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 struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table,
1046                                int oif, 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 EXPORT_SYMBOL_GPL(ip6_pol_route);
1143
1144 static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
1145                                             struct flowi6 *fl6, int flags)
1146 {
1147         return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, flags);
1148 }
1149
1150 static struct dst_entry *ip6_route_input_lookup(struct net *net,
1151                                                 struct net_device *dev,
1152                                                 struct flowi6 *fl6, int flags)
1153 {
1154         if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
1155                 flags |= RT6_LOOKUP_F_IFACE;
1156
1157         return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_input);
1158 }
1159
1160 void ip6_route_input(struct sk_buff *skb)
1161 {
1162         const struct ipv6hdr *iph = ipv6_hdr(skb);
1163         struct net *net = dev_net(skb->dev);
1164         int flags = RT6_LOOKUP_F_HAS_SADDR;
1165         struct ip_tunnel_info *tun_info;
1166         struct flowi6 fl6 = {
1167                 .flowi6_iif = l3mdev_fib_oif(skb->dev),
1168                 .daddr = iph->daddr,
1169                 .saddr = iph->saddr,
1170                 .flowlabel = ip6_flowinfo(iph),
1171                 .flowi6_mark = skb->mark,
1172                 .flowi6_proto = iph->nexthdr,
1173         };
1174
1175         tun_info = skb_tunnel_info(skb);
1176         if (tun_info && !(tun_info->mode & IP_TUNNEL_INFO_TX))
1177                 fl6.flowi6_tun_key.tun_id = tun_info->key.tun_id;
1178         skb_dst_drop(skb);
1179         skb_dst_set(skb, ip6_route_input_lookup(net, skb->dev, &fl6, flags));
1180 }
1181
1182 static struct rt6_info *ip6_pol_route_output(struct net *net, struct fib6_table *table,
1183                                              struct flowi6 *fl6, int flags)
1184 {
1185         return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, flags);
1186 }
1187
1188 struct dst_entry *ip6_route_output_flags(struct net *net, const struct sock *sk,
1189                                          struct flowi6 *fl6, int flags)
1190 {
1191         struct dst_entry *dst;
1192         bool any_src;
1193
1194         dst = l3mdev_get_rt6_dst(net, fl6);
1195         if (dst)
1196                 return dst;
1197
1198         fl6->flowi6_iif = LOOPBACK_IFINDEX;
1199
1200         any_src = ipv6_addr_any(&fl6->saddr);
1201         if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr) ||
1202             (fl6->flowi6_oif && any_src))
1203                 flags |= RT6_LOOKUP_F_IFACE;
1204
1205         if (!any_src)
1206                 flags |= RT6_LOOKUP_F_HAS_SADDR;
1207         else if (sk)
1208                 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
1209
1210         return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_output);
1211 }
1212 EXPORT_SYMBOL_GPL(ip6_route_output_flags);
1213
1214 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
1215 {
1216         struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
1217         struct dst_entry *new = NULL;
1218
1219         rt = dst_alloc(&ip6_dst_blackhole_ops, ort->dst.dev, 1, DST_OBSOLETE_NONE, 0);
1220         if (rt) {
1221                 rt6_info_init(rt);
1222
1223                 new = &rt->dst;
1224                 new->__use = 1;
1225                 new->input = dst_discard;
1226                 new->output = dst_discard_out;
1227
1228                 dst_copy_metrics(new, &ort->dst);
1229                 rt->rt6i_idev = ort->rt6i_idev;
1230                 if (rt->rt6i_idev)
1231                         in6_dev_hold(rt->rt6i_idev);
1232
1233                 rt->rt6i_gateway = ort->rt6i_gateway;
1234                 rt->rt6i_flags = ort->rt6i_flags & ~RTF_PCPU;
1235                 rt->rt6i_metric = 0;
1236
1237                 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1238 #ifdef CONFIG_IPV6_SUBTREES
1239                 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1240 #endif
1241
1242                 dst_free(new);
1243         }
1244
1245         dst_release(dst_orig);
1246         return new ? new : ERR_PTR(-ENOMEM);
1247 }
1248
1249 /*
1250  *      Destination cache support functions
1251  */
1252
1253 static void rt6_dst_from_metrics_check(struct rt6_info *rt)
1254 {
1255         if (rt->dst.from &&
1256             dst_metrics_ptr(&rt->dst) != dst_metrics_ptr(rt->dst.from))
1257                 dst_init_metrics(&rt->dst, dst_metrics_ptr(rt->dst.from), true);
1258 }
1259
1260 static struct dst_entry *rt6_check(struct rt6_info *rt, u32 cookie)
1261 {
1262         if (!rt->rt6i_node || (rt->rt6i_node->fn_sernum != cookie))
1263                 return NULL;
1264
1265         if (rt6_check_expired(rt))
1266                 return NULL;
1267
1268         return &rt->dst;
1269 }
1270
1271 static struct dst_entry *rt6_dst_from_check(struct rt6_info *rt, u32 cookie)
1272 {
1273         if (!__rt6_check_expired(rt) &&
1274             rt->dst.obsolete == DST_OBSOLETE_FORCE_CHK &&
1275             rt6_check((struct rt6_info *)(rt->dst.from), cookie))
1276                 return &rt->dst;
1277         else
1278                 return NULL;
1279 }
1280
1281 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
1282 {
1283         struct rt6_info *rt;
1284
1285         rt = (struct rt6_info *) dst;
1286
1287         /* All IPV6 dsts are created with ->obsolete set to the value
1288          * DST_OBSOLETE_FORCE_CHK which forces validation calls down
1289          * into this function always.
1290          */
1291
1292         rt6_dst_from_metrics_check(rt);
1293
1294         if (rt->rt6i_flags & RTF_PCPU ||
1295             (unlikely(dst->flags & DST_NOCACHE) && rt->dst.from))
1296                 return rt6_dst_from_check(rt, cookie);
1297         else
1298                 return rt6_check(rt, cookie);
1299 }
1300
1301 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
1302 {
1303         struct rt6_info *rt = (struct rt6_info *) dst;
1304
1305         if (rt) {
1306                 if (rt->rt6i_flags & RTF_CACHE) {
1307                         if (rt6_check_expired(rt)) {
1308                                 ip6_del_rt(rt);
1309                                 dst = NULL;
1310                         }
1311                 } else {
1312                         dst_release(dst);
1313                         dst = NULL;
1314                 }
1315         }
1316         return dst;
1317 }
1318
1319 static void ip6_link_failure(struct sk_buff *skb)
1320 {
1321         struct rt6_info *rt;
1322
1323         icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
1324
1325         rt = (struct rt6_info *) skb_dst(skb);
1326         if (rt) {
1327                 if (rt->rt6i_flags & RTF_CACHE) {
1328                         dst_hold(&rt->dst);
1329                         ip6_del_rt(rt);
1330                 } else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT)) {
1331                         rt->rt6i_node->fn_sernum = -1;
1332                 }
1333         }
1334 }
1335
1336 static void rt6_do_update_pmtu(struct rt6_info *rt, u32 mtu)
1337 {
1338         struct net *net = dev_net(rt->dst.dev);
1339
1340         rt->rt6i_flags |= RTF_MODIFIED;
1341         rt->rt6i_pmtu = mtu;
1342         rt6_update_expires(rt, net->ipv6.sysctl.ip6_rt_mtu_expires);
1343 }
1344
1345 static bool rt6_cache_allowed_for_pmtu(const struct rt6_info *rt)
1346 {
1347         return !(rt->rt6i_flags & RTF_CACHE) &&
1348                 (rt->rt6i_flags & RTF_PCPU || rt->rt6i_node);
1349 }
1350
1351 static void __ip6_rt_update_pmtu(struct dst_entry *dst, const struct sock *sk,
1352                                  const struct ipv6hdr *iph, u32 mtu)
1353 {
1354         struct rt6_info *rt6 = (struct rt6_info *)dst;
1355
1356         if (rt6->rt6i_flags & RTF_LOCAL)
1357                 return;
1358
1359         dst_confirm(dst);
1360         mtu = max_t(u32, mtu, IPV6_MIN_MTU);
1361         if (mtu >= dst_mtu(dst))
1362                 return;
1363
1364         if (!rt6_cache_allowed_for_pmtu(rt6)) {
1365                 rt6_do_update_pmtu(rt6, mtu);
1366         } else {
1367                 const struct in6_addr *daddr, *saddr;
1368                 struct rt6_info *nrt6;
1369
1370                 if (iph) {
1371                         daddr = &iph->daddr;
1372                         saddr = &iph->saddr;
1373                 } else if (sk) {
1374                         daddr = &sk->sk_v6_daddr;
1375                         saddr = &inet6_sk(sk)->saddr;
1376                 } else {
1377                         return;
1378                 }
1379                 nrt6 = ip6_rt_cache_alloc(rt6, daddr, saddr);
1380                 if (nrt6) {
1381                         rt6_do_update_pmtu(nrt6, mtu);
1382
1383                         /* ip6_ins_rt(nrt6) will bump the
1384                          * rt6->rt6i_node->fn_sernum
1385                          * which will fail the next rt6_check() and
1386                          * invalidate the sk->sk_dst_cache.
1387                          */
1388                         ip6_ins_rt(nrt6);
1389                 }
1390         }
1391 }
1392
1393 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
1394                                struct sk_buff *skb, u32 mtu)
1395 {
1396         __ip6_rt_update_pmtu(dst, sk, skb ? ipv6_hdr(skb) : NULL, mtu);
1397 }
1398
1399 void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
1400                      int oif, u32 mark)
1401 {
1402         const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1403         struct dst_entry *dst;
1404         struct flowi6 fl6;
1405
1406         memset(&fl6, 0, sizeof(fl6));
1407         fl6.flowi6_oif = oif;
1408         fl6.flowi6_mark = mark ? mark : IP6_REPLY_MARK(net, skb->mark);
1409         fl6.daddr = iph->daddr;
1410         fl6.saddr = iph->saddr;
1411         fl6.flowlabel = ip6_flowinfo(iph);
1412
1413         dst = ip6_route_output(net, NULL, &fl6);
1414         if (!dst->error)
1415                 __ip6_rt_update_pmtu(dst, NULL, iph, ntohl(mtu));
1416         dst_release(dst);
1417 }
1418 EXPORT_SYMBOL_GPL(ip6_update_pmtu);
1419
1420 void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
1421 {
1422         struct dst_entry *dst;
1423
1424         ip6_update_pmtu(skb, sock_net(sk), mtu,
1425                         sk->sk_bound_dev_if, sk->sk_mark);
1426
1427         dst = __sk_dst_get(sk);
1428         if (!dst || !dst->obsolete ||
1429             dst->ops->check(dst, inet6_sk(sk)->dst_cookie))
1430                 return;
1431
1432         bh_lock_sock(sk);
1433         if (!sock_owned_by_user(sk) && !ipv6_addr_v4mapped(&sk->sk_v6_daddr))
1434                 ip6_datagram_dst_update(sk, false);
1435         bh_unlock_sock(sk);
1436 }
1437 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
1438
1439 /* Handle redirects */
1440 struct ip6rd_flowi {
1441         struct flowi6 fl6;
1442         struct in6_addr gateway;
1443 };
1444
1445 static struct rt6_info *__ip6_route_redirect(struct net *net,
1446                                              struct fib6_table *table,
1447                                              struct flowi6 *fl6,
1448                                              int flags)
1449 {
1450         struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
1451         struct rt6_info *rt;
1452         struct fib6_node *fn;
1453
1454         /* Get the "current" route for this destination and
1455          * check if the redirect has come from approriate router.
1456          *
1457          * RFC 4861 specifies that redirects should only be
1458          * accepted if they come from the nexthop to the target.
1459          * Due to the way the routes are chosen, this notion
1460          * is a bit fuzzy and one might need to check all possible
1461          * routes.
1462          */
1463
1464         read_lock_bh(&table->tb6_lock);
1465         fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1466 restart:
1467         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1468                 if (rt6_check_expired(rt))
1469                         continue;
1470                 if (rt->dst.error)
1471                         break;
1472                 if (!(rt->rt6i_flags & RTF_GATEWAY))
1473                         continue;
1474                 if (fl6->flowi6_oif != rt->dst.dev->ifindex)
1475                         continue;
1476                 if (!ipv6_addr_equal(&rdfl->gateway, &rt->rt6i_gateway))
1477                         continue;
1478                 break;
1479         }
1480
1481         if (!rt)
1482                 rt = net->ipv6.ip6_null_entry;
1483         else if (rt->dst.error) {
1484                 rt = net->ipv6.ip6_null_entry;
1485                 goto out;
1486         }
1487
1488         if (rt == net->ipv6.ip6_null_entry) {
1489                 fn = fib6_backtrack(fn, &fl6->saddr);
1490                 if (fn)
1491                         goto restart;
1492         }
1493
1494 out:
1495         dst_hold(&rt->dst);
1496
1497         read_unlock_bh(&table->tb6_lock);
1498
1499         trace_fib6_table_lookup(net, rt, table->tb6_id, fl6);
1500         return rt;
1501 };
1502
1503 static struct dst_entry *ip6_route_redirect(struct net *net,
1504                                         const struct flowi6 *fl6,
1505                                         const struct in6_addr *gateway)
1506 {
1507         int flags = RT6_LOOKUP_F_HAS_SADDR;
1508         struct ip6rd_flowi rdfl;
1509
1510         rdfl.fl6 = *fl6;
1511         rdfl.gateway = *gateway;
1512
1513         return fib6_rule_lookup(net, &rdfl.fl6,
1514                                 flags, __ip6_route_redirect);
1515 }
1516
1517 void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark)
1518 {
1519         const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1520         struct dst_entry *dst;
1521         struct flowi6 fl6;
1522
1523         memset(&fl6, 0, sizeof(fl6));
1524         fl6.flowi6_iif = LOOPBACK_IFINDEX;
1525         fl6.flowi6_oif = oif;
1526         fl6.flowi6_mark = mark;
1527         fl6.daddr = iph->daddr;
1528         fl6.saddr = iph->saddr;
1529         fl6.flowlabel = ip6_flowinfo(iph);
1530
1531         dst = ip6_route_redirect(net, &fl6, &ipv6_hdr(skb)->saddr);
1532         rt6_do_redirect(dst, NULL, skb);
1533         dst_release(dst);
1534 }
1535 EXPORT_SYMBOL_GPL(ip6_redirect);
1536
1537 void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif,
1538                             u32 mark)
1539 {
1540         const struct ipv6hdr *iph = ipv6_hdr(skb);
1541         const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
1542         struct dst_entry *dst;
1543         struct flowi6 fl6;
1544
1545         memset(&fl6, 0, sizeof(fl6));
1546         fl6.flowi6_iif = LOOPBACK_IFINDEX;
1547         fl6.flowi6_oif = oif;
1548         fl6.flowi6_mark = mark;
1549         fl6.daddr = msg->dest;
1550         fl6.saddr = iph->daddr;
1551
1552         dst = ip6_route_redirect(net, &fl6, &iph->saddr);
1553         rt6_do_redirect(dst, NULL, skb);
1554         dst_release(dst);
1555 }
1556
1557 void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
1558 {
1559         ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark);
1560 }
1561 EXPORT_SYMBOL_GPL(ip6_sk_redirect);
1562
1563 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1564 {
1565         struct net_device *dev = dst->dev;
1566         unsigned int mtu = dst_mtu(dst);
1567         struct net *net = dev_net(dev);
1568
1569         mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
1570
1571         if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
1572                 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1573
1574         /*
1575          * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1576          * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1577          * IPV6_MAXPLEN is also valid and means: "any MSS,
1578          * rely only on pmtu discovery"
1579          */
1580         if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
1581                 mtu = IPV6_MAXPLEN;
1582         return mtu;
1583 }
1584
1585 static unsigned int ip6_mtu(const struct dst_entry *dst)
1586 {
1587         const struct rt6_info *rt = (const struct rt6_info *)dst;
1588         unsigned int mtu = rt->rt6i_pmtu;
1589         struct inet6_dev *idev;
1590
1591         if (mtu)
1592                 goto out;
1593
1594         mtu = dst_metric_raw(dst, RTAX_MTU);
1595         if (mtu)
1596                 goto out;
1597
1598         mtu = IPV6_MIN_MTU;
1599
1600         rcu_read_lock();
1601         idev = __in6_dev_get(dst->dev);
1602         if (idev)
1603                 mtu = idev->cnf.mtu6;
1604         rcu_read_unlock();
1605
1606 out:
1607         return min_t(unsigned int, mtu, IP6_MAX_MTU);
1608 }
1609
1610 static struct dst_entry *icmp6_dst_gc_list;
1611 static DEFINE_SPINLOCK(icmp6_dst_lock);
1612
1613 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
1614                                   struct flowi6 *fl6)
1615 {
1616         struct dst_entry *dst;
1617         struct rt6_info *rt;
1618         struct inet6_dev *idev = in6_dev_get(dev);
1619         struct net *net = dev_net(dev);
1620
1621         if (unlikely(!idev))
1622                 return ERR_PTR(-ENODEV);
1623
1624         rt = ip6_dst_alloc(net, dev, 0);
1625         if (unlikely(!rt)) {
1626                 in6_dev_put(idev);
1627                 dst = ERR_PTR(-ENOMEM);
1628                 goto out;
1629         }
1630
1631         rt->dst.flags |= DST_HOST;
1632         rt->dst.output  = ip6_output;
1633         atomic_set(&rt->dst.__refcnt, 1);
1634         rt->rt6i_gateway  = fl6->daddr;
1635         rt->rt6i_dst.addr = fl6->daddr;
1636         rt->rt6i_dst.plen = 128;
1637         rt->rt6i_idev     = idev;
1638         dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
1639
1640         spin_lock_bh(&icmp6_dst_lock);
1641         rt->dst.next = icmp6_dst_gc_list;
1642         icmp6_dst_gc_list = &rt->dst;
1643         spin_unlock_bh(&icmp6_dst_lock);
1644
1645         fib6_force_start_gc(net);
1646
1647         dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
1648
1649 out:
1650         return dst;
1651 }
1652
1653 int icmp6_dst_gc(void)
1654 {
1655         struct dst_entry *dst, **pprev;
1656         int more = 0;
1657
1658         spin_lock_bh(&icmp6_dst_lock);
1659         pprev = &icmp6_dst_gc_list;
1660
1661         while ((dst = *pprev) != NULL) {
1662                 if (!atomic_read(&dst->__refcnt)) {
1663                         *pprev = dst->next;
1664                         dst_free(dst);
1665                 } else {
1666                         pprev = &dst->next;
1667                         ++more;
1668                 }
1669         }
1670
1671         spin_unlock_bh(&icmp6_dst_lock);
1672
1673         return more;
1674 }
1675
1676 static void icmp6_clean_all(int (*func)(struct rt6_info *rt, void *arg),
1677                             void *arg)
1678 {
1679         struct dst_entry *dst, **pprev;
1680
1681         spin_lock_bh(&icmp6_dst_lock);
1682         pprev = &icmp6_dst_gc_list;
1683         while ((dst = *pprev) != NULL) {
1684                 struct rt6_info *rt = (struct rt6_info *) dst;
1685                 if (func(rt, arg)) {
1686                         *pprev = dst->next;
1687                         dst_free(dst);
1688                 } else {
1689                         pprev = &dst->next;
1690                 }
1691         }
1692         spin_unlock_bh(&icmp6_dst_lock);
1693 }
1694
1695 static int ip6_dst_gc(struct dst_ops *ops)
1696 {
1697         struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
1698         int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1699         int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1700         int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1701         int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1702         unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1703         int entries;
1704
1705         entries = dst_entries_get_fast(ops);
1706         if (time_after(rt_last_gc + rt_min_interval, jiffies) &&
1707             entries <= rt_max_size)
1708                 goto out;
1709
1710         net->ipv6.ip6_rt_gc_expire++;
1711         fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net, true);
1712         entries = dst_entries_get_slow(ops);
1713         if (entries < ops->gc_thresh)
1714                 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1715 out:
1716         net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1717         return entries > rt_max_size;
1718 }
1719
1720 static int ip6_convert_metrics(struct mx6_config *mxc,
1721                                const struct fib6_config *cfg)
1722 {
1723         bool ecn_ca = false;
1724         struct nlattr *nla;
1725         int remaining;
1726         u32 *mp;
1727
1728         if (!cfg->fc_mx)
1729                 return 0;
1730
1731         mp = kzalloc(sizeof(u32) * RTAX_MAX, GFP_KERNEL);
1732         if (unlikely(!mp))
1733                 return -ENOMEM;
1734
1735         nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1736                 int type = nla_type(nla);
1737                 u32 val;
1738
1739                 if (!type)
1740                         continue;
1741                 if (unlikely(type > RTAX_MAX))
1742                         goto err;
1743
1744                 if (type == RTAX_CC_ALGO) {
1745                         char tmp[TCP_CA_NAME_MAX];
1746
1747                         nla_strlcpy(tmp, nla, sizeof(tmp));
1748                         val = tcp_ca_get_key_by_name(tmp, &ecn_ca);
1749                         if (val == TCP_CA_UNSPEC)
1750                                 goto err;
1751                 } else {
1752                         val = nla_get_u32(nla);
1753                 }
1754                 if (type == RTAX_HOPLIMIT && val > 255)
1755                         val = 255;
1756                 if (type == RTAX_FEATURES && (val & ~RTAX_FEATURE_MASK))
1757                         goto err;
1758
1759                 mp[type - 1] = val;
1760                 __set_bit(type - 1, mxc->mx_valid);
1761         }
1762
1763         if (ecn_ca) {
1764                 __set_bit(RTAX_FEATURES - 1, mxc->mx_valid);
1765                 mp[RTAX_FEATURES - 1] |= DST_FEATURE_ECN_CA;
1766         }
1767
1768         mxc->mx = mp;
1769         return 0;
1770  err:
1771         kfree(mp);
1772         return -EINVAL;
1773 }
1774
1775 static struct rt6_info *ip6_nh_lookup_table(struct net *net,
1776                                             struct fib6_config *cfg,
1777                                             const struct in6_addr *gw_addr)
1778 {
1779         struct flowi6 fl6 = {
1780                 .flowi6_oif = cfg->fc_ifindex,
1781                 .daddr = *gw_addr,
1782                 .saddr = cfg->fc_prefsrc,
1783         };
1784         struct fib6_table *table;
1785         struct rt6_info *rt;
1786         int flags = RT6_LOOKUP_F_IFACE;
1787
1788         table = fib6_get_table(net, cfg->fc_table);
1789         if (!table)
1790                 return NULL;
1791
1792         if (!ipv6_addr_any(&cfg->fc_prefsrc))
1793                 flags |= RT6_LOOKUP_F_HAS_SADDR;
1794
1795         rt = ip6_pol_route(net, table, cfg->fc_ifindex, &fl6, flags);
1796
1797         /* if table lookup failed, fall back to full lookup */
1798         if (rt == net->ipv6.ip6_null_entry) {
1799                 ip6_rt_put(rt);
1800                 rt = NULL;
1801         }
1802
1803         return rt;
1804 }
1805
1806 static struct rt6_info *ip6_route_info_create(struct fib6_config *cfg)
1807 {
1808         struct net *net = cfg->fc_nlinfo.nl_net;
1809         struct rt6_info *rt = NULL;
1810         struct net_device *dev = NULL;
1811         struct inet6_dev *idev = NULL;
1812         struct fib6_table *table;
1813         int addr_type;
1814         int err = -EINVAL;
1815
1816         if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1817                 goto out;
1818 #ifndef CONFIG_IPV6_SUBTREES
1819         if (cfg->fc_src_len)
1820                 goto out;
1821 #endif
1822         if (cfg->fc_ifindex) {
1823                 err = -ENODEV;
1824                 dev = dev_get_by_index(net, cfg->fc_ifindex);
1825                 if (!dev)
1826                         goto out;
1827                 idev = in6_dev_get(dev);
1828                 if (!idev)
1829                         goto out;
1830         }
1831
1832         if (cfg->fc_metric == 0)
1833                 cfg->fc_metric = IP6_RT_PRIO_USER;
1834
1835         err = -ENOBUFS;
1836         if (cfg->fc_nlinfo.nlh &&
1837             !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
1838                 table = fib6_get_table(net, cfg->fc_table);
1839                 if (!table) {
1840                         pr_warn("NLM_F_CREATE should be specified when creating new route\n");
1841                         table = fib6_new_table(net, cfg->fc_table);
1842                 }
1843         } else {
1844                 table = fib6_new_table(net, cfg->fc_table);
1845         }
1846
1847         if (!table)
1848                 goto out;
1849
1850         rt = ip6_dst_alloc(net, NULL,
1851                            (cfg->fc_flags & RTF_ADDRCONF) ? 0 : DST_NOCOUNT);
1852
1853         if (!rt) {
1854                 err = -ENOMEM;
1855                 goto out;
1856         }
1857
1858         if (cfg->fc_flags & RTF_EXPIRES)
1859                 rt6_set_expires(rt, jiffies +
1860                                 clock_t_to_jiffies(cfg->fc_expires));
1861         else
1862                 rt6_clean_expires(rt);
1863
1864         if (cfg->fc_protocol == RTPROT_UNSPEC)
1865                 cfg->fc_protocol = RTPROT_BOOT;
1866         rt->rt6i_protocol = cfg->fc_protocol;
1867
1868         addr_type = ipv6_addr_type(&cfg->fc_dst);
1869
1870         if (addr_type & IPV6_ADDR_MULTICAST)
1871                 rt->dst.input = ip6_mc_input;
1872         else if (cfg->fc_flags & RTF_LOCAL)
1873                 rt->dst.input = ip6_input;
1874         else
1875                 rt->dst.input = ip6_forward;
1876
1877         rt->dst.output = ip6_output;
1878
1879         if (cfg->fc_encap) {
1880                 struct lwtunnel_state *lwtstate;
1881
1882                 err = lwtunnel_build_state(dev, cfg->fc_encap_type,
1883                                            cfg->fc_encap, AF_INET6, cfg,
1884                                            &lwtstate);
1885                 if (err)
1886                         goto out;
1887                 rt->dst.lwtstate = lwtstate_get(lwtstate);
1888                 if (lwtunnel_output_redirect(rt->dst.lwtstate)) {
1889                         rt->dst.lwtstate->orig_output = rt->dst.output;
1890                         rt->dst.output = lwtunnel_output;
1891                 }
1892                 if (lwtunnel_input_redirect(rt->dst.lwtstate)) {
1893                         rt->dst.lwtstate->orig_input = rt->dst.input;
1894                         rt->dst.input = lwtunnel_input;
1895                 }
1896         }
1897
1898         ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1899         rt->rt6i_dst.plen = cfg->fc_dst_len;
1900         if (rt->rt6i_dst.plen == 128)
1901                 rt->dst.flags |= DST_HOST;
1902
1903 #ifdef CONFIG_IPV6_SUBTREES
1904         ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1905         rt->rt6i_src.plen = cfg->fc_src_len;
1906 #endif
1907
1908         rt->rt6i_metric = cfg->fc_metric;
1909
1910         /* We cannot add true routes via loopback here,
1911            they would result in kernel looping; promote them to reject routes
1912          */
1913         if ((cfg->fc_flags & RTF_REJECT) ||
1914             (dev && (dev->flags & IFF_LOOPBACK) &&
1915              !(addr_type & IPV6_ADDR_LOOPBACK) &&
1916              !(cfg->fc_flags & RTF_LOCAL))) {
1917                 /* hold loopback dev/idev if we haven't done so. */
1918                 if (dev != net->loopback_dev) {
1919                         if (dev) {
1920                                 dev_put(dev);
1921                                 in6_dev_put(idev);
1922                         }
1923                         dev = net->loopback_dev;
1924                         dev_hold(dev);
1925                         idev = in6_dev_get(dev);
1926                         if (!idev) {
1927                                 err = -ENODEV;
1928                                 goto out;
1929                         }
1930                 }
1931                 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1932                 switch (cfg->fc_type) {
1933                 case RTN_BLACKHOLE:
1934                         rt->dst.error = -EINVAL;
1935                         rt->dst.output = dst_discard_out;
1936                         rt->dst.input = dst_discard;
1937                         break;
1938                 case RTN_PROHIBIT:
1939                         rt->dst.error = -EACCES;
1940                         rt->dst.output = ip6_pkt_prohibit_out;
1941                         rt->dst.input = ip6_pkt_prohibit;
1942                         break;
1943                 case RTN_THROW:
1944                 case RTN_UNREACHABLE:
1945                 default:
1946                         rt->dst.error = (cfg->fc_type == RTN_THROW) ? -EAGAIN
1947                                         : (cfg->fc_type == RTN_UNREACHABLE)
1948                                         ? -EHOSTUNREACH : -ENETUNREACH;
1949                         rt->dst.output = ip6_pkt_discard_out;
1950                         rt->dst.input = ip6_pkt_discard;
1951                         break;
1952                 }
1953                 goto install_route;
1954         }
1955
1956         if (cfg->fc_flags & RTF_GATEWAY) {
1957                 const struct in6_addr *gw_addr;
1958                 int gwa_type;
1959
1960                 gw_addr = &cfg->fc_gateway;
1961                 gwa_type = ipv6_addr_type(gw_addr);
1962
1963                 /* if gw_addr is local we will fail to detect this in case
1964                  * address is still TENTATIVE (DAD in progress). rt6_lookup()
1965                  * will return already-added prefix route via interface that
1966                  * prefix route was assigned to, which might be non-loopback.
1967                  */
1968                 err = -EINVAL;
1969                 if (ipv6_chk_addr_and_flags(net, gw_addr,
1970                                             gwa_type & IPV6_ADDR_LINKLOCAL ?
1971                                             dev : NULL, 0, 0))
1972                         goto out;
1973
1974                 rt->rt6i_gateway = *gw_addr;
1975
1976                 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1977                         struct rt6_info *grt = NULL;
1978
1979                         /* IPv6 strictly inhibits using not link-local
1980                            addresses as nexthop address.
1981                            Otherwise, router will not able to send redirects.
1982                            It is very good, but in some (rare!) circumstances
1983                            (SIT, PtP, NBMA NOARP links) it is handy to allow
1984                            some exceptions. --ANK
1985                          */
1986                         if (!(gwa_type & IPV6_ADDR_UNICAST))
1987                                 goto out;
1988
1989                         if (cfg->fc_table) {
1990                                 grt = ip6_nh_lookup_table(net, cfg, gw_addr);
1991
1992                                 if (grt) {
1993                                         if (grt->rt6i_flags & RTF_GATEWAY ||
1994                                             (dev && dev != grt->dst.dev)) {
1995                                                 ip6_rt_put(grt);
1996                                                 grt = NULL;
1997                                         }
1998                                 }
1999                         }
2000
2001                         if (!grt)
2002                                 grt = rt6_lookup(net, gw_addr, NULL,
2003                                                  cfg->fc_ifindex, 1);
2004
2005                         err = -EHOSTUNREACH;
2006                         if (!grt)
2007                                 goto out;
2008                         if (dev) {
2009                                 if (dev != grt->dst.dev) {
2010                                         ip6_rt_put(grt);
2011                                         goto out;
2012                                 }
2013                         } else {
2014                                 dev = grt->dst.dev;
2015                                 idev = grt->rt6i_idev;
2016                                 dev_hold(dev);
2017                                 in6_dev_hold(grt->rt6i_idev);
2018                         }
2019                         if (!(grt->rt6i_flags & RTF_GATEWAY))
2020                                 err = 0;
2021                         ip6_rt_put(grt);
2022
2023                         if (err)
2024                                 goto out;
2025                 }
2026                 err = -EINVAL;
2027                 if (!dev || (dev->flags & IFF_LOOPBACK))
2028                         goto out;
2029         }
2030
2031         err = -ENODEV;
2032         if (!dev)
2033                 goto out;
2034
2035         if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
2036                 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
2037                         err = -EINVAL;
2038                         goto out;
2039                 }
2040                 rt->rt6i_prefsrc.addr = cfg->fc_prefsrc;
2041                 rt->rt6i_prefsrc.plen = 128;
2042         } else
2043                 rt->rt6i_prefsrc.plen = 0;
2044
2045         rt->rt6i_flags = cfg->fc_flags;
2046
2047 install_route:
2048         rt->dst.dev = dev;
2049         rt->rt6i_idev = idev;
2050         rt->rt6i_table = table;
2051
2052         cfg->fc_nlinfo.nl_net = dev_net(dev);
2053
2054         return rt;
2055 out:
2056         if (dev)
2057                 dev_put(dev);
2058         if (idev)
2059                 in6_dev_put(idev);
2060         if (rt)
2061                 dst_free(&rt->dst);
2062
2063         return ERR_PTR(err);
2064 }
2065
2066 int ip6_route_add(struct fib6_config *cfg)
2067 {
2068         struct mx6_config mxc = { .mx = NULL, };
2069         struct rt6_info *rt;
2070         int err;
2071
2072         rt = ip6_route_info_create(cfg);
2073         if (IS_ERR(rt)) {
2074                 err = PTR_ERR(rt);
2075                 rt = NULL;
2076                 goto out;
2077         }
2078
2079         err = ip6_convert_metrics(&mxc, cfg);
2080         if (err)
2081                 goto out;
2082
2083         err = __ip6_ins_rt(rt, &cfg->fc_nlinfo, &mxc);
2084
2085         kfree(mxc.mx);
2086
2087         return err;
2088 out:
2089         if (rt)
2090                 dst_free(&rt->dst);
2091
2092         return err;
2093 }
2094
2095 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
2096 {
2097         int err;
2098         struct fib6_table *table;
2099         struct net *net = dev_net(rt->dst.dev);
2100
2101         if (rt == net->ipv6.ip6_null_entry ||
2102             rt->dst.flags & DST_NOCACHE) {
2103                 err = -ENOENT;
2104                 goto out;
2105         }
2106
2107         table = rt->rt6i_table;
2108         write_lock_bh(&table->tb6_lock);
2109         err = fib6_del(rt, info);
2110         write_unlock_bh(&table->tb6_lock);
2111
2112 out:
2113         ip6_rt_put(rt);
2114         return err;
2115 }
2116
2117 int ip6_del_rt(struct rt6_info *rt)
2118 {
2119         struct nl_info info = {
2120                 .nl_net = dev_net(rt->dst.dev),
2121         };
2122         return __ip6_del_rt(rt, &info);
2123 }
2124
2125 static int ip6_route_del(struct fib6_config *cfg)
2126 {
2127         struct fib6_table *table;
2128         struct fib6_node *fn;
2129         struct rt6_info *rt;
2130         int err = -ESRCH;
2131
2132         table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
2133         if (!table)
2134                 return err;
2135
2136         read_lock_bh(&table->tb6_lock);
2137
2138         fn = fib6_locate(&table->tb6_root,
2139                          &cfg->fc_dst, cfg->fc_dst_len,
2140                          &cfg->fc_src, cfg->fc_src_len);
2141
2142         if (fn) {
2143                 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2144                         if ((rt->rt6i_flags & RTF_CACHE) &&
2145                             !(cfg->fc_flags & RTF_CACHE))
2146                                 continue;
2147                         if (cfg->fc_ifindex &&
2148                             (!rt->dst.dev ||
2149                              rt->dst.dev->ifindex != cfg->fc_ifindex))
2150                                 continue;
2151                         if (cfg->fc_flags & RTF_GATEWAY &&
2152                             !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
2153                                 continue;
2154                         if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
2155                                 continue;
2156                         dst_hold(&rt->dst);
2157                         read_unlock_bh(&table->tb6_lock);
2158
2159                         return __ip6_del_rt(rt, &cfg->fc_nlinfo);
2160                 }
2161         }
2162         read_unlock_bh(&table->tb6_lock);
2163
2164         return err;
2165 }
2166
2167 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
2168 {
2169         struct netevent_redirect netevent;
2170         struct rt6_info *rt, *nrt = NULL;
2171         struct ndisc_options ndopts;
2172         struct inet6_dev *in6_dev;
2173         struct neighbour *neigh;
2174         struct rd_msg *msg;
2175         int optlen, on_link;
2176         u8 *lladdr;
2177
2178         optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
2179         optlen -= sizeof(*msg);
2180
2181         if (optlen < 0) {
2182                 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
2183                 return;
2184         }
2185
2186         msg = (struct rd_msg *)icmp6_hdr(skb);
2187
2188         if (ipv6_addr_is_multicast(&msg->dest)) {
2189                 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
2190                 return;
2191         }
2192
2193         on_link = 0;
2194         if (ipv6_addr_equal(&msg->dest, &msg->target)) {
2195                 on_link = 1;
2196         } else if (ipv6_addr_type(&msg->target) !=
2197                    (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
2198                 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
2199                 return;
2200         }
2201
2202         in6_dev = __in6_dev_get(skb->dev);
2203         if (!in6_dev)
2204                 return;
2205         if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
2206                 return;
2207
2208         /* RFC2461 8.1:
2209          *      The IP source address of the Redirect MUST be the same as the current
2210          *      first-hop router for the specified ICMP Destination Address.
2211          */
2212
2213         if (!ndisc_parse_options(skb->dev, msg->opt, optlen, &ndopts)) {
2214                 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
2215                 return;
2216         }
2217
2218         lladdr = NULL;
2219         if (ndopts.nd_opts_tgt_lladdr) {
2220                 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
2221                                              skb->dev);
2222                 if (!lladdr) {
2223                         net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
2224                         return;
2225                 }
2226         }
2227
2228         rt = (struct rt6_info *) dst;
2229         if (rt->rt6i_flags & RTF_REJECT) {
2230                 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
2231                 return;
2232         }
2233
2234         /* Redirect received -> path was valid.
2235          * Look, redirects are sent only in response to data packets,
2236          * so that this nexthop apparently is reachable. --ANK
2237          */
2238         dst_confirm(&rt->dst);
2239
2240         neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
2241         if (!neigh)
2242                 return;
2243
2244         /*
2245          *      We have finally decided to accept it.
2246          */
2247
2248         ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
2249                      NEIGH_UPDATE_F_WEAK_OVERRIDE|
2250                      NEIGH_UPDATE_F_OVERRIDE|
2251                      (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
2252                                      NEIGH_UPDATE_F_ISROUTER)),
2253                      NDISC_REDIRECT, &ndopts);
2254
2255         nrt = ip6_rt_cache_alloc(rt, &msg->dest, NULL);
2256         if (!nrt)
2257                 goto out;
2258
2259         nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
2260         if (on_link)
2261                 nrt->rt6i_flags &= ~RTF_GATEWAY;
2262
2263         nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
2264
2265         if (ip6_ins_rt(nrt))
2266                 goto out;
2267
2268         netevent.old = &rt->dst;
2269         netevent.new = &nrt->dst;
2270         netevent.daddr = &msg->dest;
2271         netevent.neigh = neigh;
2272         call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
2273
2274         if (rt->rt6i_flags & RTF_CACHE) {
2275                 rt = (struct rt6_info *) dst_clone(&rt->dst);
2276                 ip6_del_rt(rt);
2277         }
2278
2279 out:
2280         neigh_release(neigh);
2281 }
2282
2283 /*
2284  *      Misc support functions
2285  */
2286
2287 static void rt6_set_from(struct rt6_info *rt, struct rt6_info *from)
2288 {
2289         BUG_ON(from->dst.from);
2290
2291         rt->rt6i_flags &= ~RTF_EXPIRES;
2292         dst_hold(&from->dst);
2293         rt->dst.from = &from->dst;
2294         dst_init_metrics(&rt->dst, dst_metrics_ptr(&from->dst), true);
2295 }
2296
2297 static void ip6_rt_copy_init(struct rt6_info *rt, struct rt6_info *ort)
2298 {
2299         rt->dst.input = ort->dst.input;
2300         rt->dst.output = ort->dst.output;
2301         rt->rt6i_dst = ort->rt6i_dst;
2302         rt->dst.error = ort->dst.error;
2303         rt->rt6i_idev = ort->rt6i_idev;
2304         if (rt->rt6i_idev)
2305                 in6_dev_hold(rt->rt6i_idev);
2306         rt->dst.lastuse = jiffies;
2307         rt->rt6i_gateway = ort->rt6i_gateway;
2308         rt->rt6i_flags = ort->rt6i_flags;
2309         rt6_set_from(rt, ort);
2310         rt->rt6i_metric = ort->rt6i_metric;
2311 #ifdef CONFIG_IPV6_SUBTREES
2312         rt->rt6i_src = ort->rt6i_src;
2313 #endif
2314         rt->rt6i_prefsrc = ort->rt6i_prefsrc;
2315         rt->rt6i_table = ort->rt6i_table;
2316         rt->dst.lwtstate = lwtstate_get(ort->dst.lwtstate);
2317 }
2318
2319 #ifdef CONFIG_IPV6_ROUTE_INFO
2320 static struct rt6_info *rt6_get_route_info(struct net *net,
2321                                            const struct in6_addr *prefix, int prefixlen,
2322                                            const struct in6_addr *gwaddr, int ifindex)
2323 {
2324         struct fib6_node *fn;
2325         struct rt6_info *rt = NULL;
2326         struct fib6_table *table;
2327
2328         table = fib6_get_table(net, RT6_TABLE_INFO);
2329         if (!table)
2330                 return NULL;
2331
2332         read_lock_bh(&table->tb6_lock);
2333         fn = fib6_locate(&table->tb6_root, prefix, prefixlen, NULL, 0);
2334         if (!fn)
2335                 goto out;
2336
2337         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2338                 if (rt->dst.dev->ifindex != ifindex)
2339                         continue;
2340                 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
2341                         continue;
2342                 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
2343                         continue;
2344                 dst_hold(&rt->dst);
2345                 break;
2346         }
2347 out:
2348         read_unlock_bh(&table->tb6_lock);
2349         return rt;
2350 }
2351
2352 static struct rt6_info *rt6_add_route_info(struct net *net,
2353                                            const struct in6_addr *prefix, int prefixlen,
2354                                            const struct in6_addr *gwaddr, int ifindex,
2355                                            unsigned int pref)
2356 {
2357         struct fib6_config cfg = {
2358                 .fc_metric      = IP6_RT_PRIO_USER,
2359                 .fc_ifindex     = ifindex,
2360                 .fc_dst_len     = prefixlen,
2361                 .fc_flags       = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
2362                                   RTF_UP | RTF_PREF(pref),
2363                 .fc_nlinfo.portid = 0,
2364                 .fc_nlinfo.nlh = NULL,
2365                 .fc_nlinfo.nl_net = net,
2366         };
2367
2368         cfg.fc_table = l3mdev_fib_table_by_index(net, ifindex) ? : RT6_TABLE_INFO;
2369         cfg.fc_dst = *prefix;
2370         cfg.fc_gateway = *gwaddr;
2371
2372         /* We should treat it as a default route if prefix length is 0. */
2373         if (!prefixlen)
2374                 cfg.fc_flags |= RTF_DEFAULT;
2375
2376         ip6_route_add(&cfg);
2377
2378         return rt6_get_route_info(net, prefix, prefixlen, gwaddr, ifindex);
2379 }
2380 #endif
2381
2382 struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
2383 {
2384         struct rt6_info *rt;
2385         struct fib6_table *table;
2386
2387         table = fib6_get_table(dev_net(dev), RT6_TABLE_DFLT);
2388         if (!table)
2389                 return NULL;
2390
2391         read_lock_bh(&table->tb6_lock);
2392         for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2393                 if (dev == rt->dst.dev &&
2394                     ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
2395                     ipv6_addr_equal(&rt->rt6i_gateway, addr))
2396                         break;
2397         }
2398         if (rt)
2399                 dst_hold(&rt->dst);
2400         read_unlock_bh(&table->tb6_lock);
2401         return rt;
2402 }
2403
2404 struct rt6_info *rt6_add_dflt_router(const struct in6_addr *gwaddr,
2405                                      struct net_device *dev,
2406                                      unsigned int pref)
2407 {
2408         struct fib6_config cfg = {
2409                 .fc_table       = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT,
2410                 .fc_metric      = IP6_RT_PRIO_USER,
2411                 .fc_ifindex     = dev->ifindex,
2412                 .fc_flags       = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
2413                                   RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
2414                 .fc_nlinfo.portid = 0,
2415                 .fc_nlinfo.nlh = NULL,
2416                 .fc_nlinfo.nl_net = dev_net(dev),
2417         };
2418
2419         cfg.fc_gateway = *gwaddr;
2420
2421         ip6_route_add(&cfg);
2422
2423         return rt6_get_dflt_router(gwaddr, dev);
2424 }
2425
2426 void rt6_purge_dflt_routers(struct net *net)
2427 {
2428         struct rt6_info *rt;
2429         struct fib6_table *table;
2430
2431         /* NOTE: Keep consistent with rt6_get_dflt_router */
2432         table = fib6_get_table(net, RT6_TABLE_DFLT);
2433         if (!table)
2434                 return;
2435
2436 restart:
2437         read_lock_bh(&table->tb6_lock);
2438         for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2439                 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
2440                     (!rt->rt6i_idev || rt->rt6i_idev->cnf.accept_ra != 2)) {
2441                         dst_hold(&rt->dst);
2442                         read_unlock_bh(&table->tb6_lock);
2443                         ip6_del_rt(rt);
2444                         goto restart;
2445                 }
2446         }
2447         read_unlock_bh(&table->tb6_lock);
2448 }
2449
2450 static void rtmsg_to_fib6_config(struct net *net,
2451                                  struct in6_rtmsg *rtmsg,
2452                                  struct fib6_config *cfg)
2453 {
2454         memset(cfg, 0, sizeof(*cfg));
2455
2456         cfg->fc_table = l3mdev_fib_table_by_index(net, rtmsg->rtmsg_ifindex) ?
2457                          : RT6_TABLE_MAIN;
2458         cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
2459         cfg->fc_metric = rtmsg->rtmsg_metric;
2460         cfg->fc_expires = rtmsg->rtmsg_info;
2461         cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
2462         cfg->fc_src_len = rtmsg->rtmsg_src_len;
2463         cfg->fc_flags = rtmsg->rtmsg_flags;
2464
2465         cfg->fc_nlinfo.nl_net = net;
2466
2467         cfg->fc_dst = rtmsg->rtmsg_dst;
2468         cfg->fc_src = rtmsg->rtmsg_src;
2469         cfg->fc_gateway = rtmsg->rtmsg_gateway;
2470 }
2471
2472 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
2473 {
2474         struct fib6_config cfg;
2475         struct in6_rtmsg rtmsg;
2476         int err;
2477
2478         switch (cmd) {
2479         case SIOCADDRT:         /* Add a route */
2480         case SIOCDELRT:         /* Delete a route */
2481                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2482                         return -EPERM;
2483                 err = copy_from_user(&rtmsg, arg,
2484                                      sizeof(struct in6_rtmsg));
2485                 if (err)
2486                         return -EFAULT;
2487
2488                 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
2489
2490                 rtnl_lock();
2491                 switch (cmd) {
2492                 case SIOCADDRT:
2493                         err = ip6_route_add(&cfg);
2494                         break;
2495                 case SIOCDELRT:
2496                         err = ip6_route_del(&cfg);
2497                         break;
2498                 default:
2499                         err = -EINVAL;
2500                 }
2501                 rtnl_unlock();
2502
2503                 return err;
2504         }
2505
2506         return -EINVAL;
2507 }
2508
2509 /*
2510  *      Drop the packet on the floor
2511  */
2512
2513 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
2514 {
2515         int type;
2516         struct dst_entry *dst = skb_dst(skb);
2517         switch (ipstats_mib_noroutes) {
2518         case IPSTATS_MIB_INNOROUTES:
2519                 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
2520                 if (type == IPV6_ADDR_ANY) {
2521                         IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2522                                       IPSTATS_MIB_INADDRERRORS);
2523                         break;
2524                 }
2525                 /* FALLTHROUGH */
2526         case IPSTATS_MIB_OUTNOROUTES:
2527                 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2528                               ipstats_mib_noroutes);
2529                 break;
2530         }
2531         icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
2532         kfree_skb(skb);
2533         return 0;
2534 }
2535
2536 static int ip6_pkt_discard(struct sk_buff *skb)
2537 {
2538         return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
2539 }
2540
2541 static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb)
2542 {
2543         skb->dev = skb_dst(skb)->dev;
2544         return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
2545 }
2546
2547 static int ip6_pkt_prohibit(struct sk_buff *skb)
2548 {
2549         return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
2550 }
2551
2552 static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb)
2553 {
2554         skb->dev = skb_dst(skb)->dev;
2555         return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
2556 }
2557
2558 /*
2559  *      Allocate a dst for local (unicast / anycast) address.
2560  */
2561
2562 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
2563                                     const struct in6_addr *addr,
2564                                     bool anycast)
2565 {
2566         u32 tb_id;
2567         struct net *net = dev_net(idev->dev);
2568         struct rt6_info *rt = ip6_dst_alloc(net, net->loopback_dev,
2569                                             DST_NOCOUNT);
2570         if (!rt)
2571                 return ERR_PTR(-ENOMEM);
2572
2573         in6_dev_hold(idev);
2574
2575         rt->dst.flags |= DST_HOST;
2576         rt->dst.input = ip6_input;
2577         rt->dst.output = ip6_output;
2578         rt->rt6i_idev = idev;
2579
2580         rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
2581         if (anycast)
2582                 rt->rt6i_flags |= RTF_ANYCAST;
2583         else
2584                 rt->rt6i_flags |= RTF_LOCAL;
2585
2586         rt->rt6i_gateway  = *addr;
2587         rt->rt6i_dst.addr = *addr;
2588         rt->rt6i_dst.plen = 128;
2589         tb_id = l3mdev_fib_table(idev->dev) ? : RT6_TABLE_LOCAL;
2590         rt->rt6i_table = fib6_get_table(net, tb_id);
2591         rt->dst.flags |= DST_NOCACHE;
2592
2593         atomic_set(&rt->dst.__refcnt, 1);
2594
2595         return rt;
2596 }
2597
2598 /* remove deleted ip from prefsrc entries */
2599 struct arg_dev_net_ip {
2600         struct net_device *dev;
2601         struct net *net;
2602         struct in6_addr *addr;
2603 };
2604
2605 static int fib6_remove_prefsrc(struct rt6_info *rt, void *arg)
2606 {
2607         struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
2608         struct net *net = ((struct arg_dev_net_ip *)arg)->net;
2609         struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
2610
2611         if (((void *)rt->dst.dev == dev || !dev) &&
2612             rt != net->ipv6.ip6_null_entry &&
2613             ipv6_addr_equal(addr, &rt->rt6i_prefsrc.addr)) {
2614                 /* remove prefsrc entry */
2615                 rt->rt6i_prefsrc.plen = 0;
2616         }
2617         return 0;
2618 }
2619
2620 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
2621 {
2622         struct net *net = dev_net(ifp->idev->dev);
2623         struct arg_dev_net_ip adni = {
2624                 .dev = ifp->idev->dev,
2625                 .net = net,
2626                 .addr = &ifp->addr,
2627         };
2628         fib6_clean_all(net, fib6_remove_prefsrc, &adni);
2629 }
2630
2631 #define RTF_RA_ROUTER           (RTF_ADDRCONF | RTF_DEFAULT | RTF_GATEWAY)
2632 #define RTF_CACHE_GATEWAY       (RTF_GATEWAY | RTF_CACHE)
2633
2634 /* Remove routers and update dst entries when gateway turn into host. */
2635 static int fib6_clean_tohost(struct rt6_info *rt, void *arg)
2636 {
2637         struct in6_addr *gateway = (struct in6_addr *)arg;
2638
2639         if ((((rt->rt6i_flags & RTF_RA_ROUTER) == RTF_RA_ROUTER) ||
2640              ((rt->rt6i_flags & RTF_CACHE_GATEWAY) == RTF_CACHE_GATEWAY)) &&
2641              ipv6_addr_equal(gateway, &rt->rt6i_gateway)) {
2642                 return -1;
2643         }
2644         return 0;
2645 }
2646
2647 void rt6_clean_tohost(struct net *net, struct in6_addr *gateway)
2648 {
2649         fib6_clean_all(net, fib6_clean_tohost, gateway);
2650 }
2651
2652 struct arg_dev_net {
2653         struct net_device *dev;
2654         struct net *net;
2655 };
2656
2657 static int fib6_ifdown(struct rt6_info *rt, void *arg)
2658 {
2659         const struct arg_dev_net *adn = arg;
2660         const struct net_device *dev = adn->dev;
2661
2662         if ((rt->dst.dev == dev || !dev) &&
2663             rt != adn->net->ipv6.ip6_null_entry)
2664                 return -1;
2665
2666         return 0;
2667 }
2668
2669 void rt6_ifdown(struct net *net, struct net_device *dev)
2670 {
2671         struct arg_dev_net adn = {
2672                 .dev = dev,
2673                 .net = net,
2674         };
2675
2676         fib6_clean_all(net, fib6_ifdown, &adn);
2677         icmp6_clean_all(fib6_ifdown, &adn);
2678         if (dev)
2679                 rt6_uncached_list_flush_dev(net, dev);
2680 }
2681
2682 struct rt6_mtu_change_arg {
2683         struct net_device *dev;
2684         unsigned int mtu;
2685 };
2686
2687 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2688 {
2689         struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2690         struct inet6_dev *idev;
2691
2692         /* In IPv6 pmtu discovery is not optional,
2693            so that RTAX_MTU lock cannot disable it.
2694            We still use this lock to block changes
2695            caused by addrconf/ndisc.
2696         */
2697
2698         idev = __in6_dev_get(arg->dev);
2699         if (!idev)
2700                 return 0;
2701
2702         /* For administrative MTU increase, there is no way to discover
2703            IPv6 PMTU increase, so PMTU increase should be updated here.
2704            Since RFC 1981 doesn't include administrative MTU increase
2705            update PMTU increase is a MUST. (i.e. jumbo frame)
2706          */
2707         /*
2708            If new MTU is less than route PMTU, this new MTU will be the
2709            lowest MTU in the path, update the route PMTU to reflect PMTU
2710            decreases; if new MTU is greater than route PMTU, and the
2711            old MTU is the lowest MTU in the path, update the route PMTU
2712            to reflect the increase. In this case if the other nodes' MTU
2713            also have the lowest MTU, TOO BIG MESSAGE will be lead to
2714            PMTU discouvery.
2715          */
2716         if (rt->dst.dev == arg->dev &&
2717             !dst_metric_locked(&rt->dst, RTAX_MTU)) {
2718                 if (rt->rt6i_flags & RTF_CACHE) {
2719                         /* For RTF_CACHE with rt6i_pmtu == 0
2720                          * (i.e. a redirected route),
2721                          * the metrics of its rt->dst.from has already
2722                          * been updated.
2723                          */
2724                         if (rt->rt6i_pmtu && rt->rt6i_pmtu > arg->mtu)
2725                                 rt->rt6i_pmtu = arg->mtu;
2726                 } else if (dst_mtu(&rt->dst) >= arg->mtu ||
2727                            (dst_mtu(&rt->dst) < arg->mtu &&
2728                             dst_mtu(&rt->dst) == idev->cnf.mtu6)) {
2729                         dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
2730                 }
2731         }
2732         return 0;
2733 }
2734
2735 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
2736 {
2737         struct rt6_mtu_change_arg arg = {
2738                 .dev = dev,
2739                 .mtu = mtu,
2740         };
2741
2742         fib6_clean_all(dev_net(dev), rt6_mtu_change_route, &arg);
2743 }
2744
2745 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2746         [RTA_GATEWAY]           = { .len = sizeof(struct in6_addr) },
2747         [RTA_OIF]               = { .type = NLA_U32 },
2748         [RTA_IIF]               = { .type = NLA_U32 },
2749         [RTA_PRIORITY]          = { .type = NLA_U32 },
2750         [RTA_METRICS]           = { .type = NLA_NESTED },
2751         [RTA_MULTIPATH]         = { .len = sizeof(struct rtnexthop) },
2752         [RTA_PREF]              = { .type = NLA_U8 },
2753         [RTA_ENCAP_TYPE]        = { .type = NLA_U16 },
2754         [RTA_ENCAP]             = { .type = NLA_NESTED },
2755         [RTA_EXPIRES]           = { .type = NLA_U32 },
2756 };
2757
2758 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2759                               struct fib6_config *cfg)
2760 {
2761         struct rtmsg *rtm;
2762         struct nlattr *tb[RTA_MAX+1];
2763         unsigned int pref;
2764         int err;
2765
2766         err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2767         if (err < 0)
2768                 goto errout;
2769
2770         err = -EINVAL;
2771         rtm = nlmsg_data(nlh);
2772         memset(cfg, 0, sizeof(*cfg));
2773
2774         cfg->fc_table = rtm->rtm_table;
2775         cfg->fc_dst_len = rtm->rtm_dst_len;
2776         cfg->fc_src_len = rtm->rtm_src_len;
2777         cfg->fc_flags = RTF_UP;
2778         cfg->fc_protocol = rtm->rtm_protocol;
2779         cfg->fc_type = rtm->rtm_type;
2780
2781         if (rtm->rtm_type == RTN_UNREACHABLE ||
2782             rtm->rtm_type == RTN_BLACKHOLE ||
2783             rtm->rtm_type == RTN_PROHIBIT ||
2784             rtm->rtm_type == RTN_THROW)
2785                 cfg->fc_flags |= RTF_REJECT;
2786
2787         if (rtm->rtm_type == RTN_LOCAL)
2788                 cfg->fc_flags |= RTF_LOCAL;
2789
2790         if (rtm->rtm_flags & RTM_F_CLONED)
2791                 cfg->fc_flags |= RTF_CACHE;
2792
2793         cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid;
2794         cfg->fc_nlinfo.nlh = nlh;
2795         cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2796
2797         if (tb[RTA_GATEWAY]) {
2798                 cfg->fc_gateway = nla_get_in6_addr(tb[RTA_GATEWAY]);
2799                 cfg->fc_flags |= RTF_GATEWAY;
2800         }
2801
2802         if (tb[RTA_DST]) {
2803                 int plen = (rtm->rtm_dst_len + 7) >> 3;
2804
2805                 if (nla_len(tb[RTA_DST]) < plen)
2806                         goto errout;
2807
2808                 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2809         }
2810
2811         if (tb[RTA_SRC]) {
2812                 int plen = (rtm->rtm_src_len + 7) >> 3;
2813
2814                 if (nla_len(tb[RTA_SRC]) < plen)
2815                         goto errout;
2816
2817                 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2818         }
2819
2820         if (tb[RTA_PREFSRC])
2821                 cfg->fc_prefsrc = nla_get_in6_addr(tb[RTA_PREFSRC]);
2822
2823         if (tb[RTA_OIF])
2824                 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2825
2826         if (tb[RTA_PRIORITY])
2827                 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2828
2829         if (tb[RTA_METRICS]) {
2830                 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2831                 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2832         }
2833
2834         if (tb[RTA_TABLE])
2835                 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2836
2837         if (tb[RTA_MULTIPATH]) {
2838                 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
2839                 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
2840         }
2841
2842         if (tb[RTA_PREF]) {
2843                 pref = nla_get_u8(tb[RTA_PREF]);
2844                 if (pref != ICMPV6_ROUTER_PREF_LOW &&
2845                     pref != ICMPV6_ROUTER_PREF_HIGH)
2846                         pref = ICMPV6_ROUTER_PREF_MEDIUM;
2847                 cfg->fc_flags |= RTF_PREF(pref);
2848         }
2849
2850         if (tb[RTA_ENCAP])
2851                 cfg->fc_encap = tb[RTA_ENCAP];
2852
2853         if (tb[RTA_ENCAP_TYPE])
2854                 cfg->fc_encap_type = nla_get_u16(tb[RTA_ENCAP_TYPE]);
2855
2856         if (tb[RTA_EXPIRES]) {
2857                 unsigned long timeout = addrconf_timeout_fixup(nla_get_u32(tb[RTA_EXPIRES]), HZ);
2858
2859                 if (addrconf_finite_timeout(timeout)) {
2860                         cfg->fc_expires = jiffies_to_clock_t(timeout * HZ);
2861                         cfg->fc_flags |= RTF_EXPIRES;
2862                 }
2863         }
2864
2865         err = 0;
2866 errout:
2867         return err;
2868 }
2869
2870 struct rt6_nh {
2871         struct rt6_info *rt6_info;
2872         struct fib6_config r_cfg;
2873         struct mx6_config mxc;
2874         struct list_head next;
2875 };
2876
2877 static void ip6_print_replace_route_err(struct list_head *rt6_nh_list)
2878 {
2879         struct rt6_nh *nh;
2880
2881         list_for_each_entry(nh, rt6_nh_list, next) {
2882                 pr_warn("IPV6: multipath route replace failed (check consistency of installed routes): %pI6 nexthop %pI6 ifi %d\n",
2883                         &nh->r_cfg.fc_dst, &nh->r_cfg.fc_gateway,
2884                         nh->r_cfg.fc_ifindex);
2885         }
2886 }
2887
2888 static int ip6_route_info_append(struct list_head *rt6_nh_list,
2889                                  struct rt6_info *rt, struct fib6_config *r_cfg)
2890 {
2891         struct rt6_nh *nh;
2892         struct rt6_info *rtnh;
2893         int err = -EEXIST;
2894
2895         list_for_each_entry(nh, rt6_nh_list, next) {
2896                 /* check if rt6_info already exists */
2897                 rtnh = nh->rt6_info;
2898
2899                 if (rtnh->dst.dev == rt->dst.dev &&
2900                     rtnh->rt6i_idev == rt->rt6i_idev &&
2901                     ipv6_addr_equal(&rtnh->rt6i_gateway,
2902                                     &rt->rt6i_gateway))
2903                         return err;
2904         }
2905
2906         nh = kzalloc(sizeof(*nh), GFP_KERNEL);
2907         if (!nh)
2908                 return -ENOMEM;
2909         nh->rt6_info = rt;
2910         err = ip6_convert_metrics(&nh->mxc, r_cfg);
2911         if (err) {
2912                 kfree(nh);
2913                 return err;
2914         }
2915         memcpy(&nh->r_cfg, r_cfg, sizeof(*r_cfg));
2916         list_add_tail(&nh->next, rt6_nh_list);
2917
2918         return 0;
2919 }
2920
2921 static int ip6_route_multipath_add(struct fib6_config *cfg)
2922 {
2923         struct fib6_config r_cfg;
2924         struct rtnexthop *rtnh;
2925         struct rt6_info *rt;
2926         struct rt6_nh *err_nh;
2927         struct rt6_nh *nh, *nh_safe;
2928         int remaining;
2929         int attrlen;
2930         int err = 1;
2931         int nhn = 0;
2932         int replace = (cfg->fc_nlinfo.nlh &&
2933                        (cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_REPLACE));
2934         LIST_HEAD(rt6_nh_list);
2935
2936         remaining = cfg->fc_mp_len;
2937         rtnh = (struct rtnexthop *)cfg->fc_mp;
2938
2939         /* Parse a Multipath Entry and build a list (rt6_nh_list) of
2940          * rt6_info structs per nexthop
2941          */
2942         while (rtnh_ok(rtnh, remaining)) {
2943                 memcpy(&r_cfg, cfg, sizeof(*cfg));
2944                 if (rtnh->rtnh_ifindex)
2945                         r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
2946
2947                 attrlen = rtnh_attrlen(rtnh);
2948                 if (attrlen > 0) {
2949                         struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
2950
2951                         nla = nla_find(attrs, attrlen, RTA_GATEWAY);
2952                         if (nla) {
2953                                 r_cfg.fc_gateway = nla_get_in6_addr(nla);
2954                                 r_cfg.fc_flags |= RTF_GATEWAY;
2955                         }
2956                         r_cfg.fc_encap = nla_find(attrs, attrlen, RTA_ENCAP);
2957                         nla = nla_find(attrs, attrlen, RTA_ENCAP_TYPE);
2958                         if (nla)
2959                                 r_cfg.fc_encap_type = nla_get_u16(nla);
2960                 }
2961
2962                 rt = ip6_route_info_create(&r_cfg);
2963                 if (IS_ERR(rt)) {
2964                         err = PTR_ERR(rt);
2965                         rt = NULL;
2966                         goto cleanup;
2967                 }
2968
2969                 err = ip6_route_info_append(&rt6_nh_list, rt, &r_cfg);
2970                 if (err) {
2971                         dst_free(&rt->dst);
2972                         goto cleanup;
2973                 }
2974
2975                 rtnh = rtnh_next(rtnh, &remaining);
2976         }
2977
2978         err_nh = NULL;
2979         list_for_each_entry(nh, &rt6_nh_list, next) {
2980                 err = __ip6_ins_rt(nh->rt6_info, &cfg->fc_nlinfo, &nh->mxc);
2981                 /* nh->rt6_info is used or freed at this point, reset to NULL*/
2982                 nh->rt6_info = NULL;
2983                 if (err) {
2984                         if (replace && nhn)
2985                                 ip6_print_replace_route_err(&rt6_nh_list);
2986                         err_nh = nh;
2987                         goto add_errout;
2988                 }
2989
2990                 /* Because each route is added like a single route we remove
2991                  * these flags after the first nexthop: if there is a collision,
2992                  * we have already failed to add the first nexthop:
2993                  * fib6_add_rt2node() has rejected it; when replacing, old
2994                  * nexthops have been replaced by first new, the rest should
2995                  * be added to it.
2996                  */
2997                 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~(NLM_F_EXCL |
2998                                                      NLM_F_REPLACE);
2999                 nhn++;
3000         }
3001
3002         goto cleanup;
3003
3004 add_errout:
3005         /* Delete routes that were already added */
3006         list_for_each_entry(nh, &rt6_nh_list, next) {
3007                 if (err_nh == nh)
3008                         break;
3009                 ip6_route_del(&nh->r_cfg);
3010         }
3011
3012 cleanup:
3013         list_for_each_entry_safe(nh, nh_safe, &rt6_nh_list, next) {
3014                 if (nh->rt6_info)
3015                         dst_free(&nh->rt6_info->dst);
3016                 kfree(nh->mxc.mx);
3017                 list_del(&nh->next);
3018                 kfree(nh);
3019         }
3020
3021         return err;
3022 }
3023
3024 static int ip6_route_multipath_del(struct fib6_config *cfg)
3025 {
3026         struct fib6_config r_cfg;
3027         struct rtnexthop *rtnh;
3028         int remaining;
3029         int attrlen;
3030         int err = 1, last_err = 0;
3031
3032         remaining = cfg->fc_mp_len;
3033         rtnh = (struct rtnexthop *)cfg->fc_mp;
3034
3035         /* Parse a Multipath Entry */
3036         while (rtnh_ok(rtnh, remaining)) {
3037                 memcpy(&r_cfg, cfg, sizeof(*cfg));
3038                 if (rtnh->rtnh_ifindex)
3039                         r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
3040
3041                 attrlen = rtnh_attrlen(rtnh);
3042                 if (attrlen > 0) {
3043                         struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
3044
3045                         nla = nla_find(attrs, attrlen, RTA_GATEWAY);
3046                         if (nla) {
3047                                 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
3048                                 r_cfg.fc_flags |= RTF_GATEWAY;
3049                         }
3050                 }
3051                 err = ip6_route_del(&r_cfg);
3052                 if (err)
3053                         last_err = err;
3054
3055                 rtnh = rtnh_next(rtnh, &remaining);
3056         }
3057
3058         return last_err;
3059 }
3060
3061 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh)
3062 {
3063         struct fib6_config cfg;
3064         int err;
3065
3066         err = rtm_to_fib6_config(skb, nlh, &cfg);
3067         if (err < 0)
3068                 return err;
3069
3070         if (cfg.fc_mp)
3071                 return ip6_route_multipath_del(&cfg);
3072         else
3073                 return ip6_route_del(&cfg);
3074 }
3075
3076 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh)
3077 {
3078         struct fib6_config cfg;
3079         int err;
3080
3081         err = rtm_to_fib6_config(skb, nlh, &cfg);
3082         if (err < 0)
3083                 return err;
3084
3085         if (cfg.fc_mp)
3086                 return ip6_route_multipath_add(&cfg);
3087         else
3088                 return ip6_route_add(&cfg);
3089 }
3090
3091 static inline size_t rt6_nlmsg_size(struct rt6_info *rt)
3092 {
3093         return NLMSG_ALIGN(sizeof(struct rtmsg))
3094                + nla_total_size(16) /* RTA_SRC */
3095                + nla_total_size(16) /* RTA_DST */
3096                + nla_total_size(16) /* RTA_GATEWAY */
3097                + nla_total_size(16) /* RTA_PREFSRC */
3098                + nla_total_size(4) /* RTA_TABLE */
3099                + nla_total_size(4) /* RTA_IIF */
3100                + nla_total_size(4) /* RTA_OIF */
3101                + nla_total_size(4) /* RTA_PRIORITY */
3102                + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
3103                + nla_total_size(sizeof(struct rta_cacheinfo))
3104                + nla_total_size(TCP_CA_NAME_MAX) /* RTAX_CC_ALGO */
3105                + nla_total_size(1) /* RTA_PREF */
3106                + lwtunnel_get_encap_size(rt->dst.lwtstate);
3107 }
3108
3109 static int rt6_fill_node(struct net *net,
3110                          struct sk_buff *skb, struct rt6_info *rt,
3111                          struct in6_addr *dst, struct in6_addr *src,
3112                          int iif, int type, u32 portid, u32 seq,
3113                          int prefix, int nowait, unsigned int flags)
3114 {
3115         u32 metrics[RTAX_MAX];
3116         struct rtmsg *rtm;
3117         struct nlmsghdr *nlh;
3118         long expires;
3119         u32 table;
3120
3121         if (prefix) {   /* user wants prefix routes only */
3122                 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
3123                         /* success since this is not a prefix route */
3124                         return 1;
3125                 }
3126         }
3127
3128         nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
3129         if (!nlh)
3130                 return -EMSGSIZE;
3131
3132         rtm = nlmsg_data(nlh);
3133         rtm->rtm_family = AF_INET6;
3134         rtm->rtm_dst_len = rt->rt6i_dst.plen;
3135         rtm->rtm_src_len = rt->rt6i_src.plen;
3136         rtm->rtm_tos = 0;
3137         if (rt->rt6i_table)
3138                 table = rt->rt6i_table->tb6_id;
3139         else
3140                 table = RT6_TABLE_UNSPEC;
3141         rtm->rtm_table = table;
3142         if (nla_put_u32(skb, RTA_TABLE, table))
3143                 goto nla_put_failure;
3144         if (rt->rt6i_flags & RTF_REJECT) {
3145                 switch (rt->dst.error) {
3146                 case -EINVAL:
3147                         rtm->rtm_type = RTN_BLACKHOLE;
3148                         break;
3149                 case -EACCES:
3150                         rtm->rtm_type = RTN_PROHIBIT;
3151                         break;
3152                 case -EAGAIN:
3153                         rtm->rtm_type = RTN_THROW;
3154                         break;
3155                 default:
3156                         rtm->rtm_type = RTN_UNREACHABLE;
3157                         break;
3158                 }
3159         }
3160         else if (rt->rt6i_flags & RTF_LOCAL)
3161                 rtm->rtm_type = RTN_LOCAL;
3162         else if (rt->dst.dev && (rt->dst.dev->flags & IFF_LOOPBACK))
3163                 rtm->rtm_type = RTN_LOCAL;
3164         else
3165                 rtm->rtm_type = RTN_UNICAST;
3166         rtm->rtm_flags = 0;
3167         if (!netif_carrier_ok(rt->dst.dev)) {
3168                 rtm->rtm_flags |= RTNH_F_LINKDOWN;
3169                 if (rt->rt6i_idev->cnf.ignore_routes_with_linkdown)
3170                         rtm->rtm_flags |= RTNH_F_DEAD;
3171         }
3172         rtm->rtm_scope = RT_SCOPE_UNIVERSE;
3173         rtm->rtm_protocol = rt->rt6i_protocol;
3174         if (rt->rt6i_flags & RTF_DYNAMIC)
3175                 rtm->rtm_protocol = RTPROT_REDIRECT;
3176         else if (rt->rt6i_flags & RTF_ADDRCONF) {
3177                 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ROUTEINFO))
3178                         rtm->rtm_protocol = RTPROT_RA;
3179                 else
3180                         rtm->rtm_protocol = RTPROT_KERNEL;
3181         }
3182
3183         if (rt->rt6i_flags & RTF_CACHE)
3184                 rtm->rtm_flags |= RTM_F_CLONED;
3185
3186         if (dst) {
3187                 if (nla_put_in6_addr(skb, RTA_DST, dst))
3188                         goto nla_put_failure;
3189                 rtm->rtm_dst_len = 128;
3190         } else if (rtm->rtm_dst_len)
3191                 if (nla_put_in6_addr(skb, RTA_DST, &rt->rt6i_dst.addr))
3192                         goto nla_put_failure;
3193 #ifdef CONFIG_IPV6_SUBTREES
3194         if (src) {
3195                 if (nla_put_in6_addr(skb, RTA_SRC, src))
3196                         goto nla_put_failure;
3197                 rtm->rtm_src_len = 128;
3198         } else if (rtm->rtm_src_len &&
3199                    nla_put_in6_addr(skb, RTA_SRC, &rt->rt6i_src.addr))
3200                 goto nla_put_failure;
3201 #endif
3202         if (iif) {
3203 #ifdef CONFIG_IPV6_MROUTE
3204                 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
3205                         int err = ip6mr_get_route(net, skb, rtm, nowait);
3206                         if (err <= 0) {
3207                                 if (!nowait) {
3208                                         if (err == 0)
3209                                                 return 0;
3210                                         goto nla_put_failure;
3211                                 } else {
3212                                         if (err == -EMSGSIZE)
3213                                                 goto nla_put_failure;
3214                                 }
3215                         }
3216                 } else
3217 #endif
3218                         if (nla_put_u32(skb, RTA_IIF, iif))
3219                                 goto nla_put_failure;
3220         } else if (dst) {
3221                 struct in6_addr saddr_buf;
3222                 if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0 &&
3223                     nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
3224                         goto nla_put_failure;
3225         }
3226
3227         if (rt->rt6i_prefsrc.plen) {
3228                 struct in6_addr saddr_buf;
3229                 saddr_buf = rt->rt6i_prefsrc.addr;
3230                 if (nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
3231                         goto nla_put_failure;
3232         }
3233
3234         memcpy(metrics, dst_metrics_ptr(&rt->dst), sizeof(metrics));
3235         if (rt->rt6i_pmtu)
3236                 metrics[RTAX_MTU - 1] = rt->rt6i_pmtu;
3237         if (rtnetlink_put_metrics(skb, metrics) < 0)
3238                 goto nla_put_failure;
3239
3240         if (rt->rt6i_flags & RTF_GATEWAY) {
3241                 if (nla_put_in6_addr(skb, RTA_GATEWAY, &rt->rt6i_gateway) < 0)
3242                         goto nla_put_failure;
3243         }
3244
3245         if (rt->dst.dev &&
3246             nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex))
3247                 goto nla_put_failure;
3248         if (nla_put_u32(skb, RTA_PRIORITY, rt->rt6i_metric))
3249                 goto nla_put_failure;
3250
3251         expires = (rt->rt6i_flags & RTF_EXPIRES) ? rt->dst.expires - jiffies : 0;
3252
3253         if (rtnl_put_cacheinfo(skb, &rt->dst, 0, expires, rt->dst.error) < 0)
3254                 goto nla_put_failure;
3255
3256         if (nla_put_u8(skb, RTA_PREF, IPV6_EXTRACT_PREF(rt->rt6i_flags)))
3257                 goto nla_put_failure;
3258
3259         lwtunnel_fill_encap(skb, rt->dst.lwtstate);
3260
3261         nlmsg_end(skb, nlh);
3262         return 0;
3263
3264 nla_put_failure:
3265         nlmsg_cancel(skb, nlh);
3266         return -EMSGSIZE;
3267 }
3268
3269 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
3270 {
3271         struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
3272         int prefix;
3273
3274         if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
3275                 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
3276                 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
3277         } else
3278                 prefix = 0;
3279
3280         return rt6_fill_node(arg->net,
3281                      arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
3282                      NETLINK_CB(arg->cb->skb).portid, arg->cb->nlh->nlmsg_seq,
3283                      prefix, 0, NLM_F_MULTI);
3284 }
3285
3286 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh)
3287 {
3288         struct net *net = sock_net(in_skb->sk);
3289         struct nlattr *tb[RTA_MAX+1];
3290         struct rt6_info *rt;
3291         struct sk_buff *skb;
3292         struct rtmsg *rtm;
3293         struct flowi6 fl6;
3294         int err, iif = 0, oif = 0;
3295
3296         err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
3297         if (err < 0)
3298                 goto errout;
3299
3300         err = -EINVAL;
3301         memset(&fl6, 0, sizeof(fl6));
3302         rtm = nlmsg_data(nlh);
3303         fl6.flowlabel = ip6_make_flowinfo(rtm->rtm_tos, 0);
3304
3305         if (tb[RTA_SRC]) {
3306                 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
3307                         goto errout;
3308
3309                 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
3310         }
3311
3312         if (tb[RTA_DST]) {
3313                 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
3314                         goto errout;
3315
3316                 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
3317         }
3318
3319         if (tb[RTA_IIF])
3320                 iif = nla_get_u32(tb[RTA_IIF]);
3321
3322         if (tb[RTA_OIF])
3323                 oif = nla_get_u32(tb[RTA_OIF]);
3324
3325         if (tb[RTA_MARK])
3326                 fl6.flowi6_mark = nla_get_u32(tb[RTA_MARK]);
3327
3328         if (iif) {
3329                 struct net_device *dev;
3330                 int flags = 0;
3331
3332                 dev = __dev_get_by_index(net, iif);
3333                 if (!dev) {
3334                         err = -ENODEV;
3335                         goto errout;
3336                 }
3337
3338                 fl6.flowi6_iif = iif;
3339
3340                 if (!ipv6_addr_any(&fl6.saddr))
3341                         flags |= RT6_LOOKUP_F_HAS_SADDR;
3342
3343                 rt = (struct rt6_info *)ip6_route_input_lookup(net, dev, &fl6,
3344                                                                flags);
3345         } else {
3346                 fl6.flowi6_oif = oif;
3347
3348                 if (netif_index_is_l3_master(net, oif)) {
3349                         fl6.flowi6_flags = FLOWI_FLAG_L3MDEV_SRC |
3350                                            FLOWI_FLAG_SKIP_NH_OIF;
3351                 }
3352
3353                 rt = (struct rt6_info *)ip6_route_output(net, NULL, &fl6);
3354         }
3355
3356         skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
3357         if (!skb) {
3358                 ip6_rt_put(rt);
3359                 err = -ENOBUFS;
3360                 goto errout;
3361         }
3362
3363         /* Reserve room for dummy headers, this skb can pass
3364            through good chunk of routing engine.
3365          */
3366         skb_reset_mac_header(skb);
3367         skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
3368
3369         skb_dst_set(skb, &rt->dst);
3370
3371         err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
3372                             RTM_NEWROUTE, NETLINK_CB(in_skb).portid,
3373                             nlh->nlmsg_seq, 0, 0, 0);
3374         if (err < 0) {
3375                 kfree_skb(skb);
3376                 goto errout;
3377         }
3378
3379         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
3380 errout:
3381         return err;
3382 }
3383
3384 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info,
3385                      unsigned int nlm_flags)
3386 {
3387         struct sk_buff *skb;
3388         struct net *net = info->nl_net;
3389         u32 seq;
3390         int err;
3391
3392         err = -ENOBUFS;
3393         seq = info->nlh ? info->nlh->nlmsg_seq : 0;
3394
3395         skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
3396         if (!skb)
3397                 goto errout;
3398
3399         err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
3400                                 event, info->portid, seq, 0, 0, nlm_flags);
3401         if (err < 0) {
3402                 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
3403                 WARN_ON(err == -EMSGSIZE);
3404                 kfree_skb(skb);
3405                 goto errout;
3406         }
3407         rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
3408                     info->nlh, gfp_any());
3409         return;
3410 errout:
3411         if (err < 0)
3412                 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
3413 }
3414
3415 static int ip6_route_dev_notify(struct notifier_block *this,
3416                                 unsigned long event, void *ptr)
3417 {
3418         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3419         struct net *net = dev_net(dev);
3420
3421         if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
3422                 net->ipv6.ip6_null_entry->dst.dev = dev;
3423                 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
3424 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3425                 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
3426                 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
3427                 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
3428                 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
3429 #endif
3430         }
3431
3432         return NOTIFY_OK;
3433 }
3434
3435 /*
3436  *      /proc
3437  */
3438
3439 #ifdef CONFIG_PROC_FS
3440
3441 static const struct file_operations ipv6_route_proc_fops = {
3442         .owner          = THIS_MODULE,
3443         .open           = ipv6_route_open,
3444         .read           = seq_read,
3445         .llseek         = seq_lseek,
3446         .release        = seq_release_net,
3447 };
3448
3449 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
3450 {
3451         struct net *net = (struct net *)seq->private;
3452         seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
3453                    net->ipv6.rt6_stats->fib_nodes,
3454                    net->ipv6.rt6_stats->fib_route_nodes,
3455                    net->ipv6.rt6_stats->fib_rt_alloc,
3456                    net->ipv6.rt6_stats->fib_rt_entries,
3457                    net->ipv6.rt6_stats->fib_rt_cache,
3458                    dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
3459                    net->ipv6.rt6_stats->fib_discarded_routes);
3460
3461         return 0;
3462 }
3463
3464 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
3465 {
3466         return single_open_net(inode, file, rt6_stats_seq_show);
3467 }
3468
3469 static const struct file_operations rt6_stats_seq_fops = {
3470         .owner   = THIS_MODULE,
3471         .open    = rt6_stats_seq_open,
3472         .read    = seq_read,
3473         .llseek  = seq_lseek,
3474         .release = single_release_net,
3475 };
3476 #endif  /* CONFIG_PROC_FS */
3477
3478 #ifdef CONFIG_SYSCTL
3479
3480 static
3481 int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
3482                               void __user *buffer, size_t *lenp, loff_t *ppos)
3483 {
3484         struct net *net;
3485         int delay;
3486         if (!write)
3487                 return -EINVAL;
3488
3489         net = (struct net *)ctl->extra1;
3490         delay = net->ipv6.sysctl.flush_delay;
3491         proc_dointvec(ctl, write, buffer, lenp, ppos);
3492         fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
3493         return 0;
3494 }
3495
3496 struct ctl_table ipv6_route_table_template[] = {
3497         {
3498                 .procname       =       "flush",
3499                 .data           =       &init_net.ipv6.sysctl.flush_delay,
3500                 .maxlen         =       sizeof(int),
3501                 .mode           =       0200,
3502                 .proc_handler   =       ipv6_sysctl_rtcache_flush
3503         },
3504         {
3505                 .procname       =       "gc_thresh",
3506                 .data           =       &ip6_dst_ops_template.gc_thresh,
3507                 .maxlen         =       sizeof(int),
3508                 .mode           =       0644,
3509                 .proc_handler   =       proc_dointvec,
3510         },
3511         {
3512                 .procname       =       "max_size",
3513                 .data           =       &init_net.ipv6.sysctl.ip6_rt_max_size,
3514                 .maxlen         =       sizeof(int),
3515                 .mode           =       0644,
3516                 .proc_handler   =       proc_dointvec,
3517         },
3518         {
3519                 .procname       =       "gc_min_interval",
3520                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
3521                 .maxlen         =       sizeof(int),
3522                 .mode           =       0644,
3523                 .proc_handler   =       proc_dointvec_jiffies,
3524         },
3525         {
3526                 .procname       =       "gc_timeout",
3527                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
3528                 .maxlen         =       sizeof(int),
3529                 .mode           =       0644,
3530                 .proc_handler   =       proc_dointvec_jiffies,
3531         },
3532         {
3533                 .procname       =       "gc_interval",
3534                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_interval,
3535                 .maxlen         =       sizeof(int),
3536                 .mode           =       0644,
3537                 .proc_handler   =       proc_dointvec_jiffies,
3538         },
3539         {
3540                 .procname       =       "gc_elasticity",
3541                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
3542                 .maxlen         =       sizeof(int),
3543                 .mode           =       0644,
3544                 .proc_handler   =       proc_dointvec,
3545         },
3546         {
3547                 .procname       =       "mtu_expires",
3548                 .data           =       &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
3549                 .maxlen         =       sizeof(int),
3550                 .mode           =       0644,
3551                 .proc_handler   =       proc_dointvec_jiffies,
3552         },
3553         {
3554                 .procname       =       "min_adv_mss",
3555                 .data           =       &init_net.ipv6.sysctl.ip6_rt_min_advmss,
3556                 .maxlen         =       sizeof(int),
3557                 .mode           =       0644,
3558                 .proc_handler   =       proc_dointvec,
3559         },
3560         {
3561                 .procname       =       "gc_min_interval_ms",
3562                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
3563                 .maxlen         =       sizeof(int),
3564                 .mode           =       0644,
3565                 .proc_handler   =       proc_dointvec_ms_jiffies,
3566         },
3567         { }
3568 };
3569
3570 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
3571 {
3572         struct ctl_table *table;
3573
3574         table = kmemdup(ipv6_route_table_template,
3575                         sizeof(ipv6_route_table_template),
3576                         GFP_KERNEL);
3577
3578         if (table) {
3579                 table[0].data = &net->ipv6.sysctl.flush_delay;
3580                 table[0].extra1 = net;
3581                 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
3582                 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
3583                 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
3584                 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
3585                 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
3586                 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
3587                 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
3588                 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
3589                 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
3590
3591                 /* Don't export sysctls to unprivileged users */
3592                 if (net->user_ns != &init_user_ns)
3593                         table[0].procname = NULL;
3594         }
3595
3596         return table;
3597 }
3598 #endif
3599
3600 static int __net_init ip6_route_net_init(struct net *net)
3601 {
3602         int ret = -ENOMEM;
3603
3604         memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
3605                sizeof(net->ipv6.ip6_dst_ops));
3606
3607         if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
3608                 goto out_ip6_dst_ops;
3609
3610         net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
3611                                            sizeof(*net->ipv6.ip6_null_entry),
3612                                            GFP_KERNEL);
3613         if (!net->ipv6.ip6_null_entry)
3614                 goto out_ip6_dst_entries;
3615         net->ipv6.ip6_null_entry->dst.path =
3616                 (struct dst_entry *)net->ipv6.ip6_null_entry;
3617         net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3618         dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
3619                          ip6_template_metrics, true);
3620
3621 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3622         net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
3623                                                sizeof(*net->ipv6.ip6_prohibit_entry),
3624                                                GFP_KERNEL);
3625         if (!net->ipv6.ip6_prohibit_entry)
3626                 goto out_ip6_null_entry;
3627         net->ipv6.ip6_prohibit_entry->dst.path =
3628                 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
3629         net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3630         dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
3631                          ip6_template_metrics, true);
3632
3633         net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
3634                                                sizeof(*net->ipv6.ip6_blk_hole_entry),
3635                                                GFP_KERNEL);
3636         if (!net->ipv6.ip6_blk_hole_entry)
3637                 goto out_ip6_prohibit_entry;
3638         net->ipv6.ip6_blk_hole_entry->dst.path =
3639                 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
3640         net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3641         dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
3642                          ip6_template_metrics, true);
3643 #endif
3644
3645         net->ipv6.sysctl.flush_delay = 0;
3646         net->ipv6.sysctl.ip6_rt_max_size = 4096;
3647         net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
3648         net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
3649         net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
3650         net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
3651         net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
3652         net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
3653
3654         net->ipv6.ip6_rt_gc_expire = 30*HZ;
3655
3656         ret = 0;
3657 out:
3658         return ret;
3659
3660 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3661 out_ip6_prohibit_entry:
3662         kfree(net->ipv6.ip6_prohibit_entry);
3663 out_ip6_null_entry:
3664         kfree(net->ipv6.ip6_null_entry);
3665 #endif
3666 out_ip6_dst_entries:
3667         dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3668 out_ip6_dst_ops:
3669         goto out;
3670 }
3671
3672 static void __net_exit ip6_route_net_exit(struct net *net)
3673 {
3674         kfree(net->ipv6.ip6_null_entry);
3675 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3676         kfree(net->ipv6.ip6_prohibit_entry);
3677         kfree(net->ipv6.ip6_blk_hole_entry);
3678 #endif
3679         dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3680 }
3681
3682 static int __net_init ip6_route_net_init_late(struct net *net)
3683 {
3684 #ifdef CONFIG_PROC_FS
3685         proc_create("ipv6_route", 0, net->proc_net, &ipv6_route_proc_fops);
3686         proc_create("rt6_stats", S_IRUGO, net->proc_net, &rt6_stats_seq_fops);
3687 #endif
3688         return 0;
3689 }
3690
3691 static void __net_exit ip6_route_net_exit_late(struct net *net)
3692 {
3693 #ifdef CONFIG_PROC_FS
3694         remove_proc_entry("ipv6_route", net->proc_net);
3695         remove_proc_entry("rt6_stats", net->proc_net);
3696 #endif
3697 }
3698
3699 static struct pernet_operations ip6_route_net_ops = {
3700         .init = ip6_route_net_init,
3701         .exit = ip6_route_net_exit,
3702 };
3703
3704 static int __net_init ipv6_inetpeer_init(struct net *net)
3705 {
3706         struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
3707
3708         if (!bp)
3709                 return -ENOMEM;
3710         inet_peer_base_init(bp);
3711         net->ipv6.peers = bp;
3712         return 0;
3713 }
3714
3715 static void __net_exit ipv6_inetpeer_exit(struct net *net)
3716 {
3717         struct inet_peer_base *bp = net->ipv6.peers;
3718
3719         net->ipv6.peers = NULL;
3720         inetpeer_invalidate_tree(bp);
3721         kfree(bp);
3722 }
3723
3724 static struct pernet_operations ipv6_inetpeer_ops = {
3725         .init   =       ipv6_inetpeer_init,
3726         .exit   =       ipv6_inetpeer_exit,
3727 };
3728
3729 static struct pernet_operations ip6_route_net_late_ops = {
3730         .init = ip6_route_net_init_late,
3731         .exit = ip6_route_net_exit_late,
3732 };
3733
3734 static struct notifier_block ip6_route_dev_notifier = {
3735         .notifier_call = ip6_route_dev_notify,
3736         .priority = 0,
3737 };
3738
3739 int __init ip6_route_init(void)
3740 {
3741         int ret;
3742         int cpu;
3743
3744         ret = -ENOMEM;
3745         ip6_dst_ops_template.kmem_cachep =
3746                 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
3747                                   SLAB_HWCACHE_ALIGN, NULL);
3748         if (!ip6_dst_ops_template.kmem_cachep)
3749                 goto out;
3750
3751         ret = dst_entries_init(&ip6_dst_blackhole_ops);
3752         if (ret)
3753                 goto out_kmem_cache;
3754
3755         ret = register_pernet_subsys(&ipv6_inetpeer_ops);
3756         if (ret)
3757                 goto out_dst_entries;
3758
3759         ret = register_pernet_subsys(&ip6_route_net_ops);
3760         if (ret)
3761                 goto out_register_inetpeer;
3762
3763         ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
3764
3765         /* Registering of the loopback is done before this portion of code,
3766          * the loopback reference in rt6_info will not be taken, do it
3767          * manually for init_net */
3768         init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
3769         init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3770   #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3771         init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
3772         init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3773         init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
3774         init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3775   #endif
3776         ret = fib6_init();
3777         if (ret)
3778                 goto out_register_subsys;
3779
3780         ret = xfrm6_init();
3781         if (ret)
3782                 goto out_fib6_init;
3783
3784         ret = fib6_rules_init();
3785         if (ret)
3786                 goto xfrm6_init;
3787
3788         ret = register_pernet_subsys(&ip6_route_net_late_ops);
3789         if (ret)
3790                 goto fib6_rules_init;
3791
3792         ret = -ENOBUFS;
3793         if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL, NULL) ||
3794             __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL, NULL) ||
3795             __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL, NULL))
3796                 goto out_register_late_subsys;
3797
3798         ret = register_netdevice_notifier(&ip6_route_dev_notifier);
3799         if (ret)
3800                 goto out_register_late_subsys;
3801
3802         for_each_possible_cpu(cpu) {
3803                 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
3804
3805                 INIT_LIST_HEAD(&ul->head);
3806                 spin_lock_init(&ul->lock);
3807         }
3808
3809 out:
3810         return ret;
3811
3812 out_register_late_subsys:
3813         unregister_pernet_subsys(&ip6_route_net_late_ops);
3814 fib6_rules_init:
3815         fib6_rules_cleanup();
3816 xfrm6_init:
3817         xfrm6_fini();
3818 out_fib6_init:
3819         fib6_gc_cleanup();
3820 out_register_subsys:
3821         unregister_pernet_subsys(&ip6_route_net_ops);
3822 out_register_inetpeer:
3823         unregister_pernet_subsys(&ipv6_inetpeer_ops);
3824 out_dst_entries:
3825         dst_entries_destroy(&ip6_dst_blackhole_ops);
3826 out_kmem_cache:
3827         kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3828         goto out;
3829 }
3830
3831 void ip6_route_cleanup(void)
3832 {
3833         unregister_netdevice_notifier(&ip6_route_dev_notifier);
3834         unregister_pernet_subsys(&ip6_route_net_late_ops);
3835         fib6_rules_cleanup();
3836         xfrm6_fini();
3837         fib6_gc_cleanup();
3838         unregister_pernet_subsys(&ipv6_inetpeer_ops);
3839         unregister_pernet_subsys(&ip6_route_net_ops);
3840         dst_entries_destroy(&ip6_dst_blackhole_ops);
3841         kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3842 }