net/ipv6/route.c

   1 /*
   2  *      Linux INET6 implementation
   3  *      FIB front-end.
   4  *
   5  *      Authors:
   6  *      Pedro Roque             <roque@di.fc.ul.pt>
   7  *
   8  *      This program is free software; you can redistribute it and/or
   9  *      modify it under the terms of the GNU General Public License
  10  *      as published by the Free Software Foundation; either version
  11  *      2 of the License, or (at your option) any later version.
  12  */
  13
  14 /*      Changes:
  15  *
  16  *      YOSHIFUJI Hideaki @USAGI
  17  *              reworked default router selection.
  18  *              - respect outgoing interface
  19  *              - select from (probably) reachable routers (i.e.
  20  *              routers in REACHABLE, STALE, DELAY or PROBE states).
  21  *              - always select the same router if it is (probably)
  22  *              reachable.  otherwise, round-robin the list.
  23  *      Ville Nuorvala
  24  *              Fixed routing subtrees.
  25  */
  26
  27 #define pr_fmt(fmt) "IPv6: " fmt
  28
  29 #include <linux/capability.h>
  30 #include <linux/errno.h>
  31 #include <linux/export.h>
  32 #include <linux/types.h>
  33 #include <linux/times.h>
  34 #include <linux/socket.h>
  35 #include <linux/sockios.h>
  36 #include <linux/net.h>
  37 #include <linux/route.h>
  38 #include <linux/netdevice.h>
  39 #include <linux/in6.h>
  40 #include <linux/mroute6.h>
  41 #include <linux/init.h>
  42 #include <linux/if_arp.h>
  43 #include <linux/proc_fs.h>
  44 #include <linux/seq_file.h>
  45 #include <linux/nsproxy.h>
  46 #include <linux/slab.h>
  47 #include <net/net_namespace.h>
  48 #include <net/snmp.h>
  49 #include <net/ipv6.h>
  50 #include <net/ip6_fib.h>
  51 #include <net/ip6_route.h>
  52 #include <net/ndisc.h>
  53 #include <net/addrconf.h>
  54 #include <net/tcp.h>
  55 #include <linux/rtnetlink.h>
  56 #include <net/dst.h>
  57 #include <net/dst_metadata.h>
  58 #include <net/xfrm.h>
  59 #include <net/netevent.h>
  60 #include <net/netlink.h>
  61 #include <net/nexthop.h>
  62 #include <net/lwtunnel.h>
  63 #include <net/ip_tunnels.h>
  64 #include <net/l3mdev.h>
  65 #include <trace/events/fib6.h>
  66
  67 #include <asm/uaccess.h>
  68
  69 #ifdef CONFIG_SYSCTL
  70 #include <linux/sysctl.h>
  71 #endif
  72
  73 enum rt6_nud_state {
  74         RT6_NUD_FAIL_HARD = -3,
  75         RT6_NUD_FAIL_PROBE = -2,
  76         RT6_NUD_FAIL_DO_RR = -1,
  77         RT6_NUD_SUCCEED = 1
  78 };
  79
  80 static void ip6_rt_copy_init(struct rt6_info *rt, struct rt6_info *ort);
  81 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie);
  82 static unsigned int      ip6_default_advmss(const struct dst_entry *dst);
  83 static unsigned int      ip6_mtu(const struct dst_entry *dst);
  84 static struct dst_entry *ip6_negative_advice(struct dst_entry *);
  85 static void             ip6_dst_destroy(struct dst_entry *);
  86 static void             ip6_dst_ifdown(struct dst_entry *,
  87                                        struct net_device *dev, int how);
  88 static int               ip6_dst_gc(struct dst_ops *ops);
  89
  90 static int              ip6_pkt_discard(struct sk_buff *skb);
  91 static int              ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb);
  92 static int              ip6_pkt_prohibit(struct sk_buff *skb);
  93 static int              ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb);
  94 static void             ip6_link_failure(struct sk_buff *skb);
  95 static void             ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
  96                                            struct sk_buff *skb, u32 mtu);
  97 static void             rt6_do_redirect(struct dst_entry *dst, struct sock *sk,
  98                                         struct sk_buff *skb);
  99 static void             rt6_dst_from_metrics_check(struct rt6_info *rt);
 100 static int rt6_score_route(struct rt6_info *rt, int oif, int strict);
 101
 102 #ifdef CONFIG_IPV6_ROUTE_INFO
 103 static struct rt6_info *rt6_add_route_info(struct net *net,
 104                                            const struct in6_addr *prefix, int prefixlen,
 105                                            const struct in6_addr *gwaddr, int ifindex,
 106                                            unsigned int pref);
 107 static struct rt6_info *rt6_get_route_info(struct net *net,
 108                                            const struct in6_addr *prefix, int prefixlen,
 109                                            const struct in6_addr *gwaddr, int ifindex);
 110 #endif
 111
 112 struct uncached_list {
 113         spinlock_t              lock;
 114         struct list_head        head;
 115 };
 116
 117 static DEFINE_PER_CPU_ALIGNED(struct uncached_list, rt6_uncached_list);
 118
 119 static void rt6_uncached_list_add(struct rt6_info *rt)
 120 {
 121         struct uncached_list *ul = raw_cpu_ptr(&rt6_uncached_list);
 122
 123         rt->dst.flags |= DST_NOCACHE;
 124         rt->rt6i_uncached_list = ul;
 125
 126         spin_lock_bh(&ul->lock);
 127         list_add_tail(&rt->rt6i_uncached, &ul->head);
 128         spin_unlock_bh(&ul->lock);
 129 }
 130
 131 static void rt6_uncached_list_del(struct rt6_info *rt)
 132 {
 133         if (!list_empty(&rt->rt6i_uncached)) {
 134                 struct uncached_list *ul = rt->rt6i_uncached_list;
 135
 136                 spin_lock_bh(&ul->lock);
 137                 list_del(&rt->rt6i_uncached);
 138                 spin_unlock_bh(&ul->lock);
 139         }
 140 }
 141
 142 static void rt6_uncached_list_flush_dev(struct net *net, struct net_device *dev)
 143 {
 144         struct net_device *loopback_dev = net->loopback_dev;
 145         int cpu;
 146
 147         if (dev == loopback_dev)
 148                 return;
 149
 150         for_each_possible_cpu(cpu) {
 151                 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
 152                 struct rt6_info *rt;
 153
 154                 spin_lock_bh(&ul->lock);
 155                 list_for_each_entry(rt, &ul->head, rt6i_uncached) {
 156                         struct inet6_dev *rt_idev = rt->rt6i_idev;
 157                         struct net_device *rt_dev = rt->dst.dev;
 158
 159                         if (rt_idev->dev == dev) {
 160                                 rt->rt6i_idev = in6_dev_get(loopback_dev);
 161                                 in6_dev_put(rt_idev);
 162                         }
 163
 164                         if (rt_dev == dev) {
 165                                 rt->dst.dev = loopback_dev;
 166                                 dev_hold(rt->dst.dev);
 167                                 dev_put(rt_dev);
 168                         }
 169                 }
 170                 spin_unlock_bh(&ul->lock);
 171         }
 172 }
 173
 174 static u32 *rt6_pcpu_cow_metrics(struct rt6_info *rt)
 175 {
 176         return dst_metrics_write_ptr(rt->dst.from);
 177 }
 178
 179 static u32 *ipv6_cow_metrics(struct dst_entry *dst, unsigned long old)
 180 {
 181         struct rt6_info *rt = (struct rt6_info *)dst;
 182
 183         if (rt->rt6i_flags & RTF_PCPU)
 184                 return rt6_pcpu_cow_metrics(rt);
 185         else if (rt->rt6i_flags & RTF_CACHE)
 186                 return NULL;
 187         else
 188                 return dst_cow_metrics_generic(dst, old);
 189 }
 190
 191 static inline const void *choose_neigh_daddr(struct rt6_info *rt,
 192                                              struct sk_buff *skb,
 193                                              const void *daddr)
 194 {
 195         struct in6_addr *p = &rt->rt6i_gateway;
 196
 197         if (!ipv6_addr_any(p))
 198                 return (const void *) p;
 199         else if (skb)
 200                 return &ipv6_hdr(skb)->daddr;
 201         return daddr;
 202 }
 203
 204 static struct neighbour *ip6_neigh_lookup(const struct dst_entry *dst,
 205                                           struct sk_buff *skb,
 206                                           const void *daddr)
 207 {
 208         struct rt6_info *rt = (struct rt6_info *) dst;
 209         struct neighbour *n;
 210
 211         daddr = choose_neigh_daddr(rt, skb, daddr);
 212         n = __ipv6_neigh_lookup(dst->dev, daddr);
 213         if (n)
 214                 return n;
 215         return neigh_create(&nd_tbl, daddr, dst->dev);
 216 }
 217
 218 static struct dst_ops ip6_dst_ops_template = {
 219         .family                 =       AF_INET6,
 220         .gc                     =       ip6_dst_gc,
 221         .gc_thresh              =       1024,
 222         .check                  =       ip6_dst_check,
 223         .default_advmss         =       ip6_default_advmss,
 224         .mtu                    =       ip6_mtu,
 225         .cow_metrics            =       ipv6_cow_metrics,
 226         .destroy                =       ip6_dst_destroy,
 227         .ifdown                 =       ip6_dst_ifdown,
 228         .negative_advice        =       ip6_negative_advice,
 229         .link_failure           =       ip6_link_failure,
 230         .update_pmtu            =       ip6_rt_update_pmtu,
 231         .redirect               =       rt6_do_redirect,
 232         .local_out              =       __ip6_local_out,
 233         .neigh_lookup           =       ip6_neigh_lookup,
 234 };
 235
 236 static unsigned int ip6_blackhole_mtu(const struct dst_entry *dst)
 237 {
 238         unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
 239
 240         return mtu ? : dst->dev->mtu;
 241 }
 242
 243 static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, struct sock *sk,
 244                                          struct sk_buff *skb, u32 mtu)
 245 {
 246 }
 247
 248 static void ip6_rt_blackhole_redirect(struct dst_entry *dst, struct sock *sk,
 249                                       struct sk_buff *skb)
 250 {
 251 }
 252
 253 static struct dst_ops ip6_dst_blackhole_ops = {
 254         .family                 =       AF_INET6,
 255         .destroy                =       ip6_dst_destroy,
 256         .check                  =       ip6_dst_check,
 257         .mtu                    =       ip6_blackhole_mtu,
 258         .default_advmss         =       ip6_default_advmss,
 259         .update_pmtu            =       ip6_rt_blackhole_update_pmtu,
 260         .redirect               =       ip6_rt_blackhole_redirect,
 261         .cow_metrics            =       dst_cow_metrics_generic,
 262         .neigh_lookup           =       ip6_neigh_lookup,
 263 };
 264
 265 static const u32 ip6_template_metrics[RTAX_MAX] = {
 266         [RTAX_HOPLIMIT - 1] = 0,
 267 };
 268
 269 static const struct rt6_info ip6_null_entry_template = {
 270         .dst = {
 271                 .__refcnt       = ATOMIC_INIT(1),
 272                 .__use          = 1,
 273                 .obsolete       = DST_OBSOLETE_FORCE_CHK,
 274                 .error          = -ENETUNREACH,
 275                 .input          = ip6_pkt_discard,
 276                 .output         = ip6_pkt_discard_out,
 277         },
 278         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
 279         .rt6i_protocol  = RTPROT_KERNEL,
 280         .rt6i_metric    = ~(u32) 0,
 281         .rt6i_ref       = ATOMIC_INIT(1),
 282 };
 283
 284 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
 285
 286 static const struct rt6_info ip6_prohibit_entry_template = {
 287         .dst = {
 288                 .__refcnt       = ATOMIC_INIT(1),
 289                 .__use          = 1,
 290                 .obsolete       = DST_OBSOLETE_FORCE_CHK,
 291                 .error          = -EACCES,
 292                 .input          = ip6_pkt_prohibit,
 293                 .output         = ip6_pkt_prohibit_out,
 294         },
 295         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
 296         .rt6i_protocol  = RTPROT_KERNEL,
 297         .rt6i_metric    = ~(u32) 0,
 298         .rt6i_ref       = ATOMIC_INIT(1),
 299 };
 300
 301 static const struct rt6_info ip6_blk_hole_entry_template = {
 302         .dst = {
 303                 .__refcnt       = ATOMIC_INIT(1),
 304                 .__use          = 1,
 305                 .obsolete       = DST_OBSOLETE_FORCE_CHK,
 306                 .error          = -EINVAL,
 307                 .input          = dst_discard,
 308                 .output         = dst_discard_out,
 309         },
 310         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
 311         .rt6i_protocol  = RTPROT_KERNEL,
 312         .rt6i_metric    = ~(u32) 0,
 313         .rt6i_ref       = ATOMIC_INIT(1),
 314 };
 315
 316 #endif
 317
 318 static void rt6_info_init(struct rt6_info *rt)
 319 {
 320         struct dst_entry *dst = &rt->dst;
 321
 322         memset(dst + 1, 0, sizeof(*rt) - sizeof(*dst));
 323         INIT_LIST_HEAD(&rt->rt6i_siblings);
 324         INIT_LIST_HEAD(&rt->rt6i_uncached);
 325 }
 326
 327 /* allocate dst with ip6_dst_ops */
 328 static struct rt6_info *__ip6_dst_alloc(struct net *net,
 329                                         struct net_device *dev,
 330                                         int flags)
 331 {
 332         struct rt6_info *rt = dst_alloc(&net->ipv6.ip6_dst_ops, dev,
 333                                         0, DST_OBSOLETE_FORCE_CHK, flags);
 334
 335         if (rt)
 336                 rt6_info_init(rt);
 337
 338         return rt;
 339 }
 340
 341 struct rt6_info *ip6_dst_alloc(struct net *net,
 342                                struct net_device *dev,
 343                                int flags)
 344 {
 345         struct rt6_info *rt = __ip6_dst_alloc(net, dev, flags);
 346
 347         if (rt) {
 348                 rt->rt6i_pcpu = alloc_percpu_gfp(struct rt6_info *, GFP_ATOMIC);
 349                 if (rt->rt6i_pcpu) {
 350                         int cpu;
 351
 352                         for_each_possible_cpu(cpu) {
 353                                 struct rt6_info **p;
 354
 355                                 p = per_cpu_ptr(rt->rt6i_pcpu, cpu);
 356                                 /* no one shares rt */
 357                                 *p =  NULL;
 358                         }
 359                 } else {
 360                         dst_destroy((struct dst_entry *)rt);
 361                         return NULL;
 362                 }
 363         }
 364
 365         return rt;
 366 }
 367 EXPORT_SYMBOL(ip6_dst_alloc);
 368
 369 static void ip6_dst_destroy(struct dst_entry *dst)
 370 {
 371         struct rt6_info *rt = (struct rt6_info *)dst;
 372         struct dst_entry *from = dst->from;
 373         struct inet6_dev *idev;
 374
 375         dst_destroy_metrics_generic(dst);
 376         free_percpu(rt->rt6i_pcpu);
 377         rt6_uncached_list_del(rt);
 378
 379         idev = rt->rt6i_idev;
 380         if (idev) {
 381                 rt->rt6i_idev = NULL;
 382                 in6_dev_put(idev);
 383         }
 384
 385         dst->from = NULL;
 386         dst_release(from);
 387 }
 388
 389 static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
 390                            int how)
 391 {
 392         struct rt6_info *rt = (struct rt6_info *)dst;
 393         struct inet6_dev *idev = rt->rt6i_idev;
 394         struct net_device *loopback_dev =
 395                 dev_net(dev)->loopback_dev;
 396
 397         if (dev != loopback_dev) {
 398                 if (idev && idev->dev == dev) {
 399                         struct inet6_dev *loopback_idev =
 400                                 in6_dev_get(loopback_dev);
 401                         if (loopback_idev) {
 402                                 rt->rt6i_idev = loopback_idev;
 403                                 in6_dev_put(idev);
 404                         }
 405                 }
 406         }
 407 }
 408
 409 static bool __rt6_check_expired(const struct rt6_info *rt)
 410 {
 411         if (rt->rt6i_flags & RTF_EXPIRES)
 412                 return time_after(jiffies, rt->dst.expires);
 413         else
 414                 return false;
 415 }
 416
 417 static bool rt6_check_expired(const struct rt6_info *rt)
 418 {
 419         if (rt->rt6i_flags & RTF_EXPIRES) {
 420                 if (time_after(jiffies, rt->dst.expires))
 421                         return true;
 422         } else if (rt->dst.from) {
 423                 return rt6_check_expired((struct rt6_info *) rt->dst.from);
 424         }
 425         return false;
 426 }
 427
 428 /* Multipath route selection:
 429  *   Hash based function using packet header and flowlabel.
 430  * Adapted from fib_info_hashfn()
 431  */
 432 static int rt6_info_hash_nhsfn(unsigned int candidate_count,
 433                                const struct flowi6 *fl6)
 434 {
 435         return get_hash_from_flowi6(fl6) % candidate_count;
 436 }
 437
 438 static struct rt6_info *rt6_multipath_select(struct rt6_info *match,
 439                                              struct flowi6 *fl6, int oif,
 440                                              int strict)
 441 {
 442         struct rt6_info *sibling, *next_sibling;
 443         int route_choosen;
 444
 445         route_choosen = rt6_info_hash_nhsfn(match->rt6i_nsiblings + 1, fl6);
 446         /* Don't change the route, if route_choosen == 0
 447          * (siblings does not include ourself)
 448          */
 449         if (route_choosen)
 450                 list_for_each_entry_safe(sibling, next_sibling,
 451                                 &match->rt6i_siblings, rt6i_siblings) {
 452                         route_choosen--;
 453                         if (route_choosen == 0) {
 454                                 if (rt6_score_route(sibling, oif, strict) < 0)
 455                                         break;
 456                                 match = sibling;
 457                                 break;
 458                         }
 459                 }
 460         return match;
 461 }
 462
 463 /*
 464  *      Route lookup. Any table->tb6_lock is implied.
 465  */
 466
 467 static inline struct rt6_info *rt6_device_match(struct net *net,
 468                                                     struct rt6_info *rt,
 469                                                     const struct in6_addr *saddr,
 470                                                     int oif,
 471                                                     int flags)
 472 {
 473         struct rt6_info *local = NULL;
 474         struct rt6_info *sprt;
 475
 476         if (!oif && ipv6_addr_any(saddr))
 477                 goto out;
 478
 479         for (sprt = rt; sprt; sprt = sprt->dst.rt6_next) {
 480                 struct net_device *dev = sprt->dst.dev;
 481
 482                 if (oif) {
 483                         if (dev->ifindex == oif)
 484                                 return sprt;
 485                         if (dev->flags & IFF_LOOPBACK) {
 486                                 if (!sprt->rt6i_idev ||
 487                                     sprt->rt6i_idev->dev->ifindex != oif) {
 488                                         if (flags & RT6_LOOKUP_F_IFACE)
 489                                                 continue;
 490                                         if (local &&
 491                                             local->rt6i_idev->dev->ifindex == oif)
 492                                                 continue;
 493                                 }
 494                                 local = sprt;
 495                         }
 496                 } else {
 497                         if (ipv6_chk_addr(net, saddr, dev,
 498                                           flags & RT6_LOOKUP_F_IFACE))
 499                                 return sprt;
 500                 }
 501         }
 502
 503         if (oif) {
 504                 if (local)
 505                         return local;
 506
 507                 if (flags & RT6_LOOKUP_F_IFACE)
 508                         return net->ipv6.ip6_null_entry;
 509         }
 510 out:
 511         return rt;
 512 }
 513
 514 #ifdef CONFIG_IPV6_ROUTER_PREF
 515 struct __rt6_probe_work {
 516         struct work_struct work;
 517         struct in6_addr target;
 518         struct net_device *dev;
 519 };
 520
 521 static void rt6_probe_deferred(struct work_struct *w)
 522 {
 523         struct in6_addr mcaddr;
 524         struct __rt6_probe_work *work =
 525                 container_of(w, struct __rt6_probe_work, work);
 526
 527         addrconf_addr_solict_mult(&work->target, &mcaddr);
 528         ndisc_send_ns(work->dev, &work->target, &mcaddr, NULL);
 529         dev_put(work->dev);
 530         kfree(work);
 531 }
 532
 533 static void rt6_probe(struct rt6_info *rt)
 534 {
 535         struct __rt6_probe_work *work;
 536         struct neighbour *neigh;
 537         /*
 538          * Okay, this does not seem to be appropriate
 539          * for now, however, we need to check if it
 540          * is really so; aka Router Reachability Probing.
 541          *
 542          * Router Reachability Probe MUST be rate-limited
 543          * to no more than one per minute.
 544          */
 545         if (!rt || !(rt->rt6i_flags & RTF_GATEWAY))
 546                 return;
 547         rcu_read_lock_bh();
 548         neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
 549         if (neigh) {
 550                 if (neigh->nud_state & NUD_VALID)
 551                         goto out;
 552
 553                 work = NULL;
 554                 write_lock(&neigh->lock);
 555                 if (!(neigh->nud_state & NUD_VALID) &&
 556                     time_after(jiffies,
 557                                neigh->updated +
 558                                rt->rt6i_idev->cnf.rtr_probe_interval)) {
 559                         work = kmalloc(sizeof(*work), GFP_ATOMIC);
 560                         if (work)
 561                                 __neigh_set_probe_once(neigh);
 562                 }
 563                 write_unlock(&neigh->lock);
 564         } else {
 565                 work = kmalloc(sizeof(*work), GFP_ATOMIC);
 566         }
 567
 568         if (work) {
 569                 INIT_WORK(&work->work, rt6_probe_deferred);
 570                 work->target = rt->rt6i_gateway;
 571                 dev_hold(rt->dst.dev);
 572                 work->dev = rt->dst.dev;
 573                 schedule_work(&work->work);
 574         }
 575
 576 out:
 577         rcu_read_unlock_bh();
 578 }
 579 #else
 580 static inline void rt6_probe(struct rt6_info *rt)
 581 {
 582 }
 583 #endif
 584
 585 /*
 586  * Default Router Selection (RFC 2461 6.3.6)
 587  */
 588 static inline int rt6_check_dev(struct rt6_info *rt, int oif)
 589 {
 590         struct net_device *dev = rt->dst.dev;
 591         if (!oif || dev->ifindex == oif)
 592                 return 2;
 593         if ((dev->flags & IFF_LOOPBACK) &&
 594             rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
 595                 return 1;
 596         return 0;
 597 }
 598
 599 static inline enum rt6_nud_state rt6_check_neigh(struct rt6_info *rt)
 600 {
 601         struct neighbour *neigh;
 602         enum rt6_nud_state ret = RT6_NUD_FAIL_HARD;
 603
 604         if (rt->rt6i_flags & RTF_NONEXTHOP ||
 605             !(rt->rt6i_flags & RTF_GATEWAY))
 606                 return RT6_NUD_SUCCEED;
 607
 608         rcu_read_lock_bh();
 609         neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
 610         if (neigh) {
 611                 read_lock(&neigh->lock);
 612                 if (neigh->nud_state & NUD_VALID)
 613                         ret = RT6_NUD_SUCCEED;
 614 #ifdef CONFIG_IPV6_ROUTER_PREF
 615                 else if (!(neigh->nud_state & NUD_FAILED))
 616                         ret = RT6_NUD_SUCCEED;
 617                 else
 618                         ret = RT6_NUD_FAIL_PROBE;
 619 #endif
 620                 read_unlock(&neigh->lock);
 621         } else {
 622                 ret = IS_ENABLED(CONFIG_IPV6_ROUTER_PREF) ?
 623                       RT6_NUD_SUCCEED : RT6_NUD_FAIL_DO_RR;
 624         }
 625         rcu_read_unlock_bh();
 626
 627         return ret;
 628 }
 629
 630 static int rt6_score_route(struct rt6_info *rt, int oif,
 631                            int strict)
 632 {
 633         int m;
 634
 635         m = rt6_check_dev(rt, oif);
 636         if (!m && (strict & RT6_LOOKUP_F_IFACE))
 637                 return RT6_NUD_FAIL_HARD;
 638 #ifdef CONFIG_IPV6_ROUTER_PREF
 639         m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
 640 #endif
 641         if (strict & RT6_LOOKUP_F_REACHABLE) {
 642                 int n = rt6_check_neigh(rt);
 643                 if (n < 0)
 644                         return n;
 645         }
 646         return m;
 647 }
 648
 649 static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
 650                                    int *mpri, struct rt6_info *match,
 651                                    bool *do_rr)
 652 {
 653         int m;
 654         bool match_do_rr = false;
 655         struct inet6_dev *idev = rt->rt6i_idev;
 656         struct net_device *dev = rt->dst.dev;
 657
 658         if (dev && !netif_carrier_ok(dev) &&
 659             idev->cnf.ignore_routes_with_linkdown)
 660                 goto out;
 661
 662         if (rt6_check_expired(rt))
 663                 goto out;
 664
 665         m = rt6_score_route(rt, oif, strict);
 666         if (m == RT6_NUD_FAIL_DO_RR) {
 667                 match_do_rr = true;
 668                 m = 0; /* lowest valid score */
 669         } else if (m == RT6_NUD_FAIL_HARD) {
 670                 goto out;
 671         }
 672
 673         if (strict & RT6_LOOKUP_F_REACHABLE)
 674                 rt6_probe(rt);
 675
 676         /* note that m can be RT6_NUD_FAIL_PROBE at this point */
 677         if (m > *mpri) {
 678                 *do_rr = match_do_rr;
 679                 *mpri = m;
 680                 match = rt;
 681         }
 682 out:
 683         return match;
 684 }
 685
 686 static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
 687                                      struct rt6_info *rr_head,
 688                                      u32 metric, int oif, int strict,
 689                                      bool *do_rr)
 690 {
 691         struct rt6_info *rt, *match, *cont;
 692         int mpri = -1;
 693
 694         match = NULL;
 695         cont = NULL;
 696         for (rt = rr_head; rt; rt = rt->dst.rt6_next) {
 697                 if (rt->rt6i_metric != metric) {
 698                         cont = rt;
 699                         break;
 700                 }
 701
 702                 match = find_match(rt, oif, strict, &mpri, match, do_rr);
 703         }
 704
 705         for (rt = fn->leaf; rt && rt != rr_head; rt = rt->dst.rt6_next) {
 706                 if (rt->rt6i_metric != metric) {
 707                         cont = rt;
 708                         break;
 709                 }
 710
 711                 match = find_match(rt, oif, strict, &mpri, match, do_rr);
 712         }
 713
 714         if (match || !cont)
 715                 return match;
 716
 717         for (rt = cont; rt; rt = rt->dst.rt6_next)
 718                 match = find_match(rt, oif, strict, &mpri, match, do_rr);
 719
 720         return match;
 721 }
 722
 723 static struct rt6_info *rt6_select(struct fib6_node *fn, int oif, int strict)
 724 {
 725         struct rt6_info *match, *rt0;
 726         struct net *net;
 727         bool do_rr = false;
 728
 729         rt0 = fn->rr_ptr;
 730         if (!rt0)
 731                 fn->rr_ptr = rt0 = fn->leaf;
 732
 733         match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict,
 734                              &do_rr);
 735
 736         if (do_rr) {
 737                 struct rt6_info *next = rt0->dst.rt6_next;
 738
 739                 /* no entries matched; do round-robin */
 740                 if (!next || next->rt6i_metric != rt0->rt6i_metric)
 741                         next = fn->leaf;
 742
 743                 if (next != rt0)
 744                         fn->rr_ptr = next;
 745         }
 746
 747         net = dev_net(rt0->dst.dev);
 748         return match ? match : net->ipv6.ip6_null_entry;
 749 }
 750
 751 static bool rt6_is_gw_or_nonexthop(const struct rt6_info *rt)
 752 {
 753         return (rt->rt6i_flags & (RTF_NONEXTHOP | RTF_GATEWAY));
 754 }
 755
 756 #ifdef CONFIG_IPV6_ROUTE_INFO
 757 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
 758                   const struct in6_addr *gwaddr)
 759 {
 760         struct net *net = dev_net(dev);
 761         struct route_info *rinfo = (struct route_info *) opt;
 762         struct in6_addr prefix_buf, *prefix;
 763         unsigned int pref;
 764         unsigned long lifetime;
 765         struct rt6_info *rt;
 766
 767         if (len < sizeof(struct route_info)) {
 768                 return -EINVAL;
 769         }
 770
 771         /* Sanity check for prefix_len and length */
 772         if (rinfo->length > 3) {
 773                 return -EINVAL;
 774         } else if (rinfo->prefix_len > 128) {
 775                 return -EINVAL;
 776         } else if (rinfo->prefix_len > 64) {
 777                 if (rinfo->length < 2) {
 778                         return -EINVAL;
 779                 }
 780         } else if (rinfo->prefix_len > 0) {
 781                 if (rinfo->length < 1) {
 782                         return -EINVAL;
 783                 }
 784         }
 785
 786         pref = rinfo->route_pref;
 787         if (pref == ICMPV6_ROUTER_PREF_INVALID)
 788                 return -EINVAL;
 789
 790         lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
 791
 792         if (rinfo->length == 3)
 793                 prefix = (struct in6_addr *)rinfo->prefix;
 794         else {
 795                 /* this function is safe */
 796                 ipv6_addr_prefix(&prefix_buf,
 797                                  (struct in6_addr *)rinfo->prefix,
 798                                  rinfo->prefix_len);
 799                 prefix = &prefix_buf;
 800         }
 801
 802         if (rinfo->prefix_len == 0)
 803                 rt = rt6_get_dflt_router(gwaddr, dev);
 804         else
 805                 rt = rt6_get_route_info(net, prefix, rinfo->prefix_len,
 806                                         gwaddr, dev->ifindex);
 807
 808         if (rt && !lifetime) {
 809                 ip6_del_rt(rt);
 810                 rt = NULL;
 811         }
 812
 813         if (!rt && lifetime)
 814                 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr, dev->ifindex,
 815                                         pref);
 816         else if (rt)
 817                 rt->rt6i_flags = RTF_ROUTEINFO |
 818                                  (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
 819
 820         if (rt) {
 821                 if (!addrconf_finite_timeout(lifetime))
 822                         rt6_clean_expires(rt);
 823                 else
 824                         rt6_set_expires(rt, jiffies + HZ * lifetime);
 825
 826                 ip6_rt_put(rt);
 827         }
 828         return 0;
 829 }
 830 #endif
 831
 832 static struct fib6_node* fib6_backtrack(struct fib6_node *fn,
 833                                         struct in6_addr *saddr)
 834 {
 835         struct fib6_node *pn;
 836         while (1) {
 837                 if (fn->fn_flags & RTN_TL_ROOT)
 838                         return NULL;
 839                 pn = fn->parent;
 840                 if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn)
 841                         fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr);
 842                 else
 843                         fn = pn;
 844                 if (fn->fn_flags & RTN_RTINFO)
 845                         return fn;
 846         }
 847 }
 848
 849 static struct rt6_info *ip6_pol_route_lookup(struct net *net,
 850                                              struct fib6_table *table,
 851                                              struct flowi6 *fl6, int flags)
 852 {
 853         struct fib6_node *fn;
 854         struct rt6_info *rt;
 855
 856         read_lock_bh(&table->tb6_lock);
 857         fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
 858 restart:
 859         rt = fn->leaf;
 860         rt = rt6_device_match(net, rt, &fl6->saddr, fl6->flowi6_oif, flags);
 861         if (rt->rt6i_nsiblings && fl6->flowi6_oif == 0)
 862                 rt = rt6_multipath_select(rt, fl6, fl6->flowi6_oif, flags);
 863         if (rt == net->ipv6.ip6_null_entry) {
 864                 fn = fib6_backtrack(fn, &fl6->saddr);
 865                 if (fn)
 866                         goto restart;
 867         }
 868         dst_use(&rt->dst, jiffies);
 869         read_unlock_bh(&table->tb6_lock);
 870
 871         trace_fib6_table_lookup(net, rt, table->tb6_id, fl6);
 872
 873         return rt;
 874
 875 }
 876
 877 struct dst_entry *ip6_route_lookup(struct net *net, struct flowi6 *fl6,
 878                                     int flags)
 879 {
 880         return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_lookup);
 881 }
 882 EXPORT_SYMBOL_GPL(ip6_route_lookup);
 883
 884 struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
 885                             const struct in6_addr *saddr, int oif, int strict)
 886 {
 887         struct flowi6 fl6 = {
 888                 .flowi6_oif = oif,
 889                 .daddr = *daddr,
 890         };
 891         struct dst_entry *dst;
 892         int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
 893
 894         if (saddr) {
 895                 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
 896                 flags |= RT6_LOOKUP_F_HAS_SADDR;
 897         }
 898
 899         dst = fib6_rule_lookup(net, &fl6, flags, ip6_pol_route_lookup);
 900         if (dst->error == 0)
 901                 return (struct rt6_info *) dst;
 902
 903         dst_release(dst);
 904
 905         return NULL;
 906 }
 907 EXPORT_SYMBOL(rt6_lookup);
 908
 909 /* ip6_ins_rt is called with FREE table->tb6_lock.
 910    It takes new route entry, the addition fails by any reason the
 911    route is freed. In any case, if caller does not hold it, it may
 912    be destroyed.
 913  */
 914
 915 static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info,
 916                         struct mx6_config *mxc)
 917 {
 918         int err;
 919         struct fib6_table *table;
 920
 921         table = rt->rt6i_table;
 922         write_lock_bh(&table->tb6_lock);
 923         err = fib6_add(&table->tb6_root, rt, info, mxc);
 924         write_unlock_bh(&table->tb6_lock);
 925
 926         return err;
 927 }
 928
 929 int ip6_ins_rt(struct rt6_info *rt)
 930 {
 931         struct nl_info info = { .nl_net = dev_net(rt->dst.dev), };
 932         struct mx6_config mxc = { .mx = NULL, };
 933
 934         return __ip6_ins_rt(rt, &info, &mxc);
 935 }
 936
 937 static struct rt6_info *ip6_rt_cache_alloc(struct rt6_info *ort,
 938                                            const struct in6_addr *daddr,
 939                                            const struct in6_addr *saddr)
 940 {
 941         struct rt6_info *rt;
 942
 943         /*
 944          *      Clone the route.
 945          */
 946
 947         if (ort->rt6i_flags & (RTF_CACHE | RTF_PCPU))
 948                 ort = (struct rt6_info *)ort->dst.from;
 949
 950         rt = __ip6_dst_alloc(dev_net(ort->dst.dev), ort->dst.dev, 0);
 951
 952         if (!rt)
 953                 return NULL;
 954
 955         ip6_rt_copy_init(rt, ort);
 956         rt->rt6i_flags |= RTF_CACHE;
 957         rt->rt6i_metric = 0;
 958         rt->dst.flags |= DST_HOST;
 959         rt->rt6i_dst.addr = *daddr;
 960         rt->rt6i_dst.plen = 128;
 961
 962         if (!rt6_is_gw_or_nonexthop(ort)) {
 963                 if (ort->rt6i_dst.plen != 128 &&
 964                     ipv6_addr_equal(&ort->rt6i_dst.addr, daddr))
 965                         rt->rt6i_flags |= RTF_ANYCAST;
 966 #ifdef CONFIG_IPV6_SUBTREES
 967                 if (rt->rt6i_src.plen && saddr) {
 968                         rt->rt6i_src.addr = *saddr;
 969                         rt->rt6i_src.plen = 128;
 970                 }
 971 #endif
 972         }
 973
 974         return rt;
 975 }
 976
 977 static struct rt6_info *ip6_rt_pcpu_alloc(struct rt6_info *rt)
 978 {
 979         struct rt6_info *pcpu_rt;
 980
 981         pcpu_rt = __ip6_dst_alloc(dev_net(rt->dst.dev),
 982                                   rt->dst.dev, rt->dst.flags);
 983
 984         if (!pcpu_rt)
 985                 return NULL;
 986         ip6_rt_copy_init(pcpu_rt, rt);
 987         pcpu_rt->rt6i_protocol = rt->rt6i_protocol;
 988         pcpu_rt->rt6i_flags |= RTF_PCPU;
 989         return pcpu_rt;
 990 }
 991
 992 /* It should be called with read_lock_bh(&tb6_lock) acquired */
 993 static struct rt6_info *rt6_get_pcpu_route(struct rt6_info *rt)
 994 {
 995         struct rt6_info *pcpu_rt, **p;
 996
 997         p = this_cpu_ptr(rt->rt6i_pcpu);
 998         pcpu_rt = *p;
 999
1000         if (pcpu_rt) {
1001                 dst_hold(&pcpu_rt->dst);
1002                 rt6_dst_from_metrics_check(pcpu_rt);
1003         }
1004         return pcpu_rt;
1005 }
1006
1007 static struct rt6_info *rt6_make_pcpu_route(struct rt6_info *rt)
1008 {
1009         struct fib6_table *table = rt->rt6i_table;
1010         struct rt6_info *pcpu_rt, *prev, **p;
1011
1012         pcpu_rt = ip6_rt_pcpu_alloc(rt);
1013         if (!pcpu_rt) {
1014                 struct net *net = dev_net(rt->dst.dev);
1015
1016                 dst_hold(&net->ipv6.ip6_null_entry->dst);
1017                 return net->ipv6.ip6_null_entry;
1018         }
1019
1020         read_lock_bh(&table->tb6_lock);
1021         if (rt->rt6i_pcpu) {
1022                 p = this_cpu_ptr(rt->rt6i_pcpu);
1023                 prev = cmpxchg(p, NULL, pcpu_rt);
1024                 if (prev) {
1025                         /* If someone did it before us, return prev instead */
1026                         dst_destroy(&pcpu_rt->dst);
1027                         pcpu_rt = prev;
1028                 }
1029         } else {
1030                 /* rt has been removed from the fib6 tree
1031                  * before we have a chance to acquire the read_lock.
1032                  * In this case, don't brother to create a pcpu rt
1033                  * since rt is going away anyway.  The next
1034                  * dst_check() will trigger a re-lookup.
1035                  */
1036                 dst_destroy(&pcpu_rt->dst);
1037                 pcpu_rt = rt;
1038         }
1039         dst_hold(&pcpu_rt->dst);
1040         rt6_dst_from_metrics_check(pcpu_rt);
1041         read_unlock_bh(&table->tb6_lock);
1042         return pcpu_rt;
1043 }
1044
1045 struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table,
1046                                int oif, struct flowi6 *fl6, int flags)
1047 {
1048         struct fib6_node *fn, *saved_fn;
1049         struct rt6_info *rt;
1050         int strict = 0;
1051
1052         strict |= flags & RT6_LOOKUP_F_IFACE;
1053         if (net->ipv6.devconf_all->forwarding == 0)
1054                 strict |= RT6_LOOKUP_F_REACHABLE;
1055
1056         read_lock_bh(&table->tb6_lock);
1057
1058         fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1059         saved_fn = fn;
1060
1061         if (fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF)
1062                 oif = 0;
1063
1064 redo_rt6_select:
1065         rt = rt6_select(fn, oif, strict);
1066         if (rt->rt6i_nsiblings)
1067                 rt = rt6_multipath_select(rt, fl6, oif, strict);
1068         if (rt == net->ipv6.ip6_null_entry) {
1069                 fn = fib6_backtrack(fn, &fl6->saddr);
1070                 if (fn)
1071                         goto redo_rt6_select;
1072                 else if (strict & RT6_LOOKUP_F_REACHABLE) {
1073                         /* also consider unreachable route */
1074                         strict &= ~RT6_LOOKUP_F_REACHABLE;
1075                         fn = saved_fn;
1076                         goto redo_rt6_select;
1077                 }
1078         }
1079
1080
1081         if (rt == net->ipv6.ip6_null_entry || (rt->rt6i_flags & RTF_CACHE)) {
1082                 dst_use(&rt->dst, jiffies);
1083                 read_unlock_bh(&table->tb6_lock);
1084
1085                 rt6_dst_from_metrics_check(rt);
1086
1087                 trace_fib6_table_lookup(net, rt, table->tb6_id, fl6);
1088                 return rt;
1089         } else if (unlikely((fl6->flowi6_flags & FLOWI_FLAG_KNOWN_NH) &&
1090                             !(rt->rt6i_flags & RTF_GATEWAY))) {
1091                 /* Create a RTF_CACHE clone which will not be
1092                  * owned by the fib6 tree.  It is for the special case where
1093                  * the daddr in the skb during the neighbor look-up is different
1094                  * from the fl6->daddr used to look-up route here.
1095                  */
1096
1097                 struct rt6_info *uncached_rt;
1098
1099                 dst_use(&rt->dst, jiffies);
1100                 read_unlock_bh(&table->tb6_lock);
1101
1102                 uncached_rt = ip6_rt_cache_alloc(rt, &fl6->daddr, NULL);
1103                 dst_release(&rt->dst);
1104
1105                 if (uncached_rt)
1106                         rt6_uncached_list_add(uncached_rt);
1107                 else
1108                         uncached_rt = net->ipv6.ip6_null_entry;
1109
1110                 dst_hold(&uncached_rt->dst);
1111
1112                 trace_fib6_table_lookup(net, uncached_rt, table->tb6_id, fl6);
1113                 return uncached_rt;
1114
1115         } else {
1116                 /* Get a percpu copy */
1117
1118                 struct rt6_info *pcpu_rt;
1119
1120                 rt->dst.lastuse = jiffies;
1121                 rt->dst.__use++;
1122                 pcpu_rt = rt6_get_pcpu_route(rt);
1123
1124                 if (pcpu_rt) {
1125                         read_unlock_bh(&table->tb6_lock);
1126                 } else {
1127                         /* We have to do the read_unlock first
1128                          * because rt6_make_pcpu_route() may trigger
1129                          * ip6_dst_gc() which will take the write_lock.
1130                          */
1131                         dst_hold(&rt->dst);
1132                         read_unlock_bh(&table->tb6_lock);
1133                         pcpu_rt = rt6_make_pcpu_route(rt);
1134                         dst_release(&rt->dst);
1135                 }
1136
1137                 trace_fib6_table_lookup(net, pcpu_rt, table->tb6_id, fl6);
1138                 return pcpu_rt;
1139
1140         }
1141 }
1142 EXPORT_SYMBOL_GPL(ip6_pol_route);
1143
1144 static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
1145                                             struct flowi6 *fl6, int flags)
1146 {
1147         return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, flags);
1148 }
1149
1150 struct dst_entry *ip6_route_input_lookup(struct net *net,
1151                                          struct net_device *dev,
1152                                          struct flowi6 *fl6, int flags)
1153 {
1154         if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
1155                 flags |= RT6_LOOKUP_F_IFACE;
1156
1157         return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_input);
1158 }
1159 EXPORT_SYMBOL_GPL(ip6_route_input_lookup);
1160
1161 void ip6_route_input(struct sk_buff *skb)
1162 {
1163         const struct ipv6hdr *iph = ipv6_hdr(skb);
1164         struct net *net = dev_net(skb->dev);
1165         int flags = RT6_LOOKUP_F_HAS_SADDR;
1166         struct ip_tunnel_info *tun_info;
1167         struct flowi6 fl6 = {
1168                 .flowi6_iif = skb->dev->ifindex,
1169                 .daddr = iph->daddr,
1170                 .saddr = iph->saddr,
1171                 .flowlabel = ip6_flowinfo(iph),
1172                 .flowi6_mark = skb->mark,
1173                 .flowi6_proto = iph->nexthdr,
1174         };
1175
1176         tun_info = skb_tunnel_info(skb);
1177         if (tun_info && !(tun_info->mode & IP_TUNNEL_INFO_TX))
1178                 fl6.flowi6_tun_key.tun_id = tun_info->key.tun_id;
1179         skb_dst_drop(skb);
1180         skb_dst_set(skb, ip6_route_input_lookup(net, skb->dev, &fl6, flags));
1181 }
1182
1183 static struct rt6_info *ip6_pol_route_output(struct net *net, struct fib6_table *table,
1184                                              struct flowi6 *fl6, int flags)
1185 {
1186         return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, flags);
1187 }
1188
1189 struct dst_entry *ip6_route_output_flags(struct net *net, const struct sock *sk,
1190                                          struct flowi6 *fl6, int flags)
1191 {
1192         bool any_src;
1193
1194         if (rt6_need_strict(&fl6->daddr)) {
1195                 struct dst_entry *dst;
1196
1197                 dst = l3mdev_link_scope_lookup(net, fl6);
1198                 if (dst)
1199                         return dst;
1200         }
1201
1202         fl6->flowi6_iif = LOOPBACK_IFINDEX;
1203
1204         any_src = ipv6_addr_any(&fl6->saddr);
1205         if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr) ||
1206             (fl6->flowi6_oif && any_src))
1207                 flags |= RT6_LOOKUP_F_IFACE;
1208
1209         if (!any_src)
1210                 flags |= RT6_LOOKUP_F_HAS_SADDR;
1211         else if (sk)
1212                 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
1213
1214         return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_output);
1215 }
1216 EXPORT_SYMBOL_GPL(ip6_route_output_flags);
1217
1218 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
1219 {
1220         struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
1221         struct dst_entry *new = NULL;
1222
1223         rt = dst_alloc(&ip6_dst_blackhole_ops, ort->dst.dev, 1, DST_OBSOLETE_NONE, 0);
1224         if (rt) {
1225                 rt6_info_init(rt);
1226
1227                 new = &rt->dst;
1228                 new->__use = 1;
1229                 new->input = dst_discard;
1230                 new->output = dst_discard_out;
1231
1232                 dst_copy_metrics(new, &ort->dst);
1233                 rt->rt6i_idev = ort->rt6i_idev;
1234                 if (rt->rt6i_idev)
1235                         in6_dev_hold(rt->rt6i_idev);
1236
1237                 rt->rt6i_gateway = ort->rt6i_gateway;
1238                 rt->rt6i_flags = ort->rt6i_flags & ~RTF_PCPU;
1239                 rt->rt6i_metric = 0;
1240
1241                 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1242 #ifdef CONFIG_IPV6_SUBTREES
1243                 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1244 #endif
1245
1246                 dst_free(new);
1247         }
1248
1249         dst_release(dst_orig);
1250         return new ? new : ERR_PTR(-ENOMEM);
1251 }
1252
1253 /*
1254  *      Destination cache support functions
1255  */
1256
1257 static void rt6_dst_from_metrics_check(struct rt6_info *rt)
1258 {
1259         if (rt->dst.from &&
1260             dst_metrics_ptr(&rt->dst) != dst_metrics_ptr(rt->dst.from))
1261                 dst_init_metrics(&rt->dst, dst_metrics_ptr(rt->dst.from), true);
1262 }
1263
1264 static struct dst_entry *rt6_check(struct rt6_info *rt, u32 cookie)
1265 {
1266         if (!rt->rt6i_node || (rt->rt6i_node->fn_sernum != cookie))
1267                 return NULL;
1268
1269         if (rt6_check_expired(rt))
1270                 return NULL;
1271
1272         return &rt->dst;
1273 }
1274
1275 static struct dst_entry *rt6_dst_from_check(struct rt6_info *rt, u32 cookie)
1276 {
1277         if (!__rt6_check_expired(rt) &&
1278             rt->dst.obsolete == DST_OBSOLETE_FORCE_CHK &&
1279             rt6_check((struct rt6_info *)(rt->dst.from), cookie))
1280                 return &rt->dst;
1281         else
1282                 return NULL;
1283 }
1284
1285 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
1286 {
1287         struct rt6_info *rt;
1288
1289         rt = (struct rt6_info *) dst;
1290
1291         /* All IPV6 dsts are created with ->obsolete set to the value
1292          * DST_OBSOLETE_FORCE_CHK which forces validation calls down
1293          * into this function always.
1294          */
1295
1296         rt6_dst_from_metrics_check(rt);
1297
1298         if (rt->rt6i_flags & RTF_PCPU ||
1299             (unlikely(dst->flags & DST_NOCACHE) && rt->dst.from))
1300                 return rt6_dst_from_check(rt, cookie);
1301         else
1302                 return rt6_check(rt, cookie);
1303 }
1304
1305 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
1306 {
1307         struct rt6_info *rt = (struct rt6_info *) dst;
1308
1309         if (rt) {
1310                 if (rt->rt6i_flags & RTF_CACHE) {
1311                         if (rt6_check_expired(rt)) {
1312                                 ip6_del_rt(rt);
1313                                 dst = NULL;
1314                         }
1315                 } else {
1316                         dst_release(dst);
1317                         dst = NULL;
1318                 }
1319         }
1320         return dst;
1321 }
1322
1323 static void ip6_link_failure(struct sk_buff *skb)
1324 {
1325         struct rt6_info *rt;
1326
1327         icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
1328
1329         rt = (struct rt6_info *) skb_dst(skb);
1330         if (rt) {
1331                 if (rt->rt6i_flags & RTF_CACHE) {
1332                         dst_hold(&rt->dst);
1333                         ip6_del_rt(rt);
1334                 } else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT)) {
1335                         rt->rt6i_node->fn_sernum = -1;
1336                 }
1337         }
1338 }
1339
1340 static void rt6_do_update_pmtu(struct rt6_info *rt, u32 mtu)
1341 {
1342         struct net *net = dev_net(rt->dst.dev);
1343
1344         rt->rt6i_flags |= RTF_MODIFIED;
1345         rt->rt6i_pmtu = mtu;
1346         rt6_update_expires(rt, net->ipv6.sysctl.ip6_rt_mtu_expires);
1347 }
1348
1349 static bool rt6_cache_allowed_for_pmtu(const struct rt6_info *rt)
1350 {
1351         return !(rt->rt6i_flags & RTF_CACHE) &&
1352                 (rt->rt6i_flags & RTF_PCPU || rt->rt6i_node);
1353 }
1354
1355 static void __ip6_rt_update_pmtu(struct dst_entry *dst, const struct sock *sk,
1356                                  const struct ipv6hdr *iph, u32 mtu)
1357 {
1358         struct rt6_info *rt6 = (struct rt6_info *)dst;
1359
1360         if (rt6->rt6i_flags & RTF_LOCAL)
1361                 return;
1362
1363         dst_confirm(dst);
1364         mtu = max_t(u32, mtu, IPV6_MIN_MTU);
1365         if (mtu >= dst_mtu(dst))
1366                 return;
1367
1368         if (!rt6_cache_allowed_for_pmtu(rt6)) {
1369                 rt6_do_update_pmtu(rt6, mtu);
1370         } else {
1371                 const struct in6_addr *daddr, *saddr;
1372                 struct rt6_info *nrt6;
1373
1374                 if (iph) {
1375                         daddr = &iph->daddr;
1376                         saddr = &iph->saddr;
1377                 } else if (sk) {
1378                         daddr = &sk->sk_v6_daddr;
1379                         saddr = &inet6_sk(sk)->saddr;
1380                 } else {
1381                         return;
1382                 }
1383                 nrt6 = ip6_rt_cache_alloc(rt6, daddr, saddr);
1384                 if (nrt6) {
1385                         rt6_do_update_pmtu(nrt6, mtu);
1386
1387                         /* ip6_ins_rt(nrt6) will bump the
1388                          * rt6->rt6i_node->fn_sernum
1389                          * which will fail the next rt6_check() and
1390                          * invalidate the sk->sk_dst_cache.
1391                          */
1392                         ip6_ins_rt(nrt6);
1393                 }
1394         }
1395 }
1396
1397 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
1398                                struct sk_buff *skb, u32 mtu)
1399 {
1400         __ip6_rt_update_pmtu(dst, sk, skb ? ipv6_hdr(skb) : NULL, mtu);
1401 }
1402
1403 void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
1404                      int oif, u32 mark)
1405 {
1406         const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1407         struct dst_entry *dst;
1408         struct flowi6 fl6;
1409
1410         memset(&fl6, 0, sizeof(fl6));
1411         fl6.flowi6_oif = oif;
1412         fl6.flowi6_mark = mark ? mark : IP6_REPLY_MARK(net, skb->mark);
1413         fl6.daddr = iph->daddr;
1414         fl6.saddr = iph->saddr;
1415         fl6.flowlabel = ip6_flowinfo(iph);
1416
1417         dst = ip6_route_output(net, NULL, &fl6);
1418         if (!dst->error)
1419                 __ip6_rt_update_pmtu(dst, NULL, iph, ntohl(mtu));
1420         dst_release(dst);
1421 }
1422 EXPORT_SYMBOL_GPL(ip6_update_pmtu);
1423
1424 void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
1425 {
1426         struct dst_entry *dst;
1427
1428         ip6_update_pmtu(skb, sock_net(sk), mtu,
1429                         sk->sk_bound_dev_if, sk->sk_mark);
1430
1431         dst = __sk_dst_get(sk);
1432         if (!dst || !dst->obsolete ||
1433             dst->ops->check(dst, inet6_sk(sk)->dst_cookie))
1434                 return;
1435
1436         bh_lock_sock(sk);
1437         if (!sock_owned_by_user(sk) && !ipv6_addr_v4mapped(&sk->sk_v6_daddr))
1438                 ip6_datagram_dst_update(sk, false);
1439         bh_unlock_sock(sk);
1440 }
1441 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
1442
1443 /* Handle redirects */
1444 struct ip6rd_flowi {
1445         struct flowi6 fl6;
1446         struct in6_addr gateway;
1447 };
1448
1449 static struct rt6_info *__ip6_route_redirect(struct net *net,
1450                                              struct fib6_table *table,
1451                                              struct flowi6 *fl6,
1452                                              int flags)
1453 {
1454         struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
1455         struct rt6_info *rt;
1456         struct fib6_node *fn;
1457
1458         /* Get the "current" route for this destination and
1459          * check if the redirect has come from approriate router.
1460          *
1461          * RFC 4861 specifies that redirects should only be
1462          * accepted if they come from the nexthop to the target.
1463          * Due to the way the routes are chosen, this notion
1464          * is a bit fuzzy and one might need to check all possible
1465          * routes.
1466          */
1467
1468         read_lock_bh(&table->tb6_lock);
1469         fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1470 restart:
1471         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1472                 if (rt6_check_expired(rt))
1473                         continue;
1474                 if (rt->dst.error)
1475                         break;
1476                 if (!(rt->rt6i_flags & RTF_GATEWAY))
1477                         continue;
1478                 if (fl6->flowi6_oif != rt->dst.dev->ifindex)
1479                         continue;
1480                 if (!ipv6_addr_equal(&rdfl->gateway, &rt->rt6i_gateway))
1481                         continue;
1482                 break;
1483         }
1484
1485         if (!rt)
1486                 rt = net->ipv6.ip6_null_entry;
1487         else if (rt->dst.error) {
1488                 rt = net->ipv6.ip6_null_entry;
1489                 goto out;
1490         }
1491
1492         if (rt == net->ipv6.ip6_null_entry) {
1493                 fn = fib6_backtrack(fn, &fl6->saddr);
1494                 if (fn)
1495                         goto restart;
1496         }
1497
1498 out:
1499         dst_hold(&rt->dst);
1500
1501         read_unlock_bh(&table->tb6_lock);
1502
1503         trace_fib6_table_lookup(net, rt, table->tb6_id, fl6);
1504         return rt;
1505 };
1506
1507 static struct dst_entry *ip6_route_redirect(struct net *net,
1508                                         const struct flowi6 *fl6,
1509                                         const struct in6_addr *gateway)
1510 {
1511         int flags = RT6_LOOKUP_F_HAS_SADDR;
1512         struct ip6rd_flowi rdfl;
1513
1514         rdfl.fl6 = *fl6;
1515         rdfl.gateway = *gateway;
1516
1517         return fib6_rule_lookup(net, &rdfl.fl6,
1518                                 flags, __ip6_route_redirect);
1519 }
1520
1521 void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark)
1522 {
1523         const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1524         struct dst_entry *dst;
1525         struct flowi6 fl6;
1526
1527         memset(&fl6, 0, sizeof(fl6));
1528         fl6.flowi6_iif = LOOPBACK_IFINDEX;
1529         fl6.flowi6_oif = oif;
1530         fl6.flowi6_mark = mark;
1531         fl6.daddr = iph->daddr;
1532         fl6.saddr = iph->saddr;
1533         fl6.flowlabel = ip6_flowinfo(iph);
1534
1535         dst = ip6_route_redirect(net, &fl6, &ipv6_hdr(skb)->saddr);
1536         rt6_do_redirect(dst, NULL, skb);
1537         dst_release(dst);
1538 }
1539 EXPORT_SYMBOL_GPL(ip6_redirect);
1540
1541 void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif,
1542                             u32 mark)
1543 {
1544         const struct ipv6hdr *iph = ipv6_hdr(skb);
1545         const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
1546         struct dst_entry *dst;
1547         struct flowi6 fl6;
1548
1549         memset(&fl6, 0, sizeof(fl6));
1550         fl6.flowi6_iif = LOOPBACK_IFINDEX;
1551         fl6.flowi6_oif = oif;
1552         fl6.flowi6_mark = mark;
1553         fl6.daddr = msg->dest;
1554         fl6.saddr = iph->daddr;
1555
1556         dst = ip6_route_redirect(net, &fl6, &iph->saddr);
1557         rt6_do_redirect(dst, NULL, skb);
1558         dst_release(dst);
1559 }
1560
1561 void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
1562 {
1563         ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark);
1564 }
1565 EXPORT_SYMBOL_GPL(ip6_sk_redirect);
1566
1567 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1568 {
1569         struct net_device *dev = dst->dev;
1570         unsigned int mtu = dst_mtu(dst);
1571         struct net *net = dev_net(dev);
1572
1573         mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
1574
1575         if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
1576                 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1577
1578         /*
1579          * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1580          * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1581          * IPV6_MAXPLEN is also valid and means: "any MSS,
1582          * rely only on pmtu discovery"
1583          */
1584         if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
1585                 mtu = IPV6_MAXPLEN;
1586         return mtu;
1587 }
1588
1589 static unsigned int ip6_mtu(const struct dst_entry *dst)
1590 {
1591         const struct rt6_info *rt = (const struct rt6_info *)dst;
1592         unsigned int mtu = rt->rt6i_pmtu;
1593         struct inet6_dev *idev;
1594
1595         if (mtu)
1596                 goto out;
1597
1598         mtu = dst_metric_raw(dst, RTAX_MTU);
1599         if (mtu)
1600                 goto out;
1601
1602         mtu = IPV6_MIN_MTU;
1603
1604         rcu_read_lock();
1605         idev = __in6_dev_get(dst->dev);
1606         if (idev)
1607                 mtu = idev->cnf.mtu6;
1608         rcu_read_unlock();
1609
1610 out:
1611         mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
1612
1613         return mtu - lwtunnel_headroom(dst->lwtstate, mtu);
1614 }
1615
1616 static struct dst_entry *icmp6_dst_gc_list;
1617 static DEFINE_SPINLOCK(icmp6_dst_lock);
1618
1619 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
1620                                   struct flowi6 *fl6)
1621 {
1622         struct dst_entry *dst;
1623         struct rt6_info *rt;
1624         struct inet6_dev *idev = in6_dev_get(dev);
1625         struct net *net = dev_net(dev);
1626
1627         if (unlikely(!idev))
1628                 return ERR_PTR(-ENODEV);
1629
1630         rt = ip6_dst_alloc(net, dev, 0);
1631         if (unlikely(!rt)) {
1632                 in6_dev_put(idev);
1633                 dst = ERR_PTR(-ENOMEM);
1634                 goto out;
1635         }
1636
1637         rt->dst.flags |= DST_HOST;
1638         rt->dst.output  = ip6_output;
1639         atomic_set(&rt->dst.__refcnt, 1);
1640         rt->rt6i_gateway  = fl6->daddr;
1641         rt->rt6i_dst.addr = fl6->daddr;
1642         rt->rt6i_dst.plen = 128;
1643         rt->rt6i_idev     = idev;
1644         dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
1645
1646         spin_lock_bh(&icmp6_dst_lock);
1647         rt->dst.next = icmp6_dst_gc_list;
1648         icmp6_dst_gc_list = &rt->dst;
1649         spin_unlock_bh(&icmp6_dst_lock);
1650
1651         fib6_force_start_gc(net);
1652
1653         dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
1654
1655 out:
1656         return dst;
1657 }
1658
1659 int icmp6_dst_gc(void)
1660 {
1661         struct dst_entry *dst, **pprev;
1662         int more = 0;
1663
1664         spin_lock_bh(&icmp6_dst_lock);
1665         pprev = &icmp6_dst_gc_list;
1666
1667         while ((dst = *pprev) != NULL) {
1668                 if (!atomic_read(&dst->__refcnt)) {
1669                         *pprev = dst->next;
1670                         dst_free(dst);
1671                 } else {
1672                         pprev = &dst->next;
1673                         ++more;
1674                 }
1675         }
1676
1677         spin_unlock_bh(&icmp6_dst_lock);
1678
1679         return more;
1680 }
1681
1682 static void icmp6_clean_all(int (*func)(struct rt6_info *rt, void *arg),
1683                             void *arg)
1684 {
1685         struct dst_entry *dst, **pprev;
1686
1687         spin_lock_bh(&icmp6_dst_lock);
1688         pprev = &icmp6_dst_gc_list;
1689         while ((dst = *pprev) != NULL) {
1690                 struct rt6_info *rt = (struct rt6_info *) dst;
1691                 if (func(rt, arg)) {
1692                         *pprev = dst->next;
1693                         dst_free(dst);
1694                 } else {
1695                         pprev = &dst->next;
1696                 }
1697         }
1698         spin_unlock_bh(&icmp6_dst_lock);
1699 }
1700
1701 static int ip6_dst_gc(struct dst_ops *ops)
1702 {
1703         struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
1704         int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1705         int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1706         int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1707         int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1708         unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1709         int entries;
1710
1711         entries = dst_entries_get_fast(ops);
1712         if (time_after(rt_last_gc + rt_min_interval, jiffies) &&
1713             entries <= rt_max_size)
1714                 goto out;
1715
1716         net->ipv6.ip6_rt_gc_expire++;
1717         fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net, true);
1718         entries = dst_entries_get_slow(ops);
1719         if (entries < ops->gc_thresh)
1720                 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1721 out:
1722         net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1723         return entries > rt_max_size;
1724 }
1725
1726 static int ip6_convert_metrics(struct mx6_config *mxc,
1727                                const struct fib6_config *cfg)
1728 {
1729         bool ecn_ca = false;
1730         struct nlattr *nla;
1731         int remaining;
1732         u32 *mp;
1733
1734         if (!cfg->fc_mx)
1735                 return 0;
1736
1737         mp = kzalloc(sizeof(u32) * RTAX_MAX, GFP_KERNEL);
1738         if (unlikely(!mp))
1739                 return -ENOMEM;
1740
1741         nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1742                 int type = nla_type(nla);
1743                 u32 val;
1744
1745                 if (!type)
1746                         continue;
1747                 if (unlikely(type > RTAX_MAX))
1748                         goto err;
1749
1750                 if (type == RTAX_CC_ALGO) {
1751                         char tmp[TCP_CA_NAME_MAX];
1752
1753                         nla_strlcpy(tmp, nla, sizeof(tmp));
1754                         val = tcp_ca_get_key_by_name(tmp, &ecn_ca);
1755                         if (val == TCP_CA_UNSPEC)
1756                                 goto err;
1757                 } else {
1758                         val = nla_get_u32(nla);
1759                 }
1760                 if (type == RTAX_HOPLIMIT && val > 255)
1761                         val = 255;
1762                 if (type == RTAX_FEATURES && (val & ~RTAX_FEATURE_MASK))
1763                         goto err;
1764
1765                 mp[type - 1] = val;
1766                 __set_bit(type - 1, mxc->mx_valid);
1767         }
1768
1769         if (ecn_ca) {
1770                 __set_bit(RTAX_FEATURES - 1, mxc->mx_valid);
1771                 mp[RTAX_FEATURES - 1] |= DST_FEATURE_ECN_CA;
1772         }
1773
1774         mxc->mx = mp;
1775         return 0;
1776  err:
1777         kfree(mp);
1778         return -EINVAL;
1779 }
1780
1781 static struct rt6_info *ip6_nh_lookup_table(struct net *net,
1782                                             struct fib6_config *cfg,
1783                                             const struct in6_addr *gw_addr)
1784 {
1785         struct flowi6 fl6 = {
1786                 .flowi6_oif = cfg->fc_ifindex,
1787                 .daddr = *gw_addr,
1788                 .saddr = cfg->fc_prefsrc,
1789         };
1790         struct fib6_table *table;
1791         struct rt6_info *rt;
1792         int flags = RT6_LOOKUP_F_IFACE;
1793
1794         table = fib6_get_table(net, cfg->fc_table);
1795         if (!table)
1796                 return NULL;
1797
1798         if (!ipv6_addr_any(&cfg->fc_prefsrc))
1799                 flags |= RT6_LOOKUP_F_HAS_SADDR;
1800
1801         rt = ip6_pol_route(net, table, cfg->fc_ifindex, &fl6, flags);
1802
1803         /* if table lookup failed, fall back to full lookup */
1804         if (rt == net->ipv6.ip6_null_entry) {
1805                 ip6_rt_put(rt);
1806                 rt = NULL;
1807         }
1808
1809         return rt;
1810 }
1811
1812 static struct rt6_info *ip6_route_info_create(struct fib6_config *cfg)
1813 {
1814         struct net *net = cfg->fc_nlinfo.nl_net;
1815         struct rt6_info *rt = NULL;
1816         struct net_device *dev = NULL;
1817         struct inet6_dev *idev = NULL;
1818         struct fib6_table *table;
1819         int addr_type;
1820         int err = -EINVAL;
1821
1822         if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1823                 goto out;
1824 #ifndef CONFIG_IPV6_SUBTREES
1825         if (cfg->fc_src_len)
1826                 goto out;
1827 #endif
1828         if (cfg->fc_ifindex) {
1829                 err = -ENODEV;
1830                 dev = dev_get_by_index(net, cfg->fc_ifindex);
1831                 if (!dev)
1832                         goto out;
1833                 idev = in6_dev_get(dev);
1834                 if (!idev)
1835                         goto out;
1836         }
1837
1838         if (cfg->fc_metric == 0)
1839                 cfg->fc_metric = IP6_RT_PRIO_USER;
1840
1841         err = -ENOBUFS;
1842         if (cfg->fc_nlinfo.nlh &&
1843             !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
1844                 table = fib6_get_table(net, cfg->fc_table);
1845                 if (!table) {
1846                         pr_warn("NLM_F_CREATE should be specified when creating new route\n");
1847                         table = fib6_new_table(net, cfg->fc_table);
1848                 }
1849         } else {
1850                 table = fib6_new_table(net, cfg->fc_table);
1851         }
1852
1853         if (!table)
1854                 goto out;
1855
1856         rt = ip6_dst_alloc(net, NULL,
1857                            (cfg->fc_flags & RTF_ADDRCONF) ? 0 : DST_NOCOUNT);
1858
1859         if (!rt) {
1860                 err = -ENOMEM;
1861                 goto out;
1862         }
1863
1864         if (cfg->fc_flags & RTF_EXPIRES)
1865                 rt6_set_expires(rt, jiffies +
1866                                 clock_t_to_jiffies(cfg->fc_expires));
1867         else
1868                 rt6_clean_expires(rt);
1869
1870         if (cfg->fc_protocol == RTPROT_UNSPEC)
1871                 cfg->fc_protocol = RTPROT_BOOT;
1872         rt->rt6i_protocol = cfg->fc_protocol;
1873
1874         addr_type = ipv6_addr_type(&cfg->fc_dst);
1875
1876         if (addr_type & IPV6_ADDR_MULTICAST)
1877                 rt->dst.input = ip6_mc_input;
1878         else if (cfg->fc_flags & RTF_LOCAL)
1879                 rt->dst.input = ip6_input;
1880         else
1881                 rt->dst.input = ip6_forward;
1882
1883         rt->dst.output = ip6_output;
1884
1885         if (cfg->fc_encap) {
1886                 struct lwtunnel_state *lwtstate;
1887
1888                 err = lwtunnel_build_state(dev, cfg->fc_encap_type,
1889                                            cfg->fc_encap, AF_INET6, cfg,
1890                                            &lwtstate);
1891                 if (err)
1892                         goto out;
1893                 rt->dst.lwtstate = lwtstate_get(lwtstate);
1894                 if (lwtunnel_output_redirect(rt->dst.lwtstate)) {
1895                         rt->dst.lwtstate->orig_output = rt->dst.output;
1896                         rt->dst.output = lwtunnel_output;
1897                 }
1898                 if (lwtunnel_input_redirect(rt->dst.lwtstate)) {
1899                         rt->dst.lwtstate->orig_input = rt->dst.input;
1900                         rt->dst.input = lwtunnel_input;
1901                 }
1902         }
1903
1904         ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1905         rt->rt6i_dst.plen = cfg->fc_dst_len;
1906         if (rt->rt6i_dst.plen == 128)
1907                 rt->dst.flags |= DST_HOST;
1908
1909 #ifdef CONFIG_IPV6_SUBTREES
1910         ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1911         rt->rt6i_src.plen = cfg->fc_src_len;
1912 #endif
1913
1914         rt->rt6i_metric = cfg->fc_metric;
1915
1916         /* We cannot add true routes via loopback here,
1917            they would result in kernel looping; promote them to reject routes
1918          */
1919         if ((cfg->fc_flags & RTF_REJECT) ||
1920             (dev && (dev->flags & IFF_LOOPBACK) &&
1921              !(addr_type & IPV6_ADDR_LOOPBACK) &&
1922              !(cfg->fc_flags & RTF_LOCAL))) {
1923                 /* hold loopback dev/idev if we haven't done so. */
1924                 if (dev != net->loopback_dev) {
1925                         if (dev) {
1926                                 dev_put(dev);
1927                                 in6_dev_put(idev);
1928                         }
1929                         dev = net->loopback_dev;
1930                         dev_hold(dev);
1931                         idev = in6_dev_get(dev);
1932                         if (!idev) {
1933                                 err = -ENODEV;
1934                                 goto out;
1935                         }
1936                 }
1937                 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1938                 switch (cfg->fc_type) {
1939                 case RTN_BLACKHOLE:
1940                         rt->dst.error = -EINVAL;
1941                         rt->dst.output = dst_discard_out;
1942                         rt->dst.input = dst_discard;
1943                         break;
1944                 case RTN_PROHIBIT:
1945                         rt->dst.error = -EACCES;
1946                         rt->dst.output = ip6_pkt_prohibit_out;
1947                         rt->dst.input = ip6_pkt_prohibit;
1948                         break;
1949                 case RTN_THROW:
1950                 case RTN_UNREACHABLE:
1951                 default:
1952                         rt->dst.error = (cfg->fc_type == RTN_THROW) ? -EAGAIN
1953                                         : (cfg->fc_type == RTN_UNREACHABLE)
1954                                         ? -EHOSTUNREACH : -ENETUNREACH;
1955                         rt->dst.output = ip6_pkt_discard_out;
1956                         rt->dst.input = ip6_pkt_discard;
1957                         break;
1958                 }
1959                 goto install_route;
1960         }
1961
1962         if (cfg->fc_flags & RTF_GATEWAY) {
1963                 const struct in6_addr *gw_addr;
1964                 int gwa_type;
1965
1966                 gw_addr = &cfg->fc_gateway;
1967                 gwa_type = ipv6_addr_type(gw_addr);
1968
1969                 /* if gw_addr is local we will fail to detect this in case
1970                  * address is still TENTATIVE (DAD in progress). rt6_lookup()
1971                  * will return already-added prefix route via interface that
1972                  * prefix route was assigned to, which might be non-loopback.
1973                  */
1974                 err = -EINVAL;
1975                 if (ipv6_chk_addr_and_flags(net, gw_addr,
1976                                             gwa_type & IPV6_ADDR_LINKLOCAL ?
1977                                             dev : NULL, 0, 0))
1978                         goto out;
1979
1980                 rt->rt6i_gateway = *gw_addr;
1981
1982                 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1983                         struct rt6_info *grt = NULL;
1984
1985                         /* IPv6 strictly inhibits using not link-local
1986                            addresses as nexthop address.
1987                            Otherwise, router will not able to send redirects.
1988                            It is very good, but in some (rare!) circumstances
1989                            (SIT, PtP, NBMA NOARP links) it is handy to allow
1990                            some exceptions. --ANK
1991                          */
1992                         if (!(gwa_type & IPV6_ADDR_UNICAST))
1993                                 goto out;
1994
1995                         if (cfg->fc_table) {
1996                                 grt = ip6_nh_lookup_table(net, cfg, gw_addr);
1997
1998                                 if (grt) {
1999                                         if (grt->rt6i_flags & RTF_GATEWAY ||
2000                                             (dev && dev != grt->dst.dev)) {
2001                                                 ip6_rt_put(grt);
2002                                                 grt = NULL;
2003                                         }
2004                                 }
2005                         }
2006
2007                         if (!grt)
2008                                 grt = rt6_lookup(net, gw_addr, NULL,
2009                                                  cfg->fc_ifindex, 1);
2010
2011                         err = -EHOSTUNREACH;
2012                         if (!grt)
2013                                 goto out;
2014                         if (dev) {
2015                                 if (dev != grt->dst.dev) {
2016                                         ip6_rt_put(grt);
2017                                         goto out;
2018                                 }
2019                         } else {
2020                                 dev = grt->dst.dev;
2021                                 idev = grt->rt6i_idev;
2022                                 dev_hold(dev);
2023                                 in6_dev_hold(grt->rt6i_idev);
2024                         }
2025                         if (!(grt->rt6i_flags & RTF_GATEWAY))
2026                                 err = 0;
2027                         ip6_rt_put(grt);
2028
2029                         if (err)
2030                                 goto out;
2031                 }
2032                 err = -EINVAL;
2033                 if (!dev || (dev->flags & IFF_LOOPBACK))
2034                         goto out;
2035         }
2036
2037         err = -ENODEV;
2038         if (!dev)
2039                 goto out;
2040
2041         if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
2042                 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
2043                         err = -EINVAL;
2044                         goto out;
2045                 }
2046                 rt->rt6i_prefsrc.addr = cfg->fc_prefsrc;
2047                 rt->rt6i_prefsrc.plen = 128;
2048         } else
2049                 rt->rt6i_prefsrc.plen = 0;
2050
2051         rt->rt6i_flags = cfg->fc_flags;
2052
2053 install_route:
2054         rt->dst.dev = dev;
2055         rt->rt6i_idev = idev;
2056         rt->rt6i_table = table;
2057
2058         cfg->fc_nlinfo.nl_net = dev_net(dev);
2059
2060         return rt;
2061 out:
2062         if (dev)
2063                 dev_put(dev);
2064         if (idev)
2065                 in6_dev_put(idev);
2066         if (rt)
2067                 dst_free(&rt->dst);
2068
2069         return ERR_PTR(err);
2070 }
2071
2072 int ip6_route_add(struct fib6_config *cfg)
2073 {
2074         struct mx6_config mxc = { .mx = NULL, };
2075         struct rt6_info *rt;
2076         int err;
2077
2078         rt = ip6_route_info_create(cfg);
2079         if (IS_ERR(rt)) {
2080                 err = PTR_ERR(rt);
2081                 rt = NULL;
2082                 goto out;
2083         }
2084
2085         err = ip6_convert_metrics(&mxc, cfg);
2086         if (err)
2087                 goto out;
2088
2089         err = __ip6_ins_rt(rt, &cfg->fc_nlinfo, &mxc);
2090
2091         kfree(mxc.mx);
2092
2093         return err;
2094 out:
2095         if (rt)
2096                 dst_free(&rt->dst);
2097
2098         return err;
2099 }
2100
2101 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
2102 {
2103         int err;
2104         struct fib6_table *table;
2105         struct net *net = dev_net(rt->dst.dev);
2106
2107         if (rt == net->ipv6.ip6_null_entry ||
2108             rt->dst.flags & DST_NOCACHE) {
2109                 err = -ENOENT;
2110                 goto out;
2111         }
2112
2113         table = rt->rt6i_table;
2114         write_lock_bh(&table->tb6_lock);
2115         err = fib6_del(rt, info);
2116         write_unlock_bh(&table->tb6_lock);
2117
2118 out:
2119         ip6_rt_put(rt);
2120         return err;
2121 }
2122
2123 int ip6_del_rt(struct rt6_info *rt)
2124 {
2125         struct nl_info info = {
2126                 .nl_net = dev_net(rt->dst.dev),
2127         };
2128         return __ip6_del_rt(rt, &info);
2129 }
2130
2131 static int ip6_route_del(struct fib6_config *cfg)
2132 {
2133         struct fib6_table *table;
2134         struct fib6_node *fn;
2135         struct rt6_info *rt;
2136         int err = -ESRCH;
2137
2138         table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
2139         if (!table)
2140                 return err;
2141
2142         read_lock_bh(&table->tb6_lock);
2143
2144         fn = fib6_locate(&table->tb6_root,
2145                          &cfg->fc_dst, cfg->fc_dst_len,
2146                          &cfg->fc_src, cfg->fc_src_len);
2147
2148         if (fn) {
2149                 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2150                         if ((rt->rt6i_flags & RTF_CACHE) &&
2151                             !(cfg->fc_flags & RTF_CACHE))
2152                                 continue;
2153                         if (cfg->fc_ifindex &&
2154                             (!rt->dst.dev ||
2155                              rt->dst.dev->ifindex != cfg->fc_ifindex))
2156                                 continue;
2157                         if (cfg->fc_flags & RTF_GATEWAY &&
2158                             !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
2159                                 continue;
2160                         if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
2161                                 continue;
2162                         dst_hold(&rt->dst);
2163                         read_unlock_bh(&table->tb6_lock);
2164
2165                         return __ip6_del_rt(rt, &cfg->fc_nlinfo);
2166                 }
2167         }
2168         read_unlock_bh(&table->tb6_lock);
2169
2170         return err;
2171 }
2172
2173 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
2174 {
2175         struct netevent_redirect netevent;
2176         struct rt6_info *rt, *nrt = NULL;
2177         struct ndisc_options ndopts;
2178         struct inet6_dev *in6_dev;
2179         struct neighbour *neigh;
2180         struct rd_msg *msg;
2181         int optlen, on_link;
2182         u8 *lladdr;
2183
2184         optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
2185         optlen -= sizeof(*msg);
2186
2187         if (optlen < 0) {
2188                 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
2189                 return;
2190         }
2191
2192         msg = (struct rd_msg *)icmp6_hdr(skb);
2193
2194         if (ipv6_addr_is_multicast(&msg->dest)) {
2195                 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
2196                 return;
2197         }
2198
2199         on_link = 0;
2200         if (ipv6_addr_equal(&msg->dest, &msg->target)) {
2201                 on_link = 1;
2202         } else if (ipv6_addr_type(&msg->target) !=
2203                    (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
2204                 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
2205                 return;
2206         }
2207
2208         in6_dev = __in6_dev_get(skb->dev);
2209         if (!in6_dev)
2210                 return;
2211         if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
2212                 return;
2213
2214         /* RFC2461 8.1:
2215          *      The IP source address of the Redirect MUST be the same as the current
2216          *      first-hop router for the specified ICMP Destination Address.
2217          */
2218
2219         if (!ndisc_parse_options(skb->dev, msg->opt, optlen, &ndopts)) {
2220                 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
2221                 return;
2222         }
2223
2224         lladdr = NULL;
2225         if (ndopts.nd_opts_tgt_lladdr) {
2226                 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
2227                                              skb->dev);
2228                 if (!lladdr) {
2229                         net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
2230                         return;
2231                 }
2232         }
2233
2234         rt = (struct rt6_info *) dst;
2235         if (rt->rt6i_flags & RTF_REJECT) {
2236                 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
2237                 return;
2238         }
2239
2240         /* Redirect received -> path was valid.
2241          * Look, redirects are sent only in response to data packets,
2242          * so that this nexthop apparently is reachable. --ANK
2243          */
2244         dst_confirm(&rt->dst);
2245
2246         neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
2247         if (!neigh)
2248                 return;
2249
2250         /*
2251          *      We have finally decided to accept it.
2252          */
2253
2254         ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
2255                      NEIGH_UPDATE_F_WEAK_OVERRIDE|
2256                      NEIGH_UPDATE_F_OVERRIDE|
2257                      (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
2258                                      NEIGH_UPDATE_F_ISROUTER)),
2259                      NDISC_REDIRECT, &ndopts);
2260
2261         nrt = ip6_rt_cache_alloc(rt, &msg->dest, NULL);
2262         if (!nrt)
2263                 goto out;
2264
2265         nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
2266         if (on_link)
2267                 nrt->rt6i_flags &= ~RTF_GATEWAY;
2268
2269         nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
2270
2271         if (ip6_ins_rt(nrt))
2272                 goto out;
2273
2274         netevent.old = &rt->dst;
2275         netevent.new = &nrt->dst;
2276         netevent.daddr = &msg->dest;
2277         netevent.neigh = neigh;
2278         call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
2279
2280         if (rt->rt6i_flags & RTF_CACHE) {
2281                 rt = (struct rt6_info *) dst_clone(&rt->dst);
2282                 ip6_del_rt(rt);
2283         }
2284
2285 out:
2286         neigh_release(neigh);
2287 }
2288
2289 /*
2290  *      Misc support functions
2291  */
2292
2293 static void rt6_set_from(struct rt6_info *rt, struct rt6_info *from)
2294 {
2295         BUG_ON(from->dst.from);
2296
2297         rt->rt6i_flags &= ~RTF_EXPIRES;
2298         dst_hold(&from->dst);
2299         rt->dst.from = &from->dst;
2300         dst_init_metrics(&rt->dst, dst_metrics_ptr(&from->dst), true);
2301 }
2302
2303 static void ip6_rt_copy_init(struct rt6_info *rt, struct rt6_info *ort)
2304 {
2305         rt->dst.input = ort->dst.input;
2306         rt->dst.output = ort->dst.output;
2307         rt->rt6i_dst = ort->rt6i_dst;
2308         rt->dst.error = ort->dst.error;
2309         rt->rt6i_idev = ort->rt6i_idev;
2310         if (rt->rt6i_idev)
2311                 in6_dev_hold(rt->rt6i_idev);
2312         rt->dst.lastuse = jiffies;
2313         rt->rt6i_gateway = ort->rt6i_gateway;
2314         rt->rt6i_flags = ort->rt6i_flags;
2315         rt6_set_from(rt, ort);
2316         rt->rt6i_metric = ort->rt6i_metric;
2317 #ifdef CONFIG_IPV6_SUBTREES
2318         rt->rt6i_src = ort->rt6i_src;
2319 #endif
2320         rt->rt6i_prefsrc = ort->rt6i_prefsrc;
2321         rt->rt6i_table = ort->rt6i_table;
2322         rt->dst.lwtstate = lwtstate_get(ort->dst.lwtstate);
2323 }
2324
2325 #ifdef CONFIG_IPV6_ROUTE_INFO
2326 static struct rt6_info *rt6_get_route_info(struct net *net,
2327                                            const struct in6_addr *prefix, int prefixlen,
2328                                            const struct in6_addr *gwaddr, int ifindex)
2329 {
2330         struct fib6_node *fn;
2331         struct rt6_info *rt = NULL;
2332         struct fib6_table *table;
2333
2334         table = fib6_get_table(net, RT6_TABLE_INFO);
2335         if (!table)
2336                 return NULL;
2337
2338         read_lock_bh(&table->tb6_lock);
2339         fn = fib6_locate(&table->tb6_root, prefix, prefixlen, NULL, 0);
2340         if (!fn)
2341                 goto out;
2342
2343         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2344                 if (rt->dst.dev->ifindex != ifindex)
2345                         continue;
2346                 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
2347                         continue;
2348                 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
2349                         continue;
2350                 dst_hold(&rt->dst);
2351                 break;
2352         }
2353 out:
2354         read_unlock_bh(&table->tb6_lock);
2355         return rt;
2356 }
2357
2358 static struct rt6_info *rt6_add_route_info(struct net *net,
2359                                            const struct in6_addr *prefix, int prefixlen,
2360                                            const struct in6_addr *gwaddr, int ifindex,
2361                                            unsigned int pref)
2362 {
2363         struct fib6_config cfg = {
2364                 .fc_metric      = IP6_RT_PRIO_USER,
2365                 .fc_ifindex     = ifindex,
2366                 .fc_dst_len     = prefixlen,
2367                 .fc_flags       = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
2368                                   RTF_UP | RTF_PREF(pref),
2369                 .fc_nlinfo.portid = 0,
2370                 .fc_nlinfo.nlh = NULL,
2371                 .fc_nlinfo.nl_net = net,
2372         };
2373
2374         cfg.fc_table = l3mdev_fib_table_by_index(net, ifindex) ? : RT6_TABLE_INFO;
2375         cfg.fc_dst = *prefix;
2376         cfg.fc_gateway = *gwaddr;
2377
2378         /* We should treat it as a default route if prefix length is 0. */
2379         if (!prefixlen)
2380                 cfg.fc_flags |= RTF_DEFAULT;
2381
2382         ip6_route_add(&cfg);
2383
2384         return rt6_get_route_info(net, prefix, prefixlen, gwaddr, ifindex);
2385 }
2386 #endif
2387
2388 struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
2389 {
2390         struct rt6_info *rt;
2391         struct fib6_table *table;
2392
2393         table = fib6_get_table(dev_net(dev), RT6_TABLE_DFLT);
2394         if (!table)
2395                 return NULL;
2396
2397         read_lock_bh(&table->tb6_lock);
2398         for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2399                 if (dev == rt->dst.dev &&
2400                     ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
2401                     ipv6_addr_equal(&rt->rt6i_gateway, addr))
2402                         break;
2403         }
2404         if (rt)
2405                 dst_hold(&rt->dst);
2406         read_unlock_bh(&table->tb6_lock);
2407         return rt;
2408 }
2409
2410 struct rt6_info *rt6_add_dflt_router(const struct in6_addr *gwaddr,
2411                                      struct net_device *dev,
2412                                      unsigned int pref)
2413 {
2414         struct fib6_config cfg = {
2415                 .fc_table       = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT,
2416                 .fc_metric      = IP6_RT_PRIO_USER,
2417                 .fc_ifindex     = dev->ifindex,
2418                 .fc_flags       = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
2419                                   RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
2420                 .fc_nlinfo.portid = 0,
2421                 .fc_nlinfo.nlh = NULL,
2422                 .fc_nlinfo.nl_net = dev_net(dev),
2423         };
2424
2425         cfg.fc_gateway = *gwaddr;
2426
2427         ip6_route_add(&cfg);
2428
2429         return rt6_get_dflt_router(gwaddr, dev);
2430 }
2431
2432 void rt6_purge_dflt_routers(struct net *net)
2433 {
2434         struct rt6_info *rt;
2435         struct fib6_table *table;
2436
2437         /* NOTE: Keep consistent with rt6_get_dflt_router */
2438         table = fib6_get_table(net, RT6_TABLE_DFLT);
2439         if (!table)
2440                 return;
2441
2442 restart:
2443         read_lock_bh(&table->tb6_lock);
2444         for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2445                 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
2446                     (!rt->rt6i_idev || rt->rt6i_idev->cnf.accept_ra != 2)) {
2447                         dst_hold(&rt->dst);
2448                         read_unlock_bh(&table->tb6_lock);
2449                         ip6_del_rt(rt);
2450                         goto restart;
2451                 }
2452         }
2453         read_unlock_bh(&table->tb6_lock);
2454 }
2455
2456 static void rtmsg_to_fib6_config(struct net *net,
2457                                  struct in6_rtmsg *rtmsg,
2458                                  struct fib6_config *cfg)
2459 {
2460         memset(cfg, 0, sizeof(*cfg));
2461
2462         cfg->fc_table = l3mdev_fib_table_by_index(net, rtmsg->rtmsg_ifindex) ?
2463                          : RT6_TABLE_MAIN;
2464         cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
2465         cfg->fc_metric = rtmsg->rtmsg_metric;
2466         cfg->fc_expires = rtmsg->rtmsg_info;
2467         cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
2468         cfg->fc_src_len = rtmsg->rtmsg_src_len;
2469         cfg->fc_flags = rtmsg->rtmsg_flags;
2470
2471         cfg->fc_nlinfo.nl_net = net;
2472
2473         cfg->fc_dst = rtmsg->rtmsg_dst;
2474         cfg->fc_src = rtmsg->rtmsg_src;
2475         cfg->fc_gateway = rtmsg->rtmsg_gateway;
2476 }
2477
2478 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
2479 {
2480         struct fib6_config cfg;
2481         struct in6_rtmsg rtmsg;
2482         int err;
2483
2484         switch (cmd) {
2485         case SIOCADDRT:         /* Add a route */
2486         case SIOCDELRT:         /* Delete a route */
2487                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2488                         return -EPERM;
2489                 err = copy_from_user(&rtmsg, arg,
2490                                      sizeof(struct in6_rtmsg));
2491                 if (err)
2492                         return -EFAULT;
2493
2494                 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
2495
2496                 rtnl_lock();
2497                 switch (cmd) {
2498                 case SIOCADDRT:
2499                         err = ip6_route_add(&cfg);
2500                         break;
2501                 case SIOCDELRT:
2502                         err = ip6_route_del(&cfg);
2503                         break;
2504                 default:
2505                         err = -EINVAL;
2506                 }
2507                 rtnl_unlock();
2508
2509                 return err;
2510         }
2511
2512         return -EINVAL;
2513 }
2514
2515 /*
2516  *      Drop the packet on the floor
2517  */
2518
2519 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
2520 {
2521         int type;
2522         struct dst_entry *dst = skb_dst(skb);
2523         switch (ipstats_mib_noroutes) {
2524         case IPSTATS_MIB_INNOROUTES:
2525                 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
2526                 if (type == IPV6_ADDR_ANY) {
2527                         IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2528                                       IPSTATS_MIB_INADDRERRORS);
2529                         break;
2530                 }
2531                 /* FALLTHROUGH */
2532         case IPSTATS_MIB_OUTNOROUTES:
2533                 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2534                               ipstats_mib_noroutes);
2535                 break;
2536         }
2537         icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
2538         kfree_skb(skb);
2539         return 0;
2540 }
2541
2542 static int ip6_pkt_discard(struct sk_buff *skb)
2543 {
2544         return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
2545 }
2546
2547 static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb)
2548 {
2549         skb->dev = skb_dst(skb)->dev;
2550         return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
2551 }
2552
2553 static int ip6_pkt_prohibit(struct sk_buff *skb)
2554 {
2555         return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
2556 }
2557
2558 static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb)
2559 {
2560         skb->dev = skb_dst(skb)->dev;
2561         return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
2562 }
2563
2564 /*
2565  *      Allocate a dst for local (unicast / anycast) address.
2566  */
2567
2568 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
2569                                     const struct in6_addr *addr,
2570                                     bool anycast)
2571 {
2572         u32 tb_id;
2573         struct net *net = dev_net(idev->dev);
2574         struct net_device *dev = net->loopback_dev;
2575         struct rt6_info *rt;
2576
2577         /* use L3 Master device as loopback for host routes if device
2578          * is enslaved and address is not link local or multicast
2579          */
2580         if (!rt6_need_strict(addr))
2581                 dev = l3mdev_master_dev_rcu(idev->dev) ? : dev;
2582
2583         rt = ip6_dst_alloc(net, dev, DST_NOCOUNT);
2584         if (!rt)
2585                 return ERR_PTR(-ENOMEM);
2586
2587         in6_dev_hold(idev);
2588
2589         rt->dst.flags |= DST_HOST;
2590         rt->dst.input = ip6_input;
2591         rt->dst.output = ip6_output;
2592         rt->rt6i_idev = idev;
2593
2594         rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
2595         if (anycast)
2596                 rt->rt6i_flags |= RTF_ANYCAST;
2597         else
2598                 rt->rt6i_flags |= RTF_LOCAL;
2599
2600         rt->rt6i_gateway  = *addr;
2601         rt->rt6i_dst.addr = *addr;
2602         rt->rt6i_dst.plen = 128;
2603         tb_id = l3mdev_fib_table(idev->dev) ? : RT6_TABLE_LOCAL;
2604         rt->rt6i_table = fib6_get_table(net, tb_id);
2605         rt->dst.flags |= DST_NOCACHE;
2606
2607         atomic_set(&rt->dst.__refcnt, 1);
2608
2609         return rt;
2610 }
2611
2612 /* remove deleted ip from prefsrc entries */
2613 struct arg_dev_net_ip {
2614         struct net_device *dev;
2615         struct net *net;
2616         struct in6_addr *addr;
2617 };
2618
2619 static int fib6_remove_prefsrc(struct rt6_info *rt, void *arg)
2620 {
2621         struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
2622         struct net *net = ((struct arg_dev_net_ip *)arg)->net;
2623         struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
2624
2625         if (((void *)rt->dst.dev == dev || !dev) &&
2626             rt != net->ipv6.ip6_null_entry &&
2627             ipv6_addr_equal(addr, &rt->rt6i_prefsrc.addr)) {
2628                 /* remove prefsrc entry */
2629                 rt->rt6i_prefsrc.plen = 0;
2630         }
2631         return 0;
2632 }
2633
2634 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
2635 {
2636         struct net *net = dev_net(ifp->idev->dev);
2637         struct arg_dev_net_ip adni = {
2638                 .dev = ifp->idev->dev,
2639                 .net = net,
2640                 .addr = &ifp->addr,
2641         };
2642         fib6_clean_all(net, fib6_remove_prefsrc, &adni);
2643 }
2644
2645 #define RTF_RA_ROUTER           (RTF_ADDRCONF | RTF_DEFAULT | RTF_GATEWAY)
2646 #define RTF_CACHE_GATEWAY       (RTF_GATEWAY | RTF_CACHE)
2647
2648 /* Remove routers and update dst entries when gateway turn into host. */
2649 static int fib6_clean_tohost(struct rt6_info *rt, void *arg)
2650 {
2651         struct in6_addr *gateway = (struct in6_addr *)arg;
2652
2653         if ((((rt->rt6i_flags & RTF_RA_ROUTER) == RTF_RA_ROUTER) ||
2654              ((rt->rt6i_flags & RTF_CACHE_GATEWAY) == RTF_CACHE_GATEWAY)) &&
2655              ipv6_addr_equal(gateway, &rt->rt6i_gateway)) {
2656                 return -1;
2657         }
2658         return 0;
2659 }
2660
2661 void rt6_clean_tohost(struct net *net, struct in6_addr *gateway)
2662 {
2663         fib6_clean_all(net, fib6_clean_tohost, gateway);
2664 }
2665
2666 struct arg_dev_net {
2667         struct net_device *dev;
2668         struct net *net;
2669 };
2670
2671 static int fib6_ifdown(struct rt6_info *rt, void *arg)
2672 {
2673         const struct arg_dev_net *adn = arg;
2674         const struct net_device *dev = adn->dev;
2675
2676         if ((rt->dst.dev == dev || !dev) &&
2677             rt != adn->net->ipv6.ip6_null_entry)
2678                 return -1;
2679
2680         return 0;
2681 }
2682
2683 void rt6_ifdown(struct net *net, struct net_device *dev)
2684 {
2685         struct arg_dev_net adn = {
2686                 .dev = dev,
2687                 .net = net,
2688         };
2689
2690         fib6_clean_all(net, fib6_ifdown, &adn);
2691         icmp6_clean_all(fib6_ifdown, &adn);
2692         if (dev)
2693                 rt6_uncached_list_flush_dev(net, dev);
2694 }
2695
2696 struct rt6_mtu_change_arg {
2697         struct net_device *dev;
2698         unsigned int mtu;
2699 };
2700
2701 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2702 {
2703         struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2704         struct inet6_dev *idev;
2705
2706         /* In IPv6 pmtu discovery is not optional,
2707            so that RTAX_MTU lock cannot disable it.
2708            We still use this lock to block changes
2709            caused by addrconf/ndisc.
2710         */
2711
2712         idev = __in6_dev_get(arg->dev);
2713         if (!idev)
2714                 return 0;
2715
2716         /* For administrative MTU increase, there is no way to discover
2717            IPv6 PMTU increase, so PMTU increase should be updated here.
2718            Since RFC 1981 doesn't include administrative MTU increase
2719            update PMTU increase is a MUST. (i.e. jumbo frame)
2720          */
2721         /*
2722            If new MTU is less than route PMTU, this new MTU will be the
2723            lowest MTU in the path, update the route PMTU to reflect PMTU
2724            decreases; if new MTU is greater than route PMTU, and the
2725            old MTU is the lowest MTU in the path, update the route PMTU
2726            to reflect the increase. In this case if the other nodes' MTU
2727            also have the lowest MTU, TOO BIG MESSAGE will be lead to
2728            PMTU discouvery.
2729          */
2730         if (rt->dst.dev == arg->dev &&
2731             !dst_metric_locked(&rt->dst, RTAX_MTU)) {
2732                 if (rt->rt6i_flags & RTF_CACHE) {
2733                         /* For RTF_CACHE with rt6i_pmtu == 0
2734                          * (i.e. a redirected route),
2735                          * the metrics of its rt->dst.from has already
2736                          * been updated.
2737                          */
2738                         if (rt->rt6i_pmtu && rt->rt6i_pmtu > arg->mtu)
2739                                 rt->rt6i_pmtu = arg->mtu;
2740                 } else if (dst_mtu(&rt->dst) >= arg->mtu ||
2741                            (dst_mtu(&rt->dst) < arg->mtu &&
2742                             dst_mtu(&rt->dst) == idev->cnf.mtu6)) {
2743                         dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
2744                 }
2745         }
2746         return 0;
2747 }
2748
2749 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
2750 {
2751         struct rt6_mtu_change_arg arg = {
2752                 .dev = dev,
2753                 .mtu = mtu,
2754         };
2755
2756         fib6_clean_all(dev_net(dev), rt6_mtu_change_route, &arg);
2757 }
2758
2759 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2760         [RTA_GATEWAY]           = { .len = sizeof(struct in6_addr) },
2761         [RTA_OIF]               = { .type = NLA_U32 },
2762         [RTA_IIF]               = { .type = NLA_U32 },
2763         [RTA_PRIORITY]          = { .type = NLA_U32 },
2764         [RTA_METRICS]           = { .type = NLA_NESTED },
2765         [RTA_MULTIPATH]         = { .len = sizeof(struct rtnexthop) },
2766         [RTA_PREF]              = { .type = NLA_U8 },
2767         [RTA_ENCAP_TYPE]        = { .type = NLA_U16 },
2768         [RTA_ENCAP]             = { .type = NLA_NESTED },
2769         [RTA_EXPIRES]           = { .type = NLA_U32 },
2770 };
2771
2772 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2773                               struct fib6_config *cfg)
2774 {
2775         struct rtmsg *rtm;
2776         struct nlattr *tb[RTA_MAX+1];
2777         unsigned int pref;
2778         int err;
2779
2780         err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2781         if (err < 0)
2782                 goto errout;
2783
2784         err = -EINVAL;
2785         rtm = nlmsg_data(nlh);
2786         memset(cfg, 0, sizeof(*cfg));
2787
2788         cfg->fc_table = rtm->rtm_table;
2789         cfg->fc_dst_len = rtm->rtm_dst_len;
2790         cfg->fc_src_len = rtm->rtm_src_len;
2791         cfg->fc_flags = RTF_UP;
2792         cfg->fc_protocol = rtm->rtm_protocol;
2793         cfg->fc_type = rtm->rtm_type;
2794
2795         if (rtm->rtm_type == RTN_UNREACHABLE ||
2796             rtm->rtm_type == RTN_BLACKHOLE ||
2797             rtm->rtm_type == RTN_PROHIBIT ||
2798             rtm->rtm_type == RTN_THROW)
2799                 cfg->fc_flags |= RTF_REJECT;
2800
2801         if (rtm->rtm_type == RTN_LOCAL)
2802                 cfg->fc_flags |= RTF_LOCAL;
2803
2804         if (rtm->rtm_flags & RTM_F_CLONED)
2805                 cfg->fc_flags |= RTF_CACHE;
2806
2807         cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid;
2808         cfg->fc_nlinfo.nlh = nlh;
2809         cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2810
2811         if (tb[RTA_GATEWAY]) {
2812                 cfg->fc_gateway = nla_get_in6_addr(tb[RTA_GATEWAY]);
2813                 cfg->fc_flags |= RTF_GATEWAY;
2814         }
2815
2816         if (tb[RTA_DST]) {
2817                 int plen = (rtm->rtm_dst_len + 7) >> 3;
2818
2819                 if (nla_len(tb[RTA_DST]) < plen)
2820                         goto errout;
2821
2822                 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2823         }
2824
2825         if (tb[RTA_SRC]) {
2826                 int plen = (rtm->rtm_src_len + 7) >> 3;
2827
2828                 if (nla_len(tb[RTA_SRC]) < plen)
2829                         goto errout;
2830
2831                 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2832         }
2833
2834         if (tb[RTA_PREFSRC])
2835                 cfg->fc_prefsrc = nla_get_in6_addr(tb[RTA_PREFSRC]);
2836
2837         if (tb[RTA_OIF])
2838                 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2839
2840         if (tb[RTA_PRIORITY])
2841                 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2842
2843         if (tb[RTA_METRICS]) {
2844                 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2845                 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2846         }
2847
2848         if (tb[RTA_TABLE])
2849                 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2850
2851         if (tb[RTA_MULTIPATH]) {
2852                 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
2853                 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
2854         }
2855
2856         if (tb[RTA_PREF]) {
2857                 pref = nla_get_u8(tb[RTA_PREF]);
2858                 if (pref != ICMPV6_ROUTER_PREF_LOW &&
2859                     pref != ICMPV6_ROUTER_PREF_HIGH)
2860                         pref = ICMPV6_ROUTER_PREF_MEDIUM;
2861                 cfg->fc_flags |= RTF_PREF(pref);
2862         }
2863
2864         if (tb[RTA_ENCAP])
2865                 cfg->fc_encap = tb[RTA_ENCAP];
2866
2867         if (tb[RTA_ENCAP_TYPE])
2868                 cfg->fc_encap_type = nla_get_u16(tb[RTA_ENCAP_TYPE]);
2869
2870         if (tb[RTA_EXPIRES]) {
2871                 unsigned long timeout = addrconf_timeout_fixup(nla_get_u32(tb[RTA_EXPIRES]), HZ);
2872
2873                 if (addrconf_finite_timeout(timeout)) {
2874                         cfg->fc_expires = jiffies_to_clock_t(timeout * HZ);
2875                         cfg->fc_flags |= RTF_EXPIRES;
2876                 }
2877         }
2878
2879         err = 0;
2880 errout:
2881         return err;
2882 }
2883
2884 struct rt6_nh {
2885         struct rt6_info *rt6_info;
2886         struct fib6_config r_cfg;
2887         struct mx6_config mxc;
2888         struct list_head next;
2889 };
2890
2891 static void ip6_print_replace_route_err(struct list_head *rt6_nh_list)
2892 {
2893         struct rt6_nh *nh;
2894
2895         list_for_each_entry(nh, rt6_nh_list, next) {
2896                 pr_warn("IPV6: multipath route replace failed (check consistency of installed routes): %pI6 nexthop %pI6 ifi %d\n",
2897                         &nh->r_cfg.fc_dst, &nh->r_cfg.fc_gateway,
2898                         nh->r_cfg.fc_ifindex);
2899         }
2900 }
2901
2902 static int ip6_route_info_append(struct list_head *rt6_nh_list,
2903                                  struct rt6_info *rt, struct fib6_config *r_cfg)
2904 {
2905         struct rt6_nh *nh;
2906         struct rt6_info *rtnh;
2907         int err = -EEXIST;
2908
2909         list_for_each_entry(nh, rt6_nh_list, next) {
2910                 /* check if rt6_info already exists */
2911                 rtnh = nh->rt6_info;
2912
2913                 if (rtnh->dst.dev == rt->dst.dev &&
2914                     rtnh->rt6i_idev == rt->rt6i_idev &&
2915                     ipv6_addr_equal(&rtnh->rt6i_gateway,
2916                                     &rt->rt6i_gateway))
2917                         return err;
2918         }
2919
2920         nh = kzalloc(sizeof(*nh), GFP_KERNEL);
2921         if (!nh)
2922                 return -ENOMEM;
2923         nh->rt6_info = rt;
2924         err = ip6_convert_metrics(&nh->mxc, r_cfg);
2925         if (err) {
2926                 kfree(nh);
2927                 return err;
2928         }
2929         memcpy(&nh->r_cfg, r_cfg, sizeof(*r_cfg));
2930         list_add_tail(&nh->next, rt6_nh_list);
2931
2932         return 0;
2933 }
2934
2935 static int ip6_route_multipath_add(struct fib6_config *cfg)
2936 {
2937         struct fib6_config r_cfg;
2938         struct rtnexthop *rtnh;
2939         struct rt6_info *rt;
2940         struct rt6_nh *err_nh;
2941         struct rt6_nh *nh, *nh_safe;
2942         int remaining;
2943         int attrlen;
2944         int err = 1;
2945         int nhn = 0;
2946         int replace = (cfg->fc_nlinfo.nlh &&
2947                        (cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_REPLACE));
2948         LIST_HEAD(rt6_nh_list);
2949
2950         remaining = cfg->fc_mp_len;
2951         rtnh = (struct rtnexthop *)cfg->fc_mp;
2952
2953         /* Parse a Multipath Entry and build a list (rt6_nh_list) of
2954          * rt6_info structs per nexthop
2955          */
2956         while (rtnh_ok(rtnh, remaining)) {
2957                 memcpy(&r_cfg, cfg, sizeof(*cfg));
2958                 if (rtnh->rtnh_ifindex)
2959                         r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
2960
2961                 attrlen = rtnh_attrlen(rtnh);
2962                 if (attrlen > 0) {
2963                         struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
2964
2965                         nla = nla_find(attrs, attrlen, RTA_GATEWAY);
2966                         if (nla) {
2967                                 r_cfg.fc_gateway = nla_get_in6_addr(nla);
2968                                 r_cfg.fc_flags |= RTF_GATEWAY;
2969                         }
2970                         r_cfg.fc_encap = nla_find(attrs, attrlen, RTA_ENCAP);
2971                         nla = nla_find(attrs, attrlen, RTA_ENCAP_TYPE);
2972                         if (nla)
2973                                 r_cfg.fc_encap_type = nla_get_u16(nla);
2974                 }
2975
2976                 rt = ip6_route_info_create(&r_cfg);
2977                 if (IS_ERR(rt)) {
2978                         err = PTR_ERR(rt);
2979                         rt = NULL;
2980                         goto cleanup;
2981                 }
2982
2983                 err = ip6_route_info_append(&rt6_nh_list, rt, &r_cfg);
2984                 if (err) {
2985                         dst_free(&rt->dst);
2986                         goto cleanup;
2987                 }
2988
2989                 rtnh = rtnh_next(rtnh, &remaining);
2990         }
2991
2992         err_nh = NULL;
2993         list_for_each_entry(nh, &rt6_nh_list, next) {
2994                 err = __ip6_ins_rt(nh->rt6_info, &cfg->fc_nlinfo, &nh->mxc);
2995                 /* nh->rt6_info is used or freed at this point, reset to NULL*/
2996                 nh->rt6_info = NULL;
2997                 if (err) {
2998                         if (replace && nhn)
2999                                 ip6_print_replace_route_err(&rt6_nh_list);
3000                         err_nh = nh;
3001                         goto add_errout;
3002                 }
3003
3004                 /* Because each route is added like a single route we remove
3005                  * these flags after the first nexthop: if there is a collision,
3006                  * we have already failed to add the first nexthop:
3007                  * fib6_add_rt2node() has rejected it; when replacing, old
3008                  * nexthops have been replaced by first new, the rest should
3009                  * be added to it.
3010                  */
3011                 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~(NLM_F_EXCL |
3012                                                      NLM_F_REPLACE);
3013                 nhn++;
3014         }
3015
3016         goto cleanup;
3017
3018 add_errout:
3019         /* Delete routes that were already added */
3020         list_for_each_entry(nh, &rt6_nh_list, next) {
3021                 if (err_nh == nh)
3022                         break;
3023                 ip6_route_del(&nh->r_cfg);
3024         }
3025
3026 cleanup:
3027         list_for_each_entry_safe(nh, nh_safe, &rt6_nh_list, next) {
3028                 if (nh->rt6_info)
3029                         dst_free(&nh->rt6_info->dst);
3030                 kfree(nh->mxc.mx);
3031                 list_del(&nh->next);
3032                 kfree(nh);
3033         }
3034
3035         return err;
3036 }
3037
3038 static int ip6_route_multipath_del(struct fib6_config *cfg)
3039 {
3040         struct fib6_config r_cfg;
3041         struct rtnexthop *rtnh;
3042         int remaining;
3043         int attrlen;
3044         int err = 1, last_err = 0;
3045
3046         remaining = cfg->fc_mp_len;
3047         rtnh = (struct rtnexthop *)cfg->fc_mp;
3048
3049         /* Parse a Multipath Entry */
3050         while (rtnh_ok(rtnh, remaining)) {
3051                 memcpy(&r_cfg, cfg, sizeof(*cfg));
3052                 if (rtnh->rtnh_ifindex)
3053                         r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
3054
3055                 attrlen = rtnh_attrlen(rtnh);
3056                 if (attrlen > 0) {
3057                         struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
3058
3059                         nla = nla_find(attrs, attrlen, RTA_GATEWAY);
3060                         if (nla) {
3061                                 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
3062                                 r_cfg.fc_flags |= RTF_GATEWAY;
3063                         }
3064                 }
3065                 err = ip6_route_del(&r_cfg);
3066                 if (err)
3067                         last_err = err;
3068
3069                 rtnh = rtnh_next(rtnh, &remaining);
3070         }
3071
3072         return last_err;
3073 }
3074
3075 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh)
3076 {
3077         struct fib6_config cfg;
3078         int err;
3079
3080         err = rtm_to_fib6_config(skb, nlh, &cfg);
3081         if (err < 0)
3082                 return err;
3083
3084         if (cfg.fc_mp)
3085                 return ip6_route_multipath_del(&cfg);
3086         else
3087                 return ip6_route_del(&cfg);
3088 }
3089
3090 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh)
3091 {
3092         struct fib6_config cfg;
3093         int err;
3094
3095         err = rtm_to_fib6_config(skb, nlh, &cfg);
3096         if (err < 0)
3097                 return err;
3098
3099         if (cfg.fc_mp)
3100                 return ip6_route_multipath_add(&cfg);
3101         else
3102                 return ip6_route_add(&cfg);
3103 }
3104
3105 static inline size_t rt6_nlmsg_size(struct rt6_info *rt)
3106 {
3107         return NLMSG_ALIGN(sizeof(struct rtmsg))
3108                + nla_total_size(16) /* RTA_SRC */
3109                + nla_total_size(16) /* RTA_DST */
3110                + nla_total_size(16) /* RTA_GATEWAY */
3111                + nla_total_size(16) /* RTA_PREFSRC */
3112                + nla_total_size(4) /* RTA_TABLE */
3113                + nla_total_size(4) /* RTA_IIF */
3114                + nla_total_size(4) /* RTA_OIF */
3115                + nla_total_size(4) /* RTA_PRIORITY */
3116                + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
3117                + nla_total_size(sizeof(struct rta_cacheinfo))
3118                + nla_total_size(TCP_CA_NAME_MAX) /* RTAX_CC_ALGO */
3119                + nla_total_size(1) /* RTA_PREF */
3120                + lwtunnel_get_encap_size(rt->dst.lwtstate);
3121 }
3122
3123 static int rt6_fill_node(struct net *net,
3124                          struct sk_buff *skb, struct rt6_info *rt,
3125                          struct in6_addr *dst, struct in6_addr *src,
3126                          int iif, int type, u32 portid, u32 seq,
3127                          int prefix, int nowait, unsigned int flags)
3128 {
3129         u32 metrics[RTAX_MAX];
3130         struct rtmsg *rtm;
3131         struct nlmsghdr *nlh;
3132         long expires;
3133         u32 table;
3134
3135         if (prefix) {   /* user wants prefix routes only */
3136                 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
3137                         /* success since this is not a prefix route */
3138                         return 1;
3139                 }
3140         }
3141
3142         nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
3143         if (!nlh)
3144                 return -EMSGSIZE;
3145
3146         rtm = nlmsg_data(nlh);
3147         rtm->rtm_family = AF_INET6;
3148         rtm->rtm_dst_len = rt->rt6i_dst.plen;
3149         rtm->rtm_src_len = rt->rt6i_src.plen;
3150         rtm->rtm_tos = 0;
3151         if (rt->rt6i_table)
3152                 table = rt->rt6i_table->tb6_id;
3153         else
3154                 table = RT6_TABLE_UNSPEC;
3155         rtm->rtm_table = table;
3156         if (nla_put_u32(skb, RTA_TABLE, table))
3157                 goto nla_put_failure;
3158         if (rt->rt6i_flags & RTF_REJECT) {
3159                 switch (rt->dst.error) {
3160                 case -EINVAL:
3161                         rtm->rtm_type = RTN_BLACKHOLE;
3162                         break;
3163                 case -EACCES:
3164                         rtm->rtm_type = RTN_PROHIBIT;
3165                         break;
3166                 case -EAGAIN:
3167                         rtm->rtm_type = RTN_THROW;
3168                         break;
3169                 default:
3170                         rtm->rtm_type = RTN_UNREACHABLE;
3171                         break;
3172                 }
3173         }
3174         else if (rt->rt6i_flags & RTF_LOCAL)
3175                 rtm->rtm_type = RTN_LOCAL;
3176         else if (rt->dst.dev && (rt->dst.dev->flags & IFF_LOOPBACK))
3177                 rtm->rtm_type = RTN_LOCAL;
3178         else
3179                 rtm->rtm_type = RTN_UNICAST;
3180         rtm->rtm_flags = 0;
3181         if (!netif_carrier_ok(rt->dst.dev)) {
3182                 rtm->rtm_flags |= RTNH_F_LINKDOWN;
3183                 if (rt->rt6i_idev->cnf.ignore_routes_with_linkdown)
3184                         rtm->rtm_flags |= RTNH_F_DEAD;
3185         }
3186         rtm->rtm_scope = RT_SCOPE_UNIVERSE;
3187         rtm->rtm_protocol = rt->rt6i_protocol;
3188         if (rt->rt6i_flags & RTF_DYNAMIC)
3189                 rtm->rtm_protocol = RTPROT_REDIRECT;
3190         else if (rt->rt6i_flags & RTF_ADDRCONF) {
3191                 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ROUTEINFO))
3192                         rtm->rtm_protocol = RTPROT_RA;
3193                 else
3194                         rtm->rtm_protocol = RTPROT_KERNEL;
3195         }
3196
3197         if (rt->rt6i_flags & RTF_CACHE)
3198                 rtm->rtm_flags |= RTM_F_CLONED;
3199
3200         if (dst) {
3201                 if (nla_put_in6_addr(skb, RTA_DST, dst))
3202                         goto nla_put_failure;
3203                 rtm->rtm_dst_len = 128;
3204         } else if (rtm->rtm_dst_len)
3205                 if (nla_put_in6_addr(skb, RTA_DST, &rt->rt6i_dst.addr))
3206                         goto nla_put_failure;
3207 #ifdef CONFIG_IPV6_SUBTREES
3208         if (src) {
3209                 if (nla_put_in6_addr(skb, RTA_SRC, src))
3210                         goto nla_put_failure;
3211                 rtm->rtm_src_len = 128;
3212         } else if (rtm->rtm_src_len &&
3213                    nla_put_in6_addr(skb, RTA_SRC, &rt->rt6i_src.addr))
3214                 goto nla_put_failure;
3215 #endif
3216         if (iif) {
3217 #ifdef CONFIG_IPV6_MROUTE
3218                 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
3219                         int err = ip6mr_get_route(net, skb, rtm, nowait);
3220                         if (err <= 0) {
3221                                 if (!nowait) {
3222                                         if (err == 0)
3223                                                 return 0;
3224                                         goto nla_put_failure;
3225                                 } else {
3226                                         if (err == -EMSGSIZE)
3227                                                 goto nla_put_failure;
3228                                 }
3229                         }
3230                 } else
3231 #endif
3232                         if (nla_put_u32(skb, RTA_IIF, iif))
3233                                 goto nla_put_failure;
3234         } else if (dst) {
3235                 struct in6_addr saddr_buf;
3236                 if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0 &&
3237                     nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
3238                         goto nla_put_failure;
3239         }
3240
3241         if (rt->rt6i_prefsrc.plen) {
3242                 struct in6_addr saddr_buf;
3243                 saddr_buf = rt->rt6i_prefsrc.addr;
3244                 if (nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
3245                         goto nla_put_failure;
3246         }
3247
3248         memcpy(metrics, dst_metrics_ptr(&rt->dst), sizeof(metrics));
3249         if (rt->rt6i_pmtu)
3250                 metrics[RTAX_MTU - 1] = rt->rt6i_pmtu;
3251         if (rtnetlink_put_metrics(skb, metrics) < 0)
3252                 goto nla_put_failure;
3253
3254         if (rt->rt6i_flags & RTF_GATEWAY) {
3255                 if (nla_put_in6_addr(skb, RTA_GATEWAY, &rt->rt6i_gateway) < 0)
3256                         goto nla_put_failure;
3257         }
3258
3259         if (rt->dst.dev &&
3260             nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex))
3261                 goto nla_put_failure;
3262         if (nla_put_u32(skb, RTA_PRIORITY, rt->rt6i_metric))
3263                 goto nla_put_failure;
3264
3265         expires = (rt->rt6i_flags & RTF_EXPIRES) ? rt->dst.expires - jiffies : 0;
3266
3267         if (rtnl_put_cacheinfo(skb, &rt->dst, 0, expires, rt->dst.error) < 0)
3268                 goto nla_put_failure;
3269
3270         if (nla_put_u8(skb, RTA_PREF, IPV6_EXTRACT_PREF(rt->rt6i_flags)))
3271                 goto nla_put_failure;
3272
3273         lwtunnel_fill_encap(skb, rt->dst.lwtstate);
3274
3275         nlmsg_end(skb, nlh);
3276         return 0;
3277
3278 nla_put_failure:
3279         nlmsg_cancel(skb, nlh);
3280         return -EMSGSIZE;
3281 }
3282
3283 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
3284 {
3285         struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
3286         int prefix;
3287
3288         if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
3289                 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
3290                 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
3291         } else
3292                 prefix = 0;
3293
3294         return rt6_fill_node(arg->net,
3295                      arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
3296                      NETLINK_CB(arg->cb->skb).portid, arg->cb->nlh->nlmsg_seq,
3297                      prefix, 0, NLM_F_MULTI);
3298 }
3299
3300 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh)
3301 {
3302         struct net *net = sock_net(in_skb->sk);
3303         struct nlattr *tb[RTA_MAX+1];
3304         struct rt6_info *rt;
3305         struct sk_buff *skb;
3306         struct rtmsg *rtm;
3307         struct flowi6 fl6;
3308         int err, iif = 0, oif = 0;
3309
3310         err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
3311         if (err < 0)
3312                 goto errout;
3313
3314         err = -EINVAL;
3315         memset(&fl6, 0, sizeof(fl6));
3316         rtm = nlmsg_data(nlh);
3317         fl6.flowlabel = ip6_make_flowinfo(rtm->rtm_tos, 0);
3318
3319         if (tb[RTA_SRC]) {
3320                 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
3321                         goto errout;
3322
3323                 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
3324         }
3325
3326         if (tb[RTA_DST]) {
3327                 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
3328                         goto errout;
3329
3330                 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
3331         }
3332
3333         if (tb[RTA_IIF])
3334                 iif = nla_get_u32(tb[RTA_IIF]);
3335
3336         if (tb[RTA_OIF])
3337                 oif = nla_get_u32(tb[RTA_OIF]);
3338
3339         if (tb[RTA_MARK])
3340                 fl6.flowi6_mark = nla_get_u32(tb[RTA_MARK]);
3341
3342         if (iif) {
3343                 struct net_device *dev;
3344                 int flags = 0;
3345
3346                 dev = __dev_get_by_index(net, iif);
3347                 if (!dev) {
3348                         err = -ENODEV;
3349                         goto errout;
3350                 }
3351
3352                 fl6.flowi6_iif = iif;
3353
3354                 if (!ipv6_addr_any(&fl6.saddr))
3355                         flags |= RT6_LOOKUP_F_HAS_SADDR;
3356
3357                 rt = (struct rt6_info *)ip6_route_input_lookup(net, dev, &fl6,
3358                                                                flags);
3359         } else {
3360                 fl6.flowi6_oif = oif;
3361
3362                 rt = (struct rt6_info *)ip6_route_output(net, NULL, &fl6);
3363         }
3364
3365         skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
3366         if (!skb) {
3367                 ip6_rt_put(rt);
3368                 err = -ENOBUFS;
3369                 goto errout;
3370         }
3371
3372         /* Reserve room for dummy headers, this skb can pass
3373            through good chunk of routing engine.
3374          */
3375         skb_reset_mac_header(skb);
3376         skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
3377
3378         skb_dst_set(skb, &rt->dst);
3379
3380         err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
3381                             RTM_NEWROUTE, NETLINK_CB(in_skb).portid,
3382                             nlh->nlmsg_seq, 0, 0, 0);
3383         if (err < 0) {
3384                 kfree_skb(skb);
3385                 goto errout;
3386         }
3387
3388         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
3389 errout:
3390         return err;
3391 }
3392
3393 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info,
3394                      unsigned int nlm_flags)
3395 {
3396         struct sk_buff *skb;
3397         struct net *net = info->nl_net;
3398         u32 seq;
3399         int err;
3400
3401         err = -ENOBUFS;
3402         seq = info->nlh ? info->nlh->nlmsg_seq : 0;
3403
3404         skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
3405         if (!skb)
3406                 goto errout;
3407
3408         err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
3409                                 event, info->portid, seq, 0, 0, nlm_flags);
3410         if (err < 0) {
3411                 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
3412                 WARN_ON(err == -EMSGSIZE);
3413                 kfree_skb(skb);
3414                 goto errout;
3415         }
3416         rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
3417                     info->nlh, gfp_any());
3418         return;
3419 errout:
3420         if (err < 0)
3421                 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
3422 }
3423
3424 static int ip6_route_dev_notify(struct notifier_block *this,
3425                                 unsigned long event, void *ptr)
3426 {
3427         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3428         struct net *net = dev_net(dev);
3429
3430         if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
3431                 net->ipv6.ip6_null_entry->dst.dev = dev;
3432                 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
3433 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3434                 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
3435                 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
3436                 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
3437                 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
3438 #endif
3439         }
3440
3441         return NOTIFY_OK;
3442 }
3443
3444 /*
3445  *      /proc
3446  */
3447
3448 #ifdef CONFIG_PROC_FS
3449
3450 static const struct file_operations ipv6_route_proc_fops = {
3451         .owner          = THIS_MODULE,
3452         .open           = ipv6_route_open,
3453         .read           = seq_read,
3454         .llseek         = seq_lseek,
3455         .release        = seq_release_net,
3456 };
3457
3458 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
3459 {
3460         struct net *net = (struct net *)seq->private;
3461         seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
3462                    net->ipv6.rt6_stats->fib_nodes,
3463                    net->ipv6.rt6_stats->fib_route_nodes,
3464                    net->ipv6.rt6_stats->fib_rt_alloc,
3465                    net->ipv6.rt6_stats->fib_rt_entries,
3466                    net->ipv6.rt6_stats->fib_rt_cache,
3467                    dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
3468                    net->ipv6.rt6_stats->fib_discarded_routes);
3469
3470         return 0;
3471 }
3472
3473 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
3474 {
3475         return single_open_net(inode, file, rt6_stats_seq_show);
3476 }
3477
3478 static const struct file_operations rt6_stats_seq_fops = {
3479         .owner   = THIS_MODULE,
3480         .open    = rt6_stats_seq_open,
3481         .read    = seq_read,
3482         .llseek  = seq_lseek,
3483         .release = single_release_net,
3484 };
3485 #endif  /* CONFIG_PROC_FS */
3486
3487 #ifdef CONFIG_SYSCTL
3488
3489 static
3490 int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
3491                               void __user *buffer, size_t *lenp, loff_t *ppos)
3492 {
3493         struct net *net;
3494         int delay;
3495         if (!write)
3496                 return -EINVAL;
3497
3498         net = (struct net *)ctl->extra1;
3499         delay = net->ipv6.sysctl.flush_delay;
3500         proc_dointvec(ctl, write, buffer, lenp, ppos);
3501         fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
3502         return 0;
3503 }
3504
3505 struct ctl_table ipv6_route_table_template[] = {
3506         {
3507                 .procname       =       "flush",
3508                 .data           =       &init_net.ipv6.sysctl.flush_delay,
3509                 .maxlen         =       sizeof(int),
3510                 .mode           =       0200,
3511                 .proc_handler   =       ipv6_sysctl_rtcache_flush
3512         },
3513         {
3514                 .procname       =       "gc_thresh",
3515                 .data           =       &ip6_dst_ops_template.gc_thresh,
3516                 .maxlen         =       sizeof(int),
3517                 .mode           =       0644,
3518                 .proc_handler   =       proc_dointvec,
3519         },
3520         {
3521                 .procname       =       "max_size",
3522                 .data           =       &init_net.ipv6.sysctl.ip6_rt_max_size,
3523                 .maxlen         =       sizeof(int),
3524                 .mode           =       0644,
3525                 .proc_handler   =       proc_dointvec,
3526         },
3527         {
3528                 .procname       =       "gc_min_interval",
3529                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
3530                 .maxlen         =       sizeof(int),
3531                 .mode           =       0644,
3532                 .proc_handler   =       proc_dointvec_jiffies,
3533         },
3534         {
3535                 .procname       =       "gc_timeout",
3536                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
3537                 .maxlen         =       sizeof(int),
3538                 .mode           =       0644,
3539                 .proc_handler   =       proc_dointvec_jiffies,
3540         },
3541         {
3542                 .procname       =       "gc_interval",
3543                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_interval,
3544                 .maxlen         =       sizeof(int),
3545                 .mode           =       0644,
3546                 .proc_handler   =       proc_dointvec_jiffies,
3547         },
3548         {
3549                 .procname       =       "gc_elasticity",
3550                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
3551                 .maxlen         =       sizeof(int),
3552                 .mode           =       0644,
3553                 .proc_handler   =       proc_dointvec,
3554         },
3555         {
3556                 .procname       =       "mtu_expires",
3557                 .data           =       &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
3558                 .maxlen         =       sizeof(int),
3559                 .mode           =       0644,
3560                 .proc_handler   =       proc_dointvec_jiffies,
3561         },
3562         {
3563                 .procname       =       "min_adv_mss",
3564                 .data           =       &init_net.ipv6.sysctl.ip6_rt_min_advmss,
3565                 .maxlen         =       sizeof(int),
3566                 .mode           =       0644,
3567                 .proc_handler   =       proc_dointvec,
3568         },
3569         {
3570                 .procname       =       "gc_min_interval_ms",
3571                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
3572                 .maxlen         =       sizeof(int),
3573                 .mode           =       0644,
3574                 .proc_handler   =       proc_dointvec_ms_jiffies,
3575         },
3576         { }
3577 };
3578
3579 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
3580 {
3581         struct ctl_table *table;
3582
3583         table = kmemdup(ipv6_route_table_template,
3584                         sizeof(ipv6_route_table_template),
3585                         GFP_KERNEL);
3586
3587         if (table) {
3588                 table[0].data = &net->ipv6.sysctl.flush_delay;
3589                 table[0].extra1 = net;
3590                 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
3591                 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
3592                 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
3593                 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
3594                 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
3595                 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
3596                 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
3597                 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
3598                 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
3599
3600                 /* Don't export sysctls to unprivileged users */
3601                 if (net->user_ns != &init_user_ns)
3602                         table[0].procname = NULL;
3603         }
3604
3605         return table;
3606 }
3607 #endif
3608
3609 static int __net_init ip6_route_net_init(struct net *net)
3610 {
3611         int ret = -ENOMEM;
3612
3613         memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
3614                sizeof(net->ipv6.ip6_dst_ops));
3615
3616         if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
3617                 goto out_ip6_dst_ops;
3618
3619         net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
3620                                            sizeof(*net->ipv6.ip6_null_entry),
3621                                            GFP_KERNEL);
3622         if (!net->ipv6.ip6_null_entry)
3623                 goto out_ip6_dst_entries;
3624         net->ipv6.ip6_null_entry->dst.path =
3625                 (struct dst_entry *)net->ipv6.ip6_null_entry;
3626         net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3627         dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
3628                          ip6_template_metrics, true);
3629
3630 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3631         net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
3632                                                sizeof(*net->ipv6.ip6_prohibit_entry),
3633                                                GFP_KERNEL);
3634         if (!net->ipv6.ip6_prohibit_entry)
3635                 goto out_ip6_null_entry;
3636         net->ipv6.ip6_prohibit_entry->dst.path =
3637                 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
3638         net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3639         dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
3640                          ip6_template_metrics, true);
3641
3642         net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
3643                                                sizeof(*net->ipv6.ip6_blk_hole_entry),
3644                                                GFP_KERNEL);
3645         if (!net->ipv6.ip6_blk_hole_entry)
3646                 goto out_ip6_prohibit_entry;
3647         net->ipv6.ip6_blk_hole_entry->dst.path =
3648                 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
3649         net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3650         dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
3651                          ip6_template_metrics, true);
3652 #endif
3653
3654         net->ipv6.sysctl.flush_delay = 0;
3655         net->ipv6.sysctl.ip6_rt_max_size = 4096;
3656         net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
3657         net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
3658         net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
3659         net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
3660         net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
3661         net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
3662
3663         net->ipv6.ip6_rt_gc_expire = 30*HZ;
3664
3665         ret = 0;
3666 out:
3667         return ret;
3668
3669 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3670 out_ip6_prohibit_entry:
3671         kfree(net->ipv6.ip6_prohibit_entry);
3672 out_ip6_null_entry:
3673         kfree(net->ipv6.ip6_null_entry);
3674 #endif
3675 out_ip6_dst_entries:
3676         dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3677 out_ip6_dst_ops:
3678         goto out;
3679 }
3680
3681 static void __net_exit ip6_route_net_exit(struct net *net)
3682 {
3683         kfree(net->ipv6.ip6_null_entry);
3684 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3685         kfree(net->ipv6.ip6_prohibit_entry);
3686         kfree(net->ipv6.ip6_blk_hole_entry);
3687 #endif
3688         dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3689 }
3690
3691 static int __net_init ip6_route_net_init_late(struct net *net)
3692 {
3693 #ifdef CONFIG_PROC_FS
3694         proc_create("ipv6_route", 0, net->proc_net, &ipv6_route_proc_fops);
3695         proc_create("rt6_stats", S_IRUGO, net->proc_net, &rt6_stats_seq_fops);
3696 #endif
3697         return 0;
3698 }
3699
3700 static void __net_exit ip6_route_net_exit_late(struct net *net)
3701 {
3702 #ifdef CONFIG_PROC_FS
3703         remove_proc_entry("ipv6_route", net->proc_net);
3704         remove_proc_entry("rt6_stats", net->proc_net);
3705 #endif
3706 }
3707
3708 static struct pernet_operations ip6_route_net_ops = {
3709         .init = ip6_route_net_init,
3710         .exit = ip6_route_net_exit,
3711 };
3712
3713 static int __net_init ipv6_inetpeer_init(struct net *net)
3714 {
3715         struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
3716
3717         if (!bp)
3718                 return -ENOMEM;
3719         inet_peer_base_init(bp);
3720         net->ipv6.peers = bp;
3721         return 0;
3722 }
3723
3724 static void __net_exit ipv6_inetpeer_exit(struct net *net)
3725 {
3726         struct inet_peer_base *bp = net->ipv6.peers;
3727
3728         net->ipv6.peers = NULL;
3729         inetpeer_invalidate_tree(bp);
3730         kfree(bp);
3731 }
3732
3733 static struct pernet_operations ipv6_inetpeer_ops = {
3734         .init   =       ipv6_inetpeer_init,
3735         .exit   =       ipv6_inetpeer_exit,
3736 };
3737
3738 static struct pernet_operations ip6_route_net_late_ops = {
3739         .init = ip6_route_net_init_late,
3740         .exit = ip6_route_net_exit_late,
3741 };
3742
3743 static struct notifier_block ip6_route_dev_notifier = {
3744         .notifier_call = ip6_route_dev_notify,
3745         .priority = 0,
3746 };
3747
3748 int __init ip6_route_init(void)
3749 {
3750         int ret;
3751         int cpu;
3752
3753         ret = -ENOMEM;
3754         ip6_dst_ops_template.kmem_cachep =
3755                 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
3756                                   SLAB_HWCACHE_ALIGN, NULL);
3757         if (!ip6_dst_ops_template.kmem_cachep)
3758                 goto out;
3759
3760         ret = dst_entries_init(&ip6_dst_blackhole_ops);
3761         if (ret)
3762                 goto out_kmem_cache;
3763
3764         ret = register_pernet_subsys(&ipv6_inetpeer_ops);
3765         if (ret)
3766                 goto out_dst_entries;
3767
3768         ret = register_pernet_subsys(&ip6_route_net_ops);
3769         if (ret)
3770                 goto out_register_inetpeer;
3771
3772         ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
3773
3774         /* Registering of the loopback is done before this portion of code,
3775          * the loopback reference in rt6_info will not be taken, do it
3776          * manually for init_net */
3777         init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
3778         init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3779   #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3780         init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
3781         init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3782         init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
3783         init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3784   #endif
3785         ret = fib6_init();
3786         if (ret)
3787                 goto out_register_subsys;
3788
3789         ret = xfrm6_init();
3790         if (ret)
3791                 goto out_fib6_init;
3792
3793         ret = fib6_rules_init();
3794         if (ret)
3795                 goto xfrm6_init;
3796
3797         ret = register_pernet_subsys(&ip6_route_net_late_ops);
3798         if (ret)
3799                 goto fib6_rules_init;
3800
3801         ret = -ENOBUFS;
3802         if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL, NULL) ||
3803             __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL, NULL) ||
3804             __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL, NULL))
3805                 goto out_register_late_subsys;
3806
3807         ret = register_netdevice_notifier(&ip6_route_dev_notifier);
3808         if (ret)
3809                 goto out_register_late_subsys;
3810
3811         for_each_possible_cpu(cpu) {
3812                 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
3813
3814                 INIT_LIST_HEAD(&ul->head);
3815                 spin_lock_init(&ul->lock);
3816         }
3817
3818 out:
3819         return ret;
3820
3821 out_register_late_subsys:
3822         unregister_pernet_subsys(&ip6_route_net_late_ops);
3823 fib6_rules_init:
3824         fib6_rules_cleanup();
3825 xfrm6_init:
3826         xfrm6_fini();
3827 out_fib6_init:
3828         fib6_gc_cleanup();
3829 out_register_subsys:
3830         unregister_pernet_subsys(&ip6_route_net_ops);
3831 out_register_inetpeer:
3832         unregister_pernet_subsys(&ipv6_inetpeer_ops);
3833 out_dst_entries:
3834         dst_entries_destroy(&ip6_dst_blackhole_ops);
3835 out_kmem_cache:
3836         kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3837         goto out;
3838 }
3839
3840 void ip6_route_cleanup(void)
3841 {
3842         unregister_netdevice_notifier(&ip6_route_dev_notifier);
3843         unregister_pernet_subsys(&ip6_route_net_late_ops);
3844         fib6_rules_cleanup();
3845         xfrm6_fini();
3846         fib6_gc_cleanup();
3847         unregister_pernet_subsys(&ipv6_inetpeer_ops);
3848         unregister_pernet_subsys(&ip6_route_net_ops);
3849         dst_entries_destroy(&ip6_dst_blackhole_ops);
3850         kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3851 }