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_RX)
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                         if (ip6_del_rt(rt))
1326                                 dst_free(&rt->dst);
1327                 } else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT)) {
1328                         rt->rt6i_node->fn_sernum = -1;
1329                 }
1330         }
1331 }
1332
1333 static void rt6_do_update_pmtu(struct rt6_info *rt, u32 mtu)
1334 {
1335         struct net *net = dev_net(rt->dst.dev);
1336
1337         rt->rt6i_flags |= RTF_MODIFIED;
1338         rt->rt6i_pmtu = mtu;
1339         rt6_update_expires(rt, net->ipv6.sysctl.ip6_rt_mtu_expires);
1340 }
1341
1342 static void __ip6_rt_update_pmtu(struct dst_entry *dst, const struct sock *sk,
1343                                  const struct ipv6hdr *iph, u32 mtu)
1344 {
1345         struct rt6_info *rt6 = (struct rt6_info *)dst;
1346
1347         if (rt6->rt6i_flags & RTF_LOCAL)
1348                 return;
1349
1350         dst_confirm(dst);
1351         mtu = max_t(u32, mtu, IPV6_MIN_MTU);
1352         if (mtu >= dst_mtu(dst))
1353                 return;
1354
1355         if (rt6->rt6i_flags & RTF_CACHE) {
1356                 rt6_do_update_pmtu(rt6, mtu);
1357         } else {
1358                 const struct in6_addr *daddr, *saddr;
1359                 struct rt6_info *nrt6;
1360
1361                 if (iph) {
1362                         daddr = &iph->daddr;
1363                         saddr = &iph->saddr;
1364                 } else if (sk) {
1365                         daddr = &sk->sk_v6_daddr;
1366                         saddr = &inet6_sk(sk)->saddr;
1367                 } else {
1368                         return;
1369                 }
1370                 nrt6 = ip6_rt_cache_alloc(rt6, daddr, saddr);
1371                 if (nrt6) {
1372                         rt6_do_update_pmtu(nrt6, mtu);
1373
1374                         /* ip6_ins_rt(nrt6) will bump the
1375                          * rt6->rt6i_node->fn_sernum
1376                          * which will fail the next rt6_check() and
1377                          * invalidate the sk->sk_dst_cache.
1378                          */
1379                         ip6_ins_rt(nrt6);
1380                 }
1381         }
1382 }
1383
1384 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
1385                                struct sk_buff *skb, u32 mtu)
1386 {
1387         __ip6_rt_update_pmtu(dst, sk, skb ? ipv6_hdr(skb) : NULL, mtu);
1388 }
1389
1390 void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
1391                      int oif, u32 mark)
1392 {
1393         const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1394         struct dst_entry *dst;
1395         struct flowi6 fl6;
1396
1397         memset(&fl6, 0, sizeof(fl6));
1398         fl6.flowi6_oif = oif;
1399         fl6.flowi6_mark = mark ? mark : IP6_REPLY_MARK(net, skb->mark);
1400         fl6.daddr = iph->daddr;
1401         fl6.saddr = iph->saddr;
1402         fl6.flowlabel = ip6_flowinfo(iph);
1403
1404         dst = ip6_route_output(net, NULL, &fl6);
1405         if (!dst->error)
1406                 __ip6_rt_update_pmtu(dst, NULL, iph, ntohl(mtu));
1407         dst_release(dst);
1408 }
1409 EXPORT_SYMBOL_GPL(ip6_update_pmtu);
1410
1411 void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
1412 {
1413         ip6_update_pmtu(skb, sock_net(sk), mtu,
1414                         sk->sk_bound_dev_if, sk->sk_mark);
1415 }
1416 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
1417
1418 /* Handle redirects */
1419 struct ip6rd_flowi {
1420         struct flowi6 fl6;
1421         struct in6_addr gateway;
1422 };
1423
1424 static struct rt6_info *__ip6_route_redirect(struct net *net,
1425                                              struct fib6_table *table,
1426                                              struct flowi6 *fl6,
1427                                              int flags)
1428 {
1429         struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
1430         struct rt6_info *rt;
1431         struct fib6_node *fn;
1432
1433         /* Get the "current" route for this destination and
1434          * check if the redirect has come from approriate router.
1435          *
1436          * RFC 4861 specifies that redirects should only be
1437          * accepted if they come from the nexthop to the target.
1438          * Due to the way the routes are chosen, this notion
1439          * is a bit fuzzy and one might need to check all possible
1440          * routes.
1441          */
1442
1443         read_lock_bh(&table->tb6_lock);
1444         fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1445 restart:
1446         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1447                 if (rt6_check_expired(rt))
1448                         continue;
1449                 if (rt->dst.error)
1450                         break;
1451                 if (!(rt->rt6i_flags & RTF_GATEWAY))
1452                         continue;
1453                 if (fl6->flowi6_oif != rt->dst.dev->ifindex)
1454                         continue;
1455                 if (!ipv6_addr_equal(&rdfl->gateway, &rt->rt6i_gateway))
1456                         continue;
1457                 break;
1458         }
1459
1460         if (!rt)
1461                 rt = net->ipv6.ip6_null_entry;
1462         else if (rt->dst.error) {
1463                 rt = net->ipv6.ip6_null_entry;
1464                 goto out;
1465         }
1466
1467         if (rt == net->ipv6.ip6_null_entry) {
1468                 fn = fib6_backtrack(fn, &fl6->saddr);
1469                 if (fn)
1470                         goto restart;
1471         }
1472
1473 out:
1474         dst_hold(&rt->dst);
1475
1476         read_unlock_bh(&table->tb6_lock);
1477
1478         return rt;
1479 };
1480
1481 static struct dst_entry *ip6_route_redirect(struct net *net,
1482                                         const struct flowi6 *fl6,
1483                                         const struct in6_addr *gateway)
1484 {
1485         int flags = RT6_LOOKUP_F_HAS_SADDR;
1486         struct ip6rd_flowi rdfl;
1487
1488         rdfl.fl6 = *fl6;
1489         rdfl.gateway = *gateway;
1490
1491         return fib6_rule_lookup(net, &rdfl.fl6,
1492                                 flags, __ip6_route_redirect);
1493 }
1494
1495 void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark)
1496 {
1497         const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1498         struct dst_entry *dst;
1499         struct flowi6 fl6;
1500
1501         memset(&fl6, 0, sizeof(fl6));
1502         fl6.flowi6_iif = LOOPBACK_IFINDEX;
1503         fl6.flowi6_oif = oif;
1504         fl6.flowi6_mark = mark;
1505         fl6.daddr = iph->daddr;
1506         fl6.saddr = iph->saddr;
1507         fl6.flowlabel = ip6_flowinfo(iph);
1508
1509         dst = ip6_route_redirect(net, &fl6, &ipv6_hdr(skb)->saddr);
1510         rt6_do_redirect(dst, NULL, skb);
1511         dst_release(dst);
1512 }
1513 EXPORT_SYMBOL_GPL(ip6_redirect);
1514
1515 void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif,
1516                             u32 mark)
1517 {
1518         const struct ipv6hdr *iph = ipv6_hdr(skb);
1519         const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
1520         struct dst_entry *dst;
1521         struct flowi6 fl6;
1522
1523         memset(&fl6, 0, sizeof(fl6));
1524         fl6.flowi6_iif = LOOPBACK_IFINDEX;
1525         fl6.flowi6_oif = oif;
1526         fl6.flowi6_mark = mark;
1527         fl6.daddr = msg->dest;
1528         fl6.saddr = iph->daddr;
1529
1530         dst = ip6_route_redirect(net, &fl6, &iph->saddr);
1531         rt6_do_redirect(dst, NULL, skb);
1532         dst_release(dst);
1533 }
1534
1535 void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
1536 {
1537         ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark);
1538 }
1539 EXPORT_SYMBOL_GPL(ip6_sk_redirect);
1540
1541 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1542 {
1543         struct net_device *dev = dst->dev;
1544         unsigned int mtu = dst_mtu(dst);
1545         struct net *net = dev_net(dev);
1546
1547         mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
1548
1549         if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
1550                 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1551
1552         /*
1553          * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1554          * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1555          * IPV6_MAXPLEN is also valid and means: "any MSS,
1556          * rely only on pmtu discovery"
1557          */
1558         if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
1559                 mtu = IPV6_MAXPLEN;
1560         return mtu;
1561 }
1562
1563 static unsigned int ip6_mtu(const struct dst_entry *dst)
1564 {
1565         const struct rt6_info *rt = (const struct rt6_info *)dst;
1566         unsigned int mtu = rt->rt6i_pmtu;
1567         struct inet6_dev *idev;
1568
1569         if (mtu)
1570                 goto out;
1571
1572         mtu = dst_metric_raw(dst, RTAX_MTU);
1573         if (mtu)
1574                 goto out;
1575
1576         mtu = IPV6_MIN_MTU;
1577
1578         rcu_read_lock();
1579         idev = __in6_dev_get(dst->dev);
1580         if (idev)
1581                 mtu = idev->cnf.mtu6;
1582         rcu_read_unlock();
1583
1584 out:
1585         return min_t(unsigned int, mtu, IP6_MAX_MTU);
1586 }
1587
1588 static struct dst_entry *icmp6_dst_gc_list;
1589 static DEFINE_SPINLOCK(icmp6_dst_lock);
1590
1591 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
1592                                   struct flowi6 *fl6)
1593 {
1594         struct dst_entry *dst;
1595         struct rt6_info *rt;
1596         struct inet6_dev *idev = in6_dev_get(dev);
1597         struct net *net = dev_net(dev);
1598
1599         if (unlikely(!idev))
1600                 return ERR_PTR(-ENODEV);
1601
1602         rt = ip6_dst_alloc(net, dev, 0);
1603         if (unlikely(!rt)) {
1604                 in6_dev_put(idev);
1605                 dst = ERR_PTR(-ENOMEM);
1606                 goto out;
1607         }
1608
1609         rt->dst.flags |= DST_HOST;
1610         rt->dst.output  = ip6_output;
1611         atomic_set(&rt->dst.__refcnt, 1);
1612         rt->rt6i_gateway  = fl6->daddr;
1613         rt->rt6i_dst.addr = fl6->daddr;
1614         rt->rt6i_dst.plen = 128;
1615         rt->rt6i_idev     = idev;
1616         dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
1617
1618         spin_lock_bh(&icmp6_dst_lock);
1619         rt->dst.next = icmp6_dst_gc_list;
1620         icmp6_dst_gc_list = &rt->dst;
1621         spin_unlock_bh(&icmp6_dst_lock);
1622
1623         fib6_force_start_gc(net);
1624
1625         dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
1626
1627 out:
1628         return dst;
1629 }
1630
1631 int icmp6_dst_gc(void)
1632 {
1633         struct dst_entry *dst, **pprev;
1634         int more = 0;
1635
1636         spin_lock_bh(&icmp6_dst_lock);
1637         pprev = &icmp6_dst_gc_list;
1638
1639         while ((dst = *pprev) != NULL) {
1640                 if (!atomic_read(&dst->__refcnt)) {
1641                         *pprev = dst->next;
1642                         dst_free(dst);
1643                 } else {
1644                         pprev = &dst->next;
1645                         ++more;
1646                 }
1647         }
1648
1649         spin_unlock_bh(&icmp6_dst_lock);
1650
1651         return more;
1652 }
1653
1654 static void icmp6_clean_all(int (*func)(struct rt6_info *rt, void *arg),
1655                             void *arg)
1656 {
1657         struct dst_entry *dst, **pprev;
1658
1659         spin_lock_bh(&icmp6_dst_lock);
1660         pprev = &icmp6_dst_gc_list;
1661         while ((dst = *pprev) != NULL) {
1662                 struct rt6_info *rt = (struct rt6_info *) dst;
1663                 if (func(rt, arg)) {
1664                         *pprev = dst->next;
1665                         dst_free(dst);
1666                 } else {
1667                         pprev = &dst->next;
1668                 }
1669         }
1670         spin_unlock_bh(&icmp6_dst_lock);
1671 }
1672
1673 static int ip6_dst_gc(struct dst_ops *ops)
1674 {
1675         struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
1676         int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1677         int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1678         int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1679         int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1680         unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1681         int entries;
1682
1683         entries = dst_entries_get_fast(ops);
1684         if (time_after(rt_last_gc + rt_min_interval, jiffies) &&
1685             entries <= rt_max_size)
1686                 goto out;
1687
1688         net->ipv6.ip6_rt_gc_expire++;
1689         fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net, true);
1690         entries = dst_entries_get_slow(ops);
1691         if (entries < ops->gc_thresh)
1692                 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1693 out:
1694         net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1695         return entries > rt_max_size;
1696 }
1697
1698 static int ip6_convert_metrics(struct mx6_config *mxc,
1699                                const struct fib6_config *cfg)
1700 {
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
1715                 if (type) {
1716                         u32 val;
1717
1718                         if (unlikely(type > RTAX_MAX))
1719                                 goto err;
1720                         if (type == RTAX_CC_ALGO) {
1721                                 char tmp[TCP_CA_NAME_MAX];
1722
1723                                 nla_strlcpy(tmp, nla, sizeof(tmp));
1724                                 val = tcp_ca_get_key_by_name(tmp);
1725                                 if (val == TCP_CA_UNSPEC)
1726                                         goto err;
1727                         } else {
1728                                 val = nla_get_u32(nla);
1729                         }
1730
1731                         mp[type - 1] = val;
1732                         __set_bit(type - 1, mxc->mx_valid);
1733                 }
1734         }
1735
1736         mxc->mx = mp;
1737
1738         return 0;
1739  err:
1740         kfree(mp);
1741         return -EINVAL;
1742 }
1743
1744 int ip6_route_add(struct fib6_config *cfg)
1745 {
1746         int err;
1747         struct net *net = cfg->fc_nlinfo.nl_net;
1748         struct rt6_info *rt = NULL;
1749         struct net_device *dev = NULL;
1750         struct inet6_dev *idev = NULL;
1751         struct fib6_table *table;
1752         struct mx6_config mxc = { .mx = NULL, };
1753         int addr_type;
1754
1755         if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1756                 return -EINVAL;
1757 #ifndef CONFIG_IPV6_SUBTREES
1758         if (cfg->fc_src_len)
1759                 return -EINVAL;
1760 #endif
1761         if (cfg->fc_ifindex) {
1762                 err = -ENODEV;
1763                 dev = dev_get_by_index(net, cfg->fc_ifindex);
1764                 if (!dev)
1765                         goto out;
1766                 idev = in6_dev_get(dev);
1767                 if (!idev)
1768                         goto out;
1769         }
1770
1771         if (cfg->fc_metric == 0)
1772                 cfg->fc_metric = IP6_RT_PRIO_USER;
1773
1774         err = -ENOBUFS;
1775         if (cfg->fc_nlinfo.nlh &&
1776             !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
1777                 table = fib6_get_table(net, cfg->fc_table);
1778                 if (!table) {
1779                         pr_warn("NLM_F_CREATE should be specified when creating new route\n");
1780                         table = fib6_new_table(net, cfg->fc_table);
1781                 }
1782         } else {
1783                 table = fib6_new_table(net, cfg->fc_table);
1784         }
1785
1786         if (!table)
1787                 goto out;
1788
1789         rt = ip6_dst_alloc(net, NULL,
1790                            (cfg->fc_flags & RTF_ADDRCONF) ? 0 : DST_NOCOUNT);
1791
1792         if (!rt) {
1793                 err = -ENOMEM;
1794                 goto out;
1795         }
1796
1797         if (cfg->fc_flags & RTF_EXPIRES)
1798                 rt6_set_expires(rt, jiffies +
1799                                 clock_t_to_jiffies(cfg->fc_expires));
1800         else
1801                 rt6_clean_expires(rt);
1802
1803         if (cfg->fc_protocol == RTPROT_UNSPEC)
1804                 cfg->fc_protocol = RTPROT_BOOT;
1805         rt->rt6i_protocol = cfg->fc_protocol;
1806
1807         addr_type = ipv6_addr_type(&cfg->fc_dst);
1808
1809         if (addr_type & IPV6_ADDR_MULTICAST)
1810                 rt->dst.input = ip6_mc_input;
1811         else if (cfg->fc_flags & RTF_LOCAL)
1812                 rt->dst.input = ip6_input;
1813         else
1814                 rt->dst.input = ip6_forward;
1815
1816         rt->dst.output = ip6_output;
1817
1818         if (cfg->fc_encap) {
1819                 struct lwtunnel_state *lwtstate;
1820
1821                 err = lwtunnel_build_state(dev, cfg->fc_encap_type,
1822                                            cfg->fc_encap, &lwtstate);
1823                 if (err)
1824                         goto out;
1825                 rt->dst.lwtstate = lwtstate_get(lwtstate);
1826                 if (lwtunnel_output_redirect(rt->dst.lwtstate)) {
1827                         rt->dst.lwtstate->orig_output = rt->dst.output;
1828                         rt->dst.output = lwtunnel_output;
1829                 }
1830                 if (lwtunnel_input_redirect(rt->dst.lwtstate)) {
1831                         rt->dst.lwtstate->orig_input = rt->dst.input;
1832                         rt->dst.input = lwtunnel_input;
1833                 }
1834         }
1835
1836         ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1837         rt->rt6i_dst.plen = cfg->fc_dst_len;
1838         if (rt->rt6i_dst.plen == 128)
1839                 rt->dst.flags |= DST_HOST;
1840
1841 #ifdef CONFIG_IPV6_SUBTREES
1842         ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1843         rt->rt6i_src.plen = cfg->fc_src_len;
1844 #endif
1845
1846         rt->rt6i_metric = cfg->fc_metric;
1847
1848         /* We cannot add true routes via loopback here,
1849            they would result in kernel looping; promote them to reject routes
1850          */
1851         if ((cfg->fc_flags & RTF_REJECT) ||
1852             (dev && (dev->flags & IFF_LOOPBACK) &&
1853              !(addr_type & IPV6_ADDR_LOOPBACK) &&
1854              !(cfg->fc_flags & RTF_LOCAL))) {
1855                 /* hold loopback dev/idev if we haven't done so. */
1856                 if (dev != net->loopback_dev) {
1857                         if (dev) {
1858                                 dev_put(dev);
1859                                 in6_dev_put(idev);
1860                         }
1861                         dev = net->loopback_dev;
1862                         dev_hold(dev);
1863                         idev = in6_dev_get(dev);
1864                         if (!idev) {
1865                                 err = -ENODEV;
1866                                 goto out;
1867                         }
1868                 }
1869                 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1870                 switch (cfg->fc_type) {
1871                 case RTN_BLACKHOLE:
1872                         rt->dst.error = -EINVAL;
1873                         rt->dst.output = dst_discard_sk;
1874                         rt->dst.input = dst_discard;
1875                         break;
1876                 case RTN_PROHIBIT:
1877                         rt->dst.error = -EACCES;
1878                         rt->dst.output = ip6_pkt_prohibit_out;
1879                         rt->dst.input = ip6_pkt_prohibit;
1880                         break;
1881                 case RTN_THROW:
1882                 default:
1883                         rt->dst.error = (cfg->fc_type == RTN_THROW) ? -EAGAIN
1884                                         : -ENETUNREACH;
1885                         rt->dst.output = ip6_pkt_discard_out;
1886                         rt->dst.input = ip6_pkt_discard;
1887                         break;
1888                 }
1889                 goto install_route;
1890         }
1891
1892         if (cfg->fc_flags & RTF_GATEWAY) {
1893                 const struct in6_addr *gw_addr;
1894                 int gwa_type;
1895
1896                 gw_addr = &cfg->fc_gateway;
1897                 gwa_type = ipv6_addr_type(gw_addr);
1898
1899                 /* if gw_addr is local we will fail to detect this in case
1900                  * address is still TENTATIVE (DAD in progress). rt6_lookup()
1901                  * will return already-added prefix route via interface that
1902                  * prefix route was assigned to, which might be non-loopback.
1903                  */
1904                 err = -EINVAL;
1905                 if (ipv6_chk_addr_and_flags(net, gw_addr,
1906                                             gwa_type & IPV6_ADDR_LINKLOCAL ?
1907                                             dev : NULL, 0, 0))
1908                         goto out;
1909
1910                 rt->rt6i_gateway = *gw_addr;
1911
1912                 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1913                         struct rt6_info *grt;
1914
1915                         /* IPv6 strictly inhibits using not link-local
1916                            addresses as nexthop address.
1917                            Otherwise, router will not able to send redirects.
1918                            It is very good, but in some (rare!) circumstances
1919                            (SIT, PtP, NBMA NOARP links) it is handy to allow
1920                            some exceptions. --ANK
1921                          */
1922                         if (!(gwa_type & IPV6_ADDR_UNICAST))
1923                                 goto out;
1924
1925                         grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, 1);
1926
1927                         err = -EHOSTUNREACH;
1928                         if (!grt)
1929                                 goto out;
1930                         if (dev) {
1931                                 if (dev != grt->dst.dev) {
1932                                         ip6_rt_put(grt);
1933                                         goto out;
1934                                 }
1935                         } else {
1936                                 dev = grt->dst.dev;
1937                                 idev = grt->rt6i_idev;
1938                                 dev_hold(dev);
1939                                 in6_dev_hold(grt->rt6i_idev);
1940                         }
1941                         if (!(grt->rt6i_flags & RTF_GATEWAY))
1942                                 err = 0;
1943                         ip6_rt_put(grt);
1944
1945                         if (err)
1946                                 goto out;
1947                 }
1948                 err = -EINVAL;
1949                 if (!dev || (dev->flags & IFF_LOOPBACK))
1950                         goto out;
1951         }
1952
1953         err = -ENODEV;
1954         if (!dev)
1955                 goto out;
1956
1957         if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
1958                 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
1959                         err = -EINVAL;
1960                         goto out;
1961                 }
1962                 rt->rt6i_prefsrc.addr = cfg->fc_prefsrc;
1963                 rt->rt6i_prefsrc.plen = 128;
1964         } else
1965                 rt->rt6i_prefsrc.plen = 0;
1966
1967         rt->rt6i_flags = cfg->fc_flags;
1968
1969 install_route:
1970         rt->dst.dev = dev;
1971         rt->rt6i_idev = idev;
1972         rt->rt6i_table = table;
1973
1974         cfg->fc_nlinfo.nl_net = dev_net(dev);
1975
1976         err = ip6_convert_metrics(&mxc, cfg);
1977         if (err)
1978                 goto out;
1979
1980         err = __ip6_ins_rt(rt, &cfg->fc_nlinfo, &mxc);
1981
1982         kfree(mxc.mx);
1983         return err;
1984 out:
1985         if (dev)
1986                 dev_put(dev);
1987         if (idev)
1988                 in6_dev_put(idev);
1989         if (rt)
1990                 dst_free(&rt->dst);
1991         return err;
1992 }
1993
1994 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
1995 {
1996         int err;
1997         struct fib6_table *table;
1998         struct net *net = dev_net(rt->dst.dev);
1999
2000         if (rt == net->ipv6.ip6_null_entry) {
2001                 err = -ENOENT;
2002                 goto out;
2003         }
2004
2005         table = rt->rt6i_table;
2006         write_lock_bh(&table->tb6_lock);
2007         err = fib6_del(rt, info);
2008         write_unlock_bh(&table->tb6_lock);
2009
2010 out:
2011         ip6_rt_put(rt);
2012         return err;
2013 }
2014
2015 int ip6_del_rt(struct rt6_info *rt)
2016 {
2017         struct nl_info info = {
2018                 .nl_net = dev_net(rt->dst.dev),
2019         };
2020         return __ip6_del_rt(rt, &info);
2021 }
2022
2023 static int ip6_route_del(struct fib6_config *cfg)
2024 {
2025         struct fib6_table *table;
2026         struct fib6_node *fn;
2027         struct rt6_info *rt;
2028         int err = -ESRCH;
2029
2030         table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
2031         if (!table)
2032                 return err;
2033
2034         read_lock_bh(&table->tb6_lock);
2035
2036         fn = fib6_locate(&table->tb6_root,
2037                          &cfg->fc_dst, cfg->fc_dst_len,
2038                          &cfg->fc_src, cfg->fc_src_len);
2039
2040         if (fn) {
2041                 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2042                         if ((rt->rt6i_flags & RTF_CACHE) &&
2043                             !(cfg->fc_flags & RTF_CACHE))
2044                                 continue;
2045                         if (cfg->fc_ifindex &&
2046                             (!rt->dst.dev ||
2047                              rt->dst.dev->ifindex != cfg->fc_ifindex))
2048                                 continue;
2049                         if (cfg->fc_flags & RTF_GATEWAY &&
2050                             !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
2051                                 continue;
2052                         if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
2053                                 continue;
2054                         dst_hold(&rt->dst);
2055                         read_unlock_bh(&table->tb6_lock);
2056
2057                         return __ip6_del_rt(rt, &cfg->fc_nlinfo);
2058                 }
2059         }
2060         read_unlock_bh(&table->tb6_lock);
2061
2062         return err;
2063 }
2064
2065 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
2066 {
2067         struct net *net = dev_net(skb->dev);
2068         struct netevent_redirect netevent;
2069         struct rt6_info *rt, *nrt = NULL;
2070         struct ndisc_options ndopts;
2071         struct inet6_dev *in6_dev;
2072         struct neighbour *neigh;
2073         struct rd_msg *msg;
2074         int optlen, on_link;
2075         u8 *lladdr;
2076
2077         optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
2078         optlen -= sizeof(*msg);
2079
2080         if (optlen < 0) {
2081                 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
2082                 return;
2083         }
2084
2085         msg = (struct rd_msg *)icmp6_hdr(skb);
2086
2087         if (ipv6_addr_is_multicast(&msg->dest)) {
2088                 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
2089                 return;
2090         }
2091
2092         on_link = 0;
2093         if (ipv6_addr_equal(&msg->dest, &msg->target)) {
2094                 on_link = 1;
2095         } else if (ipv6_addr_type(&msg->target) !=
2096                    (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
2097                 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
2098                 return;
2099         }
2100
2101         in6_dev = __in6_dev_get(skb->dev);
2102         if (!in6_dev)
2103                 return;
2104         if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
2105                 return;
2106
2107         /* RFC2461 8.1:
2108          *      The IP source address of the Redirect MUST be the same as the current
2109          *      first-hop router for the specified ICMP Destination Address.
2110          */
2111
2112         if (!ndisc_parse_options(msg->opt, optlen, &ndopts)) {
2113                 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
2114                 return;
2115         }
2116
2117         lladdr = NULL;
2118         if (ndopts.nd_opts_tgt_lladdr) {
2119                 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
2120                                              skb->dev);
2121                 if (!lladdr) {
2122                         net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
2123                         return;
2124                 }
2125         }
2126
2127         rt = (struct rt6_info *) dst;
2128         if (rt == net->ipv6.ip6_null_entry) {
2129                 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
2130                 return;
2131         }
2132
2133         /* Redirect received -> path was valid.
2134          * Look, redirects are sent only in response to data packets,
2135          * so that this nexthop apparently is reachable. --ANK
2136          */
2137         dst_confirm(&rt->dst);
2138
2139         neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
2140         if (!neigh)
2141                 return;
2142
2143         /*
2144          *      We have finally decided to accept it.
2145          */
2146
2147         neigh_update(neigh, lladdr, NUD_STALE,
2148                      NEIGH_UPDATE_F_WEAK_OVERRIDE|
2149                      NEIGH_UPDATE_F_OVERRIDE|
2150                      (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
2151                                      NEIGH_UPDATE_F_ISROUTER))
2152                      );
2153
2154         nrt = ip6_rt_cache_alloc(rt, &msg->dest, NULL);
2155         if (!nrt)
2156                 goto out;
2157
2158         nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
2159         if (on_link)
2160                 nrt->rt6i_flags &= ~RTF_GATEWAY;
2161
2162         nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
2163
2164         if (ip6_ins_rt(nrt))
2165                 goto out;
2166
2167         netevent.old = &rt->dst;
2168         netevent.new = &nrt->dst;
2169         netevent.daddr = &msg->dest;
2170         netevent.neigh = neigh;
2171         call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
2172
2173         if (rt->rt6i_flags & RTF_CACHE) {
2174                 rt = (struct rt6_info *) dst_clone(&rt->dst);
2175                 ip6_del_rt(rt);
2176         }
2177
2178 out:
2179         neigh_release(neigh);
2180 }
2181
2182 /*
2183  *      Misc support functions
2184  */
2185
2186 static void rt6_set_from(struct rt6_info *rt, struct rt6_info *from)
2187 {
2188         BUG_ON(from->dst.from);
2189
2190         rt->rt6i_flags &= ~RTF_EXPIRES;
2191         dst_hold(&from->dst);
2192         rt->dst.from = &from->dst;
2193         dst_init_metrics(&rt->dst, dst_metrics_ptr(&from->dst), true);
2194 }
2195
2196 static void ip6_rt_copy_init(struct rt6_info *rt, struct rt6_info *ort)
2197 {
2198         rt->dst.input = ort->dst.input;
2199         rt->dst.output = ort->dst.output;
2200         rt->rt6i_dst = ort->rt6i_dst;
2201         rt->dst.error = ort->dst.error;
2202         rt->rt6i_idev = ort->rt6i_idev;
2203         if (rt->rt6i_idev)
2204                 in6_dev_hold(rt->rt6i_idev);
2205         rt->dst.lastuse = jiffies;
2206         rt->rt6i_gateway = ort->rt6i_gateway;
2207         rt->rt6i_flags = ort->rt6i_flags;
2208         rt6_set_from(rt, ort);
2209         rt->rt6i_metric = ort->rt6i_metric;
2210 #ifdef CONFIG_IPV6_SUBTREES
2211         rt->rt6i_src = ort->rt6i_src;
2212 #endif
2213         rt->rt6i_prefsrc = ort->rt6i_prefsrc;
2214         rt->rt6i_table = ort->rt6i_table;
2215         rt->dst.lwtstate = lwtstate_get(ort->dst.lwtstate);
2216 }
2217
2218 #ifdef CONFIG_IPV6_ROUTE_INFO
2219 static struct rt6_info *rt6_get_route_info(struct net *net,
2220                                            const struct in6_addr *prefix, int prefixlen,
2221                                            const struct in6_addr *gwaddr, int ifindex)
2222 {
2223         struct fib6_node *fn;
2224         struct rt6_info *rt = NULL;
2225         struct fib6_table *table;
2226
2227         table = fib6_get_table(net, RT6_TABLE_INFO);
2228         if (!table)
2229                 return NULL;
2230
2231         read_lock_bh(&table->tb6_lock);
2232         fn = fib6_locate(&table->tb6_root, prefix, prefixlen, NULL, 0);
2233         if (!fn)
2234                 goto out;
2235
2236         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2237                 if (rt->dst.dev->ifindex != ifindex)
2238                         continue;
2239                 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
2240                         continue;
2241                 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
2242                         continue;
2243                 dst_hold(&rt->dst);
2244                 break;
2245         }
2246 out:
2247         read_unlock_bh(&table->tb6_lock);
2248         return rt;
2249 }
2250
2251 static struct rt6_info *rt6_add_route_info(struct net *net,
2252                                            const struct in6_addr *prefix, int prefixlen,
2253                                            const struct in6_addr *gwaddr, int ifindex,
2254                                            unsigned int pref)
2255 {
2256         struct fib6_config cfg = {
2257                 .fc_table       = RT6_TABLE_INFO,
2258                 .fc_metric      = IP6_RT_PRIO_USER,
2259                 .fc_ifindex     = ifindex,
2260                 .fc_dst_len     = prefixlen,
2261                 .fc_flags       = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
2262                                   RTF_UP | RTF_PREF(pref),
2263                 .fc_nlinfo.portid = 0,
2264                 .fc_nlinfo.nlh = NULL,
2265                 .fc_nlinfo.nl_net = net,
2266         };
2267
2268         cfg.fc_dst = *prefix;
2269         cfg.fc_gateway = *gwaddr;
2270
2271         /* We should treat it as a default route if prefix length is 0. */
2272         if (!prefixlen)
2273                 cfg.fc_flags |= RTF_DEFAULT;
2274
2275         ip6_route_add(&cfg);
2276
2277         return rt6_get_route_info(net, prefix, prefixlen, gwaddr, ifindex);
2278 }
2279 #endif
2280
2281 struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
2282 {
2283         struct rt6_info *rt;
2284         struct fib6_table *table;
2285
2286         table = fib6_get_table(dev_net(dev), RT6_TABLE_DFLT);
2287         if (!table)
2288                 return NULL;
2289
2290         read_lock_bh(&table->tb6_lock);
2291         for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2292                 if (dev == rt->dst.dev &&
2293                     ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
2294                     ipv6_addr_equal(&rt->rt6i_gateway, addr))
2295                         break;
2296         }
2297         if (rt)
2298                 dst_hold(&rt->dst);
2299         read_unlock_bh(&table->tb6_lock);
2300         return rt;
2301 }
2302
2303 struct rt6_info *rt6_add_dflt_router(const struct in6_addr *gwaddr,
2304                                      struct net_device *dev,
2305                                      unsigned int pref)
2306 {
2307         struct fib6_config cfg = {
2308                 .fc_table       = RT6_TABLE_DFLT,
2309                 .fc_metric      = IP6_RT_PRIO_USER,
2310                 .fc_ifindex     = dev->ifindex,
2311                 .fc_flags       = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
2312                                   RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
2313                 .fc_nlinfo.portid = 0,
2314                 .fc_nlinfo.nlh = NULL,
2315                 .fc_nlinfo.nl_net = dev_net(dev),
2316         };
2317
2318         cfg.fc_gateway = *gwaddr;
2319
2320         ip6_route_add(&cfg);
2321
2322         return rt6_get_dflt_router(gwaddr, dev);
2323 }
2324
2325 void rt6_purge_dflt_routers(struct net *net)
2326 {
2327         struct rt6_info *rt;
2328         struct fib6_table *table;
2329
2330         /* NOTE: Keep consistent with rt6_get_dflt_router */
2331         table = fib6_get_table(net, RT6_TABLE_DFLT);
2332         if (!table)
2333                 return;
2334
2335 restart:
2336         read_lock_bh(&table->tb6_lock);
2337         for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2338                 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
2339                     (!rt->rt6i_idev || rt->rt6i_idev->cnf.accept_ra != 2)) {
2340                         dst_hold(&rt->dst);
2341                         read_unlock_bh(&table->tb6_lock);
2342                         ip6_del_rt(rt);
2343                         goto restart;
2344                 }
2345         }
2346         read_unlock_bh(&table->tb6_lock);
2347 }
2348
2349 static void rtmsg_to_fib6_config(struct net *net,
2350                                  struct in6_rtmsg *rtmsg,
2351                                  struct fib6_config *cfg)
2352 {
2353         memset(cfg, 0, sizeof(*cfg));
2354
2355         cfg->fc_table = RT6_TABLE_MAIN;
2356         cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
2357         cfg->fc_metric = rtmsg->rtmsg_metric;
2358         cfg->fc_expires = rtmsg->rtmsg_info;
2359         cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
2360         cfg->fc_src_len = rtmsg->rtmsg_src_len;
2361         cfg->fc_flags = rtmsg->rtmsg_flags;
2362
2363         cfg->fc_nlinfo.nl_net = net;
2364
2365         cfg->fc_dst = rtmsg->rtmsg_dst;
2366         cfg->fc_src = rtmsg->rtmsg_src;
2367         cfg->fc_gateway = rtmsg->rtmsg_gateway;
2368 }
2369
2370 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
2371 {
2372         struct fib6_config cfg;
2373         struct in6_rtmsg rtmsg;
2374         int err;
2375
2376         switch (cmd) {
2377         case SIOCADDRT:         /* Add a route */
2378         case SIOCDELRT:         /* Delete a route */
2379                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2380                         return -EPERM;
2381                 err = copy_from_user(&rtmsg, arg,
2382                                      sizeof(struct in6_rtmsg));
2383                 if (err)
2384                         return -EFAULT;
2385
2386                 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
2387
2388                 rtnl_lock();
2389                 switch (cmd) {
2390                 case SIOCADDRT:
2391                         err = ip6_route_add(&cfg);
2392                         break;
2393                 case SIOCDELRT:
2394                         err = ip6_route_del(&cfg);
2395                         break;
2396                 default:
2397                         err = -EINVAL;
2398                 }
2399                 rtnl_unlock();
2400
2401                 return err;
2402         }
2403
2404         return -EINVAL;
2405 }
2406
2407 /*
2408  *      Drop the packet on the floor
2409  */
2410
2411 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
2412 {
2413         int type;
2414         struct dst_entry *dst = skb_dst(skb);
2415         switch (ipstats_mib_noroutes) {
2416         case IPSTATS_MIB_INNOROUTES:
2417                 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
2418                 if (type == IPV6_ADDR_ANY) {
2419                         IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2420                                       IPSTATS_MIB_INADDRERRORS);
2421                         break;
2422                 }
2423                 /* FALLTHROUGH */
2424         case IPSTATS_MIB_OUTNOROUTES:
2425                 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2426                               ipstats_mib_noroutes);
2427                 break;
2428         }
2429         icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
2430         kfree_skb(skb);
2431         return 0;
2432 }
2433
2434 static int ip6_pkt_discard(struct sk_buff *skb)
2435 {
2436         return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
2437 }
2438
2439 static int ip6_pkt_discard_out(struct sock *sk, struct sk_buff *skb)
2440 {
2441         skb->dev = skb_dst(skb)->dev;
2442         return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
2443 }
2444
2445 static int ip6_pkt_prohibit(struct sk_buff *skb)
2446 {
2447         return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
2448 }
2449
2450 static int ip6_pkt_prohibit_out(struct sock *sk, struct sk_buff *skb)
2451 {
2452         skb->dev = skb_dst(skb)->dev;
2453         return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
2454 }
2455
2456 /*
2457  *      Allocate a dst for local (unicast / anycast) address.
2458  */
2459
2460 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
2461                                     const struct in6_addr *addr,
2462                                     bool anycast)
2463 {
2464         struct net *net = dev_net(idev->dev);
2465         struct rt6_info *rt = ip6_dst_alloc(net, net->loopback_dev,
2466                                             DST_NOCOUNT);
2467         if (!rt)
2468                 return ERR_PTR(-ENOMEM);
2469
2470         in6_dev_hold(idev);
2471
2472         rt->dst.flags |= DST_HOST;
2473         rt->dst.input = ip6_input;
2474         rt->dst.output = ip6_output;
2475         rt->rt6i_idev = idev;
2476
2477         rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
2478         if (anycast)
2479                 rt->rt6i_flags |= RTF_ANYCAST;
2480         else
2481                 rt->rt6i_flags |= RTF_LOCAL;
2482
2483         rt->rt6i_gateway  = *addr;
2484         rt->rt6i_dst.addr = *addr;
2485         rt->rt6i_dst.plen = 128;
2486         rt->rt6i_table = fib6_get_table(net, RT6_TABLE_LOCAL);
2487
2488         atomic_set(&rt->dst.__refcnt, 1);
2489
2490         return rt;
2491 }
2492
2493 int ip6_route_get_saddr(struct net *net,
2494                         struct rt6_info *rt,
2495                         const struct in6_addr *daddr,
2496                         unsigned int prefs,
2497                         struct in6_addr *saddr)
2498 {
2499         struct inet6_dev *idev =
2500                 rt ? ip6_dst_idev((struct dst_entry *)rt) : NULL;
2501         int err = 0;
2502         if (rt && rt->rt6i_prefsrc.plen)
2503                 *saddr = rt->rt6i_prefsrc.addr;
2504         else
2505                 err = ipv6_dev_get_saddr(net, idev ? idev->dev : NULL,
2506                                          daddr, prefs, saddr);
2507         return err;
2508 }
2509
2510 /* remove deleted ip from prefsrc entries */
2511 struct arg_dev_net_ip {
2512         struct net_device *dev;
2513         struct net *net;
2514         struct in6_addr *addr;
2515 };
2516
2517 static int fib6_remove_prefsrc(struct rt6_info *rt, void *arg)
2518 {
2519         struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
2520         struct net *net = ((struct arg_dev_net_ip *)arg)->net;
2521         struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
2522
2523         if (((void *)rt->dst.dev == dev || !dev) &&
2524             rt != net->ipv6.ip6_null_entry &&
2525             ipv6_addr_equal(addr, &rt->rt6i_prefsrc.addr)) {
2526                 /* remove prefsrc entry */
2527                 rt->rt6i_prefsrc.plen = 0;
2528         }
2529         return 0;
2530 }
2531
2532 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
2533 {
2534         struct net *net = dev_net(ifp->idev->dev);
2535         struct arg_dev_net_ip adni = {
2536                 .dev = ifp->idev->dev,
2537                 .net = net,
2538                 .addr = &ifp->addr,
2539         };
2540         fib6_clean_all(net, fib6_remove_prefsrc, &adni);
2541 }
2542
2543 #define RTF_RA_ROUTER           (RTF_ADDRCONF | RTF_DEFAULT | RTF_GATEWAY)
2544 #define RTF_CACHE_GATEWAY       (RTF_GATEWAY | RTF_CACHE)
2545
2546 /* Remove routers and update dst entries when gateway turn into host. */
2547 static int fib6_clean_tohost(struct rt6_info *rt, void *arg)
2548 {
2549         struct in6_addr *gateway = (struct in6_addr *)arg;
2550
2551         if ((((rt->rt6i_flags & RTF_RA_ROUTER) == RTF_RA_ROUTER) ||
2552              ((rt->rt6i_flags & RTF_CACHE_GATEWAY) == RTF_CACHE_GATEWAY)) &&
2553              ipv6_addr_equal(gateway, &rt->rt6i_gateway)) {
2554                 return -1;
2555         }
2556         return 0;
2557 }
2558
2559 void rt6_clean_tohost(struct net *net, struct in6_addr *gateway)
2560 {
2561         fib6_clean_all(net, fib6_clean_tohost, gateway);
2562 }
2563
2564 struct arg_dev_net {
2565         struct net_device *dev;
2566         struct net *net;
2567 };
2568
2569 static int fib6_ifdown(struct rt6_info *rt, void *arg)
2570 {
2571         const struct arg_dev_net *adn = arg;
2572         const struct net_device *dev = adn->dev;
2573
2574         if ((rt->dst.dev == dev || !dev) &&
2575             rt != adn->net->ipv6.ip6_null_entry)
2576                 return -1;
2577
2578         return 0;
2579 }
2580
2581 void rt6_ifdown(struct net *net, struct net_device *dev)
2582 {
2583         struct arg_dev_net adn = {
2584                 .dev = dev,
2585                 .net = net,
2586         };
2587
2588         fib6_clean_all(net, fib6_ifdown, &adn);
2589         icmp6_clean_all(fib6_ifdown, &adn);
2590         rt6_uncached_list_flush_dev(net, dev);
2591 }
2592
2593 struct rt6_mtu_change_arg {
2594         struct net_device *dev;
2595         unsigned int mtu;
2596 };
2597
2598 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2599 {
2600         struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2601         struct inet6_dev *idev;
2602
2603         /* In IPv6 pmtu discovery is not optional,
2604            so that RTAX_MTU lock cannot disable it.
2605            We still use this lock to block changes
2606            caused by addrconf/ndisc.
2607         */
2608
2609         idev = __in6_dev_get(arg->dev);
2610         if (!idev)
2611                 return 0;
2612
2613         /* For administrative MTU increase, there is no way to discover
2614            IPv6 PMTU increase, so PMTU increase should be updated here.
2615            Since RFC 1981 doesn't include administrative MTU increase
2616            update PMTU increase is a MUST. (i.e. jumbo frame)
2617          */
2618         /*
2619            If new MTU is less than route PMTU, this new MTU will be the
2620            lowest MTU in the path, update the route PMTU to reflect PMTU
2621            decreases; if new MTU is greater than route PMTU, and the
2622            old MTU is the lowest MTU in the path, update the route PMTU
2623            to reflect the increase. In this case if the other nodes' MTU
2624            also have the lowest MTU, TOO BIG MESSAGE will be lead to
2625            PMTU discouvery.
2626          */
2627         if (rt->dst.dev == arg->dev &&
2628             !dst_metric_locked(&rt->dst, RTAX_MTU)) {
2629                 if (rt->rt6i_flags & RTF_CACHE) {
2630                         /* For RTF_CACHE with rt6i_pmtu == 0
2631                          * (i.e. a redirected route),
2632                          * the metrics of its rt->dst.from has already
2633                          * been updated.
2634                          */
2635                         if (rt->rt6i_pmtu && rt->rt6i_pmtu > arg->mtu)
2636                                 rt->rt6i_pmtu = arg->mtu;
2637                 } else if (dst_mtu(&rt->dst) >= arg->mtu ||
2638                            (dst_mtu(&rt->dst) < arg->mtu &&
2639                             dst_mtu(&rt->dst) == idev->cnf.mtu6)) {
2640                         dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
2641                 }
2642         }
2643         return 0;
2644 }
2645
2646 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
2647 {
2648         struct rt6_mtu_change_arg arg = {
2649                 .dev = dev,
2650                 .mtu = mtu,
2651         };
2652
2653         fib6_clean_all(dev_net(dev), rt6_mtu_change_route, &arg);
2654 }
2655
2656 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2657         [RTA_GATEWAY]           = { .len = sizeof(struct in6_addr) },
2658         [RTA_OIF]               = { .type = NLA_U32 },
2659         [RTA_IIF]               = { .type = NLA_U32 },
2660         [RTA_PRIORITY]          = { .type = NLA_U32 },
2661         [RTA_METRICS]           = { .type = NLA_NESTED },
2662         [RTA_MULTIPATH]         = { .len = sizeof(struct rtnexthop) },
2663         [RTA_PREF]              = { .type = NLA_U8 },
2664         [RTA_ENCAP_TYPE]        = { .type = NLA_U16 },
2665         [RTA_ENCAP]             = { .type = NLA_NESTED },
2666 };
2667
2668 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2669                               struct fib6_config *cfg)
2670 {
2671         struct rtmsg *rtm;
2672         struct nlattr *tb[RTA_MAX+1];
2673         unsigned int pref;
2674         int err;
2675
2676         err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2677         if (err < 0)
2678                 goto errout;
2679
2680         err = -EINVAL;
2681         rtm = nlmsg_data(nlh);
2682         memset(cfg, 0, sizeof(*cfg));
2683
2684         cfg->fc_table = rtm->rtm_table;
2685         cfg->fc_dst_len = rtm->rtm_dst_len;
2686         cfg->fc_src_len = rtm->rtm_src_len;
2687         cfg->fc_flags = RTF_UP;
2688         cfg->fc_protocol = rtm->rtm_protocol;
2689         cfg->fc_type = rtm->rtm_type;
2690
2691         if (rtm->rtm_type == RTN_UNREACHABLE ||
2692             rtm->rtm_type == RTN_BLACKHOLE ||
2693             rtm->rtm_type == RTN_PROHIBIT ||
2694             rtm->rtm_type == RTN_THROW)
2695                 cfg->fc_flags |= RTF_REJECT;
2696
2697         if (rtm->rtm_type == RTN_LOCAL)
2698                 cfg->fc_flags |= RTF_LOCAL;
2699
2700         if (rtm->rtm_flags & RTM_F_CLONED)
2701                 cfg->fc_flags |= RTF_CACHE;
2702
2703         cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid;
2704         cfg->fc_nlinfo.nlh = nlh;
2705         cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2706
2707         if (tb[RTA_GATEWAY]) {
2708                 cfg->fc_gateway = nla_get_in6_addr(tb[RTA_GATEWAY]);
2709                 cfg->fc_flags |= RTF_GATEWAY;
2710         }
2711
2712         if (tb[RTA_DST]) {
2713                 int plen = (rtm->rtm_dst_len + 7) >> 3;
2714
2715                 if (nla_len(tb[RTA_DST]) < plen)
2716                         goto errout;
2717
2718                 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2719         }
2720
2721         if (tb[RTA_SRC]) {
2722                 int plen = (rtm->rtm_src_len + 7) >> 3;
2723
2724                 if (nla_len(tb[RTA_SRC]) < plen)
2725                         goto errout;
2726
2727                 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2728         }
2729
2730         if (tb[RTA_PREFSRC])
2731                 cfg->fc_prefsrc = nla_get_in6_addr(tb[RTA_PREFSRC]);
2732
2733         if (tb[RTA_OIF])
2734                 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2735
2736         if (tb[RTA_PRIORITY])
2737                 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2738
2739         if (tb[RTA_METRICS]) {
2740                 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2741                 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2742         }
2743
2744         if (tb[RTA_TABLE])
2745                 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2746
2747         if (tb[RTA_MULTIPATH]) {
2748                 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
2749                 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
2750         }
2751
2752         if (tb[RTA_PREF]) {
2753                 pref = nla_get_u8(tb[RTA_PREF]);
2754                 if (pref != ICMPV6_ROUTER_PREF_LOW &&
2755                     pref != ICMPV6_ROUTER_PREF_HIGH)
2756                         pref = ICMPV6_ROUTER_PREF_MEDIUM;
2757                 cfg->fc_flags |= RTF_PREF(pref);
2758         }
2759
2760         if (tb[RTA_ENCAP])
2761                 cfg->fc_encap = tb[RTA_ENCAP];
2762
2763         if (tb[RTA_ENCAP_TYPE])
2764                 cfg->fc_encap_type = nla_get_u16(tb[RTA_ENCAP_TYPE]);
2765
2766         err = 0;
2767 errout:
2768         return err;
2769 }
2770
2771 static int ip6_route_multipath(struct fib6_config *cfg, int add)
2772 {
2773         struct fib6_config r_cfg;
2774         struct rtnexthop *rtnh;
2775         int remaining;
2776         int attrlen;
2777         int err = 0, last_err = 0;
2778
2779         remaining = cfg->fc_mp_len;
2780 beginning:
2781         rtnh = (struct rtnexthop *)cfg->fc_mp;
2782
2783         /* Parse a Multipath Entry */
2784         while (rtnh_ok(rtnh, remaining)) {
2785                 memcpy(&r_cfg, cfg, sizeof(*cfg));
2786                 if (rtnh->rtnh_ifindex)
2787                         r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
2788
2789                 attrlen = rtnh_attrlen(rtnh);
2790                 if (attrlen > 0) {
2791                         struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
2792
2793                         nla = nla_find(attrs, attrlen, RTA_GATEWAY);
2794                         if (nla) {
2795                                 r_cfg.fc_gateway = nla_get_in6_addr(nla);
2796                                 r_cfg.fc_flags |= RTF_GATEWAY;
2797                         }
2798                         r_cfg.fc_encap = nla_find(attrs, attrlen, RTA_ENCAP);
2799                         nla = nla_find(attrs, attrlen, RTA_ENCAP_TYPE);
2800                         if (nla)
2801                                 r_cfg.fc_encap_type = nla_get_u16(nla);
2802                 }
2803                 err = add ? ip6_route_add(&r_cfg) : ip6_route_del(&r_cfg);
2804                 if (err) {
2805                         last_err = err;
2806                         /* If we are trying to remove a route, do not stop the
2807                          * loop when ip6_route_del() fails (because next hop is
2808                          * already gone), we should try to remove all next hops.
2809                          */
2810                         if (add) {
2811                                 /* If add fails, we should try to delete all
2812                                  * next hops that have been already added.
2813                                  */
2814                                 add = 0;
2815                                 remaining = cfg->fc_mp_len - remaining;
2816                                 goto beginning;
2817                         }
2818                 }
2819                 /* Because each route is added like a single route we remove
2820                  * these flags after the first nexthop: if there is a collision,
2821                  * we have already failed to add the first nexthop:
2822                  * fib6_add_rt2node() has rejected it; when replacing, old
2823                  * nexthops have been replaced by first new, the rest should
2824                  * be added to it.
2825                  */
2826                 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~(NLM_F_EXCL |
2827                                                      NLM_F_REPLACE);
2828                 rtnh = rtnh_next(rtnh, &remaining);
2829         }
2830
2831         return last_err;
2832 }
2833
2834 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh)
2835 {
2836         struct fib6_config cfg;
2837         int err;
2838
2839         err = rtm_to_fib6_config(skb, nlh, &cfg);
2840         if (err < 0)
2841                 return err;
2842
2843         if (cfg.fc_mp)
2844                 return ip6_route_multipath(&cfg, 0);
2845         else
2846                 return ip6_route_del(&cfg);
2847 }
2848
2849 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh)
2850 {
2851         struct fib6_config cfg;
2852         int err;
2853
2854         err = rtm_to_fib6_config(skb, nlh, &cfg);
2855         if (err < 0)
2856                 return err;
2857
2858         if (cfg.fc_mp)
2859                 return ip6_route_multipath(&cfg, 1);
2860         else
2861                 return ip6_route_add(&cfg);
2862 }
2863
2864 static inline size_t rt6_nlmsg_size(struct rt6_info *rt)
2865 {
2866         return NLMSG_ALIGN(sizeof(struct rtmsg))
2867                + nla_total_size(16) /* RTA_SRC */
2868                + nla_total_size(16) /* RTA_DST */
2869                + nla_total_size(16) /* RTA_GATEWAY */
2870                + nla_total_size(16) /* RTA_PREFSRC */
2871                + nla_total_size(4) /* RTA_TABLE */
2872                + nla_total_size(4) /* RTA_IIF */
2873                + nla_total_size(4) /* RTA_OIF */
2874                + nla_total_size(4) /* RTA_PRIORITY */
2875                + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
2876                + nla_total_size(sizeof(struct rta_cacheinfo))
2877                + nla_total_size(TCP_CA_NAME_MAX) /* RTAX_CC_ALGO */
2878                + nla_total_size(1) /* RTA_PREF */
2879                + lwtunnel_get_encap_size(rt->dst.lwtstate);
2880 }
2881
2882 static int rt6_fill_node(struct net *net,
2883                          struct sk_buff *skb, struct rt6_info *rt,
2884                          struct in6_addr *dst, struct in6_addr *src,
2885                          int iif, int type, u32 portid, u32 seq,
2886                          int prefix, int nowait, unsigned int flags)
2887 {
2888         u32 metrics[RTAX_MAX];
2889         struct rtmsg *rtm;
2890         struct nlmsghdr *nlh;
2891         long expires;
2892         u32 table;
2893
2894         if (prefix) {   /* user wants prefix routes only */
2895                 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
2896                         /* success since this is not a prefix route */
2897                         return 1;
2898                 }
2899         }
2900
2901         nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
2902         if (!nlh)
2903                 return -EMSGSIZE;
2904
2905         rtm = nlmsg_data(nlh);
2906         rtm->rtm_family = AF_INET6;
2907         rtm->rtm_dst_len = rt->rt6i_dst.plen;
2908         rtm->rtm_src_len = rt->rt6i_src.plen;
2909         rtm->rtm_tos = 0;
2910         if (rt->rt6i_table)
2911                 table = rt->rt6i_table->tb6_id;
2912         else
2913                 table = RT6_TABLE_UNSPEC;
2914         rtm->rtm_table = table;
2915         if (nla_put_u32(skb, RTA_TABLE, table))
2916                 goto nla_put_failure;
2917         if (rt->rt6i_flags & RTF_REJECT) {
2918                 switch (rt->dst.error) {
2919                 case -EINVAL:
2920                         rtm->rtm_type = RTN_BLACKHOLE;
2921                         break;
2922                 case -EACCES:
2923                         rtm->rtm_type = RTN_PROHIBIT;
2924                         break;
2925                 case -EAGAIN:
2926                         rtm->rtm_type = RTN_THROW;
2927                         break;
2928                 default:
2929                         rtm->rtm_type = RTN_UNREACHABLE;
2930                         break;
2931                 }
2932         }
2933         else if (rt->rt6i_flags & RTF_LOCAL)
2934                 rtm->rtm_type = RTN_LOCAL;
2935         else if (rt->dst.dev && (rt->dst.dev->flags & IFF_LOOPBACK))
2936                 rtm->rtm_type = RTN_LOCAL;
2937         else
2938                 rtm->rtm_type = RTN_UNICAST;
2939         rtm->rtm_flags = 0;
2940         if (!netif_carrier_ok(rt->dst.dev)) {
2941                 rtm->rtm_flags |= RTNH_F_LINKDOWN;
2942                 if (rt->rt6i_idev->cnf.ignore_routes_with_linkdown)
2943                         rtm->rtm_flags |= RTNH_F_DEAD;
2944         }
2945         rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2946         rtm->rtm_protocol = rt->rt6i_protocol;
2947         if (rt->rt6i_flags & RTF_DYNAMIC)
2948                 rtm->rtm_protocol = RTPROT_REDIRECT;
2949         else if (rt->rt6i_flags & RTF_ADDRCONF) {
2950                 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ROUTEINFO))
2951                         rtm->rtm_protocol = RTPROT_RA;
2952                 else
2953                         rtm->rtm_protocol = RTPROT_KERNEL;
2954         }
2955
2956         if (rt->rt6i_flags & RTF_CACHE)
2957                 rtm->rtm_flags |= RTM_F_CLONED;
2958
2959         if (dst) {
2960                 if (nla_put_in6_addr(skb, RTA_DST, dst))
2961                         goto nla_put_failure;
2962                 rtm->rtm_dst_len = 128;
2963         } else if (rtm->rtm_dst_len)
2964                 if (nla_put_in6_addr(skb, RTA_DST, &rt->rt6i_dst.addr))
2965                         goto nla_put_failure;
2966 #ifdef CONFIG_IPV6_SUBTREES
2967         if (src) {
2968                 if (nla_put_in6_addr(skb, RTA_SRC, src))
2969                         goto nla_put_failure;
2970                 rtm->rtm_src_len = 128;
2971         } else if (rtm->rtm_src_len &&
2972                    nla_put_in6_addr(skb, RTA_SRC, &rt->rt6i_src.addr))
2973                 goto nla_put_failure;
2974 #endif
2975         if (iif) {
2976 #ifdef CONFIG_IPV6_MROUTE
2977                 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
2978                         int err = ip6mr_get_route(net, skb, rtm, nowait);
2979                         if (err <= 0) {
2980                                 if (!nowait) {
2981                                         if (err == 0)
2982                                                 return 0;
2983                                         goto nla_put_failure;
2984                                 } else {
2985                                         if (err == -EMSGSIZE)
2986                                                 goto nla_put_failure;
2987                                 }
2988                         }
2989                 } else
2990 #endif
2991                         if (nla_put_u32(skb, RTA_IIF, iif))
2992                                 goto nla_put_failure;
2993         } else if (dst) {
2994                 struct in6_addr saddr_buf;
2995                 if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0 &&
2996                     nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
2997                         goto nla_put_failure;
2998         }
2999
3000         if (rt->rt6i_prefsrc.plen) {
3001                 struct in6_addr saddr_buf;
3002                 saddr_buf = rt->rt6i_prefsrc.addr;
3003                 if (nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
3004                         goto nla_put_failure;
3005         }
3006
3007         memcpy(metrics, dst_metrics_ptr(&rt->dst), sizeof(metrics));
3008         if (rt->rt6i_pmtu)
3009                 metrics[RTAX_MTU - 1] = rt->rt6i_pmtu;
3010         if (rtnetlink_put_metrics(skb, metrics) < 0)
3011                 goto nla_put_failure;
3012
3013         if (rt->rt6i_flags & RTF_GATEWAY) {
3014                 if (nla_put_in6_addr(skb, RTA_GATEWAY, &rt->rt6i_gateway) < 0)
3015                         goto nla_put_failure;
3016         }
3017
3018         if (rt->dst.dev &&
3019             nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex))
3020                 goto nla_put_failure;
3021         if (nla_put_u32(skb, RTA_PRIORITY, rt->rt6i_metric))
3022                 goto nla_put_failure;
3023
3024         expires = (rt->rt6i_flags & RTF_EXPIRES) ? rt->dst.expires - jiffies : 0;
3025
3026         if (rtnl_put_cacheinfo(skb, &rt->dst, 0, expires, rt->dst.error) < 0)
3027                 goto nla_put_failure;
3028
3029         if (nla_put_u8(skb, RTA_PREF, IPV6_EXTRACT_PREF(rt->rt6i_flags)))
3030                 goto nla_put_failure;
3031
3032         lwtunnel_fill_encap(skb, rt->dst.lwtstate);
3033
3034         nlmsg_end(skb, nlh);
3035         return 0;
3036
3037 nla_put_failure:
3038         nlmsg_cancel(skb, nlh);
3039         return -EMSGSIZE;
3040 }
3041
3042 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
3043 {
3044         struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
3045         int prefix;
3046
3047         if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
3048                 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
3049                 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
3050         } else
3051                 prefix = 0;
3052
3053         return rt6_fill_node(arg->net,
3054                      arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
3055                      NETLINK_CB(arg->cb->skb).portid, arg->cb->nlh->nlmsg_seq,
3056                      prefix, 0, NLM_F_MULTI);
3057 }
3058
3059 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh)
3060 {
3061         struct net *net = sock_net(in_skb->sk);
3062         struct nlattr *tb[RTA_MAX+1];
3063         struct rt6_info *rt;
3064         struct sk_buff *skb;
3065         struct rtmsg *rtm;
3066         struct flowi6 fl6;
3067         int err, iif = 0, oif = 0;
3068
3069         err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
3070         if (err < 0)
3071                 goto errout;
3072
3073         err = -EINVAL;
3074         memset(&fl6, 0, sizeof(fl6));
3075
3076         if (tb[RTA_SRC]) {
3077                 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
3078                         goto errout;
3079
3080                 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
3081         }
3082
3083         if (tb[RTA_DST]) {
3084                 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
3085                         goto errout;
3086
3087                 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
3088         }
3089
3090         if (tb[RTA_IIF])
3091                 iif = nla_get_u32(tb[RTA_IIF]);
3092
3093         if (tb[RTA_OIF])
3094                 oif = nla_get_u32(tb[RTA_OIF]);
3095
3096         if (tb[RTA_MARK])
3097                 fl6.flowi6_mark = nla_get_u32(tb[RTA_MARK]);
3098
3099         if (iif) {
3100                 struct net_device *dev;
3101                 int flags = 0;
3102
3103                 dev = __dev_get_by_index(net, iif);
3104                 if (!dev) {
3105                         err = -ENODEV;
3106                         goto errout;
3107                 }
3108
3109                 fl6.flowi6_iif = iif;
3110
3111                 if (!ipv6_addr_any(&fl6.saddr))
3112                         flags |= RT6_LOOKUP_F_HAS_SADDR;
3113
3114                 rt = (struct rt6_info *)ip6_route_input_lookup(net, dev, &fl6,
3115                                                                flags);
3116         } else {
3117                 fl6.flowi6_oif = oif;
3118
3119                 rt = (struct rt6_info *)ip6_route_output(net, NULL, &fl6);
3120         }
3121
3122         skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
3123         if (!skb) {
3124                 ip6_rt_put(rt);
3125                 err = -ENOBUFS;
3126                 goto errout;
3127         }
3128
3129         /* Reserve room for dummy headers, this skb can pass
3130            through good chunk of routing engine.
3131          */
3132         skb_reset_mac_header(skb);
3133         skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
3134
3135         skb_dst_set(skb, &rt->dst);
3136
3137         err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
3138                             RTM_NEWROUTE, NETLINK_CB(in_skb).portid,
3139                             nlh->nlmsg_seq, 0, 0, 0);
3140         if (err < 0) {
3141                 kfree_skb(skb);
3142                 goto errout;
3143         }
3144
3145         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
3146 errout:
3147         return err;
3148 }
3149
3150 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info)
3151 {
3152         struct sk_buff *skb;
3153         struct net *net = info->nl_net;
3154         u32 seq;
3155         int err;
3156
3157         err = -ENOBUFS;
3158         seq = info->nlh ? info->nlh->nlmsg_seq : 0;
3159
3160         skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
3161         if (!skb)
3162                 goto errout;
3163
3164         err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
3165                                 event, info->portid, seq, 0, 0, 0);
3166         if (err < 0) {
3167                 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
3168                 WARN_ON(err == -EMSGSIZE);
3169                 kfree_skb(skb);
3170                 goto errout;
3171         }
3172         rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
3173                     info->nlh, gfp_any());
3174         return;
3175 errout:
3176         if (err < 0)
3177                 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
3178 }
3179
3180 static int ip6_route_dev_notify(struct notifier_block *this,
3181                                 unsigned long event, void *ptr)
3182 {
3183         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3184         struct net *net = dev_net(dev);
3185
3186         if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
3187                 net->ipv6.ip6_null_entry->dst.dev = dev;
3188                 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
3189 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3190                 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
3191                 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
3192                 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
3193                 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
3194 #endif
3195         }
3196
3197         return NOTIFY_OK;
3198 }
3199
3200 /*
3201  *      /proc
3202  */
3203
3204 #ifdef CONFIG_PROC_FS
3205
3206 static const struct file_operations ipv6_route_proc_fops = {
3207         .owner          = THIS_MODULE,
3208         .open           = ipv6_route_open,
3209         .read           = seq_read,
3210         .llseek         = seq_lseek,
3211         .release        = seq_release_net,
3212 };
3213
3214 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
3215 {
3216         struct net *net = (struct net *)seq->private;
3217         seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
3218                    net->ipv6.rt6_stats->fib_nodes,
3219                    net->ipv6.rt6_stats->fib_route_nodes,
3220                    net->ipv6.rt6_stats->fib_rt_alloc,
3221                    net->ipv6.rt6_stats->fib_rt_entries,
3222                    net->ipv6.rt6_stats->fib_rt_cache,
3223                    dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
3224                    net->ipv6.rt6_stats->fib_discarded_routes);
3225
3226         return 0;
3227 }
3228
3229 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
3230 {
3231         return single_open_net(inode, file, rt6_stats_seq_show);
3232 }
3233
3234 static const struct file_operations rt6_stats_seq_fops = {
3235         .owner   = THIS_MODULE,
3236         .open    = rt6_stats_seq_open,
3237         .read    = seq_read,
3238         .llseek  = seq_lseek,
3239         .release = single_release_net,
3240 };
3241 #endif  /* CONFIG_PROC_FS */
3242
3243 #ifdef CONFIG_SYSCTL
3244
3245 static
3246 int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
3247                               void __user *buffer, size_t *lenp, loff_t *ppos)
3248 {
3249         struct net *net;
3250         int delay;
3251         if (!write)
3252                 return -EINVAL;
3253
3254         net = (struct net *)ctl->extra1;
3255         delay = net->ipv6.sysctl.flush_delay;
3256         proc_dointvec(ctl, write, buffer, lenp, ppos);
3257         fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
3258         return 0;
3259 }
3260
3261 struct ctl_table ipv6_route_table_template[] = {
3262         {
3263                 .procname       =       "flush",
3264                 .data           =       &init_net.ipv6.sysctl.flush_delay,
3265                 .maxlen         =       sizeof(int),
3266                 .mode           =       0200,
3267                 .proc_handler   =       ipv6_sysctl_rtcache_flush
3268         },
3269         {
3270                 .procname       =       "gc_thresh",
3271                 .data           =       &ip6_dst_ops_template.gc_thresh,
3272                 .maxlen         =       sizeof(int),
3273                 .mode           =       0644,
3274                 .proc_handler   =       proc_dointvec,
3275         },
3276         {
3277                 .procname       =       "max_size",
3278                 .data           =       &init_net.ipv6.sysctl.ip6_rt_max_size,
3279                 .maxlen         =       sizeof(int),
3280                 .mode           =       0644,
3281                 .proc_handler   =       proc_dointvec,
3282         },
3283         {
3284                 .procname       =       "gc_min_interval",
3285                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
3286                 .maxlen         =       sizeof(int),
3287                 .mode           =       0644,
3288                 .proc_handler   =       proc_dointvec_jiffies,
3289         },
3290         {
3291                 .procname       =       "gc_timeout",
3292                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
3293                 .maxlen         =       sizeof(int),
3294                 .mode           =       0644,
3295                 .proc_handler   =       proc_dointvec_jiffies,
3296         },
3297         {
3298                 .procname       =       "gc_interval",
3299                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_interval,
3300                 .maxlen         =       sizeof(int),
3301                 .mode           =       0644,
3302                 .proc_handler   =       proc_dointvec_jiffies,
3303         },
3304         {
3305                 .procname       =       "gc_elasticity",
3306                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
3307                 .maxlen         =       sizeof(int),
3308                 .mode           =       0644,
3309                 .proc_handler   =       proc_dointvec,
3310         },
3311         {
3312                 .procname       =       "mtu_expires",
3313                 .data           =       &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
3314                 .maxlen         =       sizeof(int),
3315                 .mode           =       0644,
3316                 .proc_handler   =       proc_dointvec_jiffies,
3317         },
3318         {
3319                 .procname       =       "min_adv_mss",
3320                 .data           =       &init_net.ipv6.sysctl.ip6_rt_min_advmss,
3321                 .maxlen         =       sizeof(int),
3322                 .mode           =       0644,
3323                 .proc_handler   =       proc_dointvec,
3324         },
3325         {
3326                 .procname       =       "gc_min_interval_ms",
3327                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
3328                 .maxlen         =       sizeof(int),
3329                 .mode           =       0644,
3330                 .proc_handler   =       proc_dointvec_ms_jiffies,
3331         },
3332         { }
3333 };
3334
3335 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
3336 {
3337         struct ctl_table *table;
3338
3339         table = kmemdup(ipv6_route_table_template,
3340                         sizeof(ipv6_route_table_template),
3341                         GFP_KERNEL);
3342
3343         if (table) {
3344                 table[0].data = &net->ipv6.sysctl.flush_delay;
3345                 table[0].extra1 = net;
3346                 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
3347                 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
3348                 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
3349                 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
3350                 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
3351                 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
3352                 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
3353                 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
3354                 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
3355
3356                 /* Don't export sysctls to unprivileged users */
3357                 if (net->user_ns != &init_user_ns)
3358                         table[0].procname = NULL;
3359         }
3360
3361         return table;
3362 }
3363 #endif
3364
3365 static int __net_init ip6_route_net_init(struct net *net)
3366 {
3367         int ret = -ENOMEM;
3368
3369         memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
3370                sizeof(net->ipv6.ip6_dst_ops));
3371
3372         if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
3373                 goto out_ip6_dst_ops;
3374
3375         net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
3376                                            sizeof(*net->ipv6.ip6_null_entry),
3377                                            GFP_KERNEL);
3378         if (!net->ipv6.ip6_null_entry)
3379                 goto out_ip6_dst_entries;
3380         net->ipv6.ip6_null_entry->dst.path =
3381                 (struct dst_entry *)net->ipv6.ip6_null_entry;
3382         net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3383         dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
3384                          ip6_template_metrics, true);
3385
3386 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3387         net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
3388                                                sizeof(*net->ipv6.ip6_prohibit_entry),
3389                                                GFP_KERNEL);
3390         if (!net->ipv6.ip6_prohibit_entry)
3391                 goto out_ip6_null_entry;
3392         net->ipv6.ip6_prohibit_entry->dst.path =
3393                 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
3394         net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3395         dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
3396                          ip6_template_metrics, true);
3397
3398         net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
3399                                                sizeof(*net->ipv6.ip6_blk_hole_entry),
3400                                                GFP_KERNEL);
3401         if (!net->ipv6.ip6_blk_hole_entry)
3402                 goto out_ip6_prohibit_entry;
3403         net->ipv6.ip6_blk_hole_entry->dst.path =
3404                 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
3405         net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3406         dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
3407                          ip6_template_metrics, true);
3408 #endif
3409
3410         net->ipv6.sysctl.flush_delay = 0;
3411         net->ipv6.sysctl.ip6_rt_max_size = 4096;
3412         net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
3413         net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
3414         net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
3415         net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
3416         net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
3417         net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
3418
3419         net->ipv6.ip6_rt_gc_expire = 30*HZ;
3420
3421         ret = 0;
3422 out:
3423         return ret;
3424
3425 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3426 out_ip6_prohibit_entry:
3427         kfree(net->ipv6.ip6_prohibit_entry);
3428 out_ip6_null_entry:
3429         kfree(net->ipv6.ip6_null_entry);
3430 #endif
3431 out_ip6_dst_entries:
3432         dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3433 out_ip6_dst_ops:
3434         goto out;
3435 }
3436
3437 static void __net_exit ip6_route_net_exit(struct net *net)
3438 {
3439         kfree(net->ipv6.ip6_null_entry);
3440 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3441         kfree(net->ipv6.ip6_prohibit_entry);
3442         kfree(net->ipv6.ip6_blk_hole_entry);
3443 #endif
3444         dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3445 }
3446
3447 static int __net_init ip6_route_net_init_late(struct net *net)
3448 {
3449 #ifdef CONFIG_PROC_FS
3450         proc_create("ipv6_route", 0, net->proc_net, &ipv6_route_proc_fops);
3451         proc_create("rt6_stats", S_IRUGO, net->proc_net, &rt6_stats_seq_fops);
3452 #endif
3453         return 0;
3454 }
3455
3456 static void __net_exit ip6_route_net_exit_late(struct net *net)
3457 {
3458 #ifdef CONFIG_PROC_FS
3459         remove_proc_entry("ipv6_route", net->proc_net);
3460         remove_proc_entry("rt6_stats", net->proc_net);
3461 #endif
3462 }
3463
3464 static struct pernet_operations ip6_route_net_ops = {
3465         .init = ip6_route_net_init,
3466         .exit = ip6_route_net_exit,
3467 };
3468
3469 static int __net_init ipv6_inetpeer_init(struct net *net)
3470 {
3471         struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
3472
3473         if (!bp)
3474                 return -ENOMEM;
3475         inet_peer_base_init(bp);
3476         net->ipv6.peers = bp;
3477         return 0;
3478 }
3479
3480 static void __net_exit ipv6_inetpeer_exit(struct net *net)
3481 {
3482         struct inet_peer_base *bp = net->ipv6.peers;
3483
3484         net->ipv6.peers = NULL;
3485         inetpeer_invalidate_tree(bp);
3486         kfree(bp);
3487 }
3488
3489 static struct pernet_operations ipv6_inetpeer_ops = {
3490         .init   =       ipv6_inetpeer_init,
3491         .exit   =       ipv6_inetpeer_exit,
3492 };
3493
3494 static struct pernet_operations ip6_route_net_late_ops = {
3495         .init = ip6_route_net_init_late,
3496         .exit = ip6_route_net_exit_late,
3497 };
3498
3499 static struct notifier_block ip6_route_dev_notifier = {
3500         .notifier_call = ip6_route_dev_notify,
3501         .priority = 0,
3502 };
3503
3504 int __init ip6_route_init(void)
3505 {
3506         int ret;
3507         int cpu;
3508
3509         ret = -ENOMEM;
3510         ip6_dst_ops_template.kmem_cachep =
3511                 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
3512                                   SLAB_HWCACHE_ALIGN, NULL);
3513         if (!ip6_dst_ops_template.kmem_cachep)
3514                 goto out;
3515
3516         ret = dst_entries_init(&ip6_dst_blackhole_ops);
3517         if (ret)
3518                 goto out_kmem_cache;
3519
3520         ret = register_pernet_subsys(&ipv6_inetpeer_ops);
3521         if (ret)
3522                 goto out_dst_entries;
3523
3524         ret = register_pernet_subsys(&ip6_route_net_ops);
3525         if (ret)
3526                 goto out_register_inetpeer;
3527
3528         ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
3529
3530         /* Registering of the loopback is done before this portion of code,
3531          * the loopback reference in rt6_info will not be taken, do it
3532          * manually for init_net */
3533         init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
3534         init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3535   #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3536         init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
3537         init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3538         init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
3539         init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3540   #endif
3541         ret = fib6_init();
3542         if (ret)
3543                 goto out_register_subsys;
3544
3545         ret = xfrm6_init();
3546         if (ret)
3547                 goto out_fib6_init;
3548
3549         ret = fib6_rules_init();
3550         if (ret)
3551                 goto xfrm6_init;
3552
3553         ret = register_pernet_subsys(&ip6_route_net_late_ops);
3554         if (ret)
3555                 goto fib6_rules_init;
3556
3557         ret = -ENOBUFS;
3558         if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL, NULL) ||
3559             __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL, NULL) ||
3560             __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL, NULL))
3561                 goto out_register_late_subsys;
3562
3563         ret = register_netdevice_notifier(&ip6_route_dev_notifier);
3564         if (ret)
3565                 goto out_register_late_subsys;
3566
3567         for_each_possible_cpu(cpu) {
3568                 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
3569
3570                 INIT_LIST_HEAD(&ul->head);
3571                 spin_lock_init(&ul->lock);
3572         }
3573
3574 out:
3575         return ret;
3576
3577 out_register_late_subsys:
3578         unregister_pernet_subsys(&ip6_route_net_late_ops);
3579 fib6_rules_init:
3580         fib6_rules_cleanup();
3581 xfrm6_init:
3582         xfrm6_fini();
3583 out_fib6_init:
3584         fib6_gc_cleanup();
3585 out_register_subsys:
3586         unregister_pernet_subsys(&ip6_route_net_ops);
3587 out_register_inetpeer:
3588         unregister_pernet_subsys(&ipv6_inetpeer_ops);
3589 out_dst_entries:
3590         dst_entries_destroy(&ip6_dst_blackhole_ops);
3591 out_kmem_cache:
3592         kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3593         goto out;
3594 }
3595
3596 void ip6_route_cleanup(void)
3597 {
3598         unregister_netdevice_notifier(&ip6_route_dev_notifier);
3599         unregister_pernet_subsys(&ip6_route_net_late_ops);
3600         fib6_rules_cleanup();
3601         xfrm6_fini();
3602         fib6_gc_cleanup();
3603         unregister_pernet_subsys(&ipv6_inetpeer_ops);
3604         unregister_pernet_subsys(&ip6_route_net_ops);
3605         dst_entries_destroy(&ip6_dst_blackhole_ops);
3606         kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3607 }