
2446                 int plen = (rtm->rtm_dst_len + 7) >> 3;
2447
2448                 if (nla_len(tb[RTA_DST]) < plen)
2449                         goto errout;
2450
2451                 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2452         }
2453
2454         if (tb[RTA_SRC]) {
2455                 int plen = (rtm->rtm_src_len + 7) >> 3;
2456
2457                 if (nla_len(tb[RTA_SRC]) < plen)
2458                         goto errout;
2459
2460                 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2461         }
2462
2463         if (tb[RTA_PREFSRC])
2464                 nla_memcpy(&cfg->fc_prefsrc, tb[RTA_PREFSRC], 16);
2465
2466         if (tb[RTA_OIF])
2467                 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2468
2469         if (tb[RTA_PRIORITY])
2470                 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2471
2472         if (tb[RTA_METRICS]) {
2473                 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2474                 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2475         }
2476
2477         if (tb[RTA_TABLE])
2478                 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2479
2480         if (tb[RTA_MULTIPATH]) {
2481                 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
2482                 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
2483         }
2484
2485         err = 0;
2486 errout:
2487         return err;
2488 }
2489
2490 static int ip6_route_multipath(struct fib6_config *cfg, int add)
2491 {
2492         struct fib6_config r_cfg;
2493         struct rtnexthop *rtnh;
2494         int remaining;
2495         int attrlen;
2496         int err = 0, last_err = 0;
2497
2498 beginning:
2499         rtnh = (struct rtnexthop *)cfg->fc_mp;
2500         remaining = cfg->fc_mp_len;
2501
2502         /* Parse a Multipath Entry */
2503         while (rtnh_ok(rtnh, remaining)) {
2504                 memcpy(&r_cfg, cfg, sizeof(*cfg));
2505                 if (rtnh->rtnh_ifindex)
2506                         r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
2507
2508                 attrlen = rtnh_attrlen(rtnh);
2509                 if (attrlen > 0) {
2510                         struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
2511
2512                         nla = nla_find(attrs, attrlen, RTA_GATEWAY);
2513                         if (nla) {
2514                                 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
2515                                 r_cfg.fc_flags |= RTF_GATEWAY;
2516                         }
2517                 }
2518                 err = add ? ip6_route_add(&r_cfg) : ip6_route_del(&r_cfg);
2519                 if (err) {
2520                         last_err = err;
2521                         /* If we are trying to remove a route, do not stop the
2522                          * loop when ip6_route_del() fails (because next hop is
2523                          * already gone), we should try to remove all next hops.
2524                          */
2525                         if (add) {
2526                                 /* If add fails, we should try to delete all
2527                                  * next hops that have been already added.
2528                                  */
2529                                 add = 0;
2530                                 goto beginning;
2531                         }
2532                 }
2533                 /* Because each route is added like a single route we remove
2534                  * this flag after the first nexthop (if there is a collision,
2535                  * we have already fail to add the first nexthop:
2536                  * fib6_add_rt2node() has reject it).
2537                  */
2538                 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~NLM_F_EXCL;
2539                 rtnh = rtnh_next(rtnh, &remaining);
2540         }
2541
2542         return last_err;
2543 }
2544
2545 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh)
2546 {
2547         struct fib6_config cfg;
2548         int err;
2549
2550         err = rtm_to_fib6_config(skb, nlh, &cfg);
2551         if (err < 0)
2552                 return err;
2553
2554         if (cfg.fc_mp)
2555                 return ip6_route_multipath(&cfg, 0);
2556         else
2557                 return ip6_route_del(&cfg);
2558 }
2559
2560 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh)
2561 {
2562         struct fib6_config cfg;
2563         int err;
2564
2565         err = rtm_to_fib6_config(skb, nlh, &cfg);
2566         if (err < 0)
2567                 return err;
2568
2569         if (cfg.fc_mp)
2570                 return ip6_route_multipath(&cfg, 1);
2571         else
2572                 return ip6_route_add(&cfg);
2573 }
2574
2575 static inline size_t rt6_nlmsg_size(void)
2576 {
2577         return NLMSG_ALIGN(sizeof(struct rtmsg))
2578                + nla_total_size(16) /* RTA_SRC */
2579                + nla_total_size(16) /* RTA_DST */
2580                + nla_total_size(16) /* RTA_GATEWAY */
2581                + nla_total_size(16) /* RTA_PREFSRC */
2582                + nla_total_size(4) /* RTA_TABLE */
2583                + nla_total_size(4) /* RTA_IIF */
2584                + nla_total_size(4) /* RTA_OIF */
2585                + nla_total_size(4) /* RTA_PRIORITY */
2586                + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
2587                + nla_total_size(sizeof(struct rta_cacheinfo))
2588                + nla_total_size(TCP_CA_NAME_MAX); /* RTAX_CC_ALGO */
2589 }
2590
2591 static int rt6_fill_node(struct net *net,
2592                          struct sk_buff *skb, struct rt6_info *rt,
2593                          struct in6_addr *dst, struct in6_addr *src,
2594                          int iif, int type, u32 portid, u32 seq,
2595                          int prefix, int nowait, unsigned int flags)
2596 {
2597         struct rtmsg *rtm;
2598         struct nlmsghdr *nlh;
2599         long expires;
2600         u32 table;
2601
2602         if (prefix) {   /* user wants prefix routes only */
2603                 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
2604                         /* success since this is not a prefix route */
2605                         return 1;
2606                 }
2607         }
2608
2609         nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
2610         if (!nlh)
2611                 return -EMSGSIZE;
2612
2613         rtm = nlmsg_data(nlh);
2614         rtm->rtm_family = AF_INET6;
2615         rtm->rtm_dst_len = rt->rt6i_dst.plen;
2616         rtm->rtm_src_len = rt->rt6i_src.plen;
2617         rtm->rtm_tos = 0;
2618         if (rt->rt6i_table)
2619                 table = rt->rt6i_table->tb6_id;
2620         else
2621                 table = RT6_TABLE_UNSPEC;
2622         rtm->rtm_table = table;
2623         if (nla_put_u32(skb, RTA_TABLE, table))
2624                 goto nla_put_failure;
2625         if (rt->rt6i_flags & RTF_REJECT) {
2626                 switch (rt->dst.error) {
2627                 case -EINVAL:
2628                         rtm->rtm_type = RTN_BLACKHOLE;
2629                         break;
2630                 case -EACCES:
2631                         rtm->rtm_type = RTN_PROHIBIT;
2632                         break;
2633                 case -EAGAIN:
2634                         rtm->rtm_type = RTN_THROW;
2635                         break;
2636                 default:
2637                         rtm->rtm_type = RTN_UNREACHABLE;
2638                         break;
2639                 }
2640         }
2641         else if (rt->rt6i_flags & RTF_LOCAL)
2642                 rtm->rtm_type = RTN_LOCAL;
2643         else if (rt->dst.dev && (rt->dst.dev->flags & IFF_LOOPBACK))
2644                 rtm->rtm_type = RTN_LOCAL;
2645         else
2646                 rtm->rtm_type = RTN_UNICAST;
2647         rtm->rtm_flags = 0;
2648         rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2649         rtm->rtm_protocol = rt->rt6i_protocol;
2650         if (rt->rt6i_flags & RTF_DYNAMIC)
2651                 rtm->rtm_protocol = RTPROT_REDIRECT;
2652         else if (rt->rt6i_flags & RTF_ADDRCONF) {
2653                 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ROUTEINFO))
2654                         rtm->rtm_protocol = RTPROT_RA;
2655                 else
2656                         rtm->rtm_protocol = RTPROT_KERNEL;
2657         }
2658
2659         if (rt->rt6i_flags & RTF_CACHE)
2660                 rtm->rtm_flags |= RTM_F_CLONED;
2661
2662         if (dst) {
2663                 if (nla_put(skb, RTA_DST, 16, dst))
2664                         goto nla_put_failure;
2665                 rtm->rtm_dst_len = 128;
2666         } else if (rtm->rtm_dst_len)
2667                 if (nla_put(skb, RTA_DST, 16, &rt->rt6i_dst.addr))
2668                         goto nla_put_failure;
2669 #ifdef CONFIG_IPV6_SUBTREES
2670         if (src) {
2671                 if (nla_put(skb, RTA_SRC, 16, src))
2672                         goto nla_put_failure;
2673                 rtm->rtm_src_len = 128;
2674         } else if (rtm->rtm_src_len &&
2675                    nla_put(skb, RTA_SRC, 16, &rt->rt6i_src.addr))
2676                 goto nla_put_failure;
2677 #endif
2678         if (iif) {
2679 #ifdef CONFIG_IPV6_MROUTE
2680                 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
2681                         int err = ip6mr_get_route(net, skb, rtm, nowait);
2682                         if (err <= 0) {
2683                                 if (!nowait) {
2684                                         if (err == 0)
2685                                                 return 0;
2686                                         goto nla_put_failure;
2687                                 } else {
2688                                         if (err == -EMSGSIZE)
2689                                                 goto nla_put_failure;
2690                                 }
2691                         }
2692                 } else
2693 #endif
2694                         if (nla_put_u32(skb, RTA_IIF, iif))
2695                                 goto nla_put_failure;
2696         } else if (dst) {
2697                 struct in6_addr saddr_buf;
2698                 if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0 &&
2699                     nla_put(skb, RTA_PREFSRC, 16, &saddr_buf))
2700                         goto nla_put_failure;
2701         }
2702
2703         if (rt->rt6i_prefsrc.plen) {
2704                 struct in6_addr saddr_buf;
2705                 saddr_buf = rt->rt6i_prefsrc.addr;
2706                 if (nla_put(skb, RTA_PREFSRC, 16, &saddr_buf))
2707                         goto nla_put_failure;
2708         }
2709
2710         if (rtnetlink_put_metrics(skb, dst_metrics_ptr(&rt->dst)) < 0)
2711                 goto nla_put_failure;
2712
2713         if (rt->rt6i_flags & RTF_GATEWAY) {
2714                 if (nla_put(skb, RTA_GATEWAY, 16, &rt->rt6i_gateway) < 0)
2715                         goto nla_put_failure;
2716         }
2717
2718         if (rt->dst.dev &&
2719             nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex))
2720                 goto nla_put_failure;
2721         if (nla_put_u32(skb, RTA_PRIORITY, rt->rt6i_metric))
2722                 goto nla_put_failure;
2723
2724         expires = (rt->rt6i_flags & RTF_EXPIRES) ? rt->dst.expires - jiffies : 0;
2725
2726         if (rtnl_put_cacheinfo(skb, &rt->dst, 0, expires, rt->dst.error) < 0)
2727                 goto nla_put_failure;
2728
2729         nlmsg_end(skb, nlh);
2730         return 0;
2731
2732 nla_put_failure:
2733         nlmsg_cancel(skb, nlh);
2734         return -EMSGSIZE;
2735 }
2736
2737 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
2738 {
2739         struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
2740         int prefix;
2741
2742         if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
2743                 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
2744                 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
2745         } else
2746                 prefix = 0;
2747
2748         return rt6_fill_node(arg->net,
2749                      arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
2750                      NETLINK_CB(arg->cb->skb).portid, arg->cb->nlh->nlmsg_seq,
2751                      prefix, 0, NLM_F_MULTI);
2752 }
2753
2754 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh)
2755 {
2756         struct net *net = sock_net(in_skb->sk);
2757         struct nlattr *tb[RTA_MAX+1];
2758         struct rt6_info *rt;
2759         struct sk_buff *skb;
2760         struct rtmsg *rtm;
2761         struct flowi6 fl6;
2762         int err, iif = 0, oif = 0;
2763
2764         err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2765         if (err < 0)
2766                 goto errout;
2767
2768         err = -EINVAL;
2769         memset(&fl6, 0, sizeof(fl6));
2770
2771         if (tb[RTA_SRC]) {
2772                 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
2773                         goto errout;
2774
2775                 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
2776         }
2777
2778         if (tb[RTA_DST]) {
2779                 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
2780                         goto errout;
2781
2782                 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
2783         }
2784
2785         if (tb[RTA_IIF])
2786                 iif = nla_get_u32(tb[RTA_IIF]);
2787
2788         if (tb[RTA_OIF])
2789                 oif = nla_get_u32(tb[RTA_OIF]);
2790
2791         if (tb[RTA_MARK])
2792                 fl6.flowi6_mark = nla_get_u32(tb[RTA_MARK]);
2793
2794         if (iif) {
2795                 struct net_device *dev;
2796                 int flags = 0;
2797
2798                 dev = __dev_get_by_index(net, iif);
2799                 if (!dev) {
2800                         err = -ENODEV;
2801                         goto errout;
2802                 }
2803
2804                 fl6.flowi6_iif = iif;
2805
2806                 if (!ipv6_addr_any(&fl6.saddr))
2807                         flags |= RT6_LOOKUP_F_HAS_SADDR;
2808
2809                 rt = (struct rt6_info *)ip6_route_input_lookup(net, dev, &fl6,
2810                                                                flags);
2811         } else {
2812                 fl6.flowi6_oif = oif;
2813
2814                 rt = (struct rt6_info *)ip6_route_output(net, NULL, &fl6);
2815         }
2816
2817         skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
2818         if (!skb) {
2819                 ip6_rt_put(rt);
2820                 err = -ENOBUFS;
2821                 goto errout;
2822         }
2823
2824         /* Reserve room for dummy headers, this skb can pass
2825            through good chunk of routing engine.
2826          */
2827         skb_reset_mac_header(skb);
2828         skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
2829
2830         skb_dst_set(skb, &rt->dst);
2831
2832         err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
2833                             RTM_NEWROUTE, NETLINK_CB(in_skb).portid,
2834                             nlh->nlmsg_seq, 0, 0, 0);
2835         if (err < 0) {
2836                 kfree_skb(skb);
2837                 goto errout;
2838         }
2839
2840         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
2841 errout:
2842         return err;
2843 }
2844
2845 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info)
2846 {
2847         struct sk_buff *skb;
2848         struct net *net = info->nl_net;
2849         u32 seq;
2850         int err;
2851
2852         err = -ENOBUFS;
2853         seq = info->nlh ? info->nlh->nlmsg_seq : 0;
2854
2855         skb = nlmsg_new(rt6_nlmsg_size(), gfp_any());
2856         if (!skb)
2857                 goto errout;
2858
2859         err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
2860                                 event, info->portid, seq, 0, 0, 0);
2861         if (err < 0) {
2862                 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
2863                 WARN_ON(err == -EMSGSIZE);
2864                 kfree_skb(skb);
2865                 goto errout;
2866         }
2867         rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
2868                     info->nlh, gfp_any());
2869         return;
2870 errout:
2871         if (err < 0)
2872                 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
2873 }
2874
2875 static int ip6_route_dev_notify(struct notifier_block *this,
2876                                 unsigned long event, void *ptr)
2877 {
2878         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2879         struct net *net = dev_net(dev);
2880
2881         if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
2882                 net->ipv6.ip6_null_entry->dst.dev = dev;
2883                 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
2884 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2885                 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
2886                 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
2887                 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
2888                 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
2889 #endif
2890         }
2891
2892         return NOTIFY_OK;
2893 }
2894
2895 /*
2896  *      /proc
2897  */
2898
2899 #ifdef CONFIG_PROC_FS
2900
2901 static const struct file_operations ipv6_route_proc_fops = {
2902         .owner          = THIS_MODULE,
2903         .open           = ipv6_route_open,
2904         .read           = seq_read,
2905         .llseek         = seq_lseek,
2906         .release        = seq_release_net,
2907 };
2908
2909 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
2910 {
2911         struct net *net = (struct net *)seq->private;
2912         seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
2913                    net->ipv6.rt6_stats->fib_nodes,
2914                    net->ipv6.rt6_stats->fib_route_nodes,
2915                    net->ipv6.rt6_stats->fib_rt_alloc,
2916                    net->ipv6.rt6_stats->fib_rt_entries,
2917                    net->ipv6.rt6_stats->fib_rt_cache,
2918                    dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
2919                    net->ipv6.rt6_stats->fib_discarded_routes);
2920
2921         return 0;
2922 }
2923
2924 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
2925 {
2926         return single_open_net(inode, file, rt6_stats_seq_show);
2927 }
2928
2929 static const struct file_operations rt6_stats_seq_fops = {
2930         .owner   = THIS_MODULE,
2931         .open    = rt6_stats_seq_open,
2932         .read    = seq_read,
2933         .llseek  = seq_lseek,
2934         .release = single_release_net,
2935 };
2936 #endif  /* CONFIG_PROC_FS */
2937
2938 #ifdef CONFIG_SYSCTL
2939
2940 static
2941 int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
2942                               void __user *buffer, size_t *lenp, loff_t *ppos)
2943 {
2944         struct net *net;
2945         int delay;
2946         if (!write)
2947                 return -EINVAL;
2948
2949         net = (struct net *)ctl->extra1;
2950         delay = net->ipv6.sysctl.flush_delay;
2951         proc_dointvec(ctl, write, buffer, lenp, ppos);
2952         fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
2953         return 0;
2954 }
2955
2956 struct ctl_table ipv6_route_table_template[] = {
2957         {
2958                 .procname       =       "flush",
2959                 .data           =       &init_net.ipv6.sysctl.flush_delay,
2960                 .maxlen         =       sizeof(int),
2961                 .mode           =       0200,
2962                 .proc_handler   =       ipv6_sysctl_rtcache_flush
2963         },
2964         {
2965                 .procname       =       "gc_thresh",
2966                 .data           =       &ip6_dst_ops_template.gc_thresh,
2967                 .maxlen         =       sizeof(int),
2968                 .mode           =       0644,
2969                 .proc_handler   =       proc_dointvec,
2970         },
2971         {
2972                 .procname       =       "max_size",
2973                 .data           =       &init_net.ipv6.sysctl.ip6_rt_max_size,
2974                 .maxlen         =       sizeof(int),
2975                 .mode           =       0644,
2976                 .proc_handler   =       proc_dointvec,
2977         },
2978         {
2979                 .procname       =       "gc_min_interval",
2980                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2981                 .maxlen         =       sizeof(int),
2982                 .mode           =       0644,
2983                 .proc_handler   =       proc_dointvec_jiffies,
2984         },
2985         {
2986                 .procname       =       "gc_timeout",
2987                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
2988                 .maxlen         =       sizeof(int),
2989                 .mode           =       0644,
2990                 .proc_handler   =       proc_dointvec_jiffies,
2991         },
2992         {
2993                 .procname       =       "gc_interval",
2994                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_interval,
2995                 .maxlen         =       sizeof(int),
2996                 .mode           =       0644,
2997                 .proc_handler   =       proc_dointvec_jiffies,
2998         },
2999         {
3000                 .procname       =       "gc_elasticity",
3001                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
3002                 .maxlen         =       sizeof(int),
3003                 .mode           =       0644,
3004                 .proc_handler   =       proc_dointvec,
3005         },
3006         {
3007                 .procname       =       "mtu_expires",
3008                 .data           =       &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
3009                 .maxlen         =       sizeof(int),
3010                 .mode           =       0644,
3011                 .proc_handler   =       proc_dointvec_jiffies,
3012         },
3013         {
3014                 .procname       =       "min_adv_mss",
3015                 .data           =       &init_net.ipv6.sysctl.ip6_rt_min_advmss,
3016                 .maxlen         =       sizeof(int),
3017                 .mode           =       0644,
3018                 .proc_handler   =       proc_dointvec,
3019         },
3020         {
3021                 .procname       =       "gc_min_interval_ms",
3022                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
3023                 .maxlen         =       sizeof(int),
3024                 .mode           =       0644,
3025                 .proc_handler   =       proc_dointvec_ms_jiffies,
3026         },
3027         { }
3028 };
3029
3030 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
3031 {
3032         struct ctl_table *table;
3033
3034         table = kmemdup(ipv6_route_table_template,
3035                         sizeof(ipv6_route_table_template),
3036                         GFP_KERNEL);
3037
3038         if (table) {
3039                 table[0].data = &net->ipv6.sysctl.flush_delay;
3040                 table[0].extra1 = net;
3041                 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
3042                 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
3043                 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
3044                 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
3045                 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
3046                 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
3047                 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
3048                 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
3049                 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
3050
3051                 /* Don't export sysctls to unprivileged users */
3052                 if (net->user_ns != &init_user_ns)
3053                         table[0].procname = NULL;
3054         }
3055
3056         return table;
3057 }
3058 #endif
3059
3060 static int __net_init ip6_route_net_init(struct net *net)
3061 {
3062         int ret = -ENOMEM;
3063
3064         memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
3065                sizeof(net->ipv6.ip6_dst_ops));
3066
3067         if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
3068                 goto out_ip6_dst_ops;
3069
3070         net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
3071                                            sizeof(*net->ipv6.ip6_null_entry),
3072                                            GFP_KERNEL);
3073         if (!net->ipv6.ip6_null_entry)
3074                 goto out_ip6_dst_entries;
3075         net->ipv6.ip6_null_entry->dst.path =
3076                 (struct dst_entry *)net->ipv6.ip6_null_entry;
3077         net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3078         dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
3079                          ip6_template_metrics, true);
3080
3081 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3082         net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
3083                                                sizeof(*net->ipv6.ip6_prohibit_entry),
3084                                                GFP_KERNEL);
3085         if (!net->ipv6.ip6_prohibit_entry)
3086                 goto out_ip6_null_entry;
3087         net->ipv6.ip6_prohibit_entry->dst.path =
3088                 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
3089         net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3090         dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
3091                          ip6_template_metrics, true);
3092
3093         net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
3094                                                sizeof(*net->ipv6.ip6_blk_hole_entry),
3095                                                GFP_KERNEL);
3096         if (!net->ipv6.ip6_blk_hole_entry)
3097                 goto out_ip6_prohibit_entry;
3098         net->ipv6.ip6_blk_hole_entry->dst.path =
3099                 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
3100         net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3101         dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
3102                          ip6_template_metrics, true);
3103 #endif
3104
3105         net->ipv6.sysctl.flush_delay = 0;
3106         net->ipv6.sysctl.ip6_rt_max_size = 4096;
3107         net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
3108         net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
3109         net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
3110         net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
3111         net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
3112         net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
3113
3114         net->ipv6.ip6_rt_gc_expire = 30*HZ;
3115
3116         ret = 0;
3117 out:
3118         return ret;
3119
3120 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3121 out_ip6_prohibit_entry:
3122         kfree(net->ipv6.ip6_prohibit_entry);
3123 out_ip6_null_entry:
3124         kfree(net->ipv6.ip6_null_entry);
3125 #endif
3126 out_ip6_dst_entries:
3127         dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3128 out_ip6_dst_ops:
3129         goto out;
3130 }
3131
3132 static void __net_exit ip6_route_net_exit(struct net *net)
3133 {
3134         kfree(net->ipv6.ip6_null_entry);
3135 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3136         kfree(net->ipv6.ip6_prohibit_entry);
3137         kfree(net->ipv6.ip6_blk_hole_entry);
3138 #endif
3139         dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3140 }
3141
3142 static int __net_init ip6_route_net_init_late(struct net *net)
3143 {
3144 #ifdef CONFIG_PROC_FS
3145         proc_create("ipv6_route", 0, net->proc_net, &ipv6_route_proc_fops);
3146         proc_create("rt6_stats", S_IRUGO, net->proc_net, &rt6_stats_seq_fops);
3147 #endif
3148         return 0;
3149 }
3150
3151 static void __net_exit ip6_route_net_exit_late(struct net *net)
3152 {
3153 #ifdef CONFIG_PROC_FS
3154         remove_proc_entry("ipv6_route", net->proc_net);
3155         remove_proc_entry("rt6_stats", net->proc_net);
3156 #endif
3157 }
3158
3159 static struct pernet_operations ip6_route_net_ops = {
3160         .init = ip6_route_net_init,
3161         .exit = ip6_route_net_exit,
3162 };
3163
3164 static int __net_init ipv6_inetpeer_init(struct net *net)
3165 {
3166         struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
3167
3168         if (!bp)
3169                 return -ENOMEM;
3170         inet_peer_base_init(bp);
3171         net->ipv6.peers = bp;
3172         return 0;
3173 }
3174
3175 static void __net_exit ipv6_inetpeer_exit(struct net *net)
3176 {
3177         struct inet_peer_base *bp = net->ipv6.peers;
3178
3179         net->ipv6.peers = NULL;
3180         inetpeer_invalidate_tree(bp);
3181         kfree(bp);
3182 }
3183
3184 static struct pernet_operations ipv6_inetpeer_ops = {
3185         .init   =       ipv6_inetpeer_init,
3186         .exit   =       ipv6_inetpeer_exit,
3187 };
3188
3189 static struct pernet_operations ip6_route_net_late_ops = {
3190         .init = ip6_route_net_init_late,
3191         .exit = ip6_route_net_exit_late,
3192 };
3193
3194 static struct notifier_block ip6_route_dev_notifier = {
3195         .notifier_call = ip6_route_dev_notify,
3196         .priority = 0,
3197 };
3198
3199 int __init ip6_route_init(void)
3200 {
3201         int ret;
3202
3203         ret = -ENOMEM;
3204         ip6_dst_ops_template.kmem_cachep =
3205                 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
3206                                   SLAB_HWCACHE_ALIGN, NULL);
3207         if (!ip6_dst_ops_template.kmem_cachep)
3208                 goto out;
3209
3210         ret = dst_entries_init(&ip6_dst_blackhole_ops);
3211         if (ret)
3212                 goto out_kmem_cache;
3213
3214         ret = register_pernet_subsys(&ipv6_inetpeer_ops);
3215         if (ret)
3216                 goto out_dst_entries;
3217
3218         ret = register_pernet_subsys(&ip6_route_net_ops);
3219         if (ret)
3220                 goto out_register_inetpeer;
3221
3222         ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
3223
3224         /* Registering of the loopback is done before this portion of code,
3225          * the loopback reference in rt6_info will not be taken, do it
3226          * manually for init_net */
3227         init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
3228         init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3229   #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3230         init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
3231         init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3232         init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
3233         init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3234   #endif
3235         ret = fib6_init();
3236         if (ret)
3237                 goto out_register_subsys;
3238
3239         ret = xfrm6_init();
3240         if (ret)
3241                 goto out_fib6_init;
3242
3243         ret = fib6_rules_init();
3244         if (ret)
3245                 goto xfrm6_init;
3246
3247         ret = register_pernet_subsys(&ip6_route_net_late_ops);
3248         if (ret)
3249                 goto fib6_rules_init;
3250
3251         ret = -ENOBUFS;
3252         if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL, NULL) ||
3253             __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL, NULL) ||
3254             __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL, NULL))
3255                 goto out_register_late_subsys;
3256
3257         ret = register_netdevice_notifier(&ip6_route_dev_notifier);
3258         if (ret)
3259                 goto out_register_late_subsys;
3260
3261 out:
3262         return ret;
3263
3264 out_register_late_subsys:
3265         unregister_pernet_subsys(&ip6_route_net_late_ops);
3266 fib6_rules_init:
3267         fib6_rules_cleanup();
3268 xfrm6_init:
3269         xfrm6_fini();
3270 out_fib6_init:
3271         fib6_gc_cleanup();
3272 out_register_subsys:
3273         unregister_pernet_subsys(&ip6_route_net_ops);
3274 out_register_inetpeer:
3275         unregister_pernet_subsys(&ipv6_inetpeer_ops);
3276 out_dst_entries:
3277         dst_entries_destroy(&ip6_dst_blackhole_ops);
3278 out_kmem_cache:
3279         kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3280         goto out;
3281 }
3282
3283 void ip6_route_cleanup(void)
3284 {
3285         unregister_netdevice_notifier(&ip6_route_dev_notifier);
3286         unregister_pernet_subsys(&ip6_route_net_late_ops);
3287         fib6_rules_cleanup();
3288         xfrm6_fini();
3289         fib6_gc_cleanup();
3290         unregister_pernet_subsys(&ipv6_inetpeer_ops);
3291         unregister_pernet_subsys(&ip6_route_net_ops);
3292         dst_entries_destroy(&ip6_dst_blackhole_ops);
3293         kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3294 }