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 = -2,
  70         RT6_NUD_FAIL_SOFT = -1,
  71         RT6_NUD_SUCCEED = 1
  72 };
  73
  74 static struct rt6_info *ip6_rt_copy(struct rt6_info *ort,
  75                                     const struct in6_addr *dest);
  76 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie);
  77 static unsigned int      ip6_default_advmss(const struct dst_entry *dst);
  78 static unsigned int      ip6_mtu(const struct dst_entry *dst);
  79 static struct dst_entry *ip6_negative_advice(struct dst_entry *);
  80 static void             ip6_dst_destroy(struct dst_entry *);
  81 static void             ip6_dst_ifdown(struct dst_entry *,
  82                                        struct net_device *dev, int how);
  83 static int               ip6_dst_gc(struct dst_ops *ops);
  84
  85 static int              ip6_pkt_discard(struct sk_buff *skb);
  86 static int              ip6_pkt_discard_out(struct sk_buff *skb);
  87 static void             ip6_link_failure(struct sk_buff *skb);
  88 static void             ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
  89                                            struct sk_buff *skb, u32 mtu);
  90 static void             rt6_do_redirect(struct dst_entry *dst, struct sock *sk,
  91                                         struct sk_buff *skb);
  92 static int rt6_score_route(struct rt6_info *rt, int oif, int strict);
  93
  94 #ifdef CONFIG_IPV6_ROUTE_INFO
  95 static struct rt6_info *rt6_add_route_info(struct net *net,
  96                                            const struct in6_addr *prefix, int prefixlen,
  97                                            const struct in6_addr *gwaddr, int ifindex,
  98                                            unsigned int pref);
  99 static struct rt6_info *rt6_get_route_info(struct net *net,
 100                                            const struct in6_addr *prefix, int prefixlen,
 101                                            const struct in6_addr *gwaddr, int ifindex);
 102 #endif
 103
 104 static u32 *ipv6_cow_metrics(struct dst_entry *dst, unsigned long old)
 105 {
 106         struct rt6_info *rt = (struct rt6_info *) dst;
 107         struct inet_peer *peer;
 108         u32 *p = NULL;
 109
 110         if (!(rt->dst.flags & DST_HOST))
 111                 return NULL;
 112
 113         peer = rt6_get_peer_create(rt);
 114         if (peer) {
 115                 u32 *old_p = __DST_METRICS_PTR(old);
 116                 unsigned long prev, new;
 117
 118                 p = peer->metrics;
 119                 if (inet_metrics_new(peer))
 120                         memcpy(p, old_p, sizeof(u32) * RTAX_MAX);
 121
 122                 new = (unsigned long) p;
 123                 prev = cmpxchg(&dst->_metrics, old, new);
 124
 125                 if (prev != old) {
 126                         p = __DST_METRICS_PTR(prev);
 127                         if (prev & DST_METRICS_READ_ONLY)
 128                                 p = NULL;
 129                 }
 130         }
 131         return p;
 132 }
 133
 134 static inline const void *choose_neigh_daddr(struct rt6_info *rt,
 135                                              struct sk_buff *skb,
 136                                              const void *daddr)
 137 {
 138         struct in6_addr *p = &rt->rt6i_gateway;
 139
 140         if (!ipv6_addr_any(p))
 141                 return (const void *) p;
 142         else if (skb)
 143                 return &ipv6_hdr(skb)->daddr;
 144         return daddr;
 145 }
 146
 147 static struct neighbour *ip6_neigh_lookup(const struct dst_entry *dst,
 148                                           struct sk_buff *skb,
 149                                           const void *daddr)
 150 {
 151         struct rt6_info *rt = (struct rt6_info *) dst;
 152         struct neighbour *n;
 153
 154         daddr = choose_neigh_daddr(rt, skb, daddr);
 155         n = __ipv6_neigh_lookup(dst->dev, daddr);
 156         if (n)
 157                 return n;
 158         return neigh_create(&nd_tbl, daddr, dst->dev);
 159 }
 160
 161 static struct dst_ops ip6_dst_ops_template = {
 162         .family                 =       AF_INET6,
 163         .protocol               =       cpu_to_be16(ETH_P_IPV6),
 164         .gc                     =       ip6_dst_gc,
 165         .gc_thresh              =       1024,
 166         .check                  =       ip6_dst_check,
 167         .default_advmss         =       ip6_default_advmss,
 168         .mtu                    =       ip6_mtu,
 169         .cow_metrics            =       ipv6_cow_metrics,
 170         .destroy                =       ip6_dst_destroy,
 171         .ifdown                 =       ip6_dst_ifdown,
 172         .negative_advice        =       ip6_negative_advice,
 173         .link_failure           =       ip6_link_failure,
 174         .update_pmtu            =       ip6_rt_update_pmtu,
 175         .redirect               =       rt6_do_redirect,
 176         .local_out              =       __ip6_local_out,
 177         .neigh_lookup           =       ip6_neigh_lookup,
 178 };
 179
 180 static unsigned int ip6_blackhole_mtu(const struct dst_entry *dst)
 181 {
 182         unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
 183
 184         return mtu ? : dst->dev->mtu;
 185 }
 186
 187 static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, struct sock *sk,
 188                                          struct sk_buff *skb, u32 mtu)
 189 {
 190 }
 191
 192 static void ip6_rt_blackhole_redirect(struct dst_entry *dst, struct sock *sk,
 193                                       struct sk_buff *skb)
 194 {
 195 }
 196
 197 static u32 *ip6_rt_blackhole_cow_metrics(struct dst_entry *dst,
 198                                          unsigned long old)
 199 {
 200         return NULL;
 201 }
 202
 203 static struct dst_ops ip6_dst_blackhole_ops = {
 204         .family                 =       AF_INET6,
 205         .protocol               =       cpu_to_be16(ETH_P_IPV6),
 206         .destroy                =       ip6_dst_destroy,
 207         .check                  =       ip6_dst_check,
 208         .mtu                    =       ip6_blackhole_mtu,
 209         .default_advmss         =       ip6_default_advmss,
 210         .update_pmtu            =       ip6_rt_blackhole_update_pmtu,
 211         .redirect               =       ip6_rt_blackhole_redirect,
 212         .cow_metrics            =       ip6_rt_blackhole_cow_metrics,
 213         .neigh_lookup           =       ip6_neigh_lookup,
 214 };
 215
 216 static const u32 ip6_template_metrics[RTAX_MAX] = {
 217         [RTAX_HOPLIMIT - 1] = 0,
 218 };
 219
 220 static const struct rt6_info ip6_null_entry_template = {
 221         .dst = {
 222                 .__refcnt       = ATOMIC_INIT(1),
 223                 .__use          = 1,
 224                 .obsolete       = DST_OBSOLETE_FORCE_CHK,
 225                 .error          = -ENETUNREACH,
 226                 .input          = ip6_pkt_discard,
 227                 .output         = ip6_pkt_discard_out,
 228         },
 229         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
 230         .rt6i_protocol  = RTPROT_KERNEL,
 231         .rt6i_metric    = ~(u32) 0,
 232         .rt6i_ref       = ATOMIC_INIT(1),
 233 };
 234
 235 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
 236
 237 static int ip6_pkt_prohibit(struct sk_buff *skb);
 238 static int ip6_pkt_prohibit_out(struct sk_buff *skb);
 239
 240 static const struct rt6_info ip6_prohibit_entry_template = {
 241         .dst = {
 242                 .__refcnt       = ATOMIC_INIT(1),
 243                 .__use          = 1,
 244                 .obsolete       = DST_OBSOLETE_FORCE_CHK,
 245                 .error          = -EACCES,
 246                 .input          = ip6_pkt_prohibit,
 247                 .output         = ip6_pkt_prohibit_out,
 248         },
 249         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
 250         .rt6i_protocol  = RTPROT_KERNEL,
 251         .rt6i_metric    = ~(u32) 0,
 252         .rt6i_ref       = ATOMIC_INIT(1),
 253 };
 254
 255 static const struct rt6_info ip6_blk_hole_entry_template = {
 256         .dst = {
 257                 .__refcnt       = ATOMIC_INIT(1),
 258                 .__use          = 1,
 259                 .obsolete       = DST_OBSOLETE_FORCE_CHK,
 260                 .error          = -EINVAL,
 261                 .input          = dst_discard,
 262                 .output         = dst_discard,
 263         },
 264         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
 265         .rt6i_protocol  = RTPROT_KERNEL,
 266         .rt6i_metric    = ~(u32) 0,
 267         .rt6i_ref       = ATOMIC_INIT(1),
 268 };
 269
 270 #endif
 271
 272 /* allocate dst with ip6_dst_ops */
 273 static inline struct rt6_info *ip6_dst_alloc(struct net *net,
 274                                              struct net_device *dev,
 275                                              int flags,
 276                                              struct fib6_table *table)
 277 {
 278         struct rt6_info *rt = dst_alloc(&net->ipv6.ip6_dst_ops, dev,
 279                                         0, DST_OBSOLETE_FORCE_CHK, flags);
 280
 281         if (rt) {
 282                 struct dst_entry *dst = &rt->dst;
 283
 284                 memset(dst + 1, 0, sizeof(*rt) - sizeof(*dst));
 285                 rt6_init_peer(rt, table ? &table->tb6_peers : net->ipv6.peers);
 286                 rt->rt6i_genid = rt_genid_ipv6(net);
 287                 INIT_LIST_HEAD(&rt->rt6i_siblings);
 288         }
 289         return rt;
 290 }
 291
 292 static void ip6_dst_destroy(struct dst_entry *dst)
 293 {
 294         struct rt6_info *rt = (struct rt6_info *)dst;
 295         struct inet6_dev *idev = rt->rt6i_idev;
 296         struct dst_entry *from = dst->from;
 297
 298         if (!(rt->dst.flags & DST_HOST))
 299                 dst_destroy_metrics_generic(dst);
 300
 301         if (idev) {
 302                 rt->rt6i_idev = NULL;
 303                 in6_dev_put(idev);
 304         }
 305
 306         dst->from = NULL;
 307         dst_release(from);
 308
 309         if (rt6_has_peer(rt)) {
 310                 struct inet_peer *peer = rt6_peer_ptr(rt);
 311                 inet_putpeer(peer);
 312         }
 313 }
 314
 315 void rt6_bind_peer(struct rt6_info *rt, int create)
 316 {
 317         struct inet_peer_base *base;
 318         struct inet_peer *peer;
 319
 320         base = inetpeer_base_ptr(rt->_rt6i_peer);
 321         if (!base)
 322                 return;
 323
 324         peer = inet_getpeer_v6(base, &rt->rt6i_dst.addr, create);
 325         if (peer) {
 326                 if (!rt6_set_peer(rt, peer))
 327                         inet_putpeer(peer);
 328         }
 329 }
 330
 331 static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
 332                            int how)
 333 {
 334         struct rt6_info *rt = (struct rt6_info *)dst;
 335         struct inet6_dev *idev = rt->rt6i_idev;
 336         struct net_device *loopback_dev =
 337                 dev_net(dev)->loopback_dev;
 338
 339         if (dev != loopback_dev) {
 340                 if (idev && idev->dev == dev) {
 341                         struct inet6_dev *loopback_idev =
 342                                 in6_dev_get(loopback_dev);
 343                         if (loopback_idev) {
 344                                 rt->rt6i_idev = loopback_idev;
 345                                 in6_dev_put(idev);
 346                         }
 347                 }
 348         }
 349 }
 350
 351 static bool rt6_check_expired(const struct rt6_info *rt)
 352 {
 353         if (rt->rt6i_flags & RTF_EXPIRES) {
 354                 if (time_after(jiffies, rt->dst.expires))
 355                         return true;
 356         } else if (rt->dst.from) {
 357                 return rt6_check_expired((struct rt6_info *) rt->dst.from);
 358         }
 359         return false;
 360 }
 361
 362 static bool rt6_need_strict(const struct in6_addr *daddr)
 363 {
 364         return ipv6_addr_type(daddr) &
 365                 (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
 366 }
 367
 368 /* Multipath route selection:
 369  *   Hash based function using packet header and flowlabel.
 370  * Adapted from fib_info_hashfn()
 371  */
 372 static int rt6_info_hash_nhsfn(unsigned int candidate_count,
 373                                const struct flowi6 *fl6)
 374 {
 375         unsigned int val = fl6->flowi6_proto;
 376
 377         val ^= ipv6_addr_hash(&fl6->daddr);
 378         val ^= ipv6_addr_hash(&fl6->saddr);
 379
 380         /* Work only if this not encapsulated */
 381         switch (fl6->flowi6_proto) {
 382         case IPPROTO_UDP:
 383         case IPPROTO_TCP:
 384         case IPPROTO_SCTP:
 385                 val ^= (__force u16)fl6->fl6_sport;
 386                 val ^= (__force u16)fl6->fl6_dport;
 387                 break;
 388
 389         case IPPROTO_ICMPV6:
 390                 val ^= (__force u16)fl6->fl6_icmp_type;
 391                 val ^= (__force u16)fl6->fl6_icmp_code;
 392                 break;
 393         }
 394         /* RFC6438 recommands to use flowlabel */
 395         val ^= (__force u32)fl6->flowlabel;
 396
 397         /* Perhaps, we need to tune, this function? */
 398         val = val ^ (val >> 7) ^ (val >> 12);
 399         return val % candidate_count;
 400 }
 401
 402 static struct rt6_info *rt6_multipath_select(struct rt6_info *match,
 403                                              struct flowi6 *fl6, int oif,
 404                                              int strict)
 405 {
 406         struct rt6_info *sibling, *next_sibling;
 407         int route_choosen;
 408
 409         route_choosen = rt6_info_hash_nhsfn(match->rt6i_nsiblings + 1, fl6);
 410         /* Don't change the route, if route_choosen == 0
 411          * (siblings does not include ourself)
 412          */
 413         if (route_choosen)
 414                 list_for_each_entry_safe(sibling, next_sibling,
 415                                 &match->rt6i_siblings, rt6i_siblings) {
 416                         route_choosen--;
 417                         if (route_choosen == 0) {
 418                                 if (rt6_score_route(sibling, oif, strict) < 0)
 419                                         break;
 420                                 match = sibling;
 421                                 break;
 422                         }
 423                 }
 424         return match;
 425 }
 426
 427 /*
 428  *      Route lookup. Any table->tb6_lock is implied.
 429  */
 430
 431 static inline struct rt6_info *rt6_device_match(struct net *net,
 432                                                     struct rt6_info *rt,
 433                                                     const struct in6_addr *saddr,
 434                                                     int oif,
 435                                                     int flags)
 436 {
 437         struct rt6_info *local = NULL;
 438         struct rt6_info *sprt;
 439
 440         if (!oif && ipv6_addr_any(saddr))
 441                 goto out;
 442
 443         for (sprt = rt; sprt; sprt = sprt->dst.rt6_next) {
 444                 struct net_device *dev = sprt->dst.dev;
 445
 446                 if (oif) {
 447                         if (dev->ifindex == oif)
 448                                 return sprt;
 449                         if (dev->flags & IFF_LOOPBACK) {
 450                                 if (!sprt->rt6i_idev ||
 451                                     sprt->rt6i_idev->dev->ifindex != oif) {
 452                                         if (flags & RT6_LOOKUP_F_IFACE && oif)
 453                                                 continue;
 454                                         if (local && (!oif ||
 455                                                       local->rt6i_idev->dev->ifindex == oif))
 456                                                 continue;
 457                                 }
 458                                 local = sprt;
 459                         }
 460                 } else {
 461                         if (ipv6_chk_addr(net, saddr, dev,
 462                                           flags & RT6_LOOKUP_F_IFACE))
 463                                 return sprt;
 464                 }
 465         }
 466
 467         if (oif) {
 468                 if (local)
 469                         return local;
 470
 471                 if (flags & RT6_LOOKUP_F_IFACE)
 472                         return net->ipv6.ip6_null_entry;
 473         }
 474 out:
 475         return rt;
 476 }
 477
 478 #ifdef CONFIG_IPV6_ROUTER_PREF
 479 struct __rt6_probe_work {
 480         struct work_struct work;
 481         struct in6_addr target;
 482         struct net_device *dev;
 483 };
 484
 485 static void rt6_probe_deferred(struct work_struct *w)
 486 {
 487         struct in6_addr mcaddr;
 488         struct __rt6_probe_work *work =
 489                 container_of(w, struct __rt6_probe_work, work);
 490
 491         addrconf_addr_solict_mult(&work->target, &mcaddr);
 492         ndisc_send_ns(work->dev, NULL, &work->target, &mcaddr, NULL);
 493         dev_put(work->dev);
 494         kfree(w);
 495 }
 496
 497 static void rt6_probe(struct rt6_info *rt)
 498 {
 499         struct neighbour *neigh;
 500         /*
 501          * Okay, this does not seem to be appropriate
 502          * for now, however, we need to check if it
 503          * is really so; aka Router Reachability Probing.
 504          *
 505          * Router Reachability Probe MUST be rate-limited
 506          * to no more than one per minute.
 507          */
 508         if (!rt || !(rt->rt6i_flags & RTF_GATEWAY))
 509                 return;
 510         rcu_read_lock_bh();
 511         neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
 512         if (neigh) {
 513                 write_lock(&neigh->lock);
 514                 if (neigh->nud_state & NUD_VALID)
 515                         goto out;
 516         }
 517
 518         if (!neigh ||
 519             time_after(jiffies, neigh->updated + rt->rt6i_idev->cnf.rtr_probe_interval)) {
 520                 struct __rt6_probe_work *work;
 521
 522                 work = kmalloc(sizeof(*work), GFP_ATOMIC);
 523
 524                 if (neigh && work)
 525                         neigh->updated = jiffies;
 526
 527                 if (neigh)
 528                         write_unlock(&neigh->lock);
 529
 530                 if (work) {
 531                         INIT_WORK(&work->work, rt6_probe_deferred);
 532                         work->target = rt->rt6i_gateway;
 533                         dev_hold(rt->dst.dev);
 534                         work->dev = rt->dst.dev;
 535                         schedule_work(&work->work);
 536                 }
 537         } else {
 538 out:
 539                 write_unlock(&neigh->lock);
 540         }
 541         rcu_read_unlock_bh();
 542 }
 543 #else
 544 static inline void rt6_probe(struct rt6_info *rt)
 545 {
 546 }
 547 #endif
 548
 549 /*
 550  * Default Router Selection (RFC 2461 6.3.6)
 551  */
 552 static inline int rt6_check_dev(struct rt6_info *rt, int oif)
 553 {
 554         struct net_device *dev = rt->dst.dev;
 555         if (!oif || dev->ifindex == oif)
 556                 return 2;
 557         if ((dev->flags & IFF_LOOPBACK) &&
 558             rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
 559                 return 1;
 560         return 0;
 561 }
 562
 563 static inline enum rt6_nud_state rt6_check_neigh(struct rt6_info *rt)
 564 {
 565         struct neighbour *neigh;
 566         enum rt6_nud_state ret = RT6_NUD_FAIL_HARD;
 567
 568         if (rt->rt6i_flags & RTF_NONEXTHOP ||
 569             !(rt->rt6i_flags & RTF_GATEWAY))
 570                 return RT6_NUD_SUCCEED;
 571
 572         rcu_read_lock_bh();
 573         neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
 574         if (neigh) {
 575                 read_lock(&neigh->lock);
 576                 if (neigh->nud_state & NUD_VALID)
 577                         ret = RT6_NUD_SUCCEED;
 578 #ifdef CONFIG_IPV6_ROUTER_PREF
 579                 else if (!(neigh->nud_state & NUD_FAILED))
 580                         ret = RT6_NUD_SUCCEED;
 581 #endif
 582                 read_unlock(&neigh->lock);
 583         } else {
 584                 ret = IS_ENABLED(CONFIG_IPV6_ROUTER_PREF) ?
 585                       RT6_NUD_SUCCEED : RT6_NUD_FAIL_SOFT;
 586         }
 587         rcu_read_unlock_bh();
 588
 589         return ret;
 590 }
 591
 592 static int rt6_score_route(struct rt6_info *rt, int oif,
 593                            int strict)
 594 {
 595         int m;
 596
 597         m = rt6_check_dev(rt, oif);
 598         if (!m && (strict & RT6_LOOKUP_F_IFACE))
 599                 return RT6_NUD_FAIL_HARD;
 600 #ifdef CONFIG_IPV6_ROUTER_PREF
 601         m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
 602 #endif
 603         if (strict & RT6_LOOKUP_F_REACHABLE) {
 604                 int n = rt6_check_neigh(rt);
 605                 if (n < 0)
 606                         return n;
 607         }
 608         return m;
 609 }
 610
 611 static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
 612                                    int *mpri, struct rt6_info *match,
 613                                    bool *do_rr)
 614 {
 615         int m;
 616         bool match_do_rr = false;
 617
 618         if (rt6_check_expired(rt))
 619                 goto out;
 620
 621         m = rt6_score_route(rt, oif, strict);
 622         if (m == RT6_NUD_FAIL_SOFT && !IS_ENABLED(CONFIG_IPV6_ROUTER_PREF)) {
 623                 match_do_rr = true;
 624                 m = 0; /* lowest valid score */
 625         } else if (m < 0) {
 626                 goto out;
 627         }
 628
 629         if (strict & RT6_LOOKUP_F_REACHABLE)
 630                 rt6_probe(rt);
 631
 632         if (m > *mpri) {
 633                 *do_rr = match_do_rr;
 634                 *mpri = m;
 635                 match = rt;
 636         }
 637 out:
 638         return match;
 639 }
 640
 641 static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
 642                                      struct rt6_info *rr_head,
 643                                      u32 metric, int oif, int strict,
 644                                      bool *do_rr)
 645 {
 646         struct rt6_info *rt, *match;
 647         int mpri = -1;
 648
 649         match = NULL;
 650         for (rt = rr_head; rt && rt->rt6i_metric == metric;
 651              rt = rt->dst.rt6_next)
 652                 match = find_match(rt, oif, strict, &mpri, match, do_rr);
 653         for (rt = fn->leaf; rt && rt != rr_head && rt->rt6i_metric == metric;
 654              rt = rt->dst.rt6_next)
 655                 match = find_match(rt, oif, strict, &mpri, match, do_rr);
 656
 657         return match;
 658 }
 659
 660 static struct rt6_info *rt6_select(struct fib6_node *fn, int oif, int strict)
 661 {
 662         struct rt6_info *match, *rt0;
 663         struct net *net;
 664         bool do_rr = false;
 665
 666         rt0 = fn->rr_ptr;
 667         if (!rt0)
 668                 fn->rr_ptr = rt0 = fn->leaf;
 669
 670         match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict,
 671                              &do_rr);
 672
 673         if (do_rr) {
 674                 struct rt6_info *next = rt0->dst.rt6_next;
 675
 676                 /* no entries matched; do round-robin */
 677                 if (!next || next->rt6i_metric != rt0->rt6i_metric)
 678                         next = fn->leaf;
 679
 680                 if (next != rt0)
 681                         fn->rr_ptr = next;
 682         }
 683
 684         net = dev_net(rt0->dst.dev);
 685         return match ? match : net->ipv6.ip6_null_entry;
 686 }
 687
 688 #ifdef CONFIG_IPV6_ROUTE_INFO
 689 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
 690                   const struct in6_addr *gwaddr)
 691 {
 692         struct net *net = dev_net(dev);
 693         struct route_info *rinfo = (struct route_info *) opt;
 694         struct in6_addr prefix_buf, *prefix;
 695         unsigned int pref;
 696         unsigned long lifetime;
 697         struct rt6_info *rt;
 698
 699         if (len < sizeof(struct route_info)) {
 700                 return -EINVAL;
 701         }
 702
 703         /* Sanity check for prefix_len and length */
 704         if (rinfo->length > 3) {
 705                 return -EINVAL;
 706         } else if (rinfo->prefix_len > 128) {
 707                 return -EINVAL;
 708         } else if (rinfo->prefix_len > 64) {
 709                 if (rinfo->length < 2) {
 710                         return -EINVAL;
 711                 }
 712         } else if (rinfo->prefix_len > 0) {
 713                 if (rinfo->length < 1) {
 714                         return -EINVAL;
 715                 }
 716         }
 717
 718         pref = rinfo->route_pref;
 719         if (pref == ICMPV6_ROUTER_PREF_INVALID)
 720                 return -EINVAL;
 721
 722         lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
 723
 724         if (rinfo->length == 3)
 725                 prefix = (struct in6_addr *)rinfo->prefix;
 726         else {
 727                 /* this function is safe */
 728                 ipv6_addr_prefix(&prefix_buf,
 729                                  (struct in6_addr *)rinfo->prefix,
 730                                  rinfo->prefix_len);
 731                 prefix = &prefix_buf;
 732         }
 733
 734         rt = rt6_get_route_info(net, prefix, rinfo->prefix_len, gwaddr,
 735                                 dev->ifindex);
 736
 737         if (rt && !lifetime) {
 738                 ip6_del_rt(rt);
 739                 rt = NULL;
 740         }
 741
 742         if (!rt && lifetime)
 743                 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr, dev->ifindex,
 744                                         pref);
 745         else if (rt)
 746                 rt->rt6i_flags = RTF_ROUTEINFO |
 747                                  (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
 748
 749         if (rt) {
 750                 if (!addrconf_finite_timeout(lifetime))
 751                         rt6_clean_expires(rt);
 752                 else
 753                         rt6_set_expires(rt, jiffies + HZ * lifetime);
 754
 755                 ip6_rt_put(rt);
 756         }
 757         return 0;
 758 }
 759 #endif
 760
 761 #define BACKTRACK(__net, saddr)                 \
 762 do { \
 763         if (rt == __net->ipv6.ip6_null_entry) { \
 764                 struct fib6_node *pn; \
 765                 while (1) { \
 766                         if (fn->fn_flags & RTN_TL_ROOT) \
 767                                 goto out; \
 768                         pn = fn->parent; \
 769                         if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn) \
 770                                 fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr); \
 771                         else \
 772                                 fn = pn; \
 773                         if (fn->fn_flags & RTN_RTINFO) \
 774                                 goto restart; \
 775                 } \
 776         } \
 777 } while (0)
 778
 779 static struct rt6_info *ip6_pol_route_lookup(struct net *net,
 780                                              struct fib6_table *table,
 781                                              struct flowi6 *fl6, int flags)
 782 {
 783         struct fib6_node *fn;
 784         struct rt6_info *rt;
 785
 786         read_lock_bh(&table->tb6_lock);
 787         fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
 788 restart:
 789         rt = fn->leaf;
 790         rt = rt6_device_match(net, rt, &fl6->saddr, fl6->flowi6_oif, flags);
 791         if (rt->rt6i_nsiblings && fl6->flowi6_oif == 0)
 792                 rt = rt6_multipath_select(rt, fl6, fl6->flowi6_oif, flags);
 793         BACKTRACK(net, &fl6->saddr);
 794 out:
 795         dst_use(&rt->dst, jiffies);
 796         read_unlock_bh(&table->tb6_lock);
 797         return rt;
 798
 799 }
 800
 801 struct dst_entry * ip6_route_lookup(struct net *net, struct flowi6 *fl6,
 802                                     int flags)
 803 {
 804         return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_lookup);
 805 }
 806 EXPORT_SYMBOL_GPL(ip6_route_lookup);
 807
 808 struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
 809                             const struct in6_addr *saddr, int oif, int strict)
 810 {
 811         struct flowi6 fl6 = {
 812                 .flowi6_oif = oif,
 813                 .daddr = *daddr,
 814         };
 815         struct dst_entry *dst;
 816         int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
 817
 818         if (saddr) {
 819                 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
 820                 flags |= RT6_LOOKUP_F_HAS_SADDR;
 821         }
 822
 823         dst = fib6_rule_lookup(net, &fl6, flags, ip6_pol_route_lookup);
 824         if (dst->error == 0)
 825                 return (struct rt6_info *) dst;
 826
 827         dst_release(dst);
 828
 829         return NULL;
 830 }
 831
 832 EXPORT_SYMBOL(rt6_lookup);
 833
 834 /* ip6_ins_rt is called with FREE table->tb6_lock.
 835    It takes new route entry, the addition fails by any reason the
 836    route is freed. In any case, if caller does not hold it, it may
 837    be destroyed.
 838  */
 839
 840 static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info)
 841 {
 842         int err;
 843         struct fib6_table *table;
 844
 845         table = rt->rt6i_table;
 846         write_lock_bh(&table->tb6_lock);
 847         err = fib6_add(&table->tb6_root, rt, info);
 848         write_unlock_bh(&table->tb6_lock);
 849
 850         return err;
 851 }
 852
 853 int ip6_ins_rt(struct rt6_info *rt)
 854 {
 855         struct nl_info info = {
 856                 .nl_net = dev_net(rt->dst.dev),
 857         };
 858         return __ip6_ins_rt(rt, &info);
 859 }
 860
 861 static struct rt6_info *rt6_alloc_cow(struct rt6_info *ort,
 862                                       const struct in6_addr *daddr,
 863                                       const struct in6_addr *saddr)
 864 {
 865         struct rt6_info *rt;
 866
 867         /*
 868          *      Clone the route.
 869          */
 870
 871         rt = ip6_rt_copy(ort, daddr);
 872
 873         if (rt) {
 874                 if (!(rt->rt6i_flags & RTF_GATEWAY)) {
 875                         if (ort->rt6i_dst.plen != 128 &&
 876                             ipv6_addr_equal(&ort->rt6i_dst.addr, daddr))
 877                                 rt->rt6i_flags |= RTF_ANYCAST;
 878                 }
 879
 880                 rt->rt6i_flags |= RTF_CACHE;
 881
 882 #ifdef CONFIG_IPV6_SUBTREES
 883                 if (rt->rt6i_src.plen && saddr) {
 884                         rt->rt6i_src.addr = *saddr;
 885                         rt->rt6i_src.plen = 128;
 886                 }
 887 #endif
 888         }
 889
 890         return rt;
 891 }
 892
 893 static struct rt6_info *rt6_alloc_clone(struct rt6_info *ort,
 894                                         const struct in6_addr *daddr)
 895 {
 896         struct rt6_info *rt = ip6_rt_copy(ort, daddr);
 897
 898         if (rt)
 899                 rt->rt6i_flags |= RTF_CACHE;
 900         return rt;
 901 }
 902
 903 static struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table, int oif,
 904                                       struct flowi6 *fl6, int flags)
 905 {
 906         struct fib6_node *fn;
 907         struct rt6_info *rt, *nrt;
 908         int strict = 0;
 909         int attempts = 3;
 910         int err;
 911         int reachable = net->ipv6.devconf_all->forwarding ? 0 : RT6_LOOKUP_F_REACHABLE;
 912
 913         strict |= flags & RT6_LOOKUP_F_IFACE;
 914
 915 relookup:
 916         read_lock_bh(&table->tb6_lock);
 917
 918 restart_2:
 919         fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
 920
 921 restart:
 922         rt = rt6_select(fn, oif, strict | reachable);
 923         if (rt->rt6i_nsiblings)
 924                 rt = rt6_multipath_select(rt, fl6, oif, strict | reachable);
 925         BACKTRACK(net, &fl6->saddr);
 926         if (rt == net->ipv6.ip6_null_entry ||
 927             rt->rt6i_flags & RTF_CACHE)
 928                 goto out;
 929
 930         dst_hold(&rt->dst);
 931         read_unlock_bh(&table->tb6_lock);
 932
 933         if (!(rt->rt6i_flags & (RTF_NONEXTHOP | RTF_GATEWAY)))
 934                 nrt = rt6_alloc_cow(rt, &fl6->daddr, &fl6->saddr);
 935         else if (!(rt->dst.flags & DST_HOST))
 936                 nrt = rt6_alloc_clone(rt, &fl6->daddr);
 937         else
 938                 goto out2;
 939
 940         ip6_rt_put(rt);
 941         rt = nrt ? : net->ipv6.ip6_null_entry;
 942
 943         dst_hold(&rt->dst);
 944         if (nrt) {
 945                 err = ip6_ins_rt(nrt);
 946                 if (!err)
 947                         goto out2;
 948         }
 949
 950         if (--attempts <= 0)
 951                 goto out2;
 952
 953         /*
 954          * Race condition! In the gap, when table->tb6_lock was
 955          * released someone could insert this route.  Relookup.
 956          */
 957         ip6_rt_put(rt);
 958         goto relookup;
 959
 960 out:
 961         if (reachable) {
 962                 reachable = 0;
 963                 goto restart_2;
 964         }
 965         dst_hold(&rt->dst);
 966         read_unlock_bh(&table->tb6_lock);
 967 out2:
 968         rt->dst.lastuse = jiffies;
 969         rt->dst.__use++;
 970
 971         return rt;
 972 }
 973
 974 static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
 975                                             struct flowi6 *fl6, int flags)
 976 {
 977         return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, flags);
 978 }
 979
 980 static struct dst_entry *ip6_route_input_lookup(struct net *net,
 981                                                 struct net_device *dev,
 982                                                 struct flowi6 *fl6, int flags)
 983 {
 984         if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
 985                 flags |= RT6_LOOKUP_F_IFACE;
 986
 987         return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_input);
 988 }
 989
 990 void ip6_route_input(struct sk_buff *skb)
 991 {
 992         const struct ipv6hdr *iph = ipv6_hdr(skb);
 993         struct net *net = dev_net(skb->dev);
 994         int flags = RT6_LOOKUP_F_HAS_SADDR;
 995         struct flowi6 fl6 = {
 996                 .flowi6_iif = skb->dev->ifindex,
 997                 .daddr = iph->daddr,
 998                 .saddr = iph->saddr,
 999                 .flowlabel = ip6_flowinfo(iph),
1000                 .flowi6_mark = skb->mark,
1001                 .flowi6_proto = iph->nexthdr,
1002         };
1003
1004         skb_dst_set(skb, ip6_route_input_lookup(net, skb->dev, &fl6, flags));
1005 }
1006
1007 static struct rt6_info *ip6_pol_route_output(struct net *net, struct fib6_table *table,
1008                                              struct flowi6 *fl6, int flags)
1009 {
1010         return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, flags);
1011 }
1012
1013 struct dst_entry * ip6_route_output(struct net *net, const struct sock *sk,
1014                                     struct flowi6 *fl6)
1015 {
1016         int flags = 0;
1017
1018         fl6->flowi6_iif = LOOPBACK_IFINDEX;
1019
1020         if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr))
1021                 flags |= RT6_LOOKUP_F_IFACE;
1022
1023         if (!ipv6_addr_any(&fl6->saddr))
1024                 flags |= RT6_LOOKUP_F_HAS_SADDR;
1025         else if (sk)
1026                 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
1027
1028         return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_output);
1029 }
1030
1031 EXPORT_SYMBOL(ip6_route_output);
1032
1033 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
1034 {
1035         struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
1036         struct dst_entry *new = NULL;
1037
1038         rt = dst_alloc(&ip6_dst_blackhole_ops, ort->dst.dev, 1, DST_OBSOLETE_NONE, 0);
1039         if (rt) {
1040                 new = &rt->dst;
1041
1042                 memset(new + 1, 0, sizeof(*rt) - sizeof(*new));
1043                 rt6_init_peer(rt, net->ipv6.peers);
1044
1045                 new->__use = 1;
1046                 new->input = dst_discard;
1047                 new->output = dst_discard;
1048
1049                 if (dst_metrics_read_only(&ort->dst))
1050                         new->_metrics = ort->dst._metrics;
1051                 else
1052                         dst_copy_metrics(new, &ort->dst);
1053                 rt->rt6i_idev = ort->rt6i_idev;
1054                 if (rt->rt6i_idev)
1055                         in6_dev_hold(rt->rt6i_idev);
1056
1057                 rt->rt6i_gateway = ort->rt6i_gateway;
1058                 rt->rt6i_flags = ort->rt6i_flags;
1059                 rt->rt6i_metric = 0;
1060
1061                 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1062 #ifdef CONFIG_IPV6_SUBTREES
1063                 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1064 #endif
1065
1066                 dst_free(new);
1067         }
1068
1069         dst_release(dst_orig);
1070         return new ? new : ERR_PTR(-ENOMEM);
1071 }
1072
1073 /*
1074  *      Destination cache support functions
1075  */
1076
1077 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
1078 {
1079         struct rt6_info *rt;
1080
1081         rt = (struct rt6_info *) dst;
1082
1083         /* All IPV6 dsts are created with ->obsolete set to the value
1084          * DST_OBSOLETE_FORCE_CHK which forces validation calls down
1085          * into this function always.
1086          */
1087         if (rt->rt6i_genid != rt_genid_ipv6(dev_net(rt->dst.dev)))
1088                 return NULL;
1089
1090         if (rt->rt6i_node && (rt->rt6i_node->fn_sernum == cookie))
1091                 return dst;
1092
1093         return NULL;
1094 }
1095
1096 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
1097 {
1098         struct rt6_info *rt = (struct rt6_info *) dst;
1099
1100         if (rt) {
1101                 if (rt->rt6i_flags & RTF_CACHE) {
1102                         if (rt6_check_expired(rt)) {
1103                                 ip6_del_rt(rt);
1104                                 dst = NULL;
1105                         }
1106                 } else {
1107                         dst_release(dst);
1108                         dst = NULL;
1109                 }
1110         }
1111         return dst;
1112 }
1113
1114 static void ip6_link_failure(struct sk_buff *skb)
1115 {
1116         struct rt6_info *rt;
1117
1118         icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
1119
1120         rt = (struct rt6_info *) skb_dst(skb);
1121         if (rt) {
1122                 if (rt->rt6i_flags & RTF_CACHE) {
1123                         dst_hold(&rt->dst);
1124                         if (ip6_del_rt(rt))
1125                                 dst_free(&rt->dst);
1126                 } else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT)) {
1127                         rt->rt6i_node->fn_sernum = -1;
1128                 }
1129         }
1130 }
1131
1132 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
1133                                struct sk_buff *skb, u32 mtu)
1134 {
1135         struct rt6_info *rt6 = (struct rt6_info*)dst;
1136
1137         dst_confirm(dst);
1138         if (mtu < dst_mtu(dst) && rt6->rt6i_dst.plen == 128) {
1139                 struct net *net = dev_net(dst->dev);
1140
1141                 rt6->rt6i_flags |= RTF_MODIFIED;
1142                 if (mtu < IPV6_MIN_MTU) {
1143                         u32 features = dst_metric(dst, RTAX_FEATURES);
1144                         mtu = IPV6_MIN_MTU;
1145                         features |= RTAX_FEATURE_ALLFRAG;
1146                         dst_metric_set(dst, RTAX_FEATURES, features);
1147                 }
1148                 dst_metric_set(dst, RTAX_MTU, mtu);
1149                 rt6_update_expires(rt6, net->ipv6.sysctl.ip6_rt_mtu_expires);
1150         }
1151 }
1152
1153 void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
1154                      int oif, u32 mark)
1155 {
1156         const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1157         struct dst_entry *dst;
1158         struct flowi6 fl6;
1159
1160         memset(&fl6, 0, sizeof(fl6));
1161         fl6.flowi6_oif = oif;
1162         fl6.flowi6_mark = mark;
1163         fl6.daddr = iph->daddr;
1164         fl6.saddr = iph->saddr;
1165         fl6.flowlabel = ip6_flowinfo(iph);
1166
1167         dst = ip6_route_output(net, NULL, &fl6);
1168         if (!dst->error)
1169                 ip6_rt_update_pmtu(dst, NULL, skb, ntohl(mtu));
1170         dst_release(dst);
1171 }
1172 EXPORT_SYMBOL_GPL(ip6_update_pmtu);
1173
1174 void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
1175 {
1176         ip6_update_pmtu(skb, sock_net(sk), mtu,
1177                         sk->sk_bound_dev_if, sk->sk_mark);
1178 }
1179 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
1180
1181 /* Handle redirects */
1182 struct ip6rd_flowi {
1183         struct flowi6 fl6;
1184         struct in6_addr gateway;
1185 };
1186
1187 static struct rt6_info *__ip6_route_redirect(struct net *net,
1188                                              struct fib6_table *table,
1189                                              struct flowi6 *fl6,
1190                                              int flags)
1191 {
1192         struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
1193         struct rt6_info *rt;
1194         struct fib6_node *fn;
1195
1196         /* Get the "current" route for this destination and
1197          * check if the redirect has come from approriate router.
1198          *
1199          * RFC 4861 specifies that redirects should only be
1200          * accepted if they come from the nexthop to the target.
1201          * Due to the way the routes are chosen, this notion
1202          * is a bit fuzzy and one might need to check all possible
1203          * routes.
1204          */
1205
1206         read_lock_bh(&table->tb6_lock);
1207         fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1208 restart:
1209         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1210                 if (rt6_check_expired(rt))
1211                         continue;
1212                 if (rt->dst.error)
1213                         break;
1214                 if (!(rt->rt6i_flags & RTF_GATEWAY))
1215                         continue;
1216                 if (fl6->flowi6_oif != rt->dst.dev->ifindex)
1217                         continue;
1218                 if (!ipv6_addr_equal(&rdfl->gateway, &rt->rt6i_gateway))
1219                         continue;
1220                 break;
1221         }
1222
1223         if (!rt)
1224                 rt = net->ipv6.ip6_null_entry;
1225         else if (rt->dst.error) {
1226                 rt = net->ipv6.ip6_null_entry;
1227                 goto out;
1228         }
1229         BACKTRACK(net, &fl6->saddr);
1230 out:
1231         dst_hold(&rt->dst);
1232
1233         read_unlock_bh(&table->tb6_lock);
1234
1235         return rt;
1236 };
1237
1238 static struct dst_entry *ip6_route_redirect(struct net *net,
1239                                         const struct flowi6 *fl6,
1240                                         const struct in6_addr *gateway)
1241 {
1242         int flags = RT6_LOOKUP_F_HAS_SADDR;
1243         struct ip6rd_flowi rdfl;
1244
1245         rdfl.fl6 = *fl6;
1246         rdfl.gateway = *gateway;
1247
1248         return fib6_rule_lookup(net, &rdfl.fl6,
1249                                 flags, __ip6_route_redirect);
1250 }
1251
1252 void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark)
1253 {
1254         const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1255         struct dst_entry *dst;
1256         struct flowi6 fl6;
1257
1258         memset(&fl6, 0, sizeof(fl6));
1259         fl6.flowi6_oif = oif;
1260         fl6.flowi6_mark = mark;
1261         fl6.daddr = iph->daddr;
1262         fl6.saddr = iph->saddr;
1263         fl6.flowlabel = ip6_flowinfo(iph);
1264
1265         dst = ip6_route_redirect(net, &fl6, &ipv6_hdr(skb)->saddr);
1266         rt6_do_redirect(dst, NULL, skb);
1267         dst_release(dst);
1268 }
1269 EXPORT_SYMBOL_GPL(ip6_redirect);
1270
1271 void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif,
1272                             u32 mark)
1273 {
1274         const struct ipv6hdr *iph = ipv6_hdr(skb);
1275         const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
1276         struct dst_entry *dst;
1277         struct flowi6 fl6;
1278
1279         memset(&fl6, 0, sizeof(fl6));
1280         fl6.flowi6_oif = oif;
1281         fl6.flowi6_mark = mark;
1282         fl6.daddr = msg->dest;
1283         fl6.saddr = iph->daddr;
1284
1285         dst = ip6_route_redirect(net, &fl6, &iph->saddr);
1286         rt6_do_redirect(dst, NULL, skb);
1287         dst_release(dst);
1288 }
1289
1290 void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
1291 {
1292         ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark);
1293 }
1294 EXPORT_SYMBOL_GPL(ip6_sk_redirect);
1295
1296 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1297 {
1298         struct net_device *dev = dst->dev;
1299         unsigned int mtu = dst_mtu(dst);
1300         struct net *net = dev_net(dev);
1301
1302         mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
1303
1304         if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
1305                 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1306
1307         /*
1308          * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1309          * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1310          * IPV6_MAXPLEN is also valid and means: "any MSS,
1311          * rely only on pmtu discovery"
1312          */
1313         if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
1314                 mtu = IPV6_MAXPLEN;
1315         return mtu;
1316 }
1317
1318 static unsigned int ip6_mtu(const struct dst_entry *dst)
1319 {
1320         struct inet6_dev *idev;
1321         unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
1322
1323         if (mtu)
1324                 return mtu;
1325
1326         mtu = IPV6_MIN_MTU;
1327
1328         rcu_read_lock();
1329         idev = __in6_dev_get(dst->dev);
1330         if (idev)
1331                 mtu = idev->cnf.mtu6;
1332         rcu_read_unlock();
1333
1334         return mtu;
1335 }
1336
1337 static struct dst_entry *icmp6_dst_gc_list;
1338 static DEFINE_SPINLOCK(icmp6_dst_lock);
1339
1340 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
1341                                   struct flowi6 *fl6)
1342 {
1343         struct dst_entry *dst;
1344         struct rt6_info *rt;
1345         struct inet6_dev *idev = in6_dev_get(dev);
1346         struct net *net = dev_net(dev);
1347
1348         if (unlikely(!idev))
1349                 return ERR_PTR(-ENODEV);
1350
1351         rt = ip6_dst_alloc(net, dev, 0, NULL);
1352         if (unlikely(!rt)) {
1353                 in6_dev_put(idev);
1354                 dst = ERR_PTR(-ENOMEM);
1355                 goto out;
1356         }
1357
1358         rt->dst.flags |= DST_HOST;
1359         rt->dst.output  = ip6_output;
1360         atomic_set(&rt->dst.__refcnt, 1);
1361         rt->rt6i_gateway  = fl6->daddr;
1362         rt->rt6i_dst.addr = fl6->daddr;
1363         rt->rt6i_dst.plen = 128;
1364         rt->rt6i_idev     = idev;
1365         dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
1366
1367         spin_lock_bh(&icmp6_dst_lock);
1368         rt->dst.next = icmp6_dst_gc_list;
1369         icmp6_dst_gc_list = &rt->dst;
1370         spin_unlock_bh(&icmp6_dst_lock);
1371
1372         fib6_force_start_gc(net);
1373
1374         dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
1375
1376 out:
1377         return dst;
1378 }
1379
1380 int icmp6_dst_gc(void)
1381 {
1382         struct dst_entry *dst, **pprev;
1383         int more = 0;
1384
1385         spin_lock_bh(&icmp6_dst_lock);
1386         pprev = &icmp6_dst_gc_list;
1387
1388         while ((dst = *pprev) != NULL) {
1389                 if (!atomic_read(&dst->__refcnt)) {
1390                         *pprev = dst->next;
1391                         dst_free(dst);
1392                 } else {
1393                         pprev = &dst->next;
1394                         ++more;
1395                 }
1396         }
1397
1398         spin_unlock_bh(&icmp6_dst_lock);
1399
1400         return more;
1401 }
1402
1403 static void icmp6_clean_all(int (*func)(struct rt6_info *rt, void *arg),
1404                             void *arg)
1405 {
1406         struct dst_entry *dst, **pprev;
1407
1408         spin_lock_bh(&icmp6_dst_lock);
1409         pprev = &icmp6_dst_gc_list;
1410         while ((dst = *pprev) != NULL) {
1411                 struct rt6_info *rt = (struct rt6_info *) dst;
1412                 if (func(rt, arg)) {
1413                         *pprev = dst->next;
1414                         dst_free(dst);
1415                 } else {
1416                         pprev = &dst->next;
1417                 }
1418         }
1419         spin_unlock_bh(&icmp6_dst_lock);
1420 }
1421
1422 static int ip6_dst_gc(struct dst_ops *ops)
1423 {
1424         struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
1425         int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1426         int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1427         int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1428         int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1429         unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1430         int entries;
1431
1432         entries = dst_entries_get_fast(ops);
1433         if (time_after(rt_last_gc + rt_min_interval, jiffies) &&
1434             entries <= rt_max_size)
1435                 goto out;
1436
1437         net->ipv6.ip6_rt_gc_expire++;
1438         fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net, entries > rt_max_size);
1439         entries = dst_entries_get_slow(ops);
1440         if (entries < ops->gc_thresh)
1441                 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1442 out:
1443         net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1444         return entries > rt_max_size;
1445 }
1446
1447 /*
1448  *
1449  */
1450
1451 int ip6_route_add(struct fib6_config *cfg)
1452 {
1453         int err;
1454         struct net *net = cfg->fc_nlinfo.nl_net;
1455         struct rt6_info *rt = NULL;
1456         struct net_device *dev = NULL;
1457         struct inet6_dev *idev = NULL;
1458         struct fib6_table *table;
1459         int addr_type;
1460
1461         if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1462                 return -EINVAL;
1463 #ifndef CONFIG_IPV6_SUBTREES
1464         if (cfg->fc_src_len)
1465                 return -EINVAL;
1466 #endif
1467         if (cfg->fc_ifindex) {
1468                 err = -ENODEV;
1469                 dev = dev_get_by_index(net, cfg->fc_ifindex);
1470                 if (!dev)
1471                         goto out;
1472                 idev = in6_dev_get(dev);
1473                 if (!idev)
1474                         goto out;
1475         }
1476
1477         if (cfg->fc_metric == 0)
1478                 cfg->fc_metric = IP6_RT_PRIO_USER;
1479
1480         err = -ENOBUFS;
1481         if (cfg->fc_nlinfo.nlh &&
1482             !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
1483                 table = fib6_get_table(net, cfg->fc_table);
1484                 if (!table) {
1485                         pr_warn("NLM_F_CREATE should be specified when creating new route\n");
1486                         table = fib6_new_table(net, cfg->fc_table);
1487                 }
1488         } else {
1489                 table = fib6_new_table(net, cfg->fc_table);
1490         }
1491
1492         if (!table)
1493                 goto out;
1494
1495         rt = ip6_dst_alloc(net, NULL, DST_NOCOUNT, table);
1496
1497         if (!rt) {
1498                 err = -ENOMEM;
1499                 goto out;
1500         }
1501
1502         if (cfg->fc_flags & RTF_EXPIRES)
1503                 rt6_set_expires(rt, jiffies +
1504                                 clock_t_to_jiffies(cfg->fc_expires));
1505         else
1506                 rt6_clean_expires(rt);
1507
1508         if (cfg->fc_protocol == RTPROT_UNSPEC)
1509                 cfg->fc_protocol = RTPROT_BOOT;
1510         rt->rt6i_protocol = cfg->fc_protocol;
1511
1512         addr_type = ipv6_addr_type(&cfg->fc_dst);
1513
1514         if (addr_type & IPV6_ADDR_MULTICAST)
1515                 rt->dst.input = ip6_mc_input;
1516         else if (cfg->fc_flags & RTF_LOCAL)
1517                 rt->dst.input = ip6_input;
1518         else
1519                 rt->dst.input = ip6_forward;
1520
1521         rt->dst.output = ip6_output;
1522
1523         ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1524         rt->rt6i_dst.plen = cfg->fc_dst_len;
1525         if (rt->rt6i_dst.plen == 128)
1526                rt->dst.flags |= DST_HOST;
1527
1528         if (!(rt->dst.flags & DST_HOST) && cfg->fc_mx) {
1529                 u32 *metrics = kzalloc(sizeof(u32) * RTAX_MAX, GFP_KERNEL);
1530                 if (!metrics) {
1531                         err = -ENOMEM;
1532                         goto out;
1533                 }
1534                 dst_init_metrics(&rt->dst, metrics, 0);
1535         }
1536 #ifdef CONFIG_IPV6_SUBTREES
1537         ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1538         rt->rt6i_src.plen = cfg->fc_src_len;
1539 #endif
1540
1541         rt->rt6i_metric = cfg->fc_metric;
1542
1543         /* We cannot add true routes via loopback here,
1544            they would result in kernel looping; promote them to reject routes
1545          */
1546         if ((cfg->fc_flags & RTF_REJECT) ||
1547             (dev && (dev->flags & IFF_LOOPBACK) &&
1548              !(addr_type & IPV6_ADDR_LOOPBACK) &&
1549              !(cfg->fc_flags & RTF_LOCAL))) {
1550                 /* hold loopback dev/idev if we haven't done so. */
1551                 if (dev != net->loopback_dev) {
1552                         if (dev) {
1553                                 dev_put(dev);
1554                                 in6_dev_put(idev);
1555                         }
1556                         dev = net->loopback_dev;
1557                         dev_hold(dev);
1558                         idev = in6_dev_get(dev);
1559                         if (!idev) {
1560                                 err = -ENODEV;
1561                                 goto out;
1562                         }
1563                 }
1564                 rt->dst.output = ip6_pkt_discard_out;
1565                 rt->dst.input = ip6_pkt_discard;
1566                 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1567                 switch (cfg->fc_type) {
1568                 case RTN_BLACKHOLE:
1569                         rt->dst.error = -EINVAL;
1570                         break;
1571                 case RTN_PROHIBIT:
1572                         rt->dst.error = -EACCES;
1573                         break;
1574                 case RTN_THROW:
1575                         rt->dst.error = -EAGAIN;
1576                         break;
1577                 default:
1578                         rt->dst.error = -ENETUNREACH;
1579                         break;
1580                 }
1581                 goto install_route;
1582         }
1583
1584         if (cfg->fc_flags & RTF_GATEWAY) {
1585                 const struct in6_addr *gw_addr;
1586                 int gwa_type;
1587
1588                 gw_addr = &cfg->fc_gateway;
1589                 rt->rt6i_gateway = *gw_addr;
1590                 gwa_type = ipv6_addr_type(gw_addr);
1591
1592                 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1593                         struct rt6_info *grt;
1594
1595                         /* IPv6 strictly inhibits using not link-local
1596                            addresses as nexthop address.
1597                            Otherwise, router will not able to send redirects.
1598                            It is very good, but in some (rare!) circumstances
1599                            (SIT, PtP, NBMA NOARP links) it is handy to allow
1600                            some exceptions. --ANK
1601                          */
1602                         err = -EINVAL;
1603                         if (!(gwa_type & IPV6_ADDR_UNICAST))
1604                                 goto out;
1605
1606                         grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, 1);
1607
1608                         err = -EHOSTUNREACH;
1609                         if (!grt)
1610                                 goto out;
1611                         if (dev) {
1612                                 if (dev != grt->dst.dev) {
1613                                         ip6_rt_put(grt);
1614                                         goto out;
1615                                 }
1616                         } else {
1617                                 dev = grt->dst.dev;
1618                                 idev = grt->rt6i_idev;
1619                                 dev_hold(dev);
1620                                 in6_dev_hold(grt->rt6i_idev);
1621                         }
1622                         if (!(grt->rt6i_flags & RTF_GATEWAY))
1623                                 err = 0;
1624                         ip6_rt_put(grt);
1625
1626                         if (err)
1627                                 goto out;
1628                 }
1629                 err = -EINVAL;
1630                 if (!dev || (dev->flags & IFF_LOOPBACK))
1631                         goto out;
1632         }
1633
1634         err = -ENODEV;
1635         if (!dev)
1636                 goto out;
1637
1638         if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
1639                 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
1640                         err = -EINVAL;
1641                         goto out;
1642                 }
1643                 rt->rt6i_prefsrc.addr = cfg->fc_prefsrc;
1644                 rt->rt6i_prefsrc.plen = 128;
1645         } else
1646                 rt->rt6i_prefsrc.plen = 0;
1647
1648         rt->rt6i_flags = cfg->fc_flags;
1649
1650 install_route:
1651         if (cfg->fc_mx) {
1652                 struct nlattr *nla;
1653                 int remaining;
1654
1655                 nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1656                         int type = nla_type(nla);
1657
1658                         if (type) {
1659                                 if (type > RTAX_MAX) {
1660                                         err = -EINVAL;
1661                                         goto out;
1662                                 }
1663
1664                                 dst_metric_set(&rt->dst, type, nla_get_u32(nla));
1665                         }
1666                 }
1667         }
1668
1669         rt->dst.dev = dev;
1670         rt->rt6i_idev = idev;
1671         rt->rt6i_table = table;
1672
1673         cfg->fc_nlinfo.nl_net = dev_net(dev);
1674
1675         return __ip6_ins_rt(rt, &cfg->fc_nlinfo);
1676
1677 out:
1678         if (dev)
1679                 dev_put(dev);
1680         if (idev)
1681                 in6_dev_put(idev);
1682         if (rt)
1683                 dst_free(&rt->dst);
1684         return err;
1685 }
1686
1687 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
1688 {
1689         int err;
1690         struct fib6_table *table;
1691         struct net *net = dev_net(rt->dst.dev);
1692
1693         if (rt == net->ipv6.ip6_null_entry) {
1694                 err = -ENOENT;
1695                 goto out;
1696         }
1697
1698         table = rt->rt6i_table;
1699         write_lock_bh(&table->tb6_lock);
1700         err = fib6_del(rt, info);
1701         write_unlock_bh(&table->tb6_lock);
1702
1703 out:
1704         ip6_rt_put(rt);
1705         return err;
1706 }
1707
1708 int ip6_del_rt(struct rt6_info *rt)
1709 {
1710         struct nl_info info = {
1711                 .nl_net = dev_net(rt->dst.dev),
1712         };
1713         return __ip6_del_rt(rt, &info);
1714 }
1715
1716 static int ip6_route_del(struct fib6_config *cfg)
1717 {
1718         struct fib6_table *table;
1719         struct fib6_node *fn;
1720         struct rt6_info *rt;
1721         int err = -ESRCH;
1722
1723         table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
1724         if (!table)
1725                 return err;
1726
1727         read_lock_bh(&table->tb6_lock);
1728
1729         fn = fib6_locate(&table->tb6_root,
1730                          &cfg->fc_dst, cfg->fc_dst_len,
1731                          &cfg->fc_src, cfg->fc_src_len);
1732
1733         if (fn) {
1734                 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1735                         if (cfg->fc_ifindex &&
1736                             (!rt->dst.dev ||
1737                              rt->dst.dev->ifindex != cfg->fc_ifindex))
1738                                 continue;
1739                         if (cfg->fc_flags & RTF_GATEWAY &&
1740                             !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
1741                                 continue;
1742                         if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
1743                                 continue;
1744                         dst_hold(&rt->dst);
1745                         read_unlock_bh(&table->tb6_lock);
1746
1747                         return __ip6_del_rt(rt, &cfg->fc_nlinfo);
1748                 }
1749         }
1750         read_unlock_bh(&table->tb6_lock);
1751
1752         return err;
1753 }
1754
1755 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
1756 {
1757         struct net *net = dev_net(skb->dev);
1758         struct netevent_redirect netevent;
1759         struct rt6_info *rt, *nrt = NULL;
1760         struct ndisc_options ndopts;
1761         struct inet6_dev *in6_dev;
1762         struct neighbour *neigh;
1763         struct rd_msg *msg;
1764         int optlen, on_link;
1765         u8 *lladdr;
1766
1767         optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
1768         optlen -= sizeof(*msg);
1769
1770         if (optlen < 0) {
1771                 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
1772                 return;
1773         }
1774
1775         msg = (struct rd_msg *)icmp6_hdr(skb);
1776
1777         if (ipv6_addr_is_multicast(&msg->dest)) {
1778                 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
1779                 return;
1780         }
1781
1782         on_link = 0;
1783         if (ipv6_addr_equal(&msg->dest, &msg->target)) {
1784                 on_link = 1;
1785         } else if (ipv6_addr_type(&msg->target) !=
1786                    (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
1787                 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
1788                 return;
1789         }
1790
1791         in6_dev = __in6_dev_get(skb->dev);
1792         if (!in6_dev)
1793                 return;
1794         if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
1795                 return;
1796
1797         /* RFC2461 8.1:
1798          *      The IP source address of the Redirect MUST be the same as the current
1799          *      first-hop router for the specified ICMP Destination Address.
1800          */
1801
1802         if (!ndisc_parse_options(msg->opt, optlen, &ndopts)) {
1803                 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
1804                 return;
1805         }
1806
1807         lladdr = NULL;
1808         if (ndopts.nd_opts_tgt_lladdr) {
1809                 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
1810                                              skb->dev);
1811                 if (!lladdr) {
1812                         net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
1813                         return;
1814                 }
1815         }
1816
1817         rt = (struct rt6_info *) dst;
1818         if (rt == net->ipv6.ip6_null_entry) {
1819                 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
1820                 return;
1821         }
1822
1823         /* Redirect received -> path was valid.
1824          * Look, redirects are sent only in response to data packets,
1825          * so that this nexthop apparently is reachable. --ANK
1826          */
1827         dst_confirm(&rt->dst);
1828
1829         neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
1830         if (!neigh)
1831                 return;
1832
1833         /*
1834          *      We have finally decided to accept it.
1835          */
1836
1837         neigh_update(neigh, lladdr, NUD_STALE,
1838                      NEIGH_UPDATE_F_WEAK_OVERRIDE|
1839                      NEIGH_UPDATE_F_OVERRIDE|
1840                      (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1841                                      NEIGH_UPDATE_F_ISROUTER))
1842                      );
1843
1844         nrt = ip6_rt_copy(rt, &msg->dest);
1845         if (!nrt)
1846                 goto out;
1847
1848         nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
1849         if (on_link)
1850                 nrt->rt6i_flags &= ~RTF_GATEWAY;
1851
1852         nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
1853
1854         if (ip6_ins_rt(nrt))
1855                 goto out;
1856
1857         netevent.old = &rt->dst;
1858         netevent.new = &nrt->dst;
1859         netevent.daddr = &msg->dest;
1860         netevent.neigh = neigh;
1861         call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
1862
1863         if (rt->rt6i_flags & RTF_CACHE) {
1864                 rt = (struct rt6_info *) dst_clone(&rt->dst);
1865                 ip6_del_rt(rt);
1866         }
1867
1868 out:
1869         neigh_release(neigh);
1870 }
1871
1872 /*
1873  *      Misc support functions
1874  */
1875
1876 static struct rt6_info *ip6_rt_copy(struct rt6_info *ort,
1877                                     const struct in6_addr *dest)
1878 {
1879         struct net *net = dev_net(ort->dst.dev);
1880         struct rt6_info *rt = ip6_dst_alloc(net, ort->dst.dev, 0,
1881                                             ort->rt6i_table);
1882
1883         if (rt) {
1884                 rt->dst.input = ort->dst.input;
1885                 rt->dst.output = ort->dst.output;
1886                 rt->dst.flags |= DST_HOST;
1887
1888                 rt->rt6i_dst.addr = *dest;
1889                 rt->rt6i_dst.plen = 128;
1890                 dst_copy_metrics(&rt->dst, &ort->dst);
1891                 rt->dst.error = ort->dst.error;
1892                 rt->rt6i_idev = ort->rt6i_idev;
1893                 if (rt->rt6i_idev)
1894                         in6_dev_hold(rt->rt6i_idev);
1895                 rt->dst.lastuse = jiffies;
1896
1897                 if (ort->rt6i_flags & RTF_GATEWAY)
1898                         rt->rt6i_gateway = ort->rt6i_gateway;
1899                 else
1900                         rt->rt6i_gateway = *dest;
1901                 rt->rt6i_flags = ort->rt6i_flags;
1902                 if ((ort->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF)) ==
1903                     (RTF_DEFAULT | RTF_ADDRCONF))
1904                         rt6_set_from(rt, ort);
1905                 rt->rt6i_metric = 0;
1906
1907 #ifdef CONFIG_IPV6_SUBTREES
1908                 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1909 #endif
1910                 memcpy(&rt->rt6i_prefsrc, &ort->rt6i_prefsrc, sizeof(struct rt6key));
1911                 rt->rt6i_table = ort->rt6i_table;
1912         }
1913         return rt;
1914 }
1915
1916 #ifdef CONFIG_IPV6_ROUTE_INFO
1917 static struct rt6_info *rt6_get_route_info(struct net *net,
1918                                            const struct in6_addr *prefix, int prefixlen,
1919                                            const struct in6_addr *gwaddr, int ifindex)
1920 {
1921         struct fib6_node *fn;
1922         struct rt6_info *rt = NULL;
1923         struct fib6_table *table;
1924
1925         table = fib6_get_table(net, RT6_TABLE_INFO);
1926         if (!table)
1927                 return NULL;
1928
1929         read_lock_bh(&table->tb6_lock);
1930         fn = fib6_locate(&table->tb6_root, prefix ,prefixlen, NULL, 0);
1931         if (!fn)
1932                 goto out;
1933
1934         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1935                 if (rt->dst.dev->ifindex != ifindex)
1936                         continue;
1937                 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
1938                         continue;
1939                 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
1940                         continue;
1941                 dst_hold(&rt->dst);
1942                 break;
1943         }
1944 out:
1945         read_unlock_bh(&table->tb6_lock);
1946         return rt;
1947 }
1948
1949 static struct rt6_info *rt6_add_route_info(struct net *net,
1950                                            const struct in6_addr *prefix, int prefixlen,
1951                                            const struct in6_addr *gwaddr, int ifindex,
1952                                            unsigned int pref)
1953 {
1954         struct fib6_config cfg = {
1955                 .fc_table       = RT6_TABLE_INFO,
1956                 .fc_metric      = IP6_RT_PRIO_USER,
1957                 .fc_ifindex     = ifindex,
1958                 .fc_dst_len     = prefixlen,
1959                 .fc_flags       = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
1960                                   RTF_UP | RTF_PREF(pref),
1961                 .fc_nlinfo.portid = 0,
1962                 .fc_nlinfo.nlh = NULL,
1963                 .fc_nlinfo.nl_net = net,
1964         };
1965
1966         cfg.fc_dst = *prefix;
1967         cfg.fc_gateway = *gwaddr;
1968
1969         /* We should treat it as a default route if prefix length is 0. */
1970         if (!prefixlen)
1971                 cfg.fc_flags |= RTF_DEFAULT;
1972
1973         ip6_route_add(&cfg);
1974
1975         return rt6_get_route_info(net, prefix, prefixlen, gwaddr, ifindex);
1976 }
1977 #endif
1978
1979 struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
1980 {
1981         struct rt6_info *rt;
1982         struct fib6_table *table;
1983
1984         table = fib6_get_table(dev_net(dev), RT6_TABLE_DFLT);
1985         if (!table)
1986                 return NULL;
1987
1988         read_lock_bh(&table->tb6_lock);
1989         for (rt = table->tb6_root.leaf; rt; rt=rt->dst.rt6_next) {
1990                 if (dev == rt->dst.dev &&
1991                     ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
1992                     ipv6_addr_equal(&rt->rt6i_gateway, addr))
1993                         break;
1994         }
1995         if (rt)
1996                 dst_hold(&rt->dst);
1997         read_unlock_bh(&table->tb6_lock);
1998         return rt;
1999 }
2000
2001 struct rt6_info *rt6_add_dflt_router(const struct in6_addr *gwaddr,
2002                                      struct net_device *dev,
2003                                      unsigned int pref)
2004 {
2005         struct fib6_config cfg = {
2006                 .fc_table       = RT6_TABLE_DFLT,
2007                 .fc_metric      = IP6_RT_PRIO_USER,
2008                 .fc_ifindex     = dev->ifindex,
2009                 .fc_flags       = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
2010                                   RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
2011                 .fc_nlinfo.portid = 0,
2012                 .fc_nlinfo.nlh = NULL,
2013                 .fc_nlinfo.nl_net = dev_net(dev),
2014         };
2015
2016         cfg.fc_gateway = *gwaddr;
2017
2018         ip6_route_add(&cfg);
2019
2020         return rt6_get_dflt_router(gwaddr, dev);
2021 }
2022
2023 void rt6_purge_dflt_routers(struct net *net)
2024 {
2025         struct rt6_info *rt;
2026         struct fib6_table *table;
2027
2028         /* NOTE: Keep consistent with rt6_get_dflt_router */
2029         table = fib6_get_table(net, RT6_TABLE_DFLT);
2030         if (!table)
2031                 return;
2032
2033 restart:
2034         read_lock_bh(&table->tb6_lock);
2035         for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2036                 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
2037                     (!rt->rt6i_idev || rt->rt6i_idev->cnf.accept_ra != 2)) {
2038                         dst_hold(&rt->dst);
2039                         read_unlock_bh(&table->tb6_lock);
2040                         ip6_del_rt(rt);
2041                         goto restart;
2042                 }
2043         }
2044         read_unlock_bh(&table->tb6_lock);
2045 }
2046
2047 static void rtmsg_to_fib6_config(struct net *net,
2048                                  struct in6_rtmsg *rtmsg,
2049                                  struct fib6_config *cfg)
2050 {
2051         memset(cfg, 0, sizeof(*cfg));
2052
2053         cfg->fc_table = RT6_TABLE_MAIN;
2054         cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
2055         cfg->fc_metric = rtmsg->rtmsg_metric;
2056         cfg->fc_expires = rtmsg->rtmsg_info;
2057         cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
2058         cfg->fc_src_len = rtmsg->rtmsg_src_len;
2059         cfg->fc_flags = rtmsg->rtmsg_flags;
2060
2061         cfg->fc_nlinfo.nl_net = net;
2062
2063         cfg->fc_dst = rtmsg->rtmsg_dst;
2064         cfg->fc_src = rtmsg->rtmsg_src;
2065         cfg->fc_gateway = rtmsg->rtmsg_gateway;
2066 }
2067
2068 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
2069 {
2070         struct fib6_config cfg;
2071         struct in6_rtmsg rtmsg;
2072         int err;
2073
2074         switch(cmd) {
2075         case SIOCADDRT:         /* Add a route */
2076         case SIOCDELRT:         /* Delete a route */
2077                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2078                         return -EPERM;
2079                 err = copy_from_user(&rtmsg, arg,
2080                                      sizeof(struct in6_rtmsg));
2081                 if (err)
2082                         return -EFAULT;
2083
2084                 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
2085
2086                 rtnl_lock();
2087                 switch (cmd) {
2088                 case SIOCADDRT:
2089                         err = ip6_route_add(&cfg);
2090                         break;
2091                 case SIOCDELRT:
2092                         err = ip6_route_del(&cfg);
2093                         break;
2094                 default:
2095                         err = -EINVAL;
2096                 }
2097                 rtnl_unlock();
2098
2099                 return err;
2100         }
2101
2102         return -EINVAL;
2103 }
2104
2105 /*
2106  *      Drop the packet on the floor
2107  */
2108
2109 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
2110 {
2111         int type;
2112         struct dst_entry *dst = skb_dst(skb);
2113         switch (ipstats_mib_noroutes) {
2114         case IPSTATS_MIB_INNOROUTES:
2115                 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
2116                 if (type == IPV6_ADDR_ANY) {
2117                         IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2118                                       IPSTATS_MIB_INADDRERRORS);
2119                         break;
2120                 }
2121                 /* FALLTHROUGH */
2122         case IPSTATS_MIB_OUTNOROUTES:
2123                 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2124                               ipstats_mib_noroutes);
2125                 break;
2126         }
2127         icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
2128         kfree_skb(skb);
2129         return 0;
2130 }
2131
2132 static int ip6_pkt_discard(struct sk_buff *skb)
2133 {
2134         return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
2135 }
2136
2137 static int ip6_pkt_discard_out(struct sk_buff *skb)
2138 {
2139         skb->dev = skb_dst(skb)->dev;
2140         return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
2141 }
2142
2143 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2144
2145 static int ip6_pkt_prohibit(struct sk_buff *skb)
2146 {
2147         return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
2148 }
2149
2150 static int ip6_pkt_prohibit_out(struct sk_buff *skb)
2151 {
2152         skb->dev = skb_dst(skb)->dev;
2153         return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
2154 }
2155
2156 #endif
2157
2158 /*
2159  *      Allocate a dst for local (unicast / anycast) address.
2160  */
2161
2162 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
2163                                     const struct in6_addr *addr,
2164                                     bool anycast)
2165 {
2166         struct net *net = dev_net(idev->dev);
2167         struct rt6_info *rt = ip6_dst_alloc(net, net->loopback_dev, 0, NULL);
2168
2169         if (!rt) {
2170                 net_warn_ratelimited("Maximum number of routes reached, consider increasing route/max_size\n");
2171                 return ERR_PTR(-ENOMEM);
2172         }
2173
2174         in6_dev_hold(idev);
2175
2176         rt->dst.flags |= DST_HOST;
2177         rt->dst.input = ip6_input;
2178         rt->dst.output = ip6_output;
2179         rt->rt6i_idev = idev;
2180
2181         rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
2182         if (anycast)
2183                 rt->rt6i_flags |= RTF_ANYCAST;
2184         else
2185                 rt->rt6i_flags |= RTF_LOCAL;
2186
2187         rt->rt6i_gateway  = *addr;
2188         rt->rt6i_dst.addr = *addr;
2189         rt->rt6i_dst.plen = 128;
2190         rt->rt6i_table = fib6_get_table(net, RT6_TABLE_LOCAL);
2191
2192         atomic_set(&rt->dst.__refcnt, 1);
2193
2194         return rt;
2195 }
2196
2197 int ip6_route_get_saddr(struct net *net,
2198                         struct rt6_info *rt,
2199                         const struct in6_addr *daddr,
2200                         unsigned int prefs,
2201                         struct in6_addr *saddr)
2202 {
2203         struct inet6_dev *idev = ip6_dst_idev((struct dst_entry*)rt);
2204         int err = 0;
2205         if (rt->rt6i_prefsrc.plen)
2206                 *saddr = rt->rt6i_prefsrc.addr;
2207         else
2208                 err = ipv6_dev_get_saddr(net, idev ? idev->dev : NULL,
2209                                          daddr, prefs, saddr);
2210         return err;
2211 }
2212
2213 /* remove deleted ip from prefsrc entries */
2214 struct arg_dev_net_ip {
2215         struct net_device *dev;
2216         struct net *net;
2217         struct in6_addr *addr;
2218 };
2219
2220 static int fib6_remove_prefsrc(struct rt6_info *rt, void *arg)
2221 {
2222         struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
2223         struct net *net = ((struct arg_dev_net_ip *)arg)->net;
2224         struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
2225
2226         if (((void *)rt->dst.dev == dev || !dev) &&
2227             rt != net->ipv6.ip6_null_entry &&
2228             ipv6_addr_equal(addr, &rt->rt6i_prefsrc.addr)) {
2229                 /* remove prefsrc entry */
2230                 rt->rt6i_prefsrc.plen = 0;
2231         }
2232         return 0;
2233 }
2234
2235 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
2236 {
2237         struct net *net = dev_net(ifp->idev->dev);
2238         struct arg_dev_net_ip adni = {
2239                 .dev = ifp->idev->dev,
2240                 .net = net,
2241                 .addr = &ifp->addr,
2242         };
2243         fib6_clean_all(net, fib6_remove_prefsrc, 0, &adni);
2244 }
2245
2246 struct arg_dev_net {
2247         struct net_device *dev;
2248         struct net *net;
2249 };
2250
2251 static int fib6_ifdown(struct rt6_info *rt, void *arg)
2252 {
2253         const struct arg_dev_net *adn = arg;
2254         const struct net_device *dev = adn->dev;
2255
2256         if ((rt->dst.dev == dev || !dev) &&
2257             rt != adn->net->ipv6.ip6_null_entry)
2258                 return -1;
2259
2260         return 0;
2261 }
2262
2263 void rt6_ifdown(struct net *net, struct net_device *dev)
2264 {
2265         struct arg_dev_net adn = {
2266                 .dev = dev,
2267                 .net = net,
2268         };
2269
2270         fib6_clean_all(net, fib6_ifdown, 0, &adn);
2271         icmp6_clean_all(fib6_ifdown, &adn);
2272 }
2273
2274 struct rt6_mtu_change_arg {
2275         struct net_device *dev;
2276         unsigned int mtu;
2277 };
2278
2279 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2280 {
2281         struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2282         struct inet6_dev *idev;
2283
2284         /* In IPv6 pmtu discovery is not optional,
2285            so that RTAX_MTU lock cannot disable it.
2286            We still use this lock to block changes
2287            caused by addrconf/ndisc.
2288         */
2289
2290         idev = __in6_dev_get(arg->dev);
2291         if (!idev)
2292                 return 0;
2293
2294         /* For administrative MTU increase, there is no way to discover
2295            IPv6 PMTU increase, so PMTU increase should be updated here.
2296            Since RFC 1981 doesn't include administrative MTU increase
2297            update PMTU increase is a MUST. (i.e. jumbo frame)
2298          */
2299         /*
2300            If new MTU is less than route PMTU, this new MTU will be the
2301            lowest MTU in the path, update the route PMTU to reflect PMTU
2302            decreases; if new MTU is greater than route PMTU, and the
2303            old MTU is the lowest MTU in the path, update the route PMTU
2304            to reflect the increase. In this case if the other nodes' MTU
2305            also have the lowest MTU, TOO BIG MESSAGE will be lead to
2306            PMTU discouvery.
2307          */
2308         if (rt->dst.dev == arg->dev &&
2309             !dst_metric_locked(&rt->dst, RTAX_MTU) &&
2310             (dst_mtu(&rt->dst) >= arg->mtu ||
2311              (dst_mtu(&rt->dst) < arg->mtu &&
2312               dst_mtu(&rt->dst) == idev->cnf.mtu6))) {
2313                 dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
2314         }
2315         return 0;
2316 }
2317
2318 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
2319 {
2320         struct rt6_mtu_change_arg arg = {
2321                 .dev = dev,
2322                 .mtu = mtu,
2323         };
2324
2325         fib6_clean_all(dev_net(dev), rt6_mtu_change_route, 0, &arg);
2326 }
2327
2328 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2329         [RTA_GATEWAY]           = { .len = sizeof(struct in6_addr) },
2330         [RTA_OIF]               = { .type = NLA_U32 },
2331         [RTA_IIF]               = { .type = NLA_U32 },
2332         [RTA_PRIORITY]          = { .type = NLA_U32 },
2333         [RTA_METRICS]           = { .type = NLA_NESTED },
2334         [RTA_MULTIPATH]         = { .len = sizeof(struct rtnexthop) },
2335 };
2336
2337 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2338                               struct fib6_config *cfg)
2339 {
2340         struct rtmsg *rtm;
2341         struct nlattr *tb[RTA_MAX+1];
2342         int err;
2343
2344         err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2345         if (err < 0)
2346                 goto errout;
2347
2348         err = -EINVAL;
2349         rtm = nlmsg_data(nlh);
2350         memset(cfg, 0, sizeof(*cfg));
2351
2352         cfg->fc_table = rtm->rtm_table;
2353         cfg->fc_dst_len = rtm->rtm_dst_len;
2354         cfg->fc_src_len = rtm->rtm_src_len;
2355         cfg->fc_flags = RTF_UP;
2356         cfg->fc_protocol = rtm->rtm_protocol;
2357         cfg->fc_type = rtm->rtm_type;
2358
2359         if (rtm->rtm_type == RTN_UNREACHABLE ||
2360             rtm->rtm_type == RTN_BLACKHOLE ||
2361             rtm->rtm_type == RTN_PROHIBIT ||
2362             rtm->rtm_type == RTN_THROW)
2363                 cfg->fc_flags |= RTF_REJECT;
2364
2365         if (rtm->rtm_type == RTN_LOCAL)
2366                 cfg->fc_flags |= RTF_LOCAL;
2367
2368         cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid;
2369         cfg->fc_nlinfo.nlh = nlh;
2370         cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2371
2372         if (tb[RTA_GATEWAY]) {
2373                 nla_memcpy(&cfg->fc_gateway, tb[RTA_GATEWAY], 16);
2374                 cfg->fc_flags |= RTF_GATEWAY;
2375         }
2376
2377         if (tb[RTA_DST]) {
2378                 int plen = (rtm->rtm_dst_len + 7) >> 3;
2379
2380                 if (nla_len(tb[RTA_DST]) < plen)
2381                         goto errout;
2382
2383                 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2384         }
2385
2386         if (tb[RTA_SRC]) {
2387                 int plen = (rtm->rtm_src_len + 7) >> 3;
2388
2389                 if (nla_len(tb[RTA_SRC]) < plen)
2390                         goto errout;
2391
2392                 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2393         }
2394
2395         if (tb[RTA_PREFSRC])
2396                 nla_memcpy(&cfg->fc_prefsrc, tb[RTA_PREFSRC], 16);
2397
2398         if (tb[RTA_OIF])
2399                 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2400
2401         if (tb[RTA_PRIORITY])
2402                 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2403
2404         if (tb[RTA_METRICS]) {
2405                 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2406                 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2407         }
2408
2409         if (tb[RTA_TABLE])
2410                 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2411
2412         if (tb[RTA_MULTIPATH]) {
2413                 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
2414                 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
2415         }
2416
2417         err = 0;
2418 errout:
2419         return err;
2420 }
2421
2422 static int ip6_route_multipath(struct fib6_config *cfg, int add)
2423 {
2424         struct fib6_config r_cfg;
2425         struct rtnexthop *rtnh;
2426         int remaining;
2427         int attrlen;
2428         int err = 0, last_err = 0;
2429
2430 beginning:
2431         rtnh = (struct rtnexthop *)cfg->fc_mp;
2432         remaining = cfg->fc_mp_len;
2433
2434         /* Parse a Multipath Entry */
2435         while (rtnh_ok(rtnh, remaining)) {
2436                 memcpy(&r_cfg, cfg, sizeof(*cfg));
2437                 if (rtnh->rtnh_ifindex)
2438                         r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
2439
2440                 attrlen = rtnh_attrlen(rtnh);
2441                 if (attrlen > 0) {
2442                         struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
2443
2444                         nla = nla_find(attrs, attrlen, RTA_GATEWAY);
2445                         if (nla) {
2446                                 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
2447                                 r_cfg.fc_flags |= RTF_GATEWAY;
2448                         }
2449                 }
2450                 err = add ? ip6_route_add(&r_cfg) : ip6_route_del(&r_cfg);
2451                 if (err) {
2452                         last_err = err;
2453                         /* If we are trying to remove a route, do not stop the
2454                          * loop when ip6_route_del() fails (because next hop is
2455                          * already gone), we should try to remove all next hops.
2456                          */
2457                         if (add) {
2458                                 /* If add fails, we should try to delete all
2459                                  * next hops that have been already added.
2460                                  */
2461                                 add = 0;
2462                                 goto beginning;
2463                         }
2464                 }
2465                 /* Because each route is added like a single route we remove
2466                  * this flag after the first nexthop (if there is a collision,
2467                  * we have already fail to add the first nexthop:
2468                  * fib6_add_rt2node() has reject it).
2469                  */
2470                 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~NLM_F_EXCL;
2471                 rtnh = rtnh_next(rtnh, &remaining);
2472         }
2473
2474         return last_err;
2475 }
2476
2477 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr* nlh)
2478 {
2479         struct fib6_config cfg;
2480         int err;
2481
2482         err = rtm_to_fib6_config(skb, nlh, &cfg);
2483         if (err < 0)
2484                 return err;
2485
2486         if (cfg.fc_mp)
2487                 return ip6_route_multipath(&cfg, 0);
2488         else
2489                 return ip6_route_del(&cfg);
2490 }
2491
2492 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr* nlh)
2493 {
2494         struct fib6_config cfg;
2495         int err;
2496
2497         err = rtm_to_fib6_config(skb, nlh, &cfg);
2498         if (err < 0)
2499                 return err;
2500
2501         if (cfg.fc_mp)
2502                 return ip6_route_multipath(&cfg, 1);
2503         else
2504                 return ip6_route_add(&cfg);
2505 }
2506
2507 static inline size_t rt6_nlmsg_size(void)
2508 {
2509         return NLMSG_ALIGN(sizeof(struct rtmsg))
2510                + nla_total_size(16) /* RTA_SRC */
2511                + nla_total_size(16) /* RTA_DST */
2512                + nla_total_size(16) /* RTA_GATEWAY */
2513                + nla_total_size(16) /* RTA_PREFSRC */
2514                + nla_total_size(4) /* RTA_TABLE */
2515                + nla_total_size(4) /* RTA_IIF */
2516                + nla_total_size(4) /* RTA_OIF */
2517                + nla_total_size(4) /* RTA_PRIORITY */
2518                + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
2519                + nla_total_size(sizeof(struct rta_cacheinfo));
2520 }
2521
2522 static int rt6_fill_node(struct net *net,
2523                          struct sk_buff *skb, struct rt6_info *rt,
2524                          struct in6_addr *dst, struct in6_addr *src,
2525                          int iif, int type, u32 portid, u32 seq,
2526                          int prefix, int nowait, unsigned int flags)
2527 {
2528         struct rtmsg *rtm;
2529         struct nlmsghdr *nlh;
2530         long expires;
2531         u32 table;
2532
2533         if (prefix) {   /* user wants prefix routes only */
2534                 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
2535                         /* success since this is not a prefix route */
2536                         return 1;
2537                 }
2538         }
2539
2540         nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
2541         if (!nlh)
2542                 return -EMSGSIZE;
2543
2544         rtm = nlmsg_data(nlh);
2545         rtm->rtm_family = AF_INET6;
2546         rtm->rtm_dst_len = rt->rt6i_dst.plen;
2547         rtm->rtm_src_len = rt->rt6i_src.plen;
2548         rtm->rtm_tos = 0;
2549         if (rt->rt6i_table)
2550                 table = rt->rt6i_table->tb6_id;
2551         else
2552                 table = RT6_TABLE_UNSPEC;
2553         rtm->rtm_table = table;
2554         if (nla_put_u32(skb, RTA_TABLE, table))
2555                 goto nla_put_failure;
2556         if (rt->rt6i_flags & RTF_REJECT) {
2557                 switch (rt->dst.error) {
2558                 case -EINVAL:
2559                         rtm->rtm_type = RTN_BLACKHOLE;
2560                         break;
2561                 case -EACCES:
2562                         rtm->rtm_type = RTN_PROHIBIT;
2563                         break;
2564                 case -EAGAIN:
2565                         rtm->rtm_type = RTN_THROW;
2566                         break;
2567                 default:
2568                         rtm->rtm_type = RTN_UNREACHABLE;
2569                         break;
2570                 }
2571         }
2572         else if (rt->rt6i_flags & RTF_LOCAL)
2573                 rtm->rtm_type = RTN_LOCAL;
2574         else if (rt->dst.dev && (rt->dst.dev->flags & IFF_LOOPBACK))
2575                 rtm->rtm_type = RTN_LOCAL;
2576         else
2577                 rtm->rtm_type = RTN_UNICAST;
2578         rtm->rtm_flags = 0;
2579         rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2580         rtm->rtm_protocol = rt->rt6i_protocol;
2581         if (rt->rt6i_flags & RTF_DYNAMIC)
2582                 rtm->rtm_protocol = RTPROT_REDIRECT;
2583         else if (rt->rt6i_flags & RTF_ADDRCONF) {
2584                 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ROUTEINFO))
2585                         rtm->rtm_protocol = RTPROT_RA;
2586                 else
2587                         rtm->rtm_protocol = RTPROT_KERNEL;
2588         }
2589
2590         if (rt->rt6i_flags & RTF_CACHE)
2591                 rtm->rtm_flags |= RTM_F_CLONED;
2592
2593         if (dst) {
2594                 if (nla_put(skb, RTA_DST, 16, dst))
2595                         goto nla_put_failure;
2596                 rtm->rtm_dst_len = 128;
2597         } else if (rtm->rtm_dst_len)
2598                 if (nla_put(skb, RTA_DST, 16, &rt->rt6i_dst.addr))
2599                         goto nla_put_failure;
2600 #ifdef CONFIG_IPV6_SUBTREES
2601         if (src) {
2602                 if (nla_put(skb, RTA_SRC, 16, src))
2603                         goto nla_put_failure;
2604                 rtm->rtm_src_len = 128;
2605         } else if (rtm->rtm_src_len &&
2606                    nla_put(skb, RTA_SRC, 16, &rt->rt6i_src.addr))
2607                 goto nla_put_failure;
2608 #endif
2609         if (iif) {
2610 #ifdef CONFIG_IPV6_MROUTE
2611                 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
2612                         int err = ip6mr_get_route(net, skb, rtm, nowait);
2613                         if (err <= 0) {
2614                                 if (!nowait) {
2615                                         if (err == 0)
2616                                                 return 0;
2617                                         goto nla_put_failure;
2618                                 } else {
2619                                         if (err == -EMSGSIZE)
2620                                                 goto nla_put_failure;
2621                                 }
2622                         }
2623                 } else
2624 #endif
2625                         if (nla_put_u32(skb, RTA_IIF, iif))
2626                                 goto nla_put_failure;
2627         } else if (dst) {
2628                 struct in6_addr saddr_buf;
2629                 if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0 &&
2630                     nla_put(skb, RTA_PREFSRC, 16, &saddr_buf))
2631                         goto nla_put_failure;
2632         }
2633
2634         if (rt->rt6i_prefsrc.plen) {
2635                 struct in6_addr saddr_buf;
2636                 saddr_buf = rt->rt6i_prefsrc.addr;
2637                 if (nla_put(skb, RTA_PREFSRC, 16, &saddr_buf))
2638                         goto nla_put_failure;
2639         }
2640
2641         if (rtnetlink_put_metrics(skb, dst_metrics_ptr(&rt->dst)) < 0)
2642                 goto nla_put_failure;
2643
2644         if (rt->rt6i_flags & RTF_GATEWAY) {
2645                 if (nla_put(skb, RTA_GATEWAY, 16, &rt->rt6i_gateway) < 0)
2646                         goto nla_put_failure;
2647         }
2648
2649         if (rt->dst.dev &&
2650             nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex))
2651                 goto nla_put_failure;
2652         if (nla_put_u32(skb, RTA_PRIORITY, rt->rt6i_metric))
2653                 goto nla_put_failure;
2654
2655         expires = (rt->rt6i_flags & RTF_EXPIRES) ? rt->dst.expires - jiffies : 0;
2656
2657         if (rtnl_put_cacheinfo(skb, &rt->dst, 0, expires, rt->dst.error) < 0)
2658                 goto nla_put_failure;
2659
2660         return nlmsg_end(skb, nlh);
2661
2662 nla_put_failure:
2663         nlmsg_cancel(skb, nlh);
2664         return -EMSGSIZE;
2665 }
2666
2667 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
2668 {
2669         struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
2670         int prefix;
2671
2672         if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
2673                 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
2674                 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
2675         } else
2676                 prefix = 0;
2677
2678         return rt6_fill_node(arg->net,
2679                      arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
2680                      NETLINK_CB(arg->cb->skb).portid, arg->cb->nlh->nlmsg_seq,
2681                      prefix, 0, NLM_F_MULTI);
2682 }
2683
2684 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr* nlh)
2685 {
2686         struct net *net = sock_net(in_skb->sk);
2687         struct nlattr *tb[RTA_MAX+1];
2688         struct rt6_info *rt;
2689         struct sk_buff *skb;
2690         struct rtmsg *rtm;
2691         struct flowi6 fl6;
2692         int err, iif = 0, oif = 0;
2693
2694         err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2695         if (err < 0)
2696                 goto errout;
2697
2698         err = -EINVAL;
2699         memset(&fl6, 0, sizeof(fl6));
2700
2701         if (tb[RTA_SRC]) {
2702                 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
2703                         goto errout;
2704
2705                 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
2706         }
2707
2708         if (tb[RTA_DST]) {
2709                 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
2710                         goto errout;
2711
2712                 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
2713         }
2714
2715         if (tb[RTA_IIF])
2716                 iif = nla_get_u32(tb[RTA_IIF]);
2717
2718         if (tb[RTA_OIF])
2719                 oif = nla_get_u32(tb[RTA_OIF]);
2720
2721         if (iif) {
2722                 struct net_device *dev;
2723                 int flags = 0;
2724
2725                 dev = __dev_get_by_index(net, iif);
2726                 if (!dev) {
2727                         err = -ENODEV;
2728                         goto errout;
2729                 }
2730
2731                 fl6.flowi6_iif = iif;
2732
2733                 if (!ipv6_addr_any(&fl6.saddr))
2734                         flags |= RT6_LOOKUP_F_HAS_SADDR;
2735
2736                 rt = (struct rt6_info *)ip6_route_input_lookup(net, dev, &fl6,
2737                                                                flags);
2738         } else {
2739                 fl6.flowi6_oif = oif;
2740
2741                 rt = (struct rt6_info *)ip6_route_output(net, NULL, &fl6);
2742         }
2743
2744         skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
2745         if (!skb) {
2746                 ip6_rt_put(rt);
2747                 err = -ENOBUFS;
2748                 goto errout;
2749         }
2750
2751         /* Reserve room for dummy headers, this skb can pass
2752            through good chunk of routing engine.
2753          */
2754         skb_reset_mac_header(skb);
2755         skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
2756
2757         skb_dst_set(skb, &rt->dst);
2758
2759         err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
2760                             RTM_NEWROUTE, NETLINK_CB(in_skb).portid,
2761                             nlh->nlmsg_seq, 0, 0, 0);
2762         if (err < 0) {
2763                 kfree_skb(skb);
2764                 goto errout;
2765         }
2766
2767         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
2768 errout:
2769         return err;
2770 }
2771
2772 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info)
2773 {
2774         struct sk_buff *skb;
2775         struct net *net = info->nl_net;
2776         u32 seq;
2777         int err;
2778
2779         err = -ENOBUFS;
2780         seq = info->nlh ? info->nlh->nlmsg_seq : 0;
2781
2782         skb = nlmsg_new(rt6_nlmsg_size(), gfp_any());
2783         if (!skb)
2784                 goto errout;
2785
2786         err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
2787                                 event, info->portid, seq, 0, 0, 0);
2788         if (err < 0) {
2789                 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
2790                 WARN_ON(err == -EMSGSIZE);
2791                 kfree_skb(skb);
2792                 goto errout;
2793         }
2794         rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
2795                     info->nlh, gfp_any());
2796         return;
2797 errout:
2798         if (err < 0)
2799                 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
2800 }
2801
2802 static int ip6_route_dev_notify(struct notifier_block *this,
2803                                 unsigned long event, void *ptr)
2804 {
2805         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2806         struct net *net = dev_net(dev);
2807
2808         if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
2809                 net->ipv6.ip6_null_entry->dst.dev = dev;
2810                 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
2811 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2812                 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
2813                 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
2814                 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
2815                 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
2816 #endif
2817         }
2818
2819         return NOTIFY_OK;
2820 }
2821
2822 /*
2823  *      /proc
2824  */
2825
2826 #ifdef CONFIG_PROC_FS
2827
2828 static const struct file_operations ipv6_route_proc_fops = {
2829         .owner          = THIS_MODULE,
2830         .open           = ipv6_route_open,
2831         .read           = seq_read,
2832         .llseek         = seq_lseek,
2833         .release        = seq_release_net,
2834 };
2835
2836 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
2837 {
2838         struct net *net = (struct net *)seq->private;
2839         seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
2840                    net->ipv6.rt6_stats->fib_nodes,
2841                    net->ipv6.rt6_stats->fib_route_nodes,
2842                    net->ipv6.rt6_stats->fib_rt_alloc,
2843                    net->ipv6.rt6_stats->fib_rt_entries,
2844                    net->ipv6.rt6_stats->fib_rt_cache,
2845                    dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
2846                    net->ipv6.rt6_stats->fib_discarded_routes);
2847
2848         return 0;
2849 }
2850
2851 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
2852 {
2853         return single_open_net(inode, file, rt6_stats_seq_show);
2854 }
2855
2856 static const struct file_operations rt6_stats_seq_fops = {
2857         .owner   = THIS_MODULE,
2858         .open    = rt6_stats_seq_open,
2859         .read    = seq_read,
2860         .llseek  = seq_lseek,
2861         .release = single_release_net,
2862 };
2863 #endif  /* CONFIG_PROC_FS */
2864
2865 #ifdef CONFIG_SYSCTL
2866
2867 static
2868 int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
2869                               void __user *buffer, size_t *lenp, loff_t *ppos)
2870 {
2871         struct net *net;
2872         int delay;
2873         if (!write)
2874                 return -EINVAL;
2875
2876         net = (struct net *)ctl->extra1;
2877         delay = net->ipv6.sysctl.flush_delay;
2878         proc_dointvec(ctl, write, buffer, lenp, ppos);
2879         fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
2880         return 0;
2881 }
2882
2883 struct ctl_table ipv6_route_table_template[] = {
2884         {
2885                 .procname       =       "flush",
2886                 .data           =       &init_net.ipv6.sysctl.flush_delay,
2887                 .maxlen         =       sizeof(int),
2888                 .mode           =       0200,
2889                 .proc_handler   =       ipv6_sysctl_rtcache_flush
2890         },
2891         {
2892                 .procname       =       "gc_thresh",
2893                 .data           =       &ip6_dst_ops_template.gc_thresh,
2894                 .maxlen         =       sizeof(int),
2895                 .mode           =       0644,
2896                 .proc_handler   =       proc_dointvec,
2897         },
2898         {
2899                 .procname       =       "max_size",
2900                 .data           =       &init_net.ipv6.sysctl.ip6_rt_max_size,
2901                 .maxlen         =       sizeof(int),
2902                 .mode           =       0644,
2903                 .proc_handler   =       proc_dointvec,
2904         },
2905         {
2906                 .procname       =       "gc_min_interval",
2907                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2908                 .maxlen         =       sizeof(int),
2909                 .mode           =       0644,
2910                 .proc_handler   =       proc_dointvec_jiffies,
2911         },
2912         {
2913                 .procname       =       "gc_timeout",
2914                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
2915                 .maxlen         =       sizeof(int),
2916                 .mode           =       0644,
2917                 .proc_handler   =       proc_dointvec_jiffies,
2918         },
2919         {
2920                 .procname       =       "gc_interval",
2921                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_interval,
2922                 .maxlen         =       sizeof(int),
2923                 .mode           =       0644,
2924                 .proc_handler   =       proc_dointvec_jiffies,
2925         },
2926         {
2927                 .procname       =       "gc_elasticity",
2928                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
2929                 .maxlen         =       sizeof(int),
2930                 .mode           =       0644,
2931                 .proc_handler   =       proc_dointvec,
2932         },
2933         {
2934                 .procname       =       "mtu_expires",
2935                 .data           =       &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
2936                 .maxlen         =       sizeof(int),
2937                 .mode           =       0644,
2938                 .proc_handler   =       proc_dointvec_jiffies,
2939         },
2940         {
2941                 .procname       =       "min_adv_mss",
2942                 .data           =       &init_net.ipv6.sysctl.ip6_rt_min_advmss,
2943                 .maxlen         =       sizeof(int),
2944                 .mode           =       0644,
2945                 .proc_handler   =       proc_dointvec,
2946         },
2947         {
2948                 .procname       =       "gc_min_interval_ms",
2949                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2950                 .maxlen         =       sizeof(int),
2951                 .mode           =       0644,
2952                 .proc_handler   =       proc_dointvec_ms_jiffies,
2953         },
2954         { }
2955 };
2956
2957 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
2958 {
2959         struct ctl_table *table;
2960
2961         table = kmemdup(ipv6_route_table_template,
2962                         sizeof(ipv6_route_table_template),
2963                         GFP_KERNEL);
2964
2965         if (table) {
2966                 table[0].data = &net->ipv6.sysctl.flush_delay;
2967                 table[0].extra1 = net;
2968                 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
2969                 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
2970                 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2971                 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
2972                 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
2973                 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
2974                 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
2975                 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
2976                 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2977
2978                 /* Don't export sysctls to unprivileged users */
2979                 if (net->user_ns != &init_user_ns)
2980                         table[0].procname = NULL;
2981         }
2982
2983         return table;
2984 }
2985 #endif
2986
2987 static int __net_init ip6_route_net_init(struct net *net)
2988 {
2989         int ret = -ENOMEM;
2990
2991         memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
2992                sizeof(net->ipv6.ip6_dst_ops));
2993
2994         if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
2995                 goto out_ip6_dst_ops;
2996
2997         net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
2998                                            sizeof(*net->ipv6.ip6_null_entry),
2999                                            GFP_KERNEL);
3000         if (!net->ipv6.ip6_null_entry)
3001                 goto out_ip6_dst_entries;
3002         net->ipv6.ip6_null_entry->dst.path =
3003                 (struct dst_entry *)net->ipv6.ip6_null_entry;
3004         net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3005         dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
3006                          ip6_template_metrics, true);
3007
3008 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3009         net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
3010                                                sizeof(*net->ipv6.ip6_prohibit_entry),
3011                                                GFP_KERNEL);
3012         if (!net->ipv6.ip6_prohibit_entry)
3013                 goto out_ip6_null_entry;
3014         net->ipv6.ip6_prohibit_entry->dst.path =
3015                 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
3016         net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3017         dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
3018                          ip6_template_metrics, true);
3019
3020         net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
3021                                                sizeof(*net->ipv6.ip6_blk_hole_entry),
3022                                                GFP_KERNEL);
3023         if (!net->ipv6.ip6_blk_hole_entry)
3024                 goto out_ip6_prohibit_entry;
3025         net->ipv6.ip6_blk_hole_entry->dst.path =
3026                 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
3027         net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3028         dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
3029                          ip6_template_metrics, true);
3030 #endif
3031
3032         net->ipv6.sysctl.flush_delay = 0;
3033         net->ipv6.sysctl.ip6_rt_max_size = 4096;
3034         net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
3035         net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
3036         net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
3037         net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
3038         net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
3039         net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
3040
3041         net->ipv6.ip6_rt_gc_expire = 30*HZ;
3042
3043         ret = 0;
3044 out:
3045         return ret;
3046
3047 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3048 out_ip6_prohibit_entry:
3049         kfree(net->ipv6.ip6_prohibit_entry);
3050 out_ip6_null_entry:
3051         kfree(net->ipv6.ip6_null_entry);
3052 #endif
3053 out_ip6_dst_entries:
3054         dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3055 out_ip6_dst_ops:
3056         goto out;
3057 }
3058
3059 static void __net_exit ip6_route_net_exit(struct net *net)
3060 {
3061         kfree(net->ipv6.ip6_null_entry);
3062 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3063         kfree(net->ipv6.ip6_prohibit_entry);
3064         kfree(net->ipv6.ip6_blk_hole_entry);
3065 #endif
3066         dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3067 }
3068
3069 static int __net_init ip6_route_net_init_late(struct net *net)
3070 {
3071 #ifdef CONFIG_PROC_FS
3072         proc_create("ipv6_route", 0, net->proc_net, &ipv6_route_proc_fops);
3073         proc_create("rt6_stats", S_IRUGO, net->proc_net, &rt6_stats_seq_fops);
3074 #endif
3075         return 0;
3076 }
3077
3078 static void __net_exit ip6_route_net_exit_late(struct net *net)
3079 {
3080 #ifdef CONFIG_PROC_FS
3081         remove_proc_entry("ipv6_route", net->proc_net);
3082         remove_proc_entry("rt6_stats", net->proc_net);
3083 #endif
3084 }
3085
3086 static struct pernet_operations ip6_route_net_ops = {
3087         .init = ip6_route_net_init,
3088         .exit = ip6_route_net_exit,
3089 };
3090
3091 static int __net_init ipv6_inetpeer_init(struct net *net)
3092 {
3093         struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
3094
3095         if (!bp)
3096                 return -ENOMEM;
3097         inet_peer_base_init(bp);
3098         net->ipv6.peers = bp;
3099         return 0;
3100 }
3101
3102 static void __net_exit ipv6_inetpeer_exit(struct net *net)
3103 {
3104         struct inet_peer_base *bp = net->ipv6.peers;
3105
3106         net->ipv6.peers = NULL;
3107         inetpeer_invalidate_tree(bp);
3108         kfree(bp);
3109 }
3110
3111 static struct pernet_operations ipv6_inetpeer_ops = {
3112         .init   =       ipv6_inetpeer_init,
3113         .exit   =       ipv6_inetpeer_exit,
3114 };
3115
3116 static struct pernet_operations ip6_route_net_late_ops = {
3117         .init = ip6_route_net_init_late,
3118         .exit = ip6_route_net_exit_late,
3119 };
3120
3121 static struct notifier_block ip6_route_dev_notifier = {
3122         .notifier_call = ip6_route_dev_notify,
3123         .priority = 0,
3124 };
3125
3126 int __init ip6_route_init(void)
3127 {
3128         int ret;
3129
3130         ret = -ENOMEM;
3131         ip6_dst_ops_template.kmem_cachep =
3132                 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
3133                                   SLAB_HWCACHE_ALIGN, NULL);
3134         if (!ip6_dst_ops_template.kmem_cachep)
3135                 goto out;
3136
3137         ret = dst_entries_init(&ip6_dst_blackhole_ops);
3138         if (ret)
3139                 goto out_kmem_cache;
3140
3141         ret = register_pernet_subsys(&ipv6_inetpeer_ops);
3142         if (ret)
3143                 goto out_dst_entries;
3144
3145         ret = register_pernet_subsys(&ip6_route_net_ops);
3146         if (ret)
3147                 goto out_register_inetpeer;
3148
3149         ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
3150
3151         /* Registering of the loopback is done before this portion of code,
3152          * the loopback reference in rt6_info will not be taken, do it
3153          * manually for init_net */
3154         init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
3155         init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3156   #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3157         init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
3158         init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3159         init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
3160         init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3161   #endif
3162         ret = fib6_init();
3163         if (ret)
3164                 goto out_register_subsys;
3165
3166         ret = xfrm6_init();
3167         if (ret)
3168                 goto out_fib6_init;
3169
3170         ret = fib6_rules_init();
3171         if (ret)
3172                 goto xfrm6_init;
3173
3174         ret = register_pernet_subsys(&ip6_route_net_late_ops);
3175         if (ret)
3176                 goto fib6_rules_init;
3177
3178         ret = -ENOBUFS;
3179         if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL, NULL) ||
3180             __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL, NULL) ||
3181             __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL, NULL))
3182                 goto out_register_late_subsys;
3183
3184         ret = register_netdevice_notifier(&ip6_route_dev_notifier);
3185         if (ret)
3186                 goto out_register_late_subsys;
3187
3188 out:
3189         return ret;
3190
3191 out_register_late_subsys:
3192         unregister_pernet_subsys(&ip6_route_net_late_ops);
3193 fib6_rules_init:
3194         fib6_rules_cleanup();
3195 xfrm6_init:
3196         xfrm6_fini();
3197 out_fib6_init:
3198         fib6_gc_cleanup();
3199 out_register_subsys:
3200         unregister_pernet_subsys(&ip6_route_net_ops);
3201 out_register_inetpeer:
3202         unregister_pernet_subsys(&ipv6_inetpeer_ops);
3203 out_dst_entries:
3204         dst_entries_destroy(&ip6_dst_blackhole_ops);
3205 out_kmem_cache:
3206         kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3207         goto out;
3208 }
3209
3210 void ip6_route_cleanup(void)
3211 {
3212         unregister_netdevice_notifier(&ip6_route_dev_notifier);
3213         unregister_pernet_subsys(&ip6_route_net_late_ops);
3214         fib6_rules_cleanup();
3215         xfrm6_fini();
3216         fib6_gc_cleanup();
3217         unregister_pernet_subsys(&ipv6_inetpeer_ops);
3218         unregister_pernet_subsys(&ip6_route_net_ops);
3219         dst_entries_destroy(&ip6_dst_blackhole_ops);
3220         kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3221 }