Pileus Git - ~andy/linux/blob - net/ipv6/route.c

   1 /*
   2  *      Linux INET6 implementation
   3  *      FIB front-end.
   4  *
   5  *      Authors:
   6  *      Pedro Roque             <roque@di.fc.ul.pt>
   7  *
   8  *      This program is free software; you can redistribute it and/or
   9  *      modify it under the terms of the GNU General Public License
  10  *      as published by the Free Software Foundation; either version
  11  *      2 of the License, or (at your option) any later version.
  12  */
  13
  14 /*      Changes:
  15  *
  16  *      YOSHIFUJI Hideaki @USAGI
  17  *              reworked default router selection.
  18  *              - respect outgoing interface
  19  *              - select from (probably) reachable routers (i.e.
  20  *              routers in REACHABLE, STALE, DELAY or PROBE states).
  21  *              - always select the same router if it is (probably)
  22  *              reachable.  otherwise, round-robin the list.
  23  *      Ville Nuorvala
  24  *              Fixed routing subtrees.
  25  */
  26
  27 #define pr_fmt(fmt) "IPv6: " fmt
  28
  29 #include <linux/capability.h>
  30 #include <linux/errno.h>
  31 #include <linux/export.h>
  32 #include <linux/types.h>
  33 #include <linux/times.h>
  34 #include <linux/socket.h>
  35 #include <linux/sockios.h>
  36 #include <linux/net.h>
  37 #include <linux/route.h>
  38 #include <linux/netdevice.h>
  39 #include <linux/in6.h>
  40 #include <linux/mroute6.h>
  41 #include <linux/init.h>
  42 #include <linux/if_arp.h>
  43 #include <linux/proc_fs.h>
  44 #include <linux/seq_file.h>
  45 #include <linux/nsproxy.h>
  46 #include <linux/slab.h>
  47 #include <net/net_namespace.h>
  48 #include <net/snmp.h>
  49 #include <net/ipv6.h>
  50 #include <net/ip6_fib.h>
  51 #include <net/ip6_route.h>
  52 #include <net/ndisc.h>
  53 #include <net/addrconf.h>
  54 #include <net/tcp.h>
  55 #include <linux/rtnetlink.h>
  56 #include <net/dst.h>
  57 #include <net/xfrm.h>
  58 #include <net/netevent.h>
  59 #include <net/netlink.h>
  60 #include <net/nexthop.h>
  61
  62 #include <asm/uaccess.h>
  63
  64 #ifdef CONFIG_SYSCTL
  65 #include <linux/sysctl.h>
  66 #endif
  67
  68 enum rt6_nud_state {
  69         RT6_NUD_FAIL_HARD = -3,
  70         RT6_NUD_FAIL_PROBE = -2,
  71         RT6_NUD_FAIL_DO_RR = -1,
  72         RT6_NUD_SUCCEED = 1
  73 };
  74
  75 static struct rt6_info *ip6_rt_copy(struct rt6_info *ort,
  76                                     const struct in6_addr *dest);
  77 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie);
  78 static unsigned int      ip6_default_advmss(const struct dst_entry *dst);
  79 static unsigned int      ip6_mtu(const struct dst_entry *dst);
  80 static struct dst_entry *ip6_negative_advice(struct dst_entry *);
  81 static void             ip6_dst_destroy(struct dst_entry *);
  82 static void             ip6_dst_ifdown(struct dst_entry *,
  83                                        struct net_device *dev, int how);
  84 static int               ip6_dst_gc(struct dst_ops *ops);
  85
  86 static int              ip6_pkt_discard(struct sk_buff *skb);
  87 static int              ip6_pkt_discard_out(struct sk_buff *skb);
  88 static int              ip6_pkt_prohibit(struct sk_buff *skb);
  89 static int              ip6_pkt_prohibit_out(struct sk_buff *skb);
  90 static void             ip6_link_failure(struct sk_buff *skb);
  91 static void             ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
  92                                            struct sk_buff *skb, u32 mtu);
  93 static void             rt6_do_redirect(struct dst_entry *dst, struct sock *sk,
  94                                         struct sk_buff *skb);
  95 static int rt6_score_route(struct rt6_info *rt, int oif, int strict);
  96
  97 #ifdef CONFIG_IPV6_ROUTE_INFO
  98 static struct rt6_info *rt6_add_route_info(struct net *net,
  99                                            const struct in6_addr *prefix, int prefixlen,
 100                                            const struct in6_addr *gwaddr, int ifindex,
 101                                            unsigned int pref);
 102 static struct rt6_info *rt6_get_route_info(struct net *net,
 103                                            const struct in6_addr *prefix, int prefixlen,
 104                                            const struct in6_addr *gwaddr, int ifindex);
 105 #endif
 106
 107 static void rt6_bind_peer(struct rt6_info *rt, int create)
 108 {
 109         struct inet_peer_base *base;
 110         struct inet_peer *peer;
 111
 112         base = inetpeer_base_ptr(rt->_rt6i_peer);
 113         if (!base)
 114                 return;
 115
 116         peer = inet_getpeer_v6(base, &rt->rt6i_dst.addr, create);
 117         if (peer) {
 118                 if (!rt6_set_peer(rt, peer))
 119                         inet_putpeer(peer);
 120         }
 121 }
 122
 123 static struct inet_peer *__rt6_get_peer(struct rt6_info *rt, int create)
 124 {
 125         if (rt6_has_peer(rt))
 126                 return rt6_peer_ptr(rt);
 127
 128         rt6_bind_peer(rt, create);
 129         return (rt6_has_peer(rt) ? rt6_peer_ptr(rt) : NULL);
 130 }
 131
 132 static struct inet_peer *rt6_get_peer_create(struct rt6_info *rt)
 133 {
 134         return __rt6_get_peer(rt, 1);
 135 }
 136
 137 static u32 *ipv6_cow_metrics(struct dst_entry *dst, unsigned long old)
 138 {
 139         struct rt6_info *rt = (struct rt6_info *) dst;
 140         struct inet_peer *peer;
 141         u32 *p = NULL;
 142
 143         if (!(rt->dst.flags & DST_HOST))
 144                 return NULL;
 145
 146         peer = rt6_get_peer_create(rt);
 147         if (peer) {
 148                 u32 *old_p = __DST_METRICS_PTR(old);
 149                 unsigned long prev, new;
 150
 151                 p = peer->metrics;
 152                 if (inet_metrics_new(peer))
 153                         memcpy(p, old_p, sizeof(u32) * RTAX_MAX);
 154
 155                 new = (unsigned long) p;
 156                 prev = cmpxchg(&dst->_metrics, old, new);
 157
 158                 if (prev != old) {
 159                         p = __DST_METRICS_PTR(prev);
 160                         if (prev & DST_METRICS_READ_ONLY)
 161                                 p = NULL;
 162                 }
 163         }
 164         return p;
 165 }
 166
 167 static inline const void *choose_neigh_daddr(struct rt6_info *rt,
 168                                              struct sk_buff *skb,
 169                                              const void *daddr)
 170 {
 171         struct in6_addr *p = &rt->rt6i_gateway;
 172
 173         if (!ipv6_addr_any(p))
 174                 return (const void *) p;
 175         else if (skb)
 176                 return &ipv6_hdr(skb)->daddr;
 177         return daddr;
 178 }
 179
 180 static struct neighbour *ip6_neigh_lookup(const struct dst_entry *dst,
 181                                           struct sk_buff *skb,
 182                                           const void *daddr)
 183 {
 184         struct rt6_info *rt = (struct rt6_info *) dst;
 185         struct neighbour *n;
 186
 187         daddr = choose_neigh_daddr(rt, skb, daddr);
 188         n = __ipv6_neigh_lookup(dst->dev, daddr);
 189         if (n)
 190                 return n;
 191         return neigh_create(&nd_tbl, daddr, dst->dev);
 192 }
 193
 194 static struct dst_ops ip6_dst_ops_template = {
 195         .family                 =       AF_INET6,
 196         .protocol               =       cpu_to_be16(ETH_P_IPV6),
 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         .protocol               =       cpu_to_be16(ETH_P_IPV6),
 239         .destroy                =       ip6_dst_destroy,
 240         .check                  =       ip6_dst_check,
 241         .mtu                    =       ip6_blackhole_mtu,
 242         .default_advmss         =       ip6_default_advmss,
 243         .update_pmtu            =       ip6_rt_blackhole_update_pmtu,
 244         .redirect               =       ip6_rt_blackhole_redirect,
 245         .cow_metrics            =       ip6_rt_blackhole_cow_metrics,
 246         .neigh_lookup           =       ip6_neigh_lookup,
 247 };
 248
 249 static const u32 ip6_template_metrics[RTAX_MAX] = {
 250         [RTAX_HOPLIMIT - 1] = 0,
 251 };
 252
 253 static const struct rt6_info ip6_null_entry_template = {
 254         .dst = {
 255                 .__refcnt       = ATOMIC_INIT(1),
 256                 .__use          = 1,
 257                 .obsolete       = DST_OBSOLETE_FORCE_CHK,
 258                 .error          = -ENETUNREACH,
 259                 .input          = ip6_pkt_discard,
 260                 .output         = ip6_pkt_discard_out,
 261         },
 262         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
 263         .rt6i_protocol  = RTPROT_KERNEL,
 264         .rt6i_metric    = ~(u32) 0,
 265         .rt6i_ref       = ATOMIC_INIT(1),
 266 };
 267
 268 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
 269
 270 static const struct rt6_info ip6_prohibit_entry_template = {
 271         .dst = {
 272                 .__refcnt       = ATOMIC_INIT(1),
 273                 .__use          = 1,
 274                 .obsolete       = DST_OBSOLETE_FORCE_CHK,
 275                 .error          = -EACCES,
 276                 .input          = ip6_pkt_prohibit,
 277                 .output         = ip6_pkt_prohibit_out,
 278         },
 279         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
 280         .rt6i_protocol  = RTPROT_KERNEL,
 281         .rt6i_metric    = ~(u32) 0,
 282         .rt6i_ref       = ATOMIC_INIT(1),
 283 };
 284
 285 static const struct rt6_info ip6_blk_hole_entry_template = {
 286         .dst = {
 287                 .__refcnt       = ATOMIC_INIT(1),
 288                 .__use          = 1,
 289                 .obsolete       = DST_OBSOLETE_FORCE_CHK,
 290                 .error          = -EINVAL,
 291                 .input          = dst_discard,
 292                 .output         = dst_discard,
 293         },
 294         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
 295         .rt6i_protocol  = RTPROT_KERNEL,
 296         .rt6i_metric    = ~(u32) 0,
 297         .rt6i_ref       = ATOMIC_INIT(1),
 298 };
 299
 300 #endif
 301
 302 /* allocate dst with ip6_dst_ops */
 303 static inline struct rt6_info *ip6_dst_alloc(struct net *net,
 304                                              struct net_device *dev,
 305                                              int flags,
 306                                              struct fib6_table *table)
 307 {
 308         struct rt6_info *rt = dst_alloc(&net->ipv6.ip6_dst_ops, dev,
 309                                         0, DST_OBSOLETE_FORCE_CHK, flags);
 310
 311         if (rt) {
 312                 struct dst_entry *dst = &rt->dst;
 313
 314                 memset(dst + 1, 0, sizeof(*rt) - sizeof(*dst));
 315                 rt6_init_peer(rt, table ? &table->tb6_peers : net->ipv6.peers);
 316                 rt->rt6i_genid = rt_genid_ipv6(net);
 317                 INIT_LIST_HEAD(&rt->rt6i_siblings);
 318         }
 319         return rt;
 320 }
 321
 322 static void ip6_dst_destroy(struct dst_entry *dst)
 323 {
 324         struct rt6_info *rt = (struct rt6_info *)dst;
 325         struct inet6_dev *idev = rt->rt6i_idev;
 326         struct dst_entry *from = dst->from;
 327
 328         if (!(rt->dst.flags & DST_HOST))
 329                 dst_destroy_metrics_generic(dst);
 330
 331         if (idev) {
 332                 rt->rt6i_idev = NULL;
 333                 in6_dev_put(idev);
 334         }
 335
 336         dst->from = NULL;
 337         dst_release(from);
 338
 339         if (rt6_has_peer(rt)) {
 340                 struct inet_peer *peer = rt6_peer_ptr(rt);
 341                 inet_putpeer(peer);
 342         }
 343 }
 344
 345 static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
 346                            int how)
 347 {
 348         struct rt6_info *rt = (struct rt6_info *)dst;
 349         struct inet6_dev *idev = rt->rt6i_idev;
 350         struct net_device *loopback_dev =
 351                 dev_net(dev)->loopback_dev;
 352
 353         if (dev != loopback_dev) {
 354                 if (idev && idev->dev == dev) {
 355                         struct inet6_dev *loopback_idev =
 356                                 in6_dev_get(loopback_dev);
 357                         if (loopback_idev) {
 358                                 rt->rt6i_idev = loopback_idev;
 359                                 in6_dev_put(idev);
 360                         }
 361                 }
 362         }
 363 }
 364
 365 static bool rt6_check_expired(const struct rt6_info *rt)
 366 {
 367         if (rt->rt6i_flags & RTF_EXPIRES) {
 368                 if (time_after(jiffies, rt->dst.expires))
 369                         return true;
 370         } else if (rt->dst.from) {
 371                 return rt6_check_expired((struct rt6_info *) rt->dst.from);
 372         }
 373         return false;
 374 }
 375
 376 static bool rt6_need_strict(const struct in6_addr *daddr)
 377 {
 378         return ipv6_addr_type(daddr) &
 379                 (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
 380 }
 381
 382 /* Multipath route selection:
 383  *   Hash based function using packet header and flowlabel.
 384  * Adapted from fib_info_hashfn()
 385  */
 386 static int rt6_info_hash_nhsfn(unsigned int candidate_count,
 387                                const struct flowi6 *fl6)
 388 {
 389         unsigned int val = fl6->flowi6_proto;
 390
 391         val ^= ipv6_addr_hash(&fl6->daddr);
 392         val ^= ipv6_addr_hash(&fl6->saddr);
 393
 394         /* Work only if this not encapsulated */
 395         switch (fl6->flowi6_proto) {
 396         case IPPROTO_UDP:
 397         case IPPROTO_TCP:
 398         case IPPROTO_SCTP:
 399                 val ^= (__force u16)fl6->fl6_sport;
 400                 val ^= (__force u16)fl6->fl6_dport;
 401                 break;
 402
 403         case IPPROTO_ICMPV6:
 404                 val ^= (__force u16)fl6->fl6_icmp_type;
 405                 val ^= (__force u16)fl6->fl6_icmp_code;
 406                 break;
 407         }
 408         /* RFC6438 recommands to use flowlabel */
 409         val ^= (__force u32)fl6->flowlabel;
 410
 411         /* Perhaps, we need to tune, this function? */
 412         val = val ^ (val >> 7) ^ (val >> 12);
 413         return val % candidate_count;
 414 }
 415
 416 static struct rt6_info *rt6_multipath_select(struct rt6_info *match,
 417                                              struct flowi6 *fl6, int oif,
 418                                              int strict)
 419 {
 420         struct rt6_info *sibling, *next_sibling;
 421         int route_choosen;
 422
 423         route_choosen = rt6_info_hash_nhsfn(match->rt6i_nsiblings + 1, fl6);
 424         /* Don't change the route, if route_choosen == 0
 425          * (siblings does not include ourself)
 426          */
 427         if (route_choosen)
 428                 list_for_each_entry_safe(sibling, next_sibling,
 429                                 &match->rt6i_siblings, rt6i_siblings) {
 430                         route_choosen--;
 431                         if (route_choosen == 0) {
 432                                 if (rt6_score_route(sibling, oif, strict) < 0)
 433                                         break;
 434                                 match = sibling;
 435                                 break;
 436                         }
 437                 }
 438         return match;
 439 }
 440
 441 /*
 442  *      Route lookup. Any table->tb6_lock is implied.
 443  */
 444
 445 static inline struct rt6_info *rt6_device_match(struct net *net,
 446                                                     struct rt6_info *rt,
 447                                                     const struct in6_addr *saddr,
 448                                                     int oif,
 449                                                     int flags)
 450 {
 451         struct rt6_info *local = NULL;
 452         struct rt6_info *sprt;
 453
 454         if (!oif && ipv6_addr_any(saddr))
 455                 goto out;
 456
 457         for (sprt = rt; sprt; sprt = sprt->dst.rt6_next) {
 458                 struct net_device *dev = sprt->dst.dev;
 459
 460                 if (oif) {
 461                         if (dev->ifindex == oif)
 462                                 return sprt;
 463                         if (dev->flags & IFF_LOOPBACK) {
 464                                 if (!sprt->rt6i_idev ||
 465                                     sprt->rt6i_idev->dev->ifindex != oif) {
 466                                         if (flags & RT6_LOOKUP_F_IFACE && oif)
 467                                                 continue;
 468                                         if (local && (!oif ||
 469                                                       local->rt6i_idev->dev->ifindex == oif))
 470                                                 continue;
 471                                 }
 472                                 local = sprt;
 473                         }
 474                 } else {
 475                         if (ipv6_chk_addr(net, saddr, dev,
 476                                           flags & RT6_LOOKUP_F_IFACE))
 477                                 return sprt;
 478                 }
 479         }
 480
 481         if (oif) {
 482                 if (local)
 483                         return local;
 484
 485                 if (flags & RT6_LOOKUP_F_IFACE)
 486                         return net->ipv6.ip6_null_entry;
 487         }
 488 out:
 489         return rt;
 490 }
 491
 492 #ifdef CONFIG_IPV6_ROUTER_PREF
 493 struct __rt6_probe_work {
 494         struct work_struct work;
 495         struct in6_addr target;
 496         struct net_device *dev;
 497 };
 498
 499 static void rt6_probe_deferred(struct work_struct *w)
 500 {
 501         struct in6_addr mcaddr;
 502         struct __rt6_probe_work *work =
 503                 container_of(w, struct __rt6_probe_work, work);
 504
 505         addrconf_addr_solict_mult(&work->target, &mcaddr);
 506         ndisc_send_ns(work->dev, NULL, &work->target, &mcaddr, NULL);
 507         dev_put(work->dev);
 508         kfree(w);
 509 }
 510
 511 static void rt6_probe(struct rt6_info *rt)
 512 {
 513         struct neighbour *neigh;
 514         /*
 515          * Okay, this does not seem to be appropriate
 516          * for now, however, we need to check if it
 517          * is really so; aka Router Reachability Probing.
 518          *
 519          * Router Reachability Probe MUST be rate-limited
 520          * to no more than one per minute.
 521          */
 522         if (!rt || !(rt->rt6i_flags & RTF_GATEWAY))
 523                 return;
 524         rcu_read_lock_bh();
 525         neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
 526         if (neigh) {
 527                 write_lock(&neigh->lock);
 528                 if (neigh->nud_state & NUD_VALID)
 529                         goto out;
 530         }
 531
 532         if (!neigh ||
 533             time_after(jiffies, neigh->updated + rt->rt6i_idev->cnf.rtr_probe_interval)) {
 534                 struct __rt6_probe_work *work;
 535
 536                 work = kmalloc(sizeof(*work), GFP_ATOMIC);
 537
 538                 if (neigh && work)
 539                         __neigh_set_probe_once(neigh);
 540
 541                 if (neigh)
 542                         write_unlock(&neigh->lock);
 543
 544                 if (work) {
 545                         INIT_WORK(&work->work, rt6_probe_deferred);
 546                         work->target = rt->rt6i_gateway;
 547                         dev_hold(rt->dst.dev);
 548                         work->dev = rt->dst.dev;
 549                         schedule_work(&work->work);
 550                 }
 551         } else {
 552 out:
 553                 write_unlock(&neigh->lock);
 554         }
 555         rcu_read_unlock_bh();
 556 }
 557 #else
 558 static inline void rt6_probe(struct rt6_info *rt)
 559 {
 560 }
 561 #endif
 562
 563 /*
 564  * Default Router Selection (RFC 2461 6.3.6)
 565  */
 566 static inline int rt6_check_dev(struct rt6_info *rt, int oif)
 567 {
 568         struct net_device *dev = rt->dst.dev;
 569         if (!oif || dev->ifindex == oif)
 570                 return 2;
 571         if ((dev->flags & IFF_LOOPBACK) &&
 572             rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
 573                 return 1;
 574         return 0;
 575 }
 576
 577 static inline enum rt6_nud_state rt6_check_neigh(struct rt6_info *rt)
 578 {
 579         struct neighbour *neigh;
 580         enum rt6_nud_state ret = RT6_NUD_FAIL_HARD;
 581
 582         if (rt->rt6i_flags & RTF_NONEXTHOP ||
 583             !(rt->rt6i_flags & RTF_GATEWAY))
 584                 return RT6_NUD_SUCCEED;
 585
 586         rcu_read_lock_bh();
 587         neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
 588         if (neigh) {
 589                 read_lock(&neigh->lock);
 590                 if (neigh->nud_state & NUD_VALID)
 591                         ret = RT6_NUD_SUCCEED;
 592 #ifdef CONFIG_IPV6_ROUTER_PREF
 593                 else if (!(neigh->nud_state & NUD_FAILED))
 594                         ret = RT6_NUD_SUCCEED;
 595                 else
 596                         ret = RT6_NUD_FAIL_PROBE;
 597 #endif
 598                 read_unlock(&neigh->lock);
 599         } else {
 600                 ret = IS_ENABLED(CONFIG_IPV6_ROUTER_PREF) ?
 601                       RT6_NUD_SUCCEED : RT6_NUD_FAIL_DO_RR;
 602         }
 603         rcu_read_unlock_bh();
 604
 605         return ret;
 606 }
 607
 608 static int rt6_score_route(struct rt6_info *rt, int oif,
 609                            int strict)
 610 {
 611         int m;
 612
 613         m = rt6_check_dev(rt, oif);
 614         if (!m && (strict & RT6_LOOKUP_F_IFACE))
 615                 return RT6_NUD_FAIL_HARD;
 616 #ifdef CONFIG_IPV6_ROUTER_PREF
 617         m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
 618 #endif
 619         if (strict & RT6_LOOKUP_F_REACHABLE) {
 620                 int n = rt6_check_neigh(rt);
 621                 if (n < 0)
 622                         return n;
 623         }
 624         return m;
 625 }
 626
 627 static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
 628                                    int *mpri, struct rt6_info *match,
 629                                    bool *do_rr)
 630 {
 631         int m;
 632         bool match_do_rr = false;
 633
 634         if (rt6_check_expired(rt))
 635                 goto out;
 636
 637         m = rt6_score_route(rt, oif, strict);
 638         if (m == RT6_NUD_FAIL_DO_RR) {
 639                 match_do_rr = true;
 640                 m = 0; /* lowest valid score */
 641         } else if (m == RT6_NUD_FAIL_HARD) {
 642                 goto out;
 643         }
 644
 645         if (strict & RT6_LOOKUP_F_REACHABLE)
 646                 rt6_probe(rt);
 647
 648         /* note that m can be RT6_NUD_FAIL_PROBE at this point */
 649         if (m > *mpri) {
 650                 *do_rr = match_do_rr;
 651                 *mpri = m;
 652                 match = rt;
 653         }
 654 out:
 655         return match;
 656 }
 657
 658 static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
 659                                      struct rt6_info *rr_head,
 660                                      u32 metric, int oif, int strict,
 661                                      bool *do_rr)
 662 {
 663         struct rt6_info *rt, *match;
 664         int mpri = -1;
 665
 666         match = NULL;
 667         for (rt = rr_head; rt && rt->rt6i_metric == metric;
 668              rt = rt->dst.rt6_next)
 669                 match = find_match(rt, oif, strict, &mpri, match, do_rr);
 670         for (rt = fn->leaf; rt && rt != rr_head && rt->rt6i_metric == metric;
 671              rt = rt->dst.rt6_next)
 672                 match = find_match(rt, oif, strict, &mpri, match, do_rr);
 673
 674         return match;
 675 }
 676
 677 static struct rt6_info *rt6_select(struct fib6_node *fn, int oif, int strict)
 678 {
 679         struct rt6_info *match, *rt0;
 680         struct net *net;
 681         bool do_rr = false;
 682
 683         rt0 = fn->rr_ptr;
 684         if (!rt0)
 685                 fn->rr_ptr = rt0 = fn->leaf;
 686
 687         match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict,
 688                              &do_rr);
 689
 690         if (do_rr) {
 691                 struct rt6_info *next = rt0->dst.rt6_next;
 692
 693                 /* no entries matched; do round-robin */
 694                 if (!next || next->rt6i_metric != rt0->rt6i_metric)
 695                         next = fn->leaf;
 696
 697                 if (next != rt0)
 698                         fn->rr_ptr = next;
 699         }
 700
 701         net = dev_net(rt0->dst.dev);
 702         return match ? match : net->ipv6.ip6_null_entry;
 703 }
 704
 705 #ifdef CONFIG_IPV6_ROUTE_INFO
 706 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
 707                   const struct in6_addr *gwaddr)
 708 {
 709         struct net *net = dev_net(dev);
 710         struct route_info *rinfo = (struct route_info *) opt;
 711         struct in6_addr prefix_buf, *prefix;
 712         unsigned int pref;
 713         unsigned long lifetime;
 714         struct rt6_info *rt;
 715
 716         if (len < sizeof(struct route_info)) {
 717                 return -EINVAL;
 718         }
 719
 720         /* Sanity check for prefix_len and length */
 721         if (rinfo->length > 3) {
 722                 return -EINVAL;
 723         } else if (rinfo->prefix_len > 128) {
 724                 return -EINVAL;
 725         } else if (rinfo->prefix_len > 64) {
 726                 if (rinfo->length < 2) {
 727                         return -EINVAL;
 728                 }
 729         } else if (rinfo->prefix_len > 0) {
 730                 if (rinfo->length < 1) {
 731                         return -EINVAL;
 732                 }
 733         }
 734
 735         pref = rinfo->route_pref;
 736         if (pref == ICMPV6_ROUTER_PREF_INVALID)
 737                 return -EINVAL;
 738
 739         lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
 740
 741         if (rinfo->length == 3)
 742                 prefix = (struct in6_addr *)rinfo->prefix;
 743         else {
 744                 /* this function is safe */
 745                 ipv6_addr_prefix(&prefix_buf,
 746                                  (struct in6_addr *)rinfo->prefix,
 747                                  rinfo->prefix_len);
 748                 prefix = &prefix_buf;
 749         }
 750
 751         if (rinfo->prefix_len == 0)
 752                 rt = rt6_get_dflt_router(gwaddr, dev);
 753         else
 754                 rt = rt6_get_route_info(net, prefix, rinfo->prefix_len,
 755                                         gwaddr, dev->ifindex);
 756
 757         if (rt && !lifetime) {
 758                 ip6_del_rt(rt);
 759                 rt = NULL;
 760         }
 761
 762         if (!rt && lifetime)
 763                 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr, dev->ifindex,
 764                                         pref);
 765         else if (rt)
 766                 rt->rt6i_flags = RTF_ROUTEINFO |
 767                                  (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
 768
 769         if (rt) {
 770                 if (!addrconf_finite_timeout(lifetime))
 771                         rt6_clean_expires(rt);
 772                 else
 773                         rt6_set_expires(rt, jiffies + HZ * lifetime);
 774
 775                 ip6_rt_put(rt);
 776         }
 777         return 0;
 778 }
 779 #endif
 780
 781 #define BACKTRACK(__net, saddr)                 \
 782 do { \
 783         if (rt == __net->ipv6.ip6_null_entry) { \
 784                 struct fib6_node *pn; \
 785                 while (1) { \
 786                         if (fn->fn_flags & RTN_TL_ROOT) \
 787                                 goto out; \
 788                         pn = fn->parent; \
 789                         if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn) \
 790                                 fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr); \
 791                         else \
 792                                 fn = pn; \
 793                         if (fn->fn_flags & RTN_RTINFO) \
 794                                 goto restart; \
 795                 } \
 796         } \
 797 } while (0)
 798
 799 static struct rt6_info *ip6_pol_route_lookup(struct net *net,
 800                                              struct fib6_table *table,
 801                                              struct flowi6 *fl6, int flags)
 802 {
 803         struct fib6_node *fn;
 804         struct rt6_info *rt;
 805
 806         read_lock_bh(&table->tb6_lock);
 807         fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
 808 restart:
 809         rt = fn->leaf;
 810         rt = rt6_device_match(net, rt, &fl6->saddr, fl6->flowi6_oif, flags);
 811         if (rt->rt6i_nsiblings && fl6->flowi6_oif == 0)
 812                 rt = rt6_multipath_select(rt, fl6, fl6->flowi6_oif, flags);
 813         BACKTRACK(net, &fl6->saddr);
 814 out:
 815         dst_use(&rt->dst, jiffies);
 816         read_unlock_bh(&table->tb6_lock);
 817         return rt;
 818
 819 }
 820
 821 struct dst_entry * ip6_route_lookup(struct net *net, struct flowi6 *fl6,
 822                                     int flags)
 823 {
 824         return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_lookup);
 825 }
 826 EXPORT_SYMBOL_GPL(ip6_route_lookup);
 827
 828 struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
 829                             const struct in6_addr *saddr, int oif, int strict)
 830 {
 831         struct flowi6 fl6 = {
 832                 .flowi6_oif = oif,
 833                 .daddr = *daddr,
 834         };
 835         struct dst_entry *dst;
 836         int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
 837
 838         if (saddr) {
 839                 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
 840                 flags |= RT6_LOOKUP_F_HAS_SADDR;
 841         }
 842
 843         dst = fib6_rule_lookup(net, &fl6, flags, ip6_pol_route_lookup);
 844         if (dst->error == 0)
 845                 return (struct rt6_info *) dst;
 846
 847         dst_release(dst);
 848
 849         return NULL;
 850 }
 851
 852 EXPORT_SYMBOL(rt6_lookup);
 853
 854 /* ip6_ins_rt is called with FREE table->tb6_lock.
 855    It takes new route entry, the addition fails by any reason the
 856    route is freed. In any case, if caller does not hold it, it may
 857    be destroyed.
 858  */
 859
 860 static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info)
 861 {
 862         int err;
 863         struct fib6_table *table;
 864
 865         table = rt->rt6i_table;
 866         write_lock_bh(&table->tb6_lock);
 867         err = fib6_add(&table->tb6_root, rt, info);
 868         write_unlock_bh(&table->tb6_lock);
 869
 870         return err;
 871 }
 872
 873 int ip6_ins_rt(struct rt6_info *rt)
 874 {
 875         struct nl_info info = {
 876                 .nl_net = dev_net(rt->dst.dev),
 877         };
 878         return __ip6_ins_rt(rt, &info);
 879 }
 880
 881 static struct rt6_info *rt6_alloc_cow(struct rt6_info *ort,
 882                                       const struct in6_addr *daddr,
 883                                       const struct in6_addr *saddr)
 884 {
 885         struct rt6_info *rt;
 886
 887         /*
 888          *      Clone the route.
 889          */
 890
 891         rt = ip6_rt_copy(ort, daddr);
 892
 893         if (rt) {
 894                 if (ort->rt6i_dst.plen != 128 &&
 895                     ipv6_addr_equal(&ort->rt6i_dst.addr, daddr))
 896                         rt->rt6i_flags |= RTF_ANYCAST;
 897
 898                 rt->rt6i_flags |= RTF_CACHE;
 899
 900 #ifdef CONFIG_IPV6_SUBTREES
 901                 if (rt->rt6i_src.plen && saddr) {
 902                         rt->rt6i_src.addr = *saddr;
 903                         rt->rt6i_src.plen = 128;
 904                 }
 905 #endif
 906         }
 907
 908         return rt;
 909 }
 910
 911 static struct rt6_info *rt6_alloc_clone(struct rt6_info *ort,
 912                                         const struct in6_addr *daddr)
 913 {
 914         struct rt6_info *rt = ip6_rt_copy(ort, daddr);
 915
 916         if (rt)
 917                 rt->rt6i_flags |= RTF_CACHE;
 918         return rt;
 919 }
 920
 921 static struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table, int oif,
 922                                       struct flowi6 *fl6, int flags)
 923 {
 924         struct fib6_node *fn;
 925         struct rt6_info *rt, *nrt;
 926         int strict = 0;
 927         int attempts = 3;
 928         int err;
 929         int reachable = net->ipv6.devconf_all->forwarding ? 0 : RT6_LOOKUP_F_REACHABLE;
 930
 931         strict |= flags & RT6_LOOKUP_F_IFACE;
 932
 933 relookup:
 934         read_lock_bh(&table->tb6_lock);
 935
 936 restart_2:
 937         fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
 938
 939 restart:
 940         rt = rt6_select(fn, oif, strict | reachable);
 941         if (rt->rt6i_nsiblings)
 942                 rt = rt6_multipath_select(rt, fl6, oif, strict | reachable);
 943         BACKTRACK(net, &fl6->saddr);
 944         if (rt == net->ipv6.ip6_null_entry ||
 945             rt->rt6i_flags & RTF_CACHE)
 946                 goto out;
 947
 948         dst_hold(&rt->dst);
 949         read_unlock_bh(&table->tb6_lock);
 950
 951         if (!(rt->rt6i_flags & (RTF_NONEXTHOP | RTF_GATEWAY)))
 952                 nrt = rt6_alloc_cow(rt, &fl6->daddr, &fl6->saddr);
 953         else if (!(rt->dst.flags & DST_HOST))
 954                 nrt = rt6_alloc_clone(rt, &fl6->daddr);
 955         else
 956                 goto out2;
 957
 958         ip6_rt_put(rt);
 959         rt = nrt ? : net->ipv6.ip6_null_entry;
 960
 961         dst_hold(&rt->dst);
 962         if (nrt) {
 963                 err = ip6_ins_rt(nrt);
 964                 if (!err)
 965                         goto out2;
 966         }
 967
 968         if (--attempts <= 0)
 969                 goto out2;
 970
 971         /*
 972          * Race condition! In the gap, when table->tb6_lock was
 973          * released someone could insert this route.  Relookup.
 974          */
 975         ip6_rt_put(rt);
 976         goto relookup;
 977
 978 out:
 979         if (reachable) {
 980                 reachable = 0;
 981                 goto restart_2;
 982         }
 983         dst_hold(&rt->dst);
 984         read_unlock_bh(&table->tb6_lock);
 985 out2:
 986         rt->dst.lastuse = jiffies;
 987         rt->dst.__use++;
 988
 989         return rt;
 990 }
 991
 992 static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
 993                                             struct flowi6 *fl6, int flags)
 994 {
 995         return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, flags);
 996 }
 997
 998 static struct dst_entry *ip6_route_input_lookup(struct net *net,
 999                                                 struct net_device *dev,
1000                                                 struct flowi6 *fl6, int flags)
1001 {
1002         if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
1003                 flags |= RT6_LOOKUP_F_IFACE;
1004
1005         return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_input);
1006 }
1007
1008 void ip6_route_input(struct sk_buff *skb)
1009 {
1010         const struct ipv6hdr *iph = ipv6_hdr(skb);
1011         struct net *net = dev_net(skb->dev);
1012         int flags = RT6_LOOKUP_F_HAS_SADDR;
1013         struct flowi6 fl6 = {
1014                 .flowi6_iif = skb->dev->ifindex,
1015                 .daddr = iph->daddr,
1016                 .saddr = iph->saddr,
1017                 .flowlabel = ip6_flowinfo(iph),
1018                 .flowi6_mark = skb->mark,
1019                 .flowi6_proto = iph->nexthdr,
1020         };
1021
1022         skb_dst_set(skb, ip6_route_input_lookup(net, skb->dev, &fl6, flags));
1023 }
1024
1025 static struct rt6_info *ip6_pol_route_output(struct net *net, struct fib6_table *table,
1026                                              struct flowi6 *fl6, int flags)
1027 {
1028         return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, flags);
1029 }
1030
1031 struct dst_entry * ip6_route_output(struct net *net, const struct sock *sk,
1032                                     struct flowi6 *fl6)
1033 {
1034         int flags = 0;
1035
1036         fl6->flowi6_iif = LOOPBACK_IFINDEX;
1037
1038         if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr))
1039                 flags |= RT6_LOOKUP_F_IFACE;
1040
1041         if (!ipv6_addr_any(&fl6->saddr))
1042                 flags |= RT6_LOOKUP_F_HAS_SADDR;
1043         else if (sk)
1044                 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
1045
1046         return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_output);
1047 }
1048
1049 EXPORT_SYMBOL(ip6_route_output);
1050
1051 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
1052 {
1053         struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
1054         struct dst_entry *new = NULL;
1055
1056         rt = dst_alloc(&ip6_dst_blackhole_ops, ort->dst.dev, 1, DST_OBSOLETE_NONE, 0);
1057         if (rt) {
1058                 new = &rt->dst;
1059
1060                 memset(new + 1, 0, sizeof(*rt) - sizeof(*new));
1061                 rt6_init_peer(rt, net->ipv6.peers);
1062
1063                 new->__use = 1;
1064                 new->input = dst_discard;
1065                 new->output = dst_discard;
1066
1067                 if (dst_metrics_read_only(&ort->dst))
1068                         new->_metrics = ort->dst._metrics;
1069                 else
1070                         dst_copy_metrics(new, &ort->dst);
1071                 rt->rt6i_idev = ort->rt6i_idev;
1072                 if (rt->rt6i_idev)
1073                         in6_dev_hold(rt->rt6i_idev);
1074
1075                 rt->rt6i_gateway = ort->rt6i_gateway;
1076                 rt->rt6i_flags = ort->rt6i_flags;
1077                 rt->rt6i_metric = 0;
1078
1079                 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1080 #ifdef CONFIG_IPV6_SUBTREES
1081                 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1082 #endif
1083
1084                 dst_free(new);
1085         }
1086
1087         dst_release(dst_orig);
1088         return new ? new : ERR_PTR(-ENOMEM);
1089 }
1090
1091 /*
1092  *      Destination cache support functions
1093  */
1094
1095 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
1096 {
1097         struct rt6_info *rt;
1098
1099         rt = (struct rt6_info *) dst;
1100
1101         /* All IPV6 dsts are created with ->obsolete set to the value
1102          * DST_OBSOLETE_FORCE_CHK which forces validation calls down
1103          * into this function always.
1104          */
1105         if (rt->rt6i_genid != rt_genid_ipv6(dev_net(rt->dst.dev)))
1106                 return NULL;
1107
1108         if (!rt->rt6i_node || (rt->rt6i_node->fn_sernum != cookie))
1109                 return NULL;
1110
1111         if (rt6_check_expired(rt))
1112                 return NULL;
1113
1114         return dst;
1115 }
1116
1117 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
1118 {
1119         struct rt6_info *rt = (struct rt6_info *) dst;
1120
1121         if (rt) {
1122                 if (rt->rt6i_flags & RTF_CACHE) {
1123                         if (rt6_check_expired(rt)) {
1124                                 ip6_del_rt(rt);
1125                                 dst = NULL;
1126                         }
1127                 } else {
1128                         dst_release(dst);
1129                         dst = NULL;
1130                 }
1131         }
1132         return dst;
1133 }
1134
1135 static void ip6_link_failure(struct sk_buff *skb)
1136 {
1137         struct rt6_info *rt;
1138
1139         icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
1140
1141         rt = (struct rt6_info *) skb_dst(skb);
1142         if (rt) {
1143                 if (rt->rt6i_flags & RTF_CACHE) {
1144                         dst_hold(&rt->dst);
1145                         if (ip6_del_rt(rt))
1146                                 dst_free(&rt->dst);
1147                 } else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT)) {
1148                         rt->rt6i_node->fn_sernum = -1;
1149                 }
1150         }
1151 }
1152
1153 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
1154                                struct sk_buff *skb, u32 mtu)
1155 {
1156         struct rt6_info *rt6 = (struct rt6_info*)dst;
1157
1158         dst_confirm(dst);
1159         if (mtu < dst_mtu(dst) && rt6->rt6i_dst.plen == 128) {
1160                 struct net *net = dev_net(dst->dev);
1161
1162                 rt6->rt6i_flags |= RTF_MODIFIED;
1163                 if (mtu < IPV6_MIN_MTU) {
1164                         u32 features = dst_metric(dst, RTAX_FEATURES);
1165                         mtu = IPV6_MIN_MTU;
1166                         features |= RTAX_FEATURE_ALLFRAG;
1167                         dst_metric_set(dst, RTAX_FEATURES, features);
1168                 }
1169                 dst_metric_set(dst, RTAX_MTU, mtu);
1170                 rt6_update_expires(rt6, net->ipv6.sysctl.ip6_rt_mtu_expires);
1171         }
1172 }
1173
1174 void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
1175                      int oif, u32 mark)
1176 {
1177         const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1178         struct dst_entry *dst;
1179         struct flowi6 fl6;
1180
1181         memset(&fl6, 0, sizeof(fl6));
1182         fl6.flowi6_oif = oif;
1183         fl6.flowi6_mark = mark;
1184         fl6.daddr = iph->daddr;
1185         fl6.saddr = iph->saddr;
1186         fl6.flowlabel = ip6_flowinfo(iph);
1187
1188         dst = ip6_route_output(net, NULL, &fl6);
1189         if (!dst->error)
1190                 ip6_rt_update_pmtu(dst, NULL, skb, ntohl(mtu));
1191         dst_release(dst);
1192 }
1193 EXPORT_SYMBOL_GPL(ip6_update_pmtu);
1194
1195 void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
1196 {
1197         ip6_update_pmtu(skb, sock_net(sk), mtu,
1198                         sk->sk_bound_dev_if, sk->sk_mark);
1199 }
1200 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
1201
1202 /* Handle redirects */
1203 struct ip6rd_flowi {
1204         struct flowi6 fl6;
1205         struct in6_addr gateway;
1206 };
1207
1208 static struct rt6_info *__ip6_route_redirect(struct net *net,
1209                                              struct fib6_table *table,
1210                                              struct flowi6 *fl6,
1211                                              int flags)
1212 {
1213         struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
1214         struct rt6_info *rt;
1215         struct fib6_node *fn;
1216
1217         /* Get the "current" route for this destination and
1218          * check if the redirect has come from approriate router.
1219          *
1220          * RFC 4861 specifies that redirects should only be
1221          * accepted if they come from the nexthop to the target.
1222          * Due to the way the routes are chosen, this notion
1223          * is a bit fuzzy and one might need to check all possible
1224          * routes.
1225          */
1226
1227         read_lock_bh(&table->tb6_lock);
1228         fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1229 restart:
1230         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1231                 if (rt6_check_expired(rt))
1232                         continue;
1233                 if (rt->dst.error)
1234                         break;
1235                 if (!(rt->rt6i_flags & RTF_GATEWAY))
1236                         continue;
1237                 if (fl6->flowi6_oif != rt->dst.dev->ifindex)
1238                         continue;
1239                 if (!ipv6_addr_equal(&rdfl->gateway, &rt->rt6i_gateway))
1240                         continue;
1241                 break;
1242         }
1243
1244         if (!rt)
1245                 rt = net->ipv6.ip6_null_entry;
1246         else if (rt->dst.error) {
1247                 rt = net->ipv6.ip6_null_entry;
1248                 goto out;
1249         }
1250         BACKTRACK(net, &fl6->saddr);
1251 out:
1252         dst_hold(&rt->dst);
1253
1254         read_unlock_bh(&table->tb6_lock);
1255
1256         return rt;
1257 };
1258
1259 static struct dst_entry *ip6_route_redirect(struct net *net,
1260                                         const struct flowi6 *fl6,
1261                                         const struct in6_addr *gateway)
1262 {
1263         int flags = RT6_LOOKUP_F_HAS_SADDR;
1264         struct ip6rd_flowi rdfl;
1265
1266         rdfl.fl6 = *fl6;
1267         rdfl.gateway = *gateway;
1268
1269         return fib6_rule_lookup(net, &rdfl.fl6,
1270                                 flags, __ip6_route_redirect);
1271 }
1272
1273 void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark)
1274 {
1275         const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1276         struct dst_entry *dst;
1277         struct flowi6 fl6;
1278
1279         memset(&fl6, 0, sizeof(fl6));
1280         fl6.flowi6_oif = oif;
1281         fl6.flowi6_mark = mark;
1282         fl6.daddr = iph->daddr;
1283         fl6.saddr = iph->saddr;
1284         fl6.flowlabel = ip6_flowinfo(iph);
1285
1286         dst = ip6_route_redirect(net, &fl6, &ipv6_hdr(skb)->saddr);
1287         rt6_do_redirect(dst, NULL, skb);
1288         dst_release(dst);
1289 }
1290 EXPORT_SYMBOL_GPL(ip6_redirect);
1291
1292 void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif,
1293                             u32 mark)
1294 {
1295         const struct ipv6hdr *iph = ipv6_hdr(skb);
1296         const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
1297         struct dst_entry *dst;
1298         struct flowi6 fl6;
1299
1300         memset(&fl6, 0, sizeof(fl6));
1301         fl6.flowi6_oif = oif;
1302         fl6.flowi6_mark = mark;
1303         fl6.daddr = msg->dest;
1304         fl6.saddr = iph->daddr;
1305
1306         dst = ip6_route_redirect(net, &fl6, &iph->saddr);
1307         rt6_do_redirect(dst, NULL, skb);
1308         dst_release(dst);
1309 }
1310
1311 void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
1312 {
1313         ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark);
1314 }
1315 EXPORT_SYMBOL_GPL(ip6_sk_redirect);
1316
1317 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1318 {
1319         struct net_device *dev = dst->dev;
1320         unsigned int mtu = dst_mtu(dst);
1321         struct net *net = dev_net(dev);
1322
1323         mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
1324
1325         if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
1326                 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1327
1328         /*
1329          * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1330          * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1331          * IPV6_MAXPLEN is also valid and means: "any MSS,
1332          * rely only on pmtu discovery"
1333          */
1334         if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
1335                 mtu = IPV6_MAXPLEN;
1336         return mtu;
1337 }
1338
1339 static unsigned int ip6_mtu(const struct dst_entry *dst)
1340 {
1341         struct inet6_dev *idev;
1342         unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
1343
1344         if (mtu)
1345                 return mtu;
1346
1347         mtu = IPV6_MIN_MTU;
1348
1349         rcu_read_lock();
1350         idev = __in6_dev_get(dst->dev);
1351         if (idev)
1352                 mtu = idev->cnf.mtu6;
1353         rcu_read_unlock();
1354
1355         return mtu;
1356 }
1357
1358 static struct dst_entry *icmp6_dst_gc_list;
1359 static DEFINE_SPINLOCK(icmp6_dst_lock);
1360
1361 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
1362                                   struct flowi6 *fl6)
1363 {
1364         struct dst_entry *dst;
1365         struct rt6_info *rt;
1366         struct inet6_dev *idev = in6_dev_get(dev);
1367         struct net *net = dev_net(dev);
1368
1369         if (unlikely(!idev))
1370                 return ERR_PTR(-ENODEV);
1371
1372         rt = ip6_dst_alloc(net, dev, 0, NULL);
1373         if (unlikely(!rt)) {
1374                 in6_dev_put(idev);
1375                 dst = ERR_PTR(-ENOMEM);
1376                 goto out;
1377         }
1378
1379         rt->dst.flags |= DST_HOST;
1380         rt->dst.output  = ip6_output;
1381         atomic_set(&rt->dst.__refcnt, 1);
1382         rt->rt6i_gateway  = fl6->daddr;
1383         rt->rt6i_dst.addr = fl6->daddr;
1384         rt->rt6i_dst.plen = 128;
1385         rt->rt6i_idev     = idev;
1386         dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
1387
1388         spin_lock_bh(&icmp6_dst_lock);
1389         rt->dst.next = icmp6_dst_gc_list;
1390         icmp6_dst_gc_list = &rt->dst;
1391         spin_unlock_bh(&icmp6_dst_lock);
1392
1393         fib6_force_start_gc(net);
1394
1395         dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
1396
1397 out:
1398         return dst;
1399 }
1400
1401 int icmp6_dst_gc(void)
1402 {
1403         struct dst_entry *dst, **pprev;
1404         int more = 0;
1405
1406         spin_lock_bh(&icmp6_dst_lock);
1407         pprev = &icmp6_dst_gc_list;
1408
1409         while ((dst = *pprev) != NULL) {
1410                 if (!atomic_read(&dst->__refcnt)) {
1411                         *pprev = dst->next;
1412                         dst_free(dst);
1413                 } else {
1414                         pprev = &dst->next;
1415                         ++more;
1416                 }
1417         }
1418
1419         spin_unlock_bh(&icmp6_dst_lock);
1420
1421         return more;
1422 }
1423
1424 static void icmp6_clean_all(int (*func)(struct rt6_info *rt, void *arg),
1425                             void *arg)
1426 {
1427         struct dst_entry *dst, **pprev;
1428
1429         spin_lock_bh(&icmp6_dst_lock);
1430         pprev = &icmp6_dst_gc_list;
1431         while ((dst = *pprev) != NULL) {
1432                 struct rt6_info *rt = (struct rt6_info *) dst;
1433                 if (func(rt, arg)) {
1434                         *pprev = dst->next;
1435                         dst_free(dst);
1436                 } else {
1437                         pprev = &dst->next;
1438                 }
1439         }
1440         spin_unlock_bh(&icmp6_dst_lock);
1441 }
1442
1443 static int ip6_dst_gc(struct dst_ops *ops)
1444 {
1445         struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
1446         int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1447         int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1448         int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1449         int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1450         unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1451         int entries;
1452
1453         entries = dst_entries_get_fast(ops);
1454         if (time_after(rt_last_gc + rt_min_interval, jiffies) &&
1455             entries <= rt_max_size)
1456                 goto out;
1457
1458         net->ipv6.ip6_rt_gc_expire++;
1459         fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net, entries > rt_max_size);
1460         entries = dst_entries_get_slow(ops);
1461         if (entries < ops->gc_thresh)
1462                 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1463 out:
1464         net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1465         return entries > rt_max_size;
1466 }
1467
1468 /*
1469  *
1470  */
1471
1472 int ip6_route_add(struct fib6_config *cfg)
1473 {
1474         int err;
1475         struct net *net = cfg->fc_nlinfo.nl_net;
1476         struct rt6_info *rt = NULL;
1477         struct net_device *dev = NULL;
1478         struct inet6_dev *idev = NULL;
1479         struct fib6_table *table;
1480         int addr_type;
1481
1482         if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1483                 return -EINVAL;
1484 #ifndef CONFIG_IPV6_SUBTREES
1485         if (cfg->fc_src_len)
1486                 return -EINVAL;
1487 #endif
1488         if (cfg->fc_ifindex) {
1489                 err = -ENODEV;
1490                 dev = dev_get_by_index(net, cfg->fc_ifindex);
1491                 if (!dev)
1492                         goto out;
1493                 idev = in6_dev_get(dev);
1494                 if (!idev)
1495                         goto out;
1496         }
1497
1498         if (cfg->fc_metric == 0)
1499                 cfg->fc_metric = IP6_RT_PRIO_USER;
1500
1501         err = -ENOBUFS;
1502         if (cfg->fc_nlinfo.nlh &&
1503             !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
1504                 table = fib6_get_table(net, cfg->fc_table);
1505                 if (!table) {
1506                         pr_warn("NLM_F_CREATE should be specified when creating new route\n");
1507                         table = fib6_new_table(net, cfg->fc_table);
1508                 }
1509         } else {
1510                 table = fib6_new_table(net, cfg->fc_table);
1511         }
1512
1513         if (!table)
1514                 goto out;
1515
1516         rt = ip6_dst_alloc(net, NULL, (cfg->fc_flags & RTF_ADDRCONF) ? 0 : DST_NOCOUNT, table);
1517
1518         if (!rt) {
1519                 err = -ENOMEM;
1520                 goto out;
1521         }
1522
1523         if (cfg->fc_flags & RTF_EXPIRES)
1524                 rt6_set_expires(rt, jiffies +
1525                                 clock_t_to_jiffies(cfg->fc_expires));
1526         else
1527                 rt6_clean_expires(rt);
1528
1529         if (cfg->fc_protocol == RTPROT_UNSPEC)
1530                 cfg->fc_protocol = RTPROT_BOOT;
1531         rt->rt6i_protocol = cfg->fc_protocol;
1532
1533         addr_type = ipv6_addr_type(&cfg->fc_dst);
1534
1535         if (addr_type & IPV6_ADDR_MULTICAST)
1536                 rt->dst.input = ip6_mc_input;
1537         else if (cfg->fc_flags & RTF_LOCAL)
1538                 rt->dst.input = ip6_input;
1539         else
1540                 rt->dst.input = ip6_forward;
1541
1542         rt->dst.output = ip6_output;
1543
1544         ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1545         rt->rt6i_dst.plen = cfg->fc_dst_len;
1546         if (rt->rt6i_dst.plen == 128)
1547                rt->dst.flags |= DST_HOST;
1548
1549         if (!(rt->dst.flags & DST_HOST) && cfg->fc_mx) {
1550                 u32 *metrics = kzalloc(sizeof(u32) * RTAX_MAX, GFP_KERNEL);
1551                 if (!metrics) {
1552                         err = -ENOMEM;
1553                         goto out;
1554                 }
1555                 dst_init_metrics(&rt->dst, metrics, 0);
1556         }
1557 #ifdef CONFIG_IPV6_SUBTREES
1558         ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1559         rt->rt6i_src.plen = cfg->fc_src_len;
1560 #endif
1561
1562         rt->rt6i_metric = cfg->fc_metric;
1563
1564         /* We cannot add true routes via loopback here,
1565            they would result in kernel looping; promote them to reject routes
1566          */
1567         if ((cfg->fc_flags & RTF_REJECT) ||
1568             (dev && (dev->flags & IFF_LOOPBACK) &&
1569              !(addr_type & IPV6_ADDR_LOOPBACK) &&
1570              !(cfg->fc_flags & RTF_LOCAL))) {
1571                 /* hold loopback dev/idev if we haven't done so. */
1572                 if (dev != net->loopback_dev) {
1573                         if (dev) {
1574                                 dev_put(dev);
1575                                 in6_dev_put(idev);
1576                         }
1577                         dev = net->loopback_dev;
1578                         dev_hold(dev);
1579                         idev = in6_dev_get(dev);
1580                         if (!idev) {
1581                                 err = -ENODEV;
1582                                 goto out;
1583                         }
1584                 }
1585                 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1586                 switch (cfg->fc_type) {
1587                 case RTN_BLACKHOLE:
1588                         rt->dst.error = -EINVAL;
1589                         rt->dst.output = dst_discard;
1590                         rt->dst.input = dst_discard;
1591                         break;
1592                 case RTN_PROHIBIT:
1593                         rt->dst.error = -EACCES;
1594                         rt->dst.output = ip6_pkt_prohibit_out;
1595                         rt->dst.input = ip6_pkt_prohibit;
1596                         break;
1597                 case RTN_THROW:
1598                 default:
1599                         rt->dst.error = (cfg->fc_type == RTN_THROW) ? -EAGAIN
1600                                         : -ENETUNREACH;
1601                         rt->dst.output = ip6_pkt_discard_out;
1602                         rt->dst.input = ip6_pkt_discard;
1603                         break;
1604                 }
1605                 goto install_route;
1606         }
1607
1608         if (cfg->fc_flags & RTF_GATEWAY) {
1609                 const struct in6_addr *gw_addr;
1610                 int gwa_type;
1611
1612                 gw_addr = &cfg->fc_gateway;
1613                 rt->rt6i_gateway = *gw_addr;
1614                 gwa_type = ipv6_addr_type(gw_addr);
1615
1616                 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1617                         struct rt6_info *grt;
1618
1619                         /* IPv6 strictly inhibits using not link-local
1620                            addresses as nexthop address.
1621                            Otherwise, router will not able to send redirects.
1622                            It is very good, but in some (rare!) circumstances
1623                            (SIT, PtP, NBMA NOARP links) it is handy to allow
1624                            some exceptions. --ANK
1625                          */
1626                         err = -EINVAL;
1627                         if (!(gwa_type & IPV6_ADDR_UNICAST))
1628                                 goto out;
1629
1630                         grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, 1);
1631
1632                         err = -EHOSTUNREACH;
1633                         if (!grt)
1634                                 goto out;
1635                         if (dev) {
1636                                 if (dev != grt->dst.dev) {
1637                                         ip6_rt_put(grt);
1638                                         goto out;
1639                                 }
1640                         } else {
1641                                 dev = grt->dst.dev;
1642                                 idev = grt->rt6i_idev;
1643                                 dev_hold(dev);
1644                                 in6_dev_hold(grt->rt6i_idev);
1645                         }
1646                         if (!(grt->rt6i_flags & RTF_GATEWAY))
1647                                 err = 0;
1648                         ip6_rt_put(grt);
1649
1650                         if (err)
1651                                 goto out;
1652                 }
1653                 err = -EINVAL;
1654                 if (!dev || (dev->flags & IFF_LOOPBACK))
1655                         goto out;
1656         }
1657
1658         err = -ENODEV;
1659         if (!dev)
1660                 goto out;
1661
1662         if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
1663                 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
1664                         err = -EINVAL;
1665                         goto out;
1666                 }
1667                 rt->rt6i_prefsrc.addr = cfg->fc_prefsrc;
1668                 rt->rt6i_prefsrc.plen = 128;
1669         } else
1670                 rt->rt6i_prefsrc.plen = 0;
1671
1672         rt->rt6i_flags = cfg->fc_flags;
1673
1674 install_route:
1675         if (cfg->fc_mx) {
1676                 struct nlattr *nla;
1677                 int remaining;
1678
1679                 nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1680                         int type = nla_type(nla);
1681
1682                         if (type) {
1683                                 if (type > RTAX_MAX) {
1684                                         err = -EINVAL;
1685                                         goto out;
1686                                 }
1687
1688                                 dst_metric_set(&rt->dst, type, nla_get_u32(nla));
1689                         }
1690                 }
1691         }
1692
1693         rt->dst.dev = dev;
1694         rt->rt6i_idev = idev;
1695         rt->rt6i_table = table;
1696
1697         cfg->fc_nlinfo.nl_net = dev_net(dev);
1698
1699         return __ip6_ins_rt(rt, &cfg->fc_nlinfo);
1700
1701 out:
1702         if (dev)
1703                 dev_put(dev);
1704         if (idev)
1705                 in6_dev_put(idev);
1706         if (rt)
1707                 dst_free(&rt->dst);
1708         return err;
1709 }
1710
1711 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
1712 {
1713         int err;
1714         struct fib6_table *table;
1715         struct net *net = dev_net(rt->dst.dev);
1716
1717         if (rt == net->ipv6.ip6_null_entry) {
1718                 err = -ENOENT;
1719                 goto out;
1720         }
1721
1722         table = rt->rt6i_table;
1723         write_lock_bh(&table->tb6_lock);
1724         err = fib6_del(rt, info);
1725         write_unlock_bh(&table->tb6_lock);
1726
1727 out:
1728         ip6_rt_put(rt);
1729         return err;
1730 }
1731
1732 int ip6_del_rt(struct rt6_info *rt)
1733 {
1734         struct nl_info info = {
1735                 .nl_net = dev_net(rt->dst.dev),
1736         };
1737         return __ip6_del_rt(rt, &info);
1738 }
1739
1740 static int ip6_route_del(struct fib6_config *cfg)
1741 {
1742         struct fib6_table *table;
1743         struct fib6_node *fn;
1744         struct rt6_info *rt;
1745         int err = -ESRCH;
1746
1747         table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
1748         if (!table)
1749                 return err;
1750
1751         read_lock_bh(&table->tb6_lock);
1752
1753         fn = fib6_locate(&table->tb6_root,
1754                          &cfg->fc_dst, cfg->fc_dst_len,
1755                          &cfg->fc_src, cfg->fc_src_len);
1756
1757         if (fn) {
1758                 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1759                         if (cfg->fc_ifindex &&
1760                             (!rt->dst.dev ||
1761                              rt->dst.dev->ifindex != cfg->fc_ifindex))
1762                                 continue;
1763                         if (cfg->fc_flags & RTF_GATEWAY &&
1764                             !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
1765                                 continue;
1766                         if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
1767                                 continue;
1768                         dst_hold(&rt->dst);
1769                         read_unlock_bh(&table->tb6_lock);
1770
1771                         return __ip6_del_rt(rt, &cfg->fc_nlinfo);
1772                 }
1773         }
1774         read_unlock_bh(&table->tb6_lock);
1775
1776         return err;
1777 }
1778
1779 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
1780 {
1781         struct net *net = dev_net(skb->dev);
1782         struct netevent_redirect netevent;
1783         struct rt6_info *rt, *nrt = NULL;
1784         struct ndisc_options ndopts;
1785         struct inet6_dev *in6_dev;
1786         struct neighbour *neigh;
1787         struct rd_msg *msg;
1788         int optlen, on_link;
1789         u8 *lladdr;
1790
1791         optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
1792         optlen -= sizeof(*msg);
1793
1794         if (optlen < 0) {
1795                 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
1796                 return;
1797         }
1798
1799         msg = (struct rd_msg *)icmp6_hdr(skb);
1800
1801         if (ipv6_addr_is_multicast(&msg->dest)) {
1802                 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
1803                 return;
1804         }
1805
1806         on_link = 0;
1807         if (ipv6_addr_equal(&msg->dest, &msg->target)) {
1808                 on_link = 1;
1809         } else if (ipv6_addr_type(&msg->target) !=
1810                    (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
1811                 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
1812                 return;
1813         }
1814
1815         in6_dev = __in6_dev_get(skb->dev);
1816         if (!in6_dev)
1817                 return;
1818         if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
1819                 return;
1820
1821         /* RFC2461 8.1:
1822          *      The IP source address of the Redirect MUST be the same as the current
1823          *      first-hop router for the specified ICMP Destination Address.
1824          */
1825
1826         if (!ndisc_parse_options(msg->opt, optlen, &ndopts)) {
1827                 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
1828                 return;
1829         }
1830
1831         lladdr = NULL;
1832         if (ndopts.nd_opts_tgt_lladdr) {
1833                 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
1834                                              skb->dev);
1835                 if (!lladdr) {
1836                         net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
1837                         return;
1838                 }
1839         }
1840
1841         rt = (struct rt6_info *) dst;
1842         if (rt == net->ipv6.ip6_null_entry) {
1843                 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
1844                 return;
1845         }
1846
1847         /* Redirect received -> path was valid.
1848          * Look, redirects are sent only in response to data packets,
1849          * so that this nexthop apparently is reachable. --ANK
1850          */
1851         dst_confirm(&rt->dst);
1852
1853         neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
1854         if (!neigh)
1855                 return;
1856
1857         /*
1858          *      We have finally decided to accept it.
1859          */
1860
1861         neigh_update(neigh, lladdr, NUD_STALE,
1862                      NEIGH_UPDATE_F_WEAK_OVERRIDE|
1863                      NEIGH_UPDATE_F_OVERRIDE|
1864                      (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1865                                      NEIGH_UPDATE_F_ISROUTER))
1866                      );
1867
1868         nrt = ip6_rt_copy(rt, &msg->dest);
1869         if (!nrt)
1870                 goto out;
1871
1872         nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
1873         if (on_link)
1874                 nrt->rt6i_flags &= ~RTF_GATEWAY;
1875
1876         nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
1877
1878         if (ip6_ins_rt(nrt))
1879                 goto out;
1880
1881         netevent.old = &rt->dst;
1882         netevent.new = &nrt->dst;
1883         netevent.daddr = &msg->dest;
1884         netevent.neigh = neigh;
1885         call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
1886
1887         if (rt->rt6i_flags & RTF_CACHE) {
1888                 rt = (struct rt6_info *) dst_clone(&rt->dst);
1889                 ip6_del_rt(rt);
1890         }
1891
1892 out:
1893         neigh_release(neigh);
1894 }
1895
1896 /*
1897  *      Misc support functions
1898  */
1899
1900 static struct rt6_info *ip6_rt_copy(struct rt6_info *ort,
1901                                     const struct in6_addr *dest)
1902 {
1903         struct net *net = dev_net(ort->dst.dev);
1904         struct rt6_info *rt = ip6_dst_alloc(net, ort->dst.dev, 0,
1905                                             ort->rt6i_table);
1906
1907         if (rt) {
1908                 rt->dst.input = ort->dst.input;
1909                 rt->dst.output = ort->dst.output;
1910                 rt->dst.flags |= DST_HOST;
1911
1912                 rt->rt6i_dst.addr = *dest;
1913                 rt->rt6i_dst.plen = 128;
1914                 dst_copy_metrics(&rt->dst, &ort->dst);
1915                 rt->dst.error = ort->dst.error;
1916                 rt->rt6i_idev = ort->rt6i_idev;
1917                 if (rt->rt6i_idev)
1918                         in6_dev_hold(rt->rt6i_idev);
1919                 rt->dst.lastuse = jiffies;
1920
1921                 if (ort->rt6i_flags & RTF_GATEWAY)
1922                         rt->rt6i_gateway = ort->rt6i_gateway;
1923                 else
1924                         rt->rt6i_gateway = *dest;
1925                 rt->rt6i_flags = ort->rt6i_flags;
1926                 rt6_set_from(rt, ort);
1927                 rt->rt6i_metric = 0;
1928
1929 #ifdef CONFIG_IPV6_SUBTREES
1930                 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1931 #endif
1932                 memcpy(&rt->rt6i_prefsrc, &ort->rt6i_prefsrc, sizeof(struct rt6key));
1933                 rt->rt6i_table = ort->rt6i_table;
1934         }
1935         return rt;
1936 }
1937
1938 #ifdef CONFIG_IPV6_ROUTE_INFO
1939 static struct rt6_info *rt6_get_route_info(struct net *net,
1940                                            const struct in6_addr *prefix, int prefixlen,
1941                                            const struct in6_addr *gwaddr, int ifindex)
1942 {
1943         struct fib6_node *fn;
1944         struct rt6_info *rt = NULL;
1945         struct fib6_table *table;
1946
1947         table = fib6_get_table(net, RT6_TABLE_INFO);
1948         if (!table)
1949                 return NULL;
1950
1951         read_lock_bh(&table->tb6_lock);
1952         fn = fib6_locate(&table->tb6_root, prefix ,prefixlen, NULL, 0);
1953         if (!fn)
1954                 goto out;
1955
1956         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1957                 if (rt->dst.dev->ifindex != ifindex)
1958                         continue;
1959                 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
1960                         continue;
1961                 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
1962                         continue;
1963                 dst_hold(&rt->dst);
1964                 break;
1965         }
1966 out:
1967         read_unlock_bh(&table->tb6_lock);
1968         return rt;
1969 }
1970
1971 static struct rt6_info *rt6_add_route_info(struct net *net,
1972                                            const struct in6_addr *prefix, int prefixlen,
1973                                            const struct in6_addr *gwaddr, int ifindex,
1974                                            unsigned int pref)
1975 {
1976         struct fib6_config cfg = {
1977                 .fc_table       = RT6_TABLE_INFO,
1978                 .fc_metric      = IP6_RT_PRIO_USER,
1979                 .fc_ifindex     = ifindex,
1980                 .fc_dst_len     = prefixlen,
1981                 .fc_flags       = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
1982                                   RTF_UP | RTF_PREF(pref),
1983                 .fc_nlinfo.portid = 0,
1984                 .fc_nlinfo.nlh = NULL,
1985                 .fc_nlinfo.nl_net = net,
1986         };
1987
1988         cfg.fc_dst = *prefix;
1989         cfg.fc_gateway = *gwaddr;
1990
1991         /* We should treat it as a default route if prefix length is 0. */
1992         if (!prefixlen)
1993                 cfg.fc_flags |= RTF_DEFAULT;
1994
1995         ip6_route_add(&cfg);
1996
1997         return rt6_get_route_info(net, prefix, prefixlen, gwaddr, ifindex);
1998 }
1999 #endif
2000
2001 struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
2002 {
2003         struct rt6_info *rt;
2004         struct fib6_table *table;
2005
2006         table = fib6_get_table(dev_net(dev), RT6_TABLE_DFLT);
2007         if (!table)
2008                 return NULL;
2009
2010         read_lock_bh(&table->tb6_lock);
2011         for (rt = table->tb6_root.leaf; rt; rt=rt->dst.rt6_next) {
2012                 if (dev == rt->dst.dev &&
2013                     ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
2014                     ipv6_addr_equal(&rt->rt6i_gateway, addr))
2015                         break;
2016         }
2017         if (rt)
2018                 dst_hold(&rt->dst);
2019         read_unlock_bh(&table->tb6_lock);
2020         return rt;
2021 }
2022
2023 struct rt6_info *rt6_add_dflt_router(const struct in6_addr *gwaddr,
2024                                      struct net_device *dev,
2025                                      unsigned int pref)
2026 {
2027         struct fib6_config cfg = {
2028                 .fc_table       = RT6_TABLE_DFLT,
2029                 .fc_metric      = IP6_RT_PRIO_USER,
2030                 .fc_ifindex     = dev->ifindex,
2031                 .fc_flags       = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
2032                                   RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
2033                 .fc_nlinfo.portid = 0,
2034                 .fc_nlinfo.nlh = NULL,
2035                 .fc_nlinfo.nl_net = dev_net(dev),
2036         };
2037
2038         cfg.fc_gateway = *gwaddr;
2039
2040         ip6_route_add(&cfg);
2041
2042         return rt6_get_dflt_router(gwaddr, dev);
2043 }
2044
2045 void rt6_purge_dflt_routers(struct net *net)
2046 {
2047         struct rt6_info *rt;
2048         struct fib6_table *table;
2049
2050         /* NOTE: Keep consistent with rt6_get_dflt_router */
2051         table = fib6_get_table(net, RT6_TABLE_DFLT);
2052         if (!table)
2053                 return;
2054
2055 restart:
2056         read_lock_bh(&table->tb6_lock);
2057         for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2058                 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
2059                     (!rt->rt6i_idev || rt->rt6i_idev->cnf.accept_ra != 2)) {
2060                         dst_hold(&rt->dst);
2061                         read_unlock_bh(&table->tb6_lock);
2062                         ip6_del_rt(rt);
2063                         goto restart;
2064                 }
2065         }
2066         read_unlock_bh(&table->tb6_lock);
2067 }
2068
2069 static void rtmsg_to_fib6_config(struct net *net,
2070                                  struct in6_rtmsg *rtmsg,
2071                                  struct fib6_config *cfg)
2072 {
2073         memset(cfg, 0, sizeof(*cfg));
2074
2075         cfg->fc_table = RT6_TABLE_MAIN;
2076         cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
2077         cfg->fc_metric = rtmsg->rtmsg_metric;
2078         cfg->fc_expires = rtmsg->rtmsg_info;
2079         cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
2080         cfg->fc_src_len = rtmsg->rtmsg_src_len;
2081         cfg->fc_flags = rtmsg->rtmsg_flags;
2082
2083         cfg->fc_nlinfo.nl_net = net;
2084
2085         cfg->fc_dst = rtmsg->rtmsg_dst;
2086         cfg->fc_src = rtmsg->rtmsg_src;
2087         cfg->fc_gateway = rtmsg->rtmsg_gateway;
2088 }
2089
2090 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
2091 {
2092         struct fib6_config cfg;
2093         struct in6_rtmsg rtmsg;
2094         int err;
2095
2096         switch(cmd) {
2097         case SIOCADDRT:         /* Add a route */
2098         case SIOCDELRT:         /* Delete a route */
2099                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2100                         return -EPERM;
2101                 err = copy_from_user(&rtmsg, arg,
2102                                      sizeof(struct in6_rtmsg));
2103                 if (err)
2104                         return -EFAULT;
2105
2106                 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
2107
2108                 rtnl_lock();
2109                 switch (cmd) {
2110                 case SIOCADDRT:
2111                         err = ip6_route_add(&cfg);
2112                         break;
2113                 case SIOCDELRT:
2114                         err = ip6_route_del(&cfg);
2115                         break;
2116                 default:
2117                         err = -EINVAL;
2118                 }
2119                 rtnl_unlock();
2120
2121                 return err;
2122         }
2123
2124         return -EINVAL;
2125 }
2126
2127 /*
2128  *      Drop the packet on the floor
2129  */
2130
2131 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
2132 {
2133         int type;
2134         struct dst_entry *dst = skb_dst(skb);
2135         switch (ipstats_mib_noroutes) {
2136         case IPSTATS_MIB_INNOROUTES:
2137                 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
2138                 if (type == IPV6_ADDR_ANY) {
2139                         IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2140                                       IPSTATS_MIB_INADDRERRORS);
2141                         break;
2142                 }
2143                 /* FALLTHROUGH */
2144         case IPSTATS_MIB_OUTNOROUTES:
2145                 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2146                               ipstats_mib_noroutes);
2147                 break;
2148         }
2149         icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
2150         kfree_skb(skb);
2151         return 0;
2152 }
2153
2154 static int ip6_pkt_discard(struct sk_buff *skb)
2155 {
2156         return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
2157 }
2158
2159 static int ip6_pkt_discard_out(struct sk_buff *skb)
2160 {
2161         skb->dev = skb_dst(skb)->dev;
2162         return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
2163 }
2164
2165 static int ip6_pkt_prohibit(struct sk_buff *skb)
2166 {
2167         return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
2168 }
2169
2170 static int ip6_pkt_prohibit_out(struct sk_buff *skb)
2171 {
2172         skb->dev = skb_dst(skb)->dev;
2173         return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
2174 }
2175
2176 /*
2177  *      Allocate a dst for local (unicast / anycast) address.
2178  */
2179
2180 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
2181                                     const struct in6_addr *addr,
2182                                     bool anycast)
2183 {
2184         struct net *net = dev_net(idev->dev);
2185         struct rt6_info *rt = ip6_dst_alloc(net, net->loopback_dev,
2186                                             DST_NOCOUNT, NULL);
2187         if (!rt)
2188                 return ERR_PTR(-ENOMEM);
2189
2190         in6_dev_hold(idev);
2191
2192         rt->dst.flags |= DST_HOST;
2193         rt->dst.input = ip6_input;
2194         rt->dst.output = ip6_output;
2195         rt->rt6i_idev = idev;
2196
2197         rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
2198         if (anycast)
2199                 rt->rt6i_flags |= RTF_ANYCAST;
2200         else
2201                 rt->rt6i_flags |= RTF_LOCAL;
2202
2203         rt->rt6i_gateway  = *addr;
2204         rt->rt6i_dst.addr = *addr;
2205         rt->rt6i_dst.plen = 128;
2206         rt->rt6i_table = fib6_get_table(net, RT6_TABLE_LOCAL);
2207
2208         atomic_set(&rt->dst.__refcnt, 1);
2209
2210         return rt;
2211 }
2212
2213 int ip6_route_get_saddr(struct net *net,
2214                         struct rt6_info *rt,
2215                         const struct in6_addr *daddr,
2216                         unsigned int prefs,
2217                         struct in6_addr *saddr)
2218 {
2219         struct inet6_dev *idev = ip6_dst_idev((struct dst_entry*)rt);
2220         int err = 0;
2221         if (rt->rt6i_prefsrc.plen)
2222                 *saddr = rt->rt6i_prefsrc.addr;
2223         else
2224                 err = ipv6_dev_get_saddr(net, idev ? idev->dev : NULL,
2225                                          daddr, prefs, saddr);
2226         return err;
2227 }
2228
2229 /* remove deleted ip from prefsrc entries */
2230 struct arg_dev_net_ip {
2231         struct net_device *dev;
2232         struct net *net;
2233         struct in6_addr *addr;
2234 };
2235
2236 static int fib6_remove_prefsrc(struct rt6_info *rt, void *arg)
2237 {
2238         struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
2239         struct net *net = ((struct arg_dev_net_ip *)arg)->net;
2240         struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
2241
2242         if (((void *)rt->dst.dev == dev || !dev) &&
2243             rt != net->ipv6.ip6_null_entry &&
2244             ipv6_addr_equal(addr, &rt->rt6i_prefsrc.addr)) {
2245                 /* remove prefsrc entry */
2246                 rt->rt6i_prefsrc.plen = 0;
2247         }
2248         return 0;
2249 }
2250
2251 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
2252 {
2253         struct net *net = dev_net(ifp->idev->dev);
2254         struct arg_dev_net_ip adni = {
2255                 .dev = ifp->idev->dev,
2256                 .net = net,
2257                 .addr = &ifp->addr,
2258         };
2259         fib6_clean_all(net, fib6_remove_prefsrc, &adni);
2260 }
2261
2262 struct arg_dev_net {
2263         struct net_device *dev;
2264         struct net *net;
2265 };
2266
2267 static int fib6_ifdown(struct rt6_info *rt, void *arg)
2268 {
2269         const struct arg_dev_net *adn = arg;
2270         const struct net_device *dev = adn->dev;
2271
2272         if ((rt->dst.dev == dev || !dev) &&
2273             rt != adn->net->ipv6.ip6_null_entry)
2274                 return -1;
2275
2276         return 0;
2277 }
2278
2279 void rt6_ifdown(struct net *net, struct net_device *dev)
2280 {
2281         struct arg_dev_net adn = {
2282                 .dev = dev,
2283                 .net = net,
2284         };
2285
2286         fib6_clean_all(net, fib6_ifdown, &adn);
2287         icmp6_clean_all(fib6_ifdown, &adn);
2288 }
2289
2290 struct rt6_mtu_change_arg {
2291         struct net_device *dev;
2292         unsigned int mtu;
2293 };
2294
2295 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2296 {
2297         struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2298         struct inet6_dev *idev;
2299
2300         /* In IPv6 pmtu discovery is not optional,
2301            so that RTAX_MTU lock cannot disable it.
2302            We still use this lock to block changes
2303            caused by addrconf/ndisc.
2304         */
2305
2306         idev = __in6_dev_get(arg->dev);
2307         if (!idev)
2308                 return 0;
2309
2310         /* For administrative MTU increase, there is no way to discover
2311            IPv6 PMTU increase, so PMTU increase should be updated here.
2312            Since RFC 1981 doesn't include administrative MTU increase
2313            update PMTU increase is a MUST. (i.e. jumbo frame)
2314          */
2315         /*
2316            If new MTU is less than route PMTU, this new MTU will be the
2317            lowest MTU in the path, update the route PMTU to reflect PMTU
2318            decreases; if new MTU is greater than route PMTU, and the
2319            old MTU is the lowest MTU in the path, update the route PMTU
2320            to reflect the increase. In this case if the other nodes' MTU
2321            also have the lowest MTU, TOO BIG MESSAGE will be lead to
2322            PMTU discouvery.
2323          */
2324         if (rt->dst.dev == arg->dev &&
2325             !dst_metric_locked(&rt->dst, RTAX_MTU) &&
2326             (dst_mtu(&rt->dst) >= arg->mtu ||
2327              (dst_mtu(&rt->dst) < arg->mtu &&
2328               dst_mtu(&rt->dst) == idev->cnf.mtu6))) {
2329                 dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
2330         }
2331         return 0;
2332 }
2333
2334 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
2335 {
2336         struct rt6_mtu_change_arg arg = {
2337                 .dev = dev,
2338                 .mtu = mtu,
2339         };
2340
2341         fib6_clean_all(dev_net(dev), rt6_mtu_change_route, &arg);
2342 }
2343
2344 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2345         [RTA_GATEWAY]           = { .len = sizeof(struct in6_addr) },
2346         [RTA_OIF]               = { .type = NLA_U32 },
2347         [RTA_IIF]               = { .type = NLA_U32 },
2348         [RTA_PRIORITY]          = { .type = NLA_U32 },
2349         [RTA_METRICS]           = { .type = NLA_NESTED },
2350         [RTA_MULTIPATH]         = { .len = sizeof(struct rtnexthop) },
2351 };
2352
2353 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2354                               struct fib6_config *cfg)
2355 {
2356         struct rtmsg *rtm;
2357         struct nlattr *tb[RTA_MAX+1];
2358         int err;
2359
2360         err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2361         if (err < 0)
2362                 goto errout;
2363
2364         err = -EINVAL;
2365         rtm = nlmsg_data(nlh);
2366         memset(cfg, 0, sizeof(*cfg));
2367
2368         cfg->fc_table = rtm->rtm_table;
2369         cfg->fc_dst_len = rtm->rtm_dst_len;
2370         cfg->fc_src_len = rtm->rtm_src_len;
2371         cfg->fc_flags = RTF_UP;
2372         cfg->fc_protocol = rtm->rtm_protocol;
2373         cfg->fc_type = rtm->rtm_type;
2374
2375         if (rtm->rtm_type == RTN_UNREACHABLE ||
2376             rtm->rtm_type == RTN_BLACKHOLE ||
2377             rtm->rtm_type == RTN_PROHIBIT ||
2378             rtm->rtm_type == RTN_THROW)
2379                 cfg->fc_flags |= RTF_REJECT;
2380
2381         if (rtm->rtm_type == RTN_LOCAL)
2382                 cfg->fc_flags |= RTF_LOCAL;
2383
2384         cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid;
2385         cfg->fc_nlinfo.nlh = nlh;
2386         cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2387
2388         if (tb[RTA_GATEWAY]) {
2389                 nla_memcpy(&cfg->fc_gateway, tb[RTA_GATEWAY], 16);
2390                 cfg->fc_flags |= RTF_GATEWAY;
2391         }
2392
2393         if (tb[RTA_DST]) {
2394                 int plen = (rtm->rtm_dst_len + 7) >> 3;
2395
2396                 if (nla_len(tb[RTA_DST]) < plen)
2397                         goto errout;
2398
2399                 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2400         }
2401
2402         if (tb[RTA_SRC]) {
2403                 int plen = (rtm->rtm_src_len + 7) >> 3;
2404
2405                 if (nla_len(tb[RTA_SRC]) < plen)
2406                         goto errout;
2407
2408                 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2409         }
2410
2411         if (tb[RTA_PREFSRC])
2412                 nla_memcpy(&cfg->fc_prefsrc, tb[RTA_PREFSRC], 16);
2413
2414         if (tb[RTA_OIF])
2415                 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2416
2417         if (tb[RTA_PRIORITY])
2418                 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2419
2420         if (tb[RTA_METRICS]) {
2421                 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2422                 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2423         }
2424
2425         if (tb[RTA_TABLE])
2426                 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2427
2428         if (tb[RTA_MULTIPATH]) {
2429                 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
2430                 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
2431         }
2432
2433         err = 0;
2434 errout:
2435         return err;
2436 }
2437
2438 static int ip6_route_multipath(struct fib6_config *cfg, int add)
2439 {
2440         struct fib6_config r_cfg;
2441         struct rtnexthop *rtnh;
2442         int remaining;
2443         int attrlen;
2444         int err = 0, last_err = 0;
2445
2446 beginning:
2447         rtnh = (struct rtnexthop *)cfg->fc_mp;
2448         remaining = cfg->fc_mp_len;
2449
2450         /* Parse a Multipath Entry */
2451         while (rtnh_ok(rtnh, remaining)) {
2452                 memcpy(&r_cfg, cfg, sizeof(*cfg));
2453                 if (rtnh->rtnh_ifindex)
2454                         r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
2455
2456                 attrlen = rtnh_attrlen(rtnh);
2457                 if (attrlen > 0) {
2458                         struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
2459
2460                         nla = nla_find(attrs, attrlen, RTA_GATEWAY);
2461                         if (nla) {
2462                                 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
2463                                 r_cfg.fc_flags |= RTF_GATEWAY;
2464                         }
2465                 }
2466                 err = add ? ip6_route_add(&r_cfg) : ip6_route_del(&r_cfg);
2467                 if (err) {
2468                         last_err = err;
2469                         /* If we are trying to remove a route, do not stop the
2470                          * loop when ip6_route_del() fails (because next hop is
2471                          * already gone), we should try to remove all next hops.
2472                          */
2473                         if (add) {
2474                                 /* If add fails, we should try to delete all
2475                                  * next hops that have been already added.
2476                                  */
2477                                 add = 0;
2478                                 goto beginning;
2479                         }
2480                 }
2481                 /* Because each route is added like a single route we remove
2482                  * this flag after the first nexthop (if there is a collision,
2483                  * we have already fail to add the first nexthop:
2484                  * fib6_add_rt2node() has reject it).
2485                  */
2486                 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~NLM_F_EXCL;
2487                 rtnh = rtnh_next(rtnh, &remaining);
2488         }
2489
2490         return last_err;
2491 }
2492
2493 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr* nlh)
2494 {
2495         struct fib6_config cfg;
2496         int err;
2497
2498         err = rtm_to_fib6_config(skb, nlh, &cfg);
2499         if (err < 0)
2500                 return err;
2501
2502         if (cfg.fc_mp)
2503                 return ip6_route_multipath(&cfg, 0);
2504         else
2505                 return ip6_route_del(&cfg);
2506 }
2507
2508 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr* nlh)
2509 {
2510         struct fib6_config cfg;
2511         int err;
2512
2513         err = rtm_to_fib6_config(skb, nlh, &cfg);
2514         if (err < 0)
2515                 return err;
2516
2517         if (cfg.fc_mp)
2518                 return ip6_route_multipath(&cfg, 1);
2519         else
2520                 return ip6_route_add(&cfg);
2521 }
2522
2523 static inline size_t rt6_nlmsg_size(void)
2524 {
2525         return NLMSG_ALIGN(sizeof(struct rtmsg))
2526                + nla_total_size(16) /* RTA_SRC */
2527                + nla_total_size(16) /* RTA_DST */
2528                + nla_total_size(16) /* RTA_GATEWAY */
2529                + nla_total_size(16) /* RTA_PREFSRC */
2530                + nla_total_size(4) /* RTA_TABLE */
2531                + nla_total_size(4) /* RTA_IIF */
2532                + nla_total_size(4) /* RTA_OIF */
2533                + nla_total_size(4) /* RTA_PRIORITY */
2534                + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
2535                + nla_total_size(sizeof(struct rta_cacheinfo));
2536 }
2537
2538 static int rt6_fill_node(struct net *net,
2539                          struct sk_buff *skb, struct rt6_info *rt,
2540                          struct in6_addr *dst, struct in6_addr *src,
2541                          int iif, int type, u32 portid, u32 seq,
2542                          int prefix, int nowait, unsigned int flags)
2543 {
2544         struct rtmsg *rtm;
2545         struct nlmsghdr *nlh;
2546         long expires;
2547         u32 table;
2548
2549         if (prefix) {   /* user wants prefix routes only */
2550                 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
2551                         /* success since this is not a prefix route */
2552                         return 1;
2553                 }
2554         }
2555
2556         nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
2557         if (!nlh)
2558                 return -EMSGSIZE;
2559
2560         rtm = nlmsg_data(nlh);
2561         rtm->rtm_family = AF_INET6;
2562         rtm->rtm_dst_len = rt->rt6i_dst.plen;
2563         rtm->rtm_src_len = rt->rt6i_src.plen;
2564         rtm->rtm_tos = 0;
2565         if (rt->rt6i_table)
2566                 table = rt->rt6i_table->tb6_id;
2567         else
2568                 table = RT6_TABLE_UNSPEC;
2569         rtm->rtm_table = table;
2570         if (nla_put_u32(skb, RTA_TABLE, table))
2571                 goto nla_put_failure;
2572         if (rt->rt6i_flags & RTF_REJECT) {
2573                 switch (rt->dst.error) {
2574                 case -EINVAL:
2575                         rtm->rtm_type = RTN_BLACKHOLE;
2576                         break;
2577                 case -EACCES:
2578                         rtm->rtm_type = RTN_PROHIBIT;
2579                         break;
2580                 case -EAGAIN:
2581                         rtm->rtm_type = RTN_THROW;
2582                         break;
2583                 default:
2584                         rtm->rtm_type = RTN_UNREACHABLE;
2585                         break;
2586                 }
2587         }
2588         else if (rt->rt6i_flags & RTF_LOCAL)
2589                 rtm->rtm_type = RTN_LOCAL;
2590         else if (rt->dst.dev && (rt->dst.dev->flags & IFF_LOOPBACK))
2591                 rtm->rtm_type = RTN_LOCAL;
2592         else
2593                 rtm->rtm_type = RTN_UNICAST;
2594         rtm->rtm_flags = 0;
2595         rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2596         rtm->rtm_protocol = rt->rt6i_protocol;
2597         if (rt->rt6i_flags & RTF_DYNAMIC)
2598                 rtm->rtm_protocol = RTPROT_REDIRECT;
2599         else if (rt->rt6i_flags & RTF_ADDRCONF) {
2600                 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ROUTEINFO))
2601                         rtm->rtm_protocol = RTPROT_RA;
2602                 else
2603                         rtm->rtm_protocol = RTPROT_KERNEL;
2604         }
2605
2606         if (rt->rt6i_flags & RTF_CACHE)
2607                 rtm->rtm_flags |= RTM_F_CLONED;
2608
2609         if (dst) {
2610                 if (nla_put(skb, RTA_DST, 16, dst))
2611                         goto nla_put_failure;
2612                 rtm->rtm_dst_len = 128;
2613         } else if (rtm->rtm_dst_len)
2614                 if (nla_put(skb, RTA_DST, 16, &rt->rt6i_dst.addr))
2615                         goto nla_put_failure;
2616 #ifdef CONFIG_IPV6_SUBTREES
2617         if (src) {
2618                 if (nla_put(skb, RTA_SRC, 16, src))
2619                         goto nla_put_failure;
2620                 rtm->rtm_src_len = 128;
2621         } else if (rtm->rtm_src_len &&
2622                    nla_put(skb, RTA_SRC, 16, &rt->rt6i_src.addr))
2623                 goto nla_put_failure;
2624 #endif
2625         if (iif) {
2626 #ifdef CONFIG_IPV6_MROUTE
2627                 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
2628                         int err = ip6mr_get_route(net, skb, rtm, nowait);
2629                         if (err <= 0) {
2630                                 if (!nowait) {
2631                                         if (err == 0)
2632                                                 return 0;
2633                                         goto nla_put_failure;
2634                                 } else {
2635                                         if (err == -EMSGSIZE)
2636                                                 goto nla_put_failure;
2637                                 }
2638                         }
2639                 } else
2640 #endif
2641                         if (nla_put_u32(skb, RTA_IIF, iif))
2642                                 goto nla_put_failure;
2643         } else if (dst) {
2644                 struct in6_addr saddr_buf;
2645                 if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0 &&
2646                     nla_put(skb, RTA_PREFSRC, 16, &saddr_buf))
2647                         goto nla_put_failure;
2648         }
2649
2650         if (rt->rt6i_prefsrc.plen) {
2651                 struct in6_addr saddr_buf;
2652                 saddr_buf = rt->rt6i_prefsrc.addr;
2653                 if (nla_put(skb, RTA_PREFSRC, 16, &saddr_buf))
2654                         goto nla_put_failure;
2655         }
2656
2657         if (rtnetlink_put_metrics(skb, dst_metrics_ptr(&rt->dst)) < 0)
2658                 goto nla_put_failure;
2659
2660         if (rt->rt6i_flags & RTF_GATEWAY) {
2661                 if (nla_put(skb, RTA_GATEWAY, 16, &rt->rt6i_gateway) < 0)
2662                         goto nla_put_failure;
2663         }
2664
2665         if (rt->dst.dev &&
2666             nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex))
2667                 goto nla_put_failure;
2668         if (nla_put_u32(skb, RTA_PRIORITY, rt->rt6i_metric))
2669                 goto nla_put_failure;
2670
2671         expires = (rt->rt6i_flags & RTF_EXPIRES) ? rt->dst.expires - jiffies : 0;
2672
2673         if (rtnl_put_cacheinfo(skb, &rt->dst, 0, expires, rt->dst.error) < 0)
2674                 goto nla_put_failure;
2675
2676         return nlmsg_end(skb, nlh);
2677
2678 nla_put_failure:
2679         nlmsg_cancel(skb, nlh);
2680         return -EMSGSIZE;
2681 }
2682
2683 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
2684 {
2685         struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
2686         int prefix;
2687
2688         if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
2689                 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
2690                 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
2691         } else
2692                 prefix = 0;
2693
2694         return rt6_fill_node(arg->net,
2695                      arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
2696                      NETLINK_CB(arg->cb->skb).portid, arg->cb->nlh->nlmsg_seq,
2697                      prefix, 0, NLM_F_MULTI);
2698 }
2699
2700 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr* nlh)
2701 {
2702         struct net *net = sock_net(in_skb->sk);
2703         struct nlattr *tb[RTA_MAX+1];
2704         struct rt6_info *rt;
2705         struct sk_buff *skb;
2706         struct rtmsg *rtm;
2707         struct flowi6 fl6;
2708         int err, iif = 0, oif = 0;
2709
2710         err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2711         if (err < 0)
2712                 goto errout;
2713
2714         err = -EINVAL;
2715         memset(&fl6, 0, sizeof(fl6));
2716
2717         if (tb[RTA_SRC]) {
2718                 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
2719                         goto errout;
2720
2721                 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
2722         }
2723
2724         if (tb[RTA_DST]) {
2725                 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
2726                         goto errout;
2727
2728                 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
2729         }
2730
2731         if (tb[RTA_IIF])
2732                 iif = nla_get_u32(tb[RTA_IIF]);
2733
2734         if (tb[RTA_OIF])
2735                 oif = nla_get_u32(tb[RTA_OIF]);
2736
2737         if (iif) {
2738                 struct net_device *dev;
2739                 int flags = 0;
2740
2741                 dev = __dev_get_by_index(net, iif);
2742                 if (!dev) {
2743                         err = -ENODEV;
2744                         goto errout;
2745                 }
2746
2747                 fl6.flowi6_iif = iif;
2748
2749                 if (!ipv6_addr_any(&fl6.saddr))
2750                         flags |= RT6_LOOKUP_F_HAS_SADDR;
2751
2752                 rt = (struct rt6_info *)ip6_route_input_lookup(net, dev, &fl6,
2753                                                                flags);
2754         } else {
2755                 fl6.flowi6_oif = oif;
2756
2757                 rt = (struct rt6_info *)ip6_route_output(net, NULL, &fl6);
2758         }
2759
2760         skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
2761         if (!skb) {
2762                 ip6_rt_put(rt);
2763                 err = -ENOBUFS;
2764                 goto errout;
2765         }
2766
2767         /* Reserve room for dummy headers, this skb can pass
2768            through good chunk of routing engine.
2769          */
2770         skb_reset_mac_header(skb);
2771         skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
2772
2773         skb_dst_set(skb, &rt->dst);
2774
2775         err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
2776                             RTM_NEWROUTE, NETLINK_CB(in_skb).portid,
2777                             nlh->nlmsg_seq, 0, 0, 0);
2778         if (err < 0) {
2779                 kfree_skb(skb);
2780                 goto errout;
2781         }
2782
2783         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
2784 errout:
2785         return err;
2786 }
2787
2788 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info)
2789 {
2790         struct sk_buff *skb;
2791         struct net *net = info->nl_net;
2792         u32 seq;
2793         int err;
2794
2795         err = -ENOBUFS;
2796         seq = info->nlh ? info->nlh->nlmsg_seq : 0;
2797
2798         skb = nlmsg_new(rt6_nlmsg_size(), gfp_any());
2799         if (!skb)
2800                 goto errout;
2801
2802         err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
2803                                 event, info->portid, seq, 0, 0, 0);
2804         if (err < 0) {
2805                 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
2806                 WARN_ON(err == -EMSGSIZE);
2807                 kfree_skb(skb);
2808                 goto errout;
2809         }
2810         rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
2811                     info->nlh, gfp_any());
2812         return;
2813 errout:
2814         if (err < 0)
2815                 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
2816 }
2817
2818 static int ip6_route_dev_notify(struct notifier_block *this,
2819                                 unsigned long event, void *ptr)
2820 {
2821         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2822         struct net *net = dev_net(dev);
2823
2824         if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
2825                 net->ipv6.ip6_null_entry->dst.dev = dev;
2826                 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
2827 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2828                 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
2829                 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
2830                 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
2831                 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
2832 #endif
2833         }
2834
2835         return NOTIFY_OK;
2836 }
2837
2838 /*
2839  *      /proc
2840  */
2841
2842 #ifdef CONFIG_PROC_FS
2843
2844 static const struct file_operations ipv6_route_proc_fops = {
2845         .owner          = THIS_MODULE,
2846         .open           = ipv6_route_open,
2847         .read           = seq_read,
2848         .llseek         = seq_lseek,
2849         .release        = seq_release_net,
2850 };
2851
2852 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
2853 {
2854         struct net *net = (struct net *)seq->private;
2855         seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
2856                    net->ipv6.rt6_stats->fib_nodes,
2857                    net->ipv6.rt6_stats->fib_route_nodes,
2858                    net->ipv6.rt6_stats->fib_rt_alloc,
2859                    net->ipv6.rt6_stats->fib_rt_entries,
2860                    net->ipv6.rt6_stats->fib_rt_cache,
2861                    dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
2862                    net->ipv6.rt6_stats->fib_discarded_routes);
2863
2864         return 0;
2865 }
2866
2867 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
2868 {
2869         return single_open_net(inode, file, rt6_stats_seq_show);
2870 }
2871
2872 static const struct file_operations rt6_stats_seq_fops = {
2873         .owner   = THIS_MODULE,
2874         .open    = rt6_stats_seq_open,
2875         .read    = seq_read,
2876         .llseek  = seq_lseek,
2877         .release = single_release_net,
2878 };
2879 #endif  /* CONFIG_PROC_FS */
2880
2881 #ifdef CONFIG_SYSCTL
2882
2883 static
2884 int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
2885                               void __user *buffer, size_t *lenp, loff_t *ppos)
2886 {
2887         struct net *net;
2888         int delay;
2889         if (!write)
2890                 return -EINVAL;
2891
2892         net = (struct net *)ctl->extra1;
2893         delay = net->ipv6.sysctl.flush_delay;
2894         proc_dointvec(ctl, write, buffer, lenp, ppos);
2895         fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
2896         return 0;
2897 }
2898
2899 struct ctl_table ipv6_route_table_template[] = {
2900         {
2901                 .procname       =       "flush",
2902                 .data           =       &init_net.ipv6.sysctl.flush_delay,
2903                 .maxlen         =       sizeof(int),
2904                 .mode           =       0200,
2905                 .proc_handler   =       ipv6_sysctl_rtcache_flush
2906         },
2907         {
2908                 .procname       =       "gc_thresh",
2909                 .data           =       &ip6_dst_ops_template.gc_thresh,
2910                 .maxlen         =       sizeof(int),
2911                 .mode           =       0644,
2912                 .proc_handler   =       proc_dointvec,
2913         },
2914         {
2915                 .procname       =       "max_size",
2916                 .data           =       &init_net.ipv6.sysctl.ip6_rt_max_size,
2917                 .maxlen         =       sizeof(int),
2918                 .mode           =       0644,
2919                 .proc_handler   =       proc_dointvec,
2920         },
2921         {
2922                 .procname       =       "gc_min_interval",
2923                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2924                 .maxlen         =       sizeof(int),
2925                 .mode           =       0644,
2926                 .proc_handler   =       proc_dointvec_jiffies,
2927         },
2928         {
2929                 .procname       =       "gc_timeout",
2930                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
2931                 .maxlen         =       sizeof(int),
2932                 .mode           =       0644,
2933                 .proc_handler   =       proc_dointvec_jiffies,
2934         },
2935         {
2936                 .procname       =       "gc_interval",
2937                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_interval,
2938                 .maxlen         =       sizeof(int),
2939                 .mode           =       0644,
2940                 .proc_handler   =       proc_dointvec_jiffies,
2941         },
2942         {
2943                 .procname       =       "gc_elasticity",
2944                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
2945                 .maxlen         =       sizeof(int),
2946                 .mode           =       0644,
2947                 .proc_handler   =       proc_dointvec,
2948         },
2949         {
2950                 .procname       =       "mtu_expires",
2951                 .data           =       &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
2952                 .maxlen         =       sizeof(int),
2953                 .mode           =       0644,
2954                 .proc_handler   =       proc_dointvec_jiffies,
2955         },
2956         {
2957                 .procname       =       "min_adv_mss",
2958                 .data           =       &init_net.ipv6.sysctl.ip6_rt_min_advmss,
2959                 .maxlen         =       sizeof(int),
2960                 .mode           =       0644,
2961                 .proc_handler   =       proc_dointvec,
2962         },
2963         {
2964                 .procname       =       "gc_min_interval_ms",
2965                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2966                 .maxlen         =       sizeof(int),
2967                 .mode           =       0644,
2968                 .proc_handler   =       proc_dointvec_ms_jiffies,
2969         },
2970         { }
2971 };
2972
2973 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
2974 {
2975         struct ctl_table *table;
2976
2977         table = kmemdup(ipv6_route_table_template,
2978                         sizeof(ipv6_route_table_template),
2979                         GFP_KERNEL);
2980
2981         if (table) {
2982                 table[0].data = &net->ipv6.sysctl.flush_delay;
2983                 table[0].extra1 = net;
2984                 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
2985                 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
2986                 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2987                 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
2988                 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
2989                 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
2990                 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
2991                 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
2992                 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2993
2994                 /* Don't export sysctls to unprivileged users */
2995                 if (net->user_ns != &init_user_ns)
2996                         table[0].procname = NULL;
2997         }
2998
2999         return table;
3000 }
3001 #endif
3002
3003 static int __net_init ip6_route_net_init(struct net *net)
3004 {
3005         int ret = -ENOMEM;
3006
3007         memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
3008                sizeof(net->ipv6.ip6_dst_ops));
3009
3010         if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
3011                 goto out_ip6_dst_ops;
3012
3013         net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
3014                                            sizeof(*net->ipv6.ip6_null_entry),
3015                                            GFP_KERNEL);
3016         if (!net->ipv6.ip6_null_entry)
3017                 goto out_ip6_dst_entries;
3018         net->ipv6.ip6_null_entry->dst.path =
3019                 (struct dst_entry *)net->ipv6.ip6_null_entry;
3020         net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3021         dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
3022                          ip6_template_metrics, true);
3023
3024 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3025         net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
3026                                                sizeof(*net->ipv6.ip6_prohibit_entry),
3027                                                GFP_KERNEL);
3028         if (!net->ipv6.ip6_prohibit_entry)
3029                 goto out_ip6_null_entry;
3030         net->ipv6.ip6_prohibit_entry->dst.path =
3031                 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
3032         net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3033         dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
3034                          ip6_template_metrics, true);
3035
3036         net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
3037                                                sizeof(*net->ipv6.ip6_blk_hole_entry),
3038                                                GFP_KERNEL);
3039         if (!net->ipv6.ip6_blk_hole_entry)
3040                 goto out_ip6_prohibit_entry;
3041         net->ipv6.ip6_blk_hole_entry->dst.path =
3042                 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
3043         net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3044         dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
3045                          ip6_template_metrics, true);
3046 #endif
3047
3048         net->ipv6.sysctl.flush_delay = 0;
3049         net->ipv6.sysctl.ip6_rt_max_size = 4096;
3050         net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
3051         net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
3052         net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
3053         net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
3054         net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
3055         net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
3056
3057         net->ipv6.ip6_rt_gc_expire = 30*HZ;
3058
3059         ret = 0;
3060 out:
3061         return ret;
3062
3063 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3064 out_ip6_prohibit_entry:
3065         kfree(net->ipv6.ip6_prohibit_entry);
3066 out_ip6_null_entry:
3067         kfree(net->ipv6.ip6_null_entry);
3068 #endif
3069 out_ip6_dst_entries:
3070         dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3071 out_ip6_dst_ops:
3072         goto out;
3073 }
3074
3075 static void __net_exit ip6_route_net_exit(struct net *net)
3076 {
3077         kfree(net->ipv6.ip6_null_entry);
3078 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3079         kfree(net->ipv6.ip6_prohibit_entry);
3080         kfree(net->ipv6.ip6_blk_hole_entry);
3081 #endif
3082         dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3083 }
3084
3085 static int __net_init ip6_route_net_init_late(struct net *net)
3086 {
3087 #ifdef CONFIG_PROC_FS
3088         proc_create("ipv6_route", 0, net->proc_net, &ipv6_route_proc_fops);
3089         proc_create("rt6_stats", S_IRUGO, net->proc_net, &rt6_stats_seq_fops);
3090 #endif
3091         return 0;
3092 }
3093
3094 static void __net_exit ip6_route_net_exit_late(struct net *net)
3095 {
3096 #ifdef CONFIG_PROC_FS
3097         remove_proc_entry("ipv6_route", net->proc_net);
3098         remove_proc_entry("rt6_stats", net->proc_net);
3099 #endif
3100 }
3101
3102 static struct pernet_operations ip6_route_net_ops = {
3103         .init = ip6_route_net_init,
3104         .exit = ip6_route_net_exit,
3105 };
3106
3107 static int __net_init ipv6_inetpeer_init(struct net *net)
3108 {
3109         struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
3110
3111         if (!bp)
3112                 return -ENOMEM;
3113         inet_peer_base_init(bp);
3114         net->ipv6.peers = bp;
3115         return 0;
3116 }
3117
3118 static void __net_exit ipv6_inetpeer_exit(struct net *net)
3119 {
3120         struct inet_peer_base *bp = net->ipv6.peers;
3121
3122         net->ipv6.peers = NULL;
3123         inetpeer_invalidate_tree(bp);
3124         kfree(bp);
3125 }
3126
3127 static struct pernet_operations ipv6_inetpeer_ops = {
3128         .init   =       ipv6_inetpeer_init,
3129         .exit   =       ipv6_inetpeer_exit,
3130 };
3131
3132 static struct pernet_operations ip6_route_net_late_ops = {
3133         .init = ip6_route_net_init_late,
3134         .exit = ip6_route_net_exit_late,
3135 };
3136
3137 static struct notifier_block ip6_route_dev_notifier = {
3138         .notifier_call = ip6_route_dev_notify,
3139         .priority = 0,
3140 };
3141
3142 int __init ip6_route_init(void)
3143 {
3144         int ret;
3145
3146         ret = -ENOMEM;
3147         ip6_dst_ops_template.kmem_cachep =
3148                 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
3149                                   SLAB_HWCACHE_ALIGN, NULL);
3150         if (!ip6_dst_ops_template.kmem_cachep)
3151                 goto out;
3152
3153         ret = dst_entries_init(&ip6_dst_blackhole_ops);
3154         if (ret)
3155                 goto out_kmem_cache;
3156
3157         ret = register_pernet_subsys(&ipv6_inetpeer_ops);
3158         if (ret)
3159                 goto out_dst_entries;
3160
3161         ret = register_pernet_subsys(&ip6_route_net_ops);
3162         if (ret)
3163                 goto out_register_inetpeer;
3164
3165         ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
3166
3167         /* Registering of the loopback is done before this portion of code,
3168          * the loopback reference in rt6_info will not be taken, do it
3169          * manually for init_net */
3170         init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
3171         init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3172   #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3173         init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
3174         init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3175         init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
3176         init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3177   #endif
3178         ret = fib6_init();
3179         if (ret)
3180                 goto out_register_subsys;
3181
3182         ret = xfrm6_init();
3183         if (ret)
3184                 goto out_fib6_init;
3185
3186         ret = fib6_rules_init();
3187         if (ret)
3188                 goto xfrm6_init;
3189
3190         ret = register_pernet_subsys(&ip6_route_net_late_ops);
3191         if (ret)
3192                 goto fib6_rules_init;
3193
3194         ret = -ENOBUFS;
3195         if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL, NULL) ||
3196             __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL, NULL) ||
3197             __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL, NULL))
3198                 goto out_register_late_subsys;
3199
3200         ret = register_netdevice_notifier(&ip6_route_dev_notifier);
3201         if (ret)
3202                 goto out_register_late_subsys;
3203
3204 out:
3205         return ret;
3206
3207 out_register_late_subsys:
3208         unregister_pernet_subsys(&ip6_route_net_late_ops);
3209 fib6_rules_init:
3210         fib6_rules_cleanup();
3211 xfrm6_init:
3212         xfrm6_fini();
3213 out_fib6_init:
3214         fib6_gc_cleanup();
3215 out_register_subsys:
3216         unregister_pernet_subsys(&ip6_route_net_ops);
3217 out_register_inetpeer:
3218         unregister_pernet_subsys(&ipv6_inetpeer_ops);
3219 out_dst_entries:
3220         dst_entries_destroy(&ip6_dst_blackhole_ops);
3221 out_kmem_cache:
3222         kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3223         goto out;
3224 }
3225
3226 void ip6_route_cleanup(void)
3227 {
3228         unregister_netdevice_notifier(&ip6_route_dev_notifier);
3229         unregister_pernet_subsys(&ip6_route_net_late_ops);
3230         fib6_rules_cleanup();
3231         xfrm6_fini();
3232         fib6_gc_cleanup();
3233         unregister_pernet_subsys(&ipv6_inetpeer_ops);
3234         unregister_pernet_subsys(&ip6_route_net_ops);
3235         dst_entries_destroy(&ip6_dst_blackhole_ops);
3236         kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3237 }