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         if (dev == loopback_dev)
 146                 return;
 147
 148         for_each_possible_cpu(cpu) {
 149                 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
 150                 struct rt6_info *rt;
 151
 152                 spin_lock_bh(&ul->lock);
 153                 list_for_each_entry(rt, &ul->head, rt6i_uncached) {
 154                         struct inet6_dev *rt_idev = rt->rt6i_idev;
 155                         struct net_device *rt_dev = rt->dst.dev;
 156
 157                         if (rt_idev->dev == dev) {
 158                                 rt->rt6i_idev = in6_dev_get(loopback_dev);
 159                                 in6_dev_put(rt_idev);
 160                         }
 161
 162                         if (rt_dev == dev) {
 163                                 rt->dst.dev = loopback_dev;
 164                                 dev_hold(rt->dst.dev);
 165                                 dev_put(rt_dev);
 166                         }
 167                 }
 168                 spin_unlock_bh(&ul->lock);
 169         }
 170 }
 171
 172 static u32 *rt6_pcpu_cow_metrics(struct rt6_info *rt)
 173 {
 174         return dst_metrics_write_ptr(rt->dst.from);
 175 }
 176
 177 static u32 *ipv6_cow_metrics(struct dst_entry *dst, unsigned long old)
 178 {
 179         struct rt6_info *rt = (struct rt6_info *)dst;
 180
 181         if (rt->rt6i_flags & RTF_PCPU)
 182                 return rt6_pcpu_cow_metrics(rt);
 183         else if (rt->rt6i_flags & RTF_CACHE)
 184                 return NULL;
 185         else
 186                 return dst_cow_metrics_generic(dst, old);
 187 }
 188
 189 static inline const void *choose_neigh_daddr(struct rt6_info *rt,
 190                                              struct sk_buff *skb,
 191                                              const void *daddr)
 192 {
 193         struct in6_addr *p = &rt->rt6i_gateway;
 194
 195         if (!ipv6_addr_any(p))
 196                 return (const void *) p;
 197         else if (skb)
 198                 return &ipv6_hdr(skb)->daddr;
 199         return daddr;
 200 }
 201
 202 static struct neighbour *ip6_neigh_lookup(const struct dst_entry *dst,
 203                                           struct sk_buff *skb,
 204                                           const void *daddr)
 205 {
 206         struct rt6_info *rt = (struct rt6_info *) dst;
 207         struct neighbour *n;
 208
 209         daddr = choose_neigh_daddr(rt, skb, daddr);
 210         n = __ipv6_neigh_lookup(dst->dev, daddr);
 211         if (n)
 212                 return n;
 213         return neigh_create(&nd_tbl, daddr, dst->dev);
 214 }
 215
 216 static struct dst_ops ip6_dst_ops_template = {
 217         .family                 =       AF_INET6,
 218         .gc                     =       ip6_dst_gc,
 219         .gc_thresh              =       1024,
 220         .check                  =       ip6_dst_check,
 221         .default_advmss         =       ip6_default_advmss,
 222         .mtu                    =       ip6_mtu,
 223         .cow_metrics            =       ipv6_cow_metrics,
 224         .destroy                =       ip6_dst_destroy,
 225         .ifdown                 =       ip6_dst_ifdown,
 226         .negative_advice        =       ip6_negative_advice,
 227         .link_failure           =       ip6_link_failure,
 228         .update_pmtu            =       ip6_rt_update_pmtu,
 229         .redirect               =       rt6_do_redirect,
 230         .local_out              =       __ip6_local_out,
 231         .neigh_lookup           =       ip6_neigh_lookup,
 232 };
 233
 234 static unsigned int ip6_blackhole_mtu(const struct dst_entry *dst)
 235 {
 236         unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
 237
 238         return mtu ? : dst->dev->mtu;
 239 }
 240
 241 static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, struct sock *sk,
 242                                          struct sk_buff *skb, u32 mtu)
 243 {
 244 }
 245
 246 static void ip6_rt_blackhole_redirect(struct dst_entry *dst, struct sock *sk,
 247                                       struct sk_buff *skb)
 248 {
 249 }
 250
 251 static u32 *ip6_rt_blackhole_cow_metrics(struct dst_entry *dst,
 252                                          unsigned long old)
 253 {
 254         return NULL;
 255 }
 256
 257 static struct dst_ops ip6_dst_blackhole_ops = {
 258         .family                 =       AF_INET6,
 259         .destroy                =       ip6_dst_destroy,
 260         .check                  =       ip6_dst_check,
 261         .mtu                    =       ip6_blackhole_mtu,
 262         .default_advmss         =       ip6_default_advmss,
 263         .update_pmtu            =       ip6_rt_blackhole_update_pmtu,
 264         .redirect               =       ip6_rt_blackhole_redirect,
 265         .cow_metrics            =       ip6_rt_blackhole_cow_metrics,
 266         .neigh_lookup           =       ip6_neigh_lookup,
 267 };
 268
 269 static const u32 ip6_template_metrics[RTAX_MAX] = {
 270         [RTAX_HOPLIMIT - 1] = 0,
 271 };
 272
 273 static const struct rt6_info ip6_null_entry_template = {
 274         .dst = {
 275                 .__refcnt       = ATOMIC_INIT(1),
 276                 .__use          = 1,
 277                 .obsolete       = DST_OBSOLETE_FORCE_CHK,
 278                 .error          = -ENETUNREACH,
 279                 .input          = ip6_pkt_discard,
 280                 .output         = ip6_pkt_discard_out,
 281         },
 282         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
 283         .rt6i_protocol  = RTPROT_KERNEL,
 284         .rt6i_metric    = ~(u32) 0,
 285         .rt6i_ref       = ATOMIC_INIT(1),
 286 };
 287
 288 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
 289
 290 static const struct rt6_info ip6_prohibit_entry_template = {
 291         .dst = {
 292                 .__refcnt       = ATOMIC_INIT(1),
 293                 .__use          = 1,
 294                 .obsolete       = DST_OBSOLETE_FORCE_CHK,
 295                 .error          = -EACCES,
 296                 .input          = ip6_pkt_prohibit,
 297                 .output         = ip6_pkt_prohibit_out,
 298         },
 299         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
 300         .rt6i_protocol  = RTPROT_KERNEL,
 301         .rt6i_metric    = ~(u32) 0,
 302         .rt6i_ref       = ATOMIC_INIT(1),
 303 };
 304
 305 static const struct rt6_info ip6_blk_hole_entry_template = {
 306         .dst = {
 307                 .__refcnt       = ATOMIC_INIT(1),
 308                 .__use          = 1,
 309                 .obsolete       = DST_OBSOLETE_FORCE_CHK,
 310                 .error          = -EINVAL,
 311                 .input          = dst_discard,
 312                 .output         = dst_discard_sk,
 313         },
 314         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
 315         .rt6i_protocol  = RTPROT_KERNEL,
 316         .rt6i_metric    = ~(u32) 0,
 317         .rt6i_ref       = ATOMIC_INIT(1),
 318 };
 319
 320 #endif
 321
 322 /* allocate dst with ip6_dst_ops */
 323 static struct rt6_info *__ip6_dst_alloc(struct net *net,
 324                                         struct net_device *dev,
 325                                         int flags)
 326 {
 327         struct rt6_info *rt = dst_alloc(&net->ipv6.ip6_dst_ops, dev,
 328                                         0, DST_OBSOLETE_FORCE_CHK, flags);
 329
 330         if (rt) {
 331                 struct dst_entry *dst = &rt->dst;
 332
 333                 memset(dst + 1, 0, sizeof(*rt) - sizeof(*dst));
 334                 INIT_LIST_HEAD(&rt->rt6i_siblings);
 335                 INIT_LIST_HEAD(&rt->rt6i_uncached);
 336         }
 337         return rt;
 338 }
 339
 340 static struct rt6_info *ip6_dst_alloc(struct net *net,
 341                                       struct net_device *dev,
 342                                       int flags)
 343 {
 344         struct rt6_info *rt = __ip6_dst_alloc(net, dev, flags);
 345
 346         if (rt) {
 347                 rt->rt6i_pcpu = alloc_percpu_gfp(struct rt6_info *, GFP_ATOMIC);
 348                 if (rt->rt6i_pcpu) {
 349                         int cpu;
 350
 351                         for_each_possible_cpu(cpu) {
 352                                 struct rt6_info **p;
 353
 354                                 p = per_cpu_ptr(rt->rt6i_pcpu, cpu);
 355                                 /* no one shares rt */
 356                                 *p =  NULL;
 357                         }
 358                 } else {
 359                         dst_destroy((struct dst_entry *)rt);
 360                         return NULL;
 361                 }
 362         }
 363
 364         return rt;
 365 }
 366
 367 static void ip6_dst_destroy(struct dst_entry *dst)
 368 {
 369         struct rt6_info *rt = (struct rt6_info *)dst;
 370         struct dst_entry *from = dst->from;
 371         struct inet6_dev *idev;
 372
 373         dst_destroy_metrics_generic(dst);
 374         free_percpu(rt->rt6i_pcpu);
 375         rt6_uncached_list_del(rt);
 376
 377         idev = rt->rt6i_idev;
 378         if (idev) {
 379                 rt->rt6i_idev = NULL;
 380                 in6_dev_put(idev);
 381         }
 382
 383         dst->from = NULL;
 384         dst_release(from);
 385 }
 386
 387 static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
 388                            int how)
 389 {
 390         struct rt6_info *rt = (struct rt6_info *)dst;
 391         struct inet6_dev *idev = rt->rt6i_idev;
 392         struct net_device *loopback_dev =
 393                 dev_net(dev)->loopback_dev;
 394
 395         if (dev != loopback_dev) {
 396                 if (idev && idev->dev == dev) {
 397                         struct inet6_dev *loopback_idev =
 398                                 in6_dev_get(loopback_dev);
 399                         if (loopback_idev) {
 400                                 rt->rt6i_idev = loopback_idev;
 401                                 in6_dev_put(idev);
 402                         }
 403                 }
 404         }
 405 }
 406
 407 static bool rt6_check_expired(const struct rt6_info *rt)
 408 {
 409         if (rt->rt6i_flags & RTF_EXPIRES) {
 410                 if (time_after(jiffies, rt->dst.expires))
 411                         return true;
 412         } else if (rt->dst.from) {
 413                 return rt6_check_expired((struct rt6_info *) rt->dst.from);
 414         }
 415         return false;
 416 }
 417
 418 /* Multipath route selection:
 419  *   Hash based function using packet header and flowlabel.
 420  * Adapted from fib_info_hashfn()
 421  */
 422 static int rt6_info_hash_nhsfn(unsigned int candidate_count,
 423                                const struct flowi6 *fl6)
 424 {
 425         unsigned int val = fl6->flowi6_proto;
 426
 427         val ^= ipv6_addr_hash(&fl6->daddr);
 428         val ^= ipv6_addr_hash(&fl6->saddr);
 429
 430         /* Work only if this not encapsulated */
 431         switch (fl6->flowi6_proto) {
 432         case IPPROTO_UDP:
 433         case IPPROTO_TCP:
 434         case IPPROTO_SCTP:
 435                 val ^= (__force u16)fl6->fl6_sport;
 436                 val ^= (__force u16)fl6->fl6_dport;
 437                 break;
 438
 439         case IPPROTO_ICMPV6:
 440                 val ^= (__force u16)fl6->fl6_icmp_type;
 441                 val ^= (__force u16)fl6->fl6_icmp_code;
 442                 break;
 443         }
 444         /* RFC6438 recommands to use flowlabel */
 445         val ^= (__force u32)fl6->flowlabel;
 446
 447         /* Perhaps, we need to tune, this function? */
 448         val = val ^ (val >> 7) ^ (val >> 12);
 449         return val % candidate_count;
 450 }
 451
 452 static struct rt6_info *rt6_multipath_select(struct rt6_info *match,
 453                                              struct flowi6 *fl6, int oif,
 454                                              int strict)
 455 {
 456         struct rt6_info *sibling, *next_sibling;
 457         int route_choosen;
 458
 459         route_choosen = rt6_info_hash_nhsfn(match->rt6i_nsiblings + 1, fl6);
 460         /* Don't change the route, if route_choosen == 0
 461          * (siblings does not include ourself)
 462          */
 463         if (route_choosen)
 464                 list_for_each_entry_safe(sibling, next_sibling,
 465                                 &match->rt6i_siblings, rt6i_siblings) {
 466                         route_choosen--;
 467                         if (route_choosen == 0) {
 468                                 if (rt6_score_route(sibling, oif, strict) < 0)
 469                                         break;
 470                                 match = sibling;
 471                                 break;
 472                         }
 473                 }
 474         return match;
 475 }
 476
 477 /*
 478  *      Route lookup. Any table->tb6_lock is implied.
 479  */
 480
 481 static inline struct rt6_info *rt6_device_match(struct net *net,
 482                                                     struct rt6_info *rt,
 483                                                     const struct in6_addr *saddr,
 484                                                     int oif,
 485                                                     int flags)
 486 {
 487         struct rt6_info *local = NULL;
 488         struct rt6_info *sprt;
 489
 490         if (!oif && ipv6_addr_any(saddr))
 491                 goto out;
 492
 493         for (sprt = rt; sprt; sprt = sprt->dst.rt6_next) {
 494                 struct net_device *dev = sprt->dst.dev;
 495
 496                 if (oif) {
 497                         if (dev->ifindex == oif)
 498                                 return sprt;
 499                         if (dev->flags & IFF_LOOPBACK) {
 500                                 if (!sprt->rt6i_idev ||
 501                                     sprt->rt6i_idev->dev->ifindex != oif) {
 502                                         if (flags & RT6_LOOKUP_F_IFACE && oif)
 503                                                 continue;
 504                                         if (local && (!oif ||
 505                                                       local->rt6i_idev->dev->ifindex == oif))
 506                                                 continue;
 507                                 }
 508                                 local = sprt;
 509                         }
 510                 } else {
 511                         if (ipv6_chk_addr(net, saddr, dev,
 512                                           flags & RT6_LOOKUP_F_IFACE))
 513                                 return sprt;
 514                 }
 515         }
 516
 517         if (oif) {
 518                 if (local)
 519                         return local;
 520
 521                 if (flags & RT6_LOOKUP_F_IFACE)
 522                         return net->ipv6.ip6_null_entry;
 523         }
 524 out:
 525         return rt;
 526 }
 527
 528 #ifdef CONFIG_IPV6_ROUTER_PREF
 529 struct __rt6_probe_work {
 530         struct work_struct work;
 531         struct in6_addr target;
 532         struct net_device *dev;
 533 };
 534
 535 static void rt6_probe_deferred(struct work_struct *w)
 536 {
 537         struct in6_addr mcaddr;
 538         struct __rt6_probe_work *work =
 539                 container_of(w, struct __rt6_probe_work, work);
 540
 541         addrconf_addr_solict_mult(&work->target, &mcaddr);
 542         ndisc_send_ns(work->dev, NULL, &work->target, &mcaddr, NULL, NULL);
 543         dev_put(work->dev);
 544         kfree(work);
 545 }
 546
 547 static void rt6_probe(struct rt6_info *rt)
 548 {
 549         struct __rt6_probe_work *work;
 550         struct neighbour *neigh;
 551         /*
 552          * Okay, this does not seem to be appropriate
 553          * for now, however, we need to check if it
 554          * is really so; aka Router Reachability Probing.
 555          *
 556          * Router Reachability Probe MUST be rate-limited
 557          * to no more than one per minute.
 558          */
 559         if (!rt || !(rt->rt6i_flags & RTF_GATEWAY))
 560                 return;
 561         rcu_read_lock_bh();
 562         neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
 563         if (neigh) {
 564                 if (neigh->nud_state & NUD_VALID)
 565                         goto out;
 566
 567                 work = NULL;
 568                 write_lock(&neigh->lock);
 569                 if (!(neigh->nud_state & NUD_VALID) &&
 570                     time_after(jiffies,
 571                                neigh->updated +
 572                                rt->rt6i_idev->cnf.rtr_probe_interval)) {
 573                         work = kmalloc(sizeof(*work), GFP_ATOMIC);
 574                         if (work)
 575                                 __neigh_set_probe_once(neigh);
 576                 }
 577                 write_unlock(&neigh->lock);
 578         } else {
 579                 work = kmalloc(sizeof(*work), GFP_ATOMIC);
 580         }
 581
 582         if (work) {
 583                 INIT_WORK(&work->work, rt6_probe_deferred);
 584                 work->target = rt->rt6i_gateway;
 585                 dev_hold(rt->dst.dev);
 586                 work->dev = rt->dst.dev;
 587                 schedule_work(&work->work);
 588         }
 589
 590 out:
 591         rcu_read_unlock_bh();
 592 }
 593 #else
 594 static inline void rt6_probe(struct rt6_info *rt)
 595 {
 596 }
 597 #endif
 598
 599 /*
 600  * Default Router Selection (RFC 2461 6.3.6)
 601  */
 602 static inline int rt6_check_dev(struct rt6_info *rt, int oif)
 603 {
 604         struct net_device *dev = rt->dst.dev;
 605         if (!oif || dev->ifindex == oif)
 606                 return 2;
 607         if ((dev->flags & IFF_LOOPBACK) &&
 608             rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
 609                 return 1;
 610         return 0;
 611 }
 612
 613 static inline enum rt6_nud_state rt6_check_neigh(struct rt6_info *rt)
 614 {
 615         struct neighbour *neigh;
 616         enum rt6_nud_state ret = RT6_NUD_FAIL_HARD;
 617
 618         if (rt->rt6i_flags & RTF_NONEXTHOP ||
 619             !(rt->rt6i_flags & RTF_GATEWAY))
 620                 return RT6_NUD_SUCCEED;
 621
 622         rcu_read_lock_bh();
 623         neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
 624         if (neigh) {
 625                 read_lock(&neigh->lock);
 626                 if (neigh->nud_state & NUD_VALID)
 627                         ret = RT6_NUD_SUCCEED;
 628 #ifdef CONFIG_IPV6_ROUTER_PREF
 629                 else if (!(neigh->nud_state & NUD_FAILED))
 630                         ret = RT6_NUD_SUCCEED;
 631                 else
 632                         ret = RT6_NUD_FAIL_PROBE;
 633 #endif
 634                 read_unlock(&neigh->lock);
 635         } else {
 636                 ret = IS_ENABLED(CONFIG_IPV6_ROUTER_PREF) ?
 637                       RT6_NUD_SUCCEED : RT6_NUD_FAIL_DO_RR;
 638         }
 639         rcu_read_unlock_bh();
 640
 641         return ret;
 642 }
 643
 644 static int rt6_score_route(struct rt6_info *rt, int oif,
 645                            int strict)
 646 {
 647         int m;
 648
 649         m = rt6_check_dev(rt, oif);
 650         if (!m && (strict & RT6_LOOKUP_F_IFACE))
 651                 return RT6_NUD_FAIL_HARD;
 652 #ifdef CONFIG_IPV6_ROUTER_PREF
 653         m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
 654 #endif
 655         if (strict & RT6_LOOKUP_F_REACHABLE) {
 656                 int n = rt6_check_neigh(rt);
 657                 if (n < 0)
 658                         return n;
 659         }
 660         return m;
 661 }
 662
 663 static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
 664                                    int *mpri, struct rt6_info *match,
 665                                    bool *do_rr)
 666 {
 667         int m;
 668         bool match_do_rr = false;
 669         struct inet6_dev *idev = rt->rt6i_idev;
 670         struct net_device *dev = rt->dst.dev;
 671
 672         if (dev && !netif_carrier_ok(dev) &&
 673             idev->cnf.ignore_routes_with_linkdown)
 674                 goto out;
 675
 676         if (rt6_check_expired(rt))
 677                 goto out;
 678
 679         m = rt6_score_route(rt, oif, strict);
 680         if (m == RT6_NUD_FAIL_DO_RR) {
 681                 match_do_rr = true;
 682                 m = 0; /* lowest valid score */
 683         } else if (m == RT6_NUD_FAIL_HARD) {
 684                 goto out;
 685         }
 686
 687         if (strict & RT6_LOOKUP_F_REACHABLE)
 688                 rt6_probe(rt);
 689
 690         /* note that m can be RT6_NUD_FAIL_PROBE at this point */
 691         if (m > *mpri) {
 692                 *do_rr = match_do_rr;
 693                 *mpri = m;
 694                 match = rt;
 695         }
 696 out:
 697         return match;
 698 }
 699
 700 static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
 701                                      struct rt6_info *rr_head,
 702                                      u32 metric, int oif, int strict,
 703                                      bool *do_rr)
 704 {
 705         struct rt6_info *rt, *match, *cont;
 706         int mpri = -1;
 707
 708         match = NULL;
 709         cont = NULL;
 710         for (rt = rr_head; rt; rt = rt->dst.rt6_next) {
 711                 if (rt->rt6i_metric != metric) {
 712                         cont = rt;
 713                         break;
 714                 }
 715
 716                 match = find_match(rt, oif, strict, &mpri, match, do_rr);
 717         }
 718
 719         for (rt = fn->leaf; rt && rt != rr_head; rt = rt->dst.rt6_next) {
 720                 if (rt->rt6i_metric != metric) {
 721                         cont = rt;
 722                         break;
 723                 }
 724
 725                 match = find_match(rt, oif, strict, &mpri, match, do_rr);
 726         }
 727
 728         if (match || !cont)
 729                 return match;
 730
 731         for (rt = cont; rt; rt = rt->dst.rt6_next)
 732                 match = find_match(rt, oif, strict, &mpri, match, do_rr);
 733
 734         return match;
 735 }
 736
 737 static struct rt6_info *rt6_select(struct fib6_node *fn, int oif, int strict)
 738 {
 739         struct rt6_info *match, *rt0;
 740         struct net *net;
 741         bool do_rr = false;
 742
 743         rt0 = fn->rr_ptr;
 744         if (!rt0)
 745                 fn->rr_ptr = rt0 = fn->leaf;
 746
 747         match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict,
 748                              &do_rr);
 749
 750         if (do_rr) {
 751                 struct rt6_info *next = rt0->dst.rt6_next;
 752
 753                 /* no entries matched; do round-robin */
 754                 if (!next || next->rt6i_metric != rt0->rt6i_metric)
 755                         next = fn->leaf;
 756
 757                 if (next != rt0)
 758                         fn->rr_ptr = next;
 759         }
 760
 761         net = dev_net(rt0->dst.dev);
 762         return match ? match : net->ipv6.ip6_null_entry;
 763 }
 764
 765 static bool rt6_is_gw_or_nonexthop(const struct rt6_info *rt)
 766 {
 767         return (rt->rt6i_flags & (RTF_NONEXTHOP | RTF_GATEWAY));
 768 }
 769
 770 #ifdef CONFIG_IPV6_ROUTE_INFO
 771 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
 772                   const struct in6_addr *gwaddr)
 773 {
 774         struct net *net = dev_net(dev);
 775         struct route_info *rinfo = (struct route_info *) opt;
 776         struct in6_addr prefix_buf, *prefix;
 777         unsigned int pref;
 778         unsigned long lifetime;
 779         struct rt6_info *rt;
 780
 781         if (len < sizeof(struct route_info)) {
 782                 return -EINVAL;
 783         }
 784
 785         /* Sanity check for prefix_len and length */
 786         if (rinfo->length > 3) {
 787                 return -EINVAL;
 788         } else if (rinfo->prefix_len > 128) {
 789                 return -EINVAL;
 790         } else if (rinfo->prefix_len > 64) {
 791                 if (rinfo->length < 2) {
 792                         return -EINVAL;
 793                 }
 794         } else if (rinfo->prefix_len > 0) {
 795                 if (rinfo->length < 1) {
 796                         return -EINVAL;
 797                 }
 798         }
 799
 800         pref = rinfo->route_pref;
 801         if (pref == ICMPV6_ROUTER_PREF_INVALID)
 802                 return -EINVAL;
 803
 804         lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
 805
 806         if (rinfo->length == 3)
 807                 prefix = (struct in6_addr *)rinfo->prefix;
 808         else {
 809                 /* this function is safe */
 810                 ipv6_addr_prefix(&prefix_buf,
 811                                  (struct in6_addr *)rinfo->prefix,
 812                                  rinfo->prefix_len);
 813                 prefix = &prefix_buf;
 814         }
 815
 816         if (rinfo->prefix_len == 0)
 817                 rt = rt6_get_dflt_router(gwaddr, dev);
 818         else
 819                 rt = rt6_get_route_info(net, prefix, rinfo->prefix_len,
 820                                         gwaddr, dev->ifindex);
 821
 822         if (rt && !lifetime) {
 823                 ip6_del_rt(rt);
 824                 rt = NULL;
 825         }
 826
 827         if (!rt && lifetime)
 828                 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr, dev->ifindex,
 829                                         pref);
 830         else if (rt)
 831                 rt->rt6i_flags = RTF_ROUTEINFO |
 832                                  (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
 833
 834         if (rt) {
 835                 if (!addrconf_finite_timeout(lifetime))
 836                         rt6_clean_expires(rt);
 837                 else
 838                         rt6_set_expires(rt, jiffies + HZ * lifetime);
 839
 840                 ip6_rt_put(rt);
 841         }
 842         return 0;
 843 }
 844 #endif
 845
 846 static struct fib6_node* fib6_backtrack(struct fib6_node *fn,
 847                                         struct in6_addr *saddr)
 848 {
 849         struct fib6_node *pn;
 850         while (1) {
 851                 if (fn->fn_flags & RTN_TL_ROOT)
 852                         return NULL;
 853                 pn = fn->parent;
 854                 if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn)
 855                         fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr);
 856                 else
 857                         fn = pn;
 858                 if (fn->fn_flags & RTN_RTINFO)
 859                         return fn;
 860         }
 861 }
 862
 863 static struct rt6_info *ip6_pol_route_lookup(struct net *net,
 864                                              struct fib6_table *table,
 865                                              struct flowi6 *fl6, int flags)
 866 {
 867         struct fib6_node *fn;
 868         struct rt6_info *rt;
 869
 870         read_lock_bh(&table->tb6_lock);
 871         fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
 872 restart:
 873         rt = fn->leaf;
 874         rt = rt6_device_match(net, rt, &fl6->saddr, fl6->flowi6_oif, flags);
 875         if (rt->rt6i_nsiblings && fl6->flowi6_oif == 0)
 876                 rt = rt6_multipath_select(rt, fl6, fl6->flowi6_oif, flags);
 877         if (rt == net->ipv6.ip6_null_entry) {
 878                 fn = fib6_backtrack(fn, &fl6->saddr);
 879                 if (fn)
 880                         goto restart;
 881         }
 882         dst_use(&rt->dst, jiffies);
 883         read_unlock_bh(&table->tb6_lock);
 884         return rt;
 885
 886 }
 887
 888 struct dst_entry *ip6_route_lookup(struct net *net, struct flowi6 *fl6,
 889                                     int flags)
 890 {
 891         return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_lookup);
 892 }
 893 EXPORT_SYMBOL_GPL(ip6_route_lookup);
 894
 895 struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
 896                             const struct in6_addr *saddr, int oif, int strict)
 897 {
 898         struct flowi6 fl6 = {
 899                 .flowi6_oif = oif,
 900                 .daddr = *daddr,
 901         };
 902         struct dst_entry *dst;
 903         int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
 904
 905         if (saddr) {
 906                 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
 907                 flags |= RT6_LOOKUP_F_HAS_SADDR;
 908         }
 909
 910         dst = fib6_rule_lookup(net, &fl6, flags, ip6_pol_route_lookup);
 911         if (dst->error == 0)
 912                 return (struct rt6_info *) dst;
 913
 914         dst_release(dst);
 915
 916         return NULL;
 917 }
 918 EXPORT_SYMBOL(rt6_lookup);
 919
 920 /* ip6_ins_rt is called with FREE table->tb6_lock.
 921    It takes new route entry, the addition fails by any reason the
 922    route is freed. In any case, if caller does not hold it, it may
 923    be destroyed.
 924  */
 925
 926 static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info,
 927                         struct mx6_config *mxc)
 928 {
 929         int err;
 930         struct fib6_table *table;
 931
 932         table = rt->rt6i_table;
 933         write_lock_bh(&table->tb6_lock);
 934         err = fib6_add(&table->tb6_root, rt, info, mxc);
 935         write_unlock_bh(&table->tb6_lock);
 936
 937         return err;
 938 }
 939
 940 int ip6_ins_rt(struct rt6_info *rt)
 941 {
 942         struct nl_info info = { .nl_net = dev_net(rt->dst.dev), };
 943         struct mx6_config mxc = { .mx = NULL, };
 944
 945         return __ip6_ins_rt(rt, &info, &mxc);
 946 }
 947
 948 static struct rt6_info *ip6_rt_cache_alloc(struct rt6_info *ort,
 949                                            const struct in6_addr *daddr,
 950                                            const struct in6_addr *saddr)
 951 {
 952         struct rt6_info *rt;
 953
 954         /*
 955          *      Clone the route.
 956          */
 957
 958         if (ort->rt6i_flags & (RTF_CACHE | RTF_PCPU))
 959                 ort = (struct rt6_info *)ort->dst.from;
 960
 961         rt = __ip6_dst_alloc(dev_net(ort->dst.dev), ort->dst.dev, 0);
 962
 963         if (!rt)
 964                 return NULL;
 965
 966         ip6_rt_copy_init(rt, ort);
 967         rt->rt6i_flags |= RTF_CACHE;
 968         rt->rt6i_metric = 0;
 969         rt->dst.flags |= DST_HOST;
 970         rt->rt6i_dst.addr = *daddr;
 971         rt->rt6i_dst.plen = 128;
 972
 973         if (!rt6_is_gw_or_nonexthop(ort)) {
 974                 if (ort->rt6i_dst.plen != 128 &&
 975                     ipv6_addr_equal(&ort->rt6i_dst.addr, daddr))
 976                         rt->rt6i_flags |= RTF_ANYCAST;
 977 #ifdef CONFIG_IPV6_SUBTREES
 978                 if (rt->rt6i_src.plen && saddr) {
 979                         rt->rt6i_src.addr = *saddr;
 980                         rt->rt6i_src.plen = 128;
 981                 }
 982 #endif
 983         }
 984
 985         return rt;
 986 }
 987
 988 static struct rt6_info *ip6_rt_pcpu_alloc(struct rt6_info *rt)
 989 {
 990         struct rt6_info *pcpu_rt;
 991
 992         pcpu_rt = __ip6_dst_alloc(dev_net(rt->dst.dev),
 993                                   rt->dst.dev, rt->dst.flags);
 994
 995         if (!pcpu_rt)
 996                 return NULL;
 997         ip6_rt_copy_init(pcpu_rt, rt);
 998         pcpu_rt->rt6i_protocol = rt->rt6i_protocol;
 999         pcpu_rt->rt6i_flags |= RTF_PCPU;
1000         return pcpu_rt;
1001 }
1002
1003 /* It should be called with read_lock_bh(&tb6_lock) acquired */
1004 static struct rt6_info *rt6_get_pcpu_route(struct rt6_info *rt)
1005 {
1006         struct rt6_info *pcpu_rt, **p;
1007
1008         p = this_cpu_ptr(rt->rt6i_pcpu);
1009         pcpu_rt = *p;
1010
1011         if (pcpu_rt) {
1012                 dst_hold(&pcpu_rt->dst);
1013                 rt6_dst_from_metrics_check(pcpu_rt);
1014         }
1015         return pcpu_rt;
1016 }
1017
1018 static struct rt6_info *rt6_make_pcpu_route(struct rt6_info *rt)
1019 {
1020         struct fib6_table *table = rt->rt6i_table;
1021         struct rt6_info *pcpu_rt, *prev, **p;
1022
1023         pcpu_rt = ip6_rt_pcpu_alloc(rt);
1024         if (!pcpu_rt) {
1025                 struct net *net = dev_net(rt->dst.dev);
1026
1027                 dst_hold(&net->ipv6.ip6_null_entry->dst);
1028                 return net->ipv6.ip6_null_entry;
1029         }
1030
1031         read_lock_bh(&table->tb6_lock);
1032         if (rt->rt6i_pcpu) {
1033                 p = this_cpu_ptr(rt->rt6i_pcpu);
1034                 prev = cmpxchg(p, NULL, pcpu_rt);
1035                 if (prev) {
1036                         /* If someone did it before us, return prev instead */
1037                         dst_destroy(&pcpu_rt->dst);
1038                         pcpu_rt = prev;
1039                 }
1040         } else {
1041                 /* rt has been removed from the fib6 tree
1042                  * before we have a chance to acquire the read_lock.
1043                  * In this case, don't brother to create a pcpu rt
1044                  * since rt is going away anyway.  The next
1045                  * dst_check() will trigger a re-lookup.
1046                  */
1047                 dst_destroy(&pcpu_rt->dst);
1048                 pcpu_rt = rt;
1049         }
1050         dst_hold(&pcpu_rt->dst);
1051         rt6_dst_from_metrics_check(pcpu_rt);
1052         read_unlock_bh(&table->tb6_lock);
1053         return pcpu_rt;
1054 }
1055
1056 static struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table, int oif,
1057                                       struct flowi6 *fl6, int flags)
1058 {
1059         struct fib6_node *fn, *saved_fn;
1060         struct rt6_info *rt;
1061         int strict = 0;
1062
1063         strict |= flags & RT6_LOOKUP_F_IFACE;
1064         if (net->ipv6.devconf_all->forwarding == 0)
1065                 strict |= RT6_LOOKUP_F_REACHABLE;
1066
1067         read_lock_bh(&table->tb6_lock);
1068
1069         fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1070         saved_fn = fn;
1071
1072 redo_rt6_select:
1073         rt = rt6_select(fn, oif, strict);
1074         if (rt->rt6i_nsiblings)
1075                 rt = rt6_multipath_select(rt, fl6, oif, strict);
1076         if (rt == net->ipv6.ip6_null_entry) {
1077                 fn = fib6_backtrack(fn, &fl6->saddr);
1078                 if (fn)
1079                         goto redo_rt6_select;
1080                 else if (strict & RT6_LOOKUP_F_REACHABLE) {
1081                         /* also consider unreachable route */
1082                         strict &= ~RT6_LOOKUP_F_REACHABLE;
1083                         fn = saved_fn;
1084                         goto redo_rt6_select;
1085                 }
1086         }
1087
1088
1089         if (rt == net->ipv6.ip6_null_entry || (rt->rt6i_flags & RTF_CACHE)) {
1090                 dst_use(&rt->dst, jiffies);
1091                 read_unlock_bh(&table->tb6_lock);
1092
1093                 rt6_dst_from_metrics_check(rt);
1094                 return rt;
1095         } else if (unlikely((fl6->flowi6_flags & FLOWI_FLAG_KNOWN_NH) &&
1096                             !(rt->rt6i_flags & RTF_GATEWAY))) {
1097                 /* Create a RTF_CACHE clone which will not be
1098                  * owned by the fib6 tree.  It is for the special case where
1099                  * the daddr in the skb during the neighbor look-up is different
1100                  * from the fl6->daddr used to look-up route here.
1101                  */
1102
1103                 struct rt6_info *uncached_rt;
1104
1105                 dst_use(&rt->dst, jiffies);
1106                 read_unlock_bh(&table->tb6_lock);
1107
1108                 uncached_rt = ip6_rt_cache_alloc(rt, &fl6->daddr, NULL);
1109                 dst_release(&rt->dst);
1110
1111                 if (uncached_rt)
1112                         rt6_uncached_list_add(uncached_rt);
1113                 else
1114                         uncached_rt = net->ipv6.ip6_null_entry;
1115
1116                 dst_hold(&uncached_rt->dst);
1117                 return uncached_rt;
1118
1119         } else {
1120                 /* Get a percpu copy */
1121
1122                 struct rt6_info *pcpu_rt;
1123
1124                 rt->dst.lastuse = jiffies;
1125                 rt->dst.__use++;
1126                 pcpu_rt = rt6_get_pcpu_route(rt);
1127
1128                 if (pcpu_rt) {
1129                         read_unlock_bh(&table->tb6_lock);
1130                 } else {
1131                         /* We have to do the read_unlock first
1132                          * because rt6_make_pcpu_route() may trigger
1133                          * ip6_dst_gc() which will take the write_lock.
1134                          */
1135                         dst_hold(&rt->dst);
1136                         read_unlock_bh(&table->tb6_lock);
1137                         pcpu_rt = rt6_make_pcpu_route(rt);
1138                         dst_release(&rt->dst);
1139                 }
1140
1141                 return pcpu_rt;
1142
1143         }
1144 }
1145
1146 static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
1147                                             struct flowi6 *fl6, int flags)
1148 {
1149         return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, flags);
1150 }
1151
1152 static struct dst_entry *ip6_route_input_lookup(struct net *net,
1153                                                 struct net_device *dev,
1154                                                 struct flowi6 *fl6, int flags)
1155 {
1156         if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
1157                 flags |= RT6_LOOKUP_F_IFACE;
1158
1159         return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_input);
1160 }
1161
1162 void ip6_route_input(struct sk_buff *skb)
1163 {
1164         const struct ipv6hdr *iph = ipv6_hdr(skb);
1165         struct net *net = dev_net(skb->dev);
1166         int flags = RT6_LOOKUP_F_HAS_SADDR;
1167         struct ip_tunnel_info *tun_info;
1168         struct flowi6 fl6 = {
1169                 .flowi6_iif = skb->dev->ifindex,
1170                 .daddr = iph->daddr,
1171                 .saddr = iph->saddr,
1172                 .flowlabel = ip6_flowinfo(iph),
1173                 .flowi6_mark = skb->mark,
1174                 .flowi6_proto = iph->nexthdr,
1175         };
1176
1177         tun_info = skb_tunnel_info(skb);
1178         if (tun_info && !(tun_info->mode & IP_TUNNEL_INFO_TX))
1179                 fl6.flowi6_tun_key.tun_id = tun_info->key.tun_id;
1180         skb_dst_drop(skb);
1181         skb_dst_set(skb, ip6_route_input_lookup(net, skb->dev, &fl6, flags));
1182 }
1183
1184 static struct rt6_info *ip6_pol_route_output(struct net *net, struct fib6_table *table,
1185                                              struct flowi6 *fl6, int flags)
1186 {
1187         return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, flags);
1188 }
1189
1190 struct dst_entry *ip6_route_output(struct net *net, const struct sock *sk,
1191                                     struct flowi6 *fl6)
1192 {
1193         int flags = 0;
1194
1195         fl6->flowi6_iif = LOOPBACK_IFINDEX;
1196
1197         if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr) ||
1198             fl6->flowi6_oif)
1199                 flags |= RT6_LOOKUP_F_IFACE;
1200
1201         if (!ipv6_addr_any(&fl6->saddr))
1202                 flags |= RT6_LOOKUP_F_HAS_SADDR;
1203         else if (sk)
1204                 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
1205
1206         return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_output);
1207 }
1208 EXPORT_SYMBOL(ip6_route_output);
1209
1210 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
1211 {
1212         struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
1213         struct dst_entry *new = NULL;
1214
1215         rt = dst_alloc(&ip6_dst_blackhole_ops, ort->dst.dev, 1, DST_OBSOLETE_NONE, 0);
1216         if (rt) {
1217                 new = &rt->dst;
1218
1219                 memset(new + 1, 0, sizeof(*rt) - sizeof(*new));
1220
1221                 new->__use = 1;
1222                 new->input = dst_discard;
1223                 new->output = dst_discard_sk;
1224
1225                 if (dst_metrics_read_only(&ort->dst))
1226                         new->_metrics = ort->dst._metrics;
1227                 else
1228                         dst_copy_metrics(new, &ort->dst);
1229                 rt->rt6i_idev = ort->rt6i_idev;
1230                 if (rt->rt6i_idev)
1231                         in6_dev_hold(rt->rt6i_idev);
1232
1233                 rt->rt6i_gateway = ort->rt6i_gateway;
1234                 rt->rt6i_flags = ort->rt6i_flags;
1235                 rt->rt6i_metric = 0;
1236
1237                 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1238 #ifdef CONFIG_IPV6_SUBTREES
1239                 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1240 #endif
1241
1242                 dst_free(new);
1243         }
1244
1245         dst_release(dst_orig);
1246         return new ? new : ERR_PTR(-ENOMEM);
1247 }
1248
1249 /*
1250  *      Destination cache support functions
1251  */
1252
1253 static void rt6_dst_from_metrics_check(struct rt6_info *rt)
1254 {
1255         if (rt->dst.from &&
1256             dst_metrics_ptr(&rt->dst) != dst_metrics_ptr(rt->dst.from))
1257                 dst_init_metrics(&rt->dst, dst_metrics_ptr(rt->dst.from), true);
1258 }
1259
1260 static struct dst_entry *rt6_check(struct rt6_info *rt, u32 cookie)
1261 {
1262         if (!rt->rt6i_node || (rt->rt6i_node->fn_sernum != cookie))
1263                 return NULL;
1264
1265         if (rt6_check_expired(rt))
1266                 return NULL;
1267
1268         return &rt->dst;
1269 }
1270
1271 static struct dst_entry *rt6_dst_from_check(struct rt6_info *rt, u32 cookie)
1272 {
1273         if (rt->dst.obsolete == DST_OBSOLETE_FORCE_CHK &&
1274             rt6_check((struct rt6_info *)(rt->dst.from), cookie))
1275                 return &rt->dst;
1276         else
1277                 return NULL;
1278 }
1279
1280 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
1281 {
1282         struct rt6_info *rt;
1283
1284         rt = (struct rt6_info *) dst;
1285
1286         /* All IPV6 dsts are created with ->obsolete set to the value
1287          * DST_OBSOLETE_FORCE_CHK which forces validation calls down
1288          * into this function always.
1289          */
1290
1291         rt6_dst_from_metrics_check(rt);
1292
1293         if ((rt->rt6i_flags & RTF_PCPU) || unlikely(dst->flags & DST_NOCACHE))
1294                 return rt6_dst_from_check(rt, cookie);
1295         else
1296                 return rt6_check(rt, cookie);
1297 }
1298
1299 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
1300 {
1301         struct rt6_info *rt = (struct rt6_info *) dst;
1302
1303         if (rt) {
1304                 if (rt->rt6i_flags & RTF_CACHE) {
1305                         if (rt6_check_expired(rt)) {
1306                                 ip6_del_rt(rt);
1307                                 dst = NULL;
1308                         }
1309                 } else {
1310                         dst_release(dst);
1311                         dst = NULL;
1312                 }
1313         }
1314         return dst;
1315 }
1316
1317 static void ip6_link_failure(struct sk_buff *skb)
1318 {
1319         struct rt6_info *rt;
1320
1321         icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
1322
1323         rt = (struct rt6_info *) skb_dst(skb);
1324         if (rt) {
1325                 if (rt->rt6i_flags & RTF_CACHE) {
1326                         dst_hold(&rt->dst);
1327                         ip6_del_rt(rt);
1328                 } else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT)) {
1329                         rt->rt6i_node->fn_sernum = -1;
1330                 }
1331         }
1332 }
1333
1334 static void rt6_do_update_pmtu(struct rt6_info *rt, u32 mtu)
1335 {
1336         struct net *net = dev_net(rt->dst.dev);
1337
1338         rt->rt6i_flags |= RTF_MODIFIED;
1339         rt->rt6i_pmtu = mtu;
1340         rt6_update_expires(rt, net->ipv6.sysctl.ip6_rt_mtu_expires);
1341 }
1342
1343 static void __ip6_rt_update_pmtu(struct dst_entry *dst, const struct sock *sk,
1344                                  const struct ipv6hdr *iph, u32 mtu)
1345 {
1346         struct rt6_info *rt6 = (struct rt6_info *)dst;
1347
1348         if (rt6->rt6i_flags & RTF_LOCAL)
1349                 return;
1350
1351         dst_confirm(dst);
1352         mtu = max_t(u32, mtu, IPV6_MIN_MTU);
1353         if (mtu >= dst_mtu(dst))
1354                 return;
1355
1356         if (rt6->rt6i_flags & RTF_CACHE) {
1357                 rt6_do_update_pmtu(rt6, mtu);
1358         } else {
1359                 const struct in6_addr *daddr, *saddr;
1360                 struct rt6_info *nrt6;
1361
1362                 if (iph) {
1363                         daddr = &iph->daddr;
1364                         saddr = &iph->saddr;
1365                 } else if (sk) {
1366                         daddr = &sk->sk_v6_daddr;
1367                         saddr = &inet6_sk(sk)->saddr;
1368                 } else {
1369                         return;
1370                 }
1371                 nrt6 = ip6_rt_cache_alloc(rt6, daddr, saddr);
1372                 if (nrt6) {
1373                         rt6_do_update_pmtu(nrt6, mtu);
1374
1375                         /* ip6_ins_rt(nrt6) will bump the
1376                          * rt6->rt6i_node->fn_sernum
1377                          * which will fail the next rt6_check() and
1378                          * invalidate the sk->sk_dst_cache.
1379                          */
1380                         ip6_ins_rt(nrt6);
1381                 }
1382         }
1383 }
1384
1385 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
1386                                struct sk_buff *skb, u32 mtu)
1387 {
1388         __ip6_rt_update_pmtu(dst, sk, skb ? ipv6_hdr(skb) : NULL, mtu);
1389 }
1390
1391 void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
1392                      int oif, u32 mark)
1393 {
1394         const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1395         struct dst_entry *dst;
1396         struct flowi6 fl6;
1397
1398         memset(&fl6, 0, sizeof(fl6));
1399         fl6.flowi6_oif = oif;
1400         fl6.flowi6_mark = mark ? mark : IP6_REPLY_MARK(net, skb->mark);
1401         fl6.daddr = iph->daddr;
1402         fl6.saddr = iph->saddr;
1403         fl6.flowlabel = ip6_flowinfo(iph);
1404
1405         dst = ip6_route_output(net, NULL, &fl6);
1406         if (!dst->error)
1407                 __ip6_rt_update_pmtu(dst, NULL, iph, ntohl(mtu));
1408         dst_release(dst);
1409 }
1410 EXPORT_SYMBOL_GPL(ip6_update_pmtu);
1411
1412 void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
1413 {
1414         ip6_update_pmtu(skb, sock_net(sk), mtu,
1415                         sk->sk_bound_dev_if, sk->sk_mark);
1416 }
1417 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
1418
1419 /* Handle redirects */
1420 struct ip6rd_flowi {
1421         struct flowi6 fl6;
1422         struct in6_addr gateway;
1423 };
1424
1425 static struct rt6_info *__ip6_route_redirect(struct net *net,
1426                                              struct fib6_table *table,
1427                                              struct flowi6 *fl6,
1428                                              int flags)
1429 {
1430         struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
1431         struct rt6_info *rt;
1432         struct fib6_node *fn;
1433
1434         /* Get the "current" route for this destination and
1435          * check if the redirect has come from approriate router.
1436          *
1437          * RFC 4861 specifies that redirects should only be
1438          * accepted if they come from the nexthop to the target.
1439          * Due to the way the routes are chosen, this notion
1440          * is a bit fuzzy and one might need to check all possible
1441          * routes.
1442          */
1443
1444         read_lock_bh(&table->tb6_lock);
1445         fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1446 restart:
1447         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1448                 if (rt6_check_expired(rt))
1449                         continue;
1450                 if (rt->dst.error)
1451                         break;
1452                 if (!(rt->rt6i_flags & RTF_GATEWAY))
1453                         continue;
1454                 if (fl6->flowi6_oif != rt->dst.dev->ifindex)
1455                         continue;
1456                 if (!ipv6_addr_equal(&rdfl->gateway, &rt->rt6i_gateway))
1457                         continue;
1458                 break;
1459         }
1460
1461         if (!rt)
1462                 rt = net->ipv6.ip6_null_entry;
1463         else if (rt->dst.error) {
1464                 rt = net->ipv6.ip6_null_entry;
1465                 goto out;
1466         }
1467
1468         if (rt == net->ipv6.ip6_null_entry) {
1469                 fn = fib6_backtrack(fn, &fl6->saddr);
1470                 if (fn)
1471                         goto restart;
1472         }
1473
1474 out:
1475         dst_hold(&rt->dst);
1476
1477         read_unlock_bh(&table->tb6_lock);
1478
1479         return rt;
1480 };
1481
1482 static struct dst_entry *ip6_route_redirect(struct net *net,
1483                                         const struct flowi6 *fl6,
1484                                         const struct in6_addr *gateway)
1485 {
1486         int flags = RT6_LOOKUP_F_HAS_SADDR;
1487         struct ip6rd_flowi rdfl;
1488
1489         rdfl.fl6 = *fl6;
1490         rdfl.gateway = *gateway;
1491
1492         return fib6_rule_lookup(net, &rdfl.fl6,
1493                                 flags, __ip6_route_redirect);
1494 }
1495
1496 void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark)
1497 {
1498         const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1499         struct dst_entry *dst;
1500         struct flowi6 fl6;
1501
1502         memset(&fl6, 0, sizeof(fl6));
1503         fl6.flowi6_iif = LOOPBACK_IFINDEX;
1504         fl6.flowi6_oif = oif;
1505         fl6.flowi6_mark = mark;
1506         fl6.daddr = iph->daddr;
1507         fl6.saddr = iph->saddr;
1508         fl6.flowlabel = ip6_flowinfo(iph);
1509
1510         dst = ip6_route_redirect(net, &fl6, &ipv6_hdr(skb)->saddr);
1511         rt6_do_redirect(dst, NULL, skb);
1512         dst_release(dst);
1513 }
1514 EXPORT_SYMBOL_GPL(ip6_redirect);
1515
1516 void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif,
1517                             u32 mark)
1518 {
1519         const struct ipv6hdr *iph = ipv6_hdr(skb);
1520         const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
1521         struct dst_entry *dst;
1522         struct flowi6 fl6;
1523
1524         memset(&fl6, 0, sizeof(fl6));
1525         fl6.flowi6_iif = LOOPBACK_IFINDEX;
1526         fl6.flowi6_oif = oif;
1527         fl6.flowi6_mark = mark;
1528         fl6.daddr = msg->dest;
1529         fl6.saddr = iph->daddr;
1530
1531         dst = ip6_route_redirect(net, &fl6, &iph->saddr);
1532         rt6_do_redirect(dst, NULL, skb);
1533         dst_release(dst);
1534 }
1535
1536 void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
1537 {
1538         ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark);
1539 }
1540 EXPORT_SYMBOL_GPL(ip6_sk_redirect);
1541
1542 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1543 {
1544         struct net_device *dev = dst->dev;
1545         unsigned int mtu = dst_mtu(dst);
1546         struct net *net = dev_net(dev);
1547
1548         mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
1549
1550         if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
1551                 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1552
1553         /*
1554          * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1555          * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1556          * IPV6_MAXPLEN is also valid and means: "any MSS,
1557          * rely only on pmtu discovery"
1558          */
1559         if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
1560                 mtu = IPV6_MAXPLEN;
1561         return mtu;
1562 }
1563
1564 static unsigned int ip6_mtu(const struct dst_entry *dst)
1565 {
1566         const struct rt6_info *rt = (const struct rt6_info *)dst;
1567         unsigned int mtu = rt->rt6i_pmtu;
1568         struct inet6_dev *idev;
1569
1570         if (mtu)
1571                 goto out;
1572
1573         mtu = dst_metric_raw(dst, RTAX_MTU);
1574         if (mtu)
1575                 goto out;
1576
1577         mtu = IPV6_MIN_MTU;
1578
1579         rcu_read_lock();
1580         idev = __in6_dev_get(dst->dev);
1581         if (idev)
1582                 mtu = idev->cnf.mtu6;
1583         rcu_read_unlock();
1584
1585 out:
1586         return min_t(unsigned int, mtu, IP6_MAX_MTU);
1587 }
1588
1589 static struct dst_entry *icmp6_dst_gc_list;
1590 static DEFINE_SPINLOCK(icmp6_dst_lock);
1591
1592 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
1593                                   struct flowi6 *fl6)
1594 {
1595         struct dst_entry *dst;
1596         struct rt6_info *rt;
1597         struct inet6_dev *idev = in6_dev_get(dev);
1598         struct net *net = dev_net(dev);
1599
1600         if (unlikely(!idev))
1601                 return ERR_PTR(-ENODEV);
1602
1603         rt = ip6_dst_alloc(net, dev, 0);
1604         if (unlikely(!rt)) {
1605                 in6_dev_put(idev);
1606                 dst = ERR_PTR(-ENOMEM);
1607                 goto out;
1608         }
1609
1610         rt->dst.flags |= DST_HOST;
1611         rt->dst.output  = ip6_output;
1612         atomic_set(&rt->dst.__refcnt, 1);
1613         rt->rt6i_gateway  = fl6->daddr;
1614         rt->rt6i_dst.addr = fl6->daddr;
1615         rt->rt6i_dst.plen = 128;
1616         rt->rt6i_idev     = idev;
1617         dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
1618
1619         spin_lock_bh(&icmp6_dst_lock);
1620         rt->dst.next = icmp6_dst_gc_list;
1621         icmp6_dst_gc_list = &rt->dst;
1622         spin_unlock_bh(&icmp6_dst_lock);
1623
1624         fib6_force_start_gc(net);
1625
1626         dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
1627
1628 out:
1629         return dst;
1630 }
1631
1632 int icmp6_dst_gc(void)
1633 {
1634         struct dst_entry *dst, **pprev;
1635         int more = 0;
1636
1637         spin_lock_bh(&icmp6_dst_lock);
1638         pprev = &icmp6_dst_gc_list;
1639
1640         while ((dst = *pprev) != NULL) {
1641                 if (!atomic_read(&dst->__refcnt)) {
1642                         *pprev = dst->next;
1643                         dst_free(dst);
1644                 } else {
1645                         pprev = &dst->next;
1646                         ++more;
1647                 }
1648         }
1649
1650         spin_unlock_bh(&icmp6_dst_lock);
1651
1652         return more;
1653 }
1654
1655 static void icmp6_clean_all(int (*func)(struct rt6_info *rt, void *arg),
1656                             void *arg)
1657 {
1658         struct dst_entry *dst, **pprev;
1659
1660         spin_lock_bh(&icmp6_dst_lock);
1661         pprev = &icmp6_dst_gc_list;
1662         while ((dst = *pprev) != NULL) {
1663                 struct rt6_info *rt = (struct rt6_info *) dst;
1664                 if (func(rt, arg)) {
1665                         *pprev = dst->next;
1666                         dst_free(dst);
1667                 } else {
1668                         pprev = &dst->next;
1669                 }
1670         }
1671         spin_unlock_bh(&icmp6_dst_lock);
1672 }
1673
1674 static int ip6_dst_gc(struct dst_ops *ops)
1675 {
1676         struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
1677         int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1678         int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1679         int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1680         int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1681         unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1682         int entries;
1683
1684         entries = dst_entries_get_fast(ops);
1685         if (time_after(rt_last_gc + rt_min_interval, jiffies) &&
1686             entries <= rt_max_size)
1687                 goto out;
1688
1689         net->ipv6.ip6_rt_gc_expire++;
1690         fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net, true);
1691         entries = dst_entries_get_slow(ops);
1692         if (entries < ops->gc_thresh)
1693                 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1694 out:
1695         net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1696         return entries > rt_max_size;
1697 }
1698
1699 static int ip6_convert_metrics(struct mx6_config *mxc,
1700                                const struct fib6_config *cfg)
1701 {
1702         bool ecn_ca = false;
1703         struct nlattr *nla;
1704         int remaining;
1705         u32 *mp;
1706
1707         if (!cfg->fc_mx)
1708                 return 0;
1709
1710         mp = kzalloc(sizeof(u32) * RTAX_MAX, GFP_KERNEL);
1711         if (unlikely(!mp))
1712                 return -ENOMEM;
1713
1714         nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1715                 int type = nla_type(nla);
1716                 u32 val;
1717
1718                 if (!type)
1719                         continue;
1720                 if (unlikely(type > RTAX_MAX))
1721                         goto err;
1722
1723                 if (type == RTAX_CC_ALGO) {
1724                         char tmp[TCP_CA_NAME_MAX];
1725
1726                         nla_strlcpy(tmp, nla, sizeof(tmp));
1727                         val = tcp_ca_get_key_by_name(tmp, &ecn_ca);
1728                         if (val == TCP_CA_UNSPEC)
1729                                 goto err;
1730                 } else {
1731                         val = nla_get_u32(nla);
1732                 }
1733                 if (type == RTAX_FEATURES && (val & ~RTAX_FEATURE_MASK))
1734                         goto err;
1735
1736                 mp[type - 1] = val;
1737                 __set_bit(type - 1, mxc->mx_valid);
1738         }
1739
1740         if (ecn_ca) {
1741                 __set_bit(RTAX_FEATURES - 1, mxc->mx_valid);
1742                 mp[RTAX_FEATURES - 1] |= DST_FEATURE_ECN_CA;
1743         }
1744
1745         mxc->mx = mp;
1746         return 0;
1747  err:
1748         kfree(mp);
1749         return -EINVAL;
1750 }
1751
1752 int ip6_route_info_create(struct fib6_config *cfg, struct rt6_info **rt_ret)
1753 {
1754         int err;
1755         struct net *net = cfg->fc_nlinfo.nl_net;
1756         struct rt6_info *rt = NULL;
1757         struct net_device *dev = NULL;
1758         struct inet6_dev *idev = NULL;
1759         struct fib6_table *table;
1760         int addr_type;
1761
1762         if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1763                 return -EINVAL;
1764 #ifndef CONFIG_IPV6_SUBTREES
1765         if (cfg->fc_src_len)
1766                 return -EINVAL;
1767 #endif
1768         if (cfg->fc_ifindex) {
1769                 err = -ENODEV;
1770                 dev = dev_get_by_index(net, cfg->fc_ifindex);
1771                 if (!dev)
1772                         goto out;
1773                 idev = in6_dev_get(dev);
1774                 if (!idev)
1775                         goto out;
1776         }
1777
1778         if (cfg->fc_metric == 0)
1779                 cfg->fc_metric = IP6_RT_PRIO_USER;
1780
1781         err = -ENOBUFS;
1782         if (cfg->fc_nlinfo.nlh &&
1783             !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
1784                 table = fib6_get_table(net, cfg->fc_table);
1785                 if (!table) {
1786                         pr_warn("NLM_F_CREATE should be specified when creating new route\n");
1787                         table = fib6_new_table(net, cfg->fc_table);
1788                 }
1789         } else {
1790                 table = fib6_new_table(net, cfg->fc_table);
1791         }
1792
1793         if (!table)
1794                 goto out;
1795
1796         rt = ip6_dst_alloc(net, NULL,
1797                            (cfg->fc_flags & RTF_ADDRCONF) ? 0 : DST_NOCOUNT);
1798
1799         if (!rt) {
1800                 err = -ENOMEM;
1801                 goto out;
1802         }
1803
1804         if (cfg->fc_flags & RTF_EXPIRES)
1805                 rt6_set_expires(rt, jiffies +
1806                                 clock_t_to_jiffies(cfg->fc_expires));
1807         else
1808                 rt6_clean_expires(rt);
1809
1810         if (cfg->fc_protocol == RTPROT_UNSPEC)
1811                 cfg->fc_protocol = RTPROT_BOOT;
1812         rt->rt6i_protocol = cfg->fc_protocol;
1813
1814         addr_type = ipv6_addr_type(&cfg->fc_dst);
1815
1816         if (addr_type & IPV6_ADDR_MULTICAST)
1817                 rt->dst.input = ip6_mc_input;
1818         else if (cfg->fc_flags & RTF_LOCAL)
1819                 rt->dst.input = ip6_input;
1820         else
1821                 rt->dst.input = ip6_forward;
1822
1823         rt->dst.output = ip6_output;
1824
1825         if (cfg->fc_encap) {
1826                 struct lwtunnel_state *lwtstate;
1827
1828                 err = lwtunnel_build_state(dev, cfg->fc_encap_type,
1829                                            cfg->fc_encap, AF_INET6, cfg,
1830                                            &lwtstate);
1831                 if (err)
1832                         goto out;
1833                 rt->dst.lwtstate = lwtstate_get(lwtstate);
1834                 if (lwtunnel_output_redirect(rt->dst.lwtstate)) {
1835                         rt->dst.lwtstate->orig_output = rt->dst.output;
1836                         rt->dst.output = lwtunnel_output;
1837                 }
1838                 if (lwtunnel_input_redirect(rt->dst.lwtstate)) {
1839                         rt->dst.lwtstate->orig_input = rt->dst.input;
1840                         rt->dst.input = lwtunnel_input;
1841                 }
1842         }
1843
1844         ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1845         rt->rt6i_dst.plen = cfg->fc_dst_len;
1846         if (rt->rt6i_dst.plen == 128)
1847                 rt->dst.flags |= DST_HOST;
1848
1849 #ifdef CONFIG_IPV6_SUBTREES
1850         ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1851         rt->rt6i_src.plen = cfg->fc_src_len;
1852 #endif
1853
1854         rt->rt6i_metric = cfg->fc_metric;
1855
1856         /* We cannot add true routes via loopback here,
1857            they would result in kernel looping; promote them to reject routes
1858          */
1859         if ((cfg->fc_flags & RTF_REJECT) ||
1860             (dev && (dev->flags & IFF_LOOPBACK) &&
1861              !(addr_type & IPV6_ADDR_LOOPBACK) &&
1862              !(cfg->fc_flags & RTF_LOCAL))) {
1863                 /* hold loopback dev/idev if we haven't done so. */
1864                 if (dev != net->loopback_dev) {
1865                         if (dev) {
1866                                 dev_put(dev);
1867                                 in6_dev_put(idev);
1868                         }
1869                         dev = net->loopback_dev;
1870                         dev_hold(dev);
1871                         idev = in6_dev_get(dev);
1872                         if (!idev) {
1873                                 err = -ENODEV;
1874                                 goto out;
1875                         }
1876                 }
1877                 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1878                 switch (cfg->fc_type) {
1879                 case RTN_BLACKHOLE:
1880                         rt->dst.error = -EINVAL;
1881                         rt->dst.output = dst_discard_sk;
1882                         rt->dst.input = dst_discard;
1883                         break;
1884                 case RTN_PROHIBIT:
1885                         rt->dst.error = -EACCES;
1886                         rt->dst.output = ip6_pkt_prohibit_out;
1887                         rt->dst.input = ip6_pkt_prohibit;
1888                         break;
1889                 case RTN_THROW:
1890                 case RTN_UNREACHABLE:
1891                 default:
1892                         rt->dst.error = (cfg->fc_type == RTN_THROW) ? -EAGAIN
1893                                         : (cfg->fc_type == RTN_UNREACHABLE)
1894                                         ? -EHOSTUNREACH : -ENETUNREACH;
1895                         rt->dst.output = ip6_pkt_discard_out;
1896                         rt->dst.input = ip6_pkt_discard;
1897                         break;
1898                 }
1899                 goto install_route;
1900         }
1901
1902         if (cfg->fc_flags & RTF_GATEWAY) {
1903                 const struct in6_addr *gw_addr;
1904                 int gwa_type;
1905
1906                 gw_addr = &cfg->fc_gateway;
1907                 gwa_type = ipv6_addr_type(gw_addr);
1908
1909                 /* if gw_addr is local we will fail to detect this in case
1910                  * address is still TENTATIVE (DAD in progress). rt6_lookup()
1911                  * will return already-added prefix route via interface that
1912                  * prefix route was assigned to, which might be non-loopback.
1913                  */
1914                 err = -EINVAL;
1915                 if (ipv6_chk_addr_and_flags(net, gw_addr,
1916                                             gwa_type & IPV6_ADDR_LINKLOCAL ?
1917                                             dev : NULL, 0, 0))
1918                         goto out;
1919
1920                 rt->rt6i_gateway = *gw_addr;
1921
1922                 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1923                         struct rt6_info *grt;
1924
1925                         /* IPv6 strictly inhibits using not link-local
1926                            addresses as nexthop address.
1927                            Otherwise, router will not able to send redirects.
1928                            It is very good, but in some (rare!) circumstances
1929                            (SIT, PtP, NBMA NOARP links) it is handy to allow
1930                            some exceptions. --ANK
1931                          */
1932                         if (!(gwa_type & IPV6_ADDR_UNICAST))
1933                                 goto out;
1934
1935                         grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, 1);
1936
1937                         err = -EHOSTUNREACH;
1938                         if (!grt)
1939                                 goto out;
1940                         if (dev) {
1941                                 if (dev != grt->dst.dev) {
1942                                         ip6_rt_put(grt);
1943                                         goto out;
1944                                 }
1945                         } else {
1946                                 dev = grt->dst.dev;
1947                                 idev = grt->rt6i_idev;
1948                                 dev_hold(dev);
1949                                 in6_dev_hold(grt->rt6i_idev);
1950                         }
1951                         if (!(grt->rt6i_flags & RTF_GATEWAY))
1952                                 err = 0;
1953                         ip6_rt_put(grt);
1954
1955                         if (err)
1956                                 goto out;
1957                 }
1958                 err = -EINVAL;
1959                 if (!dev || (dev->flags & IFF_LOOPBACK))
1960                         goto out;
1961         }
1962
1963         err = -ENODEV;
1964         if (!dev)
1965                 goto out;
1966
1967         if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
1968                 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
1969                         err = -EINVAL;
1970                         goto out;
1971                 }
1972                 rt->rt6i_prefsrc.addr = cfg->fc_prefsrc;
1973                 rt->rt6i_prefsrc.plen = 128;
1974         } else
1975                 rt->rt6i_prefsrc.plen = 0;
1976
1977         rt->rt6i_flags = cfg->fc_flags;
1978
1979 install_route:
1980         rt->dst.dev = dev;
1981         rt->rt6i_idev = idev;
1982         rt->rt6i_table = table;
1983
1984         cfg->fc_nlinfo.nl_net = dev_net(dev);
1985
1986         *rt_ret = rt;
1987
1988         return 0;
1989 out:
1990         if (dev)
1991                 dev_put(dev);
1992         if (idev)
1993                 in6_dev_put(idev);
1994         if (rt)
1995                 dst_free(&rt->dst);
1996
1997         *rt_ret = NULL;
1998
1999         return err;
2000 }
2001
2002 int ip6_route_add(struct fib6_config *cfg)
2003 {
2004         struct mx6_config mxc = { .mx = NULL, };
2005         struct rt6_info *rt = NULL;
2006         int err;
2007
2008         err = ip6_route_info_create(cfg, &rt);
2009         if (err)
2010                 goto out;
2011
2012         err = ip6_convert_metrics(&mxc, cfg);
2013         if (err)
2014                 goto out;
2015
2016         err = __ip6_ins_rt(rt, &cfg->fc_nlinfo, &mxc);
2017
2018         kfree(mxc.mx);
2019
2020         return err;
2021 out:
2022         if (rt)
2023                 dst_free(&rt->dst);
2024
2025         return err;
2026 }
2027
2028 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
2029 {
2030         int err;
2031         struct fib6_table *table;
2032         struct net *net = dev_net(rt->dst.dev);
2033
2034         if (rt == net->ipv6.ip6_null_entry ||
2035             rt->dst.flags & DST_NOCACHE) {
2036                 err = -ENOENT;
2037                 goto out;
2038         }
2039
2040         table = rt->rt6i_table;
2041         write_lock_bh(&table->tb6_lock);
2042         err = fib6_del(rt, info);
2043         write_unlock_bh(&table->tb6_lock);
2044
2045 out:
2046         ip6_rt_put(rt);
2047         return err;
2048 }
2049
2050 int ip6_del_rt(struct rt6_info *rt)
2051 {
2052         struct nl_info info = {
2053                 .nl_net = dev_net(rt->dst.dev),
2054         };
2055         return __ip6_del_rt(rt, &info);
2056 }
2057
2058 static int ip6_route_del(struct fib6_config *cfg)
2059 {
2060         struct fib6_table *table;
2061         struct fib6_node *fn;
2062         struct rt6_info *rt;
2063         int err = -ESRCH;
2064
2065         table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
2066         if (!table)
2067                 return err;
2068
2069         read_lock_bh(&table->tb6_lock);
2070
2071         fn = fib6_locate(&table->tb6_root,
2072                          &cfg->fc_dst, cfg->fc_dst_len,
2073                          &cfg->fc_src, cfg->fc_src_len);
2074
2075         if (fn) {
2076                 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2077                         if ((rt->rt6i_flags & RTF_CACHE) &&
2078                             !(cfg->fc_flags & RTF_CACHE))
2079                                 continue;
2080                         if (cfg->fc_ifindex &&
2081                             (!rt->dst.dev ||
2082                              rt->dst.dev->ifindex != cfg->fc_ifindex))
2083                                 continue;
2084                         if (cfg->fc_flags & RTF_GATEWAY &&
2085                             !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
2086                                 continue;
2087                         if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
2088                                 continue;
2089                         dst_hold(&rt->dst);
2090                         read_unlock_bh(&table->tb6_lock);
2091
2092                         return __ip6_del_rt(rt, &cfg->fc_nlinfo);
2093                 }
2094         }
2095         read_unlock_bh(&table->tb6_lock);
2096
2097         return err;
2098 }
2099
2100 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
2101 {
2102         struct net *net = dev_net(skb->dev);
2103         struct netevent_redirect netevent;
2104         struct rt6_info *rt, *nrt = NULL;
2105         struct ndisc_options ndopts;
2106         struct inet6_dev *in6_dev;
2107         struct neighbour *neigh;
2108         struct rd_msg *msg;
2109         int optlen, on_link;
2110         u8 *lladdr;
2111
2112         optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
2113         optlen -= sizeof(*msg);
2114
2115         if (optlen < 0) {
2116                 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
2117                 return;
2118         }
2119
2120         msg = (struct rd_msg *)icmp6_hdr(skb);
2121
2122         if (ipv6_addr_is_multicast(&msg->dest)) {
2123                 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
2124                 return;
2125         }
2126
2127         on_link = 0;
2128         if (ipv6_addr_equal(&msg->dest, &msg->target)) {
2129                 on_link = 1;
2130         } else if (ipv6_addr_type(&msg->target) !=
2131                    (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
2132                 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
2133                 return;
2134         }
2135
2136         in6_dev = __in6_dev_get(skb->dev);
2137         if (!in6_dev)
2138                 return;
2139         if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
2140                 return;
2141
2142         /* RFC2461 8.1:
2143          *      The IP source address of the Redirect MUST be the same as the current
2144          *      first-hop router for the specified ICMP Destination Address.
2145          */
2146
2147         if (!ndisc_parse_options(msg->opt, optlen, &ndopts)) {
2148                 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
2149                 return;
2150         }
2151
2152         lladdr = NULL;
2153         if (ndopts.nd_opts_tgt_lladdr) {
2154                 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
2155                                              skb->dev);
2156                 if (!lladdr) {
2157                         net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
2158                         return;
2159                 }
2160         }
2161
2162         rt = (struct rt6_info *) dst;
2163         if (rt == net->ipv6.ip6_null_entry) {
2164                 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
2165                 return;
2166         }
2167
2168         /* Redirect received -> path was valid.
2169          * Look, redirects are sent only in response to data packets,
2170          * so that this nexthop apparently is reachable. --ANK
2171          */
2172         dst_confirm(&rt->dst);
2173
2174         neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
2175         if (!neigh)
2176                 return;
2177
2178         /*
2179          *      We have finally decided to accept it.
2180          */
2181
2182         neigh_update(neigh, lladdr, NUD_STALE,
2183                      NEIGH_UPDATE_F_WEAK_OVERRIDE|
2184                      NEIGH_UPDATE_F_OVERRIDE|
2185                      (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
2186                                      NEIGH_UPDATE_F_ISROUTER))
2187                      );
2188
2189         nrt = ip6_rt_cache_alloc(rt, &msg->dest, NULL);
2190         if (!nrt)
2191                 goto out;
2192
2193         nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
2194         if (on_link)
2195                 nrt->rt6i_flags &= ~RTF_GATEWAY;
2196
2197         nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
2198
2199         if (ip6_ins_rt(nrt))
2200                 goto out;
2201
2202         netevent.old = &rt->dst;
2203         netevent.new = &nrt->dst;
2204         netevent.daddr = &msg->dest;
2205         netevent.neigh = neigh;
2206         call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
2207
2208         if (rt->rt6i_flags & RTF_CACHE) {
2209                 rt = (struct rt6_info *) dst_clone(&rt->dst);
2210                 ip6_del_rt(rt);
2211         }
2212
2213 out:
2214         neigh_release(neigh);
2215 }
2216
2217 /*
2218  *      Misc support functions
2219  */
2220
2221 static void rt6_set_from(struct rt6_info *rt, struct rt6_info *from)
2222 {
2223         BUG_ON(from->dst.from);
2224
2225         rt->rt6i_flags &= ~RTF_EXPIRES;
2226         dst_hold(&from->dst);
2227         rt->dst.from = &from->dst;
2228         dst_init_metrics(&rt->dst, dst_metrics_ptr(&from->dst), true);
2229 }
2230
2231 static void ip6_rt_copy_init(struct rt6_info *rt, struct rt6_info *ort)
2232 {
2233         rt->dst.input = ort->dst.input;
2234         rt->dst.output = ort->dst.output;
2235         rt->rt6i_dst = ort->rt6i_dst;
2236         rt->dst.error = ort->dst.error;
2237         rt->rt6i_idev = ort->rt6i_idev;
2238         if (rt->rt6i_idev)
2239                 in6_dev_hold(rt->rt6i_idev);
2240         rt->dst.lastuse = jiffies;
2241         rt->rt6i_gateway = ort->rt6i_gateway;
2242         rt->rt6i_flags = ort->rt6i_flags;
2243         rt6_set_from(rt, ort);
2244         rt->rt6i_metric = ort->rt6i_metric;
2245 #ifdef CONFIG_IPV6_SUBTREES
2246         rt->rt6i_src = ort->rt6i_src;
2247 #endif
2248         rt->rt6i_prefsrc = ort->rt6i_prefsrc;
2249         rt->rt6i_table = ort->rt6i_table;
2250         rt->dst.lwtstate = lwtstate_get(ort->dst.lwtstate);
2251 }
2252
2253 #ifdef CONFIG_IPV6_ROUTE_INFO
2254 static struct rt6_info *rt6_get_route_info(struct net *net,
2255                                            const struct in6_addr *prefix, int prefixlen,
2256                                            const struct in6_addr *gwaddr, int ifindex)
2257 {
2258         struct fib6_node *fn;
2259         struct rt6_info *rt = NULL;
2260         struct fib6_table *table;
2261
2262         table = fib6_get_table(net, RT6_TABLE_INFO);
2263         if (!table)
2264                 return NULL;
2265
2266         read_lock_bh(&table->tb6_lock);
2267         fn = fib6_locate(&table->tb6_root, prefix, prefixlen, NULL, 0);
2268         if (!fn)
2269                 goto out;
2270
2271         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2272                 if (rt->dst.dev->ifindex != ifindex)
2273                         continue;
2274                 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
2275                         continue;
2276                 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
2277                         continue;
2278                 dst_hold(&rt->dst);
2279                 break;
2280         }
2281 out:
2282         read_unlock_bh(&table->tb6_lock);
2283         return rt;
2284 }
2285
2286 static struct rt6_info *rt6_add_route_info(struct net *net,
2287                                            const struct in6_addr *prefix, int prefixlen,
2288                                            const struct in6_addr *gwaddr, int ifindex,
2289                                            unsigned int pref)
2290 {
2291         struct fib6_config cfg = {
2292                 .fc_table       = RT6_TABLE_INFO,
2293                 .fc_metric      = IP6_RT_PRIO_USER,
2294                 .fc_ifindex     = ifindex,
2295                 .fc_dst_len     = prefixlen,
2296                 .fc_flags       = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
2297                                   RTF_UP | RTF_PREF(pref),
2298                 .fc_nlinfo.portid = 0,
2299                 .fc_nlinfo.nlh = NULL,
2300                 .fc_nlinfo.nl_net = net,
2301         };
2302
2303         cfg.fc_dst = *prefix;
2304         cfg.fc_gateway = *gwaddr;
2305
2306         /* We should treat it as a default route if prefix length is 0. */
2307         if (!prefixlen)
2308                 cfg.fc_flags |= RTF_DEFAULT;
2309
2310         ip6_route_add(&cfg);
2311
2312         return rt6_get_route_info(net, prefix, prefixlen, gwaddr, ifindex);
2313 }
2314 #endif
2315
2316 struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
2317 {
2318         struct rt6_info *rt;
2319         struct fib6_table *table;
2320
2321         table = fib6_get_table(dev_net(dev), RT6_TABLE_DFLT);
2322         if (!table)
2323                 return NULL;
2324
2325         read_lock_bh(&table->tb6_lock);
2326         for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2327                 if (dev == rt->dst.dev &&
2328                     ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
2329                     ipv6_addr_equal(&rt->rt6i_gateway, addr))
2330                         break;
2331         }
2332         if (rt)
2333                 dst_hold(&rt->dst);
2334         read_unlock_bh(&table->tb6_lock);
2335         return rt;
2336 }
2337
2338 struct rt6_info *rt6_add_dflt_router(const struct in6_addr *gwaddr,
2339                                      struct net_device *dev,
2340                                      unsigned int pref)
2341 {
2342         struct fib6_config cfg = {
2343                 .fc_table       = RT6_TABLE_DFLT,
2344                 .fc_metric      = IP6_RT_PRIO_USER,
2345                 .fc_ifindex     = dev->ifindex,
2346                 .fc_flags       = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
2347                                   RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
2348                 .fc_nlinfo.portid = 0,
2349                 .fc_nlinfo.nlh = NULL,
2350                 .fc_nlinfo.nl_net = dev_net(dev),
2351         };
2352
2353         cfg.fc_gateway = *gwaddr;
2354
2355         ip6_route_add(&cfg);
2356
2357         return rt6_get_dflt_router(gwaddr, dev);
2358 }
2359
2360 void rt6_purge_dflt_routers(struct net *net)
2361 {
2362         struct rt6_info *rt;
2363         struct fib6_table *table;
2364
2365         /* NOTE: Keep consistent with rt6_get_dflt_router */
2366         table = fib6_get_table(net, RT6_TABLE_DFLT);
2367         if (!table)
2368                 return;
2369
2370 restart:
2371         read_lock_bh(&table->tb6_lock);
2372         for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2373                 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
2374                     (!rt->rt6i_idev || rt->rt6i_idev->cnf.accept_ra != 2)) {
2375                         dst_hold(&rt->dst);
2376                         read_unlock_bh(&table->tb6_lock);
2377                         ip6_del_rt(rt);
2378                         goto restart;
2379                 }
2380         }
2381         read_unlock_bh(&table->tb6_lock);
2382 }
2383
2384 static void rtmsg_to_fib6_config(struct net *net,
2385                                  struct in6_rtmsg *rtmsg,
2386                                  struct fib6_config *cfg)
2387 {
2388         memset(cfg, 0, sizeof(*cfg));
2389
2390         cfg->fc_table = RT6_TABLE_MAIN;
2391         cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
2392         cfg->fc_metric = rtmsg->rtmsg_metric;
2393         cfg->fc_expires = rtmsg->rtmsg_info;
2394         cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
2395         cfg->fc_src_len = rtmsg->rtmsg_src_len;
2396         cfg->fc_flags = rtmsg->rtmsg_flags;
2397
2398         cfg->fc_nlinfo.nl_net = net;
2399
2400         cfg->fc_dst = rtmsg->rtmsg_dst;
2401         cfg->fc_src = rtmsg->rtmsg_src;
2402         cfg->fc_gateway = rtmsg->rtmsg_gateway;
2403 }
2404
2405 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
2406 {
2407         struct fib6_config cfg;
2408         struct in6_rtmsg rtmsg;
2409         int err;
2410
2411         switch (cmd) {
2412         case SIOCADDRT:         /* Add a route */
2413         case SIOCDELRT:         /* Delete a route */
2414                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2415                         return -EPERM;
2416                 err = copy_from_user(&rtmsg, arg,
2417                                      sizeof(struct in6_rtmsg));
2418                 if (err)
2419                         return -EFAULT;
2420
2421                 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
2422
2423                 rtnl_lock();
2424                 switch (cmd) {
2425                 case SIOCADDRT:
2426                         err = ip6_route_add(&cfg);
2427                         break;
2428                 case SIOCDELRT:
2429                         err = ip6_route_del(&cfg);
2430                         break;
2431                 default:
2432                         err = -EINVAL;
2433                 }
2434                 rtnl_unlock();
2435
2436                 return err;
2437         }
2438
2439         return -EINVAL;
2440 }
2441
2442 /*
2443  *      Drop the packet on the floor
2444  */
2445
2446 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
2447 {
2448         int type;
2449         struct dst_entry *dst = skb_dst(skb);
2450         switch (ipstats_mib_noroutes) {
2451         case IPSTATS_MIB_INNOROUTES:
2452                 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
2453                 if (type == IPV6_ADDR_ANY) {
2454                         IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2455                                       IPSTATS_MIB_INADDRERRORS);
2456                         break;
2457                 }
2458                 /* FALLTHROUGH */
2459         case IPSTATS_MIB_OUTNOROUTES:
2460                 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2461                               ipstats_mib_noroutes);
2462                 break;
2463         }
2464         icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
2465         kfree_skb(skb);
2466         return 0;
2467 }
2468
2469 static int ip6_pkt_discard(struct sk_buff *skb)
2470 {
2471         return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
2472 }
2473
2474 static int ip6_pkt_discard_out(struct sock *sk, struct sk_buff *skb)
2475 {
2476         skb->dev = skb_dst(skb)->dev;
2477         return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
2478 }
2479
2480 static int ip6_pkt_prohibit(struct sk_buff *skb)
2481 {
2482         return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
2483 }
2484
2485 static int ip6_pkt_prohibit_out(struct sock *sk, struct sk_buff *skb)
2486 {
2487         skb->dev = skb_dst(skb)->dev;
2488         return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
2489 }
2490
2491 /*
2492  *      Allocate a dst for local (unicast / anycast) address.
2493  */
2494
2495 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
2496                                     const struct in6_addr *addr,
2497                                     bool anycast)
2498 {
2499         struct net *net = dev_net(idev->dev);
2500         struct rt6_info *rt = ip6_dst_alloc(net, net->loopback_dev,
2501                                             DST_NOCOUNT);
2502         if (!rt)
2503                 return ERR_PTR(-ENOMEM);
2504
2505         in6_dev_hold(idev);
2506
2507         rt->dst.flags |= DST_HOST;
2508         rt->dst.input = ip6_input;
2509         rt->dst.output = ip6_output;
2510         rt->rt6i_idev = idev;
2511
2512         rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
2513         if (anycast)
2514                 rt->rt6i_flags |= RTF_ANYCAST;
2515         else
2516                 rt->rt6i_flags |= RTF_LOCAL;
2517
2518         rt->rt6i_gateway  = *addr;
2519         rt->rt6i_dst.addr = *addr;
2520         rt->rt6i_dst.plen = 128;
2521         rt->rt6i_table = fib6_get_table(net, RT6_TABLE_LOCAL);
2522         rt->dst.flags |= DST_NOCACHE;
2523
2524         atomic_set(&rt->dst.__refcnt, 1);
2525
2526         return rt;
2527 }
2528
2529 int ip6_route_get_saddr(struct net *net,
2530                         struct rt6_info *rt,
2531                         const struct in6_addr *daddr,
2532                         unsigned int prefs,
2533                         struct in6_addr *saddr)
2534 {
2535         struct inet6_dev *idev =
2536                 rt ? ip6_dst_idev((struct dst_entry *)rt) : NULL;
2537         int err = 0;
2538         if (rt && rt->rt6i_prefsrc.plen)
2539                 *saddr = rt->rt6i_prefsrc.addr;
2540         else
2541                 err = ipv6_dev_get_saddr(net, idev ? idev->dev : NULL,
2542                                          daddr, prefs, saddr);
2543         return err;
2544 }
2545
2546 /* remove deleted ip from prefsrc entries */
2547 struct arg_dev_net_ip {
2548         struct net_device *dev;
2549         struct net *net;
2550         struct in6_addr *addr;
2551 };
2552
2553 static int fib6_remove_prefsrc(struct rt6_info *rt, void *arg)
2554 {
2555         struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
2556         struct net *net = ((struct arg_dev_net_ip *)arg)->net;
2557         struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
2558
2559         if (((void *)rt->dst.dev == dev || !dev) &&
2560             rt != net->ipv6.ip6_null_entry &&
2561             ipv6_addr_equal(addr, &rt->rt6i_prefsrc.addr)) {
2562                 /* remove prefsrc entry */
2563                 rt->rt6i_prefsrc.plen = 0;
2564         }
2565         return 0;
2566 }
2567
2568 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
2569 {
2570         struct net *net = dev_net(ifp->idev->dev);
2571         struct arg_dev_net_ip adni = {
2572                 .dev = ifp->idev->dev,
2573                 .net = net,
2574                 .addr = &ifp->addr,
2575         };
2576         fib6_clean_all(net, fib6_remove_prefsrc, &adni);
2577 }
2578
2579 #define RTF_RA_ROUTER           (RTF_ADDRCONF | RTF_DEFAULT | RTF_GATEWAY)
2580 #define RTF_CACHE_GATEWAY       (RTF_GATEWAY | RTF_CACHE)
2581
2582 /* Remove routers and update dst entries when gateway turn into host. */
2583 static int fib6_clean_tohost(struct rt6_info *rt, void *arg)
2584 {
2585         struct in6_addr *gateway = (struct in6_addr *)arg;
2586
2587         if ((((rt->rt6i_flags & RTF_RA_ROUTER) == RTF_RA_ROUTER) ||
2588              ((rt->rt6i_flags & RTF_CACHE_GATEWAY) == RTF_CACHE_GATEWAY)) &&
2589              ipv6_addr_equal(gateway, &rt->rt6i_gateway)) {
2590                 return -1;
2591         }
2592         return 0;
2593 }
2594
2595 void rt6_clean_tohost(struct net *net, struct in6_addr *gateway)
2596 {
2597         fib6_clean_all(net, fib6_clean_tohost, gateway);
2598 }
2599
2600 struct arg_dev_net {
2601         struct net_device *dev;
2602         struct net *net;
2603 };
2604
2605 static int fib6_ifdown(struct rt6_info *rt, void *arg)
2606 {
2607         const struct arg_dev_net *adn = arg;
2608         const struct net_device *dev = adn->dev;
2609
2610         if ((rt->dst.dev == dev || !dev) &&
2611             rt != adn->net->ipv6.ip6_null_entry)
2612                 return -1;
2613
2614         return 0;
2615 }
2616
2617 void rt6_ifdown(struct net *net, struct net_device *dev)
2618 {
2619         struct arg_dev_net adn = {
2620                 .dev = dev,
2621                 .net = net,
2622         };
2623
2624         fib6_clean_all(net, fib6_ifdown, &adn);
2625         icmp6_clean_all(fib6_ifdown, &adn);
2626         if (dev)
2627                 rt6_uncached_list_flush_dev(net, dev);
2628 }
2629
2630 struct rt6_mtu_change_arg {
2631         struct net_device *dev;
2632         unsigned int mtu;
2633 };
2634
2635 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2636 {
2637         struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2638         struct inet6_dev *idev;
2639
2640         /* In IPv6 pmtu discovery is not optional,
2641            so that RTAX_MTU lock cannot disable it.
2642            We still use this lock to block changes
2643            caused by addrconf/ndisc.
2644         */
2645
2646         idev = __in6_dev_get(arg->dev);
2647         if (!idev)
2648                 return 0;
2649
2650         /* For administrative MTU increase, there is no way to discover
2651            IPv6 PMTU increase, so PMTU increase should be updated here.
2652            Since RFC 1981 doesn't include administrative MTU increase
2653            update PMTU increase is a MUST. (i.e. jumbo frame)
2654          */
2655         /*
2656            If new MTU is less than route PMTU, this new MTU will be the
2657            lowest MTU in the path, update the route PMTU to reflect PMTU
2658            decreases; if new MTU is greater than route PMTU, and the
2659            old MTU is the lowest MTU in the path, update the route PMTU
2660            to reflect the increase. In this case if the other nodes' MTU
2661            also have the lowest MTU, TOO BIG MESSAGE will be lead to
2662            PMTU discouvery.
2663          */
2664         if (rt->dst.dev == arg->dev &&
2665             !dst_metric_locked(&rt->dst, RTAX_MTU)) {
2666                 if (rt->rt6i_flags & RTF_CACHE) {
2667                         /* For RTF_CACHE with rt6i_pmtu == 0
2668                          * (i.e. a redirected route),
2669                          * the metrics of its rt->dst.from has already
2670                          * been updated.
2671                          */
2672                         if (rt->rt6i_pmtu && rt->rt6i_pmtu > arg->mtu)
2673                                 rt->rt6i_pmtu = arg->mtu;
2674                 } else if (dst_mtu(&rt->dst) >= arg->mtu ||
2675                            (dst_mtu(&rt->dst) < arg->mtu &&
2676                             dst_mtu(&rt->dst) == idev->cnf.mtu6)) {
2677                         dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
2678                 }
2679         }
2680         return 0;
2681 }
2682
2683 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
2684 {
2685         struct rt6_mtu_change_arg arg = {
2686                 .dev = dev,
2687                 .mtu = mtu,
2688         };
2689
2690         fib6_clean_all(dev_net(dev), rt6_mtu_change_route, &arg);
2691 }
2692
2693 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2694         [RTA_GATEWAY]           = { .len = sizeof(struct in6_addr) },
2695         [RTA_OIF]               = { .type = NLA_U32 },
2696         [RTA_IIF]               = { .type = NLA_U32 },
2697         [RTA_PRIORITY]          = { .type = NLA_U32 },
2698         [RTA_METRICS]           = { .type = NLA_NESTED },
2699         [RTA_MULTIPATH]         = { .len = sizeof(struct rtnexthop) },
2700         [RTA_PREF]              = { .type = NLA_U8 },
2701         [RTA_ENCAP_TYPE]        = { .type = NLA_U16 },
2702         [RTA_ENCAP]             = { .type = NLA_NESTED },
2703 };
2704
2705 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2706                               struct fib6_config *cfg)
2707 {
2708         struct rtmsg *rtm;
2709         struct nlattr *tb[RTA_MAX+1];
2710         unsigned int pref;
2711         int err;
2712
2713         err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2714         if (err < 0)
2715                 goto errout;
2716
2717         err = -EINVAL;
2718         rtm = nlmsg_data(nlh);
2719         memset(cfg, 0, sizeof(*cfg));
2720
2721         cfg->fc_table = rtm->rtm_table;
2722         cfg->fc_dst_len = rtm->rtm_dst_len;
2723         cfg->fc_src_len = rtm->rtm_src_len;
2724         cfg->fc_flags = RTF_UP;
2725         cfg->fc_protocol = rtm->rtm_protocol;
2726         cfg->fc_type = rtm->rtm_type;
2727
2728         if (rtm->rtm_type == RTN_UNREACHABLE ||
2729             rtm->rtm_type == RTN_BLACKHOLE ||
2730             rtm->rtm_type == RTN_PROHIBIT ||
2731             rtm->rtm_type == RTN_THROW)
2732                 cfg->fc_flags |= RTF_REJECT;
2733
2734         if (rtm->rtm_type == RTN_LOCAL)
2735                 cfg->fc_flags |= RTF_LOCAL;
2736
2737         if (rtm->rtm_flags & RTM_F_CLONED)
2738                 cfg->fc_flags |= RTF_CACHE;
2739
2740         cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid;
2741         cfg->fc_nlinfo.nlh = nlh;
2742         cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2743
2744         if (tb[RTA_GATEWAY]) {
2745                 cfg->fc_gateway = nla_get_in6_addr(tb[RTA_GATEWAY]);
2746                 cfg->fc_flags |= RTF_GATEWAY;
2747         }
2748
2749         if (tb[RTA_DST]) {
2750                 int plen = (rtm->rtm_dst_len + 7) >> 3;
2751
2752                 if (nla_len(tb[RTA_DST]) < plen)
2753                         goto errout;
2754
2755                 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2756         }
2757
2758         if (tb[RTA_SRC]) {
2759                 int plen = (rtm->rtm_src_len + 7) >> 3;
2760
2761                 if (nla_len(tb[RTA_SRC]) < plen)
2762                         goto errout;
2763
2764                 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2765         }
2766
2767         if (tb[RTA_PREFSRC])
2768                 cfg->fc_prefsrc = nla_get_in6_addr(tb[RTA_PREFSRC]);
2769
2770         if (tb[RTA_OIF])
2771                 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2772
2773         if (tb[RTA_PRIORITY])
2774                 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2775
2776         if (tb[RTA_METRICS]) {
2777                 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2778                 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2779         }
2780
2781         if (tb[RTA_TABLE])
2782                 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2783
2784         if (tb[RTA_MULTIPATH]) {
2785                 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
2786                 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
2787         }
2788
2789         if (tb[RTA_PREF]) {
2790                 pref = nla_get_u8(tb[RTA_PREF]);
2791                 if (pref != ICMPV6_ROUTER_PREF_LOW &&
2792                     pref != ICMPV6_ROUTER_PREF_HIGH)
2793                         pref = ICMPV6_ROUTER_PREF_MEDIUM;
2794                 cfg->fc_flags |= RTF_PREF(pref);
2795         }
2796
2797         if (tb[RTA_ENCAP])
2798                 cfg->fc_encap = tb[RTA_ENCAP];
2799
2800         if (tb[RTA_ENCAP_TYPE])
2801                 cfg->fc_encap_type = nla_get_u16(tb[RTA_ENCAP_TYPE]);
2802
2803         err = 0;
2804 errout:
2805         return err;
2806 }
2807
2808 struct rt6_nh {
2809         struct rt6_info *rt6_info;
2810         struct fib6_config r_cfg;
2811         struct mx6_config mxc;
2812         struct list_head next;
2813 };
2814
2815 static void ip6_print_replace_route_err(struct list_head *rt6_nh_list)
2816 {
2817         struct rt6_nh *nh;
2818
2819         list_for_each_entry(nh, rt6_nh_list, next) {
2820                 pr_warn("IPV6: multipath route replace failed (check consistency of installed routes): %pI6 nexthop %pI6 ifi %d\n",
2821                         &nh->r_cfg.fc_dst, &nh->r_cfg.fc_gateway,
2822                         nh->r_cfg.fc_ifindex);
2823         }
2824 }
2825
2826 static int ip6_route_info_append(struct list_head *rt6_nh_list,
2827                                  struct rt6_info *rt, struct fib6_config *r_cfg)
2828 {
2829         struct rt6_nh *nh;
2830         struct rt6_info *rtnh;
2831         int err = -EEXIST;
2832
2833         list_for_each_entry(nh, rt6_nh_list, next) {
2834                 /* check if rt6_info already exists */
2835                 rtnh = nh->rt6_info;
2836
2837                 if (rtnh->dst.dev == rt->dst.dev &&
2838                     rtnh->rt6i_idev == rt->rt6i_idev &&
2839                     ipv6_addr_equal(&rtnh->rt6i_gateway,
2840                                     &rt->rt6i_gateway))
2841                         return err;
2842         }
2843
2844         nh = kzalloc(sizeof(*nh), GFP_KERNEL);
2845         if (!nh)
2846                 return -ENOMEM;
2847         nh->rt6_info = rt;
2848         err = ip6_convert_metrics(&nh->mxc, r_cfg);
2849         if (err) {
2850                 kfree(nh);
2851                 return err;
2852         }
2853         memcpy(&nh->r_cfg, r_cfg, sizeof(*r_cfg));
2854         list_add_tail(&nh->next, rt6_nh_list);
2855
2856         return 0;
2857 }
2858
2859 static int ip6_route_multipath_add(struct fib6_config *cfg)
2860 {
2861         struct fib6_config r_cfg;
2862         struct rtnexthop *rtnh;
2863         struct rt6_info *rt;
2864         struct rt6_nh *err_nh;
2865         struct rt6_nh *nh, *nh_safe;
2866         int remaining;
2867         int attrlen;
2868         int err = 1;
2869         int nhn = 0;
2870         int replace = (cfg->fc_nlinfo.nlh &&
2871                        (cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_REPLACE));
2872         LIST_HEAD(rt6_nh_list);
2873
2874         remaining = cfg->fc_mp_len;
2875         rtnh = (struct rtnexthop *)cfg->fc_mp;
2876
2877         /* Parse a Multipath Entry and build a list (rt6_nh_list) of
2878          * rt6_info structs per nexthop
2879          */
2880         while (rtnh_ok(rtnh, remaining)) {
2881                 memcpy(&r_cfg, cfg, sizeof(*cfg));
2882                 if (rtnh->rtnh_ifindex)
2883                         r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
2884
2885                 attrlen = rtnh_attrlen(rtnh);
2886                 if (attrlen > 0) {
2887                         struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
2888
2889                         nla = nla_find(attrs, attrlen, RTA_GATEWAY);
2890                         if (nla) {
2891                                 r_cfg.fc_gateway = nla_get_in6_addr(nla);
2892                                 r_cfg.fc_flags |= RTF_GATEWAY;
2893                         }
2894                         r_cfg.fc_encap = nla_find(attrs, attrlen, RTA_ENCAP);
2895                         nla = nla_find(attrs, attrlen, RTA_ENCAP_TYPE);
2896                         if (nla)
2897                                 r_cfg.fc_encap_type = nla_get_u16(nla);
2898                 }
2899
2900                 err = ip6_route_info_create(&r_cfg, &rt);
2901                 if (err)
2902                         goto cleanup;
2903
2904                 err = ip6_route_info_append(&rt6_nh_list, rt, &r_cfg);
2905                 if (err) {
2906                         dst_free(&rt->dst);
2907                         goto cleanup;
2908                 }
2909
2910                 rtnh = rtnh_next(rtnh, &remaining);
2911         }
2912
2913         err_nh = NULL;
2914         list_for_each_entry(nh, &rt6_nh_list, next) {
2915                 err = __ip6_ins_rt(nh->rt6_info, &cfg->fc_nlinfo, &nh->mxc);
2916                 /* nh->rt6_info is used or freed at this point, reset to NULL*/
2917                 nh->rt6_info = NULL;
2918                 if (err) {
2919                         if (replace && nhn)
2920                                 ip6_print_replace_route_err(&rt6_nh_list);
2921                         err_nh = nh;
2922                         goto add_errout;
2923                 }
2924
2925                 /* Because each route is added like a single route we remove
2926                  * these flags after the first nexthop: if there is a collision,
2927                  * we have already failed to add the first nexthop:
2928                  * fib6_add_rt2node() has rejected it; when replacing, old
2929                  * nexthops have been replaced by first new, the rest should
2930                  * be added to it.
2931                  */
2932                 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~(NLM_F_EXCL |
2933                                                      NLM_F_REPLACE);
2934                 nhn++;
2935         }
2936
2937         goto cleanup;
2938
2939 add_errout:
2940         /* Delete routes that were already added */
2941         list_for_each_entry(nh, &rt6_nh_list, next) {
2942                 if (err_nh == nh)
2943                         break;
2944                 ip6_route_del(&nh->r_cfg);
2945         }
2946
2947 cleanup:
2948         list_for_each_entry_safe(nh, nh_safe, &rt6_nh_list, next) {
2949                 if (nh->rt6_info)
2950                         dst_free(&nh->rt6_info->dst);
2951                 kfree(nh->mxc.mx);
2952                 list_del(&nh->next);
2953                 kfree(nh);
2954         }
2955
2956         return err;
2957 }
2958
2959 static int ip6_route_multipath_del(struct fib6_config *cfg)
2960 {
2961         struct fib6_config r_cfg;
2962         struct rtnexthop *rtnh;
2963         int remaining;
2964         int attrlen;
2965         int err = 1, last_err = 0;
2966
2967         remaining = cfg->fc_mp_len;
2968         rtnh = (struct rtnexthop *)cfg->fc_mp;
2969
2970         /* Parse a Multipath Entry */
2971         while (rtnh_ok(rtnh, remaining)) {
2972                 memcpy(&r_cfg, cfg, sizeof(*cfg));
2973                 if (rtnh->rtnh_ifindex)
2974                         r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
2975
2976                 attrlen = rtnh_attrlen(rtnh);
2977                 if (attrlen > 0) {
2978                         struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
2979
2980                         nla = nla_find(attrs, attrlen, RTA_GATEWAY);
2981                         if (nla) {
2982                                 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
2983                                 r_cfg.fc_flags |= RTF_GATEWAY;
2984                         }
2985                 }
2986                 err = ip6_route_del(&r_cfg);
2987                 if (err)
2988                         last_err = err;
2989
2990                 rtnh = rtnh_next(rtnh, &remaining);
2991         }
2992
2993         return last_err;
2994 }
2995
2996 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh)
2997 {
2998         struct fib6_config cfg;
2999         int err;
3000
3001         err = rtm_to_fib6_config(skb, nlh, &cfg);
3002         if (err < 0)
3003                 return err;
3004
3005         if (cfg.fc_mp)
3006                 return ip6_route_multipath_del(&cfg);
3007         else
3008                 return ip6_route_del(&cfg);
3009 }
3010
3011 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh)
3012 {
3013         struct fib6_config cfg;
3014         int err;
3015
3016         err = rtm_to_fib6_config(skb, nlh, &cfg);
3017         if (err < 0)
3018                 return err;
3019
3020         if (cfg.fc_mp)
3021                 return ip6_route_multipath_add(&cfg);
3022         else
3023                 return ip6_route_add(&cfg);
3024 }
3025
3026 static inline size_t rt6_nlmsg_size(struct rt6_info *rt)
3027 {
3028         return NLMSG_ALIGN(sizeof(struct rtmsg))
3029                + nla_total_size(16) /* RTA_SRC */
3030                + nla_total_size(16) /* RTA_DST */
3031                + nla_total_size(16) /* RTA_GATEWAY */
3032                + nla_total_size(16) /* RTA_PREFSRC */
3033                + nla_total_size(4) /* RTA_TABLE */
3034                + nla_total_size(4) /* RTA_IIF */
3035                + nla_total_size(4) /* RTA_OIF */
3036                + nla_total_size(4) /* RTA_PRIORITY */
3037                + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
3038                + nla_total_size(sizeof(struct rta_cacheinfo))
3039                + nla_total_size(TCP_CA_NAME_MAX) /* RTAX_CC_ALGO */
3040                + nla_total_size(1) /* RTA_PREF */
3041                + lwtunnel_get_encap_size(rt->dst.lwtstate);
3042 }
3043
3044 static int rt6_fill_node(struct net *net,
3045                          struct sk_buff *skb, struct rt6_info *rt,
3046                          struct in6_addr *dst, struct in6_addr *src,
3047                          int iif, int type, u32 portid, u32 seq,
3048                          int prefix, int nowait, unsigned int flags)
3049 {
3050         u32 metrics[RTAX_MAX];
3051         struct rtmsg *rtm;
3052         struct nlmsghdr *nlh;
3053         long expires;
3054         u32 table;
3055
3056         if (prefix) {   /* user wants prefix routes only */
3057                 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
3058                         /* success since this is not a prefix route */
3059                         return 1;
3060                 }
3061         }
3062
3063         nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
3064         if (!nlh)
3065                 return -EMSGSIZE;
3066
3067         rtm = nlmsg_data(nlh);
3068         rtm->rtm_family = AF_INET6;
3069         rtm->rtm_dst_len = rt->rt6i_dst.plen;
3070         rtm->rtm_src_len = rt->rt6i_src.plen;
3071         rtm->rtm_tos = 0;
3072         if (rt->rt6i_table)
3073                 table = rt->rt6i_table->tb6_id;
3074         else
3075                 table = RT6_TABLE_UNSPEC;
3076         rtm->rtm_table = table;
3077         if (nla_put_u32(skb, RTA_TABLE, table))
3078                 goto nla_put_failure;
3079         if (rt->rt6i_flags & RTF_REJECT) {
3080                 switch (rt->dst.error) {
3081                 case -EINVAL:
3082                         rtm->rtm_type = RTN_BLACKHOLE;
3083                         break;
3084                 case -EACCES:
3085                         rtm->rtm_type = RTN_PROHIBIT;
3086                         break;
3087                 case -EAGAIN:
3088                         rtm->rtm_type = RTN_THROW;
3089                         break;
3090                 default:
3091                         rtm->rtm_type = RTN_UNREACHABLE;
3092                         break;
3093                 }
3094         }
3095         else if (rt->rt6i_flags & RTF_LOCAL)
3096                 rtm->rtm_type = RTN_LOCAL;
3097         else if (rt->dst.dev && (rt->dst.dev->flags & IFF_LOOPBACK))
3098                 rtm->rtm_type = RTN_LOCAL;
3099         else
3100                 rtm->rtm_type = RTN_UNICAST;
3101         rtm->rtm_flags = 0;
3102         if (!netif_carrier_ok(rt->dst.dev)) {
3103                 rtm->rtm_flags |= RTNH_F_LINKDOWN;
3104                 if (rt->rt6i_idev->cnf.ignore_routes_with_linkdown)
3105                         rtm->rtm_flags |= RTNH_F_DEAD;
3106         }
3107         rtm->rtm_scope = RT_SCOPE_UNIVERSE;
3108         rtm->rtm_protocol = rt->rt6i_protocol;
3109         if (rt->rt6i_flags & RTF_DYNAMIC)
3110                 rtm->rtm_protocol = RTPROT_REDIRECT;
3111         else if (rt->rt6i_flags & RTF_ADDRCONF) {
3112                 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ROUTEINFO))
3113                         rtm->rtm_protocol = RTPROT_RA;
3114                 else
3115                         rtm->rtm_protocol = RTPROT_KERNEL;
3116         }
3117
3118         if (rt->rt6i_flags & RTF_CACHE)
3119                 rtm->rtm_flags |= RTM_F_CLONED;
3120
3121         if (dst) {
3122                 if (nla_put_in6_addr(skb, RTA_DST, dst))
3123                         goto nla_put_failure;
3124                 rtm->rtm_dst_len = 128;
3125         } else if (rtm->rtm_dst_len)
3126                 if (nla_put_in6_addr(skb, RTA_DST, &rt->rt6i_dst.addr))
3127                         goto nla_put_failure;
3128 #ifdef CONFIG_IPV6_SUBTREES
3129         if (src) {
3130                 if (nla_put_in6_addr(skb, RTA_SRC, src))
3131                         goto nla_put_failure;
3132                 rtm->rtm_src_len = 128;
3133         } else if (rtm->rtm_src_len &&
3134                    nla_put_in6_addr(skb, RTA_SRC, &rt->rt6i_src.addr))
3135                 goto nla_put_failure;
3136 #endif
3137         if (iif) {
3138 #ifdef CONFIG_IPV6_MROUTE
3139                 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
3140                         int err = ip6mr_get_route(net, skb, rtm, nowait);
3141                         if (err <= 0) {
3142                                 if (!nowait) {
3143                                         if (err == 0)
3144                                                 return 0;
3145                                         goto nla_put_failure;
3146                                 } else {
3147                                         if (err == -EMSGSIZE)
3148                                                 goto nla_put_failure;
3149                                 }
3150                         }
3151                 } else
3152 #endif
3153                         if (nla_put_u32(skb, RTA_IIF, iif))
3154                                 goto nla_put_failure;
3155         } else if (dst) {
3156                 struct in6_addr saddr_buf;
3157                 if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0 &&
3158                     nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
3159                         goto nla_put_failure;
3160         }
3161
3162         if (rt->rt6i_prefsrc.plen) {
3163                 struct in6_addr saddr_buf;
3164                 saddr_buf = rt->rt6i_prefsrc.addr;
3165                 if (nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
3166                         goto nla_put_failure;
3167         }
3168
3169         memcpy(metrics, dst_metrics_ptr(&rt->dst), sizeof(metrics));
3170         if (rt->rt6i_pmtu)
3171                 metrics[RTAX_MTU - 1] = rt->rt6i_pmtu;
3172         if (rtnetlink_put_metrics(skb, metrics) < 0)
3173                 goto nla_put_failure;
3174
3175         if (rt->rt6i_flags & RTF_GATEWAY) {
3176                 if (nla_put_in6_addr(skb, RTA_GATEWAY, &rt->rt6i_gateway) < 0)
3177                         goto nla_put_failure;
3178         }
3179
3180         if (rt->dst.dev &&
3181             nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex))
3182                 goto nla_put_failure;
3183         if (nla_put_u32(skb, RTA_PRIORITY, rt->rt6i_metric))
3184                 goto nla_put_failure;
3185
3186         expires = (rt->rt6i_flags & RTF_EXPIRES) ? rt->dst.expires - jiffies : 0;
3187
3188         if (rtnl_put_cacheinfo(skb, &rt->dst, 0, expires, rt->dst.error) < 0)
3189                 goto nla_put_failure;
3190
3191         if (nla_put_u8(skb, RTA_PREF, IPV6_EXTRACT_PREF(rt->rt6i_flags)))
3192                 goto nla_put_failure;
3193
3194         lwtunnel_fill_encap(skb, rt->dst.lwtstate);
3195
3196         nlmsg_end(skb, nlh);
3197         return 0;
3198
3199 nla_put_failure:
3200         nlmsg_cancel(skb, nlh);
3201         return -EMSGSIZE;
3202 }
3203
3204 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
3205 {
3206         struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
3207         int prefix;
3208
3209         if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
3210                 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
3211                 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
3212         } else
3213                 prefix = 0;
3214
3215         return rt6_fill_node(arg->net,
3216                      arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
3217                      NETLINK_CB(arg->cb->skb).portid, arg->cb->nlh->nlmsg_seq,
3218                      prefix, 0, NLM_F_MULTI);
3219 }
3220
3221 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh)
3222 {
3223         struct net *net = sock_net(in_skb->sk);
3224         struct nlattr *tb[RTA_MAX+1];
3225         struct rt6_info *rt;
3226         struct sk_buff *skb;
3227         struct rtmsg *rtm;
3228         struct flowi6 fl6;
3229         int err, iif = 0, oif = 0;
3230
3231         err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
3232         if (err < 0)
3233                 goto errout;
3234
3235         err = -EINVAL;
3236         memset(&fl6, 0, sizeof(fl6));
3237
3238         if (tb[RTA_SRC]) {
3239                 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
3240                         goto errout;
3241
3242                 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
3243         }
3244
3245         if (tb[RTA_DST]) {
3246                 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
3247                         goto errout;
3248
3249                 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
3250         }
3251
3252         if (tb[RTA_IIF])
3253                 iif = nla_get_u32(tb[RTA_IIF]);
3254
3255         if (tb[RTA_OIF])
3256                 oif = nla_get_u32(tb[RTA_OIF]);
3257
3258         if (tb[RTA_MARK])
3259                 fl6.flowi6_mark = nla_get_u32(tb[RTA_MARK]);
3260
3261         if (iif) {
3262                 struct net_device *dev;
3263                 int flags = 0;
3264
3265                 dev = __dev_get_by_index(net, iif);
3266                 if (!dev) {
3267                         err = -ENODEV;
3268                         goto errout;
3269                 }
3270
3271                 fl6.flowi6_iif = iif;
3272
3273                 if (!ipv6_addr_any(&fl6.saddr))
3274                         flags |= RT6_LOOKUP_F_HAS_SADDR;
3275
3276                 rt = (struct rt6_info *)ip6_route_input_lookup(net, dev, &fl6,
3277                                                                flags);
3278         } else {
3279                 fl6.flowi6_oif = oif;
3280
3281                 rt = (struct rt6_info *)ip6_route_output(net, NULL, &fl6);
3282         }
3283
3284         skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
3285         if (!skb) {
3286                 ip6_rt_put(rt);
3287                 err = -ENOBUFS;
3288                 goto errout;
3289         }
3290
3291         /* Reserve room for dummy headers, this skb can pass
3292            through good chunk of routing engine.
3293          */
3294         skb_reset_mac_header(skb);
3295         skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
3296
3297         skb_dst_set(skb, &rt->dst);
3298
3299         err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
3300                             RTM_NEWROUTE, NETLINK_CB(in_skb).portid,
3301                             nlh->nlmsg_seq, 0, 0, 0);
3302         if (err < 0) {
3303                 kfree_skb(skb);
3304                 goto errout;
3305         }
3306
3307         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
3308 errout:
3309         return err;
3310 }
3311
3312 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info,
3313                      unsigned int nlm_flags)
3314 {
3315         struct sk_buff *skb;
3316         struct net *net = info->nl_net;
3317         u32 seq;
3318         int err;
3319
3320         err = -ENOBUFS;
3321         seq = info->nlh ? info->nlh->nlmsg_seq : 0;
3322
3323         skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
3324         if (!skb)
3325                 goto errout;
3326
3327         err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
3328                                 event, info->portid, seq, 0, 0, nlm_flags);
3329         if (err < 0) {
3330                 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
3331                 WARN_ON(err == -EMSGSIZE);
3332                 kfree_skb(skb);
3333                 goto errout;
3334         }
3335         rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
3336                     info->nlh, gfp_any());
3337         return;
3338 errout:
3339         if (err < 0)
3340                 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
3341 }
3342
3343 static int ip6_route_dev_notify(struct notifier_block *this,
3344                                 unsigned long event, void *ptr)
3345 {
3346         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3347         struct net *net = dev_net(dev);
3348
3349         if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
3350                 net->ipv6.ip6_null_entry->dst.dev = dev;
3351                 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
3352 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3353                 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
3354                 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
3355                 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
3356                 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
3357 #endif
3358         }
3359
3360         return NOTIFY_OK;
3361 }
3362
3363 /*
3364  *      /proc
3365  */
3366
3367 #ifdef CONFIG_PROC_FS
3368
3369 static const struct file_operations ipv6_route_proc_fops = {
3370         .owner          = THIS_MODULE,
3371         .open           = ipv6_route_open,
3372         .read           = seq_read,
3373         .llseek         = seq_lseek,
3374         .release        = seq_release_net,
3375 };
3376
3377 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
3378 {
3379         struct net *net = (struct net *)seq->private;
3380         seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
3381                    net->ipv6.rt6_stats->fib_nodes,
3382                    net->ipv6.rt6_stats->fib_route_nodes,
3383                    net->ipv6.rt6_stats->fib_rt_alloc,
3384                    net->ipv6.rt6_stats->fib_rt_entries,
3385                    net->ipv6.rt6_stats->fib_rt_cache,
3386                    dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
3387                    net->ipv6.rt6_stats->fib_discarded_routes);
3388
3389         return 0;
3390 }
3391
3392 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
3393 {
3394         return single_open_net(inode, file, rt6_stats_seq_show);
3395 }
3396
3397 static const struct file_operations rt6_stats_seq_fops = {
3398         .owner   = THIS_MODULE,
3399         .open    = rt6_stats_seq_open,
3400         .read    = seq_read,
3401         .llseek  = seq_lseek,
3402         .release = single_release_net,
3403 };
3404 #endif  /* CONFIG_PROC_FS */
3405
3406 #ifdef CONFIG_SYSCTL
3407
3408 static
3409 int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
3410                               void __user *buffer, size_t *lenp, loff_t *ppos)
3411 {
3412         struct net *net;
3413         int delay;
3414         if (!write)
3415                 return -EINVAL;
3416
3417         net = (struct net *)ctl->extra1;
3418         delay = net->ipv6.sysctl.flush_delay;
3419         proc_dointvec(ctl, write, buffer, lenp, ppos);
3420         fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
3421         return 0;
3422 }
3423
3424 struct ctl_table ipv6_route_table_template[] = {
3425         {
3426                 .procname       =       "flush",
3427                 .data           =       &init_net.ipv6.sysctl.flush_delay,
3428                 .maxlen         =       sizeof(int),
3429                 .mode           =       0200,
3430                 .proc_handler   =       ipv6_sysctl_rtcache_flush
3431         },
3432         {
3433                 .procname       =       "gc_thresh",
3434                 .data           =       &ip6_dst_ops_template.gc_thresh,
3435                 .maxlen         =       sizeof(int),
3436                 .mode           =       0644,
3437                 .proc_handler   =       proc_dointvec,
3438         },
3439         {
3440                 .procname       =       "max_size",
3441                 .data           =       &init_net.ipv6.sysctl.ip6_rt_max_size,
3442                 .maxlen         =       sizeof(int),
3443                 .mode           =       0644,
3444                 .proc_handler   =       proc_dointvec,
3445         },
3446         {
3447                 .procname       =       "gc_min_interval",
3448                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
3449                 .maxlen         =       sizeof(int),
3450                 .mode           =       0644,
3451                 .proc_handler   =       proc_dointvec_jiffies,
3452         },
3453         {
3454                 .procname       =       "gc_timeout",
3455                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
3456                 .maxlen         =       sizeof(int),
3457                 .mode           =       0644,
3458                 .proc_handler   =       proc_dointvec_jiffies,
3459         },
3460         {
3461                 .procname       =       "gc_interval",
3462                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_interval,
3463                 .maxlen         =       sizeof(int),
3464                 .mode           =       0644,
3465                 .proc_handler   =       proc_dointvec_jiffies,
3466         },
3467         {
3468                 .procname       =       "gc_elasticity",
3469                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
3470                 .maxlen         =       sizeof(int),
3471                 .mode           =       0644,
3472                 .proc_handler   =       proc_dointvec,
3473         },
3474         {
3475                 .procname       =       "mtu_expires",
3476                 .data           =       &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
3477                 .maxlen         =       sizeof(int),
3478                 .mode           =       0644,
3479                 .proc_handler   =       proc_dointvec_jiffies,
3480         },
3481         {
3482                 .procname       =       "min_adv_mss",
3483                 .data           =       &init_net.ipv6.sysctl.ip6_rt_min_advmss,
3484                 .maxlen         =       sizeof(int),
3485                 .mode           =       0644,
3486                 .proc_handler   =       proc_dointvec,
3487         },
3488         {
3489                 .procname       =       "gc_min_interval_ms",
3490                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
3491                 .maxlen         =       sizeof(int),
3492                 .mode           =       0644,
3493                 .proc_handler   =       proc_dointvec_ms_jiffies,
3494         },
3495         { }
3496 };
3497
3498 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
3499 {
3500         struct ctl_table *table;
3501
3502         table = kmemdup(ipv6_route_table_template,
3503                         sizeof(ipv6_route_table_template),
3504                         GFP_KERNEL);
3505
3506         if (table) {
3507                 table[0].data = &net->ipv6.sysctl.flush_delay;
3508                 table[0].extra1 = net;
3509                 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
3510                 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
3511                 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
3512                 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
3513                 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
3514                 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
3515                 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
3516                 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
3517                 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
3518
3519                 /* Don't export sysctls to unprivileged users */
3520                 if (net->user_ns != &init_user_ns)
3521                         table[0].procname = NULL;
3522         }
3523
3524         return table;
3525 }
3526 #endif
3527
3528 static int __net_init ip6_route_net_init(struct net *net)
3529 {
3530         int ret = -ENOMEM;
3531
3532         memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
3533                sizeof(net->ipv6.ip6_dst_ops));
3534
3535         if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
3536                 goto out_ip6_dst_ops;
3537
3538         net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
3539                                            sizeof(*net->ipv6.ip6_null_entry),
3540                                            GFP_KERNEL);
3541         if (!net->ipv6.ip6_null_entry)
3542                 goto out_ip6_dst_entries;
3543         net->ipv6.ip6_null_entry->dst.path =
3544                 (struct dst_entry *)net->ipv6.ip6_null_entry;
3545         net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3546         dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
3547                          ip6_template_metrics, true);
3548
3549 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3550         net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
3551                                                sizeof(*net->ipv6.ip6_prohibit_entry),
3552                                                GFP_KERNEL);
3553         if (!net->ipv6.ip6_prohibit_entry)
3554                 goto out_ip6_null_entry;
3555         net->ipv6.ip6_prohibit_entry->dst.path =
3556                 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
3557         net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3558         dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
3559                          ip6_template_metrics, true);
3560
3561         net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
3562                                                sizeof(*net->ipv6.ip6_blk_hole_entry),
3563                                                GFP_KERNEL);
3564         if (!net->ipv6.ip6_blk_hole_entry)
3565                 goto out_ip6_prohibit_entry;
3566         net->ipv6.ip6_blk_hole_entry->dst.path =
3567                 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
3568         net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3569         dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
3570                          ip6_template_metrics, true);
3571 #endif
3572
3573         net->ipv6.sysctl.flush_delay = 0;
3574         net->ipv6.sysctl.ip6_rt_max_size = 4096;
3575         net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
3576         net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
3577         net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
3578         net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
3579         net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
3580         net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
3581
3582         net->ipv6.ip6_rt_gc_expire = 30*HZ;
3583
3584         ret = 0;
3585 out:
3586         return ret;
3587
3588 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3589 out_ip6_prohibit_entry:
3590         kfree(net->ipv6.ip6_prohibit_entry);
3591 out_ip6_null_entry:
3592         kfree(net->ipv6.ip6_null_entry);
3593 #endif
3594 out_ip6_dst_entries:
3595         dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3596 out_ip6_dst_ops:
3597         goto out;
3598 }
3599
3600 static void __net_exit ip6_route_net_exit(struct net *net)
3601 {
3602         kfree(net->ipv6.ip6_null_entry);
3603 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3604         kfree(net->ipv6.ip6_prohibit_entry);
3605         kfree(net->ipv6.ip6_blk_hole_entry);
3606 #endif
3607         dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3608 }
3609
3610 static int __net_init ip6_route_net_init_late(struct net *net)
3611 {
3612 #ifdef CONFIG_PROC_FS
3613         proc_create("ipv6_route", 0, net->proc_net, &ipv6_route_proc_fops);
3614         proc_create("rt6_stats", S_IRUGO, net->proc_net, &rt6_stats_seq_fops);
3615 #endif
3616         return 0;
3617 }
3618
3619 static void __net_exit ip6_route_net_exit_late(struct net *net)
3620 {
3621 #ifdef CONFIG_PROC_FS
3622         remove_proc_entry("ipv6_route", net->proc_net);
3623         remove_proc_entry("rt6_stats", net->proc_net);
3624 #endif
3625 }
3626
3627 static struct pernet_operations ip6_route_net_ops = {
3628         .init = ip6_route_net_init,
3629         .exit = ip6_route_net_exit,
3630 };
3631
3632 static int __net_init ipv6_inetpeer_init(struct net *net)
3633 {
3634         struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
3635
3636         if (!bp)
3637                 return -ENOMEM;
3638         inet_peer_base_init(bp);
3639         net->ipv6.peers = bp;
3640         return 0;
3641 }
3642
3643 static void __net_exit ipv6_inetpeer_exit(struct net *net)
3644 {
3645         struct inet_peer_base *bp = net->ipv6.peers;
3646
3647         net->ipv6.peers = NULL;
3648         inetpeer_invalidate_tree(bp);
3649         kfree(bp);
3650 }
3651
3652 static struct pernet_operations ipv6_inetpeer_ops = {
3653         .init   =       ipv6_inetpeer_init,
3654         .exit   =       ipv6_inetpeer_exit,
3655 };
3656
3657 static struct pernet_operations ip6_route_net_late_ops = {
3658         .init = ip6_route_net_init_late,
3659         .exit = ip6_route_net_exit_late,
3660 };
3661
3662 static struct notifier_block ip6_route_dev_notifier = {
3663         .notifier_call = ip6_route_dev_notify,
3664         .priority = 0,
3665 };
3666
3667 int __init ip6_route_init(void)
3668 {
3669         int ret;
3670         int cpu;
3671
3672         ret = -ENOMEM;
3673         ip6_dst_ops_template.kmem_cachep =
3674                 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
3675                                   SLAB_HWCACHE_ALIGN, NULL);
3676         if (!ip6_dst_ops_template.kmem_cachep)
3677                 goto out;
3678
3679         ret = dst_entries_init(&ip6_dst_blackhole_ops);
3680         if (ret)
3681                 goto out_kmem_cache;
3682
3683         ret = register_pernet_subsys(&ipv6_inetpeer_ops);
3684         if (ret)
3685                 goto out_dst_entries;
3686
3687         ret = register_pernet_subsys(&ip6_route_net_ops);
3688         if (ret)
3689                 goto out_register_inetpeer;
3690
3691         ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
3692
3693         /* Registering of the loopback is done before this portion of code,
3694          * the loopback reference in rt6_info will not be taken, do it
3695          * manually for init_net */
3696         init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
3697         init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3698   #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3699         init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
3700         init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3701         init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
3702         init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3703   #endif
3704         ret = fib6_init();
3705         if (ret)
3706                 goto out_register_subsys;
3707
3708         ret = xfrm6_init();
3709         if (ret)
3710                 goto out_fib6_init;
3711
3712         ret = fib6_rules_init();
3713         if (ret)
3714                 goto xfrm6_init;
3715
3716         ret = register_pernet_subsys(&ip6_route_net_late_ops);
3717         if (ret)
3718                 goto fib6_rules_init;
3719
3720         ret = -ENOBUFS;
3721         if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL, NULL) ||
3722             __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL, NULL) ||
3723             __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL, NULL))
3724                 goto out_register_late_subsys;
3725
3726         ret = register_netdevice_notifier(&ip6_route_dev_notifier);
3727         if (ret)
3728                 goto out_register_late_subsys;
3729
3730         for_each_possible_cpu(cpu) {
3731                 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
3732
3733                 INIT_LIST_HEAD(&ul->head);
3734                 spin_lock_init(&ul->lock);
3735         }
3736
3737 out:
3738         return ret;
3739
3740 out_register_late_subsys:
3741         unregister_pernet_subsys(&ip6_route_net_late_ops);
3742 fib6_rules_init:
3743         fib6_rules_cleanup();
3744 xfrm6_init:
3745         xfrm6_fini();
3746 out_fib6_init:
3747         fib6_gc_cleanup();
3748 out_register_subsys:
3749         unregister_pernet_subsys(&ip6_route_net_ops);
3750 out_register_inetpeer:
3751         unregister_pernet_subsys(&ipv6_inetpeer_ops);
3752 out_dst_entries:
3753         dst_entries_destroy(&ip6_dst_blackhole_ops);
3754 out_kmem_cache:
3755         kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3756         goto out;
3757 }
3758
3759 void ip6_route_cleanup(void)
3760 {
3761         unregister_netdevice_notifier(&ip6_route_dev_notifier);
3762         unregister_pernet_subsys(&ip6_route_net_late_ops);
3763         fib6_rules_cleanup();
3764         xfrm6_fini();
3765         fib6_gc_cleanup();
3766         unregister_pernet_subsys(&ipv6_inetpeer_ops);
3767         unregister_pernet_subsys(&ip6_route_net_ops);
3768         dst_entries_destroy(&ip6_dst_blackhole_ops);
3769         kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3770 }