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 <linux/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,
 106                                            struct net_device *dev,
 107                                            unsigned int pref);
 108 static struct rt6_info *rt6_get_route_info(struct net *net,
 109                                            const struct in6_addr *prefix, int prefixlen,
 110                                            const struct in6_addr *gwaddr,
 111                                            struct net_device *dev);
 112 #endif
 113
 114 struct uncached_list {
 115         spinlock_t              lock;
 116         struct list_head        head;
 117 };
 118
 119 static DEFINE_PER_CPU_ALIGNED(struct uncached_list, rt6_uncached_list);
 120
 121 static void rt6_uncached_list_add(struct rt6_info *rt)
 122 {
 123         struct uncached_list *ul = raw_cpu_ptr(&rt6_uncached_list);
 124
 125         rt->dst.flags |= DST_NOCACHE;
 126         rt->rt6i_uncached_list = ul;
 127
 128         spin_lock_bh(&ul->lock);
 129         list_add_tail(&rt->rt6i_uncached, &ul->head);
 130         spin_unlock_bh(&ul->lock);
 131 }
 132
 133 static void rt6_uncached_list_del(struct rt6_info *rt)
 134 {
 135         if (!list_empty(&rt->rt6i_uncached)) {
 136                 struct uncached_list *ul = rt->rt6i_uncached_list;
 137
 138                 spin_lock_bh(&ul->lock);
 139                 list_del(&rt->rt6i_uncached);
 140                 spin_unlock_bh(&ul->lock);
 141         }
 142 }
 143
 144 static void rt6_uncached_list_flush_dev(struct net *net, struct net_device *dev)
 145 {
 146         struct net_device *loopback_dev = net->loopback_dev;
 147         int cpu;
 148
 149         if (dev == loopback_dev)
 150                 return;
 151
 152         for_each_possible_cpu(cpu) {
 153                 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
 154                 struct rt6_info *rt;
 155
 156                 spin_lock_bh(&ul->lock);
 157                 list_for_each_entry(rt, &ul->head, rt6i_uncached) {
 158                         struct inet6_dev *rt_idev = rt->rt6i_idev;
 159                         struct net_device *rt_dev = rt->dst.dev;
 160
 161                         if (rt_idev->dev == dev) {
 162                                 rt->rt6i_idev = in6_dev_get(loopback_dev);
 163                                 in6_dev_put(rt_idev);
 164                         }
 165
 166                         if (rt_dev == dev) {
 167                                 rt->dst.dev = loopback_dev;
 168                                 dev_hold(rt->dst.dev);
 169                                 dev_put(rt_dev);
 170                         }
 171                 }
 172                 spin_unlock_bh(&ul->lock);
 173         }
 174 }
 175
 176 static u32 *rt6_pcpu_cow_metrics(struct rt6_info *rt)
 177 {
 178         return dst_metrics_write_ptr(rt->dst.from);
 179 }
 180
 181 static u32 *ipv6_cow_metrics(struct dst_entry *dst, unsigned long old)
 182 {
 183         struct rt6_info *rt = (struct rt6_info *)dst;
 184
 185         if (rt->rt6i_flags & RTF_PCPU)
 186                 return rt6_pcpu_cow_metrics(rt);
 187         else if (rt->rt6i_flags & RTF_CACHE)
 188                 return NULL;
 189         else
 190                 return dst_cow_metrics_generic(dst, old);
 191 }
 192
 193 static inline const void *choose_neigh_daddr(struct rt6_info *rt,
 194                                              struct sk_buff *skb,
 195                                              const void *daddr)
 196 {
 197         struct in6_addr *p = &rt->rt6i_gateway;
 198
 199         if (!ipv6_addr_any(p))
 200                 return (const void *) p;
 201         else if (skb)
 202                 return &ipv6_hdr(skb)->daddr;
 203         return daddr;
 204 }
 205
 206 static struct neighbour *ip6_neigh_lookup(const struct dst_entry *dst,
 207                                           struct sk_buff *skb,
 208                                           const void *daddr)
 209 {
 210         struct rt6_info *rt = (struct rt6_info *) dst;
 211         struct neighbour *n;
 212
 213         daddr = choose_neigh_daddr(rt, skb, daddr);
 214         n = __ipv6_neigh_lookup(dst->dev, daddr);
 215         if (n)
 216                 return n;
 217         return neigh_create(&nd_tbl, daddr, dst->dev);
 218 }
 219
 220 static struct dst_ops ip6_dst_ops_template = {
 221         .family                 =       AF_INET6,
 222         .gc                     =       ip6_dst_gc,
 223         .gc_thresh              =       1024,
 224         .check                  =       ip6_dst_check,
 225         .default_advmss         =       ip6_default_advmss,
 226         .mtu                    =       ip6_mtu,
 227         .cow_metrics            =       ipv6_cow_metrics,
 228         .destroy                =       ip6_dst_destroy,
 229         .ifdown                 =       ip6_dst_ifdown,
 230         .negative_advice        =       ip6_negative_advice,
 231         .link_failure           =       ip6_link_failure,
 232         .update_pmtu            =       ip6_rt_update_pmtu,
 233         .redirect               =       rt6_do_redirect,
 234         .local_out              =       __ip6_local_out,
 235         .neigh_lookup           =       ip6_neigh_lookup,
 236 };
 237
 238 static unsigned int ip6_blackhole_mtu(const struct dst_entry *dst)
 239 {
 240         unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
 241
 242         return mtu ? : dst->dev->mtu;
 243 }
 244
 245 static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, struct sock *sk,
 246                                          struct sk_buff *skb, u32 mtu)
 247 {
 248 }
 249
 250 static void ip6_rt_blackhole_redirect(struct dst_entry *dst, struct sock *sk,
 251                                       struct sk_buff *skb)
 252 {
 253 }
 254
 255 static struct dst_ops ip6_dst_blackhole_ops = {
 256         .family                 =       AF_INET6,
 257         .destroy                =       ip6_dst_destroy,
 258         .check                  =       ip6_dst_check,
 259         .mtu                    =       ip6_blackhole_mtu,
 260         .default_advmss         =       ip6_default_advmss,
 261         .update_pmtu            =       ip6_rt_blackhole_update_pmtu,
 262         .redirect               =       ip6_rt_blackhole_redirect,
 263         .cow_metrics            =       dst_cow_metrics_generic,
 264         .neigh_lookup           =       ip6_neigh_lookup,
 265 };
 266
 267 static const u32 ip6_template_metrics[RTAX_MAX] = {
 268         [RTAX_HOPLIMIT - 1] = 0,
 269 };
 270
 271 static const struct rt6_info ip6_null_entry_template = {
 272         .dst = {
 273                 .__refcnt       = ATOMIC_INIT(1),
 274                 .__use          = 1,
 275                 .obsolete       = DST_OBSOLETE_FORCE_CHK,
 276                 .error          = -ENETUNREACH,
 277                 .input          = ip6_pkt_discard,
 278                 .output         = ip6_pkt_discard_out,
 279         },
 280         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
 281         .rt6i_protocol  = RTPROT_KERNEL,
 282         .rt6i_metric    = ~(u32) 0,
 283         .rt6i_ref       = ATOMIC_INIT(1),
 284 };
 285
 286 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
 287
 288 static const struct rt6_info ip6_prohibit_entry_template = {
 289         .dst = {
 290                 .__refcnt       = ATOMIC_INIT(1),
 291                 .__use          = 1,
 292                 .obsolete       = DST_OBSOLETE_FORCE_CHK,
 293                 .error          = -EACCES,
 294                 .input          = ip6_pkt_prohibit,
 295                 .output         = ip6_pkt_prohibit_out,
 296         },
 297         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
 298         .rt6i_protocol  = RTPROT_KERNEL,
 299         .rt6i_metric    = ~(u32) 0,
 300         .rt6i_ref       = ATOMIC_INIT(1),
 301 };
 302
 303 static const struct rt6_info ip6_blk_hole_entry_template = {
 304         .dst = {
 305                 .__refcnt       = ATOMIC_INIT(1),
 306                 .__use          = 1,
 307                 .obsolete       = DST_OBSOLETE_FORCE_CHK,
 308                 .error          = -EINVAL,
 309                 .input          = dst_discard,
 310                 .output         = dst_discard_out,
 311         },
 312         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
 313         .rt6i_protocol  = RTPROT_KERNEL,
 314         .rt6i_metric    = ~(u32) 0,
 315         .rt6i_ref       = ATOMIC_INIT(1),
 316 };
 317
 318 #endif
 319
 320 static void rt6_info_init(struct rt6_info *rt)
 321 {
 322         struct dst_entry *dst = &rt->dst;
 323
 324         memset(dst + 1, 0, sizeof(*rt) - sizeof(*dst));
 325         INIT_LIST_HEAD(&rt->rt6i_siblings);
 326         INIT_LIST_HEAD(&rt->rt6i_uncached);
 327 }
 328
 329 /* allocate dst with ip6_dst_ops */
 330 static struct rt6_info *__ip6_dst_alloc(struct net *net,
 331                                         struct net_device *dev,
 332                                         int flags)
 333 {
 334         struct rt6_info *rt = dst_alloc(&net->ipv6.ip6_dst_ops, dev,
 335                                         0, DST_OBSOLETE_FORCE_CHK, flags);
 336
 337         if (rt)
 338                 rt6_info_init(rt);
 339
 340         return rt;
 341 }
 342
 343 struct rt6_info *ip6_dst_alloc(struct net *net,
 344                                struct net_device *dev,
 345                                int flags)
 346 {
 347         struct rt6_info *rt = __ip6_dst_alloc(net, dev, flags);
 348
 349         if (rt) {
 350                 rt->rt6i_pcpu = alloc_percpu_gfp(struct rt6_info *, GFP_ATOMIC);
 351                 if (rt->rt6i_pcpu) {
 352                         int cpu;
 353
 354                         for_each_possible_cpu(cpu) {
 355                                 struct rt6_info **p;
 356
 357                                 p = per_cpu_ptr(rt->rt6i_pcpu, cpu);
 358                                 /* no one shares rt */
 359                                 *p =  NULL;
 360                         }
 361                 } else {
 362                         dst_destroy((struct dst_entry *)rt);
 363                         return NULL;
 364                 }
 365         }
 366
 367         return rt;
 368 }
 369 EXPORT_SYMBOL(ip6_dst_alloc);
 370
 371 static void ip6_dst_destroy(struct dst_entry *dst)
 372 {
 373         struct rt6_info *rt = (struct rt6_info *)dst;
 374         struct dst_entry *from = dst->from;
 375         struct inet6_dev *idev;
 376
 377         dst_destroy_metrics_generic(dst);
 378         free_percpu(rt->rt6i_pcpu);
 379         rt6_uncached_list_del(rt);
 380
 381         idev = rt->rt6i_idev;
 382         if (idev) {
 383                 rt->rt6i_idev = NULL;
 384                 in6_dev_put(idev);
 385         }
 386
 387         dst->from = NULL;
 388         dst_release(from);
 389 }
 390
 391 static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
 392                            int how)
 393 {
 394         struct rt6_info *rt = (struct rt6_info *)dst;
 395         struct inet6_dev *idev = rt->rt6i_idev;
 396         struct net_device *loopback_dev =
 397                 dev_net(dev)->loopback_dev;
 398
 399         if (dev != loopback_dev) {
 400                 if (idev && idev->dev == dev) {
 401                         struct inet6_dev *loopback_idev =
 402                                 in6_dev_get(loopback_dev);
 403                         if (loopback_idev) {
 404                                 rt->rt6i_idev = loopback_idev;
 405                                 in6_dev_put(idev);
 406                         }
 407                 }
 408         }
 409 }
 410
 411 static bool __rt6_check_expired(const struct rt6_info *rt)
 412 {
 413         if (rt->rt6i_flags & RTF_EXPIRES)
 414                 return time_after(jiffies, rt->dst.expires);
 415         else
 416                 return false;
 417 }
 418
 419 static bool rt6_check_expired(const struct rt6_info *rt)
 420 {
 421         if (rt->rt6i_flags & RTF_EXPIRES) {
 422                 if (time_after(jiffies, rt->dst.expires))
 423                         return true;
 424         } else if (rt->dst.from) {
 425                 return rt6_check_expired((struct rt6_info *) rt->dst.from);
 426         }
 427         return false;
 428 }
 429
 430 /* Multipath route selection:
 431  *   Hash based function using packet header and flowlabel.
 432  * Adapted from fib_info_hashfn()
 433  */
 434 static int rt6_info_hash_nhsfn(unsigned int candidate_count,
 435                                const struct flowi6 *fl6)
 436 {
 437         return get_hash_from_flowi6(fl6) % candidate_count;
 438 }
 439
 440 static struct rt6_info *rt6_multipath_select(struct rt6_info *match,
 441                                              struct flowi6 *fl6, int oif,
 442                                              int strict)
 443 {
 444         struct rt6_info *sibling, *next_sibling;
 445         int route_choosen;
 446
 447         route_choosen = rt6_info_hash_nhsfn(match->rt6i_nsiblings + 1, fl6);
 448         /* Don't change the route, if route_choosen == 0
 449          * (siblings does not include ourself)
 450          */
 451         if (route_choosen)
 452                 list_for_each_entry_safe(sibling, next_sibling,
 453                                 &match->rt6i_siblings, rt6i_siblings) {
 454                         route_choosen--;
 455                         if (route_choosen == 0) {
 456                                 if (rt6_score_route(sibling, oif, strict) < 0)
 457                                         break;
 458                                 match = sibling;
 459                                 break;
 460                         }
 461                 }
 462         return match;
 463 }
 464
 465 /*
 466  *      Route lookup. Any table->tb6_lock is implied.
 467  */
 468
 469 static inline struct rt6_info *rt6_device_match(struct net *net,
 470                                                     struct rt6_info *rt,
 471                                                     const struct in6_addr *saddr,
 472                                                     int oif,
 473                                                     int flags)
 474 {
 475         struct rt6_info *local = NULL;
 476         struct rt6_info *sprt;
 477
 478         if (!oif && ipv6_addr_any(saddr))
 479                 goto out;
 480
 481         for (sprt = rt; sprt; sprt = sprt->dst.rt6_next) {
 482                 struct net_device *dev = sprt->dst.dev;
 483
 484                 if (oif) {
 485                         if (dev->ifindex == oif)
 486                                 return sprt;
 487                         if (dev->flags & IFF_LOOPBACK) {
 488                                 if (!sprt->rt6i_idev ||
 489                                     sprt->rt6i_idev->dev->ifindex != oif) {
 490                                         if (flags & RT6_LOOKUP_F_IFACE)
 491                                                 continue;
 492                                         if (local &&
 493                                             local->rt6i_idev->dev->ifindex == oif)
 494                                                 continue;
 495                                 }
 496                                 local = sprt;
 497                         }
 498                 } else {
 499                         if (ipv6_chk_addr(net, saddr, dev,
 500                                           flags & RT6_LOOKUP_F_IFACE))
 501                                 return sprt;
 502                 }
 503         }
 504
 505         if (oif) {
 506                 if (local)
 507                         return local;
 508
 509                 if (flags & RT6_LOOKUP_F_IFACE)
 510                         return net->ipv6.ip6_null_entry;
 511         }
 512 out:
 513         return rt;
 514 }
 515
 516 #ifdef CONFIG_IPV6_ROUTER_PREF
 517 struct __rt6_probe_work {
 518         struct work_struct work;
 519         struct in6_addr target;
 520         struct net_device *dev;
 521 };
 522
 523 static void rt6_probe_deferred(struct work_struct *w)
 524 {
 525         struct in6_addr mcaddr;
 526         struct __rt6_probe_work *work =
 527                 container_of(w, struct __rt6_probe_work, work);
 528
 529         addrconf_addr_solict_mult(&work->target, &mcaddr);
 530         ndisc_send_ns(work->dev, &work->target, &mcaddr, NULL, 0);
 531         dev_put(work->dev);
 532         kfree(work);
 533 }
 534
 535 static void rt6_probe(struct rt6_info *rt)
 536 {
 537         struct __rt6_probe_work *work;
 538         struct neighbour *neigh;
 539         /*
 540          * Okay, this does not seem to be appropriate
 541          * for now, however, we need to check if it
 542          * is really so; aka Router Reachability Probing.
 543          *
 544          * Router Reachability Probe MUST be rate-limited
 545          * to no more than one per minute.
 546          */
 547         if (!rt || !(rt->rt6i_flags & RTF_GATEWAY))
 548                 return;
 549         rcu_read_lock_bh();
 550         neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
 551         if (neigh) {
 552                 if (neigh->nud_state & NUD_VALID)
 553                         goto out;
 554
 555                 work = NULL;
 556                 write_lock(&neigh->lock);
 557                 if (!(neigh->nud_state & NUD_VALID) &&
 558                     time_after(jiffies,
 559                                neigh->updated +
 560                                rt->rt6i_idev->cnf.rtr_probe_interval)) {
 561                         work = kmalloc(sizeof(*work), GFP_ATOMIC);
 562                         if (work)
 563                                 __neigh_set_probe_once(neigh);
 564                 }
 565                 write_unlock(&neigh->lock);
 566         } else {
 567                 work = kmalloc(sizeof(*work), GFP_ATOMIC);
 568         }
 569
 570         if (work) {
 571                 INIT_WORK(&work->work, rt6_probe_deferred);
 572                 work->target = rt->rt6i_gateway;
 573                 dev_hold(rt->dst.dev);
 574                 work->dev = rt->dst.dev;
 575                 schedule_work(&work->work);
 576         }
 577
 578 out:
 579         rcu_read_unlock_bh();
 580 }
 581 #else
 582 static inline void rt6_probe(struct rt6_info *rt)
 583 {
 584 }
 585 #endif
 586
 587 /*
 588  * Default Router Selection (RFC 2461 6.3.6)
 589  */
 590 static inline int rt6_check_dev(struct rt6_info *rt, int oif)
 591 {
 592         struct net_device *dev = rt->dst.dev;
 593         if (!oif || dev->ifindex == oif)
 594                 return 2;
 595         if ((dev->flags & IFF_LOOPBACK) &&
 596             rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
 597                 return 1;
 598         return 0;
 599 }
 600
 601 static inline enum rt6_nud_state rt6_check_neigh(struct rt6_info *rt)
 602 {
 603         struct neighbour *neigh;
 604         enum rt6_nud_state ret = RT6_NUD_FAIL_HARD;
 605
 606         if (rt->rt6i_flags & RTF_NONEXTHOP ||
 607             !(rt->rt6i_flags & RTF_GATEWAY))
 608                 return RT6_NUD_SUCCEED;
 609
 610         rcu_read_lock_bh();
 611         neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
 612         if (neigh) {
 613                 read_lock(&neigh->lock);
 614                 if (neigh->nud_state & NUD_VALID)
 615                         ret = RT6_NUD_SUCCEED;
 616 #ifdef CONFIG_IPV6_ROUTER_PREF
 617                 else if (!(neigh->nud_state & NUD_FAILED))
 618                         ret = RT6_NUD_SUCCEED;
 619                 else
 620                         ret = RT6_NUD_FAIL_PROBE;
 621 #endif
 622                 read_unlock(&neigh->lock);
 623         } else {
 624                 ret = IS_ENABLED(CONFIG_IPV6_ROUTER_PREF) ?
 625                       RT6_NUD_SUCCEED : RT6_NUD_FAIL_DO_RR;
 626         }
 627         rcu_read_unlock_bh();
 628
 629         return ret;
 630 }
 631
 632 static int rt6_score_route(struct rt6_info *rt, int oif,
 633                            int strict)
 634 {
 635         int m;
 636
 637         m = rt6_check_dev(rt, oif);
 638         if (!m && (strict & RT6_LOOKUP_F_IFACE))
 639                 return RT6_NUD_FAIL_HARD;
 640 #ifdef CONFIG_IPV6_ROUTER_PREF
 641         m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
 642 #endif
 643         if (strict & RT6_LOOKUP_F_REACHABLE) {
 644                 int n = rt6_check_neigh(rt);
 645                 if (n < 0)
 646                         return n;
 647         }
 648         return m;
 649 }
 650
 651 static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
 652                                    int *mpri, struct rt6_info *match,
 653                                    bool *do_rr)
 654 {
 655         int m;
 656         bool match_do_rr = false;
 657         struct inet6_dev *idev = rt->rt6i_idev;
 658         struct net_device *dev = rt->dst.dev;
 659
 660         if (dev && !netif_carrier_ok(dev) &&
 661             idev->cnf.ignore_routes_with_linkdown &&
 662             !(strict & RT6_LOOKUP_F_IGNORE_LINKSTATE))
 663                 goto out;
 664
 665         if (rt6_check_expired(rt))
 666                 goto out;
 667
 668         m = rt6_score_route(rt, oif, strict);
 669         if (m == RT6_NUD_FAIL_DO_RR) {
 670                 match_do_rr = true;
 671                 m = 0; /* lowest valid score */
 672         } else if (m == RT6_NUD_FAIL_HARD) {
 673                 goto out;
 674         }
 675
 676         if (strict & RT6_LOOKUP_F_REACHABLE)
 677                 rt6_probe(rt);
 678
 679         /* note that m can be RT6_NUD_FAIL_PROBE at this point */
 680         if (m > *mpri) {
 681                 *do_rr = match_do_rr;
 682                 *mpri = m;
 683                 match = rt;
 684         }
 685 out:
 686         return match;
 687 }
 688
 689 static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
 690                                      struct rt6_info *rr_head,
 691                                      u32 metric, int oif, int strict,
 692                                      bool *do_rr)
 693 {
 694         struct rt6_info *rt, *match, *cont;
 695         int mpri = -1;
 696
 697         match = NULL;
 698         cont = NULL;
 699         for (rt = rr_head; rt; rt = rt->dst.rt6_next) {
 700                 if (rt->rt6i_metric != metric) {
 701                         cont = rt;
 702                         break;
 703                 }
 704
 705                 match = find_match(rt, oif, strict, &mpri, match, do_rr);
 706         }
 707
 708         for (rt = fn->leaf; rt && rt != rr_head; rt = rt->dst.rt6_next) {
 709                 if (rt->rt6i_metric != metric) {
 710                         cont = rt;
 711                         break;
 712                 }
 713
 714                 match = find_match(rt, oif, strict, &mpri, match, do_rr);
 715         }
 716
 717         if (match || !cont)
 718                 return match;
 719
 720         for (rt = cont; rt; rt = rt->dst.rt6_next)
 721                 match = find_match(rt, oif, strict, &mpri, match, do_rr);
 722
 723         return match;
 724 }
 725
 726 static struct rt6_info *rt6_select(struct fib6_node *fn, int oif, int strict)
 727 {
 728         struct rt6_info *match, *rt0;
 729         struct net *net;
 730         bool do_rr = false;
 731
 732         rt0 = fn->rr_ptr;
 733         if (!rt0)
 734                 fn->rr_ptr = rt0 = fn->leaf;
 735
 736         match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict,
 737                              &do_rr);
 738
 739         if (do_rr) {
 740                 struct rt6_info *next = rt0->dst.rt6_next;
 741
 742                 /* no entries matched; do round-robin */
 743                 if (!next || next->rt6i_metric != rt0->rt6i_metric)
 744                         next = fn->leaf;
 745
 746                 if (next != rt0)
 747                         fn->rr_ptr = next;
 748         }
 749
 750         net = dev_net(rt0->dst.dev);
 751         return match ? match : net->ipv6.ip6_null_entry;
 752 }
 753
 754 static bool rt6_is_gw_or_nonexthop(const struct rt6_info *rt)
 755 {
 756         return (rt->rt6i_flags & (RTF_NONEXTHOP | RTF_GATEWAY));
 757 }
 758
 759 #ifdef CONFIG_IPV6_ROUTE_INFO
 760 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
 761                   const struct in6_addr *gwaddr)
 762 {
 763         struct net *net = dev_net(dev);
 764         struct route_info *rinfo = (struct route_info *) opt;
 765         struct in6_addr prefix_buf, *prefix;
 766         unsigned int pref;
 767         unsigned long lifetime;
 768         struct rt6_info *rt;
 769
 770         if (len < sizeof(struct route_info)) {
 771                 return -EINVAL;
 772         }
 773
 774         /* Sanity check for prefix_len and length */
 775         if (rinfo->length > 3) {
 776                 return -EINVAL;
 777         } else if (rinfo->prefix_len > 128) {
 778                 return -EINVAL;
 779         } else if (rinfo->prefix_len > 64) {
 780                 if (rinfo->length < 2) {
 781                         return -EINVAL;
 782                 }
 783         } else if (rinfo->prefix_len > 0) {
 784                 if (rinfo->length < 1) {
 785                         return -EINVAL;
 786                 }
 787         }
 788
 789         pref = rinfo->route_pref;
 790         if (pref == ICMPV6_ROUTER_PREF_INVALID)
 791                 return -EINVAL;
 792
 793         lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
 794
 795         if (rinfo->length == 3)
 796                 prefix = (struct in6_addr *)rinfo->prefix;
 797         else {
 798                 /* this function is safe */
 799                 ipv6_addr_prefix(&prefix_buf,
 800                                  (struct in6_addr *)rinfo->prefix,
 801                                  rinfo->prefix_len);
 802                 prefix = &prefix_buf;
 803         }
 804
 805         if (rinfo->prefix_len == 0)
 806                 rt = rt6_get_dflt_router(gwaddr, dev);
 807         else
 808                 rt = rt6_get_route_info(net, prefix, rinfo->prefix_len,
 809                                         gwaddr, dev);
 810
 811         if (rt && !lifetime) {
 812                 ip6_del_rt(rt);
 813                 rt = NULL;
 814         }
 815
 816         if (!rt && lifetime)
 817                 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr,
 818                                         dev, pref);
 819         else if (rt)
 820                 rt->rt6i_flags = RTF_ROUTEINFO |
 821                                  (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
 822
 823         if (rt) {
 824                 if (!addrconf_finite_timeout(lifetime))
 825                         rt6_clean_expires(rt);
 826                 else
 827                         rt6_set_expires(rt, jiffies + HZ * lifetime);
 828
 829                 ip6_rt_put(rt);
 830         }
 831         return 0;
 832 }
 833 #endif
 834
 835 static struct fib6_node* fib6_backtrack(struct fib6_node *fn,
 836                                         struct in6_addr *saddr)
 837 {
 838         struct fib6_node *pn;
 839         while (1) {
 840                 if (fn->fn_flags & RTN_TL_ROOT)
 841                         return NULL;
 842                 pn = fn->parent;
 843                 if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn)
 844                         fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr);
 845                 else
 846                         fn = pn;
 847                 if (fn->fn_flags & RTN_RTINFO)
 848                         return fn;
 849         }
 850 }
 851
 852 static struct rt6_info *ip6_pol_route_lookup(struct net *net,
 853                                              struct fib6_table *table,
 854                                              struct flowi6 *fl6, int flags)
 855 {
 856         struct fib6_node *fn;
 857         struct rt6_info *rt;
 858
 859         read_lock_bh(&table->tb6_lock);
 860         fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
 861 restart:
 862         rt = fn->leaf;
 863         rt = rt6_device_match(net, rt, &fl6->saddr, fl6->flowi6_oif, flags);
 864         if (rt->rt6i_nsiblings && fl6->flowi6_oif == 0)
 865                 rt = rt6_multipath_select(rt, fl6, fl6->flowi6_oif, flags);
 866         if (rt == net->ipv6.ip6_null_entry) {
 867                 fn = fib6_backtrack(fn, &fl6->saddr);
 868                 if (fn)
 869                         goto restart;
 870         }
 871         dst_use(&rt->dst, jiffies);
 872         read_unlock_bh(&table->tb6_lock);
 873
 874         trace_fib6_table_lookup(net, rt, table->tb6_id, fl6);
 875
 876         return rt;
 877
 878 }
 879
 880 struct dst_entry *ip6_route_lookup(struct net *net, struct flowi6 *fl6,
 881                                     int flags)
 882 {
 883         return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_lookup);
 884 }
 885 EXPORT_SYMBOL_GPL(ip6_route_lookup);
 886
 887 struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
 888                             const struct in6_addr *saddr, int oif, int strict)
 889 {
 890         struct flowi6 fl6 = {
 891                 .flowi6_oif = oif,
 892                 .daddr = *daddr,
 893         };
 894         struct dst_entry *dst;
 895         int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
 896
 897         if (saddr) {
 898                 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
 899                 flags |= RT6_LOOKUP_F_HAS_SADDR;
 900         }
 901
 902         dst = fib6_rule_lookup(net, &fl6, flags, ip6_pol_route_lookup);
 903         if (dst->error == 0)
 904                 return (struct rt6_info *) dst;
 905
 906         dst_release(dst);
 907
 908         return NULL;
 909 }
 910 EXPORT_SYMBOL(rt6_lookup);
 911
 912 /* ip6_ins_rt is called with FREE table->tb6_lock.
 913    It takes new route entry, the addition fails by any reason the
 914    route is freed. In any case, if caller does not hold it, it may
 915    be destroyed.
 916  */
 917
 918 static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info,
 919                         struct mx6_config *mxc)
 920 {
 921         int err;
 922         struct fib6_table *table;
 923
 924         table = rt->rt6i_table;
 925         write_lock_bh(&table->tb6_lock);
 926         err = fib6_add(&table->tb6_root, rt, info, mxc);
 927         write_unlock_bh(&table->tb6_lock);
 928
 929         return err;
 930 }
 931
 932 int ip6_ins_rt(struct rt6_info *rt)
 933 {
 934         struct nl_info info = { .nl_net = dev_net(rt->dst.dev), };
 935         struct mx6_config mxc = { .mx = NULL, };
 936
 937         return __ip6_ins_rt(rt, &info, &mxc);
 938 }
 939
 940 static struct rt6_info *ip6_rt_cache_alloc(struct rt6_info *ort,
 941                                            const struct in6_addr *daddr,
 942                                            const struct in6_addr *saddr)
 943 {
 944         struct rt6_info *rt;
 945
 946         /*
 947          *      Clone the route.
 948          */
 949
 950         if (ort->rt6i_flags & (RTF_CACHE | RTF_PCPU))
 951                 ort = (struct rt6_info *)ort->dst.from;
 952
 953         rt = __ip6_dst_alloc(dev_net(ort->dst.dev), ort->dst.dev, 0);
 954
 955         if (!rt)
 956                 return NULL;
 957
 958         ip6_rt_copy_init(rt, ort);
 959         rt->rt6i_flags |= RTF_CACHE;
 960         rt->rt6i_metric = 0;
 961         rt->dst.flags |= DST_HOST;
 962         rt->rt6i_dst.addr = *daddr;
 963         rt->rt6i_dst.plen = 128;
 964
 965         if (!rt6_is_gw_or_nonexthop(ort)) {
 966                 if (ort->rt6i_dst.plen != 128 &&
 967                     ipv6_addr_equal(&ort->rt6i_dst.addr, daddr))
 968                         rt->rt6i_flags |= RTF_ANYCAST;
 969 #ifdef CONFIG_IPV6_SUBTREES
 970                 if (rt->rt6i_src.plen && saddr) {
 971                         rt->rt6i_src.addr = *saddr;
 972                         rt->rt6i_src.plen = 128;
 973                 }
 974 #endif
 975         }
 976
 977         return rt;
 978 }
 979
 980 static struct rt6_info *ip6_rt_pcpu_alloc(struct rt6_info *rt)
 981 {
 982         struct rt6_info *pcpu_rt;
 983
 984         pcpu_rt = __ip6_dst_alloc(dev_net(rt->dst.dev),
 985                                   rt->dst.dev, rt->dst.flags);
 986
 987         if (!pcpu_rt)
 988                 return NULL;
 989         ip6_rt_copy_init(pcpu_rt, rt);
 990         pcpu_rt->rt6i_protocol = rt->rt6i_protocol;
 991         pcpu_rt->rt6i_flags |= RTF_PCPU;
 992         return pcpu_rt;
 993 }
 994
 995 /* It should be called with read_lock_bh(&tb6_lock) acquired */
 996 static struct rt6_info *rt6_get_pcpu_route(struct rt6_info *rt)
 997 {
 998         struct rt6_info *pcpu_rt, **p;
 999
1000         p = this_cpu_ptr(rt->rt6i_pcpu);
1001         pcpu_rt = *p;
1002
1003         if (pcpu_rt) {
1004                 dst_hold(&pcpu_rt->dst);
1005                 rt6_dst_from_metrics_check(pcpu_rt);
1006         }
1007         return pcpu_rt;
1008 }
1009
1010 static struct rt6_info *rt6_make_pcpu_route(struct rt6_info *rt)
1011 {
1012         struct fib6_table *table = rt->rt6i_table;
1013         struct rt6_info *pcpu_rt, *prev, **p;
1014
1015         pcpu_rt = ip6_rt_pcpu_alloc(rt);
1016         if (!pcpu_rt) {
1017                 struct net *net = dev_net(rt->dst.dev);
1018
1019                 dst_hold(&net->ipv6.ip6_null_entry->dst);
1020                 return net->ipv6.ip6_null_entry;
1021         }
1022
1023         read_lock_bh(&table->tb6_lock);
1024         if (rt->rt6i_pcpu) {
1025                 p = this_cpu_ptr(rt->rt6i_pcpu);
1026                 prev = cmpxchg(p, NULL, pcpu_rt);
1027                 if (prev) {
1028                         /* If someone did it before us, return prev instead */
1029                         dst_destroy(&pcpu_rt->dst);
1030                         pcpu_rt = prev;
1031                 }
1032         } else {
1033                 /* rt has been removed from the fib6 tree
1034                  * before we have a chance to acquire the read_lock.
1035                  * In this case, don't brother to create a pcpu rt
1036                  * since rt is going away anyway.  The next
1037                  * dst_check() will trigger a re-lookup.
1038                  */
1039                 dst_destroy(&pcpu_rt->dst);
1040                 pcpu_rt = rt;
1041         }
1042         dst_hold(&pcpu_rt->dst);
1043         rt6_dst_from_metrics_check(pcpu_rt);
1044         read_unlock_bh(&table->tb6_lock);
1045         return pcpu_rt;
1046 }
1047
1048 struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table,
1049                                int oif, struct flowi6 *fl6, int flags)
1050 {
1051         struct fib6_node *fn, *saved_fn;
1052         struct rt6_info *rt;
1053         int strict = 0;
1054
1055         strict |= flags & RT6_LOOKUP_F_IFACE;
1056         strict |= flags & RT6_LOOKUP_F_IGNORE_LINKSTATE;
1057         if (net->ipv6.devconf_all->forwarding == 0)
1058                 strict |= RT6_LOOKUP_F_REACHABLE;
1059
1060         read_lock_bh(&table->tb6_lock);
1061
1062         fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1063         saved_fn = fn;
1064
1065         if (fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF)
1066                 oif = 0;
1067
1068 redo_rt6_select:
1069         rt = rt6_select(fn, oif, strict);
1070         if (rt->rt6i_nsiblings)
1071                 rt = rt6_multipath_select(rt, fl6, oif, strict);
1072         if (rt == net->ipv6.ip6_null_entry) {
1073                 fn = fib6_backtrack(fn, &fl6->saddr);
1074                 if (fn)
1075                         goto redo_rt6_select;
1076                 else if (strict & RT6_LOOKUP_F_REACHABLE) {
1077                         /* also consider unreachable route */
1078                         strict &= ~RT6_LOOKUP_F_REACHABLE;
1079                         fn = saved_fn;
1080                         goto redo_rt6_select;
1081                 }
1082         }
1083
1084
1085         if (rt == net->ipv6.ip6_null_entry || (rt->rt6i_flags & RTF_CACHE)) {
1086                 dst_use(&rt->dst, jiffies);
1087                 read_unlock_bh(&table->tb6_lock);
1088
1089                 rt6_dst_from_metrics_check(rt);
1090
1091                 trace_fib6_table_lookup(net, rt, table->tb6_id, fl6);
1092                 return rt;
1093         } else if (unlikely((fl6->flowi6_flags & FLOWI_FLAG_KNOWN_NH) &&
1094                             !(rt->rt6i_flags & RTF_GATEWAY))) {
1095                 /* Create a RTF_CACHE clone which will not be
1096                  * owned by the fib6 tree.  It is for the special case where
1097                  * the daddr in the skb during the neighbor look-up is different
1098                  * from the fl6->daddr used to look-up route here.
1099                  */
1100
1101                 struct rt6_info *uncached_rt;
1102
1103                 dst_use(&rt->dst, jiffies);
1104                 read_unlock_bh(&table->tb6_lock);
1105
1106                 uncached_rt = ip6_rt_cache_alloc(rt, &fl6->daddr, NULL);
1107                 dst_release(&rt->dst);
1108
1109                 if (uncached_rt)
1110                         rt6_uncached_list_add(uncached_rt);
1111                 else
1112                         uncached_rt = net->ipv6.ip6_null_entry;
1113
1114                 dst_hold(&uncached_rt->dst);
1115
1116                 trace_fib6_table_lookup(net, uncached_rt, table->tb6_id, fl6);
1117                 return uncached_rt;
1118
1119         } else {
1120                 /* Get a percpu copy */
1121
1122                 struct rt6_info *pcpu_rt;
1123
1124                 rt->dst.lastuse = jiffies;
1125                 rt->dst.__use++;
1126                 pcpu_rt = rt6_get_pcpu_route(rt);
1127
1128                 if (pcpu_rt) {
1129                         read_unlock_bh(&table->tb6_lock);
1130                 } else {
1131                         /* We have to do the read_unlock first
1132                          * because rt6_make_pcpu_route() may trigger
1133                          * ip6_dst_gc() which will take the write_lock.
1134                          */
1135                         dst_hold(&rt->dst);
1136                         read_unlock_bh(&table->tb6_lock);
1137                         pcpu_rt = rt6_make_pcpu_route(rt);
1138                         dst_release(&rt->dst);
1139                 }
1140
1141                 trace_fib6_table_lookup(net, pcpu_rt, table->tb6_id, fl6);
1142                 return pcpu_rt;
1143
1144         }
1145 }
1146 EXPORT_SYMBOL_GPL(ip6_pol_route);
1147
1148 static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
1149                                             struct flowi6 *fl6, int flags)
1150 {
1151         return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, flags);
1152 }
1153
1154 struct dst_entry *ip6_route_input_lookup(struct net *net,
1155                                          struct net_device *dev,
1156                                          struct flowi6 *fl6, int flags)
1157 {
1158         if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
1159                 flags |= RT6_LOOKUP_F_IFACE;
1160
1161         return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_input);
1162 }
1163 EXPORT_SYMBOL_GPL(ip6_route_input_lookup);
1164
1165 void ip6_route_input(struct sk_buff *skb)
1166 {
1167         const struct ipv6hdr *iph = ipv6_hdr(skb);
1168         struct net *net = dev_net(skb->dev);
1169         int flags = RT6_LOOKUP_F_HAS_SADDR;
1170         struct ip_tunnel_info *tun_info;
1171         struct flowi6 fl6 = {
1172                 .flowi6_iif = skb->dev->ifindex,
1173                 .daddr = iph->daddr,
1174                 .saddr = iph->saddr,
1175                 .flowlabel = ip6_flowinfo(iph),
1176                 .flowi6_mark = skb->mark,
1177                 .flowi6_proto = iph->nexthdr,
1178         };
1179
1180         tun_info = skb_tunnel_info(skb);
1181         if (tun_info && !(tun_info->mode & IP_TUNNEL_INFO_TX))
1182                 fl6.flowi6_tun_key.tun_id = tun_info->key.tun_id;
1183         skb_dst_drop(skb);
1184         skb_dst_set(skb, ip6_route_input_lookup(net, skb->dev, &fl6, flags));
1185 }
1186
1187 static struct rt6_info *ip6_pol_route_output(struct net *net, struct fib6_table *table,
1188                                              struct flowi6 *fl6, int flags)
1189 {
1190         return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, flags);
1191 }
1192
1193 struct dst_entry *ip6_route_output_flags(struct net *net, const struct sock *sk,
1194                                          struct flowi6 *fl6, int flags)
1195 {
1196         bool any_src;
1197
1198         if (rt6_need_strict(&fl6->daddr)) {
1199                 struct dst_entry *dst;
1200
1201                 dst = l3mdev_link_scope_lookup(net, fl6);
1202                 if (dst)
1203                         return dst;
1204         }
1205
1206         fl6->flowi6_iif = LOOPBACK_IFINDEX;
1207
1208         any_src = ipv6_addr_any(&fl6->saddr);
1209         if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr) ||
1210             (fl6->flowi6_oif && any_src))
1211                 flags |= RT6_LOOKUP_F_IFACE;
1212
1213         if (!any_src)
1214                 flags |= RT6_LOOKUP_F_HAS_SADDR;
1215         else if (sk)
1216                 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
1217
1218         return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_output);
1219 }
1220 EXPORT_SYMBOL_GPL(ip6_route_output_flags);
1221
1222 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
1223 {
1224         struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
1225         struct dst_entry *new = NULL;
1226
1227         rt = dst_alloc(&ip6_dst_blackhole_ops, ort->dst.dev, 1, DST_OBSOLETE_NONE, 0);
1228         if (rt) {
1229                 rt6_info_init(rt);
1230
1231                 new = &rt->dst;
1232                 new->__use = 1;
1233                 new->input = dst_discard;
1234                 new->output = dst_discard_out;
1235
1236                 dst_copy_metrics(new, &ort->dst);
1237                 rt->rt6i_idev = ort->rt6i_idev;
1238                 if (rt->rt6i_idev)
1239                         in6_dev_hold(rt->rt6i_idev);
1240
1241                 rt->rt6i_gateway = ort->rt6i_gateway;
1242                 rt->rt6i_flags = ort->rt6i_flags & ~RTF_PCPU;
1243                 rt->rt6i_metric = 0;
1244
1245                 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1246 #ifdef CONFIG_IPV6_SUBTREES
1247                 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1248 #endif
1249
1250                 dst_free(new);
1251         }
1252
1253         dst_release(dst_orig);
1254         return new ? new : ERR_PTR(-ENOMEM);
1255 }
1256
1257 /*
1258  *      Destination cache support functions
1259  */
1260
1261 static void rt6_dst_from_metrics_check(struct rt6_info *rt)
1262 {
1263         if (rt->dst.from &&
1264             dst_metrics_ptr(&rt->dst) != dst_metrics_ptr(rt->dst.from))
1265                 dst_init_metrics(&rt->dst, dst_metrics_ptr(rt->dst.from), true);
1266 }
1267
1268 static struct dst_entry *rt6_check(struct rt6_info *rt, u32 cookie)
1269 {
1270         if (!rt->rt6i_node || (rt->rt6i_node->fn_sernum != cookie))
1271                 return NULL;
1272
1273         if (rt6_check_expired(rt))
1274                 return NULL;
1275
1276         return &rt->dst;
1277 }
1278
1279 static struct dst_entry *rt6_dst_from_check(struct rt6_info *rt, u32 cookie)
1280 {
1281         if (!__rt6_check_expired(rt) &&
1282             rt->dst.obsolete == DST_OBSOLETE_FORCE_CHK &&
1283             rt6_check((struct rt6_info *)(rt->dst.from), cookie))
1284                 return &rt->dst;
1285         else
1286                 return NULL;
1287 }
1288
1289 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
1290 {
1291         struct rt6_info *rt;
1292
1293         rt = (struct rt6_info *) dst;
1294
1295         /* All IPV6 dsts are created with ->obsolete set to the value
1296          * DST_OBSOLETE_FORCE_CHK which forces validation calls down
1297          * into this function always.
1298          */
1299
1300         rt6_dst_from_metrics_check(rt);
1301
1302         if (rt->rt6i_flags & RTF_PCPU ||
1303             (unlikely(dst->flags & DST_NOCACHE) && rt->dst.from))
1304                 return rt6_dst_from_check(rt, cookie);
1305         else
1306                 return rt6_check(rt, cookie);
1307 }
1308
1309 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
1310 {
1311         struct rt6_info *rt = (struct rt6_info *) dst;
1312
1313         if (rt) {
1314                 if (rt->rt6i_flags & RTF_CACHE) {
1315                         if (rt6_check_expired(rt)) {
1316                                 ip6_del_rt(rt);
1317                                 dst = NULL;
1318                         }
1319                 } else {
1320                         dst_release(dst);
1321                         dst = NULL;
1322                 }
1323         }
1324         return dst;
1325 }
1326
1327 static void ip6_link_failure(struct sk_buff *skb)
1328 {
1329         struct rt6_info *rt;
1330
1331         icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
1332
1333         rt = (struct rt6_info *) skb_dst(skb);
1334         if (rt) {
1335                 if (rt->rt6i_flags & RTF_CACHE) {
1336                         dst_hold(&rt->dst);
1337                         ip6_del_rt(rt);
1338                 } else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT)) {
1339                         rt->rt6i_node->fn_sernum = -1;
1340                 }
1341         }
1342 }
1343
1344 static void rt6_do_update_pmtu(struct rt6_info *rt, u32 mtu)
1345 {
1346         struct net *net = dev_net(rt->dst.dev);
1347
1348         rt->rt6i_flags |= RTF_MODIFIED;
1349         rt->rt6i_pmtu = mtu;
1350         rt6_update_expires(rt, net->ipv6.sysctl.ip6_rt_mtu_expires);
1351 }
1352
1353 static bool rt6_cache_allowed_for_pmtu(const struct rt6_info *rt)
1354 {
1355         return !(rt->rt6i_flags & RTF_CACHE) &&
1356                 (rt->rt6i_flags & RTF_PCPU || rt->rt6i_node);
1357 }
1358
1359 static void __ip6_rt_update_pmtu(struct dst_entry *dst, const struct sock *sk,
1360                                  const struct ipv6hdr *iph, u32 mtu)
1361 {
1362         struct rt6_info *rt6 = (struct rt6_info *)dst;
1363
1364         if (rt6->rt6i_flags & RTF_LOCAL)
1365                 return;
1366
1367         if (dst_metric_locked(dst, RTAX_MTU))
1368                 return;
1369
1370         dst_confirm(dst);
1371         mtu = max_t(u32, mtu, IPV6_MIN_MTU);
1372         if (mtu >= dst_mtu(dst))
1373                 return;
1374
1375         if (!rt6_cache_allowed_for_pmtu(rt6)) {
1376                 rt6_do_update_pmtu(rt6, mtu);
1377         } else {
1378                 const struct in6_addr *daddr, *saddr;
1379                 struct rt6_info *nrt6;
1380
1381                 if (iph) {
1382                         daddr = &iph->daddr;
1383                         saddr = &iph->saddr;
1384                 } else if (sk) {
1385                         daddr = &sk->sk_v6_daddr;
1386                         saddr = &inet6_sk(sk)->saddr;
1387                 } else {
1388                         return;
1389                 }
1390                 nrt6 = ip6_rt_cache_alloc(rt6, daddr, saddr);
1391                 if (nrt6) {
1392                         rt6_do_update_pmtu(nrt6, mtu);
1393
1394                         /* ip6_ins_rt(nrt6) will bump the
1395                          * rt6->rt6i_node->fn_sernum
1396                          * which will fail the next rt6_check() and
1397                          * invalidate the sk->sk_dst_cache.
1398                          */
1399                         ip6_ins_rt(nrt6);
1400                 }
1401         }
1402 }
1403
1404 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
1405                                struct sk_buff *skb, u32 mtu)
1406 {
1407         __ip6_rt_update_pmtu(dst, sk, skb ? ipv6_hdr(skb) : NULL, mtu);
1408 }
1409
1410 void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
1411                      int oif, u32 mark, kuid_t uid)
1412 {
1413         const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1414         struct dst_entry *dst;
1415         struct flowi6 fl6;
1416
1417         memset(&fl6, 0, sizeof(fl6));
1418         fl6.flowi6_oif = oif;
1419         fl6.flowi6_mark = mark ? mark : IP6_REPLY_MARK(net, skb->mark);
1420         fl6.daddr = iph->daddr;
1421         fl6.saddr = iph->saddr;
1422         fl6.flowlabel = ip6_flowinfo(iph);
1423         fl6.flowi6_uid = uid;
1424
1425         dst = ip6_route_output(net, NULL, &fl6);
1426         if (!dst->error)
1427                 __ip6_rt_update_pmtu(dst, NULL, iph, ntohl(mtu));
1428         dst_release(dst);
1429 }
1430 EXPORT_SYMBOL_GPL(ip6_update_pmtu);
1431
1432 void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
1433 {
1434         struct dst_entry *dst;
1435
1436         ip6_update_pmtu(skb, sock_net(sk), mtu,
1437                         sk->sk_bound_dev_if, sk->sk_mark, sk->sk_uid);
1438
1439         dst = __sk_dst_get(sk);
1440         if (!dst || !dst->obsolete ||
1441             dst->ops->check(dst, inet6_sk(sk)->dst_cookie))
1442                 return;
1443
1444         bh_lock_sock(sk);
1445         if (!sock_owned_by_user(sk) && !ipv6_addr_v4mapped(&sk->sk_v6_daddr))
1446                 ip6_datagram_dst_update(sk, false);
1447         bh_unlock_sock(sk);
1448 }
1449 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
1450
1451 /* Handle redirects */
1452 struct ip6rd_flowi {
1453         struct flowi6 fl6;
1454         struct in6_addr gateway;
1455 };
1456
1457 static struct rt6_info *__ip6_route_redirect(struct net *net,
1458                                              struct fib6_table *table,
1459                                              struct flowi6 *fl6,
1460                                              int flags)
1461 {
1462         struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
1463         struct rt6_info *rt;
1464         struct fib6_node *fn;
1465
1466         /* Get the "current" route for this destination and
1467          * check if the redirect has come from appropriate router.
1468          *
1469          * RFC 4861 specifies that redirects should only be
1470          * accepted if they come from the nexthop to the target.
1471          * Due to the way the routes are chosen, this notion
1472          * is a bit fuzzy and one might need to check all possible
1473          * routes.
1474          */
1475
1476         read_lock_bh(&table->tb6_lock);
1477         fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1478 restart:
1479         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1480                 if (rt6_check_expired(rt))
1481                         continue;
1482                 if (rt->dst.error)
1483                         break;
1484                 if (!(rt->rt6i_flags & RTF_GATEWAY))
1485                         continue;
1486                 if (fl6->flowi6_oif != rt->dst.dev->ifindex)
1487                         continue;
1488                 if (!ipv6_addr_equal(&rdfl->gateway, &rt->rt6i_gateway))
1489                         continue;
1490                 break;
1491         }
1492
1493         if (!rt)
1494                 rt = net->ipv6.ip6_null_entry;
1495         else if (rt->dst.error) {
1496                 rt = net->ipv6.ip6_null_entry;
1497                 goto out;
1498         }
1499
1500         if (rt == net->ipv6.ip6_null_entry) {
1501                 fn = fib6_backtrack(fn, &fl6->saddr);
1502                 if (fn)
1503                         goto restart;
1504         }
1505
1506 out:
1507         dst_hold(&rt->dst);
1508
1509         read_unlock_bh(&table->tb6_lock);
1510
1511         trace_fib6_table_lookup(net, rt, table->tb6_id, fl6);
1512         return rt;
1513 };
1514
1515 static struct dst_entry *ip6_route_redirect(struct net *net,
1516                                         const struct flowi6 *fl6,
1517                                         const struct in6_addr *gateway)
1518 {
1519         int flags = RT6_LOOKUP_F_HAS_SADDR;
1520         struct ip6rd_flowi rdfl;
1521
1522         rdfl.fl6 = *fl6;
1523         rdfl.gateway = *gateway;
1524
1525         return fib6_rule_lookup(net, &rdfl.fl6,
1526                                 flags, __ip6_route_redirect);
1527 }
1528
1529 void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark,
1530                   kuid_t uid)
1531 {
1532         const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1533         struct dst_entry *dst;
1534         struct flowi6 fl6;
1535
1536         memset(&fl6, 0, sizeof(fl6));
1537         fl6.flowi6_iif = LOOPBACK_IFINDEX;
1538         fl6.flowi6_oif = oif;
1539         fl6.flowi6_mark = mark;
1540         fl6.daddr = iph->daddr;
1541         fl6.saddr = iph->saddr;
1542         fl6.flowlabel = ip6_flowinfo(iph);
1543         fl6.flowi6_uid = uid;
1544
1545         dst = ip6_route_redirect(net, &fl6, &ipv6_hdr(skb)->saddr);
1546         rt6_do_redirect(dst, NULL, skb);
1547         dst_release(dst);
1548 }
1549 EXPORT_SYMBOL_GPL(ip6_redirect);
1550
1551 void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif,
1552                             u32 mark)
1553 {
1554         const struct ipv6hdr *iph = ipv6_hdr(skb);
1555         const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
1556         struct dst_entry *dst;
1557         struct flowi6 fl6;
1558
1559         memset(&fl6, 0, sizeof(fl6));
1560         fl6.flowi6_iif = LOOPBACK_IFINDEX;
1561         fl6.flowi6_oif = oif;
1562         fl6.flowi6_mark = mark;
1563         fl6.daddr = msg->dest;
1564         fl6.saddr = iph->daddr;
1565         fl6.flowi6_uid = sock_net_uid(net, NULL);
1566
1567         dst = ip6_route_redirect(net, &fl6, &iph->saddr);
1568         rt6_do_redirect(dst, NULL, skb);
1569         dst_release(dst);
1570 }
1571
1572 void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
1573 {
1574         ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark,
1575                      sk->sk_uid);
1576 }
1577 EXPORT_SYMBOL_GPL(ip6_sk_redirect);
1578
1579 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1580 {
1581         struct net_device *dev = dst->dev;
1582         unsigned int mtu = dst_mtu(dst);
1583         struct net *net = dev_net(dev);
1584
1585         mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
1586
1587         if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
1588                 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1589
1590         /*
1591          * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1592          * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1593          * IPV6_MAXPLEN is also valid and means: "any MSS,
1594          * rely only on pmtu discovery"
1595          */
1596         if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
1597                 mtu = IPV6_MAXPLEN;
1598         return mtu;
1599 }
1600
1601 static unsigned int ip6_mtu(const struct dst_entry *dst)
1602 {
1603         const struct rt6_info *rt = (const struct rt6_info *)dst;
1604         unsigned int mtu = rt->rt6i_pmtu;
1605         struct inet6_dev *idev;
1606
1607         if (mtu)
1608                 goto out;
1609
1610         mtu = dst_metric_raw(dst, RTAX_MTU);
1611         if (mtu)
1612                 goto out;
1613
1614         mtu = IPV6_MIN_MTU;
1615
1616         rcu_read_lock();
1617         idev = __in6_dev_get(dst->dev);
1618         if (idev)
1619                 mtu = idev->cnf.mtu6;
1620         rcu_read_unlock();
1621
1622 out:
1623         mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
1624
1625         return mtu - lwtunnel_headroom(dst->lwtstate, mtu);
1626 }
1627
1628 static struct dst_entry *icmp6_dst_gc_list;
1629 static DEFINE_SPINLOCK(icmp6_dst_lock);
1630
1631 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
1632                                   struct flowi6 *fl6)
1633 {
1634         struct dst_entry *dst;
1635         struct rt6_info *rt;
1636         struct inet6_dev *idev = in6_dev_get(dev);
1637         struct net *net = dev_net(dev);
1638
1639         if (unlikely(!idev))
1640                 return ERR_PTR(-ENODEV);
1641
1642         rt = ip6_dst_alloc(net, dev, 0);
1643         if (unlikely(!rt)) {
1644                 in6_dev_put(idev);
1645                 dst = ERR_PTR(-ENOMEM);
1646                 goto out;
1647         }
1648
1649         rt->dst.flags |= DST_HOST;
1650         rt->dst.output  = ip6_output;
1651         atomic_set(&rt->dst.__refcnt, 1);
1652         rt->rt6i_gateway  = fl6->daddr;
1653         rt->rt6i_dst.addr = fl6->daddr;
1654         rt->rt6i_dst.plen = 128;
1655         rt->rt6i_idev     = idev;
1656         dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
1657
1658         spin_lock_bh(&icmp6_dst_lock);
1659         rt->dst.next = icmp6_dst_gc_list;
1660         icmp6_dst_gc_list = &rt->dst;
1661         spin_unlock_bh(&icmp6_dst_lock);
1662
1663         fib6_force_start_gc(net);
1664
1665         dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
1666
1667 out:
1668         return dst;
1669 }
1670
1671 int icmp6_dst_gc(void)
1672 {
1673         struct dst_entry *dst, **pprev;
1674         int more = 0;
1675
1676         spin_lock_bh(&icmp6_dst_lock);
1677         pprev = &icmp6_dst_gc_list;
1678
1679         while ((dst = *pprev) != NULL) {
1680                 if (!atomic_read(&dst->__refcnt)) {
1681                         *pprev = dst->next;
1682                         dst_free(dst);
1683                 } else {
1684                         pprev = &dst->next;
1685                         ++more;
1686                 }
1687         }
1688
1689         spin_unlock_bh(&icmp6_dst_lock);
1690
1691         return more;
1692 }
1693
1694 static void icmp6_clean_all(int (*func)(struct rt6_info *rt, void *arg),
1695                             void *arg)
1696 {
1697         struct dst_entry *dst, **pprev;
1698
1699         spin_lock_bh(&icmp6_dst_lock);
1700         pprev = &icmp6_dst_gc_list;
1701         while ((dst = *pprev) != NULL) {
1702                 struct rt6_info *rt = (struct rt6_info *) dst;
1703                 if (func(rt, arg)) {
1704                         *pprev = dst->next;
1705                         dst_free(dst);
1706                 } else {
1707                         pprev = &dst->next;
1708                 }
1709         }
1710         spin_unlock_bh(&icmp6_dst_lock);
1711 }
1712
1713 static int ip6_dst_gc(struct dst_ops *ops)
1714 {
1715         struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
1716         int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1717         int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1718         int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1719         int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1720         unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1721         int entries;
1722
1723         entries = dst_entries_get_fast(ops);
1724         if (time_after(rt_last_gc + rt_min_interval, jiffies) &&
1725             entries <= rt_max_size)
1726                 goto out;
1727
1728         net->ipv6.ip6_rt_gc_expire++;
1729         fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net, true);
1730         entries = dst_entries_get_slow(ops);
1731         if (entries < ops->gc_thresh)
1732                 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1733 out:
1734         net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1735         return entries > rt_max_size;
1736 }
1737
1738 static int ip6_convert_metrics(struct mx6_config *mxc,
1739                                const struct fib6_config *cfg)
1740 {
1741         bool ecn_ca = false;
1742         struct nlattr *nla;
1743         int remaining;
1744         u32 *mp;
1745
1746         if (!cfg->fc_mx)
1747                 return 0;
1748
1749         mp = kzalloc(sizeof(u32) * RTAX_MAX, GFP_KERNEL);
1750         if (unlikely(!mp))
1751                 return -ENOMEM;
1752
1753         nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1754                 int type = nla_type(nla);
1755                 u32 val;
1756
1757                 if (!type)
1758                         continue;
1759                 if (unlikely(type > RTAX_MAX))
1760                         goto err;
1761
1762                 if (type == RTAX_CC_ALGO) {
1763                         char tmp[TCP_CA_NAME_MAX];
1764
1765                         nla_strlcpy(tmp, nla, sizeof(tmp));
1766                         val = tcp_ca_get_key_by_name(tmp, &ecn_ca);
1767                         if (val == TCP_CA_UNSPEC)
1768                                 goto err;
1769                 } else {
1770                         val = nla_get_u32(nla);
1771                 }
1772                 if (type == RTAX_HOPLIMIT && val > 255)
1773                         val = 255;
1774                 if (type == RTAX_FEATURES && (val & ~RTAX_FEATURE_MASK))
1775                         goto err;
1776
1777                 mp[type - 1] = val;
1778                 __set_bit(type - 1, mxc->mx_valid);
1779         }
1780
1781         if (ecn_ca) {
1782                 __set_bit(RTAX_FEATURES - 1, mxc->mx_valid);
1783                 mp[RTAX_FEATURES - 1] |= DST_FEATURE_ECN_CA;
1784         }
1785
1786         mxc->mx = mp;
1787         return 0;
1788  err:
1789         kfree(mp);
1790         return -EINVAL;
1791 }
1792
1793 static struct rt6_info *ip6_nh_lookup_table(struct net *net,
1794                                             struct fib6_config *cfg,
1795                                             const struct in6_addr *gw_addr)
1796 {
1797         struct flowi6 fl6 = {
1798                 .flowi6_oif = cfg->fc_ifindex,
1799                 .daddr = *gw_addr,
1800                 .saddr = cfg->fc_prefsrc,
1801         };
1802         struct fib6_table *table;
1803         struct rt6_info *rt;
1804         int flags = RT6_LOOKUP_F_IFACE | RT6_LOOKUP_F_IGNORE_LINKSTATE;
1805
1806         table = fib6_get_table(net, cfg->fc_table);
1807         if (!table)
1808                 return NULL;
1809
1810         if (!ipv6_addr_any(&cfg->fc_prefsrc))
1811                 flags |= RT6_LOOKUP_F_HAS_SADDR;
1812
1813         rt = ip6_pol_route(net, table, cfg->fc_ifindex, &fl6, flags);
1814
1815         /* if table lookup failed, fall back to full lookup */
1816         if (rt == net->ipv6.ip6_null_entry) {
1817                 ip6_rt_put(rt);
1818                 rt = NULL;
1819         }
1820
1821         return rt;
1822 }
1823
1824 static struct rt6_info *ip6_route_info_create(struct fib6_config *cfg)
1825 {
1826         struct net *net = cfg->fc_nlinfo.nl_net;
1827         struct rt6_info *rt = NULL;
1828         struct net_device *dev = NULL;
1829         struct inet6_dev *idev = NULL;
1830         struct fib6_table *table;
1831         int addr_type;
1832         int err = -EINVAL;
1833
1834         if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1835                 goto out;
1836 #ifndef CONFIG_IPV6_SUBTREES
1837         if (cfg->fc_src_len)
1838                 goto out;
1839 #endif
1840         if (cfg->fc_ifindex) {
1841                 err = -ENODEV;
1842                 dev = dev_get_by_index(net, cfg->fc_ifindex);
1843                 if (!dev)
1844                         goto out;
1845                 idev = in6_dev_get(dev);
1846                 if (!idev)
1847                         goto out;
1848         }
1849
1850         if (cfg->fc_metric == 0)
1851                 cfg->fc_metric = IP6_RT_PRIO_USER;
1852
1853         err = -ENOBUFS;
1854         if (cfg->fc_nlinfo.nlh &&
1855             !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
1856                 table = fib6_get_table(net, cfg->fc_table);
1857                 if (!table) {
1858                         pr_warn("NLM_F_CREATE should be specified when creating new route\n");
1859                         table = fib6_new_table(net, cfg->fc_table);
1860                 }
1861         } else {
1862                 table = fib6_new_table(net, cfg->fc_table);
1863         }
1864
1865         if (!table)
1866                 goto out;
1867
1868         rt = ip6_dst_alloc(net, NULL,
1869                            (cfg->fc_flags & RTF_ADDRCONF) ? 0 : DST_NOCOUNT);
1870
1871         if (!rt) {
1872                 err = -ENOMEM;
1873                 goto out;
1874         }
1875
1876         if (cfg->fc_flags & RTF_EXPIRES)
1877                 rt6_set_expires(rt, jiffies +
1878                                 clock_t_to_jiffies(cfg->fc_expires));
1879         else
1880                 rt6_clean_expires(rt);
1881
1882         if (cfg->fc_protocol == RTPROT_UNSPEC)
1883                 cfg->fc_protocol = RTPROT_BOOT;
1884         rt->rt6i_protocol = cfg->fc_protocol;
1885
1886         addr_type = ipv6_addr_type(&cfg->fc_dst);
1887
1888         if (addr_type & IPV6_ADDR_MULTICAST)
1889                 rt->dst.input = ip6_mc_input;
1890         else if (cfg->fc_flags & RTF_LOCAL)
1891                 rt->dst.input = ip6_input;
1892         else
1893                 rt->dst.input = ip6_forward;
1894
1895         rt->dst.output = ip6_output;
1896
1897         if (cfg->fc_encap) {
1898                 struct lwtunnel_state *lwtstate;
1899
1900                 err = lwtunnel_build_state(dev, cfg->fc_encap_type,
1901                                            cfg->fc_encap, AF_INET6, cfg,
1902                                            &lwtstate);
1903                 if (err)
1904                         goto out;
1905                 rt->dst.lwtstate = lwtstate_get(lwtstate);
1906                 if (lwtunnel_output_redirect(rt->dst.lwtstate)) {
1907                         rt->dst.lwtstate->orig_output = rt->dst.output;
1908                         rt->dst.output = lwtunnel_output;
1909                 }
1910                 if (lwtunnel_input_redirect(rt->dst.lwtstate)) {
1911                         rt->dst.lwtstate->orig_input = rt->dst.input;
1912                         rt->dst.input = lwtunnel_input;
1913                 }
1914         }
1915
1916         ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1917         rt->rt6i_dst.plen = cfg->fc_dst_len;
1918         if (rt->rt6i_dst.plen == 128)
1919                 rt->dst.flags |= DST_HOST;
1920
1921 #ifdef CONFIG_IPV6_SUBTREES
1922         ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1923         rt->rt6i_src.plen = cfg->fc_src_len;
1924 #endif
1925
1926         rt->rt6i_metric = cfg->fc_metric;
1927
1928         /* We cannot add true routes via loopback here,
1929            they would result in kernel looping; promote them to reject routes
1930          */
1931         if ((cfg->fc_flags & RTF_REJECT) ||
1932             (dev && (dev->flags & IFF_LOOPBACK) &&
1933              !(addr_type & IPV6_ADDR_LOOPBACK) &&
1934              !(cfg->fc_flags & RTF_LOCAL))) {
1935                 /* hold loopback dev/idev if we haven't done so. */
1936                 if (dev != net->loopback_dev) {
1937                         if (dev) {
1938                                 dev_put(dev);
1939                                 in6_dev_put(idev);
1940                         }
1941                         dev = net->loopback_dev;
1942                         dev_hold(dev);
1943                         idev = in6_dev_get(dev);
1944                         if (!idev) {
1945                                 err = -ENODEV;
1946                                 goto out;
1947                         }
1948                 }
1949                 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1950                 switch (cfg->fc_type) {
1951                 case RTN_BLACKHOLE:
1952                         rt->dst.error = -EINVAL;
1953                         rt->dst.output = dst_discard_out;
1954                         rt->dst.input = dst_discard;
1955                         break;
1956                 case RTN_PROHIBIT:
1957                         rt->dst.error = -EACCES;
1958                         rt->dst.output = ip6_pkt_prohibit_out;
1959                         rt->dst.input = ip6_pkt_prohibit;
1960                         break;
1961                 case RTN_THROW:
1962                 case RTN_UNREACHABLE:
1963                 default:
1964                         rt->dst.error = (cfg->fc_type == RTN_THROW) ? -EAGAIN
1965                                         : (cfg->fc_type == RTN_UNREACHABLE)
1966                                         ? -EHOSTUNREACH : -ENETUNREACH;
1967                         rt->dst.output = ip6_pkt_discard_out;
1968                         rt->dst.input = ip6_pkt_discard;
1969                         break;
1970                 }
1971                 goto install_route;
1972         }
1973
1974         if (cfg->fc_flags & RTF_GATEWAY) {
1975                 const struct in6_addr *gw_addr;
1976                 int gwa_type;
1977
1978                 gw_addr = &cfg->fc_gateway;
1979                 gwa_type = ipv6_addr_type(gw_addr);
1980
1981                 /* if gw_addr is local we will fail to detect this in case
1982                  * address is still TENTATIVE (DAD in progress). rt6_lookup()
1983                  * will return already-added prefix route via interface that
1984                  * prefix route was assigned to, which might be non-loopback.
1985                  */
1986                 err = -EINVAL;
1987                 if (ipv6_chk_addr_and_flags(net, gw_addr,
1988                                             gwa_type & IPV6_ADDR_LINKLOCAL ?
1989                                             dev : NULL, 0, 0))
1990                         goto out;
1991
1992                 rt->rt6i_gateway = *gw_addr;
1993
1994                 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1995                         struct rt6_info *grt = NULL;
1996
1997                         /* IPv6 strictly inhibits using not link-local
1998                            addresses as nexthop address.
1999                            Otherwise, router will not able to send redirects.
2000                            It is very good, but in some (rare!) circumstances
2001                            (SIT, PtP, NBMA NOARP links) it is handy to allow
2002                            some exceptions. --ANK
2003                            We allow IPv4-mapped nexthops to support RFC4798-type
2004                            addressing
2005                          */
2006                         if (!(gwa_type & (IPV6_ADDR_UNICAST |
2007                                           IPV6_ADDR_MAPPED)))
2008                                 goto out;
2009
2010                         if (cfg->fc_table) {
2011                                 grt = ip6_nh_lookup_table(net, cfg, gw_addr);
2012
2013                                 if (grt) {
2014                                         if (grt->rt6i_flags & RTF_GATEWAY ||
2015                                             (dev && dev != grt->dst.dev)) {
2016                                                 ip6_rt_put(grt);
2017                                                 grt = NULL;
2018                                         }
2019                                 }
2020                         }
2021
2022                         if (!grt)
2023                                 grt = rt6_lookup(net, gw_addr, NULL,
2024                                                  cfg->fc_ifindex, 1);
2025
2026                         err = -EHOSTUNREACH;
2027                         if (!grt)
2028                                 goto out;
2029                         if (dev) {
2030                                 if (dev != grt->dst.dev) {
2031                                         ip6_rt_put(grt);
2032                                         goto out;
2033                                 }
2034                         } else {
2035                                 dev = grt->dst.dev;
2036                                 idev = grt->rt6i_idev;
2037                                 dev_hold(dev);
2038                                 in6_dev_hold(grt->rt6i_idev);
2039                         }
2040                         if (!(grt->rt6i_flags & RTF_GATEWAY))
2041                                 err = 0;
2042                         ip6_rt_put(grt);
2043
2044                         if (err)
2045                                 goto out;
2046                 }
2047                 err = -EINVAL;
2048                 if (!dev || (dev->flags & IFF_LOOPBACK))
2049                         goto out;
2050         }
2051
2052         err = -ENODEV;
2053         if (!dev)
2054                 goto out;
2055
2056         if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
2057                 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
2058                         err = -EINVAL;
2059                         goto out;
2060                 }
2061                 rt->rt6i_prefsrc.addr = cfg->fc_prefsrc;
2062                 rt->rt6i_prefsrc.plen = 128;
2063         } else
2064                 rt->rt6i_prefsrc.plen = 0;
2065
2066         rt->rt6i_flags = cfg->fc_flags;
2067
2068 install_route:
2069         rt->dst.dev = dev;
2070         rt->rt6i_idev = idev;
2071         rt->rt6i_table = table;
2072
2073         cfg->fc_nlinfo.nl_net = dev_net(dev);
2074
2075         return rt;
2076 out:
2077         if (dev)
2078                 dev_put(dev);
2079         if (idev)
2080                 in6_dev_put(idev);
2081         if (rt)
2082                 dst_free(&rt->dst);
2083
2084         return ERR_PTR(err);
2085 }
2086
2087 int ip6_route_add(struct fib6_config *cfg)
2088 {
2089         struct mx6_config mxc = { .mx = NULL, };
2090         struct rt6_info *rt;
2091         int err;
2092
2093         rt = ip6_route_info_create(cfg);
2094         if (IS_ERR(rt)) {
2095                 err = PTR_ERR(rt);
2096                 rt = NULL;
2097                 goto out;
2098         }
2099
2100         err = ip6_convert_metrics(&mxc, cfg);
2101         if (err)
2102                 goto out;
2103
2104         err = __ip6_ins_rt(rt, &cfg->fc_nlinfo, &mxc);
2105
2106         kfree(mxc.mx);
2107
2108         return err;
2109 out:
2110         if (rt)
2111                 dst_free(&rt->dst);
2112
2113         return err;
2114 }
2115
2116 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
2117 {
2118         int err;
2119         struct fib6_table *table;
2120         struct net *net = dev_net(rt->dst.dev);
2121
2122         if (rt == net->ipv6.ip6_null_entry ||
2123             rt->dst.flags & DST_NOCACHE) {
2124                 err = -ENOENT;
2125                 goto out;
2126         }
2127
2128         table = rt->rt6i_table;
2129         write_lock_bh(&table->tb6_lock);
2130         err = fib6_del(rt, info);
2131         write_unlock_bh(&table->tb6_lock);
2132
2133 out:
2134         ip6_rt_put(rt);
2135         return err;
2136 }
2137
2138 int ip6_del_rt(struct rt6_info *rt)
2139 {
2140         struct nl_info info = {
2141                 .nl_net = dev_net(rt->dst.dev),
2142         };
2143         return __ip6_del_rt(rt, &info);
2144 }
2145
2146 static int ip6_route_del(struct fib6_config *cfg)
2147 {
2148         struct fib6_table *table;
2149         struct fib6_node *fn;
2150         struct rt6_info *rt;
2151         int err = -ESRCH;
2152
2153         table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
2154         if (!table)
2155                 return err;
2156
2157         read_lock_bh(&table->tb6_lock);
2158
2159         fn = fib6_locate(&table->tb6_root,
2160                          &cfg->fc_dst, cfg->fc_dst_len,
2161                          &cfg->fc_src, cfg->fc_src_len);
2162
2163         if (fn) {
2164                 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2165                         if ((rt->rt6i_flags & RTF_CACHE) &&
2166                             !(cfg->fc_flags & RTF_CACHE))
2167                                 continue;
2168                         if (cfg->fc_ifindex &&
2169                             (!rt->dst.dev ||
2170                              rt->dst.dev->ifindex != cfg->fc_ifindex))
2171                                 continue;
2172                         if (cfg->fc_flags & RTF_GATEWAY &&
2173                             !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
2174                                 continue;
2175                         if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
2176                                 continue;
2177                         if (cfg->fc_protocol && cfg->fc_protocol != rt->rt6i_protocol)
2178                                 continue;
2179                         dst_hold(&rt->dst);
2180                         read_unlock_bh(&table->tb6_lock);
2181
2182                         return __ip6_del_rt(rt, &cfg->fc_nlinfo);
2183                 }
2184         }
2185         read_unlock_bh(&table->tb6_lock);
2186
2187         return err;
2188 }
2189
2190 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
2191 {
2192         struct netevent_redirect netevent;
2193         struct rt6_info *rt, *nrt = NULL;
2194         struct ndisc_options ndopts;
2195         struct inet6_dev *in6_dev;
2196         struct neighbour *neigh;
2197         struct rd_msg *msg;
2198         int optlen, on_link;
2199         u8 *lladdr;
2200
2201         optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
2202         optlen -= sizeof(*msg);
2203
2204         if (optlen < 0) {
2205                 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
2206                 return;
2207         }
2208
2209         msg = (struct rd_msg *)icmp6_hdr(skb);
2210
2211         if (ipv6_addr_is_multicast(&msg->dest)) {
2212                 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
2213                 return;
2214         }
2215
2216         on_link = 0;
2217         if (ipv6_addr_equal(&msg->dest, &msg->target)) {
2218                 on_link = 1;
2219         } else if (ipv6_addr_type(&msg->target) !=
2220                    (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
2221                 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
2222                 return;
2223         }
2224
2225         in6_dev = __in6_dev_get(skb->dev);
2226         if (!in6_dev)
2227                 return;
2228         if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
2229                 return;
2230
2231         /* RFC2461 8.1:
2232          *      The IP source address of the Redirect MUST be the same as the current
2233          *      first-hop router for the specified ICMP Destination Address.
2234          */
2235
2236         if (!ndisc_parse_options(skb->dev, msg->opt, optlen, &ndopts)) {
2237                 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
2238                 return;
2239         }
2240
2241         lladdr = NULL;
2242         if (ndopts.nd_opts_tgt_lladdr) {
2243                 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
2244                                              skb->dev);
2245                 if (!lladdr) {
2246                         net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
2247                         return;
2248                 }
2249         }
2250
2251         rt = (struct rt6_info *) dst;
2252         if (rt->rt6i_flags & RTF_REJECT) {
2253                 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
2254                 return;
2255         }
2256
2257         /* Redirect received -> path was valid.
2258          * Look, redirects are sent only in response to data packets,
2259          * so that this nexthop apparently is reachable. --ANK
2260          */
2261         dst_confirm(&rt->dst);
2262
2263         neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
2264         if (!neigh)
2265                 return;
2266
2267         /*
2268          *      We have finally decided to accept it.
2269          */
2270
2271         ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
2272                      NEIGH_UPDATE_F_WEAK_OVERRIDE|
2273                      NEIGH_UPDATE_F_OVERRIDE|
2274                      (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
2275                                      NEIGH_UPDATE_F_ISROUTER)),
2276                      NDISC_REDIRECT, &ndopts);
2277
2278         nrt = ip6_rt_cache_alloc(rt, &msg->dest, NULL);
2279         if (!nrt)
2280                 goto out;
2281
2282         nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
2283         if (on_link)
2284                 nrt->rt6i_flags &= ~RTF_GATEWAY;
2285
2286         nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
2287
2288         if (ip6_ins_rt(nrt))
2289                 goto out;
2290
2291         netevent.old = &rt->dst;
2292         netevent.new = &nrt->dst;
2293         netevent.daddr = &msg->dest;
2294         netevent.neigh = neigh;
2295         call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
2296
2297         if (rt->rt6i_flags & RTF_CACHE) {
2298                 rt = (struct rt6_info *) dst_clone(&rt->dst);
2299                 ip6_del_rt(rt);
2300         }
2301
2302 out:
2303         neigh_release(neigh);
2304 }
2305
2306 /*
2307  *      Misc support functions
2308  */
2309
2310 static void rt6_set_from(struct rt6_info *rt, struct rt6_info *from)
2311 {
2312         BUG_ON(from->dst.from);
2313
2314         rt->rt6i_flags &= ~RTF_EXPIRES;
2315         dst_hold(&from->dst);
2316         rt->dst.from = &from->dst;
2317         dst_init_metrics(&rt->dst, dst_metrics_ptr(&from->dst), true);
2318 }
2319
2320 static void ip6_rt_copy_init(struct rt6_info *rt, struct rt6_info *ort)
2321 {
2322         rt->dst.input = ort->dst.input;
2323         rt->dst.output = ort->dst.output;
2324         rt->rt6i_dst = ort->rt6i_dst;
2325         rt->dst.error = ort->dst.error;
2326         rt->rt6i_idev = ort->rt6i_idev;
2327         if (rt->rt6i_idev)
2328                 in6_dev_hold(rt->rt6i_idev);
2329         rt->dst.lastuse = jiffies;
2330         rt->rt6i_gateway = ort->rt6i_gateway;
2331         rt->rt6i_flags = ort->rt6i_flags;
2332         rt6_set_from(rt, ort);
2333         rt->rt6i_metric = ort->rt6i_metric;
2334 #ifdef CONFIG_IPV6_SUBTREES
2335         rt->rt6i_src = ort->rt6i_src;
2336 #endif
2337         rt->rt6i_prefsrc = ort->rt6i_prefsrc;
2338         rt->rt6i_table = ort->rt6i_table;
2339         rt->dst.lwtstate = lwtstate_get(ort->dst.lwtstate);
2340 }
2341
2342 #ifdef CONFIG_IPV6_ROUTE_INFO
2343 static struct rt6_info *rt6_get_route_info(struct net *net,
2344                                            const struct in6_addr *prefix, int prefixlen,
2345                                            const struct in6_addr *gwaddr,
2346                                            struct net_device *dev)
2347 {
2348         u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO;
2349         int ifindex = dev->ifindex;
2350         struct fib6_node *fn;
2351         struct rt6_info *rt = NULL;
2352         struct fib6_table *table;
2353
2354         table = fib6_get_table(net, tb_id);
2355         if (!table)
2356                 return NULL;
2357
2358         read_lock_bh(&table->tb6_lock);
2359         fn = fib6_locate(&table->tb6_root, prefix, prefixlen, NULL, 0);
2360         if (!fn)
2361                 goto out;
2362
2363         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2364                 if (rt->dst.dev->ifindex != ifindex)
2365                         continue;
2366                 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
2367                         continue;
2368                 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
2369                         continue;
2370                 dst_hold(&rt->dst);
2371                 break;
2372         }
2373 out:
2374         read_unlock_bh(&table->tb6_lock);
2375         return rt;
2376 }
2377
2378 static struct rt6_info *rt6_add_route_info(struct net *net,
2379                                            const struct in6_addr *prefix, int prefixlen,
2380                                            const struct in6_addr *gwaddr,
2381                                            struct net_device *dev,
2382                                            unsigned int pref)
2383 {
2384         struct fib6_config cfg = {
2385                 .fc_metric      = IP6_RT_PRIO_USER,
2386                 .fc_ifindex     = dev->ifindex,
2387                 .fc_dst_len     = prefixlen,
2388                 .fc_flags       = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
2389                                   RTF_UP | RTF_PREF(pref),
2390                 .fc_nlinfo.portid = 0,
2391                 .fc_nlinfo.nlh = NULL,
2392                 .fc_nlinfo.nl_net = net,
2393         };
2394
2395         cfg.fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO,
2396         cfg.fc_dst = *prefix;
2397         cfg.fc_gateway = *gwaddr;
2398
2399         /* We should treat it as a default route if prefix length is 0. */
2400         if (!prefixlen)
2401                 cfg.fc_flags |= RTF_DEFAULT;
2402
2403         ip6_route_add(&cfg);
2404
2405         return rt6_get_route_info(net, prefix, prefixlen, gwaddr, dev);
2406 }
2407 #endif
2408
2409 struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
2410 {
2411         u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT;
2412         struct rt6_info *rt;
2413         struct fib6_table *table;
2414
2415         table = fib6_get_table(dev_net(dev), tb_id);
2416         if (!table)
2417                 return NULL;
2418
2419         read_lock_bh(&table->tb6_lock);
2420         for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2421                 if (dev == rt->dst.dev &&
2422                     ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
2423                     ipv6_addr_equal(&rt->rt6i_gateway, addr))
2424                         break;
2425         }
2426         if (rt)
2427                 dst_hold(&rt->dst);
2428         read_unlock_bh(&table->tb6_lock);
2429         return rt;
2430 }
2431
2432 struct rt6_info *rt6_add_dflt_router(const struct in6_addr *gwaddr,
2433                                      struct net_device *dev,
2434                                      unsigned int pref)
2435 {
2436         struct fib6_config cfg = {
2437                 .fc_table       = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT,
2438                 .fc_metric      = IP6_RT_PRIO_USER,
2439                 .fc_ifindex     = dev->ifindex,
2440                 .fc_flags       = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
2441                                   RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
2442                 .fc_nlinfo.portid = 0,
2443                 .fc_nlinfo.nlh = NULL,
2444                 .fc_nlinfo.nl_net = dev_net(dev),
2445         };
2446
2447         cfg.fc_gateway = *gwaddr;
2448
2449         if (!ip6_route_add(&cfg)) {
2450                 struct fib6_table *table;
2451
2452                 table = fib6_get_table(dev_net(dev), cfg.fc_table);
2453                 if (table)
2454                         table->flags |= RT6_TABLE_HAS_DFLT_ROUTER;
2455         }
2456
2457         return rt6_get_dflt_router(gwaddr, dev);
2458 }
2459
2460 static void __rt6_purge_dflt_routers(struct fib6_table *table)
2461 {
2462         struct rt6_info *rt;
2463
2464 restart:
2465         read_lock_bh(&table->tb6_lock);
2466         for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2467                 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
2468                     (!rt->rt6i_idev || rt->rt6i_idev->cnf.accept_ra != 2)) {
2469                         dst_hold(&rt->dst);
2470                         read_unlock_bh(&table->tb6_lock);
2471                         ip6_del_rt(rt);
2472                         goto restart;
2473                 }
2474         }
2475         read_unlock_bh(&table->tb6_lock);
2476
2477         table->flags &= ~RT6_TABLE_HAS_DFLT_ROUTER;
2478 }
2479
2480 void rt6_purge_dflt_routers(struct net *net)
2481 {
2482         struct fib6_table *table;
2483         struct hlist_head *head;
2484         unsigned int h;
2485
2486         rcu_read_lock();
2487
2488         for (h = 0; h < FIB6_TABLE_HASHSZ; h++) {
2489                 head = &net->ipv6.fib_table_hash[h];
2490                 hlist_for_each_entry_rcu(table, head, tb6_hlist) {
2491                         if (table->flags & RT6_TABLE_HAS_DFLT_ROUTER)
2492                                 __rt6_purge_dflt_routers(table);
2493                 }
2494         }
2495
2496         rcu_read_unlock();
2497 }
2498
2499 static void rtmsg_to_fib6_config(struct net *net,
2500                                  struct in6_rtmsg *rtmsg,
2501                                  struct fib6_config *cfg)
2502 {
2503         memset(cfg, 0, sizeof(*cfg));
2504
2505         cfg->fc_table = l3mdev_fib_table_by_index(net, rtmsg->rtmsg_ifindex) ?
2506                          : RT6_TABLE_MAIN;
2507         cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
2508         cfg->fc_metric = rtmsg->rtmsg_metric;
2509         cfg->fc_expires = rtmsg->rtmsg_info;
2510         cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
2511         cfg->fc_src_len = rtmsg->rtmsg_src_len;
2512         cfg->fc_flags = rtmsg->rtmsg_flags;
2513
2514         cfg->fc_nlinfo.nl_net = net;
2515
2516         cfg->fc_dst = rtmsg->rtmsg_dst;
2517         cfg->fc_src = rtmsg->rtmsg_src;
2518         cfg->fc_gateway = rtmsg->rtmsg_gateway;
2519 }
2520
2521 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
2522 {
2523         struct fib6_config cfg;
2524         struct in6_rtmsg rtmsg;
2525         int err;
2526
2527         switch (cmd) {
2528         case SIOCADDRT:         /* Add a route */
2529         case SIOCDELRT:         /* Delete a route */
2530                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2531                         return -EPERM;
2532                 err = copy_from_user(&rtmsg, arg,
2533                                      sizeof(struct in6_rtmsg));
2534                 if (err)
2535                         return -EFAULT;
2536
2537                 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
2538
2539                 rtnl_lock();
2540                 switch (cmd) {
2541                 case SIOCADDRT:
2542                         err = ip6_route_add(&cfg);
2543                         break;
2544                 case SIOCDELRT:
2545                         err = ip6_route_del(&cfg);
2546                         break;
2547                 default:
2548                         err = -EINVAL;
2549                 }
2550                 rtnl_unlock();
2551
2552                 return err;
2553         }
2554
2555         return -EINVAL;
2556 }
2557
2558 /*
2559  *      Drop the packet on the floor
2560  */
2561
2562 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
2563 {
2564         int type;
2565         struct dst_entry *dst = skb_dst(skb);
2566         switch (ipstats_mib_noroutes) {
2567         case IPSTATS_MIB_INNOROUTES:
2568                 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
2569                 if (type == IPV6_ADDR_ANY) {
2570                         IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2571                                       IPSTATS_MIB_INADDRERRORS);
2572                         break;
2573                 }
2574                 /* FALLTHROUGH */
2575         case IPSTATS_MIB_OUTNOROUTES:
2576                 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2577                               ipstats_mib_noroutes);
2578                 break;
2579         }
2580         icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
2581         kfree_skb(skb);
2582         return 0;
2583 }
2584
2585 static int ip6_pkt_discard(struct sk_buff *skb)
2586 {
2587         return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
2588 }
2589
2590 static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb)
2591 {
2592         skb->dev = skb_dst(skb)->dev;
2593         return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
2594 }
2595
2596 static int ip6_pkt_prohibit(struct sk_buff *skb)
2597 {
2598         return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
2599 }
2600
2601 static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb)
2602 {
2603         skb->dev = skb_dst(skb)->dev;
2604         return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
2605 }
2606
2607 /*
2608  *      Allocate a dst for local (unicast / anycast) address.
2609  */
2610
2611 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
2612                                     const struct in6_addr *addr,
2613                                     bool anycast)
2614 {
2615         u32 tb_id;
2616         struct net *net = dev_net(idev->dev);
2617         struct net_device *dev = net->loopback_dev;
2618         struct rt6_info *rt;
2619
2620         /* use L3 Master device as loopback for host routes if device
2621          * is enslaved and address is not link local or multicast
2622          */
2623         if (!rt6_need_strict(addr))
2624                 dev = l3mdev_master_dev_rcu(idev->dev) ? : dev;
2625
2626         rt = ip6_dst_alloc(net, dev, DST_NOCOUNT);
2627         if (!rt)
2628                 return ERR_PTR(-ENOMEM);
2629
2630         in6_dev_hold(idev);
2631
2632         rt->dst.flags |= DST_HOST;
2633         rt->dst.input = ip6_input;
2634         rt->dst.output = ip6_output;
2635         rt->rt6i_idev = idev;
2636
2637         rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
2638         if (anycast)
2639                 rt->rt6i_flags |= RTF_ANYCAST;
2640         else
2641                 rt->rt6i_flags |= RTF_LOCAL;
2642
2643         rt->rt6i_gateway  = *addr;
2644         rt->rt6i_dst.addr = *addr;
2645         rt->rt6i_dst.plen = 128;
2646         tb_id = l3mdev_fib_table(idev->dev) ? : RT6_TABLE_LOCAL;
2647         rt->rt6i_table = fib6_get_table(net, tb_id);
2648         rt->dst.flags |= DST_NOCACHE;
2649
2650         atomic_set(&rt->dst.__refcnt, 1);
2651
2652         return rt;
2653 }
2654
2655 /* remove deleted ip from prefsrc entries */
2656 struct arg_dev_net_ip {
2657         struct net_device *dev;
2658         struct net *net;
2659         struct in6_addr *addr;
2660 };
2661
2662 static int fib6_remove_prefsrc(struct rt6_info *rt, void *arg)
2663 {
2664         struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
2665         struct net *net = ((struct arg_dev_net_ip *)arg)->net;
2666         struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
2667
2668         if (((void *)rt->dst.dev == dev || !dev) &&
2669             rt != net->ipv6.ip6_null_entry &&
2670             ipv6_addr_equal(addr, &rt->rt6i_prefsrc.addr)) {
2671                 /* remove prefsrc entry */
2672                 rt->rt6i_prefsrc.plen = 0;
2673         }
2674         return 0;
2675 }
2676
2677 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
2678 {
2679         struct net *net = dev_net(ifp->idev->dev);
2680         struct arg_dev_net_ip adni = {
2681                 .dev = ifp->idev->dev,
2682                 .net = net,
2683                 .addr = &ifp->addr,
2684         };
2685         fib6_clean_all(net, fib6_remove_prefsrc, &adni);
2686 }
2687
2688 #define RTF_RA_ROUTER           (RTF_ADDRCONF | RTF_DEFAULT | RTF_GATEWAY)
2689 #define RTF_CACHE_GATEWAY       (RTF_GATEWAY | RTF_CACHE)
2690
2691 /* Remove routers and update dst entries when gateway turn into host. */
2692 static int fib6_clean_tohost(struct rt6_info *rt, void *arg)
2693 {
2694         struct in6_addr *gateway = (struct in6_addr *)arg;
2695
2696         if ((((rt->rt6i_flags & RTF_RA_ROUTER) == RTF_RA_ROUTER) ||
2697              ((rt->rt6i_flags & RTF_CACHE_GATEWAY) == RTF_CACHE_GATEWAY)) &&
2698              ipv6_addr_equal(gateway, &rt->rt6i_gateway)) {
2699                 return -1;
2700         }
2701         return 0;
2702 }
2703
2704 void rt6_clean_tohost(struct net *net, struct in6_addr *gateway)
2705 {
2706         fib6_clean_all(net, fib6_clean_tohost, gateway);
2707 }
2708
2709 struct arg_dev_net {
2710         struct net_device *dev;
2711         struct net *net;
2712 };
2713
2714 static int fib6_ifdown(struct rt6_info *rt, void *arg)
2715 {
2716         const struct arg_dev_net *adn = arg;
2717         const struct net_device *dev = adn->dev;
2718
2719         if ((rt->dst.dev == dev || !dev) &&
2720             rt != adn->net->ipv6.ip6_null_entry)
2721                 return -1;
2722
2723         return 0;
2724 }
2725
2726 void rt6_ifdown(struct net *net, struct net_device *dev)
2727 {
2728         struct arg_dev_net adn = {
2729                 .dev = dev,
2730                 .net = net,
2731         };
2732
2733         fib6_clean_all(net, fib6_ifdown, &adn);
2734         icmp6_clean_all(fib6_ifdown, &adn);
2735         if (dev)
2736                 rt6_uncached_list_flush_dev(net, dev);
2737 }
2738
2739 struct rt6_mtu_change_arg {
2740         struct net_device *dev;
2741         unsigned int mtu;
2742 };
2743
2744 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2745 {
2746         struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2747         struct inet6_dev *idev;
2748
2749         /* In IPv6 pmtu discovery is not optional,
2750            so that RTAX_MTU lock cannot disable it.
2751            We still use this lock to block changes
2752            caused by addrconf/ndisc.
2753         */
2754
2755         idev = __in6_dev_get(arg->dev);
2756         if (!idev)
2757                 return 0;
2758
2759         /* For administrative MTU increase, there is no way to discover
2760            IPv6 PMTU increase, so PMTU increase should be updated here.
2761            Since RFC 1981 doesn't include administrative MTU increase
2762            update PMTU increase is a MUST. (i.e. jumbo frame)
2763          */
2764         /*
2765            If new MTU is less than route PMTU, this new MTU will be the
2766            lowest MTU in the path, update the route PMTU to reflect PMTU
2767            decreases; if new MTU is greater than route PMTU, and the
2768            old MTU is the lowest MTU in the path, update the route PMTU
2769            to reflect the increase. In this case if the other nodes' MTU
2770            also have the lowest MTU, TOO BIG MESSAGE will be lead to
2771            PMTU discovery.
2772          */
2773         if (rt->dst.dev == arg->dev &&
2774             dst_metric_raw(&rt->dst, RTAX_MTU) &&
2775             !dst_metric_locked(&rt->dst, RTAX_MTU)) {
2776                 if (rt->rt6i_flags & RTF_CACHE) {
2777                         /* For RTF_CACHE with rt6i_pmtu == 0
2778                          * (i.e. a redirected route),
2779                          * the metrics of its rt->dst.from has already
2780                          * been updated.
2781                          */
2782                         if (rt->rt6i_pmtu && rt->rt6i_pmtu > arg->mtu)
2783                                 rt->rt6i_pmtu = arg->mtu;
2784                 } else if (dst_mtu(&rt->dst) >= arg->mtu ||
2785                            (dst_mtu(&rt->dst) < arg->mtu &&
2786                             dst_mtu(&rt->dst) == idev->cnf.mtu6)) {
2787                         dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
2788                 }
2789         }
2790         return 0;
2791 }
2792
2793 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
2794 {
2795         struct rt6_mtu_change_arg arg = {
2796                 .dev = dev,
2797                 .mtu = mtu,
2798         };
2799
2800         fib6_clean_all(dev_net(dev), rt6_mtu_change_route, &arg);
2801 }
2802
2803 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2804         [RTA_GATEWAY]           = { .len = sizeof(struct in6_addr) },
2805         [RTA_OIF]               = { .type = NLA_U32 },
2806         [RTA_IIF]               = { .type = NLA_U32 },
2807         [RTA_PRIORITY]          = { .type = NLA_U32 },
2808         [RTA_METRICS]           = { .type = NLA_NESTED },
2809         [RTA_MULTIPATH]         = { .len = sizeof(struct rtnexthop) },
2810         [RTA_PREF]              = { .type = NLA_U8 },
2811         [RTA_ENCAP_TYPE]        = { .type = NLA_U16 },
2812         [RTA_ENCAP]             = { .type = NLA_NESTED },
2813         [RTA_EXPIRES]           = { .type = NLA_U32 },
2814         [RTA_UID]               = { .type = NLA_U32 },
2815 };
2816
2817 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2818                               struct fib6_config *cfg)
2819 {
2820         struct rtmsg *rtm;
2821         struct nlattr *tb[RTA_MAX+1];
2822         unsigned int pref;
2823         int err;
2824
2825         err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2826         if (err < 0)
2827                 goto errout;
2828
2829         err = -EINVAL;
2830         rtm = nlmsg_data(nlh);
2831         memset(cfg, 0, sizeof(*cfg));
2832
2833         cfg->fc_table = rtm->rtm_table;
2834         cfg->fc_dst_len = rtm->rtm_dst_len;
2835         cfg->fc_src_len = rtm->rtm_src_len;
2836         cfg->fc_flags = RTF_UP;
2837         cfg->fc_protocol = rtm->rtm_protocol;
2838         cfg->fc_type = rtm->rtm_type;
2839
2840         if (rtm->rtm_type == RTN_UNREACHABLE ||
2841             rtm->rtm_type == RTN_BLACKHOLE ||
2842             rtm->rtm_type == RTN_PROHIBIT ||
2843             rtm->rtm_type == RTN_THROW)
2844                 cfg->fc_flags |= RTF_REJECT;
2845
2846         if (rtm->rtm_type == RTN_LOCAL)
2847                 cfg->fc_flags |= RTF_LOCAL;
2848
2849         if (rtm->rtm_flags & RTM_F_CLONED)
2850                 cfg->fc_flags |= RTF_CACHE;
2851
2852         cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid;
2853         cfg->fc_nlinfo.nlh = nlh;
2854         cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2855
2856         if (tb[RTA_GATEWAY]) {
2857                 cfg->fc_gateway = nla_get_in6_addr(tb[RTA_GATEWAY]);
2858                 cfg->fc_flags |= RTF_GATEWAY;
2859         }
2860
2861         if (tb[RTA_DST]) {
2862                 int plen = (rtm->rtm_dst_len + 7) >> 3;
2863
2864                 if (nla_len(tb[RTA_DST]) < plen)
2865                         goto errout;
2866
2867                 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2868         }
2869
2870         if (tb[RTA_SRC]) {
2871                 int plen = (rtm->rtm_src_len + 7) >> 3;
2872
2873                 if (nla_len(tb[RTA_SRC]) < plen)
2874                         goto errout;
2875
2876                 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2877         }
2878
2879         if (tb[RTA_PREFSRC])
2880                 cfg->fc_prefsrc = nla_get_in6_addr(tb[RTA_PREFSRC]);
2881
2882         if (tb[RTA_OIF])
2883                 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2884
2885         if (tb[RTA_PRIORITY])
2886                 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2887
2888         if (tb[RTA_METRICS]) {
2889                 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2890                 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2891         }
2892
2893         if (tb[RTA_TABLE])
2894                 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2895
2896         if (tb[RTA_MULTIPATH]) {
2897                 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
2898                 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
2899
2900                 err = lwtunnel_valid_encap_type_attr(cfg->fc_mp,
2901                                                      cfg->fc_mp_len);
2902                 if (err < 0)
2903                         goto errout;
2904         }
2905
2906         if (tb[RTA_PREF]) {
2907                 pref = nla_get_u8(tb[RTA_PREF]);
2908                 if (pref != ICMPV6_ROUTER_PREF_LOW &&
2909                     pref != ICMPV6_ROUTER_PREF_HIGH)
2910                         pref = ICMPV6_ROUTER_PREF_MEDIUM;
2911                 cfg->fc_flags |= RTF_PREF(pref);
2912         }
2913
2914         if (tb[RTA_ENCAP])
2915                 cfg->fc_encap = tb[RTA_ENCAP];
2916
2917         if (tb[RTA_ENCAP_TYPE]) {
2918                 cfg->fc_encap_type = nla_get_u16(tb[RTA_ENCAP_TYPE]);
2919
2920                 err = lwtunnel_valid_encap_type(cfg->fc_encap_type);
2921                 if (err < 0)
2922                         goto errout;
2923         }
2924
2925         if (tb[RTA_EXPIRES]) {
2926                 unsigned long timeout = addrconf_timeout_fixup(nla_get_u32(tb[RTA_EXPIRES]), HZ);
2927
2928                 if (addrconf_finite_timeout(timeout)) {
2929                         cfg->fc_expires = jiffies_to_clock_t(timeout * HZ);
2930                         cfg->fc_flags |= RTF_EXPIRES;
2931                 }
2932         }
2933
2934         err = 0;
2935 errout:
2936         return err;
2937 }
2938
2939 struct rt6_nh {
2940         struct rt6_info *rt6_info;
2941         struct fib6_config r_cfg;
2942         struct mx6_config mxc;
2943         struct list_head next;
2944 };
2945
2946 static void ip6_print_replace_route_err(struct list_head *rt6_nh_list)
2947 {
2948         struct rt6_nh *nh;
2949
2950         list_for_each_entry(nh, rt6_nh_list, next) {
2951                 pr_warn("IPV6: multipath route replace failed (check consistency of installed routes): %pI6 nexthop %pI6 ifi %d\n",
2952                         &nh->r_cfg.fc_dst, &nh->r_cfg.fc_gateway,
2953                         nh->r_cfg.fc_ifindex);
2954         }
2955 }
2956
2957 static int ip6_route_info_append(struct list_head *rt6_nh_list,
2958                                  struct rt6_info *rt, struct fib6_config *r_cfg)
2959 {
2960         struct rt6_nh *nh;
2961         struct rt6_info *rtnh;
2962         int err = -EEXIST;
2963
2964         list_for_each_entry(nh, rt6_nh_list, next) {
2965                 /* check if rt6_info already exists */
2966                 rtnh = nh->rt6_info;
2967
2968                 if (rtnh->dst.dev == rt->dst.dev &&
2969                     rtnh->rt6i_idev == rt->rt6i_idev &&
2970                     ipv6_addr_equal(&rtnh->rt6i_gateway,
2971                                     &rt->rt6i_gateway))
2972                         return err;
2973         }
2974
2975         nh = kzalloc(sizeof(*nh), GFP_KERNEL);
2976         if (!nh)
2977                 return -ENOMEM;
2978         nh->rt6_info = rt;
2979         err = ip6_convert_metrics(&nh->mxc, r_cfg);
2980         if (err) {
2981                 kfree(nh);
2982                 return err;
2983         }
2984         memcpy(&nh->r_cfg, r_cfg, sizeof(*r_cfg));
2985         list_add_tail(&nh->next, rt6_nh_list);
2986
2987         return 0;
2988 }
2989
2990 static int ip6_route_multipath_add(struct fib6_config *cfg)
2991 {
2992         struct fib6_config r_cfg;
2993         struct rtnexthop *rtnh;
2994         struct rt6_info *rt;
2995         struct rt6_nh *err_nh;
2996         struct rt6_nh *nh, *nh_safe;
2997         int remaining;
2998         int attrlen;
2999         int err = 1;
3000         int nhn = 0;
3001         int replace = (cfg->fc_nlinfo.nlh &&
3002                        (cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_REPLACE));
3003         LIST_HEAD(rt6_nh_list);
3004
3005         remaining = cfg->fc_mp_len;
3006         rtnh = (struct rtnexthop *)cfg->fc_mp;
3007
3008         /* Parse a Multipath Entry and build a list (rt6_nh_list) of
3009          * rt6_info structs per nexthop
3010          */
3011         while (rtnh_ok(rtnh, remaining)) {
3012                 memcpy(&r_cfg, cfg, sizeof(*cfg));
3013                 if (rtnh->rtnh_ifindex)
3014                         r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
3015
3016                 attrlen = rtnh_attrlen(rtnh);
3017                 if (attrlen > 0) {
3018                         struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
3019
3020                         nla = nla_find(attrs, attrlen, RTA_GATEWAY);
3021                         if (nla) {
3022                                 r_cfg.fc_gateway = nla_get_in6_addr(nla);
3023                                 r_cfg.fc_flags |= RTF_GATEWAY;
3024                         }
3025                         r_cfg.fc_encap = nla_find(attrs, attrlen, RTA_ENCAP);
3026                         nla = nla_find(attrs, attrlen, RTA_ENCAP_TYPE);
3027                         if (nla)
3028                                 r_cfg.fc_encap_type = nla_get_u16(nla);
3029                 }
3030
3031                 rt = ip6_route_info_create(&r_cfg);
3032                 if (IS_ERR(rt)) {
3033                         err = PTR_ERR(rt);
3034                         rt = NULL;
3035                         goto cleanup;
3036                 }
3037
3038                 err = ip6_route_info_append(&rt6_nh_list, rt, &r_cfg);
3039                 if (err) {
3040                         dst_free(&rt->dst);
3041                         goto cleanup;
3042                 }
3043
3044                 rtnh = rtnh_next(rtnh, &remaining);
3045         }
3046
3047         err_nh = NULL;
3048         list_for_each_entry(nh, &rt6_nh_list, next) {
3049                 err = __ip6_ins_rt(nh->rt6_info, &cfg->fc_nlinfo, &nh->mxc);
3050                 /* nh->rt6_info is used or freed at this point, reset to NULL*/
3051                 nh->rt6_info = NULL;
3052                 if (err) {
3053                         if (replace && nhn)
3054                                 ip6_print_replace_route_err(&rt6_nh_list);
3055                         err_nh = nh;
3056                         goto add_errout;
3057                 }
3058
3059                 /* Because each route is added like a single route we remove
3060                  * these flags after the first nexthop: if there is a collision,
3061                  * we have already failed to add the first nexthop:
3062                  * fib6_add_rt2node() has rejected it; when replacing, old
3063                  * nexthops have been replaced by first new, the rest should
3064                  * be added to it.
3065                  */
3066                 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~(NLM_F_EXCL |
3067                                                      NLM_F_REPLACE);
3068                 nhn++;
3069         }
3070
3071         goto cleanup;
3072
3073 add_errout:
3074         /* Delete routes that were already added */
3075         list_for_each_entry(nh, &rt6_nh_list, next) {
3076                 if (err_nh == nh)
3077                         break;
3078                 ip6_route_del(&nh->r_cfg);
3079         }
3080
3081 cleanup:
3082         list_for_each_entry_safe(nh, nh_safe, &rt6_nh_list, next) {
3083                 if (nh->rt6_info)
3084                         dst_free(&nh->rt6_info->dst);
3085                 kfree(nh->mxc.mx);
3086                 list_del(&nh->next);
3087                 kfree(nh);
3088         }
3089
3090         return err;
3091 }
3092
3093 static int ip6_route_multipath_del(struct fib6_config *cfg)
3094 {
3095         struct fib6_config r_cfg;
3096         struct rtnexthop *rtnh;
3097         int remaining;
3098         int attrlen;
3099         int err = 1, last_err = 0;
3100
3101         remaining = cfg->fc_mp_len;
3102         rtnh = (struct rtnexthop *)cfg->fc_mp;
3103
3104         /* Parse a Multipath Entry */
3105         while (rtnh_ok(rtnh, remaining)) {
3106                 memcpy(&r_cfg, cfg, sizeof(*cfg));
3107                 if (rtnh->rtnh_ifindex)
3108                         r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
3109
3110                 attrlen = rtnh_attrlen(rtnh);
3111                 if (attrlen > 0) {
3112                         struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
3113
3114                         nla = nla_find(attrs, attrlen, RTA_GATEWAY);
3115                         if (nla) {
3116                                 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
3117                                 r_cfg.fc_flags |= RTF_GATEWAY;
3118                         }
3119                 }
3120                 err = ip6_route_del(&r_cfg);
3121                 if (err)
3122                         last_err = err;
3123
3124                 rtnh = rtnh_next(rtnh, &remaining);
3125         }
3126
3127         return last_err;
3128 }
3129
3130 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh)
3131 {
3132         struct fib6_config cfg;
3133         int err;
3134
3135         err = rtm_to_fib6_config(skb, nlh, &cfg);
3136         if (err < 0)
3137                 return err;
3138
3139         if (cfg.fc_mp)
3140                 return ip6_route_multipath_del(&cfg);
3141         else
3142                 return ip6_route_del(&cfg);
3143 }
3144
3145 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh)
3146 {
3147         struct fib6_config cfg;
3148         int err;
3149
3150         err = rtm_to_fib6_config(skb, nlh, &cfg);
3151         if (err < 0)
3152                 return err;
3153
3154         if (cfg.fc_mp)
3155                 return ip6_route_multipath_add(&cfg);
3156         else
3157                 return ip6_route_add(&cfg);
3158 }
3159
3160 static inline size_t rt6_nlmsg_size(struct rt6_info *rt)
3161 {
3162         return NLMSG_ALIGN(sizeof(struct rtmsg))
3163                + nla_total_size(16) /* RTA_SRC */
3164                + nla_total_size(16) /* RTA_DST */
3165                + nla_total_size(16) /* RTA_GATEWAY */
3166                + nla_total_size(16) /* RTA_PREFSRC */
3167                + nla_total_size(4) /* RTA_TABLE */
3168                + nla_total_size(4) /* RTA_IIF */
3169                + nla_total_size(4) /* RTA_OIF */
3170                + nla_total_size(4) /* RTA_PRIORITY */
3171                + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
3172                + nla_total_size(sizeof(struct rta_cacheinfo))
3173                + nla_total_size(TCP_CA_NAME_MAX) /* RTAX_CC_ALGO */
3174                + nla_total_size(1) /* RTA_PREF */
3175                + lwtunnel_get_encap_size(rt->dst.lwtstate);
3176 }
3177
3178 static int rt6_fill_node(struct net *net,
3179                          struct sk_buff *skb, struct rt6_info *rt,
3180                          struct in6_addr *dst, struct in6_addr *src,
3181                          int iif, int type, u32 portid, u32 seq,
3182                          int prefix, int nowait, unsigned int flags)
3183 {
3184         u32 metrics[RTAX_MAX];
3185         struct rtmsg *rtm;
3186         struct nlmsghdr *nlh;
3187         long expires;
3188         u32 table;
3189
3190         if (prefix) {   /* user wants prefix routes only */
3191                 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
3192                         /* success since this is not a prefix route */
3193                         return 1;
3194                 }
3195         }
3196
3197         nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
3198         if (!nlh)
3199                 return -EMSGSIZE;
3200
3201         rtm = nlmsg_data(nlh);
3202         rtm->rtm_family = AF_INET6;
3203         rtm->rtm_dst_len = rt->rt6i_dst.plen;
3204         rtm->rtm_src_len = rt->rt6i_src.plen;
3205         rtm->rtm_tos = 0;
3206         if (rt->rt6i_table)
3207                 table = rt->rt6i_table->tb6_id;
3208         else
3209                 table = RT6_TABLE_UNSPEC;
3210         rtm->rtm_table = table;
3211         if (nla_put_u32(skb, RTA_TABLE, table))
3212                 goto nla_put_failure;
3213         if (rt->rt6i_flags & RTF_REJECT) {
3214                 switch (rt->dst.error) {
3215                 case -EINVAL:
3216                         rtm->rtm_type = RTN_BLACKHOLE;
3217                         break;
3218                 case -EACCES:
3219                         rtm->rtm_type = RTN_PROHIBIT;
3220                         break;
3221                 case -EAGAIN:
3222                         rtm->rtm_type = RTN_THROW;
3223                         break;
3224                 default:
3225                         rtm->rtm_type = RTN_UNREACHABLE;
3226                         break;
3227                 }
3228         }
3229         else if (rt->rt6i_flags & RTF_LOCAL)
3230                 rtm->rtm_type = RTN_LOCAL;
3231         else if (rt->dst.dev && (rt->dst.dev->flags & IFF_LOOPBACK))
3232                 rtm->rtm_type = RTN_LOCAL;
3233         else
3234                 rtm->rtm_type = RTN_UNICAST;
3235         rtm->rtm_flags = 0;
3236         if (!netif_carrier_ok(rt->dst.dev)) {
3237                 rtm->rtm_flags |= RTNH_F_LINKDOWN;
3238                 if (rt->rt6i_idev->cnf.ignore_routes_with_linkdown)
3239                         rtm->rtm_flags |= RTNH_F_DEAD;
3240         }
3241         rtm->rtm_scope = RT_SCOPE_UNIVERSE;
3242         rtm->rtm_protocol = rt->rt6i_protocol;
3243         if (rt->rt6i_flags & RTF_DYNAMIC)
3244                 rtm->rtm_protocol = RTPROT_REDIRECT;
3245         else if (rt->rt6i_flags & RTF_ADDRCONF) {
3246                 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ROUTEINFO))
3247                         rtm->rtm_protocol = RTPROT_RA;
3248                 else
3249                         rtm->rtm_protocol = RTPROT_KERNEL;
3250         }
3251
3252         if (rt->rt6i_flags & RTF_CACHE)
3253                 rtm->rtm_flags |= RTM_F_CLONED;
3254
3255         if (dst) {
3256                 if (nla_put_in6_addr(skb, RTA_DST, dst))
3257                         goto nla_put_failure;
3258                 rtm->rtm_dst_len = 128;
3259         } else if (rtm->rtm_dst_len)
3260                 if (nla_put_in6_addr(skb, RTA_DST, &rt->rt6i_dst.addr))
3261                         goto nla_put_failure;
3262 #ifdef CONFIG_IPV6_SUBTREES
3263         if (src) {
3264                 if (nla_put_in6_addr(skb, RTA_SRC, src))
3265                         goto nla_put_failure;
3266                 rtm->rtm_src_len = 128;
3267         } else if (rtm->rtm_src_len &&
3268                    nla_put_in6_addr(skb, RTA_SRC, &rt->rt6i_src.addr))
3269                 goto nla_put_failure;
3270 #endif
3271         if (iif) {
3272 #ifdef CONFIG_IPV6_MROUTE
3273                 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
3274                         int err = ip6mr_get_route(net, skb, rtm, nowait,
3275                                                   portid);
3276
3277                         if (err <= 0) {
3278                                 if (!nowait) {
3279                                         if (err == 0)
3280                                                 return 0;
3281                                         goto nla_put_failure;
3282                                 } else {
3283                                         if (err == -EMSGSIZE)
3284                                                 goto nla_put_failure;
3285                                 }
3286                         }
3287                 } else
3288 #endif
3289                         if (nla_put_u32(skb, RTA_IIF, iif))
3290                                 goto nla_put_failure;
3291         } else if (dst) {
3292                 struct in6_addr saddr_buf;
3293                 if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0 &&
3294                     nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
3295                         goto nla_put_failure;
3296         }
3297
3298         if (rt->rt6i_prefsrc.plen) {
3299                 struct in6_addr saddr_buf;
3300                 saddr_buf = rt->rt6i_prefsrc.addr;
3301                 if (nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
3302                         goto nla_put_failure;
3303         }
3304
3305         memcpy(metrics, dst_metrics_ptr(&rt->dst), sizeof(metrics));
3306         if (rt->rt6i_pmtu)
3307                 metrics[RTAX_MTU - 1] = rt->rt6i_pmtu;
3308         if (rtnetlink_put_metrics(skb, metrics) < 0)
3309                 goto nla_put_failure;
3310
3311         if (rt->rt6i_flags & RTF_GATEWAY) {
3312                 if (nla_put_in6_addr(skb, RTA_GATEWAY, &rt->rt6i_gateway) < 0)
3313                         goto nla_put_failure;
3314         }
3315
3316         if (rt->dst.dev &&
3317             nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex))
3318                 goto nla_put_failure;
3319         if (nla_put_u32(skb, RTA_PRIORITY, rt->rt6i_metric))
3320                 goto nla_put_failure;
3321
3322         expires = (rt->rt6i_flags & RTF_EXPIRES) ? rt->dst.expires - jiffies : 0;
3323
3324         if (rtnl_put_cacheinfo(skb, &rt->dst, 0, expires, rt->dst.error) < 0)
3325                 goto nla_put_failure;
3326
3327         if (nla_put_u8(skb, RTA_PREF, IPV6_EXTRACT_PREF(rt->rt6i_flags)))
3328                 goto nla_put_failure;
3329
3330         if (lwtunnel_fill_encap(skb, rt->dst.lwtstate) < 0)
3331                 goto nla_put_failure;
3332
3333         nlmsg_end(skb, nlh);
3334         return 0;
3335
3336 nla_put_failure:
3337         nlmsg_cancel(skb, nlh);
3338         return -EMSGSIZE;
3339 }
3340
3341 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
3342 {
3343         struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
3344         int prefix;
3345
3346         if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
3347                 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
3348                 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
3349         } else
3350                 prefix = 0;
3351
3352         return rt6_fill_node(arg->net,
3353                      arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
3354                      NETLINK_CB(arg->cb->skb).portid, arg->cb->nlh->nlmsg_seq,
3355                      prefix, 0, NLM_F_MULTI);
3356 }
3357
3358 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh)
3359 {
3360         struct net *net = sock_net(in_skb->sk);
3361         struct nlattr *tb[RTA_MAX+1];
3362         struct rt6_info *rt;
3363         struct sk_buff *skb;
3364         struct rtmsg *rtm;
3365         struct flowi6 fl6;
3366         int err, iif = 0, oif = 0;
3367
3368         err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
3369         if (err < 0)
3370                 goto errout;
3371
3372         err = -EINVAL;
3373         memset(&fl6, 0, sizeof(fl6));
3374         rtm = nlmsg_data(nlh);
3375         fl6.flowlabel = ip6_make_flowinfo(rtm->rtm_tos, 0);
3376
3377         if (tb[RTA_SRC]) {
3378                 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
3379                         goto errout;
3380
3381                 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
3382         }
3383
3384         if (tb[RTA_DST]) {
3385                 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
3386                         goto errout;
3387
3388                 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
3389         }
3390
3391         if (tb[RTA_IIF])
3392                 iif = nla_get_u32(tb[RTA_IIF]);
3393
3394         if (tb[RTA_OIF])
3395                 oif = nla_get_u32(tb[RTA_OIF]);
3396
3397         if (tb[RTA_MARK])
3398                 fl6.flowi6_mark = nla_get_u32(tb[RTA_MARK]);
3399
3400         if (tb[RTA_UID])
3401                 fl6.flowi6_uid = make_kuid(current_user_ns(),
3402                                            nla_get_u32(tb[RTA_UID]));
3403         else
3404                 fl6.flowi6_uid = iif ? INVALID_UID : current_uid();
3405
3406         if (iif) {
3407                 struct net_device *dev;
3408                 int flags = 0;
3409
3410                 dev = __dev_get_by_index(net, iif);
3411                 if (!dev) {
3412                         err = -ENODEV;
3413                         goto errout;
3414                 }
3415
3416                 fl6.flowi6_iif = iif;
3417
3418                 if (!ipv6_addr_any(&fl6.saddr))
3419                         flags |= RT6_LOOKUP_F_HAS_SADDR;
3420
3421                 rt = (struct rt6_info *)ip6_route_input_lookup(net, dev, &fl6,
3422                                                                flags);
3423         } else {
3424                 fl6.flowi6_oif = oif;
3425
3426                 rt = (struct rt6_info *)ip6_route_output(net, NULL, &fl6);
3427         }
3428
3429         skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
3430         if (!skb) {
3431                 ip6_rt_put(rt);
3432                 err = -ENOBUFS;
3433                 goto errout;
3434         }
3435
3436         /* Reserve room for dummy headers, this skb can pass
3437            through good chunk of routing engine.
3438          */
3439         skb_reset_mac_header(skb);
3440         skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
3441
3442         skb_dst_set(skb, &rt->dst);
3443
3444         err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
3445                             RTM_NEWROUTE, NETLINK_CB(in_skb).portid,
3446                             nlh->nlmsg_seq, 0, 0, 0);
3447         if (err < 0) {
3448                 kfree_skb(skb);
3449                 goto errout;
3450         }
3451
3452         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
3453 errout:
3454         return err;
3455 }
3456
3457 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info,
3458                      unsigned int nlm_flags)
3459 {
3460         struct sk_buff *skb;
3461         struct net *net = info->nl_net;
3462         u32 seq;
3463         int err;
3464
3465         err = -ENOBUFS;
3466         seq = info->nlh ? info->nlh->nlmsg_seq : 0;
3467
3468         skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
3469         if (!skb)
3470                 goto errout;
3471
3472         err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
3473                                 event, info->portid, seq, 0, 0, nlm_flags);
3474         if (err < 0) {
3475                 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
3476                 WARN_ON(err == -EMSGSIZE);
3477                 kfree_skb(skb);
3478                 goto errout;
3479         }
3480         rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
3481                     info->nlh, gfp_any());
3482         return;
3483 errout:
3484         if (err < 0)
3485                 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
3486 }
3487
3488 static int ip6_route_dev_notify(struct notifier_block *this,
3489                                 unsigned long event, void *ptr)
3490 {
3491         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3492         struct net *net = dev_net(dev);
3493
3494         if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
3495                 net->ipv6.ip6_null_entry->dst.dev = dev;
3496                 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
3497 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3498                 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
3499                 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
3500                 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
3501                 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
3502 #endif
3503         }
3504
3505         return NOTIFY_OK;
3506 }
3507
3508 /*
3509  *      /proc
3510  */
3511
3512 #ifdef CONFIG_PROC_FS
3513
3514 static const struct file_operations ipv6_route_proc_fops = {
3515         .owner          = THIS_MODULE,
3516         .open           = ipv6_route_open,
3517         .read           = seq_read,
3518         .llseek         = seq_lseek,
3519         .release        = seq_release_net,
3520 };
3521
3522 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
3523 {
3524         struct net *net = (struct net *)seq->private;
3525         seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
3526                    net->ipv6.rt6_stats->fib_nodes,
3527                    net->ipv6.rt6_stats->fib_route_nodes,
3528                    net->ipv6.rt6_stats->fib_rt_alloc,
3529                    net->ipv6.rt6_stats->fib_rt_entries,
3530                    net->ipv6.rt6_stats->fib_rt_cache,
3531                    dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
3532                    net->ipv6.rt6_stats->fib_discarded_routes);
3533
3534         return 0;
3535 }
3536
3537 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
3538 {
3539         return single_open_net(inode, file, rt6_stats_seq_show);
3540 }
3541
3542 static const struct file_operations rt6_stats_seq_fops = {
3543         .owner   = THIS_MODULE,
3544         .open    = rt6_stats_seq_open,
3545         .read    = seq_read,
3546         .llseek  = seq_lseek,
3547         .release = single_release_net,
3548 };
3549 #endif  /* CONFIG_PROC_FS */
3550
3551 #ifdef CONFIG_SYSCTL
3552
3553 static
3554 int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
3555                               void __user *buffer, size_t *lenp, loff_t *ppos)
3556 {
3557         struct net *net;
3558         int delay;
3559         if (!write)
3560                 return -EINVAL;
3561
3562         net = (struct net *)ctl->extra1;
3563         delay = net->ipv6.sysctl.flush_delay;
3564         proc_dointvec(ctl, write, buffer, lenp, ppos);
3565         fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
3566         return 0;
3567 }
3568
3569 struct ctl_table ipv6_route_table_template[] = {
3570         {
3571                 .procname       =       "flush",
3572                 .data           =       &init_net.ipv6.sysctl.flush_delay,
3573                 .maxlen         =       sizeof(int),
3574                 .mode           =       0200,
3575                 .proc_handler   =       ipv6_sysctl_rtcache_flush
3576         },
3577         {
3578                 .procname       =       "gc_thresh",
3579                 .data           =       &ip6_dst_ops_template.gc_thresh,
3580                 .maxlen         =       sizeof(int),
3581                 .mode           =       0644,
3582                 .proc_handler   =       proc_dointvec,
3583         },
3584         {
3585                 .procname       =       "max_size",
3586                 .data           =       &init_net.ipv6.sysctl.ip6_rt_max_size,
3587                 .maxlen         =       sizeof(int),
3588                 .mode           =       0644,
3589                 .proc_handler   =       proc_dointvec,
3590         },
3591         {
3592                 .procname       =       "gc_min_interval",
3593                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
3594                 .maxlen         =       sizeof(int),
3595                 .mode           =       0644,
3596                 .proc_handler   =       proc_dointvec_jiffies,
3597         },
3598         {
3599                 .procname       =       "gc_timeout",
3600                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
3601                 .maxlen         =       sizeof(int),
3602                 .mode           =       0644,
3603                 .proc_handler   =       proc_dointvec_jiffies,
3604         },
3605         {
3606                 .procname       =       "gc_interval",
3607                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_interval,
3608                 .maxlen         =       sizeof(int),
3609                 .mode           =       0644,
3610                 .proc_handler   =       proc_dointvec_jiffies,
3611         },
3612         {
3613                 .procname       =       "gc_elasticity",
3614                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
3615                 .maxlen         =       sizeof(int),
3616                 .mode           =       0644,
3617                 .proc_handler   =       proc_dointvec,
3618         },
3619         {
3620                 .procname       =       "mtu_expires",
3621                 .data           =       &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
3622                 .maxlen         =       sizeof(int),
3623                 .mode           =       0644,
3624                 .proc_handler   =       proc_dointvec_jiffies,
3625         },
3626         {
3627                 .procname       =       "min_adv_mss",
3628                 .data           =       &init_net.ipv6.sysctl.ip6_rt_min_advmss,
3629                 .maxlen         =       sizeof(int),
3630                 .mode           =       0644,
3631                 .proc_handler   =       proc_dointvec,
3632         },
3633         {
3634                 .procname       =       "gc_min_interval_ms",
3635                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
3636                 .maxlen         =       sizeof(int),
3637                 .mode           =       0644,
3638                 .proc_handler   =       proc_dointvec_ms_jiffies,
3639         },
3640         { }
3641 };
3642
3643 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
3644 {
3645         struct ctl_table *table;
3646
3647         table = kmemdup(ipv6_route_table_template,
3648                         sizeof(ipv6_route_table_template),
3649                         GFP_KERNEL);
3650
3651         if (table) {
3652                 table[0].data = &net->ipv6.sysctl.flush_delay;
3653                 table[0].extra1 = net;
3654                 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
3655                 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
3656                 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
3657                 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
3658                 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
3659                 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
3660                 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
3661                 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
3662                 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
3663
3664                 /* Don't export sysctls to unprivileged users */
3665                 if (net->user_ns != &init_user_ns)
3666                         table[0].procname = NULL;
3667         }
3668
3669         return table;
3670 }
3671 #endif
3672
3673 static int __net_init ip6_route_net_init(struct net *net)
3674 {
3675         int ret = -ENOMEM;
3676
3677         memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
3678                sizeof(net->ipv6.ip6_dst_ops));
3679
3680         if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
3681                 goto out_ip6_dst_ops;
3682
3683         net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
3684                                            sizeof(*net->ipv6.ip6_null_entry),
3685                                            GFP_KERNEL);
3686         if (!net->ipv6.ip6_null_entry)
3687                 goto out_ip6_dst_entries;
3688         net->ipv6.ip6_null_entry->dst.path =
3689                 (struct dst_entry *)net->ipv6.ip6_null_entry;
3690         net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3691         dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
3692                          ip6_template_metrics, true);
3693
3694 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3695         net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
3696                                                sizeof(*net->ipv6.ip6_prohibit_entry),
3697                                                GFP_KERNEL);
3698         if (!net->ipv6.ip6_prohibit_entry)
3699                 goto out_ip6_null_entry;
3700         net->ipv6.ip6_prohibit_entry->dst.path =
3701                 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
3702         net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3703         dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
3704                          ip6_template_metrics, true);
3705
3706         net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
3707                                                sizeof(*net->ipv6.ip6_blk_hole_entry),
3708                                                GFP_KERNEL);
3709         if (!net->ipv6.ip6_blk_hole_entry)
3710                 goto out_ip6_prohibit_entry;
3711         net->ipv6.ip6_blk_hole_entry->dst.path =
3712                 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
3713         net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3714         dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
3715                          ip6_template_metrics, true);
3716 #endif
3717
3718         net->ipv6.sysctl.flush_delay = 0;
3719         net->ipv6.sysctl.ip6_rt_max_size = 4096;
3720         net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
3721         net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
3722         net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
3723         net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
3724         net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
3725         net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
3726
3727         net->ipv6.ip6_rt_gc_expire = 30*HZ;
3728
3729         ret = 0;
3730 out:
3731         return ret;
3732
3733 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3734 out_ip6_prohibit_entry:
3735         kfree(net->ipv6.ip6_prohibit_entry);
3736 out_ip6_null_entry:
3737         kfree(net->ipv6.ip6_null_entry);
3738 #endif
3739 out_ip6_dst_entries:
3740         dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3741 out_ip6_dst_ops:
3742         goto out;
3743 }
3744
3745 static void __net_exit ip6_route_net_exit(struct net *net)
3746 {
3747         kfree(net->ipv6.ip6_null_entry);
3748 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3749         kfree(net->ipv6.ip6_prohibit_entry);
3750         kfree(net->ipv6.ip6_blk_hole_entry);
3751 #endif
3752         dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3753 }
3754
3755 static int __net_init ip6_route_net_init_late(struct net *net)
3756 {
3757 #ifdef CONFIG_PROC_FS
3758         proc_create("ipv6_route", 0, net->proc_net, &ipv6_route_proc_fops);
3759         proc_create("rt6_stats", S_IRUGO, net->proc_net, &rt6_stats_seq_fops);
3760 #endif
3761         return 0;
3762 }
3763
3764 static void __net_exit ip6_route_net_exit_late(struct net *net)
3765 {
3766 #ifdef CONFIG_PROC_FS
3767         remove_proc_entry("ipv6_route", net->proc_net);
3768         remove_proc_entry("rt6_stats", net->proc_net);
3769 #endif
3770 }
3771
3772 static struct pernet_operations ip6_route_net_ops = {
3773         .init = ip6_route_net_init,
3774         .exit = ip6_route_net_exit,
3775 };
3776
3777 static int __net_init ipv6_inetpeer_init(struct net *net)
3778 {
3779         struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
3780
3781         if (!bp)
3782                 return -ENOMEM;
3783         inet_peer_base_init(bp);
3784         net->ipv6.peers = bp;
3785         return 0;
3786 }
3787
3788 static void __net_exit ipv6_inetpeer_exit(struct net *net)
3789 {
3790         struct inet_peer_base *bp = net->ipv6.peers;
3791
3792         net->ipv6.peers = NULL;
3793         inetpeer_invalidate_tree(bp);
3794         kfree(bp);
3795 }
3796
3797 static struct pernet_operations ipv6_inetpeer_ops = {
3798         .init   =       ipv6_inetpeer_init,
3799         .exit   =       ipv6_inetpeer_exit,
3800 };
3801
3802 static struct pernet_operations ip6_route_net_late_ops = {
3803         .init = ip6_route_net_init_late,
3804         .exit = ip6_route_net_exit_late,
3805 };
3806
3807 static struct notifier_block ip6_route_dev_notifier = {
3808         .notifier_call = ip6_route_dev_notify,
3809         .priority = 0,
3810 };
3811
3812 int __init ip6_route_init(void)
3813 {
3814         int ret;
3815         int cpu;
3816
3817         ret = -ENOMEM;
3818         ip6_dst_ops_template.kmem_cachep =
3819                 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
3820                                   SLAB_HWCACHE_ALIGN, NULL);
3821         if (!ip6_dst_ops_template.kmem_cachep)
3822                 goto out;
3823
3824         ret = dst_entries_init(&ip6_dst_blackhole_ops);
3825         if (ret)
3826                 goto out_kmem_cache;
3827
3828         ret = register_pernet_subsys(&ipv6_inetpeer_ops);
3829         if (ret)
3830                 goto out_dst_entries;
3831
3832         ret = register_pernet_subsys(&ip6_route_net_ops);
3833         if (ret)
3834                 goto out_register_inetpeer;
3835
3836         ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
3837
3838         /* Registering of the loopback is done before this portion of code,
3839          * the loopback reference in rt6_info will not be taken, do it
3840          * manually for init_net */
3841         init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
3842         init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3843   #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3844         init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
3845         init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3846         init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
3847         init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3848   #endif
3849         ret = fib6_init();
3850         if (ret)
3851                 goto out_register_subsys;
3852
3853         ret = xfrm6_init();
3854         if (ret)
3855                 goto out_fib6_init;
3856
3857         ret = fib6_rules_init();
3858         if (ret)
3859                 goto xfrm6_init;
3860
3861         ret = register_pernet_subsys(&ip6_route_net_late_ops);
3862         if (ret)
3863                 goto fib6_rules_init;
3864
3865         ret = -ENOBUFS;
3866         if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL, NULL) ||
3867             __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL, NULL) ||
3868             __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL, NULL))
3869                 goto out_register_late_subsys;
3870
3871         ret = register_netdevice_notifier(&ip6_route_dev_notifier);
3872         if (ret)
3873                 goto out_register_late_subsys;
3874
3875         for_each_possible_cpu(cpu) {
3876                 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
3877
3878                 INIT_LIST_HEAD(&ul->head);
3879                 spin_lock_init(&ul->lock);
3880         }
3881
3882 out:
3883         return ret;
3884
3885 out_register_late_subsys:
3886         unregister_pernet_subsys(&ip6_route_net_late_ops);
3887 fib6_rules_init:
3888         fib6_rules_cleanup();
3889 xfrm6_init:
3890         xfrm6_fini();
3891 out_fib6_init:
3892         fib6_gc_cleanup();
3893 out_register_subsys:
3894         unregister_pernet_subsys(&ip6_route_net_ops);
3895 out_register_inetpeer:
3896         unregister_pernet_subsys(&ipv6_inetpeer_ops);
3897 out_dst_entries:
3898         dst_entries_destroy(&ip6_dst_blackhole_ops);
3899 out_kmem_cache:
3900         kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3901         goto out;
3902 }
3903
3904 void ip6_route_cleanup(void)
3905 {
3906         unregister_netdevice_notifier(&ip6_route_dev_notifier);
3907         unregister_pernet_subsys(&ip6_route_net_late_ops);
3908         fib6_rules_cleanup();
3909         xfrm6_fini();
3910         fib6_gc_cleanup();
3911         unregister_pernet_subsys(&ipv6_inetpeer_ops);
3912         unregister_pernet_subsys(&ip6_route_net_ops);
3913         dst_entries_destroy(&ip6_dst_blackhole_ops);
3914         kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3915 }