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 struct dst_ops ip6_dst_blackhole_ops = {
 252         .family                 =       AF_INET6,
 253         .destroy                =       ip6_dst_destroy,
 254         .check                  =       ip6_dst_check,
 255         .mtu                    =       ip6_blackhole_mtu,
 256         .default_advmss         =       ip6_default_advmss,
 257         .update_pmtu            =       ip6_rt_blackhole_update_pmtu,
 258         .redirect               =       ip6_rt_blackhole_redirect,
 259         .cow_metrics            =       dst_cow_metrics_generic,
 260         .neigh_lookup           =       ip6_neigh_lookup,
 261 };
 262
 263 static const u32 ip6_template_metrics[RTAX_MAX] = {
 264         [RTAX_HOPLIMIT - 1] = 0,
 265 };
 266
 267 static const struct rt6_info ip6_null_entry_template = {
 268         .dst = {
 269                 .__refcnt       = ATOMIC_INIT(1),
 270                 .__use          = 1,
 271                 .obsolete       = DST_OBSOLETE_FORCE_CHK,
 272                 .error          = -ENETUNREACH,
 273                 .input          = ip6_pkt_discard,
 274                 .output         = ip6_pkt_discard_out,
 275         },
 276         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
 277         .rt6i_protocol  = RTPROT_KERNEL,
 278         .rt6i_metric    = ~(u32) 0,
 279         .rt6i_ref       = ATOMIC_INIT(1),
 280 };
 281
 282 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
 283
 284 static const struct rt6_info ip6_prohibit_entry_template = {
 285         .dst = {
 286                 .__refcnt       = ATOMIC_INIT(1),
 287                 .__use          = 1,
 288                 .obsolete       = DST_OBSOLETE_FORCE_CHK,
 289                 .error          = -EACCES,
 290                 .input          = ip6_pkt_prohibit,
 291                 .output         = ip6_pkt_prohibit_out,
 292         },
 293         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
 294         .rt6i_protocol  = RTPROT_KERNEL,
 295         .rt6i_metric    = ~(u32) 0,
 296         .rt6i_ref       = ATOMIC_INIT(1),
 297 };
 298
 299 static const struct rt6_info ip6_blk_hole_entry_template = {
 300         .dst = {
 301                 .__refcnt       = ATOMIC_INIT(1),
 302                 .__use          = 1,
 303                 .obsolete       = DST_OBSOLETE_FORCE_CHK,
 304                 .error          = -EINVAL,
 305                 .input          = dst_discard,
 306                 .output         = dst_discard_sk,
 307         },
 308         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
 309         .rt6i_protocol  = RTPROT_KERNEL,
 310         .rt6i_metric    = ~(u32) 0,
 311         .rt6i_ref       = ATOMIC_INIT(1),
 312 };
 313
 314 #endif
 315
 316 static void rt6_info_init(struct rt6_info *rt)
 317 {
 318         struct dst_entry *dst = &rt->dst;
 319
 320         memset(dst + 1, 0, sizeof(*rt) - sizeof(*dst));
 321         INIT_LIST_HEAD(&rt->rt6i_siblings);
 322         INIT_LIST_HEAD(&rt->rt6i_uncached);
 323 }
 324
 325 /* allocate dst with ip6_dst_ops */
 326 static struct rt6_info *__ip6_dst_alloc(struct net *net,
 327                                         struct net_device *dev,
 328                                         int flags)
 329 {
 330         struct rt6_info *rt = dst_alloc(&net->ipv6.ip6_dst_ops, dev,
 331                                         0, DST_OBSOLETE_FORCE_CHK, flags);
 332
 333         if (rt)
 334                 rt6_info_init(rt);
 335
 336         return rt;
 337 }
 338
 339 static struct rt6_info *ip6_dst_alloc(struct net *net,
 340                                       struct net_device *dev,
 341                                       int flags)
 342 {
 343         struct rt6_info *rt = __ip6_dst_alloc(net, dev, flags);
 344
 345         if (rt) {
 346                 rt->rt6i_pcpu = alloc_percpu_gfp(struct rt6_info *, GFP_ATOMIC);
 347                 if (rt->rt6i_pcpu) {
 348                         int cpu;
 349
 350                         for_each_possible_cpu(cpu) {
 351                                 struct rt6_info **p;
 352
 353                                 p = per_cpu_ptr(rt->rt6i_pcpu, cpu);
 354                                 /* no one shares rt */
 355                                 *p =  NULL;
 356                         }
 357                 } else {
 358                         dst_destroy((struct dst_entry *)rt);
 359                         return NULL;
 360                 }
 361         }
 362
 363         return rt;
 364 }
 365
 366 static void ip6_dst_destroy(struct dst_entry *dst)
 367 {
 368         struct rt6_info *rt = (struct rt6_info *)dst;
 369         struct dst_entry *from = dst->from;
 370         struct inet6_dev *idev;
 371
 372         dst_destroy_metrics_generic(dst);
 373         free_percpu(rt->rt6i_pcpu);
 374         rt6_uncached_list_del(rt);
 375
 376         idev = rt->rt6i_idev;
 377         if (idev) {
 378                 rt->rt6i_idev = NULL;
 379                 in6_dev_put(idev);
 380         }
 381
 382         dst->from = NULL;
 383         dst_release(from);
 384 }
 385
 386 static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
 387                            int how)
 388 {
 389         struct rt6_info *rt = (struct rt6_info *)dst;
 390         struct inet6_dev *idev = rt->rt6i_idev;
 391         struct net_device *loopback_dev =
 392                 dev_net(dev)->loopback_dev;
 393
 394         if (dev != loopback_dev) {
 395                 if (idev && idev->dev == dev) {
 396                         struct inet6_dev *loopback_idev =
 397                                 in6_dev_get(loopback_dev);
 398                         if (loopback_idev) {
 399                                 rt->rt6i_idev = loopback_idev;
 400                                 in6_dev_put(idev);
 401                         }
 402                 }
 403         }
 404 }
 405
 406 static bool rt6_check_expired(const struct rt6_info *rt)
 407 {
 408         if (rt->rt6i_flags & RTF_EXPIRES) {
 409                 if (time_after(jiffies, rt->dst.expires))
 410                         return true;
 411         } else if (rt->dst.from) {
 412                 return rt6_check_expired((struct rt6_info *) rt->dst.from);
 413         }
 414         return false;
 415 }
 416
 417 /* Multipath route selection:
 418  *   Hash based function using packet header and flowlabel.
 419  * Adapted from fib_info_hashfn()
 420  */
 421 static int rt6_info_hash_nhsfn(unsigned int candidate_count,
 422                                const struct flowi6 *fl6)
 423 {
 424         unsigned int val = fl6->flowi6_proto;
 425
 426         val ^= ipv6_addr_hash(&fl6->daddr);
 427         val ^= ipv6_addr_hash(&fl6->saddr);
 428
 429         /* Work only if this not encapsulated */
 430         switch (fl6->flowi6_proto) {
 431         case IPPROTO_UDP:
 432         case IPPROTO_TCP:
 433         case IPPROTO_SCTP:
 434                 val ^= (__force u16)fl6->fl6_sport;
 435                 val ^= (__force u16)fl6->fl6_dport;
 436                 break;
 437
 438         case IPPROTO_ICMPV6:
 439                 val ^= (__force u16)fl6->fl6_icmp_type;
 440                 val ^= (__force u16)fl6->fl6_icmp_code;
 441                 break;
 442         }
 443         /* RFC6438 recommands to use flowlabel */
 444         val ^= (__force u32)fl6->flowlabel;
 445
 446         /* Perhaps, we need to tune, this function? */
 447         val = val ^ (val >> 7) ^ (val >> 12);
 448         return val % candidate_count;
 449 }
 450
 451 static struct rt6_info *rt6_multipath_select(struct rt6_info *match,
 452                                              struct flowi6 *fl6, int oif,
 453                                              int strict)
 454 {
 455         struct rt6_info *sibling, *next_sibling;
 456         int route_choosen;
 457
 458         route_choosen = rt6_info_hash_nhsfn(match->rt6i_nsiblings + 1, fl6);
 459         /* Don't change the route, if route_choosen == 0
 460          * (siblings does not include ourself)
 461          */
 462         if (route_choosen)
 463                 list_for_each_entry_safe(sibling, next_sibling,
 464                                 &match->rt6i_siblings, rt6i_siblings) {
 465                         route_choosen--;
 466                         if (route_choosen == 0) {
 467                                 if (rt6_score_route(sibling, oif, strict) < 0)
 468                                         break;
 469                                 match = sibling;
 470                                 break;
 471                         }
 472                 }
 473         return match;
 474 }
 475
 476 /*
 477  *      Route lookup. Any table->tb6_lock is implied.
 478  */
 479
 480 static inline struct rt6_info *rt6_device_match(struct net *net,
 481                                                     struct rt6_info *rt,
 482                                                     const struct in6_addr *saddr,
 483                                                     int oif,
 484                                                     int flags)
 485 {
 486         struct rt6_info *local = NULL;
 487         struct rt6_info *sprt;
 488
 489         if (!oif && ipv6_addr_any(saddr))
 490                 goto out;
 491
 492         for (sprt = rt; sprt; sprt = sprt->dst.rt6_next) {
 493                 struct net_device *dev = sprt->dst.dev;
 494
 495                 if (oif) {
 496                         if (dev->ifindex == oif)
 497                                 return sprt;
 498                         if (dev->flags & IFF_LOOPBACK) {
 499                                 if (!sprt->rt6i_idev ||
 500                                     sprt->rt6i_idev->dev->ifindex != oif) {
 501                                         if (flags & RT6_LOOKUP_F_IFACE && oif)
 502                                                 continue;
 503                                         if (local && (!oif ||
 504                                                       local->rt6i_idev->dev->ifindex == oif))
 505                                                 continue;
 506                                 }
 507                                 local = sprt;
 508                         }
 509                 } else {
 510                         if (ipv6_chk_addr(net, saddr, dev,
 511                                           flags & RT6_LOOKUP_F_IFACE))
 512                                 return sprt;
 513                 }
 514         }
 515
 516         if (oif) {
 517                 if (local)
 518                         return local;
 519
 520                 if (flags & RT6_LOOKUP_F_IFACE)
 521                         return net->ipv6.ip6_null_entry;
 522         }
 523 out:
 524         return rt;
 525 }
 526
 527 #ifdef CONFIG_IPV6_ROUTER_PREF
 528 struct __rt6_probe_work {
 529         struct work_struct work;
 530         struct in6_addr target;
 531         struct net_device *dev;
 532 };
 533
 534 static void rt6_probe_deferred(struct work_struct *w)
 535 {
 536         struct in6_addr mcaddr;
 537         struct __rt6_probe_work *work =
 538                 container_of(w, struct __rt6_probe_work, work);
 539
 540         addrconf_addr_solict_mult(&work->target, &mcaddr);
 541         ndisc_send_ns(work->dev, NULL, &work->target, &mcaddr, NULL, NULL);
 542         dev_put(work->dev);
 543         kfree(work);
 544 }
 545
 546 static void rt6_probe(struct rt6_info *rt)
 547 {
 548         struct __rt6_probe_work *work;
 549         struct neighbour *neigh;
 550         /*
 551          * Okay, this does not seem to be appropriate
 552          * for now, however, we need to check if it
 553          * is really so; aka Router Reachability Probing.
 554          *
 555          * Router Reachability Probe MUST be rate-limited
 556          * to no more than one per minute.
 557          */
 558         if (!rt || !(rt->rt6i_flags & RTF_GATEWAY))
 559                 return;
 560         rcu_read_lock_bh();
 561         neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
 562         if (neigh) {
 563                 if (neigh->nud_state & NUD_VALID)
 564                         goto out;
 565
 566                 work = NULL;
 567                 write_lock(&neigh->lock);
 568                 if (!(neigh->nud_state & NUD_VALID) &&
 569                     time_after(jiffies,
 570                                neigh->updated +
 571                                rt->rt6i_idev->cnf.rtr_probe_interval)) {
 572                         work = kmalloc(sizeof(*work), GFP_ATOMIC);
 573                         if (work)
 574                                 __neigh_set_probe_once(neigh);
 575                 }
 576                 write_unlock(&neigh->lock);
 577         } else {
 578                 work = kmalloc(sizeof(*work), GFP_ATOMIC);
 579         }
 580
 581         if (work) {
 582                 INIT_WORK(&work->work, rt6_probe_deferred);
 583                 work->target = rt->rt6i_gateway;
 584                 dev_hold(rt->dst.dev);
 585                 work->dev = rt->dst.dev;
 586                 schedule_work(&work->work);
 587         }
 588
 589 out:
 590         rcu_read_unlock_bh();
 591 }
 592 #else
 593 static inline void rt6_probe(struct rt6_info *rt)
 594 {
 595 }
 596 #endif
 597
 598 /*
 599  * Default Router Selection (RFC 2461 6.3.6)
 600  */
 601 static inline int rt6_check_dev(struct rt6_info *rt, int oif)
 602 {
 603         struct net_device *dev = rt->dst.dev;
 604         if (!oif || dev->ifindex == oif)
 605                 return 2;
 606         if ((dev->flags & IFF_LOOPBACK) &&
 607             rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
 608                 return 1;
 609         return 0;
 610 }
 611
 612 static inline enum rt6_nud_state rt6_check_neigh(struct rt6_info *rt)
 613 {
 614         struct neighbour *neigh;
 615         enum rt6_nud_state ret = RT6_NUD_FAIL_HARD;
 616
 617         if (rt->rt6i_flags & RTF_NONEXTHOP ||
 618             !(rt->rt6i_flags & RTF_GATEWAY))
 619                 return RT6_NUD_SUCCEED;
 620
 621         rcu_read_lock_bh();
 622         neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
 623         if (neigh) {
 624                 read_lock(&neigh->lock);
 625                 if (neigh->nud_state & NUD_VALID)
 626                         ret = RT6_NUD_SUCCEED;
 627 #ifdef CONFIG_IPV6_ROUTER_PREF
 628                 else if (!(neigh->nud_state & NUD_FAILED))
 629                         ret = RT6_NUD_SUCCEED;
 630                 else
 631                         ret = RT6_NUD_FAIL_PROBE;
 632 #endif
 633                 read_unlock(&neigh->lock);
 634         } else {
 635                 ret = IS_ENABLED(CONFIG_IPV6_ROUTER_PREF) ?
 636                       RT6_NUD_SUCCEED : RT6_NUD_FAIL_DO_RR;
 637         }
 638         rcu_read_unlock_bh();
 639
 640         return ret;
 641 }
 642
 643 static int rt6_score_route(struct rt6_info *rt, int oif,
 644                            int strict)
 645 {
 646         int m;
 647
 648         m = rt6_check_dev(rt, oif);
 649         if (!m && (strict & RT6_LOOKUP_F_IFACE))
 650                 return RT6_NUD_FAIL_HARD;
 651 #ifdef CONFIG_IPV6_ROUTER_PREF
 652         m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
 653 #endif
 654         if (strict & RT6_LOOKUP_F_REACHABLE) {
 655                 int n = rt6_check_neigh(rt);
 656                 if (n < 0)
 657                         return n;
 658         }
 659         return m;
 660 }
 661
 662 static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
 663                                    int *mpri, struct rt6_info *match,
 664                                    bool *do_rr)
 665 {
 666         int m;
 667         bool match_do_rr = false;
 668         struct inet6_dev *idev = rt->rt6i_idev;
 669         struct net_device *dev = rt->dst.dev;
 670
 671         if (dev && !netif_carrier_ok(dev) &&
 672             idev->cnf.ignore_routes_with_linkdown)
 673                 goto out;
 674
 675         if (rt6_check_expired(rt))
 676                 goto out;
 677
 678         m = rt6_score_route(rt, oif, strict);
 679         if (m == RT6_NUD_FAIL_DO_RR) {
 680                 match_do_rr = true;
 681                 m = 0; /* lowest valid score */
 682         } else if (m == RT6_NUD_FAIL_HARD) {
 683                 goto out;
 684         }
 685
 686         if (strict & RT6_LOOKUP_F_REACHABLE)
 687                 rt6_probe(rt);
 688
 689         /* note that m can be RT6_NUD_FAIL_PROBE at this point */
 690         if (m > *mpri) {
 691                 *do_rr = match_do_rr;
 692                 *mpri = m;
 693                 match = rt;
 694         }
 695 out:
 696         return match;
 697 }
 698
 699 static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
 700                                      struct rt6_info *rr_head,
 701                                      u32 metric, int oif, int strict,
 702                                      bool *do_rr)
 703 {
 704         struct rt6_info *rt, *match, *cont;
 705         int mpri = -1;
 706
 707         match = NULL;
 708         cont = NULL;
 709         for (rt = rr_head; rt; rt = rt->dst.rt6_next) {
 710                 if (rt->rt6i_metric != metric) {
 711                         cont = rt;
 712                         break;
 713                 }
 714
 715                 match = find_match(rt, oif, strict, &mpri, match, do_rr);
 716         }
 717
 718         for (rt = fn->leaf; rt && rt != rr_head; rt = rt->dst.rt6_next) {
 719                 if (rt->rt6i_metric != metric) {
 720                         cont = rt;
 721                         break;
 722                 }
 723
 724                 match = find_match(rt, oif, strict, &mpri, match, do_rr);
 725         }
 726
 727         if (match || !cont)
 728                 return match;
 729
 730         for (rt = cont; rt; rt = rt->dst.rt6_next)
 731                 match = find_match(rt, oif, strict, &mpri, match, do_rr);
 732
 733         return match;
 734 }
 735
 736 static struct rt6_info *rt6_select(struct fib6_node *fn, int oif, int strict)
 737 {
 738         struct rt6_info *match, *rt0;
 739         struct net *net;
 740         bool do_rr = false;
 741
 742         rt0 = fn->rr_ptr;
 743         if (!rt0)
 744                 fn->rr_ptr = rt0 = fn->leaf;
 745
 746         match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict,
 747                              &do_rr);
 748
 749         if (do_rr) {
 750                 struct rt6_info *next = rt0->dst.rt6_next;
 751
 752                 /* no entries matched; do round-robin */
 753                 if (!next || next->rt6i_metric != rt0->rt6i_metric)
 754                         next = fn->leaf;
 755
 756                 if (next != rt0)
 757                         fn->rr_ptr = next;
 758         }
 759
 760         net = dev_net(rt0->dst.dev);
 761         return match ? match : net->ipv6.ip6_null_entry;
 762 }
 763
 764 static bool rt6_is_gw_or_nonexthop(const struct rt6_info *rt)
 765 {
 766         return (rt->rt6i_flags & (RTF_NONEXTHOP | RTF_GATEWAY));
 767 }
 768
 769 #ifdef CONFIG_IPV6_ROUTE_INFO
 770 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
 771                   const struct in6_addr *gwaddr)
 772 {
 773         struct net *net = dev_net(dev);
 774         struct route_info *rinfo = (struct route_info *) opt;
 775         struct in6_addr prefix_buf, *prefix;
 776         unsigned int pref;
 777         unsigned long lifetime;
 778         struct rt6_info *rt;
 779
 780         if (len < sizeof(struct route_info)) {
 781                 return -EINVAL;
 782         }
 783
 784         /* Sanity check for prefix_len and length */
 785         if (rinfo->length > 3) {
 786                 return -EINVAL;
 787         } else if (rinfo->prefix_len > 128) {
 788                 return -EINVAL;
 789         } else if (rinfo->prefix_len > 64) {
 790                 if (rinfo->length < 2) {
 791                         return -EINVAL;
 792                 }
 793         } else if (rinfo->prefix_len > 0) {
 794                 if (rinfo->length < 1) {
 795                         return -EINVAL;
 796                 }
 797         }
 798
 799         pref = rinfo->route_pref;
 800         if (pref == ICMPV6_ROUTER_PREF_INVALID)
 801                 return -EINVAL;
 802
 803         lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
 804
 805         if (rinfo->length == 3)
 806                 prefix = (struct in6_addr *)rinfo->prefix;
 807         else {
 808                 /* this function is safe */
 809                 ipv6_addr_prefix(&prefix_buf,
 810                                  (struct in6_addr *)rinfo->prefix,
 811                                  rinfo->prefix_len);
 812                 prefix = &prefix_buf;
 813         }
 814
 815         if (rinfo->prefix_len == 0)
 816                 rt = rt6_get_dflt_router(gwaddr, dev);
 817         else
 818                 rt = rt6_get_route_info(net, prefix, rinfo->prefix_len,
 819                                         gwaddr, dev->ifindex);
 820
 821         if (rt && !lifetime) {
 822                 ip6_del_rt(rt);
 823                 rt = NULL;
 824         }
 825
 826         if (!rt && lifetime)
 827                 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr, dev->ifindex,
 828                                         pref);
 829         else if (rt)
 830                 rt->rt6i_flags = RTF_ROUTEINFO |
 831                                  (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
 832
 833         if (rt) {
 834                 if (!addrconf_finite_timeout(lifetime))
 835                         rt6_clean_expires(rt);
 836                 else
 837                         rt6_set_expires(rt, jiffies + HZ * lifetime);
 838
 839                 ip6_rt_put(rt);
 840         }
 841         return 0;
 842 }
 843 #endif
 844
 845 static struct fib6_node* fib6_backtrack(struct fib6_node *fn,
 846                                         struct in6_addr *saddr)
 847 {
 848         struct fib6_node *pn;
 849         while (1) {
 850                 if (fn->fn_flags & RTN_TL_ROOT)
 851                         return NULL;
 852                 pn = fn->parent;
 853                 if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn)
 854                         fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr);
 855                 else
 856                         fn = pn;
 857                 if (fn->fn_flags & RTN_RTINFO)
 858                         return fn;
 859         }
 860 }
 861
 862 static struct rt6_info *ip6_pol_route_lookup(struct net *net,
 863                                              struct fib6_table *table,
 864                                              struct flowi6 *fl6, int flags)
 865 {
 866         struct fib6_node *fn;
 867         struct rt6_info *rt;
 868
 869         read_lock_bh(&table->tb6_lock);
 870         fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
 871 restart:
 872         rt = fn->leaf;
 873         rt = rt6_device_match(net, rt, &fl6->saddr, fl6->flowi6_oif, flags);
 874         if (rt->rt6i_nsiblings && fl6->flowi6_oif == 0)
 875                 rt = rt6_multipath_select(rt, fl6, fl6->flowi6_oif, flags);
 876         if (rt == net->ipv6.ip6_null_entry) {
 877                 fn = fib6_backtrack(fn, &fl6->saddr);
 878                 if (fn)
 879                         goto restart;
 880         }
 881         dst_use(&rt->dst, jiffies);
 882         read_unlock_bh(&table->tb6_lock);
 883         return rt;
 884
 885 }
 886
 887 struct dst_entry *ip6_route_lookup(struct net *net, struct flowi6 *fl6,
 888                                     int flags)
 889 {
 890         return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_lookup);
 891 }
 892 EXPORT_SYMBOL_GPL(ip6_route_lookup);
 893
 894 struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
 895                             const struct in6_addr *saddr, int oif, int strict)
 896 {
 897         struct flowi6 fl6 = {
 898                 .flowi6_oif = oif,
 899                 .daddr = *daddr,
 900         };
 901         struct dst_entry *dst;
 902         int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
 903
 904         if (saddr) {
 905                 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
 906                 flags |= RT6_LOOKUP_F_HAS_SADDR;
 907         }
 908
 909         dst = fib6_rule_lookup(net, &fl6, flags, ip6_pol_route_lookup);
 910         if (dst->error == 0)
 911                 return (struct rt6_info *) dst;
 912
 913         dst_release(dst);
 914
 915         return NULL;
 916 }
 917 EXPORT_SYMBOL(rt6_lookup);
 918
 919 /* ip6_ins_rt is called with FREE table->tb6_lock.
 920    It takes new route entry, the addition fails by any reason the
 921    route is freed. In any case, if caller does not hold it, it may
 922    be destroyed.
 923  */
 924
 925 static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info,
 926                         struct mx6_config *mxc)
 927 {
 928         int err;
 929         struct fib6_table *table;
 930
 931         table = rt->rt6i_table;
 932         write_lock_bh(&table->tb6_lock);
 933         err = fib6_add(&table->tb6_root, rt, info, mxc);
 934         write_unlock_bh(&table->tb6_lock);
 935
 936         return err;
 937 }
 938
 939 int ip6_ins_rt(struct rt6_info *rt)
 940 {
 941         struct nl_info info = { .nl_net = dev_net(rt->dst.dev), };
 942         struct mx6_config mxc = { .mx = NULL, };
 943
 944         return __ip6_ins_rt(rt, &info, &mxc);
 945 }
 946
 947 static struct rt6_info *ip6_rt_cache_alloc(struct rt6_info *ort,
 948                                            const struct in6_addr *daddr,
 949                                            const struct in6_addr *saddr)
 950 {
 951         struct rt6_info *rt;
 952
 953         /*
 954          *      Clone the route.
 955          */
 956
 957         if (ort->rt6i_flags & (RTF_CACHE | RTF_PCPU))
 958                 ort = (struct rt6_info *)ort->dst.from;
 959
 960         rt = __ip6_dst_alloc(dev_net(ort->dst.dev), ort->dst.dev, 0);
 961
 962         if (!rt)
 963                 return NULL;
 964
 965         ip6_rt_copy_init(rt, ort);
 966         rt->rt6i_flags |= RTF_CACHE;
 967         rt->rt6i_metric = 0;
 968         rt->dst.flags |= DST_HOST;
 969         rt->rt6i_dst.addr = *daddr;
 970         rt->rt6i_dst.plen = 128;
 971
 972         if (!rt6_is_gw_or_nonexthop(ort)) {
 973                 if (ort->rt6i_dst.plen != 128 &&
 974                     ipv6_addr_equal(&ort->rt6i_dst.addr, daddr))
 975                         rt->rt6i_flags |= RTF_ANYCAST;
 976 #ifdef CONFIG_IPV6_SUBTREES
 977                 if (rt->rt6i_src.plen && saddr) {
 978                         rt->rt6i_src.addr = *saddr;
 979                         rt->rt6i_src.plen = 128;
 980                 }
 981 #endif
 982         }
 983
 984         return rt;
 985 }
 986
 987 static struct rt6_info *ip6_rt_pcpu_alloc(struct rt6_info *rt)
 988 {
 989         struct rt6_info *pcpu_rt;
 990
 991         pcpu_rt = __ip6_dst_alloc(dev_net(rt->dst.dev),
 992                                   rt->dst.dev, rt->dst.flags);
 993
 994         if (!pcpu_rt)
 995                 return NULL;
 996         ip6_rt_copy_init(pcpu_rt, rt);
 997         pcpu_rt->rt6i_protocol = rt->rt6i_protocol;
 998         pcpu_rt->rt6i_flags |= RTF_PCPU;
 999         return pcpu_rt;
1000 }
1001
1002 /* It should be called with read_lock_bh(&tb6_lock) acquired */
1003 static struct rt6_info *rt6_get_pcpu_route(struct rt6_info *rt)
1004 {
1005         struct rt6_info *pcpu_rt, **p;
1006
1007         p = this_cpu_ptr(rt->rt6i_pcpu);
1008         pcpu_rt = *p;
1009
1010         if (pcpu_rt) {
1011                 dst_hold(&pcpu_rt->dst);
1012                 rt6_dst_from_metrics_check(pcpu_rt);
1013         }
1014         return pcpu_rt;
1015 }
1016
1017 static struct rt6_info *rt6_make_pcpu_route(struct rt6_info *rt)
1018 {
1019         struct fib6_table *table = rt->rt6i_table;
1020         struct rt6_info *pcpu_rt, *prev, **p;
1021
1022         pcpu_rt = ip6_rt_pcpu_alloc(rt);
1023         if (!pcpu_rt) {
1024                 struct net *net = dev_net(rt->dst.dev);
1025
1026                 dst_hold(&net->ipv6.ip6_null_entry->dst);
1027                 return net->ipv6.ip6_null_entry;
1028         }
1029
1030         read_lock_bh(&table->tb6_lock);
1031         if (rt->rt6i_pcpu) {
1032                 p = this_cpu_ptr(rt->rt6i_pcpu);
1033                 prev = cmpxchg(p, NULL, pcpu_rt);
1034                 if (prev) {
1035                         /* If someone did it before us, return prev instead */
1036                         dst_destroy(&pcpu_rt->dst);
1037                         pcpu_rt = prev;
1038                 }
1039         } else {
1040                 /* rt has been removed from the fib6 tree
1041                  * before we have a chance to acquire the read_lock.
1042                  * In this case, don't brother to create a pcpu rt
1043                  * since rt is going away anyway.  The next
1044                  * dst_check() will trigger a re-lookup.
1045                  */
1046                 dst_destroy(&pcpu_rt->dst);
1047                 pcpu_rt = rt;
1048         }
1049         dst_hold(&pcpu_rt->dst);
1050         rt6_dst_from_metrics_check(pcpu_rt);
1051         read_unlock_bh(&table->tb6_lock);
1052         return pcpu_rt;
1053 }
1054
1055 static struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table, int oif,
1056                                       struct flowi6 *fl6, int flags)
1057 {
1058         struct fib6_node *fn, *saved_fn;
1059         struct rt6_info *rt;
1060         int strict = 0;
1061
1062         strict |= flags & RT6_LOOKUP_F_IFACE;
1063         if (net->ipv6.devconf_all->forwarding == 0)
1064                 strict |= RT6_LOOKUP_F_REACHABLE;
1065
1066         read_lock_bh(&table->tb6_lock);
1067
1068         fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1069         saved_fn = fn;
1070
1071         if (fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF)
1072                 oif = 0;
1073
1074 redo_rt6_select:
1075         rt = rt6_select(fn, oif, strict);
1076         if (rt->rt6i_nsiblings)
1077                 rt = rt6_multipath_select(rt, fl6, oif, strict);
1078         if (rt == net->ipv6.ip6_null_entry) {
1079                 fn = fib6_backtrack(fn, &fl6->saddr);
1080                 if (fn)
1081                         goto redo_rt6_select;
1082                 else if (strict & RT6_LOOKUP_F_REACHABLE) {
1083                         /* also consider unreachable route */
1084                         strict &= ~RT6_LOOKUP_F_REACHABLE;
1085                         fn = saved_fn;
1086                         goto redo_rt6_select;
1087                 }
1088         }
1089
1090
1091         if (rt == net->ipv6.ip6_null_entry || (rt->rt6i_flags & RTF_CACHE)) {
1092                 dst_use(&rt->dst, jiffies);
1093                 read_unlock_bh(&table->tb6_lock);
1094
1095                 rt6_dst_from_metrics_check(rt);
1096                 return rt;
1097         } else if (unlikely((fl6->flowi6_flags & FLOWI_FLAG_KNOWN_NH) &&
1098                             !(rt->rt6i_flags & RTF_GATEWAY))) {
1099                 /* Create a RTF_CACHE clone which will not be
1100                  * owned by the fib6 tree.  It is for the special case where
1101                  * the daddr in the skb during the neighbor look-up is different
1102                  * from the fl6->daddr used to look-up route here.
1103                  */
1104
1105                 struct rt6_info *uncached_rt;
1106
1107                 dst_use(&rt->dst, jiffies);
1108                 read_unlock_bh(&table->tb6_lock);
1109
1110                 uncached_rt = ip6_rt_cache_alloc(rt, &fl6->daddr, NULL);
1111                 dst_release(&rt->dst);
1112
1113                 if (uncached_rt)
1114                         rt6_uncached_list_add(uncached_rt);
1115                 else
1116                         uncached_rt = net->ipv6.ip6_null_entry;
1117
1118                 dst_hold(&uncached_rt->dst);
1119                 return uncached_rt;
1120
1121         } else {
1122                 /* Get a percpu copy */
1123
1124                 struct rt6_info *pcpu_rt;
1125
1126                 rt->dst.lastuse = jiffies;
1127                 rt->dst.__use++;
1128                 pcpu_rt = rt6_get_pcpu_route(rt);
1129
1130                 if (pcpu_rt) {
1131                         read_unlock_bh(&table->tb6_lock);
1132                 } else {
1133                         /* We have to do the read_unlock first
1134                          * because rt6_make_pcpu_route() may trigger
1135                          * ip6_dst_gc() which will take the write_lock.
1136                          */
1137                         dst_hold(&rt->dst);
1138                         read_unlock_bh(&table->tb6_lock);
1139                         pcpu_rt = rt6_make_pcpu_route(rt);
1140                         dst_release(&rt->dst);
1141                 }
1142
1143                 return pcpu_rt;
1144
1145         }
1146 }
1147
1148 static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
1149                                             struct flowi6 *fl6, int flags)
1150 {
1151         return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, flags);
1152 }
1153
1154 static struct dst_entry *ip6_route_input_lookup(struct net *net,
1155                                                 struct net_device *dev,
1156                                                 struct flowi6 *fl6, int flags)
1157 {
1158         if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
1159                 flags |= RT6_LOOKUP_F_IFACE;
1160
1161         return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_input);
1162 }
1163
1164 void ip6_route_input(struct sk_buff *skb)
1165 {
1166         const struct ipv6hdr *iph = ipv6_hdr(skb);
1167         struct net *net = dev_net(skb->dev);
1168         int flags = RT6_LOOKUP_F_HAS_SADDR;
1169         struct ip_tunnel_info *tun_info;
1170         struct flowi6 fl6 = {
1171                 .flowi6_iif = skb->dev->ifindex,
1172                 .daddr = iph->daddr,
1173                 .saddr = iph->saddr,
1174                 .flowlabel = ip6_flowinfo(iph),
1175                 .flowi6_mark = skb->mark,
1176                 .flowi6_proto = iph->nexthdr,
1177         };
1178
1179         tun_info = skb_tunnel_info(skb);
1180         if (tun_info && !(tun_info->mode & IP_TUNNEL_INFO_TX))
1181                 fl6.flowi6_tun_key.tun_id = tun_info->key.tun_id;
1182         skb_dst_drop(skb);
1183         skb_dst_set(skb, ip6_route_input_lookup(net, skb->dev, &fl6, flags));
1184 }
1185
1186 static struct rt6_info *ip6_pol_route_output(struct net *net, struct fib6_table *table,
1187                                              struct flowi6 *fl6, int flags)
1188 {
1189         return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, flags);
1190 }
1191
1192 struct dst_entry *ip6_route_output(struct net *net, const struct sock *sk,
1193                                     struct flowi6 *fl6)
1194 {
1195         int flags = 0;
1196         bool any_src;
1197
1198         fl6->flowi6_iif = LOOPBACK_IFINDEX;
1199
1200         any_src = ipv6_addr_any(&fl6->saddr);
1201         if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr) ||
1202             (fl6->flowi6_oif && any_src))
1203                 flags |= RT6_LOOKUP_F_IFACE;
1204
1205         if (!any_src)
1206                 flags |= RT6_LOOKUP_F_HAS_SADDR;
1207         else if (sk)
1208                 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
1209
1210         return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_output);
1211 }
1212 EXPORT_SYMBOL(ip6_route_output);
1213
1214 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
1215 {
1216         struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
1217         struct dst_entry *new = NULL;
1218
1219         rt = dst_alloc(&ip6_dst_blackhole_ops, ort->dst.dev, 1, DST_OBSOLETE_NONE, 0);
1220         if (rt) {
1221                 rt6_info_init(rt);
1222
1223                 new = &rt->dst;
1224                 new->__use = 1;
1225                 new->input = dst_discard;
1226                 new->output = dst_discard_sk;
1227
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 & ~RTF_PCPU;
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 netevent_redirect netevent;
2103         struct rt6_info *rt, *nrt = NULL;
2104         struct ndisc_options ndopts;
2105         struct inet6_dev *in6_dev;
2106         struct neighbour *neigh;
2107         struct rd_msg *msg;
2108         int optlen, on_link;
2109         u8 *lladdr;
2110
2111         optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
2112         optlen -= sizeof(*msg);
2113
2114         if (optlen < 0) {
2115                 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
2116                 return;
2117         }
2118
2119         msg = (struct rd_msg *)icmp6_hdr(skb);
2120
2121         if (ipv6_addr_is_multicast(&msg->dest)) {
2122                 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
2123                 return;
2124         }
2125
2126         on_link = 0;
2127         if (ipv6_addr_equal(&msg->dest, &msg->target)) {
2128                 on_link = 1;
2129         } else if (ipv6_addr_type(&msg->target) !=
2130                    (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
2131                 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
2132                 return;
2133         }
2134
2135         in6_dev = __in6_dev_get(skb->dev);
2136         if (!in6_dev)
2137                 return;
2138         if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
2139                 return;
2140
2141         /* RFC2461 8.1:
2142          *      The IP source address of the Redirect MUST be the same as the current
2143          *      first-hop router for the specified ICMP Destination Address.
2144          */
2145
2146         if (!ndisc_parse_options(msg->opt, optlen, &ndopts)) {
2147                 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
2148                 return;
2149         }
2150
2151         lladdr = NULL;
2152         if (ndopts.nd_opts_tgt_lladdr) {
2153                 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
2154                                              skb->dev);
2155                 if (!lladdr) {
2156                         net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
2157                         return;
2158                 }
2159         }
2160
2161         rt = (struct rt6_info *) dst;
2162         if (rt->rt6i_flags & RTF_REJECT) {
2163                 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
2164                 return;
2165         }
2166
2167         /* Redirect received -> path was valid.
2168          * Look, redirects are sent only in response to data packets,
2169          * so that this nexthop apparently is reachable. --ANK
2170          */
2171         dst_confirm(&rt->dst);
2172
2173         neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
2174         if (!neigh)
2175                 return;
2176
2177         /*
2178          *      We have finally decided to accept it.
2179          */
2180
2181         neigh_update(neigh, lladdr, NUD_STALE,
2182                      NEIGH_UPDATE_F_WEAK_OVERRIDE|
2183                      NEIGH_UPDATE_F_OVERRIDE|
2184                      (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
2185                                      NEIGH_UPDATE_F_ISROUTER))
2186                      );
2187
2188         nrt = ip6_rt_cache_alloc(rt, &msg->dest, NULL);
2189         if (!nrt)
2190                 goto out;
2191
2192         nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
2193         if (on_link)
2194                 nrt->rt6i_flags &= ~RTF_GATEWAY;
2195
2196         nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
2197
2198         if (ip6_ins_rt(nrt))
2199                 goto out;
2200
2201         netevent.old = &rt->dst;
2202         netevent.new = &nrt->dst;
2203         netevent.daddr = &msg->dest;
2204         netevent.neigh = neigh;
2205         call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
2206
2207         if (rt->rt6i_flags & RTF_CACHE) {
2208                 rt = (struct rt6_info *) dst_clone(&rt->dst);
2209                 ip6_del_rt(rt);
2210         }
2211
2212 out:
2213         neigh_release(neigh);
2214 }
2215
2216 /*
2217  *      Misc support functions
2218  */
2219
2220 static void rt6_set_from(struct rt6_info *rt, struct rt6_info *from)
2221 {
2222         BUG_ON(from->dst.from);
2223
2224         rt->rt6i_flags &= ~RTF_EXPIRES;
2225         dst_hold(&from->dst);
2226         rt->dst.from = &from->dst;
2227         dst_init_metrics(&rt->dst, dst_metrics_ptr(&from->dst), true);
2228 }
2229
2230 static void ip6_rt_copy_init(struct rt6_info *rt, struct rt6_info *ort)
2231 {
2232         rt->dst.input = ort->dst.input;
2233         rt->dst.output = ort->dst.output;
2234         rt->rt6i_dst = ort->rt6i_dst;
2235         rt->dst.error = ort->dst.error;
2236         rt->rt6i_idev = ort->rt6i_idev;
2237         if (rt->rt6i_idev)
2238                 in6_dev_hold(rt->rt6i_idev);
2239         rt->dst.lastuse = jiffies;
2240         rt->rt6i_gateway = ort->rt6i_gateway;
2241         rt->rt6i_flags = ort->rt6i_flags;
2242         rt6_set_from(rt, ort);
2243         rt->rt6i_metric = ort->rt6i_metric;
2244 #ifdef CONFIG_IPV6_SUBTREES
2245         rt->rt6i_src = ort->rt6i_src;
2246 #endif
2247         rt->rt6i_prefsrc = ort->rt6i_prefsrc;
2248         rt->rt6i_table = ort->rt6i_table;
2249         rt->dst.lwtstate = lwtstate_get(ort->dst.lwtstate);
2250 }
2251
2252 #ifdef CONFIG_IPV6_ROUTE_INFO
2253 static struct rt6_info *rt6_get_route_info(struct net *net,
2254                                            const struct in6_addr *prefix, int prefixlen,
2255                                            const struct in6_addr *gwaddr, int ifindex)
2256 {
2257         struct fib6_node *fn;
2258         struct rt6_info *rt = NULL;
2259         struct fib6_table *table;
2260
2261         table = fib6_get_table(net, RT6_TABLE_INFO);
2262         if (!table)
2263                 return NULL;
2264
2265         read_lock_bh(&table->tb6_lock);
2266         fn = fib6_locate(&table->tb6_root, prefix, prefixlen, NULL, 0);
2267         if (!fn)
2268                 goto out;
2269
2270         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2271                 if (rt->dst.dev->ifindex != ifindex)
2272                         continue;
2273                 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
2274                         continue;
2275                 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
2276                         continue;
2277                 dst_hold(&rt->dst);
2278                 break;
2279         }
2280 out:
2281         read_unlock_bh(&table->tb6_lock);
2282         return rt;
2283 }
2284
2285 static struct rt6_info *rt6_add_route_info(struct net *net,
2286                                            const struct in6_addr *prefix, int prefixlen,
2287                                            const struct in6_addr *gwaddr, int ifindex,
2288                                            unsigned int pref)
2289 {
2290         struct fib6_config cfg = {
2291                 .fc_table       = RT6_TABLE_INFO,
2292                 .fc_metric      = IP6_RT_PRIO_USER,
2293                 .fc_ifindex     = ifindex,
2294                 .fc_dst_len     = prefixlen,
2295                 .fc_flags       = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
2296                                   RTF_UP | RTF_PREF(pref),
2297                 .fc_nlinfo.portid = 0,
2298                 .fc_nlinfo.nlh = NULL,
2299                 .fc_nlinfo.nl_net = net,
2300         };
2301
2302         cfg.fc_dst = *prefix;
2303         cfg.fc_gateway = *gwaddr;
2304
2305         /* We should treat it as a default route if prefix length is 0. */
2306         if (!prefixlen)
2307                 cfg.fc_flags |= RTF_DEFAULT;
2308
2309         ip6_route_add(&cfg);
2310
2311         return rt6_get_route_info(net, prefix, prefixlen, gwaddr, ifindex);
2312 }
2313 #endif
2314
2315 struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
2316 {
2317         struct rt6_info *rt;
2318         struct fib6_table *table;
2319
2320         table = fib6_get_table(dev_net(dev), RT6_TABLE_DFLT);
2321         if (!table)
2322                 return NULL;
2323
2324         read_lock_bh(&table->tb6_lock);
2325         for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2326                 if (dev == rt->dst.dev &&
2327                     ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
2328                     ipv6_addr_equal(&rt->rt6i_gateway, addr))
2329                         break;
2330         }
2331         if (rt)
2332                 dst_hold(&rt->dst);
2333         read_unlock_bh(&table->tb6_lock);
2334         return rt;
2335 }
2336
2337 struct rt6_info *rt6_add_dflt_router(const struct in6_addr *gwaddr,
2338                                      struct net_device *dev,
2339                                      unsigned int pref)
2340 {
2341         struct fib6_config cfg = {
2342                 .fc_table       = RT6_TABLE_DFLT,
2343                 .fc_metric      = IP6_RT_PRIO_USER,
2344                 .fc_ifindex     = dev->ifindex,
2345                 .fc_flags       = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
2346                                   RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
2347                 .fc_nlinfo.portid = 0,
2348                 .fc_nlinfo.nlh = NULL,
2349                 .fc_nlinfo.nl_net = dev_net(dev),
2350         };
2351
2352         cfg.fc_gateway = *gwaddr;
2353
2354         ip6_route_add(&cfg);
2355
2356         return rt6_get_dflt_router(gwaddr, dev);
2357 }
2358
2359 void rt6_purge_dflt_routers(struct net *net)
2360 {
2361         struct rt6_info *rt;
2362         struct fib6_table *table;
2363
2364         /* NOTE: Keep consistent with rt6_get_dflt_router */
2365         table = fib6_get_table(net, RT6_TABLE_DFLT);
2366         if (!table)
2367                 return;
2368
2369 restart:
2370         read_lock_bh(&table->tb6_lock);
2371         for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2372                 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
2373                     (!rt->rt6i_idev || rt->rt6i_idev->cnf.accept_ra != 2)) {
2374                         dst_hold(&rt->dst);
2375                         read_unlock_bh(&table->tb6_lock);
2376                         ip6_del_rt(rt);
2377                         goto restart;
2378                 }
2379         }
2380         read_unlock_bh(&table->tb6_lock);
2381 }
2382
2383 static void rtmsg_to_fib6_config(struct net *net,
2384                                  struct in6_rtmsg *rtmsg,
2385                                  struct fib6_config *cfg)
2386 {
2387         memset(cfg, 0, sizeof(*cfg));
2388
2389         cfg->fc_table = RT6_TABLE_MAIN;
2390         cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
2391         cfg->fc_metric = rtmsg->rtmsg_metric;
2392         cfg->fc_expires = rtmsg->rtmsg_info;
2393         cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
2394         cfg->fc_src_len = rtmsg->rtmsg_src_len;
2395         cfg->fc_flags = rtmsg->rtmsg_flags;
2396
2397         cfg->fc_nlinfo.nl_net = net;
2398
2399         cfg->fc_dst = rtmsg->rtmsg_dst;
2400         cfg->fc_src = rtmsg->rtmsg_src;
2401         cfg->fc_gateway = rtmsg->rtmsg_gateway;
2402 }
2403
2404 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
2405 {
2406         struct fib6_config cfg;
2407         struct in6_rtmsg rtmsg;
2408         int err;
2409
2410         switch (cmd) {
2411         case SIOCADDRT:         /* Add a route */
2412         case SIOCDELRT:         /* Delete a route */
2413                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2414                         return -EPERM;
2415                 err = copy_from_user(&rtmsg, arg,
2416                                      sizeof(struct in6_rtmsg));
2417                 if (err)
2418                         return -EFAULT;
2419
2420                 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
2421
2422                 rtnl_lock();
2423                 switch (cmd) {
2424                 case SIOCADDRT:
2425                         err = ip6_route_add(&cfg);
2426                         break;
2427                 case SIOCDELRT:
2428                         err = ip6_route_del(&cfg);
2429                         break;
2430                 default:
2431                         err = -EINVAL;
2432                 }
2433                 rtnl_unlock();
2434
2435                 return err;
2436         }
2437
2438         return -EINVAL;
2439 }
2440
2441 /*
2442  *      Drop the packet on the floor
2443  */
2444
2445 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
2446 {
2447         int type;
2448         struct dst_entry *dst = skb_dst(skb);
2449         switch (ipstats_mib_noroutes) {
2450         case IPSTATS_MIB_INNOROUTES:
2451                 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
2452                 if (type == IPV6_ADDR_ANY) {
2453                         IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2454                                       IPSTATS_MIB_INADDRERRORS);
2455                         break;
2456                 }
2457                 /* FALLTHROUGH */
2458         case IPSTATS_MIB_OUTNOROUTES:
2459                 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2460                               ipstats_mib_noroutes);
2461                 break;
2462         }
2463         icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
2464         kfree_skb(skb);
2465         return 0;
2466 }
2467
2468 static int ip6_pkt_discard(struct sk_buff *skb)
2469 {
2470         return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
2471 }
2472
2473 static int ip6_pkt_discard_out(struct sock *sk, struct sk_buff *skb)
2474 {
2475         skb->dev = skb_dst(skb)->dev;
2476         return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
2477 }
2478
2479 static int ip6_pkt_prohibit(struct sk_buff *skb)
2480 {
2481         return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
2482 }
2483
2484 static int ip6_pkt_prohibit_out(struct sock *sk, struct sk_buff *skb)
2485 {
2486         skb->dev = skb_dst(skb)->dev;
2487         return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
2488 }
2489
2490 /*
2491  *      Allocate a dst for local (unicast / anycast) address.
2492  */
2493
2494 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
2495                                     const struct in6_addr *addr,
2496                                     bool anycast)
2497 {
2498         struct net *net = dev_net(idev->dev);
2499         struct rt6_info *rt = ip6_dst_alloc(net, net->loopback_dev,
2500                                             DST_NOCOUNT);
2501         if (!rt)
2502                 return ERR_PTR(-ENOMEM);
2503
2504         in6_dev_hold(idev);
2505
2506         rt->dst.flags |= DST_HOST;
2507         rt->dst.input = ip6_input;
2508         rt->dst.output = ip6_output;
2509         rt->rt6i_idev = idev;
2510
2511         rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
2512         if (anycast)
2513                 rt->rt6i_flags |= RTF_ANYCAST;
2514         else
2515                 rt->rt6i_flags |= RTF_LOCAL;
2516
2517         rt->rt6i_gateway  = *addr;
2518         rt->rt6i_dst.addr = *addr;
2519         rt->rt6i_dst.plen = 128;
2520         rt->rt6i_table = fib6_get_table(net, RT6_TABLE_LOCAL);
2521         rt->dst.flags |= DST_NOCACHE;
2522
2523         atomic_set(&rt->dst.__refcnt, 1);
2524
2525         return rt;
2526 }
2527
2528 int ip6_route_get_saddr(struct net *net,
2529                         struct rt6_info *rt,
2530                         const struct in6_addr *daddr,
2531                         unsigned int prefs,
2532                         struct in6_addr *saddr)
2533 {
2534         struct inet6_dev *idev =
2535                 rt ? ip6_dst_idev((struct dst_entry *)rt) : NULL;
2536         int err = 0;
2537         if (rt && rt->rt6i_prefsrc.plen)
2538                 *saddr = rt->rt6i_prefsrc.addr;
2539         else
2540                 err = ipv6_dev_get_saddr(net, idev ? idev->dev : NULL,
2541                                          daddr, prefs, saddr);
2542         return err;
2543 }
2544
2545 /* remove deleted ip from prefsrc entries */
2546 struct arg_dev_net_ip {
2547         struct net_device *dev;
2548         struct net *net;
2549         struct in6_addr *addr;
2550 };
2551
2552 static int fib6_remove_prefsrc(struct rt6_info *rt, void *arg)
2553 {
2554         struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
2555         struct net *net = ((struct arg_dev_net_ip *)arg)->net;
2556         struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
2557
2558         if (((void *)rt->dst.dev == dev || !dev) &&
2559             rt != net->ipv6.ip6_null_entry &&
2560             ipv6_addr_equal(addr, &rt->rt6i_prefsrc.addr)) {
2561                 /* remove prefsrc entry */
2562                 rt->rt6i_prefsrc.plen = 0;
2563         }
2564         return 0;
2565 }
2566
2567 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
2568 {
2569         struct net *net = dev_net(ifp->idev->dev);
2570         struct arg_dev_net_ip adni = {
2571                 .dev = ifp->idev->dev,
2572                 .net = net,
2573                 .addr = &ifp->addr,
2574         };
2575         fib6_clean_all(net, fib6_remove_prefsrc, &adni);
2576 }
2577
2578 #define RTF_RA_ROUTER           (RTF_ADDRCONF | RTF_DEFAULT | RTF_GATEWAY)
2579 #define RTF_CACHE_GATEWAY       (RTF_GATEWAY | RTF_CACHE)
2580
2581 /* Remove routers and update dst entries when gateway turn into host. */
2582 static int fib6_clean_tohost(struct rt6_info *rt, void *arg)
2583 {
2584         struct in6_addr *gateway = (struct in6_addr *)arg;
2585
2586         if ((((rt->rt6i_flags & RTF_RA_ROUTER) == RTF_RA_ROUTER) ||
2587              ((rt->rt6i_flags & RTF_CACHE_GATEWAY) == RTF_CACHE_GATEWAY)) &&
2588              ipv6_addr_equal(gateway, &rt->rt6i_gateway)) {
2589                 return -1;
2590         }
2591         return 0;
2592 }
2593
2594 void rt6_clean_tohost(struct net *net, struct in6_addr *gateway)
2595 {
2596         fib6_clean_all(net, fib6_clean_tohost, gateway);
2597 }
2598
2599 struct arg_dev_net {
2600         struct net_device *dev;
2601         struct net *net;
2602 };
2603
2604 static int fib6_ifdown(struct rt6_info *rt, void *arg)
2605 {
2606         const struct arg_dev_net *adn = arg;
2607         const struct net_device *dev = adn->dev;
2608
2609         if ((rt->dst.dev == dev || !dev) &&
2610             rt != adn->net->ipv6.ip6_null_entry)
2611                 return -1;
2612
2613         return 0;
2614 }
2615
2616 void rt6_ifdown(struct net *net, struct net_device *dev)
2617 {
2618         struct arg_dev_net adn = {
2619                 .dev = dev,
2620                 .net = net,
2621         };
2622
2623         fib6_clean_all(net, fib6_ifdown, &adn);
2624         icmp6_clean_all(fib6_ifdown, &adn);
2625         if (dev)
2626                 rt6_uncached_list_flush_dev(net, dev);
2627 }
2628
2629 struct rt6_mtu_change_arg {
2630         struct net_device *dev;
2631         unsigned int mtu;
2632 };
2633
2634 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2635 {
2636         struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2637         struct inet6_dev *idev;
2638
2639         /* In IPv6 pmtu discovery is not optional,
2640            so that RTAX_MTU lock cannot disable it.
2641            We still use this lock to block changes
2642            caused by addrconf/ndisc.
2643         */
2644
2645         idev = __in6_dev_get(arg->dev);
2646         if (!idev)
2647                 return 0;
2648
2649         /* For administrative MTU increase, there is no way to discover
2650            IPv6 PMTU increase, so PMTU increase should be updated here.
2651            Since RFC 1981 doesn't include administrative MTU increase
2652            update PMTU increase is a MUST. (i.e. jumbo frame)
2653          */
2654         /*
2655            If new MTU is less than route PMTU, this new MTU will be the
2656            lowest MTU in the path, update the route PMTU to reflect PMTU
2657            decreases; if new MTU is greater than route PMTU, and the
2658            old MTU is the lowest MTU in the path, update the route PMTU
2659            to reflect the increase. In this case if the other nodes' MTU
2660            also have the lowest MTU, TOO BIG MESSAGE will be lead to
2661            PMTU discouvery.
2662          */
2663         if (rt->dst.dev == arg->dev &&
2664             !dst_metric_locked(&rt->dst, RTAX_MTU)) {
2665                 if (rt->rt6i_flags & RTF_CACHE) {
2666                         /* For RTF_CACHE with rt6i_pmtu == 0
2667                          * (i.e. a redirected route),
2668                          * the metrics of its rt->dst.from has already
2669                          * been updated.
2670                          */
2671                         if (rt->rt6i_pmtu && rt->rt6i_pmtu > arg->mtu)
2672                                 rt->rt6i_pmtu = arg->mtu;
2673                 } else if (dst_mtu(&rt->dst) >= arg->mtu ||
2674                            (dst_mtu(&rt->dst) < arg->mtu &&
2675                             dst_mtu(&rt->dst) == idev->cnf.mtu6)) {
2676                         dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
2677                 }
2678         }
2679         return 0;
2680 }
2681
2682 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
2683 {
2684         struct rt6_mtu_change_arg arg = {
2685                 .dev = dev,
2686                 .mtu = mtu,
2687         };
2688
2689         fib6_clean_all(dev_net(dev), rt6_mtu_change_route, &arg);
2690 }
2691
2692 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2693         [RTA_GATEWAY]           = { .len = sizeof(struct in6_addr) },
2694         [RTA_OIF]               = { .type = NLA_U32 },
2695         [RTA_IIF]               = { .type = NLA_U32 },
2696         [RTA_PRIORITY]          = { .type = NLA_U32 },
2697         [RTA_METRICS]           = { .type = NLA_NESTED },
2698         [RTA_MULTIPATH]         = { .len = sizeof(struct rtnexthop) },
2699         [RTA_PREF]              = { .type = NLA_U8 },
2700         [RTA_ENCAP_TYPE]        = { .type = NLA_U16 },
2701         [RTA_ENCAP]             = { .type = NLA_NESTED },
2702 };
2703
2704 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2705                               struct fib6_config *cfg)
2706 {
2707         struct rtmsg *rtm;
2708         struct nlattr *tb[RTA_MAX+1];
2709         unsigned int pref;
2710         int err;
2711
2712         err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2713         if (err < 0)
2714                 goto errout;
2715
2716         err = -EINVAL;
2717         rtm = nlmsg_data(nlh);
2718         memset(cfg, 0, sizeof(*cfg));
2719
2720         cfg->fc_table = rtm->rtm_table;
2721         cfg->fc_dst_len = rtm->rtm_dst_len;
2722         cfg->fc_src_len = rtm->rtm_src_len;
2723         cfg->fc_flags = RTF_UP;
2724         cfg->fc_protocol = rtm->rtm_protocol;
2725         cfg->fc_type = rtm->rtm_type;
2726
2727         if (rtm->rtm_type == RTN_UNREACHABLE ||
2728             rtm->rtm_type == RTN_BLACKHOLE ||
2729             rtm->rtm_type == RTN_PROHIBIT ||
2730             rtm->rtm_type == RTN_THROW)
2731                 cfg->fc_flags |= RTF_REJECT;
2732
2733         if (rtm->rtm_type == RTN_LOCAL)
2734                 cfg->fc_flags |= RTF_LOCAL;
2735
2736         if (rtm->rtm_flags & RTM_F_CLONED)
2737                 cfg->fc_flags |= RTF_CACHE;
2738
2739         cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid;
2740         cfg->fc_nlinfo.nlh = nlh;
2741         cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2742
2743         if (tb[RTA_GATEWAY]) {
2744                 cfg->fc_gateway = nla_get_in6_addr(tb[RTA_GATEWAY]);
2745                 cfg->fc_flags |= RTF_GATEWAY;
2746         }
2747
2748         if (tb[RTA_DST]) {
2749                 int plen = (rtm->rtm_dst_len + 7) >> 3;
2750
2751                 if (nla_len(tb[RTA_DST]) < plen)
2752                         goto errout;
2753
2754                 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2755         }
2756
2757         if (tb[RTA_SRC]) {
2758                 int plen = (rtm->rtm_src_len + 7) >> 3;
2759
2760                 if (nla_len(tb[RTA_SRC]) < plen)
2761                         goto errout;
2762
2763                 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2764         }
2765
2766         if (tb[RTA_PREFSRC])
2767                 cfg->fc_prefsrc = nla_get_in6_addr(tb[RTA_PREFSRC]);
2768
2769         if (tb[RTA_OIF])
2770                 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2771
2772         if (tb[RTA_PRIORITY])
2773                 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2774
2775         if (tb[RTA_METRICS]) {
2776                 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2777                 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2778         }
2779
2780         if (tb[RTA_TABLE])
2781                 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2782
2783         if (tb[RTA_MULTIPATH]) {
2784                 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
2785                 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
2786         }
2787
2788         if (tb[RTA_PREF]) {
2789                 pref = nla_get_u8(tb[RTA_PREF]);
2790                 if (pref != ICMPV6_ROUTER_PREF_LOW &&
2791                     pref != ICMPV6_ROUTER_PREF_HIGH)
2792                         pref = ICMPV6_ROUTER_PREF_MEDIUM;
2793                 cfg->fc_flags |= RTF_PREF(pref);
2794         }
2795
2796         if (tb[RTA_ENCAP])
2797                 cfg->fc_encap = tb[RTA_ENCAP];
2798
2799         if (tb[RTA_ENCAP_TYPE])
2800                 cfg->fc_encap_type = nla_get_u16(tb[RTA_ENCAP_TYPE]);
2801
2802         err = 0;
2803 errout:
2804         return err;
2805 }
2806
2807 struct rt6_nh {
2808         struct rt6_info *rt6_info;
2809         struct fib6_config r_cfg;
2810         struct mx6_config mxc;
2811         struct list_head next;
2812 };
2813
2814 static void ip6_print_replace_route_err(struct list_head *rt6_nh_list)
2815 {
2816         struct rt6_nh *nh;
2817
2818         list_for_each_entry(nh, rt6_nh_list, next) {
2819                 pr_warn("IPV6: multipath route replace failed (check consistency of installed routes): %pI6 nexthop %pI6 ifi %d\n",
2820                         &nh->r_cfg.fc_dst, &nh->r_cfg.fc_gateway,
2821                         nh->r_cfg.fc_ifindex);
2822         }
2823 }
2824
2825 static int ip6_route_info_append(struct list_head *rt6_nh_list,
2826                                  struct rt6_info *rt, struct fib6_config *r_cfg)
2827 {
2828         struct rt6_nh *nh;
2829         struct rt6_info *rtnh;
2830         int err = -EEXIST;
2831
2832         list_for_each_entry(nh, rt6_nh_list, next) {
2833                 /* check if rt6_info already exists */
2834                 rtnh = nh->rt6_info;
2835
2836                 if (rtnh->dst.dev == rt->dst.dev &&
2837                     rtnh->rt6i_idev == rt->rt6i_idev &&
2838                     ipv6_addr_equal(&rtnh->rt6i_gateway,
2839                                     &rt->rt6i_gateway))
2840                         return err;
2841         }
2842
2843         nh = kzalloc(sizeof(*nh), GFP_KERNEL);
2844         if (!nh)
2845                 return -ENOMEM;
2846         nh->rt6_info = rt;
2847         err = ip6_convert_metrics(&nh->mxc, r_cfg);
2848         if (err) {
2849                 kfree(nh);
2850                 return err;
2851         }
2852         memcpy(&nh->r_cfg, r_cfg, sizeof(*r_cfg));
2853         list_add_tail(&nh->next, rt6_nh_list);
2854
2855         return 0;
2856 }
2857
2858 static int ip6_route_multipath_add(struct fib6_config *cfg)
2859 {
2860         struct fib6_config r_cfg;
2861         struct rtnexthop *rtnh;
2862         struct rt6_info *rt;
2863         struct rt6_nh *err_nh;
2864         struct rt6_nh *nh, *nh_safe;
2865         int remaining;
2866         int attrlen;
2867         int err = 1;
2868         int nhn = 0;
2869         int replace = (cfg->fc_nlinfo.nlh &&
2870                        (cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_REPLACE));
2871         LIST_HEAD(rt6_nh_list);
2872
2873         remaining = cfg->fc_mp_len;
2874         rtnh = (struct rtnexthop *)cfg->fc_mp;
2875
2876         /* Parse a Multipath Entry and build a list (rt6_nh_list) of
2877          * rt6_info structs per nexthop
2878          */
2879         while (rtnh_ok(rtnh, remaining)) {
2880                 memcpy(&r_cfg, cfg, sizeof(*cfg));
2881                 if (rtnh->rtnh_ifindex)
2882                         r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
2883
2884                 attrlen = rtnh_attrlen(rtnh);
2885                 if (attrlen > 0) {
2886                         struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
2887
2888                         nla = nla_find(attrs, attrlen, RTA_GATEWAY);
2889                         if (nla) {
2890                                 r_cfg.fc_gateway = nla_get_in6_addr(nla);
2891                                 r_cfg.fc_flags |= RTF_GATEWAY;
2892                         }
2893                         r_cfg.fc_encap = nla_find(attrs, attrlen, RTA_ENCAP);
2894                         nla = nla_find(attrs, attrlen, RTA_ENCAP_TYPE);
2895                         if (nla)
2896                                 r_cfg.fc_encap_type = nla_get_u16(nla);
2897                 }
2898
2899                 err = ip6_route_info_create(&r_cfg, &rt);
2900                 if (err)
2901                         goto cleanup;
2902
2903                 err = ip6_route_info_append(&rt6_nh_list, rt, &r_cfg);
2904                 if (err) {
2905                         dst_free(&rt->dst);
2906                         goto cleanup;
2907                 }
2908
2909                 rtnh = rtnh_next(rtnh, &remaining);
2910         }
2911
2912         err_nh = NULL;
2913         list_for_each_entry(nh, &rt6_nh_list, next) {
2914                 err = __ip6_ins_rt(nh->rt6_info, &cfg->fc_nlinfo, &nh->mxc);
2915                 /* nh->rt6_info is used or freed at this point, reset to NULL*/
2916                 nh->rt6_info = NULL;
2917                 if (err) {
2918                         if (replace && nhn)
2919                                 ip6_print_replace_route_err(&rt6_nh_list);
2920                         err_nh = nh;
2921                         goto add_errout;
2922                 }
2923
2924                 /* Because each route is added like a single route we remove
2925                  * these flags after the first nexthop: if there is a collision,
2926                  * we have already failed to add the first nexthop:
2927                  * fib6_add_rt2node() has rejected it; when replacing, old
2928                  * nexthops have been replaced by first new, the rest should
2929                  * be added to it.
2930                  */
2931                 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~(NLM_F_EXCL |
2932                                                      NLM_F_REPLACE);
2933                 nhn++;
2934         }
2935
2936         goto cleanup;
2937
2938 add_errout:
2939         /* Delete routes that were already added */
2940         list_for_each_entry(nh, &rt6_nh_list, next) {
2941                 if (err_nh == nh)
2942                         break;
2943                 ip6_route_del(&nh->r_cfg);
2944         }
2945
2946 cleanup:
2947         list_for_each_entry_safe(nh, nh_safe, &rt6_nh_list, next) {
2948                 if (nh->rt6_info)
2949                         dst_free(&nh->rt6_info->dst);
2950                 kfree(nh->mxc.mx);
2951                 list_del(&nh->next);
2952                 kfree(nh);
2953         }
2954
2955         return err;
2956 }
2957
2958 static int ip6_route_multipath_del(struct fib6_config *cfg)
2959 {
2960         struct fib6_config r_cfg;
2961         struct rtnexthop *rtnh;
2962         int remaining;
2963         int attrlen;
2964         int err = 1, last_err = 0;
2965
2966         remaining = cfg->fc_mp_len;
2967         rtnh = (struct rtnexthop *)cfg->fc_mp;
2968
2969         /* Parse a Multipath Entry */
2970         while (rtnh_ok(rtnh, remaining)) {
2971                 memcpy(&r_cfg, cfg, sizeof(*cfg));
2972                 if (rtnh->rtnh_ifindex)
2973                         r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
2974
2975                 attrlen = rtnh_attrlen(rtnh);
2976                 if (attrlen > 0) {
2977                         struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
2978
2979                         nla = nla_find(attrs, attrlen, RTA_GATEWAY);
2980                         if (nla) {
2981                                 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
2982                                 r_cfg.fc_flags |= RTF_GATEWAY;
2983                         }
2984                 }
2985                 err = ip6_route_del(&r_cfg);
2986                 if (err)
2987                         last_err = err;
2988
2989                 rtnh = rtnh_next(rtnh, &remaining);
2990         }
2991
2992         return last_err;
2993 }
2994
2995 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh)
2996 {
2997         struct fib6_config cfg;
2998         int err;
2999
3000         err = rtm_to_fib6_config(skb, nlh, &cfg);
3001         if (err < 0)
3002                 return err;
3003
3004         if (cfg.fc_mp)
3005                 return ip6_route_multipath_del(&cfg);
3006         else
3007                 return ip6_route_del(&cfg);
3008 }
3009
3010 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh)
3011 {
3012         struct fib6_config cfg;
3013         int err;
3014
3015         err = rtm_to_fib6_config(skb, nlh, &cfg);
3016         if (err < 0)
3017                 return err;
3018
3019         if (cfg.fc_mp)
3020                 return ip6_route_multipath_add(&cfg);
3021         else
3022                 return ip6_route_add(&cfg);
3023 }
3024
3025 static inline size_t rt6_nlmsg_size(struct rt6_info *rt)
3026 {
3027         return NLMSG_ALIGN(sizeof(struct rtmsg))
3028                + nla_total_size(16) /* RTA_SRC */
3029                + nla_total_size(16) /* RTA_DST */
3030                + nla_total_size(16) /* RTA_GATEWAY */
3031                + nla_total_size(16) /* RTA_PREFSRC */
3032                + nla_total_size(4) /* RTA_TABLE */
3033                + nla_total_size(4) /* RTA_IIF */
3034                + nla_total_size(4) /* RTA_OIF */
3035                + nla_total_size(4) /* RTA_PRIORITY */
3036                + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
3037                + nla_total_size(sizeof(struct rta_cacheinfo))
3038                + nla_total_size(TCP_CA_NAME_MAX) /* RTAX_CC_ALGO */
3039                + nla_total_size(1) /* RTA_PREF */
3040                + lwtunnel_get_encap_size(rt->dst.lwtstate);
3041 }
3042
3043 static int rt6_fill_node(struct net *net,
3044                          struct sk_buff *skb, struct rt6_info *rt,
3045                          struct in6_addr *dst, struct in6_addr *src,
3046                          int iif, int type, u32 portid, u32 seq,
3047                          int prefix, int nowait, unsigned int flags)
3048 {
3049         u32 metrics[RTAX_MAX];
3050         struct rtmsg *rtm;
3051         struct nlmsghdr *nlh;
3052         long expires;
3053         u32 table;
3054
3055         if (prefix) {   /* user wants prefix routes only */
3056                 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
3057                         /* success since this is not a prefix route */
3058                         return 1;
3059                 }
3060         }
3061
3062         nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
3063         if (!nlh)
3064                 return -EMSGSIZE;
3065
3066         rtm = nlmsg_data(nlh);
3067         rtm->rtm_family = AF_INET6;
3068         rtm->rtm_dst_len = rt->rt6i_dst.plen;
3069         rtm->rtm_src_len = rt->rt6i_src.plen;
3070         rtm->rtm_tos = 0;
3071         if (rt->rt6i_table)
3072                 table = rt->rt6i_table->tb6_id;
3073         else
3074                 table = RT6_TABLE_UNSPEC;
3075         rtm->rtm_table = table;
3076         if (nla_put_u32(skb, RTA_TABLE, table))
3077                 goto nla_put_failure;
3078         if (rt->rt6i_flags & RTF_REJECT) {
3079                 switch (rt->dst.error) {
3080                 case -EINVAL:
3081                         rtm->rtm_type = RTN_BLACKHOLE;
3082                         break;
3083                 case -EACCES:
3084                         rtm->rtm_type = RTN_PROHIBIT;
3085                         break;
3086                 case -EAGAIN:
3087                         rtm->rtm_type = RTN_THROW;
3088                         break;
3089                 default:
3090                         rtm->rtm_type = RTN_UNREACHABLE;
3091                         break;
3092                 }
3093         }
3094         else if (rt->rt6i_flags & RTF_LOCAL)
3095                 rtm->rtm_type = RTN_LOCAL;
3096         else if (rt->dst.dev && (rt->dst.dev->flags & IFF_LOOPBACK))
3097                 rtm->rtm_type = RTN_LOCAL;
3098         else
3099                 rtm->rtm_type = RTN_UNICAST;
3100         rtm->rtm_flags = 0;
3101         if (!netif_carrier_ok(rt->dst.dev)) {
3102                 rtm->rtm_flags |= RTNH_F_LINKDOWN;
3103                 if (rt->rt6i_idev->cnf.ignore_routes_with_linkdown)
3104                         rtm->rtm_flags |= RTNH_F_DEAD;
3105         }
3106         rtm->rtm_scope = RT_SCOPE_UNIVERSE;
3107         rtm->rtm_protocol = rt->rt6i_protocol;
3108         if (rt->rt6i_flags & RTF_DYNAMIC)
3109                 rtm->rtm_protocol = RTPROT_REDIRECT;
3110         else if (rt->rt6i_flags & RTF_ADDRCONF) {
3111                 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ROUTEINFO))
3112                         rtm->rtm_protocol = RTPROT_RA;
3113                 else
3114                         rtm->rtm_protocol = RTPROT_KERNEL;
3115         }
3116
3117         if (rt->rt6i_flags & RTF_CACHE)
3118                 rtm->rtm_flags |= RTM_F_CLONED;
3119
3120         if (dst) {
3121                 if (nla_put_in6_addr(skb, RTA_DST, dst))
3122                         goto nla_put_failure;
3123                 rtm->rtm_dst_len = 128;
3124         } else if (rtm->rtm_dst_len)
3125                 if (nla_put_in6_addr(skb, RTA_DST, &rt->rt6i_dst.addr))
3126                         goto nla_put_failure;
3127 #ifdef CONFIG_IPV6_SUBTREES
3128         if (src) {
3129                 if (nla_put_in6_addr(skb, RTA_SRC, src))
3130                         goto nla_put_failure;
3131                 rtm->rtm_src_len = 128;
3132         } else if (rtm->rtm_src_len &&
3133                    nla_put_in6_addr(skb, RTA_SRC, &rt->rt6i_src.addr))
3134                 goto nla_put_failure;
3135 #endif
3136         if (iif) {
3137 #ifdef CONFIG_IPV6_MROUTE
3138                 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
3139                         int err = ip6mr_get_route(net, skb, rtm, nowait);
3140                         if (err <= 0) {
3141                                 if (!nowait) {
3142                                         if (err == 0)
3143                                                 return 0;
3144                                         goto nla_put_failure;
3145                                 } else {
3146                                         if (err == -EMSGSIZE)
3147                                                 goto nla_put_failure;
3148                                 }
3149                         }
3150                 } else
3151 #endif
3152                         if (nla_put_u32(skb, RTA_IIF, iif))
3153                                 goto nla_put_failure;
3154         } else if (dst) {
3155                 struct in6_addr saddr_buf;
3156                 if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0 &&
3157                     nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
3158                         goto nla_put_failure;
3159         }
3160
3161         if (rt->rt6i_prefsrc.plen) {
3162                 struct in6_addr saddr_buf;
3163                 saddr_buf = rt->rt6i_prefsrc.addr;
3164                 if (nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
3165                         goto nla_put_failure;
3166         }
3167
3168         memcpy(metrics, dst_metrics_ptr(&rt->dst), sizeof(metrics));
3169         if (rt->rt6i_pmtu)
3170                 metrics[RTAX_MTU - 1] = rt->rt6i_pmtu;
3171         if (rtnetlink_put_metrics(skb, metrics) < 0)
3172                 goto nla_put_failure;
3173
3174         if (rt->rt6i_flags & RTF_GATEWAY) {
3175                 if (nla_put_in6_addr(skb, RTA_GATEWAY, &rt->rt6i_gateway) < 0)
3176                         goto nla_put_failure;
3177         }
3178
3179         if (rt->dst.dev &&
3180             nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex))
3181                 goto nla_put_failure;
3182         if (nla_put_u32(skb, RTA_PRIORITY, rt->rt6i_metric))
3183                 goto nla_put_failure;
3184
3185         expires = (rt->rt6i_flags & RTF_EXPIRES) ? rt->dst.expires - jiffies : 0;
3186
3187         if (rtnl_put_cacheinfo(skb, &rt->dst, 0, expires, rt->dst.error) < 0)
3188                 goto nla_put_failure;
3189
3190         if (nla_put_u8(skb, RTA_PREF, IPV6_EXTRACT_PREF(rt->rt6i_flags)))
3191                 goto nla_put_failure;
3192
3193         lwtunnel_fill_encap(skb, rt->dst.lwtstate);
3194
3195         nlmsg_end(skb, nlh);
3196         return 0;
3197
3198 nla_put_failure:
3199         nlmsg_cancel(skb, nlh);
3200         return -EMSGSIZE;
3201 }
3202
3203 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
3204 {
3205         struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
3206         int prefix;
3207
3208         if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
3209                 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
3210                 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
3211         } else
3212                 prefix = 0;
3213
3214         return rt6_fill_node(arg->net,
3215                      arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
3216                      NETLINK_CB(arg->cb->skb).portid, arg->cb->nlh->nlmsg_seq,
3217                      prefix, 0, NLM_F_MULTI);
3218 }
3219
3220 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh)
3221 {
3222         struct net *net = sock_net(in_skb->sk);
3223         struct nlattr *tb[RTA_MAX+1];
3224         struct rt6_info *rt;
3225         struct sk_buff *skb;
3226         struct rtmsg *rtm;
3227         struct flowi6 fl6;
3228         int err, iif = 0, oif = 0;
3229
3230         err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
3231         if (err < 0)
3232                 goto errout;
3233
3234         err = -EINVAL;
3235         memset(&fl6, 0, sizeof(fl6));
3236
3237         if (tb[RTA_SRC]) {
3238                 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
3239                         goto errout;
3240
3241                 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
3242         }
3243
3244         if (tb[RTA_DST]) {
3245                 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
3246                         goto errout;
3247
3248                 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
3249         }
3250
3251         if (tb[RTA_IIF])
3252                 iif = nla_get_u32(tb[RTA_IIF]);
3253
3254         if (tb[RTA_OIF])
3255                 oif = nla_get_u32(tb[RTA_OIF]);
3256
3257         if (tb[RTA_MARK])
3258                 fl6.flowi6_mark = nla_get_u32(tb[RTA_MARK]);
3259
3260         if (iif) {
3261                 struct net_device *dev;
3262                 int flags = 0;
3263
3264                 dev = __dev_get_by_index(net, iif);
3265                 if (!dev) {
3266                         err = -ENODEV;
3267                         goto errout;
3268                 }
3269
3270                 fl6.flowi6_iif = iif;
3271
3272                 if (!ipv6_addr_any(&fl6.saddr))
3273                         flags |= RT6_LOOKUP_F_HAS_SADDR;
3274
3275                 rt = (struct rt6_info *)ip6_route_input_lookup(net, dev, &fl6,
3276                                                                flags);
3277         } else {
3278                 fl6.flowi6_oif = oif;
3279
3280                 rt = (struct rt6_info *)ip6_route_output(net, NULL, &fl6);
3281         }
3282
3283         skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
3284         if (!skb) {
3285                 ip6_rt_put(rt);
3286                 err = -ENOBUFS;
3287                 goto errout;
3288         }
3289
3290         /* Reserve room for dummy headers, this skb can pass
3291            through good chunk of routing engine.
3292          */
3293         skb_reset_mac_header(skb);
3294         skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
3295
3296         skb_dst_set(skb, &rt->dst);
3297
3298         err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
3299                             RTM_NEWROUTE, NETLINK_CB(in_skb).portid,
3300                             nlh->nlmsg_seq, 0, 0, 0);
3301         if (err < 0) {
3302                 kfree_skb(skb);
3303                 goto errout;
3304         }
3305
3306         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
3307 errout:
3308         return err;
3309 }
3310
3311 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info,
3312                      unsigned int nlm_flags)
3313 {
3314         struct sk_buff *skb;
3315         struct net *net = info->nl_net;
3316         u32 seq;
3317         int err;
3318
3319         err = -ENOBUFS;
3320         seq = info->nlh ? info->nlh->nlmsg_seq : 0;
3321
3322         skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
3323         if (!skb)
3324                 goto errout;
3325
3326         err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
3327                                 event, info->portid, seq, 0, 0, nlm_flags);
3328         if (err < 0) {
3329                 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
3330                 WARN_ON(err == -EMSGSIZE);
3331                 kfree_skb(skb);
3332                 goto errout;
3333         }
3334         rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
3335                     info->nlh, gfp_any());
3336         return;
3337 errout:
3338         if (err < 0)
3339                 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
3340 }
3341
3342 static int ip6_route_dev_notify(struct notifier_block *this,
3343                                 unsigned long event, void *ptr)
3344 {
3345         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3346         struct net *net = dev_net(dev);
3347
3348         if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
3349                 net->ipv6.ip6_null_entry->dst.dev = dev;
3350                 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
3351 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3352                 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
3353                 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
3354                 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
3355                 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
3356 #endif
3357         }
3358
3359         return NOTIFY_OK;
3360 }
3361
3362 /*
3363  *      /proc
3364  */
3365
3366 #ifdef CONFIG_PROC_FS
3367
3368 static const struct file_operations ipv6_route_proc_fops = {
3369         .owner          = THIS_MODULE,
3370         .open           = ipv6_route_open,
3371         .read           = seq_read,
3372         .llseek         = seq_lseek,
3373         .release        = seq_release_net,
3374 };
3375
3376 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
3377 {
3378         struct net *net = (struct net *)seq->private;
3379         seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
3380                    net->ipv6.rt6_stats->fib_nodes,
3381                    net->ipv6.rt6_stats->fib_route_nodes,
3382                    net->ipv6.rt6_stats->fib_rt_alloc,
3383                    net->ipv6.rt6_stats->fib_rt_entries,
3384                    net->ipv6.rt6_stats->fib_rt_cache,
3385                    dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
3386                    net->ipv6.rt6_stats->fib_discarded_routes);
3387
3388         return 0;
3389 }
3390
3391 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
3392 {
3393         return single_open_net(inode, file, rt6_stats_seq_show);
3394 }
3395
3396 static const struct file_operations rt6_stats_seq_fops = {
3397         .owner   = THIS_MODULE,
3398         .open    = rt6_stats_seq_open,
3399         .read    = seq_read,
3400         .llseek  = seq_lseek,
3401         .release = single_release_net,
3402 };
3403 #endif  /* CONFIG_PROC_FS */
3404
3405 #ifdef CONFIG_SYSCTL
3406
3407 static
3408 int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
3409                               void __user *buffer, size_t *lenp, loff_t *ppos)
3410 {
3411         struct net *net;
3412         int delay;
3413         if (!write)
3414                 return -EINVAL;
3415
3416         net = (struct net *)ctl->extra1;
3417         delay = net->ipv6.sysctl.flush_delay;
3418         proc_dointvec(ctl, write, buffer, lenp, ppos);
3419         fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
3420         return 0;
3421 }
3422
3423 struct ctl_table ipv6_route_table_template[] = {
3424         {
3425                 .procname       =       "flush",
3426                 .data           =       &init_net.ipv6.sysctl.flush_delay,
3427                 .maxlen         =       sizeof(int),
3428                 .mode           =       0200,
3429                 .proc_handler   =       ipv6_sysctl_rtcache_flush
3430         },
3431         {
3432                 .procname       =       "gc_thresh",
3433                 .data           =       &ip6_dst_ops_template.gc_thresh,
3434                 .maxlen         =       sizeof(int),
3435                 .mode           =       0644,
3436                 .proc_handler   =       proc_dointvec,
3437         },
3438         {
3439                 .procname       =       "max_size",
3440                 .data           =       &init_net.ipv6.sysctl.ip6_rt_max_size,
3441                 .maxlen         =       sizeof(int),
3442                 .mode           =       0644,
3443                 .proc_handler   =       proc_dointvec,
3444         },
3445         {
3446                 .procname       =       "gc_min_interval",
3447                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
3448                 .maxlen         =       sizeof(int),
3449                 .mode           =       0644,
3450                 .proc_handler   =       proc_dointvec_jiffies,
3451         },
3452         {
3453                 .procname       =       "gc_timeout",
3454                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
3455                 .maxlen         =       sizeof(int),
3456                 .mode           =       0644,
3457                 .proc_handler   =       proc_dointvec_jiffies,
3458         },
3459         {
3460                 .procname       =       "gc_interval",
3461                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_interval,
3462                 .maxlen         =       sizeof(int),
3463                 .mode           =       0644,
3464                 .proc_handler   =       proc_dointvec_jiffies,
3465         },
3466         {
3467                 .procname       =       "gc_elasticity",
3468                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
3469                 .maxlen         =       sizeof(int),
3470                 .mode           =       0644,
3471                 .proc_handler   =       proc_dointvec,
3472         },
3473         {
3474                 .procname       =       "mtu_expires",
3475                 .data           =       &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
3476                 .maxlen         =       sizeof(int),
3477                 .mode           =       0644,
3478                 .proc_handler   =       proc_dointvec_jiffies,
3479         },
3480         {
3481                 .procname       =       "min_adv_mss",
3482                 .data           =       &init_net.ipv6.sysctl.ip6_rt_min_advmss,
3483                 .maxlen         =       sizeof(int),
3484                 .mode           =       0644,
3485                 .proc_handler   =       proc_dointvec,
3486         },
3487         {
3488                 .procname       =       "gc_min_interval_ms",
3489                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
3490                 .maxlen         =       sizeof(int),
3491                 .mode           =       0644,
3492                 .proc_handler   =       proc_dointvec_ms_jiffies,
3493         },
3494         { }
3495 };
3496
3497 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
3498 {
3499         struct ctl_table *table;
3500
3501         table = kmemdup(ipv6_route_table_template,
3502                         sizeof(ipv6_route_table_template),
3503                         GFP_KERNEL);
3504
3505         if (table) {
3506                 table[0].data = &net->ipv6.sysctl.flush_delay;
3507                 table[0].extra1 = net;
3508                 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
3509                 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
3510                 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
3511                 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
3512                 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
3513                 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
3514                 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
3515                 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
3516                 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
3517
3518                 /* Don't export sysctls to unprivileged users */
3519                 if (net->user_ns != &init_user_ns)
3520                         table[0].procname = NULL;
3521         }
3522
3523         return table;
3524 }
3525 #endif
3526
3527 static int __net_init ip6_route_net_init(struct net *net)
3528 {
3529         int ret = -ENOMEM;
3530
3531         memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
3532                sizeof(net->ipv6.ip6_dst_ops));
3533
3534         if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
3535                 goto out_ip6_dst_ops;
3536
3537         net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
3538                                            sizeof(*net->ipv6.ip6_null_entry),
3539                                            GFP_KERNEL);
3540         if (!net->ipv6.ip6_null_entry)
3541                 goto out_ip6_dst_entries;
3542         net->ipv6.ip6_null_entry->dst.path =
3543                 (struct dst_entry *)net->ipv6.ip6_null_entry;
3544         net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3545         dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
3546                          ip6_template_metrics, true);
3547
3548 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3549         net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
3550                                                sizeof(*net->ipv6.ip6_prohibit_entry),
3551                                                GFP_KERNEL);
3552         if (!net->ipv6.ip6_prohibit_entry)
3553                 goto out_ip6_null_entry;
3554         net->ipv6.ip6_prohibit_entry->dst.path =
3555                 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
3556         net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3557         dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
3558                          ip6_template_metrics, true);
3559
3560         net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
3561                                                sizeof(*net->ipv6.ip6_blk_hole_entry),
3562                                                GFP_KERNEL);
3563         if (!net->ipv6.ip6_blk_hole_entry)
3564                 goto out_ip6_prohibit_entry;
3565         net->ipv6.ip6_blk_hole_entry->dst.path =
3566                 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
3567         net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3568         dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
3569                          ip6_template_metrics, true);
3570 #endif
3571
3572         net->ipv6.sysctl.flush_delay = 0;
3573         net->ipv6.sysctl.ip6_rt_max_size = 4096;
3574         net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
3575         net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
3576         net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
3577         net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
3578         net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
3579         net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
3580
3581         net->ipv6.ip6_rt_gc_expire = 30*HZ;
3582
3583         ret = 0;
3584 out:
3585         return ret;
3586
3587 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3588 out_ip6_prohibit_entry:
3589         kfree(net->ipv6.ip6_prohibit_entry);
3590 out_ip6_null_entry:
3591         kfree(net->ipv6.ip6_null_entry);
3592 #endif
3593 out_ip6_dst_entries:
3594         dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3595 out_ip6_dst_ops:
3596         goto out;
3597 }
3598
3599 static void __net_exit ip6_route_net_exit(struct net *net)
3600 {
3601         kfree(net->ipv6.ip6_null_entry);
3602 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3603         kfree(net->ipv6.ip6_prohibit_entry);
3604         kfree(net->ipv6.ip6_blk_hole_entry);
3605 #endif
3606         dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3607 }
3608
3609 static int __net_init ip6_route_net_init_late(struct net *net)
3610 {
3611 #ifdef CONFIG_PROC_FS
3612         proc_create("ipv6_route", 0, net->proc_net, &ipv6_route_proc_fops);
3613         proc_create("rt6_stats", S_IRUGO, net->proc_net, &rt6_stats_seq_fops);
3614 #endif
3615         return 0;
3616 }
3617
3618 static void __net_exit ip6_route_net_exit_late(struct net *net)
3619 {
3620 #ifdef CONFIG_PROC_FS
3621         remove_proc_entry("ipv6_route", net->proc_net);
3622         remove_proc_entry("rt6_stats", net->proc_net);
3623 #endif
3624 }
3625
3626 static struct pernet_operations ip6_route_net_ops = {
3627         .init = ip6_route_net_init,
3628         .exit = ip6_route_net_exit,
3629 };
3630
3631 static int __net_init ipv6_inetpeer_init(struct net *net)
3632 {
3633         struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
3634
3635         if (!bp)
3636                 return -ENOMEM;
3637         inet_peer_base_init(bp);
3638         net->ipv6.peers = bp;
3639         return 0;
3640 }
3641
3642 static void __net_exit ipv6_inetpeer_exit(struct net *net)
3643 {
3644         struct inet_peer_base *bp = net->ipv6.peers;
3645
3646         net->ipv6.peers = NULL;
3647         inetpeer_invalidate_tree(bp);
3648         kfree(bp);
3649 }
3650
3651 static struct pernet_operations ipv6_inetpeer_ops = {
3652         .init   =       ipv6_inetpeer_init,
3653         .exit   =       ipv6_inetpeer_exit,
3654 };
3655
3656 static struct pernet_operations ip6_route_net_late_ops = {
3657         .init = ip6_route_net_init_late,
3658         .exit = ip6_route_net_exit_late,
3659 };
3660
3661 static struct notifier_block ip6_route_dev_notifier = {
3662         .notifier_call = ip6_route_dev_notify,
3663         .priority = 0,
3664 };
3665
3666 int __init ip6_route_init(void)
3667 {
3668         int ret;
3669         int cpu;
3670
3671         ret = -ENOMEM;
3672         ip6_dst_ops_template.kmem_cachep =
3673                 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
3674                                   SLAB_HWCACHE_ALIGN, NULL);
3675         if (!ip6_dst_ops_template.kmem_cachep)
3676                 goto out;
3677
3678         ret = dst_entries_init(&ip6_dst_blackhole_ops);
3679         if (ret)
3680                 goto out_kmem_cache;
3681
3682         ret = register_pernet_subsys(&ipv6_inetpeer_ops);
3683         if (ret)
3684                 goto out_dst_entries;
3685
3686         ret = register_pernet_subsys(&ip6_route_net_ops);
3687         if (ret)
3688                 goto out_register_inetpeer;
3689
3690         ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
3691
3692         /* Registering of the loopback is done before this portion of code,
3693          * the loopback reference in rt6_info will not be taken, do it
3694          * manually for init_net */
3695         init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
3696         init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3697   #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3698         init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
3699         init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3700         init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
3701         init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3702   #endif
3703         ret = fib6_init();
3704         if (ret)
3705                 goto out_register_subsys;
3706
3707         ret = xfrm6_init();
3708         if (ret)
3709                 goto out_fib6_init;
3710
3711         ret = fib6_rules_init();
3712         if (ret)
3713                 goto xfrm6_init;
3714
3715         ret = register_pernet_subsys(&ip6_route_net_late_ops);
3716         if (ret)
3717                 goto fib6_rules_init;
3718
3719         ret = -ENOBUFS;
3720         if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL, NULL) ||
3721             __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL, NULL) ||
3722             __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL, NULL))
3723                 goto out_register_late_subsys;
3724
3725         ret = register_netdevice_notifier(&ip6_route_dev_notifier);
3726         if (ret)
3727                 goto out_register_late_subsys;
3728
3729         for_each_possible_cpu(cpu) {
3730                 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
3731
3732                 INIT_LIST_HEAD(&ul->head);
3733                 spin_lock_init(&ul->lock);
3734         }
3735
3736 out:
3737         return ret;
3738
3739 out_register_late_subsys:
3740         unregister_pernet_subsys(&ip6_route_net_late_ops);
3741 fib6_rules_init:
3742         fib6_rules_cleanup();
3743 xfrm6_init:
3744         xfrm6_fini();
3745 out_fib6_init:
3746         fib6_gc_cleanup();
3747 out_register_subsys:
3748         unregister_pernet_subsys(&ip6_route_net_ops);
3749 out_register_inetpeer:
3750         unregister_pernet_subsys(&ipv6_inetpeer_ops);
3751 out_dst_entries:
3752         dst_entries_destroy(&ip6_dst_blackhole_ops);
3753 out_kmem_cache:
3754         kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3755         goto out;
3756 }
3757
3758 void ip6_route_cleanup(void)
3759 {
3760         unregister_netdevice_notifier(&ip6_route_dev_notifier);
3761         unregister_pernet_subsys(&ip6_route_net_late_ops);
3762         fib6_rules_cleanup();
3763         xfrm6_fini();
3764         fib6_gc_cleanup();
3765         unregister_pernet_subsys(&ipv6_inetpeer_ops);
3766         unregister_pernet_subsys(&ip6_route_net_ops);
3767         dst_entries_destroy(&ip6_dst_blackhole_ops);
3768         kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3769 }