net/ipv6/route.c

   1 /*
   2  *      Linux INET6 implementation
   3  *      FIB front-end.
   4  *
   5  *      Authors:
   6  *      Pedro Roque             <roque@di.fc.ul.pt>
   7  *
   8  *      This program is free software; you can redistribute it and/or
   9  *      modify it under the terms of the GNU General Public License
  10  *      as published by the Free Software Foundation; either version
  11  *      2 of the License, or (at your option) any later version.
  12  */
  13
  14 /*      Changes:
  15  *
  16  *      YOSHIFUJI Hideaki @USAGI
  17  *              reworked default router selection.
  18  *              - respect outgoing interface
  19  *              - select from (probably) reachable routers (i.e.
  20  *              routers in REACHABLE, STALE, DELAY or PROBE states).
  21  *              - always select the same router if it is (probably)
  22  *              reachable.  otherwise, round-robin the list.
  23  *      Ville Nuorvala
  24  *              Fixed routing subtrees.
  25  */
  26
  27 #define pr_fmt(fmt) "IPv6: " fmt
  28
  29 #include <linux/capability.h>
  30 #include <linux/errno.h>
  31 #include <linux/export.h>
  32 #include <linux/types.h>
  33 #include <linux/times.h>
  34 #include <linux/socket.h>
  35 #include <linux/sockios.h>
  36 #include <linux/net.h>
  37 #include <linux/route.h>
  38 #include <linux/netdevice.h>
  39 #include <linux/in6.h>
  40 #include <linux/mroute6.h>
  41 #include <linux/init.h>
  42 #include <linux/if_arp.h>
  43 #include <linux/proc_fs.h>
  44 #include <linux/seq_file.h>
  45 #include <linux/nsproxy.h>
  46 #include <linux/slab.h>
  47 #include <net/net_namespace.h>
  48 #include <net/snmp.h>
  49 #include <net/ipv6.h>
  50 #include <net/ip6_fib.h>
  51 #include <net/ip6_route.h>
  52 #include <net/ndisc.h>
  53 #include <net/addrconf.h>
  54 #include <net/tcp.h>
  55 #include <linux/rtnetlink.h>
  56 #include <net/dst.h>
  57 #include <net/xfrm.h>
  58 #include <net/netevent.h>
  59 #include <net/netlink.h>
  60 #include <net/nexthop.h>
  61
  62 #include <asm/uaccess.h>
  63
  64 #ifdef CONFIG_SYSCTL
  65 #include <linux/sysctl.h>
  66 #endif
  67
  68 enum rt6_nud_state {
  69         RT6_NUD_FAIL_HARD = -3,
  70         RT6_NUD_FAIL_PROBE = -2,
  71         RT6_NUD_FAIL_DO_RR = -1,
  72         RT6_NUD_SUCCEED = 1
  73 };
  74
  75 static struct rt6_info *ip6_rt_copy(struct rt6_info *ort,
  76                                     const struct in6_addr *dest);
  77 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie);
  78 static unsigned int      ip6_default_advmss(const struct dst_entry *dst);
  79 static unsigned int      ip6_mtu(const struct dst_entry *dst);
  80 static struct dst_entry *ip6_negative_advice(struct dst_entry *);
  81 static void             ip6_dst_destroy(struct dst_entry *);
  82 static void             ip6_dst_ifdown(struct dst_entry *,
  83                                        struct net_device *dev, int how);
  84 static int               ip6_dst_gc(struct dst_ops *ops);
  85
  86 static int              ip6_pkt_discard(struct sk_buff *skb);
  87 static int              ip6_pkt_discard_out(struct sock *sk, struct sk_buff *skb);
  88 static int              ip6_pkt_prohibit(struct sk_buff *skb);
  89 static int              ip6_pkt_prohibit_out(struct sock *sk, struct sk_buff *skb);
  90 static void             ip6_link_failure(struct sk_buff *skb);
  91 static void             ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
  92                                            struct sk_buff *skb, u32 mtu);
  93 static void             rt6_do_redirect(struct dst_entry *dst, struct sock *sk,
  94                                         struct sk_buff *skb);
  95 static int rt6_score_route(struct rt6_info *rt, int oif, int strict);
  96
  97 #ifdef CONFIG_IPV6_ROUTE_INFO
  98 static struct rt6_info *rt6_add_route_info(struct net *net,
  99                                            const struct in6_addr *prefix, int prefixlen,
 100                                            const struct in6_addr *gwaddr, int ifindex,
 101                                            unsigned int pref);
 102 static struct rt6_info *rt6_get_route_info(struct net *net,
 103                                            const struct in6_addr *prefix, int prefixlen,
 104                                            const struct in6_addr *gwaddr, int ifindex);
 105 #endif
 106
 107 static void rt6_bind_peer(struct rt6_info *rt, int create)
 108 {
 109         struct inet_peer_base *base;
 110         struct inet_peer *peer;
 111
 112         base = inetpeer_base_ptr(rt->_rt6i_peer);
 113         if (!base)
 114                 return;
 115
 116         peer = inet_getpeer_v6(base, &rt->rt6i_dst.addr, create);
 117         if (peer) {
 118                 if (!rt6_set_peer(rt, peer))
 119                         inet_putpeer(peer);
 120         }
 121 }
 122
 123 static struct inet_peer *__rt6_get_peer(struct rt6_info *rt, int create)
 124 {
 125         if (rt6_has_peer(rt))
 126                 return rt6_peer_ptr(rt);
 127
 128         rt6_bind_peer(rt, create);
 129         return (rt6_has_peer(rt) ? rt6_peer_ptr(rt) : NULL);
 130 }
 131
 132 static struct inet_peer *rt6_get_peer_create(struct rt6_info *rt)
 133 {
 134         return __rt6_get_peer(rt, 1);
 135 }
 136
 137 static u32 *ipv6_cow_metrics(struct dst_entry *dst, unsigned long old)
 138 {
 139         struct rt6_info *rt = (struct rt6_info *) dst;
 140         struct inet_peer *peer;
 141         u32 *p = NULL;
 142
 143         if (!(rt->dst.flags & DST_HOST))
 144                 return NULL;
 145
 146         peer = rt6_get_peer_create(rt);
 147         if (peer) {
 148                 u32 *old_p = __DST_METRICS_PTR(old);
 149                 unsigned long prev, new;
 150
 151                 p = peer->metrics;
 152                 if (inet_metrics_new(peer) ||
 153                     (old & DST_METRICS_FORCE_OVERWRITE))
 154                         memcpy(p, old_p, sizeof(u32) * RTAX_MAX);
 155
 156                 new = (unsigned long) p;
 157                 prev = cmpxchg(&dst->_metrics, old, new);
 158
 159                 if (prev != old) {
 160                         p = __DST_METRICS_PTR(prev);
 161                         if (prev & DST_METRICS_READ_ONLY)
 162                                 p = NULL;
 163                 }
 164         }
 165         return p;
 166 }
 167
 168 static inline const void *choose_neigh_daddr(struct rt6_info *rt,
 169                                              struct sk_buff *skb,
 170                                              const void *daddr)
 171 {
 172         struct in6_addr *p = &rt->rt6i_gateway;
 173
 174         if (!ipv6_addr_any(p))
 175                 return (const void *) p;
 176         else if (skb)
 177                 return &ipv6_hdr(skb)->daddr;
 178         return daddr;
 179 }
 180
 181 static struct neighbour *ip6_neigh_lookup(const struct dst_entry *dst,
 182                                           struct sk_buff *skb,
 183                                           const void *daddr)
 184 {
 185         struct rt6_info *rt = (struct rt6_info *) dst;
 186         struct neighbour *n;
 187
 188         daddr = choose_neigh_daddr(rt, skb, daddr);
 189         n = __ipv6_neigh_lookup(dst->dev, daddr);
 190         if (n)
 191                 return n;
 192         return neigh_create(&nd_tbl, daddr, dst->dev);
 193 }
 194
 195 static struct dst_ops ip6_dst_ops_template = {
 196         .family                 =       AF_INET6,
 197         .protocol               =       cpu_to_be16(ETH_P_IPV6),
 198         .gc                     =       ip6_dst_gc,
 199         .gc_thresh              =       1024,
 200         .check                  =       ip6_dst_check,
 201         .default_advmss         =       ip6_default_advmss,
 202         .mtu                    =       ip6_mtu,
 203         .cow_metrics            =       ipv6_cow_metrics,
 204         .destroy                =       ip6_dst_destroy,
 205         .ifdown                 =       ip6_dst_ifdown,
 206         .negative_advice        =       ip6_negative_advice,
 207         .link_failure           =       ip6_link_failure,
 208         .update_pmtu            =       ip6_rt_update_pmtu,
 209         .redirect               =       rt6_do_redirect,
 210         .local_out              =       __ip6_local_out,
 211         .neigh_lookup           =       ip6_neigh_lookup,
 212 };
 213
 214 static unsigned int ip6_blackhole_mtu(const struct dst_entry *dst)
 215 {
 216         unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
 217
 218         return mtu ? : dst->dev->mtu;
 219 }
 220
 221 static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, struct sock *sk,
 222                                          struct sk_buff *skb, u32 mtu)
 223 {
 224 }
 225
 226 static void ip6_rt_blackhole_redirect(struct dst_entry *dst, struct sock *sk,
 227                                       struct sk_buff *skb)
 228 {
 229 }
 230
 231 static u32 *ip6_rt_blackhole_cow_metrics(struct dst_entry *dst,
 232                                          unsigned long old)
 233 {
 234         return NULL;
 235 }
 236
 237 static struct dst_ops ip6_dst_blackhole_ops = {
 238         .family                 =       AF_INET6,
 239         .protocol               =       cpu_to_be16(ETH_P_IPV6),
 240         .destroy                =       ip6_dst_destroy,
 241         .check                  =       ip6_dst_check,
 242         .mtu                    =       ip6_blackhole_mtu,
 243         .default_advmss         =       ip6_default_advmss,
 244         .update_pmtu            =       ip6_rt_blackhole_update_pmtu,
 245         .redirect               =       ip6_rt_blackhole_redirect,
 246         .cow_metrics            =       ip6_rt_blackhole_cow_metrics,
 247         .neigh_lookup           =       ip6_neigh_lookup,
 248 };
 249
 250 static const u32 ip6_template_metrics[RTAX_MAX] = {
 251         [RTAX_HOPLIMIT - 1] = 0,
 252 };
 253
 254 static const struct rt6_info ip6_null_entry_template = {
 255         .dst = {
 256                 .__refcnt       = ATOMIC_INIT(1),
 257                 .__use          = 1,
 258                 .obsolete       = DST_OBSOLETE_FORCE_CHK,
 259                 .error          = -ENETUNREACH,
 260                 .input          = ip6_pkt_discard,
 261                 .output         = ip6_pkt_discard_out,
 262         },
 263         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
 264         .rt6i_protocol  = RTPROT_KERNEL,
 265         .rt6i_metric    = ~(u32) 0,
 266         .rt6i_ref       = ATOMIC_INIT(1),
 267 };
 268
 269 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
 270
 271 static const struct rt6_info ip6_prohibit_entry_template = {
 272         .dst = {
 273                 .__refcnt       = ATOMIC_INIT(1),
 274                 .__use          = 1,
 275                 .obsolete       = DST_OBSOLETE_FORCE_CHK,
 276                 .error          = -EACCES,
 277                 .input          = ip6_pkt_prohibit,
 278                 .output         = ip6_pkt_prohibit_out,
 279         },
 280         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
 281         .rt6i_protocol  = RTPROT_KERNEL,
 282         .rt6i_metric    = ~(u32) 0,
 283         .rt6i_ref       = ATOMIC_INIT(1),
 284 };
 285
 286 static const struct rt6_info ip6_blk_hole_entry_template = {
 287         .dst = {
 288                 .__refcnt       = ATOMIC_INIT(1),
 289                 .__use          = 1,
 290                 .obsolete       = DST_OBSOLETE_FORCE_CHK,
 291                 .error          = -EINVAL,
 292                 .input          = dst_discard,
 293                 .output         = dst_discard_sk,
 294         },
 295         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
 296         .rt6i_protocol  = RTPROT_KERNEL,
 297         .rt6i_metric    = ~(u32) 0,
 298         .rt6i_ref       = ATOMIC_INIT(1),
 299 };
 300
 301 #endif
 302
 303 /* allocate dst with ip6_dst_ops */
 304 static inline struct rt6_info *ip6_dst_alloc(struct net *net,
 305                                              struct net_device *dev,
 306                                              int flags,
 307                                              struct fib6_table *table)
 308 {
 309         struct rt6_info *rt = dst_alloc(&net->ipv6.ip6_dst_ops, dev,
 310                                         0, DST_OBSOLETE_FORCE_CHK, flags);
 311
 312         if (rt) {
 313                 struct dst_entry *dst = &rt->dst;
 314
 315                 memset(dst + 1, 0, sizeof(*rt) - sizeof(*dst));
 316                 rt6_init_peer(rt, table ? &table->tb6_peers : net->ipv6.peers);
 317                 INIT_LIST_HEAD(&rt->rt6i_siblings);
 318         }
 319         return rt;
 320 }
 321
 322 static void ip6_dst_destroy(struct dst_entry *dst)
 323 {
 324         struct rt6_info *rt = (struct rt6_info *)dst;
 325         struct inet6_dev *idev = rt->rt6i_idev;
 326         struct dst_entry *from = dst->from;
 327
 328         if (!(rt->dst.flags & DST_HOST))
 329                 dst_destroy_metrics_generic(dst);
 330
 331         if (idev) {
 332                 rt->rt6i_idev = NULL;
 333                 in6_dev_put(idev);
 334         }
 335
 336         dst->from = NULL;
 337         dst_release(from);
 338
 339         if (rt6_has_peer(rt)) {
 340                 struct inet_peer *peer = rt6_peer_ptr(rt);
 341                 inet_putpeer(peer);
 342         }
 343 }
 344
 345 static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
 346                            int how)
 347 {
 348         struct rt6_info *rt = (struct rt6_info *)dst;
 349         struct inet6_dev *idev = rt->rt6i_idev;
 350         struct net_device *loopback_dev =
 351                 dev_net(dev)->loopback_dev;
 352
 353         if (dev != loopback_dev) {
 354                 if (idev && idev->dev == dev) {
 355                         struct inet6_dev *loopback_idev =
 356                                 in6_dev_get(loopback_dev);
 357                         if (loopback_idev) {
 358                                 rt->rt6i_idev = loopback_idev;
 359                                 in6_dev_put(idev);
 360                         }
 361                 }
 362         }
 363 }
 364
 365 static bool rt6_check_expired(const struct rt6_info *rt)
 366 {
 367         if (rt->rt6i_flags & RTF_EXPIRES) {
 368                 if (time_after(jiffies, rt->dst.expires))
 369                         return true;
 370         } else if (rt->dst.from) {
 371                 return rt6_check_expired((struct rt6_info *) rt->dst.from);
 372         }
 373         return false;
 374 }
 375
 376 /* Multipath route selection:
 377  *   Hash based function using packet header and flowlabel.
 378  * Adapted from fib_info_hashfn()
 379  */
 380 static int rt6_info_hash_nhsfn(unsigned int candidate_count,
 381                                const struct flowi6 *fl6)
 382 {
 383         unsigned int val = fl6->flowi6_proto;
 384
 385         val ^= ipv6_addr_hash(&fl6->daddr);
 386         val ^= ipv6_addr_hash(&fl6->saddr);
 387
 388         /* Work only if this not encapsulated */
 389         switch (fl6->flowi6_proto) {
 390         case IPPROTO_UDP:
 391         case IPPROTO_TCP:
 392         case IPPROTO_SCTP:
 393                 val ^= (__force u16)fl6->fl6_sport;
 394                 val ^= (__force u16)fl6->fl6_dport;
 395                 break;
 396
 397         case IPPROTO_ICMPV6:
 398                 val ^= (__force u16)fl6->fl6_icmp_type;
 399                 val ^= (__force u16)fl6->fl6_icmp_code;
 400                 break;
 401         }
 402         /* RFC6438 recommands to use flowlabel */
 403         val ^= (__force u32)fl6->flowlabel;
 404
 405         /* Perhaps, we need to tune, this function? */
 406         val = val ^ (val >> 7) ^ (val >> 12);
 407         return val % candidate_count;
 408 }
 409
 410 static struct rt6_info *rt6_multipath_select(struct rt6_info *match,
 411                                              struct flowi6 *fl6, int oif,
 412                                              int strict)
 413 {
 414         struct rt6_info *sibling, *next_sibling;
 415         int route_choosen;
 416
 417         route_choosen = rt6_info_hash_nhsfn(match->rt6i_nsiblings + 1, fl6);
 418         /* Don't change the route, if route_choosen == 0
 419          * (siblings does not include ourself)
 420          */
 421         if (route_choosen)
 422                 list_for_each_entry_safe(sibling, next_sibling,
 423                                 &match->rt6i_siblings, rt6i_siblings) {
 424                         route_choosen--;
 425                         if (route_choosen == 0) {
 426                                 if (rt6_score_route(sibling, oif, strict) < 0)
 427                                         break;
 428                                 match = sibling;
 429                                 break;
 430                         }
 431                 }
 432         return match;
 433 }
 434
 435 /*
 436  *      Route lookup. Any table->tb6_lock is implied.
 437  */
 438
 439 static inline struct rt6_info *rt6_device_match(struct net *net,
 440                                                     struct rt6_info *rt,
 441                                                     const struct in6_addr *saddr,
 442                                                     int oif,
 443                                                     int flags)
 444 {
 445         struct rt6_info *local = NULL;
 446         struct rt6_info *sprt;
 447
 448         if (!oif && ipv6_addr_any(saddr))
 449                 goto out;
 450
 451         for (sprt = rt; sprt; sprt = sprt->dst.rt6_next) {
 452                 struct net_device *dev = sprt->dst.dev;
 453
 454                 if (oif) {
 455                         if (dev->ifindex == oif)
 456                                 return sprt;
 457                         if (dev->flags & IFF_LOOPBACK) {
 458                                 if (!sprt->rt6i_idev ||
 459                                     sprt->rt6i_idev->dev->ifindex != oif) {
 460                                         if (flags & RT6_LOOKUP_F_IFACE && oif)
 461                                                 continue;
 462                                         if (local && (!oif ||
 463                                                       local->rt6i_idev->dev->ifindex == oif))
 464                                                 continue;
 465                                 }
 466                                 local = sprt;
 467                         }
 468                 } else {
 469                         if (ipv6_chk_addr(net, saddr, dev,
 470                                           flags & RT6_LOOKUP_F_IFACE))
 471                                 return sprt;
 472                 }
 473         }
 474
 475         if (oif) {
 476                 if (local)
 477                         return local;
 478
 479                 if (flags & RT6_LOOKUP_F_IFACE)
 480                         return net->ipv6.ip6_null_entry;
 481         }
 482 out:
 483         return rt;
 484 }
 485
 486 #ifdef CONFIG_IPV6_ROUTER_PREF
 487 struct __rt6_probe_work {
 488         struct work_struct work;
 489         struct in6_addr target;
 490         struct net_device *dev;
 491 };
 492
 493 static void rt6_probe_deferred(struct work_struct *w)
 494 {
 495         struct in6_addr mcaddr;
 496         struct __rt6_probe_work *work =
 497                 container_of(w, struct __rt6_probe_work, work);
 498
 499         addrconf_addr_solict_mult(&work->target, &mcaddr);
 500         ndisc_send_ns(work->dev, NULL, &work->target, &mcaddr, NULL);
 501         dev_put(work->dev);
 502         kfree(w);
 503 }
 504
 505 static void rt6_probe(struct rt6_info *rt)
 506 {
 507         struct neighbour *neigh;
 508         /*
 509          * Okay, this does not seem to be appropriate
 510          * for now, however, we need to check if it
 511          * is really so; aka Router Reachability Probing.
 512          *
 513          * Router Reachability Probe MUST be rate-limited
 514          * to no more than one per minute.
 515          */
 516         if (!rt || !(rt->rt6i_flags & RTF_GATEWAY))
 517                 return;
 518         rcu_read_lock_bh();
 519         neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
 520         if (neigh) {
 521                 write_lock(&neigh->lock);
 522                 if (neigh->nud_state & NUD_VALID)
 523                         goto out;
 524         }
 525
 526         if (!neigh ||
 527             time_after(jiffies, neigh->updated + rt->rt6i_idev->cnf.rtr_probe_interval)) {
 528                 struct __rt6_probe_work *work;
 529
 530                 work = kmalloc(sizeof(*work), GFP_ATOMIC);
 531
 532                 if (neigh && work)
 533                         __neigh_set_probe_once(neigh);
 534
 535                 if (neigh)
 536                         write_unlock(&neigh->lock);
 537
 538                 if (work) {
 539                         INIT_WORK(&work->work, rt6_probe_deferred);
 540                         work->target = rt->rt6i_gateway;
 541                         dev_hold(rt->dst.dev);
 542                         work->dev = rt->dst.dev;
 543                         schedule_work(&work->work);
 544                 }
 545         } else {
 546 out:
 547                 write_unlock(&neigh->lock);
 548         }
 549         rcu_read_unlock_bh();
 550 }
 551 #else
 552 static inline void rt6_probe(struct rt6_info *rt)
 553 {
 554 }
 555 #endif
 556
 557 /*
 558  * Default Router Selection (RFC 2461 6.3.6)
 559  */
 560 static inline int rt6_check_dev(struct rt6_info *rt, int oif)
 561 {
 562         struct net_device *dev = rt->dst.dev;
 563         if (!oif || dev->ifindex == oif)
 564                 return 2;
 565         if ((dev->flags & IFF_LOOPBACK) &&
 566             rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
 567                 return 1;
 568         return 0;
 569 }
 570
 571 static inline enum rt6_nud_state rt6_check_neigh(struct rt6_info *rt)
 572 {
 573         struct neighbour *neigh;
 574         enum rt6_nud_state ret = RT6_NUD_FAIL_HARD;
 575
 576         if (rt->rt6i_flags & RTF_NONEXTHOP ||
 577             !(rt->rt6i_flags & RTF_GATEWAY))
 578                 return RT6_NUD_SUCCEED;
 579
 580         rcu_read_lock_bh();
 581         neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
 582         if (neigh) {
 583                 read_lock(&neigh->lock);
 584                 if (neigh->nud_state & NUD_VALID)
 585                         ret = RT6_NUD_SUCCEED;
 586 #ifdef CONFIG_IPV6_ROUTER_PREF
 587                 else if (!(neigh->nud_state & NUD_FAILED))
 588                         ret = RT6_NUD_SUCCEED;
 589                 else
 590                         ret = RT6_NUD_FAIL_PROBE;
 591 #endif
 592                 read_unlock(&neigh->lock);
 593         } else {
 594                 ret = IS_ENABLED(CONFIG_IPV6_ROUTER_PREF) ?
 595                       RT6_NUD_SUCCEED : RT6_NUD_FAIL_DO_RR;
 596         }
 597         rcu_read_unlock_bh();
 598
 599         return ret;
 600 }
 601
 602 static int rt6_score_route(struct rt6_info *rt, int oif,
 603                            int strict)
 604 {
 605         int m;
 606
 607         m = rt6_check_dev(rt, oif);
 608         if (!m && (strict & RT6_LOOKUP_F_IFACE))
 609                 return RT6_NUD_FAIL_HARD;
 610 #ifdef CONFIG_IPV6_ROUTER_PREF
 611         m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
 612 #endif
 613         if (strict & RT6_LOOKUP_F_REACHABLE) {
 614                 int n = rt6_check_neigh(rt);
 615                 if (n < 0)
 616                         return n;
 617         }
 618         return m;
 619 }
 620
 621 static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
 622                                    int *mpri, struct rt6_info *match,
 623                                    bool *do_rr)
 624 {
 625         int m;
 626         bool match_do_rr = false;
 627
 628         if (rt6_check_expired(rt))
 629                 goto out;
 630
 631         m = rt6_score_route(rt, oif, strict);
 632         if (m == RT6_NUD_FAIL_DO_RR) {
 633                 match_do_rr = true;
 634                 m = 0; /* lowest valid score */
 635         } else if (m == RT6_NUD_FAIL_HARD) {
 636                 goto out;
 637         }
 638
 639         if (strict & RT6_LOOKUP_F_REACHABLE)
 640                 rt6_probe(rt);
 641
 642         /* note that m can be RT6_NUD_FAIL_PROBE at this point */
 643         if (m > *mpri) {
 644                 *do_rr = match_do_rr;
 645                 *mpri = m;
 646                 match = rt;
 647         }
 648 out:
 649         return match;
 650 }
 651
 652 static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
 653                                      struct rt6_info *rr_head,
 654                                      u32 metric, int oif, int strict,
 655                                      bool *do_rr)
 656 {
 657         struct rt6_info *rt, *match;
 658         int mpri = -1;
 659
 660         match = NULL;
 661         for (rt = rr_head; rt && rt->rt6i_metric == metric;
 662              rt = rt->dst.rt6_next)
 663                 match = find_match(rt, oif, strict, &mpri, match, do_rr);
 664         for (rt = fn->leaf; rt && rt != rr_head && rt->rt6i_metric == metric;
 665              rt = rt->dst.rt6_next)
 666                 match = find_match(rt, oif, strict, &mpri, match, do_rr);
 667
 668         return match;
 669 }
 670
 671 static struct rt6_info *rt6_select(struct fib6_node *fn, int oif, int strict)
 672 {
 673         struct rt6_info *match, *rt0;
 674         struct net *net;
 675         bool do_rr = false;
 676
 677         rt0 = fn->rr_ptr;
 678         if (!rt0)
 679                 fn->rr_ptr = rt0 = fn->leaf;
 680
 681         match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict,
 682                              &do_rr);
 683
 684         if (do_rr) {
 685                 struct rt6_info *next = rt0->dst.rt6_next;
 686
 687                 /* no entries matched; do round-robin */
 688                 if (!next || next->rt6i_metric != rt0->rt6i_metric)
 689                         next = fn->leaf;
 690
 691                 if (next != rt0)
 692                         fn->rr_ptr = next;
 693         }
 694
 695         net = dev_net(rt0->dst.dev);
 696         return match ? match : net->ipv6.ip6_null_entry;
 697 }
 698
 699 #ifdef CONFIG_IPV6_ROUTE_INFO
 700 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
 701                   const struct in6_addr *gwaddr)
 702 {
 703         struct net *net = dev_net(dev);
 704         struct route_info *rinfo = (struct route_info *) opt;
 705         struct in6_addr prefix_buf, *prefix;
 706         unsigned int pref;
 707         unsigned long lifetime;
 708         struct rt6_info *rt;
 709
 710         if (len < sizeof(struct route_info)) {
 711                 return -EINVAL;
 712         }
 713
 714         /* Sanity check for prefix_len and length */
 715         if (rinfo->length > 3) {
 716                 return -EINVAL;
 717         } else if (rinfo->prefix_len > 128) {
 718                 return -EINVAL;
 719         } else if (rinfo->prefix_len > 64) {
 720                 if (rinfo->length < 2) {
 721                         return -EINVAL;
 722                 }
 723         } else if (rinfo->prefix_len > 0) {
 724                 if (rinfo->length < 1) {
 725                         return -EINVAL;
 726                 }
 727         }
 728
 729         pref = rinfo->route_pref;
 730         if (pref == ICMPV6_ROUTER_PREF_INVALID)
 731                 return -EINVAL;
 732
 733         lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
 734
 735         if (rinfo->length == 3)
 736                 prefix = (struct in6_addr *)rinfo->prefix;
 737         else {
 738                 /* this function is safe */
 739                 ipv6_addr_prefix(&prefix_buf,
 740                                  (struct in6_addr *)rinfo->prefix,
 741                                  rinfo->prefix_len);
 742                 prefix = &prefix_buf;
 743         }
 744
 745         if (rinfo->prefix_len == 0)
 746                 rt = rt6_get_dflt_router(gwaddr, dev);
 747         else
 748                 rt = rt6_get_route_info(net, prefix, rinfo->prefix_len,
 749                                         gwaddr, dev->ifindex);
 750
 751         if (rt && !lifetime) {
 752                 ip6_del_rt(rt);
 753                 rt = NULL;
 754         }
 755
 756         if (!rt && lifetime)
 757                 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr, dev->ifindex,
 758                                         pref);
 759         else if (rt)
 760                 rt->rt6i_flags = RTF_ROUTEINFO |
 761                                  (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
 762
 763         if (rt) {
 764                 if (!addrconf_finite_timeout(lifetime))
 765                         rt6_clean_expires(rt);
 766                 else
 767                         rt6_set_expires(rt, jiffies + HZ * lifetime);
 768
 769                 ip6_rt_put(rt);
 770         }
 771         return 0;
 772 }
 773 #endif
 774
 775 static struct fib6_node* fib6_backtrack(struct fib6_node *fn,
 776                                         struct in6_addr *saddr)
 777 {
 778         struct fib6_node *pn;
 779         while (1) {
 780                 if (fn->fn_flags & RTN_TL_ROOT)
 781                         return NULL;
 782                 pn = fn->parent;
 783                 if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn)
 784                         fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr);
 785                 else
 786                         fn = pn;
 787                 if (fn->fn_flags & RTN_RTINFO)
 788                         return fn;
 789         }
 790 }
 791
 792 static struct rt6_info *ip6_pol_route_lookup(struct net *net,
 793                                              struct fib6_table *table,
 794                                              struct flowi6 *fl6, int flags)
 795 {
 796         struct fib6_node *fn;
 797         struct rt6_info *rt;
 798
 799         read_lock_bh(&table->tb6_lock);
 800         fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
 801 restart:
 802         rt = fn->leaf;
 803         rt = rt6_device_match(net, rt, &fl6->saddr, fl6->flowi6_oif, flags);
 804         if (rt->rt6i_nsiblings && fl6->flowi6_oif == 0)
 805                 rt = rt6_multipath_select(rt, fl6, fl6->flowi6_oif, flags);
 806         if (rt == net->ipv6.ip6_null_entry) {
 807                 fn = fib6_backtrack(fn, &fl6->saddr);
 808                 if (fn)
 809                         goto restart;
 810         }
 811         dst_use(&rt->dst, jiffies);
 812         read_unlock_bh(&table->tb6_lock);
 813         return rt;
 814
 815 }
 816
 817 struct dst_entry *ip6_route_lookup(struct net *net, struct flowi6 *fl6,
 818                                     int flags)
 819 {
 820         return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_lookup);
 821 }
 822 EXPORT_SYMBOL_GPL(ip6_route_lookup);
 823
 824 struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
 825                             const struct in6_addr *saddr, int oif, int strict)
 826 {
 827         struct flowi6 fl6 = {
 828                 .flowi6_oif = oif,
 829                 .daddr = *daddr,
 830         };
 831         struct dst_entry *dst;
 832         int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
 833
 834         if (saddr) {
 835                 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
 836                 flags |= RT6_LOOKUP_F_HAS_SADDR;
 837         }
 838
 839         dst = fib6_rule_lookup(net, &fl6, flags, ip6_pol_route_lookup);
 840         if (dst->error == 0)
 841                 return (struct rt6_info *) dst;
 842
 843         dst_release(dst);
 844
 845         return NULL;
 846 }
 847 EXPORT_SYMBOL(rt6_lookup);
 848
 849 /* ip6_ins_rt is called with FREE table->tb6_lock.
 850    It takes new route entry, the addition fails by any reason the
 851    route is freed. In any case, if caller does not hold it, it may
 852    be destroyed.
 853  */
 854
 855 static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info,
 856                         struct nlattr *mx, int mx_len)
 857 {
 858         int err;
 859         struct fib6_table *table;
 860
 861         table = rt->rt6i_table;
 862         write_lock_bh(&table->tb6_lock);
 863         err = fib6_add(&table->tb6_root, rt, info, mx, mx_len);
 864         write_unlock_bh(&table->tb6_lock);
 865
 866         return err;
 867 }
 868
 869 int ip6_ins_rt(struct rt6_info *rt)
 870 {
 871         struct nl_info info = {
 872                 .nl_net = dev_net(rt->dst.dev),
 873         };
 874         return __ip6_ins_rt(rt, &info, NULL, 0);
 875 }
 876
 877 static struct rt6_info *rt6_alloc_cow(struct rt6_info *ort,
 878                                       const struct in6_addr *daddr,
 879                                       const struct in6_addr *saddr)
 880 {
 881         struct rt6_info *rt;
 882
 883         /*
 884          *      Clone the route.
 885          */
 886
 887         rt = ip6_rt_copy(ort, daddr);
 888
 889         if (rt) {
 890                 if (ort->rt6i_dst.plen != 128 &&
 891                     ipv6_addr_equal(&ort->rt6i_dst.addr, daddr))
 892                         rt->rt6i_flags |= RTF_ANYCAST;
 893
 894                 rt->rt6i_flags |= RTF_CACHE;
 895
 896 #ifdef CONFIG_IPV6_SUBTREES
 897                 if (rt->rt6i_src.plen && saddr) {
 898                         rt->rt6i_src.addr = *saddr;
 899                         rt->rt6i_src.plen = 128;
 900                 }
 901 #endif
 902         }
 903
 904         return rt;
 905 }
 906
 907 static struct rt6_info *rt6_alloc_clone(struct rt6_info *ort,
 908                                         const struct in6_addr *daddr)
 909 {
 910         struct rt6_info *rt = ip6_rt_copy(ort, daddr);
 911
 912         if (rt)
 913                 rt->rt6i_flags |= RTF_CACHE;
 914         return rt;
 915 }
 916
 917 static struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table, int oif,
 918                                       struct flowi6 *fl6, int flags)
 919 {
 920         struct fib6_node *fn, *saved_fn;
 921         struct rt6_info *rt, *nrt;
 922         int strict = 0;
 923         int attempts = 3;
 924         int err;
 925
 926         strict |= flags & RT6_LOOKUP_F_IFACE;
 927         if (net->ipv6.devconf_all->forwarding == 0)
 928                 strict |= RT6_LOOKUP_F_REACHABLE;
 929
 930 redo_fib6_lookup_lock:
 931         read_lock_bh(&table->tb6_lock);
 932
 933         fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
 934         saved_fn = fn;
 935
 936 redo_rt6_select:
 937         rt = rt6_select(fn, oif, strict);
 938         if (rt->rt6i_nsiblings)
 939                 rt = rt6_multipath_select(rt, fl6, oif, strict);
 940         if (rt == net->ipv6.ip6_null_entry) {
 941                 fn = fib6_backtrack(fn, &fl6->saddr);
 942                 if (fn)
 943                         goto redo_rt6_select;
 944                 else if (strict & RT6_LOOKUP_F_REACHABLE) {
 945                         /* also consider unreachable route */
 946                         strict &= ~RT6_LOOKUP_F_REACHABLE;
 947                         fn = saved_fn;
 948                         goto redo_rt6_select;
 949                 } else {
 950                         dst_hold(&rt->dst);
 951                         read_unlock_bh(&table->tb6_lock);
 952                         goto out2;
 953                 }
 954         }
 955
 956         dst_hold(&rt->dst);
 957         read_unlock_bh(&table->tb6_lock);
 958
 959         if (rt->rt6i_flags & RTF_CACHE)
 960                 goto out2;
 961
 962         if (!(rt->rt6i_flags & (RTF_NONEXTHOP | RTF_GATEWAY)))
 963                 nrt = rt6_alloc_cow(rt, &fl6->daddr, &fl6->saddr);
 964         else if (!(rt->dst.flags & DST_HOST))
 965                 nrt = rt6_alloc_clone(rt, &fl6->daddr);
 966         else
 967                 goto out2;
 968
 969         ip6_rt_put(rt);
 970         rt = nrt ? : net->ipv6.ip6_null_entry;
 971
 972         dst_hold(&rt->dst);
 973         if (nrt) {
 974                 err = ip6_ins_rt(nrt);
 975                 if (!err)
 976                         goto out2;
 977         }
 978
 979         if (--attempts <= 0)
 980                 goto out2;
 981
 982         /*
 983          * Race condition! In the gap, when table->tb6_lock was
 984          * released someone could insert this route.  Relookup.
 985          */
 986         ip6_rt_put(rt);
 987         goto redo_fib6_lookup_lock;
 988
 989 out2:
 990         rt->dst.lastuse = jiffies;
 991         rt->dst.__use++;
 992
 993         return rt;
 994 }
 995
 996 static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
 997                                             struct flowi6 *fl6, int flags)
 998 {
 999         return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, flags);
1000 }
1001
1002 static struct dst_entry *ip6_route_input_lookup(struct net *net,
1003                                                 struct net_device *dev,
1004                                                 struct flowi6 *fl6, int flags)
1005 {
1006         if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
1007                 flags |= RT6_LOOKUP_F_IFACE;
1008
1009         return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_input);
1010 }
1011
1012 void ip6_route_input(struct sk_buff *skb)
1013 {
1014         const struct ipv6hdr *iph = ipv6_hdr(skb);
1015         struct net *net = dev_net(skb->dev);
1016         int flags = RT6_LOOKUP_F_HAS_SADDR;
1017         struct flowi6 fl6 = {
1018                 .flowi6_iif = skb->dev->ifindex,
1019                 .daddr = iph->daddr,
1020                 .saddr = iph->saddr,
1021                 .flowlabel = ip6_flowinfo(iph),
1022                 .flowi6_mark = skb->mark,
1023                 .flowi6_proto = iph->nexthdr,
1024         };
1025
1026         skb_dst_set(skb, ip6_route_input_lookup(net, skb->dev, &fl6, flags));
1027 }
1028
1029 static struct rt6_info *ip6_pol_route_output(struct net *net, struct fib6_table *table,
1030                                              struct flowi6 *fl6, int flags)
1031 {
1032         return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, flags);
1033 }
1034
1035 struct dst_entry *ip6_route_output(struct net *net, const struct sock *sk,
1036                                     struct flowi6 *fl6)
1037 {
1038         int flags = 0;
1039
1040         fl6->flowi6_iif = LOOPBACK_IFINDEX;
1041
1042         if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr))
1043                 flags |= RT6_LOOKUP_F_IFACE;
1044
1045         if (!ipv6_addr_any(&fl6->saddr))
1046                 flags |= RT6_LOOKUP_F_HAS_SADDR;
1047         else if (sk)
1048                 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
1049
1050         return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_output);
1051 }
1052 EXPORT_SYMBOL(ip6_route_output);
1053
1054 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
1055 {
1056         struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
1057         struct dst_entry *new = NULL;
1058
1059         rt = dst_alloc(&ip6_dst_blackhole_ops, ort->dst.dev, 1, DST_OBSOLETE_NONE, 0);
1060         if (rt) {
1061                 new = &rt->dst;
1062
1063                 memset(new + 1, 0, sizeof(*rt) - sizeof(*new));
1064                 rt6_init_peer(rt, net->ipv6.peers);
1065
1066                 new->__use = 1;
1067                 new->input = dst_discard;
1068                 new->output = dst_discard_sk;
1069
1070                 if (dst_metrics_read_only(&ort->dst))
1071                         new->_metrics = ort->dst._metrics;
1072                 else
1073                         dst_copy_metrics(new, &ort->dst);
1074                 rt->rt6i_idev = ort->rt6i_idev;
1075                 if (rt->rt6i_idev)
1076                         in6_dev_hold(rt->rt6i_idev);
1077
1078                 rt->rt6i_gateway = ort->rt6i_gateway;
1079                 rt->rt6i_flags = ort->rt6i_flags;
1080                 rt->rt6i_metric = 0;
1081
1082                 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1083 #ifdef CONFIG_IPV6_SUBTREES
1084                 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1085 #endif
1086
1087                 dst_free(new);
1088         }
1089
1090         dst_release(dst_orig);
1091         return new ? new : ERR_PTR(-ENOMEM);
1092 }
1093
1094 /*
1095  *      Destination cache support functions
1096  */
1097
1098 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
1099 {
1100         struct rt6_info *rt;
1101
1102         rt = (struct rt6_info *) dst;
1103
1104         /* All IPV6 dsts are created with ->obsolete set to the value
1105          * DST_OBSOLETE_FORCE_CHK which forces validation calls down
1106          * into this function always.
1107          */
1108         if (!rt->rt6i_node || (rt->rt6i_node->fn_sernum != cookie))
1109                 return NULL;
1110
1111         if (rt6_check_expired(rt))
1112                 return NULL;
1113
1114         return dst;
1115 }
1116
1117 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
1118 {
1119         struct rt6_info *rt = (struct rt6_info *) dst;
1120
1121         if (rt) {
1122                 if (rt->rt6i_flags & RTF_CACHE) {
1123                         if (rt6_check_expired(rt)) {
1124                                 ip6_del_rt(rt);
1125                                 dst = NULL;
1126                         }
1127                 } else {
1128                         dst_release(dst);
1129                         dst = NULL;
1130                 }
1131         }
1132         return dst;
1133 }
1134
1135 static void ip6_link_failure(struct sk_buff *skb)
1136 {
1137         struct rt6_info *rt;
1138
1139         icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
1140
1141         rt = (struct rt6_info *) skb_dst(skb);
1142         if (rt) {
1143                 if (rt->rt6i_flags & RTF_CACHE) {
1144                         dst_hold(&rt->dst);
1145                         if (ip6_del_rt(rt))
1146                                 dst_free(&rt->dst);
1147                 } else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT)) {
1148                         rt->rt6i_node->fn_sernum = -1;
1149                 }
1150         }
1151 }
1152
1153 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
1154                                struct sk_buff *skb, u32 mtu)
1155 {
1156         struct rt6_info *rt6 = (struct rt6_info *)dst;
1157
1158         dst_confirm(dst);
1159         if (mtu < dst_mtu(dst) && rt6->rt6i_dst.plen == 128) {
1160                 struct net *net = dev_net(dst->dev);
1161
1162                 rt6->rt6i_flags |= RTF_MODIFIED;
1163                 if (mtu < IPV6_MIN_MTU) {
1164                         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 ? mark : IP6_REPLY_MARK(net, skb->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         } else if (rt == net->ipv6.ip6_null_entry) {
1249                 fn = fib6_backtrack(fn, &fl6->saddr);
1250                 if (fn)
1251                         goto restart;
1252         }
1253
1254         dst_hold(&rt->dst);
1255
1256         read_unlock_bh(&table->tb6_lock);
1257
1258         return rt;
1259 };
1260
1261 static struct dst_entry *ip6_route_redirect(struct net *net,
1262                                         const struct flowi6 *fl6,
1263                                         const struct in6_addr *gateway)
1264 {
1265         int flags = RT6_LOOKUP_F_HAS_SADDR;
1266         struct ip6rd_flowi rdfl;
1267
1268         rdfl.fl6 = *fl6;
1269         rdfl.gateway = *gateway;
1270
1271         return fib6_rule_lookup(net, &rdfl.fl6,
1272                                 flags, __ip6_route_redirect);
1273 }
1274
1275 void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark)
1276 {
1277         const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1278         struct dst_entry *dst;
1279         struct flowi6 fl6;
1280
1281         memset(&fl6, 0, sizeof(fl6));
1282         fl6.flowi6_iif = LOOPBACK_IFINDEX;
1283         fl6.flowi6_oif = oif;
1284         fl6.flowi6_mark = mark;
1285         fl6.daddr = iph->daddr;
1286         fl6.saddr = iph->saddr;
1287         fl6.flowlabel = ip6_flowinfo(iph);
1288
1289         dst = ip6_route_redirect(net, &fl6, &ipv6_hdr(skb)->saddr);
1290         rt6_do_redirect(dst, NULL, skb);
1291         dst_release(dst);
1292 }
1293 EXPORT_SYMBOL_GPL(ip6_redirect);
1294
1295 void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif,
1296                             u32 mark)
1297 {
1298         const struct ipv6hdr *iph = ipv6_hdr(skb);
1299         const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
1300         struct dst_entry *dst;
1301         struct flowi6 fl6;
1302
1303         memset(&fl6, 0, sizeof(fl6));
1304         fl6.flowi6_iif = LOOPBACK_IFINDEX;
1305         fl6.flowi6_oif = oif;
1306         fl6.flowi6_mark = mark;
1307         fl6.daddr = msg->dest;
1308         fl6.saddr = iph->daddr;
1309
1310         dst = ip6_route_redirect(net, &fl6, &iph->saddr);
1311         rt6_do_redirect(dst, NULL, skb);
1312         dst_release(dst);
1313 }
1314
1315 void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
1316 {
1317         ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark);
1318 }
1319 EXPORT_SYMBOL_GPL(ip6_sk_redirect);
1320
1321 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1322 {
1323         struct net_device *dev = dst->dev;
1324         unsigned int mtu = dst_mtu(dst);
1325         struct net *net = dev_net(dev);
1326
1327         mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
1328
1329         if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
1330                 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1331
1332         /*
1333          * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1334          * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1335          * IPV6_MAXPLEN is also valid and means: "any MSS,
1336          * rely only on pmtu discovery"
1337          */
1338         if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
1339                 mtu = IPV6_MAXPLEN;
1340         return mtu;
1341 }
1342
1343 static unsigned int ip6_mtu(const struct dst_entry *dst)
1344 {
1345         struct inet6_dev *idev;
1346         unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
1347
1348         if (mtu)
1349                 goto out;
1350
1351         mtu = IPV6_MIN_MTU;
1352
1353         rcu_read_lock();
1354         idev = __in6_dev_get(dst->dev);
1355         if (idev)
1356                 mtu = idev->cnf.mtu6;
1357         rcu_read_unlock();
1358
1359 out:
1360         return min_t(unsigned int, mtu, IP6_MAX_MTU);
1361 }
1362
1363 static struct dst_entry *icmp6_dst_gc_list;
1364 static DEFINE_SPINLOCK(icmp6_dst_lock);
1365
1366 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
1367                                   struct flowi6 *fl6)
1368 {
1369         struct dst_entry *dst;
1370         struct rt6_info *rt;
1371         struct inet6_dev *idev = in6_dev_get(dev);
1372         struct net *net = dev_net(dev);
1373
1374         if (unlikely(!idev))
1375                 return ERR_PTR(-ENODEV);
1376
1377         rt = ip6_dst_alloc(net, dev, 0, NULL);
1378         if (unlikely(!rt)) {
1379                 in6_dev_put(idev);
1380                 dst = ERR_PTR(-ENOMEM);
1381                 goto out;
1382         }
1383
1384         rt->dst.flags |= DST_HOST;
1385         rt->dst.output  = ip6_output;
1386         atomic_set(&rt->dst.__refcnt, 1);
1387         rt->rt6i_gateway  = fl6->daddr;
1388         rt->rt6i_dst.addr = fl6->daddr;
1389         rt->rt6i_dst.plen = 128;
1390         rt->rt6i_idev     = idev;
1391         dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
1392
1393         spin_lock_bh(&icmp6_dst_lock);
1394         rt->dst.next = icmp6_dst_gc_list;
1395         icmp6_dst_gc_list = &rt->dst;
1396         spin_unlock_bh(&icmp6_dst_lock);
1397
1398         fib6_force_start_gc(net);
1399
1400         dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
1401
1402 out:
1403         return dst;
1404 }
1405
1406 int icmp6_dst_gc(void)
1407 {
1408         struct dst_entry *dst, **pprev;
1409         int more = 0;
1410
1411         spin_lock_bh(&icmp6_dst_lock);
1412         pprev = &icmp6_dst_gc_list;
1413
1414         while ((dst = *pprev) != NULL) {
1415                 if (!atomic_read(&dst->__refcnt)) {
1416                         *pprev = dst->next;
1417                         dst_free(dst);
1418                 } else {
1419                         pprev = &dst->next;
1420                         ++more;
1421                 }
1422         }
1423
1424         spin_unlock_bh(&icmp6_dst_lock);
1425
1426         return more;
1427 }
1428
1429 static void icmp6_clean_all(int (*func)(struct rt6_info *rt, void *arg),
1430                             void *arg)
1431 {
1432         struct dst_entry *dst, **pprev;
1433
1434         spin_lock_bh(&icmp6_dst_lock);
1435         pprev = &icmp6_dst_gc_list;
1436         while ((dst = *pprev) != NULL) {
1437                 struct rt6_info *rt = (struct rt6_info *) dst;
1438                 if (func(rt, arg)) {
1439                         *pprev = dst->next;
1440                         dst_free(dst);
1441                 } else {
1442                         pprev = &dst->next;
1443                 }
1444         }
1445         spin_unlock_bh(&icmp6_dst_lock);
1446 }
1447
1448 static int ip6_dst_gc(struct dst_ops *ops)
1449 {
1450         struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
1451         int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1452         int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1453         int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1454         int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1455         unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1456         int entries;
1457
1458         entries = dst_entries_get_fast(ops);
1459         if (time_after(rt_last_gc + rt_min_interval, jiffies) &&
1460             entries <= rt_max_size)
1461                 goto out;
1462
1463         net->ipv6.ip6_rt_gc_expire++;
1464         fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net, true);
1465         entries = dst_entries_get_slow(ops);
1466         if (entries < ops->gc_thresh)
1467                 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1468 out:
1469         net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1470         return entries > rt_max_size;
1471 }
1472
1473 /*
1474  *
1475  */
1476
1477 int ip6_route_add(struct fib6_config *cfg)
1478 {
1479         int err;
1480         struct net *net = cfg->fc_nlinfo.nl_net;
1481         struct rt6_info *rt = NULL;
1482         struct net_device *dev = NULL;
1483         struct inet6_dev *idev = NULL;
1484         struct fib6_table *table;
1485         int addr_type;
1486
1487         if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1488                 return -EINVAL;
1489 #ifndef CONFIG_IPV6_SUBTREES
1490         if (cfg->fc_src_len)
1491                 return -EINVAL;
1492 #endif
1493         if (cfg->fc_ifindex) {
1494                 err = -ENODEV;
1495                 dev = dev_get_by_index(net, cfg->fc_ifindex);
1496                 if (!dev)
1497                         goto out;
1498                 idev = in6_dev_get(dev);
1499                 if (!idev)
1500                         goto out;
1501         }
1502
1503         if (cfg->fc_metric == 0)
1504                 cfg->fc_metric = IP6_RT_PRIO_USER;
1505
1506         err = -ENOBUFS;
1507         if (cfg->fc_nlinfo.nlh &&
1508             !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
1509                 table = fib6_get_table(net, cfg->fc_table);
1510                 if (!table) {
1511                         pr_warn("NLM_F_CREATE should be specified when creating new route\n");
1512                         table = fib6_new_table(net, cfg->fc_table);
1513                 }
1514         } else {
1515                 table = fib6_new_table(net, cfg->fc_table);
1516         }
1517
1518         if (!table)
1519                 goto out;
1520
1521         rt = ip6_dst_alloc(net, NULL, (cfg->fc_flags & RTF_ADDRCONF) ? 0 : DST_NOCOUNT, table);
1522
1523         if (!rt) {
1524                 err = -ENOMEM;
1525                 goto out;
1526         }
1527
1528         if (cfg->fc_flags & RTF_EXPIRES)
1529                 rt6_set_expires(rt, jiffies +
1530                                 clock_t_to_jiffies(cfg->fc_expires));
1531         else
1532                 rt6_clean_expires(rt);
1533
1534         if (cfg->fc_protocol == RTPROT_UNSPEC)
1535                 cfg->fc_protocol = RTPROT_BOOT;
1536         rt->rt6i_protocol = cfg->fc_protocol;
1537
1538         addr_type = ipv6_addr_type(&cfg->fc_dst);
1539
1540         if (addr_type & IPV6_ADDR_MULTICAST)
1541                 rt->dst.input = ip6_mc_input;
1542         else if (cfg->fc_flags & RTF_LOCAL)
1543                 rt->dst.input = ip6_input;
1544         else
1545                 rt->dst.input = ip6_forward;
1546
1547         rt->dst.output = ip6_output;
1548
1549         ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1550         rt->rt6i_dst.plen = cfg->fc_dst_len;
1551         if (rt->rt6i_dst.plen == 128) {
1552                 rt->dst.flags |= DST_HOST;
1553                 dst_metrics_set_force_overwrite(&rt->dst);
1554         }
1555
1556 #ifdef CONFIG_IPV6_SUBTREES
1557         ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1558         rt->rt6i_src.plen = cfg->fc_src_len;
1559 #endif
1560
1561         rt->rt6i_metric = cfg->fc_metric;
1562
1563         /* We cannot add true routes via loopback here,
1564            they would result in kernel looping; promote them to reject routes
1565          */
1566         if ((cfg->fc_flags & RTF_REJECT) ||
1567             (dev && (dev->flags & IFF_LOOPBACK) &&
1568              !(addr_type & IPV6_ADDR_LOOPBACK) &&
1569              !(cfg->fc_flags & RTF_LOCAL))) {
1570                 /* hold loopback dev/idev if we haven't done so. */
1571                 if (dev != net->loopback_dev) {
1572                         if (dev) {
1573                                 dev_put(dev);
1574                                 in6_dev_put(idev);
1575                         }
1576                         dev = net->loopback_dev;
1577                         dev_hold(dev);
1578                         idev = in6_dev_get(dev);
1579                         if (!idev) {
1580                                 err = -ENODEV;
1581                                 goto out;
1582                         }
1583                 }
1584                 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1585                 switch (cfg->fc_type) {
1586                 case RTN_BLACKHOLE:
1587                         rt->dst.error = -EINVAL;
1588                         rt->dst.output = dst_discard_sk;
1589                         rt->dst.input = dst_discard;
1590                         break;
1591                 case RTN_PROHIBIT:
1592                         rt->dst.error = -EACCES;
1593                         rt->dst.output = ip6_pkt_prohibit_out;
1594                         rt->dst.input = ip6_pkt_prohibit;
1595                         break;
1596                 case RTN_THROW:
1597                 default:
1598                         rt->dst.error = (cfg->fc_type == RTN_THROW) ? -EAGAIN
1599                                         : -ENETUNREACH;
1600                         rt->dst.output = ip6_pkt_discard_out;
1601                         rt->dst.input = ip6_pkt_discard;
1602                         break;
1603                 }
1604                 goto install_route;
1605         }
1606
1607         if (cfg->fc_flags & RTF_GATEWAY) {
1608                 const struct in6_addr *gw_addr;
1609                 int gwa_type;
1610
1611                 gw_addr = &cfg->fc_gateway;
1612                 rt->rt6i_gateway = *gw_addr;
1613                 gwa_type = ipv6_addr_type(gw_addr);
1614
1615                 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1616                         struct rt6_info *grt;
1617
1618                         /* IPv6 strictly inhibits using not link-local
1619                            addresses as nexthop address.
1620                            Otherwise, router will not able to send redirects.
1621                            It is very good, but in some (rare!) circumstances
1622                            (SIT, PtP, NBMA NOARP links) it is handy to allow
1623                            some exceptions. --ANK
1624                          */
1625                         err = -EINVAL;
1626                         if (!(gwa_type & IPV6_ADDR_UNICAST))
1627                                 goto out;
1628
1629                         grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, 1);
1630
1631                         err = -EHOSTUNREACH;
1632                         if (!grt)
1633                                 goto out;
1634                         if (dev) {
1635                                 if (dev != grt->dst.dev) {
1636                                         ip6_rt_put(grt);
1637                                         goto out;
1638                                 }
1639                         } else {
1640                                 dev = grt->dst.dev;
1641                                 idev = grt->rt6i_idev;
1642                                 dev_hold(dev);
1643                                 in6_dev_hold(grt->rt6i_idev);
1644                         }
1645                         if (!(grt->rt6i_flags & RTF_GATEWAY))
1646                                 err = 0;
1647                         ip6_rt_put(grt);
1648
1649                         if (err)
1650                                 goto out;
1651                 }
1652                 err = -EINVAL;
1653                 if (!dev || (dev->flags & IFF_LOOPBACK))
1654                         goto out;
1655         }
1656
1657         err = -ENODEV;
1658         if (!dev)
1659                 goto out;
1660
1661         if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
1662                 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
1663                         err = -EINVAL;
1664                         goto out;
1665                 }
1666                 rt->rt6i_prefsrc.addr = cfg->fc_prefsrc;
1667                 rt->rt6i_prefsrc.plen = 128;
1668         } else
1669                 rt->rt6i_prefsrc.plen = 0;
1670
1671         rt->rt6i_flags = cfg->fc_flags;
1672
1673 install_route:
1674         rt->dst.dev = dev;
1675         rt->rt6i_idev = idev;
1676         rt->rt6i_table = table;
1677
1678         cfg->fc_nlinfo.nl_net = dev_net(dev);
1679
1680         return __ip6_ins_rt(rt, &cfg->fc_nlinfo, cfg->fc_mx, cfg->fc_mx_len);
1681
1682 out:
1683         if (dev)
1684                 dev_put(dev);
1685         if (idev)
1686                 in6_dev_put(idev);
1687         if (rt)
1688                 dst_free(&rt->dst);
1689         return err;
1690 }
1691
1692 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
1693 {
1694         int err;
1695         struct fib6_table *table;
1696         struct net *net = dev_net(rt->dst.dev);
1697
1698         if (rt == net->ipv6.ip6_null_entry) {
1699                 err = -ENOENT;
1700                 goto out;
1701         }
1702
1703         table = rt->rt6i_table;
1704         write_lock_bh(&table->tb6_lock);
1705         err = fib6_del(rt, info);
1706         write_unlock_bh(&table->tb6_lock);
1707
1708 out:
1709         ip6_rt_put(rt);
1710         return err;
1711 }
1712
1713 int ip6_del_rt(struct rt6_info *rt)
1714 {
1715         struct nl_info info = {
1716                 .nl_net = dev_net(rt->dst.dev),
1717         };
1718         return __ip6_del_rt(rt, &info);
1719 }
1720
1721 static int ip6_route_del(struct fib6_config *cfg)
1722 {
1723         struct fib6_table *table;
1724         struct fib6_node *fn;
1725         struct rt6_info *rt;
1726         int err = -ESRCH;
1727
1728         table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
1729         if (!table)
1730                 return err;
1731
1732         read_lock_bh(&table->tb6_lock);
1733
1734         fn = fib6_locate(&table->tb6_root,
1735                          &cfg->fc_dst, cfg->fc_dst_len,
1736                          &cfg->fc_src, cfg->fc_src_len);
1737
1738         if (fn) {
1739                 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1740                         if (cfg->fc_ifindex &&
1741                             (!rt->dst.dev ||
1742                              rt->dst.dev->ifindex != cfg->fc_ifindex))
1743                                 continue;
1744                         if (cfg->fc_flags & RTF_GATEWAY &&
1745                             !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
1746                                 continue;
1747                         if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
1748                                 continue;
1749                         dst_hold(&rt->dst);
1750                         read_unlock_bh(&table->tb6_lock);
1751
1752                         return __ip6_del_rt(rt, &cfg->fc_nlinfo);
1753                 }
1754         }
1755         read_unlock_bh(&table->tb6_lock);
1756
1757         return err;
1758 }
1759
1760 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
1761 {
1762         struct net *net = dev_net(skb->dev);
1763         struct netevent_redirect netevent;
1764         struct rt6_info *rt, *nrt = NULL;
1765         struct ndisc_options ndopts;
1766         struct inet6_dev *in6_dev;
1767         struct neighbour *neigh;
1768         struct rd_msg *msg;
1769         int optlen, on_link;
1770         u8 *lladdr;
1771
1772         optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
1773         optlen -= sizeof(*msg);
1774
1775         if (optlen < 0) {
1776                 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
1777                 return;
1778         }
1779
1780         msg = (struct rd_msg *)icmp6_hdr(skb);
1781
1782         if (ipv6_addr_is_multicast(&msg->dest)) {
1783                 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
1784                 return;
1785         }
1786
1787         on_link = 0;
1788         if (ipv6_addr_equal(&msg->dest, &msg->target)) {
1789                 on_link = 1;
1790         } else if (ipv6_addr_type(&msg->target) !=
1791                    (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
1792                 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
1793                 return;
1794         }
1795
1796         in6_dev = __in6_dev_get(skb->dev);
1797         if (!in6_dev)
1798                 return;
1799         if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
1800                 return;
1801
1802         /* RFC2461 8.1:
1803          *      The IP source address of the Redirect MUST be the same as the current
1804          *      first-hop router for the specified ICMP Destination Address.
1805          */
1806
1807         if (!ndisc_parse_options(msg->opt, optlen, &ndopts)) {
1808                 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
1809                 return;
1810         }
1811
1812         lladdr = NULL;
1813         if (ndopts.nd_opts_tgt_lladdr) {
1814                 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
1815                                              skb->dev);
1816                 if (!lladdr) {
1817                         net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
1818                         return;
1819                 }
1820         }
1821
1822         rt = (struct rt6_info *) dst;
1823         if (rt == net->ipv6.ip6_null_entry) {
1824                 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
1825                 return;
1826         }
1827
1828         /* Redirect received -> path was valid.
1829          * Look, redirects are sent only in response to data packets,
1830          * so that this nexthop apparently is reachable. --ANK
1831          */
1832         dst_confirm(&rt->dst);
1833
1834         neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
1835         if (!neigh)
1836                 return;
1837
1838         /*
1839          *      We have finally decided to accept it.
1840          */
1841
1842         neigh_update(neigh, lladdr, NUD_STALE,
1843                      NEIGH_UPDATE_F_WEAK_OVERRIDE|
1844                      NEIGH_UPDATE_F_OVERRIDE|
1845                      (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1846                                      NEIGH_UPDATE_F_ISROUTER))
1847                      );
1848
1849         nrt = ip6_rt_copy(rt, &msg->dest);
1850         if (!nrt)
1851                 goto out;
1852
1853         nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
1854         if (on_link)
1855                 nrt->rt6i_flags &= ~RTF_GATEWAY;
1856
1857         nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
1858
1859         if (ip6_ins_rt(nrt))
1860                 goto out;
1861
1862         netevent.old = &rt->dst;
1863         netevent.new = &nrt->dst;
1864         netevent.daddr = &msg->dest;
1865         netevent.neigh = neigh;
1866         call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
1867
1868         if (rt->rt6i_flags & RTF_CACHE) {
1869                 rt = (struct rt6_info *) dst_clone(&rt->dst);
1870                 ip6_del_rt(rt);
1871         }
1872
1873 out:
1874         neigh_release(neigh);
1875 }
1876
1877 /*
1878  *      Misc support functions
1879  */
1880
1881 static struct rt6_info *ip6_rt_copy(struct rt6_info *ort,
1882                                     const struct in6_addr *dest)
1883 {
1884         struct net *net = dev_net(ort->dst.dev);
1885         struct rt6_info *rt = ip6_dst_alloc(net, ort->dst.dev, 0,
1886                                             ort->rt6i_table);
1887
1888         if (rt) {
1889                 rt->dst.input = ort->dst.input;
1890                 rt->dst.output = ort->dst.output;
1891                 rt->dst.flags |= DST_HOST;
1892
1893                 rt->rt6i_dst.addr = *dest;
1894                 rt->rt6i_dst.plen = 128;
1895                 dst_copy_metrics(&rt->dst, &ort->dst);
1896                 rt->dst.error = ort->dst.error;
1897                 rt->rt6i_idev = ort->rt6i_idev;
1898                 if (rt->rt6i_idev)
1899                         in6_dev_hold(rt->rt6i_idev);
1900                 rt->dst.lastuse = jiffies;
1901
1902                 if (ort->rt6i_flags & RTF_GATEWAY)
1903                         rt->rt6i_gateway = ort->rt6i_gateway;
1904                 else
1905                         rt->rt6i_gateway = *dest;
1906                 rt->rt6i_flags = ort->rt6i_flags;
1907                 rt6_set_from(rt, ort);
1908                 rt->rt6i_metric = 0;
1909
1910 #ifdef CONFIG_IPV6_SUBTREES
1911                 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1912 #endif
1913                 memcpy(&rt->rt6i_prefsrc, &ort->rt6i_prefsrc, sizeof(struct rt6key));
1914                 rt->rt6i_table = ort->rt6i_table;
1915         }
1916         return rt;
1917 }
1918
1919 #ifdef CONFIG_IPV6_ROUTE_INFO
1920 static struct rt6_info *rt6_get_route_info(struct net *net,
1921                                            const struct in6_addr *prefix, int prefixlen,
1922                                            const struct in6_addr *gwaddr, int ifindex)
1923 {
1924         struct fib6_node *fn;
1925         struct rt6_info *rt = NULL;
1926         struct fib6_table *table;
1927
1928         table = fib6_get_table(net, RT6_TABLE_INFO);
1929         if (!table)
1930                 return NULL;
1931
1932         read_lock_bh(&table->tb6_lock);
1933         fn = fib6_locate(&table->tb6_root, prefix, prefixlen, NULL, 0);
1934         if (!fn)
1935                 goto out;
1936
1937         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1938                 if (rt->dst.dev->ifindex != ifindex)
1939                         continue;
1940                 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
1941                         continue;
1942                 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
1943                         continue;
1944                 dst_hold(&rt->dst);
1945                 break;
1946         }
1947 out:
1948         read_unlock_bh(&table->tb6_lock);
1949         return rt;
1950 }
1951
1952 static struct rt6_info *rt6_add_route_info(struct net *net,
1953                                            const struct in6_addr *prefix, int prefixlen,
1954                                            const struct in6_addr *gwaddr, int ifindex,
1955                                            unsigned int pref)
1956 {
1957         struct fib6_config cfg = {
1958                 .fc_table       = RT6_TABLE_INFO,
1959                 .fc_metric      = IP6_RT_PRIO_USER,
1960                 .fc_ifindex     = ifindex,
1961                 .fc_dst_len     = prefixlen,
1962                 .fc_flags       = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
1963                                   RTF_UP | RTF_PREF(pref),
1964                 .fc_nlinfo.portid = 0,
1965                 .fc_nlinfo.nlh = NULL,
1966                 .fc_nlinfo.nl_net = net,
1967         };
1968
1969         cfg.fc_dst = *prefix;
1970         cfg.fc_gateway = *gwaddr;
1971
1972         /* We should treat it as a default route if prefix length is 0. */
1973         if (!prefixlen)
1974                 cfg.fc_flags |= RTF_DEFAULT;
1975
1976         ip6_route_add(&cfg);
1977
1978         return rt6_get_route_info(net, prefix, prefixlen, gwaddr, ifindex);
1979 }
1980 #endif
1981
1982 struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
1983 {
1984         struct rt6_info *rt;
1985         struct fib6_table *table;
1986
1987         table = fib6_get_table(dev_net(dev), RT6_TABLE_DFLT);
1988         if (!table)
1989                 return NULL;
1990
1991         read_lock_bh(&table->tb6_lock);
1992         for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
1993                 if (dev == rt->dst.dev &&
1994                     ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
1995                     ipv6_addr_equal(&rt->rt6i_gateway, addr))
1996                         break;
1997         }
1998         if (rt)
1999                 dst_hold(&rt->dst);
2000         read_unlock_bh(&table->tb6_lock);
2001         return rt;
2002 }
2003
2004 struct rt6_info *rt6_add_dflt_router(const struct in6_addr *gwaddr,
2005                                      struct net_device *dev,
2006                                      unsigned int pref)
2007 {
2008         struct fib6_config cfg = {
2009                 .fc_table       = RT6_TABLE_DFLT,
2010                 .fc_metric      = IP6_RT_PRIO_USER,
2011                 .fc_ifindex     = dev->ifindex,
2012                 .fc_flags       = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
2013                                   RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
2014                 .fc_nlinfo.portid = 0,
2015                 .fc_nlinfo.nlh = NULL,
2016                 .fc_nlinfo.nl_net = dev_net(dev),
2017         };
2018
2019         cfg.fc_gateway = *gwaddr;
2020
2021         ip6_route_add(&cfg);
2022
2023         return rt6_get_dflt_router(gwaddr, dev);
2024 }
2025
2026 void rt6_purge_dflt_routers(struct net *net)
2027 {
2028         struct rt6_info *rt;
2029         struct fib6_table *table;
2030
2031         /* NOTE: Keep consistent with rt6_get_dflt_router */
2032         table = fib6_get_table(net, RT6_TABLE_DFLT);
2033         if (!table)
2034                 return;
2035
2036 restart:
2037         read_lock_bh(&table->tb6_lock);
2038         for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2039                 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
2040                     (!rt->rt6i_idev || rt->rt6i_idev->cnf.accept_ra != 2)) {
2041                         dst_hold(&rt->dst);
2042                         read_unlock_bh(&table->tb6_lock);
2043                         ip6_del_rt(rt);
2044                         goto restart;
2045                 }
2046         }
2047         read_unlock_bh(&table->tb6_lock);
2048 }
2049
2050 static void rtmsg_to_fib6_config(struct net *net,
2051                                  struct in6_rtmsg *rtmsg,
2052                                  struct fib6_config *cfg)
2053 {
2054         memset(cfg, 0, sizeof(*cfg));
2055
2056         cfg->fc_table = RT6_TABLE_MAIN;
2057         cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
2058         cfg->fc_metric = rtmsg->rtmsg_metric;
2059         cfg->fc_expires = rtmsg->rtmsg_info;
2060         cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
2061         cfg->fc_src_len = rtmsg->rtmsg_src_len;
2062         cfg->fc_flags = rtmsg->rtmsg_flags;
2063
2064         cfg->fc_nlinfo.nl_net = net;
2065
2066         cfg->fc_dst = rtmsg->rtmsg_dst;
2067         cfg->fc_src = rtmsg->rtmsg_src;
2068         cfg->fc_gateway = rtmsg->rtmsg_gateway;
2069 }
2070
2071 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
2072 {
2073         struct fib6_config cfg;
2074         struct in6_rtmsg rtmsg;
2075         int err;
2076
2077         switch (cmd) {
2078         case SIOCADDRT:         /* Add a route */
2079         case SIOCDELRT:         /* Delete a route */
2080                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2081                         return -EPERM;
2082                 err = copy_from_user(&rtmsg, arg,
2083                                      sizeof(struct in6_rtmsg));
2084                 if (err)
2085                         return -EFAULT;
2086
2087                 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
2088
2089                 rtnl_lock();
2090                 switch (cmd) {
2091                 case SIOCADDRT:
2092                         err = ip6_route_add(&cfg);
2093                         break;
2094                 case SIOCDELRT:
2095                         err = ip6_route_del(&cfg);
2096                         break;
2097                 default:
2098                         err = -EINVAL;
2099                 }
2100                 rtnl_unlock();
2101
2102                 return err;
2103         }
2104
2105         return -EINVAL;
2106 }
2107
2108 /*
2109  *      Drop the packet on the floor
2110  */
2111
2112 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
2113 {
2114         int type;
2115         struct dst_entry *dst = skb_dst(skb);
2116         switch (ipstats_mib_noroutes) {
2117         case IPSTATS_MIB_INNOROUTES:
2118                 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
2119                 if (type == IPV6_ADDR_ANY) {
2120                         IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2121                                       IPSTATS_MIB_INADDRERRORS);
2122                         break;
2123                 }
2124                 /* FALLTHROUGH */
2125         case IPSTATS_MIB_OUTNOROUTES:
2126                 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2127                               ipstats_mib_noroutes);
2128                 break;
2129         }
2130         icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
2131         kfree_skb(skb);
2132         return 0;
2133 }
2134
2135 static int ip6_pkt_discard(struct sk_buff *skb)
2136 {
2137         return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
2138 }
2139
2140 static int ip6_pkt_discard_out(struct sock *sk, struct sk_buff *skb)
2141 {
2142         skb->dev = skb_dst(skb)->dev;
2143         return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
2144 }
2145
2146 static int ip6_pkt_prohibit(struct sk_buff *skb)
2147 {
2148         return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
2149 }
2150
2151 static int ip6_pkt_prohibit_out(struct sock *sk, struct sk_buff *skb)
2152 {
2153         skb->dev = skb_dst(skb)->dev;
2154         return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
2155 }
2156
2157 /*
2158  *      Allocate a dst for local (unicast / anycast) address.
2159  */
2160
2161 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
2162                                     const struct in6_addr *addr,
2163                                     bool anycast)
2164 {
2165         struct net *net = dev_net(idev->dev);
2166         struct rt6_info *rt = ip6_dst_alloc(net, net->loopback_dev,
2167                                             DST_NOCOUNT, NULL);
2168         if (!rt)
2169                 return ERR_PTR(-ENOMEM);
2170
2171         in6_dev_hold(idev);
2172
2173         rt->dst.flags |= DST_HOST;
2174         rt->dst.input = ip6_input;
2175         rt->dst.output = ip6_output;
2176         rt->rt6i_idev = idev;
2177
2178         rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
2179         if (anycast)
2180                 rt->rt6i_flags |= RTF_ANYCAST;
2181         else
2182                 rt->rt6i_flags |= RTF_LOCAL;
2183
2184         rt->rt6i_gateway  = *addr;
2185         rt->rt6i_dst.addr = *addr;
2186         rt->rt6i_dst.plen = 128;
2187         rt->rt6i_table = fib6_get_table(net, RT6_TABLE_LOCAL);
2188
2189         atomic_set(&rt->dst.__refcnt, 1);
2190
2191         return rt;
2192 }
2193
2194 int ip6_route_get_saddr(struct net *net,
2195                         struct rt6_info *rt,
2196                         const struct in6_addr *daddr,
2197                         unsigned int prefs,
2198                         struct in6_addr *saddr)
2199 {
2200         struct inet6_dev *idev = ip6_dst_idev((struct dst_entry *)rt);
2201         int err = 0;
2202         if (rt->rt6i_prefsrc.plen)
2203                 *saddr = rt->rt6i_prefsrc.addr;
2204         else
2205                 err = ipv6_dev_get_saddr(net, idev ? idev->dev : NULL,
2206                                          daddr, prefs, saddr);
2207         return err;
2208 }
2209
2210 /* remove deleted ip from prefsrc entries */
2211 struct arg_dev_net_ip {
2212         struct net_device *dev;
2213         struct net *net;
2214         struct in6_addr *addr;
2215 };
2216
2217 static int fib6_remove_prefsrc(struct rt6_info *rt, void *arg)
2218 {
2219         struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
2220         struct net *net = ((struct arg_dev_net_ip *)arg)->net;
2221         struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
2222
2223         if (((void *)rt->dst.dev == dev || !dev) &&
2224             rt != net->ipv6.ip6_null_entry &&
2225             ipv6_addr_equal(addr, &rt->rt6i_prefsrc.addr)) {
2226                 /* remove prefsrc entry */
2227                 rt->rt6i_prefsrc.plen = 0;
2228         }
2229         return 0;
2230 }
2231
2232 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
2233 {
2234         struct net *net = dev_net(ifp->idev->dev);
2235         struct arg_dev_net_ip adni = {
2236                 .dev = ifp->idev->dev,
2237                 .net = net,
2238                 .addr = &ifp->addr,
2239         };
2240         fib6_clean_all(net, fib6_remove_prefsrc, &adni);
2241 }
2242
2243 #define RTF_RA_ROUTER           (RTF_ADDRCONF | RTF_DEFAULT | RTF_GATEWAY)
2244 #define RTF_CACHE_GATEWAY       (RTF_GATEWAY | RTF_CACHE)
2245
2246 /* Remove routers and update dst entries when gateway turn into host. */
2247 static int fib6_clean_tohost(struct rt6_info *rt, void *arg)
2248 {
2249         struct in6_addr *gateway = (struct in6_addr *)arg;
2250
2251         if ((((rt->rt6i_flags & RTF_RA_ROUTER) == RTF_RA_ROUTER) ||
2252              ((rt->rt6i_flags & RTF_CACHE_GATEWAY) == RTF_CACHE_GATEWAY)) &&
2253              ipv6_addr_equal(gateway, &rt->rt6i_gateway)) {
2254                 return -1;
2255         }
2256         return 0;
2257 }
2258
2259 void rt6_clean_tohost(struct net *net, struct in6_addr *gateway)
2260 {
2261         fib6_clean_all(net, fib6_clean_tohost, gateway);
2262 }
2263
2264 struct arg_dev_net {
2265         struct net_device *dev;
2266         struct net *net;
2267 };
2268
2269 static int fib6_ifdown(struct rt6_info *rt, void *arg)
2270 {
2271         const struct arg_dev_net *adn = arg;
2272         const struct net_device *dev = adn->dev;
2273
2274         if ((rt->dst.dev == dev || !dev) &&
2275             rt != adn->net->ipv6.ip6_null_entry)
2276                 return -1;
2277
2278         return 0;
2279 }
2280
2281 void rt6_ifdown(struct net *net, struct net_device *dev)
2282 {
2283         struct arg_dev_net adn = {
2284                 .dev = dev,
2285                 .net = net,
2286         };
2287
2288         fib6_clean_all(net, fib6_ifdown, &adn);
2289         icmp6_clean_all(fib6_ifdown, &adn);
2290 }
2291
2292 struct rt6_mtu_change_arg {
2293         struct net_device *dev;
2294         unsigned int mtu;
2295 };
2296
2297 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2298 {
2299         struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2300         struct inet6_dev *idev;
2301
2302         /* In IPv6 pmtu discovery is not optional,
2303            so that RTAX_MTU lock cannot disable it.
2304            We still use this lock to block changes
2305            caused by addrconf/ndisc.
2306         */
2307
2308         idev = __in6_dev_get(arg->dev);
2309         if (!idev)
2310                 return 0;
2311
2312         /* For administrative MTU increase, there is no way to discover
2313            IPv6 PMTU increase, so PMTU increase should be updated here.
2314            Since RFC 1981 doesn't include administrative MTU increase
2315            update PMTU increase is a MUST. (i.e. jumbo frame)
2316          */
2317         /*
2318            If new MTU is less than route PMTU, this new MTU will be the
2319            lowest MTU in the path, update the route PMTU to reflect PMTU
2320            decreases; if new MTU is greater than route PMTU, and the
2321            old MTU is the lowest MTU in the path, update the route PMTU
2322            to reflect the increase. In this case if the other nodes' MTU
2323            also have the lowest MTU, TOO BIG MESSAGE will be lead to
2324            PMTU discouvery.
2325          */
2326         if (rt->dst.dev == arg->dev &&
2327             !dst_metric_locked(&rt->dst, RTAX_MTU) &&
2328             (dst_mtu(&rt->dst) >= arg->mtu ||
2329              (dst_mtu(&rt->dst) < arg->mtu &&
2330               dst_mtu(&rt->dst) == idev->cnf.mtu6))) {
2331                 dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
2332         }
2333         return 0;
2334 }
2335
2336 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
2337 {
2338         struct rt6_mtu_change_arg arg = {
2339                 .dev = dev,
2340                 .mtu = mtu,
2341         };
2342
2343         fib6_clean_all(dev_net(dev), rt6_mtu_change_route, &arg);
2344 }
2345
2346 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2347         [RTA_GATEWAY]           = { .len = sizeof(struct in6_addr) },
2348         [RTA_OIF]               = { .type = NLA_U32 },
2349         [RTA_IIF]               = { .type = NLA_U32 },
2350         [RTA_PRIORITY]          = { .type = NLA_U32 },
2351         [RTA_METRICS]           = { .type = NLA_NESTED },
2352         [RTA_MULTIPATH]         = { .len = sizeof(struct rtnexthop) },
2353 };
2354
2355 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2356                               struct fib6_config *cfg)
2357 {
2358         struct rtmsg *rtm;
2359         struct nlattr *tb[RTA_MAX+1];
2360         int err;
2361
2362         err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2363         if (err < 0)
2364                 goto errout;
2365
2366         err = -EINVAL;
2367         rtm = nlmsg_data(nlh);
2368         memset(cfg, 0, sizeof(*cfg));
2369
2370         cfg->fc_table = rtm->rtm_table;
2371         cfg->fc_dst_len = rtm->rtm_dst_len;
2372         cfg->fc_src_len = rtm->rtm_src_len;
2373         cfg->fc_flags = RTF_UP;
2374         cfg->fc_protocol = rtm->rtm_protocol;
2375         cfg->fc_type = rtm->rtm_type;
2376
2377         if (rtm->rtm_type == RTN_UNREACHABLE ||
2378             rtm->rtm_type == RTN_BLACKHOLE ||
2379             rtm->rtm_type == RTN_PROHIBIT ||
2380             rtm->rtm_type == RTN_THROW)
2381                 cfg->fc_flags |= RTF_REJECT;
2382
2383         if (rtm->rtm_type == RTN_LOCAL)
2384                 cfg->fc_flags |= RTF_LOCAL;
2385
2386         cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid;
2387         cfg->fc_nlinfo.nlh = nlh;
2388         cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2389
2390         if (tb[RTA_GATEWAY]) {
2391                 nla_memcpy(&cfg->fc_gateway, tb[RTA_GATEWAY], 16);
2392                 cfg->fc_flags |= RTF_GATEWAY;
2393         }
2394
2395         if (tb[RTA_DST]) {
2396                 int plen = (rtm->rtm_dst_len + 7) >> 3;
2397
2398                 if (nla_len(tb[RTA_DST]) < plen)
2399                         goto errout;
2400
2401                 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2402         }
2403
2404         if (tb[RTA_SRC]) {
2405                 int plen = (rtm->rtm_src_len + 7) >> 3;
2406
2407                 if (nla_len(tb[RTA_SRC]) < plen)
2408                         goto errout;
2409
2410                 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2411         }
2412
2413         if (tb[RTA_PREFSRC])
2414                 nla_memcpy(&cfg->fc_prefsrc, tb[RTA_PREFSRC], 16);
2415
2416         if (tb[RTA_OIF])
2417                 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2418
2419         if (tb[RTA_PRIORITY])
2420                 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2421
2422         if (tb[RTA_METRICS]) {
2423                 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2424                 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2425         }
2426
2427         if (tb[RTA_TABLE])
2428                 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2429
2430         if (tb[RTA_MULTIPATH]) {
2431                 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
2432                 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
2433         }
2434
2435         err = 0;
2436 errout:
2437         return err;
2438 }
2439
2440 static int ip6_route_multipath(struct fib6_config *cfg, int add)
2441 {
2442         struct fib6_config r_cfg;
2443         struct rtnexthop *rtnh;
2444         int remaining;
2445         int attrlen;
2446         int err = 0, last_err = 0;
2447
2448 beginning:
2449         rtnh = (struct rtnexthop *)cfg->fc_mp;
2450         remaining = cfg->fc_mp_len;
2451
2452         /* Parse a Multipath Entry */
2453         while (rtnh_ok(rtnh, remaining)) {
2454                 memcpy(&r_cfg, cfg, sizeof(*cfg));
2455                 if (rtnh->rtnh_ifindex)
2456                         r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
2457
2458                 attrlen = rtnh_attrlen(rtnh);
2459                 if (attrlen > 0) {
2460                         struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
2461
2462                         nla = nla_find(attrs, attrlen, RTA_GATEWAY);
2463                         if (nla) {
2464                                 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
2465                                 r_cfg.fc_flags |= RTF_GATEWAY;
2466                         }
2467                 }
2468                 err = add ? ip6_route_add(&r_cfg) : ip6_route_del(&r_cfg);
2469                 if (err) {
2470                         last_err = err;
2471                         /* If we are trying to remove a route, do not stop the
2472                          * loop when ip6_route_del() fails (because next hop is
2473                          * already gone), we should try to remove all next hops.
2474                          */
2475                         if (add) {
2476                                 /* If add fails, we should try to delete all
2477                                  * next hops that have been already added.
2478                                  */
2479                                 add = 0;
2480                                 goto beginning;
2481                         }
2482                 }
2483                 /* Because each route is added like a single route we remove
2484                  * this flag after the first nexthop (if there is a collision,
2485                  * we have already fail to add the first nexthop:
2486                  * fib6_add_rt2node() has reject it).
2487                  */
2488                 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~NLM_F_EXCL;
2489                 rtnh = rtnh_next(rtnh, &remaining);
2490         }
2491
2492         return last_err;
2493 }
2494
2495 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh)
2496 {
2497         struct fib6_config cfg;
2498         int err;
2499
2500         err = rtm_to_fib6_config(skb, nlh, &cfg);
2501         if (err < 0)
2502                 return err;
2503
2504         if (cfg.fc_mp)
2505                 return ip6_route_multipath(&cfg, 0);
2506         else
2507                 return ip6_route_del(&cfg);
2508 }
2509
2510 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh)
2511 {
2512         struct fib6_config cfg;
2513         int err;
2514
2515         err = rtm_to_fib6_config(skb, nlh, &cfg);
2516         if (err < 0)
2517                 return err;
2518
2519         if (cfg.fc_mp)
2520                 return ip6_route_multipath(&cfg, 1);
2521         else
2522                 return ip6_route_add(&cfg);
2523 }
2524
2525 static inline size_t rt6_nlmsg_size(void)
2526 {
2527         return NLMSG_ALIGN(sizeof(struct rtmsg))
2528                + nla_total_size(16) /* RTA_SRC */
2529                + nla_total_size(16) /* RTA_DST */
2530                + nla_total_size(16) /* RTA_GATEWAY */
2531                + nla_total_size(16) /* RTA_PREFSRC */
2532                + nla_total_size(4) /* RTA_TABLE */
2533                + nla_total_size(4) /* RTA_IIF */
2534                + nla_total_size(4) /* RTA_OIF */
2535                + nla_total_size(4) /* RTA_PRIORITY */
2536                + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
2537                + nla_total_size(sizeof(struct rta_cacheinfo));
2538 }
2539
2540 static int rt6_fill_node(struct net *net,
2541                          struct sk_buff *skb, struct rt6_info *rt,
2542                          struct in6_addr *dst, struct in6_addr *src,
2543                          int iif, int type, u32 portid, u32 seq,
2544                          int prefix, int nowait, unsigned int flags)
2545 {
2546         struct rtmsg *rtm;
2547         struct nlmsghdr *nlh;
2548         long expires;
2549         u32 table;
2550
2551         if (prefix) {   /* user wants prefix routes only */
2552                 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
2553                         /* success since this is not a prefix route */
2554                         return 1;
2555                 }
2556         }
2557
2558         nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
2559         if (!nlh)
2560                 return -EMSGSIZE;
2561
2562         rtm = nlmsg_data(nlh);
2563         rtm->rtm_family = AF_INET6;
2564         rtm->rtm_dst_len = rt->rt6i_dst.plen;
2565         rtm->rtm_src_len = rt->rt6i_src.plen;
2566         rtm->rtm_tos = 0;
2567         if (rt->rt6i_table)
2568                 table = rt->rt6i_table->tb6_id;
2569         else
2570                 table = RT6_TABLE_UNSPEC;
2571         rtm->rtm_table = table;
2572         if (nla_put_u32(skb, RTA_TABLE, table))
2573                 goto nla_put_failure;
2574         if (rt->rt6i_flags & RTF_REJECT) {
2575                 switch (rt->dst.error) {
2576                 case -EINVAL:
2577                         rtm->rtm_type = RTN_BLACKHOLE;
2578                         break;
2579                 case -EACCES:
2580                         rtm->rtm_type = RTN_PROHIBIT;
2581                         break;
2582                 case -EAGAIN:
2583                         rtm->rtm_type = RTN_THROW;
2584                         break;
2585                 default:
2586                         rtm->rtm_type = RTN_UNREACHABLE;
2587                         break;
2588                 }
2589         }
2590         else if (rt->rt6i_flags & RTF_LOCAL)
2591                 rtm->rtm_type = RTN_LOCAL;
2592         else if (rt->dst.dev && (rt->dst.dev->flags & IFF_LOOPBACK))
2593                 rtm->rtm_type = RTN_LOCAL;
2594         else
2595                 rtm->rtm_type = RTN_UNICAST;
2596         rtm->rtm_flags = 0;
2597         rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2598         rtm->rtm_protocol = rt->rt6i_protocol;
2599         if (rt->rt6i_flags & RTF_DYNAMIC)
2600                 rtm->rtm_protocol = RTPROT_REDIRECT;
2601         else if (rt->rt6i_flags & RTF_ADDRCONF) {
2602                 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ROUTEINFO))
2603                         rtm->rtm_protocol = RTPROT_RA;
2604                 else
2605                         rtm->rtm_protocol = RTPROT_KERNEL;
2606         }
2607
2608         if (rt->rt6i_flags & RTF_CACHE)
2609                 rtm->rtm_flags |= RTM_F_CLONED;
2610
2611         if (dst) {
2612                 if (nla_put(skb, RTA_DST, 16, dst))
2613                         goto nla_put_failure;
2614                 rtm->rtm_dst_len = 128;
2615         } else if (rtm->rtm_dst_len)
2616                 if (nla_put(skb, RTA_DST, 16, &rt->rt6i_dst.addr))
2617                         goto nla_put_failure;
2618 #ifdef CONFIG_IPV6_SUBTREES
2619         if (src) {
2620                 if (nla_put(skb, RTA_SRC, 16, src))
2621                         goto nla_put_failure;
2622                 rtm->rtm_src_len = 128;
2623         } else if (rtm->rtm_src_len &&
2624                    nla_put(skb, RTA_SRC, 16, &rt->rt6i_src.addr))
2625                 goto nla_put_failure;
2626 #endif
2627         if (iif) {
2628 #ifdef CONFIG_IPV6_MROUTE
2629                 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
2630                         int err = ip6mr_get_route(net, skb, rtm, nowait);
2631                         if (err <= 0) {
2632                                 if (!nowait) {
2633                                         if (err == 0)
2634                                                 return 0;
2635                                         goto nla_put_failure;
2636                                 } else {
2637                                         if (err == -EMSGSIZE)
2638                                                 goto nla_put_failure;
2639                                 }
2640                         }
2641                 } else
2642 #endif
2643                         if (nla_put_u32(skb, RTA_IIF, iif))
2644                                 goto nla_put_failure;
2645         } else if (dst) {
2646                 struct in6_addr saddr_buf;
2647                 if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0 &&
2648                     nla_put(skb, RTA_PREFSRC, 16, &saddr_buf))
2649                         goto nla_put_failure;
2650         }
2651
2652         if (rt->rt6i_prefsrc.plen) {
2653                 struct in6_addr saddr_buf;
2654                 saddr_buf = rt->rt6i_prefsrc.addr;
2655                 if (nla_put(skb, RTA_PREFSRC, 16, &saddr_buf))
2656                         goto nla_put_failure;
2657         }
2658
2659         if (rtnetlink_put_metrics(skb, dst_metrics_ptr(&rt->dst)) < 0)
2660                 goto nla_put_failure;
2661
2662         if (rt->rt6i_flags & RTF_GATEWAY) {
2663                 if (nla_put(skb, RTA_GATEWAY, 16, &rt->rt6i_gateway) < 0)
2664                         goto nla_put_failure;
2665         }
2666
2667         if (rt->dst.dev &&
2668             nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex))
2669                 goto nla_put_failure;
2670         if (nla_put_u32(skb, RTA_PRIORITY, rt->rt6i_metric))
2671                 goto nla_put_failure;
2672
2673         expires = (rt->rt6i_flags & RTF_EXPIRES) ? rt->dst.expires - jiffies : 0;
2674
2675         if (rtnl_put_cacheinfo(skb, &rt->dst, 0, expires, rt->dst.error) < 0)
2676                 goto nla_put_failure;
2677
2678         return nlmsg_end(skb, nlh);
2679
2680 nla_put_failure:
2681         nlmsg_cancel(skb, nlh);
2682         return -EMSGSIZE;
2683 }
2684
2685 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
2686 {
2687         struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
2688         int prefix;
2689
2690         if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
2691                 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
2692                 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
2693         } else
2694                 prefix = 0;
2695
2696         return rt6_fill_node(arg->net,
2697                      arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
2698                      NETLINK_CB(arg->cb->skb).portid, arg->cb->nlh->nlmsg_seq,
2699                      prefix, 0, NLM_F_MULTI);
2700 }
2701
2702 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh)
2703 {
2704         struct net *net = sock_net(in_skb->sk);
2705         struct nlattr *tb[RTA_MAX+1];
2706         struct rt6_info *rt;
2707         struct sk_buff *skb;
2708         struct rtmsg *rtm;
2709         struct flowi6 fl6;
2710         int err, iif = 0, oif = 0;
2711
2712         err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2713         if (err < 0)
2714                 goto errout;
2715
2716         err = -EINVAL;
2717         memset(&fl6, 0, sizeof(fl6));
2718
2719         if (tb[RTA_SRC]) {
2720                 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
2721                         goto errout;
2722
2723                 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
2724         }
2725
2726         if (tb[RTA_DST]) {
2727                 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
2728                         goto errout;
2729
2730                 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
2731         }
2732
2733         if (tb[RTA_IIF])
2734                 iif = nla_get_u32(tb[RTA_IIF]);
2735
2736         if (tb[RTA_OIF])
2737                 oif = nla_get_u32(tb[RTA_OIF]);
2738
2739         if (tb[RTA_MARK])
2740                 fl6.flowi6_mark = nla_get_u32(tb[RTA_MARK]);
2741
2742         if (iif) {
2743                 struct net_device *dev;
2744                 int flags = 0;
2745
2746                 dev = __dev_get_by_index(net, iif);
2747                 if (!dev) {
2748                         err = -ENODEV;
2749                         goto errout;
2750                 }
2751
2752                 fl6.flowi6_iif = iif;
2753
2754                 if (!ipv6_addr_any(&fl6.saddr))
2755                         flags |= RT6_LOOKUP_F_HAS_SADDR;
2756
2757                 rt = (struct rt6_info *)ip6_route_input_lookup(net, dev, &fl6,
2758                                                                flags);
2759         } else {
2760                 fl6.flowi6_oif = oif;
2761
2762                 rt = (struct rt6_info *)ip6_route_output(net, NULL, &fl6);
2763         }
2764
2765         skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
2766         if (!skb) {
2767                 ip6_rt_put(rt);
2768                 err = -ENOBUFS;
2769                 goto errout;
2770         }
2771
2772         /* Reserve room for dummy headers, this skb can pass
2773            through good chunk of routing engine.
2774          */
2775         skb_reset_mac_header(skb);
2776         skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
2777
2778         skb_dst_set(skb, &rt->dst);
2779
2780         err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
2781                             RTM_NEWROUTE, NETLINK_CB(in_skb).portid,
2782                             nlh->nlmsg_seq, 0, 0, 0);
2783         if (err < 0) {
2784                 kfree_skb(skb);
2785                 goto errout;
2786         }
2787
2788         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
2789 errout:
2790         return err;
2791 }
2792
2793 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info)
2794 {
2795         struct sk_buff *skb;
2796         struct net *net = info->nl_net;
2797         u32 seq;
2798         int err;
2799
2800         err = -ENOBUFS;
2801         seq = info->nlh ? info->nlh->nlmsg_seq : 0;
2802
2803         skb = nlmsg_new(rt6_nlmsg_size(), gfp_any());
2804         if (!skb)
2805                 goto errout;
2806
2807         err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
2808                                 event, info->portid, seq, 0, 0, 0);
2809         if (err < 0) {
2810                 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
2811                 WARN_ON(err == -EMSGSIZE);
2812                 kfree_skb(skb);
2813                 goto errout;
2814         }
2815         rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
2816                     info->nlh, gfp_any());
2817         return;
2818 errout:
2819         if (err < 0)
2820                 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
2821 }
2822
2823 static int ip6_route_dev_notify(struct notifier_block *this,
2824                                 unsigned long event, void *ptr)
2825 {
2826         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2827         struct net *net = dev_net(dev);
2828
2829         if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
2830                 net->ipv6.ip6_null_entry->dst.dev = dev;
2831                 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
2832 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2833                 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
2834                 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
2835                 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
2836                 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
2837 #endif
2838         }
2839
2840         return NOTIFY_OK;
2841 }
2842
2843 /*
2844  *      /proc
2845  */
2846
2847 #ifdef CONFIG_PROC_FS
2848
2849 static const struct file_operations ipv6_route_proc_fops = {
2850         .owner          = THIS_MODULE,
2851         .open           = ipv6_route_open,
2852         .read           = seq_read,
2853         .llseek         = seq_lseek,
2854         .release        = seq_release_net,
2855 };
2856
2857 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
2858 {
2859         struct net *net = (struct net *)seq->private;
2860         seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
2861                    net->ipv6.rt6_stats->fib_nodes,
2862                    net->ipv6.rt6_stats->fib_route_nodes,
2863                    net->ipv6.rt6_stats->fib_rt_alloc,
2864                    net->ipv6.rt6_stats->fib_rt_entries,
2865                    net->ipv6.rt6_stats->fib_rt_cache,
2866                    dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
2867                    net->ipv6.rt6_stats->fib_discarded_routes);
2868
2869         return 0;
2870 }
2871
2872 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
2873 {
2874         return single_open_net(inode, file, rt6_stats_seq_show);
2875 }
2876
2877 static const struct file_operations rt6_stats_seq_fops = {
2878         .owner   = THIS_MODULE,
2879         .open    = rt6_stats_seq_open,
2880         .read    = seq_read,
2881         .llseek  = seq_lseek,
2882         .release = single_release_net,
2883 };
2884 #endif  /* CONFIG_PROC_FS */
2885
2886 #ifdef CONFIG_SYSCTL
2887
2888 static
2889 int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
2890                               void __user *buffer, size_t *lenp, loff_t *ppos)
2891 {
2892         struct net *net;
2893         int delay;
2894         if (!write)
2895                 return -EINVAL;
2896
2897         net = (struct net *)ctl->extra1;
2898         delay = net->ipv6.sysctl.flush_delay;
2899         proc_dointvec(ctl, write, buffer, lenp, ppos);
2900         fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
2901         return 0;
2902 }
2903
2904 struct ctl_table ipv6_route_table_template[] = {
2905         {
2906                 .procname       =       "flush",
2907                 .data           =       &init_net.ipv6.sysctl.flush_delay,
2908                 .maxlen         =       sizeof(int),
2909                 .mode           =       0200,
2910                 .proc_handler   =       ipv6_sysctl_rtcache_flush
2911         },
2912         {
2913                 .procname       =       "gc_thresh",
2914                 .data           =       &ip6_dst_ops_template.gc_thresh,
2915                 .maxlen         =       sizeof(int),
2916                 .mode           =       0644,
2917                 .proc_handler   =       proc_dointvec,
2918         },
2919         {
2920                 .procname       =       "max_size",
2921                 .data           =       &init_net.ipv6.sysctl.ip6_rt_max_size,
2922                 .maxlen         =       sizeof(int),
2923                 .mode           =       0644,
2924                 .proc_handler   =       proc_dointvec,
2925         },
2926         {
2927                 .procname       =       "gc_min_interval",
2928                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2929                 .maxlen         =       sizeof(int),
2930                 .mode           =       0644,
2931                 .proc_handler   =       proc_dointvec_jiffies,
2932         },
2933         {
2934                 .procname       =       "gc_timeout",
2935                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
2936                 .maxlen         =       sizeof(int),
2937                 .mode           =       0644,
2938                 .proc_handler   =       proc_dointvec_jiffies,
2939         },
2940         {
2941                 .procname       =       "gc_interval",
2942                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_interval,
2943                 .maxlen         =       sizeof(int),
2944                 .mode           =       0644,
2945                 .proc_handler   =       proc_dointvec_jiffies,
2946         },
2947         {
2948                 .procname       =       "gc_elasticity",
2949                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
2950                 .maxlen         =       sizeof(int),
2951                 .mode           =       0644,
2952                 .proc_handler   =       proc_dointvec,
2953         },
2954         {
2955                 .procname       =       "mtu_expires",
2956                 .data           =       &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
2957                 .maxlen         =       sizeof(int),
2958                 .mode           =       0644,
2959                 .proc_handler   =       proc_dointvec_jiffies,
2960         },
2961         {
2962                 .procname       =       "min_adv_mss",
2963                 .data           =       &init_net.ipv6.sysctl.ip6_rt_min_advmss,
2964                 .maxlen         =       sizeof(int),
2965                 .mode           =       0644,
2966                 .proc_handler   =       proc_dointvec,
2967         },
2968         {
2969                 .procname       =       "gc_min_interval_ms",
2970                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2971                 .maxlen         =       sizeof(int),
2972                 .mode           =       0644,
2973                 .proc_handler   =       proc_dointvec_ms_jiffies,
2974         },
2975         { }
2976 };
2977
2978 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
2979 {
2980         struct ctl_table *table;
2981
2982         table = kmemdup(ipv6_route_table_template,
2983                         sizeof(ipv6_route_table_template),
2984                         GFP_KERNEL);
2985
2986         if (table) {
2987                 table[0].data = &net->ipv6.sysctl.flush_delay;
2988                 table[0].extra1 = net;
2989                 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
2990                 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
2991                 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2992                 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
2993                 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
2994                 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
2995                 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
2996                 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
2997                 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2998
2999                 /* Don't export sysctls to unprivileged users */
3000                 if (net->user_ns != &init_user_ns)
3001                         table[0].procname = NULL;
3002         }
3003
3004         return table;
3005 }
3006 #endif
3007
3008 static int __net_init ip6_route_net_init(struct net *net)
3009 {
3010         int ret = -ENOMEM;
3011
3012         memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
3013                sizeof(net->ipv6.ip6_dst_ops));
3014
3015         if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
3016                 goto out_ip6_dst_ops;
3017
3018         net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
3019                                            sizeof(*net->ipv6.ip6_null_entry),
3020                                            GFP_KERNEL);
3021         if (!net->ipv6.ip6_null_entry)
3022                 goto out_ip6_dst_entries;
3023         net->ipv6.ip6_null_entry->dst.path =
3024                 (struct dst_entry *)net->ipv6.ip6_null_entry;
3025         net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3026         dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
3027                          ip6_template_metrics, true);
3028
3029 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3030         net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
3031                                                sizeof(*net->ipv6.ip6_prohibit_entry),
3032                                                GFP_KERNEL);
3033         if (!net->ipv6.ip6_prohibit_entry)
3034                 goto out_ip6_null_entry;
3035         net->ipv6.ip6_prohibit_entry->dst.path =
3036                 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
3037         net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3038         dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
3039                          ip6_template_metrics, true);
3040
3041         net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
3042                                                sizeof(*net->ipv6.ip6_blk_hole_entry),
3043                                                GFP_KERNEL);
3044         if (!net->ipv6.ip6_blk_hole_entry)
3045                 goto out_ip6_prohibit_entry;
3046         net->ipv6.ip6_blk_hole_entry->dst.path =
3047                 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
3048         net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3049         dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
3050                          ip6_template_metrics, true);
3051 #endif
3052
3053         net->ipv6.sysctl.flush_delay = 0;
3054         net->ipv6.sysctl.ip6_rt_max_size = 4096;
3055         net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
3056         net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
3057         net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
3058         net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
3059         net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
3060         net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
3061
3062         net->ipv6.ip6_rt_gc_expire = 30*HZ;
3063
3064         ret = 0;
3065 out:
3066         return ret;
3067
3068 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3069 out_ip6_prohibit_entry:
3070         kfree(net->ipv6.ip6_prohibit_entry);
3071 out_ip6_null_entry:
3072         kfree(net->ipv6.ip6_null_entry);
3073 #endif
3074 out_ip6_dst_entries:
3075         dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3076 out_ip6_dst_ops:
3077         goto out;
3078 }
3079
3080 static void __net_exit ip6_route_net_exit(struct net *net)
3081 {
3082         kfree(net->ipv6.ip6_null_entry);
3083 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3084         kfree(net->ipv6.ip6_prohibit_entry);
3085         kfree(net->ipv6.ip6_blk_hole_entry);
3086 #endif
3087         dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3088 }
3089
3090 static int __net_init ip6_route_net_init_late(struct net *net)
3091 {
3092 #ifdef CONFIG_PROC_FS
3093         proc_create("ipv6_route", 0, net->proc_net, &ipv6_route_proc_fops);
3094         proc_create("rt6_stats", S_IRUGO, net->proc_net, &rt6_stats_seq_fops);
3095 #endif
3096         return 0;
3097 }
3098
3099 static void __net_exit ip6_route_net_exit_late(struct net *net)
3100 {
3101 #ifdef CONFIG_PROC_FS
3102         remove_proc_entry("ipv6_route", net->proc_net);
3103         remove_proc_entry("rt6_stats", net->proc_net);
3104 #endif
3105 }
3106
3107 static struct pernet_operations ip6_route_net_ops = {
3108         .init = ip6_route_net_init,
3109         .exit = ip6_route_net_exit,
3110 };
3111
3112 static int __net_init ipv6_inetpeer_init(struct net *net)
3113 {
3114         struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
3115
3116         if (!bp)
3117                 return -ENOMEM;
3118         inet_peer_base_init(bp);
3119         net->ipv6.peers = bp;
3120         return 0;
3121 }
3122
3123 static void __net_exit ipv6_inetpeer_exit(struct net *net)
3124 {
3125         struct inet_peer_base *bp = net->ipv6.peers;
3126
3127         net->ipv6.peers = NULL;
3128         inetpeer_invalidate_tree(bp);
3129         kfree(bp);
3130 }
3131
3132 static struct pernet_operations ipv6_inetpeer_ops = {
3133         .init   =       ipv6_inetpeer_init,
3134         .exit   =       ipv6_inetpeer_exit,
3135 };
3136
3137 static struct pernet_operations ip6_route_net_late_ops = {
3138         .init = ip6_route_net_init_late,
3139         .exit = ip6_route_net_exit_late,
3140 };
3141
3142 static struct notifier_block ip6_route_dev_notifier = {
3143         .notifier_call = ip6_route_dev_notify,
3144         .priority = 0,
3145 };
3146
3147 int __init ip6_route_init(void)
3148 {
3149         int ret;
3150
3151         ret = -ENOMEM;
3152         ip6_dst_ops_template.kmem_cachep =
3153                 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
3154                                   SLAB_HWCACHE_ALIGN, NULL);
3155         if (!ip6_dst_ops_template.kmem_cachep)
3156                 goto out;
3157
3158         ret = dst_entries_init(&ip6_dst_blackhole_ops);
3159         if (ret)
3160                 goto out_kmem_cache;
3161
3162         ret = register_pernet_subsys(&ipv6_inetpeer_ops);
3163         if (ret)
3164                 goto out_dst_entries;
3165
3166         ret = register_pernet_subsys(&ip6_route_net_ops);
3167         if (ret)
3168                 goto out_register_inetpeer;
3169
3170         ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
3171
3172         /* Registering of the loopback is done before this portion of code,
3173          * the loopback reference in rt6_info will not be taken, do it
3174          * manually for init_net */
3175         init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
3176         init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3177   #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3178         init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
3179         init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3180         init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
3181         init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3182   #endif
3183         ret = fib6_init();
3184         if (ret)
3185                 goto out_register_subsys;
3186
3187         ret = xfrm6_init();
3188         if (ret)
3189                 goto out_fib6_init;
3190
3191         ret = fib6_rules_init();
3192         if (ret)
3193                 goto xfrm6_init;
3194
3195         ret = register_pernet_subsys(&ip6_route_net_late_ops);
3196         if (ret)
3197                 goto fib6_rules_init;
3198
3199         ret = -ENOBUFS;
3200         if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL, NULL) ||
3201             __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL, NULL) ||
3202             __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL, NULL))
3203                 goto out_register_late_subsys;
3204
3205         ret = register_netdevice_notifier(&ip6_route_dev_notifier);
3206         if (ret)
3207                 goto out_register_late_subsys;
3208
3209 out:
3210         return ret;
3211
3212 out_register_late_subsys:
3213         unregister_pernet_subsys(&ip6_route_net_late_ops);
3214 fib6_rules_init:
3215         fib6_rules_cleanup();
3216 xfrm6_init:
3217         xfrm6_fini();
3218 out_fib6_init:
3219         fib6_gc_cleanup();
3220 out_register_subsys:
3221         unregister_pernet_subsys(&ip6_route_net_ops);
3222 out_register_inetpeer:
3223         unregister_pernet_subsys(&ipv6_inetpeer_ops);
3224 out_dst_entries:
3225         dst_entries_destroy(&ip6_dst_blackhole_ops);
3226 out_kmem_cache:
3227         kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3228         goto out;
3229 }
3230
3231 void ip6_route_cleanup(void)
3232 {
3233         unregister_netdevice_notifier(&ip6_route_dev_notifier);
3234         unregister_pernet_subsys(&ip6_route_net_late_ops);
3235         fib6_rules_cleanup();
3236         xfrm6_fini();
3237         fib6_gc_cleanup();
3238         unregister_pernet_subsys(&ipv6_inetpeer_ops);
3239         unregister_pernet_subsys(&ip6_route_net_ops);
3240         dst_entries_destroy(&ip6_dst_blackhole_ops);
3241         kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3242 }