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
  65 #include <asm/uaccess.h>
  66
  67 #ifdef CONFIG_SYSCTL
  68 #include <linux/sysctl.h>
  69 #endif
  70
  71 enum rt6_nud_state {
  72         RT6_NUD_FAIL_HARD = -3,
  73         RT6_NUD_FAIL_PROBE = -2,
  74         RT6_NUD_FAIL_DO_RR = -1,
  75         RT6_NUD_SUCCEED = 1
  76 };
  77
  78 static void ip6_rt_copy_init(struct rt6_info *rt, struct rt6_info *ort);
  79 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie);
  80 static unsigned int      ip6_default_advmss(const struct dst_entry *dst);
  81 static unsigned int      ip6_mtu(const struct dst_entry *dst);
  82 static struct dst_entry *ip6_negative_advice(struct dst_entry *);
  83 static void             ip6_dst_destroy(struct dst_entry *);
  84 static void             ip6_dst_ifdown(struct dst_entry *,
  85                                        struct net_device *dev, int how);
  86 static int               ip6_dst_gc(struct dst_ops *ops);
  87
  88 static int              ip6_pkt_discard(struct sk_buff *skb);
  89 static int              ip6_pkt_discard_out(struct sock *sk, struct sk_buff *skb);
  90 static int              ip6_pkt_prohibit(struct sk_buff *skb);
  91 static int              ip6_pkt_prohibit_out(struct sock *sk, struct sk_buff *skb);
  92 static void             ip6_link_failure(struct sk_buff *skb);
  93 static void             ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
  94                                            struct sk_buff *skb, u32 mtu);
  95 static void             rt6_do_redirect(struct dst_entry *dst, struct sock *sk,
  96                                         struct sk_buff *skb);
  97 static void             rt6_dst_from_metrics_check(struct rt6_info *rt);
  98 static int rt6_score_route(struct rt6_info *rt, int oif, int strict);
  99
 100 #ifdef CONFIG_IPV6_ROUTE_INFO
 101 static struct rt6_info *rt6_add_route_info(struct net *net,
 102                                            const struct in6_addr *prefix, int prefixlen,
 103                                            const struct in6_addr *gwaddr, int ifindex,
 104                                            unsigned int pref);
 105 static struct rt6_info *rt6_get_route_info(struct net *net,
 106                                            const struct in6_addr *prefix, int prefixlen,
 107                                            const struct in6_addr *gwaddr, int ifindex);
 108 #endif
 109
 110 struct uncached_list {
 111         spinlock_t              lock;
 112         struct list_head        head;
 113 };
 114
 115 static DEFINE_PER_CPU_ALIGNED(struct uncached_list, rt6_uncached_list);
 116
 117 static void rt6_uncached_list_add(struct rt6_info *rt)
 118 {
 119         struct uncached_list *ul = raw_cpu_ptr(&rt6_uncached_list);
 120
 121         rt->dst.flags |= DST_NOCACHE;
 122         rt->rt6i_uncached_list = ul;
 123
 124         spin_lock_bh(&ul->lock);
 125         list_add_tail(&rt->rt6i_uncached, &ul->head);
 126         spin_unlock_bh(&ul->lock);
 127 }
 128
 129 static void rt6_uncached_list_del(struct rt6_info *rt)
 130 {
 131         if (!list_empty(&rt->rt6i_uncached)) {
 132                 struct uncached_list *ul = rt->rt6i_uncached_list;
 133
 134                 spin_lock_bh(&ul->lock);
 135                 list_del(&rt->rt6i_uncached);
 136                 spin_unlock_bh(&ul->lock);
 137         }
 138 }
 139
 140 static void rt6_uncached_list_flush_dev(struct net *net, struct net_device *dev)
 141 {
 142         struct net_device *loopback_dev = net->loopback_dev;
 143         int cpu;
 144
 145         for_each_possible_cpu(cpu) {
 146                 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
 147                 struct rt6_info *rt;
 148
 149                 spin_lock_bh(&ul->lock);
 150                 list_for_each_entry(rt, &ul->head, rt6i_uncached) {
 151                         struct inet6_dev *rt_idev = rt->rt6i_idev;
 152                         struct net_device *rt_dev = rt->dst.dev;
 153
 154                         if (rt_idev && (rt_idev->dev == dev || !dev) &&
 155                             rt_idev->dev != loopback_dev) {
 156                                 rt->rt6i_idev = in6_dev_get(loopback_dev);
 157                                 in6_dev_put(rt_idev);
 158                         }
 159
 160                         if (rt_dev && (rt_dev == dev || !dev) &&
 161                             rt_dev != loopback_dev) {
 162                                 rt->dst.dev = loopback_dev;
 163                                 dev_hold(rt->dst.dev);
 164                                 dev_put(rt_dev);
 165                         }
 166                 }
 167                 spin_unlock_bh(&ul->lock);
 168         }
 169 }
 170
 171 static u32 *rt6_pcpu_cow_metrics(struct rt6_info *rt)
 172 {
 173         return dst_metrics_write_ptr(rt->dst.from);
 174 }
 175
 176 static u32 *ipv6_cow_metrics(struct dst_entry *dst, unsigned long old)
 177 {
 178         struct rt6_info *rt = (struct rt6_info *)dst;
 179
 180         if (rt->rt6i_flags & RTF_PCPU)
 181                 return rt6_pcpu_cow_metrics(rt);
 182         else if (rt->rt6i_flags & RTF_CACHE)
 183                 return NULL;
 184         else
 185                 return dst_cow_metrics_generic(dst, old);
 186 }
 187
 188 static inline const void *choose_neigh_daddr(struct rt6_info *rt,
 189                                              struct sk_buff *skb,
 190                                              const void *daddr)
 191 {
 192         struct in6_addr *p = &rt->rt6i_gateway;
 193
 194         if (!ipv6_addr_any(p))
 195                 return (const void *) p;
 196         else if (skb)
 197                 return &ipv6_hdr(skb)->daddr;
 198         return daddr;
 199 }
 200
 201 static struct neighbour *ip6_neigh_lookup(const struct dst_entry *dst,
 202                                           struct sk_buff *skb,
 203                                           const void *daddr)
 204 {
 205         struct rt6_info *rt = (struct rt6_info *) dst;
 206         struct neighbour *n;
 207
 208         daddr = choose_neigh_daddr(rt, skb, daddr);
 209         n = __ipv6_neigh_lookup(dst->dev, daddr);
 210         if (n)
 211                 return n;
 212         return neigh_create(&nd_tbl, daddr, dst->dev);
 213 }
 214
 215 static struct dst_ops ip6_dst_ops_template = {
 216         .family                 =       AF_INET6,
 217         .gc                     =       ip6_dst_gc,
 218         .gc_thresh              =       1024,
 219         .check                  =       ip6_dst_check,
 220         .default_advmss         =       ip6_default_advmss,
 221         .mtu                    =       ip6_mtu,
 222         .cow_metrics            =       ipv6_cow_metrics,
 223         .destroy                =       ip6_dst_destroy,
 224         .ifdown                 =       ip6_dst_ifdown,
 225         .negative_advice        =       ip6_negative_advice,
 226         .link_failure           =       ip6_link_failure,
 227         .update_pmtu            =       ip6_rt_update_pmtu,
 228         .redirect               =       rt6_do_redirect,
 229         .local_out              =       __ip6_local_out,
 230         .neigh_lookup           =       ip6_neigh_lookup,
 231 };
 232
 233 static unsigned int ip6_blackhole_mtu(const struct dst_entry *dst)
 234 {
 235         unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
 236
 237         return mtu ? : dst->dev->mtu;
 238 }
 239
 240 static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, struct sock *sk,
 241                                          struct sk_buff *skb, u32 mtu)
 242 {
 243 }
 244
 245 static void ip6_rt_blackhole_redirect(struct dst_entry *dst, struct sock *sk,
 246                                       struct sk_buff *skb)
 247 {
 248 }
 249
 250 static u32 *ip6_rt_blackhole_cow_metrics(struct dst_entry *dst,
 251                                          unsigned long old)
 252 {
 253         return NULL;
 254 }
 255
 256 static struct dst_ops ip6_dst_blackhole_ops = {
 257         .family                 =       AF_INET6,
 258         .destroy                =       ip6_dst_destroy,
 259         .check                  =       ip6_dst_check,
 260         .mtu                    =       ip6_blackhole_mtu,
 261         .default_advmss         =       ip6_default_advmss,
 262         .update_pmtu            =       ip6_rt_blackhole_update_pmtu,
 263         .redirect               =       ip6_rt_blackhole_redirect,
 264         .cow_metrics            =       ip6_rt_blackhole_cow_metrics,
 265         .neigh_lookup           =       ip6_neigh_lookup,
 266 };
 267
 268 static const u32 ip6_template_metrics[RTAX_MAX] = {
 269         [RTAX_HOPLIMIT - 1] = 0,
 270 };
 271
 272 static const struct rt6_info ip6_null_entry_template = {
 273         .dst = {
 274                 .__refcnt       = ATOMIC_INIT(1),
 275                 .__use          = 1,
 276                 .obsolete       = DST_OBSOLETE_FORCE_CHK,
 277                 .error          = -ENETUNREACH,
 278                 .input          = ip6_pkt_discard,
 279                 .output         = ip6_pkt_discard_out,
 280         },
 281         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
 282         .rt6i_protocol  = RTPROT_KERNEL,
 283         .rt6i_metric    = ~(u32) 0,
 284         .rt6i_ref       = ATOMIC_INIT(1),
 285 };
 286
 287 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
 288
 289 static const struct rt6_info ip6_prohibit_entry_template = {
 290         .dst = {
 291                 .__refcnt       = ATOMIC_INIT(1),
 292                 .__use          = 1,
 293                 .obsolete       = DST_OBSOLETE_FORCE_CHK,
 294                 .error          = -EACCES,
 295                 .input          = ip6_pkt_prohibit,
 296                 .output         = ip6_pkt_prohibit_out,
 297         },
 298         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
 299         .rt6i_protocol  = RTPROT_KERNEL,
 300         .rt6i_metric    = ~(u32) 0,
 301         .rt6i_ref       = ATOMIC_INIT(1),
 302 };
 303
 304 static const struct rt6_info ip6_blk_hole_entry_template = {
 305         .dst = {
 306                 .__refcnt       = ATOMIC_INIT(1),
 307                 .__use          = 1,
 308                 .obsolete       = DST_OBSOLETE_FORCE_CHK,
 309                 .error          = -EINVAL,
 310                 .input          = dst_discard,
 311                 .output         = dst_discard_sk,
 312         },
 313         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
 314         .rt6i_protocol  = RTPROT_KERNEL,
 315         .rt6i_metric    = ~(u32) 0,
 316         .rt6i_ref       = ATOMIC_INIT(1),
 317 };
 318
 319 #endif
 320
 321 /* allocate dst with ip6_dst_ops */
 322 static struct rt6_info *__ip6_dst_alloc(struct net *net,
 323                                         struct net_device *dev,
 324                                         int flags)
 325 {
 326         struct rt6_info *rt = dst_alloc(&net->ipv6.ip6_dst_ops, dev,
 327                                         0, DST_OBSOLETE_FORCE_CHK, flags);
 328
 329         if (rt) {
 330                 struct dst_entry *dst = &rt->dst;
 331
 332                 memset(dst + 1, 0, sizeof(*rt) - sizeof(*dst));
 333                 INIT_LIST_HEAD(&rt->rt6i_siblings);
 334                 INIT_LIST_HEAD(&rt->rt6i_uncached);
 335         }
 336         return rt;
 337 }
 338
 339 static struct rt6_info *ip6_dst_alloc(struct net *net,
 340                                       struct net_device *dev,
 341                                       int flags)
 342 {
 343         struct rt6_info *rt = __ip6_dst_alloc(net, dev, flags);
 344
 345         if (rt) {
 346                 rt->rt6i_pcpu = alloc_percpu_gfp(struct rt6_info *, GFP_ATOMIC);
 347                 if (rt->rt6i_pcpu) {
 348                         int cpu;
 349
 350                         for_each_possible_cpu(cpu) {
 351                                 struct rt6_info **p;
 352
 353                                 p = per_cpu_ptr(rt->rt6i_pcpu, cpu);
 354                                 /* no one shares rt */
 355                                 *p =  NULL;
 356                         }
 357                 } else {
 358                         dst_destroy((struct dst_entry *)rt);
 359                         return NULL;
 360                 }
 361         }
 362
 363         return rt;
 364 }
 365
 366 static void ip6_dst_destroy(struct dst_entry *dst)
 367 {
 368         struct rt6_info *rt = (struct rt6_info *)dst;
 369         struct dst_entry *from = dst->from;
 370         struct inet6_dev *idev;
 371
 372         dst_destroy_metrics_generic(dst);
 373         free_percpu(rt->rt6i_pcpu);
 374         rt6_uncached_list_del(rt);
 375
 376         idev = rt->rt6i_idev;
 377         if (idev) {
 378                 rt->rt6i_idev = NULL;
 379                 in6_dev_put(idev);
 380         }
 381
 382         dst->from = NULL;
 383         dst_release(from);
 384 }
 385
 386 static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
 387                            int how)
 388 {
 389         struct rt6_info *rt = (struct rt6_info *)dst;
 390         struct inet6_dev *idev = rt->rt6i_idev;
 391         struct net_device *loopback_dev =
 392                 dev_net(dev)->loopback_dev;
 393
 394         if (dev != loopback_dev) {
 395                 if (idev && idev->dev == dev) {
 396                         struct inet6_dev *loopback_idev =
 397                                 in6_dev_get(loopback_dev);
 398                         if (loopback_idev) {
 399                                 rt->rt6i_idev = loopback_idev;
 400                                 in6_dev_put(idev);
 401                         }
 402                 }
 403         }
 404 }
 405
 406 static bool rt6_check_expired(const struct rt6_info *rt)
 407 {
 408         if (rt->rt6i_flags & RTF_EXPIRES) {
 409                 if (time_after(jiffies, rt->dst.expires))
 410                         return true;
 411         } else if (rt->dst.from) {
 412                 return rt6_check_expired((struct rt6_info *) rt->dst.from);
 413         }
 414         return false;
 415 }
 416
 417 /* Multipath route selection:
 418  *   Hash based function using packet header and flowlabel.
 419  * Adapted from fib_info_hashfn()
 420  */
 421 static int rt6_info_hash_nhsfn(unsigned int candidate_count,
 422                                const struct flowi6 *fl6)
 423 {
 424         unsigned int val = fl6->flowi6_proto;
 425
 426         val ^= ipv6_addr_hash(&fl6->daddr);
 427         val ^= ipv6_addr_hash(&fl6->saddr);
 428
 429         /* Work only if this not encapsulated */
 430         switch (fl6->flowi6_proto) {
 431         case IPPROTO_UDP:
 432         case IPPROTO_TCP:
 433         case IPPROTO_SCTP:
 434                 val ^= (__force u16)fl6->fl6_sport;
 435                 val ^= (__force u16)fl6->fl6_dport;
 436                 break;
 437
 438         case IPPROTO_ICMPV6:
 439                 val ^= (__force u16)fl6->fl6_icmp_type;
 440                 val ^= (__force u16)fl6->fl6_icmp_code;
 441                 break;
 442         }
 443         /* RFC6438 recommands to use flowlabel */
 444         val ^= (__force u32)fl6->flowlabel;
 445
 446         /* Perhaps, we need to tune, this function? */
 447         val = val ^ (val >> 7) ^ (val >> 12);
 448         return val % candidate_count;
 449 }
 450
 451 static struct rt6_info *rt6_multipath_select(struct rt6_info *match,
 452                                              struct flowi6 *fl6, int oif,
 453                                              int strict)
 454 {
 455         struct rt6_info *sibling, *next_sibling;
 456         int route_choosen;
 457
 458         route_choosen = rt6_info_hash_nhsfn(match->rt6i_nsiblings + 1, fl6);
 459         /* Don't change the route, if route_choosen == 0
 460          * (siblings does not include ourself)
 461          */
 462         if (route_choosen)
 463                 list_for_each_entry_safe(sibling, next_sibling,
 464                                 &match->rt6i_siblings, rt6i_siblings) {
 465                         route_choosen--;
 466                         if (route_choosen == 0) {
 467                                 if (rt6_score_route(sibling, oif, strict) < 0)
 468                                         break;
 469                                 match = sibling;
 470                                 break;
 471                         }
 472                 }
 473         return match;
 474 }
 475
 476 /*
 477  *      Route lookup. Any table->tb6_lock is implied.
 478  */
 479
 480 static inline struct rt6_info *rt6_device_match(struct net *net,
 481                                                     struct rt6_info *rt,
 482                                                     const struct in6_addr *saddr,
 483                                                     int oif,
 484                                                     int flags)
 485 {
 486         struct rt6_info *local = NULL;
 487         struct rt6_info *sprt;
 488
 489         if (!oif && ipv6_addr_any(saddr))
 490                 goto out;
 491
 492         for (sprt = rt; sprt; sprt = sprt->dst.rt6_next) {
 493                 struct net_device *dev = sprt->dst.dev;
 494
 495                 if (oif) {
 496                         if (dev->ifindex == oif)
 497                                 return sprt;
 498                         if (dev->flags & IFF_LOOPBACK) {
 499                                 if (!sprt->rt6i_idev ||
 500                                     sprt->rt6i_idev->dev->ifindex != oif) {
 501                                         if (flags & RT6_LOOKUP_F_IFACE && oif)
 502                                                 continue;
 503                                         if (local && (!oif ||
 504                                                       local->rt6i_idev->dev->ifindex == oif))
 505                                                 continue;
 506                                 }
 507                                 local = sprt;
 508                         }
 509                 } else {
 510                         if (ipv6_chk_addr(net, saddr, dev,
 511                                           flags & RT6_LOOKUP_F_IFACE))
 512                                 return sprt;
 513                 }
 514         }
 515
 516         if (oif) {
 517                 if (local)
 518                         return local;
 519
 520                 if (flags & RT6_LOOKUP_F_IFACE)
 521                         return net->ipv6.ip6_null_entry;
 522         }
 523 out:
 524         return rt;
 525 }
 526
 527 #ifdef CONFIG_IPV6_ROUTER_PREF
 528 struct __rt6_probe_work {
 529         struct work_struct work;
 530         struct in6_addr target;
 531         struct net_device *dev;
 532 };
 533
 534 static void rt6_probe_deferred(struct work_struct *w)
 535 {
 536         struct in6_addr mcaddr;
 537         struct __rt6_probe_work *work =
 538                 container_of(w, struct __rt6_probe_work, work);
 539
 540         addrconf_addr_solict_mult(&work->target, &mcaddr);
 541         ndisc_send_ns(work->dev, NULL, &work->target, &mcaddr, NULL, NULL);
 542         dev_put(work->dev);
 543         kfree(work);
 544 }
 545
 546 static void rt6_probe(struct rt6_info *rt)
 547 {
 548         struct __rt6_probe_work *work;
 549         struct neighbour *neigh;
 550         /*
 551          * Okay, this does not seem to be appropriate
 552          * for now, however, we need to check if it
 553          * is really so; aka Router Reachability Probing.
 554          *
 555          * Router Reachability Probe MUST be rate-limited
 556          * to no more than one per minute.
 557          */
 558         if (!rt || !(rt->rt6i_flags & RTF_GATEWAY))
 559                 return;
 560         rcu_read_lock_bh();
 561         neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
 562         if (neigh) {
 563                 if (neigh->nud_state & NUD_VALID)
 564                         goto out;
 565
 566                 work = NULL;
 567                 write_lock(&neigh->lock);
 568                 if (!(neigh->nud_state & NUD_VALID) &&
 569                     time_after(jiffies,
 570                                neigh->updated +
 571                                rt->rt6i_idev->cnf.rtr_probe_interval)) {
 572                         work = kmalloc(sizeof(*work), GFP_ATOMIC);
 573                         if (work)
 574                                 __neigh_set_probe_once(neigh);
 575                 }
 576                 write_unlock(&neigh->lock);
 577         } else {
 578                 work = kmalloc(sizeof(*work), GFP_ATOMIC);
 579         }
 580
 581         if (work) {
 582                 INIT_WORK(&work->work, rt6_probe_deferred);
 583                 work->target = rt->rt6i_gateway;
 584                 dev_hold(rt->dst.dev);
 585                 work->dev = rt->dst.dev;
 586                 schedule_work(&work->work);
 587         }
 588
 589 out:
 590         rcu_read_unlock_bh();
 591 }
 592 #else
 593 static inline void rt6_probe(struct rt6_info *rt)
 594 {
 595 }
 596 #endif
 597
 598 /*
 599  * Default Router Selection (RFC 2461 6.3.6)
 600  */
 601 static inline int rt6_check_dev(struct rt6_info *rt, int oif)
 602 {
 603         struct net_device *dev = rt->dst.dev;
 604         if (!oif || dev->ifindex == oif)
 605                 return 2;
 606         if ((dev->flags & IFF_LOOPBACK) &&
 607             rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
 608                 return 1;
 609         return 0;
 610 }
 611
 612 static inline enum rt6_nud_state rt6_check_neigh(struct rt6_info *rt)
 613 {
 614         struct neighbour *neigh;
 615         enum rt6_nud_state ret = RT6_NUD_FAIL_HARD;
 616
 617         if (rt->rt6i_flags & RTF_NONEXTHOP ||
 618             !(rt->rt6i_flags & RTF_GATEWAY))
 619                 return RT6_NUD_SUCCEED;
 620
 621         rcu_read_lock_bh();
 622         neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
 623         if (neigh) {
 624                 read_lock(&neigh->lock);
 625                 if (neigh->nud_state & NUD_VALID)
 626                         ret = RT6_NUD_SUCCEED;
 627 #ifdef CONFIG_IPV6_ROUTER_PREF
 628                 else if (!(neigh->nud_state & NUD_FAILED))
 629                         ret = RT6_NUD_SUCCEED;
 630                 else
 631                         ret = RT6_NUD_FAIL_PROBE;
 632 #endif
 633                 read_unlock(&neigh->lock);
 634         } else {
 635                 ret = IS_ENABLED(CONFIG_IPV6_ROUTER_PREF) ?
 636                       RT6_NUD_SUCCEED : RT6_NUD_FAIL_DO_RR;
 637         }
 638         rcu_read_unlock_bh();
 639
 640         return ret;
 641 }
 642
 643 static int rt6_score_route(struct rt6_info *rt, int oif,
 644                            int strict)
 645 {
 646         int m;
 647
 648         m = rt6_check_dev(rt, oif);
 649         if (!m && (strict & RT6_LOOKUP_F_IFACE))
 650                 return RT6_NUD_FAIL_HARD;
 651 #ifdef CONFIG_IPV6_ROUTER_PREF
 652         m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
 653 #endif
 654         if (strict & RT6_LOOKUP_F_REACHABLE) {
 655                 int n = rt6_check_neigh(rt);
 656                 if (n < 0)
 657                         return n;
 658         }
 659         return m;
 660 }
 661
 662 static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
 663                                    int *mpri, struct rt6_info *match,
 664                                    bool *do_rr)
 665 {
 666         int m;
 667         bool match_do_rr = false;
 668         struct inet6_dev *idev = rt->rt6i_idev;
 669         struct net_device *dev = rt->dst.dev;
 670
 671         if (dev && !netif_carrier_ok(dev) &&
 672             idev->cnf.ignore_routes_with_linkdown)
 673                 goto out;
 674
 675         if (rt6_check_expired(rt))
 676                 goto out;
 677
 678         m = rt6_score_route(rt, oif, strict);
 679         if (m == RT6_NUD_FAIL_DO_RR) {
 680                 match_do_rr = true;
 681                 m = 0; /* lowest valid score */
 682         } else if (m == RT6_NUD_FAIL_HARD) {
 683                 goto out;
 684         }
 685
 686         if (strict & RT6_LOOKUP_F_REACHABLE)
 687                 rt6_probe(rt);
 688
 689         /* note that m can be RT6_NUD_FAIL_PROBE at this point */
 690         if (m > *mpri) {
 691                 *do_rr = match_do_rr;
 692                 *mpri = m;
 693                 match = rt;
 694         }
 695 out:
 696         return match;
 697 }
 698
 699 static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
 700                                      struct rt6_info *rr_head,
 701                                      u32 metric, int oif, int strict,
 702                                      bool *do_rr)
 703 {
 704         struct rt6_info *rt, *match, *cont;
 705         int mpri = -1;
 706
 707         match = NULL;
 708         cont = NULL;
 709         for (rt = rr_head; rt; rt = rt->dst.rt6_next) {
 710                 if (rt->rt6i_metric != metric) {
 711                         cont = rt;
 712                         break;
 713                 }
 714
 715                 match = find_match(rt, oif, strict, &mpri, match, do_rr);
 716         }
 717
 718         for (rt = fn->leaf; rt && rt != rr_head; rt = rt->dst.rt6_next) {
 719                 if (rt->rt6i_metric != metric) {
 720                         cont = rt;
 721                         break;
 722                 }
 723
 724                 match = find_match(rt, oif, strict, &mpri, match, do_rr);
 725         }
 726
 727         if (match || !cont)
 728                 return match;
 729
 730         for (rt = cont; rt; rt = rt->dst.rt6_next)
 731                 match = find_match(rt, oif, strict, &mpri, match, do_rr);
 732
 733         return match;
 734 }
 735
 736 static struct rt6_info *rt6_select(struct fib6_node *fn, int oif, int strict)
 737 {
 738         struct rt6_info *match, *rt0;
 739         struct net *net;
 740         bool do_rr = false;
 741
 742         rt0 = fn->rr_ptr;
 743         if (!rt0)
 744                 fn->rr_ptr = rt0 = fn->leaf;
 745
 746         match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict,
 747                              &do_rr);
 748
 749         if (do_rr) {
 750                 struct rt6_info *next = rt0->dst.rt6_next;
 751
 752                 /* no entries matched; do round-robin */
 753                 if (!next || next->rt6i_metric != rt0->rt6i_metric)
 754                         next = fn->leaf;
 755
 756                 if (next != rt0)
 757                         fn->rr_ptr = next;
 758         }
 759
 760         net = dev_net(rt0->dst.dev);
 761         return match ? match : net->ipv6.ip6_null_entry;
 762 }
 763
 764 static bool rt6_is_gw_or_nonexthop(const struct rt6_info *rt)
 765 {
 766         return (rt->rt6i_flags & (RTF_NONEXTHOP | RTF_GATEWAY));
 767 }
 768
 769 #ifdef CONFIG_IPV6_ROUTE_INFO
 770 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
 771                   const struct in6_addr *gwaddr)
 772 {
 773         struct net *net = dev_net(dev);
 774         struct route_info *rinfo = (struct route_info *) opt;
 775         struct in6_addr prefix_buf, *prefix;
 776         unsigned int pref;
 777         unsigned long lifetime;
 778         struct rt6_info *rt;
 779
 780         if (len < sizeof(struct route_info)) {
 781                 return -EINVAL;
 782         }
 783
 784         /* Sanity check for prefix_len and length */
 785         if (rinfo->length > 3) {
 786                 return -EINVAL;
 787         } else if (rinfo->prefix_len > 128) {
 788                 return -EINVAL;
 789         } else if (rinfo->prefix_len > 64) {
 790                 if (rinfo->length < 2) {
 791                         return -EINVAL;
 792                 }
 793         } else if (rinfo->prefix_len > 0) {
 794                 if (rinfo->length < 1) {
 795                         return -EINVAL;
 796                 }
 797         }
 798
 799         pref = rinfo->route_pref;
 800         if (pref == ICMPV6_ROUTER_PREF_INVALID)
 801                 return -EINVAL;
 802
 803         lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
 804
 805         if (rinfo->length == 3)
 806                 prefix = (struct in6_addr *)rinfo->prefix;
 807         else {
 808                 /* this function is safe */
 809                 ipv6_addr_prefix(&prefix_buf,
 810                                  (struct in6_addr *)rinfo->prefix,
 811                                  rinfo->prefix_len);
 812                 prefix = &prefix_buf;
 813         }
 814
 815         if (rinfo->prefix_len == 0)
 816                 rt = rt6_get_dflt_router(gwaddr, dev);
 817         else
 818                 rt = rt6_get_route_info(net, prefix, rinfo->prefix_len,
 819                                         gwaddr, dev->ifindex);
 820
 821         if (rt && !lifetime) {
 822                 ip6_del_rt(rt);
 823                 rt = NULL;
 824         }
 825
 826         if (!rt && lifetime)
 827                 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr, dev->ifindex,
 828                                         pref);
 829         else if (rt)
 830                 rt->rt6i_flags = RTF_ROUTEINFO |
 831                                  (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
 832
 833         if (rt) {
 834                 if (!addrconf_finite_timeout(lifetime))
 835                         rt6_clean_expires(rt);
 836                 else
 837                         rt6_set_expires(rt, jiffies + HZ * lifetime);
 838
 839                 ip6_rt_put(rt);
 840         }
 841         return 0;
 842 }
 843 #endif
 844
 845 static struct fib6_node* fib6_backtrack(struct fib6_node *fn,
 846                                         struct in6_addr *saddr)
 847 {
 848         struct fib6_node *pn;
 849         while (1) {
 850                 if (fn->fn_flags & RTN_TL_ROOT)
 851                         return NULL;
 852                 pn = fn->parent;
 853                 if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn)
 854                         fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr);
 855                 else
 856                         fn = pn;
 857                 if (fn->fn_flags & RTN_RTINFO)
 858                         return fn;
 859         }
 860 }
 861
 862 static struct rt6_info *ip6_pol_route_lookup(struct net *net,
 863                                              struct fib6_table *table,
 864                                              struct flowi6 *fl6, int flags)
 865 {
 866         struct fib6_node *fn;
 867         struct rt6_info *rt;
 868
 869         read_lock_bh(&table->tb6_lock);
 870         fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
 871 restart:
 872         rt = fn->leaf;
 873         rt = rt6_device_match(net, rt, &fl6->saddr, fl6->flowi6_oif, flags);
 874         if (rt->rt6i_nsiblings && fl6->flowi6_oif == 0)
 875                 rt = rt6_multipath_select(rt, fl6, fl6->flowi6_oif, flags);
 876         if (rt == net->ipv6.ip6_null_entry) {
 877                 fn = fib6_backtrack(fn, &fl6->saddr);
 878                 if (fn)
 879                         goto restart;
 880         }
 881         dst_use(&rt->dst, jiffies);
 882         read_unlock_bh(&table->tb6_lock);
 883         return rt;
 884
 885 }
 886
 887 struct dst_entry *ip6_route_lookup(struct net *net, struct flowi6 *fl6,
 888                                     int flags)
 889 {
 890         return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_lookup);
 891 }
 892 EXPORT_SYMBOL_GPL(ip6_route_lookup);
 893
 894 struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
 895                             const struct in6_addr *saddr, int oif, int strict)
 896 {
 897         struct flowi6 fl6 = {
 898                 .flowi6_oif = oif,
 899                 .daddr = *daddr,
 900         };
 901         struct dst_entry *dst;
 902         int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
 903
 904         if (saddr) {
 905                 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
 906                 flags |= RT6_LOOKUP_F_HAS_SADDR;
 907         }
 908
 909         dst = fib6_rule_lookup(net, &fl6, flags, ip6_pol_route_lookup);
 910         if (dst->error == 0)
 911                 return (struct rt6_info *) dst;
 912
 913         dst_release(dst);
 914
 915         return NULL;
 916 }
 917 EXPORT_SYMBOL(rt6_lookup);
 918
 919 /* ip6_ins_rt is called with FREE table->tb6_lock.
 920    It takes new route entry, the addition fails by any reason the
 921    route is freed. In any case, if caller does not hold it, it may
 922    be destroyed.
 923  */
 924
 925 static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info,
 926                         struct mx6_config *mxc)
 927 {
 928         int err;
 929         struct fib6_table *table;
 930
 931         table = rt->rt6i_table;
 932         write_lock_bh(&table->tb6_lock);
 933         err = fib6_add(&table->tb6_root, rt, info, mxc);
 934         write_unlock_bh(&table->tb6_lock);
 935
 936         return err;
 937 }
 938
 939 int ip6_ins_rt(struct rt6_info *rt)
 940 {
 941         struct nl_info info = { .nl_net = dev_net(rt->dst.dev), };
 942         struct mx6_config mxc = { .mx = NULL, };
 943
 944         return __ip6_ins_rt(rt, &info, &mxc);
 945 }
 946
 947 static struct rt6_info *ip6_rt_cache_alloc(struct rt6_info *ort,
 948                                            const struct in6_addr *daddr,
 949                                            const struct in6_addr *saddr)
 950 {
 951         struct rt6_info *rt;
 952
 953         /*
 954          *      Clone the route.
 955          */
 956
 957         if (ort->rt6i_flags & (RTF_CACHE | RTF_PCPU))
 958                 ort = (struct rt6_info *)ort->dst.from;
 959
 960         rt = __ip6_dst_alloc(dev_net(ort->dst.dev), ort->dst.dev, 0);
 961
 962         if (!rt)
 963                 return NULL;
 964
 965         ip6_rt_copy_init(rt, ort);
 966         rt->rt6i_flags |= RTF_CACHE;
 967         rt->rt6i_metric = 0;
 968         rt->dst.flags |= DST_HOST;
 969         rt->rt6i_dst.addr = *daddr;
 970         rt->rt6i_dst.plen = 128;
 971
 972         if (!rt6_is_gw_or_nonexthop(ort)) {
 973                 if (ort->rt6i_dst.plen != 128 &&
 974                     ipv6_addr_equal(&ort->rt6i_dst.addr, daddr))
 975                         rt->rt6i_flags |= RTF_ANYCAST;
 976 #ifdef CONFIG_IPV6_SUBTREES
 977                 if (rt->rt6i_src.plen && saddr) {
 978                         rt->rt6i_src.addr = *saddr;
 979                         rt->rt6i_src.plen = 128;
 980                 }
 981 #endif
 982         }
 983
 984         return rt;
 985 }
 986
 987 static struct rt6_info *ip6_rt_pcpu_alloc(struct rt6_info *rt)
 988 {
 989         struct rt6_info *pcpu_rt;
 990
 991         pcpu_rt = __ip6_dst_alloc(dev_net(rt->dst.dev),
 992                                   rt->dst.dev, rt->dst.flags);
 993
 994         if (!pcpu_rt)
 995                 return NULL;
 996         ip6_rt_copy_init(pcpu_rt, rt);
 997         pcpu_rt->rt6i_protocol = rt->rt6i_protocol;
 998         pcpu_rt->rt6i_flags |= RTF_PCPU;
 999         return pcpu_rt;
1000 }
1001
1002 /* It should be called with read_lock_bh(&tb6_lock) acquired */
1003 static struct rt6_info *rt6_get_pcpu_route(struct rt6_info *rt)
1004 {
1005         struct rt6_info *pcpu_rt, **p;
1006
1007         p = this_cpu_ptr(rt->rt6i_pcpu);
1008         pcpu_rt = *p;
1009
1010         if (pcpu_rt) {
1011                 dst_hold(&pcpu_rt->dst);
1012                 rt6_dst_from_metrics_check(pcpu_rt);
1013         }
1014         return pcpu_rt;
1015 }
1016
1017 static struct rt6_info *rt6_make_pcpu_route(struct rt6_info *rt)
1018 {
1019         struct fib6_table *table = rt->rt6i_table;
1020         struct rt6_info *pcpu_rt, *prev, **p;
1021
1022         pcpu_rt = ip6_rt_pcpu_alloc(rt);
1023         if (!pcpu_rt) {
1024                 struct net *net = dev_net(rt->dst.dev);
1025
1026                 dst_hold(&net->ipv6.ip6_null_entry->dst);
1027                 return net->ipv6.ip6_null_entry;
1028         }
1029
1030         read_lock_bh(&table->tb6_lock);
1031         if (rt->rt6i_pcpu) {
1032                 p = this_cpu_ptr(rt->rt6i_pcpu);
1033                 prev = cmpxchg(p, NULL, pcpu_rt);
1034                 if (prev) {
1035                         /* If someone did it before us, return prev instead */
1036                         dst_destroy(&pcpu_rt->dst);
1037                         pcpu_rt = prev;
1038                 }
1039         } else {
1040                 /* rt has been removed from the fib6 tree
1041                  * before we have a chance to acquire the read_lock.
1042                  * In this case, don't brother to create a pcpu rt
1043                  * since rt is going away anyway.  The next
1044                  * dst_check() will trigger a re-lookup.
1045                  */
1046                 dst_destroy(&pcpu_rt->dst);
1047                 pcpu_rt = rt;
1048         }
1049         dst_hold(&pcpu_rt->dst);
1050         rt6_dst_from_metrics_check(pcpu_rt);
1051         read_unlock_bh(&table->tb6_lock);
1052         return pcpu_rt;
1053 }
1054
1055 static struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table, int oif,
1056                                       struct flowi6 *fl6, int flags)
1057 {
1058         struct fib6_node *fn, *saved_fn;
1059         struct rt6_info *rt;
1060         int strict = 0;
1061
1062         strict |= flags & RT6_LOOKUP_F_IFACE;
1063         if (net->ipv6.devconf_all->forwarding == 0)
1064                 strict |= RT6_LOOKUP_F_REACHABLE;
1065
1066         read_lock_bh(&table->tb6_lock);
1067
1068         fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1069         saved_fn = fn;
1070
1071 redo_rt6_select:
1072         rt = rt6_select(fn, oif, strict);
1073         if (rt->rt6i_nsiblings)
1074                 rt = rt6_multipath_select(rt, fl6, oif, strict);
1075         if (rt == net->ipv6.ip6_null_entry) {
1076                 fn = fib6_backtrack(fn, &fl6->saddr);
1077                 if (fn)
1078                         goto redo_rt6_select;
1079                 else if (strict & RT6_LOOKUP_F_REACHABLE) {
1080                         /* also consider unreachable route */
1081                         strict &= ~RT6_LOOKUP_F_REACHABLE;
1082                         fn = saved_fn;
1083                         goto redo_rt6_select;
1084                 }
1085         }
1086
1087
1088         if (rt == net->ipv6.ip6_null_entry || (rt->rt6i_flags & RTF_CACHE)) {
1089                 dst_use(&rt->dst, jiffies);
1090                 read_unlock_bh(&table->tb6_lock);
1091
1092                 rt6_dst_from_metrics_check(rt);
1093                 return rt;
1094         } else if (unlikely((fl6->flowi6_flags & FLOWI_FLAG_KNOWN_NH) &&
1095                             !(rt->rt6i_flags & RTF_GATEWAY))) {
1096                 /* Create a RTF_CACHE clone which will not be
1097                  * owned by the fib6 tree.  It is for the special case where
1098                  * the daddr in the skb during the neighbor look-up is different
1099                  * from the fl6->daddr used to look-up route here.
1100                  */
1101
1102                 struct rt6_info *uncached_rt;
1103
1104                 dst_use(&rt->dst, jiffies);
1105                 read_unlock_bh(&table->tb6_lock);
1106
1107                 uncached_rt = ip6_rt_cache_alloc(rt, &fl6->daddr, NULL);
1108                 dst_release(&rt->dst);
1109
1110                 if (uncached_rt)
1111                         rt6_uncached_list_add(uncached_rt);
1112                 else
1113                         uncached_rt = net->ipv6.ip6_null_entry;
1114
1115                 dst_hold(&uncached_rt->dst);
1116                 return uncached_rt;
1117
1118         } else {
1119                 /* Get a percpu copy */
1120
1121                 struct rt6_info *pcpu_rt;
1122
1123                 rt->dst.lastuse = jiffies;
1124                 rt->dst.__use++;
1125                 pcpu_rt = rt6_get_pcpu_route(rt);
1126
1127                 if (pcpu_rt) {
1128                         read_unlock_bh(&table->tb6_lock);
1129                 } else {
1130                         /* We have to do the read_unlock first
1131                          * because rt6_make_pcpu_route() may trigger
1132                          * ip6_dst_gc() which will take the write_lock.
1133                          */
1134                         dst_hold(&rt->dst);
1135                         read_unlock_bh(&table->tb6_lock);
1136                         pcpu_rt = rt6_make_pcpu_route(rt);
1137                         dst_release(&rt->dst);
1138                 }
1139
1140                 return pcpu_rt;
1141
1142         }
1143 }
1144
1145 static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
1146                                             struct flowi6 *fl6, int flags)
1147 {
1148         return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, flags);
1149 }
1150
1151 static struct dst_entry *ip6_route_input_lookup(struct net *net,
1152                                                 struct net_device *dev,
1153                                                 struct flowi6 *fl6, int flags)
1154 {
1155         if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
1156                 flags |= RT6_LOOKUP_F_IFACE;
1157
1158         return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_input);
1159 }
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(struct net *net, const struct sock *sk,
1190                                     struct flowi6 *fl6)
1191 {
1192         int flags = 0;
1193
1194         fl6->flowi6_iif = LOOPBACK_IFINDEX;
1195
1196         if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr))
1197                 flags |= RT6_LOOKUP_F_IFACE;
1198
1199         if (!ipv6_addr_any(&fl6->saddr))
1200                 flags |= RT6_LOOKUP_F_HAS_SADDR;
1201         else if (sk)
1202                 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
1203
1204         return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_output);
1205 }
1206 EXPORT_SYMBOL(ip6_route_output);
1207
1208 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
1209 {
1210         struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
1211         struct dst_entry *new = NULL;
1212
1213         rt = dst_alloc(&ip6_dst_blackhole_ops, ort->dst.dev, 1, DST_OBSOLETE_NONE, 0);
1214         if (rt) {
1215                 new = &rt->dst;
1216
1217                 memset(new + 1, 0, sizeof(*rt) - sizeof(*new));
1218
1219                 new->__use = 1;
1220                 new->input = dst_discard;
1221                 new->output = dst_discard_sk;
1222
1223                 if (dst_metrics_read_only(&ort->dst))
1224                         new->_metrics = ort->dst._metrics;
1225                 else
1226                         dst_copy_metrics(new, &ort->dst);
1227                 rt->rt6i_idev = ort->rt6i_idev;
1228                 if (rt->rt6i_idev)
1229                         in6_dev_hold(rt->rt6i_idev);
1230
1231                 rt->rt6i_gateway = ort->rt6i_gateway;
1232                 rt->rt6i_flags = ort->rt6i_flags;
1233                 rt->rt6i_metric = 0;
1234
1235                 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1236 #ifdef CONFIG_IPV6_SUBTREES
1237                 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1238 #endif
1239
1240                 dst_free(new);
1241         }
1242
1243         dst_release(dst_orig);
1244         return new ? new : ERR_PTR(-ENOMEM);
1245 }
1246
1247 /*
1248  *      Destination cache support functions
1249  */
1250
1251 static void rt6_dst_from_metrics_check(struct rt6_info *rt)
1252 {
1253         if (rt->dst.from &&
1254             dst_metrics_ptr(&rt->dst) != dst_metrics_ptr(rt->dst.from))
1255                 dst_init_metrics(&rt->dst, dst_metrics_ptr(rt->dst.from), true);
1256 }
1257
1258 static struct dst_entry *rt6_check(struct rt6_info *rt, u32 cookie)
1259 {
1260         if (!rt->rt6i_node || (rt->rt6i_node->fn_sernum != cookie))
1261                 return NULL;
1262
1263         if (rt6_check_expired(rt))
1264                 return NULL;
1265
1266         return &rt->dst;
1267 }
1268
1269 static struct dst_entry *rt6_dst_from_check(struct rt6_info *rt, u32 cookie)
1270 {
1271         if (rt->dst.obsolete == DST_OBSOLETE_FORCE_CHK &&
1272             rt6_check((struct rt6_info *)(rt->dst.from), cookie))
1273                 return &rt->dst;
1274         else
1275                 return NULL;
1276 }
1277
1278 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
1279 {
1280         struct rt6_info *rt;
1281
1282         rt = (struct rt6_info *) dst;
1283
1284         /* All IPV6 dsts are created with ->obsolete set to the value
1285          * DST_OBSOLETE_FORCE_CHK which forces validation calls down
1286          * into this function always.
1287          */
1288
1289         rt6_dst_from_metrics_check(rt);
1290
1291         if ((rt->rt6i_flags & RTF_PCPU) || unlikely(dst->flags & DST_NOCACHE))
1292                 return rt6_dst_from_check(rt, cookie);
1293         else
1294                 return rt6_check(rt, cookie);
1295 }
1296
1297 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
1298 {
1299         struct rt6_info *rt = (struct rt6_info *) dst;
1300
1301         if (rt) {
1302                 if (rt->rt6i_flags & RTF_CACHE) {
1303                         if (rt6_check_expired(rt)) {
1304                                 ip6_del_rt(rt);
1305                                 dst = NULL;
1306                         }
1307                 } else {
1308                         dst_release(dst);
1309                         dst = NULL;
1310                 }
1311         }
1312         return dst;
1313 }
1314
1315 static void ip6_link_failure(struct sk_buff *skb)
1316 {
1317         struct rt6_info *rt;
1318
1319         icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
1320
1321         rt = (struct rt6_info *) skb_dst(skb);
1322         if (rt) {
1323                 if (rt->rt6i_flags & RTF_CACHE) {
1324                         dst_hold(&rt->dst);
1325                         ip6_del_rt(rt);
1326                 } else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT)) {
1327                         rt->rt6i_node->fn_sernum = -1;
1328                 }
1329         }
1330 }
1331
1332 static void rt6_do_update_pmtu(struct rt6_info *rt, u32 mtu)
1333 {
1334         struct net *net = dev_net(rt->dst.dev);
1335
1336         rt->rt6i_flags |= RTF_MODIFIED;
1337         rt->rt6i_pmtu = mtu;
1338         rt6_update_expires(rt, net->ipv6.sysctl.ip6_rt_mtu_expires);
1339 }
1340
1341 static void __ip6_rt_update_pmtu(struct dst_entry *dst, const struct sock *sk,
1342                                  const struct ipv6hdr *iph, u32 mtu)
1343 {
1344         struct rt6_info *rt6 = (struct rt6_info *)dst;
1345
1346         if (rt6->rt6i_flags & RTF_LOCAL)
1347                 return;
1348
1349         dst_confirm(dst);
1350         mtu = max_t(u32, mtu, IPV6_MIN_MTU);
1351         if (mtu >= dst_mtu(dst))
1352                 return;
1353
1354         if (rt6->rt6i_flags & RTF_CACHE) {
1355                 rt6_do_update_pmtu(rt6, mtu);
1356         } else {
1357                 const struct in6_addr *daddr, *saddr;
1358                 struct rt6_info *nrt6;
1359
1360                 if (iph) {
1361                         daddr = &iph->daddr;
1362                         saddr = &iph->saddr;
1363                 } else if (sk) {
1364                         daddr = &sk->sk_v6_daddr;
1365                         saddr = &inet6_sk(sk)->saddr;
1366                 } else {
1367                         return;
1368                 }
1369                 nrt6 = ip6_rt_cache_alloc(rt6, daddr, saddr);
1370                 if (nrt6) {
1371                         rt6_do_update_pmtu(nrt6, mtu);
1372
1373                         /* ip6_ins_rt(nrt6) will bump the
1374                          * rt6->rt6i_node->fn_sernum
1375                          * which will fail the next rt6_check() and
1376                          * invalidate the sk->sk_dst_cache.
1377                          */
1378                         ip6_ins_rt(nrt6);
1379                 }
1380         }
1381 }
1382
1383 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
1384                                struct sk_buff *skb, u32 mtu)
1385 {
1386         __ip6_rt_update_pmtu(dst, sk, skb ? ipv6_hdr(skb) : NULL, mtu);
1387 }
1388
1389 void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
1390                      int oif, u32 mark)
1391 {
1392         const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1393         struct dst_entry *dst;
1394         struct flowi6 fl6;
1395
1396         memset(&fl6, 0, sizeof(fl6));
1397         fl6.flowi6_oif = oif;
1398         fl6.flowi6_mark = mark ? mark : IP6_REPLY_MARK(net, skb->mark);
1399         fl6.daddr = iph->daddr;
1400         fl6.saddr = iph->saddr;
1401         fl6.flowlabel = ip6_flowinfo(iph);
1402
1403         dst = ip6_route_output(net, NULL, &fl6);
1404         if (!dst->error)
1405                 __ip6_rt_update_pmtu(dst, NULL, iph, ntohl(mtu));
1406         dst_release(dst);
1407 }
1408 EXPORT_SYMBOL_GPL(ip6_update_pmtu);
1409
1410 void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
1411 {
1412         ip6_update_pmtu(skb, sock_net(sk), mtu,
1413                         sk->sk_bound_dev_if, sk->sk_mark);
1414 }
1415 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
1416
1417 /* Handle redirects */
1418 struct ip6rd_flowi {
1419         struct flowi6 fl6;
1420         struct in6_addr gateway;
1421 };
1422
1423 static struct rt6_info *__ip6_route_redirect(struct net *net,
1424                                              struct fib6_table *table,
1425                                              struct flowi6 *fl6,
1426                                              int flags)
1427 {
1428         struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
1429         struct rt6_info *rt;
1430         struct fib6_node *fn;
1431
1432         /* Get the "current" route for this destination and
1433          * check if the redirect has come from approriate router.
1434          *
1435          * RFC 4861 specifies that redirects should only be
1436          * accepted if they come from the nexthop to the target.
1437          * Due to the way the routes are chosen, this notion
1438          * is a bit fuzzy and one might need to check all possible
1439          * routes.
1440          */
1441
1442         read_lock_bh(&table->tb6_lock);
1443         fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1444 restart:
1445         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1446                 if (rt6_check_expired(rt))
1447                         continue;
1448                 if (rt->dst.error)
1449                         break;
1450                 if (!(rt->rt6i_flags & RTF_GATEWAY))
1451                         continue;
1452                 if (fl6->flowi6_oif != rt->dst.dev->ifindex)
1453                         continue;
1454                 if (!ipv6_addr_equal(&rdfl->gateway, &rt->rt6i_gateway))
1455                         continue;
1456                 break;
1457         }
1458
1459         if (!rt)
1460                 rt = net->ipv6.ip6_null_entry;
1461         else if (rt->dst.error) {
1462                 rt = net->ipv6.ip6_null_entry;
1463                 goto out;
1464         }
1465
1466         if (rt == net->ipv6.ip6_null_entry) {
1467                 fn = fib6_backtrack(fn, &fl6->saddr);
1468                 if (fn)
1469                         goto restart;
1470         }
1471
1472 out:
1473         dst_hold(&rt->dst);
1474
1475         read_unlock_bh(&table->tb6_lock);
1476
1477         return rt;
1478 };
1479
1480 static struct dst_entry *ip6_route_redirect(struct net *net,
1481                                         const struct flowi6 *fl6,
1482                                         const struct in6_addr *gateway)
1483 {
1484         int flags = RT6_LOOKUP_F_HAS_SADDR;
1485         struct ip6rd_flowi rdfl;
1486
1487         rdfl.fl6 = *fl6;
1488         rdfl.gateway = *gateway;
1489
1490         return fib6_rule_lookup(net, &rdfl.fl6,
1491                                 flags, __ip6_route_redirect);
1492 }
1493
1494 void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark)
1495 {
1496         const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1497         struct dst_entry *dst;
1498         struct flowi6 fl6;
1499
1500         memset(&fl6, 0, sizeof(fl6));
1501         fl6.flowi6_iif = LOOPBACK_IFINDEX;
1502         fl6.flowi6_oif = oif;
1503         fl6.flowi6_mark = mark;
1504         fl6.daddr = iph->daddr;
1505         fl6.saddr = iph->saddr;
1506         fl6.flowlabel = ip6_flowinfo(iph);
1507
1508         dst = ip6_route_redirect(net, &fl6, &ipv6_hdr(skb)->saddr);
1509         rt6_do_redirect(dst, NULL, skb);
1510         dst_release(dst);
1511 }
1512 EXPORT_SYMBOL_GPL(ip6_redirect);
1513
1514 void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif,
1515                             u32 mark)
1516 {
1517         const struct ipv6hdr *iph = ipv6_hdr(skb);
1518         const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
1519         struct dst_entry *dst;
1520         struct flowi6 fl6;
1521
1522         memset(&fl6, 0, sizeof(fl6));
1523         fl6.flowi6_iif = LOOPBACK_IFINDEX;
1524         fl6.flowi6_oif = oif;
1525         fl6.flowi6_mark = mark;
1526         fl6.daddr = msg->dest;
1527         fl6.saddr = iph->daddr;
1528
1529         dst = ip6_route_redirect(net, &fl6, &iph->saddr);
1530         rt6_do_redirect(dst, NULL, skb);
1531         dst_release(dst);
1532 }
1533
1534 void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
1535 {
1536         ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark);
1537 }
1538 EXPORT_SYMBOL_GPL(ip6_sk_redirect);
1539
1540 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1541 {
1542         struct net_device *dev = dst->dev;
1543         unsigned int mtu = dst_mtu(dst);
1544         struct net *net = dev_net(dev);
1545
1546         mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
1547
1548         if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
1549                 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1550
1551         /*
1552          * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1553          * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1554          * IPV6_MAXPLEN is also valid and means: "any MSS,
1555          * rely only on pmtu discovery"
1556          */
1557         if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
1558                 mtu = IPV6_MAXPLEN;
1559         return mtu;
1560 }
1561
1562 static unsigned int ip6_mtu(const struct dst_entry *dst)
1563 {
1564         const struct rt6_info *rt = (const struct rt6_info *)dst;
1565         unsigned int mtu = rt->rt6i_pmtu;
1566         struct inet6_dev *idev;
1567
1568         if (mtu)
1569                 goto out;
1570
1571         mtu = dst_metric_raw(dst, RTAX_MTU);
1572         if (mtu)
1573                 goto out;
1574
1575         mtu = IPV6_MIN_MTU;
1576
1577         rcu_read_lock();
1578         idev = __in6_dev_get(dst->dev);
1579         if (idev)
1580                 mtu = idev->cnf.mtu6;
1581         rcu_read_unlock();
1582
1583 out:
1584         return min_t(unsigned int, mtu, IP6_MAX_MTU);
1585 }
1586
1587 static struct dst_entry *icmp6_dst_gc_list;
1588 static DEFINE_SPINLOCK(icmp6_dst_lock);
1589
1590 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
1591                                   struct flowi6 *fl6)
1592 {
1593         struct dst_entry *dst;
1594         struct rt6_info *rt;
1595         struct inet6_dev *idev = in6_dev_get(dev);
1596         struct net *net = dev_net(dev);
1597
1598         if (unlikely(!idev))
1599                 return ERR_PTR(-ENODEV);
1600
1601         rt = ip6_dst_alloc(net, dev, 0);
1602         if (unlikely(!rt)) {
1603                 in6_dev_put(idev);
1604                 dst = ERR_PTR(-ENOMEM);
1605                 goto out;
1606         }
1607
1608         rt->dst.flags |= DST_HOST;
1609         rt->dst.output  = ip6_output;
1610         atomic_set(&rt->dst.__refcnt, 1);
1611         rt->rt6i_gateway  = fl6->daddr;
1612         rt->rt6i_dst.addr = fl6->daddr;
1613         rt->rt6i_dst.plen = 128;
1614         rt->rt6i_idev     = idev;
1615         dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
1616
1617         spin_lock_bh(&icmp6_dst_lock);
1618         rt->dst.next = icmp6_dst_gc_list;
1619         icmp6_dst_gc_list = &rt->dst;
1620         spin_unlock_bh(&icmp6_dst_lock);
1621
1622         fib6_force_start_gc(net);
1623
1624         dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
1625
1626 out:
1627         return dst;
1628 }
1629
1630 int icmp6_dst_gc(void)
1631 {
1632         struct dst_entry *dst, **pprev;
1633         int more = 0;
1634
1635         spin_lock_bh(&icmp6_dst_lock);
1636         pprev = &icmp6_dst_gc_list;
1637
1638         while ((dst = *pprev) != NULL) {
1639                 if (!atomic_read(&dst->__refcnt)) {
1640                         *pprev = dst->next;
1641                         dst_free(dst);
1642                 } else {
1643                         pprev = &dst->next;
1644                         ++more;
1645                 }
1646         }
1647
1648         spin_unlock_bh(&icmp6_dst_lock);
1649
1650         return more;
1651 }
1652
1653 static void icmp6_clean_all(int (*func)(struct rt6_info *rt, void *arg),
1654                             void *arg)
1655 {
1656         struct dst_entry *dst, **pprev;
1657
1658         spin_lock_bh(&icmp6_dst_lock);
1659         pprev = &icmp6_dst_gc_list;
1660         while ((dst = *pprev) != NULL) {
1661                 struct rt6_info *rt = (struct rt6_info *) dst;
1662                 if (func(rt, arg)) {
1663                         *pprev = dst->next;
1664                         dst_free(dst);
1665                 } else {
1666                         pprev = &dst->next;
1667                 }
1668         }
1669         spin_unlock_bh(&icmp6_dst_lock);
1670 }
1671
1672 static int ip6_dst_gc(struct dst_ops *ops)
1673 {
1674         struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
1675         int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1676         int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1677         int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1678         int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1679         unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1680         int entries;
1681
1682         entries = dst_entries_get_fast(ops);
1683         if (time_after(rt_last_gc + rt_min_interval, jiffies) &&
1684             entries <= rt_max_size)
1685                 goto out;
1686
1687         net->ipv6.ip6_rt_gc_expire++;
1688         fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net, true);
1689         entries = dst_entries_get_slow(ops);
1690         if (entries < ops->gc_thresh)
1691                 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1692 out:
1693         net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1694         return entries > rt_max_size;
1695 }
1696
1697 static int ip6_convert_metrics(struct mx6_config *mxc,
1698                                const struct fib6_config *cfg)
1699 {
1700         bool ecn_ca = false;
1701         struct nlattr *nla;
1702         int remaining;
1703         u32 *mp;
1704
1705         if (!cfg->fc_mx)
1706                 return 0;
1707
1708         mp = kzalloc(sizeof(u32) * RTAX_MAX, GFP_KERNEL);
1709         if (unlikely(!mp))
1710                 return -ENOMEM;
1711
1712         nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1713                 int type = nla_type(nla);
1714                 u32 val;
1715
1716                 if (!type)
1717                         continue;
1718                 if (unlikely(type > RTAX_MAX))
1719                         goto err;
1720
1721                 if (type == RTAX_CC_ALGO) {
1722                         char tmp[TCP_CA_NAME_MAX];
1723
1724                         nla_strlcpy(tmp, nla, sizeof(tmp));
1725                         val = tcp_ca_get_key_by_name(tmp, &ecn_ca);
1726                         if (val == TCP_CA_UNSPEC)
1727                                 goto err;
1728                 } else {
1729                         val = nla_get_u32(nla);
1730                 }
1731                 if (type == RTAX_FEATURES && (val & ~RTAX_FEATURE_MASK))
1732                         goto err;
1733
1734                 mp[type - 1] = val;
1735                 __set_bit(type - 1, mxc->mx_valid);
1736         }
1737
1738         if (ecn_ca) {
1739                 __set_bit(RTAX_FEATURES - 1, mxc->mx_valid);
1740                 mp[RTAX_FEATURES - 1] |= DST_FEATURE_ECN_CA;
1741         }
1742
1743         mxc->mx = mp;
1744         return 0;
1745  err:
1746         kfree(mp);
1747         return -EINVAL;
1748 }
1749
1750 int ip6_route_info_create(struct fib6_config *cfg, struct rt6_info **rt_ret)
1751 {
1752         int err;
1753         struct net *net = cfg->fc_nlinfo.nl_net;
1754         struct rt6_info *rt = NULL;
1755         struct net_device *dev = NULL;
1756         struct inet6_dev *idev = NULL;
1757         struct fib6_table *table;
1758         int addr_type;
1759
1760         if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1761                 return -EINVAL;
1762 #ifndef CONFIG_IPV6_SUBTREES
1763         if (cfg->fc_src_len)
1764                 return -EINVAL;
1765 #endif
1766         if (cfg->fc_ifindex) {
1767                 err = -ENODEV;
1768                 dev = dev_get_by_index(net, cfg->fc_ifindex);
1769                 if (!dev)
1770                         goto out;
1771                 idev = in6_dev_get(dev);
1772                 if (!idev)
1773                         goto out;
1774         }
1775
1776         if (cfg->fc_metric == 0)
1777                 cfg->fc_metric = IP6_RT_PRIO_USER;
1778
1779         err = -ENOBUFS;
1780         if (cfg->fc_nlinfo.nlh &&
1781             !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
1782                 table = fib6_get_table(net, cfg->fc_table);
1783                 if (!table) {
1784                         pr_warn("NLM_F_CREATE should be specified when creating new route\n");
1785                         table = fib6_new_table(net, cfg->fc_table);
1786                 }
1787         } else {
1788                 table = fib6_new_table(net, cfg->fc_table);
1789         }
1790
1791         if (!table)
1792                 goto out;
1793
1794         rt = ip6_dst_alloc(net, NULL,
1795                            (cfg->fc_flags & RTF_ADDRCONF) ? 0 : DST_NOCOUNT);
1796
1797         if (!rt) {
1798                 err = -ENOMEM;
1799                 goto out;
1800         }
1801
1802         if (cfg->fc_flags & RTF_EXPIRES)
1803                 rt6_set_expires(rt, jiffies +
1804                                 clock_t_to_jiffies(cfg->fc_expires));
1805         else
1806                 rt6_clean_expires(rt);
1807
1808         if (cfg->fc_protocol == RTPROT_UNSPEC)
1809                 cfg->fc_protocol = RTPROT_BOOT;
1810         rt->rt6i_protocol = cfg->fc_protocol;
1811
1812         addr_type = ipv6_addr_type(&cfg->fc_dst);
1813
1814         if (addr_type & IPV6_ADDR_MULTICAST)
1815                 rt->dst.input = ip6_mc_input;
1816         else if (cfg->fc_flags & RTF_LOCAL)
1817                 rt->dst.input = ip6_input;
1818         else
1819                 rt->dst.input = ip6_forward;
1820
1821         rt->dst.output = ip6_output;
1822
1823         if (cfg->fc_encap) {
1824                 struct lwtunnel_state *lwtstate;
1825
1826                 err = lwtunnel_build_state(dev, cfg->fc_encap_type,
1827                                            cfg->fc_encap, AF_INET6, cfg,
1828                                            &lwtstate);
1829                 if (err)
1830                         goto out;
1831                 rt->dst.lwtstate = lwtstate_get(lwtstate);
1832                 if (lwtunnel_output_redirect(rt->dst.lwtstate)) {
1833                         rt->dst.lwtstate->orig_output = rt->dst.output;
1834                         rt->dst.output = lwtunnel_output;
1835                 }
1836                 if (lwtunnel_input_redirect(rt->dst.lwtstate)) {
1837                         rt->dst.lwtstate->orig_input = rt->dst.input;
1838                         rt->dst.input = lwtunnel_input;
1839                 }
1840         }
1841
1842         ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1843         rt->rt6i_dst.plen = cfg->fc_dst_len;
1844         if (rt->rt6i_dst.plen == 128)
1845                 rt->dst.flags |= DST_HOST;
1846
1847 #ifdef CONFIG_IPV6_SUBTREES
1848         ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1849         rt->rt6i_src.plen = cfg->fc_src_len;
1850 #endif
1851
1852         rt->rt6i_metric = cfg->fc_metric;
1853
1854         /* We cannot add true routes via loopback here,
1855            they would result in kernel looping; promote them to reject routes
1856          */
1857         if ((cfg->fc_flags & RTF_REJECT) ||
1858             (dev && (dev->flags & IFF_LOOPBACK) &&
1859              !(addr_type & IPV6_ADDR_LOOPBACK) &&
1860              !(cfg->fc_flags & RTF_LOCAL))) {
1861                 /* hold loopback dev/idev if we haven't done so. */
1862                 if (dev != net->loopback_dev) {
1863                         if (dev) {
1864                                 dev_put(dev);
1865                                 in6_dev_put(idev);
1866                         }
1867                         dev = net->loopback_dev;
1868                         dev_hold(dev);
1869                         idev = in6_dev_get(dev);
1870                         if (!idev) {
1871                                 err = -ENODEV;
1872                                 goto out;
1873                         }
1874                 }
1875                 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1876                 switch (cfg->fc_type) {
1877                 case RTN_BLACKHOLE:
1878                         rt->dst.error = -EINVAL;
1879                         rt->dst.output = dst_discard_sk;
1880                         rt->dst.input = dst_discard;
1881                         break;
1882                 case RTN_PROHIBIT:
1883                         rt->dst.error = -EACCES;
1884                         rt->dst.output = ip6_pkt_prohibit_out;
1885                         rt->dst.input = ip6_pkt_prohibit;
1886                         break;
1887                 case RTN_THROW:
1888                 case RTN_UNREACHABLE:
1889                 default:
1890                         rt->dst.error = (cfg->fc_type == RTN_THROW) ? -EAGAIN
1891                                         : (cfg->fc_type == RTN_UNREACHABLE)
1892                                         ? -EHOSTUNREACH : -ENETUNREACH;
1893                         rt->dst.output = ip6_pkt_discard_out;
1894                         rt->dst.input = ip6_pkt_discard;
1895                         break;
1896                 }
1897                 goto install_route;
1898         }
1899
1900         if (cfg->fc_flags & RTF_GATEWAY) {
1901                 const struct in6_addr *gw_addr;
1902                 int gwa_type;
1903
1904                 gw_addr = &cfg->fc_gateway;
1905                 gwa_type = ipv6_addr_type(gw_addr);
1906
1907                 /* if gw_addr is local we will fail to detect this in case
1908                  * address is still TENTATIVE (DAD in progress). rt6_lookup()
1909                  * will return already-added prefix route via interface that
1910                  * prefix route was assigned to, which might be non-loopback.
1911                  */
1912                 err = -EINVAL;
1913                 if (ipv6_chk_addr_and_flags(net, gw_addr,
1914                                             gwa_type & IPV6_ADDR_LINKLOCAL ?
1915                                             dev : NULL, 0, 0))
1916                         goto out;
1917
1918                 rt->rt6i_gateway = *gw_addr;
1919
1920                 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1921                         struct rt6_info *grt;
1922
1923                         /* IPv6 strictly inhibits using not link-local
1924                            addresses as nexthop address.
1925                            Otherwise, router will not able to send redirects.
1926                            It is very good, but in some (rare!) circumstances
1927                            (SIT, PtP, NBMA NOARP links) it is handy to allow
1928                            some exceptions. --ANK
1929                          */
1930                         if (!(gwa_type & IPV6_ADDR_UNICAST))
1931                                 goto out;
1932
1933                         grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, 1);
1934
1935                         err = -EHOSTUNREACH;
1936                         if (!grt)
1937                                 goto out;
1938                         if (dev) {
1939                                 if (dev != grt->dst.dev) {
1940                                         ip6_rt_put(grt);
1941                                         goto out;
1942                                 }
1943                         } else {
1944                                 dev = grt->dst.dev;
1945                                 idev = grt->rt6i_idev;
1946                                 dev_hold(dev);
1947                                 in6_dev_hold(grt->rt6i_idev);
1948                         }
1949                         if (!(grt->rt6i_flags & RTF_GATEWAY))
1950                                 err = 0;
1951                         ip6_rt_put(grt);
1952
1953                         if (err)
1954                                 goto out;
1955                 }
1956                 err = -EINVAL;
1957                 if (!dev || (dev->flags & IFF_LOOPBACK))
1958                         goto out;
1959         }
1960
1961         err = -ENODEV;
1962         if (!dev)
1963                 goto out;
1964
1965         if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
1966                 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
1967                         err = -EINVAL;
1968                         goto out;
1969                 }
1970                 rt->rt6i_prefsrc.addr = cfg->fc_prefsrc;
1971                 rt->rt6i_prefsrc.plen = 128;
1972         } else
1973                 rt->rt6i_prefsrc.plen = 0;
1974
1975         rt->rt6i_flags = cfg->fc_flags;
1976
1977 install_route:
1978         rt->dst.dev = dev;
1979         rt->rt6i_idev = idev;
1980         rt->rt6i_table = table;
1981
1982         cfg->fc_nlinfo.nl_net = dev_net(dev);
1983
1984         *rt_ret = rt;
1985
1986         return 0;
1987 out:
1988         if (dev)
1989                 dev_put(dev);
1990         if (idev)
1991                 in6_dev_put(idev);
1992         if (rt)
1993                 dst_free(&rt->dst);
1994
1995         *rt_ret = NULL;
1996
1997         return err;
1998 }
1999
2000 int ip6_route_add(struct fib6_config *cfg)
2001 {
2002         struct mx6_config mxc = { .mx = NULL, };
2003         struct rt6_info *rt = NULL;
2004         int err;
2005
2006         err = ip6_route_info_create(cfg, &rt);
2007         if (err)
2008                 goto out;
2009
2010         err = ip6_convert_metrics(&mxc, cfg);
2011         if (err)
2012                 goto out;
2013
2014         err = __ip6_ins_rt(rt, &cfg->fc_nlinfo, &mxc);
2015
2016         kfree(mxc.mx);
2017
2018         return err;
2019 out:
2020         if (rt)
2021                 dst_free(&rt->dst);
2022
2023         return err;
2024 }
2025
2026 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
2027 {
2028         int err;
2029         struct fib6_table *table;
2030         struct net *net = dev_net(rt->dst.dev);
2031
2032         if (rt == net->ipv6.ip6_null_entry ||
2033             rt->dst.flags & DST_NOCACHE) {
2034                 err = -ENOENT;
2035                 goto out;
2036         }
2037
2038         table = rt->rt6i_table;
2039         write_lock_bh(&table->tb6_lock);
2040         err = fib6_del(rt, info);
2041         write_unlock_bh(&table->tb6_lock);
2042
2043 out:
2044         ip6_rt_put(rt);
2045         return err;
2046 }
2047
2048 int ip6_del_rt(struct rt6_info *rt)
2049 {
2050         struct nl_info info = {
2051                 .nl_net = dev_net(rt->dst.dev),
2052         };
2053         return __ip6_del_rt(rt, &info);
2054 }
2055
2056 static int ip6_route_del(struct fib6_config *cfg)
2057 {
2058         struct fib6_table *table;
2059         struct fib6_node *fn;
2060         struct rt6_info *rt;
2061         int err = -ESRCH;
2062
2063         table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
2064         if (!table)
2065                 return err;
2066
2067         read_lock_bh(&table->tb6_lock);
2068
2069         fn = fib6_locate(&table->tb6_root,
2070                          &cfg->fc_dst, cfg->fc_dst_len,
2071                          &cfg->fc_src, cfg->fc_src_len);
2072
2073         if (fn) {
2074                 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2075                         if ((rt->rt6i_flags & RTF_CACHE) &&
2076                             !(cfg->fc_flags & RTF_CACHE))
2077                                 continue;
2078                         if (cfg->fc_ifindex &&
2079                             (!rt->dst.dev ||
2080                              rt->dst.dev->ifindex != cfg->fc_ifindex))
2081                                 continue;
2082                         if (cfg->fc_flags & RTF_GATEWAY &&
2083                             !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
2084                                 continue;
2085                         if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
2086                                 continue;
2087                         dst_hold(&rt->dst);
2088                         read_unlock_bh(&table->tb6_lock);
2089
2090                         return __ip6_del_rt(rt, &cfg->fc_nlinfo);
2091                 }
2092         }
2093         read_unlock_bh(&table->tb6_lock);
2094
2095         return err;
2096 }
2097
2098 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
2099 {
2100         struct net *net = dev_net(skb->dev);
2101         struct netevent_redirect netevent;
2102         struct rt6_info *rt, *nrt = NULL;
2103         struct ndisc_options ndopts;
2104         struct inet6_dev *in6_dev;
2105         struct neighbour *neigh;
2106         struct rd_msg *msg;
2107         int optlen, on_link;
2108         u8 *lladdr;
2109
2110         optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
2111         optlen -= sizeof(*msg);
2112
2113         if (optlen < 0) {
2114                 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
2115                 return;
2116         }
2117
2118         msg = (struct rd_msg *)icmp6_hdr(skb);
2119
2120         if (ipv6_addr_is_multicast(&msg->dest)) {
2121                 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
2122                 return;
2123         }
2124
2125         on_link = 0;
2126         if (ipv6_addr_equal(&msg->dest, &msg->target)) {
2127                 on_link = 1;
2128         } else if (ipv6_addr_type(&msg->target) !=
2129                    (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
2130                 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
2131                 return;
2132         }
2133
2134         in6_dev = __in6_dev_get(skb->dev);
2135         if (!in6_dev)
2136                 return;
2137         if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
2138                 return;
2139
2140         /* RFC2461 8.1:
2141          *      The IP source address of the Redirect MUST be the same as the current
2142          *      first-hop router for the specified ICMP Destination Address.
2143          */
2144
2145         if (!ndisc_parse_options(msg->opt, optlen, &ndopts)) {
2146                 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
2147                 return;
2148         }
2149
2150         lladdr = NULL;
2151         if (ndopts.nd_opts_tgt_lladdr) {
2152                 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
2153                                              skb->dev);
2154                 if (!lladdr) {
2155                         net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
2156                         return;
2157                 }
2158         }
2159
2160         rt = (struct rt6_info *) dst;
2161         if (rt == net->ipv6.ip6_null_entry) {
2162                 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
2163                 return;
2164         }
2165
2166         /* Redirect received -> path was valid.
2167          * Look, redirects are sent only in response to data packets,
2168          * so that this nexthop apparently is reachable. --ANK
2169          */
2170         dst_confirm(&rt->dst);
2171
2172         neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
2173         if (!neigh)
2174                 return;
2175
2176         /*
2177          *      We have finally decided to accept it.
2178          */
2179
2180         neigh_update(neigh, lladdr, NUD_STALE,
2181                      NEIGH_UPDATE_F_WEAK_OVERRIDE|
2182                      NEIGH_UPDATE_F_OVERRIDE|
2183                      (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
2184                                      NEIGH_UPDATE_F_ISROUTER))
2185                      );
2186
2187         nrt = ip6_rt_cache_alloc(rt, &msg->dest, NULL);
2188         if (!nrt)
2189                 goto out;
2190
2191         nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
2192         if (on_link)
2193                 nrt->rt6i_flags &= ~RTF_GATEWAY;
2194
2195         nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
2196
2197         if (ip6_ins_rt(nrt))
2198                 goto out;
2199
2200         netevent.old = &rt->dst;
2201         netevent.new = &nrt->dst;
2202         netevent.daddr = &msg->dest;
2203         netevent.neigh = neigh;
2204         call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
2205
2206         if (rt->rt6i_flags & RTF_CACHE) {
2207                 rt = (struct rt6_info *) dst_clone(&rt->dst);
2208                 ip6_del_rt(rt);
2209         }
2210
2211 out:
2212         neigh_release(neigh);
2213 }
2214
2215 /*
2216  *      Misc support functions
2217  */
2218
2219 static void rt6_set_from(struct rt6_info *rt, struct rt6_info *from)
2220 {
2221         BUG_ON(from->dst.from);
2222
2223         rt->rt6i_flags &= ~RTF_EXPIRES;
2224         dst_hold(&from->dst);
2225         rt->dst.from = &from->dst;
2226         dst_init_metrics(&rt->dst, dst_metrics_ptr(&from->dst), true);
2227 }
2228
2229 static void ip6_rt_copy_init(struct rt6_info *rt, struct rt6_info *ort)
2230 {
2231         rt->dst.input = ort->dst.input;
2232         rt->dst.output = ort->dst.output;
2233         rt->rt6i_dst = ort->rt6i_dst;
2234         rt->dst.error = ort->dst.error;
2235         rt->rt6i_idev = ort->rt6i_idev;
2236         if (rt->rt6i_idev)
2237                 in6_dev_hold(rt->rt6i_idev);
2238         rt->dst.lastuse = jiffies;
2239         rt->rt6i_gateway = ort->rt6i_gateway;
2240         rt->rt6i_flags = ort->rt6i_flags;
2241         rt6_set_from(rt, ort);
2242         rt->rt6i_metric = ort->rt6i_metric;
2243 #ifdef CONFIG_IPV6_SUBTREES
2244         rt->rt6i_src = ort->rt6i_src;
2245 #endif
2246         rt->rt6i_prefsrc = ort->rt6i_prefsrc;
2247         rt->rt6i_table = ort->rt6i_table;
2248         rt->dst.lwtstate = lwtstate_get(ort->dst.lwtstate);
2249 }
2250
2251 #ifdef CONFIG_IPV6_ROUTE_INFO
2252 static struct rt6_info *rt6_get_route_info(struct net *net,
2253                                            const struct in6_addr *prefix, int prefixlen,
2254                                            const struct in6_addr *gwaddr, int ifindex)
2255 {
2256         struct fib6_node *fn;
2257         struct rt6_info *rt = NULL;
2258         struct fib6_table *table;
2259
2260         table = fib6_get_table(net, RT6_TABLE_INFO);
2261         if (!table)
2262                 return NULL;
2263
2264         read_lock_bh(&table->tb6_lock);
2265         fn = fib6_locate(&table->tb6_root, prefix, prefixlen, NULL, 0);
2266         if (!fn)
2267                 goto out;
2268
2269         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2270                 if (rt->dst.dev->ifindex != ifindex)
2271                         continue;
2272                 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
2273                         continue;
2274                 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
2275                         continue;
2276                 dst_hold(&rt->dst);
2277                 break;
2278         }
2279 out:
2280         read_unlock_bh(&table->tb6_lock);
2281         return rt;
2282 }
2283
2284 static struct rt6_info *rt6_add_route_info(struct net *net,
2285                                            const struct in6_addr *prefix, int prefixlen,
2286                                            const struct in6_addr *gwaddr, int ifindex,
2287                                            unsigned int pref)
2288 {
2289         struct fib6_config cfg = {
2290                 .fc_table       = RT6_TABLE_INFO,
2291                 .fc_metric      = IP6_RT_PRIO_USER,
2292                 .fc_ifindex     = ifindex,
2293                 .fc_dst_len     = prefixlen,
2294                 .fc_flags       = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
2295                                   RTF_UP | RTF_PREF(pref),
2296                 .fc_nlinfo.portid = 0,
2297                 .fc_nlinfo.nlh = NULL,
2298                 .fc_nlinfo.nl_net = net,
2299         };
2300
2301         cfg.fc_dst = *prefix;
2302         cfg.fc_gateway = *gwaddr;
2303
2304         /* We should treat it as a default route if prefix length is 0. */
2305         if (!prefixlen)
2306                 cfg.fc_flags |= RTF_DEFAULT;
2307
2308         ip6_route_add(&cfg);
2309
2310         return rt6_get_route_info(net, prefix, prefixlen, gwaddr, ifindex);
2311 }
2312 #endif
2313
2314 struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
2315 {
2316         struct rt6_info *rt;
2317         struct fib6_table *table;
2318
2319         table = fib6_get_table(dev_net(dev), RT6_TABLE_DFLT);
2320         if (!table)
2321                 return NULL;
2322
2323         read_lock_bh(&table->tb6_lock);
2324         for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2325                 if (dev == rt->dst.dev &&
2326                     ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
2327                     ipv6_addr_equal(&rt->rt6i_gateway, addr))
2328                         break;
2329         }
2330         if (rt)
2331                 dst_hold(&rt->dst);
2332         read_unlock_bh(&table->tb6_lock);
2333         return rt;
2334 }
2335
2336 struct rt6_info *rt6_add_dflt_router(const struct in6_addr *gwaddr,
2337                                      struct net_device *dev,
2338                                      unsigned int pref)
2339 {
2340         struct fib6_config cfg = {
2341                 .fc_table       = RT6_TABLE_DFLT,
2342                 .fc_metric      = IP6_RT_PRIO_USER,
2343                 .fc_ifindex     = dev->ifindex,
2344                 .fc_flags       = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
2345                                   RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
2346                 .fc_nlinfo.portid = 0,
2347                 .fc_nlinfo.nlh = NULL,
2348                 .fc_nlinfo.nl_net = dev_net(dev),
2349         };
2350
2351         cfg.fc_gateway = *gwaddr;
2352
2353         ip6_route_add(&cfg);
2354
2355         return rt6_get_dflt_router(gwaddr, dev);
2356 }
2357
2358 void rt6_purge_dflt_routers(struct net *net)
2359 {
2360         struct rt6_info *rt;
2361         struct fib6_table *table;
2362
2363         /* NOTE: Keep consistent with rt6_get_dflt_router */
2364         table = fib6_get_table(net, RT6_TABLE_DFLT);
2365         if (!table)
2366                 return;
2367
2368 restart:
2369         read_lock_bh(&table->tb6_lock);
2370         for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2371                 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
2372                     (!rt->rt6i_idev || rt->rt6i_idev->cnf.accept_ra != 2)) {
2373                         dst_hold(&rt->dst);
2374                         read_unlock_bh(&table->tb6_lock);
2375                         ip6_del_rt(rt);
2376                         goto restart;
2377                 }
2378         }
2379         read_unlock_bh(&table->tb6_lock);
2380 }
2381
2382 static void rtmsg_to_fib6_config(struct net *net,
2383                                  struct in6_rtmsg *rtmsg,
2384                                  struct fib6_config *cfg)
2385 {
2386         memset(cfg, 0, sizeof(*cfg));
2387
2388         cfg->fc_table = RT6_TABLE_MAIN;
2389         cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
2390         cfg->fc_metric = rtmsg->rtmsg_metric;
2391         cfg->fc_expires = rtmsg->rtmsg_info;
2392         cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
2393         cfg->fc_src_len = rtmsg->rtmsg_src_len;
2394         cfg->fc_flags = rtmsg->rtmsg_flags;
2395
2396         cfg->fc_nlinfo.nl_net = net;
2397
2398         cfg->fc_dst = rtmsg->rtmsg_dst;
2399         cfg->fc_src = rtmsg->rtmsg_src;
2400         cfg->fc_gateway = rtmsg->rtmsg_gateway;
2401 }
2402
2403 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
2404 {
2405         struct fib6_config cfg;
2406         struct in6_rtmsg rtmsg;
2407         int err;
2408
2409         switch (cmd) {
2410         case SIOCADDRT:         /* Add a route */
2411         case SIOCDELRT:         /* Delete a route */
2412                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2413                         return -EPERM;
2414                 err = copy_from_user(&rtmsg, arg,
2415                                      sizeof(struct in6_rtmsg));
2416                 if (err)
2417                         return -EFAULT;
2418
2419                 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
2420
2421                 rtnl_lock();
2422                 switch (cmd) {
2423                 case SIOCADDRT:
2424                         err = ip6_route_add(&cfg);
2425                         break;
2426                 case SIOCDELRT:
2427                         err = ip6_route_del(&cfg);
2428                         break;
2429                 default:
2430                         err = -EINVAL;
2431                 }
2432                 rtnl_unlock();
2433
2434                 return err;
2435         }
2436
2437         return -EINVAL;
2438 }
2439
2440 /*
2441  *      Drop the packet on the floor
2442  */
2443
2444 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
2445 {
2446         int type;
2447         struct dst_entry *dst = skb_dst(skb);
2448         switch (ipstats_mib_noroutes) {
2449         case IPSTATS_MIB_INNOROUTES:
2450                 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
2451                 if (type == IPV6_ADDR_ANY) {
2452                         IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2453                                       IPSTATS_MIB_INADDRERRORS);
2454                         break;
2455                 }
2456                 /* FALLTHROUGH */
2457         case IPSTATS_MIB_OUTNOROUTES:
2458                 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2459                               ipstats_mib_noroutes);
2460                 break;
2461         }
2462         icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
2463         kfree_skb(skb);
2464         return 0;
2465 }
2466
2467 static int ip6_pkt_discard(struct sk_buff *skb)
2468 {
2469         return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
2470 }
2471
2472 static int ip6_pkt_discard_out(struct sock *sk, struct sk_buff *skb)
2473 {
2474         skb->dev = skb_dst(skb)->dev;
2475         return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
2476 }
2477
2478 static int ip6_pkt_prohibit(struct sk_buff *skb)
2479 {
2480         return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
2481 }
2482
2483 static int ip6_pkt_prohibit_out(struct sock *sk, struct sk_buff *skb)
2484 {
2485         skb->dev = skb_dst(skb)->dev;
2486         return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
2487 }
2488
2489 /*
2490  *      Allocate a dst for local (unicast / anycast) address.
2491  */
2492
2493 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
2494                                     const struct in6_addr *addr,
2495                                     bool anycast)
2496 {
2497         struct net *net = dev_net(idev->dev);
2498         struct rt6_info *rt = ip6_dst_alloc(net, net->loopback_dev,
2499                                             DST_NOCOUNT);
2500         if (!rt)
2501                 return ERR_PTR(-ENOMEM);
2502
2503         in6_dev_hold(idev);
2504
2505         rt->dst.flags |= DST_HOST;
2506         rt->dst.input = ip6_input;
2507         rt->dst.output = ip6_output;
2508         rt->rt6i_idev = idev;
2509
2510         rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
2511         if (anycast)
2512                 rt->rt6i_flags |= RTF_ANYCAST;
2513         else
2514                 rt->rt6i_flags |= RTF_LOCAL;
2515
2516         rt->rt6i_gateway  = *addr;
2517         rt->rt6i_dst.addr = *addr;
2518         rt->rt6i_dst.plen = 128;
2519         rt->rt6i_table = fib6_get_table(net, RT6_TABLE_LOCAL);
2520         rt->dst.flags |= DST_NOCACHE;
2521
2522         atomic_set(&rt->dst.__refcnt, 1);
2523
2524         return rt;
2525 }
2526
2527 int ip6_route_get_saddr(struct net *net,
2528                         struct rt6_info *rt,
2529                         const struct in6_addr *daddr,
2530                         unsigned int prefs,
2531                         struct in6_addr *saddr)
2532 {
2533         struct inet6_dev *idev =
2534                 rt ? ip6_dst_idev((struct dst_entry *)rt) : NULL;
2535         int err = 0;
2536         if (rt && rt->rt6i_prefsrc.plen)
2537                 *saddr = rt->rt6i_prefsrc.addr;
2538         else
2539                 err = ipv6_dev_get_saddr(net, idev ? idev->dev : NULL,
2540                                          daddr, prefs, saddr);
2541         return err;
2542 }
2543
2544 /* remove deleted ip from prefsrc entries */
2545 struct arg_dev_net_ip {
2546         struct net_device *dev;
2547         struct net *net;
2548         struct in6_addr *addr;
2549 };
2550
2551 static int fib6_remove_prefsrc(struct rt6_info *rt, void *arg)
2552 {
2553         struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
2554         struct net *net = ((struct arg_dev_net_ip *)arg)->net;
2555         struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
2556
2557         if (((void *)rt->dst.dev == dev || !dev) &&
2558             rt != net->ipv6.ip6_null_entry &&
2559             ipv6_addr_equal(addr, &rt->rt6i_prefsrc.addr)) {
2560                 /* remove prefsrc entry */
2561                 rt->rt6i_prefsrc.plen = 0;
2562         }
2563         return 0;
2564 }
2565
2566 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
2567 {
2568         struct net *net = dev_net(ifp->idev->dev);
2569         struct arg_dev_net_ip adni = {
2570                 .dev = ifp->idev->dev,
2571                 .net = net,
2572                 .addr = &ifp->addr,
2573         };
2574         fib6_clean_all(net, fib6_remove_prefsrc, &adni);
2575 }
2576
2577 #define RTF_RA_ROUTER           (RTF_ADDRCONF | RTF_DEFAULT | RTF_GATEWAY)
2578 #define RTF_CACHE_GATEWAY       (RTF_GATEWAY | RTF_CACHE)
2579
2580 /* Remove routers and update dst entries when gateway turn into host. */
2581 static int fib6_clean_tohost(struct rt6_info *rt, void *arg)
2582 {
2583         struct in6_addr *gateway = (struct in6_addr *)arg;
2584
2585         if ((((rt->rt6i_flags & RTF_RA_ROUTER) == RTF_RA_ROUTER) ||
2586              ((rt->rt6i_flags & RTF_CACHE_GATEWAY) == RTF_CACHE_GATEWAY)) &&
2587              ipv6_addr_equal(gateway, &rt->rt6i_gateway)) {
2588                 return -1;
2589         }
2590         return 0;
2591 }
2592
2593 void rt6_clean_tohost(struct net *net, struct in6_addr *gateway)
2594 {
2595         fib6_clean_all(net, fib6_clean_tohost, gateway);
2596 }
2597
2598 struct arg_dev_net {
2599         struct net_device *dev;
2600         struct net *net;
2601 };
2602
2603 static int fib6_ifdown(struct rt6_info *rt, void *arg)
2604 {
2605         const struct arg_dev_net *adn = arg;
2606         const struct net_device *dev = adn->dev;
2607
2608         if ((rt->dst.dev == dev || !dev) &&
2609             rt != adn->net->ipv6.ip6_null_entry)
2610                 return -1;
2611
2612         return 0;
2613 }
2614
2615 void rt6_ifdown(struct net *net, struct net_device *dev)
2616 {
2617         struct arg_dev_net adn = {
2618                 .dev = dev,
2619                 .net = net,
2620         };
2621
2622         fib6_clean_all(net, fib6_ifdown, &adn);
2623         icmp6_clean_all(fib6_ifdown, &adn);
2624         rt6_uncached_list_flush_dev(net, dev);
2625 }
2626
2627 struct rt6_mtu_change_arg {
2628         struct net_device *dev;
2629         unsigned int mtu;
2630 };
2631
2632 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2633 {
2634         struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2635         struct inet6_dev *idev;
2636
2637         /* In IPv6 pmtu discovery is not optional,
2638            so that RTAX_MTU lock cannot disable it.
2639            We still use this lock to block changes
2640            caused by addrconf/ndisc.
2641         */
2642
2643         idev = __in6_dev_get(arg->dev);
2644         if (!idev)
2645                 return 0;
2646
2647         /* For administrative MTU increase, there is no way to discover
2648            IPv6 PMTU increase, so PMTU increase should be updated here.
2649            Since RFC 1981 doesn't include administrative MTU increase
2650            update PMTU increase is a MUST. (i.e. jumbo frame)
2651          */
2652         /*
2653            If new MTU is less than route PMTU, this new MTU will be the
2654            lowest MTU in the path, update the route PMTU to reflect PMTU
2655            decreases; if new MTU is greater than route PMTU, and the
2656            old MTU is the lowest MTU in the path, update the route PMTU
2657            to reflect the increase. In this case if the other nodes' MTU
2658            also have the lowest MTU, TOO BIG MESSAGE will be lead to
2659            PMTU discouvery.
2660          */
2661         if (rt->dst.dev == arg->dev &&
2662             !dst_metric_locked(&rt->dst, RTAX_MTU)) {
2663                 if (rt->rt6i_flags & RTF_CACHE) {
2664                         /* For RTF_CACHE with rt6i_pmtu == 0
2665                          * (i.e. a redirected route),
2666                          * the metrics of its rt->dst.from has already
2667                          * been updated.
2668                          */
2669                         if (rt->rt6i_pmtu && rt->rt6i_pmtu > arg->mtu)
2670                                 rt->rt6i_pmtu = arg->mtu;
2671                 } else if (dst_mtu(&rt->dst) >= arg->mtu ||
2672                            (dst_mtu(&rt->dst) < arg->mtu &&
2673                             dst_mtu(&rt->dst) == idev->cnf.mtu6)) {
2674                         dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
2675                 }
2676         }
2677         return 0;
2678 }
2679
2680 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
2681 {
2682         struct rt6_mtu_change_arg arg = {
2683                 .dev = dev,
2684                 .mtu = mtu,
2685         };
2686
2687         fib6_clean_all(dev_net(dev), rt6_mtu_change_route, &arg);
2688 }
2689
2690 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2691         [RTA_GATEWAY]           = { .len = sizeof(struct in6_addr) },
2692         [RTA_OIF]               = { .type = NLA_U32 },
2693         [RTA_IIF]               = { .type = NLA_U32 },
2694         [RTA_PRIORITY]          = { .type = NLA_U32 },
2695         [RTA_METRICS]           = { .type = NLA_NESTED },
2696         [RTA_MULTIPATH]         = { .len = sizeof(struct rtnexthop) },
2697         [RTA_PREF]              = { .type = NLA_U8 },
2698         [RTA_ENCAP_TYPE]        = { .type = NLA_U16 },
2699         [RTA_ENCAP]             = { .type = NLA_NESTED },
2700 };
2701
2702 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2703                               struct fib6_config *cfg)
2704 {
2705         struct rtmsg *rtm;
2706         struct nlattr *tb[RTA_MAX+1];
2707         unsigned int pref;
2708         int err;
2709
2710         err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2711         if (err < 0)
2712                 goto errout;
2713
2714         err = -EINVAL;
2715         rtm = nlmsg_data(nlh);
2716         memset(cfg, 0, sizeof(*cfg));
2717
2718         cfg->fc_table = rtm->rtm_table;
2719         cfg->fc_dst_len = rtm->rtm_dst_len;
2720         cfg->fc_src_len = rtm->rtm_src_len;
2721         cfg->fc_flags = RTF_UP;
2722         cfg->fc_protocol = rtm->rtm_protocol;
2723         cfg->fc_type = rtm->rtm_type;
2724
2725         if (rtm->rtm_type == RTN_UNREACHABLE ||
2726             rtm->rtm_type == RTN_BLACKHOLE ||
2727             rtm->rtm_type == RTN_PROHIBIT ||
2728             rtm->rtm_type == RTN_THROW)
2729                 cfg->fc_flags |= RTF_REJECT;
2730
2731         if (rtm->rtm_type == RTN_LOCAL)
2732                 cfg->fc_flags |= RTF_LOCAL;
2733
2734         if (rtm->rtm_flags & RTM_F_CLONED)
2735                 cfg->fc_flags |= RTF_CACHE;
2736
2737         cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid;
2738         cfg->fc_nlinfo.nlh = nlh;
2739         cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2740
2741         if (tb[RTA_GATEWAY]) {
2742                 cfg->fc_gateway = nla_get_in6_addr(tb[RTA_GATEWAY]);
2743                 cfg->fc_flags |= RTF_GATEWAY;
2744         }
2745
2746         if (tb[RTA_DST]) {
2747                 int plen = (rtm->rtm_dst_len + 7) >> 3;
2748
2749                 if (nla_len(tb[RTA_DST]) < plen)
2750                         goto errout;
2751
2752                 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2753         }
2754
2755         if (tb[RTA_SRC]) {
2756                 int plen = (rtm->rtm_src_len + 7) >> 3;
2757
2758                 if (nla_len(tb[RTA_SRC]) < plen)
2759                         goto errout;
2760
2761                 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2762         }
2763
2764         if (tb[RTA_PREFSRC])
2765                 cfg->fc_prefsrc = nla_get_in6_addr(tb[RTA_PREFSRC]);
2766
2767         if (tb[RTA_OIF])
2768                 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2769
2770         if (tb[RTA_PRIORITY])
2771                 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2772
2773         if (tb[RTA_METRICS]) {
2774                 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2775                 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2776         }
2777
2778         if (tb[RTA_TABLE])
2779                 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2780
2781         if (tb[RTA_MULTIPATH]) {
2782                 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
2783                 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
2784         }
2785
2786         if (tb[RTA_PREF]) {
2787                 pref = nla_get_u8(tb[RTA_PREF]);
2788                 if (pref != ICMPV6_ROUTER_PREF_LOW &&
2789                     pref != ICMPV6_ROUTER_PREF_HIGH)
2790                         pref = ICMPV6_ROUTER_PREF_MEDIUM;
2791                 cfg->fc_flags |= RTF_PREF(pref);
2792         }
2793
2794         if (tb[RTA_ENCAP])
2795                 cfg->fc_encap = tb[RTA_ENCAP];
2796
2797         if (tb[RTA_ENCAP_TYPE])
2798                 cfg->fc_encap_type = nla_get_u16(tb[RTA_ENCAP_TYPE]);
2799
2800         err = 0;
2801 errout:
2802         return err;
2803 }
2804
2805 struct rt6_nh {
2806         struct rt6_info *rt6_info;
2807         struct fib6_config r_cfg;
2808         struct mx6_config mxc;
2809         struct list_head next;
2810 };
2811
2812 static void ip6_print_replace_route_err(struct list_head *rt6_nh_list)
2813 {
2814         struct rt6_nh *nh;
2815
2816         list_for_each_entry(nh, rt6_nh_list, next) {
2817                 pr_warn("IPV6: multipath route replace failed (check consistency of installed routes): %pI6 nexthop %pI6 ifi %d\n",
2818                         &nh->r_cfg.fc_dst, &nh->r_cfg.fc_gateway,
2819                         nh->r_cfg.fc_ifindex);
2820         }
2821 }
2822
2823 static int ip6_route_info_append(struct list_head *rt6_nh_list,
2824                                  struct rt6_info *rt, struct fib6_config *r_cfg)
2825 {
2826         struct rt6_nh *nh;
2827         struct rt6_info *rtnh;
2828         int err = -EEXIST;
2829
2830         list_for_each_entry(nh, rt6_nh_list, next) {
2831                 /* check if rt6_info already exists */
2832                 rtnh = nh->rt6_info;
2833
2834                 if (rtnh->dst.dev == rt->dst.dev &&
2835                     rtnh->rt6i_idev == rt->rt6i_idev &&
2836                     ipv6_addr_equal(&rtnh->rt6i_gateway,
2837                                     &rt->rt6i_gateway))
2838                         return err;
2839         }
2840
2841         nh = kzalloc(sizeof(*nh), GFP_KERNEL);
2842         if (!nh)
2843                 return -ENOMEM;
2844         nh->rt6_info = rt;
2845         err = ip6_convert_metrics(&nh->mxc, r_cfg);
2846         if (err) {
2847                 kfree(nh);
2848                 return err;
2849         }
2850         memcpy(&nh->r_cfg, r_cfg, sizeof(*r_cfg));
2851         list_add_tail(&nh->next, rt6_nh_list);
2852
2853         return 0;
2854 }
2855
2856 static int ip6_route_multipath_add(struct fib6_config *cfg)
2857 {
2858         struct fib6_config r_cfg;
2859         struct rtnexthop *rtnh;
2860         struct rt6_info *rt;
2861         struct rt6_nh *err_nh;
2862         struct rt6_nh *nh, *nh_safe;
2863         int remaining;
2864         int attrlen;
2865         int err = 1;
2866         int nhn = 0;
2867         int replace = (cfg->fc_nlinfo.nlh &&
2868                        (cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_REPLACE));
2869         LIST_HEAD(rt6_nh_list);
2870
2871         remaining = cfg->fc_mp_len;
2872         rtnh = (struct rtnexthop *)cfg->fc_mp;
2873
2874         /* Parse a Multipath Entry and build a list (rt6_nh_list) of
2875          * rt6_info structs per nexthop
2876          */
2877         while (rtnh_ok(rtnh, remaining)) {
2878                 memcpy(&r_cfg, cfg, sizeof(*cfg));
2879                 if (rtnh->rtnh_ifindex)
2880                         r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
2881
2882                 attrlen = rtnh_attrlen(rtnh);
2883                 if (attrlen > 0) {
2884                         struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
2885
2886                         nla = nla_find(attrs, attrlen, RTA_GATEWAY);
2887                         if (nla) {
2888                                 r_cfg.fc_gateway = nla_get_in6_addr(nla);
2889                                 r_cfg.fc_flags |= RTF_GATEWAY;
2890                         }
2891                         r_cfg.fc_encap = nla_find(attrs, attrlen, RTA_ENCAP);
2892                         nla = nla_find(attrs, attrlen, RTA_ENCAP_TYPE);
2893                         if (nla)
2894                                 r_cfg.fc_encap_type = nla_get_u16(nla);
2895                 }
2896
2897                 err = ip6_route_info_create(&r_cfg, &rt);
2898                 if (err)
2899                         goto cleanup;
2900
2901                 err = ip6_route_info_append(&rt6_nh_list, rt, &r_cfg);
2902                 if (err) {
2903                         dst_free(&rt->dst);
2904                         goto cleanup;
2905                 }
2906
2907                 rtnh = rtnh_next(rtnh, &remaining);
2908         }
2909
2910         err_nh = NULL;
2911         list_for_each_entry(nh, &rt6_nh_list, next) {
2912                 err = __ip6_ins_rt(nh->rt6_info, &cfg->fc_nlinfo, &nh->mxc);
2913                 /* nh->rt6_info is used or freed at this point, reset to NULL*/
2914                 nh->rt6_info = NULL;
2915                 if (err) {
2916                         if (replace && nhn)
2917                                 ip6_print_replace_route_err(&rt6_nh_list);
2918                         err_nh = nh;
2919                         goto add_errout;
2920                 }
2921
2922                 /* Because each route is added like a single route we remove
2923                  * these flags after the first nexthop: if there is a collision,
2924                  * we have already failed to add the first nexthop:
2925                  * fib6_add_rt2node() has rejected it; when replacing, old
2926                  * nexthops have been replaced by first new, the rest should
2927                  * be added to it.
2928                  */
2929                 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~(NLM_F_EXCL |
2930                                                      NLM_F_REPLACE);
2931                 nhn++;
2932         }
2933
2934         goto cleanup;
2935
2936 add_errout:
2937         /* Delete routes that were already added */
2938         list_for_each_entry(nh, &rt6_nh_list, next) {
2939                 if (err_nh == nh)
2940                         break;
2941                 ip6_route_del(&nh->r_cfg);
2942         }
2943
2944 cleanup:
2945         list_for_each_entry_safe(nh, nh_safe, &rt6_nh_list, next) {
2946                 if (nh->rt6_info)
2947                         dst_free(&nh->rt6_info->dst);
2948                 kfree(nh->mxc.mx);
2949                 list_del(&nh->next);
2950                 kfree(nh);
2951         }
2952
2953         return err;
2954 }
2955
2956 static int ip6_route_multipath_del(struct fib6_config *cfg)
2957 {
2958         struct fib6_config r_cfg;
2959         struct rtnexthop *rtnh;
2960         int remaining;
2961         int attrlen;
2962         int err = 1, last_err = 0;
2963
2964         remaining = cfg->fc_mp_len;
2965         rtnh = (struct rtnexthop *)cfg->fc_mp;
2966
2967         /* Parse a Multipath Entry */
2968         while (rtnh_ok(rtnh, remaining)) {
2969                 memcpy(&r_cfg, cfg, sizeof(*cfg));
2970                 if (rtnh->rtnh_ifindex)
2971                         r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
2972
2973                 attrlen = rtnh_attrlen(rtnh);
2974                 if (attrlen > 0) {
2975                         struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
2976
2977                         nla = nla_find(attrs, attrlen, RTA_GATEWAY);
2978                         if (nla) {
2979                                 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
2980                                 r_cfg.fc_flags |= RTF_GATEWAY;
2981                         }
2982                 }
2983                 err = ip6_route_del(&r_cfg);
2984                 if (err)
2985                         last_err = err;
2986
2987                 rtnh = rtnh_next(rtnh, &remaining);
2988         }
2989
2990         return last_err;
2991 }
2992
2993 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh)
2994 {
2995         struct fib6_config cfg;
2996         int err;
2997
2998         err = rtm_to_fib6_config(skb, nlh, &cfg);
2999         if (err < 0)
3000                 return err;
3001
3002         if (cfg.fc_mp)
3003                 return ip6_route_multipath_del(&cfg);
3004         else
3005                 return ip6_route_del(&cfg);
3006 }
3007
3008 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh)
3009 {
3010         struct fib6_config cfg;
3011         int err;
3012
3013         err = rtm_to_fib6_config(skb, nlh, &cfg);
3014         if (err < 0)
3015                 return err;
3016
3017         if (cfg.fc_mp)
3018                 return ip6_route_multipath_add(&cfg);
3019         else
3020                 return ip6_route_add(&cfg);
3021 }
3022
3023 static inline size_t rt6_nlmsg_size(struct rt6_info *rt)
3024 {
3025         return NLMSG_ALIGN(sizeof(struct rtmsg))
3026                + nla_total_size(16) /* RTA_SRC */
3027                + nla_total_size(16) /* RTA_DST */
3028                + nla_total_size(16) /* RTA_GATEWAY */
3029                + nla_total_size(16) /* RTA_PREFSRC */
3030                + nla_total_size(4) /* RTA_TABLE */
3031                + nla_total_size(4) /* RTA_IIF */
3032                + nla_total_size(4) /* RTA_OIF */
3033                + nla_total_size(4) /* RTA_PRIORITY */
3034                + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
3035                + nla_total_size(sizeof(struct rta_cacheinfo))
3036                + nla_total_size(TCP_CA_NAME_MAX) /* RTAX_CC_ALGO */
3037                + nla_total_size(1) /* RTA_PREF */
3038                + lwtunnel_get_encap_size(rt->dst.lwtstate);
3039 }
3040
3041 static int rt6_fill_node(struct net *net,
3042                          struct sk_buff *skb, struct rt6_info *rt,
3043                          struct in6_addr *dst, struct in6_addr *src,
3044                          int iif, int type, u32 portid, u32 seq,
3045                          int prefix, int nowait, unsigned int flags)
3046 {
3047         u32 metrics[RTAX_MAX];
3048         struct rtmsg *rtm;
3049         struct nlmsghdr *nlh;
3050         long expires;
3051         u32 table;
3052
3053         if (prefix) {   /* user wants prefix routes only */
3054                 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
3055                         /* success since this is not a prefix route */
3056                         return 1;
3057                 }
3058         }
3059
3060         nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
3061         if (!nlh)
3062                 return -EMSGSIZE;
3063
3064         rtm = nlmsg_data(nlh);
3065         rtm->rtm_family = AF_INET6;
3066         rtm->rtm_dst_len = rt->rt6i_dst.plen;
3067         rtm->rtm_src_len = rt->rt6i_src.plen;
3068         rtm->rtm_tos = 0;
3069         if (rt->rt6i_table)
3070                 table = rt->rt6i_table->tb6_id;
3071         else
3072                 table = RT6_TABLE_UNSPEC;
3073         rtm->rtm_table = table;
3074         if (nla_put_u32(skb, RTA_TABLE, table))
3075                 goto nla_put_failure;
3076         if (rt->rt6i_flags & RTF_REJECT) {
3077                 switch (rt->dst.error) {
3078                 case -EINVAL:
3079                         rtm->rtm_type = RTN_BLACKHOLE;
3080                         break;
3081                 case -EACCES:
3082                         rtm->rtm_type = RTN_PROHIBIT;
3083                         break;
3084                 case -EAGAIN:
3085                         rtm->rtm_type = RTN_THROW;
3086                         break;
3087                 default:
3088                         rtm->rtm_type = RTN_UNREACHABLE;
3089                         break;
3090                 }
3091         }
3092         else if (rt->rt6i_flags & RTF_LOCAL)
3093                 rtm->rtm_type = RTN_LOCAL;
3094         else if (rt->dst.dev && (rt->dst.dev->flags & IFF_LOOPBACK))
3095                 rtm->rtm_type = RTN_LOCAL;
3096         else
3097                 rtm->rtm_type = RTN_UNICAST;
3098         rtm->rtm_flags = 0;
3099         if (!netif_carrier_ok(rt->dst.dev)) {
3100                 rtm->rtm_flags |= RTNH_F_LINKDOWN;
3101                 if (rt->rt6i_idev->cnf.ignore_routes_with_linkdown)
3102                         rtm->rtm_flags |= RTNH_F_DEAD;
3103         }
3104         rtm->rtm_scope = RT_SCOPE_UNIVERSE;
3105         rtm->rtm_protocol = rt->rt6i_protocol;
3106         if (rt->rt6i_flags & RTF_DYNAMIC)
3107                 rtm->rtm_protocol = RTPROT_REDIRECT;
3108         else if (rt->rt6i_flags & RTF_ADDRCONF) {
3109                 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ROUTEINFO))
3110                         rtm->rtm_protocol = RTPROT_RA;
3111                 else
3112                         rtm->rtm_protocol = RTPROT_KERNEL;
3113         }
3114
3115         if (rt->rt6i_flags & RTF_CACHE)
3116                 rtm->rtm_flags |= RTM_F_CLONED;
3117
3118         if (dst) {
3119                 if (nla_put_in6_addr(skb, RTA_DST, dst))
3120                         goto nla_put_failure;
3121                 rtm->rtm_dst_len = 128;
3122         } else if (rtm->rtm_dst_len)
3123                 if (nla_put_in6_addr(skb, RTA_DST, &rt->rt6i_dst.addr))
3124                         goto nla_put_failure;
3125 #ifdef CONFIG_IPV6_SUBTREES
3126         if (src) {
3127                 if (nla_put_in6_addr(skb, RTA_SRC, src))
3128                         goto nla_put_failure;
3129                 rtm->rtm_src_len = 128;
3130         } else if (rtm->rtm_src_len &&
3131                    nla_put_in6_addr(skb, RTA_SRC, &rt->rt6i_src.addr))
3132                 goto nla_put_failure;
3133 #endif
3134         if (iif) {
3135 #ifdef CONFIG_IPV6_MROUTE
3136                 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
3137                         int err = ip6mr_get_route(net, skb, rtm, nowait);
3138                         if (err <= 0) {
3139                                 if (!nowait) {
3140                                         if (err == 0)
3141                                                 return 0;
3142                                         goto nla_put_failure;
3143                                 } else {
3144                                         if (err == -EMSGSIZE)
3145                                                 goto nla_put_failure;
3146                                 }
3147                         }
3148                 } else
3149 #endif
3150                         if (nla_put_u32(skb, RTA_IIF, iif))
3151                                 goto nla_put_failure;
3152         } else if (dst) {
3153                 struct in6_addr saddr_buf;
3154                 if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0 &&
3155                     nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
3156                         goto nla_put_failure;
3157         }
3158
3159         if (rt->rt6i_prefsrc.plen) {
3160                 struct in6_addr saddr_buf;
3161                 saddr_buf = rt->rt6i_prefsrc.addr;
3162                 if (nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
3163                         goto nla_put_failure;
3164         }
3165
3166         memcpy(metrics, dst_metrics_ptr(&rt->dst), sizeof(metrics));
3167         if (rt->rt6i_pmtu)
3168                 metrics[RTAX_MTU - 1] = rt->rt6i_pmtu;
3169         if (rtnetlink_put_metrics(skb, metrics) < 0)
3170                 goto nla_put_failure;
3171
3172         if (rt->rt6i_flags & RTF_GATEWAY) {
3173                 if (nla_put_in6_addr(skb, RTA_GATEWAY, &rt->rt6i_gateway) < 0)
3174                         goto nla_put_failure;
3175         }
3176
3177         if (rt->dst.dev &&
3178             nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex))
3179                 goto nla_put_failure;
3180         if (nla_put_u32(skb, RTA_PRIORITY, rt->rt6i_metric))
3181                 goto nla_put_failure;
3182
3183         expires = (rt->rt6i_flags & RTF_EXPIRES) ? rt->dst.expires - jiffies : 0;
3184
3185         if (rtnl_put_cacheinfo(skb, &rt->dst, 0, expires, rt->dst.error) < 0)
3186                 goto nla_put_failure;
3187
3188         if (nla_put_u8(skb, RTA_PREF, IPV6_EXTRACT_PREF(rt->rt6i_flags)))
3189                 goto nla_put_failure;
3190
3191         lwtunnel_fill_encap(skb, rt->dst.lwtstate);
3192
3193         nlmsg_end(skb, nlh);
3194         return 0;
3195
3196 nla_put_failure:
3197         nlmsg_cancel(skb, nlh);
3198         return -EMSGSIZE;
3199 }
3200
3201 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
3202 {
3203         struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
3204         int prefix;
3205
3206         if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
3207                 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
3208                 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
3209         } else
3210                 prefix = 0;
3211
3212         return rt6_fill_node(arg->net,
3213                      arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
3214                      NETLINK_CB(arg->cb->skb).portid, arg->cb->nlh->nlmsg_seq,
3215                      prefix, 0, NLM_F_MULTI);
3216 }
3217
3218 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh)
3219 {
3220         struct net *net = sock_net(in_skb->sk);
3221         struct nlattr *tb[RTA_MAX+1];
3222         struct rt6_info *rt;
3223         struct sk_buff *skb;
3224         struct rtmsg *rtm;
3225         struct flowi6 fl6;
3226         int err, iif = 0, oif = 0;
3227
3228         err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
3229         if (err < 0)
3230                 goto errout;
3231
3232         err = -EINVAL;
3233         memset(&fl6, 0, sizeof(fl6));
3234
3235         if (tb[RTA_SRC]) {
3236                 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
3237                         goto errout;
3238
3239                 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
3240         }
3241
3242         if (tb[RTA_DST]) {
3243                 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
3244                         goto errout;
3245
3246                 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
3247         }
3248
3249         if (tb[RTA_IIF])
3250                 iif = nla_get_u32(tb[RTA_IIF]);
3251
3252         if (tb[RTA_OIF])
3253                 oif = nla_get_u32(tb[RTA_OIF]);
3254
3255         if (tb[RTA_MARK])
3256                 fl6.flowi6_mark = nla_get_u32(tb[RTA_MARK]);
3257
3258         if (iif) {
3259                 struct net_device *dev;
3260                 int flags = 0;
3261
3262                 dev = __dev_get_by_index(net, iif);
3263                 if (!dev) {
3264                         err = -ENODEV;
3265                         goto errout;
3266                 }
3267
3268                 fl6.flowi6_iif = iif;
3269
3270                 if (!ipv6_addr_any(&fl6.saddr))
3271                         flags |= RT6_LOOKUP_F_HAS_SADDR;
3272
3273                 rt = (struct rt6_info *)ip6_route_input_lookup(net, dev, &fl6,
3274                                                                flags);
3275         } else {
3276                 fl6.flowi6_oif = oif;
3277
3278                 rt = (struct rt6_info *)ip6_route_output(net, NULL, &fl6);
3279         }
3280
3281         skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
3282         if (!skb) {
3283                 ip6_rt_put(rt);
3284                 err = -ENOBUFS;
3285                 goto errout;
3286         }
3287
3288         /* Reserve room for dummy headers, this skb can pass
3289            through good chunk of routing engine.
3290          */
3291         skb_reset_mac_header(skb);
3292         skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
3293
3294         skb_dst_set(skb, &rt->dst);
3295
3296         err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
3297                             RTM_NEWROUTE, NETLINK_CB(in_skb).portid,
3298                             nlh->nlmsg_seq, 0, 0, 0);
3299         if (err < 0) {
3300                 kfree_skb(skb);
3301                 goto errout;
3302         }
3303
3304         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
3305 errout:
3306         return err;
3307 }
3308
3309 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info,
3310                      unsigned int nlm_flags)
3311 {
3312         struct sk_buff *skb;
3313         struct net *net = info->nl_net;
3314         u32 seq;
3315         int err;
3316
3317         err = -ENOBUFS;
3318         seq = info->nlh ? info->nlh->nlmsg_seq : 0;
3319
3320         skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
3321         if (!skb)
3322                 goto errout;
3323
3324         err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
3325                                 event, info->portid, seq, 0, 0, nlm_flags);
3326         if (err < 0) {
3327                 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
3328                 WARN_ON(err == -EMSGSIZE);
3329                 kfree_skb(skb);
3330                 goto errout;
3331         }
3332         rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
3333                     info->nlh, gfp_any());
3334         return;
3335 errout:
3336         if (err < 0)
3337                 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
3338 }
3339
3340 static int ip6_route_dev_notify(struct notifier_block *this,
3341                                 unsigned long event, void *ptr)
3342 {
3343         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3344         struct net *net = dev_net(dev);
3345
3346         if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
3347                 net->ipv6.ip6_null_entry->dst.dev = dev;
3348                 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
3349 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3350                 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
3351                 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
3352                 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
3353                 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
3354 #endif
3355         }
3356
3357         return NOTIFY_OK;
3358 }
3359
3360 /*
3361  *      /proc
3362  */
3363
3364 #ifdef CONFIG_PROC_FS
3365
3366 static const struct file_operations ipv6_route_proc_fops = {
3367         .owner          = THIS_MODULE,
3368         .open           = ipv6_route_open,
3369         .read           = seq_read,
3370         .llseek         = seq_lseek,
3371         .release        = seq_release_net,
3372 };
3373
3374 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
3375 {
3376         struct net *net = (struct net *)seq->private;
3377         seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
3378                    net->ipv6.rt6_stats->fib_nodes,
3379                    net->ipv6.rt6_stats->fib_route_nodes,
3380                    net->ipv6.rt6_stats->fib_rt_alloc,
3381                    net->ipv6.rt6_stats->fib_rt_entries,
3382                    net->ipv6.rt6_stats->fib_rt_cache,
3383                    dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
3384                    net->ipv6.rt6_stats->fib_discarded_routes);
3385
3386         return 0;
3387 }
3388
3389 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
3390 {
3391         return single_open_net(inode, file, rt6_stats_seq_show);
3392 }
3393
3394 static const struct file_operations rt6_stats_seq_fops = {
3395         .owner   = THIS_MODULE,
3396         .open    = rt6_stats_seq_open,
3397         .read    = seq_read,
3398         .llseek  = seq_lseek,
3399         .release = single_release_net,
3400 };
3401 #endif  /* CONFIG_PROC_FS */
3402
3403 #ifdef CONFIG_SYSCTL
3404
3405 static
3406 int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
3407                               void __user *buffer, size_t *lenp, loff_t *ppos)
3408 {
3409         struct net *net;
3410         int delay;
3411         if (!write)
3412                 return -EINVAL;
3413
3414         net = (struct net *)ctl->extra1;
3415         delay = net->ipv6.sysctl.flush_delay;
3416         proc_dointvec(ctl, write, buffer, lenp, ppos);
3417         fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
3418         return 0;
3419 }
3420
3421 struct ctl_table ipv6_route_table_template[] = {
3422         {
3423                 .procname       =       "flush",
3424                 .data           =       &init_net.ipv6.sysctl.flush_delay,
3425                 .maxlen         =       sizeof(int),
3426                 .mode           =       0200,
3427                 .proc_handler   =       ipv6_sysctl_rtcache_flush
3428         },
3429         {
3430                 .procname       =       "gc_thresh",
3431                 .data           =       &ip6_dst_ops_template.gc_thresh,
3432                 .maxlen         =       sizeof(int),
3433                 .mode           =       0644,
3434                 .proc_handler   =       proc_dointvec,
3435         },
3436         {
3437                 .procname       =       "max_size",
3438                 .data           =       &init_net.ipv6.sysctl.ip6_rt_max_size,
3439                 .maxlen         =       sizeof(int),
3440                 .mode           =       0644,
3441                 .proc_handler   =       proc_dointvec,
3442         },
3443         {
3444                 .procname       =       "gc_min_interval",
3445                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
3446                 .maxlen         =       sizeof(int),
3447                 .mode           =       0644,
3448                 .proc_handler   =       proc_dointvec_jiffies,
3449         },
3450         {
3451                 .procname       =       "gc_timeout",
3452                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
3453                 .maxlen         =       sizeof(int),
3454                 .mode           =       0644,
3455                 .proc_handler   =       proc_dointvec_jiffies,
3456         },
3457         {
3458                 .procname       =       "gc_interval",
3459                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_interval,
3460                 .maxlen         =       sizeof(int),
3461                 .mode           =       0644,
3462                 .proc_handler   =       proc_dointvec_jiffies,
3463         },
3464         {
3465                 .procname       =       "gc_elasticity",
3466                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
3467                 .maxlen         =       sizeof(int),
3468                 .mode           =       0644,
3469                 .proc_handler   =       proc_dointvec,
3470         },
3471         {
3472                 .procname       =       "mtu_expires",
3473                 .data           =       &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
3474                 .maxlen         =       sizeof(int),
3475                 .mode           =       0644,
3476                 .proc_handler   =       proc_dointvec_jiffies,
3477         },
3478         {
3479                 .procname       =       "min_adv_mss",
3480                 .data           =       &init_net.ipv6.sysctl.ip6_rt_min_advmss,
3481                 .maxlen         =       sizeof(int),
3482                 .mode           =       0644,
3483                 .proc_handler   =       proc_dointvec,
3484         },
3485         {
3486                 .procname       =       "gc_min_interval_ms",
3487                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
3488                 .maxlen         =       sizeof(int),
3489                 .mode           =       0644,
3490                 .proc_handler   =       proc_dointvec_ms_jiffies,
3491         },
3492         { }
3493 };
3494
3495 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
3496 {
3497         struct ctl_table *table;
3498
3499         table = kmemdup(ipv6_route_table_template,
3500                         sizeof(ipv6_route_table_template),
3501                         GFP_KERNEL);
3502
3503         if (table) {
3504                 table[0].data = &net->ipv6.sysctl.flush_delay;
3505                 table[0].extra1 = net;
3506                 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
3507                 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
3508                 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
3509                 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
3510                 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
3511                 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
3512                 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
3513                 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
3514                 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
3515
3516                 /* Don't export sysctls to unprivileged users */
3517                 if (net->user_ns != &init_user_ns)
3518                         table[0].procname = NULL;
3519         }
3520
3521         return table;
3522 }
3523 #endif
3524
3525 static int __net_init ip6_route_net_init(struct net *net)
3526 {
3527         int ret = -ENOMEM;
3528
3529         memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
3530                sizeof(net->ipv6.ip6_dst_ops));
3531
3532         if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
3533                 goto out_ip6_dst_ops;
3534
3535         net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
3536                                            sizeof(*net->ipv6.ip6_null_entry),
3537                                            GFP_KERNEL);
3538         if (!net->ipv6.ip6_null_entry)
3539                 goto out_ip6_dst_entries;
3540         net->ipv6.ip6_null_entry->dst.path =
3541                 (struct dst_entry *)net->ipv6.ip6_null_entry;
3542         net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3543         dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
3544                          ip6_template_metrics, true);
3545
3546 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3547         net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
3548                                                sizeof(*net->ipv6.ip6_prohibit_entry),
3549                                                GFP_KERNEL);
3550         if (!net->ipv6.ip6_prohibit_entry)
3551                 goto out_ip6_null_entry;
3552         net->ipv6.ip6_prohibit_entry->dst.path =
3553                 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
3554         net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3555         dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
3556                          ip6_template_metrics, true);
3557
3558         net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
3559                                                sizeof(*net->ipv6.ip6_blk_hole_entry),
3560                                                GFP_KERNEL);
3561         if (!net->ipv6.ip6_blk_hole_entry)
3562                 goto out_ip6_prohibit_entry;
3563         net->ipv6.ip6_blk_hole_entry->dst.path =
3564                 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
3565         net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3566         dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
3567                          ip6_template_metrics, true);
3568 #endif
3569
3570         net->ipv6.sysctl.flush_delay = 0;
3571         net->ipv6.sysctl.ip6_rt_max_size = 4096;
3572         net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
3573         net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
3574         net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
3575         net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
3576         net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
3577         net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
3578
3579         net->ipv6.ip6_rt_gc_expire = 30*HZ;
3580
3581         ret = 0;
3582 out:
3583         return ret;
3584
3585 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3586 out_ip6_prohibit_entry:
3587         kfree(net->ipv6.ip6_prohibit_entry);
3588 out_ip6_null_entry:
3589         kfree(net->ipv6.ip6_null_entry);
3590 #endif
3591 out_ip6_dst_entries:
3592         dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3593 out_ip6_dst_ops:
3594         goto out;
3595 }
3596
3597 static void __net_exit ip6_route_net_exit(struct net *net)
3598 {
3599         kfree(net->ipv6.ip6_null_entry);
3600 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3601         kfree(net->ipv6.ip6_prohibit_entry);
3602         kfree(net->ipv6.ip6_blk_hole_entry);
3603 #endif
3604         dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3605 }
3606
3607 static int __net_init ip6_route_net_init_late(struct net *net)
3608 {
3609 #ifdef CONFIG_PROC_FS
3610         proc_create("ipv6_route", 0, net->proc_net, &ipv6_route_proc_fops);
3611         proc_create("rt6_stats", S_IRUGO, net->proc_net, &rt6_stats_seq_fops);
3612 #endif
3613         return 0;
3614 }
3615
3616 static void __net_exit ip6_route_net_exit_late(struct net *net)
3617 {
3618 #ifdef CONFIG_PROC_FS
3619         remove_proc_entry("ipv6_route", net->proc_net);
3620         remove_proc_entry("rt6_stats", net->proc_net);
3621 #endif
3622 }
3623
3624 static struct pernet_operations ip6_route_net_ops = {
3625         .init = ip6_route_net_init,
3626         .exit = ip6_route_net_exit,
3627 };
3628
3629 static int __net_init ipv6_inetpeer_init(struct net *net)
3630 {
3631         struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
3632
3633         if (!bp)
3634                 return -ENOMEM;
3635         inet_peer_base_init(bp);
3636         net->ipv6.peers = bp;
3637         return 0;
3638 }
3639
3640 static void __net_exit ipv6_inetpeer_exit(struct net *net)
3641 {
3642         struct inet_peer_base *bp = net->ipv6.peers;
3643
3644         net->ipv6.peers = NULL;
3645         inetpeer_invalidate_tree(bp);
3646         kfree(bp);
3647 }
3648
3649 static struct pernet_operations ipv6_inetpeer_ops = {
3650         .init   =       ipv6_inetpeer_init,
3651         .exit   =       ipv6_inetpeer_exit,
3652 };
3653
3654 static struct pernet_operations ip6_route_net_late_ops = {
3655         .init = ip6_route_net_init_late,
3656         .exit = ip6_route_net_exit_late,
3657 };
3658
3659 static struct notifier_block ip6_route_dev_notifier = {
3660         .notifier_call = ip6_route_dev_notify,
3661         .priority = 0,
3662 };
3663
3664 int __init ip6_route_init(void)
3665 {
3666         int ret;
3667         int cpu;
3668
3669         ret = -ENOMEM;
3670         ip6_dst_ops_template.kmem_cachep =
3671                 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
3672                                   SLAB_HWCACHE_ALIGN, NULL);
3673         if (!ip6_dst_ops_template.kmem_cachep)
3674                 goto out;
3675
3676         ret = dst_entries_init(&ip6_dst_blackhole_ops);
3677         if (ret)
3678                 goto out_kmem_cache;
3679
3680         ret = register_pernet_subsys(&ipv6_inetpeer_ops);
3681         if (ret)
3682                 goto out_dst_entries;
3683
3684         ret = register_pernet_subsys(&ip6_route_net_ops);
3685         if (ret)
3686                 goto out_register_inetpeer;
3687
3688         ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
3689
3690         /* Registering of the loopback is done before this portion of code,
3691          * the loopback reference in rt6_info will not be taken, do it
3692          * manually for init_net */
3693         init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
3694         init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3695   #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3696         init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
3697         init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3698         init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
3699         init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3700   #endif
3701         ret = fib6_init();
3702         if (ret)
3703                 goto out_register_subsys;
3704
3705         ret = xfrm6_init();
3706         if (ret)
3707                 goto out_fib6_init;
3708
3709         ret = fib6_rules_init();
3710         if (ret)
3711                 goto xfrm6_init;
3712
3713         ret = register_pernet_subsys(&ip6_route_net_late_ops);
3714         if (ret)
3715                 goto fib6_rules_init;
3716
3717         ret = -ENOBUFS;
3718         if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL, NULL) ||
3719             __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL, NULL) ||
3720             __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL, NULL))
3721                 goto out_register_late_subsys;
3722
3723         ret = register_netdevice_notifier(&ip6_route_dev_notifier);
3724         if (ret)
3725                 goto out_register_late_subsys;
3726
3727         for_each_possible_cpu(cpu) {
3728                 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
3729
3730                 INIT_LIST_HEAD(&ul->head);
3731                 spin_lock_init(&ul->lock);
3732         }
3733
3734 out:
3735         return ret;
3736
3737 out_register_late_subsys:
3738         unregister_pernet_subsys(&ip6_route_net_late_ops);
3739 fib6_rules_init:
3740         fib6_rules_cleanup();
3741 xfrm6_init:
3742         xfrm6_fini();
3743 out_fib6_init:
3744         fib6_gc_cleanup();
3745 out_register_subsys:
3746         unregister_pernet_subsys(&ip6_route_net_ops);
3747 out_register_inetpeer:
3748         unregister_pernet_subsys(&ipv6_inetpeer_ops);
3749 out_dst_entries:
3750         dst_entries_destroy(&ip6_dst_blackhole_ops);
3751 out_kmem_cache:
3752         kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3753         goto out;
3754 }
3755
3756 void ip6_route_cleanup(void)
3757 {
3758         unregister_netdevice_notifier(&ip6_route_dev_notifier);
3759         unregister_pernet_subsys(&ip6_route_net_late_ops);
3760         fib6_rules_cleanup();
3761         xfrm6_fini();
3762         fib6_gc_cleanup();
3763         unregister_pernet_subsys(&ipv6_inetpeer_ops);
3764         unregister_pernet_subsys(&ip6_route_net_ops);
3765         dst_entries_destroy(&ip6_dst_blackhole_ops);
3766         kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3767 }