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/dst_metadata.h>
  58 #include <net/xfrm.h>
  59 #include <net/netevent.h>
  60 #include <net/netlink.h>
  61 #include <net/nexthop.h>
  62 #include <net/lwtunnel.h>
  63 #include <net/ip_tunnels.h>
  64 #include <net/l3mdev.h>
  65 #include <trace/events/fib6.h>
  66
  67 #include <asm/uaccess.h>
  68
  69 #ifdef CONFIG_SYSCTL
  70 #include <linux/sysctl.h>
  71 #endif
  72
  73 enum rt6_nud_state {
  74         RT6_NUD_FAIL_HARD = -3,
  75         RT6_NUD_FAIL_PROBE = -2,
  76         RT6_NUD_FAIL_DO_RR = -1,
  77         RT6_NUD_SUCCEED = 1
  78 };
  79
  80 static void ip6_rt_copy_init(struct rt6_info *rt, struct rt6_info *ort);
  81 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie);
  82 static unsigned int      ip6_default_advmss(const struct dst_entry *dst);
  83 static unsigned int      ip6_mtu(const struct dst_entry *dst);
  84 static struct dst_entry *ip6_negative_advice(struct dst_entry *);
  85 static void             ip6_dst_destroy(struct dst_entry *);
  86 static void             ip6_dst_ifdown(struct dst_entry *,
  87                                        struct net_device *dev, int how);
  88 static int               ip6_dst_gc(struct dst_ops *ops);
  89
  90 static int              ip6_pkt_discard(struct sk_buff *skb);
  91 static int              ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb);
  92 static int              ip6_pkt_prohibit(struct sk_buff *skb);
  93 static int              ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb);
  94 static void             ip6_link_failure(struct sk_buff *skb);
  95 static void             ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
  96                                            struct sk_buff *skb, u32 mtu);
  97 static void             rt6_do_redirect(struct dst_entry *dst, struct sock *sk,
  98                                         struct sk_buff *skb);
  99 static void             rt6_dst_from_metrics_check(struct rt6_info *rt);
 100 static int rt6_score_route(struct rt6_info *rt, int oif, int strict);
 101
 102 #ifdef CONFIG_IPV6_ROUTE_INFO
 103 static struct rt6_info *rt6_add_route_info(struct net *net,
 104                                            const struct in6_addr *prefix, int prefixlen,
 105                                            const struct in6_addr *gwaddr, int ifindex,
 106                                            unsigned int pref);
 107 static struct rt6_info *rt6_get_route_info(struct net *net,
 108                                            const struct in6_addr *prefix, int prefixlen,
 109                                            const struct in6_addr *gwaddr, int ifindex);
 110 #endif
 111
 112 struct uncached_list {
 113         spinlock_t              lock;
 114         struct list_head        head;
 115 };
 116
 117 static DEFINE_PER_CPU_ALIGNED(struct uncached_list, rt6_uncached_list);
 118
 119 static void rt6_uncached_list_add(struct rt6_info *rt)
 120 {
 121         struct uncached_list *ul = raw_cpu_ptr(&rt6_uncached_list);
 122
 123         rt->dst.flags |= DST_NOCACHE;
 124         rt->rt6i_uncached_list = ul;
 125
 126         spin_lock_bh(&ul->lock);
 127         list_add_tail(&rt->rt6i_uncached, &ul->head);
 128         spin_unlock_bh(&ul->lock);
 129 }
 130
 131 static void rt6_uncached_list_del(struct rt6_info *rt)
 132 {
 133         if (!list_empty(&rt->rt6i_uncached)) {
 134                 struct uncached_list *ul = rt->rt6i_uncached_list;
 135
 136                 spin_lock_bh(&ul->lock);
 137                 list_del(&rt->rt6i_uncached);
 138                 spin_unlock_bh(&ul->lock);
 139         }
 140 }
 141
 142 static void rt6_uncached_list_flush_dev(struct net *net, struct net_device *dev)
 143 {
 144         struct net_device *loopback_dev = net->loopback_dev;
 145         int cpu;
 146
 147         if (dev == loopback_dev)
 148                 return;
 149
 150         for_each_possible_cpu(cpu) {
 151                 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
 152                 struct rt6_info *rt;
 153
 154                 spin_lock_bh(&ul->lock);
 155                 list_for_each_entry(rt, &ul->head, rt6i_uncached) {
 156                         struct inet6_dev *rt_idev = rt->rt6i_idev;
 157                         struct net_device *rt_dev = rt->dst.dev;
 158
 159                         if (rt_idev->dev == dev) {
 160                                 rt->rt6i_idev = in6_dev_get(loopback_dev);
 161                                 in6_dev_put(rt_idev);
 162                         }
 163
 164                         if (rt_dev == dev) {
 165                                 rt->dst.dev = loopback_dev;
 166                                 dev_hold(rt->dst.dev);
 167                                 dev_put(rt_dev);
 168                         }
 169                 }
 170                 spin_unlock_bh(&ul->lock);
 171         }
 172 }
 173
 174 static u32 *rt6_pcpu_cow_metrics(struct rt6_info *rt)
 175 {
 176         return dst_metrics_write_ptr(rt->dst.from);
 177 }
 178
 179 static u32 *ipv6_cow_metrics(struct dst_entry *dst, unsigned long old)
 180 {
 181         struct rt6_info *rt = (struct rt6_info *)dst;
 182
 183         if (rt->rt6i_flags & RTF_PCPU)
 184                 return rt6_pcpu_cow_metrics(rt);
 185         else if (rt->rt6i_flags & RTF_CACHE)
 186                 return NULL;
 187         else
 188                 return dst_cow_metrics_generic(dst, old);
 189 }
 190
 191 static inline const void *choose_neigh_daddr(struct rt6_info *rt,
 192                                              struct sk_buff *skb,
 193                                              const void *daddr)
 194 {
 195         struct in6_addr *p = &rt->rt6i_gateway;
 196
 197         if (!ipv6_addr_any(p))
 198                 return (const void *) p;
 199         else if (skb)
 200                 return &ipv6_hdr(skb)->daddr;
 201         return daddr;
 202 }
 203
 204 static struct neighbour *ip6_neigh_lookup(const struct dst_entry *dst,
 205                                           struct sk_buff *skb,
 206                                           const void *daddr)
 207 {
 208         struct rt6_info *rt = (struct rt6_info *) dst;
 209         struct neighbour *n;
 210
 211         daddr = choose_neigh_daddr(rt, skb, daddr);
 212         n = __ipv6_neigh_lookup(dst->dev, daddr);
 213         if (n)
 214                 return n;
 215         return neigh_create(&nd_tbl, daddr, dst->dev);
 216 }
 217
 218 static struct dst_ops ip6_dst_ops_template = {
 219         .family                 =       AF_INET6,
 220         .gc                     =       ip6_dst_gc,
 221         .gc_thresh              =       1024,
 222         .check                  =       ip6_dst_check,
 223         .default_advmss         =       ip6_default_advmss,
 224         .mtu                    =       ip6_mtu,
 225         .cow_metrics            =       ipv6_cow_metrics,
 226         .destroy                =       ip6_dst_destroy,
 227         .ifdown                 =       ip6_dst_ifdown,
 228         .negative_advice        =       ip6_negative_advice,
 229         .link_failure           =       ip6_link_failure,
 230         .update_pmtu            =       ip6_rt_update_pmtu,
 231         .redirect               =       rt6_do_redirect,
 232         .local_out              =       __ip6_local_out,
 233         .neigh_lookup           =       ip6_neigh_lookup,
 234 };
 235
 236 static unsigned int ip6_blackhole_mtu(const struct dst_entry *dst)
 237 {
 238         unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
 239
 240         return mtu ? : dst->dev->mtu;
 241 }
 242
 243 static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, struct sock *sk,
 244                                          struct sk_buff *skb, u32 mtu)
 245 {
 246 }
 247
 248 static void ip6_rt_blackhole_redirect(struct dst_entry *dst, struct sock *sk,
 249                                       struct sk_buff *skb)
 250 {
 251 }
 252
 253 static struct dst_ops ip6_dst_blackhole_ops = {
 254         .family                 =       AF_INET6,
 255         .destroy                =       ip6_dst_destroy,
 256         .check                  =       ip6_dst_check,
 257         .mtu                    =       ip6_blackhole_mtu,
 258         .default_advmss         =       ip6_default_advmss,
 259         .update_pmtu            =       ip6_rt_blackhole_update_pmtu,
 260         .redirect               =       ip6_rt_blackhole_redirect,
 261         .cow_metrics            =       dst_cow_metrics_generic,
 262         .neigh_lookup           =       ip6_neigh_lookup,
 263 };
 264
 265 static const u32 ip6_template_metrics[RTAX_MAX] = {
 266         [RTAX_HOPLIMIT - 1] = 0,
 267 };
 268
 269 static const struct rt6_info ip6_null_entry_template = {
 270         .dst = {
 271                 .__refcnt       = ATOMIC_INIT(1),
 272                 .__use          = 1,
 273                 .obsolete       = DST_OBSOLETE_FORCE_CHK,
 274                 .error          = -ENETUNREACH,
 275                 .input          = ip6_pkt_discard,
 276                 .output         = ip6_pkt_discard_out,
 277         },
 278         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
 279         .rt6i_protocol  = RTPROT_KERNEL,
 280         .rt6i_metric    = ~(u32) 0,
 281         .rt6i_ref       = ATOMIC_INIT(1),
 282 };
 283
 284 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
 285
 286 static const struct rt6_info ip6_prohibit_entry_template = {
 287         .dst = {
 288                 .__refcnt       = ATOMIC_INIT(1),
 289                 .__use          = 1,
 290                 .obsolete       = DST_OBSOLETE_FORCE_CHK,
 291                 .error          = -EACCES,
 292                 .input          = ip6_pkt_prohibit,
 293                 .output         = ip6_pkt_prohibit_out,
 294         },
 295         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
 296         .rt6i_protocol  = RTPROT_KERNEL,
 297         .rt6i_metric    = ~(u32) 0,
 298         .rt6i_ref       = ATOMIC_INIT(1),
 299 };
 300
 301 static const struct rt6_info ip6_blk_hole_entry_template = {
 302         .dst = {
 303                 .__refcnt       = ATOMIC_INIT(1),
 304                 .__use          = 1,
 305                 .obsolete       = DST_OBSOLETE_FORCE_CHK,
 306                 .error          = -EINVAL,
 307                 .input          = dst_discard,
 308                 .output         = dst_discard_out,
 309         },
 310         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
 311         .rt6i_protocol  = RTPROT_KERNEL,
 312         .rt6i_metric    = ~(u32) 0,
 313         .rt6i_ref       = ATOMIC_INIT(1),
 314 };
 315
 316 #endif
 317
 318 static void rt6_info_init(struct rt6_info *rt)
 319 {
 320         struct dst_entry *dst = &rt->dst;
 321
 322         memset(dst + 1, 0, sizeof(*rt) - sizeof(*dst));
 323         INIT_LIST_HEAD(&rt->rt6i_siblings);
 324         INIT_LIST_HEAD(&rt->rt6i_uncached);
 325 }
 326
 327 /* allocate dst with ip6_dst_ops */
 328 static struct rt6_info *__ip6_dst_alloc(struct net *net,
 329                                         struct net_device *dev,
 330                                         int flags)
 331 {
 332         struct rt6_info *rt = dst_alloc(&net->ipv6.ip6_dst_ops, dev,
 333                                         0, DST_OBSOLETE_FORCE_CHK, flags);
 334
 335         if (rt)
 336                 rt6_info_init(rt);
 337
 338         return rt;
 339 }
 340
 341 struct rt6_info *ip6_dst_alloc(struct net *net,
 342                                struct net_device *dev,
 343                                int flags)
 344 {
 345         struct rt6_info *rt = __ip6_dst_alloc(net, dev, flags);
 346
 347         if (rt) {
 348                 rt->rt6i_pcpu = alloc_percpu_gfp(struct rt6_info *, GFP_ATOMIC);
 349                 if (rt->rt6i_pcpu) {
 350                         int cpu;
 351
 352                         for_each_possible_cpu(cpu) {
 353                                 struct rt6_info **p;
 354
 355                                 p = per_cpu_ptr(rt->rt6i_pcpu, cpu);
 356                                 /* no one shares rt */
 357                                 *p =  NULL;
 358                         }
 359                 } else {
 360                         dst_destroy((struct dst_entry *)rt);
 361                         return NULL;
 362                 }
 363         }
 364
 365         return rt;
 366 }
 367 EXPORT_SYMBOL(ip6_dst_alloc);
 368
 369 static void ip6_dst_destroy(struct dst_entry *dst)
 370 {
 371         struct rt6_info *rt = (struct rt6_info *)dst;
 372         struct dst_entry *from = dst->from;
 373         struct inet6_dev *idev;
 374
 375         dst_destroy_metrics_generic(dst);
 376         free_percpu(rt->rt6i_pcpu);
 377         rt6_uncached_list_del(rt);
 378
 379         idev = rt->rt6i_idev;
 380         if (idev) {
 381                 rt->rt6i_idev = NULL;
 382                 in6_dev_put(idev);
 383         }
 384
 385         dst->from = NULL;
 386         dst_release(from);
 387 }
 388
 389 static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
 390                            int how)
 391 {
 392         struct rt6_info *rt = (struct rt6_info *)dst;
 393         struct inet6_dev *idev = rt->rt6i_idev;
 394         struct net_device *loopback_dev =
 395                 dev_net(dev)->loopback_dev;
 396
 397         if (dev != loopback_dev) {
 398                 if (idev && idev->dev == dev) {
 399                         struct inet6_dev *loopback_idev =
 400                                 in6_dev_get(loopback_dev);
 401                         if (loopback_idev) {
 402                                 rt->rt6i_idev = loopback_idev;
 403                                 in6_dev_put(idev);
 404                         }
 405                 }
 406         }
 407 }
 408
 409 static bool __rt6_check_expired(const struct rt6_info *rt)
 410 {
 411         if (rt->rt6i_flags & RTF_EXPIRES)
 412                 return time_after(jiffies, rt->dst.expires);
 413         else
 414                 return false;
 415 }
 416
 417 static bool rt6_check_expired(const struct rt6_info *rt)
 418 {
 419         if (rt->rt6i_flags & RTF_EXPIRES) {
 420                 if (time_after(jiffies, rt->dst.expires))
 421                         return true;
 422         } else if (rt->dst.from) {
 423                 return rt6_check_expired((struct rt6_info *) rt->dst.from);
 424         }
 425         return false;
 426 }
 427
 428 /* Multipath route selection:
 429  *   Hash based function using packet header and flowlabel.
 430  * Adapted from fib_info_hashfn()
 431  */
 432 static int rt6_info_hash_nhsfn(unsigned int candidate_count,
 433                                const struct flowi6 *fl6)
 434 {
 435         return get_hash_from_flowi6(fl6) % candidate_count;
 436 }
 437
 438 static struct rt6_info *rt6_multipath_select(struct rt6_info *match,
 439                                              struct flowi6 *fl6, int oif,
 440                                              int strict)
 441 {
 442         struct rt6_info *sibling, *next_sibling;
 443         int route_choosen;
 444
 445         route_choosen = rt6_info_hash_nhsfn(match->rt6i_nsiblings + 1, fl6);
 446         /* Don't change the route, if route_choosen == 0
 447          * (siblings does not include ourself)
 448          */
 449         if (route_choosen)
 450                 list_for_each_entry_safe(sibling, next_sibling,
 451                                 &match->rt6i_siblings, rt6i_siblings) {
 452                         route_choosen--;
 453                         if (route_choosen == 0) {
 454                                 if (rt6_score_route(sibling, oif, strict) < 0)
 455                                         break;
 456                                 match = sibling;
 457                                 break;
 458                         }
 459                 }
 460         return match;
 461 }
 462
 463 /*
 464  *      Route lookup. Any table->tb6_lock is implied.
 465  */
 466
 467 static inline struct rt6_info *rt6_device_match(struct net *net,
 468                                                     struct rt6_info *rt,
 469                                                     const struct in6_addr *saddr,
 470                                                     int oif,
 471                                                     int flags)
 472 {
 473         struct rt6_info *local = NULL;
 474         struct rt6_info *sprt;
 475
 476         if (!oif && ipv6_addr_any(saddr))
 477                 goto out;
 478
 479         for (sprt = rt; sprt; sprt = sprt->dst.rt6_next) {
 480                 struct net_device *dev = sprt->dst.dev;
 481
 482                 if (oif) {
 483                         if (dev->ifindex == oif)
 484                                 return sprt;
 485                         if (dev->flags & IFF_LOOPBACK) {
 486                                 if (!sprt->rt6i_idev ||
 487                                     sprt->rt6i_idev->dev->ifindex != oif) {
 488                                         if (flags & RT6_LOOKUP_F_IFACE)
 489                                                 continue;
 490                                         if (local &&
 491                                             local->rt6i_idev->dev->ifindex == oif)
 492                                                 continue;
 493                                 }
 494                                 local = sprt;
 495                         }
 496                 } else {
 497                         if (ipv6_chk_addr(net, saddr, dev,
 498                                           flags & RT6_LOOKUP_F_IFACE))
 499                                 return sprt;
 500                 }
 501         }
 502
 503         if (oif) {
 504                 if (local)
 505                         return local;
 506
 507                 if (flags & RT6_LOOKUP_F_IFACE)
 508                         return net->ipv6.ip6_null_entry;
 509         }
 510 out:
 511         return rt;
 512 }
 513
 514 #ifdef CONFIG_IPV6_ROUTER_PREF
 515 struct __rt6_probe_work {
 516         struct work_struct work;
 517         struct in6_addr target;
 518         struct net_device *dev;
 519 };
 520
 521 static void rt6_probe_deferred(struct work_struct *w)
 522 {
 523         struct in6_addr mcaddr;
 524         struct __rt6_probe_work *work =
 525                 container_of(w, struct __rt6_probe_work, work);
 526
 527         addrconf_addr_solict_mult(&work->target, &mcaddr);
 528         ndisc_send_ns(work->dev, &work->target, &mcaddr, NULL);
 529         dev_put(work->dev);
 530         kfree(work);
 531 }
 532
 533 static void rt6_probe(struct rt6_info *rt)
 534 {
 535         struct __rt6_probe_work *work;
 536         struct neighbour *neigh;
 537         /*
 538          * Okay, this does not seem to be appropriate
 539          * for now, however, we need to check if it
 540          * is really so; aka Router Reachability Probing.
 541          *
 542          * Router Reachability Probe MUST be rate-limited
 543          * to no more than one per minute.
 544          */
 545         if (!rt || !(rt->rt6i_flags & RTF_GATEWAY))
 546                 return;
 547         rcu_read_lock_bh();
 548         neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
 549         if (neigh) {
 550                 if (neigh->nud_state & NUD_VALID)
 551                         goto out;
 552
 553                 work = NULL;
 554                 write_lock(&neigh->lock);
 555                 if (!(neigh->nud_state & NUD_VALID) &&
 556                     time_after(jiffies,
 557                                neigh->updated +
 558                                rt->rt6i_idev->cnf.rtr_probe_interval)) {
 559                         work = kmalloc(sizeof(*work), GFP_ATOMIC);
 560                         if (work)
 561                                 __neigh_set_probe_once(neigh);
 562                 }
 563                 write_unlock(&neigh->lock);
 564         } else {
 565                 work = kmalloc(sizeof(*work), GFP_ATOMIC);
 566         }
 567
 568         if (work) {
 569                 INIT_WORK(&work->work, rt6_probe_deferred);
 570                 work->target = rt->rt6i_gateway;
 571                 dev_hold(rt->dst.dev);
 572                 work->dev = rt->dst.dev;
 573                 schedule_work(&work->work);
 574         }
 575
 576 out:
 577         rcu_read_unlock_bh();
 578 }
 579 #else
 580 static inline void rt6_probe(struct rt6_info *rt)
 581 {
 582 }
 583 #endif
 584
 585 /*
 586  * Default Router Selection (RFC 2461 6.3.6)
 587  */
 588 static inline int rt6_check_dev(struct rt6_info *rt, int oif)
 589 {
 590         struct net_device *dev = rt->dst.dev;
 591         if (!oif || dev->ifindex == oif)
 592                 return 2;
 593         if ((dev->flags & IFF_LOOPBACK) &&
 594             rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
 595                 return 1;
 596         return 0;
 597 }
 598
 599 static inline enum rt6_nud_state rt6_check_neigh(struct rt6_info *rt)
 600 {
 601         struct neighbour *neigh;
 602         enum rt6_nud_state ret = RT6_NUD_FAIL_HARD;
 603
 604         if (rt->rt6i_flags & RTF_NONEXTHOP ||
 605             !(rt->rt6i_flags & RTF_GATEWAY))
 606                 return RT6_NUD_SUCCEED;
 607
 608         rcu_read_lock_bh();
 609         neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
 610         if (neigh) {
 611                 read_lock(&neigh->lock);
 612                 if (neigh->nud_state & NUD_VALID)
 613                         ret = RT6_NUD_SUCCEED;
 614 #ifdef CONFIG_IPV6_ROUTER_PREF
 615                 else if (!(neigh->nud_state & NUD_FAILED))
 616                         ret = RT6_NUD_SUCCEED;
 617                 else
 618                         ret = RT6_NUD_FAIL_PROBE;
 619 #endif
 620                 read_unlock(&neigh->lock);
 621         } else {
 622                 ret = IS_ENABLED(CONFIG_IPV6_ROUTER_PREF) ?
 623                       RT6_NUD_SUCCEED : RT6_NUD_FAIL_DO_RR;
 624         }
 625         rcu_read_unlock_bh();
 626
 627         return ret;
 628 }
 629
 630 static int rt6_score_route(struct rt6_info *rt, int oif,
 631                            int strict)
 632 {
 633         int m;
 634
 635         m = rt6_check_dev(rt, oif);
 636         if (!m && (strict & RT6_LOOKUP_F_IFACE))
 637                 return RT6_NUD_FAIL_HARD;
 638 #ifdef CONFIG_IPV6_ROUTER_PREF
 639         m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
 640 #endif
 641         if (strict & RT6_LOOKUP_F_REACHABLE) {
 642                 int n = rt6_check_neigh(rt);
 643                 if (n < 0)
 644                         return n;
 645         }
 646         return m;
 647 }
 648
 649 static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
 650                                    int *mpri, struct rt6_info *match,
 651                                    bool *do_rr)
 652 {
 653         int m;
 654         bool match_do_rr = false;
 655         struct inet6_dev *idev = rt->rt6i_idev;
 656         struct net_device *dev = rt->dst.dev;
 657
 658         if (dev && !netif_carrier_ok(dev) &&
 659             idev->cnf.ignore_routes_with_linkdown)
 660                 goto out;
 661
 662         if (rt6_check_expired(rt))
 663                 goto out;
 664
 665         m = rt6_score_route(rt, oif, strict);
 666         if (m == RT6_NUD_FAIL_DO_RR) {
 667                 match_do_rr = true;
 668                 m = 0; /* lowest valid score */
 669         } else if (m == RT6_NUD_FAIL_HARD) {
 670                 goto out;
 671         }
 672
 673         if (strict & RT6_LOOKUP_F_REACHABLE)
 674                 rt6_probe(rt);
 675
 676         /* note that m can be RT6_NUD_FAIL_PROBE at this point */
 677         if (m > *mpri) {
 678                 *do_rr = match_do_rr;
 679                 *mpri = m;
 680                 match = rt;
 681         }
 682 out:
 683         return match;
 684 }
 685
 686 static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
 687                                      struct rt6_info *rr_head,
 688                                      u32 metric, int oif, int strict,
 689                                      bool *do_rr)
 690 {
 691         struct rt6_info *rt, *match, *cont;
 692         int mpri = -1;
 693
 694         match = NULL;
 695         cont = NULL;
 696         for (rt = rr_head; rt; rt = rt->dst.rt6_next) {
 697                 if (rt->rt6i_metric != metric) {
 698                         cont = rt;
 699                         break;
 700                 }
 701
 702                 match = find_match(rt, oif, strict, &mpri, match, do_rr);
 703         }
 704
 705         for (rt = fn->leaf; rt && rt != rr_head; rt = rt->dst.rt6_next) {
 706                 if (rt->rt6i_metric != metric) {
 707                         cont = rt;
 708                         break;
 709                 }
 710
 711                 match = find_match(rt, oif, strict, &mpri, match, do_rr);
 712         }
 713
 714         if (match || !cont)
 715                 return match;
 716
 717         for (rt = cont; rt; rt = rt->dst.rt6_next)
 718                 match = find_match(rt, oif, strict, &mpri, match, do_rr);
 719
 720         return match;
 721 }
 722
 723 static struct rt6_info *rt6_select(struct fib6_node *fn, int oif, int strict)
 724 {
 725         struct rt6_info *match, *rt0;
 726         struct net *net;
 727         bool do_rr = false;
 728
 729         rt0 = fn->rr_ptr;
 730         if (!rt0)
 731                 fn->rr_ptr = rt0 = fn->leaf;
 732
 733         match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict,
 734                              &do_rr);
 735
 736         if (do_rr) {
 737                 struct rt6_info *next = rt0->dst.rt6_next;
 738
 739                 /* no entries matched; do round-robin */
 740                 if (!next || next->rt6i_metric != rt0->rt6i_metric)
 741                         next = fn->leaf;
 742
 743                 if (next != rt0)
 744                         fn->rr_ptr = next;
 745         }
 746
 747         net = dev_net(rt0->dst.dev);
 748         return match ? match : net->ipv6.ip6_null_entry;
 749 }
 750
 751 static bool rt6_is_gw_or_nonexthop(const struct rt6_info *rt)
 752 {
 753         return (rt->rt6i_flags & (RTF_NONEXTHOP | RTF_GATEWAY));
 754 }
 755
 756 #ifdef CONFIG_IPV6_ROUTE_INFO
 757 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
 758                   const struct in6_addr *gwaddr)
 759 {
 760         struct net *net = dev_net(dev);
 761         struct route_info *rinfo = (struct route_info *) opt;
 762         struct in6_addr prefix_buf, *prefix;
 763         unsigned int pref;
 764         unsigned long lifetime;
 765         struct rt6_info *rt;
 766
 767         if (len < sizeof(struct route_info)) {
 768                 return -EINVAL;
 769         }
 770
 771         /* Sanity check for prefix_len and length */
 772         if (rinfo->length > 3) {
 773                 return -EINVAL;
 774         } else if (rinfo->prefix_len > 128) {
 775                 return -EINVAL;
 776         } else if (rinfo->prefix_len > 64) {
 777                 if (rinfo->length < 2) {
 778                         return -EINVAL;
 779                 }
 780         } else if (rinfo->prefix_len > 0) {
 781                 if (rinfo->length < 1) {
 782                         return -EINVAL;
 783                 }
 784         }
 785
 786         pref = rinfo->route_pref;
 787         if (pref == ICMPV6_ROUTER_PREF_INVALID)
 788                 return -EINVAL;
 789
 790         lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
 791
 792         if (rinfo->length == 3)
 793                 prefix = (struct in6_addr *)rinfo->prefix;
 794         else {
 795                 /* this function is safe */
 796                 ipv6_addr_prefix(&prefix_buf,
 797                                  (struct in6_addr *)rinfo->prefix,
 798                                  rinfo->prefix_len);
 799                 prefix = &prefix_buf;
 800         }
 801
 802         if (rinfo->prefix_len == 0)
 803                 rt = rt6_get_dflt_router(gwaddr, dev);
 804         else
 805                 rt = rt6_get_route_info(net, prefix, rinfo->prefix_len,
 806                                         gwaddr, dev->ifindex);
 807
 808         if (rt && !lifetime) {
 809                 ip6_del_rt(rt);
 810                 rt = NULL;
 811         }
 812
 813         if (!rt && lifetime)
 814                 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr, dev->ifindex,
 815                                         pref);
 816         else if (rt)
 817                 rt->rt6i_flags = RTF_ROUTEINFO |
 818                                  (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
 819
 820         if (rt) {
 821                 if (!addrconf_finite_timeout(lifetime))
 822                         rt6_clean_expires(rt);
 823                 else
 824                         rt6_set_expires(rt, jiffies + HZ * lifetime);
 825
 826                 ip6_rt_put(rt);
 827         }
 828         return 0;
 829 }
 830 #endif
 831
 832 static struct fib6_node* fib6_backtrack(struct fib6_node *fn,
 833                                         struct in6_addr *saddr)
 834 {
 835         struct fib6_node *pn;
 836         while (1) {
 837                 if (fn->fn_flags & RTN_TL_ROOT)
 838                         return NULL;
 839                 pn = fn->parent;
 840                 if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn)
 841                         fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr);
 842                 else
 843                         fn = pn;
 844                 if (fn->fn_flags & RTN_RTINFO)
 845                         return fn;
 846         }
 847 }
 848
 849 static struct rt6_info *ip6_pol_route_lookup(struct net *net,
 850                                              struct fib6_table *table,
 851                                              struct flowi6 *fl6, int flags)
 852 {
 853         struct fib6_node *fn;
 854         struct rt6_info *rt;
 855
 856         read_lock_bh(&table->tb6_lock);
 857         fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
 858 restart:
 859         rt = fn->leaf;
 860         rt = rt6_device_match(net, rt, &fl6->saddr, fl6->flowi6_oif, flags);
 861         if (rt->rt6i_nsiblings && fl6->flowi6_oif == 0)
 862                 rt = rt6_multipath_select(rt, fl6, fl6->flowi6_oif, flags);
 863         if (rt == net->ipv6.ip6_null_entry) {
 864                 fn = fib6_backtrack(fn, &fl6->saddr);
 865                 if (fn)
 866                         goto restart;
 867         }
 868         dst_use(&rt->dst, jiffies);
 869         read_unlock_bh(&table->tb6_lock);
 870
 871         trace_fib6_table_lookup(net, rt, table->tb6_id, fl6);
 872
 873         return rt;
 874
 875 }
 876
 877 struct dst_entry *ip6_route_lookup(struct net *net, struct flowi6 *fl6,
 878                                     int flags)
 879 {
 880         return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_lookup);
 881 }
 882 EXPORT_SYMBOL_GPL(ip6_route_lookup);
 883
 884 struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
 885                             const struct in6_addr *saddr, int oif, int strict)
 886 {
 887         struct flowi6 fl6 = {
 888                 .flowi6_oif = oif,
 889                 .daddr = *daddr,
 890         };
 891         struct dst_entry *dst;
 892         int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
 893
 894         if (saddr) {
 895                 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
 896                 flags |= RT6_LOOKUP_F_HAS_SADDR;
 897         }
 898
 899         dst = fib6_rule_lookup(net, &fl6, flags, ip6_pol_route_lookup);
 900         if (dst->error == 0)
 901                 return (struct rt6_info *) dst;
 902
 903         dst_release(dst);
 904
 905         return NULL;
 906 }
 907 EXPORT_SYMBOL(rt6_lookup);
 908
 909 /* ip6_ins_rt is called with FREE table->tb6_lock.
 910    It takes new route entry, the addition fails by any reason the
 911    route is freed. In any case, if caller does not hold it, it may
 912    be destroyed.
 913  */
 914
 915 static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info,
 916                         struct mx6_config *mxc)
 917 {
 918         int err;
 919         struct fib6_table *table;
 920
 921         table = rt->rt6i_table;
 922         write_lock_bh(&table->tb6_lock);
 923         err = fib6_add(&table->tb6_root, rt, info, mxc);
 924         write_unlock_bh(&table->tb6_lock);
 925
 926         return err;
 927 }
 928
 929 int ip6_ins_rt(struct rt6_info *rt)
 930 {
 931         struct nl_info info = { .nl_net = dev_net(rt->dst.dev), };
 932         struct mx6_config mxc = { .mx = NULL, };
 933
 934         return __ip6_ins_rt(rt, &info, &mxc);
 935 }
 936
 937 static struct rt6_info *ip6_rt_cache_alloc(struct rt6_info *ort,
 938                                            const struct in6_addr *daddr,
 939                                            const struct in6_addr *saddr)
 940 {
 941         struct rt6_info *rt;
 942
 943         /*
 944          *      Clone the route.
 945          */
 946
 947         if (ort->rt6i_flags & (RTF_CACHE | RTF_PCPU))
 948                 ort = (struct rt6_info *)ort->dst.from;
 949
 950         rt = __ip6_dst_alloc(dev_net(ort->dst.dev), ort->dst.dev, 0);
 951
 952         if (!rt)
 953                 return NULL;
 954
 955         ip6_rt_copy_init(rt, ort);
 956         rt->rt6i_flags |= RTF_CACHE;
 957         rt->rt6i_metric = 0;
 958         rt->dst.flags |= DST_HOST;
 959         rt->rt6i_dst.addr = *daddr;
 960         rt->rt6i_dst.plen = 128;
 961
 962         if (!rt6_is_gw_or_nonexthop(ort)) {
 963                 if (ort->rt6i_dst.plen != 128 &&
 964                     ipv6_addr_equal(&ort->rt6i_dst.addr, daddr))
 965                         rt->rt6i_flags |= RTF_ANYCAST;
 966 #ifdef CONFIG_IPV6_SUBTREES
 967                 if (rt->rt6i_src.plen && saddr) {
 968                         rt->rt6i_src.addr = *saddr;
 969                         rt->rt6i_src.plen = 128;
 970                 }
 971 #endif
 972         }
 973
 974         return rt;
 975 }
 976
 977 static struct rt6_info *ip6_rt_pcpu_alloc(struct rt6_info *rt)
 978 {
 979         struct rt6_info *pcpu_rt;
 980
 981         pcpu_rt = __ip6_dst_alloc(dev_net(rt->dst.dev),
 982                                   rt->dst.dev, rt->dst.flags);
 983
 984         if (!pcpu_rt)
 985                 return NULL;
 986         ip6_rt_copy_init(pcpu_rt, rt);
 987         pcpu_rt->rt6i_protocol = rt->rt6i_protocol;
 988         pcpu_rt->rt6i_flags |= RTF_PCPU;
 989         return pcpu_rt;
 990 }
 991
 992 /* It should be called with read_lock_bh(&tb6_lock) acquired */
 993 static struct rt6_info *rt6_get_pcpu_route(struct rt6_info *rt)
 994 {
 995         struct rt6_info *pcpu_rt, **p;
 996
 997         p = this_cpu_ptr(rt->rt6i_pcpu);
 998         pcpu_rt = *p;
 999
1000         if (pcpu_rt) {
1001                 dst_hold(&pcpu_rt->dst);
1002                 rt6_dst_from_metrics_check(pcpu_rt);
1003         }
1004         return pcpu_rt;
1005 }
1006
1007 static struct rt6_info *rt6_make_pcpu_route(struct rt6_info *rt)
1008 {
1009         struct fib6_table *table = rt->rt6i_table;
1010         struct rt6_info *pcpu_rt, *prev, **p;
1011
1012         pcpu_rt = ip6_rt_pcpu_alloc(rt);
1013         if (!pcpu_rt) {
1014                 struct net *net = dev_net(rt->dst.dev);
1015
1016                 dst_hold(&net->ipv6.ip6_null_entry->dst);
1017                 return net->ipv6.ip6_null_entry;
1018         }
1019
1020         read_lock_bh(&table->tb6_lock);
1021         if (rt->rt6i_pcpu) {
1022                 p = this_cpu_ptr(rt->rt6i_pcpu);
1023                 prev = cmpxchg(p, NULL, pcpu_rt);
1024                 if (prev) {
1025                         /* If someone did it before us, return prev instead */
1026                         dst_destroy(&pcpu_rt->dst);
1027                         pcpu_rt = prev;
1028                 }
1029         } else {
1030                 /* rt has been removed from the fib6 tree
1031                  * before we have a chance to acquire the read_lock.
1032                  * In this case, don't brother to create a pcpu rt
1033                  * since rt is going away anyway.  The next
1034                  * dst_check() will trigger a re-lookup.
1035                  */
1036                 dst_destroy(&pcpu_rt->dst);
1037                 pcpu_rt = rt;
1038         }
1039         dst_hold(&pcpu_rt->dst);
1040         rt6_dst_from_metrics_check(pcpu_rt);
1041         read_unlock_bh(&table->tb6_lock);
1042         return pcpu_rt;
1043 }
1044
1045 static struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table, int oif,
1046                                       struct flowi6 *fl6, int flags)
1047 {
1048         struct fib6_node *fn, *saved_fn;
1049         struct rt6_info *rt;
1050         int strict = 0;
1051
1052         strict |= flags & RT6_LOOKUP_F_IFACE;
1053         if (net->ipv6.devconf_all->forwarding == 0)
1054                 strict |= RT6_LOOKUP_F_REACHABLE;
1055
1056         read_lock_bh(&table->tb6_lock);
1057
1058         fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1059         saved_fn = fn;
1060
1061         if (fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF)
1062                 oif = 0;
1063
1064 redo_rt6_select:
1065         rt = rt6_select(fn, oif, strict);
1066         if (rt->rt6i_nsiblings)
1067                 rt = rt6_multipath_select(rt, fl6, oif, strict);
1068         if (rt == net->ipv6.ip6_null_entry) {
1069                 fn = fib6_backtrack(fn, &fl6->saddr);
1070                 if (fn)
1071                         goto redo_rt6_select;
1072                 else if (strict & RT6_LOOKUP_F_REACHABLE) {
1073                         /* also consider unreachable route */
1074                         strict &= ~RT6_LOOKUP_F_REACHABLE;
1075                         fn = saved_fn;
1076                         goto redo_rt6_select;
1077                 }
1078         }
1079
1080
1081         if (rt == net->ipv6.ip6_null_entry || (rt->rt6i_flags & RTF_CACHE)) {
1082                 dst_use(&rt->dst, jiffies);
1083                 read_unlock_bh(&table->tb6_lock);
1084
1085                 rt6_dst_from_metrics_check(rt);
1086
1087                 trace_fib6_table_lookup(net, rt, table->tb6_id, fl6);
1088                 return rt;
1089         } else if (unlikely((fl6->flowi6_flags & FLOWI_FLAG_KNOWN_NH) &&
1090                             !(rt->rt6i_flags & RTF_GATEWAY))) {
1091                 /* Create a RTF_CACHE clone which will not be
1092                  * owned by the fib6 tree.  It is for the special case where
1093                  * the daddr in the skb during the neighbor look-up is different
1094                  * from the fl6->daddr used to look-up route here.
1095                  */
1096
1097                 struct rt6_info *uncached_rt;
1098
1099                 dst_use(&rt->dst, jiffies);
1100                 read_unlock_bh(&table->tb6_lock);
1101
1102                 uncached_rt = ip6_rt_cache_alloc(rt, &fl6->daddr, NULL);
1103                 dst_release(&rt->dst);
1104
1105                 if (uncached_rt)
1106                         rt6_uncached_list_add(uncached_rt);
1107                 else
1108                         uncached_rt = net->ipv6.ip6_null_entry;
1109
1110                 dst_hold(&uncached_rt->dst);
1111
1112                 trace_fib6_table_lookup(net, uncached_rt, table->tb6_id, fl6);
1113                 return uncached_rt;
1114
1115         } else {
1116                 /* Get a percpu copy */
1117
1118                 struct rt6_info *pcpu_rt;
1119
1120                 rt->dst.lastuse = jiffies;
1121                 rt->dst.__use++;
1122                 pcpu_rt = rt6_get_pcpu_route(rt);
1123
1124                 if (pcpu_rt) {
1125                         read_unlock_bh(&table->tb6_lock);
1126                 } else {
1127                         /* We have to do the read_unlock first
1128                          * because rt6_make_pcpu_route() may trigger
1129                          * ip6_dst_gc() which will take the write_lock.
1130                          */
1131                         dst_hold(&rt->dst);
1132                         read_unlock_bh(&table->tb6_lock);
1133                         pcpu_rt = rt6_make_pcpu_route(rt);
1134                         dst_release(&rt->dst);
1135                 }
1136
1137                 trace_fib6_table_lookup(net, pcpu_rt, table->tb6_id, fl6);
1138                 return pcpu_rt;
1139
1140         }
1141 }
1142
1143 static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
1144                                             struct flowi6 *fl6, int flags)
1145 {
1146         return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, flags);
1147 }
1148
1149 static struct dst_entry *ip6_route_input_lookup(struct net *net,
1150                                                 struct net_device *dev,
1151                                                 struct flowi6 *fl6, int flags)
1152 {
1153         if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
1154                 flags |= RT6_LOOKUP_F_IFACE;
1155
1156         return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_input);
1157 }
1158
1159 void ip6_route_input(struct sk_buff *skb)
1160 {
1161         const struct ipv6hdr *iph = ipv6_hdr(skb);
1162         struct net *net = dev_net(skb->dev);
1163         int flags = RT6_LOOKUP_F_HAS_SADDR;
1164         struct ip_tunnel_info *tun_info;
1165         struct flowi6 fl6 = {
1166                 .flowi6_iif = l3mdev_fib_oif(skb->dev),
1167                 .daddr = iph->daddr,
1168                 .saddr = iph->saddr,
1169                 .flowlabel = ip6_flowinfo(iph),
1170                 .flowi6_mark = skb->mark,
1171                 .flowi6_proto = iph->nexthdr,
1172         };
1173
1174         tun_info = skb_tunnel_info(skb);
1175         if (tun_info && !(tun_info->mode & IP_TUNNEL_INFO_TX))
1176                 fl6.flowi6_tun_key.tun_id = tun_info->key.tun_id;
1177         skb_dst_drop(skb);
1178         skb_dst_set(skb, ip6_route_input_lookup(net, skb->dev, &fl6, flags));
1179 }
1180
1181 static struct rt6_info *ip6_pol_route_output(struct net *net, struct fib6_table *table,
1182                                              struct flowi6 *fl6, int flags)
1183 {
1184         return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, flags);
1185 }
1186
1187 struct dst_entry *ip6_route_output_flags(struct net *net, const struct sock *sk,
1188                                          struct flowi6 *fl6, int flags)
1189 {
1190         struct dst_entry *dst;
1191         bool any_src;
1192
1193         dst = l3mdev_rt6_dst_by_oif(net, fl6);
1194         if (dst)
1195                 return dst;
1196
1197         fl6->flowi6_iif = LOOPBACK_IFINDEX;
1198
1199         any_src = ipv6_addr_any(&fl6->saddr);
1200         if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr) ||
1201             (fl6->flowi6_oif && any_src))
1202                 flags |= RT6_LOOKUP_F_IFACE;
1203
1204         if (!any_src)
1205                 flags |= RT6_LOOKUP_F_HAS_SADDR;
1206         else if (sk)
1207                 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
1208
1209         return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_output);
1210 }
1211 EXPORT_SYMBOL_GPL(ip6_route_output_flags);
1212
1213 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
1214 {
1215         struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
1216         struct dst_entry *new = NULL;
1217
1218         rt = dst_alloc(&ip6_dst_blackhole_ops, ort->dst.dev, 1, DST_OBSOLETE_NONE, 0);
1219         if (rt) {
1220                 rt6_info_init(rt);
1221
1222                 new = &rt->dst;
1223                 new->__use = 1;
1224                 new->input = dst_discard;
1225                 new->output = dst_discard_out;
1226
1227                 dst_copy_metrics(new, &ort->dst);
1228                 rt->rt6i_idev = ort->rt6i_idev;
1229                 if (rt->rt6i_idev)
1230                         in6_dev_hold(rt->rt6i_idev);
1231
1232                 rt->rt6i_gateway = ort->rt6i_gateway;
1233                 rt->rt6i_flags = ort->rt6i_flags & ~RTF_PCPU;
1234                 rt->rt6i_metric = 0;
1235
1236                 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1237 #ifdef CONFIG_IPV6_SUBTREES
1238                 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1239 #endif
1240
1241                 dst_free(new);
1242         }
1243
1244         dst_release(dst_orig);
1245         return new ? new : ERR_PTR(-ENOMEM);
1246 }
1247
1248 /*
1249  *      Destination cache support functions
1250  */
1251
1252 static void rt6_dst_from_metrics_check(struct rt6_info *rt)
1253 {
1254         if (rt->dst.from &&
1255             dst_metrics_ptr(&rt->dst) != dst_metrics_ptr(rt->dst.from))
1256                 dst_init_metrics(&rt->dst, dst_metrics_ptr(rt->dst.from), true);
1257 }
1258
1259 static struct dst_entry *rt6_check(struct rt6_info *rt, u32 cookie)
1260 {
1261         if (!rt->rt6i_node || (rt->rt6i_node->fn_sernum != cookie))
1262                 return NULL;
1263
1264         if (rt6_check_expired(rt))
1265                 return NULL;
1266
1267         return &rt->dst;
1268 }
1269
1270 static struct dst_entry *rt6_dst_from_check(struct rt6_info *rt, u32 cookie)
1271 {
1272         if (!__rt6_check_expired(rt) &&
1273             rt->dst.obsolete == DST_OBSOLETE_FORCE_CHK &&
1274             rt6_check((struct rt6_info *)(rt->dst.from), cookie))
1275                 return &rt->dst;
1276         else
1277                 return NULL;
1278 }
1279
1280 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
1281 {
1282         struct rt6_info *rt;
1283
1284         rt = (struct rt6_info *) dst;
1285
1286         /* All IPV6 dsts are created with ->obsolete set to the value
1287          * DST_OBSOLETE_FORCE_CHK which forces validation calls down
1288          * into this function always.
1289          */
1290
1291         rt6_dst_from_metrics_check(rt);
1292
1293         if (rt->rt6i_flags & RTF_PCPU ||
1294             (unlikely(dst->flags & DST_NOCACHE) && rt->dst.from))
1295                 return rt6_dst_from_check(rt, cookie);
1296         else
1297                 return rt6_check(rt, cookie);
1298 }
1299
1300 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
1301 {
1302         struct rt6_info *rt = (struct rt6_info *) dst;
1303
1304         if (rt) {
1305                 if (rt->rt6i_flags & RTF_CACHE) {
1306                         if (rt6_check_expired(rt)) {
1307                                 ip6_del_rt(rt);
1308                                 dst = NULL;
1309                         }
1310                 } else {
1311                         dst_release(dst);
1312                         dst = NULL;
1313                 }
1314         }
1315         return dst;
1316 }
1317
1318 static void ip6_link_failure(struct sk_buff *skb)
1319 {
1320         struct rt6_info *rt;
1321
1322         icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
1323
1324         rt = (struct rt6_info *) skb_dst(skb);
1325         if (rt) {
1326                 if (rt->rt6i_flags & RTF_CACHE) {
1327                         dst_hold(&rt->dst);
1328                         ip6_del_rt(rt);
1329                 } else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT)) {
1330                         rt->rt6i_node->fn_sernum = -1;
1331                 }
1332         }
1333 }
1334
1335 static void rt6_do_update_pmtu(struct rt6_info *rt, u32 mtu)
1336 {
1337         struct net *net = dev_net(rt->dst.dev);
1338
1339         rt->rt6i_flags |= RTF_MODIFIED;
1340         rt->rt6i_pmtu = mtu;
1341         rt6_update_expires(rt, net->ipv6.sysctl.ip6_rt_mtu_expires);
1342 }
1343
1344 static bool rt6_cache_allowed_for_pmtu(const struct rt6_info *rt)
1345 {
1346         return !(rt->rt6i_flags & RTF_CACHE) &&
1347                 (rt->rt6i_flags & RTF_PCPU || rt->rt6i_node);
1348 }
1349
1350 static void __ip6_rt_update_pmtu(struct dst_entry *dst, const struct sock *sk,
1351                                  const struct ipv6hdr *iph, u32 mtu)
1352 {
1353         struct rt6_info *rt6 = (struct rt6_info *)dst;
1354
1355         if (rt6->rt6i_flags & RTF_LOCAL)
1356                 return;
1357
1358         dst_confirm(dst);
1359         mtu = max_t(u32, mtu, IPV6_MIN_MTU);
1360         if (mtu >= dst_mtu(dst))
1361                 return;
1362
1363         if (!rt6_cache_allowed_for_pmtu(rt6)) {
1364                 rt6_do_update_pmtu(rt6, mtu);
1365         } else {
1366                 const struct in6_addr *daddr, *saddr;
1367                 struct rt6_info *nrt6;
1368
1369                 if (iph) {
1370                         daddr = &iph->daddr;
1371                         saddr = &iph->saddr;
1372                 } else if (sk) {
1373                         daddr = &sk->sk_v6_daddr;
1374                         saddr = &inet6_sk(sk)->saddr;
1375                 } else {
1376                         return;
1377                 }
1378                 nrt6 = ip6_rt_cache_alloc(rt6, daddr, saddr);
1379                 if (nrt6) {
1380                         rt6_do_update_pmtu(nrt6, mtu);
1381
1382                         /* ip6_ins_rt(nrt6) will bump the
1383                          * rt6->rt6i_node->fn_sernum
1384                          * which will fail the next rt6_check() and
1385                          * invalidate the sk->sk_dst_cache.
1386                          */
1387                         ip6_ins_rt(nrt6);
1388                 }
1389         }
1390 }
1391
1392 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
1393                                struct sk_buff *skb, u32 mtu)
1394 {
1395         __ip6_rt_update_pmtu(dst, sk, skb ? ipv6_hdr(skb) : NULL, mtu);
1396 }
1397
1398 void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
1399                      int oif, u32 mark)
1400 {
1401         const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1402         struct dst_entry *dst;
1403         struct flowi6 fl6;
1404
1405         memset(&fl6, 0, sizeof(fl6));
1406         fl6.flowi6_oif = oif;
1407         fl6.flowi6_mark = mark ? mark : IP6_REPLY_MARK(net, skb->mark);
1408         fl6.daddr = iph->daddr;
1409         fl6.saddr = iph->saddr;
1410         fl6.flowlabel = ip6_flowinfo(iph);
1411
1412         dst = ip6_route_output(net, NULL, &fl6);
1413         if (!dst->error)
1414                 __ip6_rt_update_pmtu(dst, NULL, iph, ntohl(mtu));
1415         dst_release(dst);
1416 }
1417 EXPORT_SYMBOL_GPL(ip6_update_pmtu);
1418
1419 void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
1420 {
1421         struct dst_entry *dst;
1422
1423         ip6_update_pmtu(skb, sock_net(sk), mtu,
1424                         sk->sk_bound_dev_if, sk->sk_mark);
1425
1426         dst = __sk_dst_get(sk);
1427         if (!dst || !dst->obsolete ||
1428             dst->ops->check(dst, inet6_sk(sk)->dst_cookie))
1429                 return;
1430
1431         bh_lock_sock(sk);
1432         if (!sock_owned_by_user(sk) && !ipv6_addr_v4mapped(&sk->sk_v6_daddr))
1433                 ip6_datagram_dst_update(sk, false);
1434         bh_unlock_sock(sk);
1435 }
1436 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
1437
1438 /* Handle redirects */
1439 struct ip6rd_flowi {
1440         struct flowi6 fl6;
1441         struct in6_addr gateway;
1442 };
1443
1444 static struct rt6_info *__ip6_route_redirect(struct net *net,
1445                                              struct fib6_table *table,
1446                                              struct flowi6 *fl6,
1447                                              int flags)
1448 {
1449         struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
1450         struct rt6_info *rt;
1451         struct fib6_node *fn;
1452
1453         /* Get the "current" route for this destination and
1454          * check if the redirect has come from approriate router.
1455          *
1456          * RFC 4861 specifies that redirects should only be
1457          * accepted if they come from the nexthop to the target.
1458          * Due to the way the routes are chosen, this notion
1459          * is a bit fuzzy and one might need to check all possible
1460          * routes.
1461          */
1462
1463         read_lock_bh(&table->tb6_lock);
1464         fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1465 restart:
1466         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1467                 if (rt6_check_expired(rt))
1468                         continue;
1469                 if (rt->dst.error)
1470                         break;
1471                 if (!(rt->rt6i_flags & RTF_GATEWAY))
1472                         continue;
1473                 if (fl6->flowi6_oif != rt->dst.dev->ifindex)
1474                         continue;
1475                 if (!ipv6_addr_equal(&rdfl->gateway, &rt->rt6i_gateway))
1476                         continue;
1477                 break;
1478         }
1479
1480         if (!rt)
1481                 rt = net->ipv6.ip6_null_entry;
1482         else if (rt->dst.error) {
1483                 rt = net->ipv6.ip6_null_entry;
1484                 goto out;
1485         }
1486
1487         if (rt == net->ipv6.ip6_null_entry) {
1488                 fn = fib6_backtrack(fn, &fl6->saddr);
1489                 if (fn)
1490                         goto restart;
1491         }
1492
1493 out:
1494         dst_hold(&rt->dst);
1495
1496         read_unlock_bh(&table->tb6_lock);
1497
1498         trace_fib6_table_lookup(net, rt, table->tb6_id, fl6);
1499         return rt;
1500 };
1501
1502 static struct dst_entry *ip6_route_redirect(struct net *net,
1503                                         const struct flowi6 *fl6,
1504                                         const struct in6_addr *gateway)
1505 {
1506         int flags = RT6_LOOKUP_F_HAS_SADDR;
1507         struct ip6rd_flowi rdfl;
1508
1509         rdfl.fl6 = *fl6;
1510         rdfl.gateway = *gateway;
1511
1512         return fib6_rule_lookup(net, &rdfl.fl6,
1513                                 flags, __ip6_route_redirect);
1514 }
1515
1516 void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark)
1517 {
1518         const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1519         struct dst_entry *dst;
1520         struct flowi6 fl6;
1521
1522         memset(&fl6, 0, sizeof(fl6));
1523         fl6.flowi6_iif = LOOPBACK_IFINDEX;
1524         fl6.flowi6_oif = oif;
1525         fl6.flowi6_mark = mark;
1526         fl6.daddr = iph->daddr;
1527         fl6.saddr = iph->saddr;
1528         fl6.flowlabel = ip6_flowinfo(iph);
1529
1530         dst = ip6_route_redirect(net, &fl6, &ipv6_hdr(skb)->saddr);
1531         rt6_do_redirect(dst, NULL, skb);
1532         dst_release(dst);
1533 }
1534 EXPORT_SYMBOL_GPL(ip6_redirect);
1535
1536 void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif,
1537                             u32 mark)
1538 {
1539         const struct ipv6hdr *iph = ipv6_hdr(skb);
1540         const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
1541         struct dst_entry *dst;
1542         struct flowi6 fl6;
1543
1544         memset(&fl6, 0, sizeof(fl6));
1545         fl6.flowi6_iif = LOOPBACK_IFINDEX;
1546         fl6.flowi6_oif = oif;
1547         fl6.flowi6_mark = mark;
1548         fl6.daddr = msg->dest;
1549         fl6.saddr = iph->daddr;
1550
1551         dst = ip6_route_redirect(net, &fl6, &iph->saddr);
1552         rt6_do_redirect(dst, NULL, skb);
1553         dst_release(dst);
1554 }
1555
1556 void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
1557 {
1558         ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark);
1559 }
1560 EXPORT_SYMBOL_GPL(ip6_sk_redirect);
1561
1562 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1563 {
1564         struct net_device *dev = dst->dev;
1565         unsigned int mtu = dst_mtu(dst);
1566         struct net *net = dev_net(dev);
1567
1568         mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
1569
1570         if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
1571                 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1572
1573         /*
1574          * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1575          * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1576          * IPV6_MAXPLEN is also valid and means: "any MSS,
1577          * rely only on pmtu discovery"
1578          */
1579         if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
1580                 mtu = IPV6_MAXPLEN;
1581         return mtu;
1582 }
1583
1584 static unsigned int ip6_mtu(const struct dst_entry *dst)
1585 {
1586         const struct rt6_info *rt = (const struct rt6_info *)dst;
1587         unsigned int mtu = rt->rt6i_pmtu;
1588         struct inet6_dev *idev;
1589
1590         if (mtu)
1591                 goto out;
1592
1593         mtu = dst_metric_raw(dst, RTAX_MTU);
1594         if (mtu)
1595                 goto out;
1596
1597         mtu = IPV6_MIN_MTU;
1598
1599         rcu_read_lock();
1600         idev = __in6_dev_get(dst->dev);
1601         if (idev)
1602                 mtu = idev->cnf.mtu6;
1603         rcu_read_unlock();
1604
1605 out:
1606         return min_t(unsigned int, mtu, IP6_MAX_MTU);
1607 }
1608
1609 static struct dst_entry *icmp6_dst_gc_list;
1610 static DEFINE_SPINLOCK(icmp6_dst_lock);
1611
1612 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
1613                                   struct flowi6 *fl6)
1614 {
1615         struct dst_entry *dst;
1616         struct rt6_info *rt;
1617         struct inet6_dev *idev = in6_dev_get(dev);
1618         struct net *net = dev_net(dev);
1619
1620         if (unlikely(!idev))
1621                 return ERR_PTR(-ENODEV);
1622
1623         rt = ip6_dst_alloc(net, dev, 0);
1624         if (unlikely(!rt)) {
1625                 in6_dev_put(idev);
1626                 dst = ERR_PTR(-ENOMEM);
1627                 goto out;
1628         }
1629
1630         rt->dst.flags |= DST_HOST;
1631         rt->dst.output  = ip6_output;
1632         atomic_set(&rt->dst.__refcnt, 1);
1633         rt->rt6i_gateway  = fl6->daddr;
1634         rt->rt6i_dst.addr = fl6->daddr;
1635         rt->rt6i_dst.plen = 128;
1636         rt->rt6i_idev     = idev;
1637         dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
1638
1639         spin_lock_bh(&icmp6_dst_lock);
1640         rt->dst.next = icmp6_dst_gc_list;
1641         icmp6_dst_gc_list = &rt->dst;
1642         spin_unlock_bh(&icmp6_dst_lock);
1643
1644         fib6_force_start_gc(net);
1645
1646         dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
1647
1648 out:
1649         return dst;
1650 }
1651
1652 int icmp6_dst_gc(void)
1653 {
1654         struct dst_entry *dst, **pprev;
1655         int more = 0;
1656
1657         spin_lock_bh(&icmp6_dst_lock);
1658         pprev = &icmp6_dst_gc_list;
1659
1660         while ((dst = *pprev) != NULL) {
1661                 if (!atomic_read(&dst->__refcnt)) {
1662                         *pprev = dst->next;
1663                         dst_free(dst);
1664                 } else {
1665                         pprev = &dst->next;
1666                         ++more;
1667                 }
1668         }
1669
1670         spin_unlock_bh(&icmp6_dst_lock);
1671
1672         return more;
1673 }
1674
1675 static void icmp6_clean_all(int (*func)(struct rt6_info *rt, void *arg),
1676                             void *arg)
1677 {
1678         struct dst_entry *dst, **pprev;
1679
1680         spin_lock_bh(&icmp6_dst_lock);
1681         pprev = &icmp6_dst_gc_list;
1682         while ((dst = *pprev) != NULL) {
1683                 struct rt6_info *rt = (struct rt6_info *) dst;
1684                 if (func(rt, arg)) {
1685                         *pprev = dst->next;
1686                         dst_free(dst);
1687                 } else {
1688                         pprev = &dst->next;
1689                 }
1690         }
1691         spin_unlock_bh(&icmp6_dst_lock);
1692 }
1693
1694 static int ip6_dst_gc(struct dst_ops *ops)
1695 {
1696         struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
1697         int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1698         int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1699         int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1700         int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1701         unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1702         int entries;
1703
1704         entries = dst_entries_get_fast(ops);
1705         if (time_after(rt_last_gc + rt_min_interval, jiffies) &&
1706             entries <= rt_max_size)
1707                 goto out;
1708
1709         net->ipv6.ip6_rt_gc_expire++;
1710         fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net, true);
1711         entries = dst_entries_get_slow(ops);
1712         if (entries < ops->gc_thresh)
1713                 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1714 out:
1715         net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1716         return entries > rt_max_size;
1717 }
1718
1719 static int ip6_convert_metrics(struct mx6_config *mxc,
1720                                const struct fib6_config *cfg)
1721 {
1722         bool ecn_ca = false;
1723         struct nlattr *nla;
1724         int remaining;
1725         u32 *mp;
1726
1727         if (!cfg->fc_mx)
1728                 return 0;
1729
1730         mp = kzalloc(sizeof(u32) * RTAX_MAX, GFP_KERNEL);
1731         if (unlikely(!mp))
1732                 return -ENOMEM;
1733
1734         nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1735                 int type = nla_type(nla);
1736                 u32 val;
1737
1738                 if (!type)
1739                         continue;
1740                 if (unlikely(type > RTAX_MAX))
1741                         goto err;
1742
1743                 if (type == RTAX_CC_ALGO) {
1744                         char tmp[TCP_CA_NAME_MAX];
1745
1746                         nla_strlcpy(tmp, nla, sizeof(tmp));
1747                         val = tcp_ca_get_key_by_name(tmp, &ecn_ca);
1748                         if (val == TCP_CA_UNSPEC)
1749                                 goto err;
1750                 } else {
1751                         val = nla_get_u32(nla);
1752                 }
1753                 if (type == RTAX_FEATURES && (val & ~RTAX_FEATURE_MASK))
1754                         goto err;
1755
1756                 mp[type - 1] = val;
1757                 __set_bit(type - 1, mxc->mx_valid);
1758         }
1759
1760         if (ecn_ca) {
1761                 __set_bit(RTAX_FEATURES - 1, mxc->mx_valid);
1762                 mp[RTAX_FEATURES - 1] |= DST_FEATURE_ECN_CA;
1763         }
1764
1765         mxc->mx = mp;
1766         return 0;
1767  err:
1768         kfree(mp);
1769         return -EINVAL;
1770 }
1771
1772 static struct rt6_info *ip6_nh_lookup_table(struct net *net,
1773                                             struct fib6_config *cfg,
1774                                             const struct in6_addr *gw_addr)
1775 {
1776         struct flowi6 fl6 = {
1777                 .flowi6_oif = cfg->fc_ifindex,
1778                 .daddr = *gw_addr,
1779                 .saddr = cfg->fc_prefsrc,
1780         };
1781         struct fib6_table *table;
1782         struct rt6_info *rt;
1783         int flags = 0;
1784
1785         table = fib6_get_table(net, cfg->fc_table);
1786         if (!table)
1787                 return NULL;
1788
1789         if (!ipv6_addr_any(&cfg->fc_prefsrc))
1790                 flags |= RT6_LOOKUP_F_HAS_SADDR;
1791
1792         rt = ip6_pol_route(net, table, cfg->fc_ifindex, &fl6, flags);
1793
1794         /* if table lookup failed, fall back to full lookup */
1795         if (rt == net->ipv6.ip6_null_entry) {
1796                 ip6_rt_put(rt);
1797                 rt = NULL;
1798         }
1799
1800         return rt;
1801 }
1802
1803 static struct rt6_info *ip6_route_info_create(struct fib6_config *cfg)
1804 {
1805         struct net *net = cfg->fc_nlinfo.nl_net;
1806         struct rt6_info *rt = NULL;
1807         struct net_device *dev = NULL;
1808         struct inet6_dev *idev = NULL;
1809         struct fib6_table *table;
1810         int addr_type;
1811         int err = -EINVAL;
1812
1813         if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1814                 goto out;
1815 #ifndef CONFIG_IPV6_SUBTREES
1816         if (cfg->fc_src_len)
1817                 goto out;
1818 #endif
1819         if (cfg->fc_ifindex) {
1820                 err = -ENODEV;
1821                 dev = dev_get_by_index(net, cfg->fc_ifindex);
1822                 if (!dev)
1823                         goto out;
1824                 idev = in6_dev_get(dev);
1825                 if (!idev)
1826                         goto out;
1827         }
1828
1829         if (cfg->fc_metric == 0)
1830                 cfg->fc_metric = IP6_RT_PRIO_USER;
1831
1832         err = -ENOBUFS;
1833         if (cfg->fc_nlinfo.nlh &&
1834             !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
1835                 table = fib6_get_table(net, cfg->fc_table);
1836                 if (!table) {
1837                         pr_warn("NLM_F_CREATE should be specified when creating new route\n");
1838                         table = fib6_new_table(net, cfg->fc_table);
1839                 }
1840         } else {
1841                 table = fib6_new_table(net, cfg->fc_table);
1842         }
1843
1844         if (!table)
1845                 goto out;
1846
1847         rt = ip6_dst_alloc(net, NULL,
1848                            (cfg->fc_flags & RTF_ADDRCONF) ? 0 : DST_NOCOUNT);
1849
1850         if (!rt) {
1851                 err = -ENOMEM;
1852                 goto out;
1853         }
1854
1855         if (cfg->fc_flags & RTF_EXPIRES)
1856                 rt6_set_expires(rt, jiffies +
1857                                 clock_t_to_jiffies(cfg->fc_expires));
1858         else
1859                 rt6_clean_expires(rt);
1860
1861         if (cfg->fc_protocol == RTPROT_UNSPEC)
1862                 cfg->fc_protocol = RTPROT_BOOT;
1863         rt->rt6i_protocol = cfg->fc_protocol;
1864
1865         addr_type = ipv6_addr_type(&cfg->fc_dst);
1866
1867         if (addr_type & IPV6_ADDR_MULTICAST)
1868                 rt->dst.input = ip6_mc_input;
1869         else if (cfg->fc_flags & RTF_LOCAL)
1870                 rt->dst.input = ip6_input;
1871         else
1872                 rt->dst.input = ip6_forward;
1873
1874         rt->dst.output = ip6_output;
1875
1876         if (cfg->fc_encap) {
1877                 struct lwtunnel_state *lwtstate;
1878
1879                 err = lwtunnel_build_state(dev, cfg->fc_encap_type,
1880                                            cfg->fc_encap, AF_INET6, cfg,
1881                                            &lwtstate);
1882                 if (err)
1883                         goto out;
1884                 rt->dst.lwtstate = lwtstate_get(lwtstate);
1885                 if (lwtunnel_output_redirect(rt->dst.lwtstate)) {
1886                         rt->dst.lwtstate->orig_output = rt->dst.output;
1887                         rt->dst.output = lwtunnel_output;
1888                 }
1889                 if (lwtunnel_input_redirect(rt->dst.lwtstate)) {
1890                         rt->dst.lwtstate->orig_input = rt->dst.input;
1891                         rt->dst.input = lwtunnel_input;
1892                 }
1893         }
1894
1895         ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1896         rt->rt6i_dst.plen = cfg->fc_dst_len;
1897         if (rt->rt6i_dst.plen == 128)
1898                 rt->dst.flags |= DST_HOST;
1899
1900 #ifdef CONFIG_IPV6_SUBTREES
1901         ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1902         rt->rt6i_src.plen = cfg->fc_src_len;
1903 #endif
1904
1905         rt->rt6i_metric = cfg->fc_metric;
1906
1907         /* We cannot add true routes via loopback here,
1908            they would result in kernel looping; promote them to reject routes
1909          */
1910         if ((cfg->fc_flags & RTF_REJECT) ||
1911             (dev && (dev->flags & IFF_LOOPBACK) &&
1912              !(addr_type & IPV6_ADDR_LOOPBACK) &&
1913              !(cfg->fc_flags & RTF_LOCAL))) {
1914                 /* hold loopback dev/idev if we haven't done so. */
1915                 if (dev != net->loopback_dev) {
1916                         if (dev) {
1917                                 dev_put(dev);
1918                                 in6_dev_put(idev);
1919                         }
1920                         dev = net->loopback_dev;
1921                         dev_hold(dev);
1922                         idev = in6_dev_get(dev);
1923                         if (!idev) {
1924                                 err = -ENODEV;
1925                                 goto out;
1926                         }
1927                 }
1928                 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1929                 switch (cfg->fc_type) {
1930                 case RTN_BLACKHOLE:
1931                         rt->dst.error = -EINVAL;
1932                         rt->dst.output = dst_discard_out;
1933                         rt->dst.input = dst_discard;
1934                         break;
1935                 case RTN_PROHIBIT:
1936                         rt->dst.error = -EACCES;
1937                         rt->dst.output = ip6_pkt_prohibit_out;
1938                         rt->dst.input = ip6_pkt_prohibit;
1939                         break;
1940                 case RTN_THROW:
1941                 case RTN_UNREACHABLE:
1942                 default:
1943                         rt->dst.error = (cfg->fc_type == RTN_THROW) ? -EAGAIN
1944                                         : (cfg->fc_type == RTN_UNREACHABLE)
1945                                         ? -EHOSTUNREACH : -ENETUNREACH;
1946                         rt->dst.output = ip6_pkt_discard_out;
1947                         rt->dst.input = ip6_pkt_discard;
1948                         break;
1949                 }
1950                 goto install_route;
1951         }
1952
1953         if (cfg->fc_flags & RTF_GATEWAY) {
1954                 const struct in6_addr *gw_addr;
1955                 int gwa_type;
1956
1957                 gw_addr = &cfg->fc_gateway;
1958                 gwa_type = ipv6_addr_type(gw_addr);
1959
1960                 /* if gw_addr is local we will fail to detect this in case
1961                  * address is still TENTATIVE (DAD in progress). rt6_lookup()
1962                  * will return already-added prefix route via interface that
1963                  * prefix route was assigned to, which might be non-loopback.
1964                  */
1965                 err = -EINVAL;
1966                 if (ipv6_chk_addr_and_flags(net, gw_addr,
1967                                             gwa_type & IPV6_ADDR_LINKLOCAL ?
1968                                             dev : NULL, 0, 0))
1969                         goto out;
1970
1971                 rt->rt6i_gateway = *gw_addr;
1972
1973                 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1974                         struct rt6_info *grt = NULL;
1975
1976                         /* IPv6 strictly inhibits using not link-local
1977                            addresses as nexthop address.
1978                            Otherwise, router will not able to send redirects.
1979                            It is very good, but in some (rare!) circumstances
1980                            (SIT, PtP, NBMA NOARP links) it is handy to allow
1981                            some exceptions. --ANK
1982                          */
1983                         if (!(gwa_type & IPV6_ADDR_UNICAST))
1984                                 goto out;
1985
1986                         if (cfg->fc_table)
1987                                 grt = ip6_nh_lookup_table(net, cfg, gw_addr);
1988
1989                         if (!grt)
1990                                 grt = rt6_lookup(net, gw_addr, NULL,
1991                                                  cfg->fc_ifindex, 1);
1992
1993                         err = -EHOSTUNREACH;
1994                         if (!grt)
1995                                 goto out;
1996                         if (dev) {
1997                                 if (dev != grt->dst.dev) {
1998                                         ip6_rt_put(grt);
1999                                         goto out;
2000                                 }
2001                         } else {
2002                                 dev = grt->dst.dev;
2003                                 idev = grt->rt6i_idev;
2004                                 dev_hold(dev);
2005                                 in6_dev_hold(grt->rt6i_idev);
2006                         }
2007                         if (!(grt->rt6i_flags & RTF_GATEWAY))
2008                                 err = 0;
2009                         ip6_rt_put(grt);
2010
2011                         if (err)
2012                                 goto out;
2013                 }
2014                 err = -EINVAL;
2015                 if (!dev || (dev->flags & IFF_LOOPBACK))
2016                         goto out;
2017         }
2018
2019         err = -ENODEV;
2020         if (!dev)
2021                 goto out;
2022
2023         if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
2024                 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
2025                         err = -EINVAL;
2026                         goto out;
2027                 }
2028                 rt->rt6i_prefsrc.addr = cfg->fc_prefsrc;
2029                 rt->rt6i_prefsrc.plen = 128;
2030         } else
2031                 rt->rt6i_prefsrc.plen = 0;
2032
2033         rt->rt6i_flags = cfg->fc_flags;
2034
2035 install_route:
2036         rt->dst.dev = dev;
2037         rt->rt6i_idev = idev;
2038         rt->rt6i_table = table;
2039
2040         cfg->fc_nlinfo.nl_net = dev_net(dev);
2041
2042         return rt;
2043 out:
2044         if (dev)
2045                 dev_put(dev);
2046         if (idev)
2047                 in6_dev_put(idev);
2048         if (rt)
2049                 dst_free(&rt->dst);
2050
2051         return ERR_PTR(err);
2052 }
2053
2054 int ip6_route_add(struct fib6_config *cfg)
2055 {
2056         struct mx6_config mxc = { .mx = NULL, };
2057         struct rt6_info *rt;
2058         int err;
2059
2060         rt = ip6_route_info_create(cfg);
2061         if (IS_ERR(rt)) {
2062                 err = PTR_ERR(rt);
2063                 rt = NULL;
2064                 goto out;
2065         }
2066
2067         err = ip6_convert_metrics(&mxc, cfg);
2068         if (err)
2069                 goto out;
2070
2071         err = __ip6_ins_rt(rt, &cfg->fc_nlinfo, &mxc);
2072
2073         kfree(mxc.mx);
2074
2075         return err;
2076 out:
2077         if (rt)
2078                 dst_free(&rt->dst);
2079
2080         return err;
2081 }
2082
2083 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
2084 {
2085         int err;
2086         struct fib6_table *table;
2087         struct net *net = dev_net(rt->dst.dev);
2088
2089         if (rt == net->ipv6.ip6_null_entry ||
2090             rt->dst.flags & DST_NOCACHE) {
2091                 err = -ENOENT;
2092                 goto out;
2093         }
2094
2095         table = rt->rt6i_table;
2096         write_lock_bh(&table->tb6_lock);
2097         err = fib6_del(rt, info);
2098         write_unlock_bh(&table->tb6_lock);
2099
2100 out:
2101         ip6_rt_put(rt);
2102         return err;
2103 }
2104
2105 int ip6_del_rt(struct rt6_info *rt)
2106 {
2107         struct nl_info info = {
2108                 .nl_net = dev_net(rt->dst.dev),
2109         };
2110         return __ip6_del_rt(rt, &info);
2111 }
2112
2113 static int ip6_route_del(struct fib6_config *cfg)
2114 {
2115         struct fib6_table *table;
2116         struct fib6_node *fn;
2117         struct rt6_info *rt;
2118         int err = -ESRCH;
2119
2120         table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
2121         if (!table)
2122                 return err;
2123
2124         read_lock_bh(&table->tb6_lock);
2125
2126         fn = fib6_locate(&table->tb6_root,
2127                          &cfg->fc_dst, cfg->fc_dst_len,
2128                          &cfg->fc_src, cfg->fc_src_len);
2129
2130         if (fn) {
2131                 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2132                         if ((rt->rt6i_flags & RTF_CACHE) &&
2133                             !(cfg->fc_flags & RTF_CACHE))
2134                                 continue;
2135                         if (cfg->fc_ifindex &&
2136                             (!rt->dst.dev ||
2137                              rt->dst.dev->ifindex != cfg->fc_ifindex))
2138                                 continue;
2139                         if (cfg->fc_flags & RTF_GATEWAY &&
2140                             !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
2141                                 continue;
2142                         if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
2143                                 continue;
2144                         dst_hold(&rt->dst);
2145                         read_unlock_bh(&table->tb6_lock);
2146
2147                         return __ip6_del_rt(rt, &cfg->fc_nlinfo);
2148                 }
2149         }
2150         read_unlock_bh(&table->tb6_lock);
2151
2152         return err;
2153 }
2154
2155 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
2156 {
2157         struct netevent_redirect netevent;
2158         struct rt6_info *rt, *nrt = NULL;
2159         struct ndisc_options ndopts;
2160         struct inet6_dev *in6_dev;
2161         struct neighbour *neigh;
2162         struct rd_msg *msg;
2163         int optlen, on_link;
2164         u8 *lladdr;
2165
2166         optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
2167         optlen -= sizeof(*msg);
2168
2169         if (optlen < 0) {
2170                 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
2171                 return;
2172         }
2173
2174         msg = (struct rd_msg *)icmp6_hdr(skb);
2175
2176         if (ipv6_addr_is_multicast(&msg->dest)) {
2177                 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
2178                 return;
2179         }
2180
2181         on_link = 0;
2182         if (ipv6_addr_equal(&msg->dest, &msg->target)) {
2183                 on_link = 1;
2184         } else if (ipv6_addr_type(&msg->target) !=
2185                    (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
2186                 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
2187                 return;
2188         }
2189
2190         in6_dev = __in6_dev_get(skb->dev);
2191         if (!in6_dev)
2192                 return;
2193         if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
2194                 return;
2195
2196         /* RFC2461 8.1:
2197          *      The IP source address of the Redirect MUST be the same as the current
2198          *      first-hop router for the specified ICMP Destination Address.
2199          */
2200
2201         if (!ndisc_parse_options(msg->opt, optlen, &ndopts)) {
2202                 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
2203                 return;
2204         }
2205
2206         lladdr = NULL;
2207         if (ndopts.nd_opts_tgt_lladdr) {
2208                 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
2209                                              skb->dev);
2210                 if (!lladdr) {
2211                         net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
2212                         return;
2213                 }
2214         }
2215
2216         rt = (struct rt6_info *) dst;
2217         if (rt->rt6i_flags & RTF_REJECT) {
2218                 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
2219                 return;
2220         }
2221
2222         /* Redirect received -> path was valid.
2223          * Look, redirects are sent only in response to data packets,
2224          * so that this nexthop apparently is reachable. --ANK
2225          */
2226         dst_confirm(&rt->dst);
2227
2228         neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
2229         if (!neigh)
2230                 return;
2231
2232         /*
2233          *      We have finally decided to accept it.
2234          */
2235
2236         neigh_update(neigh, lladdr, NUD_STALE,
2237                      NEIGH_UPDATE_F_WEAK_OVERRIDE|
2238                      NEIGH_UPDATE_F_OVERRIDE|
2239                      (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
2240                                      NEIGH_UPDATE_F_ISROUTER))
2241                      );
2242
2243         nrt = ip6_rt_cache_alloc(rt, &msg->dest, NULL);
2244         if (!nrt)
2245                 goto out;
2246
2247         nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
2248         if (on_link)
2249                 nrt->rt6i_flags &= ~RTF_GATEWAY;
2250
2251         nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
2252
2253         if (ip6_ins_rt(nrt))
2254                 goto out;
2255
2256         netevent.old = &rt->dst;
2257         netevent.new = &nrt->dst;
2258         netevent.daddr = &msg->dest;
2259         netevent.neigh = neigh;
2260         call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
2261
2262         if (rt->rt6i_flags & RTF_CACHE) {
2263                 rt = (struct rt6_info *) dst_clone(&rt->dst);
2264                 ip6_del_rt(rt);
2265         }
2266
2267 out:
2268         neigh_release(neigh);
2269 }
2270
2271 /*
2272  *      Misc support functions
2273  */
2274
2275 static void rt6_set_from(struct rt6_info *rt, struct rt6_info *from)
2276 {
2277         BUG_ON(from->dst.from);
2278
2279         rt->rt6i_flags &= ~RTF_EXPIRES;
2280         dst_hold(&from->dst);
2281         rt->dst.from = &from->dst;
2282         dst_init_metrics(&rt->dst, dst_metrics_ptr(&from->dst), true);
2283 }
2284
2285 static void ip6_rt_copy_init(struct rt6_info *rt, struct rt6_info *ort)
2286 {
2287         rt->dst.input = ort->dst.input;
2288         rt->dst.output = ort->dst.output;
2289         rt->rt6i_dst = ort->rt6i_dst;
2290         rt->dst.error = ort->dst.error;
2291         rt->rt6i_idev = ort->rt6i_idev;
2292         if (rt->rt6i_idev)
2293                 in6_dev_hold(rt->rt6i_idev);
2294         rt->dst.lastuse = jiffies;
2295         rt->rt6i_gateway = ort->rt6i_gateway;
2296         rt->rt6i_flags = ort->rt6i_flags;
2297         rt6_set_from(rt, ort);
2298         rt->rt6i_metric = ort->rt6i_metric;
2299 #ifdef CONFIG_IPV6_SUBTREES
2300         rt->rt6i_src = ort->rt6i_src;
2301 #endif
2302         rt->rt6i_prefsrc = ort->rt6i_prefsrc;
2303         rt->rt6i_table = ort->rt6i_table;
2304         rt->dst.lwtstate = lwtstate_get(ort->dst.lwtstate);
2305 }
2306
2307 #ifdef CONFIG_IPV6_ROUTE_INFO
2308 static struct rt6_info *rt6_get_route_info(struct net *net,
2309                                            const struct in6_addr *prefix, int prefixlen,
2310                                            const struct in6_addr *gwaddr, int ifindex)
2311 {
2312         struct fib6_node *fn;
2313         struct rt6_info *rt = NULL;
2314         struct fib6_table *table;
2315
2316         table = fib6_get_table(net, RT6_TABLE_INFO);
2317         if (!table)
2318                 return NULL;
2319
2320         read_lock_bh(&table->tb6_lock);
2321         fn = fib6_locate(&table->tb6_root, prefix, prefixlen, NULL, 0);
2322         if (!fn)
2323                 goto out;
2324
2325         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2326                 if (rt->dst.dev->ifindex != ifindex)
2327                         continue;
2328                 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
2329                         continue;
2330                 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
2331                         continue;
2332                 dst_hold(&rt->dst);
2333                 break;
2334         }
2335 out:
2336         read_unlock_bh(&table->tb6_lock);
2337         return rt;
2338 }
2339
2340 static struct rt6_info *rt6_add_route_info(struct net *net,
2341                                            const struct in6_addr *prefix, int prefixlen,
2342                                            const struct in6_addr *gwaddr, int ifindex,
2343                                            unsigned int pref)
2344 {
2345         struct fib6_config cfg = {
2346                 .fc_metric      = IP6_RT_PRIO_USER,
2347                 .fc_ifindex     = ifindex,
2348                 .fc_dst_len     = prefixlen,
2349                 .fc_flags       = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
2350                                   RTF_UP | RTF_PREF(pref),
2351                 .fc_nlinfo.portid = 0,
2352                 .fc_nlinfo.nlh = NULL,
2353                 .fc_nlinfo.nl_net = net,
2354         };
2355
2356         cfg.fc_table = l3mdev_fib_table_by_index(net, ifindex) ? : RT6_TABLE_INFO;
2357         cfg.fc_dst = *prefix;
2358         cfg.fc_gateway = *gwaddr;
2359
2360         /* We should treat it as a default route if prefix length is 0. */
2361         if (!prefixlen)
2362                 cfg.fc_flags |= RTF_DEFAULT;
2363
2364         ip6_route_add(&cfg);
2365
2366         return rt6_get_route_info(net, prefix, prefixlen, gwaddr, ifindex);
2367 }
2368 #endif
2369
2370 struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
2371 {
2372         struct rt6_info *rt;
2373         struct fib6_table *table;
2374
2375         table = fib6_get_table(dev_net(dev), RT6_TABLE_DFLT);
2376         if (!table)
2377                 return NULL;
2378
2379         read_lock_bh(&table->tb6_lock);
2380         for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2381                 if (dev == rt->dst.dev &&
2382                     ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
2383                     ipv6_addr_equal(&rt->rt6i_gateway, addr))
2384                         break;
2385         }
2386         if (rt)
2387                 dst_hold(&rt->dst);
2388         read_unlock_bh(&table->tb6_lock);
2389         return rt;
2390 }
2391
2392 struct rt6_info *rt6_add_dflt_router(const struct in6_addr *gwaddr,
2393                                      struct net_device *dev,
2394                                      unsigned int pref)
2395 {
2396         struct fib6_config cfg = {
2397                 .fc_table       = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT,
2398                 .fc_metric      = IP6_RT_PRIO_USER,
2399                 .fc_ifindex     = dev->ifindex,
2400                 .fc_flags       = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
2401                                   RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
2402                 .fc_nlinfo.portid = 0,
2403                 .fc_nlinfo.nlh = NULL,
2404                 .fc_nlinfo.nl_net = dev_net(dev),
2405         };
2406
2407         cfg.fc_gateway = *gwaddr;
2408
2409         ip6_route_add(&cfg);
2410
2411         return rt6_get_dflt_router(gwaddr, dev);
2412 }
2413
2414 void rt6_purge_dflt_routers(struct net *net)
2415 {
2416         struct rt6_info *rt;
2417         struct fib6_table *table;
2418
2419         /* NOTE: Keep consistent with rt6_get_dflt_router */
2420         table = fib6_get_table(net, RT6_TABLE_DFLT);
2421         if (!table)
2422                 return;
2423
2424 restart:
2425         read_lock_bh(&table->tb6_lock);
2426         for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2427                 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
2428                     (!rt->rt6i_idev || rt->rt6i_idev->cnf.accept_ra != 2)) {
2429                         dst_hold(&rt->dst);
2430                         read_unlock_bh(&table->tb6_lock);
2431                         ip6_del_rt(rt);
2432                         goto restart;
2433                 }
2434         }
2435         read_unlock_bh(&table->tb6_lock);
2436 }
2437
2438 static void rtmsg_to_fib6_config(struct net *net,
2439                                  struct in6_rtmsg *rtmsg,
2440                                  struct fib6_config *cfg)
2441 {
2442         memset(cfg, 0, sizeof(*cfg));
2443
2444         cfg->fc_table = l3mdev_fib_table_by_index(net, rtmsg->rtmsg_ifindex) ?
2445                          : RT6_TABLE_MAIN;
2446         cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
2447         cfg->fc_metric = rtmsg->rtmsg_metric;
2448         cfg->fc_expires = rtmsg->rtmsg_info;
2449         cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
2450         cfg->fc_src_len = rtmsg->rtmsg_src_len;
2451         cfg->fc_flags = rtmsg->rtmsg_flags;
2452
2453         cfg->fc_nlinfo.nl_net = net;
2454
2455         cfg->fc_dst = rtmsg->rtmsg_dst;
2456         cfg->fc_src = rtmsg->rtmsg_src;
2457         cfg->fc_gateway = rtmsg->rtmsg_gateway;
2458 }
2459
2460 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
2461 {
2462         struct fib6_config cfg;
2463         struct in6_rtmsg rtmsg;
2464         int err;
2465
2466         switch (cmd) {
2467         case SIOCADDRT:         /* Add a route */
2468         case SIOCDELRT:         /* Delete a route */
2469                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2470                         return -EPERM;
2471                 err = copy_from_user(&rtmsg, arg,
2472                                      sizeof(struct in6_rtmsg));
2473                 if (err)
2474                         return -EFAULT;
2475
2476                 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
2477
2478                 rtnl_lock();
2479                 switch (cmd) {
2480                 case SIOCADDRT:
2481                         err = ip6_route_add(&cfg);
2482                         break;
2483                 case SIOCDELRT:
2484                         err = ip6_route_del(&cfg);
2485                         break;
2486                 default:
2487                         err = -EINVAL;
2488                 }
2489                 rtnl_unlock();
2490
2491                 return err;
2492         }
2493
2494         return -EINVAL;
2495 }
2496
2497 /*
2498  *      Drop the packet on the floor
2499  */
2500
2501 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
2502 {
2503         int type;
2504         struct dst_entry *dst = skb_dst(skb);
2505         switch (ipstats_mib_noroutes) {
2506         case IPSTATS_MIB_INNOROUTES:
2507                 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
2508                 if (type == IPV6_ADDR_ANY) {
2509                         IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2510                                       IPSTATS_MIB_INADDRERRORS);
2511                         break;
2512                 }
2513                 /* FALLTHROUGH */
2514         case IPSTATS_MIB_OUTNOROUTES:
2515                 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2516                               ipstats_mib_noroutes);
2517                 break;
2518         }
2519         icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
2520         kfree_skb(skb);
2521         return 0;
2522 }
2523
2524 static int ip6_pkt_discard(struct sk_buff *skb)
2525 {
2526         return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
2527 }
2528
2529 static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb)
2530 {
2531         skb->dev = skb_dst(skb)->dev;
2532         return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
2533 }
2534
2535 static int ip6_pkt_prohibit(struct sk_buff *skb)
2536 {
2537         return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
2538 }
2539
2540 static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb)
2541 {
2542         skb->dev = skb_dst(skb)->dev;
2543         return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
2544 }
2545
2546 /*
2547  *      Allocate a dst for local (unicast / anycast) address.
2548  */
2549
2550 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
2551                                     const struct in6_addr *addr,
2552                                     bool anycast)
2553 {
2554         u32 tb_id;
2555         struct net *net = dev_net(idev->dev);
2556         struct rt6_info *rt = ip6_dst_alloc(net, net->loopback_dev,
2557                                             DST_NOCOUNT);
2558         if (!rt)
2559                 return ERR_PTR(-ENOMEM);
2560
2561         in6_dev_hold(idev);
2562
2563         rt->dst.flags |= DST_HOST;
2564         rt->dst.input = ip6_input;
2565         rt->dst.output = ip6_output;
2566         rt->rt6i_idev = idev;
2567
2568         rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
2569         if (anycast)
2570                 rt->rt6i_flags |= RTF_ANYCAST;
2571         else
2572                 rt->rt6i_flags |= RTF_LOCAL;
2573
2574         rt->rt6i_gateway  = *addr;
2575         rt->rt6i_dst.addr = *addr;
2576         rt->rt6i_dst.plen = 128;
2577         tb_id = l3mdev_fib_table(idev->dev) ? : RT6_TABLE_LOCAL;
2578         rt->rt6i_table = fib6_get_table(net, tb_id);
2579         rt->dst.flags |= DST_NOCACHE;
2580
2581         atomic_set(&rt->dst.__refcnt, 1);
2582
2583         return rt;
2584 }
2585
2586 int ip6_route_get_saddr(struct net *net,
2587                         struct rt6_info *rt,
2588                         const struct in6_addr *daddr,
2589                         unsigned int prefs,
2590                         struct in6_addr *saddr)
2591 {
2592         struct inet6_dev *idev =
2593                 rt ? ip6_dst_idev((struct dst_entry *)rt) : NULL;
2594         int err = 0;
2595         if (rt && rt->rt6i_prefsrc.plen)
2596                 *saddr = rt->rt6i_prefsrc.addr;
2597         else
2598                 err = ipv6_dev_get_saddr(net, idev ? idev->dev : NULL,
2599                                          daddr, prefs, saddr);
2600         return err;
2601 }
2602
2603 /* remove deleted ip from prefsrc entries */
2604 struct arg_dev_net_ip {
2605         struct net_device *dev;
2606         struct net *net;
2607         struct in6_addr *addr;
2608 };
2609
2610 static int fib6_remove_prefsrc(struct rt6_info *rt, void *arg)
2611 {
2612         struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
2613         struct net *net = ((struct arg_dev_net_ip *)arg)->net;
2614         struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
2615
2616         if (((void *)rt->dst.dev == dev || !dev) &&
2617             rt != net->ipv6.ip6_null_entry &&
2618             ipv6_addr_equal(addr, &rt->rt6i_prefsrc.addr)) {
2619                 /* remove prefsrc entry */
2620                 rt->rt6i_prefsrc.plen = 0;
2621         }
2622         return 0;
2623 }
2624
2625 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
2626 {
2627         struct net *net = dev_net(ifp->idev->dev);
2628         struct arg_dev_net_ip adni = {
2629                 .dev = ifp->idev->dev,
2630                 .net = net,
2631                 .addr = &ifp->addr,
2632         };
2633         fib6_clean_all(net, fib6_remove_prefsrc, &adni);
2634 }
2635
2636 #define RTF_RA_ROUTER           (RTF_ADDRCONF | RTF_DEFAULT | RTF_GATEWAY)
2637 #define RTF_CACHE_GATEWAY       (RTF_GATEWAY | RTF_CACHE)
2638
2639 /* Remove routers and update dst entries when gateway turn into host. */
2640 static int fib6_clean_tohost(struct rt6_info *rt, void *arg)
2641 {
2642         struct in6_addr *gateway = (struct in6_addr *)arg;
2643
2644         if ((((rt->rt6i_flags & RTF_RA_ROUTER) == RTF_RA_ROUTER) ||
2645              ((rt->rt6i_flags & RTF_CACHE_GATEWAY) == RTF_CACHE_GATEWAY)) &&
2646              ipv6_addr_equal(gateway, &rt->rt6i_gateway)) {
2647                 return -1;
2648         }
2649         return 0;
2650 }
2651
2652 void rt6_clean_tohost(struct net *net, struct in6_addr *gateway)
2653 {
2654         fib6_clean_all(net, fib6_clean_tohost, gateway);
2655 }
2656
2657 struct arg_dev_net {
2658         struct net_device *dev;
2659         struct net *net;
2660 };
2661
2662 static int fib6_ifdown(struct rt6_info *rt, void *arg)
2663 {
2664         const struct arg_dev_net *adn = arg;
2665         const struct net_device *dev = adn->dev;
2666
2667         if ((rt->dst.dev == dev || !dev) &&
2668             rt != adn->net->ipv6.ip6_null_entry)
2669                 return -1;
2670
2671         return 0;
2672 }
2673
2674 void rt6_ifdown(struct net *net, struct net_device *dev)
2675 {
2676         struct arg_dev_net adn = {
2677                 .dev = dev,
2678                 .net = net,
2679         };
2680
2681         fib6_clean_all(net, fib6_ifdown, &adn);
2682         icmp6_clean_all(fib6_ifdown, &adn);
2683         if (dev)
2684                 rt6_uncached_list_flush_dev(net, dev);
2685 }
2686
2687 struct rt6_mtu_change_arg {
2688         struct net_device *dev;
2689         unsigned int mtu;
2690 };
2691
2692 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2693 {
2694         struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2695         struct inet6_dev *idev;
2696
2697         /* In IPv6 pmtu discovery is not optional,
2698            so that RTAX_MTU lock cannot disable it.
2699            We still use this lock to block changes
2700            caused by addrconf/ndisc.
2701         */
2702
2703         idev = __in6_dev_get(arg->dev);
2704         if (!idev)
2705                 return 0;
2706
2707         /* For administrative MTU increase, there is no way to discover
2708            IPv6 PMTU increase, so PMTU increase should be updated here.
2709            Since RFC 1981 doesn't include administrative MTU increase
2710            update PMTU increase is a MUST. (i.e. jumbo frame)
2711          */
2712         /*
2713            If new MTU is less than route PMTU, this new MTU will be the
2714            lowest MTU in the path, update the route PMTU to reflect PMTU
2715            decreases; if new MTU is greater than route PMTU, and the
2716            old MTU is the lowest MTU in the path, update the route PMTU
2717            to reflect the increase. In this case if the other nodes' MTU
2718            also have the lowest MTU, TOO BIG MESSAGE will be lead to
2719            PMTU discouvery.
2720          */
2721         if (rt->dst.dev == arg->dev &&
2722             !dst_metric_locked(&rt->dst, RTAX_MTU)) {
2723                 if (rt->rt6i_flags & RTF_CACHE) {
2724                         /* For RTF_CACHE with rt6i_pmtu == 0
2725                          * (i.e. a redirected route),
2726                          * the metrics of its rt->dst.from has already
2727                          * been updated.
2728                          */
2729                         if (rt->rt6i_pmtu && rt->rt6i_pmtu > arg->mtu)
2730                                 rt->rt6i_pmtu = arg->mtu;
2731                 } else if (dst_mtu(&rt->dst) >= arg->mtu ||
2732                            (dst_mtu(&rt->dst) < arg->mtu &&
2733                             dst_mtu(&rt->dst) == idev->cnf.mtu6)) {
2734                         dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
2735                 }
2736         }
2737         return 0;
2738 }
2739
2740 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
2741 {
2742         struct rt6_mtu_change_arg arg = {
2743                 .dev = dev,
2744                 .mtu = mtu,
2745         };
2746
2747         fib6_clean_all(dev_net(dev), rt6_mtu_change_route, &arg);
2748 }
2749
2750 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2751         [RTA_GATEWAY]           = { .len = sizeof(struct in6_addr) },
2752         [RTA_OIF]               = { .type = NLA_U32 },
2753         [RTA_IIF]               = { .type = NLA_U32 },
2754         [RTA_PRIORITY]          = { .type = NLA_U32 },
2755         [RTA_METRICS]           = { .type = NLA_NESTED },
2756         [RTA_MULTIPATH]         = { .len = sizeof(struct rtnexthop) },
2757         [RTA_PREF]              = { .type = NLA_U8 },
2758         [RTA_ENCAP_TYPE]        = { .type = NLA_U16 },
2759         [RTA_ENCAP]             = { .type = NLA_NESTED },
2760         [RTA_EXPIRES]           = { .type = NLA_U32 },
2761 };
2762
2763 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2764                               struct fib6_config *cfg)
2765 {
2766         struct rtmsg *rtm;
2767         struct nlattr *tb[RTA_MAX+1];
2768         unsigned int pref;
2769         int err;
2770
2771         err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2772         if (err < 0)
2773                 goto errout;
2774
2775         err = -EINVAL;
2776         rtm = nlmsg_data(nlh);
2777         memset(cfg, 0, sizeof(*cfg));
2778
2779         cfg->fc_table = rtm->rtm_table;
2780         cfg->fc_dst_len = rtm->rtm_dst_len;
2781         cfg->fc_src_len = rtm->rtm_src_len;
2782         cfg->fc_flags = RTF_UP;
2783         cfg->fc_protocol = rtm->rtm_protocol;
2784         cfg->fc_type = rtm->rtm_type;
2785
2786         if (rtm->rtm_type == RTN_UNREACHABLE ||
2787             rtm->rtm_type == RTN_BLACKHOLE ||
2788             rtm->rtm_type == RTN_PROHIBIT ||
2789             rtm->rtm_type == RTN_THROW)
2790                 cfg->fc_flags |= RTF_REJECT;
2791
2792         if (rtm->rtm_type == RTN_LOCAL)
2793                 cfg->fc_flags |= RTF_LOCAL;
2794
2795         if (rtm->rtm_flags & RTM_F_CLONED)
2796                 cfg->fc_flags |= RTF_CACHE;
2797
2798         cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid;
2799         cfg->fc_nlinfo.nlh = nlh;
2800         cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2801
2802         if (tb[RTA_GATEWAY]) {
2803                 cfg->fc_gateway = nla_get_in6_addr(tb[RTA_GATEWAY]);
2804                 cfg->fc_flags |= RTF_GATEWAY;
2805         }
2806
2807         if (tb[RTA_DST]) {
2808                 int plen = (rtm->rtm_dst_len + 7) >> 3;
2809
2810                 if (nla_len(tb[RTA_DST]) < plen)
2811                         goto errout;
2812
2813                 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2814         }
2815
2816         if (tb[RTA_SRC]) {
2817                 int plen = (rtm->rtm_src_len + 7) >> 3;
2818
2819                 if (nla_len(tb[RTA_SRC]) < plen)
2820                         goto errout;
2821
2822                 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2823         }
2824
2825         if (tb[RTA_PREFSRC])
2826                 cfg->fc_prefsrc = nla_get_in6_addr(tb[RTA_PREFSRC]);
2827
2828         if (tb[RTA_OIF])
2829                 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2830
2831         if (tb[RTA_PRIORITY])
2832                 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2833
2834         if (tb[RTA_METRICS]) {
2835                 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2836                 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2837         }
2838
2839         if (tb[RTA_TABLE])
2840                 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2841
2842         if (tb[RTA_MULTIPATH]) {
2843                 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
2844                 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
2845         }
2846
2847         if (tb[RTA_PREF]) {
2848                 pref = nla_get_u8(tb[RTA_PREF]);
2849                 if (pref != ICMPV6_ROUTER_PREF_LOW &&
2850                     pref != ICMPV6_ROUTER_PREF_HIGH)
2851                         pref = ICMPV6_ROUTER_PREF_MEDIUM;
2852                 cfg->fc_flags |= RTF_PREF(pref);
2853         }
2854
2855         if (tb[RTA_ENCAP])
2856                 cfg->fc_encap = tb[RTA_ENCAP];
2857
2858         if (tb[RTA_ENCAP_TYPE])
2859                 cfg->fc_encap_type = nla_get_u16(tb[RTA_ENCAP_TYPE]);
2860
2861         if (tb[RTA_EXPIRES]) {
2862                 unsigned long timeout = addrconf_timeout_fixup(nla_get_u32(tb[RTA_EXPIRES]), HZ);
2863
2864                 if (addrconf_finite_timeout(timeout)) {
2865                         cfg->fc_expires = jiffies_to_clock_t(timeout * HZ);
2866                         cfg->fc_flags |= RTF_EXPIRES;
2867                 }
2868         }
2869
2870         err = 0;
2871 errout:
2872         return err;
2873 }
2874
2875 struct rt6_nh {
2876         struct rt6_info *rt6_info;
2877         struct fib6_config r_cfg;
2878         struct mx6_config mxc;
2879         struct list_head next;
2880 };
2881
2882 static void ip6_print_replace_route_err(struct list_head *rt6_nh_list)
2883 {
2884         struct rt6_nh *nh;
2885
2886         list_for_each_entry(nh, rt6_nh_list, next) {
2887                 pr_warn("IPV6: multipath route replace failed (check consistency of installed routes): %pI6 nexthop %pI6 ifi %d\n",
2888                         &nh->r_cfg.fc_dst, &nh->r_cfg.fc_gateway,
2889                         nh->r_cfg.fc_ifindex);
2890         }
2891 }
2892
2893 static int ip6_route_info_append(struct list_head *rt6_nh_list,
2894                                  struct rt6_info *rt, struct fib6_config *r_cfg)
2895 {
2896         struct rt6_nh *nh;
2897         struct rt6_info *rtnh;
2898         int err = -EEXIST;
2899
2900         list_for_each_entry(nh, rt6_nh_list, next) {
2901                 /* check if rt6_info already exists */
2902                 rtnh = nh->rt6_info;
2903
2904                 if (rtnh->dst.dev == rt->dst.dev &&
2905                     rtnh->rt6i_idev == rt->rt6i_idev &&
2906                     ipv6_addr_equal(&rtnh->rt6i_gateway,
2907                                     &rt->rt6i_gateway))
2908                         return err;
2909         }
2910
2911         nh = kzalloc(sizeof(*nh), GFP_KERNEL);
2912         if (!nh)
2913                 return -ENOMEM;
2914         nh->rt6_info = rt;
2915         err = ip6_convert_metrics(&nh->mxc, r_cfg);
2916         if (err) {
2917                 kfree(nh);
2918                 return err;
2919         }
2920         memcpy(&nh->r_cfg, r_cfg, sizeof(*r_cfg));
2921         list_add_tail(&nh->next, rt6_nh_list);
2922
2923         return 0;
2924 }
2925
2926 static int ip6_route_multipath_add(struct fib6_config *cfg)
2927 {
2928         struct fib6_config r_cfg;
2929         struct rtnexthop *rtnh;
2930         struct rt6_info *rt;
2931         struct rt6_nh *err_nh;
2932         struct rt6_nh *nh, *nh_safe;
2933         int remaining;
2934         int attrlen;
2935         int err = 1;
2936         int nhn = 0;
2937         int replace = (cfg->fc_nlinfo.nlh &&
2938                        (cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_REPLACE));
2939         LIST_HEAD(rt6_nh_list);
2940
2941         remaining = cfg->fc_mp_len;
2942         rtnh = (struct rtnexthop *)cfg->fc_mp;
2943
2944         /* Parse a Multipath Entry and build a list (rt6_nh_list) of
2945          * rt6_info structs per nexthop
2946          */
2947         while (rtnh_ok(rtnh, remaining)) {
2948                 memcpy(&r_cfg, cfg, sizeof(*cfg));
2949                 if (rtnh->rtnh_ifindex)
2950                         r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
2951
2952                 attrlen = rtnh_attrlen(rtnh);
2953                 if (attrlen > 0) {
2954                         struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
2955
2956                         nla = nla_find(attrs, attrlen, RTA_GATEWAY);
2957                         if (nla) {
2958                                 r_cfg.fc_gateway = nla_get_in6_addr(nla);
2959                                 r_cfg.fc_flags |= RTF_GATEWAY;
2960                         }
2961                         r_cfg.fc_encap = nla_find(attrs, attrlen, RTA_ENCAP);
2962                         nla = nla_find(attrs, attrlen, RTA_ENCAP_TYPE);
2963                         if (nla)
2964                                 r_cfg.fc_encap_type = nla_get_u16(nla);
2965                 }
2966
2967                 rt = ip6_route_info_create(&r_cfg);
2968                 if (IS_ERR(rt)) {
2969                         err = PTR_ERR(rt);
2970                         rt = NULL;
2971                         goto cleanup;
2972                 }
2973
2974                 err = ip6_route_info_append(&rt6_nh_list, rt, &r_cfg);
2975                 if (err) {
2976                         dst_free(&rt->dst);
2977                         goto cleanup;
2978                 }
2979
2980                 rtnh = rtnh_next(rtnh, &remaining);
2981         }
2982
2983         err_nh = NULL;
2984         list_for_each_entry(nh, &rt6_nh_list, next) {
2985                 err = __ip6_ins_rt(nh->rt6_info, &cfg->fc_nlinfo, &nh->mxc);
2986                 /* nh->rt6_info is used or freed at this point, reset to NULL*/
2987                 nh->rt6_info = NULL;
2988                 if (err) {
2989                         if (replace && nhn)
2990                                 ip6_print_replace_route_err(&rt6_nh_list);
2991                         err_nh = nh;
2992                         goto add_errout;
2993                 }
2994
2995                 /* Because each route is added like a single route we remove
2996                  * these flags after the first nexthop: if there is a collision,
2997                  * we have already failed to add the first nexthop:
2998                  * fib6_add_rt2node() has rejected it; when replacing, old
2999                  * nexthops have been replaced by first new, the rest should
3000                  * be added to it.
3001                  */
3002                 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~(NLM_F_EXCL |
3003                                                      NLM_F_REPLACE);
3004                 nhn++;
3005         }
3006
3007         goto cleanup;
3008
3009 add_errout:
3010         /* Delete routes that were already added */
3011         list_for_each_entry(nh, &rt6_nh_list, next) {
3012                 if (err_nh == nh)
3013                         break;
3014                 ip6_route_del(&nh->r_cfg);
3015         }
3016
3017 cleanup:
3018         list_for_each_entry_safe(nh, nh_safe, &rt6_nh_list, next) {
3019                 if (nh->rt6_info)
3020                         dst_free(&nh->rt6_info->dst);
3021                 kfree(nh->mxc.mx);
3022                 list_del(&nh->next);
3023                 kfree(nh);
3024         }
3025
3026         return err;
3027 }
3028
3029 static int ip6_route_multipath_del(struct fib6_config *cfg)
3030 {
3031         struct fib6_config r_cfg;
3032         struct rtnexthop *rtnh;
3033         int remaining;
3034         int attrlen;
3035         int err = 1, last_err = 0;
3036
3037         remaining = cfg->fc_mp_len;
3038         rtnh = (struct rtnexthop *)cfg->fc_mp;
3039
3040         /* Parse a Multipath Entry */
3041         while (rtnh_ok(rtnh, remaining)) {
3042                 memcpy(&r_cfg, cfg, sizeof(*cfg));
3043                 if (rtnh->rtnh_ifindex)
3044                         r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
3045
3046                 attrlen = rtnh_attrlen(rtnh);
3047                 if (attrlen > 0) {
3048                         struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
3049
3050                         nla = nla_find(attrs, attrlen, RTA_GATEWAY);
3051                         if (nla) {
3052                                 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
3053                                 r_cfg.fc_flags |= RTF_GATEWAY;
3054                         }
3055                 }
3056                 err = ip6_route_del(&r_cfg);
3057                 if (err)
3058                         last_err = err;
3059
3060                 rtnh = rtnh_next(rtnh, &remaining);
3061         }
3062
3063         return last_err;
3064 }
3065
3066 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh)
3067 {
3068         struct fib6_config cfg;
3069         int err;
3070
3071         err = rtm_to_fib6_config(skb, nlh, &cfg);
3072         if (err < 0)
3073                 return err;
3074
3075         if (cfg.fc_mp)
3076                 return ip6_route_multipath_del(&cfg);
3077         else
3078                 return ip6_route_del(&cfg);
3079 }
3080
3081 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh)
3082 {
3083         struct fib6_config cfg;
3084         int err;
3085
3086         err = rtm_to_fib6_config(skb, nlh, &cfg);
3087         if (err < 0)
3088                 return err;
3089
3090         if (cfg.fc_mp)
3091                 return ip6_route_multipath_add(&cfg);
3092         else
3093                 return ip6_route_add(&cfg);
3094 }
3095
3096 static inline size_t rt6_nlmsg_size(struct rt6_info *rt)
3097 {
3098         return NLMSG_ALIGN(sizeof(struct rtmsg))
3099                + nla_total_size(16) /* RTA_SRC */
3100                + nla_total_size(16) /* RTA_DST */
3101                + nla_total_size(16) /* RTA_GATEWAY */
3102                + nla_total_size(16) /* RTA_PREFSRC */
3103                + nla_total_size(4) /* RTA_TABLE */
3104                + nla_total_size(4) /* RTA_IIF */
3105                + nla_total_size(4) /* RTA_OIF */
3106                + nla_total_size(4) /* RTA_PRIORITY */
3107                + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
3108                + nla_total_size(sizeof(struct rta_cacheinfo))
3109                + nla_total_size(TCP_CA_NAME_MAX) /* RTAX_CC_ALGO */
3110                + nla_total_size(1) /* RTA_PREF */
3111                + lwtunnel_get_encap_size(rt->dst.lwtstate);
3112 }
3113
3114 static int rt6_fill_node(struct net *net,
3115                          struct sk_buff *skb, struct rt6_info *rt,
3116                          struct in6_addr *dst, struct in6_addr *src,
3117                          int iif, int type, u32 portid, u32 seq,
3118                          int prefix, int nowait, unsigned int flags)
3119 {
3120         u32 metrics[RTAX_MAX];
3121         struct rtmsg *rtm;
3122         struct nlmsghdr *nlh;
3123         long expires;
3124         u32 table;
3125
3126         if (prefix) {   /* user wants prefix routes only */
3127                 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
3128                         /* success since this is not a prefix route */
3129                         return 1;
3130                 }
3131         }
3132
3133         nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
3134         if (!nlh)
3135                 return -EMSGSIZE;
3136
3137         rtm = nlmsg_data(nlh);
3138         rtm->rtm_family = AF_INET6;
3139         rtm->rtm_dst_len = rt->rt6i_dst.plen;
3140         rtm->rtm_src_len = rt->rt6i_src.plen;
3141         rtm->rtm_tos = 0;
3142         if (rt->rt6i_table)
3143                 table = rt->rt6i_table->tb6_id;
3144         else
3145                 table = RT6_TABLE_UNSPEC;
3146         rtm->rtm_table = table;
3147         if (nla_put_u32(skb, RTA_TABLE, table))
3148                 goto nla_put_failure;
3149         if (rt->rt6i_flags & RTF_REJECT) {
3150                 switch (rt->dst.error) {
3151                 case -EINVAL:
3152                         rtm->rtm_type = RTN_BLACKHOLE;
3153                         break;
3154                 case -EACCES:
3155                         rtm->rtm_type = RTN_PROHIBIT;
3156                         break;
3157                 case -EAGAIN:
3158                         rtm->rtm_type = RTN_THROW;
3159                         break;
3160                 default:
3161                         rtm->rtm_type = RTN_UNREACHABLE;
3162                         break;
3163                 }
3164         }
3165         else if (rt->rt6i_flags & RTF_LOCAL)
3166                 rtm->rtm_type = RTN_LOCAL;
3167         else if (rt->dst.dev && (rt->dst.dev->flags & IFF_LOOPBACK))
3168                 rtm->rtm_type = RTN_LOCAL;
3169         else
3170                 rtm->rtm_type = RTN_UNICAST;
3171         rtm->rtm_flags = 0;
3172         if (!netif_carrier_ok(rt->dst.dev)) {
3173                 rtm->rtm_flags |= RTNH_F_LINKDOWN;
3174                 if (rt->rt6i_idev->cnf.ignore_routes_with_linkdown)
3175                         rtm->rtm_flags |= RTNH_F_DEAD;
3176         }
3177         rtm->rtm_scope = RT_SCOPE_UNIVERSE;
3178         rtm->rtm_protocol = rt->rt6i_protocol;
3179         if (rt->rt6i_flags & RTF_DYNAMIC)
3180                 rtm->rtm_protocol = RTPROT_REDIRECT;
3181         else if (rt->rt6i_flags & RTF_ADDRCONF) {
3182                 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ROUTEINFO))
3183                         rtm->rtm_protocol = RTPROT_RA;
3184                 else
3185                         rtm->rtm_protocol = RTPROT_KERNEL;
3186         }
3187
3188         if (rt->rt6i_flags & RTF_CACHE)
3189                 rtm->rtm_flags |= RTM_F_CLONED;
3190
3191         if (dst) {
3192                 if (nla_put_in6_addr(skb, RTA_DST, dst))
3193                         goto nla_put_failure;
3194                 rtm->rtm_dst_len = 128;
3195         } else if (rtm->rtm_dst_len)
3196                 if (nla_put_in6_addr(skb, RTA_DST, &rt->rt6i_dst.addr))
3197                         goto nla_put_failure;
3198 #ifdef CONFIG_IPV6_SUBTREES
3199         if (src) {
3200                 if (nla_put_in6_addr(skb, RTA_SRC, src))
3201                         goto nla_put_failure;
3202                 rtm->rtm_src_len = 128;
3203         } else if (rtm->rtm_src_len &&
3204                    nla_put_in6_addr(skb, RTA_SRC, &rt->rt6i_src.addr))
3205                 goto nla_put_failure;
3206 #endif
3207         if (iif) {
3208 #ifdef CONFIG_IPV6_MROUTE
3209                 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
3210                         int err = ip6mr_get_route(net, skb, rtm, nowait);
3211                         if (err <= 0) {
3212                                 if (!nowait) {
3213                                         if (err == 0)
3214                                                 return 0;
3215                                         goto nla_put_failure;
3216                                 } else {
3217                                         if (err == -EMSGSIZE)
3218                                                 goto nla_put_failure;
3219                                 }
3220                         }
3221                 } else
3222 #endif
3223                         if (nla_put_u32(skb, RTA_IIF, iif))
3224                                 goto nla_put_failure;
3225         } else if (dst) {
3226                 struct in6_addr saddr_buf;
3227                 if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0 &&
3228                     nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
3229                         goto nla_put_failure;
3230         }
3231
3232         if (rt->rt6i_prefsrc.plen) {
3233                 struct in6_addr saddr_buf;
3234                 saddr_buf = rt->rt6i_prefsrc.addr;
3235                 if (nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
3236                         goto nla_put_failure;
3237         }
3238
3239         memcpy(metrics, dst_metrics_ptr(&rt->dst), sizeof(metrics));
3240         if (rt->rt6i_pmtu)
3241                 metrics[RTAX_MTU - 1] = rt->rt6i_pmtu;
3242         if (rtnetlink_put_metrics(skb, metrics) < 0)
3243                 goto nla_put_failure;
3244
3245         if (rt->rt6i_flags & RTF_GATEWAY) {
3246                 if (nla_put_in6_addr(skb, RTA_GATEWAY, &rt->rt6i_gateway) < 0)
3247                         goto nla_put_failure;
3248         }
3249
3250         if (rt->dst.dev &&
3251             nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex))
3252                 goto nla_put_failure;
3253         if (nla_put_u32(skb, RTA_PRIORITY, rt->rt6i_metric))
3254                 goto nla_put_failure;
3255
3256         expires = (rt->rt6i_flags & RTF_EXPIRES) ? rt->dst.expires - jiffies : 0;
3257
3258         if (rtnl_put_cacheinfo(skb, &rt->dst, 0, expires, rt->dst.error) < 0)
3259                 goto nla_put_failure;
3260
3261         if (nla_put_u8(skb, RTA_PREF, IPV6_EXTRACT_PREF(rt->rt6i_flags)))
3262                 goto nla_put_failure;
3263
3264         lwtunnel_fill_encap(skb, rt->dst.lwtstate);
3265
3266         nlmsg_end(skb, nlh);
3267         return 0;
3268
3269 nla_put_failure:
3270         nlmsg_cancel(skb, nlh);
3271         return -EMSGSIZE;
3272 }
3273
3274 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
3275 {
3276         struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
3277         int prefix;
3278
3279         if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
3280                 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
3281                 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
3282         } else
3283                 prefix = 0;
3284
3285         return rt6_fill_node(arg->net,
3286                      arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
3287                      NETLINK_CB(arg->cb->skb).portid, arg->cb->nlh->nlmsg_seq,
3288                      prefix, 0, NLM_F_MULTI);
3289 }
3290
3291 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh)
3292 {
3293         struct net *net = sock_net(in_skb->sk);
3294         struct nlattr *tb[RTA_MAX+1];
3295         struct rt6_info *rt;
3296         struct sk_buff *skb;
3297         struct rtmsg *rtm;
3298         struct flowi6 fl6;
3299         int err, iif = 0, oif = 0;
3300
3301         err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
3302         if (err < 0)
3303                 goto errout;
3304
3305         err = -EINVAL;
3306         memset(&fl6, 0, sizeof(fl6));
3307
3308         if (tb[RTA_SRC]) {
3309                 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
3310                         goto errout;
3311
3312                 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
3313         }
3314
3315         if (tb[RTA_DST]) {
3316                 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
3317                         goto errout;
3318
3319                 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
3320         }
3321
3322         if (tb[RTA_IIF])
3323                 iif = nla_get_u32(tb[RTA_IIF]);
3324
3325         if (tb[RTA_OIF])
3326                 oif = nla_get_u32(tb[RTA_OIF]);
3327
3328         if (tb[RTA_MARK])
3329                 fl6.flowi6_mark = nla_get_u32(tb[RTA_MARK]);
3330
3331         if (iif) {
3332                 struct net_device *dev;
3333                 int flags = 0;
3334
3335                 dev = __dev_get_by_index(net, iif);
3336                 if (!dev) {
3337                         err = -ENODEV;
3338                         goto errout;
3339                 }
3340
3341                 fl6.flowi6_iif = iif;
3342
3343                 if (!ipv6_addr_any(&fl6.saddr))
3344                         flags |= RT6_LOOKUP_F_HAS_SADDR;
3345
3346                 rt = (struct rt6_info *)ip6_route_input_lookup(net, dev, &fl6,
3347                                                                flags);
3348         } else {
3349                 fl6.flowi6_oif = oif;
3350
3351                 if (netif_index_is_l3_master(net, oif)) {
3352                         fl6.flowi6_flags = FLOWI_FLAG_L3MDEV_SRC |
3353                                            FLOWI_FLAG_SKIP_NH_OIF;
3354                 }
3355
3356                 rt = (struct rt6_info *)ip6_route_output(net, NULL, &fl6);
3357         }
3358
3359         skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
3360         if (!skb) {
3361                 ip6_rt_put(rt);
3362                 err = -ENOBUFS;
3363                 goto errout;
3364         }
3365
3366         /* Reserve room for dummy headers, this skb can pass
3367            through good chunk of routing engine.
3368          */
3369         skb_reset_mac_header(skb);
3370         skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
3371
3372         skb_dst_set(skb, &rt->dst);
3373
3374         err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
3375                             RTM_NEWROUTE, NETLINK_CB(in_skb).portid,
3376                             nlh->nlmsg_seq, 0, 0, 0);
3377         if (err < 0) {
3378                 kfree_skb(skb);
3379                 goto errout;
3380         }
3381
3382         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
3383 errout:
3384         return err;
3385 }
3386
3387 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info,
3388                      unsigned int nlm_flags)
3389 {
3390         struct sk_buff *skb;
3391         struct net *net = info->nl_net;
3392         u32 seq;
3393         int err;
3394
3395         err = -ENOBUFS;
3396         seq = info->nlh ? info->nlh->nlmsg_seq : 0;
3397
3398         skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
3399         if (!skb)
3400                 goto errout;
3401
3402         err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
3403                                 event, info->portid, seq, 0, 0, nlm_flags);
3404         if (err < 0) {
3405                 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
3406                 WARN_ON(err == -EMSGSIZE);
3407                 kfree_skb(skb);
3408                 goto errout;
3409         }
3410         rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
3411                     info->nlh, gfp_any());
3412         return;
3413 errout:
3414         if (err < 0)
3415                 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
3416 }
3417
3418 static int ip6_route_dev_notify(struct notifier_block *this,
3419                                 unsigned long event, void *ptr)
3420 {
3421         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3422         struct net *net = dev_net(dev);
3423
3424         if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
3425                 net->ipv6.ip6_null_entry->dst.dev = dev;
3426                 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
3427 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3428                 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
3429                 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
3430                 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
3431                 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
3432 #endif
3433         }
3434
3435         return NOTIFY_OK;
3436 }
3437
3438 /*
3439  *      /proc
3440  */
3441
3442 #ifdef CONFIG_PROC_FS
3443
3444 static const struct file_operations ipv6_route_proc_fops = {
3445         .owner          = THIS_MODULE,
3446         .open           = ipv6_route_open,
3447         .read           = seq_read,
3448         .llseek         = seq_lseek,
3449         .release        = seq_release_net,
3450 };
3451
3452 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
3453 {
3454         struct net *net = (struct net *)seq->private;
3455         seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
3456                    net->ipv6.rt6_stats->fib_nodes,
3457                    net->ipv6.rt6_stats->fib_route_nodes,
3458                    net->ipv6.rt6_stats->fib_rt_alloc,
3459                    net->ipv6.rt6_stats->fib_rt_entries,
3460                    net->ipv6.rt6_stats->fib_rt_cache,
3461                    dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
3462                    net->ipv6.rt6_stats->fib_discarded_routes);
3463
3464         return 0;
3465 }
3466
3467 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
3468 {
3469         return single_open_net(inode, file, rt6_stats_seq_show);
3470 }
3471
3472 static const struct file_operations rt6_stats_seq_fops = {
3473         .owner   = THIS_MODULE,
3474         .open    = rt6_stats_seq_open,
3475         .read    = seq_read,
3476         .llseek  = seq_lseek,
3477         .release = single_release_net,
3478 };
3479 #endif  /* CONFIG_PROC_FS */
3480
3481 #ifdef CONFIG_SYSCTL
3482
3483 static
3484 int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
3485                               void __user *buffer, size_t *lenp, loff_t *ppos)
3486 {
3487         struct net *net;
3488         int delay;
3489         if (!write)
3490                 return -EINVAL;
3491
3492         net = (struct net *)ctl->extra1;
3493         delay = net->ipv6.sysctl.flush_delay;
3494         proc_dointvec(ctl, write, buffer, lenp, ppos);
3495         fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
3496         return 0;
3497 }
3498
3499 struct ctl_table ipv6_route_table_template[] = {
3500         {
3501                 .procname       =       "flush",
3502                 .data           =       &init_net.ipv6.sysctl.flush_delay,
3503                 .maxlen         =       sizeof(int),
3504                 .mode           =       0200,
3505                 .proc_handler   =       ipv6_sysctl_rtcache_flush
3506         },
3507         {
3508                 .procname       =       "gc_thresh",
3509                 .data           =       &ip6_dst_ops_template.gc_thresh,
3510                 .maxlen         =       sizeof(int),
3511                 .mode           =       0644,
3512                 .proc_handler   =       proc_dointvec,
3513         },
3514         {
3515                 .procname       =       "max_size",
3516                 .data           =       &init_net.ipv6.sysctl.ip6_rt_max_size,
3517                 .maxlen         =       sizeof(int),
3518                 .mode           =       0644,
3519                 .proc_handler   =       proc_dointvec,
3520         },
3521         {
3522                 .procname       =       "gc_min_interval",
3523                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
3524                 .maxlen         =       sizeof(int),
3525                 .mode           =       0644,
3526                 .proc_handler   =       proc_dointvec_jiffies,
3527         },
3528         {
3529                 .procname       =       "gc_timeout",
3530                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
3531                 .maxlen         =       sizeof(int),
3532                 .mode           =       0644,
3533                 .proc_handler   =       proc_dointvec_jiffies,
3534         },
3535         {
3536                 .procname       =       "gc_interval",
3537                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_interval,
3538                 .maxlen         =       sizeof(int),
3539                 .mode           =       0644,
3540                 .proc_handler   =       proc_dointvec_jiffies,
3541         },
3542         {
3543                 .procname       =       "gc_elasticity",
3544                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
3545                 .maxlen         =       sizeof(int),
3546                 .mode           =       0644,
3547                 .proc_handler   =       proc_dointvec,
3548         },
3549         {
3550                 .procname       =       "mtu_expires",
3551                 .data           =       &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
3552                 .maxlen         =       sizeof(int),
3553                 .mode           =       0644,
3554                 .proc_handler   =       proc_dointvec_jiffies,
3555         },
3556         {
3557                 .procname       =       "min_adv_mss",
3558                 .data           =       &init_net.ipv6.sysctl.ip6_rt_min_advmss,
3559                 .maxlen         =       sizeof(int),
3560                 .mode           =       0644,
3561                 .proc_handler   =       proc_dointvec,
3562         },
3563         {
3564                 .procname       =       "gc_min_interval_ms",
3565                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
3566                 .maxlen         =       sizeof(int),
3567                 .mode           =       0644,
3568                 .proc_handler   =       proc_dointvec_ms_jiffies,
3569         },
3570         { }
3571 };
3572
3573 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
3574 {
3575         struct ctl_table *table;
3576
3577         table = kmemdup(ipv6_route_table_template,
3578                         sizeof(ipv6_route_table_template),
3579                         GFP_KERNEL);
3580
3581         if (table) {
3582                 table[0].data = &net->ipv6.sysctl.flush_delay;
3583                 table[0].extra1 = net;
3584                 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
3585                 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
3586                 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
3587                 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
3588                 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
3589                 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
3590                 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
3591                 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
3592                 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
3593
3594                 /* Don't export sysctls to unprivileged users */
3595                 if (net->user_ns != &init_user_ns)
3596                         table[0].procname = NULL;
3597         }
3598
3599         return table;
3600 }
3601 #endif
3602
3603 static int __net_init ip6_route_net_init(struct net *net)
3604 {
3605         int ret = -ENOMEM;
3606
3607         memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
3608                sizeof(net->ipv6.ip6_dst_ops));
3609
3610         if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
3611                 goto out_ip6_dst_ops;
3612
3613         net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
3614                                            sizeof(*net->ipv6.ip6_null_entry),
3615                                            GFP_KERNEL);
3616         if (!net->ipv6.ip6_null_entry)
3617                 goto out_ip6_dst_entries;
3618         net->ipv6.ip6_null_entry->dst.path =
3619                 (struct dst_entry *)net->ipv6.ip6_null_entry;
3620         net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3621         dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
3622                          ip6_template_metrics, true);
3623
3624 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3625         net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
3626                                                sizeof(*net->ipv6.ip6_prohibit_entry),
3627                                                GFP_KERNEL);
3628         if (!net->ipv6.ip6_prohibit_entry)
3629                 goto out_ip6_null_entry;
3630         net->ipv6.ip6_prohibit_entry->dst.path =
3631                 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
3632         net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3633         dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
3634                          ip6_template_metrics, true);
3635
3636         net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
3637                                                sizeof(*net->ipv6.ip6_blk_hole_entry),
3638                                                GFP_KERNEL);
3639         if (!net->ipv6.ip6_blk_hole_entry)
3640                 goto out_ip6_prohibit_entry;
3641         net->ipv6.ip6_blk_hole_entry->dst.path =
3642                 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
3643         net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3644         dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
3645                          ip6_template_metrics, true);
3646 #endif
3647
3648         net->ipv6.sysctl.flush_delay = 0;
3649         net->ipv6.sysctl.ip6_rt_max_size = 4096;
3650         net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
3651         net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
3652         net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
3653         net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
3654         net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
3655         net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
3656
3657         net->ipv6.ip6_rt_gc_expire = 30*HZ;
3658
3659         ret = 0;
3660 out:
3661         return ret;
3662
3663 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3664 out_ip6_prohibit_entry:
3665         kfree(net->ipv6.ip6_prohibit_entry);
3666 out_ip6_null_entry:
3667         kfree(net->ipv6.ip6_null_entry);
3668 #endif
3669 out_ip6_dst_entries:
3670         dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3671 out_ip6_dst_ops:
3672         goto out;
3673 }
3674
3675 static void __net_exit ip6_route_net_exit(struct net *net)
3676 {
3677         kfree(net->ipv6.ip6_null_entry);
3678 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3679         kfree(net->ipv6.ip6_prohibit_entry);
3680         kfree(net->ipv6.ip6_blk_hole_entry);
3681 #endif
3682         dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3683 }
3684
3685 static int __net_init ip6_route_net_init_late(struct net *net)
3686 {
3687 #ifdef CONFIG_PROC_FS
3688         proc_create("ipv6_route", 0, net->proc_net, &ipv6_route_proc_fops);
3689         proc_create("rt6_stats", S_IRUGO, net->proc_net, &rt6_stats_seq_fops);
3690 #endif
3691         return 0;
3692 }
3693
3694 static void __net_exit ip6_route_net_exit_late(struct net *net)
3695 {
3696 #ifdef CONFIG_PROC_FS
3697         remove_proc_entry("ipv6_route", net->proc_net);
3698         remove_proc_entry("rt6_stats", net->proc_net);
3699 #endif
3700 }
3701
3702 static struct pernet_operations ip6_route_net_ops = {
3703         .init = ip6_route_net_init,
3704         .exit = ip6_route_net_exit,
3705 };
3706
3707 static int __net_init ipv6_inetpeer_init(struct net *net)
3708 {
3709         struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
3710
3711         if (!bp)
3712                 return -ENOMEM;
3713         inet_peer_base_init(bp);
3714         net->ipv6.peers = bp;
3715         return 0;
3716 }
3717
3718 static void __net_exit ipv6_inetpeer_exit(struct net *net)
3719 {
3720         struct inet_peer_base *bp = net->ipv6.peers;
3721
3722         net->ipv6.peers = NULL;
3723         inetpeer_invalidate_tree(bp);
3724         kfree(bp);
3725 }
3726
3727 static struct pernet_operations ipv6_inetpeer_ops = {
3728         .init   =       ipv6_inetpeer_init,
3729         .exit   =       ipv6_inetpeer_exit,
3730 };
3731
3732 static struct pernet_operations ip6_route_net_late_ops = {
3733         .init = ip6_route_net_init_late,
3734         .exit = ip6_route_net_exit_late,
3735 };
3736
3737 static struct notifier_block ip6_route_dev_notifier = {
3738         .notifier_call = ip6_route_dev_notify,
3739         .priority = 0,
3740 };
3741
3742 int __init ip6_route_init(void)
3743 {
3744         int ret;
3745         int cpu;
3746
3747         ret = -ENOMEM;
3748         ip6_dst_ops_template.kmem_cachep =
3749                 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
3750                                   SLAB_HWCACHE_ALIGN, NULL);
3751         if (!ip6_dst_ops_template.kmem_cachep)
3752                 goto out;
3753
3754         ret = dst_entries_init(&ip6_dst_blackhole_ops);
3755         if (ret)
3756                 goto out_kmem_cache;
3757
3758         ret = register_pernet_subsys(&ipv6_inetpeer_ops);
3759         if (ret)
3760                 goto out_dst_entries;
3761
3762         ret = register_pernet_subsys(&ip6_route_net_ops);
3763         if (ret)
3764                 goto out_register_inetpeer;
3765
3766         ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
3767
3768         /* Registering of the loopback is done before this portion of code,
3769          * the loopback reference in rt6_info will not be taken, do it
3770          * manually for init_net */
3771         init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
3772         init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3773   #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3774         init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
3775         init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3776         init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
3777         init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3778   #endif
3779         ret = fib6_init();
3780         if (ret)
3781                 goto out_register_subsys;
3782
3783         ret = xfrm6_init();
3784         if (ret)
3785                 goto out_fib6_init;
3786
3787         ret = fib6_rules_init();
3788         if (ret)
3789                 goto xfrm6_init;
3790
3791         ret = register_pernet_subsys(&ip6_route_net_late_ops);
3792         if (ret)
3793                 goto fib6_rules_init;
3794
3795         ret = -ENOBUFS;
3796         if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL, NULL) ||
3797             __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL, NULL) ||
3798             __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL, NULL))
3799                 goto out_register_late_subsys;
3800
3801         ret = register_netdevice_notifier(&ip6_route_dev_notifier);
3802         if (ret)
3803                 goto out_register_late_subsys;
3804
3805         for_each_possible_cpu(cpu) {
3806                 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
3807
3808                 INIT_LIST_HEAD(&ul->head);
3809                 spin_lock_init(&ul->lock);
3810         }
3811
3812 out:
3813         return ret;
3814
3815 out_register_late_subsys:
3816         unregister_pernet_subsys(&ip6_route_net_late_ops);
3817 fib6_rules_init:
3818         fib6_rules_cleanup();
3819 xfrm6_init:
3820         xfrm6_fini();
3821 out_fib6_init:
3822         fib6_gc_cleanup();
3823 out_register_subsys:
3824         unregister_pernet_subsys(&ip6_route_net_ops);
3825 out_register_inetpeer:
3826         unregister_pernet_subsys(&ipv6_inetpeer_ops);
3827 out_dst_entries:
3828         dst_entries_destroy(&ip6_dst_blackhole_ops);
3829 out_kmem_cache:
3830         kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3831         goto out;
3832 }
3833
3834 void ip6_route_cleanup(void)
3835 {
3836         unregister_netdevice_notifier(&ip6_route_dev_notifier);
3837         unregister_pernet_subsys(&ip6_route_net_late_ops);
3838         fib6_rules_cleanup();
3839         xfrm6_fini();
3840         fib6_gc_cleanup();
3841         unregister_pernet_subsys(&ipv6_inetpeer_ops);
3842         unregister_pernet_subsys(&ip6_route_net_ops);
3843         dst_entries_destroy(&ip6_dst_blackhole_ops);
3844         kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3845 }