net/ipv6/route.c

   1 /*
   2  *      Linux INET6 implementation
   3  *      FIB front-end.
   4  *
   5  *      Authors:
   6  *      Pedro Roque             <roque@di.fc.ul.pt>
   7  *
   8  *      This program is free software; you can redistribute it and/or
   9  *      modify it under the terms of the GNU General Public License
  10  *      as published by the Free Software Foundation; either version
  11  *      2 of the License, or (at your option) any later version.
  12  */
  13
  14 /*      Changes:
  15  *
  16  *      YOSHIFUJI Hideaki @USAGI
  17  *              reworked default router selection.
  18  *              - respect outgoing interface
  19  *              - select from (probably) reachable routers (i.e.
  20  *              routers in REACHABLE, STALE, DELAY or PROBE states).
  21  *              - always select the same router if it is (probably)
  22  *              reachable.  otherwise, round-robin the list.
  23  *      Ville Nuorvala
  24  *              Fixed routing subtrees.
  25  */
  26
  27 #define pr_fmt(fmt) "IPv6: " fmt
  28
  29 #include <linux/capability.h>
  30 #include <linux/errno.h>
  31 #include <linux/export.h>
  32 #include <linux/types.h>
  33 #include <linux/times.h>
  34 #include <linux/socket.h>
  35 #include <linux/sockios.h>
  36 #include <linux/net.h>
  37 #include <linux/route.h>
  38 #include <linux/netdevice.h>
  39 #include <linux/in6.h>
  40 #include <linux/mroute6.h>
  41 #include <linux/init.h>
  42 #include <linux/if_arp.h>
  43 #include <linux/proc_fs.h>
  44 #include <linux/seq_file.h>
  45 #include <linux/nsproxy.h>
  46 #include <linux/slab.h>
  47 #include <net/net_namespace.h>
  48 #include <net/snmp.h>
  49 #include <net/ipv6.h>
  50 #include <net/ip6_fib.h>
  51 #include <net/ip6_route.h>
  52 #include <net/ndisc.h>
  53 #include <net/addrconf.h>
  54 #include <net/tcp.h>
  55 #include <linux/rtnetlink.h>
  56 #include <net/dst.h>
  57 #include <net/xfrm.h>
  58 #include <net/netevent.h>
  59 #include <net/netlink.h>
  60 #include <net/nexthop.h>
  61
  62 #include <asm/uaccess.h>
  63
  64 #ifdef CONFIG_SYSCTL
  65 #include <linux/sysctl.h>
  66 #endif
  67
  68 enum rt6_nud_state {
  69         RT6_NUD_FAIL_HARD = -2,
  70         RT6_NUD_FAIL_SOFT = -1,
  71         RT6_NUD_SUCCEED = 1
  72 };
  73
  74 static struct rt6_info *ip6_rt_copy(struct rt6_info *ort,
  75                                     const struct in6_addr *dest);
  76 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie);
  77 static unsigned int      ip6_default_advmss(const struct dst_entry *dst);
  78 static unsigned int      ip6_mtu(const struct dst_entry *dst);
  79 static struct dst_entry *ip6_negative_advice(struct dst_entry *);
  80 static void             ip6_dst_destroy(struct dst_entry *);
  81 static void             ip6_dst_ifdown(struct dst_entry *,
  82                                        struct net_device *dev, int how);
  83 static int               ip6_dst_gc(struct dst_ops *ops);
  84
  85 static int              ip6_pkt_discard(struct sk_buff *skb);
  86 static int              ip6_pkt_discard_out(struct sk_buff *skb);
  87 static void             ip6_link_failure(struct sk_buff *skb);
  88 static void             ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
  89                                            struct sk_buff *skb, u32 mtu);
  90 static void             rt6_do_redirect(struct dst_entry *dst, struct sock *sk,
  91                                         struct sk_buff *skb);
  92 static int rt6_score_route(struct rt6_info *rt, int oif, int strict);
  93
  94 #ifdef CONFIG_IPV6_ROUTE_INFO
  95 static struct rt6_info *rt6_add_route_info(struct net *net,
  96                                            const struct in6_addr *prefix, int prefixlen,
  97                                            const struct in6_addr *gwaddr, int ifindex,
  98                                            unsigned int pref);
  99 static struct rt6_info *rt6_get_route_info(struct net *net,
 100                                            const struct in6_addr *prefix, int prefixlen,
 101                                            const struct in6_addr *gwaddr, int ifindex);
 102 #endif
 103
 104 static u32 *ipv6_cow_metrics(struct dst_entry *dst, unsigned long old)
 105 {
 106         struct rt6_info *rt = (struct rt6_info *) dst;
 107         struct inet_peer *peer;
 108         u32 *p = NULL;
 109
 110         if (!(rt->dst.flags & DST_HOST))
 111                 return NULL;
 112
 113         peer = rt6_get_peer_create(rt);
 114         if (peer) {
 115                 u32 *old_p = __DST_METRICS_PTR(old);
 116                 unsigned long prev, new;
 117
 118                 p = peer->metrics;
 119                 if (inet_metrics_new(peer))
 120                         memcpy(p, old_p, sizeof(u32) * RTAX_MAX);
 121
 122                 new = (unsigned long) p;
 123                 prev = cmpxchg(&dst->_metrics, old, new);
 124
 125                 if (prev != old) {
 126                         p = __DST_METRICS_PTR(prev);
 127                         if (prev & DST_METRICS_READ_ONLY)
 128                                 p = NULL;
 129                 }
 130         }
 131         return p;
 132 }
 133
 134 static inline const void *choose_neigh_daddr(struct rt6_info *rt,
 135                                              struct sk_buff *skb,
 136                                              const void *daddr)
 137 {
 138         struct in6_addr *p = &rt->rt6i_gateway;
 139
 140         if (!ipv6_addr_any(p))
 141                 return (const void *) p;
 142         else if (skb)
 143                 return &ipv6_hdr(skb)->daddr;
 144         return daddr;
 145 }
 146
 147 static struct neighbour *ip6_neigh_lookup(const struct dst_entry *dst,
 148                                           struct sk_buff *skb,
 149                                           const void *daddr)
 150 {
 151         struct rt6_info *rt = (struct rt6_info *) dst;
 152         struct neighbour *n;
 153
 154         daddr = choose_neigh_daddr(rt, skb, daddr);
 155         n = __ipv6_neigh_lookup(dst->dev, daddr);
 156         if (n)
 157                 return n;
 158         return neigh_create(&nd_tbl, daddr, dst->dev);
 159 }
 160
 161 static struct dst_ops ip6_dst_ops_template = {
 162         .family                 =       AF_INET6,
 163         .protocol               =       cpu_to_be16(ETH_P_IPV6),
 164         .gc                     =       ip6_dst_gc,
 165         .gc_thresh              =       1024,
 166         .check                  =       ip6_dst_check,
 167         .default_advmss         =       ip6_default_advmss,
 168         .mtu                    =       ip6_mtu,
 169         .cow_metrics            =       ipv6_cow_metrics,
 170         .destroy                =       ip6_dst_destroy,
 171         .ifdown                 =       ip6_dst_ifdown,
 172         .negative_advice        =       ip6_negative_advice,
 173         .link_failure           =       ip6_link_failure,
 174         .update_pmtu            =       ip6_rt_update_pmtu,
 175         .redirect               =       rt6_do_redirect,
 176         .local_out              =       __ip6_local_out,
 177         .neigh_lookup           =       ip6_neigh_lookup,
 178 };
 179
 180 static unsigned int ip6_blackhole_mtu(const struct dst_entry *dst)
 181 {
 182         unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
 183
 184         return mtu ? : dst->dev->mtu;
 185 }
 186
 187 static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, struct sock *sk,
 188                                          struct sk_buff *skb, u32 mtu)
 189 {
 190 }
 191
 192 static void ip6_rt_blackhole_redirect(struct dst_entry *dst, struct sock *sk,
 193                                       struct sk_buff *skb)
 194 {
 195 }
 196
 197 static u32 *ip6_rt_blackhole_cow_metrics(struct dst_entry *dst,
 198                                          unsigned long old)
 199 {
 200         return NULL;
 201 }
 202
 203 static struct dst_ops ip6_dst_blackhole_ops = {
 204         .family                 =       AF_INET6,
 205         .protocol               =       cpu_to_be16(ETH_P_IPV6),
 206         .destroy                =       ip6_dst_destroy,
 207         .check                  =       ip6_dst_check,
 208         .mtu                    =       ip6_blackhole_mtu,
 209         .default_advmss         =       ip6_default_advmss,
 210         .update_pmtu            =       ip6_rt_blackhole_update_pmtu,
 211         .redirect               =       ip6_rt_blackhole_redirect,
 212         .cow_metrics            =       ip6_rt_blackhole_cow_metrics,
 213         .neigh_lookup           =       ip6_neigh_lookup,
 214 };
 215
 216 static const u32 ip6_template_metrics[RTAX_MAX] = {
 217         [RTAX_HOPLIMIT - 1] = 0,
 218 };
 219
 220 static const struct rt6_info ip6_null_entry_template = {
 221         .dst = {
 222                 .__refcnt       = ATOMIC_INIT(1),
 223                 .__use          = 1,
 224                 .obsolete       = DST_OBSOLETE_FORCE_CHK,
 225                 .error          = -ENETUNREACH,
 226                 .input          = ip6_pkt_discard,
 227                 .output         = ip6_pkt_discard_out,
 228         },
 229         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
 230         .rt6i_protocol  = RTPROT_KERNEL,
 231         .rt6i_metric    = ~(u32) 0,
 232         .rt6i_ref       = ATOMIC_INIT(1),
 233 };
 234
 235 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
 236
 237 static int ip6_pkt_prohibit(struct sk_buff *skb);
 238 static int ip6_pkt_prohibit_out(struct sk_buff *skb);
 239
 240 static const struct rt6_info ip6_prohibit_entry_template = {
 241         .dst = {
 242                 .__refcnt       = ATOMIC_INIT(1),
 243                 .__use          = 1,
 244                 .obsolete       = DST_OBSOLETE_FORCE_CHK,
 245                 .error          = -EACCES,
 246                 .input          = ip6_pkt_prohibit,
 247                 .output         = ip6_pkt_prohibit_out,
 248         },
 249         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
 250         .rt6i_protocol  = RTPROT_KERNEL,
 251         .rt6i_metric    = ~(u32) 0,
 252         .rt6i_ref       = ATOMIC_INIT(1),
 253 };
 254
 255 static const struct rt6_info ip6_blk_hole_entry_template = {
 256         .dst = {
 257                 .__refcnt       = ATOMIC_INIT(1),
 258                 .__use          = 1,
 259                 .obsolete       = DST_OBSOLETE_FORCE_CHK,
 260                 .error          = -EINVAL,
 261                 .input          = dst_discard,
 262                 .output         = dst_discard,
 263         },
 264         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
 265         .rt6i_protocol  = RTPROT_KERNEL,
 266         .rt6i_metric    = ~(u32) 0,
 267         .rt6i_ref       = ATOMIC_INIT(1),
 268 };
 269
 270 #endif
 271
 272 /* allocate dst with ip6_dst_ops */
 273 static inline struct rt6_info *ip6_dst_alloc(struct net *net,
 274                                              struct net_device *dev,
 275                                              int flags,
 276                                              struct fib6_table *table)
 277 {
 278         struct rt6_info *rt = dst_alloc(&net->ipv6.ip6_dst_ops, dev,
 279                                         0, DST_OBSOLETE_FORCE_CHK, flags);
 280
 281         if (rt) {
 282                 struct dst_entry *dst = &rt->dst;
 283
 284                 memset(dst + 1, 0, sizeof(*rt) - sizeof(*dst));
 285                 rt6_init_peer(rt, table ? &table->tb6_peers : net->ipv6.peers);
 286                 rt->rt6i_genid = rt_genid_ipv6(net);
 287                 INIT_LIST_HEAD(&rt->rt6i_siblings);
 288         }
 289         return rt;
 290 }
 291
 292 static void ip6_dst_destroy(struct dst_entry *dst)
 293 {
 294         struct rt6_info *rt = (struct rt6_info *)dst;
 295         struct inet6_dev *idev = rt->rt6i_idev;
 296         struct dst_entry *from = dst->from;
 297
 298         if (!(rt->dst.flags & DST_HOST))
 299                 dst_destroy_metrics_generic(dst);
 300
 301         if (idev) {
 302                 rt->rt6i_idev = NULL;
 303                 in6_dev_put(idev);
 304         }
 305
 306         dst->from = NULL;
 307         dst_release(from);
 308
 309         if (rt6_has_peer(rt)) {
 310                 struct inet_peer *peer = rt6_peer_ptr(rt);
 311                 inet_putpeer(peer);
 312         }
 313 }
 314
 315 void rt6_bind_peer(struct rt6_info *rt, int create)
 316 {
 317         struct inet_peer_base *base;
 318         struct inet_peer *peer;
 319
 320         base = inetpeer_base_ptr(rt->_rt6i_peer);
 321         if (!base)
 322                 return;
 323
 324         peer = inet_getpeer_v6(base, &rt->rt6i_dst.addr, create);
 325         if (peer) {
 326                 if (!rt6_set_peer(rt, peer))
 327                         inet_putpeer(peer);
 328         }
 329 }
 330
 331 static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
 332                            int how)
 333 {
 334         struct rt6_info *rt = (struct rt6_info *)dst;
 335         struct inet6_dev *idev = rt->rt6i_idev;
 336         struct net_device *loopback_dev =
 337                 dev_net(dev)->loopback_dev;
 338
 339         if (dev != loopback_dev) {
 340                 if (idev && idev->dev == dev) {
 341                         struct inet6_dev *loopback_idev =
 342                                 in6_dev_get(loopback_dev);
 343                         if (loopback_idev) {
 344                                 rt->rt6i_idev = loopback_idev;
 345                                 in6_dev_put(idev);
 346                         }
 347                 }
 348         }
 349 }
 350
 351 static bool rt6_check_expired(const struct rt6_info *rt)
 352 {
 353         if (rt->rt6i_flags & RTF_EXPIRES) {
 354                 if (time_after(jiffies, rt->dst.expires))
 355                         return true;
 356         } else if (rt->dst.from) {
 357                 return rt6_check_expired((struct rt6_info *) rt->dst.from);
 358         }
 359         return false;
 360 }
 361
 362 static bool rt6_need_strict(const struct in6_addr *daddr)
 363 {
 364         return ipv6_addr_type(daddr) &
 365                 (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
 366 }
 367
 368 /* Multipath route selection:
 369  *   Hash based function using packet header and flowlabel.
 370  * Adapted from fib_info_hashfn()
 371  */
 372 static int rt6_info_hash_nhsfn(unsigned int candidate_count,
 373                                const struct flowi6 *fl6)
 374 {
 375         unsigned int val = fl6->flowi6_proto;
 376
 377         val ^= ipv6_addr_hash(&fl6->daddr);
 378         val ^= ipv6_addr_hash(&fl6->saddr);
 379
 380         /* Work only if this not encapsulated */
 381         switch (fl6->flowi6_proto) {
 382         case IPPROTO_UDP:
 383         case IPPROTO_TCP:
 384         case IPPROTO_SCTP:
 385                 val ^= (__force u16)fl6->fl6_sport;
 386                 val ^= (__force u16)fl6->fl6_dport;
 387                 break;
 388
 389         case IPPROTO_ICMPV6:
 390                 val ^= (__force u16)fl6->fl6_icmp_type;
 391                 val ^= (__force u16)fl6->fl6_icmp_code;
 392                 break;
 393         }
 394         /* RFC6438 recommands to use flowlabel */
 395         val ^= (__force u32)fl6->flowlabel;
 396
 397         /* Perhaps, we need to tune, this function? */
 398         val = val ^ (val >> 7) ^ (val >> 12);
 399         return val % candidate_count;
 400 }
 401
 402 static struct rt6_info *rt6_multipath_select(struct rt6_info *match,
 403                                              struct flowi6 *fl6, int oif,
 404                                              int strict)
 405 {
 406         struct rt6_info *sibling, *next_sibling;
 407         int route_choosen;
 408
 409         route_choosen = rt6_info_hash_nhsfn(match->rt6i_nsiblings + 1, fl6);
 410         /* Don't change the route, if route_choosen == 0
 411          * (siblings does not include ourself)
 412          */
 413         if (route_choosen)
 414                 list_for_each_entry_safe(sibling, next_sibling,
 415                                 &match->rt6i_siblings, rt6i_siblings) {
 416                         route_choosen--;
 417                         if (route_choosen == 0) {
 418                                 if (rt6_score_route(sibling, oif, strict) < 0)
 419                                         break;
 420                                 match = sibling;
 421                                 break;
 422                         }
 423                 }
 424         return match;
 425 }
 426
 427 /*
 428  *      Route lookup. Any table->tb6_lock is implied.
 429  */
 430
 431 static inline struct rt6_info *rt6_device_match(struct net *net,
 432                                                     struct rt6_info *rt,
 433                                                     const struct in6_addr *saddr,
 434                                                     int oif,
 435                                                     int flags)
 436 {
 437         struct rt6_info *local = NULL;
 438         struct rt6_info *sprt;
 439
 440         if (!oif && ipv6_addr_any(saddr))
 441                 goto out;
 442
 443         for (sprt = rt; sprt; sprt = sprt->dst.rt6_next) {
 444                 struct net_device *dev = sprt->dst.dev;
 445
 446                 if (oif) {
 447                         if (dev->ifindex == oif)
 448                                 return sprt;
 449                         if (dev->flags & IFF_LOOPBACK) {
 450                                 if (!sprt->rt6i_idev ||
 451                                     sprt->rt6i_idev->dev->ifindex != oif) {
 452                                         if (flags & RT6_LOOKUP_F_IFACE && oif)
 453                                                 continue;
 454                                         if (local && (!oif ||
 455                                                       local->rt6i_idev->dev->ifindex == oif))
 456                                                 continue;
 457                                 }
 458                                 local = sprt;
 459                         }
 460                 } else {
 461                         if (ipv6_chk_addr(net, saddr, dev,
 462                                           flags & RT6_LOOKUP_F_IFACE))
 463                                 return sprt;
 464                 }
 465         }
 466
 467         if (oif) {
 468                 if (local)
 469                         return local;
 470
 471                 if (flags & RT6_LOOKUP_F_IFACE)
 472                         return net->ipv6.ip6_null_entry;
 473         }
 474 out:
 475         return rt;
 476 }
 477
 478 #ifdef CONFIG_IPV6_ROUTER_PREF
 479 struct __rt6_probe_work {
 480         struct work_struct work;
 481         struct in6_addr target;
 482         struct net_device *dev;
 483 };
 484
 485 static void rt6_probe_deferred(struct work_struct *w)
 486 {
 487         struct in6_addr mcaddr;
 488         struct __rt6_probe_work *work =
 489                 container_of(w, struct __rt6_probe_work, work);
 490
 491         addrconf_addr_solict_mult(&work->target, &mcaddr);
 492         ndisc_send_ns(work->dev, NULL, &work->target, &mcaddr, NULL);
 493         dev_put(work->dev);
 494         kfree(w);
 495 }
 496
 497 static void rt6_probe(struct rt6_info *rt)
 498 {
 499         struct neighbour *neigh;
 500         /*
 501          * Okay, this does not seem to be appropriate
 502          * for now, however, we need to check if it
 503          * is really so; aka Router Reachability Probing.
 504          *
 505          * Router Reachability Probe MUST be rate-limited
 506          * to no more than one per minute.
 507          */
 508         if (!rt || !(rt->rt6i_flags & RTF_GATEWAY))
 509                 return;
 510         rcu_read_lock_bh();
 511         neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
 512         if (neigh) {
 513                 write_lock(&neigh->lock);
 514                 if (neigh->nud_state & NUD_VALID)
 515                         goto out;
 516         }
 517
 518         if (!neigh ||
 519             time_after(jiffies, neigh->updated + rt->rt6i_idev->cnf.rtr_probe_interval)) {
 520                 struct __rt6_probe_work *work;
 521
 522                 work = kmalloc(sizeof(*work), GFP_ATOMIC);
 523
 524                 if (neigh && work)
 525                         neigh->updated = jiffies;
 526
 527                 if (neigh)
 528                         write_unlock(&neigh->lock);
 529
 530                 if (work) {
 531                         INIT_WORK(&work->work, rt6_probe_deferred);
 532                         work->target = rt->rt6i_gateway;
 533                         dev_hold(rt->dst.dev);
 534                         work->dev = rt->dst.dev;
 535                         schedule_work(&work->work);
 536                 }
 537         } else {
 538 out:
 539                 write_unlock(&neigh->lock);
 540         }
 541         rcu_read_unlock_bh();
 542 }
 543 #else
 544 static inline void rt6_probe(struct rt6_info *rt)
 545 {
 546 }
 547 #endif
 548
 549 /*
 550  * Default Router Selection (RFC 2461 6.3.6)
 551  */
 552 static inline int rt6_check_dev(struct rt6_info *rt, int oif)
 553 {
 554         struct net_device *dev = rt->dst.dev;
 555         if (!oif || dev->ifindex == oif)
 556                 return 2;
 557         if ((dev->flags & IFF_LOOPBACK) &&
 558             rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
 559                 return 1;
 560         return 0;
 561 }
 562
 563 static inline enum rt6_nud_state rt6_check_neigh(struct rt6_info *rt)
 564 {
 565         struct neighbour *neigh;
 566         enum rt6_nud_state ret = RT6_NUD_FAIL_HARD;
 567
 568         if (rt->rt6i_flags & RTF_NONEXTHOP ||
 569             !(rt->rt6i_flags & RTF_GATEWAY))
 570                 return RT6_NUD_SUCCEED;
 571
 572         rcu_read_lock_bh();
 573         neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
 574         if (neigh) {
 575                 read_lock(&neigh->lock);
 576                 if (neigh->nud_state & NUD_VALID)
 577                         ret = RT6_NUD_SUCCEED;
 578 #ifdef CONFIG_IPV6_ROUTER_PREF
 579                 else if (!(neigh->nud_state & NUD_FAILED))
 580                         ret = RT6_NUD_SUCCEED;
 581 #endif
 582                 read_unlock(&neigh->lock);
 583         } else {
 584                 ret = IS_ENABLED(CONFIG_IPV6_ROUTER_PREF) ?
 585                       RT6_NUD_SUCCEED : RT6_NUD_FAIL_SOFT;
 586         }
 587         rcu_read_unlock_bh();
 588
 589         return ret;
 590 }
 591
 592 static int rt6_score_route(struct rt6_info *rt, int oif,
 593                            int strict)
 594 {
 595         int m;
 596
 597         m = rt6_check_dev(rt, oif);
 598         if (!m && (strict & RT6_LOOKUP_F_IFACE))
 599                 return RT6_NUD_FAIL_HARD;
 600 #ifdef CONFIG_IPV6_ROUTER_PREF
 601         m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
 602 #endif
 603         if (strict & RT6_LOOKUP_F_REACHABLE) {
 604                 int n = rt6_check_neigh(rt);
 605                 if (n < 0)
 606                         return n;
 607         }
 608         return m;
 609 }
 610
 611 static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
 612                                    int *mpri, struct rt6_info *match,
 613                                    bool *do_rr)
 614 {
 615         int m;
 616         bool match_do_rr = false;
 617
 618         if (rt6_check_expired(rt))
 619                 goto out;
 620
 621         m = rt6_score_route(rt, oif, strict);
 622         if (m == RT6_NUD_FAIL_SOFT && !IS_ENABLED(CONFIG_IPV6_ROUTER_PREF)) {
 623                 match_do_rr = true;
 624                 m = 0; /* lowest valid score */
 625         } else if (m < 0) {
 626                 goto out;
 627         }
 628
 629         if (strict & RT6_LOOKUP_F_REACHABLE)
 630                 rt6_probe(rt);
 631
 632         if (m > *mpri) {
 633                 *do_rr = match_do_rr;
 634                 *mpri = m;
 635                 match = rt;
 636         }
 637 out:
 638         return match;
 639 }
 640
 641 static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
 642                                      struct rt6_info *rr_head,
 643                                      u32 metric, int oif, int strict,
 644                                      bool *do_rr)
 645 {
 646         struct rt6_info *rt, *match;
 647         int mpri = -1;
 648
 649         match = NULL;
 650         for (rt = rr_head; rt && rt->rt6i_metric == metric;
 651              rt = rt->dst.rt6_next)
 652                 match = find_match(rt, oif, strict, &mpri, match, do_rr);
 653         for (rt = fn->leaf; rt && rt != rr_head && rt->rt6i_metric == metric;
 654              rt = rt->dst.rt6_next)
 655                 match = find_match(rt, oif, strict, &mpri, match, do_rr);
 656
 657         return match;
 658 }
 659
 660 static struct rt6_info *rt6_select(struct fib6_node *fn, int oif, int strict)
 661 {
 662         struct rt6_info *match, *rt0;
 663         struct net *net;
 664         bool do_rr = false;
 665
 666         rt0 = fn->rr_ptr;
 667         if (!rt0)
 668                 fn->rr_ptr = rt0 = fn->leaf;
 669
 670         match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict,
 671                              &do_rr);
 672
 673         if (do_rr) {
 674                 struct rt6_info *next = rt0->dst.rt6_next;
 675
 676                 /* no entries matched; do round-robin */
 677                 if (!next || next->rt6i_metric != rt0->rt6i_metric)
 678                         next = fn->leaf;
 679
 680                 if (next != rt0)
 681                         fn->rr_ptr = next;
 682         }
 683
 684         net = dev_net(rt0->dst.dev);
 685         return match ? match : net->ipv6.ip6_null_entry;
 686 }
 687
 688 #ifdef CONFIG_IPV6_ROUTE_INFO
 689 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
 690                   const struct in6_addr *gwaddr)
 691 {
 692         struct net *net = dev_net(dev);
 693         struct route_info *rinfo = (struct route_info *) opt;
 694         struct in6_addr prefix_buf, *prefix;
 695         unsigned int pref;
 696         unsigned long lifetime;
 697         struct rt6_info *rt;
 698
 699         if (len < sizeof(struct route_info)) {
 700                 return -EINVAL;
 701         }
 702
 703         /* Sanity check for prefix_len and length */
 704         if (rinfo->length > 3) {
 705                 return -EINVAL;
 706         } else if (rinfo->prefix_len > 128) {
 707                 return -EINVAL;
 708         } else if (rinfo->prefix_len > 64) {
 709                 if (rinfo->length < 2) {
 710                         return -EINVAL;
 711                 }
 712         } else if (rinfo->prefix_len > 0) {
 713                 if (rinfo->length < 1) {
 714                         return -EINVAL;
 715                 }
 716         }
 717
 718         pref = rinfo->route_pref;
 719         if (pref == ICMPV6_ROUTER_PREF_INVALID)
 720                 return -EINVAL;
 721
 722         lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
 723
 724         if (rinfo->length == 3)
 725                 prefix = (struct in6_addr *)rinfo->prefix;
 726         else {
 727                 /* this function is safe */
 728                 ipv6_addr_prefix(&prefix_buf,
 729                                  (struct in6_addr *)rinfo->prefix,
 730                                  rinfo->prefix_len);
 731                 prefix = &prefix_buf;
 732         }
 733
 734         rt = rt6_get_route_info(net, prefix, rinfo->prefix_len, gwaddr,
 735                                 dev->ifindex);
 736
 737         if (rt && !lifetime) {
 738                 ip6_del_rt(rt);
 739                 rt = NULL;
 740         }
 741
 742         if (!rt && lifetime)
 743                 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr, dev->ifindex,
 744                                         pref);
 745         else if (rt)
 746                 rt->rt6i_flags = RTF_ROUTEINFO |
 747                                  (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
 748
 749         if (rt) {
 750                 if (!addrconf_finite_timeout(lifetime))
 751                         rt6_clean_expires(rt);
 752                 else
 753                         rt6_set_expires(rt, jiffies + HZ * lifetime);
 754
 755                 ip6_rt_put(rt);
 756         }
 757         return 0;
 758 }
 759 #endif
 760
 761 #define BACKTRACK(__net, saddr)                 \
 762 do { \
 763         if (rt == __net->ipv6.ip6_null_entry) { \
 764                 struct fib6_node *pn; \
 765                 while (1) { \
 766                         if (fn->fn_flags & RTN_TL_ROOT) \
 767                                 goto out; \
 768                         pn = fn->parent; \
 769                         if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn) \
 770                                 fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr); \
 771                         else \
 772                                 fn = pn; \
 773                         if (fn->fn_flags & RTN_RTINFO) \
 774                                 goto restart; \
 775                 } \
 776         } \
 777 } while (0)
 778
 779 static struct rt6_info *ip6_pol_route_lookup(struct net *net,
 780                                              struct fib6_table *table,
 781                                              struct flowi6 *fl6, int flags)
 782 {
 783         struct fib6_node *fn;
 784         struct rt6_info *rt;
 785
 786         read_lock_bh(&table->tb6_lock);
 787         fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
 788 restart:
 789         rt = fn->leaf;
 790         rt = rt6_device_match(net, rt, &fl6->saddr, fl6->flowi6_oif, flags);
 791         if (rt->rt6i_nsiblings && fl6->flowi6_oif == 0)
 792                 rt = rt6_multipath_select(rt, fl6, fl6->flowi6_oif, flags);
 793         BACKTRACK(net, &fl6->saddr);
 794 out:
 795         dst_use(&rt->dst, jiffies);
 796         read_unlock_bh(&table->tb6_lock);
 797         return rt;
 798
 799 }
 800
 801 struct dst_entry * ip6_route_lookup(struct net *net, struct flowi6 *fl6,
 802                                     int flags)
 803 {
 804         return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_lookup);
 805 }
 806 EXPORT_SYMBOL_GPL(ip6_route_lookup);
 807
 808 struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
 809                             const struct in6_addr *saddr, int oif, int strict)
 810 {
 811         struct flowi6 fl6 = {
 812                 .flowi6_oif = oif,
 813                 .daddr = *daddr,
 814         };
 815         struct dst_entry *dst;
 816         int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
 817
 818         if (saddr) {
 819                 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
 820                 flags |= RT6_LOOKUP_F_HAS_SADDR;
 821         }
 822
 823         dst = fib6_rule_lookup(net, &fl6, flags, ip6_pol_route_lookup);
 824         if (dst->error == 0)
 825                 return (struct rt6_info *) dst;
 826
 827         dst_release(dst);
 828
 829         return NULL;
 830 }
 831
 832 EXPORT_SYMBOL(rt6_lookup);
 833
 834 /* ip6_ins_rt is called with FREE table->tb6_lock.
 835    It takes new route entry, the addition fails by any reason the
 836    route is freed. In any case, if caller does not hold it, it may
 837    be destroyed.
 838  */
 839
 840 static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info)
 841 {
 842         int err;
 843         struct fib6_table *table;
 844
 845         table = rt->rt6i_table;
 846         write_lock_bh(&table->tb6_lock);
 847         err = fib6_add(&table->tb6_root, rt, info);
 848         write_unlock_bh(&table->tb6_lock);
 849
 850         return err;
 851 }
 852
 853 int ip6_ins_rt(struct rt6_info *rt)
 854 {
 855         struct nl_info info = {
 856                 .nl_net = dev_net(rt->dst.dev),
 857         };
 858         return __ip6_ins_rt(rt, &info);
 859 }
 860
 861 static struct rt6_info *rt6_alloc_cow(struct rt6_info *ort,
 862                                       const struct in6_addr *daddr,
 863                                       const struct in6_addr *saddr)
 864 {
 865         struct rt6_info *rt;
 866
 867         /*
 868          *      Clone the route.
 869          */
 870
 871         rt = ip6_rt_copy(ort, daddr);
 872
 873         if (rt) {
 874                 if (!(rt->rt6i_flags & RTF_GATEWAY)) {
 875                         if (ort->rt6i_dst.plen != 128 &&
 876                             ipv6_addr_equal(&ort->rt6i_dst.addr, daddr))
 877                                 rt->rt6i_flags |= RTF_ANYCAST;
 878                 }
 879
 880                 rt->rt6i_flags |= RTF_CACHE;
 881
 882 #ifdef CONFIG_IPV6_SUBTREES
 883                 if (rt->rt6i_src.plen && saddr) {
 884                         rt->rt6i_src.addr = *saddr;
 885                         rt->rt6i_src.plen = 128;
 886                 }
 887 #endif
 888         }
 889
 890         return rt;
 891 }
 892
 893 static struct rt6_info *rt6_alloc_clone(struct rt6_info *ort,
 894                                         const struct in6_addr *daddr)
 895 {
 896         struct rt6_info *rt = ip6_rt_copy(ort, daddr);
 897
 898         if (rt)
 899                 rt->rt6i_flags |= RTF_CACHE;
 900         return rt;
 901 }
 902
 903 static struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table, int oif,
 904                                       struct flowi6 *fl6, int flags)
 905 {
 906         struct fib6_node *fn;
 907         struct rt6_info *rt, *nrt;
 908         int strict = 0;
 909         int attempts = 3;
 910         int err;
 911         int reachable = net->ipv6.devconf_all->forwarding ? 0 : RT6_LOOKUP_F_REACHABLE;
 912
 913         strict |= flags & RT6_LOOKUP_F_IFACE;
 914
 915 relookup:
 916         read_lock_bh(&table->tb6_lock);
 917
 918 restart_2:
 919         fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
 920
 921 restart:
 922         rt = rt6_select(fn, oif, strict | reachable);
 923         if (rt->rt6i_nsiblings)
 924                 rt = rt6_multipath_select(rt, fl6, oif, strict | reachable);
 925         BACKTRACK(net, &fl6->saddr);
 926         if (rt == net->ipv6.ip6_null_entry ||
 927             rt->rt6i_flags & RTF_CACHE)
 928                 goto out;
 929
 930         dst_hold(&rt->dst);
 931         read_unlock_bh(&table->tb6_lock);
 932
 933         if (!(rt->rt6i_flags & (RTF_NONEXTHOP | RTF_GATEWAY)))
 934                 nrt = rt6_alloc_cow(rt, &fl6->daddr, &fl6->saddr);
 935         else if (!(rt->dst.flags & DST_HOST))
 936                 nrt = rt6_alloc_clone(rt, &fl6->daddr);
 937         else
 938                 goto out2;
 939
 940         ip6_rt_put(rt);
 941         rt = nrt ? : net->ipv6.ip6_null_entry;
 942
 943         dst_hold(&rt->dst);
 944         if (nrt) {
 945                 err = ip6_ins_rt(nrt);
 946                 if (!err)
 947                         goto out2;
 948         }
 949
 950         if (--attempts <= 0)
 951                 goto out2;
 952
 953         /*
 954          * Race condition! In the gap, when table->tb6_lock was
 955          * released someone could insert this route.  Relookup.
 956          */
 957         ip6_rt_put(rt);
 958         goto relookup;
 959
 960 out:
 961         if (reachable) {
 962                 reachable = 0;
 963                 goto restart_2;
 964         }
 965         dst_hold(&rt->dst);
 966         read_unlock_bh(&table->tb6_lock);
 967 out2:
 968         rt->dst.lastuse = jiffies;
 969         rt->dst.__use++;
 970
 971         return rt;
 972 }
 973
 974 static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
 975                                             struct flowi6 *fl6, int flags)
 976 {
 977         return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, flags);
 978 }
 979
 980 static struct dst_entry *ip6_route_input_lookup(struct net *net,
 981                                                 struct net_device *dev,
 982                                                 struct flowi6 *fl6, int flags)
 983 {
 984         if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
 985                 flags |= RT6_LOOKUP_F_IFACE;
 986
 987         return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_input);
 988 }
 989
 990 void ip6_route_input(struct sk_buff *skb)
 991 {
 992         const struct ipv6hdr *iph = ipv6_hdr(skb);
 993         struct net *net = dev_net(skb->dev);
 994         int flags = RT6_LOOKUP_F_HAS_SADDR;
 995         struct flowi6 fl6 = {
 996                 .flowi6_iif = skb->dev->ifindex,
 997                 .daddr = iph->daddr,
 998                 .saddr = iph->saddr,
 999                 .flowlabel = ip6_flowinfo(iph),
1000                 .flowi6_mark = skb->mark,
1001                 .flowi6_proto = iph->nexthdr,
1002         };
1003
1004         skb_dst_set(skb, ip6_route_input_lookup(net, skb->dev, &fl6, flags));
1005 }
1006
1007 static struct rt6_info *ip6_pol_route_output(struct net *net, struct fib6_table *table,
1008                                              struct flowi6 *fl6, int flags)
1009 {
1010         return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, flags);
1011 }
1012
1013 struct dst_entry * ip6_route_output(struct net *net, const struct sock *sk,
1014                                     struct flowi6 *fl6)
1015 {
1016         int flags = 0;
1017
1018         fl6->flowi6_iif = LOOPBACK_IFINDEX;
1019
1020         if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr))
1021                 flags |= RT6_LOOKUP_F_IFACE;
1022
1023         if (!ipv6_addr_any(&fl6->saddr))
1024                 flags |= RT6_LOOKUP_F_HAS_SADDR;
1025         else if (sk)
1026                 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
1027
1028         return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_output);
1029 }
1030
1031 EXPORT_SYMBOL(ip6_route_output);
1032
1033 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
1034 {
1035         struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
1036         struct dst_entry *new = NULL;
1037
1038         rt = dst_alloc(&ip6_dst_blackhole_ops, ort->dst.dev, 1, DST_OBSOLETE_NONE, 0);
1039         if (rt) {
1040                 new = &rt->dst;
1041
1042                 memset(new + 1, 0, sizeof(*rt) - sizeof(*new));
1043                 rt6_init_peer(rt, net->ipv6.peers);
1044
1045                 new->__use = 1;
1046                 new->input = dst_discard;
1047                 new->output = dst_discard;
1048
1049                 if (dst_metrics_read_only(&ort->dst))
1050                         new->_metrics = ort->dst._metrics;
1051                 else
1052                         dst_copy_metrics(new, &ort->dst);
1053                 rt->rt6i_idev = ort->rt6i_idev;
1054                 if (rt->rt6i_idev)
1055                         in6_dev_hold(rt->rt6i_idev);
1056
1057                 rt->rt6i_gateway = ort->rt6i_gateway;
1058                 rt->rt6i_flags = ort->rt6i_flags;
1059                 rt->rt6i_metric = 0;
1060
1061                 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1062 #ifdef CONFIG_IPV6_SUBTREES
1063                 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1064 #endif
1065
1066                 dst_free(new);
1067         }
1068
1069         dst_release(dst_orig);
1070         return new ? new : ERR_PTR(-ENOMEM);
1071 }
1072
1073 /*
1074  *      Destination cache support functions
1075  */
1076
1077 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
1078 {
1079         struct rt6_info *rt;
1080
1081         rt = (struct rt6_info *) dst;
1082
1083         /* All IPV6 dsts are created with ->obsolete set to the value
1084          * DST_OBSOLETE_FORCE_CHK which forces validation calls down
1085          * into this function always.
1086          */
1087         if (rt->rt6i_genid != rt_genid_ipv6(dev_net(rt->dst.dev)))
1088                 return NULL;
1089
1090         if (rt->rt6i_node && (rt->rt6i_node->fn_sernum == cookie))
1091                 return dst;
1092
1093         return NULL;
1094 }
1095
1096 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
1097 {
1098         struct rt6_info *rt = (struct rt6_info *) dst;
1099
1100         if (rt) {
1101                 if (rt->rt6i_flags & RTF_CACHE) {
1102                         if (rt6_check_expired(rt)) {
1103                                 ip6_del_rt(rt);
1104                                 dst = NULL;
1105                         }
1106                 } else {
1107                         dst_release(dst);
1108                         dst = NULL;
1109                 }
1110         }
1111         return dst;
1112 }
1113
1114 static void ip6_link_failure(struct sk_buff *skb)
1115 {
1116         struct rt6_info *rt;
1117
1118         icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
1119
1120         rt = (struct rt6_info *) skb_dst(skb);
1121         if (rt) {
1122                 if (rt->rt6i_flags & RTF_CACHE) {
1123                         dst_hold(&rt->dst);
1124                         if (ip6_del_rt(rt))
1125                                 dst_free(&rt->dst);
1126                 } else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT)) {
1127                         rt->rt6i_node->fn_sernum = -1;
1128                 }
1129         }
1130 }
1131
1132 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
1133                                struct sk_buff *skb, u32 mtu)
1134 {
1135         struct rt6_info *rt6 = (struct rt6_info*)dst;
1136
1137         dst_confirm(dst);
1138         if (mtu < dst_mtu(dst) && rt6->rt6i_dst.plen == 128) {
1139                 struct net *net = dev_net(dst->dev);
1140
1141                 rt6->rt6i_flags |= RTF_MODIFIED;
1142                 if (mtu < IPV6_MIN_MTU) {
1143                         u32 features = dst_metric(dst, RTAX_FEATURES);
1144                         mtu = IPV6_MIN_MTU;
1145                         features |= RTAX_FEATURE_ALLFRAG;
1146                         dst_metric_set(dst, RTAX_FEATURES, features);
1147                 }
1148                 dst_metric_set(dst, RTAX_MTU, mtu);
1149                 rt6_update_expires(rt6, net->ipv6.sysctl.ip6_rt_mtu_expires);
1150         }
1151 }
1152
1153 void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
1154                      int oif, u32 mark)
1155 {
1156         const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1157         struct dst_entry *dst;
1158         struct flowi6 fl6;
1159
1160         memset(&fl6, 0, sizeof(fl6));
1161         fl6.flowi6_oif = oif;
1162         fl6.flowi6_mark = mark;
1163         fl6.flowi6_flags = 0;
1164         fl6.daddr = iph->daddr;
1165         fl6.saddr = iph->saddr;
1166         fl6.flowlabel = ip6_flowinfo(iph);
1167
1168         dst = ip6_route_output(net, NULL, &fl6);
1169         if (!dst->error)
1170                 ip6_rt_update_pmtu(dst, NULL, skb, ntohl(mtu));
1171         dst_release(dst);
1172 }
1173 EXPORT_SYMBOL_GPL(ip6_update_pmtu);
1174
1175 void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
1176 {
1177         ip6_update_pmtu(skb, sock_net(sk), mtu,
1178                         sk->sk_bound_dev_if, sk->sk_mark);
1179 }
1180 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
1181
1182 /* Handle redirects */
1183 struct ip6rd_flowi {
1184         struct flowi6 fl6;
1185         struct in6_addr gateway;
1186 };
1187
1188 static struct rt6_info *__ip6_route_redirect(struct net *net,
1189                                              struct fib6_table *table,
1190                                              struct flowi6 *fl6,
1191                                              int flags)
1192 {
1193         struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
1194         struct rt6_info *rt;
1195         struct fib6_node *fn;
1196
1197         /* Get the "current" route for this destination and
1198          * check if the redirect has come from approriate router.
1199          *
1200          * RFC 4861 specifies that redirects should only be
1201          * accepted if they come from the nexthop to the target.
1202          * Due to the way the routes are chosen, this notion
1203          * is a bit fuzzy and one might need to check all possible
1204          * routes.
1205          */
1206
1207         read_lock_bh(&table->tb6_lock);
1208         fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1209 restart:
1210         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1211                 if (rt6_check_expired(rt))
1212                         continue;
1213                 if (rt->dst.error)
1214                         break;
1215                 if (!(rt->rt6i_flags & RTF_GATEWAY))
1216                         continue;
1217                 if (fl6->flowi6_oif != rt->dst.dev->ifindex)
1218                         continue;
1219                 if (!ipv6_addr_equal(&rdfl->gateway, &rt->rt6i_gateway))
1220                         continue;
1221                 break;
1222         }
1223
1224         if (!rt)
1225                 rt = net->ipv6.ip6_null_entry;
1226         else if (rt->dst.error) {
1227                 rt = net->ipv6.ip6_null_entry;
1228                 goto out;
1229         }
1230         BACKTRACK(net, &fl6->saddr);
1231 out:
1232         dst_hold(&rt->dst);
1233
1234         read_unlock_bh(&table->tb6_lock);
1235
1236         return rt;
1237 };
1238
1239 static struct dst_entry *ip6_route_redirect(struct net *net,
1240                                         const struct flowi6 *fl6,
1241                                         const struct in6_addr *gateway)
1242 {
1243         int flags = RT6_LOOKUP_F_HAS_SADDR;
1244         struct ip6rd_flowi rdfl;
1245
1246         rdfl.fl6 = *fl6;
1247         rdfl.gateway = *gateway;
1248
1249         return fib6_rule_lookup(net, &rdfl.fl6,
1250                                 flags, __ip6_route_redirect);
1251 }
1252
1253 void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark)
1254 {
1255         const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1256         struct dst_entry *dst;
1257         struct flowi6 fl6;
1258
1259         memset(&fl6, 0, sizeof(fl6));
1260         fl6.flowi6_oif = oif;
1261         fl6.flowi6_mark = mark;
1262         fl6.flowi6_flags = 0;
1263         fl6.daddr = iph->daddr;
1264         fl6.saddr = iph->saddr;
1265         fl6.flowlabel = ip6_flowinfo(iph);
1266
1267         dst = ip6_route_redirect(net, &fl6, &ipv6_hdr(skb)->saddr);
1268         rt6_do_redirect(dst, NULL, skb);
1269         dst_release(dst);
1270 }
1271 EXPORT_SYMBOL_GPL(ip6_redirect);
1272
1273 void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif,
1274                             u32 mark)
1275 {
1276         const struct ipv6hdr *iph = ipv6_hdr(skb);
1277         const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
1278         struct dst_entry *dst;
1279         struct flowi6 fl6;
1280
1281         memset(&fl6, 0, sizeof(fl6));
1282         fl6.flowi6_oif = oif;
1283         fl6.flowi6_mark = mark;
1284         fl6.flowi6_flags = 0;
1285         fl6.daddr = msg->dest;
1286         fl6.saddr = iph->daddr;
1287
1288         dst = ip6_route_redirect(net, &fl6, &iph->saddr);
1289         rt6_do_redirect(dst, NULL, skb);
1290         dst_release(dst);
1291 }
1292
1293 void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
1294 {
1295         ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark);
1296 }
1297 EXPORT_SYMBOL_GPL(ip6_sk_redirect);
1298
1299 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1300 {
1301         struct net_device *dev = dst->dev;
1302         unsigned int mtu = dst_mtu(dst);
1303         struct net *net = dev_net(dev);
1304
1305         mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
1306
1307         if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
1308                 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1309
1310         /*
1311          * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1312          * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1313          * IPV6_MAXPLEN is also valid and means: "any MSS,
1314          * rely only on pmtu discovery"
1315          */
1316         if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
1317                 mtu = IPV6_MAXPLEN;
1318         return mtu;
1319 }
1320
1321 static unsigned int ip6_mtu(const struct dst_entry *dst)
1322 {
1323         struct inet6_dev *idev;
1324         unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
1325
1326         if (mtu)
1327                 return mtu;
1328
1329         mtu = IPV6_MIN_MTU;
1330
1331         rcu_read_lock();
1332         idev = __in6_dev_get(dst->dev);
1333         if (idev)
1334                 mtu = idev->cnf.mtu6;
1335         rcu_read_unlock();
1336
1337         return mtu;
1338 }
1339
1340 static struct dst_entry *icmp6_dst_gc_list;
1341 static DEFINE_SPINLOCK(icmp6_dst_lock);
1342
1343 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
1344                                   struct flowi6 *fl6)
1345 {
1346         struct dst_entry *dst;
1347         struct rt6_info *rt;
1348         struct inet6_dev *idev = in6_dev_get(dev);
1349         struct net *net = dev_net(dev);
1350
1351         if (unlikely(!idev))
1352                 return ERR_PTR(-ENODEV);
1353
1354         rt = ip6_dst_alloc(net, dev, 0, NULL);
1355         if (unlikely(!rt)) {
1356                 in6_dev_put(idev);
1357                 dst = ERR_PTR(-ENOMEM);
1358                 goto out;
1359         }
1360
1361         rt->dst.flags |= DST_HOST;
1362         rt->dst.output  = ip6_output;
1363         atomic_set(&rt->dst.__refcnt, 1);
1364         rt->rt6i_gateway  = fl6->daddr;
1365         rt->rt6i_dst.addr = fl6->daddr;
1366         rt->rt6i_dst.plen = 128;
1367         rt->rt6i_idev     = idev;
1368         dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
1369
1370         spin_lock_bh(&icmp6_dst_lock);
1371         rt->dst.next = icmp6_dst_gc_list;
1372         icmp6_dst_gc_list = &rt->dst;
1373         spin_unlock_bh(&icmp6_dst_lock);
1374
1375         fib6_force_start_gc(net);
1376
1377         dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
1378
1379 out:
1380         return dst;
1381 }
1382
1383 int icmp6_dst_gc(void)
1384 {
1385         struct dst_entry *dst, **pprev;
1386         int more = 0;
1387
1388         spin_lock_bh(&icmp6_dst_lock);
1389         pprev = &icmp6_dst_gc_list;
1390
1391         while ((dst = *pprev) != NULL) {
1392                 if (!atomic_read(&dst->__refcnt)) {
1393                         *pprev = dst->next;
1394                         dst_free(dst);
1395                 } else {
1396                         pprev = &dst->next;
1397                         ++more;
1398                 }
1399         }
1400
1401         spin_unlock_bh(&icmp6_dst_lock);
1402
1403         return more;
1404 }
1405
1406 static void icmp6_clean_all(int (*func)(struct rt6_info *rt, void *arg),
1407                             void *arg)
1408 {
1409         struct dst_entry *dst, **pprev;
1410
1411         spin_lock_bh(&icmp6_dst_lock);
1412         pprev = &icmp6_dst_gc_list;
1413         while ((dst = *pprev) != NULL) {
1414                 struct rt6_info *rt = (struct rt6_info *) dst;
1415                 if (func(rt, arg)) {
1416                         *pprev = dst->next;
1417                         dst_free(dst);
1418                 } else {
1419                         pprev = &dst->next;
1420                 }
1421         }
1422         spin_unlock_bh(&icmp6_dst_lock);
1423 }
1424
1425 static int ip6_dst_gc(struct dst_ops *ops)
1426 {
1427         struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
1428         int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1429         int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1430         int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1431         int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1432         unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1433         int entries;
1434
1435         entries = dst_entries_get_fast(ops);
1436         if (time_after(rt_last_gc + rt_min_interval, jiffies) &&
1437             entries <= rt_max_size)
1438                 goto out;
1439
1440         net->ipv6.ip6_rt_gc_expire++;
1441         fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net, entries > rt_max_size);
1442         entries = dst_entries_get_slow(ops);
1443         if (entries < ops->gc_thresh)
1444                 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1445 out:
1446         net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1447         return entries > rt_max_size;
1448 }
1449
1450 /*
1451  *
1452  */
1453
1454 int ip6_route_add(struct fib6_config *cfg)
1455 {
1456         int err;
1457         struct net *net = cfg->fc_nlinfo.nl_net;
1458         struct rt6_info *rt = NULL;
1459         struct net_device *dev = NULL;
1460         struct inet6_dev *idev = NULL;
1461         struct fib6_table *table;
1462         int addr_type;
1463
1464         if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1465                 return -EINVAL;
1466 #ifndef CONFIG_IPV6_SUBTREES
1467         if (cfg->fc_src_len)
1468                 return -EINVAL;
1469 #endif
1470         if (cfg->fc_ifindex) {
1471                 err = -ENODEV;
1472                 dev = dev_get_by_index(net, cfg->fc_ifindex);
1473                 if (!dev)
1474                         goto out;
1475                 idev = in6_dev_get(dev);
1476                 if (!idev)
1477                         goto out;
1478         }
1479
1480         if (cfg->fc_metric == 0)
1481                 cfg->fc_metric = IP6_RT_PRIO_USER;
1482
1483         err = -ENOBUFS;
1484         if (cfg->fc_nlinfo.nlh &&
1485             !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
1486                 table = fib6_get_table(net, cfg->fc_table);
1487                 if (!table) {
1488                         pr_warn("NLM_F_CREATE should be specified when creating new route\n");
1489                         table = fib6_new_table(net, cfg->fc_table);
1490                 }
1491         } else {
1492                 table = fib6_new_table(net, cfg->fc_table);
1493         }
1494
1495         if (!table)
1496                 goto out;
1497
1498         rt = ip6_dst_alloc(net, NULL, DST_NOCOUNT, table);
1499
1500         if (!rt) {
1501                 err = -ENOMEM;
1502                 goto out;
1503         }
1504
1505         if (cfg->fc_flags & RTF_EXPIRES)
1506                 rt6_set_expires(rt, jiffies +
1507                                 clock_t_to_jiffies(cfg->fc_expires));
1508         else
1509                 rt6_clean_expires(rt);
1510
1511         if (cfg->fc_protocol == RTPROT_UNSPEC)
1512                 cfg->fc_protocol = RTPROT_BOOT;
1513         rt->rt6i_protocol = cfg->fc_protocol;
1514
1515         addr_type = ipv6_addr_type(&cfg->fc_dst);
1516
1517         if (addr_type & IPV6_ADDR_MULTICAST)
1518                 rt->dst.input = ip6_mc_input;
1519         else if (cfg->fc_flags & RTF_LOCAL)
1520                 rt->dst.input = ip6_input;
1521         else
1522                 rt->dst.input = ip6_forward;
1523
1524         rt->dst.output = ip6_output;
1525
1526         ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1527         rt->rt6i_dst.plen = cfg->fc_dst_len;
1528         if (rt->rt6i_dst.plen == 128)
1529                rt->dst.flags |= DST_HOST;
1530
1531         if (!(rt->dst.flags & DST_HOST) && cfg->fc_mx) {
1532                 u32 *metrics = kzalloc(sizeof(u32) * RTAX_MAX, GFP_KERNEL);
1533                 if (!metrics) {
1534                         err = -ENOMEM;
1535                         goto out;
1536                 }
1537                 dst_init_metrics(&rt->dst, metrics, 0);
1538         }
1539 #ifdef CONFIG_IPV6_SUBTREES
1540         ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1541         rt->rt6i_src.plen = cfg->fc_src_len;
1542 #endif
1543
1544         rt->rt6i_metric = cfg->fc_metric;
1545
1546         /* We cannot add true routes via loopback here,
1547            they would result in kernel looping; promote them to reject routes
1548          */
1549         if ((cfg->fc_flags & RTF_REJECT) ||
1550             (dev && (dev->flags & IFF_LOOPBACK) &&
1551              !(addr_type & IPV6_ADDR_LOOPBACK) &&
1552              !(cfg->fc_flags & RTF_LOCAL))) {
1553                 /* hold loopback dev/idev if we haven't done so. */
1554                 if (dev != net->loopback_dev) {
1555                         if (dev) {
1556                                 dev_put(dev);
1557                                 in6_dev_put(idev);
1558                         }
1559                         dev = net->loopback_dev;
1560                         dev_hold(dev);
1561                         idev = in6_dev_get(dev);
1562                         if (!idev) {
1563                                 err = -ENODEV;
1564                                 goto out;
1565                         }
1566                 }
1567                 rt->dst.output = ip6_pkt_discard_out;
1568                 rt->dst.input = ip6_pkt_discard;
1569                 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1570                 switch (cfg->fc_type) {
1571                 case RTN_BLACKHOLE:
1572                         rt->dst.error = -EINVAL;
1573                         break;
1574                 case RTN_PROHIBIT:
1575                         rt->dst.error = -EACCES;
1576                         break;
1577                 case RTN_THROW:
1578                         rt->dst.error = -EAGAIN;
1579                         break;
1580                 default:
1581                         rt->dst.error = -ENETUNREACH;
1582                         break;
1583                 }
1584                 goto install_route;
1585         }
1586
1587         if (cfg->fc_flags & RTF_GATEWAY) {
1588                 const struct in6_addr *gw_addr;
1589                 int gwa_type;
1590
1591                 gw_addr = &cfg->fc_gateway;
1592                 rt->rt6i_gateway = *gw_addr;
1593                 gwa_type = ipv6_addr_type(gw_addr);
1594
1595                 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1596                         struct rt6_info *grt;
1597
1598                         /* IPv6 strictly inhibits using not link-local
1599                            addresses as nexthop address.
1600                            Otherwise, router will not able to send redirects.
1601                            It is very good, but in some (rare!) circumstances
1602                            (SIT, PtP, NBMA NOARP links) it is handy to allow
1603                            some exceptions. --ANK
1604                          */
1605                         err = -EINVAL;
1606                         if (!(gwa_type & IPV6_ADDR_UNICAST))
1607                                 goto out;
1608
1609                         grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, 1);
1610
1611                         err = -EHOSTUNREACH;
1612                         if (!grt)
1613                                 goto out;
1614                         if (dev) {
1615                                 if (dev != grt->dst.dev) {
1616                                         ip6_rt_put(grt);
1617                                         goto out;
1618                                 }
1619                         } else {
1620                                 dev = grt->dst.dev;
1621                                 idev = grt->rt6i_idev;
1622                                 dev_hold(dev);
1623                                 in6_dev_hold(grt->rt6i_idev);
1624                         }
1625                         if (!(grt->rt6i_flags & RTF_GATEWAY))
1626                                 err = 0;
1627                         ip6_rt_put(grt);
1628
1629                         if (err)
1630                                 goto out;
1631                 }
1632                 err = -EINVAL;
1633                 if (!dev || (dev->flags & IFF_LOOPBACK))
1634                         goto out;
1635         }
1636
1637         err = -ENODEV;
1638         if (!dev)
1639                 goto out;
1640
1641         if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
1642                 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
1643                         err = -EINVAL;
1644                         goto out;
1645                 }
1646                 rt->rt6i_prefsrc.addr = cfg->fc_prefsrc;
1647                 rt->rt6i_prefsrc.plen = 128;
1648         } else
1649                 rt->rt6i_prefsrc.plen = 0;
1650
1651         rt->rt6i_flags = cfg->fc_flags;
1652
1653 install_route:
1654         if (cfg->fc_mx) {
1655                 struct nlattr *nla;
1656                 int remaining;
1657
1658                 nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1659                         int type = nla_type(nla);
1660
1661                         if (type) {
1662                                 if (type > RTAX_MAX) {
1663                                         err = -EINVAL;
1664                                         goto out;
1665                                 }
1666
1667                                 dst_metric_set(&rt->dst, type, nla_get_u32(nla));
1668                         }
1669                 }
1670         }
1671
1672         rt->dst.dev = dev;
1673         rt->rt6i_idev = idev;
1674         rt->rt6i_table = table;
1675
1676         cfg->fc_nlinfo.nl_net = dev_net(dev);
1677
1678         return __ip6_ins_rt(rt, &cfg->fc_nlinfo);
1679
1680 out:
1681         if (dev)
1682                 dev_put(dev);
1683         if (idev)
1684                 in6_dev_put(idev);
1685         if (rt)
1686                 dst_free(&rt->dst);
1687         return err;
1688 }
1689
1690 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
1691 {
1692         int err;
1693         struct fib6_table *table;
1694         struct net *net = dev_net(rt->dst.dev);
1695
1696         if (rt == net->ipv6.ip6_null_entry) {
1697                 err = -ENOENT;
1698                 goto out;
1699         }
1700
1701         table = rt->rt6i_table;
1702         write_lock_bh(&table->tb6_lock);
1703         err = fib6_del(rt, info);
1704         write_unlock_bh(&table->tb6_lock);
1705
1706 out:
1707         ip6_rt_put(rt);
1708         return err;
1709 }
1710
1711 int ip6_del_rt(struct rt6_info *rt)
1712 {
1713         struct nl_info info = {
1714                 .nl_net = dev_net(rt->dst.dev),
1715         };
1716         return __ip6_del_rt(rt, &info);
1717 }
1718
1719 static int ip6_route_del(struct fib6_config *cfg)
1720 {
1721         struct fib6_table *table;
1722         struct fib6_node *fn;
1723         struct rt6_info *rt;
1724         int err = -ESRCH;
1725
1726         table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
1727         if (!table)
1728                 return err;
1729
1730         read_lock_bh(&table->tb6_lock);
1731
1732         fn = fib6_locate(&table->tb6_root,
1733                          &cfg->fc_dst, cfg->fc_dst_len,
1734                          &cfg->fc_src, cfg->fc_src_len);
1735
1736         if (fn) {
1737                 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1738                         if (cfg->fc_ifindex &&
1739                             (!rt->dst.dev ||
1740                              rt->dst.dev->ifindex != cfg->fc_ifindex))
1741                                 continue;
1742                         if (cfg->fc_flags & RTF_GATEWAY &&
1743                             !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
1744                                 continue;
1745                         if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
1746                                 continue;
1747                         dst_hold(&rt->dst);
1748                         read_unlock_bh(&table->tb6_lock);
1749
1750                         return __ip6_del_rt(rt, &cfg->fc_nlinfo);
1751                 }
1752         }
1753         read_unlock_bh(&table->tb6_lock);
1754
1755         return err;
1756 }
1757
1758 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
1759 {
1760         struct net *net = dev_net(skb->dev);
1761         struct netevent_redirect netevent;
1762         struct rt6_info *rt, *nrt = NULL;
1763         struct ndisc_options ndopts;
1764         struct inet6_dev *in6_dev;
1765         struct neighbour *neigh;
1766         struct rd_msg *msg;
1767         int optlen, on_link;
1768         u8 *lladdr;
1769
1770         optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
1771         optlen -= sizeof(*msg);
1772
1773         if (optlen < 0) {
1774                 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
1775                 return;
1776         }
1777
1778         msg = (struct rd_msg *)icmp6_hdr(skb);
1779
1780         if (ipv6_addr_is_multicast(&msg->dest)) {
1781                 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
1782                 return;
1783         }
1784
1785         on_link = 0;
1786         if (ipv6_addr_equal(&msg->dest, &msg->target)) {
1787                 on_link = 1;
1788         } else if (ipv6_addr_type(&msg->target) !=
1789                    (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
1790                 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
1791                 return;
1792         }
1793
1794         in6_dev = __in6_dev_get(skb->dev);
1795         if (!in6_dev)
1796                 return;
1797         if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
1798                 return;
1799
1800         /* RFC2461 8.1:
1801          *      The IP source address of the Redirect MUST be the same as the current
1802          *      first-hop router for the specified ICMP Destination Address.
1803          */
1804
1805         if (!ndisc_parse_options(msg->opt, optlen, &ndopts)) {
1806                 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
1807                 return;
1808         }
1809
1810         lladdr = NULL;
1811         if (ndopts.nd_opts_tgt_lladdr) {
1812                 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
1813                                              skb->dev);
1814                 if (!lladdr) {
1815                         net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
1816                         return;
1817                 }
1818         }
1819
1820         rt = (struct rt6_info *) dst;
1821         if (rt == net->ipv6.ip6_null_entry) {
1822                 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
1823                 return;
1824         }
1825
1826         /* Redirect received -> path was valid.
1827          * Look, redirects are sent only in response to data packets,
1828          * so that this nexthop apparently is reachable. --ANK
1829          */
1830         dst_confirm(&rt->dst);
1831
1832         neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
1833         if (!neigh)
1834                 return;
1835
1836         /*
1837          *      We have finally decided to accept it.
1838          */
1839
1840         neigh_update(neigh, lladdr, NUD_STALE,
1841                      NEIGH_UPDATE_F_WEAK_OVERRIDE|
1842                      NEIGH_UPDATE_F_OVERRIDE|
1843                      (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1844                                      NEIGH_UPDATE_F_ISROUTER))
1845                      );
1846
1847         nrt = ip6_rt_copy(rt, &msg->dest);
1848         if (!nrt)
1849                 goto out;
1850
1851         nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
1852         if (on_link)
1853                 nrt->rt6i_flags &= ~RTF_GATEWAY;
1854
1855         nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
1856
1857         if (ip6_ins_rt(nrt))
1858                 goto out;
1859
1860         netevent.old = &rt->dst;
1861         netevent.new = &nrt->dst;
1862         netevent.daddr = &msg->dest;
1863         netevent.neigh = neigh;
1864         call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
1865
1866         if (rt->rt6i_flags & RTF_CACHE) {
1867                 rt = (struct rt6_info *) dst_clone(&rt->dst);
1868                 ip6_del_rt(rt);
1869         }
1870
1871 out:
1872         neigh_release(neigh);
1873 }
1874
1875 /*
1876  *      Misc support functions
1877  */
1878
1879 static struct rt6_info *ip6_rt_copy(struct rt6_info *ort,
1880                                     const struct in6_addr *dest)
1881 {
1882         struct net *net = dev_net(ort->dst.dev);
1883         struct rt6_info *rt = ip6_dst_alloc(net, ort->dst.dev, 0,
1884                                             ort->rt6i_table);
1885
1886         if (rt) {
1887                 rt->dst.input = ort->dst.input;
1888                 rt->dst.output = ort->dst.output;
1889                 rt->dst.flags |= DST_HOST;
1890
1891                 rt->rt6i_dst.addr = *dest;
1892                 rt->rt6i_dst.plen = 128;
1893                 dst_copy_metrics(&rt->dst, &ort->dst);
1894                 rt->dst.error = ort->dst.error;
1895                 rt->rt6i_idev = ort->rt6i_idev;
1896                 if (rt->rt6i_idev)
1897                         in6_dev_hold(rt->rt6i_idev);
1898                 rt->dst.lastuse = jiffies;
1899
1900                 if (ort->rt6i_flags & RTF_GATEWAY)
1901                         rt->rt6i_gateway = ort->rt6i_gateway;
1902                 else
1903                         rt->rt6i_gateway = *dest;
1904                 rt->rt6i_flags = ort->rt6i_flags;
1905                 if ((ort->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF)) ==
1906                     (RTF_DEFAULT | RTF_ADDRCONF))
1907                         rt6_set_from(rt, ort);
1908                 rt->rt6i_metric = 0;
1909
1910 #ifdef CONFIG_IPV6_SUBTREES
1911                 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1912 #endif
1913                 memcpy(&rt->rt6i_prefsrc, &ort->rt6i_prefsrc, sizeof(struct rt6key));
1914                 rt->rt6i_table = ort->rt6i_table;
1915         }
1916         return rt;
1917 }
1918
1919 #ifdef CONFIG_IPV6_ROUTE_INFO
1920 static struct rt6_info *rt6_get_route_info(struct net *net,
1921                                            const struct in6_addr *prefix, int prefixlen,
1922                                            const struct in6_addr *gwaddr, int ifindex)
1923 {
1924         struct fib6_node *fn;
1925         struct rt6_info *rt = NULL;
1926         struct fib6_table *table;
1927
1928         table = fib6_get_table(net, RT6_TABLE_INFO);
1929         if (!table)
1930                 return NULL;
1931
1932         read_lock_bh(&table->tb6_lock);
1933         fn = fib6_locate(&table->tb6_root, prefix ,prefixlen, NULL, 0);
1934         if (!fn)
1935                 goto out;
1936
1937         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1938                 if (rt->dst.dev->ifindex != ifindex)
1939                         continue;
1940                 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
1941                         continue;
1942                 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
1943                         continue;
1944                 dst_hold(&rt->dst);
1945                 break;
1946         }
1947 out:
1948         read_unlock_bh(&table->tb6_lock);
1949         return rt;
1950 }
1951
1952 static struct rt6_info *rt6_add_route_info(struct net *net,
1953                                            const struct in6_addr *prefix, int prefixlen,
1954                                            const struct in6_addr *gwaddr, int ifindex,
1955                                            unsigned int pref)
1956 {
1957         struct fib6_config cfg = {
1958                 .fc_table       = RT6_TABLE_INFO,
1959                 .fc_metric      = IP6_RT_PRIO_USER,
1960                 .fc_ifindex     = ifindex,
1961                 .fc_dst_len     = prefixlen,
1962                 .fc_flags       = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
1963                                   RTF_UP | RTF_PREF(pref),
1964                 .fc_nlinfo.portid = 0,
1965                 .fc_nlinfo.nlh = NULL,
1966                 .fc_nlinfo.nl_net = net,
1967         };
1968
1969         cfg.fc_dst = *prefix;
1970         cfg.fc_gateway = *gwaddr;
1971
1972         /* We should treat it as a default route if prefix length is 0. */
1973         if (!prefixlen)
1974                 cfg.fc_flags |= RTF_DEFAULT;
1975
1976         ip6_route_add(&cfg);
1977
1978         return rt6_get_route_info(net, prefix, prefixlen, gwaddr, ifindex);
1979 }
1980 #endif
1981
1982 struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
1983 {
1984         struct rt6_info *rt;
1985         struct fib6_table *table;
1986
1987         table = fib6_get_table(dev_net(dev), RT6_TABLE_DFLT);
1988         if (!table)
1989                 return NULL;
1990
1991         read_lock_bh(&table->tb6_lock);
1992         for (rt = table->tb6_root.leaf; rt; rt=rt->dst.rt6_next) {
1993                 if (dev == rt->dst.dev &&
1994                     ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
1995                     ipv6_addr_equal(&rt->rt6i_gateway, addr))
1996                         break;
1997         }
1998         if (rt)
1999                 dst_hold(&rt->dst);
2000         read_unlock_bh(&table->tb6_lock);
2001         return rt;
2002 }
2003
2004 struct rt6_info *rt6_add_dflt_router(const struct in6_addr *gwaddr,
2005                                      struct net_device *dev,
2006                                      unsigned int pref)
2007 {
2008         struct fib6_config cfg = {
2009                 .fc_table       = RT6_TABLE_DFLT,
2010                 .fc_metric      = IP6_RT_PRIO_USER,
2011                 .fc_ifindex     = dev->ifindex,
2012                 .fc_flags       = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
2013                                   RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
2014                 .fc_nlinfo.portid = 0,
2015                 .fc_nlinfo.nlh = NULL,
2016                 .fc_nlinfo.nl_net = dev_net(dev),
2017         };
2018
2019         cfg.fc_gateway = *gwaddr;
2020
2021         ip6_route_add(&cfg);
2022
2023         return rt6_get_dflt_router(gwaddr, dev);
2024 }
2025
2026 void rt6_purge_dflt_routers(struct net *net)
2027 {
2028         struct rt6_info *rt;
2029         struct fib6_table *table;
2030
2031         /* NOTE: Keep consistent with rt6_get_dflt_router */
2032         table = fib6_get_table(net, RT6_TABLE_DFLT);
2033         if (!table)
2034                 return;
2035
2036 restart:
2037         read_lock_bh(&table->tb6_lock);
2038         for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2039                 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
2040                     (!rt->rt6i_idev || rt->rt6i_idev->cnf.accept_ra != 2)) {
2041                         dst_hold(&rt->dst);
2042                         read_unlock_bh(&table->tb6_lock);
2043                         ip6_del_rt(rt);
2044                         goto restart;
2045                 }
2046         }
2047         read_unlock_bh(&table->tb6_lock);
2048 }
2049
2050 static void rtmsg_to_fib6_config(struct net *net,
2051                                  struct in6_rtmsg *rtmsg,
2052                                  struct fib6_config *cfg)
2053 {
2054         memset(cfg, 0, sizeof(*cfg));
2055
2056         cfg->fc_table = RT6_TABLE_MAIN;
2057         cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
2058         cfg->fc_metric = rtmsg->rtmsg_metric;
2059         cfg->fc_expires = rtmsg->rtmsg_info;
2060         cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
2061         cfg->fc_src_len = rtmsg->rtmsg_src_len;
2062         cfg->fc_flags = rtmsg->rtmsg_flags;
2063
2064         cfg->fc_nlinfo.nl_net = net;
2065
2066         cfg->fc_dst = rtmsg->rtmsg_dst;
2067         cfg->fc_src = rtmsg->rtmsg_src;
2068         cfg->fc_gateway = rtmsg->rtmsg_gateway;
2069 }
2070
2071 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
2072 {
2073         struct fib6_config cfg;
2074         struct in6_rtmsg rtmsg;
2075         int err;
2076
2077         switch(cmd) {
2078         case SIOCADDRT:         /* Add a route */
2079         case SIOCDELRT:         /* Delete a route */
2080                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2081                         return -EPERM;
2082                 err = copy_from_user(&rtmsg, arg,
2083                                      sizeof(struct in6_rtmsg));
2084                 if (err)
2085                         return -EFAULT;
2086
2087                 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
2088
2089                 rtnl_lock();
2090                 switch (cmd) {
2091                 case SIOCADDRT:
2092                         err = ip6_route_add(&cfg);
2093                         break;
2094                 case SIOCDELRT:
2095                         err = ip6_route_del(&cfg);
2096                         break;
2097                 default:
2098                         err = -EINVAL;
2099                 }
2100                 rtnl_unlock();
2101
2102                 return err;
2103         }
2104
2105         return -EINVAL;
2106 }
2107
2108 /*
2109  *      Drop the packet on the floor
2110  */
2111
2112 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
2113 {
2114         int type;
2115         struct dst_entry *dst = skb_dst(skb);
2116         switch (ipstats_mib_noroutes) {
2117         case IPSTATS_MIB_INNOROUTES:
2118                 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
2119                 if (type == IPV6_ADDR_ANY) {
2120                         IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2121                                       IPSTATS_MIB_INADDRERRORS);
2122                         break;
2123                 }
2124                 /* FALLTHROUGH */
2125         case IPSTATS_MIB_OUTNOROUTES:
2126                 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2127                               ipstats_mib_noroutes);
2128                 break;
2129         }
2130         icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
2131         kfree_skb(skb);
2132         return 0;
2133 }
2134
2135 static int ip6_pkt_discard(struct sk_buff *skb)
2136 {
2137         return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
2138 }
2139
2140 static int ip6_pkt_discard_out(struct sk_buff *skb)
2141 {
2142         skb->dev = skb_dst(skb)->dev;
2143         return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
2144 }
2145
2146 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2147
2148 static int ip6_pkt_prohibit(struct sk_buff *skb)
2149 {
2150         return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
2151 }
2152
2153 static int ip6_pkt_prohibit_out(struct sk_buff *skb)
2154 {
2155         skb->dev = skb_dst(skb)->dev;
2156         return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
2157 }
2158
2159 #endif
2160
2161 /*
2162  *      Allocate a dst for local (unicast / anycast) address.
2163  */
2164
2165 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
2166                                     const struct in6_addr *addr,
2167                                     bool anycast)
2168 {
2169         struct net *net = dev_net(idev->dev);
2170         struct rt6_info *rt = ip6_dst_alloc(net, net->loopback_dev, 0, NULL);
2171
2172         if (!rt) {
2173                 net_warn_ratelimited("Maximum number of routes reached, consider increasing route/max_size\n");
2174                 return ERR_PTR(-ENOMEM);
2175         }
2176
2177         in6_dev_hold(idev);
2178
2179         rt->dst.flags |= DST_HOST;
2180         rt->dst.input = ip6_input;
2181         rt->dst.output = ip6_output;
2182         rt->rt6i_idev = idev;
2183
2184         rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
2185         if (anycast)
2186                 rt->rt6i_flags |= RTF_ANYCAST;
2187         else
2188                 rt->rt6i_flags |= RTF_LOCAL;
2189
2190         rt->rt6i_gateway  = *addr;
2191         rt->rt6i_dst.addr = *addr;
2192         rt->rt6i_dst.plen = 128;
2193         rt->rt6i_table = fib6_get_table(net, RT6_TABLE_LOCAL);
2194
2195         atomic_set(&rt->dst.__refcnt, 1);
2196
2197         return rt;
2198 }
2199
2200 int ip6_route_get_saddr(struct net *net,
2201                         struct rt6_info *rt,
2202                         const struct in6_addr *daddr,
2203                         unsigned int prefs,
2204                         struct in6_addr *saddr)
2205 {
2206         struct inet6_dev *idev = ip6_dst_idev((struct dst_entry*)rt);
2207         int err = 0;
2208         if (rt->rt6i_prefsrc.plen)
2209                 *saddr = rt->rt6i_prefsrc.addr;
2210         else
2211                 err = ipv6_dev_get_saddr(net, idev ? idev->dev : NULL,
2212                                          daddr, prefs, saddr);
2213         return err;
2214 }
2215
2216 /* remove deleted ip from prefsrc entries */
2217 struct arg_dev_net_ip {
2218         struct net_device *dev;
2219         struct net *net;
2220         struct in6_addr *addr;
2221 };
2222
2223 static int fib6_remove_prefsrc(struct rt6_info *rt, void *arg)
2224 {
2225         struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
2226         struct net *net = ((struct arg_dev_net_ip *)arg)->net;
2227         struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
2228
2229         if (((void *)rt->dst.dev == dev || !dev) &&
2230             rt != net->ipv6.ip6_null_entry &&
2231             ipv6_addr_equal(addr, &rt->rt6i_prefsrc.addr)) {
2232                 /* remove prefsrc entry */
2233                 rt->rt6i_prefsrc.plen = 0;
2234         }
2235         return 0;
2236 }
2237
2238 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
2239 {
2240         struct net *net = dev_net(ifp->idev->dev);
2241         struct arg_dev_net_ip adni = {
2242                 .dev = ifp->idev->dev,
2243                 .net = net,
2244                 .addr = &ifp->addr,
2245         };
2246         fib6_clean_all(net, fib6_remove_prefsrc, 0, &adni);
2247 }
2248
2249 struct arg_dev_net {
2250         struct net_device *dev;
2251         struct net *net;
2252 };
2253
2254 static int fib6_ifdown(struct rt6_info *rt, void *arg)
2255 {
2256         const struct arg_dev_net *adn = arg;
2257         const struct net_device *dev = adn->dev;
2258
2259         if ((rt->dst.dev == dev || !dev) &&
2260             rt != adn->net->ipv6.ip6_null_entry)
2261                 return -1;
2262
2263         return 0;
2264 }
2265
2266 void rt6_ifdown(struct net *net, struct net_device *dev)
2267 {
2268         struct arg_dev_net adn = {
2269                 .dev = dev,
2270                 .net = net,
2271         };
2272
2273         fib6_clean_all(net, fib6_ifdown, 0, &adn);
2274         icmp6_clean_all(fib6_ifdown, &adn);
2275 }
2276
2277 struct rt6_mtu_change_arg {
2278         struct net_device *dev;
2279         unsigned int mtu;
2280 };
2281
2282 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2283 {
2284         struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2285         struct inet6_dev *idev;
2286
2287         /* In IPv6 pmtu discovery is not optional,
2288            so that RTAX_MTU lock cannot disable it.
2289            We still use this lock to block changes
2290            caused by addrconf/ndisc.
2291         */
2292
2293         idev = __in6_dev_get(arg->dev);
2294         if (!idev)
2295                 return 0;
2296
2297         /* For administrative MTU increase, there is no way to discover
2298            IPv6 PMTU increase, so PMTU increase should be updated here.
2299            Since RFC 1981 doesn't include administrative MTU increase
2300            update PMTU increase is a MUST. (i.e. jumbo frame)
2301          */
2302         /*
2303            If new MTU is less than route PMTU, this new MTU will be the
2304            lowest MTU in the path, update the route PMTU to reflect PMTU
2305            decreases; if new MTU is greater than route PMTU, and the
2306            old MTU is the lowest MTU in the path, update the route PMTU
2307            to reflect the increase. In this case if the other nodes' MTU
2308            also have the lowest MTU, TOO BIG MESSAGE will be lead to
2309            PMTU discouvery.
2310          */
2311         if (rt->dst.dev == arg->dev &&
2312             !dst_metric_locked(&rt->dst, RTAX_MTU) &&
2313             (dst_mtu(&rt->dst) >= arg->mtu ||
2314              (dst_mtu(&rt->dst) < arg->mtu &&
2315               dst_mtu(&rt->dst) == idev->cnf.mtu6))) {
2316                 dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
2317         }
2318         return 0;
2319 }
2320
2321 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
2322 {
2323         struct rt6_mtu_change_arg arg = {
2324                 .dev = dev,
2325                 .mtu = mtu,
2326         };
2327
2328         fib6_clean_all(dev_net(dev), rt6_mtu_change_route, 0, &arg);
2329 }
2330
2331 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2332         [RTA_GATEWAY]           = { .len = sizeof(struct in6_addr) },
2333         [RTA_OIF]               = { .type = NLA_U32 },
2334         [RTA_IIF]               = { .type = NLA_U32 },
2335         [RTA_PRIORITY]          = { .type = NLA_U32 },
2336         [RTA_METRICS]           = { .type = NLA_NESTED },
2337         [RTA_MULTIPATH]         = { .len = sizeof(struct rtnexthop) },
2338 };
2339
2340 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2341                               struct fib6_config *cfg)
2342 {
2343         struct rtmsg *rtm;
2344         struct nlattr *tb[RTA_MAX+1];
2345         int err;
2346
2347         err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2348         if (err < 0)
2349                 goto errout;
2350
2351         err = -EINVAL;
2352         rtm = nlmsg_data(nlh);
2353         memset(cfg, 0, sizeof(*cfg));
2354
2355         cfg->fc_table = rtm->rtm_table;
2356         cfg->fc_dst_len = rtm->rtm_dst_len;
2357         cfg->fc_src_len = rtm->rtm_src_len;
2358         cfg->fc_flags = RTF_UP;
2359         cfg->fc_protocol = rtm->rtm_protocol;
2360         cfg->fc_type = rtm->rtm_type;
2361
2362         if (rtm->rtm_type == RTN_UNREACHABLE ||
2363             rtm->rtm_type == RTN_BLACKHOLE ||
2364             rtm->rtm_type == RTN_PROHIBIT ||
2365             rtm->rtm_type == RTN_THROW)
2366                 cfg->fc_flags |= RTF_REJECT;
2367
2368         if (rtm->rtm_type == RTN_LOCAL)
2369                 cfg->fc_flags |= RTF_LOCAL;
2370
2371         cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid;
2372         cfg->fc_nlinfo.nlh = nlh;
2373         cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2374
2375         if (tb[RTA_GATEWAY]) {
2376                 nla_memcpy(&cfg->fc_gateway, tb[RTA_GATEWAY], 16);
2377                 cfg->fc_flags |= RTF_GATEWAY;
2378         }
2379
2380         if (tb[RTA_DST]) {
2381                 int plen = (rtm->rtm_dst_len + 7) >> 3;
2382
2383                 if (nla_len(tb[RTA_DST]) < plen)
2384                         goto errout;
2385
2386                 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2387         }
2388
2389         if (tb[RTA_SRC]) {
2390                 int plen = (rtm->rtm_src_len + 7) >> 3;
2391
2392                 if (nla_len(tb[RTA_SRC]) < plen)
2393                         goto errout;
2394
2395                 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2396         }
2397
2398         if (tb[RTA_PREFSRC])
2399                 nla_memcpy(&cfg->fc_prefsrc, tb[RTA_PREFSRC], 16);
2400
2401         if (tb[RTA_OIF])
2402                 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2403
2404         if (tb[RTA_PRIORITY])
2405                 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2406
2407         if (tb[RTA_METRICS]) {
2408                 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2409                 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2410         }
2411
2412         if (tb[RTA_TABLE])
2413                 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2414
2415         if (tb[RTA_MULTIPATH]) {
2416                 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
2417                 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
2418         }
2419
2420         err = 0;
2421 errout:
2422         return err;
2423 }
2424
2425 static int ip6_route_multipath(struct fib6_config *cfg, int add)
2426 {
2427         struct fib6_config r_cfg;
2428         struct rtnexthop *rtnh;
2429         int remaining;
2430         int attrlen;
2431         int err = 0, last_err = 0;
2432
2433 beginning:
2434         rtnh = (struct rtnexthop *)cfg->fc_mp;
2435         remaining = cfg->fc_mp_len;
2436
2437         /* Parse a Multipath Entry */
2438         while (rtnh_ok(rtnh, remaining)) {
2439                 memcpy(&r_cfg, cfg, sizeof(*cfg));
2440                 if (rtnh->rtnh_ifindex)
2441                         r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
2442
2443                 attrlen = rtnh_attrlen(rtnh);
2444                 if (attrlen > 0) {
2445                         struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
2446
2447                         nla = nla_find(attrs, attrlen, RTA_GATEWAY);
2448                         if (nla) {
2449                                 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
2450                                 r_cfg.fc_flags |= RTF_GATEWAY;
2451                         }
2452                 }
2453                 err = add ? ip6_route_add(&r_cfg) : ip6_route_del(&r_cfg);
2454                 if (err) {
2455                         last_err = err;
2456                         /* If we are trying to remove a route, do not stop the
2457                          * loop when ip6_route_del() fails (because next hop is
2458                          * already gone), we should try to remove all next hops.
2459                          */
2460                         if (add) {
2461                                 /* If add fails, we should try to delete all
2462                                  * next hops that have been already added.
2463                                  */
2464                                 add = 0;
2465                                 goto beginning;
2466                         }
2467                 }
2468                 /* Because each route is added like a single route we remove
2469                  * this flag after the first nexthop (if there is a collision,
2470                  * we have already fail to add the first nexthop:
2471                  * fib6_add_rt2node() has reject it).
2472                  */
2473                 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~NLM_F_EXCL;
2474                 rtnh = rtnh_next(rtnh, &remaining);
2475         }
2476
2477         return last_err;
2478 }
2479
2480 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr* nlh)
2481 {
2482         struct fib6_config cfg;
2483         int err;
2484
2485         err = rtm_to_fib6_config(skb, nlh, &cfg);
2486         if (err < 0)
2487                 return err;
2488
2489         if (cfg.fc_mp)
2490                 return ip6_route_multipath(&cfg, 0);
2491         else
2492                 return ip6_route_del(&cfg);
2493 }
2494
2495 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr* nlh)
2496 {
2497         struct fib6_config cfg;
2498         int err;
2499
2500         err = rtm_to_fib6_config(skb, nlh, &cfg);
2501         if (err < 0)
2502                 return err;
2503
2504         if (cfg.fc_mp)
2505                 return ip6_route_multipath(&cfg, 1);
2506         else
2507                 return ip6_route_add(&cfg);
2508 }
2509
2510 static inline size_t rt6_nlmsg_size(void)
2511 {
2512         return NLMSG_ALIGN(sizeof(struct rtmsg))
2513                + nla_total_size(16) /* RTA_SRC */
2514                + nla_total_size(16) /* RTA_DST */
2515                + nla_total_size(16) /* RTA_GATEWAY */
2516                + nla_total_size(16) /* RTA_PREFSRC */
2517                + nla_total_size(4) /* RTA_TABLE */
2518                + nla_total_size(4) /* RTA_IIF */
2519                + nla_total_size(4) /* RTA_OIF */
2520                + nla_total_size(4) /* RTA_PRIORITY */
2521                + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
2522                + nla_total_size(sizeof(struct rta_cacheinfo));
2523 }
2524
2525 static int rt6_fill_node(struct net *net,
2526                          struct sk_buff *skb, struct rt6_info *rt,
2527                          struct in6_addr *dst, struct in6_addr *src,
2528                          int iif, int type, u32 portid, u32 seq,
2529                          int prefix, int nowait, unsigned int flags)
2530 {
2531         struct rtmsg *rtm;
2532         struct nlmsghdr *nlh;
2533         long expires;
2534         u32 table;
2535
2536         if (prefix) {   /* user wants prefix routes only */
2537                 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
2538                         /* success since this is not a prefix route */
2539                         return 1;
2540                 }
2541         }
2542
2543         nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
2544         if (!nlh)
2545                 return -EMSGSIZE;
2546
2547         rtm = nlmsg_data(nlh);
2548         rtm->rtm_family = AF_INET6;
2549         rtm->rtm_dst_len = rt->rt6i_dst.plen;
2550         rtm->rtm_src_len = rt->rt6i_src.plen;
2551         rtm->rtm_tos = 0;
2552         if (rt->rt6i_table)
2553                 table = rt->rt6i_table->tb6_id;
2554         else
2555                 table = RT6_TABLE_UNSPEC;
2556         rtm->rtm_table = table;
2557         if (nla_put_u32(skb, RTA_TABLE, table))
2558                 goto nla_put_failure;
2559         if (rt->rt6i_flags & RTF_REJECT) {
2560                 switch (rt->dst.error) {
2561                 case -EINVAL:
2562                         rtm->rtm_type = RTN_BLACKHOLE;
2563                         break;
2564                 case -EACCES:
2565                         rtm->rtm_type = RTN_PROHIBIT;
2566                         break;
2567                 case -EAGAIN:
2568                         rtm->rtm_type = RTN_THROW;
2569                         break;
2570                 default:
2571                         rtm->rtm_type = RTN_UNREACHABLE;
2572                         break;
2573                 }
2574         }
2575         else if (rt->rt6i_flags & RTF_LOCAL)
2576                 rtm->rtm_type = RTN_LOCAL;
2577         else if (rt->dst.dev && (rt->dst.dev->flags & IFF_LOOPBACK))
2578                 rtm->rtm_type = RTN_LOCAL;
2579         else
2580                 rtm->rtm_type = RTN_UNICAST;
2581         rtm->rtm_flags = 0;
2582         rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2583         rtm->rtm_protocol = rt->rt6i_protocol;
2584         if (rt->rt6i_flags & RTF_DYNAMIC)
2585                 rtm->rtm_protocol = RTPROT_REDIRECT;
2586         else if (rt->rt6i_flags & RTF_ADDRCONF) {
2587                 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ROUTEINFO))
2588                         rtm->rtm_protocol = RTPROT_RA;
2589                 else
2590                         rtm->rtm_protocol = RTPROT_KERNEL;
2591         }
2592
2593         if (rt->rt6i_flags & RTF_CACHE)
2594                 rtm->rtm_flags |= RTM_F_CLONED;
2595
2596         if (dst) {
2597                 if (nla_put(skb, RTA_DST, 16, dst))
2598                         goto nla_put_failure;
2599                 rtm->rtm_dst_len = 128;
2600         } else if (rtm->rtm_dst_len)
2601                 if (nla_put(skb, RTA_DST, 16, &rt->rt6i_dst.addr))
2602                         goto nla_put_failure;
2603 #ifdef CONFIG_IPV6_SUBTREES
2604         if (src) {
2605                 if (nla_put(skb, RTA_SRC, 16, src))
2606                         goto nla_put_failure;
2607                 rtm->rtm_src_len = 128;
2608         } else if (rtm->rtm_src_len &&
2609                    nla_put(skb, RTA_SRC, 16, &rt->rt6i_src.addr))
2610                 goto nla_put_failure;
2611 #endif
2612         if (iif) {
2613 #ifdef CONFIG_IPV6_MROUTE
2614                 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
2615                         int err = ip6mr_get_route(net, skb, rtm, nowait);
2616                         if (err <= 0) {
2617                                 if (!nowait) {
2618                                         if (err == 0)
2619                                                 return 0;
2620                                         goto nla_put_failure;
2621                                 } else {
2622                                         if (err == -EMSGSIZE)
2623                                                 goto nla_put_failure;
2624                                 }
2625                         }
2626                 } else
2627 #endif
2628                         if (nla_put_u32(skb, RTA_IIF, iif))
2629                                 goto nla_put_failure;
2630         } else if (dst) {
2631                 struct in6_addr saddr_buf;
2632                 if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0 &&
2633                     nla_put(skb, RTA_PREFSRC, 16, &saddr_buf))
2634                         goto nla_put_failure;
2635         }
2636
2637         if (rt->rt6i_prefsrc.plen) {
2638                 struct in6_addr saddr_buf;
2639                 saddr_buf = rt->rt6i_prefsrc.addr;
2640                 if (nla_put(skb, RTA_PREFSRC, 16, &saddr_buf))
2641                         goto nla_put_failure;
2642         }
2643
2644         if (rtnetlink_put_metrics(skb, dst_metrics_ptr(&rt->dst)) < 0)
2645                 goto nla_put_failure;
2646
2647         if (rt->rt6i_flags & RTF_GATEWAY) {
2648                 if (nla_put(skb, RTA_GATEWAY, 16, &rt->rt6i_gateway) < 0)
2649                         goto nla_put_failure;
2650         }
2651
2652         if (rt->dst.dev &&
2653             nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex))
2654                 goto nla_put_failure;
2655         if (nla_put_u32(skb, RTA_PRIORITY, rt->rt6i_metric))
2656                 goto nla_put_failure;
2657
2658         expires = (rt->rt6i_flags & RTF_EXPIRES) ? rt->dst.expires - jiffies : 0;
2659
2660         if (rtnl_put_cacheinfo(skb, &rt->dst, 0, expires, rt->dst.error) < 0)
2661                 goto nla_put_failure;
2662
2663         return nlmsg_end(skb, nlh);
2664
2665 nla_put_failure:
2666         nlmsg_cancel(skb, nlh);
2667         return -EMSGSIZE;
2668 }
2669
2670 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
2671 {
2672         struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
2673         int prefix;
2674
2675         if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
2676                 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
2677                 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
2678         } else
2679                 prefix = 0;
2680
2681         return rt6_fill_node(arg->net,
2682                      arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
2683                      NETLINK_CB(arg->cb->skb).portid, arg->cb->nlh->nlmsg_seq,
2684                      prefix, 0, NLM_F_MULTI);
2685 }
2686
2687 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr* nlh)
2688 {
2689         struct net *net = sock_net(in_skb->sk);
2690         struct nlattr *tb[RTA_MAX+1];
2691         struct rt6_info *rt;
2692         struct sk_buff *skb;
2693         struct rtmsg *rtm;
2694         struct flowi6 fl6;
2695         int err, iif = 0, oif = 0;
2696
2697         err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2698         if (err < 0)
2699                 goto errout;
2700
2701         err = -EINVAL;
2702         memset(&fl6, 0, sizeof(fl6));
2703
2704         if (tb[RTA_SRC]) {
2705                 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
2706                         goto errout;
2707
2708                 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
2709         }
2710
2711         if (tb[RTA_DST]) {
2712                 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
2713                         goto errout;
2714
2715                 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
2716         }
2717
2718         if (tb[RTA_IIF])
2719                 iif = nla_get_u32(tb[RTA_IIF]);
2720
2721         if (tb[RTA_OIF])
2722                 oif = nla_get_u32(tb[RTA_OIF]);
2723
2724         if (iif) {
2725                 struct net_device *dev;
2726                 int flags = 0;
2727
2728                 dev = __dev_get_by_index(net, iif);
2729                 if (!dev) {
2730                         err = -ENODEV;
2731                         goto errout;
2732                 }
2733
2734                 fl6.flowi6_iif = iif;
2735
2736                 if (!ipv6_addr_any(&fl6.saddr))
2737                         flags |= RT6_LOOKUP_F_HAS_SADDR;
2738
2739                 rt = (struct rt6_info *)ip6_route_input_lookup(net, dev, &fl6,
2740                                                                flags);
2741         } else {
2742                 fl6.flowi6_oif = oif;
2743
2744                 rt = (struct rt6_info *)ip6_route_output(net, NULL, &fl6);
2745         }
2746
2747         skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
2748         if (!skb) {
2749                 ip6_rt_put(rt);
2750                 err = -ENOBUFS;
2751                 goto errout;
2752         }
2753
2754         /* Reserve room for dummy headers, this skb can pass
2755            through good chunk of routing engine.
2756          */
2757         skb_reset_mac_header(skb);
2758         skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
2759
2760         skb_dst_set(skb, &rt->dst);
2761
2762         err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
2763                             RTM_NEWROUTE, NETLINK_CB(in_skb).portid,
2764                             nlh->nlmsg_seq, 0, 0, 0);
2765         if (err < 0) {
2766                 kfree_skb(skb);
2767                 goto errout;
2768         }
2769
2770         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
2771 errout:
2772         return err;
2773 }
2774
2775 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info)
2776 {
2777         struct sk_buff *skb;
2778         struct net *net = info->nl_net;
2779         u32 seq;
2780         int err;
2781
2782         err = -ENOBUFS;
2783         seq = info->nlh ? info->nlh->nlmsg_seq : 0;
2784
2785         skb = nlmsg_new(rt6_nlmsg_size(), gfp_any());
2786         if (!skb)
2787                 goto errout;
2788
2789         err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
2790                                 event, info->portid, seq, 0, 0, 0);
2791         if (err < 0) {
2792                 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
2793                 WARN_ON(err == -EMSGSIZE);
2794                 kfree_skb(skb);
2795                 goto errout;
2796         }
2797         rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
2798                     info->nlh, gfp_any());
2799         return;
2800 errout:
2801         if (err < 0)
2802                 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
2803 }
2804
2805 static int ip6_route_dev_notify(struct notifier_block *this,
2806                                 unsigned long event, void *ptr)
2807 {
2808         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2809         struct net *net = dev_net(dev);
2810
2811         if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
2812                 net->ipv6.ip6_null_entry->dst.dev = dev;
2813                 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
2814 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2815                 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
2816                 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
2817                 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
2818                 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
2819 #endif
2820         }
2821
2822         return NOTIFY_OK;
2823 }
2824
2825 /*
2826  *      /proc
2827  */
2828
2829 #ifdef CONFIG_PROC_FS
2830
2831 struct rt6_proc_arg
2832 {
2833         char *buffer;
2834         int offset;
2835         int length;
2836         int skip;
2837         int len;
2838 };
2839
2840 static int rt6_info_route(struct rt6_info *rt, void *p_arg)
2841 {
2842         struct seq_file *m = p_arg;
2843
2844         seq_printf(m, "%pi6 %02x ", &rt->rt6i_dst.addr, rt->rt6i_dst.plen);
2845
2846 #ifdef CONFIG_IPV6_SUBTREES
2847         seq_printf(m, "%pi6 %02x ", &rt->rt6i_src.addr, rt->rt6i_src.plen);
2848 #else
2849         seq_puts(m, "00000000000000000000000000000000 00 ");
2850 #endif
2851         if (rt->rt6i_flags & RTF_GATEWAY) {
2852                 seq_printf(m, "%pi6", &rt->rt6i_gateway);
2853         } else {
2854                 seq_puts(m, "00000000000000000000000000000000");
2855         }
2856         seq_printf(m, " %08x %08x %08x %08x %8s\n",
2857                    rt->rt6i_metric, atomic_read(&rt->dst.__refcnt),
2858                    rt->dst.__use, rt->rt6i_flags,
2859                    rt->dst.dev ? rt->dst.dev->name : "");
2860         return 0;
2861 }
2862
2863 static int ipv6_route_show(struct seq_file *m, void *v)
2864 {
2865         struct net *net = (struct net *)m->private;
2866         fib6_clean_all_ro(net, rt6_info_route, 0, m);
2867         return 0;
2868 }
2869
2870 static int ipv6_route_open(struct inode *inode, struct file *file)
2871 {
2872         return single_open_net(inode, file, ipv6_route_show);
2873 }
2874
2875 static const struct file_operations ipv6_route_proc_fops = {
2876         .owner          = THIS_MODULE,
2877         .open           = ipv6_route_open,
2878         .read           = seq_read,
2879         .llseek         = seq_lseek,
2880         .release        = single_release_net,
2881 };
2882
2883 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
2884 {
2885         struct net *net = (struct net *)seq->private;
2886         seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
2887                    net->ipv6.rt6_stats->fib_nodes,
2888                    net->ipv6.rt6_stats->fib_route_nodes,
2889                    net->ipv6.rt6_stats->fib_rt_alloc,
2890                    net->ipv6.rt6_stats->fib_rt_entries,
2891                    net->ipv6.rt6_stats->fib_rt_cache,
2892                    dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
2893                    net->ipv6.rt6_stats->fib_discarded_routes);
2894
2895         return 0;
2896 }
2897
2898 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
2899 {
2900         return single_open_net(inode, file, rt6_stats_seq_show);
2901 }
2902
2903 static const struct file_operations rt6_stats_seq_fops = {
2904         .owner   = THIS_MODULE,
2905         .open    = rt6_stats_seq_open,
2906         .read    = seq_read,
2907         .llseek  = seq_lseek,
2908         .release = single_release_net,
2909 };
2910 #endif  /* CONFIG_PROC_FS */
2911
2912 #ifdef CONFIG_SYSCTL
2913
2914 static
2915 int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
2916                               void __user *buffer, size_t *lenp, loff_t *ppos)
2917 {
2918         struct net *net;
2919         int delay;
2920         if (!write)
2921                 return -EINVAL;
2922
2923         net = (struct net *)ctl->extra1;
2924         delay = net->ipv6.sysctl.flush_delay;
2925         proc_dointvec(ctl, write, buffer, lenp, ppos);
2926         fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
2927         return 0;
2928 }
2929
2930 struct ctl_table ipv6_route_table_template[] = {
2931         {
2932                 .procname       =       "flush",
2933                 .data           =       &init_net.ipv6.sysctl.flush_delay,
2934                 .maxlen         =       sizeof(int),
2935                 .mode           =       0200,
2936                 .proc_handler   =       ipv6_sysctl_rtcache_flush
2937         },
2938         {
2939                 .procname       =       "gc_thresh",
2940                 .data           =       &ip6_dst_ops_template.gc_thresh,
2941                 .maxlen         =       sizeof(int),
2942                 .mode           =       0644,
2943                 .proc_handler   =       proc_dointvec,
2944         },
2945         {
2946                 .procname       =       "max_size",
2947                 .data           =       &init_net.ipv6.sysctl.ip6_rt_max_size,
2948                 .maxlen         =       sizeof(int),
2949                 .mode           =       0644,
2950                 .proc_handler   =       proc_dointvec,
2951         },
2952         {
2953                 .procname       =       "gc_min_interval",
2954                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2955                 .maxlen         =       sizeof(int),
2956                 .mode           =       0644,
2957                 .proc_handler   =       proc_dointvec_jiffies,
2958         },
2959         {
2960                 .procname       =       "gc_timeout",
2961                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
2962                 .maxlen         =       sizeof(int),
2963                 .mode           =       0644,
2964                 .proc_handler   =       proc_dointvec_jiffies,
2965         },
2966         {
2967                 .procname       =       "gc_interval",
2968                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_interval,
2969                 .maxlen         =       sizeof(int),
2970                 .mode           =       0644,
2971                 .proc_handler   =       proc_dointvec_jiffies,
2972         },
2973         {
2974                 .procname       =       "gc_elasticity",
2975                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
2976                 .maxlen         =       sizeof(int),
2977                 .mode           =       0644,
2978                 .proc_handler   =       proc_dointvec,
2979         },
2980         {
2981                 .procname       =       "mtu_expires",
2982                 .data           =       &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
2983                 .maxlen         =       sizeof(int),
2984                 .mode           =       0644,
2985                 .proc_handler   =       proc_dointvec_jiffies,
2986         },
2987         {
2988                 .procname       =       "min_adv_mss",
2989                 .data           =       &init_net.ipv6.sysctl.ip6_rt_min_advmss,
2990                 .maxlen         =       sizeof(int),
2991                 .mode           =       0644,
2992                 .proc_handler   =       proc_dointvec,
2993         },
2994         {
2995                 .procname       =       "gc_min_interval_ms",
2996                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2997                 .maxlen         =       sizeof(int),
2998                 .mode           =       0644,
2999                 .proc_handler   =       proc_dointvec_ms_jiffies,
3000         },
3001         { }
3002 };
3003
3004 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
3005 {
3006         struct ctl_table *table;
3007
3008         table = kmemdup(ipv6_route_table_template,
3009                         sizeof(ipv6_route_table_template),
3010                         GFP_KERNEL);
3011
3012         if (table) {
3013                 table[0].data = &net->ipv6.sysctl.flush_delay;
3014                 table[0].extra1 = net;
3015                 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
3016                 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
3017                 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
3018                 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
3019                 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
3020                 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
3021                 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
3022                 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
3023                 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
3024
3025                 /* Don't export sysctls to unprivileged users */
3026                 if (net->user_ns != &init_user_ns)
3027                         table[0].procname = NULL;
3028         }
3029
3030         return table;
3031 }
3032 #endif
3033
3034 static int __net_init ip6_route_net_init(struct net *net)
3035 {
3036         int ret = -ENOMEM;
3037
3038         memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
3039                sizeof(net->ipv6.ip6_dst_ops));
3040
3041         if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
3042                 goto out_ip6_dst_ops;
3043
3044         net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
3045                                            sizeof(*net->ipv6.ip6_null_entry),
3046                                            GFP_KERNEL);
3047         if (!net->ipv6.ip6_null_entry)
3048                 goto out_ip6_dst_entries;
3049         net->ipv6.ip6_null_entry->dst.path =
3050                 (struct dst_entry *)net->ipv6.ip6_null_entry;
3051         net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3052         dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
3053                          ip6_template_metrics, true);
3054
3055 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3056         net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
3057                                                sizeof(*net->ipv6.ip6_prohibit_entry),
3058                                                GFP_KERNEL);
3059         if (!net->ipv6.ip6_prohibit_entry)
3060                 goto out_ip6_null_entry;
3061         net->ipv6.ip6_prohibit_entry->dst.path =
3062                 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
3063         net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3064         dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
3065                          ip6_template_metrics, true);
3066
3067         net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
3068                                                sizeof(*net->ipv6.ip6_blk_hole_entry),
3069                                                GFP_KERNEL);
3070         if (!net->ipv6.ip6_blk_hole_entry)
3071                 goto out_ip6_prohibit_entry;
3072         net->ipv6.ip6_blk_hole_entry->dst.path =
3073                 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
3074         net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3075         dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
3076                          ip6_template_metrics, true);
3077 #endif
3078
3079         net->ipv6.sysctl.flush_delay = 0;
3080         net->ipv6.sysctl.ip6_rt_max_size = 4096;
3081         net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
3082         net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
3083         net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
3084         net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
3085         net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
3086         net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
3087
3088         net->ipv6.ip6_rt_gc_expire = 30*HZ;
3089
3090         ret = 0;
3091 out:
3092         return ret;
3093
3094 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3095 out_ip6_prohibit_entry:
3096         kfree(net->ipv6.ip6_prohibit_entry);
3097 out_ip6_null_entry:
3098         kfree(net->ipv6.ip6_null_entry);
3099 #endif
3100 out_ip6_dst_entries:
3101         dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3102 out_ip6_dst_ops:
3103         goto out;
3104 }
3105
3106 static void __net_exit ip6_route_net_exit(struct net *net)
3107 {
3108         kfree(net->ipv6.ip6_null_entry);
3109 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3110         kfree(net->ipv6.ip6_prohibit_entry);
3111         kfree(net->ipv6.ip6_blk_hole_entry);
3112 #endif
3113         dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3114 }
3115
3116 static int __net_init ip6_route_net_init_late(struct net *net)
3117 {
3118 #ifdef CONFIG_PROC_FS
3119         proc_create("ipv6_route", 0, net->proc_net, &ipv6_route_proc_fops);
3120         proc_create("rt6_stats", S_IRUGO, net->proc_net, &rt6_stats_seq_fops);
3121 #endif
3122         return 0;
3123 }
3124
3125 static void __net_exit ip6_route_net_exit_late(struct net *net)
3126 {
3127 #ifdef CONFIG_PROC_FS
3128         remove_proc_entry("ipv6_route", net->proc_net);
3129         remove_proc_entry("rt6_stats", net->proc_net);
3130 #endif
3131 }
3132
3133 static struct pernet_operations ip6_route_net_ops = {
3134         .init = ip6_route_net_init,
3135         .exit = ip6_route_net_exit,
3136 };
3137
3138 static int __net_init ipv6_inetpeer_init(struct net *net)
3139 {
3140         struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
3141
3142         if (!bp)
3143                 return -ENOMEM;
3144         inet_peer_base_init(bp);
3145         net->ipv6.peers = bp;
3146         return 0;
3147 }
3148
3149 static void __net_exit ipv6_inetpeer_exit(struct net *net)
3150 {
3151         struct inet_peer_base *bp = net->ipv6.peers;
3152
3153         net->ipv6.peers = NULL;
3154         inetpeer_invalidate_tree(bp);
3155         kfree(bp);
3156 }
3157
3158 static struct pernet_operations ipv6_inetpeer_ops = {
3159         .init   =       ipv6_inetpeer_init,
3160         .exit   =       ipv6_inetpeer_exit,
3161 };
3162
3163 static struct pernet_operations ip6_route_net_late_ops = {
3164         .init = ip6_route_net_init_late,
3165         .exit = ip6_route_net_exit_late,
3166 };
3167
3168 static struct notifier_block ip6_route_dev_notifier = {
3169         .notifier_call = ip6_route_dev_notify,
3170         .priority = 0,
3171 };
3172
3173 int __init ip6_route_init(void)
3174 {
3175         int ret;
3176
3177         ret = -ENOMEM;
3178         ip6_dst_ops_template.kmem_cachep =
3179                 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
3180                                   SLAB_HWCACHE_ALIGN, NULL);
3181         if (!ip6_dst_ops_template.kmem_cachep)
3182                 goto out;
3183
3184         ret = dst_entries_init(&ip6_dst_blackhole_ops);
3185         if (ret)
3186                 goto out_kmem_cache;
3187
3188         ret = register_pernet_subsys(&ipv6_inetpeer_ops);
3189         if (ret)
3190                 goto out_dst_entries;
3191
3192         ret = register_pernet_subsys(&ip6_route_net_ops);
3193         if (ret)
3194                 goto out_register_inetpeer;
3195
3196         ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
3197
3198         /* Registering of the loopback is done before this portion of code,
3199          * the loopback reference in rt6_info will not be taken, do it
3200          * manually for init_net */
3201         init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
3202         init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3203   #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3204         init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
3205         init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3206         init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
3207         init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3208   #endif
3209         ret = fib6_init();
3210         if (ret)
3211                 goto out_register_subsys;
3212
3213         ret = xfrm6_init();
3214         if (ret)
3215                 goto out_fib6_init;
3216
3217         ret = fib6_rules_init();
3218         if (ret)
3219                 goto xfrm6_init;
3220
3221         ret = register_pernet_subsys(&ip6_route_net_late_ops);
3222         if (ret)
3223                 goto fib6_rules_init;
3224
3225         ret = -ENOBUFS;
3226         if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL, NULL) ||
3227             __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL, NULL) ||
3228             __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL, NULL))
3229                 goto out_register_late_subsys;
3230
3231         ret = register_netdevice_notifier(&ip6_route_dev_notifier);
3232         if (ret)
3233                 goto out_register_late_subsys;
3234
3235 out:
3236         return ret;
3237
3238 out_register_late_subsys:
3239         unregister_pernet_subsys(&ip6_route_net_late_ops);
3240 fib6_rules_init:
3241         fib6_rules_cleanup();
3242 xfrm6_init:
3243         xfrm6_fini();
3244 out_fib6_init:
3245         fib6_gc_cleanup();
3246 out_register_subsys:
3247         unregister_pernet_subsys(&ip6_route_net_ops);
3248 out_register_inetpeer:
3249         unregister_pernet_subsys(&ipv6_inetpeer_ops);
3250 out_dst_entries:
3251         dst_entries_destroy(&ip6_dst_blackhole_ops);
3252 out_kmem_cache:
3253         kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3254         goto out;
3255 }
3256
3257 void ip6_route_cleanup(void)
3258 {
3259         unregister_netdevice_notifier(&ip6_route_dev_notifier);
3260         unregister_pernet_subsys(&ip6_route_net_late_ops);
3261         fib6_rules_cleanup();
3262         xfrm6_fini();
3263         fib6_gc_cleanup();
3264         unregister_pernet_subsys(&ipv6_inetpeer_ops);
3265         unregister_pernet_subsys(&ip6_route_net_ops);
3266         dst_entries_destroy(&ip6_dst_blackhole_ops);
3267         kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3268 }