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 sk_buff *skb);
  88 static int              ip6_pkt_prohibit(struct sk_buff *skb);
  89 static int              ip6_pkt_prohibit_out(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 u32 *ipv6_cow_metrics(struct dst_entry *dst, unsigned long old)
 108 {
 109         struct rt6_info *rt = (struct rt6_info *) dst;
 110         struct inet_peer *peer;
 111         u32 *p = NULL;
 112
 113         if (!(rt->dst.flags & DST_HOST))
 114                 return NULL;
 115
 116         peer = rt6_get_peer_create(rt);
 117         if (peer) {
 118                 u32 *old_p = __DST_METRICS_PTR(old);
 119                 unsigned long prev, new;
 120
 121                 p = peer->metrics;
 122                 if (inet_metrics_new(peer))
 123                         memcpy(p, old_p, sizeof(u32) * RTAX_MAX);
 124
 125                 new = (unsigned long) p;
 126                 prev = cmpxchg(&dst->_metrics, old, new);
 127
 128                 if (prev != old) {
 129                         p = __DST_METRICS_PTR(prev);
 130                         if (prev & DST_METRICS_READ_ONLY)
 131                                 p = NULL;
 132                 }
 133         }
 134         return p;
 135 }
 136
 137 static inline const void *choose_neigh_daddr(struct rt6_info *rt,
 138                                              struct sk_buff *skb,
 139                                              const void *daddr)
 140 {
 141         struct in6_addr *p = &rt->rt6i_gateway;
 142
 143         if (!ipv6_addr_any(p))
 144                 return (const void *) p;
 145         else if (skb)
 146                 return &ipv6_hdr(skb)->daddr;
 147         return daddr;
 148 }
 149
 150 static struct neighbour *ip6_neigh_lookup(const struct dst_entry *dst,
 151                                           struct sk_buff *skb,
 152                                           const void *daddr)
 153 {
 154         struct rt6_info *rt = (struct rt6_info *) dst;
 155         struct neighbour *n;
 156
 157         daddr = choose_neigh_daddr(rt, skb, daddr);
 158         n = __ipv6_neigh_lookup(dst->dev, daddr);
 159         if (n)
 160                 return n;
 161         return neigh_create(&nd_tbl, daddr, dst->dev);
 162 }
 163
 164 static struct dst_ops ip6_dst_ops_template = {
 165         .family                 =       AF_INET6,
 166         .protocol               =       cpu_to_be16(ETH_P_IPV6),
 167         .gc                     =       ip6_dst_gc,
 168         .gc_thresh              =       1024,
 169         .check                  =       ip6_dst_check,
 170         .default_advmss         =       ip6_default_advmss,
 171         .mtu                    =       ip6_mtu,
 172         .cow_metrics            =       ipv6_cow_metrics,
 173         .destroy                =       ip6_dst_destroy,
 174         .ifdown                 =       ip6_dst_ifdown,
 175         .negative_advice        =       ip6_negative_advice,
 176         .link_failure           =       ip6_link_failure,
 177         .update_pmtu            =       ip6_rt_update_pmtu,
 178         .redirect               =       rt6_do_redirect,
 179         .local_out              =       __ip6_local_out,
 180         .neigh_lookup           =       ip6_neigh_lookup,
 181 };
 182
 183 static unsigned int ip6_blackhole_mtu(const struct dst_entry *dst)
 184 {
 185         unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
 186
 187         return mtu ? : dst->dev->mtu;
 188 }
 189
 190 static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, struct sock *sk,
 191                                          struct sk_buff *skb, u32 mtu)
 192 {
 193 }
 194
 195 static void ip6_rt_blackhole_redirect(struct dst_entry *dst, struct sock *sk,
 196                                       struct sk_buff *skb)
 197 {
 198 }
 199
 200 static u32 *ip6_rt_blackhole_cow_metrics(struct dst_entry *dst,
 201                                          unsigned long old)
 202 {
 203         return NULL;
 204 }
 205
 206 static struct dst_ops ip6_dst_blackhole_ops = {
 207         .family                 =       AF_INET6,
 208         .protocol               =       cpu_to_be16(ETH_P_IPV6),
 209         .destroy                =       ip6_dst_destroy,
 210         .check                  =       ip6_dst_check,
 211         .mtu                    =       ip6_blackhole_mtu,
 212         .default_advmss         =       ip6_default_advmss,
 213         .update_pmtu            =       ip6_rt_blackhole_update_pmtu,
 214         .redirect               =       ip6_rt_blackhole_redirect,
 215         .cow_metrics            =       ip6_rt_blackhole_cow_metrics,
 216         .neigh_lookup           =       ip6_neigh_lookup,
 217 };
 218
 219 static const u32 ip6_template_metrics[RTAX_MAX] = {
 220         [RTAX_HOPLIMIT - 1] = 0,
 221 };
 222
 223 static const struct rt6_info ip6_null_entry_template = {
 224         .dst = {
 225                 .__refcnt       = ATOMIC_INIT(1),
 226                 .__use          = 1,
 227                 .obsolete       = DST_OBSOLETE_FORCE_CHK,
 228                 .error          = -ENETUNREACH,
 229                 .input          = ip6_pkt_discard,
 230                 .output         = ip6_pkt_discard_out,
 231         },
 232         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
 233         .rt6i_protocol  = RTPROT_KERNEL,
 234         .rt6i_metric    = ~(u32) 0,
 235         .rt6i_ref       = ATOMIC_INIT(1),
 236 };
 237
 238 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
 239
 240 static const struct rt6_info ip6_prohibit_entry_template = {
 241         .dst = {
 242                 .__refcnt       = ATOMIC_INIT(1),
 243                 .__use          = 1,
 244                 .obsolete       = DST_OBSOLETE_FORCE_CHK,
 245                 .error          = -EACCES,
 246                 .input          = ip6_pkt_prohibit,
 247                 .output         = ip6_pkt_prohibit_out,
 248         },
 249         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
 250         .rt6i_protocol  = RTPROT_KERNEL,
 251         .rt6i_metric    = ~(u32) 0,
 252         .rt6i_ref       = ATOMIC_INIT(1),
 253 };
 254
 255 static const struct rt6_info ip6_blk_hole_entry_template = {
 256         .dst = {
 257                 .__refcnt       = ATOMIC_INIT(1),
 258                 .__use          = 1,
 259                 .obsolete       = DST_OBSOLETE_FORCE_CHK,
 260                 .error          = -EINVAL,
 261                 .input          = dst_discard,
 262                 .output         = dst_discard,
 263         },
 264         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
 265         .rt6i_protocol  = RTPROT_KERNEL,
 266         .rt6i_metric    = ~(u32) 0,
 267         .rt6i_ref       = ATOMIC_INIT(1),
 268 };
 269
 270 #endif
 271
 272 /* allocate dst with ip6_dst_ops */
 273 static inline struct rt6_info *ip6_dst_alloc(struct net *net,
 274                                              struct net_device *dev,
 275                                              int flags,
 276                                              struct fib6_table *table)
 277 {
 278         struct rt6_info *rt = dst_alloc(&net->ipv6.ip6_dst_ops, dev,
 279                                         0, DST_OBSOLETE_FORCE_CHK, flags);
 280
 281         if (rt) {
 282                 struct dst_entry *dst = &rt->dst;
 283
 284                 memset(dst + 1, 0, sizeof(*rt) - sizeof(*dst));
 285                 rt6_init_peer(rt, table ? &table->tb6_peers : net->ipv6.peers);
 286                 rt->rt6i_genid = rt_genid_ipv6(net);
 287                 INIT_LIST_HEAD(&rt->rt6i_siblings);
 288         }
 289         return rt;
 290 }
 291
 292 static void ip6_dst_destroy(struct dst_entry *dst)
 293 {
 294         struct rt6_info *rt = (struct rt6_info *)dst;
 295         struct inet6_dev *idev = rt->rt6i_idev;
 296         struct dst_entry *from = dst->from;
 297
 298         if (!(rt->dst.flags & DST_HOST))
 299                 dst_destroy_metrics_generic(dst);
 300
 301         if (idev) {
 302                 rt->rt6i_idev = NULL;
 303                 in6_dev_put(idev);
 304         }
 305
 306         dst->from = NULL;
 307         dst_release(from);
 308
 309         if (rt6_has_peer(rt)) {
 310                 struct inet_peer *peer = rt6_peer_ptr(rt);
 311                 inet_putpeer(peer);
 312         }
 313 }
 314
 315 void rt6_bind_peer(struct rt6_info *rt, int create)
 316 {
 317         struct inet_peer_base *base;
 318         struct inet_peer *peer;
 319
 320         base = inetpeer_base_ptr(rt->_rt6i_peer);
 321         if (!base)
 322                 return;
 323
 324         peer = inet_getpeer_v6(base, &rt->rt6i_dst.addr, create);
 325         if (peer) {
 326                 if (!rt6_set_peer(rt, peer))
 327                         inet_putpeer(peer);
 328         }
 329 }
 330
 331 static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
 332                            int how)
 333 {
 334         struct rt6_info *rt = (struct rt6_info *)dst;
 335         struct inet6_dev *idev = rt->rt6i_idev;
 336         struct net_device *loopback_dev =
 337                 dev_net(dev)->loopback_dev;
 338
 339         if (dev != loopback_dev) {
 340                 if (idev && idev->dev == dev) {
 341                         struct inet6_dev *loopback_idev =
 342                                 in6_dev_get(loopback_dev);
 343                         if (loopback_idev) {
 344                                 rt->rt6i_idev = loopback_idev;
 345                                 in6_dev_put(idev);
 346                         }
 347                 }
 348         }
 349 }
 350
 351 static bool rt6_check_expired(const struct rt6_info *rt)
 352 {
 353         if (rt->rt6i_flags & RTF_EXPIRES) {
 354                 if (time_after(jiffies, rt->dst.expires))
 355                         return true;
 356         } else if (rt->dst.from) {
 357                 return rt6_check_expired((struct rt6_info *) rt->dst.from);
 358         }
 359         return false;
 360 }
 361
 362 static bool rt6_need_strict(const struct in6_addr *daddr)
 363 {
 364         return ipv6_addr_type(daddr) &
 365                 (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
 366 }
 367
 368 /* Multipath route selection:
 369  *   Hash based function using packet header and flowlabel.
 370  * Adapted from fib_info_hashfn()
 371  */
 372 static int rt6_info_hash_nhsfn(unsigned int candidate_count,
 373                                const struct flowi6 *fl6)
 374 {
 375         unsigned int val = fl6->flowi6_proto;
 376
 377         val ^= ipv6_addr_hash(&fl6->daddr);
 378         val ^= ipv6_addr_hash(&fl6->saddr);
 379
 380         /* Work only if this not encapsulated */
 381         switch (fl6->flowi6_proto) {
 382         case IPPROTO_UDP:
 383         case IPPROTO_TCP:
 384         case IPPROTO_SCTP:
 385                 val ^= (__force u16)fl6->fl6_sport;
 386                 val ^= (__force u16)fl6->fl6_dport;
 387                 break;
 388
 389         case IPPROTO_ICMPV6:
 390                 val ^= (__force u16)fl6->fl6_icmp_type;
 391                 val ^= (__force u16)fl6->fl6_icmp_code;
 392                 break;
 393         }
 394         /* RFC6438 recommands to use flowlabel */
 395         val ^= (__force u32)fl6->flowlabel;
 396
 397         /* Perhaps, we need to tune, this function? */
 398         val = val ^ (val >> 7) ^ (val >> 12);
 399         return val % candidate_count;
 400 }
 401
 402 static struct rt6_info *rt6_multipath_select(struct rt6_info *match,
 403                                              struct flowi6 *fl6, int oif,
 404                                              int strict)
 405 {
 406         struct rt6_info *sibling, *next_sibling;
 407         int route_choosen;
 408
 409         route_choosen = rt6_info_hash_nhsfn(match->rt6i_nsiblings + 1, fl6);
 410         /* Don't change the route, if route_choosen == 0
 411          * (siblings does not include ourself)
 412          */
 413         if (route_choosen)
 414                 list_for_each_entry_safe(sibling, next_sibling,
 415                                 &match->rt6i_siblings, rt6i_siblings) {
 416                         route_choosen--;
 417                         if (route_choosen == 0) {
 418                                 if (rt6_score_route(sibling, oif, strict) < 0)
 419                                         break;
 420                                 match = sibling;
 421                                 break;
 422                         }
 423                 }
 424         return match;
 425 }
 426
 427 /*
 428  *      Route lookup. Any table->tb6_lock is implied.
 429  */
 430
 431 static inline struct rt6_info *rt6_device_match(struct net *net,
 432                                                     struct rt6_info *rt,
 433                                                     const struct in6_addr *saddr,
 434                                                     int oif,
 435                                                     int flags)
 436 {
 437         struct rt6_info *local = NULL;
 438         struct rt6_info *sprt;
 439
 440         if (!oif && ipv6_addr_any(saddr))
 441                 goto out;
 442
 443         for (sprt = rt; sprt; sprt = sprt->dst.rt6_next) {
 444                 struct net_device *dev = sprt->dst.dev;
 445
 446                 if (oif) {
 447                         if (dev->ifindex == oif)
 448                                 return sprt;
 449                         if (dev->flags & IFF_LOOPBACK) {
 450                                 if (!sprt->rt6i_idev ||
 451                                     sprt->rt6i_idev->dev->ifindex != oif) {
 452                                         if (flags & RT6_LOOKUP_F_IFACE && oif)
 453                                                 continue;
 454                                         if (local && (!oif ||
 455                                                       local->rt6i_idev->dev->ifindex == oif))
 456                                                 continue;
 457                                 }
 458                                 local = sprt;
 459                         }
 460                 } else {
 461                         if (ipv6_chk_addr(net, saddr, dev,
 462                                           flags & RT6_LOOKUP_F_IFACE))
 463                                 return sprt;
 464                 }
 465         }
 466
 467         if (oif) {
 468                 if (local)
 469                         return local;
 470
 471                 if (flags & RT6_LOOKUP_F_IFACE)
 472                         return net->ipv6.ip6_null_entry;
 473         }
 474 out:
 475         return rt;
 476 }
 477
 478 #ifdef CONFIG_IPV6_ROUTER_PREF
 479 struct __rt6_probe_work {
 480         struct work_struct work;
 481         struct in6_addr target;
 482         struct net_device *dev;
 483 };
 484
 485 static void rt6_probe_deferred(struct work_struct *w)
 486 {
 487         struct in6_addr mcaddr;
 488         struct __rt6_probe_work *work =
 489                 container_of(w, struct __rt6_probe_work, work);
 490
 491         addrconf_addr_solict_mult(&work->target, &mcaddr);
 492         ndisc_send_ns(work->dev, NULL, &work->target, &mcaddr, NULL);
 493         dev_put(work->dev);
 494         kfree(w);
 495 }
 496
 497 static void rt6_probe(struct rt6_info *rt)
 498 {
 499         struct neighbour *neigh;
 500         /*
 501          * Okay, this does not seem to be appropriate
 502          * for now, however, we need to check if it
 503          * is really so; aka Router Reachability Probing.
 504          *
 505          * Router Reachability Probe MUST be rate-limited
 506          * to no more than one per minute.
 507          */
 508         if (!rt || !(rt->rt6i_flags & RTF_GATEWAY))
 509                 return;
 510         rcu_read_lock_bh();
 511         neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
 512         if (neigh) {
 513                 write_lock(&neigh->lock);
 514                 if (neigh->nud_state & NUD_VALID)
 515                         goto out;
 516         }
 517
 518         if (!neigh ||
 519             time_after(jiffies, neigh->updated + rt->rt6i_idev->cnf.rtr_probe_interval)) {
 520                 struct __rt6_probe_work *work;
 521
 522                 work = kmalloc(sizeof(*work), GFP_ATOMIC);
 523
 524                 if (neigh && work)
 525                         __neigh_set_probe_once(neigh);
 526
 527                 if (neigh)
 528                         write_unlock(&neigh->lock);
 529
 530                 if (work) {
 531                         INIT_WORK(&work->work, rt6_probe_deferred);
 532                         work->target = rt->rt6i_gateway;
 533                         dev_hold(rt->dst.dev);
 534                         work->dev = rt->dst.dev;
 535                         schedule_work(&work->work);
 536                 }
 537         } else {
 538 out:
 539                 write_unlock(&neigh->lock);
 540         }
 541         rcu_read_unlock_bh();
 542 }
 543 #else
 544 static inline void rt6_probe(struct rt6_info *rt)
 545 {
 546 }
 547 #endif
 548
 549 /*
 550  * Default Router Selection (RFC 2461 6.3.6)
 551  */
 552 static inline int rt6_check_dev(struct rt6_info *rt, int oif)
 553 {
 554         struct net_device *dev = rt->dst.dev;
 555         if (!oif || dev->ifindex == oif)
 556                 return 2;
 557         if ((dev->flags & IFF_LOOPBACK) &&
 558             rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
 559                 return 1;
 560         return 0;
 561 }
 562
 563 static inline enum rt6_nud_state rt6_check_neigh(struct rt6_info *rt)
 564 {
 565         struct neighbour *neigh;
 566         enum rt6_nud_state ret = RT6_NUD_FAIL_HARD;
 567
 568         if (rt->rt6i_flags & RTF_NONEXTHOP ||
 569             !(rt->rt6i_flags & RTF_GATEWAY))
 570                 return RT6_NUD_SUCCEED;
 571
 572         rcu_read_lock_bh();
 573         neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
 574         if (neigh) {
 575                 read_lock(&neigh->lock);
 576                 if (neigh->nud_state & NUD_VALID)
 577                         ret = RT6_NUD_SUCCEED;
 578 #ifdef CONFIG_IPV6_ROUTER_PREF
 579                 else if (!(neigh->nud_state & NUD_FAILED))
 580                         ret = RT6_NUD_SUCCEED;
 581                 else
 582                         ret = RT6_NUD_FAIL_PROBE;
 583 #endif
 584                 read_unlock(&neigh->lock);
 585         } else {
 586                 ret = IS_ENABLED(CONFIG_IPV6_ROUTER_PREF) ?
 587                       RT6_NUD_SUCCEED : RT6_NUD_FAIL_DO_RR;
 588         }
 589         rcu_read_unlock_bh();
 590
 591         return ret;
 592 }
 593
 594 static int rt6_score_route(struct rt6_info *rt, int oif,
 595                            int strict)
 596 {
 597         int m;
 598
 599         m = rt6_check_dev(rt, oif);
 600         if (!m && (strict & RT6_LOOKUP_F_IFACE))
 601                 return RT6_NUD_FAIL_HARD;
 602 #ifdef CONFIG_IPV6_ROUTER_PREF
 603         m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
 604 #endif
 605         if (strict & RT6_LOOKUP_F_REACHABLE) {
 606                 int n = rt6_check_neigh(rt);
 607                 if (n < 0)
 608                         return n;
 609         }
 610         return m;
 611 }
 612
 613 static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
 614                                    int *mpri, struct rt6_info *match,
 615                                    bool *do_rr)
 616 {
 617         int m;
 618         bool match_do_rr = false;
 619
 620         if (rt6_check_expired(rt))
 621                 goto out;
 622
 623         m = rt6_score_route(rt, oif, strict);
 624         if (m == RT6_NUD_FAIL_DO_RR) {
 625                 match_do_rr = true;
 626                 m = 0; /* lowest valid score */
 627         } else if (m == RT6_NUD_FAIL_HARD) {
 628                 goto out;
 629         }
 630
 631         if (strict & RT6_LOOKUP_F_REACHABLE)
 632                 rt6_probe(rt);
 633
 634         /* note that m can be RT6_NUD_FAIL_PROBE at this point */
 635         if (m > *mpri) {
 636                 *do_rr = match_do_rr;
 637                 *mpri = m;
 638                 match = rt;
 639         }
 640 out:
 641         return match;
 642 }
 643
 644 static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
 645                                      struct rt6_info *rr_head,
 646                                      u32 metric, int oif, int strict,
 647                                      bool *do_rr)
 648 {
 649         struct rt6_info *rt, *match;
 650         int mpri = -1;
 651
 652         match = NULL;
 653         for (rt = rr_head; rt && rt->rt6i_metric == metric;
 654              rt = rt->dst.rt6_next)
 655                 match = find_match(rt, oif, strict, &mpri, match, do_rr);
 656         for (rt = fn->leaf; rt && rt != rr_head && rt->rt6i_metric == metric;
 657              rt = rt->dst.rt6_next)
 658                 match = find_match(rt, oif, strict, &mpri, match, do_rr);
 659
 660         return match;
 661 }
 662
 663 static struct rt6_info *rt6_select(struct fib6_node *fn, int oif, int strict)
 664 {
 665         struct rt6_info *match, *rt0;
 666         struct net *net;
 667         bool do_rr = false;
 668
 669         rt0 = fn->rr_ptr;
 670         if (!rt0)
 671                 fn->rr_ptr = rt0 = fn->leaf;
 672
 673         match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict,
 674                              &do_rr);
 675
 676         if (do_rr) {
 677                 struct rt6_info *next = rt0->dst.rt6_next;
 678
 679                 /* no entries matched; do round-robin */
 680                 if (!next || next->rt6i_metric != rt0->rt6i_metric)
 681                         next = fn->leaf;
 682
 683                 if (next != rt0)
 684                         fn->rr_ptr = next;
 685         }
 686
 687         net = dev_net(rt0->dst.dev);
 688         return match ? match : net->ipv6.ip6_null_entry;
 689 }
 690
 691 #ifdef CONFIG_IPV6_ROUTE_INFO
 692 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
 693                   const struct in6_addr *gwaddr)
 694 {
 695         struct net *net = dev_net(dev);
 696         struct route_info *rinfo = (struct route_info *) opt;
 697         struct in6_addr prefix_buf, *prefix;
 698         unsigned int pref;
 699         unsigned long lifetime;
 700         struct rt6_info *rt;
 701
 702         if (len < sizeof(struct route_info)) {
 703                 return -EINVAL;
 704         }
 705
 706         /* Sanity check for prefix_len and length */
 707         if (rinfo->length > 3) {
 708                 return -EINVAL;
 709         } else if (rinfo->prefix_len > 128) {
 710                 return -EINVAL;
 711         } else if (rinfo->prefix_len > 64) {
 712                 if (rinfo->length < 2) {
 713                         return -EINVAL;
 714                 }
 715         } else if (rinfo->prefix_len > 0) {
 716                 if (rinfo->length < 1) {
 717                         return -EINVAL;
 718                 }
 719         }
 720
 721         pref = rinfo->route_pref;
 722         if (pref == ICMPV6_ROUTER_PREF_INVALID)
 723                 return -EINVAL;
 724
 725         lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
 726
 727         if (rinfo->length == 3)
 728                 prefix = (struct in6_addr *)rinfo->prefix;
 729         else {
 730                 /* this function is safe */
 731                 ipv6_addr_prefix(&prefix_buf,
 732                                  (struct in6_addr *)rinfo->prefix,
 733                                  rinfo->prefix_len);
 734                 prefix = &prefix_buf;
 735         }
 736
 737         if (rinfo->prefix_len == 0)
 738                 rt = rt6_get_dflt_router(gwaddr, dev);
 739         else
 740                 rt = rt6_get_route_info(net, prefix, rinfo->prefix_len,
 741                                         gwaddr, dev->ifindex);
 742
 743         if (rt && !lifetime) {
 744                 ip6_del_rt(rt);
 745                 rt = NULL;
 746         }
 747
 748         if (!rt && lifetime)
 749                 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr, dev->ifindex,
 750                                         pref);
 751         else if (rt)
 752                 rt->rt6i_flags = RTF_ROUTEINFO |
 753                                  (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
 754
 755         if (rt) {
 756                 if (!addrconf_finite_timeout(lifetime))
 757                         rt6_clean_expires(rt);
 758                 else
 759                         rt6_set_expires(rt, jiffies + HZ * lifetime);
 760
 761                 ip6_rt_put(rt);
 762         }
 763         return 0;
 764 }
 765 #endif
 766
 767 #define BACKTRACK(__net, saddr)                 \
 768 do { \
 769         if (rt == __net->ipv6.ip6_null_entry) { \
 770                 struct fib6_node *pn; \
 771                 while (1) { \
 772                         if (fn->fn_flags & RTN_TL_ROOT) \
 773                                 goto out; \
 774                         pn = fn->parent; \
 775                         if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn) \
 776                                 fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr); \
 777                         else \
 778                                 fn = pn; \
 779                         if (fn->fn_flags & RTN_RTINFO) \
 780                                 goto restart; \
 781                 } \
 782         } \
 783 } while (0)
 784
 785 static struct rt6_info *ip6_pol_route_lookup(struct net *net,
 786                                              struct fib6_table *table,
 787                                              struct flowi6 *fl6, int flags)
 788 {
 789         struct fib6_node *fn;
 790         struct rt6_info *rt;
 791
 792         read_lock_bh(&table->tb6_lock);
 793         fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
 794 restart:
 795         rt = fn->leaf;
 796         rt = rt6_device_match(net, rt, &fl6->saddr, fl6->flowi6_oif, flags);
 797         if (rt->rt6i_nsiblings && fl6->flowi6_oif == 0)
 798                 rt = rt6_multipath_select(rt, fl6, fl6->flowi6_oif, flags);
 799         BACKTRACK(net, &fl6->saddr);
 800 out:
 801         dst_use(&rt->dst, jiffies);
 802         read_unlock_bh(&table->tb6_lock);
 803         return rt;
 804
 805 }
 806
 807 struct dst_entry * ip6_route_lookup(struct net *net, struct flowi6 *fl6,
 808                                     int flags)
 809 {
 810         return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_lookup);
 811 }
 812 EXPORT_SYMBOL_GPL(ip6_route_lookup);
 813
 814 struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
 815                             const struct in6_addr *saddr, int oif, int strict)
 816 {
 817         struct flowi6 fl6 = {
 818                 .flowi6_oif = oif,
 819                 .daddr = *daddr,
 820         };
 821         struct dst_entry *dst;
 822         int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
 823
 824         if (saddr) {
 825                 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
 826                 flags |= RT6_LOOKUP_F_HAS_SADDR;
 827         }
 828
 829         dst = fib6_rule_lookup(net, &fl6, flags, ip6_pol_route_lookup);
 830         if (dst->error == 0)
 831                 return (struct rt6_info *) dst;
 832
 833         dst_release(dst);
 834
 835         return NULL;
 836 }
 837
 838 EXPORT_SYMBOL(rt6_lookup);
 839
 840 /* ip6_ins_rt is called with FREE table->tb6_lock.
 841    It takes new route entry, the addition fails by any reason the
 842    route is freed. In any case, if caller does not hold it, it may
 843    be destroyed.
 844  */
 845
 846 static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info)
 847 {
 848         int err;
 849         struct fib6_table *table;
 850
 851         table = rt->rt6i_table;
 852         write_lock_bh(&table->tb6_lock);
 853         err = fib6_add(&table->tb6_root, rt, info);
 854         write_unlock_bh(&table->tb6_lock);
 855
 856         return err;
 857 }
 858
 859 int ip6_ins_rt(struct rt6_info *rt)
 860 {
 861         struct nl_info info = {
 862                 .nl_net = dev_net(rt->dst.dev),
 863         };
 864         return __ip6_ins_rt(rt, &info);
 865 }
 866
 867 static struct rt6_info *rt6_alloc_cow(struct rt6_info *ort,
 868                                       const struct in6_addr *daddr,
 869                                       const struct in6_addr *saddr)
 870 {
 871         struct rt6_info *rt;
 872
 873         /*
 874          *      Clone the route.
 875          */
 876
 877         rt = ip6_rt_copy(ort, daddr);
 878
 879         if (rt) {
 880                 if (ort->rt6i_dst.plen != 128 &&
 881                     ipv6_addr_equal(&ort->rt6i_dst.addr, daddr))
 882                         rt->rt6i_flags |= RTF_ANYCAST;
 883
 884                 rt->rt6i_flags |= RTF_CACHE;
 885
 886 #ifdef CONFIG_IPV6_SUBTREES
 887                 if (rt->rt6i_src.plen && saddr) {
 888                         rt->rt6i_src.addr = *saddr;
 889                         rt->rt6i_src.plen = 128;
 890                 }
 891 #endif
 892         }
 893
 894         return rt;
 895 }
 896
 897 static struct rt6_info *rt6_alloc_clone(struct rt6_info *ort,
 898                                         const struct in6_addr *daddr)
 899 {
 900         struct rt6_info *rt = ip6_rt_copy(ort, daddr);
 901
 902         if (rt)
 903                 rt->rt6i_flags |= RTF_CACHE;
 904         return rt;
 905 }
 906
 907 static struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table, int oif,
 908                                       struct flowi6 *fl6, int flags)
 909 {
 910         struct fib6_node *fn;
 911         struct rt6_info *rt, *nrt;
 912         int strict = 0;
 913         int attempts = 3;
 914         int err;
 915         int reachable = net->ipv6.devconf_all->forwarding ? 0 : RT6_LOOKUP_F_REACHABLE;
 916
 917         strict |= flags & RT6_LOOKUP_F_IFACE;
 918
 919 relookup:
 920         read_lock_bh(&table->tb6_lock);
 921
 922 restart_2:
 923         fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
 924
 925 restart:
 926         rt = rt6_select(fn, oif, strict | reachable);
 927         if (rt->rt6i_nsiblings)
 928                 rt = rt6_multipath_select(rt, fl6, oif, strict | reachable);
 929         BACKTRACK(net, &fl6->saddr);
 930         if (rt == net->ipv6.ip6_null_entry ||
 931             rt->rt6i_flags & RTF_CACHE)
 932                 goto out;
 933
 934         dst_hold(&rt->dst);
 935         read_unlock_bh(&table->tb6_lock);
 936
 937         if (!(rt->rt6i_flags & (RTF_NONEXTHOP | RTF_GATEWAY)))
 938                 nrt = rt6_alloc_cow(rt, &fl6->daddr, &fl6->saddr);
 939         else if (!(rt->dst.flags & DST_HOST))
 940                 nrt = rt6_alloc_clone(rt, &fl6->daddr);
 941         else
 942                 goto out2;
 943
 944         ip6_rt_put(rt);
 945         rt = nrt ? : net->ipv6.ip6_null_entry;
 946
 947         dst_hold(&rt->dst);
 948         if (nrt) {
 949                 err = ip6_ins_rt(nrt);
 950                 if (!err)
 951                         goto out2;
 952         }
 953
 954         if (--attempts <= 0)
 955                 goto out2;
 956
 957         /*
 958          * Race condition! In the gap, when table->tb6_lock was
 959          * released someone could insert this route.  Relookup.
 960          */
 961         ip6_rt_put(rt);
 962         goto relookup;
 963
 964 out:
 965         if (reachable) {
 966                 reachable = 0;
 967                 goto restart_2;
 968         }
 969         dst_hold(&rt->dst);
 970         read_unlock_bh(&table->tb6_lock);
 971 out2:
 972         rt->dst.lastuse = jiffies;
 973         rt->dst.__use++;
 974
 975         return rt;
 976 }
 977
 978 static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
 979                                             struct flowi6 *fl6, int flags)
 980 {
 981         return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, flags);
 982 }
 983
 984 static struct dst_entry *ip6_route_input_lookup(struct net *net,
 985                                                 struct net_device *dev,
 986                                                 struct flowi6 *fl6, int flags)
 987 {
 988         if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
 989                 flags |= RT6_LOOKUP_F_IFACE;
 990
 991         return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_input);
 992 }
 993
 994 void ip6_route_input(struct sk_buff *skb)
 995 {
 996         const struct ipv6hdr *iph = ipv6_hdr(skb);
 997         struct net *net = dev_net(skb->dev);
 998         int flags = RT6_LOOKUP_F_HAS_SADDR;
 999         struct flowi6 fl6 = {
1000                 .flowi6_iif = skb->dev->ifindex,
1001                 .daddr = iph->daddr,
1002                 .saddr = iph->saddr,
1003                 .flowlabel = ip6_flowinfo(iph),
1004                 .flowi6_mark = skb->mark,
1005                 .flowi6_proto = iph->nexthdr,
1006         };
1007
1008         skb_dst_set(skb, ip6_route_input_lookup(net, skb->dev, &fl6, flags));
1009 }
1010
1011 static struct rt6_info *ip6_pol_route_output(struct net *net, struct fib6_table *table,
1012                                              struct flowi6 *fl6, int flags)
1013 {
1014         return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, flags);
1015 }
1016
1017 struct dst_entry * ip6_route_output(struct net *net, const struct sock *sk,
1018                                     struct flowi6 *fl6)
1019 {
1020         int flags = 0;
1021
1022         fl6->flowi6_iif = LOOPBACK_IFINDEX;
1023
1024         if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr))
1025                 flags |= RT6_LOOKUP_F_IFACE;
1026
1027         if (!ipv6_addr_any(&fl6->saddr))
1028                 flags |= RT6_LOOKUP_F_HAS_SADDR;
1029         else if (sk)
1030                 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
1031
1032         return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_output);
1033 }
1034
1035 EXPORT_SYMBOL(ip6_route_output);
1036
1037 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
1038 {
1039         struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
1040         struct dst_entry *new = NULL;
1041
1042         rt = dst_alloc(&ip6_dst_blackhole_ops, ort->dst.dev, 1, DST_OBSOLETE_NONE, 0);
1043         if (rt) {
1044                 new = &rt->dst;
1045
1046                 memset(new + 1, 0, sizeof(*rt) - sizeof(*new));
1047                 rt6_init_peer(rt, net->ipv6.peers);
1048
1049                 new->__use = 1;
1050                 new->input = dst_discard;
1051                 new->output = dst_discard;
1052
1053                 if (dst_metrics_read_only(&ort->dst))
1054                         new->_metrics = ort->dst._metrics;
1055                 else
1056                         dst_copy_metrics(new, &ort->dst);
1057                 rt->rt6i_idev = ort->rt6i_idev;
1058                 if (rt->rt6i_idev)
1059                         in6_dev_hold(rt->rt6i_idev);
1060
1061                 rt->rt6i_gateway = ort->rt6i_gateway;
1062                 rt->rt6i_flags = ort->rt6i_flags;
1063                 rt->rt6i_metric = 0;
1064
1065                 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1066 #ifdef CONFIG_IPV6_SUBTREES
1067                 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1068 #endif
1069
1070                 dst_free(new);
1071         }
1072
1073         dst_release(dst_orig);
1074         return new ? new : ERR_PTR(-ENOMEM);
1075 }
1076
1077 /*
1078  *      Destination cache support functions
1079  */
1080
1081 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
1082 {
1083         struct rt6_info *rt;
1084
1085         rt = (struct rt6_info *) dst;
1086
1087         /* All IPV6 dsts are created with ->obsolete set to the value
1088          * DST_OBSOLETE_FORCE_CHK which forces validation calls down
1089          * into this function always.
1090          */
1091         if (rt->rt6i_genid != rt_genid_ipv6(dev_net(rt->dst.dev)))
1092                 return NULL;
1093
1094         if (!rt->rt6i_node || (rt->rt6i_node->fn_sernum != cookie))
1095                 return NULL;
1096
1097         if (rt6_check_expired(rt))
1098                 return NULL;
1099
1100         return dst;
1101 }
1102
1103 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
1104 {
1105         struct rt6_info *rt = (struct rt6_info *) dst;
1106
1107         if (rt) {
1108                 if (rt->rt6i_flags & RTF_CACHE) {
1109                         if (rt6_check_expired(rt)) {
1110                                 ip6_del_rt(rt);
1111                                 dst = NULL;
1112                         }
1113                 } else {
1114                         dst_release(dst);
1115                         dst = NULL;
1116                 }
1117         }
1118         return dst;
1119 }
1120
1121 static void ip6_link_failure(struct sk_buff *skb)
1122 {
1123         struct rt6_info *rt;
1124
1125         icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
1126
1127         rt = (struct rt6_info *) skb_dst(skb);
1128         if (rt) {
1129                 if (rt->rt6i_flags & RTF_CACHE) {
1130                         dst_hold(&rt->dst);
1131                         if (ip6_del_rt(rt))
1132                                 dst_free(&rt->dst);
1133                 } else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT)) {
1134                         rt->rt6i_node->fn_sernum = -1;
1135                 }
1136         }
1137 }
1138
1139 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
1140                                struct sk_buff *skb, u32 mtu)
1141 {
1142         struct rt6_info *rt6 = (struct rt6_info*)dst;
1143
1144         dst_confirm(dst);
1145         if (mtu < dst_mtu(dst) && rt6->rt6i_dst.plen == 128) {
1146                 struct net *net = dev_net(dst->dev);
1147
1148                 rt6->rt6i_flags |= RTF_MODIFIED;
1149                 if (mtu < IPV6_MIN_MTU) {
1150                         u32 features = dst_metric(dst, RTAX_FEATURES);
1151                         mtu = IPV6_MIN_MTU;
1152                         features |= RTAX_FEATURE_ALLFRAG;
1153                         dst_metric_set(dst, RTAX_FEATURES, features);
1154                 }
1155                 dst_metric_set(dst, RTAX_MTU, mtu);
1156                 rt6_update_expires(rt6, net->ipv6.sysctl.ip6_rt_mtu_expires);
1157         }
1158 }
1159
1160 void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
1161                      int oif, u32 mark)
1162 {
1163         const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1164         struct dst_entry *dst;
1165         struct flowi6 fl6;
1166
1167         memset(&fl6, 0, sizeof(fl6));
1168         fl6.flowi6_oif = oif;
1169         fl6.flowi6_mark = mark;
1170         fl6.daddr = iph->daddr;
1171         fl6.saddr = iph->saddr;
1172         fl6.flowlabel = ip6_flowinfo(iph);
1173
1174         dst = ip6_route_output(net, NULL, &fl6);
1175         if (!dst->error)
1176                 ip6_rt_update_pmtu(dst, NULL, skb, ntohl(mtu));
1177         dst_release(dst);
1178 }
1179 EXPORT_SYMBOL_GPL(ip6_update_pmtu);
1180
1181 void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
1182 {
1183         ip6_update_pmtu(skb, sock_net(sk), mtu,
1184                         sk->sk_bound_dev_if, sk->sk_mark);
1185 }
1186 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
1187
1188 /* Handle redirects */
1189 struct ip6rd_flowi {
1190         struct flowi6 fl6;
1191         struct in6_addr gateway;
1192 };
1193
1194 static struct rt6_info *__ip6_route_redirect(struct net *net,
1195                                              struct fib6_table *table,
1196                                              struct flowi6 *fl6,
1197                                              int flags)
1198 {
1199         struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
1200         struct rt6_info *rt;
1201         struct fib6_node *fn;
1202
1203         /* Get the "current" route for this destination and
1204          * check if the redirect has come from approriate router.
1205          *
1206          * RFC 4861 specifies that redirects should only be
1207          * accepted if they come from the nexthop to the target.
1208          * Due to the way the routes are chosen, this notion
1209          * is a bit fuzzy and one might need to check all possible
1210          * routes.
1211          */
1212
1213         read_lock_bh(&table->tb6_lock);
1214         fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1215 restart:
1216         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1217                 if (rt6_check_expired(rt))
1218                         continue;
1219                 if (rt->dst.error)
1220                         break;
1221                 if (!(rt->rt6i_flags & RTF_GATEWAY))
1222                         continue;
1223                 if (fl6->flowi6_oif != rt->dst.dev->ifindex)
1224                         continue;
1225                 if (!ipv6_addr_equal(&rdfl->gateway, &rt->rt6i_gateway))
1226                         continue;
1227                 break;
1228         }
1229
1230         if (!rt)
1231                 rt = net->ipv6.ip6_null_entry;
1232         else if (rt->dst.error) {
1233                 rt = net->ipv6.ip6_null_entry;
1234                 goto out;
1235         }
1236         BACKTRACK(net, &fl6->saddr);
1237 out:
1238         dst_hold(&rt->dst);
1239
1240         read_unlock_bh(&table->tb6_lock);
1241
1242         return rt;
1243 };
1244
1245 static struct dst_entry *ip6_route_redirect(struct net *net,
1246                                         const struct flowi6 *fl6,
1247                                         const struct in6_addr *gateway)
1248 {
1249         int flags = RT6_LOOKUP_F_HAS_SADDR;
1250         struct ip6rd_flowi rdfl;
1251
1252         rdfl.fl6 = *fl6;
1253         rdfl.gateway = *gateway;
1254
1255         return fib6_rule_lookup(net, &rdfl.fl6,
1256                                 flags, __ip6_route_redirect);
1257 }
1258
1259 void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark)
1260 {
1261         const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1262         struct dst_entry *dst;
1263         struct flowi6 fl6;
1264
1265         memset(&fl6, 0, sizeof(fl6));
1266         fl6.flowi6_oif = oif;
1267         fl6.flowi6_mark = mark;
1268         fl6.daddr = iph->daddr;
1269         fl6.saddr = iph->saddr;
1270         fl6.flowlabel = ip6_flowinfo(iph);
1271
1272         dst = ip6_route_redirect(net, &fl6, &ipv6_hdr(skb)->saddr);
1273         rt6_do_redirect(dst, NULL, skb);
1274         dst_release(dst);
1275 }
1276 EXPORT_SYMBOL_GPL(ip6_redirect);
1277
1278 void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif,
1279                             u32 mark)
1280 {
1281         const struct ipv6hdr *iph = ipv6_hdr(skb);
1282         const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
1283         struct dst_entry *dst;
1284         struct flowi6 fl6;
1285
1286         memset(&fl6, 0, sizeof(fl6));
1287         fl6.flowi6_oif = oif;
1288         fl6.flowi6_mark = mark;
1289         fl6.daddr = msg->dest;
1290         fl6.saddr = iph->daddr;
1291
1292         dst = ip6_route_redirect(net, &fl6, &iph->saddr);
1293         rt6_do_redirect(dst, NULL, skb);
1294         dst_release(dst);
1295 }
1296
1297 void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
1298 {
1299         ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark);
1300 }
1301 EXPORT_SYMBOL_GPL(ip6_sk_redirect);
1302
1303 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1304 {
1305         struct net_device *dev = dst->dev;
1306         unsigned int mtu = dst_mtu(dst);
1307         struct net *net = dev_net(dev);
1308
1309         mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
1310
1311         if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
1312                 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1313
1314         /*
1315          * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1316          * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1317          * IPV6_MAXPLEN is also valid and means: "any MSS,
1318          * rely only on pmtu discovery"
1319          */
1320         if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
1321                 mtu = IPV6_MAXPLEN;
1322         return mtu;
1323 }
1324
1325 static unsigned int ip6_mtu(const struct dst_entry *dst)
1326 {
1327         struct inet6_dev *idev;
1328         unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
1329
1330         if (mtu)
1331                 return mtu;
1332
1333         mtu = IPV6_MIN_MTU;
1334
1335         rcu_read_lock();
1336         idev = __in6_dev_get(dst->dev);
1337         if (idev)
1338                 mtu = idev->cnf.mtu6;
1339         rcu_read_unlock();
1340
1341         return mtu;
1342 }
1343
1344 static struct dst_entry *icmp6_dst_gc_list;
1345 static DEFINE_SPINLOCK(icmp6_dst_lock);
1346
1347 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
1348                                   struct flowi6 *fl6)
1349 {
1350         struct dst_entry *dst;
1351         struct rt6_info *rt;
1352         struct inet6_dev *idev = in6_dev_get(dev);
1353         struct net *net = dev_net(dev);
1354
1355         if (unlikely(!idev))
1356                 return ERR_PTR(-ENODEV);
1357
1358         rt = ip6_dst_alloc(net, dev, 0, NULL);
1359         if (unlikely(!rt)) {
1360                 in6_dev_put(idev);
1361                 dst = ERR_PTR(-ENOMEM);
1362                 goto out;
1363         }
1364
1365         rt->dst.flags |= DST_HOST;
1366         rt->dst.output  = ip6_output;
1367         atomic_set(&rt->dst.__refcnt, 1);
1368         rt->rt6i_gateway  = fl6->daddr;
1369         rt->rt6i_dst.addr = fl6->daddr;
1370         rt->rt6i_dst.plen = 128;
1371         rt->rt6i_idev     = idev;
1372         dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
1373
1374         spin_lock_bh(&icmp6_dst_lock);
1375         rt->dst.next = icmp6_dst_gc_list;
1376         icmp6_dst_gc_list = &rt->dst;
1377         spin_unlock_bh(&icmp6_dst_lock);
1378
1379         fib6_force_start_gc(net);
1380
1381         dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
1382
1383 out:
1384         return dst;
1385 }
1386
1387 int icmp6_dst_gc(void)
1388 {
1389         struct dst_entry *dst, **pprev;
1390         int more = 0;
1391
1392         spin_lock_bh(&icmp6_dst_lock);
1393         pprev = &icmp6_dst_gc_list;
1394
1395         while ((dst = *pprev) != NULL) {
1396                 if (!atomic_read(&dst->__refcnt)) {
1397                         *pprev = dst->next;
1398                         dst_free(dst);
1399                 } else {
1400                         pprev = &dst->next;
1401                         ++more;
1402                 }
1403         }
1404
1405         spin_unlock_bh(&icmp6_dst_lock);
1406
1407         return more;
1408 }
1409
1410 static void icmp6_clean_all(int (*func)(struct rt6_info *rt, void *arg),
1411                             void *arg)
1412 {
1413         struct dst_entry *dst, **pprev;
1414
1415         spin_lock_bh(&icmp6_dst_lock);
1416         pprev = &icmp6_dst_gc_list;
1417         while ((dst = *pprev) != NULL) {
1418                 struct rt6_info *rt = (struct rt6_info *) dst;
1419                 if (func(rt, arg)) {
1420                         *pprev = dst->next;
1421                         dst_free(dst);
1422                 } else {
1423                         pprev = &dst->next;
1424                 }
1425         }
1426         spin_unlock_bh(&icmp6_dst_lock);
1427 }
1428
1429 static int ip6_dst_gc(struct dst_ops *ops)
1430 {
1431         struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
1432         int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1433         int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1434         int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1435         int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1436         unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1437         int entries;
1438
1439         entries = dst_entries_get_fast(ops);
1440         if (time_after(rt_last_gc + rt_min_interval, jiffies) &&
1441             entries <= rt_max_size)
1442                 goto out;
1443
1444         net->ipv6.ip6_rt_gc_expire++;
1445         fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net, entries > rt_max_size);
1446         entries = dst_entries_get_slow(ops);
1447         if (entries < ops->gc_thresh)
1448                 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1449 out:
1450         net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1451         return entries > rt_max_size;
1452 }
1453
1454 /*
1455  *
1456  */
1457
1458 int ip6_route_add(struct fib6_config *cfg)
1459 {
1460         int err;
1461         struct net *net = cfg->fc_nlinfo.nl_net;
1462         struct rt6_info *rt = NULL;
1463         struct net_device *dev = NULL;
1464         struct inet6_dev *idev = NULL;
1465         struct fib6_table *table;
1466         int addr_type;
1467
1468         if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1469                 return -EINVAL;
1470 #ifndef CONFIG_IPV6_SUBTREES
1471         if (cfg->fc_src_len)
1472                 return -EINVAL;
1473 #endif
1474         if (cfg->fc_ifindex) {
1475                 err = -ENODEV;
1476                 dev = dev_get_by_index(net, cfg->fc_ifindex);
1477                 if (!dev)
1478                         goto out;
1479                 idev = in6_dev_get(dev);
1480                 if (!idev)
1481                         goto out;
1482         }
1483
1484         if (cfg->fc_metric == 0)
1485                 cfg->fc_metric = IP6_RT_PRIO_USER;
1486
1487         err = -ENOBUFS;
1488         if (cfg->fc_nlinfo.nlh &&
1489             !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
1490                 table = fib6_get_table(net, cfg->fc_table);
1491                 if (!table) {
1492                         pr_warn("NLM_F_CREATE should be specified when creating new route\n");
1493                         table = fib6_new_table(net, cfg->fc_table);
1494                 }
1495         } else {
1496                 table = fib6_new_table(net, cfg->fc_table);
1497         }
1498
1499         if (!table)
1500                 goto out;
1501
1502         rt = ip6_dst_alloc(net, NULL, DST_NOCOUNT, table);
1503
1504         if (!rt) {
1505                 err = -ENOMEM;
1506                 goto out;
1507         }
1508
1509         if (cfg->fc_flags & RTF_EXPIRES)
1510                 rt6_set_expires(rt, jiffies +
1511                                 clock_t_to_jiffies(cfg->fc_expires));
1512         else
1513                 rt6_clean_expires(rt);
1514
1515         if (cfg->fc_protocol == RTPROT_UNSPEC)
1516                 cfg->fc_protocol = RTPROT_BOOT;
1517         rt->rt6i_protocol = cfg->fc_protocol;
1518
1519         addr_type = ipv6_addr_type(&cfg->fc_dst);
1520
1521         if (addr_type & IPV6_ADDR_MULTICAST)
1522                 rt->dst.input = ip6_mc_input;
1523         else if (cfg->fc_flags & RTF_LOCAL)
1524                 rt->dst.input = ip6_input;
1525         else
1526                 rt->dst.input = ip6_forward;
1527
1528         rt->dst.output = ip6_output;
1529
1530         ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1531         rt->rt6i_dst.plen = cfg->fc_dst_len;
1532         if (rt->rt6i_dst.plen == 128)
1533                rt->dst.flags |= DST_HOST;
1534
1535         if (!(rt->dst.flags & DST_HOST) && cfg->fc_mx) {
1536                 u32 *metrics = kzalloc(sizeof(u32) * RTAX_MAX, GFP_KERNEL);
1537                 if (!metrics) {
1538                         err = -ENOMEM;
1539                         goto out;
1540                 }
1541                 dst_init_metrics(&rt->dst, metrics, 0);
1542         }
1543 #ifdef CONFIG_IPV6_SUBTREES
1544         ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1545         rt->rt6i_src.plen = cfg->fc_src_len;
1546 #endif
1547
1548         rt->rt6i_metric = cfg->fc_metric;
1549
1550         /* We cannot add true routes via loopback here,
1551            they would result in kernel looping; promote them to reject routes
1552          */
1553         if ((cfg->fc_flags & RTF_REJECT) ||
1554             (dev && (dev->flags & IFF_LOOPBACK) &&
1555              !(addr_type & IPV6_ADDR_LOOPBACK) &&
1556              !(cfg->fc_flags & RTF_LOCAL))) {
1557                 /* hold loopback dev/idev if we haven't done so. */
1558                 if (dev != net->loopback_dev) {
1559                         if (dev) {
1560                                 dev_put(dev);
1561                                 in6_dev_put(idev);
1562                         }
1563                         dev = net->loopback_dev;
1564                         dev_hold(dev);
1565                         idev = in6_dev_get(dev);
1566                         if (!idev) {
1567                                 err = -ENODEV;
1568                                 goto out;
1569                         }
1570                 }
1571                 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1572                 switch (cfg->fc_type) {
1573                 case RTN_BLACKHOLE:
1574                         rt->dst.error = -EINVAL;
1575                         rt->dst.output = dst_discard;
1576                         rt->dst.input = dst_discard;
1577                         break;
1578                 case RTN_PROHIBIT:
1579                         rt->dst.error = -EACCES;
1580                         rt->dst.output = ip6_pkt_prohibit_out;
1581                         rt->dst.input = ip6_pkt_prohibit;
1582                         break;
1583                 case RTN_THROW:
1584                 default:
1585                         rt->dst.error = (cfg->fc_type == RTN_THROW) ? -EAGAIN
1586                                         : -ENETUNREACH;
1587                         rt->dst.output = ip6_pkt_discard_out;
1588                         rt->dst.input = ip6_pkt_discard;
1589                         break;
1590                 }
1591                 goto install_route;
1592         }
1593
1594         if (cfg->fc_flags & RTF_GATEWAY) {
1595                 const struct in6_addr *gw_addr;
1596                 int gwa_type;
1597
1598                 gw_addr = &cfg->fc_gateway;
1599                 rt->rt6i_gateway = *gw_addr;
1600                 gwa_type = ipv6_addr_type(gw_addr);
1601
1602                 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1603                         struct rt6_info *grt;
1604
1605                         /* IPv6 strictly inhibits using not link-local
1606                            addresses as nexthop address.
1607                            Otherwise, router will not able to send redirects.
1608                            It is very good, but in some (rare!) circumstances
1609                            (SIT, PtP, NBMA NOARP links) it is handy to allow
1610                            some exceptions. --ANK
1611                          */
1612                         err = -EINVAL;
1613                         if (!(gwa_type & IPV6_ADDR_UNICAST))
1614                                 goto out;
1615
1616                         grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, 1);
1617
1618                         err = -EHOSTUNREACH;
1619                         if (!grt)
1620                                 goto out;
1621                         if (dev) {
1622                                 if (dev != grt->dst.dev) {
1623                                         ip6_rt_put(grt);
1624                                         goto out;
1625                                 }
1626                         } else {
1627                                 dev = grt->dst.dev;
1628                                 idev = grt->rt6i_idev;
1629                                 dev_hold(dev);
1630                                 in6_dev_hold(grt->rt6i_idev);
1631                         }
1632                         if (!(grt->rt6i_flags & RTF_GATEWAY))
1633                                 err = 0;
1634                         ip6_rt_put(grt);
1635
1636                         if (err)
1637                                 goto out;
1638                 }
1639                 err = -EINVAL;
1640                 if (!dev || (dev->flags & IFF_LOOPBACK))
1641                         goto out;
1642         }
1643
1644         err = -ENODEV;
1645         if (!dev)
1646                 goto out;
1647
1648         if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
1649                 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
1650                         err = -EINVAL;
1651                         goto out;
1652                 }
1653                 rt->rt6i_prefsrc.addr = cfg->fc_prefsrc;
1654                 rt->rt6i_prefsrc.plen = 128;
1655         } else
1656                 rt->rt6i_prefsrc.plen = 0;
1657
1658         rt->rt6i_flags = cfg->fc_flags;
1659
1660 install_route:
1661         if (cfg->fc_mx) {
1662                 struct nlattr *nla;
1663                 int remaining;
1664
1665                 nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1666                         int type = nla_type(nla);
1667
1668                         if (type) {
1669                                 if (type > RTAX_MAX) {
1670                                         err = -EINVAL;
1671                                         goto out;
1672                                 }
1673
1674                                 dst_metric_set(&rt->dst, type, nla_get_u32(nla));
1675                         }
1676                 }
1677         }
1678
1679         rt->dst.dev = dev;
1680         rt->rt6i_idev = idev;
1681         rt->rt6i_table = table;
1682
1683         cfg->fc_nlinfo.nl_net = dev_net(dev);
1684
1685         return __ip6_ins_rt(rt, &cfg->fc_nlinfo);
1686
1687 out:
1688         if (dev)
1689                 dev_put(dev);
1690         if (idev)
1691                 in6_dev_put(idev);
1692         if (rt)
1693                 dst_free(&rt->dst);
1694         return err;
1695 }
1696
1697 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
1698 {
1699         int err;
1700         struct fib6_table *table;
1701         struct net *net = dev_net(rt->dst.dev);
1702
1703         if (rt == net->ipv6.ip6_null_entry) {
1704                 err = -ENOENT;
1705                 goto out;
1706         }
1707
1708         table = rt->rt6i_table;
1709         write_lock_bh(&table->tb6_lock);
1710         err = fib6_del(rt, info);
1711         write_unlock_bh(&table->tb6_lock);
1712
1713 out:
1714         ip6_rt_put(rt);
1715         return err;
1716 }
1717
1718 int ip6_del_rt(struct rt6_info *rt)
1719 {
1720         struct nl_info info = {
1721                 .nl_net = dev_net(rt->dst.dev),
1722         };
1723         return __ip6_del_rt(rt, &info);
1724 }
1725
1726 static int ip6_route_del(struct fib6_config *cfg)
1727 {
1728         struct fib6_table *table;
1729         struct fib6_node *fn;
1730         struct rt6_info *rt;
1731         int err = -ESRCH;
1732
1733         table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
1734         if (!table)
1735                 return err;
1736
1737         read_lock_bh(&table->tb6_lock);
1738
1739         fn = fib6_locate(&table->tb6_root,
1740                          &cfg->fc_dst, cfg->fc_dst_len,
1741                          &cfg->fc_src, cfg->fc_src_len);
1742
1743         if (fn) {
1744                 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1745                         if (cfg->fc_ifindex &&
1746                             (!rt->dst.dev ||
1747                              rt->dst.dev->ifindex != cfg->fc_ifindex))
1748                                 continue;
1749                         if (cfg->fc_flags & RTF_GATEWAY &&
1750                             !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
1751                                 continue;
1752                         if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
1753                                 continue;
1754                         dst_hold(&rt->dst);
1755                         read_unlock_bh(&table->tb6_lock);
1756
1757                         return __ip6_del_rt(rt, &cfg->fc_nlinfo);
1758                 }
1759         }
1760         read_unlock_bh(&table->tb6_lock);
1761
1762         return err;
1763 }
1764
1765 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
1766 {
1767         struct net *net = dev_net(skb->dev);
1768         struct netevent_redirect netevent;
1769         struct rt6_info *rt, *nrt = NULL;
1770         struct ndisc_options ndopts;
1771         struct inet6_dev *in6_dev;
1772         struct neighbour *neigh;
1773         struct rd_msg *msg;
1774         int optlen, on_link;
1775         u8 *lladdr;
1776
1777         optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
1778         optlen -= sizeof(*msg);
1779
1780         if (optlen < 0) {
1781                 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
1782                 return;
1783         }
1784
1785         msg = (struct rd_msg *)icmp6_hdr(skb);
1786
1787         if (ipv6_addr_is_multicast(&msg->dest)) {
1788                 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
1789                 return;
1790         }
1791
1792         on_link = 0;
1793         if (ipv6_addr_equal(&msg->dest, &msg->target)) {
1794                 on_link = 1;
1795         } else if (ipv6_addr_type(&msg->target) !=
1796                    (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
1797                 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
1798                 return;
1799         }
1800
1801         in6_dev = __in6_dev_get(skb->dev);
1802         if (!in6_dev)
1803                 return;
1804         if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
1805                 return;
1806
1807         /* RFC2461 8.1:
1808          *      The IP source address of the Redirect MUST be the same as the current
1809          *      first-hop router for the specified ICMP Destination Address.
1810          */
1811
1812         if (!ndisc_parse_options(msg->opt, optlen, &ndopts)) {
1813                 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
1814                 return;
1815         }
1816
1817         lladdr = NULL;
1818         if (ndopts.nd_opts_tgt_lladdr) {
1819                 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
1820                                              skb->dev);
1821                 if (!lladdr) {
1822                         net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
1823                         return;
1824                 }
1825         }
1826
1827         rt = (struct rt6_info *) dst;
1828         if (rt == net->ipv6.ip6_null_entry) {
1829                 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
1830                 return;
1831         }
1832
1833         /* Redirect received -> path was valid.
1834          * Look, redirects are sent only in response to data packets,
1835          * so that this nexthop apparently is reachable. --ANK
1836          */
1837         dst_confirm(&rt->dst);
1838
1839         neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
1840         if (!neigh)
1841                 return;
1842
1843         /*
1844          *      We have finally decided to accept it.
1845          */
1846
1847         neigh_update(neigh, lladdr, NUD_STALE,
1848                      NEIGH_UPDATE_F_WEAK_OVERRIDE|
1849                      NEIGH_UPDATE_F_OVERRIDE|
1850                      (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1851                                      NEIGH_UPDATE_F_ISROUTER))
1852                      );
1853
1854         nrt = ip6_rt_copy(rt, &msg->dest);
1855         if (!nrt)
1856                 goto out;
1857
1858         nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
1859         if (on_link)
1860                 nrt->rt6i_flags &= ~RTF_GATEWAY;
1861
1862         nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
1863
1864         if (ip6_ins_rt(nrt))
1865                 goto out;
1866
1867         netevent.old = &rt->dst;
1868         netevent.new = &nrt->dst;
1869         netevent.daddr = &msg->dest;
1870         netevent.neigh = neigh;
1871         call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
1872
1873         if (rt->rt6i_flags & RTF_CACHE) {
1874                 rt = (struct rt6_info *) dst_clone(&rt->dst);
1875                 ip6_del_rt(rt);
1876         }
1877
1878 out:
1879         neigh_release(neigh);
1880 }
1881
1882 /*
1883  *      Misc support functions
1884  */
1885
1886 static struct rt6_info *ip6_rt_copy(struct rt6_info *ort,
1887                                     const struct in6_addr *dest)
1888 {
1889         struct net *net = dev_net(ort->dst.dev);
1890         struct rt6_info *rt = ip6_dst_alloc(net, ort->dst.dev, 0,
1891                                             ort->rt6i_table);
1892
1893         if (rt) {
1894                 rt->dst.input = ort->dst.input;
1895                 rt->dst.output = ort->dst.output;
1896                 rt->dst.flags |= DST_HOST;
1897
1898                 rt->rt6i_dst.addr = *dest;
1899                 rt->rt6i_dst.plen = 128;
1900                 dst_copy_metrics(&rt->dst, &ort->dst);
1901                 rt->dst.error = ort->dst.error;
1902                 rt->rt6i_idev = ort->rt6i_idev;
1903                 if (rt->rt6i_idev)
1904                         in6_dev_hold(rt->rt6i_idev);
1905                 rt->dst.lastuse = jiffies;
1906
1907                 if (ort->rt6i_flags & RTF_GATEWAY)
1908                         rt->rt6i_gateway = ort->rt6i_gateway;
1909                 else
1910                         rt->rt6i_gateway = *dest;
1911                 rt->rt6i_flags = ort->rt6i_flags;
1912                 if ((ort->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF)) ==
1913                     (RTF_DEFAULT | RTF_ADDRCONF))
1914                         rt6_set_from(rt, ort);
1915                 rt->rt6i_metric = 0;
1916
1917 #ifdef CONFIG_IPV6_SUBTREES
1918                 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1919 #endif
1920                 memcpy(&rt->rt6i_prefsrc, &ort->rt6i_prefsrc, sizeof(struct rt6key));
1921                 rt->rt6i_table = ort->rt6i_table;
1922         }
1923         return rt;
1924 }
1925
1926 #ifdef CONFIG_IPV6_ROUTE_INFO
1927 static struct rt6_info *rt6_get_route_info(struct net *net,
1928                                            const struct in6_addr *prefix, int prefixlen,
1929                                            const struct in6_addr *gwaddr, int ifindex)
1930 {
1931         struct fib6_node *fn;
1932         struct rt6_info *rt = NULL;
1933         struct fib6_table *table;
1934
1935         table = fib6_get_table(net, RT6_TABLE_INFO);
1936         if (!table)
1937                 return NULL;
1938
1939         read_lock_bh(&table->tb6_lock);
1940         fn = fib6_locate(&table->tb6_root, prefix ,prefixlen, NULL, 0);
1941         if (!fn)
1942                 goto out;
1943
1944         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1945                 if (rt->dst.dev->ifindex != ifindex)
1946                         continue;
1947                 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
1948                         continue;
1949                 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
1950                         continue;
1951                 dst_hold(&rt->dst);
1952                 break;
1953         }
1954 out:
1955         read_unlock_bh(&table->tb6_lock);
1956         return rt;
1957 }
1958
1959 static struct rt6_info *rt6_add_route_info(struct net *net,
1960                                            const struct in6_addr *prefix, int prefixlen,
1961                                            const struct in6_addr *gwaddr, int ifindex,
1962                                            unsigned int pref)
1963 {
1964         struct fib6_config cfg = {
1965                 .fc_table       = RT6_TABLE_INFO,
1966                 .fc_metric      = IP6_RT_PRIO_USER,
1967                 .fc_ifindex     = ifindex,
1968                 .fc_dst_len     = prefixlen,
1969                 .fc_flags       = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
1970                                   RTF_UP | RTF_PREF(pref),
1971                 .fc_nlinfo.portid = 0,
1972                 .fc_nlinfo.nlh = NULL,
1973                 .fc_nlinfo.nl_net = net,
1974         };
1975
1976         cfg.fc_dst = *prefix;
1977         cfg.fc_gateway = *gwaddr;
1978
1979         /* We should treat it as a default route if prefix length is 0. */
1980         if (!prefixlen)
1981                 cfg.fc_flags |= RTF_DEFAULT;
1982
1983         ip6_route_add(&cfg);
1984
1985         return rt6_get_route_info(net, prefix, prefixlen, gwaddr, ifindex);
1986 }
1987 #endif
1988
1989 struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
1990 {
1991         struct rt6_info *rt;
1992         struct fib6_table *table;
1993
1994         table = fib6_get_table(dev_net(dev), RT6_TABLE_DFLT);
1995         if (!table)
1996                 return NULL;
1997
1998         read_lock_bh(&table->tb6_lock);
1999         for (rt = table->tb6_root.leaf; rt; rt=rt->dst.rt6_next) {
2000                 if (dev == rt->dst.dev &&
2001                     ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
2002                     ipv6_addr_equal(&rt->rt6i_gateway, addr))
2003                         break;
2004         }
2005         if (rt)
2006                 dst_hold(&rt->dst);
2007         read_unlock_bh(&table->tb6_lock);
2008         return rt;
2009 }
2010
2011 struct rt6_info *rt6_add_dflt_router(const struct in6_addr *gwaddr,
2012                                      struct net_device *dev,
2013                                      unsigned int pref)
2014 {
2015         struct fib6_config cfg = {
2016                 .fc_table       = RT6_TABLE_DFLT,
2017                 .fc_metric      = IP6_RT_PRIO_USER,
2018                 .fc_ifindex     = dev->ifindex,
2019                 .fc_flags       = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
2020                                   RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
2021                 .fc_nlinfo.portid = 0,
2022                 .fc_nlinfo.nlh = NULL,
2023                 .fc_nlinfo.nl_net = dev_net(dev),
2024         };
2025
2026         cfg.fc_gateway = *gwaddr;
2027
2028         ip6_route_add(&cfg);
2029
2030         return rt6_get_dflt_router(gwaddr, dev);
2031 }
2032
2033 void rt6_purge_dflt_routers(struct net *net)
2034 {
2035         struct rt6_info *rt;
2036         struct fib6_table *table;
2037
2038         /* NOTE: Keep consistent with rt6_get_dflt_router */
2039         table = fib6_get_table(net, RT6_TABLE_DFLT);
2040         if (!table)
2041                 return;
2042
2043 restart:
2044         read_lock_bh(&table->tb6_lock);
2045         for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2046                 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
2047                     (!rt->rt6i_idev || rt->rt6i_idev->cnf.accept_ra != 2)) {
2048                         dst_hold(&rt->dst);
2049                         read_unlock_bh(&table->tb6_lock);
2050                         ip6_del_rt(rt);
2051                         goto restart;
2052                 }
2053         }
2054         read_unlock_bh(&table->tb6_lock);
2055 }
2056
2057 static void rtmsg_to_fib6_config(struct net *net,
2058                                  struct in6_rtmsg *rtmsg,
2059                                  struct fib6_config *cfg)
2060 {
2061         memset(cfg, 0, sizeof(*cfg));
2062
2063         cfg->fc_table = RT6_TABLE_MAIN;
2064         cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
2065         cfg->fc_metric = rtmsg->rtmsg_metric;
2066         cfg->fc_expires = rtmsg->rtmsg_info;
2067         cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
2068         cfg->fc_src_len = rtmsg->rtmsg_src_len;
2069         cfg->fc_flags = rtmsg->rtmsg_flags;
2070
2071         cfg->fc_nlinfo.nl_net = net;
2072
2073         cfg->fc_dst = rtmsg->rtmsg_dst;
2074         cfg->fc_src = rtmsg->rtmsg_src;
2075         cfg->fc_gateway = rtmsg->rtmsg_gateway;
2076 }
2077
2078 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
2079 {
2080         struct fib6_config cfg;
2081         struct in6_rtmsg rtmsg;
2082         int err;
2083
2084         switch(cmd) {
2085         case SIOCADDRT:         /* Add a route */
2086         case SIOCDELRT:         /* Delete a route */
2087                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2088                         return -EPERM;
2089                 err = copy_from_user(&rtmsg, arg,
2090                                      sizeof(struct in6_rtmsg));
2091                 if (err)
2092                         return -EFAULT;
2093
2094                 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
2095
2096                 rtnl_lock();
2097                 switch (cmd) {
2098                 case SIOCADDRT:
2099                         err = ip6_route_add(&cfg);
2100                         break;
2101                 case SIOCDELRT:
2102                         err = ip6_route_del(&cfg);
2103                         break;
2104                 default:
2105                         err = -EINVAL;
2106                 }
2107                 rtnl_unlock();
2108
2109                 return err;
2110         }
2111
2112         return -EINVAL;
2113 }
2114
2115 /*
2116  *      Drop the packet on the floor
2117  */
2118
2119 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
2120 {
2121         int type;
2122         struct dst_entry *dst = skb_dst(skb);
2123         switch (ipstats_mib_noroutes) {
2124         case IPSTATS_MIB_INNOROUTES:
2125                 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
2126                 if (type == IPV6_ADDR_ANY) {
2127                         IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2128                                       IPSTATS_MIB_INADDRERRORS);
2129                         break;
2130                 }
2131                 /* FALLTHROUGH */
2132         case IPSTATS_MIB_OUTNOROUTES:
2133                 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2134                               ipstats_mib_noroutes);
2135                 break;
2136         }
2137         icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
2138         kfree_skb(skb);
2139         return 0;
2140 }
2141
2142 static int ip6_pkt_discard(struct sk_buff *skb)
2143 {
2144         return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
2145 }
2146
2147 static int ip6_pkt_discard_out(struct sk_buff *skb)
2148 {
2149         skb->dev = skb_dst(skb)->dev;
2150         return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
2151 }
2152
2153 static int ip6_pkt_prohibit(struct sk_buff *skb)
2154 {
2155         return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
2156 }
2157
2158 static int ip6_pkt_prohibit_out(struct sk_buff *skb)
2159 {
2160         skb->dev = skb_dst(skb)->dev;
2161         return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
2162 }
2163
2164 /*
2165  *      Allocate a dst for local (unicast / anycast) address.
2166  */
2167
2168 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
2169                                     const struct in6_addr *addr,
2170                                     bool anycast)
2171 {
2172         struct net *net = dev_net(idev->dev);
2173         struct rt6_info *rt = ip6_dst_alloc(net, net->loopback_dev, 0, NULL);
2174
2175         if (!rt) {
2176                 net_warn_ratelimited("Maximum number of routes reached, consider increasing route/max_size\n");
2177                 return ERR_PTR(-ENOMEM);
2178         }
2179
2180         in6_dev_hold(idev);
2181
2182         rt->dst.flags |= DST_HOST;
2183         rt->dst.input = ip6_input;
2184         rt->dst.output = ip6_output;
2185         rt->rt6i_idev = idev;
2186
2187         rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
2188         if (anycast)
2189                 rt->rt6i_flags |= RTF_ANYCAST;
2190         else
2191                 rt->rt6i_flags |= RTF_LOCAL;
2192
2193         rt->rt6i_gateway  = *addr;
2194         rt->rt6i_dst.addr = *addr;
2195         rt->rt6i_dst.plen = 128;
2196         rt->rt6i_table = fib6_get_table(net, RT6_TABLE_LOCAL);
2197
2198         atomic_set(&rt->dst.__refcnt, 1);
2199
2200         return rt;
2201 }
2202
2203 int ip6_route_get_saddr(struct net *net,
2204                         struct rt6_info *rt,
2205                         const struct in6_addr *daddr,
2206                         unsigned int prefs,
2207                         struct in6_addr *saddr)
2208 {
2209         struct inet6_dev *idev = ip6_dst_idev((struct dst_entry*)rt);
2210         int err = 0;
2211         if (rt->rt6i_prefsrc.plen)
2212                 *saddr = rt->rt6i_prefsrc.addr;
2213         else
2214                 err = ipv6_dev_get_saddr(net, idev ? idev->dev : NULL,
2215                                          daddr, prefs, saddr);
2216         return err;
2217 }
2218
2219 /* remove deleted ip from prefsrc entries */
2220 struct arg_dev_net_ip {
2221         struct net_device *dev;
2222         struct net *net;
2223         struct in6_addr *addr;
2224 };
2225
2226 static int fib6_remove_prefsrc(struct rt6_info *rt, void *arg)
2227 {
2228         struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
2229         struct net *net = ((struct arg_dev_net_ip *)arg)->net;
2230         struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
2231
2232         if (((void *)rt->dst.dev == dev || !dev) &&
2233             rt != net->ipv6.ip6_null_entry &&
2234             ipv6_addr_equal(addr, &rt->rt6i_prefsrc.addr)) {
2235                 /* remove prefsrc entry */
2236                 rt->rt6i_prefsrc.plen = 0;
2237         }
2238         return 0;
2239 }
2240
2241 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
2242 {
2243         struct net *net = dev_net(ifp->idev->dev);
2244         struct arg_dev_net_ip adni = {
2245                 .dev = ifp->idev->dev,
2246                 .net = net,
2247                 .addr = &ifp->addr,
2248         };
2249         fib6_clean_all(net, fib6_remove_prefsrc, 0, &adni);
2250 }
2251
2252 struct arg_dev_net {
2253         struct net_device *dev;
2254         struct net *net;
2255 };
2256
2257 static int fib6_ifdown(struct rt6_info *rt, void *arg)
2258 {
2259         const struct arg_dev_net *adn = arg;
2260         const struct net_device *dev = adn->dev;
2261
2262         if ((rt->dst.dev == dev || !dev) &&
2263             rt != adn->net->ipv6.ip6_null_entry)
2264                 return -1;
2265
2266         return 0;
2267 }
2268
2269 void rt6_ifdown(struct net *net, struct net_device *dev)
2270 {
2271         struct arg_dev_net adn = {
2272                 .dev = dev,
2273                 .net = net,
2274         };
2275
2276         fib6_clean_all(net, fib6_ifdown, 0, &adn);
2277         icmp6_clean_all(fib6_ifdown, &adn);
2278 }
2279
2280 struct rt6_mtu_change_arg {
2281         struct net_device *dev;
2282         unsigned int mtu;
2283 };
2284
2285 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2286 {
2287         struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2288         struct inet6_dev *idev;
2289
2290         /* In IPv6 pmtu discovery is not optional,
2291            so that RTAX_MTU lock cannot disable it.
2292            We still use this lock to block changes
2293            caused by addrconf/ndisc.
2294         */
2295
2296         idev = __in6_dev_get(arg->dev);
2297         if (!idev)
2298                 return 0;
2299
2300         /* For administrative MTU increase, there is no way to discover
2301            IPv6 PMTU increase, so PMTU increase should be updated here.
2302            Since RFC 1981 doesn't include administrative MTU increase
2303            update PMTU increase is a MUST. (i.e. jumbo frame)
2304          */
2305         /*
2306            If new MTU is less than route PMTU, this new MTU will be the
2307            lowest MTU in the path, update the route PMTU to reflect PMTU
2308            decreases; if new MTU is greater than route PMTU, and the
2309            old MTU is the lowest MTU in the path, update the route PMTU
2310            to reflect the increase. In this case if the other nodes' MTU
2311            also have the lowest MTU, TOO BIG MESSAGE will be lead to
2312            PMTU discouvery.
2313          */
2314         if (rt->dst.dev == arg->dev &&
2315             !dst_metric_locked(&rt->dst, RTAX_MTU) &&
2316             (dst_mtu(&rt->dst) >= arg->mtu ||
2317              (dst_mtu(&rt->dst) < arg->mtu &&
2318               dst_mtu(&rt->dst) == idev->cnf.mtu6))) {
2319                 dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
2320         }
2321         return 0;
2322 }
2323
2324 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
2325 {
2326         struct rt6_mtu_change_arg arg = {
2327                 .dev = dev,
2328                 .mtu = mtu,
2329         };
2330
2331         fib6_clean_all(dev_net(dev), rt6_mtu_change_route, 0, &arg);
2332 }
2333
2334 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2335         [RTA_GATEWAY]           = { .len = sizeof(struct in6_addr) },
2336         [RTA_OIF]               = { .type = NLA_U32 },
2337         [RTA_IIF]               = { .type = NLA_U32 },
2338         [RTA_PRIORITY]          = { .type = NLA_U32 },
2339         [RTA_METRICS]           = { .type = NLA_NESTED },
2340         [RTA_MULTIPATH]         = { .len = sizeof(struct rtnexthop) },
2341 };
2342
2343 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2344                               struct fib6_config *cfg)
2345 {
2346         struct rtmsg *rtm;
2347         struct nlattr *tb[RTA_MAX+1];
2348         int err;
2349
2350         err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2351         if (err < 0)
2352                 goto errout;
2353
2354         err = -EINVAL;
2355         rtm = nlmsg_data(nlh);
2356         memset(cfg, 0, sizeof(*cfg));
2357
2358         cfg->fc_table = rtm->rtm_table;
2359         cfg->fc_dst_len = rtm->rtm_dst_len;
2360         cfg->fc_src_len = rtm->rtm_src_len;
2361         cfg->fc_flags = RTF_UP;
2362         cfg->fc_protocol = rtm->rtm_protocol;
2363         cfg->fc_type = rtm->rtm_type;
2364
2365         if (rtm->rtm_type == RTN_UNREACHABLE ||
2366             rtm->rtm_type == RTN_BLACKHOLE ||
2367             rtm->rtm_type == RTN_PROHIBIT ||
2368             rtm->rtm_type == RTN_THROW)
2369                 cfg->fc_flags |= RTF_REJECT;
2370
2371         if (rtm->rtm_type == RTN_LOCAL)
2372                 cfg->fc_flags |= RTF_LOCAL;
2373
2374         cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid;
2375         cfg->fc_nlinfo.nlh = nlh;
2376         cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2377
2378         if (tb[RTA_GATEWAY]) {
2379                 nla_memcpy(&cfg->fc_gateway, tb[RTA_GATEWAY], 16);
2380                 cfg->fc_flags |= RTF_GATEWAY;
2381         }
2382
2383         if (tb[RTA_DST]) {
2384                 int plen = (rtm->rtm_dst_len + 7) >> 3;
2385
2386                 if (nla_len(tb[RTA_DST]) < plen)
2387                         goto errout;
2388
2389                 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2390         }
2391
2392         if (tb[RTA_SRC]) {
2393                 int plen = (rtm->rtm_src_len + 7) >> 3;
2394
2395                 if (nla_len(tb[RTA_SRC]) < plen)
2396                         goto errout;
2397
2398                 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2399         }
2400
2401         if (tb[RTA_PREFSRC])
2402                 nla_memcpy(&cfg->fc_prefsrc, tb[RTA_PREFSRC], 16);
2403
2404         if (tb[RTA_OIF])
2405                 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2406
2407         if (tb[RTA_PRIORITY])
2408                 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2409
2410         if (tb[RTA_METRICS]) {
2411                 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2412                 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2413         }
2414
2415         if (tb[RTA_TABLE])
2416                 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2417
2418         if (tb[RTA_MULTIPATH]) {
2419                 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
2420                 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
2421         }
2422
2423         err = 0;
2424 errout:
2425         return err;
2426 }
2427
2428 static int ip6_route_multipath(struct fib6_config *cfg, int add)
2429 {
2430         struct fib6_config r_cfg;
2431         struct rtnexthop *rtnh;
2432         int remaining;
2433         int attrlen;
2434         int err = 0, last_err = 0;
2435
2436 beginning:
2437         rtnh = (struct rtnexthop *)cfg->fc_mp;
2438         remaining = cfg->fc_mp_len;
2439
2440         /* Parse a Multipath Entry */
2441         while (rtnh_ok(rtnh, remaining)) {
2442                 memcpy(&r_cfg, cfg, sizeof(*cfg));
2443                 if (rtnh->rtnh_ifindex)
2444                         r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
2445
2446                 attrlen = rtnh_attrlen(rtnh);
2447                 if (attrlen > 0) {
2448                         struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
2449
2450                         nla = nla_find(attrs, attrlen, RTA_GATEWAY);
2451                         if (nla) {
2452                                 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
2453                                 r_cfg.fc_flags |= RTF_GATEWAY;
2454                         }
2455                 }
2456                 err = add ? ip6_route_add(&r_cfg) : ip6_route_del(&r_cfg);
2457                 if (err) {
2458                         last_err = err;
2459                         /* If we are trying to remove a route, do not stop the
2460                          * loop when ip6_route_del() fails (because next hop is
2461                          * already gone), we should try to remove all next hops.
2462                          */
2463                         if (add) {
2464                                 /* If add fails, we should try to delete all
2465                                  * next hops that have been already added.
2466                                  */
2467                                 add = 0;
2468                                 goto beginning;
2469                         }
2470                 }
2471                 /* Because each route is added like a single route we remove
2472                  * this flag after the first nexthop (if there is a collision,
2473                  * we have already fail to add the first nexthop:
2474                  * fib6_add_rt2node() has reject it).
2475                  */
2476                 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~NLM_F_EXCL;
2477                 rtnh = rtnh_next(rtnh, &remaining);
2478         }
2479
2480         return last_err;
2481 }
2482
2483 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr* nlh)
2484 {
2485         struct fib6_config cfg;
2486         int err;
2487
2488         err = rtm_to_fib6_config(skb, nlh, &cfg);
2489         if (err < 0)
2490                 return err;
2491
2492         if (cfg.fc_mp)
2493                 return ip6_route_multipath(&cfg, 0);
2494         else
2495                 return ip6_route_del(&cfg);
2496 }
2497
2498 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr* nlh)
2499 {
2500         struct fib6_config cfg;
2501         int err;
2502
2503         err = rtm_to_fib6_config(skb, nlh, &cfg);
2504         if (err < 0)
2505                 return err;
2506
2507         if (cfg.fc_mp)
2508                 return ip6_route_multipath(&cfg, 1);
2509         else
2510                 return ip6_route_add(&cfg);
2511 }
2512
2513 static inline size_t rt6_nlmsg_size(void)
2514 {
2515         return NLMSG_ALIGN(sizeof(struct rtmsg))
2516                + nla_total_size(16) /* RTA_SRC */
2517                + nla_total_size(16) /* RTA_DST */
2518                + nla_total_size(16) /* RTA_GATEWAY */
2519                + nla_total_size(16) /* RTA_PREFSRC */
2520                + nla_total_size(4) /* RTA_TABLE */
2521                + nla_total_size(4) /* RTA_IIF */
2522                + nla_total_size(4) /* RTA_OIF */
2523                + nla_total_size(4) /* RTA_PRIORITY */
2524                + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
2525                + nla_total_size(sizeof(struct rta_cacheinfo));
2526 }
2527
2528 static int rt6_fill_node(struct net *net,
2529                          struct sk_buff *skb, struct rt6_info *rt,
2530                          struct in6_addr *dst, struct in6_addr *src,
2531                          int iif, int type, u32 portid, u32 seq,
2532                          int prefix, int nowait, unsigned int flags)
2533 {
2534         struct rtmsg *rtm;
2535         struct nlmsghdr *nlh;
2536         long expires;
2537         u32 table;
2538
2539         if (prefix) {   /* user wants prefix routes only */
2540                 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
2541                         /* success since this is not a prefix route */
2542                         return 1;
2543                 }
2544         }
2545
2546         nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
2547         if (!nlh)
2548                 return -EMSGSIZE;
2549
2550         rtm = nlmsg_data(nlh);
2551         rtm->rtm_family = AF_INET6;
2552         rtm->rtm_dst_len = rt->rt6i_dst.plen;
2553         rtm->rtm_src_len = rt->rt6i_src.plen;
2554         rtm->rtm_tos = 0;
2555         if (rt->rt6i_table)
2556                 table = rt->rt6i_table->tb6_id;
2557         else
2558                 table = RT6_TABLE_UNSPEC;
2559         rtm->rtm_table = table;
2560         if (nla_put_u32(skb, RTA_TABLE, table))
2561                 goto nla_put_failure;
2562         if (rt->rt6i_flags & RTF_REJECT) {
2563                 switch (rt->dst.error) {
2564                 case -EINVAL:
2565                         rtm->rtm_type = RTN_BLACKHOLE;
2566                         break;
2567                 case -EACCES:
2568                         rtm->rtm_type = RTN_PROHIBIT;
2569                         break;
2570                 case -EAGAIN:
2571                         rtm->rtm_type = RTN_THROW;
2572                         break;
2573                 default:
2574                         rtm->rtm_type = RTN_UNREACHABLE;
2575                         break;
2576                 }
2577         }
2578         else if (rt->rt6i_flags & RTF_LOCAL)
2579                 rtm->rtm_type = RTN_LOCAL;
2580         else if (rt->dst.dev && (rt->dst.dev->flags & IFF_LOOPBACK))
2581                 rtm->rtm_type = RTN_LOCAL;
2582         else
2583                 rtm->rtm_type = RTN_UNICAST;
2584         rtm->rtm_flags = 0;
2585         rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2586         rtm->rtm_protocol = rt->rt6i_protocol;
2587         if (rt->rt6i_flags & RTF_DYNAMIC)
2588                 rtm->rtm_protocol = RTPROT_REDIRECT;
2589         else if (rt->rt6i_flags & RTF_ADDRCONF) {
2590                 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ROUTEINFO))
2591                         rtm->rtm_protocol = RTPROT_RA;
2592                 else
2593                         rtm->rtm_protocol = RTPROT_KERNEL;
2594         }
2595
2596         if (rt->rt6i_flags & RTF_CACHE)
2597                 rtm->rtm_flags |= RTM_F_CLONED;
2598
2599         if (dst) {
2600                 if (nla_put(skb, RTA_DST, 16, dst))
2601                         goto nla_put_failure;
2602                 rtm->rtm_dst_len = 128;
2603         } else if (rtm->rtm_dst_len)
2604                 if (nla_put(skb, RTA_DST, 16, &rt->rt6i_dst.addr))
2605                         goto nla_put_failure;
2606 #ifdef CONFIG_IPV6_SUBTREES
2607         if (src) {
2608                 if (nla_put(skb, RTA_SRC, 16, src))
2609                         goto nla_put_failure;
2610                 rtm->rtm_src_len = 128;
2611         } else if (rtm->rtm_src_len &&
2612                    nla_put(skb, RTA_SRC, 16, &rt->rt6i_src.addr))
2613                 goto nla_put_failure;
2614 #endif
2615         if (iif) {
2616 #ifdef CONFIG_IPV6_MROUTE
2617                 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
2618                         int err = ip6mr_get_route(net, skb, rtm, nowait);
2619                         if (err <= 0) {
2620                                 if (!nowait) {
2621                                         if (err == 0)
2622                                                 return 0;
2623                                         goto nla_put_failure;
2624                                 } else {
2625                                         if (err == -EMSGSIZE)
2626                                                 goto nla_put_failure;
2627                                 }
2628                         }
2629                 } else
2630 #endif
2631                         if (nla_put_u32(skb, RTA_IIF, iif))
2632                                 goto nla_put_failure;
2633         } else if (dst) {
2634                 struct in6_addr saddr_buf;
2635                 if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0 &&
2636                     nla_put(skb, RTA_PREFSRC, 16, &saddr_buf))
2637                         goto nla_put_failure;
2638         }
2639
2640         if (rt->rt6i_prefsrc.plen) {
2641                 struct in6_addr saddr_buf;
2642                 saddr_buf = rt->rt6i_prefsrc.addr;
2643                 if (nla_put(skb, RTA_PREFSRC, 16, &saddr_buf))
2644                         goto nla_put_failure;
2645         }
2646
2647         if (rtnetlink_put_metrics(skb, dst_metrics_ptr(&rt->dst)) < 0)
2648                 goto nla_put_failure;
2649
2650         if (rt->rt6i_flags & RTF_GATEWAY) {
2651                 if (nla_put(skb, RTA_GATEWAY, 16, &rt->rt6i_gateway) < 0)
2652                         goto nla_put_failure;
2653         }
2654
2655         if (rt->dst.dev &&
2656             nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex))
2657                 goto nla_put_failure;
2658         if (nla_put_u32(skb, RTA_PRIORITY, rt->rt6i_metric))
2659                 goto nla_put_failure;
2660
2661         expires = (rt->rt6i_flags & RTF_EXPIRES) ? rt->dst.expires - jiffies : 0;
2662
2663         if (rtnl_put_cacheinfo(skb, &rt->dst, 0, expires, rt->dst.error) < 0)
2664                 goto nla_put_failure;
2665
2666         return nlmsg_end(skb, nlh);
2667
2668 nla_put_failure:
2669         nlmsg_cancel(skb, nlh);
2670         return -EMSGSIZE;
2671 }
2672
2673 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
2674 {
2675         struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
2676         int prefix;
2677
2678         if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
2679                 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
2680                 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
2681         } else
2682                 prefix = 0;
2683
2684         return rt6_fill_node(arg->net,
2685                      arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
2686                      NETLINK_CB(arg->cb->skb).portid, arg->cb->nlh->nlmsg_seq,
2687                      prefix, 0, NLM_F_MULTI);
2688 }
2689
2690 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr* nlh)
2691 {
2692         struct net *net = sock_net(in_skb->sk);
2693         struct nlattr *tb[RTA_MAX+1];
2694         struct rt6_info *rt;
2695         struct sk_buff *skb;
2696         struct rtmsg *rtm;
2697         struct flowi6 fl6;
2698         int err, iif = 0, oif = 0;
2699
2700         err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2701         if (err < 0)
2702                 goto errout;
2703
2704         err = -EINVAL;
2705         memset(&fl6, 0, sizeof(fl6));
2706
2707         if (tb[RTA_SRC]) {
2708                 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
2709                         goto errout;
2710
2711                 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
2712         }
2713
2714         if (tb[RTA_DST]) {
2715                 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
2716                         goto errout;
2717
2718                 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
2719         }
2720
2721         if (tb[RTA_IIF])
2722                 iif = nla_get_u32(tb[RTA_IIF]);
2723
2724         if (tb[RTA_OIF])
2725                 oif = nla_get_u32(tb[RTA_OIF]);
2726
2727         if (iif) {
2728                 struct net_device *dev;
2729                 int flags = 0;
2730
2731                 dev = __dev_get_by_index(net, iif);
2732                 if (!dev) {
2733                         err = -ENODEV;
2734                         goto errout;
2735                 }
2736
2737                 fl6.flowi6_iif = iif;
2738
2739                 if (!ipv6_addr_any(&fl6.saddr))
2740                         flags |= RT6_LOOKUP_F_HAS_SADDR;
2741
2742                 rt = (struct rt6_info *)ip6_route_input_lookup(net, dev, &fl6,
2743                                                                flags);
2744         } else {
2745                 fl6.flowi6_oif = oif;
2746
2747                 rt = (struct rt6_info *)ip6_route_output(net, NULL, &fl6);
2748         }
2749
2750         skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
2751         if (!skb) {
2752                 ip6_rt_put(rt);
2753                 err = -ENOBUFS;
2754                 goto errout;
2755         }
2756
2757         /* Reserve room for dummy headers, this skb can pass
2758            through good chunk of routing engine.
2759          */
2760         skb_reset_mac_header(skb);
2761         skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
2762
2763         skb_dst_set(skb, &rt->dst);
2764
2765         err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
2766                             RTM_NEWROUTE, NETLINK_CB(in_skb).portid,
2767                             nlh->nlmsg_seq, 0, 0, 0);
2768         if (err < 0) {
2769                 kfree_skb(skb);
2770                 goto errout;
2771         }
2772
2773         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
2774 errout:
2775         return err;
2776 }
2777
2778 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info)
2779 {
2780         struct sk_buff *skb;
2781         struct net *net = info->nl_net;
2782         u32 seq;
2783         int err;
2784
2785         err = -ENOBUFS;
2786         seq = info->nlh ? info->nlh->nlmsg_seq : 0;
2787
2788         skb = nlmsg_new(rt6_nlmsg_size(), gfp_any());
2789         if (!skb)
2790                 goto errout;
2791
2792         err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
2793                                 event, info->portid, seq, 0, 0, 0);
2794         if (err < 0) {
2795                 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
2796                 WARN_ON(err == -EMSGSIZE);
2797                 kfree_skb(skb);
2798                 goto errout;
2799         }
2800         rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
2801                     info->nlh, gfp_any());
2802         return;
2803 errout:
2804         if (err < 0)
2805                 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
2806 }
2807
2808 static int ip6_route_dev_notify(struct notifier_block *this,
2809                                 unsigned long event, void *ptr)
2810 {
2811         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2812         struct net *net = dev_net(dev);
2813
2814         if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
2815                 net->ipv6.ip6_null_entry->dst.dev = dev;
2816                 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
2817 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2818                 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
2819                 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
2820                 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
2821                 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
2822 #endif
2823         }
2824
2825         return NOTIFY_OK;
2826 }
2827
2828 /*
2829  *      /proc
2830  */
2831
2832 #ifdef CONFIG_PROC_FS
2833
2834 static const struct file_operations ipv6_route_proc_fops = {
2835         .owner          = THIS_MODULE,
2836         .open           = ipv6_route_open,
2837         .read           = seq_read,
2838         .llseek         = seq_lseek,
2839         .release        = seq_release_net,
2840 };
2841
2842 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
2843 {
2844         struct net *net = (struct net *)seq->private;
2845         seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
2846                    net->ipv6.rt6_stats->fib_nodes,
2847                    net->ipv6.rt6_stats->fib_route_nodes,
2848                    net->ipv6.rt6_stats->fib_rt_alloc,
2849                    net->ipv6.rt6_stats->fib_rt_entries,
2850                    net->ipv6.rt6_stats->fib_rt_cache,
2851                    dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
2852                    net->ipv6.rt6_stats->fib_discarded_routes);
2853
2854         return 0;
2855 }
2856
2857 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
2858 {
2859         return single_open_net(inode, file, rt6_stats_seq_show);
2860 }
2861
2862 static const struct file_operations rt6_stats_seq_fops = {
2863         .owner   = THIS_MODULE,
2864         .open    = rt6_stats_seq_open,
2865         .read    = seq_read,
2866         .llseek  = seq_lseek,
2867         .release = single_release_net,
2868 };
2869 #endif  /* CONFIG_PROC_FS */
2870
2871 #ifdef CONFIG_SYSCTL
2872
2873 static
2874 int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
2875                               void __user *buffer, size_t *lenp, loff_t *ppos)
2876 {
2877         struct net *net;
2878         int delay;
2879         if (!write)
2880                 return -EINVAL;
2881
2882         net = (struct net *)ctl->extra1;
2883         delay = net->ipv6.sysctl.flush_delay;
2884         proc_dointvec(ctl, write, buffer, lenp, ppos);
2885         fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
2886         return 0;
2887 }
2888
2889 struct ctl_table ipv6_route_table_template[] = {
2890         {
2891                 .procname       =       "flush",
2892                 .data           =       &init_net.ipv6.sysctl.flush_delay,
2893                 .maxlen         =       sizeof(int),
2894                 .mode           =       0200,
2895                 .proc_handler   =       ipv6_sysctl_rtcache_flush
2896         },
2897         {
2898                 .procname       =       "gc_thresh",
2899                 .data           =       &ip6_dst_ops_template.gc_thresh,
2900                 .maxlen         =       sizeof(int),
2901                 .mode           =       0644,
2902                 .proc_handler   =       proc_dointvec,
2903         },
2904         {
2905                 .procname       =       "max_size",
2906                 .data           =       &init_net.ipv6.sysctl.ip6_rt_max_size,
2907                 .maxlen         =       sizeof(int),
2908                 .mode           =       0644,
2909                 .proc_handler   =       proc_dointvec,
2910         },
2911         {
2912                 .procname       =       "gc_min_interval",
2913                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2914                 .maxlen         =       sizeof(int),
2915                 .mode           =       0644,
2916                 .proc_handler   =       proc_dointvec_jiffies,
2917         },
2918         {
2919                 .procname       =       "gc_timeout",
2920                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
2921                 .maxlen         =       sizeof(int),
2922                 .mode           =       0644,
2923                 .proc_handler   =       proc_dointvec_jiffies,
2924         },
2925         {
2926                 .procname       =       "gc_interval",
2927                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_interval,
2928                 .maxlen         =       sizeof(int),
2929                 .mode           =       0644,
2930                 .proc_handler   =       proc_dointvec_jiffies,
2931         },
2932         {
2933                 .procname       =       "gc_elasticity",
2934                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
2935                 .maxlen         =       sizeof(int),
2936                 .mode           =       0644,
2937                 .proc_handler   =       proc_dointvec,
2938         },
2939         {
2940                 .procname       =       "mtu_expires",
2941                 .data           =       &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
2942                 .maxlen         =       sizeof(int),
2943                 .mode           =       0644,
2944                 .proc_handler   =       proc_dointvec_jiffies,
2945         },
2946         {
2947                 .procname       =       "min_adv_mss",
2948                 .data           =       &init_net.ipv6.sysctl.ip6_rt_min_advmss,
2949                 .maxlen         =       sizeof(int),
2950                 .mode           =       0644,
2951                 .proc_handler   =       proc_dointvec,
2952         },
2953         {
2954                 .procname       =       "gc_min_interval_ms",
2955                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2956                 .maxlen         =       sizeof(int),
2957                 .mode           =       0644,
2958                 .proc_handler   =       proc_dointvec_ms_jiffies,
2959         },
2960         { }
2961 };
2962
2963 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
2964 {
2965         struct ctl_table *table;
2966
2967         table = kmemdup(ipv6_route_table_template,
2968                         sizeof(ipv6_route_table_template),
2969                         GFP_KERNEL);
2970
2971         if (table) {
2972                 table[0].data = &net->ipv6.sysctl.flush_delay;
2973                 table[0].extra1 = net;
2974                 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
2975                 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
2976                 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2977                 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
2978                 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
2979                 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
2980                 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
2981                 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
2982                 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2983
2984                 /* Don't export sysctls to unprivileged users */
2985                 if (net->user_ns != &init_user_ns)
2986                         table[0].procname = NULL;
2987         }
2988
2989         return table;
2990 }
2991 #endif
2992
2993 static int __net_init ip6_route_net_init(struct net *net)
2994 {
2995         int ret = -ENOMEM;
2996
2997         memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
2998                sizeof(net->ipv6.ip6_dst_ops));
2999
3000         if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
3001                 goto out_ip6_dst_ops;
3002
3003         net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
3004                                            sizeof(*net->ipv6.ip6_null_entry),
3005                                            GFP_KERNEL);
3006         if (!net->ipv6.ip6_null_entry)
3007                 goto out_ip6_dst_entries;
3008         net->ipv6.ip6_null_entry->dst.path =
3009                 (struct dst_entry *)net->ipv6.ip6_null_entry;
3010         net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3011         dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
3012                          ip6_template_metrics, true);
3013
3014 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3015         net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
3016                                                sizeof(*net->ipv6.ip6_prohibit_entry),
3017                                                GFP_KERNEL);
3018         if (!net->ipv6.ip6_prohibit_entry)
3019                 goto out_ip6_null_entry;
3020         net->ipv6.ip6_prohibit_entry->dst.path =
3021                 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
3022         net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3023         dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
3024                          ip6_template_metrics, true);
3025
3026         net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
3027                                                sizeof(*net->ipv6.ip6_blk_hole_entry),
3028                                                GFP_KERNEL);
3029         if (!net->ipv6.ip6_blk_hole_entry)
3030                 goto out_ip6_prohibit_entry;
3031         net->ipv6.ip6_blk_hole_entry->dst.path =
3032                 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
3033         net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3034         dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
3035                          ip6_template_metrics, true);
3036 #endif
3037
3038         net->ipv6.sysctl.flush_delay = 0;
3039         net->ipv6.sysctl.ip6_rt_max_size = 4096;
3040         net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
3041         net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
3042         net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
3043         net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
3044         net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
3045         net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
3046
3047         net->ipv6.ip6_rt_gc_expire = 30*HZ;
3048
3049         ret = 0;
3050 out:
3051         return ret;
3052
3053 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3054 out_ip6_prohibit_entry:
3055         kfree(net->ipv6.ip6_prohibit_entry);
3056 out_ip6_null_entry:
3057         kfree(net->ipv6.ip6_null_entry);
3058 #endif
3059 out_ip6_dst_entries:
3060         dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3061 out_ip6_dst_ops:
3062         goto out;
3063 }
3064
3065 static void __net_exit ip6_route_net_exit(struct net *net)
3066 {
3067         kfree(net->ipv6.ip6_null_entry);
3068 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3069         kfree(net->ipv6.ip6_prohibit_entry);
3070         kfree(net->ipv6.ip6_blk_hole_entry);
3071 #endif
3072         dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3073 }
3074
3075 static int __net_init ip6_route_net_init_late(struct net *net)
3076 {
3077 #ifdef CONFIG_PROC_FS
3078         proc_create("ipv6_route", 0, net->proc_net, &ipv6_route_proc_fops);
3079         proc_create("rt6_stats", S_IRUGO, net->proc_net, &rt6_stats_seq_fops);
3080 #endif
3081         return 0;
3082 }
3083
3084 static void __net_exit ip6_route_net_exit_late(struct net *net)
3085 {
3086 #ifdef CONFIG_PROC_FS
3087         remove_proc_entry("ipv6_route", net->proc_net);
3088         remove_proc_entry("rt6_stats", net->proc_net);
3089 #endif
3090 }
3091
3092 static struct pernet_operations ip6_route_net_ops = {
3093         .init = ip6_route_net_init,
3094         .exit = ip6_route_net_exit,
3095 };
3096
3097 static int __net_init ipv6_inetpeer_init(struct net *net)
3098 {
3099         struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
3100
3101         if (!bp)
3102                 return -ENOMEM;
3103         inet_peer_base_init(bp);
3104         net->ipv6.peers = bp;
3105         return 0;
3106 }
3107
3108 static void __net_exit ipv6_inetpeer_exit(struct net *net)
3109 {
3110         struct inet_peer_base *bp = net->ipv6.peers;
3111
3112         net->ipv6.peers = NULL;
3113         inetpeer_invalidate_tree(bp);
3114         kfree(bp);
3115 }
3116
3117 static struct pernet_operations ipv6_inetpeer_ops = {
3118         .init   =       ipv6_inetpeer_init,
3119         .exit   =       ipv6_inetpeer_exit,
3120 };
3121
3122 static struct pernet_operations ip6_route_net_late_ops = {
3123         .init = ip6_route_net_init_late,
3124         .exit = ip6_route_net_exit_late,
3125 };
3126
3127 static struct notifier_block ip6_route_dev_notifier = {
3128         .notifier_call = ip6_route_dev_notify,
3129         .priority = 0,
3130 };
3131
3132 int __init ip6_route_init(void)
3133 {
3134         int ret;
3135
3136         ret = -ENOMEM;
3137         ip6_dst_ops_template.kmem_cachep =
3138                 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
3139                                   SLAB_HWCACHE_ALIGN, NULL);
3140         if (!ip6_dst_ops_template.kmem_cachep)
3141                 goto out;
3142
3143         ret = dst_entries_init(&ip6_dst_blackhole_ops);
3144         if (ret)
3145                 goto out_kmem_cache;
3146
3147         ret = register_pernet_subsys(&ipv6_inetpeer_ops);
3148         if (ret)
3149                 goto out_dst_entries;
3150
3151         ret = register_pernet_subsys(&ip6_route_net_ops);
3152         if (ret)
3153                 goto out_register_inetpeer;
3154
3155         ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
3156
3157         /* Registering of the loopback is done before this portion of code,
3158          * the loopback reference in rt6_info will not be taken, do it
3159          * manually for init_net */
3160         init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
3161         init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3162   #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3163         init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
3164         init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3165         init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
3166         init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3167   #endif
3168         ret = fib6_init();
3169         if (ret)
3170                 goto out_register_subsys;
3171
3172         ret = xfrm6_init();
3173         if (ret)
3174                 goto out_fib6_init;
3175
3176         ret = fib6_rules_init();
3177         if (ret)
3178                 goto xfrm6_init;
3179
3180         ret = register_pernet_subsys(&ip6_route_net_late_ops);
3181         if (ret)
3182                 goto fib6_rules_init;
3183
3184         ret = -ENOBUFS;
3185         if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL, NULL) ||
3186             __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL, NULL) ||
3187             __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL, NULL))
3188                 goto out_register_late_subsys;
3189
3190         ret = register_netdevice_notifier(&ip6_route_dev_notifier);
3191         if (ret)
3192                 goto out_register_late_subsys;
3193
3194 out:
3195         return ret;
3196
3197 out_register_late_subsys:
3198         unregister_pernet_subsys(&ip6_route_net_late_ops);
3199 fib6_rules_init:
3200         fib6_rules_cleanup();
3201 xfrm6_init:
3202         xfrm6_fini();
3203 out_fib6_init:
3204         fib6_gc_cleanup();
3205 out_register_subsys:
3206         unregister_pernet_subsys(&ip6_route_net_ops);
3207 out_register_inetpeer:
3208         unregister_pernet_subsys(&ipv6_inetpeer_ops);
3209 out_dst_entries:
3210         dst_entries_destroy(&ip6_dst_blackhole_ops);
3211 out_kmem_cache:
3212         kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3213         goto out;
3214 }
3215
3216 void ip6_route_cleanup(void)
3217 {
3218         unregister_netdevice_notifier(&ip6_route_dev_notifier);
3219         unregister_pernet_subsys(&ip6_route_net_late_ops);
3220         fib6_rules_cleanup();
3221         xfrm6_fini();
3222         fib6_gc_cleanup();
3223         unregister_pernet_subsys(&ipv6_inetpeer_ops);
3224         unregister_pernet_subsys(&ip6_route_net_ops);
3225         dst_entries_destroy(&ip6_dst_blackhole_ops);
3226         kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3227 }