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