net/ipv6/route.c

   1 /*
   2  *      Linux INET6 implementation
   3  *      FIB front-end.
   4  *
   5  *      Authors:
   6  *      Pedro Roque             <roque@di.fc.ul.pt>
   7  *
   8  *      This program is free software; you can redistribute it and/or
   9  *      modify it under the terms of the GNU General Public License
  10  *      as published by the Free Software Foundation; either version
  11  *      2 of the License, or (at your option) any later version.
  12  */
  13
  14 /*      Changes:
  15  *
  16  *      YOSHIFUJI Hideaki @USAGI
  17  *              reworked default router selection.
  18  *              - respect outgoing interface
  19  *              - select from (probably) reachable routers (i.e.
  20  *              routers in REACHABLE, STALE, DELAY or PROBE states).
  21  *              - always select the same router if it is (probably)
  22  *              reachable.  otherwise, round-robin the list.
  23  *      Ville Nuorvala
  24  *              Fixed routing subtrees.
  25  */
  26
  27 #define pr_fmt(fmt) "IPv6: " fmt
  28
  29 #include <linux/capability.h>
  30 #include <linux/errno.h>
  31 #include <linux/export.h>
  32 #include <linux/types.h>
  33 #include <linux/times.h>
  34 #include <linux/socket.h>
  35 #include <linux/sockios.h>
  36 #include <linux/net.h>
  37 #include <linux/route.h>
  38 #include <linux/netdevice.h>
  39 #include <linux/in6.h>
  40 #include <linux/mroute6.h>
  41 #include <linux/init.h>
  42 #include <linux/if_arp.h>
  43 #include <linux/proc_fs.h>
  44 #include <linux/seq_file.h>
  45 #include <linux/nsproxy.h>
  46 #include <linux/slab.h>
  47 #include <net/net_namespace.h>
  48 #include <net/snmp.h>
  49 #include <net/ipv6.h>
  50 #include <net/ip6_fib.h>
  51 #include <net/ip6_route.h>
  52 #include <net/ndisc.h>
  53 #include <net/addrconf.h>
  54 #include <net/tcp.h>
  55 #include <linux/rtnetlink.h>
  56 #include <net/dst.h>
  57 #include <net/xfrm.h>
  58 #include <net/netevent.h>
  59 #include <net/netlink.h>
  60 #include <net/nexthop.h>
  61
  62 #include <asm/uaccess.h>
  63
  64 #ifdef CONFIG_SYSCTL
  65 #include <linux/sysctl.h>
  66 #endif
  67
  68 enum rt6_nud_state {
  69         RT6_NUD_FAIL_HARD = -3,
  70         RT6_NUD_FAIL_PROBE = -2,
  71         RT6_NUD_FAIL_DO_RR = -1,
  72         RT6_NUD_SUCCEED = 1
  73 };
  74
  75 static struct rt6_info *ip6_rt_copy(struct rt6_info *ort,
  76                                     const struct in6_addr *dest);
  77 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie);
  78 static unsigned int      ip6_default_advmss(const struct dst_entry *dst);
  79 static unsigned int      ip6_mtu(const struct dst_entry *dst);
  80 static struct dst_entry *ip6_negative_advice(struct dst_entry *);
  81 static void             ip6_dst_destroy(struct dst_entry *);
  82 static void             ip6_dst_ifdown(struct dst_entry *,
  83                                        struct net_device *dev, int how);
  84 static int               ip6_dst_gc(struct dst_ops *ops);
  85
  86 static int              ip6_pkt_discard(struct sk_buff *skb);
  87 static int              ip6_pkt_discard_out(struct sock *sk, struct sk_buff *skb);
  88 static int              ip6_pkt_prohibit(struct sk_buff *skb);
  89 static int              ip6_pkt_prohibit_out(struct sock *sk, struct sk_buff *skb);
  90 static void             ip6_link_failure(struct sk_buff *skb);
  91 static void             ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
  92                                            struct sk_buff *skb, u32 mtu);
  93 static void             rt6_do_redirect(struct dst_entry *dst, struct sock *sk,
  94                                         struct sk_buff *skb);
  95 static int rt6_score_route(struct rt6_info *rt, int oif, int strict);
  96
  97 #ifdef CONFIG_IPV6_ROUTE_INFO
  98 static struct rt6_info *rt6_add_route_info(struct net *net,
  99                                            const struct in6_addr *prefix, int prefixlen,
 100                                            const struct in6_addr *gwaddr, int ifindex,
 101                                            unsigned int pref);
 102 static struct rt6_info *rt6_get_route_info(struct net *net,
 103                                            const struct in6_addr *prefix, int prefixlen,
 104                                            const struct in6_addr *gwaddr, int ifindex);
 105 #endif
 106
 107 static void rt6_bind_peer(struct rt6_info *rt, int create)
 108 {
 109         struct inet_peer_base *base;
 110         struct inet_peer *peer;
 111
 112         base = inetpeer_base_ptr(rt->_rt6i_peer);
 113         if (!base)
 114                 return;
 115
 116         peer = inet_getpeer_v6(base, &rt->rt6i_dst.addr, create);
 117         if (peer) {
 118                 if (!rt6_set_peer(rt, peer))
 119                         inet_putpeer(peer);
 120         }
 121 }
 122
 123 static struct inet_peer *__rt6_get_peer(struct rt6_info *rt, int create)
 124 {
 125         if (rt6_has_peer(rt))
 126                 return rt6_peer_ptr(rt);
 127
 128         rt6_bind_peer(rt, create);
 129         return (rt6_has_peer(rt) ? rt6_peer_ptr(rt) : NULL);
 130 }
 131
 132 static struct inet_peer *rt6_get_peer_create(struct rt6_info *rt)
 133 {
 134         return __rt6_get_peer(rt, 1);
 135 }
 136
 137 static u32 *ipv6_cow_metrics(struct dst_entry *dst, unsigned long old)
 138 {
 139         struct rt6_info *rt = (struct rt6_info *) dst;
 140         struct inet_peer *peer;
 141         u32 *p = NULL;
 142
 143         if (!(rt->dst.flags & DST_HOST))
 144                 return dst_cow_metrics_generic(dst, old);
 145
 146         peer = rt6_get_peer_create(rt);
 147         if (peer) {
 148                 u32 *old_p = __DST_METRICS_PTR(old);
 149                 unsigned long prev, new;
 150
 151                 p = peer->metrics;
 152                 if (inet_metrics_new(peer) ||
 153                     (old & DST_METRICS_FORCE_OVERWRITE))
 154                         memcpy(p, old_p, sizeof(u32) * RTAX_MAX);
 155
 156                 new = (unsigned long) p;
 157                 prev = cmpxchg(&dst->_metrics, old, new);
 158
 159                 if (prev != old) {
 160                         p = __DST_METRICS_PTR(prev);
 161                         if (prev & DST_METRICS_READ_ONLY)
 162                                 p = NULL;
 163                 }
 164         }
 165         return p;
 166 }
 167
 168 static inline const void *choose_neigh_daddr(struct rt6_info *rt,
 169                                              struct sk_buff *skb,
 170                                              const void *daddr)
 171 {
 172         struct in6_addr *p = &rt->rt6i_gateway;
 173
 174         if (!ipv6_addr_any(p))
 175                 return (const void *) p;
 176         else if (skb)
 177                 return &ipv6_hdr(skb)->daddr;
 178         return daddr;
 179 }
 180
 181 static struct neighbour *ip6_neigh_lookup(const struct dst_entry *dst,
 182                                           struct sk_buff *skb,
 183                                           const void *daddr)
 184 {
 185         struct rt6_info *rt = (struct rt6_info *) dst;
 186         struct neighbour *n;
 187
 188         daddr = choose_neigh_daddr(rt, skb, daddr);
 189         n = __ipv6_neigh_lookup(dst->dev, daddr);
 190         if (n)
 191                 return n;
 192         return neigh_create(&nd_tbl, daddr, dst->dev);
 193 }
 194
 195 static struct dst_ops ip6_dst_ops_template = {
 196         .family                 =       AF_INET6,
 197         .gc                     =       ip6_dst_gc,
 198         .gc_thresh              =       1024,
 199         .check                  =       ip6_dst_check,
 200         .default_advmss         =       ip6_default_advmss,
 201         .mtu                    =       ip6_mtu,
 202         .cow_metrics            =       ipv6_cow_metrics,
 203         .destroy                =       ip6_dst_destroy,
 204         .ifdown                 =       ip6_dst_ifdown,
 205         .negative_advice        =       ip6_negative_advice,
 206         .link_failure           =       ip6_link_failure,
 207         .update_pmtu            =       ip6_rt_update_pmtu,
 208         .redirect               =       rt6_do_redirect,
 209         .local_out              =       __ip6_local_out,
 210         .neigh_lookup           =       ip6_neigh_lookup,
 211 };
 212
 213 static unsigned int ip6_blackhole_mtu(const struct dst_entry *dst)
 214 {
 215         unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
 216
 217         return mtu ? : dst->dev->mtu;
 218 }
 219
 220 static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, struct sock *sk,
 221                                          struct sk_buff *skb, u32 mtu)
 222 {
 223 }
 224
 225 static void ip6_rt_blackhole_redirect(struct dst_entry *dst, struct sock *sk,
 226                                       struct sk_buff *skb)
 227 {
 228 }
 229
 230 static u32 *ip6_rt_blackhole_cow_metrics(struct dst_entry *dst,
 231                                          unsigned long old)
 232 {
 233         return NULL;
 234 }
 235
 236 static struct dst_ops ip6_dst_blackhole_ops = {
 237         .family                 =       AF_INET6,
 238         .destroy                =       ip6_dst_destroy,
 239         .check                  =       ip6_dst_check,
 240         .mtu                    =       ip6_blackhole_mtu,
 241         .default_advmss         =       ip6_default_advmss,
 242         .update_pmtu            =       ip6_rt_blackhole_update_pmtu,
 243         .redirect               =       ip6_rt_blackhole_redirect,
 244         .cow_metrics            =       ip6_rt_blackhole_cow_metrics,
 245         .neigh_lookup           =       ip6_neigh_lookup,
 246 };
 247
 248 static const u32 ip6_template_metrics[RTAX_MAX] = {
 249         [RTAX_HOPLIMIT - 1] = 0,
 250 };
 251
 252 static const struct rt6_info ip6_null_entry_template = {
 253         .dst = {
 254                 .__refcnt       = ATOMIC_INIT(1),
 255                 .__use          = 1,
 256                 .obsolete       = DST_OBSOLETE_FORCE_CHK,
 257                 .error          = -ENETUNREACH,
 258                 .input          = ip6_pkt_discard,
 259                 .output         = ip6_pkt_discard_out,
 260         },
 261         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
 262         .rt6i_protocol  = RTPROT_KERNEL,
 263         .rt6i_metric    = ~(u32) 0,
 264         .rt6i_ref       = ATOMIC_INIT(1),
 265 };
 266
 267 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
 268
 269 static const struct rt6_info ip6_prohibit_entry_template = {
 270         .dst = {
 271                 .__refcnt       = ATOMIC_INIT(1),
 272                 .__use          = 1,
 273                 .obsolete       = DST_OBSOLETE_FORCE_CHK,
 274                 .error          = -EACCES,
 275                 .input          = ip6_pkt_prohibit,
 276                 .output         = ip6_pkt_prohibit_out,
 277         },
 278         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
 279         .rt6i_protocol  = RTPROT_KERNEL,
 280         .rt6i_metric    = ~(u32) 0,
 281         .rt6i_ref       = ATOMIC_INIT(1),
 282 };
 283
 284 static const struct rt6_info ip6_blk_hole_entry_template = {
 285         .dst = {
 286                 .__refcnt       = ATOMIC_INIT(1),
 287                 .__use          = 1,
 288                 .obsolete       = DST_OBSOLETE_FORCE_CHK,
 289                 .error          = -EINVAL,
 290                 .input          = dst_discard,
 291                 .output         = dst_discard_sk,
 292         },
 293         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
 294         .rt6i_protocol  = RTPROT_KERNEL,
 295         .rt6i_metric    = ~(u32) 0,
 296         .rt6i_ref       = ATOMIC_INIT(1),
 297 };
 298
 299 #endif
 300
 301 /* allocate dst with ip6_dst_ops */
 302 static inline struct rt6_info *ip6_dst_alloc(struct net *net,
 303                                              struct net_device *dev,
 304                                              int flags,
 305                                              struct fib6_table *table)
 306 {
 307         struct rt6_info *rt = dst_alloc(&net->ipv6.ip6_dst_ops, dev,
 308                                         0, DST_OBSOLETE_FORCE_CHK, flags);
 309
 310         if (rt) {
 311                 struct dst_entry *dst = &rt->dst;
 312
 313                 memset(dst + 1, 0, sizeof(*rt) - sizeof(*dst));
 314                 rt6_init_peer(rt, table ? &table->tb6_peers : net->ipv6.peers);
 315                 INIT_LIST_HEAD(&rt->rt6i_siblings);
 316         }
 317         return rt;
 318 }
 319
 320 static void ip6_dst_destroy(struct dst_entry *dst)
 321 {
 322         struct rt6_info *rt = (struct rt6_info *)dst;
 323         struct inet6_dev *idev = rt->rt6i_idev;
 324         struct dst_entry *from = dst->from;
 325
 326         if (!(rt->dst.flags & DST_HOST))
 327                 dst_destroy_metrics_generic(dst);
 328
 329         if (idev) {
 330                 rt->rt6i_idev = NULL;
 331                 in6_dev_put(idev);
 332         }
 333
 334         dst->from = NULL;
 335         dst_release(from);
 336
 337         if (rt6_has_peer(rt)) {
 338                 struct inet_peer *peer = rt6_peer_ptr(rt);
 339                 inet_putpeer(peer);
 340         }
 341 }
 342
 343 static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
 344                            int how)
 345 {
 346         struct rt6_info *rt = (struct rt6_info *)dst;
 347         struct inet6_dev *idev = rt->rt6i_idev;
 348         struct net_device *loopback_dev =
 349                 dev_net(dev)->loopback_dev;
 350
 351         if (dev != loopback_dev) {
 352                 if (idev && idev->dev == dev) {
 353                         struct inet6_dev *loopback_idev =
 354                                 in6_dev_get(loopback_dev);
 355                         if (loopback_idev) {
 356                                 rt->rt6i_idev = loopback_idev;
 357                                 in6_dev_put(idev);
 358                         }
 359                 }
 360         }
 361 }
 362
 363 static bool rt6_check_expired(const struct rt6_info *rt)
 364 {
 365         if (rt->rt6i_flags & RTF_EXPIRES) {
 366                 if (time_after(jiffies, rt->dst.expires))
 367                         return true;
 368         } else if (rt->dst.from) {
 369                 return rt6_check_expired((struct rt6_info *) rt->dst.from);
 370         }
 371         return false;
 372 }
 373
 374 /* Multipath route selection:
 375  *   Hash based function using packet header and flowlabel.
 376  * Adapted from fib_info_hashfn()
 377  */
 378 static int rt6_info_hash_nhsfn(unsigned int candidate_count,
 379                                const struct flowi6 *fl6)
 380 {
 381         unsigned int val = fl6->flowi6_proto;
 382
 383         val ^= ipv6_addr_hash(&fl6->daddr);
 384         val ^= ipv6_addr_hash(&fl6->saddr);
 385
 386         /* Work only if this not encapsulated */
 387         switch (fl6->flowi6_proto) {
 388         case IPPROTO_UDP:
 389         case IPPROTO_TCP:
 390         case IPPROTO_SCTP:
 391                 val ^= (__force u16)fl6->fl6_sport;
 392                 val ^= (__force u16)fl6->fl6_dport;
 393                 break;
 394
 395         case IPPROTO_ICMPV6:
 396                 val ^= (__force u16)fl6->fl6_icmp_type;
 397                 val ^= (__force u16)fl6->fl6_icmp_code;
 398                 break;
 399         }
 400         /* RFC6438 recommands to use flowlabel */
 401         val ^= (__force u32)fl6->flowlabel;
 402
 403         /* Perhaps, we need to tune, this function? */
 404         val = val ^ (val >> 7) ^ (val >> 12);
 405         return val % candidate_count;
 406 }
 407
 408 static struct rt6_info *rt6_multipath_select(struct rt6_info *match,
 409                                              struct flowi6 *fl6, int oif,
 410                                              int strict)
 411 {
 412         struct rt6_info *sibling, *next_sibling;
 413         int route_choosen;
 414
 415         route_choosen = rt6_info_hash_nhsfn(match->rt6i_nsiblings + 1, fl6);
 416         /* Don't change the route, if route_choosen == 0
 417          * (siblings does not include ourself)
 418          */
 419         if (route_choosen)
 420                 list_for_each_entry_safe(sibling, next_sibling,
 421                                 &match->rt6i_siblings, rt6i_siblings) {
 422                         route_choosen--;
 423                         if (route_choosen == 0) {
 424                                 if (rt6_score_route(sibling, oif, strict) < 0)
 425                                         break;
 426                                 match = sibling;
 427                                 break;
 428                         }
 429                 }
 430         return match;
 431 }
 432
 433 /*
 434  *      Route lookup. Any table->tb6_lock is implied.
 435  */
 436
 437 static inline struct rt6_info *rt6_device_match(struct net *net,
 438                                                     struct rt6_info *rt,
 439                                                     const struct in6_addr *saddr,
 440                                                     int oif,
 441                                                     int flags)
 442 {
 443         struct rt6_info *local = NULL;
 444         struct rt6_info *sprt;
 445
 446         if (!oif && ipv6_addr_any(saddr))
 447                 goto out;
 448
 449         for (sprt = rt; sprt; sprt = sprt->dst.rt6_next) {
 450                 struct net_device *dev = sprt->dst.dev;
 451
 452                 if (oif) {
 453                         if (dev->ifindex == oif)
 454                                 return sprt;
 455                         if (dev->flags & IFF_LOOPBACK) {
 456                                 if (!sprt->rt6i_idev ||
 457                                     sprt->rt6i_idev->dev->ifindex != oif) {
 458                                         if (flags & RT6_LOOKUP_F_IFACE && oif)
 459                                                 continue;
 460                                         if (local && (!oif ||
 461                                                       local->rt6i_idev->dev->ifindex == oif))
 462                                                 continue;
 463                                 }
 464                                 local = sprt;
 465                         }
 466                 } else {
 467                         if (ipv6_chk_addr(net, saddr, dev,
 468                                           flags & RT6_LOOKUP_F_IFACE))
 469                                 return sprt;
 470                 }
 471         }
 472
 473         if (oif) {
 474                 if (local)
 475                         return local;
 476
 477                 if (flags & RT6_LOOKUP_F_IFACE)
 478                         return net->ipv6.ip6_null_entry;
 479         }
 480 out:
 481         return rt;
 482 }
 483
 484 #ifdef CONFIG_IPV6_ROUTER_PREF
 485 struct __rt6_probe_work {
 486         struct work_struct work;
 487         struct in6_addr target;
 488         struct net_device *dev;
 489 };
 490
 491 static void rt6_probe_deferred(struct work_struct *w)
 492 {
 493         struct in6_addr mcaddr;
 494         struct __rt6_probe_work *work =
 495                 container_of(w, struct __rt6_probe_work, work);
 496
 497         addrconf_addr_solict_mult(&work->target, &mcaddr);
 498         ndisc_send_ns(work->dev, NULL, &work->target, &mcaddr, NULL);
 499         dev_put(work->dev);
 500         kfree(work);
 501 }
 502
 503 static void rt6_probe(struct rt6_info *rt)
 504 {
 505         struct neighbour *neigh;
 506         /*
 507          * Okay, this does not seem to be appropriate
 508          * for now, however, we need to check if it
 509          * is really so; aka Router Reachability Probing.
 510          *
 511          * Router Reachability Probe MUST be rate-limited
 512          * to no more than one per minute.
 513          */
 514         if (!rt || !(rt->rt6i_flags & RTF_GATEWAY))
 515                 return;
 516         rcu_read_lock_bh();
 517         neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
 518         if (neigh) {
 519                 write_lock(&neigh->lock);
 520                 if (neigh->nud_state & NUD_VALID)
 521                         goto out;
 522         }
 523
 524         if (!neigh ||
 525             time_after(jiffies, neigh->updated + rt->rt6i_idev->cnf.rtr_probe_interval)) {
 526                 struct __rt6_probe_work *work;
 527
 528                 work = kmalloc(sizeof(*work), GFP_ATOMIC);
 529
 530                 if (neigh && work)
 531                         __neigh_set_probe_once(neigh);
 532
 533                 if (neigh)
 534                         write_unlock(&neigh->lock);
 535
 536                 if (work) {
 537                         INIT_WORK(&work->work, rt6_probe_deferred);
 538                         work->target = rt->rt6i_gateway;
 539                         dev_hold(rt->dst.dev);
 540                         work->dev = rt->dst.dev;
 541                         schedule_work(&work->work);
 542                 }
 543         } else {
 544 out:
 545                 write_unlock(&neigh->lock);
 546         }
 547         rcu_read_unlock_bh();
 548 }
 549 #else
 550 static inline void rt6_probe(struct rt6_info *rt)
 551 {
 552 }
 553 #endif
 554
 555 /*
 556  * Default Router Selection (RFC 2461 6.3.6)
 557  */
 558 static inline int rt6_check_dev(struct rt6_info *rt, int oif)
 559 {
 560         struct net_device *dev = rt->dst.dev;
 561         if (!oif || dev->ifindex == oif)
 562                 return 2;
 563         if ((dev->flags & IFF_LOOPBACK) &&
 564             rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
 565                 return 1;
 566         return 0;
 567 }
 568
 569 static inline enum rt6_nud_state rt6_check_neigh(struct rt6_info *rt)
 570 {
 571         struct neighbour *neigh;
 572         enum rt6_nud_state ret = RT6_NUD_FAIL_HARD;
 573
 574         if (rt->rt6i_flags & RTF_NONEXTHOP ||
 575             !(rt->rt6i_flags & RTF_GATEWAY))
 576                 return RT6_NUD_SUCCEED;
 577
 578         rcu_read_lock_bh();
 579         neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
 580         if (neigh) {
 581                 read_lock(&neigh->lock);
 582                 if (neigh->nud_state & NUD_VALID)
 583                         ret = RT6_NUD_SUCCEED;
 584 #ifdef CONFIG_IPV6_ROUTER_PREF
 585                 else if (!(neigh->nud_state & NUD_FAILED))
 586                         ret = RT6_NUD_SUCCEED;
 587                 else
 588                         ret = RT6_NUD_FAIL_PROBE;
 589 #endif
 590                 read_unlock(&neigh->lock);
 591         } else {
 592                 ret = IS_ENABLED(CONFIG_IPV6_ROUTER_PREF) ?
 593                       RT6_NUD_SUCCEED : RT6_NUD_FAIL_DO_RR;
 594         }
 595         rcu_read_unlock_bh();
 596
 597         return ret;
 598 }
 599
 600 static int rt6_score_route(struct rt6_info *rt, int oif,
 601                            int strict)
 602 {
 603         int m;
 604
 605         m = rt6_check_dev(rt, oif);
 606         if (!m && (strict & RT6_LOOKUP_F_IFACE))
 607                 return RT6_NUD_FAIL_HARD;
 608 #ifdef CONFIG_IPV6_ROUTER_PREF
 609         m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
 610 #endif
 611         if (strict & RT6_LOOKUP_F_REACHABLE) {
 612                 int n = rt6_check_neigh(rt);
 613                 if (n < 0)
 614                         return n;
 615         }
 616         return m;
 617 }
 618
 619 static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
 620                                    int *mpri, struct rt6_info *match,
 621                                    bool *do_rr)
 622 {
 623         int m;
 624         bool match_do_rr = false;
 625
 626         if (rt6_check_expired(rt))
 627                 goto out;
 628
 629         m = rt6_score_route(rt, oif, strict);
 630         if (m == RT6_NUD_FAIL_DO_RR) {
 631                 match_do_rr = true;
 632                 m = 0; /* lowest valid score */
 633         } else if (m == RT6_NUD_FAIL_HARD) {
 634                 goto out;
 635         }
 636
 637         if (strict & RT6_LOOKUP_F_REACHABLE)
 638                 rt6_probe(rt);
 639
 640         /* note that m can be RT6_NUD_FAIL_PROBE at this point */
 641         if (m > *mpri) {
 642                 *do_rr = match_do_rr;
 643                 *mpri = m;
 644                 match = rt;
 645         }
 646 out:
 647         return match;
 648 }
 649
 650 static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
 651                                      struct rt6_info *rr_head,
 652                                      u32 metric, int oif, int strict,
 653                                      bool *do_rr)
 654 {
 655         struct rt6_info *rt, *match;
 656         int mpri = -1;
 657
 658         match = NULL;
 659         for (rt = rr_head; rt && rt->rt6i_metric == metric;
 660              rt = rt->dst.rt6_next)
 661                 match = find_match(rt, oif, strict, &mpri, match, do_rr);
 662         for (rt = fn->leaf; rt && rt != rr_head && rt->rt6i_metric == metric;
 663              rt = rt->dst.rt6_next)
 664                 match = find_match(rt, oif, strict, &mpri, match, do_rr);
 665
 666         return match;
 667 }
 668
 669 static struct rt6_info *rt6_select(struct fib6_node *fn, int oif, int strict)
 670 {
 671         struct rt6_info *match, *rt0;
 672         struct net *net;
 673         bool do_rr = false;
 674
 675         rt0 = fn->rr_ptr;
 676         if (!rt0)
 677                 fn->rr_ptr = rt0 = fn->leaf;
 678
 679         match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict,
 680                              &do_rr);
 681
 682         if (do_rr) {
 683                 struct rt6_info *next = rt0->dst.rt6_next;
 684
 685                 /* no entries matched; do round-robin */
 686                 if (!next || next->rt6i_metric != rt0->rt6i_metric)
 687                         next = fn->leaf;
 688
 689                 if (next != rt0)
 690                         fn->rr_ptr = next;
 691         }
 692
 693         net = dev_net(rt0->dst.dev);
 694         return match ? match : net->ipv6.ip6_null_entry;
 695 }
 696
 697 #ifdef CONFIG_IPV6_ROUTE_INFO
 698 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
 699                   const struct in6_addr *gwaddr)
 700 {
 701         struct net *net = dev_net(dev);
 702         struct route_info *rinfo = (struct route_info *) opt;
 703         struct in6_addr prefix_buf, *prefix;
 704         unsigned int pref;
 705         unsigned long lifetime;
 706         struct rt6_info *rt;
 707
 708         if (len < sizeof(struct route_info)) {
 709                 return -EINVAL;
 710         }
 711
 712         /* Sanity check for prefix_len and length */
 713         if (rinfo->length > 3) {
 714                 return -EINVAL;
 715         } else if (rinfo->prefix_len > 128) {
 716                 return -EINVAL;
 717         } else if (rinfo->prefix_len > 64) {
 718                 if (rinfo->length < 2) {
 719                         return -EINVAL;
 720                 }
 721         } else if (rinfo->prefix_len > 0) {
 722                 if (rinfo->length < 1) {
 723                         return -EINVAL;
 724                 }
 725         }
 726
 727         pref = rinfo->route_pref;
 728         if (pref == ICMPV6_ROUTER_PREF_INVALID)
 729                 return -EINVAL;
 730
 731         lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
 732
 733         if (rinfo->length == 3)
 734                 prefix = (struct in6_addr *)rinfo->prefix;
 735         else {
 736                 /* this function is safe */
 737                 ipv6_addr_prefix(&prefix_buf,
 738                                  (struct in6_addr *)rinfo->prefix,
 739                                  rinfo->prefix_len);
 740                 prefix = &prefix_buf;
 741         }
 742
 743         if (rinfo->prefix_len == 0)
 744                 rt = rt6_get_dflt_router(gwaddr, dev);
 745         else
 746                 rt = rt6_get_route_info(net, prefix, rinfo->prefix_len,
 747                                         gwaddr, dev->ifindex);
 748
 749         if (rt && !lifetime) {
 750                 ip6_del_rt(rt);
 751                 rt = NULL;
 752         }
 753
 754         if (!rt && lifetime)
 755                 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr, dev->ifindex,
 756                                         pref);
 757         else if (rt)
 758                 rt->rt6i_flags = RTF_ROUTEINFO |
 759                                  (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
 760
 761         if (rt) {
 762                 if (!addrconf_finite_timeout(lifetime))
 763                         rt6_clean_expires(rt);
 764                 else
 765                         rt6_set_expires(rt, jiffies + HZ * lifetime);
 766
 767                 ip6_rt_put(rt);
 768         }
 769         return 0;
 770 }
 771 #endif
 772
 773 static struct fib6_node* fib6_backtrack(struct fib6_node *fn,
 774                                         struct in6_addr *saddr)
 775 {
 776         struct fib6_node *pn;
 777         while (1) {
 778                 if (fn->fn_flags & RTN_TL_ROOT)
 779                         return NULL;
 780                 pn = fn->parent;
 781                 if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn)
 782                         fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr);
 783                 else
 784                         fn = pn;
 785                 if (fn->fn_flags & RTN_RTINFO)
 786                         return fn;
 787         }
 788 }
 789
 790 static struct rt6_info *ip6_pol_route_lookup(struct net *net,
 791                                              struct fib6_table *table,
 792                                              struct flowi6 *fl6, int flags)
 793 {
 794         struct fib6_node *fn;
 795         struct rt6_info *rt;
 796
 797         read_lock_bh(&table->tb6_lock);
 798         fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
 799 restart:
 800         rt = fn->leaf;
 801         rt = rt6_device_match(net, rt, &fl6->saddr, fl6->flowi6_oif, flags);
 802         if (rt->rt6i_nsiblings && fl6->flowi6_oif == 0)
 803                 rt = rt6_multipath_select(rt, fl6, fl6->flowi6_oif, flags);
 804         if (rt == net->ipv6.ip6_null_entry) {
 805                 fn = fib6_backtrack(fn, &fl6->saddr);
 806                 if (fn)
 807                         goto restart;
 808         }
 809         dst_use(&rt->dst, jiffies);
 810         read_unlock_bh(&table->tb6_lock);
 811         return rt;
 812
 813 }
 814
 815 struct dst_entry *ip6_route_lookup(struct net *net, struct flowi6 *fl6,
 816                                     int flags)
 817 {
 818         return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_lookup);
 819 }
 820 EXPORT_SYMBOL_GPL(ip6_route_lookup);
 821
 822 struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
 823                             const struct in6_addr *saddr, int oif, int strict)
 824 {
 825         struct flowi6 fl6 = {
 826                 .flowi6_oif = oif,
 827                 .daddr = *daddr,
 828         };
 829         struct dst_entry *dst;
 830         int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
 831
 832         if (saddr) {
 833                 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
 834                 flags |= RT6_LOOKUP_F_HAS_SADDR;
 835         }
 836
 837         dst = fib6_rule_lookup(net, &fl6, flags, ip6_pol_route_lookup);
 838         if (dst->error == 0)
 839                 return (struct rt6_info *) dst;
 840
 841         dst_release(dst);
 842
 843         return NULL;
 844 }
 845 EXPORT_SYMBOL(rt6_lookup);
 846
 847 /* ip6_ins_rt is called with FREE table->tb6_lock.
 848    It takes new route entry, the addition fails by any reason the
 849    route is freed. In any case, if caller does not hold it, it may
 850    be destroyed.
 851  */
 852
 853 static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info,
 854                         struct mx6_config *mxc)
 855 {
 856         int err;
 857         struct fib6_table *table;
 858
 859         table = rt->rt6i_table;
 860         write_lock_bh(&table->tb6_lock);
 861         err = fib6_add(&table->tb6_root, rt, info, mxc);
 862         write_unlock_bh(&table->tb6_lock);
 863
 864         return err;
 865 }
 866
 867 int ip6_ins_rt(struct rt6_info *rt)
 868 {
 869         struct nl_info info = { .nl_net = dev_net(rt->dst.dev), };
 870         struct mx6_config mxc = { .mx = NULL, };
 871
 872         return __ip6_ins_rt(rt, &info, &mxc);
 873 }
 874
 875 static struct rt6_info *rt6_alloc_cow(struct rt6_info *ort,
 876                                       const struct in6_addr *daddr,
 877                                       const struct in6_addr *saddr)
 878 {
 879         struct rt6_info *rt;
 880
 881         /*
 882          *      Clone the route.
 883          */
 884
 885         rt = ip6_rt_copy(ort, daddr);
 886
 887         if (rt) {
 888                 if (ort->rt6i_dst.plen != 128 &&
 889                     ipv6_addr_equal(&ort->rt6i_dst.addr, daddr))
 890                         rt->rt6i_flags |= RTF_ANYCAST;
 891
 892                 rt->rt6i_flags |= RTF_CACHE;
 893
 894 #ifdef CONFIG_IPV6_SUBTREES
 895                 if (rt->rt6i_src.plen && saddr) {
 896                         rt->rt6i_src.addr = *saddr;
 897                         rt->rt6i_src.plen = 128;
 898                 }
 899 #endif
 900         }
 901
 902         return rt;
 903 }
 904
 905 static struct rt6_info *rt6_alloc_clone(struct rt6_info *ort,
 906                                         const struct in6_addr *daddr)
 907 {
 908         struct rt6_info *rt = ip6_rt_copy(ort, daddr);
 909
 910         if (rt)
 911                 rt->rt6i_flags |= RTF_CACHE;
 912         return rt;
 913 }
 914
 915 static struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table, int oif,
 916                                       struct flowi6 *fl6, int flags)
 917 {
 918         struct fib6_node *fn, *saved_fn;
 919         struct rt6_info *rt, *nrt;
 920         int strict = 0;
 921         int attempts = 3;
 922         int err;
 923
 924         strict |= flags & RT6_LOOKUP_F_IFACE;
 925         if (net->ipv6.devconf_all->forwarding == 0)
 926                 strict |= RT6_LOOKUP_F_REACHABLE;
 927
 928 redo_fib6_lookup_lock:
 929         read_lock_bh(&table->tb6_lock);
 930
 931         fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
 932         saved_fn = fn;
 933
 934 redo_rt6_select:
 935         rt = rt6_select(fn, oif, strict);
 936         if (rt->rt6i_nsiblings)
 937                 rt = rt6_multipath_select(rt, fl6, oif, strict);
 938         if (rt == net->ipv6.ip6_null_entry) {
 939                 fn = fib6_backtrack(fn, &fl6->saddr);
 940                 if (fn)
 941                         goto redo_rt6_select;
 942                 else if (strict & RT6_LOOKUP_F_REACHABLE) {
 943                         /* also consider unreachable route */
 944                         strict &= ~RT6_LOOKUP_F_REACHABLE;
 945                         fn = saved_fn;
 946                         goto redo_rt6_select;
 947                 } else {
 948                         dst_hold(&rt->dst);
 949                         read_unlock_bh(&table->tb6_lock);
 950                         goto out2;
 951                 }
 952         }
 953
 954         dst_hold(&rt->dst);
 955         read_unlock_bh(&table->tb6_lock);
 956
 957         if (rt->rt6i_flags & RTF_CACHE)
 958                 goto out2;
 959
 960         if (!(rt->rt6i_flags & (RTF_NONEXTHOP | RTF_GATEWAY)))
 961                 nrt = rt6_alloc_cow(rt, &fl6->daddr, &fl6->saddr);
 962         else if (!(rt->dst.flags & DST_HOST))
 963                 nrt = rt6_alloc_clone(rt, &fl6->daddr);
 964         else
 965                 goto out2;
 966
 967         ip6_rt_put(rt);
 968         rt = nrt ? : net->ipv6.ip6_null_entry;
 969
 970         dst_hold(&rt->dst);
 971         if (nrt) {
 972                 err = ip6_ins_rt(nrt);
 973                 if (!err)
 974                         goto out2;
 975         }
 976
 977         if (--attempts <= 0)
 978                 goto out2;
 979
 980         /*
 981          * Race condition! In the gap, when table->tb6_lock was
 982          * released someone could insert this route.  Relookup.
 983          */
 984         ip6_rt_put(rt);
 985         goto redo_fib6_lookup_lock;
 986
 987 out2:
 988         rt->dst.lastuse = jiffies;
 989         rt->dst.__use++;
 990
 991         return rt;
 992 }
 993
 994 static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
 995                                             struct flowi6 *fl6, int flags)
 996 {
 997         return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, flags);
 998 }
 999
1000 static struct dst_entry *ip6_route_input_lookup(struct net *net,
1001                                                 struct net_device *dev,
1002                                                 struct flowi6 *fl6, int flags)
1003 {
1004         if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
1005                 flags |= RT6_LOOKUP_F_IFACE;
1006
1007         return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_input);
1008 }
1009
1010 void ip6_route_input(struct sk_buff *skb)
1011 {
1012         const struct ipv6hdr *iph = ipv6_hdr(skb);
1013         struct net *net = dev_net(skb->dev);
1014         int flags = RT6_LOOKUP_F_HAS_SADDR;
1015         struct flowi6 fl6 = {
1016                 .flowi6_iif = skb->dev->ifindex,
1017                 .daddr = iph->daddr,
1018                 .saddr = iph->saddr,
1019                 .flowlabel = ip6_flowinfo(iph),
1020                 .flowi6_mark = skb->mark,
1021                 .flowi6_proto = iph->nexthdr,
1022         };
1023
1024         skb_dst_set(skb, ip6_route_input_lookup(net, skb->dev, &fl6, flags));
1025 }
1026
1027 static struct rt6_info *ip6_pol_route_output(struct net *net, struct fib6_table *table,
1028                                              struct flowi6 *fl6, int flags)
1029 {
1030         return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, flags);
1031 }
1032
1033 struct dst_entry *ip6_route_output(struct net *net, const struct sock *sk,
1034                                     struct flowi6 *fl6)
1035 {
1036         int flags = 0;
1037
1038         fl6->flowi6_iif = LOOPBACK_IFINDEX;
1039
1040         if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr))
1041                 flags |= RT6_LOOKUP_F_IFACE;
1042
1043         if (!ipv6_addr_any(&fl6->saddr))
1044                 flags |= RT6_LOOKUP_F_HAS_SADDR;
1045         else if (sk)
1046                 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
1047
1048         return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_output);
1049 }
1050 EXPORT_SYMBOL(ip6_route_output);
1051
1052 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
1053 {
1054         struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
1055         struct dst_entry *new = NULL;
1056
1057         rt = dst_alloc(&ip6_dst_blackhole_ops, ort->dst.dev, 1, DST_OBSOLETE_NONE, 0);
1058         if (rt) {
1059                 new = &rt->dst;
1060
1061                 memset(new + 1, 0, sizeof(*rt) - sizeof(*new));
1062                 rt6_init_peer(rt, net->ipv6.peers);
1063
1064                 new->__use = 1;
1065                 new->input = dst_discard;
1066                 new->output = dst_discard_sk;
1067
1068                 if (dst_metrics_read_only(&ort->dst))
1069                         new->_metrics = ort->dst._metrics;
1070                 else
1071                         dst_copy_metrics(new, &ort->dst);
1072                 rt->rt6i_idev = ort->rt6i_idev;
1073                 if (rt->rt6i_idev)
1074                         in6_dev_hold(rt->rt6i_idev);
1075
1076                 rt->rt6i_gateway = ort->rt6i_gateway;
1077                 rt->rt6i_flags = ort->rt6i_flags;
1078                 rt->rt6i_metric = 0;
1079
1080                 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1081 #ifdef CONFIG_IPV6_SUBTREES
1082                 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1083 #endif
1084
1085                 dst_free(new);
1086         }
1087
1088         dst_release(dst_orig);
1089         return new ? new : ERR_PTR(-ENOMEM);
1090 }
1091
1092 /*
1093  *      Destination cache support functions
1094  */
1095
1096 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
1097 {
1098         struct rt6_info *rt;
1099
1100         rt = (struct rt6_info *) dst;
1101
1102         /* All IPV6 dsts are created with ->obsolete set to the value
1103          * DST_OBSOLETE_FORCE_CHK which forces validation calls down
1104          * into this function always.
1105          */
1106         if (!rt->rt6i_node || (rt->rt6i_node->fn_sernum != cookie))
1107                 return NULL;
1108
1109         if (rt6_check_expired(rt))
1110                 return NULL;
1111
1112         return dst;
1113 }
1114
1115 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
1116 {
1117         struct rt6_info *rt = (struct rt6_info *) dst;
1118
1119         if (rt) {
1120                 if (rt->rt6i_flags & RTF_CACHE) {
1121                         if (rt6_check_expired(rt)) {
1122                                 ip6_del_rt(rt);
1123                                 dst = NULL;
1124                         }
1125                 } else {
1126                         dst_release(dst);
1127                         dst = NULL;
1128                 }
1129         }
1130         return dst;
1131 }
1132
1133 static void ip6_link_failure(struct sk_buff *skb)
1134 {
1135         struct rt6_info *rt;
1136
1137         icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
1138
1139         rt = (struct rt6_info *) skb_dst(skb);
1140         if (rt) {
1141                 if (rt->rt6i_flags & RTF_CACHE) {
1142                         dst_hold(&rt->dst);
1143                         if (ip6_del_rt(rt))
1144                                 dst_free(&rt->dst);
1145                 } else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT)) {
1146                         rt->rt6i_node->fn_sernum = -1;
1147                 }
1148         }
1149 }
1150
1151 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
1152                                struct sk_buff *skb, u32 mtu)
1153 {
1154         struct rt6_info *rt6 = (struct rt6_info *)dst;
1155
1156         dst_confirm(dst);
1157         if (mtu < dst_mtu(dst) && rt6->rt6i_dst.plen == 128) {
1158                 struct net *net = dev_net(dst->dev);
1159
1160                 rt6->rt6i_flags |= RTF_MODIFIED;
1161                 if (mtu < IPV6_MIN_MTU)
1162                         mtu = IPV6_MIN_MTU;
1163
1164                 dst_metric_set(dst, RTAX_MTU, mtu);
1165                 rt6_update_expires(rt6, net->ipv6.sysctl.ip6_rt_mtu_expires);
1166         }
1167 }
1168
1169 void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
1170                      int oif, u32 mark)
1171 {
1172         const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1173         struct dst_entry *dst;
1174         struct flowi6 fl6;
1175
1176         memset(&fl6, 0, sizeof(fl6));
1177         fl6.flowi6_oif = oif;
1178         fl6.flowi6_mark = mark ? mark : IP6_REPLY_MARK(net, skb->mark);
1179         fl6.daddr = iph->daddr;
1180         fl6.saddr = iph->saddr;
1181         fl6.flowlabel = ip6_flowinfo(iph);
1182
1183         dst = ip6_route_output(net, NULL, &fl6);
1184         if (!dst->error)
1185                 ip6_rt_update_pmtu(dst, NULL, skb, ntohl(mtu));
1186         dst_release(dst);
1187 }
1188 EXPORT_SYMBOL_GPL(ip6_update_pmtu);
1189
1190 void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
1191 {
1192         ip6_update_pmtu(skb, sock_net(sk), mtu,
1193                         sk->sk_bound_dev_if, sk->sk_mark);
1194 }
1195 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
1196
1197 /* Handle redirects */
1198 struct ip6rd_flowi {
1199         struct flowi6 fl6;
1200         struct in6_addr gateway;
1201 };
1202
1203 static struct rt6_info *__ip6_route_redirect(struct net *net,
1204                                              struct fib6_table *table,
1205                                              struct flowi6 *fl6,
1206                                              int flags)
1207 {
1208         struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
1209         struct rt6_info *rt;
1210         struct fib6_node *fn;
1211
1212         /* Get the "current" route for this destination and
1213          * check if the redirect has come from approriate router.
1214          *
1215          * RFC 4861 specifies that redirects should only be
1216          * accepted if they come from the nexthop to the target.
1217          * Due to the way the routes are chosen, this notion
1218          * is a bit fuzzy and one might need to check all possible
1219          * routes.
1220          */
1221
1222         read_lock_bh(&table->tb6_lock);
1223         fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1224 restart:
1225         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1226                 if (rt6_check_expired(rt))
1227                         continue;
1228                 if (rt->dst.error)
1229                         break;
1230                 if (!(rt->rt6i_flags & RTF_GATEWAY))
1231                         continue;
1232                 if (fl6->flowi6_oif != rt->dst.dev->ifindex)
1233                         continue;
1234                 if (!ipv6_addr_equal(&rdfl->gateway, &rt->rt6i_gateway))
1235                         continue;
1236                 break;
1237         }
1238
1239         if (!rt)
1240                 rt = net->ipv6.ip6_null_entry;
1241         else if (rt->dst.error) {
1242                 rt = net->ipv6.ip6_null_entry;
1243                 goto out;
1244         }
1245
1246         if (rt == net->ipv6.ip6_null_entry) {
1247                 fn = fib6_backtrack(fn, &fl6->saddr);
1248                 if (fn)
1249                         goto restart;
1250         }
1251
1252 out:
1253         dst_hold(&rt->dst);
1254
1255         read_unlock_bh(&table->tb6_lock);
1256
1257         return rt;
1258 };
1259
1260 static struct dst_entry *ip6_route_redirect(struct net *net,
1261                                         const struct flowi6 *fl6,
1262                                         const struct in6_addr *gateway)
1263 {
1264         int flags = RT6_LOOKUP_F_HAS_SADDR;
1265         struct ip6rd_flowi rdfl;
1266
1267         rdfl.fl6 = *fl6;
1268         rdfl.gateway = *gateway;
1269
1270         return fib6_rule_lookup(net, &rdfl.fl6,
1271                                 flags, __ip6_route_redirect);
1272 }
1273
1274 void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark)
1275 {
1276         const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1277         struct dst_entry *dst;
1278         struct flowi6 fl6;
1279
1280         memset(&fl6, 0, sizeof(fl6));
1281         fl6.flowi6_iif = LOOPBACK_IFINDEX;
1282         fl6.flowi6_oif = oif;
1283         fl6.flowi6_mark = mark;
1284         fl6.daddr = iph->daddr;
1285         fl6.saddr = iph->saddr;
1286         fl6.flowlabel = ip6_flowinfo(iph);
1287
1288         dst = ip6_route_redirect(net, &fl6, &ipv6_hdr(skb)->saddr);
1289         rt6_do_redirect(dst, NULL, skb);
1290         dst_release(dst);
1291 }
1292 EXPORT_SYMBOL_GPL(ip6_redirect);
1293
1294 void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif,
1295                             u32 mark)
1296 {
1297         const struct ipv6hdr *iph = ipv6_hdr(skb);
1298         const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
1299         struct dst_entry *dst;
1300         struct flowi6 fl6;
1301
1302         memset(&fl6, 0, sizeof(fl6));
1303         fl6.flowi6_iif = LOOPBACK_IFINDEX;
1304         fl6.flowi6_oif = oif;
1305         fl6.flowi6_mark = mark;
1306         fl6.daddr = msg->dest;
1307         fl6.saddr = iph->daddr;
1308
1309         dst = ip6_route_redirect(net, &fl6, &iph->saddr);
1310         rt6_do_redirect(dst, NULL, skb);
1311         dst_release(dst);
1312 }
1313
1314 void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
1315 {
1316         ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark);
1317 }
1318 EXPORT_SYMBOL_GPL(ip6_sk_redirect);
1319
1320 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1321 {
1322         struct net_device *dev = dst->dev;
1323         unsigned int mtu = dst_mtu(dst);
1324         struct net *net = dev_net(dev);
1325
1326         mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
1327
1328         if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
1329                 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1330
1331         /*
1332          * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1333          * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1334          * IPV6_MAXPLEN is also valid and means: "any MSS,
1335          * rely only on pmtu discovery"
1336          */
1337         if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
1338                 mtu = IPV6_MAXPLEN;
1339         return mtu;
1340 }
1341
1342 static unsigned int ip6_mtu(const struct dst_entry *dst)
1343 {
1344         struct inet6_dev *idev;
1345         unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
1346
1347         if (mtu)
1348                 goto out;
1349
1350         mtu = IPV6_MIN_MTU;
1351
1352         rcu_read_lock();
1353         idev = __in6_dev_get(dst->dev);
1354         if (idev)
1355                 mtu = idev->cnf.mtu6;
1356         rcu_read_unlock();
1357
1358 out:
1359         return min_t(unsigned int, mtu, IP6_MAX_MTU);
1360 }
1361
1362 static struct dst_entry *icmp6_dst_gc_list;
1363 static DEFINE_SPINLOCK(icmp6_dst_lock);
1364
1365 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
1366                                   struct flowi6 *fl6)
1367 {
1368         struct dst_entry *dst;
1369         struct rt6_info *rt;
1370         struct inet6_dev *idev = in6_dev_get(dev);
1371         struct net *net = dev_net(dev);
1372
1373         if (unlikely(!idev))
1374                 return ERR_PTR(-ENODEV);
1375
1376         rt = ip6_dst_alloc(net, dev, 0, NULL);
1377         if (unlikely(!rt)) {
1378                 in6_dev_put(idev);
1379                 dst = ERR_PTR(-ENOMEM);
1380                 goto out;
1381         }
1382
1383         rt->dst.flags |= DST_HOST;
1384         rt->dst.output  = ip6_output;
1385         atomic_set(&rt->dst.__refcnt, 1);
1386         rt->rt6i_gateway  = fl6->daddr;
1387         rt->rt6i_dst.addr = fl6->daddr;
1388         rt->rt6i_dst.plen = 128;
1389         rt->rt6i_idev     = idev;
1390         dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
1391
1392         spin_lock_bh(&icmp6_dst_lock);
1393         rt->dst.next = icmp6_dst_gc_list;
1394         icmp6_dst_gc_list = &rt->dst;
1395         spin_unlock_bh(&icmp6_dst_lock);
1396
1397         fib6_force_start_gc(net);
1398
1399         dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
1400
1401 out:
1402         return dst;
1403 }
1404
1405 int icmp6_dst_gc(void)
1406 {
1407         struct dst_entry *dst, **pprev;
1408         int more = 0;
1409
1410         spin_lock_bh(&icmp6_dst_lock);
1411         pprev = &icmp6_dst_gc_list;
1412
1413         while ((dst = *pprev) != NULL) {
1414                 if (!atomic_read(&dst->__refcnt)) {
1415                         *pprev = dst->next;
1416                         dst_free(dst);
1417                 } else {
1418                         pprev = &dst->next;
1419                         ++more;
1420                 }
1421         }
1422
1423         spin_unlock_bh(&icmp6_dst_lock);
1424
1425         return more;
1426 }
1427
1428 static void icmp6_clean_all(int (*func)(struct rt6_info *rt, void *arg),
1429                             void *arg)
1430 {
1431         struct dst_entry *dst, **pprev;
1432
1433         spin_lock_bh(&icmp6_dst_lock);
1434         pprev = &icmp6_dst_gc_list;
1435         while ((dst = *pprev) != NULL) {
1436                 struct rt6_info *rt = (struct rt6_info *) dst;
1437                 if (func(rt, arg)) {
1438                         *pprev = dst->next;
1439                         dst_free(dst);
1440                 } else {
1441                         pprev = &dst->next;
1442                 }
1443         }
1444         spin_unlock_bh(&icmp6_dst_lock);
1445 }
1446
1447 static int ip6_dst_gc(struct dst_ops *ops)
1448 {
1449         struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
1450         int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1451         int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1452         int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1453         int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1454         unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1455         int entries;
1456
1457         entries = dst_entries_get_fast(ops);
1458         if (time_after(rt_last_gc + rt_min_interval, jiffies) &&
1459             entries <= rt_max_size)
1460                 goto out;
1461
1462         net->ipv6.ip6_rt_gc_expire++;
1463         fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net, true);
1464         entries = dst_entries_get_slow(ops);
1465         if (entries < ops->gc_thresh)
1466                 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1467 out:
1468         net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1469         return entries > rt_max_size;
1470 }
1471
1472 static int ip6_convert_metrics(struct mx6_config *mxc,
1473                                const struct fib6_config *cfg)
1474 {
1475         struct nlattr *nla;
1476         int remaining;
1477         u32 *mp;
1478
1479         if (!cfg->fc_mx)
1480                 return 0;
1481
1482         mp = kzalloc(sizeof(u32) * RTAX_MAX, GFP_KERNEL);
1483         if (unlikely(!mp))
1484                 return -ENOMEM;
1485
1486         nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1487                 int type = nla_type(nla);
1488
1489                 if (type) {
1490                         u32 val;
1491
1492                         if (unlikely(type > RTAX_MAX))
1493                                 goto err;
1494                         if (type == RTAX_CC_ALGO) {
1495                                 char tmp[TCP_CA_NAME_MAX];
1496
1497                                 nla_strlcpy(tmp, nla, sizeof(tmp));
1498                                 val = tcp_ca_get_key_by_name(tmp);
1499                                 if (val == TCP_CA_UNSPEC)
1500                                         goto err;
1501                         } else {
1502                                 val = nla_get_u32(nla);
1503                         }
1504
1505                         mp[type - 1] = val;
1506                         __set_bit(type - 1, mxc->mx_valid);
1507                 }
1508         }
1509
1510         mxc->mx = mp;
1511
1512         return 0;
1513  err:
1514         kfree(mp);
1515         return -EINVAL;
1516 }
1517
1518 int ip6_route_add(struct fib6_config *cfg)
1519 {
1520         int err;
1521         struct net *net = cfg->fc_nlinfo.nl_net;
1522         struct rt6_info *rt = NULL;
1523         struct net_device *dev = NULL;
1524         struct inet6_dev *idev = NULL;
1525         struct fib6_table *table;
1526         struct mx6_config mxc = { .mx = NULL, };
1527         int addr_type;
1528
1529         if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1530                 return -EINVAL;
1531 #ifndef CONFIG_IPV6_SUBTREES
1532         if (cfg->fc_src_len)
1533                 return -EINVAL;
1534 #endif
1535         if (cfg->fc_ifindex) {
1536                 err = -ENODEV;
1537                 dev = dev_get_by_index(net, cfg->fc_ifindex);
1538                 if (!dev)
1539                         goto out;
1540                 idev = in6_dev_get(dev);
1541                 if (!idev)
1542                         goto out;
1543         }
1544
1545         if (cfg->fc_metric == 0)
1546                 cfg->fc_metric = IP6_RT_PRIO_USER;
1547
1548         err = -ENOBUFS;
1549         if (cfg->fc_nlinfo.nlh &&
1550             !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
1551                 table = fib6_get_table(net, cfg->fc_table);
1552                 if (!table) {
1553                         pr_warn("NLM_F_CREATE should be specified when creating new route\n");
1554                         table = fib6_new_table(net, cfg->fc_table);
1555                 }
1556         } else {
1557                 table = fib6_new_table(net, cfg->fc_table);
1558         }
1559
1560         if (!table)
1561                 goto out;
1562
1563         rt = ip6_dst_alloc(net, NULL, (cfg->fc_flags & RTF_ADDRCONF) ? 0 : DST_NOCOUNT, table);
1564
1565         if (!rt) {
1566                 err = -ENOMEM;
1567                 goto out;
1568         }
1569
1570         if (cfg->fc_flags & RTF_EXPIRES)
1571                 rt6_set_expires(rt, jiffies +
1572                                 clock_t_to_jiffies(cfg->fc_expires));
1573         else
1574                 rt6_clean_expires(rt);
1575
1576         if (cfg->fc_protocol == RTPROT_UNSPEC)
1577                 cfg->fc_protocol = RTPROT_BOOT;
1578         rt->rt6i_protocol = cfg->fc_protocol;
1579
1580         addr_type = ipv6_addr_type(&cfg->fc_dst);
1581
1582         if (addr_type & IPV6_ADDR_MULTICAST)
1583                 rt->dst.input = ip6_mc_input;
1584         else if (cfg->fc_flags & RTF_LOCAL)
1585                 rt->dst.input = ip6_input;
1586         else
1587                 rt->dst.input = ip6_forward;
1588
1589         rt->dst.output = ip6_output;
1590
1591         ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1592         rt->rt6i_dst.plen = cfg->fc_dst_len;
1593         if (rt->rt6i_dst.plen == 128) {
1594                 rt->dst.flags |= DST_HOST;
1595                 dst_metrics_set_force_overwrite(&rt->dst);
1596         }
1597
1598 #ifdef CONFIG_IPV6_SUBTREES
1599         ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1600         rt->rt6i_src.plen = cfg->fc_src_len;
1601 #endif
1602
1603         rt->rt6i_metric = cfg->fc_metric;
1604
1605         /* We cannot add true routes via loopback here,
1606            they would result in kernel looping; promote them to reject routes
1607          */
1608         if ((cfg->fc_flags & RTF_REJECT) ||
1609             (dev && (dev->flags & IFF_LOOPBACK) &&
1610              !(addr_type & IPV6_ADDR_LOOPBACK) &&
1611              !(cfg->fc_flags & RTF_LOCAL))) {
1612                 /* hold loopback dev/idev if we haven't done so. */
1613                 if (dev != net->loopback_dev) {
1614                         if (dev) {
1615                                 dev_put(dev);
1616                                 in6_dev_put(idev);
1617                         }
1618                         dev = net->loopback_dev;
1619                         dev_hold(dev);
1620                         idev = in6_dev_get(dev);
1621                         if (!idev) {
1622                                 err = -ENODEV;
1623                                 goto out;
1624                         }
1625                 }
1626                 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1627                 switch (cfg->fc_type) {
1628                 case RTN_BLACKHOLE:
1629                         rt->dst.error = -EINVAL;
1630                         rt->dst.output = dst_discard_sk;
1631                         rt->dst.input = dst_discard;
1632                         break;
1633                 case RTN_PROHIBIT:
1634                         rt->dst.error = -EACCES;
1635                         rt->dst.output = ip6_pkt_prohibit_out;
1636                         rt->dst.input = ip6_pkt_prohibit;
1637                         break;
1638                 case RTN_THROW:
1639                 default:
1640                         rt->dst.error = (cfg->fc_type == RTN_THROW) ? -EAGAIN
1641                                         : -ENETUNREACH;
1642                         rt->dst.output = ip6_pkt_discard_out;
1643                         rt->dst.input = ip6_pkt_discard;
1644                         break;
1645                 }
1646                 goto install_route;
1647         }
1648
1649         if (cfg->fc_flags & RTF_GATEWAY) {
1650                 const struct in6_addr *gw_addr;
1651                 int gwa_type;
1652
1653                 gw_addr = &cfg->fc_gateway;
1654                 rt->rt6i_gateway = *gw_addr;
1655                 gwa_type = ipv6_addr_type(gw_addr);
1656
1657                 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1658                         struct rt6_info *grt;
1659
1660                         /* IPv6 strictly inhibits using not link-local
1661                            addresses as nexthop address.
1662                            Otherwise, router will not able to send redirects.
1663                            It is very good, but in some (rare!) circumstances
1664                            (SIT, PtP, NBMA NOARP links) it is handy to allow
1665                            some exceptions. --ANK
1666                          */
1667                         err = -EINVAL;
1668                         if (!(gwa_type & IPV6_ADDR_UNICAST))
1669                                 goto out;
1670
1671                         grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, 1);
1672
1673                         err = -EHOSTUNREACH;
1674                         if (!grt)
1675                                 goto out;
1676                         if (dev) {
1677                                 if (dev != grt->dst.dev) {
1678                                         ip6_rt_put(grt);
1679                                         goto out;
1680                                 }
1681                         } else {
1682                                 dev = grt->dst.dev;
1683                                 idev = grt->rt6i_idev;
1684                                 dev_hold(dev);
1685                                 in6_dev_hold(grt->rt6i_idev);
1686                         }
1687                         if (!(grt->rt6i_flags & RTF_GATEWAY))
1688                                 err = 0;
1689                         ip6_rt_put(grt);
1690
1691                         if (err)
1692                                 goto out;
1693                 }
1694                 err = -EINVAL;
1695                 if (!dev || (dev->flags & IFF_LOOPBACK))
1696                         goto out;
1697         }
1698
1699         err = -ENODEV;
1700         if (!dev)
1701                 goto out;
1702
1703         if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
1704                 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
1705                         err = -EINVAL;
1706                         goto out;
1707                 }
1708                 rt->rt6i_prefsrc.addr = cfg->fc_prefsrc;
1709                 rt->rt6i_prefsrc.plen = 128;
1710         } else
1711                 rt->rt6i_prefsrc.plen = 0;
1712
1713         rt->rt6i_flags = cfg->fc_flags;
1714
1715 install_route:
1716         rt->dst.dev = dev;
1717         rt->rt6i_idev = idev;
1718         rt->rt6i_table = table;
1719
1720         cfg->fc_nlinfo.nl_net = dev_net(dev);
1721
1722         err = ip6_convert_metrics(&mxc, cfg);
1723         if (err)
1724                 goto out;
1725
1726         err = __ip6_ins_rt(rt, &cfg->fc_nlinfo, &mxc);
1727
1728         kfree(mxc.mx);
1729         return err;
1730 out:
1731         if (dev)
1732                 dev_put(dev);
1733         if (idev)
1734                 in6_dev_put(idev);
1735         if (rt)
1736                 dst_free(&rt->dst);
1737         return err;
1738 }
1739
1740 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
1741 {
1742         int err;
1743         struct fib6_table *table;
1744         struct net *net = dev_net(rt->dst.dev);
1745
1746         if (rt == net->ipv6.ip6_null_entry) {
1747                 err = -ENOENT;
1748                 goto out;
1749         }
1750
1751         table = rt->rt6i_table;
1752         write_lock_bh(&table->tb6_lock);
1753         err = fib6_del(rt, info);
1754         write_unlock_bh(&table->tb6_lock);
1755
1756 out:
1757         ip6_rt_put(rt);
1758         return err;
1759 }
1760
1761 int ip6_del_rt(struct rt6_info *rt)
1762 {
1763         struct nl_info info = {
1764                 .nl_net = dev_net(rt->dst.dev),
1765         };
1766         return __ip6_del_rt(rt, &info);
1767 }
1768
1769 static int ip6_route_del(struct fib6_config *cfg)
1770 {
1771         struct fib6_table *table;
1772         struct fib6_node *fn;
1773         struct rt6_info *rt;
1774         int err = -ESRCH;
1775
1776         table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
1777         if (!table)
1778                 return err;
1779
1780         read_lock_bh(&table->tb6_lock);
1781
1782         fn = fib6_locate(&table->tb6_root,
1783                          &cfg->fc_dst, cfg->fc_dst_len,
1784                          &cfg->fc_src, cfg->fc_src_len);
1785
1786         if (fn) {
1787                 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1788                         if (cfg->fc_ifindex &&
1789                             (!rt->dst.dev ||
1790                              rt->dst.dev->ifindex != cfg->fc_ifindex))
1791                                 continue;
1792                         if (cfg->fc_flags & RTF_GATEWAY &&
1793                             !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
1794                                 continue;
1795                         if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
1796                                 continue;
1797                         dst_hold(&rt->dst);
1798                         read_unlock_bh(&table->tb6_lock);
1799
1800                         return __ip6_del_rt(rt, &cfg->fc_nlinfo);
1801                 }
1802         }
1803         read_unlock_bh(&table->tb6_lock);
1804
1805         return err;
1806 }
1807
1808 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
1809 {
1810         struct net *net = dev_net(skb->dev);
1811         struct netevent_redirect netevent;
1812         struct rt6_info *rt, *nrt = NULL;
1813         struct ndisc_options ndopts;
1814         struct inet6_dev *in6_dev;
1815         struct neighbour *neigh;
1816         struct rd_msg *msg;
1817         int optlen, on_link;
1818         u8 *lladdr;
1819
1820         optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
1821         optlen -= sizeof(*msg);
1822
1823         if (optlen < 0) {
1824                 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
1825                 return;
1826         }
1827
1828         msg = (struct rd_msg *)icmp6_hdr(skb);
1829
1830         if (ipv6_addr_is_multicast(&msg->dest)) {
1831                 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
1832                 return;
1833         }
1834
1835         on_link = 0;
1836         if (ipv6_addr_equal(&msg->dest, &msg->target)) {
1837                 on_link = 1;
1838         } else if (ipv6_addr_type(&msg->target) !=
1839                    (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
1840                 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
1841                 return;
1842         }
1843
1844         in6_dev = __in6_dev_get(skb->dev);
1845         if (!in6_dev)
1846                 return;
1847         if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
1848                 return;
1849
1850         /* RFC2461 8.1:
1851          *      The IP source address of the Redirect MUST be the same as the current
1852          *      first-hop router for the specified ICMP Destination Address.
1853          */
1854
1855         if (!ndisc_parse_options(msg->opt, optlen, &ndopts)) {
1856                 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
1857                 return;
1858         }
1859
1860         lladdr = NULL;
1861         if (ndopts.nd_opts_tgt_lladdr) {
1862                 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
1863                                              skb->dev);
1864                 if (!lladdr) {
1865                         net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
1866                         return;
1867                 }
1868         }
1869
1870         rt = (struct rt6_info *) dst;
1871         if (rt == net->ipv6.ip6_null_entry) {
1872                 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
1873                 return;
1874         }
1875
1876         /* Redirect received -> path was valid.
1877          * Look, redirects are sent only in response to data packets,
1878          * so that this nexthop apparently is reachable. --ANK
1879          */
1880         dst_confirm(&rt->dst);
1881
1882         neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
1883         if (!neigh)
1884                 return;
1885
1886         /*
1887          *      We have finally decided to accept it.
1888          */
1889
1890         neigh_update(neigh, lladdr, NUD_STALE,
1891                      NEIGH_UPDATE_F_WEAK_OVERRIDE|
1892                      NEIGH_UPDATE_F_OVERRIDE|
1893                      (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1894                                      NEIGH_UPDATE_F_ISROUTER))
1895                      );
1896
1897         nrt = ip6_rt_copy(rt, &msg->dest);
1898         if (!nrt)
1899                 goto out;
1900
1901         nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
1902         if (on_link)
1903                 nrt->rt6i_flags &= ~RTF_GATEWAY;
1904
1905         nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
1906
1907         if (ip6_ins_rt(nrt))
1908                 goto out;
1909
1910         netevent.old = &rt->dst;
1911         netevent.new = &nrt->dst;
1912         netevent.daddr = &msg->dest;
1913         netevent.neigh = neigh;
1914         call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
1915
1916         if (rt->rt6i_flags & RTF_CACHE) {
1917                 rt = (struct rt6_info *) dst_clone(&rt->dst);
1918                 ip6_del_rt(rt);
1919         }
1920
1921 out:
1922         neigh_release(neigh);
1923 }
1924
1925 /*
1926  *      Misc support functions
1927  */
1928
1929 static struct rt6_info *ip6_rt_copy(struct rt6_info *ort,
1930                                     const struct in6_addr *dest)
1931 {
1932         struct net *net = dev_net(ort->dst.dev);
1933         struct rt6_info *rt = ip6_dst_alloc(net, ort->dst.dev, 0,
1934                                             ort->rt6i_table);
1935
1936         if (rt) {
1937                 rt->dst.input = ort->dst.input;
1938                 rt->dst.output = ort->dst.output;
1939                 rt->dst.flags |= DST_HOST;
1940
1941                 rt->rt6i_dst.addr = *dest;
1942                 rt->rt6i_dst.plen = 128;
1943                 dst_copy_metrics(&rt->dst, &ort->dst);
1944                 rt->dst.error = ort->dst.error;
1945                 rt->rt6i_idev = ort->rt6i_idev;
1946                 if (rt->rt6i_idev)
1947                         in6_dev_hold(rt->rt6i_idev);
1948                 rt->dst.lastuse = jiffies;
1949
1950                 if (ort->rt6i_flags & RTF_GATEWAY)
1951                         rt->rt6i_gateway = ort->rt6i_gateway;
1952                 else
1953                         rt->rt6i_gateway = *dest;
1954                 rt->rt6i_flags = ort->rt6i_flags;
1955                 rt6_set_from(rt, ort);
1956                 rt->rt6i_metric = 0;
1957
1958 #ifdef CONFIG_IPV6_SUBTREES
1959                 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1960 #endif
1961                 memcpy(&rt->rt6i_prefsrc, &ort->rt6i_prefsrc, sizeof(struct rt6key));
1962                 rt->rt6i_table = ort->rt6i_table;
1963         }
1964         return rt;
1965 }
1966
1967 #ifdef CONFIG_IPV6_ROUTE_INFO
1968 static struct rt6_info *rt6_get_route_info(struct net *net,
1969                                            const struct in6_addr *prefix, int prefixlen,
1970                                            const struct in6_addr *gwaddr, int ifindex)
1971 {
1972         struct fib6_node *fn;
1973         struct rt6_info *rt = NULL;
1974         struct fib6_table *table;
1975
1976         table = fib6_get_table(net, RT6_TABLE_INFO);
1977         if (!table)
1978                 return NULL;
1979
1980         read_lock_bh(&table->tb6_lock);
1981         fn = fib6_locate(&table->tb6_root, prefix, prefixlen, NULL, 0);
1982         if (!fn)
1983                 goto out;
1984
1985         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1986                 if (rt->dst.dev->ifindex != ifindex)
1987                         continue;
1988                 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
1989                         continue;
1990                 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
1991                         continue;
1992                 dst_hold(&rt->dst);
1993                 break;
1994         }
1995 out:
1996         read_unlock_bh(&table->tb6_lock);
1997         return rt;
1998 }
1999
2000 static struct rt6_info *rt6_add_route_info(struct net *net,
2001                                            const struct in6_addr *prefix, int prefixlen,
2002                                            const struct in6_addr *gwaddr, int ifindex,
2003                                            unsigned int pref)
2004 {
2005         struct fib6_config cfg = {
2006                 .fc_table       = RT6_TABLE_INFO,
2007                 .fc_metric      = IP6_RT_PRIO_USER,
2008                 .fc_ifindex     = ifindex,
2009                 .fc_dst_len     = prefixlen,
2010                 .fc_flags       = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
2011                                   RTF_UP | RTF_PREF(pref),
2012                 .fc_nlinfo.portid = 0,
2013                 .fc_nlinfo.nlh = NULL,
2014                 .fc_nlinfo.nl_net = net,
2015         };
2016
2017         cfg.fc_dst = *prefix;
2018         cfg.fc_gateway = *gwaddr;
2019
2020         /* We should treat it as a default route if prefix length is 0. */
2021         if (!prefixlen)
2022                 cfg.fc_flags |= RTF_DEFAULT;
2023
2024         ip6_route_add(&cfg);
2025
2026         return rt6_get_route_info(net, prefix, prefixlen, gwaddr, ifindex);
2027 }
2028 #endif
2029
2030 struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
2031 {
2032         struct rt6_info *rt;
2033         struct fib6_table *table;
2034
2035         table = fib6_get_table(dev_net(dev), RT6_TABLE_DFLT);
2036         if (!table)
2037                 return NULL;
2038
2039         read_lock_bh(&table->tb6_lock);
2040         for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2041                 if (dev == rt->dst.dev &&
2042                     ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
2043                     ipv6_addr_equal(&rt->rt6i_gateway, addr))
2044                         break;
2045         }
2046         if (rt)
2047                 dst_hold(&rt->dst);
2048         read_unlock_bh(&table->tb6_lock);
2049         return rt;
2050 }
2051
2052 struct rt6_info *rt6_add_dflt_router(const struct in6_addr *gwaddr,
2053                                      struct net_device *dev,
2054                                      unsigned int pref)
2055 {
2056         struct fib6_config cfg = {
2057                 .fc_table       = RT6_TABLE_DFLT,
2058                 .fc_metric      = IP6_RT_PRIO_USER,
2059                 .fc_ifindex     = dev->ifindex,
2060                 .fc_flags       = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
2061                                   RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
2062                 .fc_nlinfo.portid = 0,
2063                 .fc_nlinfo.nlh = NULL,
2064                 .fc_nlinfo.nl_net = dev_net(dev),
2065         };
2066
2067         cfg.fc_gateway = *gwaddr;
2068
2069         ip6_route_add(&cfg);
2070
2071         return rt6_get_dflt_router(gwaddr, dev);
2072 }
2073
2074 void rt6_purge_dflt_routers(struct net *net)
2075 {
2076         struct rt6_info *rt;
2077         struct fib6_table *table;
2078
2079         /* NOTE: Keep consistent with rt6_get_dflt_router */
2080         table = fib6_get_table(net, RT6_TABLE_DFLT);
2081         if (!table)
2082                 return;
2083
2084 restart:
2085         read_lock_bh(&table->tb6_lock);
2086         for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2087                 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
2088                     (!rt->rt6i_idev || rt->rt6i_idev->cnf.accept_ra != 2)) {
2089                         dst_hold(&rt->dst);
2090                         read_unlock_bh(&table->tb6_lock);
2091                         ip6_del_rt(rt);
2092                         goto restart;
2093                 }
2094         }
2095         read_unlock_bh(&table->tb6_lock);
2096 }
2097
2098 static void rtmsg_to_fib6_config(struct net *net,
2099                                  struct in6_rtmsg *rtmsg,
2100                                  struct fib6_config *cfg)
2101 {
2102         memset(cfg, 0, sizeof(*cfg));
2103
2104         cfg->fc_table = RT6_TABLE_MAIN;
2105         cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
2106         cfg->fc_metric = rtmsg->rtmsg_metric;
2107         cfg->fc_expires = rtmsg->rtmsg_info;
2108         cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
2109         cfg->fc_src_len = rtmsg->rtmsg_src_len;
2110         cfg->fc_flags = rtmsg->rtmsg_flags;
2111
2112         cfg->fc_nlinfo.nl_net = net;
2113
2114         cfg->fc_dst = rtmsg->rtmsg_dst;
2115         cfg->fc_src = rtmsg->rtmsg_src;
2116         cfg->fc_gateway = rtmsg->rtmsg_gateway;
2117 }
2118
2119 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
2120 {
2121         struct fib6_config cfg;
2122         struct in6_rtmsg rtmsg;
2123         int err;
2124
2125         switch (cmd) {
2126         case SIOCADDRT:         /* Add a route */
2127         case SIOCDELRT:         /* Delete a route */
2128                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2129                         return -EPERM;
2130                 err = copy_from_user(&rtmsg, arg,
2131                                      sizeof(struct in6_rtmsg));
2132                 if (err)
2133                         return -EFAULT;
2134
2135                 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
2136
2137                 rtnl_lock();
2138                 switch (cmd) {
2139                 case SIOCADDRT:
2140                         err = ip6_route_add(&cfg);
2141                         break;
2142                 case SIOCDELRT:
2143                         err = ip6_route_del(&cfg);
2144                         break;
2145                 default:
2146                         err = -EINVAL;
2147                 }
2148                 rtnl_unlock();
2149
2150                 return err;
2151         }
2152
2153         return -EINVAL;
2154 }
2155
2156 /*
2157  *      Drop the packet on the floor
2158  */
2159
2160 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
2161 {
2162         int type;
2163         struct dst_entry *dst = skb_dst(skb);
2164         switch (ipstats_mib_noroutes) {
2165         case IPSTATS_MIB_INNOROUTES:
2166                 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
2167                 if (type == IPV6_ADDR_ANY) {
2168                         IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2169                                       IPSTATS_MIB_INADDRERRORS);
2170                         break;
2171                 }
2172                 /* FALLTHROUGH */
2173         case IPSTATS_MIB_OUTNOROUTES:
2174                 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2175                               ipstats_mib_noroutes);
2176                 break;
2177         }
2178         icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
2179         kfree_skb(skb);
2180         return 0;
2181 }
2182
2183 static int ip6_pkt_discard(struct sk_buff *skb)
2184 {
2185         return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
2186 }
2187
2188 static int ip6_pkt_discard_out(struct sock *sk, struct sk_buff *skb)
2189 {
2190         skb->dev = skb_dst(skb)->dev;
2191         return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
2192 }
2193
2194 static int ip6_pkt_prohibit(struct sk_buff *skb)
2195 {
2196         return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
2197 }
2198
2199 static int ip6_pkt_prohibit_out(struct sock *sk, struct sk_buff *skb)
2200 {
2201         skb->dev = skb_dst(skb)->dev;
2202         return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
2203 }
2204
2205 /*
2206  *      Allocate a dst for local (unicast / anycast) address.
2207  */
2208
2209 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
2210                                     const struct in6_addr *addr,
2211                                     bool anycast)
2212 {
2213         struct net *net = dev_net(idev->dev);
2214         struct rt6_info *rt = ip6_dst_alloc(net, net->loopback_dev,
2215                                             DST_NOCOUNT, NULL);
2216         if (!rt)
2217                 return ERR_PTR(-ENOMEM);
2218
2219         in6_dev_hold(idev);
2220
2221         rt->dst.flags |= DST_HOST;
2222         rt->dst.input = ip6_input;
2223         rt->dst.output = ip6_output;
2224         rt->rt6i_idev = idev;
2225
2226         rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
2227         if (anycast)
2228                 rt->rt6i_flags |= RTF_ANYCAST;
2229         else
2230                 rt->rt6i_flags |= RTF_LOCAL;
2231
2232         rt->rt6i_gateway  = *addr;
2233         rt->rt6i_dst.addr = *addr;
2234         rt->rt6i_dst.plen = 128;
2235         rt->rt6i_table = fib6_get_table(net, RT6_TABLE_LOCAL);
2236
2237         atomic_set(&rt->dst.__refcnt, 1);
2238
2239         return rt;
2240 }
2241
2242 int ip6_route_get_saddr(struct net *net,
2243                         struct rt6_info *rt,
2244                         const struct in6_addr *daddr,
2245                         unsigned int prefs,
2246                         struct in6_addr *saddr)
2247 {
2248         struct inet6_dev *idev =
2249                 rt ? ip6_dst_idev((struct dst_entry *)rt) : NULL;
2250         int err = 0;
2251         if (rt && rt->rt6i_prefsrc.plen)
2252                 *saddr = rt->rt6i_prefsrc.addr;
2253         else
2254                 err = ipv6_dev_get_saddr(net, idev ? idev->dev : NULL,
2255                                          daddr, prefs, saddr);
2256         return err;
2257 }
2258
2259 /* remove deleted ip from prefsrc entries */
2260 struct arg_dev_net_ip {
2261         struct net_device *dev;
2262         struct net *net;
2263         struct in6_addr *addr;
2264 };
2265
2266 static int fib6_remove_prefsrc(struct rt6_info *rt, void *arg)
2267 {
2268         struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
2269         struct net *net = ((struct arg_dev_net_ip *)arg)->net;
2270         struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
2271
2272         if (((void *)rt->dst.dev == dev || !dev) &&
2273             rt != net->ipv6.ip6_null_entry &&
2274             ipv6_addr_equal(addr, &rt->rt6i_prefsrc.addr)) {
2275                 /* remove prefsrc entry */
2276                 rt->rt6i_prefsrc.plen = 0;
2277         }
2278         return 0;
2279 }
2280
2281 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
2282 {
2283         struct net *net = dev_net(ifp->idev->dev);
2284         struct arg_dev_net_ip adni = {
2285                 .dev = ifp->idev->dev,
2286                 .net = net,
2287                 .addr = &ifp->addr,
2288         };
2289         fib6_clean_all(net, fib6_remove_prefsrc, &adni);
2290 }
2291
2292 #define RTF_RA_ROUTER           (RTF_ADDRCONF | RTF_DEFAULT | RTF_GATEWAY)
2293 #define RTF_CACHE_GATEWAY       (RTF_GATEWAY | RTF_CACHE)
2294
2295 /* Remove routers and update dst entries when gateway turn into host. */
2296 static int fib6_clean_tohost(struct rt6_info *rt, void *arg)
2297 {
2298         struct in6_addr *gateway = (struct in6_addr *)arg;
2299
2300         if ((((rt->rt6i_flags & RTF_RA_ROUTER) == RTF_RA_ROUTER) ||
2301              ((rt->rt6i_flags & RTF_CACHE_GATEWAY) == RTF_CACHE_GATEWAY)) &&
2302              ipv6_addr_equal(gateway, &rt->rt6i_gateway)) {
2303                 return -1;
2304         }
2305         return 0;
2306 }
2307
2308 void rt6_clean_tohost(struct net *net, struct in6_addr *gateway)
2309 {
2310         fib6_clean_all(net, fib6_clean_tohost, gateway);
2311 }
2312
2313 struct arg_dev_net {
2314         struct net_device *dev;
2315         struct net *net;
2316 };
2317
2318 static int fib6_ifdown(struct rt6_info *rt, void *arg)
2319 {
2320         const struct arg_dev_net *adn = arg;
2321         const struct net_device *dev = adn->dev;
2322
2323         if ((rt->dst.dev == dev || !dev) &&
2324             rt != adn->net->ipv6.ip6_null_entry)
2325                 return -1;
2326
2327         return 0;
2328 }
2329
2330 void rt6_ifdown(struct net *net, struct net_device *dev)
2331 {
2332         struct arg_dev_net adn = {
2333                 .dev = dev,
2334                 .net = net,
2335         };
2336
2337         fib6_clean_all(net, fib6_ifdown, &adn);
2338         icmp6_clean_all(fib6_ifdown, &adn);
2339 }
2340
2341 struct rt6_mtu_change_arg {
2342         struct net_device *dev;
2343         unsigned int mtu;
2344 };
2345
2346 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2347 {
2348         struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2349         struct inet6_dev *idev;
2350
2351         /* In IPv6 pmtu discovery is not optional,
2352            so that RTAX_MTU lock cannot disable it.
2353            We still use this lock to block changes
2354            caused by addrconf/ndisc.
2355         */
2356
2357         idev = __in6_dev_get(arg->dev);
2358         if (!idev)
2359                 return 0;
2360
2361         /* For administrative MTU increase, there is no way to discover
2362            IPv6 PMTU increase, so PMTU increase should be updated here.
2363            Since RFC 1981 doesn't include administrative MTU increase
2364            update PMTU increase is a MUST. (i.e. jumbo frame)
2365          */
2366         /*
2367            If new MTU is less than route PMTU, this new MTU will be the
2368            lowest MTU in the path, update the route PMTU to reflect PMTU
2369            decreases; if new MTU is greater than route PMTU, and the
2370            old MTU is the lowest MTU in the path, update the route PMTU
2371            to reflect the increase. In this case if the other nodes' MTU
2372            also have the lowest MTU, TOO BIG MESSAGE will be lead to
2373            PMTU discouvery.
2374          */
2375         if (rt->dst.dev == arg->dev &&
2376             !dst_metric_locked(&rt->dst, RTAX_MTU) &&
2377             (dst_mtu(&rt->dst) >= arg->mtu ||
2378              (dst_mtu(&rt->dst) < arg->mtu &&
2379               dst_mtu(&rt->dst) == idev->cnf.mtu6))) {
2380                 dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
2381         }
2382         return 0;
2383 }
2384
2385 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
2386 {
2387         struct rt6_mtu_change_arg arg = {
2388                 .dev = dev,
2389                 .mtu = mtu,
2390         };
2391
2392         fib6_clean_all(dev_net(dev), rt6_mtu_change_route, &arg);
2393 }
2394
2395 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2396         [RTA_GATEWAY]           = { .len = sizeof(struct in6_addr) },
2397         [RTA_OIF]               = { .type = NLA_U32 },
2398         [RTA_IIF]               = { .type = NLA_U32 },
2399         [RTA_PRIORITY]          = { .type = NLA_U32 },
2400         [RTA_METRICS]           = { .type = NLA_NESTED },
2401         [RTA_MULTIPATH]         = { .len = sizeof(struct rtnexthop) },
2402         [RTA_PREF]              = { .type = NLA_U8 },
2403 };
2404
2405 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2406                               struct fib6_config *cfg)
2407 {
2408         struct rtmsg *rtm;
2409         struct nlattr *tb[RTA_MAX+1];
2410         unsigned int pref;
2411         int err;
2412
2413         err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2414         if (err < 0)
2415                 goto errout;
2416
2417         err = -EINVAL;
2418         rtm = nlmsg_data(nlh);
2419         memset(cfg, 0, sizeof(*cfg));
2420
2421         cfg->fc_table = rtm->rtm_table;
2422         cfg->fc_dst_len = rtm->rtm_dst_len;
2423         cfg->fc_src_len = rtm->rtm_src_len;
2424         cfg->fc_flags = RTF_UP;
2425         cfg->fc_protocol = rtm->rtm_protocol;
2426         cfg->fc_type = rtm->rtm_type;
2427
2428         if (rtm->rtm_type == RTN_UNREACHABLE ||
2429             rtm->rtm_type == RTN_BLACKHOLE ||
2430             rtm->rtm_type == RTN_PROHIBIT ||
2431             rtm->rtm_type == RTN_THROW)
2432                 cfg->fc_flags |= RTF_REJECT;
2433
2434         if (rtm->rtm_type == RTN_LOCAL)
2435                 cfg->fc_flags |= RTF_LOCAL;
2436
2437         cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid;
2438         cfg->fc_nlinfo.nlh = nlh;
2439         cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2440
2441         if (tb[RTA_GATEWAY]) {
2442                 cfg->fc_gateway = nla_get_in6_addr(tb[RTA_GATEWAY]);
2443                 cfg->fc_flags |= RTF_GATEWAY;
2444         }
2445
2446         if (tb[RTA_DST]) {
2447                 int plen = (rtm->rtm_dst_len + 7) >> 3;
2448
2449                 if (nla_len(tb[RTA_DST]) < plen)
2450                         goto errout;
2451
2452                 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2453         }
2454
2455         if (tb[RTA_SRC]) {
2456                 int plen = (rtm->rtm_src_len + 7) >> 3;
2457
2458                 if (nla_len(tb[RTA_SRC]) < plen)
2459                         goto errout;
2460
2461                 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2462         }
2463
2464         if (tb[RTA_PREFSRC])
2465                 cfg->fc_prefsrc = nla_get_in6_addr(tb[RTA_PREFSRC]);
2466
2467         if (tb[RTA_OIF])
2468                 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2469
2470         if (tb[RTA_PRIORITY])
2471                 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2472
2473         if (tb[RTA_METRICS]) {
2474                 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2475                 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2476         }
2477
2478         if (tb[RTA_TABLE])
2479                 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2480
2481         if (tb[RTA_MULTIPATH]) {
2482                 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
2483                 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
2484         }
2485
2486         if (tb[RTA_PREF]) {
2487                 pref = nla_get_u8(tb[RTA_PREF]);
2488                 if (pref != ICMPV6_ROUTER_PREF_LOW &&
2489                     pref != ICMPV6_ROUTER_PREF_HIGH)
2490                         pref = ICMPV6_ROUTER_PREF_MEDIUM;
2491                 cfg->fc_flags |= RTF_PREF(pref);
2492         }
2493
2494         err = 0;
2495 errout:
2496         return err;
2497 }
2498
2499 static int ip6_route_multipath(struct fib6_config *cfg, int add)
2500 {
2501         struct fib6_config r_cfg;
2502         struct rtnexthop *rtnh;
2503         int remaining;
2504         int attrlen;
2505         int err = 0, last_err = 0;
2506
2507         remaining = cfg->fc_mp_len;
2508 beginning:
2509         rtnh = (struct rtnexthop *)cfg->fc_mp;
2510
2511         /* Parse a Multipath Entry */
2512         while (rtnh_ok(rtnh, remaining)) {
2513                 memcpy(&r_cfg, cfg, sizeof(*cfg));
2514                 if (rtnh->rtnh_ifindex)
2515                         r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
2516
2517                 attrlen = rtnh_attrlen(rtnh);
2518                 if (attrlen > 0) {
2519                         struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
2520
2521                         nla = nla_find(attrs, attrlen, RTA_GATEWAY);
2522                         if (nla) {
2523                                 r_cfg.fc_gateway = nla_get_in6_addr(nla);
2524                                 r_cfg.fc_flags |= RTF_GATEWAY;
2525                         }
2526                 }
2527                 err = add ? ip6_route_add(&r_cfg) : ip6_route_del(&r_cfg);
2528                 if (err) {
2529                         last_err = err;
2530                         /* If we are trying to remove a route, do not stop the
2531                          * loop when ip6_route_del() fails (because next hop is
2532                          * already gone), we should try to remove all next hops.
2533                          */
2534                         if (add) {
2535                                 /* If add fails, we should try to delete all
2536                                  * next hops that have been already added.
2537                                  */
2538                                 add = 0;
2539                                 remaining = cfg->fc_mp_len - remaining;
2540                                 goto beginning;
2541                         }
2542                 }
2543                 /* Because each route is added like a single route we remove
2544                  * these flags after the first nexthop: if there is a collision,
2545                  * we have already failed to add the first nexthop:
2546                  * fib6_add_rt2node() has rejected it; when replacing, old
2547                  * nexthops have been replaced by first new, the rest should
2548                  * be added to it.
2549                  */
2550                 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~(NLM_F_EXCL |
2551                                                      NLM_F_REPLACE);
2552                 rtnh = rtnh_next(rtnh, &remaining);
2553         }
2554
2555         return last_err;
2556 }
2557
2558 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh)
2559 {
2560         struct fib6_config cfg;
2561         int err;
2562
2563         err = rtm_to_fib6_config(skb, nlh, &cfg);
2564         if (err < 0)
2565                 return err;
2566
2567         if (cfg.fc_mp)
2568                 return ip6_route_multipath(&cfg, 0);
2569         else
2570                 return ip6_route_del(&cfg);
2571 }
2572
2573 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh)
2574 {
2575         struct fib6_config cfg;
2576         int err;
2577
2578         err = rtm_to_fib6_config(skb, nlh, &cfg);
2579         if (err < 0)
2580                 return err;
2581
2582         if (cfg.fc_mp)
2583                 return ip6_route_multipath(&cfg, 1);
2584         else
2585                 return ip6_route_add(&cfg);
2586 }
2587
2588 static inline size_t rt6_nlmsg_size(void)
2589 {
2590         return NLMSG_ALIGN(sizeof(struct rtmsg))
2591                + nla_total_size(16) /* RTA_SRC */
2592                + nla_total_size(16) /* RTA_DST */
2593                + nla_total_size(16) /* RTA_GATEWAY */
2594                + nla_total_size(16) /* RTA_PREFSRC */
2595                + nla_total_size(4) /* RTA_TABLE */
2596                + nla_total_size(4) /* RTA_IIF */
2597                + nla_total_size(4) /* RTA_OIF */
2598                + nla_total_size(4) /* RTA_PRIORITY */
2599                + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
2600                + nla_total_size(sizeof(struct rta_cacheinfo))
2601                + nla_total_size(TCP_CA_NAME_MAX) /* RTAX_CC_ALGO */
2602                + nla_total_size(1); /* RTA_PREF */
2603 }
2604
2605 static int rt6_fill_node(struct net *net,
2606                          struct sk_buff *skb, struct rt6_info *rt,
2607                          struct in6_addr *dst, struct in6_addr *src,
2608                          int iif, int type, u32 portid, u32 seq,
2609                          int prefix, int nowait, unsigned int flags)
2610 {
2611         struct rtmsg *rtm;
2612         struct nlmsghdr *nlh;
2613         long expires;
2614         u32 table;
2615
2616         if (prefix) {   /* user wants prefix routes only */
2617                 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
2618                         /* success since this is not a prefix route */
2619                         return 1;
2620                 }
2621         }
2622
2623         nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
2624         if (!nlh)
2625                 return -EMSGSIZE;
2626
2627         rtm = nlmsg_data(nlh);
2628         rtm->rtm_family = AF_INET6;
2629         rtm->rtm_dst_len = rt->rt6i_dst.plen;
2630         rtm->rtm_src_len = rt->rt6i_src.plen;
2631         rtm->rtm_tos = 0;
2632         if (rt->rt6i_table)
2633                 table = rt->rt6i_table->tb6_id;
2634         else
2635                 table = RT6_TABLE_UNSPEC;
2636         rtm->rtm_table = table;
2637         if (nla_put_u32(skb, RTA_TABLE, table))
2638                 goto nla_put_failure;
2639         if (rt->rt6i_flags & RTF_REJECT) {
2640                 switch (rt->dst.error) {
2641                 case -EINVAL:
2642                         rtm->rtm_type = RTN_BLACKHOLE;
2643                         break;
2644                 case -EACCES:
2645                         rtm->rtm_type = RTN_PROHIBIT;
2646                         break;
2647                 case -EAGAIN:
2648                         rtm->rtm_type = RTN_THROW;
2649                         break;
2650                 default:
2651                         rtm->rtm_type = RTN_UNREACHABLE;
2652                         break;
2653                 }
2654         }
2655         else if (rt->rt6i_flags & RTF_LOCAL)
2656                 rtm->rtm_type = RTN_LOCAL;
2657         else if (rt->dst.dev && (rt->dst.dev->flags & IFF_LOOPBACK))
2658                 rtm->rtm_type = RTN_LOCAL;
2659         else
2660                 rtm->rtm_type = RTN_UNICAST;
2661         rtm->rtm_flags = 0;
2662         rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2663         rtm->rtm_protocol = rt->rt6i_protocol;
2664         if (rt->rt6i_flags & RTF_DYNAMIC)
2665                 rtm->rtm_protocol = RTPROT_REDIRECT;
2666         else if (rt->rt6i_flags & RTF_ADDRCONF) {
2667                 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ROUTEINFO))
2668                         rtm->rtm_protocol = RTPROT_RA;
2669                 else
2670                         rtm->rtm_protocol = RTPROT_KERNEL;
2671         }
2672
2673         if (rt->rt6i_flags & RTF_CACHE)
2674                 rtm->rtm_flags |= RTM_F_CLONED;
2675
2676         if (dst) {
2677                 if (nla_put_in6_addr(skb, RTA_DST, dst))
2678                         goto nla_put_failure;
2679                 rtm->rtm_dst_len = 128;
2680         } else if (rtm->rtm_dst_len)
2681                 if (nla_put_in6_addr(skb, RTA_DST, &rt->rt6i_dst.addr))
2682                         goto nla_put_failure;
2683 #ifdef CONFIG_IPV6_SUBTREES
2684         if (src) {
2685                 if (nla_put_in6_addr(skb, RTA_SRC, src))
2686                         goto nla_put_failure;
2687                 rtm->rtm_src_len = 128;
2688         } else if (rtm->rtm_src_len &&
2689                    nla_put_in6_addr(skb, RTA_SRC, &rt->rt6i_src.addr))
2690                 goto nla_put_failure;
2691 #endif
2692         if (iif) {
2693 #ifdef CONFIG_IPV6_MROUTE
2694                 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
2695                         int err = ip6mr_get_route(net, skb, rtm, nowait);
2696                         if (err <= 0) {
2697                                 if (!nowait) {
2698                                         if (err == 0)
2699                                                 return 0;
2700                                         goto nla_put_failure;
2701                                 } else {
2702                                         if (err == -EMSGSIZE)
2703                                                 goto nla_put_failure;
2704                                 }
2705                         }
2706                 } else
2707 #endif
2708                         if (nla_put_u32(skb, RTA_IIF, iif))
2709                                 goto nla_put_failure;
2710         } else if (dst) {
2711                 struct in6_addr saddr_buf;
2712                 if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0 &&
2713                     nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
2714                         goto nla_put_failure;
2715         }
2716
2717         if (rt->rt6i_prefsrc.plen) {
2718                 struct in6_addr saddr_buf;
2719                 saddr_buf = rt->rt6i_prefsrc.addr;
2720                 if (nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
2721                         goto nla_put_failure;
2722         }
2723
2724         if (rtnetlink_put_metrics(skb, dst_metrics_ptr(&rt->dst)) < 0)
2725                 goto nla_put_failure;
2726
2727         if (rt->rt6i_flags & RTF_GATEWAY) {
2728                 if (nla_put_in6_addr(skb, RTA_GATEWAY, &rt->rt6i_gateway) < 0)
2729                         goto nla_put_failure;
2730         }
2731
2732         if (rt->dst.dev &&
2733             nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex))
2734                 goto nla_put_failure;
2735         if (nla_put_u32(skb, RTA_PRIORITY, rt->rt6i_metric))
2736                 goto nla_put_failure;
2737
2738         expires = (rt->rt6i_flags & RTF_EXPIRES) ? rt->dst.expires - jiffies : 0;
2739
2740         if (rtnl_put_cacheinfo(skb, &rt->dst, 0, expires, rt->dst.error) < 0)
2741                 goto nla_put_failure;
2742
2743         if (nla_put_u8(skb, RTA_PREF, IPV6_EXTRACT_PREF(rt->rt6i_flags)))
2744                 goto nla_put_failure;
2745
2746         nlmsg_end(skb, nlh);
2747         return 0;
2748
2749 nla_put_failure:
2750         nlmsg_cancel(skb, nlh);
2751         return -EMSGSIZE;
2752 }
2753
2754 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
2755 {
2756         struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
2757         int prefix;
2758
2759         if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
2760                 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
2761                 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
2762         } else
2763                 prefix = 0;
2764
2765         return rt6_fill_node(arg->net,
2766                      arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
2767                      NETLINK_CB(arg->cb->skb).portid, arg->cb->nlh->nlmsg_seq,
2768                      prefix, 0, NLM_F_MULTI);
2769 }
2770
2771 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh)
2772 {
2773         struct net *net = sock_net(in_skb->sk);
2774         struct nlattr *tb[RTA_MAX+1];
2775         struct rt6_info *rt;
2776         struct sk_buff *skb;
2777         struct rtmsg *rtm;
2778         struct flowi6 fl6;
2779         int err, iif = 0, oif = 0;
2780
2781         err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2782         if (err < 0)
2783                 goto errout;
2784
2785         err = -EINVAL;
2786         memset(&fl6, 0, sizeof(fl6));
2787
2788         if (tb[RTA_SRC]) {
2789                 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
2790                         goto errout;
2791
2792                 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
2793         }
2794
2795         if (tb[RTA_DST]) {
2796                 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
2797                         goto errout;
2798
2799                 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
2800         }
2801
2802         if (tb[RTA_IIF])
2803                 iif = nla_get_u32(tb[RTA_IIF]);
2804
2805         if (tb[RTA_OIF])
2806                 oif = nla_get_u32(tb[RTA_OIF]);
2807
2808         if (tb[RTA_MARK])
2809                 fl6.flowi6_mark = nla_get_u32(tb[RTA_MARK]);
2810
2811         if (iif) {
2812                 struct net_device *dev;
2813                 int flags = 0;
2814
2815                 dev = __dev_get_by_index(net, iif);
2816                 if (!dev) {
2817                         err = -ENODEV;
2818                         goto errout;
2819                 }
2820
2821                 fl6.flowi6_iif = iif;
2822
2823                 if (!ipv6_addr_any(&fl6.saddr))
2824                         flags |= RT6_LOOKUP_F_HAS_SADDR;
2825
2826                 rt = (struct rt6_info *)ip6_route_input_lookup(net, dev, &fl6,
2827                                                                flags);
2828         } else {
2829                 fl6.flowi6_oif = oif;
2830
2831                 rt = (struct rt6_info *)ip6_route_output(net, NULL, &fl6);
2832         }
2833
2834         skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
2835         if (!skb) {
2836                 ip6_rt_put(rt);
2837                 err = -ENOBUFS;
2838                 goto errout;
2839         }
2840
2841         /* Reserve room for dummy headers, this skb can pass
2842            through good chunk of routing engine.
2843          */
2844         skb_reset_mac_header(skb);
2845         skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
2846
2847         skb_dst_set(skb, &rt->dst);
2848
2849         err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
2850                             RTM_NEWROUTE, NETLINK_CB(in_skb).portid,
2851                             nlh->nlmsg_seq, 0, 0, 0);
2852         if (err < 0) {
2853                 kfree_skb(skb);
2854                 goto errout;
2855         }
2856
2857         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
2858 errout:
2859         return err;
2860 }
2861
2862 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info)
2863 {
2864         struct sk_buff *skb;
2865         struct net *net = info->nl_net;
2866         u32 seq;
2867         int err;
2868
2869         err = -ENOBUFS;
2870         seq = info->nlh ? info->nlh->nlmsg_seq : 0;
2871
2872         skb = nlmsg_new(rt6_nlmsg_size(), gfp_any());
2873         if (!skb)
2874                 goto errout;
2875
2876         err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
2877                                 event, info->portid, seq, 0, 0, 0);
2878         if (err < 0) {
2879                 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
2880                 WARN_ON(err == -EMSGSIZE);
2881                 kfree_skb(skb);
2882                 goto errout;
2883         }
2884         rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
2885                     info->nlh, gfp_any());
2886         return;
2887 errout:
2888         if (err < 0)
2889                 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
2890 }
2891
2892 static int ip6_route_dev_notify(struct notifier_block *this,
2893                                 unsigned long event, void *ptr)
2894 {
2895         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2896         struct net *net = dev_net(dev);
2897
2898         if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
2899                 net->ipv6.ip6_null_entry->dst.dev = dev;
2900                 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
2901 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2902                 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
2903                 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
2904                 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
2905                 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
2906 #endif
2907         }
2908
2909         return NOTIFY_OK;
2910 }
2911
2912 /*
2913  *      /proc
2914  */
2915
2916 #ifdef CONFIG_PROC_FS
2917
2918 static const struct file_operations ipv6_route_proc_fops = {
2919         .owner          = THIS_MODULE,
2920         .open           = ipv6_route_open,
2921         .read           = seq_read,
2922         .llseek         = seq_lseek,
2923         .release        = seq_release_net,
2924 };
2925
2926 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
2927 {
2928         struct net *net = (struct net *)seq->private;
2929         seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
2930                    net->ipv6.rt6_stats->fib_nodes,
2931                    net->ipv6.rt6_stats->fib_route_nodes,
2932                    net->ipv6.rt6_stats->fib_rt_alloc,
2933                    net->ipv6.rt6_stats->fib_rt_entries,
2934                    net->ipv6.rt6_stats->fib_rt_cache,
2935                    dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
2936                    net->ipv6.rt6_stats->fib_discarded_routes);
2937
2938         return 0;
2939 }
2940
2941 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
2942 {
2943         return single_open_net(inode, file, rt6_stats_seq_show);
2944 }
2945
2946 static const struct file_operations rt6_stats_seq_fops = {
2947         .owner   = THIS_MODULE,
2948         .open    = rt6_stats_seq_open,
2949         .read    = seq_read,
2950         .llseek  = seq_lseek,
2951         .release = single_release_net,
2952 };
2953 #endif  /* CONFIG_PROC_FS */
2954
2955 #ifdef CONFIG_SYSCTL
2956
2957 static
2958 int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
2959                               void __user *buffer, size_t *lenp, loff_t *ppos)
2960 {
2961         struct net *net;
2962         int delay;
2963         if (!write)
2964                 return -EINVAL;
2965
2966         net = (struct net *)ctl->extra1;
2967         delay = net->ipv6.sysctl.flush_delay;
2968         proc_dointvec(ctl, write, buffer, lenp, ppos);
2969         fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
2970         return 0;
2971 }
2972
2973 struct ctl_table ipv6_route_table_template[] = {
2974         {
2975                 .procname       =       "flush",
2976                 .data           =       &init_net.ipv6.sysctl.flush_delay,
2977                 .maxlen         =       sizeof(int),
2978                 .mode           =       0200,
2979                 .proc_handler   =       ipv6_sysctl_rtcache_flush
2980         },
2981         {
2982                 .procname       =       "gc_thresh",
2983                 .data           =       &ip6_dst_ops_template.gc_thresh,
2984                 .maxlen         =       sizeof(int),
2985                 .mode           =       0644,
2986                 .proc_handler   =       proc_dointvec,
2987         },
2988         {
2989                 .procname       =       "max_size",
2990                 .data           =       &init_net.ipv6.sysctl.ip6_rt_max_size,
2991                 .maxlen         =       sizeof(int),
2992                 .mode           =       0644,
2993                 .proc_handler   =       proc_dointvec,
2994         },
2995         {
2996                 .procname       =       "gc_min_interval",
2997                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2998                 .maxlen         =       sizeof(int),
2999                 .mode           =       0644,
3000                 .proc_handler   =       proc_dointvec_jiffies,
3001         },
3002         {
3003                 .procname       =       "gc_timeout",
3004                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
3005                 .maxlen         =       sizeof(int),
3006                 .mode           =       0644,
3007                 .proc_handler   =       proc_dointvec_jiffies,
3008         },
3009         {
3010                 .procname       =       "gc_interval",
3011                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_interval,
3012                 .maxlen         =       sizeof(int),
3013                 .mode           =       0644,
3014                 .proc_handler   =       proc_dointvec_jiffies,
3015         },
3016         {
3017                 .procname       =       "gc_elasticity",
3018                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
3019                 .maxlen         =       sizeof(int),
3020                 .mode           =       0644,
3021                 .proc_handler   =       proc_dointvec,
3022         },
3023         {
3024                 .procname       =       "mtu_expires",
3025                 .data           =       &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
3026                 .maxlen         =       sizeof(int),
3027                 .mode           =       0644,
3028                 .proc_handler   =       proc_dointvec_jiffies,
3029         },
3030         {
3031                 .procname       =       "min_adv_mss",
3032                 .data           =       &init_net.ipv6.sysctl.ip6_rt_min_advmss,
3033                 .maxlen         =       sizeof(int),
3034                 .mode           =       0644,
3035                 .proc_handler   =       proc_dointvec,
3036         },
3037         {
3038                 .procname       =       "gc_min_interval_ms",
3039                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
3040                 .maxlen         =       sizeof(int),
3041                 .mode           =       0644,
3042                 .proc_handler   =       proc_dointvec_ms_jiffies,
3043         },
3044         { }
3045 };
3046
3047 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
3048 {
3049         struct ctl_table *table;
3050
3051         table = kmemdup(ipv6_route_table_template,
3052                         sizeof(ipv6_route_table_template),
3053                         GFP_KERNEL);
3054
3055         if (table) {
3056                 table[0].data = &net->ipv6.sysctl.flush_delay;
3057                 table[0].extra1 = net;
3058                 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
3059                 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
3060                 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
3061                 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
3062                 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
3063                 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
3064                 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
3065                 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
3066                 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
3067
3068                 /* Don't export sysctls to unprivileged users */
3069                 if (net->user_ns != &init_user_ns)
3070                         table[0].procname = NULL;
3071         }
3072
3073         return table;
3074 }
3075 #endif
3076
3077 static int __net_init ip6_route_net_init(struct net *net)
3078 {
3079         int ret = -ENOMEM;
3080
3081         memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
3082                sizeof(net->ipv6.ip6_dst_ops));
3083
3084         if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
3085                 goto out_ip6_dst_ops;
3086
3087         net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
3088                                            sizeof(*net->ipv6.ip6_null_entry),
3089                                            GFP_KERNEL);
3090         if (!net->ipv6.ip6_null_entry)
3091                 goto out_ip6_dst_entries;
3092         net->ipv6.ip6_null_entry->dst.path =
3093                 (struct dst_entry *)net->ipv6.ip6_null_entry;
3094         net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3095         dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
3096                          ip6_template_metrics, true);
3097
3098 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3099         net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
3100                                                sizeof(*net->ipv6.ip6_prohibit_entry),
3101                                                GFP_KERNEL);
3102         if (!net->ipv6.ip6_prohibit_entry)
3103                 goto out_ip6_null_entry;
3104         net->ipv6.ip6_prohibit_entry->dst.path =
3105                 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
3106         net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3107         dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
3108                          ip6_template_metrics, true);
3109
3110         net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
3111                                                sizeof(*net->ipv6.ip6_blk_hole_entry),
3112                                                GFP_KERNEL);
3113         if (!net->ipv6.ip6_blk_hole_entry)
3114                 goto out_ip6_prohibit_entry;
3115         net->ipv6.ip6_blk_hole_entry->dst.path =
3116                 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
3117         net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3118         dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
3119                          ip6_template_metrics, true);
3120 #endif
3121
3122         net->ipv6.sysctl.flush_delay = 0;
3123         net->ipv6.sysctl.ip6_rt_max_size = 4096;
3124         net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
3125         net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
3126         net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
3127         net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
3128         net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
3129         net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
3130
3131         net->ipv6.ip6_rt_gc_expire = 30*HZ;
3132
3133         ret = 0;
3134 out:
3135         return ret;
3136
3137 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3138 out_ip6_prohibit_entry:
3139         kfree(net->ipv6.ip6_prohibit_entry);
3140 out_ip6_null_entry:
3141         kfree(net->ipv6.ip6_null_entry);
3142 #endif
3143 out_ip6_dst_entries:
3144         dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3145 out_ip6_dst_ops:
3146         goto out;
3147 }
3148
3149 static void __net_exit ip6_route_net_exit(struct net *net)
3150 {
3151         kfree(net->ipv6.ip6_null_entry);
3152 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3153         kfree(net->ipv6.ip6_prohibit_entry);
3154         kfree(net->ipv6.ip6_blk_hole_entry);
3155 #endif
3156         dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3157 }
3158
3159 static int __net_init ip6_route_net_init_late(struct net *net)
3160 {
3161 #ifdef CONFIG_PROC_FS
3162         proc_create("ipv6_route", 0, net->proc_net, &ipv6_route_proc_fops);
3163         proc_create("rt6_stats", S_IRUGO, net->proc_net, &rt6_stats_seq_fops);
3164 #endif
3165         return 0;
3166 }
3167
3168 static void __net_exit ip6_route_net_exit_late(struct net *net)
3169 {
3170 #ifdef CONFIG_PROC_FS
3171         remove_proc_entry("ipv6_route", net->proc_net);
3172         remove_proc_entry("rt6_stats", net->proc_net);
3173 #endif
3174 }
3175
3176 static struct pernet_operations ip6_route_net_ops = {
3177         .init = ip6_route_net_init,
3178         .exit = ip6_route_net_exit,
3179 };
3180
3181 static int __net_init ipv6_inetpeer_init(struct net *net)
3182 {
3183         struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
3184
3185         if (!bp)
3186                 return -ENOMEM;
3187         inet_peer_base_init(bp);
3188         net->ipv6.peers = bp;
3189         return 0;
3190 }
3191
3192 static void __net_exit ipv6_inetpeer_exit(struct net *net)
3193 {
3194         struct inet_peer_base *bp = net->ipv6.peers;
3195
3196         net->ipv6.peers = NULL;
3197         inetpeer_invalidate_tree(bp);
3198         kfree(bp);
3199 }
3200
3201 static struct pernet_operations ipv6_inetpeer_ops = {
3202         .init   =       ipv6_inetpeer_init,
3203         .exit   =       ipv6_inetpeer_exit,
3204 };
3205
3206 static struct pernet_operations ip6_route_net_late_ops = {
3207         .init = ip6_route_net_init_late,
3208         .exit = ip6_route_net_exit_late,
3209 };
3210
3211 static struct notifier_block ip6_route_dev_notifier = {
3212         .notifier_call = ip6_route_dev_notify,
3213         .priority = 0,
3214 };
3215
3216 int __init ip6_route_init(void)
3217 {
3218         int ret;
3219
3220         ret = -ENOMEM;
3221         ip6_dst_ops_template.kmem_cachep =
3222                 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
3223                                   SLAB_HWCACHE_ALIGN, NULL);
3224         if (!ip6_dst_ops_template.kmem_cachep)
3225                 goto out;
3226
3227         ret = dst_entries_init(&ip6_dst_blackhole_ops);
3228         if (ret)
3229                 goto out_kmem_cache;
3230
3231         ret = register_pernet_subsys(&ipv6_inetpeer_ops);
3232         if (ret)
3233                 goto out_dst_entries;
3234
3235         ret = register_pernet_subsys(&ip6_route_net_ops);
3236         if (ret)
3237                 goto out_register_inetpeer;
3238
3239         ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
3240
3241         /* Registering of the loopback is done before this portion of code,
3242          * the loopback reference in rt6_info will not be taken, do it
3243          * manually for init_net */
3244         init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
3245         init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3246   #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3247         init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
3248         init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3249         init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
3250         init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3251   #endif
3252         ret = fib6_init();
3253         if (ret)
3254                 goto out_register_subsys;
3255
3256         ret = xfrm6_init();
3257         if (ret)
3258                 goto out_fib6_init;
3259
3260         ret = fib6_rules_init();
3261         if (ret)
3262                 goto xfrm6_init;
3263
3264         ret = register_pernet_subsys(&ip6_route_net_late_ops);
3265         if (ret)
3266                 goto fib6_rules_init;
3267
3268         ret = -ENOBUFS;
3269         if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL, NULL) ||
3270             __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL, NULL) ||
3271             __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL, NULL))
3272                 goto out_register_late_subsys;
3273
3274         ret = register_netdevice_notifier(&ip6_route_dev_notifier);
3275         if (ret)
3276                 goto out_register_late_subsys;
3277
3278 out:
3279         return ret;
3280
3281 out_register_late_subsys:
3282         unregister_pernet_subsys(&ip6_route_net_late_ops);
3283 fib6_rules_init:
3284         fib6_rules_cleanup();
3285 xfrm6_init:
3286         xfrm6_fini();
3287 out_fib6_init:
3288         fib6_gc_cleanup();
3289 out_register_subsys:
3290         unregister_pernet_subsys(&ip6_route_net_ops);
3291 out_register_inetpeer:
3292         unregister_pernet_subsys(&ipv6_inetpeer_ops);
3293 out_dst_entries:
3294         dst_entries_destroy(&ip6_dst_blackhole_ops);
3295 out_kmem_cache:
3296         kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3297         goto out;
3298 }
3299
3300 void ip6_route_cleanup(void)
3301 {
3302         unregister_netdevice_notifier(&ip6_route_dev_notifier);
3303         unregister_pernet_subsys(&ip6_route_net_late_ops);
3304         fib6_rules_cleanup();
3305         xfrm6_fini();
3306         fib6_gc_cleanup();
3307         unregister_pernet_subsys(&ipv6_inetpeer_ops);
3308         unregister_pernet_subsys(&ip6_route_net_ops);
3309         dst_entries_destroy(&ip6_dst_blackhole_ops);
3310         kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3311 }